{
  "schemaVersion": 1,
  "deviceProperties": [
    {
      "id": 0, "name": "NVIDIA A800 80GB PCIe", "totalGlobalMem": 84987740160,
      "computeMajor": 8, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 167936,
      "numSms": 108, "sharedMemPerBlockOptin": 166912
    },
    {
      "id": 1, "name": "NVIDIA A800 80GB PCIe", "totalGlobalMem": 84987740160,
      "computeMajor": 8, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 167936,
      "numSms": 108, "sharedMemPerBlockOptin": 166912
    },
    {
      "id": 2, "name": "NVIDIA A800 80GB PCIe", "totalGlobalMem": 84987740160,
      "computeMajor": 8, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 167936,
      "numSms": 108, "sharedMemPerBlockOptin": 166912
    },
    {
      "id": 3, "name": "NVIDIA A800 80GB PCIe", "totalGlobalMem": 84987740160,
      "computeMajor": 8, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 167936,
      "numSms": 108, "sharedMemPerBlockOptin": 166912
    },
    {
      "id": 4, "name": "Tesla V100S-PCIE-32GB", "totalGlobalMem": 34079899648,
      "computeMajor": 7, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 98304,
      "numSms": 80, "sharedMemPerBlockOptin": 98304
    },
    {
      "id": 5, "name": "Tesla V100S-PCIE-32GB", "totalGlobalMem": 34079899648,
      "computeMajor": 7, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 98304,
      "numSms": 80, "sharedMemPerBlockOptin": 98304
    },
    {
      "id": 6, "name": "Tesla V100S-PCIE-32GB", "totalGlobalMem": 34079899648,
      "computeMajor": 7, "computeMinor": 0,
      "maxThreadsPerBlock": 1024, "maxThreadsPerMultiprocessor": 2048,
      "regsPerBlock": 65536, "regsPerMultiprocessor": 65536, "warpSize": 32,
      "sharedMemPerBlock": 49152, "sharedMemPerMultiprocessor": 98304,
      "numSms": 80, "sharedMemPerBlockOptin": 98304
    }
  ],
  "traceEvents": [
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454215754427, "dur": 7,
    "args": {
      "External id": 8,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 8, "pid": 5, "tid": 7, "ts": 1716454215754427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454215754164, "dur": 267,
    "args": {
      "External id": 8,
      "cbid": 211, "correlation": 8
    }
  },
  {
    "ph": "s", "id": 8, "pid": 76337, "tid": -914061504, "ts": 1716454215754164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454215754750, "dur": 5,
    "args": {
      "External id": 19,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.15,
      "warps per SM": 0.6,
      "grid": [12, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 19, "pid": 5, "tid": 7, "ts": 1716454215754750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454215754668, "dur": 83,
    "args": {
      "External id": 19,
      "cbid": 211, "correlation": 19
    }
  },
  {
    "ph": "s", "id": 19, "pid": 76337, "tid": -914061504, "ts": 1716454215754668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454215755400, "dur": 4,
    "args": {
      "External id": 29,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29,
      "registers per thread": 26,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29, "pid": 5, "tid": 7, "ts": 1716454215755400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454215755286, "dur": 114,
    "args": {
      "External id": 29,
      "cbid": 211, "correlation": 29
    }
  },
  {
    "ph": "s", "id": 29, "pid": 76337, "tid": -914061504, "ts": 1716454215755286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454215779029, "dur": 1,
    "args": {
      "External id": 39,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 39,
      "bytes": 160, "memory bandwidth (GB/s)": 0.08064516129032258
    }
  },
  {
    "ph": "f", "id": 39, "pid": 5, "tid": 7, "ts": 1716454215779029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454215778938, "dur": 91,
    "args": {
      "External id": 39,
      "cbid": 41, "correlation": 39
    }
  },
  {
    "ph": "s", "id": 39, "pid": 76337, "tid": -914061504, "ts": 1716454215778938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454215779031, "dur": 20,
    "args": {
      "External id": 40,
      "cbid": 131, "correlation": 40
    }
  },
  {
    "ph": "f", "id": 40, "pid": 76337, "tid": -914061504, "ts": 1716454215779031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454215780009, "dur": 94,
    "args": {
      "External id": 54,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 54,
      "bytes": 1179648, "memory bandwidth (GB/s)": 12.53464525932144
    }
  },
  {
    "ph": "f", "id": 54, "pid": 5, "tid": 7, "ts": 1716454215780009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454215779738, "dur": 286,
    "args": {
      "External id": 54,
      "cbid": 41, "correlation": 54
    }
  },
  {
    "ph": "s", "id": 54, "pid": 76337, "tid": -914061504, "ts": 1716454215779738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454215780025, "dur": 85,
    "args": {
      "External id": 55,
      "cbid": 131, "correlation": 55
    }
  },
  {
    "ph": "f", "id": 55, "pid": 76337, "tid": -914061504, "ts": 1716454215780025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454215780178, "dur": 7,
    "args": {
      "External id": 59,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 59, "pid": 5, "tid": 7, "ts": 1716454215780178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454215780131, "dur": 49,
    "args": {
      "External id": 59,
      "cbid": 211, "correlation": 59
    }
  },
  {
    "ph": "s", "id": 59, "pid": 76337, "tid": -914061504, "ts": 1716454215780131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454215780719, "dur": 2,
    "args": {
      "External id": 68,
      "cbid": 317, "correlation": 68
    }
  },
  {
    "ph": "f", "id": 68, "pid": 76337, "tid": -914061504, "ts": 1716454215780719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454215780722, "dur": 221920,
    "args": {
      "External id": 69,
      "cbid": 20, "correlation": 69
    }
  },
  {
    "ph": "f", "id": 69, "pid": 76337, "tid": -914061504, "ts": 1716454215780722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216002883, "dur": 6,
    "args": {
      "External id": 78,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0875,
      "warps per SM": 0.35,
      "grid": [7, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 78, "pid": 5, "tid": 7, "ts": 1716454216002883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216002804, "dur": 83,
    "args": {
      "External id": 78,
      "cbid": 211, "correlation": 78
    }
  },
  {
    "ph": "s", "id": 78, "pid": 76337, "tid": -914061504, "ts": 1716454216002804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216002924, "dur": 1,
    "args": {
      "External id": 85,
      "cbid": 317, "correlation": 85
    }
  },
  {
    "ph": "f", "id": 85, "pid": 76337, "tid": -914061504, "ts": 1716454216002924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216002927, "dur": 1,
    "args": {
      "External id": 86,
      "cbid": 203, "correlation": 86
    }
  },
  {
    "ph": "f", "id": 86, "pid": 76337, "tid": -914061504, "ts": 1716454216002927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216002929, "dur": 1,
    "args": {
      "External id": 87,
      "cbid": 205, "correlation": 87
    }
  },
  {
    "ph": "f", "id": 87, "pid": 76337, "tid": -914061504, "ts": 1716454216002929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_wo_smem_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x32x16_stage1_warpsize4x1x1_g1_tensor8x8x4_aligna2_alignc8_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216008035, "dur": 190,
    "args": {
      "External id": 91,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91,
      "registers per thread": 111,
      "shared memory": 4096,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [4, 1536, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 91, "pid": 5, "tid": 7, "ts": 1716454216008035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216007881, "dur": 155,
    "args": {
      "External id": 91,
      "cbid": 211, "correlation": 91
    }
  },
  {
    "ph": "s", "id": 91, "pid": 76337, "tid": -914061504, "ts": 1716454216007881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216008851, "dur": 178,
    "args": {
      "External id": 97,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97, "pid": 5, "tid": 7, "ts": 1716454216008851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216008756, "dur": 92,
    "args": {
      "External id": 97,
      "cbid": 211, "correlation": 97
    }
  },
  {
    "ph": "s", "id": 97, "pid": 76337, "tid": -914061504, "ts": 1716454216008756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216009361, "dur": 2,
    "args": {
      "External id": 105,
      "cbid": 317, "correlation": 105
    }
  },
  {
    "ph": "f", "id": 105, "pid": 76337, "tid": -914061504, "ts": 1716454216009361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216009364, "dur": 394997,
    "args": {
      "External id": 106,
      "cbid": 20, "correlation": 106
    }
  },
  {
    "ph": "f", "id": 106, "pid": 76337, "tid": -914061504, "ts": 1716454216009364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216404536, "dur": 1802,
    "args": {
      "External id": 111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111, "pid": 5, "tid": 7, "ts": 1716454216404536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216404465, "dur": 76,
    "args": {
      "External id": 111,
      "cbid": 211, "correlation": 111
    }
  },
  {
    "ph": "s", "id": 111, "pid": 76337, "tid": -914061504, "ts": 1716454216404465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216404651, "dur": 2,
    "args": {
      "External id": 121,
      "cbid": 317, "correlation": 121
    }
  },
  {
    "ph": "f", "id": 121, "pid": 76337, "tid": -914061504, "ts": 1716454216404651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216404654, "dur": 16045,
    "args": {
      "External id": 122,
      "cbid": 20, "correlation": 122
    }
  },
  {
    "ph": "f", "id": 122, "pid": 76337, "tid": -914061504, "ts": 1716454216404654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216420919, "dur": 703,
    "args": {
      "External id": 136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 136, "pid": 5, "tid": 7, "ts": 1716454216420919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216420761, "dur": 160,
    "args": {
      "External id": 136,
      "cbid": 211, "correlation": 136
    }
  },
  {
    "ph": "s", "id": 136, "pid": 76337, "tid": -914061504, "ts": 1716454216420761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216421629, "dur": 5,
    "args": {
      "External id": 148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 148, "pid": 5, "tid": 7, "ts": 1716454216421629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216420955, "dur": 75,
    "args": {
      "External id": 148,
      "cbid": 211, "correlation": 148
    }
  },
  {
    "ph": "s", "id": 148, "pid": 76337, "tid": -914061504, "ts": 1716454216420955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216421644, "dur": 176,
    "args": {
      "External id": 151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151, "pid": 5, "tid": 7, "ts": 1716454216421644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216421073, "dur": 78,
    "args": {
      "External id": 151,
      "cbid": 211, "correlation": 151
    }
  },
  {
    "ph": "s", "id": 151, "pid": 76337, "tid": -914061504, "ts": 1716454216421073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216421828, "dur": 108,
    "args": {
      "External id": 160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160, "pid": 5, "tid": 7, "ts": 1716454216421828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216421444, "dur": 77,
    "args": {
      "External id": 160,
      "cbid": 211, "correlation": 160
    }
  },
  {
    "ph": "s", "id": 160, "pid": 76337, "tid": -914061504, "ts": 1716454216421444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216421664, "dur": 0,
    "args": {
      "External id": 170,
      "cbid": 317, "correlation": 170
    }
  },
  {
    "ph": "f", "id": 170, "pid": 76337, "tid": -914061504, "ts": 1716454216421664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216421665, "dur": 2,
    "args": {
      "External id": 171,
      "cbid": 203, "correlation": 171
    }
  },
  {
    "ph": "f", "id": 171, "pid": 76337, "tid": -914061504, "ts": 1716454216421665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216421668, "dur": 1,
    "args": {
      "External id": 172,
      "cbid": 205, "correlation": 172
    }
  },
  {
    "ph": "f", "id": 172, "pid": 76337, "tid": -914061504, "ts": 1716454216421668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216425170, "dur": 4,
    "args": {
      "External id": 178,
      "cbid": 317, "correlation": 178
    }
  },
  {
    "ph": "f", "id": 178, "pid": 76337, "tid": -914061504, "ts": 1716454216425170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216425175, "dur": 6658,
    "args": {
      "External id": 179,
      "cbid": 20, "correlation": 179
    }
  },
  {
    "ph": "f", "id": 179, "pid": 76337, "tid": -914061504, "ts": 1716454216425175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216431989, "dur": 144,
    "args": {
      "External id": 180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180, "pid": 5, "tid": 7, "ts": 1716454216431989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216431881, "dur": 111,
    "args": {
      "External id": 180,
      "cbid": 211, "correlation": 180
    }
  },
  {
    "ph": "s", "id": 180, "pid": 76337, "tid": -914061504, "ts": 1716454216431881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216432135, "dur": 6,
    "args": {
      "External id": 182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182, "pid": 5, "tid": 7, "ts": 1716454216432135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216431996, "dur": 8,
    "args": {
      "External id": 182,
      "cbid": 211, "correlation": 182
    }
  },
  {
    "ph": "s", "id": 182, "pid": 76337, "tid": -914061504, "ts": 1716454216431996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216432142, "dur": 6,
    "args": {
      "External id": 184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184, "pid": 5, "tid": 7, "ts": 1716454216432142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216432020, "dur": 12,
    "args": {
      "External id": 184,
      "cbid": 211, "correlation": 184
    }
  },
  {
    "ph": "s", "id": 184, "pid": 76337, "tid": -914061504, "ts": 1716454216432020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216432036, "dur": 0,
    "args": {
      "External id": 185,
      "cbid": 51, "correlation": 185
    }
  },
  {
    "ph": "s", "id": 185, "pid": 76337, "tid": -914061504, "ts": 1716454216432036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216432158, "dur": 818,
    "args": {
      "External id": 186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186, "pid": 5, "tid": 7, "ts": 1716454216432158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216432038, "dur": 103,
    "args": {
      "External id": 186,
      "cbid": 211, "correlation": 186
    }
  },
  {
    "ph": "s", "id": 186, "pid": 76337, "tid": -914061504, "ts": 1716454216432038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216432978, "dur": 178,
    "args": {
      "External id": 191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191, "pid": 5, "tid": 7, "ts": 1716454216432978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216432910, "dur": 14,
    "args": {
      "External id": 191,
      "cbid": 211, "correlation": 191
    }
  },
  {
    "ph": "s", "id": 191, "pid": 76337, "tid": -914061504, "ts": 1716454216432910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216433162, "dur": 708,
    "args": {
      "External id": 211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 211, "pid": 5, "tid": 7, "ts": 1716454216433162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433143, "dur": 17,
    "args": {
      "External id": 211,
      "cbid": 211, "correlation": 211
    }
  },
  {
    "ph": "s", "id": 211, "pid": 76337, "tid": -914061504, "ts": 1716454216433143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216433871, "dur": 5,
    "args": {
      "External id": 223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 223, "pid": 5, "tid": 7, "ts": 1716454216433871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433173, "dur": 8,
    "args": {
      "External id": 223,
      "cbid": 211, "correlation": 223
    }
  },
  {
    "ph": "s", "id": 223, "pid": 76337, "tid": -914061504, "ts": 1716454216433173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216433878, "dur": 178,
    "args": {
      "External id": 226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226, "pid": 5, "tid": 7, "ts": 1716454216433878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433209, "dur": 8,
    "args": {
      "External id": 226,
      "cbid": 211, "correlation": 226
    }
  },
  {
    "ph": "s", "id": 226, "pid": 76337, "tid": -914061504, "ts": 1716454216433209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216434057, "dur": 109,
    "args": {
      "External id": 235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235, "pid": 5, "tid": 7, "ts": 1716454216434057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433271, "dur": 11,
    "args": {
      "External id": 235,
      "cbid": 211, "correlation": 235
    }
  },
  {
    "ph": "s", "id": 235, "pid": 76337, "tid": -914061504, "ts": 1716454216433271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433446, "dur": 0,
    "args": {
      "External id": 245,
      "cbid": 317, "correlation": 245
    }
  },
  {
    "ph": "f", "id": 245, "pid": 76337, "tid": -914061504, "ts": 1716454216433446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433448, "dur": 0,
    "args": {
      "External id": 246,
      "cbid": 203, "correlation": 246
    }
  },
  {
    "ph": "f", "id": 246, "pid": 76337, "tid": -914061504, "ts": 1716454216433448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433449, "dur": 0,
    "args": {
      "External id": 247,
      "cbid": 205, "correlation": 247
    }
  },
  {
    "ph": "f", "id": 247, "pid": 76337, "tid": -914061504, "ts": 1716454216433449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216434167, "dur": 147,
    "args": {
      "External id": 251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251, "pid": 5, "tid": 7, "ts": 1716454216434167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433465, "dur": 14,
    "args": {
      "External id": 251,
      "cbid": 211, "correlation": 251
    }
  },
  {
    "ph": "s", "id": 251, "pid": 76337, "tid": -914061504, "ts": 1716454216433465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216434316, "dur": 6,
    "args": {
      "External id": 253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 253, "pid": 5, "tid": 7, "ts": 1716454216434316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433482, "dur": 5,
    "args": {
      "External id": 253,
      "cbid": 211, "correlation": 253
    }
  },
  {
    "ph": "s", "id": 253, "pid": 76337, "tid": -914061504, "ts": 1716454216433482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216434323, "dur": 5,
    "args": {
      "External id": 255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255, "pid": 5, "tid": 7, "ts": 1716454216434323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433492, "dur": 6,
    "args": {
      "External id": 255,
      "cbid": 211, "correlation": 255
    }
  },
  {
    "ph": "s", "id": 255, "pid": 76337, "tid": -914061504, "ts": 1716454216433492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433501, "dur": 0,
    "args": {
      "External id": 256,
      "cbid": 51, "correlation": 256
    }
  },
  {
    "ph": "s", "id": 256, "pid": 76337, "tid": -914061504, "ts": 1716454216433501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216434330, "dur": 816,
    "args": {
      "External id": 257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257, "pid": 5, "tid": 7, "ts": 1716454216434330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433502, "dur": 5,
    "args": {
      "External id": 257,
      "cbid": 211, "correlation": 257
    }
  },
  {
    "ph": "s", "id": 257, "pid": 76337, "tid": -914061504, "ts": 1716454216433502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216435147, "dur": 176,
    "args": {
      "External id": 262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262, "pid": 5, "tid": 7, "ts": 1716454216435147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433531, "dur": 10,
    "args": {
      "External id": 262,
      "cbid": 211, "correlation": 262
    }
  },
  {
    "ph": "s", "id": 262, "pid": 76337, "tid": -914061504, "ts": 1716454216433531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216435324, "dur": 751,
    "args": {
      "External id": 270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270, "pid": 5, "tid": 7, "ts": 1716454216435324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433578, "dur": 9,
    "args": {
      "External id": 270,
      "cbid": 211, "correlation": 270
    }
  },
  {
    "ph": "s", "id": 270, "pid": 76337, "tid": -914061504, "ts": 1716454216433578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216436082, "dur": 103,
    "args": {
      "External id": 278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278, "pid": 5, "tid": 7, "ts": 1716454216436082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433692, "dur": 107,
    "args": {
      "External id": 278,
      "cbid": 211, "correlation": 278
    }
  },
  {
    "ph": "s", "id": 278, "pid": 76337, "tid": -914061504, "ts": 1716454216433692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216436186, "dur": 1802,
    "args": {
      "External id": 288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288, "pid": 5, "tid": 7, "ts": 1716454216436186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433897, "dur": 16,
    "args": {
      "External id": 288,
      "cbid": 211, "correlation": 288
    }
  },
  {
    "ph": "s", "id": 288, "pid": 76337, "tid": -914061504, "ts": 1716454216433897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216437990, "dur": 704,
    "args": {
      "External id": 309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 309,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 309, "pid": 5, "tid": 7, "ts": 1716454216437990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433938, "dur": 8,
    "args": {
      "External id": 309,
      "cbid": 211, "correlation": 309
    }
  },
  {
    "ph": "s", "id": 309, "pid": 76337, "tid": -914061504, "ts": 1716454216433938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216438695, "dur": 5,
    "args": {
      "External id": 321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 321,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 321, "pid": 5, "tid": 7, "ts": 1716454216438695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433964, "dur": 8,
    "args": {
      "External id": 321,
      "cbid": 211, "correlation": 321
    }
  },
  {
    "ph": "s", "id": 321, "pid": 76337, "tid": -914061504, "ts": 1716454216433964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216438701, "dur": 176,
    "args": {
      "External id": 324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 324, "pid": 5, "tid": 7, "ts": 1716454216438701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216433995, "dur": 7,
    "args": {
      "External id": 324,
      "cbid": 211, "correlation": 324
    }
  },
  {
    "ph": "s", "id": 324, "pid": 76337, "tid": -914061504, "ts": 1716454216433995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216438879, "dur": 110,
    "args": {
      "External id": 333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 333,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 333, "pid": 5, "tid": 7, "ts": 1716454216438879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216434038, "dur": 11,
    "args": {
      "External id": 333,
      "cbid": 211, "correlation": 333
    }
  },
  {
    "ph": "s", "id": 333, "pid": 76337, "tid": -914061504, "ts": 1716454216434038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216434110, "dur": 0,
    "args": {
      "External id": 343,
      "cbid": 317, "correlation": 343
    }
  },
  {
    "ph": "f", "id": 343, "pid": 76337, "tid": -914061504, "ts": 1716454216434110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216434111, "dur": 0,
    "args": {
      "External id": 344,
      "cbid": 203, "correlation": 344
    }
  },
  {
    "ph": "f", "id": 344, "pid": 76337, "tid": -914061504, "ts": 1716454216434111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216434111, "dur": 0,
    "args": {
      "External id": 345,
      "cbid": 205, "correlation": 345
    }
  },
  {
    "ph": "f", "id": 345, "pid": 76337, "tid": -914061504, "ts": 1716454216434111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216434119, "dur": 1,
    "args": {
      "External id": 351,
      "cbid": 317, "correlation": 351
    }
  },
  {
    "ph": "f", "id": 351, "pid": 76337, "tid": -914061504, "ts": 1716454216434119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216434121, "dur": 11381,
    "args": {
      "External id": 352,
      "cbid": 20, "correlation": 352
    }
  },
  {
    "ph": "f", "id": 352, "pid": 76337, "tid": -914061504, "ts": 1716454216434121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216445537, "dur": 141,
    "args": {
      "External id": 353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 353,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 353, "pid": 5, "tid": 7, "ts": 1716454216445537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445518, "dur": 19,
    "args": {
      "External id": 353,
      "cbid": 211, "correlation": 353
    }
  },
  {
    "ph": "s", "id": 353, "pid": 76337, "tid": -914061504, "ts": 1716454216445518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216445680, "dur": 6,
    "args": {
      "External id": 355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 355,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 355, "pid": 5, "tid": 7, "ts": 1716454216445680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445540, "dur": 6,
    "args": {
      "External id": 355,
      "cbid": 211, "correlation": 355
    }
  },
  {
    "ph": "s", "id": 355, "pid": 76337, "tid": -914061504, "ts": 1716454216445540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216445688, "dur": 5,
    "args": {
      "External id": 357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 357, "pid": 5, "tid": 7, "ts": 1716454216445688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445550, "dur": 6,
    "args": {
      "External id": 357,
      "cbid": 211, "correlation": 357
    }
  },
  {
    "ph": "s", "id": 357, "pid": 76337, "tid": -914061504, "ts": 1716454216445550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445559, "dur": 0,
    "args": {
      "External id": 358,
      "cbid": 51, "correlation": 358
    }
  },
  {
    "ph": "s", "id": 358, "pid": 76337, "tid": -914061504, "ts": 1716454216445559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216445695, "dur": 822,
    "args": {
      "External id": 359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 359,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 359, "pid": 5, "tid": 7, "ts": 1716454216445695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445560, "dur": 5,
    "args": {
      "External id": 359,
      "cbid": 211, "correlation": 359
    }
  },
  {
    "ph": "s", "id": 359, "pid": 76337, "tid": -914061504, "ts": 1716454216445560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216446518, "dur": 178,
    "args": {
      "External id": 364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 364, "pid": 5, "tid": 7, "ts": 1716454216446518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445595, "dur": 10,
    "args": {
      "External id": 364,
      "cbid": 211, "correlation": 364
    }
  },
  {
    "ph": "s", "id": 364, "pid": 76337, "tid": -914061504, "ts": 1716454216445595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216446697, "dur": 713,
    "args": {
      "External id": 384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 384,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 384, "pid": 5, "tid": 7, "ts": 1716454216446697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445706, "dur": 13,
    "args": {
      "External id": 384,
      "cbid": 211, "correlation": 384
    }
  },
  {
    "ph": "s", "id": 384, "pid": 76337, "tid": -914061504, "ts": 1716454216445706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216447412, "dur": 4,
    "args": {
      "External id": 396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 396,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 396, "pid": 5, "tid": 7, "ts": 1716454216447412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445729, "dur": 7,
    "args": {
      "External id": 396,
      "cbid": 211, "correlation": 396
    }
  },
  {
    "ph": "s", "id": 396, "pid": 76337, "tid": -914061504, "ts": 1716454216445729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216447418, "dur": 177,
    "args": {
      "External id": 399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 399,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 399, "pid": 5, "tid": 7, "ts": 1716454216447418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445750, "dur": 7,
    "args": {
      "External id": 399,
      "cbid": 211, "correlation": 399
    }
  },
  {
    "ph": "s", "id": 399, "pid": 76337, "tid": -914061504, "ts": 1716454216445750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216447596, "dur": 109,
    "args": {
      "External id": 408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 408,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 408, "pid": 5, "tid": 7, "ts": 1716454216447596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445791, "dur": 10,
    "args": {
      "External id": 408,
      "cbid": 211, "correlation": 408
    }
  },
  {
    "ph": "s", "id": 408, "pid": 76337, "tid": -914061504, "ts": 1716454216445791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445883, "dur": 0,
    "args": {
      "External id": 418,
      "cbid": 317, "correlation": 418
    }
  },
  {
    "ph": "f", "id": 418, "pid": 76337, "tid": -914061504, "ts": 1716454216445883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445884, "dur": 0,
    "args": {
      "External id": 419,
      "cbid": 203, "correlation": 419
    }
  },
  {
    "ph": "f", "id": 419, "pid": 76337, "tid": -914061504, "ts": 1716454216445884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445885, "dur": 0,
    "args": {
      "External id": 420,
      "cbid": 205, "correlation": 420
    }
  },
  {
    "ph": "f", "id": 420, "pid": 76337, "tid": -914061504, "ts": 1716454216445885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216447707, "dur": 134,
    "args": {
      "External id": 424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 424,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 424, "pid": 5, "tid": 7, "ts": 1716454216447707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445898, "dur": 13,
    "args": {
      "External id": 424,
      "cbid": 211, "correlation": 424
    }
  },
  {
    "ph": "s", "id": 424, "pid": 76337, "tid": -914061504, "ts": 1716454216445898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216447843, "dur": 6,
    "args": {
      "External id": 426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 426,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 426, "pid": 5, "tid": 7, "ts": 1716454216447843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445914, "dur": 5,
    "args": {
      "External id": 426,
      "cbid": 211, "correlation": 426
    }
  },
  {
    "ph": "s", "id": 426, "pid": 76337, "tid": -914061504, "ts": 1716454216445914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216447850, "dur": 5,
    "args": {
      "External id": 428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 428, "pid": 5, "tid": 7, "ts": 1716454216447850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445923, "dur": 5,
    "args": {
      "External id": 428,
      "cbid": 211, "correlation": 428
    }
  },
  {
    "ph": "s", "id": 428, "pid": 76337, "tid": -914061504, "ts": 1716454216445923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445931, "dur": 0,
    "args": {
      "External id": 429,
      "cbid": 51, "correlation": 429
    }
  },
  {
    "ph": "s", "id": 429, "pid": 76337, "tid": -914061504, "ts": 1716454216445931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216447857, "dur": 818,
    "args": {
      "External id": 430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 430,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 430, "pid": 5, "tid": 7, "ts": 1716454216447857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445932, "dur": 5,
    "args": {
      "External id": 430,
      "cbid": 211, "correlation": 430
    }
  },
  {
    "ph": "s", "id": 430, "pid": 76337, "tid": -914061504, "ts": 1716454216445932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216448676, "dur": 177,
    "args": {
      "External id": 435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 435, "pid": 5, "tid": 7, "ts": 1716454216448676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216445960, "dur": 9,
    "args": {
      "External id": 435,
      "cbid": 211, "correlation": 435
    }
  },
  {
    "ph": "s", "id": 435, "pid": 76337, "tid": -914061504, "ts": 1716454216445960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216448854, "dur": 751,
    "args": {
      "External id": 443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 443, "pid": 5, "tid": 7, "ts": 1716454216448854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446008, "dur": 10,
    "args": {
      "External id": 443,
      "cbid": 211, "correlation": 443
    }
  },
  {
    "ph": "s", "id": 443, "pid": 76337, "tid": -914061504, "ts": 1716454216446008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216449606, "dur": 103,
    "args": {
      "External id": 451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 451,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 451, "pid": 5, "tid": 7, "ts": 1716454216449606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446053, "dur": 10,
    "args": {
      "External id": 451,
      "cbid": 211, "correlation": 451
    }
  },
  {
    "ph": "s", "id": 451, "pid": 76337, "tid": -914061504, "ts": 1716454216446053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216449716, "dur": 88,
    "args": {
      "External id": 461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 461,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 617.2125,
      "warps per SM": 2468.85,
      "grid": [49377, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 461, "pid": 5, "tid": 7, "ts": 1716454216449716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446305, "dur": 64,
    "args": {
      "External id": 461,
      "cbid": 211, "correlation": 461
    }
  },
  {
    "ph": "s", "id": 461, "pid": 76337, "tid": -914061504, "ts": 1716454216446305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216449806, "dur": 166,
    "args": {
      "External id": 466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 466, "pid": 5, "tid": 7, "ts": 1716454216449806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446407, "dur": 10,
    "args": {
      "External id": 466,
      "cbid": 211, "correlation": 466
    }
  },
  {
    "ph": "s", "id": 466, "pid": 76337, "tid": -914061504, "ts": 1716454216446407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216449973, "dur": 6,
    "args": {
      "External id": 481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 481,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 3.6,
      "warps per SM": 14.4,
      "grid": [288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 22
    }
  },
  {
    "ph": "f", "id": 481, "pid": 5, "tid": 7, "ts": 1716454216449973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446522, "dur": 16,
    "args": {
      "External id": 481,
      "cbid": 211, "correlation": 481
    }
  },
  {
    "ph": "s", "id": 481, "pid": 76337, "tid": -914061504, "ts": 1716454216446522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446547, "dur": 0,
    "args": {
      "External id": 488,
      "cbid": 317, "correlation": 488
    }
  },
  {
    "ph": "f", "id": 488, "pid": 76337, "tid": -914061504, "ts": 1716454216446547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446548, "dur": 0,
    "args": {
      "External id": 489,
      "cbid": 203, "correlation": 489
    }
  },
  {
    "ph": "f", "id": 489, "pid": 76337, "tid": -914061504, "ts": 1716454216446548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216446549, "dur": 0,
    "args": {
      "External id": 490,
      "cbid": 205, "correlation": 490
    }
  },
  {
    "ph": "f", "id": 490, "pid": 76337, "tid": -914061504, "ts": 1716454216446549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216450020, "dur": 255,
    "args": {
      "External id": 494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 494,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [1, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 494, "pid": 5, "tid": 7, "ts": 1716454216450020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216449771, "dur": 251,
    "args": {
      "External id": 494,
      "cbid": 211, "correlation": 494
    }
  },
  {
    "ph": "s", "id": 494, "pid": 76337, "tid": -914061504, "ts": 1716454216449771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216450803, "dur": 50,
    "args": {
      "External id": 500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 500, "pid": 5, "tid": 7, "ts": 1716454216450803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216450788, "dur": 14,
    "args": {
      "External id": 500,
      "cbid": 211, "correlation": 500
    }
  },
  {
    "ph": "s", "id": 500, "pid": 76337, "tid": -914061504, "ts": 1716454216450788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216451030, "dur": 144,
    "args": {
      "External id": 510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 510, "pid": 5, "tid": 7, "ts": 1716454216451030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451014, "dur": 16,
    "args": {
      "External id": 510,
      "cbid": 211, "correlation": 510
    }
  },
  {
    "ph": "s", "id": 510, "pid": 76337, "tid": -914061504, "ts": 1716454216451014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216451176, "dur": 179,
    "args": {
      "External id": 531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 531,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 531, "pid": 5, "tid": 7, "ts": 1716454216451176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451075, "dur": 11,
    "args": {
      "External id": 531,
      "cbid": 211, "correlation": 531
    }
  },
  {
    "ph": "s", "id": 531, "pid": 76337, "tid": -914061504, "ts": 1716454216451075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216451356, "dur": 5,
    "args": {
      "External id": 543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 543,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 543, "pid": 5, "tid": 7, "ts": 1716454216451356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451097, "dur": 7,
    "args": {
      "External id": 543,
      "cbid": 211, "correlation": 543
    }
  },
  {
    "ph": "s", "id": 543, "pid": 76337, "tid": -914061504, "ts": 1716454216451097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216451362, "dur": 49,
    "args": {
      "External id": 546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 546, "pid": 5, "tid": 7, "ts": 1716454216451362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451128, "dur": 8,
    "args": {
      "External id": 546,
      "cbid": 211, "correlation": 546
    }
  },
  {
    "ph": "s", "id": 546, "pid": 76337, "tid": -914061504, "ts": 1716454216451128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216451413, "dur": 31,
    "args": {
      "External id": 555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 555,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 555, "pid": 5, "tid": 7, "ts": 1716454216451413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451188, "dur": 11,
    "args": {
      "External id": 555,
      "cbid": 211, "correlation": 555
    }
  },
  {
    "ph": "s", "id": 555, "pid": 76337, "tid": -914061504, "ts": 1716454216451188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451287, "dur": 0,
    "args": {
      "External id": 565,
      "cbid": 317, "correlation": 565
    }
  },
  {
    "ph": "f", "id": 565, "pid": 76337, "tid": -914061504, "ts": 1716454216451287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451288, "dur": 0,
    "args": {
      "External id": 566,
      "cbid": 203, "correlation": 566
    }
  },
  {
    "ph": "f", "id": 566, "pid": 76337, "tid": -914061504, "ts": 1716454216451288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216451290, "dur": 0,
    "args": {
      "External id": 567,
      "cbid": 205, "correlation": 567
    }
  },
  {
    "ph": "f", "id": 567, "pid": 76337, "tid": -914061504, "ts": 1716454216451290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216454315, "dur": 38,
    "args": {
      "External id": 571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 571,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [1536, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 571, "pid": 5, "tid": 7, "ts": 1716454216454315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216454288, "dur": 28,
    "args": {
      "External id": 571,
      "cbid": 211, "correlation": 571
    }
  },
  {
    "ph": "s", "id": 571, "pid": 76337, "tid": -914061504, "ts": 1716454216454288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216454354, "dur": 8,
    "args": {
      "External id": 573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 573,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12.8,
      "warps per SM": 102.4,
      "grid": [1, 4, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 573, "pid": 5, "tid": 7, "ts": 1716454216454354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216454319, "dur": 6,
    "args": {
      "External id": 573,
      "cbid": 211, "correlation": 573
    }
  },
  {
    "ph": "s", "id": 573, "pid": 76337, "tid": -914061504, "ts": 1716454216454319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216454364, "dur": 4,
    "args": {
      "External id": 575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 575, "pid": 5, "tid": 7, "ts": 1716454216454364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216454332, "dur": 9,
    "args": {
      "External id": 575,
      "cbid": 211, "correlation": 575
    }
  },
  {
    "ph": "s", "id": 575, "pid": 76337, "tid": -914061504, "ts": 1716454216454332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216454345, "dur": 0,
    "args": {
      "External id": 576,
      "cbid": 51, "correlation": 576
    }
  },
  {
    "ph": "s", "id": 576, "pid": 76337, "tid": -914061504, "ts": 1716454216454345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216454482, "dur": 428,
    "args": {
      "External id": 577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 577,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [192, 4, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 577, "pid": 5, "tid": 7, "ts": 1716454216454482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216454369, "dur": 113,
    "args": {
      "External id": 577,
      "cbid": 211, "correlation": 577
    }
  },
  {
    "ph": "s", "id": 577, "pid": 76337, "tid": -914061504, "ts": 1716454216454369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216455262, "dur": 92,
    "args": {
      "External id": 582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 582, "pid": 5, "tid": 7, "ts": 1716454216455262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455247, "dur": 14,
    "args": {
      "External id": 582,
      "cbid": 211, "correlation": 582
    }
  },
  {
    "ph": "s", "id": 582, "pid": 76337, "tid": -914061504, "ts": 1716454216455247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216455405, "dur": 362,
    "args": {
      "External id": 602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 602,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 602, "pid": 5, "tid": 7, "ts": 1716454216455405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455391, "dur": 14,
    "args": {
      "External id": 602,
      "cbid": 211, "correlation": 602
    }
  },
  {
    "ph": "s", "id": 602, "pid": 76337, "tid": -914061504, "ts": 1716454216455391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216455769, "dur": 5,
    "args": {
      "External id": 614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 614,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 614, "pid": 5, "tid": 7, "ts": 1716454216455769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455425, "dur": 10,
    "args": {
      "External id": 614,
      "cbid": 211, "correlation": 614
    }
  },
  {
    "ph": "s", "id": 614, "pid": 76337, "tid": -914061504, "ts": 1716454216455425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216455775, "dur": 92,
    "args": {
      "External id": 617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 617,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 617, "pid": 5, "tid": 7, "ts": 1716454216455775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455455, "dur": 8,
    "args": {
      "External id": 617,
      "cbid": 211, "correlation": 617
    }
  },
  {
    "ph": "s", "id": 617, "pid": 76337, "tid": -914061504, "ts": 1716454216455455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216455868, "dur": 58,
    "args": {
      "External id": 626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 626,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 626, "pid": 5, "tid": 7, "ts": 1716454216455868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455524, "dur": 12,
    "args": {
      "External id": 626,
      "cbid": 211, "correlation": 626
    }
  },
  {
    "ph": "s", "id": 626, "pid": 76337, "tid": -914061504, "ts": 1716454216455524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455640, "dur": 0,
    "args": {
      "External id": 636,
      "cbid": 317, "correlation": 636
    }
  },
  {
    "ph": "f", "id": 636, "pid": 76337, "tid": -914061504, "ts": 1716454216455640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455641, "dur": 1,
    "args": {
      "External id": 637,
      "cbid": 203, "correlation": 637
    }
  },
  {
    "ph": "f", "id": 637, "pid": 76337, "tid": -914061504, "ts": 1716454216455641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216455642, "dur": 1,
    "args": {
      "External id": 638,
      "cbid": 205, "correlation": 638
    }
  },
  {
    "ph": "f", "id": 638, "pid": 76337, "tid": -914061504, "ts": 1716454216455642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216458605, "dur": 104,
    "args": {
      "External id": 642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 642,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 642, "pid": 5, "tid": 7, "ts": 1716454216458605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216458579, "dur": 27,
    "args": {
      "External id": 642,
      "cbid": 211, "correlation": 642
    }
  },
  {
    "ph": "s", "id": 642, "pid": 76337, "tid": -914061504, "ts": 1716454216458579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216458711, "dur": 12,
    "args": {
      "External id": 644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 644,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 644, "pid": 5, "tid": 7, "ts": 1716454216458711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216458609, "dur": 6,
    "args": {
      "External id": 644,
      "cbid": 211, "correlation": 644
    }
  },
  {
    "ph": "s", "id": 644, "pid": 76337, "tid": -914061504, "ts": 1716454216458609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216458724, "dur": 4,
    "args": {
      "External id": 646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 646, "pid": 5, "tid": 7, "ts": 1716454216458724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216458622, "dur": 9,
    "args": {
      "External id": 646,
      "cbid": 211, "correlation": 646
    }
  },
  {
    "ph": "s", "id": 646, "pid": 76337, "tid": -914061504, "ts": 1716454216458622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216458635, "dur": 0,
    "args": {
      "External id": 647,
      "cbid": 51, "correlation": 647
    }
  },
  {
    "ph": "s", "id": 647, "pid": 76337, "tid": -914061504, "ts": 1716454216458635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216458729, "dur": 775,
    "args": {
      "External id": 648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 648,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 648, "pid": 5, "tid": 7, "ts": 1716454216458729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216458636, "dur": 7,
    "args": {
      "External id": 648,
      "cbid": 211, "correlation": 648
    }
  },
  {
    "ph": "s", "id": 648, "pid": 76337, "tid": -914061504, "ts": 1716454216458636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216459506, "dur": 91,
    "args": {
      "External id": 653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 653,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 653, "pid": 5, "tid": 7, "ts": 1716454216459506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216459386, "dur": 14,
    "args": {
      "External id": 653,
      "cbid": 211, "correlation": 653
    }
  },
  {
    "ph": "s", "id": 653, "pid": 76337, "tid": -914061504, "ts": 1716454216459386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216459489, "dur": 0,
    "args": {
      "External id": 663,
      "cbid": 317, "correlation": 663
    }
  },
  {
    "ph": "f", "id": 663, "pid": 76337, "tid": -914061504, "ts": 1716454216459489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216459490, "dur": 1,
    "args": {
      "External id": 664,
      "cbid": 203, "correlation": 664
    }
  },
  {
    "ph": "f", "id": 664, "pid": 76337, "tid": -914061504, "ts": 1716454216459490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216459491, "dur": 0,
    "args": {
      "External id": 665,
      "cbid": 205, "correlation": 665
    }
  },
  {
    "ph": "f", "id": 665, "pid": 76337, "tid": -914061504, "ts": 1716454216459491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216461618, "dur": 147,
    "args": {
      "External id": 669,
      "cbid": 273, "correlation": 669
    }
  },
  {
    "ph": "f", "id": 669, "pid": 76337, "tid": -914061504, "ts": 1716454216461618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216461767, "dur": 71,
    "args": {
      "External id": 670,
      "cbid": 273, "correlation": 670
    }
  },
  {
    "ph": "f", "id": 670, "pid": 76337, "tid": -914061504, "ts": 1716454216461767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216461838, "dur": 72,
    "args": {
      "External id": 671,
      "cbid": 273, "correlation": 671
    }
  },
  {
    "ph": "f", "id": 671, "pid": 76337, "tid": -914061504, "ts": 1716454216461838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216461912, "dur": 50,
    "args": {
      "External id": 672,
      "cbid": 273, "correlation": 672
    }
  },
  {
    "ph": "f", "id": 672, "pid": 76337, "tid": -914061504, "ts": 1716454216461912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216461962, "dur": 69,
    "args": {
      "External id": 673,
      "cbid": 273, "correlation": 673
    }
  },
  {
    "ph": "f", "id": 673, "pid": 76337, "tid": -914061504, "ts": 1716454216461962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462033, "dur": 48,
    "args": {
      "External id": 674,
      "cbid": 273, "correlation": 674
    }
  },
  {
    "ph": "f", "id": 674, "pid": 76337, "tid": -914061504, "ts": 1716454216462033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462082, "dur": 55,
    "args": {
      "External id": 675,
      "cbid": 273, "correlation": 675
    }
  },
  {
    "ph": "f", "id": 675, "pid": 76337, "tid": -914061504, "ts": 1716454216462082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462138, "dur": 43,
    "args": {
      "External id": 676,
      "cbid": 273, "correlation": 676
    }
  },
  {
    "ph": "f", "id": 676, "pid": 76337, "tid": -914061504, "ts": 1716454216462138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462182, "dur": 59,
    "args": {
      "External id": 677,
      "cbid": 273, "correlation": 677
    }
  },
  {
    "ph": "f", "id": 677, "pid": 76337, "tid": -914061504, "ts": 1716454216462182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462242, "dur": 50,
    "args": {
      "External id": 678,
      "cbid": 273, "correlation": 678
    }
  },
  {
    "ph": "f", "id": 678, "pid": 76337, "tid": -914061504, "ts": 1716454216462242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462292, "dur": 49,
    "args": {
      "External id": 679,
      "cbid": 273, "correlation": 679
    }
  },
  {
    "ph": "f", "id": 679, "pid": 76337, "tid": -914061504, "ts": 1716454216462292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462343, "dur": 43,
    "args": {
      "External id": 680,
      "cbid": 273, "correlation": 680
    }
  },
  {
    "ph": "f", "id": 680, "pid": 76337, "tid": -914061504, "ts": 1716454216462343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462386, "dur": 42,
    "args": {
      "External id": 681,
      "cbid": 273, "correlation": 681
    }
  },
  {
    "ph": "f", "id": 681, "pid": 76337, "tid": -914061504, "ts": 1716454216462386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462429, "dur": 52,
    "args": {
      "External id": 682,
      "cbid": 273, "correlation": 682
    }
  },
  {
    "ph": "f", "id": 682, "pid": 76337, "tid": -914061504, "ts": 1716454216462429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462482, "dur": 62,
    "args": {
      "External id": 683,
      "cbid": 273, "correlation": 683
    }
  },
  {
    "ph": "f", "id": 683, "pid": 76337, "tid": -914061504, "ts": 1716454216462482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462546, "dur": 53,
    "args": {
      "External id": 684,
      "cbid": 273, "correlation": 684
    }
  },
  {
    "ph": "f", "id": 684, "pid": 76337, "tid": -914061504, "ts": 1716454216462546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462600, "dur": 58,
    "args": {
      "External id": 685,
      "cbid": 273, "correlation": 685
    }
  },
  {
    "ph": "f", "id": 685, "pid": 76337, "tid": -914061504, "ts": 1716454216462600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462659, "dur": 44,
    "args": {
      "External id": 686,
      "cbid": 273, "correlation": 686
    }
  },
  {
    "ph": "f", "id": 686, "pid": 76337, "tid": -914061504, "ts": 1716454216462659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462703, "dur": 41,
    "args": {
      "External id": 687,
      "cbid": 273, "correlation": 687
    }
  },
  {
    "ph": "f", "id": 687, "pid": 76337, "tid": -914061504, "ts": 1716454216462703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462746, "dur": 46,
    "args": {
      "External id": 688,
      "cbid": 273, "correlation": 688
    }
  },
  {
    "ph": "f", "id": 688, "pid": 76337, "tid": -914061504, "ts": 1716454216462746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462792, "dur": 44,
    "args": {
      "External id": 689,
      "cbid": 273, "correlation": 689
    }
  },
  {
    "ph": "f", "id": 689, "pid": 76337, "tid": -914061504, "ts": 1716454216462792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462837, "dur": 51,
    "args": {
      "External id": 690,
      "cbid": 273, "correlation": 690
    }
  },
  {
    "ph": "f", "id": 690, "pid": 76337, "tid": -914061504, "ts": 1716454216462837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462890, "dur": 48,
    "args": {
      "External id": 691,
      "cbid": 273, "correlation": 691
    }
  },
  {
    "ph": "f", "id": 691, "pid": 76337, "tid": -914061504, "ts": 1716454216462890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462939, "dur": 53,
    "args": {
      "External id": 692,
      "cbid": 273, "correlation": 692
    }
  },
  {
    "ph": "f", "id": 692, "pid": 76337, "tid": -914061504, "ts": 1716454216462939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216462994, "dur": 57,
    "args": {
      "External id": 693,
      "cbid": 273, "correlation": 693
    }
  },
  {
    "ph": "f", "id": 693, "pid": 76337, "tid": -914061504, "ts": 1716454216462994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216463053, "dur": 44,
    "args": {
      "External id": 694,
      "cbid": 273, "correlation": 694
    }
  },
  {
    "ph": "f", "id": 694, "pid": 76337, "tid": -914061504, "ts": 1716454216463053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216463098, "dur": 60,
    "args": {
      "External id": 695,
      "cbid": 273, "correlation": 695
    }
  },
  {
    "ph": "f", "id": 695, "pid": 76337, "tid": -914061504, "ts": 1716454216463098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216463158, "dur": 43,
    "args": {
      "External id": 696,
      "cbid": 273, "correlation": 696
    }
  },
  {
    "ph": "f", "id": 696, "pid": 76337, "tid": -914061504, "ts": 1716454216463158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216463202, "dur": 10711,
    "args": {
      "External id": 697,
      "cbid": 273, "correlation": 697
    }
  },
  {
    "ph": "f", "id": 697, "pid": 76337, "tid": -914061504, "ts": 1716454216463202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216473915, "dur": 52,
    "args": {
      "External id": 698,
      "cbid": 273, "correlation": 698
    }
  },
  {
    "ph": "f", "id": 698, "pid": 76337, "tid": -914061504, "ts": 1716454216473915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216473968, "dur": 69,
    "args": {
      "External id": 699,
      "cbid": 273, "correlation": 699
    }
  },
  {
    "ph": "f", "id": 699, "pid": 76337, "tid": -914061504, "ts": 1716454216473968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474038, "dur": 52,
    "args": {
      "External id": 700,
      "cbid": 273, "correlation": 700
    }
  },
  {
    "ph": "f", "id": 700, "pid": 76337, "tid": -914061504, "ts": 1716454216474038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474091, "dur": 43,
    "args": {
      "External id": 701,
      "cbid": 273, "correlation": 701
    }
  },
  {
    "ph": "f", "id": 701, "pid": 76337, "tid": -914061504, "ts": 1716454216474091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474134, "dur": 44,
    "args": {
      "External id": 702,
      "cbid": 273, "correlation": 702
    }
  },
  {
    "ph": "f", "id": 702, "pid": 76337, "tid": -914061504, "ts": 1716454216474134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474179, "dur": 45,
    "args": {
      "External id": 703,
      "cbid": 273, "correlation": 703
    }
  },
  {
    "ph": "f", "id": 703, "pid": 76337, "tid": -914061504, "ts": 1716454216474179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474225, "dur": 43,
    "args": {
      "External id": 704,
      "cbid": 273, "correlation": 704
    }
  },
  {
    "ph": "f", "id": 704, "pid": 76337, "tid": -914061504, "ts": 1716454216474225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474270, "dur": 45,
    "args": {
      "External id": 705,
      "cbid": 273, "correlation": 705
    }
  },
  {
    "ph": "f", "id": 705, "pid": 76337, "tid": -914061504, "ts": 1716454216474270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474316, "dur": 43,
    "args": {
      "External id": 706,
      "cbid": 273, "correlation": 706
    }
  },
  {
    "ph": "f", "id": 706, "pid": 76337, "tid": -914061504, "ts": 1716454216474316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474360, "dur": 46,
    "args": {
      "External id": 707,
      "cbid": 273, "correlation": 707
    }
  },
  {
    "ph": "f", "id": 707, "pid": 76337, "tid": -914061504, "ts": 1716454216474360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474407, "dur": 45,
    "args": {
      "External id": 708,
      "cbid": 273, "correlation": 708
    }
  },
  {
    "ph": "f", "id": 708, "pid": 76337, "tid": -914061504, "ts": 1716454216474407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474454, "dur": 47,
    "args": {
      "External id": 709,
      "cbid": 273, "correlation": 709
    }
  },
  {
    "ph": "f", "id": 709, "pid": 76337, "tid": -914061504, "ts": 1716454216474454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474501, "dur": 41,
    "args": {
      "External id": 710,
      "cbid": 273, "correlation": 710
    }
  },
  {
    "ph": "f", "id": 710, "pid": 76337, "tid": -914061504, "ts": 1716454216474501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474543, "dur": 43,
    "args": {
      "External id": 711,
      "cbid": 273, "correlation": 711
    }
  },
  {
    "ph": "f", "id": 711, "pid": 76337, "tid": -914061504, "ts": 1716454216474543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474587, "dur": 45,
    "args": {
      "External id": 712,
      "cbid": 273, "correlation": 712
    }
  },
  {
    "ph": "f", "id": 712, "pid": 76337, "tid": -914061504, "ts": 1716454216474587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474633, "dur": 49,
    "args": {
      "External id": 713,
      "cbid": 273, "correlation": 713
    }
  },
  {
    "ph": "f", "id": 713, "pid": 76337, "tid": -914061504, "ts": 1716454216474633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474683, "dur": 46,
    "args": {
      "External id": 714,
      "cbid": 273, "correlation": 714
    }
  },
  {
    "ph": "f", "id": 714, "pid": 76337, "tid": -914061504, "ts": 1716454216474683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474730, "dur": 48,
    "args": {
      "External id": 715,
      "cbid": 273, "correlation": 715
    }
  },
  {
    "ph": "f", "id": 715, "pid": 76337, "tid": -914061504, "ts": 1716454216474730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncSetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474779, "dur": 47,
    "args": {
      "External id": 716,
      "cbid": 273, "correlation": 716
    }
  },
  {
    "ph": "f", "id": 716, "pid": 76337, "tid": -914061504, "ts": 1716454216474779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216474958, "dur": 72,
    "args": {
      "External id": 717,
      "cbid": 251, "correlation": 717
    }
  },
  {
    "ph": "f", "id": 717, "pid": 76337, "tid": -914061504, "ts": 1716454216474958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216475092, "dur": 80,
    "args": {
      "External id": 718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 718,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [2, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 718, "pid": 5, "tid": 7, "ts": 1716454216475092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216475052, "dur": 43,
    "args": {
      "External id": 718,
      "cbid": 211, "correlation": 718
    }
  },
  {
    "ph": "s", "id": 718, "pid": 76337, "tid": -914061504, "ts": 1716454216475052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216475823, "dur": 92,
    "args": {
      "External id": 724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 724,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 724, "pid": 5, "tid": 7, "ts": 1716454216475823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216475806, "dur": 15,
    "args": {
      "External id": 724,
      "cbid": 211, "correlation": 724
    }
  },
  {
    "ph": "s", "id": 724, "pid": 76337, "tid": -914061504, "ts": 1716454216475806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216475923, "dur": 253,
    "args": {
      "External id": 732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 732,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 732, "pid": 5, "tid": 7, "ts": 1716454216475923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216475910, "dur": 11,
    "args": {
      "External id": 732,
      "cbid": 211, "correlation": 732
    }
  },
  {
    "ph": "s", "id": 732, "pid": 76337, "tid": -914061504, "ts": 1716454216475910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216476178, "dur": 53,
    "args": {
      "External id": 740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 740,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 740, "pid": 5, "tid": 7, "ts": 1716454216476178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216475988, "dur": 12,
    "args": {
      "External id": 740,
      "cbid": 211, "correlation": 740
    }
  },
  {
    "ph": "s", "id": 740, "pid": 76337, "tid": -914061504, "ts": 1716454216475988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216476232, "dur": 371,
    "args": {
      "External id": 750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 750, "pid": 5, "tid": 7, "ts": 1716454216476232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476201, "dur": 19,
    "args": {
      "External id": 750,
      "cbid": 211, "correlation": 750
    }
  },
  {
    "ph": "s", "id": 750, "pid": 76337, "tid": -914061504, "ts": 1716454216476201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216476605, "dur": 375,
    "args": {
      "External id": 771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 771,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 771, "pid": 5, "tid": 7, "ts": 1716454216476605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476257, "dur": 10,
    "args": {
      "External id": 771,
      "cbid": 211, "correlation": 771
    }
  },
  {
    "ph": "s", "id": 771, "pid": 76337, "tid": -914061504, "ts": 1716454216476257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216476981, "dur": 5,
    "args": {
      "External id": 783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 783,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 783, "pid": 5, "tid": 7, "ts": 1716454216476981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476281, "dur": 7,
    "args": {
      "External id": 783,
      "cbid": 211, "correlation": 783
    }
  },
  {
    "ph": "s", "id": 783, "pid": 76337, "tid": -914061504, "ts": 1716454216476281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216476987, "dur": 91,
    "args": {
      "External id": 786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 786, "pid": 5, "tid": 7, "ts": 1716454216476987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476304, "dur": 7,
    "args": {
      "External id": 786,
      "cbid": 211, "correlation": 786
    }
  },
  {
    "ph": "s", "id": 786, "pid": 76337, "tid": -914061504, "ts": 1716454216476304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216477080, "dur": 58,
    "args": {
      "External id": 795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 795,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 795, "pid": 5, "tid": 7, "ts": 1716454216477080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476365, "dur": 12,
    "args": {
      "External id": 795,
      "cbid": 211, "correlation": 795
    }
  },
  {
    "ph": "s", "id": 795, "pid": 76337, "tid": -914061504, "ts": 1716454216476365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476460, "dur": 1,
    "args": {
      "External id": 805,
      "cbid": 317, "correlation": 805
    }
  },
  {
    "ph": "f", "id": 805, "pid": 76337, "tid": -914061504, "ts": 1716454216476460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476462, "dur": 1,
    "args": {
      "External id": 806,
      "cbid": 203, "correlation": 806
    }
  },
  {
    "ph": "f", "id": 806, "pid": 76337, "tid": -914061504, "ts": 1716454216476462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476464, "dur": 1,
    "args": {
      "External id": 807,
      "cbid": 205, "correlation": 807
    }
  },
  {
    "ph": "f", "id": 807, "pid": 76337, "tid": -914061504, "ts": 1716454216476464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216477139, "dur": 100,
    "args": {
      "External id": 811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 811,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 811, "pid": 5, "tid": 7, "ts": 1716454216477139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476497, "dur": 14,
    "args": {
      "External id": 811,
      "cbid": 211, "correlation": 811
    }
  },
  {
    "ph": "s", "id": 811, "pid": 76337, "tid": -914061504, "ts": 1716454216476497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216477241, "dur": 12,
    "args": {
      "External id": 813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 813, "pid": 5, "tid": 7, "ts": 1716454216477241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476514, "dur": 5,
    "args": {
      "External id": 813,
      "cbid": 211, "correlation": 813
    }
  },
  {
    "ph": "s", "id": 813, "pid": 76337, "tid": -914061504, "ts": 1716454216476514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216477254, "dur": 4,
    "args": {
      "External id": 815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 815, "pid": 5, "tid": 7, "ts": 1716454216477254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476530, "dur": 9,
    "args": {
      "External id": 815,
      "cbid": 211, "correlation": 815
    }
  },
  {
    "ph": "s", "id": 815, "pid": 76337, "tid": -914061504, "ts": 1716454216476530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476546, "dur": 0,
    "args": {
      "External id": 816,
      "cbid": 51, "correlation": 816
    }
  },
  {
    "ph": "s", "id": 816, "pid": 76337, "tid": -914061504, "ts": 1716454216476546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216477260, "dur": 777,
    "args": {
      "External id": 817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 817,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 817, "pid": 5, "tid": 7, "ts": 1716454216477260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476547, "dur": 8,
    "args": {
      "External id": 817,
      "cbid": 211, "correlation": 817
    }
  },
  {
    "ph": "s", "id": 817, "pid": 76337, "tid": -914061504, "ts": 1716454216476547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216478038, "dur": 91,
    "args": {
      "External id": 822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 822, "pid": 5, "tid": 7, "ts": 1716454216478038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476579, "dur": 9,
    "args": {
      "External id": 822,
      "cbid": 211, "correlation": 822
    }
  },
  {
    "ph": "s", "id": 822, "pid": 76337, "tid": -914061504, "ts": 1716454216476579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216478131, "dur": 364,
    "args": {
      "External id": 842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 842,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 842, "pid": 5, "tid": 7, "ts": 1716454216478131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476666, "dur": 13,
    "args": {
      "External id": 842,
      "cbid": 211, "correlation": 842
    }
  },
  {
    "ph": "s", "id": 842, "pid": 76337, "tid": -914061504, "ts": 1716454216476666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216478497, "dur": 4,
    "args": {
      "External id": 854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 854,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 854, "pid": 5, "tid": 7, "ts": 1716454216478497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476689, "dur": 6,
    "args": {
      "External id": 854,
      "cbid": 211, "correlation": 854
    }
  },
  {
    "ph": "s", "id": 854, "pid": 76337, "tid": -914061504, "ts": 1716454216476689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216478502, "dur": 92,
    "args": {
      "External id": 857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 857, "pid": 5, "tid": 7, "ts": 1716454216478502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476707, "dur": 7,
    "args": {
      "External id": 857,
      "cbid": 211, "correlation": 857
    }
  },
  {
    "ph": "s", "id": 857, "pid": 76337, "tid": -914061504, "ts": 1716454216476707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216478595, "dur": 57,
    "args": {
      "External id": 866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 866,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 866, "pid": 5, "tid": 7, "ts": 1716454216478595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476748, "dur": 10,
    "args": {
      "External id": 866,
      "cbid": 211, "correlation": 866
    }
  },
  {
    "ph": "s", "id": 866, "pid": 76337, "tid": -914061504, "ts": 1716454216476748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476837, "dur": 0,
    "args": {
      "External id": 876,
      "cbid": 317, "correlation": 876
    }
  },
  {
    "ph": "f", "id": 876, "pid": 76337, "tid": -914061504, "ts": 1716454216476837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476837, "dur": 0,
    "args": {
      "External id": 877,
      "cbid": 203, "correlation": 877
    }
  },
  {
    "ph": "f", "id": 877, "pid": 76337, "tid": -914061504, "ts": 1716454216476837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476838, "dur": 0,
    "args": {
      "External id": 878,
      "cbid": 205, "correlation": 878
    }
  },
  {
    "ph": "f", "id": 878, "pid": 76337, "tid": -914061504, "ts": 1716454216476838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216478654, "dur": 103,
    "args": {
      "External id": 882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 882,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 882, "pid": 5, "tid": 7, "ts": 1716454216478654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476852, "dur": 13,
    "args": {
      "External id": 882,
      "cbid": 211, "correlation": 882
    }
  },
  {
    "ph": "s", "id": 882, "pid": 76337, "tid": -914061504, "ts": 1716454216476852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216478758, "dur": 12,
    "args": {
      "External id": 884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 884,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 884, "pid": 5, "tid": 7, "ts": 1716454216478758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476867, "dur": 5,
    "args": {
      "External id": 884,
      "cbid": 211, "correlation": 884
    }
  },
  {
    "ph": "s", "id": 884, "pid": 76337, "tid": -914061504, "ts": 1716454216476867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216478772, "dur": 4,
    "args": {
      "External id": 886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 886, "pid": 5, "tid": 7, "ts": 1716454216478772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476876, "dur": 5,
    "args": {
      "External id": 886,
      "cbid": 211, "correlation": 886
    }
  },
  {
    "ph": "s", "id": 886, "pid": 76337, "tid": -914061504, "ts": 1716454216476876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476885, "dur": 0,
    "args": {
      "External id": 887,
      "cbid": 51, "correlation": 887
    }
  },
  {
    "ph": "s", "id": 887, "pid": 76337, "tid": -914061504, "ts": 1716454216476885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216478777, "dur": 771,
    "args": {
      "External id": 888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 888,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 888, "pid": 5, "tid": 7, "ts": 1716454216478777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476886, "dur": 6,
    "args": {
      "External id": 888,
      "cbid": 211, "correlation": 888
    }
  },
  {
    "ph": "s", "id": 888, "pid": 76337, "tid": -914061504, "ts": 1716454216476886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216479550, "dur": 91,
    "args": {
      "External id": 893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 893,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 893, "pid": 5, "tid": 7, "ts": 1716454216479550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476914, "dur": 9,
    "args": {
      "External id": 893,
      "cbid": 211, "correlation": 893
    }
  },
  {
    "ph": "s", "id": 893, "pid": 76337, "tid": -914061504, "ts": 1716454216476914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216479642, "dur": 255,
    "args": {
      "External id": 901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 901, "pid": 5, "tid": 7, "ts": 1716454216479642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476950, "dur": 9,
    "args": {
      "External id": 901,
      "cbid": 211, "correlation": 901
    }
  },
  {
    "ph": "s", "id": 901, "pid": 76337, "tid": -914061504, "ts": 1716454216476950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216479899, "dur": 54,
    "args": {
      "External id": 909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 909,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 909, "pid": 5, "tid": 7, "ts": 1716454216479899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216476987, "dur": 9,
    "args": {
      "External id": 909,
      "cbid": 211, "correlation": 909
    }
  },
  {
    "ph": "s", "id": 909, "pid": 76337, "tid": -914061504, "ts": 1716454216476987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216479954, "dur": 46,
    "args": {
      "External id": 919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 919,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 310.0125,
      "warps per SM": 1240.05,
      "grid": [24801, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 919, "pid": 5, "tid": 7, "ts": 1716454216479954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216477099, "dur": 15,
    "args": {
      "External id": 919,
      "cbid": 211, "correlation": 919
    }
  },
  {
    "ph": "s", "id": 919, "pid": 76337, "tid": -914061504, "ts": 1716454216477099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216480001, "dur": 87,
    "args": {
      "External id": 924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 924, "pid": 5, "tid": 7, "ts": 1716454216480001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216477149, "dur": 8,
    "args": {
      "External id": 924,
      "cbid": 211, "correlation": 924
    }
  },
  {
    "ph": "s", "id": 924, "pid": 76337, "tid": -914061504, "ts": 1716454216477149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216480090, "dur": 10,
    "args": {
      "External id": 939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 939,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 939, "pid": 5, "tid": 7, "ts": 1716454216480090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216477218, "dur": 13,
    "args": {
      "External id": 939,
      "cbid": 211, "correlation": 939
    }
  },
  {
    "ph": "s", "id": 939, "pid": 76337, "tid": -914061504, "ts": 1716454216477218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216477239, "dur": 0,
    "args": {
      "External id": 946,
      "cbid": 317, "correlation": 946
    }
  },
  {
    "ph": "f", "id": 946, "pid": 76337, "tid": -914061504, "ts": 1716454216477239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216477240, "dur": 0,
    "args": {
      "External id": 947,
      "cbid": 203, "correlation": 947
    }
  },
  {
    "ph": "f", "id": 947, "pid": 76337, "tid": -914061504, "ts": 1716454216477240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216477240, "dur": 0,
    "args": {
      "External id": 948,
      "cbid": 205, "correlation": 948
    }
  },
  {
    "ph": "f", "id": 948, "pid": 76337, "tid": -914061504, "ts": 1716454216477240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216480485, "dur": 305,
    "args": {
      "External id": 952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 952,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 1.2,
      "warps per SM": 9.6,
      "grid": [1, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 952, "pid": 5, "tid": 7, "ts": 1716454216480485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216480181, "dur": 308,
    "args": {
      "External id": 952,
      "cbid": 211, "correlation": 952
    }
  },
  {
    "ph": "s", "id": 952, "pid": 76337, "tid": -914061504, "ts": 1716454216480181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216481275, "dur": 26,
    "args": {
      "External id": 958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 958, "pid": 5, "tid": 7, "ts": 1716454216481275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481260, "dur": 13,
    "args": {
      "External id": 958,
      "cbid": 211, "correlation": 958
    }
  },
  {
    "ph": "s", "id": 958, "pid": 76337, "tid": -914061504, "ts": 1716454216481260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216481457, "dur": 69,
    "args": {
      "External id": 968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 968, "pid": 5, "tid": 7, "ts": 1716454216481457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481441, "dur": 16,
    "args": {
      "External id": 968,
      "cbid": 211, "correlation": 968
    }
  },
  {
    "ph": "s", "id": 968, "pid": 76337, "tid": -914061504, "ts": 1716454216481441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216481527, "dur": 95,
    "args": {
      "External id": 989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 989,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 989, "pid": 5, "tid": 7, "ts": 1716454216481527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481484, "dur": 9,
    "args": {
      "External id": 989,
      "cbid": 211, "correlation": 989
    }
  },
  {
    "ph": "s", "id": 989, "pid": 76337, "tid": -914061504, "ts": 1716454216481484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216481624, "dur": 5,
    "args": {
      "External id": 1001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1001,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1001, "pid": 5, "tid": 7, "ts": 1716454216481624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481504, "dur": 8,
    "args": {
      "External id": 1001,
      "cbid": 211, "correlation": 1001
    }
  },
  {
    "ph": "s", "id": 1001, "pid": 76337, "tid": -914061504, "ts": 1716454216481504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216481631, "dur": 26,
    "args": {
      "External id": 1004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1004, "pid": 5, "tid": 7, "ts": 1716454216481631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481528, "dur": 7,
    "args": {
      "External id": 1004,
      "cbid": 211, "correlation": 1004
    }
  },
  {
    "ph": "s", "id": 1004, "pid": 76337, "tid": -914061504, "ts": 1716454216481528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216481659, "dur": 18,
    "args": {
      "External id": 1013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1013,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1013, "pid": 5, "tid": 7, "ts": 1716454216481659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481580, "dur": 10,
    "args": {
      "External id": 1013,
      "cbid": 211, "correlation": 1013
    }
  },
  {
    "ph": "s", "id": 1013, "pid": 76337, "tid": -914061504, "ts": 1716454216481580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481675, "dur": 0,
    "args": {
      "External id": 1023,
      "cbid": 317, "correlation": 1023
    }
  },
  {
    "ph": "f", "id": 1023, "pid": 76337, "tid": -914061504, "ts": 1716454216481675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481676, "dur": 1,
    "args": {
      "External id": 1024,
      "cbid": 203, "correlation": 1024
    }
  },
  {
    "ph": "f", "id": 1024, "pid": 76337, "tid": -914061504, "ts": 1716454216481676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216481678, "dur": 0,
    "args": {
      "External id": 1025,
      "cbid": 205, "correlation": 1025
    }
  },
  {
    "ph": "f", "id": 1025, "pid": 76337, "tid": -914061504, "ts": 1716454216481678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216484413, "dur": 20,
    "args": {
      "External id": 1029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1029,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 38.4,
      "warps per SM": 307.2,
      "grid": [384, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1029, "pid": 5, "tid": 7, "ts": 1716454216484413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216484389, "dur": 25,
    "args": {
      "External id": 1029,
      "cbid": 211, "correlation": 1029
    }
  },
  {
    "ph": "s", "id": 1029, "pid": 76337, "tid": -914061504, "ts": 1716454216484389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216484435, "dur": 19,
    "args": {
      "External id": 1031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1031,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 51.2,
      "warps per SM": 409.6,
      "grid": [1, 8, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1031, "pid": 5, "tid": 7, "ts": 1716454216484435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216484417, "dur": 6,
    "args": {
      "External id": 1031,
      "cbid": 211, "correlation": 1031
    }
  },
  {
    "ph": "s", "id": 1031, "pid": 76337, "tid": -914061504, "ts": 1716454216484417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216484461, "dur": 2,
    "args": {
      "External id": 1033,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1033,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.4328358208955223
    }
  },
  {
    "ph": "f", "id": 1033, "pid": 5, "tid": 7, "ts": 1716454216484461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216484433, "dur": 25,
    "args": {
      "External id": 1033,
      "cbid": 51, "correlation": 1033
    }
  },
  {
    "ph": "s", "id": 1033, "pid": 76337, "tid": -914061504, "ts": 1716454216484433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216484739, "dur": 459,
    "args": {
      "External id": 1034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1034,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [4, 96, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1034, "pid": 5, "tid": 7, "ts": 1716454216484739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216484480, "dur": 259,
    "args": {
      "External id": 1034,
      "cbid": 211, "correlation": 1034
    }
  },
  {
    "ph": "s", "id": 1034, "pid": 76337, "tid": -914061504, "ts": 1716454216484480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216485206, "dur": 36,
    "args": {
      "External id": 1036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1036,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1036, "pid": 5, "tid": 7, "ts": 1716454216485206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216484756, "dur": 74,
    "args": {
      "External id": 1036,
      "cbid": 211, "correlation": 1036
    }
  },
  {
    "ph": "s", "id": 1036, "pid": 76337, "tid": -914061504, "ts": 1716454216484756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216485576, "dur": 49,
    "args": {
      "External id": 1042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1042,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1042, "pid": 5, "tid": 7, "ts": 1716454216485576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485561, "dur": 14,
    "args": {
      "External id": 1042,
      "cbid": 211, "correlation": 1042
    }
  },
  {
    "ph": "s", "id": 1042, "pid": 76337, "tid": -914061504, "ts": 1716454216485561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216485721, "dur": 189,
    "args": {
      "External id": 1062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1062,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1062, "pid": 5, "tid": 7, "ts": 1716454216485721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485707, "dur": 14,
    "args": {
      "External id": 1062,
      "cbid": 211, "correlation": 1062
    }
  },
  {
    "ph": "s", "id": 1062, "pid": 76337, "tid": -914061504, "ts": 1716454216485707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216485912, "dur": 5,
    "args": {
      "External id": 1074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1074,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1074, "pid": 5, "tid": 7, "ts": 1716454216485912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485733, "dur": 7,
    "args": {
      "External id": 1074,
      "cbid": 211, "correlation": 1074
    }
  },
  {
    "ph": "s", "id": 1074, "pid": 76337, "tid": -914061504, "ts": 1716454216485733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216485918, "dur": 47,
    "args": {
      "External id": 1077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1077,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1077, "pid": 5, "tid": 7, "ts": 1716454216485918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485759, "dur": 7,
    "args": {
      "External id": 1077,
      "cbid": 211, "correlation": 1077
    }
  },
  {
    "ph": "s", "id": 1077, "pid": 76337, "tid": -914061504, "ts": 1716454216485759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216485966, "dur": 32,
    "args": {
      "External id": 1086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1086,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1086, "pid": 5, "tid": 7, "ts": 1716454216485966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485812, "dur": 11,
    "args": {
      "External id": 1086,
      "cbid": 211, "correlation": 1086
    }
  },
  {
    "ph": "s", "id": 1086, "pid": 76337, "tid": -914061504, "ts": 1716454216485812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485910, "dur": 0,
    "args": {
      "External id": 1096,
      "cbid": 317, "correlation": 1096
    }
  },
  {
    "ph": "f", "id": 1096, "pid": 76337, "tid": -914061504, "ts": 1716454216485910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485911, "dur": 0,
    "args": {
      "External id": 1097,
      "cbid": 203, "correlation": 1097
    }
  },
  {
    "ph": "f", "id": 1097, "pid": 76337, "tid": -914061504, "ts": 1716454216485911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216485912, "dur": 1,
    "args": {
      "External id": 1098,
      "cbid": 205, "correlation": 1098
    }
  },
  {
    "ph": "f", "id": 1098, "pid": 76337, "tid": -914061504, "ts": 1716454216485912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216488468, "dur": 3,
    "args": {
      "External id": 1104,
      "cbid": 317, "correlation": 1104
    }
  },
  {
    "ph": "f", "id": 1104, "pid": 76337, "tid": -914061504, "ts": 1716454216488468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216488473, "dur": 7294,
    "args": {
      "External id": 1105,
      "cbid": 20, "correlation": 1105
    }
  },
  {
    "ph": "f", "id": 1105, "pid": 76337, "tid": -914061504, "ts": 1716454216488473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216495826, "dur": 34,
    "args": {
      "External id": 1106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1106,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1106, "pid": 5, "tid": 7, "ts": 1716454216495826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216495799, "dur": 28,
    "args": {
      "External id": 1106,
      "cbid": 211, "correlation": 1106
    }
  },
  {
    "ph": "s", "id": 1106, "pid": 76337, "tid": -914061504, "ts": 1716454216495799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216495861, "dur": 33,
    "args": {
      "External id": 1108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1108,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1108, "pid": 5, "tid": 7, "ts": 1716454216495861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216495829, "dur": 7,
    "args": {
      "External id": 1108,
      "cbid": 211, "correlation": 1108
    }
  },
  {
    "ph": "s", "id": 1108, "pid": 76337, "tid": -914061504, "ts": 1716454216495829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216495897, "dur": 2,
    "args": {
      "External id": 1110,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1110,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.4328358208955223
    }
  },
  {
    "ph": "f", "id": 1110, "pid": 5, "tid": 7, "ts": 1716454216495897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216495848, "dur": 12,
    "args": {
      "External id": 1110,
      "cbid": 51, "correlation": 1110
    }
  },
  {
    "ph": "s", "id": 1110, "pid": 76337, "tid": -914061504, "ts": 1716454216495848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216495901, "dur": 909,
    "args": {
      "External id": 1111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1111,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1111, "pid": 5, "tid": 7, "ts": 1716454216495901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216495862, "dur": 9,
    "args": {
      "External id": 1111,
      "cbid": 211, "correlation": 1111
    }
  },
  {
    "ph": "s", "id": 1111, "pid": 76337, "tid": -914061504, "ts": 1716454216495862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216496812, "dur": 40,
    "args": {
      "External id": 1113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1113,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1113, "pid": 5, "tid": 7, "ts": 1716454216496812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216495877, "dur": 8,
    "args": {
      "External id": 1113,
      "cbid": 211, "correlation": 1113
    }
  },
  {
    "ph": "s", "id": 1113, "pid": 76337, "tid": -914061504, "ts": 1716454216495877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216496853, "dur": 48,
    "args": {
      "External id": 1119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1119, "pid": 5, "tid": 7, "ts": 1716454216496853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216496624, "dur": 14,
    "args": {
      "External id": 1119,
      "cbid": 211, "correlation": 1119
    }
  },
  {
    "ph": "s", "id": 1119, "pid": 76337, "tid": -914061504, "ts": 1716454216496624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216496764, "dur": 0,
    "args": {
      "External id": 1129,
      "cbid": 317, "correlation": 1129
    }
  },
  {
    "ph": "f", "id": 1129, "pid": 76337, "tid": -914061504, "ts": 1716454216496764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216496765, "dur": 0,
    "args": {
      "External id": 1130,
      "cbid": 203, "correlation": 1130
    }
  },
  {
    "ph": "f", "id": 1130, "pid": 76337, "tid": -914061504, "ts": 1716454216496765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216496766, "dur": 0,
    "args": {
      "External id": 1131,
      "cbid": 205, "correlation": 1131
    }
  },
  {
    "ph": "f", "id": 1131, "pid": 76337, "tid": -914061504, "ts": 1716454216496766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216498615, "dur": 112,
    "args": {
      "External id": 1135,
      "cbid": 251, "correlation": 1135
    }
  },
  {
    "ph": "f", "id": 1135, "pid": 76337, "tid": -914061504, "ts": 1716454216498615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216498753, "dur": 62,
    "args": {
      "External id": 1136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1136,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [4, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1136, "pid": 5, "tid": 7, "ts": 1716454216498753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216498735, "dur": 27,
    "args": {
      "External id": 1136,
      "cbid": 211, "correlation": 1136
    }
  },
  {
    "ph": "s", "id": 1136, "pid": 76337, "tid": -914061504, "ts": 1716454216498735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216499360, "dur": 50,
    "args": {
      "External id": 1142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1142, "pid": 5, "tid": 7, "ts": 1716454216499360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499345, "dur": 14,
    "args": {
      "External id": 1142,
      "cbid": 211, "correlation": 1142
    }
  },
  {
    "ph": "s", "id": 1142, "pid": 76337, "tid": -914061504, "ts": 1716454216499345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216499416, "dur": 138,
    "args": {
      "External id": 1150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1150,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1150, "pid": 5, "tid": 7, "ts": 1716454216499416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499404, "dur": 10,
    "args": {
      "External id": 1150,
      "cbid": 211, "correlation": 1150
    }
  },
  {
    "ph": "s", "id": 1150, "pid": 76337, "tid": -914061504, "ts": 1716454216499404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216499555, "dur": 29,
    "args": {
      "External id": 1158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1158,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1158, "pid": 5, "tid": 7, "ts": 1716454216499555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499459, "dur": 12,
    "args": {
      "External id": 1158,
      "cbid": 211, "correlation": 1158
    }
  },
  {
    "ph": "s", "id": 1158, "pid": 76337, "tid": -914061504, "ts": 1716454216499459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216499644, "dur": 135,
    "args": {
      "External id": 1168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1168, "pid": 5, "tid": 7, "ts": 1716454216499644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499626, "dur": 18,
    "args": {
      "External id": 1168,
      "cbid": 211, "correlation": 1168
    }
  },
  {
    "ph": "s", "id": 1168, "pid": 76337, "tid": -914061504, "ts": 1716454216499626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216499780, "dur": 191,
    "args": {
      "External id": 1189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1189,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1189, "pid": 5, "tid": 7, "ts": 1716454216499780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499679, "dur": 11,
    "args": {
      "External id": 1189,
      "cbid": 211, "correlation": 1189
    }
  },
  {
    "ph": "s", "id": 1189, "pid": 76337, "tid": -914061504, "ts": 1716454216499679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216499972, "dur": 5,
    "args": {
      "External id": 1201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1201,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1201, "pid": 5, "tid": 7, "ts": 1716454216499972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499702, "dur": 7,
    "args": {
      "External id": 1201,
      "cbid": 211, "correlation": 1201
    }
  },
  {
    "ph": "s", "id": 1201, "pid": 76337, "tid": -914061504, "ts": 1716454216499702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216499978, "dur": 47,
    "args": {
      "External id": 1204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1204, "pid": 5, "tid": 7, "ts": 1716454216499978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499727, "dur": 8,
    "args": {
      "External id": 1204,
      "cbid": 211, "correlation": 1204
    }
  },
  {
    "ph": "s", "id": 1204, "pid": 76337, "tid": -914061504, "ts": 1716454216499727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216500027, "dur": 31,
    "args": {
      "External id": 1213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1213,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1213, "pid": 5, "tid": 7, "ts": 1716454216500027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499785, "dur": 13,
    "args": {
      "External id": 1213,
      "cbid": 211, "correlation": 1213
    }
  },
  {
    "ph": "s", "id": 1213, "pid": 76337, "tid": -914061504, "ts": 1716454216499785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499883, "dur": 0,
    "args": {
      "External id": 1223,
      "cbid": 317, "correlation": 1223
    }
  },
  {
    "ph": "f", "id": 1223, "pid": 76337, "tid": -914061504, "ts": 1716454216499883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499884, "dur": 0,
    "args": {
      "External id": 1224,
      "cbid": 203, "correlation": 1224
    }
  },
  {
    "ph": "f", "id": 1224, "pid": 76337, "tid": -914061504, "ts": 1716454216499884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499885, "dur": 0,
    "args": {
      "External id": 1225,
      "cbid": 205, "correlation": 1225
    }
  },
  {
    "ph": "f", "id": 1225, "pid": 76337, "tid": -914061504, "ts": 1716454216499885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216500059, "dur": 34,
    "args": {
      "External id": 1229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1229, "pid": 5, "tid": 7, "ts": 1716454216500059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499905, "dur": 15,
    "args": {
      "External id": 1229,
      "cbid": 211, "correlation": 1229
    }
  },
  {
    "ph": "s", "id": 1229, "pid": 76337, "tid": -914061504, "ts": 1716454216499905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216500095, "dur": 33,
    "args": {
      "External id": 1231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1231, "pid": 5, "tid": 7, "ts": 1716454216500095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499923, "dur": 5,
    "args": {
      "External id": 1231,
      "cbid": 211, "correlation": 1231
    }
  },
  {
    "ph": "s", "id": 1231, "pid": 76337, "tid": -914061504, "ts": 1716454216499923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216500131, "dur": 1,
    "args": {
      "External id": 1233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1233,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.6
    }
  },
  {
    "ph": "f", "id": 1233, "pid": 5, "tid": 7, "ts": 1716454216500131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499938, "dur": 12,
    "args": {
      "External id": 1233,
      "cbid": 51, "correlation": 1233
    }
  },
  {
    "ph": "s", "id": 1233, "pid": 76337, "tid": -914061504, "ts": 1716454216499938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216500135, "dur": 905,
    "args": {
      "External id": 1234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1234,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1234, "pid": 5, "tid": 7, "ts": 1716454216500135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499952, "dur": 9,
    "args": {
      "External id": 1234,
      "cbid": 211, "correlation": 1234
    }
  },
  {
    "ph": "s", "id": 1234, "pid": 76337, "tid": -914061504, "ts": 1716454216499952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216501042, "dur": 37,
    "args": {
      "External id": 1236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1236, "pid": 5, "tid": 7, "ts": 1716454216501042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216499967, "dur": 14,
    "args": {
      "External id": 1236,
      "cbid": 211, "correlation": 1236
    }
  },
  {
    "ph": "s", "id": 1236, "pid": 76337, "tid": -914061504, "ts": 1716454216499967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216501081, "dur": 49,
    "args": {
      "External id": 1242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1242, "pid": 5, "tid": 7, "ts": 1716454216501081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500006, "dur": 10,
    "args": {
      "External id": 1242,
      "cbid": 211, "correlation": 1242
    }
  },
  {
    "ph": "s", "id": 1242, "pid": 76337, "tid": -914061504, "ts": 1716454216500006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216501131, "dur": 189,
    "args": {
      "External id": 1262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1262,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1262, "pid": 5, "tid": 7, "ts": 1716454216501131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500099, "dur": 13,
    "args": {
      "External id": 1262,
      "cbid": 211, "correlation": 1262
    }
  },
  {
    "ph": "s", "id": 1262, "pid": 76337, "tid": -914061504, "ts": 1716454216500099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216501321, "dur": 4,
    "args": {
      "External id": 1274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1274,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1274, "pid": 5, "tid": 7, "ts": 1716454216501321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500130, "dur": 8,
    "args": {
      "External id": 1274,
      "cbid": 211, "correlation": 1274
    }
  },
  {
    "ph": "s", "id": 1274, "pid": 76337, "tid": -914061504, "ts": 1716454216500130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216501326, "dur": 47,
    "args": {
      "External id": 1277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1277, "pid": 5, "tid": 7, "ts": 1716454216501326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500150, "dur": 6,
    "args": {
      "External id": 1277,
      "cbid": 211, "correlation": 1277
    }
  },
  {
    "ph": "s", "id": 1277, "pid": 76337, "tid": -914061504, "ts": 1716454216500150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216501375, "dur": 31,
    "args": {
      "External id": 1286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1286,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1286, "pid": 5, "tid": 7, "ts": 1716454216501375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500195, "dur": 11,
    "args": {
      "External id": 1286,
      "cbid": 211, "correlation": 1286
    }
  },
  {
    "ph": "s", "id": 1286, "pid": 76337, "tid": -914061504, "ts": 1716454216500195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500281, "dur": 0,
    "args": {
      "External id": 1296,
      "cbid": 317, "correlation": 1296
    }
  },
  {
    "ph": "f", "id": 1296, "pid": 76337, "tid": -914061504, "ts": 1716454216500281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500282, "dur": 0,
    "args": {
      "External id": 1297,
      "cbid": 203, "correlation": 1297
    }
  },
  {
    "ph": "f", "id": 1297, "pid": 76337, "tid": -914061504, "ts": 1716454216500282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500283, "dur": 0,
    "args": {
      "External id": 1298,
      "cbid": 205, "correlation": 1298
    }
  },
  {
    "ph": "f", "id": 1298, "pid": 76337, "tid": -914061504, "ts": 1716454216500283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216501407, "dur": 34,
    "args": {
      "External id": 1302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1302, "pid": 5, "tid": 7, "ts": 1716454216501407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500297, "dur": 13,
    "args": {
      "External id": 1302,
      "cbid": 211, "correlation": 1302
    }
  },
  {
    "ph": "s", "id": 1302, "pid": 76337, "tid": -914061504, "ts": 1716454216500297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216501442, "dur": 32,
    "args": {
      "External id": 1304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1304, "pid": 5, "tid": 7, "ts": 1716454216501442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500312, "dur": 5,
    "args": {
      "External id": 1304,
      "cbid": 211, "correlation": 1304
    }
  },
  {
    "ph": "s", "id": 1304, "pid": 76337, "tid": -914061504, "ts": 1716454216500312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216501476, "dur": 1,
    "args": {
      "External id": 1306,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1306,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.5737704918032787
    }
  },
  {
    "ph": "f", "id": 1306, "pid": 5, "tid": 7, "ts": 1716454216501476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500323, "dur": 6,
    "args": {
      "External id": 1306,
      "cbid": 51, "correlation": 1306
    }
  },
  {
    "ph": "s", "id": 1306, "pid": 76337, "tid": -914061504, "ts": 1716454216500323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216501481, "dur": 892,
    "args": {
      "External id": 1307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1307,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1307, "pid": 5, "tid": 7, "ts": 1716454216501481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500331, "dur": 6,
    "args": {
      "External id": 1307,
      "cbid": 211, "correlation": 1307
    }
  },
  {
    "ph": "s", "id": 1307, "pid": 76337, "tid": -914061504, "ts": 1716454216500331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216502374, "dur": 36,
    "args": {
      "External id": 1309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1309, "pid": 5, "tid": 7, "ts": 1716454216502374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500341, "dur": 5,
    "args": {
      "External id": 1309,
      "cbid": 211, "correlation": 1309
    }
  },
  {
    "ph": "s", "id": 1309, "pid": 76337, "tid": -914061504, "ts": 1716454216500341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216502411, "dur": 48,
    "args": {
      "External id": 1315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1315, "pid": 5, "tid": 7, "ts": 1716454216502411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500370, "dur": 9,
    "args": {
      "External id": 1315,
      "cbid": 211, "correlation": 1315
    }
  },
  {
    "ph": "s", "id": 1315, "pid": 76337, "tid": -914061504, "ts": 1716454216500370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216502461, "dur": 140,
    "args": {
      "External id": 1323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1323, "pid": 5, "tid": 7, "ts": 1716454216502461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500404, "dur": 9,
    "args": {
      "External id": 1323,
      "cbid": 211, "correlation": 1323
    }
  },
  {
    "ph": "s", "id": 1323, "pid": 76337, "tid": -914061504, "ts": 1716454216500404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216502603, "dur": 29,
    "args": {
      "External id": 1331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1331,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1331, "pid": 5, "tid": 7, "ts": 1716454216502603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500435, "dur": 8,
    "args": {
      "External id": 1331,
      "cbid": 211, "correlation": 1331
    }
  },
  {
    "ph": "s", "id": 1331, "pid": 76337, "tid": -914061504, "ts": 1716454216500435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216502633, "dur": 25,
    "args": {
      "External id": 1341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1341,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 156.4125,
      "warps per SM": 625.65,
      "grid": [12513, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1341, "pid": 5, "tid": 7, "ts": 1716454216502633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500539, "dur": 15,
    "args": {
      "External id": 1341,
      "cbid": 211, "correlation": 1341
    }
  },
  {
    "ph": "s", "id": 1341, "pid": 76337, "tid": -914061504, "ts": 1716454216500539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216502659, "dur": 47,
    "args": {
      "External id": 1346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1346, "pid": 5, "tid": 7, "ts": 1716454216502659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500587, "dur": 8,
    "args": {
      "External id": 1346,
      "cbid": 211, "correlation": 1346
    }
  },
  {
    "ph": "s", "id": 1346, "pid": 76337, "tid": -914061504, "ts": 1716454216500587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216502708, "dur": 25,
    "args": {
      "External id": 1361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 57.6,
      "warps per SM": 230.4,
      "grid": [4608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1361, "pid": 5, "tid": 7, "ts": 1716454216502708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500671, "dur": 13,
    "args": {
      "External id": 1361,
      "cbid": 211, "correlation": 1361
    }
  },
  {
    "ph": "s", "id": 1361, "pid": 76337, "tid": -914061504, "ts": 1716454216500671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500692, "dur": 0,
    "args": {
      "External id": 1368,
      "cbid": 317, "correlation": 1368
    }
  },
  {
    "ph": "f", "id": 1368, "pid": 76337, "tid": -914061504, "ts": 1716454216500692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500693, "dur": 0,
    "args": {
      "External id": 1369,
      "cbid": 203, "correlation": 1369
    }
  },
  {
    "ph": "f", "id": 1369, "pid": 76337, "tid": -914061504, "ts": 1716454216500693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216500694, "dur": 0,
    "args": {
      "External id": 1370,
      "cbid": 205, "correlation": 1370
    }
  },
  {
    "ph": "f", "id": 1370, "pid": 76337, "tid": -914061504, "ts": 1716454216500694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216503346, "dur": 310,
    "args": {
      "External id": 1374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1374,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [4, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 1374, "pid": 5, "tid": 7, "ts": 1716454216503346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216503323, "dur": 23,
    "args": {
      "External id": 1374,
      "cbid": 211, "correlation": 1374
    }
  },
  {
    "ph": "s", "id": 1374, "pid": 76337, "tid": -914061504, "ts": 1716454216503323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216504060, "dur": 12,
    "args": {
      "External id": 1380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1380, "pid": 5, "tid": 7, "ts": 1716454216504060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504045, "dur": 13,
    "args": {
      "External id": 1380,
      "cbid": 211, "correlation": 1380
    }
  },
  {
    "ph": "s", "id": 1380, "pid": 76337, "tid": -914061504, "ts": 1716454216504045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216504256, "dur": 38,
    "args": {
      "External id": 1390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1390,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1390, "pid": 5, "tid": 7, "ts": 1716454216504256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504240, "dur": 16,
    "args": {
      "External id": 1390,
      "cbid": 211, "correlation": 1390
    }
  },
  {
    "ph": "s", "id": 1390, "pid": 76337, "tid": -914061504, "ts": 1716454216504240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216504302, "dur": 48,
    "args": {
      "External id": 1411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1411,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1411, "pid": 5, "tid": 7, "ts": 1716454216504302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504291, "dur": 10,
    "args": {
      "External id": 1411,
      "cbid": 211, "correlation": 1411
    }
  },
  {
    "ph": "s", "id": 1411, "pid": 76337, "tid": -914061504, "ts": 1716454216504291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216504352, "dur": 4,
    "args": {
      "External id": 1423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1423,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1423, "pid": 5, "tid": 7, "ts": 1716454216504352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504312, "dur": 7,
    "args": {
      "External id": 1423,
      "cbid": 211, "correlation": 1423
    }
  },
  {
    "ph": "s", "id": 1423, "pid": 76337, "tid": -914061504, "ts": 1716454216504312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216504357, "dur": 15,
    "args": {
      "External id": 1426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1426, "pid": 5, "tid": 7, "ts": 1716454216504357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504337, "dur": 7,
    "args": {
      "External id": 1426,
      "cbid": 211, "correlation": 1426
    }
  },
  {
    "ph": "s", "id": 1426, "pid": 76337, "tid": -914061504, "ts": 1716454216504337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216504399, "dur": 9,
    "args": {
      "External id": 1435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1435,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1435, "pid": 5, "tid": 7, "ts": 1716454216504399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504388, "dur": 10,
    "args": {
      "External id": 1435,
      "cbid": 211, "correlation": 1435
    }
  },
  {
    "ph": "s", "id": 1435, "pid": 76337, "tid": -914061504, "ts": 1716454216504388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504471, "dur": 0,
    "args": {
      "External id": 1445,
      "cbid": 317, "correlation": 1445
    }
  },
  {
    "ph": "f", "id": 1445, "pid": 76337, "tid": -914061504, "ts": 1716454216504471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504472, "dur": 1,
    "args": {
      "External id": 1446,
      "cbid": 203, "correlation": 1446
    }
  },
  {
    "ph": "f", "id": 1446, "pid": 76337, "tid": -914061504, "ts": 1716454216504472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216504474, "dur": 0,
    "args": {
      "External id": 1447,
      "cbid": 205, "correlation": 1447
    }
  },
  {
    "ph": "f", "id": 1447, "pid": 76337, "tid": -914061504, "ts": 1716454216504474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216507858, "dur": 11,
    "args": {
      "External id": 1451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1451, "pid": 5, "tid": 7, "ts": 1716454216507858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216507836, "dur": 24,
    "args": {
      "External id": 1451,
      "cbid": 211, "correlation": 1451
    }
  },
  {
    "ph": "s", "id": 1451, "pid": 76337, "tid": -914061504, "ts": 1716454216507836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216507872, "dur": 32,
    "args": {
      "External id": 1453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1453,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1453, "pid": 5, "tid": 7, "ts": 1716454216507872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216507863, "dur": 6,
    "args": {
      "External id": 1453,
      "cbid": 211, "correlation": 1453
    }
  },
  {
    "ph": "s", "id": 1453, "pid": 76337, "tid": -914061504, "ts": 1716454216507863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216507906, "dur": 2,
    "args": {
      "External id": 1455,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1455,
      "bytes": 768, "memory bandwidth (GB/s)": 0.375
    }
  },
  {
    "ph": "f", "id": 1455, "pid": 5, "tid": 7, "ts": 1716454216507906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216507877, "dur": 16,
    "args": {
      "External id": 1455,
      "cbid": 51, "correlation": 1455
    }
  },
  {
    "ph": "s", "id": 1455, "pid": 76337, "tid": -914061504, "ts": 1716454216507877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216507911, "dur": 276,
    "args": {
      "External id": 1456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1456,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1456, "pid": 5, "tid": 7, "ts": 1716454216507911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216507894, "dur": 9,
    "args": {
      "External id": 1456,
      "cbid": 211, "correlation": 1456
    }
  },
  {
    "ph": "s", "id": 1456, "pid": 76337, "tid": -914061504, "ts": 1716454216507894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216508188, "dur": 11,
    "args": {
      "External id": 1458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1458,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1458, "pid": 5, "tid": 7, "ts": 1716454216508188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216507910, "dur": 7,
    "args": {
      "External id": 1458,
      "cbid": 211, "correlation": 1458
    }
  },
  {
    "ph": "s", "id": 1458, "pid": 76337, "tid": -914061504, "ts": 1716454216507910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216508693, "dur": 13,
    "args": {
      "External id": 1464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1464, "pid": 5, "tid": 7, "ts": 1716454216508693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216508679, "dur": 13,
    "args": {
      "External id": 1464,
      "cbid": 211, "correlation": 1464
    }
  },
  {
    "ph": "s", "id": 1464, "pid": 76337, "tid": -914061504, "ts": 1716454216508679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216508829, "dur": 30,
    "args": {
      "External id": 1484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1484,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1484, "pid": 5, "tid": 7, "ts": 1716454216508829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216508815, "dur": 14,
    "args": {
      "External id": 1484,
      "cbid": 211, "correlation": 1484
    }
  },
  {
    "ph": "s", "id": 1484, "pid": 76337, "tid": -914061504, "ts": 1716454216508815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216508861, "dur": 4,
    "args": {
      "External id": 1496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1496,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1496, "pid": 5, "tid": 7, "ts": 1716454216508861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216508840, "dur": 8,
    "args": {
      "External id": 1496,
      "cbid": 211, "correlation": 1496
    }
  },
  {
    "ph": "s", "id": 1496, "pid": 76337, "tid": -914061504, "ts": 1716454216508840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216508875, "dur": 13,
    "args": {
      "External id": 1499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1499, "pid": 5, "tid": 7, "ts": 1716454216508875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216508866, "dur": 8,
    "args": {
      "External id": 1499,
      "cbid": 211, "correlation": 1499
    }
  },
  {
    "ph": "s", "id": 1499, "pid": 76337, "tid": -914061504, "ts": 1716454216508866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216508931, "dur": 9,
    "args": {
      "External id": 1508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1508,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1508, "pid": 5, "tid": 7, "ts": 1716454216508931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216508919, "dur": 11,
    "args": {
      "External id": 1508,
      "cbid": 211, "correlation": 1508
    }
  },
  {
    "ph": "s", "id": 1508, "pid": 76337, "tid": -914061504, "ts": 1716454216508919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509040, "dur": 0,
    "args": {
      "External id": 1518,
      "cbid": 317, "correlation": 1518
    }
  },
  {
    "ph": "f", "id": 1518, "pid": 76337, "tid": -914061504, "ts": 1716454216509040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509041, "dur": 0,
    "args": {
      "External id": 1519,
      "cbid": 203, "correlation": 1519
    }
  },
  {
    "ph": "f", "id": 1519, "pid": 76337, "tid": -914061504, "ts": 1716454216509041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509042, "dur": 0,
    "args": {
      "External id": 1520,
      "cbid": 205, "correlation": 1520
    }
  },
  {
    "ph": "f", "id": 1520, "pid": 76337, "tid": -914061504, "ts": 1716454216509042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509072, "dur": 9,
    "args": {
      "External id": 1524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1524, "pid": 5, "tid": 7, "ts": 1716454216509072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509058, "dur": 13,
    "args": {
      "External id": 1524,
      "cbid": 211, "correlation": 1524
    }
  },
  {
    "ph": "s", "id": 1524, "pid": 76337, "tid": -914061504, "ts": 1716454216509058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509083, "dur": 32,
    "args": {
      "External id": 1526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1526,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1526, "pid": 5, "tid": 7, "ts": 1716454216509083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509073, "dur": 6,
    "args": {
      "External id": 1526,
      "cbid": 211, "correlation": 1526
    }
  },
  {
    "ph": "s", "id": 1526, "pid": 76337, "tid": -914061504, "ts": 1716454216509073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216509117, "dur": 1,
    "args": {
      "External id": 1528,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1528,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 1528, "pid": 5, "tid": 7, "ts": 1716454216509117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509085, "dur": 8,
    "args": {
      "External id": 1528,
      "cbid": 51, "correlation": 1528
    }
  },
  {
    "ph": "s", "id": 1528, "pid": 76337, "tid": -914061504, "ts": 1716454216509085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216509121, "dur": 266,
    "args": {
      "External id": 1529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1529,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1529, "pid": 5, "tid": 7, "ts": 1716454216509121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509094, "dur": 7,
    "args": {
      "External id": 1529,
      "cbid": 211, "correlation": 1529
    }
  },
  {
    "ph": "s", "id": 1529, "pid": 76337, "tid": -914061504, "ts": 1716454216509094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509388, "dur": 10,
    "args": {
      "External id": 1531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1531,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1531, "pid": 5, "tid": 7, "ts": 1716454216509388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509106, "dur": 6,
    "args": {
      "External id": 1531,
      "cbid": 211, "correlation": 1531
    }
  },
  {
    "ph": "s", "id": 1531, "pid": 76337, "tid": -914061504, "ts": 1716454216509106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216509400, "dur": 12,
    "args": {
      "External id": 1537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1537,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1537, "pid": 5, "tid": 7, "ts": 1716454216509400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509136, "dur": 9,
    "args": {
      "External id": 1537,
      "cbid": 211, "correlation": 1537
    }
  },
  {
    "ph": "s", "id": 1537, "pid": 76337, "tid": -914061504, "ts": 1716454216509136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216509413, "dur": 39,
    "args": {
      "External id": 1545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1545, "pid": 5, "tid": 7, "ts": 1716454216509413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509180, "dur": 9,
    "args": {
      "External id": 1545,
      "cbid": 211, "correlation": 1545
    }
  },
  {
    "ph": "s", "id": 1545, "pid": 76337, "tid": -914061504, "ts": 1716454216509180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216509454, "dur": 10,
    "args": {
      "External id": 1553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1553,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1553, "pid": 5, "tid": 7, "ts": 1716454216509454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509225, "dur": 9,
    "args": {
      "External id": 1553,
      "cbid": 211, "correlation": 1553
    }
  },
  {
    "ph": "s", "id": 1553, "pid": 76337, "tid": -914061504, "ts": 1716454216509225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216509465, "dur": 38,
    "args": {
      "External id": 1563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1563, "pid": 5, "tid": 7, "ts": 1716454216509465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509320, "dur": 16,
    "args": {
      "External id": 1563,
      "cbid": 211, "correlation": 1563
    }
  },
  {
    "ph": "s", "id": 1563, "pid": 76337, "tid": -914061504, "ts": 1716454216509320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509504, "dur": 48,
    "args": {
      "External id": 1584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1584,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1584, "pid": 5, "tid": 7, "ts": 1716454216509504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509356, "dur": 8,
    "args": {
      "External id": 1584,
      "cbid": 211, "correlation": 1584
    }
  },
  {
    "ph": "s", "id": 1584, "pid": 76337, "tid": -914061504, "ts": 1716454216509356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216509553, "dur": 4,
    "args": {
      "External id": 1596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1596,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1596, "pid": 5, "tid": 7, "ts": 1716454216509553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509385, "dur": 8,
    "args": {
      "External id": 1596,
      "cbid": 211, "correlation": 1596
    }
  },
  {
    "ph": "s", "id": 1596, "pid": 76337, "tid": -914061504, "ts": 1716454216509385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216509559, "dur": 13,
    "args": {
      "External id": 1599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1599, "pid": 5, "tid": 7, "ts": 1716454216509559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509404, "dur": 7,
    "args": {
      "External id": 1599,
      "cbid": 211, "correlation": 1599
    }
  },
  {
    "ph": "s", "id": 1599, "pid": 76337, "tid": -914061504, "ts": 1716454216509404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216509574, "dur": 10,
    "args": {
      "External id": 1608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1608,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1608, "pid": 5, "tid": 7, "ts": 1716454216509574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509447, "dur": 10,
    "args": {
      "External id": 1608,
      "cbid": 211, "correlation": 1608
    }
  },
  {
    "ph": "s", "id": 1608, "pid": 76337, "tid": -914061504, "ts": 1716454216509447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509518, "dur": 0,
    "args": {
      "External id": 1618,
      "cbid": 317, "correlation": 1618
    }
  },
  {
    "ph": "f", "id": 1618, "pid": 76337, "tid": -914061504, "ts": 1716454216509518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509519, "dur": 0,
    "args": {
      "External id": 1619,
      "cbid": 203, "correlation": 1619
    }
  },
  {
    "ph": "f", "id": 1619, "pid": 76337, "tid": -914061504, "ts": 1716454216509519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509520, "dur": 0,
    "args": {
      "External id": 1620,
      "cbid": 205, "correlation": 1620
    }
  },
  {
    "ph": "f", "id": 1620, "pid": 76337, "tid": -914061504, "ts": 1716454216509520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509585, "dur": 9,
    "args": {
      "External id": 1624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1624, "pid": 5, "tid": 7, "ts": 1716454216509585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509534, "dur": 13,
    "args": {
      "External id": 1624,
      "cbid": 211, "correlation": 1624
    }
  },
  {
    "ph": "s", "id": 1624, "pid": 76337, "tid": -914061504, "ts": 1716454216509534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509595, "dur": 32,
    "args": {
      "External id": 1626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1626, "pid": 5, "tid": 7, "ts": 1716454216509595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509549, "dur": 5,
    "args": {
      "External id": 1626,
      "cbid": 211, "correlation": 1626
    }
  },
  {
    "ph": "s", "id": 1626, "pid": 76337, "tid": -914061504, "ts": 1716454216509549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216509630, "dur": 1,
    "args": {
      "External id": 1628,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1628,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46181599518941674
    }
  },
  {
    "ph": "f", "id": 1628, "pid": 5, "tid": 7, "ts": 1716454216509630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509560, "dur": 7,
    "args": {
      "External id": 1628,
      "cbid": 51, "correlation": 1628
    }
  },
  {
    "ph": "s", "id": 1628, "pid": 76337, "tid": -914061504, "ts": 1716454216509560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216509634, "dur": 265,
    "args": {
      "External id": 1629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1629,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1629, "pid": 5, "tid": 7, "ts": 1716454216509634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509568, "dur": 7,
    "args": {
      "External id": 1629,
      "cbid": 211, "correlation": 1629
    }
  },
  {
    "ph": "s", "id": 1629, "pid": 76337, "tid": -914061504, "ts": 1716454216509568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509901, "dur": 10,
    "args": {
      "External id": 1631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1631, "pid": 5, "tid": 7, "ts": 1716454216509901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509578, "dur": 5,
    "args": {
      "External id": 1631,
      "cbid": 211, "correlation": 1631
    }
  },
  {
    "ph": "s", "id": 1631, "pid": 76337, "tid": -914061504, "ts": 1716454216509578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216509913, "dur": 12,
    "args": {
      "External id": 1637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1637, "pid": 5, "tid": 7, "ts": 1716454216509913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509607, "dur": 9,
    "args": {
      "External id": 1637,
      "cbid": 211, "correlation": 1637
    }
  },
  {
    "ph": "s", "id": 1637, "pid": 76337, "tid": -914061504, "ts": 1716454216509607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509926, "dur": 30,
    "args": {
      "External id": 1657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1657,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1657, "pid": 5, "tid": 7, "ts": 1716454216509926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509700, "dur": 13,
    "args": {
      "External id": 1657,
      "cbid": 211, "correlation": 1657
    }
  },
  {
    "ph": "s", "id": 1657, "pid": 76337, "tid": -914061504, "ts": 1716454216509700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216509958, "dur": 4,
    "args": {
      "External id": 1669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1669,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1669, "pid": 5, "tid": 7, "ts": 1716454216509958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509723, "dur": 7,
    "args": {
      "External id": 1669,
      "cbid": 211, "correlation": 1669
    }
  },
  {
    "ph": "s", "id": 1669, "pid": 76337, "tid": -914061504, "ts": 1716454216509723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216509963, "dur": 14,
    "args": {
      "External id": 1672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1672, "pid": 5, "tid": 7, "ts": 1716454216509963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509742, "dur": 6,
    "args": {
      "External id": 1672,
      "cbid": 211, "correlation": 1672
    }
  },
  {
    "ph": "s", "id": 1672, "pid": 76337, "tid": -914061504, "ts": 1716454216509742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216509978, "dur": 9,
    "args": {
      "External id": 1681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1681,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1681, "pid": 5, "tid": 7, "ts": 1716454216509978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509784, "dur": 15,
    "args": {
      "External id": 1681,
      "cbid": 211, "correlation": 1681
    }
  },
  {
    "ph": "s", "id": 1681, "pid": 76337, "tid": -914061504, "ts": 1716454216509784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509852, "dur": 0,
    "args": {
      "External id": 1691,
      "cbid": 317, "correlation": 1691
    }
  },
  {
    "ph": "f", "id": 1691, "pid": 76337, "tid": -914061504, "ts": 1716454216509852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509853, "dur": 0,
    "args": {
      "External id": 1692,
      "cbid": 203, "correlation": 1692
    }
  },
  {
    "ph": "f", "id": 1692, "pid": 76337, "tid": -914061504, "ts": 1716454216509853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509853, "dur": 0,
    "args": {
      "External id": 1693,
      "cbid": 205, "correlation": 1693
    }
  },
  {
    "ph": "f", "id": 1693, "pid": 76337, "tid": -914061504, "ts": 1716454216509853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509989, "dur": 9,
    "args": {
      "External id": 1697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1697,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1697, "pid": 5, "tid": 7, "ts": 1716454216509989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509867, "dur": 13,
    "args": {
      "External id": 1697,
      "cbid": 211, "correlation": 1697
    }
  },
  {
    "ph": "s", "id": 1697, "pid": 76337, "tid": -914061504, "ts": 1716454216509867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216509999, "dur": 32,
    "args": {
      "External id": 1699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1699, "pid": 5, "tid": 7, "ts": 1716454216509999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509882, "dur": 5,
    "args": {
      "External id": 1699,
      "cbid": 211, "correlation": 1699
    }
  },
  {
    "ph": "s", "id": 1699, "pid": 76337, "tid": -914061504, "ts": 1716454216509882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216510033, "dur": 1,
    "args": {
      "External id": 1701,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1701,
      "bytes": 768, "memory bandwidth (GB/s)": 0.4528301886792453
    }
  },
  {
    "ph": "f", "id": 1701, "pid": 5, "tid": 7, "ts": 1716454216510033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509892, "dur": 6,
    "args": {
      "External id": 1701,
      "cbid": 51, "correlation": 1701
    }
  },
  {
    "ph": "s", "id": 1701, "pid": 76337, "tid": -914061504, "ts": 1716454216509892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216510037, "dur": 265,
    "args": {
      "External id": 1702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1702,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1702, "pid": 5, "tid": 7, "ts": 1716454216510037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509900, "dur": 6,
    "args": {
      "External id": 1702,
      "cbid": 211, "correlation": 1702
    }
  },
  {
    "ph": "s", "id": 1702, "pid": 76337, "tid": -914061504, "ts": 1716454216509900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510304, "dur": 10,
    "args": {
      "External id": 1704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1704, "pid": 5, "tid": 7, "ts": 1716454216510304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509909, "dur": 5,
    "args": {
      "External id": 1704,
      "cbid": 211, "correlation": 1704
    }
  },
  {
    "ph": "s", "id": 1704, "pid": 76337, "tid": -914061504, "ts": 1716454216509909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216510315, "dur": 12,
    "args": {
      "External id": 1710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1710, "pid": 5, "tid": 7, "ts": 1716454216510315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509937, "dur": 8,
    "args": {
      "External id": 1710,
      "cbid": 211, "correlation": 1710
    }
  },
  {
    "ph": "s", "id": 1710, "pid": 76337, "tid": -914061504, "ts": 1716454216509937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216510329, "dur": 39,
    "args": {
      "External id": 1718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1718, "pid": 5, "tid": 7, "ts": 1716454216510329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216509971, "dur": 18,
    "args": {
      "External id": 1718,
      "cbid": 211, "correlation": 1718
    }
  },
  {
    "ph": "s", "id": 1718, "pid": 76337, "tid": -914061504, "ts": 1716454216509971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216510370, "dur": 9,
    "args": {
      "External id": 1726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1726,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1726, "pid": 5, "tid": 7, "ts": 1716454216510370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510011, "dur": 8,
    "args": {
      "External id": 1726,
      "cbid": 211, "correlation": 1726
    }
  },
  {
    "ph": "s", "id": 1726, "pid": 76337, "tid": -914061504, "ts": 1716454216510011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216510380, "dur": 38,
    "args": {
      "External id": 1736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1736,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1736, "pid": 5, "tid": 7, "ts": 1716454216510380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510245, "dur": 15,
    "args": {
      "External id": 1736,
      "cbid": 211, "correlation": 1736
    }
  },
  {
    "ph": "s", "id": 1736, "pid": 76337, "tid": -914061504, "ts": 1716454216510245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510420, "dur": 34,
    "args": {
      "External id": 1757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1757,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1757, "pid": 5, "tid": 7, "ts": 1716454216510420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510293, "dur": 9,
    "args": {
      "External id": 1757,
      "cbid": 211, "correlation": 1757
    }
  },
  {
    "ph": "s", "id": 1757, "pid": 76337, "tid": -914061504, "ts": 1716454216510293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216510456, "dur": 5,
    "args": {
      "External id": 1769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1769,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1769, "pid": 5, "tid": 7, "ts": 1716454216510456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510311, "dur": 6,
    "args": {
      "External id": 1769,
      "cbid": 211, "correlation": 1769
    }
  },
  {
    "ph": "s", "id": 1769, "pid": 76337, "tid": -914061504, "ts": 1716454216510311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216510462, "dur": 13,
    "args": {
      "External id": 1772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1772, "pid": 5, "tid": 7, "ts": 1716454216510462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510329, "dur": 7,
    "args": {
      "External id": 1772,
      "cbid": 211, "correlation": 1772
    }
  },
  {
    "ph": "s", "id": 1772, "pid": 76337, "tid": -914061504, "ts": 1716454216510329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216510476, "dur": 10,
    "args": {
      "External id": 1781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1781,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1781, "pid": 5, "tid": 7, "ts": 1716454216510476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510371, "dur": 11,
    "args": {
      "External id": 1781,
      "cbid": 211, "correlation": 1781
    }
  },
  {
    "ph": "s", "id": 1781, "pid": 76337, "tid": -914061504, "ts": 1716454216510371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510448, "dur": 0,
    "args": {
      "External id": 1791,
      "cbid": 317, "correlation": 1791
    }
  },
  {
    "ph": "f", "id": 1791, "pid": 76337, "tid": -914061504, "ts": 1716454216510448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510449, "dur": 0,
    "args": {
      "External id": 1792,
      "cbid": 203, "correlation": 1792
    }
  },
  {
    "ph": "f", "id": 1792, "pid": 76337, "tid": -914061504, "ts": 1716454216510449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510449, "dur": 0,
    "args": {
      "External id": 1793,
      "cbid": 205, "correlation": 1793
    }
  },
  {
    "ph": "f", "id": 1793, "pid": 76337, "tid": -914061504, "ts": 1716454216510449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510488, "dur": 9,
    "args": {
      "External id": 1797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1797,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1797, "pid": 5, "tid": 7, "ts": 1716454216510488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510470, "dur": 14,
    "args": {
      "External id": 1797,
      "cbid": 211, "correlation": 1797
    }
  },
  {
    "ph": "s", "id": 1797, "pid": 76337, "tid": -914061504, "ts": 1716454216510470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510498, "dur": 33,
    "args": {
      "External id": 1799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1799,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1799, "pid": 5, "tid": 7, "ts": 1716454216510498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510486, "dur": 5,
    "args": {
      "External id": 1799,
      "cbid": 211, "correlation": 1799
    }
  },
  {
    "ph": "s", "id": 1799, "pid": 76337, "tid": -914061504, "ts": 1716454216510486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216510533, "dur": 1,
    "args": {
      "External id": 1801,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1801,
      "bytes": 768, "memory bandwidth (GB/s)": 0.4528301886792453
    }
  },
  {
    "ph": "f", "id": 1801, "pid": 5, "tid": 7, "ts": 1716454216510533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510497, "dur": 6,
    "args": {
      "External id": 1801,
      "cbid": 51, "correlation": 1801
    }
  },
  {
    "ph": "s", "id": 1801, "pid": 76337, "tid": -914061504, "ts": 1716454216510497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216510537, "dur": 267,
    "args": {
      "External id": 1802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1802,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1802, "pid": 5, "tid": 7, "ts": 1716454216510537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510504, "dur": 6,
    "args": {
      "External id": 1802,
      "cbid": 211, "correlation": 1802
    }
  },
  {
    "ph": "s", "id": 1802, "pid": 76337, "tid": -914061504, "ts": 1716454216510504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510805, "dur": 10,
    "args": {
      "External id": 1804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1804,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1804, "pid": 5, "tid": 7, "ts": 1716454216510805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510515, "dur": 5,
    "args": {
      "External id": 1804,
      "cbid": 211, "correlation": 1804
    }
  },
  {
    "ph": "s", "id": 1804, "pid": 76337, "tid": -914061504, "ts": 1716454216510515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216510816, "dur": 12,
    "args": {
      "External id": 1810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1810, "pid": 5, "tid": 7, "ts": 1716454216510816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510544, "dur": 9,
    "args": {
      "External id": 1810,
      "cbid": 211, "correlation": 1810
    }
  },
  {
    "ph": "s", "id": 1810, "pid": 76337, "tid": -914061504, "ts": 1716454216510544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510830, "dur": 30,
    "args": {
      "External id": 1830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1830,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1830, "pid": 5, "tid": 7, "ts": 1716454216510830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510625, "dur": 12,
    "args": {
      "External id": 1830,
      "cbid": 211, "correlation": 1830
    }
  },
  {
    "ph": "s", "id": 1830, "pid": 76337, "tid": -914061504, "ts": 1716454216510625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216510861, "dur": 4,
    "args": {
      "External id": 1842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1842,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1842, "pid": 5, "tid": 7, "ts": 1716454216510861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510648, "dur": 6,
    "args": {
      "External id": 1842,
      "cbid": 211, "correlation": 1842
    }
  },
  {
    "ph": "s", "id": 1842, "pid": 76337, "tid": -914061504, "ts": 1716454216510648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216510866, "dur": 13,
    "args": {
      "External id": 1845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1845, "pid": 5, "tid": 7, "ts": 1716454216510866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510666, "dur": 6,
    "args": {
      "External id": 1845,
      "cbid": 211, "correlation": 1845
    }
  },
  {
    "ph": "s", "id": 1845, "pid": 76337, "tid": -914061504, "ts": 1716454216510666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216510881, "dur": 9,
    "args": {
      "External id": 1854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1854,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1854, "pid": 5, "tid": 7, "ts": 1716454216510881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510706, "dur": 9,
    "args": {
      "External id": 1854,
      "cbid": 211, "correlation": 1854
    }
  },
  {
    "ph": "s", "id": 1854, "pid": 76337, "tid": -914061504, "ts": 1716454216510706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510788, "dur": 0,
    "args": {
      "External id": 1864,
      "cbid": 317, "correlation": 1864
    }
  },
  {
    "ph": "f", "id": 1864, "pid": 76337, "tid": -914061504, "ts": 1716454216510788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510789, "dur": 0,
    "args": {
      "External id": 1865,
      "cbid": 203, "correlation": 1865
    }
  },
  {
    "ph": "f", "id": 1865, "pid": 76337, "tid": -914061504, "ts": 1716454216510789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510790, "dur": 0,
    "args": {
      "External id": 1866,
      "cbid": 205, "correlation": 1866
    }
  },
  {
    "ph": "f", "id": 1866, "pid": 76337, "tid": -914061504, "ts": 1716454216510790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510892, "dur": 9,
    "args": {
      "External id": 1870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1870, "pid": 5, "tid": 7, "ts": 1716454216510892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510803, "dur": 13,
    "args": {
      "External id": 1870,
      "cbid": 211, "correlation": 1870
    }
  },
  {
    "ph": "s", "id": 1870, "pid": 76337, "tid": -914061504, "ts": 1716454216510803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216510902, "dur": 33,
    "args": {
      "External id": 1872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1872,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1872, "pid": 5, "tid": 7, "ts": 1716454216510902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510818, "dur": 5,
    "args": {
      "External id": 1872,
      "cbid": 211, "correlation": 1872
    }
  },
  {
    "ph": "s", "id": 1872, "pid": 76337, "tid": -914061504, "ts": 1716454216510818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216510937, "dur": 1,
    "args": {
      "External id": 1874,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 1874,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 1874, "pid": 5, "tid": 7, "ts": 1716454216510937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510829, "dur": 6,
    "args": {
      "External id": 1874,
      "cbid": 51, "correlation": 1874
    }
  },
  {
    "ph": "s", "id": 1874, "pid": 76337, "tid": -914061504, "ts": 1716454216510829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216510941, "dur": 265,
    "args": {
      "External id": 1875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1875,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 1875, "pid": 5, "tid": 7, "ts": 1716454216510941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510836, "dur": 6,
    "args": {
      "External id": 1875,
      "cbid": 211, "correlation": 1875
    }
  },
  {
    "ph": "s", "id": 1875, "pid": 76337, "tid": -914061504, "ts": 1716454216510836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216511207, "dur": 10,
    "args": {
      "External id": 1877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1877,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1877, "pid": 5, "tid": 7, "ts": 1716454216511207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510846, "dur": 5,
    "args": {
      "External id": 1877,
      "cbid": 211, "correlation": 1877
    }
  },
  {
    "ph": "s", "id": 1877, "pid": 76337, "tid": -914061504, "ts": 1716454216510846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216511218, "dur": 12,
    "args": {
      "External id": 1883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1883, "pid": 5, "tid": 7, "ts": 1716454216511218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510875, "dur": 8,
    "args": {
      "External id": 1883,
      "cbid": 211, "correlation": 1883
    }
  },
  {
    "ph": "s", "id": 1883, "pid": 76337, "tid": -914061504, "ts": 1716454216510875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216511232, "dur": 39,
    "args": {
      "External id": 1891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1891, "pid": 5, "tid": 7, "ts": 1716454216511232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510909, "dur": 9,
    "args": {
      "External id": 1891,
      "cbid": 211, "correlation": 1891
    }
  },
  {
    "ph": "s", "id": 1891, "pid": 76337, "tid": -914061504, "ts": 1716454216510909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216511272, "dur": 9,
    "args": {
      "External id": 1899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1899,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1899, "pid": 5, "tid": 7, "ts": 1716454216511272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216510939, "dur": 8,
    "args": {
      "External id": 1899,
      "cbid": 211, "correlation": 1899
    }
  },
  {
    "ph": "s", "id": 1899, "pid": 76337, "tid": -914061504, "ts": 1716454216510939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216511336, "dur": 38,
    "args": {
      "External id": 1909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1909, "pid": 5, "tid": 7, "ts": 1716454216511336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511319, "dur": 17,
    "args": {
      "External id": 1909,
      "cbid": 211, "correlation": 1909
    }
  },
  {
    "ph": "s", "id": 1909, "pid": 76337, "tid": -914061504, "ts": 1716454216511319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216511375, "dur": 31,
    "args": {
      "External id": 1930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1930,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 1930, "pid": 5, "tid": 7, "ts": 1716454216511375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511357, "dur": 9,
    "args": {
      "External id": 1930,
      "cbid": 211, "correlation": 1930
    }
  },
  {
    "ph": "s", "id": 1930, "pid": 76337, "tid": -914061504, "ts": 1716454216511357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216511408, "dur": 4,
    "args": {
      "External id": 1942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1942,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 1942, "pid": 5, "tid": 7, "ts": 1716454216511408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511376, "dur": 6,
    "args": {
      "External id": 1942,
      "cbid": 211, "correlation": 1942
    }
  },
  {
    "ph": "s", "id": 1942, "pid": 76337, "tid": -914061504, "ts": 1716454216511376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216511414, "dur": 14,
    "args": {
      "External id": 1945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1945, "pid": 5, "tid": 7, "ts": 1716454216511414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511393, "dur": 6,
    "args": {
      "External id": 1945,
      "cbid": 211, "correlation": 1945
    }
  },
  {
    "ph": "s", "id": 1945, "pid": 76337, "tid": -914061504, "ts": 1716454216511393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511635, "dur": 93,
    "args": {
      "External id": 1957,
      "cbid": 251, "correlation": 1957
    }
  },
  {
    "ph": "f", "id": 1957, "pid": 76337, "tid": -914061504, "ts": 1716454216511635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216511755, "dur": 37,
    "args": {
      "External id": 1958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1958,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 1958, "pid": 5, "tid": 7, "ts": 1716454216511755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511738, "dur": 17,
    "args": {
      "External id": 1958,
      "cbid": 211, "correlation": 1958
    }
  },
  {
    "ph": "s", "id": 1958, "pid": 76337, "tid": -914061504, "ts": 1716454216511738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216511804, "dur": 17,
    "args": {
      "External id": 1963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1963,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1963, "pid": 5, "tid": 7, "ts": 1716454216511804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511792, "dur": 10,
    "args": {
      "External id": 1963,
      "cbid": 211, "correlation": 1963
    }
  },
  {
    "ph": "s", "id": 1963, "pid": 76337, "tid": -914061504, "ts": 1716454216511792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511893, "dur": 1,
    "args": {
      "External id": 1974,
      "cbid": 251, "correlation": 1974
    }
  },
  {
    "ph": "f", "id": 1974, "pid": 76337, "tid": -914061504, "ts": 1716454216511893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216511912, "dur": 37,
    "args": {
      "External id": 1975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1975,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 1975, "pid": 5, "tid": 7, "ts": 1716454216511912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511898, "dur": 13,
    "args": {
      "External id": 1975,
      "cbid": 211, "correlation": 1975
    }
  },
  {
    "ph": "s", "id": 1975, "pid": 76337, "tid": -914061504, "ts": 1716454216511898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216511952, "dur": 14,
    "args": {
      "External id": 1980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1980,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1980, "pid": 5, "tid": 7, "ts": 1716454216511952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216511939, "dur": 11,
    "args": {
      "External id": 1980,
      "cbid": 211, "correlation": 1980
    }
  },
  {
    "ph": "s", "id": 1980, "pid": 76337, "tid": -914061504, "ts": 1716454216511939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512036, "dur": 1,
    "args": {
      "External id": 1991,
      "cbid": 251, "correlation": 1991
    }
  },
  {
    "ph": "f", "id": 1991, "pid": 76337, "tid": -914061504, "ts": 1716454216512036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216512056, "dur": 37,
    "args": {
      "External id": 1992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1992,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 1992, "pid": 5, "tid": 7, "ts": 1716454216512056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512041, "dur": 15,
    "args": {
      "External id": 1992,
      "cbid": 211, "correlation": 1992
    }
  },
  {
    "ph": "s", "id": 1992, "pid": 76337, "tid": -914061504, "ts": 1716454216512041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216512094, "dur": 13,
    "args": {
      "External id": 1997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 1997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 1997, "pid": 5, "tid": 7, "ts": 1716454216512094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512073, "dur": 9,
    "args": {
      "External id": 1997,
      "cbid": 211, "correlation": 1997
    }
  },
  {
    "ph": "s", "id": 1997, "pid": 76337, "tid": -914061504, "ts": 1716454216512073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216512545, "dur": 1205,
    "args": {
      "External id": 2022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2022,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 2022, "pid": 5, "tid": 7, "ts": 1716454216512545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512383, "dur": 163,
    "args": {
      "External id": 2022,
      "cbid": 211, "correlation": 2022
    }
  },
  {
    "ph": "s", "id": 2022, "pid": 76337, "tid": -914061504, "ts": 1716454216512383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512712, "dur": 59,
    "args": {
      "External id": 2040,
      "cbid": 251, "correlation": 2040
    }
  },
  {
    "ph": "f", "id": 2040, "pid": 76337, "tid": -914061504, "ts": 1716454216512712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216513759, "dur": 40,
    "args": {
      "External id": 2042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2042,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 2042, "pid": 5, "tid": 7, "ts": 1716454216513759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512778, "dur": 14,
    "args": {
      "External id": 2042,
      "cbid": 211, "correlation": 2042
    }
  },
  {
    "ph": "s", "id": 2042, "pid": 76337, "tid": -914061504, "ts": 1716454216512778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216513806, "dur": 12,
    "args": {
      "External id": 2050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2050,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2050, "pid": 5, "tid": 7, "ts": 1716454216513806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512889, "dur": 58,
    "args": {
      "External id": 2050,
      "cbid": 211, "correlation": 2050
    }
  },
  {
    "ph": "s", "id": 2050, "pid": 76337, "tid": -914061504, "ts": 1716454216512889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216513819, "dur": 9,
    "args": {
      "External id": 2058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2058,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2058, "pid": 5, "tid": 7, "ts": 1716454216513819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216512997, "dur": 13,
    "args": {
      "External id": 2058,
      "cbid": 211, "correlation": 2058
    }
  },
  {
    "ph": "s", "id": 2058, "pid": 76337, "tid": -914061504, "ts": 1716454216512997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216513830, "dur": 38,
    "args": {
      "External id": 2068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2068,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2068, "pid": 5, "tid": 7, "ts": 1716454216513830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513099, "dur": 15,
    "args": {
      "External id": 2068,
      "cbid": 211, "correlation": 2068
    }
  },
  {
    "ph": "s", "id": 2068, "pid": 76337, "tid": -914061504, "ts": 1716454216513099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216513870, "dur": 34,
    "args": {
      "External id": 2089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2089,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2089, "pid": 5, "tid": 7, "ts": 1716454216513870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513135, "dur": 8,
    "args": {
      "External id": 2089,
      "cbid": 211, "correlation": 2089
    }
  },
  {
    "ph": "s", "id": 2089, "pid": 76337, "tid": -914061504, "ts": 1716454216513135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216513905, "dur": 5,
    "args": {
      "External id": 2101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2101,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2101, "pid": 5, "tid": 7, "ts": 1716454216513905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513160, "dur": 9,
    "args": {
      "External id": 2101,
      "cbid": 211, "correlation": 2101
    }
  },
  {
    "ph": "s", "id": 2101, "pid": 76337, "tid": -914061504, "ts": 1716454216513160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216513911, "dur": 15,
    "args": {
      "External id": 2104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2104, "pid": 5, "tid": 7, "ts": 1716454216513911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513181, "dur": 7,
    "args": {
      "External id": 2104,
      "cbid": 211, "correlation": 2104
    }
  },
  {
    "ph": "s", "id": 2104, "pid": 76337, "tid": -914061504, "ts": 1716454216513181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216513928, "dur": 10,
    "args": {
      "External id": 2113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2113,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2113, "pid": 5, "tid": 7, "ts": 1716454216513928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513225, "dur": 9,
    "args": {
      "External id": 2113,
      "cbid": 211, "correlation": 2113
    }
  },
  {
    "ph": "s", "id": 2113, "pid": 76337, "tid": -914061504, "ts": 1716454216513225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513305, "dur": 0,
    "args": {
      "External id": 2123,
      "cbid": 317, "correlation": 2123
    }
  },
  {
    "ph": "f", "id": 2123, "pid": 76337, "tid": -914061504, "ts": 1716454216513305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513306, "dur": 0,
    "args": {
      "External id": 2124,
      "cbid": 203, "correlation": 2124
    }
  },
  {
    "ph": "f", "id": 2124, "pid": 76337, "tid": -914061504, "ts": 1716454216513306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513307, "dur": 0,
    "args": {
      "External id": 2125,
      "cbid": 205, "correlation": 2125
    }
  },
  {
    "ph": "f", "id": 2125, "pid": 76337, "tid": -914061504, "ts": 1716454216513307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216513940, "dur": 11,
    "args": {
      "External id": 2129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2129, "pid": 5, "tid": 7, "ts": 1716454216513940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513325, "dur": 14,
    "args": {
      "External id": 2129,
      "cbid": 211, "correlation": 2129
    }
  },
  {
    "ph": "s", "id": 2129, "pid": 76337, "tid": -914061504, "ts": 1716454216513325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216513952, "dur": 32,
    "args": {
      "External id": 2131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2131,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2131, "pid": 5, "tid": 7, "ts": 1716454216513952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513341, "dur": 5,
    "args": {
      "External id": 2131,
      "cbid": 211, "correlation": 2131
    }
  },
  {
    "ph": "s", "id": 2131, "pid": 76337, "tid": -914061504, "ts": 1716454216513341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216513986, "dur": 2,
    "args": {
      "External id": 2133,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 2133,
      "bytes": 768, "memory bandwidth (GB/s)": 0.38095238095238093
    }
  },
  {
    "ph": "f", "id": 2133, "pid": 5, "tid": 7, "ts": 1716454216513986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513353, "dur": 8,
    "args": {
      "External id": 2133,
      "cbid": 51, "correlation": 2133
    }
  },
  {
    "ph": "s", "id": 2133, "pid": 76337, "tid": -914061504, "ts": 1716454216513353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216513990, "dur": 275,
    "args": {
      "External id": 2134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2134,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2134, "pid": 5, "tid": 7, "ts": 1716454216513990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513361, "dur": 6,
    "args": {
      "External id": 2134,
      "cbid": 211, "correlation": 2134
    }
  },
  {
    "ph": "s", "id": 2134, "pid": 76337, "tid": -914061504, "ts": 1716454216513361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216514267, "dur": 11,
    "args": {
      "External id": 2136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2136,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2136, "pid": 5, "tid": 7, "ts": 1716454216514267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513372, "dur": 6,
    "args": {
      "External id": 2136,
      "cbid": 211, "correlation": 2136
    }
  },
  {
    "ph": "s", "id": 2136, "pid": 76337, "tid": -914061504, "ts": 1716454216513372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216514280, "dur": 14,
    "args": {
      "External id": 2142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2142, "pid": 5, "tid": 7, "ts": 1716454216514280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513402, "dur": 9,
    "args": {
      "External id": 2142,
      "cbid": 211, "correlation": 2142
    }
  },
  {
    "ph": "s", "id": 2142, "pid": 76337, "tid": -914061504, "ts": 1716454216513402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216514295, "dur": 31,
    "args": {
      "External id": 2162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2162,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2162, "pid": 5, "tid": 7, "ts": 1716454216514295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513482, "dur": 13,
    "args": {
      "External id": 2162,
      "cbid": 211, "correlation": 2162
    }
  },
  {
    "ph": "s", "id": 2162, "pid": 76337, "tid": -914061504, "ts": 1716454216513482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216514327, "dur": 4,
    "args": {
      "External id": 2174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2174,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2174, "pid": 5, "tid": 7, "ts": 1716454216514327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513505, "dur": 6,
    "args": {
      "External id": 2174,
      "cbid": 211, "correlation": 2174
    }
  },
  {
    "ph": "s", "id": 2174, "pid": 76337, "tid": -914061504, "ts": 1716454216513505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216514332, "dur": 14,
    "args": {
      "External id": 2177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2177, "pid": 5, "tid": 7, "ts": 1716454216514332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513523, "dur": 6,
    "args": {
      "External id": 2177,
      "cbid": 211, "correlation": 2177
    }
  },
  {
    "ph": "s", "id": 2177, "pid": 76337, "tid": -914061504, "ts": 1716454216513523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216514347, "dur": 9,
    "args": {
      "External id": 2186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2186,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2186, "pid": 5, "tid": 7, "ts": 1716454216514347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513563, "dur": 10,
    "args": {
      "External id": 2186,
      "cbid": 211, "correlation": 2186
    }
  },
  {
    "ph": "s", "id": 2186, "pid": 76337, "tid": -914061504, "ts": 1716454216513563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513623, "dur": 0,
    "args": {
      "External id": 2196,
      "cbid": 317, "correlation": 2196
    }
  },
  {
    "ph": "f", "id": 2196, "pid": 76337, "tid": -914061504, "ts": 1716454216513623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513624, "dur": 0,
    "args": {
      "External id": 2197,
      "cbid": 203, "correlation": 2197
    }
  },
  {
    "ph": "f", "id": 2197, "pid": 76337, "tid": -914061504, "ts": 1716454216513624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513625, "dur": 0,
    "args": {
      "External id": 2198,
      "cbid": 205, "correlation": 2198
    }
  },
  {
    "ph": "f", "id": 2198, "pid": 76337, "tid": -914061504, "ts": 1716454216513625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216514358, "dur": 8,
    "args": {
      "External id": 2202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2202, "pid": 5, "tid": 7, "ts": 1716454216514358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513639, "dur": 12,
    "args": {
      "External id": 2202,
      "cbid": 211, "correlation": 2202
    }
  },
  {
    "ph": "s", "id": 2202, "pid": 76337, "tid": -914061504, "ts": 1716454216513639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216514368, "dur": 32,
    "args": {
      "External id": 2204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2204,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2204, "pid": 5, "tid": 7, "ts": 1716454216514368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513653, "dur": 6,
    "args": {
      "External id": 2204,
      "cbid": 211, "correlation": 2204
    }
  },
  {
    "ph": "s", "id": 2204, "pid": 76337, "tid": -914061504, "ts": 1716454216513653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216514402, "dur": 1,
    "args": {
      "External id": 2206,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 2206,
      "bytes": 768, "memory bandwidth (GB/s)": 0.4528301886792453
    }
  },
  {
    "ph": "f", "id": 2206, "pid": 5, "tid": 7, "ts": 1716454216514402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513664, "dur": 6,
    "args": {
      "External id": 2206,
      "cbid": 51, "correlation": 2206
    }
  },
  {
    "ph": "s", "id": 2206, "pid": 76337, "tid": -914061504, "ts": 1716454216513664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216514406, "dur": 264,
    "args": {
      "External id": 2207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2207,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2207, "pid": 5, "tid": 7, "ts": 1716454216514406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513671, "dur": 6,
    "args": {
      "External id": 2207,
      "cbid": 211, "correlation": 2207
    }
  },
  {
    "ph": "s", "id": 2207, "pid": 76337, "tid": -914061504, "ts": 1716454216513671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216514671, "dur": 11,
    "args": {
      "External id": 2209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2209,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2209, "pid": 5, "tid": 7, "ts": 1716454216514671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513681, "dur": 5,
    "args": {
      "External id": 2209,
      "cbid": 211, "correlation": 2209
    }
  },
  {
    "ph": "s", "id": 2209, "pid": 76337, "tid": -914061504, "ts": 1716454216513681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216514683, "dur": 12,
    "args": {
      "External id": 2215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2215,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2215, "pid": 5, "tid": 7, "ts": 1716454216514683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513709, "dur": 8,
    "args": {
      "External id": 2215,
      "cbid": 211, "correlation": 2215
    }
  },
  {
    "ph": "s", "id": 2215, "pid": 76337, "tid": -914061504, "ts": 1716454216513709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216514697, "dur": 40,
    "args": {
      "External id": 2223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2223, "pid": 5, "tid": 7, "ts": 1716454216514697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513742, "dur": 9,
    "args": {
      "External id": 2223,
      "cbid": 211, "correlation": 2223
    }
  },
  {
    "ph": "s", "id": 2223, "pid": 76337, "tid": -914061504, "ts": 1716454216513742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216514738, "dur": 9,
    "args": {
      "External id": 2231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2231,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2231, "pid": 5, "tid": 7, "ts": 1716454216514738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513773, "dur": 8,
    "args": {
      "External id": 2231,
      "cbid": 211, "correlation": 2231
    }
  },
  {
    "ph": "s", "id": 2231, "pid": 76337, "tid": -914061504, "ts": 1716454216513773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216514749, "dur": 38,
    "args": {
      "External id": 2241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2241, "pid": 5, "tid": 7, "ts": 1716454216514749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513842, "dur": 12,
    "args": {
      "External id": 2241,
      "cbid": 211, "correlation": 2241
    }
  },
  {
    "ph": "s", "id": 2241, "pid": 76337, "tid": -914061504, "ts": 1716454216513842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216514788, "dur": 47,
    "args": {
      "External id": 2262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2262,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2262, "pid": 5, "tid": 7, "ts": 1716454216514788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513887, "dur": 9,
    "args": {
      "External id": 2262,
      "cbid": 211, "correlation": 2262
    }
  },
  {
    "ph": "s", "id": 2262, "pid": 76337, "tid": -914061504, "ts": 1716454216513887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216514836, "dur": 4,
    "args": {
      "External id": 2274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2274,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2274, "pid": 5, "tid": 7, "ts": 1716454216514836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513905, "dur": 6,
    "args": {
      "External id": 2274,
      "cbid": 211, "correlation": 2274
    }
  },
  {
    "ph": "s", "id": 2274, "pid": 76337, "tid": -914061504, "ts": 1716454216513905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216514842, "dur": 14,
    "args": {
      "External id": 2277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2277, "pid": 5, "tid": 7, "ts": 1716454216514842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513922, "dur": 7,
    "args": {
      "External id": 2277,
      "cbid": 211, "correlation": 2277
    }
  },
  {
    "ph": "s", "id": 2277, "pid": 76337, "tid": -914061504, "ts": 1716454216513922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216514857, "dur": 9,
    "args": {
      "External id": 2286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2286,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2286, "pid": 5, "tid": 7, "ts": 1716454216514857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216513990, "dur": 12,
    "args": {
      "External id": 2286,
      "cbid": 211, "correlation": 2286
    }
  },
  {
    "ph": "s", "id": 2286, "pid": 76337, "tid": -914061504, "ts": 1716454216513990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216514064, "dur": 0,
    "args": {
      "External id": 2296,
      "cbid": 317, "correlation": 2296
    }
  },
  {
    "ph": "f", "id": 2296, "pid": 76337, "tid": -914061504, "ts": 1716454216514064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216514065, "dur": 0,
    "args": {
      "External id": 2297,
      "cbid": 203, "correlation": 2297
    }
  },
  {
    "ph": "f", "id": 2297, "pid": 76337, "tid": -914061504, "ts": 1716454216514065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216514066, "dur": 0,
    "args": {
      "External id": 2298,
      "cbid": 205, "correlation": 2298
    }
  },
  {
    "ph": "f", "id": 2298, "pid": 76337, "tid": -914061504, "ts": 1716454216514066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216516857, "dur": 8,
    "args": {
      "External id": 2302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2302, "pid": 5, "tid": 7, "ts": 1716454216516857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216516833, "dur": 24,
    "args": {
      "External id": 2302,
      "cbid": 211, "correlation": 2302
    }
  },
  {
    "ph": "s", "id": 2302, "pid": 76337, "tid": -914061504, "ts": 1716454216516833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216516870, "dur": 4,
    "args": {
      "External id": 2304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1.6,
      "warps per SM": 12.8,
      "grid": [1, 16, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 20
    }
  },
  {
    "ph": "f", "id": 2304, "pid": 5, "tid": 7, "ts": 1716454216516870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216516860, "dur": 6,
    "args": {
      "External id": 2304,
      "cbid": 211, "correlation": 2304
    }
  },
  {
    "ph": "s", "id": 2304, "pid": 76337, "tid": -914061504, "ts": 1716454216516860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216516906, "dur": 1,
    "args": {
      "External id": 2306,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 2306,
      "bytes": 2688, "memory bandwidth (GB/s)": 1.423728813559322
    }
  },
  {
    "ph": "f", "id": 2306, "pid": 5, "tid": 7, "ts": 1716454216516906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216516875, "dur": 28,
    "args": {
      "External id": 2306,
      "cbid": 51, "correlation": 2306
    }
  },
  {
    "ph": "s", "id": 2306, "pid": 76337, "tid": -914061504, "ts": 1716454216516875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x32x64_stage1_warpsize2x1x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216517106, "dur": 58,
    "args": {
      "External id": 2307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2307,
      "registers per thread": 128,
      "shared memory": 12288,
      "blocks per SM": 8.4,
      "warps per SM": 33.6,
      "grid": [1, 48, 14],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 2307, "pid": 5, "tid": 7, "ts": 1716454216517106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216516914, "dur": 197,
    "args": {
      "External id": 2307,
      "cbid": 211, "correlation": 2307
    }
  },
  {
    "ph": "s", "id": 2307, "pid": 76337, "tid": -914061504, "ts": 1716454216516914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216517166, "dur": 6,
    "args": {
      "External id": 2309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1.2,
      "warps per SM": 9.6,
      "grid": [96, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 2309, "pid": 5, "tid": 7, "ts": 1716454216517166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216517118, "dur": 8,
    "args": {
      "External id": 2309,
      "cbid": 211, "correlation": 2309
    }
  },
  {
    "ph": "s", "id": 2309, "pid": 76337, "tid": -914061504, "ts": 1716454216517118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216517851, "dur": 7,
    "args": {
      "External id": 2315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 2315, "pid": 5, "tid": 7, "ts": 1716454216517851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216517838, "dur": 12,
    "args": {
      "External id": 2315,
      "cbid": 211, "correlation": 2315
    }
  },
  {
    "ph": "s", "id": 2315, "pid": 76337, "tid": -914061504, "ts": 1716454216517838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216518013, "dur": 0,
    "args": {
      "External id": 2325,
      "cbid": 317, "correlation": 2325
    }
  },
  {
    "ph": "f", "id": 2325, "pid": 76337, "tid": -914061504, "ts": 1716454216518013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216518014, "dur": 0,
    "args": {
      "External id": 2326,
      "cbid": 203, "correlation": 2326
    }
  },
  {
    "ph": "f", "id": 2326, "pid": 76337, "tid": -914061504, "ts": 1716454216518014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216518015, "dur": 0,
    "args": {
      "External id": 2327,
      "cbid": 205, "correlation": 2327
    }
  },
  {
    "ph": "f", "id": 2327, "pid": 76337, "tid": -914061504, "ts": 1716454216518015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216519949, "dur": 2,
    "args": {
      "External id": 2333,
      "cbid": 317, "correlation": 2333
    }
  },
  {
    "ph": "f", "id": 2333, "pid": 76337, "tid": -914061504, "ts": 1716454216519949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216519952, "dur": 6784,
    "args": {
      "External id": 2334,
      "cbid": 20, "correlation": 2334
    }
  },
  {
    "ph": "f", "id": 2334, "pid": 76337, "tid": -914061504, "ts": 1716454216519952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216526790, "dur": 69,
    "args": {
      "External id": 2335,
      "cbid": 251, "correlation": 2335
    }
  },
  {
    "ph": "f", "id": 2335, "pid": 76337, "tid": -914061504, "ts": 1716454216526790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216526887, "dur": 7,
    "args": {
      "External id": 2336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2336,
      "registers per thread": 106,
      "shared memory": 16640,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 2336, "pid": 5, "tid": 7, "ts": 1716454216526887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216526866, "dur": 22,
    "args": {
      "External id": 2336,
      "cbid": 211, "correlation": 2336
    }
  },
  {
    "ph": "s", "id": 2336, "pid": 76337, "tid": -914061504, "ts": 1716454216526866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216527553, "dur": 5,
    "args": {
      "External id": 2342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2342,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 2342, "pid": 5, "tid": 7, "ts": 1716454216527553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216527540, "dur": 13,
    "args": {
      "External id": 2342,
      "cbid": 211, "correlation": 2342
    }
  },
  {
    "ph": "s", "id": 2342, "pid": 76337, "tid": -914061504, "ts": 1716454216527540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216527952, "dur": 3,
    "args": {
      "External id": 2350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2350,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 2350, "pid": 5, "tid": 7, "ts": 1716454216527952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216527878, "dur": 75,
    "args": {
      "External id": 2350,
      "cbid": 211, "correlation": 2350
    }
  },
  {
    "ph": "s", "id": 2350, "pid": 76337, "tid": -914061504, "ts": 1716454216527878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216528081, "dur": 3,
    "args": {
      "External id": 2358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2358,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 2358, "pid": 5, "tid": 7, "ts": 1716454216528081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216528018, "dur": 63,
    "args": {
      "External id": 2358,
      "cbid": 211, "correlation": 2358
    }
  },
  {
    "ph": "s", "id": 2358, "pid": 76337, "tid": -914061504, "ts": 1716454216528018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216528221, "dur": 3,
    "args": {
      "External id": 2366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2366,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 2366, "pid": 5, "tid": 7, "ts": 1716454216528221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216528159, "dur": 62,
    "args": {
      "External id": 2366,
      "cbid": 211, "correlation": 2366
    }
  },
  {
    "ph": "s", "id": 2366, "pid": 76337, "tid": -914061504, "ts": 1716454216528159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216528254, "dur": 3,
    "args": {
      "External id": 2374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2374,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 2374, "pid": 5, "tid": 7, "ts": 1716454216528254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216528244, "dur": 9,
    "args": {
      "External id": 2374,
      "cbid": 211, "correlation": 2374
    }
  },
  {
    "ph": "s", "id": 2374, "pid": 76337, "tid": -914061504, "ts": 1716454216528244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216528528, "dur": 5,
    "args": {
      "External id": 2383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2383,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 8,
      "warps per SM": 128,
      "grid": [160, 4, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2383, "pid": 5, "tid": 7, "ts": 1716454216528528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216528447, "dur": 85,
    "args": {
      "External id": 2383,
      "cbid": 211, "correlation": 2383
    }
  },
  {
    "ph": "s", "id": 2383, "pid": 76337, "tid": -914061504, "ts": 1716454216528447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216528672, "dur": 3,
    "args": {
      "External id": 2394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 2394, "pid": 5, "tid": 7, "ts": 1716454216528672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216528657, "dur": 15,
    "args": {
      "External id": 2394,
      "cbid": 211, "correlation": 2394
    }
  },
  {
    "ph": "s", "id": 2394, "pid": 76337, "tid": -914061504, "ts": 1716454216528657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216528708, "dur": 5,
    "args": {
      "External id": 2404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2404,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2404, "pid": 5, "tid": 7, "ts": 1716454216528708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216528698, "dur": 9,
    "args": {
      "External id": 2404,
      "cbid": 211, "correlation": 2404
    }
  },
  {
    "ph": "s", "id": 2404, "pid": 76337, "tid": -914061504, "ts": 1716454216528698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454216529589, "dur": 95,
    "args": {
      "External id": 2419,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 2419,
      "bytes": 1179648, "memory bandwidth (GB/s)": 12.362302589522443
    }
  },
  {
    "ph": "f", "id": 2419, "pid": 5, "tid": 7, "ts": 1716454216529589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216529336, "dur": 271,
    "args": {
      "External id": 2419,
      "cbid": 41, "correlation": 2419
    }
  },
  {
    "ph": "s", "id": 2419, "pid": 76337, "tid": -914061504, "ts": 1716454216529336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454216529609, "dur": 87,
    "args": {
      "External id": 2420,
      "cbid": 131, "correlation": 2420
    }
  },
  {
    "ph": "f", "id": 2420, "pid": 76337, "tid": -914061504, "ts": 1716454216529609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216529729, "dur": 8,
    "args": {
      "External id": 2424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2424,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 2424, "pid": 5, "tid": 7, "ts": 1716454216529729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216529710, "dur": 18,
    "args": {
      "External id": 2424,
      "cbid": 211, "correlation": 2424
    }
  },
  {
    "ph": "s", "id": 2424, "pid": 76337, "tid": -914061504, "ts": 1716454216529710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454216530735, "dur": 94,
    "args": {
      "External id": 2440,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 2440,
      "bytes": 1179648, "memory bandwidth (GB/s)": 12.496403563597072
    }
  },
  {
    "ph": "f", "id": 2440, "pid": 5, "tid": 7, "ts": 1716454216530735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216530511, "dur": 241,
    "args": {
      "External id": 2440,
      "cbid": 41, "correlation": 2440
    }
  },
  {
    "ph": "s", "id": 2440, "pid": 76337, "tid": -914061504, "ts": 1716454216530511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454216530753, "dur": 83,
    "args": {
      "External id": 2441,
      "cbid": 131, "correlation": 2441
    }
  },
  {
    "ph": "f", "id": 2441, "pid": 76337, "tid": -914061504, "ts": 1716454216530753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216530861, "dur": 6,
    "args": {
      "External id": 2445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 2445, "pid": 5, "tid": 7, "ts": 1716454216530861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216530846, "dur": 14,
    "args": {
      "External id": 2445,
      "cbid": 211, "correlation": 2445
    }
  },
  {
    "ph": "s", "id": 2445, "pid": 76337, "tid": -914061504, "ts": 1716454216530846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216531096, "dur": 5,
    "args": {
      "External id": 2460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0875,
      "warps per SM": 0.35,
      "grid": [7, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 2460, "pid": 5, "tid": 7, "ts": 1716454216531096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531080, "dur": 16,
    "args": {
      "External id": 2460,
      "cbid": 211, "correlation": 2460
    }
  },
  {
    "ph": "s", "id": 2460, "pid": 76337, "tid": -914061504, "ts": 1716454216531080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531113, "dur": 1,
    "args": {
      "External id": 2467,
      "cbid": 317, "correlation": 2467
    }
  },
  {
    "ph": "f", "id": 2467, "pid": 76337, "tid": -914061504, "ts": 1716454216531113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531115, "dur": 1,
    "args": {
      "External id": 2468,
      "cbid": 203, "correlation": 2468
    }
  },
  {
    "ph": "f", "id": 2468, "pid": 76337, "tid": -914061504, "ts": 1716454216531115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531116, "dur": 1,
    "args": {
      "External id": 2469,
      "cbid": 205, "correlation": 2469
    }
  },
  {
    "ph": "f", "id": 2469, "pid": 76337, "tid": -914061504, "ts": 1716454216531116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_wo_smem_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x32x16_stage1_warpsize4x1x1_g1_tensor8x8x4_aligna2_alignc8_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216531167, "dur": 199,
    "args": {
      "External id": 2473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2473,
      "registers per thread": 111,
      "shared memory": 4096,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [4, 1536, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 2473, "pid": 5, "tid": 7, "ts": 1716454216531167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531153, "dur": 14,
    "args": {
      "External id": 2473,
      "cbid": 211, "correlation": 2473
    }
  },
  {
    "ph": "s", "id": 2473, "pid": 76337, "tid": -914061504, "ts": 1716454216531153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216531368, "dur": 180,
    "args": {
      "External id": 2479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2479, "pid": 5, "tid": 7, "ts": 1716454216531368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531203, "dur": 10,
    "args": {
      "External id": 2479,
      "cbid": 211, "correlation": 2479
    }
  },
  {
    "ph": "s", "id": 2479, "pid": 76337, "tid": -914061504, "ts": 1716454216531203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216531549, "dur": 1803,
    "args": {
      "External id": 2489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2489,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2489, "pid": 5, "tid": 7, "ts": 1716454216531549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531327, "dur": 12,
    "args": {
      "External id": 2489,
      "cbid": 211, "correlation": 2489
    }
  },
  {
    "ph": "s", "id": 2489, "pid": 76337, "tid": -914061504, "ts": 1716454216531327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216533354, "dur": 704,
    "args": {
      "External id": 2510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2510,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2510, "pid": 5, "tid": 7, "ts": 1716454216533354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531372, "dur": 10,
    "args": {
      "External id": 2510,
      "cbid": 211, "correlation": 2510
    }
  },
  {
    "ph": "s", "id": 2510, "pid": 76337, "tid": -914061504, "ts": 1716454216531372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216534059, "dur": 5,
    "args": {
      "External id": 2522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2522,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2522, "pid": 5, "tid": 7, "ts": 1716454216534059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531395, "dur": 7,
    "args": {
      "External id": 2522,
      "cbid": 211, "correlation": 2522
    }
  },
  {
    "ph": "s", "id": 2522, "pid": 76337, "tid": -914061504, "ts": 1716454216531395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216534065, "dur": 178,
    "args": {
      "External id": 2525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2525,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2525, "pid": 5, "tid": 7, "ts": 1716454216534065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531419, "dur": 8,
    "args": {
      "External id": 2525,
      "cbid": 211, "correlation": 2525
    }
  },
  {
    "ph": "s", "id": 2525, "pid": 76337, "tid": -914061504, "ts": 1716454216531419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216534244, "dur": 110,
    "args": {
      "External id": 2534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2534,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2534, "pid": 5, "tid": 7, "ts": 1716454216534244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531476, "dur": 11,
    "args": {
      "External id": 2534,
      "cbid": 211, "correlation": 2534
    }
  },
  {
    "ph": "s", "id": 2534, "pid": 76337, "tid": -914061504, "ts": 1716454216531476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531533, "dur": 0,
    "args": {
      "External id": 2544,
      "cbid": 317, "correlation": 2544
    }
  },
  {
    "ph": "f", "id": 2544, "pid": 76337, "tid": -914061504, "ts": 1716454216531533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531533, "dur": 0,
    "args": {
      "External id": 2545,
      "cbid": 203, "correlation": 2545
    }
  },
  {
    "ph": "f", "id": 2545, "pid": 76337, "tid": -914061504, "ts": 1716454216531533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531534, "dur": 0,
    "args": {
      "External id": 2546,
      "cbid": 205, "correlation": 2546
    }
  },
  {
    "ph": "f", "id": 2546, "pid": 76337, "tid": -914061504, "ts": 1716454216531534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216534356, "dur": 143,
    "args": {
      "External id": 2550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2550,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2550, "pid": 5, "tid": 7, "ts": 1716454216534356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531559, "dur": 14,
    "args": {
      "External id": 2550,
      "cbid": 211, "correlation": 2550
    }
  },
  {
    "ph": "s", "id": 2550, "pid": 76337, "tid": -914061504, "ts": 1716454216531559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216534501, "dur": 6,
    "args": {
      "External id": 2552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2552,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 2552, "pid": 5, "tid": 7, "ts": 1716454216534501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531576, "dur": 5,
    "args": {
      "External id": 2552,
      "cbid": 211, "correlation": 2552
    }
  },
  {
    "ph": "s", "id": 2552, "pid": 76337, "tid": -914061504, "ts": 1716454216531576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216534509, "dur": 5,
    "args": {
      "External id": 2554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2554,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2554, "pid": 5, "tid": 7, "ts": 1716454216534509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531589, "dur": 8,
    "args": {
      "External id": 2554,
      "cbid": 211, "correlation": 2554
    }
  },
  {
    "ph": "s", "id": 2554, "pid": 76337, "tid": -914061504, "ts": 1716454216531589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531602, "dur": 0,
    "args": {
      "External id": 2555,
      "cbid": 51, "correlation": 2555
    }
  },
  {
    "ph": "s", "id": 2555, "pid": 76337, "tid": -914061504, "ts": 1716454216531602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216534515, "dur": 824,
    "args": {
      "External id": 2556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2556,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2556, "pid": 5, "tid": 7, "ts": 1716454216534515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531604, "dur": 7,
    "args": {
      "External id": 2556,
      "cbid": 211, "correlation": 2556
    }
  },
  {
    "ph": "s", "id": 2556, "pid": 76337, "tid": -914061504, "ts": 1716454216531604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216535340, "dur": 176,
    "args": {
      "External id": 2561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2561,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2561, "pid": 5, "tid": 7, "ts": 1716454216535340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531634, "dur": 8,
    "args": {
      "External id": 2561,
      "cbid": 211, "correlation": 2561
    }
  },
  {
    "ph": "s", "id": 2561, "pid": 76337, "tid": -914061504, "ts": 1716454216531634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216535518, "dur": 699,
    "args": {
      "External id": 2581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2581,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2581, "pid": 5, "tid": 7, "ts": 1716454216535518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531702, "dur": 11,
    "args": {
      "External id": 2581,
      "cbid": 211, "correlation": 2581
    }
  },
  {
    "ph": "s", "id": 2581, "pid": 76337, "tid": -914061504, "ts": 1716454216531702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216536218, "dur": 4,
    "args": {
      "External id": 2593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2593,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2593, "pid": 5, "tid": 7, "ts": 1716454216536218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531723, "dur": 6,
    "args": {
      "External id": 2593,
      "cbid": 211, "correlation": 2593
    }
  },
  {
    "ph": "s", "id": 2593, "pid": 76337, "tid": -914061504, "ts": 1716454216531723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216536224, "dur": 177,
    "args": {
      "External id": 2596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2596,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2596, "pid": 5, "tid": 7, "ts": 1716454216536224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531741, "dur": 7,
    "args": {
      "External id": 2596,
      "cbid": 211, "correlation": 2596
    }
  },
  {
    "ph": "s", "id": 2596, "pid": 76337, "tid": -914061504, "ts": 1716454216531741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216536403, "dur": 109,
    "args": {
      "External id": 2605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2605,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2605, "pid": 5, "tid": 7, "ts": 1716454216536403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531779, "dur": 9,
    "args": {
      "External id": 2605,
      "cbid": 211, "correlation": 2605
    }
  },
  {
    "ph": "s", "id": 2605, "pid": 76337, "tid": -914061504, "ts": 1716454216531779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531854, "dur": 0,
    "args": {
      "External id": 2615,
      "cbid": 317, "correlation": 2615
    }
  },
  {
    "ph": "f", "id": 2615, "pid": 76337, "tid": -914061504, "ts": 1716454216531854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531855, "dur": 0,
    "args": {
      "External id": 2616,
      "cbid": 203, "correlation": 2616
    }
  },
  {
    "ph": "f", "id": 2616, "pid": 76337, "tid": -914061504, "ts": 1716454216531855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531856, "dur": 0,
    "args": {
      "External id": 2617,
      "cbid": 205, "correlation": 2617
    }
  },
  {
    "ph": "f", "id": 2617, "pid": 76337, "tid": -914061504, "ts": 1716454216531856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216536514, "dur": 128,
    "args": {
      "External id": 2621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2621,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2621, "pid": 5, "tid": 7, "ts": 1716454216536514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531869, "dur": 11,
    "args": {
      "External id": 2621,
      "cbid": 211, "correlation": 2621
    }
  },
  {
    "ph": "s", "id": 2621, "pid": 76337, "tid": -914061504, "ts": 1716454216531869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216536643, "dur": 6,
    "args": {
      "External id": 2623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2623,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 2623, "pid": 5, "tid": 7, "ts": 1716454216536643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531882, "dur": 5,
    "args": {
      "External id": 2623,
      "cbid": 211, "correlation": 2623
    }
  },
  {
    "ph": "s", "id": 2623, "pid": 76337, "tid": -914061504, "ts": 1716454216531882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216536651, "dur": 5,
    "args": {
      "External id": 2625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2625,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2625, "pid": 5, "tid": 7, "ts": 1716454216536651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531892, "dur": 5,
    "args": {
      "External id": 2625,
      "cbid": 211, "correlation": 2625
    }
  },
  {
    "ph": "s", "id": 2625, "pid": 76337, "tid": -914061504, "ts": 1716454216531892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531900, "dur": 0,
    "args": {
      "External id": 2626,
      "cbid": 51, "correlation": 2626
    }
  },
  {
    "ph": "s", "id": 2626, "pid": 76337, "tid": -914061504, "ts": 1716454216531900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216536657, "dur": 818,
    "args": {
      "External id": 2627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2627,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2627, "pid": 5, "tid": 7, "ts": 1716454216536657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531901, "dur": 5,
    "args": {
      "External id": 2627,
      "cbid": 211, "correlation": 2627
    }
  },
  {
    "ph": "s", "id": 2627, "pid": 76337, "tid": -914061504, "ts": 1716454216531901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216537476, "dur": 177,
    "args": {
      "External id": 2632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2632,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2632, "pid": 5, "tid": 7, "ts": 1716454216537476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531928, "dur": 8,
    "args": {
      "External id": 2632,
      "cbid": 211, "correlation": 2632
    }
  },
  {
    "ph": "s", "id": 2632, "pid": 76337, "tid": -914061504, "ts": 1716454216531928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216537655, "dur": 750,
    "args": {
      "External id": 2640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2640, "pid": 5, "tid": 7, "ts": 1716454216537655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216531967, "dur": 19,
    "args": {
      "External id": 2640,
      "cbid": 211, "correlation": 2640
    }
  },
  {
    "ph": "s", "id": 2640, "pid": 76337, "tid": -914061504, "ts": 1716454216531967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216538406, "dur": 104,
    "args": {
      "External id": 2648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2648,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2648, "pid": 5, "tid": 7, "ts": 1716454216538406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532020, "dur": 12,
    "args": {
      "External id": 2648,
      "cbid": 211, "correlation": 2648
    }
  },
  {
    "ph": "s", "id": 2648, "pid": 76337, "tid": -914061504, "ts": 1716454216532020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216538511, "dur": 1797,
    "args": {
      "External id": 2658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2658,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2658, "pid": 5, "tid": 7, "ts": 1716454216538511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532094, "dur": 12,
    "args": {
      "External id": 2658,
      "cbid": 211, "correlation": 2658
    }
  },
  {
    "ph": "s", "id": 2658, "pid": 76337, "tid": -914061504, "ts": 1716454216532094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216540310, "dur": 702,
    "args": {
      "External id": 2679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2679,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2679, "pid": 5, "tid": 7, "ts": 1716454216540310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532125, "dur": 7,
    "args": {
      "External id": 2679,
      "cbid": 211, "correlation": 2679
    }
  },
  {
    "ph": "s", "id": 2679, "pid": 76337, "tid": -914061504, "ts": 1716454216532125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216541013, "dur": 4,
    "args": {
      "External id": 2691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2691,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2691, "pid": 5, "tid": 7, "ts": 1716454216541013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532142, "dur": 6,
    "args": {
      "External id": 2691,
      "cbid": 211, "correlation": 2691
    }
  },
  {
    "ph": "s", "id": 2691, "pid": 76337, "tid": -914061504, "ts": 1716454216532142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216541019, "dur": 178,
    "args": {
      "External id": 2694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2694, "pid": 5, "tid": 7, "ts": 1716454216541019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532158, "dur": 6,
    "args": {
      "External id": 2694,
      "cbid": 211, "correlation": 2694
    }
  },
  {
    "ph": "s", "id": 2694, "pid": 76337, "tid": -914061504, "ts": 1716454216532158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216541198, "dur": 109,
    "args": {
      "External id": 2703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2703,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2703, "pid": 5, "tid": 7, "ts": 1716454216541198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532197, "dur": 10,
    "args": {
      "External id": 2703,
      "cbid": 211, "correlation": 2703
    }
  },
  {
    "ph": "s", "id": 2703, "pid": 76337, "tid": -914061504, "ts": 1716454216532197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532250, "dur": 0,
    "args": {
      "External id": 2713,
      "cbid": 317, "correlation": 2713
    }
  },
  {
    "ph": "f", "id": 2713, "pid": 76337, "tid": -914061504, "ts": 1716454216532250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532250, "dur": 0,
    "args": {
      "External id": 2714,
      "cbid": 203, "correlation": 2714
    }
  },
  {
    "ph": "f", "id": 2714, "pid": 76337, "tid": -914061504, "ts": 1716454216532250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532251, "dur": 0,
    "args": {
      "External id": 2715,
      "cbid": 205, "correlation": 2715
    }
  },
  {
    "ph": "f", "id": 2715, "pid": 76337, "tid": -914061504, "ts": 1716454216532251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216541309, "dur": 135,
    "args": {
      "External id": 2719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2719,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2719, "pid": 5, "tid": 7, "ts": 1716454216541309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532264, "dur": 12,
    "args": {
      "External id": 2719,
      "cbid": 211, "correlation": 2719
    }
  },
  {
    "ph": "s", "id": 2719, "pid": 76337, "tid": -914061504, "ts": 1716454216532264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216541445, "dur": 6,
    "args": {
      "External id": 2721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2721,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 2721, "pid": 5, "tid": 7, "ts": 1716454216541445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532279, "dur": 5,
    "args": {
      "External id": 2721,
      "cbid": 211, "correlation": 2721
    }
  },
  {
    "ph": "s", "id": 2721, "pid": 76337, "tid": -914061504, "ts": 1716454216532279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216541453, "dur": 5,
    "args": {
      "External id": 2723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2723,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2723, "pid": 5, "tid": 7, "ts": 1716454216541453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532288, "dur": 5,
    "args": {
      "External id": 2723,
      "cbid": 211, "correlation": 2723
    }
  },
  {
    "ph": "s", "id": 2723, "pid": 76337, "tid": -914061504, "ts": 1716454216532288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532296, "dur": 0,
    "args": {
      "External id": 2724,
      "cbid": 51, "correlation": 2724
    }
  },
  {
    "ph": "s", "id": 2724, "pid": 76337, "tid": -914061504, "ts": 1716454216532296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216541460, "dur": 817,
    "args": {
      "External id": 2725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2725,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2725, "pid": 5, "tid": 7, "ts": 1716454216541460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532297, "dur": 5,
    "args": {
      "External id": 2725,
      "cbid": 211, "correlation": 2725
    }
  },
  {
    "ph": "s", "id": 2725, "pid": 76337, "tid": -914061504, "ts": 1716454216532297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216542278, "dur": 175,
    "args": {
      "External id": 2730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2730, "pid": 5, "tid": 7, "ts": 1716454216542278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532322, "dur": 9,
    "args": {
      "External id": 2730,
      "cbid": 211, "correlation": 2730
    }
  },
  {
    "ph": "s", "id": 2730, "pid": 76337, "tid": -914061504, "ts": 1716454216532322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216542454, "dur": 698,
    "args": {
      "External id": 2750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2750,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2750, "pid": 5, "tid": 7, "ts": 1716454216542454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532390, "dur": 11,
    "args": {
      "External id": 2750,
      "cbid": 211, "correlation": 2750
    }
  },
  {
    "ph": "s", "id": 2750, "pid": 76337, "tid": -914061504, "ts": 1716454216532390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216543154, "dur": 4,
    "args": {
      "External id": 2762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2762,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2762, "pid": 5, "tid": 7, "ts": 1716454216543154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532410, "dur": 6,
    "args": {
      "External id": 2762,
      "cbid": 211, "correlation": 2762
    }
  },
  {
    "ph": "s", "id": 2762, "pid": 76337, "tid": -914061504, "ts": 1716454216532410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216543159, "dur": 175,
    "args": {
      "External id": 2765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2765, "pid": 5, "tid": 7, "ts": 1716454216543159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532427, "dur": 6,
    "args": {
      "External id": 2765,
      "cbid": 211, "correlation": 2765
    }
  },
  {
    "ph": "s", "id": 2765, "pid": 76337, "tid": -914061504, "ts": 1716454216532427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216543336, "dur": 109,
    "args": {
      "External id": 2774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2774,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2774, "pid": 5, "tid": 7, "ts": 1716454216543336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532466, "dur": 9,
    "args": {
      "External id": 2774,
      "cbid": 211, "correlation": 2774
    }
  },
  {
    "ph": "s", "id": 2774, "pid": 76337, "tid": -914061504, "ts": 1716454216532466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532527, "dur": 0,
    "args": {
      "External id": 2784,
      "cbid": 317, "correlation": 2784
    }
  },
  {
    "ph": "f", "id": 2784, "pid": 76337, "tid": -914061504, "ts": 1716454216532527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532527, "dur": 0,
    "args": {
      "External id": 2785,
      "cbid": 203, "correlation": 2785
    }
  },
  {
    "ph": "f", "id": 2785, "pid": 76337, "tid": -914061504, "ts": 1716454216532527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532528, "dur": 0,
    "args": {
      "External id": 2786,
      "cbid": 205, "correlation": 2786
    }
  },
  {
    "ph": "f", "id": 2786, "pid": 76337, "tid": -914061504, "ts": 1716454216532528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216543446, "dur": 132,
    "args": {
      "External id": 2790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2790, "pid": 5, "tid": 7, "ts": 1716454216543446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532540, "dur": 11,
    "args": {
      "External id": 2790,
      "cbid": 211, "correlation": 2790
    }
  },
  {
    "ph": "s", "id": 2790, "pid": 76337, "tid": -914061504, "ts": 1716454216532540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216543580, "dur": 6,
    "args": {
      "External id": 2792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2792,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 2792, "pid": 5, "tid": 7, "ts": 1716454216543580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532554, "dur": 5,
    "args": {
      "External id": 2792,
      "cbid": 211, "correlation": 2792
    }
  },
  {
    "ph": "s", "id": 2792, "pid": 76337, "tid": -914061504, "ts": 1716454216532554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216543588, "dur": 5,
    "args": {
      "External id": 2794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2794,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2794, "pid": 5, "tid": 7, "ts": 1716454216543588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532563, "dur": 6,
    "args": {
      "External id": 2794,
      "cbid": 211, "correlation": 2794
    }
  },
  {
    "ph": "s", "id": 2794, "pid": 76337, "tid": -914061504, "ts": 1716454216532563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532572, "dur": 0,
    "args": {
      "External id": 2795,
      "cbid": 51, "correlation": 2795
    }
  },
  {
    "ph": "s", "id": 2795, "pid": 76337, "tid": -914061504, "ts": 1716454216532572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216543594, "dur": 817,
    "args": {
      "External id": 2796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2796,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2796, "pid": 5, "tid": 7, "ts": 1716454216543594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532573, "dur": 5,
    "args": {
      "External id": 2796,
      "cbid": 211, "correlation": 2796
    }
  },
  {
    "ph": "s", "id": 2796, "pid": 76337, "tid": -914061504, "ts": 1716454216532573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216544413, "dur": 177,
    "args": {
      "External id": 2801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2801, "pid": 5, "tid": 7, "ts": 1716454216544413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532598, "dur": 8,
    "args": {
      "External id": 2801,
      "cbid": 211, "correlation": 2801
    }
  },
  {
    "ph": "s", "id": 2801, "pid": 76337, "tid": -914061504, "ts": 1716454216532598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216544591, "dur": 746,
    "args": {
      "External id": 2809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2809, "pid": 5, "tid": 7, "ts": 1716454216544591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532632, "dur": 9,
    "args": {
      "External id": 2809,
      "cbid": 211, "correlation": 2809
    }
  },
  {
    "ph": "s", "id": 2809, "pid": 76337, "tid": -914061504, "ts": 1716454216532632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216545339, "dur": 103,
    "args": {
      "External id": 2817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2817,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2817, "pid": 5, "tid": 7, "ts": 1716454216545339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532662, "dur": 8,
    "args": {
      "External id": 2817,
      "cbid": 211, "correlation": 2817
    }
  },
  {
    "ph": "s", "id": 2817, "pid": 76337, "tid": -914061504, "ts": 1716454216532662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216545443, "dur": 88,
    "args": {
      "External id": 2827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2827,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 617.2125,
      "warps per SM": 2468.85,
      "grid": [49377, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2827, "pid": 5, "tid": 7, "ts": 1716454216545443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532738, "dur": 12,
    "args": {
      "External id": 2827,
      "cbid": 211, "correlation": 2827
    }
  },
  {
    "ph": "s", "id": 2827, "pid": 76337, "tid": -914061504, "ts": 1716454216532738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216545533, "dur": 166,
    "args": {
      "External id": 2832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2832, "pid": 5, "tid": 7, "ts": 1716454216545533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532778, "dur": 7,
    "args": {
      "External id": 2832,
      "cbid": 211, "correlation": 2832
    }
  },
  {
    "ph": "s", "id": 2832, "pid": 76337, "tid": -914061504, "ts": 1716454216532778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216545700, "dur": 6,
    "args": {
      "External id": 2847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 3.6,
      "warps per SM": 14.4,
      "grid": [288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 22
    }
  },
  {
    "ph": "f", "id": 2847, "pid": 5, "tid": 7, "ts": 1716454216545700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532845, "dur": 12,
    "args": {
      "External id": 2847,
      "cbid": 211, "correlation": 2847
    }
  },
  {
    "ph": "s", "id": 2847, "pid": 76337, "tid": -914061504, "ts": 1716454216532845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532866, "dur": 0,
    "args": {
      "External id": 2854,
      "cbid": 317, "correlation": 2854
    }
  },
  {
    "ph": "f", "id": 2854, "pid": 76337, "tid": -914061504, "ts": 1716454216532866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532867, "dur": 0,
    "args": {
      "External id": 2855,
      "cbid": 203, "correlation": 2855
    }
  },
  {
    "ph": "f", "id": 2855, "pid": 76337, "tid": -914061504, "ts": 1716454216532867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532868, "dur": 0,
    "args": {
      "External id": 2856,
      "cbid": 205, "correlation": 2856
    }
  },
  {
    "ph": "f", "id": 2856, "pid": 76337, "tid": -914061504, "ts": 1716454216532868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216545707, "dur": 265,
    "args": {
      "External id": 2860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2860,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [1, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2860, "pid": 5, "tid": 7, "ts": 1716454216545707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532887, "dur": 12,
    "args": {
      "External id": 2860,
      "cbid": 211, "correlation": 2860
    }
  },
  {
    "ph": "s", "id": 2860, "pid": 76337, "tid": -914061504, "ts": 1716454216532887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216545973, "dur": 49,
    "args": {
      "External id": 2866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2866, "pid": 5, "tid": 7, "ts": 1716454216545973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216532921, "dur": 9,
    "args": {
      "External id": 2866,
      "cbid": 211, "correlation": 2866
    }
  },
  {
    "ph": "s", "id": 2866, "pid": 76337, "tid": -914061504, "ts": 1716454216532921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216546023, "dur": 144,
    "args": {
      "External id": 2876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2876, "pid": 5, "tid": 7, "ts": 1716454216546023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533014, "dur": 13,
    "args": {
      "External id": 2876,
      "cbid": 211, "correlation": 2876
    }
  },
  {
    "ph": "s", "id": 2876, "pid": 76337, "tid": -914061504, "ts": 1716454216533014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216546169, "dur": 180,
    "args": {
      "External id": 2897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2897,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2897, "pid": 5, "tid": 7, "ts": 1716454216546169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533050, "dur": 7,
    "args": {
      "External id": 2897,
      "cbid": 211, "correlation": 2897
    }
  },
  {
    "ph": "s", "id": 2897, "pid": 76337, "tid": -914061504, "ts": 1716454216533050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216546351, "dur": 4,
    "args": {
      "External id": 2909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2909,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2909, "pid": 5, "tid": 7, "ts": 1716454216546351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533066, "dur": 6,
    "args": {
      "External id": 2909,
      "cbid": 211, "correlation": 2909
    }
  },
  {
    "ph": "s", "id": 2909, "pid": 76337, "tid": -914061504, "ts": 1716454216533066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216546356, "dur": 48,
    "args": {
      "External id": 2912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2912, "pid": 5, "tid": 7, "ts": 1716454216546356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533083, "dur": 6,
    "args": {
      "External id": 2912,
      "cbid": 211, "correlation": 2912
    }
  },
  {
    "ph": "s", "id": 2912, "pid": 76337, "tid": -914061504, "ts": 1716454216533083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216546406, "dur": 31,
    "args": {
      "External id": 2921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2921,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2921, "pid": 5, "tid": 7, "ts": 1716454216546406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533123, "dur": 9,
    "args": {
      "External id": 2921,
      "cbid": 211, "correlation": 2921
    }
  },
  {
    "ph": "s", "id": 2921, "pid": 76337, "tid": -914061504, "ts": 1716454216533123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533176, "dur": 0,
    "args": {
      "External id": 2931,
      "cbid": 317, "correlation": 2931
    }
  },
  {
    "ph": "f", "id": 2931, "pid": 76337, "tid": -914061504, "ts": 1716454216533176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533177, "dur": 0,
    "args": {
      "External id": 2932,
      "cbid": 203, "correlation": 2932
    }
  },
  {
    "ph": "f", "id": 2932, "pid": 76337, "tid": -914061504, "ts": 1716454216533177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533177, "dur": 0,
    "args": {
      "External id": 2933,
      "cbid": 205, "correlation": 2933
    }
  },
  {
    "ph": "f", "id": 2933, "pid": 76337, "tid": -914061504, "ts": 1716454216533177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216546439, "dur": 39,
    "args": {
      "External id": 2937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2937,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [1536, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2937, "pid": 5, "tid": 7, "ts": 1716454216546439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533192, "dur": 11,
    "args": {
      "External id": 2937,
      "cbid": 211, "correlation": 2937
    }
  },
  {
    "ph": "s", "id": 2937, "pid": 76337, "tid": -914061504, "ts": 1716454216533192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216546480, "dur": 8,
    "args": {
      "External id": 2939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2939,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12.8,
      "warps per SM": 102.4,
      "grid": [1, 4, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2939, "pid": 5, "tid": 7, "ts": 1716454216546480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533207, "dur": 5,
    "args": {
      "External id": 2939,
      "cbid": 211, "correlation": 2939
    }
  },
  {
    "ph": "s", "id": 2939, "pid": 76337, "tid": -914061504, "ts": 1716454216533207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216546489, "dur": 4,
    "args": {
      "External id": 2941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 2941, "pid": 5, "tid": 7, "ts": 1716454216546489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533217, "dur": 6,
    "args": {
      "External id": 2941,
      "cbid": 211, "correlation": 2941
    }
  },
  {
    "ph": "s", "id": 2941, "pid": 76337, "tid": -914061504, "ts": 1716454216533217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533227, "dur": 0,
    "args": {
      "External id": 2942,
      "cbid": 51, "correlation": 2942
    }
  },
  {
    "ph": "s", "id": 2942, "pid": 76337, "tid": -914061504, "ts": 1716454216533227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216546495, "dur": 435,
    "args": {
      "External id": 2943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2943,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [192, 4, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 2943, "pid": 5, "tid": 7, "ts": 1716454216546495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533228, "dur": 6,
    "args": {
      "External id": 2943,
      "cbid": 211, "correlation": 2943
    }
  },
  {
    "ph": "s", "id": 2943, "pid": 76337, "tid": -914061504, "ts": 1716454216533228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216546931, "dur": 91,
    "args": {
      "External id": 2948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2948, "pid": 5, "tid": 7, "ts": 1716454216546931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533257, "dur": 9,
    "args": {
      "External id": 2948,
      "cbid": 211, "correlation": 2948
    }
  },
  {
    "ph": "s", "id": 2948, "pid": 76337, "tid": -914061504, "ts": 1716454216533257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216547024, "dur": 367,
    "args": {
      "External id": 2968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2968,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 2968, "pid": 5, "tid": 7, "ts": 1716454216547024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533326, "dur": 11,
    "args": {
      "External id": 2968,
      "cbid": 211, "correlation": 2968
    }
  },
  {
    "ph": "s", "id": 2968, "pid": 76337, "tid": -914061504, "ts": 1716454216533326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216547393, "dur": 4,
    "args": {
      "External id": 2980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2980,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 2980, "pid": 5, "tid": 7, "ts": 1716454216547393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533349, "dur": 6,
    "args": {
      "External id": 2980,
      "cbid": 211, "correlation": 2980
    }
  },
  {
    "ph": "s", "id": 2980, "pid": 76337, "tid": -914061504, "ts": 1716454216533349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216547398, "dur": 90,
    "args": {
      "External id": 2983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2983, "pid": 5, "tid": 7, "ts": 1716454216547398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533366, "dur": 8,
    "args": {
      "External id": 2983,
      "cbid": 211, "correlation": 2983
    }
  },
  {
    "ph": "s", "id": 2983, "pid": 76337, "tid": -914061504, "ts": 1716454216533366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216547490, "dur": 58,
    "args": {
      "External id": 2992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 2992,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 2992, "pid": 5, "tid": 7, "ts": 1716454216547490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533406, "dur": 9,
    "args": {
      "External id": 2992,
      "cbid": 211, "correlation": 2992
    }
  },
  {
    "ph": "s", "id": 2992, "pid": 76337, "tid": -914061504, "ts": 1716454216533406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533468, "dur": 0,
    "args": {
      "External id": 3002,
      "cbid": 317, "correlation": 3002
    }
  },
  {
    "ph": "f", "id": 3002, "pid": 76337, "tid": -914061504, "ts": 1716454216533468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533469, "dur": 0,
    "args": {
      "External id": 3003,
      "cbid": 203, "correlation": 3003
    }
  },
  {
    "ph": "f", "id": 3003, "pid": 76337, "tid": -914061504, "ts": 1716454216533469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533470, "dur": 0,
    "args": {
      "External id": 3004,
      "cbid": 205, "correlation": 3004
    }
  },
  {
    "ph": "f", "id": 3004, "pid": 76337, "tid": -914061504, "ts": 1716454216533470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216547549, "dur": 102,
    "args": {
      "External id": 3008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3008,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3008, "pid": 5, "tid": 7, "ts": 1716454216547549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533485, "dur": 12,
    "args": {
      "External id": 3008,
      "cbid": 211, "correlation": 3008
    }
  },
  {
    "ph": "s", "id": 3008, "pid": 76337, "tid": -914061504, "ts": 1716454216533485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216547653, "dur": 12,
    "args": {
      "External id": 3010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3010,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3010, "pid": 5, "tid": 7, "ts": 1716454216547653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533499, "dur": 5,
    "args": {
      "External id": 3010,
      "cbid": 211, "correlation": 3010
    }
  },
  {
    "ph": "s", "id": 3010, "pid": 76337, "tid": -914061504, "ts": 1716454216533499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216547666, "dur": 4,
    "args": {
      "External id": 3012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3012,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 3012, "pid": 5, "tid": 7, "ts": 1716454216547666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533509, "dur": 5,
    "args": {
      "External id": 3012,
      "cbid": 211, "correlation": 3012
    }
  },
  {
    "ph": "s", "id": 3012, "pid": 76337, "tid": -914061504, "ts": 1716454216533509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533517, "dur": 0,
    "args": {
      "External id": 3013,
      "cbid": 51, "correlation": 3013
    }
  },
  {
    "ph": "s", "id": 3013, "pid": 76337, "tid": -914061504, "ts": 1716454216533517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216547671, "dur": 777,
    "args": {
      "External id": 3014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3014,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3014, "pid": 5, "tid": 7, "ts": 1716454216547671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533518, "dur": 5,
    "args": {
      "External id": 3014,
      "cbid": 211, "correlation": 3014
    }
  },
  {
    "ph": "s", "id": 3014, "pid": 76337, "tid": -914061504, "ts": 1716454216533518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216548449, "dur": 91,
    "args": {
      "External id": 3019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3019, "pid": 5, "tid": 7, "ts": 1716454216548449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533546, "dur": 9,
    "args": {
      "External id": 3019,
      "cbid": 211, "correlation": 3019
    }
  },
  {
    "ph": "s", "id": 3019, "pid": 76337, "tid": -914061504, "ts": 1716454216533546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533604, "dur": 0,
    "args": {
      "External id": 3029,
      "cbid": 317, "correlation": 3029
    }
  },
  {
    "ph": "f", "id": 3029, "pid": 76337, "tid": -914061504, "ts": 1716454216533604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533604, "dur": 0,
    "args": {
      "External id": 3030,
      "cbid": 203, "correlation": 3030
    }
  },
  {
    "ph": "f", "id": 3030, "pid": 76337, "tid": -914061504, "ts": 1716454216533604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533605, "dur": 0,
    "args": {
      "External id": 3031,
      "cbid": 205, "correlation": 3031
    }
  },
  {
    "ph": "f", "id": 3031, "pid": 76337, "tid": -914061504, "ts": 1716454216533605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533667, "dur": 4,
    "args": {
      "External id": 3035,
      "cbid": 251, "correlation": 3035
    }
  },
  {
    "ph": "f", "id": 3035, "pid": 76337, "tid": -914061504, "ts": 1716454216533667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216548541, "dur": 85,
    "args": {
      "External id": 3036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3036,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [2, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3036, "pid": 5, "tid": 7, "ts": 1716454216548541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533677, "dur": 15,
    "args": {
      "External id": 3036,
      "cbid": 211, "correlation": 3036
    }
  },
  {
    "ph": "s", "id": 3036, "pid": 76337, "tid": -914061504, "ts": 1716454216533677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216548627, "dur": 92,
    "args": {
      "External id": 3042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3042,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3042, "pid": 5, "tid": 7, "ts": 1716454216548627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533714, "dur": 9,
    "args": {
      "External id": 3042,
      "cbid": 211, "correlation": 3042
    }
  },
  {
    "ph": "s", "id": 3042, "pid": 76337, "tid": -914061504, "ts": 1716454216533714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216548720, "dur": 254,
    "args": {
      "External id": 3050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3050, "pid": 5, "tid": 7, "ts": 1716454216548720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533744, "dur": 8,
    "args": {
      "External id": 3050,
      "cbid": 211, "correlation": 3050
    }
  },
  {
    "ph": "s", "id": 3050, "pid": 76337, "tid": -914061504, "ts": 1716454216533744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216548976, "dur": 54,
    "args": {
      "External id": 3058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3058,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3058, "pid": 5, "tid": 7, "ts": 1716454216548976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533774, "dur": 8,
    "args": {
      "External id": 3058,
      "cbid": 211, "correlation": 3058
    }
  },
  {
    "ph": "s", "id": 3058, "pid": 76337, "tid": -914061504, "ts": 1716454216533774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216549031, "dur": 370,
    "args": {
      "External id": 3068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3068,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3068, "pid": 5, "tid": 7, "ts": 1716454216549031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533843, "dur": 12,
    "args": {
      "External id": 3068,
      "cbid": 211, "correlation": 3068
    }
  },
  {
    "ph": "s", "id": 3068, "pid": 76337, "tid": -914061504, "ts": 1716454216533843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216549403, "dur": 377,
    "args": {
      "External id": 3089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3089,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3089, "pid": 5, "tid": 7, "ts": 1716454216549403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533876, "dur": 8,
    "args": {
      "External id": 3089,
      "cbid": 211, "correlation": 3089
    }
  },
  {
    "ph": "s", "id": 3089, "pid": 76337, "tid": -914061504, "ts": 1716454216533876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216549781, "dur": 4,
    "args": {
      "External id": 3101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3101,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3101, "pid": 5, "tid": 7, "ts": 1716454216549781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533892, "dur": 6,
    "args": {
      "External id": 3101,
      "cbid": 211, "correlation": 3101
    }
  },
  {
    "ph": "s", "id": 3101, "pid": 76337, "tid": -914061504, "ts": 1716454216533892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216549787, "dur": 90,
    "args": {
      "External id": 3104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3104, "pid": 5, "tid": 7, "ts": 1716454216549787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533909, "dur": 6,
    "args": {
      "External id": 3104,
      "cbid": 211, "correlation": 3104
    }
  },
  {
    "ph": "s", "id": 3104, "pid": 76337, "tid": -914061504, "ts": 1716454216533909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216549878, "dur": 57,
    "args": {
      "External id": 3113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3113,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3113, "pid": 5, "tid": 7, "ts": 1716454216549878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216533948, "dur": 9,
    "args": {
      "External id": 3113,
      "cbid": 211, "correlation": 3113
    }
  },
  {
    "ph": "s", "id": 3113, "pid": 76337, "tid": -914061504, "ts": 1716454216533948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534008, "dur": 0,
    "args": {
      "External id": 3123,
      "cbid": 317, "correlation": 3123
    }
  },
  {
    "ph": "f", "id": 3123, "pid": 76337, "tid": -914061504, "ts": 1716454216534008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534009, "dur": 0,
    "args": {
      "External id": 3124,
      "cbid": 203, "correlation": 3124
    }
  },
  {
    "ph": "f", "id": 3124, "pid": 76337, "tid": -914061504, "ts": 1716454216534009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534010, "dur": 0,
    "args": {
      "External id": 3125,
      "cbid": 205, "correlation": 3125
    }
  },
  {
    "ph": "f", "id": 3125, "pid": 76337, "tid": -914061504, "ts": 1716454216534010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216549937, "dur": 103,
    "args": {
      "External id": 3129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3129, "pid": 5, "tid": 7, "ts": 1716454216549937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534023, "dur": 12,
    "args": {
      "External id": 3129,
      "cbid": 211, "correlation": 3129
    }
  },
  {
    "ph": "s", "id": 3129, "pid": 76337, "tid": -914061504, "ts": 1716454216534023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216550042, "dur": 12,
    "args": {
      "External id": 3131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3131,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3131, "pid": 5, "tid": 7, "ts": 1716454216550042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534037, "dur": 5,
    "args": {
      "External id": 3131,
      "cbid": 211, "correlation": 3131
    }
  },
  {
    "ph": "s", "id": 3131, "pid": 76337, "tid": -914061504, "ts": 1716454216534037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216550055, "dur": 4,
    "args": {
      "External id": 3133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 3133, "pid": 5, "tid": 7, "ts": 1716454216550055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534046, "dur": 6,
    "args": {
      "External id": 3133,
      "cbid": 211, "correlation": 3133
    }
  },
  {
    "ph": "s", "id": 3133, "pid": 76337, "tid": -914061504, "ts": 1716454216534046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534055, "dur": 0,
    "args": {
      "External id": 3134,
      "cbid": 51, "correlation": 3134
    }
  },
  {
    "ph": "s", "id": 3134, "pid": 76337, "tid": -914061504, "ts": 1716454216534055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216550060, "dur": 772,
    "args": {
      "External id": 3135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3135,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3135, "pid": 5, "tid": 7, "ts": 1716454216550060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534056, "dur": 5,
    "args": {
      "External id": 3135,
      "cbid": 211, "correlation": 3135
    }
  },
  {
    "ph": "s", "id": 3135, "pid": 76337, "tid": -914061504, "ts": 1716454216534056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216550834, "dur": 92,
    "args": {
      "External id": 3140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3140, "pid": 5, "tid": 7, "ts": 1716454216550834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534084, "dur": 9,
    "args": {
      "External id": 3140,
      "cbid": 211, "correlation": 3140
    }
  },
  {
    "ph": "s", "id": 3140, "pid": 76337, "tid": -914061504, "ts": 1716454216534084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216550927, "dur": 366,
    "args": {
      "External id": 3160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3160,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3160, "pid": 5, "tid": 7, "ts": 1716454216550927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534151, "dur": 12,
    "args": {
      "External id": 3160,
      "cbid": 211, "correlation": 3160
    }
  },
  {
    "ph": "s", "id": 3160, "pid": 76337, "tid": -914061504, "ts": 1716454216534151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216551295, "dur": 4,
    "args": {
      "External id": 3172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3172,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3172, "pid": 5, "tid": 7, "ts": 1716454216551295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534173, "dur": 6,
    "args": {
      "External id": 3172,
      "cbid": 211, "correlation": 3172
    }
  },
  {
    "ph": "s", "id": 3172, "pid": 76337, "tid": -914061504, "ts": 1716454216534173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216551300, "dur": 90,
    "args": {
      "External id": 3175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3175, "pid": 5, "tid": 7, "ts": 1716454216551300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534190, "dur": 6,
    "args": {
      "External id": 3175,
      "cbid": 211, "correlation": 3175
    }
  },
  {
    "ph": "s", "id": 3175, "pid": 76337, "tid": -914061504, "ts": 1716454216534190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216551391, "dur": 58,
    "args": {
      "External id": 3184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3184,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3184, "pid": 5, "tid": 7, "ts": 1716454216551391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534227, "dur": 10,
    "args": {
      "External id": 3184,
      "cbid": 211, "correlation": 3184
    }
  },
  {
    "ph": "s", "id": 3184, "pid": 76337, "tid": -914061504, "ts": 1716454216534227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534290, "dur": 0,
    "args": {
      "External id": 3194,
      "cbid": 317, "correlation": 3194
    }
  },
  {
    "ph": "f", "id": 3194, "pid": 76337, "tid": -914061504, "ts": 1716454216534290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534291, "dur": 0,
    "args": {
      "External id": 3195,
      "cbid": 203, "correlation": 3195
    }
  },
  {
    "ph": "f", "id": 3195, "pid": 76337, "tid": -914061504, "ts": 1716454216534291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534292, "dur": 0,
    "args": {
      "External id": 3196,
      "cbid": 205, "correlation": 3196
    }
  },
  {
    "ph": "f", "id": 3196, "pid": 76337, "tid": -914061504, "ts": 1716454216534292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216551450, "dur": 105,
    "args": {
      "External id": 3200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3200,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3200, "pid": 5, "tid": 7, "ts": 1716454216551450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534305, "dur": 12,
    "args": {
      "External id": 3200,
      "cbid": 211, "correlation": 3200
    }
  },
  {
    "ph": "s", "id": 3200, "pid": 76337, "tid": -914061504, "ts": 1716454216534305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216551557, "dur": 12,
    "args": {
      "External id": 3202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3202, "pid": 5, "tid": 7, "ts": 1716454216551557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534319, "dur": 5,
    "args": {
      "External id": 3202,
      "cbid": 211, "correlation": 3202
    }
  },
  {
    "ph": "s", "id": 3202, "pid": 76337, "tid": -914061504, "ts": 1716454216534319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216551570, "dur": 4,
    "args": {
      "External id": 3204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 3204, "pid": 5, "tid": 7, "ts": 1716454216551570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534327, "dur": 5,
    "args": {
      "External id": 3204,
      "cbid": 211, "correlation": 3204
    }
  },
  {
    "ph": "s", "id": 3204, "pid": 76337, "tid": -914061504, "ts": 1716454216534327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534335, "dur": 0,
    "args": {
      "External id": 3205,
      "cbid": 51, "correlation": 3205
    }
  },
  {
    "ph": "s", "id": 3205, "pid": 76337, "tid": -914061504, "ts": 1716454216534335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216551575, "dur": 770,
    "args": {
      "External id": 3206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3206,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3206, "pid": 5, "tid": 7, "ts": 1716454216551575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534336, "dur": 5,
    "args": {
      "External id": 3206,
      "cbid": 211, "correlation": 3206
    }
  },
  {
    "ph": "s", "id": 3206, "pid": 76337, "tid": -914061504, "ts": 1716454216534336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216552346, "dur": 91,
    "args": {
      "External id": 3211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3211, "pid": 5, "tid": 7, "ts": 1716454216552346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534364, "dur": 8,
    "args": {
      "External id": 3211,
      "cbid": 211, "correlation": 3211
    }
  },
  {
    "ph": "s", "id": 3211, "pid": 76337, "tid": -914061504, "ts": 1716454216534364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216552439, "dur": 254,
    "args": {
      "External id": 3219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3219, "pid": 5, "tid": 7, "ts": 1716454216552439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534397, "dur": 10,
    "args": {
      "External id": 3219,
      "cbid": 211, "correlation": 3219
    }
  },
  {
    "ph": "s", "id": 3219, "pid": 76337, "tid": -914061504, "ts": 1716454216534397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216552695, "dur": 53,
    "args": {
      "External id": 3227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3227,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3227, "pid": 5, "tid": 7, "ts": 1716454216552695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534427, "dur": 8,
    "args": {
      "External id": 3227,
      "cbid": 211, "correlation": 3227
    }
  },
  {
    "ph": "s", "id": 3227, "pid": 76337, "tid": -914061504, "ts": 1716454216534427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216552749, "dur": 46,
    "args": {
      "External id": 3237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 310.0125,
      "warps per SM": 1240.05,
      "grid": [24801, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3237, "pid": 5, "tid": 7, "ts": 1716454216552749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534481, "dur": 11,
    "args": {
      "External id": 3237,
      "cbid": 211, "correlation": 3237
    }
  },
  {
    "ph": "s", "id": 3237, "pid": 76337, "tid": -914061504, "ts": 1716454216534481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216552797, "dur": 85,
    "args": {
      "External id": 3242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3242, "pid": 5, "tid": 7, "ts": 1716454216552797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534510, "dur": 8,
    "args": {
      "External id": 3242,
      "cbid": 211, "correlation": 3242
    }
  },
  {
    "ph": "s", "id": 3242, "pid": 76337, "tid": -914061504, "ts": 1716454216534510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216552883, "dur": 9,
    "args": {
      "External id": 3257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 3257, "pid": 5, "tid": 7, "ts": 1716454216552883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534575, "dur": 13,
    "args": {
      "External id": 3257,
      "cbid": 211, "correlation": 3257
    }
  },
  {
    "ph": "s", "id": 3257, "pid": 76337, "tid": -914061504, "ts": 1716454216534575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534595, "dur": 0,
    "args": {
      "External id": 3264,
      "cbid": 317, "correlation": 3264
    }
  },
  {
    "ph": "f", "id": 3264, "pid": 76337, "tid": -914061504, "ts": 1716454216534595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534596, "dur": 0,
    "args": {
      "External id": 3265,
      "cbid": 203, "correlation": 3265
    }
  },
  {
    "ph": "f", "id": 3265, "pid": 76337, "tid": -914061504, "ts": 1716454216534596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534597, "dur": 0,
    "args": {
      "External id": 3266,
      "cbid": 205, "correlation": 3266
    }
  },
  {
    "ph": "f", "id": 3266, "pid": 76337, "tid": -914061504, "ts": 1716454216534597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216552893, "dur": 314,
    "args": {
      "External id": 3270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3270,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 1.2,
      "warps per SM": 9.6,
      "grid": [1, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3270, "pid": 5, "tid": 7, "ts": 1716454216552893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534615, "dur": 10,
    "args": {
      "External id": 3270,
      "cbid": 211, "correlation": 3270
    }
  },
  {
    "ph": "s", "id": 3270, "pid": 76337, "tid": -914061504, "ts": 1716454216534615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216553209, "dur": 26,
    "args": {
      "External id": 3276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3276, "pid": 5, "tid": 7, "ts": 1716454216553209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534647, "dur": 8,
    "args": {
      "External id": 3276,
      "cbid": 211, "correlation": 3276
    }
  },
  {
    "ph": "s", "id": 3276, "pid": 76337, "tid": -914061504, "ts": 1716454216534647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216553236, "dur": 70,
    "args": {
      "External id": 3286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3286, "pid": 5, "tid": 7, "ts": 1716454216553236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534730, "dur": 12,
    "args": {
      "External id": 3286,
      "cbid": 211, "correlation": 3286
    }
  },
  {
    "ph": "s", "id": 3286, "pid": 76337, "tid": -914061504, "ts": 1716454216534730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216553307, "dur": 95,
    "args": {
      "External id": 3307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3307,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3307, "pid": 5, "tid": 7, "ts": 1716454216553307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534764, "dur": 7,
    "args": {
      "External id": 3307,
      "cbid": 211, "correlation": 3307
    }
  },
  {
    "ph": "s", "id": 3307, "pid": 76337, "tid": -914061504, "ts": 1716454216534764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216553404, "dur": 5,
    "args": {
      "External id": 3319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3319,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3319, "pid": 5, "tid": 7, "ts": 1716454216553404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534781, "dur": 6,
    "args": {
      "External id": 3319,
      "cbid": 211, "correlation": 3319
    }
  },
  {
    "ph": "s", "id": 3319, "pid": 76337, "tid": -914061504, "ts": 1716454216534781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216553409, "dur": 26,
    "args": {
      "External id": 3322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3322,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3322, "pid": 5, "tid": 7, "ts": 1716454216553409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534796, "dur": 6,
    "args": {
      "External id": 3322,
      "cbid": 211, "correlation": 3322
    }
  },
  {
    "ph": "s", "id": 3322, "pid": 76337, "tid": -914061504, "ts": 1716454216534796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216553437, "dur": 18,
    "args": {
      "External id": 3331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3331,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3331, "pid": 5, "tid": 7, "ts": 1716454216553437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534835, "dur": 9,
    "args": {
      "External id": 3331,
      "cbid": 211, "correlation": 3331
    }
  },
  {
    "ph": "s", "id": 3331, "pid": 76337, "tid": -914061504, "ts": 1716454216534835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534887, "dur": 0,
    "args": {
      "External id": 3341,
      "cbid": 317, "correlation": 3341
    }
  },
  {
    "ph": "f", "id": 3341, "pid": 76337, "tid": -914061504, "ts": 1716454216534887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534888, "dur": 0,
    "args": {
      "External id": 3342,
      "cbid": 203, "correlation": 3342
    }
  },
  {
    "ph": "f", "id": 3342, "pid": 76337, "tid": -914061504, "ts": 1716454216534888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534888, "dur": 0,
    "args": {
      "External id": 3343,
      "cbid": 205, "correlation": 3343
    }
  },
  {
    "ph": "f", "id": 3343, "pid": 76337, "tid": -914061504, "ts": 1716454216534888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216553456, "dur": 20,
    "args": {
      "External id": 3347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3347,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 38.4,
      "warps per SM": 307.2,
      "grid": [384, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3347, "pid": 5, "tid": 7, "ts": 1716454216553456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534902, "dur": 11,
    "args": {
      "External id": 3347,
      "cbid": 211, "correlation": 3347
    }
  },
  {
    "ph": "s", "id": 3347, "pid": 76337, "tid": -914061504, "ts": 1716454216534902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216553477, "dur": 19,
    "args": {
      "External id": 3349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 51.2,
      "warps per SM": 409.6,
      "grid": [1, 8, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3349, "pid": 5, "tid": 7, "ts": 1716454216553477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534916, "dur": 5,
    "args": {
      "External id": 3349,
      "cbid": 211, "correlation": 3349
    }
  },
  {
    "ph": "s", "id": 3349, "pid": 76337, "tid": -914061504, "ts": 1716454216534916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216553499, "dur": 1,
    "args": {
      "External id": 3351,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3351,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.6
    }
  },
  {
    "ph": "f", "id": 3351, "pid": 5, "tid": 7, "ts": 1716454216553499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534929, "dur": 15,
    "args": {
      "External id": 3351,
      "cbid": 51, "correlation": 3351
    }
  },
  {
    "ph": "s", "id": 3351, "pid": 76337, "tid": -914061504, "ts": 1716454216534929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216553503, "dur": 465,
    "args": {
      "External id": 3352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3352,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [4, 96, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3352, "pid": 5, "tid": 7, "ts": 1716454216553503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534945, "dur": 8,
    "args": {
      "External id": 3352,
      "cbid": 211, "correlation": 3352
    }
  },
  {
    "ph": "s", "id": 3352, "pid": 76337, "tid": -914061504, "ts": 1716454216534945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216553969, "dur": 38,
    "args": {
      "External id": 3354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3354,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3354, "pid": 5, "tid": 7, "ts": 1716454216553969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534959, "dur": 7,
    "args": {
      "External id": 3354,
      "cbid": 211, "correlation": 3354
    }
  },
  {
    "ph": "s", "id": 3354, "pid": 76337, "tid": -914061504, "ts": 1716454216534959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216554009, "dur": 49,
    "args": {
      "External id": 3360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3360, "pid": 5, "tid": 7, "ts": 1716454216554009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216534999, "dur": 9,
    "args": {
      "External id": 3360,
      "cbid": 211, "correlation": 3360
    }
  },
  {
    "ph": "s", "id": 3360, "pid": 76337, "tid": -914061504, "ts": 1716454216534999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216554059, "dur": 189,
    "args": {
      "External id": 3380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3380,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3380, "pid": 5, "tid": 7, "ts": 1716454216554059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535112, "dur": 13,
    "args": {
      "External id": 3380,
      "cbid": 211, "correlation": 3380
    }
  },
  {
    "ph": "s", "id": 3380, "pid": 76337, "tid": -914061504, "ts": 1716454216535112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216554250, "dur": 4,
    "args": {
      "External id": 3392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3392,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3392, "pid": 5, "tid": 7, "ts": 1716454216554250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535135, "dur": 6,
    "args": {
      "External id": 3392,
      "cbid": 211, "correlation": 3392
    }
  },
  {
    "ph": "s", "id": 3392, "pid": 76337, "tid": -914061504, "ts": 1716454216535135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216554255, "dur": 47,
    "args": {
      "External id": 3395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3395, "pid": 5, "tid": 7, "ts": 1716454216554255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535153, "dur": 6,
    "args": {
      "External id": 3395,
      "cbid": 211, "correlation": 3395
    }
  },
  {
    "ph": "s", "id": 3395, "pid": 76337, "tid": -914061504, "ts": 1716454216535153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216554304, "dur": 31,
    "args": {
      "External id": 3404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3404,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3404, "pid": 5, "tid": 7, "ts": 1716454216554304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535194, "dur": 10,
    "args": {
      "External id": 3404,
      "cbid": 211, "correlation": 3404
    }
  },
  {
    "ph": "s", "id": 3404, "pid": 76337, "tid": -914061504, "ts": 1716454216535194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535259, "dur": 0,
    "args": {
      "External id": 3414,
      "cbid": 317, "correlation": 3414
    }
  },
  {
    "ph": "f", "id": 3414, "pid": 76337, "tid": -914061504, "ts": 1716454216535259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535260, "dur": 0,
    "args": {
      "External id": 3415,
      "cbid": 203, "correlation": 3415
    }
  },
  {
    "ph": "f", "id": 3415, "pid": 76337, "tid": -914061504, "ts": 1716454216535260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535261, "dur": 0,
    "args": {
      "External id": 3416,
      "cbid": 205, "correlation": 3416
    }
  },
  {
    "ph": "f", "id": 3416, "pid": 76337, "tid": -914061504, "ts": 1716454216535261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216554336, "dur": 33,
    "args": {
      "External id": 3420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3420, "pid": 5, "tid": 7, "ts": 1716454216554336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535278, "dur": 12,
    "args": {
      "External id": 3420,
      "cbid": 211, "correlation": 3420
    }
  },
  {
    "ph": "s", "id": 3420, "pid": 76337, "tid": -914061504, "ts": 1716454216535278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216554371, "dur": 32,
    "args": {
      "External id": 3422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3422, "pid": 5, "tid": 7, "ts": 1716454216554371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535292, "dur": 5,
    "args": {
      "External id": 3422,
      "cbid": 211, "correlation": 3422
    }
  },
  {
    "ph": "s", "id": 3422, "pid": 76337, "tid": -914061504, "ts": 1716454216535292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216554406, "dur": 2,
    "args": {
      "External id": 3424,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3424,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.4545454545454546
    }
  },
  {
    "ph": "f", "id": 3424, "pid": 5, "tid": 7, "ts": 1716454216554406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535303, "dur": 7,
    "args": {
      "External id": 3424,
      "cbid": 51, "correlation": 3424
    }
  },
  {
    "ph": "s", "id": 3424, "pid": 76337, "tid": -914061504, "ts": 1716454216535303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216554410, "dur": 891,
    "args": {
      "External id": 3425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3425,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3425, "pid": 5, "tid": 7, "ts": 1716454216554410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535312, "dur": 6,
    "args": {
      "External id": 3425,
      "cbid": 211, "correlation": 3425
    }
  },
  {
    "ph": "s", "id": 3425, "pid": 76337, "tid": -914061504, "ts": 1716454216535312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216555302, "dur": 37,
    "args": {
      "External id": 3427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3427,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3427, "pid": 5, "tid": 7, "ts": 1716454216555302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535322, "dur": 5,
    "args": {
      "External id": 3427,
      "cbid": 211, "correlation": 3427
    }
  },
  {
    "ph": "s", "id": 3427, "pid": 76337, "tid": -914061504, "ts": 1716454216535322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216555340, "dur": 49,
    "args": {
      "External id": 3433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3433,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3433, "pid": 5, "tid": 7, "ts": 1716454216555340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535351, "dur": 8,
    "args": {
      "External id": 3433,
      "cbid": 211, "correlation": 3433
    }
  },
  {
    "ph": "s", "id": 3433, "pid": 76337, "tid": -914061504, "ts": 1716454216535351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535410, "dur": 0,
    "args": {
      "External id": 3443,
      "cbid": 317, "correlation": 3443
    }
  },
  {
    "ph": "f", "id": 3443, "pid": 76337, "tid": -914061504, "ts": 1716454216535410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535411, "dur": 0,
    "args": {
      "External id": 3444,
      "cbid": 203, "correlation": 3444
    }
  },
  {
    "ph": "f", "id": 3444, "pid": 76337, "tid": -914061504, "ts": 1716454216535411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535412, "dur": 0,
    "args": {
      "External id": 3445,
      "cbid": 205, "correlation": 3445
    }
  },
  {
    "ph": "f", "id": 3445, "pid": 76337, "tid": -914061504, "ts": 1716454216535412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535442, "dur": 1,
    "args": {
      "External id": 3449,
      "cbid": 251, "correlation": 3449
    }
  },
  {
    "ph": "f", "id": 3449, "pid": 76337, "tid": -914061504, "ts": 1716454216535442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216555391, "dur": 68,
    "args": {
      "External id": 3450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3450,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [4, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3450, "pid": 5, "tid": 7, "ts": 1716454216555391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535447, "dur": 13,
    "args": {
      "External id": 3450,
      "cbid": 211, "correlation": 3450
    }
  },
  {
    "ph": "s", "id": 3450, "pid": 76337, "tid": -914061504, "ts": 1716454216535447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216555460, "dur": 49,
    "args": {
      "External id": 3456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3456, "pid": 5, "tid": 7, "ts": 1716454216555460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535480, "dur": 9,
    "args": {
      "External id": 3456,
      "cbid": 211, "correlation": 3456
    }
  },
  {
    "ph": "s", "id": 3456, "pid": 76337, "tid": -914061504, "ts": 1716454216535480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216555510, "dur": 139,
    "args": {
      "External id": 3464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3464, "pid": 5, "tid": 7, "ts": 1716454216555510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535512, "dur": 8,
    "args": {
      "External id": 3464,
      "cbid": 211, "correlation": 3464
    }
  },
  {
    "ph": "s", "id": 3464, "pid": 76337, "tid": -914061504, "ts": 1716454216535512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216555650, "dur": 29,
    "args": {
      "External id": 3472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3472,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3472, "pid": 5, "tid": 7, "ts": 1716454216555650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535543, "dur": 8,
    "args": {
      "External id": 3472,
      "cbid": 211, "correlation": 3472
    }
  },
  {
    "ph": "s", "id": 3472, "pid": 76337, "tid": -914061504, "ts": 1716454216535543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216555680, "dur": 135,
    "args": {
      "External id": 3482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3482,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3482, "pid": 5, "tid": 7, "ts": 1716454216555680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535614, "dur": 12,
    "args": {
      "External id": 3482,
      "cbid": 211, "correlation": 3482
    }
  },
  {
    "ph": "s", "id": 3482, "pid": 76337, "tid": -914061504, "ts": 1716454216535614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216555817, "dur": 192,
    "args": {
      "External id": 3503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3503,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3503, "pid": 5, "tid": 7, "ts": 1716454216555817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535648, "dur": 8,
    "args": {
      "External id": 3503,
      "cbid": 211, "correlation": 3503
    }
  },
  {
    "ph": "s", "id": 3503, "pid": 76337, "tid": -914061504, "ts": 1716454216535648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216556011, "dur": 4,
    "args": {
      "External id": 3515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3515,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3515, "pid": 5, "tid": 7, "ts": 1716454216556011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535664, "dur": 6,
    "args": {
      "External id": 3515,
      "cbid": 211, "correlation": 3515
    }
  },
  {
    "ph": "s", "id": 3515, "pid": 76337, "tid": -914061504, "ts": 1716454216535664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216556016, "dur": 48,
    "args": {
      "External id": 3518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3518,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3518, "pid": 5, "tid": 7, "ts": 1716454216556016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535681, "dur": 6,
    "args": {
      "External id": 3518,
      "cbid": 211, "correlation": 3518
    }
  },
  {
    "ph": "s", "id": 3518, "pid": 76337, "tid": -914061504, "ts": 1716454216535681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216556066, "dur": 31,
    "args": {
      "External id": 3527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3527,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3527, "pid": 5, "tid": 7, "ts": 1716454216556066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535720, "dur": 9,
    "args": {
      "External id": 3527,
      "cbid": 211, "correlation": 3527
    }
  },
  {
    "ph": "s", "id": 3527, "pid": 76337, "tid": -914061504, "ts": 1716454216535720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535772, "dur": 0,
    "args": {
      "External id": 3537,
      "cbid": 317, "correlation": 3537
    }
  },
  {
    "ph": "f", "id": 3537, "pid": 76337, "tid": -914061504, "ts": 1716454216535772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535773, "dur": 0,
    "args": {
      "External id": 3538,
      "cbid": 203, "correlation": 3538
    }
  },
  {
    "ph": "f", "id": 3538, "pid": 76337, "tid": -914061504, "ts": 1716454216535773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535774, "dur": 0,
    "args": {
      "External id": 3539,
      "cbid": 205, "correlation": 3539
    }
  },
  {
    "ph": "f", "id": 3539, "pid": 76337, "tid": -914061504, "ts": 1716454216535774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216556098, "dur": 34,
    "args": {
      "External id": 3543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3543,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3543, "pid": 5, "tid": 7, "ts": 1716454216556098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535787, "dur": 11,
    "args": {
      "External id": 3543,
      "cbid": 211, "correlation": 3543
    }
  },
  {
    "ph": "s", "id": 3543, "pid": 76337, "tid": -914061504, "ts": 1716454216535787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216556134, "dur": 33,
    "args": {
      "External id": 3545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3545,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3545, "pid": 5, "tid": 7, "ts": 1716454216556134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535801, "dur": 6,
    "args": {
      "External id": 3545,
      "cbid": 211, "correlation": 3545
    }
  },
  {
    "ph": "s", "id": 3545, "pid": 76337, "tid": -914061504, "ts": 1716454216535801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216556169, "dur": 2,
    "args": {
      "External id": 3547,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3547,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.476923076923077
    }
  },
  {
    "ph": "f", "id": 3547, "pid": 5, "tid": 7, "ts": 1716454216556169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535812, "dur": 6,
    "args": {
      "External id": 3547,
      "cbid": 51, "correlation": 3547
    }
  },
  {
    "ph": "s", "id": 3547, "pid": 76337, "tid": -914061504, "ts": 1716454216535812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216556173, "dur": 897,
    "args": {
      "External id": 3548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3548,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3548, "pid": 5, "tid": 7, "ts": 1716454216556173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535820, "dur": 6,
    "args": {
      "External id": 3548,
      "cbid": 211, "correlation": 3548
    }
  },
  {
    "ph": "s", "id": 3548, "pid": 76337, "tid": -914061504, "ts": 1716454216535820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216557071, "dur": 37,
    "args": {
      "External id": 3550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3550,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3550, "pid": 5, "tid": 7, "ts": 1716454216557071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535830, "dur": 5,
    "args": {
      "External id": 3550,
      "cbid": 211, "correlation": 3550
    }
  },
  {
    "ph": "s", "id": 3550, "pid": 76337, "tid": -914061504, "ts": 1716454216535830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216557110, "dur": 50,
    "args": {
      "External id": 3556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3556, "pid": 5, "tid": 7, "ts": 1716454216557110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535858, "dur": 9,
    "args": {
      "External id": 3556,
      "cbid": 211, "correlation": 3556
    }
  },
  {
    "ph": "s", "id": 3556, "pid": 76337, "tid": -914061504, "ts": 1716454216535858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216557161, "dur": 191,
    "args": {
      "External id": 3576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3576,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3576, "pid": 5, "tid": 7, "ts": 1716454216557161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535926, "dur": 11,
    "args": {
      "External id": 3576,
      "cbid": 211, "correlation": 3576
    }
  },
  {
    "ph": "s", "id": 3576, "pid": 76337, "tid": -914061504, "ts": 1716454216535926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216557353, "dur": 4,
    "args": {
      "External id": 3588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3588,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3588, "pid": 5, "tid": 7, "ts": 1716454216557353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535947, "dur": 6,
    "args": {
      "External id": 3588,
      "cbid": 211, "correlation": 3588
    }
  },
  {
    "ph": "s", "id": 3588, "pid": 76337, "tid": -914061504, "ts": 1716454216535947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216557359, "dur": 47,
    "args": {
      "External id": 3591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3591,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3591, "pid": 5, "tid": 7, "ts": 1716454216557359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216535964, "dur": 7,
    "args": {
      "External id": 3591,
      "cbid": 211, "correlation": 3591
    }
  },
  {
    "ph": "s", "id": 3591, "pid": 76337, "tid": -914061504, "ts": 1716454216535964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216557407, "dur": 31,
    "args": {
      "External id": 3600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3600,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3600, "pid": 5, "tid": 7, "ts": 1716454216557407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536013, "dur": 11,
    "args": {
      "External id": 3600,
      "cbid": 211, "correlation": 3600
    }
  },
  {
    "ph": "s", "id": 3600, "pid": 76337, "tid": -914061504, "ts": 1716454216536013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536078, "dur": 0,
    "args": {
      "External id": 3610,
      "cbid": 317, "correlation": 3610
    }
  },
  {
    "ph": "f", "id": 3610, "pid": 76337, "tid": -914061504, "ts": 1716454216536078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536078, "dur": 0,
    "args": {
      "External id": 3611,
      "cbid": 203, "correlation": 3611
    }
  },
  {
    "ph": "f", "id": 3611, "pid": 76337, "tid": -914061504, "ts": 1716454216536078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536079, "dur": 0,
    "args": {
      "External id": 3612,
      "cbid": 205, "correlation": 3612
    }
  },
  {
    "ph": "f", "id": 3612, "pid": 76337, "tid": -914061504, "ts": 1716454216536079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216557439, "dur": 33,
    "args": {
      "External id": 3616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3616,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3616, "pid": 5, "tid": 7, "ts": 1716454216557439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536092, "dur": 11,
    "args": {
      "External id": 3616,
      "cbid": 211, "correlation": 3616
    }
  },
  {
    "ph": "s", "id": 3616, "pid": 76337, "tid": -914061504, "ts": 1716454216536092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216557473, "dur": 32,
    "args": {
      "External id": 3618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3618,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3618, "pid": 5, "tid": 7, "ts": 1716454216557473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536106, "dur": 5,
    "args": {
      "External id": 3618,
      "cbid": 211, "correlation": 3618
    }
  },
  {
    "ph": "s", "id": 3618, "pid": 76337, "tid": -914061504, "ts": 1716454216536106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216557508, "dur": 1,
    "args": {
      "External id": 3620,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3620,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.5737704918032787
    }
  },
  {
    "ph": "f", "id": 3620, "pid": 5, "tid": 7, "ts": 1716454216557508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536116, "dur": 7,
    "args": {
      "External id": 3620,
      "cbid": 51, "correlation": 3620
    }
  },
  {
    "ph": "s", "id": 3620, "pid": 76337, "tid": -914061504, "ts": 1716454216536116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216557512, "dur": 889,
    "args": {
      "External id": 3621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3621,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3621, "pid": 5, "tid": 7, "ts": 1716454216557512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536124, "dur": 6,
    "args": {
      "External id": 3621,
      "cbid": 211, "correlation": 3621
    }
  },
  {
    "ph": "s", "id": 3621, "pid": 76337, "tid": -914061504, "ts": 1716454216536124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216558402, "dur": 37,
    "args": {
      "External id": 3623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3623,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3623, "pid": 5, "tid": 7, "ts": 1716454216558402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536134, "dur": 5,
    "args": {
      "External id": 3623,
      "cbid": 211, "correlation": 3623
    }
  },
  {
    "ph": "s", "id": 3623, "pid": 76337, "tid": -914061504, "ts": 1716454216536134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216558441, "dur": 49,
    "args": {
      "External id": 3629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3629, "pid": 5, "tid": 7, "ts": 1716454216558441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536161, "dur": 8,
    "args": {
      "External id": 3629,
      "cbid": 211, "correlation": 3629
    }
  },
  {
    "ph": "s", "id": 3629, "pid": 76337, "tid": -914061504, "ts": 1716454216536161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216558491, "dur": 140,
    "args": {
      "External id": 3637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3637, "pid": 5, "tid": 7, "ts": 1716454216558491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536194, "dur": 9,
    "args": {
      "External id": 3637,
      "cbid": 211, "correlation": 3637
    }
  },
  {
    "ph": "s", "id": 3637, "pid": 76337, "tid": -914061504, "ts": 1716454216536194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216558632, "dur": 28,
    "args": {
      "External id": 3645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3645,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3645, "pid": 5, "tid": 7, "ts": 1716454216558632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536224, "dur": 8,
    "args": {
      "External id": 3645,
      "cbid": 211, "correlation": 3645
    }
  },
  {
    "ph": "s", "id": 3645, "pid": 76337, "tid": -914061504, "ts": 1716454216536224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216558662, "dur": 25,
    "args": {
      "External id": 3655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 156.4125,
      "warps per SM": 625.65,
      "grid": [12513, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3655, "pid": 5, "tid": 7, "ts": 1716454216558662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536279, "dur": 10,
    "args": {
      "External id": 3655,
      "cbid": 211, "correlation": 3655
    }
  },
  {
    "ph": "s", "id": 3655, "pid": 76337, "tid": -914061504, "ts": 1716454216536279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216558688, "dur": 45,
    "args": {
      "External id": 3660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3660, "pid": 5, "tid": 7, "ts": 1716454216558688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536308, "dur": 7,
    "args": {
      "External id": 3660,
      "cbid": 211, "correlation": 3660
    }
  },
  {
    "ph": "s", "id": 3660, "pid": 76337, "tid": -914061504, "ts": 1716454216536308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216558735, "dur": 25,
    "args": {
      "External id": 3675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 57.6,
      "warps per SM": 230.4,
      "grid": [4608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3675, "pid": 5, "tid": 7, "ts": 1716454216558735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536371, "dur": 12,
    "args": {
      "External id": 3675,
      "cbid": 211, "correlation": 3675
    }
  },
  {
    "ph": "s", "id": 3675, "pid": 76337, "tid": -914061504, "ts": 1716454216536371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536391, "dur": 0,
    "args": {
      "External id": 3682,
      "cbid": 317, "correlation": 3682
    }
  },
  {
    "ph": "f", "id": 3682, "pid": 76337, "tid": -914061504, "ts": 1716454216536391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536392, "dur": 0,
    "args": {
      "External id": 3683,
      "cbid": 203, "correlation": 3683
    }
  },
  {
    "ph": "f", "id": 3683, "pid": 76337, "tid": -914061504, "ts": 1716454216536392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536393, "dur": 0,
    "args": {
      "External id": 3684,
      "cbid": 205, "correlation": 3684
    }
  },
  {
    "ph": "f", "id": 3684, "pid": 76337, "tid": -914061504, "ts": 1716454216536393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216558761, "dur": 310,
    "args": {
      "External id": 3688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3688,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [4, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 3688, "pid": 5, "tid": 7, "ts": 1716454216558761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536409, "dur": 8,
    "args": {
      "External id": 3688,
      "cbid": 211, "correlation": 3688
    }
  },
  {
    "ph": "s", "id": 3688, "pid": 76337, "tid": -914061504, "ts": 1716454216536409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559072, "dur": 14,
    "args": {
      "External id": 3694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3694, "pid": 5, "tid": 7, "ts": 1716454216559072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536440, "dur": 8,
    "args": {
      "External id": 3694,
      "cbid": 211, "correlation": 3694
    }
  },
  {
    "ph": "s", "id": 3694, "pid": 76337, "tid": -914061504, "ts": 1716454216536440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559088, "dur": 37,
    "args": {
      "External id": 3704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3704,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3704, "pid": 5, "tid": 7, "ts": 1716454216559088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536522, "dur": 13,
    "args": {
      "External id": 3704,
      "cbid": 211, "correlation": 3704
    }
  },
  {
    "ph": "s", "id": 3704, "pid": 76337, "tid": -914061504, "ts": 1716454216536522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559127, "dur": 47,
    "args": {
      "External id": 3725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3725,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3725, "pid": 5, "tid": 7, "ts": 1716454216559127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536556, "dur": 7,
    "args": {
      "External id": 3725,
      "cbid": 211, "correlation": 3725
    }
  },
  {
    "ph": "s", "id": 3725, "pid": 76337, "tid": -914061504, "ts": 1716454216536556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216559176, "dur": 4,
    "args": {
      "External id": 3737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3737,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3737, "pid": 5, "tid": 7, "ts": 1716454216559176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536572, "dur": 6,
    "args": {
      "External id": 3737,
      "cbid": 211, "correlation": 3737
    }
  },
  {
    "ph": "s", "id": 3737, "pid": 76337, "tid": -914061504, "ts": 1716454216536572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559181, "dur": 16,
    "args": {
      "External id": 3740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3740, "pid": 5, "tid": 7, "ts": 1716454216559181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536590, "dur": 6,
    "args": {
      "External id": 3740,
      "cbid": 211, "correlation": 3740
    }
  },
  {
    "ph": "s", "id": 3740, "pid": 76337, "tid": -914061504, "ts": 1716454216536590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216559198, "dur": 10,
    "args": {
      "External id": 3749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3749,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3749, "pid": 5, "tid": 7, "ts": 1716454216559198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536628, "dur": 10,
    "args": {
      "External id": 3749,
      "cbid": 211, "correlation": 3749
    }
  },
  {
    "ph": "s", "id": 3749, "pid": 76337, "tid": -914061504, "ts": 1716454216536628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536680, "dur": 0,
    "args": {
      "External id": 3759,
      "cbid": 317, "correlation": 3759
    }
  },
  {
    "ph": "f", "id": 3759, "pid": 76337, "tid": -914061504, "ts": 1716454216536680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536681, "dur": 0,
    "args": {
      "External id": 3760,
      "cbid": 203, "correlation": 3760
    }
  },
  {
    "ph": "f", "id": 3760, "pid": 76337, "tid": -914061504, "ts": 1716454216536681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536681, "dur": 0,
    "args": {
      "External id": 3761,
      "cbid": 205, "correlation": 3761
    }
  },
  {
    "ph": "f", "id": 3761, "pid": 76337, "tid": -914061504, "ts": 1716454216536681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559209, "dur": 11,
    "args": {
      "External id": 3765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3765,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3765, "pid": 5, "tid": 7, "ts": 1716454216559209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536696, "dur": 11,
    "args": {
      "External id": 3765,
      "cbid": 211, "correlation": 3765
    }
  },
  {
    "ph": "s", "id": 3765, "pid": 76337, "tid": -914061504, "ts": 1716454216536696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559222, "dur": 32,
    "args": {
      "External id": 3767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3767,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3767, "pid": 5, "tid": 7, "ts": 1716454216559222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536710, "dur": 5,
    "args": {
      "External id": 3767,
      "cbid": 211, "correlation": 3767
    }
  },
  {
    "ph": "s", "id": 3767, "pid": 76337, "tid": -914061504, "ts": 1716454216536710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216559256, "dur": 1,
    "args": {
      "External id": 3769,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3769,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 3769, "pid": 5, "tid": 7, "ts": 1716454216559256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536721, "dur": 7,
    "args": {
      "External id": 3769,
      "cbid": 51, "correlation": 3769
    }
  },
  {
    "ph": "s", "id": 3769, "pid": 76337, "tid": -914061504, "ts": 1716454216536721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216559260, "dur": 273,
    "args": {
      "External id": 3770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3770,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3770, "pid": 5, "tid": 7, "ts": 1716454216559260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536729, "dur": 6,
    "args": {
      "External id": 3770,
      "cbid": 211, "correlation": 3770
    }
  },
  {
    "ph": "s", "id": 3770, "pid": 76337, "tid": -914061504, "ts": 1716454216536729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559535, "dur": 11,
    "args": {
      "External id": 3772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3772, "pid": 5, "tid": 7, "ts": 1716454216559535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536739, "dur": 6,
    "args": {
      "External id": 3772,
      "cbid": 211, "correlation": 3772
    }
  },
  {
    "ph": "s", "id": 3772, "pid": 76337, "tid": -914061504, "ts": 1716454216536739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559547, "dur": 12,
    "args": {
      "External id": 3778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3778, "pid": 5, "tid": 7, "ts": 1716454216559547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536767, "dur": 8,
    "args": {
      "External id": 3778,
      "cbid": 211, "correlation": 3778
    }
  },
  {
    "ph": "s", "id": 3778, "pid": 76337, "tid": -914061504, "ts": 1716454216536767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559561, "dur": 30,
    "args": {
      "External id": 3798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3798,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3798, "pid": 5, "tid": 7, "ts": 1716454216559561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536835, "dur": 12,
    "args": {
      "External id": 3798,
      "cbid": 211, "correlation": 3798
    }
  },
  {
    "ph": "s", "id": 3798, "pid": 76337, "tid": -914061504, "ts": 1716454216536835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216559592, "dur": 4,
    "args": {
      "External id": 3810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3810,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3810, "pid": 5, "tid": 7, "ts": 1716454216559592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536857, "dur": 6,
    "args": {
      "External id": 3810,
      "cbid": 211, "correlation": 3810
    }
  },
  {
    "ph": "s", "id": 3810, "pid": 76337, "tid": -914061504, "ts": 1716454216536857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559598, "dur": 13,
    "args": {
      "External id": 3813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3813, "pid": 5, "tid": 7, "ts": 1716454216559598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536874, "dur": 6,
    "args": {
      "External id": 3813,
      "cbid": 211, "correlation": 3813
    }
  },
  {
    "ph": "s", "id": 3813, "pid": 76337, "tid": -914061504, "ts": 1716454216536874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216559612, "dur": 9,
    "args": {
      "External id": 3822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3822,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3822, "pid": 5, "tid": 7, "ts": 1716454216559612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536913, "dur": 10,
    "args": {
      "External id": 3822,
      "cbid": 211, "correlation": 3822
    }
  },
  {
    "ph": "s", "id": 3822, "pid": 76337, "tid": -914061504, "ts": 1716454216536913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536983, "dur": 0,
    "args": {
      "External id": 3832,
      "cbid": 317, "correlation": 3832
    }
  },
  {
    "ph": "f", "id": 3832, "pid": 76337, "tid": -914061504, "ts": 1716454216536983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536984, "dur": 0,
    "args": {
      "External id": 3833,
      "cbid": 203, "correlation": 3833
    }
  },
  {
    "ph": "f", "id": 3833, "pid": 76337, "tid": -914061504, "ts": 1716454216536984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536984, "dur": 0,
    "args": {
      "External id": 3834,
      "cbid": 205, "correlation": 3834
    }
  },
  {
    "ph": "f", "id": 3834, "pid": 76337, "tid": -914061504, "ts": 1716454216536984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559623, "dur": 9,
    "args": {
      "External id": 3838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3838, "pid": 5, "tid": 7, "ts": 1716454216559623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216536998, "dur": 13,
    "args": {
      "External id": 3838,
      "cbid": 211, "correlation": 3838
    }
  },
  {
    "ph": "s", "id": 3838, "pid": 76337, "tid": -914061504, "ts": 1716454216536998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559633, "dur": 32,
    "args": {
      "External id": 3840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3840,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3840, "pid": 5, "tid": 7, "ts": 1716454216559633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537013, "dur": 5,
    "args": {
      "External id": 3840,
      "cbid": 211, "correlation": 3840
    }
  },
  {
    "ph": "s", "id": 3840, "pid": 76337, "tid": -914061504, "ts": 1716454216537013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216559667, "dur": 1,
    "args": {
      "External id": 3842,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3842,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 3842, "pid": 5, "tid": 7, "ts": 1716454216559667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537023, "dur": 6,
    "args": {
      "External id": 3842,
      "cbid": 51, "correlation": 3842
    }
  },
  {
    "ph": "s", "id": 3842, "pid": 76337, "tid": -914061504, "ts": 1716454216537023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216559671, "dur": 265,
    "args": {
      "External id": 3843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3843,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3843, "pid": 5, "tid": 7, "ts": 1716454216559671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537031, "dur": 6,
    "args": {
      "External id": 3843,
      "cbid": 211, "correlation": 3843
    }
  },
  {
    "ph": "s", "id": 3843, "pid": 76337, "tid": -914061504, "ts": 1716454216537031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216559938, "dur": 10,
    "args": {
      "External id": 3845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3845,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3845, "pid": 5, "tid": 7, "ts": 1716454216559938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537040, "dur": 5,
    "args": {
      "External id": 3845,
      "cbid": 211, "correlation": 3845
    }
  },
  {
    "ph": "s", "id": 3845, "pid": 76337, "tid": -914061504, "ts": 1716454216537040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559949, "dur": 12,
    "args": {
      "External id": 3851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3851,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3851, "pid": 5, "tid": 7, "ts": 1716454216559949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537068, "dur": 8,
    "args": {
      "External id": 3851,
      "cbid": 211, "correlation": 3851
    }
  },
  {
    "ph": "s", "id": 3851, "pid": 76337, "tid": -914061504, "ts": 1716454216537068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216559962, "dur": 39,
    "args": {
      "External id": 3859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3859,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3859, "pid": 5, "tid": 7, "ts": 1716454216559962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537102, "dur": 10,
    "args": {
      "External id": 3859,
      "cbid": 211, "correlation": 3859
    }
  },
  {
    "ph": "s", "id": 3859, "pid": 76337, "tid": -914061504, "ts": 1716454216537102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216560003, "dur": 10,
    "args": {
      "External id": 3867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3867,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3867, "pid": 5, "tid": 7, "ts": 1716454216560003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537132, "dur": 8,
    "args": {
      "External id": 3867,
      "cbid": 211, "correlation": 3867
    }
  },
  {
    "ph": "s", "id": 3867, "pid": 76337, "tid": -914061504, "ts": 1716454216537132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560014, "dur": 38,
    "args": {
      "External id": 3877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3877,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3877, "pid": 5, "tid": 7, "ts": 1716454216560014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537200, "dur": 12,
    "args": {
      "External id": 3877,
      "cbid": 211, "correlation": 3877
    }
  },
  {
    "ph": "s", "id": 3877, "pid": 76337, "tid": -914061504, "ts": 1716454216537200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560053, "dur": 48,
    "args": {
      "External id": 3898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3898,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3898, "pid": 5, "tid": 7, "ts": 1716454216560053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537232, "dur": 7,
    "args": {
      "External id": 3898,
      "cbid": 211, "correlation": 3898
    }
  },
  {
    "ph": "s", "id": 3898, "pid": 76337, "tid": -914061504, "ts": 1716454216537232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216560102, "dur": 4,
    "args": {
      "External id": 3910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3910,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3910, "pid": 5, "tid": 7, "ts": 1716454216560102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537248, "dur": 6,
    "args": {
      "External id": 3910,
      "cbid": 211, "correlation": 3910
    }
  },
  {
    "ph": "s", "id": 3910, "pid": 76337, "tid": -914061504, "ts": 1716454216537248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560108, "dur": 14,
    "args": {
      "External id": 3913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3913,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3913, "pid": 5, "tid": 7, "ts": 1716454216560108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537264, "dur": 7,
    "args": {
      "External id": 3913,
      "cbid": 211, "correlation": 3913
    }
  },
  {
    "ph": "s", "id": 3913, "pid": 76337, "tid": -914061504, "ts": 1716454216537264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216560124, "dur": 9,
    "args": {
      "External id": 3922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3922,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3922, "pid": 5, "tid": 7, "ts": 1716454216560124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537303, "dur": 10,
    "args": {
      "External id": 3922,
      "cbid": 211, "correlation": 3922
    }
  },
  {
    "ph": "s", "id": 3922, "pid": 76337, "tid": -914061504, "ts": 1716454216537303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537354, "dur": 0,
    "args": {
      "External id": 3932,
      "cbid": 317, "correlation": 3932
    }
  },
  {
    "ph": "f", "id": 3932, "pid": 76337, "tid": -914061504, "ts": 1716454216537354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537355, "dur": 0,
    "args": {
      "External id": 3933,
      "cbid": 203, "correlation": 3933
    }
  },
  {
    "ph": "f", "id": 3933, "pid": 76337, "tid": -914061504, "ts": 1716454216537355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537356, "dur": 0,
    "args": {
      "External id": 3934,
      "cbid": 205, "correlation": 3934
    }
  },
  {
    "ph": "f", "id": 3934, "pid": 76337, "tid": -914061504, "ts": 1716454216537356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560135, "dur": 9,
    "args": {
      "External id": 3938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3938,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3938, "pid": 5, "tid": 7, "ts": 1716454216560135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537370, "dur": 12,
    "args": {
      "External id": 3938,
      "cbid": 211, "correlation": 3938
    }
  },
  {
    "ph": "s", "id": 3938, "pid": 76337, "tid": -914061504, "ts": 1716454216537370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560145, "dur": 32,
    "args": {
      "External id": 3940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3940,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3940, "pid": 5, "tid": 7, "ts": 1716454216560145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537384, "dur": 5,
    "args": {
      "External id": 3940,
      "cbid": 211, "correlation": 3940
    }
  },
  {
    "ph": "s", "id": 3940, "pid": 76337, "tid": -914061504, "ts": 1716454216537384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216560179, "dur": 1,
    "args": {
      "External id": 3942,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 3942,
      "bytes": 768, "memory bandwidth (GB/s)": 0.42105263157894735
    }
  },
  {
    "ph": "f", "id": 3942, "pid": 5, "tid": 7, "ts": 1716454216560179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537394, "dur": 6,
    "args": {
      "External id": 3942,
      "cbid": 51, "correlation": 3942
    }
  },
  {
    "ph": "s", "id": 3942, "pid": 76337, "tid": -914061504, "ts": 1716454216537394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216560183, "dur": 266,
    "args": {
      "External id": 3943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3943,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 3943, "pid": 5, "tid": 7, "ts": 1716454216560183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537401, "dur": 6,
    "args": {
      "External id": 3943,
      "cbid": 211, "correlation": 3943
    }
  },
  {
    "ph": "s", "id": 3943, "pid": 76337, "tid": -914061504, "ts": 1716454216537401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560451, "dur": 10,
    "args": {
      "External id": 3945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3945,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3945, "pid": 5, "tid": 7, "ts": 1716454216560451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537411, "dur": 5,
    "args": {
      "External id": 3945,
      "cbid": 211, "correlation": 3945
    }
  },
  {
    "ph": "s", "id": 3945, "pid": 76337, "tid": -914061504, "ts": 1716454216537411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560462, "dur": 12,
    "args": {
      "External id": 3951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3951,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3951, "pid": 5, "tid": 7, "ts": 1716454216560462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537440, "dur": 9,
    "args": {
      "External id": 3951,
      "cbid": 211, "correlation": 3951
    }
  },
  {
    "ph": "s", "id": 3951, "pid": 76337, "tid": -914061504, "ts": 1716454216537440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560476, "dur": 30,
    "args": {
      "External id": 3971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3971,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 3971, "pid": 5, "tid": 7, "ts": 1716454216560476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537508, "dur": 11,
    "args": {
      "External id": 3971,
      "cbid": 211, "correlation": 3971
    }
  },
  {
    "ph": "s", "id": 3971, "pid": 76337, "tid": -914061504, "ts": 1716454216537508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216560507, "dur": 4,
    "args": {
      "External id": 3983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3983,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 3983, "pid": 5, "tid": 7, "ts": 1716454216560507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537530, "dur": 6,
    "args": {
      "External id": 3983,
      "cbid": 211, "correlation": 3983
    }
  },
  {
    "ph": "s", "id": 3983, "pid": 76337, "tid": -914061504, "ts": 1716454216537530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560513, "dur": 14,
    "args": {
      "External id": 3986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3986,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3986, "pid": 5, "tid": 7, "ts": 1716454216560513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537548, "dur": 6,
    "args": {
      "External id": 3986,
      "cbid": 211, "correlation": 3986
    }
  },
  {
    "ph": "s", "id": 3986, "pid": 76337, "tid": -914061504, "ts": 1716454216537548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216560528, "dur": 9,
    "args": {
      "External id": 3995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 3995,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 3995, "pid": 5, "tid": 7, "ts": 1716454216560528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537587, "dur": 10,
    "args": {
      "External id": 3995,
      "cbid": 211, "correlation": 3995
    }
  },
  {
    "ph": "s", "id": 3995, "pid": 76337, "tid": -914061504, "ts": 1716454216537587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537647, "dur": 0,
    "args": {
      "External id": 4005,
      "cbid": 317, "correlation": 4005
    }
  },
  {
    "ph": "f", "id": 4005, "pid": 76337, "tid": -914061504, "ts": 1716454216537647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537648, "dur": 0,
    "args": {
      "External id": 4006,
      "cbid": 203, "correlation": 4006
    }
  },
  {
    "ph": "f", "id": 4006, "pid": 76337, "tid": -914061504, "ts": 1716454216537648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537649, "dur": 0,
    "args": {
      "External id": 4007,
      "cbid": 205, "correlation": 4007
    }
  },
  {
    "ph": "f", "id": 4007, "pid": 76337, "tid": -914061504, "ts": 1716454216537649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560538, "dur": 9,
    "args": {
      "External id": 4011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4011,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4011, "pid": 5, "tid": 7, "ts": 1716454216560538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537662, "dur": 11,
    "args": {
      "External id": 4011,
      "cbid": 211, "correlation": 4011
    }
  },
  {
    "ph": "s", "id": 4011, "pid": 76337, "tid": -914061504, "ts": 1716454216537662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560549, "dur": 33,
    "args": {
      "External id": 4013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4013,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4013, "pid": 5, "tid": 7, "ts": 1716454216560549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537675, "dur": 5,
    "args": {
      "External id": 4013,
      "cbid": 211, "correlation": 4013
    }
  },
  {
    "ph": "s", "id": 4013, "pid": 76337, "tid": -914061504, "ts": 1716454216537675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216560584, "dur": 1,
    "args": {
      "External id": 4015,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 4015,
      "bytes": 768, "memory bandwidth (GB/s)": 0.4528301886792453
    }
  },
  {
    "ph": "f", "id": 4015, "pid": 5, "tid": 7, "ts": 1716454216560584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537686, "dur": 6,
    "args": {
      "External id": 4015,
      "cbid": 51, "correlation": 4015
    }
  },
  {
    "ph": "s", "id": 4015, "pid": 76337, "tid": -914061504, "ts": 1716454216537686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216560588, "dur": 265,
    "args": {
      "External id": 4016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4016,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4016, "pid": 5, "tid": 7, "ts": 1716454216560588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537693, "dur": 6,
    "args": {
      "External id": 4016,
      "cbid": 211, "correlation": 4016
    }
  },
  {
    "ph": "s", "id": 4016, "pid": 76337, "tid": -914061504, "ts": 1716454216537693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560854, "dur": 10,
    "args": {
      "External id": 4018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4018,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4018, "pid": 5, "tid": 7, "ts": 1716454216560854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537703, "dur": 5,
    "args": {
      "External id": 4018,
      "cbid": 211, "correlation": 4018
    }
  },
  {
    "ph": "s", "id": 4018, "pid": 76337, "tid": -914061504, "ts": 1716454216537703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560866, "dur": 12,
    "args": {
      "External id": 4024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4024,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4024, "pid": 5, "tid": 7, "ts": 1716454216560866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537730, "dur": 8,
    "args": {
      "External id": 4024,
      "cbid": 211, "correlation": 4024
    }
  },
  {
    "ph": "s", "id": 4024, "pid": 76337, "tid": -914061504, "ts": 1716454216537730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560879, "dur": 40,
    "args": {
      "External id": 4032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4032, "pid": 5, "tid": 7, "ts": 1716454216560879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537763, "dur": 9,
    "args": {
      "External id": 4032,
      "cbid": 211, "correlation": 4032
    }
  },
  {
    "ph": "s", "id": 4032, "pid": 76337, "tid": -914061504, "ts": 1716454216537763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216560920, "dur": 10,
    "args": {
      "External id": 4040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4040,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4040, "pid": 5, "tid": 7, "ts": 1716454216560920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537792, "dur": 9,
    "args": {
      "External id": 4040,
      "cbid": 211, "correlation": 4040
    }
  },
  {
    "ph": "s", "id": 4040, "pid": 76337, "tid": -914061504, "ts": 1716454216537792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216560931, "dur": 38,
    "args": {
      "External id": 4050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4050, "pid": 5, "tid": 7, "ts": 1716454216560931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537887, "dur": 12,
    "args": {
      "External id": 4050,
      "cbid": 211, "correlation": 4050
    }
  },
  {
    "ph": "s", "id": 4050, "pid": 76337, "tid": -914061504, "ts": 1716454216537887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216560971, "dur": 34,
    "args": {
      "External id": 4071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4071,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4071, "pid": 5, "tid": 7, "ts": 1716454216560971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537921, "dur": 7,
    "args": {
      "External id": 4071,
      "cbid": 211, "correlation": 4071
    }
  },
  {
    "ph": "s", "id": 4071, "pid": 76337, "tid": -914061504, "ts": 1716454216537921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216561006, "dur": 4,
    "args": {
      "External id": 4083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4083,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4083, "pid": 5, "tid": 7, "ts": 1716454216561006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537937, "dur": 6,
    "args": {
      "External id": 4083,
      "cbid": 211, "correlation": 4083
    }
  },
  {
    "ph": "s", "id": 4083, "pid": 76337, "tid": -914061504, "ts": 1716454216537937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561011, "dur": 13,
    "args": {
      "External id": 4086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4086, "pid": 5, "tid": 7, "ts": 1716454216561011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216537953, "dur": 7,
    "args": {
      "External id": 4086,
      "cbid": 211, "correlation": 4086
    }
  },
  {
    "ph": "s", "id": 4086, "pid": 76337, "tid": -914061504, "ts": 1716454216537953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216561026, "dur": 10,
    "args": {
      "External id": 4095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4095,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4095, "pid": 5, "tid": 7, "ts": 1716454216561026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538001, "dur": 10,
    "args": {
      "External id": 4095,
      "cbid": 211, "correlation": 4095
    }
  },
  {
    "ph": "s", "id": 4095, "pid": 76337, "tid": -914061504, "ts": 1716454216538001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538054, "dur": 0,
    "args": {
      "External id": 4105,
      "cbid": 317, "correlation": 4105
    }
  },
  {
    "ph": "f", "id": 4105, "pid": 76337, "tid": -914061504, "ts": 1716454216538054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538054, "dur": 0,
    "args": {
      "External id": 4106,
      "cbid": 203, "correlation": 4106
    }
  },
  {
    "ph": "f", "id": 4106, "pid": 76337, "tid": -914061504, "ts": 1716454216538054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538055, "dur": 0,
    "args": {
      "External id": 4107,
      "cbid": 205, "correlation": 4107
    }
  },
  {
    "ph": "f", "id": 4107, "pid": 76337, "tid": -914061504, "ts": 1716454216538055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561037, "dur": 9,
    "args": {
      "External id": 4111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4111,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4111, "pid": 5, "tid": 7, "ts": 1716454216561037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538069, "dur": 12,
    "args": {
      "External id": 4111,
      "cbid": 211, "correlation": 4111
    }
  },
  {
    "ph": "s", "id": 4111, "pid": 76337, "tid": -914061504, "ts": 1716454216538069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561047, "dur": 32,
    "args": {
      "External id": 4113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4113,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4113, "pid": 5, "tid": 7, "ts": 1716454216561047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538083, "dur": 5,
    "args": {
      "External id": 4113,
      "cbid": 211, "correlation": 4113
    }
  },
  {
    "ph": "s", "id": 4113, "pid": 76337, "tid": -914061504, "ts": 1716454216538083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216561082, "dur": 1,
    "args": {
      "External id": 4115,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 4115,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 4115, "pid": 5, "tid": 7, "ts": 1716454216561082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538094, "dur": 6,
    "args": {
      "External id": 4115,
      "cbid": 51, "correlation": 4115
    }
  },
  {
    "ph": "s", "id": 4115, "pid": 76337, "tid": -914061504, "ts": 1716454216538094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216561085, "dur": 266,
    "args": {
      "External id": 4116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4116,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4116, "pid": 5, "tid": 7, "ts": 1716454216561085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538101, "dur": 6,
    "args": {
      "External id": 4116,
      "cbid": 211, "correlation": 4116
    }
  },
  {
    "ph": "s", "id": 4116, "pid": 76337, "tid": -914061504, "ts": 1716454216538101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561353, "dur": 10,
    "args": {
      "External id": 4118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4118, "pid": 5, "tid": 7, "ts": 1716454216561353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538110, "dur": 5,
    "args": {
      "External id": 4118,
      "cbid": 211, "correlation": 4118
    }
  },
  {
    "ph": "s", "id": 4118, "pid": 76337, "tid": -914061504, "ts": 1716454216538110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561365, "dur": 12,
    "args": {
      "External id": 4124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4124,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4124, "pid": 5, "tid": 7, "ts": 1716454216561365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538138, "dur": 8,
    "args": {
      "External id": 4124,
      "cbid": 211, "correlation": 4124
    }
  },
  {
    "ph": "s", "id": 4124, "pid": 76337, "tid": -914061504, "ts": 1716454216538138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561379, "dur": 30,
    "args": {
      "External id": 4144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4144,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4144, "pid": 5, "tid": 7, "ts": 1716454216561379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538207, "dur": 11,
    "args": {
      "External id": 4144,
      "cbid": 211, "correlation": 4144
    }
  },
  {
    "ph": "s", "id": 4144, "pid": 76337, "tid": -914061504, "ts": 1716454216538207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216561410, "dur": 4,
    "args": {
      "External id": 4156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4156,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4156, "pid": 5, "tid": 7, "ts": 1716454216561410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538229, "dur": 6,
    "args": {
      "External id": 4156,
      "cbid": 211, "correlation": 4156
    }
  },
  {
    "ph": "s", "id": 4156, "pid": 76337, "tid": -914061504, "ts": 1716454216538229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561416, "dur": 14,
    "args": {
      "External id": 4159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4159,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4159, "pid": 5, "tid": 7, "ts": 1716454216561416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538246, "dur": 6,
    "args": {
      "External id": 4159,
      "cbid": 211, "correlation": 4159
    }
  },
  {
    "ph": "s", "id": 4159, "pid": 76337, "tid": -914061504, "ts": 1716454216538246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216561431, "dur": 9,
    "args": {
      "External id": 4168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4168,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4168, "pid": 5, "tid": 7, "ts": 1716454216561431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538286, "dur": 9,
    "args": {
      "External id": 4168,
      "cbid": 211, "correlation": 4168
    }
  },
  {
    "ph": "s", "id": 4168, "pid": 76337, "tid": -914061504, "ts": 1716454216538286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538348, "dur": 0,
    "args": {
      "External id": 4178,
      "cbid": 317, "correlation": 4178
    }
  },
  {
    "ph": "f", "id": 4178, "pid": 76337, "tid": -914061504, "ts": 1716454216538348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538348, "dur": 0,
    "args": {
      "External id": 4179,
      "cbid": 203, "correlation": 4179
    }
  },
  {
    "ph": "f", "id": 4179, "pid": 76337, "tid": -914061504, "ts": 1716454216538348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538349, "dur": 0,
    "args": {
      "External id": 4180,
      "cbid": 205, "correlation": 4180
    }
  },
  {
    "ph": "f", "id": 4180, "pid": 76337, "tid": -914061504, "ts": 1716454216538349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561442, "dur": 9,
    "args": {
      "External id": 4184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4184,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4184, "pid": 5, "tid": 7, "ts": 1716454216561442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538361, "dur": 12,
    "args": {
      "External id": 4184,
      "cbid": 211, "correlation": 4184
    }
  },
  {
    "ph": "s", "id": 4184, "pid": 76337, "tid": -914061504, "ts": 1716454216538361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561452, "dur": 33,
    "args": {
      "External id": 4186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4186, "pid": 5, "tid": 7, "ts": 1716454216561452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538376, "dur": 5,
    "args": {
      "External id": 4186,
      "cbid": 211, "correlation": 4186
    }
  },
  {
    "ph": "s", "id": 4186, "pid": 76337, "tid": -914061504, "ts": 1716454216538376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216561488, "dur": 1,
    "args": {
      "External id": 4188,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 4188,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 4188, "pid": 5, "tid": 7, "ts": 1716454216561488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538386, "dur": 7,
    "args": {
      "External id": 4188,
      "cbid": 51, "correlation": 4188
    }
  },
  {
    "ph": "s", "id": 4188, "pid": 76337, "tid": -914061504, "ts": 1716454216538386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216561491, "dur": 267,
    "args": {
      "External id": 4189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4189,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4189, "pid": 5, "tid": 7, "ts": 1716454216561491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538394, "dur": 6,
    "args": {
      "External id": 4189,
      "cbid": 211, "correlation": 4189
    }
  },
  {
    "ph": "s", "id": 4189, "pid": 76337, "tid": -914061504, "ts": 1716454216538394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561760, "dur": 10,
    "args": {
      "External id": 4191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4191,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4191, "pid": 5, "tid": 7, "ts": 1716454216561760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538403, "dur": 5,
    "args": {
      "External id": 4191,
      "cbid": 211, "correlation": 4191
    }
  },
  {
    "ph": "s", "id": 4191, "pid": 76337, "tid": -914061504, "ts": 1716454216538403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561772, "dur": 12,
    "args": {
      "External id": 4197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4197, "pid": 5, "tid": 7, "ts": 1716454216561772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538431, "dur": 8,
    "args": {
      "External id": 4197,
      "cbid": 211, "correlation": 4197
    }
  },
  {
    "ph": "s", "id": 4197, "pid": 76337, "tid": -914061504, "ts": 1716454216538431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561785, "dur": 39,
    "args": {
      "External id": 4205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4205, "pid": 5, "tid": 7, "ts": 1716454216561785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538464, "dur": 8,
    "args": {
      "External id": 4205,
      "cbid": 211, "correlation": 4205
    }
  },
  {
    "ph": "s", "id": 4205, "pid": 76337, "tid": -914061504, "ts": 1716454216538464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216561826, "dur": 10,
    "args": {
      "External id": 4213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4213,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4213, "pid": 5, "tid": 7, "ts": 1716454216561826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538493, "dur": 9,
    "args": {
      "External id": 4213,
      "cbid": 211, "correlation": 4213
    }
  },
  {
    "ph": "s", "id": 4213, "pid": 76337, "tid": -914061504, "ts": 1716454216538493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561837, "dur": 38,
    "args": {
      "External id": 4223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4223, "pid": 5, "tid": 7, "ts": 1716454216561837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538685, "dur": 16,
    "args": {
      "External id": 4223,
      "cbid": 211, "correlation": 4223
    }
  },
  {
    "ph": "s", "id": 4223, "pid": 76337, "tid": -914061504, "ts": 1716454216538685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216561876, "dur": 31,
    "args": {
      "External id": 4244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4244,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4244, "pid": 5, "tid": 7, "ts": 1716454216561876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538722, "dur": 7,
    "args": {
      "External id": 4244,
      "cbid": 211, "correlation": 4244
    }
  },
  {
    "ph": "s", "id": 4244, "pid": 76337, "tid": -914061504, "ts": 1716454216538722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216561908, "dur": 4,
    "args": {
      "External id": 4256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4256,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4256, "pid": 5, "tid": 7, "ts": 1716454216561908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538739, "dur": 6,
    "args": {
      "External id": 4256,
      "cbid": 211, "correlation": 4256
    }
  },
  {
    "ph": "s", "id": 4256, "pid": 76337, "tid": -914061504, "ts": 1716454216538739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561914, "dur": 14,
    "args": {
      "External id": 4259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4259,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4259, "pid": 5, "tid": 7, "ts": 1716454216561914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538756, "dur": 6,
    "args": {
      "External id": 4259,
      "cbid": 211, "correlation": 4259
    }
  },
  {
    "ph": "s", "id": 4259, "pid": 76337, "tid": -914061504, "ts": 1716454216538756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538901, "dur": 2,
    "args": {
      "External id": 4271,
      "cbid": 251, "correlation": 4271
    }
  },
  {
    "ph": "f", "id": 4271, "pid": 76337, "tid": -914061504, "ts": 1716454216538901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216561929, "dur": 39,
    "args": {
      "External id": 4272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4272,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 4272, "pid": 5, "tid": 7, "ts": 1716454216561929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538909, "dur": 14,
    "args": {
      "External id": 4272,
      "cbid": 211, "correlation": 4272
    }
  },
  {
    "ph": "s", "id": 4272, "pid": 76337, "tid": -914061504, "ts": 1716454216538909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216561969, "dur": 16,
    "args": {
      "External id": 4277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4277, "pid": 5, "tid": 7, "ts": 1716454216561969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216538948, "dur": 8,
    "args": {
      "External id": 4277,
      "cbid": 211, "correlation": 4277
    }
  },
  {
    "ph": "s", "id": 4277, "pid": 76337, "tid": -914061504, "ts": 1716454216538948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539033, "dur": 1,
    "args": {
      "External id": 4288,
      "cbid": 251, "correlation": 4288
    }
  },
  {
    "ph": "f", "id": 4288, "pid": 76337, "tid": -914061504, "ts": 1716454216539033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216561987, "dur": 36,
    "args": {
      "External id": 4289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4289,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 4289, "pid": 5, "tid": 7, "ts": 1716454216561987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539037, "dur": 13,
    "args": {
      "External id": 4289,
      "cbid": 211, "correlation": 4289
    }
  },
  {
    "ph": "s", "id": 4289, "pid": 76337, "tid": -914061504, "ts": 1716454216539037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216562024, "dur": 15,
    "args": {
      "External id": 4294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4294, "pid": 5, "tid": 7, "ts": 1716454216562024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539067, "dur": 8,
    "args": {
      "External id": 4294,
      "cbid": 211, "correlation": 4294
    }
  },
  {
    "ph": "s", "id": 4294, "pid": 76337, "tid": -914061504, "ts": 1716454216539067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539141, "dur": 1,
    "args": {
      "External id": 4305,
      "cbid": 251, "correlation": 4305
    }
  },
  {
    "ph": "f", "id": 4305, "pid": 76337, "tid": -914061504, "ts": 1716454216539141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216562041, "dur": 37,
    "args": {
      "External id": 4306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4306,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 4306, "pid": 5, "tid": 7, "ts": 1716454216562041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539145, "dur": 12,
    "args": {
      "External id": 4306,
      "cbid": 211, "correlation": 4306
    }
  },
  {
    "ph": "s", "id": 4306, "pid": 76337, "tid": -914061504, "ts": 1716454216539145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216562079, "dur": 15,
    "args": {
      "External id": 4311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4311, "pid": 5, "tid": 7, "ts": 1716454216562079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539173, "dur": 8,
    "args": {
      "External id": 4311,
      "cbid": 211, "correlation": 4311
    }
  },
  {
    "ph": "s", "id": 4311, "pid": 76337, "tid": -914061504, "ts": 1716454216539173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216562096, "dur": 1221,
    "args": {
      "External id": 4336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4336,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 4336, "pid": 5, "tid": 7, "ts": 1716454216562096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539274, "dur": 13,
    "args": {
      "External id": 4336,
      "cbid": 211, "correlation": 4336
    }
  },
  {
    "ph": "s", "id": 4336, "pid": 76337, "tid": -914061504, "ts": 1716454216539274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539402, "dur": 2,
    "args": {
      "External id": 4354,
      "cbid": 251, "correlation": 4354
    }
  },
  {
    "ph": "f", "id": 4354, "pid": 76337, "tid": -914061504, "ts": 1716454216539402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216563318, "dur": 42,
    "args": {
      "External id": 4356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4356,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 4356, "pid": 5, "tid": 7, "ts": 1716454216563318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539409, "dur": 14,
    "args": {
      "External id": 4356,
      "cbid": 211, "correlation": 4356
    }
  },
  {
    "ph": "s", "id": 4356, "pid": 76337, "tid": -914061504, "ts": 1716454216539409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216563361, "dur": 12,
    "args": {
      "External id": 4364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4364,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4364, "pid": 5, "tid": 7, "ts": 1716454216563361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539495, "dur": 14,
    "args": {
      "External id": 4364,
      "cbid": 211, "correlation": 4364
    }
  },
  {
    "ph": "s", "id": 4364, "pid": 76337, "tid": -914061504, "ts": 1716454216539495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216563375, "dur": 9,
    "args": {
      "External id": 4372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4372,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4372, "pid": 5, "tid": 7, "ts": 1716454216563375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539533, "dur": 9,
    "args": {
      "External id": 4372,
      "cbid": 211, "correlation": 4372
    }
  },
  {
    "ph": "s", "id": 4372, "pid": 76337, "tid": -914061504, "ts": 1716454216539533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216563385, "dur": 38,
    "args": {
      "External id": 4382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4382, "pid": 5, "tid": 7, "ts": 1716454216563385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539609, "dur": 13,
    "args": {
      "External id": 4382,
      "cbid": 211, "correlation": 4382
    }
  },
  {
    "ph": "s", "id": 4382, "pid": 76337, "tid": -914061504, "ts": 1716454216539609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563425, "dur": 33,
    "args": {
      "External id": 4403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4403,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4403, "pid": 5, "tid": 7, "ts": 1716454216563425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539642, "dur": 7,
    "args": {
      "External id": 4403,
      "cbid": 211, "correlation": 4403
    }
  },
  {
    "ph": "s", "id": 4403, "pid": 76337, "tid": -914061504, "ts": 1716454216539642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216563459, "dur": 4,
    "args": {
      "External id": 4415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4415,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4415, "pid": 5, "tid": 7, "ts": 1716454216563459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539659, "dur": 7,
    "args": {
      "External id": 4415,
      "cbid": 211, "correlation": 4415
    }
  },
  {
    "ph": "s", "id": 4415, "pid": 76337, "tid": -914061504, "ts": 1716454216539659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216563465, "dur": 14,
    "args": {
      "External id": 4418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4418, "pid": 5, "tid": 7, "ts": 1716454216563465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539676, "dur": 6,
    "args": {
      "External id": 4418,
      "cbid": 211, "correlation": 4418
    }
  },
  {
    "ph": "s", "id": 4418, "pid": 76337, "tid": -914061504, "ts": 1716454216539676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216563481, "dur": 10,
    "args": {
      "External id": 4427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4427,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4427, "pid": 5, "tid": 7, "ts": 1716454216563481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539715, "dur": 10,
    "args": {
      "External id": 4427,
      "cbid": 211, "correlation": 4427
    }
  },
  {
    "ph": "s", "id": 4427, "pid": 76337, "tid": -914061504, "ts": 1716454216539715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539770, "dur": 0,
    "args": {
      "External id": 4437,
      "cbid": 317, "correlation": 4437
    }
  },
  {
    "ph": "f", "id": 4437, "pid": 76337, "tid": -914061504, "ts": 1716454216539770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539771, "dur": 0,
    "args": {
      "External id": 4438,
      "cbid": 203, "correlation": 4438
    }
  },
  {
    "ph": "f", "id": 4438, "pid": 76337, "tid": -914061504, "ts": 1716454216539771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539772, "dur": 0,
    "args": {
      "External id": 4439,
      "cbid": 205, "correlation": 4439
    }
  },
  {
    "ph": "f", "id": 4439, "pid": 76337, "tid": -914061504, "ts": 1716454216539772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563492, "dur": 11,
    "args": {
      "External id": 4443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4443,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4443, "pid": 5, "tid": 7, "ts": 1716454216563492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539786, "dur": 12,
    "args": {
      "External id": 4443,
      "cbid": 211, "correlation": 4443
    }
  },
  {
    "ph": "s", "id": 4443, "pid": 76337, "tid": -914061504, "ts": 1716454216539786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563505, "dur": 32,
    "args": {
      "External id": 4445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4445,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4445, "pid": 5, "tid": 7, "ts": 1716454216563505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539800, "dur": 5,
    "args": {
      "External id": 4445,
      "cbid": 211, "correlation": 4445
    }
  },
  {
    "ph": "s", "id": 4445, "pid": 76337, "tid": -914061504, "ts": 1716454216539800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216563539, "dur": 1,
    "args": {
      "External id": 4447,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 4447,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 4447, "pid": 5, "tid": 7, "ts": 1716454216563539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539811, "dur": 6,
    "args": {
      "External id": 4447,
      "cbid": 51, "correlation": 4447
    }
  },
  {
    "ph": "s", "id": 4447, "pid": 76337, "tid": -914061504, "ts": 1716454216539811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216563543, "dur": 274,
    "args": {
      "External id": 4448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4448,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4448, "pid": 5, "tid": 7, "ts": 1716454216563543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539819, "dur": 7,
    "args": {
      "External id": 4448,
      "cbid": 211, "correlation": 4448
    }
  },
  {
    "ph": "s", "id": 4448, "pid": 76337, "tid": -914061504, "ts": 1716454216539819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563818, "dur": 11,
    "args": {
      "External id": 4450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4450,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4450, "pid": 5, "tid": 7, "ts": 1716454216563818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539829, "dur": 5,
    "args": {
      "External id": 4450,
      "cbid": 211, "correlation": 4450
    }
  },
  {
    "ph": "s", "id": 4450, "pid": 76337, "tid": -914061504, "ts": 1716454216539829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216563830, "dur": 13,
    "args": {
      "External id": 4456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4456, "pid": 5, "tid": 7, "ts": 1716454216563830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539857, "dur": 8,
    "args": {
      "External id": 4456,
      "cbid": 211, "correlation": 4456
    }
  },
  {
    "ph": "s", "id": 4456, "pid": 76337, "tid": -914061504, "ts": 1716454216539857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563845, "dur": 31,
    "args": {
      "External id": 4476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4476,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4476, "pid": 5, "tid": 7, "ts": 1716454216563845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539925, "dur": 12,
    "args": {
      "External id": 4476,
      "cbid": 211, "correlation": 4476
    }
  },
  {
    "ph": "s", "id": 4476, "pid": 76337, "tid": -914061504, "ts": 1716454216539925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216563877, "dur": 4,
    "args": {
      "External id": 4488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4488,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4488, "pid": 5, "tid": 7, "ts": 1716454216563877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539947, "dur": 6,
    "args": {
      "External id": 4488,
      "cbid": 211, "correlation": 4488
    }
  },
  {
    "ph": "s", "id": 4488, "pid": 76337, "tid": -914061504, "ts": 1716454216539947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216563882, "dur": 14,
    "args": {
      "External id": 4491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4491, "pid": 5, "tid": 7, "ts": 1716454216563882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216539964, "dur": 6,
    "args": {
      "External id": 4491,
      "cbid": 211, "correlation": 4491
    }
  },
  {
    "ph": "s", "id": 4491, "pid": 76337, "tid": -914061504, "ts": 1716454216539964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216563898, "dur": 9,
    "args": {
      "External id": 4500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4500,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4500, "pid": 5, "tid": 7, "ts": 1716454216563898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540013, "dur": 10,
    "args": {
      "External id": 4500,
      "cbid": 211, "correlation": 4500
    }
  },
  {
    "ph": "s", "id": 4500, "pid": 76337, "tid": -914061504, "ts": 1716454216540013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540078, "dur": 0,
    "args": {
      "External id": 4510,
      "cbid": 317, "correlation": 4510
    }
  },
  {
    "ph": "f", "id": 4510, "pid": 76337, "tid": -914061504, "ts": 1716454216540078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540078, "dur": 0,
    "args": {
      "External id": 4511,
      "cbid": 203, "correlation": 4511
    }
  },
  {
    "ph": "f", "id": 4511, "pid": 76337, "tid": -914061504, "ts": 1716454216540078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540079, "dur": 0,
    "args": {
      "External id": 4512,
      "cbid": 205, "correlation": 4512
    }
  },
  {
    "ph": "f", "id": 4512, "pid": 76337, "tid": -914061504, "ts": 1716454216540079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563908, "dur": 8,
    "args": {
      "External id": 4516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4516,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4516, "pid": 5, "tid": 7, "ts": 1716454216563908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540092, "dur": 12,
    "args": {
      "External id": 4516,
      "cbid": 211, "correlation": 4516
    }
  },
  {
    "ph": "s", "id": 4516, "pid": 76337, "tid": -914061504, "ts": 1716454216540092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216563918, "dur": 32,
    "args": {
      "External id": 4518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4518,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4518, "pid": 5, "tid": 7, "ts": 1716454216563918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540107, "dur": 5,
    "args": {
      "External id": 4518,
      "cbid": 211, "correlation": 4518
    }
  },
  {
    "ph": "s", "id": 4518, "pid": 76337, "tid": -914061504, "ts": 1716454216540107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216563953, "dur": 1,
    "args": {
      "External id": 4520,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 4520,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 4520, "pid": 5, "tid": 7, "ts": 1716454216563953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540117, "dur": 6,
    "args": {
      "External id": 4520,
      "cbid": 51, "correlation": 4520
    }
  },
  {
    "ph": "s", "id": 4520, "pid": 76337, "tid": -914061504, "ts": 1716454216540117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216563957, "dur": 265,
    "args": {
      "External id": 4521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4521,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4521, "pid": 5, "tid": 7, "ts": 1716454216563957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540125, "dur": 6,
    "args": {
      "External id": 4521,
      "cbid": 211, "correlation": 4521
    }
  },
  {
    "ph": "s", "id": 4521, "pid": 76337, "tid": -914061504, "ts": 1716454216540125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216564223, "dur": 11,
    "args": {
      "External id": 4523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4523,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4523, "pid": 5, "tid": 7, "ts": 1716454216564223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540134, "dur": 5,
    "args": {
      "External id": 4523,
      "cbid": 211, "correlation": 4523
    }
  },
  {
    "ph": "s", "id": 4523, "pid": 76337, "tid": -914061504, "ts": 1716454216540134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564235, "dur": 12,
    "args": {
      "External id": 4529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4529, "pid": 5, "tid": 7, "ts": 1716454216564235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540162, "dur": 8,
    "args": {
      "External id": 4529,
      "cbid": 211, "correlation": 4529
    }
  },
  {
    "ph": "s", "id": 4529, "pid": 76337, "tid": -914061504, "ts": 1716454216540162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564248, "dur": 40,
    "args": {
      "External id": 4537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4537,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4537, "pid": 5, "tid": 7, "ts": 1716454216564248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540196, "dur": 9,
    "args": {
      "External id": 4537,
      "cbid": 211, "correlation": 4537
    }
  },
  {
    "ph": "s", "id": 4537, "pid": 76337, "tid": -914061504, "ts": 1716454216540196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216564290, "dur": 9,
    "args": {
      "External id": 4545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4545,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4545, "pid": 5, "tid": 7, "ts": 1716454216564290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540226, "dur": 8,
    "args": {
      "External id": 4545,
      "cbid": 211, "correlation": 4545
    }
  },
  {
    "ph": "s", "id": 4545, "pid": 76337, "tid": -914061504, "ts": 1716454216540226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564301, "dur": 38,
    "args": {
      "External id": 4555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4555, "pid": 5, "tid": 7, "ts": 1716454216564301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540294, "dur": 13,
    "args": {
      "External id": 4555,
      "cbid": 211, "correlation": 4555
    }
  },
  {
    "ph": "s", "id": 4555, "pid": 76337, "tid": -914061504, "ts": 1716454216540294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216564340, "dur": 47,
    "args": {
      "External id": 4576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4576,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4576, "pid": 5, "tid": 7, "ts": 1716454216564340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540328, "dur": 7,
    "args": {
      "External id": 4576,
      "cbid": 211, "correlation": 4576
    }
  },
  {
    "ph": "s", "id": 4576, "pid": 76337, "tid": -914061504, "ts": 1716454216540328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216564389, "dur": 4,
    "args": {
      "External id": 4588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4588,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4588, "pid": 5, "tid": 7, "ts": 1716454216564389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540344, "dur": 5,
    "args": {
      "External id": 4588,
      "cbid": 211, "correlation": 4588
    }
  },
  {
    "ph": "s", "id": 4588, "pid": 76337, "tid": -914061504, "ts": 1716454216540344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564395, "dur": 13,
    "args": {
      "External id": 4591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4591,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4591, "pid": 5, "tid": 7, "ts": 1716454216564395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540361, "dur": 6,
    "args": {
      "External id": 4591,
      "cbid": 211, "correlation": 4591
    }
  },
  {
    "ph": "s", "id": 4591, "pid": 76337, "tid": -914061504, "ts": 1716454216540361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216564409, "dur": 9,
    "args": {
      "External id": 4600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4600,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4600, "pid": 5, "tid": 7, "ts": 1716454216564409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540403, "dur": 11,
    "args": {
      "External id": 4600,
      "cbid": 211, "correlation": 4600
    }
  },
  {
    "ph": "s", "id": 4600, "pid": 76337, "tid": -914061504, "ts": 1716454216540403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540455, "dur": 0,
    "args": {
      "External id": 4610,
      "cbid": 317, "correlation": 4610
    }
  },
  {
    "ph": "f", "id": 4610, "pid": 76337, "tid": -914061504, "ts": 1716454216540455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540456, "dur": 0,
    "args": {
      "External id": 4611,
      "cbid": 203, "correlation": 4611
    }
  },
  {
    "ph": "f", "id": 4611, "pid": 76337, "tid": -914061504, "ts": 1716454216540456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540457, "dur": 0,
    "args": {
      "External id": 4612,
      "cbid": 205, "correlation": 4612
    }
  },
  {
    "ph": "f", "id": 4612, "pid": 76337, "tid": -914061504, "ts": 1716454216540457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216564420, "dur": 9,
    "args": {
      "External id": 4616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4616,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4616, "pid": 5, "tid": 7, "ts": 1716454216564420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540472, "dur": 11,
    "args": {
      "External id": 4616,
      "cbid": 211, "correlation": 4616
    }
  },
  {
    "ph": "s", "id": 4616, "pid": 76337, "tid": -914061504, "ts": 1716454216540472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216564430, "dur": 4,
    "args": {
      "External id": 4618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4618,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1.6,
      "warps per SM": 12.8,
      "grid": [1, 16, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 20
    }
  },
  {
    "ph": "f", "id": 4618, "pid": 5, "tid": 7, "ts": 1716454216564430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540487, "dur": 5,
    "args": {
      "External id": 4618,
      "cbid": 211, "correlation": 4618
    }
  },
  {
    "ph": "s", "id": 4618, "pid": 76337, "tid": -914061504, "ts": 1716454216540487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216564437, "dur": 1,
    "args": {
      "External id": 4620,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 4620,
      "bytes": 2688, "memory bandwidth (GB/s)": 1.423728813559322
    }
  },
  {
    "ph": "f", "id": 4620, "pid": 5, "tid": 7, "ts": 1716454216564437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540498, "dur": 7,
    "args": {
      "External id": 4620,
      "cbid": 51, "correlation": 4620
    }
  },
  {
    "ph": "s", "id": 4620, "pid": 76337, "tid": -914061504, "ts": 1716454216540498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x32x64_stage1_warpsize2x1x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216564441, "dur": 60,
    "args": {
      "External id": 4621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4621,
      "registers per thread": 128,
      "shared memory": 12288,
      "blocks per SM": 8.4,
      "warps per SM": 33.6,
      "grid": [1, 48, 14],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 4621, "pid": 5, "tid": 7, "ts": 1716454216564441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540507, "dur": 9,
    "args": {
      "External id": 4621,
      "cbid": 211, "correlation": 4621
    }
  },
  {
    "ph": "s", "id": 4621, "pid": 76337, "tid": -914061504, "ts": 1716454216540507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216564503, "dur": 5,
    "args": {
      "External id": 4623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4623,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1.2,
      "warps per SM": 9.6,
      "grid": [96, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 4623, "pid": 5, "tid": 7, "ts": 1716454216564503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540521, "dur": 5,
    "args": {
      "External id": 4623,
      "cbid": 211, "correlation": 4623
    }
  },
  {
    "ph": "s", "id": 4623, "pid": 76337, "tid": -914061504, "ts": 1716454216540521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564509, "dur": 5,
    "args": {
      "External id": 4629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 4629, "pid": 5, "tid": 7, "ts": 1716454216564509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540548, "dur": 8,
    "args": {
      "External id": 4629,
      "cbid": 211, "correlation": 4629
    }
  },
  {
    "ph": "s", "id": 4629, "pid": 76337, "tid": -914061504, "ts": 1716454216540548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540613, "dur": 0,
    "args": {
      "External id": 4639,
      "cbid": 317, "correlation": 4639
    }
  },
  {
    "ph": "f", "id": 4639, "pid": 76337, "tid": -914061504, "ts": 1716454216540613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540613, "dur": 0,
    "args": {
      "External id": 4640,
      "cbid": 203, "correlation": 4640
    }
  },
  {
    "ph": "f", "id": 4640, "pid": 76337, "tid": -914061504, "ts": 1716454216540613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540614, "dur": 0,
    "args": {
      "External id": 4641,
      "cbid": 205, "correlation": 4641
    }
  },
  {
    "ph": "f", "id": 4641, "pid": 76337, "tid": -914061504, "ts": 1716454216540614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540635, "dur": 1,
    "args": {
      "External id": 4645,
      "cbid": 251, "correlation": 4645
    }
  },
  {
    "ph": "f", "id": 4645, "pid": 76337, "tid": -914061504, "ts": 1716454216540635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216564515, "dur": 11,
    "args": {
      "External id": 4646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4646,
      "registers per thread": 106,
      "shared memory": 16640,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 4646, "pid": 5, "tid": 7, "ts": 1716454216564515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540639, "dur": 13,
    "args": {
      "External id": 4646,
      "cbid": 211, "correlation": 4646
    }
  },
  {
    "ph": "s", "id": 4646, "pid": 76337, "tid": -914061504, "ts": 1716454216540639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564528, "dur": 4,
    "args": {
      "External id": 4652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4652,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 4652, "pid": 5, "tid": 7, "ts": 1716454216564528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540674, "dur": 9,
    "args": {
      "External id": 4652,
      "cbid": 211, "correlation": 4652
    }
  },
  {
    "ph": "s", "id": 4652, "pid": 76337, "tid": -914061504, "ts": 1716454216540674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216564533, "dur": 4,
    "args": {
      "External id": 4660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4660,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 4660, "pid": 5, "tid": 7, "ts": 1716454216564533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540746, "dur": 13,
    "args": {
      "External id": 4660,
      "cbid": 211, "correlation": 4660
    }
  },
  {
    "ph": "s", "id": 4660, "pid": 76337, "tid": -914061504, "ts": 1716454216540746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216564538, "dur": 3,
    "args": {
      "External id": 4668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4668,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 4668, "pid": 5, "tid": 7, "ts": 1716454216564538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540792, "dur": 12,
    "args": {
      "External id": 4668,
      "cbid": 211, "correlation": 4668
    }
  },
  {
    "ph": "s", "id": 4668, "pid": 76337, "tid": -914061504, "ts": 1716454216540792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216564543, "dur": 3,
    "args": {
      "External id": 4676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4676,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 4676, "pid": 5, "tid": 7, "ts": 1716454216564543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540823, "dur": 10,
    "args": {
      "External id": 4676,
      "cbid": 211, "correlation": 4676
    }
  },
  {
    "ph": "s", "id": 4676, "pid": 76337, "tid": -914061504, "ts": 1716454216540823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216564547, "dur": 3,
    "args": {
      "External id": 4684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4684,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 4684, "pid": 5, "tid": 7, "ts": 1716454216564547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540847, "dur": 7,
    "args": {
      "External id": 4684,
      "cbid": 211, "correlation": 4684
    }
  },
  {
    "ph": "s", "id": 4684, "pid": 76337, "tid": -914061504, "ts": 1716454216540847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564552, "dur": 5,
    "args": {
      "External id": 4698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0875,
      "warps per SM": 0.35,
      "grid": [7, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 4698, "pid": 5, "tid": 7, "ts": 1716454216564552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216540969, "dur": 23,
    "args": {
      "External id": 4698,
      "cbid": 211, "correlation": 4698
    }
  },
  {
    "ph": "s", "id": 4698, "pid": 76337, "tid": -914061504, "ts": 1716454216540969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541000, "dur": 0,
    "args": {
      "External id": 4705,
      "cbid": 317, "correlation": 4705
    }
  },
  {
    "ph": "f", "id": 4705, "pid": 76337, "tid": -914061504, "ts": 1716454216541000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541001, "dur": 0,
    "args": {
      "External id": 4706,
      "cbid": 203, "correlation": 4706
    }
  },
  {
    "ph": "f", "id": 4706, "pid": 76337, "tid": -914061504, "ts": 1716454216541001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541002, "dur": 0,
    "args": {
      "External id": 4707,
      "cbid": 205, "correlation": 4707
    }
  },
  {
    "ph": "f", "id": 4707, "pid": 76337, "tid": -914061504, "ts": 1716454216541002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_wo_smem_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x32x16_stage1_warpsize4x1x1_g1_tensor8x8x4_aligna2_alignc8_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216564558, "dur": 199,
    "args": {
      "External id": 4711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4711,
      "registers per thread": 111,
      "shared memory": 4096,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [4, 1536, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 4711, "pid": 5, "tid": 7, "ts": 1716454216564558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541018, "dur": 11,
    "args": {
      "External id": 4711,
      "cbid": 211, "correlation": 4711
    }
  },
  {
    "ph": "s", "id": 4711, "pid": 76337, "tid": -914061504, "ts": 1716454216541018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564759, "dur": 177,
    "args": {
      "External id": 4717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4717, "pid": 5, "tid": 7, "ts": 1716454216564759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541052, "dur": 8,
    "args": {
      "External id": 4717,
      "cbid": 211, "correlation": 4717
    }
  },
  {
    "ph": "s", "id": 4717, "pid": 76337, "tid": -914061504, "ts": 1716454216541052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216564938, "dur": 1807,
    "args": {
      "External id": 4727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4727, "pid": 5, "tid": 7, "ts": 1716454216564938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541128, "dur": 12,
    "args": {
      "External id": 4727,
      "cbid": 211, "correlation": 4727
    }
  },
  {
    "ph": "s", "id": 4727, "pid": 76337, "tid": -914061504, "ts": 1716454216541128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216566745, "dur": 700,
    "args": {
      "External id": 4748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4748,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4748, "pid": 5, "tid": 7, "ts": 1716454216566745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541161, "dur": 8,
    "args": {
      "External id": 4748,
      "cbid": 211, "correlation": 4748
    }
  },
  {
    "ph": "s", "id": 4748, "pid": 76337, "tid": -914061504, "ts": 1716454216541161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216567447, "dur": 4,
    "args": {
      "External id": 4760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4760,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4760, "pid": 5, "tid": 7, "ts": 1716454216567447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541178, "dur": 6,
    "args": {
      "External id": 4760,
      "cbid": 211, "correlation": 4760
    }
  },
  {
    "ph": "s", "id": 4760, "pid": 76337, "tid": -914061504, "ts": 1716454216541178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216567453, "dur": 176,
    "args": {
      "External id": 4763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4763, "pid": 5, "tid": 7, "ts": 1716454216567453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541194, "dur": 7,
    "args": {
      "External id": 4763,
      "cbid": 211, "correlation": 4763
    }
  },
  {
    "ph": "s", "id": 4763, "pid": 76337, "tid": -914061504, "ts": 1716454216541194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216567630, "dur": 109,
    "args": {
      "External id": 4772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4772,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4772, "pid": 5, "tid": 7, "ts": 1716454216567630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541233, "dur": 10,
    "args": {
      "External id": 4772,
      "cbid": 211, "correlation": 4772
    }
  },
  {
    "ph": "s", "id": 4772, "pid": 76337, "tid": -914061504, "ts": 1716454216541233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541283, "dur": 0,
    "args": {
      "External id": 4782,
      "cbid": 317, "correlation": 4782
    }
  },
  {
    "ph": "f", "id": 4782, "pid": 76337, "tid": -914061504, "ts": 1716454216541283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541284, "dur": 0,
    "args": {
      "External id": 4783,
      "cbid": 203, "correlation": 4783
    }
  },
  {
    "ph": "f", "id": 4783, "pid": 76337, "tid": -914061504, "ts": 1716454216541284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541285, "dur": 0,
    "args": {
      "External id": 4784,
      "cbid": 205, "correlation": 4784
    }
  },
  {
    "ph": "f", "id": 4784, "pid": 76337, "tid": -914061504, "ts": 1716454216541285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216567741, "dur": 137,
    "args": {
      "External id": 4788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4788, "pid": 5, "tid": 7, "ts": 1716454216567741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541299, "dur": 11,
    "args": {
      "External id": 4788,
      "cbid": 211, "correlation": 4788
    }
  },
  {
    "ph": "s", "id": 4788, "pid": 76337, "tid": -914061504, "ts": 1716454216541299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216567879, "dur": 6,
    "args": {
      "External id": 4790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 4790, "pid": 5, "tid": 7, "ts": 1716454216567879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541313, "dur": 5,
    "args": {
      "External id": 4790,
      "cbid": 211, "correlation": 4790
    }
  },
  {
    "ph": "s", "id": 4790, "pid": 76337, "tid": -914061504, "ts": 1716454216541313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216567887, "dur": 5,
    "args": {
      "External id": 4792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4792,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4792, "pid": 5, "tid": 7, "ts": 1716454216567887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541322, "dur": 6,
    "args": {
      "External id": 4792,
      "cbid": 211, "correlation": 4792
    }
  },
  {
    "ph": "s", "id": 4792, "pid": 76337, "tid": -914061504, "ts": 1716454216541322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541331, "dur": 0,
    "args": {
      "External id": 4793,
      "cbid": 51, "correlation": 4793
    }
  },
  {
    "ph": "s", "id": 4793, "pid": 76337, "tid": -914061504, "ts": 1716454216541331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216567893, "dur": 823,
    "args": {
      "External id": 4794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4794,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4794, "pid": 5, "tid": 7, "ts": 1716454216567893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541331, "dur": 5,
    "args": {
      "External id": 4794,
      "cbid": 211, "correlation": 4794
    }
  },
  {
    "ph": "s", "id": 4794, "pid": 76337, "tid": -914061504, "ts": 1716454216541331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216568718, "dur": 176,
    "args": {
      "External id": 4799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4799, "pid": 5, "tid": 7, "ts": 1716454216568718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541358, "dur": 8,
    "args": {
      "External id": 4799,
      "cbid": 211, "correlation": 4799
    }
  },
  {
    "ph": "s", "id": 4799, "pid": 76337, "tid": -914061504, "ts": 1716454216541358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216568895, "dur": 700,
    "args": {
      "External id": 4819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4819,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4819, "pid": 5, "tid": 7, "ts": 1716454216568895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541426, "dur": 11,
    "args": {
      "External id": 4819,
      "cbid": 211, "correlation": 4819
    }
  },
  {
    "ph": "s", "id": 4819, "pid": 76337, "tid": -914061504, "ts": 1716454216541426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216569597, "dur": 4,
    "args": {
      "External id": 4831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4831,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4831, "pid": 5, "tid": 7, "ts": 1716454216569597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541446, "dur": 7,
    "args": {
      "External id": 4831,
      "cbid": 211, "correlation": 4831
    }
  },
  {
    "ph": "s", "id": 4831, "pid": 76337, "tid": -914061504, "ts": 1716454216541446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216569602, "dur": 178,
    "args": {
      "External id": 4834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4834,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4834, "pid": 5, "tid": 7, "ts": 1716454216569602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541463, "dur": 6,
    "args": {
      "External id": 4834,
      "cbid": 211, "correlation": 4834
    }
  },
  {
    "ph": "s", "id": 4834, "pid": 76337, "tid": -914061504, "ts": 1716454216541463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216569781, "dur": 109,
    "args": {
      "External id": 4843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4843,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4843, "pid": 5, "tid": 7, "ts": 1716454216569781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541501, "dur": 10,
    "args": {
      "External id": 4843,
      "cbid": 211, "correlation": 4843
    }
  },
  {
    "ph": "s", "id": 4843, "pid": 76337, "tid": -914061504, "ts": 1716454216541501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541561, "dur": 0,
    "args": {
      "External id": 4853,
      "cbid": 317, "correlation": 4853
    }
  },
  {
    "ph": "f", "id": 4853, "pid": 76337, "tid": -914061504, "ts": 1716454216541561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541562, "dur": 0,
    "args": {
      "External id": 4854,
      "cbid": 203, "correlation": 4854
    }
  },
  {
    "ph": "f", "id": 4854, "pid": 76337, "tid": -914061504, "ts": 1716454216541562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541563, "dur": 0,
    "args": {
      "External id": 4855,
      "cbid": 205, "correlation": 4855
    }
  },
  {
    "ph": "f", "id": 4855, "pid": 76337, "tid": -914061504, "ts": 1716454216541563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216569892, "dur": 131,
    "args": {
      "External id": 4859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4859,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4859, "pid": 5, "tid": 7, "ts": 1716454216569892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541575, "dur": 12,
    "args": {
      "External id": 4859,
      "cbid": 211, "correlation": 4859
    }
  },
  {
    "ph": "s", "id": 4859, "pid": 76337, "tid": -914061504, "ts": 1716454216541575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216570024, "dur": 6,
    "args": {
      "External id": 4861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4861,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 4861, "pid": 5, "tid": 7, "ts": 1716454216570024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541589, "dur": 5,
    "args": {
      "External id": 4861,
      "cbid": 211, "correlation": 4861
    }
  },
  {
    "ph": "s", "id": 4861, "pid": 76337, "tid": -914061504, "ts": 1716454216541589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216570032, "dur": 5,
    "args": {
      "External id": 4863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4863, "pid": 5, "tid": 7, "ts": 1716454216570032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541598, "dur": 5,
    "args": {
      "External id": 4863,
      "cbid": 211, "correlation": 4863
    }
  },
  {
    "ph": "s", "id": 4863, "pid": 76337, "tid": -914061504, "ts": 1716454216541598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541606, "dur": 0,
    "args": {
      "External id": 4864,
      "cbid": 51, "correlation": 4864
    }
  },
  {
    "ph": "s", "id": 4864, "pid": 76337, "tid": -914061504, "ts": 1716454216541606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216570039, "dur": 819,
    "args": {
      "External id": 4865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4865,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4865, "pid": 5, "tid": 7, "ts": 1716454216570039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541606, "dur": 5,
    "args": {
      "External id": 4865,
      "cbid": 211, "correlation": 4865
    }
  },
  {
    "ph": "s", "id": 4865, "pid": 76337, "tid": -914061504, "ts": 1716454216541606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216570859, "dur": 177,
    "args": {
      "External id": 4870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4870, "pid": 5, "tid": 7, "ts": 1716454216570859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541633, "dur": 8,
    "args": {
      "External id": 4870,
      "cbid": 211, "correlation": 4870
    }
  },
  {
    "ph": "s", "id": 4870, "pid": 76337, "tid": -914061504, "ts": 1716454216541633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216571037, "dur": 748,
    "args": {
      "External id": 4878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4878, "pid": 5, "tid": 7, "ts": 1716454216571037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541666, "dur": 10,
    "args": {
      "External id": 4878,
      "cbid": 211, "correlation": 4878
    }
  },
  {
    "ph": "s", "id": 4878, "pid": 76337, "tid": -914061504, "ts": 1716454216541666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216571787, "dur": 103,
    "args": {
      "External id": 4886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4886,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4886, "pid": 5, "tid": 7, "ts": 1716454216571787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541697, "dur": 8,
    "args": {
      "External id": 4886,
      "cbid": 211, "correlation": 4886
    }
  },
  {
    "ph": "s", "id": 4886, "pid": 76337, "tid": -914061504, "ts": 1716454216541697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216571891, "dur": 1797,
    "args": {
      "External id": 4896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4896,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4896, "pid": 5, "tid": 7, "ts": 1716454216571891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541764, "dur": 15,
    "args": {
      "External id": 4896,
      "cbid": 211, "correlation": 4896
    }
  },
  {
    "ph": "s", "id": 4896, "pid": 76337, "tid": -914061504, "ts": 1716454216541764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216573690, "dur": 699,
    "args": {
      "External id": 4917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4917,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4917, "pid": 5, "tid": 7, "ts": 1716454216573690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541800, "dur": 8,
    "args": {
      "External id": 4917,
      "cbid": 211, "correlation": 4917
    }
  },
  {
    "ph": "s", "id": 4917, "pid": 76337, "tid": -914061504, "ts": 1716454216541800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216574390, "dur": 4,
    "args": {
      "External id": 4929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4929,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 4929, "pid": 5, "tid": 7, "ts": 1716454216574390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541817, "dur": 6,
    "args": {
      "External id": 4929,
      "cbid": 211, "correlation": 4929
    }
  },
  {
    "ph": "s", "id": 4929, "pid": 76337, "tid": -914061504, "ts": 1716454216541817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216574395, "dur": 177,
    "args": {
      "External id": 4932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4932, "pid": 5, "tid": 7, "ts": 1716454216574395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541833, "dur": 7,
    "args": {
      "External id": 4932,
      "cbid": 211, "correlation": 4932
    }
  },
  {
    "ph": "s", "id": 4932, "pid": 76337, "tid": -914061504, "ts": 1716454216541833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216574573, "dur": 110,
    "args": {
      "External id": 4941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4941,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4941, "pid": 5, "tid": 7, "ts": 1716454216574573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541871, "dur": 9,
    "args": {
      "External id": 4941,
      "cbid": 211, "correlation": 4941
    }
  },
  {
    "ph": "s", "id": 4941, "pid": 76337, "tid": -914061504, "ts": 1716454216541871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541922, "dur": 0,
    "args": {
      "External id": 4951,
      "cbid": 317, "correlation": 4951
    }
  },
  {
    "ph": "f", "id": 4951, "pid": 76337, "tid": -914061504, "ts": 1716454216541922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541923, "dur": 0,
    "args": {
      "External id": 4952,
      "cbid": 203, "correlation": 4952
    }
  },
  {
    "ph": "f", "id": 4952, "pid": 76337, "tid": -914061504, "ts": 1716454216541923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541923, "dur": 0,
    "args": {
      "External id": 4953,
      "cbid": 205, "correlation": 4953
    }
  },
  {
    "ph": "f", "id": 4953, "pid": 76337, "tid": -914061504, "ts": 1716454216541923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216574685, "dur": 137,
    "args": {
      "External id": 4957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4957,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4957, "pid": 5, "tid": 7, "ts": 1716454216574685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541936, "dur": 12,
    "args": {
      "External id": 4957,
      "cbid": 211, "correlation": 4957
    }
  },
  {
    "ph": "s", "id": 4957, "pid": 76337, "tid": -914061504, "ts": 1716454216541936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216574823, "dur": 6,
    "args": {
      "External id": 4959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4959,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 4959, "pid": 5, "tid": 7, "ts": 1716454216574823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541951, "dur": 5,
    "args": {
      "External id": 4959,
      "cbid": 211, "correlation": 4959
    }
  },
  {
    "ph": "s", "id": 4959, "pid": 76337, "tid": -914061504, "ts": 1716454216541951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216574831, "dur": 5,
    "args": {
      "External id": 4961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4961, "pid": 5, "tid": 7, "ts": 1716454216574831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541959, "dur": 5,
    "args": {
      "External id": 4961,
      "cbid": 211, "correlation": 4961
    }
  },
  {
    "ph": "s", "id": 4961, "pid": 76337, "tid": -914061504, "ts": 1716454216541959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541967, "dur": 0,
    "args": {
      "External id": 4962,
      "cbid": 51, "correlation": 4962
    }
  },
  {
    "ph": "s", "id": 4962, "pid": 76337, "tid": -914061504, "ts": 1716454216541967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216574837, "dur": 817,
    "args": {
      "External id": 4963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4963,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 4963, "pid": 5, "tid": 7, "ts": 1716454216574837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216541968, "dur": 14,
    "args": {
      "External id": 4963,
      "cbid": 211, "correlation": 4963
    }
  },
  {
    "ph": "s", "id": 4963, "pid": 76337, "tid": -914061504, "ts": 1716454216541968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216575656, "dur": 176,
    "args": {
      "External id": 4968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 4968, "pid": 5, "tid": 7, "ts": 1716454216575656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542005, "dur": 9,
    "args": {
      "External id": 4968,
      "cbid": 211, "correlation": 4968
    }
  },
  {
    "ph": "s", "id": 4968, "pid": 76337, "tid": -914061504, "ts": 1716454216542005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216575833, "dur": 702,
    "args": {
      "External id": 4988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 4988,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 4988, "pid": 5, "tid": 7, "ts": 1716454216575833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542073, "dur": 11,
    "args": {
      "External id": 4988,
      "cbid": 211, "correlation": 4988
    }
  },
  {
    "ph": "s", "id": 4988, "pid": 76337, "tid": -914061504, "ts": 1716454216542073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216576536, "dur": 4,
    "args": {
      "External id": 5000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5000,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5000, "pid": 5, "tid": 7, "ts": 1716454216576536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542093, "dur": 6,
    "args": {
      "External id": 5000,
      "cbid": 211, "correlation": 5000
    }
  },
  {
    "ph": "s", "id": 5000, "pid": 76337, "tid": -914061504, "ts": 1716454216542093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216576542, "dur": 176,
    "args": {
      "External id": 5003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5003, "pid": 5, "tid": 7, "ts": 1716454216576542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542110, "dur": 6,
    "args": {
      "External id": 5003,
      "cbid": 211, "correlation": 5003
    }
  },
  {
    "ph": "s", "id": 5003, "pid": 76337, "tid": -914061504, "ts": 1716454216542110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216576719, "dur": 110,
    "args": {
      "External id": 5012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5012,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5012, "pid": 5, "tid": 7, "ts": 1716454216576719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542148, "dur": 10,
    "args": {
      "External id": 5012,
      "cbid": 211, "correlation": 5012
    }
  },
  {
    "ph": "s", "id": 5012, "pid": 76337, "tid": -914061504, "ts": 1716454216542148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542208, "dur": 0,
    "args": {
      "External id": 5022,
      "cbid": 317, "correlation": 5022
    }
  },
  {
    "ph": "f", "id": 5022, "pid": 76337, "tid": -914061504, "ts": 1716454216542208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542209, "dur": 0,
    "args": {
      "External id": 5023,
      "cbid": 203, "correlation": 5023
    }
  },
  {
    "ph": "f", "id": 5023, "pid": 76337, "tid": -914061504, "ts": 1716454216542209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542209, "dur": 0,
    "args": {
      "External id": 5024,
      "cbid": 205, "correlation": 5024
    }
  },
  {
    "ph": "f", "id": 5024, "pid": 76337, "tid": -914061504, "ts": 1716454216542209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216576830, "dur": 141,
    "args": {
      "External id": 5028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5028, "pid": 5, "tid": 7, "ts": 1716454216576830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542222, "dur": 12,
    "args": {
      "External id": 5028,
      "cbid": 211, "correlation": 5028
    }
  },
  {
    "ph": "s", "id": 5028, "pid": 76337, "tid": -914061504, "ts": 1716454216542222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216576973, "dur": 6,
    "args": {
      "External id": 5030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5030,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 5030, "pid": 5, "tid": 7, "ts": 1716454216576973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542237, "dur": 5,
    "args": {
      "External id": 5030,
      "cbid": 211, "correlation": 5030
    }
  },
  {
    "ph": "s", "id": 5030, "pid": 76337, "tid": -914061504, "ts": 1716454216542237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216576980, "dur": 5,
    "args": {
      "External id": 5032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5032, "pid": 5, "tid": 7, "ts": 1716454216576980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542246, "dur": 5,
    "args": {
      "External id": 5032,
      "cbid": 211, "correlation": 5032
    }
  },
  {
    "ph": "s", "id": 5032, "pid": 76337, "tid": -914061504, "ts": 1716454216542246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542253, "dur": 0,
    "args": {
      "External id": 5033,
      "cbid": 51, "correlation": 5033
    }
  },
  {
    "ph": "s", "id": 5033, "pid": 76337, "tid": -914061504, "ts": 1716454216542253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216576986, "dur": 815,
    "args": {
      "External id": 5034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5034,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5034, "pid": 5, "tid": 7, "ts": 1716454216576986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542254, "dur": 5,
    "args": {
      "External id": 5034,
      "cbid": 211, "correlation": 5034
    }
  },
  {
    "ph": "s", "id": 5034, "pid": 76337, "tid": -914061504, "ts": 1716454216542254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216577803, "dur": 177,
    "args": {
      "External id": 5039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5039, "pid": 5, "tid": 7, "ts": 1716454216577803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542280, "dur": 8,
    "args": {
      "External id": 5039,
      "cbid": 211, "correlation": 5039
    }
  },
  {
    "ph": "s", "id": 5039, "pid": 76337, "tid": -914061504, "ts": 1716454216542280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216577982, "dur": 754,
    "args": {
      "External id": 5047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5047, "pid": 5, "tid": 7, "ts": 1716454216577982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542314, "dur": 10,
    "args": {
      "External id": 5047,
      "cbid": 211, "correlation": 5047
    }
  },
  {
    "ph": "s", "id": 5047, "pid": 76337, "tid": -914061504, "ts": 1716454216542314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216578737, "dur": 104,
    "args": {
      "External id": 5055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5055,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5055, "pid": 5, "tid": 7, "ts": 1716454216578737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542344, "dur": 8,
    "args": {
      "External id": 5055,
      "cbid": 211, "correlation": 5055
    }
  },
  {
    "ph": "s", "id": 5055, "pid": 76337, "tid": -914061504, "ts": 1716454216542344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216578842, "dur": 88,
    "args": {
      "External id": 5065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 617.2125,
      "warps per SM": 2468.85,
      "grid": [49377, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5065, "pid": 5, "tid": 7, "ts": 1716454216578842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542401, "dur": 10,
    "args": {
      "External id": 5065,
      "cbid": 211, "correlation": 5065
    }
  },
  {
    "ph": "s", "id": 5065, "pid": 76337, "tid": -914061504, "ts": 1716454216542401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216578932, "dur": 165,
    "args": {
      "External id": 5070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5070,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5070, "pid": 5, "tid": 7, "ts": 1716454216578932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542431, "dur": 7,
    "args": {
      "External id": 5070,
      "cbid": 211, "correlation": 5070
    }
  },
  {
    "ph": "s", "id": 5070, "pid": 76337, "tid": -914061504, "ts": 1716454216542431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216579098, "dur": 5,
    "args": {
      "External id": 5085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 3.6,
      "warps per SM": 14.4,
      "grid": [288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 22
    }
  },
  {
    "ph": "f", "id": 5085, "pid": 5, "tid": 7, "ts": 1716454216579098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542494, "dur": 12,
    "args": {
      "External id": 5085,
      "cbid": 211, "correlation": 5085
    }
  },
  {
    "ph": "s", "id": 5085, "pid": 76337, "tid": -914061504, "ts": 1716454216542494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542514, "dur": 0,
    "args": {
      "External id": 5092,
      "cbid": 317, "correlation": 5092
    }
  },
  {
    "ph": "f", "id": 5092, "pid": 76337, "tid": -914061504, "ts": 1716454216542514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542514, "dur": 0,
    "args": {
      "External id": 5093,
      "cbid": 203, "correlation": 5093
    }
  },
  {
    "ph": "f", "id": 5093, "pid": 76337, "tid": -914061504, "ts": 1716454216542514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542515, "dur": 0,
    "args": {
      "External id": 5094,
      "cbid": 205, "correlation": 5094
    }
  },
  {
    "ph": "f", "id": 5094, "pid": 76337, "tid": -914061504, "ts": 1716454216542515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216579105, "dur": 265,
    "args": {
      "External id": 5098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5098,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [1, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5098, "pid": 5, "tid": 7, "ts": 1716454216579105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542530, "dur": 9,
    "args": {
      "External id": 5098,
      "cbid": 211, "correlation": 5098
    }
  },
  {
    "ph": "s", "id": 5098, "pid": 76337, "tid": -914061504, "ts": 1716454216542530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216579372, "dur": 48,
    "args": {
      "External id": 5104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5104, "pid": 5, "tid": 7, "ts": 1716454216579372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542561, "dur": 8,
    "args": {
      "External id": 5104,
      "cbid": 211, "correlation": 5104
    }
  },
  {
    "ph": "s", "id": 5104, "pid": 76337, "tid": -914061504, "ts": 1716454216542561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216579421, "dur": 146,
    "args": {
      "External id": 5114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5114,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5114, "pid": 5, "tid": 7, "ts": 1716454216579421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542641, "dur": 13,
    "args": {
      "External id": 5114,
      "cbid": 211, "correlation": 5114
    }
  },
  {
    "ph": "s", "id": 5114, "pid": 76337, "tid": -914061504, "ts": 1716454216542641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216579568, "dur": 180,
    "args": {
      "External id": 5135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5135,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5135, "pid": 5, "tid": 7, "ts": 1716454216579568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542675, "dur": 7,
    "args": {
      "External id": 5135,
      "cbid": 211, "correlation": 5135
    }
  },
  {
    "ph": "s", "id": 5135, "pid": 76337, "tid": -914061504, "ts": 1716454216542675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216579750, "dur": 4,
    "args": {
      "External id": 5147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5147,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5147, "pid": 5, "tid": 7, "ts": 1716454216579750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542691, "dur": 6,
    "args": {
      "External id": 5147,
      "cbid": 211, "correlation": 5147
    }
  },
  {
    "ph": "s", "id": 5147, "pid": 76337, "tid": -914061504, "ts": 1716454216542691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216579755, "dur": 48,
    "args": {
      "External id": 5150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5150,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5150, "pid": 5, "tid": 7, "ts": 1716454216579755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542707, "dur": 7,
    "args": {
      "External id": 5150,
      "cbid": 211, "correlation": 5150
    }
  },
  {
    "ph": "s", "id": 5150, "pid": 76337, "tid": -914061504, "ts": 1716454216542707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216579805, "dur": 31,
    "args": {
      "External id": 5159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5159,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5159, "pid": 5, "tid": 7, "ts": 1716454216579805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542747, "dur": 9,
    "args": {
      "External id": 5159,
      "cbid": 211, "correlation": 5159
    }
  },
  {
    "ph": "s", "id": 5159, "pid": 76337, "tid": -914061504, "ts": 1716454216542747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542797, "dur": 0,
    "args": {
      "External id": 5169,
      "cbid": 317, "correlation": 5169
    }
  },
  {
    "ph": "f", "id": 5169, "pid": 76337, "tid": -914061504, "ts": 1716454216542797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542798, "dur": 0,
    "args": {
      "External id": 5170,
      "cbid": 203, "correlation": 5170
    }
  },
  {
    "ph": "f", "id": 5170, "pid": 76337, "tid": -914061504, "ts": 1716454216542798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542799, "dur": 0,
    "args": {
      "External id": 5171,
      "cbid": 205, "correlation": 5171
    }
  },
  {
    "ph": "f", "id": 5171, "pid": 76337, "tid": -914061504, "ts": 1716454216542799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216579837, "dur": 37,
    "args": {
      "External id": 5175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5175,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [1536, 4, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5175, "pid": 5, "tid": 7, "ts": 1716454216579837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542812, "dur": 15,
    "args": {
      "External id": 5175,
      "cbid": 211, "correlation": 5175
    }
  },
  {
    "ph": "s", "id": 5175, "pid": 76337, "tid": -914061504, "ts": 1716454216542812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216579875, "dur": 8,
    "args": {
      "External id": 5177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5177,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12.8,
      "warps per SM": 102.4,
      "grid": [1, 4, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5177, "pid": 5, "tid": 7, "ts": 1716454216579875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542830, "dur": 5,
    "args": {
      "External id": 5177,
      "cbid": 211, "correlation": 5177
    }
  },
  {
    "ph": "s", "id": 5177, "pid": 76337, "tid": -914061504, "ts": 1716454216542830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216579884, "dur": 4,
    "args": {
      "External id": 5179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5179,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 5179, "pid": 5, "tid": 7, "ts": 1716454216579884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542839, "dur": 5,
    "args": {
      "External id": 5179,
      "cbid": 211, "correlation": 5179
    }
  },
  {
    "ph": "s", "id": 5179, "pid": 76337, "tid": -914061504, "ts": 1716454216542839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542847, "dur": 0,
    "args": {
      "External id": 5180,
      "cbid": 51, "correlation": 5180
    }
  },
  {
    "ph": "s", "id": 5180, "pid": 76337, "tid": -914061504, "ts": 1716454216542847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216579890, "dur": 434,
    "args": {
      "External id": 5181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5181,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [192, 4, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5181, "pid": 5, "tid": 7, "ts": 1716454216579890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542848, "dur": 6,
    "args": {
      "External id": 5181,
      "cbid": 211, "correlation": 5181
    }
  },
  {
    "ph": "s", "id": 5181, "pid": 76337, "tid": -914061504, "ts": 1716454216542848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216580326, "dur": 91,
    "args": {
      "External id": 5186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5186, "pid": 5, "tid": 7, "ts": 1716454216580326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542876, "dur": 8,
    "args": {
      "External id": 5186,
      "cbid": 211, "correlation": 5186
    }
  },
  {
    "ph": "s", "id": 5186, "pid": 76337, "tid": -914061504, "ts": 1716454216542876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216580419, "dur": 366,
    "args": {
      "External id": 5206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5206,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5206, "pid": 5, "tid": 7, "ts": 1716454216580419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542944, "dur": 12,
    "args": {
      "External id": 5206,
      "cbid": 211, "correlation": 5206
    }
  },
  {
    "ph": "s", "id": 5206, "pid": 76337, "tid": -914061504, "ts": 1716454216542944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216580786, "dur": 4,
    "args": {
      "External id": 5218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5218,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5218, "pid": 5, "tid": 7, "ts": 1716454216580786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542965, "dur": 6,
    "args": {
      "External id": 5218,
      "cbid": 211, "correlation": 5218
    }
  },
  {
    "ph": "s", "id": 5218, "pid": 76337, "tid": -914061504, "ts": 1716454216542965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216580792, "dur": 91,
    "args": {
      "External id": 5221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5221,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5221, "pid": 5, "tid": 7, "ts": 1716454216580792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216542991, "dur": 8,
    "args": {
      "External id": 5221,
      "cbid": 211, "correlation": 5221
    }
  },
  {
    "ph": "s", "id": 5221, "pid": 76337, "tid": -914061504, "ts": 1716454216542991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216580884, "dur": 58,
    "args": {
      "External id": 5230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5230,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5230, "pid": 5, "tid": 7, "ts": 1716454216580884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543031, "dur": 10,
    "args": {
      "External id": 5230,
      "cbid": 211, "correlation": 5230
    }
  },
  {
    "ph": "s", "id": 5230, "pid": 76337, "tid": -914061504, "ts": 1716454216543031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543091, "dur": 0,
    "args": {
      "External id": 5240,
      "cbid": 317, "correlation": 5240
    }
  },
  {
    "ph": "f", "id": 5240, "pid": 76337, "tid": -914061504, "ts": 1716454216543091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543092, "dur": 0,
    "args": {
      "External id": 5241,
      "cbid": 203, "correlation": 5241
    }
  },
  {
    "ph": "f", "id": 5241, "pid": 76337, "tid": -914061504, "ts": 1716454216543092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543093, "dur": 0,
    "args": {
      "External id": 5242,
      "cbid": 205, "correlation": 5242
    }
  },
  {
    "ph": "f", "id": 5242, "pid": 76337, "tid": -914061504, "ts": 1716454216543093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216580944, "dur": 101,
    "args": {
      "External id": 5246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5246,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5246, "pid": 5, "tid": 7, "ts": 1716454216580944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543106, "dur": 11,
    "args": {
      "External id": 5246,
      "cbid": 211, "correlation": 5246
    }
  },
  {
    "ph": "s", "id": 5246, "pid": 76337, "tid": -914061504, "ts": 1716454216543106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216581046, "dur": 12,
    "args": {
      "External id": 5248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5248,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5248, "pid": 5, "tid": 7, "ts": 1716454216581046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543120, "dur": 5,
    "args": {
      "External id": 5248,
      "cbid": 211, "correlation": 5248
    }
  },
  {
    "ph": "s", "id": 5248, "pid": 76337, "tid": -914061504, "ts": 1716454216543120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216581059, "dur": 4,
    "args": {
      "External id": 5250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5250,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 5250, "pid": 5, "tid": 7, "ts": 1716454216581059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543129, "dur": 5,
    "args": {
      "External id": 5250,
      "cbid": 211, "correlation": 5250
    }
  },
  {
    "ph": "s", "id": 5250, "pid": 76337, "tid": -914061504, "ts": 1716454216543129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543137, "dur": 0,
    "args": {
      "External id": 5251,
      "cbid": 51, "correlation": 5251
    }
  },
  {
    "ph": "s", "id": 5251, "pid": 76337, "tid": -914061504, "ts": 1716454216543137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216581064, "dur": 774,
    "args": {
      "External id": 5252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5252,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5252, "pid": 5, "tid": 7, "ts": 1716454216581064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543138, "dur": 6,
    "args": {
      "External id": 5252,
      "cbid": 211, "correlation": 5252
    }
  },
  {
    "ph": "s", "id": 5252, "pid": 76337, "tid": -914061504, "ts": 1716454216543138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216581839, "dur": 91,
    "args": {
      "External id": 5257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5257, "pid": 5, "tid": 7, "ts": 1716454216581839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543166, "dur": 8,
    "args": {
      "External id": 5257,
      "cbid": 211, "correlation": 5257
    }
  },
  {
    "ph": "s", "id": 5257, "pid": 76337, "tid": -914061504, "ts": 1716454216543166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543223, "dur": 0,
    "args": {
      "External id": 5267,
      "cbid": 317, "correlation": 5267
    }
  },
  {
    "ph": "f", "id": 5267, "pid": 76337, "tid": -914061504, "ts": 1716454216543223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543223, "dur": 0,
    "args": {
      "External id": 5268,
      "cbid": 203, "correlation": 5268
    }
  },
  {
    "ph": "f", "id": 5268, "pid": 76337, "tid": -914061504, "ts": 1716454216543223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543224, "dur": 0,
    "args": {
      "External id": 5269,
      "cbid": 205, "correlation": 5269
    }
  },
  {
    "ph": "f", "id": 5269, "pid": 76337, "tid": -914061504, "ts": 1716454216543224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543254, "dur": 1,
    "args": {
      "External id": 5273,
      "cbid": 251, "correlation": 5273
    }
  },
  {
    "ph": "f", "id": 5273, "pid": 76337, "tid": -914061504, "ts": 1716454216543254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216581932, "dur": 86,
    "args": {
      "External id": 5274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5274,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [2, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5274, "pid": 5, "tid": 7, "ts": 1716454216581932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543258, "dur": 12,
    "args": {
      "External id": 5274,
      "cbid": 211, "correlation": 5274
    }
  },
  {
    "ph": "s", "id": 5274, "pid": 76337, "tid": -914061504, "ts": 1716454216543258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216582019, "dur": 91,
    "args": {
      "External id": 5280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5280, "pid": 5, "tid": 7, "ts": 1716454216582019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543291, "dur": 9,
    "args": {
      "External id": 5280,
      "cbid": 211, "correlation": 5280
    }
  },
  {
    "ph": "s", "id": 5280, "pid": 76337, "tid": -914061504, "ts": 1716454216543291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216582112, "dur": 253,
    "args": {
      "External id": 5288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5288, "pid": 5, "tid": 7, "ts": 1716454216582112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543322, "dur": 8,
    "args": {
      "External id": 5288,
      "cbid": 211, "correlation": 5288
    }
  },
  {
    "ph": "s", "id": 5288, "pid": 76337, "tid": -914061504, "ts": 1716454216543322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216582366, "dur": 54,
    "args": {
      "External id": 5296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5296,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5296, "pid": 5, "tid": 7, "ts": 1716454216582366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543350, "dur": 9,
    "args": {
      "External id": 5296,
      "cbid": 211, "correlation": 5296
    }
  },
  {
    "ph": "s", "id": 5296, "pid": 76337, "tid": -914061504, "ts": 1716454216543350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216582421, "dur": 373,
    "args": {
      "External id": 5306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5306, "pid": 5, "tid": 7, "ts": 1716454216582421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543419, "dur": 12,
    "args": {
      "External id": 5306,
      "cbid": 211, "correlation": 5306
    }
  },
  {
    "ph": "s", "id": 5306, "pid": 76337, "tid": -914061504, "ts": 1716454216543419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216582796, "dur": 375,
    "args": {
      "External id": 5327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5327,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5327, "pid": 5, "tid": 7, "ts": 1716454216582796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543450, "dur": 7,
    "args": {
      "External id": 5327,
      "cbid": 211, "correlation": 5327
    }
  },
  {
    "ph": "s", "id": 5327, "pid": 76337, "tid": -914061504, "ts": 1716454216543450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216583172, "dur": 4,
    "args": {
      "External id": 5339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5339,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5339, "pid": 5, "tid": 7, "ts": 1716454216583172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543466, "dur": 6,
    "args": {
      "External id": 5339,
      "cbid": 211, "correlation": 5339
    }
  },
  {
    "ph": "s", "id": 5339, "pid": 76337, "tid": -914061504, "ts": 1716454216543466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216583177, "dur": 90,
    "args": {
      "External id": 5342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5342,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5342, "pid": 5, "tid": 7, "ts": 1716454216583177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543483, "dur": 7,
    "args": {
      "External id": 5342,
      "cbid": 211, "correlation": 5342
    }
  },
  {
    "ph": "s", "id": 5342, "pid": 76337, "tid": -914061504, "ts": 1716454216543483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216583269, "dur": 57,
    "args": {
      "External id": 5351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5351,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5351, "pid": 5, "tid": 7, "ts": 1716454216583269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543521, "dur": 10,
    "args": {
      "External id": 5351,
      "cbid": 211, "correlation": 5351
    }
  },
  {
    "ph": "s", "id": 5351, "pid": 76337, "tid": -914061504, "ts": 1716454216543521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543571, "dur": 0,
    "args": {
      "External id": 5361,
      "cbid": 317, "correlation": 5361
    }
  },
  {
    "ph": "f", "id": 5361, "pid": 76337, "tid": -914061504, "ts": 1716454216543571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543571, "dur": 0,
    "args": {
      "External id": 5362,
      "cbid": 203, "correlation": 5362
    }
  },
  {
    "ph": "f", "id": 5362, "pid": 76337, "tid": -914061504, "ts": 1716454216543571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543572, "dur": 0,
    "args": {
      "External id": 5363,
      "cbid": 205, "correlation": 5363
    }
  },
  {
    "ph": "f", "id": 5363, "pid": 76337, "tid": -914061504, "ts": 1716454216543572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216583327, "dur": 104,
    "args": {
      "External id": 5367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5367,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5367, "pid": 5, "tid": 7, "ts": 1716454216583327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543586, "dur": 11,
    "args": {
      "External id": 5367,
      "cbid": 211, "correlation": 5367
    }
  },
  {
    "ph": "s", "id": 5367, "pid": 76337, "tid": -914061504, "ts": 1716454216543586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216583433, "dur": 12,
    "args": {
      "External id": 5369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5369,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5369, "pid": 5, "tid": 7, "ts": 1716454216583433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543600, "dur": 5,
    "args": {
      "External id": 5369,
      "cbid": 211, "correlation": 5369
    }
  },
  {
    "ph": "s", "id": 5369, "pid": 76337, "tid": -914061504, "ts": 1716454216543600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216583446, "dur": 4,
    "args": {
      "External id": 5371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5371,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 5371, "pid": 5, "tid": 7, "ts": 1716454216583446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543609, "dur": 5,
    "args": {
      "External id": 5371,
      "cbid": 211, "correlation": 5371
    }
  },
  {
    "ph": "s", "id": 5371, "pid": 76337, "tid": -914061504, "ts": 1716454216543609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543617, "dur": 0,
    "args": {
      "External id": 5372,
      "cbid": 51, "correlation": 5372
    }
  },
  {
    "ph": "s", "id": 5372, "pid": 76337, "tid": -914061504, "ts": 1716454216543617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216583451, "dur": 772,
    "args": {
      "External id": 5373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5373,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5373, "pid": 5, "tid": 7, "ts": 1716454216583451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543618, "dur": 5,
    "args": {
      "External id": 5373,
      "cbid": 211, "correlation": 5373
    }
  },
  {
    "ph": "s", "id": 5373, "pid": 76337, "tid": -914061504, "ts": 1716454216543618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216584224, "dur": 91,
    "args": {
      "External id": 5378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5378,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5378, "pid": 5, "tid": 7, "ts": 1716454216584224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543646, "dur": 8,
    "args": {
      "External id": 5378,
      "cbid": 211, "correlation": 5378
    }
  },
  {
    "ph": "s", "id": 5378, "pid": 76337, "tid": -914061504, "ts": 1716454216543646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216584317, "dur": 364,
    "args": {
      "External id": 5398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5398,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5398, "pid": 5, "tid": 7, "ts": 1716454216584317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543712, "dur": 11,
    "args": {
      "External id": 5398,
      "cbid": 211, "correlation": 5398
    }
  },
  {
    "ph": "s", "id": 5398, "pid": 76337, "tid": -914061504, "ts": 1716454216543712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216584683, "dur": 4,
    "args": {
      "External id": 5410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5410,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5410, "pid": 5, "tid": 7, "ts": 1716454216584683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543732, "dur": 6,
    "args": {
      "External id": 5410,
      "cbid": 211, "correlation": 5410
    }
  },
  {
    "ph": "s", "id": 5410, "pid": 76337, "tid": -914061504, "ts": 1716454216543732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216584689, "dur": 92,
    "args": {
      "External id": 5413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5413, "pid": 5, "tid": 7, "ts": 1716454216584689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543750, "dur": 6,
    "args": {
      "External id": 5413,
      "cbid": 211, "correlation": 5413
    }
  },
  {
    "ph": "s", "id": 5413, "pid": 76337, "tid": -914061504, "ts": 1716454216543750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216584781, "dur": 57,
    "args": {
      "External id": 5422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5422,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5422, "pid": 5, "tid": 7, "ts": 1716454216584781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543787, "dur": 10,
    "args": {
      "External id": 5422,
      "cbid": 211, "correlation": 5422
    }
  },
  {
    "ph": "s", "id": 5422, "pid": 76337, "tid": -914061504, "ts": 1716454216543787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543847, "dur": 0,
    "args": {
      "External id": 5432,
      "cbid": 317, "correlation": 5432
    }
  },
  {
    "ph": "f", "id": 5432, "pid": 76337, "tid": -914061504, "ts": 1716454216543847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543848, "dur": 0,
    "args": {
      "External id": 5433,
      "cbid": 203, "correlation": 5433
    }
  },
  {
    "ph": "f", "id": 5433, "pid": 76337, "tid": -914061504, "ts": 1716454216543848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543849, "dur": 0,
    "args": {
      "External id": 5434,
      "cbid": 205, "correlation": 5434
    }
  },
  {
    "ph": "f", "id": 5434, "pid": 76337, "tid": -914061504, "ts": 1716454216543849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216584840, "dur": 99,
    "args": {
      "External id": 5438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 153.6,
      "warps per SM": 1228.8,
      "grid": [1536, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5438, "pid": 5, "tid": 7, "ts": 1716454216584840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543862, "dur": 12,
    "args": {
      "External id": 5438,
      "cbid": 211, "correlation": 5438
    }
  },
  {
    "ph": "s", "id": 5438, "pid": 76337, "tid": -914061504, "ts": 1716454216543862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216584940, "dur": 11,
    "args": {
      "External id": 5440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5440,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5440, "pid": 5, "tid": 7, "ts": 1716454216584940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543876, "dur": 5,
    "args": {
      "External id": 5440,
      "cbid": 211, "correlation": 5440
    }
  },
  {
    "ph": "s", "id": 5440, "pid": 76337, "tid": -914061504, "ts": 1716454216543876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216584953, "dur": 4,
    "args": {
      "External id": 5442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 5442, "pid": 5, "tid": 7, "ts": 1716454216584953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543885, "dur": 5,
    "args": {
      "External id": 5442,
      "cbid": 211, "correlation": 5442
    }
  },
  {
    "ph": "s", "id": 5442, "pid": 76337, "tid": -914061504, "ts": 1716454216543885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543893, "dur": 0,
    "args": {
      "External id": 5443,
      "cbid": 51, "correlation": 5443
    }
  },
  {
    "ph": "s", "id": 5443, "pid": 76337, "tid": -914061504, "ts": 1716454216543893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216584958, "dur": 771,
    "args": {
      "External id": 5444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5444,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [384, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5444, "pid": 5, "tid": 7, "ts": 1716454216584958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543893, "dur": 5,
    "args": {
      "External id": 5444,
      "cbid": 211, "correlation": 5444
    }
  },
  {
    "ph": "s", "id": 5444, "pid": 76337, "tid": -914061504, "ts": 1716454216543893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216585731, "dur": 91,
    "args": {
      "External id": 5449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5449, "pid": 5, "tid": 7, "ts": 1716454216585731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543921, "dur": 8,
    "args": {
      "External id": 5449,
      "cbid": 211, "correlation": 5449
    }
  },
  {
    "ph": "s", "id": 5449, "pid": 76337, "tid": -914061504, "ts": 1716454216543921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216585824, "dur": 254,
    "args": {
      "External id": 5457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5457, "pid": 5, "tid": 7, "ts": 1716454216585824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543953, "dur": 9,
    "args": {
      "External id": 5457,
      "cbid": 211, "correlation": 5457
    }
  },
  {
    "ph": "s", "id": 5457, "pid": 76337, "tid": -914061504, "ts": 1716454216543953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216586079, "dur": 54,
    "args": {
      "External id": 5465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5465,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5465, "pid": 5, "tid": 7, "ts": 1716454216586079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216543992, "dur": 9,
    "args": {
      "External id": 5465,
      "cbid": 211, "correlation": 5465
    }
  },
  {
    "ph": "s", "id": 5465, "pid": 76337, "tid": -914061504, "ts": 1716454216543992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216586134, "dur": 46,
    "args": {
      "External id": 5475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 310.0125,
      "warps per SM": 1240.05,
      "grid": [24801, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5475, "pid": 5, "tid": 7, "ts": 1716454216586134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544048, "dur": 11,
    "args": {
      "External id": 5475,
      "cbid": 211, "correlation": 5475
    }
  },
  {
    "ph": "s", "id": 5475, "pid": 76337, "tid": -914061504, "ts": 1716454216544048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216586182, "dur": 87,
    "args": {
      "External id": 5480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5480, "pid": 5, "tid": 7, "ts": 1716454216586182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544078, "dur": 8,
    "args": {
      "External id": 5480,
      "cbid": 211, "correlation": 5480
    }
  },
  {
    "ph": "s", "id": 5480, "pid": 76337, "tid": -914061504, "ts": 1716454216544078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216586271, "dur": 9,
    "args": {
      "External id": 5495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5495,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 5495, "pid": 5, "tid": 7, "ts": 1716454216586271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544140, "dur": 12,
    "args": {
      "External id": 5495,
      "cbid": 211, "correlation": 5495
    }
  },
  {
    "ph": "s", "id": 5495, "pid": 76337, "tid": -914061504, "ts": 1716454216544140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544160, "dur": 0,
    "args": {
      "External id": 5502,
      "cbid": 317, "correlation": 5502
    }
  },
  {
    "ph": "f", "id": 5502, "pid": 76337, "tid": -914061504, "ts": 1716454216544160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544161, "dur": 0,
    "args": {
      "External id": 5503,
      "cbid": 203, "correlation": 5503
    }
  },
  {
    "ph": "f", "id": 5503, "pid": 76337, "tid": -914061504, "ts": 1716454216544161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544162, "dur": 0,
    "args": {
      "External id": 5504,
      "cbid": 205, "correlation": 5504
    }
  },
  {
    "ph": "f", "id": 5504, "pid": 76337, "tid": -914061504, "ts": 1716454216544162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216586281, "dur": 314,
    "args": {
      "External id": 5508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5508,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 1.2,
      "warps per SM": 9.6,
      "grid": [1, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5508, "pid": 5, "tid": 7, "ts": 1716454216586281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544177, "dur": 9,
    "args": {
      "External id": 5508,
      "cbid": 211, "correlation": 5508
    }
  },
  {
    "ph": "s", "id": 5508, "pid": 76337, "tid": -914061504, "ts": 1716454216544177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216586596, "dur": 26,
    "args": {
      "External id": 5514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5514, "pid": 5, "tid": 7, "ts": 1716454216586596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544208, "dur": 8,
    "args": {
      "External id": 5514,
      "cbid": 211, "correlation": 5514
    }
  },
  {
    "ph": "s", "id": 5514, "pid": 76337, "tid": -914061504, "ts": 1716454216544208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216586623, "dur": 69,
    "args": {
      "External id": 5524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5524, "pid": 5, "tid": 7, "ts": 1716454216586623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544287, "dur": 12,
    "args": {
      "External id": 5524,
      "cbid": 211, "correlation": 5524
    }
  },
  {
    "ph": "s", "id": 5524, "pid": 76337, "tid": -914061504, "ts": 1716454216544287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216586694, "dur": 94,
    "args": {
      "External id": 5545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5545,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5545, "pid": 5, "tid": 7, "ts": 1716454216586694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544320, "dur": 7,
    "args": {
      "External id": 5545,
      "cbid": 211, "correlation": 5545
    }
  },
  {
    "ph": "s", "id": 5545, "pid": 76337, "tid": -914061504, "ts": 1716454216544320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216586790, "dur": 5,
    "args": {
      "External id": 5557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5557,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5557, "pid": 5, "tid": 7, "ts": 1716454216586790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544336, "dur": 6,
    "args": {
      "External id": 5557,
      "cbid": 211, "correlation": 5557
    }
  },
  {
    "ph": "s", "id": 5557, "pid": 76337, "tid": -914061504, "ts": 1716454216544336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216586796, "dur": 26,
    "args": {
      "External id": 5560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5560, "pid": 5, "tid": 7, "ts": 1716454216586796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544353, "dur": 6,
    "args": {
      "External id": 5560,
      "cbid": 211, "correlation": 5560
    }
  },
  {
    "ph": "s", "id": 5560, "pid": 76337, "tid": -914061504, "ts": 1716454216544353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216586824, "dur": 18,
    "args": {
      "External id": 5569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5569,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5569, "pid": 5, "tid": 7, "ts": 1716454216586824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544390, "dur": 10,
    "args": {
      "External id": 5569,
      "cbid": 211, "correlation": 5569
    }
  },
  {
    "ph": "s", "id": 5569, "pid": 76337, "tid": -914061504, "ts": 1716454216544390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544441, "dur": 0,
    "args": {
      "External id": 5579,
      "cbid": 317, "correlation": 5579
    }
  },
  {
    "ph": "f", "id": 5579, "pid": 76337, "tid": -914061504, "ts": 1716454216544441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544442, "dur": 0,
    "args": {
      "External id": 5580,
      "cbid": 203, "correlation": 5580
    }
  },
  {
    "ph": "f", "id": 5580, "pid": 76337, "tid": -914061504, "ts": 1716454216544442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544443, "dur": 0,
    "args": {
      "External id": 5581,
      "cbid": 205, "correlation": 5581
    }
  },
  {
    "ph": "f", "id": 5581, "pid": 76337, "tid": -914061504, "ts": 1716454216544443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216586843, "dur": 20,
    "args": {
      "External id": 5585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 38.4,
      "warps per SM": 307.2,
      "grid": [384, 8, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5585, "pid": 5, "tid": 7, "ts": 1716454216586843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544456, "dur": 11,
    "args": {
      "External id": 5585,
      "cbid": 211, "correlation": 5585
    }
  },
  {
    "ph": "s", "id": 5585, "pid": 76337, "tid": -914061504, "ts": 1716454216544456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216586865, "dur": 19,
    "args": {
      "External id": 5587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 51.2,
      "warps per SM": 409.6,
      "grid": [1, 8, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5587, "pid": 5, "tid": 7, "ts": 1716454216586865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544470, "dur": 5,
    "args": {
      "External id": 5587,
      "cbid": 211, "correlation": 5587
    }
  },
  {
    "ph": "s", "id": 5587, "pid": 76337, "tid": -914061504, "ts": 1716454216544470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216586886, "dur": 1,
    "args": {
      "External id": 5589,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 5589,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.6
    }
  },
  {
    "ph": "f", "id": 5589, "pid": 5, "tid": 7, "ts": 1716454216586886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544481, "dur": 7,
    "args": {
      "External id": 5589,
      "cbid": 51, "correlation": 5589
    }
  },
  {
    "ph": "s", "id": 5589, "pid": 76337, "tid": -914061504, "ts": 1716454216544481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216586890, "dur": 465,
    "args": {
      "External id": 5590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5590,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [4, 96, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5590, "pid": 5, "tid": 7, "ts": 1716454216586890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544489, "dur": 6,
    "args": {
      "External id": 5590,
      "cbid": 211, "correlation": 5590
    }
  },
  {
    "ph": "s", "id": 5590, "pid": 76337, "tid": -914061504, "ts": 1716454216544489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216587357, "dur": 38,
    "args": {
      "External id": 5592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5592, "pid": 5, "tid": 7, "ts": 1716454216587357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544499, "dur": 6,
    "args": {
      "External id": 5592,
      "cbid": 211, "correlation": 5592
    }
  },
  {
    "ph": "s", "id": 5592, "pid": 76337, "tid": -914061504, "ts": 1716454216544499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216587396, "dur": 49,
    "args": {
      "External id": 5598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5598, "pid": 5, "tid": 7, "ts": 1716454216587396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544527, "dur": 8,
    "args": {
      "External id": 5598,
      "cbid": 211, "correlation": 5598
    }
  },
  {
    "ph": "s", "id": 5598, "pid": 76337, "tid": -914061504, "ts": 1716454216544527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216587447, "dur": 190,
    "args": {
      "External id": 5618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5618,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5618, "pid": 5, "tid": 7, "ts": 1716454216587447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544595, "dur": 11,
    "args": {
      "External id": 5618,
      "cbid": 211, "correlation": 5618
    }
  },
  {
    "ph": "s", "id": 5618, "pid": 76337, "tid": -914061504, "ts": 1716454216544595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216587638, "dur": 4,
    "args": {
      "External id": 5630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5630,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5630, "pid": 5, "tid": 7, "ts": 1716454216587638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544615, "dur": 7,
    "args": {
      "External id": 5630,
      "cbid": 211, "correlation": 5630
    }
  },
  {
    "ph": "s", "id": 5630, "pid": 76337, "tid": -914061504, "ts": 1716454216544615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216587644, "dur": 46,
    "args": {
      "External id": 5633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5633, "pid": 5, "tid": 7, "ts": 1716454216587644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544633, "dur": 6,
    "args": {
      "External id": 5633,
      "cbid": 211, "correlation": 5633
    }
  },
  {
    "ph": "s", "id": 5633, "pid": 76337, "tid": -914061504, "ts": 1716454216544633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216587692, "dur": 31,
    "args": {
      "External id": 5642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5642,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5642, "pid": 5, "tid": 7, "ts": 1716454216587692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544672, "dur": 11,
    "args": {
      "External id": 5642,
      "cbid": 211, "correlation": 5642
    }
  },
  {
    "ph": "s", "id": 5642, "pid": 76337, "tid": -914061504, "ts": 1716454216544672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544734, "dur": 0,
    "args": {
      "External id": 5652,
      "cbid": 317, "correlation": 5652
    }
  },
  {
    "ph": "f", "id": 5652, "pid": 76337, "tid": -914061504, "ts": 1716454216544734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544735, "dur": 0,
    "args": {
      "External id": 5653,
      "cbid": 203, "correlation": 5653
    }
  },
  {
    "ph": "f", "id": 5653, "pid": 76337, "tid": -914061504, "ts": 1716454216544735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544736, "dur": 0,
    "args": {
      "External id": 5654,
      "cbid": 205, "correlation": 5654
    }
  },
  {
    "ph": "f", "id": 5654, "pid": 76337, "tid": -914061504, "ts": 1716454216544736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216587724, "dur": 34,
    "args": {
      "External id": 5658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5658, "pid": 5, "tid": 7, "ts": 1716454216587724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544749, "dur": 12,
    "args": {
      "External id": 5658,
      "cbid": 211, "correlation": 5658
    }
  },
  {
    "ph": "s", "id": 5658, "pid": 76337, "tid": -914061504, "ts": 1716454216544749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216587760, "dur": 32,
    "args": {
      "External id": 5660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5660, "pid": 5, "tid": 7, "ts": 1716454216587760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544763, "dur": 5,
    "args": {
      "External id": 5660,
      "cbid": 211, "correlation": 5660
    }
  },
  {
    "ph": "s", "id": 5660, "pid": 76337, "tid": -914061504, "ts": 1716454216544763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216587794, "dur": 1,
    "args": {
      "External id": 5662,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 5662,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.5483870967741935
    }
  },
  {
    "ph": "f", "id": 5662, "pid": 5, "tid": 7, "ts": 1716454216587794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544774, "dur": 6,
    "args": {
      "External id": 5662,
      "cbid": 51, "correlation": 5662
    }
  },
  {
    "ph": "s", "id": 5662, "pid": 76337, "tid": -914061504, "ts": 1716454216544774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216587798, "dur": 884,
    "args": {
      "External id": 5663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5663,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5663, "pid": 5, "tid": 7, "ts": 1716454216587798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544781, "dur": 6,
    "args": {
      "External id": 5663,
      "cbid": 211, "correlation": 5663
    }
  },
  {
    "ph": "s", "id": 5663, "pid": 76337, "tid": -914061504, "ts": 1716454216544781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216588684, "dur": 36,
    "args": {
      "External id": 5665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5665, "pid": 5, "tid": 7, "ts": 1716454216588684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544792, "dur": 5,
    "args": {
      "External id": 5665,
      "cbid": 211, "correlation": 5665
    }
  },
  {
    "ph": "s", "id": 5665, "pid": 76337, "tid": -914061504, "ts": 1716454216544792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216588722, "dur": 49,
    "args": {
      "External id": 5671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5671, "pid": 5, "tid": 7, "ts": 1716454216588722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544820, "dur": 8,
    "args": {
      "External id": 5671,
      "cbid": 211, "correlation": 5671
    }
  },
  {
    "ph": "s", "id": 5671, "pid": 76337, "tid": -914061504, "ts": 1716454216544820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544877, "dur": 0,
    "args": {
      "External id": 5681,
      "cbid": 317, "correlation": 5681
    }
  },
  {
    "ph": "f", "id": 5681, "pid": 76337, "tid": -914061504, "ts": 1716454216544877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544878, "dur": 0,
    "args": {
      "External id": 5682,
      "cbid": 203, "correlation": 5682
    }
  },
  {
    "ph": "f", "id": 5682, "pid": 76337, "tid": -914061504, "ts": 1716454216544878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544879, "dur": 0,
    "args": {
      "External id": 5683,
      "cbid": 205, "correlation": 5683
    }
  },
  {
    "ph": "f", "id": 5683, "pid": 76337, "tid": -914061504, "ts": 1716454216544879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544906, "dur": 1,
    "args": {
      "External id": 5687,
      "cbid": 251, "correlation": 5687
    }
  },
  {
    "ph": "f", "id": 5687, "pid": 76337, "tid": -914061504, "ts": 1716454216544906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216588773, "dur": 68,
    "args": {
      "External id": 5688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5688,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [4, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5688, "pid": 5, "tid": 7, "ts": 1716454216588773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544909, "dur": 12,
    "args": {
      "External id": 5688,
      "cbid": 211, "correlation": 5688
    }
  },
  {
    "ph": "s", "id": 5688, "pid": 76337, "tid": -914061504, "ts": 1716454216544909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216588842, "dur": 47,
    "args": {
      "External id": 5694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5694, "pid": 5, "tid": 7, "ts": 1716454216588842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544941, "dur": 9,
    "args": {
      "External id": 5694,
      "cbid": 211, "correlation": 5694
    }
  },
  {
    "ph": "s", "id": 5694, "pid": 76337, "tid": -914061504, "ts": 1716454216544941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216588891, "dur": 140,
    "args": {
      "External id": 5702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5702, "pid": 5, "tid": 7, "ts": 1716454216588891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216544972, "dur": 17,
    "args": {
      "External id": 5702,
      "cbid": 211, "correlation": 5702
    }
  },
  {
    "ph": "s", "id": 5702, "pid": 76337, "tid": -914061504, "ts": 1716454216544972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216589032, "dur": 29,
    "args": {
      "External id": 5710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5710,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5710, "pid": 5, "tid": 7, "ts": 1716454216589032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545013, "dur": 8,
    "args": {
      "External id": 5710,
      "cbid": 211, "correlation": 5710
    }
  },
  {
    "ph": "s", "id": 5710, "pid": 76337, "tid": -914061504, "ts": 1716454216545013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216589062, "dur": 135,
    "args": {
      "External id": 5720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5720,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5720, "pid": 5, "tid": 7, "ts": 1716454216589062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545083, "dur": 12,
    "args": {
      "External id": 5720,
      "cbid": 211, "correlation": 5720
    }
  },
  {
    "ph": "s", "id": 5720, "pid": 76337, "tid": -914061504, "ts": 1716454216545083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216589199, "dur": 191,
    "args": {
      "External id": 5741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5741,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5741, "pid": 5, "tid": 7, "ts": 1716454216589199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545115, "dur": 7,
    "args": {
      "External id": 5741,
      "cbid": 211, "correlation": 5741
    }
  },
  {
    "ph": "s", "id": 5741, "pid": 76337, "tid": -914061504, "ts": 1716454216545115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216589392, "dur": 5,
    "args": {
      "External id": 5753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5753,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5753, "pid": 5, "tid": 7, "ts": 1716454216589392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545132, "dur": 6,
    "args": {
      "External id": 5753,
      "cbid": 211, "correlation": 5753
    }
  },
  {
    "ph": "s", "id": 5753, "pid": 76337, "tid": -914061504, "ts": 1716454216545132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216589397, "dur": 48,
    "args": {
      "External id": 5756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5756,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5756, "pid": 5, "tid": 7, "ts": 1716454216589397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545148, "dur": 7,
    "args": {
      "External id": 5756,
      "cbid": 211, "correlation": 5756
    }
  },
  {
    "ph": "s", "id": 5756, "pid": 76337, "tid": -914061504, "ts": 1716454216545148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216589447, "dur": 30,
    "args": {
      "External id": 5765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5765,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5765, "pid": 5, "tid": 7, "ts": 1716454216589447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545188, "dur": 9,
    "args": {
      "External id": 5765,
      "cbid": 211, "correlation": 5765
    }
  },
  {
    "ph": "s", "id": 5765, "pid": 76337, "tid": -914061504, "ts": 1716454216545188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545239, "dur": 0,
    "args": {
      "External id": 5775,
      "cbid": 317, "correlation": 5775
    }
  },
  {
    "ph": "f", "id": 5775, "pid": 76337, "tid": -914061504, "ts": 1716454216545239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545240, "dur": 0,
    "args": {
      "External id": 5776,
      "cbid": 203, "correlation": 5776
    }
  },
  {
    "ph": "f", "id": 5776, "pid": 76337, "tid": -914061504, "ts": 1716454216545240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545241, "dur": 0,
    "args": {
      "External id": 5777,
      "cbid": 205, "correlation": 5777
    }
  },
  {
    "ph": "f", "id": 5777, "pid": 76337, "tid": -914061504, "ts": 1716454216545241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216589478, "dur": 33,
    "args": {
      "External id": 5781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5781,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5781, "pid": 5, "tid": 7, "ts": 1716454216589478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545254, "dur": 12,
    "args": {
      "External id": 5781,
      "cbid": 211, "correlation": 5781
    }
  },
  {
    "ph": "s", "id": 5781, "pid": 76337, "tid": -914061504, "ts": 1716454216545254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216589513, "dur": 32,
    "args": {
      "External id": 5783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5783,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5783, "pid": 5, "tid": 7, "ts": 1716454216589513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545269, "dur": 5,
    "args": {
      "External id": 5783,
      "cbid": 211, "correlation": 5783
    }
  },
  {
    "ph": "s", "id": 5783, "pid": 76337, "tid": -914061504, "ts": 1716454216545269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216589548, "dur": 1,
    "args": {
      "External id": 5785,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 5785,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.5737704918032787
    }
  },
  {
    "ph": "f", "id": 5785, "pid": 5, "tid": 7, "ts": 1716454216589548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545279, "dur": 7,
    "args": {
      "External id": 5785,
      "cbid": 51, "correlation": 5785
    }
  },
  {
    "ph": "s", "id": 5785, "pid": 76337, "tid": -914061504, "ts": 1716454216545279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216589552, "dur": 894,
    "args": {
      "External id": 5786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5786,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5786, "pid": 5, "tid": 7, "ts": 1716454216589552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545287, "dur": 6,
    "args": {
      "External id": 5786,
      "cbid": 211, "correlation": 5786
    }
  },
  {
    "ph": "s", "id": 5786, "pid": 76337, "tid": -914061504, "ts": 1716454216545287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216590447, "dur": 38,
    "args": {
      "External id": 5788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5788, "pid": 5, "tid": 7, "ts": 1716454216590447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545296, "dur": 5,
    "args": {
      "External id": 5788,
      "cbid": 211, "correlation": 5788
    }
  },
  {
    "ph": "s", "id": 5788, "pid": 76337, "tid": -914061504, "ts": 1716454216545296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216590486, "dur": 49,
    "args": {
      "External id": 5794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5794,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5794, "pid": 5, "tid": 7, "ts": 1716454216590486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545325, "dur": 8,
    "args": {
      "External id": 5794,
      "cbid": 211, "correlation": 5794
    }
  },
  {
    "ph": "s", "id": 5794, "pid": 76337, "tid": -914061504, "ts": 1716454216545325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216590537, "dur": 189,
    "args": {
      "External id": 5814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5814,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5814, "pid": 5, "tid": 7, "ts": 1716454216590537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545392, "dur": 11,
    "args": {
      "External id": 5814,
      "cbid": 211, "correlation": 5814
    }
  },
  {
    "ph": "s", "id": 5814, "pid": 76337, "tid": -914061504, "ts": 1716454216545392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216590728, "dur": 4,
    "args": {
      "External id": 5826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5826,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5826, "pid": 5, "tid": 7, "ts": 1716454216590728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545412, "dur": 6,
    "args": {
      "External id": 5826,
      "cbid": 211, "correlation": 5826
    }
  },
  {
    "ph": "s", "id": 5826, "pid": 76337, "tid": -914061504, "ts": 1716454216545412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216590733, "dur": 47,
    "args": {
      "External id": 5829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5829, "pid": 5, "tid": 7, "ts": 1716454216590733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545430, "dur": 6,
    "args": {
      "External id": 5829,
      "cbid": 211, "correlation": 5829
    }
  },
  {
    "ph": "s", "id": 5829, "pid": 76337, "tid": -914061504, "ts": 1716454216545430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216590782, "dur": 31,
    "args": {
      "External id": 5838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5838,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5838, "pid": 5, "tid": 7, "ts": 1716454216590782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545469, "dur": 10,
    "args": {
      "External id": 5838,
      "cbid": 211, "correlation": 5838
    }
  },
  {
    "ph": "s", "id": 5838, "pid": 76337, "tid": -914061504, "ts": 1716454216545469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545530, "dur": 0,
    "args": {
      "External id": 5848,
      "cbid": 317, "correlation": 5848
    }
  },
  {
    "ph": "f", "id": 5848, "pid": 76337, "tid": -914061504, "ts": 1716454216545530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545531, "dur": 0,
    "args": {
      "External id": 5849,
      "cbid": 203, "correlation": 5849
    }
  },
  {
    "ph": "f", "id": 5849, "pid": 76337, "tid": -914061504, "ts": 1716454216545531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545532, "dur": 0,
    "args": {
      "External id": 5850,
      "cbid": 205, "correlation": 5850
    }
  },
  {
    "ph": "f", "id": 5850, "pid": 76337, "tid": -914061504, "ts": 1716454216545532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216590814, "dur": 34,
    "args": {
      "External id": 5854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5854,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5854, "pid": 5, "tid": 7, "ts": 1716454216590814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545545, "dur": 12,
    "args": {
      "External id": 5854,
      "cbid": 211, "correlation": 5854
    }
  },
  {
    "ph": "s", "id": 5854, "pid": 76337, "tid": -914061504, "ts": 1716454216545545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216590849, "dur": 33,
    "args": {
      "External id": 5856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5856, "pid": 5, "tid": 7, "ts": 1716454216590849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545559, "dur": 5,
    "args": {
      "External id": 5856,
      "cbid": 211, "correlation": 5856
    }
  },
  {
    "ph": "s", "id": 5856, "pid": 76337, "tid": -914061504, "ts": 1716454216545559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216590884, "dur": 1,
    "args": {
      "External id": 5858,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 5858,
      "bytes": 3072, "memory bandwidth (GB/s)": 1.5737704918032787
    }
  },
  {
    "ph": "f", "id": 5858, "pid": 5, "tid": 7, "ts": 1716454216590884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545569, "dur": 6,
    "args": {
      "External id": 5858,
      "cbid": 51, "correlation": 5858
    }
  },
  {
    "ph": "s", "id": 5858, "pid": 76337, "tid": -914061504, "ts": 1716454216545569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216590888, "dur": 890,
    "args": {
      "External id": 5859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5859,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 96, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 5859, "pid": 5, "tid": 7, "ts": 1716454216590888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545577, "dur": 7,
    "args": {
      "External id": 5859,
      "cbid": 211, "correlation": 5859
    }
  },
  {
    "ph": "s", "id": 5859, "pid": 76337, "tid": -914061504, "ts": 1716454216545577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216591779, "dur": 37,
    "args": {
      "External id": 5861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5861,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [384, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5861, "pid": 5, "tid": 7, "ts": 1716454216591779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545587, "dur": 5,
    "args": {
      "External id": 5861,
      "cbid": 211, "correlation": 5861
    }
  },
  {
    "ph": "s", "id": 5861, "pid": 76337, "tid": -914061504, "ts": 1716454216545587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216591818, "dur": 49,
    "args": {
      "External id": 5867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5867, "pid": 5, "tid": 7, "ts": 1716454216591818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545614, "dur": 8,
    "args": {
      "External id": 5867,
      "cbid": 211, "correlation": 5867
    }
  },
  {
    "ph": "s", "id": 5867, "pid": 76337, "tid": -914061504, "ts": 1716454216545614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216591868, "dur": 141,
    "args": {
      "External id": 5875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5875, "pid": 5, "tid": 7, "ts": 1716454216591868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545649, "dur": 9,
    "args": {
      "External id": 5875,
      "cbid": 211, "correlation": 5875
    }
  },
  {
    "ph": "s", "id": 5875, "pid": 76337, "tid": -914061504, "ts": 1716454216545649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216592010, "dur": 28,
    "args": {
      "External id": 5883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5883,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5883, "pid": 5, "tid": 7, "ts": 1716454216592010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545677, "dur": 8,
    "args": {
      "External id": 5883,
      "cbid": 211, "correlation": 5883
    }
  },
  {
    "ph": "s", "id": 5883, "pid": 76337, "tid": -914061504, "ts": 1716454216545677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<c10::Half>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454216592039, "dur": 25,
    "args": {
      "External id": 5893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5893,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 156.4125,
      "warps per SM": 625.65,
      "grid": [12513, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5893, "pid": 5, "tid": 7, "ts": 1716454216592039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545731, "dur": 11,
    "args": {
      "External id": 5893,
      "cbid": 211, "correlation": 5893
    }
  },
  {
    "ph": "s", "id": 5893, "pid": 76337, "tid": -914061504, "ts": 1716454216545731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592065, "dur": 47,
    "args": {
      "External id": 5898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5898, "pid": 5, "tid": 7, "ts": 1716454216592065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545761, "dur": 8,
    "args": {
      "External id": 5898,
      "cbid": 211, "correlation": 5898
    }
  },
  {
    "ph": "s", "id": 5898, "pid": 76337, "tid": -914061504, "ts": 1716454216545761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592113, "dur": 25,
    "args": {
      "External id": 5913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5913,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 57.6,
      "warps per SM": 230.4,
      "grid": [4608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5913, "pid": 5, "tid": 7, "ts": 1716454216592113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545823, "dur": 12,
    "args": {
      "External id": 5913,
      "cbid": 211, "correlation": 5913
    }
  },
  {
    "ph": "s", "id": 5913, "pid": 76337, "tid": -914061504, "ts": 1716454216545823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545842, "dur": 0,
    "args": {
      "External id": 5920,
      "cbid": 317, "correlation": 5920
    }
  },
  {
    "ph": "f", "id": 5920, "pid": 76337, "tid": -914061504, "ts": 1716454216545842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545843, "dur": 0,
    "args": {
      "External id": 5921,
      "cbid": 203, "correlation": 5921
    }
  },
  {
    "ph": "f", "id": 5921, "pid": 76337, "tid": -914061504, "ts": 1716454216545843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545844, "dur": 0,
    "args": {
      "External id": 5922,
      "cbid": 205, "correlation": 5922
    }
  },
  {
    "ph": "f", "id": 5922, "pid": 76337, "tid": -914061504, "ts": 1716454216545844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216592139, "dur": 312,
    "args": {
      "External id": 5926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5926,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [4, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 5926, "pid": 5, "tid": 7, "ts": 1716454216592139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545859, "dur": 8,
    "args": {
      "External id": 5926,
      "cbid": 211, "correlation": 5926
    }
  },
  {
    "ph": "s", "id": 5926, "pid": 76337, "tid": -914061504, "ts": 1716454216545859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592453, "dur": 14,
    "args": {
      "External id": 5932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5932, "pid": 5, "tid": 7, "ts": 1716454216592453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545889, "dur": 8,
    "args": {
      "External id": 5932,
      "cbid": 211, "correlation": 5932
    }
  },
  {
    "ph": "s", "id": 5932, "pid": 76337, "tid": -914061504, "ts": 1716454216545889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592468, "dur": 37,
    "args": {
      "External id": 5942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5942,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5942, "pid": 5, "tid": 7, "ts": 1716454216592468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216545969, "dur": 20,
    "args": {
      "External id": 5942,
      "cbid": 211, "correlation": 5942
    }
  },
  {
    "ph": "s", "id": 5942, "pid": 76337, "tid": -914061504, "ts": 1716454216545969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216592507, "dur": 49,
    "args": {
      "External id": 5963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5963,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 5963, "pid": 5, "tid": 7, "ts": 1716454216592507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546010, "dur": 8,
    "args": {
      "External id": 5963,
      "cbid": 211, "correlation": 5963
    }
  },
  {
    "ph": "s", "id": 5963, "pid": 76337, "tid": -914061504, "ts": 1716454216546010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216592557, "dur": 4,
    "args": {
      "External id": 5975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5975,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 5975, "pid": 5, "tid": 7, "ts": 1716454216592557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546028, "dur": 6,
    "args": {
      "External id": 5975,
      "cbid": 211, "correlation": 5975
    }
  },
  {
    "ph": "s", "id": 5975, "pid": 76337, "tid": -914061504, "ts": 1716454216546028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592562, "dur": 15,
    "args": {
      "External id": 5978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5978,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5978, "pid": 5, "tid": 7, "ts": 1716454216592562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546044, "dur": 6,
    "args": {
      "External id": 5978,
      "cbid": 211, "correlation": 5978
    }
  },
  {
    "ph": "s", "id": 5978, "pid": 76337, "tid": -914061504, "ts": 1716454216546044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216592579, "dur": 9,
    "args": {
      "External id": 5987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 5987,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 5987, "pid": 5, "tid": 7, "ts": 1716454216592579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546085, "dur": 9,
    "args": {
      "External id": 5987,
      "cbid": 211, "correlation": 5987
    }
  },
  {
    "ph": "s", "id": 5987, "pid": 76337, "tid": -914061504, "ts": 1716454216546085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546136, "dur": 0,
    "args": {
      "External id": 5997,
      "cbid": 317, "correlation": 5997
    }
  },
  {
    "ph": "f", "id": 5997, "pid": 76337, "tid": -914061504, "ts": 1716454216546136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546136, "dur": 0,
    "args": {
      "External id": 5998,
      "cbid": 203, "correlation": 5998
    }
  },
  {
    "ph": "f", "id": 5998, "pid": 76337, "tid": -914061504, "ts": 1716454216546136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546137, "dur": 0,
    "args": {
      "External id": 5999,
      "cbid": 205, "correlation": 5999
    }
  },
  {
    "ph": "f", "id": 5999, "pid": 76337, "tid": -914061504, "ts": 1716454216546137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216592590, "dur": 11,
    "args": {
      "External id": 6003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6003,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6003, "pid": 5, "tid": 7, "ts": 1716454216592590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546150, "dur": 12,
    "args": {
      "External id": 6003,
      "cbid": 211, "correlation": 6003
    }
  },
  {
    "ph": "s", "id": 6003, "pid": 76337, "tid": -914061504, "ts": 1716454216546150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216592602, "dur": 33,
    "args": {
      "External id": 6005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6005,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6005, "pid": 5, "tid": 7, "ts": 1716454216592602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546164, "dur": 5,
    "args": {
      "External id": 6005,
      "cbid": 211, "correlation": 6005
    }
  },
  {
    "ph": "s", "id": 6005, "pid": 76337, "tid": -914061504, "ts": 1716454216546164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216592637, "dur": 1,
    "args": {
      "External id": 6007,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6007,
      "bytes": 768, "memory bandwidth (GB/s)": 0.4525633470830878
    }
  },
  {
    "ph": "f", "id": 6007, "pid": 5, "tid": 7, "ts": 1716454216592637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546175, "dur": 7,
    "args": {
      "External id": 6007,
      "cbid": 51, "correlation": 6007
    }
  },
  {
    "ph": "s", "id": 6007, "pid": 76337, "tid": -914061504, "ts": 1716454216546175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216592641, "dur": 270,
    "args": {
      "External id": 6008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6008,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6008, "pid": 5, "tid": 7, "ts": 1716454216592641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546184, "dur": 6,
    "args": {
      "External id": 6008,
      "cbid": 211, "correlation": 6008
    }
  },
  {
    "ph": "s", "id": 6008, "pid": 76337, "tid": -914061504, "ts": 1716454216546184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216592913, "dur": 11,
    "args": {
      "External id": 6010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6010,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6010, "pid": 5, "tid": 7, "ts": 1716454216592913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546193, "dur": 5,
    "args": {
      "External id": 6010,
      "cbid": 211, "correlation": 6010
    }
  },
  {
    "ph": "s", "id": 6010, "pid": 76337, "tid": -914061504, "ts": 1716454216546193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592926, "dur": 13,
    "args": {
      "External id": 6016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6016, "pid": 5, "tid": 7, "ts": 1716454216592926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546220, "dur": 9,
    "args": {
      "External id": 6016,
      "cbid": 211, "correlation": 6016
    }
  },
  {
    "ph": "s", "id": 6016, "pid": 76337, "tid": -914061504, "ts": 1716454216546220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216592940, "dur": 30,
    "args": {
      "External id": 6036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6036,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6036, "pid": 5, "tid": 7, "ts": 1716454216592940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546289, "dur": 11,
    "args": {
      "External id": 6036,
      "cbid": 211, "correlation": 6036
    }
  },
  {
    "ph": "s", "id": 6036, "pid": 76337, "tid": -914061504, "ts": 1716454216546289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216592971, "dur": 4,
    "args": {
      "External id": 6048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6048,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6048, "pid": 5, "tid": 7, "ts": 1716454216592971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546310, "dur": 6,
    "args": {
      "External id": 6048,
      "cbid": 211, "correlation": 6048
    }
  },
  {
    "ph": "s", "id": 6048, "pid": 76337, "tid": -914061504, "ts": 1716454216546310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216592977, "dur": 13,
    "args": {
      "External id": 6051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6051, "pid": 5, "tid": 7, "ts": 1716454216592977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546326, "dur": 7,
    "args": {
      "External id": 6051,
      "cbid": 211, "correlation": 6051
    }
  },
  {
    "ph": "s", "id": 6051, "pid": 76337, "tid": -914061504, "ts": 1716454216546326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216592991, "dur": 9,
    "args": {
      "External id": 6060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6060,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6060, "pid": 5, "tid": 7, "ts": 1716454216592991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546367, "dur": 10,
    "args": {
      "External id": 6060,
      "cbid": 211, "correlation": 6060
    }
  },
  {
    "ph": "s", "id": 6060, "pid": 76337, "tid": -914061504, "ts": 1716454216546367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546427, "dur": 0,
    "args": {
      "External id": 6070,
      "cbid": 317, "correlation": 6070
    }
  },
  {
    "ph": "f", "id": 6070, "pid": 76337, "tid": -914061504, "ts": 1716454216546427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546428, "dur": 0,
    "args": {
      "External id": 6071,
      "cbid": 203, "correlation": 6071
    }
  },
  {
    "ph": "f", "id": 6071, "pid": 76337, "tid": -914061504, "ts": 1716454216546428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546429, "dur": 0,
    "args": {
      "External id": 6072,
      "cbid": 205, "correlation": 6072
    }
  },
  {
    "ph": "f", "id": 6072, "pid": 76337, "tid": -914061504, "ts": 1716454216546429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593002, "dur": 9,
    "args": {
      "External id": 6076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6076, "pid": 5, "tid": 7, "ts": 1716454216593002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546442, "dur": 13,
    "args": {
      "External id": 6076,
      "cbid": 211, "correlation": 6076
    }
  },
  {
    "ph": "s", "id": 6076, "pid": 76337, "tid": -914061504, "ts": 1716454216546442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593012, "dur": 32,
    "args": {
      "External id": 6078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6078,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6078, "pid": 5, "tid": 7, "ts": 1716454216593012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546457, "dur": 5,
    "args": {
      "External id": 6078,
      "cbid": 211, "correlation": 6078
    }
  },
  {
    "ph": "s", "id": 6078, "pid": 76337, "tid": -914061504, "ts": 1716454216546457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216593047, "dur": 1,
    "args": {
      "External id": 6080,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6080,
      "bytes": 768, "memory bandwidth (GB/s)": 0.4528301886792453
    }
  },
  {
    "ph": "f", "id": 6080, "pid": 5, "tid": 7, "ts": 1716454216593047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546468, "dur": 6,
    "args": {
      "External id": 6080,
      "cbid": 51, "correlation": 6080
    }
  },
  {
    "ph": "s", "id": 6080, "pid": 76337, "tid": -914061504, "ts": 1716454216546468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216593051, "dur": 264,
    "args": {
      "External id": 6081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6081,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6081, "pid": 5, "tid": 7, "ts": 1716454216593051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546475, "dur": 6,
    "args": {
      "External id": 6081,
      "cbid": 211, "correlation": 6081
    }
  },
  {
    "ph": "s", "id": 6081, "pid": 76337, "tid": -914061504, "ts": 1716454216546475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593317, "dur": 11,
    "args": {
      "External id": 6083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6083,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6083, "pid": 5, "tid": 7, "ts": 1716454216593317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546484, "dur": 5,
    "args": {
      "External id": 6083,
      "cbid": 211, "correlation": 6083
    }
  },
  {
    "ph": "s", "id": 6083, "pid": 76337, "tid": -914061504, "ts": 1716454216546484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216593329, "dur": 12,
    "args": {
      "External id": 6089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6089, "pid": 5, "tid": 7, "ts": 1716454216593329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546512, "dur": 9,
    "args": {
      "External id": 6089,
      "cbid": 211, "correlation": 6089
    }
  },
  {
    "ph": "s", "id": 6089, "pid": 76337, "tid": -914061504, "ts": 1716454216546512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216593342, "dur": 39,
    "args": {
      "External id": 6097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6097, "pid": 5, "tid": 7, "ts": 1716454216593342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546546, "dur": 9,
    "args": {
      "External id": 6097,
      "cbid": 211, "correlation": 6097
    }
  },
  {
    "ph": "s", "id": 6097, "pid": 76337, "tid": -914061504, "ts": 1716454216546546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216593382, "dur": 10,
    "args": {
      "External id": 6105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6105,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6105, "pid": 5, "tid": 7, "ts": 1716454216593382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546575, "dur": 8,
    "args": {
      "External id": 6105,
      "cbid": 211, "correlation": 6105
    }
  },
  {
    "ph": "s", "id": 6105, "pid": 76337, "tid": -914061504, "ts": 1716454216546575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216593393, "dur": 38,
    "args": {
      "External id": 6115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6115,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6115, "pid": 5, "tid": 7, "ts": 1716454216593393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546643, "dur": 12,
    "args": {
      "External id": 6115,
      "cbid": 211, "correlation": 6115
    }
  },
  {
    "ph": "s", "id": 6115, "pid": 76337, "tid": -914061504, "ts": 1716454216546643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593433, "dur": 49,
    "args": {
      "External id": 6136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6136,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6136, "pid": 5, "tid": 7, "ts": 1716454216593433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546676, "dur": 7,
    "args": {
      "External id": 6136,
      "cbid": 211, "correlation": 6136
    }
  },
  {
    "ph": "s", "id": 6136, "pid": 76337, "tid": -914061504, "ts": 1716454216546676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216593483, "dur": 4,
    "args": {
      "External id": 6148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6148,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6148, "pid": 5, "tid": 7, "ts": 1716454216593483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546692, "dur": 6,
    "args": {
      "External id": 6148,
      "cbid": 211, "correlation": 6148
    }
  },
  {
    "ph": "s", "id": 6148, "pid": 76337, "tid": -914061504, "ts": 1716454216546692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216593489, "dur": 14,
    "args": {
      "External id": 6151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6151, "pid": 5, "tid": 7, "ts": 1716454216593489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546708, "dur": 6,
    "args": {
      "External id": 6151,
      "cbid": 211, "correlation": 6151
    }
  },
  {
    "ph": "s", "id": 6151, "pid": 76337, "tid": -914061504, "ts": 1716454216546708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216593504, "dur": 9,
    "args": {
      "External id": 6160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6160,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6160, "pid": 5, "tid": 7, "ts": 1716454216593504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546746, "dur": 9,
    "args": {
      "External id": 6160,
      "cbid": 211, "correlation": 6160
    }
  },
  {
    "ph": "s", "id": 6160, "pid": 76337, "tid": -914061504, "ts": 1716454216546746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546796, "dur": 0,
    "args": {
      "External id": 6170,
      "cbid": 317, "correlation": 6170
    }
  },
  {
    "ph": "f", "id": 6170, "pid": 76337, "tid": -914061504, "ts": 1716454216546796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546797, "dur": 0,
    "args": {
      "External id": 6171,
      "cbid": 203, "correlation": 6171
    }
  },
  {
    "ph": "f", "id": 6171, "pid": 76337, "tid": -914061504, "ts": 1716454216546797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546797, "dur": 0,
    "args": {
      "External id": 6172,
      "cbid": 205, "correlation": 6172
    }
  },
  {
    "ph": "f", "id": 6172, "pid": 76337, "tid": -914061504, "ts": 1716454216546797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593515, "dur": 10,
    "args": {
      "External id": 6176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6176,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6176, "pid": 5, "tid": 7, "ts": 1716454216593515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546810, "dur": 12,
    "args": {
      "External id": 6176,
      "cbid": 211, "correlation": 6176
    }
  },
  {
    "ph": "s", "id": 6176, "pid": 76337, "tid": -914061504, "ts": 1716454216546810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593526, "dur": 32,
    "args": {
      "External id": 6178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6178, "pid": 5, "tid": 7, "ts": 1716454216593526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546824, "dur": 5,
    "args": {
      "External id": 6178,
      "cbid": 211, "correlation": 6178
    }
  },
  {
    "ph": "s", "id": 6178, "pid": 76337, "tid": -914061504, "ts": 1716454216546824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216593560, "dur": 1,
    "args": {
      "External id": 6180,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6180,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 6180, "pid": 5, "tid": 7, "ts": 1716454216593560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546835, "dur": 6,
    "args": {
      "External id": 6180,
      "cbid": 51, "correlation": 6180
    }
  },
  {
    "ph": "s", "id": 6180, "pid": 76337, "tid": -914061504, "ts": 1716454216546835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216593564, "dur": 314,
    "args": {
      "External id": 6181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6181,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6181, "pid": 5, "tid": 7, "ts": 1716454216593564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546842, "dur": 6,
    "args": {
      "External id": 6181,
      "cbid": 211, "correlation": 6181
    }
  },
  {
    "ph": "s", "id": 6181, "pid": 76337, "tid": -914061504, "ts": 1716454216546842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593880, "dur": 10,
    "args": {
      "External id": 6183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6183,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6183, "pid": 5, "tid": 7, "ts": 1716454216593880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546852, "dur": 5,
    "args": {
      "External id": 6183,
      "cbid": 211, "correlation": 6183
    }
  },
  {
    "ph": "s", "id": 6183, "pid": 76337, "tid": -914061504, "ts": 1716454216546852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216593891, "dur": 12,
    "args": {
      "External id": 6189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6189, "pid": 5, "tid": 7, "ts": 1716454216593891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546879, "dur": 9,
    "args": {
      "External id": 6189,
      "cbid": 211, "correlation": 6189
    }
  },
  {
    "ph": "s", "id": 6189, "pid": 76337, "tid": -914061504, "ts": 1716454216546879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593905, "dur": 30,
    "args": {
      "External id": 6209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6209,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6209, "pid": 5, "tid": 7, "ts": 1716454216593905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546950, "dur": 11,
    "args": {
      "External id": 6209,
      "cbid": 211, "correlation": 6209
    }
  },
  {
    "ph": "s", "id": 6209, "pid": 76337, "tid": -914061504, "ts": 1716454216546950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216593937, "dur": 4,
    "args": {
      "External id": 6221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6221,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6221, "pid": 5, "tid": 7, "ts": 1716454216593937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546971, "dur": 16,
    "args": {
      "External id": 6221,
      "cbid": 211, "correlation": 6221
    }
  },
  {
    "ph": "s", "id": 6221, "pid": 76337, "tid": -914061504, "ts": 1716454216546971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216593942, "dur": 13,
    "args": {
      "External id": 6224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6224,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6224, "pid": 5, "tid": 7, "ts": 1716454216593942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216546999, "dur": 7,
    "args": {
      "External id": 6224,
      "cbid": 211, "correlation": 6224
    }
  },
  {
    "ph": "s", "id": 6224, "pid": 76337, "tid": -914061504, "ts": 1716454216546999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216593957, "dur": 9,
    "args": {
      "External id": 6233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6233,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6233, "pid": 5, "tid": 7, "ts": 1716454216593957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547040, "dur": 10,
    "args": {
      "External id": 6233,
      "cbid": 211, "correlation": 6233
    }
  },
  {
    "ph": "s", "id": 6233, "pid": 76337, "tid": -914061504, "ts": 1716454216547040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547101, "dur": 0,
    "args": {
      "External id": 6243,
      "cbid": 317, "correlation": 6243
    }
  },
  {
    "ph": "f", "id": 6243, "pid": 76337, "tid": -914061504, "ts": 1716454216547101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547102, "dur": 0,
    "args": {
      "External id": 6244,
      "cbid": 203, "correlation": 6244
    }
  },
  {
    "ph": "f", "id": 6244, "pid": 76337, "tid": -914061504, "ts": 1716454216547102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547102, "dur": 0,
    "args": {
      "External id": 6245,
      "cbid": 205, "correlation": 6245
    }
  },
  {
    "ph": "f", "id": 6245, "pid": 76337, "tid": -914061504, "ts": 1716454216547102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593967, "dur": 9,
    "args": {
      "External id": 6249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6249,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6249, "pid": 5, "tid": 7, "ts": 1716454216593967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547116, "dur": 11,
    "args": {
      "External id": 6249,
      "cbid": 211, "correlation": 6249
    }
  },
  {
    "ph": "s", "id": 6249, "pid": 76337, "tid": -914061504, "ts": 1716454216547116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216593978, "dur": 32,
    "args": {
      "External id": 6251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6251,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6251, "pid": 5, "tid": 7, "ts": 1716454216593978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547129, "dur": 5,
    "args": {
      "External id": 6251,
      "cbid": 211, "correlation": 6251
    }
  },
  {
    "ph": "s", "id": 6251, "pid": 76337, "tid": -914061504, "ts": 1716454216547129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216594012, "dur": 1,
    "args": {
      "External id": 6253,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6253,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 6253, "pid": 5, "tid": 7, "ts": 1716454216594012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547140, "dur": 6,
    "args": {
      "External id": 6253,
      "cbid": 51, "correlation": 6253
    }
  },
  {
    "ph": "s", "id": 6253, "pid": 76337, "tid": -914061504, "ts": 1716454216547140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216594016, "dur": 266,
    "args": {
      "External id": 6254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6254,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6254, "pid": 5, "tid": 7, "ts": 1716454216594016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547147, "dur": 6,
    "args": {
      "External id": 6254,
      "cbid": 211, "correlation": 6254
    }
  },
  {
    "ph": "s", "id": 6254, "pid": 76337, "tid": -914061504, "ts": 1716454216547147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594283, "dur": 10,
    "args": {
      "External id": 6256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6256,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6256, "pid": 5, "tid": 7, "ts": 1716454216594283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547157, "dur": 6,
    "args": {
      "External id": 6256,
      "cbid": 211, "correlation": 6256
    }
  },
  {
    "ph": "s", "id": 6256, "pid": 76337, "tid": -914061504, "ts": 1716454216547157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216594295, "dur": 12,
    "args": {
      "External id": 6262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6262,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6262, "pid": 5, "tid": 7, "ts": 1716454216594295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547185, "dur": 8,
    "args": {
      "External id": 6262,
      "cbid": 211, "correlation": 6262
    }
  },
  {
    "ph": "s", "id": 6262, "pid": 76337, "tid": -914061504, "ts": 1716454216547185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216594308, "dur": 39,
    "args": {
      "External id": 6270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6270, "pid": 5, "tid": 7, "ts": 1716454216594308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547219, "dur": 9,
    "args": {
      "External id": 6270,
      "cbid": 211, "correlation": 6270
    }
  },
  {
    "ph": "s", "id": 6270, "pid": 76337, "tid": -914061504, "ts": 1716454216547219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216594348, "dur": 9,
    "args": {
      "External id": 6278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6278,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6278, "pid": 5, "tid": 7, "ts": 1716454216594348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547248, "dur": 8,
    "args": {
      "External id": 6278,
      "cbid": 211, "correlation": 6278
    }
  },
  {
    "ph": "s", "id": 6278, "pid": 76337, "tid": -914061504, "ts": 1716454216547248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216594359, "dur": 39,
    "args": {
      "External id": 6288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6288, "pid": 5, "tid": 7, "ts": 1716454216594359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547331, "dur": 13,
    "args": {
      "External id": 6288,
      "cbid": 211, "correlation": 6288
    }
  },
  {
    "ph": "s", "id": 6288, "pid": 76337, "tid": -914061504, "ts": 1716454216547331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594399, "dur": 34,
    "args": {
      "External id": 6309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6309,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6309, "pid": 5, "tid": 7, "ts": 1716454216594399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547364, "dur": 8,
    "args": {
      "External id": 6309,
      "cbid": 211, "correlation": 6309
    }
  },
  {
    "ph": "s", "id": 6309, "pid": 76337, "tid": -914061504, "ts": 1716454216547364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216594435, "dur": 4,
    "args": {
      "External id": 6321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6321,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6321, "pid": 5, "tid": 7, "ts": 1716454216594435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547381, "dur": 6,
    "args": {
      "External id": 6321,
      "cbid": 211, "correlation": 6321
    }
  },
  {
    "ph": "s", "id": 6321, "pid": 76337, "tid": -914061504, "ts": 1716454216547381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216594441, "dur": 14,
    "args": {
      "External id": 6324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6324, "pid": 5, "tid": 7, "ts": 1716454216594441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547397, "dur": 6,
    "args": {
      "External id": 6324,
      "cbid": 211, "correlation": 6324
    }
  },
  {
    "ph": "s", "id": 6324, "pid": 76337, "tid": -914061504, "ts": 1716454216547397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216594456, "dur": 9,
    "args": {
      "External id": 6333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6333,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6333, "pid": 5, "tid": 7, "ts": 1716454216594456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547437, "dur": 9,
    "args": {
      "External id": 6333,
      "cbid": 211, "correlation": 6333
    }
  },
  {
    "ph": "s", "id": 6333, "pid": 76337, "tid": -914061504, "ts": 1716454216547437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547488, "dur": 0,
    "args": {
      "External id": 6343,
      "cbid": 317, "correlation": 6343
    }
  },
  {
    "ph": "f", "id": 6343, "pid": 76337, "tid": -914061504, "ts": 1716454216547488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547489, "dur": 0,
    "args": {
      "External id": 6344,
      "cbid": 203, "correlation": 6344
    }
  },
  {
    "ph": "f", "id": 6344, "pid": 76337, "tid": -914061504, "ts": 1716454216547489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547489, "dur": 0,
    "args": {
      "External id": 6345,
      "cbid": 205, "correlation": 6345
    }
  },
  {
    "ph": "f", "id": 6345, "pid": 76337, "tid": -914061504, "ts": 1716454216547489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594466, "dur": 8,
    "args": {
      "External id": 6349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6349, "pid": 5, "tid": 7, "ts": 1716454216594466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547502, "dur": 11,
    "args": {
      "External id": 6349,
      "cbid": 211, "correlation": 6349
    }
  },
  {
    "ph": "s", "id": 6349, "pid": 76337, "tid": -914061504, "ts": 1716454216547502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594476, "dur": 32,
    "args": {
      "External id": 6351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6351,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6351, "pid": 5, "tid": 7, "ts": 1716454216594476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547515, "dur": 5,
    "args": {
      "External id": 6351,
      "cbid": 211, "correlation": 6351
    }
  },
  {
    "ph": "s", "id": 6351, "pid": 76337, "tid": -914061504, "ts": 1716454216547515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216594511, "dur": 1,
    "args": {
      "External id": 6353,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6353,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 6353, "pid": 5, "tid": 7, "ts": 1716454216594511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547526, "dur": 7,
    "args": {
      "External id": 6353,
      "cbid": 51, "correlation": 6353
    }
  },
  {
    "ph": "s", "id": 6353, "pid": 76337, "tid": -914061504, "ts": 1716454216547526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216594515, "dur": 266,
    "args": {
      "External id": 6354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6354,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6354, "pid": 5, "tid": 7, "ts": 1716454216594515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547534, "dur": 6,
    "args": {
      "External id": 6354,
      "cbid": 211, "correlation": 6354
    }
  },
  {
    "ph": "s", "id": 6354, "pid": 76337, "tid": -914061504, "ts": 1716454216547534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594782, "dur": 10,
    "args": {
      "External id": 6356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6356, "pid": 5, "tid": 7, "ts": 1716454216594782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547544, "dur": 5,
    "args": {
      "External id": 6356,
      "cbid": 211, "correlation": 6356
    }
  },
  {
    "ph": "s", "id": 6356, "pid": 76337, "tid": -914061504, "ts": 1716454216547544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216594793, "dur": 13,
    "args": {
      "External id": 6362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6362, "pid": 5, "tid": 7, "ts": 1716454216594793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547571, "dur": 9,
    "args": {
      "External id": 6362,
      "cbid": 211, "correlation": 6362
    }
  },
  {
    "ph": "s", "id": 6362, "pid": 76337, "tid": -914061504, "ts": 1716454216547571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594808, "dur": 30,
    "args": {
      "External id": 6382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6382,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6382, "pid": 5, "tid": 7, "ts": 1716454216594808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547641, "dur": 11,
    "args": {
      "External id": 6382,
      "cbid": 211, "correlation": 6382
    }
  },
  {
    "ph": "s", "id": 6382, "pid": 76337, "tid": -914061504, "ts": 1716454216547641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216594840, "dur": 4,
    "args": {
      "External id": 6394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6394,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6394, "pid": 5, "tid": 7, "ts": 1716454216594840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547661, "dur": 6,
    "args": {
      "External id": 6394,
      "cbid": 211, "correlation": 6394
    }
  },
  {
    "ph": "s", "id": 6394, "pid": 76337, "tid": -914061504, "ts": 1716454216547661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216594845, "dur": 14,
    "args": {
      "External id": 6397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6397, "pid": 5, "tid": 7, "ts": 1716454216594845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547678, "dur": 6,
    "args": {
      "External id": 6397,
      "cbid": 211, "correlation": 6397
    }
  },
  {
    "ph": "s", "id": 6397, "pid": 76337, "tid": -914061504, "ts": 1716454216547678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216594860, "dur": 9,
    "args": {
      "External id": 6406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6406,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6406, "pid": 5, "tid": 7, "ts": 1716454216594860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547717, "dur": 10,
    "args": {
      "External id": 6406,
      "cbid": 211, "correlation": 6406
    }
  },
  {
    "ph": "s", "id": 6406, "pid": 76337, "tid": -914061504, "ts": 1716454216547717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547778, "dur": 0,
    "args": {
      "External id": 6416,
      "cbid": 317, "correlation": 6416
    }
  },
  {
    "ph": "f", "id": 6416, "pid": 76337, "tid": -914061504, "ts": 1716454216547778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547778, "dur": 0,
    "args": {
      "External id": 6417,
      "cbid": 203, "correlation": 6417
    }
  },
  {
    "ph": "f", "id": 6417, "pid": 76337, "tid": -914061504, "ts": 1716454216547778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547779, "dur": 0,
    "args": {
      "External id": 6418,
      "cbid": 205, "correlation": 6418
    }
  },
  {
    "ph": "f", "id": 6418, "pid": 76337, "tid": -914061504, "ts": 1716454216547779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594871, "dur": 9,
    "args": {
      "External id": 6422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6422, "pid": 5, "tid": 7, "ts": 1716454216594871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547792, "dur": 12,
    "args": {
      "External id": 6422,
      "cbid": 211, "correlation": 6422
    }
  },
  {
    "ph": "s", "id": 6422, "pid": 76337, "tid": -914061504, "ts": 1716454216547792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216594881, "dur": 32,
    "args": {
      "External id": 6424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6424,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6424, "pid": 5, "tid": 7, "ts": 1716454216594881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547806, "dur": 5,
    "args": {
      "External id": 6424,
      "cbid": 211, "correlation": 6424
    }
  },
  {
    "ph": "s", "id": 6424, "pid": 76337, "tid": -914061504, "ts": 1716454216547806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216594916, "dur": 1,
    "args": {
      "External id": 6426,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6426,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 6426, "pid": 5, "tid": 7, "ts": 1716454216594916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547817, "dur": 6,
    "args": {
      "External id": 6426,
      "cbid": 51, "correlation": 6426
    }
  },
  {
    "ph": "s", "id": 6426, "pid": 76337, "tid": -914061504, "ts": 1716454216547817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216594920, "dur": 265,
    "args": {
      "External id": 6427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6427,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6427, "pid": 5, "tid": 7, "ts": 1716454216594920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547824, "dur": 6,
    "args": {
      "External id": 6427,
      "cbid": 211, "correlation": 6427
    }
  },
  {
    "ph": "s", "id": 6427, "pid": 76337, "tid": -914061504, "ts": 1716454216547824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216595187, "dur": 10,
    "args": {
      "External id": 6429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6429,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6429, "pid": 5, "tid": 7, "ts": 1716454216595187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547834, "dur": 5,
    "args": {
      "External id": 6429,
      "cbid": 211, "correlation": 6429
    }
  },
  {
    "ph": "s", "id": 6429, "pid": 76337, "tid": -914061504, "ts": 1716454216547834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595199, "dur": 12,
    "args": {
      "External id": 6435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6435, "pid": 5, "tid": 7, "ts": 1716454216595199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547862, "dur": 9,
    "args": {
      "External id": 6435,
      "cbid": 211, "correlation": 6435
    }
  },
  {
    "ph": "s", "id": 6435, "pid": 76337, "tid": -914061504, "ts": 1716454216547862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595212, "dur": 40,
    "args": {
      "External id": 6443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6443, "pid": 5, "tid": 7, "ts": 1716454216595212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547896, "dur": 9,
    "args": {
      "External id": 6443,
      "cbid": 211, "correlation": 6443
    }
  },
  {
    "ph": "s", "id": 6443, "pid": 76337, "tid": -914061504, "ts": 1716454216547896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216595254, "dur": 10,
    "args": {
      "External id": 6451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6451,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6451, "pid": 5, "tid": 7, "ts": 1716454216595254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216547925, "dur": 8,
    "args": {
      "External id": 6451,
      "cbid": 211, "correlation": 6451
    }
  },
  {
    "ph": "s", "id": 6451, "pid": 76337, "tid": -914061504, "ts": 1716454216547925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595265, "dur": 38,
    "args": {
      "External id": 6461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6461,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6461, "pid": 5, "tid": 7, "ts": 1716454216595265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548079, "dur": 15,
    "args": {
      "External id": 6461,
      "cbid": 211, "correlation": 6461
    }
  },
  {
    "ph": "s", "id": 6461, "pid": 76337, "tid": -914061504, "ts": 1716454216548079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216595304, "dur": 31,
    "args": {
      "External id": 6482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6482,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6482, "pid": 5, "tid": 7, "ts": 1716454216595304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548115, "dur": 7,
    "args": {
      "External id": 6482,
      "cbid": 211, "correlation": 6482
    }
  },
  {
    "ph": "s", "id": 6482, "pid": 76337, "tid": -914061504, "ts": 1716454216548115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216595337, "dur": 4,
    "args": {
      "External id": 6494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6494,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6494, "pid": 5, "tid": 7, "ts": 1716454216595337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548133, "dur": 6,
    "args": {
      "External id": 6494,
      "cbid": 211, "correlation": 6494
    }
  },
  {
    "ph": "s", "id": 6494, "pid": 76337, "tid": -914061504, "ts": 1716454216548133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595343, "dur": 14,
    "args": {
      "External id": 6497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6497, "pid": 5, "tid": 7, "ts": 1716454216595343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548149, "dur": 6,
    "args": {
      "External id": 6497,
      "cbid": 211, "correlation": 6497
    }
  },
  {
    "ph": "s", "id": 6497, "pid": 76337, "tid": -914061504, "ts": 1716454216548149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548246, "dur": 1,
    "args": {
      "External id": 6509,
      "cbid": 251, "correlation": 6509
    }
  },
  {
    "ph": "f", "id": 6509, "pid": 76337, "tid": -914061504, "ts": 1716454216548246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216595358, "dur": 38,
    "args": {
      "External id": 6510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6510,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 6510, "pid": 5, "tid": 7, "ts": 1716454216595358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548252, "dur": 13,
    "args": {
      "External id": 6510,
      "cbid": 211, "correlation": 6510
    }
  },
  {
    "ph": "s", "id": 6510, "pid": 76337, "tid": -914061504, "ts": 1716454216548252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595398, "dur": 15,
    "args": {
      "External id": 6515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6515,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6515, "pid": 5, "tid": 7, "ts": 1716454216595398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548284, "dur": 9,
    "args": {
      "External id": 6515,
      "cbid": 211, "correlation": 6515
    }
  },
  {
    "ph": "s", "id": 6515, "pid": 76337, "tid": -914061504, "ts": 1716454216548284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548358, "dur": 1,
    "args": {
      "External id": 6526,
      "cbid": 251, "correlation": 6526
    }
  },
  {
    "ph": "f", "id": 6526, "pid": 76337, "tid": -914061504, "ts": 1716454216548358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216595414, "dur": 36,
    "args": {
      "External id": 6527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6527,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 6527, "pid": 5, "tid": 7, "ts": 1716454216595414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548363, "dur": 12,
    "args": {
      "External id": 6527,
      "cbid": 211, "correlation": 6527
    }
  },
  {
    "ph": "s", "id": 6527, "pid": 76337, "tid": -914061504, "ts": 1716454216548363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595451, "dur": 14,
    "args": {
      "External id": 6532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6532, "pid": 5, "tid": 7, "ts": 1716454216595451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548391, "dur": 8,
    "args": {
      "External id": 6532,
      "cbid": 211, "correlation": 6532
    }
  },
  {
    "ph": "s", "id": 6532, "pid": 76337, "tid": -914061504, "ts": 1716454216548391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548462, "dur": 1,
    "args": {
      "External id": 6543,
      "cbid": 251, "correlation": 6543
    }
  },
  {
    "ph": "f", "id": 6543, "pid": 76337, "tid": -914061504, "ts": 1716454216548462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454216595467, "dur": 36,
    "args": {
      "External id": 6544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6544,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 6544, "pid": 5, "tid": 7, "ts": 1716454216595467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548467, "dur": 12,
    "args": {
      "External id": 6544,
      "cbid": 211, "correlation": 6544
    }
  },
  {
    "ph": "s", "id": 6544, "pid": 76337, "tid": -914061504, "ts": 1716454216548467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216595504, "dur": 14,
    "args": {
      "External id": 6549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6549,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6549, "pid": 5, "tid": 7, "ts": 1716454216595504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548495, "dur": 9,
    "args": {
      "External id": 6549,
      "cbid": 211, "correlation": 6549
    }
  },
  {
    "ph": "s", "id": 6549, "pid": 76337, "tid": -914061504, "ts": 1716454216548495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216595520, "dur": 1210,
    "args": {
      "External id": 6574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6574,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 6574, "pid": 5, "tid": 7, "ts": 1716454216595520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548576, "dur": 12,
    "args": {
      "External id": 6574,
      "cbid": 211, "correlation": 6574
    }
  },
  {
    "ph": "s", "id": 6574, "pid": 76337, "tid": -914061504, "ts": 1716454216548576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548682, "dur": 1,
    "args": {
      "External id": 6592,
      "cbid": 251, "correlation": 6592
    }
  },
  {
    "ph": "f", "id": 6592, "pid": 76337, "tid": -914061504, "ts": 1716454216548682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216596731, "dur": 42,
    "args": {
      "External id": 6594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6594,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [8, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 6594, "pid": 5, "tid": 7, "ts": 1716454216596731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548688, "dur": 14,
    "args": {
      "External id": 6594,
      "cbid": 211, "correlation": 6594
    }
  },
  {
    "ph": "s", "id": 6594, "pid": 76337, "tid": -914061504, "ts": 1716454216548688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216596775, "dur": 12,
    "args": {
      "External id": 6602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6602,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6602, "pid": 5, "tid": 7, "ts": 1716454216596775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548771, "dur": 13,
    "args": {
      "External id": 6602,
      "cbid": 211, "correlation": 6602
    }
  },
  {
    "ph": "s", "id": 6602, "pid": 76337, "tid": -914061504, "ts": 1716454216548771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216596789, "dur": 9,
    "args": {
      "External id": 6610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6610,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6610, "pid": 5, "tid": 7, "ts": 1716454216596789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548809, "dur": 9,
    "args": {
      "External id": 6610,
      "cbid": 211, "correlation": 6610
    }
  },
  {
    "ph": "s", "id": 6610, "pid": 76337, "tid": -914061504, "ts": 1716454216548809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216596800, "dur": 38,
    "args": {
      "External id": 6620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6620,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6620, "pid": 5, "tid": 7, "ts": 1716454216596800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548882, "dur": 12,
    "args": {
      "External id": 6620,
      "cbid": 211, "correlation": 6620
    }
  },
  {
    "ph": "s", "id": 6620, "pid": 76337, "tid": -914061504, "ts": 1716454216548882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216596839, "dur": 33,
    "args": {
      "External id": 6641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6641,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6641, "pid": 5, "tid": 7, "ts": 1716454216596839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548916, "dur": 8,
    "args": {
      "External id": 6641,
      "cbid": 211, "correlation": 6641
    }
  },
  {
    "ph": "s", "id": 6641, "pid": 76337, "tid": -914061504, "ts": 1716454216548916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216596874, "dur": 4,
    "args": {
      "External id": 6653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6653,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6653, "pid": 5, "tid": 7, "ts": 1716454216596874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548933, "dur": 6,
    "args": {
      "External id": 6653,
      "cbid": 211, "correlation": 6653
    }
  },
  {
    "ph": "s", "id": 6653, "pid": 76337, "tid": -914061504, "ts": 1716454216548933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216596879, "dur": 14,
    "args": {
      "External id": 6656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6656,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6656, "pid": 5, "tid": 7, "ts": 1716454216596879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548949, "dur": 6,
    "args": {
      "External id": 6656,
      "cbid": 211, "correlation": 6656
    }
  },
  {
    "ph": "s", "id": 6656, "pid": 76337, "tid": -914061504, "ts": 1716454216548949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216596895, "dur": 10,
    "args": {
      "External id": 6665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6665,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6665, "pid": 5, "tid": 7, "ts": 1716454216596895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216548997, "dur": 10,
    "args": {
      "External id": 6665,
      "cbid": 211, "correlation": 6665
    }
  },
  {
    "ph": "s", "id": 6665, "pid": 76337, "tid": -914061504, "ts": 1716454216548997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549050, "dur": 0,
    "args": {
      "External id": 6675,
      "cbid": 317, "correlation": 6675
    }
  },
  {
    "ph": "f", "id": 6675, "pid": 76337, "tid": -914061504, "ts": 1716454216549050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549051, "dur": 0,
    "args": {
      "External id": 6676,
      "cbid": 203, "correlation": 6676
    }
  },
  {
    "ph": "f", "id": 6676, "pid": 76337, "tid": -914061504, "ts": 1716454216549051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549052, "dur": 0,
    "args": {
      "External id": 6677,
      "cbid": 205, "correlation": 6677
    }
  },
  {
    "ph": "f", "id": 6677, "pid": 76337, "tid": -914061504, "ts": 1716454216549052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216596907, "dur": 10,
    "args": {
      "External id": 6681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6681,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6681, "pid": 5, "tid": 7, "ts": 1716454216596907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549066, "dur": 12,
    "args": {
      "External id": 6681,
      "cbid": 211, "correlation": 6681
    }
  },
  {
    "ph": "s", "id": 6681, "pid": 76337, "tid": -914061504, "ts": 1716454216549066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216596919, "dur": 33,
    "args": {
      "External id": 6683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6683,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6683, "pid": 5, "tid": 7, "ts": 1716454216596919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549080, "dur": 5,
    "args": {
      "External id": 6683,
      "cbid": 211, "correlation": 6683
    }
  },
  {
    "ph": "s", "id": 6683, "pid": 76337, "tid": -914061504, "ts": 1716454216549080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216596954, "dur": 1,
    "args": {
      "External id": 6685,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6685,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 6685, "pid": 5, "tid": 7, "ts": 1716454216596954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549091, "dur": 7,
    "args": {
      "External id": 6685,
      "cbid": 51, "correlation": 6685
    }
  },
  {
    "ph": "s", "id": 6685, "pid": 76337, "tid": -914061504, "ts": 1716454216549091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216596957, "dur": 273,
    "args": {
      "External id": 6686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6686,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6686, "pid": 5, "tid": 7, "ts": 1716454216596957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549099, "dur": 6,
    "args": {
      "External id": 6686,
      "cbid": 211, "correlation": 6686
    }
  },
  {
    "ph": "s", "id": 6686, "pid": 76337, "tid": -914061504, "ts": 1716454216549099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597232, "dur": 11,
    "args": {
      "External id": 6688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6688,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6688, "pid": 5, "tid": 7, "ts": 1716454216597232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549108, "dur": 5,
    "args": {
      "External id": 6688,
      "cbid": 211, "correlation": 6688
    }
  },
  {
    "ph": "s", "id": 6688, "pid": 76337, "tid": -914061504, "ts": 1716454216549108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597244, "dur": 14,
    "args": {
      "External id": 6694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6694, "pid": 5, "tid": 7, "ts": 1716454216597244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549137, "dur": 8,
    "args": {
      "External id": 6694,
      "cbid": 211, "correlation": 6694
    }
  },
  {
    "ph": "s", "id": 6694, "pid": 76337, "tid": -914061504, "ts": 1716454216549137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597259, "dur": 30,
    "args": {
      "External id": 6714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6714,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6714, "pid": 5, "tid": 7, "ts": 1716454216597259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549204, "dur": 11,
    "args": {
      "External id": 6714,
      "cbid": 211, "correlation": 6714
    }
  },
  {
    "ph": "s", "id": 6714, "pid": 76337, "tid": -914061504, "ts": 1716454216549204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216597291, "dur": 4,
    "args": {
      "External id": 6726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6726,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6726, "pid": 5, "tid": 7, "ts": 1716454216597291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549225, "dur": 7,
    "args": {
      "External id": 6726,
      "cbid": 211, "correlation": 6726
    }
  },
  {
    "ph": "s", "id": 6726, "pid": 76337, "tid": -914061504, "ts": 1716454216549225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597296, "dur": 13,
    "args": {
      "External id": 6729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6729,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6729, "pid": 5, "tid": 7, "ts": 1716454216597296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549243, "dur": 6,
    "args": {
      "External id": 6729,
      "cbid": 211, "correlation": 6729
    }
  },
  {
    "ph": "s", "id": 6729, "pid": 76337, "tid": -914061504, "ts": 1716454216549243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597311, "dur": 9,
    "args": {
      "External id": 6738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6738,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6738, "pid": 5, "tid": 7, "ts": 1716454216597311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549281, "dur": 9,
    "args": {
      "External id": 6738,
      "cbid": 211, "correlation": 6738
    }
  },
  {
    "ph": "s", "id": 6738, "pid": 76337, "tid": -914061504, "ts": 1716454216549281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549340, "dur": 0,
    "args": {
      "External id": 6748,
      "cbid": 317, "correlation": 6748
    }
  },
  {
    "ph": "f", "id": 6748, "pid": 76337, "tid": -914061504, "ts": 1716454216549340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549341, "dur": 0,
    "args": {
      "External id": 6749,
      "cbid": 203, "correlation": 6749
    }
  },
  {
    "ph": "f", "id": 6749, "pid": 76337, "tid": -914061504, "ts": 1716454216549341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549342, "dur": 0,
    "args": {
      "External id": 6750,
      "cbid": 205, "correlation": 6750
    }
  },
  {
    "ph": "f", "id": 6750, "pid": 76337, "tid": -914061504, "ts": 1716454216549342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597322, "dur": 9,
    "args": {
      "External id": 6754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6754,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6754, "pid": 5, "tid": 7, "ts": 1716454216597322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549354, "dur": 12,
    "args": {
      "External id": 6754,
      "cbid": 211, "correlation": 6754
    }
  },
  {
    "ph": "s", "id": 6754, "pid": 76337, "tid": -914061504, "ts": 1716454216549354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597332, "dur": 32,
    "args": {
      "External id": 6756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6756,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6756, "pid": 5, "tid": 7, "ts": 1716454216597332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549368, "dur": 5,
    "args": {
      "External id": 6756,
      "cbid": 211, "correlation": 6756
    }
  },
  {
    "ph": "s", "id": 6756, "pid": 76337, "tid": -914061504, "ts": 1716454216549368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216597366, "dur": 1,
    "args": {
      "External id": 6758,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6758,
      "bytes": 768, "memory bandwidth (GB/s)": 0.46153846153846156
    }
  },
  {
    "ph": "f", "id": 6758, "pid": 5, "tid": 7, "ts": 1716454216597366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549379, "dur": 6,
    "args": {
      "External id": 6758,
      "cbid": 51, "correlation": 6758
    }
  },
  {
    "ph": "s", "id": 6758, "pid": 76337, "tid": -914061504, "ts": 1716454216549379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216597370, "dur": 266,
    "args": {
      "External id": 6759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6759,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [4, 24, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 6759, "pid": 5, "tid": 7, "ts": 1716454216597370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549386, "dur": 6,
    "args": {
      "External id": 6759,
      "cbid": 211, "correlation": 6759
    }
  },
  {
    "ph": "s", "id": 6759, "pid": 76337, "tid": -914061504, "ts": 1716454216549386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597637, "dur": 11,
    "args": {
      "External id": 6761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6761, "pid": 5, "tid": 7, "ts": 1716454216597637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549396, "dur": 5,
    "args": {
      "External id": 6761,
      "cbid": 211, "correlation": 6761
    }
  },
  {
    "ph": "s", "id": 6761, "pid": 76337, "tid": -914061504, "ts": 1716454216549396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597649, "dur": 12,
    "args": {
      "External id": 6767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6767, "pid": 5, "tid": 7, "ts": 1716454216597649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549423, "dur": 8,
    "args": {
      "External id": 6767,
      "cbid": 211, "correlation": 6767
    }
  },
  {
    "ph": "s", "id": 6767, "pid": 76337, "tid": -914061504, "ts": 1716454216549423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597663, "dur": 40,
    "args": {
      "External id": 6775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6775, "pid": 5, "tid": 7, "ts": 1716454216597663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549457, "dur": 9,
    "args": {
      "External id": 6775,
      "cbid": 211, "correlation": 6775
    }
  },
  {
    "ph": "s", "id": 6775, "pid": 76337, "tid": -914061504, "ts": 1716454216549457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597704, "dur": 10,
    "args": {
      "External id": 6783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6783,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6783, "pid": 5, "tid": 7, "ts": 1716454216597704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549487, "dur": 9,
    "args": {
      "External id": 6783,
      "cbid": 211, "correlation": 6783
    }
  },
  {
    "ph": "s", "id": 6783, "pid": 76337, "tid": -914061504, "ts": 1716454216549487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597716, "dur": 38,
    "args": {
      "External id": 6793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6793,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6793, "pid": 5, "tid": 7, "ts": 1716454216597716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549557, "dur": 12,
    "args": {
      "External id": 6793,
      "cbid": 211, "correlation": 6793
    }
  },
  {
    "ph": "s", "id": 6793, "pid": 76337, "tid": -914061504, "ts": 1716454216549557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597755, "dur": 47,
    "args": {
      "External id": 6814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6814,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 0.4,
      "warps per SM": 6.4,
      "grid": [32, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 6814, "pid": 5, "tid": 7, "ts": 1716454216597755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549589, "dur": 7,
    "args": {
      "External id": 6814,
      "cbid": 211, "correlation": 6814
    }
  },
  {
    "ph": "s", "id": 6814, "pid": 76337, "tid": -914061504, "ts": 1716454216549589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216597804, "dur": 5,
    "args": {
      "External id": 6826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6826,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 6826, "pid": 5, "tid": 7, "ts": 1716454216597804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549605, "dur": 6,
    "args": {
      "External id": 6826,
      "cbid": 211, "correlation": 6826
    }
  },
  {
    "ph": "s", "id": 6826, "pid": 76337, "tid": -914061504, "ts": 1716454216549605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597809, "dur": 14,
    "args": {
      "External id": 6829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6829, "pid": 5, "tid": 7, "ts": 1716454216597809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549622, "dur": 6,
    "args": {
      "External id": 6829,
      "cbid": 211, "correlation": 6829
    }
  },
  {
    "ph": "s", "id": 6829, "pid": 76337, "tid": -914061504, "ts": 1716454216549622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597825, "dur": 9,
    "args": {
      "External id": 6838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6838,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6838, "pid": 5, "tid": 7, "ts": 1716454216597825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549664, "dur": 10,
    "args": {
      "External id": 6838,
      "cbid": 211, "correlation": 6838
    }
  },
  {
    "ph": "s", "id": 6838, "pid": 76337, "tid": -914061504, "ts": 1716454216549664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549714, "dur": 0,
    "args": {
      "External id": 6848,
      "cbid": 317, "correlation": 6848
    }
  },
  {
    "ph": "f", "id": 6848, "pid": 76337, "tid": -914061504, "ts": 1716454216549714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549715, "dur": 0,
    "args": {
      "External id": 6849,
      "cbid": 203, "correlation": 6849
    }
  },
  {
    "ph": "f", "id": 6849, "pid": 76337, "tid": -914061504, "ts": 1716454216549715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549715, "dur": 0,
    "args": {
      "External id": 6850,
      "cbid": 205, "correlation": 6850
    }
  },
  {
    "ph": "f", "id": 6850, "pid": 76337, "tid": -914061504, "ts": 1716454216549715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597835, "dur": 8,
    "args": {
      "External id": 6854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6854,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [96, 16, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6854, "pid": 5, "tid": 7, "ts": 1716454216597835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549730, "dur": 11,
    "args": {
      "External id": 6854,
      "cbid": 211, "correlation": 6854
    }
  },
  {
    "ph": "s", "id": 6854, "pid": 76337, "tid": -914061504, "ts": 1716454216549730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597844, "dur": 5,
    "args": {
      "External id": 6856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1.6,
      "warps per SM": 12.8,
      "grid": [1, 16, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 20
    }
  },
  {
    "ph": "f", "id": 6856, "pid": 5, "tid": 7, "ts": 1716454216597844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549744, "dur": 5,
    "args": {
      "External id": 6856,
      "cbid": 211, "correlation": 6856
    }
  },
  {
    "ph": "s", "id": 6856, "pid": 76337, "tid": -914061504, "ts": 1716454216549744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216597851, "dur": 1,
    "args": {
      "External id": 6858,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6858,
      "bytes": 2688, "memory bandwidth (GB/s)": 1.4
    }
  },
  {
    "ph": "f", "id": 6858, "pid": 5, "tid": 7, "ts": 1716454216597851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549755, "dur": 7,
    "args": {
      "External id": 6858,
      "cbid": 51, "correlation": 6858
    }
  },
  {
    "ph": "s", "id": 6858, "pid": 76337, "tid": -914061504, "ts": 1716454216549755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x32x64_stage1_warpsize2x1x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216597856, "dur": 61,
    "args": {
      "External id": 6859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6859,
      "registers per thread": 128,
      "shared memory": 12288,
      "blocks per SM": 8.4,
      "warps per SM": 33.6,
      "grid": [1, 48, 14],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 6859, "pid": 5, "tid": 7, "ts": 1716454216597856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549762, "dur": 7,
    "args": {
      "External id": 6859,
      "cbid": 211, "correlation": 6859
    }
  },
  {
    "ph": "s", "id": 6859, "pid": 76337, "tid": -914061504, "ts": 1716454216549762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216597918, "dur": 4,
    "args": {
      "External id": 6861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6861,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1.2,
      "warps per SM": 9.6,
      "grid": [96, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 6861, "pid": 5, "tid": 7, "ts": 1716454216597918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549774, "dur": 5,
    "args": {
      "External id": 6861,
      "cbid": 211, "correlation": 6861
    }
  },
  {
    "ph": "s", "id": 6861, "pid": 76337, "tid": -914061504, "ts": 1716454216549774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597923, "dur": 5,
    "args": {
      "External id": 6867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 6867, "pid": 5, "tid": 7, "ts": 1716454216597923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549801, "dur": 8,
    "args": {
      "External id": 6867,
      "cbid": 211, "correlation": 6867
    }
  },
  {
    "ph": "s", "id": 6867, "pid": 76337, "tid": -914061504, "ts": 1716454216549801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549861, "dur": 0,
    "args": {
      "External id": 6877,
      "cbid": 317, "correlation": 6877
    }
  },
  {
    "ph": "f", "id": 6877, "pid": 76337, "tid": -914061504, "ts": 1716454216549861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549862, "dur": 0,
    "args": {
      "External id": 6878,
      "cbid": 203, "correlation": 6878
    }
  },
  {
    "ph": "f", "id": 6878, "pid": 76337, "tid": -914061504, "ts": 1716454216549862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549863, "dur": 0,
    "args": {
      "External id": 6879,
      "cbid": 205, "correlation": 6879
    }
  },
  {
    "ph": "f", "id": 6879, "pid": 76337, "tid": -914061504, "ts": 1716454216549863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549882, "dur": 1,
    "args": {
      "External id": 6883,
      "cbid": 251, "correlation": 6883
    }
  },
  {
    "ph": "f", "id": 6883, "pid": 76337, "tid": -914061504, "ts": 1716454216549882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216597930, "dur": 11,
    "args": {
      "External id": 6884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6884,
      "registers per thread": 106,
      "shared memory": 16640,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 6884, "pid": 5, "tid": 7, "ts": 1716454216597930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549886, "dur": 12,
    "args": {
      "External id": 6884,
      "cbid": 211, "correlation": 6884
    }
  },
  {
    "ph": "s", "id": 6884, "pid": 76337, "tid": -914061504, "ts": 1716454216549886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216597942, "dur": 4,
    "args": {
      "External id": 6890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.6,
      "warps per SM": 2.4,
      "grid": [48, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 4
    }
  },
  {
    "ph": "f", "id": 6890, "pid": 5, "tid": 7, "ts": 1716454216597942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549919, "dur": 9,
    "args": {
      "External id": 6890,
      "cbid": 211, "correlation": 6890
    }
  },
  {
    "ph": "s", "id": 6890, "pid": 76337, "tid": -914061504, "ts": 1716454216549919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597948, "dur": 4,
    "args": {
      "External id": 6898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6898,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 6898, "pid": 5, "tid": 7, "ts": 1716454216597948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216549985, "dur": 12,
    "args": {
      "External id": 6898,
      "cbid": 211, "correlation": 6898
    }
  },
  {
    "ph": "s", "id": 6898, "pid": 76337, "tid": -914061504, "ts": 1716454216549985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597953, "dur": 3,
    "args": {
      "External id": 6906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6906,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 6906, "pid": 5, "tid": 7, "ts": 1716454216597953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216550025, "dur": 11,
    "args": {
      "External id": 6906,
      "cbid": 211, "correlation": 6906
    }
  },
  {
    "ph": "s", "id": 6906, "pid": 76337, "tid": -914061504, "ts": 1716454216550025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597957, "dur": 3,
    "args": {
      "External id": 6914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6914,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 6914, "pid": 5, "tid": 7, "ts": 1716454216597957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216550051, "dur": 8,
    "args": {
      "External id": 6914,
      "cbid": 211, "correlation": 6914
    }
  },
  {
    "ph": "s", "id": 6914, "pid": 76337, "tid": -914061504, "ts": 1716454216550051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216597962, "dur": 3,
    "args": {
      "External id": 6922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6922,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 6922, "pid": 5, "tid": 7, "ts": 1716454216597962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216550073, "dur": 7,
    "args": {
      "External id": 6922,
      "cbid": 211, "correlation": 6922
    }
  },
  {
    "ph": "s", "id": 6922, "pid": 76337, "tid": -914061504, "ts": 1716454216550073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454216597967, "dur": 2,
    "args": {
      "External id": 6932,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6932,
      "bytes": 6144, "memory bandwidth (GB/s)": 2.6666666666666665
    }
  },
  {
    "ph": "f", "id": 6932, "pid": 5, "tid": 7, "ts": 1716454216597967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216550367, "dur": 45,
    "args": {
      "External id": 6932,
      "cbid": 41, "correlation": 6932
    }
  },
  {
    "ph": "s", "id": 6932, "pid": 76337, "tid": -914061504, "ts": 1716454216550367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454216550413, "dur": 47573,
    "args": {
      "External id": 6933,
      "cbid": 131, "correlation": 6933
    }
  },
  {
    "ph": "f", "id": 6933, "pid": 76337, "tid": -914061504, "ts": 1716454216550413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216598064, "dur": 5,
    "args": {
      "External id": 6941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6941,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 8,
      "warps per SM": 128,
      "grid": [160, 4, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6941, "pid": 5, "tid": 7, "ts": 1716454216598064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216598044, "dur": 21,
    "args": {
      "External id": 6941,
      "cbid": 211, "correlation": 6941
    }
  },
  {
    "ph": "s", "id": 6941, "pid": 76337, "tid": -914061504, "ts": 1716454216598044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216598102, "dur": 4,
    "args": {
      "External id": 6950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6950,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 8,
      "warps per SM": 128,
      "grid": [160, 4, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6950, "pid": 5, "tid": 7, "ts": 1716454216598102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216598092, "dur": 9,
    "args": {
      "External id": 6950,
      "cbid": 211, "correlation": 6950
    }
  },
  {
    "ph": "s", "id": 6950, "pid": 76337, "tid": -914061504, "ts": 1716454216598092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216598130, "dur": 5,
    "args": {
      "External id": 6959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6959,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 8,
      "warps per SM": 128,
      "grid": [160, 4, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6959, "pid": 5, "tid": 7, "ts": 1716454216598130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216598119, "dur": 10,
    "args": {
      "External id": 6959,
      "cbid": 211, "correlation": 6959
    }
  },
  {
    "ph": "s", "id": 6959, "pid": 76337, "tid": -914061504, "ts": 1716454216598119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216598158, "dur": 5,
    "args": {
      "External id": 6968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6968,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 6968, "pid": 5, "tid": 7, "ts": 1716454216598158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216598148, "dur": 8,
    "args": {
      "External id": 6968,
      "cbid": 211, "correlation": 6968
    }
  },
  {
    "ph": "s", "id": 6968, "pid": 76337, "tid": -914061504, "ts": 1716454216598148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454216602686, "dur": 10,
    "args": {
      "External id": 6978,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6978,
      "bytes": 98304, "memory bandwidth (GB/s)": 9.540372670807454
    }
  },
  {
    "ph": "f", "id": 6978, "pid": 5, "tid": 7, "ts": 1716454216602686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216602639, "dur": 47,
    "args": {
      "External id": 6978,
      "cbid": 41, "correlation": 6978
    }
  },
  {
    "ph": "s", "id": 6978, "pid": 76337, "tid": -914061504, "ts": 1716454216602639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454216602687, "dur": 17,
    "args": {
      "External id": 6979,
      "cbid": 131, "correlation": 6979
    }
  },
  {
    "ph": "f", "id": 6979, "pid": 76337, "tid": -914061504, "ts": 1716454216602687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216602756, "dur": 3,
    "args": {
      "External id": 6986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 6986,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 6986, "pid": 5, "tid": 7, "ts": 1716454216602756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216602739, "dur": 17,
    "args": {
      "External id": 6986,
      "cbid": 211, "correlation": 6986
    }
  },
  {
    "ph": "s", "id": 6986, "pid": 76337, "tid": -914061504, "ts": 1716454216602739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454216602827, "dur": 3,
    "args": {
      "External id": 6995,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 6995,
      "bytes": 98304, "memory bandwidth (GB/s)": 26.94736842105263
    }
  },
  {
    "ph": "f", "id": 6995, "pid": 5, "tid": 7, "ts": 1716454216602827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216602780, "dur": 45,
    "args": {
      "External id": 6995,
      "cbid": 41, "correlation": 6995
    }
  },
  {
    "ph": "s", "id": 6995, "pid": 76337, "tid": -914061504, "ts": 1716454216602780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454216603206, "dur": 1,
    "args": {
      "External id": 7005,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 7005,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004032258064516129
    }
  },
  {
    "ph": "f", "id": 7005, "pid": 5, "tid": 7, "ts": 1716454216603206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603177, "dur": 27,
    "args": {
      "External id": 7005,
      "cbid": 41, "correlation": 7005
    }
  },
  {
    "ph": "s", "id": 7005, "pid": 76337, "tid": -914061504, "ts": 1716454216603177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603206, "dur": 9,
    "args": {
      "External id": 7006,
      "cbid": 131, "correlation": 7006
    }
  },
  {
    "ph": "f", "id": 7006, "pid": 76337, "tid": -914061504, "ts": 1716454216603206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454216603476, "dur": 3,
    "args": {
      "External id": 7019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7019, "pid": 5, "tid": 7, "ts": 1716454216603476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603394, "dur": 88,
    "args": {
      "External id": 7019,
      "cbid": 211, "correlation": 7019
    }
  },
  {
    "ph": "s", "id": 7019, "pid": 76337, "tid": -914061504, "ts": 1716454216603394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216603524, "dur": 3,
    "args": {
      "External id": 7027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7027, "pid": 5, "tid": 7, "ts": 1716454216603524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603511, "dur": 13,
    "args": {
      "External id": 7027,
      "cbid": 211, "correlation": 7027
    }
  },
  {
    "ph": "s", "id": 7027, "pid": 76337, "tid": -914061504, "ts": 1716454216603511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216603643, "dur": 3,
    "args": {
      "External id": 7035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7035,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7035, "pid": 5, "tid": 7, "ts": 1716454216603643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603581, "dur": 63,
    "args": {
      "External id": 7035,
      "cbid": 211, "correlation": 7035
    }
  },
  {
    "ph": "s", "id": 7035, "pid": 76337, "tid": -914061504, "ts": 1716454216603581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216603727, "dur": 3,
    "args": {
      "External id": 7043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7043,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7043, "pid": 5, "tid": 7, "ts": 1716454216603727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603676, "dur": 51,
    "args": {
      "External id": 7043,
      "cbid": 211, "correlation": 7043
    }
  },
  {
    "ph": "s", "id": 7043, "pid": 76337, "tid": -914061504, "ts": 1716454216603676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454216603903, "dur": 4,
    "args": {
      "External id": 7055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7055,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7055, "pid": 5, "tid": 7, "ts": 1716454216603903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603802, "dur": 100,
    "args": {
      "External id": 7055,
      "cbid": 211, "correlation": 7055
    }
  },
  {
    "ph": "s", "id": 7055, "pid": 76337, "tid": -914061504, "ts": 1716454216603802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216604028, "dur": 4,
    "args": {
      "External id": 7066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7066, "pid": 5, "tid": 7, "ts": 1716454216604028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216603963, "dur": 66,
    "args": {
      "External id": 7066,
      "cbid": 211, "correlation": 7066
    }
  },
  {
    "ph": "s", "id": 7066, "pid": 76337, "tid": -914061504, "ts": 1716454216603963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216604069, "dur": 3,
    "args": {
      "External id": 7074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7074, "pid": 5, "tid": 7, "ts": 1716454216604069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216604059, "dur": 9,
    "args": {
      "External id": 7074,
      "cbid": 211, "correlation": 7074
    }
  },
  {
    "ph": "s", "id": 7074, "pid": 76337, "tid": -914061504, "ts": 1716454216604059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216604397, "dur": 4,
    "args": {
      "External id": 7082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7082,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7082, "pid": 5, "tid": 7, "ts": 1716454216604397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216604147, "dur": 251,
    "args": {
      "External id": 7082,
      "cbid": 211, "correlation": 7082
    }
  },
  {
    "ph": "s", "id": 7082, "pid": 76337, "tid": -914061504, "ts": 1716454216604147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216604559, "dur": 4,
    "args": {
      "External id": 7090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7090,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7090, "pid": 5, "tid": 7, "ts": 1716454216604559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216604481, "dur": 79,
    "args": {
      "External id": 7090,
      "cbid": 211, "correlation": 7090
    }
  },
  {
    "ph": "s", "id": 7090, "pid": 76337, "tid": -914061504, "ts": 1716454216604481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216604663, "dur": 4,
    "args": {
      "External id": 7099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7099,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7099, "pid": 5, "tid": 7, "ts": 1716454216604663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216604599, "dur": 65,
    "args": {
      "External id": 7099,
      "cbid": 211, "correlation": 7099
    }
  },
  {
    "ph": "s", "id": 7099, "pid": 76337, "tid": -914061504, "ts": 1716454216604599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216604785, "dur": 5,
    "args": {
      "External id": 7112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7112,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7112, "pid": 5, "tid": 7, "ts": 1716454216604785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216604729, "dur": 62,
    "args": {
      "External id": 7112,
      "cbid": 211, "correlation": 7112
    }
  },
  {
    "ph": "s", "id": 7112, "pid": 76337, "tid": -914061504, "ts": 1716454216604729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454216604847, "dur": 8,
    "args": {
      "External id": 7122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7122,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7122, "pid": 5, "tid": 7, "ts": 1716454216604847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216604833, "dur": 14,
    "args": {
      "External id": 7122,
      "cbid": 211, "correlation": 7122
    }
  },
  {
    "ph": "s", "id": 7122, "pid": 76337, "tid": -914061504, "ts": 1716454216604833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216605049, "dur": 98,
    "args": {
      "External id": 7139,
      "cbid": 251, "correlation": 7139
    }
  },
  {
    "ph": "f", "id": 7139, "pid": 76337, "tid": -914061504, "ts": 1716454216605049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454216605174, "dur": 11,
    "args": {
      "External id": 7141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7141,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7141, "pid": 5, "tid": 7, "ts": 1716454216605174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216605157, "dur": 24,
    "args": {
      "External id": 7141,
      "cbid": 211, "correlation": 7141
    }
  },
  {
    "ph": "s", "id": 7141, "pid": 76337, "tid": -914061504, "ts": 1716454216605157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216605255, "dur": 4,
    "args": {
      "External id": 7149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7149,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7149, "pid": 5, "tid": 7, "ts": 1716454216605255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216605242, "dur": 12,
    "args": {
      "External id": 7149,
      "cbid": 211, "correlation": 7149
    }
  },
  {
    "ph": "s", "id": 7149, "pid": 76337, "tid": -914061504, "ts": 1716454216605242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216647829, "dur": 106,
    "args": {
      "External id": 7165,
      "cbid": 251, "correlation": 7165
    }
  },
  {
    "ph": "f", "id": 7165, "pid": 76337, "tid": -914061504, "ts": 1716454216647829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216647944, "dur": 0,
    "args": {
      "External id": 7167,
      "cbid": 251, "correlation": 7167
    }
  },
  {
    "ph": "f", "id": 7167, "pid": 76337, "tid": -914061504, "ts": 1716454216647944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216647964, "dur": 12,
    "args": {
      "External id": 7168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7168,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7168, "pid": 5, "tid": 7, "ts": 1716454216647964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216647947, "dur": 18,
    "args": {
      "External id": 7168,
      "cbid": 211, "correlation": 7168
    }
  },
  {
    "ph": "s", "id": 7168, "pid": 76337, "tid": -914061504, "ts": 1716454216647947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216648121, "dur": 4,
    "args": {
      "External id": 7170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7170,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7170, "pid": 5, "tid": 7, "ts": 1716454216648121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216647992, "dur": 129,
    "args": {
      "External id": 7170,
      "cbid": 211, "correlation": 7170
    }
  },
  {
    "ph": "s", "id": 7170, "pid": 76337, "tid": -914061504, "ts": 1716454216647992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216648233, "dur": 1,
    "args": {
      "External id": 7180,
      "cbid": 317, "correlation": 7180
    }
  },
  {
    "ph": "f", "id": 7180, "pid": 76337, "tid": -914061504, "ts": 1716454216648233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216648235, "dur": 0,
    "args": {
      "External id": 7181,
      "cbid": 203, "correlation": 7181
    }
  },
  {
    "ph": "f", "id": 7181, "pid": 76337, "tid": -914061504, "ts": 1716454216648235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216648236, "dur": 1,
    "args": {
      "External id": 7182,
      "cbid": 205, "correlation": 7182
    }
  },
  {
    "ph": "f", "id": 7182, "pid": 76337, "tid": -914061504, "ts": 1716454216648236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216652109, "dur": 3,
    "args": {
      "External id": 7186,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 7186,
      "bytes": 393216, "memory bandwidth (GB/s)": 107.78947368421052
    }
  },
  {
    "ph": "f", "id": 7186, "pid": 5, "tid": 7, "ts": 1716454216652109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652049, "dur": 60,
    "args": {
      "External id": 7186,
      "cbid": 51, "correlation": 7186
    }
  },
  {
    "ph": "s", "id": 7186, "pid": 76337, "tid": -914061504, "ts": 1716454216652049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216652131, "dur": 8,
    "args": {
      "External id": 7187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7187,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7187, "pid": 5, "tid": 7, "ts": 1716454216652131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652116, "dur": 15,
    "args": {
      "External id": 7187,
      "cbid": 211, "correlation": 7187
    }
  },
  {
    "ph": "s", "id": 7187, "pid": 76337, "tid": -914061504, "ts": 1716454216652116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216652145, "dur": 3,
    "args": {
      "External id": 7189,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 7189,
      "bytes": 46080, "memory bandwidth (GB/s)": 14.117647058823529
    }
  },
  {
    "ph": "f", "id": 7189, "pid": 5, "tid": 7, "ts": 1716454216652145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652135, "dur": 9,
    "args": {
      "External id": 7189,
      "cbid": 51, "correlation": 7189
    }
  },
  {
    "ph": "s", "id": 7189, "pid": 76337, "tid": -914061504, "ts": 1716454216652135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216652152, "dur": 4,
    "args": {
      "External id": 7190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 7190, "pid": 5, "tid": 7, "ts": 1716454216652152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652144, "dur": 6,
    "args": {
      "External id": 7190,
      "cbid": 211, "correlation": 7190
    }
  },
  {
    "ph": "s", "id": 7190, "pid": 76337, "tid": -914061504, "ts": 1716454216652144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216652171, "dur": 4,
    "args": {
      "External id": 7192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7192, "pid": 5, "tid": 7, "ts": 1716454216652171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652159, "dur": 10,
    "args": {
      "External id": 7192,
      "cbid": 211, "correlation": 7192
    }
  },
  {
    "ph": "s", "id": 7192, "pid": 76337, "tid": -914061504, "ts": 1716454216652159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652174, "dur": 0,
    "args": {
      "External id": 7193,
      "cbid": 51, "correlation": 7193
    }
  },
  {
    "ph": "s", "id": 7193, "pid": 76337, "tid": -914061504, "ts": 1716454216652174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216652185, "dur": 95,
    "args": {
      "External id": 7194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7194,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7194, "pid": 5, "tid": 7, "ts": 1716454216652185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652176, "dur": 7,
    "args": {
      "External id": 7194,
      "cbid": 211, "correlation": 7194
    }
  },
  {
    "ph": "s", "id": 7194, "pid": 76337, "tid": -914061504, "ts": 1716454216652176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216652973, "dur": 64,
    "args": {
      "External id": 7199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7199, "pid": 5, "tid": 7, "ts": 1716454216652973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216652956, "dur": 25,
    "args": {
      "External id": 7199,
      "cbid": 211, "correlation": 7199
    }
  },
  {
    "ph": "s", "id": 7199, "pid": 76337, "tid": -914061504, "ts": 1716454216652956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216671816, "dur": 56,
    "args": {
      "External id": 7219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7219,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 7219, "pid": 5, "tid": 7, "ts": 1716454216671816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216671746, "dur": 75,
    "args": {
      "External id": 7219,
      "cbid": 211, "correlation": 7219
    }
  },
  {
    "ph": "s", "id": 7219, "pid": 76337, "tid": -914061504, "ts": 1716454216671746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216671874, "dur": 5,
    "args": {
      "External id": 7231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7231,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7231, "pid": 5, "tid": 7, "ts": 1716454216671874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216671837, "dur": 10,
    "args": {
      "External id": 7231,
      "cbid": 211, "correlation": 7231
    }
  },
  {
    "ph": "s", "id": 7231, "pid": 76337, "tid": -914061504, "ts": 1716454216671837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216671901, "dur": 60,
    "args": {
      "External id": 7234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7234, "pid": 5, "tid": 7, "ts": 1716454216671901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216671890, "dur": 10,
    "args": {
      "External id": 7234,
      "cbid": 211, "correlation": 7234
    }
  },
  {
    "ph": "s", "id": 7234, "pid": 76337, "tid": -914061504, "ts": 1716454216671890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216671986, "dur": 38,
    "args": {
      "External id": 7243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7243,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7243, "pid": 5, "tid": 7, "ts": 1716454216671986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216671964, "dur": 22,
    "args": {
      "External id": 7243,
      "cbid": 211, "correlation": 7243
    }
  },
  {
    "ph": "s", "id": 7243, "pid": 76337, "tid": -914061504, "ts": 1716454216671964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216672111, "dur": 1,
    "args": {
      "External id": 7253,
      "cbid": 317, "correlation": 7253
    }
  },
  {
    "ph": "f", "id": 7253, "pid": 76337, "tid": -914061504, "ts": 1716454216672111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216672113, "dur": 1,
    "args": {
      "External id": 7254,
      "cbid": 203, "correlation": 7254
    }
  },
  {
    "ph": "f", "id": 7254, "pid": 76337, "tid": -914061504, "ts": 1716454216672113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216672116, "dur": 1,
    "args": {
      "External id": 7255,
      "cbid": 205, "correlation": 7255
    }
  },
  {
    "ph": "f", "id": 7255, "pid": 76337, "tid": -914061504, "ts": 1716454216672116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216675584, "dur": 41,
    "args": {
      "External id": 7259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7259,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7259, "pid": 5, "tid": 7, "ts": 1716454216675584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216675559, "dur": 25,
    "args": {
      "External id": 7259,
      "cbid": 211, "correlation": 7259
    }
  },
  {
    "ph": "s", "id": 7259, "pid": 76337, "tid": -914061504, "ts": 1716454216675559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216675626, "dur": 15,
    "args": {
      "External id": 7261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7261,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7261, "pid": 5, "tid": 7, "ts": 1716454216675626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216675588, "dur": 7,
    "args": {
      "External id": 7261,
      "cbid": 211, "correlation": 7261
    }
  },
  {
    "ph": "s", "id": 7261, "pid": 76337, "tid": -914061504, "ts": 1716454216675588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216675643, "dur": 3,
    "args": {
      "External id": 7263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7263, "pid": 5, "tid": 7, "ts": 1716454216675643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216675604, "dur": 11,
    "args": {
      "External id": 7263,
      "cbid": 211, "correlation": 7263
    }
  },
  {
    "ph": "s", "id": 7263, "pid": 76337, "tid": -914061504, "ts": 1716454216675604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216675619, "dur": 0,
    "args": {
      "External id": 7264,
      "cbid": 51, "correlation": 7264
    }
  },
  {
    "ph": "s", "id": 7264, "pid": 76337, "tid": -914061504, "ts": 1716454216675619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216675647, "dur": 772,
    "args": {
      "External id": 7265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7265,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7265, "pid": 5, "tid": 7, "ts": 1716454216675647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216675621, "dur": 8,
    "args": {
      "External id": 7265,
      "cbid": 211, "correlation": 7265
    }
  },
  {
    "ph": "s", "id": 7265, "pid": 76337, "tid": -914061504, "ts": 1716454216675621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216676442, "dur": 63,
    "args": {
      "External id": 7270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7270, "pid": 5, "tid": 7, "ts": 1716454216676442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676423, "dur": 18,
    "args": {
      "External id": 7270,
      "cbid": 211, "correlation": 7270
    }
  },
  {
    "ph": "s", "id": 7270, "pid": 76337, "tid": -914061504, "ts": 1716454216676423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216676524, "dur": 4,
    "args": {
      "External id": 7278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7278,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7278, "pid": 5, "tid": 7, "ts": 1716454216676524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676512, "dur": 11,
    "args": {
      "External id": 7278,
      "cbid": 211, "correlation": 7278
    }
  },
  {
    "ph": "s", "id": 7278, "pid": 76337, "tid": -914061504, "ts": 1716454216676512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676699, "dur": 6,
    "args": {
      "External id": 7294,
      "cbid": 251, "correlation": 7294
    }
  },
  {
    "ph": "f", "id": 7294, "pid": 76337, "tid": -914061504, "ts": 1716454216676699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676711, "dur": 0,
    "args": {
      "External id": 7296,
      "cbid": 251, "correlation": 7296
    }
  },
  {
    "ph": "f", "id": 7296, "pid": 76337, "tid": -914061504, "ts": 1716454216676711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216676732, "dur": 12,
    "args": {
      "External id": 7297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7297,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 7297, "pid": 5, "tid": 7, "ts": 1716454216676732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676715, "dur": 17,
    "args": {
      "External id": 7297,
      "cbid": 211, "correlation": 7297
    }
  },
  {
    "ph": "s", "id": 7297, "pid": 76337, "tid": -914061504, "ts": 1716454216676715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216676750, "dur": 4,
    "args": {
      "External id": 7299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7299,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 7299, "pid": 5, "tid": 7, "ts": 1716454216676750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676738, "dur": 10,
    "args": {
      "External id": 7299,
      "cbid": 211, "correlation": 7299
    }
  },
  {
    "ph": "s", "id": 7299, "pid": 76337, "tid": -914061504, "ts": 1716454216676738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216676853, "dur": 57,
    "args": {
      "External id": 7309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7309, "pid": 5, "tid": 7, "ts": 1716454216676853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676839, "dur": 13,
    "args": {
      "External id": 7309,
      "cbid": 211, "correlation": 7309
    }
  },
  {
    "ph": "s", "id": 7309, "pid": 76337, "tid": -914061504, "ts": 1716454216676839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216676952, "dur": 56,
    "args": {
      "External id": 7329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7329,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 7329, "pid": 5, "tid": 7, "ts": 1716454216676952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676938, "dur": 13,
    "args": {
      "External id": 7329,
      "cbid": 211, "correlation": 7329
    }
  },
  {
    "ph": "s", "id": 7329, "pid": 76337, "tid": -914061504, "ts": 1716454216676938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216677009, "dur": 4,
    "args": {
      "External id": 7341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7341,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7341, "pid": 5, "tid": 7, "ts": 1716454216677009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676962, "dur": 8,
    "args": {
      "External id": 7341,
      "cbid": 211, "correlation": 7341
    }
  },
  {
    "ph": "s", "id": 7341, "pid": 76337, "tid": -914061504, "ts": 1716454216676962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216677015, "dur": 57,
    "args": {
      "External id": 7344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7344,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7344, "pid": 5, "tid": 7, "ts": 1716454216677015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216676998, "dur": 8,
    "args": {
      "External id": 7344,
      "cbid": 211, "correlation": 7344
    }
  },
  {
    "ph": "s", "id": 7344, "pid": 76337, "tid": -914061504, "ts": 1716454216676998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216677074, "dur": 37,
    "args": {
      "External id": 7353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7353,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7353, "pid": 5, "tid": 7, "ts": 1716454216677074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677045, "dur": 12,
    "args": {
      "External id": 7353,
      "cbid": 211, "correlation": 7353
    }
  },
  {
    "ph": "s", "id": 7353, "pid": 76337, "tid": -914061504, "ts": 1716454216677045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677158, "dur": 0,
    "args": {
      "External id": 7363,
      "cbid": 317, "correlation": 7363
    }
  },
  {
    "ph": "f", "id": 7363, "pid": 76337, "tid": -914061504, "ts": 1716454216677158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677159, "dur": 1,
    "args": {
      "External id": 7364,
      "cbid": 203, "correlation": 7364
    }
  },
  {
    "ph": "f", "id": 7364, "pid": 76337, "tid": -914061504, "ts": 1716454216677159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677161, "dur": 0,
    "args": {
      "External id": 7365,
      "cbid": 205, "correlation": 7365
    }
  },
  {
    "ph": "f", "id": 7365, "pid": 76337, "tid": -914061504, "ts": 1716454216677161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216677190, "dur": 40,
    "args": {
      "External id": 7369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7369,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7369, "pid": 5, "tid": 7, "ts": 1716454216677190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677176, "dur": 13,
    "args": {
      "External id": 7369,
      "cbid": 211, "correlation": 7369
    }
  },
  {
    "ph": "s", "id": 7369, "pid": 76337, "tid": -914061504, "ts": 1716454216677176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216677232, "dur": 16,
    "args": {
      "External id": 7371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7371, "pid": 5, "tid": 7, "ts": 1716454216677232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677192, "dur": 6,
    "args": {
      "External id": 7371,
      "cbid": 211, "correlation": 7371
    }
  },
  {
    "ph": "s", "id": 7371, "pid": 76337, "tid": -914061504, "ts": 1716454216677192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216677249, "dur": 3,
    "args": {
      "External id": 7373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7373, "pid": 5, "tid": 7, "ts": 1716454216677249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677203, "dur": 6,
    "args": {
      "External id": 7373,
      "cbid": 211, "correlation": 7373
    }
  },
  {
    "ph": "s", "id": 7373, "pid": 76337, "tid": -914061504, "ts": 1716454216677203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677212, "dur": 0,
    "args": {
      "External id": 7374,
      "cbid": 51, "correlation": 7374
    }
  },
  {
    "ph": "s", "id": 7374, "pid": 76337, "tid": -914061504, "ts": 1716454216677212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216677254, "dur": 767,
    "args": {
      "External id": 7375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7375,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7375, "pid": 5, "tid": 7, "ts": 1716454216677254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677213, "dur": 5,
    "args": {
      "External id": 7375,
      "cbid": 211, "correlation": 7375
    }
  },
  {
    "ph": "s", "id": 7375, "pid": 76337, "tid": -914061504, "ts": 1716454216677213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216678022, "dur": 63,
    "args": {
      "External id": 7380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7380, "pid": 5, "tid": 7, "ts": 1716454216678022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677242, "dur": 10,
    "args": {
      "External id": 7380,
      "cbid": 211, "correlation": 7380
    }
  },
  {
    "ph": "s", "id": 7380, "pid": 76337, "tid": -914061504, "ts": 1716454216677242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216678086, "dur": 51,
    "args": {
      "External id": 7388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7388,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7388, "pid": 5, "tid": 7, "ts": 1716454216678086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677278, "dur": 9,
    "args": {
      "External id": 7388,
      "cbid": 211, "correlation": 7388
    }
  },
  {
    "ph": "s", "id": 7388, "pid": 76337, "tid": -914061504, "ts": 1716454216677278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216678138, "dur": 36,
    "args": {
      "External id": 7396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7396,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7396, "pid": 5, "tid": 7, "ts": 1716454216678138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677328, "dur": 10,
    "args": {
      "External id": 7396,
      "cbid": 211, "correlation": 7396
    }
  },
  {
    "ph": "s", "id": 7396, "pid": 76337, "tid": -914061504, "ts": 1716454216677328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216678176, "dur": 56,
    "args": {
      "External id": 7416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7416,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 7416, "pid": 5, "tid": 7, "ts": 1716454216678176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677494, "dur": 15,
    "args": {
      "External id": 7416,
      "cbid": 211, "correlation": 7416
    }
  },
  {
    "ph": "s", "id": 7416, "pid": 76337, "tid": -914061504, "ts": 1716454216677494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216678233, "dur": 4,
    "args": {
      "External id": 7428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7428,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7428, "pid": 5, "tid": 7, "ts": 1716454216678233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677519, "dur": 7,
    "args": {
      "External id": 7428,
      "cbid": 211, "correlation": 7428
    }
  },
  {
    "ph": "s", "id": 7428, "pid": 76337, "tid": -914061504, "ts": 1716454216677519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216678239, "dur": 59,
    "args": {
      "External id": 7431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7431,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7431, "pid": 5, "tid": 7, "ts": 1716454216678239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677538, "dur": 6,
    "args": {
      "External id": 7431,
      "cbid": 211, "correlation": 7431
    }
  },
  {
    "ph": "s", "id": 7431, "pid": 76337, "tid": -914061504, "ts": 1716454216677538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677615, "dur": 0,
    "args": {
      "External id": 7442,
      "cbid": 317, "correlation": 7442
    }
  },
  {
    "ph": "f", "id": 7442, "pid": 76337, "tid": -914061504, "ts": 1716454216677615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677616, "dur": 0,
    "args": {
      "External id": 7443,
      "cbid": 203, "correlation": 7443
    }
  },
  {
    "ph": "f", "id": 7443, "pid": 76337, "tid": -914061504, "ts": 1716454216677616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216677617, "dur": 0,
    "args": {
      "External id": 7444,
      "cbid": 205, "correlation": 7444
    }
  },
  {
    "ph": "f", "id": 7444, "pid": 76337, "tid": -914061504, "ts": 1716454216677617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679525, "dur": 127,
    "args": {
      "External id": 7448,
      "cbid": 251, "correlation": 7448
    }
  },
  {
    "ph": "f", "id": 7448, "pid": 76337, "tid": -914061504, "ts": 1716454216679525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679654, "dur": 1,
    "args": {
      "External id": 7449,
      "cbid": 251, "correlation": 7449
    }
  },
  {
    "ph": "f", "id": 7449, "pid": 76337, "tid": -914061504, "ts": 1716454216679654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679655, "dur": 47,
    "args": {
      "External id": 7450,
      "cbid": 251, "correlation": 7450
    }
  },
  {
    "ph": "f", "id": 7450, "pid": 76337, "tid": -914061504, "ts": 1716454216679655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679704, "dur": 1,
    "args": {
      "External id": 7451,
      "cbid": 251, "correlation": 7451
    }
  },
  {
    "ph": "f", "id": 7451, "pid": 76337, "tid": -914061504, "ts": 1716454216679704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679706, "dur": 59,
    "args": {
      "External id": 7452,
      "cbid": 251, "correlation": 7452
    }
  },
  {
    "ph": "f", "id": 7452, "pid": 76337, "tid": -914061504, "ts": 1716454216679706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679766, "dur": 1,
    "args": {
      "External id": 7453,
      "cbid": 251, "correlation": 7453
    }
  },
  {
    "ph": "f", "id": 7453, "pid": 76337, "tid": -914061504, "ts": 1716454216679766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679768, "dur": 35,
    "args": {
      "External id": 7454,
      "cbid": 251, "correlation": 7454
    }
  },
  {
    "ph": "f", "id": 7454, "pid": 76337, "tid": -914061504, "ts": 1716454216679768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679804, "dur": 1,
    "args": {
      "External id": 7455,
      "cbid": 251, "correlation": 7455
    }
  },
  {
    "ph": "f", "id": 7455, "pid": 76337, "tid": -914061504, "ts": 1716454216679804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679807, "dur": 0,
    "args": {
      "External id": 7456,
      "cbid": 251, "correlation": 7456
    }
  },
  {
    "ph": "f", "id": 7456, "pid": 76337, "tid": -914061504, "ts": 1716454216679807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216679831, "dur": 123,
    "args": {
      "External id": 7457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7457,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 7457, "pid": 5, "tid": 7, "ts": 1716454216679831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216679813, "dur": 19,
    "args": {
      "External id": 7457,
      "cbid": 211, "correlation": 7457
    }
  },
  {
    "ph": "s", "id": 7457, "pid": 76337, "tid": -914061504, "ts": 1716454216679813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216680393, "dur": 64,
    "args": {
      "External id": 7463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7463, "pid": 5, "tid": 7, "ts": 1716454216680393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680379, "dur": 14,
    "args": {
      "External id": 7463,
      "cbid": 211, "correlation": 7463
    }
  },
  {
    "ph": "s", "id": 7463, "pid": 76337, "tid": -914061504, "ts": 1716454216680379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216680642, "dur": 164,
    "args": {
      "External id": 7474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7474, "pid": 5, "tid": 7, "ts": 1716454216680642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680623, "dur": 19,
    "args": {
      "External id": 7474,
      "cbid": 211, "correlation": 7474
    }
  },
  {
    "ph": "s", "id": 7474, "pid": 76337, "tid": -914061504, "ts": 1716454216680623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216680814, "dur": 102,
    "args": {
      "External id": 7496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7496,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7496, "pid": 5, "tid": 7, "ts": 1716454216680814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680688, "dur": 56,
    "args": {
      "External id": 7496,
      "cbid": 211, "correlation": 7496
    }
  },
  {
    "ph": "s", "id": 7496, "pid": 76337, "tid": -914061504, "ts": 1716454216680688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680892, "dur": 2,
    "args": {
      "External id": 7507,
      "cbid": 251, "correlation": 7507
    }
  },
  {
    "ph": "f", "id": 7507, "pid": 76337, "tid": -914061504, "ts": 1716454216680892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216680918, "dur": 119,
    "args": {
      "External id": 7508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7508,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7508, "pid": 5, "tid": 7, "ts": 1716454216680918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680901, "dur": 15,
    "args": {
      "External id": 7508,
      "cbid": 211, "correlation": 7508
    }
  },
  {
    "ph": "s", "id": 7508, "pid": 76337, "tid": -914061504, "ts": 1716454216680901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680986, "dur": 1,
    "args": {
      "External id": 7519,
      "cbid": 251, "correlation": 7519
    }
  },
  {
    "ph": "f", "id": 7519, "pid": 76337, "tid": -914061504, "ts": 1716454216680986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216681039, "dur": 116,
    "args": {
      "External id": 7520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7520,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7520, "pid": 5, "tid": 7, "ts": 1716454216681039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216680990, "dur": 13,
    "args": {
      "External id": 7520,
      "cbid": 211, "correlation": 7520
    }
  },
  {
    "ph": "s", "id": 7520, "pid": 76337, "tid": -914061504, "ts": 1716454216680990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681073, "dur": 1,
    "args": {
      "External id": 7531,
      "cbid": 251, "correlation": 7531
    }
  },
  {
    "ph": "f", "id": 7531, "pid": 76337, "tid": -914061504, "ts": 1716454216681073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216681156, "dur": 111,
    "args": {
      "External id": 7532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7532,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7532, "pid": 5, "tid": 7, "ts": 1716454216681156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681077, "dur": 12,
    "args": {
      "External id": 7532,
      "cbid": 211, "correlation": 7532
    }
  },
  {
    "ph": "s", "id": 7532, "pid": 76337, "tid": -914061504, "ts": 1716454216681077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216681337, "dur": 5218,
    "args": {
      "External id": 7553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7553,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 7553, "pid": 5, "tid": 7, "ts": 1716454216681337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681219, "dur": 117,
    "args": {
      "External id": 7553,
      "cbid": 211, "correlation": 7553
    }
  },
  {
    "ph": "s", "id": 7553, "pid": 76337, "tid": -914061504, "ts": 1716454216681219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681464, "dur": 2,
    "args": {
      "External id": 7571,
      "cbid": 251, "correlation": 7571
    }
  },
  {
    "ph": "f", "id": 7571, "pid": 76337, "tid": -914061504, "ts": 1716454216681464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216686556, "dur": 114,
    "args": {
      "External id": 7573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7573,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 7573, "pid": 5, "tid": 7, "ts": 1716454216686556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681471, "dur": 14,
    "args": {
      "External id": 7573,
      "cbid": 211, "correlation": 7573
    }
  },
  {
    "ph": "s", "id": 7573, "pid": 76337, "tid": -914061504, "ts": 1716454216681471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216686671, "dur": 35,
    "args": {
      "External id": 7581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7581,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7581, "pid": 5, "tid": 7, "ts": 1716454216686671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681567, "dur": 15,
    "args": {
      "External id": 7581,
      "cbid": 211, "correlation": 7581
    }
  },
  {
    "ph": "s", "id": 7581, "pid": 76337, "tid": -914061504, "ts": 1716454216681567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216686708, "dur": 172,
    "args": {
      "External id": 7589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7589, "pid": 5, "tid": 7, "ts": 1716454216686708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681617, "dur": 10,
    "args": {
      "External id": 7589,
      "cbid": 211, "correlation": 7589
    }
  },
  {
    "ph": "s", "id": 7589, "pid": 76337, "tid": -914061504, "ts": 1716454216681617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216686881, "dur": 103,
    "args": {
      "External id": 7611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7611,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7611, "pid": 5, "tid": 7, "ts": 1716454216686881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681686, "dur": 12,
    "args": {
      "External id": 7611,
      "cbid": 211, "correlation": 7611
    }
  },
  {
    "ph": "s", "id": 7611, "pid": 76337, "tid": -914061504, "ts": 1716454216681686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681777, "dur": 1,
    "args": {
      "External id": 7622,
      "cbid": 251, "correlation": 7622
    }
  },
  {
    "ph": "f", "id": 7622, "pid": 76337, "tid": -914061504, "ts": 1716454216681777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216686986, "dur": 118,
    "args": {
      "External id": 7623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7623,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7623, "pid": 5, "tid": 7, "ts": 1716454216686986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681782, "dur": 13,
    "args": {
      "External id": 7623,
      "cbid": 211, "correlation": 7623
    }
  },
  {
    "ph": "s", "id": 7623, "pid": 76337, "tid": -914061504, "ts": 1716454216681782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681921, "dur": 48,
    "args": {
      "External id": 7634,
      "cbid": 251, "correlation": 7634
    }
  },
  {
    "ph": "f", "id": 7634, "pid": 76337, "tid": -914061504, "ts": 1716454216681921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681982, "dur": 0,
    "args": {
      "External id": 7635,
      "cbid": 251, "correlation": 7635
    }
  },
  {
    "ph": "f", "id": 7635, "pid": 76337, "tid": -914061504, "ts": 1716454216681982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216687112, "dur": 8,
    "args": {
      "External id": 7636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7636,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 7636, "pid": 5, "tid": 7, "ts": 1716454216687112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216681985, "dur": 22,
    "args": {
      "External id": 7636,
      "cbid": 211, "correlation": 7636
    }
  },
  {
    "ph": "s", "id": 7636, "pid": 76337, "tid": -914061504, "ts": 1716454216681985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216687126, "dur": 5,
    "args": {
      "External id": 7638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7638,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 7638, "pid": 5, "tid": 7, "ts": 1716454216687126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682015, "dur": 98,
    "args": {
      "External id": 7638,
      "cbid": 211, "correlation": 7638
    }
  },
  {
    "ph": "s", "id": 7638, "pid": 76337, "tid": -914061504, "ts": 1716454216682015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682192, "dur": 1,
    "args": {
      "External id": 7649,
      "cbid": 251, "correlation": 7649
    }
  },
  {
    "ph": "f", "id": 7649, "pid": 76337, "tid": -914061504, "ts": 1716454216682192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682196, "dur": 0,
    "args": {
      "External id": 7650,
      "cbid": 251, "correlation": 7650
    }
  },
  {
    "ph": "f", "id": 7650, "pid": 76337, "tid": -914061504, "ts": 1716454216682196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216687132, "dur": 7,
    "args": {
      "External id": 7651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7651,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 7651, "pid": 5, "tid": 7, "ts": 1716454216687132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682198, "dur": 14,
    "args": {
      "External id": 7651,
      "cbid": 211, "correlation": 7651
    }
  },
  {
    "ph": "s", "id": 7651, "pid": 76337, "tid": -914061504, "ts": 1716454216682198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216687141, "dur": 4,
    "args": {
      "External id": 7653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7653,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 7653, "pid": 5, "tid": 7, "ts": 1716454216687141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682214, "dur": 6,
    "args": {
      "External id": 7653,
      "cbid": 211, "correlation": 7653
    }
  },
  {
    "ph": "s", "id": 7653, "pid": 76337, "tid": -914061504, "ts": 1716454216682214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216687146, "dur": 177,
    "args": {
      "External id": 7674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7674,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 7674, "pid": 5, "tid": 7, "ts": 1716454216687146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682292, "dur": 13,
    "args": {
      "External id": 7674,
      "cbid": 211, "correlation": 7674
    }
  },
  {
    "ph": "s", "id": 7674, "pid": 76337, "tid": -914061504, "ts": 1716454216682292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682405, "dur": 1,
    "args": {
      "External id": 7692,
      "cbid": 251, "correlation": 7692
    }
  },
  {
    "ph": "f", "id": 7692, "pid": 76337, "tid": -914061504, "ts": 1716454216682405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216687325, "dur": 117,
    "args": {
      "External id": 7694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7694,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 7694, "pid": 5, "tid": 7, "ts": 1716454216687325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682411, "dur": 15,
    "args": {
      "External id": 7694,
      "cbid": 211, "correlation": 7694
    }
  },
  {
    "ph": "s", "id": 7694, "pid": 76337, "tid": -914061504, "ts": 1716454216682411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216687443, "dur": 35,
    "args": {
      "External id": 7702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7702,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7702, "pid": 5, "tid": 7, "ts": 1716454216687443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682486, "dur": 12,
    "args": {
      "External id": 7702,
      "cbid": 211, "correlation": 7702
    }
  },
  {
    "ph": "s", "id": 7702, "pid": 76337, "tid": -914061504, "ts": 1716454216682486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216687479, "dur": 51,
    "args": {
      "External id": 7710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7710,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7710, "pid": 5, "tid": 7, "ts": 1716454216687479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682528, "dur": 9,
    "args": {
      "External id": 7710,
      "cbid": 211, "correlation": 7710
    }
  },
  {
    "ph": "s", "id": 7710, "pid": 76337, "tid": -914061504, "ts": 1716454216682528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216687531, "dur": 103,
    "args": {
      "External id": 7732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7732,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7732, "pid": 5, "tid": 7, "ts": 1716454216687531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682581, "dur": 10,
    "args": {
      "External id": 7732,
      "cbid": 211, "correlation": 7732
    }
  },
  {
    "ph": "s", "id": 7732, "pid": 76337, "tid": -914061504, "ts": 1716454216682581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682731, "dur": 74,
    "args": {
      "External id": 7748,
      "cbid": 251, "correlation": 7748
    }
  },
  {
    "ph": "f", "id": 7748, "pid": 76337, "tid": -914061504, "ts": 1716454216682731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216687646, "dur": 624,
    "args": {
      "External id": 7750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7750,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7750, "pid": 5, "tid": 7, "ts": 1716454216687646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216682812, "dur": 20,
    "args": {
      "External id": 7750,
      "cbid": 211, "correlation": 7750
    }
  },
  {
    "ph": "s", "id": 7750, "pid": 76337, "tid": -914061504, "ts": 1716454216682812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216683002, "dur": 2,
    "args": {
      "External id": 7758,
      "cbid": 317, "correlation": 7758
    }
  },
  {
    "ph": "f", "id": 7758, "pid": 76337, "tid": -914061504, "ts": 1716454216683002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216683005, "dur": 9502,
    "args": {
      "External id": 7759,
      "cbid": 20, "correlation": 7759
    }
  },
  {
    "ph": "f", "id": 7759, "pid": 76337, "tid": -914061504, "ts": 1716454216683005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216692659, "dur": 260,
    "args": {
      "External id": 7762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7762,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7762, "pid": 5, "tid": 7, "ts": 1716454216692659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216692571, "dur": 88,
    "args": {
      "External id": 7762,
      "cbid": 211, "correlation": 7762
    }
  },
  {
    "ph": "s", "id": 7762, "pid": 76337, "tid": -914061504, "ts": 1716454216692571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216692705, "dur": 1,
    "args": {
      "External id": 7770,
      "cbid": 317, "correlation": 7770
    }
  },
  {
    "ph": "f", "id": 7770, "pid": 76337, "tid": -914061504, "ts": 1716454216692705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216692707, "dur": 3543,
    "args": {
      "External id": 7771,
      "cbid": 20, "correlation": 7771
    }
  },
  {
    "ph": "f", "id": 7771, "pid": 76337, "tid": -914061504, "ts": 1716454216692707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216696349, "dur": 261,
    "args": {
      "External id": 7774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7774, "pid": 5, "tid": 7, "ts": 1716454216696349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696271, "dur": 78,
    "args": {
      "External id": 7774,
      "cbid": 211, "correlation": 7774
    }
  },
  {
    "ph": "s", "id": 7774, "pid": 76337, "tid": -914061504, "ts": 1716454216696271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696506, "dur": 3,
    "args": {
      "External id": 7790,
      "cbid": 251, "correlation": 7790
    }
  },
  {
    "ph": "f", "id": 7790, "pid": 76337, "tid": -914061504, "ts": 1716454216696506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696513, "dur": 0,
    "args": {
      "External id": 7792,
      "cbid": 251, "correlation": 7792
    }
  },
  {
    "ph": "f", "id": 7792, "pid": 76337, "tid": -914061504, "ts": 1716454216696513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216696611, "dur": 377,
    "args": {
      "External id": 7793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7793,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 7793, "pid": 5, "tid": 7, "ts": 1716454216696611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696531, "dur": 16,
    "args": {
      "External id": 7793,
      "cbid": 211, "correlation": 7793
    }
  },
  {
    "ph": "s", "id": 7793, "pid": 76337, "tid": -914061504, "ts": 1716454216696531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216696990, "dur": 50,
    "args": {
      "External id": 7801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7801,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7801, "pid": 5, "tid": 7, "ts": 1716454216696990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696589, "dur": 11,
    "args": {
      "External id": 7801,
      "cbid": 211, "correlation": 7801
    }
  },
  {
    "ph": "s", "id": 7801, "pid": 76337, "tid": -914061504, "ts": 1716454216696589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216697042, "dur": 172,
    "args": {
      "External id": 7812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7812, "pid": 5, "tid": 7, "ts": 1716454216697042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696687, "dur": 14,
    "args": {
      "External id": 7812,
      "cbid": 211, "correlation": 7812
    }
  },
  {
    "ph": "s", "id": 7812, "pid": 76337, "tid": -914061504, "ts": 1716454216696687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696793, "dur": 0,
    "args": {
      "External id": 7824,
      "cbid": 317, "correlation": 7824
    }
  },
  {
    "ph": "f", "id": 7824, "pid": 76337, "tid": -914061504, "ts": 1716454216696793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696794, "dur": 1,
    "args": {
      "External id": 7825,
      "cbid": 203, "correlation": 7825
    }
  },
  {
    "ph": "f", "id": 7825, "pid": 76337, "tid": -914061504, "ts": 1716454216696794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696796, "dur": 1,
    "args": {
      "External id": 7826,
      "cbid": 205, "correlation": 7826
    }
  },
  {
    "ph": "f", "id": 7826, "pid": 76337, "tid": -914061504, "ts": 1716454216696796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696836, "dur": 1,
    "args": {
      "External id": 7830,
      "cbid": 251, "correlation": 7830
    }
  },
  {
    "ph": "f", "id": 7830, "pid": 76337, "tid": -914061504, "ts": 1716454216696836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696839, "dur": 1,
    "args": {
      "External id": 7831,
      "cbid": 251, "correlation": 7831
    }
  },
  {
    "ph": "f", "id": 7831, "pid": 76337, "tid": -914061504, "ts": 1716454216696839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696840, "dur": 0,
    "args": {
      "External id": 7832,
      "cbid": 251, "correlation": 7832
    }
  },
  {
    "ph": "f", "id": 7832, "pid": 76337, "tid": -914061504, "ts": 1716454216696840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696842, "dur": 0,
    "args": {
      "External id": 7833,
      "cbid": 251, "correlation": 7833
    }
  },
  {
    "ph": "f", "id": 7833, "pid": 76337, "tid": -914061504, "ts": 1716454216696842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696843, "dur": 1,
    "args": {
      "External id": 7834,
      "cbid": 251, "correlation": 7834
    }
  },
  {
    "ph": "f", "id": 7834, "pid": 76337, "tid": -914061504, "ts": 1716454216696843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696845, "dur": 1,
    "args": {
      "External id": 7835,
      "cbid": 251, "correlation": 7835
    }
  },
  {
    "ph": "f", "id": 7835, "pid": 76337, "tid": -914061504, "ts": 1716454216696845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696847, "dur": 1,
    "args": {
      "External id": 7836,
      "cbid": 251, "correlation": 7836
    }
  },
  {
    "ph": "f", "id": 7836, "pid": 76337, "tid": -914061504, "ts": 1716454216696847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696848, "dur": 0,
    "args": {
      "External id": 7837,
      "cbid": 251, "correlation": 7837
    }
  },
  {
    "ph": "f", "id": 7837, "pid": 76337, "tid": -914061504, "ts": 1716454216696848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696850, "dur": 0,
    "args": {
      "External id": 7838,
      "cbid": 251, "correlation": 7838
    }
  },
  {
    "ph": "f", "id": 7838, "pid": 76337, "tid": -914061504, "ts": 1716454216696850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216697216, "dur": 124,
    "args": {
      "External id": 7839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7839,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 7839, "pid": 5, "tid": 7, "ts": 1716454216697216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696853, "dur": 13,
    "args": {
      "External id": 7839,
      "cbid": 211, "correlation": 7839
    }
  },
  {
    "ph": "s", "id": 7839, "pid": 76337, "tid": -914061504, "ts": 1716454216696853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216697342, "dur": 64,
    "args": {
      "External id": 7845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7845, "pid": 5, "tid": 7, "ts": 1716454216697342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696907, "dur": 10,
    "args": {
      "External id": 7845,
      "cbid": 211, "correlation": 7845
    }
  },
  {
    "ph": "s", "id": 7845, "pid": 76337, "tid": -914061504, "ts": 1716454216696907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216697407, "dur": 50,
    "args": {
      "External id": 7853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7853,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7853, "pid": 5, "tid": 7, "ts": 1716454216697407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216696943, "dur": 9,
    "args": {
      "External id": 7853,
      "cbid": 211, "correlation": 7853
    }
  },
  {
    "ph": "s", "id": 7853, "pid": 76337, "tid": -914061504, "ts": 1716454216696943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216697458, "dur": 58,
    "args": {
      "External id": 7873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7873,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 7873, "pid": 5, "tid": 7, "ts": 1716454216697458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697079, "dur": 15,
    "args": {
      "External id": 7873,
      "cbid": 211, "correlation": 7873
    }
  },
  {
    "ph": "s", "id": 7873, "pid": 76337, "tid": -914061504, "ts": 1716454216697079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216697518, "dur": 5,
    "args": {
      "External id": 7885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7885,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7885, "pid": 5, "tid": 7, "ts": 1716454216697518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697116, "dur": 9,
    "args": {
      "External id": 7885,
      "cbid": 211, "correlation": 7885
    }
  },
  {
    "ph": "s", "id": 7885, "pid": 76337, "tid": -914061504, "ts": 1716454216697116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216697524, "dur": 59,
    "args": {
      "External id": 7888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7888, "pid": 5, "tid": 7, "ts": 1716454216697524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697140, "dur": 8,
    "args": {
      "External id": 7888,
      "cbid": 211, "correlation": 7888
    }
  },
  {
    "ph": "s", "id": 7888, "pid": 76337, "tid": -914061504, "ts": 1716454216697140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216697584, "dur": 38,
    "args": {
      "External id": 7897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7897,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7897, "pid": 5, "tid": 7, "ts": 1716454216697584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697196, "dur": 11,
    "args": {
      "External id": 7897,
      "cbid": 211, "correlation": 7897
    }
  },
  {
    "ph": "s", "id": 7897, "pid": 76337, "tid": -914061504, "ts": 1716454216697196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697268, "dur": 0,
    "args": {
      "External id": 7907,
      "cbid": 317, "correlation": 7907
    }
  },
  {
    "ph": "f", "id": 7907, "pid": 76337, "tid": -914061504, "ts": 1716454216697268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697269, "dur": 0,
    "args": {
      "External id": 7908,
      "cbid": 203, "correlation": 7908
    }
  },
  {
    "ph": "f", "id": 7908, "pid": 76337, "tid": -914061504, "ts": 1716454216697269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697270, "dur": 0,
    "args": {
      "External id": 7909,
      "cbid": 205, "correlation": 7909
    }
  },
  {
    "ph": "f", "id": 7909, "pid": 76337, "tid": -914061504, "ts": 1716454216697270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216697623, "dur": 42,
    "args": {
      "External id": 7913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7913, "pid": 5, "tid": 7, "ts": 1716454216697623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697293, "dur": 15,
    "args": {
      "External id": 7913,
      "cbid": 211, "correlation": 7913
    }
  },
  {
    "ph": "s", "id": 7913, "pid": 76337, "tid": -914061504, "ts": 1716454216697293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216697667, "dur": 16,
    "args": {
      "External id": 7915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7915, "pid": 5, "tid": 7, "ts": 1716454216697667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697311, "dur": 5,
    "args": {
      "External id": 7915,
      "cbid": 211, "correlation": 7915
    }
  },
  {
    "ph": "s", "id": 7915, "pid": 76337, "tid": -914061504, "ts": 1716454216697311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216697684, "dur": 4,
    "args": {
      "External id": 7917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7917,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7917, "pid": 5, "tid": 7, "ts": 1716454216697684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697329, "dur": 8,
    "args": {
      "External id": 7917,
      "cbid": 211, "correlation": 7917
    }
  },
  {
    "ph": "s", "id": 7917, "pid": 76337, "tid": -914061504, "ts": 1716454216697329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697343, "dur": 0,
    "args": {
      "External id": 7918,
      "cbid": 51, "correlation": 7918
    }
  },
  {
    "ph": "s", "id": 7918, "pid": 76337, "tid": -914061504, "ts": 1716454216697343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216697689, "dur": 772,
    "args": {
      "External id": 7919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7919,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 7919, "pid": 5, "tid": 7, "ts": 1716454216697689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697345, "dur": 7,
    "args": {
      "External id": 7919,
      "cbid": 211, "correlation": 7919
    }
  },
  {
    "ph": "s", "id": 7919, "pid": 76337, "tid": -914061504, "ts": 1716454216697345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216698462, "dur": 63,
    "args": {
      "External id": 7924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7924, "pid": 5, "tid": 7, "ts": 1716454216698462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697376, "dur": 10,
    "args": {
      "External id": 7924,
      "cbid": 211, "correlation": 7924
    }
  },
  {
    "ph": "s", "id": 7924, "pid": 76337, "tid": -914061504, "ts": 1716454216697376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216698527, "dur": 4,
    "args": {
      "External id": 7932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7932,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7932, "pid": 5, "tid": 7, "ts": 1716454216698527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697421, "dur": 9,
    "args": {
      "External id": 7932,
      "cbid": 211, "correlation": 7932
    }
  },
  {
    "ph": "s", "id": 7932, "pid": 76337, "tid": -914061504, "ts": 1716454216697421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697513, "dur": 2,
    "args": {
      "External id": 7948,
      "cbid": 251, "correlation": 7948
    }
  },
  {
    "ph": "f", "id": 7948, "pid": 76337, "tid": -914061504, "ts": 1716454216697513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697518, "dur": 0,
    "args": {
      "External id": 7950,
      "cbid": 251, "correlation": 7950
    }
  },
  {
    "ph": "f", "id": 7950, "pid": 76337, "tid": -914061504, "ts": 1716454216697518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216698532, "dur": 12,
    "args": {
      "External id": 7951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7951,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 7951, "pid": 5, "tid": 7, "ts": 1716454216698532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697521, "dur": 14,
    "args": {
      "External id": 7951,
      "cbid": 211, "correlation": 7951
    }
  },
  {
    "ph": "s", "id": 7951, "pid": 76337, "tid": -914061504, "ts": 1716454216697521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216698546, "dur": 5,
    "args": {
      "External id": 7953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7953,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 7953, "pid": 5, "tid": 7, "ts": 1716454216698546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697539, "dur": 7,
    "args": {
      "External id": 7953,
      "cbid": 211, "correlation": 7953
    }
  },
  {
    "ph": "s", "id": 7953, "pid": 76337, "tid": -914061504, "ts": 1716454216697539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216698553, "dur": 57,
    "args": {
      "External id": 7963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7963,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7963, "pid": 5, "tid": 7, "ts": 1716454216698553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697615, "dur": 12,
    "args": {
      "External id": 7963,
      "cbid": 211, "correlation": 7963
    }
  },
  {
    "ph": "s", "id": 7963, "pid": 76337, "tid": -914061504, "ts": 1716454216697615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216698612, "dur": 56,
    "args": {
      "External id": 7983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7983,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 7983, "pid": 5, "tid": 7, "ts": 1716454216698612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697684, "dur": 11,
    "args": {
      "External id": 7983,
      "cbid": 211, "correlation": 7983
    }
  },
  {
    "ph": "s", "id": 7983, "pid": 76337, "tid": -914061504, "ts": 1716454216697684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216698669, "dur": 4,
    "args": {
      "External id": 7995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7995,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 7995, "pid": 5, "tid": 7, "ts": 1716454216698669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697705, "dur": 6,
    "args": {
      "External id": 7995,
      "cbid": 211, "correlation": 7995
    }
  },
  {
    "ph": "s", "id": 7995, "pid": 76337, "tid": -914061504, "ts": 1716454216697705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216698675, "dur": 58,
    "args": {
      "External id": 7998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 7998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 7998, "pid": 5, "tid": 7, "ts": 1716454216698675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697723, "dur": 7,
    "args": {
      "External id": 7998,
      "cbid": 211, "correlation": 7998
    }
  },
  {
    "ph": "s", "id": 7998, "pid": 76337, "tid": -914061504, "ts": 1716454216697723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216698734, "dur": 38,
    "args": {
      "External id": 8007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8007,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8007, "pid": 5, "tid": 7, "ts": 1716454216698734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697764, "dur": 10,
    "args": {
      "External id": 8007,
      "cbid": 211, "correlation": 8007
    }
  },
  {
    "ph": "s", "id": 8007, "pid": 76337, "tid": -914061504, "ts": 1716454216697764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697842, "dur": 0,
    "args": {
      "External id": 8017,
      "cbid": 317, "correlation": 8017
    }
  },
  {
    "ph": "f", "id": 8017, "pid": 76337, "tid": -914061504, "ts": 1716454216697842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697843, "dur": 0,
    "args": {
      "External id": 8018,
      "cbid": 203, "correlation": 8018
    }
  },
  {
    "ph": "f", "id": 8018, "pid": 76337, "tid": -914061504, "ts": 1716454216697843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697844, "dur": 0,
    "args": {
      "External id": 8019,
      "cbid": 205, "correlation": 8019
    }
  },
  {
    "ph": "f", "id": 8019, "pid": 76337, "tid": -914061504, "ts": 1716454216697844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216698774, "dur": 41,
    "args": {
      "External id": 8023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8023, "pid": 5, "tid": 7, "ts": 1716454216698774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697858, "dur": 13,
    "args": {
      "External id": 8023,
      "cbid": 211, "correlation": 8023
    }
  },
  {
    "ph": "s", "id": 8023, "pid": 76337, "tid": -914061504, "ts": 1716454216697858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216698816, "dur": 15,
    "args": {
      "External id": 8025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8025,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8025, "pid": 5, "tid": 7, "ts": 1716454216698816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697874, "dur": 5,
    "args": {
      "External id": 8025,
      "cbid": 211, "correlation": 8025
    }
  },
  {
    "ph": "s", "id": 8025, "pid": 76337, "tid": -914061504, "ts": 1716454216697874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216698833, "dur": 3,
    "args": {
      "External id": 8027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 8027, "pid": 5, "tid": 7, "ts": 1716454216698833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697883, "dur": 5,
    "args": {
      "External id": 8027,
      "cbid": 211, "correlation": 8027
    }
  },
  {
    "ph": "s", "id": 8027, "pid": 76337, "tid": -914061504, "ts": 1716454216697883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697891, "dur": 0,
    "args": {
      "External id": 8028,
      "cbid": 51, "correlation": 8028
    }
  },
  {
    "ph": "s", "id": 8028, "pid": 76337, "tid": -914061504, "ts": 1716454216697891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216698838, "dur": 766,
    "args": {
      "External id": 8029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8029,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8029, "pid": 5, "tid": 7, "ts": 1716454216698838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697892, "dur": 5,
    "args": {
      "External id": 8029,
      "cbid": 211, "correlation": 8029
    }
  },
  {
    "ph": "s", "id": 8029, "pid": 76337, "tid": -914061504, "ts": 1716454216697892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216699606, "dur": 63,
    "args": {
      "External id": 8034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8034, "pid": 5, "tid": 7, "ts": 1716454216699606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697920, "dur": 9,
    "args": {
      "External id": 8034,
      "cbid": 211, "correlation": 8034
    }
  },
  {
    "ph": "s", "id": 8034, "pid": 76337, "tid": -914061504, "ts": 1716454216697920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216699670, "dur": 50,
    "args": {
      "External id": 8042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8042,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8042, "pid": 5, "tid": 7, "ts": 1716454216699670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697953, "dur": 9,
    "args": {
      "External id": 8042,
      "cbid": 211, "correlation": 8042
    }
  },
  {
    "ph": "s", "id": 8042, "pid": 76337, "tid": -914061504, "ts": 1716454216697953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216699721, "dur": 35,
    "args": {
      "External id": 8050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8050,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8050, "pid": 5, "tid": 7, "ts": 1716454216699721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216697999, "dur": 10,
    "args": {
      "External id": 8050,
      "cbid": 211, "correlation": 8050
    }
  },
  {
    "ph": "s", "id": 8050, "pid": 76337, "tid": -914061504, "ts": 1716454216697999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216699758, "dur": 57,
    "args": {
      "External id": 8070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8070,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 8070, "pid": 5, "tid": 7, "ts": 1716454216699758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698099, "dur": 14,
    "args": {
      "External id": 8070,
      "cbid": 211, "correlation": 8070
    }
  },
  {
    "ph": "s", "id": 8070, "pid": 76337, "tid": -914061504, "ts": 1716454216698099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216699816, "dur": 4,
    "args": {
      "External id": 8082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8082,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 8082, "pid": 5, "tid": 7, "ts": 1716454216699816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698129, "dur": 8,
    "args": {
      "External id": 8082,
      "cbid": 211, "correlation": 8082
    }
  },
  {
    "ph": "s", "id": 8082, "pid": 76337, "tid": -914061504, "ts": 1716454216698129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216699822, "dur": 60,
    "args": {
      "External id": 8085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8085, "pid": 5, "tid": 7, "ts": 1716454216699822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698149, "dur": 7,
    "args": {
      "External id": 8085,
      "cbid": 211, "correlation": 8085
    }
  },
  {
    "ph": "s", "id": 8085, "pid": 76337, "tid": -914061504, "ts": 1716454216698149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698218, "dur": 0,
    "args": {
      "External id": 8096,
      "cbid": 317, "correlation": 8096
    }
  },
  {
    "ph": "f", "id": 8096, "pid": 76337, "tid": -914061504, "ts": 1716454216698218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698219, "dur": 0,
    "args": {
      "External id": 8097,
      "cbid": 203, "correlation": 8097
    }
  },
  {
    "ph": "f", "id": 8097, "pid": 76337, "tid": -914061504, "ts": 1716454216698219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698220, "dur": 0,
    "args": {
      "External id": 8098,
      "cbid": 205, "correlation": 8098
    }
  },
  {
    "ph": "f", "id": 8098, "pid": 76337, "tid": -914061504, "ts": 1716454216698220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698242, "dur": 1,
    "args": {
      "External id": 8102,
      "cbid": 251, "correlation": 8102
    }
  },
  {
    "ph": "f", "id": 8102, "pid": 76337, "tid": -914061504, "ts": 1716454216698242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698244, "dur": 0,
    "args": {
      "External id": 8103,
      "cbid": 251, "correlation": 8103
    }
  },
  {
    "ph": "f", "id": 8103, "pid": 76337, "tid": -914061504, "ts": 1716454216698244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698245, "dur": 0,
    "args": {
      "External id": 8104,
      "cbid": 251, "correlation": 8104
    }
  },
  {
    "ph": "f", "id": 8104, "pid": 76337, "tid": -914061504, "ts": 1716454216698245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698246, "dur": 0,
    "args": {
      "External id": 8105,
      "cbid": 251, "correlation": 8105
    }
  },
  {
    "ph": "f", "id": 8105, "pid": 76337, "tid": -914061504, "ts": 1716454216698246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698246, "dur": 0,
    "args": {
      "External id": 8106,
      "cbid": 251, "correlation": 8106
    }
  },
  {
    "ph": "f", "id": 8106, "pid": 76337, "tid": -914061504, "ts": 1716454216698246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698247, "dur": 0,
    "args": {
      "External id": 8107,
      "cbid": 251, "correlation": 8107
    }
  },
  {
    "ph": "f", "id": 8107, "pid": 76337, "tid": -914061504, "ts": 1716454216698247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698248, "dur": 0,
    "args": {
      "External id": 8108,
      "cbid": 251, "correlation": 8108
    }
  },
  {
    "ph": "f", "id": 8108, "pid": 76337, "tid": -914061504, "ts": 1716454216698248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698248, "dur": 0,
    "args": {
      "External id": 8109,
      "cbid": 251, "correlation": 8109
    }
  },
  {
    "ph": "f", "id": 8109, "pid": 76337, "tid": -914061504, "ts": 1716454216698248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698250, "dur": 0,
    "args": {
      "External id": 8110,
      "cbid": 251, "correlation": 8110
    }
  },
  {
    "ph": "f", "id": 8110, "pid": 76337, "tid": -914061504, "ts": 1716454216698250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216699883, "dur": 122,
    "args": {
      "External id": 8111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8111,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8111, "pid": 5, "tid": 7, "ts": 1716454216699883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698252, "dur": 13,
    "args": {
      "External id": 8111,
      "cbid": 211, "correlation": 8111
    }
  },
  {
    "ph": "s", "id": 8111, "pid": 76337, "tid": -914061504, "ts": 1716454216698252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216700007, "dur": 63,
    "args": {
      "External id": 8117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8117,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8117, "pid": 5, "tid": 7, "ts": 1716454216700007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698288, "dur": 9,
    "args": {
      "External id": 8117,
      "cbid": 211, "correlation": 8117
    }
  },
  {
    "ph": "s", "id": 8117, "pid": 76337, "tid": -914061504, "ts": 1716454216698288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216700071, "dur": 163,
    "args": {
      "External id": 8128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8128,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8128, "pid": 5, "tid": 7, "ts": 1716454216700071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698399, "dur": 15,
    "args": {
      "External id": 8128,
      "cbid": 211, "correlation": 8128
    }
  },
  {
    "ph": "s", "id": 8128, "pid": 76337, "tid": -914061504, "ts": 1716454216698399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216700235, "dur": 103,
    "args": {
      "External id": 8150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8150,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8150, "pid": 5, "tid": 7, "ts": 1716454216700235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698435, "dur": 8,
    "args": {
      "External id": 8150,
      "cbid": 211, "correlation": 8150
    }
  },
  {
    "ph": "s", "id": 8150, "pid": 76337, "tid": -914061504, "ts": 1716454216698435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698539, "dur": 1,
    "args": {
      "External id": 8161,
      "cbid": 251, "correlation": 8161
    }
  },
  {
    "ph": "f", "id": 8161, "pid": 76337, "tid": -914061504, "ts": 1716454216698539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216700340, "dur": 121,
    "args": {
      "External id": 8162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8162,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8162, "pid": 5, "tid": 7, "ts": 1716454216700340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698544, "dur": 13,
    "args": {
      "External id": 8162,
      "cbid": 211, "correlation": 8162
    }
  },
  {
    "ph": "s", "id": 8162, "pid": 76337, "tid": -914061504, "ts": 1716454216698544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698628, "dur": 1,
    "args": {
      "External id": 8173,
      "cbid": 251, "correlation": 8173
    }
  },
  {
    "ph": "f", "id": 8173, "pid": 76337, "tid": -914061504, "ts": 1716454216698628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216700462, "dur": 113,
    "args": {
      "External id": 8174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8174,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8174, "pid": 5, "tid": 7, "ts": 1716454216700462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698632, "dur": 12,
    "args": {
      "External id": 8174,
      "cbid": 211, "correlation": 8174
    }
  },
  {
    "ph": "s", "id": 8174, "pid": 76337, "tid": -914061504, "ts": 1716454216698632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698710, "dur": 1,
    "args": {
      "External id": 8185,
      "cbid": 251, "correlation": 8185
    }
  },
  {
    "ph": "f", "id": 8185, "pid": 76337, "tid": -914061504, "ts": 1716454216698710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216700576, "dur": 113,
    "args": {
      "External id": 8186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8186,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8186, "pid": 5, "tid": 7, "ts": 1716454216700576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698715, "dur": 12,
    "args": {
      "External id": 8186,
      "cbid": 211, "correlation": 8186
    }
  },
  {
    "ph": "s", "id": 8186, "pid": 76337, "tid": -914061504, "ts": 1716454216698715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216700691, "dur": 5237,
    "args": {
      "External id": 8207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8207,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8207, "pid": 5, "tid": 7, "ts": 1716454216700691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698813, "dur": 15,
    "args": {
      "External id": 8207,
      "cbid": 211, "correlation": 8207
    }
  },
  {
    "ph": "s", "id": 8207, "pid": 76337, "tid": -914061504, "ts": 1716454216698813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698941, "dur": 1,
    "args": {
      "External id": 8225,
      "cbid": 251, "correlation": 8225
    }
  },
  {
    "ph": "f", "id": 8225, "pid": 76337, "tid": -914061504, "ts": 1716454216698941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216705929, "dur": 114,
    "args": {
      "External id": 8227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8227,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8227, "pid": 5, "tid": 7, "ts": 1716454216705929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216698947, "dur": 14,
    "args": {
      "External id": 8227,
      "cbid": 211, "correlation": 8227
    }
  },
  {
    "ph": "s", "id": 8227, "pid": 76337, "tid": -914061504, "ts": 1716454216698947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216706045, "dur": 35,
    "args": {
      "External id": 8235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8235,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8235, "pid": 5, "tid": 7, "ts": 1716454216706045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699035, "dur": 13,
    "args": {
      "External id": 8235,
      "cbid": 211, "correlation": 8235
    }
  },
  {
    "ph": "s", "id": 8235, "pid": 76337, "tid": -914061504, "ts": 1716454216699035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216706081, "dur": 171,
    "args": {
      "External id": 8243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8243, "pid": 5, "tid": 7, "ts": 1716454216706081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699078, "dur": 9,
    "args": {
      "External id": 8243,
      "cbid": 211, "correlation": 8243
    }
  },
  {
    "ph": "s", "id": 8243, "pid": 76337, "tid": -914061504, "ts": 1716454216699078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216706254, "dur": 102,
    "args": {
      "External id": 8265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8265,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8265, "pid": 5, "tid": 7, "ts": 1716454216706254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699132, "dur": 10,
    "args": {
      "External id": 8265,
      "cbid": 211, "correlation": 8265
    }
  },
  {
    "ph": "s", "id": 8265, "pid": 76337, "tid": -914061504, "ts": 1716454216699132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699218, "dur": 1,
    "args": {
      "External id": 8276,
      "cbid": 251, "correlation": 8276
    }
  },
  {
    "ph": "f", "id": 8276, "pid": 76337, "tid": -914061504, "ts": 1716454216699218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216706357, "dur": 115,
    "args": {
      "External id": 8277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8277,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8277, "pid": 5, "tid": 7, "ts": 1716454216706357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699224, "dur": 13,
    "args": {
      "External id": 8277,
      "cbid": 211, "correlation": 8277
    }
  },
  {
    "ph": "s", "id": 8277, "pid": 76337, "tid": -914061504, "ts": 1716454216699224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699302, "dur": 1,
    "args": {
      "External id": 8288,
      "cbid": 251, "correlation": 8288
    }
  },
  {
    "ph": "f", "id": 8288, "pid": 76337, "tid": -914061504, "ts": 1716454216699302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699305, "dur": 0,
    "args": {
      "External id": 8289,
      "cbid": 251, "correlation": 8289
    }
  },
  {
    "ph": "f", "id": 8289, "pid": 76337, "tid": -914061504, "ts": 1716454216699305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216706473, "dur": 10,
    "args": {
      "External id": 8290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8290,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 8290, "pid": 5, "tid": 7, "ts": 1716454216706473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699307, "dur": 14,
    "args": {
      "External id": 8290,
      "cbid": 211, "correlation": 8290
    }
  },
  {
    "ph": "s", "id": 8290, "pid": 76337, "tid": -914061504, "ts": 1716454216699307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216706485, "dur": 5,
    "args": {
      "External id": 8292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8292,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 8292, "pid": 5, "tid": 7, "ts": 1716454216706485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699324, "dur": 6,
    "args": {
      "External id": 8292,
      "cbid": 211, "correlation": 8292
    }
  },
  {
    "ph": "s", "id": 8292, "pid": 76337, "tid": -914061504, "ts": 1716454216699324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699385, "dur": 1,
    "args": {
      "External id": 8303,
      "cbid": 251, "correlation": 8303
    }
  },
  {
    "ph": "f", "id": 8303, "pid": 76337, "tid": -914061504, "ts": 1716454216699385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699389, "dur": 0,
    "args": {
      "External id": 8304,
      "cbid": 251, "correlation": 8304
    }
  },
  {
    "ph": "f", "id": 8304, "pid": 76337, "tid": -914061504, "ts": 1716454216699389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216706492, "dur": 7,
    "args": {
      "External id": 8305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8305,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 8305, "pid": 5, "tid": 7, "ts": 1716454216706492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699390, "dur": 12,
    "args": {
      "External id": 8305,
      "cbid": 211, "correlation": 8305
    }
  },
  {
    "ph": "s", "id": 8305, "pid": 76337, "tid": -914061504, "ts": 1716454216699390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216706500, "dur": 4,
    "args": {
      "External id": 8307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8307,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 8307, "pid": 5, "tid": 7, "ts": 1716454216706500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699404, "dur": 5,
    "args": {
      "External id": 8307,
      "cbid": 211, "correlation": 8307
    }
  },
  {
    "ph": "s", "id": 8307, "pid": 76337, "tid": -914061504, "ts": 1716454216699404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216706505, "dur": 169,
    "args": {
      "External id": 8328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8328,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8328, "pid": 5, "tid": 7, "ts": 1716454216706505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699478, "dur": 12,
    "args": {
      "External id": 8328,
      "cbid": 211, "correlation": 8328
    }
  },
  {
    "ph": "s", "id": 8328, "pid": 76337, "tid": -914061504, "ts": 1716454216699478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699596, "dur": 1,
    "args": {
      "External id": 8346,
      "cbid": 251, "correlation": 8346
    }
  },
  {
    "ph": "f", "id": 8346, "pid": 76337, "tid": -914061504, "ts": 1716454216699596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216706675, "dur": 114,
    "args": {
      "External id": 8348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8348,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8348, "pid": 5, "tid": 7, "ts": 1716454216706675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699602, "dur": 14,
    "args": {
      "External id": 8348,
      "cbid": 211, "correlation": 8348
    }
  },
  {
    "ph": "s", "id": 8348, "pid": 76337, "tid": -914061504, "ts": 1716454216699602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216706791, "dur": 35,
    "args": {
      "External id": 8356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8356,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8356, "pid": 5, "tid": 7, "ts": 1716454216706791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699676, "dur": 12,
    "args": {
      "External id": 8356,
      "cbid": 211, "correlation": 8356
    }
  },
  {
    "ph": "s", "id": 8356, "pid": 76337, "tid": -914061504, "ts": 1716454216699676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216706828, "dur": 51,
    "args": {
      "External id": 8364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8364,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8364, "pid": 5, "tid": 7, "ts": 1716454216706828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699716, "dur": 9,
    "args": {
      "External id": 8364,
      "cbid": 211, "correlation": 8364
    }
  },
  {
    "ph": "s", "id": 8364, "pid": 76337, "tid": -914061504, "ts": 1716454216699716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216706880, "dur": 102,
    "args": {
      "External id": 8386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8386,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8386, "pid": 5, "tid": 7, "ts": 1716454216706880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699768, "dur": 10,
    "args": {
      "External id": 8386,
      "cbid": 211, "correlation": 8386
    }
  },
  {
    "ph": "s", "id": 8386, "pid": 76337, "tid": -914061504, "ts": 1716454216699768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699859, "dur": 2,
    "args": {
      "External id": 8402,
      "cbid": 251, "correlation": 8402
    }
  },
  {
    "ph": "f", "id": 8402, "pid": 76337, "tid": -914061504, "ts": 1716454216699859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216706983, "dur": 629,
    "args": {
      "External id": 8404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8404,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8404, "pid": 5, "tid": 7, "ts": 1716454216706983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699865, "dur": 14,
    "args": {
      "External id": 8404,
      "cbid": 211, "correlation": 8404
    }
  },
  {
    "ph": "s", "id": 8404, "pid": 76337, "tid": -914061504, "ts": 1716454216699865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216707613, "dur": 260,
    "args": {
      "External id": 8412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8412,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8412, "pid": 5, "tid": 7, "ts": 1716454216707613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699953, "dur": 13,
    "args": {
      "External id": 8412,
      "cbid": 211, "correlation": 8412
    }
  },
  {
    "ph": "s", "id": 8412, "pid": 76337, "tid": -914061504, "ts": 1716454216699953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216707875, "dur": 261,
    "args": {
      "External id": 8420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8420, "pid": 5, "tid": 7, "ts": 1716454216707875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216699999, "dur": 10,
    "args": {
      "External id": 8420,
      "cbid": 211, "correlation": 8420
    }
  },
  {
    "ph": "s", "id": 8420, "pid": 76337, "tid": -914061504, "ts": 1716454216699999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700099, "dur": 2,
    "args": {
      "External id": 8436,
      "cbid": 251, "correlation": 8436
    }
  },
  {
    "ph": "f", "id": 8436, "pid": 76337, "tid": -914061504, "ts": 1716454216700099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700104, "dur": 0,
    "args": {
      "External id": 8438,
      "cbid": 251, "correlation": 8438
    }
  },
  {
    "ph": "f", "id": 8438, "pid": 76337, "tid": -914061504, "ts": 1716454216700104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216708138, "dur": 380,
    "args": {
      "External id": 8439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8439,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 8439, "pid": 5, "tid": 7, "ts": 1716454216708138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700108, "dur": 14,
    "args": {
      "External id": 8439,
      "cbid": 211, "correlation": 8439
    }
  },
  {
    "ph": "s", "id": 8439, "pid": 76337, "tid": -914061504, "ts": 1716454216700108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216708519, "dur": 51,
    "args": {
      "External id": 8447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8447,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8447, "pid": 5, "tid": 7, "ts": 1716454216708519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700152, "dur": 11,
    "args": {
      "External id": 8447,
      "cbid": 211, "correlation": 8447
    }
  },
  {
    "ph": "s", "id": 8447, "pid": 76337, "tid": -914061504, "ts": 1716454216700152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216708572, "dur": 173,
    "args": {
      "External id": 8458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8458, "pid": 5, "tid": 7, "ts": 1716454216708572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700218, "dur": 12,
    "args": {
      "External id": 8458,
      "cbid": 211, "correlation": 8458
    }
  },
  {
    "ph": "s", "id": 8458, "pid": 76337, "tid": -914061504, "ts": 1716454216700218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700296, "dur": 0,
    "args": {
      "External id": 8470,
      "cbid": 317, "correlation": 8470
    }
  },
  {
    "ph": "f", "id": 8470, "pid": 76337, "tid": -914061504, "ts": 1716454216700296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700297, "dur": 0,
    "args": {
      "External id": 8471,
      "cbid": 203, "correlation": 8471
    }
  },
  {
    "ph": "f", "id": 8471, "pid": 76337, "tid": -914061504, "ts": 1716454216700297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700298, "dur": 0,
    "args": {
      "External id": 8472,
      "cbid": 205, "correlation": 8472
    }
  },
  {
    "ph": "f", "id": 8472, "pid": 76337, "tid": -914061504, "ts": 1716454216700298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700321, "dur": 1,
    "args": {
      "External id": 8476,
      "cbid": 251, "correlation": 8476
    }
  },
  {
    "ph": "f", "id": 8476, "pid": 76337, "tid": -914061504, "ts": 1716454216700321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700323, "dur": 0,
    "args": {
      "External id": 8477,
      "cbid": 251, "correlation": 8477
    }
  },
  {
    "ph": "f", "id": 8477, "pid": 76337, "tid": -914061504, "ts": 1716454216700323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700324, "dur": 0,
    "args": {
      "External id": 8478,
      "cbid": 251, "correlation": 8478
    }
  },
  {
    "ph": "f", "id": 8478, "pid": 76337, "tid": -914061504, "ts": 1716454216700324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700324, "dur": 0,
    "args": {
      "External id": 8479,
      "cbid": 251, "correlation": 8479
    }
  },
  {
    "ph": "f", "id": 8479, "pid": 76337, "tid": -914061504, "ts": 1716454216700324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700325, "dur": 0,
    "args": {
      "External id": 8480,
      "cbid": 251, "correlation": 8480
    }
  },
  {
    "ph": "f", "id": 8480, "pid": 76337, "tid": -914061504, "ts": 1716454216700325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700326, "dur": 0,
    "args": {
      "External id": 8481,
      "cbid": 251, "correlation": 8481
    }
  },
  {
    "ph": "f", "id": 8481, "pid": 76337, "tid": -914061504, "ts": 1716454216700326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700327, "dur": 0,
    "args": {
      "External id": 8482,
      "cbid": 251, "correlation": 8482
    }
  },
  {
    "ph": "f", "id": 8482, "pid": 76337, "tid": -914061504, "ts": 1716454216700327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700328, "dur": 0,
    "args": {
      "External id": 8483,
      "cbid": 251, "correlation": 8483
    }
  },
  {
    "ph": "f", "id": 8483, "pid": 76337, "tid": -914061504, "ts": 1716454216700328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700329, "dur": 0,
    "args": {
      "External id": 8484,
      "cbid": 251, "correlation": 8484
    }
  },
  {
    "ph": "f", "id": 8484, "pid": 76337, "tid": -914061504, "ts": 1716454216700329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216708746, "dur": 125,
    "args": {
      "External id": 8485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8485,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8485, "pid": 5, "tid": 7, "ts": 1716454216708746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700331, "dur": 14,
    "args": {
      "External id": 8485,
      "cbid": 211, "correlation": 8485
    }
  },
  {
    "ph": "s", "id": 8485, "pid": 76337, "tid": -914061504, "ts": 1716454216700331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216708873, "dur": 64,
    "args": {
      "External id": 8491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8491, "pid": 5, "tid": 7, "ts": 1716454216708873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700368, "dur": 9,
    "args": {
      "External id": 8491,
      "cbid": 211, "correlation": 8491
    }
  },
  {
    "ph": "s", "id": 8491, "pid": 76337, "tid": -914061504, "ts": 1716454216700368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216708938, "dur": 51,
    "args": {
      "External id": 8499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8499,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8499, "pid": 5, "tid": 7, "ts": 1716454216708938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700401, "dur": 9,
    "args": {
      "External id": 8499,
      "cbid": 211, "correlation": 8499
    }
  },
  {
    "ph": "s", "id": 8499, "pid": 76337, "tid": -914061504, "ts": 1716454216700401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700477, "dur": 0,
    "args": {
      "External id": 8509,
      "cbid": 317, "correlation": 8509
    }
  },
  {
    "ph": "f", "id": 8509, "pid": 76337, "tid": -914061504, "ts": 1716454216700477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700477, "dur": 0,
    "args": {
      "External id": 8510,
      "cbid": 203, "correlation": 8510
    }
  },
  {
    "ph": "f", "id": 8510, "pid": 76337, "tid": -914061504, "ts": 1716454216700477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216700478, "dur": 0,
    "args": {
      "External id": 8511,
      "cbid": 205, "correlation": 8511
    }
  },
  {
    "ph": "f", "id": 8511, "pid": 76337, "tid": -914061504, "ts": 1716454216700478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216708991, "dur": 42,
    "args": {
      "External id": 8515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8515,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8515, "pid": 5, "tid": 7, "ts": 1716454216708991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216703922, "dur": 31,
    "args": {
      "External id": 8515,
      "cbid": 211, "correlation": 8515
    }
  },
  {
    "ph": "s", "id": 8515, "pid": 76337, "tid": -914061504, "ts": 1716454216703922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216709034, "dur": 15,
    "args": {
      "External id": 8517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8517,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8517, "pid": 5, "tid": 7, "ts": 1716454216709034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216703957, "dur": 6,
    "args": {
      "External id": 8517,
      "cbid": 211, "correlation": 8517
    }
  },
  {
    "ph": "s", "id": 8517, "pid": 76337, "tid": -914061504, "ts": 1716454216703957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216709052, "dur": 2,
    "args": {
      "External id": 8519,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 8519,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.6153846153846154
    }
  },
  {
    "ph": "f", "id": 8519, "pid": 5, "tid": 7, "ts": 1716454216709052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216703982, "dur": 18,
    "args": {
      "External id": 8519,
      "cbid": 51, "correlation": 8519
    }
  },
  {
    "ph": "s", "id": 8519, "pid": 76337, "tid": -914061504, "ts": 1716454216703982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216709071, "dur": 388,
    "args": {
      "External id": 8520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8520,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8520, "pid": 5, "tid": 7, "ts": 1716454216709071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216704003, "dur": 171,
    "args": {
      "External id": 8520,
      "cbid": 211, "correlation": 8520
    }
  },
  {
    "ph": "s", "id": 8520, "pid": 76337, "tid": -914061504, "ts": 1716454216704003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216709461, "dur": 13,
    "args": {
      "External id": 8522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8522,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8522, "pid": 5, "tid": 7, "ts": 1716454216709461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216704183, "dur": 8,
    "args": {
      "External id": 8522,
      "cbid": 211, "correlation": 8522
    }
  },
  {
    "ph": "s", "id": 8522, "pid": 76337, "tid": -914061504, "ts": 1716454216704183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216709476, "dur": 18,
    "args": {
      "External id": 8528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8528,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8528, "pid": 5, "tid": 7, "ts": 1716454216709476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216704987, "dur": 13,
    "args": {
      "External id": 8528,
      "cbid": 211, "correlation": 8528
    }
  },
  {
    "ph": "s", "id": 8528, "pid": 76337, "tid": -914061504, "ts": 1716454216704987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216709495, "dur": 20,
    "args": {
      "External id": 8548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8548,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 8548, "pid": 5, "tid": 7, "ts": 1716454216709495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705175, "dur": 15,
    "args": {
      "External id": 8548,
      "cbid": 211, "correlation": 8548
    }
  },
  {
    "ph": "s", "id": 8548, "pid": 76337, "tid": -914061504, "ts": 1716454216705175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216709516, "dur": 5,
    "args": {
      "External id": 8560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8560,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 8560, "pid": 5, "tid": 7, "ts": 1716454216709516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705235, "dur": 10,
    "args": {
      "External id": 8560,
      "cbid": 211, "correlation": 8560
    }
  },
  {
    "ph": "s", "id": 8560, "pid": 76337, "tid": -914061504, "ts": 1716454216705235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216709523, "dur": 19,
    "args": {
      "External id": 8563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8563, "pid": 5, "tid": 7, "ts": 1716454216709523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705265, "dur": 7,
    "args": {
      "External id": 8563,
      "cbid": 211, "correlation": 8563
    }
  },
  {
    "ph": "s", "id": 8563, "pid": 76337, "tid": -914061504, "ts": 1716454216705265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216709543, "dur": 12,
    "args": {
      "External id": 8572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8572,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8572, "pid": 5, "tid": 7, "ts": 1716454216709543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705320, "dur": 12,
    "args": {
      "External id": 8572,
      "cbid": 211, "correlation": 8572
    }
  },
  {
    "ph": "s", "id": 8572, "pid": 76337, "tid": -914061504, "ts": 1716454216705320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705420, "dur": 0,
    "args": {
      "External id": 8582,
      "cbid": 317, "correlation": 8582
    }
  },
  {
    "ph": "f", "id": 8582, "pid": 76337, "tid": -914061504, "ts": 1716454216705420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705421, "dur": 1,
    "args": {
      "External id": 8583,
      "cbid": 203, "correlation": 8583
    }
  },
  {
    "ph": "f", "id": 8583, "pid": 76337, "tid": -914061504, "ts": 1716454216705421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216705423, "dur": 0,
    "args": {
      "External id": 8584,
      "cbid": 205, "correlation": 8584
    }
  },
  {
    "ph": "f", "id": 8584, "pid": 76337, "tid": -914061504, "ts": 1716454216705423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216709557, "dur": 14,
    "args": {
      "External id": 8588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8588,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8588, "pid": 5, "tid": 7, "ts": 1716454216709557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216708166, "dur": 23,
    "args": {
      "External id": 8588,
      "cbid": 211, "correlation": 8588
    }
  },
  {
    "ph": "s", "id": 8588, "pid": 76337, "tid": -914061504, "ts": 1716454216708166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216709572, "dur": 26,
    "args": {
      "External id": 8590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8590,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8590, "pid": 5, "tid": 7, "ts": 1716454216709572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216708192, "dur": 6,
    "args": {
      "External id": 8590,
      "cbid": 211, "correlation": 8590
    }
  },
  {
    "ph": "s", "id": 8590, "pid": 76337, "tid": -914061504, "ts": 1716454216708192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216709600, "dur": 4,
    "args": {
      "External id": 8592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 8592, "pid": 5, "tid": 7, "ts": 1716454216709600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216708205, "dur": 10,
    "args": {
      "External id": 8592,
      "cbid": 211, "correlation": 8592
    }
  },
  {
    "ph": "s", "id": 8592, "pid": 76337, "tid": -914061504, "ts": 1716454216708205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216708218, "dur": 0,
    "args": {
      "External id": 8593,
      "cbid": 51, "correlation": 8593
    }
  },
  {
    "ph": "s", "id": 8593, "pid": 76337, "tid": -914061504, "ts": 1716454216708218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216709615, "dur": 384,
    "args": {
      "External id": 8594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8594,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8594, "pid": 5, "tid": 7, "ts": 1716454216709615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216708235, "dur": 120,
    "args": {
      "External id": 8594,
      "cbid": 211, "correlation": 8594
    }
  },
  {
    "ph": "s", "id": 8594, "pid": 76337, "tid": -914061504, "ts": 1716454216708235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216710000, "dur": 22,
    "args": {
      "External id": 8595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8595,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8595, "pid": 5, "tid": 7, "ts": 1716454216710000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216708360, "dur": 7,
    "args": {
      "External id": 8595,
      "cbid": 211, "correlation": 8595
    }
  },
  {
    "ph": "s", "id": 8595, "pid": 76337, "tid": -914061504, "ts": 1716454216708360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216710023, "dur": 35,
    "args": {
      "External id": 8601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8601,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8601, "pid": 5, "tid": 7, "ts": 1716454216710023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709074, "dur": 13,
    "args": {
      "External id": 8601,
      "cbid": 211, "correlation": 8601
    }
  },
  {
    "ph": "s", "id": 8601, "pid": 76337, "tid": -914061504, "ts": 1716454216709074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216710060, "dur": 4,
    "args": {
      "External id": 8609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8609,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 8609, "pid": 5, "tid": 7, "ts": 1716454216710060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709149, "dur": 10,
    "args": {
      "External id": 8609,
      "cbid": 211, "correlation": 8609
    }
  },
  {
    "ph": "s", "id": 8609, "pid": 76337, "tid": -914061504, "ts": 1716454216709149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709304, "dur": 5,
    "args": {
      "External id": 8625,
      "cbid": 251, "correlation": 8625
    }
  },
  {
    "ph": "f", "id": 8625, "pid": 76337, "tid": -914061504, "ts": 1716454216709304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709315, "dur": 0,
    "args": {
      "External id": 8627,
      "cbid": 251, "correlation": 8627
    }
  },
  {
    "ph": "f", "id": 8627, "pid": 76337, "tid": -914061504, "ts": 1716454216709315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216710065, "dur": 13,
    "args": {
      "External id": 8628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8628,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 8628, "pid": 5, "tid": 7, "ts": 1716454216710065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709320, "dur": 17,
    "args": {
      "External id": 8628,
      "cbid": 211, "correlation": 8628
    }
  },
  {
    "ph": "s", "id": 8628, "pid": 76337, "tid": -914061504, "ts": 1716454216709320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216710080, "dur": 5,
    "args": {
      "External id": 8630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8630,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 8630, "pid": 5, "tid": 7, "ts": 1716454216710080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709342, "dur": 8,
    "args": {
      "External id": 8630,
      "cbid": 211, "correlation": 8630
    }
  },
  {
    "ph": "s", "id": 8630, "pid": 76337, "tid": -914061504, "ts": 1716454216709342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216710087, "dur": 31,
    "args": {
      "External id": 8640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8640, "pid": 5, "tid": 7, "ts": 1716454216710087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709437, "dur": 12,
    "args": {
      "External id": 8640,
      "cbid": 211, "correlation": 8640
    }
  },
  {
    "ph": "s", "id": 8640, "pid": 76337, "tid": -914061504, "ts": 1716454216709437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216710119, "dur": 34,
    "args": {
      "External id": 8660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8660,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 8660, "pid": 5, "tid": 7, "ts": 1716454216710119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709529, "dur": 12,
    "args": {
      "External id": 8660,
      "cbid": 211, "correlation": 8660
    }
  },
  {
    "ph": "s", "id": 8660, "pid": 76337, "tid": -914061504, "ts": 1716454216709529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216710155, "dur": 5,
    "args": {
      "External id": 8672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8672,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 8672, "pid": 5, "tid": 7, "ts": 1716454216710155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709564, "dur": 9,
    "args": {
      "External id": 8672,
      "cbid": 211, "correlation": 8672
    }
  },
  {
    "ph": "s", "id": 8672, "pid": 76337, "tid": -914061504, "ts": 1716454216709564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216710162, "dur": 33,
    "args": {
      "External id": 8675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8675, "pid": 5, "tid": 7, "ts": 1716454216710162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709592, "dur": 7,
    "args": {
      "External id": 8675,
      "cbid": 211, "correlation": 8675
    }
  },
  {
    "ph": "s", "id": 8675, "pid": 76337, "tid": -914061504, "ts": 1716454216709592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216710196, "dur": 23,
    "args": {
      "External id": 8684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8684,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8684, "pid": 5, "tid": 7, "ts": 1716454216710196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709639, "dur": 10,
    "args": {
      "External id": 8684,
      "cbid": 211, "correlation": 8684
    }
  },
  {
    "ph": "s", "id": 8684, "pid": 76337, "tid": -914061504, "ts": 1716454216709639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709747, "dur": 0,
    "args": {
      "External id": 8694,
      "cbid": 317, "correlation": 8694
    }
  },
  {
    "ph": "f", "id": 8694, "pid": 76337, "tid": -914061504, "ts": 1716454216709747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709748, "dur": 1,
    "args": {
      "External id": 8695,
      "cbid": 203, "correlation": 8695
    }
  },
  {
    "ph": "f", "id": 8695, "pid": 76337, "tid": -914061504, "ts": 1716454216709748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216709750, "dur": 1,
    "args": {
      "External id": 8696,
      "cbid": 205, "correlation": 8696
    }
  },
  {
    "ph": "f", "id": 8696, "pid": 76337, "tid": -914061504, "ts": 1716454216709750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216712348, "dur": 26,
    "args": {
      "External id": 8700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8700, "pid": 5, "tid": 7, "ts": 1716454216712348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216712327, "dur": 21,
    "args": {
      "External id": 8700,
      "cbid": 211, "correlation": 8700
    }
  },
  {
    "ph": "s", "id": 8700, "pid": 76337, "tid": -914061504, "ts": 1716454216712327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216712375, "dur": 49,
    "args": {
      "External id": 8702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8702,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8702, "pid": 5, "tid": 7, "ts": 1716454216712375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216712351, "dur": 6,
    "args": {
      "External id": 8702,
      "cbid": 211, "correlation": 8702
    }
  },
  {
    "ph": "s", "id": 8702, "pid": 76337, "tid": -914061504, "ts": 1716454216712351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216712425, "dur": 723,
    "args": {
      "External id": 8704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8704,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8704, "pid": 5, "tid": 7, "ts": 1716454216712425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216712369, "dur": 11,
    "args": {
      "External id": 8704,
      "cbid": 211, "correlation": 8704
    }
  },
  {
    "ph": "s", "id": 8704, "pid": 76337, "tid": -914061504, "ts": 1716454216712369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216713149, "dur": 21,
    "args": {
      "External id": 8706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8706, "pid": 5, "tid": 7, "ts": 1716454216713149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216712386, "dur": 7,
    "args": {
      "External id": 8706,
      "cbid": 211, "correlation": 8706
    }
  },
  {
    "ph": "s", "id": 8706, "pid": 76337, "tid": -914061504, "ts": 1716454216712386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216713172, "dur": 35,
    "args": {
      "External id": 8712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8712, "pid": 5, "tid": 7, "ts": 1716454216713172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216713094, "dur": 14,
    "args": {
      "External id": 8712,
      "cbid": 211, "correlation": 8712
    }
  },
  {
    "ph": "s", "id": 8712, "pid": 76337, "tid": -914061504, "ts": 1716454216713094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216713186, "dur": 0,
    "args": {
      "External id": 8722,
      "cbid": 317, "correlation": 8722
    }
  },
  {
    "ph": "f", "id": 8722, "pid": 76337, "tid": -914061504, "ts": 1716454216713186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216713187, "dur": 0,
    "args": {
      "External id": 8723,
      "cbid": 203, "correlation": 8723
    }
  },
  {
    "ph": "f", "id": 8723, "pid": 76337, "tid": -914061504, "ts": 1716454216713187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216713188, "dur": 1,
    "args": {
      "External id": 8724,
      "cbid": 205, "correlation": 8724
    }
  },
  {
    "ph": "f", "id": 8724, "pid": 76337, "tid": -914061504, "ts": 1716454216713188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715052, "dur": 4,
    "args": {
      "External id": 8728,
      "cbid": 251, "correlation": 8728
    }
  },
  {
    "ph": "f", "id": 8728, "pid": 76337, "tid": -914061504, "ts": 1716454216715052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715058, "dur": 1,
    "args": {
      "External id": 8729,
      "cbid": 251, "correlation": 8729
    }
  },
  {
    "ph": "f", "id": 8729, "pid": 76337, "tid": -914061504, "ts": 1716454216715058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715060, "dur": 1,
    "args": {
      "External id": 8730,
      "cbid": 251, "correlation": 8730
    }
  },
  {
    "ph": "f", "id": 8730, "pid": 76337, "tid": -914061504, "ts": 1716454216715060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715062, "dur": 1,
    "args": {
      "External id": 8731,
      "cbid": 251, "correlation": 8731
    }
  },
  {
    "ph": "f", "id": 8731, "pid": 76337, "tid": -914061504, "ts": 1716454216715062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715064, "dur": 1,
    "args": {
      "External id": 8732,
      "cbid": 251, "correlation": 8732
    }
  },
  {
    "ph": "f", "id": 8732, "pid": 76337, "tid": -914061504, "ts": 1716454216715064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715066, "dur": 1,
    "args": {
      "External id": 8733,
      "cbid": 251, "correlation": 8733
    }
  },
  {
    "ph": "f", "id": 8733, "pid": 76337, "tid": -914061504, "ts": 1716454216715066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715068, "dur": 1,
    "args": {
      "External id": 8734,
      "cbid": 251, "correlation": 8734
    }
  },
  {
    "ph": "f", "id": 8734, "pid": 76337, "tid": -914061504, "ts": 1716454216715068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715070, "dur": 1,
    "args": {
      "External id": 8735,
      "cbid": 251, "correlation": 8735
    }
  },
  {
    "ph": "f", "id": 8735, "pid": 76337, "tid": -914061504, "ts": 1716454216715070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715072, "dur": 0,
    "args": {
      "External id": 8736,
      "cbid": 251, "correlation": 8736
    }
  },
  {
    "ph": "f", "id": 8736, "pid": 76337, "tid": -914061504, "ts": 1716454216715072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216715098, "dur": 56,
    "args": {
      "External id": 8737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8737,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8737, "pid": 5, "tid": 7, "ts": 1716454216715098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715077, "dur": 22,
    "args": {
      "External id": 8737,
      "cbid": 211, "correlation": 8737
    }
  },
  {
    "ph": "s", "id": 8737, "pid": 76337, "tid": -914061504, "ts": 1716454216715077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216715687, "dur": 34,
    "args": {
      "External id": 8743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8743, "pid": 5, "tid": 7, "ts": 1716454216715687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715672, "dur": 14,
    "args": {
      "External id": 8743,
      "cbid": 211, "correlation": 8743
    }
  },
  {
    "ph": "s", "id": 8743, "pid": 76337, "tid": -914061504, "ts": 1716454216715672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216715733, "dur": 27,
    "args": {
      "External id": 8751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8751,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8751, "pid": 5, "tid": 7, "ts": 1716454216715733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715722, "dur": 10,
    "args": {
      "External id": 8751,
      "cbid": 211, "correlation": 8751
    }
  },
  {
    "ph": "s", "id": 8751, "pid": 76337, "tid": -914061504, "ts": 1716454216715722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216715787, "dur": 20,
    "args": {
      "External id": 8759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8759,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8759, "pid": 5, "tid": 7, "ts": 1716454216715787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715774, "dur": 12,
    "args": {
      "External id": 8759,
      "cbid": 211, "correlation": 8759
    }
  },
  {
    "ph": "s", "id": 8759, "pid": 76337, "tid": -914061504, "ts": 1716454216715774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216715934, "dur": 34,
    "args": {
      "External id": 8779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8779,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 8779, "pid": 5, "tid": 7, "ts": 1716454216715934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715920, "dur": 14,
    "args": {
      "External id": 8779,
      "cbid": 211, "correlation": 8779
    }
  },
  {
    "ph": "s", "id": 8779, "pid": 76337, "tid": -914061504, "ts": 1716454216715920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216715969, "dur": 4,
    "args": {
      "External id": 8791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8791,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 8791, "pid": 5, "tid": 7, "ts": 1716454216715969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715945, "dur": 8,
    "args": {
      "External id": 8791,
      "cbid": 211, "correlation": 8791
    }
  },
  {
    "ph": "s", "id": 8791, "pid": 76337, "tid": -914061504, "ts": 1716454216715945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216715990, "dur": 32,
    "args": {
      "External id": 8794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8794,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8794, "pid": 5, "tid": 7, "ts": 1716454216715990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216715972, "dur": 17,
    "args": {
      "External id": 8794,
      "cbid": 211, "correlation": 8794
    }
  },
  {
    "ph": "s", "id": 8794, "pid": 76337, "tid": -914061504, "ts": 1716454216715972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216716059, "dur": 0,
    "args": {
      "External id": 8805,
      "cbid": 317, "correlation": 8805
    }
  },
  {
    "ph": "f", "id": 8805, "pid": 76337, "tid": -914061504, "ts": 1716454216716059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216716060, "dur": 1,
    "args": {
      "External id": 8806,
      "cbid": 203, "correlation": 8806
    }
  },
  {
    "ph": "f", "id": 8806, "pid": 76337, "tid": -914061504, "ts": 1716454216716060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216716062, "dur": 0,
    "args": {
      "External id": 8807,
      "cbid": 205, "correlation": 8807
    }
  },
  {
    "ph": "f", "id": 8807, "pid": 76337, "tid": -914061504, "ts": 1716454216716062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216717878, "dur": 24,
    "args": {
      "External id": 8811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8811,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8811, "pid": 5, "tid": 7, "ts": 1716454216717878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216717859, "dur": 19,
    "args": {
      "External id": 8811,
      "cbid": 211, "correlation": 8811
    }
  },
  {
    "ph": "s", "id": 8811, "pid": 76337, "tid": -914061504, "ts": 1716454216717859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216718071, "dur": 122,
    "args": {
      "External id": 8813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8813,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8813, "pid": 5, "tid": 7, "ts": 1716454216718071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216717914, "dur": 158,
    "args": {
      "External id": 8813,
      "cbid": 211, "correlation": 8813
    }
  },
  {
    "ph": "s", "id": 8813, "pid": 76337, "tid": -914061504, "ts": 1716454216717914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216718194, "dur": 23,
    "args": {
      "External id": 8815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8815,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8815, "pid": 5, "tid": 7, "ts": 1716454216718194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216718079, "dur": 8,
    "args": {
      "External id": 8815,
      "cbid": 211, "correlation": 8815
    }
  },
  {
    "ph": "s", "id": 8815, "pid": 76337, "tid": -914061504, "ts": 1716454216718079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216718608, "dur": 35,
    "args": {
      "External id": 8821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8821,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8821, "pid": 5, "tid": 7, "ts": 1716454216718608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216718596, "dur": 11,
    "args": {
      "External id": 8821,
      "cbid": 211, "correlation": 8821
    }
  },
  {
    "ph": "s", "id": 8821, "pid": 76337, "tid": -914061504, "ts": 1716454216718596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216718810, "dur": 89,
    "args": {
      "External id": 8832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8832, "pid": 5, "tid": 7, "ts": 1716454216718810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216718793, "dur": 18,
    "args": {
      "External id": 8832,
      "cbid": 211, "correlation": 8832
    }
  },
  {
    "ph": "s", "id": 8832, "pid": 76337, "tid": -914061504, "ts": 1716454216718793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216718901, "dur": 40,
    "args": {
      "External id": 8854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8854,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8854, "pid": 5, "tid": 7, "ts": 1716454216718901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216718837, "dur": 11,
    "args": {
      "External id": 8854,
      "cbid": 211, "correlation": 8854
    }
  },
  {
    "ph": "s", "id": 8854, "pid": 76337, "tid": -914061504, "ts": 1716454216718837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719006, "dur": 3,
    "args": {
      "External id": 8865,
      "cbid": 251, "correlation": 8865
    }
  },
  {
    "ph": "f", "id": 8865, "pid": 76337, "tid": -914061504, "ts": 1716454216719006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216719032, "dur": 105,
    "args": {
      "External id": 8866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8866,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8866, "pid": 5, "tid": 7, "ts": 1716454216719032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719016, "dur": 16,
    "args": {
      "External id": 8866,
      "cbid": 211, "correlation": 8866
    }
  },
  {
    "ph": "s", "id": 8866, "pid": 76337, "tid": -914061504, "ts": 1716454216719016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719099, "dur": 1,
    "args": {
      "External id": 8877,
      "cbid": 251, "correlation": 8877
    }
  },
  {
    "ph": "f", "id": 8877, "pid": 76337, "tid": -914061504, "ts": 1716454216719099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216719139, "dur": 98,
    "args": {
      "External id": 8878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8878,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8878, "pid": 5, "tid": 7, "ts": 1716454216719139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719103, "dur": 11,
    "args": {
      "External id": 8878,
      "cbid": 211, "correlation": 8878
    }
  },
  {
    "ph": "s", "id": 8878, "pid": 76337, "tid": -914061504, "ts": 1716454216719103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719187, "dur": 1,
    "args": {
      "External id": 8889,
      "cbid": 251, "correlation": 8889
    }
  },
  {
    "ph": "f", "id": 8889, "pid": 76337, "tid": -914061504, "ts": 1716454216719187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216719238, "dur": 96,
    "args": {
      "External id": 8890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8890,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8890, "pid": 5, "tid": 7, "ts": 1716454216719238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719192, "dur": 13,
    "args": {
      "External id": 8890,
      "cbid": 211, "correlation": 8890
    }
  },
  {
    "ph": "s", "id": 8890, "pid": 76337, "tid": -914061504, "ts": 1716454216719192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216719467, "dur": 588,
    "args": {
      "External id": 8911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8911,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 8911, "pid": 5, "tid": 7, "ts": 1716454216719467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719314, "dur": 155,
    "args": {
      "External id": 8911,
      "cbid": 211, "correlation": 8911
    }
  },
  {
    "ph": "s", "id": 8911, "pid": 76337, "tid": -914061504, "ts": 1716454216719314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719610, "dur": 3,
    "args": {
      "External id": 8929,
      "cbid": 251, "correlation": 8929
    }
  },
  {
    "ph": "f", "id": 8929, "pid": 76337, "tid": -914061504, "ts": 1716454216719610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216720056, "dur": 105,
    "args": {
      "External id": 8931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8931,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8931, "pid": 5, "tid": 7, "ts": 1716454216720056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719619, "dur": 14,
    "args": {
      "External id": 8931,
      "cbid": 211, "correlation": 8931
    }
  },
  {
    "ph": "s", "id": 8931, "pid": 76337, "tid": -914061504, "ts": 1716454216719619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216720163, "dur": 20,
    "args": {
      "External id": 8939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8939,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8939, "pid": 5, "tid": 7, "ts": 1716454216720163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719722, "dur": 16,
    "args": {
      "External id": 8939,
      "cbid": 211, "correlation": 8939
    }
  },
  {
    "ph": "s", "id": 8939, "pid": 76337, "tid": -914061504, "ts": 1716454216719722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216720184, "dur": 94,
    "args": {
      "External id": 8947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8947,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8947, "pid": 5, "tid": 7, "ts": 1716454216720184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719770, "dur": 10,
    "args": {
      "External id": 8947,
      "cbid": 211, "correlation": 8947
    }
  },
  {
    "ph": "s", "id": 8947, "pid": 76337, "tid": -914061504, "ts": 1716454216719770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216720279, "dur": 40,
    "args": {
      "External id": 8969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8969,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 8969, "pid": 5, "tid": 7, "ts": 1716454216720279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719842, "dur": 12,
    "args": {
      "External id": 8969,
      "cbid": 211, "correlation": 8969
    }
  },
  {
    "ph": "s", "id": 8969, "pid": 76337, "tid": -914061504, "ts": 1716454216719842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719947, "dur": 1,
    "args": {
      "External id": 8980,
      "cbid": 251, "correlation": 8980
    }
  },
  {
    "ph": "f", "id": 8980, "pid": 76337, "tid": -914061504, "ts": 1716454216719947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216720320, "dur": 89,
    "args": {
      "External id": 8981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8981,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 8981, "pid": 5, "tid": 7, "ts": 1716454216720320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216719952, "dur": 14,
    "args": {
      "External id": 8981,
      "cbid": 211, "correlation": 8981
    }
  },
  {
    "ph": "s", "id": 8981, "pid": 76337, "tid": -914061504, "ts": 1716454216719952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720055, "dur": 2,
    "args": {
      "External id": 8992,
      "cbid": 251, "correlation": 8992
    }
  },
  {
    "ph": "f", "id": 8992, "pid": 76337, "tid": -914061504, "ts": 1716454216720055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720061, "dur": 0,
    "args": {
      "External id": 8993,
      "cbid": 251, "correlation": 8993
    }
  },
  {
    "ph": "f", "id": 8993, "pid": 76337, "tid": -914061504, "ts": 1716454216720061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216720411, "dur": 12,
    "args": {
      "External id": 8994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8994,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 8994, "pid": 5, "tid": 7, "ts": 1716454216720411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720063, "dur": 15,
    "args": {
      "External id": 8994,
      "cbid": 211, "correlation": 8994
    }
  },
  {
    "ph": "s", "id": 8994, "pid": 76337, "tid": -914061504, "ts": 1716454216720063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216720430, "dur": 4,
    "args": {
      "External id": 8996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 8996,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 8996, "pid": 5, "tid": 7, "ts": 1716454216720430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720082, "dur": 72,
    "args": {
      "External id": 8996,
      "cbid": 211, "correlation": 8996
    }
  },
  {
    "ph": "s", "id": 8996, "pid": 76337, "tid": -914061504, "ts": 1716454216720082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720226, "dur": 1,
    "args": {
      "External id": 9007,
      "cbid": 251, "correlation": 9007
    }
  },
  {
    "ph": "f", "id": 9007, "pid": 76337, "tid": -914061504, "ts": 1716454216720226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720230, "dur": 0,
    "args": {
      "External id": 9008,
      "cbid": 251, "correlation": 9008
    }
  },
  {
    "ph": "f", "id": 9008, "pid": 76337, "tid": -914061504, "ts": 1716454216720230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216720436, "dur": 9,
    "args": {
      "External id": 9009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9009,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 9009, "pid": 5, "tid": 7, "ts": 1716454216720436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720232, "dur": 13,
    "args": {
      "External id": 9009,
      "cbid": 211, "correlation": 9009
    }
  },
  {
    "ph": "s", "id": 9009, "pid": 76337, "tid": -914061504, "ts": 1716454216720232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216720446, "dur": 3,
    "args": {
      "External id": 9011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9011,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 9011, "pid": 5, "tid": 7, "ts": 1716454216720446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720247, "dur": 6,
    "args": {
      "External id": 9011,
      "cbid": 211, "correlation": 9011
    }
  },
  {
    "ph": "s", "id": 9011, "pid": 76337, "tid": -914061504, "ts": 1716454216720247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216720450, "dur": 110,
    "args": {
      "External id": 9032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9032,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 9032, "pid": 5, "tid": 7, "ts": 1716454216720450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720324, "dur": 12,
    "args": {
      "External id": 9032,
      "cbid": 211, "correlation": 9032
    }
  },
  {
    "ph": "s", "id": 9032, "pid": 76337, "tid": -914061504, "ts": 1716454216720324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720437, "dur": 1,
    "args": {
      "External id": 9050,
      "cbid": 251, "correlation": 9050
    }
  },
  {
    "ph": "f", "id": 9050, "pid": 76337, "tid": -914061504, "ts": 1716454216720437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216720562, "dur": 106,
    "args": {
      "External id": 9052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9052,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9052, "pid": 5, "tid": 7, "ts": 1716454216720562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720443, "dur": 14,
    "args": {
      "External id": 9052,
      "cbid": 211, "correlation": 9052
    }
  },
  {
    "ph": "s", "id": 9052, "pid": 76337, "tid": -914061504, "ts": 1716454216720443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216720670, "dur": 20,
    "args": {
      "External id": 9060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9060,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9060, "pid": 5, "tid": 7, "ts": 1716454216720670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720518, "dur": 12,
    "args": {
      "External id": 9060,
      "cbid": 211, "correlation": 9060
    }
  },
  {
    "ph": "s", "id": 9060, "pid": 76337, "tid": -914061504, "ts": 1716454216720518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216720691, "dur": 28,
    "args": {
      "External id": 9068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9068,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9068, "pid": 5, "tid": 7, "ts": 1716454216720691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720559, "dur": 9,
    "args": {
      "External id": 9068,
      "cbid": 211, "correlation": 9068
    }
  },
  {
    "ph": "s", "id": 9068, "pid": 76337, "tid": -914061504, "ts": 1716454216720559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216720721, "dur": 40,
    "args": {
      "External id": 9090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9090,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9090, "pid": 5, "tid": 7, "ts": 1716454216720721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720611, "dur": 10,
    "args": {
      "External id": 9090,
      "cbid": 211, "correlation": 9090
    }
  },
  {
    "ph": "s", "id": 9090, "pid": 76337, "tid": -914061504, "ts": 1716454216720611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720709, "dur": 2,
    "args": {
      "External id": 9106,
      "cbid": 251, "correlation": 9106
    }
  },
  {
    "ph": "f", "id": 9106, "pid": 76337, "tid": -914061504, "ts": 1716454216720709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720715, "dur": 0,
    "args": {
      "External id": 9108,
      "cbid": 251, "correlation": 9108
    }
  },
  {
    "ph": "f", "id": 9108, "pid": 76337, "tid": -914061504, "ts": 1716454216720715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216720763, "dur": 590,
    "args": {
      "External id": 9109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9109,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 9109, "pid": 5, "tid": 7, "ts": 1716454216720763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720721, "dur": 13,
    "args": {
      "External id": 9109,
      "cbid": 211, "correlation": 9109
    }
  },
  {
    "ph": "s", "id": 9109, "pid": 76337, "tid": -914061504, "ts": 1716454216720721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216721354, "dur": 134,
    "args": {
      "External id": 9117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9117,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9117, "pid": 5, "tid": 7, "ts": 1716454216721354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720819, "dur": 15,
    "args": {
      "External id": 9117,
      "cbid": 211, "correlation": 9117
    }
  },
  {
    "ph": "s", "id": 9117, "pid": 76337, "tid": -914061504, "ts": 1716454216720819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216721489, "dur": 131,
    "args": {
      "External id": 9125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9125, "pid": 5, "tid": 7, "ts": 1716454216721489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720862, "dur": 10,
    "args": {
      "External id": 9125,
      "cbid": 211, "correlation": 9125
    }
  },
  {
    "ph": "s", "id": 9125, "pid": 76337, "tid": -914061504, "ts": 1716454216720862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720945, "dur": 1,
    "args": {
      "External id": 9141,
      "cbid": 251, "correlation": 9141
    }
  },
  {
    "ph": "f", "id": 9141, "pid": 76337, "tid": -914061504, "ts": 1716454216720945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216721622, "dur": 327,
    "args": {
      "External id": 9143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9143,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9143, "pid": 5, "tid": 7, "ts": 1716454216721622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216720951, "dur": 13,
    "args": {
      "External id": 9143,
      "cbid": 211, "correlation": 9143
    }
  },
  {
    "ph": "s", "id": 9143, "pid": 76337, "tid": -914061504, "ts": 1716454216720951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216724405, "dur": 27,
    "args": {
      "External id": 9151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9151,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9151, "pid": 5, "tid": 7, "ts": 1716454216724405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724380, "dur": 25,
    "args": {
      "External id": 9151,
      "cbid": 211, "correlation": 9151
    }
  },
  {
    "ph": "s", "id": 9151, "pid": 76337, "tid": -914061504, "ts": 1716454216724380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216724527, "dur": 90,
    "args": {
      "External id": 9162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9162, "pid": 5, "tid": 7, "ts": 1716454216724527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724510, "dur": 16,
    "args": {
      "External id": 9162,
      "cbid": 211, "correlation": 9162
    }
  },
  {
    "ph": "s", "id": 9162, "pid": 76337, "tid": -914061504, "ts": 1716454216724510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724639, "dur": 1,
    "args": {
      "External id": 9174,
      "cbid": 317, "correlation": 9174
    }
  },
  {
    "ph": "f", "id": 9174, "pid": 76337, "tid": -914061504, "ts": 1716454216724639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724641, "dur": 1,
    "args": {
      "External id": 9175,
      "cbid": 203, "correlation": 9175
    }
  },
  {
    "ph": "f", "id": 9175, "pid": 76337, "tid": -914061504, "ts": 1716454216724641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724643, "dur": 1,
    "args": {
      "External id": 9176,
      "cbid": 205, "correlation": 9176
    }
  },
  {
    "ph": "f", "id": 9176, "pid": 76337, "tid": -914061504, "ts": 1716454216724643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216724700, "dur": 24,
    "args": {
      "External id": 9180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9180,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9180, "pid": 5, "tid": 7, "ts": 1716454216724700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724684, "dur": 16,
    "args": {
      "External id": 9180,
      "cbid": 211, "correlation": 9180
    }
  },
  {
    "ph": "s", "id": 9180, "pid": 76337, "tid": -914061504, "ts": 1716454216724684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216724732, "dur": 131,
    "args": {
      "External id": 9182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9182,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9182, "pid": 5, "tid": 7, "ts": 1716454216724732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724719, "dur": 12,
    "args": {
      "External id": 9182,
      "cbid": 211, "correlation": 9182
    }
  },
  {
    "ph": "s", "id": 9182, "pid": 76337, "tid": -914061504, "ts": 1716454216724719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216724864, "dur": 24,
    "args": {
      "External id": 9184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9184,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9184, "pid": 5, "tid": 7, "ts": 1716454216724864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724739, "dur": 21,
    "args": {
      "External id": 9184,
      "cbid": 211, "correlation": 9184
    }
  },
  {
    "ph": "s", "id": 9184, "pid": 76337, "tid": -914061504, "ts": 1716454216724739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216724890, "dur": 35,
    "args": {
      "External id": 9190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9190,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9190, "pid": 5, "tid": 7, "ts": 1716454216724890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724790, "dur": 9,
    "args": {
      "External id": 9190,
      "cbid": 211, "correlation": 9190
    }
  },
  {
    "ph": "s", "id": 9190, "pid": 76337, "tid": -914061504, "ts": 1716454216724790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216724926, "dur": 27,
    "args": {
      "External id": 9198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9198,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9198, "pid": 5, "tid": 7, "ts": 1716454216724926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724824, "dur": 8,
    "args": {
      "External id": 9198,
      "cbid": 211, "correlation": 9198
    }
  },
  {
    "ph": "s", "id": 9198, "pid": 76337, "tid": -914061504, "ts": 1716454216724824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216724986, "dur": 32,
    "args": {
      "External id": 9218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9218,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 9218, "pid": 5, "tid": 7, "ts": 1716454216724986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216724962, "dur": 24,
    "args": {
      "External id": 9218,
      "cbid": 211, "correlation": 9218
    }
  },
  {
    "ph": "s", "id": 9218, "pid": 76337, "tid": -914061504, "ts": 1716454216724962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216725020, "dur": 5,
    "args": {
      "External id": 9230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9230,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 9230, "pid": 5, "tid": 7, "ts": 1716454216725020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725005, "dur": 9,
    "args": {
      "External id": 9230,
      "cbid": 211, "correlation": 9230
    }
  },
  {
    "ph": "s", "id": 9230, "pid": 76337, "tid": -914061504, "ts": 1716454216725005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216725040, "dur": 33,
    "args": {
      "External id": 9233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9233,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9233, "pid": 5, "tid": 7, "ts": 1716454216725040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725030, "dur": 8,
    "args": {
      "External id": 9233,
      "cbid": 211, "correlation": 9233
    }
  },
  {
    "ph": "s", "id": 9233, "pid": 76337, "tid": -914061504, "ts": 1716454216725030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216725109, "dur": 23,
    "args": {
      "External id": 9242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9242,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9242, "pid": 5, "tid": 7, "ts": 1716454216725109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725089, "dur": 29,
    "args": {
      "External id": 9242,
      "cbid": 211, "correlation": 9242
    }
  },
  {
    "ph": "s", "id": 9242, "pid": 76337, "tid": -914061504, "ts": 1716454216725089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725172, "dur": 0,
    "args": {
      "External id": 9252,
      "cbid": 317, "correlation": 9252
    }
  },
  {
    "ph": "f", "id": 9252, "pid": 76337, "tid": -914061504, "ts": 1716454216725172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725173, "dur": 0,
    "args": {
      "External id": 9253,
      "cbid": 203, "correlation": 9253
    }
  },
  {
    "ph": "f", "id": 9253, "pid": 76337, "tid": -914061504, "ts": 1716454216725173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725173, "dur": 0,
    "args": {
      "External id": 9254,
      "cbid": 205, "correlation": 9254
    }
  },
  {
    "ph": "f", "id": 9254, "pid": 76337, "tid": -914061504, "ts": 1716454216725173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216725210, "dur": 22,
    "args": {
      "External id": 9258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9258,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9258, "pid": 5, "tid": 7, "ts": 1716454216725210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725196, "dur": 13,
    "args": {
      "External id": 9258,
      "cbid": 211, "correlation": 9258
    }
  },
  {
    "ph": "s", "id": 9258, "pid": 76337, "tid": -914061504, "ts": 1716454216725196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216725233, "dur": 48,
    "args": {
      "External id": 9260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9260,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9260, "pid": 5, "tid": 7, "ts": 1716454216725233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725212, "dur": 6,
    "args": {
      "External id": 9260,
      "cbid": 211, "correlation": 9260
    }
  },
  {
    "ph": "s", "id": 9260, "pid": 76337, "tid": -914061504, "ts": 1716454216725212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216725283, "dur": 719,
    "args": {
      "External id": 9262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9262,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9262, "pid": 5, "tid": 7, "ts": 1716454216725283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725226, "dur": 10,
    "args": {
      "External id": 9262,
      "cbid": 211, "correlation": 9262
    }
  },
  {
    "ph": "s", "id": 9262, "pid": 76337, "tid": -914061504, "ts": 1716454216725226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216726003, "dur": 22,
    "args": {
      "External id": 9264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9264,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9264, "pid": 5, "tid": 7, "ts": 1716454216726003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725240, "dur": 5,
    "args": {
      "External id": 9264,
      "cbid": 211, "correlation": 9264
    }
  },
  {
    "ph": "s", "id": 9264, "pid": 76337, "tid": -914061504, "ts": 1716454216725240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216726027, "dur": 35,
    "args": {
      "External id": 9270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9270, "pid": 5, "tid": 7, "ts": 1716454216726027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725277, "dur": 10,
    "args": {
      "External id": 9270,
      "cbid": 211, "correlation": 9270
    }
  },
  {
    "ph": "s", "id": 9270, "pid": 76337, "tid": -914061504, "ts": 1716454216725277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216726063, "dur": 4,
    "args": {
      "External id": 9278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9278,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 9278, "pid": 5, "tid": 7, "ts": 1716454216726063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725326, "dur": 10,
    "args": {
      "External id": 9278,
      "cbid": 211, "correlation": 9278
    }
  },
  {
    "ph": "s", "id": 9278, "pid": 76337, "tid": -914061504, "ts": 1716454216725326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725441, "dur": 3,
    "args": {
      "External id": 9294,
      "cbid": 251, "correlation": 9294
    }
  },
  {
    "ph": "f", "id": 9294, "pid": 76337, "tid": -914061504, "ts": 1716454216725441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725448, "dur": 0,
    "args": {
      "External id": 9296,
      "cbid": 251, "correlation": 9296
    }
  },
  {
    "ph": "f", "id": 9296, "pid": 76337, "tid": -914061504, "ts": 1716454216725448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216726068, "dur": 14,
    "args": {
      "External id": 9297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9297,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 9297, "pid": 5, "tid": 7, "ts": 1716454216726068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725451, "dur": 17,
    "args": {
      "External id": 9297,
      "cbid": 211, "correlation": 9297
    }
  },
  {
    "ph": "s", "id": 9297, "pid": 76337, "tid": -914061504, "ts": 1716454216725451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216726083, "dur": 5,
    "args": {
      "External id": 9299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9299,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 9299, "pid": 5, "tid": 7, "ts": 1716454216726083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725473, "dur": 8,
    "args": {
      "External id": 9299,
      "cbid": 211, "correlation": 9299
    }
  },
  {
    "ph": "s", "id": 9299, "pid": 76337, "tid": -914061504, "ts": 1716454216725473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216726090, "dur": 31,
    "args": {
      "External id": 9309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9309, "pid": 5, "tid": 7, "ts": 1716454216726090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725552, "dur": 12,
    "args": {
      "External id": 9309,
      "cbid": 211, "correlation": 9309
    }
  },
  {
    "ph": "s", "id": 9309, "pid": 76337, "tid": -914061504, "ts": 1716454216725552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216726123, "dur": 32,
    "args": {
      "External id": 9329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9329,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 9329, "pid": 5, "tid": 7, "ts": 1716454216726123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725634, "dur": 12,
    "args": {
      "External id": 9329,
      "cbid": 211, "correlation": 9329
    }
  },
  {
    "ph": "s", "id": 9329, "pid": 76337, "tid": -914061504, "ts": 1716454216725634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216726157, "dur": 4,
    "args": {
      "External id": 9341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9341,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 9341, "pid": 5, "tid": 7, "ts": 1716454216726157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725662, "dur": 8,
    "args": {
      "External id": 9341,
      "cbid": 211, "correlation": 9341
    }
  },
  {
    "ph": "s", "id": 9341, "pid": 76337, "tid": -914061504, "ts": 1716454216725662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216726162, "dur": 32,
    "args": {
      "External id": 9344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9344,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9344, "pid": 5, "tid": 7, "ts": 1716454216726162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725684, "dur": 7,
    "args": {
      "External id": 9344,
      "cbid": 211, "correlation": 9344
    }
  },
  {
    "ph": "s", "id": 9344, "pid": 76337, "tid": -914061504, "ts": 1716454216725684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216726195, "dur": 23,
    "args": {
      "External id": 9353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9353,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9353, "pid": 5, "tid": 7, "ts": 1716454216726195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725726, "dur": 10,
    "args": {
      "External id": 9353,
      "cbid": 211, "correlation": 9353
    }
  },
  {
    "ph": "s", "id": 9353, "pid": 76337, "tid": -914061504, "ts": 1716454216725726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725806, "dur": 0,
    "args": {
      "External id": 9363,
      "cbid": 317, "correlation": 9363
    }
  },
  {
    "ph": "f", "id": 9363, "pid": 76337, "tid": -914061504, "ts": 1716454216725806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725807, "dur": 0,
    "args": {
      "External id": 9364,
      "cbid": 203, "correlation": 9364
    }
  },
  {
    "ph": "f", "id": 9364, "pid": 76337, "tid": -914061504, "ts": 1716454216725807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725808, "dur": 0,
    "args": {
      "External id": 9365,
      "cbid": 205, "correlation": 9365
    }
  },
  {
    "ph": "f", "id": 9365, "pid": 76337, "tid": -914061504, "ts": 1716454216725808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216726219, "dur": 24,
    "args": {
      "External id": 9369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9369,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9369, "pid": 5, "tid": 7, "ts": 1716454216726219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725821, "dur": 13,
    "args": {
      "External id": 9369,
      "cbid": 211, "correlation": 9369
    }
  },
  {
    "ph": "s", "id": 9369, "pid": 76337, "tid": -914061504, "ts": 1716454216725821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216726245, "dur": 48,
    "args": {
      "External id": 9371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9371, "pid": 5, "tid": 7, "ts": 1716454216726245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725836, "dur": 6,
    "args": {
      "External id": 9371,
      "cbid": 211, "correlation": 9371
    }
  },
  {
    "ph": "s", "id": 9371, "pid": 76337, "tid": -914061504, "ts": 1716454216725836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216726295, "dur": 709,
    "args": {
      "External id": 9373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9373,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9373, "pid": 5, "tid": 7, "ts": 1716454216726295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725848, "dur": 6,
    "args": {
      "External id": 9373,
      "cbid": 211, "correlation": 9373
    }
  },
  {
    "ph": "s", "id": 9373, "pid": 76337, "tid": -914061504, "ts": 1716454216725848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216727006, "dur": 22,
    "args": {
      "External id": 9375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9375,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9375, "pid": 5, "tid": 7, "ts": 1716454216727006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725858, "dur": 5,
    "args": {
      "External id": 9375,
      "cbid": 211, "correlation": 9375
    }
  },
  {
    "ph": "s", "id": 9375, "pid": 76337, "tid": -914061504, "ts": 1716454216725858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216727029, "dur": 35,
    "args": {
      "External id": 9381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9381, "pid": 5, "tid": 7, "ts": 1716454216727029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725886, "dur": 9,
    "args": {
      "External id": 9381,
      "cbid": 211, "correlation": 9381
    }
  },
  {
    "ph": "s", "id": 9381, "pid": 76337, "tid": -914061504, "ts": 1716454216725886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216727065, "dur": 27,
    "args": {
      "External id": 9389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9389,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9389, "pid": 5, "tid": 7, "ts": 1716454216727065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725920, "dur": 8,
    "args": {
      "External id": 9389,
      "cbid": 211, "correlation": 9389
    }
  },
  {
    "ph": "s", "id": 9389, "pid": 76337, "tid": -914061504, "ts": 1716454216725920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216727094, "dur": 20,
    "args": {
      "External id": 9397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9397,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9397, "pid": 5, "tid": 7, "ts": 1716454216727094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216725962, "dur": 10,
    "args": {
      "External id": 9397,
      "cbid": 211, "correlation": 9397
    }
  },
  {
    "ph": "s", "id": 9397, "pid": 76337, "tid": -914061504, "ts": 1716454216725962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216727115, "dur": 32,
    "args": {
      "External id": 9417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9417,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 9417, "pid": 5, "tid": 7, "ts": 1716454216727115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726057, "dur": 13,
    "args": {
      "External id": 9417,
      "cbid": 211, "correlation": 9417
    }
  },
  {
    "ph": "s", "id": 9417, "pid": 76337, "tid": -914061504, "ts": 1716454216726057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216727149, "dur": 5,
    "args": {
      "External id": 9429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9429,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 9429, "pid": 5, "tid": 7, "ts": 1716454216727149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726091, "dur": 8,
    "args": {
      "External id": 9429,
      "cbid": 211, "correlation": 9429
    }
  },
  {
    "ph": "s", "id": 9429, "pid": 76337, "tid": -914061504, "ts": 1716454216726091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216727155, "dur": 32,
    "args": {
      "External id": 9432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9432, "pid": 5, "tid": 7, "ts": 1716454216727155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726112, "dur": 7,
    "args": {
      "External id": 9432,
      "cbid": 211, "correlation": 9432
    }
  },
  {
    "ph": "s", "id": 9432, "pid": 76337, "tid": -914061504, "ts": 1716454216726112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726180, "dur": 0,
    "args": {
      "External id": 9443,
      "cbid": 317, "correlation": 9443
    }
  },
  {
    "ph": "f", "id": 9443, "pid": 76337, "tid": -914061504, "ts": 1716454216726180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726181, "dur": 0,
    "args": {
      "External id": 9444,
      "cbid": 203, "correlation": 9444
    }
  },
  {
    "ph": "f", "id": 9444, "pid": 76337, "tid": -914061504, "ts": 1716454216726181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726182, "dur": 0,
    "args": {
      "External id": 9445,
      "cbid": 205, "correlation": 9445
    }
  },
  {
    "ph": "f", "id": 9445, "pid": 76337, "tid": -914061504, "ts": 1716454216726182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216727189, "dur": 23,
    "args": {
      "External id": 9449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9449, "pid": 5, "tid": 7, "ts": 1716454216727189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726195, "dur": 13,
    "args": {
      "External id": 9449,
      "cbid": 211, "correlation": 9449
    }
  },
  {
    "ph": "s", "id": 9449, "pid": 76337, "tid": -914061504, "ts": 1716454216726195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216727213, "dur": 129,
    "args": {
      "External id": 9451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9451,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9451, "pid": 5, "tid": 7, "ts": 1716454216727213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726215, "dur": 6,
    "args": {
      "External id": 9451,
      "cbid": 211, "correlation": 9451
    }
  },
  {
    "ph": "s", "id": 9451, "pid": 76337, "tid": -914061504, "ts": 1716454216726215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216727343, "dur": 22,
    "args": {
      "External id": 9453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9453,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9453, "pid": 5, "tid": 7, "ts": 1716454216727343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726225, "dur": 5,
    "args": {
      "External id": 9453,
      "cbid": 211, "correlation": 9453
    }
  },
  {
    "ph": "s", "id": 9453, "pid": 76337, "tid": -914061504, "ts": 1716454216726225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216727367, "dur": 35,
    "args": {
      "External id": 9459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9459,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9459, "pid": 5, "tid": 7, "ts": 1716454216727367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726253, "dur": 9,
    "args": {
      "External id": 9459,
      "cbid": 211, "correlation": 9459
    }
  },
  {
    "ph": "s", "id": 9459, "pid": 76337, "tid": -914061504, "ts": 1716454216726253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216727404, "dur": 88,
    "args": {
      "External id": 9470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9470,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9470, "pid": 5, "tid": 7, "ts": 1716454216727404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726365, "dur": 14,
    "args": {
      "External id": 9470,
      "cbid": 211, "correlation": 9470
    }
  },
  {
    "ph": "s", "id": 9470, "pid": 76337, "tid": -914061504, "ts": 1716454216726365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216727493, "dur": 39,
    "args": {
      "External id": 9492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9492,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9492, "pid": 5, "tid": 7, "ts": 1716454216727493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726399, "dur": 10,
    "args": {
      "External id": 9492,
      "cbid": 211, "correlation": 9492
    }
  },
  {
    "ph": "s", "id": 9492, "pid": 76337, "tid": -914061504, "ts": 1716454216726399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726512, "dur": 2,
    "args": {
      "External id": 9503,
      "cbid": 251, "correlation": 9503
    }
  },
  {
    "ph": "f", "id": 9503, "pid": 76337, "tid": -914061504, "ts": 1716454216726512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216727534, "dur": 105,
    "args": {
      "External id": 9504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9504,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9504, "pid": 5, "tid": 7, "ts": 1716454216727534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726519, "dur": 19,
    "args": {
      "External id": 9504,
      "cbid": 211, "correlation": 9504
    }
  },
  {
    "ph": "s", "id": 9504, "pid": 76337, "tid": -914061504, "ts": 1716454216726519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726613, "dur": 1,
    "args": {
      "External id": 9515,
      "cbid": 251, "correlation": 9515
    }
  },
  {
    "ph": "f", "id": 9515, "pid": 76337, "tid": -914061504, "ts": 1716454216726613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216727641, "dur": 98,
    "args": {
      "External id": 9516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9516,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9516, "pid": 5, "tid": 7, "ts": 1716454216727641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726618, "dur": 12,
    "args": {
      "External id": 9516,
      "cbid": 211, "correlation": 9516
    }
  },
  {
    "ph": "s", "id": 9516, "pid": 76337, "tid": -914061504, "ts": 1716454216726618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726698, "dur": 1,
    "args": {
      "External id": 9527,
      "cbid": 251, "correlation": 9527
    }
  },
  {
    "ph": "f", "id": 9527, "pid": 76337, "tid": -914061504, "ts": 1716454216726698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216727740, "dur": 97,
    "args": {
      "External id": 9528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9528,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9528, "pid": 5, "tid": 7, "ts": 1716454216727740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726702, "dur": 13,
    "args": {
      "External id": 9528,
      "cbid": 211, "correlation": 9528
    }
  },
  {
    "ph": "s", "id": 9528, "pid": 76337, "tid": -914061504, "ts": 1716454216726702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216727838, "dur": 610,
    "args": {
      "External id": 9549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9549,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 9549, "pid": 5, "tid": 7, "ts": 1716454216727838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726805, "dur": 17,
    "args": {
      "External id": 9549,
      "cbid": 211, "correlation": 9549
    }
  },
  {
    "ph": "s", "id": 9549, "pid": 76337, "tid": -914061504, "ts": 1716454216726805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726922, "dur": 14,
    "args": {
      "External id": 9567,
      "cbid": 251, "correlation": 9567
    }
  },
  {
    "ph": "f", "id": 9567, "pid": 76337, "tid": -914061504, "ts": 1716454216726922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216728450, "dur": 105,
    "args": {
      "External id": 9569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9569,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9569, "pid": 5, "tid": 7, "ts": 1716454216728450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216726942, "dur": 14,
    "args": {
      "External id": 9569,
      "cbid": 211, "correlation": 9569
    }
  },
  {
    "ph": "s", "id": 9569, "pid": 76337, "tid": -914061504, "ts": 1716454216726942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216728556, "dur": 19,
    "args": {
      "External id": 9577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9577,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9577, "pid": 5, "tid": 7, "ts": 1716454216728556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727026, "dur": 13,
    "args": {
      "External id": 9577,
      "cbid": 211, "correlation": 9577
    }
  },
  {
    "ph": "s", "id": 9577, "pid": 76337, "tid": -914061504, "ts": 1716454216727026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216728577, "dur": 93,
    "args": {
      "External id": 9585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9585,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9585, "pid": 5, "tid": 7, "ts": 1716454216728577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727069, "dur": 9,
    "args": {
      "External id": 9585,
      "cbid": 211, "correlation": 9585
    }
  },
  {
    "ph": "s", "id": 9585, "pid": 76337, "tid": -914061504, "ts": 1716454216727069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216728671, "dur": 37,
    "args": {
      "External id": 9607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9607,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9607, "pid": 5, "tid": 7, "ts": 1716454216728671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727123, "dur": 10,
    "args": {
      "External id": 9607,
      "cbid": 211, "correlation": 9607
    }
  },
  {
    "ph": "s", "id": 9607, "pid": 76337, "tid": -914061504, "ts": 1716454216727123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727208, "dur": 1,
    "args": {
      "External id": 9618,
      "cbid": 251, "correlation": 9618
    }
  },
  {
    "ph": "f", "id": 9618, "pid": 76337, "tid": -914061504, "ts": 1716454216727208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216728710, "dur": 95,
    "args": {
      "External id": 9619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9619,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9619, "pid": 5, "tid": 7, "ts": 1716454216728710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727213, "dur": 15,
    "args": {
      "External id": 9619,
      "cbid": 211, "correlation": 9619
    }
  },
  {
    "ph": "s", "id": 9619, "pid": 76337, "tid": -914061504, "ts": 1716454216727213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727293, "dur": 1,
    "args": {
      "External id": 9630,
      "cbid": 251, "correlation": 9630
    }
  },
  {
    "ph": "f", "id": 9630, "pid": 76337, "tid": -914061504, "ts": 1716454216727293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727297, "dur": 0,
    "args": {
      "External id": 9631,
      "cbid": 251, "correlation": 9631
    }
  },
  {
    "ph": "f", "id": 9631, "pid": 76337, "tid": -914061504, "ts": 1716454216727297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216728807, "dur": 12,
    "args": {
      "External id": 9632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9632,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 9632, "pid": 5, "tid": 7, "ts": 1716454216728807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727299, "dur": 12,
    "args": {
      "External id": 9632,
      "cbid": 211, "correlation": 9632
    }
  },
  {
    "ph": "s", "id": 9632, "pid": 76337, "tid": -914061504, "ts": 1716454216727299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216728820, "dur": 5,
    "args": {
      "External id": 9634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9634,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 9634, "pid": 5, "tid": 7, "ts": 1716454216728820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727314, "dur": 7,
    "args": {
      "External id": 9634,
      "cbid": 211, "correlation": 9634
    }
  },
  {
    "ph": "s", "id": 9634, "pid": 76337, "tid": -914061504, "ts": 1716454216727314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727388, "dur": 1,
    "args": {
      "External id": 9645,
      "cbid": 251, "correlation": 9645
    }
  },
  {
    "ph": "f", "id": 9645, "pid": 76337, "tid": -914061504, "ts": 1716454216727388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727392, "dur": 0,
    "args": {
      "External id": 9646,
      "cbid": 251, "correlation": 9646
    }
  },
  {
    "ph": "f", "id": 9646, "pid": 76337, "tid": -914061504, "ts": 1716454216727392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216728826, "dur": 8,
    "args": {
      "External id": 9647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9647,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 9647, "pid": 5, "tid": 7, "ts": 1716454216728826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727394, "dur": 12,
    "args": {
      "External id": 9647,
      "cbid": 211, "correlation": 9647
    }
  },
  {
    "ph": "s", "id": 9647, "pid": 76337, "tid": -914061504, "ts": 1716454216727394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216728835, "dur": 3,
    "args": {
      "External id": 9649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9649,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 9649, "pid": 5, "tid": 7, "ts": 1716454216728835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727408, "dur": 6,
    "args": {
      "External id": 9649,
      "cbid": 211, "correlation": 9649
    }
  },
  {
    "ph": "s", "id": 9649, "pid": 76337, "tid": -914061504, "ts": 1716454216727408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216728840, "dur": 99,
    "args": {
      "External id": 9670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9670,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 9670, "pid": 5, "tid": 7, "ts": 1716454216728840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727483, "dur": 13,
    "args": {
      "External id": 9670,
      "cbid": 211, "correlation": 9670
    }
  },
  {
    "ph": "s", "id": 9670, "pid": 76337, "tid": -914061504, "ts": 1716454216727483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727599, "dur": 1,
    "args": {
      "External id": 9688,
      "cbid": 251, "correlation": 9688
    }
  },
  {
    "ph": "f", "id": 9688, "pid": 76337, "tid": -914061504, "ts": 1716454216727599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216728941, "dur": 100,
    "args": {
      "External id": 9690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9690,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9690, "pid": 5, "tid": 7, "ts": 1716454216728941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727605, "dur": 14,
    "args": {
      "External id": 9690,
      "cbid": 211, "correlation": 9690
    }
  },
  {
    "ph": "s", "id": 9690, "pid": 76337, "tid": -914061504, "ts": 1716454216727605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216729042, "dur": 19,
    "args": {
      "External id": 9698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9698,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9698, "pid": 5, "tid": 7, "ts": 1716454216729042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727677, "dur": 13,
    "args": {
      "External id": 9698,
      "cbid": 211, "correlation": 9698
    }
  },
  {
    "ph": "s", "id": 9698, "pid": 76337, "tid": -914061504, "ts": 1716454216727677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216729063, "dur": 28,
    "args": {
      "External id": 9706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9706,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9706, "pid": 5, "tid": 7, "ts": 1716454216729063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727717, "dur": 8,
    "args": {
      "External id": 9706,
      "cbid": 211, "correlation": 9706
    }
  },
  {
    "ph": "s", "id": 9706, "pid": 76337, "tid": -914061504, "ts": 1716454216727717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216729092, "dur": 38,
    "args": {
      "External id": 9728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9728,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9728, "pid": 5, "tid": 7, "ts": 1716454216729092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727767, "dur": 10,
    "args": {
      "External id": 9728,
      "cbid": 211, "correlation": 9728
    }
  },
  {
    "ph": "s", "id": 9728, "pid": 76337, "tid": -914061504, "ts": 1716454216727767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727860, "dur": 2,
    "args": {
      "External id": 9744,
      "cbid": 251, "correlation": 9744
    }
  },
  {
    "ph": "f", "id": 9744, "pid": 76337, "tid": -914061504, "ts": 1716454216727860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727865, "dur": 0,
    "args": {
      "External id": 9746,
      "cbid": 251, "correlation": 9746
    }
  },
  {
    "ph": "f", "id": 9746, "pid": 76337, "tid": -914061504, "ts": 1716454216727865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216729132, "dur": 589,
    "args": {
      "External id": 9747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9747,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 9747, "pid": 5, "tid": 7, "ts": 1716454216729132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727871, "dur": 14,
    "args": {
      "External id": 9747,
      "cbid": 211, "correlation": 9747
    }
  },
  {
    "ph": "s", "id": 9747, "pid": 76337, "tid": -914061504, "ts": 1716454216727871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216729722, "dur": 133,
    "args": {
      "External id": 9755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9755,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9755, "pid": 5, "tid": 7, "ts": 1716454216729722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216727967, "dur": 28,
    "args": {
      "External id": 9755,
      "cbid": 211, "correlation": 9755
    }
  },
  {
    "ph": "s", "id": 9755, "pid": 76337, "tid": -914061504, "ts": 1716454216727967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216729856, "dur": 133,
    "args": {
      "External id": 9763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9763, "pid": 5, "tid": 7, "ts": 1716454216729856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728021, "dur": 13,
    "args": {
      "External id": 9763,
      "cbid": 211, "correlation": 9763
    }
  },
  {
    "ph": "s", "id": 9763, "pid": 76337, "tid": -914061504, "ts": 1716454216728021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728107, "dur": 2,
    "args": {
      "External id": 9779,
      "cbid": 251, "correlation": 9779
    }
  },
  {
    "ph": "f", "id": 9779, "pid": 76337, "tid": -914061504, "ts": 1716454216728107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216729991, "dur": 318,
    "args": {
      "External id": 9781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9781,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9781, "pid": 5, "tid": 7, "ts": 1716454216729991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728113, "dur": 12,
    "args": {
      "External id": 9781,
      "cbid": 211, "correlation": 9781
    }
  },
  {
    "ph": "s", "id": 9781, "pid": 76337, "tid": -914061504, "ts": 1716454216728113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216730310, "dur": 27,
    "args": {
      "External id": 9789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9789, "pid": 5, "tid": 7, "ts": 1716454216730310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728156, "dur": 10,
    "args": {
      "External id": 9789,
      "cbid": 211, "correlation": 9789
    }
  },
  {
    "ph": "s", "id": 9789, "pid": 76337, "tid": -914061504, "ts": 1716454216728156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216730337, "dur": 82,
    "args": {
      "External id": 9800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9800, "pid": 5, "tid": 7, "ts": 1716454216730337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728219, "dur": 13,
    "args": {
      "External id": 9800,
      "cbid": 211, "correlation": 9800
    }
  },
  {
    "ph": "s", "id": 9800, "pid": 76337, "tid": -914061504, "ts": 1716454216728219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728286, "dur": 0,
    "args": {
      "External id": 9812,
      "cbid": 317, "correlation": 9812
    }
  },
  {
    "ph": "f", "id": 9812, "pid": 76337, "tid": -914061504, "ts": 1716454216728286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728287, "dur": 0,
    "args": {
      "External id": 9813,
      "cbid": 203, "correlation": 9813
    }
  },
  {
    "ph": "f", "id": 9813, "pid": 76337, "tid": -914061504, "ts": 1716454216728287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728288, "dur": 0,
    "args": {
      "External id": 9814,
      "cbid": 205, "correlation": 9814
    }
  },
  {
    "ph": "f", "id": 9814, "pid": 76337, "tid": -914061504, "ts": 1716454216728288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216730421, "dur": 22,
    "args": {
      "External id": 9818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9818,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9818, "pid": 5, "tid": 7, "ts": 1716454216730421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728302, "dur": 12,
    "args": {
      "External id": 9818,
      "cbid": 211, "correlation": 9818
    }
  },
  {
    "ph": "s", "id": 9818, "pid": 76337, "tid": -914061504, "ts": 1716454216728302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216730445, "dur": 121,
    "args": {
      "External id": 9820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9820,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9820, "pid": 5, "tid": 7, "ts": 1716454216730445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728321, "dur": 6,
    "args": {
      "External id": 9820,
      "cbid": 211, "correlation": 9820
    }
  },
  {
    "ph": "s", "id": 9820, "pid": 76337, "tid": -914061504, "ts": 1716454216728321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216730567, "dur": 22,
    "args": {
      "External id": 9822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9822,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9822, "pid": 5, "tid": 7, "ts": 1716454216730567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728331, "dur": 5,
    "args": {
      "External id": 9822,
      "cbid": 211, "correlation": 9822
    }
  },
  {
    "ph": "s", "id": 9822, "pid": 76337, "tid": -914061504, "ts": 1716454216728331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216730590, "dur": 33,
    "args": {
      "External id": 9828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9828,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9828, "pid": 5, "tid": 7, "ts": 1716454216730590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728358, "dur": 8,
    "args": {
      "External id": 9828,
      "cbid": 211, "correlation": 9828
    }
  },
  {
    "ph": "s", "id": 9828, "pid": 76337, "tid": -914061504, "ts": 1716454216728358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216730625, "dur": 26,
    "args": {
      "External id": 9836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9836,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9836, "pid": 5, "tid": 7, "ts": 1716454216730625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728389, "dur": 8,
    "args": {
      "External id": 9836,
      "cbid": 211, "correlation": 9836
    }
  },
  {
    "ph": "s", "id": 9836, "pid": 76337, "tid": -914061504, "ts": 1716454216728389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728478, "dur": 0,
    "args": {
      "External id": 9846,
      "cbid": 317, "correlation": 9846
    }
  },
  {
    "ph": "f", "id": 9846, "pid": 76337, "tid": -914061504, "ts": 1716454216728478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728479, "dur": 0,
    "args": {
      "External id": 9847,
      "cbid": 203, "correlation": 9847
    }
  },
  {
    "ph": "f", "id": 9847, "pid": 76337, "tid": -914061504, "ts": 1716454216728479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216728480, "dur": 0,
    "args": {
      "External id": 9848,
      "cbid": 205, "correlation": 9848
    }
  },
  {
    "ph": "f", "id": 9848, "pid": 76337, "tid": -914061504, "ts": 1716454216728480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216731383, "dur": 23,
    "args": {
      "External id": 9852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9852,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9852, "pid": 5, "tid": 7, "ts": 1716454216731383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216731357, "dur": 27,
    "args": {
      "External id": 9852,
      "cbid": 211, "correlation": 9852
    }
  },
  {
    "ph": "s", "id": 9852, "pid": 76337, "tid": -914061504, "ts": 1716454216731357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216731407, "dur": 45,
    "args": {
      "External id": 9854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9854,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9854, "pid": 5, "tid": 7, "ts": 1716454216731407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216731387, "dur": 6,
    "args": {
      "External id": 9854,
      "cbid": 211, "correlation": 9854
    }
  },
  {
    "ph": "s", "id": 9854, "pid": 76337, "tid": -914061504, "ts": 1716454216731387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216731453, "dur": 236,
    "args": {
      "External id": 9856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9856,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 9856, "pid": 5, "tid": 7, "ts": 1716454216731453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216731403, "dur": 10,
    "args": {
      "External id": 9856,
      "cbid": 211, "correlation": 9856
    }
  },
  {
    "ph": "s", "id": 9856, "pid": 76337, "tid": -914061504, "ts": 1716454216731403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216731690, "dur": 7,
    "args": {
      "External id": 9858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9858,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9858, "pid": 5, "tid": 7, "ts": 1716454216731690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216731418, "dur": 7,
    "args": {
      "External id": 9858,
      "cbid": 211, "correlation": 9858
    }
  },
  {
    "ph": "s", "id": 9858, "pid": 76337, "tid": -914061504, "ts": 1716454216731418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216732158, "dur": 9,
    "args": {
      "External id": 9864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9864, "pid": 5, "tid": 7, "ts": 1716454216732158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732145, "dur": 12,
    "args": {
      "External id": 9864,
      "cbid": 211, "correlation": 9864
    }
  },
  {
    "ph": "s", "id": 9864, "pid": 76337, "tid": -914061504, "ts": 1716454216732145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216732325, "dur": 12,
    "args": {
      "External id": 9884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9884,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 9884, "pid": 5, "tid": 7, "ts": 1716454216732325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732310, "dur": 14,
    "args": {
      "External id": 9884,
      "cbid": 211, "correlation": 9884
    }
  },
  {
    "ph": "s", "id": 9884, "pid": 76337, "tid": -914061504, "ts": 1716454216732310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216732345, "dur": 4,
    "args": {
      "External id": 9896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9896,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 9896, "pid": 5, "tid": 7, "ts": 1716454216732345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732336, "dur": 8,
    "args": {
      "External id": 9896,
      "cbid": 211, "correlation": 9896
    }
  },
  {
    "ph": "s", "id": 9896, "pid": 76337, "tid": -914061504, "ts": 1716454216732336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216732372, "dur": 13,
    "args": {
      "External id": 9899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9899, "pid": 5, "tid": 7, "ts": 1716454216732372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732364, "dur": 7,
    "args": {
      "External id": 9899,
      "cbid": 211, "correlation": 9899
    }
  },
  {
    "ph": "s", "id": 9899, "pid": 76337, "tid": -914061504, "ts": 1716454216732364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216732428, "dur": 7,
    "args": {
      "External id": 9908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9908,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9908, "pid": 5, "tid": 7, "ts": 1716454216732428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732416, "dur": 10,
    "args": {
      "External id": 9908,
      "cbid": 211, "correlation": 9908
    }
  },
  {
    "ph": "s", "id": 9908, "pid": 76337, "tid": -914061504, "ts": 1716454216732416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732520, "dur": 0,
    "args": {
      "External id": 9918,
      "cbid": 317, "correlation": 9918
    }
  },
  {
    "ph": "f", "id": 9918, "pid": 76337, "tid": -914061504, "ts": 1716454216732520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732521, "dur": 1,
    "args": {
      "External id": 9919,
      "cbid": 203, "correlation": 9919
    }
  },
  {
    "ph": "f", "id": 9919, "pid": 76337, "tid": -914061504, "ts": 1716454216732521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216732523, "dur": 1,
    "args": {
      "External id": 9920,
      "cbid": 205, "correlation": 9920
    }
  },
  {
    "ph": "f", "id": 9920, "pid": 76337, "tid": -914061504, "ts": 1716454216732523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216735326, "dur": 6,
    "args": {
      "External id": 9924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9924,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9924, "pid": 5, "tid": 7, "ts": 1716454216735326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216735303, "dur": 23,
    "args": {
      "External id": 9924,
      "cbid": 211, "correlation": 9924
    }
  },
  {
    "ph": "s", "id": 9924, "pid": 76337, "tid": -914061504, "ts": 1716454216735303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216735338, "dur": 84,
    "args": {
      "External id": 9926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9926,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9926, "pid": 5, "tid": 7, "ts": 1716454216735338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216735329, "dur": 7,
    "args": {
      "External id": 9926,
      "cbid": 211, "correlation": 9926
    }
  },
  {
    "ph": "s", "id": 9926, "pid": 76337, "tid": -914061504, "ts": 1716454216735329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216735425, "dur": 1,
    "args": {
      "External id": 9928,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 9928,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5
    }
  },
  {
    "ph": "f", "id": 9928, "pid": 5, "tid": 7, "ts": 1716454216735425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216735344, "dur": 30,
    "args": {
      "External id": 9928,
      "cbid": 51, "correlation": 9928
    }
  },
  {
    "ph": "s", "id": 9928, "pid": 76337, "tid": -914061504, "ts": 1716454216735344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216735429, "dur": 541,
    "args": {
      "External id": 9929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9929,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9929, "pid": 5, "tid": 7, "ts": 1716454216735429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216735377, "dur": 13,
    "args": {
      "External id": 9929,
      "cbid": 211, "correlation": 9929
    }
  },
  {
    "ph": "s", "id": 9929, "pid": 76337, "tid": -914061504, "ts": 1716454216735377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216735971, "dur": 11,
    "args": {
      "External id": 9931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9931,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9931, "pid": 5, "tid": 7, "ts": 1716454216735971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216735397, "dur": 7,
    "args": {
      "External id": 9931,
      "cbid": 211, "correlation": 9931
    }
  },
  {
    "ph": "s", "id": 9931, "pid": 76337, "tid": -914061504, "ts": 1716454216735397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216736169, "dur": 15,
    "args": {
      "External id": 9937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9937,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9937, "pid": 5, "tid": 7, "ts": 1716454216736169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736156, "dur": 12,
    "args": {
      "External id": 9937,
      "cbid": 211, "correlation": 9937
    }
  },
  {
    "ph": "s", "id": 9937, "pid": 76337, "tid": -914061504, "ts": 1716454216736156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216736243, "dur": 4,
    "args": {
      "External id": 9945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9945,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 9945, "pid": 5, "tid": 7, "ts": 1716454216736243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736233, "dur": 10,
    "args": {
      "External id": 9945,
      "cbid": 211, "correlation": 9945
    }
  },
  {
    "ph": "s", "id": 9945, "pid": 76337, "tid": -914061504, "ts": 1716454216736233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736366, "dur": 4,
    "args": {
      "External id": 9961,
      "cbid": 251, "correlation": 9961
    }
  },
  {
    "ph": "f", "id": 9961, "pid": 76337, "tid": -914061504, "ts": 1716454216736366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736376, "dur": 0,
    "args": {
      "External id": 9963,
      "cbid": 251, "correlation": 9963
    }
  },
  {
    "ph": "f", "id": 9963, "pid": 76337, "tid": -914061504, "ts": 1716454216736376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216736394, "dur": 13,
    "args": {
      "External id": 9964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9964,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9964, "pid": 5, "tid": 7, "ts": 1716454216736394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736379, "dur": 15,
    "args": {
      "External id": 9964,
      "cbid": 211, "correlation": 9964
    }
  },
  {
    "ph": "s", "id": 9964, "pid": 76337, "tid": -914061504, "ts": 1716454216736379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216736410, "dur": 5,
    "args": {
      "External id": 9966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9966,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 9966, "pid": 5, "tid": 7, "ts": 1716454216736410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736399, "dur": 9,
    "args": {
      "External id": 9966,
      "cbid": 211, "correlation": 9966
    }
  },
  {
    "ph": "s", "id": 9966, "pid": 76337, "tid": -914061504, "ts": 1716454216736399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216736504, "dur": 17,
    "args": {
      "External id": 9976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9976,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 9976, "pid": 5, "tid": 7, "ts": 1716454216736504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736492, "dur": 12,
    "args": {
      "External id": 9976,
      "cbid": 211, "correlation": 9976
    }
  },
  {
    "ph": "s", "id": 9976, "pid": 76337, "tid": -914061504, "ts": 1716454216736492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216736590, "dur": 17,
    "args": {
      "External id": 9996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 9996,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 9996, "pid": 5, "tid": 7, "ts": 1716454216736590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736578, "dur": 12,
    "args": {
      "External id": 9996,
      "cbid": 211, "correlation": 9996
    }
  },
  {
    "ph": "s", "id": 9996, "pid": 76337, "tid": -914061504, "ts": 1716454216736578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216736611, "dur": 4,
    "args": {
      "External id": 10008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10008,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 10008, "pid": 5, "tid": 7, "ts": 1716454216736611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736601, "dur": 7,
    "args": {
      "External id": 10008,
      "cbid": 211, "correlation": 10008
    }
  },
  {
    "ph": "s", "id": 10008, "pid": 76337, "tid": -914061504, "ts": 1716454216736601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216736635, "dur": 17,
    "args": {
      "External id": 10011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10011, "pid": 5, "tid": 7, "ts": 1716454216736635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736626, "dur": 8,
    "args": {
      "External id": 10011,
      "cbid": 211, "correlation": 10011
    }
  },
  {
    "ph": "s", "id": 10011, "pid": 76337, "tid": -914061504, "ts": 1716454216736626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216736683, "dur": 11,
    "args": {
      "External id": 10020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10020,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10020, "pid": 5, "tid": 7, "ts": 1716454216736683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736672, "dur": 10,
    "args": {
      "External id": 10020,
      "cbid": 211, "correlation": 10020
    }
  },
  {
    "ph": "s", "id": 10020, "pid": 76337, "tid": -914061504, "ts": 1716454216736672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736787, "dur": 0,
    "args": {
      "External id": 10030,
      "cbid": 317, "correlation": 10030
    }
  },
  {
    "ph": "f", "id": 10030, "pid": 76337, "tid": -914061504, "ts": 1716454216736787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736789, "dur": 0,
    "args": {
      "External id": 10031,
      "cbid": 203, "correlation": 10031
    }
  },
  {
    "ph": "f", "id": 10031, "pid": 76337, "tid": -914061504, "ts": 1716454216736789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216736790, "dur": 0,
    "args": {
      "External id": 10032,
      "cbid": 205, "correlation": 10032
    }
  },
  {
    "ph": "f", "id": 10032, "pid": 76337, "tid": -914061504, "ts": 1716454216736790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216739475, "dur": 11,
    "args": {
      "External id": 10036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10036,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10036, "pid": 5, "tid": 7, "ts": 1716454216739475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216739451, "dur": 25,
    "args": {
      "External id": 10036,
      "cbid": 211, "correlation": 10036
    }
  },
  {
    "ph": "s", "id": 10036, "pid": 76337, "tid": -914061504, "ts": 1716454216739451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216739488, "dur": 165,
    "args": {
      "External id": 10038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10038,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10038, "pid": 5, "tid": 7, "ts": 1716454216739488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216739478, "dur": 6,
    "args": {
      "External id": 10038,
      "cbid": 211, "correlation": 10038
    }
  },
  {
    "ph": "s", "id": 10038, "pid": 76337, "tid": -914061504, "ts": 1716454216739478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216739655, "dur": 1,
    "args": {
      "External id": 10040,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 10040,
      "bytes": 960, "memory bandwidth (GB/s)": 0.4918032786885246
    }
  },
  {
    "ph": "f", "id": 10040, "pid": 5, "tid": 7, "ts": 1716454216739655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216739493, "dur": 16,
    "args": {
      "External id": 10040,
      "cbid": 51, "correlation": 10040
    }
  },
  {
    "ph": "s", "id": 10040, "pid": 76337, "tid": -914061504, "ts": 1716454216739493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216739659, "dur": 668,
    "args": {
      "External id": 10041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10041,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10041, "pid": 5, "tid": 7, "ts": 1716454216739659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216739512, "dur": 11,
    "args": {
      "External id": 10041,
      "cbid": 211, "correlation": 10041
    }
  },
  {
    "ph": "s", "id": 10041, "pid": 76337, "tid": -914061504, "ts": 1716454216739512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216740329, "dur": 12,
    "args": {
      "External id": 10043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10043,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10043, "pid": 5, "tid": 7, "ts": 1716454216740329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216739528, "dur": 7,
    "args": {
      "External id": 10043,
      "cbid": 211, "correlation": 10043
    }
  },
  {
    "ph": "s", "id": 10043, "pid": 76337, "tid": -914061504, "ts": 1716454216739528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216740343, "dur": 15,
    "args": {
      "External id": 10049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10049, "pid": 5, "tid": 7, "ts": 1716454216740343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216740254, "dur": 13,
    "args": {
      "External id": 10049,
      "cbid": 211, "correlation": 10049
    }
  },
  {
    "ph": "s", "id": 10049, "pid": 76337, "tid": -914061504, "ts": 1716454216740254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216740348, "dur": 0,
    "args": {
      "External id": 10059,
      "cbid": 317, "correlation": 10059
    }
  },
  {
    "ph": "f", "id": 10059, "pid": 76337, "tid": -914061504, "ts": 1716454216740348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216740349, "dur": 0,
    "args": {
      "External id": 10060,
      "cbid": 203, "correlation": 10060
    }
  },
  {
    "ph": "f", "id": 10060, "pid": 76337, "tid": -914061504, "ts": 1716454216740349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216740350, "dur": 1,
    "args": {
      "External id": 10061,
      "cbid": 205, "correlation": 10061
    }
  },
  {
    "ph": "f", "id": 10061, "pid": 76337, "tid": -914061504, "ts": 1716454216740350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216742328, "dur": 8,
    "args": {
      "External id": 10065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10065,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10065, "pid": 5, "tid": 7, "ts": 1716454216742328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216742307, "dur": 21,
    "args": {
      "External id": 10065,
      "cbid": 211, "correlation": 10065
    }
  },
  {
    "ph": "s", "id": 10065, "pid": 76337, "tid": -914061504, "ts": 1716454216742307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216742345, "dur": 4,
    "args": {
      "External id": 10067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10067, "pid": 5, "tid": 7, "ts": 1716454216742345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216742335, "dur": 10,
    "args": {
      "External id": 10067,
      "cbid": 211, "correlation": 10067
    }
  },
  {
    "ph": "s", "id": 10067, "pid": 76337, "tid": -914061504, "ts": 1716454216742335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216742349, "dur": 0,
    "args": {
      "External id": 10068,
      "cbid": 51, "correlation": 10068
    }
  },
  {
    "ph": "s", "id": 10068, "pid": 76337, "tid": -914061504, "ts": 1716454216742349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216742360, "dur": 58,
    "args": {
      "External id": 10069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10069,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 10069, "pid": 5, "tid": 7, "ts": 1716454216742360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216742350, "dur": 8,
    "args": {
      "External id": 10069,
      "cbid": 211, "correlation": 10069
    }
  },
  {
    "ph": "s", "id": 10069, "pid": 76337, "tid": -914061504, "ts": 1716454216742350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216742924, "dur": 14,
    "args": {
      "External id": 10074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10074, "pid": 5, "tid": 7, "ts": 1716454216742924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216742913, "dur": 10,
    "args": {
      "External id": 10074,
      "cbid": 211, "correlation": 10074
    }
  },
  {
    "ph": "s", "id": 10074, "pid": 76337, "tid": -914061504, "ts": 1716454216742913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216742967, "dur": 12,
    "args": {
      "External id": 10082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10082,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10082, "pid": 5, "tid": 7, "ts": 1716454216742967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216742957, "dur": 9,
    "args": {
      "External id": 10082,
      "cbid": 211, "correlation": 10082
    }
  },
  {
    "ph": "s", "id": 10082, "pid": 76337, "tid": -914061504, "ts": 1716454216742957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216743035, "dur": 10,
    "args": {
      "External id": 10090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10090,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10090, "pid": 5, "tid": 7, "ts": 1716454216743035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743021, "dur": 12,
    "args": {
      "External id": 10090,
      "cbid": 211, "correlation": 10090
    }
  },
  {
    "ph": "s", "id": 10090, "pid": 76337, "tid": -914061504, "ts": 1716454216743021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216743181, "dur": 18,
    "args": {
      "External id": 10110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10110,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 10110, "pid": 5, "tid": 7, "ts": 1716454216743181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743165, "dur": 15,
    "args": {
      "External id": 10110,
      "cbid": 211, "correlation": 10110
    }
  },
  {
    "ph": "s", "id": 10110, "pid": 76337, "tid": -914061504, "ts": 1716454216743165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216743234, "dur": 4,
    "args": {
      "External id": 10122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10122,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 10122, "pid": 5, "tid": 7, "ts": 1716454216743234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743224, "dur": 10,
    "args": {
      "External id": 10122,
      "cbid": 211, "correlation": 10122
    }
  },
  {
    "ph": "s", "id": 10122, "pid": 76337, "tid": -914061504, "ts": 1716454216743224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216743262, "dur": 17,
    "args": {
      "External id": 10125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10125, "pid": 5, "tid": 7, "ts": 1716454216743262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743254, "dur": 7,
    "args": {
      "External id": 10125,
      "cbid": 211, "correlation": 10125
    }
  },
  {
    "ph": "s", "id": 10125, "pid": 76337, "tid": -914061504, "ts": 1716454216743254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743326, "dur": 0,
    "args": {
      "External id": 10136,
      "cbid": 317, "correlation": 10136
    }
  },
  {
    "ph": "f", "id": 10136, "pid": 76337, "tid": -914061504, "ts": 1716454216743326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743326, "dur": 1,
    "args": {
      "External id": 10137,
      "cbid": 203, "correlation": 10137
    }
  },
  {
    "ph": "f", "id": 10137, "pid": 76337, "tid": -914061504, "ts": 1716454216743326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216743328, "dur": 1,
    "args": {
      "External id": 10138,
      "cbid": 205, "correlation": 10138
    }
  },
  {
    "ph": "f", "id": 10138, "pid": 76337, "tid": -914061504, "ts": 1716454216743328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216745067, "dur": 11,
    "args": {
      "External id": 10142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10142,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10142, "pid": 5, "tid": 7, "ts": 1716454216745067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745048, "dur": 19,
    "args": {
      "External id": 10142,
      "cbid": 211, "correlation": 10142
    }
  },
  {
    "ph": "s", "id": 10142, "pid": 76337, "tid": -914061504, "ts": 1716454216745048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216745084, "dur": 3,
    "args": {
      "External id": 10144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10144, "pid": 5, "tid": 7, "ts": 1716454216745084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745074, "dur": 8,
    "args": {
      "External id": 10144,
      "cbid": 211, "correlation": 10144
    }
  },
  {
    "ph": "s", "id": 10144, "pid": 76337, "tid": -914061504, "ts": 1716454216745074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745085, "dur": 0,
    "args": {
      "External id": 10145,
      "cbid": 51, "correlation": 10145
    }
  },
  {
    "ph": "s", "id": 10145, "pid": 76337, "tid": -914061504, "ts": 1716454216745085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216745095, "dur": 98,
    "args": {
      "External id": 10146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10146,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 10146, "pid": 5, "tid": 7, "ts": 1716454216745095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745086, "dur": 7,
    "args": {
      "External id": 10146,
      "cbid": 211, "correlation": 10146
    }
  },
  {
    "ph": "s", "id": 10146, "pid": 76337, "tid": -914061504, "ts": 1716454216745086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216745596, "dur": 16,
    "args": {
      "External id": 10151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10151, "pid": 5, "tid": 7, "ts": 1716454216745596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745583, "dur": 12,
    "args": {
      "External id": 10151,
      "cbid": 211, "correlation": 10151
    }
  },
  {
    "ph": "s", "id": 10151, "pid": 76337, "tid": -914061504, "ts": 1716454216745583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216745775, "dur": 44,
    "args": {
      "External id": 10162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10162, "pid": 5, "tid": 7, "ts": 1716454216745775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745758, "dur": 17,
    "args": {
      "External id": 10162,
      "cbid": 211, "correlation": 10162
    }
  },
  {
    "ph": "s", "id": 10162, "pid": 76337, "tid": -914061504, "ts": 1716454216745758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216745827, "dur": 19,
    "args": {
      "External id": 10184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10184,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10184, "pid": 5, "tid": 7, "ts": 1716454216745827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745814, "dur": 13,
    "args": {
      "External id": 10184,
      "cbid": 211, "correlation": 10184
    }
  },
  {
    "ph": "s", "id": 10184, "pid": 76337, "tid": -914061504, "ts": 1716454216745814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216745989, "dur": 4,
    "args": {
      "External id": 10195,
      "cbid": 251, "correlation": 10195
    }
  },
  {
    "ph": "f", "id": 10195, "pid": 76337, "tid": -914061504, "ts": 1716454216745989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216746018, "dur": 90,
    "args": {
      "External id": 10196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10196,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10196, "pid": 5, "tid": 7, "ts": 1716454216746018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746002, "dur": 31,
    "args": {
      "External id": 10196,
      "cbid": 211, "correlation": 10196
    }
  },
  {
    "ph": "s", "id": 10196, "pid": 76337, "tid": -914061504, "ts": 1716454216746002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746110, "dur": 1,
    "args": {
      "External id": 10207,
      "cbid": 251, "correlation": 10207
    }
  },
  {
    "ph": "f", "id": 10207, "pid": 76337, "tid": -914061504, "ts": 1716454216746110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216746127, "dur": 86,
    "args": {
      "External id": 10208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10208,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10208, "pid": 5, "tid": 7, "ts": 1716454216746127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746114, "dur": 13,
    "args": {
      "External id": 10208,
      "cbid": 211, "correlation": 10208
    }
  },
  {
    "ph": "s", "id": 10208, "pid": 76337, "tid": -914061504, "ts": 1716454216746114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746182, "dur": 1,
    "args": {
      "External id": 10219,
      "cbid": 251, "correlation": 10219
    }
  },
  {
    "ph": "f", "id": 10219, "pid": 76337, "tid": -914061504, "ts": 1716454216746182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216746215, "dur": 85,
    "args": {
      "External id": 10220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10220,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10220, "pid": 5, "tid": 7, "ts": 1716454216746215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746186, "dur": 13,
    "args": {
      "External id": 10220,
      "cbid": 211, "correlation": 10220
    }
  },
  {
    "ph": "s", "id": 10220, "pid": 76337, "tid": -914061504, "ts": 1716454216746186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216746324, "dur": 140,
    "args": {
      "External id": 10245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10245,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10245, "pid": 5, "tid": 7, "ts": 1716454216746324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746309, "dur": 15,
    "args": {
      "External id": 10245,
      "cbid": 211, "correlation": 10245
    }
  },
  {
    "ph": "s", "id": 10245, "pid": 76337, "tid": -914061504, "ts": 1716454216746309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746447, "dur": 2,
    "args": {
      "External id": 10263,
      "cbid": 251, "correlation": 10263
    }
  },
  {
    "ph": "f", "id": 10263, "pid": 76337, "tid": -914061504, "ts": 1716454216746447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216746471, "dur": 92,
    "args": {
      "External id": 10265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10265,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10265, "pid": 5, "tid": 7, "ts": 1716454216746471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746454, "dur": 16,
    "args": {
      "External id": 10265,
      "cbid": 211, "correlation": 10265
    }
  },
  {
    "ph": "s", "id": 10265, "pid": 76337, "tid": -914061504, "ts": 1716454216746454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216746569, "dur": 10,
    "args": {
      "External id": 10273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10273,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10273, "pid": 5, "tid": 7, "ts": 1716454216746569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746553, "dur": 15,
    "args": {
      "External id": 10273,
      "cbid": 211, "correlation": 10273
    }
  },
  {
    "ph": "s", "id": 10273, "pid": 76337, "tid": -914061504, "ts": 1716454216746553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216746609, "dur": 46,
    "args": {
      "External id": 10281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10281, "pid": 5, "tid": 7, "ts": 1716454216746609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746599, "dur": 10,
    "args": {
      "External id": 10281,
      "cbid": 211, "correlation": 10281
    }
  },
  {
    "ph": "s", "id": 10281, "pid": 76337, "tid": -914061504, "ts": 1716454216746599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216746700, "dur": 17,
    "args": {
      "External id": 10303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10303,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10303, "pid": 5, "tid": 7, "ts": 1716454216746700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746686, "dur": 13,
    "args": {
      "External id": 10303,
      "cbid": 211, "correlation": 10303
    }
  },
  {
    "ph": "s", "id": 10303, "pid": 76337, "tid": -914061504, "ts": 1716454216746686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746789, "dur": 1,
    "args": {
      "External id": 10314,
      "cbid": 251, "correlation": 10314
    }
  },
  {
    "ph": "f", "id": 10314, "pid": 76337, "tid": -914061504, "ts": 1716454216746789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216746809, "dur": 85,
    "args": {
      "External id": 10315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10315,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10315, "pid": 5, "tid": 7, "ts": 1716454216746809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746794, "dur": 14,
    "args": {
      "External id": 10315,
      "cbid": 211, "correlation": 10315
    }
  },
  {
    "ph": "s", "id": 10315, "pid": 76337, "tid": -914061504, "ts": 1716454216746794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746880, "dur": 1,
    "args": {
      "External id": 10326,
      "cbid": 251, "correlation": 10326
    }
  },
  {
    "ph": "f", "id": 10326, "pid": 76337, "tid": -914061504, "ts": 1716454216746880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746885, "dur": 0,
    "args": {
      "External id": 10327,
      "cbid": 251, "correlation": 10327
    }
  },
  {
    "ph": "f", "id": 10327, "pid": 76337, "tid": -914061504, "ts": 1716454216746885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216746903, "dur": 12,
    "args": {
      "External id": 10328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10328,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10328, "pid": 5, "tid": 7, "ts": 1716454216746903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746888, "dur": 14,
    "args": {
      "External id": 10328,
      "cbid": 211, "correlation": 10328
    }
  },
  {
    "ph": "s", "id": 10328, "pid": 76337, "tid": -914061504, "ts": 1716454216746888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216746918, "dur": 5,
    "args": {
      "External id": 10330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10330,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10330, "pid": 5, "tid": 7, "ts": 1716454216746918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746906, "dur": 9,
    "args": {
      "External id": 10330,
      "cbid": 211, "correlation": 10330
    }
  },
  {
    "ph": "s", "id": 10330, "pid": 76337, "tid": -914061504, "ts": 1716454216746906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746987, "dur": 1,
    "args": {
      "External id": 10341,
      "cbid": 251, "correlation": 10341
    }
  },
  {
    "ph": "f", "id": 10341, "pid": 76337, "tid": -914061504, "ts": 1716454216746987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746991, "dur": 0,
    "args": {
      "External id": 10342,
      "cbid": 251, "correlation": 10342
    }
  },
  {
    "ph": "f", "id": 10342, "pid": 76337, "tid": -914061504, "ts": 1716454216746991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216747008, "dur": 9,
    "args": {
      "External id": 10343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10343,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10343, "pid": 5, "tid": 7, "ts": 1716454216747008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216746993, "dur": 14,
    "args": {
      "External id": 10343,
      "cbid": 211, "correlation": 10343
    }
  },
  {
    "ph": "s", "id": 10343, "pid": 76337, "tid": -914061504, "ts": 1716454216746993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216747018, "dur": 3,
    "args": {
      "External id": 10345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10345,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10345, "pid": 5, "tid": 7, "ts": 1716454216747018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747009, "dur": 6,
    "args": {
      "External id": 10345,
      "cbid": 211, "correlation": 10345
    }
  },
  {
    "ph": "s", "id": 10345, "pid": 76337, "tid": -914061504, "ts": 1716454216747009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216747102, "dur": 57,
    "args": {
      "External id": 10370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10370,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10370, "pid": 5, "tid": 7, "ts": 1716454216747102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747089, "dur": 13,
    "args": {
      "External id": 10370,
      "cbid": 211, "correlation": 10370
    }
  },
  {
    "ph": "s", "id": 10370, "pid": 76337, "tid": -914061504, "ts": 1716454216747089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747205, "dur": 2,
    "args": {
      "External id": 10388,
      "cbid": 251, "correlation": 10388
    }
  },
  {
    "ph": "f", "id": 10388, "pid": 76337, "tid": -914061504, "ts": 1716454216747205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216747225, "dur": 88,
    "args": {
      "External id": 10390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10390,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10390, "pid": 5, "tid": 7, "ts": 1716454216747225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747211, "dur": 14,
    "args": {
      "External id": 10390,
      "cbid": 211, "correlation": 10390
    }
  },
  {
    "ph": "s", "id": 10390, "pid": 76337, "tid": -914061504, "ts": 1716454216747211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216747314, "dur": 10,
    "args": {
      "External id": 10398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10398,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10398, "pid": 5, "tid": 7, "ts": 1716454216747314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747285, "dur": 12,
    "args": {
      "External id": 10398,
      "cbid": 211, "correlation": 10398
    }
  },
  {
    "ph": "s", "id": 10398, "pid": 76337, "tid": -914061504, "ts": 1716454216747285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216747336, "dur": 14,
    "args": {
      "External id": 10406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10406,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10406, "pid": 5, "tid": 7, "ts": 1716454216747336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747326, "dur": 9,
    "args": {
      "External id": 10406,
      "cbid": 211, "correlation": 10406
    }
  },
  {
    "ph": "s", "id": 10406, "pid": 76337, "tid": -914061504, "ts": 1716454216747326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216747388, "dur": 17,
    "args": {
      "External id": 10428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10428,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10428, "pid": 5, "tid": 7, "ts": 1716454216747388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216747377, "dur": 10,
    "args": {
      "External id": 10428,
      "cbid": 211, "correlation": 10428
    }
  },
  {
    "ph": "s", "id": 10428, "pid": 76337, "tid": -914061504, "ts": 1716454216747377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216765669, "dur": 231,
    "args": {
      "External id": 10444,
      "cbid": 251, "correlation": 10444
    }
  },
  {
    "ph": "f", "id": 10444, "pid": 76337, "tid": -914061504, "ts": 1716454216765669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216765908, "dur": 0,
    "args": {
      "External id": 10446,
      "cbid": 251, "correlation": 10446
    }
  },
  {
    "ph": "f", "id": 10446, "pid": 76337, "tid": -914061504, "ts": 1716454216765908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216765940, "dur": 487,
    "args": {
      "External id": 10447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10447,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10447, "pid": 5, "tid": 7, "ts": 1716454216765940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216765911, "dur": 30,
    "args": {
      "External id": 10447,
      "cbid": 211, "correlation": 10447
    }
  },
  {
    "ph": "s", "id": 10447, "pid": 76337, "tid": -914061504, "ts": 1716454216765911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216766429, "dur": 66,
    "args": {
      "External id": 10455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10455, "pid": 5, "tid": 7, "ts": 1716454216766429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766063, "dur": 16,
    "args": {
      "External id": 10455,
      "cbid": 211, "correlation": 10455
    }
  },
  {
    "ph": "s", "id": 10455, "pid": 76337, "tid": -914061504, "ts": 1716454216766063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216766497, "dur": 66,
    "args": {
      "External id": 10463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10463, "pid": 5, "tid": 7, "ts": 1716454216766497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766109, "dur": 11,
    "args": {
      "External id": 10463,
      "cbid": 211, "correlation": 10463
    }
  },
  {
    "ph": "s", "id": 10463, "pid": 76337, "tid": -914061504, "ts": 1716454216766109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766207, "dur": 1,
    "args": {
      "External id": 10479,
      "cbid": 251, "correlation": 10479
    }
  },
  {
    "ph": "f", "id": 10479, "pid": 76337, "tid": -914061504, "ts": 1716454216766207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216766564, "dur": 1,
    "args": {
      "External id": 10481,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 10481,
      "bytes": 240, "memory bandwidth (GB/s)": 0.125
    }
  },
  {
    "ph": "f", "id": 10481, "pid": 5, "tid": 7, "ts": 1716454216766564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766213, "dur": 18,
    "args": {
      "External id": 10481,
      "cbid": 51, "correlation": 10481
    }
  },
  {
    "ph": "s", "id": 10481, "pid": 76337, "tid": -914061504, "ts": 1716454216766213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216766568, "dur": 280,
    "args": {
      "External id": 10482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10482,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10482, "pid": 5, "tid": 7, "ts": 1716454216766568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766232, "dur": 11,
    "args": {
      "External id": 10482,
      "cbid": 211, "correlation": 10482
    }
  },
  {
    "ph": "s", "id": 10482, "pid": 76337, "tid": -914061504, "ts": 1716454216766232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216766850, "dur": 14,
    "args": {
      "External id": 10490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10490,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10490, "pid": 5, "tid": 7, "ts": 1716454216766850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766276, "dur": 10,
    "args": {
      "External id": 10490,
      "cbid": 211, "correlation": 10490
    }
  },
  {
    "ph": "s", "id": 10490, "pid": 76337, "tid": -914061504, "ts": 1716454216766276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216766865, "dur": 38,
    "args": {
      "External id": 10501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10501, "pid": 5, "tid": 7, "ts": 1716454216766865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766353, "dur": 13,
    "args": {
      "External id": 10501,
      "cbid": 211, "correlation": 10501
    }
  },
  {
    "ph": "s", "id": 10501, "pid": 76337, "tid": -914061504, "ts": 1716454216766353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766444, "dur": 1,
    "args": {
      "External id": 10513,
      "cbid": 317, "correlation": 10513
    }
  },
  {
    "ph": "f", "id": 10513, "pid": 76337, "tid": -914061504, "ts": 1716454216766444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766446, "dur": 1,
    "args": {
      "External id": 10514,
      "cbid": 203, "correlation": 10514
    }
  },
  {
    "ph": "f", "id": 10514, "pid": 76337, "tid": -914061504, "ts": 1716454216766446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766447, "dur": 1,
    "args": {
      "External id": 10515,
      "cbid": 205, "correlation": 10515
    }
  },
  {
    "ph": "f", "id": 10515, "pid": 76337, "tid": -914061504, "ts": 1716454216766447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216766904, "dur": 13,
    "args": {
      "External id": 10519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10519,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10519, "pid": 5, "tid": 7, "ts": 1716454216766904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766486, "dur": 15,
    "args": {
      "External id": 10519,
      "cbid": 211, "correlation": 10519
    }
  },
  {
    "ph": "s", "id": 10519, "pid": 76337, "tid": -914061504, "ts": 1716454216766486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216766919, "dur": 4,
    "args": {
      "External id": 10521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10521, "pid": 5, "tid": 7, "ts": 1716454216766919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766508, "dur": 7,
    "args": {
      "External id": 10521,
      "cbid": 211, "correlation": 10521
    }
  },
  {
    "ph": "s", "id": 10521, "pid": 76337, "tid": -914061504, "ts": 1716454216766508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766519, "dur": 0,
    "args": {
      "External id": 10522,
      "cbid": 51, "correlation": 10522
    }
  },
  {
    "ph": "s", "id": 10522, "pid": 76337, "tid": -914061504, "ts": 1716454216766519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216766924, "dur": 97,
    "args": {
      "External id": 10523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10523,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 10523, "pid": 5, "tid": 7, "ts": 1716454216766924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766520, "dur": 7,
    "args": {
      "External id": 10523,
      "cbid": 211, "correlation": 10523
    }
  },
  {
    "ph": "s", "id": 10523, "pid": 76337, "tid": -914061504, "ts": 1716454216766520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216767023, "dur": 17,
    "args": {
      "External id": 10528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10528,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10528, "pid": 5, "tid": 7, "ts": 1716454216767023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766553, "dur": 9,
    "args": {
      "External id": 10528,
      "cbid": 211, "correlation": 10528
    }
  },
  {
    "ph": "s", "id": 10528, "pid": 76337, "tid": -914061504, "ts": 1716454216766553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216767041, "dur": 11,
    "args": {
      "External id": 10536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10536,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10536, "pid": 5, "tid": 7, "ts": 1716454216767041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766585, "dur": 9,
    "args": {
      "External id": 10536,
      "cbid": 211, "correlation": 10536
    }
  },
  {
    "ph": "s", "id": 10536, "pid": 76337, "tid": -914061504, "ts": 1716454216766585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216767053, "dur": 18,
    "args": {
      "External id": 10556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10556,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 10556, "pid": 5, "tid": 7, "ts": 1716454216767053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766685, "dur": 13,
    "args": {
      "External id": 10556,
      "cbid": 211, "correlation": 10556
    }
  },
  {
    "ph": "s", "id": 10556, "pid": 76337, "tid": -914061504, "ts": 1716454216766685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216767073, "dur": 5,
    "args": {
      "External id": 10568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10568,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 10568, "pid": 5, "tid": 7, "ts": 1716454216767073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766709, "dur": 7,
    "args": {
      "External id": 10568,
      "cbid": 211, "correlation": 10568
    }
  },
  {
    "ph": "s", "id": 10568, "pid": 76337, "tid": -914061504, "ts": 1716454216766709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216767079, "dur": 18,
    "args": {
      "External id": 10571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10571,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10571, "pid": 5, "tid": 7, "ts": 1716454216767079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766731, "dur": 7,
    "args": {
      "External id": 10571,
      "cbid": 211, "correlation": 10571
    }
  },
  {
    "ph": "s", "id": 10571, "pid": 76337, "tid": -914061504, "ts": 1716454216766731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216767098, "dur": 11,
    "args": {
      "External id": 10580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10580,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10580, "pid": 5, "tid": 7, "ts": 1716454216767098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766781, "dur": 11,
    "args": {
      "External id": 10580,
      "cbid": 211, "correlation": 10580
    }
  },
  {
    "ph": "s", "id": 10580, "pid": 76337, "tid": -914061504, "ts": 1716454216766781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766853, "dur": 0,
    "args": {
      "External id": 10590,
      "cbid": 317, "correlation": 10590
    }
  },
  {
    "ph": "f", "id": 10590, "pid": 76337, "tid": -914061504, "ts": 1716454216766853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766853, "dur": 0,
    "args": {
      "External id": 10591,
      "cbid": 203, "correlation": 10591
    }
  },
  {
    "ph": "f", "id": 10591, "pid": 76337, "tid": -914061504, "ts": 1716454216766853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766854, "dur": 0,
    "args": {
      "External id": 10592,
      "cbid": 205, "correlation": 10592
    }
  },
  {
    "ph": "f", "id": 10592, "pid": 76337, "tid": -914061504, "ts": 1716454216766854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216767111, "dur": 11,
    "args": {
      "External id": 10596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10596, "pid": 5, "tid": 7, "ts": 1716454216767111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766870, "dur": 14,
    "args": {
      "External id": 10596,
      "cbid": 211, "correlation": 10596
    }
  },
  {
    "ph": "s", "id": 10596, "pid": 76337, "tid": -914061504, "ts": 1716454216766870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216767123, "dur": 165,
    "args": {
      "External id": 10598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10598, "pid": 5, "tid": 7, "ts": 1716454216767123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766886, "dur": 5,
    "args": {
      "External id": 10598,
      "cbid": 211, "correlation": 10598
    }
  },
  {
    "ph": "s", "id": 10598, "pid": 76337, "tid": -914061504, "ts": 1716454216766886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216767290, "dur": 1,
    "args": {
      "External id": 10600,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 10600,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 10600, "pid": 5, "tid": 7, "ts": 1716454216767290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766903, "dur": 8,
    "args": {
      "External id": 10600,
      "cbid": 51, "correlation": 10600
    }
  },
  {
    "ph": "s", "id": 10600, "pid": 76337, "tid": -914061504, "ts": 1716454216766903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216767294, "dur": 667,
    "args": {
      "External id": 10601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10601,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10601, "pid": 5, "tid": 7, "ts": 1716454216767294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766913, "dur": 9,
    "args": {
      "External id": 10601,
      "cbid": 211, "correlation": 10601
    }
  },
  {
    "ph": "s", "id": 10601, "pid": 76337, "tid": -914061504, "ts": 1716454216766913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216767962, "dur": 13,
    "args": {
      "External id": 10603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10603,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10603, "pid": 5, "tid": 7, "ts": 1716454216767962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766930, "dur": 7,
    "args": {
      "External id": 10603,
      "cbid": 211, "correlation": 10603
    }
  },
  {
    "ph": "s", "id": 10603, "pid": 76337, "tid": -914061504, "ts": 1716454216766930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216767977, "dur": 15,
    "args": {
      "External id": 10609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10609, "pid": 5, "tid": 7, "ts": 1716454216767977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216766961, "dur": 8,
    "args": {
      "External id": 10609,
      "cbid": 211, "correlation": 10609
    }
  },
  {
    "ph": "s", "id": 10609, "pid": 76337, "tid": -914061504, "ts": 1716454216766961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216767993, "dur": 3,
    "args": {
      "External id": 10617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10617,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 10617, "pid": 5, "tid": 7, "ts": 1716454216767993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767016, "dur": 11,
    "args": {
      "External id": 10617,
      "cbid": 211, "correlation": 10617
    }
  },
  {
    "ph": "s", "id": 10617, "pid": 76337, "tid": -914061504, "ts": 1716454216767016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767102, "dur": 1,
    "args": {
      "External id": 10633,
      "cbid": 251, "correlation": 10633
    }
  },
  {
    "ph": "f", "id": 10633, "pid": 76337, "tid": -914061504, "ts": 1716454216767102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767107, "dur": 0,
    "args": {
      "External id": 10635,
      "cbid": 251, "correlation": 10635
    }
  },
  {
    "ph": "f", "id": 10635, "pid": 76337, "tid": -914061504, "ts": 1716454216767107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216767998, "dur": 13,
    "args": {
      "External id": 10636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10636,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10636, "pid": 5, "tid": 7, "ts": 1716454216767998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767109, "dur": 13,
    "args": {
      "External id": 10636,
      "cbid": 211, "correlation": 10636
    }
  },
  {
    "ph": "s", "id": 10636, "pid": 76337, "tid": -914061504, "ts": 1716454216767109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216768013, "dur": 5,
    "args": {
      "External id": 10638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10638,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10638, "pid": 5, "tid": 7, "ts": 1716454216768013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767125, "dur": 7,
    "args": {
      "External id": 10638,
      "cbid": 211, "correlation": 10638
    }
  },
  {
    "ph": "s", "id": 10638, "pid": 76337, "tid": -914061504, "ts": 1716454216767125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216768019, "dur": 17,
    "args": {
      "External id": 10648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10648, "pid": 5, "tid": 7, "ts": 1716454216768019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767196, "dur": 13,
    "args": {
      "External id": 10648,
      "cbid": 211, "correlation": 10648
    }
  },
  {
    "ph": "s", "id": 10648, "pid": 76337, "tid": -914061504, "ts": 1716454216767196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216768038, "dur": 18,
    "args": {
      "External id": 10668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10668,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 10668, "pid": 5, "tid": 7, "ts": 1716454216768038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767263, "dur": 12,
    "args": {
      "External id": 10668,
      "cbid": 211, "correlation": 10668
    }
  },
  {
    "ph": "s", "id": 10668, "pid": 76337, "tid": -914061504, "ts": 1716454216767263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216768057, "dur": 4,
    "args": {
      "External id": 10680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10680,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 10680, "pid": 5, "tid": 7, "ts": 1716454216768057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767284, "dur": 6,
    "args": {
      "External id": 10680,
      "cbid": 211, "correlation": 10680
    }
  },
  {
    "ph": "s", "id": 10680, "pid": 76337, "tid": -914061504, "ts": 1716454216767284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216768063, "dur": 17,
    "args": {
      "External id": 10683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10683, "pid": 5, "tid": 7, "ts": 1716454216768063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767303, "dur": 6,
    "args": {
      "External id": 10683,
      "cbid": 211, "correlation": 10683
    }
  },
  {
    "ph": "s", "id": 10683, "pid": 76337, "tid": -914061504, "ts": 1716454216767303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216768081, "dur": 11,
    "args": {
      "External id": 10692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10692,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10692, "pid": 5, "tid": 7, "ts": 1716454216768081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767345, "dur": 10,
    "args": {
      "External id": 10692,
      "cbid": 211, "correlation": 10692
    }
  },
  {
    "ph": "s", "id": 10692, "pid": 76337, "tid": -914061504, "ts": 1716454216767345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767422, "dur": 0,
    "args": {
      "External id": 10702,
      "cbid": 317, "correlation": 10702
    }
  },
  {
    "ph": "f", "id": 10702, "pid": 76337, "tid": -914061504, "ts": 1716454216767422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767422, "dur": 0,
    "args": {
      "External id": 10703,
      "cbid": 203, "correlation": 10703
    }
  },
  {
    "ph": "f", "id": 10703, "pid": 76337, "tid": -914061504, "ts": 1716454216767422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767423, "dur": 0,
    "args": {
      "External id": 10704,
      "cbid": 205, "correlation": 10704
    }
  },
  {
    "ph": "f", "id": 10704, "pid": 76337, "tid": -914061504, "ts": 1716454216767423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216768093, "dur": 13,
    "args": {
      "External id": 10708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10708,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10708, "pid": 5, "tid": 7, "ts": 1716454216768093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767437, "dur": 14,
    "args": {
      "External id": 10708,
      "cbid": 211, "correlation": 10708
    }
  },
  {
    "ph": "s", "id": 10708, "pid": 76337, "tid": -914061504, "ts": 1716454216767437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216768107, "dur": 165,
    "args": {
      "External id": 10710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10710,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10710, "pid": 5, "tid": 7, "ts": 1716454216768107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767453, "dur": 5,
    "args": {
      "External id": 10710,
      "cbid": 211, "correlation": 10710
    }
  },
  {
    "ph": "s", "id": 10710, "pid": 76337, "tid": -914061504, "ts": 1716454216767453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216768275, "dur": 1,
    "args": {
      "External id": 10712,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 10712,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 10712, "pid": 5, "tid": 7, "ts": 1716454216768275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767464, "dur": 6,
    "args": {
      "External id": 10712,
      "cbid": 51, "correlation": 10712
    }
  },
  {
    "ph": "s", "id": 10712, "pid": 76337, "tid": -914061504, "ts": 1716454216767464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216768278, "dur": 653,
    "args": {
      "External id": 10713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10713,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10713, "pid": 5, "tid": 7, "ts": 1716454216768278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767472, "dur": 6,
    "args": {
      "External id": 10713,
      "cbid": 211, "correlation": 10713
    }
  },
  {
    "ph": "s", "id": 10713, "pid": 76337, "tid": -914061504, "ts": 1716454216767472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216768933, "dur": 12,
    "args": {
      "External id": 10715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10715,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10715, "pid": 5, "tid": 7, "ts": 1716454216768933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767482, "dur": 5,
    "args": {
      "External id": 10715,
      "cbid": 211, "correlation": 10715
    }
  },
  {
    "ph": "s", "id": 10715, "pid": 76337, "tid": -914061504, "ts": 1716454216767482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216768946, "dur": 15,
    "args": {
      "External id": 10721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10721,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10721, "pid": 5, "tid": 7, "ts": 1716454216768946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767511, "dur": 8,
    "args": {
      "External id": 10721,
      "cbid": 211, "correlation": 10721
    }
  },
  {
    "ph": "s", "id": 10721, "pid": 76337, "tid": -914061504, "ts": 1716454216767511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216768962, "dur": 12,
    "args": {
      "External id": 10729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10729,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10729, "pid": 5, "tid": 7, "ts": 1716454216768962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767543, "dur": 8,
    "args": {
      "External id": 10729,
      "cbid": 211, "correlation": 10729
    }
  },
  {
    "ph": "s", "id": 10729, "pid": 76337, "tid": -914061504, "ts": 1716454216767543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216768975, "dur": 10,
    "args": {
      "External id": 10737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10737,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10737, "pid": 5, "tid": 7, "ts": 1716454216768975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767575, "dur": 8,
    "args": {
      "External id": 10737,
      "cbid": 211, "correlation": 10737
    }
  },
  {
    "ph": "s", "id": 10737, "pid": 76337, "tid": -914061504, "ts": 1716454216767575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216768987, "dur": 18,
    "args": {
      "External id": 10757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10757,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 10757, "pid": 5, "tid": 7, "ts": 1716454216768987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767656, "dur": 14,
    "args": {
      "External id": 10757,
      "cbid": 211, "correlation": 10757
    }
  },
  {
    "ph": "s", "id": 10757, "pid": 76337, "tid": -914061504, "ts": 1716454216767656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216769006, "dur": 4,
    "args": {
      "External id": 10769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10769,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 10769, "pid": 5, "tid": 7, "ts": 1716454216769006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767680, "dur": 6,
    "args": {
      "External id": 10769,
      "cbid": 211, "correlation": 10769
    }
  },
  {
    "ph": "s", "id": 10769, "pid": 76337, "tid": -914061504, "ts": 1716454216767680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216769011, "dur": 17,
    "args": {
      "External id": 10772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10772, "pid": 5, "tid": 7, "ts": 1716454216769011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767698, "dur": 6,
    "args": {
      "External id": 10772,
      "cbid": 211, "correlation": 10772
    }
  },
  {
    "ph": "s", "id": 10772, "pid": 76337, "tid": -914061504, "ts": 1716454216767698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767768, "dur": 0,
    "args": {
      "External id": 10783,
      "cbid": 317, "correlation": 10783
    }
  },
  {
    "ph": "f", "id": 10783, "pid": 76337, "tid": -914061504, "ts": 1716454216767768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767768, "dur": 0,
    "args": {
      "External id": 10784,
      "cbid": 203, "correlation": 10784
    }
  },
  {
    "ph": "f", "id": 10784, "pid": 76337, "tid": -914061504, "ts": 1716454216767768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767769, "dur": 0,
    "args": {
      "External id": 10785,
      "cbid": 205, "correlation": 10785
    }
  },
  {
    "ph": "f", "id": 10785, "pid": 76337, "tid": -914061504, "ts": 1716454216767769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216769030, "dur": 11,
    "args": {
      "External id": 10789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10789,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10789, "pid": 5, "tid": 7, "ts": 1716454216769030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767782, "dur": 13,
    "args": {
      "External id": 10789,
      "cbid": 211, "correlation": 10789
    }
  },
  {
    "ph": "s", "id": 10789, "pid": 76337, "tid": -914061504, "ts": 1716454216767782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216769042, "dur": 4,
    "args": {
      "External id": 10791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10791,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10791, "pid": 5, "tid": 7, "ts": 1716454216769042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767800, "dur": 6,
    "args": {
      "External id": 10791,
      "cbid": 211, "correlation": 10791
    }
  },
  {
    "ph": "s", "id": 10791, "pid": 76337, "tid": -914061504, "ts": 1716454216767800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767808, "dur": 0,
    "args": {
      "External id": 10792,
      "cbid": 51, "correlation": 10792
    }
  },
  {
    "ph": "s", "id": 10792, "pid": 76337, "tid": -914061504, "ts": 1716454216767808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216769047, "dur": 96,
    "args": {
      "External id": 10793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10793,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 10793, "pid": 5, "tid": 7, "ts": 1716454216769047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767809, "dur": 5,
    "args": {
      "External id": 10793,
      "cbid": 211, "correlation": 10793
    }
  },
  {
    "ph": "s", "id": 10793, "pid": 76337, "tid": -914061504, "ts": 1716454216767809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216769145, "dur": 15,
    "args": {
      "External id": 10798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10798,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10798, "pid": 5, "tid": 7, "ts": 1716454216769145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767837, "dur": 9,
    "args": {
      "External id": 10798,
      "cbid": 211, "correlation": 10798
    }
  },
  {
    "ph": "s", "id": 10798, "pid": 76337, "tid": -914061504, "ts": 1716454216767837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216769162, "dur": 43,
    "args": {
      "External id": 10809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10809, "pid": 5, "tid": 7, "ts": 1716454216769162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767937, "dur": 14,
    "args": {
      "External id": 10809,
      "cbid": 211, "correlation": 10809
    }
  },
  {
    "ph": "s", "id": 10809, "pid": 76337, "tid": -914061504, "ts": 1716454216767937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216769206, "dur": 19,
    "args": {
      "External id": 10831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10831,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10831, "pid": 5, "tid": 7, "ts": 1716454216769206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216767992, "dur": 10,
    "args": {
      "External id": 10831,
      "cbid": 211, "correlation": 10831
    }
  },
  {
    "ph": "s", "id": 10831, "pid": 76337, "tid": -914061504, "ts": 1716454216767992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768103, "dur": 2,
    "args": {
      "External id": 10842,
      "cbid": 251, "correlation": 10842
    }
  },
  {
    "ph": "f", "id": 10842, "pid": 76337, "tid": -914061504, "ts": 1716454216768103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216769226, "dur": 91,
    "args": {
      "External id": 10843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10843,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10843, "pid": 5, "tid": 7, "ts": 1716454216769226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768108, "dur": 14,
    "args": {
      "External id": 10843,
      "cbid": 211, "correlation": 10843
    }
  },
  {
    "ph": "s", "id": 10843, "pid": 76337, "tid": -914061504, "ts": 1716454216768108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768193, "dur": 1,
    "args": {
      "External id": 10854,
      "cbid": 251, "correlation": 10854
    }
  },
  {
    "ph": "f", "id": 10854, "pid": 76337, "tid": -914061504, "ts": 1716454216768193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216769319, "dur": 86,
    "args": {
      "External id": 10855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10855,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10855, "pid": 5, "tid": 7, "ts": 1716454216769319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768197, "dur": 12,
    "args": {
      "External id": 10855,
      "cbid": 211, "correlation": 10855
    }
  },
  {
    "ph": "s", "id": 10855, "pid": 76337, "tid": -914061504, "ts": 1716454216768197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768265, "dur": 1,
    "args": {
      "External id": 10866,
      "cbid": 251, "correlation": 10866
    }
  },
  {
    "ph": "f", "id": 10866, "pid": 76337, "tid": -914061504, "ts": 1716454216768265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216769406, "dur": 84,
    "args": {
      "External id": 10867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10867,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10867, "pid": 5, "tid": 7, "ts": 1716454216769406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768269, "dur": 11,
    "args": {
      "External id": 10867,
      "cbid": 211, "correlation": 10867
    }
  },
  {
    "ph": "s", "id": 10867, "pid": 76337, "tid": -914061504, "ts": 1716454216768269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216769492, "dur": 140,
    "args": {
      "External id": 10892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10892,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10892, "pid": 5, "tid": 7, "ts": 1716454216769492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768356, "dur": 13,
    "args": {
      "External id": 10892,
      "cbid": 211, "correlation": 10892
    }
  },
  {
    "ph": "s", "id": 10892, "pid": 76337, "tid": -914061504, "ts": 1716454216768356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768464, "dur": 1,
    "args": {
      "External id": 10910,
      "cbid": 251, "correlation": 10910
    }
  },
  {
    "ph": "f", "id": 10910, "pid": 76337, "tid": -914061504, "ts": 1716454216768464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216769633, "dur": 93,
    "args": {
      "External id": 10912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10912,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10912, "pid": 5, "tid": 7, "ts": 1716454216769633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768470, "dur": 14,
    "args": {
      "External id": 10912,
      "cbid": 211, "correlation": 10912
    }
  },
  {
    "ph": "s", "id": 10912, "pid": 76337, "tid": -914061504, "ts": 1716454216768470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216769727, "dur": 10,
    "args": {
      "External id": 10920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10920,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10920, "pid": 5, "tid": 7, "ts": 1716454216769727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768542, "dur": 13,
    "args": {
      "External id": 10920,
      "cbid": 211, "correlation": 10920
    }
  },
  {
    "ph": "s", "id": 10920, "pid": 76337, "tid": -914061504, "ts": 1716454216768542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216769738, "dur": 46,
    "args": {
      "External id": 10928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10928, "pid": 5, "tid": 7, "ts": 1716454216769738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768583, "dur": 10,
    "args": {
      "External id": 10928,
      "cbid": 211, "correlation": 10928
    }
  },
  {
    "ph": "s", "id": 10928, "pid": 76337, "tid": -914061504, "ts": 1716454216768583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216769786, "dur": 17,
    "args": {
      "External id": 10950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10950,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 10950, "pid": 5, "tid": 7, "ts": 1716454216769786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768635, "dur": 10,
    "args": {
      "External id": 10950,
      "cbid": 211, "correlation": 10950
    }
  },
  {
    "ph": "s", "id": 10950, "pid": 76337, "tid": -914061504, "ts": 1716454216768635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768727, "dur": 1,
    "args": {
      "External id": 10961,
      "cbid": 251, "correlation": 10961
    }
  },
  {
    "ph": "f", "id": 10961, "pid": 76337, "tid": -914061504, "ts": 1716454216768727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216769804, "dur": 86,
    "args": {
      "External id": 10962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10962,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 10962, "pid": 5, "tid": 7, "ts": 1716454216769804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768732, "dur": 12,
    "args": {
      "External id": 10962,
      "cbid": 211, "correlation": 10962
    }
  },
  {
    "ph": "s", "id": 10962, "pid": 76337, "tid": -914061504, "ts": 1716454216768732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768804, "dur": 1,
    "args": {
      "External id": 10973,
      "cbid": 251, "correlation": 10973
    }
  },
  {
    "ph": "f", "id": 10973, "pid": 76337, "tid": -914061504, "ts": 1716454216768804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768807, "dur": 0,
    "args": {
      "External id": 10974,
      "cbid": 251, "correlation": 10974
    }
  },
  {
    "ph": "f", "id": 10974, "pid": 76337, "tid": -914061504, "ts": 1716454216768807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216769892, "dur": 12,
    "args": {
      "External id": 10975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10975,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10975, "pid": 5, "tid": 7, "ts": 1716454216769892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768809, "dur": 12,
    "args": {
      "External id": 10975,
      "cbid": 211, "correlation": 10975
    }
  },
  {
    "ph": "s", "id": 10975, "pid": 76337, "tid": -914061504, "ts": 1716454216768809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216769906, "dur": 5,
    "args": {
      "External id": 10977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10977,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10977, "pid": 5, "tid": 7, "ts": 1716454216769906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768823, "dur": 6,
    "args": {
      "External id": 10977,
      "cbid": 211, "correlation": 10977
    }
  },
  {
    "ph": "s", "id": 10977, "pid": 76337, "tid": -914061504, "ts": 1716454216768823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768882, "dur": 1,
    "args": {
      "External id": 10988,
      "cbid": 251, "correlation": 10988
    }
  },
  {
    "ph": "f", "id": 10988, "pid": 76337, "tid": -914061504, "ts": 1716454216768882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768885, "dur": 0,
    "args": {
      "External id": 10989,
      "cbid": 251, "correlation": 10989
    }
  },
  {
    "ph": "f", "id": 10989, "pid": 76337, "tid": -914061504, "ts": 1716454216768885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216769912, "dur": 9,
    "args": {
      "External id": 10990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10990,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10990, "pid": 5, "tid": 7, "ts": 1716454216769912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768887, "dur": 11,
    "args": {
      "External id": 10990,
      "cbid": 211, "correlation": 10990
    }
  },
  {
    "ph": "s", "id": 10990, "pid": 76337, "tid": -914061504, "ts": 1716454216768887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216769922, "dur": 3,
    "args": {
      "External id": 10992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 10992,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 10992, "pid": 5, "tid": 7, "ts": 1716454216769922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768900, "dur": 5,
    "args": {
      "External id": 10992,
      "cbid": 211, "correlation": 10992
    }
  },
  {
    "ph": "s", "id": 10992, "pid": 76337, "tid": -914061504, "ts": 1716454216768900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216769927, "dur": 57,
    "args": {
      "External id": 11017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11017,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11017, "pid": 5, "tid": 7, "ts": 1716454216769927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216768987, "dur": 14,
    "args": {
      "External id": 11017,
      "cbid": 211, "correlation": 11017
    }
  },
  {
    "ph": "s", "id": 11017, "pid": 76337, "tid": -914061504, "ts": 1716454216768987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769102, "dur": 1,
    "args": {
      "External id": 11035,
      "cbid": 251, "correlation": 11035
    }
  },
  {
    "ph": "f", "id": 11035, "pid": 76337, "tid": -914061504, "ts": 1716454216769102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216769984, "dur": 87,
    "args": {
      "External id": 11037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11037,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 11037, "pid": 5, "tid": 7, "ts": 1716454216769984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769108, "dur": 15,
    "args": {
      "External id": 11037,
      "cbid": 211, "correlation": 11037
    }
  },
  {
    "ph": "s", "id": 11037, "pid": 76337, "tid": -914061504, "ts": 1716454216769108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216770073, "dur": 9,
    "args": {
      "External id": 11045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11045,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11045, "pid": 5, "tid": 7, "ts": 1716454216770073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769181, "dur": 12,
    "args": {
      "External id": 11045,
      "cbid": 211, "correlation": 11045
    }
  },
  {
    "ph": "s", "id": 11045, "pid": 76337, "tid": -914061504, "ts": 1716454216769181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216770084, "dur": 14,
    "args": {
      "External id": 11053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11053,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11053, "pid": 5, "tid": 7, "ts": 1716454216770084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769220, "dur": 9,
    "args": {
      "External id": 11053,
      "cbid": 211, "correlation": 11053
    }
  },
  {
    "ph": "s", "id": 11053, "pid": 76337, "tid": -914061504, "ts": 1716454216769220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216770098, "dur": 17,
    "args": {
      "External id": 11075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11075,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11075, "pid": 5, "tid": 7, "ts": 1716454216770098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769272, "dur": 10,
    "args": {
      "External id": 11075,
      "cbid": 211, "correlation": 11075
    }
  },
  {
    "ph": "s", "id": 11075, "pid": 76337, "tid": -914061504, "ts": 1716454216769272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769359, "dur": 1,
    "args": {
      "External id": 11091,
      "cbid": 251, "correlation": 11091
    }
  },
  {
    "ph": "f", "id": 11091, "pid": 76337, "tid": -914061504, "ts": 1716454216769359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769364, "dur": 0,
    "args": {
      "External id": 11093,
      "cbid": 251, "correlation": 11093
    }
  },
  {
    "ph": "f", "id": 11093, "pid": 76337, "tid": -914061504, "ts": 1716454216769364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216770117, "dur": 494,
    "args": {
      "External id": 11094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11094,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11094, "pid": 5, "tid": 7, "ts": 1716454216770117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769366, "dur": 13,
    "args": {
      "External id": 11094,
      "cbid": 211, "correlation": 11094
    }
  },
  {
    "ph": "s", "id": 11094, "pid": 76337, "tid": -914061504, "ts": 1716454216769366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216770612, "dur": 66,
    "args": {
      "External id": 11102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11102, "pid": 5, "tid": 7, "ts": 1716454216770612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769435, "dur": 13,
    "args": {
      "External id": 11102,
      "cbid": 211, "correlation": 11102
    }
  },
  {
    "ph": "s", "id": 11102, "pid": 76337, "tid": -914061504, "ts": 1716454216769435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216770680, "dur": 66,
    "args": {
      "External id": 11110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11110, "pid": 5, "tid": 7, "ts": 1716454216770680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769467, "dur": 8,
    "args": {
      "External id": 11110,
      "cbid": 211, "correlation": 11110
    }
  },
  {
    "ph": "s", "id": 11110, "pid": 76337, "tid": -914061504, "ts": 1716454216769467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769547, "dur": 1,
    "args": {
      "External id": 11126,
      "cbid": 251, "correlation": 11126
    }
  },
  {
    "ph": "f", "id": 11126, "pid": 76337, "tid": -914061504, "ts": 1716454216769547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216770748, "dur": 1,
    "args": {
      "External id": 11128,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11128,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 11128, "pid": 5, "tid": 7, "ts": 1716454216770748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769552, "dur": 10,
    "args": {
      "External id": 11128,
      "cbid": 51, "correlation": 11128
    }
  },
  {
    "ph": "s", "id": 11128, "pid": 76337, "tid": -914061504, "ts": 1716454216769552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216770752, "dur": 271,
    "args": {
      "External id": 11129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11129,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 11129, "pid": 5, "tid": 7, "ts": 1716454216770752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769564, "dur": 12,
    "args": {
      "External id": 11129,
      "cbid": 211, "correlation": 11129
    }
  },
  {
    "ph": "s", "id": 11129, "pid": 76337, "tid": -914061504, "ts": 1716454216769564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216771024, "dur": 13,
    "args": {
      "External id": 11137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11137,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11137, "pid": 5, "tid": 7, "ts": 1716454216771024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769606, "dur": 10,
    "args": {
      "External id": 11137,
      "cbid": 211, "correlation": 11137
    }
  },
  {
    "ph": "s", "id": 11137, "pid": 76337, "tid": -914061504, "ts": 1716454216769606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216771039, "dur": 38,
    "args": {
      "External id": 11148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11148,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11148, "pid": 5, "tid": 7, "ts": 1716454216771039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769669, "dur": 12,
    "args": {
      "External id": 11148,
      "cbid": 211, "correlation": 11148
    }
  },
  {
    "ph": "s", "id": 11148, "pid": 76337, "tid": -914061504, "ts": 1716454216769669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769734, "dur": 0,
    "args": {
      "External id": 11160,
      "cbid": 317, "correlation": 11160
    }
  },
  {
    "ph": "f", "id": 11160, "pid": 76337, "tid": -914061504, "ts": 1716454216769734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769735, "dur": 0,
    "args": {
      "External id": 11161,
      "cbid": 203, "correlation": 11161
    }
  },
  {
    "ph": "f", "id": 11161, "pid": 76337, "tid": -914061504, "ts": 1716454216769735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769736, "dur": 0,
    "args": {
      "External id": 11162,
      "cbid": 205, "correlation": 11162
    }
  },
  {
    "ph": "f", "id": 11162, "pid": 76337, "tid": -914061504, "ts": 1716454216769736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216771078, "dur": 13,
    "args": {
      "External id": 11166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11166,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11166, "pid": 5, "tid": 7, "ts": 1716454216771078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769751, "dur": 12,
    "args": {
      "External id": 11166,
      "cbid": 211, "correlation": 11166
    }
  },
  {
    "ph": "s", "id": 11166, "pid": 76337, "tid": -914061504, "ts": 1716454216769751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216771093, "dur": 4,
    "args": {
      "External id": 11168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 11168, "pid": 5, "tid": 7, "ts": 1716454216771093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769768, "dur": 9,
    "args": {
      "External id": 11168,
      "cbid": 211, "correlation": 11168
    }
  },
  {
    "ph": "s", "id": 11168, "pid": 76337, "tid": -914061504, "ts": 1716454216769768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769780, "dur": 0,
    "args": {
      "External id": 11169,
      "cbid": 51, "correlation": 11169
    }
  },
  {
    "ph": "s", "id": 11169, "pid": 76337, "tid": -914061504, "ts": 1716454216769780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216771098, "dur": 99,
    "args": {
      "External id": 11170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11170,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 11170, "pid": 5, "tid": 7, "ts": 1716454216771098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769781, "dur": 5,
    "args": {
      "External id": 11170,
      "cbid": 211, "correlation": 11170
    }
  },
  {
    "ph": "s", "id": 11170, "pid": 76337, "tid": -914061504, "ts": 1716454216769781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216771198, "dur": 17,
    "args": {
      "External id": 11175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11175, "pid": 5, "tid": 7, "ts": 1716454216771198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769808, "dur": 9,
    "args": {
      "External id": 11175,
      "cbid": 211, "correlation": 11175
    }
  },
  {
    "ph": "s", "id": 11175, "pid": 76337, "tid": -914061504, "ts": 1716454216769808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216771216, "dur": 12,
    "args": {
      "External id": 11183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11183,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11183, "pid": 5, "tid": 7, "ts": 1716454216771216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769839, "dur": 8,
    "args": {
      "External id": 11183,
      "cbid": 211, "correlation": 11183
    }
  },
  {
    "ph": "s", "id": 11183, "pid": 76337, "tid": -914061504, "ts": 1716454216769839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769913, "dur": 0,
    "args": {
      "External id": 11193,
      "cbid": 317, "correlation": 11193
    }
  },
  {
    "ph": "f", "id": 11193, "pid": 76337, "tid": -914061504, "ts": 1716454216769913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769913, "dur": 0,
    "args": {
      "External id": 11194,
      "cbid": 203, "correlation": 11194
    }
  },
  {
    "ph": "f", "id": 11194, "pid": 76337, "tid": -914061504, "ts": 1716454216769913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216769914, "dur": 0,
    "args": {
      "External id": 11195,
      "cbid": 205, "correlation": 11195
    }
  },
  {
    "ph": "f", "id": 11195, "pid": 76337, "tid": -914061504, "ts": 1716454216769914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216773089, "dur": 12,
    "args": {
      "External id": 11199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11199,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11199, "pid": 5, "tid": 7, "ts": 1716454216773089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216773058, "dur": 32,
    "args": {
      "External id": 11199,
      "cbid": 211, "correlation": 11199
    }
  },
  {
    "ph": "s", "id": 11199, "pid": 76337, "tid": -914061504, "ts": 1716454216773058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216773103, "dur": 166,
    "args": {
      "External id": 11201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11201,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11201, "pid": 5, "tid": 7, "ts": 1716454216773103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216773094, "dur": 6,
    "args": {
      "External id": 11201,
      "cbid": 211, "correlation": 11201
    }
  },
  {
    "ph": "s", "id": 11201, "pid": 76337, "tid": -914061504, "ts": 1716454216773094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216773271, "dur": 1,
    "args": {
      "External id": 11203,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11203,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 11203, "pid": 5, "tid": 7, "ts": 1716454216773271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216773109, "dur": 12,
    "args": {
      "External id": 11203,
      "cbid": 51, "correlation": 11203
    }
  },
  {
    "ph": "s", "id": 11203, "pid": 76337, "tid": -914061504, "ts": 1716454216773109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216773295, "dur": 196,
    "args": {
      "External id": 11204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11204,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 11204, "pid": 5, "tid": 7, "ts": 1716454216773295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216773123, "dur": 172,
    "args": {
      "External id": 11204,
      "cbid": 211, "correlation": 11204
    }
  },
  {
    "ph": "s", "id": 11204, "pid": 76337, "tid": -914061504, "ts": 1716454216773123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216773492, "dur": 6,
    "args": {
      "External id": 11206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11206,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11206, "pid": 5, "tid": 7, "ts": 1716454216773492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216773303, "dur": 9,
    "args": {
      "External id": 11206,
      "cbid": 211, "correlation": 11206
    }
  },
  {
    "ph": "s", "id": 11206, "pid": 76337, "tid": -914061504, "ts": 1716454216773303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216774107, "dur": 8,
    "args": {
      "External id": 11212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11212, "pid": 5, "tid": 7, "ts": 1716454216774107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774094, "dur": 13,
    "args": {
      "External id": 11212,
      "cbid": 211, "correlation": 11212
    }
  },
  {
    "ph": "s", "id": 11212, "pid": 76337, "tid": -914061504, "ts": 1716454216774094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216774315, "dur": 11,
    "args": {
      "External id": 11232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11232,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11232, "pid": 5, "tid": 7, "ts": 1716454216774315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774301, "dur": 14,
    "args": {
      "External id": 11232,
      "cbid": 211, "correlation": 11232
    }
  },
  {
    "ph": "s", "id": 11232, "pid": 76337, "tid": -914061504, "ts": 1716454216774301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216774336, "dur": 4,
    "args": {
      "External id": 11244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11244,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11244, "pid": 5, "tid": 7, "ts": 1716454216774336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774327, "dur": 9,
    "args": {
      "External id": 11244,
      "cbid": 211, "correlation": 11244
    }
  },
  {
    "ph": "s", "id": 11244, "pid": 76337, "tid": -914061504, "ts": 1716454216774327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216774365, "dur": 8,
    "args": {
      "External id": 11247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11247, "pid": 5, "tid": 7, "ts": 1716454216774365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774355, "dur": 8,
    "args": {
      "External id": 11247,
      "cbid": 211, "correlation": 11247
    }
  },
  {
    "ph": "s", "id": 11247, "pid": 76337, "tid": -914061504, "ts": 1716454216774355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216774423, "dur": 5,
    "args": {
      "External id": 11256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11256,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11256, "pid": 5, "tid": 7, "ts": 1716454216774423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774412, "dur": 10,
    "args": {
      "External id": 11256,
      "cbid": 211, "correlation": 11256
    }
  },
  {
    "ph": "s", "id": 11256, "pid": 76337, "tid": -914061504, "ts": 1716454216774412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774502, "dur": 0,
    "args": {
      "External id": 11266,
      "cbid": 317, "correlation": 11266
    }
  },
  {
    "ph": "f", "id": 11266, "pid": 76337, "tid": -914061504, "ts": 1716454216774502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774503, "dur": 0,
    "args": {
      "External id": 11267,
      "cbid": 203, "correlation": 11267
    }
  },
  {
    "ph": "f", "id": 11267, "pid": 76337, "tid": -914061504, "ts": 1716454216774503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216774504, "dur": 0,
    "args": {
      "External id": 11268,
      "cbid": 205, "correlation": 11268
    }
  },
  {
    "ph": "f", "id": 11268, "pid": 76337, "tid": -914061504, "ts": 1716454216774504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216777609, "dur": 6,
    "args": {
      "External id": 11272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11272,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11272, "pid": 5, "tid": 7, "ts": 1716454216777609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216777585, "dur": 26,
    "args": {
      "External id": 11272,
      "cbid": 211, "correlation": 11272
    }
  },
  {
    "ph": "s", "id": 11272, "pid": 76337, "tid": -914061504, "ts": 1716454216777585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216777621, "dur": 164,
    "args": {
      "External id": 11274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11274,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11274, "pid": 5, "tid": 7, "ts": 1716454216777621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216777614, "dur": 6,
    "args": {
      "External id": 11274,
      "cbid": 211, "correlation": 11274
    }
  },
  {
    "ph": "s", "id": 11274, "pid": 76337, "tid": -914061504, "ts": 1716454216777614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216777788, "dur": 1,
    "args": {
      "External id": 11276,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11276,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 11276, "pid": 5, "tid": 7, "ts": 1716454216777788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216777629, "dur": 14,
    "args": {
      "External id": 11276,
      "cbid": 51, "correlation": 11276
    }
  },
  {
    "ph": "s", "id": 11276, "pid": 76337, "tid": -914061504, "ts": 1716454216777629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216777791, "dur": 273,
    "args": {
      "External id": 11277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11277,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11277, "pid": 5, "tid": 7, "ts": 1716454216777791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216777645, "dur": 12,
    "args": {
      "External id": 11277,
      "cbid": 211, "correlation": 11277
    }
  },
  {
    "ph": "s", "id": 11277, "pid": 76337, "tid": -914061504, "ts": 1716454216777645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216778066, "dur": 6,
    "args": {
      "External id": 11279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11279,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11279, "pid": 5, "tid": 7, "ts": 1716454216778066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216777663, "dur": 7,
    "args": {
      "External id": 11279,
      "cbid": 211, "correlation": 11279
    }
  },
  {
    "ph": "s", "id": 11279, "pid": 76337, "tid": -914061504, "ts": 1716454216777663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216778435, "dur": 6,
    "args": {
      "External id": 11285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11285,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11285, "pid": 5, "tid": 7, "ts": 1716454216778435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778420, "dur": 15,
    "args": {
      "External id": 11285,
      "cbid": 211, "correlation": 11285
    }
  },
  {
    "ph": "s", "id": 11285, "pid": 76337, "tid": -914061504, "ts": 1716454216778420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216778530, "dur": 3,
    "args": {
      "External id": 11293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11293,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 11293, "pid": 5, "tid": 7, "ts": 1716454216778530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778517, "dur": 12,
    "args": {
      "External id": 11293,
      "cbid": 211, "correlation": 11293
    }
  },
  {
    "ph": "s", "id": 11293, "pid": 76337, "tid": -914061504, "ts": 1716454216778517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778668, "dur": 4,
    "args": {
      "External id": 11309,
      "cbid": 251, "correlation": 11309
    }
  },
  {
    "ph": "f", "id": 11309, "pid": 76337, "tid": -914061504, "ts": 1716454216778668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778679, "dur": 0,
    "args": {
      "External id": 11311,
      "cbid": 251, "correlation": 11311
    }
  },
  {
    "ph": "f", "id": 11311, "pid": 76337, "tid": -914061504, "ts": 1716454216778679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216778699, "dur": 13,
    "args": {
      "External id": 11312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11312,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11312, "pid": 5, "tid": 7, "ts": 1716454216778699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778684, "dur": 16,
    "args": {
      "External id": 11312,
      "cbid": 211, "correlation": 11312
    }
  },
  {
    "ph": "s", "id": 11312, "pid": 76337, "tid": -914061504, "ts": 1716454216778684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216778715, "dur": 5,
    "args": {
      "External id": 11314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11314,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11314, "pid": 5, "tid": 7, "ts": 1716454216778715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778704, "dur": 8,
    "args": {
      "External id": 11314,
      "cbid": 211, "correlation": 11314
    }
  },
  {
    "ph": "s", "id": 11314, "pid": 76337, "tid": -914061504, "ts": 1716454216778704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216778811, "dur": 6,
    "args": {
      "External id": 11324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11324, "pid": 5, "tid": 7, "ts": 1716454216778811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778798, "dur": 13,
    "args": {
      "External id": 11324,
      "cbid": 211, "correlation": 11324
    }
  },
  {
    "ph": "s", "id": 11324, "pid": 76337, "tid": -914061504, "ts": 1716454216778798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216778904, "dur": 10,
    "args": {
      "External id": 11344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11344,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11344, "pid": 5, "tid": 7, "ts": 1716454216778904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778891, "dur": 12,
    "args": {
      "External id": 11344,
      "cbid": 211, "correlation": 11344
    }
  },
  {
    "ph": "s", "id": 11344, "pid": 76337, "tid": -914061504, "ts": 1716454216778891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216778923, "dur": 4,
    "args": {
      "External id": 11356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11356,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11356, "pid": 5, "tid": 7, "ts": 1716454216778923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778915, "dur": 8,
    "args": {
      "External id": 11356,
      "cbid": 211, "correlation": 11356
    }
  },
  {
    "ph": "s", "id": 11356, "pid": 76337, "tid": -914061504, "ts": 1716454216778915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216778950, "dur": 7,
    "args": {
      "External id": 11359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11359, "pid": 5, "tid": 7, "ts": 1716454216778950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778941, "dur": 7,
    "args": {
      "External id": 11359,
      "cbid": 211, "correlation": 11359
    }
  },
  {
    "ph": "s", "id": 11359, "pid": 76337, "tid": -914061504, "ts": 1716454216778941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216779008, "dur": 5,
    "args": {
      "External id": 11368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11368,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11368, "pid": 5, "tid": 7, "ts": 1716454216779008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216778996, "dur": 11,
    "args": {
      "External id": 11368,
      "cbid": 211, "correlation": 11368
    }
  },
  {
    "ph": "s", "id": 11368, "pid": 76337, "tid": -914061504, "ts": 1716454216778996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779105, "dur": 0,
    "args": {
      "External id": 11378,
      "cbid": 317, "correlation": 11378
    }
  },
  {
    "ph": "f", "id": 11378, "pid": 76337, "tid": -914061504, "ts": 1716454216779105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779106, "dur": 1,
    "args": {
      "External id": 11379,
      "cbid": 203, "correlation": 11379
    }
  },
  {
    "ph": "f", "id": 11379, "pid": 76337, "tid": -914061504, "ts": 1716454216779106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779108, "dur": 0,
    "args": {
      "External id": 11380,
      "cbid": 205, "correlation": 11380
    }
  },
  {
    "ph": "f", "id": 11380, "pid": 76337, "tid": -914061504, "ts": 1716454216779108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216779138, "dur": 5,
    "args": {
      "External id": 11384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11384,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11384, "pid": 5, "tid": 7, "ts": 1716454216779138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779124, "dur": 14,
    "args": {
      "External id": 11384,
      "cbid": 211, "correlation": 11384
    }
  },
  {
    "ph": "s", "id": 11384, "pid": 76337, "tid": -914061504, "ts": 1716454216779124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216779148, "dur": 163,
    "args": {
      "External id": 11386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11386,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11386, "pid": 5, "tid": 7, "ts": 1716454216779148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779140, "dur": 6,
    "args": {
      "External id": 11386,
      "cbid": 211, "correlation": 11386
    }
  },
  {
    "ph": "s", "id": 11386, "pid": 76337, "tid": -914061504, "ts": 1716454216779140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216779313, "dur": 1,
    "args": {
      "External id": 11388,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11388,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 11388, "pid": 5, "tid": 7, "ts": 1716454216779313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779152, "dur": 8,
    "args": {
      "External id": 11388,
      "cbid": 51, "correlation": 11388
    }
  },
  {
    "ph": "s", "id": 11388, "pid": 76337, "tid": -914061504, "ts": 1716454216779152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216779317, "dur": 263,
    "args": {
      "External id": 11389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11389,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11389, "pid": 5, "tid": 7, "ts": 1716454216779317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779161, "dur": 7,
    "args": {
      "External id": 11389,
      "cbid": 211, "correlation": 11389
    }
  },
  {
    "ph": "s", "id": 11389, "pid": 76337, "tid": -914061504, "ts": 1716454216779161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216779581, "dur": 6,
    "args": {
      "External id": 11391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11391,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11391, "pid": 5, "tid": 7, "ts": 1716454216779581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779171, "dur": 6,
    "args": {
      "External id": 11391,
      "cbid": 211, "correlation": 11391
    }
  },
  {
    "ph": "s", "id": 11391, "pid": 76337, "tid": -914061504, "ts": 1716454216779171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216779588, "dur": 6,
    "args": {
      "External id": 11397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11397, "pid": 5, "tid": 7, "ts": 1716454216779588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779202, "dur": 8,
    "args": {
      "External id": 11397,
      "cbid": 211, "correlation": 11397
    }
  },
  {
    "ph": "s", "id": 11397, "pid": 76337, "tid": -914061504, "ts": 1716454216779202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216779596, "dur": 5,
    "args": {
      "External id": 11405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11405,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11405, "pid": 5, "tid": 7, "ts": 1716454216779596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779236, "dur": 9,
    "args": {
      "External id": 11405,
      "cbid": 211, "correlation": 11405
    }
  },
  {
    "ph": "s", "id": 11405, "pid": 76337, "tid": -914061504, "ts": 1716454216779236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779280, "dur": 1,
    "args": {
      "External id": 11413,
      "cbid": 317, "correlation": 11413
    }
  },
  {
    "ph": "f", "id": 11413, "pid": 76337, "tid": -914061504, "ts": 1716454216779280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216779282, "dur": 10016,
    "args": {
      "External id": 11414,
      "cbid": 20, "correlation": 11414
    }
  },
  {
    "ph": "f", "id": 11414, "pid": 76337, "tid": -914061504, "ts": 1716454216779282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216789338, "dur": 5,
    "args": {
      "External id": 11417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11417,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11417, "pid": 5, "tid": 7, "ts": 1716454216789338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789315, "dur": 24,
    "args": {
      "External id": 11417,
      "cbid": 211, "correlation": 11417
    }
  },
  {
    "ph": "s", "id": 11417, "pid": 76337, "tid": -914061504, "ts": 1716454216789315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216789443, "dur": 10,
    "args": {
      "External id": 11437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11437,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11437, "pid": 5, "tid": 7, "ts": 1716454216789443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789430, "dur": 12,
    "args": {
      "External id": 11437,
      "cbid": 211, "correlation": 11437
    }
  },
  {
    "ph": "s", "id": 11437, "pid": 76337, "tid": -914061504, "ts": 1716454216789430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216789461, "dur": 4,
    "args": {
      "External id": 11449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11449,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11449, "pid": 5, "tid": 7, "ts": 1716454216789461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789453, "dur": 7,
    "args": {
      "External id": 11449,
      "cbid": 211, "correlation": 11449
    }
  },
  {
    "ph": "s", "id": 11449, "pid": 76337, "tid": -914061504, "ts": 1716454216789453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216789482, "dur": 6,
    "args": {
      "External id": 11452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11452, "pid": 5, "tid": 7, "ts": 1716454216789482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789474, "dur": 7,
    "args": {
      "External id": 11452,
      "cbid": 211, "correlation": 11452
    }
  },
  {
    "ph": "s", "id": 11452, "pid": 76337, "tid": -914061504, "ts": 1716454216789474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216789526, "dur": 5,
    "args": {
      "External id": 11461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11461,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11461, "pid": 5, "tid": 7, "ts": 1716454216789526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789515, "dur": 10,
    "args": {
      "External id": 11461,
      "cbid": 211, "correlation": 11461
    }
  },
  {
    "ph": "s", "id": 11461, "pid": 76337, "tid": -914061504, "ts": 1716454216789515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789587, "dur": 0,
    "args": {
      "External id": 11471,
      "cbid": 317, "correlation": 11471
    }
  },
  {
    "ph": "f", "id": 11471, "pid": 76337, "tid": -914061504, "ts": 1716454216789587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789588, "dur": 0,
    "args": {
      "External id": 11472,
      "cbid": 203, "correlation": 11472
    }
  },
  {
    "ph": "f", "id": 11472, "pid": 76337, "tid": -914061504, "ts": 1716454216789588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789589, "dur": 0,
    "args": {
      "External id": 11473,
      "cbid": 205, "correlation": 11473
    }
  },
  {
    "ph": "f", "id": 11473, "pid": 76337, "tid": -914061504, "ts": 1716454216789589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216789616, "dur": 5,
    "args": {
      "External id": 11477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11477,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11477, "pid": 5, "tid": 7, "ts": 1716454216789616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789604, "dur": 12,
    "args": {
      "External id": 11477,
      "cbid": 211, "correlation": 11477
    }
  },
  {
    "ph": "s", "id": 11477, "pid": 76337, "tid": -914061504, "ts": 1716454216789604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216789626, "dur": 164,
    "args": {
      "External id": 11479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11479,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11479, "pid": 5, "tid": 7, "ts": 1716454216789626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789619, "dur": 6,
    "args": {
      "External id": 11479,
      "cbid": 211, "correlation": 11479
    }
  },
  {
    "ph": "s", "id": 11479, "pid": 76337, "tid": -914061504, "ts": 1716454216789619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216789792, "dur": 1,
    "args": {
      "External id": 11481,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11481,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 11481, "pid": 5, "tid": 7, "ts": 1716454216789792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789631, "dur": 8,
    "args": {
      "External id": 11481,
      "cbid": 51, "correlation": 11481
    }
  },
  {
    "ph": "s", "id": 11481, "pid": 76337, "tid": -914061504, "ts": 1716454216789631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216789796, "dur": 261,
    "args": {
      "External id": 11482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11482,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11482, "pid": 5, "tid": 7, "ts": 1716454216789796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789640, "dur": 6,
    "args": {
      "External id": 11482,
      "cbid": 211, "correlation": 11482
    }
  },
  {
    "ph": "s", "id": 11482, "pid": 76337, "tid": -914061504, "ts": 1716454216789640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790058, "dur": 5,
    "args": {
      "External id": 11484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11484,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11484, "pid": 5, "tid": 7, "ts": 1716454216790058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789651, "dur": 5,
    "args": {
      "External id": 11484,
      "cbid": 211, "correlation": 11484
    }
  },
  {
    "ph": "s", "id": 11484, "pid": 76337, "tid": -914061504, "ts": 1716454216789651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216790065, "dur": 6,
    "args": {
      "External id": 11490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11490, "pid": 5, "tid": 7, "ts": 1716454216790065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789680, "dur": 10,
    "args": {
      "External id": 11490,
      "cbid": 211, "correlation": 11490
    }
  },
  {
    "ph": "s", "id": 11490, "pid": 76337, "tid": -914061504, "ts": 1716454216789680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216790072, "dur": 3,
    "args": {
      "External id": 11498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11498,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 11498, "pid": 5, "tid": 7, "ts": 1716454216790072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789726, "dur": 10,
    "args": {
      "External id": 11498,
      "cbid": 211, "correlation": 11498
    }
  },
  {
    "ph": "s", "id": 11498, "pid": 76337, "tid": -914061504, "ts": 1716454216789726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789823, "dur": 2,
    "args": {
      "External id": 11514,
      "cbid": 251, "correlation": 11514
    }
  },
  {
    "ph": "f", "id": 11514, "pid": 76337, "tid": -914061504, "ts": 1716454216789823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789829, "dur": 0,
    "args": {
      "External id": 11516,
      "cbid": 251, "correlation": 11516
    }
  },
  {
    "ph": "f", "id": 11516, "pid": 76337, "tid": -914061504, "ts": 1716454216789829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216790077, "dur": 11,
    "args": {
      "External id": 11517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11517,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11517, "pid": 5, "tid": 7, "ts": 1716454216790077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789831, "dur": 13,
    "args": {
      "External id": 11517,
      "cbid": 211, "correlation": 11517
    }
  },
  {
    "ph": "s", "id": 11517, "pid": 76337, "tid": -914061504, "ts": 1716454216789831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216790089, "dur": 4,
    "args": {
      "External id": 11519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11519,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11519, "pid": 5, "tid": 7, "ts": 1716454216790089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789847, "dur": 6,
    "args": {
      "External id": 11519,
      "cbid": 211, "correlation": 11519
    }
  },
  {
    "ph": "s", "id": 11519, "pid": 76337, "tid": -914061504, "ts": 1716454216789847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216790093, "dur": 6,
    "args": {
      "External id": 11529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11529, "pid": 5, "tid": 7, "ts": 1716454216790093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789910, "dur": 13,
    "args": {
      "External id": 11529,
      "cbid": 211, "correlation": 11529
    }
  },
  {
    "ph": "s", "id": 11529, "pid": 76337, "tid": -914061504, "ts": 1716454216789910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790100, "dur": 10,
    "args": {
      "External id": 11549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11549,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11549, "pid": 5, "tid": 7, "ts": 1716454216790100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216789987, "dur": 11,
    "args": {
      "External id": 11549,
      "cbid": 211, "correlation": 11549
    }
  },
  {
    "ph": "s", "id": 11549, "pid": 76337, "tid": -914061504, "ts": 1716454216789987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216790112, "dur": 4,
    "args": {
      "External id": 11561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11561,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11561, "pid": 5, "tid": 7, "ts": 1716454216790112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790009, "dur": 6,
    "args": {
      "External id": 11561,
      "cbid": 211, "correlation": 11561
    }
  },
  {
    "ph": "s", "id": 11561, "pid": 76337, "tid": -914061504, "ts": 1716454216790009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216790117, "dur": 7,
    "args": {
      "External id": 11564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11564, "pid": 5, "tid": 7, "ts": 1716454216790117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790027, "dur": 7,
    "args": {
      "External id": 11564,
      "cbid": 211, "correlation": 11564
    }
  },
  {
    "ph": "s", "id": 11564, "pid": 76337, "tid": -914061504, "ts": 1716454216790027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216790125, "dur": 5,
    "args": {
      "External id": 11573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11573,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11573, "pid": 5, "tid": 7, "ts": 1716454216790125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790068, "dur": 10,
    "args": {
      "External id": 11573,
      "cbid": 211, "correlation": 11573
    }
  },
  {
    "ph": "s", "id": 11573, "pid": 76337, "tid": -914061504, "ts": 1716454216790068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790149, "dur": 0,
    "args": {
      "External id": 11583,
      "cbid": 317, "correlation": 11583
    }
  },
  {
    "ph": "f", "id": 11583, "pid": 76337, "tid": -914061504, "ts": 1716454216790149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790150, "dur": 0,
    "args": {
      "External id": 11584,
      "cbid": 203, "correlation": 11584
    }
  },
  {
    "ph": "f", "id": 11584, "pid": 76337, "tid": -914061504, "ts": 1716454216790150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790151, "dur": 0,
    "args": {
      "External id": 11585,
      "cbid": 205, "correlation": 11585
    }
  },
  {
    "ph": "f", "id": 11585, "pid": 76337, "tid": -914061504, "ts": 1716454216790151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790178, "dur": 5,
    "args": {
      "External id": 11589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11589,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11589, "pid": 5, "tid": 7, "ts": 1716454216790178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790165, "dur": 13,
    "args": {
      "External id": 11589,
      "cbid": 211, "correlation": 11589
    }
  },
  {
    "ph": "s", "id": 11589, "pid": 76337, "tid": -914061504, "ts": 1716454216790165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790188, "dur": 163,
    "args": {
      "External id": 11591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11591, "pid": 5, "tid": 7, "ts": 1716454216790188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790180, "dur": 6,
    "args": {
      "External id": 11591,
      "cbid": 211, "correlation": 11591
    }
  },
  {
    "ph": "s", "id": 11591, "pid": 76337, "tid": -914061504, "ts": 1716454216790180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216790354, "dur": 1,
    "args": {
      "External id": 11593,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11593,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 11593, "pid": 5, "tid": 7, "ts": 1716454216790354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790191, "dur": 7,
    "args": {
      "External id": 11593,
      "cbid": 51, "correlation": 11593
    }
  },
  {
    "ph": "s", "id": 11593, "pid": 76337, "tid": -914061504, "ts": 1716454216790191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216790357, "dur": 260,
    "args": {
      "External id": 11594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11594,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11594, "pid": 5, "tid": 7, "ts": 1716454216790357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790199, "dur": 6,
    "args": {
      "External id": 11594,
      "cbid": 211, "correlation": 11594
    }
  },
  {
    "ph": "s", "id": 11594, "pid": 76337, "tid": -914061504, "ts": 1716454216790199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790619, "dur": 6,
    "args": {
      "External id": 11596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11596, "pid": 5, "tid": 7, "ts": 1716454216790619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790209, "dur": 6,
    "args": {
      "External id": 11596,
      "cbid": 211, "correlation": 11596
    }
  },
  {
    "ph": "s", "id": 11596, "pid": 76337, "tid": -914061504, "ts": 1716454216790209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216790627, "dur": 7,
    "args": {
      "External id": 11602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11602, "pid": 5, "tid": 7, "ts": 1716454216790627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790239, "dur": 8,
    "args": {
      "External id": 11602,
      "cbid": 211, "correlation": 11602
    }
  },
  {
    "ph": "s", "id": 11602, "pid": 76337, "tid": -914061504, "ts": 1716454216790239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216790634, "dur": 5,
    "args": {
      "External id": 11610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11610,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11610, "pid": 5, "tid": 7, "ts": 1716454216790634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790272, "dur": 10,
    "args": {
      "External id": 11610,
      "cbid": 211, "correlation": 11610
    }
  },
  {
    "ph": "s", "id": 11610, "pid": 76337, "tid": -914061504, "ts": 1716454216790272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216790641, "dur": 4,
    "args": {
      "External id": 11618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11618,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11618, "pid": 5, "tid": 7, "ts": 1716454216790641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790304, "dur": 8,
    "args": {
      "External id": 11618,
      "cbid": 211, "correlation": 11618
    }
  },
  {
    "ph": "s", "id": 11618, "pid": 76337, "tid": -914061504, "ts": 1716454216790304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790647, "dur": 10,
    "args": {
      "External id": 11638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11638,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11638, "pid": 5, "tid": 7, "ts": 1716454216790647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790470, "dur": 14,
    "args": {
      "External id": 11638,
      "cbid": 211, "correlation": 11638
    }
  },
  {
    "ph": "s", "id": 11638, "pid": 76337, "tid": -914061504, "ts": 1716454216790470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216790658, "dur": 4,
    "args": {
      "External id": 11650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11650,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11650, "pid": 5, "tid": 7, "ts": 1716454216790658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790494, "dur": 7,
    "args": {
      "External id": 11650,
      "cbid": 211, "correlation": 11650
    }
  },
  {
    "ph": "s", "id": 11650, "pid": 76337, "tid": -914061504, "ts": 1716454216790494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216790663, "dur": 7,
    "args": {
      "External id": 11653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11653,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11653, "pid": 5, "tid": 7, "ts": 1716454216790663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790513, "dur": 7,
    "args": {
      "External id": 11653,
      "cbid": 211, "correlation": 11653
    }
  },
  {
    "ph": "s", "id": 11653, "pid": 76337, "tid": -914061504, "ts": 1716454216790513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216790671, "dur": 5,
    "args": {
      "External id": 11662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11662,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11662, "pid": 5, "tid": 7, "ts": 1716454216790671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790552, "dur": 10,
    "args": {
      "External id": 11662,
      "cbid": 211, "correlation": 11662
    }
  },
  {
    "ph": "s", "id": 11662, "pid": 76337, "tid": -914061504, "ts": 1716454216790552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790618, "dur": 0,
    "args": {
      "External id": 11672,
      "cbid": 317, "correlation": 11672
    }
  },
  {
    "ph": "f", "id": 11672, "pid": 76337, "tid": -914061504, "ts": 1716454216790618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790618, "dur": 0,
    "args": {
      "External id": 11673,
      "cbid": 203, "correlation": 11673
    }
  },
  {
    "ph": "f", "id": 11673, "pid": 76337, "tid": -914061504, "ts": 1716454216790618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790619, "dur": 0,
    "args": {
      "External id": 11674,
      "cbid": 205, "correlation": 11674
    }
  },
  {
    "ph": "f", "id": 11674, "pid": 76337, "tid": -914061504, "ts": 1716454216790619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790678, "dur": 5,
    "args": {
      "External id": 11678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11678, "pid": 5, "tid": 7, "ts": 1716454216790678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790643, "dur": 14,
    "args": {
      "External id": 11678,
      "cbid": 211, "correlation": 11678
    }
  },
  {
    "ph": "s", "id": 11678, "pid": 76337, "tid": -914061504, "ts": 1716454216790643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216790685, "dur": 163,
    "args": {
      "External id": 11680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11680,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11680, "pid": 5, "tid": 7, "ts": 1716454216790685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790659, "dur": 5,
    "args": {
      "External id": 11680,
      "cbid": 211, "correlation": 11680
    }
  },
  {
    "ph": "s", "id": 11680, "pid": 76337, "tid": -914061504, "ts": 1716454216790659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216790850, "dur": 1,
    "args": {
      "External id": 11682,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11682,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 11682, "pid": 5, "tid": 7, "ts": 1716454216790850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790670, "dur": 7,
    "args": {
      "External id": 11682,
      "cbid": 51, "correlation": 11682
    }
  },
  {
    "ph": "s", "id": 11682, "pid": 76337, "tid": -914061504, "ts": 1716454216790670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216790854, "dur": 261,
    "args": {
      "External id": 11683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11683,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11683, "pid": 5, "tid": 7, "ts": 1716454216790854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790677, "dur": 6,
    "args": {
      "External id": 11683,
      "cbid": 211, "correlation": 11683
    }
  },
  {
    "ph": "s", "id": 11683, "pid": 76337, "tid": -914061504, "ts": 1716454216790677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216791116, "dur": 6,
    "args": {
      "External id": 11685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11685,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11685, "pid": 5, "tid": 7, "ts": 1716454216791116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790687, "dur": 5,
    "args": {
      "External id": 11685,
      "cbid": 211, "correlation": 11685
    }
  },
  {
    "ph": "s", "id": 11685, "pid": 76337, "tid": -914061504, "ts": 1716454216790687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216791123, "dur": 6,
    "args": {
      "External id": 11691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11691, "pid": 5, "tid": 7, "ts": 1716454216791123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790717, "dur": 9,
    "args": {
      "External id": 11691,
      "cbid": 211, "correlation": 11691
    }
  },
  {
    "ph": "s", "id": 11691, "pid": 76337, "tid": -914061504, "ts": 1716454216790717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216791130, "dur": 3,
    "args": {
      "External id": 11699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11699,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 11699, "pid": 5, "tid": 7, "ts": 1716454216791130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790763, "dur": 10,
    "args": {
      "External id": 11699,
      "cbid": 211, "correlation": 11699
    }
  },
  {
    "ph": "s", "id": 11699, "pid": 76337, "tid": -914061504, "ts": 1716454216790763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790839, "dur": 1,
    "args": {
      "External id": 11715,
      "cbid": 251, "correlation": 11715
    }
  },
  {
    "ph": "f", "id": 11715, "pid": 76337, "tid": -914061504, "ts": 1716454216790839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790844, "dur": 0,
    "args": {
      "External id": 11717,
      "cbid": 251, "correlation": 11717
    }
  },
  {
    "ph": "f", "id": 11717, "pid": 76337, "tid": -914061504, "ts": 1716454216790844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216791135, "dur": 10,
    "args": {
      "External id": 11718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11718,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11718, "pid": 5, "tid": 7, "ts": 1716454216791135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790846, "dur": 13,
    "args": {
      "External id": 11718,
      "cbid": 211, "correlation": 11718
    }
  },
  {
    "ph": "s", "id": 11718, "pid": 76337, "tid": -914061504, "ts": 1716454216790846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216791147, "dur": 4,
    "args": {
      "External id": 11720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11720,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11720, "pid": 5, "tid": 7, "ts": 1716454216791147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790861, "dur": 6,
    "args": {
      "External id": 11720,
      "cbid": 211, "correlation": 11720
    }
  },
  {
    "ph": "s", "id": 11720, "pid": 76337, "tid": -914061504, "ts": 1716454216790861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216791151, "dur": 6,
    "args": {
      "External id": 11730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11730, "pid": 5, "tid": 7, "ts": 1716454216791151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790920, "dur": 12,
    "args": {
      "External id": 11730,
      "cbid": 211, "correlation": 11730
    }
  },
  {
    "ph": "s", "id": 11730, "pid": 76337, "tid": -914061504, "ts": 1716454216790920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216791158, "dur": 10,
    "args": {
      "External id": 11750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11750,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11750, "pid": 5, "tid": 7, "ts": 1716454216791158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216790995, "dur": 11,
    "args": {
      "External id": 11750,
      "cbid": 211, "correlation": 11750
    }
  },
  {
    "ph": "s", "id": 11750, "pid": 76337, "tid": -914061504, "ts": 1716454216790995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216791170, "dur": 4,
    "args": {
      "External id": 11762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11762,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11762, "pid": 5, "tid": 7, "ts": 1716454216791170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791027, "dur": 7,
    "args": {
      "External id": 11762,
      "cbid": 211, "correlation": 11762
    }
  },
  {
    "ph": "s", "id": 11762, "pid": 76337, "tid": -914061504, "ts": 1716454216791027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216791175, "dur": 7,
    "args": {
      "External id": 11765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11765, "pid": 5, "tid": 7, "ts": 1716454216791175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791047, "dur": 6,
    "args": {
      "External id": 11765,
      "cbid": 211, "correlation": 11765
    }
  },
  {
    "ph": "s", "id": 11765, "pid": 76337, "tid": -914061504, "ts": 1716454216791047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216791183, "dur": 5,
    "args": {
      "External id": 11774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11774,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11774, "pid": 5, "tid": 7, "ts": 1716454216791183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791090, "dur": 10,
    "args": {
      "External id": 11774,
      "cbid": 211, "correlation": 11774
    }
  },
  {
    "ph": "s", "id": 11774, "pid": 76337, "tid": -914061504, "ts": 1716454216791090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791154, "dur": 0,
    "args": {
      "External id": 11784,
      "cbid": 317, "correlation": 11784
    }
  },
  {
    "ph": "f", "id": 11784, "pid": 76337, "tid": -914061504, "ts": 1716454216791154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791155, "dur": 0,
    "args": {
      "External id": 11785,
      "cbid": 203, "correlation": 11785
    }
  },
  {
    "ph": "f", "id": 11785, "pid": 76337, "tid": -914061504, "ts": 1716454216791155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791155, "dur": 0,
    "args": {
      "External id": 11786,
      "cbid": 205, "correlation": 11786
    }
  },
  {
    "ph": "f", "id": 11786, "pid": 76337, "tid": -914061504, "ts": 1716454216791155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216791189, "dur": 5,
    "args": {
      "External id": 11790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11790, "pid": 5, "tid": 7, "ts": 1716454216791189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791169, "dur": 12,
    "args": {
      "External id": 11790,
      "cbid": 211, "correlation": 11790
    }
  },
  {
    "ph": "s", "id": 11790, "pid": 76337, "tid": -914061504, "ts": 1716454216791169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216791196, "dur": 164,
    "args": {
      "External id": 11792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11792,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11792, "pid": 5, "tid": 7, "ts": 1716454216791196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791184, "dur": 5,
    "args": {
      "External id": 11792,
      "cbid": 211, "correlation": 11792
    }
  },
  {
    "ph": "s", "id": 11792, "pid": 76337, "tid": -914061504, "ts": 1716454216791184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216791362, "dur": 1,
    "args": {
      "External id": 11794,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11794,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14423076923076922
    }
  },
  {
    "ph": "f", "id": 11794, "pid": 5, "tid": 7, "ts": 1716454216791362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791195, "dur": 7,
    "args": {
      "External id": 11794,
      "cbid": 51, "correlation": 11794
    }
  },
  {
    "ph": "s", "id": 11794, "pid": 76337, "tid": -914061504, "ts": 1716454216791195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216791366, "dur": 261,
    "args": {
      "External id": 11795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11795,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 11795, "pid": 5, "tid": 7, "ts": 1716454216791366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791202, "dur": 6,
    "args": {
      "External id": 11795,
      "cbid": 211, "correlation": 11795
    }
  },
  {
    "ph": "s", "id": 11795, "pid": 76337, "tid": -914061504, "ts": 1716454216791202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216791628, "dur": 6,
    "args": {
      "External id": 11797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11797,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11797, "pid": 5, "tid": 7, "ts": 1716454216791628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791212, "dur": 5,
    "args": {
      "External id": 11797,
      "cbid": 211, "correlation": 11797
    }
  },
  {
    "ph": "s", "id": 11797, "pid": 76337, "tid": -914061504, "ts": 1716454216791212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216791635, "dur": 6,
    "args": {
      "External id": 11803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11803,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11803, "pid": 5, "tid": 7, "ts": 1716454216791635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791241, "dur": 8,
    "args": {
      "External id": 11803,
      "cbid": 211, "correlation": 11803
    }
  },
  {
    "ph": "s", "id": 11803, "pid": 76337, "tid": -914061504, "ts": 1716454216791241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216791643, "dur": 5,
    "args": {
      "External id": 11811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11811,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11811, "pid": 5, "tid": 7, "ts": 1716454216791643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791274, "dur": 8,
    "args": {
      "External id": 11811,
      "cbid": 211, "correlation": 11811
    }
  },
  {
    "ph": "s", "id": 11811, "pid": 76337, "tid": -914061504, "ts": 1716454216791274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791301, "dur": 1,
    "args": {
      "External id": 11819,
      "cbid": 317, "correlation": 11819
    }
  },
  {
    "ph": "f", "id": 11819, "pid": 76337, "tid": -914061504, "ts": 1716454216791301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216791302, "dur": 1231,
    "args": {
      "External id": 11820,
      "cbid": 20, "correlation": 11820
    }
  },
  {
    "ph": "f", "id": 11820, "pid": 76337, "tid": -914061504, "ts": 1716454216791302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216792562, "dur": 5,
    "args": {
      "External id": 11823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11823,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11823, "pid": 5, "tid": 7, "ts": 1716454216792562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792545, "dur": 16,
    "args": {
      "External id": 11823,
      "cbid": 211, "correlation": 11823
    }
  },
  {
    "ph": "s", "id": 11823, "pid": 76337, "tid": -914061504, "ts": 1716454216792545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216792736, "dur": 10,
    "args": {
      "External id": 11843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11843,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 11843, "pid": 5, "tid": 7, "ts": 1716454216792736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792721, "dur": 14,
    "args": {
      "External id": 11843,
      "cbid": 211, "correlation": 11843
    }
  },
  {
    "ph": "s", "id": 11843, "pid": 76337, "tid": -914061504, "ts": 1716454216792721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216792754, "dur": 4,
    "args": {
      "External id": 11855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11855,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 11855, "pid": 5, "tid": 7, "ts": 1716454216792754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792745, "dur": 7,
    "args": {
      "External id": 11855,
      "cbid": 211, "correlation": 11855
    }
  },
  {
    "ph": "s", "id": 11855, "pid": 76337, "tid": -914061504, "ts": 1716454216792745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216792773, "dur": 6,
    "args": {
      "External id": 11858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11858, "pid": 5, "tid": 7, "ts": 1716454216792773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792765, "dur": 6,
    "args": {
      "External id": 11858,
      "cbid": 211, "correlation": 11858
    }
  },
  {
    "ph": "s", "id": 11858, "pid": 76337, "tid": -914061504, "ts": 1716454216792765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792840, "dur": 0,
    "args": {
      "External id": 11869,
      "cbid": 317, "correlation": 11869
    }
  },
  {
    "ph": "f", "id": 11869, "pid": 76337, "tid": -914061504, "ts": 1716454216792840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792841, "dur": 0,
    "args": {
      "External id": 11870,
      "cbid": 203, "correlation": 11870
    }
  },
  {
    "ph": "f", "id": 11870, "pid": 76337, "tid": -914061504, "ts": 1716454216792841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216792842, "dur": 0,
    "args": {
      "External id": 11871,
      "cbid": 205, "correlation": 11871
    }
  },
  {
    "ph": "f", "id": 11871, "pid": 76337, "tid": -914061504, "ts": 1716454216792842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216794800, "dur": 5,
    "args": {
      "External id": 11875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11875, "pid": 5, "tid": 7, "ts": 1716454216794800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216794774, "dur": 26,
    "args": {
      "External id": 11875,
      "cbid": 211, "correlation": 11875
    }
  },
  {
    "ph": "s", "id": 11875, "pid": 76337, "tid": -914061504, "ts": 1716454216794774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216794935, "dur": 35,
    "args": {
      "External id": 11877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11877,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 11877, "pid": 5, "tid": 7, "ts": 1716454216794935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216794819, "dur": 118,
    "args": {
      "External id": 11877,
      "cbid": 211, "correlation": 11877
    }
  },
  {
    "ph": "s", "id": 11877, "pid": 76337, "tid": -914061504, "ts": 1716454216794819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216794971, "dur": 6,
    "args": {
      "External id": 11879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11879,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 11879, "pid": 5, "tid": 7, "ts": 1716454216794971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216794944, "dur": 8,
    "args": {
      "External id": 11879,
      "cbid": 211, "correlation": 11879
    }
  },
  {
    "ph": "s", "id": 11879, "pid": 76337, "tid": -914061504, "ts": 1716454216794944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216795477, "dur": 8,
    "args": {
      "External id": 11885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11885,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11885, "pid": 5, "tid": 7, "ts": 1716454216795477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216795465, "dur": 12,
    "args": {
      "External id": 11885,
      "cbid": 211, "correlation": 11885
    }
  },
  {
    "ph": "s", "id": 11885, "pid": 76337, "tid": -914061504, "ts": 1716454216795465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216795675, "dur": 12,
    "args": {
      "External id": 11896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11896,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 11896, "pid": 5, "tid": 7, "ts": 1716454216795675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216795657, "dur": 18,
    "args": {
      "External id": 11896,
      "cbid": 211, "correlation": 11896
    }
  },
  {
    "ph": "s", "id": 11896, "pid": 76337, "tid": -914061504, "ts": 1716454216795657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216795713, "dur": 10,
    "args": {
      "External id": 11918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11918,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 11918, "pid": 5, "tid": 7, "ts": 1716454216795713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216795702, "dur": 10,
    "args": {
      "External id": 11918,
      "cbid": 211, "correlation": 11918
    }
  },
  {
    "ph": "s", "id": 11918, "pid": 76337, "tid": -914061504, "ts": 1716454216795702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216795869, "dur": 58,
    "args": {
      "External id": 11929,
      "cbid": 251, "correlation": 11929
    }
  },
  {
    "ph": "f", "id": 11929, "pid": 76337, "tid": -914061504, "ts": 1716454216795869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216795963, "dur": 1,
    "args": {
      "External id": 11930,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11930,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 11930, "pid": 5, "tid": 7, "ts": 1716454216795963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216795934, "dur": 27,
    "args": {
      "External id": 11930,
      "cbid": 51, "correlation": 11930
    }
  },
  {
    "ph": "s", "id": 11930, "pid": 76337, "tid": -914061504, "ts": 1716454216795934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216795988, "dur": 35,
    "args": {
      "External id": 11931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11931,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 11931, "pid": 5, "tid": 7, "ts": 1716454216795988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216795964, "dur": 25,
    "args": {
      "External id": 11931,
      "cbid": 211, "correlation": 11931
    }
  },
  {
    "ph": "s", "id": 11931, "pid": 76337, "tid": -914061504, "ts": 1716454216795964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216796047, "dur": 2,
    "args": {
      "External id": 11939,
      "cbid": 317, "correlation": 11939
    }
  },
  {
    "ph": "f", "id": 11939, "pid": 76337, "tid": -914061504, "ts": 1716454216796047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216796050, "dur": 12343,
    "args": {
      "External id": 11940,
      "cbid": 20, "correlation": 11940
    }
  },
  {
    "ph": "f", "id": 11940, "pid": 76337, "tid": -914061504, "ts": 1716454216796050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808426, "dur": 1,
    "args": {
      "External id": 11946,
      "cbid": 251, "correlation": 11946
    }
  },
  {
    "ph": "f", "id": 11946, "pid": 76337, "tid": -914061504, "ts": 1716454216808426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216808443, "dur": 1,
    "args": {
      "External id": 11947,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11947,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 11947, "pid": 5, "tid": 7, "ts": 1716454216808443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808431, "dur": 9,
    "args": {
      "External id": 11947,
      "cbid": 51, "correlation": 11947
    }
  },
  {
    "ph": "s", "id": 11947, "pid": 76337, "tid": -914061504, "ts": 1716454216808431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216808456, "dur": 36,
    "args": {
      "External id": 11948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11948,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 11948, "pid": 5, "tid": 7, "ts": 1716454216808456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808442, "dur": 13,
    "args": {
      "External id": 11948,
      "cbid": 211, "correlation": 11948
    }
  },
  {
    "ph": "s", "id": 11948, "pid": 76337, "tid": -914061504, "ts": 1716454216808442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808529, "dur": 1,
    "args": {
      "External id": 11959,
      "cbid": 251, "correlation": 11959
    }
  },
  {
    "ph": "f", "id": 11959, "pid": 76337, "tid": -914061504, "ts": 1716454216808529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216808543, "dur": 1,
    "args": {
      "External id": 11960,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 11960,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 11960, "pid": 5, "tid": 7, "ts": 1716454216808543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808533, "dur": 7,
    "args": {
      "External id": 11960,
      "cbid": 51, "correlation": 11960
    }
  },
  {
    "ph": "s", "id": 11960, "pid": 76337, "tid": -914061504, "ts": 1716454216808533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216808553, "dur": 35,
    "args": {
      "External id": 11961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11961,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 11961, "pid": 5, "tid": 7, "ts": 1716454216808553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808541, "dur": 11,
    "args": {
      "External id": 11961,
      "cbid": 211, "correlation": 11961
    }
  },
  {
    "ph": "s", "id": 11961, "pid": 76337, "tid": -914061504, "ts": 1716454216808541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808651, "dur": 1,
    "args": {
      "External id": 11976,
      "cbid": 317, "correlation": 11976
    }
  },
  {
    "ph": "f", "id": 11976, "pid": 76337, "tid": -914061504, "ts": 1716454216808651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216808653, "dur": 2117,
    "args": {
      "External id": 11977,
      "cbid": 20, "correlation": 11977
    }
  },
  {
    "ph": "f", "id": 11977, "pid": 76337, "tid": -914061504, "ts": 1716454216808653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216810812, "dur": 35,
    "args": {
      "External id": 11990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 11990,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 11990, "pid": 5, "tid": 7, "ts": 1716454216810812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216810792, "dur": 20,
    "args": {
      "External id": 11990,
      "cbid": 211, "correlation": 11990
    }
  },
  {
    "ph": "s", "id": 11990, "pid": 76337, "tid": -914061504, "ts": 1716454216810792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216810921, "dur": 1,
    "args": {
      "External id": 12005,
      "cbid": 317, "correlation": 12005
    }
  },
  {
    "ph": "f", "id": 12005, "pid": 76337, "tid": -914061504, "ts": 1716454216810921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454216810922, "dur": 826,
    "args": {
      "External id": 12006,
      "cbid": 20, "correlation": 12006
    }
  },
  {
    "ph": "f", "id": 12006, "pid": 76337, "tid": -914061504, "ts": 1716454216810922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216811785, "dur": 64,
    "args": {
      "External id": 12012,
      "cbid": 251, "correlation": 12012
    }
  },
  {
    "ph": "f", "id": 12012, "pid": 76337, "tid": -914061504, "ts": 1716454216811785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216811864, "dur": 2,
    "args": {
      "External id": 12014,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12014,
      "bytes": 480, "memory bandwidth (GB/s)": 0.22727272727272727
    }
  },
  {
    "ph": "f", "id": 12014, "pid": 5, "tid": 7, "ts": 1716454216811864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216811855, "dur": 7,
    "args": {
      "External id": 12014,
      "cbid": 51, "correlation": 12014
    }
  },
  {
    "ph": "s", "id": 12014, "pid": 76337, "tid": -914061504, "ts": 1716454216811855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216811876, "dur": 36,
    "args": {
      "External id": 12015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12015,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 12015, "pid": 5, "tid": 7, "ts": 1716454216811876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216811863, "dur": 14,
    "args": {
      "External id": 12015,
      "cbid": 211, "correlation": 12015
    }
  },
  {
    "ph": "s", "id": 12015, "pid": 76337, "tid": -914061504, "ts": 1716454216811863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216812005, "dur": 5,
    "args": {
      "External id": 12023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12023,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12023, "pid": 5, "tid": 7, "ts": 1716454216812005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216811989, "dur": 17,
    "args": {
      "External id": 12023,
      "cbid": 211, "correlation": 12023
    }
  },
  {
    "ph": "s", "id": 12023, "pid": 76337, "tid": -914061504, "ts": 1716454216811989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216812056, "dur": 12,
    "args": {
      "External id": 12031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12031, "pid": 5, "tid": 7, "ts": 1716454216812056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812045, "dur": 11,
    "args": {
      "External id": 12031,
      "cbid": 211, "correlation": 12031
    }
  },
  {
    "ph": "s", "id": 12031, "pid": 76337, "tid": -914061504, "ts": 1716454216812045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216812117, "dur": 10,
    "args": {
      "External id": 12053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12053,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 12053, "pid": 5, "tid": 7, "ts": 1716454216812117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812105, "dur": 12,
    "args": {
      "External id": 12053,
      "cbid": 211, "correlation": 12053
    }
  },
  {
    "ph": "s", "id": 12053, "pid": 76337, "tid": -914061504, "ts": 1716454216812105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812198, "dur": 1,
    "args": {
      "External id": 12064,
      "cbid": 251, "correlation": 12064
    }
  },
  {
    "ph": "f", "id": 12064, "pid": 76337, "tid": -914061504, "ts": 1716454216812198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216812214, "dur": 1,
    "args": {
      "External id": 12065,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12065,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3
    }
  },
  {
    "ph": "f", "id": 12065, "pid": 5, "tid": 7, "ts": 1716454216812214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812203, "dur": 8,
    "args": {
      "External id": 12065,
      "cbid": 51, "correlation": 12065
    }
  },
  {
    "ph": "s", "id": 12065, "pid": 76337, "tid": -914061504, "ts": 1716454216812203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216812225, "dur": 36,
    "args": {
      "External id": 12066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12066,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 12066, "pid": 5, "tid": 7, "ts": 1716454216812225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812213, "dur": 12,
    "args": {
      "External id": 12066,
      "cbid": 211, "correlation": 12066
    }
  },
  {
    "ph": "s", "id": 12066, "pid": 76337, "tid": -914061504, "ts": 1716454216812213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812296, "dur": 1,
    "args": {
      "External id": 12077,
      "cbid": 251, "correlation": 12077
    }
  },
  {
    "ph": "f", "id": 12077, "pid": 76337, "tid": -914061504, "ts": 1716454216812296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812300, "dur": 0,
    "args": {
      "External id": 12078,
      "cbid": 251, "correlation": 12078
    }
  },
  {
    "ph": "f", "id": 12078, "pid": 76337, "tid": -914061504, "ts": 1716454216812300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216812317, "dur": 12,
    "args": {
      "External id": 12079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12079,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12079, "pid": 5, "tid": 7, "ts": 1716454216812317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812303, "dur": 14,
    "args": {
      "External id": 12079,
      "cbid": 211, "correlation": 12079
    }
  },
  {
    "ph": "s", "id": 12079, "pid": 76337, "tid": -914061504, "ts": 1716454216812303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216812332, "dur": 5,
    "args": {
      "External id": 12081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12081,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12081, "pid": 5, "tid": 7, "ts": 1716454216812332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812320, "dur": 10,
    "args": {
      "External id": 12081,
      "cbid": 211, "correlation": 12081
    }
  },
  {
    "ph": "s", "id": 12081, "pid": 76337, "tid": -914061504, "ts": 1716454216812320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812383, "dur": 1,
    "args": {
      "External id": 12092,
      "cbid": 251, "correlation": 12092
    }
  },
  {
    "ph": "f", "id": 12092, "pid": 76337, "tid": -914061504, "ts": 1716454216812383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812387, "dur": 0,
    "args": {
      "External id": 12093,
      "cbid": 251, "correlation": 12093
    }
  },
  {
    "ph": "f", "id": 12093, "pid": 76337, "tid": -914061504, "ts": 1716454216812387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216812401, "dur": 9,
    "args": {
      "External id": 12094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12094,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12094, "pid": 5, "tid": 7, "ts": 1716454216812401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812388, "dur": 11,
    "args": {
      "External id": 12094,
      "cbid": 211, "correlation": 12094
    }
  },
  {
    "ph": "s", "id": 12094, "pid": 76337, "tid": -914061504, "ts": 1716454216812388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216812411, "dur": 3,
    "args": {
      "External id": 12096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12096, "pid": 5, "tid": 7, "ts": 1716454216812411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812401, "dur": 6,
    "args": {
      "External id": 12096,
      "cbid": 211, "correlation": 12096
    }
  },
  {
    "ph": "s", "id": 12096, "pid": 76337, "tid": -914061504, "ts": 1716454216812401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216812493, "dur": 32,
    "args": {
      "External id": 12121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12121,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 12121, "pid": 5, "tid": 7, "ts": 1716454216812493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812480, "dur": 13,
    "args": {
      "External id": 12121,
      "cbid": 211, "correlation": 12121
    }
  },
  {
    "ph": "s", "id": 12121, "pid": 76337, "tid": -914061504, "ts": 1716454216812480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812588, "dur": 1,
    "args": {
      "External id": 12139,
      "cbid": 251, "correlation": 12139
    }
  },
  {
    "ph": "f", "id": 12139, "pid": 76337, "tid": -914061504, "ts": 1716454216812588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216812605, "dur": 1,
    "args": {
      "External id": 12141,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12141,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3
    }
  },
  {
    "ph": "f", "id": 12141, "pid": 5, "tid": 7, "ts": 1716454216812605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812593, "dur": 8,
    "args": {
      "External id": 12141,
      "cbid": 51, "correlation": 12141
    }
  },
  {
    "ph": "s", "id": 12141, "pid": 76337, "tid": -914061504, "ts": 1716454216812593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216812616, "dur": 35,
    "args": {
      "External id": 12142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12142,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 12142, "pid": 5, "tid": 7, "ts": 1716454216812616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812604, "dur": 12,
    "args": {
      "External id": 12142,
      "cbid": 211, "correlation": 12142
    }
  },
  {
    "ph": "s", "id": 12142, "pid": 76337, "tid": -914061504, "ts": 1716454216812604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216812689, "dur": 4,
    "args": {
      "External id": 12150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12150,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12150, "pid": 5, "tid": 7, "ts": 1716454216812689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812676, "dur": 12,
    "args": {
      "External id": 12150,
      "cbid": 211, "correlation": 12150
    }
  },
  {
    "ph": "s", "id": 12150, "pid": 76337, "tid": -914061504, "ts": 1716454216812676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216812727, "dur": 6,
    "args": {
      "External id": 12158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12158,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12158, "pid": 5, "tid": 7, "ts": 1716454216812727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812717, "dur": 9,
    "args": {
      "External id": 12158,
      "cbid": 211, "correlation": 12158
    }
  },
  {
    "ph": "s", "id": 12158, "pid": 76337, "tid": -914061504, "ts": 1716454216812717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216812795, "dur": 9,
    "args": {
      "External id": 12180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12180,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 12180, "pid": 5, "tid": 7, "ts": 1716454216812795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812782, "dur": 12,
    "args": {
      "External id": 12180,
      "cbid": 211, "correlation": 12180
    }
  },
  {
    "ph": "s", "id": 12180, "pid": 76337, "tid": -914061504, "ts": 1716454216812782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812905, "dur": 2,
    "args": {
      "External id": 12196,
      "cbid": 251, "correlation": 12196
    }
  },
  {
    "ph": "f", "id": 12196, "pid": 76337, "tid": -914061504, "ts": 1716454216812905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812911, "dur": 0,
    "args": {
      "External id": 12198,
      "cbid": 251, "correlation": 12198
    }
  },
  {
    "ph": "f", "id": 12198, "pid": 76337, "tid": -914061504, "ts": 1716454216812911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216812931, "dur": 195,
    "args": {
      "External id": 12199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12199,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12199, "pid": 5, "tid": 7, "ts": 1716454216812931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216812914, "dur": 17,
    "args": {
      "External id": 12199,
      "cbid": 211, "correlation": 12199
    }
  },
  {
    "ph": "s", "id": 12199, "pid": 76337, "tid": -914061504, "ts": 1716454216812914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216813127, "dur": 21,
    "args": {
      "External id": 12207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12207,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12207, "pid": 5, "tid": 7, "ts": 1716454216813127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813028, "dur": 15,
    "args": {
      "External id": 12207,
      "cbid": 211, "correlation": 12207
    }
  },
  {
    "ph": "s", "id": 12207, "pid": 76337, "tid": -914061504, "ts": 1716454216813028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216813150, "dur": 21,
    "args": {
      "External id": 12215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12215,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12215, "pid": 5, "tid": 7, "ts": 1716454216813150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813071, "dur": 10,
    "args": {
      "External id": 12215,
      "cbid": 211, "correlation": 12215
    }
  },
  {
    "ph": "s", "id": 12215, "pid": 76337, "tid": -914061504, "ts": 1716454216813071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813156, "dur": 64,
    "args": {
      "External id": 12231,
      "cbid": 251, "correlation": 12231
    }
  },
  {
    "ph": "f", "id": 12231, "pid": 76337, "tid": -914061504, "ts": 1716454216813156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216813235, "dur": 1,
    "args": {
      "External id": 12233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12233,
      "bytes": 120, "memory bandwidth (GB/s)": 0.07653061224489796
    }
  },
  {
    "ph": "f", "id": 12233, "pid": 5, "tid": 7, "ts": 1716454216813235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813225, "dur": 7,
    "args": {
      "External id": 12233,
      "cbid": 51, "correlation": 12233
    }
  },
  {
    "ph": "s", "id": 12233, "pid": 76337, "tid": -914061504, "ts": 1716454216813225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216813259, "dur": 110,
    "args": {
      "External id": 12234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12234,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 12234, "pid": 5, "tid": 7, "ts": 1716454216813259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813234, "dur": 25,
    "args": {
      "External id": 12234,
      "cbid": 211, "correlation": 12234
    }
  },
  {
    "ph": "s", "id": 12234, "pid": 76337, "tid": -914061504, "ts": 1716454216813234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216813370, "dur": 5,
    "args": {
      "External id": 12242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12242,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12242, "pid": 5, "tid": 7, "ts": 1716454216813370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813296, "dur": 11,
    "args": {
      "External id": 12242,
      "cbid": 211, "correlation": 12242
    }
  },
  {
    "ph": "s", "id": 12242, "pid": 76337, "tid": -914061504, "ts": 1716454216813296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216813388, "dur": 10,
    "args": {
      "External id": 12253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12253, "pid": 5, "tid": 7, "ts": 1716454216813388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813374, "dur": 14,
    "args": {
      "External id": 12253,
      "cbid": 211, "correlation": 12253
    }
  },
  {
    "ph": "s", "id": 12253, "pid": 76337, "tid": -914061504, "ts": 1716454216813374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813484, "dur": 0,
    "args": {
      "External id": 12265,
      "cbid": 317, "correlation": 12265
    }
  },
  {
    "ph": "f", "id": 12265, "pid": 76337, "tid": -914061504, "ts": 1716454216813484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813486, "dur": 0,
    "args": {
      "External id": 12266,
      "cbid": 203, "correlation": 12266
    }
  },
  {
    "ph": "f", "id": 12266, "pid": 76337, "tid": -914061504, "ts": 1716454216813486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813487, "dur": 1,
    "args": {
      "External id": 12267,
      "cbid": 205, "correlation": 12267
    }
  },
  {
    "ph": "f", "id": 12267, "pid": 76337, "tid": -914061504, "ts": 1716454216813487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216813526, "dur": 6,
    "args": {
      "External id": 12271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12271, "pid": 5, "tid": 7, "ts": 1716454216813526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813511, "dur": 14,
    "args": {
      "External id": 12271,
      "cbid": 211, "correlation": 12271
    }
  },
  {
    "ph": "s", "id": 12271, "pid": 76337, "tid": -914061504, "ts": 1716454216813511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216813547, "dur": 37,
    "args": {
      "External id": 12273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12273,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 12273, "pid": 5, "tid": 7, "ts": 1716454216813547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813537, "dur": 8,
    "args": {
      "External id": 12273,
      "cbid": 211, "correlation": 12273
    }
  },
  {
    "ph": "s", "id": 12273, "pid": 76337, "tid": -914061504, "ts": 1716454216813537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216813585, "dur": 6,
    "args": {
      "External id": 12275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12275,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12275, "pid": 5, "tid": 7, "ts": 1716454216813585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813551, "dur": 8,
    "args": {
      "External id": 12275,
      "cbid": 211, "correlation": 12275
    }
  },
  {
    "ph": "s", "id": 12275, "pid": 76337, "tid": -914061504, "ts": 1716454216813551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216813592, "dur": 8,
    "args": {
      "External id": 12281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12281, "pid": 5, "tid": 7, "ts": 1716454216813592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813582, "dur": 9,
    "args": {
      "External id": 12281,
      "cbid": 211, "correlation": 12281
    }
  },
  {
    "ph": "s", "id": 12281, "pid": 76337, "tid": -914061504, "ts": 1716454216813582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216813626, "dur": 4,
    "args": {
      "External id": 12289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12289,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12289, "pid": 5, "tid": 7, "ts": 1716454216813626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813616, "dur": 9,
    "args": {
      "External id": 12289,
      "cbid": 211, "correlation": 12289
    }
  },
  {
    "ph": "s", "id": 12289, "pid": 76337, "tid": -914061504, "ts": 1716454216813616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216813727, "dur": 11,
    "args": {
      "External id": 12309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12309,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12309, "pid": 5, "tid": 7, "ts": 1716454216813727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813713, "dur": 13,
    "args": {
      "External id": 12309,
      "cbid": 211, "correlation": 12309
    }
  },
  {
    "ph": "s", "id": 12309, "pid": 76337, "tid": -914061504, "ts": 1716454216813713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216813746, "dur": 5,
    "args": {
      "External id": 12321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12321,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 12321, "pid": 5, "tid": 7, "ts": 1716454216813746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813738, "dur": 8,
    "args": {
      "External id": 12321,
      "cbid": 211, "correlation": 12321
    }
  },
  {
    "ph": "s", "id": 12321, "pid": 76337, "tid": -914061504, "ts": 1716454216813738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216813769, "dur": 9,
    "args": {
      "External id": 12324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12324, "pid": 5, "tid": 7, "ts": 1716454216813769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813760, "dur": 8,
    "args": {
      "External id": 12324,
      "cbid": 211, "correlation": 12324
    }
  },
  {
    "ph": "s", "id": 12324, "pid": 76337, "tid": -914061504, "ts": 1716454216813760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216813823, "dur": 5,
    "args": {
      "External id": 12333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12333,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12333, "pid": 5, "tid": 7, "ts": 1716454216813823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813812, "dur": 11,
    "args": {
      "External id": 12333,
      "cbid": 211, "correlation": 12333
    }
  },
  {
    "ph": "s", "id": 12333, "pid": 76337, "tid": -914061504, "ts": 1716454216813812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813867, "dur": 0,
    "args": {
      "External id": 12343,
      "cbid": 317, "correlation": 12343
    }
  },
  {
    "ph": "f", "id": 12343, "pid": 76337, "tid": -914061504, "ts": 1716454216813867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813868, "dur": 0,
    "args": {
      "External id": 12344,
      "cbid": 203, "correlation": 12344
    }
  },
  {
    "ph": "f", "id": 12344, "pid": 76337, "tid": -914061504, "ts": 1716454216813868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813869, "dur": 0,
    "args": {
      "External id": 12345,
      "cbid": 205, "correlation": 12345
    }
  },
  {
    "ph": "f", "id": 12345, "pid": 76337, "tid": -914061504, "ts": 1716454216813869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216813897, "dur": 5,
    "args": {
      "External id": 12349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12349, "pid": 5, "tid": 7, "ts": 1716454216813897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813885, "dur": 11,
    "args": {
      "External id": 12349,
      "cbid": 211, "correlation": 12349
    }
  },
  {
    "ph": "s", "id": 12349, "pid": 76337, "tid": -914061504, "ts": 1716454216813885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216813907, "dur": 163,
    "args": {
      "External id": 12351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12351,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12351, "pid": 5, "tid": 7, "ts": 1716454216813907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813899, "dur": 6,
    "args": {
      "External id": 12351,
      "cbid": 211, "correlation": 12351
    }
  },
  {
    "ph": "s", "id": 12351, "pid": 76337, "tid": -914061504, "ts": 1716454216813899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216814072, "dur": 1,
    "args": {
      "External id": 12353,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12353,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 12353, "pid": 5, "tid": 7, "ts": 1716454216814072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813912, "dur": 8,
    "args": {
      "External id": 12353,
      "cbid": 51, "correlation": 12353
    }
  },
  {
    "ph": "s", "id": 12353, "pid": 76337, "tid": -914061504, "ts": 1716454216813912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216814076, "dur": 272,
    "args": {
      "External id": 12354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12354,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12354, "pid": 5, "tid": 7, "ts": 1716454216814076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813922, "dur": 9,
    "args": {
      "External id": 12354,
      "cbid": 211, "correlation": 12354
    }
  },
  {
    "ph": "s", "id": 12354, "pid": 76337, "tid": -914061504, "ts": 1716454216813922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216814349, "dur": 6,
    "args": {
      "External id": 12356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12356, "pid": 5, "tid": 7, "ts": 1716454216814349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813935, "dur": 5,
    "args": {
      "External id": 12356,
      "cbid": 211, "correlation": 12356
    }
  },
  {
    "ph": "s", "id": 12356, "pid": 76337, "tid": -914061504, "ts": 1716454216813935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216814357, "dur": 7,
    "args": {
      "External id": 12362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12362, "pid": 5, "tid": 7, "ts": 1716454216814357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216813965, "dur": 18,
    "args": {
      "External id": 12362,
      "cbid": 211, "correlation": 12362
    }
  },
  {
    "ph": "s", "id": 12362, "pid": 76337, "tid": -914061504, "ts": 1716454216813965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216814365, "dur": 3,
    "args": {
      "External id": 12370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12370,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 12370, "pid": 5, "tid": 7, "ts": 1716454216814365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814019, "dur": 10,
    "args": {
      "External id": 12370,
      "cbid": 211, "correlation": 12370
    }
  },
  {
    "ph": "s", "id": 12370, "pid": 76337, "tid": -914061504, "ts": 1716454216814019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814088, "dur": 1,
    "args": {
      "External id": 12386,
      "cbid": 251, "correlation": 12386
    }
  },
  {
    "ph": "f", "id": 12386, "pid": 76337, "tid": -914061504, "ts": 1716454216814088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814094, "dur": 0,
    "args": {
      "External id": 12388,
      "cbid": 251, "correlation": 12388
    }
  },
  {
    "ph": "f", "id": 12388, "pid": 76337, "tid": -914061504, "ts": 1716454216814094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216814369, "dur": 13,
    "args": {
      "External id": 12389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12389,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12389, "pid": 5, "tid": 7, "ts": 1716454216814369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814096, "dur": 11,
    "args": {
      "External id": 12389,
      "cbid": 211, "correlation": 12389
    }
  },
  {
    "ph": "s", "id": 12389, "pid": 76337, "tid": -914061504, "ts": 1716454216814096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216814384, "dur": 5,
    "args": {
      "External id": 12391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12391,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12391, "pid": 5, "tid": 7, "ts": 1716454216814384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814110, "dur": 6,
    "args": {
      "External id": 12391,
      "cbid": 211, "correlation": 12391
    }
  },
  {
    "ph": "s", "id": 12391, "pid": 76337, "tid": -914061504, "ts": 1716454216814110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216814390, "dur": 6,
    "args": {
      "External id": 12401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12401, "pid": 5, "tid": 7, "ts": 1716454216814390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814178, "dur": 13,
    "args": {
      "External id": 12401,
      "cbid": 211, "correlation": 12401
    }
  },
  {
    "ph": "s", "id": 12401, "pid": 76337, "tid": -914061504, "ts": 1716454216814178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216814397, "dur": 10,
    "args": {
      "External id": 12421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12421,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12421, "pid": 5, "tid": 7, "ts": 1716454216814397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814246, "dur": 11,
    "args": {
      "External id": 12421,
      "cbid": 211, "correlation": 12421
    }
  },
  {
    "ph": "s", "id": 12421, "pid": 76337, "tid": -914061504, "ts": 1716454216814246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216814408, "dur": 4,
    "args": {
      "External id": 12433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12433,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 12433, "pid": 5, "tid": 7, "ts": 1716454216814408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814266, "dur": 6,
    "args": {
      "External id": 12433,
      "cbid": 211, "correlation": 12433
    }
  },
  {
    "ph": "s", "id": 12433, "pid": 76337, "tid": -914061504, "ts": 1716454216814266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216814414, "dur": 7,
    "args": {
      "External id": 12436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12436, "pid": 5, "tid": 7, "ts": 1716454216814414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814284, "dur": 6,
    "args": {
      "External id": 12436,
      "cbid": 211, "correlation": 12436
    }
  },
  {
    "ph": "s", "id": 12436, "pid": 76337, "tid": -914061504, "ts": 1716454216814284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216814422, "dur": 5,
    "args": {
      "External id": 12445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12445,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12445, "pid": 5, "tid": 7, "ts": 1716454216814422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814325, "dur": 10,
    "args": {
      "External id": 12445,
      "cbid": 211, "correlation": 12445
    }
  },
  {
    "ph": "s", "id": 12445, "pid": 76337, "tid": -914061504, "ts": 1716454216814325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814389, "dur": 0,
    "args": {
      "External id": 12455,
      "cbid": 317, "correlation": 12455
    }
  },
  {
    "ph": "f", "id": 12455, "pid": 76337, "tid": -914061504, "ts": 1716454216814389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814390, "dur": 0,
    "args": {
      "External id": 12456,
      "cbid": 203, "correlation": 12456
    }
  },
  {
    "ph": "f", "id": 12456, "pid": 76337, "tid": -914061504, "ts": 1716454216814390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814391, "dur": 0,
    "args": {
      "External id": 12457,
      "cbid": 205, "correlation": 12457
    }
  },
  {
    "ph": "f", "id": 12457, "pid": 76337, "tid": -914061504, "ts": 1716454216814391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216814428, "dur": 5,
    "args": {
      "External id": 12461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12461,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12461, "pid": 5, "tid": 7, "ts": 1716454216814428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814404, "dur": 12,
    "args": {
      "External id": 12461,
      "cbid": 211, "correlation": 12461
    }
  },
  {
    "ph": "s", "id": 12461, "pid": 76337, "tid": -914061504, "ts": 1716454216814404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216814435, "dur": 164,
    "args": {
      "External id": 12463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12463,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12463, "pid": 5, "tid": 7, "ts": 1716454216814435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814419, "dur": 6,
    "args": {
      "External id": 12463,
      "cbid": 211, "correlation": 12463
    }
  },
  {
    "ph": "s", "id": 12463, "pid": 76337, "tid": -914061504, "ts": 1716454216814419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216814601, "dur": 1,
    "args": {
      "External id": 12465,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12465,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 12465, "pid": 5, "tid": 7, "ts": 1716454216814601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814430, "dur": 7,
    "args": {
      "External id": 12465,
      "cbid": 51, "correlation": 12465
    }
  },
  {
    "ph": "s", "id": 12465, "pid": 76337, "tid": -914061504, "ts": 1716454216814430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216814604, "dur": 262,
    "args": {
      "External id": 12466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12466,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12466, "pid": 5, "tid": 7, "ts": 1716454216814604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814438, "dur": 6,
    "args": {
      "External id": 12466,
      "cbid": 211, "correlation": 12466
    }
  },
  {
    "ph": "s", "id": 12466, "pid": 76337, "tid": -914061504, "ts": 1716454216814438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216814867, "dur": 6,
    "args": {
      "External id": 12468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12468,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12468, "pid": 5, "tid": 7, "ts": 1716454216814867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814448, "dur": 5,
    "args": {
      "External id": 12468,
      "cbid": 211, "correlation": 12468
    }
  },
  {
    "ph": "s", "id": 12468, "pid": 76337, "tid": -914061504, "ts": 1716454216814448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216814875, "dur": 6,
    "args": {
      "External id": 12474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12474, "pid": 5, "tid": 7, "ts": 1716454216814875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814476, "dur": 9,
    "args": {
      "External id": 12474,
      "cbid": 211, "correlation": 12474
    }
  },
  {
    "ph": "s", "id": 12474, "pid": 76337, "tid": -914061504, "ts": 1716454216814476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216814882, "dur": 5,
    "args": {
      "External id": 12482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12482,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12482, "pid": 5, "tid": 7, "ts": 1716454216814882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814509, "dur": 8,
    "args": {
      "External id": 12482,
      "cbid": 211, "correlation": 12482
    }
  },
  {
    "ph": "s", "id": 12482, "pid": 76337, "tid": -914061504, "ts": 1716454216814509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216814888, "dur": 5,
    "args": {
      "External id": 12490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12490,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12490, "pid": 5, "tid": 7, "ts": 1716454216814888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814539, "dur": 8,
    "args": {
      "External id": 12490,
      "cbid": 211, "correlation": 12490
    }
  },
  {
    "ph": "s", "id": 12490, "pid": 76337, "tid": -914061504, "ts": 1716454216814539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216814894, "dur": 11,
    "args": {
      "External id": 12499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12499,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12499, "pid": 5, "tid": 7, "ts": 1716454216814894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814721, "dur": 22,
    "args": {
      "External id": 12499,
      "cbid": 211, "correlation": 12499
    }
  },
  {
    "ph": "s", "id": 12499, "pid": 76337, "tid": -914061504, "ts": 1716454216814721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216814907, "dur": 12,
    "args": {
      "External id": 12519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12519,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12519, "pid": 5, "tid": 7, "ts": 1716454216814907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814813, "dur": 13,
    "args": {
      "External id": 12519,
      "cbid": 211, "correlation": 12519
    }
  },
  {
    "ph": "s", "id": 12519, "pid": 76337, "tid": -914061504, "ts": 1716454216814813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216814921, "dur": 4,
    "args": {
      "External id": 12531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12531,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12531, "pid": 5, "tid": 7, "ts": 1716454216814921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814836, "dur": 6,
    "args": {
      "External id": 12531,
      "cbid": 211, "correlation": 12531
    }
  },
  {
    "ph": "s", "id": 12531, "pid": 76337, "tid": -914061504, "ts": 1716454216814836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216814926, "dur": 10,
    "args": {
      "External id": 12534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12534, "pid": 5, "tid": 7, "ts": 1716454216814926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814855, "dur": 6,
    "args": {
      "External id": 12534,
      "cbid": 211, "correlation": 12534
    }
  },
  {
    "ph": "s", "id": 12534, "pid": 76337, "tid": -914061504, "ts": 1716454216814855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216814938, "dur": 6,
    "args": {
      "External id": 12543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12543,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12543, "pid": 5, "tid": 7, "ts": 1716454216814938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814895, "dur": 9,
    "args": {
      "External id": 12543,
      "cbid": 211, "correlation": 12543
    }
  },
  {
    "ph": "s", "id": 12543, "pid": 76337, "tid": -914061504, "ts": 1716454216814895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814961, "dur": 0,
    "args": {
      "External id": 12553,
      "cbid": 317, "correlation": 12553
    }
  },
  {
    "ph": "f", "id": 12553, "pid": 76337, "tid": -914061504, "ts": 1716454216814961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814962, "dur": 0,
    "args": {
      "External id": 12554,
      "cbid": 203, "correlation": 12554
    }
  },
  {
    "ph": "f", "id": 12554, "pid": 76337, "tid": -914061504, "ts": 1716454216814962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216814963, "dur": 0,
    "args": {
      "External id": 12555,
      "cbid": 205, "correlation": 12555
    }
  },
  {
    "ph": "f", "id": 12555, "pid": 76337, "tid": -914061504, "ts": 1716454216814963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216818191, "dur": 7,
    "args": {
      "External id": 12559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12559, "pid": 5, "tid": 7, "ts": 1716454216818191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216818161, "dur": 31,
    "args": {
      "External id": 12559,
      "cbid": 211, "correlation": 12559
    }
  },
  {
    "ph": "s", "id": 12559, "pid": 76337, "tid": -914061504, "ts": 1716454216818161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216818203, "dur": 323,
    "args": {
      "External id": 12561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12561,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12561, "pid": 5, "tid": 7, "ts": 1716454216818203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216818195, "dur": 6,
    "args": {
      "External id": 12561,
      "cbid": 211, "correlation": 12561
    }
  },
  {
    "ph": "s", "id": 12561, "pid": 76337, "tid": -914061504, "ts": 1716454216818195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216818528, "dur": 1,
    "args": {
      "External id": 12563,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12563,
      "bytes": 240, "memory bandwidth (GB/s)": 0.125
    }
  },
  {
    "ph": "f", "id": 12563, "pid": 5, "tid": 7, "ts": 1716454216818528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216818209, "dur": 14,
    "args": {
      "External id": 12563,
      "cbid": 51, "correlation": 12563
    }
  },
  {
    "ph": "s", "id": 12563, "pid": 76337, "tid": -914061504, "ts": 1716454216818209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216818532, "dur": 500,
    "args": {
      "External id": 12564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12564,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12564, "pid": 5, "tid": 7, "ts": 1716454216818532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216818225, "dur": 9,
    "args": {
      "External id": 12564,
      "cbid": 211, "correlation": 12564
    }
  },
  {
    "ph": "s", "id": 12564, "pid": 76337, "tid": -914061504, "ts": 1716454216818225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216819033, "dur": 5,
    "args": {
      "External id": 12566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12566,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12566, "pid": 5, "tid": 7, "ts": 1716454216819033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216818239, "dur": 6,
    "args": {
      "External id": 12566,
      "cbid": 211, "correlation": 12566
    }
  },
  {
    "ph": "s", "id": 12566, "pid": 76337, "tid": -914061504, "ts": 1716454216818239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216819039, "dur": 6,
    "args": {
      "External id": 12572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12572,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12572, "pid": 5, "tid": 7, "ts": 1716454216819039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819002, "dur": 14,
    "args": {
      "External id": 12572,
      "cbid": 211, "correlation": 12572
    }
  },
  {
    "ph": "s", "id": 12572, "pid": 76337, "tid": -914061504, "ts": 1716454216819002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216819099, "dur": 3,
    "args": {
      "External id": 12580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12580,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 12580, "pid": 5, "tid": 7, "ts": 1716454216819099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819088, "dur": 11,
    "args": {
      "External id": 12580,
      "cbid": 211, "correlation": 12580
    }
  },
  {
    "ph": "s", "id": 12580, "pid": 76337, "tid": -914061504, "ts": 1716454216819088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819211, "dur": 4,
    "args": {
      "External id": 12596,
      "cbid": 251, "correlation": 12596
    }
  },
  {
    "ph": "f", "id": 12596, "pid": 76337, "tid": -914061504, "ts": 1716454216819211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819220, "dur": 0,
    "args": {
      "External id": 12598,
      "cbid": 251, "correlation": 12598
    }
  },
  {
    "ph": "f", "id": 12598, "pid": 76337, "tid": -914061504, "ts": 1716454216819220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216819238, "dur": 10,
    "args": {
      "External id": 12599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12599,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12599, "pid": 5, "tid": 7, "ts": 1716454216819238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819223, "dur": 16,
    "args": {
      "External id": 12599,
      "cbid": 211, "correlation": 12599
    }
  },
  {
    "ph": "s", "id": 12599, "pid": 76337, "tid": -914061504, "ts": 1716454216819223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216819253, "dur": 4,
    "args": {
      "External id": 12601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12601,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12601, "pid": 5, "tid": 7, "ts": 1716454216819253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819242, "dur": 9,
    "args": {
      "External id": 12601,
      "cbid": 211, "correlation": 12601
    }
  },
  {
    "ph": "s", "id": 12601, "pid": 76337, "tid": -914061504, "ts": 1716454216819242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216819340, "dur": 6,
    "args": {
      "External id": 12611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12611, "pid": 5, "tid": 7, "ts": 1716454216819340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819327, "dur": 12,
    "args": {
      "External id": 12611,
      "cbid": 211, "correlation": 12611
    }
  },
  {
    "ph": "s", "id": 12611, "pid": 76337, "tid": -914061504, "ts": 1716454216819327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216819432, "dur": 10,
    "args": {
      "External id": 12631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12631,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12631, "pid": 5, "tid": 7, "ts": 1716454216819432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819419, "dur": 12,
    "args": {
      "External id": 12631,
      "cbid": 211, "correlation": 12631
    }
  },
  {
    "ph": "s", "id": 12631, "pid": 76337, "tid": -914061504, "ts": 1716454216819419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216819451, "dur": 4,
    "args": {
      "External id": 12643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12643,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 12643, "pid": 5, "tid": 7, "ts": 1716454216819451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819443, "dur": 8,
    "args": {
      "External id": 12643,
      "cbid": 211, "correlation": 12643
    }
  },
  {
    "ph": "s", "id": 12643, "pid": 76337, "tid": -914061504, "ts": 1716454216819443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216819478, "dur": 7,
    "args": {
      "External id": 12646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12646, "pid": 5, "tid": 7, "ts": 1716454216819478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819470, "dur": 7,
    "args": {
      "External id": 12646,
      "cbid": 211, "correlation": 12646
    }
  },
  {
    "ph": "s", "id": 12646, "pid": 76337, "tid": -914061504, "ts": 1716454216819470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216819524, "dur": 5,
    "args": {
      "External id": 12655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12655,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12655, "pid": 5, "tid": 7, "ts": 1716454216819524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819513, "dur": 10,
    "args": {
      "External id": 12655,
      "cbid": 211, "correlation": 12655
    }
  },
  {
    "ph": "s", "id": 12655, "pid": 76337, "tid": -914061504, "ts": 1716454216819513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819600, "dur": 0,
    "args": {
      "External id": 12665,
      "cbid": 317, "correlation": 12665
    }
  },
  {
    "ph": "f", "id": 12665, "pid": 76337, "tid": -914061504, "ts": 1716454216819600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819601, "dur": 0,
    "args": {
      "External id": 12666,
      "cbid": 203, "correlation": 12666
    }
  },
  {
    "ph": "f", "id": 12666, "pid": 76337, "tid": -914061504, "ts": 1716454216819601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819603, "dur": 0,
    "args": {
      "External id": 12667,
      "cbid": 205, "correlation": 12667
    }
  },
  {
    "ph": "f", "id": 12667, "pid": 76337, "tid": -914061504, "ts": 1716454216819603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216819632, "dur": 5,
    "args": {
      "External id": 12671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12671, "pid": 5, "tid": 7, "ts": 1716454216819632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819620, "dur": 12,
    "args": {
      "External id": 12671,
      "cbid": 211, "correlation": 12671
    }
  },
  {
    "ph": "s", "id": 12671, "pid": 76337, "tid": -914061504, "ts": 1716454216819620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216819642, "dur": 164,
    "args": {
      "External id": 12673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12673,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12673, "pid": 5, "tid": 7, "ts": 1716454216819642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819634, "dur": 6,
    "args": {
      "External id": 12673,
      "cbid": 211, "correlation": 12673
    }
  },
  {
    "ph": "s", "id": 12673, "pid": 76337, "tid": -914061504, "ts": 1716454216819634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216819809, "dur": 1,
    "args": {
      "External id": 12675,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12675,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 12675, "pid": 5, "tid": 7, "ts": 1716454216819809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819646, "dur": 8,
    "args": {
      "External id": 12675,
      "cbid": 51, "correlation": 12675
    }
  },
  {
    "ph": "s", "id": 12675, "pid": 76337, "tid": -914061504, "ts": 1716454216819646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216819812, "dur": 261,
    "args": {
      "External id": 12676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12676,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12676, "pid": 5, "tid": 7, "ts": 1716454216819812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819656, "dur": 6,
    "args": {
      "External id": 12676,
      "cbid": 211, "correlation": 12676
    }
  },
  {
    "ph": "s", "id": 12676, "pid": 76337, "tid": -914061504, "ts": 1716454216819656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216820075, "dur": 6,
    "args": {
      "External id": 12678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12678, "pid": 5, "tid": 7, "ts": 1716454216820075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819666, "dur": 5,
    "args": {
      "External id": 12678,
      "cbid": 211, "correlation": 12678
    }
  },
  {
    "ph": "s", "id": 12678, "pid": 76337, "tid": -914061504, "ts": 1716454216819666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216820082, "dur": 6,
    "args": {
      "External id": 12684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12684, "pid": 5, "tid": 7, "ts": 1716454216820082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819695, "dur": 9,
    "args": {
      "External id": 12684,
      "cbid": 211, "correlation": 12684
    }
  },
  {
    "ph": "s", "id": 12684, "pid": 76337, "tid": -914061504, "ts": 1716454216819695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819756, "dur": 0,
    "args": {
      "External id": 12694,
      "cbid": 317, "correlation": 12694
    }
  },
  {
    "ph": "f", "id": 12694, "pid": 76337, "tid": -914061504, "ts": 1716454216819756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819757, "dur": 0,
    "args": {
      "External id": 12695,
      "cbid": 203, "correlation": 12695
    }
  },
  {
    "ph": "f", "id": 12695, "pid": 76337, "tid": -914061504, "ts": 1716454216819757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216819758, "dur": 0,
    "args": {
      "External id": 12696,
      "cbid": 205, "correlation": 12696
    }
  },
  {
    "ph": "f", "id": 12696, "pid": 76337, "tid": -914061504, "ts": 1716454216819758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216821688, "dur": 8,
    "args": {
      "External id": 12700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12700, "pid": 5, "tid": 7, "ts": 1716454216821688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216821666, "dur": 21,
    "args": {
      "External id": 12700,
      "cbid": 211, "correlation": 12700
    }
  },
  {
    "ph": "s", "id": 12700, "pid": 76337, "tid": -914061504, "ts": 1716454216821666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216821712, "dur": 3,
    "args": {
      "External id": 12702,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12702,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 12702, "pid": 5, "tid": 7, "ts": 1716454216821712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216821694, "dur": 16,
    "args": {
      "External id": 12702,
      "cbid": 51, "correlation": 12702
    }
  },
  {
    "ph": "s", "id": 12702, "pid": 76337, "tid": -914061504, "ts": 1716454216821694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216821722, "dur": 97,
    "args": {
      "External id": 12703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12703,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 12703, "pid": 5, "tid": 7, "ts": 1716454216821722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216821712, "dur": 10,
    "args": {
      "External id": 12703,
      "cbid": 211, "correlation": 12703
    }
  },
  {
    "ph": "s", "id": 12703, "pid": 76337, "tid": -914061504, "ts": 1716454216821712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216821821, "dur": 5,
    "args": {
      "External id": 12705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12705,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12705, "pid": 5, "tid": 7, "ts": 1716454216821821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216821727, "dur": 6,
    "args": {
      "External id": 12705,
      "cbid": 211, "correlation": 12705
    }
  },
  {
    "ph": "s", "id": 12705, "pid": 76337, "tid": -914061504, "ts": 1716454216821727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216822262, "dur": 6,
    "args": {
      "External id": 12711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12711, "pid": 5, "tid": 7, "ts": 1716454216822262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822249, "dur": 12,
    "args": {
      "External id": 12711,
      "cbid": 211, "correlation": 12711
    }
  },
  {
    "ph": "s", "id": 12711, "pid": 76337, "tid": -914061504, "ts": 1716454216822249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216822303, "dur": 5,
    "args": {
      "External id": 12719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12719,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12719, "pid": 5, "tid": 7, "ts": 1716454216822303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822292, "dur": 9,
    "args": {
      "External id": 12719,
      "cbid": 211, "correlation": 12719
    }
  },
  {
    "ph": "s", "id": 12719, "pid": 76337, "tid": -914061504, "ts": 1716454216822292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216822355, "dur": 4,
    "args": {
      "External id": 12727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12727,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12727, "pid": 5, "tid": 7, "ts": 1716454216822355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822344, "dur": 10,
    "args": {
      "External id": 12727,
      "cbid": 211, "correlation": 12727
    }
  },
  {
    "ph": "s", "id": 12727, "pid": 76337, "tid": -914061504, "ts": 1716454216822344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216822432, "dur": 11,
    "args": {
      "External id": 12736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12736,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12736, "pid": 5, "tid": 7, "ts": 1716454216822432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822419, "dur": 13,
    "args": {
      "External id": 12736,
      "cbid": 211, "correlation": 12736
    }
  },
  {
    "ph": "s", "id": 12736, "pid": 76337, "tid": -914061504, "ts": 1716454216822419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216822528, "dur": 12,
    "args": {
      "External id": 12756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12756,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12756, "pid": 5, "tid": 7, "ts": 1716454216822528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822515, "dur": 12,
    "args": {
      "External id": 12756,
      "cbid": 211, "correlation": 12756
    }
  },
  {
    "ph": "s", "id": 12756, "pid": 76337, "tid": -914061504, "ts": 1716454216822515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216822547, "dur": 4,
    "args": {
      "External id": 12768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12768,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12768, "pid": 5, "tid": 7, "ts": 1716454216822547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822538, "dur": 8,
    "args": {
      "External id": 12768,
      "cbid": 211, "correlation": 12768
    }
  },
  {
    "ph": "s", "id": 12768, "pid": 76337, "tid": -914061504, "ts": 1716454216822538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216822568, "dur": 10,
    "args": {
      "External id": 12771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12771, "pid": 5, "tid": 7, "ts": 1716454216822568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822560, "dur": 6,
    "args": {
      "External id": 12771,
      "cbid": 211, "correlation": 12771
    }
  },
  {
    "ph": "s", "id": 12771, "pid": 76337, "tid": -914061504, "ts": 1716454216822560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216822613, "dur": 6,
    "args": {
      "External id": 12780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12780,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12780, "pid": 5, "tid": 7, "ts": 1716454216822613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822601, "dur": 11,
    "args": {
      "External id": 12780,
      "cbid": 211, "correlation": 12780
    }
  },
  {
    "ph": "s", "id": 12780, "pid": 76337, "tid": -914061504, "ts": 1716454216822601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822679, "dur": 0,
    "args": {
      "External id": 12790,
      "cbid": 317, "correlation": 12790
    }
  },
  {
    "ph": "f", "id": 12790, "pid": 76337, "tid": -914061504, "ts": 1716454216822679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822679, "dur": 0,
    "args": {
      "External id": 12791,
      "cbid": 203, "correlation": 12791
    }
  },
  {
    "ph": "f", "id": 12791, "pid": 76337, "tid": -914061504, "ts": 1716454216822679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822680, "dur": 1,
    "args": {
      "External id": 12792,
      "cbid": 205, "correlation": 12792
    }
  },
  {
    "ph": "f", "id": 12792, "pid": 76337, "tid": -914061504, "ts": 1716454216822680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216822709, "dur": 7,
    "args": {
      "External id": 12796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12796, "pid": 5, "tid": 7, "ts": 1716454216822709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822696, "dur": 13,
    "args": {
      "External id": 12796,
      "cbid": 211, "correlation": 12796
    }
  },
  {
    "ph": "s", "id": 12796, "pid": 76337, "tid": -914061504, "ts": 1716454216822696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216822720, "dur": 323,
    "args": {
      "External id": 12798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12798,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12798, "pid": 5, "tid": 7, "ts": 1716454216822720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822711, "dur": 6,
    "args": {
      "External id": 12798,
      "cbid": 211, "correlation": 12798
    }
  },
  {
    "ph": "s", "id": 12798, "pid": 76337, "tid": -914061504, "ts": 1716454216822711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216823046, "dur": 1,
    "args": {
      "External id": 12800,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12800,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 12800, "pid": 5, "tid": 7, "ts": 1716454216823046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822724, "dur": 8,
    "args": {
      "External id": 12800,
      "cbid": 51, "correlation": 12800
    }
  },
  {
    "ph": "s", "id": 12800, "pid": 76337, "tid": -914061504, "ts": 1716454216822724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216823050, "dur": 503,
    "args": {
      "External id": 12801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12801,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12801, "pid": 5, "tid": 7, "ts": 1716454216823050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822733, "dur": 7,
    "args": {
      "External id": 12801,
      "cbid": 211, "correlation": 12801
    }
  },
  {
    "ph": "s", "id": 12801, "pid": 76337, "tid": -914061504, "ts": 1716454216822733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216823554, "dur": 6,
    "args": {
      "External id": 12803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12803,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12803, "pid": 5, "tid": 7, "ts": 1716454216823554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822743, "dur": 6,
    "args": {
      "External id": 12803,
      "cbid": 211, "correlation": 12803
    }
  },
  {
    "ph": "s", "id": 12803, "pid": 76337, "tid": -914061504, "ts": 1716454216822743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216823561, "dur": 6,
    "args": {
      "External id": 12809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12809, "pid": 5, "tid": 7, "ts": 1716454216823561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822773, "dur": 10,
    "args": {
      "External id": 12809,
      "cbid": 211, "correlation": 12809
    }
  },
  {
    "ph": "s", "id": 12809, "pid": 76337, "tid": -914061504, "ts": 1716454216822773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216823569, "dur": 3,
    "args": {
      "External id": 12817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12817,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 12817, "pid": 5, "tid": 7, "ts": 1716454216823569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822819, "dur": 10,
    "args": {
      "External id": 12817,
      "cbid": 211, "correlation": 12817
    }
  },
  {
    "ph": "s", "id": 12817, "pid": 76337, "tid": -914061504, "ts": 1716454216822819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822932, "dur": 2,
    "args": {
      "External id": 12833,
      "cbid": 251, "correlation": 12833
    }
  },
  {
    "ph": "f", "id": 12833, "pid": 76337, "tid": -914061504, "ts": 1716454216822932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822939, "dur": 0,
    "args": {
      "External id": 12835,
      "cbid": 251, "correlation": 12835
    }
  },
  {
    "ph": "f", "id": 12835, "pid": 76337, "tid": -914061504, "ts": 1716454216822939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216823573, "dur": 13,
    "args": {
      "External id": 12836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12836,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12836, "pid": 5, "tid": 7, "ts": 1716454216823573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822941, "dur": 15,
    "args": {
      "External id": 12836,
      "cbid": 211, "correlation": 12836
    }
  },
  {
    "ph": "s", "id": 12836, "pid": 76337, "tid": -914061504, "ts": 1716454216822941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216823587, "dur": 5,
    "args": {
      "External id": 12838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12838,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12838, "pid": 5, "tid": 7, "ts": 1716454216823587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216822959, "dur": 8,
    "args": {
      "External id": 12838,
      "cbid": 211, "correlation": 12838
    }
  },
  {
    "ph": "s", "id": 12838, "pid": 76337, "tid": -914061504, "ts": 1716454216822959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216823593, "dur": 6,
    "args": {
      "External id": 12848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12848, "pid": 5, "tid": 7, "ts": 1716454216823593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823036, "dur": 13,
    "args": {
      "External id": 12848,
      "cbid": 211, "correlation": 12848
    }
  },
  {
    "ph": "s", "id": 12848, "pid": 76337, "tid": -914061504, "ts": 1716454216823036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216823600, "dur": 10,
    "args": {
      "External id": 12868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12868,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12868, "pid": 5, "tid": 7, "ts": 1716454216823600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823105, "dur": 11,
    "args": {
      "External id": 12868,
      "cbid": 211, "correlation": 12868
    }
  },
  {
    "ph": "s", "id": 12868, "pid": 76337, "tid": -914061504, "ts": 1716454216823105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216823611, "dur": 4,
    "args": {
      "External id": 12880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12880,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 12880, "pid": 5, "tid": 7, "ts": 1716454216823611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823126, "dur": 6,
    "args": {
      "External id": 12880,
      "cbid": 211, "correlation": 12880
    }
  },
  {
    "ph": "s", "id": 12880, "pid": 76337, "tid": -914061504, "ts": 1716454216823126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216823616, "dur": 7,
    "args": {
      "External id": 12883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12883, "pid": 5, "tid": 7, "ts": 1716454216823616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823144, "dur": 6,
    "args": {
      "External id": 12883,
      "cbid": 211, "correlation": 12883
    }
  },
  {
    "ph": "s", "id": 12883, "pid": 76337, "tid": -914061504, "ts": 1716454216823144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216823624, "dur": 5,
    "args": {
      "External id": 12892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12892,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12892, "pid": 5, "tid": 7, "ts": 1716454216823624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823185, "dur": 9,
    "args": {
      "External id": 12892,
      "cbid": 211, "correlation": 12892
    }
  },
  {
    "ph": "s", "id": 12892, "pid": 76337, "tid": -914061504, "ts": 1716454216823185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823266, "dur": 0,
    "args": {
      "External id": 12902,
      "cbid": 317, "correlation": 12902
    }
  },
  {
    "ph": "f", "id": 12902, "pid": 76337, "tid": -914061504, "ts": 1716454216823266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823267, "dur": 0,
    "args": {
      "External id": 12903,
      "cbid": 203, "correlation": 12903
    }
  },
  {
    "ph": "f", "id": 12903, "pid": 76337, "tid": -914061504, "ts": 1716454216823267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823268, "dur": 0,
    "args": {
      "External id": 12904,
      "cbid": 205, "correlation": 12904
    }
  },
  {
    "ph": "f", "id": 12904, "pid": 76337, "tid": -914061504, "ts": 1716454216823268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216823630, "dur": 5,
    "args": {
      "External id": 12908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12908,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12908, "pid": 5, "tid": 7, "ts": 1716454216823630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823283, "dur": 13,
    "args": {
      "External id": 12908,
      "cbid": 211, "correlation": 12908
    }
  },
  {
    "ph": "s", "id": 12908, "pid": 76337, "tid": -914061504, "ts": 1716454216823283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216823637, "dur": 163,
    "args": {
      "External id": 12910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12910,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12910, "pid": 5, "tid": 7, "ts": 1716454216823637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823299, "dur": 6,
    "args": {
      "External id": 12910,
      "cbid": 211, "correlation": 12910
    }
  },
  {
    "ph": "s", "id": 12910, "pid": 76337, "tid": -914061504, "ts": 1716454216823299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216823802, "dur": 1,
    "args": {
      "External id": 12912,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12912,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 12912, "pid": 5, "tid": 7, "ts": 1716454216823802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823310, "dur": 7,
    "args": {
      "External id": 12912,
      "cbid": 51, "correlation": 12912
    }
  },
  {
    "ph": "s", "id": 12912, "pid": 76337, "tid": -914061504, "ts": 1716454216823310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216823806, "dur": 262,
    "args": {
      "External id": 12913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12913,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 12913, "pid": 5, "tid": 7, "ts": 1716454216823806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823318, "dur": 6,
    "args": {
      "External id": 12913,
      "cbid": 211, "correlation": 12913
    }
  },
  {
    "ph": "s", "id": 12913, "pid": 76337, "tid": -914061504, "ts": 1716454216823318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216824069, "dur": 6,
    "args": {
      "External id": 12915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12915, "pid": 5, "tid": 7, "ts": 1716454216824069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823328, "dur": 5,
    "args": {
      "External id": 12915,
      "cbid": 211, "correlation": 12915
    }
  },
  {
    "ph": "s", "id": 12915, "pid": 76337, "tid": -914061504, "ts": 1716454216823328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216824076, "dur": 6,
    "args": {
      "External id": 12921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12921, "pid": 5, "tid": 7, "ts": 1716454216824076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823358, "dur": 8,
    "args": {
      "External id": 12921,
      "cbid": 211, "correlation": 12921
    }
  },
  {
    "ph": "s", "id": 12921, "pid": 76337, "tid": -914061504, "ts": 1716454216823358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823419, "dur": 0,
    "args": {
      "External id": 12931,
      "cbid": 317, "correlation": 12931
    }
  },
  {
    "ph": "f", "id": 12931, "pid": 76337, "tid": -914061504, "ts": 1716454216823419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823420, "dur": 0,
    "args": {
      "External id": 12932,
      "cbid": 203, "correlation": 12932
    }
  },
  {
    "ph": "f", "id": 12932, "pid": 76337, "tid": -914061504, "ts": 1716454216823420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823420, "dur": 0,
    "args": {
      "External id": 12933,
      "cbid": 205, "correlation": 12933
    }
  },
  {
    "ph": "f", "id": 12933, "pid": 76337, "tid": -914061504, "ts": 1716454216823420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216824083, "dur": 8,
    "args": {
      "External id": 12937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12937,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12937, "pid": 5, "tid": 7, "ts": 1716454216824083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823432, "dur": 11,
    "args": {
      "External id": 12937,
      "cbid": 211, "correlation": 12937
    }
  },
  {
    "ph": "s", "id": 12937, "pid": 76337, "tid": -914061504, "ts": 1716454216823432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216824092, "dur": 3,
    "args": {
      "External id": 12939,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 12939,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.293103448275862
    }
  },
  {
    "ph": "f", "id": 12939, "pid": 5, "tid": 7, "ts": 1716454216824092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823449, "dur": 10,
    "args": {
      "External id": 12939,
      "cbid": 51, "correlation": 12939
    }
  },
  {
    "ph": "s", "id": 12939, "pid": 76337, "tid": -914061504, "ts": 1716454216823449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216824097, "dur": 95,
    "args": {
      "External id": 12940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12940,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 12940, "pid": 5, "tid": 7, "ts": 1716454216824097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823460, "dur": 6,
    "args": {
      "External id": 12940,
      "cbid": 211, "correlation": 12940
    }
  },
  {
    "ph": "s", "id": 12940, "pid": 76337, "tid": -914061504, "ts": 1716454216823460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216824193, "dur": 6,
    "args": {
      "External id": 12942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12942,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12942, "pid": 5, "tid": 7, "ts": 1716454216824193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823470, "dur": 5,
    "args": {
      "External id": 12942,
      "cbid": 211, "correlation": 12942
    }
  },
  {
    "ph": "s", "id": 12942, "pid": 76337, "tid": -914061504, "ts": 1716454216823470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216824200, "dur": 6,
    "args": {
      "External id": 12948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12948, "pid": 5, "tid": 7, "ts": 1716454216824200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823497, "dur": 8,
    "args": {
      "External id": 12948,
      "cbid": 211, "correlation": 12948
    }
  },
  {
    "ph": "s", "id": 12948, "pid": 76337, "tid": -914061504, "ts": 1716454216823497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216824208, "dur": 5,
    "args": {
      "External id": 12956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12956,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12956, "pid": 5, "tid": 7, "ts": 1716454216824208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823527, "dur": 8,
    "args": {
      "External id": 12956,
      "cbid": 211, "correlation": 12956
    }
  },
  {
    "ph": "s", "id": 12956, "pid": 76337, "tid": -914061504, "ts": 1716454216823527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216824214, "dur": 4,
    "args": {
      "External id": 12964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12964,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 12964, "pid": 5, "tid": 7, "ts": 1716454216824214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823556, "dur": 9,
    "args": {
      "External id": 12964,
      "cbid": 211, "correlation": 12964
    }
  },
  {
    "ph": "s", "id": 12964, "pid": 76337, "tid": -914061504, "ts": 1716454216823556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216824220, "dur": 11,
    "args": {
      "External id": 12973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12973,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 12973, "pid": 5, "tid": 7, "ts": 1716454216824220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823603, "dur": 10,
    "args": {
      "External id": 12973,
      "cbid": 211, "correlation": 12973
    }
  },
  {
    "ph": "s", "id": 12973, "pid": 76337, "tid": -914061504, "ts": 1716454216823603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216824232, "dur": 12,
    "args": {
      "External id": 12993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 12993,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 12993, "pid": 5, "tid": 7, "ts": 1716454216824232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823674, "dur": 11,
    "args": {
      "External id": 12993,
      "cbid": 211, "correlation": 12993
    }
  },
  {
    "ph": "s", "id": 12993, "pid": 76337, "tid": -914061504, "ts": 1716454216823674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216824246, "dur": 4,
    "args": {
      "External id": 13005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13005,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13005, "pid": 5, "tid": 7, "ts": 1716454216824246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823697, "dur": 6,
    "args": {
      "External id": 13005,
      "cbid": 211, "correlation": 13005
    }
  },
  {
    "ph": "s", "id": 13005, "pid": 76337, "tid": -914061504, "ts": 1716454216823697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216824251, "dur": 10,
    "args": {
      "External id": 13008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13008,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13008, "pid": 5, "tid": 7, "ts": 1716454216824251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823715, "dur": 7,
    "args": {
      "External id": 13008,
      "cbid": 211, "correlation": 13008
    }
  },
  {
    "ph": "s", "id": 13008, "pid": 76337, "tid": -914061504, "ts": 1716454216823715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216824263, "dur": 6,
    "args": {
      "External id": 13017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13017,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13017, "pid": 5, "tid": 7, "ts": 1716454216824263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823753, "dur": 10,
    "args": {
      "External id": 13017,
      "cbid": 211, "correlation": 13017
    }
  },
  {
    "ph": "s", "id": 13017, "pid": 76337, "tid": -914061504, "ts": 1716454216823753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823804, "dur": 0,
    "args": {
      "External id": 13027,
      "cbid": 317, "correlation": 13027
    }
  },
  {
    "ph": "f", "id": 13027, "pid": 76337, "tid": -914061504, "ts": 1716454216823804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823805, "dur": 0,
    "args": {
      "External id": 13028,
      "cbid": 203, "correlation": 13028
    }
  },
  {
    "ph": "f", "id": 13028, "pid": 76337, "tid": -914061504, "ts": 1716454216823805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823805, "dur": 0,
    "args": {
      "External id": 13029,
      "cbid": 205, "correlation": 13029
    }
  },
  {
    "ph": "f", "id": 13029, "pid": 76337, "tid": -914061504, "ts": 1716454216823805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216824270, "dur": 7,
    "args": {
      "External id": 13033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13033,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13033, "pid": 5, "tid": 7, "ts": 1716454216824270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823820, "dur": 12,
    "args": {
      "External id": 13033,
      "cbid": 211, "correlation": 13033
    }
  },
  {
    "ph": "s", "id": 13033, "pid": 76337, "tid": -914061504, "ts": 1716454216823820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216824278, "dur": 324,
    "args": {
      "External id": 13035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13035,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13035, "pid": 5, "tid": 7, "ts": 1716454216824278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823834, "dur": 5,
    "args": {
      "External id": 13035,
      "cbid": 211, "correlation": 13035
    }
  },
  {
    "ph": "s", "id": 13035, "pid": 76337, "tid": -914061504, "ts": 1716454216823834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216824604, "dur": 1,
    "args": {
      "External id": 13037,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13037,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 13037, "pid": 5, "tid": 7, "ts": 1716454216824604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823844, "dur": 6,
    "args": {
      "External id": 13037,
      "cbid": 51, "correlation": 13037
    }
  },
  {
    "ph": "s", "id": 13037, "pid": 76337, "tid": -914061504, "ts": 1716454216823844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216824608, "dur": 501,
    "args": {
      "External id": 13038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13038,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13038, "pid": 5, "tid": 7, "ts": 1716454216824608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823851, "dur": 6,
    "args": {
      "External id": 13038,
      "cbid": 211, "correlation": 13038
    }
  },
  {
    "ph": "s", "id": 13038, "pid": 76337, "tid": -914061504, "ts": 1716454216823851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825110, "dur": 5,
    "args": {
      "External id": 13040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13040,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13040, "pid": 5, "tid": 7, "ts": 1716454216825110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823861, "dur": 5,
    "args": {
      "External id": 13040,
      "cbid": 211, "correlation": 13040
    }
  },
  {
    "ph": "s", "id": 13040, "pid": 76337, "tid": -914061504, "ts": 1716454216823861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216825117, "dur": 6,
    "args": {
      "External id": 13046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13046,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13046, "pid": 5, "tid": 7, "ts": 1716454216825117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823889, "dur": 8,
    "args": {
      "External id": 13046,
      "cbid": 211, "correlation": 13046
    }
  },
  {
    "ph": "s", "id": 13046, "pid": 76337, "tid": -914061504, "ts": 1716454216823889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216825125, "dur": 3,
    "args": {
      "External id": 13054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13054,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 13054, "pid": 5, "tid": 7, "ts": 1716454216825125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216823933, "dur": 9,
    "args": {
      "External id": 13054,
      "cbid": 211, "correlation": 13054
    }
  },
  {
    "ph": "s", "id": 13054, "pid": 76337, "tid": -914061504, "ts": 1716454216823933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824028, "dur": 1,
    "args": {
      "External id": 13070,
      "cbid": 251, "correlation": 13070
    }
  },
  {
    "ph": "f", "id": 13070, "pid": 76337, "tid": -914061504, "ts": 1716454216824028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824033, "dur": 0,
    "args": {
      "External id": 13072,
      "cbid": 251, "correlation": 13072
    }
  },
  {
    "ph": "f", "id": 13072, "pid": 76337, "tid": -914061504, "ts": 1716454216824033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216825129, "dur": 12,
    "args": {
      "External id": 13073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13073,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13073, "pid": 5, "tid": 7, "ts": 1716454216825129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824035, "dur": 13,
    "args": {
      "External id": 13073,
      "cbid": 211, "correlation": 13073
    }
  },
  {
    "ph": "s", "id": 13073, "pid": 76337, "tid": -914061504, "ts": 1716454216824035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216825143, "dur": 5,
    "args": {
      "External id": 13075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13075,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13075, "pid": 5, "tid": 7, "ts": 1716454216825143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824050, "dur": 6,
    "args": {
      "External id": 13075,
      "cbid": 211, "correlation": 13075
    }
  },
  {
    "ph": "s", "id": 13075, "pid": 76337, "tid": -914061504, "ts": 1716454216824050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216825149, "dur": 6,
    "args": {
      "External id": 13085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13085, "pid": 5, "tid": 7, "ts": 1716454216825149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824108, "dur": 12,
    "args": {
      "External id": 13085,
      "cbid": 211, "correlation": 13085
    }
  },
  {
    "ph": "s", "id": 13085, "pid": 76337, "tid": -914061504, "ts": 1716454216824108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825157, "dur": 10,
    "args": {
      "External id": 13105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13105,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 13105, "pid": 5, "tid": 7, "ts": 1716454216825157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824173, "dur": 10,
    "args": {
      "External id": 13105,
      "cbid": 211, "correlation": 13105
    }
  },
  {
    "ph": "s", "id": 13105, "pid": 76337, "tid": -914061504, "ts": 1716454216824173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216825168, "dur": 4,
    "args": {
      "External id": 13117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13117,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 13117, "pid": 5, "tid": 7, "ts": 1716454216825168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824203, "dur": 8,
    "args": {
      "External id": 13117,
      "cbid": 211, "correlation": 13117
    }
  },
  {
    "ph": "s", "id": 13117, "pid": 76337, "tid": -914061504, "ts": 1716454216824203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216825173, "dur": 7,
    "args": {
      "External id": 13120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13120, "pid": 5, "tid": 7, "ts": 1716454216825173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824224, "dur": 7,
    "args": {
      "External id": 13120,
      "cbid": 211, "correlation": 13120
    }
  },
  {
    "ph": "s", "id": 13120, "pid": 76337, "tid": -914061504, "ts": 1716454216824224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216825181, "dur": 5,
    "args": {
      "External id": 13129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13129,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13129, "pid": 5, "tid": 7, "ts": 1716454216825181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824267, "dur": 10,
    "args": {
      "External id": 13129,
      "cbid": 211, "correlation": 13129
    }
  },
  {
    "ph": "s", "id": 13129, "pid": 76337, "tid": -914061504, "ts": 1716454216824267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824341, "dur": 0,
    "args": {
      "External id": 13139,
      "cbid": 317, "correlation": 13139
    }
  },
  {
    "ph": "f", "id": 13139, "pid": 76337, "tid": -914061504, "ts": 1716454216824341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824342, "dur": 0,
    "args": {
      "External id": 13140,
      "cbid": 203, "correlation": 13140
    }
  },
  {
    "ph": "f", "id": 13140, "pid": 76337, "tid": -914061504, "ts": 1716454216824342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824343, "dur": 0,
    "args": {
      "External id": 13141,
      "cbid": 205, "correlation": 13141
    }
  },
  {
    "ph": "f", "id": 13141, "pid": 76337, "tid": -914061504, "ts": 1716454216824343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825187, "dur": 5,
    "args": {
      "External id": 13145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13145,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13145, "pid": 5, "tid": 7, "ts": 1716454216825187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824356, "dur": 13,
    "args": {
      "External id": 13145,
      "cbid": 211, "correlation": 13145
    }
  },
  {
    "ph": "s", "id": 13145, "pid": 76337, "tid": -914061504, "ts": 1716454216824356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825193, "dur": 163,
    "args": {
      "External id": 13147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13147,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13147, "pid": 5, "tid": 7, "ts": 1716454216825193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824372, "dur": 5,
    "args": {
      "External id": 13147,
      "cbid": 211, "correlation": 13147
    }
  },
  {
    "ph": "s", "id": 13147, "pid": 76337, "tid": -914061504, "ts": 1716454216824372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216825358, "dur": 1,
    "args": {
      "External id": 13149,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13149,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 13149, "pid": 5, "tid": 7, "ts": 1716454216825358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824383, "dur": 6,
    "args": {
      "External id": 13149,
      "cbid": 51, "correlation": 13149
    }
  },
  {
    "ph": "s", "id": 13149, "pid": 76337, "tid": -914061504, "ts": 1716454216824383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216825362, "dur": 261,
    "args": {
      "External id": 13150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13150,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13150, "pid": 5, "tid": 7, "ts": 1716454216825362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824391, "dur": 6,
    "args": {
      "External id": 13150,
      "cbid": 211, "correlation": 13150
    }
  },
  {
    "ph": "s", "id": 13150, "pid": 76337, "tid": -914061504, "ts": 1716454216824391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825624, "dur": 6,
    "args": {
      "External id": 13152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13152,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13152, "pid": 5, "tid": 7, "ts": 1716454216825624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824400, "dur": 5,
    "args": {
      "External id": 13152,
      "cbid": 211, "correlation": 13152
    }
  },
  {
    "ph": "s", "id": 13152, "pid": 76337, "tid": -914061504, "ts": 1716454216824400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216825632, "dur": 6,
    "args": {
      "External id": 13158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13158, "pid": 5, "tid": 7, "ts": 1716454216825632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824429, "dur": 9,
    "args": {
      "External id": 13158,
      "cbid": 211, "correlation": 13158
    }
  },
  {
    "ph": "s", "id": 13158, "pid": 76337, "tid": -914061504, "ts": 1716454216824429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824489, "dur": 0,
    "args": {
      "External id": 13168,
      "cbid": 317, "correlation": 13168
    }
  },
  {
    "ph": "f", "id": 13168, "pid": 76337, "tid": -914061504, "ts": 1716454216824489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824490, "dur": 0,
    "args": {
      "External id": 13169,
      "cbid": 203, "correlation": 13169
    }
  },
  {
    "ph": "f", "id": 13169, "pid": 76337, "tid": -914061504, "ts": 1716454216824490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824491, "dur": 0,
    "args": {
      "External id": 13170,
      "cbid": 205, "correlation": 13170
    }
  },
  {
    "ph": "f", "id": 13170, "pid": 76337, "tid": -914061504, "ts": 1716454216824491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825639, "dur": 8,
    "args": {
      "External id": 13174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13174,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13174, "pid": 5, "tid": 7, "ts": 1716454216825639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824503, "dur": 11,
    "args": {
      "External id": 13174,
      "cbid": 211, "correlation": 13174
    }
  },
  {
    "ph": "s", "id": 13174, "pid": 76337, "tid": -914061504, "ts": 1716454216824503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216825649, "dur": 3,
    "args": {
      "External id": 13176,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13176,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 13176, "pid": 5, "tid": 7, "ts": 1716454216825649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824520, "dur": 10,
    "args": {
      "External id": 13176,
      "cbid": 51, "correlation": 13176
    }
  },
  {
    "ph": "s", "id": 13176, "pid": 76337, "tid": -914061504, "ts": 1716454216824520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216825653, "dur": 95,
    "args": {
      "External id": 13177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13177,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 13177, "pid": 5, "tid": 7, "ts": 1716454216825653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824531, "dur": 6,
    "args": {
      "External id": 13177,
      "cbid": 211, "correlation": 13177
    }
  },
  {
    "ph": "s", "id": 13177, "pid": 76337, "tid": -914061504, "ts": 1716454216824531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825749, "dur": 5,
    "args": {
      "External id": 13179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13179,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13179, "pid": 5, "tid": 7, "ts": 1716454216825749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824541, "dur": 5,
    "args": {
      "External id": 13179,
      "cbid": 211, "correlation": 13179
    }
  },
  {
    "ph": "s", "id": 13179, "pid": 76337, "tid": -914061504, "ts": 1716454216824541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216825756, "dur": 6,
    "args": {
      "External id": 13185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13185, "pid": 5, "tid": 7, "ts": 1716454216825756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824567, "dur": 9,
    "args": {
      "External id": 13185,
      "cbid": 211, "correlation": 13185
    }
  },
  {
    "ph": "s", "id": 13185, "pid": 76337, "tid": -914061504, "ts": 1716454216824567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216825764, "dur": 5,
    "args": {
      "External id": 13193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13193,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13193, "pid": 5, "tid": 7, "ts": 1716454216825764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824597, "dur": 8,
    "args": {
      "External id": 13193,
      "cbid": 211, "correlation": 13193
    }
  },
  {
    "ph": "s", "id": 13193, "pid": 76337, "tid": -914061504, "ts": 1716454216824597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216825769, "dur": 4,
    "args": {
      "External id": 13201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13201,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 13201, "pid": 5, "tid": 7, "ts": 1716454216825769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824627, "dur": 8,
    "args": {
      "External id": 13201,
      "cbid": 211, "correlation": 13201
    }
  },
  {
    "ph": "s", "id": 13201, "pid": 76337, "tid": -914061504, "ts": 1716454216824627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454216825780, "dur": 14,
    "args": {
      "External id": 13212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13212,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13212, "pid": 5, "tid": 7, "ts": 1716454216825780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824780, "dur": 100,
    "args": {
      "External id": 13212,
      "cbid": 211, "correlation": 13212
    }
  },
  {
    "ph": "s", "id": 13212, "pid": 76337, "tid": -914061504, "ts": 1716454216824780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824946, "dur": 0,
    "args": {
      "External id": 13222,
      "cbid": 317, "correlation": 13222
    }
  },
  {
    "ph": "f", "id": 13222, "pid": 76337, "tid": -914061504, "ts": 1716454216824946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824947, "dur": 0,
    "args": {
      "External id": 13223,
      "cbid": 203, "correlation": 13223
    }
  },
  {
    "ph": "f", "id": 13223, "pid": 76337, "tid": -914061504, "ts": 1716454216824947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824948, "dur": 0,
    "args": {
      "External id": 13224,
      "cbid": 205, "correlation": 13224
    }
  },
  {
    "ph": "f", "id": 13224, "pid": 76337, "tid": -914061504, "ts": 1716454216824948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825796, "dur": 10,
    "args": {
      "External id": 13228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13228,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13228, "pid": 5, "tid": 7, "ts": 1716454216825796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824972, "dur": 20,
    "args": {
      "External id": 13228,
      "cbid": 211, "correlation": 13228
    }
  },
  {
    "ph": "s", "id": 13228, "pid": 76337, "tid": -914061504, "ts": 1716454216824972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216825808, "dur": 164,
    "args": {
      "External id": 13230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13230,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13230, "pid": 5, "tid": 7, "ts": 1716454216825808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216824996, "dur": 5,
    "args": {
      "External id": 13230,
      "cbid": 211, "correlation": 13230
    }
  },
  {
    "ph": "s", "id": 13230, "pid": 76337, "tid": -914061504, "ts": 1716454216824996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216825974, "dur": 1,
    "args": {
      "External id": 13232,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13232,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5357142857142857
    }
  },
  {
    "ph": "f", "id": 13232, "pid": 5, "tid": 7, "ts": 1716454216825974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825008, "dur": 7,
    "args": {
      "External id": 13232,
      "cbid": 51, "correlation": 13232
    }
  },
  {
    "ph": "s", "id": 13232, "pid": 76337, "tid": -914061504, "ts": 1716454216825008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216825978, "dur": 666,
    "args": {
      "External id": 13233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13233,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13233, "pid": 5, "tid": 7, "ts": 1716454216825978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825016, "dur": 7,
    "args": {
      "External id": 13233,
      "cbid": 211, "correlation": 13233
    }
  },
  {
    "ph": "s", "id": 13233, "pid": 76337, "tid": -914061504, "ts": 1716454216825016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216826646, "dur": 14,
    "args": {
      "External id": 13235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13235, "pid": 5, "tid": 7, "ts": 1716454216826646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825027, "dur": 5,
    "args": {
      "External id": 13235,
      "cbid": 211, "correlation": 13235
    }
  },
  {
    "ph": "s", "id": 13235, "pid": 76337, "tid": -914061504, "ts": 1716454216825027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216826661, "dur": 16,
    "args": {
      "External id": 13241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13241, "pid": 5, "tid": 7, "ts": 1716454216826661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825057, "dur": 9,
    "args": {
      "External id": 13241,
      "cbid": 211, "correlation": 13241
    }
  },
  {
    "ph": "s", "id": 13241, "pid": 76337, "tid": -914061504, "ts": 1716454216825057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216826678, "dur": 30,
    "args": {
      "External id": 13250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13250,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13250, "pid": 5, "tid": 7, "ts": 1716454216826678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825198, "dur": 14,
    "args": {
      "External id": 13250,
      "cbid": 211, "correlation": 13250
    }
  },
  {
    "ph": "s", "id": 13250, "pid": 76337, "tid": -914061504, "ts": 1716454216825198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216826710, "dur": 31,
    "args": {
      "External id": 13270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13270,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 13270, "pid": 5, "tid": 7, "ts": 1716454216826710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825282, "dur": 12,
    "args": {
      "External id": 13270,
      "cbid": 211, "correlation": 13270
    }
  },
  {
    "ph": "s", "id": 13270, "pid": 76337, "tid": -914061504, "ts": 1716454216825282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216826742, "dur": 5,
    "args": {
      "External id": 13282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13282,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13282, "pid": 5, "tid": 7, "ts": 1716454216826742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825305, "dur": 6,
    "args": {
      "External id": 13282,
      "cbid": 211, "correlation": 13282
    }
  },
  {
    "ph": "s", "id": 13282, "pid": 76337, "tid": -914061504, "ts": 1716454216825305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216826749, "dur": 31,
    "args": {
      "External id": 13285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13285,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13285, "pid": 5, "tid": 7, "ts": 1716454216826749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825323, "dur": 7,
    "args": {
      "External id": 13285,
      "cbid": 211, "correlation": 13285
    }
  },
  {
    "ph": "s", "id": 13285, "pid": 76337, "tid": -914061504, "ts": 1716454216825323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216826781, "dur": 22,
    "args": {
      "External id": 13294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13294,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13294, "pid": 5, "tid": 7, "ts": 1716454216826781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825364, "dur": 10,
    "args": {
      "External id": 13294,
      "cbid": 211, "correlation": 13294
    }
  },
  {
    "ph": "s", "id": 13294, "pid": 76337, "tid": -914061504, "ts": 1716454216825364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825434, "dur": 0,
    "args": {
      "External id": 13304,
      "cbid": 317, "correlation": 13304
    }
  },
  {
    "ph": "f", "id": 13304, "pid": 76337, "tid": -914061504, "ts": 1716454216825434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825435, "dur": 0,
    "args": {
      "External id": 13305,
      "cbid": 203, "correlation": 13305
    }
  },
  {
    "ph": "f", "id": 13305, "pid": 76337, "tid": -914061504, "ts": 1716454216825435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216825436, "dur": 0,
    "args": {
      "External id": 13306,
      "cbid": 205, "correlation": 13306
    }
  },
  {
    "ph": "f", "id": 13306, "pid": 76337, "tid": -914061504, "ts": 1716454216825436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216828720, "dur": 21,
    "args": {
      "External id": 13310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13310, "pid": 5, "tid": 7, "ts": 1716454216828720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216828690, "dur": 30,
    "args": {
      "External id": 13310,
      "cbid": 211, "correlation": 13310
    }
  },
  {
    "ph": "s", "id": 13310, "pid": 76337, "tid": -914061504, "ts": 1716454216828690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216828742, "dur": 323,
    "args": {
      "External id": 13312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13312,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13312, "pid": 5, "tid": 7, "ts": 1716454216828742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216828723, "dur": 6,
    "args": {
      "External id": 13312,
      "cbid": 211, "correlation": 13312
    }
  },
  {
    "ph": "s", "id": 13312, "pid": 76337, "tid": -914061504, "ts": 1716454216828723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216829067, "dur": 1,
    "args": {
      "External id": 13314,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13314,
      "bytes": 960, "memory bandwidth (GB/s)": 0.4918032786885246
    }
  },
  {
    "ph": "f", "id": 13314, "pid": 5, "tid": 7, "ts": 1716454216829067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216828738, "dur": 11,
    "args": {
      "External id": 13314,
      "cbid": 51, "correlation": 13314
    }
  },
  {
    "ph": "s", "id": 13314, "pid": 76337, "tid": -914061504, "ts": 1716454216828738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216829071, "dur": 1249,
    "args": {
      "External id": 13315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13315,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13315, "pid": 5, "tid": 7, "ts": 1716454216829071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216828750, "dur": 8,
    "args": {
      "External id": 13315,
      "cbid": 211, "correlation": 13315
    }
  },
  {
    "ph": "s", "id": 13315, "pid": 76337, "tid": -914061504, "ts": 1716454216828750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216830322, "dur": 13,
    "args": {
      "External id": 13317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13317,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13317, "pid": 5, "tid": 7, "ts": 1716454216830322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216828763, "dur": 6,
    "args": {
      "External id": 13317,
      "cbid": 211, "correlation": 13317
    }
  },
  {
    "ph": "s", "id": 13317, "pid": 76337, "tid": -914061504, "ts": 1716454216828763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216830336, "dur": 15,
    "args": {
      "External id": 13323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13323, "pid": 5, "tid": 7, "ts": 1716454216830336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829536, "dur": 12,
    "args": {
      "External id": 13323,
      "cbid": 211, "correlation": 13323
    }
  },
  {
    "ph": "s", "id": 13323, "pid": 76337, "tid": -914061504, "ts": 1716454216829536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216830352, "dur": 3,
    "args": {
      "External id": 13331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13331,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 13331, "pid": 5, "tid": 7, "ts": 1716454216830352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829622, "dur": 10,
    "args": {
      "External id": 13331,
      "cbid": 211, "correlation": 13331
    }
  },
  {
    "ph": "s", "id": 13331, "pid": 76337, "tid": -914061504, "ts": 1716454216829622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829737, "dur": 3,
    "args": {
      "External id": 13347,
      "cbid": 251, "correlation": 13347
    }
  },
  {
    "ph": "f", "id": 13347, "pid": 76337, "tid": -914061504, "ts": 1716454216829737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829747, "dur": 0,
    "args": {
      "External id": 13349,
      "cbid": 251, "correlation": 13349
    }
  },
  {
    "ph": "f", "id": 13349, "pid": 76337, "tid": -914061504, "ts": 1716454216829747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216830357, "dur": 13,
    "args": {
      "External id": 13350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13350,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13350, "pid": 5, "tid": 7, "ts": 1716454216830357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829750, "dur": 15,
    "args": {
      "External id": 13350,
      "cbid": 211, "correlation": 13350
    }
  },
  {
    "ph": "s", "id": 13350, "pid": 76337, "tid": -914061504, "ts": 1716454216829750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216830372, "dur": 6,
    "args": {
      "External id": 13352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13352,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13352, "pid": 5, "tid": 7, "ts": 1716454216830372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829768, "dur": 8,
    "args": {
      "External id": 13352,
      "cbid": 211, "correlation": 13352
    }
  },
  {
    "ph": "s", "id": 13352, "pid": 76337, "tid": -914061504, "ts": 1716454216829768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216830378, "dur": 17,
    "args": {
      "External id": 13362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13362, "pid": 5, "tid": 7, "ts": 1716454216830378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829849, "dur": 12,
    "args": {
      "External id": 13362,
      "cbid": 211, "correlation": 13362
    }
  },
  {
    "ph": "s", "id": 13362, "pid": 76337, "tid": -914061504, "ts": 1716454216829849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216830397, "dur": 18,
    "args": {
      "External id": 13382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13382,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 13382, "pid": 5, "tid": 7, "ts": 1716454216830397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829939, "dur": 13,
    "args": {
      "External id": 13382,
      "cbid": 211, "correlation": 13382
    }
  },
  {
    "ph": "s", "id": 13382, "pid": 76337, "tid": -914061504, "ts": 1716454216829939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216830416, "dur": 4,
    "args": {
      "External id": 13394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13394,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 13394, "pid": 5, "tid": 7, "ts": 1716454216830416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829963, "dur": 7,
    "args": {
      "External id": 13394,
      "cbid": 211, "correlation": 13394
    }
  },
  {
    "ph": "s", "id": 13394, "pid": 76337, "tid": -914061504, "ts": 1716454216829963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216830421, "dur": 17,
    "args": {
      "External id": 13397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13397, "pid": 5, "tid": 7, "ts": 1716454216830421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216829997, "dur": 9,
    "args": {
      "External id": 13397,
      "cbid": 211, "correlation": 13397
    }
  },
  {
    "ph": "s", "id": 13397, "pid": 76337, "tid": -914061504, "ts": 1716454216829997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216830440, "dur": 11,
    "args": {
      "External id": 13406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13406,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13406, "pid": 5, "tid": 7, "ts": 1716454216830440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830043, "dur": 10,
    "args": {
      "External id": 13406,
      "cbid": 211, "correlation": 13406
    }
  },
  {
    "ph": "s", "id": 13406, "pid": 76337, "tid": -914061504, "ts": 1716454216830043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830133, "dur": 0,
    "args": {
      "External id": 13416,
      "cbid": 317, "correlation": 13416
    }
  },
  {
    "ph": "f", "id": 13416, "pid": 76337, "tid": -914061504, "ts": 1716454216830133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830134, "dur": 0,
    "args": {
      "External id": 13417,
      "cbid": 203, "correlation": 13417
    }
  },
  {
    "ph": "f", "id": 13417, "pid": 76337, "tid": -914061504, "ts": 1716454216830134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830135, "dur": 1,
    "args": {
      "External id": 13418,
      "cbid": 205, "correlation": 13418
    }
  },
  {
    "ph": "f", "id": 13418, "pid": 76337, "tid": -914061504, "ts": 1716454216830135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216830452, "dur": 11,
    "args": {
      "External id": 13422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13422, "pid": 5, "tid": 7, "ts": 1716454216830452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830151, "dur": 12,
    "args": {
      "External id": 13422,
      "cbid": 211, "correlation": 13422
    }
  },
  {
    "ph": "s", "id": 13422, "pid": 76337, "tid": -914061504, "ts": 1716454216830151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216830464, "dur": 164,
    "args": {
      "External id": 13424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13424,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13424, "pid": 5, "tid": 7, "ts": 1716454216830464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830166, "dur": 6,
    "args": {
      "External id": 13424,
      "cbid": 211, "correlation": 13424
    }
  },
  {
    "ph": "s", "id": 13424, "pid": 76337, "tid": -914061504, "ts": 1716454216830166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216830631, "dur": 1,
    "args": {
      "External id": 13426,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13426,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 13426, "pid": 5, "tid": 7, "ts": 1716454216830631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830178, "dur": 7,
    "args": {
      "External id": 13426,
      "cbid": 51, "correlation": 13426
    }
  },
  {
    "ph": "s", "id": 13426, "pid": 76337, "tid": -914061504, "ts": 1716454216830178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216830635, "dur": 653,
    "args": {
      "External id": 13427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13427,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13427, "pid": 5, "tid": 7, "ts": 1716454216830635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830187, "dur": 6,
    "args": {
      "External id": 13427,
      "cbid": 211, "correlation": 13427
    }
  },
  {
    "ph": "s", "id": 13427, "pid": 76337, "tid": -914061504, "ts": 1716454216830187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216831289, "dur": 13,
    "args": {
      "External id": 13429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13429,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13429, "pid": 5, "tid": 7, "ts": 1716454216831289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830197, "dur": 5,
    "args": {
      "External id": 13429,
      "cbid": 211, "correlation": 13429
    }
  },
  {
    "ph": "s", "id": 13429, "pid": 76337, "tid": -914061504, "ts": 1716454216830197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216831303, "dur": 15,
    "args": {
      "External id": 13435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13435, "pid": 5, "tid": 7, "ts": 1716454216831303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830226, "dur": 9,
    "args": {
      "External id": 13435,
      "cbid": 211, "correlation": 13435
    }
  },
  {
    "ph": "s", "id": 13435, "pid": 76337, "tid": -914061504, "ts": 1716454216830226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830286, "dur": 0,
    "args": {
      "External id": 13445,
      "cbid": 317, "correlation": 13445
    }
  },
  {
    "ph": "f", "id": 13445, "pid": 76337, "tid": -914061504, "ts": 1716454216830286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830286, "dur": 0,
    "args": {
      "External id": 13446,
      "cbid": 203, "correlation": 13446
    }
  },
  {
    "ph": "f", "id": 13446, "pid": 76337, "tid": -914061504, "ts": 1716454216830286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216830287, "dur": 0,
    "args": {
      "External id": 13447,
      "cbid": 205, "correlation": 13447
    }
  },
  {
    "ph": "f", "id": 13447, "pid": 76337, "tid": -914061504, "ts": 1716454216830287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216832016, "dur": 21,
    "args": {
      "External id": 13451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13451, "pid": 5, "tid": 7, "ts": 1716454216832016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216831998, "dur": 19,
    "args": {
      "External id": 13451,
      "cbid": 211, "correlation": 13451
    }
  },
  {
    "ph": "s", "id": 13451, "pid": 76337, "tid": -914061504, "ts": 1716454216831998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216832038, "dur": 4,
    "args": {
      "External id": 13453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13453, "pid": 5, "tid": 7, "ts": 1716454216832038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832023, "dur": 11,
    "args": {
      "External id": 13453,
      "cbid": 211, "correlation": 13453
    }
  },
  {
    "ph": "s", "id": 13453, "pid": 76337, "tid": -914061504, "ts": 1716454216832023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832037, "dur": 0,
    "args": {
      "External id": 13454,
      "cbid": 51, "correlation": 13454
    }
  },
  {
    "ph": "s", "id": 13454, "pid": 76337, "tid": -914061504, "ts": 1716454216832037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216832048, "dur": 177,
    "args": {
      "External id": 13455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13455,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 13455, "pid": 5, "tid": 7, "ts": 1716454216832048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832038, "dur": 8,
    "args": {
      "External id": 13455,
      "cbid": 211, "correlation": 13455
    }
  },
  {
    "ph": "s", "id": 13455, "pid": 76337, "tid": -914061504, "ts": 1716454216832038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216832495, "dur": 16,
    "args": {
      "External id": 13460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13460, "pid": 5, "tid": 7, "ts": 1716454216832495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832483, "dur": 10,
    "args": {
      "External id": 13460,
      "cbid": 211, "correlation": 13460
    }
  },
  {
    "ph": "s", "id": 13460, "pid": 76337, "tid": -914061504, "ts": 1716454216832483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216832531, "dur": 13,
    "args": {
      "External id": 13468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13468,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13468, "pid": 5, "tid": 7, "ts": 1716454216832531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832521, "dur": 10,
    "args": {
      "External id": 13468,
      "cbid": 211, "correlation": 13468
    }
  },
  {
    "ph": "s", "id": 13468, "pid": 76337, "tid": -914061504, "ts": 1716454216832521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216832581, "dur": 10,
    "args": {
      "External id": 13476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13476,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13476, "pid": 5, "tid": 7, "ts": 1716454216832581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832571, "dur": 10,
    "args": {
      "External id": 13476,
      "cbid": 211, "correlation": 13476
    }
  },
  {
    "ph": "s", "id": 13476, "pid": 76337, "tid": -914061504, "ts": 1716454216832571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216832705, "dur": 18,
    "args": {
      "External id": 13496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13496,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 13496, "pid": 5, "tid": 7, "ts": 1716454216832705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832691, "dur": 14,
    "args": {
      "External id": 13496,
      "cbid": 211, "correlation": 13496
    }
  },
  {
    "ph": "s", "id": 13496, "pid": 76337, "tid": -914061504, "ts": 1716454216832691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216832747, "dur": 4,
    "args": {
      "External id": 13508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13508,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 13508, "pid": 5, "tid": 7, "ts": 1716454216832747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832737, "dur": 9,
    "args": {
      "External id": 13508,
      "cbid": 211, "correlation": 13508
    }
  },
  {
    "ph": "s", "id": 13508, "pid": 76337, "tid": -914061504, "ts": 1716454216832737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216832769, "dur": 17,
    "args": {
      "External id": 13511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13511, "pid": 5, "tid": 7, "ts": 1716454216832769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832761, "dur": 7,
    "args": {
      "External id": 13511,
      "cbid": 211, "correlation": 13511
    }
  },
  {
    "ph": "s", "id": 13511, "pid": 76337, "tid": -914061504, "ts": 1716454216832761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832839, "dur": 0,
    "args": {
      "External id": 13522,
      "cbid": 317, "correlation": 13522
    }
  },
  {
    "ph": "f", "id": 13522, "pid": 76337, "tid": -914061504, "ts": 1716454216832839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832840, "dur": 0,
    "args": {
      "External id": 13523,
      "cbid": 203, "correlation": 13523
    }
  },
  {
    "ph": "f", "id": 13523, "pid": 76337, "tid": -914061504, "ts": 1716454216832840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832841, "dur": 0,
    "args": {
      "External id": 13524,
      "cbid": 205, "correlation": 13524
    }
  },
  {
    "ph": "f", "id": 13524, "pid": 76337, "tid": -914061504, "ts": 1716454216832841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216832870, "dur": 11,
    "args": {
      "External id": 13528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13528,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13528, "pid": 5, "tid": 7, "ts": 1716454216832870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832857, "dur": 13,
    "args": {
      "External id": 13528,
      "cbid": 211, "correlation": 13528
    }
  },
  {
    "ph": "s", "id": 13528, "pid": 76337, "tid": -914061504, "ts": 1716454216832857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216832883, "dur": 3,
    "args": {
      "External id": 13530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13530, "pid": 5, "tid": 7, "ts": 1716454216832883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832874, "dur": 6,
    "args": {
      "External id": 13530,
      "cbid": 211, "correlation": 13530
    }
  },
  {
    "ph": "s", "id": 13530, "pid": 76337, "tid": -914061504, "ts": 1716454216832874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832884, "dur": 0,
    "args": {
      "External id": 13531,
      "cbid": 51, "correlation": 13531
    }
  },
  {
    "ph": "s", "id": 13531, "pid": 76337, "tid": -914061504, "ts": 1716454216832884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216832892, "dur": 93,
    "args": {
      "External id": 13532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13532,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 13532, "pid": 5, "tid": 7, "ts": 1716454216832892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832885, "dur": 6,
    "args": {
      "External id": 13532,
      "cbid": 211, "correlation": 13532
    }
  },
  {
    "ph": "s", "id": 13532, "pid": 76337, "tid": -914061504, "ts": 1716454216832885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216832986, "dur": 15,
    "args": {
      "External id": 13537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13537,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13537, "pid": 5, "tid": 7, "ts": 1716454216832986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216832914, "dur": 9,
    "args": {
      "External id": 13537,
      "cbid": 211, "correlation": 13537
    }
  },
  {
    "ph": "s", "id": 13537, "pid": 76337, "tid": -914061504, "ts": 1716454216832914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216833088, "dur": 44,
    "args": {
      "External id": 13548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13548, "pid": 5, "tid": 7, "ts": 1716454216833088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833071, "dur": 19,
    "args": {
      "External id": 13548,
      "cbid": 211, "correlation": 13548
    }
  },
  {
    "ph": "s", "id": 13548, "pid": 76337, "tid": -914061504, "ts": 1716454216833071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216833133, "dur": 18,
    "args": {
      "External id": 13570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13570,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13570, "pid": 5, "tid": 7, "ts": 1716454216833133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833113, "dur": 10,
    "args": {
      "External id": 13570,
      "cbid": 211, "correlation": 13570
    }
  },
  {
    "ph": "s", "id": 13570, "pid": 76337, "tid": -914061504, "ts": 1716454216833113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833285, "dur": 3,
    "args": {
      "External id": 13581,
      "cbid": 251, "correlation": 13581
    }
  },
  {
    "ph": "f", "id": 13581, "pid": 76337, "tid": -914061504, "ts": 1716454216833285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216833313, "dur": 91,
    "args": {
      "External id": 13582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13582,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13582, "pid": 5, "tid": 7, "ts": 1716454216833313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833296, "dur": 18,
    "args": {
      "External id": 13582,
      "cbid": 211, "correlation": 13582
    }
  },
  {
    "ph": "s", "id": 13582, "pid": 76337, "tid": -914061504, "ts": 1716454216833296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833393, "dur": 1,
    "args": {
      "External id": 13593,
      "cbid": 251, "correlation": 13593
    }
  },
  {
    "ph": "f", "id": 13593, "pid": 76337, "tid": -914061504, "ts": 1716454216833393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216833412, "dur": 86,
    "args": {
      "External id": 13594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13594,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13594, "pid": 5, "tid": 7, "ts": 1716454216833412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833398, "dur": 13,
    "args": {
      "External id": 13594,
      "cbid": 211, "correlation": 13594
    }
  },
  {
    "ph": "s", "id": 13594, "pid": 76337, "tid": -914061504, "ts": 1716454216833398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833478, "dur": 1,
    "args": {
      "External id": 13605,
      "cbid": 251, "correlation": 13605
    }
  },
  {
    "ph": "f", "id": 13605, "pid": 76337, "tid": -914061504, "ts": 1716454216833478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216833499, "dur": 85,
    "args": {
      "External id": 13606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13606,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13606, "pid": 5, "tid": 7, "ts": 1716454216833499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833482, "dur": 13,
    "args": {
      "External id": 13606,
      "cbid": 211, "correlation": 13606
    }
  },
  {
    "ph": "s", "id": 13606, "pid": 76337, "tid": -914061504, "ts": 1716454216833482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216833619, "dur": 140,
    "args": {
      "External id": 13631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13631,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13631, "pid": 5, "tid": 7, "ts": 1716454216833619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833603, "dur": 16,
    "args": {
      "External id": 13631,
      "cbid": 211, "correlation": 13631
    }
  },
  {
    "ph": "s", "id": 13631, "pid": 76337, "tid": -914061504, "ts": 1716454216833603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833748, "dur": 2,
    "args": {
      "External id": 13649,
      "cbid": 251, "correlation": 13649
    }
  },
  {
    "ph": "f", "id": 13649, "pid": 76337, "tid": -914061504, "ts": 1716454216833748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216833770, "dur": 92,
    "args": {
      "External id": 13651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13651,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13651, "pid": 5, "tid": 7, "ts": 1716454216833770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833755, "dur": 16,
    "args": {
      "External id": 13651,
      "cbid": 211, "correlation": 13651
    }
  },
  {
    "ph": "s", "id": 13651, "pid": 76337, "tid": -914061504, "ts": 1716454216833755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216833864, "dur": 10,
    "args": {
      "External id": 13659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13659,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13659, "pid": 5, "tid": 7, "ts": 1716454216833864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833842, "dur": 13,
    "args": {
      "External id": 13659,
      "cbid": 211, "correlation": 13659
    }
  },
  {
    "ph": "s", "id": 13659, "pid": 76337, "tid": -914061504, "ts": 1716454216833842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216833895, "dur": 46,
    "args": {
      "External id": 13667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13667, "pid": 5, "tid": 7, "ts": 1716454216833895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833884, "dur": 10,
    "args": {
      "External id": 13667,
      "cbid": 211, "correlation": 13667
    }
  },
  {
    "ph": "s", "id": 13667, "pid": 76337, "tid": -914061504, "ts": 1716454216833884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216833964, "dur": 17,
    "args": {
      "External id": 13689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13689,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13689, "pid": 5, "tid": 7, "ts": 1716454216833964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216833952, "dur": 11,
    "args": {
      "External id": 13689,
      "cbid": 211, "correlation": 13689
    }
  },
  {
    "ph": "s", "id": 13689, "pid": 76337, "tid": -914061504, "ts": 1716454216833952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834061, "dur": 1,
    "args": {
      "External id": 13700,
      "cbid": 251, "correlation": 13700
    }
  },
  {
    "ph": "f", "id": 13700, "pid": 76337, "tid": -914061504, "ts": 1716454216834061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216834081, "dur": 86,
    "args": {
      "External id": 13701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13701,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13701, "pid": 5, "tid": 7, "ts": 1716454216834081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834066, "dur": 14,
    "args": {
      "External id": 13701,
      "cbid": 211, "correlation": 13701
    }
  },
  {
    "ph": "s", "id": 13701, "pid": 76337, "tid": -914061504, "ts": 1716454216834066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834160, "dur": 2,
    "args": {
      "External id": 13712,
      "cbid": 251, "correlation": 13712
    }
  },
  {
    "ph": "f", "id": 13712, "pid": 76337, "tid": -914061504, "ts": 1716454216834160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834165, "dur": 0,
    "args": {
      "External id": 13713,
      "cbid": 251, "correlation": 13713
    }
  },
  {
    "ph": "f", "id": 13713, "pid": 76337, "tid": -914061504, "ts": 1716454216834165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216834181, "dur": 12,
    "args": {
      "External id": 13714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13714,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13714, "pid": 5, "tid": 7, "ts": 1716454216834181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834167, "dur": 14,
    "args": {
      "External id": 13714,
      "cbid": 211, "correlation": 13714
    }
  },
  {
    "ph": "s", "id": 13714, "pid": 76337, "tid": -914061504, "ts": 1716454216834167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216834195, "dur": 5,
    "args": {
      "External id": 13716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13716,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13716, "pid": 5, "tid": 7, "ts": 1716454216834195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834184, "dur": 9,
    "args": {
      "External id": 13716,
      "cbid": 211, "correlation": 13716
    }
  },
  {
    "ph": "s", "id": 13716, "pid": 76337, "tid": -914061504, "ts": 1716454216834184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834246, "dur": 1,
    "args": {
      "External id": 13727,
      "cbid": 251, "correlation": 13727
    }
  },
  {
    "ph": "f", "id": 13727, "pid": 76337, "tid": -914061504, "ts": 1716454216834246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834250, "dur": 0,
    "args": {
      "External id": 13728,
      "cbid": 251, "correlation": 13728
    }
  },
  {
    "ph": "f", "id": 13728, "pid": 76337, "tid": -914061504, "ts": 1716454216834250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216834264, "dur": 8,
    "args": {
      "External id": 13729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13729,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13729, "pid": 5, "tid": 7, "ts": 1716454216834264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834251, "dur": 12,
    "args": {
      "External id": 13729,
      "cbid": 211, "correlation": 13729
    }
  },
  {
    "ph": "s", "id": 13729, "pid": 76337, "tid": -914061504, "ts": 1716454216834251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216834274, "dur": 3,
    "args": {
      "External id": 13731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13731,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13731, "pid": 5, "tid": 7, "ts": 1716454216834274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834265, "dur": 6,
    "args": {
      "External id": 13731,
      "cbid": 211, "correlation": 13731
    }
  },
  {
    "ph": "s", "id": 13731, "pid": 76337, "tid": -914061504, "ts": 1716454216834265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216834356, "dur": 57,
    "args": {
      "External id": 13756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13756,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13756, "pid": 5, "tid": 7, "ts": 1716454216834356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834343, "dur": 12,
    "args": {
      "External id": 13756,
      "cbid": 211, "correlation": 13756
    }
  },
  {
    "ph": "s", "id": 13756, "pid": 76337, "tid": -914061504, "ts": 1716454216834343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834460, "dur": 1,
    "args": {
      "External id": 13774,
      "cbid": 251, "correlation": 13774
    }
  },
  {
    "ph": "f", "id": 13774, "pid": 76337, "tid": -914061504, "ts": 1716454216834460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216834480, "dur": 87,
    "args": {
      "External id": 13776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13776,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13776, "pid": 5, "tid": 7, "ts": 1716454216834480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834466, "dur": 14,
    "args": {
      "External id": 13776,
      "cbid": 211, "correlation": 13776
    }
  },
  {
    "ph": "s", "id": 13776, "pid": 76337, "tid": -914061504, "ts": 1716454216834466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216834569, "dur": 10,
    "args": {
      "External id": 13784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13784,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13784, "pid": 5, "tid": 7, "ts": 1716454216834569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834539, "dur": 12,
    "args": {
      "External id": 13784,
      "cbid": 211, "correlation": 13784
    }
  },
  {
    "ph": "s", "id": 13784, "pid": 76337, "tid": -914061504, "ts": 1716454216834539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216834589, "dur": 13,
    "args": {
      "External id": 13792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13792,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13792, "pid": 5, "tid": 7, "ts": 1716454216834589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834579, "dur": 9,
    "args": {
      "External id": 13792,
      "cbid": 211, "correlation": 13792
    }
  },
  {
    "ph": "s", "id": 13792, "pid": 76337, "tid": -914061504, "ts": 1716454216834579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216834654, "dur": 17,
    "args": {
      "External id": 13814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13814,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13814, "pid": 5, "tid": 7, "ts": 1716454216834654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834642, "dur": 12,
    "args": {
      "External id": 13814,
      "cbid": 211, "correlation": 13814
    }
  },
  {
    "ph": "s", "id": 13814, "pid": 76337, "tid": -914061504, "ts": 1716454216834642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834747, "dur": 2,
    "args": {
      "External id": 13830,
      "cbid": 251, "correlation": 13830
    }
  },
  {
    "ph": "f", "id": 13830, "pid": 76337, "tid": -914061504, "ts": 1716454216834747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834753, "dur": 0,
    "args": {
      "External id": 13832,
      "cbid": 251, "correlation": 13832
    }
  },
  {
    "ph": "f", "id": 13832, "pid": 76337, "tid": -914061504, "ts": 1716454216834753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216834771, "dur": 495,
    "args": {
      "External id": 13833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13833,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13833, "pid": 5, "tid": 7, "ts": 1716454216834771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834756, "dur": 15,
    "args": {
      "External id": 13833,
      "cbid": 211, "correlation": 13833
    }
  },
  {
    "ph": "s", "id": 13833, "pid": 76337, "tid": -914061504, "ts": 1716454216834756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216835268, "dur": 66,
    "args": {
      "External id": 13841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13841,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13841, "pid": 5, "tid": 7, "ts": 1716454216835268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834854, "dur": 15,
    "args": {
      "External id": 13841,
      "cbid": 211, "correlation": 13841
    }
  },
  {
    "ph": "s", "id": 13841, "pid": 76337, "tid": -914061504, "ts": 1716454216834854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216835335, "dur": 68,
    "args": {
      "External id": 13849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13849, "pid": 5, "tid": 7, "ts": 1716454216835335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834896, "dur": 11,
    "args": {
      "External id": 13849,
      "cbid": 211, "correlation": 13849
    }
  },
  {
    "ph": "s", "id": 13849, "pid": 76337, "tid": -914061504, "ts": 1716454216834896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834989, "dur": 1,
    "args": {
      "External id": 13865,
      "cbid": 251, "correlation": 13865
    }
  },
  {
    "ph": "f", "id": 13865, "pid": 76337, "tid": -914061504, "ts": 1716454216834989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216835405, "dur": 1,
    "args": {
      "External id": 13867,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13867,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 13867, "pid": 5, "tid": 7, "ts": 1716454216835405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216834994, "dur": 15,
    "args": {
      "External id": 13867,
      "cbid": 51, "correlation": 13867
    }
  },
  {
    "ph": "s", "id": 13867, "pid": 76337, "tid": -914061504, "ts": 1716454216834994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216835409, "dur": 270,
    "args": {
      "External id": 13868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13868,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13868, "pid": 5, "tid": 7, "ts": 1716454216835409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835011, "dur": 12,
    "args": {
      "External id": 13868,
      "cbid": 211, "correlation": 13868
    }
  },
  {
    "ph": "s", "id": 13868, "pid": 76337, "tid": -914061504, "ts": 1716454216835011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216835680, "dur": 14,
    "args": {
      "External id": 13876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13876,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13876, "pid": 5, "tid": 7, "ts": 1716454216835680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835056, "dur": 11,
    "args": {
      "External id": 13876,
      "cbid": 211, "correlation": 13876
    }
  },
  {
    "ph": "s", "id": 13876, "pid": 76337, "tid": -914061504, "ts": 1716454216835056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216835695, "dur": 38,
    "args": {
      "External id": 13887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13887, "pid": 5, "tid": 7, "ts": 1716454216835695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835127, "dur": 12,
    "args": {
      "External id": 13887,
      "cbid": 211, "correlation": 13887
    }
  },
  {
    "ph": "s", "id": 13887, "pid": 76337, "tid": -914061504, "ts": 1716454216835127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835195, "dur": 0,
    "args": {
      "External id": 13899,
      "cbid": 317, "correlation": 13899
    }
  },
  {
    "ph": "f", "id": 13899, "pid": 76337, "tid": -914061504, "ts": 1716454216835195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835196, "dur": 0,
    "args": {
      "External id": 13900,
      "cbid": 203, "correlation": 13900
    }
  },
  {
    "ph": "f", "id": 13900, "pid": 76337, "tid": -914061504, "ts": 1716454216835196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835197, "dur": 0,
    "args": {
      "External id": 13901,
      "cbid": 205, "correlation": 13901
    }
  },
  {
    "ph": "f", "id": 13901, "pid": 76337, "tid": -914061504, "ts": 1716454216835197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216835734, "dur": 13,
    "args": {
      "External id": 13905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13905,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13905, "pid": 5, "tid": 7, "ts": 1716454216835734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835211, "dur": 12,
    "args": {
      "External id": 13905,
      "cbid": 211, "correlation": 13905
    }
  },
  {
    "ph": "s", "id": 13905, "pid": 76337, "tid": -914061504, "ts": 1716454216835211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216835749, "dur": 4,
    "args": {
      "External id": 13907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13907,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 13907, "pid": 5, "tid": 7, "ts": 1716454216835749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835229, "dur": 6,
    "args": {
      "External id": 13907,
      "cbid": 211, "correlation": 13907
    }
  },
  {
    "ph": "s", "id": 13907, "pid": 76337, "tid": -914061504, "ts": 1716454216835229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835238, "dur": 0,
    "args": {
      "External id": 13908,
      "cbid": 51, "correlation": 13908
    }
  },
  {
    "ph": "s", "id": 13908, "pid": 76337, "tid": -914061504, "ts": 1716454216835238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216835754, "dur": 100,
    "args": {
      "External id": 13909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13909,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 13909, "pid": 5, "tid": 7, "ts": 1716454216835754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835239, "dur": 6,
    "args": {
      "External id": 13909,
      "cbid": 211, "correlation": 13909
    }
  },
  {
    "ph": "s", "id": 13909, "pid": 76337, "tid": -914061504, "ts": 1716454216835239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216835855, "dur": 17,
    "args": {
      "External id": 13914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13914, "pid": 5, "tid": 7, "ts": 1716454216835855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835267, "dur": 9,
    "args": {
      "External id": 13914,
      "cbid": 211, "correlation": 13914
    }
  },
  {
    "ph": "s", "id": 13914, "pid": 76337, "tid": -914061504, "ts": 1716454216835267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216835873, "dur": 12,
    "args": {
      "External id": 13922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13922,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13922, "pid": 5, "tid": 7, "ts": 1716454216835873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835300, "dur": 8,
    "args": {
      "External id": 13922,
      "cbid": 211, "correlation": 13922
    }
  },
  {
    "ph": "s", "id": 13922, "pid": 76337, "tid": -914061504, "ts": 1716454216835300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216835887, "dur": 31,
    "args": {
      "External id": 13931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13931,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13931, "pid": 5, "tid": 7, "ts": 1716454216835887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835357, "dur": 13,
    "args": {
      "External id": 13931,
      "cbid": 211, "correlation": 13931
    }
  },
  {
    "ph": "s", "id": 13931, "pid": 76337, "tid": -914061504, "ts": 1716454216835357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216835918, "dur": 31,
    "args": {
      "External id": 13951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13951,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 13951, "pid": 5, "tid": 7, "ts": 1716454216835918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835450, "dur": 13,
    "args": {
      "External id": 13951,
      "cbid": 211, "correlation": 13951
    }
  },
  {
    "ph": "s", "id": 13951, "pid": 76337, "tid": -914061504, "ts": 1716454216835450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216835950, "dur": 5,
    "args": {
      "External id": 13963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13963,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13963, "pid": 5, "tid": 7, "ts": 1716454216835950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835473, "dur": 6,
    "args": {
      "External id": 13963,
      "cbid": 211, "correlation": 13963
    }
  },
  {
    "ph": "s", "id": 13963, "pid": 76337, "tid": -914061504, "ts": 1716454216835473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216835956, "dur": 31,
    "args": {
      "External id": 13966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13966,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13966, "pid": 5, "tid": 7, "ts": 1716454216835956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835492, "dur": 8,
    "args": {
      "External id": 13966,
      "cbid": 211, "correlation": 13966
    }
  },
  {
    "ph": "s", "id": 13966, "pid": 76337, "tid": -914061504, "ts": 1716454216835492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216835989, "dur": 21,
    "args": {
      "External id": 13975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13975,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13975, "pid": 5, "tid": 7, "ts": 1716454216835989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835538, "dur": 11,
    "args": {
      "External id": 13975,
      "cbid": 211, "correlation": 13975
    }
  },
  {
    "ph": "s", "id": 13975, "pid": 76337, "tid": -914061504, "ts": 1716454216835538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835606, "dur": 0,
    "args": {
      "External id": 13985,
      "cbid": 317, "correlation": 13985
    }
  },
  {
    "ph": "f", "id": 13985, "pid": 76337, "tid": -914061504, "ts": 1716454216835606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835607, "dur": 0,
    "args": {
      "External id": 13986,
      "cbid": 203, "correlation": 13986
    }
  },
  {
    "ph": "f", "id": 13986, "pid": 76337, "tid": -914061504, "ts": 1716454216835607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835608, "dur": 0,
    "args": {
      "External id": 13987,
      "cbid": 205, "correlation": 13987
    }
  },
  {
    "ph": "f", "id": 13987, "pid": 76337, "tid": -914061504, "ts": 1716454216835608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216836011, "dur": 23,
    "args": {
      "External id": 13991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13991,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13991, "pid": 5, "tid": 7, "ts": 1716454216836011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835623, "dur": 12,
    "args": {
      "External id": 13991,
      "cbid": 211, "correlation": 13991
    }
  },
  {
    "ph": "s", "id": 13991, "pid": 76337, "tid": -914061504, "ts": 1716454216835623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216836036, "dur": 321,
    "args": {
      "External id": 13993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13993,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13993, "pid": 5, "tid": 7, "ts": 1716454216836036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835638, "dur": 5,
    "args": {
      "External id": 13993,
      "cbid": 211, "correlation": 13993
    }
  },
  {
    "ph": "s", "id": 13993, "pid": 76337, "tid": -914061504, "ts": 1716454216835638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216836359, "dur": 1,
    "args": {
      "External id": 13995,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 13995,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 13995, "pid": 5, "tid": 7, "ts": 1716454216836359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835652, "dur": 8,
    "args": {
      "External id": 13995,
      "cbid": 51, "correlation": 13995
    }
  },
  {
    "ph": "s", "id": 13995, "pid": 76337, "tid": -914061504, "ts": 1716454216835652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216836363, "dur": 1264,
    "args": {
      "External id": 13996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13996,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 13996, "pid": 5, "tid": 7, "ts": 1716454216836363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835662, "dur": 7,
    "args": {
      "External id": 13996,
      "cbid": 211, "correlation": 13996
    }
  },
  {
    "ph": "s", "id": 13996, "pid": 76337, "tid": -914061504, "ts": 1716454216835662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216837628, "dur": 13,
    "args": {
      "External id": 13998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 13998,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 13998, "pid": 5, "tid": 7, "ts": 1716454216837628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835675, "dur": 6,
    "args": {
      "External id": 13998,
      "cbid": 211, "correlation": 13998
    }
  },
  {
    "ph": "s", "id": 13998, "pid": 76337, "tid": -914061504, "ts": 1716454216835675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216837642, "dur": 15,
    "args": {
      "External id": 14004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14004, "pid": 5, "tid": 7, "ts": 1716454216837642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835705, "dur": 8,
    "args": {
      "External id": 14004,
      "cbid": 211, "correlation": 14004
    }
  },
  {
    "ph": "s", "id": 14004, "pid": 76337, "tid": -914061504, "ts": 1716454216835705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216837658, "dur": 3,
    "args": {
      "External id": 14012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14012,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 14012, "pid": 5, "tid": 7, "ts": 1716454216837658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835750, "dur": 10,
    "args": {
      "External id": 14012,
      "cbid": 211, "correlation": 14012
    }
  },
  {
    "ph": "s", "id": 14012, "pid": 76337, "tid": -914061504, "ts": 1716454216835750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835817, "dur": 1,
    "args": {
      "External id": 14028,
      "cbid": 251, "correlation": 14028
    }
  },
  {
    "ph": "f", "id": 14028, "pid": 76337, "tid": -914061504, "ts": 1716454216835817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835822, "dur": 0,
    "args": {
      "External id": 14030,
      "cbid": 251, "correlation": 14030
    }
  },
  {
    "ph": "f", "id": 14030, "pid": 76337, "tid": -914061504, "ts": 1716454216835822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216837663, "dur": 14,
    "args": {
      "External id": 14031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14031,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14031, "pid": 5, "tid": 7, "ts": 1716454216837663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835824, "dur": 11,
    "args": {
      "External id": 14031,
      "cbid": 211, "correlation": 14031
    }
  },
  {
    "ph": "s", "id": 14031, "pid": 76337, "tid": -914061504, "ts": 1716454216835824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216837678, "dur": 5,
    "args": {
      "External id": 14033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14033,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14033, "pid": 5, "tid": 7, "ts": 1716454216837678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835838, "dur": 6,
    "args": {
      "External id": 14033,
      "cbid": 211, "correlation": 14033
    }
  },
  {
    "ph": "s", "id": 14033, "pid": 76337, "tid": -914061504, "ts": 1716454216835838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216837685, "dur": 18,
    "args": {
      "External id": 14043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14043,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14043, "pid": 5, "tid": 7, "ts": 1716454216837685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835903, "dur": 12,
    "args": {
      "External id": 14043,
      "cbid": 211, "correlation": 14043
    }
  },
  {
    "ph": "s", "id": 14043, "pid": 76337, "tid": -914061504, "ts": 1716454216835903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216837704, "dur": 18,
    "args": {
      "External id": 14063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14063,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 14063, "pid": 5, "tid": 7, "ts": 1716454216837704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216835969, "dur": 19,
    "args": {
      "External id": 14063,
      "cbid": 211, "correlation": 14063
    }
  },
  {
    "ph": "s", "id": 14063, "pid": 76337, "tid": -914061504, "ts": 1716454216835969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216837723, "dur": 4,
    "args": {
      "External id": 14075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14075,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 14075, "pid": 5, "tid": 7, "ts": 1716454216837723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836011, "dur": 8,
    "args": {
      "External id": 14075,
      "cbid": 211, "correlation": 14075
    }
  },
  {
    "ph": "s", "id": 14075, "pid": 76337, "tid": -914061504, "ts": 1716454216836011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216837729, "dur": 17,
    "args": {
      "External id": 14078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14078,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14078, "pid": 5, "tid": 7, "ts": 1716454216837729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836033, "dur": 7,
    "args": {
      "External id": 14078,
      "cbid": 211, "correlation": 14078
    }
  },
  {
    "ph": "s", "id": 14078, "pid": 76337, "tid": -914061504, "ts": 1716454216836033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216837747, "dur": 12,
    "args": {
      "External id": 14087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14087,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14087, "pid": 5, "tid": 7, "ts": 1716454216837747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836077, "dur": 10,
    "args": {
      "External id": 14087,
      "cbid": 211, "correlation": 14087
    }
  },
  {
    "ph": "s", "id": 14087, "pid": 76337, "tid": -914061504, "ts": 1716454216836077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836141, "dur": 0,
    "args": {
      "External id": 14097,
      "cbid": 317, "correlation": 14097
    }
  },
  {
    "ph": "f", "id": 14097, "pid": 76337, "tid": -914061504, "ts": 1716454216836141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836141, "dur": 0,
    "args": {
      "External id": 14098,
      "cbid": 203, "correlation": 14098
    }
  },
  {
    "ph": "f", "id": 14098, "pid": 76337, "tid": -914061504, "ts": 1716454216836141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836142, "dur": 0,
    "args": {
      "External id": 14099,
      "cbid": 205, "correlation": 14099
    }
  },
  {
    "ph": "f", "id": 14099, "pid": 76337, "tid": -914061504, "ts": 1716454216836142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216837760, "dur": 12,
    "args": {
      "External id": 14103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14103,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14103, "pid": 5, "tid": 7, "ts": 1716454216837760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836158, "dur": 13,
    "args": {
      "External id": 14103,
      "cbid": 211, "correlation": 14103
    }
  },
  {
    "ph": "s", "id": 14103, "pid": 76337, "tid": -914061504, "ts": 1716454216836158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216837773, "dur": 166,
    "args": {
      "External id": 14105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14105,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14105, "pid": 5, "tid": 7, "ts": 1716454216837773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836173, "dur": 5,
    "args": {
      "External id": 14105,
      "cbid": 211, "correlation": 14105
    }
  },
  {
    "ph": "s", "id": 14105, "pid": 76337, "tid": -914061504, "ts": 1716454216836173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216837941, "dur": 1,
    "args": {
      "External id": 14107,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 14107,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 14107, "pid": 5, "tid": 7, "ts": 1716454216837941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836184, "dur": 7,
    "args": {
      "External id": 14107,
      "cbid": 51, "correlation": 14107
    }
  },
  {
    "ph": "s", "id": 14107, "pid": 76337, "tid": -914061504, "ts": 1716454216836184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216837945, "dur": 653,
    "args": {
      "External id": 14108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14108,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14108, "pid": 5, "tid": 7, "ts": 1716454216837945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836192, "dur": 6,
    "args": {
      "External id": 14108,
      "cbid": 211, "correlation": 14108
    }
  },
  {
    "ph": "s", "id": 14108, "pid": 76337, "tid": -914061504, "ts": 1716454216836192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216838599, "dur": 13,
    "args": {
      "External id": 14110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14110, "pid": 5, "tid": 7, "ts": 1716454216838599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836202, "dur": 5,
    "args": {
      "External id": 14110,
      "cbid": 211, "correlation": 14110
    }
  },
  {
    "ph": "s", "id": 14110, "pid": 76337, "tid": -914061504, "ts": 1716454216836202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216838614, "dur": 15,
    "args": {
      "External id": 14116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14116,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14116, "pid": 5, "tid": 7, "ts": 1716454216838614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836231, "dur": 8,
    "args": {
      "External id": 14116,
      "cbid": 211, "correlation": 14116
    }
  },
  {
    "ph": "s", "id": 14116, "pid": 76337, "tid": -914061504, "ts": 1716454216836231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836291, "dur": 0,
    "args": {
      "External id": 14126,
      "cbid": 317, "correlation": 14126
    }
  },
  {
    "ph": "f", "id": 14126, "pid": 76337, "tid": -914061504, "ts": 1716454216836291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836291, "dur": 0,
    "args": {
      "External id": 14127,
      "cbid": 203, "correlation": 14127
    }
  },
  {
    "ph": "f", "id": 14127, "pid": 76337, "tid": -914061504, "ts": 1716454216836291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836292, "dur": 0,
    "args": {
      "External id": 14128,
      "cbid": 205, "correlation": 14128
    }
  },
  {
    "ph": "f", "id": 14128, "pid": 76337, "tid": -914061504, "ts": 1716454216836292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216838630, "dur": 22,
    "args": {
      "External id": 14132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14132, "pid": 5, "tid": 7, "ts": 1716454216838630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836303, "dur": 12,
    "args": {
      "External id": 14132,
      "cbid": 211, "correlation": 14132
    }
  },
  {
    "ph": "s", "id": 14132, "pid": 76337, "tid": -914061504, "ts": 1716454216836303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216838653, "dur": 4,
    "args": {
      "External id": 14134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14134, "pid": 5, "tid": 7, "ts": 1716454216838653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836320, "dur": 6,
    "args": {
      "External id": 14134,
      "cbid": 211, "correlation": 14134
    }
  },
  {
    "ph": "s", "id": 14134, "pid": 76337, "tid": -914061504, "ts": 1716454216836320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836328, "dur": 0,
    "args": {
      "External id": 14135,
      "cbid": 51, "correlation": 14135
    }
  },
  {
    "ph": "s", "id": 14135, "pid": 76337, "tid": -914061504, "ts": 1716454216836328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216838659, "dur": 174,
    "args": {
      "External id": 14136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14136,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 14136, "pid": 5, "tid": 7, "ts": 1716454216838659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836329, "dur": 5,
    "args": {
      "External id": 14136,
      "cbid": 211, "correlation": 14136
    }
  },
  {
    "ph": "s", "id": 14136, "pid": 76337, "tid": -914061504, "ts": 1716454216836329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216838834, "dur": 16,
    "args": {
      "External id": 14141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14141, "pid": 5, "tid": 7, "ts": 1716454216838834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836353, "dur": 9,
    "args": {
      "External id": 14141,
      "cbid": 211, "correlation": 14141
    }
  },
  {
    "ph": "s", "id": 14141, "pid": 76337, "tid": -914061504, "ts": 1716454216836353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216838852, "dur": 12,
    "args": {
      "External id": 14149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14149,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14149, "pid": 5, "tid": 7, "ts": 1716454216838852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836383, "dur": 8,
    "args": {
      "External id": 14149,
      "cbid": 211, "correlation": 14149
    }
  },
  {
    "ph": "s", "id": 14149, "pid": 76337, "tid": -914061504, "ts": 1716454216836383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216838865, "dur": 10,
    "args": {
      "External id": 14157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14157,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14157, "pid": 5, "tid": 7, "ts": 1716454216838865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836412, "dur": 9,
    "args": {
      "External id": 14157,
      "cbid": 211, "correlation": 14157
    }
  },
  {
    "ph": "s", "id": 14157, "pid": 76337, "tid": -914061504, "ts": 1716454216836412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216838877, "dur": 18,
    "args": {
      "External id": 14177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14177,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 14177, "pid": 5, "tid": 7, "ts": 1716454216838877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836498, "dur": 13,
    "args": {
      "External id": 14177,
      "cbid": 211, "correlation": 14177
    }
  },
  {
    "ph": "s", "id": 14177, "pid": 76337, "tid": -914061504, "ts": 1716454216836498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216838896, "dur": 5,
    "args": {
      "External id": 14189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14189,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 14189, "pid": 5, "tid": 7, "ts": 1716454216838896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836521, "dur": 6,
    "args": {
      "External id": 14189,
      "cbid": 211, "correlation": 14189
    }
  },
  {
    "ph": "s", "id": 14189, "pid": 76337, "tid": -914061504, "ts": 1716454216836521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216838903, "dur": 17,
    "args": {
      "External id": 14192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14192, "pid": 5, "tid": 7, "ts": 1716454216838903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836539, "dur": 6,
    "args": {
      "External id": 14192,
      "cbid": 211, "correlation": 14192
    }
  },
  {
    "ph": "s", "id": 14192, "pid": 76337, "tid": -914061504, "ts": 1716454216836539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836610, "dur": 0,
    "args": {
      "External id": 14203,
      "cbid": 317, "correlation": 14203
    }
  },
  {
    "ph": "f", "id": 14203, "pid": 76337, "tid": -914061504, "ts": 1716454216836610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836611, "dur": 0,
    "args": {
      "External id": 14204,
      "cbid": 203, "correlation": 14204
    }
  },
  {
    "ph": "f", "id": 14204, "pid": 76337, "tid": -914061504, "ts": 1716454216836611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836612, "dur": 0,
    "args": {
      "External id": 14205,
      "cbid": 205, "correlation": 14205
    }
  },
  {
    "ph": "f", "id": 14205, "pid": 76337, "tid": -914061504, "ts": 1716454216836612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216838921, "dur": 11,
    "args": {
      "External id": 14209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14209,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14209, "pid": 5, "tid": 7, "ts": 1716454216838921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836625, "dur": 13,
    "args": {
      "External id": 14209,
      "cbid": 211, "correlation": 14209
    }
  },
  {
    "ph": "s", "id": 14209, "pid": 76337, "tid": -914061504, "ts": 1716454216836625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216838934, "dur": 3,
    "args": {
      "External id": 14211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14211, "pid": 5, "tid": 7, "ts": 1716454216838934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836642, "dur": 6,
    "args": {
      "External id": 14211,
      "cbid": 211, "correlation": 14211
    }
  },
  {
    "ph": "s", "id": 14211, "pid": 76337, "tid": -914061504, "ts": 1716454216836642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836651, "dur": 0,
    "args": {
      "External id": 14212,
      "cbid": 51, "correlation": 14212
    }
  },
  {
    "ph": "s", "id": 14212, "pid": 76337, "tid": -914061504, "ts": 1716454216836651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216838938, "dur": 92,
    "args": {
      "External id": 14213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14213,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 14213, "pid": 5, "tid": 7, "ts": 1716454216838938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836651, "dur": 5,
    "args": {
      "External id": 14213,
      "cbid": 211, "correlation": 14213
    }
  },
  {
    "ph": "s", "id": 14213, "pid": 76337, "tid": -914061504, "ts": 1716454216836651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216839032, "dur": 15,
    "args": {
      "External id": 14218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14218,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14218, "pid": 5, "tid": 7, "ts": 1716454216839032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836679, "dur": 9,
    "args": {
      "External id": 14218,
      "cbid": 211, "correlation": 14218
    }
  },
  {
    "ph": "s", "id": 14218, "pid": 76337, "tid": -914061504, "ts": 1716454216836679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216839049, "dur": 43,
    "args": {
      "External id": 14229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14229,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14229, "pid": 5, "tid": 7, "ts": 1716454216839049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836778, "dur": 15,
    "args": {
      "External id": 14229,
      "cbid": 211, "correlation": 14229
    }
  },
  {
    "ph": "s", "id": 14229, "pid": 76337, "tid": -914061504, "ts": 1716454216836778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216839093, "dur": 19,
    "args": {
      "External id": 14251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14251,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14251, "pid": 5, "tid": 7, "ts": 1716454216839093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836812, "dur": 8,
    "args": {
      "External id": 14251,
      "cbid": 211, "correlation": 14251
    }
  },
  {
    "ph": "s", "id": 14251, "pid": 76337, "tid": -914061504, "ts": 1716454216836812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836915, "dur": 2,
    "args": {
      "External id": 14262,
      "cbid": 251, "correlation": 14262
    }
  },
  {
    "ph": "f", "id": 14262, "pid": 76337, "tid": -914061504, "ts": 1716454216836915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216839113, "dur": 90,
    "args": {
      "External id": 14263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14263,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14263, "pid": 5, "tid": 7, "ts": 1716454216839113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216836921, "dur": 14,
    "args": {
      "External id": 14263,
      "cbid": 211, "correlation": 14263
    }
  },
  {
    "ph": "s", "id": 14263, "pid": 76337, "tid": -914061504, "ts": 1716454216836921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837020, "dur": 1,
    "args": {
      "External id": 14274,
      "cbid": 251, "correlation": 14274
    }
  },
  {
    "ph": "f", "id": 14274, "pid": 76337, "tid": -914061504, "ts": 1716454216837020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216839205, "dur": 85,
    "args": {
      "External id": 14275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14275,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14275, "pid": 5, "tid": 7, "ts": 1716454216839205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837024, "dur": 13,
    "args": {
      "External id": 14275,
      "cbid": 211, "correlation": 14275
    }
  },
  {
    "ph": "s", "id": 14275, "pid": 76337, "tid": -914061504, "ts": 1716454216837024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837104, "dur": 1,
    "args": {
      "External id": 14286,
      "cbid": 251, "correlation": 14286
    }
  },
  {
    "ph": "f", "id": 14286, "pid": 76337, "tid": -914061504, "ts": 1716454216837104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216839292, "dur": 86,
    "args": {
      "External id": 14287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14287,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14287, "pid": 5, "tid": 7, "ts": 1716454216839292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837108, "dur": 13,
    "args": {
      "External id": 14287,
      "cbid": 211, "correlation": 14287
    }
  },
  {
    "ph": "s", "id": 14287, "pid": 76337, "tid": -914061504, "ts": 1716454216837108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216839379, "dur": 139,
    "args": {
      "External id": 14312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14312,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14312, "pid": 5, "tid": 7, "ts": 1716454216839379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837194, "dur": 14,
    "args": {
      "External id": 14312,
      "cbid": 211, "correlation": 14312
    }
  },
  {
    "ph": "s", "id": 14312, "pid": 76337, "tid": -914061504, "ts": 1716454216837194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837296, "dur": 1,
    "args": {
      "External id": 14330,
      "cbid": 251, "correlation": 14330
    }
  },
  {
    "ph": "f", "id": 14330, "pid": 76337, "tid": -914061504, "ts": 1716454216837296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216839518, "dur": 92,
    "args": {
      "External id": 14332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14332,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14332, "pid": 5, "tid": 7, "ts": 1716454216839518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837302, "dur": 14,
    "args": {
      "External id": 14332,
      "cbid": 211, "correlation": 14332
    }
  },
  {
    "ph": "s", "id": 14332, "pid": 76337, "tid": -914061504, "ts": 1716454216837302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216839612, "dur": 10,
    "args": {
      "External id": 14340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14340,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14340, "pid": 5, "tid": 7, "ts": 1716454216839612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837374, "dur": 12,
    "args": {
      "External id": 14340,
      "cbid": 211, "correlation": 14340
    }
  },
  {
    "ph": "s", "id": 14340, "pid": 76337, "tid": -914061504, "ts": 1716454216837374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216839623, "dur": 46,
    "args": {
      "External id": 14348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14348,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14348, "pid": 5, "tid": 7, "ts": 1716454216839623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837415, "dur": 10,
    "args": {
      "External id": 14348,
      "cbid": 211, "correlation": 14348
    }
  },
  {
    "ph": "s", "id": 14348, "pid": 76337, "tid": -914061504, "ts": 1716454216837415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216839671, "dur": 19,
    "args": {
      "External id": 14370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14370,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14370, "pid": 5, "tid": 7, "ts": 1716454216839671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837481, "dur": 11,
    "args": {
      "External id": 14370,
      "cbid": 211, "correlation": 14370
    }
  },
  {
    "ph": "s", "id": 14370, "pid": 76337, "tid": -914061504, "ts": 1716454216837481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837570, "dur": 1,
    "args": {
      "External id": 14381,
      "cbid": 251, "correlation": 14381
    }
  },
  {
    "ph": "f", "id": 14381, "pid": 76337, "tid": -914061504, "ts": 1716454216837570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216839691, "dur": 86,
    "args": {
      "External id": 14382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14382,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14382, "pid": 5, "tid": 7, "ts": 1716454216839691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837575, "dur": 13,
    "args": {
      "External id": 14382,
      "cbid": 211, "correlation": 14382
    }
  },
  {
    "ph": "s", "id": 14382, "pid": 76337, "tid": -914061504, "ts": 1716454216837575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837659, "dur": 1,
    "args": {
      "External id": 14393,
      "cbid": 251, "correlation": 14393
    }
  },
  {
    "ph": "f", "id": 14393, "pid": 76337, "tid": -914061504, "ts": 1716454216837659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837663, "dur": 0,
    "args": {
      "External id": 14394,
      "cbid": 251, "correlation": 14394
    }
  },
  {
    "ph": "f", "id": 14394, "pid": 76337, "tid": -914061504, "ts": 1716454216837663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216839778, "dur": 12,
    "args": {
      "External id": 14395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14395,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14395, "pid": 5, "tid": 7, "ts": 1716454216839778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837665, "dur": 13,
    "args": {
      "External id": 14395,
      "cbid": 211, "correlation": 14395
    }
  },
  {
    "ph": "s", "id": 14395, "pid": 76337, "tid": -914061504, "ts": 1716454216837665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216839791, "dur": 5,
    "args": {
      "External id": 14397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14397,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14397, "pid": 5, "tid": 7, "ts": 1716454216839791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837680, "dur": 6,
    "args": {
      "External id": 14397,
      "cbid": 211, "correlation": 14397
    }
  },
  {
    "ph": "s", "id": 14397, "pid": 76337, "tid": -914061504, "ts": 1716454216837680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837740, "dur": 1,
    "args": {
      "External id": 14408,
      "cbid": 251, "correlation": 14408
    }
  },
  {
    "ph": "f", "id": 14408, "pid": 76337, "tid": -914061504, "ts": 1716454216837740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837743, "dur": 0,
    "args": {
      "External id": 14409,
      "cbid": 251, "correlation": 14409
    }
  },
  {
    "ph": "f", "id": 14409, "pid": 76337, "tid": -914061504, "ts": 1716454216837743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216839798, "dur": 8,
    "args": {
      "External id": 14410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14410,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14410, "pid": 5, "tid": 7, "ts": 1716454216839798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837745, "dur": 12,
    "args": {
      "External id": 14410,
      "cbid": 211, "correlation": 14410
    }
  },
  {
    "ph": "s", "id": 14410, "pid": 76337, "tid": -914061504, "ts": 1716454216837745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216839808, "dur": 3,
    "args": {
      "External id": 14412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14412,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14412, "pid": 5, "tid": 7, "ts": 1716454216839808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837758, "dur": 6,
    "args": {
      "External id": 14412,
      "cbid": 211, "correlation": 14412
    }
  },
  {
    "ph": "s", "id": 14412, "pid": 76337, "tid": -914061504, "ts": 1716454216837758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216839812, "dur": 56,
    "args": {
      "External id": 14437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14437,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14437, "pid": 5, "tid": 7, "ts": 1716454216839812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837835, "dur": 13,
    "args": {
      "External id": 14437,
      "cbid": 211, "correlation": 14437
    }
  },
  {
    "ph": "s", "id": 14437, "pid": 76337, "tid": -914061504, "ts": 1716454216837835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837934, "dur": 1,
    "args": {
      "External id": 14455,
      "cbid": 251, "correlation": 14455
    }
  },
  {
    "ph": "f", "id": 14455, "pid": 76337, "tid": -914061504, "ts": 1716454216837934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216839869, "dur": 87,
    "args": {
      "External id": 14457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14457,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14457, "pid": 5, "tid": 7, "ts": 1716454216839869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216837940, "dur": 14,
    "args": {
      "External id": 14457,
      "cbid": 211, "correlation": 14457
    }
  },
  {
    "ph": "s", "id": 14457, "pid": 76337, "tid": -914061504, "ts": 1716454216837940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216839957, "dur": 10,
    "args": {
      "External id": 14465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14465,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14465, "pid": 5, "tid": 7, "ts": 1716454216839957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838020, "dur": 12,
    "args": {
      "External id": 14465,
      "cbid": 211, "correlation": 14465
    }
  },
  {
    "ph": "s", "id": 14465, "pid": 76337, "tid": -914061504, "ts": 1716454216838020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216839968, "dur": 13,
    "args": {
      "External id": 14473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14473,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14473, "pid": 5, "tid": 7, "ts": 1716454216839968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838060, "dur": 9,
    "args": {
      "External id": 14473,
      "cbid": 211, "correlation": 14473
    }
  },
  {
    "ph": "s", "id": 14473, "pid": 76337, "tid": -914061504, "ts": 1716454216838060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216839981, "dur": 17,
    "args": {
      "External id": 14495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14495,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14495, "pid": 5, "tid": 7, "ts": 1716454216839981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838111, "dur": 10,
    "args": {
      "External id": 14495,
      "cbid": 211, "correlation": 14495
    }
  },
  {
    "ph": "s", "id": 14495, "pid": 76337, "tid": -914061504, "ts": 1716454216838111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838200, "dur": 1,
    "args": {
      "External id": 14511,
      "cbid": 251, "correlation": 14511
    }
  },
  {
    "ph": "f", "id": 14511, "pid": 76337, "tid": -914061504, "ts": 1716454216838200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838205, "dur": 0,
    "args": {
      "External id": 14513,
      "cbid": 251, "correlation": 14513
    }
  },
  {
    "ph": "f", "id": 14513, "pid": 76337, "tid": -914061504, "ts": 1716454216838205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216840000, "dur": 492,
    "args": {
      "External id": 14514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14514,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14514, "pid": 5, "tid": 7, "ts": 1716454216840000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838207, "dur": 13,
    "args": {
      "External id": 14514,
      "cbid": 211, "correlation": 14514
    }
  },
  {
    "ph": "s", "id": 14514, "pid": 76337, "tid": -914061504, "ts": 1716454216838207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216840493, "dur": 68,
    "args": {
      "External id": 14522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14522, "pid": 5, "tid": 7, "ts": 1716454216840493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838274, "dur": 12,
    "args": {
      "External id": 14522,
      "cbid": 211, "correlation": 14522
    }
  },
  {
    "ph": "s", "id": 14522, "pid": 76337, "tid": -914061504, "ts": 1716454216838274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216840562, "dur": 67,
    "args": {
      "External id": 14530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14530, "pid": 5, "tid": 7, "ts": 1716454216840562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838305, "dur": 9,
    "args": {
      "External id": 14530,
      "cbid": 211, "correlation": 14530
    }
  },
  {
    "ph": "s", "id": 14530, "pid": 76337, "tid": -914061504, "ts": 1716454216838305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838386, "dur": 1,
    "args": {
      "External id": 14546,
      "cbid": 251, "correlation": 14546
    }
  },
  {
    "ph": "f", "id": 14546, "pid": 76337, "tid": -914061504, "ts": 1716454216838386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216840632, "dur": 1,
    "args": {
      "External id": 14548,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 14548,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 14548, "pid": 5, "tid": 7, "ts": 1716454216840632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838391, "dur": 9,
    "args": {
      "External id": 14548,
      "cbid": 51, "correlation": 14548
    }
  },
  {
    "ph": "s", "id": 14548, "pid": 76337, "tid": -914061504, "ts": 1716454216838391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216840635, "dur": 271,
    "args": {
      "External id": 14549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14549,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14549, "pid": 5, "tid": 7, "ts": 1716454216840635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838401, "dur": 12,
    "args": {
      "External id": 14549,
      "cbid": 211, "correlation": 14549
    }
  },
  {
    "ph": "s", "id": 14549, "pid": 76337, "tid": -914061504, "ts": 1716454216838401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216840908, "dur": 13,
    "args": {
      "External id": 14557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14557,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14557, "pid": 5, "tid": 7, "ts": 1716454216840908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838444, "dur": 10,
    "args": {
      "External id": 14557,
      "cbid": 211, "correlation": 14557
    }
  },
  {
    "ph": "s", "id": 14557, "pid": 76337, "tid": -914061504, "ts": 1716454216838444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216840922, "dur": 38,
    "args": {
      "External id": 14568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14568, "pid": 5, "tid": 7, "ts": 1716454216840922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838509, "dur": 13,
    "args": {
      "External id": 14568,
      "cbid": 211, "correlation": 14568
    }
  },
  {
    "ph": "s", "id": 14568, "pid": 76337, "tid": -914061504, "ts": 1716454216838509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838574, "dur": 0,
    "args": {
      "External id": 14580,
      "cbid": 317, "correlation": 14580
    }
  },
  {
    "ph": "f", "id": 14580, "pid": 76337, "tid": -914061504, "ts": 1716454216838574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838575, "dur": 0,
    "args": {
      "External id": 14581,
      "cbid": 203, "correlation": 14581
    }
  },
  {
    "ph": "f", "id": 14581, "pid": 76337, "tid": -914061504, "ts": 1716454216838575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838575, "dur": 0,
    "args": {
      "External id": 14582,
      "cbid": 205, "correlation": 14582
    }
  },
  {
    "ph": "f", "id": 14582, "pid": 76337, "tid": -914061504, "ts": 1716454216838575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216840962, "dur": 14,
    "args": {
      "External id": 14586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14586,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14586, "pid": 5, "tid": 7, "ts": 1716454216840962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838590, "dur": 12,
    "args": {
      "External id": 14586,
      "cbid": 211, "correlation": 14586
    }
  },
  {
    "ph": "s", "id": 14586, "pid": 76337, "tid": -914061504, "ts": 1716454216838590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216840978, "dur": 4,
    "args": {
      "External id": 14588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14588,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14588, "pid": 5, "tid": 7, "ts": 1716454216840978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838606, "dur": 6,
    "args": {
      "External id": 14588,
      "cbid": 211, "correlation": 14588
    }
  },
  {
    "ph": "s", "id": 14588, "pid": 76337, "tid": -914061504, "ts": 1716454216838606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838615, "dur": 0,
    "args": {
      "External id": 14589,
      "cbid": 51, "correlation": 14589
    }
  },
  {
    "ph": "s", "id": 14589, "pid": 76337, "tid": -914061504, "ts": 1716454216838615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216840983, "dur": 99,
    "args": {
      "External id": 14590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14590,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 14590, "pid": 5, "tid": 7, "ts": 1716454216840983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838615, "dur": 5,
    "args": {
      "External id": 14590,
      "cbid": 211, "correlation": 14590
    }
  },
  {
    "ph": "s", "id": 14590, "pid": 76337, "tid": -914061504, "ts": 1716454216838615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216841084, "dur": 17,
    "args": {
      "External id": 14595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14595,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14595, "pid": 5, "tid": 7, "ts": 1716454216841084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838642, "dur": 8,
    "args": {
      "External id": 14595,
      "cbid": 211, "correlation": 14595
    }
  },
  {
    "ph": "s", "id": 14595, "pid": 76337, "tid": -914061504, "ts": 1716454216838642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216841102, "dur": 12,
    "args": {
      "External id": 14603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14603,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14603, "pid": 5, "tid": 7, "ts": 1716454216841102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838673, "dur": 9,
    "args": {
      "External id": 14603,
      "cbid": 211, "correlation": 14603
    }
  },
  {
    "ph": "s", "id": 14603, "pid": 76337, "tid": -914061504, "ts": 1716454216838673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216841115, "dur": 25,
    "args": {
      "External id": 14612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14612,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14612, "pid": 5, "tid": 7, "ts": 1716454216841115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838715, "dur": 10,
    "args": {
      "External id": 14612,
      "cbid": 211, "correlation": 14612
    }
  },
  {
    "ph": "s", "id": 14612, "pid": 76337, "tid": -914061504, "ts": 1716454216838715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216841141, "dur": 25,
    "args": {
      "External id": 14632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14632,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 14632, "pid": 5, "tid": 7, "ts": 1716454216841141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838787, "dur": 12,
    "args": {
      "External id": 14632,
      "cbid": 211, "correlation": 14632
    }
  },
  {
    "ph": "s", "id": 14632, "pid": 76337, "tid": -914061504, "ts": 1716454216838787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216841167, "dur": 5,
    "args": {
      "External id": 14644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14644,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 14644, "pid": 5, "tid": 7, "ts": 1716454216841167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838809, "dur": 6,
    "args": {
      "External id": 14644,
      "cbid": 211, "correlation": 14644
    }
  },
  {
    "ph": "s", "id": 14644, "pid": 76337, "tid": -914061504, "ts": 1716454216838809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216841174, "dur": 26,
    "args": {
      "External id": 14647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14647, "pid": 5, "tid": 7, "ts": 1716454216841174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838827, "dur": 7,
    "args": {
      "External id": 14647,
      "cbid": 211, "correlation": 14647
    }
  },
  {
    "ph": "s", "id": 14647, "pid": 76337, "tid": -914061504, "ts": 1716454216838827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216841201, "dur": 18,
    "args": {
      "External id": 14656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14656,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14656, "pid": 5, "tid": 7, "ts": 1716454216841201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838866, "dur": 10,
    "args": {
      "External id": 14656,
      "cbid": 211, "correlation": 14656
    }
  },
  {
    "ph": "s", "id": 14656, "pid": 76337, "tid": -914061504, "ts": 1716454216838866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838932, "dur": 0,
    "args": {
      "External id": 14666,
      "cbid": 317, "correlation": 14666
    }
  },
  {
    "ph": "f", "id": 14666, "pid": 76337, "tid": -914061504, "ts": 1716454216838932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838933, "dur": 0,
    "args": {
      "External id": 14667,
      "cbid": 203, "correlation": 14667
    }
  },
  {
    "ph": "f", "id": 14667, "pid": 76337, "tid": -914061504, "ts": 1716454216838933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216838934, "dur": 0,
    "args": {
      "External id": 14668,
      "cbid": 205, "correlation": 14668
    }
  },
  {
    "ph": "f", "id": 14668, "pid": 76337, "tid": -914061504, "ts": 1716454216838934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216842148, "dur": 17,
    "args": {
      "External id": 14672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14672, "pid": 5, "tid": 7, "ts": 1716454216842148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216842115, "dur": 33,
    "args": {
      "External id": 14672,
      "cbid": 211, "correlation": 14672
    }
  },
  {
    "ph": "s", "id": 14672, "pid": 76337, "tid": -914061504, "ts": 1716454216842115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216842167, "dur": 244,
    "args": {
      "External id": 14674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14674, "pid": 5, "tid": 7, "ts": 1716454216842167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216842151, "dur": 6,
    "args": {
      "External id": 14674,
      "cbid": 211, "correlation": 14674
    }
  },
  {
    "ph": "s", "id": 14674, "pid": 76337, "tid": -914061504, "ts": 1716454216842151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216842414, "dur": 1,
    "args": {
      "External id": 14676,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 14676,
      "bytes": 960, "memory bandwidth (GB/s)": 0.4918032786885246
    }
  },
  {
    "ph": "f", "id": 14676, "pid": 5, "tid": 7, "ts": 1716454216842414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216842166, "dur": 14,
    "args": {
      "External id": 14676,
      "cbid": 51, "correlation": 14676
    }
  },
  {
    "ph": "s", "id": 14676, "pid": 76337, "tid": -914061504, "ts": 1716454216842166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216842418, "dur": 815,
    "args": {
      "External id": 14677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14677,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14677, "pid": 5, "tid": 7, "ts": 1716454216842418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216842182, "dur": 10,
    "args": {
      "External id": 14677,
      "cbid": 211, "correlation": 14677
    }
  },
  {
    "ph": "s", "id": 14677, "pid": 76337, "tid": -914061504, "ts": 1716454216842182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216843234, "dur": 13,
    "args": {
      "External id": 14679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14679,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14679, "pid": 5, "tid": 7, "ts": 1716454216843234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216842198, "dur": 7,
    "args": {
      "External id": 14679,
      "cbid": 211, "correlation": 14679
    }
  },
  {
    "ph": "s", "id": 14679, "pid": 76337, "tid": -914061504, "ts": 1716454216842198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216843248, "dur": 14,
    "args": {
      "External id": 14685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14685,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14685, "pid": 5, "tid": 7, "ts": 1716454216843248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216842982, "dur": 12,
    "args": {
      "External id": 14685,
      "cbid": 211, "correlation": 14685
    }
  },
  {
    "ph": "s", "id": 14685, "pid": 76337, "tid": -914061504, "ts": 1716454216842982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216843264, "dur": 3,
    "args": {
      "External id": 14693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14693,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 14693, "pid": 5, "tid": 7, "ts": 1716454216843264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843076, "dur": 11,
    "args": {
      "External id": 14693,
      "cbid": 211, "correlation": 14693
    }
  },
  {
    "ph": "s", "id": 14693, "pid": 76337, "tid": -914061504, "ts": 1716454216843076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843190, "dur": 3,
    "args": {
      "External id": 14709,
      "cbid": 251, "correlation": 14709
    }
  },
  {
    "ph": "f", "id": 14709, "pid": 76337, "tid": -914061504, "ts": 1716454216843190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843199, "dur": 0,
    "args": {
      "External id": 14711,
      "cbid": 251, "correlation": 14711
    }
  },
  {
    "ph": "f", "id": 14711, "pid": 76337, "tid": -914061504, "ts": 1716454216843199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216843269, "dur": 13,
    "args": {
      "External id": 14712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14712,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14712, "pid": 5, "tid": 7, "ts": 1716454216843269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843201, "dur": 14,
    "args": {
      "External id": 14712,
      "cbid": 211, "correlation": 14712
    }
  },
  {
    "ph": "s", "id": 14712, "pid": 76337, "tid": -914061504, "ts": 1716454216843201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216843283, "dur": 5,
    "args": {
      "External id": 14714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14714,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14714, "pid": 5, "tid": 7, "ts": 1716454216843283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843219, "dur": 7,
    "args": {
      "External id": 14714,
      "cbid": 211, "correlation": 14714
    }
  },
  {
    "ph": "s", "id": 14714, "pid": 76337, "tid": -914061504, "ts": 1716454216843219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216843313, "dur": 17,
    "args": {
      "External id": 14724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14724,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14724, "pid": 5, "tid": 7, "ts": 1716454216843313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843301, "dur": 12,
    "args": {
      "External id": 14724,
      "cbid": 211, "correlation": 14724
    }
  },
  {
    "ph": "s", "id": 14724, "pid": 76337, "tid": -914061504, "ts": 1716454216843301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216843406, "dur": 18,
    "args": {
      "External id": 14744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14744,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 14744, "pid": 5, "tid": 7, "ts": 1716454216843406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843394, "dur": 12,
    "args": {
      "External id": 14744,
      "cbid": 211, "correlation": 14744
    }
  },
  {
    "ph": "s", "id": 14744, "pid": 76337, "tid": -914061504, "ts": 1716454216843394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216843428, "dur": 4,
    "args": {
      "External id": 14756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14756,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 14756, "pid": 5, "tid": 7, "ts": 1716454216843428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843417, "dur": 8,
    "args": {
      "External id": 14756,
      "cbid": 211, "correlation": 14756
    }
  },
  {
    "ph": "s", "id": 14756, "pid": 76337, "tid": -914061504, "ts": 1716454216843417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216843453, "dur": 17,
    "args": {
      "External id": 14759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14759, "pid": 5, "tid": 7, "ts": 1716454216843453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843445, "dur": 8,
    "args": {
      "External id": 14759,
      "cbid": 211, "correlation": 14759
    }
  },
  {
    "ph": "s", "id": 14759, "pid": 76337, "tid": -914061504, "ts": 1716454216843445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216843500, "dur": 12,
    "args": {
      "External id": 14768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14768,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14768, "pid": 5, "tid": 7, "ts": 1716454216843500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843489, "dur": 10,
    "args": {
      "External id": 14768,
      "cbid": 211, "correlation": 14768
    }
  },
  {
    "ph": "s", "id": 14768, "pid": 76337, "tid": -914061504, "ts": 1716454216843489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843580, "dur": 1,
    "args": {
      "External id": 14778,
      "cbid": 317, "correlation": 14778
    }
  },
  {
    "ph": "f", "id": 14778, "pid": 76337, "tid": -914061504, "ts": 1716454216843580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843582, "dur": 0,
    "args": {
      "External id": 14779,
      "cbid": 203, "correlation": 14779
    }
  },
  {
    "ph": "f", "id": 14779, "pid": 76337, "tid": -914061504, "ts": 1716454216843582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843583, "dur": 1,
    "args": {
      "External id": 14780,
      "cbid": 205, "correlation": 14780
    }
  },
  {
    "ph": "f", "id": 14780, "pid": 76337, "tid": -914061504, "ts": 1716454216843583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216843614, "dur": 12,
    "args": {
      "External id": 14784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14784,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14784, "pid": 5, "tid": 7, "ts": 1716454216843614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843601, "dur": 12,
    "args": {
      "External id": 14784,
      "cbid": 211, "correlation": 14784
    }
  },
  {
    "ph": "s", "id": 14784, "pid": 76337, "tid": -914061504, "ts": 1716454216843601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216843627, "dur": 165,
    "args": {
      "External id": 14786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14786,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14786, "pid": 5, "tid": 7, "ts": 1716454216843627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843616, "dur": 6,
    "args": {
      "External id": 14786,
      "cbid": 211, "correlation": 14786
    }
  },
  {
    "ph": "s", "id": 14786, "pid": 76337, "tid": -914061504, "ts": 1716454216843616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216843794, "dur": 1,
    "args": {
      "External id": 14788,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 14788,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 14788, "pid": 5, "tid": 7, "ts": 1716454216843794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843629, "dur": 7,
    "args": {
      "External id": 14788,
      "cbid": 51, "correlation": 14788
    }
  },
  {
    "ph": "s", "id": 14788, "pid": 76337, "tid": -914061504, "ts": 1716454216843629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216843798, "dur": 653,
    "args": {
      "External id": 14789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14789,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14789, "pid": 5, "tid": 7, "ts": 1716454216843798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843638, "dur": 6,
    "args": {
      "External id": 14789,
      "cbid": 211, "correlation": 14789
    }
  },
  {
    "ph": "s", "id": 14789, "pid": 76337, "tid": -914061504, "ts": 1716454216843638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216844453, "dur": 13,
    "args": {
      "External id": 14791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14791,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14791, "pid": 5, "tid": 7, "ts": 1716454216844453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843648, "dur": 6,
    "args": {
      "External id": 14791,
      "cbid": 211, "correlation": 14791
    }
  },
  {
    "ph": "s", "id": 14791, "pid": 76337, "tid": -914061504, "ts": 1716454216843648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216844467, "dur": 15,
    "args": {
      "External id": 14797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14797, "pid": 5, "tid": 7, "ts": 1716454216844467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843679, "dur": 9,
    "args": {
      "External id": 14797,
      "cbid": 211, "correlation": 14797
    }
  },
  {
    "ph": "s", "id": 14797, "pid": 76337, "tid": -914061504, "ts": 1716454216843679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843739, "dur": 0,
    "args": {
      "External id": 14807,
      "cbid": 317, "correlation": 14807
    }
  },
  {
    "ph": "f", "id": 14807, "pid": 76337, "tid": -914061504, "ts": 1716454216843739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843740, "dur": 0,
    "args": {
      "External id": 14808,
      "cbid": 203, "correlation": 14808
    }
  },
  {
    "ph": "f", "id": 14808, "pid": 76337, "tid": -914061504, "ts": 1716454216843740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216843740, "dur": 0,
    "args": {
      "External id": 14809,
      "cbid": 205, "correlation": 14809
    }
  },
  {
    "ph": "f", "id": 14809, "pid": 76337, "tid": -914061504, "ts": 1716454216843740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216845551, "dur": 17,
    "args": {
      "External id": 14813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14813, "pid": 5, "tid": 7, "ts": 1716454216845551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216845531, "dur": 20,
    "args": {
      "External id": 14813,
      "cbid": 211, "correlation": 14813
    }
  },
  {
    "ph": "s", "id": 14813, "pid": 76337, "tid": -914061504, "ts": 1716454216845531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216845569, "dur": 4,
    "args": {
      "External id": 14815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14815, "pid": 5, "tid": 7, "ts": 1716454216845569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216845556, "dur": 9,
    "args": {
      "External id": 14815,
      "cbid": 211, "correlation": 14815
    }
  },
  {
    "ph": "s", "id": 14815, "pid": 76337, "tid": -914061504, "ts": 1716454216845556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216845568, "dur": 0,
    "args": {
      "External id": 14816,
      "cbid": 51, "correlation": 14816
    }
  },
  {
    "ph": "s", "id": 14816, "pid": 76337, "tid": -914061504, "ts": 1716454216845568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216845579, "dur": 133,
    "args": {
      "External id": 14817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14817,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 14817, "pid": 5, "tid": 7, "ts": 1716454216845579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216845570, "dur": 8,
    "args": {
      "External id": 14817,
      "cbid": 211, "correlation": 14817
    }
  },
  {
    "ph": "s", "id": 14817, "pid": 76337, "tid": -914061504, "ts": 1716454216845570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216846046, "dur": 16,
    "args": {
      "External id": 14822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14822, "pid": 5, "tid": 7, "ts": 1716454216846046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846033, "dur": 11,
    "args": {
      "External id": 14822,
      "cbid": 211, "correlation": 14822
    }
  },
  {
    "ph": "s", "id": 14822, "pid": 76337, "tid": -914061504, "ts": 1716454216846033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216846086, "dur": 12,
    "args": {
      "External id": 14830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14830,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14830, "pid": 5, "tid": 7, "ts": 1716454216846086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846076, "dur": 9,
    "args": {
      "External id": 14830,
      "cbid": 211, "correlation": 14830
    }
  },
  {
    "ph": "s", "id": 14830, "pid": 76337, "tid": -914061504, "ts": 1716454216846076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216846138, "dur": 11,
    "args": {
      "External id": 14838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14838,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14838, "pid": 5, "tid": 7, "ts": 1716454216846138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846127, "dur": 11,
    "args": {
      "External id": 14838,
      "cbid": 211, "correlation": 14838
    }
  },
  {
    "ph": "s", "id": 14838, "pid": 76337, "tid": -914061504, "ts": 1716454216846127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216846269, "dur": 19,
    "args": {
      "External id": 14858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14858,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 14858, "pid": 5, "tid": 7, "ts": 1716454216846269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846256, "dur": 13,
    "args": {
      "External id": 14858,
      "cbid": 211, "correlation": 14858
    }
  },
  {
    "ph": "s", "id": 14858, "pid": 76337, "tid": -914061504, "ts": 1716454216846256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216846289, "dur": 4,
    "args": {
      "External id": 14870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14870,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 14870, "pid": 5, "tid": 7, "ts": 1716454216846289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846279, "dur": 8,
    "args": {
      "External id": 14870,
      "cbid": 211, "correlation": 14870
    }
  },
  {
    "ph": "s", "id": 14870, "pid": 76337, "tid": -914061504, "ts": 1716454216846279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216846310, "dur": 17,
    "args": {
      "External id": 14873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14873,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14873, "pid": 5, "tid": 7, "ts": 1716454216846310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846302, "dur": 7,
    "args": {
      "External id": 14873,
      "cbid": 211, "correlation": 14873
    }
  },
  {
    "ph": "s", "id": 14873, "pid": 76337, "tid": -914061504, "ts": 1716454216846302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846385, "dur": 0,
    "args": {
      "External id": 14884,
      "cbid": 317, "correlation": 14884
    }
  },
  {
    "ph": "f", "id": 14884, "pid": 76337, "tid": -914061504, "ts": 1716454216846385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846386, "dur": 0,
    "args": {
      "External id": 14885,
      "cbid": 203, "correlation": 14885
    }
  },
  {
    "ph": "f", "id": 14885, "pid": 76337, "tid": -914061504, "ts": 1716454216846386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846387, "dur": 0,
    "args": {
      "External id": 14886,
      "cbid": 205, "correlation": 14886
    }
  },
  {
    "ph": "f", "id": 14886, "pid": 76337, "tid": -914061504, "ts": 1716454216846387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216846420, "dur": 12,
    "args": {
      "External id": 14890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14890,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14890, "pid": 5, "tid": 7, "ts": 1716454216846420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846407, "dur": 13,
    "args": {
      "External id": 14890,
      "cbid": 211, "correlation": 14890
    }
  },
  {
    "ph": "s", "id": 14890, "pid": 76337, "tid": -914061504, "ts": 1716454216846407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216846434, "dur": 3,
    "args": {
      "External id": 14892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14892, "pid": 5, "tid": 7, "ts": 1716454216846434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846425, "dur": 6,
    "args": {
      "External id": 14892,
      "cbid": 211, "correlation": 14892
    }
  },
  {
    "ph": "s", "id": 14892, "pid": 76337, "tid": -914061504, "ts": 1716454216846425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846434, "dur": 0,
    "args": {
      "External id": 14893,
      "cbid": 51, "correlation": 14893
    }
  },
  {
    "ph": "s", "id": 14893, "pid": 76337, "tid": -914061504, "ts": 1716454216846434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216846443, "dur": 93,
    "args": {
      "External id": 14894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14894,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 14894, "pid": 5, "tid": 7, "ts": 1716454216846443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846435, "dur": 7,
    "args": {
      "External id": 14894,
      "cbid": 211, "correlation": 14894
    }
  },
  {
    "ph": "s", "id": 14894, "pid": 76337, "tid": -914061504, "ts": 1716454216846435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216846538, "dur": 15,
    "args": {
      "External id": 14899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14899, "pid": 5, "tid": 7, "ts": 1716454216846538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846465, "dur": 9,
    "args": {
      "External id": 14899,
      "cbid": 211, "correlation": 14899
    }
  },
  {
    "ph": "s", "id": 14899, "pid": 76337, "tid": -914061504, "ts": 1716454216846465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216846626, "dur": 44,
    "args": {
      "External id": 14910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14910,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14910, "pid": 5, "tid": 7, "ts": 1716454216846626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846609, "dur": 17,
    "args": {
      "External id": 14910,
      "cbid": 211, "correlation": 14910
    }
  },
  {
    "ph": "s", "id": 14910, "pid": 76337, "tid": -914061504, "ts": 1716454216846609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216846676, "dur": 19,
    "args": {
      "External id": 14932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14932,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 14932, "pid": 5, "tid": 7, "ts": 1716454216846676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846663, "dur": 12,
    "args": {
      "External id": 14932,
      "cbid": 211, "correlation": 14932
    }
  },
  {
    "ph": "s", "id": 14932, "pid": 76337, "tid": -914061504, "ts": 1716454216846663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846812, "dur": 3,
    "args": {
      "External id": 14943,
      "cbid": 251, "correlation": 14943
    }
  },
  {
    "ph": "f", "id": 14943, "pid": 76337, "tid": -914061504, "ts": 1716454216846812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216846839, "dur": 91,
    "args": {
      "External id": 14944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14944,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14944, "pid": 5, "tid": 7, "ts": 1716454216846839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846823, "dur": 16,
    "args": {
      "External id": 14944,
      "cbid": 211, "correlation": 14944
    }
  },
  {
    "ph": "s", "id": 14944, "pid": 76337, "tid": -914061504, "ts": 1716454216846823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846916, "dur": 1,
    "args": {
      "External id": 14955,
      "cbid": 251, "correlation": 14955
    }
  },
  {
    "ph": "f", "id": 14955, "pid": 76337, "tid": -914061504, "ts": 1716454216846916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216846935, "dur": 85,
    "args": {
      "External id": 14956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14956,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14956, "pid": 5, "tid": 7, "ts": 1716454216846935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216846920, "dur": 13,
    "args": {
      "External id": 14956,
      "cbid": 211, "correlation": 14956
    }
  },
  {
    "ph": "s", "id": 14956, "pid": 76337, "tid": -914061504, "ts": 1716454216846920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847000, "dur": 1,
    "args": {
      "External id": 14967,
      "cbid": 251, "correlation": 14967
    }
  },
  {
    "ph": "f", "id": 14967, "pid": 76337, "tid": -914061504, "ts": 1716454216847000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216847021, "dur": 86,
    "args": {
      "External id": 14968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14968,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 14968, "pid": 5, "tid": 7, "ts": 1716454216847021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847004, "dur": 12,
    "args": {
      "External id": 14968,
      "cbid": 211, "correlation": 14968
    }
  },
  {
    "ph": "s", "id": 14968, "pid": 76337, "tid": -914061504, "ts": 1716454216847004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216847134, "dur": 139,
    "args": {
      "External id": 14993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 14993,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 14993, "pid": 5, "tid": 7, "ts": 1716454216847134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847118, "dur": 16,
    "args": {
      "External id": 14993,
      "cbid": 211, "correlation": 14993
    }
  },
  {
    "ph": "s", "id": 14993, "pid": 76337, "tid": -914061504, "ts": 1716454216847118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847249, "dur": 2,
    "args": {
      "External id": 15011,
      "cbid": 251, "correlation": 15011
    }
  },
  {
    "ph": "f", "id": 15011, "pid": 76337, "tid": -914061504, "ts": 1716454216847249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216847274, "dur": 92,
    "args": {
      "External id": 15013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15013,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 15013, "pid": 5, "tid": 7, "ts": 1716454216847274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847256, "dur": 16,
    "args": {
      "External id": 15013,
      "cbid": 211, "correlation": 15013
    }
  },
  {
    "ph": "s", "id": 15013, "pid": 76337, "tid": -914061504, "ts": 1716454216847256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216847368, "dur": 9,
    "args": {
      "External id": 15021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15021,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15021, "pid": 5, "tid": 7, "ts": 1716454216847368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847341, "dur": 12,
    "args": {
      "External id": 15021,
      "cbid": 211, "correlation": 15021
    }
  },
  {
    "ph": "s", "id": 15021, "pid": 76337, "tid": -914061504, "ts": 1716454216847341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216847394, "dur": 46,
    "args": {
      "External id": 15029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15029,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15029, "pid": 5, "tid": 7, "ts": 1716454216847394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847384, "dur": 10,
    "args": {
      "External id": 15029,
      "cbid": 211, "correlation": 15029
    }
  },
  {
    "ph": "s", "id": 15029, "pid": 76337, "tid": -914061504, "ts": 1716454216847384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216847449, "dur": 18,
    "args": {
      "External id": 15051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15051,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15051, "pid": 5, "tid": 7, "ts": 1716454216847449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847437, "dur": 11,
    "args": {
      "External id": 15051,
      "cbid": 211, "correlation": 15051
    }
  },
  {
    "ph": "s", "id": 15051, "pid": 76337, "tid": -914061504, "ts": 1716454216847437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847524, "dur": 1,
    "args": {
      "External id": 15062,
      "cbid": 251, "correlation": 15062
    }
  },
  {
    "ph": "f", "id": 15062, "pid": 76337, "tid": -914061504, "ts": 1716454216847524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216847543, "dur": 86,
    "args": {
      "External id": 15063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15063,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 15063, "pid": 5, "tid": 7, "ts": 1716454216847543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847529, "dur": 13,
    "args": {
      "External id": 15063,
      "cbid": 211, "correlation": 15063
    }
  },
  {
    "ph": "s", "id": 15063, "pid": 76337, "tid": -914061504, "ts": 1716454216847529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847608, "dur": 1,
    "args": {
      "External id": 15074,
      "cbid": 251, "correlation": 15074
    }
  },
  {
    "ph": "f", "id": 15074, "pid": 76337, "tid": -914061504, "ts": 1716454216847608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847612, "dur": 0,
    "args": {
      "External id": 15075,
      "cbid": 251, "correlation": 15075
    }
  },
  {
    "ph": "f", "id": 15075, "pid": 76337, "tid": -914061504, "ts": 1716454216847612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216847630, "dur": 12,
    "args": {
      "External id": 15076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15076,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15076, "pid": 5, "tid": 7, "ts": 1716454216847630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847614, "dur": 14,
    "args": {
      "External id": 15076,
      "cbid": 211, "correlation": 15076
    }
  },
  {
    "ph": "s", "id": 15076, "pid": 76337, "tid": -914061504, "ts": 1716454216847614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216847644, "dur": 5,
    "args": {
      "External id": 15078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15078,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15078, "pid": 5, "tid": 7, "ts": 1716454216847644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847632, "dur": 8,
    "args": {
      "External id": 15078,
      "cbid": 211, "correlation": 15078
    }
  },
  {
    "ph": "s", "id": 15078, "pid": 76337, "tid": -914061504, "ts": 1716454216847632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847708, "dur": 1,
    "args": {
      "External id": 15089,
      "cbid": 251, "correlation": 15089
    }
  },
  {
    "ph": "f", "id": 15089, "pid": 76337, "tid": -914061504, "ts": 1716454216847708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847711, "dur": 0,
    "args": {
      "External id": 15090,
      "cbid": 251, "correlation": 15090
    }
  },
  {
    "ph": "f", "id": 15090, "pid": 76337, "tid": -914061504, "ts": 1716454216847711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216847726, "dur": 9,
    "args": {
      "External id": 15091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15091,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15091, "pid": 5, "tid": 7, "ts": 1716454216847726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847713, "dur": 13,
    "args": {
      "External id": 15091,
      "cbid": 211, "correlation": 15091
    }
  },
  {
    "ph": "s", "id": 15091, "pid": 76337, "tid": -914061504, "ts": 1716454216847713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216847736, "dur": 3,
    "args": {
      "External id": 15093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15093,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15093, "pid": 5, "tid": 7, "ts": 1716454216847736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847727, "dur": 6,
    "args": {
      "External id": 15093,
      "cbid": 211, "correlation": 15093
    }
  },
  {
    "ph": "s", "id": 15093, "pid": 76337, "tid": -914061504, "ts": 1716454216847727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216847822, "dur": 56,
    "args": {
      "External id": 15118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15118,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15118, "pid": 5, "tid": 7, "ts": 1716454216847822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847809, "dur": 13,
    "args": {
      "External id": 15118,
      "cbid": 211, "correlation": 15118
    }
  },
  {
    "ph": "s", "id": 15118, "pid": 76337, "tid": -914061504, "ts": 1716454216847809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847924, "dur": 1,
    "args": {
      "External id": 15136,
      "cbid": 251, "correlation": 15136
    }
  },
  {
    "ph": "f", "id": 15136, "pid": 76337, "tid": -914061504, "ts": 1716454216847924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216847944, "dur": 88,
    "args": {
      "External id": 15138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15138,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 15138, "pid": 5, "tid": 7, "ts": 1716454216847944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216847930, "dur": 14,
    "args": {
      "External id": 15138,
      "cbid": 211, "correlation": 15138
    }
  },
  {
    "ph": "s", "id": 15138, "pid": 76337, "tid": -914061504, "ts": 1716454216847930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216848033, "dur": 10,
    "args": {
      "External id": 15146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15146,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15146, "pid": 5, "tid": 7, "ts": 1716454216848033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848012, "dur": 13,
    "args": {
      "External id": 15146,
      "cbid": 211, "correlation": 15146
    }
  },
  {
    "ph": "s", "id": 15146, "pid": 76337, "tid": -914061504, "ts": 1716454216848012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216848063, "dur": 13,
    "args": {
      "External id": 15154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15154,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15154, "pid": 5, "tid": 7, "ts": 1716454216848063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848053, "dur": 9,
    "args": {
      "External id": 15154,
      "cbid": 211, "correlation": 15154
    }
  },
  {
    "ph": "s", "id": 15154, "pid": 76337, "tid": -914061504, "ts": 1716454216848053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216848115, "dur": 18,
    "args": {
      "External id": 15176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15176,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15176, "pid": 5, "tid": 7, "ts": 1716454216848115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848104, "dur": 10,
    "args": {
      "External id": 15176,
      "cbid": 211, "correlation": 15176
    }
  },
  {
    "ph": "s", "id": 15176, "pid": 76337, "tid": -914061504, "ts": 1716454216848104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848201, "dur": 2,
    "args": {
      "External id": 15192,
      "cbid": 251, "correlation": 15192
    }
  },
  {
    "ph": "f", "id": 15192, "pid": 76337, "tid": -914061504, "ts": 1716454216848201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848207, "dur": 0,
    "args": {
      "External id": 15194,
      "cbid": 251, "correlation": 15194
    }
  },
  {
    "ph": "f", "id": 15194, "pid": 76337, "tid": -914061504, "ts": 1716454216848207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216848225, "dur": 495,
    "args": {
      "External id": 15195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15195,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15195, "pid": 5, "tid": 7, "ts": 1716454216848225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848209, "dur": 16,
    "args": {
      "External id": 15195,
      "cbid": 211, "correlation": 15195
    }
  },
  {
    "ph": "s", "id": 15195, "pid": 76337, "tid": -914061504, "ts": 1716454216848209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216848721, "dur": 67,
    "args": {
      "External id": 15203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15203, "pid": 5, "tid": 7, "ts": 1716454216848721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848303, "dur": 16,
    "args": {
      "External id": 15203,
      "cbid": 211, "correlation": 15203
    }
  },
  {
    "ph": "s", "id": 15203, "pid": 76337, "tid": -914061504, "ts": 1716454216848303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216848789, "dur": 68,
    "args": {
      "External id": 15211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15211, "pid": 5, "tid": 7, "ts": 1716454216848789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848346, "dur": 10,
    "args": {
      "External id": 15211,
      "cbid": 211, "correlation": 15211
    }
  },
  {
    "ph": "s", "id": 15211, "pid": 76337, "tid": -914061504, "ts": 1716454216848346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848429, "dur": 1,
    "args": {
      "External id": 15227,
      "cbid": 251, "correlation": 15227
    }
  },
  {
    "ph": "f", "id": 15227, "pid": 76337, "tid": -914061504, "ts": 1716454216848429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216848859, "dur": 1,
    "args": {
      "External id": 15229,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 15229,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 15229, "pid": 5, "tid": 7, "ts": 1716454216848859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848435, "dur": 16,
    "args": {
      "External id": 15229,
      "cbid": 51, "correlation": 15229
    }
  },
  {
    "ph": "s", "id": 15229, "pid": 76337, "tid": -914061504, "ts": 1716454216848435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216848863, "dur": 270,
    "args": {
      "External id": 15230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15230,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 15230, "pid": 5, "tid": 7, "ts": 1716454216848863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848452, "dur": 11,
    "args": {
      "External id": 15230,
      "cbid": 211, "correlation": 15230
    }
  },
  {
    "ph": "s", "id": 15230, "pid": 76337, "tid": -914061504, "ts": 1716454216848452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216849134, "dur": 13,
    "args": {
      "External id": 15238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15238,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15238, "pid": 5, "tid": 7, "ts": 1716454216849134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848496, "dur": 10,
    "args": {
      "External id": 15238,
      "cbid": 211, "correlation": 15238
    }
  },
  {
    "ph": "s", "id": 15238, "pid": 76337, "tid": -914061504, "ts": 1716454216848496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216849149, "dur": 38,
    "args": {
      "External id": 15249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15249,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15249, "pid": 5, "tid": 7, "ts": 1716454216849149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848564, "dur": 13,
    "args": {
      "External id": 15249,
      "cbid": 211, "correlation": 15249
    }
  },
  {
    "ph": "s", "id": 15249, "pid": 76337, "tid": -914061504, "ts": 1716454216848564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848631, "dur": 0,
    "args": {
      "External id": 15261,
      "cbid": 317, "correlation": 15261
    }
  },
  {
    "ph": "f", "id": 15261, "pid": 76337, "tid": -914061504, "ts": 1716454216848631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848632, "dur": 0,
    "args": {
      "External id": 15262,
      "cbid": 203, "correlation": 15262
    }
  },
  {
    "ph": "f", "id": 15262, "pid": 76337, "tid": -914061504, "ts": 1716454216848632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848633, "dur": 0,
    "args": {
      "External id": 15263,
      "cbid": 205, "correlation": 15263
    }
  },
  {
    "ph": "f", "id": 15263, "pid": 76337, "tid": -914061504, "ts": 1716454216848633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216849188, "dur": 13,
    "args": {
      "External id": 15267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15267,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15267, "pid": 5, "tid": 7, "ts": 1716454216849188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848649, "dur": 12,
    "args": {
      "External id": 15267,
      "cbid": 211, "correlation": 15267
    }
  },
  {
    "ph": "s", "id": 15267, "pid": 76337, "tid": -914061504, "ts": 1716454216848649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216849202, "dur": 4,
    "args": {
      "External id": 15269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15269,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 15269, "pid": 5, "tid": 7, "ts": 1716454216849202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848667, "dur": 6,
    "args": {
      "External id": 15269,
      "cbid": 211, "correlation": 15269
    }
  },
  {
    "ph": "s", "id": 15269, "pid": 76337, "tid": -914061504, "ts": 1716454216848667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848676, "dur": 0,
    "args": {
      "External id": 15270,
      "cbid": 51, "correlation": 15270
    }
  },
  {
    "ph": "s", "id": 15270, "pid": 76337, "tid": -914061504, "ts": 1716454216848676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216849207, "dur": 98,
    "args": {
      "External id": 15271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15271,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 15271, "pid": 5, "tid": 7, "ts": 1716454216849207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848677, "dur": 5,
    "args": {
      "External id": 15271,
      "cbid": 211, "correlation": 15271
    }
  },
  {
    "ph": "s", "id": 15271, "pid": 76337, "tid": -914061504, "ts": 1716454216848677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216849306, "dur": 17,
    "args": {
      "External id": 15276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15276, "pid": 5, "tid": 7, "ts": 1716454216849306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848705, "dur": 9,
    "args": {
      "External id": 15276,
      "cbid": 211, "correlation": 15276
    }
  },
  {
    "ph": "s", "id": 15276, "pid": 76337, "tid": -914061504, "ts": 1716454216848705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216849325, "dur": 13,
    "args": {
      "External id": 15284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15284,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15284, "pid": 5, "tid": 7, "ts": 1716454216849325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848737, "dur": 9,
    "args": {
      "External id": 15284,
      "cbid": 211, "correlation": 15284
    }
  },
  {
    "ph": "s", "id": 15284, "pid": 76337, "tid": -914061504, "ts": 1716454216848737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454216849338, "dur": 57,
    "args": {
      "External id": 15295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15295,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15295, "pid": 5, "tid": 7, "ts": 1716454216849338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848818, "dur": 13,
    "args": {
      "External id": 15295,
      "cbid": 211, "correlation": 15295
    }
  },
  {
    "ph": "s", "id": 15295, "pid": 76337, "tid": -914061504, "ts": 1716454216848818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848874, "dur": 0,
    "args": {
      "External id": 15305,
      "cbid": 317, "correlation": 15305
    }
  },
  {
    "ph": "f", "id": 15305, "pid": 76337, "tid": -914061504, "ts": 1716454216848874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848875, "dur": 0,
    "args": {
      "External id": 15306,
      "cbid": 203, "correlation": 15306
    }
  },
  {
    "ph": "f", "id": 15306, "pid": 76337, "tid": -914061504, "ts": 1716454216848875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216848876, "dur": 0,
    "args": {
      "External id": 15307,
      "cbid": 205, "correlation": 15307
    }
  },
  {
    "ph": "f", "id": 15307, "pid": 76337, "tid": -914061504, "ts": 1716454216848876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216851164, "dur": 39,
    "args": {
      "External id": 15311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15311,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15311, "pid": 5, "tid": 7, "ts": 1716454216851164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216851140, "dur": 24,
    "args": {
      "External id": 15311,
      "cbid": 211, "correlation": 15311
    }
  },
  {
    "ph": "s", "id": 15311, "pid": 76337, "tid": -914061504, "ts": 1716454216851140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216851204, "dur": 166,
    "args": {
      "External id": 15313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15313, "pid": 5, "tid": 7, "ts": 1716454216851204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216851167, "dur": 6,
    "args": {
      "External id": 15313,
      "cbid": 211, "correlation": 15313
    }
  },
  {
    "ph": "s", "id": 15313, "pid": 76337, "tid": -914061504, "ts": 1716454216851167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216851372, "dur": 1937,
    "args": {
      "External id": 15315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15315,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15315, "pid": 5, "tid": 7, "ts": 1716454216851372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216851184, "dur": 10,
    "args": {
      "External id": 15315,
      "cbid": 211, "correlation": 15315
    }
  },
  {
    "ph": "s", "id": 15315, "pid": 76337, "tid": -914061504, "ts": 1716454216851184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216853310, "dur": 39,
    "args": {
      "External id": 15317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15317,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15317, "pid": 5, "tid": 7, "ts": 1716454216853310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216851202, "dur": 7,
    "args": {
      "External id": 15317,
      "cbid": 211, "correlation": 15317
    }
  },
  {
    "ph": "s", "id": 15317, "pid": 76337, "tid": -914061504, "ts": 1716454216851202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216853351, "dur": 59,
    "args": {
      "External id": 15323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15323, "pid": 5, "tid": 7, "ts": 1716454216853351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216851778, "dur": 11,
    "args": {
      "External id": 15323,
      "cbid": 211, "correlation": 15323
    }
  },
  {
    "ph": "s", "id": 15323, "pid": 76337, "tid": -914061504, "ts": 1716454216851778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216853411, "dur": 85,
    "args": {
      "External id": 15332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15332,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15332, "pid": 5, "tid": 7, "ts": 1716454216853411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216851929, "dur": 17,
    "args": {
      "External id": 15332,
      "cbid": 211, "correlation": 15332
    }
  },
  {
    "ph": "s", "id": 15332, "pid": 76337, "tid": -914061504, "ts": 1716454216851929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216853497, "dur": 71,
    "args": {
      "External id": 15352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15352,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 15352, "pid": 5, "tid": 7, "ts": 1716454216853497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852040, "dur": 14,
    "args": {
      "External id": 15352,
      "cbid": 211, "correlation": 15352
    }
  },
  {
    "ph": "s", "id": 15352, "pid": 76337, "tid": -914061504, "ts": 1716454216852040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216853570, "dur": 4,
    "args": {
      "External id": 15364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15364,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 15364, "pid": 5, "tid": 7, "ts": 1716454216853570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852078, "dur": 8,
    "args": {
      "External id": 15364,
      "cbid": 211, "correlation": 15364
    }
  },
  {
    "ph": "s", "id": 15364, "pid": 76337, "tid": -914061504, "ts": 1716454216852078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216853575, "dur": 80,
    "args": {
      "External id": 15367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15367, "pid": 5, "tid": 7, "ts": 1716454216853575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852108, "dur": 8,
    "args": {
      "External id": 15367,
      "cbid": 211, "correlation": 15367
    }
  },
  {
    "ph": "s", "id": 15367, "pid": 76337, "tid": -914061504, "ts": 1716454216852108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216853657, "dur": 54,
    "args": {
      "External id": 15376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15376,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15376, "pid": 5, "tid": 7, "ts": 1716454216853657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852165, "dur": 11,
    "args": {
      "External id": 15376,
      "cbid": 211, "correlation": 15376
    }
  },
  {
    "ph": "s", "id": 15376, "pid": 76337, "tid": -914061504, "ts": 1716454216852165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852258, "dur": 0,
    "args": {
      "External id": 15386,
      "cbid": 317, "correlation": 15386
    }
  },
  {
    "ph": "f", "id": 15386, "pid": 76337, "tid": -914061504, "ts": 1716454216852258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852259, "dur": 1,
    "args": {
      "External id": 15387,
      "cbid": 203, "correlation": 15387
    }
  },
  {
    "ph": "f", "id": 15387, "pid": 76337, "tid": -914061504, "ts": 1716454216852259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216852260, "dur": 0,
    "args": {
      "External id": 15388,
      "cbid": 205, "correlation": 15388
    }
  },
  {
    "ph": "f", "id": 15388, "pid": 76337, "tid": -914061504, "ts": 1716454216852260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216855232, "dur": 58,
    "args": {
      "External id": 15392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15392,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15392, "pid": 5, "tid": 7, "ts": 1716454216855232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216855208, "dur": 24,
    "args": {
      "External id": 15392,
      "cbid": 211, "correlation": 15392
    }
  },
  {
    "ph": "s", "id": 15392, "pid": 76337, "tid": -914061504, "ts": 1716454216855208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216855291, "dur": 124,
    "args": {
      "External id": 15394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15394,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15394, "pid": 5, "tid": 7, "ts": 1716454216855291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216855235, "dur": 6,
    "args": {
      "External id": 15394,
      "cbid": 211, "correlation": 15394
    }
  },
  {
    "ph": "s", "id": 15394, "pid": 76337, "tid": -914061504, "ts": 1716454216855235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216855416, "dur": 1903,
    "args": {
      "External id": 15396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15396,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15396, "pid": 5, "tid": 7, "ts": 1716454216855416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216855252, "dur": 9,
    "args": {
      "External id": 15396,
      "cbid": 211, "correlation": 15396
    }
  },
  {
    "ph": "s", "id": 15396, "pid": 76337, "tid": -914061504, "ts": 1716454216855252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216857321, "dur": 20,
    "args": {
      "External id": 15398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15398,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15398, "pid": 5, "tid": 7, "ts": 1716454216857321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216855267, "dur": 7,
    "args": {
      "External id": 15398,
      "cbid": 211, "correlation": 15398
    }
  },
  {
    "ph": "s", "id": 15398, "pid": 76337, "tid": -914061504, "ts": 1716454216855267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216857342, "dur": 33,
    "args": {
      "External id": 15404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15404,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15404, "pid": 5, "tid": 7, "ts": 1716454216857342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216855953, "dur": 11,
    "args": {
      "External id": 15404,
      "cbid": 211, "correlation": 15404
    }
  },
  {
    "ph": "s", "id": 15404, "pid": 76337, "tid": -914061504, "ts": 1716454216855953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216857376, "dur": 3,
    "args": {
      "External id": 15412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15412,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 15412, "pid": 5, "tid": 7, "ts": 1716454216857376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856031, "dur": 11,
    "args": {
      "External id": 15412,
      "cbid": 211, "correlation": 15412
    }
  },
  {
    "ph": "s", "id": 15412, "pid": 76337, "tid": -914061504, "ts": 1716454216856031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856160, "dur": 4,
    "args": {
      "External id": 15428,
      "cbid": 251, "correlation": 15428
    }
  },
  {
    "ph": "f", "id": 15428, "pid": 76337, "tid": -914061504, "ts": 1716454216856160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856170, "dur": 0,
    "args": {
      "External id": 15430,
      "cbid": 251, "correlation": 15430
    }
  },
  {
    "ph": "f", "id": 15430, "pid": 76337, "tid": -914061504, "ts": 1716454216856170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216857381, "dur": 12,
    "args": {
      "External id": 15431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15431,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 15431, "pid": 5, "tid": 7, "ts": 1716454216857381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856174, "dur": 15,
    "args": {
      "External id": 15431,
      "cbid": 211, "correlation": 15431
    }
  },
  {
    "ph": "s", "id": 15431, "pid": 76337, "tid": -914061504, "ts": 1716454216856174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216857394, "dur": 5,
    "args": {
      "External id": 15433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15433,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 15433, "pid": 5, "tid": 7, "ts": 1716454216857394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856193, "dur": 8,
    "args": {
      "External id": 15433,
      "cbid": 211, "correlation": 15433
    }
  },
  {
    "ph": "s", "id": 15433, "pid": 76337, "tid": -914061504, "ts": 1716454216856193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216857400, "dur": 29,
    "args": {
      "External id": 15443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15443, "pid": 5, "tid": 7, "ts": 1716454216857400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856281, "dur": 12,
    "args": {
      "External id": 15443,
      "cbid": 211, "correlation": 15443
    }
  },
  {
    "ph": "s", "id": 15443, "pid": 76337, "tid": -914061504, "ts": 1716454216856281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216857430, "dur": 30,
    "args": {
      "External id": 15463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15463,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 15463, "pid": 5, "tid": 7, "ts": 1716454216857430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856372, "dur": 12,
    "args": {
      "External id": 15463,
      "cbid": 211, "correlation": 15463
    }
  },
  {
    "ph": "s", "id": 15463, "pid": 76337, "tid": -914061504, "ts": 1716454216856372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216857462, "dur": 4,
    "args": {
      "External id": 15475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15475,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 15475, "pid": 5, "tid": 7, "ts": 1716454216857462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856394, "dur": 7,
    "args": {
      "External id": 15475,
      "cbid": 211, "correlation": 15475
    }
  },
  {
    "ph": "s", "id": 15475, "pid": 76337, "tid": -914061504, "ts": 1716454216856394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216857467, "dur": 30,
    "args": {
      "External id": 15478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15478,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15478, "pid": 5, "tid": 7, "ts": 1716454216857467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856421, "dur": 7,
    "args": {
      "External id": 15478,
      "cbid": 211, "correlation": 15478
    }
  },
  {
    "ph": "s", "id": 15478, "pid": 76337, "tid": -914061504, "ts": 1716454216856421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216857499, "dur": 21,
    "args": {
      "External id": 15487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15487,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15487, "pid": 5, "tid": 7, "ts": 1716454216857499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856465, "dur": 10,
    "args": {
      "External id": 15487,
      "cbid": 211, "correlation": 15487
    }
  },
  {
    "ph": "s", "id": 15487, "pid": 76337, "tid": -914061504, "ts": 1716454216856465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856568, "dur": 1,
    "args": {
      "External id": 15497,
      "cbid": 317, "correlation": 15497
    }
  },
  {
    "ph": "f", "id": 15497, "pid": 76337, "tid": -914061504, "ts": 1716454216856568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856569, "dur": 1,
    "args": {
      "External id": 15498,
      "cbid": 203, "correlation": 15498
    }
  },
  {
    "ph": "f", "id": 15498, "pid": 76337, "tid": -914061504, "ts": 1716454216856569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856571, "dur": 0,
    "args": {
      "External id": 15499,
      "cbid": 205, "correlation": 15499
    }
  },
  {
    "ph": "f", "id": 15499, "pid": 76337, "tid": -914061504, "ts": 1716454216856571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216857521, "dur": 22,
    "args": {
      "External id": 15503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15503,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15503, "pid": 5, "tid": 7, "ts": 1716454216857521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856589, "dur": 13,
    "args": {
      "External id": 15503,
      "cbid": 211, "correlation": 15503
    }
  },
  {
    "ph": "s", "id": 15503, "pid": 76337, "tid": -914061504, "ts": 1716454216856589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216857544, "dur": 45,
    "args": {
      "External id": 15505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15505,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15505, "pid": 5, "tid": 7, "ts": 1716454216857544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856605, "dur": 5,
    "args": {
      "External id": 15505,
      "cbid": 211, "correlation": 15505
    }
  },
  {
    "ph": "s", "id": 15505, "pid": 76337, "tid": -914061504, "ts": 1716454216856605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216857591, "dur": 650,
    "args": {
      "External id": 15507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15507,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15507, "pid": 5, "tid": 7, "ts": 1716454216857591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856617, "dur": 7,
    "args": {
      "External id": 15507,
      "cbid": 211, "correlation": 15507
    }
  },
  {
    "ph": "s", "id": 15507, "pid": 76337, "tid": -914061504, "ts": 1716454216856617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216858242, "dur": 21,
    "args": {
      "External id": 15509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15509,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15509, "pid": 5, "tid": 7, "ts": 1716454216858242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856629, "dur": 5,
    "args": {
      "External id": 15509,
      "cbid": 211, "correlation": 15509
    }
  },
  {
    "ph": "s", "id": 15509, "pid": 76337, "tid": -914061504, "ts": 1716454216856629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216858264, "dur": 33,
    "args": {
      "External id": 15515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15515,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15515, "pid": 5, "tid": 7, "ts": 1716454216858264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856657, "dur": 8,
    "args": {
      "External id": 15515,
      "cbid": 211, "correlation": 15515
    }
  },
  {
    "ph": "s", "id": 15515, "pid": 76337, "tid": -914061504, "ts": 1716454216856657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856717, "dur": 0,
    "args": {
      "External id": 15525,
      "cbid": 317, "correlation": 15525
    }
  },
  {
    "ph": "f", "id": 15525, "pid": 76337, "tid": -914061504, "ts": 1716454216856717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856718, "dur": 0,
    "args": {
      "External id": 15526,
      "cbid": 203, "correlation": 15526
    }
  },
  {
    "ph": "f", "id": 15526, "pid": 76337, "tid": -914061504, "ts": 1716454216856718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216856719, "dur": 0,
    "args": {
      "External id": 15527,
      "cbid": 205, "correlation": 15527
    }
  },
  {
    "ph": "f", "id": 15527, "pid": 76337, "tid": -914061504, "ts": 1716454216856719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216858438, "dur": 56,
    "args": {
      "External id": 15531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15531,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15531, "pid": 5, "tid": 7, "ts": 1716454216858438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216858420, "dur": 19,
    "args": {
      "External id": 15531,
      "cbid": 211, "correlation": 15531
    }
  },
  {
    "ph": "s", "id": 15531, "pid": 76337, "tid": -914061504, "ts": 1716454216858420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216858495, "dur": 273,
    "args": {
      "External id": 15533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15533,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15533, "pid": 5, "tid": 7, "ts": 1716454216858495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216858448, "dur": 11,
    "args": {
      "External id": 15533,
      "cbid": 211, "correlation": 15533
    }
  },
  {
    "ph": "s", "id": 15533, "pid": 76337, "tid": -914061504, "ts": 1716454216858448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216858770, "dur": 22,
    "args": {
      "External id": 15535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15535,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15535, "pid": 5, "tid": 7, "ts": 1716454216858770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216858463, "dur": 6,
    "args": {
      "External id": 15535,
      "cbid": 211, "correlation": 15535
    }
  },
  {
    "ph": "s", "id": 15535, "pid": 76337, "tid": -914061504, "ts": 1716454216858463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216858924, "dur": 32,
    "args": {
      "External id": 15541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15541, "pid": 5, "tid": 7, "ts": 1716454216858924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216858912, "dur": 10,
    "args": {
      "External id": 15541,
      "cbid": 211, "correlation": 15541
    }
  },
  {
    "ph": "s", "id": 15541, "pid": 76337, "tid": -914061504, "ts": 1716454216858912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216858963, "dur": 26,
    "args": {
      "External id": 15549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15549,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15549, "pid": 5, "tid": 7, "ts": 1716454216858963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216858951, "dur": 10,
    "args": {
      "External id": 15549,
      "cbid": 211, "correlation": 15549
    }
  },
  {
    "ph": "s", "id": 15549, "pid": 76337, "tid": -914061504, "ts": 1716454216858951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216859024, "dur": 20,
    "args": {
      "External id": 15557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15557,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15557, "pid": 5, "tid": 7, "ts": 1716454216859024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859012, "dur": 11,
    "args": {
      "External id": 15557,
      "cbid": 211, "correlation": 15557
    }
  },
  {
    "ph": "s", "id": 15557, "pid": 76337, "tid": -914061504, "ts": 1716454216859012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216859149, "dur": 30,
    "args": {
      "External id": 15577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15577,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 15577, "pid": 5, "tid": 7, "ts": 1716454216859149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859135, "dur": 12,
    "args": {
      "External id": 15577,
      "cbid": 211, "correlation": 15577
    }
  },
  {
    "ph": "s", "id": 15577, "pid": 76337, "tid": -914061504, "ts": 1716454216859135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216859180, "dur": 4,
    "args": {
      "External id": 15589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15589,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 15589, "pid": 5, "tid": 7, "ts": 1716454216859180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859159, "dur": 8,
    "args": {
      "External id": 15589,
      "cbid": 211, "correlation": 15589
    }
  },
  {
    "ph": "s", "id": 15589, "pid": 76337, "tid": -914061504, "ts": 1716454216859159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216859188, "dur": 30,
    "args": {
      "External id": 15592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15592, "pid": 5, "tid": 7, "ts": 1716454216859188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859180, "dur": 7,
    "args": {
      "External id": 15592,
      "cbid": 211, "correlation": 15592
    }
  },
  {
    "ph": "s", "id": 15592, "pid": 76337, "tid": -914061504, "ts": 1716454216859180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859245, "dur": 0,
    "args": {
      "External id": 15603,
      "cbid": 317, "correlation": 15603
    }
  },
  {
    "ph": "f", "id": 15603, "pid": 76337, "tid": -914061504, "ts": 1716454216859245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859246, "dur": 0,
    "args": {
      "External id": 15604,
      "cbid": 203, "correlation": 15604
    }
  },
  {
    "ph": "f", "id": 15604, "pid": 76337, "tid": -914061504, "ts": 1716454216859246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859247, "dur": 0,
    "args": {
      "External id": 15605,
      "cbid": 205, "correlation": 15605
    }
  },
  {
    "ph": "f", "id": 15605, "pid": 76337, "tid": -914061504, "ts": 1716454216859247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216859275, "dur": 21,
    "args": {
      "External id": 15609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15609,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15609, "pid": 5, "tid": 7, "ts": 1716454216859275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859262, "dur": 12,
    "args": {
      "External id": 15609,
      "cbid": 211, "correlation": 15609
    }
  },
  {
    "ph": "s", "id": 15609, "pid": 76337, "tid": -914061504, "ts": 1716454216859262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216859297, "dur": 105,
    "args": {
      "External id": 15611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15611,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15611, "pid": 5, "tid": 7, "ts": 1716454216859297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859281, "dur": 7,
    "args": {
      "External id": 15611,
      "cbid": 211, "correlation": 15611
    }
  },
  {
    "ph": "s", "id": 15611, "pid": 76337, "tid": -914061504, "ts": 1716454216859281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216859404, "dur": 21,
    "args": {
      "External id": 15613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15613,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15613, "pid": 5, "tid": 7, "ts": 1716454216859404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859292, "dur": 5,
    "args": {
      "External id": 15613,
      "cbid": 211, "correlation": 15613
    }
  },
  {
    "ph": "s", "id": 15613, "pid": 76337, "tid": -914061504, "ts": 1716454216859292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216859426, "dur": 32,
    "args": {
      "External id": 15619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15619, "pid": 5, "tid": 7, "ts": 1716454216859426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859321, "dur": 8,
    "args": {
      "External id": 15619,
      "cbid": 211, "correlation": 15619
    }
  },
  {
    "ph": "s", "id": 15619, "pid": 76337, "tid": -914061504, "ts": 1716454216859321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216859501, "dur": 81,
    "args": {
      "External id": 15630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15630,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15630, "pid": 5, "tid": 7, "ts": 1716454216859501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859483, "dur": 19,
    "args": {
      "External id": 15630,
      "cbid": 211, "correlation": 15630
    }
  },
  {
    "ph": "s", "id": 15630, "pid": 76337, "tid": -914061504, "ts": 1716454216859483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216859584, "dur": 37,
    "args": {
      "External id": 15652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15652,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15652, "pid": 5, "tid": 7, "ts": 1716454216859584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859525, "dur": 10,
    "args": {
      "External id": 15652,
      "cbid": 211, "correlation": 15652
    }
  },
  {
    "ph": "s", "id": 15652, "pid": 76337, "tid": -914061504, "ts": 1716454216859525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859682, "dur": 3,
    "args": {
      "External id": 15663,
      "cbid": 251, "correlation": 15663
    }
  },
  {
    "ph": "f", "id": 15663, "pid": 76337, "tid": -914061504, "ts": 1716454216859682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216859710, "dur": 96,
    "args": {
      "External id": 15664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15664,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15664, "pid": 5, "tid": 7, "ts": 1716454216859710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859693, "dur": 17,
    "args": {
      "External id": 15664,
      "cbid": 211, "correlation": 15664
    }
  },
  {
    "ph": "s", "id": 15664, "pid": 76337, "tid": -914061504, "ts": 1716454216859693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859791, "dur": 1,
    "args": {
      "External id": 15675,
      "cbid": 251, "correlation": 15675
    }
  },
  {
    "ph": "f", "id": 15675, "pid": 76337, "tid": -914061504, "ts": 1716454216859791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216859811, "dur": 89,
    "args": {
      "External id": 15676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15676,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15676, "pid": 5, "tid": 7, "ts": 1716454216859811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859795, "dur": 13,
    "args": {
      "External id": 15676,
      "cbid": 211, "correlation": 15676
    }
  },
  {
    "ph": "s", "id": 15676, "pid": 76337, "tid": -914061504, "ts": 1716454216859795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859866, "dur": 1,
    "args": {
      "External id": 15687,
      "cbid": 251, "correlation": 15687
    }
  },
  {
    "ph": "f", "id": 15687, "pid": 76337, "tid": -914061504, "ts": 1716454216859866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216859900, "dur": 76,
    "args": {
      "External id": 15688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15688,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15688, "pid": 5, "tid": 7, "ts": 1716454216859900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859870, "dur": 12,
    "args": {
      "External id": 15688,
      "cbid": 211, "correlation": 15688
    }
  },
  {
    "ph": "s", "id": 15688, "pid": 76337, "tid": -914061504, "ts": 1716454216859870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216860010, "dur": 551,
    "args": {
      "External id": 15709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15709,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 15709, "pid": 5, "tid": 7, "ts": 1716454216860010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216859993, "dur": 17,
    "args": {
      "External id": 15709,
      "cbid": 211, "correlation": 15709
    }
  },
  {
    "ph": "s", "id": 15709, "pid": 76337, "tid": -914061504, "ts": 1716454216859993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860151, "dur": 3,
    "args": {
      "External id": 15727,
      "cbid": 251, "correlation": 15727
    }
  },
  {
    "ph": "f", "id": 15727, "pid": 76337, "tid": -914061504, "ts": 1716454216860151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216860562, "dur": 98,
    "args": {
      "External id": 15729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15729,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15729, "pid": 5, "tid": 7, "ts": 1716454216860562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860159, "dur": 15,
    "args": {
      "External id": 15729,
      "cbid": 211, "correlation": 15729
    }
  },
  {
    "ph": "s", "id": 15729, "pid": 76337, "tid": -914061504, "ts": 1716454216860159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216860661, "dur": 19,
    "args": {
      "External id": 15737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15737,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15737, "pid": 5, "tid": 7, "ts": 1716454216860661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860244, "dur": 13,
    "args": {
      "External id": 15737,
      "cbid": 211, "correlation": 15737
    }
  },
  {
    "ph": "s", "id": 15737, "pid": 76337, "tid": -914061504, "ts": 1716454216860244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216860682, "dur": 87,
    "args": {
      "External id": 15745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15745, "pid": 5, "tid": 7, "ts": 1716454216860682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860287, "dur": 9,
    "args": {
      "External id": 15745,
      "cbid": 211, "correlation": 15745
    }
  },
  {
    "ph": "s", "id": 15745, "pid": 76337, "tid": -914061504, "ts": 1716454216860287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216860770, "dur": 35,
    "args": {
      "External id": 15767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15767,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15767, "pid": 5, "tid": 7, "ts": 1716454216860770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860353, "dur": 12,
    "args": {
      "External id": 15767,
      "cbid": 211, "correlation": 15767
    }
  },
  {
    "ph": "s", "id": 15767, "pid": 76337, "tid": -914061504, "ts": 1716454216860353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860459, "dur": 1,
    "args": {
      "External id": 15778,
      "cbid": 251, "correlation": 15778
    }
  },
  {
    "ph": "f", "id": 15778, "pid": 76337, "tid": -914061504, "ts": 1716454216860459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216860806, "dur": 76,
    "args": {
      "External id": 15779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15779,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15779, "pid": 5, "tid": 7, "ts": 1716454216860806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860464, "dur": 14,
    "args": {
      "External id": 15779,
      "cbid": 211, "correlation": 15779
    }
  },
  {
    "ph": "s", "id": 15779, "pid": 76337, "tid": -914061504, "ts": 1716454216860464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860558, "dur": 1,
    "args": {
      "External id": 15790,
      "cbid": 251, "correlation": 15790
    }
  },
  {
    "ph": "f", "id": 15790, "pid": 76337, "tid": -914061504, "ts": 1716454216860558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860563, "dur": 0,
    "args": {
      "External id": 15791,
      "cbid": 251, "correlation": 15791
    }
  },
  {
    "ph": "f", "id": 15791, "pid": 76337, "tid": -914061504, "ts": 1716454216860563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216860883, "dur": 11,
    "args": {
      "External id": 15792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15792,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 15792, "pid": 5, "tid": 7, "ts": 1716454216860883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860565, "dur": 16,
    "args": {
      "External id": 15792,
      "cbid": 211, "correlation": 15792
    }
  },
  {
    "ph": "s", "id": 15792, "pid": 76337, "tid": -914061504, "ts": 1716454216860565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216860896, "dur": 5,
    "args": {
      "External id": 15794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15794,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 15794, "pid": 5, "tid": 7, "ts": 1716454216860896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860584, "dur": 9,
    "args": {
      "External id": 15794,
      "cbid": 211, "correlation": 15794
    }
  },
  {
    "ph": "s", "id": 15794, "pid": 76337, "tid": -914061504, "ts": 1716454216860584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860658, "dur": 1,
    "args": {
      "External id": 15805,
      "cbid": 251, "correlation": 15805
    }
  },
  {
    "ph": "f", "id": 15805, "pid": 76337, "tid": -914061504, "ts": 1716454216860658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860661, "dur": 0,
    "args": {
      "External id": 15806,
      "cbid": 251, "correlation": 15806
    }
  },
  {
    "ph": "f", "id": 15806, "pid": 76337, "tid": -914061504, "ts": 1716454216860661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216860902, "dur": 7,
    "args": {
      "External id": 15807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15807,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 15807, "pid": 5, "tid": 7, "ts": 1716454216860902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860663, "dur": 12,
    "args": {
      "External id": 15807,
      "cbid": 211, "correlation": 15807
    }
  },
  {
    "ph": "s", "id": 15807, "pid": 76337, "tid": -914061504, "ts": 1716454216860663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216860911, "dur": 3,
    "args": {
      "External id": 15809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15809,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 15809, "pid": 5, "tid": 7, "ts": 1716454216860911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860677, "dur": 7,
    "args": {
      "External id": 15809,
      "cbid": 211, "correlation": 15809
    }
  },
  {
    "ph": "s", "id": 15809, "pid": 76337, "tid": -914061504, "ts": 1716454216860677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216860915, "dur": 92,
    "args": {
      "External id": 15830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15830,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 15830, "pid": 5, "tid": 7, "ts": 1716454216860915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860754, "dur": 13,
    "args": {
      "External id": 15830,
      "cbid": 211, "correlation": 15830
    }
  },
  {
    "ph": "s", "id": 15830, "pid": 76337, "tid": -914061504, "ts": 1716454216860754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860864, "dur": 1,
    "args": {
      "External id": 15848,
      "cbid": 251, "correlation": 15848
    }
  },
  {
    "ph": "f", "id": 15848, "pid": 76337, "tid": -914061504, "ts": 1716454216860864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216861009, "dur": 80,
    "args": {
      "External id": 15850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15850,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15850, "pid": 5, "tid": 7, "ts": 1716454216861009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860870, "dur": 14,
    "args": {
      "External id": 15850,
      "cbid": 211, "correlation": 15850
    }
  },
  {
    "ph": "s", "id": 15850, "pid": 76337, "tid": -914061504, "ts": 1716454216860870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216861091, "dur": 19,
    "args": {
      "External id": 15858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15858,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15858, "pid": 5, "tid": 7, "ts": 1716454216861091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860944, "dur": 13,
    "args": {
      "External id": 15858,
      "cbid": 211, "correlation": 15858
    }
  },
  {
    "ph": "s", "id": 15858, "pid": 76337, "tid": -914061504, "ts": 1716454216860944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216861111, "dur": 28,
    "args": {
      "External id": 15866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15866,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15866, "pid": 5, "tid": 7, "ts": 1716454216861111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216860994, "dur": 10,
    "args": {
      "External id": 15866,
      "cbid": 211, "correlation": 15866
    }
  },
  {
    "ph": "s", "id": 15866, "pid": 76337, "tid": -914061504, "ts": 1716454216860994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216861140, "dur": 35,
    "args": {
      "External id": 15888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15888,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15888, "pid": 5, "tid": 7, "ts": 1716454216861140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861046, "dur": 11,
    "args": {
      "External id": 15888,
      "cbid": 211, "correlation": 15888
    }
  },
  {
    "ph": "s", "id": 15888, "pid": 76337, "tid": -914061504, "ts": 1716454216861046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861143, "dur": 2,
    "args": {
      "External id": 15904,
      "cbid": 251, "correlation": 15904
    }
  },
  {
    "ph": "f", "id": 15904, "pid": 76337, "tid": -914061504, "ts": 1716454216861143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861149, "dur": 0,
    "args": {
      "External id": 15906,
      "cbid": 251, "correlation": 15906
    }
  },
  {
    "ph": "f", "id": 15906, "pid": 76337, "tid": -914061504, "ts": 1716454216861149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216861177, "dur": 547,
    "args": {
      "External id": 15907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15907,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 15907, "pid": 5, "tid": 7, "ts": 1716454216861177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861155, "dur": 13,
    "args": {
      "External id": 15907,
      "cbid": 211, "correlation": 15907
    }
  },
  {
    "ph": "s", "id": 15907, "pid": 76337, "tid": -914061504, "ts": 1716454216861155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216861725, "dur": 126,
    "args": {
      "External id": 15915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15915,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15915, "pid": 5, "tid": 7, "ts": 1716454216861725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861248, "dur": 14,
    "args": {
      "External id": 15915,
      "cbid": 211, "correlation": 15915
    }
  },
  {
    "ph": "s", "id": 15915, "pid": 76337, "tid": -914061504, "ts": 1716454216861248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216861852, "dur": 129,
    "args": {
      "External id": 15923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15923, "pid": 5, "tid": 7, "ts": 1716454216861852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861288, "dur": 12,
    "args": {
      "External id": 15923,
      "cbid": 211, "correlation": 15923
    }
  },
  {
    "ph": "s", "id": 15923, "pid": 76337, "tid": -914061504, "ts": 1716454216861288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861373, "dur": 1,
    "args": {
      "External id": 15939,
      "cbid": 251, "correlation": 15939
    }
  },
  {
    "ph": "f", "id": 15939, "pid": 76337, "tid": -914061504, "ts": 1716454216861373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216861982, "dur": 307,
    "args": {
      "External id": 15941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15941,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15941, "pid": 5, "tid": 7, "ts": 1716454216861982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861378, "dur": 13,
    "args": {
      "External id": 15941,
      "cbid": 211, "correlation": 15941
    }
  },
  {
    "ph": "s", "id": 15941, "pid": 76337, "tid": -914061504, "ts": 1716454216861378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216862291, "dur": 27,
    "args": {
      "External id": 15949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15949,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15949, "pid": 5, "tid": 7, "ts": 1716454216862291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861420, "dur": 10,
    "args": {
      "External id": 15949,
      "cbid": 211, "correlation": 15949
    }
  },
  {
    "ph": "s", "id": 15949, "pid": 76337, "tid": -914061504, "ts": 1716454216861420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216862319, "dur": 82,
    "args": {
      "External id": 15960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15960, "pid": 5, "tid": 7, "ts": 1716454216862319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861488, "dur": 14,
    "args": {
      "External id": 15960,
      "cbid": 211, "correlation": 15960
    }
  },
  {
    "ph": "s", "id": 15960, "pid": 76337, "tid": -914061504, "ts": 1716454216861488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861557, "dur": 1,
    "args": {
      "External id": 15972,
      "cbid": 317, "correlation": 15972
    }
  },
  {
    "ph": "f", "id": 15972, "pid": 76337, "tid": -914061504, "ts": 1716454216861557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861558, "dur": 0,
    "args": {
      "External id": 15973,
      "cbid": 203, "correlation": 15973
    }
  },
  {
    "ph": "f", "id": 15973, "pid": 76337, "tid": -914061504, "ts": 1716454216861558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861559, "dur": 0,
    "args": {
      "External id": 15974,
      "cbid": 205, "correlation": 15974
    }
  },
  {
    "ph": "f", "id": 15974, "pid": 76337, "tid": -914061504, "ts": 1716454216861559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216862402, "dur": 23,
    "args": {
      "External id": 15978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15978,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15978, "pid": 5, "tid": 7, "ts": 1716454216862402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861575, "dur": 12,
    "args": {
      "External id": 15978,
      "cbid": 211, "correlation": 15978
    }
  },
  {
    "ph": "s", "id": 15978, "pid": 76337, "tid": -914061504, "ts": 1716454216861575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216862426, "dur": 122,
    "args": {
      "External id": 15980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15980,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 15980, "pid": 5, "tid": 7, "ts": 1716454216862426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861595, "dur": 7,
    "args": {
      "External id": 15980,
      "cbid": 211, "correlation": 15980
    }
  },
  {
    "ph": "s", "id": 15980, "pid": 76337, "tid": -914061504, "ts": 1716454216861595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216862549, "dur": 22,
    "args": {
      "External id": 15982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15982, "pid": 5, "tid": 7, "ts": 1716454216862549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861606, "dur": 5,
    "args": {
      "External id": 15982,
      "cbid": 211, "correlation": 15982
    }
  },
  {
    "ph": "s", "id": 15982, "pid": 76337, "tid": -914061504, "ts": 1716454216861606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216862572, "dur": 33,
    "args": {
      "External id": 15988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15988,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15988, "pid": 5, "tid": 7, "ts": 1716454216862572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861634, "dur": 9,
    "args": {
      "External id": 15988,
      "cbid": 211, "correlation": 15988
    }
  },
  {
    "ph": "s", "id": 15988, "pid": 76337, "tid": -914061504, "ts": 1716454216861634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216862607, "dur": 27,
    "args": {
      "External id": 15996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 15996,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 15996, "pid": 5, "tid": 7, "ts": 1716454216862607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861666, "dur": 8,
    "args": {
      "External id": 15996,
      "cbid": 211, "correlation": 15996
    }
  },
  {
    "ph": "s", "id": 15996, "pid": 76337, "tid": -914061504, "ts": 1716454216861666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216862634, "dur": 54,
    "args": {
      "External id": 16005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16005,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16005, "pid": 5, "tid": 7, "ts": 1716454216862634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861724, "dur": 12,
    "args": {
      "External id": 16005,
      "cbid": 211, "correlation": 16005
    }
  },
  {
    "ph": "s", "id": 16005, "pid": 76337, "tid": -914061504, "ts": 1716454216861724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216862690, "dur": 53,
    "args": {
      "External id": 16025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16025,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 16025, "pid": 5, "tid": 7, "ts": 1716454216862690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861816, "dur": 13,
    "args": {
      "External id": 16025,
      "cbid": 211, "correlation": 16025
    }
  },
  {
    "ph": "s", "id": 16025, "pid": 76337, "tid": -914061504, "ts": 1716454216861816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216862744, "dur": 5,
    "args": {
      "External id": 16037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16037,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 16037, "pid": 5, "tid": 7, "ts": 1716454216862744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861840, "dur": 7,
    "args": {
      "External id": 16037,
      "cbid": 211, "correlation": 16037
    }
  },
  {
    "ph": "s", "id": 16037, "pid": 76337, "tid": -914061504, "ts": 1716454216861840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216862750, "dur": 58,
    "args": {
      "External id": 16040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16040, "pid": 5, "tid": 7, "ts": 1716454216862750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861860, "dur": 7,
    "args": {
      "External id": 16040,
      "cbid": 211, "correlation": 16040
    }
  },
  {
    "ph": "s", "id": 16040, "pid": 76337, "tid": -914061504, "ts": 1716454216861860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216862809, "dur": 37,
    "args": {
      "External id": 16049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16049,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16049, "pid": 5, "tid": 7, "ts": 1716454216862809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861907, "dur": 10,
    "args": {
      "External id": 16049,
      "cbid": 211, "correlation": 16049
    }
  },
  {
    "ph": "s", "id": 16049, "pid": 76337, "tid": -914061504, "ts": 1716454216861907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861972, "dur": 9,
    "args": {
      "External id": 16059,
      "cbid": 317, "correlation": 16059
    }
  },
  {
    "ph": "f", "id": 16059, "pid": 76337, "tid": -914061504, "ts": 1716454216861972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861982, "dur": 0,
    "args": {
      "External id": 16060,
      "cbid": 203, "correlation": 16060
    }
  },
  {
    "ph": "f", "id": 16060, "pid": 76337, "tid": -914061504, "ts": 1716454216861982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216861983, "dur": 0,
    "args": {
      "External id": 16061,
      "cbid": 205, "correlation": 16061
    }
  },
  {
    "ph": "f", "id": 16061, "pid": 76337, "tid": -914061504, "ts": 1716454216861983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216865040, "dur": 39,
    "args": {
      "External id": 16065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16065,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16065, "pid": 5, "tid": 7, "ts": 1716454216865040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216865012, "dur": 28,
    "args": {
      "External id": 16065,
      "cbid": 211, "correlation": 16065
    }
  },
  {
    "ph": "s", "id": 16065, "pid": 76337, "tid": -914061504, "ts": 1716454216865012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216865080, "dur": 85,
    "args": {
      "External id": 16067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16067,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16067, "pid": 5, "tid": 7, "ts": 1716454216865080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216865043, "dur": 6,
    "args": {
      "External id": 16067,
      "cbid": 211, "correlation": 16067
    }
  },
  {
    "ph": "s", "id": 16067, "pid": 76337, "tid": -914061504, "ts": 1716454216865043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216865166, "dur": 1287,
    "args": {
      "External id": 16069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16069,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16069, "pid": 5, "tid": 7, "ts": 1716454216865166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216865058, "dur": 11,
    "args": {
      "External id": 16069,
      "cbid": 211, "correlation": 16069
    }
  },
  {
    "ph": "s", "id": 16069, "pid": 76337, "tid": -914061504, "ts": 1716454216865058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216866454, "dur": 20,
    "args": {
      "External id": 16071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16071,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16071, "pid": 5, "tid": 7, "ts": 1716454216866454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216865074, "dur": 8,
    "args": {
      "External id": 16071,
      "cbid": 211, "correlation": 16071
    }
  },
  {
    "ph": "s", "id": 16071, "pid": 76337, "tid": -914061504, "ts": 1716454216865074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216866476, "dur": 34,
    "args": {
      "External id": 16077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16077,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16077, "pid": 5, "tid": 7, "ts": 1716454216866476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216865803, "dur": 11,
    "args": {
      "External id": 16077,
      "cbid": 211, "correlation": 16077
    }
  },
  {
    "ph": "s", "id": 16077, "pid": 76337, "tid": -914061504, "ts": 1716454216865803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216866511, "dur": 3,
    "args": {
      "External id": 16085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16085,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 16085, "pid": 5, "tid": 7, "ts": 1716454216866511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216865887, "dur": 11,
    "args": {
      "External id": 16085,
      "cbid": 211, "correlation": 16085
    }
  },
  {
    "ph": "s", "id": 16085, "pid": 76337, "tid": -914061504, "ts": 1716454216865887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866026, "dur": 3,
    "args": {
      "External id": 16101,
      "cbid": 251, "correlation": 16101
    }
  },
  {
    "ph": "f", "id": 16101, "pid": 76337, "tid": -914061504, "ts": 1716454216866026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866035, "dur": 0,
    "args": {
      "External id": 16103,
      "cbid": 251, "correlation": 16103
    }
  },
  {
    "ph": "f", "id": 16103, "pid": 76337, "tid": -914061504, "ts": 1716454216866035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216866516, "dur": 12,
    "args": {
      "External id": 16104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16104,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 16104, "pid": 5, "tid": 7, "ts": 1716454216866516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866038, "dur": 16,
    "args": {
      "External id": 16104,
      "cbid": 211, "correlation": 16104
    }
  },
  {
    "ph": "s", "id": 16104, "pid": 76337, "tid": -914061504, "ts": 1716454216866038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216866529, "dur": 5,
    "args": {
      "External id": 16106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16106,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 16106, "pid": 5, "tid": 7, "ts": 1716454216866529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866058, "dur": 8,
    "args": {
      "External id": 16106,
      "cbid": 211, "correlation": 16106
    }
  },
  {
    "ph": "s", "id": 16106, "pid": 76337, "tid": -914061504, "ts": 1716454216866058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216866535, "dur": 30,
    "args": {
      "External id": 16116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16116,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16116, "pid": 5, "tid": 7, "ts": 1716454216866535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866147, "dur": 12,
    "args": {
      "External id": 16116,
      "cbid": 211, "correlation": 16116
    }
  },
  {
    "ph": "s", "id": 16116, "pid": 76337, "tid": -914061504, "ts": 1716454216866147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216866567, "dur": 31,
    "args": {
      "External id": 16136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16136,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 16136, "pid": 5, "tid": 7, "ts": 1716454216866567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866242, "dur": 12,
    "args": {
      "External id": 16136,
      "cbid": 211, "correlation": 16136
    }
  },
  {
    "ph": "s", "id": 16136, "pid": 76337, "tid": -914061504, "ts": 1716454216866242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216866599, "dur": 4,
    "args": {
      "External id": 16148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16148,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 16148, "pid": 5, "tid": 7, "ts": 1716454216866599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866265, "dur": 9,
    "args": {
      "External id": 16148,
      "cbid": 211, "correlation": 16148
    }
  },
  {
    "ph": "s", "id": 16148, "pid": 76337, "tid": -914061504, "ts": 1716454216866265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216866605, "dur": 30,
    "args": {
      "External id": 16151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16151, "pid": 5, "tid": 7, "ts": 1716454216866605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866295, "dur": 7,
    "args": {
      "External id": 16151,
      "cbid": 211, "correlation": 16151
    }
  },
  {
    "ph": "s", "id": 16151, "pid": 76337, "tid": -914061504, "ts": 1716454216866295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216866636, "dur": 21,
    "args": {
      "External id": 16160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16160,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16160, "pid": 5, "tid": 7, "ts": 1716454216866636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866340, "dur": 10,
    "args": {
      "External id": 16160,
      "cbid": 211, "correlation": 16160
    }
  },
  {
    "ph": "s", "id": 16160, "pid": 76337, "tid": -914061504, "ts": 1716454216866340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866448, "dur": 0,
    "args": {
      "External id": 16170,
      "cbid": 317, "correlation": 16170
    }
  },
  {
    "ph": "f", "id": 16170, "pid": 76337, "tid": -914061504, "ts": 1716454216866448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866449, "dur": 1,
    "args": {
      "External id": 16171,
      "cbid": 203, "correlation": 16171
    }
  },
  {
    "ph": "f", "id": 16171, "pid": 76337, "tid": -914061504, "ts": 1716454216866449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866451, "dur": 0,
    "args": {
      "External id": 16172,
      "cbid": 205, "correlation": 16172
    }
  },
  {
    "ph": "f", "id": 16172, "pid": 76337, "tid": -914061504, "ts": 1716454216866451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216866658, "dur": 23,
    "args": {
      "External id": 16176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16176,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16176, "pid": 5, "tid": 7, "ts": 1716454216866658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866469, "dur": 16,
    "args": {
      "External id": 16176,
      "cbid": 211, "correlation": 16176
    }
  },
  {
    "ph": "s", "id": 16176, "pid": 76337, "tid": -914061504, "ts": 1716454216866469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216866682, "dur": 44,
    "args": {
      "External id": 16178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16178, "pid": 5, "tid": 7, "ts": 1716454216866682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866487, "dur": 5,
    "args": {
      "External id": 16178,
      "cbid": 211, "correlation": 16178
    }
  },
  {
    "ph": "s", "id": 16178, "pid": 76337, "tid": -914061504, "ts": 1716454216866487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216866728, "dur": 647,
    "args": {
      "External id": 16180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16180,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16180, "pid": 5, "tid": 7, "ts": 1716454216866728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866500, "dur": 7,
    "args": {
      "External id": 16180,
      "cbid": 211, "correlation": 16180
    }
  },
  {
    "ph": "s", "id": 16180, "pid": 76337, "tid": -914061504, "ts": 1716454216866500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216867376, "dur": 22,
    "args": {
      "External id": 16182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16182,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16182, "pid": 5, "tid": 7, "ts": 1716454216867376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866511, "dur": 5,
    "args": {
      "External id": 16182,
      "cbid": 211, "correlation": 16182
    }
  },
  {
    "ph": "s", "id": 16182, "pid": 76337, "tid": -914061504, "ts": 1716454216866511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216867399, "dur": 33,
    "args": {
      "External id": 16188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16188,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16188, "pid": 5, "tid": 7, "ts": 1716454216867399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866543, "dur": 9,
    "args": {
      "External id": 16188,
      "cbid": 211, "correlation": 16188
    }
  },
  {
    "ph": "s", "id": 16188, "pid": 76337, "tid": -914061504, "ts": 1716454216866543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866607, "dur": 0,
    "args": {
      "External id": 16198,
      "cbid": 317, "correlation": 16198
    }
  },
  {
    "ph": "f", "id": 16198, "pid": 76337, "tid": -914061504, "ts": 1716454216866607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866608, "dur": 0,
    "args": {
      "External id": 16199,
      "cbid": 203, "correlation": 16199
    }
  },
  {
    "ph": "f", "id": 16199, "pid": 76337, "tid": -914061504, "ts": 1716454216866608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216866609, "dur": 0,
    "args": {
      "External id": 16200,
      "cbid": 205, "correlation": 16200
    }
  },
  {
    "ph": "f", "id": 16200, "pid": 76337, "tid": -914061504, "ts": 1716454216866609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216868364, "dur": 38,
    "args": {
      "External id": 16204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16204,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16204, "pid": 5, "tid": 7, "ts": 1716454216868364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216868344, "dur": 21,
    "args": {
      "External id": 16204,
      "cbid": 211, "correlation": 16204
    }
  },
  {
    "ph": "s", "id": 16204, "pid": 76337, "tid": -914061504, "ts": 1716454216868344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216868404, "dur": 192,
    "args": {
      "External id": 16206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16206,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16206, "pid": 5, "tid": 7, "ts": 1716454216868404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216868373, "dur": 10,
    "args": {
      "External id": 16206,
      "cbid": 211, "correlation": 16206
    }
  },
  {
    "ph": "s", "id": 16206, "pid": 76337, "tid": -914061504, "ts": 1716454216868373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216868598, "dur": 22,
    "args": {
      "External id": 16208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16208,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16208, "pid": 5, "tid": 7, "ts": 1716454216868598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216868387, "dur": 6,
    "args": {
      "External id": 16208,
      "cbid": 211, "correlation": 16208
    }
  },
  {
    "ph": "s", "id": 16208, "pid": 76337, "tid": -914061504, "ts": 1716454216868387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216868859, "dur": 33,
    "args": {
      "External id": 16214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16214, "pid": 5, "tid": 7, "ts": 1716454216868859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216868847, "dur": 11,
    "args": {
      "External id": 16214,
      "cbid": 211, "correlation": 16214
    }
  },
  {
    "ph": "s", "id": 16214, "pid": 76337, "tid": -914061504, "ts": 1716454216868847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216868901, "dur": 27,
    "args": {
      "External id": 16222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16222,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16222, "pid": 5, "tid": 7, "ts": 1716454216868901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216868891, "dur": 9,
    "args": {
      "External id": 16222,
      "cbid": 211, "correlation": 16222
    }
  },
  {
    "ph": "s", "id": 16222, "pid": 76337, "tid": -914061504, "ts": 1716454216868891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216868954, "dur": 19,
    "args": {
      "External id": 16230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16230,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16230, "pid": 5, "tid": 7, "ts": 1716454216868954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216868944, "dur": 10,
    "args": {
      "External id": 16230,
      "cbid": 211, "correlation": 16230
    }
  },
  {
    "ph": "s", "id": 16230, "pid": 76337, "tid": -914061504, "ts": 1716454216868944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216869084, "dur": 30,
    "args": {
      "External id": 16250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16250,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 16250, "pid": 5, "tid": 7, "ts": 1716454216869084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869070, "dur": 14,
    "args": {
      "External id": 16250,
      "cbid": 211, "correlation": 16250
    }
  },
  {
    "ph": "s", "id": 16250, "pid": 76337, "tid": -914061504, "ts": 1716454216869070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216869118, "dur": 4,
    "args": {
      "External id": 16262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16262,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 16262, "pid": 5, "tid": 7, "ts": 1716454216869118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869107, "dur": 9,
    "args": {
      "External id": 16262,
      "cbid": 211, "correlation": 16262
    }
  },
  {
    "ph": "s", "id": 16262, "pid": 76337, "tid": -914061504, "ts": 1716454216869107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216869140, "dur": 30,
    "args": {
      "External id": 16265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16265, "pid": 5, "tid": 7, "ts": 1716454216869140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869132, "dur": 7,
    "args": {
      "External id": 16265,
      "cbid": 211, "correlation": 16265
    }
  },
  {
    "ph": "s", "id": 16265, "pid": 76337, "tid": -914061504, "ts": 1716454216869132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869204, "dur": 0,
    "args": {
      "External id": 16276,
      "cbid": 317, "correlation": 16276
    }
  },
  {
    "ph": "f", "id": 16276, "pid": 76337, "tid": -914061504, "ts": 1716454216869204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869205, "dur": 0,
    "args": {
      "External id": 16277,
      "cbid": 203, "correlation": 16277
    }
  },
  {
    "ph": "f", "id": 16277, "pid": 76337, "tid": -914061504, "ts": 1716454216869205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869206, "dur": 0,
    "args": {
      "External id": 16278,
      "cbid": 205, "correlation": 16278
    }
  },
  {
    "ph": "f", "id": 16278, "pid": 76337, "tid": -914061504, "ts": 1716454216869206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216869235, "dur": 22,
    "args": {
      "External id": 16282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16282,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16282, "pid": 5, "tid": 7, "ts": 1716454216869235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869222, "dur": 12,
    "args": {
      "External id": 16282,
      "cbid": 211, "correlation": 16282
    }
  },
  {
    "ph": "s", "id": 16282, "pid": 76337, "tid": -914061504, "ts": 1716454216869222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216869258, "dur": 105,
    "args": {
      "External id": 16284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16284,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16284, "pid": 5, "tid": 7, "ts": 1716454216869258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869240, "dur": 7,
    "args": {
      "External id": 16284,
      "cbid": 211, "correlation": 16284
    }
  },
  {
    "ph": "s", "id": 16284, "pid": 76337, "tid": -914061504, "ts": 1716454216869240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216869365, "dur": 22,
    "args": {
      "External id": 16286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16286,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16286, "pid": 5, "tid": 7, "ts": 1716454216869365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869251, "dur": 5,
    "args": {
      "External id": 16286,
      "cbid": 211, "correlation": 16286
    }
  },
  {
    "ph": "s", "id": 16286, "pid": 76337, "tid": -914061504, "ts": 1716454216869251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216869388, "dur": 32,
    "args": {
      "External id": 16292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16292, "pid": 5, "tid": 7, "ts": 1716454216869388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869280, "dur": 9,
    "args": {
      "External id": 16292,
      "cbid": 211, "correlation": 16292
    }
  },
  {
    "ph": "s", "id": 16292, "pid": 76337, "tid": -914061504, "ts": 1716454216869280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216869441, "dur": 82,
    "args": {
      "External id": 16303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16303,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16303, "pid": 5, "tid": 7, "ts": 1716454216869441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869423, "dur": 18,
    "args": {
      "External id": 16303,
      "cbid": 211, "correlation": 16303
    }
  },
  {
    "ph": "s", "id": 16303, "pid": 76337, "tid": -914061504, "ts": 1716454216869423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216869523, "dur": 37,
    "args": {
      "External id": 16325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16325,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16325, "pid": 5, "tid": 7, "ts": 1716454216869523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869464, "dur": 10,
    "args": {
      "External id": 16325,
      "cbid": 211, "correlation": 16325
    }
  },
  {
    "ph": "s", "id": 16325, "pid": 76337, "tid": -914061504, "ts": 1716454216869464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869621, "dur": 3,
    "args": {
      "External id": 16336,
      "cbid": 251, "correlation": 16336
    }
  },
  {
    "ph": "f", "id": 16336, "pid": 76337, "tid": -914061504, "ts": 1716454216869621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216869648, "dur": 95,
    "args": {
      "External id": 16337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16337,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16337, "pid": 5, "tid": 7, "ts": 1716454216869648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869631, "dur": 17,
    "args": {
      "External id": 16337,
      "cbid": 211, "correlation": 16337
    }
  },
  {
    "ph": "s", "id": 16337, "pid": 76337, "tid": -914061504, "ts": 1716454216869631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869711, "dur": 1,
    "args": {
      "External id": 16348,
      "cbid": 251, "correlation": 16348
    }
  },
  {
    "ph": "f", "id": 16348, "pid": 76337, "tid": -914061504, "ts": 1716454216869711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216869745, "dur": 89,
    "args": {
      "External id": 16349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16349,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16349, "pid": 5, "tid": 7, "ts": 1716454216869745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869715, "dur": 12,
    "args": {
      "External id": 16349,
      "cbid": 211, "correlation": 16349
    }
  },
  {
    "ph": "s", "id": 16349, "pid": 76337, "tid": -914061504, "ts": 1716454216869715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869780, "dur": 1,
    "args": {
      "External id": 16360,
      "cbid": 251, "correlation": 16360
    }
  },
  {
    "ph": "f", "id": 16360, "pid": 76337, "tid": -914061504, "ts": 1716454216869780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216869835, "dur": 91,
    "args": {
      "External id": 16361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16361,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16361, "pid": 5, "tid": 7, "ts": 1716454216869835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869785, "dur": 12,
    "args": {
      "External id": 16361,
      "cbid": 211, "correlation": 16361
    }
  },
  {
    "ph": "s", "id": 16361, "pid": 76337, "tid": -914061504, "ts": 1716454216869785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216869927, "dur": 557,
    "args": {
      "External id": 16382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16382,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 16382, "pid": 5, "tid": 7, "ts": 1716454216869927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216869896, "dur": 16,
    "args": {
      "External id": 16382,
      "cbid": 211, "correlation": 16382
    }
  },
  {
    "ph": "s", "id": 16382, "pid": 76337, "tid": -914061504, "ts": 1716454216869896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870047, "dur": 2,
    "args": {
      "External id": 16400,
      "cbid": 251, "correlation": 16400
    }
  },
  {
    "ph": "f", "id": 16400, "pid": 76337, "tid": -914061504, "ts": 1716454216870047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216870485, "dur": 99,
    "args": {
      "External id": 16402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16402,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16402, "pid": 5, "tid": 7, "ts": 1716454216870485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870054, "dur": 15,
    "args": {
      "External id": 16402,
      "cbid": 211, "correlation": 16402
    }
  },
  {
    "ph": "s", "id": 16402, "pid": 76337, "tid": -914061504, "ts": 1716454216870054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216870586, "dur": 20,
    "args": {
      "External id": 16410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16410,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16410, "pid": 5, "tid": 7, "ts": 1716454216870586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870139, "dur": 13,
    "args": {
      "External id": 16410,
      "cbid": 211, "correlation": 16410
    }
  },
  {
    "ph": "s", "id": 16410, "pid": 76337, "tid": -914061504, "ts": 1716454216870139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216870607, "dur": 86,
    "args": {
      "External id": 16418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16418, "pid": 5, "tid": 7, "ts": 1716454216870607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870182, "dur": 10,
    "args": {
      "External id": 16418,
      "cbid": 211, "correlation": 16418
    }
  },
  {
    "ph": "s", "id": 16418, "pid": 76337, "tid": -914061504, "ts": 1716454216870182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216870694, "dur": 35,
    "args": {
      "External id": 16440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16440,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16440, "pid": 5, "tid": 7, "ts": 1716454216870694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870237, "dur": 10,
    "args": {
      "External id": 16440,
      "cbid": 211, "correlation": 16440
    }
  },
  {
    "ph": "s", "id": 16440, "pid": 76337, "tid": -914061504, "ts": 1716454216870237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870339, "dur": 1,
    "args": {
      "External id": 16451,
      "cbid": 251, "correlation": 16451
    }
  },
  {
    "ph": "f", "id": 16451, "pid": 76337, "tid": -914061504, "ts": 1716454216870339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216870730, "dur": 90,
    "args": {
      "External id": 16452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16452,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16452, "pid": 5, "tid": 7, "ts": 1716454216870730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870344, "dur": 14,
    "args": {
      "External id": 16452,
      "cbid": 211, "correlation": 16452
    }
  },
  {
    "ph": "s", "id": 16452, "pid": 76337, "tid": -914061504, "ts": 1716454216870344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870446, "dur": 2,
    "args": {
      "External id": 16463,
      "cbid": 251, "correlation": 16463
    }
  },
  {
    "ph": "f", "id": 16463, "pid": 76337, "tid": -914061504, "ts": 1716454216870446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870451, "dur": 0,
    "args": {
      "External id": 16464,
      "cbid": 251, "correlation": 16464
    }
  },
  {
    "ph": "f", "id": 16464, "pid": 76337, "tid": -914061504, "ts": 1716454216870451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216870822, "dur": 11,
    "args": {
      "External id": 16465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16465,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 16465, "pid": 5, "tid": 7, "ts": 1716454216870822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870454, "dur": 16,
    "args": {
      "External id": 16465,
      "cbid": 211, "correlation": 16465
    }
  },
  {
    "ph": "s", "id": 16465, "pid": 76337, "tid": -914061504, "ts": 1716454216870454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216870834, "dur": 5,
    "args": {
      "External id": 16467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16467,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 16467, "pid": 5, "tid": 7, "ts": 1716454216870834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870472, "dur": 9,
    "args": {
      "External id": 16467,
      "cbid": 211, "correlation": 16467
    }
  },
  {
    "ph": "s", "id": 16467, "pid": 76337, "tid": -914061504, "ts": 1716454216870472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870535, "dur": 1,
    "args": {
      "External id": 16478,
      "cbid": 251, "correlation": 16478
    }
  },
  {
    "ph": "f", "id": 16478, "pid": 76337, "tid": -914061504, "ts": 1716454216870535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870539, "dur": 0,
    "args": {
      "External id": 16479,
      "cbid": 251, "correlation": 16479
    }
  },
  {
    "ph": "f", "id": 16479, "pid": 76337, "tid": -914061504, "ts": 1716454216870539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216870841, "dur": 7,
    "args": {
      "External id": 16480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16480,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 16480, "pid": 5, "tid": 7, "ts": 1716454216870841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870540, "dur": 12,
    "args": {
      "External id": 16480,
      "cbid": 211, "correlation": 16480
    }
  },
  {
    "ph": "s", "id": 16480, "pid": 76337, "tid": -914061504, "ts": 1716454216870540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216870849, "dur": 3,
    "args": {
      "External id": 16482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16482,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 16482, "pid": 5, "tid": 7, "ts": 1716454216870849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870554, "dur": 5,
    "args": {
      "External id": 16482,
      "cbid": 211, "correlation": 16482
    }
  },
  {
    "ph": "s", "id": 16482, "pid": 76337, "tid": -914061504, "ts": 1716454216870554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216870853, "dur": 93,
    "args": {
      "External id": 16503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16503,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 16503, "pid": 5, "tid": 7, "ts": 1716454216870853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870629, "dur": 13,
    "args": {
      "External id": 16503,
      "cbid": 211, "correlation": 16503
    }
  },
  {
    "ph": "s", "id": 16503, "pid": 76337, "tid": -914061504, "ts": 1716454216870629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870743, "dur": 1,
    "args": {
      "External id": 16521,
      "cbid": 251, "correlation": 16521
    }
  },
  {
    "ph": "f", "id": 16521, "pid": 76337, "tid": -914061504, "ts": 1716454216870743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216870947, "dur": 80,
    "args": {
      "External id": 16523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16523,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16523, "pid": 5, "tid": 7, "ts": 1716454216870947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870749, "dur": 14,
    "args": {
      "External id": 16523,
      "cbid": 211, "correlation": 16523
    }
  },
  {
    "ph": "s", "id": 16523, "pid": 76337, "tid": -914061504, "ts": 1716454216870749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216871029, "dur": 19,
    "args": {
      "External id": 16531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16531,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16531, "pid": 5, "tid": 7, "ts": 1716454216871029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870823, "dur": 14,
    "args": {
      "External id": 16531,
      "cbid": 211, "correlation": 16531
    }
  },
  {
    "ph": "s", "id": 16531, "pid": 76337, "tid": -914061504, "ts": 1716454216870823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216871049, "dur": 27,
    "args": {
      "External id": 16539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16539,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16539, "pid": 5, "tid": 7, "ts": 1716454216871049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870865, "dur": 9,
    "args": {
      "External id": 16539,
      "cbid": 211, "correlation": 16539
    }
  },
  {
    "ph": "s", "id": 16539, "pid": 76337, "tid": -914061504, "ts": 1716454216870865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216871078, "dur": 35,
    "args": {
      "External id": 16561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16561,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16561, "pid": 5, "tid": 7, "ts": 1716454216871078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216870930, "dur": 12,
    "args": {
      "External id": 16561,
      "cbid": 211, "correlation": 16561
    }
  },
  {
    "ph": "s", "id": 16561, "pid": 76337, "tid": -914061504, "ts": 1716454216870930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871049, "dur": 2,
    "args": {
      "External id": 16577,
      "cbid": 251, "correlation": 16577
    }
  },
  {
    "ph": "f", "id": 16577, "pid": 76337, "tid": -914061504, "ts": 1716454216871049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871055, "dur": 0,
    "args": {
      "External id": 16579,
      "cbid": 251, "correlation": 16579
    }
  },
  {
    "ph": "f", "id": 16579, "pid": 76337, "tid": -914061504, "ts": 1716454216871055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216871114, "dur": 548,
    "args": {
      "External id": 16580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16580,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 16580, "pid": 5, "tid": 7, "ts": 1716454216871114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871061, "dur": 15,
    "args": {
      "External id": 16580,
      "cbid": 211, "correlation": 16580
    }
  },
  {
    "ph": "s", "id": 16580, "pid": 76337, "tid": -914061504, "ts": 1716454216871061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216871664, "dur": 126,
    "args": {
      "External id": 16588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16588,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16588, "pid": 5, "tid": 7, "ts": 1716454216871664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871158, "dur": 15,
    "args": {
      "External id": 16588,
      "cbid": 211, "correlation": 16588
    }
  },
  {
    "ph": "s", "id": 16588, "pid": 76337, "tid": -914061504, "ts": 1716454216871158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216871791, "dur": 130,
    "args": {
      "External id": 16596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16596,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16596, "pid": 5, "tid": 7, "ts": 1716454216871791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871201, "dur": 12,
    "args": {
      "External id": 16596,
      "cbid": 211, "correlation": 16596
    }
  },
  {
    "ph": "s", "id": 16596, "pid": 76337, "tid": -914061504, "ts": 1716454216871201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871286, "dur": 1,
    "args": {
      "External id": 16612,
      "cbid": 251, "correlation": 16612
    }
  },
  {
    "ph": "f", "id": 16612, "pid": 76337, "tid": -914061504, "ts": 1716454216871286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216871923, "dur": 306,
    "args": {
      "External id": 16614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16614,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16614, "pid": 5, "tid": 7, "ts": 1716454216871923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871292, "dur": 12,
    "args": {
      "External id": 16614,
      "cbid": 211, "correlation": 16614
    }
  },
  {
    "ph": "s", "id": 16614, "pid": 76337, "tid": -914061504, "ts": 1716454216871292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216872230, "dur": 26,
    "args": {
      "External id": 16622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16622,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16622, "pid": 5, "tid": 7, "ts": 1716454216872230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871333, "dur": 10,
    "args": {
      "External id": 16622,
      "cbid": 211, "correlation": 16622
    }
  },
  {
    "ph": "s", "id": 16622, "pid": 76337, "tid": -914061504, "ts": 1716454216871333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216872258, "dur": 82,
    "args": {
      "External id": 16633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16633, "pid": 5, "tid": 7, "ts": 1716454216872258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871405, "dur": 12,
    "args": {
      "External id": 16633,
      "cbid": 211, "correlation": 16633
    }
  },
  {
    "ph": "s", "id": 16633, "pid": 76337, "tid": -914061504, "ts": 1716454216871405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871474, "dur": 0,
    "args": {
      "External id": 16645,
      "cbid": 317, "correlation": 16645
    }
  },
  {
    "ph": "f", "id": 16645, "pid": 76337, "tid": -914061504, "ts": 1716454216871474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871475, "dur": 0,
    "args": {
      "External id": 16646,
      "cbid": 203, "correlation": 16646
    }
  },
  {
    "ph": "f", "id": 16646, "pid": 76337, "tid": -914061504, "ts": 1716454216871475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871475, "dur": 0,
    "args": {
      "External id": 16647,
      "cbid": 205, "correlation": 16647
    }
  },
  {
    "ph": "f", "id": 16647, "pid": 76337, "tid": -914061504, "ts": 1716454216871475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216872342, "dur": 24,
    "args": {
      "External id": 16651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16651,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16651, "pid": 5, "tid": 7, "ts": 1716454216872342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871491, "dur": 13,
    "args": {
      "External id": 16651,
      "cbid": 211, "correlation": 16651
    }
  },
  {
    "ph": "s", "id": 16651, "pid": 76337, "tid": -914061504, "ts": 1716454216871491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216872367, "dur": 121,
    "args": {
      "External id": 16653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16653,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16653, "pid": 5, "tid": 7, "ts": 1716454216872367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871511, "dur": 7,
    "args": {
      "External id": 16653,
      "cbid": 211, "correlation": 16653
    }
  },
  {
    "ph": "s", "id": 16653, "pid": 76337, "tid": -914061504, "ts": 1716454216871511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216872489, "dur": 21,
    "args": {
      "External id": 16655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16655, "pid": 5, "tid": 7, "ts": 1716454216872489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871522, "dur": 6,
    "args": {
      "External id": 16655,
      "cbid": 211, "correlation": 16655
    }
  },
  {
    "ph": "s", "id": 16655, "pid": 76337, "tid": -914061504, "ts": 1716454216871522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216872512, "dur": 34,
    "args": {
      "External id": 16661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16661, "pid": 5, "tid": 7, "ts": 1716454216872512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871551, "dur": 9,
    "args": {
      "External id": 16661,
      "cbid": 211, "correlation": 16661
    }
  },
  {
    "ph": "s", "id": 16661, "pid": 76337, "tid": -914061504, "ts": 1716454216871551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216872547, "dur": 27,
    "args": {
      "External id": 16669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16669,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16669, "pid": 5, "tid": 7, "ts": 1716454216872547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871582, "dur": 8,
    "args": {
      "External id": 16669,
      "cbid": 211, "correlation": 16669
    }
  },
  {
    "ph": "s", "id": 16669, "pid": 76337, "tid": -914061504, "ts": 1716454216871582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216872575, "dur": 45,
    "args": {
      "External id": 16678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16678,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16678, "pid": 5, "tid": 7, "ts": 1716454216872575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871637, "dur": 13,
    "args": {
      "External id": 16678,
      "cbid": 211, "correlation": 16678
    }
  },
  {
    "ph": "s", "id": 16678, "pid": 76337, "tid": -914061504, "ts": 1716454216871637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216872622, "dur": 43,
    "args": {
      "External id": 16698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16698,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 16698, "pid": 5, "tid": 7, "ts": 1716454216872622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871732, "dur": 13,
    "args": {
      "External id": 16698,
      "cbid": 211, "correlation": 16698
    }
  },
  {
    "ph": "s", "id": 16698, "pid": 76337, "tid": -914061504, "ts": 1716454216871732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216872666, "dur": 4,
    "args": {
      "External id": 16710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16710,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 16710, "pid": 5, "tid": 7, "ts": 1716454216872666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871762, "dur": 9,
    "args": {
      "External id": 16710,
      "cbid": 211, "correlation": 16710
    }
  },
  {
    "ph": "s", "id": 16710, "pid": 76337, "tid": -914061504, "ts": 1716454216871762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216872672, "dur": 45,
    "args": {
      "External id": 16713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16713, "pid": 5, "tid": 7, "ts": 1716454216872672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871785, "dur": 7,
    "args": {
      "External id": 16713,
      "cbid": 211, "correlation": 16713
    }
  },
  {
    "ph": "s", "id": 16713, "pid": 76337, "tid": -914061504, "ts": 1716454216871785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216872718, "dur": 28,
    "args": {
      "External id": 16722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16722,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16722, "pid": 5, "tid": 7, "ts": 1716454216872718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871832, "dur": 10,
    "args": {
      "External id": 16722,
      "cbid": 211, "correlation": 16722
    }
  },
  {
    "ph": "s", "id": 16722, "pid": 76337, "tid": -914061504, "ts": 1716454216871832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871900, "dur": 0,
    "args": {
      "External id": 16732,
      "cbid": 317, "correlation": 16732
    }
  },
  {
    "ph": "f", "id": 16732, "pid": 76337, "tid": -914061504, "ts": 1716454216871900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871901, "dur": 0,
    "args": {
      "External id": 16733,
      "cbid": 203, "correlation": 16733
    }
  },
  {
    "ph": "f", "id": 16733, "pid": 76337, "tid": -914061504, "ts": 1716454216871901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216871901, "dur": 0,
    "args": {
      "External id": 16734,
      "cbid": 205, "correlation": 16734
    }
  },
  {
    "ph": "f", "id": 16734, "pid": 76337, "tid": -914061504, "ts": 1716454216871901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216874954, "dur": 30,
    "args": {
      "External id": 16738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16738, "pid": 5, "tid": 7, "ts": 1716454216874954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216874926, "dur": 29,
    "args": {
      "External id": 16738,
      "cbid": 211, "correlation": 16738
    }
  },
  {
    "ph": "s", "id": 16738, "pid": 76337, "tid": -914061504, "ts": 1716454216874926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216874986, "dur": 64,
    "args": {
      "External id": 16740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16740,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16740, "pid": 5, "tid": 7, "ts": 1716454216874986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216874958, "dur": 6,
    "args": {
      "External id": 16740,
      "cbid": 211, "correlation": 16740
    }
  },
  {
    "ph": "s", "id": 16740, "pid": 76337, "tid": -914061504, "ts": 1716454216874958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216875051, "dur": 970,
    "args": {
      "External id": 16742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16742,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16742, "pid": 5, "tid": 7, "ts": 1716454216875051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216874981, "dur": 13,
    "args": {
      "External id": 16742,
      "cbid": 211, "correlation": 16742
    }
  },
  {
    "ph": "s", "id": 16742, "pid": 76337, "tid": -914061504, "ts": 1716454216874981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216876022, "dur": 21,
    "args": {
      "External id": 16744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16744, "pid": 5, "tid": 7, "ts": 1716454216876022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216874998, "dur": 6,
    "args": {
      "External id": 16744,
      "cbid": 211, "correlation": 16744
    }
  },
  {
    "ph": "s", "id": 16744, "pid": 76337, "tid": -914061504, "ts": 1716454216874998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216876044, "dur": 34,
    "args": {
      "External id": 16750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16750, "pid": 5, "tid": 7, "ts": 1716454216876044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216875726, "dur": 11,
    "args": {
      "External id": 16750,
      "cbid": 211, "correlation": 16750
    }
  },
  {
    "ph": "s", "id": 16750, "pid": 76337, "tid": -914061504, "ts": 1716454216875726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216876079, "dur": 4,
    "args": {
      "External id": 16758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16758,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 16758, "pid": 5, "tid": 7, "ts": 1716454216876079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216875808, "dur": 10,
    "args": {
      "External id": 16758,
      "cbid": 211, "correlation": 16758
    }
  },
  {
    "ph": "s", "id": 16758, "pid": 76337, "tid": -914061504, "ts": 1716454216875808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216875936, "dur": 3,
    "args": {
      "External id": 16774,
      "cbid": 251, "correlation": 16774
    }
  },
  {
    "ph": "f", "id": 16774, "pid": 76337, "tid": -914061504, "ts": 1716454216875936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216875944, "dur": 0,
    "args": {
      "External id": 16776,
      "cbid": 251, "correlation": 16776
    }
  },
  {
    "ph": "f", "id": 16776, "pid": 76337, "tid": -914061504, "ts": 1716454216875944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216876084, "dur": 12,
    "args": {
      "External id": 16777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16777,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 16777, "pid": 5, "tid": 7, "ts": 1716454216876084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216875947, "dur": 16,
    "args": {
      "External id": 16777,
      "cbid": 211, "correlation": 16777
    }
  },
  {
    "ph": "s", "id": 16777, "pid": 76337, "tid": -914061504, "ts": 1716454216875947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216876098, "dur": 5,
    "args": {
      "External id": 16779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16779,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 16779, "pid": 5, "tid": 7, "ts": 1716454216876098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216875967, "dur": 16,
    "args": {
      "External id": 16779,
      "cbid": 211, "correlation": 16779
    }
  },
  {
    "ph": "s", "id": 16779, "pid": 76337, "tid": -914061504, "ts": 1716454216875967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216876104, "dur": 30,
    "args": {
      "External id": 16789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16789,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16789, "pid": 5, "tid": 7, "ts": 1716454216876104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876063, "dur": 13,
    "args": {
      "External id": 16789,
      "cbid": 211, "correlation": 16789
    }
  },
  {
    "ph": "s", "id": 16789, "pid": 76337, "tid": -914061504, "ts": 1716454216876063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216876169, "dur": 31,
    "args": {
      "External id": 16809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16809,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 16809, "pid": 5, "tid": 7, "ts": 1716454216876169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876156, "dur": 12,
    "args": {
      "External id": 16809,
      "cbid": 211, "correlation": 16809
    }
  },
  {
    "ph": "s", "id": 16809, "pid": 76337, "tid": -914061504, "ts": 1716454216876156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216876202, "dur": 4,
    "args": {
      "External id": 16821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16821,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 16821, "pid": 5, "tid": 7, "ts": 1716454216876202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876179, "dur": 7,
    "args": {
      "External id": 16821,
      "cbid": 211, "correlation": 16821
    }
  },
  {
    "ph": "s", "id": 16821, "pid": 76337, "tid": -914061504, "ts": 1716454216876179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216876216, "dur": 30,
    "args": {
      "External id": 16824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16824, "pid": 5, "tid": 7, "ts": 1716454216876216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876208, "dur": 7,
    "args": {
      "External id": 16824,
      "cbid": 211, "correlation": 16824
    }
  },
  {
    "ph": "s", "id": 16824, "pid": 76337, "tid": -914061504, "ts": 1716454216876208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216876262, "dur": 21,
    "args": {
      "External id": 16833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16833,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16833, "pid": 5, "tid": 7, "ts": 1716454216876262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876251, "dur": 10,
    "args": {
      "External id": 16833,
      "cbid": 211, "correlation": 16833
    }
  },
  {
    "ph": "s", "id": 16833, "pid": 76337, "tid": -914061504, "ts": 1716454216876251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876354, "dur": 0,
    "args": {
      "External id": 16843,
      "cbid": 317, "correlation": 16843
    }
  },
  {
    "ph": "f", "id": 16843, "pid": 76337, "tid": -914061504, "ts": 1716454216876354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876355, "dur": 1,
    "args": {
      "External id": 16844,
      "cbid": 203, "correlation": 16844
    }
  },
  {
    "ph": "f", "id": 16844, "pid": 76337, "tid": -914061504, "ts": 1716454216876355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876356, "dur": 0,
    "args": {
      "External id": 16845,
      "cbid": 205, "correlation": 16845
    }
  },
  {
    "ph": "f", "id": 16845, "pid": 76337, "tid": -914061504, "ts": 1716454216876356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216876389, "dur": 22,
    "args": {
      "External id": 16849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16849, "pid": 5, "tid": 7, "ts": 1716454216876389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876375, "dur": 13,
    "args": {
      "External id": 16849,
      "cbid": 211, "correlation": 16849
    }
  },
  {
    "ph": "s", "id": 16849, "pid": 76337, "tid": -914061504, "ts": 1716454216876375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216876412, "dur": 44,
    "args": {
      "External id": 16851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16851, "pid": 5, "tid": 7, "ts": 1716454216876412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876391, "dur": 6,
    "args": {
      "External id": 16851,
      "cbid": 211, "correlation": 16851
    }
  },
  {
    "ph": "s", "id": 16851, "pid": 76337, "tid": -914061504, "ts": 1716454216876391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216876457, "dur": 647,
    "args": {
      "External id": 16853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16853,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16853, "pid": 5, "tid": 7, "ts": 1716454216876457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876405, "dur": 6,
    "args": {
      "External id": 16853,
      "cbid": 211, "correlation": 16853
    }
  },
  {
    "ph": "s", "id": 16853, "pid": 76337, "tid": -914061504, "ts": 1716454216876405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216877106, "dur": 21,
    "args": {
      "External id": 16855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16855,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16855, "pid": 5, "tid": 7, "ts": 1716454216877106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876415, "dur": 6,
    "args": {
      "External id": 16855,
      "cbid": 211, "correlation": 16855
    }
  },
  {
    "ph": "s", "id": 16855, "pid": 76337, "tid": -914061504, "ts": 1716454216876415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216877128, "dur": 33,
    "args": {
      "External id": 16861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16861, "pid": 5, "tid": 7, "ts": 1716454216877128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876444, "dur": 10,
    "args": {
      "External id": 16861,
      "cbid": 211, "correlation": 16861
    }
  },
  {
    "ph": "s", "id": 16861, "pid": 76337, "tid": -914061504, "ts": 1716454216876444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876504, "dur": 0,
    "args": {
      "External id": 16871,
      "cbid": 317, "correlation": 16871
    }
  },
  {
    "ph": "f", "id": 16871, "pid": 76337, "tid": -914061504, "ts": 1716454216876504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876505, "dur": 0,
    "args": {
      "External id": 16872,
      "cbid": 203, "correlation": 16872
    }
  },
  {
    "ph": "f", "id": 16872, "pid": 76337, "tid": -914061504, "ts": 1716454216876505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216876506, "dur": 0,
    "args": {
      "External id": 16873,
      "cbid": 205, "correlation": 16873
    }
  },
  {
    "ph": "f", "id": 16873, "pid": 76337, "tid": -914061504, "ts": 1716454216876506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216878243, "dur": 30,
    "args": {
      "External id": 16877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16877,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16877, "pid": 5, "tid": 7, "ts": 1716454216878243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878224, "dur": 19,
    "args": {
      "External id": 16877,
      "cbid": 211, "correlation": 16877
    }
  },
  {
    "ph": "s", "id": 16877, "pid": 76337, "tid": -914061504, "ts": 1716454216878224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216878274, "dur": 153,
    "args": {
      "External id": 16879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16879,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16879, "pid": 5, "tid": 7, "ts": 1716454216878274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878252, "dur": 10,
    "args": {
      "External id": 16879,
      "cbid": 211, "correlation": 16879
    }
  },
  {
    "ph": "s", "id": 16879, "pid": 76337, "tid": -914061504, "ts": 1716454216878252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216878428, "dur": 21,
    "args": {
      "External id": 16881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16881, "pid": 5, "tid": 7, "ts": 1716454216878428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878266, "dur": 6,
    "args": {
      "External id": 16881,
      "cbid": 211, "correlation": 16881
    }
  },
  {
    "ph": "s", "id": 16881, "pid": 76337, "tid": -914061504, "ts": 1716454216878266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216878729, "dur": 32,
    "args": {
      "External id": 16887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16887, "pid": 5, "tid": 7, "ts": 1716454216878729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878717, "dur": 10,
    "args": {
      "External id": 16887,
      "cbid": 211, "correlation": 16887
    }
  },
  {
    "ph": "s", "id": 16887, "pid": 76337, "tid": -914061504, "ts": 1716454216878717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216878767, "dur": 27,
    "args": {
      "External id": 16895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16895,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16895, "pid": 5, "tid": 7, "ts": 1716454216878767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878756, "dur": 10,
    "args": {
      "External id": 16895,
      "cbid": 211, "correlation": 16895
    }
  },
  {
    "ph": "s", "id": 16895, "pid": 76337, "tid": -914061504, "ts": 1716454216878756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216878818, "dur": 20,
    "args": {
      "External id": 16903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16903,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16903, "pid": 5, "tid": 7, "ts": 1716454216878818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878807, "dur": 10,
    "args": {
      "External id": 16903,
      "cbid": 211, "correlation": 16903
    }
  },
  {
    "ph": "s", "id": 16903, "pid": 76337, "tid": -914061504, "ts": 1716454216878807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216878937, "dur": 31,
    "args": {
      "External id": 16923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16923,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 16923, "pid": 5, "tid": 7, "ts": 1716454216878937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878923, "dur": 13,
    "args": {
      "External id": 16923,
      "cbid": 211, "correlation": 16923
    }
  },
  {
    "ph": "s", "id": 16923, "pid": 76337, "tid": -914061504, "ts": 1716454216878923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216878969, "dur": 4,
    "args": {
      "External id": 16935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16935,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 16935, "pid": 5, "tid": 7, "ts": 1716454216878969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878947, "dur": 7,
    "args": {
      "External id": 16935,
      "cbid": 211, "correlation": 16935
    }
  },
  {
    "ph": "s", "id": 16935, "pid": 76337, "tid": -914061504, "ts": 1716454216878947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216878984, "dur": 31,
    "args": {
      "External id": 16938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16938,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16938, "pid": 5, "tid": 7, "ts": 1716454216878984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216878968, "dur": 15,
    "args": {
      "External id": 16938,
      "cbid": 211, "correlation": 16938
    }
  },
  {
    "ph": "s", "id": 16938, "pid": 76337, "tid": -914061504, "ts": 1716454216878968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879041, "dur": 0,
    "args": {
      "External id": 16949,
      "cbid": 317, "correlation": 16949
    }
  },
  {
    "ph": "f", "id": 16949, "pid": 76337, "tid": -914061504, "ts": 1716454216879041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879042, "dur": 0,
    "args": {
      "External id": 16950,
      "cbid": 203, "correlation": 16950
    }
  },
  {
    "ph": "f", "id": 16950, "pid": 76337, "tid": -914061504, "ts": 1716454216879042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879043, "dur": 0,
    "args": {
      "External id": 16951,
      "cbid": 205, "correlation": 16951
    }
  },
  {
    "ph": "f", "id": 16951, "pid": 76337, "tid": -914061504, "ts": 1716454216879043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216879072, "dur": 21,
    "args": {
      "External id": 16955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16955,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16955, "pid": 5, "tid": 7, "ts": 1716454216879072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879060, "dur": 12,
    "args": {
      "External id": 16955,
      "cbid": 211, "correlation": 16955
    }
  },
  {
    "ph": "s", "id": 16955, "pid": 76337, "tid": -914061504, "ts": 1716454216879060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216879094, "dur": 106,
    "args": {
      "External id": 16957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16957,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 16957, "pid": 5, "tid": 7, "ts": 1716454216879094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879078, "dur": 7,
    "args": {
      "External id": 16957,
      "cbid": 211, "correlation": 16957
    }
  },
  {
    "ph": "s", "id": 16957, "pid": 76337, "tid": -914061504, "ts": 1716454216879078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216879201, "dur": 22,
    "args": {
      "External id": 16959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16959,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16959, "pid": 5, "tid": 7, "ts": 1716454216879201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879089, "dur": 5,
    "args": {
      "External id": 16959,
      "cbid": 211, "correlation": 16959
    }
  },
  {
    "ph": "s", "id": 16959, "pid": 76337, "tid": -914061504, "ts": 1716454216879089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216879225, "dur": 32,
    "args": {
      "External id": 16965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16965, "pid": 5, "tid": 7, "ts": 1716454216879225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879117, "dur": 9,
    "args": {
      "External id": 16965,
      "cbid": 211, "correlation": 16965
    }
  },
  {
    "ph": "s", "id": 16965, "pid": 76337, "tid": -914061504, "ts": 1716454216879117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216879294, "dur": 81,
    "args": {
      "External id": 16976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16976,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16976, "pid": 5, "tid": 7, "ts": 1716454216879294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879276, "dur": 19,
    "args": {
      "External id": 16976,
      "cbid": 211, "correlation": 16976
    }
  },
  {
    "ph": "s", "id": 16976, "pid": 76337, "tid": -914061504, "ts": 1716454216879276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216879377, "dur": 36,
    "args": {
      "External id": 16998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 16998,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 16998, "pid": 5, "tid": 7, "ts": 1716454216879377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879317, "dur": 11,
    "args": {
      "External id": 16998,
      "cbid": 211, "correlation": 16998
    }
  },
  {
    "ph": "s", "id": 16998, "pid": 76337, "tid": -914061504, "ts": 1716454216879317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879480, "dur": 2,
    "args": {
      "External id": 17009,
      "cbid": 251, "correlation": 17009
    }
  },
  {
    "ph": "f", "id": 17009, "pid": 76337, "tid": -914061504, "ts": 1716454216879480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216879506, "dur": 97,
    "args": {
      "External id": 17010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17010,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17010, "pid": 5, "tid": 7, "ts": 1716454216879506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879490, "dur": 16,
    "args": {
      "External id": 17010,
      "cbid": 211, "correlation": 17010
    }
  },
  {
    "ph": "s", "id": 17010, "pid": 76337, "tid": -914061504, "ts": 1716454216879490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879584, "dur": 1,
    "args": {
      "External id": 17021,
      "cbid": 251, "correlation": 17021
    }
  },
  {
    "ph": "f", "id": 17021, "pid": 76337, "tid": -914061504, "ts": 1716454216879584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216879604, "dur": 90,
    "args": {
      "External id": 17022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17022,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17022, "pid": 5, "tid": 7, "ts": 1716454216879604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879588, "dur": 13,
    "args": {
      "External id": 17022,
      "cbid": 211, "correlation": 17022
    }
  },
  {
    "ph": "s", "id": 17022, "pid": 76337, "tid": -914061504, "ts": 1716454216879588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879658, "dur": 1,
    "args": {
      "External id": 17033,
      "cbid": 251, "correlation": 17033
    }
  },
  {
    "ph": "f", "id": 17033, "pid": 76337, "tid": -914061504, "ts": 1716454216879658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216879696, "dur": 90,
    "args": {
      "External id": 17034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17034,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17034, "pid": 5, "tid": 7, "ts": 1716454216879696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879662, "dur": 12,
    "args": {
      "External id": 17034,
      "cbid": 211, "correlation": 17034
    }
  },
  {
    "ph": "s", "id": 17034, "pid": 76337, "tid": -914061504, "ts": 1716454216879662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216879789, "dur": 559,
    "args": {
      "External id": 17055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17055,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17055, "pid": 5, "tid": 7, "ts": 1716454216879789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879772, "dur": 17,
    "args": {
      "External id": 17055,
      "cbid": 211, "correlation": 17055
    }
  },
  {
    "ph": "s", "id": 17055, "pid": 76337, "tid": -914061504, "ts": 1716454216879772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879915, "dur": 2,
    "args": {
      "External id": 17073,
      "cbid": 251, "correlation": 17073
    }
  },
  {
    "ph": "f", "id": 17073, "pid": 76337, "tid": -914061504, "ts": 1716454216879915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216880350, "dur": 99,
    "args": {
      "External id": 17075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17075,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17075, "pid": 5, "tid": 7, "ts": 1716454216880350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216879922, "dur": 15,
    "args": {
      "External id": 17075,
      "cbid": 211, "correlation": 17075
    }
  },
  {
    "ph": "s", "id": 17075, "pid": 76337, "tid": -914061504, "ts": 1716454216879922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216880450, "dur": 20,
    "args": {
      "External id": 17083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17083,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17083, "pid": 5, "tid": 7, "ts": 1716454216880450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880017, "dur": 13,
    "args": {
      "External id": 17083,
      "cbid": 211, "correlation": 17083
    }
  },
  {
    "ph": "s", "id": 17083, "pid": 76337, "tid": -914061504, "ts": 1716454216880017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216880471, "dur": 86,
    "args": {
      "External id": 17091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17091, "pid": 5, "tid": 7, "ts": 1716454216880471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880061, "dur": 10,
    "args": {
      "External id": 17091,
      "cbid": 211, "correlation": 17091
    }
  },
  {
    "ph": "s", "id": 17091, "pid": 76337, "tid": -914061504, "ts": 1716454216880061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216880559, "dur": 35,
    "args": {
      "External id": 17113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17113,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17113, "pid": 5, "tid": 7, "ts": 1716454216880559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880115, "dur": 10,
    "args": {
      "External id": 17113,
      "cbid": 211, "correlation": 17113
    }
  },
  {
    "ph": "s", "id": 17113, "pid": 76337, "tid": -914061504, "ts": 1716454216880115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880213, "dur": 1,
    "args": {
      "External id": 17124,
      "cbid": 251, "correlation": 17124
    }
  },
  {
    "ph": "f", "id": 17124, "pid": 76337, "tid": -914061504, "ts": 1716454216880213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216880595, "dur": 91,
    "args": {
      "External id": 17125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17125,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17125, "pid": 5, "tid": 7, "ts": 1716454216880595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880218, "dur": 14,
    "args": {
      "External id": 17125,
      "cbid": 211, "correlation": 17125
    }
  },
  {
    "ph": "s", "id": 17125, "pid": 76337, "tid": -914061504, "ts": 1716454216880218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880315, "dur": 1,
    "args": {
      "External id": 17136,
      "cbid": 251, "correlation": 17136
    }
  },
  {
    "ph": "f", "id": 17136, "pid": 76337, "tid": -914061504, "ts": 1716454216880315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880320, "dur": 0,
    "args": {
      "External id": 17137,
      "cbid": 251, "correlation": 17137
    }
  },
  {
    "ph": "f", "id": 17137, "pid": 76337, "tid": -914061504, "ts": 1716454216880320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216880687, "dur": 11,
    "args": {
      "External id": 17138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17138,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 17138, "pid": 5, "tid": 7, "ts": 1716454216880687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880322, "dur": 14,
    "args": {
      "External id": 17138,
      "cbid": 211, "correlation": 17138
    }
  },
  {
    "ph": "s", "id": 17138, "pid": 76337, "tid": -914061504, "ts": 1716454216880322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216880699, "dur": 5,
    "args": {
      "External id": 17140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17140,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 17140, "pid": 5, "tid": 7, "ts": 1716454216880699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880339, "dur": 8,
    "args": {
      "External id": 17140,
      "cbid": 211, "correlation": 17140
    }
  },
  {
    "ph": "s", "id": 17140, "pid": 76337, "tid": -914061504, "ts": 1716454216880339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880401, "dur": 1,
    "args": {
      "External id": 17151,
      "cbid": 251, "correlation": 17151
    }
  },
  {
    "ph": "f", "id": 17151, "pid": 76337, "tid": -914061504, "ts": 1716454216880401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880404, "dur": 0,
    "args": {
      "External id": 17152,
      "cbid": 251, "correlation": 17152
    }
  },
  {
    "ph": "f", "id": 17152, "pid": 76337, "tid": -914061504, "ts": 1716454216880404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216880706, "dur": 7,
    "args": {
      "External id": 17153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17153,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 17153, "pid": 5, "tid": 7, "ts": 1716454216880706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880406, "dur": 13,
    "args": {
      "External id": 17153,
      "cbid": 211, "correlation": 17153
    }
  },
  {
    "ph": "s", "id": 17153, "pid": 76337, "tid": -914061504, "ts": 1716454216880406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216880714, "dur": 3,
    "args": {
      "External id": 17155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17155,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 17155, "pid": 5, "tid": 7, "ts": 1716454216880714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880420, "dur": 5,
    "args": {
      "External id": 17155,
      "cbid": 211, "correlation": 17155
    }
  },
  {
    "ph": "s", "id": 17155, "pid": 76337, "tid": -914061504, "ts": 1716454216880420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216880719, "dur": 93,
    "args": {
      "External id": 17176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17176,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17176, "pid": 5, "tid": 7, "ts": 1716454216880719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880494, "dur": 12,
    "args": {
      "External id": 17176,
      "cbid": 211, "correlation": 17176
    }
  },
  {
    "ph": "s", "id": 17176, "pid": 76337, "tid": -914061504, "ts": 1716454216880494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880593, "dur": 1,
    "args": {
      "External id": 17194,
      "cbid": 251, "correlation": 17194
    }
  },
  {
    "ph": "f", "id": 17194, "pid": 76337, "tid": -914061504, "ts": 1716454216880593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216880813, "dur": 94,
    "args": {
      "External id": 17196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17196,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17196, "pid": 5, "tid": 7, "ts": 1716454216880813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880599, "dur": 13,
    "args": {
      "External id": 17196,
      "cbid": 211, "correlation": 17196
    }
  },
  {
    "ph": "s", "id": 17196, "pid": 76337, "tid": -914061504, "ts": 1716454216880599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216880909, "dur": 19,
    "args": {
      "External id": 17204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17204,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17204, "pid": 5, "tid": 7, "ts": 1716454216880909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880670, "dur": 13,
    "args": {
      "External id": 17204,
      "cbid": 211, "correlation": 17204
    }
  },
  {
    "ph": "s", "id": 17204, "pid": 76337, "tid": -914061504, "ts": 1716454216880670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216880929, "dur": 28,
    "args": {
      "External id": 17212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17212,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17212, "pid": 5, "tid": 7, "ts": 1716454216880929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880711, "dur": 8,
    "args": {
      "External id": 17212,
      "cbid": 211, "correlation": 17212
    }
  },
  {
    "ph": "s", "id": 17212, "pid": 76337, "tid": -914061504, "ts": 1716454216880711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216880958, "dur": 35,
    "args": {
      "External id": 17234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17234,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17234, "pid": 5, "tid": 7, "ts": 1716454216880958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880763, "dur": 10,
    "args": {
      "External id": 17234,
      "cbid": 211, "correlation": 17234
    }
  },
  {
    "ph": "s", "id": 17234, "pid": 76337, "tid": -914061504, "ts": 1716454216880763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880858, "dur": 2,
    "args": {
      "External id": 17250,
      "cbid": 251, "correlation": 17250
    }
  },
  {
    "ph": "f", "id": 17250, "pid": 76337, "tid": -914061504, "ts": 1716454216880858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880864, "dur": 0,
    "args": {
      "External id": 17252,
      "cbid": 251, "correlation": 17252
    }
  },
  {
    "ph": "f", "id": 17252, "pid": 76337, "tid": -914061504, "ts": 1716454216880864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216880995, "dur": 548,
    "args": {
      "External id": 17253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17253,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 17253, "pid": 5, "tid": 7, "ts": 1716454216880995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880869, "dur": 14,
    "args": {
      "External id": 17253,
      "cbid": 211, "correlation": 17253
    }
  },
  {
    "ph": "s", "id": 17253, "pid": 76337, "tid": -914061504, "ts": 1716454216880869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216881545, "dur": 127,
    "args": {
      "External id": 17261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17261,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17261, "pid": 5, "tid": 7, "ts": 1716454216881545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216880964, "dur": 22,
    "args": {
      "External id": 17261,
      "cbid": 211, "correlation": 17261
    }
  },
  {
    "ph": "s", "id": 17261, "pid": 76337, "tid": -914061504, "ts": 1716454216880964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216881673, "dur": 130,
    "args": {
      "External id": 17269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17269,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17269, "pid": 5, "tid": 7, "ts": 1716454216881673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881014, "dur": 11,
    "args": {
      "External id": 17269,
      "cbid": 211, "correlation": 17269
    }
  },
  {
    "ph": "s", "id": 17269, "pid": 76337, "tid": -914061504, "ts": 1716454216881014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881097, "dur": 1,
    "args": {
      "External id": 17285,
      "cbid": 251, "correlation": 17285
    }
  },
  {
    "ph": "f", "id": 17285, "pid": 76337, "tid": -914061504, "ts": 1716454216881097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216881804, "dur": 307,
    "args": {
      "External id": 17287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17287,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17287, "pid": 5, "tid": 7, "ts": 1716454216881804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881103, "dur": 12,
    "args": {
      "External id": 17287,
      "cbid": 211, "correlation": 17287
    }
  },
  {
    "ph": "s", "id": 17287, "pid": 76337, "tid": -914061504, "ts": 1716454216881103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216882113, "dur": 27,
    "args": {
      "External id": 17295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17295,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17295, "pid": 5, "tid": 7, "ts": 1716454216882113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881145, "dur": 11,
    "args": {
      "External id": 17295,
      "cbid": 211, "correlation": 17295
    }
  },
  {
    "ph": "s", "id": 17295, "pid": 76337, "tid": -914061504, "ts": 1716454216881145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216882141, "dur": 83,
    "args": {
      "External id": 17306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17306, "pid": 5, "tid": 7, "ts": 1716454216882141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881215, "dur": 12,
    "args": {
      "External id": 17306,
      "cbid": 211, "correlation": 17306
    }
  },
  {
    "ph": "s", "id": 17306, "pid": 76337, "tid": -914061504, "ts": 1716454216881215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881281, "dur": 0,
    "args": {
      "External id": 17318,
      "cbid": 317, "correlation": 17318
    }
  },
  {
    "ph": "f", "id": 17318, "pid": 76337, "tid": -914061504, "ts": 1716454216881281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881282, "dur": 0,
    "args": {
      "External id": 17319,
      "cbid": 203, "correlation": 17319
    }
  },
  {
    "ph": "f", "id": 17319, "pid": 76337, "tid": -914061504, "ts": 1716454216881282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881283, "dur": 0,
    "args": {
      "External id": 17320,
      "cbid": 205, "correlation": 17320
    }
  },
  {
    "ph": "f", "id": 17320, "pid": 76337, "tid": -914061504, "ts": 1716454216881283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216882225, "dur": 23,
    "args": {
      "External id": 17324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17324,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17324, "pid": 5, "tid": 7, "ts": 1716454216882225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881299, "dur": 13,
    "args": {
      "External id": 17324,
      "cbid": 211, "correlation": 17324
    }
  },
  {
    "ph": "s", "id": 17324, "pid": 76337, "tid": -914061504, "ts": 1716454216881299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216882249, "dur": 121,
    "args": {
      "External id": 17326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17326,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17326, "pid": 5, "tid": 7, "ts": 1716454216882249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881318, "dur": 6,
    "args": {
      "External id": 17326,
      "cbid": 211, "correlation": 17326
    }
  },
  {
    "ph": "s", "id": 17326, "pid": 76337, "tid": -914061504, "ts": 1716454216881318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216882371, "dur": 22,
    "args": {
      "External id": 17328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17328,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17328, "pid": 5, "tid": 7, "ts": 1716454216882371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881329, "dur": 5,
    "args": {
      "External id": 17328,
      "cbid": 211, "correlation": 17328
    }
  },
  {
    "ph": "s", "id": 17328, "pid": 76337, "tid": -914061504, "ts": 1716454216881329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216882394, "dur": 33,
    "args": {
      "External id": 17334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17334,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17334, "pid": 5, "tid": 7, "ts": 1716454216882394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881357, "dur": 9,
    "args": {
      "External id": 17334,
      "cbid": 211, "correlation": 17334
    }
  },
  {
    "ph": "s", "id": 17334, "pid": 76337, "tid": -914061504, "ts": 1716454216881357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216882429, "dur": 27,
    "args": {
      "External id": 17342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17342,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17342, "pid": 5, "tid": 7, "ts": 1716454216882429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881389, "dur": 8,
    "args": {
      "External id": 17342,
      "cbid": 211, "correlation": 17342
    }
  },
  {
    "ph": "s", "id": 17342, "pid": 76337, "tid": -914061504, "ts": 1716454216881389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454216882457, "dur": 102,
    "args": {
      "External id": 17353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17353,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17353, "pid": 5, "tid": 7, "ts": 1716454216882457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881469, "dur": 13,
    "args": {
      "External id": 17353,
      "cbid": 211, "correlation": 17353
    }
  },
  {
    "ph": "s", "id": 17353, "pid": 76337, "tid": -914061504, "ts": 1716454216881469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881525, "dur": 0,
    "args": {
      "External id": 17363,
      "cbid": 317, "correlation": 17363
    }
  },
  {
    "ph": "f", "id": 17363, "pid": 76337, "tid": -914061504, "ts": 1716454216881525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881526, "dur": 0,
    "args": {
      "External id": 17364,
      "cbid": 203, "correlation": 17364
    }
  },
  {
    "ph": "f", "id": 17364, "pid": 76337, "tid": -914061504, "ts": 1716454216881526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216881527, "dur": 0,
    "args": {
      "External id": 17365,
      "cbid": 205, "correlation": 17365
    }
  },
  {
    "ph": "f", "id": 17365, "pid": 76337, "tid": -914061504, "ts": 1716454216881527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216884392, "dur": 75,
    "args": {
      "External id": 17369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17369,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17369, "pid": 5, "tid": 7, "ts": 1716454216884392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216884366, "dur": 26,
    "args": {
      "External id": 17369,
      "cbid": 211, "correlation": 17369
    }
  },
  {
    "ph": "s", "id": 17369, "pid": 76337, "tid": -914061504, "ts": 1716454216884366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216884468, "dur": 45,
    "args": {
      "External id": 17371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17371, "pid": 5, "tid": 7, "ts": 1716454216884468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216884394, "dur": 6,
    "args": {
      "External id": 17371,
      "cbid": 211, "correlation": 17371
    }
  },
  {
    "ph": "s", "id": 17371, "pid": 76337, "tid": -914061504, "ts": 1716454216884394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216884514, "dur": 4,
    "args": {
      "External id": 17373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17373, "pid": 5, "tid": 7, "ts": 1716454216884514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216884407, "dur": 9,
    "args": {
      "External id": 17373,
      "cbid": 211, "correlation": 17373
    }
  },
  {
    "ph": "s", "id": 17373, "pid": 76337, "tid": -914061504, "ts": 1716454216884407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216884420, "dur": 0,
    "args": {
      "External id": 17374,
      "cbid": 51, "correlation": 17374
    }
  },
  {
    "ph": "s", "id": 17374, "pid": 76337, "tid": -914061504, "ts": 1716454216884420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216884519, "dur": 2220,
    "args": {
      "External id": 17375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17375,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17375, "pid": 5, "tid": 7, "ts": 1716454216884519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216884421, "dur": 7,
    "args": {
      "External id": 17375,
      "cbid": 211, "correlation": 17375
    }
  },
  {
    "ph": "s", "id": 17375, "pid": 76337, "tid": -914061504, "ts": 1716454216884421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216886740, "dur": 114,
    "args": {
      "External id": 17380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17380, "pid": 5, "tid": 7, "ts": 1716454216886740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885099, "dur": 12,
    "args": {
      "External id": 17380,
      "cbid": 211, "correlation": 17380
    }
  },
  {
    "ph": "s", "id": 17380, "pid": 76337, "tid": -914061504, "ts": 1716454216885099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216886856, "dur": 163,
    "args": {
      "External id": 17389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17389,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17389, "pid": 5, "tid": 7, "ts": 1716454216886856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885256, "dur": 17,
    "args": {
      "External id": 17389,
      "cbid": 211, "correlation": 17389
    }
  },
  {
    "ph": "s", "id": 17389, "pid": 76337, "tid": -914061504, "ts": 1716454216885256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216887020, "dur": 128,
    "args": {
      "External id": 17409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17409,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 17409, "pid": 5, "tid": 7, "ts": 1716454216887020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885362, "dur": 14,
    "args": {
      "External id": 17409,
      "cbid": 211, "correlation": 17409
    }
  },
  {
    "ph": "s", "id": 17409, "pid": 76337, "tid": -914061504, "ts": 1716454216885362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216887150, "dur": 5,
    "args": {
      "External id": 17421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17421,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 17421, "pid": 5, "tid": 7, "ts": 1716454216887150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885387, "dur": 8,
    "args": {
      "External id": 17421,
      "cbid": 211, "correlation": 17421
    }
  },
  {
    "ph": "s", "id": 17421, "pid": 76337, "tid": -914061504, "ts": 1716454216885387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216887156, "dur": 160,
    "args": {
      "External id": 17424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17424,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17424, "pid": 5, "tid": 7, "ts": 1716454216887156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885418, "dur": 8,
    "args": {
      "External id": 17424,
      "cbid": 211, "correlation": 17424
    }
  },
  {
    "ph": "s", "id": 17424, "pid": 76337, "tid": -914061504, "ts": 1716454216885418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216887317, "dur": 102,
    "args": {
      "External id": 17433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17433,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17433, "pid": 5, "tid": 7, "ts": 1716454216887317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885475, "dur": 11,
    "args": {
      "External id": 17433,
      "cbid": 211, "correlation": 17433
    }
  },
  {
    "ph": "s", "id": 17433, "pid": 76337, "tid": -914061504, "ts": 1716454216885475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885546, "dur": 0,
    "args": {
      "External id": 17443,
      "cbid": 317, "correlation": 17443
    }
  },
  {
    "ph": "f", "id": 17443, "pid": 76337, "tid": -914061504, "ts": 1716454216885546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885547, "dur": 1,
    "args": {
      "External id": 17444,
      "cbid": 203, "correlation": 17444
    }
  },
  {
    "ph": "f", "id": 17444, "pid": 76337, "tid": -914061504, "ts": 1716454216885547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216885549, "dur": 0,
    "args": {
      "External id": 17445,
      "cbid": 205, "correlation": 17445
    }
  },
  {
    "ph": "f", "id": 17445, "pid": 76337, "tid": -914061504, "ts": 1716454216885549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216888696, "dur": 111,
    "args": {
      "External id": 17449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17449, "pid": 5, "tid": 7, "ts": 1716454216888696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216888672, "dur": 24,
    "args": {
      "External id": 17449,
      "cbid": 211, "correlation": 17449
    }
  },
  {
    "ph": "s", "id": 17449, "pid": 76337, "tid": -914061504, "ts": 1716454216888672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216888808, "dur": 34,
    "args": {
      "External id": 17451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17451, "pid": 5, "tid": 7, "ts": 1716454216888808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216888699, "dur": 6,
    "args": {
      "External id": 17451,
      "cbid": 211, "correlation": 17451
    }
  },
  {
    "ph": "s", "id": 17451, "pid": 76337, "tid": -914061504, "ts": 1716454216888699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216888844, "dur": 3,
    "args": {
      "External id": 17453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17453, "pid": 5, "tid": 7, "ts": 1716454216888844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216888710, "dur": 8,
    "args": {
      "External id": 17453,
      "cbid": 211, "correlation": 17453
    }
  },
  {
    "ph": "s", "id": 17453, "pid": 76337, "tid": -914061504, "ts": 1716454216888710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216888722, "dur": 0,
    "args": {
      "External id": 17454,
      "cbid": 51, "correlation": 17454
    }
  },
  {
    "ph": "s", "id": 17454, "pid": 76337, "tid": -914061504, "ts": 1716454216888722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216888848, "dur": 2025,
    "args": {
      "External id": 17455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17455,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17455, "pid": 5, "tid": 7, "ts": 1716454216888848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216888723, "dur": 7,
    "args": {
      "External id": 17455,
      "cbid": 211, "correlation": 17455
    }
  },
  {
    "ph": "s", "id": 17455, "pid": 76337, "tid": -914061504, "ts": 1716454216888723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216890875, "dur": 59,
    "args": {
      "External id": 17460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17460, "pid": 5, "tid": 7, "ts": 1716454216890875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889460, "dur": 12,
    "args": {
      "External id": 17460,
      "cbid": 211, "correlation": 17460
    }
  },
  {
    "ph": "s", "id": 17460, "pid": 76337, "tid": -914061504, "ts": 1716454216889460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216890936, "dur": 3,
    "args": {
      "External id": 17468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17468,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17468, "pid": 5, "tid": 7, "ts": 1716454216890936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889537, "dur": 11,
    "args": {
      "External id": 17468,
      "cbid": 211, "correlation": 17468
    }
  },
  {
    "ph": "s", "id": 17468, "pid": 76337, "tid": -914061504, "ts": 1716454216889537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889697, "dur": 4,
    "args": {
      "External id": 17484,
      "cbid": 251, "correlation": 17484
    }
  },
  {
    "ph": "f", "id": 17484, "pid": 76337, "tid": -914061504, "ts": 1716454216889697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889707, "dur": 0,
    "args": {
      "External id": 17486,
      "cbid": 251, "correlation": 17486
    }
  },
  {
    "ph": "f", "id": 17486, "pid": 76337, "tid": -914061504, "ts": 1716454216889707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216890940, "dur": 11,
    "args": {
      "External id": 17487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17487,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 17487, "pid": 5, "tid": 7, "ts": 1716454216890940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889712, "dur": 16,
    "args": {
      "External id": 17487,
      "cbid": 211, "correlation": 17487
    }
  },
  {
    "ph": "s", "id": 17487, "pid": 76337, "tid": -914061504, "ts": 1716454216889712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216890953, "dur": 5,
    "args": {
      "External id": 17489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17489,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 17489, "pid": 5, "tid": 7, "ts": 1716454216890953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889733, "dur": 8,
    "args": {
      "External id": 17489,
      "cbid": 211, "correlation": 17489
    }
  },
  {
    "ph": "s", "id": 17489, "pid": 76337, "tid": -914061504, "ts": 1716454216889733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216890959, "dur": 54,
    "args": {
      "External id": 17499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17499, "pid": 5, "tid": 7, "ts": 1716454216890959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889827, "dur": 13,
    "args": {
      "External id": 17499,
      "cbid": 211, "correlation": 17499
    }
  },
  {
    "ph": "s", "id": 17499, "pid": 76337, "tid": -914061504, "ts": 1716454216889827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216891014, "dur": 52,
    "args": {
      "External id": 17519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17519,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 17519, "pid": 5, "tid": 7, "ts": 1716454216891014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889919, "dur": 13,
    "args": {
      "External id": 17519,
      "cbid": 211, "correlation": 17519
    }
  },
  {
    "ph": "s", "id": 17519, "pid": 76337, "tid": -914061504, "ts": 1716454216889919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216891068, "dur": 4,
    "args": {
      "External id": 17531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17531,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17531, "pid": 5, "tid": 7, "ts": 1716454216891068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889944, "dur": 8,
    "args": {
      "External id": 17531,
      "cbid": 211, "correlation": 17531
    }
  },
  {
    "ph": "s", "id": 17531, "pid": 76337, "tid": -914061504, "ts": 1716454216889944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216891073, "dur": 55,
    "args": {
      "External id": 17534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17534, "pid": 5, "tid": 7, "ts": 1716454216891073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216889971, "dur": 16,
    "args": {
      "External id": 17534,
      "cbid": 211, "correlation": 17534
    }
  },
  {
    "ph": "s", "id": 17534, "pid": 76337, "tid": -914061504, "ts": 1716454216889971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216891130, "dur": 37,
    "args": {
      "External id": 17543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17543,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17543, "pid": 5, "tid": 7, "ts": 1716454216891130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890025, "dur": 11,
    "args": {
      "External id": 17543,
      "cbid": 211, "correlation": 17543
    }
  },
  {
    "ph": "s", "id": 17543, "pid": 76337, "tid": -914061504, "ts": 1716454216890025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890132, "dur": 0,
    "args": {
      "External id": 17553,
      "cbid": 317, "correlation": 17553
    }
  },
  {
    "ph": "f", "id": 17553, "pid": 76337, "tid": -914061504, "ts": 1716454216890132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890133, "dur": 1,
    "args": {
      "External id": 17554,
      "cbid": 203, "correlation": 17554
    }
  },
  {
    "ph": "f", "id": 17554, "pid": 76337, "tid": -914061504, "ts": 1716454216890133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890134, "dur": 1,
    "args": {
      "External id": 17555,
      "cbid": 205, "correlation": 17555
    }
  },
  {
    "ph": "f", "id": 17555, "pid": 76337, "tid": -914061504, "ts": 1716454216890134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216891168, "dur": 39,
    "args": {
      "External id": 17559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17559, "pid": 5, "tid": 7, "ts": 1716454216891168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890154, "dur": 13,
    "args": {
      "External id": 17559,
      "cbid": 211, "correlation": 17559
    }
  },
  {
    "ph": "s", "id": 17559, "pid": 76337, "tid": -914061504, "ts": 1716454216890154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216891209, "dur": 14,
    "args": {
      "External id": 17561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17561,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17561, "pid": 5, "tid": 7, "ts": 1716454216891209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890170, "dur": 6,
    "args": {
      "External id": 17561,
      "cbid": 211, "correlation": 17561
    }
  },
  {
    "ph": "s", "id": 17561, "pid": 76337, "tid": -914061504, "ts": 1716454216890170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216891224, "dur": 4,
    "args": {
      "External id": 17563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17563, "pid": 5, "tid": 7, "ts": 1716454216891224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890181, "dur": 6,
    "args": {
      "External id": 17563,
      "cbid": 211, "correlation": 17563
    }
  },
  {
    "ph": "s", "id": 17563, "pid": 76337, "tid": -914061504, "ts": 1716454216890181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890190, "dur": 0,
    "args": {
      "External id": 17564,
      "cbid": 51, "correlation": 17564
    }
  },
  {
    "ph": "s", "id": 17564, "pid": 76337, "tid": -914061504, "ts": 1716454216890190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216891229, "dur": 702,
    "args": {
      "External id": 17565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17565,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17565, "pid": 5, "tid": 7, "ts": 1716454216891229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890191, "dur": 5,
    "args": {
      "External id": 17565,
      "cbid": 211, "correlation": 17565
    }
  },
  {
    "ph": "s", "id": 17565, "pid": 76337, "tid": -914061504, "ts": 1716454216890191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216891932, "dur": 60,
    "args": {
      "External id": 17570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17570, "pid": 5, "tid": 7, "ts": 1716454216891932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890220, "dur": 9,
    "args": {
      "External id": 17570,
      "cbid": 211, "correlation": 17570
    }
  },
  {
    "ph": "s", "id": 17570, "pid": 76337, "tid": -914061504, "ts": 1716454216890220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890280, "dur": 0,
    "args": {
      "External id": 17580,
      "cbid": 317, "correlation": 17580
    }
  },
  {
    "ph": "f", "id": 17580, "pid": 76337, "tid": -914061504, "ts": 1716454216890280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890281, "dur": 0,
    "args": {
      "External id": 17581,
      "cbid": 203, "correlation": 17581
    }
  },
  {
    "ph": "f", "id": 17581, "pid": 76337, "tid": -914061504, "ts": 1716454216890281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216890281, "dur": 0,
    "args": {
      "External id": 17582,
      "cbid": 205, "correlation": 17582
    }
  },
  {
    "ph": "f", "id": 17582, "pid": 76337, "tid": -914061504, "ts": 1716454216890281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216891993, "dur": 4,
    "args": {
      "External id": 17586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17586,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17586, "pid": 5, "tid": 7, "ts": 1716454216891993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216891836, "dur": 19,
    "args": {
      "External id": 17586,
      "cbid": 211, "correlation": 17586
    }
  },
  {
    "ph": "s", "id": 17586, "pid": 76337, "tid": -914061504, "ts": 1716454216891836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216891858, "dur": 0,
    "args": {
      "External id": 17587,
      "cbid": 51, "correlation": 17587
    }
  },
  {
    "ph": "s", "id": 17587, "pid": 76337, "tid": -914061504, "ts": 1716454216891858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216892027, "dur": 262,
    "args": {
      "External id": 17588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17588,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17588, "pid": 5, "tid": 7, "ts": 1716454216892027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216891873, "dur": 154,
    "args": {
      "External id": 17588,
      "cbid": 211, "correlation": 17588
    }
  },
  {
    "ph": "s", "id": 17588, "pid": 76337, "tid": -914061504, "ts": 1716454216891873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216892448, "dur": 59,
    "args": {
      "External id": 17593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17593,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17593, "pid": 5, "tid": 7, "ts": 1716454216892448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892437, "dur": 11,
    "args": {
      "External id": 17593,
      "cbid": 211, "correlation": 17593
    }
  },
  {
    "ph": "s", "id": 17593, "pid": 76337, "tid": -914061504, "ts": 1716454216892437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216892509, "dur": 50,
    "args": {
      "External id": 17601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17601,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17601, "pid": 5, "tid": 7, "ts": 1716454216892509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892477, "dur": 10,
    "args": {
      "External id": 17601,
      "cbid": 211, "correlation": 17601
    }
  },
  {
    "ph": "s", "id": 17601, "pid": 76337, "tid": -914061504, "ts": 1716454216892477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216892560, "dur": 35,
    "args": {
      "External id": 17609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17609,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17609, "pid": 5, "tid": 7, "ts": 1716454216892560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892529, "dur": 11,
    "args": {
      "External id": 17609,
      "cbid": 211, "correlation": 17609
    }
  },
  {
    "ph": "s", "id": 17609, "pid": 76337, "tid": -914061504, "ts": 1716454216892529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216892666, "dur": 52,
    "args": {
      "External id": 17629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17629,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 17629, "pid": 5, "tid": 7, "ts": 1716454216892666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892653, "dur": 13,
    "args": {
      "External id": 17629,
      "cbid": 211, "correlation": 17629
    }
  },
  {
    "ph": "s", "id": 17629, "pid": 76337, "tid": -914061504, "ts": 1716454216892653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216892720, "dur": 4,
    "args": {
      "External id": 17641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17641,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 17641, "pid": 5, "tid": 7, "ts": 1716454216892720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892689, "dur": 9,
    "args": {
      "External id": 17641,
      "cbid": 211, "correlation": 17641
    }
  },
  {
    "ph": "s", "id": 17641, "pid": 76337, "tid": -914061504, "ts": 1716454216892689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216892725, "dur": 58,
    "args": {
      "External id": 17644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17644,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17644, "pid": 5, "tid": 7, "ts": 1716454216892725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892713, "dur": 8,
    "args": {
      "External id": 17644,
      "cbid": 211, "correlation": 17644
    }
  },
  {
    "ph": "s", "id": 17644, "pid": 76337, "tid": -914061504, "ts": 1716454216892713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892796, "dur": 0,
    "args": {
      "External id": 17655,
      "cbid": 317, "correlation": 17655
    }
  },
  {
    "ph": "f", "id": 17655, "pid": 76337, "tid": -914061504, "ts": 1716454216892796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892797, "dur": 1,
    "args": {
      "External id": 17656,
      "cbid": 203, "correlation": 17656
    }
  },
  {
    "ph": "f", "id": 17656, "pid": 76337, "tid": -914061504, "ts": 1716454216892797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892798, "dur": 0,
    "args": {
      "External id": 17657,
      "cbid": 205, "correlation": 17657
    }
  },
  {
    "ph": "f", "id": 17657, "pid": 76337, "tid": -914061504, "ts": 1716454216892798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892849, "dur": 3,
    "args": {
      "External id": 17661,
      "cbid": 251, "correlation": 17661
    }
  },
  {
    "ph": "f", "id": 17661, "pid": 76337, "tid": -914061504, "ts": 1716454216892849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892854, "dur": 1,
    "args": {
      "External id": 17662,
      "cbid": 251, "correlation": 17662
    }
  },
  {
    "ph": "f", "id": 17662, "pid": 76337, "tid": -914061504, "ts": 1716454216892854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892856, "dur": 1,
    "args": {
      "External id": 17663,
      "cbid": 251, "correlation": 17663
    }
  },
  {
    "ph": "f", "id": 17663, "pid": 76337, "tid": -914061504, "ts": 1716454216892856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892858, "dur": 1,
    "args": {
      "External id": 17664,
      "cbid": 251, "correlation": 17664
    }
  },
  {
    "ph": "f", "id": 17664, "pid": 76337, "tid": -914061504, "ts": 1716454216892858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892860, "dur": 1,
    "args": {
      "External id": 17665,
      "cbid": 251, "correlation": 17665
    }
  },
  {
    "ph": "f", "id": 17665, "pid": 76337, "tid": -914061504, "ts": 1716454216892860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892862, "dur": 1,
    "args": {
      "External id": 17666,
      "cbid": 251, "correlation": 17666
    }
  },
  {
    "ph": "f", "id": 17666, "pid": 76337, "tid": -914061504, "ts": 1716454216892862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892864, "dur": 1,
    "args": {
      "External id": 17667,
      "cbid": 251, "correlation": 17667
    }
  },
  {
    "ph": "f", "id": 17667, "pid": 76337, "tid": -914061504, "ts": 1716454216892864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892866, "dur": 1,
    "args": {
      "External id": 17668,
      "cbid": 251, "correlation": 17668
    }
  },
  {
    "ph": "f", "id": 17668, "pid": 76337, "tid": -914061504, "ts": 1716454216892866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892868, "dur": 0,
    "args": {
      "External id": 17669,
      "cbid": 251, "correlation": 17669
    }
  },
  {
    "ph": "f", "id": 17669, "pid": 76337, "tid": -914061504, "ts": 1716454216892868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216892892, "dur": 116,
    "args": {
      "External id": 17670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17670,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17670, "pid": 5, "tid": 7, "ts": 1716454216892892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892874, "dur": 18,
    "args": {
      "External id": 17670,
      "cbid": 211, "correlation": 17670
    }
  },
  {
    "ph": "s", "id": 17670, "pid": 76337, "tid": -914061504, "ts": 1716454216892874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216893009, "dur": 60,
    "args": {
      "External id": 17676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17676, "pid": 5, "tid": 7, "ts": 1716454216893009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216892916, "dur": 9,
    "args": {
      "External id": 17676,
      "cbid": 211, "correlation": 17676
    }
  },
  {
    "ph": "s", "id": 17676, "pid": 76337, "tid": -914061504, "ts": 1716454216892916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216893091, "dur": 149,
    "args": {
      "External id": 17687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17687,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17687, "pid": 5, "tid": 7, "ts": 1716454216893091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893072, "dur": 18,
    "args": {
      "External id": 17687,
      "cbid": 211, "correlation": 17687
    }
  },
  {
    "ph": "s", "id": 17687, "pid": 76337, "tid": -914061504, "ts": 1716454216893072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216893242, "dur": 95,
    "args": {
      "External id": 17709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17709,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17709, "pid": 5, "tid": 7, "ts": 1716454216893242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893126, "dur": 12,
    "args": {
      "External id": 17709,
      "cbid": 211, "correlation": 17709
    }
  },
  {
    "ph": "s", "id": 17709, "pid": 76337, "tid": -914061504, "ts": 1716454216893126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893266, "dur": 2,
    "args": {
      "External id": 17720,
      "cbid": 251, "correlation": 17720
    }
  },
  {
    "ph": "f", "id": 17720, "pid": 76337, "tid": -914061504, "ts": 1716454216893266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216893338, "dur": 111,
    "args": {
      "External id": 17721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17721,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17721, "pid": 5, "tid": 7, "ts": 1716454216893338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893273, "dur": 16,
    "args": {
      "External id": 17721,
      "cbid": 211, "correlation": 17721
    }
  },
  {
    "ph": "s", "id": 17721, "pid": 76337, "tid": -914061504, "ts": 1716454216893273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893365, "dur": 1,
    "args": {
      "External id": 17732,
      "cbid": 251, "correlation": 17732
    }
  },
  {
    "ph": "f", "id": 17732, "pid": 76337, "tid": -914061504, "ts": 1716454216893365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216893450, "dur": 103,
    "args": {
      "External id": 17733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17733,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17733, "pid": 5, "tid": 7, "ts": 1716454216893450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893369, "dur": 12,
    "args": {
      "External id": 17733,
      "cbid": 211, "correlation": 17733
    }
  },
  {
    "ph": "s", "id": 17733, "pid": 76337, "tid": -914061504, "ts": 1716454216893369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893448, "dur": 1,
    "args": {
      "External id": 17744,
      "cbid": 251, "correlation": 17744
    }
  },
  {
    "ph": "f", "id": 17744, "pid": 76337, "tid": -914061504, "ts": 1716454216893448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216893554, "dur": 102,
    "args": {
      "External id": 17745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17745,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17745, "pid": 5, "tid": 7, "ts": 1716454216893554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893452, "dur": 13,
    "args": {
      "External id": 17745,
      "cbid": 211, "correlation": 17745
    }
  },
  {
    "ph": "s", "id": 17745, "pid": 76337, "tid": -914061504, "ts": 1716454216893452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216893658, "dur": 4776,
    "args": {
      "External id": 17766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17766,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17766, "pid": 5, "tid": 7, "ts": 1716454216893658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893566, "dur": 16,
    "args": {
      "External id": 17766,
      "cbid": 211, "correlation": 17766
    }
  },
  {
    "ph": "s", "id": 17766, "pid": 76337, "tid": -914061504, "ts": 1716454216893566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893711, "dur": 2,
    "args": {
      "External id": 17784,
      "cbid": 251, "correlation": 17784
    }
  },
  {
    "ph": "f", "id": 17784, "pid": 76337, "tid": -914061504, "ts": 1716454216893711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216898435, "dur": 106,
    "args": {
      "External id": 17786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17786,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17786, "pid": 5, "tid": 7, "ts": 1716454216898435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893719, "dur": 15,
    "args": {
      "External id": 17786,
      "cbid": 211, "correlation": 17786
    }
  },
  {
    "ph": "s", "id": 17786, "pid": 76337, "tid": -914061504, "ts": 1716454216893719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216898542, "dur": 35,
    "args": {
      "External id": 17794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17794,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17794, "pid": 5, "tid": 7, "ts": 1716454216898542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893803, "dur": 13,
    "args": {
      "External id": 17794,
      "cbid": 211, "correlation": 17794
    }
  },
  {
    "ph": "s", "id": 17794, "pid": 76337, "tid": -914061504, "ts": 1716454216893803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216898579, "dur": 160,
    "args": {
      "External id": 17802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17802,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17802, "pid": 5, "tid": 7, "ts": 1716454216898579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893847, "dur": 10,
    "args": {
      "External id": 17802,
      "cbid": 211, "correlation": 17802
    }
  },
  {
    "ph": "s", "id": 17802, "pid": 76337, "tid": -914061504, "ts": 1716454216893847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216898741, "dur": 93,
    "args": {
      "External id": 17824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17824,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17824, "pid": 5, "tid": 7, "ts": 1716454216898741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893903, "dur": 10,
    "args": {
      "External id": 17824,
      "cbid": 211, "correlation": 17824
    }
  },
  {
    "ph": "s", "id": 17824, "pid": 76337, "tid": -914061504, "ts": 1716454216893903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216893995, "dur": 1,
    "args": {
      "External id": 17835,
      "cbid": 251, "correlation": 17835
    }
  },
  {
    "ph": "f", "id": 17835, "pid": 76337, "tid": -914061504, "ts": 1716454216893995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216898835, "dur": 104,
    "args": {
      "External id": 17836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17836,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17836, "pid": 5, "tid": 7, "ts": 1716454216898835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894000, "dur": 13,
    "args": {
      "External id": 17836,
      "cbid": 211, "correlation": 17836
    }
  },
  {
    "ph": "s", "id": 17836, "pid": 76337, "tid": -914061504, "ts": 1716454216894000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894099, "dur": 2,
    "args": {
      "External id": 17847,
      "cbid": 251, "correlation": 17847
    }
  },
  {
    "ph": "f", "id": 17847, "pid": 76337, "tid": -914061504, "ts": 1716454216894099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894104, "dur": 0,
    "args": {
      "External id": 17848,
      "cbid": 251, "correlation": 17848
    }
  },
  {
    "ph": "f", "id": 17848, "pid": 76337, "tid": -914061504, "ts": 1716454216894104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216898941, "dur": 10,
    "args": {
      "External id": 17849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17849,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 17849, "pid": 5, "tid": 7, "ts": 1716454216898941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894107, "dur": 16,
    "args": {
      "External id": 17849,
      "cbid": 211, "correlation": 17849
    }
  },
  {
    "ph": "s", "id": 17849, "pid": 76337, "tid": -914061504, "ts": 1716454216894107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216898952, "dur": 5,
    "args": {
      "External id": 17851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17851,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 17851, "pid": 5, "tid": 7, "ts": 1716454216898952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894126, "dur": 9,
    "args": {
      "External id": 17851,
      "cbid": 211, "correlation": 17851
    }
  },
  {
    "ph": "s", "id": 17851, "pid": 76337, "tid": -914061504, "ts": 1716454216894126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894203, "dur": 1,
    "args": {
      "External id": 17862,
      "cbid": 251, "correlation": 17862
    }
  },
  {
    "ph": "f", "id": 17862, "pid": 76337, "tid": -914061504, "ts": 1716454216894203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894206, "dur": 0,
    "args": {
      "External id": 17863,
      "cbid": 251, "correlation": 17863
    }
  },
  {
    "ph": "f", "id": 17863, "pid": 76337, "tid": -914061504, "ts": 1716454216894206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216898959, "dur": 6,
    "args": {
      "External id": 17864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17864,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 17864, "pid": 5, "tid": 7, "ts": 1716454216898959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894208, "dur": 12,
    "args": {
      "External id": 17864,
      "cbid": 211, "correlation": 17864
    }
  },
  {
    "ph": "s", "id": 17864, "pid": 76337, "tid": -914061504, "ts": 1716454216894208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216898966, "dur": 3,
    "args": {
      "External id": 17866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17866,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 17866, "pid": 5, "tid": 7, "ts": 1716454216898966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894222, "dur": 6,
    "args": {
      "External id": 17866,
      "cbid": 211, "correlation": 17866
    }
  },
  {
    "ph": "s", "id": 17866, "pid": 76337, "tid": -914061504, "ts": 1716454216894222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216898971, "dur": 154,
    "args": {
      "External id": 17887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17887,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17887, "pid": 5, "tid": 7, "ts": 1716454216898971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894299, "dur": 12,
    "args": {
      "External id": 17887,
      "cbid": 211, "correlation": 17887
    }
  },
  {
    "ph": "s", "id": 17887, "pid": 76337, "tid": -914061504, "ts": 1716454216894299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894409, "dur": 1,
    "args": {
      "External id": 17905,
      "cbid": 251, "correlation": 17905
    }
  },
  {
    "ph": "f", "id": 17905, "pid": 76337, "tid": -914061504, "ts": 1716454216894409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216899126, "dur": 105,
    "args": {
      "External id": 17907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17907,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 17907, "pid": 5, "tid": 7, "ts": 1716454216899126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894415, "dur": 14,
    "args": {
      "External id": 17907,
      "cbid": 211, "correlation": 17907
    }
  },
  {
    "ph": "s", "id": 17907, "pid": 76337, "tid": -914061504, "ts": 1716454216894415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216899233, "dur": 34,
    "args": {
      "External id": 17915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17915,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17915, "pid": 5, "tid": 7, "ts": 1716454216899233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894487, "dur": 13,
    "args": {
      "External id": 17915,
      "cbid": 211, "correlation": 17915
    }
  },
  {
    "ph": "s", "id": 17915, "pid": 76337, "tid": -914061504, "ts": 1716454216894487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216899268, "dur": 51,
    "args": {
      "External id": 17923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17923,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17923, "pid": 5, "tid": 7, "ts": 1716454216899268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894527, "dur": 9,
    "args": {
      "External id": 17923,
      "cbid": 211, "correlation": 17923
    }
  },
  {
    "ph": "s", "id": 17923, "pid": 76337, "tid": -914061504, "ts": 1716454216894527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216899320, "dur": 94,
    "args": {
      "External id": 17945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17945,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17945, "pid": 5, "tid": 7, "ts": 1716454216899320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894591, "dur": 12,
    "args": {
      "External id": 17945,
      "cbid": 211, "correlation": 17945
    }
  },
  {
    "ph": "s", "id": 17945, "pid": 76337, "tid": -914061504, "ts": 1716454216894591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894698, "dur": 2,
    "args": {
      "External id": 17961,
      "cbid": 251, "correlation": 17961
    }
  },
  {
    "ph": "f", "id": 17961, "pid": 76337, "tid": -914061504, "ts": 1716454216894698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216899415, "dur": 580,
    "args": {
      "External id": 17963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17963,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 17963, "pid": 5, "tid": 7, "ts": 1716454216899415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894705, "dur": 14,
    "args": {
      "External id": 17963,
      "cbid": 211, "correlation": 17963
    }
  },
  {
    "ph": "s", "id": 17963, "pid": 76337, "tid": -914061504, "ts": 1716454216894705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216899996, "dur": 245,
    "args": {
      "External id": 17971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17971, "pid": 5, "tid": 7, "ts": 1716454216899996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894803, "dur": 15,
    "args": {
      "External id": 17971,
      "cbid": 211, "correlation": 17971
    }
  },
  {
    "ph": "s", "id": 17971, "pid": 76337, "tid": -914061504, "ts": 1716454216894803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216900243, "dur": 253,
    "args": {
      "External id": 17979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 17979, "pid": 5, "tid": 7, "ts": 1716454216900243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894845, "dur": 11,
    "args": {
      "External id": 17979,
      "cbid": 211, "correlation": 17979
    }
  },
  {
    "ph": "s", "id": 17979, "pid": 76337, "tid": -914061504, "ts": 1716454216894845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894945, "dur": 2,
    "args": {
      "External id": 17995,
      "cbid": 251, "correlation": 17995
    }
  },
  {
    "ph": "f", "id": 17995, "pid": 76337, "tid": -914061504, "ts": 1716454216894945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894951, "dur": 0,
    "args": {
      "External id": 17997,
      "cbid": 251, "correlation": 17997
    }
  },
  {
    "ph": "f", "id": 17997, "pid": 76337, "tid": -914061504, "ts": 1716454216894951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216900497, "dur": 357,
    "args": {
      "External id": 17998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 17998,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 17998, "pid": 5, "tid": 7, "ts": 1716454216900497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216894956, "dur": 14,
    "args": {
      "External id": 17998,
      "cbid": 211, "correlation": 17998
    }
  },
  {
    "ph": "s", "id": 17998, "pid": 76337, "tid": -914061504, "ts": 1716454216894956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216900856, "dur": 50,
    "args": {
      "External id": 18006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18006,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18006, "pid": 5, "tid": 7, "ts": 1716454216900856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895010, "dur": 11,
    "args": {
      "External id": 18006,
      "cbid": 211, "correlation": 18006
    }
  },
  {
    "ph": "s", "id": 18006, "pid": 76337, "tid": -914061504, "ts": 1716454216895010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216900907, "dur": 158,
    "args": {
      "External id": 18017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18017, "pid": 5, "tid": 7, "ts": 1716454216900907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895081, "dur": 13,
    "args": {
      "External id": 18017,
      "cbid": 211, "correlation": 18017
    }
  },
  {
    "ph": "s", "id": 18017, "pid": 76337, "tid": -914061504, "ts": 1716454216895081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895158, "dur": 0,
    "args": {
      "External id": 18029,
      "cbid": 317, "correlation": 18029
    }
  },
  {
    "ph": "f", "id": 18029, "pid": 76337, "tid": -914061504, "ts": 1716454216895158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895159, "dur": 0,
    "args": {
      "External id": 18030,
      "cbid": 203, "correlation": 18030
    }
  },
  {
    "ph": "f", "id": 18030, "pid": 76337, "tid": -914061504, "ts": 1716454216895159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895159, "dur": 0,
    "args": {
      "External id": 18031,
      "cbid": 205, "correlation": 18031
    }
  },
  {
    "ph": "f", "id": 18031, "pid": 76337, "tid": -914061504, "ts": 1716454216895159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895186, "dur": 1,
    "args": {
      "External id": 18035,
      "cbid": 251, "correlation": 18035
    }
  },
  {
    "ph": "f", "id": 18035, "pid": 76337, "tid": -914061504, "ts": 1716454216895186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895189, "dur": 0,
    "args": {
      "External id": 18036,
      "cbid": 251, "correlation": 18036
    }
  },
  {
    "ph": "f", "id": 18036, "pid": 76337, "tid": -914061504, "ts": 1716454216895189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895190, "dur": 0,
    "args": {
      "External id": 18037,
      "cbid": 251, "correlation": 18037
    }
  },
  {
    "ph": "f", "id": 18037, "pid": 76337, "tid": -914061504, "ts": 1716454216895190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895191, "dur": 0,
    "args": {
      "External id": 18038,
      "cbid": 251, "correlation": 18038
    }
  },
  {
    "ph": "f", "id": 18038, "pid": 76337, "tid": -914061504, "ts": 1716454216895191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895192, "dur": 1,
    "args": {
      "External id": 18039,
      "cbid": 251, "correlation": 18039
    }
  },
  {
    "ph": "f", "id": 18039, "pid": 76337, "tid": -914061504, "ts": 1716454216895192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895194, "dur": 0,
    "args": {
      "External id": 18040,
      "cbid": 251, "correlation": 18040
    }
  },
  {
    "ph": "f", "id": 18040, "pid": 76337, "tid": -914061504, "ts": 1716454216895194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895194, "dur": 1,
    "args": {
      "External id": 18041,
      "cbid": 251, "correlation": 18041
    }
  },
  {
    "ph": "f", "id": 18041, "pid": 76337, "tid": -914061504, "ts": 1716454216895194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895196, "dur": 0,
    "args": {
      "External id": 18042,
      "cbid": 251, "correlation": 18042
    }
  },
  {
    "ph": "f", "id": 18042, "pid": 76337, "tid": -914061504, "ts": 1716454216895196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895197, "dur": 0,
    "args": {
      "External id": 18043,
      "cbid": 251, "correlation": 18043
    }
  },
  {
    "ph": "f", "id": 18043, "pid": 76337, "tid": -914061504, "ts": 1716454216895197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216901066, "dur": 114,
    "args": {
      "External id": 18044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18044,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18044, "pid": 5, "tid": 7, "ts": 1716454216901066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895199, "dur": 13,
    "args": {
      "External id": 18044,
      "cbid": 211, "correlation": 18044
    }
  },
  {
    "ph": "s", "id": 18044, "pid": 76337, "tid": -914061504, "ts": 1716454216895199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216901181, "dur": 61,
    "args": {
      "External id": 18050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18050, "pid": 5, "tid": 7, "ts": 1716454216901181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895235, "dur": 10,
    "args": {
      "External id": 18050,
      "cbid": 211, "correlation": 18050
    }
  },
  {
    "ph": "s", "id": 18050, "pid": 76337, "tid": -914061504, "ts": 1716454216895235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216901244, "dur": 50,
    "args": {
      "External id": 18058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18058,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18058, "pid": 5, "tid": 7, "ts": 1716454216901244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895268, "dur": 8,
    "args": {
      "External id": 18058,
      "cbid": 211, "correlation": 18058
    }
  },
  {
    "ph": "s", "id": 18058, "pid": 76337, "tid": -914061504, "ts": 1716454216895268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216901294, "dur": 98,
    "args": {
      "External id": 18067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18067,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18067, "pid": 5, "tid": 7, "ts": 1716454216901294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895324, "dur": 13,
    "args": {
      "External id": 18067,
      "cbid": 211, "correlation": 18067
    }
  },
  {
    "ph": "s", "id": 18067, "pid": 76337, "tid": -914061504, "ts": 1716454216895324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216901394, "dur": 94,
    "args": {
      "External id": 18087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18087,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 18087, "pid": 5, "tid": 7, "ts": 1716454216901394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895404, "dur": 11,
    "args": {
      "External id": 18087,
      "cbid": 211, "correlation": 18087
    }
  },
  {
    "ph": "s", "id": 18087, "pid": 76337, "tid": -914061504, "ts": 1716454216895404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216901490, "dur": 5,
    "args": {
      "External id": 18099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18099,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 18099, "pid": 5, "tid": 7, "ts": 1716454216901490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895426, "dur": 6,
    "args": {
      "External id": 18099,
      "cbid": 211, "correlation": 18099
    }
  },
  {
    "ph": "s", "id": 18099, "pid": 76337, "tid": -914061504, "ts": 1716454216895426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216901496, "dur": 107,
    "args": {
      "External id": 18102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18102, "pid": 5, "tid": 7, "ts": 1716454216901496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895445, "dur": 7,
    "args": {
      "External id": 18102,
      "cbid": 211, "correlation": 18102
    }
  },
  {
    "ph": "s", "id": 18102, "pid": 76337, "tid": -914061504, "ts": 1716454216895445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216901605, "dur": 69,
    "args": {
      "External id": 18111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18111,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18111, "pid": 5, "tid": 7, "ts": 1716454216901605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895490, "dur": 11,
    "args": {
      "External id": 18111,
      "cbid": 211, "correlation": 18111
    }
  },
  {
    "ph": "s", "id": 18111, "pid": 76337, "tid": -914061504, "ts": 1716454216895490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895544, "dur": 0,
    "args": {
      "External id": 18121,
      "cbid": 317, "correlation": 18121
    }
  },
  {
    "ph": "f", "id": 18121, "pid": 76337, "tid": -914061504, "ts": 1716454216895544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895544, "dur": 0,
    "args": {
      "External id": 18122,
      "cbid": 203, "correlation": 18122
    }
  },
  {
    "ph": "f", "id": 18122, "pid": 76337, "tid": -914061504, "ts": 1716454216895544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216895545, "dur": 0,
    "args": {
      "External id": 18123,
      "cbid": 205, "correlation": 18123
    }
  },
  {
    "ph": "f", "id": 18123, "pid": 76337, "tid": -914061504, "ts": 1716454216895545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216901675, "dur": 76,
    "args": {
      "External id": 18127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18127,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18127, "pid": 5, "tid": 7, "ts": 1716454216901675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216898727, "dur": 37,
    "args": {
      "External id": 18127,
      "cbid": 211, "correlation": 18127
    }
  },
  {
    "ph": "s", "id": 18127, "pid": 76337, "tid": -914061504, "ts": 1716454216898727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216901753, "dur": 24,
    "args": {
      "External id": 18129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18129, "pid": 5, "tid": 7, "ts": 1716454216901753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216898767, "dur": 6,
    "args": {
      "External id": 18129,
      "cbid": 211, "correlation": 18129
    }
  },
  {
    "ph": "s", "id": 18129, "pid": 76337, "tid": -914061504, "ts": 1716454216898767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216901778, "dur": 4,
    "args": {
      "External id": 18131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18131, "pid": 5, "tid": 7, "ts": 1716454216901778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216898782, "dur": 9,
    "args": {
      "External id": 18131,
      "cbid": 211, "correlation": 18131
    }
  },
  {
    "ph": "s", "id": 18131, "pid": 76337, "tid": -914061504, "ts": 1716454216898782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216898794, "dur": 0,
    "args": {
      "External id": 18132,
      "cbid": 51, "correlation": 18132
    }
  },
  {
    "ph": "s", "id": 18132, "pid": 76337, "tid": -914061504, "ts": 1716454216898794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216901783, "dur": 1371,
    "args": {
      "External id": 18133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18133,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18133, "pid": 5, "tid": 7, "ts": 1716454216901783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216898796, "dur": 8,
    "args": {
      "External id": 18133,
      "cbid": 211, "correlation": 18133
    }
  },
  {
    "ph": "s", "id": 18133, "pid": 76337, "tid": -914061504, "ts": 1716454216898796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216903156, "dur": 59,
    "args": {
      "External id": 18138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18138, "pid": 5, "tid": 7, "ts": 1716454216903156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899566, "dur": 14,
    "args": {
      "External id": 18138,
      "cbid": 211, "correlation": 18138
    }
  },
  {
    "ph": "s", "id": 18138, "pid": 76337, "tid": -914061504, "ts": 1716454216899566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216903216, "dur": 3,
    "args": {
      "External id": 18146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18146,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18146, "pid": 5, "tid": 7, "ts": 1716454216903216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899663, "dur": 13,
    "args": {
      "External id": 18146,
      "cbid": 211, "correlation": 18146
    }
  },
  {
    "ph": "s", "id": 18146, "pid": 76337, "tid": -914061504, "ts": 1716454216899663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899808, "dur": 4,
    "args": {
      "External id": 18162,
      "cbid": 251, "correlation": 18162
    }
  },
  {
    "ph": "f", "id": 18162, "pid": 76337, "tid": -914061504, "ts": 1716454216899808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899819, "dur": 0,
    "args": {
      "External id": 18164,
      "cbid": 251, "correlation": 18164
    }
  },
  {
    "ph": "f", "id": 18164, "pid": 76337, "tid": -914061504, "ts": 1716454216899819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216903221, "dur": 11,
    "args": {
      "External id": 18165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18165,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 18165, "pid": 5, "tid": 7, "ts": 1716454216903221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899823, "dur": 16,
    "args": {
      "External id": 18165,
      "cbid": 211, "correlation": 18165
    }
  },
  {
    "ph": "s", "id": 18165, "pid": 76337, "tid": -914061504, "ts": 1716454216899823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216903233, "dur": 5,
    "args": {
      "External id": 18167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18167,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 18167, "pid": 5, "tid": 7, "ts": 1716454216903233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899844, "dur": 9,
    "args": {
      "External id": 18167,
      "cbid": 211, "correlation": 18167
    }
  },
  {
    "ph": "s", "id": 18167, "pid": 76337, "tid": -914061504, "ts": 1716454216899844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216903240, "dur": 53,
    "args": {
      "External id": 18177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18177, "pid": 5, "tid": 7, "ts": 1716454216903240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216899936, "dur": 12,
    "args": {
      "External id": 18177,
      "cbid": 211, "correlation": 18177
    }
  },
  {
    "ph": "s", "id": 18177, "pid": 76337, "tid": -914061504, "ts": 1716454216899936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216903294, "dur": 52,
    "args": {
      "External id": 18197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18197,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 18197, "pid": 5, "tid": 7, "ts": 1716454216903294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900043, "dur": 12,
    "args": {
      "External id": 18197,
      "cbid": 211, "correlation": 18197
    }
  },
  {
    "ph": "s", "id": 18197, "pid": 76337, "tid": -914061504, "ts": 1716454216900043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216903348, "dur": 4,
    "args": {
      "External id": 18209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18209,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18209, "pid": 5, "tid": 7, "ts": 1716454216903348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900066, "dur": 8,
    "args": {
      "External id": 18209,
      "cbid": 211, "correlation": 18209
    }
  },
  {
    "ph": "s", "id": 18209, "pid": 76337, "tid": -914061504, "ts": 1716454216900066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216903353, "dur": 54,
    "args": {
      "External id": 18212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18212, "pid": 5, "tid": 7, "ts": 1716454216903353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900097, "dur": 8,
    "args": {
      "External id": 18212,
      "cbid": 211, "correlation": 18212
    }
  },
  {
    "ph": "s", "id": 18212, "pid": 76337, "tid": -914061504, "ts": 1716454216900097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216903409, "dur": 36,
    "args": {
      "External id": 18221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18221,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18221, "pid": 5, "tid": 7, "ts": 1716454216903409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900143, "dur": 10,
    "args": {
      "External id": 18221,
      "cbid": 211, "correlation": 18221
    }
  },
  {
    "ph": "s", "id": 18221, "pid": 76337, "tid": -914061504, "ts": 1716454216900143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900253, "dur": 0,
    "args": {
      "External id": 18231,
      "cbid": 317, "correlation": 18231
    }
  },
  {
    "ph": "f", "id": 18231, "pid": 76337, "tid": -914061504, "ts": 1716454216900253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900255, "dur": 1,
    "args": {
      "External id": 18232,
      "cbid": 203, "correlation": 18232
    }
  },
  {
    "ph": "f", "id": 18232, "pid": 76337, "tid": -914061504, "ts": 1716454216900255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900256, "dur": 1,
    "args": {
      "External id": 18233,
      "cbid": 205, "correlation": 18233
    }
  },
  {
    "ph": "f", "id": 18233, "pid": 76337, "tid": -914061504, "ts": 1716454216900256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216903447, "dur": 40,
    "args": {
      "External id": 18237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18237, "pid": 5, "tid": 7, "ts": 1716454216903447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900279, "dur": 13,
    "args": {
      "External id": 18237,
      "cbid": 211, "correlation": 18237
    }
  },
  {
    "ph": "s", "id": 18237, "pid": 76337, "tid": -914061504, "ts": 1716454216900279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216903488, "dur": 15,
    "args": {
      "External id": 18239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18239,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18239, "pid": 5, "tid": 7, "ts": 1716454216903488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900295, "dur": 5,
    "args": {
      "External id": 18239,
      "cbid": 211, "correlation": 18239
    }
  },
  {
    "ph": "s", "id": 18239, "pid": 76337, "tid": -914061504, "ts": 1716454216900295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216903504, "dur": 3,
    "args": {
      "External id": 18241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18241, "pid": 5, "tid": 7, "ts": 1716454216903504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900305, "dur": 6,
    "args": {
      "External id": 18241,
      "cbid": 211, "correlation": 18241
    }
  },
  {
    "ph": "s", "id": 18241, "pid": 76337, "tid": -914061504, "ts": 1716454216900305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900314, "dur": 0,
    "args": {
      "External id": 18242,
      "cbid": 51, "correlation": 18242
    }
  },
  {
    "ph": "s", "id": 18242, "pid": 76337, "tid": -914061504, "ts": 1716454216900314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216903508, "dur": 702,
    "args": {
      "External id": 18243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18243,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18243, "pid": 5, "tid": 7, "ts": 1716454216903508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900315, "dur": 6,
    "args": {
      "External id": 18243,
      "cbid": 211, "correlation": 18243
    }
  },
  {
    "ph": "s", "id": 18243, "pid": 76337, "tid": -914061504, "ts": 1716454216900315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216904212, "dur": 60,
    "args": {
      "External id": 18248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18248, "pid": 5, "tid": 7, "ts": 1716454216904212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900345, "dur": 8,
    "args": {
      "External id": 18248,
      "cbid": 211, "correlation": 18248
    }
  },
  {
    "ph": "s", "id": 18248, "pid": 76337, "tid": -914061504, "ts": 1716454216900345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900403, "dur": 0,
    "args": {
      "External id": 18258,
      "cbid": 317, "correlation": 18258
    }
  },
  {
    "ph": "f", "id": 18258, "pid": 76337, "tid": -914061504, "ts": 1716454216900403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900404, "dur": 0,
    "args": {
      "External id": 18259,
      "cbid": 203, "correlation": 18259
    }
  },
  {
    "ph": "f", "id": 18259, "pid": 76337, "tid": -914061504, "ts": 1716454216900404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216900405, "dur": 0,
    "args": {
      "External id": 18260,
      "cbid": 205, "correlation": 18260
    }
  },
  {
    "ph": "f", "id": 18260, "pid": 76337, "tid": -914061504, "ts": 1716454216900405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216904273, "dur": 76,
    "args": {
      "External id": 18264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18264,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18264, "pid": 5, "tid": 7, "ts": 1716454216904273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902049, "dur": 20,
    "args": {
      "External id": 18264,
      "cbid": 211, "correlation": 18264
    }
  },
  {
    "ph": "s", "id": 18264, "pid": 76337, "tid": -914061504, "ts": 1716454216902049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216904373, "dur": 201,
    "args": {
      "External id": 18266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18266,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18266, "pid": 5, "tid": 7, "ts": 1716454216904373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902094, "dur": 207,
    "args": {
      "External id": 18266,
      "cbid": 211, "correlation": 18266
    }
  },
  {
    "ph": "s", "id": 18266, "pid": 76337, "tid": -914061504, "ts": 1716454216902094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216904576, "dur": 39,
    "args": {
      "External id": 18268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18268, "pid": 5, "tid": 7, "ts": 1716454216904576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902309, "dur": 8,
    "args": {
      "External id": 18268,
      "cbid": 211, "correlation": 18268
    }
  },
  {
    "ph": "s", "id": 18268, "pid": 76337, "tid": -914061504, "ts": 1716454216902309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216904616, "dur": 61,
    "args": {
      "External id": 18274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18274,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18274, "pid": 5, "tid": 7, "ts": 1716454216904616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902740, "dur": 11,
    "args": {
      "External id": 18274,
      "cbid": 211, "correlation": 18274
    }
  },
  {
    "ph": "s", "id": 18274, "pid": 76337, "tid": -914061504, "ts": 1716454216902740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216904679, "dur": 50,
    "args": {
      "External id": 18282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18282,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18282, "pid": 5, "tid": 7, "ts": 1716454216904679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902783, "dur": 8,
    "args": {
      "External id": 18282,
      "cbid": 211, "correlation": 18282
    }
  },
  {
    "ph": "s", "id": 18282, "pid": 76337, "tid": -914061504, "ts": 1716454216902783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216904730, "dur": 35,
    "args": {
      "External id": 18290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18290,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18290, "pid": 5, "tid": 7, "ts": 1716454216904730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902835, "dur": 10,
    "args": {
      "External id": 18290,
      "cbid": 211, "correlation": 18290
    }
  },
  {
    "ph": "s", "id": 18290, "pid": 76337, "tid": -914061504, "ts": 1716454216902835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216904767, "dur": 53,
    "args": {
      "External id": 18310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18310,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 18310, "pid": 5, "tid": 7, "ts": 1716454216904767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216902986, "dur": 16,
    "args": {
      "External id": 18310,
      "cbid": 211, "correlation": 18310
    }
  },
  {
    "ph": "s", "id": 18310, "pid": 76337, "tid": -914061504, "ts": 1716454216902986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216904821, "dur": 4,
    "args": {
      "External id": 18322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18322,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18322, "pid": 5, "tid": 7, "ts": 1716454216904821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903012, "dur": 8,
    "args": {
      "External id": 18322,
      "cbid": 211, "correlation": 18322
    }
  },
  {
    "ph": "s", "id": 18322, "pid": 76337, "tid": -914061504, "ts": 1716454216903012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216904826, "dur": 57,
    "args": {
      "External id": 18325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18325, "pid": 5, "tid": 7, "ts": 1716454216904826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903036, "dur": 7,
    "args": {
      "External id": 18325,
      "cbid": 211, "correlation": 18325
    }
  },
  {
    "ph": "s", "id": 18325, "pid": 76337, "tid": -914061504, "ts": 1716454216903036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903110, "dur": 0,
    "args": {
      "External id": 18336,
      "cbid": 317, "correlation": 18336
    }
  },
  {
    "ph": "f", "id": 18336, "pid": 76337, "tid": -914061504, "ts": 1716454216903110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903110, "dur": 0,
    "args": {
      "External id": 18337,
      "cbid": 203, "correlation": 18337
    }
  },
  {
    "ph": "f", "id": 18337, "pid": 76337, "tid": -914061504, "ts": 1716454216903110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903111, "dur": 0,
    "args": {
      "External id": 18338,
      "cbid": 205, "correlation": 18338
    }
  },
  {
    "ph": "f", "id": 18338, "pid": 76337, "tid": -914061504, "ts": 1716454216903111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903149, "dur": 2,
    "args": {
      "External id": 18342,
      "cbid": 251, "correlation": 18342
    }
  },
  {
    "ph": "f", "id": 18342, "pid": 76337, "tid": -914061504, "ts": 1716454216903149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903153, "dur": 1,
    "args": {
      "External id": 18343,
      "cbid": 251, "correlation": 18343
    }
  },
  {
    "ph": "f", "id": 18343, "pid": 76337, "tid": -914061504, "ts": 1716454216903153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903155, "dur": 1,
    "args": {
      "External id": 18344,
      "cbid": 251, "correlation": 18344
    }
  },
  {
    "ph": "f", "id": 18344, "pid": 76337, "tid": -914061504, "ts": 1716454216903155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903157, "dur": 1,
    "args": {
      "External id": 18345,
      "cbid": 251, "correlation": 18345
    }
  },
  {
    "ph": "f", "id": 18345, "pid": 76337, "tid": -914061504, "ts": 1716454216903157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903158, "dur": 1,
    "args": {
      "External id": 18346,
      "cbid": 251, "correlation": 18346
    }
  },
  {
    "ph": "f", "id": 18346, "pid": 76337, "tid": -914061504, "ts": 1716454216903158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903160, "dur": 1,
    "args": {
      "External id": 18347,
      "cbid": 251, "correlation": 18347
    }
  },
  {
    "ph": "f", "id": 18347, "pid": 76337, "tid": -914061504, "ts": 1716454216903160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903162, "dur": 1,
    "args": {
      "External id": 18348,
      "cbid": 251, "correlation": 18348
    }
  },
  {
    "ph": "f", "id": 18348, "pid": 76337, "tid": -914061504, "ts": 1716454216903162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903164, "dur": 1,
    "args": {
      "External id": 18349,
      "cbid": 251, "correlation": 18349
    }
  },
  {
    "ph": "f", "id": 18349, "pid": 76337, "tid": -914061504, "ts": 1716454216903164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903167, "dur": 0,
    "args": {
      "External id": 18350,
      "cbid": 251, "correlation": 18350
    }
  },
  {
    "ph": "f", "id": 18350, "pid": 76337, "tid": -914061504, "ts": 1716454216903167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216904885, "dur": 118,
    "args": {
      "External id": 18351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18351,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18351, "pid": 5, "tid": 7, "ts": 1716454216904885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903171, "dur": 16,
    "args": {
      "External id": 18351,
      "cbid": 211, "correlation": 18351
    }
  },
  {
    "ph": "s", "id": 18351, "pid": 76337, "tid": -914061504, "ts": 1716454216903171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216905004, "dur": 59,
    "args": {
      "External id": 18357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18357, "pid": 5, "tid": 7, "ts": 1716454216905004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903222, "dur": 11,
    "args": {
      "External id": 18357,
      "cbid": 211, "correlation": 18357
    }
  },
  {
    "ph": "s", "id": 18357, "pid": 76337, "tid": -914061504, "ts": 1716454216903222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216905064, "dur": 149,
    "args": {
      "External id": 18368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18368,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18368, "pid": 5, "tid": 7, "ts": 1716454216905064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903386, "dur": 18,
    "args": {
      "External id": 18368,
      "cbid": 211, "correlation": 18368
    }
  },
  {
    "ph": "s", "id": 18368, "pid": 76337, "tid": -914061504, "ts": 1716454216903386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216905215, "dur": 95,
    "args": {
      "External id": 18390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18390,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18390, "pid": 5, "tid": 7, "ts": 1716454216905215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903430, "dur": 10,
    "args": {
      "External id": 18390,
      "cbid": 211, "correlation": 18390
    }
  },
  {
    "ph": "s", "id": 18390, "pid": 76337, "tid": -914061504, "ts": 1716454216903430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903581, "dur": 2,
    "args": {
      "External id": 18401,
      "cbid": 251, "correlation": 18401
    }
  },
  {
    "ph": "f", "id": 18401, "pid": 76337, "tid": -914061504, "ts": 1716454216903581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216905312, "dur": 110,
    "args": {
      "External id": 18402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18402,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18402, "pid": 5, "tid": 7, "ts": 1716454216905312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903588, "dur": 15,
    "args": {
      "External id": 18402,
      "cbid": 211, "correlation": 18402
    }
  },
  {
    "ph": "s", "id": 18402, "pid": 76337, "tid": -914061504, "ts": 1716454216903588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903673, "dur": 1,
    "args": {
      "External id": 18413,
      "cbid": 251, "correlation": 18413
    }
  },
  {
    "ph": "f", "id": 18413, "pid": 76337, "tid": -914061504, "ts": 1716454216903673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216905423, "dur": 104,
    "args": {
      "External id": 18414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18414,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18414, "pid": 5, "tid": 7, "ts": 1716454216905423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903678, "dur": 13,
    "args": {
      "External id": 18414,
      "cbid": 211, "correlation": 18414
    }
  },
  {
    "ph": "s", "id": 18414, "pid": 76337, "tid": -914061504, "ts": 1716454216903678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903764, "dur": 1,
    "args": {
      "External id": 18425,
      "cbid": 251, "correlation": 18425
    }
  },
  {
    "ph": "f", "id": 18425, "pid": 76337, "tid": -914061504, "ts": 1716454216903764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216905528, "dur": 105,
    "args": {
      "External id": 18426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18426,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18426, "pid": 5, "tid": 7, "ts": 1716454216905528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903769, "dur": 12,
    "args": {
      "External id": 18426,
      "cbid": 211, "correlation": 18426
    }
  },
  {
    "ph": "s", "id": 18426, "pid": 76337, "tid": -914061504, "ts": 1716454216903769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216905635, "dur": 4778,
    "args": {
      "External id": 18447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18447,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18447, "pid": 5, "tid": 7, "ts": 1716454216905635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216903883, "dur": 16,
    "args": {
      "External id": 18447,
      "cbid": 211, "correlation": 18447
    }
  },
  {
    "ph": "s", "id": 18447, "pid": 76337, "tid": -914061504, "ts": 1716454216903883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904031, "dur": 2,
    "args": {
      "External id": 18465,
      "cbid": 251, "correlation": 18465
    }
  },
  {
    "ph": "f", "id": 18465, "pid": 76337, "tid": -914061504, "ts": 1716454216904031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216910414, "dur": 108,
    "args": {
      "External id": 18467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18467,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18467, "pid": 5, "tid": 7, "ts": 1716454216910414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904038, "dur": 15,
    "args": {
      "External id": 18467,
      "cbid": 211, "correlation": 18467
    }
  },
  {
    "ph": "s", "id": 18467, "pid": 76337, "tid": -914061504, "ts": 1716454216904038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216910523, "dur": 35,
    "args": {
      "External id": 18475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18475,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18475, "pid": 5, "tid": 7, "ts": 1716454216910523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904125, "dur": 13,
    "args": {
      "External id": 18475,
      "cbid": 211, "correlation": 18475
    }
  },
  {
    "ph": "s", "id": 18475, "pid": 76337, "tid": -914061504, "ts": 1716454216904125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216910560, "dur": 158,
    "args": {
      "External id": 18483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18483, "pid": 5, "tid": 7, "ts": 1716454216910560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904168, "dur": 9,
    "args": {
      "External id": 18483,
      "cbid": 211, "correlation": 18483
    }
  },
  {
    "ph": "s", "id": 18483, "pid": 76337, "tid": -914061504, "ts": 1716454216904168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216910719, "dur": 93,
    "args": {
      "External id": 18505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18505,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18505, "pid": 5, "tid": 7, "ts": 1716454216910719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904237, "dur": 11,
    "args": {
      "External id": 18505,
      "cbid": 211, "correlation": 18505
    }
  },
  {
    "ph": "s", "id": 18505, "pid": 76337, "tid": -914061504, "ts": 1716454216904237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904343, "dur": 1,
    "args": {
      "External id": 18516,
      "cbid": 251, "correlation": 18516
    }
  },
  {
    "ph": "f", "id": 18516, "pid": 76337, "tid": -914061504, "ts": 1716454216904343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216910814, "dur": 105,
    "args": {
      "External id": 18517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18517,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18517, "pid": 5, "tid": 7, "ts": 1716454216910814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904348, "dur": 14,
    "args": {
      "External id": 18517,
      "cbid": 211, "correlation": 18517
    }
  },
  {
    "ph": "s", "id": 18517, "pid": 76337, "tid": -914061504, "ts": 1716454216904348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904434, "dur": 2,
    "args": {
      "External id": 18528,
      "cbid": 251, "correlation": 18528
    }
  },
  {
    "ph": "f", "id": 18528, "pid": 76337, "tid": -914061504, "ts": 1716454216904434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904439, "dur": 0,
    "args": {
      "External id": 18529,
      "cbid": 251, "correlation": 18529
    }
  },
  {
    "ph": "f", "id": 18529, "pid": 76337, "tid": -914061504, "ts": 1716454216904439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216910921, "dur": 10,
    "args": {
      "External id": 18530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18530,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 18530, "pid": 5, "tid": 7, "ts": 1716454216910921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904441, "dur": 14,
    "args": {
      "External id": 18530,
      "cbid": 211, "correlation": 18530
    }
  },
  {
    "ph": "s", "id": 18530, "pid": 76337, "tid": -914061504, "ts": 1716454216904441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216910932, "dur": 5,
    "args": {
      "External id": 18532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18532,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 18532, "pid": 5, "tid": 7, "ts": 1716454216910932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904459, "dur": 8,
    "args": {
      "External id": 18532,
      "cbid": 211, "correlation": 18532
    }
  },
  {
    "ph": "s", "id": 18532, "pid": 76337, "tid": -914061504, "ts": 1716454216904459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904536, "dur": 1,
    "args": {
      "External id": 18543,
      "cbid": 251, "correlation": 18543
    }
  },
  {
    "ph": "f", "id": 18543, "pid": 76337, "tid": -914061504, "ts": 1716454216904536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904539, "dur": 0,
    "args": {
      "External id": 18544,
      "cbid": 251, "correlation": 18544
    }
  },
  {
    "ph": "f", "id": 18544, "pid": 76337, "tid": -914061504, "ts": 1716454216904539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216910939, "dur": 6,
    "args": {
      "External id": 18545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18545,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 18545, "pid": 5, "tid": 7, "ts": 1716454216910939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904541, "dur": 13,
    "args": {
      "External id": 18545,
      "cbid": 211, "correlation": 18545
    }
  },
  {
    "ph": "s", "id": 18545, "pid": 76337, "tid": -914061504, "ts": 1716454216904541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216910947, "dur": 3,
    "args": {
      "External id": 18547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18547,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 18547, "pid": 5, "tid": 7, "ts": 1716454216910947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904555, "dur": 6,
    "args": {
      "External id": 18547,
      "cbid": 211, "correlation": 18547
    }
  },
  {
    "ph": "s", "id": 18547, "pid": 76337, "tid": -914061504, "ts": 1716454216904555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216910952, "dur": 156,
    "args": {
      "External id": 18568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18568,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18568, "pid": 5, "tid": 7, "ts": 1716454216910952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904632, "dur": 12,
    "args": {
      "External id": 18568,
      "cbid": 211, "correlation": 18568
    }
  },
  {
    "ph": "s", "id": 18568, "pid": 76337, "tid": -914061504, "ts": 1716454216904632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904743, "dur": 1,
    "args": {
      "External id": 18586,
      "cbid": 251, "correlation": 18586
    }
  },
  {
    "ph": "f", "id": 18586, "pid": 76337, "tid": -914061504, "ts": 1716454216904743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216911109, "dur": 105,
    "args": {
      "External id": 18588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18588,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18588, "pid": 5, "tid": 7, "ts": 1716454216911109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904749, "dur": 14,
    "args": {
      "External id": 18588,
      "cbid": 211, "correlation": 18588
    }
  },
  {
    "ph": "s", "id": 18588, "pid": 76337, "tid": -914061504, "ts": 1716454216904749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216911216, "dur": 35,
    "args": {
      "External id": 18596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18596,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18596, "pid": 5, "tid": 7, "ts": 1716454216911216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904822, "dur": 12,
    "args": {
      "External id": 18596,
      "cbid": 211, "correlation": 18596
    }
  },
  {
    "ph": "s", "id": 18596, "pid": 76337, "tid": -914061504, "ts": 1716454216904822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216911252, "dur": 51,
    "args": {
      "External id": 18604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18604,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18604, "pid": 5, "tid": 7, "ts": 1716454216911252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904863, "dur": 9,
    "args": {
      "External id": 18604,
      "cbid": 211, "correlation": 18604
    }
  },
  {
    "ph": "s", "id": 18604, "pid": 76337, "tid": -914061504, "ts": 1716454216904863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216911304, "dur": 93,
    "args": {
      "External id": 18626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18626,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18626, "pid": 5, "tid": 7, "ts": 1716454216911304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216904916, "dur": 10,
    "args": {
      "External id": 18626,
      "cbid": 211, "correlation": 18626
    }
  },
  {
    "ph": "s", "id": 18626, "pid": 76337, "tid": -914061504, "ts": 1716454216904916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905032, "dur": 3,
    "args": {
      "External id": 18642,
      "cbid": 251, "correlation": 18642
    }
  },
  {
    "ph": "f", "id": 18642, "pid": 76337, "tid": -914061504, "ts": 1716454216905032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216911399, "dur": 579,
    "args": {
      "External id": 18644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18644,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18644, "pid": 5, "tid": 7, "ts": 1716454216911399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905040, "dur": 15,
    "args": {
      "External id": 18644,
      "cbid": 211, "correlation": 18644
    }
  },
  {
    "ph": "s", "id": 18644, "pid": 76337, "tid": -914061504, "ts": 1716454216905040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216911979, "dur": 248,
    "args": {
      "External id": 18652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18652,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18652, "pid": 5, "tid": 7, "ts": 1716454216911979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905136, "dur": 15,
    "args": {
      "External id": 18652,
      "cbid": 211, "correlation": 18652
    }
  },
  {
    "ph": "s", "id": 18652, "pid": 76337, "tid": -914061504, "ts": 1716454216905136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216912228, "dur": 254,
    "args": {
      "External id": 18660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18660, "pid": 5, "tid": 7, "ts": 1716454216912228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905179, "dur": 12,
    "args": {
      "External id": 18660,
      "cbid": 211, "correlation": 18660
    }
  },
  {
    "ph": "s", "id": 18660, "pid": 76337, "tid": -914061504, "ts": 1716454216905179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905277, "dur": 2,
    "args": {
      "External id": 18676,
      "cbid": 251, "correlation": 18676
    }
  },
  {
    "ph": "f", "id": 18676, "pid": 76337, "tid": -914061504, "ts": 1716454216905277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905282, "dur": 0,
    "args": {
      "External id": 18678,
      "cbid": 251, "correlation": 18678
    }
  },
  {
    "ph": "f", "id": 18678, "pid": 76337, "tid": -914061504, "ts": 1716454216905282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216912483, "dur": 358,
    "args": {
      "External id": 18679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18679,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 18679, "pid": 5, "tid": 7, "ts": 1716454216912483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905288, "dur": 14,
    "args": {
      "External id": 18679,
      "cbid": 211, "correlation": 18679
    }
  },
  {
    "ph": "s", "id": 18679, "pid": 76337, "tid": -914061504, "ts": 1716454216905288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216912843, "dur": 49,
    "args": {
      "External id": 18687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18687,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18687, "pid": 5, "tid": 7, "ts": 1716454216912843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905332, "dur": 11,
    "args": {
      "External id": 18687,
      "cbid": 211, "correlation": 18687
    }
  },
  {
    "ph": "s", "id": 18687, "pid": 76337, "tid": -914061504, "ts": 1716454216905332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216912894, "dur": 159,
    "args": {
      "External id": 18698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18698, "pid": 5, "tid": 7, "ts": 1716454216912894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905402, "dur": 12,
    "args": {
      "External id": 18698,
      "cbid": 211, "correlation": 18698
    }
  },
  {
    "ph": "s", "id": 18698, "pid": 76337, "tid": -914061504, "ts": 1716454216905402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905482, "dur": 0,
    "args": {
      "External id": 18710,
      "cbid": 317, "correlation": 18710
    }
  },
  {
    "ph": "f", "id": 18710, "pid": 76337, "tid": -914061504, "ts": 1716454216905482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905483, "dur": 0,
    "args": {
      "External id": 18711,
      "cbid": 203, "correlation": 18711
    }
  },
  {
    "ph": "f", "id": 18711, "pid": 76337, "tid": -914061504, "ts": 1716454216905483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905483, "dur": 0,
    "args": {
      "External id": 18712,
      "cbid": 205, "correlation": 18712
    }
  },
  {
    "ph": "f", "id": 18712, "pid": 76337, "tid": -914061504, "ts": 1716454216905483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905509, "dur": 1,
    "args": {
      "External id": 18716,
      "cbid": 251, "correlation": 18716
    }
  },
  {
    "ph": "f", "id": 18716, "pid": 76337, "tid": -914061504, "ts": 1716454216905509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905511, "dur": 0,
    "args": {
      "External id": 18717,
      "cbid": 251, "correlation": 18717
    }
  },
  {
    "ph": "f", "id": 18717, "pid": 76337, "tid": -914061504, "ts": 1716454216905511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905512, "dur": 0,
    "args": {
      "External id": 18718,
      "cbid": 251, "correlation": 18718
    }
  },
  {
    "ph": "f", "id": 18718, "pid": 76337, "tid": -914061504, "ts": 1716454216905512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905513, "dur": 0,
    "args": {
      "External id": 18719,
      "cbid": 251, "correlation": 18719
    }
  },
  {
    "ph": "f", "id": 18719, "pid": 76337, "tid": -914061504, "ts": 1716454216905513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905514, "dur": 1,
    "args": {
      "External id": 18720,
      "cbid": 251, "correlation": 18720
    }
  },
  {
    "ph": "f", "id": 18720, "pid": 76337, "tid": -914061504, "ts": 1716454216905514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905515, "dur": 0,
    "args": {
      "External id": 18721,
      "cbid": 251, "correlation": 18721
    }
  },
  {
    "ph": "f", "id": 18721, "pid": 76337, "tid": -914061504, "ts": 1716454216905515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905516, "dur": 1,
    "args": {
      "External id": 18722,
      "cbid": 251, "correlation": 18722
    }
  },
  {
    "ph": "f", "id": 18722, "pid": 76337, "tid": -914061504, "ts": 1716454216905516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905517, "dur": 0,
    "args": {
      "External id": 18723,
      "cbid": 251, "correlation": 18723
    }
  },
  {
    "ph": "f", "id": 18723, "pid": 76337, "tid": -914061504, "ts": 1716454216905517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905519, "dur": 0,
    "args": {
      "External id": 18724,
      "cbid": 251, "correlation": 18724
    }
  },
  {
    "ph": "f", "id": 18724, "pid": 76337, "tid": -914061504, "ts": 1716454216905519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216913053, "dur": 115,
    "args": {
      "External id": 18725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18725,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 18725, "pid": 5, "tid": 7, "ts": 1716454216913053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905521, "dur": 13,
    "args": {
      "External id": 18725,
      "cbid": 211, "correlation": 18725
    }
  },
  {
    "ph": "s", "id": 18725, "pid": 76337, "tid": -914061504, "ts": 1716454216905521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216913170, "dur": 59,
    "args": {
      "External id": 18731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18731, "pid": 5, "tid": 7, "ts": 1716454216913170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905557, "dur": 9,
    "args": {
      "External id": 18731,
      "cbid": 211, "correlation": 18731
    }
  },
  {
    "ph": "s", "id": 18731, "pid": 76337, "tid": -914061504, "ts": 1716454216905557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216913231, "dur": 50,
    "args": {
      "External id": 18739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18739,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18739, "pid": 5, "tid": 7, "ts": 1716454216913231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905590, "dur": 9,
    "args": {
      "External id": 18739,
      "cbid": 211, "correlation": 18739
    }
  },
  {
    "ph": "s", "id": 18739, "pid": 76337, "tid": -914061504, "ts": 1716454216905590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216913282, "dur": 98,
    "args": {
      "External id": 18748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18748,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18748, "pid": 5, "tid": 7, "ts": 1716454216913282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905648, "dur": 13,
    "args": {
      "External id": 18748,
      "cbid": 211, "correlation": 18748
    }
  },
  {
    "ph": "s", "id": 18748, "pid": 76337, "tid": -914061504, "ts": 1716454216905648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216913381, "dur": 94,
    "args": {
      "External id": 18768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18768,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 18768, "pid": 5, "tid": 7, "ts": 1716454216913381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905729, "dur": 11,
    "args": {
      "External id": 18768,
      "cbid": 211, "correlation": 18768
    }
  },
  {
    "ph": "s", "id": 18768, "pid": 76337, "tid": -914061504, "ts": 1716454216905729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216913476, "dur": 4,
    "args": {
      "External id": 18780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18780,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 18780, "pid": 5, "tid": 7, "ts": 1716454216913476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905751, "dur": 7,
    "args": {
      "External id": 18780,
      "cbid": 211, "correlation": 18780
    }
  },
  {
    "ph": "s", "id": 18780, "pid": 76337, "tid": -914061504, "ts": 1716454216905751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216913482, "dur": 109,
    "args": {
      "External id": 18783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18783, "pid": 5, "tid": 7, "ts": 1716454216913482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905770, "dur": 7,
    "args": {
      "External id": 18783,
      "cbid": 211, "correlation": 18783
    }
  },
  {
    "ph": "s", "id": 18783, "pid": 76337, "tid": -914061504, "ts": 1716454216905770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216913593, "dur": 69,
    "args": {
      "External id": 18792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18792,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18792, "pid": 5, "tid": 7, "ts": 1716454216913593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905815, "dur": 11,
    "args": {
      "External id": 18792,
      "cbid": 211, "correlation": 18792
    }
  },
  {
    "ph": "s", "id": 18792, "pid": 76337, "tid": -914061504, "ts": 1716454216905815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905869, "dur": 0,
    "args": {
      "External id": 18802,
      "cbid": 317, "correlation": 18802
    }
  },
  {
    "ph": "f", "id": 18802, "pid": 76337, "tid": -914061504, "ts": 1716454216905869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905870, "dur": 0,
    "args": {
      "External id": 18803,
      "cbid": 203, "correlation": 18803
    }
  },
  {
    "ph": "f", "id": 18803, "pid": 76337, "tid": -914061504, "ts": 1716454216905870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905871, "dur": 0,
    "args": {
      "External id": 18804,
      "cbid": 205, "correlation": 18804
    }
  },
  {
    "ph": "f", "id": 18804, "pid": 76337, "tid": -914061504, "ts": 1716454216905871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216913663, "dur": 76,
    "args": {
      "External id": 18808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18808,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18808, "pid": 5, "tid": 7, "ts": 1716454216913663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905887, "dur": 12,
    "args": {
      "External id": 18808,
      "cbid": 211, "correlation": 18808
    }
  },
  {
    "ph": "s", "id": 18808, "pid": 76337, "tid": -914061504, "ts": 1716454216905887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216913741, "dur": 24,
    "args": {
      "External id": 18810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18810, "pid": 5, "tid": 7, "ts": 1716454216913741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905902, "dur": 6,
    "args": {
      "External id": 18810,
      "cbid": 211, "correlation": 18810
    }
  },
  {
    "ph": "s", "id": 18810, "pid": 76337, "tid": -914061504, "ts": 1716454216905902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216913766, "dur": 4,
    "args": {
      "External id": 18812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18812, "pid": 5, "tid": 7, "ts": 1716454216913766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905915, "dur": 6,
    "args": {
      "External id": 18812,
      "cbid": 211, "correlation": 18812
    }
  },
  {
    "ph": "s", "id": 18812, "pid": 76337, "tid": -914061504, "ts": 1716454216905915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905925, "dur": 0,
    "args": {
      "External id": 18813,
      "cbid": 51, "correlation": 18813
    }
  },
  {
    "ph": "s", "id": 18813, "pid": 76337, "tid": -914061504, "ts": 1716454216905925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216913771, "dur": 1371,
    "args": {
      "External id": 18814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18814,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18814, "pid": 5, "tid": 7, "ts": 1716454216913771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905926, "dur": 6,
    "args": {
      "External id": 18814,
      "cbid": 211, "correlation": 18814
    }
  },
  {
    "ph": "s", "id": 18814, "pid": 76337, "tid": -914061504, "ts": 1716454216905926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216915144, "dur": 60,
    "args": {
      "External id": 18819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18819, "pid": 5, "tid": 7, "ts": 1716454216915144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216905955, "dur": 9,
    "args": {
      "External id": 18819,
      "cbid": 211, "correlation": 18819
    }
  },
  {
    "ph": "s", "id": 18819, "pid": 76337, "tid": -914061504, "ts": 1716454216905955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216915205, "dur": 4,
    "args": {
      "External id": 18827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18827,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18827, "pid": 5, "tid": 7, "ts": 1716454216915205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906009, "dur": 10,
    "args": {
      "External id": 18827,
      "cbid": 211, "correlation": 18827
    }
  },
  {
    "ph": "s", "id": 18827, "pid": 76337, "tid": -914061504, "ts": 1716454216906009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906093, "dur": 2,
    "args": {
      "External id": 18843,
      "cbid": 251, "correlation": 18843
    }
  },
  {
    "ph": "f", "id": 18843, "pid": 76337, "tid": -914061504, "ts": 1716454216906093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906099, "dur": 0,
    "args": {
      "External id": 18845,
      "cbid": 251, "correlation": 18845
    }
  },
  {
    "ph": "f", "id": 18845, "pid": 76337, "tid": -914061504, "ts": 1716454216906099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216915210, "dur": 11,
    "args": {
      "External id": 18846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18846,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 18846, "pid": 5, "tid": 7, "ts": 1716454216915210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906101, "dur": 15,
    "args": {
      "External id": 18846,
      "cbid": 211, "correlation": 18846
    }
  },
  {
    "ph": "s", "id": 18846, "pid": 76337, "tid": -914061504, "ts": 1716454216906101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216915222, "dur": 5,
    "args": {
      "External id": 18848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18848,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 18848, "pid": 5, "tid": 7, "ts": 1716454216915222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906120, "dur": 7,
    "args": {
      "External id": 18848,
      "cbid": 211, "correlation": 18848
    }
  },
  {
    "ph": "s", "id": 18848, "pid": 76337, "tid": -914061504, "ts": 1716454216906120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216915229, "dur": 54,
    "args": {
      "External id": 18858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18858, "pid": 5, "tid": 7, "ts": 1716454216915229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906188, "dur": 12,
    "args": {
      "External id": 18858,
      "cbid": 211, "correlation": 18858
    }
  },
  {
    "ph": "s", "id": 18858, "pid": 76337, "tid": -914061504, "ts": 1716454216906188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216915284, "dur": 52,
    "args": {
      "External id": 18878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18878,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 18878, "pid": 5, "tid": 7, "ts": 1716454216915284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906254, "dur": 11,
    "args": {
      "External id": 18878,
      "cbid": 211, "correlation": 18878
    }
  },
  {
    "ph": "s", "id": 18878, "pid": 76337, "tid": -914061504, "ts": 1716454216906254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216915337, "dur": 4,
    "args": {
      "External id": 18890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18890,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18890, "pid": 5, "tid": 7, "ts": 1716454216915337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906274, "dur": 6,
    "args": {
      "External id": 18890,
      "cbid": 211, "correlation": 18890
    }
  },
  {
    "ph": "s", "id": 18890, "pid": 76337, "tid": -914061504, "ts": 1716454216906274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216915342, "dur": 57,
    "args": {
      "External id": 18893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18893,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18893, "pid": 5, "tid": 7, "ts": 1716454216915342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906292, "dur": 6,
    "args": {
      "External id": 18893,
      "cbid": 211, "correlation": 18893
    }
  },
  {
    "ph": "s", "id": 18893, "pid": 76337, "tid": -914061504, "ts": 1716454216906292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216915400, "dur": 36,
    "args": {
      "External id": 18902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18902,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18902, "pid": 5, "tid": 7, "ts": 1716454216915400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906333, "dur": 10,
    "args": {
      "External id": 18902,
      "cbid": 211, "correlation": 18902
    }
  },
  {
    "ph": "s", "id": 18902, "pid": 76337, "tid": -914061504, "ts": 1716454216906333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906410, "dur": 0,
    "args": {
      "External id": 18912,
      "cbid": 317, "correlation": 18912
    }
  },
  {
    "ph": "f", "id": 18912, "pid": 76337, "tid": -914061504, "ts": 1716454216906410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906411, "dur": 0,
    "args": {
      "External id": 18913,
      "cbid": 203, "correlation": 18913
    }
  },
  {
    "ph": "f", "id": 18913, "pid": 76337, "tid": -914061504, "ts": 1716454216906411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906412, "dur": 0,
    "args": {
      "External id": 18914,
      "cbid": 205, "correlation": 18914
    }
  },
  {
    "ph": "f", "id": 18914, "pid": 76337, "tid": -914061504, "ts": 1716454216906412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216915438, "dur": 40,
    "args": {
      "External id": 18918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18918,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18918, "pid": 5, "tid": 7, "ts": 1716454216915438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906429, "dur": 13,
    "args": {
      "External id": 18918,
      "cbid": 211, "correlation": 18918
    }
  },
  {
    "ph": "s", "id": 18918, "pid": 76337, "tid": -914061504, "ts": 1716454216906429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216915479, "dur": 14,
    "args": {
      "External id": 18920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18920,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18920, "pid": 5, "tid": 7, "ts": 1716454216915479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906445, "dur": 6,
    "args": {
      "External id": 18920,
      "cbid": 211, "correlation": 18920
    }
  },
  {
    "ph": "s", "id": 18920, "pid": 76337, "tid": -914061504, "ts": 1716454216906445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216915494, "dur": 3,
    "args": {
      "External id": 18922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 18922, "pid": 5, "tid": 7, "ts": 1716454216915494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906455, "dur": 6,
    "args": {
      "External id": 18922,
      "cbid": 211, "correlation": 18922
    }
  },
  {
    "ph": "s", "id": 18922, "pid": 76337, "tid": -914061504, "ts": 1716454216906455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906464, "dur": 0,
    "args": {
      "External id": 18923,
      "cbid": 51, "correlation": 18923
    }
  },
  {
    "ph": "s", "id": 18923, "pid": 76337, "tid": -914061504, "ts": 1716454216906464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216915499, "dur": 703,
    "args": {
      "External id": 18924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18924,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18924, "pid": 5, "tid": 7, "ts": 1716454216915499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906465, "dur": 6,
    "args": {
      "External id": 18924,
      "cbid": 211, "correlation": 18924
    }
  },
  {
    "ph": "s", "id": 18924, "pid": 76337, "tid": -914061504, "ts": 1716454216906465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216916204, "dur": 59,
    "args": {
      "External id": 18929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18929,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18929, "pid": 5, "tid": 7, "ts": 1716454216916204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906494, "dur": 9,
    "args": {
      "External id": 18929,
      "cbid": 211, "correlation": 18929
    }
  },
  {
    "ph": "s", "id": 18929, "pid": 76337, "tid": -914061504, "ts": 1716454216906494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906553, "dur": 0,
    "args": {
      "External id": 18939,
      "cbid": 317, "correlation": 18939
    }
  },
  {
    "ph": "f", "id": 18939, "pid": 76337, "tid": -914061504, "ts": 1716454216906553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906554, "dur": 0,
    "args": {
      "External id": 18940,
      "cbid": 203, "correlation": 18940
    }
  },
  {
    "ph": "f", "id": 18940, "pid": 76337, "tid": -914061504, "ts": 1716454216906554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906555, "dur": 0,
    "args": {
      "External id": 18941,
      "cbid": 205, "correlation": 18941
    }
  },
  {
    "ph": "f", "id": 18941, "pid": 76337, "tid": -914061504, "ts": 1716454216906555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216916264, "dur": 75,
    "args": {
      "External id": 18945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18945,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18945, "pid": 5, "tid": 7, "ts": 1716454216916264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906566, "dur": 13,
    "args": {
      "External id": 18945,
      "cbid": 211, "correlation": 18945
    }
  },
  {
    "ph": "s", "id": 18945, "pid": 76337, "tid": -914061504, "ts": 1716454216906566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454216916341, "dur": 209,
    "args": {
      "External id": 18947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18947,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 18947, "pid": 5, "tid": 7, "ts": 1716454216916341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906588, "dur": 7,
    "args": {
      "External id": 18947,
      "cbid": 211, "correlation": 18947
    }
  },
  {
    "ph": "s", "id": 18947, "pid": 76337, "tid": -914061504, "ts": 1716454216906588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216916552, "dur": 38,
    "args": {
      "External id": 18949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18949, "pid": 5, "tid": 7, "ts": 1716454216916552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906600, "dur": 6,
    "args": {
      "External id": 18949,
      "cbid": 211, "correlation": 18949
    }
  },
  {
    "ph": "s", "id": 18949, "pid": 76337, "tid": -914061504, "ts": 1716454216906600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216916591, "dur": 60,
    "args": {
      "External id": 18955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18955, "pid": 5, "tid": 7, "ts": 1716454216916591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906628, "dur": 9,
    "args": {
      "External id": 18955,
      "cbid": 211, "correlation": 18955
    }
  },
  {
    "ph": "s", "id": 18955, "pid": 76337, "tid": -914061504, "ts": 1716454216906628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216916652, "dur": 50,
    "args": {
      "External id": 18963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18963,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18963, "pid": 5, "tid": 7, "ts": 1716454216916652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906656, "dur": 8,
    "args": {
      "External id": 18963,
      "cbid": 211, "correlation": 18963
    }
  },
  {
    "ph": "s", "id": 18963, "pid": 76337, "tid": -914061504, "ts": 1716454216906656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216916703, "dur": 35,
    "args": {
      "External id": 18971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18971,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 18971, "pid": 5, "tid": 7, "ts": 1716454216916703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906686, "dur": 8,
    "args": {
      "External id": 18971,
      "cbid": 211, "correlation": 18971
    }
  },
  {
    "ph": "s", "id": 18971, "pid": 76337, "tid": -914061504, "ts": 1716454216906686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216916739, "dur": 52,
    "args": {
      "External id": 18991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 18991,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 18991, "pid": 5, "tid": 7, "ts": 1716454216916739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906769, "dur": 14,
    "args": {
      "External id": 18991,
      "cbid": 211, "correlation": 18991
    }
  },
  {
    "ph": "s", "id": 18991, "pid": 76337, "tid": -914061504, "ts": 1716454216906769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216916793, "dur": 5,
    "args": {
      "External id": 19003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19003,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19003, "pid": 5, "tid": 7, "ts": 1716454216916793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906793, "dur": 6,
    "args": {
      "External id": 19003,
      "cbid": 211, "correlation": 19003
    }
  },
  {
    "ph": "s", "id": 19003, "pid": 76337, "tid": -914061504, "ts": 1716454216906793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216916799, "dur": 57,
    "args": {
      "External id": 19006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19006, "pid": 5, "tid": 7, "ts": 1716454216916799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906810, "dur": 7,
    "args": {
      "External id": 19006,
      "cbid": 211, "correlation": 19006
    }
  },
  {
    "ph": "s", "id": 19006, "pid": 76337, "tid": -914061504, "ts": 1716454216906810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906882, "dur": 0,
    "args": {
      "External id": 19017,
      "cbid": 317, "correlation": 19017
    }
  },
  {
    "ph": "f", "id": 19017, "pid": 76337, "tid": -914061504, "ts": 1716454216906882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906882, "dur": 0,
    "args": {
      "External id": 19018,
      "cbid": 203, "correlation": 19018
    }
  },
  {
    "ph": "f", "id": 19018, "pid": 76337, "tid": -914061504, "ts": 1716454216906882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906883, "dur": 0,
    "args": {
      "External id": 19019,
      "cbid": 205, "correlation": 19019
    }
  },
  {
    "ph": "f", "id": 19019, "pid": 76337, "tid": -914061504, "ts": 1716454216906883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906906, "dur": 1,
    "args": {
      "External id": 19023,
      "cbid": 251, "correlation": 19023
    }
  },
  {
    "ph": "f", "id": 19023, "pid": 76337, "tid": -914061504, "ts": 1716454216906906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906908, "dur": 0,
    "args": {
      "External id": 19024,
      "cbid": 251, "correlation": 19024
    }
  },
  {
    "ph": "f", "id": 19024, "pid": 76337, "tid": -914061504, "ts": 1716454216906908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906909, "dur": 0,
    "args": {
      "External id": 19025,
      "cbid": 251, "correlation": 19025
    }
  },
  {
    "ph": "f", "id": 19025, "pid": 76337, "tid": -914061504, "ts": 1716454216906909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906910, "dur": 0,
    "args": {
      "External id": 19026,
      "cbid": 251, "correlation": 19026
    }
  },
  {
    "ph": "f", "id": 19026, "pid": 76337, "tid": -914061504, "ts": 1716454216906910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906910, "dur": 0,
    "args": {
      "External id": 19027,
      "cbid": 251, "correlation": 19027
    }
  },
  {
    "ph": "f", "id": 19027, "pid": 76337, "tid": -914061504, "ts": 1716454216906910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906911, "dur": 0,
    "args": {
      "External id": 19028,
      "cbid": 251, "correlation": 19028
    }
  },
  {
    "ph": "f", "id": 19028, "pid": 76337, "tid": -914061504, "ts": 1716454216906911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906912, "dur": 0,
    "args": {
      "External id": 19029,
      "cbid": 251, "correlation": 19029
    }
  },
  {
    "ph": "f", "id": 19029, "pid": 76337, "tid": -914061504, "ts": 1716454216906912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906913, "dur": 0,
    "args": {
      "External id": 19030,
      "cbid": 251, "correlation": 19030
    }
  },
  {
    "ph": "f", "id": 19030, "pid": 76337, "tid": -914061504, "ts": 1716454216906913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906914, "dur": 0,
    "args": {
      "External id": 19031,
      "cbid": 251, "correlation": 19031
    }
  },
  {
    "ph": "f", "id": 19031, "pid": 76337, "tid": -914061504, "ts": 1716454216906914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216916858, "dur": 114,
    "args": {
      "External id": 19032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19032,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19032, "pid": 5, "tid": 7, "ts": 1716454216916858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906917, "dur": 14,
    "args": {
      "External id": 19032,
      "cbid": 211, "correlation": 19032
    }
  },
  {
    "ph": "s", "id": 19032, "pid": 76337, "tid": -914061504, "ts": 1716454216906917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216916973, "dur": 60,
    "args": {
      "External id": 19038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19038, "pid": 5, "tid": 7, "ts": 1716454216916973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216906954, "dur": 9,
    "args": {
      "External id": 19038,
      "cbid": 211, "correlation": 19038
    }
  },
  {
    "ph": "s", "id": 19038, "pid": 76337, "tid": -914061504, "ts": 1716454216906954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216917034, "dur": 150,
    "args": {
      "External id": 19049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19049, "pid": 5, "tid": 7, "ts": 1716454216917034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907063, "dur": 14,
    "args": {
      "External id": 19049,
      "cbid": 211, "correlation": 19049
    }
  },
  {
    "ph": "s", "id": 19049, "pid": 76337, "tid": -914061504, "ts": 1716454216907063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216917186, "dur": 95,
    "args": {
      "External id": 19071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19071,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19071, "pid": 5, "tid": 7, "ts": 1716454216917186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907109, "dur": 10,
    "args": {
      "External id": 19071,
      "cbid": 211, "correlation": 19071
    }
  },
  {
    "ph": "s", "id": 19071, "pid": 76337, "tid": -914061504, "ts": 1716454216907109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907212, "dur": 1,
    "args": {
      "External id": 19082,
      "cbid": 251, "correlation": 19082
    }
  },
  {
    "ph": "f", "id": 19082, "pid": 76337, "tid": -914061504, "ts": 1716454216907212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216917282, "dur": 111,
    "args": {
      "External id": 19083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19083,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19083, "pid": 5, "tid": 7, "ts": 1716454216917282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907217, "dur": 14,
    "args": {
      "External id": 19083,
      "cbid": 211, "correlation": 19083
    }
  },
  {
    "ph": "s", "id": 19083, "pid": 76337, "tid": -914061504, "ts": 1716454216907217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907286, "dur": 1,
    "args": {
      "External id": 19094,
      "cbid": 251, "correlation": 19094
    }
  },
  {
    "ph": "f", "id": 19094, "pid": 76337, "tid": -914061504, "ts": 1716454216907286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216917395, "dur": 105,
    "args": {
      "External id": 19095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19095,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19095, "pid": 5, "tid": 7, "ts": 1716454216917395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907291, "dur": 11,
    "args": {
      "External id": 19095,
      "cbid": 211, "correlation": 19095
    }
  },
  {
    "ph": "s", "id": 19095, "pid": 76337, "tid": -914061504, "ts": 1716454216907291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907353, "dur": 1,
    "args": {
      "External id": 19106,
      "cbid": 251, "correlation": 19106
    }
  },
  {
    "ph": "f", "id": 19106, "pid": 76337, "tid": -914061504, "ts": 1716454216907353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216917501, "dur": 104,
    "args": {
      "External id": 19107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19107,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19107, "pid": 5, "tid": 7, "ts": 1716454216917501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907357, "dur": 11,
    "args": {
      "External id": 19107,
      "cbid": 211, "correlation": 19107
    }
  },
  {
    "ph": "s", "id": 19107, "pid": 76337, "tid": -914061504, "ts": 1716454216907357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216917606, "dur": 4781,
    "args": {
      "External id": 19128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19128,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19128, "pid": 5, "tid": 7, "ts": 1716454216917606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907438, "dur": 13,
    "args": {
      "External id": 19128,
      "cbid": 211, "correlation": 19128
    }
  },
  {
    "ph": "s", "id": 19128, "pid": 76337, "tid": -914061504, "ts": 1716454216907438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907551, "dur": 1,
    "args": {
      "External id": 19146,
      "cbid": 251, "correlation": 19146
    }
  },
  {
    "ph": "f", "id": 19146, "pid": 76337, "tid": -914061504, "ts": 1716454216907551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216922388, "dur": 107,
    "args": {
      "External id": 19148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19148,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19148, "pid": 5, "tid": 7, "ts": 1716454216922388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907556, "dur": 14,
    "args": {
      "External id": 19148,
      "cbid": 211, "correlation": 19148
    }
  },
  {
    "ph": "s", "id": 19148, "pid": 76337, "tid": -914061504, "ts": 1716454216907556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216922497, "dur": 34,
    "args": {
      "External id": 19156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19156,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19156, "pid": 5, "tid": 7, "ts": 1716454216922497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907629, "dur": 12,
    "args": {
      "External id": 19156,
      "cbid": 211, "correlation": 19156
    }
  },
  {
    "ph": "s", "id": 19156, "pid": 76337, "tid": -914061504, "ts": 1716454216907629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216922532, "dur": 159,
    "args": {
      "External id": 19164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19164, "pid": 5, "tid": 7, "ts": 1716454216922532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907672, "dur": 9,
    "args": {
      "External id": 19164,
      "cbid": 211, "correlation": 19164
    }
  },
  {
    "ph": "s", "id": 19164, "pid": 76337, "tid": -914061504, "ts": 1716454216907672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216922693, "dur": 93,
    "args": {
      "External id": 19186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19186,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19186, "pid": 5, "tid": 7, "ts": 1716454216922693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907725, "dur": 10,
    "args": {
      "External id": 19186,
      "cbid": 211, "correlation": 19186
    }
  },
  {
    "ph": "s", "id": 19186, "pid": 76337, "tid": -914061504, "ts": 1716454216907725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907808, "dur": 1,
    "args": {
      "External id": 19197,
      "cbid": 251, "correlation": 19197
    }
  },
  {
    "ph": "f", "id": 19197, "pid": 76337, "tid": -914061504, "ts": 1716454216907808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216922787, "dur": 103,
    "args": {
      "External id": 19198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19198,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19198, "pid": 5, "tid": 7, "ts": 1716454216922787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907813, "dur": 14,
    "args": {
      "External id": 19198,
      "cbid": 211, "correlation": 19198
    }
  },
  {
    "ph": "s", "id": 19198, "pid": 76337, "tid": -914061504, "ts": 1716454216907813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907900, "dur": 1,
    "args": {
      "External id": 19209,
      "cbid": 251, "correlation": 19209
    }
  },
  {
    "ph": "f", "id": 19209, "pid": 76337, "tid": -914061504, "ts": 1716454216907900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907904, "dur": 0,
    "args": {
      "External id": 19210,
      "cbid": 251, "correlation": 19210
    }
  },
  {
    "ph": "f", "id": 19210, "pid": 76337, "tid": -914061504, "ts": 1716454216907904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216922891, "dur": 10,
    "args": {
      "External id": 19211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19211,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 19211, "pid": 5, "tid": 7, "ts": 1716454216922891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907906, "dur": 13,
    "args": {
      "External id": 19211,
      "cbid": 211, "correlation": 19211
    }
  },
  {
    "ph": "s", "id": 19211, "pid": 76337, "tid": -914061504, "ts": 1716454216907906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216922903, "dur": 5,
    "args": {
      "External id": 19213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19213,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 19213, "pid": 5, "tid": 7, "ts": 1716454216922903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907921, "dur": 6,
    "args": {
      "External id": 19213,
      "cbid": 211, "correlation": 19213
    }
  },
  {
    "ph": "s", "id": 19213, "pid": 76337, "tid": -914061504, "ts": 1716454216907921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907991, "dur": 1,
    "args": {
      "External id": 19224,
      "cbid": 251, "correlation": 19224
    }
  },
  {
    "ph": "f", "id": 19224, "pid": 76337, "tid": -914061504, "ts": 1716454216907991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907994, "dur": 0,
    "args": {
      "External id": 19225,
      "cbid": 251, "correlation": 19225
    }
  },
  {
    "ph": "f", "id": 19225, "pid": 76337, "tid": -914061504, "ts": 1716454216907994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216922909, "dur": 6,
    "args": {
      "External id": 19226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19226,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 19226, "pid": 5, "tid": 7, "ts": 1716454216922909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216907996, "dur": 12,
    "args": {
      "External id": 19226,
      "cbid": 211, "correlation": 19226
    }
  },
  {
    "ph": "s", "id": 19226, "pid": 76337, "tid": -914061504, "ts": 1716454216907996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216922917, "dur": 4,
    "args": {
      "External id": 19228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19228,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 19228, "pid": 5, "tid": 7, "ts": 1716454216922917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908010, "dur": 5,
    "args": {
      "External id": 19228,
      "cbid": 211, "correlation": 19228
    }
  },
  {
    "ph": "s", "id": 19228, "pid": 76337, "tid": -914061504, "ts": 1716454216908010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216922922, "dur": 156,
    "args": {
      "External id": 19249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19249,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19249, "pid": 5, "tid": 7, "ts": 1716454216922922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908085, "dur": 12,
    "args": {
      "External id": 19249,
      "cbid": 211, "correlation": 19249
    }
  },
  {
    "ph": "s", "id": 19249, "pid": 76337, "tid": -914061504, "ts": 1716454216908085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908181, "dur": 1,
    "args": {
      "External id": 19267,
      "cbid": 251, "correlation": 19267
    }
  },
  {
    "ph": "f", "id": 19267, "pid": 76337, "tid": -914061504, "ts": 1716454216908181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216923079, "dur": 106,
    "args": {
      "External id": 19269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19269,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19269, "pid": 5, "tid": 7, "ts": 1716454216923079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908188, "dur": 13,
    "args": {
      "External id": 19269,
      "cbid": 211, "correlation": 19269
    }
  },
  {
    "ph": "s", "id": 19269, "pid": 76337, "tid": -914061504, "ts": 1716454216908188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216923187, "dur": 34,
    "args": {
      "External id": 19277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19277,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19277, "pid": 5, "tid": 7, "ts": 1716454216923187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908258, "dur": 12,
    "args": {
      "External id": 19277,
      "cbid": 211, "correlation": 19277
    }
  },
  {
    "ph": "s", "id": 19277, "pid": 76337, "tid": -914061504, "ts": 1716454216908258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216923222, "dur": 50,
    "args": {
      "External id": 19285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19285,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19285, "pid": 5, "tid": 7, "ts": 1716454216923222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908298, "dur": 9,
    "args": {
      "External id": 19285,
      "cbid": 211, "correlation": 19285
    }
  },
  {
    "ph": "s", "id": 19285, "pid": 76337, "tid": -914061504, "ts": 1716454216908298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216923274, "dur": 93,
    "args": {
      "External id": 19307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19307,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19307, "pid": 5, "tid": 7, "ts": 1716454216923274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908351, "dur": 10,
    "args": {
      "External id": 19307,
      "cbid": 211, "correlation": 19307
    }
  },
  {
    "ph": "s", "id": 19307, "pid": 76337, "tid": -914061504, "ts": 1716454216908351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908448, "dur": 2,
    "args": {
      "External id": 19323,
      "cbid": 251, "correlation": 19323
    }
  },
  {
    "ph": "f", "id": 19323, "pid": 76337, "tid": -914061504, "ts": 1716454216908448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454216923368, "dur": 580,
    "args": {
      "External id": 19325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19325,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19325, "pid": 5, "tid": 7, "ts": 1716454216923368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908454, "dur": 14,
    "args": {
      "External id": 19325,
      "cbid": 211, "correlation": 19325
    }
  },
  {
    "ph": "s", "id": 19325, "pid": 76337, "tid": -914061504, "ts": 1716454216908454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216923950, "dur": 245,
    "args": {
      "External id": 19333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19333, "pid": 5, "tid": 7, "ts": 1716454216923950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908523, "dur": 13,
    "args": {
      "External id": 19333,
      "cbid": 211, "correlation": 19333
    }
  },
  {
    "ph": "s", "id": 19333, "pid": 76337, "tid": -914061504, "ts": 1716454216908523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216924196, "dur": 254,
    "args": {
      "External id": 19341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19341,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19341, "pid": 5, "tid": 7, "ts": 1716454216924196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908555, "dur": 9,
    "args": {
      "External id": 19341,
      "cbid": 211, "correlation": 19341
    }
  },
  {
    "ph": "s", "id": 19341, "pid": 76337, "tid": -914061504, "ts": 1716454216908555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908648, "dur": 1,
    "args": {
      "External id": 19357,
      "cbid": 251, "correlation": 19357
    }
  },
  {
    "ph": "f", "id": 19357, "pid": 76337, "tid": -914061504, "ts": 1716454216908648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908654, "dur": 0,
    "args": {
      "External id": 19359,
      "cbid": 251, "correlation": 19359
    }
  },
  {
    "ph": "f", "id": 19359, "pid": 76337, "tid": -914061504, "ts": 1716454216908654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454216924451, "dur": 361,
    "args": {
      "External id": 19360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19360,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19360, "pid": 5, "tid": 7, "ts": 1716454216924451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908656, "dur": 15,
    "args": {
      "External id": 19360,
      "cbid": 211, "correlation": 19360
    }
  },
  {
    "ph": "s", "id": 19360, "pid": 76337, "tid": -914061504, "ts": 1716454216908656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216924813, "dur": 49,
    "args": {
      "External id": 19368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19368,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19368, "pid": 5, "tid": 7, "ts": 1716454216924813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908702, "dur": 11,
    "args": {
      "External id": 19368,
      "cbid": 211, "correlation": 19368
    }
  },
  {
    "ph": "s", "id": 19368, "pid": 76337, "tid": -914061504, "ts": 1716454216908702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216924864, "dur": 159,
    "args": {
      "External id": 19379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19379, "pid": 5, "tid": 7, "ts": 1716454216924864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908769, "dur": 12,
    "args": {
      "External id": 19379,
      "cbid": 211, "correlation": 19379
    }
  },
  {
    "ph": "s", "id": 19379, "pid": 76337, "tid": -914061504, "ts": 1716454216908769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908834, "dur": 0,
    "args": {
      "External id": 19391,
      "cbid": 317, "correlation": 19391
    }
  },
  {
    "ph": "f", "id": 19391, "pid": 76337, "tid": -914061504, "ts": 1716454216908834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908835, "dur": 0,
    "args": {
      "External id": 19392,
      "cbid": 203, "correlation": 19392
    }
  },
  {
    "ph": "f", "id": 19392, "pid": 76337, "tid": -914061504, "ts": 1716454216908835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908836, "dur": 0,
    "args": {
      "External id": 19393,
      "cbid": 205, "correlation": 19393
    }
  },
  {
    "ph": "f", "id": 19393, "pid": 76337, "tid": -914061504, "ts": 1716454216908836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908859, "dur": 1,
    "args": {
      "External id": 19397,
      "cbid": 251, "correlation": 19397
    }
  },
  {
    "ph": "f", "id": 19397, "pid": 76337, "tid": -914061504, "ts": 1716454216908859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908862, "dur": 0,
    "args": {
      "External id": 19398,
      "cbid": 251, "correlation": 19398
    }
  },
  {
    "ph": "f", "id": 19398, "pid": 76337, "tid": -914061504, "ts": 1716454216908862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908862, "dur": 0,
    "args": {
      "External id": 19399,
      "cbid": 251, "correlation": 19399
    }
  },
  {
    "ph": "f", "id": 19399, "pid": 76337, "tid": -914061504, "ts": 1716454216908862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908863, "dur": 0,
    "args": {
      "External id": 19400,
      "cbid": 251, "correlation": 19400
    }
  },
  {
    "ph": "f", "id": 19400, "pid": 76337, "tid": -914061504, "ts": 1716454216908863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908864, "dur": 0,
    "args": {
      "External id": 19401,
      "cbid": 251, "correlation": 19401
    }
  },
  {
    "ph": "f", "id": 19401, "pid": 76337, "tid": -914061504, "ts": 1716454216908864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908865, "dur": 0,
    "args": {
      "External id": 19402,
      "cbid": 251, "correlation": 19402
    }
  },
  {
    "ph": "f", "id": 19402, "pid": 76337, "tid": -914061504, "ts": 1716454216908865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908866, "dur": 0,
    "args": {
      "External id": 19403,
      "cbid": 251, "correlation": 19403
    }
  },
  {
    "ph": "f", "id": 19403, "pid": 76337, "tid": -914061504, "ts": 1716454216908866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908866, "dur": 0,
    "args": {
      "External id": 19404,
      "cbid": 251, "correlation": 19404
    }
  },
  {
    "ph": "f", "id": 19404, "pid": 76337, "tid": -914061504, "ts": 1716454216908866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908868, "dur": 0,
    "args": {
      "External id": 19405,
      "cbid": 251, "correlation": 19405
    }
  },
  {
    "ph": "f", "id": 19405, "pid": 76337, "tid": -914061504, "ts": 1716454216908868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216925024, "dur": 114,
    "args": {
      "External id": 19406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19406,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19406, "pid": 5, "tid": 7, "ts": 1716454216925024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908870, "dur": 12,
    "args": {
      "External id": 19406,
      "cbid": 211, "correlation": 19406
    }
  },
  {
    "ph": "s", "id": 19406, "pid": 76337, "tid": -914061504, "ts": 1716454216908870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216925140, "dur": 60,
    "args": {
      "External id": 19412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19412,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19412, "pid": 5, "tid": 7, "ts": 1716454216925140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908904, "dur": 9,
    "args": {
      "External id": 19412,
      "cbid": 211, "correlation": 19412
    }
  },
  {
    "ph": "s", "id": 19412, "pid": 76337, "tid": -914061504, "ts": 1716454216908904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216925201, "dur": 50,
    "args": {
      "External id": 19420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19420,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19420, "pid": 5, "tid": 7, "ts": 1716454216925201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216908937, "dur": 8,
    "args": {
      "External id": 19420,
      "cbid": 211, "correlation": 19420
    }
  },
  {
    "ph": "s", "id": 19420, "pid": 76337, "tid": -914061504, "ts": 1716454216908937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216925253, "dur": 52,
    "args": {
      "External id": 19440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19440,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 19440, "pid": 5, "tid": 7, "ts": 1716454216925253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909048, "dur": 14,
    "args": {
      "External id": 19440,
      "cbid": 211, "correlation": 19440
    }
  },
  {
    "ph": "s", "id": 19440, "pid": 76337, "tid": -914061504, "ts": 1716454216909048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216925306, "dur": 5,
    "args": {
      "External id": 19452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19452,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19452, "pid": 5, "tid": 7, "ts": 1716454216925306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909079, "dur": 8,
    "args": {
      "External id": 19452,
      "cbid": 211, "correlation": 19452
    }
  },
  {
    "ph": "s", "id": 19452, "pid": 76337, "tid": -914061504, "ts": 1716454216909079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216925312, "dur": 59,
    "args": {
      "External id": 19455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19455, "pid": 5, "tid": 7, "ts": 1716454216925312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909100, "dur": 7,
    "args": {
      "External id": 19455,
      "cbid": 211, "correlation": 19455
    }
  },
  {
    "ph": "s", "id": 19455, "pid": 76337, "tid": -914061504, "ts": 1716454216909100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925372, "dur": 37,
    "args": {
      "External id": 19464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19464,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19464, "pid": 5, "tid": 7, "ts": 1716454216925372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909147, "dur": 10,
    "args": {
      "External id": 19464,
      "cbid": 211, "correlation": 19464
    }
  },
  {
    "ph": "s", "id": 19464, "pid": 76337, "tid": -914061504, "ts": 1716454216909147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909215, "dur": 0,
    "args": {
      "External id": 19474,
      "cbid": 317, "correlation": 19474
    }
  },
  {
    "ph": "f", "id": 19474, "pid": 76337, "tid": -914061504, "ts": 1716454216909215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909216, "dur": 0,
    "args": {
      "External id": 19475,
      "cbid": 203, "correlation": 19475
    }
  },
  {
    "ph": "f", "id": 19475, "pid": 76337, "tid": -914061504, "ts": 1716454216909216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216909216, "dur": 0,
    "args": {
      "External id": 19476,
      "cbid": 205, "correlation": 19476
    }
  },
  {
    "ph": "f", "id": 19476, "pid": 76337, "tid": -914061504, "ts": 1716454216909216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216925411, "dur": 42,
    "args": {
      "External id": 19480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19480,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19480, "pid": 5, "tid": 7, "ts": 1716454216925411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912093, "dur": 31,
    "args": {
      "External id": 19480,
      "cbid": 211, "correlation": 19480
    }
  },
  {
    "ph": "s", "id": 19480, "pid": 76337, "tid": -914061504, "ts": 1716454216912093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454216925454, "dur": 3,
    "args": {
      "External id": 19482,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 19482,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.310980496927598
    }
  },
  {
    "ph": "f", "id": 19482, "pid": 5, "tid": 7, "ts": 1716454216925454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912128, "dur": 25,
    "args": {
      "External id": 19482,
      "cbid": 51, "correlation": 19482
    }
  },
  {
    "ph": "s", "id": 19482, "pid": 76337, "tid": -914061504, "ts": 1716454216912128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912171, "dur": 1,
    "args": {
      "External id": 19484,
      "cbid": 200, "correlation": 19484
    }
  },
  {
    "ph": "f", "id": 19484, "pid": 76337, "tid": -914061504, "ts": 1716454216912171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912173, "dur": 0,
    "args": {
      "External id": 19485,
      "cbid": 200, "correlation": 19485
    }
  },
  {
    "ph": "f", "id": 19485, "pid": 76337, "tid": -914061504, "ts": 1716454216912173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912174, "dur": 0,
    "args": {
      "External id": 19486,
      "cbid": 200, "correlation": 19486
    }
  },
  {
    "ph": "f", "id": 19486, "pid": 76337, "tid": -914061504, "ts": 1716454216912174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912174, "dur": 0,
    "args": {
      "External id": 19487,
      "cbid": 200, "correlation": 19487
    }
  },
  {
    "ph": "f", "id": 19487, "pid": 76337, "tid": -914061504, "ts": 1716454216912174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912175, "dur": 223,
    "args": {
      "External id": 19488,
      "cbid": 15, "correlation": 19488
    }
  },
  {
    "ph": "f", "id": 19488, "pid": 76337, "tid": -914061504, "ts": 1716454216912175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912400, "dur": 1,
    "args": {
      "External id": 19489,
      "cbid": 251, "correlation": 19489
    }
  },
  {
    "ph": "f", "id": 19489, "pid": 76337, "tid": -914061504, "ts": 1716454216912400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454216925473, "dur": 25,
    "args": {
      "External id": 19490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19490,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19490, "pid": 5, "tid": 7, "ts": 1716454216925473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912413, "dur": 109,
    "args": {
      "External id": 19490,
      "cbid": 211, "correlation": 19490
    }
  },
  {
    "ph": "s", "id": 19490, "pid": 76337, "tid": -914061504, "ts": 1716454216912413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216925500, "dur": 4,
    "args": {
      "External id": 19492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19492,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19492, "pid": 5, "tid": 7, "ts": 1716454216925500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912530, "dur": 9,
    "args": {
      "External id": 19492,
      "cbid": 211, "correlation": 19492
    }
  },
  {
    "ph": "s", "id": 19492, "pid": 76337, "tid": -914061504, "ts": 1716454216912530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912542, "dur": 0,
    "args": {
      "External id": 19493,
      "cbid": 51, "correlation": 19493
    }
  },
  {
    "ph": "s", "id": 19493, "pid": 76337, "tid": -914061504, "ts": 1716454216912542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216925505, "dur": 190,
    "args": {
      "External id": 19494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19494,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 19494, "pid": 5, "tid": 7, "ts": 1716454216925505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912544, "dur": 10,
    "args": {
      "External id": 19494,
      "cbid": 211, "correlation": 19494
    }
  },
  {
    "ph": "s", "id": 19494, "pid": 76337, "tid": -914061504, "ts": 1716454216912544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216925696, "dur": 6,
    "args": {
      "External id": 19495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19495,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19495, "pid": 5, "tid": 7, "ts": 1716454216925696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216912559, "dur": 6,
    "args": {
      "External id": 19495,
      "cbid": 211, "correlation": 19495
    }
  },
  {
    "ph": "s", "id": 19495, "pid": 76337, "tid": -914061504, "ts": 1716454216912559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216925704, "dur": 6,
    "args": {
      "External id": 19501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 19501, "pid": 5, "tid": 7, "ts": 1716454216925704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216913302, "dur": 13,
    "args": {
      "External id": 19501,
      "cbid": 211, "correlation": 19501
    }
  },
  {
    "ph": "s", "id": 19501, "pid": 76337, "tid": -914061504, "ts": 1716454216913302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216925712, "dur": 5,
    "args": {
      "External id": 19510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19510,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19510, "pid": 5, "tid": 7, "ts": 1716454216925712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216915904, "dur": 23,
    "args": {
      "External id": 19510,
      "cbid": 211, "correlation": 19510
    }
  },
  {
    "ph": "s", "id": 19510, "pid": 76337, "tid": -914061504, "ts": 1716454216915904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216925719, "dur": 5,
    "args": {
      "External id": 19519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19519,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19519, "pid": 5, "tid": 7, "ts": 1716454216925719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216915953, "dur": 9,
    "args": {
      "External id": 19519,
      "cbid": 211, "correlation": 19519
    }
  },
  {
    "ph": "s", "id": 19519, "pid": 76337, "tid": -914061504, "ts": 1716454216915953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454216925725, "dur": 3,
    "args": {
      "External id": 19535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19535, "pid": 5, "tid": 7, "ts": 1716454216925725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916201, "dur": 18,
    "args": {
      "External id": 19535,
      "cbid": 211, "correlation": 19535
    }
  },
  {
    "ph": "s", "id": 19535, "pid": 76337, "tid": -914061504, "ts": 1716454216916201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925729, "dur": 3,
    "args": {
      "External id": 19543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19543,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19543, "pid": 5, "tid": 7, "ts": 1716454216925729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916263, "dur": 12,
    "args": {
      "External id": 19543,
      "cbid": 211, "correlation": 19543
    }
  },
  {
    "ph": "s", "id": 19543, "pid": 76337, "tid": -914061504, "ts": 1716454216916263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925733, "dur": 3,
    "args": {
      "External id": 19551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19551,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19551, "pid": 5, "tid": 7, "ts": 1716454216925733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916305, "dur": 11,
    "args": {
      "External id": 19551,
      "cbid": 211, "correlation": 19551
    }
  },
  {
    "ph": "s", "id": 19551, "pid": 76337, "tid": -914061504, "ts": 1716454216916305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925737, "dur": 4,
    "args": {
      "External id": 19559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19559,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19559, "pid": 5, "tid": 7, "ts": 1716454216925737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916341, "dur": 11,
    "args": {
      "External id": 19559,
      "cbid": 211, "correlation": 19559
    }
  },
  {
    "ph": "s", "id": 19559, "pid": 76337, "tid": -914061504, "ts": 1716454216916341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454216925743, "dur": 4,
    "args": {
      "External id": 19571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19571,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19571, "pid": 5, "tid": 7, "ts": 1716454216925743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916425, "dur": 15,
    "args": {
      "External id": 19571,
      "cbid": 211, "correlation": 19571
    }
  },
  {
    "ph": "s", "id": 19571, "pid": 76337, "tid": -914061504, "ts": 1716454216916425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216925748, "dur": 4,
    "args": {
      "External id": 19582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19582, "pid": 5, "tid": 7, "ts": 1716454216925748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916475, "dur": 12,
    "args": {
      "External id": 19582,
      "cbid": 211, "correlation": 19582
    }
  },
  {
    "ph": "s", "id": 19582, "pid": 76337, "tid": -914061504, "ts": 1716454216916475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925753, "dur": 3,
    "args": {
      "External id": 19590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19590, "pid": 5, "tid": 7, "ts": 1716454216925753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916511, "dur": 8,
    "args": {
      "External id": 19590,
      "cbid": 211, "correlation": 19590
    }
  },
  {
    "ph": "s", "id": 19590, "pid": 76337, "tid": -914061504, "ts": 1716454216916511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925757, "dur": 5,
    "args": {
      "External id": 19598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19598,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19598, "pid": 5, "tid": 7, "ts": 1716454216925757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916542, "dur": 10,
    "args": {
      "External id": 19598,
      "cbid": 211, "correlation": 19598
    }
  },
  {
    "ph": "s", "id": 19598, "pid": 76337, "tid": -914061504, "ts": 1716454216916542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925764, "dur": 5,
    "args": {
      "External id": 19606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19606,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19606, "pid": 5, "tid": 7, "ts": 1716454216925764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916573, "dur": 12,
    "args": {
      "External id": 19606,
      "cbid": 211, "correlation": 19606
    }
  },
  {
    "ph": "s", "id": 19606, "pid": 76337, "tid": -914061504, "ts": 1716454216916573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216925770, "dur": 4,
    "args": {
      "External id": 19615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19615,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19615, "pid": 5, "tid": 7, "ts": 1716454216925770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916608, "dur": 10,
    "args": {
      "External id": 19615,
      "cbid": 211, "correlation": 19615
    }
  },
  {
    "ph": "s", "id": 19615, "pid": 76337, "tid": -914061504, "ts": 1716454216916608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454216925775, "dur": 4,
    "args": {
      "External id": 19628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19628,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19628, "pid": 5, "tid": 7, "ts": 1716454216925775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916667, "dur": 14,
    "args": {
      "External id": 19628,
      "cbid": 211, "correlation": 19628
    }
  },
  {
    "ph": "s", "id": 19628, "pid": 76337, "tid": -914061504, "ts": 1716454216916667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454216925781, "dur": 8,
    "args": {
      "External id": 19638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19638,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 19638, "pid": 5, "tid": 7, "ts": 1716454216925781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916724, "dur": 13,
    "args": {
      "External id": 19638,
      "cbid": 211, "correlation": 19638
    }
  },
  {
    "ph": "s", "id": 19638, "pid": 76337, "tid": -914061504, "ts": 1716454216916724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916903, "dur": 3,
    "args": {
      "External id": 19655,
      "cbid": 251, "correlation": 19655
    }
  },
  {
    "ph": "f", "id": 19655, "pid": 76337, "tid": -914061504, "ts": 1716454216916903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454216925790, "dur": 12,
    "args": {
      "External id": 19657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19657,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19657, "pid": 5, "tid": 7, "ts": 1716454216925790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916916, "dur": 20,
    "args": {
      "External id": 19657,
      "cbid": 211, "correlation": 19657
    }
  },
  {
    "ph": "s", "id": 19657, "pid": 76337, "tid": -914061504, "ts": 1716454216916916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216925803, "dur": 3,
    "args": {
      "External id": 19665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19665,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19665, "pid": 5, "tid": 7, "ts": 1716454216925803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216916994, "dur": 13,
    "args": {
      "External id": 19665,
      "cbid": 211, "correlation": 19665
    }
  },
  {
    "ph": "s", "id": 19665, "pid": 76337, "tid": -914061504, "ts": 1716454216916994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917087, "dur": 2,
    "args": {
      "External id": 19681,
      "cbid": 251, "correlation": 19681
    }
  },
  {
    "ph": "f", "id": 19681, "pid": 76337, "tid": -914061504, "ts": 1716454216917087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917095, "dur": 0,
    "args": {
      "External id": 19683,
      "cbid": 251, "correlation": 19683
    }
  },
  {
    "ph": "f", "id": 19683, "pid": 76337, "tid": -914061504, "ts": 1716454216917095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216925808, "dur": 14,
    "args": {
      "External id": 19684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19684,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19684, "pid": 5, "tid": 7, "ts": 1716454216925808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917098, "dur": 15,
    "args": {
      "External id": 19684,
      "cbid": 211, "correlation": 19684
    }
  },
  {
    "ph": "s", "id": 19684, "pid": 76337, "tid": -914061504, "ts": 1716454216917098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216925823, "dur": 5,
    "args": {
      "External id": 19686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19686,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19686, "pid": 5, "tid": 7, "ts": 1716454216925823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917118, "dur": 8,
    "args": {
      "External id": 19686,
      "cbid": 211, "correlation": 19686
    }
  },
  {
    "ph": "s", "id": 19686, "pid": 76337, "tid": -914061504, "ts": 1716454216917118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917229, "dur": 1,
    "args": {
      "External id": 19696,
      "cbid": 317, "correlation": 19696
    }
  },
  {
    "ph": "f", "id": 19696, "pid": 76337, "tid": -914061504, "ts": 1716454216917229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917231, "dur": 1,
    "args": {
      "External id": 19697,
      "cbid": 203, "correlation": 19697
    }
  },
  {
    "ph": "f", "id": 19697, "pid": 76337, "tid": -914061504, "ts": 1716454216917231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216917233, "dur": 1,
    "args": {
      "External id": 19698,
      "cbid": 205, "correlation": 19698
    }
  },
  {
    "ph": "f", "id": 19698, "pid": 76337, "tid": -914061504, "ts": 1716454216917233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216925830, "dur": 6,
    "args": {
      "External id": 19702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19702,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19702, "pid": 5, "tid": 7, "ts": 1716454216925830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216920791, "dur": 31,
    "args": {
      "External id": 19702,
      "cbid": 211, "correlation": 19702
    }
  },
  {
    "ph": "s", "id": 19702, "pid": 76337, "tid": -914061504, "ts": 1716454216920791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216925837, "dur": 4,
    "args": {
      "External id": 19704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 19704, "pid": 5, "tid": 7, "ts": 1716454216925837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216920825, "dur": 6,
    "args": {
      "External id": 19704,
      "cbid": 211, "correlation": 19704
    }
  },
  {
    "ph": "s", "id": 19704, "pid": 76337, "tid": -914061504, "ts": 1716454216920825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216925842, "dur": 3,
    "args": {
      "External id": 19706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19706, "pid": 5, "tid": 7, "ts": 1716454216925842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216920839, "dur": 8,
    "args": {
      "External id": 19706,
      "cbid": 211, "correlation": 19706
    }
  },
  {
    "ph": "s", "id": 19706, "pid": 76337, "tid": -914061504, "ts": 1716454216920839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216920851, "dur": 0,
    "args": {
      "External id": 19707,
      "cbid": 51, "correlation": 19707
    }
  },
  {
    "ph": "s", "id": 19707, "pid": 76337, "tid": -914061504, "ts": 1716454216920851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216925847, "dur": 83,
    "args": {
      "External id": 19708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19708,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19708, "pid": 5, "tid": 7, "ts": 1716454216925847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216920852, "dur": 8,
    "args": {
      "External id": 19708,
      "cbid": 211, "correlation": 19708
    }
  },
  {
    "ph": "s", "id": 19708, "pid": 76337, "tid": -914061504, "ts": 1716454216920852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216925932, "dur": 59,
    "args": {
      "External id": 19713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19713, "pid": 5, "tid": 7, "ts": 1716454216925932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216921694, "dur": 16,
    "args": {
      "External id": 19713,
      "cbid": 211, "correlation": 19713
    }
  },
  {
    "ph": "s", "id": 19713, "pid": 76337, "tid": -914061504, "ts": 1716454216921694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216926687, "dur": 51,
    "args": {
      "External id": 19733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19733,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 19733, "pid": 5, "tid": 7, "ts": 1716454216926687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926659, "dur": 30,
    "args": {
      "External id": 19733,
      "cbid": 211, "correlation": 19733
    }
  },
  {
    "ph": "s", "id": 19733, "pid": 76337, "tid": -914061504, "ts": 1716454216926659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216926739, "dur": 4,
    "args": {
      "External id": 19745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19745,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19745, "pid": 5, "tid": 7, "ts": 1716454216926739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926714, "dur": 11,
    "args": {
      "External id": 19745,
      "cbid": 211, "correlation": 19745
    }
  },
  {
    "ph": "s", "id": 19745, "pid": 76337, "tid": -914061504, "ts": 1716454216926714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216926761, "dur": 58,
    "args": {
      "External id": 19748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19748,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19748, "pid": 5, "tid": 7, "ts": 1716454216926761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926751, "dur": 10,
    "args": {
      "External id": 19748,
      "cbid": 211, "correlation": 19748
    }
  },
  {
    "ph": "s", "id": 19748, "pid": 76337, "tid": -914061504, "ts": 1716454216926751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216926835, "dur": 36,
    "args": {
      "External id": 19757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19757,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19757, "pid": 5, "tid": 7, "ts": 1716454216926835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926820, "dur": 14,
    "args": {
      "External id": 19757,
      "cbid": 211, "correlation": 19757
    }
  },
  {
    "ph": "s", "id": 19757, "pid": 76337, "tid": -914061504, "ts": 1716454216926820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926934, "dur": 1,
    "args": {
      "External id": 19767,
      "cbid": 317, "correlation": 19767
    }
  },
  {
    "ph": "f", "id": 19767, "pid": 76337, "tid": -914061504, "ts": 1716454216926934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926935, "dur": 1,
    "args": {
      "External id": 19768,
      "cbid": 203, "correlation": 19768
    }
  },
  {
    "ph": "f", "id": 19768, "pid": 76337, "tid": -914061504, "ts": 1716454216926935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926938, "dur": 1,
    "args": {
      "External id": 19769,
      "cbid": 205, "correlation": 19769
    }
  },
  {
    "ph": "f", "id": 19769, "pid": 76337, "tid": -914061504, "ts": 1716454216926938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216926989, "dur": 40,
    "args": {
      "External id": 19773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19773, "pid": 5, "tid": 7, "ts": 1716454216926989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926967, "dur": 23,
    "args": {
      "External id": 19773,
      "cbid": 211, "correlation": 19773
    }
  },
  {
    "ph": "s", "id": 19773, "pid": 76337, "tid": -914061504, "ts": 1716454216926967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216927031, "dur": 15,
    "args": {
      "External id": 19775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19775,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19775, "pid": 5, "tid": 7, "ts": 1716454216927031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216926993, "dur": 6,
    "args": {
      "External id": 19775,
      "cbid": 211, "correlation": 19775
    }
  },
  {
    "ph": "s", "id": 19775, "pid": 76337, "tid": -914061504, "ts": 1716454216926993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216927047, "dur": 3,
    "args": {
      "External id": 19777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19777, "pid": 5, "tid": 7, "ts": 1716454216927047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927009, "dur": 9,
    "args": {
      "External id": 19777,
      "cbid": 211, "correlation": 19777
    }
  },
  {
    "ph": "s", "id": 19777, "pid": 76337, "tid": -914061504, "ts": 1716454216927009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927023, "dur": 0,
    "args": {
      "External id": 19778,
      "cbid": 51, "correlation": 19778
    }
  },
  {
    "ph": "s", "id": 19778, "pid": 76337, "tid": -914061504, "ts": 1716454216927023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216927051, "dur": 706,
    "args": {
      "External id": 19779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19779,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19779, "pid": 5, "tid": 7, "ts": 1716454216927051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927025, "dur": 8,
    "args": {
      "External id": 19779,
      "cbid": 211, "correlation": 19779
    }
  },
  {
    "ph": "s", "id": 19779, "pid": 76337, "tid": -914061504, "ts": 1716454216927025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216927759, "dur": 58,
    "args": {
      "External id": 19784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19784, "pid": 5, "tid": 7, "ts": 1716454216927759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927068, "dur": 12,
    "args": {
      "External id": 19784,
      "cbid": 211, "correlation": 19784
    }
  },
  {
    "ph": "s", "id": 19784, "pid": 76337, "tid": -914061504, "ts": 1716454216927068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216927819, "dur": 3,
    "args": {
      "External id": 19792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19792,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19792, "pid": 5, "tid": 7, "ts": 1716454216927819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927117, "dur": 9,
    "args": {
      "External id": 19792,
      "cbid": 211, "correlation": 19792
    }
  },
  {
    "ph": "s", "id": 19792, "pid": 76337, "tid": -914061504, "ts": 1716454216927117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927266, "dur": 5,
    "args": {
      "External id": 19808,
      "cbid": 251, "correlation": 19808
    }
  },
  {
    "ph": "f", "id": 19808, "pid": 76337, "tid": -914061504, "ts": 1716454216927266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927277, "dur": 0,
    "args": {
      "External id": 19810,
      "cbid": 251, "correlation": 19810
    }
  },
  {
    "ph": "f", "id": 19810, "pid": 76337, "tid": -914061504, "ts": 1716454216927277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454216927823, "dur": 9,
    "args": {
      "External id": 19811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19811,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 19811, "pid": 5, "tid": 7, "ts": 1716454216927823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927282, "dur": 16,
    "args": {
      "External id": 19811,
      "cbid": 211, "correlation": 19811
    }
  },
  {
    "ph": "s", "id": 19811, "pid": 76337, "tid": -914061504, "ts": 1716454216927282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454216927834, "dur": 4,
    "args": {
      "External id": 19813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19813,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 19813, "pid": 5, "tid": 7, "ts": 1716454216927834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927302, "dur": 9,
    "args": {
      "External id": 19813,
      "cbid": 211, "correlation": 19813
    }
  },
  {
    "ph": "s", "id": 19813, "pid": 76337, "tid": -914061504, "ts": 1716454216927302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216927839, "dur": 54,
    "args": {
      "External id": 19823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19823,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19823, "pid": 5, "tid": 7, "ts": 1716454216927839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927393, "dur": 12,
    "args": {
      "External id": 19823,
      "cbid": 211, "correlation": 19823
    }
  },
  {
    "ph": "s", "id": 19823, "pid": 76337, "tid": -914061504, "ts": 1716454216927393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216927895, "dur": 52,
    "args": {
      "External id": 19843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19843,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 19843, "pid": 5, "tid": 7, "ts": 1716454216927895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927462, "dur": 11,
    "args": {
      "External id": 19843,
      "cbid": 211, "correlation": 19843
    }
  },
  {
    "ph": "s", "id": 19843, "pid": 76337, "tid": -914061504, "ts": 1716454216927462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216927949, "dur": 4,
    "args": {
      "External id": 19855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19855,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19855, "pid": 5, "tid": 7, "ts": 1716454216927949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927483, "dur": 6,
    "args": {
      "External id": 19855,
      "cbid": 211, "correlation": 19855
    }
  },
  {
    "ph": "s", "id": 19855, "pid": 76337, "tid": -914061504, "ts": 1716454216927483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216927954, "dur": 56,
    "args": {
      "External id": 19858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19858, "pid": 5, "tid": 7, "ts": 1716454216927954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927501, "dur": 6,
    "args": {
      "External id": 19858,
      "cbid": 211, "correlation": 19858
    }
  },
  {
    "ph": "s", "id": 19858, "pid": 76337, "tid": -914061504, "ts": 1716454216927501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216928011, "dur": 36,
    "args": {
      "External id": 19867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19867,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19867, "pid": 5, "tid": 7, "ts": 1716454216928011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927543, "dur": 10,
    "args": {
      "External id": 19867,
      "cbid": 211, "correlation": 19867
    }
  },
  {
    "ph": "s", "id": 19867, "pid": 76337, "tid": -914061504, "ts": 1716454216927543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927619, "dur": 0,
    "args": {
      "External id": 19877,
      "cbid": 317, "correlation": 19877
    }
  },
  {
    "ph": "f", "id": 19877, "pid": 76337, "tid": -914061504, "ts": 1716454216927619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927619, "dur": 0,
    "args": {
      "External id": 19878,
      "cbid": 203, "correlation": 19878
    }
  },
  {
    "ph": "f", "id": 19878, "pid": 76337, "tid": -914061504, "ts": 1716454216927619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927620, "dur": 0,
    "args": {
      "External id": 19879,
      "cbid": 205, "correlation": 19879
    }
  },
  {
    "ph": "f", "id": 19879, "pid": 76337, "tid": -914061504, "ts": 1716454216927620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216928049, "dur": 41,
    "args": {
      "External id": 19883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19883,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19883, "pid": 5, "tid": 7, "ts": 1716454216928049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927635, "dur": 12,
    "args": {
      "External id": 19883,
      "cbid": 211, "correlation": 19883
    }
  },
  {
    "ph": "s", "id": 19883, "pid": 76337, "tid": -914061504, "ts": 1716454216927635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454216928091, "dur": 14,
    "args": {
      "External id": 19885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19885, "pid": 5, "tid": 7, "ts": 1716454216928091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927649, "dur": 5,
    "args": {
      "External id": 19885,
      "cbid": 211, "correlation": 19885
    }
  },
  {
    "ph": "s", "id": 19885, "pid": 76337, "tid": -914061504, "ts": 1716454216927649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454216928107, "dur": 3,
    "args": {
      "External id": 19887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19887, "pid": 5, "tid": 7, "ts": 1716454216928107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927658, "dur": 6,
    "args": {
      "External id": 19887,
      "cbid": 211, "correlation": 19887
    }
  },
  {
    "ph": "s", "id": 19887, "pid": 76337, "tid": -914061504, "ts": 1716454216927658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927666, "dur": 0,
    "args": {
      "External id": 19888,
      "cbid": 51, "correlation": 19888
    }
  },
  {
    "ph": "s", "id": 19888, "pid": 76337, "tid": -914061504, "ts": 1716454216927666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454216928111, "dur": 699,
    "args": {
      "External id": 19889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19889,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 19889, "pid": 5, "tid": 7, "ts": 1716454216928111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927667, "dur": 5,
    "args": {
      "External id": 19889,
      "cbid": 211, "correlation": 19889
    }
  },
  {
    "ph": "s", "id": 19889, "pid": 76337, "tid": -914061504, "ts": 1716454216927667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216928812, "dur": 59,
    "args": {
      "External id": 19894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19894,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19894, "pid": 5, "tid": 7, "ts": 1716454216928812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927695, "dur": 9,
    "args": {
      "External id": 19894,
      "cbid": 211, "correlation": 19894
    }
  },
  {
    "ph": "s", "id": 19894, "pid": 76337, "tid": -914061504, "ts": 1716454216927695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454216928873, "dur": 50,
    "args": {
      "External id": 19902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19902,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19902, "pid": 5, "tid": 7, "ts": 1716454216928873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927727, "dur": 10,
    "args": {
      "External id": 19902,
      "cbid": 211, "correlation": 19902
    }
  },
  {
    "ph": "s", "id": 19902, "pid": 76337, "tid": -914061504, "ts": 1716454216927727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454216928924, "dur": 36,
    "args": {
      "External id": 19910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19910,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19910, "pid": 5, "tid": 7, "ts": 1716454216928924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927776, "dur": 11,
    "args": {
      "External id": 19910,
      "cbid": 211, "correlation": 19910
    }
  },
  {
    "ph": "s", "id": 19910, "pid": 76337, "tid": -914061504, "ts": 1716454216927776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454216928962, "dur": 52,
    "args": {
      "External id": 19930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19930,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 19930, "pid": 5, "tid": 7, "ts": 1716454216928962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927907, "dur": 14,
    "args": {
      "External id": 19930,
      "cbid": 211, "correlation": 19930
    }
  },
  {
    "ph": "s", "id": 19930, "pid": 76337, "tid": -914061504, "ts": 1716454216927907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454216929015, "dur": 4,
    "args": {
      "External id": 19942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19942,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 19942, "pid": 5, "tid": 7, "ts": 1716454216929015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927931, "dur": 25,
    "args": {
      "External id": 19942,
      "cbid": 211, "correlation": 19942
    }
  },
  {
    "ph": "s", "id": 19942, "pid": 76337, "tid": -914061504, "ts": 1716454216927931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216929020, "dur": 56,
    "args": {
      "External id": 19945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19945, "pid": 5, "tid": 7, "ts": 1716454216929020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216927969, "dur": 16,
    "args": {
      "External id": 19945,
      "cbid": 211, "correlation": 19945
    }
  },
  {
    "ph": "s", "id": 19945, "pid": 76337, "tid": -914061504, "ts": 1716454216927969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928050, "dur": 0,
    "args": {
      "External id": 19956,
      "cbid": 317, "correlation": 19956
    }
  },
  {
    "ph": "f", "id": 19956, "pid": 76337, "tid": -914061504, "ts": 1716454216928050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928051, "dur": 0,
    "args": {
      "External id": 19957,
      "cbid": 203, "correlation": 19957
    }
  },
  {
    "ph": "f", "id": 19957, "pid": 76337, "tid": -914061504, "ts": 1716454216928051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928051, "dur": 0,
    "args": {
      "External id": 19958,
      "cbid": 205, "correlation": 19958
    }
  },
  {
    "ph": "f", "id": 19958, "pid": 76337, "tid": -914061504, "ts": 1716454216928051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928087, "dur": 3,
    "args": {
      "External id": 19962,
      "cbid": 251, "correlation": 19962
    }
  },
  {
    "ph": "f", "id": 19962, "pid": 76337, "tid": -914061504, "ts": 1716454216928087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928092, "dur": 1,
    "args": {
      "External id": 19963,
      "cbid": 251, "correlation": 19963
    }
  },
  {
    "ph": "f", "id": 19963, "pid": 76337, "tid": -914061504, "ts": 1716454216928092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928094, "dur": 1,
    "args": {
      "External id": 19964,
      "cbid": 251, "correlation": 19964
    }
  },
  {
    "ph": "f", "id": 19964, "pid": 76337, "tid": -914061504, "ts": 1716454216928094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928096, "dur": 1,
    "args": {
      "External id": 19965,
      "cbid": 251, "correlation": 19965
    }
  },
  {
    "ph": "f", "id": 19965, "pid": 76337, "tid": -914061504, "ts": 1716454216928096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928098, "dur": 1,
    "args": {
      "External id": 19966,
      "cbid": 251, "correlation": 19966
    }
  },
  {
    "ph": "f", "id": 19966, "pid": 76337, "tid": -914061504, "ts": 1716454216928098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928100, "dur": 1,
    "args": {
      "External id": 19967,
      "cbid": 251, "correlation": 19967
    }
  },
  {
    "ph": "f", "id": 19967, "pid": 76337, "tid": -914061504, "ts": 1716454216928100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928102, "dur": 1,
    "args": {
      "External id": 19968,
      "cbid": 251, "correlation": 19968
    }
  },
  {
    "ph": "f", "id": 19968, "pid": 76337, "tid": -914061504, "ts": 1716454216928102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928104, "dur": 1,
    "args": {
      "External id": 19969,
      "cbid": 251, "correlation": 19969
    }
  },
  {
    "ph": "f", "id": 19969, "pid": 76337, "tid": -914061504, "ts": 1716454216928104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928106, "dur": 0,
    "args": {
      "External id": 19970,
      "cbid": 251, "correlation": 19970
    }
  },
  {
    "ph": "f", "id": 19970, "pid": 76337, "tid": -914061504, "ts": 1716454216928106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454216929077, "dur": 116,
    "args": {
      "External id": 19971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19971,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 19971, "pid": 5, "tid": 7, "ts": 1716454216929077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928110, "dur": 15,
    "args": {
      "External id": 19971,
      "cbid": 211, "correlation": 19971
    }
  },
  {
    "ph": "s", "id": 19971, "pid": 76337, "tid": -914061504, "ts": 1716454216928110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454216929195, "dur": 60,
    "args": {
      "External id": 19977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19977, "pid": 5, "tid": 7, "ts": 1716454216929195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928158, "dur": 10,
    "args": {
      "External id": 19977,
      "cbid": 211, "correlation": 19977
    }
  },
  {
    "ph": "s", "id": 19977, "pid": 76337, "tid": -914061504, "ts": 1716454216928158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217311368, "dur": 609,
    "args": {
      "External id": 19986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 19986,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 19986, "pid": 5, "tid": 7, "ts": 1716454217311368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454216928317, "dur": 383057,
    "args": {
      "External id": 19986,
      "cbid": 211, "correlation": 19986
    }
  },
  {
    "ph": "s", "id": 19986, "pid": 76337, "tid": -914061504, "ts": 1716454216928317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217311979, "dur": 177,
    "args": {
      "External id": 20008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20008,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20008, "pid": 5, "tid": 7, "ts": 1716454217311979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311544, "dur": 17,
    "args": {
      "External id": 20008,
      "cbid": 211, "correlation": 20008
    }
  },
  {
    "ph": "s", "id": 20008, "pid": 76337, "tid": -914061504, "ts": 1716454217311544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311768, "dur": 4,
    "args": {
      "External id": 20019,
      "cbid": 251, "correlation": 20019
    }
  },
  {
    "ph": "f", "id": 20019, "pid": 76337, "tid": -914061504, "ts": 1716454217311768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217312158, "dur": 195,
    "args": {
      "External id": 20020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20020,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20020, "pid": 5, "tid": 7, "ts": 1716454217312158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311780, "dur": 17,
    "args": {
      "External id": 20020,
      "cbid": 211, "correlation": 20020
    }
  },
  {
    "ph": "s", "id": 20020, "pid": 76337, "tid": -914061504, "ts": 1716454217311780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311862, "dur": 1,
    "args": {
      "External id": 20031,
      "cbid": 251, "correlation": 20031
    }
  },
  {
    "ph": "f", "id": 20031, "pid": 76337, "tid": -914061504, "ts": 1716454217311862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217312354, "dur": 188,
    "args": {
      "External id": 20032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20032,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20032, "pid": 5, "tid": 7, "ts": 1716454217312354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311866, "dur": 11,
    "args": {
      "External id": 20032,
      "cbid": 211, "correlation": 20032
    }
  },
  {
    "ph": "s", "id": 20032, "pid": 76337, "tid": -914061504, "ts": 1716454217311866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311931, "dur": 1,
    "args": {
      "External id": 20043,
      "cbid": 251, "correlation": 20043
    }
  },
  {
    "ph": "f", "id": 20043, "pid": 76337, "tid": -914061504, "ts": 1716454217311931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217312543, "dur": 185,
    "args": {
      "External id": 20044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20044,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20044, "pid": 5, "tid": 7, "ts": 1716454217312543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217311935, "dur": 12,
    "args": {
      "External id": 20044,
      "cbid": 211, "correlation": 20044
    }
  },
  {
    "ph": "s", "id": 20044, "pid": 76337, "tid": -914061504, "ts": 1716454217311935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217312729, "dur": 18232,
    "args": {
      "External id": 20065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20065,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20065, "pid": 5, "tid": 7, "ts": 1716454217312729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312067, "dur": 16,
    "args": {
      "External id": 20065,
      "cbid": 211, "correlation": 20065
    }
  },
  {
    "ph": "s", "id": 20065, "pid": 76337, "tid": -914061504, "ts": 1716454217312067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312209, "dur": 3,
    "args": {
      "External id": 20083,
      "cbid": 251, "correlation": 20083
    }
  },
  {
    "ph": "f", "id": 20083, "pid": 76337, "tid": -914061504, "ts": 1716454217312209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217330962, "dur": 199,
    "args": {
      "External id": 20085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20085,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20085, "pid": 5, "tid": 7, "ts": 1716454217330962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312217, "dur": 14,
    "args": {
      "External id": 20085,
      "cbid": 211, "correlation": 20085
    }
  },
  {
    "ph": "s", "id": 20085, "pid": 76337, "tid": -914061504, "ts": 1716454217312217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217331163, "dur": 66,
    "args": {
      "External id": 20093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20093,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20093, "pid": 5, "tid": 7, "ts": 1716454217331163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312328, "dur": 14,
    "args": {
      "External id": 20093,
      "cbid": 211, "correlation": 20093
    }
  },
  {
    "ph": "s", "id": 20093, "pid": 76337, "tid": -914061504, "ts": 1716454217312328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217331230, "dur": 97,
    "args": {
      "External id": 20101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20101,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20101, "pid": 5, "tid": 7, "ts": 1716454217331230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312379, "dur": 10,
    "args": {
      "External id": 20101,
      "cbid": 211, "correlation": 20101
    }
  },
  {
    "ph": "s", "id": 20101, "pid": 76337, "tid": -914061504, "ts": 1716454217312379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217331328, "dur": 54,
    "args": {
      "External id": 20112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20112, "pid": 5, "tid": 7, "ts": 1716454217331328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312490, "dur": 16,
    "args": {
      "External id": 20112,
      "cbid": 211, "correlation": 20112
    }
  },
  {
    "ph": "s", "id": 20112, "pid": 76337, "tid": -914061504, "ts": 1716454217312490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217331384, "dur": 90,
    "args": {
      "External id": 20134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20134,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20134, "pid": 5, "tid": 7, "ts": 1716454217331384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312537, "dur": 10,
    "args": {
      "External id": 20134,
      "cbid": 211, "correlation": 20134
    }
  },
  {
    "ph": "s", "id": 20134, "pid": 76337, "tid": -914061504, "ts": 1716454217312537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312628, "dur": 1,
    "args": {
      "External id": 20145,
      "cbid": 251, "correlation": 20145
    }
  },
  {
    "ph": "f", "id": 20145, "pid": 76337, "tid": -914061504, "ts": 1716454217312628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217331475, "dur": 104,
    "args": {
      "External id": 20146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20146,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20146, "pid": 5, "tid": 7, "ts": 1716454217331475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312634, "dur": 13,
    "args": {
      "External id": 20146,
      "cbid": 211, "correlation": 20146
    }
  },
  {
    "ph": "s", "id": 20146, "pid": 76337, "tid": -914061504, "ts": 1716454217312634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312721, "dur": 1,
    "args": {
      "External id": 20157,
      "cbid": 251, "correlation": 20157
    }
  },
  {
    "ph": "f", "id": 20157, "pid": 76337, "tid": -914061504, "ts": 1716454217312721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312726, "dur": 0,
    "args": {
      "External id": 20158,
      "cbid": 251, "correlation": 20158
    }
  },
  {
    "ph": "f", "id": 20158, "pid": 76337, "tid": -914061504, "ts": 1716454217312726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217331580, "dur": 10,
    "args": {
      "External id": 20159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20159,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 20159, "pid": 5, "tid": 7, "ts": 1716454217331580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312728, "dur": 14,
    "args": {
      "External id": 20159,
      "cbid": 211, "correlation": 20159
    }
  },
  {
    "ph": "s", "id": 20159, "pid": 76337, "tid": -914061504, "ts": 1716454217312728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217331592, "dur": 5,
    "args": {
      "External id": 20161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20161,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 20161, "pid": 5, "tid": 7, "ts": 1716454217331592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312746, "dur": 10,
    "args": {
      "External id": 20161,
      "cbid": 211, "correlation": 20161
    }
  },
  {
    "ph": "s", "id": 20161, "pid": 76337, "tid": -914061504, "ts": 1716454217312746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312812, "dur": 1,
    "args": {
      "External id": 20172,
      "cbid": 251, "correlation": 20172
    }
  },
  {
    "ph": "f", "id": 20172, "pid": 76337, "tid": -914061504, "ts": 1716454217312812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312815, "dur": 0,
    "args": {
      "External id": 20173,
      "cbid": 251, "correlation": 20173
    }
  },
  {
    "ph": "f", "id": 20173, "pid": 76337, "tid": -914061504, "ts": 1716454217312815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217331598, "dur": 6,
    "args": {
      "External id": 20174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20174,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 20174, "pid": 5, "tid": 7, "ts": 1716454217331598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312817, "dur": 12,
    "args": {
      "External id": 20174,
      "cbid": 211, "correlation": 20174
    }
  },
  {
    "ph": "s", "id": 20174, "pid": 76337, "tid": -914061504, "ts": 1716454217312817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217331605, "dur": 3,
    "args": {
      "External id": 20176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20176,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 20176, "pid": 5, "tid": 7, "ts": 1716454217331605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312831, "dur": 6,
    "args": {
      "External id": 20176,
      "cbid": 211, "correlation": 20176
    }
  },
  {
    "ph": "s", "id": 20176, "pid": 76337, "tid": -914061504, "ts": 1716454217312831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217331610, "dur": 152,
    "args": {
      "External id": 20197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20197,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20197, "pid": 5, "tid": 7, "ts": 1716454217331610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217312906, "dur": 12,
    "args": {
      "External id": 20197,
      "cbid": 211, "correlation": 20197
    }
  },
  {
    "ph": "s", "id": 20197, "pid": 76337, "tid": -914061504, "ts": 1716454217312906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313017, "dur": 2,
    "args": {
      "External id": 20215,
      "cbid": 251, "correlation": 20215
    }
  },
  {
    "ph": "f", "id": 20215, "pid": 76337, "tid": -914061504, "ts": 1716454217313017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217331763, "dur": 105,
    "args": {
      "External id": 20217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20217,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20217, "pid": 5, "tid": 7, "ts": 1716454217331763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313025, "dur": 15,
    "args": {
      "External id": 20217,
      "cbid": 211, "correlation": 20217
    }
  },
  {
    "ph": "s", "id": 20217, "pid": 76337, "tid": -914061504, "ts": 1716454217313025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217331870, "dur": 35,
    "args": {
      "External id": 20225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20225,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20225, "pid": 5, "tid": 7, "ts": 1716454217331870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313099, "dur": 12,
    "args": {
      "External id": 20225,
      "cbid": 211, "correlation": 20225
    }
  },
  {
    "ph": "s", "id": 20225, "pid": 76337, "tid": -914061504, "ts": 1716454217313099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217331906, "dur": 67,
    "args": {
      "External id": 20233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20233,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20233, "pid": 5, "tid": 7, "ts": 1716454217331906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313142, "dur": 10,
    "args": {
      "External id": 20233,
      "cbid": 211, "correlation": 20233
    }
  },
  {
    "ph": "s", "id": 20233, "pid": 76337, "tid": -914061504, "ts": 1716454217313142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217331975, "dur": 92,
    "args": {
      "External id": 20255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20255,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20255, "pid": 5, "tid": 7, "ts": 1716454217331975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313194, "dur": 11,
    "args": {
      "External id": 20255,
      "cbid": 211, "correlation": 20255
    }
  },
  {
    "ph": "s", "id": 20255, "pid": 76337, "tid": -914061504, "ts": 1716454217313194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313314, "dur": 1,
    "args": {
      "External id": 20271,
      "cbid": 251, "correlation": 20271
    }
  },
  {
    "ph": "f", "id": 20271, "pid": 76337, "tid": -914061504, "ts": 1716454217313314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217332068, "dur": 566,
    "args": {
      "External id": 20273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20273,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20273, "pid": 5, "tid": 7, "ts": 1716454217332068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313320, "dur": 14,
    "args": {
      "External id": 20273,
      "cbid": 211, "correlation": 20273
    }
  },
  {
    "ph": "s", "id": 20273, "pid": 76337, "tid": -914061504, "ts": 1716454217313320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217332635, "dur": 240,
    "args": {
      "External id": 20281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20281, "pid": 5, "tid": 7, "ts": 1716454217332635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313413, "dur": 16,
    "args": {
      "External id": 20281,
      "cbid": 211, "correlation": 20281
    }
  },
  {
    "ph": "s", "id": 20281, "pid": 76337, "tid": -914061504, "ts": 1716454217313413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313453, "dur": 1,
    "args": {
      "External id": 20289,
      "cbid": 317, "correlation": 20289
    }
  },
  {
    "ph": "f", "id": 20289, "pid": 76337, "tid": -914061504, "ts": 1716454217313453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454217313455, "dur": 292036,
    "args": {
      "External id": 20290,
      "cbid": 20, "correlation": 20290
    }
  },
  {
    "ph": "f", "id": 20290, "pid": 76337, "tid": -914061504, "ts": 1716454217313455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217605612, "dur": 252,
    "args": {
      "External id": 20293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20293,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20293, "pid": 5, "tid": 7, "ts": 1716454217605612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217605550, "dur": 64,
    "args": {
      "External id": 20293,
      "cbid": 211, "correlation": 20293
    }
  },
  {
    "ph": "s", "id": 20293, "pid": 76337, "tid": -914061504, "ts": 1716454217605550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217605864, "dur": 11,
    "args": {
      "External id": 20309,
      "cbid": 251, "correlation": 20309
    }
  },
  {
    "ph": "f", "id": 20309, "pid": 76337, "tid": -914061504, "ts": 1716454217605864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217605881, "dur": 0,
    "args": {
      "External id": 20311,
      "cbid": 251, "correlation": 20311
    }
  },
  {
    "ph": "f", "id": 20311, "pid": 76337, "tid": -914061504, "ts": 1716454217605881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217605916, "dur": 354,
    "args": {
      "External id": 20312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20312,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 20312, "pid": 5, "tid": 7, "ts": 1716454217605916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217605893, "dur": 24,
    "args": {
      "External id": 20312,
      "cbid": 211, "correlation": 20312
    }
  },
  {
    "ph": "s", "id": 20312, "pid": 76337, "tid": -914061504, "ts": 1716454217605893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217606272, "dur": 50,
    "args": {
      "External id": 20320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20320,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20320, "pid": 5, "tid": 7, "ts": 1716454217606272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217605971, "dur": 25,
    "args": {
      "External id": 20320,
      "cbid": 211, "correlation": 20320
    }
  },
  {
    "ph": "s", "id": 20320, "pid": 76337, "tid": -914061504, "ts": 1716454217605971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217606323, "dur": 155,
    "args": {
      "External id": 20331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20331,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20331, "pid": 5, "tid": 7, "ts": 1716454217606323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606133, "dur": 20,
    "args": {
      "External id": 20331,
      "cbid": 211, "correlation": 20331
    }
  },
  {
    "ph": "s", "id": 20331, "pid": 76337, "tid": -914061504, "ts": 1716454217606133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606283, "dur": 1,
    "args": {
      "External id": 20343,
      "cbid": 317, "correlation": 20343
    }
  },
  {
    "ph": "f", "id": 20343, "pid": 76337, "tid": -914061504, "ts": 1716454217606283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606284, "dur": 1,
    "args": {
      "External id": 20344,
      "cbid": 203, "correlation": 20344
    }
  },
  {
    "ph": "f", "id": 20344, "pid": 76337, "tid": -914061504, "ts": 1716454217606284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606287, "dur": 1,
    "args": {
      "External id": 20345,
      "cbid": 205, "correlation": 20345
    }
  },
  {
    "ph": "f", "id": 20345, "pid": 76337, "tid": -914061504, "ts": 1716454217606287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606338, "dur": 2,
    "args": {
      "External id": 20349,
      "cbid": 251, "correlation": 20349
    }
  },
  {
    "ph": "f", "id": 20349, "pid": 76337, "tid": -914061504, "ts": 1716454217606338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606342, "dur": 1,
    "args": {
      "External id": 20350,
      "cbid": 251, "correlation": 20350
    }
  },
  {
    "ph": "f", "id": 20350, "pid": 76337, "tid": -914061504, "ts": 1716454217606342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606343, "dur": 1,
    "args": {
      "External id": 20351,
      "cbid": 251, "correlation": 20351
    }
  },
  {
    "ph": "f", "id": 20351, "pid": 76337, "tid": -914061504, "ts": 1716454217606343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606345, "dur": 1,
    "args": {
      "External id": 20352,
      "cbid": 251, "correlation": 20352
    }
  },
  {
    "ph": "f", "id": 20352, "pid": 76337, "tid": -914061504, "ts": 1716454217606345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606347, "dur": 1,
    "args": {
      "External id": 20353,
      "cbid": 251, "correlation": 20353
    }
  },
  {
    "ph": "f", "id": 20353, "pid": 76337, "tid": -914061504, "ts": 1716454217606347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606349, "dur": 1,
    "args": {
      "External id": 20354,
      "cbid": 251, "correlation": 20354
    }
  },
  {
    "ph": "f", "id": 20354, "pid": 76337, "tid": -914061504, "ts": 1716454217606349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606351, "dur": 1,
    "args": {
      "External id": 20355,
      "cbid": 251, "correlation": 20355
    }
  },
  {
    "ph": "f", "id": 20355, "pid": 76337, "tid": -914061504, "ts": 1716454217606351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606353, "dur": 1,
    "args": {
      "External id": 20356,
      "cbid": 251, "correlation": 20356
    }
  },
  {
    "ph": "f", "id": 20356, "pid": 76337, "tid": -914061504, "ts": 1716454217606353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606356, "dur": 0,
    "args": {
      "External id": 20357,
      "cbid": 251, "correlation": 20357
    }
  },
  {
    "ph": "f", "id": 20357, "pid": 76337, "tid": -914061504, "ts": 1716454217606356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217606479, "dur": 115,
    "args": {
      "External id": 20358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20358,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20358, "pid": 5, "tid": 7, "ts": 1716454217606479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606359, "dur": 14,
    "args": {
      "External id": 20358,
      "cbid": 211, "correlation": 20358
    }
  },
  {
    "ph": "s", "id": 20358, "pid": 76337, "tid": -914061504, "ts": 1716454217606359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217606595, "dur": 59,
    "args": {
      "External id": 20364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20364, "pid": 5, "tid": 7, "ts": 1716454217606595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606404, "dur": 10,
    "args": {
      "External id": 20364,
      "cbid": 211, "correlation": 20364
    }
  },
  {
    "ph": "s", "id": 20364, "pid": 76337, "tid": -914061504, "ts": 1716454217606404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217606655, "dur": 50,
    "args": {
      "External id": 20372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20372,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20372, "pid": 5, "tid": 7, "ts": 1716454217606655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606440, "dur": 10,
    "args": {
      "External id": 20372,
      "cbid": 211, "correlation": 20372
    }
  },
  {
    "ph": "s", "id": 20372, "pid": 76337, "tid": -914061504, "ts": 1716454217606440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217606706, "dur": 52,
    "args": {
      "External id": 20392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20392,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 20392, "pid": 5, "tid": 7, "ts": 1716454217606706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606573, "dur": 18,
    "args": {
      "External id": 20392,
      "cbid": 211, "correlation": 20392
    }
  },
  {
    "ph": "s", "id": 20392, "pid": 76337, "tid": -914061504, "ts": 1716454217606573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217606760, "dur": 4,
    "args": {
      "External id": 20404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20404,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 20404, "pid": 5, "tid": 7, "ts": 1716454217606760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606610, "dur": 11,
    "args": {
      "External id": 20404,
      "cbid": 211, "correlation": 20404
    }
  },
  {
    "ph": "s", "id": 20404, "pid": 76337, "tid": -914061504, "ts": 1716454217606610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217606765, "dur": 56,
    "args": {
      "External id": 20407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20407,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20407, "pid": 5, "tid": 7, "ts": 1716454217606765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606643, "dur": 9,
    "args": {
      "External id": 20407,
      "cbid": 211, "correlation": 20407
    }
  },
  {
    "ph": "s", "id": 20407, "pid": 76337, "tid": -914061504, "ts": 1716454217606643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217606822, "dur": 37,
    "args": {
      "External id": 20416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20416,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20416, "pid": 5, "tid": 7, "ts": 1716454217606822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606701, "dur": 12,
    "args": {
      "External id": 20416,
      "cbid": 211, "correlation": 20416
    }
  },
  {
    "ph": "s", "id": 20416, "pid": 76337, "tid": -914061504, "ts": 1716454217606701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606761, "dur": 0,
    "args": {
      "External id": 20426,
      "cbid": 317, "correlation": 20426
    }
  },
  {
    "ph": "f", "id": 20426, "pid": 76337, "tid": -914061504, "ts": 1716454217606761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606762, "dur": 0,
    "args": {
      "External id": 20427,
      "cbid": 203, "correlation": 20427
    }
  },
  {
    "ph": "f", "id": 20427, "pid": 76337, "tid": -914061504, "ts": 1716454217606762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606762, "dur": 0,
    "args": {
      "External id": 20428,
      "cbid": 205, "correlation": 20428
    }
  },
  {
    "ph": "f", "id": 20428, "pid": 76337, "tid": -914061504, "ts": 1716454217606762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217606861, "dur": 41,
    "args": {
      "External id": 20432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20432, "pid": 5, "tid": 7, "ts": 1716454217606861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606802, "dur": 28,
    "args": {
      "External id": 20432,
      "cbid": 211, "correlation": 20432
    }
  },
  {
    "ph": "s", "id": 20432, "pid": 76337, "tid": -914061504, "ts": 1716454217606802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217606903, "dur": 14,
    "args": {
      "External id": 20434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20434,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20434, "pid": 5, "tid": 7, "ts": 1716454217606903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606833, "dur": 6,
    "args": {
      "External id": 20434,
      "cbid": 211, "correlation": 20434
    }
  },
  {
    "ph": "s", "id": 20434, "pid": 76337, "tid": -914061504, "ts": 1716454217606833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217606918, "dur": 3,
    "args": {
      "External id": 20436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 20436, "pid": 5, "tid": 7, "ts": 1716454217606918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606859, "dur": 25,
    "args": {
      "External id": 20436,
      "cbid": 211, "correlation": 20436
    }
  },
  {
    "ph": "s", "id": 20436, "pid": 76337, "tid": -914061504, "ts": 1716454217606859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606892, "dur": 0,
    "args": {
      "External id": 20437,
      "cbid": 51, "correlation": 20437
    }
  },
  {
    "ph": "s", "id": 20437, "pid": 76337, "tid": -914061504, "ts": 1716454217606892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217606923, "dur": 689,
    "args": {
      "External id": 20438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20438,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20438, "pid": 5, "tid": 7, "ts": 1716454217606923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606894, "dur": 17,
    "args": {
      "External id": 20438,
      "cbid": 211, "correlation": 20438
    }
  },
  {
    "ph": "s", "id": 20438, "pid": 76337, "tid": -914061504, "ts": 1716454217606894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217607613, "dur": 59,
    "args": {
      "External id": 20443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20443, "pid": 5, "tid": 7, "ts": 1716454217607613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606936, "dur": 9,
    "args": {
      "External id": 20443,
      "cbid": 211, "correlation": 20443
    }
  },
  {
    "ph": "s", "id": 20443, "pid": 76337, "tid": -914061504, "ts": 1716454217606936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217607673, "dur": 4,
    "args": {
      "External id": 20451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20451,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 20451, "pid": 5, "tid": 7, "ts": 1716454217607673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217606993, "dur": 10,
    "args": {
      "External id": 20451,
      "cbid": 211, "correlation": 20451
    }
  },
  {
    "ph": "s", "id": 20451, "pid": 76337, "tid": -914061504, "ts": 1716454217606993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607091, "dur": 2,
    "args": {
      "External id": 20467,
      "cbid": 251, "correlation": 20467
    }
  },
  {
    "ph": "f", "id": 20467, "pid": 76337, "tid": -914061504, "ts": 1716454217607091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607097, "dur": 0,
    "args": {
      "External id": 20469,
      "cbid": 251, "correlation": 20469
    }
  },
  {
    "ph": "f", "id": 20469, "pid": 76337, "tid": -914061504, "ts": 1716454217607097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217607678, "dur": 11,
    "args": {
      "External id": 20470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20470,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 20470, "pid": 5, "tid": 7, "ts": 1716454217607678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607099, "dur": 14,
    "args": {
      "External id": 20470,
      "cbid": 211, "correlation": 20470
    }
  },
  {
    "ph": "s", "id": 20470, "pid": 76337, "tid": -914061504, "ts": 1716454217607099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217607691, "dur": 5,
    "args": {
      "External id": 20472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20472,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 20472, "pid": 5, "tid": 7, "ts": 1716454217607691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607119, "dur": 8,
    "args": {
      "External id": 20472,
      "cbid": 211, "correlation": 20472
    }
  },
  {
    "ph": "s", "id": 20472, "pid": 76337, "tid": -914061504, "ts": 1716454217607119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217607697, "dur": 55,
    "args": {
      "External id": 20482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20482,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20482, "pid": 5, "tid": 7, "ts": 1716454217607697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607198, "dur": 12,
    "args": {
      "External id": 20482,
      "cbid": 211, "correlation": 20482
    }
  },
  {
    "ph": "s", "id": 20482, "pid": 76337, "tid": -914061504, "ts": 1716454217607198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217607753, "dur": 51,
    "args": {
      "External id": 20502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20502,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 20502, "pid": 5, "tid": 7, "ts": 1716454217607753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607265, "dur": 11,
    "args": {
      "External id": 20502,
      "cbid": 211, "correlation": 20502
    }
  },
  {
    "ph": "s", "id": 20502, "pid": 76337, "tid": -914061504, "ts": 1716454217607265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217607805, "dur": 4,
    "args": {
      "External id": 20514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20514,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 20514, "pid": 5, "tid": 7, "ts": 1716454217607805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607297, "dur": 8,
    "args": {
      "External id": 20514,
      "cbid": 211, "correlation": 20514
    }
  },
  {
    "ph": "s", "id": 20514, "pid": 76337, "tid": -914061504, "ts": 1716454217607297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217607810, "dur": 54,
    "args": {
      "External id": 20517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20517, "pid": 5, "tid": 7, "ts": 1716454217607810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607318, "dur": 6,
    "args": {
      "External id": 20517,
      "cbid": 211, "correlation": 20517
    }
  },
  {
    "ph": "s", "id": 20517, "pid": 76337, "tid": -914061504, "ts": 1716454217607318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217607866, "dur": 36,
    "args": {
      "External id": 20526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20526,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20526, "pid": 5, "tid": 7, "ts": 1716454217607866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607359, "dur": 10,
    "args": {
      "External id": 20526,
      "cbid": 211, "correlation": 20526
    }
  },
  {
    "ph": "s", "id": 20526, "pid": 76337, "tid": -914061504, "ts": 1716454217607359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607425, "dur": 0,
    "args": {
      "External id": 20536,
      "cbid": 317, "correlation": 20536
    }
  },
  {
    "ph": "f", "id": 20536, "pid": 76337, "tid": -914061504, "ts": 1716454217607425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607425, "dur": 0,
    "args": {
      "External id": 20537,
      "cbid": 203, "correlation": 20537
    }
  },
  {
    "ph": "f", "id": 20537, "pid": 76337, "tid": -914061504, "ts": 1716454217607425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607426, "dur": 0,
    "args": {
      "External id": 20538,
      "cbid": 205, "correlation": 20538
    }
  },
  {
    "ph": "f", "id": 20538, "pid": 76337, "tid": -914061504, "ts": 1716454217607426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217607903, "dur": 39,
    "args": {
      "External id": 20542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20542, "pid": 5, "tid": 7, "ts": 1716454217607903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607440, "dur": 12,
    "args": {
      "External id": 20542,
      "cbid": 211, "correlation": 20542
    }
  },
  {
    "ph": "s", "id": 20542, "pid": 76337, "tid": -914061504, "ts": 1716454217607440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217607944, "dur": 14,
    "args": {
      "External id": 20544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20544,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20544, "pid": 5, "tid": 7, "ts": 1716454217607944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607455, "dur": 5,
    "args": {
      "External id": 20544,
      "cbid": 211, "correlation": 20544
    }
  },
  {
    "ph": "s", "id": 20544, "pid": 76337, "tid": -914061504, "ts": 1716454217607455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217607959, "dur": 3,
    "args": {
      "External id": 20546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 20546, "pid": 5, "tid": 7, "ts": 1716454217607959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607464, "dur": 6,
    "args": {
      "External id": 20546,
      "cbid": 211, "correlation": 20546
    }
  },
  {
    "ph": "s", "id": 20546, "pid": 76337, "tid": -914061504, "ts": 1716454217607464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607473, "dur": 0,
    "args": {
      "External id": 20547,
      "cbid": 51, "correlation": 20547
    }
  },
  {
    "ph": "s", "id": 20547, "pid": 76337, "tid": -914061504, "ts": 1716454217607473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217607964, "dur": 685,
    "args": {
      "External id": 20548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20548,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20548, "pid": 5, "tid": 7, "ts": 1716454217607964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607474, "dur": 5,
    "args": {
      "External id": 20548,
      "cbid": 211, "correlation": 20548
    }
  },
  {
    "ph": "s", "id": 20548, "pid": 76337, "tid": -914061504, "ts": 1716454217607474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217608650, "dur": 58,
    "args": {
      "External id": 20553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20553, "pid": 5, "tid": 7, "ts": 1716454217608650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607501, "dur": 9,
    "args": {
      "External id": 20553,
      "cbid": 211, "correlation": 20553
    }
  },
  {
    "ph": "s", "id": 20553, "pid": 76337, "tid": -914061504, "ts": 1716454217607501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217608710, "dur": 50,
    "args": {
      "External id": 20561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20561,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20561, "pid": 5, "tid": 7, "ts": 1716454217608710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607534, "dur": 9,
    "args": {
      "External id": 20561,
      "cbid": 211, "correlation": 20561
    }
  },
  {
    "ph": "s", "id": 20561, "pid": 76337, "tid": -914061504, "ts": 1716454217607534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217608761, "dur": 35,
    "args": {
      "External id": 20569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20569,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20569, "pid": 5, "tid": 7, "ts": 1716454217608761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607578, "dur": 11,
    "args": {
      "External id": 20569,
      "cbid": 211, "correlation": 20569
    }
  },
  {
    "ph": "s", "id": 20569, "pid": 76337, "tid": -914061504, "ts": 1716454217607578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217608797, "dur": 52,
    "args": {
      "External id": 20589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20589,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 20589, "pid": 5, "tid": 7, "ts": 1716454217608797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607664, "dur": 12,
    "args": {
      "External id": 20589,
      "cbid": 211, "correlation": 20589
    }
  },
  {
    "ph": "s", "id": 20589, "pid": 76337, "tid": -914061504, "ts": 1716454217607664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217608851, "dur": 4,
    "args": {
      "External id": 20601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20601,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 20601, "pid": 5, "tid": 7, "ts": 1716454217608851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607686, "dur": 7,
    "args": {
      "External id": 20601,
      "cbid": 211, "correlation": 20601
    }
  },
  {
    "ph": "s", "id": 20601, "pid": 76337, "tid": -914061504, "ts": 1716454217607686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217608856, "dur": 54,
    "args": {
      "External id": 20604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20604, "pid": 5, "tid": 7, "ts": 1716454217608856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607703, "dur": 6,
    "args": {
      "External id": 20604,
      "cbid": 211, "correlation": 20604
    }
  },
  {
    "ph": "s", "id": 20604, "pid": 76337, "tid": -914061504, "ts": 1716454217607703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607762, "dur": 0,
    "args": {
      "External id": 20615,
      "cbid": 317, "correlation": 20615
    }
  },
  {
    "ph": "f", "id": 20615, "pid": 76337, "tid": -914061504, "ts": 1716454217607762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607762, "dur": 0,
    "args": {
      "External id": 20616,
      "cbid": 203, "correlation": 20616
    }
  },
  {
    "ph": "f", "id": 20616, "pid": 76337, "tid": -914061504, "ts": 1716454217607762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607763, "dur": 0,
    "args": {
      "External id": 20617,
      "cbid": 205, "correlation": 20617
    }
  },
  {
    "ph": "f", "id": 20617, "pid": 76337, "tid": -914061504, "ts": 1716454217607763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607784, "dur": 1,
    "args": {
      "External id": 20621,
      "cbid": 251, "correlation": 20621
    }
  },
  {
    "ph": "f", "id": 20621, "pid": 76337, "tid": -914061504, "ts": 1716454217607784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607787, "dur": 0,
    "args": {
      "External id": 20622,
      "cbid": 251, "correlation": 20622
    }
  },
  {
    "ph": "f", "id": 20622, "pid": 76337, "tid": -914061504, "ts": 1716454217607787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607787, "dur": 0,
    "args": {
      "External id": 20623,
      "cbid": 251, "correlation": 20623
    }
  },
  {
    "ph": "f", "id": 20623, "pid": 76337, "tid": -914061504, "ts": 1716454217607787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607788, "dur": 0,
    "args": {
      "External id": 20624,
      "cbid": 251, "correlation": 20624
    }
  },
  {
    "ph": "f", "id": 20624, "pid": 76337, "tid": -914061504, "ts": 1716454217607788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607789, "dur": 0,
    "args": {
      "External id": 20625,
      "cbid": 251, "correlation": 20625
    }
  },
  {
    "ph": "f", "id": 20625, "pid": 76337, "tid": -914061504, "ts": 1716454217607789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607789, "dur": 0,
    "args": {
      "External id": 20626,
      "cbid": 251, "correlation": 20626
    }
  },
  {
    "ph": "f", "id": 20626, "pid": 76337, "tid": -914061504, "ts": 1716454217607789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607790, "dur": 0,
    "args": {
      "External id": 20627,
      "cbid": 251, "correlation": 20627
    }
  },
  {
    "ph": "f", "id": 20627, "pid": 76337, "tid": -914061504, "ts": 1716454217607790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607791, "dur": 0,
    "args": {
      "External id": 20628,
      "cbid": 251, "correlation": 20628
    }
  },
  {
    "ph": "f", "id": 20628, "pid": 76337, "tid": -914061504, "ts": 1716454217607791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607792, "dur": 0,
    "args": {
      "External id": 20629,
      "cbid": 251, "correlation": 20629
    }
  },
  {
    "ph": "f", "id": 20629, "pid": 76337, "tid": -914061504, "ts": 1716454217607792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217608911, "dur": 112,
    "args": {
      "External id": 20630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20630,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20630, "pid": 5, "tid": 7, "ts": 1716454217608911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607794, "dur": 13,
    "args": {
      "External id": 20630,
      "cbid": 211, "correlation": 20630
    }
  },
  {
    "ph": "s", "id": 20630, "pid": 76337, "tid": -914061504, "ts": 1716454217607794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217609024, "dur": 59,
    "args": {
      "External id": 20636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20636, "pid": 5, "tid": 7, "ts": 1716454217609024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607830, "dur": 9,
    "args": {
      "External id": 20636,
      "cbid": 211, "correlation": 20636
    }
  },
  {
    "ph": "s", "id": 20636, "pid": 76337, "tid": -914061504, "ts": 1716454217607830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217609084, "dur": 619,
    "args": {
      "External id": 20645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20645,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20645, "pid": 5, "tid": 7, "ts": 1716454217609084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217607950, "dur": 18,
    "args": {
      "External id": 20645,
      "cbid": 211, "correlation": 20645
    }
  },
  {
    "ph": "s", "id": 20645, "pid": 76337, "tid": -914061504, "ts": 1716454217607950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217609705, "dur": 177,
    "args": {
      "External id": 20667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20667,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20667, "pid": 5, "tid": 7, "ts": 1716454217609705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608050, "dur": 14,
    "args": {
      "External id": 20667,
      "cbid": 211, "correlation": 20667
    }
  },
  {
    "ph": "s", "id": 20667, "pid": 76337, "tid": -914061504, "ts": 1716454217608050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608178, "dur": 2,
    "args": {
      "External id": 20678,
      "cbid": 251, "correlation": 20678
    }
  },
  {
    "ph": "f", "id": 20678, "pid": 76337, "tid": -914061504, "ts": 1716454217608178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217609883, "dur": 195,
    "args": {
      "External id": 20679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20679,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20679, "pid": 5, "tid": 7, "ts": 1716454217609883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608186, "dur": 16,
    "args": {
      "External id": 20679,
      "cbid": 211, "correlation": 20679
    }
  },
  {
    "ph": "s", "id": 20679, "pid": 76337, "tid": -914061504, "ts": 1716454217608186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608262, "dur": 1,
    "args": {
      "External id": 20690,
      "cbid": 251, "correlation": 20690
    }
  },
  {
    "ph": "f", "id": 20690, "pid": 76337, "tid": -914061504, "ts": 1716454217608262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217610079, "dur": 185,
    "args": {
      "External id": 20691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20691,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20691, "pid": 5, "tid": 7, "ts": 1716454217610079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608267, "dur": 11,
    "args": {
      "External id": 20691,
      "cbid": 211, "correlation": 20691
    }
  },
  {
    "ph": "s", "id": 20691, "pid": 76337, "tid": -914061504, "ts": 1716454217608267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608331, "dur": 1,
    "args": {
      "External id": 20702,
      "cbid": 251, "correlation": 20702
    }
  },
  {
    "ph": "f", "id": 20702, "pid": 76337, "tid": -914061504, "ts": 1716454217608331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217610266, "dur": 183,
    "args": {
      "External id": 20703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20703,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20703, "pid": 5, "tid": 7, "ts": 1716454217610266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608335, "dur": 11,
    "args": {
      "External id": 20703,
      "cbid": 211, "correlation": 20703
    }
  },
  {
    "ph": "s", "id": 20703, "pid": 76337, "tid": -914061504, "ts": 1716454217608335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217610450, "dur": 18204,
    "args": {
      "External id": 20724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20724,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20724, "pid": 5, "tid": 7, "ts": 1716454217610450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608440, "dur": 18,
    "args": {
      "External id": 20724,
      "cbid": 211, "correlation": 20724
    }
  },
  {
    "ph": "s", "id": 20724, "pid": 76337, "tid": -914061504, "ts": 1716454217608440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608559, "dur": 2,
    "args": {
      "External id": 20742,
      "cbid": 251, "correlation": 20742
    }
  },
  {
    "ph": "f", "id": 20742, "pid": 76337, "tid": -914061504, "ts": 1716454217608559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217628656, "dur": 199,
    "args": {
      "External id": 20744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20744,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20744, "pid": 5, "tid": 7, "ts": 1716454217628656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608566, "dur": 13,
    "args": {
      "External id": 20744,
      "cbid": 211, "correlation": 20744
    }
  },
  {
    "ph": "s", "id": 20744, "pid": 76337, "tid": -914061504, "ts": 1716454217608566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217628855, "dur": 66,
    "args": {
      "External id": 20752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20752,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20752, "pid": 5, "tid": 7, "ts": 1716454217628855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608636, "dur": 13,
    "args": {
      "External id": 20752,
      "cbid": 211, "correlation": 20752
    }
  },
  {
    "ph": "s", "id": 20752, "pid": 76337, "tid": -914061504, "ts": 1716454217608636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217628923, "dur": 98,
    "args": {
      "External id": 20760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20760,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20760, "pid": 5, "tid": 7, "ts": 1716454217628923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608676, "dur": 9,
    "args": {
      "External id": 20760,
      "cbid": 211, "correlation": 20760
    }
  },
  {
    "ph": "s", "id": 20760, "pid": 76337, "tid": -914061504, "ts": 1716454217608676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217629022, "dur": 54,
    "args": {
      "External id": 20771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20771, "pid": 5, "tid": 7, "ts": 1716454217629022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608760, "dur": 13,
    "args": {
      "External id": 20771,
      "cbid": 211, "correlation": 20771
    }
  },
  {
    "ph": "s", "id": 20771, "pid": 76337, "tid": -914061504, "ts": 1716454217608760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217629077, "dur": 90,
    "args": {
      "External id": 20793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20793,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20793, "pid": 5, "tid": 7, "ts": 1716454217629077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608792, "dur": 8,
    "args": {
      "External id": 20793,
      "cbid": 211, "correlation": 20793
    }
  },
  {
    "ph": "s", "id": 20793, "pid": 76337, "tid": -914061504, "ts": 1716454217608792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608889, "dur": 1,
    "args": {
      "External id": 20804,
      "cbid": 251, "correlation": 20804
    }
  },
  {
    "ph": "f", "id": 20804, "pid": 76337, "tid": -914061504, "ts": 1716454217608889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217629169, "dur": 103,
    "args": {
      "External id": 20805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20805,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20805, "pid": 5, "tid": 7, "ts": 1716454217629169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608894, "dur": 13,
    "args": {
      "External id": 20805,
      "cbid": 211, "correlation": 20805
    }
  },
  {
    "ph": "s", "id": 20805, "pid": 76337, "tid": -914061504, "ts": 1716454217608894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608986, "dur": 10,
    "args": {
      "External id": 20816,
      "cbid": 251, "correlation": 20816
    }
  },
  {
    "ph": "f", "id": 20816, "pid": 76337, "tid": -914061504, "ts": 1716454217608986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217608999, "dur": 0,
    "args": {
      "External id": 20817,
      "cbid": 251, "correlation": 20817
    }
  },
  {
    "ph": "f", "id": 20817, "pid": 76337, "tid": -914061504, "ts": 1716454217608999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217629273, "dur": 10,
    "args": {
      "External id": 20818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20818,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 20818, "pid": 5, "tid": 7, "ts": 1716454217629273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609002, "dur": 23,
    "args": {
      "External id": 20818,
      "cbid": 211, "correlation": 20818
    }
  },
  {
    "ph": "s", "id": 20818, "pid": 76337, "tid": -914061504, "ts": 1716454217609002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217629284, "dur": 5,
    "args": {
      "External id": 20820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20820,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 20820, "pid": 5, "tid": 7, "ts": 1716454217629284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609028, "dur": 9,
    "args": {
      "External id": 20820,
      "cbid": 211, "correlation": 20820
    }
  },
  {
    "ph": "s", "id": 20820, "pid": 76337, "tid": -914061504, "ts": 1716454217609028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609106, "dur": 1,
    "args": {
      "External id": 20831,
      "cbid": 251, "correlation": 20831
    }
  },
  {
    "ph": "f", "id": 20831, "pid": 76337, "tid": -914061504, "ts": 1716454217609106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609110, "dur": 0,
    "args": {
      "External id": 20832,
      "cbid": 251, "correlation": 20832
    }
  },
  {
    "ph": "f", "id": 20832, "pid": 76337, "tid": -914061504, "ts": 1716454217609110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217629291, "dur": 6,
    "args": {
      "External id": 20833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20833,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 20833, "pid": 5, "tid": 7, "ts": 1716454217629291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609111, "dur": 13,
    "args": {
      "External id": 20833,
      "cbid": 211, "correlation": 20833
    }
  },
  {
    "ph": "s", "id": 20833, "pid": 76337, "tid": -914061504, "ts": 1716454217609111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217629298, "dur": 3,
    "args": {
      "External id": 20835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20835,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 20835, "pid": 5, "tid": 7, "ts": 1716454217629298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609126, "dur": 6,
    "args": {
      "External id": 20835,
      "cbid": 211, "correlation": 20835
    }
  },
  {
    "ph": "s", "id": 20835, "pid": 76337, "tid": -914061504, "ts": 1716454217609126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217629303, "dur": 152,
    "args": {
      "External id": 20856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20856,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20856, "pid": 5, "tid": 7, "ts": 1716454217629303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609202, "dur": 13,
    "args": {
      "External id": 20856,
      "cbid": 211, "correlation": 20856
    }
  },
  {
    "ph": "s", "id": 20856, "pid": 76337, "tid": -914061504, "ts": 1716454217609202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609312, "dur": 2,
    "args": {
      "External id": 20874,
      "cbid": 251, "correlation": 20874
    }
  },
  {
    "ph": "f", "id": 20874, "pid": 76337, "tid": -914061504, "ts": 1716454217609312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217629457, "dur": 105,
    "args": {
      "External id": 20876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20876,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 20876, "pid": 5, "tid": 7, "ts": 1716454217629457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609319, "dur": 15,
    "args": {
      "External id": 20876,
      "cbid": 211, "correlation": 20876
    }
  },
  {
    "ph": "s", "id": 20876, "pid": 76337, "tid": -914061504, "ts": 1716454217609319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217629564, "dur": 34,
    "args": {
      "External id": 20884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20884,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20884, "pid": 5, "tid": 7, "ts": 1716454217629564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609392, "dur": 12,
    "args": {
      "External id": 20884,
      "cbid": 211, "correlation": 20884
    }
  },
  {
    "ph": "s", "id": 20884, "pid": 76337, "tid": -914061504, "ts": 1716454217609392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217629599, "dur": 68,
    "args": {
      "External id": 20892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20892, "pid": 5, "tid": 7, "ts": 1716454217629599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609434, "dur": 10,
    "args": {
      "External id": 20892,
      "cbid": 211, "correlation": 20892
    }
  },
  {
    "ph": "s", "id": 20892, "pid": 76337, "tid": -914061504, "ts": 1716454217609434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217629669, "dur": 90,
    "args": {
      "External id": 20914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20914,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20914, "pid": 5, "tid": 7, "ts": 1716454217629669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609499, "dur": 12,
    "args": {
      "External id": 20914,
      "cbid": 211, "correlation": 20914
    }
  },
  {
    "ph": "s", "id": 20914, "pid": 76337, "tid": -914061504, "ts": 1716454217609499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609598, "dur": 1,
    "args": {
      "External id": 20930,
      "cbid": 251, "correlation": 20930
    }
  },
  {
    "ph": "f", "id": 20930, "pid": 76337, "tid": -914061504, "ts": 1716454217609598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217629761, "dur": 563,
    "args": {
      "External id": 20932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20932,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 20932, "pid": 5, "tid": 7, "ts": 1716454217629761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609604, "dur": 13,
    "args": {
      "External id": 20932,
      "cbid": 211, "correlation": 20932
    }
  },
  {
    "ph": "s", "id": 20932, "pid": 76337, "tid": -914061504, "ts": 1716454217609604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217630325, "dur": 239,
    "args": {
      "External id": 20940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20940,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20940, "pid": 5, "tid": 7, "ts": 1716454217630325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609691, "dur": 14,
    "args": {
      "External id": 20940,
      "cbid": 211, "correlation": 20940
    }
  },
  {
    "ph": "s", "id": 20940, "pid": 76337, "tid": -914061504, "ts": 1716454217609691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217630565, "dur": 250,
    "args": {
      "External id": 20948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20948, "pid": 5, "tid": 7, "ts": 1716454217630565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609729, "dur": 8,
    "args": {
      "External id": 20948,
      "cbid": 211, "correlation": 20948
    }
  },
  {
    "ph": "s", "id": 20948, "pid": 76337, "tid": -914061504, "ts": 1716454217609729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609810, "dur": 2,
    "args": {
      "External id": 20964,
      "cbid": 251, "correlation": 20964
    }
  },
  {
    "ph": "f", "id": 20964, "pid": 76337, "tid": -914061504, "ts": 1716454217609810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609815, "dur": 0,
    "args": {
      "External id": 20966,
      "cbid": 251, "correlation": 20966
    }
  },
  {
    "ph": "f", "id": 20966, "pid": 76337, "tid": -914061504, "ts": 1716454217609815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217630817, "dur": 354,
    "args": {
      "External id": 20967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20967,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 20967, "pid": 5, "tid": 7, "ts": 1716454217630817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609818, "dur": 14,
    "args": {
      "External id": 20967,
      "cbid": 211, "correlation": 20967
    }
  },
  {
    "ph": "s", "id": 20967, "pid": 76337, "tid": -914061504, "ts": 1716454217609818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217631172, "dur": 50,
    "args": {
      "External id": 20975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20975,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20975, "pid": 5, "tid": 7, "ts": 1716454217631172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609861, "dur": 10,
    "args": {
      "External id": 20975,
      "cbid": 211, "correlation": 20975
    }
  },
  {
    "ph": "s", "id": 20975, "pid": 76337, "tid": -914061504, "ts": 1716454217609861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217631223, "dur": 156,
    "args": {
      "External id": 20986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 20986,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 20986, "pid": 5, "tid": 7, "ts": 1716454217631223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217609929, "dur": 13,
    "args": {
      "External id": 20986,
      "cbid": 211, "correlation": 20986
    }
  },
  {
    "ph": "s", "id": 20986, "pid": 76337, "tid": -914061504, "ts": 1716454217609929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610004, "dur": 0,
    "args": {
      "External id": 20998,
      "cbid": 317, "correlation": 20998
    }
  },
  {
    "ph": "f", "id": 20998, "pid": 76337, "tid": -914061504, "ts": 1716454217610004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610005, "dur": 0,
    "args": {
      "External id": 20999,
      "cbid": 203, "correlation": 20999
    }
  },
  {
    "ph": "f", "id": 20999, "pid": 76337, "tid": -914061504, "ts": 1716454217610005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610005, "dur": 0,
    "args": {
      "External id": 21000,
      "cbid": 205, "correlation": 21000
    }
  },
  {
    "ph": "f", "id": 21000, "pid": 76337, "tid": -914061504, "ts": 1716454217610005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610029, "dur": 1,
    "args": {
      "External id": 21004,
      "cbid": 251, "correlation": 21004
    }
  },
  {
    "ph": "f", "id": 21004, "pid": 76337, "tid": -914061504, "ts": 1716454217610029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610031, "dur": 0,
    "args": {
      "External id": 21005,
      "cbid": 251, "correlation": 21005
    }
  },
  {
    "ph": "f", "id": 21005, "pid": 76337, "tid": -914061504, "ts": 1716454217610031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610032, "dur": 0,
    "args": {
      "External id": 21006,
      "cbid": 251, "correlation": 21006
    }
  },
  {
    "ph": "f", "id": 21006, "pid": 76337, "tid": -914061504, "ts": 1716454217610032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610032, "dur": 0,
    "args": {
      "External id": 21007,
      "cbid": 251, "correlation": 21007
    }
  },
  {
    "ph": "f", "id": 21007, "pid": 76337, "tid": -914061504, "ts": 1716454217610032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610033, "dur": 0,
    "args": {
      "External id": 21008,
      "cbid": 251, "correlation": 21008
    }
  },
  {
    "ph": "f", "id": 21008, "pid": 76337, "tid": -914061504, "ts": 1716454217610033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610034, "dur": 0,
    "args": {
      "External id": 21009,
      "cbid": 251, "correlation": 21009
    }
  },
  {
    "ph": "f", "id": 21009, "pid": 76337, "tid": -914061504, "ts": 1716454217610034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610035, "dur": 0,
    "args": {
      "External id": 21010,
      "cbid": 251, "correlation": 21010
    }
  },
  {
    "ph": "f", "id": 21010, "pid": 76337, "tid": -914061504, "ts": 1716454217610035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610035, "dur": 0,
    "args": {
      "External id": 21011,
      "cbid": 251, "correlation": 21011
    }
  },
  {
    "ph": "f", "id": 21011, "pid": 76337, "tid": -914061504, "ts": 1716454217610035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610037, "dur": 0,
    "args": {
      "External id": 21012,
      "cbid": 251, "correlation": 21012
    }
  },
  {
    "ph": "f", "id": 21012, "pid": 76337, "tid": -914061504, "ts": 1716454217610037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217631380, "dur": 114,
    "args": {
      "External id": 21013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21013,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 21013, "pid": 5, "tid": 7, "ts": 1716454217631380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610039, "dur": 13,
    "args": {
      "External id": 21013,
      "cbid": 211, "correlation": 21013
    }
  },
  {
    "ph": "s", "id": 21013, "pid": 76337, "tid": -914061504, "ts": 1716454217610039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217631496, "dur": 59,
    "args": {
      "External id": 21019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21019, "pid": 5, "tid": 7, "ts": 1716454217631496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610074, "dur": 9,
    "args": {
      "External id": 21019,
      "cbid": 211, "correlation": 21019
    }
  },
  {
    "ph": "s", "id": 21019, "pid": 76337, "tid": -914061504, "ts": 1716454217610074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217631556, "dur": 50,
    "args": {
      "External id": 21027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21027,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21027, "pid": 5, "tid": 7, "ts": 1716454217631556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610106, "dur": 9,
    "args": {
      "External id": 21027,
      "cbid": 211, "correlation": 21027
    }
  },
  {
    "ph": "s", "id": 21027, "pid": 76337, "tid": -914061504, "ts": 1716454217610106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610184, "dur": 0,
    "args": {
      "External id": 21037,
      "cbid": 317, "correlation": 21037
    }
  },
  {
    "ph": "f", "id": 21037, "pid": 76337, "tid": -914061504, "ts": 1716454217610184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610184, "dur": 0,
    "args": {
      "External id": 21038,
      "cbid": 203, "correlation": 21038
    }
  },
  {
    "ph": "f", "id": 21038, "pid": 76337, "tid": -914061504, "ts": 1716454217610184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610185, "dur": 0,
    "args": {
      "External id": 21039,
      "cbid": 205, "correlation": 21039
    }
  },
  {
    "ph": "f", "id": 21039, "pid": 76337, "tid": -914061504, "ts": 1716454217610185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217631607, "dur": 40,
    "args": {
      "External id": 21043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21043,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21043, "pid": 5, "tid": 7, "ts": 1716454217631607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610202, "dur": 12,
    "args": {
      "External id": 21043,
      "cbid": 211, "correlation": 21043
    }
  },
  {
    "ph": "s", "id": 21043, "pid": 76337, "tid": -914061504, "ts": 1716454217610202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217631649, "dur": 14,
    "args": {
      "External id": 21045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21045,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21045, "pid": 5, "tid": 7, "ts": 1716454217631649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610217, "dur": 5,
    "args": {
      "External id": 21045,
      "cbid": 211, "correlation": 21045
    }
  },
  {
    "ph": "s", "id": 21045, "pid": 76337, "tid": -914061504, "ts": 1716454217610217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217631665, "dur": 1,
    "args": {
      "External id": 21047,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 21047,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.7868852459016393
    }
  },
  {
    "ph": "f", "id": 21047, "pid": 5, "tid": 7, "ts": 1716454217631665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610236, "dur": 27,
    "args": {
      "External id": 21047,
      "cbid": 51, "correlation": 21047
    }
  },
  {
    "ph": "s", "id": 21047, "pid": 76337, "tid": -914061504, "ts": 1716454217610236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217631669, "dur": 355,
    "args": {
      "External id": 21048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21048,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21048, "pid": 5, "tid": 7, "ts": 1716454217631669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610265, "dur": 11,
    "args": {
      "External id": 21048,
      "cbid": 211, "correlation": 21048
    }
  },
  {
    "ph": "s", "id": 21048, "pid": 76337, "tid": -914061504, "ts": 1716454217610265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632026, "dur": 13,
    "args": {
      "External id": 21050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21050, "pid": 5, "tid": 7, "ts": 1716454217632026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610289, "dur": 8,
    "args": {
      "External id": 21050,
      "cbid": 211, "correlation": 21050
    }
  },
  {
    "ph": "s", "id": 21050, "pid": 76337, "tid": -914061504, "ts": 1716454217610289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217632040, "dur": 14,
    "args": {
      "External id": 21056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21056,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21056, "pid": 5, "tid": 7, "ts": 1716454217632040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610322, "dur": 9,
    "args": {
      "External id": 21056,
      "cbid": 211, "correlation": 21056
    }
  },
  {
    "ph": "s", "id": 21056, "pid": 76337, "tid": -914061504, "ts": 1716454217610322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632055, "dur": 18,
    "args": {
      "External id": 21076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21076,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 21076, "pid": 5, "tid": 7, "ts": 1716454217632055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610435, "dur": 14,
    "args": {
      "External id": 21076,
      "cbid": 211, "correlation": 21076
    }
  },
  {
    "ph": "s", "id": 21076, "pid": 76337, "tid": -914061504, "ts": 1716454217610435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217632075, "dur": 4,
    "args": {
      "External id": 21088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21088,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 21088, "pid": 5, "tid": 7, "ts": 1716454217632075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610468, "dur": 8,
    "args": {
      "External id": 21088,
      "cbid": 211, "correlation": 21088
    }
  },
  {
    "ph": "s", "id": 21088, "pid": 76337, "tid": -914061504, "ts": 1716454217610468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217632080, "dur": 17,
    "args": {
      "External id": 21091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21091, "pid": 5, "tid": 7, "ts": 1716454217632080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610488, "dur": 7,
    "args": {
      "External id": 21091,
      "cbid": 211, "correlation": 21091
    }
  },
  {
    "ph": "s", "id": 21091, "pid": 76337, "tid": -914061504, "ts": 1716454217610488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217632099, "dur": 12,
    "args": {
      "External id": 21100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21100,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21100, "pid": 5, "tid": 7, "ts": 1716454217632099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610530, "dur": 10,
    "args": {
      "External id": 21100,
      "cbid": 211, "correlation": 21100
    }
  },
  {
    "ph": "s", "id": 21100, "pid": 76337, "tid": -914061504, "ts": 1716454217610530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610595, "dur": 0,
    "args": {
      "External id": 21110,
      "cbid": 317, "correlation": 21110
    }
  },
  {
    "ph": "f", "id": 21110, "pid": 76337, "tid": -914061504, "ts": 1716454217610595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610596, "dur": 0,
    "args": {
      "External id": 21111,
      "cbid": 203, "correlation": 21111
    }
  },
  {
    "ph": "f", "id": 21111, "pid": 76337, "tid": -914061504, "ts": 1716454217610596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610597, "dur": 0,
    "args": {
      "External id": 21112,
      "cbid": 205, "correlation": 21112
    }
  },
  {
    "ph": "f", "id": 21112, "pid": 76337, "tid": -914061504, "ts": 1716454217610597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632112, "dur": 11,
    "args": {
      "External id": 21116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21116, "pid": 5, "tid": 7, "ts": 1716454217632112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610612, "dur": 13,
    "args": {
      "External id": 21116,
      "cbid": 211, "correlation": 21116
    }
  },
  {
    "ph": "s", "id": 21116, "pid": 76337, "tid": -914061504, "ts": 1716454217610612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632124, "dur": 24,
    "args": {
      "External id": 21118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21118, "pid": 5, "tid": 7, "ts": 1716454217632124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610627, "dur": 5,
    "args": {
      "External id": 21118,
      "cbid": 211, "correlation": 21118
    }
  },
  {
    "ph": "s", "id": 21118, "pid": 76337, "tid": -914061504, "ts": 1716454217610627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217632150, "dur": 4,
    "args": {
      "External id": 21120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 21120, "pid": 5, "tid": 7, "ts": 1716454217632150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610639, "dur": 6,
    "args": {
      "External id": 21120,
      "cbid": 211, "correlation": 21120
    }
  },
  {
    "ph": "s", "id": 21120, "pid": 76337, "tid": -914061504, "ts": 1716454217610639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610649, "dur": 0,
    "args": {
      "External id": 21121,
      "cbid": 51, "correlation": 21121
    }
  },
  {
    "ph": "s", "id": 21121, "pid": 76337, "tid": -914061504, "ts": 1716454217610649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217632155, "dur": 352,
    "args": {
      "External id": 21122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21122,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21122, "pid": 5, "tid": 7, "ts": 1716454217632155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610650, "dur": 16,
    "args": {
      "External id": 21122,
      "cbid": 211, "correlation": 21122
    }
  },
  {
    "ph": "s", "id": 21122, "pid": 76337, "tid": -914061504, "ts": 1716454217610650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632508, "dur": 20,
    "args": {
      "External id": 21123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21123, "pid": 5, "tid": 7, "ts": 1716454217632508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610670, "dur": 6,
    "args": {
      "External id": 21123,
      "cbid": 211, "correlation": 21123
    }
  },
  {
    "ph": "s", "id": 21123, "pid": 76337, "tid": -914061504, "ts": 1716454217610670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217632529, "dur": 32,
    "args": {
      "External id": 21129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21129, "pid": 5, "tid": 7, "ts": 1716454217632529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610700, "dur": 9,
    "args": {
      "External id": 21129,
      "cbid": 211, "correlation": 21129
    }
  },
  {
    "ph": "s", "id": 21129, "pid": 76337, "tid": -914061504, "ts": 1716454217610700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217632563, "dur": 3,
    "args": {
      "External id": 21137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21137,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 21137, "pid": 5, "tid": 7, "ts": 1716454217632563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610744, "dur": 9,
    "args": {
      "External id": 21137,
      "cbid": 211, "correlation": 21137
    }
  },
  {
    "ph": "s", "id": 21137, "pid": 76337, "tid": -914061504, "ts": 1716454217610744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610812, "dur": 1,
    "args": {
      "External id": 21153,
      "cbid": 251, "correlation": 21153
    }
  },
  {
    "ph": "f", "id": 21153, "pid": 76337, "tid": -914061504, "ts": 1716454217610812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610817, "dur": 0,
    "args": {
      "External id": 21155,
      "cbid": 251, "correlation": 21155
    }
  },
  {
    "ph": "f", "id": 21155, "pid": 76337, "tid": -914061504, "ts": 1716454217610817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217632567, "dur": 12,
    "args": {
      "External id": 21156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21156,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 21156, "pid": 5, "tid": 7, "ts": 1716454217632567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610819, "dur": 11,
    "args": {
      "External id": 21156,
      "cbid": 211, "correlation": 21156
    }
  },
  {
    "ph": "s", "id": 21156, "pid": 76337, "tid": -914061504, "ts": 1716454217610819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217632581, "dur": 5,
    "args": {
      "External id": 21158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21158,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 21158, "pid": 5, "tid": 7, "ts": 1716454217632581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610833, "dur": 6,
    "args": {
      "External id": 21158,
      "cbid": 211, "correlation": 21158
    }
  },
  {
    "ph": "s", "id": 21158, "pid": 76337, "tid": -914061504, "ts": 1716454217610833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217632587, "dur": 29,
    "args": {
      "External id": 21168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21168, "pid": 5, "tid": 7, "ts": 1716454217632587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610893, "dur": 13,
    "args": {
      "External id": 21168,
      "cbid": 211, "correlation": 21168
    }
  },
  {
    "ph": "s", "id": 21168, "pid": 76337, "tid": -914061504, "ts": 1716454217610893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632618, "dur": 31,
    "args": {
      "External id": 21188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21188,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 21188, "pid": 5, "tid": 7, "ts": 1716454217632618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610959, "dur": 11,
    "args": {
      "External id": 21188,
      "cbid": 211, "correlation": 21188
    }
  },
  {
    "ph": "s", "id": 21188, "pid": 76337, "tid": -914061504, "ts": 1716454217610959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217632649, "dur": 4,
    "args": {
      "External id": 21200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21200,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 21200, "pid": 5, "tid": 7, "ts": 1716454217632649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217610988, "dur": 7,
    "args": {
      "External id": 21200,
      "cbid": 211, "correlation": 21200
    }
  },
  {
    "ph": "s", "id": 21200, "pid": 76337, "tid": -914061504, "ts": 1716454217610988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217632655, "dur": 30,
    "args": {
      "External id": 21203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21203, "pid": 5, "tid": 7, "ts": 1716454217632655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611008, "dur": 6,
    "args": {
      "External id": 21203,
      "cbid": 211, "correlation": 21203
    }
  },
  {
    "ph": "s", "id": 21203, "pid": 76337, "tid": -914061504, "ts": 1716454217611008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217632686, "dur": 20,
    "args": {
      "External id": 21212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21212,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21212, "pid": 5, "tid": 7, "ts": 1716454217632686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611049, "dur": 10,
    "args": {
      "External id": 21212,
      "cbid": 211, "correlation": 21212
    }
  },
  {
    "ph": "s", "id": 21212, "pid": 76337, "tid": -914061504, "ts": 1716454217611049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611126, "dur": 0,
    "args": {
      "External id": 21222,
      "cbid": 317, "correlation": 21222
    }
  },
  {
    "ph": "f", "id": 21222, "pid": 76337, "tid": -914061504, "ts": 1716454217611126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611127, "dur": 0,
    "args": {
      "External id": 21223,
      "cbid": 203, "correlation": 21223
    }
  },
  {
    "ph": "f", "id": 21223, "pid": 76337, "tid": -914061504, "ts": 1716454217611127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611128, "dur": 0,
    "args": {
      "External id": 21224,
      "cbid": 205, "correlation": 21224
    }
  },
  {
    "ph": "f", "id": 21224, "pid": 76337, "tid": -914061504, "ts": 1716454217611128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632708, "dur": 22,
    "args": {
      "External id": 21228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21228,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21228, "pid": 5, "tid": 7, "ts": 1716454217632708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611147, "dur": 13,
    "args": {
      "External id": 21228,
      "cbid": 211, "correlation": 21228
    }
  },
  {
    "ph": "s", "id": 21228, "pid": 76337, "tid": -914061504, "ts": 1716454217611147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217632731, "dur": 43,
    "args": {
      "External id": 21230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21230,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21230, "pid": 5, "tid": 7, "ts": 1716454217632731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611162, "dur": 5,
    "args": {
      "External id": 21230,
      "cbid": 211, "correlation": 21230
    }
  },
  {
    "ph": "s", "id": 21230, "pid": 76337, "tid": -914061504, "ts": 1716454217611162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217632775, "dur": 643,
    "args": {
      "External id": 21232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21232,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21232, "pid": 5, "tid": 7, "ts": 1716454217632775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611177, "dur": 23,
    "args": {
      "External id": 21232,
      "cbid": 211, "correlation": 21232
    }
  },
  {
    "ph": "s", "id": 21232, "pid": 76337, "tid": -914061504, "ts": 1716454217611177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217633420, "dur": 22,
    "args": {
      "External id": 21234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21234, "pid": 5, "tid": 7, "ts": 1716454217633420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611205, "dur": 5,
    "args": {
      "External id": 21234,
      "cbid": 211, "correlation": 21234
    }
  },
  {
    "ph": "s", "id": 21234, "pid": 76337, "tid": -914061504, "ts": 1716454217611205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217633443, "dur": 32,
    "args": {
      "External id": 21240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21240,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21240, "pid": 5, "tid": 7, "ts": 1716454217633443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611234, "dur": 9,
    "args": {
      "External id": 21240,
      "cbid": 211, "correlation": 21240
    }
  },
  {
    "ph": "s", "id": 21240, "pid": 76337, "tid": -914061504, "ts": 1716454217611234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611295, "dur": 0,
    "args": {
      "External id": 21250,
      "cbid": 317, "correlation": 21250
    }
  },
  {
    "ph": "f", "id": 21250, "pid": 76337, "tid": -914061504, "ts": 1716454217611295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611296, "dur": 0,
    "args": {
      "External id": 21251,
      "cbid": 203, "correlation": 21251
    }
  },
  {
    "ph": "f", "id": 21251, "pid": 76337, "tid": -914061504, "ts": 1716454217611296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611297, "dur": 0,
    "args": {
      "External id": 21252,
      "cbid": 205, "correlation": 21252
    }
  },
  {
    "ph": "f", "id": 21252, "pid": 76337, "tid": -914061504, "ts": 1716454217611297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611320, "dur": 1,
    "args": {
      "External id": 21256,
      "cbid": 251, "correlation": 21256
    }
  },
  {
    "ph": "f", "id": 21256, "pid": 76337, "tid": -914061504, "ts": 1716454217611320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611322, "dur": 0,
    "args": {
      "External id": 21257,
      "cbid": 251, "correlation": 21257
    }
  },
  {
    "ph": "f", "id": 21257, "pid": 76337, "tid": -914061504, "ts": 1716454217611322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611322, "dur": 0,
    "args": {
      "External id": 21258,
      "cbid": 251, "correlation": 21258
    }
  },
  {
    "ph": "f", "id": 21258, "pid": 76337, "tid": -914061504, "ts": 1716454217611322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611323, "dur": 0,
    "args": {
      "External id": 21259,
      "cbid": 251, "correlation": 21259
    }
  },
  {
    "ph": "f", "id": 21259, "pid": 76337, "tid": -914061504, "ts": 1716454217611323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611324, "dur": 0,
    "args": {
      "External id": 21260,
      "cbid": 251, "correlation": 21260
    }
  },
  {
    "ph": "f", "id": 21260, "pid": 76337, "tid": -914061504, "ts": 1716454217611324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611324, "dur": 0,
    "args": {
      "External id": 21261,
      "cbid": 251, "correlation": 21261
    }
  },
  {
    "ph": "f", "id": 21261, "pid": 76337, "tid": -914061504, "ts": 1716454217611324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611325, "dur": 0,
    "args": {
      "External id": 21262,
      "cbid": 251, "correlation": 21262
    }
  },
  {
    "ph": "f", "id": 21262, "pid": 76337, "tid": -914061504, "ts": 1716454217611325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611326, "dur": 0,
    "args": {
      "External id": 21263,
      "cbid": 251, "correlation": 21263
    }
  },
  {
    "ph": "f", "id": 21263, "pid": 76337, "tid": -914061504, "ts": 1716454217611326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611327, "dur": 0,
    "args": {
      "External id": 21264,
      "cbid": 251, "correlation": 21264
    }
  },
  {
    "ph": "f", "id": 21264, "pid": 76337, "tid": -914061504, "ts": 1716454217611327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217633477, "dur": 51,
    "args": {
      "External id": 21265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21265,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 21265, "pid": 5, "tid": 7, "ts": 1716454217633477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611329, "dur": 12,
    "args": {
      "External id": 21265,
      "cbid": 211, "correlation": 21265
    }
  },
  {
    "ph": "s", "id": 21265, "pid": 76337, "tid": -914061504, "ts": 1716454217611329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217633529, "dur": 32,
    "args": {
      "External id": 21271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21271, "pid": 5, "tid": 7, "ts": 1716454217633529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611362, "dur": 9,
    "args": {
      "External id": 21271,
      "cbid": 211, "correlation": 21271
    }
  },
  {
    "ph": "s", "id": 21271, "pid": 76337, "tid": -914061504, "ts": 1716454217611362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217633562, "dur": 27,
    "args": {
      "External id": 21279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21279,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21279, "pid": 5, "tid": 7, "ts": 1716454217633562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611391, "dur": 8,
    "args": {
      "External id": 21279,
      "cbid": 211, "correlation": 21279
    }
  },
  {
    "ph": "s", "id": 21279, "pid": 76337, "tid": -914061504, "ts": 1716454217611391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217633591, "dur": 20,
    "args": {
      "External id": 21287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21287,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21287, "pid": 5, "tid": 7, "ts": 1716454217633591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611421, "dur": 9,
    "args": {
      "External id": 21287,
      "cbid": 211, "correlation": 21287
    }
  },
  {
    "ph": "s", "id": 21287, "pid": 76337, "tid": -914061504, "ts": 1716454217611421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217633612, "dur": 29,
    "args": {
      "External id": 21307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21307,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 21307, "pid": 5, "tid": 7, "ts": 1716454217633612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611503, "dur": 12,
    "args": {
      "External id": 21307,
      "cbid": 211, "correlation": 21307
    }
  },
  {
    "ph": "s", "id": 21307, "pid": 76337, "tid": -914061504, "ts": 1716454217611503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217633642, "dur": 4,
    "args": {
      "External id": 21319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21319,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 21319, "pid": 5, "tid": 7, "ts": 1716454217633642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611524, "dur": 6,
    "args": {
      "External id": 21319,
      "cbid": 211, "correlation": 21319
    }
  },
  {
    "ph": "s", "id": 21319, "pid": 76337, "tid": -914061504, "ts": 1716454217611524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217633648, "dur": 29,
    "args": {
      "External id": 21322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21322,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21322, "pid": 5, "tid": 7, "ts": 1716454217633648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611543, "dur": 6,
    "args": {
      "External id": 21322,
      "cbid": 211, "correlation": 21322
    }
  },
  {
    "ph": "s", "id": 21322, "pid": 76337, "tid": -914061504, "ts": 1716454217611543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611611, "dur": 0,
    "args": {
      "External id": 21333,
      "cbid": 317, "correlation": 21333
    }
  },
  {
    "ph": "f", "id": 21333, "pid": 76337, "tid": -914061504, "ts": 1716454217611611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611612, "dur": 0,
    "args": {
      "External id": 21334,
      "cbid": 203, "correlation": 21334
    }
  },
  {
    "ph": "f", "id": 21334, "pid": 76337, "tid": -914061504, "ts": 1716454217611612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611613, "dur": 0,
    "args": {
      "External id": 21335,
      "cbid": 205, "correlation": 21335
    }
  },
  {
    "ph": "f", "id": 21335, "pid": 76337, "tid": -914061504, "ts": 1716454217611613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217633678, "dur": 21,
    "args": {
      "External id": 21339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21339,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21339, "pid": 5, "tid": 7, "ts": 1716454217633678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611635, "dur": 13,
    "args": {
      "External id": 21339,
      "cbid": 211, "correlation": 21339
    }
  },
  {
    "ph": "s", "id": 21339, "pid": 76337, "tid": -914061504, "ts": 1716454217611635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217633701, "dur": 118,
    "args": {
      "External id": 21341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21341,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21341, "pid": 5, "tid": 7, "ts": 1716454217633701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611662, "dur": 26,
    "args": {
      "External id": 21341,
      "cbid": 211, "correlation": 21341
    }
  },
  {
    "ph": "s", "id": 21341, "pid": 76337, "tid": -914061504, "ts": 1716454217611662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217633820, "dur": 21,
    "args": {
      "External id": 21343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21343, "pid": 5, "tid": 7, "ts": 1716454217633820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611692, "dur": 6,
    "args": {
      "External id": 21343,
      "cbid": 211, "correlation": 21343
    }
  },
  {
    "ph": "s", "id": 21343, "pid": 76337, "tid": -914061504, "ts": 1716454217611692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217633842, "dur": 32,
    "args": {
      "External id": 21349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21349, "pid": 5, "tid": 7, "ts": 1716454217633842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611723, "dur": 8,
    "args": {
      "External id": 21349,
      "cbid": 211, "correlation": 21349
    }
  },
  {
    "ph": "s", "id": 21349, "pid": 76337, "tid": -914061504, "ts": 1716454217611723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217633876, "dur": 177,
    "args": {
      "External id": 21358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21358,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21358, "pid": 5, "tid": 7, "ts": 1716454217633876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611809, "dur": 15,
    "args": {
      "External id": 21358,
      "cbid": 211, "correlation": 21358
    }
  },
  {
    "ph": "s", "id": 21358, "pid": 76337, "tid": -914061504, "ts": 1716454217611809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217634054, "dur": 64,
    "args": {
      "External id": 21380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21380,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21380, "pid": 5, "tid": 7, "ts": 1716454217634054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611868, "dur": 10,
    "args": {
      "External id": 21380,
      "cbid": 211, "correlation": 21380
    }
  },
  {
    "ph": "s", "id": 21380, "pid": 76337, "tid": -914061504, "ts": 1716454217611868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611968, "dur": 1,
    "args": {
      "External id": 21391,
      "cbid": 251, "correlation": 21391
    }
  },
  {
    "ph": "f", "id": 21391, "pid": 76337, "tid": -914061504, "ts": 1716454217611968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217634119, "dur": 152,
    "args": {
      "External id": 21392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21392,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21392, "pid": 5, "tid": 7, "ts": 1716454217634119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217611982, "dur": 15,
    "args": {
      "External id": 21392,
      "cbid": 211, "correlation": 21392
    }
  },
  {
    "ph": "s", "id": 21392, "pid": 76337, "tid": -914061504, "ts": 1716454217611982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612058, "dur": 1,
    "args": {
      "External id": 21403,
      "cbid": 251, "correlation": 21403
    }
  },
  {
    "ph": "f", "id": 21403, "pid": 76337, "tid": -914061504, "ts": 1716454217612058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217634272, "dur": 143,
    "args": {
      "External id": 21404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21404,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21404, "pid": 5, "tid": 7, "ts": 1716454217634272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612063, "dur": 11,
    "args": {
      "External id": 21404,
      "cbid": 211, "correlation": 21404
    }
  },
  {
    "ph": "s", "id": 21404, "pid": 76337, "tid": -914061504, "ts": 1716454217612063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612126, "dur": 1,
    "args": {
      "External id": 21415,
      "cbid": 251, "correlation": 21415
    }
  },
  {
    "ph": "f", "id": 21415, "pid": 76337, "tid": -914061504, "ts": 1716454217612126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217634417, "dur": 147,
    "args": {
      "External id": 21416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21416,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21416, "pid": 5, "tid": 7, "ts": 1716454217634417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612130, "dur": 11,
    "args": {
      "External id": 21416,
      "cbid": 211, "correlation": 21416
    }
  },
  {
    "ph": "s", "id": 21416, "pid": 76337, "tid": -914061504, "ts": 1716454217612130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217634565, "dur": 1909,
    "args": {
      "External id": 21437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21437,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 21437, "pid": 5, "tid": 7, "ts": 1716454217634565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612213, "dur": 14,
    "args": {
      "External id": 21437,
      "cbid": 211, "correlation": 21437
    }
  },
  {
    "ph": "s", "id": 21437, "pid": 76337, "tid": -914061504, "ts": 1716454217612213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612315, "dur": 1,
    "args": {
      "External id": 21455,
      "cbid": 251, "correlation": 21455
    }
  },
  {
    "ph": "f", "id": 21455, "pid": 76337, "tid": -914061504, "ts": 1716454217612315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217636475, "dur": 146,
    "args": {
      "External id": 21457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21457,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 21457, "pid": 5, "tid": 7, "ts": 1716454217636475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612320, "dur": 13,
    "args": {
      "External id": 21457,
      "cbid": 211, "correlation": 21457
    }
  },
  {
    "ph": "s", "id": 21457, "pid": 76337, "tid": -914061504, "ts": 1716454217612320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217636622, "dur": 35,
    "args": {
      "External id": 21465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21465,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21465, "pid": 5, "tid": 7, "ts": 1716454217636622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612393, "dur": 12,
    "args": {
      "External id": 21465,
      "cbid": 211, "correlation": 21465
    }
  },
  {
    "ph": "s", "id": 21465, "pid": 76337, "tid": -914061504, "ts": 1716454217612393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217636659, "dur": 51,
    "args": {
      "External id": 21473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21473,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21473, "pid": 5, "tid": 7, "ts": 1716454217636659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612432, "dur": 9,
    "args": {
      "External id": 21473,
      "cbid": 211, "correlation": 21473
    }
  },
  {
    "ph": "s", "id": 21473, "pid": 76337, "tid": -914061504, "ts": 1716454217612432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217636711, "dur": 30,
    "args": {
      "External id": 21484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21484, "pid": 5, "tid": 7, "ts": 1716454217636711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612506, "dur": 13,
    "args": {
      "External id": 21484,
      "cbid": 211, "correlation": 21484
    }
  },
  {
    "ph": "s", "id": 21484, "pid": 76337, "tid": -914061504, "ts": 1716454217612506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217636742, "dur": 34,
    "args": {
      "External id": 21506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21506,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21506, "pid": 5, "tid": 7, "ts": 1716454217636742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612538, "dur": 8,
    "args": {
      "External id": 21506,
      "cbid": 211, "correlation": 21506
    }
  },
  {
    "ph": "s", "id": 21506, "pid": 76337, "tid": -914061504, "ts": 1716454217612538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612628, "dur": 1,
    "args": {
      "External id": 21517,
      "cbid": 251, "correlation": 21517
    }
  },
  {
    "ph": "f", "id": 21517, "pid": 76337, "tid": -914061504, "ts": 1716454217612628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217636777, "dur": 88,
    "args": {
      "External id": 21518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21518,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21518, "pid": 5, "tid": 7, "ts": 1716454217636777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612633, "dur": 13,
    "args": {
      "External id": 21518,
      "cbid": 211, "correlation": 21518
    }
  },
  {
    "ph": "s", "id": 21518, "pid": 76337, "tid": -914061504, "ts": 1716454217612633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612702, "dur": 1,
    "args": {
      "External id": 21529,
      "cbid": 251, "correlation": 21529
    }
  },
  {
    "ph": "f", "id": 21529, "pid": 76337, "tid": -914061504, "ts": 1716454217612702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612706, "dur": 0,
    "args": {
      "External id": 21530,
      "cbid": 251, "correlation": 21530
    }
  },
  {
    "ph": "f", "id": 21530, "pid": 76337, "tid": -914061504, "ts": 1716454217612706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217636865, "dur": 11,
    "args": {
      "External id": 21531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21531,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 21531, "pid": 5, "tid": 7, "ts": 1716454217636865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612707, "dur": 12,
    "args": {
      "External id": 21531,
      "cbid": 211, "correlation": 21531
    }
  },
  {
    "ph": "s", "id": 21531, "pid": 76337, "tid": -914061504, "ts": 1716454217612707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217636878, "dur": 5,
    "args": {
      "External id": 21533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21533,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 21533, "pid": 5, "tid": 7, "ts": 1716454217636878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612722, "dur": 7,
    "args": {
      "External id": 21533,
      "cbid": 211, "correlation": 21533
    }
  },
  {
    "ph": "s", "id": 21533, "pid": 76337, "tid": -914061504, "ts": 1716454217612722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612781, "dur": 1,
    "args": {
      "External id": 21544,
      "cbid": 251, "correlation": 21544
    }
  },
  {
    "ph": "f", "id": 21544, "pid": 76337, "tid": -914061504, "ts": 1716454217612781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612784, "dur": 0,
    "args": {
      "External id": 21545,
      "cbid": 251, "correlation": 21545
    }
  },
  {
    "ph": "f", "id": 21545, "pid": 76337, "tid": -914061504, "ts": 1716454217612784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217636884, "dur": 7,
    "args": {
      "External id": 21546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21546,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 21546, "pid": 5, "tid": 7, "ts": 1716454217636884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612786, "dur": 11,
    "args": {
      "External id": 21546,
      "cbid": 211, "correlation": 21546
    }
  },
  {
    "ph": "s", "id": 21546, "pid": 76337, "tid": -914061504, "ts": 1716454217612786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217636892, "dur": 4,
    "args": {
      "External id": 21548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21548,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 21548, "pid": 5, "tid": 7, "ts": 1716454217636892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612799, "dur": 5,
    "args": {
      "External id": 21548,
      "cbid": 211, "correlation": 21548
    }
  },
  {
    "ph": "s", "id": 21548, "pid": 76337, "tid": -914061504, "ts": 1716454217612799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217636897, "dur": 90,
    "args": {
      "External id": 21569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21569,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 21569, "pid": 5, "tid": 7, "ts": 1716454217636897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612873, "dur": 12,
    "args": {
      "External id": 21569,
      "cbid": 211, "correlation": 21569
    }
  },
  {
    "ph": "s", "id": 21569, "pid": 76337, "tid": -914061504, "ts": 1716454217612873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217612994, "dur": 1,
    "args": {
      "External id": 21587,
      "cbid": 251, "correlation": 21587
    }
  },
  {
    "ph": "f", "id": 21587, "pid": 76337, "tid": -914061504, "ts": 1716454217612994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217636989, "dur": 97,
    "args": {
      "External id": 21589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21589,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21589, "pid": 5, "tid": 7, "ts": 1716454217636989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613000, "dur": 14,
    "args": {
      "External id": 21589,
      "cbid": 211, "correlation": 21589
    }
  },
  {
    "ph": "s", "id": 21589, "pid": 76337, "tid": -914061504, "ts": 1716454217613000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217637087, "dur": 19,
    "args": {
      "External id": 21597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21597,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21597, "pid": 5, "tid": 7, "ts": 1716454217637087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613074, "dur": 13,
    "args": {
      "External id": 21597,
      "cbid": 211, "correlation": 21597
    }
  },
  {
    "ph": "s", "id": 21597, "pid": 76337, "tid": -914061504, "ts": 1716454217613074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217637107, "dur": 38,
    "args": {
      "External id": 21605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21605,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21605, "pid": 5, "tid": 7, "ts": 1716454217637107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613119, "dur": 9,
    "args": {
      "External id": 21605,
      "cbid": 211, "correlation": 21605
    }
  },
  {
    "ph": "s", "id": 21605, "pid": 76337, "tid": -914061504, "ts": 1716454217613119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217637147, "dur": 34,
    "args": {
      "External id": 21627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21627,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21627, "pid": 5, "tid": 7, "ts": 1716454217637147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613171, "dur": 10,
    "args": {
      "External id": 21627,
      "cbid": 211, "correlation": 21627
    }
  },
  {
    "ph": "s", "id": 21627, "pid": 76337, "tid": -914061504, "ts": 1716454217613171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613274, "dur": 1,
    "args": {
      "External id": 21643,
      "cbid": 251, "correlation": 21643
    }
  },
  {
    "ph": "f", "id": 21643, "pid": 76337, "tid": -914061504, "ts": 1716454217613274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613279, "dur": 0,
    "args": {
      "External id": 21645,
      "cbid": 251, "correlation": 21645
    }
  },
  {
    "ph": "f", "id": 21645, "pid": 76337, "tid": -914061504, "ts": 1716454217613279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217637182, "dur": 531,
    "args": {
      "External id": 21646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21646,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 21646, "pid": 5, "tid": 7, "ts": 1716454217637182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613282, "dur": 13,
    "args": {
      "External id": 21646,
      "cbid": 211, "correlation": 21646
    }
  },
  {
    "ph": "s", "id": 21646, "pid": 76337, "tid": -914061504, "ts": 1716454217613282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217637715, "dur": 124,
    "args": {
      "External id": 21654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21654,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21654, "pid": 5, "tid": 7, "ts": 1716454217637715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613352, "dur": 13,
    "args": {
      "External id": 21654,
      "cbid": 211, "correlation": 21654
    }
  },
  {
    "ph": "s", "id": 21654, "pid": 76337, "tid": -914061504, "ts": 1716454217613352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217637841, "dur": 127,
    "args": {
      "External id": 21662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21662, "pid": 5, "tid": 7, "ts": 1716454217637841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613383, "dur": 9,
    "args": {
      "External id": 21662,
      "cbid": 211, "correlation": 21662
    }
  },
  {
    "ph": "s", "id": 21662, "pid": 76337, "tid": -914061504, "ts": 1716454217613383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613464, "dur": 1,
    "args": {
      "External id": 21678,
      "cbid": 251, "correlation": 21678
    }
  },
  {
    "ph": "f", "id": 21678, "pid": 76337, "tid": -914061504, "ts": 1716454217613464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217637969, "dur": 295,
    "args": {
      "External id": 21680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21680,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21680, "pid": 5, "tid": 7, "ts": 1716454217637969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613469, "dur": 12,
    "args": {
      "External id": 21680,
      "cbid": 211, "correlation": 21680
    }
  },
  {
    "ph": "s", "id": 21680, "pid": 76337, "tid": -914061504, "ts": 1716454217613469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217638266, "dur": 27,
    "args": {
      "External id": 21688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21688,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21688, "pid": 5, "tid": 7, "ts": 1716454217638266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613512, "dur": 9,
    "args": {
      "External id": 21688,
      "cbid": 211, "correlation": 21688
    }
  },
  {
    "ph": "s", "id": 21688, "pid": 76337, "tid": -914061504, "ts": 1716454217613512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217638294, "dur": 80,
    "args": {
      "External id": 21699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21699, "pid": 5, "tid": 7, "ts": 1716454217638294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613580, "dur": 12,
    "args": {
      "External id": 21699,
      "cbid": 211, "correlation": 21699
    }
  },
  {
    "ph": "s", "id": 21699, "pid": 76337, "tid": -914061504, "ts": 1716454217613580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613656, "dur": 0,
    "args": {
      "External id": 21711,
      "cbid": 317, "correlation": 21711
    }
  },
  {
    "ph": "f", "id": 21711, "pid": 76337, "tid": -914061504, "ts": 1716454217613656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613657, "dur": 0,
    "args": {
      "External id": 21712,
      "cbid": 203, "correlation": 21712
    }
  },
  {
    "ph": "f", "id": 21712, "pid": 76337, "tid": -914061504, "ts": 1716454217613657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613658, "dur": 0,
    "args": {
      "External id": 21713,
      "cbid": 205, "correlation": 21713
    }
  },
  {
    "ph": "f", "id": 21713, "pid": 76337, "tid": -914061504, "ts": 1716454217613658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217638375, "dur": 23,
    "args": {
      "External id": 21717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21717,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21717, "pid": 5, "tid": 7, "ts": 1716454217638375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613682, "dur": 13,
    "args": {
      "External id": 21717,
      "cbid": 211, "correlation": 21717
    }
  },
  {
    "ph": "s", "id": 21717, "pid": 76337, "tid": -914061504, "ts": 1716454217613682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217638399, "dur": 119,
    "args": {
      "External id": 21719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21719,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21719, "pid": 5, "tid": 7, "ts": 1716454217638399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613701, "dur": 7,
    "args": {
      "External id": 21719,
      "cbid": 211, "correlation": 21719
    }
  },
  {
    "ph": "s", "id": 21719, "pid": 76337, "tid": -914061504, "ts": 1716454217613701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217638519, "dur": 22,
    "args": {
      "External id": 21721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21721,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21721, "pid": 5, "tid": 7, "ts": 1716454217638519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613712, "dur": 5,
    "args": {
      "External id": 21721,
      "cbid": 211, "correlation": 21721
    }
  },
  {
    "ph": "s", "id": 21721, "pid": 76337, "tid": -914061504, "ts": 1716454217613712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217638543, "dur": 32,
    "args": {
      "External id": 21727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21727, "pid": 5, "tid": 7, "ts": 1716454217638543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613741, "dur": 8,
    "args": {
      "External id": 21727,
      "cbid": 211, "correlation": 21727
    }
  },
  {
    "ph": "s", "id": 21727, "pid": 76337, "tid": -914061504, "ts": 1716454217613741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217638576, "dur": 27,
    "args": {
      "External id": 21735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21735,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21735, "pid": 5, "tid": 7, "ts": 1716454217638576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613773, "dur": 8,
    "args": {
      "External id": 21735,
      "cbid": 211, "correlation": 21735
    }
  },
  {
    "ph": "s", "id": 21735, "pid": 76337, "tid": -914061504, "ts": 1716454217613773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217638604, "dur": 31,
    "args": {
      "External id": 21755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21755,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 21755, "pid": 5, "tid": 7, "ts": 1716454217638604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613846, "dur": 12,
    "args": {
      "External id": 21755,
      "cbid": 211, "correlation": 21755
    }
  },
  {
    "ph": "s", "id": 21755, "pid": 76337, "tid": -914061504, "ts": 1716454217613846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217638636, "dur": 5,
    "args": {
      "External id": 21767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21767,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 21767, "pid": 5, "tid": 7, "ts": 1716454217638636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613869, "dur": 6,
    "args": {
      "External id": 21767,
      "cbid": 211, "correlation": 21767
    }
  },
  {
    "ph": "s", "id": 21767, "pid": 76337, "tid": -914061504, "ts": 1716454217613869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217638642, "dur": 31,
    "args": {
      "External id": 21770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21770, "pid": 5, "tid": 7, "ts": 1716454217638642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613886, "dur": 6,
    "args": {
      "External id": 21770,
      "cbid": 211, "correlation": 21770
    }
  },
  {
    "ph": "s", "id": 21770, "pid": 76337, "tid": -914061504, "ts": 1716454217613886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217638675, "dur": 21,
    "args": {
      "External id": 21779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21779,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21779, "pid": 5, "tid": 7, "ts": 1716454217638675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613926, "dur": 9,
    "args": {
      "External id": 21779,
      "cbid": 211, "correlation": 21779
    }
  },
  {
    "ph": "s", "id": 21779, "pid": 76337, "tid": -914061504, "ts": 1716454217613926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613985, "dur": 0,
    "args": {
      "External id": 21789,
      "cbid": 317, "correlation": 21789
    }
  },
  {
    "ph": "f", "id": 21789, "pid": 76337, "tid": -914061504, "ts": 1716454217613985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613986, "dur": 0,
    "args": {
      "External id": 21790,
      "cbid": 203, "correlation": 21790
    }
  },
  {
    "ph": "f", "id": 21790, "pid": 76337, "tid": -914061504, "ts": 1716454217613986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217613987, "dur": 0,
    "args": {
      "External id": 21791,
      "cbid": 205, "correlation": 21791
    }
  },
  {
    "ph": "f", "id": 21791, "pid": 76337, "tid": -914061504, "ts": 1716454217613987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217638698, "dur": 21,
    "args": {
      "External id": 21795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21795,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21795, "pid": 5, "tid": 7, "ts": 1716454217638698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614000, "dur": 11,
    "args": {
      "External id": 21795,
      "cbid": 211, "correlation": 21795
    }
  },
  {
    "ph": "s", "id": 21795, "pid": 76337, "tid": -914061504, "ts": 1716454217614000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217638720, "dur": 44,
    "args": {
      "External id": 21797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21797,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21797, "pid": 5, "tid": 7, "ts": 1716454217638720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614014, "dur": 5,
    "args": {
      "External id": 21797,
      "cbid": 211, "correlation": 21797
    }
  },
  {
    "ph": "s", "id": 21797, "pid": 76337, "tid": -914061504, "ts": 1716454217614014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217638765, "dur": 644,
    "args": {
      "External id": 21799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21799,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21799, "pid": 5, "tid": 7, "ts": 1716454217638765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614026, "dur": 7,
    "args": {
      "External id": 21799,
      "cbid": 211, "correlation": 21799
    }
  },
  {
    "ph": "s", "id": 21799, "pid": 76337, "tid": -914061504, "ts": 1716454217614026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217639410, "dur": 21,
    "args": {
      "External id": 21801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21801,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21801, "pid": 5, "tid": 7, "ts": 1716454217639410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614037, "dur": 5,
    "args": {
      "External id": 21801,
      "cbid": 211, "correlation": 21801
    }
  },
  {
    "ph": "s", "id": 21801, "pid": 76337, "tid": -914061504, "ts": 1716454217614037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217639433, "dur": 33,
    "args": {
      "External id": 21807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21807,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21807, "pid": 5, "tid": 7, "ts": 1716454217639433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614064, "dur": 8,
    "args": {
      "External id": 21807,
      "cbid": 211, "correlation": 21807
    }
  },
  {
    "ph": "s", "id": 21807, "pid": 76337, "tid": -914061504, "ts": 1716454217614064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217639467, "dur": 3,
    "args": {
      "External id": 21815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21815,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 21815, "pid": 5, "tid": 7, "ts": 1716454217639467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614108, "dur": 9,
    "args": {
      "External id": 21815,
      "cbid": 211, "correlation": 21815
    }
  },
  {
    "ph": "s", "id": 21815, "pid": 76337, "tid": -914061504, "ts": 1716454217614108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614172, "dur": 1,
    "args": {
      "External id": 21831,
      "cbid": 251, "correlation": 21831
    }
  },
  {
    "ph": "f", "id": 21831, "pid": 76337, "tid": -914061504, "ts": 1716454217614172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614177, "dur": 0,
    "args": {
      "External id": 21833,
      "cbid": 251, "correlation": 21833
    }
  },
  {
    "ph": "f", "id": 21833, "pid": 76337, "tid": -914061504, "ts": 1716454217614177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217639471, "dur": 12,
    "args": {
      "External id": 21834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21834,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 21834, "pid": 5, "tid": 7, "ts": 1716454217639471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614179, "dur": 11,
    "args": {
      "External id": 21834,
      "cbid": 211, "correlation": 21834
    }
  },
  {
    "ph": "s", "id": 21834, "pid": 76337, "tid": -914061504, "ts": 1716454217614179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217639485, "dur": 5,
    "args": {
      "External id": 21836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21836,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 21836, "pid": 5, "tid": 7, "ts": 1716454217639485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614192, "dur": 5,
    "args": {
      "External id": 21836,
      "cbid": 211, "correlation": 21836
    }
  },
  {
    "ph": "s", "id": 21836, "pid": 76337, "tid": -914061504, "ts": 1716454217614192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217639491, "dur": 28,
    "args": {
      "External id": 21846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21846, "pid": 5, "tid": 7, "ts": 1716454217639491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614251, "dur": 12,
    "args": {
      "External id": 21846,
      "cbid": 211, "correlation": 21846
    }
  },
  {
    "ph": "s", "id": 21846, "pid": 76337, "tid": -914061504, "ts": 1716454217614251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217639520, "dur": 31,
    "args": {
      "External id": 21866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21866,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 21866, "pid": 5, "tid": 7, "ts": 1716454217639520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614316, "dur": 11,
    "args": {
      "External id": 21866,
      "cbid": 211, "correlation": 21866
    }
  },
  {
    "ph": "s", "id": 21866, "pid": 76337, "tid": -914061504, "ts": 1716454217614316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217639553, "dur": 5,
    "args": {
      "External id": 21878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21878,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 21878, "pid": 5, "tid": 7, "ts": 1716454217639553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614337, "dur": 6,
    "args": {
      "External id": 21878,
      "cbid": 211, "correlation": 21878
    }
  },
  {
    "ph": "s", "id": 21878, "pid": 76337, "tid": -914061504, "ts": 1716454217614337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217639559, "dur": 30,
    "args": {
      "External id": 21881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21881, "pid": 5, "tid": 7, "ts": 1716454217639559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614355, "dur": 7,
    "args": {
      "External id": 21881,
      "cbid": 211, "correlation": 21881
    }
  },
  {
    "ph": "s", "id": 21881, "pid": 76337, "tid": -914061504, "ts": 1716454217614355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217639589, "dur": 20,
    "args": {
      "External id": 21890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21890,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21890, "pid": 5, "tid": 7, "ts": 1716454217639589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614395, "dur": 10,
    "args": {
      "External id": 21890,
      "cbid": 211, "correlation": 21890
    }
  },
  {
    "ph": "s", "id": 21890, "pid": 76337, "tid": -914061504, "ts": 1716454217614395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614457, "dur": 0,
    "args": {
      "External id": 21900,
      "cbid": 317, "correlation": 21900
    }
  },
  {
    "ph": "f", "id": 21900, "pid": 76337, "tid": -914061504, "ts": 1716454217614457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614458, "dur": 0,
    "args": {
      "External id": 21901,
      "cbid": 203, "correlation": 21901
    }
  },
  {
    "ph": "f", "id": 21901, "pid": 76337, "tid": -914061504, "ts": 1716454217614458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614459, "dur": 0,
    "args": {
      "External id": 21902,
      "cbid": 205, "correlation": 21902
    }
  },
  {
    "ph": "f", "id": 21902, "pid": 76337, "tid": -914061504, "ts": 1716454217614459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217639611, "dur": 21,
    "args": {
      "External id": 21906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21906,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21906, "pid": 5, "tid": 7, "ts": 1716454217639611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614487, "dur": 13,
    "args": {
      "External id": 21906,
      "cbid": 211, "correlation": 21906
    }
  },
  {
    "ph": "s", "id": 21906, "pid": 76337, "tid": -914061504, "ts": 1716454217614487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217639633, "dur": 44,
    "args": {
      "External id": 21908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21908,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21908, "pid": 5, "tid": 7, "ts": 1716454217639633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614502, "dur": 5,
    "args": {
      "External id": 21908,
      "cbid": 211, "correlation": 21908
    }
  },
  {
    "ph": "s", "id": 21908, "pid": 76337, "tid": -914061504, "ts": 1716454217614502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217639678, "dur": 637,
    "args": {
      "External id": 21910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21910,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21910, "pid": 5, "tid": 7, "ts": 1716454217639678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614513, "dur": 6,
    "args": {
      "External id": 21910,
      "cbid": 211, "correlation": 21910
    }
  },
  {
    "ph": "s", "id": 21910, "pid": 76337, "tid": -914061504, "ts": 1716454217614513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217640316, "dur": 22,
    "args": {
      "External id": 21912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21912, "pid": 5, "tid": 7, "ts": 1716454217640316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614523, "dur": 5,
    "args": {
      "External id": 21912,
      "cbid": 211, "correlation": 21912
    }
  },
  {
    "ph": "s", "id": 21912, "pid": 76337, "tid": -914061504, "ts": 1716454217614523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217640339, "dur": 33,
    "args": {
      "External id": 21918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21918,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21918, "pid": 5, "tid": 7, "ts": 1716454217640339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614551, "dur": 8,
    "args": {
      "External id": 21918,
      "cbid": 211, "correlation": 21918
    }
  },
  {
    "ph": "s", "id": 21918, "pid": 76337, "tid": -914061504, "ts": 1716454217614551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217640374, "dur": 26,
    "args": {
      "External id": 21926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21926,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21926, "pid": 5, "tid": 7, "ts": 1716454217640374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614583, "dur": 8,
    "args": {
      "External id": 21926,
      "cbid": 211, "correlation": 21926
    }
  },
  {
    "ph": "s", "id": 21926, "pid": 76337, "tid": -914061504, "ts": 1716454217614583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217640401, "dur": 20,
    "args": {
      "External id": 21934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21934,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21934, "pid": 5, "tid": 7, "ts": 1716454217640401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614613, "dur": 8,
    "args": {
      "External id": 21934,
      "cbid": 211, "correlation": 21934
    }
  },
  {
    "ph": "s", "id": 21934, "pid": 76337, "tid": -914061504, "ts": 1716454217614613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217640422, "dur": 29,
    "args": {
      "External id": 21954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21954,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 21954, "pid": 5, "tid": 7, "ts": 1716454217640422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614693, "dur": 12,
    "args": {
      "External id": 21954,
      "cbid": 211, "correlation": 21954
    }
  },
  {
    "ph": "s", "id": 21954, "pid": 76337, "tid": -914061504, "ts": 1716454217614693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217640453, "dur": 4,
    "args": {
      "External id": 21966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21966,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 21966, "pid": 5, "tid": 7, "ts": 1716454217640453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614715, "dur": 6,
    "args": {
      "External id": 21966,
      "cbid": 211, "correlation": 21966
    }
  },
  {
    "ph": "s", "id": 21966, "pid": 76337, "tid": -914061504, "ts": 1716454217614715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217640459, "dur": 30,
    "args": {
      "External id": 21969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21969,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21969, "pid": 5, "tid": 7, "ts": 1716454217640459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614733, "dur": 6,
    "args": {
      "External id": 21969,
      "cbid": 211, "correlation": 21969
    }
  },
  {
    "ph": "s", "id": 21969, "pid": 76337, "tid": -914061504, "ts": 1716454217614733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614791, "dur": 0,
    "args": {
      "External id": 21980,
      "cbid": 317, "correlation": 21980
    }
  },
  {
    "ph": "f", "id": 21980, "pid": 76337, "tid": -914061504, "ts": 1716454217614791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614791, "dur": 0,
    "args": {
      "External id": 21981,
      "cbid": 203, "correlation": 21981
    }
  },
  {
    "ph": "f", "id": 21981, "pid": 76337, "tid": -914061504, "ts": 1716454217614791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614792, "dur": 0,
    "args": {
      "External id": 21982,
      "cbid": 205, "correlation": 21982
    }
  },
  {
    "ph": "f", "id": 21982, "pid": 76337, "tid": -914061504, "ts": 1716454217614792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217640490, "dur": 22,
    "args": {
      "External id": 21986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21986, "pid": 5, "tid": 7, "ts": 1716454217640490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614805, "dur": 11,
    "args": {
      "External id": 21986,
      "cbid": 211, "correlation": 21986
    }
  },
  {
    "ph": "s", "id": 21986, "pid": 76337, "tid": -914061504, "ts": 1716454217614805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217640513, "dur": 114,
    "args": {
      "External id": 21988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21988,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 21988, "pid": 5, "tid": 7, "ts": 1716454217640513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614822, "dur": 6,
    "args": {
      "External id": 21988,
      "cbid": 211, "correlation": 21988
    }
  },
  {
    "ph": "s", "id": 21988, "pid": 76337, "tid": -914061504, "ts": 1716454217614822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217640628, "dur": 22,
    "args": {
      "External id": 21990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21990,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21990, "pid": 5, "tid": 7, "ts": 1716454217640628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614833, "dur": 5,
    "args": {
      "External id": 21990,
      "cbid": 211, "correlation": 21990
    }
  },
  {
    "ph": "s", "id": 21990, "pid": 76337, "tid": -914061504, "ts": 1716454217614833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217640651, "dur": 32,
    "args": {
      "External id": 21996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 21996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 21996, "pid": 5, "tid": 7, "ts": 1716454217640651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614860, "dur": 8,
    "args": {
      "External id": 21996,
      "cbid": 211, "correlation": 21996
    }
  },
  {
    "ph": "s", "id": 21996, "pid": 76337, "tid": -914061504, "ts": 1716454217614860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217640685, "dur": 166,
    "args": {
      "External id": 22005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22005,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22005, "pid": 5, "tid": 7, "ts": 1716454217640685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217614942, "dur": 14,
    "args": {
      "External id": 22005,
      "cbid": 211, "correlation": 22005
    }
  },
  {
    "ph": "s", "id": 22005, "pid": 76337, "tid": -914061504, "ts": 1716454217614942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217640852, "dur": 64,
    "args": {
      "External id": 22027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22027,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22027, "pid": 5, "tid": 7, "ts": 1716454217640852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615025, "dur": 12,
    "args": {
      "External id": 22027,
      "cbid": 211, "correlation": 22027
    }
  },
  {
    "ph": "s", "id": 22027, "pid": 76337, "tid": -914061504, "ts": 1716454217615025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615129, "dur": 1,
    "args": {
      "External id": 22038,
      "cbid": 251, "correlation": 22038
    }
  },
  {
    "ph": "f", "id": 22038, "pid": 76337, "tid": -914061504, "ts": 1716454217615129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217640917, "dur": 151,
    "args": {
      "External id": 22039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22039,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22039, "pid": 5, "tid": 7, "ts": 1716454217640917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615135, "dur": 13,
    "args": {
      "External id": 22039,
      "cbid": 211, "correlation": 22039
    }
  },
  {
    "ph": "s", "id": 22039, "pid": 76337, "tid": -914061504, "ts": 1716454217615135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615222, "dur": 1,
    "args": {
      "External id": 22050,
      "cbid": 251, "correlation": 22050
    }
  },
  {
    "ph": "f", "id": 22050, "pid": 76337, "tid": -914061504, "ts": 1716454217615222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217641070, "dur": 145,
    "args": {
      "External id": 22051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22051,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22051, "pid": 5, "tid": 7, "ts": 1716454217641070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615227, "dur": 12,
    "args": {
      "External id": 22051,
      "cbid": 211, "correlation": 22051
    }
  },
  {
    "ph": "s", "id": 22051, "pid": 76337, "tid": -914061504, "ts": 1716454217615227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615306, "dur": 1,
    "args": {
      "External id": 22062,
      "cbid": 251, "correlation": 22062
    }
  },
  {
    "ph": "f", "id": 22062, "pid": 76337, "tid": -914061504, "ts": 1716454217615306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217641216, "dur": 143,
    "args": {
      "External id": 22063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22063,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22063, "pid": 5, "tid": 7, "ts": 1716454217641216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615311, "dur": 12,
    "args": {
      "External id": 22063,
      "cbid": 211, "correlation": 22063
    }
  },
  {
    "ph": "s", "id": 22063, "pid": 76337, "tid": -914061504, "ts": 1716454217615311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217641361, "dur": 1911,
    "args": {
      "External id": 22084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22084,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 22084, "pid": 5, "tid": 7, "ts": 1716454217641361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615393, "dur": 13,
    "args": {
      "External id": 22084,
      "cbid": 211, "correlation": 22084
    }
  },
  {
    "ph": "s", "id": 22084, "pid": 76337, "tid": -914061504, "ts": 1716454217615393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615504, "dur": 1,
    "args": {
      "External id": 22102,
      "cbid": 251, "correlation": 22102
    }
  },
  {
    "ph": "f", "id": 22102, "pid": 76337, "tid": -914061504, "ts": 1716454217615504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217643273, "dur": 146,
    "args": {
      "External id": 22104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22104,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 22104, "pid": 5, "tid": 7, "ts": 1716454217643273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615510, "dur": 15,
    "args": {
      "External id": 22104,
      "cbid": 211, "correlation": 22104
    }
  },
  {
    "ph": "s", "id": 22104, "pid": 76337, "tid": -914061504, "ts": 1716454217615510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217643420, "dur": 35,
    "args": {
      "External id": 22112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22112,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22112, "pid": 5, "tid": 7, "ts": 1716454217643420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615584, "dur": 12,
    "args": {
      "External id": 22112,
      "cbid": 211, "correlation": 22112
    }
  },
  {
    "ph": "s", "id": 22112, "pid": 76337, "tid": -914061504, "ts": 1716454217615584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217643457, "dur": 50,
    "args": {
      "External id": 22120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22120,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22120, "pid": 5, "tid": 7, "ts": 1716454217643457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615622, "dur": 9,
    "args": {
      "External id": 22120,
      "cbid": 211, "correlation": 22120
    }
  },
  {
    "ph": "s", "id": 22120, "pid": 76337, "tid": -914061504, "ts": 1716454217615622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217643509, "dur": 30,
    "args": {
      "External id": 22131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22131, "pid": 5, "tid": 7, "ts": 1716454217643509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615695, "dur": 12,
    "args": {
      "External id": 22131,
      "cbid": 211, "correlation": 22131
    }
  },
  {
    "ph": "s", "id": 22131, "pid": 76337, "tid": -914061504, "ts": 1716454217615695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217643540, "dur": 34,
    "args": {
      "External id": 22153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22153,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22153, "pid": 5, "tid": 7, "ts": 1716454217643540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615737, "dur": 10,
    "args": {
      "External id": 22153,
      "cbid": 211, "correlation": 22153
    }
  },
  {
    "ph": "s", "id": 22153, "pid": 76337, "tid": -914061504, "ts": 1716454217615737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615841, "dur": 1,
    "args": {
      "External id": 22164,
      "cbid": 251, "correlation": 22164
    }
  },
  {
    "ph": "f", "id": 22164, "pid": 76337, "tid": -914061504, "ts": 1716454217615841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217643576, "dur": 89,
    "args": {
      "External id": 22165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22165,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22165, "pid": 5, "tid": 7, "ts": 1716454217643576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615846, "dur": 14,
    "args": {
      "External id": 22165,
      "cbid": 211, "correlation": 22165
    }
  },
  {
    "ph": "s", "id": 22165, "pid": 76337, "tid": -914061504, "ts": 1716454217615846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615930, "dur": 1,
    "args": {
      "External id": 22176,
      "cbid": 251, "correlation": 22176
    }
  },
  {
    "ph": "f", "id": 22176, "pid": 76337, "tid": -914061504, "ts": 1716454217615930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615934, "dur": 0,
    "args": {
      "External id": 22177,
      "cbid": 251, "correlation": 22177
    }
  },
  {
    "ph": "f", "id": 22177, "pid": 76337, "tid": -914061504, "ts": 1716454217615934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217643666, "dur": 12,
    "args": {
      "External id": 22178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22178,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 22178, "pid": 5, "tid": 7, "ts": 1716454217643666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615936, "dur": 13,
    "args": {
      "External id": 22178,
      "cbid": 211, "correlation": 22178
    }
  },
  {
    "ph": "s", "id": 22178, "pid": 76337, "tid": -914061504, "ts": 1716454217615936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217643680, "dur": 5,
    "args": {
      "External id": 22180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22180,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 22180, "pid": 5, "tid": 7, "ts": 1716454217643680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217615950, "dur": 6,
    "args": {
      "External id": 22180,
      "cbid": 211, "correlation": 22180
    }
  },
  {
    "ph": "s", "id": 22180, "pid": 76337, "tid": -914061504, "ts": 1716454217615950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616028, "dur": 1,
    "args": {
      "External id": 22191,
      "cbid": 251, "correlation": 22191
    }
  },
  {
    "ph": "f", "id": 22191, "pid": 76337, "tid": -914061504, "ts": 1716454217616028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616032, "dur": 0,
    "args": {
      "External id": 22192,
      "cbid": 251, "correlation": 22192
    }
  },
  {
    "ph": "f", "id": 22192, "pid": 76337, "tid": -914061504, "ts": 1716454217616032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217643686, "dur": 7,
    "args": {
      "External id": 22193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22193,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 22193, "pid": 5, "tid": 7, "ts": 1716454217643686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616034, "dur": 13,
    "args": {
      "External id": 22193,
      "cbid": 211, "correlation": 22193
    }
  },
  {
    "ph": "s", "id": 22193, "pid": 76337, "tid": -914061504, "ts": 1716454217616034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217643694, "dur": 4,
    "args": {
      "External id": 22195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22195,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 22195, "pid": 5, "tid": 7, "ts": 1716454217643694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616049, "dur": 6,
    "args": {
      "External id": 22195,
      "cbid": 211, "correlation": 22195
    }
  },
  {
    "ph": "s", "id": 22195, "pid": 76337, "tid": -914061504, "ts": 1716454217616049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217643699, "dur": 91,
    "args": {
      "External id": 22216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22216,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 22216, "pid": 5, "tid": 7, "ts": 1716454217643699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616124, "dur": 12,
    "args": {
      "External id": 22216,
      "cbid": 211, "correlation": 22216
    }
  },
  {
    "ph": "s", "id": 22216, "pid": 76337, "tid": -914061504, "ts": 1716454217616124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616240, "dur": 1,
    "args": {
      "External id": 22234,
      "cbid": 251, "correlation": 22234
    }
  },
  {
    "ph": "f", "id": 22234, "pid": 76337, "tid": -914061504, "ts": 1716454217616240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217643791, "dur": 97,
    "args": {
      "External id": 22236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22236,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22236, "pid": 5, "tid": 7, "ts": 1716454217643791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616246, "dur": 14,
    "args": {
      "External id": 22236,
      "cbid": 211, "correlation": 22236
    }
  },
  {
    "ph": "s", "id": 22236, "pid": 76337, "tid": -914061504, "ts": 1716454217616246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217643890, "dur": 19,
    "args": {
      "External id": 22244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22244,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22244, "pid": 5, "tid": 7, "ts": 1716454217643890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616318, "dur": 13,
    "args": {
      "External id": 22244,
      "cbid": 211, "correlation": 22244
    }
  },
  {
    "ph": "s", "id": 22244, "pid": 76337, "tid": -914061504, "ts": 1716454217616318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217643910, "dur": 37,
    "args": {
      "External id": 22252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22252,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22252, "pid": 5, "tid": 7, "ts": 1716454217643910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616360, "dur": 9,
    "args": {
      "External id": 22252,
      "cbid": 211, "correlation": 22252
    }
  },
  {
    "ph": "s", "id": 22252, "pid": 76337, "tid": -914061504, "ts": 1716454217616360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217643948, "dur": 34,
    "args": {
      "External id": 22274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22274,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22274, "pid": 5, "tid": 7, "ts": 1716454217643948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616422, "dur": 11,
    "args": {
      "External id": 22274,
      "cbid": 211, "correlation": 22274
    }
  },
  {
    "ph": "s", "id": 22274, "pid": 76337, "tid": -914061504, "ts": 1716454217616422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616515, "dur": 1,
    "args": {
      "External id": 22290,
      "cbid": 251, "correlation": 22290
    }
  },
  {
    "ph": "f", "id": 22290, "pid": 76337, "tid": -914061504, "ts": 1716454217616515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616520, "dur": 0,
    "args": {
      "External id": 22292,
      "cbid": 251, "correlation": 22292
    }
  },
  {
    "ph": "f", "id": 22292, "pid": 76337, "tid": -914061504, "ts": 1716454217616520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217643984, "dur": 532,
    "args": {
      "External id": 22293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22293,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 22293, "pid": 5, "tid": 7, "ts": 1716454217643984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616523, "dur": 13,
    "args": {
      "External id": 22293,
      "cbid": 211, "correlation": 22293
    }
  },
  {
    "ph": "s", "id": 22293, "pid": 76337, "tid": -914061504, "ts": 1716454217616523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217644517, "dur": 124,
    "args": {
      "External id": 22301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22301, "pid": 5, "tid": 7, "ts": 1716454217644517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616590, "dur": 12,
    "args": {
      "External id": 22301,
      "cbid": 211, "correlation": 22301
    }
  },
  {
    "ph": "s", "id": 22301, "pid": 76337, "tid": -914061504, "ts": 1716454217616590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217644642, "dur": 126,
    "args": {
      "External id": 22309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22309, "pid": 5, "tid": 7, "ts": 1716454217644642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616621, "dur": 8,
    "args": {
      "External id": 22309,
      "cbid": 211, "correlation": 22309
    }
  },
  {
    "ph": "s", "id": 22309, "pid": 76337, "tid": -914061504, "ts": 1716454217616621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616698, "dur": 1,
    "args": {
      "External id": 22325,
      "cbid": 251, "correlation": 22325
    }
  },
  {
    "ph": "f", "id": 22325, "pid": 76337, "tid": -914061504, "ts": 1716454217616698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217644770, "dur": 303,
    "args": {
      "External id": 22327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22327,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22327, "pid": 5, "tid": 7, "ts": 1716454217644770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616703, "dur": 12,
    "args": {
      "External id": 22327,
      "cbid": 211, "correlation": 22327
    }
  },
  {
    "ph": "s", "id": 22327, "pid": 76337, "tid": -914061504, "ts": 1716454217616703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217645074, "dur": 28,
    "args": {
      "External id": 22335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22335,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22335, "pid": 5, "tid": 7, "ts": 1716454217645074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616746, "dur": 10,
    "args": {
      "External id": 22335,
      "cbid": 211, "correlation": 22335
    }
  },
  {
    "ph": "s", "id": 22335, "pid": 76337, "tid": -914061504, "ts": 1716454217616746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217645103, "dur": 80,
    "args": {
      "External id": 22346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22346, "pid": 5, "tid": 7, "ts": 1716454217645103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616813, "dur": 12,
    "args": {
      "External id": 22346,
      "cbid": 211, "correlation": 22346
    }
  },
  {
    "ph": "s", "id": 22346, "pid": 76337, "tid": -914061504, "ts": 1716454217616813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616877, "dur": 0,
    "args": {
      "External id": 22358,
      "cbid": 317, "correlation": 22358
    }
  },
  {
    "ph": "f", "id": 22358, "pid": 76337, "tid": -914061504, "ts": 1716454217616877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616877, "dur": 0,
    "args": {
      "External id": 22359,
      "cbid": 203, "correlation": 22359
    }
  },
  {
    "ph": "f", "id": 22359, "pid": 76337, "tid": -914061504, "ts": 1716454217616877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616878, "dur": 0,
    "args": {
      "External id": 22360,
      "cbid": 205, "correlation": 22360
    }
  },
  {
    "ph": "f", "id": 22360, "pid": 76337, "tid": -914061504, "ts": 1716454217616878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645184, "dur": 24,
    "args": {
      "External id": 22364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22364,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22364, "pid": 5, "tid": 7, "ts": 1716454217645184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616906, "dur": 13,
    "args": {
      "External id": 22364,
      "cbid": 211, "correlation": 22364
    }
  },
  {
    "ph": "s", "id": 22364, "pid": 76337, "tid": -914061504, "ts": 1716454217616906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217645209, "dur": 119,
    "args": {
      "External id": 22366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22366,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22366, "pid": 5, "tid": 7, "ts": 1716454217645209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616925, "dur": 6,
    "args": {
      "External id": 22366,
      "cbid": 211, "correlation": 22366
    }
  },
  {
    "ph": "s", "id": 22366, "pid": 76337, "tid": -914061504, "ts": 1716454217616925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645329, "dur": 21,
    "args": {
      "External id": 22368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22368,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22368, "pid": 5, "tid": 7, "ts": 1716454217645329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616935, "dur": 5,
    "args": {
      "External id": 22368,
      "cbid": 211, "correlation": 22368
    }
  },
  {
    "ph": "s", "id": 22368, "pid": 76337, "tid": -914061504, "ts": 1716454217616935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217645352, "dur": 33,
    "args": {
      "External id": 22374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22374, "pid": 5, "tid": 7, "ts": 1716454217645352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217616963, "dur": 9,
    "args": {
      "External id": 22374,
      "cbid": 211, "correlation": 22374
    }
  },
  {
    "ph": "s", "id": 22374, "pid": 76337, "tid": -914061504, "ts": 1716454217616963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217645386, "dur": 27,
    "args": {
      "External id": 22382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22382,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22382, "pid": 5, "tid": 7, "ts": 1716454217645386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617004, "dur": 9,
    "args": {
      "External id": 22382,
      "cbid": 211, "correlation": 22382
    }
  },
  {
    "ph": "s", "id": 22382, "pid": 76337, "tid": -914061504, "ts": 1716454217617004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617078, "dur": 0,
    "args": {
      "External id": 22392,
      "cbid": 317, "correlation": 22392
    }
  },
  {
    "ph": "f", "id": 22392, "pid": 76337, "tid": -914061504, "ts": 1716454217617078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617079, "dur": 0,
    "args": {
      "External id": 22393,
      "cbid": 203, "correlation": 22393
    }
  },
  {
    "ph": "f", "id": 22393, "pid": 76337, "tid": -914061504, "ts": 1716454217617079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617080, "dur": 0,
    "args": {
      "External id": 22394,
      "cbid": 205, "correlation": 22394
    }
  },
  {
    "ph": "f", "id": 22394, "pid": 76337, "tid": -914061504, "ts": 1716454217617080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645414, "dur": 24,
    "args": {
      "External id": 22398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22398,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22398, "pid": 5, "tid": 7, "ts": 1716454217645414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617094, "dur": 12,
    "args": {
      "External id": 22398,
      "cbid": 211, "correlation": 22398
    }
  },
  {
    "ph": "s", "id": 22398, "pid": 76337, "tid": -914061504, "ts": 1716454217617094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645439, "dur": 44,
    "args": {
      "External id": 22400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22400,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22400, "pid": 5, "tid": 7, "ts": 1716454217645439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617109, "dur": 5,
    "args": {
      "External id": 22400,
      "cbid": 211, "correlation": 22400
    }
  },
  {
    "ph": "s", "id": 22400, "pid": 76337, "tid": -914061504, "ts": 1716454217617109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217645484, "dur": 232,
    "args": {
      "External id": 22402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22402,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 22402, "pid": 5, "tid": 7, "ts": 1716454217645484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617121, "dur": 7,
    "args": {
      "External id": 22402,
      "cbid": 211, "correlation": 22402
    }
  },
  {
    "ph": "s", "id": 22402, "pid": 76337, "tid": -914061504, "ts": 1716454217617121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645717, "dur": 6,
    "args": {
      "External id": 22404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22404, "pid": 5, "tid": 7, "ts": 1716454217645717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617132, "dur": 5,
    "args": {
      "External id": 22404,
      "cbid": 211, "correlation": 22404
    }
  },
  {
    "ph": "s", "id": 22404, "pid": 76337, "tid": -914061504, "ts": 1716454217617132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217645725, "dur": 9,
    "args": {
      "External id": 22410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22410,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22410, "pid": 5, "tid": 7, "ts": 1716454217645725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617158, "dur": 8,
    "args": {
      "External id": 22410,
      "cbid": 211, "correlation": 22410
    }
  },
  {
    "ph": "s", "id": 22410, "pid": 76337, "tid": -914061504, "ts": 1716454217617158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645735, "dur": 12,
    "args": {
      "External id": 22430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22430,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 22430, "pid": 5, "tid": 7, "ts": 1716454217645735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617254, "dur": 12,
    "args": {
      "External id": 22430,
      "cbid": 211, "correlation": 22430
    }
  },
  {
    "ph": "s", "id": 22430, "pid": 76337, "tid": -914061504, "ts": 1716454217617254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217645748, "dur": 4,
    "args": {
      "External id": 22442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22442,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 22442, "pid": 5, "tid": 7, "ts": 1716454217645748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617286, "dur": 8,
    "args": {
      "External id": 22442,
      "cbid": 211, "correlation": 22442
    }
  },
  {
    "ph": "s", "id": 22442, "pid": 76337, "tid": -914061504, "ts": 1716454217617286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217645753, "dur": 12,
    "args": {
      "External id": 22445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22445, "pid": 5, "tid": 7, "ts": 1716454217645753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617307, "dur": 7,
    "args": {
      "External id": 22445,
      "cbid": 211, "correlation": 22445
    }
  },
  {
    "ph": "s", "id": 22445, "pid": 76337, "tid": -914061504, "ts": 1716454217617307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217645767, "dur": 7,
    "args": {
      "External id": 22454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22454,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22454, "pid": 5, "tid": 7, "ts": 1716454217645767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617348, "dur": 9,
    "args": {
      "External id": 22454,
      "cbid": 211, "correlation": 22454
    }
  },
  {
    "ph": "s", "id": 22454, "pid": 76337, "tid": -914061504, "ts": 1716454217617348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617410, "dur": 0,
    "args": {
      "External id": 22464,
      "cbid": 317, "correlation": 22464
    }
  },
  {
    "ph": "f", "id": 22464, "pid": 76337, "tid": -914061504, "ts": 1716454217617410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617411, "dur": 0,
    "args": {
      "External id": 22465,
      "cbid": 203, "correlation": 22465
    }
  },
  {
    "ph": "f", "id": 22465, "pid": 76337, "tid": -914061504, "ts": 1716454217617411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617411, "dur": 0,
    "args": {
      "External id": 22466,
      "cbid": 205, "correlation": 22466
    }
  },
  {
    "ph": "f", "id": 22466, "pid": 76337, "tid": -914061504, "ts": 1716454217617411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645775, "dur": 5,
    "args": {
      "External id": 22470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22470,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22470, "pid": 5, "tid": 7, "ts": 1716454217645775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617434, "dur": 12,
    "args": {
      "External id": 22470,
      "cbid": 211, "correlation": 22470
    }
  },
  {
    "ph": "s", "id": 22470, "pid": 76337, "tid": -914061504, "ts": 1716454217617434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217645782, "dur": 82,
    "args": {
      "External id": 22472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22472,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22472, "pid": 5, "tid": 7, "ts": 1716454217645782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617450, "dur": 5,
    "args": {
      "External id": 22472,
      "cbid": 211, "correlation": 22472
    }
  },
  {
    "ph": "s", "id": 22472, "pid": 76337, "tid": -914061504, "ts": 1716454217617450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217645867, "dur": 1,
    "args": {
      "External id": 22474,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 22474,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 22474, "pid": 5, "tid": 7, "ts": 1716454217645867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617463, "dur": 9,
    "args": {
      "External id": 22474,
      "cbid": 51, "correlation": 22474
    }
  },
  {
    "ph": "s", "id": 22474, "pid": 76337, "tid": -914061504, "ts": 1716454217617463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217645870, "dur": 534,
    "args": {
      "External id": 22475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22475,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22475, "pid": 5, "tid": 7, "ts": 1716454217645870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617473, "dur": 9,
    "args": {
      "External id": 22475,
      "cbid": 211, "correlation": 22475
    }
  },
  {
    "ph": "s", "id": 22475, "pid": 76337, "tid": -914061504, "ts": 1716454217617473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217646406, "dur": 11,
    "args": {
      "External id": 22477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22477,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22477, "pid": 5, "tid": 7, "ts": 1716454217646406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617487, "dur": 5,
    "args": {
      "External id": 22477,
      "cbid": 211, "correlation": 22477
    }
  },
  {
    "ph": "s", "id": 22477, "pid": 76337, "tid": -914061504, "ts": 1716454217617487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217646418, "dur": 14,
    "args": {
      "External id": 22483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22483, "pid": 5, "tid": 7, "ts": 1716454217646418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617515, "dur": 8,
    "args": {
      "External id": 22483,
      "cbid": 211, "correlation": 22483
    }
  },
  {
    "ph": "s", "id": 22483, "pid": 76337, "tid": -914061504, "ts": 1716454217617515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217646434, "dur": 3,
    "args": {
      "External id": 22491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22491,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 22491, "pid": 5, "tid": 7, "ts": 1716454217646434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617560, "dur": 9,
    "args": {
      "External id": 22491,
      "cbid": 211, "correlation": 22491
    }
  },
  {
    "ph": "s", "id": 22491, "pid": 76337, "tid": -914061504, "ts": 1716454217617560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617643, "dur": 2,
    "args": {
      "External id": 22507,
      "cbid": 251, "correlation": 22507
    }
  },
  {
    "ph": "f", "id": 22507, "pid": 76337, "tid": -914061504, "ts": 1716454217617643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617649, "dur": 0,
    "args": {
      "External id": 22509,
      "cbid": 251, "correlation": 22509
    }
  },
  {
    "ph": "f", "id": 22509, "pid": 76337, "tid": -914061504, "ts": 1716454217617649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217646439, "dur": 13,
    "args": {
      "External id": 22510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22510,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22510, "pid": 5, "tid": 7, "ts": 1716454217646439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617652, "dur": 12,
    "args": {
      "External id": 22510,
      "cbid": 211, "correlation": 22510
    }
  },
  {
    "ph": "s", "id": 22510, "pid": 76337, "tid": -914061504, "ts": 1716454217617652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217646453, "dur": 5,
    "args": {
      "External id": 22512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22512,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22512, "pid": 5, "tid": 7, "ts": 1716454217646453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617666, "dur": 5,
    "args": {
      "External id": 22512,
      "cbid": 211, "correlation": 22512
    }
  },
  {
    "ph": "s", "id": 22512, "pid": 76337, "tid": -914061504, "ts": 1716454217617666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217646459, "dur": 16,
    "args": {
      "External id": 22522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22522, "pid": 5, "tid": 7, "ts": 1716454217646459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617724, "dur": 12,
    "args": {
      "External id": 22522,
      "cbid": 211, "correlation": 22522
    }
  },
  {
    "ph": "s", "id": 22522, "pid": 76337, "tid": -914061504, "ts": 1716454217617724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217646477, "dur": 18,
    "args": {
      "External id": 22542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22542,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 22542, "pid": 5, "tid": 7, "ts": 1716454217646477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617790, "dur": 11,
    "args": {
      "External id": 22542,
      "cbid": 211, "correlation": 22542
    }
  },
  {
    "ph": "s", "id": 22542, "pid": 76337, "tid": -914061504, "ts": 1716454217617790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217646496, "dur": 4,
    "args": {
      "External id": 22554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22554,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 22554, "pid": 5, "tid": 7, "ts": 1716454217646496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617826, "dur": 9,
    "args": {
      "External id": 22554,
      "cbid": 211, "correlation": 22554
    }
  },
  {
    "ph": "s", "id": 22554, "pid": 76337, "tid": -914061504, "ts": 1716454217617826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217646502, "dur": 17,
    "args": {
      "External id": 22557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22557, "pid": 5, "tid": 7, "ts": 1716454217646502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617847, "dur": 7,
    "args": {
      "External id": 22557,
      "cbid": 211, "correlation": 22557
    }
  },
  {
    "ph": "s", "id": 22557, "pid": 76337, "tid": -914061504, "ts": 1716454217617847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217646519, "dur": 11,
    "args": {
      "External id": 22566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22566,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22566, "pid": 5, "tid": 7, "ts": 1716454217646519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617891, "dur": 10,
    "args": {
      "External id": 22566,
      "cbid": 211, "correlation": 22566
    }
  },
  {
    "ph": "s", "id": 22566, "pid": 76337, "tid": -914061504, "ts": 1716454217617891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617969, "dur": 0,
    "args": {
      "External id": 22576,
      "cbid": 317, "correlation": 22576
    }
  },
  {
    "ph": "f", "id": 22576, "pid": 76337, "tid": -914061504, "ts": 1716454217617969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617970, "dur": 0,
    "args": {
      "External id": 22577,
      "cbid": 203, "correlation": 22577
    }
  },
  {
    "ph": "f", "id": 22577, "pid": 76337, "tid": -914061504, "ts": 1716454217617970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217617970, "dur": 0,
    "args": {
      "External id": 22578,
      "cbid": 205, "correlation": 22578
    }
  },
  {
    "ph": "f", "id": 22578, "pid": 76337, "tid": -914061504, "ts": 1716454217617970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217646532, "dur": 11,
    "args": {
      "External id": 22582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22582,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22582, "pid": 5, "tid": 7, "ts": 1716454217646532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618002, "dur": 13,
    "args": {
      "External id": 22582,
      "cbid": 211, "correlation": 22582
    }
  },
  {
    "ph": "s", "id": 22582, "pid": 76337, "tid": -914061504, "ts": 1716454217618002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217646544, "dur": 160,
    "args": {
      "External id": 22584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22584,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22584, "pid": 5, "tid": 7, "ts": 1716454217646544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618018, "dur": 5,
    "args": {
      "External id": 22584,
      "cbid": 211, "correlation": 22584
    }
  },
  {
    "ph": "s", "id": 22584, "pid": 76337, "tid": -914061504, "ts": 1716454217618018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217646706, "dur": 1,
    "args": {
      "External id": 22586,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 22586,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 22586, "pid": 5, "tid": 7, "ts": 1716454217646706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618030, "dur": 7,
    "args": {
      "External id": 22586,
      "cbid": 51, "correlation": 22586
    }
  },
  {
    "ph": "s", "id": 22586, "pid": 76337, "tid": -914061504, "ts": 1716454217618030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217646710, "dur": 654,
    "args": {
      "External id": 22587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22587,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22587, "pid": 5, "tid": 7, "ts": 1716454217646710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618038, "dur": 7,
    "args": {
      "External id": 22587,
      "cbid": 211, "correlation": 22587
    }
  },
  {
    "ph": "s", "id": 22587, "pid": 76337, "tid": -914061504, "ts": 1716454217618038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217647365, "dur": 12,
    "args": {
      "External id": 22589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22589,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22589, "pid": 5, "tid": 7, "ts": 1716454217647365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618050, "dur": 5,
    "args": {
      "External id": 22589,
      "cbid": 211, "correlation": 22589
    }
  },
  {
    "ph": "s", "id": 22589, "pid": 76337, "tid": -914061504, "ts": 1716454217618050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217647378, "dur": 14,
    "args": {
      "External id": 22595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22595,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22595, "pid": 5, "tid": 7, "ts": 1716454217647378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618080, "dur": 8,
    "args": {
      "External id": 22595,
      "cbid": 211, "correlation": 22595
    }
  },
  {
    "ph": "s", "id": 22595, "pid": 76337, "tid": -914061504, "ts": 1716454217618080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618139, "dur": 0,
    "args": {
      "External id": 22605,
      "cbid": 317, "correlation": 22605
    }
  },
  {
    "ph": "f", "id": 22605, "pid": 76337, "tid": -914061504, "ts": 1716454217618139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618140, "dur": 0,
    "args": {
      "External id": 22606,
      "cbid": 203, "correlation": 22606
    }
  },
  {
    "ph": "f", "id": 22606, "pid": 76337, "tid": -914061504, "ts": 1716454217618140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618141, "dur": 0,
    "args": {
      "External id": 22607,
      "cbid": 205, "correlation": 22607
    }
  },
  {
    "ph": "f", "id": 22607, "pid": 76337, "tid": -914061504, "ts": 1716454217618141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217647394, "dur": 8,
    "args": {
      "External id": 22611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22611,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22611, "pid": 5, "tid": 7, "ts": 1716454217647394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618156, "dur": 11,
    "args": {
      "External id": 22611,
      "cbid": 211, "correlation": 22611
    }
  },
  {
    "ph": "s", "id": 22611, "pid": 76337, "tid": -914061504, "ts": 1716454217618156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217647404, "dur": 4,
    "args": {
      "External id": 22613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22613,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 22613, "pid": 5, "tid": 7, "ts": 1716454217647404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618173, "dur": 7,
    "args": {
      "External id": 22613,
      "cbid": 211, "correlation": 22613
    }
  },
  {
    "ph": "s", "id": 22613, "pid": 76337, "tid": -914061504, "ts": 1716454217618173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618183, "dur": 0,
    "args": {
      "External id": 22614,
      "cbid": 51, "correlation": 22614
    }
  },
  {
    "ph": "s", "id": 22614, "pid": 76337, "tid": -914061504, "ts": 1716454217618183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217647408, "dur": 56,
    "args": {
      "External id": 22615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22615,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 22615, "pid": 5, "tid": 7, "ts": 1716454217647408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618184, "dur": 5,
    "args": {
      "External id": 22615,
      "cbid": 211, "correlation": 22615
    }
  },
  {
    "ph": "s", "id": 22615, "pid": 76337, "tid": -914061504, "ts": 1716454217618184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217647466, "dur": 14,
    "args": {
      "External id": 22620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22620,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22620, "pid": 5, "tid": 7, "ts": 1716454217647466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618221, "dur": 10,
    "args": {
      "External id": 22620,
      "cbid": 211, "correlation": 22620
    }
  },
  {
    "ph": "s", "id": 22620, "pid": 76337, "tid": -914061504, "ts": 1716454217618221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217647481, "dur": 12,
    "args": {
      "External id": 22628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22628,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22628, "pid": 5, "tid": 7, "ts": 1716454217647481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618252, "dur": 8,
    "args": {
      "External id": 22628,
      "cbid": 211, "correlation": 22628
    }
  },
  {
    "ph": "s", "id": 22628, "pid": 76337, "tid": -914061504, "ts": 1716454217618252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217647495, "dur": 10,
    "args": {
      "External id": 22636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22636,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22636, "pid": 5, "tid": 7, "ts": 1716454217647495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618281, "dur": 8,
    "args": {
      "External id": 22636,
      "cbid": 211, "correlation": 22636
    }
  },
  {
    "ph": "s", "id": 22636, "pid": 76337, "tid": -914061504, "ts": 1716454217618281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217647506, "dur": 19,
    "args": {
      "External id": 22656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22656,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 22656, "pid": 5, "tid": 7, "ts": 1716454217647506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618363, "dur": 12,
    "args": {
      "External id": 22656,
      "cbid": 211, "correlation": 22656
    }
  },
  {
    "ph": "s", "id": 22656, "pid": 76337, "tid": -914061504, "ts": 1716454217618363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217647526, "dur": 4,
    "args": {
      "External id": 22668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22668,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 22668, "pid": 5, "tid": 7, "ts": 1716454217647526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618384, "dur": 6,
    "args": {
      "External id": 22668,
      "cbid": 211, "correlation": 22668
    }
  },
  {
    "ph": "s", "id": 22668, "pid": 76337, "tid": -914061504, "ts": 1716454217618384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217647532, "dur": 17,
    "args": {
      "External id": 22671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22671, "pid": 5, "tid": 7, "ts": 1716454217647532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618402, "dur": 7,
    "args": {
      "External id": 22671,
      "cbid": 211, "correlation": 22671
    }
  },
  {
    "ph": "s", "id": 22671, "pid": 76337, "tid": -914061504, "ts": 1716454217618402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618471, "dur": 0,
    "args": {
      "External id": 22682,
      "cbid": 317, "correlation": 22682
    }
  },
  {
    "ph": "f", "id": 22682, "pid": 76337, "tid": -914061504, "ts": 1716454217618471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618472, "dur": 0,
    "args": {
      "External id": 22683,
      "cbid": 203, "correlation": 22683
    }
  },
  {
    "ph": "f", "id": 22683, "pid": 76337, "tid": -914061504, "ts": 1716454217618472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618473, "dur": 0,
    "args": {
      "External id": 22684,
      "cbid": 205, "correlation": 22684
    }
  },
  {
    "ph": "f", "id": 22684, "pid": 76337, "tid": -914061504, "ts": 1716454217618473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217647550, "dur": 11,
    "args": {
      "External id": 22688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22688,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22688, "pid": 5, "tid": 7, "ts": 1716454217647550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618489, "dur": 13,
    "args": {
      "External id": 22688,
      "cbid": 211, "correlation": 22688
    }
  },
  {
    "ph": "s", "id": 22688, "pid": 76337, "tid": -914061504, "ts": 1716454217618489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217647562, "dur": 3,
    "args": {
      "External id": 22690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22690,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 22690, "pid": 5, "tid": 7, "ts": 1716454217647562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618507, "dur": 6,
    "args": {
      "External id": 22690,
      "cbid": 211, "correlation": 22690
    }
  },
  {
    "ph": "s", "id": 22690, "pid": 76337, "tid": -914061504, "ts": 1716454217618507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618517, "dur": 0,
    "args": {
      "External id": 22691,
      "cbid": 51, "correlation": 22691
    }
  },
  {
    "ph": "s", "id": 22691, "pid": 76337, "tid": -914061504, "ts": 1716454217618517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217647567, "dur": 96,
    "args": {
      "External id": 22692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22692,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 22692, "pid": 5, "tid": 7, "ts": 1716454217647567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618518, "dur": 7,
    "args": {
      "External id": 22692,
      "cbid": 211, "correlation": 22692
    }
  },
  {
    "ph": "s", "id": 22692, "pid": 76337, "tid": -914061504, "ts": 1716454217618518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217647664, "dur": 14,
    "args": {
      "External id": 22697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22697,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22697, "pid": 5, "tid": 7, "ts": 1716454217647664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618548, "dur": 8,
    "args": {
      "External id": 22697,
      "cbid": 211, "correlation": 22697
    }
  },
  {
    "ph": "s", "id": 22697, "pid": 76337, "tid": -914061504, "ts": 1716454217618548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217647679, "dur": 81,
    "args": {
      "External id": 22706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22706,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22706, "pid": 5, "tid": 7, "ts": 1716454217647679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618632, "dur": 14,
    "args": {
      "External id": 22706,
      "cbid": 211, "correlation": 22706
    }
  },
  {
    "ph": "s", "id": 22706, "pid": 76337, "tid": -914061504, "ts": 1716454217618632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217647762, "dur": 29,
    "args": {
      "External id": 22728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22728,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22728, "pid": 5, "tid": 7, "ts": 1716454217647762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618700, "dur": 12,
    "args": {
      "External id": 22728,
      "cbid": 211, "correlation": 22728
    }
  },
  {
    "ph": "s", "id": 22728, "pid": 76337, "tid": -914061504, "ts": 1716454217618700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618804, "dur": 2,
    "args": {
      "External id": 22739,
      "cbid": 251, "correlation": 22739
    }
  },
  {
    "ph": "f", "id": 22739, "pid": 76337, "tid": -914061504, "ts": 1716454217618804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217647792, "dur": 162,
    "args": {
      "External id": 22740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22740,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22740, "pid": 5, "tid": 7, "ts": 1716454217647792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618811, "dur": 14,
    "args": {
      "External id": 22740,
      "cbid": 211, "correlation": 22740
    }
  },
  {
    "ph": "s", "id": 22740, "pid": 76337, "tid": -914061504, "ts": 1716454217618811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618883, "dur": 1,
    "args": {
      "External id": 22751,
      "cbid": 251, "correlation": 22751
    }
  },
  {
    "ph": "f", "id": 22751, "pid": 76337, "tid": -914061504, "ts": 1716454217618883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217647956, "dur": 157,
    "args": {
      "External id": 22752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22752,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22752, "pid": 5, "tid": 7, "ts": 1716454217647956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618887, "dur": 11,
    "args": {
      "External id": 22752,
      "cbid": 211, "correlation": 22752
    }
  },
  {
    "ph": "s", "id": 22752, "pid": 76337, "tid": -914061504, "ts": 1716454217618887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618953, "dur": 1,
    "args": {
      "External id": 22763,
      "cbid": 251, "correlation": 22763
    }
  },
  {
    "ph": "f", "id": 22763, "pid": 76337, "tid": -914061504, "ts": 1716454217618953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217648114, "dur": 156,
    "args": {
      "External id": 22764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22764,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22764, "pid": 5, "tid": 7, "ts": 1716454217648114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217618957, "dur": 11,
    "args": {
      "External id": 22764,
      "cbid": 211, "correlation": 22764
    }
  },
  {
    "ph": "s", "id": 22764, "pid": 76337, "tid": -914061504, "ts": 1716454217618957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217648272, "dur": 329,
    "args": {
      "External id": 22789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22789,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22789, "pid": 5, "tid": 7, "ts": 1716454217648272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619050, "dur": 14,
    "args": {
      "External id": 22789,
      "cbid": 211, "correlation": 22789
    }
  },
  {
    "ph": "s", "id": 22789, "pid": 76337, "tid": -914061504, "ts": 1716454217619050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619171, "dur": 1,
    "args": {
      "External id": 22807,
      "cbid": 251, "correlation": 22807
    }
  },
  {
    "ph": "f", "id": 22807, "pid": 76337, "tid": -914061504, "ts": 1716454217619171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217648602, "dur": 163,
    "args": {
      "External id": 22809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22809,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22809, "pid": 5, "tid": 7, "ts": 1716454217648602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619177, "dur": 14,
    "args": {
      "External id": 22809,
      "cbid": 211, "correlation": 22809
    }
  },
  {
    "ph": "s", "id": 22809, "pid": 76337, "tid": -914061504, "ts": 1716454217619177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217648767, "dur": 19,
    "args": {
      "External id": 22817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22817,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22817, "pid": 5, "tid": 7, "ts": 1716454217648767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619249, "dur": 12,
    "args": {
      "External id": 22817,
      "cbid": 211, "correlation": 22817
    }
  },
  {
    "ph": "s", "id": 22817, "pid": 76337, "tid": -914061504, "ts": 1716454217619249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217648787, "dur": 28,
    "args": {
      "External id": 22825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22825,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22825, "pid": 5, "tid": 7, "ts": 1716454217648787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619288, "dur": 8,
    "args": {
      "External id": 22825,
      "cbid": 211, "correlation": 22825
    }
  },
  {
    "ph": "s", "id": 22825, "pid": 76337, "tid": -914061504, "ts": 1716454217619288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217648817, "dur": 18,
    "args": {
      "External id": 22836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22836,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22836, "pid": 5, "tid": 7, "ts": 1716454217648817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619360, "dur": 12,
    "args": {
      "External id": 22836,
      "cbid": 211, "correlation": 22836
    }
  },
  {
    "ph": "s", "id": 22836, "pid": 76337, "tid": -914061504, "ts": 1716454217619360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217648836, "dur": 16,
    "args": {
      "External id": 22858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22858,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22858, "pid": 5, "tid": 7, "ts": 1716454217648836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619390, "dur": 8,
    "args": {
      "External id": 22858,
      "cbid": 211, "correlation": 22858
    }
  },
  {
    "ph": "s", "id": 22858, "pid": 76337, "tid": -914061504, "ts": 1716454217619390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619489, "dur": 2,
    "args": {
      "External id": 22869,
      "cbid": 251, "correlation": 22869
    }
  },
  {
    "ph": "f", "id": 22869, "pid": 76337, "tid": -914061504, "ts": 1716454217619489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217648853, "dur": 88,
    "args": {
      "External id": 22870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22870,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 22870, "pid": 5, "tid": 7, "ts": 1716454217648853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619495, "dur": 16,
    "args": {
      "External id": 22870,
      "cbid": 211, "correlation": 22870
    }
  },
  {
    "ph": "s", "id": 22870, "pid": 76337, "tid": -914061504, "ts": 1716454217619495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619571, "dur": 1,
    "args": {
      "External id": 22881,
      "cbid": 251, "correlation": 22881
    }
  },
  {
    "ph": "f", "id": 22881, "pid": 76337, "tid": -914061504, "ts": 1716454217619571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619575, "dur": 0,
    "args": {
      "External id": 22882,
      "cbid": 251, "correlation": 22882
    }
  },
  {
    "ph": "f", "id": 22882, "pid": 76337, "tid": -914061504, "ts": 1716454217619575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217648943, "dur": 12,
    "args": {
      "External id": 22883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22883,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22883, "pid": 5, "tid": 7, "ts": 1716454217648943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619576, "dur": 12,
    "args": {
      "External id": 22883,
      "cbid": 211, "correlation": 22883
    }
  },
  {
    "ph": "s", "id": 22883, "pid": 76337, "tid": -914061504, "ts": 1716454217619576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217648956, "dur": 6,
    "args": {
      "External id": 22885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22885,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22885, "pid": 5, "tid": 7, "ts": 1716454217648956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619590, "dur": 6,
    "args": {
      "External id": 22885,
      "cbid": 211, "correlation": 22885
    }
  },
  {
    "ph": "s", "id": 22885, "pid": 76337, "tid": -914061504, "ts": 1716454217619590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619660, "dur": 1,
    "args": {
      "External id": 22896,
      "cbid": 251, "correlation": 22896
    }
  },
  {
    "ph": "f", "id": 22896, "pid": 76337, "tid": -914061504, "ts": 1716454217619660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619663, "dur": 0,
    "args": {
      "External id": 22897,
      "cbid": 251, "correlation": 22897
    }
  },
  {
    "ph": "f", "id": 22897, "pid": 76337, "tid": -914061504, "ts": 1716454217619663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217648963, "dur": 8,
    "args": {
      "External id": 22898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22898,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22898, "pid": 5, "tid": 7, "ts": 1716454217648963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619665, "dur": 13,
    "args": {
      "External id": 22898,
      "cbid": 211, "correlation": 22898
    }
  },
  {
    "ph": "s", "id": 22898, "pid": 76337, "tid": -914061504, "ts": 1716454217619665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217648973, "dur": 3,
    "args": {
      "External id": 22900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22900,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22900, "pid": 5, "tid": 7, "ts": 1716454217648973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619679, "dur": 6,
    "args": {
      "External id": 22900,
      "cbid": 211, "correlation": 22900
    }
  },
  {
    "ph": "s", "id": 22900, "pid": 76337, "tid": -914061504, "ts": 1716454217619679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217648977, "dur": 54,
    "args": {
      "External id": 22925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22925,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 22925, "pid": 5, "tid": 7, "ts": 1716454217648977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619757, "dur": 12,
    "args": {
      "External id": 22925,
      "cbid": 211, "correlation": 22925
    }
  },
  {
    "ph": "s", "id": 22925, "pid": 76337, "tid": -914061504, "ts": 1716454217619757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619877, "dur": 2,
    "args": {
      "External id": 22943,
      "cbid": 251, "correlation": 22943
    }
  },
  {
    "ph": "f", "id": 22943, "pid": 76337, "tid": -914061504, "ts": 1716454217619877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217649033, "dur": 90,
    "args": {
      "External id": 22945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22945,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 22945, "pid": 5, "tid": 7, "ts": 1716454217649033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619883, "dur": 16,
    "args": {
      "External id": 22945,
      "cbid": 211, "correlation": 22945
    }
  },
  {
    "ph": "s", "id": 22945, "pid": 76337, "tid": -914061504, "ts": 1716454217619883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217649124, "dur": 10,
    "args": {
      "External id": 22953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22953,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22953, "pid": 5, "tid": 7, "ts": 1716454217649124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217619958, "dur": 12,
    "args": {
      "External id": 22953,
      "cbid": 211, "correlation": 22953
    }
  },
  {
    "ph": "s", "id": 22953, "pid": 76337, "tid": -914061504, "ts": 1716454217619958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217649135, "dur": 21,
    "args": {
      "External id": 22961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22961, "pid": 5, "tid": 7, "ts": 1716454217649135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620008, "dur": 10,
    "args": {
      "External id": 22961,
      "cbid": 211, "correlation": 22961
    }
  },
  {
    "ph": "s", "id": 22961, "pid": 76337, "tid": -914061504, "ts": 1716454217620008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217649157, "dur": 18,
    "args": {
      "External id": 22983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 22983,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 22983, "pid": 5, "tid": 7, "ts": 1716454217649157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620060, "dur": 10,
    "args": {
      "External id": 22983,
      "cbid": 211, "correlation": 22983
    }
  },
  {
    "ph": "s", "id": 22983, "pid": 76337, "tid": -914061504, "ts": 1716454217620060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620151, "dur": 2,
    "args": {
      "External id": 22999,
      "cbid": 251, "correlation": 22999
    }
  },
  {
    "ph": "f", "id": 22999, "pid": 76337, "tid": -914061504, "ts": 1716454217620151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620156, "dur": 0,
    "args": {
      "External id": 23001,
      "cbid": 251, "correlation": 23001
    }
  },
  {
    "ph": "f", "id": 23001, "pid": 76337, "tid": -914061504, "ts": 1716454217620156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217649176, "dur": 489,
    "args": {
      "External id": 23002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23002,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23002, "pid": 5, "tid": 7, "ts": 1716454217649176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620159, "dur": 19,
    "args": {
      "External id": 23002,
      "cbid": 211, "correlation": 23002
    }
  },
  {
    "ph": "s", "id": 23002, "pid": 76337, "tid": -914061504, "ts": 1716454217620159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217649666, "dur": 66,
    "args": {
      "External id": 23010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23010,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23010, "pid": 5, "tid": 7, "ts": 1716454217649666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620232, "dur": 13,
    "args": {
      "External id": 23010,
      "cbid": 211, "correlation": 23010
    }
  },
  {
    "ph": "s", "id": 23010, "pid": 76337, "tid": -914061504, "ts": 1716454217620232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217649733, "dur": 68,
    "args": {
      "External id": 23018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23018, "pid": 5, "tid": 7, "ts": 1716454217649733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620263, "dur": 8,
    "args": {
      "External id": 23018,
      "cbid": 211, "correlation": 23018
    }
  },
  {
    "ph": "s", "id": 23018, "pid": 76337, "tid": -914061504, "ts": 1716454217620263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620355, "dur": 1,
    "args": {
      "External id": 23034,
      "cbid": 251, "correlation": 23034
    }
  },
  {
    "ph": "f", "id": 23034, "pid": 76337, "tid": -914061504, "ts": 1716454217620355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217649803, "dur": 1,
    "args": {
      "External id": 23036,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23036,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 23036, "pid": 5, "tid": 7, "ts": 1716454217649803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620360, "dur": 12,
    "args": {
      "External id": 23036,
      "cbid": 51, "correlation": 23036
    }
  },
  {
    "ph": "s", "id": 23036, "pid": 76337, "tid": -914061504, "ts": 1716454217620360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217649806, "dur": 267,
    "args": {
      "External id": 23037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23037,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23037, "pid": 5, "tid": 7, "ts": 1716454217649806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620374, "dur": 11,
    "args": {
      "External id": 23037,
      "cbid": 211, "correlation": 23037
    }
  },
  {
    "ph": "s", "id": 23037, "pid": 76337, "tid": -914061504, "ts": 1716454217620374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217650075, "dur": 14,
    "args": {
      "External id": 23045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23045,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23045, "pid": 5, "tid": 7, "ts": 1716454217650075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620417, "dur": 11,
    "args": {
      "External id": 23045,
      "cbid": 211, "correlation": 23045
    }
  },
  {
    "ph": "s", "id": 23045, "pid": 76337, "tid": -914061504, "ts": 1716454217620417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217650090, "dur": 38,
    "args": {
      "External id": 23056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23056,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23056, "pid": 5, "tid": 7, "ts": 1716454217650090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620485, "dur": 12,
    "args": {
      "External id": 23056,
      "cbid": 211, "correlation": 23056
    }
  },
  {
    "ph": "s", "id": 23056, "pid": 76337, "tid": -914061504, "ts": 1716454217620485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620564, "dur": 0,
    "args": {
      "External id": 23068,
      "cbid": 317, "correlation": 23068
    }
  },
  {
    "ph": "f", "id": 23068, "pid": 76337, "tid": -914061504, "ts": 1716454217620564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620564, "dur": 0,
    "args": {
      "External id": 23069,
      "cbid": 203, "correlation": 23069
    }
  },
  {
    "ph": "f", "id": 23069, "pid": 76337, "tid": -914061504, "ts": 1716454217620564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620565, "dur": 0,
    "args": {
      "External id": 23070,
      "cbid": 205, "correlation": 23070
    }
  },
  {
    "ph": "f", "id": 23070, "pid": 76337, "tid": -914061504, "ts": 1716454217620565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217650129, "dur": 13,
    "args": {
      "External id": 23074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23074,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23074, "pid": 5, "tid": 7, "ts": 1716454217650129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620586, "dur": 14,
    "args": {
      "External id": 23074,
      "cbid": 211, "correlation": 23074
    }
  },
  {
    "ph": "s", "id": 23074, "pid": 76337, "tid": -914061504, "ts": 1716454217620586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217650144, "dur": 4,
    "args": {
      "External id": 23076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23076,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23076, "pid": 5, "tid": 7, "ts": 1716454217650144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620604, "dur": 6,
    "args": {
      "External id": 23076,
      "cbid": 211, "correlation": 23076
    }
  },
  {
    "ph": "s", "id": 23076, "pid": 76337, "tid": -914061504, "ts": 1716454217620604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620613, "dur": 0,
    "args": {
      "External id": 23077,
      "cbid": 51, "correlation": 23077
    }
  },
  {
    "ph": "s", "id": 23077, "pid": 76337, "tid": -914061504, "ts": 1716454217620613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217650149, "dur": 96,
    "args": {
      "External id": 23078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23078,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 23078, "pid": 5, "tid": 7, "ts": 1716454217650149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620614, "dur": 5,
    "args": {
      "External id": 23078,
      "cbid": 211, "correlation": 23078
    }
  },
  {
    "ph": "s", "id": 23078, "pid": 76337, "tid": -914061504, "ts": 1716454217620614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217650246, "dur": 16,
    "args": {
      "External id": 23083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23083,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23083, "pid": 5, "tid": 7, "ts": 1716454217650246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620642, "dur": 9,
    "args": {
      "External id": 23083,
      "cbid": 211, "correlation": 23083
    }
  },
  {
    "ph": "s", "id": 23083, "pid": 76337, "tid": -914061504, "ts": 1716454217620642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217650264, "dur": 12,
    "args": {
      "External id": 23091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23091,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23091, "pid": 5, "tid": 7, "ts": 1716454217650264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620675, "dur": 8,
    "args": {
      "External id": 23091,
      "cbid": 211, "correlation": 23091
    }
  },
  {
    "ph": "s", "id": 23091, "pid": 76337, "tid": -914061504, "ts": 1716454217620675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217650277, "dur": 17,
    "args": {
      "External id": 23111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23111,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 23111, "pid": 5, "tid": 7, "ts": 1716454217650277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620748, "dur": 12,
    "args": {
      "External id": 23111,
      "cbid": 211, "correlation": 23111
    }
  },
  {
    "ph": "s", "id": 23111, "pid": 76337, "tid": -914061504, "ts": 1716454217620748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217650295, "dur": 4,
    "args": {
      "External id": 23123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23123,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 23123, "pid": 5, "tid": 7, "ts": 1716454217650295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620780, "dur": 8,
    "args": {
      "External id": 23123,
      "cbid": 211, "correlation": 23123
    }
  },
  {
    "ph": "s", "id": 23123, "pid": 76337, "tid": -914061504, "ts": 1716454217620780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217650301, "dur": 17,
    "args": {
      "External id": 23126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23126,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23126, "pid": 5, "tid": 7, "ts": 1716454217650301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620800, "dur": 7,
    "args": {
      "External id": 23126,
      "cbid": 211, "correlation": 23126
    }
  },
  {
    "ph": "s", "id": 23126, "pid": 76337, "tid": -914061504, "ts": 1716454217620800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217650319, "dur": 12,
    "args": {
      "External id": 23135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23135,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23135, "pid": 5, "tid": 7, "ts": 1716454217650319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620841, "dur": 10,
    "args": {
      "External id": 23135,
      "cbid": 211, "correlation": 23135
    }
  },
  {
    "ph": "s", "id": 23135, "pid": 76337, "tid": -914061504, "ts": 1716454217620841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620893, "dur": 0,
    "args": {
      "External id": 23145,
      "cbid": 317, "correlation": 23145
    }
  },
  {
    "ph": "f", "id": 23145, "pid": 76337, "tid": -914061504, "ts": 1716454217620893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620894, "dur": 0,
    "args": {
      "External id": 23146,
      "cbid": 203, "correlation": 23146
    }
  },
  {
    "ph": "f", "id": 23146, "pid": 76337, "tid": -914061504, "ts": 1716454217620894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620894, "dur": 0,
    "args": {
      "External id": 23147,
      "cbid": 205, "correlation": 23147
    }
  },
  {
    "ph": "f", "id": 23147, "pid": 76337, "tid": -914061504, "ts": 1716454217620894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217650333, "dur": 11,
    "args": {
      "External id": 23151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23151,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23151, "pid": 5, "tid": 7, "ts": 1716454217650333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620918, "dur": 12,
    "args": {
      "External id": 23151,
      "cbid": 211, "correlation": 23151
    }
  },
  {
    "ph": "s", "id": 23151, "pid": 76337, "tid": -914061504, "ts": 1716454217620918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217650345, "dur": 161,
    "args": {
      "External id": 23153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23153,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23153, "pid": 5, "tid": 7, "ts": 1716454217650345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620933, "dur": 5,
    "args": {
      "External id": 23153,
      "cbid": 211, "correlation": 23153
    }
  },
  {
    "ph": "s", "id": 23153, "pid": 76337, "tid": -914061504, "ts": 1716454217620933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217650509, "dur": 1,
    "args": {
      "External id": 23155,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23155,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 23155, "pid": 5, "tid": 7, "ts": 1716454217650509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620945, "dur": 8,
    "args": {
      "External id": 23155,
      "cbid": 51, "correlation": 23155
    }
  },
  {
    "ph": "s", "id": 23155, "pid": 76337, "tid": -914061504, "ts": 1716454217620945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217650512, "dur": 652,
    "args": {
      "External id": 23156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23156,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23156, "pid": 5, "tid": 7, "ts": 1716454217650512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620954, "dur": 7,
    "args": {
      "External id": 23156,
      "cbid": 211, "correlation": 23156
    }
  },
  {
    "ph": "s", "id": 23156, "pid": 76337, "tid": -914061504, "ts": 1716454217620954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217651166, "dur": 13,
    "args": {
      "External id": 23158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23158,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23158, "pid": 5, "tid": 7, "ts": 1716454217651166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217620964, "dur": 5,
    "args": {
      "External id": 23158,
      "cbid": 211, "correlation": 23158
    }
  },
  {
    "ph": "s", "id": 23158, "pid": 76337, "tid": -914061504, "ts": 1716454217620964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217651180, "dur": 14,
    "args": {
      "External id": 23164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23164, "pid": 5, "tid": 7, "ts": 1716454217651180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621013, "dur": 11,
    "args": {
      "External id": 23164,
      "cbid": 211, "correlation": 23164
    }
  },
  {
    "ph": "s", "id": 23164, "pid": 76337, "tid": -914061504, "ts": 1716454217621013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217651196, "dur": 3,
    "args": {
      "External id": 23172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23172,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 23172, "pid": 5, "tid": 7, "ts": 1716454217651196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621060, "dur": 10,
    "args": {
      "External id": 23172,
      "cbid": 211, "correlation": 23172
    }
  },
  {
    "ph": "s", "id": 23172, "pid": 76337, "tid": -914061504, "ts": 1716454217621060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621135, "dur": 1,
    "args": {
      "External id": 23188,
      "cbid": 251, "correlation": 23188
    }
  },
  {
    "ph": "f", "id": 23188, "pid": 76337, "tid": -914061504, "ts": 1716454217621135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621140, "dur": 0,
    "args": {
      "External id": 23190,
      "cbid": 251, "correlation": 23190
    }
  },
  {
    "ph": "f", "id": 23190, "pid": 76337, "tid": -914061504, "ts": 1716454217621140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217651200, "dur": 14,
    "args": {
      "External id": 23191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23191,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23191, "pid": 5, "tid": 7, "ts": 1716454217651200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621142, "dur": 12,
    "args": {
      "External id": 23191,
      "cbid": 211, "correlation": 23191
    }
  },
  {
    "ph": "s", "id": 23191, "pid": 76337, "tid": -914061504, "ts": 1716454217621142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217651215, "dur": 5,
    "args": {
      "External id": 23193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23193,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23193, "pid": 5, "tid": 7, "ts": 1716454217651215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621156, "dur": 6,
    "args": {
      "External id": 23193,
      "cbid": 211, "correlation": 23193
    }
  },
  {
    "ph": "s", "id": 23193, "pid": 76337, "tid": -914061504, "ts": 1716454217621156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217651222, "dur": 17,
    "args": {
      "External id": 23203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23203, "pid": 5, "tid": 7, "ts": 1716454217651222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621216, "dur": 12,
    "args": {
      "External id": 23203,
      "cbid": 211, "correlation": 23203
    }
  },
  {
    "ph": "s", "id": 23203, "pid": 76337, "tid": -914061504, "ts": 1716454217621216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217651240, "dur": 19,
    "args": {
      "External id": 23223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23223,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 23223, "pid": 5, "tid": 7, "ts": 1716454217651240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621281, "dur": 11,
    "args": {
      "External id": 23223,
      "cbid": 211, "correlation": 23223
    }
  },
  {
    "ph": "s", "id": 23223, "pid": 76337, "tid": -914061504, "ts": 1716454217621281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217651260, "dur": 4,
    "args": {
      "External id": 23235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23235,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 23235, "pid": 5, "tid": 7, "ts": 1716454217651260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621303, "dur": 6,
    "args": {
      "External id": 23235,
      "cbid": 211, "correlation": 23235
    }
  },
  {
    "ph": "s", "id": 23235, "pid": 76337, "tid": -914061504, "ts": 1716454217621303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217651266, "dur": 16,
    "args": {
      "External id": 23238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23238, "pid": 5, "tid": 7, "ts": 1716454217651266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621321, "dur": 6,
    "args": {
      "External id": 23238,
      "cbid": 211, "correlation": 23238
    }
  },
  {
    "ph": "s", "id": 23238, "pid": 76337, "tid": -914061504, "ts": 1716454217621321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217651283, "dur": 10,
    "args": {
      "External id": 23247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23247,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23247, "pid": 5, "tid": 7, "ts": 1716454217651283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621361, "dur": 10,
    "args": {
      "External id": 23247,
      "cbid": 211, "correlation": 23247
    }
  },
  {
    "ph": "s", "id": 23247, "pid": 76337, "tid": -914061504, "ts": 1716454217621361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621424, "dur": 0,
    "args": {
      "External id": 23257,
      "cbid": 317, "correlation": 23257
    }
  },
  {
    "ph": "f", "id": 23257, "pid": 76337, "tid": -914061504, "ts": 1716454217621424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621424, "dur": 0,
    "args": {
      "External id": 23258,
      "cbid": 203, "correlation": 23258
    }
  },
  {
    "ph": "f", "id": 23258, "pid": 76337, "tid": -914061504, "ts": 1716454217621424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621425, "dur": 0,
    "args": {
      "External id": 23259,
      "cbid": 205, "correlation": 23259
    }
  },
  {
    "ph": "f", "id": 23259, "pid": 76337, "tid": -914061504, "ts": 1716454217621425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217651295, "dur": 11,
    "args": {
      "External id": 23263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23263,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23263, "pid": 5, "tid": 7, "ts": 1716454217651295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621439, "dur": 12,
    "args": {
      "External id": 23263,
      "cbid": 211, "correlation": 23263
    }
  },
  {
    "ph": "s", "id": 23263, "pid": 76337, "tid": -914061504, "ts": 1716454217621439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217651307, "dur": 161,
    "args": {
      "External id": 23265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23265,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23265, "pid": 5, "tid": 7, "ts": 1716454217651307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621454, "dur": 5,
    "args": {
      "External id": 23265,
      "cbid": 211, "correlation": 23265
    }
  },
  {
    "ph": "s", "id": 23265, "pid": 76337, "tid": -914061504, "ts": 1716454217621454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217651470, "dur": 1,
    "args": {
      "External id": 23267,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23267,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 23267, "pid": 5, "tid": 7, "ts": 1716454217651470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621465, "dur": 6,
    "args": {
      "External id": 23267,
      "cbid": 51, "correlation": 23267
    }
  },
  {
    "ph": "s", "id": 23267, "pid": 76337, "tid": -914061504, "ts": 1716454217621465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217651474, "dur": 637,
    "args": {
      "External id": 23268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23268,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23268, "pid": 5, "tid": 7, "ts": 1716454217651474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621473, "dur": 6,
    "args": {
      "External id": 23268,
      "cbid": 211, "correlation": 23268
    }
  },
  {
    "ph": "s", "id": 23268, "pid": 76337, "tid": -914061504, "ts": 1716454217621473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217652112, "dur": 12,
    "args": {
      "External id": 23270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23270, "pid": 5, "tid": 7, "ts": 1716454217652112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621482, "dur": 5,
    "args": {
      "External id": 23270,
      "cbid": 211, "correlation": 23270
    }
  },
  {
    "ph": "s", "id": 23270, "pid": 76337, "tid": -914061504, "ts": 1716454217621482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217652126, "dur": 14,
    "args": {
      "External id": 23276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23276, "pid": 5, "tid": 7, "ts": 1716454217652126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621522, "dur": 11,
    "args": {
      "External id": 23276,
      "cbid": 211, "correlation": 23276
    }
  },
  {
    "ph": "s", "id": 23276, "pid": 76337, "tid": -914061504, "ts": 1716454217621522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217652141, "dur": 12,
    "args": {
      "External id": 23284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23284,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23284, "pid": 5, "tid": 7, "ts": 1716454217652141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621557, "dur": 8,
    "args": {
      "External id": 23284,
      "cbid": 211, "correlation": 23284
    }
  },
  {
    "ph": "s", "id": 23284, "pid": 76337, "tid": -914061504, "ts": 1716454217621557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217652154, "dur": 10,
    "args": {
      "External id": 23292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23292,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23292, "pid": 5, "tid": 7, "ts": 1716454217652154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621587, "dur": 9,
    "args": {
      "External id": 23292,
      "cbid": 211, "correlation": 23292
    }
  },
  {
    "ph": "s", "id": 23292, "pid": 76337, "tid": -914061504, "ts": 1716454217621587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217652166, "dur": 19,
    "args": {
      "External id": 23312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23312,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 23312, "pid": 5, "tid": 7, "ts": 1716454217652166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621666, "dur": 12,
    "args": {
      "External id": 23312,
      "cbid": 211, "correlation": 23312
    }
  },
  {
    "ph": "s", "id": 23312, "pid": 76337, "tid": -914061504, "ts": 1716454217621666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217652186, "dur": 4,
    "args": {
      "External id": 23324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23324,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 23324, "pid": 5, "tid": 7, "ts": 1716454217652186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621689, "dur": 6,
    "args": {
      "External id": 23324,
      "cbid": 211, "correlation": 23324
    }
  },
  {
    "ph": "s", "id": 23324, "pid": 76337, "tid": -914061504, "ts": 1716454217621689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217652192, "dur": 16,
    "args": {
      "External id": 23327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23327, "pid": 5, "tid": 7, "ts": 1716454217652192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621707, "dur": 6,
    "args": {
      "External id": 23327,
      "cbid": 211, "correlation": 23327
    }
  },
  {
    "ph": "s", "id": 23327, "pid": 76337, "tid": -914061504, "ts": 1716454217621707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621763, "dur": 0,
    "args": {
      "External id": 23338,
      "cbid": 317, "correlation": 23338
    }
  },
  {
    "ph": "f", "id": 23338, "pid": 76337, "tid": -914061504, "ts": 1716454217621763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621764, "dur": 0,
    "args": {
      "External id": 23339,
      "cbid": 203, "correlation": 23339
    }
  },
  {
    "ph": "f", "id": 23339, "pid": 76337, "tid": -914061504, "ts": 1716454217621764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621765, "dur": 0,
    "args": {
      "External id": 23340,
      "cbid": 205, "correlation": 23340
    }
  },
  {
    "ph": "f", "id": 23340, "pid": 76337, "tid": -914061504, "ts": 1716454217621765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217652209, "dur": 10,
    "args": {
      "External id": 23344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23344, "pid": 5, "tid": 7, "ts": 1716454217652209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621778, "dur": 11,
    "args": {
      "External id": 23344,
      "cbid": 211, "correlation": 23344
    }
  },
  {
    "ph": "s", "id": 23344, "pid": 76337, "tid": -914061504, "ts": 1716454217621778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217652221, "dur": 4,
    "args": {
      "External id": 23346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23346, "pid": 5, "tid": 7, "ts": 1716454217652221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621794, "dur": 6,
    "args": {
      "External id": 23346,
      "cbid": 211, "correlation": 23346
    }
  },
  {
    "ph": "s", "id": 23346, "pid": 76337, "tid": -914061504, "ts": 1716454217621794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621803, "dur": 0,
    "args": {
      "External id": 23347,
      "cbid": 51, "correlation": 23347
    }
  },
  {
    "ph": "s", "id": 23347, "pid": 76337, "tid": -914061504, "ts": 1716454217621803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217652226, "dur": 92,
    "args": {
      "External id": 23348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23348,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 23348, "pid": 5, "tid": 7, "ts": 1716454217652226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621803, "dur": 5,
    "args": {
      "External id": 23348,
      "cbid": 211, "correlation": 23348
    }
  },
  {
    "ph": "s", "id": 23348, "pid": 76337, "tid": -914061504, "ts": 1716454217621803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217652320, "dur": 15,
    "args": {
      "External id": 23353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23353, "pid": 5, "tid": 7, "ts": 1716454217652320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621831, "dur": 8,
    "args": {
      "External id": 23353,
      "cbid": 211, "correlation": 23353
    }
  },
  {
    "ph": "s", "id": 23353, "pid": 76337, "tid": -914061504, "ts": 1716454217621831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217652336, "dur": 83,
    "args": {
      "External id": 23362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23362,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23362, "pid": 5, "tid": 7, "ts": 1716454217652336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621913, "dur": 14,
    "args": {
      "External id": 23362,
      "cbid": 211, "correlation": 23362
    }
  },
  {
    "ph": "s", "id": 23362, "pid": 76337, "tid": -914061504, "ts": 1716454217621913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217652421, "dur": 30,
    "args": {
      "External id": 23384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23384,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23384, "pid": 5, "tid": 7, "ts": 1716454217652421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217621990, "dur": 13,
    "args": {
      "External id": 23384,
      "cbid": 211, "correlation": 23384
    }
  },
  {
    "ph": "s", "id": 23384, "pid": 76337, "tid": -914061504, "ts": 1716454217621990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622094, "dur": 1,
    "args": {
      "External id": 23395,
      "cbid": 251, "correlation": 23395
    }
  },
  {
    "ph": "f", "id": 23395, "pid": 76337, "tid": -914061504, "ts": 1716454217622094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217652452, "dur": 161,
    "args": {
      "External id": 23396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23396,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23396, "pid": 5, "tid": 7, "ts": 1716454217652452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622099, "dur": 15,
    "args": {
      "External id": 23396,
      "cbid": 211, "correlation": 23396
    }
  },
  {
    "ph": "s", "id": 23396, "pid": 76337, "tid": -914061504, "ts": 1716454217622099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622183, "dur": 1,
    "args": {
      "External id": 23407,
      "cbid": 251, "correlation": 23407
    }
  },
  {
    "ph": "f", "id": 23407, "pid": 76337, "tid": -914061504, "ts": 1716454217622183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217652614, "dur": 156,
    "args": {
      "External id": 23408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23408,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23408, "pid": 5, "tid": 7, "ts": 1716454217652614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622187, "dur": 12,
    "args": {
      "External id": 23408,
      "cbid": 211, "correlation": 23408
    }
  },
  {
    "ph": "s", "id": 23408, "pid": 76337, "tid": -914061504, "ts": 1716454217622187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622265, "dur": 1,
    "args": {
      "External id": 23419,
      "cbid": 251, "correlation": 23419
    }
  },
  {
    "ph": "f", "id": 23419, "pid": 76337, "tid": -914061504, "ts": 1716454217622265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217652772, "dur": 153,
    "args": {
      "External id": 23420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23420,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23420, "pid": 5, "tid": 7, "ts": 1716454217652772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622269, "dur": 13,
    "args": {
      "External id": 23420,
      "cbid": 211, "correlation": 23420
    }
  },
  {
    "ph": "s", "id": 23420, "pid": 76337, "tid": -914061504, "ts": 1716454217622269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217652926, "dur": 334,
    "args": {
      "External id": 23445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23445,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23445, "pid": 5, "tid": 7, "ts": 1716454217652926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622357, "dur": 12,
    "args": {
      "External id": 23445,
      "cbid": 211, "correlation": 23445
    }
  },
  {
    "ph": "s", "id": 23445, "pid": 76337, "tid": -914061504, "ts": 1716454217622357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622469, "dur": 1,
    "args": {
      "External id": 23463,
      "cbid": 251, "correlation": 23463
    }
  },
  {
    "ph": "f", "id": 23463, "pid": 76337, "tid": -914061504, "ts": 1716454217622469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217653261, "dur": 164,
    "args": {
      "External id": 23465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23465,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23465, "pid": 5, "tid": 7, "ts": 1716454217653261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622475, "dur": 14,
    "args": {
      "External id": 23465,
      "cbid": 211, "correlation": 23465
    }
  },
  {
    "ph": "s", "id": 23465, "pid": 76337, "tid": -914061504, "ts": 1716454217622475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217653426, "dur": 19,
    "args": {
      "External id": 23473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23473,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23473, "pid": 5, "tid": 7, "ts": 1716454217653426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622548, "dur": 12,
    "args": {
      "External id": 23473,
      "cbid": 211, "correlation": 23473
    }
  },
  {
    "ph": "s", "id": 23473, "pid": 76337, "tid": -914061504, "ts": 1716454217622548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217653446, "dur": 27,
    "args": {
      "External id": 23481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23481,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23481, "pid": 5, "tid": 7, "ts": 1716454217653446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622588, "dur": 9,
    "args": {
      "External id": 23481,
      "cbid": 211, "correlation": 23481
    }
  },
  {
    "ph": "s", "id": 23481, "pid": 76337, "tid": -914061504, "ts": 1716454217622588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217653475, "dur": 18,
    "args": {
      "External id": 23492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23492,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23492, "pid": 5, "tid": 7, "ts": 1716454217653475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622659, "dur": 12,
    "args": {
      "External id": 23492,
      "cbid": 211, "correlation": 23492
    }
  },
  {
    "ph": "s", "id": 23492, "pid": 76337, "tid": -914061504, "ts": 1716454217622659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217653494, "dur": 16,
    "args": {
      "External id": 23514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23514,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23514, "pid": 5, "tid": 7, "ts": 1716454217653494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622690, "dur": 7,
    "args": {
      "External id": 23514,
      "cbid": 211, "correlation": 23514
    }
  },
  {
    "ph": "s", "id": 23514, "pid": 76337, "tid": -914061504, "ts": 1716454217622690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622788, "dur": 1,
    "args": {
      "External id": 23525,
      "cbid": 251, "correlation": 23525
    }
  },
  {
    "ph": "f", "id": 23525, "pid": 76337, "tid": -914061504, "ts": 1716454217622788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217653511, "dur": 88,
    "args": {
      "External id": 23526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23526,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23526, "pid": 5, "tid": 7, "ts": 1716454217653511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622794, "dur": 15,
    "args": {
      "External id": 23526,
      "cbid": 211, "correlation": 23526
    }
  },
  {
    "ph": "s", "id": 23526, "pid": 76337, "tid": -914061504, "ts": 1716454217622794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622875, "dur": 1,
    "args": {
      "External id": 23537,
      "cbid": 251, "correlation": 23537
    }
  },
  {
    "ph": "f", "id": 23537, "pid": 76337, "tid": -914061504, "ts": 1716454217622875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622879, "dur": 0,
    "args": {
      "External id": 23538,
      "cbid": 251, "correlation": 23538
    }
  },
  {
    "ph": "f", "id": 23538, "pid": 76337, "tid": -914061504, "ts": 1716454217622879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217653600, "dur": 12,
    "args": {
      "External id": 23539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23539,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23539, "pid": 5, "tid": 7, "ts": 1716454217653600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622880, "dur": 12,
    "args": {
      "External id": 23539,
      "cbid": 211, "correlation": 23539
    }
  },
  {
    "ph": "s", "id": 23539, "pid": 76337, "tid": -914061504, "ts": 1716454217622880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217653614, "dur": 6,
    "args": {
      "External id": 23541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23541,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23541, "pid": 5, "tid": 7, "ts": 1716454217653614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622894, "dur": 6,
    "args": {
      "External id": 23541,
      "cbid": 211, "correlation": 23541
    }
  },
  {
    "ph": "s", "id": 23541, "pid": 76337, "tid": -914061504, "ts": 1716454217622894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622970, "dur": 1,
    "args": {
      "External id": 23552,
      "cbid": 251, "correlation": 23552
    }
  },
  {
    "ph": "f", "id": 23552, "pid": 76337, "tid": -914061504, "ts": 1716454217622970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622981, "dur": 0,
    "args": {
      "External id": 23553,
      "cbid": 251, "correlation": 23553
    }
  },
  {
    "ph": "f", "id": 23553, "pid": 76337, "tid": -914061504, "ts": 1716454217622981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217653621, "dur": 8,
    "args": {
      "External id": 23554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23554,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23554, "pid": 5, "tid": 7, "ts": 1716454217653621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622983, "dur": 13,
    "args": {
      "External id": 23554,
      "cbid": 211, "correlation": 23554
    }
  },
  {
    "ph": "s", "id": 23554, "pid": 76337, "tid": -914061504, "ts": 1716454217622983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217653630, "dur": 3,
    "args": {
      "External id": 23556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23556,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23556, "pid": 5, "tid": 7, "ts": 1716454217653630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217622998, "dur": 5,
    "args": {
      "External id": 23556,
      "cbid": 211, "correlation": 23556
    }
  },
  {
    "ph": "s", "id": 23556, "pid": 76337, "tid": -914061504, "ts": 1716454217622998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217653635, "dur": 54,
    "args": {
      "External id": 23581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23581,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23581, "pid": 5, "tid": 7, "ts": 1716454217653635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623077, "dur": 12,
    "args": {
      "External id": 23581,
      "cbid": 211, "correlation": 23581
    }
  },
  {
    "ph": "s", "id": 23581, "pid": 76337, "tid": -914061504, "ts": 1716454217623077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623188, "dur": 1,
    "args": {
      "External id": 23599,
      "cbid": 251, "correlation": 23599
    }
  },
  {
    "ph": "f", "id": 23599, "pid": 76337, "tid": -914061504, "ts": 1716454217623188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217653689, "dur": 89,
    "args": {
      "External id": 23601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23601,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23601, "pid": 5, "tid": 7, "ts": 1716454217653689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623194, "dur": 14,
    "args": {
      "External id": 23601,
      "cbid": 211, "correlation": 23601
    }
  },
  {
    "ph": "s", "id": 23601, "pid": 76337, "tid": -914061504, "ts": 1716454217623194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217653780, "dur": 10,
    "args": {
      "External id": 23609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23609,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23609, "pid": 5, "tid": 7, "ts": 1716454217653780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623265, "dur": 13,
    "args": {
      "External id": 23609,
      "cbid": 211, "correlation": 23609
    }
  },
  {
    "ph": "s", "id": 23609, "pid": 76337, "tid": -914061504, "ts": 1716454217623265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217653791, "dur": 22,
    "args": {
      "External id": 23617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23617,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23617, "pid": 5, "tid": 7, "ts": 1716454217653791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623307, "dur": 9,
    "args": {
      "External id": 23617,
      "cbid": 211, "correlation": 23617
    }
  },
  {
    "ph": "s", "id": 23617, "pid": 76337, "tid": -914061504, "ts": 1716454217623307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217653814, "dur": 18,
    "args": {
      "External id": 23639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23639,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23639, "pid": 5, "tid": 7, "ts": 1716454217653814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623358, "dur": 10,
    "args": {
      "External id": 23639,
      "cbid": 211, "correlation": 23639
    }
  },
  {
    "ph": "s", "id": 23639, "pid": 76337, "tid": -914061504, "ts": 1716454217623358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623444, "dur": 1,
    "args": {
      "External id": 23655,
      "cbid": 251, "correlation": 23655
    }
  },
  {
    "ph": "f", "id": 23655, "pid": 76337, "tid": -914061504, "ts": 1716454217623444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623449, "dur": 0,
    "args": {
      "External id": 23657,
      "cbid": 251, "correlation": 23657
    }
  },
  {
    "ph": "f", "id": 23657, "pid": 76337, "tid": -914061504, "ts": 1716454217623449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217653833, "dur": 492,
    "args": {
      "External id": 23658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23658,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23658, "pid": 5, "tid": 7, "ts": 1716454217653833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623452, "dur": 13,
    "args": {
      "External id": 23658,
      "cbid": 211, "correlation": 23658
    }
  },
  {
    "ph": "s", "id": 23658, "pid": 76337, "tid": -914061504, "ts": 1716454217623452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217654327, "dur": 65,
    "args": {
      "External id": 23666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23666,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23666, "pid": 5, "tid": 7, "ts": 1716454217654327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623518, "dur": 13,
    "args": {
      "External id": 23666,
      "cbid": 211, "correlation": 23666
    }
  },
  {
    "ph": "s", "id": 23666, "pid": 76337, "tid": -914061504, "ts": 1716454217623518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217654394, "dur": 66,
    "args": {
      "External id": 23674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23674, "pid": 5, "tid": 7, "ts": 1716454217654394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623548, "dur": 8,
    "args": {
      "External id": 23674,
      "cbid": 211, "correlation": 23674
    }
  },
  {
    "ph": "s", "id": 23674, "pid": 76337, "tid": -914061504, "ts": 1716454217623548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623628, "dur": 1,
    "args": {
      "External id": 23690,
      "cbid": 251, "correlation": 23690
    }
  },
  {
    "ph": "f", "id": 23690, "pid": 76337, "tid": -914061504, "ts": 1716454217623628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217654463, "dur": 1,
    "args": {
      "External id": 23692,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23692,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 23692, "pid": 5, "tid": 7, "ts": 1716454217654463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623633, "dur": 9,
    "args": {
      "External id": 23692,
      "cbid": 51, "correlation": 23692
    }
  },
  {
    "ph": "s", "id": 23692, "pid": 76337, "tid": -914061504, "ts": 1716454217623633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217654466, "dur": 268,
    "args": {
      "External id": 23693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23693,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23693, "pid": 5, "tid": 7, "ts": 1716454217654466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623643, "dur": 11,
    "args": {
      "External id": 23693,
      "cbid": 211, "correlation": 23693
    }
  },
  {
    "ph": "s", "id": 23693, "pid": 76337, "tid": -914061504, "ts": 1716454217623643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217654736, "dur": 13,
    "args": {
      "External id": 23701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23701,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23701, "pid": 5, "tid": 7, "ts": 1716454217654736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623686, "dur": 10,
    "args": {
      "External id": 23701,
      "cbid": 211, "correlation": 23701
    }
  },
  {
    "ph": "s", "id": 23701, "pid": 76337, "tid": -914061504, "ts": 1716454217623686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217654751, "dur": 38,
    "args": {
      "External id": 23712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23712, "pid": 5, "tid": 7, "ts": 1716454217654751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623755, "dur": 12,
    "args": {
      "External id": 23712,
      "cbid": 211, "correlation": 23712
    }
  },
  {
    "ph": "s", "id": 23712, "pid": 76337, "tid": -914061504, "ts": 1716454217623755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623819, "dur": 0,
    "args": {
      "External id": 23724,
      "cbid": 317, "correlation": 23724
    }
  },
  {
    "ph": "f", "id": 23724, "pid": 76337, "tid": -914061504, "ts": 1716454217623819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623820, "dur": 0,
    "args": {
      "External id": 23725,
      "cbid": 203, "correlation": 23725
    }
  },
  {
    "ph": "f", "id": 23725, "pid": 76337, "tid": -914061504, "ts": 1716454217623820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623821, "dur": 0,
    "args": {
      "External id": 23726,
      "cbid": 205, "correlation": 23726
    }
  },
  {
    "ph": "f", "id": 23726, "pid": 76337, "tid": -914061504, "ts": 1716454217623821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217654790, "dur": 14,
    "args": {
      "External id": 23730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23730,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23730, "pid": 5, "tid": 7, "ts": 1716454217654790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623836, "dur": 12,
    "args": {
      "External id": 23730,
      "cbid": 211, "correlation": 23730
    }
  },
  {
    "ph": "s", "id": 23730, "pid": 76337, "tid": -914061504, "ts": 1716454217623836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217654805, "dur": 4,
    "args": {
      "External id": 23732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23732,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 23732, "pid": 5, "tid": 7, "ts": 1716454217654805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623852, "dur": 6,
    "args": {
      "External id": 23732,
      "cbid": 211, "correlation": 23732
    }
  },
  {
    "ph": "s", "id": 23732, "pid": 76337, "tid": -914061504, "ts": 1716454217623852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623862, "dur": 0,
    "args": {
      "External id": 23733,
      "cbid": 51, "correlation": 23733
    }
  },
  {
    "ph": "s", "id": 23733, "pid": 76337, "tid": -914061504, "ts": 1716454217623862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217654810, "dur": 96,
    "args": {
      "External id": 23734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23734,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 23734, "pid": 5, "tid": 7, "ts": 1716454217654810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623863, "dur": 5,
    "args": {
      "External id": 23734,
      "cbid": 211, "correlation": 23734
    }
  },
  {
    "ph": "s", "id": 23734, "pid": 76337, "tid": -914061504, "ts": 1716454217623863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217654907, "dur": 16,
    "args": {
      "External id": 23739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23739, "pid": 5, "tid": 7, "ts": 1716454217654907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623889, "dur": 8,
    "args": {
      "External id": 23739,
      "cbid": 211, "correlation": 23739
    }
  },
  {
    "ph": "s", "id": 23739, "pid": 76337, "tid": -914061504, "ts": 1716454217623889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217654925, "dur": 12,
    "args": {
      "External id": 23747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23747,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23747, "pid": 5, "tid": 7, "ts": 1716454217654925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217623921, "dur": 8,
    "args": {
      "External id": 23747,
      "cbid": 211, "correlation": 23747
    }
  },
  {
    "ph": "s", "id": 23747, "pid": 76337, "tid": -914061504, "ts": 1716454217623921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624000, "dur": 0,
    "args": {
      "External id": 23757,
      "cbid": 317, "correlation": 23757
    }
  },
  {
    "ph": "f", "id": 23757, "pid": 76337, "tid": -914061504, "ts": 1716454217624000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624001, "dur": 0,
    "args": {
      "External id": 23758,
      "cbid": 203, "correlation": 23758
    }
  },
  {
    "ph": "f", "id": 23758, "pid": 76337, "tid": -914061504, "ts": 1716454217624001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624001, "dur": 0,
    "args": {
      "External id": 23759,
      "cbid": 205, "correlation": 23759
    }
  },
  {
    "ph": "f", "id": 23759, "pid": 76337, "tid": -914061504, "ts": 1716454217624001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217654938, "dur": 12,
    "args": {
      "External id": 23763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23763,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23763, "pid": 5, "tid": 7, "ts": 1716454217654938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624016, "dur": 13,
    "args": {
      "External id": 23763,
      "cbid": 211, "correlation": 23763
    }
  },
  {
    "ph": "s", "id": 23763, "pid": 76337, "tid": -914061504, "ts": 1716454217624016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217654951, "dur": 161,
    "args": {
      "External id": 23765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23765,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23765, "pid": 5, "tid": 7, "ts": 1716454217654951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624032, "dur": 6,
    "args": {
      "External id": 23765,
      "cbid": 211, "correlation": 23765
    }
  },
  {
    "ph": "s", "id": 23765, "pid": 76337, "tid": -914061504, "ts": 1716454217624032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217655114, "dur": 1,
    "args": {
      "External id": 23767,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23767,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 23767, "pid": 5, "tid": 7, "ts": 1716454217655114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624044, "dur": 6,
    "args": {
      "External id": 23767,
      "cbid": 51, "correlation": 23767
    }
  },
  {
    "ph": "s", "id": 23767, "pid": 76337, "tid": -914061504, "ts": 1716454217624044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217655118, "dur": 195,
    "args": {
      "External id": 23768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23768,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 23768, "pid": 5, "tid": 7, "ts": 1716454217655118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624061, "dur": 18,
    "args": {
      "External id": 23768,
      "cbid": 211, "correlation": 23768
    }
  },
  {
    "ph": "s", "id": 23768, "pid": 76337, "tid": -914061504, "ts": 1716454217624061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655314, "dur": 6,
    "args": {
      "External id": 23770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23770, "pid": 5, "tid": 7, "ts": 1716454217655314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624085, "dur": 7,
    "args": {
      "External id": 23770,
      "cbid": 211, "correlation": 23770
    }
  },
  {
    "ph": "s", "id": 23770, "pid": 76337, "tid": -914061504, "ts": 1716454217624085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217655321, "dur": 6,
    "args": {
      "External id": 23776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23776, "pid": 5, "tid": 7, "ts": 1716454217655321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624117, "dur": 9,
    "args": {
      "External id": 23776,
      "cbid": 211, "correlation": 23776
    }
  },
  {
    "ph": "s", "id": 23776, "pid": 76337, "tid": -914061504, "ts": 1716454217624117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655329, "dur": 10,
    "args": {
      "External id": 23796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23796,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 23796, "pid": 5, "tid": 7, "ts": 1716454217655329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624249, "dur": 14,
    "args": {
      "External id": 23796,
      "cbid": 211, "correlation": 23796
    }
  },
  {
    "ph": "s", "id": 23796, "pid": 76337, "tid": -914061504, "ts": 1716454217624249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217655340, "dur": 4,
    "args": {
      "External id": 23808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23808,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 23808, "pid": 5, "tid": 7, "ts": 1716454217655340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624273, "dur": 7,
    "args": {
      "External id": 23808,
      "cbid": 211, "correlation": 23808
    }
  },
  {
    "ph": "s", "id": 23808, "pid": 76337, "tid": -914061504, "ts": 1716454217624273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217655346, "dur": 8,
    "args": {
      "External id": 23811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23811,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23811, "pid": 5, "tid": 7, "ts": 1716454217655346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624293, "dur": 6,
    "args": {
      "External id": 23811,
      "cbid": 211, "correlation": 23811
    }
  },
  {
    "ph": "s", "id": 23811, "pid": 76337, "tid": -914061504, "ts": 1716454217624293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217655355, "dur": 5,
    "args": {
      "External id": 23820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23820,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23820, "pid": 5, "tid": 7, "ts": 1716454217655355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624333, "dur": 10,
    "args": {
      "External id": 23820,
      "cbid": 211, "correlation": 23820
    }
  },
  {
    "ph": "s", "id": 23820, "pid": 76337, "tid": -914061504, "ts": 1716454217624333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624387, "dur": 0,
    "args": {
      "External id": 23830,
      "cbid": 317, "correlation": 23830
    }
  },
  {
    "ph": "f", "id": 23830, "pid": 76337, "tid": -914061504, "ts": 1716454217624387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624388, "dur": 0,
    "args": {
      "External id": 23831,
      "cbid": 203, "correlation": 23831
    }
  },
  {
    "ph": "f", "id": 23831, "pid": 76337, "tid": -914061504, "ts": 1716454217624388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624388, "dur": 0,
    "args": {
      "External id": 23832,
      "cbid": 205, "correlation": 23832
    }
  },
  {
    "ph": "f", "id": 23832, "pid": 76337, "tid": -914061504, "ts": 1716454217624388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655362, "dur": 5,
    "args": {
      "External id": 23836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23836,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23836, "pid": 5, "tid": 7, "ts": 1716454217655362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624403, "dur": 11,
    "args": {
      "External id": 23836,
      "cbid": 211, "correlation": 23836
    }
  },
  {
    "ph": "s", "id": 23836, "pid": 76337, "tid": -914061504, "ts": 1716454217624403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655368, "dur": 160,
    "args": {
      "External id": 23838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23838, "pid": 5, "tid": 7, "ts": 1716454217655368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624417, "dur": 5,
    "args": {
      "External id": 23838,
      "cbid": 211, "correlation": 23838
    }
  },
  {
    "ph": "s", "id": 23838, "pid": 76337, "tid": -914061504, "ts": 1716454217624417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217655530, "dur": 1,
    "args": {
      "External id": 23840,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23840,
      "bytes": 240, "memory bandwidth (GB/s)": 0.125
    }
  },
  {
    "ph": "f", "id": 23840, "pid": 5, "tid": 7, "ts": 1716454217655530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624428, "dur": 7,
    "args": {
      "External id": 23840,
      "cbid": 51, "correlation": 23840
    }
  },
  {
    "ph": "s", "id": 23840, "pid": 76337, "tid": -914061504, "ts": 1716454217624428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217655534, "dur": 265,
    "args": {
      "External id": 23841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23841,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23841, "pid": 5, "tid": 7, "ts": 1716454217655534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624436, "dur": 6,
    "args": {
      "External id": 23841,
      "cbid": 211, "correlation": 23841
    }
  },
  {
    "ph": "s", "id": 23841, "pid": 76337, "tid": -914061504, "ts": 1716454217624436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655801, "dur": 6,
    "args": {
      "External id": 23843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23843,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23843, "pid": 5, "tid": 7, "ts": 1716454217655801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624446, "dur": 5,
    "args": {
      "External id": 23843,
      "cbid": 211, "correlation": 23843
    }
  },
  {
    "ph": "s", "id": 23843, "pid": 76337, "tid": -914061504, "ts": 1716454217624446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217655808, "dur": 6,
    "args": {
      "External id": 23849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23849, "pid": 5, "tid": 7, "ts": 1716454217655808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624474, "dur": 8,
    "args": {
      "External id": 23849,
      "cbid": 211, "correlation": 23849
    }
  },
  {
    "ph": "s", "id": 23849, "pid": 76337, "tid": -914061504, "ts": 1716454217624474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217655815, "dur": 3,
    "args": {
      "External id": 23857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23857,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 23857, "pid": 5, "tid": 7, "ts": 1716454217655815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624518, "dur": 9,
    "args": {
      "External id": 23857,
      "cbid": 211, "correlation": 23857
    }
  },
  {
    "ph": "s", "id": 23857, "pid": 76337, "tid": -914061504, "ts": 1716454217624518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624600, "dur": 1,
    "args": {
      "External id": 23873,
      "cbid": 251, "correlation": 23873
    }
  },
  {
    "ph": "f", "id": 23873, "pid": 76337, "tid": -914061504, "ts": 1716454217624600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624606, "dur": 0,
    "args": {
      "External id": 23875,
      "cbid": 251, "correlation": 23875
    }
  },
  {
    "ph": "f", "id": 23875, "pid": 76337, "tid": -914061504, "ts": 1716454217624606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217655820, "dur": 13,
    "args": {
      "External id": 23876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23876,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23876, "pid": 5, "tid": 7, "ts": 1716454217655820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624608, "dur": 13,
    "args": {
      "External id": 23876,
      "cbid": 211, "correlation": 23876
    }
  },
  {
    "ph": "s", "id": 23876, "pid": 76337, "tid": -914061504, "ts": 1716454217624608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217655834, "dur": 5,
    "args": {
      "External id": 23878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23878,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23878, "pid": 5, "tid": 7, "ts": 1716454217655834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624623, "dur": 6,
    "args": {
      "External id": 23878,
      "cbid": 211, "correlation": 23878
    }
  },
  {
    "ph": "s", "id": 23878, "pid": 76337, "tid": -914061504, "ts": 1716454217624623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217655841, "dur": 6,
    "args": {
      "External id": 23888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23888, "pid": 5, "tid": 7, "ts": 1716454217655841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624682, "dur": 12,
    "args": {
      "External id": 23888,
      "cbid": 211, "correlation": 23888
    }
  },
  {
    "ph": "s", "id": 23888, "pid": 76337, "tid": -914061504, "ts": 1716454217624682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655847, "dur": 10,
    "args": {
      "External id": 23908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23908,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 23908, "pid": 5, "tid": 7, "ts": 1716454217655847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624748, "dur": 11,
    "args": {
      "External id": 23908,
      "cbid": 211, "correlation": 23908
    }
  },
  {
    "ph": "s", "id": 23908, "pid": 76337, "tid": -914061504, "ts": 1716454217624748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217655858, "dur": 4,
    "args": {
      "External id": 23920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23920,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 23920, "pid": 5, "tid": 7, "ts": 1716454217655858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624769, "dur": 6,
    "args": {
      "External id": 23920,
      "cbid": 211, "correlation": 23920
    }
  },
  {
    "ph": "s", "id": 23920, "pid": 76337, "tid": -914061504, "ts": 1716454217624769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217655863, "dur": 7,
    "args": {
      "External id": 23923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23923, "pid": 5, "tid": 7, "ts": 1716454217655863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624788, "dur": 6,
    "args": {
      "External id": 23923,
      "cbid": 211, "correlation": 23923
    }
  },
  {
    "ph": "s", "id": 23923, "pid": 76337, "tid": -914061504, "ts": 1716454217624788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217655872, "dur": 4,
    "args": {
      "External id": 23932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23932,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23932, "pid": 5, "tid": 7, "ts": 1716454217655872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624827, "dur": 10,
    "args": {
      "External id": 23932,
      "cbid": 211, "correlation": 23932
    }
  },
  {
    "ph": "s", "id": 23932, "pid": 76337, "tid": -914061504, "ts": 1716454217624827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624905, "dur": 0,
    "args": {
      "External id": 23942,
      "cbid": 317, "correlation": 23942
    }
  },
  {
    "ph": "f", "id": 23942, "pid": 76337, "tid": -914061504, "ts": 1716454217624905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624906, "dur": 0,
    "args": {
      "External id": 23943,
      "cbid": 203, "correlation": 23943
    }
  },
  {
    "ph": "f", "id": 23943, "pid": 76337, "tid": -914061504, "ts": 1716454217624906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624907, "dur": 0,
    "args": {
      "External id": 23944,
      "cbid": 205, "correlation": 23944
    }
  },
  {
    "ph": "f", "id": 23944, "pid": 76337, "tid": -914061504, "ts": 1716454217624907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655877, "dur": 5,
    "args": {
      "External id": 23948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23948, "pid": 5, "tid": 7, "ts": 1716454217655877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624920, "dur": 13,
    "args": {
      "External id": 23948,
      "cbid": 211, "correlation": 23948
    }
  },
  {
    "ph": "s", "id": 23948, "pid": 76337, "tid": -914061504, "ts": 1716454217624920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217655884, "dur": 160,
    "args": {
      "External id": 23950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23950,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23950, "pid": 5, "tid": 7, "ts": 1716454217655884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624936, "dur": 5,
    "args": {
      "External id": 23950,
      "cbid": 211, "correlation": 23950
    }
  },
  {
    "ph": "s", "id": 23950, "pid": 76337, "tid": -914061504, "ts": 1716454217624936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217656046, "dur": 1,
    "args": {
      "External id": 23952,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 23952,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 23952, "pid": 5, "tid": 7, "ts": 1716454217656046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624946, "dur": 6,
    "args": {
      "External id": 23952,
      "cbid": 51, "correlation": 23952
    }
  },
  {
    "ph": "s", "id": 23952, "pid": 76337, "tid": -914061504, "ts": 1716454217624946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217656050, "dur": 254,
    "args": {
      "External id": 23953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23953,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 23953, "pid": 5, "tid": 7, "ts": 1716454217656050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624954, "dur": 7,
    "args": {
      "External id": 23953,
      "cbid": 211, "correlation": 23953
    }
  },
  {
    "ph": "s", "id": 23953, "pid": 76337, "tid": -914061504, "ts": 1716454217624954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656305, "dur": 6,
    "args": {
      "External id": 23955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23955,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 23955, "pid": 5, "tid": 7, "ts": 1716454217656305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217624964, "dur": 5,
    "args": {
      "External id": 23955,
      "cbid": 211, "correlation": 23955
    }
  },
  {
    "ph": "s", "id": 23955, "pid": 76337, "tid": -914061504, "ts": 1716454217624964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217656313, "dur": 6,
    "args": {
      "External id": 23961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23961, "pid": 5, "tid": 7, "ts": 1716454217656313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625001, "dur": 10,
    "args": {
      "External id": 23961,
      "cbid": 211, "correlation": 23961
    }
  },
  {
    "ph": "s", "id": 23961, "pid": 76337, "tid": -914061504, "ts": 1716454217625001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217656320, "dur": 5,
    "args": {
      "External id": 23969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23969,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23969, "pid": 5, "tid": 7, "ts": 1716454217656320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625035, "dur": 8,
    "args": {
      "External id": 23969,
      "cbid": 211, "correlation": 23969
    }
  },
  {
    "ph": "s", "id": 23969, "pid": 76337, "tid": -914061504, "ts": 1716454217625035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217656327, "dur": 4,
    "args": {
      "External id": 23977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23977,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 23977, "pid": 5, "tid": 7, "ts": 1716454217656327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625064, "dur": 9,
    "args": {
      "External id": 23977,
      "cbid": 211, "correlation": 23977
    }
  },
  {
    "ph": "s", "id": 23977, "pid": 76337, "tid": -914061504, "ts": 1716454217625064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656332, "dur": 9,
    "args": {
      "External id": 23997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 23997,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 23997, "pid": 5, "tid": 7, "ts": 1716454217656332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625139, "dur": 12,
    "args": {
      "External id": 23997,
      "cbid": 211, "correlation": 23997
    }
  },
  {
    "ph": "s", "id": 23997, "pid": 76337, "tid": -914061504, "ts": 1716454217625139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217656343, "dur": 4,
    "args": {
      "External id": 24009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24009,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24009, "pid": 5, "tid": 7, "ts": 1716454217656343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625161, "dur": 6,
    "args": {
      "External id": 24009,
      "cbid": 211, "correlation": 24009
    }
  },
  {
    "ph": "s", "id": 24009, "pid": 76337, "tid": -914061504, "ts": 1716454217625161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217656348, "dur": 6,
    "args": {
      "External id": 24012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24012,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24012, "pid": 5, "tid": 7, "ts": 1716454217656348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625179, "dur": 7,
    "args": {
      "External id": 24012,
      "cbid": 211, "correlation": 24012
    }
  },
  {
    "ph": "s", "id": 24012, "pid": 76337, "tid": -914061504, "ts": 1716454217625179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217656355, "dur": 4,
    "args": {
      "External id": 24021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24021,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24021, "pid": 5, "tid": 7, "ts": 1716454217656355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625218, "dur": 9,
    "args": {
      "External id": 24021,
      "cbid": 211, "correlation": 24021
    }
  },
  {
    "ph": "s", "id": 24021, "pid": 76337, "tid": -914061504, "ts": 1716454217625218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625268, "dur": 0,
    "args": {
      "External id": 24031,
      "cbid": 317, "correlation": 24031
    }
  },
  {
    "ph": "f", "id": 24031, "pid": 76337, "tid": -914061504, "ts": 1716454217625268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625269, "dur": 0,
    "args": {
      "External id": 24032,
      "cbid": 203, "correlation": 24032
    }
  },
  {
    "ph": "f", "id": 24032, "pid": 76337, "tid": -914061504, "ts": 1716454217625269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625270, "dur": 0,
    "args": {
      "External id": 24033,
      "cbid": 205, "correlation": 24033
    }
  },
  {
    "ph": "f", "id": 24033, "pid": 76337, "tid": -914061504, "ts": 1716454217625270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656361, "dur": 5,
    "args": {
      "External id": 24037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24037,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24037, "pid": 5, "tid": 7, "ts": 1716454217656361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625283, "dur": 11,
    "args": {
      "External id": 24037,
      "cbid": 211, "correlation": 24037
    }
  },
  {
    "ph": "s", "id": 24037, "pid": 76337, "tid": -914061504, "ts": 1716454217625283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656367, "dur": 160,
    "args": {
      "External id": 24039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24039,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24039, "pid": 5, "tid": 7, "ts": 1716454217656367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625297, "dur": 6,
    "args": {
      "External id": 24039,
      "cbid": 211, "correlation": 24039
    }
  },
  {
    "ph": "s", "id": 24039, "pid": 76337, "tid": -914061504, "ts": 1716454217625297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217656529, "dur": 1,
    "args": {
      "External id": 24041,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24041,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 24041, "pid": 5, "tid": 7, "ts": 1716454217656529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625308, "dur": 6,
    "args": {
      "External id": 24041,
      "cbid": 51, "correlation": 24041
    }
  },
  {
    "ph": "s", "id": 24041, "pid": 76337, "tid": -914061504, "ts": 1716454217625308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217656533, "dur": 254,
    "args": {
      "External id": 24042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24042,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24042, "pid": 5, "tid": 7, "ts": 1716454217656533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625315, "dur": 6,
    "args": {
      "External id": 24042,
      "cbid": 211, "correlation": 24042
    }
  },
  {
    "ph": "s", "id": 24042, "pid": 76337, "tid": -914061504, "ts": 1716454217625315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656788, "dur": 6,
    "args": {
      "External id": 24044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24044,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24044, "pid": 5, "tid": 7, "ts": 1716454217656788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625324, "dur": 5,
    "args": {
      "External id": 24044,
      "cbid": 211, "correlation": 24044
    }
  },
  {
    "ph": "s", "id": 24044, "pid": 76337, "tid": -914061504, "ts": 1716454217625324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217656795, "dur": 6,
    "args": {
      "External id": 24050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24050, "pid": 5, "tid": 7, "ts": 1716454217656795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625352, "dur": 8,
    "args": {
      "External id": 24050,
      "cbid": 211, "correlation": 24050
    }
  },
  {
    "ph": "s", "id": 24050, "pid": 76337, "tid": -914061504, "ts": 1716454217625352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217656802, "dur": 3,
    "args": {
      "External id": 24058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24058,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 24058, "pid": 5, "tid": 7, "ts": 1716454217656802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625395, "dur": 10,
    "args": {
      "External id": 24058,
      "cbid": 211, "correlation": 24058
    }
  },
  {
    "ph": "s", "id": 24058, "pid": 76337, "tid": -914061504, "ts": 1716454217625395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625468, "dur": 1,
    "args": {
      "External id": 24074,
      "cbid": 251, "correlation": 24074
    }
  },
  {
    "ph": "f", "id": 24074, "pid": 76337, "tid": -914061504, "ts": 1716454217625468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625473, "dur": 0,
    "args": {
      "External id": 24076,
      "cbid": 251, "correlation": 24076
    }
  },
  {
    "ph": "f", "id": 24076, "pid": 76337, "tid": -914061504, "ts": 1716454217625473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217656807, "dur": 10,
    "args": {
      "External id": 24077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24077,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24077, "pid": 5, "tid": 7, "ts": 1716454217656807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625475, "dur": 12,
    "args": {
      "External id": 24077,
      "cbid": 211, "correlation": 24077
    }
  },
  {
    "ph": "s", "id": 24077, "pid": 76337, "tid": -914061504, "ts": 1716454217625475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217656818, "dur": 4,
    "args": {
      "External id": 24079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24079,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24079, "pid": 5, "tid": 7, "ts": 1716454217656818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625489, "dur": 6,
    "args": {
      "External id": 24079,
      "cbid": 211, "correlation": 24079
    }
  },
  {
    "ph": "s", "id": 24079, "pid": 76337, "tid": -914061504, "ts": 1716454217625489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217656823, "dur": 6,
    "args": {
      "External id": 24089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24089, "pid": 5, "tid": 7, "ts": 1716454217656823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625547, "dur": 12,
    "args": {
      "External id": 24089,
      "cbid": 211, "correlation": 24089
    }
  },
  {
    "ph": "s", "id": 24089, "pid": 76337, "tid": -914061504, "ts": 1716454217625547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656830, "dur": 9,
    "args": {
      "External id": 24109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24109,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 24109, "pid": 5, "tid": 7, "ts": 1716454217656830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625612, "dur": 11,
    "args": {
      "External id": 24109,
      "cbid": 211, "correlation": 24109
    }
  },
  {
    "ph": "s", "id": 24109, "pid": 76337, "tid": -914061504, "ts": 1716454217625612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217656841, "dur": 4,
    "args": {
      "External id": 24121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24121,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24121, "pid": 5, "tid": 7, "ts": 1716454217656841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625633, "dur": 6,
    "args": {
      "External id": 24121,
      "cbid": 211, "correlation": 24121
    }
  },
  {
    "ph": "s", "id": 24121, "pid": 76337, "tid": -914061504, "ts": 1716454217625633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217656846, "dur": 7,
    "args": {
      "External id": 24124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24124,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24124, "pid": 5, "tid": 7, "ts": 1716454217656846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625650, "dur": 6,
    "args": {
      "External id": 24124,
      "cbid": 211, "correlation": 24124
    }
  },
  {
    "ph": "s", "id": 24124, "pid": 76337, "tid": -914061504, "ts": 1716454217625650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217656853, "dur": 4,
    "args": {
      "External id": 24133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24133,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24133, "pid": 5, "tid": 7, "ts": 1716454217656853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625691, "dur": 9,
    "args": {
      "External id": 24133,
      "cbid": 211, "correlation": 24133
    }
  },
  {
    "ph": "s", "id": 24133, "pid": 76337, "tid": -914061504, "ts": 1716454217625691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625753, "dur": 0,
    "args": {
      "External id": 24143,
      "cbid": 317, "correlation": 24143
    }
  },
  {
    "ph": "f", "id": 24143, "pid": 76337, "tid": -914061504, "ts": 1716454217625753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625754, "dur": 0,
    "args": {
      "External id": 24144,
      "cbid": 203, "correlation": 24144
    }
  },
  {
    "ph": "f", "id": 24144, "pid": 76337, "tid": -914061504, "ts": 1716454217625754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625754, "dur": 0,
    "args": {
      "External id": 24145,
      "cbid": 205, "correlation": 24145
    }
  },
  {
    "ph": "f", "id": 24145, "pid": 76337, "tid": -914061504, "ts": 1716454217625754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656859, "dur": 5,
    "args": {
      "External id": 24149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24149,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24149, "pid": 5, "tid": 7, "ts": 1716454217656859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625768, "dur": 13,
    "args": {
      "External id": 24149,
      "cbid": 211, "correlation": 24149
    }
  },
  {
    "ph": "s", "id": 24149, "pid": 76337, "tid": -914061504, "ts": 1716454217625768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217656865, "dur": 159,
    "args": {
      "External id": 24151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24151,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24151, "pid": 5, "tid": 7, "ts": 1716454217656865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625783, "dur": 5,
    "args": {
      "External id": 24151,
      "cbid": 211, "correlation": 24151
    }
  },
  {
    "ph": "s", "id": 24151, "pid": 76337, "tid": -914061504, "ts": 1716454217625783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217657027, "dur": 1,
    "args": {
      "External id": 24153,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24153,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 24153, "pid": 5, "tid": 7, "ts": 1716454217657027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625794, "dur": 7,
    "args": {
      "External id": 24153,
      "cbid": 51, "correlation": 24153
    }
  },
  {
    "ph": "s", "id": 24153, "pid": 76337, "tid": -914061504, "ts": 1716454217625794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217657031, "dur": 254,
    "args": {
      "External id": 24154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24154,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24154, "pid": 5, "tid": 7, "ts": 1716454217657031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625802, "dur": 6,
    "args": {
      "External id": 24154,
      "cbid": 211, "correlation": 24154
    }
  },
  {
    "ph": "s", "id": 24154, "pid": 76337, "tid": -914061504, "ts": 1716454217625802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657286, "dur": 6,
    "args": {
      "External id": 24156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24156, "pid": 5, "tid": 7, "ts": 1716454217657286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625811, "dur": 5,
    "args": {
      "External id": 24156,
      "cbid": 211, "correlation": 24156
    }
  },
  {
    "ph": "s", "id": 24156, "pid": 76337, "tid": -914061504, "ts": 1716454217625811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217657293, "dur": 6,
    "args": {
      "External id": 24162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24162, "pid": 5, "tid": 7, "ts": 1716454217657293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625840, "dur": 9,
    "args": {
      "External id": 24162,
      "cbid": 211, "correlation": 24162
    }
  },
  {
    "ph": "s", "id": 24162, "pid": 76337, "tid": -914061504, "ts": 1716454217625840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217657300, "dur": 5,
    "args": {
      "External id": 24170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24170,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24170, "pid": 5, "tid": 7, "ts": 1716454217657300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625874, "dur": 9,
    "args": {
      "External id": 24170,
      "cbid": 211, "correlation": 24170
    }
  },
  {
    "ph": "s", "id": 24170, "pid": 76337, "tid": -914061504, "ts": 1716454217625874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217657306, "dur": 4,
    "args": {
      "External id": 24178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24178,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24178, "pid": 5, "tid": 7, "ts": 1716454217657306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217625919, "dur": 10,
    "args": {
      "External id": 24178,
      "cbid": 211, "correlation": 24178
    }
  },
  {
    "ph": "s", "id": 24178, "pid": 76337, "tid": -914061504, "ts": 1716454217625919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657312, "dur": 10,
    "args": {
      "External id": 24198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24198,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 24198, "pid": 5, "tid": 7, "ts": 1716454217657312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626081, "dur": 15,
    "args": {
      "External id": 24198,
      "cbid": 211, "correlation": 24198
    }
  },
  {
    "ph": "s", "id": 24198, "pid": 76337, "tid": -914061504, "ts": 1716454217626081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217657323, "dur": 4,
    "args": {
      "External id": 24210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24210,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24210, "pid": 5, "tid": 7, "ts": 1716454217657323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626106, "dur": 6,
    "args": {
      "External id": 24210,
      "cbid": 211, "correlation": 24210
    }
  },
  {
    "ph": "s", "id": 24210, "pid": 76337, "tid": -914061504, "ts": 1716454217626106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217657328, "dur": 7,
    "args": {
      "External id": 24213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24213, "pid": 5, "tid": 7, "ts": 1716454217657328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626125, "dur": 6,
    "args": {
      "External id": 24213,
      "cbid": 211, "correlation": 24213
    }
  },
  {
    "ph": "s", "id": 24213, "pid": 76337, "tid": -914061504, "ts": 1716454217626125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217657336, "dur": 4,
    "args": {
      "External id": 24222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24222,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24222, "pid": 5, "tid": 7, "ts": 1716454217657336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626164, "dur": 10,
    "args": {
      "External id": 24222,
      "cbid": 211, "correlation": 24222
    }
  },
  {
    "ph": "s", "id": 24222, "pid": 76337, "tid": -914061504, "ts": 1716454217626164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626236, "dur": 0,
    "args": {
      "External id": 24232,
      "cbid": 317, "correlation": 24232
    }
  },
  {
    "ph": "f", "id": 24232, "pid": 76337, "tid": -914061504, "ts": 1716454217626236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626237, "dur": 0,
    "args": {
      "External id": 24233,
      "cbid": 203, "correlation": 24233
    }
  },
  {
    "ph": "f", "id": 24233, "pid": 76337, "tid": -914061504, "ts": 1716454217626237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626237, "dur": 0,
    "args": {
      "External id": 24234,
      "cbid": 205, "correlation": 24234
    }
  },
  {
    "ph": "f", "id": 24234, "pid": 76337, "tid": -914061504, "ts": 1716454217626237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657342, "dur": 5,
    "args": {
      "External id": 24238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24238,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24238, "pid": 5, "tid": 7, "ts": 1716454217657342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626252, "dur": 12,
    "args": {
      "External id": 24238,
      "cbid": 211, "correlation": 24238
    }
  },
  {
    "ph": "s", "id": 24238, "pid": 76337, "tid": -914061504, "ts": 1716454217626252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657348, "dur": 159,
    "args": {
      "External id": 24240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24240, "pid": 5, "tid": 7, "ts": 1716454217657348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626267, "dur": 6,
    "args": {
      "External id": 24240,
      "cbid": 211, "correlation": 24240
    }
  },
  {
    "ph": "s", "id": 24240, "pid": 76337, "tid": -914061504, "ts": 1716454217626267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217657509, "dur": 1,
    "args": {
      "External id": 24242,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24242,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1561483409238777
    }
  },
  {
    "ph": "f", "id": 24242, "pid": 5, "tid": 7, "ts": 1716454217657509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626278, "dur": 7,
    "args": {
      "External id": 24242,
      "cbid": 51, "correlation": 24242
    }
  },
  {
    "ph": "s", "id": 24242, "pid": 76337, "tid": -914061504, "ts": 1716454217626278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217657513, "dur": 254,
    "args": {
      "External id": 24243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24243,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24243, "pid": 5, "tid": 7, "ts": 1716454217657513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626285, "dur": 6,
    "args": {
      "External id": 24243,
      "cbid": 211, "correlation": 24243
    }
  },
  {
    "ph": "s", "id": 24243, "pid": 76337, "tid": -914061504, "ts": 1716454217626285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657768, "dur": 6,
    "args": {
      "External id": 24245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24245,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24245, "pid": 5, "tid": 7, "ts": 1716454217657768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626296, "dur": 5,
    "args": {
      "External id": 24245,
      "cbid": 211, "correlation": 24245
    }
  },
  {
    "ph": "s", "id": 24245, "pid": 76337, "tid": -914061504, "ts": 1716454217626296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217657775, "dur": 6,
    "args": {
      "External id": 24251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24251,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24251, "pid": 5, "tid": 7, "ts": 1716454217657775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626325, "dur": 8,
    "args": {
      "External id": 24251,
      "cbid": 211, "correlation": 24251
    }
  },
  {
    "ph": "s", "id": 24251, "pid": 76337, "tid": -914061504, "ts": 1716454217626325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217657783, "dur": 3,
    "args": {
      "External id": 24259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24259,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 24259, "pid": 5, "tid": 7, "ts": 1716454217657783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626370, "dur": 9,
    "args": {
      "External id": 24259,
      "cbid": 211, "correlation": 24259
    }
  },
  {
    "ph": "s", "id": 24259, "pid": 76337, "tid": -914061504, "ts": 1716454217626370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626449, "dur": 1,
    "args": {
      "External id": 24275,
      "cbid": 251, "correlation": 24275
    }
  },
  {
    "ph": "f", "id": 24275, "pid": 76337, "tid": -914061504, "ts": 1716454217626449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626454, "dur": 0,
    "args": {
      "External id": 24277,
      "cbid": 251, "correlation": 24277
    }
  },
  {
    "ph": "f", "id": 24277, "pid": 76337, "tid": -914061504, "ts": 1716454217626454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217657787, "dur": 11,
    "args": {
      "External id": 24278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24278,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24278, "pid": 5, "tid": 7, "ts": 1716454217657787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626456, "dur": 13,
    "args": {
      "External id": 24278,
      "cbid": 211, "correlation": 24278
    }
  },
  {
    "ph": "s", "id": 24278, "pid": 76337, "tid": -914061504, "ts": 1716454217626456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217657799, "dur": 4,
    "args": {
      "External id": 24280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24280,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24280, "pid": 5, "tid": 7, "ts": 1716454217657799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626470, "dur": 5,
    "args": {
      "External id": 24280,
      "cbid": 211, "correlation": 24280
    }
  },
  {
    "ph": "s", "id": 24280, "pid": 76337, "tid": -914061504, "ts": 1716454217626470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217657804, "dur": 6,
    "args": {
      "External id": 24290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24290, "pid": 5, "tid": 7, "ts": 1716454217657804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626528, "dur": 12,
    "args": {
      "External id": 24290,
      "cbid": 211, "correlation": 24290
    }
  },
  {
    "ph": "s", "id": 24290, "pid": 76337, "tid": -914061504, "ts": 1716454217626528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657811, "dur": 9,
    "args": {
      "External id": 24310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24310,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 24310, "pid": 5, "tid": 7, "ts": 1716454217657811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626595, "dur": 11,
    "args": {
      "External id": 24310,
      "cbid": 211, "correlation": 24310
    }
  },
  {
    "ph": "s", "id": 24310, "pid": 76337, "tid": -914061504, "ts": 1716454217626595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217657822, "dur": 4,
    "args": {
      "External id": 24322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24322,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24322, "pid": 5, "tid": 7, "ts": 1716454217657822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626616, "dur": 6,
    "args": {
      "External id": 24322,
      "cbid": 211, "correlation": 24322
    }
  },
  {
    "ph": "s", "id": 24322, "pid": 76337, "tid": -914061504, "ts": 1716454217626616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217657827, "dur": 7,
    "args": {
      "External id": 24325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24325, "pid": 5, "tid": 7, "ts": 1716454217657827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626634, "dur": 6,
    "args": {
      "External id": 24325,
      "cbid": 211, "correlation": 24325
    }
  },
  {
    "ph": "s", "id": 24325, "pid": 76337, "tid": -914061504, "ts": 1716454217626634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217657834, "dur": 4,
    "args": {
      "External id": 24334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24334,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24334, "pid": 5, "tid": 7, "ts": 1716454217657834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626675, "dur": 10,
    "args": {
      "External id": 24334,
      "cbid": 211, "correlation": 24334
    }
  },
  {
    "ph": "s", "id": 24334, "pid": 76337, "tid": -914061504, "ts": 1716454217626675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626751, "dur": 0,
    "args": {
      "External id": 24344,
      "cbid": 317, "correlation": 24344
    }
  },
  {
    "ph": "f", "id": 24344, "pid": 76337, "tid": -914061504, "ts": 1716454217626751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626751, "dur": 0,
    "args": {
      "External id": 24345,
      "cbid": 203, "correlation": 24345
    }
  },
  {
    "ph": "f", "id": 24345, "pid": 76337, "tid": -914061504, "ts": 1716454217626751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626752, "dur": 0,
    "args": {
      "External id": 24346,
      "cbid": 205, "correlation": 24346
    }
  },
  {
    "ph": "f", "id": 24346, "pid": 76337, "tid": -914061504, "ts": 1716454217626752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657840, "dur": 5,
    "args": {
      "External id": 24350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24350, "pid": 5, "tid": 7, "ts": 1716454217657840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626767, "dur": 13,
    "args": {
      "External id": 24350,
      "cbid": 211, "correlation": 24350
    }
  },
  {
    "ph": "s", "id": 24350, "pid": 76337, "tid": -914061504, "ts": 1716454217626767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217657846, "dur": 159,
    "args": {
      "External id": 24352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24352,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24352, "pid": 5, "tid": 7, "ts": 1716454217657846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626782, "dur": 5,
    "args": {
      "External id": 24352,
      "cbid": 211, "correlation": 24352
    }
  },
  {
    "ph": "s", "id": 24352, "pid": 76337, "tid": -914061504, "ts": 1716454217626782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217658008, "dur": 1,
    "args": {
      "External id": 24354,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24354,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 24354, "pid": 5, "tid": 7, "ts": 1716454217658008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626792, "dur": 6,
    "args": {
      "External id": 24354,
      "cbid": 51, "correlation": 24354
    }
  },
  {
    "ph": "s", "id": 24354, "pid": 76337, "tid": -914061504, "ts": 1716454217626792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217658012, "dur": 253,
    "args": {
      "External id": 24355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24355,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24355, "pid": 5, "tid": 7, "ts": 1716454217658012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626800, "dur": 6,
    "args": {
      "External id": 24355,
      "cbid": 211, "correlation": 24355
    }
  },
  {
    "ph": "s", "id": 24355, "pid": 76337, "tid": -914061504, "ts": 1716454217626800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658266, "dur": 6,
    "args": {
      "External id": 24357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24357,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24357, "pid": 5, "tid": 7, "ts": 1716454217658266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626809, "dur": 5,
    "args": {
      "External id": 24357,
      "cbid": 211, "correlation": 24357
    }
  },
  {
    "ph": "s", "id": 24357, "pid": 76337, "tid": -914061504, "ts": 1716454217626809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217658273, "dur": 6,
    "args": {
      "External id": 24363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24363,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24363, "pid": 5, "tid": 7, "ts": 1716454217658273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626838, "dur": 8,
    "args": {
      "External id": 24363,
      "cbid": 211, "correlation": 24363
    }
  },
  {
    "ph": "s", "id": 24363, "pid": 76337, "tid": -914061504, "ts": 1716454217626838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217658281, "dur": 5,
    "args": {
      "External id": 24371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24371,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24371, "pid": 5, "tid": 7, "ts": 1716454217658281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626871, "dur": 8,
    "args": {
      "External id": 24371,
      "cbid": 211, "correlation": 24371
    }
  },
  {
    "ph": "s", "id": 24371, "pid": 76337, "tid": -914061504, "ts": 1716454217626871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217658287, "dur": 4,
    "args": {
      "External id": 24379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24379,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24379, "pid": 5, "tid": 7, "ts": 1716454217658287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217626900, "dur": 8,
    "args": {
      "External id": 24379,
      "cbid": 211, "correlation": 24379
    }
  },
  {
    "ph": "s", "id": 24379, "pid": 76337, "tid": -914061504, "ts": 1716454217626900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658293, "dur": 9,
    "args": {
      "External id": 24399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24399,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 24399, "pid": 5, "tid": 7, "ts": 1716454217658293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627057, "dur": 15,
    "args": {
      "External id": 24399,
      "cbid": 211, "correlation": 24399
    }
  },
  {
    "ph": "s", "id": 24399, "pid": 76337, "tid": -914061504, "ts": 1716454217627057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217658304, "dur": 4,
    "args": {
      "External id": 24411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24411,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24411, "pid": 5, "tid": 7, "ts": 1716454217658304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627082, "dur": 7,
    "args": {
      "External id": 24411,
      "cbid": 211, "correlation": 24411
    }
  },
  {
    "ph": "s", "id": 24411, "pid": 76337, "tid": -914061504, "ts": 1716454217627082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217658309, "dur": 6,
    "args": {
      "External id": 24414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24414,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24414, "pid": 5, "tid": 7, "ts": 1716454217658309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627102, "dur": 6,
    "args": {
      "External id": 24414,
      "cbid": 211, "correlation": 24414
    }
  },
  {
    "ph": "s", "id": 24414, "pid": 76337, "tid": -914061504, "ts": 1716454217627102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627172, "dur": 0,
    "args": {
      "External id": 24425,
      "cbid": 317, "correlation": 24425
    }
  },
  {
    "ph": "f", "id": 24425, "pid": 76337, "tid": -914061504, "ts": 1716454217627172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627173, "dur": 0,
    "args": {
      "External id": 24426,
      "cbid": 203, "correlation": 24426
    }
  },
  {
    "ph": "f", "id": 24426, "pid": 76337, "tid": -914061504, "ts": 1716454217627173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627173, "dur": 0,
    "args": {
      "External id": 24427,
      "cbid": 205, "correlation": 24427
    }
  },
  {
    "ph": "f", "id": 24427, "pid": 76337, "tid": -914061504, "ts": 1716454217627173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658316, "dur": 5,
    "args": {
      "External id": 24431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24431,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24431, "pid": 5, "tid": 7, "ts": 1716454217658316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627192, "dur": 13,
    "args": {
      "External id": 24431,
      "cbid": 211, "correlation": 24431
    }
  },
  {
    "ph": "s", "id": 24431, "pid": 76337, "tid": -914061504, "ts": 1716454217627192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217658323, "dur": 37,
    "args": {
      "External id": 24433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24433,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 24433, "pid": 5, "tid": 7, "ts": 1716454217658323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627212, "dur": 24,
    "args": {
      "External id": 24433,
      "cbid": 211, "correlation": 24433
    }
  },
  {
    "ph": "s", "id": 24433, "pid": 76337, "tid": -914061504, "ts": 1716454217627212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658361, "dur": 5,
    "args": {
      "External id": 24435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24435,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24435, "pid": 5, "tid": 7, "ts": 1716454217658361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627241, "dur": 6,
    "args": {
      "External id": 24435,
      "cbid": 211, "correlation": 24435
    }
  },
  {
    "ph": "s", "id": 24435, "pid": 76337, "tid": -914061504, "ts": 1716454217627241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217658367, "dur": 6,
    "args": {
      "External id": 24441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24441,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24441, "pid": 5, "tid": 7, "ts": 1716454217658367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627270, "dur": 9,
    "args": {
      "External id": 24441,
      "cbid": 211, "correlation": 24441
    }
  },
  {
    "ph": "s", "id": 24441, "pid": 76337, "tid": -914061504, "ts": 1716454217627270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217658374, "dur": 20,
    "args": {
      "External id": 24450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24450,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24450, "pid": 5, "tid": 7, "ts": 1716454217658374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627358, "dur": 14,
    "args": {
      "External id": 24450,
      "cbid": 211, "correlation": 24450
    }
  },
  {
    "ph": "s", "id": 24450, "pid": 76337, "tid": -914061504, "ts": 1716454217627358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658395, "dur": 10,
    "args": {
      "External id": 24472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24472,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 24472, "pid": 5, "tid": 7, "ts": 1716454217658395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627416, "dur": 10,
    "args": {
      "External id": 24472,
      "cbid": 211, "correlation": 24472
    }
  },
  {
    "ph": "s", "id": 24472, "pid": 76337, "tid": -914061504, "ts": 1716454217627416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627525, "dur": 108,
    "args": {
      "External id": 24483,
      "cbid": 251, "correlation": 24483
    }
  },
  {
    "ph": "f", "id": 24483, "pid": 76337, "tid": -914061504, "ts": 1716454217627525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627638, "dur": 0,
    "args": {
      "External id": 24484,
      "cbid": 251, "correlation": 24484
    }
  },
  {
    "ph": "f", "id": 24484, "pid": 76337, "tid": -914061504, "ts": 1716454217627638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217658413, "dur": 50,
    "args": {
      "External id": 24485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24485,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 24485, "pid": 5, "tid": 7, "ts": 1716454217658413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627641, "dur": 21,
    "args": {
      "External id": 24485,
      "cbid": 211, "correlation": 24485
    }
  },
  {
    "ph": "s", "id": 24485, "pid": 76337, "tid": -914061504, "ts": 1716454217627641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627736, "dur": 1,
    "args": {
      "External id": 24496,
      "cbid": 251, "correlation": 24496
    }
  },
  {
    "ph": "f", "id": 24496, "pid": 76337, "tid": -914061504, "ts": 1716454217627736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627740, "dur": 0,
    "args": {
      "External id": 24497,
      "cbid": 251, "correlation": 24497
    }
  },
  {
    "ph": "f", "id": 24497, "pid": 76337, "tid": -914061504, "ts": 1716454217627740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217658465, "dur": 52,
    "args": {
      "External id": 24498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24498,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 24498, "pid": 5, "tid": 7, "ts": 1716454217658465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627742, "dur": 13,
    "args": {
      "External id": 24498,
      "cbid": 211, "correlation": 24498
    }
  },
  {
    "ph": "s", "id": 24498, "pid": 76337, "tid": -914061504, "ts": 1716454217627742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627822, "dur": 1,
    "args": {
      "External id": 24509,
      "cbid": 251, "correlation": 24509
    }
  },
  {
    "ph": "f", "id": 24509, "pid": 76337, "tid": -914061504, "ts": 1716454217627822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627826, "dur": 0,
    "args": {
      "External id": 24510,
      "cbid": 251, "correlation": 24510
    }
  },
  {
    "ph": "f", "id": 24510, "pid": 76337, "tid": -914061504, "ts": 1716454217627826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217658518, "dur": 52,
    "args": {
      "External id": 24511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24511,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 24511, "pid": 5, "tid": 7, "ts": 1716454217658518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627828, "dur": 12,
    "args": {
      "External id": 24511,
      "cbid": 211, "correlation": 24511
    }
  },
  {
    "ph": "s", "id": 24511, "pid": 76337, "tid": -914061504, "ts": 1716454217627828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217658571, "dur": 55,
    "args": {
      "External id": 24536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24536,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24536, "pid": 5, "tid": 7, "ts": 1716454217658571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217627916, "dur": 14,
    "args": {
      "External id": 24536,
      "cbid": 211, "correlation": 24536
    }
  },
  {
    "ph": "s", "id": 24536, "pid": 76337, "tid": -914061504, "ts": 1716454217627916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628044, "dur": 1,
    "args": {
      "External id": 24554,
      "cbid": 251, "correlation": 24554
    }
  },
  {
    "ph": "f", "id": 24554, "pid": 76337, "tid": -914061504, "ts": 1716454217628044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217658627, "dur": 61,
    "args": {
      "External id": 24556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24556,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 24556, "pid": 5, "tid": 7, "ts": 1716454217658627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628050, "dur": 14,
    "args": {
      "External id": 24556,
      "cbid": 211, "correlation": 24556
    }
  },
  {
    "ph": "s", "id": 24556, "pid": 76337, "tid": -914061504, "ts": 1716454217628050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217658690, "dur": 6,
    "args": {
      "External id": 24564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24564,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24564, "pid": 5, "tid": 7, "ts": 1716454217658690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628124, "dur": 12,
    "args": {
      "External id": 24564,
      "cbid": 211, "correlation": 24564
    }
  },
  {
    "ph": "s", "id": 24564, "pid": 76337, "tid": -914061504, "ts": 1716454217628124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217658697, "dur": 7,
    "args": {
      "External id": 24572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24572,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24572, "pid": 5, "tid": 7, "ts": 1716454217658697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628162, "dur": 9,
    "args": {
      "External id": 24572,
      "cbid": 211, "correlation": 24572
    }
  },
  {
    "ph": "s", "id": 24572, "pid": 76337, "tid": -914061504, "ts": 1716454217628162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217658705, "dur": 8,
    "args": {
      "External id": 24583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24583,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24583, "pid": 5, "tid": 7, "ts": 1716454217658705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628238, "dur": 13,
    "args": {
      "External id": 24583,
      "cbid": 211, "correlation": 24583
    }
  },
  {
    "ph": "s", "id": 24583, "pid": 76337, "tid": -914061504, "ts": 1716454217628238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658714, "dur": 10,
    "args": {
      "External id": 24605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24605,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 24605, "pid": 5, "tid": 7, "ts": 1716454217658714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628271, "dur": 8,
    "args": {
      "External id": 24605,
      "cbid": 211, "correlation": 24605
    }
  },
  {
    "ph": "s", "id": 24605, "pid": 76337, "tid": -914061504, "ts": 1716454217628271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628367, "dur": 2,
    "args": {
      "External id": 24616,
      "cbid": 251, "correlation": 24616
    }
  },
  {
    "ph": "f", "id": 24616, "pid": 76337, "tid": -914061504, "ts": 1716454217628367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217658727, "dur": 1,
    "args": {
      "External id": 24617,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24617,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 24617, "pid": 5, "tid": 7, "ts": 1716454217658727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628373, "dur": 10,
    "args": {
      "External id": 24617,
      "cbid": 51, "correlation": 24617
    }
  },
  {
    "ph": "s", "id": 24617, "pid": 76337, "tid": -914061504, "ts": 1716454217628373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217658730, "dur": 36,
    "args": {
      "External id": 24618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24618,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 24618, "pid": 5, "tid": 7, "ts": 1716454217658730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628386, "dur": 13,
    "args": {
      "External id": 24618,
      "cbid": 211, "correlation": 24618
    }
  },
  {
    "ph": "s", "id": 24618, "pid": 76337, "tid": -914061504, "ts": 1716454217628386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628473, "dur": 1,
    "args": {
      "External id": 24629,
      "cbid": 251, "correlation": 24629
    }
  },
  {
    "ph": "f", "id": 24629, "pid": 76337, "tid": -914061504, "ts": 1716454217628473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628478, "dur": 0,
    "args": {
      "External id": 24630,
      "cbid": 251, "correlation": 24630
    }
  },
  {
    "ph": "f", "id": 24630, "pid": 76337, "tid": -914061504, "ts": 1716454217628478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217658768, "dur": 11,
    "args": {
      "External id": 24631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24631,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24631, "pid": 5, "tid": 7, "ts": 1716454217658768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628480, "dur": 14,
    "args": {
      "External id": 24631,
      "cbid": 211, "correlation": 24631
    }
  },
  {
    "ph": "s", "id": 24631, "pid": 76337, "tid": -914061504, "ts": 1716454217628480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217658780, "dur": 5,
    "args": {
      "External id": 24633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24633,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24633, "pid": 5, "tid": 7, "ts": 1716454217658780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628496, "dur": 7,
    "args": {
      "External id": 24633,
      "cbid": 211, "correlation": 24633
    }
  },
  {
    "ph": "s", "id": 24633, "pid": 76337, "tid": -914061504, "ts": 1716454217628496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628568, "dur": 1,
    "args": {
      "External id": 24644,
      "cbid": 251, "correlation": 24644
    }
  },
  {
    "ph": "f", "id": 24644, "pid": 76337, "tid": -914061504, "ts": 1716454217628568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628572, "dur": 0,
    "args": {
      "External id": 24645,
      "cbid": 251, "correlation": 24645
    }
  },
  {
    "ph": "f", "id": 24645, "pid": 76337, "tid": -914061504, "ts": 1716454217628572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217658787, "dur": 8,
    "args": {
      "External id": 24646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24646,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24646, "pid": 5, "tid": 7, "ts": 1716454217658787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628573, "dur": 13,
    "args": {
      "External id": 24646,
      "cbid": 211, "correlation": 24646
    }
  },
  {
    "ph": "s", "id": 24646, "pid": 76337, "tid": -914061504, "ts": 1716454217628573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217658797, "dur": 3,
    "args": {
      "External id": 24648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24648,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24648, "pid": 5, "tid": 7, "ts": 1716454217658797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628588, "dur": 6,
    "args": {
      "External id": 24648,
      "cbid": 211, "correlation": 24648
    }
  },
  {
    "ph": "s", "id": 24648, "pid": 76337, "tid": -914061504, "ts": 1716454217628588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217658801, "dur": 17,
    "args": {
      "External id": 24673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24673,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 24673, "pid": 5, "tid": 7, "ts": 1716454217658801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628668, "dur": 12,
    "args": {
      "External id": 24673,
      "cbid": 211, "correlation": 24673
    }
  },
  {
    "ph": "s", "id": 24673, "pid": 76337, "tid": -914061504, "ts": 1716454217628668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628782, "dur": 2,
    "args": {
      "External id": 24691,
      "cbid": 251, "correlation": 24691
    }
  },
  {
    "ph": "f", "id": 24691, "pid": 76337, "tid": -914061504, "ts": 1716454217628782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217658820, "dur": 1,
    "args": {
      "External id": 24693,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24693,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 24693, "pid": 5, "tid": 7, "ts": 1716454217658820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628788, "dur": 10,
    "args": {
      "External id": 24693,
      "cbid": 51, "correlation": 24693
    }
  },
  {
    "ph": "s", "id": 24693, "pid": 76337, "tid": -914061504, "ts": 1716454217628788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217658824, "dur": 36,
    "args": {
      "External id": 24694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24694,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 24694, "pid": 5, "tid": 7, "ts": 1716454217658824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628800, "dur": 13,
    "args": {
      "External id": 24694,
      "cbid": 211, "correlation": 24694
    }
  },
  {
    "ph": "s", "id": 24694, "pid": 76337, "tid": -914061504, "ts": 1716454217628800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217658861, "dur": 4,
    "args": {
      "External id": 24702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24702,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24702, "pid": 5, "tid": 7, "ts": 1716454217658861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628873, "dur": 12,
    "args": {
      "External id": 24702,
      "cbid": 211, "correlation": 24702
    }
  },
  {
    "ph": "s", "id": 24702, "pid": 76337, "tid": -914061504, "ts": 1716454217628873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217658867, "dur": 8,
    "args": {
      "External id": 24710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24710, "pid": 5, "tid": 7, "ts": 1716454217658867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628916, "dur": 9,
    "args": {
      "External id": 24710,
      "cbid": 211, "correlation": 24710
    }
  },
  {
    "ph": "s", "id": 24710, "pid": 76337, "tid": -914061504, "ts": 1716454217628916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217658876, "dur": 9,
    "args": {
      "External id": 24732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24732,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 24732, "pid": 5, "tid": 7, "ts": 1716454217658876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217628968, "dur": 18,
    "args": {
      "External id": 24732,
      "cbid": 211, "correlation": 24732
    }
  },
  {
    "ph": "s", "id": 24732, "pid": 76337, "tid": -914061504, "ts": 1716454217628968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629071, "dur": 1,
    "args": {
      "External id": 24748,
      "cbid": 251, "correlation": 24748
    }
  },
  {
    "ph": "f", "id": 24748, "pid": 76337, "tid": -914061504, "ts": 1716454217629071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629076, "dur": 0,
    "args": {
      "External id": 24750,
      "cbid": 251, "correlation": 24750
    }
  },
  {
    "ph": "f", "id": 24750, "pid": 76337, "tid": -914061504, "ts": 1716454217629076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217658887, "dur": 188,
    "args": {
      "External id": 24751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24751,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24751, "pid": 5, "tid": 7, "ts": 1716454217658887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629078, "dur": 14,
    "args": {
      "External id": 24751,
      "cbid": 211, "correlation": 24751
    }
  },
  {
    "ph": "s", "id": 24751, "pid": 76337, "tid": -914061504, "ts": 1716454217629078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659076, "dur": 21,
    "args": {
      "External id": 24759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24759, "pid": 5, "tid": 7, "ts": 1716454217659076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629148, "dur": 12,
    "args": {
      "External id": 24759,
      "cbid": 211, "correlation": 24759
    }
  },
  {
    "ph": "s", "id": 24759, "pid": 76337, "tid": -914061504, "ts": 1716454217629148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659098, "dur": 21,
    "args": {
      "External id": 24767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24767, "pid": 5, "tid": 7, "ts": 1716454217659098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629182, "dur": 8,
    "args": {
      "External id": 24767,
      "cbid": 211, "correlation": 24767
    }
  },
  {
    "ph": "s", "id": 24767, "pid": 76337, "tid": -914061504, "ts": 1716454217629182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629264, "dur": 1,
    "args": {
      "External id": 24783,
      "cbid": 251, "correlation": 24783
    }
  },
  {
    "ph": "f", "id": 24783, "pid": 76337, "tid": -914061504, "ts": 1716454217629264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217659121, "dur": 1,
    "args": {
      "External id": 24785,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24785,
      "bytes": 120, "memory bandwidth (GB/s)": 0.07807417046193885
    }
  },
  {
    "ph": "f", "id": 24785, "pid": 5, "tid": 7, "ts": 1716454217659121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629268, "dur": 9,
    "args": {
      "External id": 24785,
      "cbid": 51, "correlation": 24785
    }
  },
  {
    "ph": "s", "id": 24785, "pid": 76337, "tid": -914061504, "ts": 1716454217629268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217659125, "dur": 109,
    "args": {
      "External id": 24786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24786,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 24786, "pid": 5, "tid": 7, "ts": 1716454217659125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629279, "dur": 12,
    "args": {
      "External id": 24786,
      "cbid": 211, "correlation": 24786
    }
  },
  {
    "ph": "s", "id": 24786, "pid": 76337, "tid": -914061504, "ts": 1716454217629279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217659236, "dur": 5,
    "args": {
      "External id": 24794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24794,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24794, "pid": 5, "tid": 7, "ts": 1716454217659236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629322, "dur": 10,
    "args": {
      "External id": 24794,
      "cbid": 211, "correlation": 24794
    }
  },
  {
    "ph": "s", "id": 24794, "pid": 76337, "tid": -914061504, "ts": 1716454217629322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659242, "dur": 9,
    "args": {
      "External id": 24805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24805, "pid": 5, "tid": 7, "ts": 1716454217659242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629392, "dur": 12,
    "args": {
      "External id": 24805,
      "cbid": 211, "correlation": 24805
    }
  },
  {
    "ph": "s", "id": 24805, "pid": 76337, "tid": -914061504, "ts": 1716454217629392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629457, "dur": 0,
    "args": {
      "External id": 24817,
      "cbid": 317, "correlation": 24817
    }
  },
  {
    "ph": "f", "id": 24817, "pid": 76337, "tid": -914061504, "ts": 1716454217629457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629458, "dur": 0,
    "args": {
      "External id": 24818,
      "cbid": 203, "correlation": 24818
    }
  },
  {
    "ph": "f", "id": 24818, "pid": 76337, "tid": -914061504, "ts": 1716454217629458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629459, "dur": 0,
    "args": {
      "External id": 24819,
      "cbid": 205, "correlation": 24819
    }
  },
  {
    "ph": "f", "id": 24819, "pid": 76337, "tid": -914061504, "ts": 1716454217629459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659252, "dur": 5,
    "args": {
      "External id": 24823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24823,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24823, "pid": 5, "tid": 7, "ts": 1716454217659252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629474, "dur": 13,
    "args": {
      "External id": 24823,
      "cbid": 211, "correlation": 24823
    }
  },
  {
    "ph": "s", "id": 24823, "pid": 76337, "tid": -914061504, "ts": 1716454217629474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217659259, "dur": 36,
    "args": {
      "External id": 24825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24825,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 24825, "pid": 5, "tid": 7, "ts": 1716454217659259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629495, "dur": 7,
    "args": {
      "External id": 24825,
      "cbid": 211, "correlation": 24825
    }
  },
  {
    "ph": "s", "id": 24825, "pid": 76337, "tid": -914061504, "ts": 1716454217629495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659296, "dur": 6,
    "args": {
      "External id": 24827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24827,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24827, "pid": 5, "tid": 7, "ts": 1716454217659296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629506, "dur": 5,
    "args": {
      "External id": 24827,
      "cbid": 211, "correlation": 24827
    }
  },
  {
    "ph": "s", "id": 24827, "pid": 76337, "tid": -914061504, "ts": 1716454217629506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659304, "dur": 7,
    "args": {
      "External id": 24833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24833, "pid": 5, "tid": 7, "ts": 1716454217659304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629533, "dur": 9,
    "args": {
      "External id": 24833,
      "cbid": 211, "correlation": 24833
    }
  },
  {
    "ph": "s", "id": 24833, "pid": 76337, "tid": -914061504, "ts": 1716454217629533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217659312, "dur": 5,
    "args": {
      "External id": 24841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24841,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24841, "pid": 5, "tid": 7, "ts": 1716454217659312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629565, "dur": 8,
    "args": {
      "External id": 24841,
      "cbid": 211, "correlation": 24841
    }
  },
  {
    "ph": "s", "id": 24841, "pid": 76337, "tid": -914061504, "ts": 1716454217629565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659318, "dur": 11,
    "args": {
      "External id": 24861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24861,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 24861, "pid": 5, "tid": 7, "ts": 1716454217659318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629639, "dur": 12,
    "args": {
      "External id": 24861,
      "cbid": 211, "correlation": 24861
    }
  },
  {
    "ph": "s", "id": 24861, "pid": 76337, "tid": -914061504, "ts": 1716454217629639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217659330, "dur": 4,
    "args": {
      "External id": 24873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24873,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24873, "pid": 5, "tid": 7, "ts": 1716454217659330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629661, "dur": 6,
    "args": {
      "External id": 24873,
      "cbid": 211, "correlation": 24873
    }
  },
  {
    "ph": "s", "id": 24873, "pid": 76337, "tid": -914061504, "ts": 1716454217629661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659336, "dur": 9,
    "args": {
      "External id": 24876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24876, "pid": 5, "tid": 7, "ts": 1716454217659336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629679, "dur": 7,
    "args": {
      "External id": 24876,
      "cbid": 211, "correlation": 24876
    }
  },
  {
    "ph": "s", "id": 24876, "pid": 76337, "tid": -914061504, "ts": 1716454217629679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217659345, "dur": 5,
    "args": {
      "External id": 24885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24885,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24885, "pid": 5, "tid": 7, "ts": 1716454217659345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629719, "dur": 10,
    "args": {
      "External id": 24885,
      "cbid": 211, "correlation": 24885
    }
  },
  {
    "ph": "s", "id": 24885, "pid": 76337, "tid": -914061504, "ts": 1716454217629719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629785, "dur": 0,
    "args": {
      "External id": 24895,
      "cbid": 317, "correlation": 24895
    }
  },
  {
    "ph": "f", "id": 24895, "pid": 76337, "tid": -914061504, "ts": 1716454217629785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629785, "dur": 0,
    "args": {
      "External id": 24896,
      "cbid": 203, "correlation": 24896
    }
  },
  {
    "ph": "f", "id": 24896, "pid": 76337, "tid": -914061504, "ts": 1716454217629785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629786, "dur": 0,
    "args": {
      "External id": 24897,
      "cbid": 205, "correlation": 24897
    }
  },
  {
    "ph": "f", "id": 24897, "pid": 76337, "tid": -914061504, "ts": 1716454217629786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659352, "dur": 5,
    "args": {
      "External id": 24901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24901,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24901, "pid": 5, "tid": 7, "ts": 1716454217659352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629800, "dur": 13,
    "args": {
      "External id": 24901,
      "cbid": 211, "correlation": 24901
    }
  },
  {
    "ph": "s", "id": 24901, "pid": 76337, "tid": -914061504, "ts": 1716454217629800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659358, "dur": 160,
    "args": {
      "External id": 24903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24903,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24903, "pid": 5, "tid": 7, "ts": 1716454217659358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629816, "dur": 5,
    "args": {
      "External id": 24903,
      "cbid": 211, "correlation": 24903
    }
  },
  {
    "ph": "s", "id": 24903, "pid": 76337, "tid": -914061504, "ts": 1716454217629816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217659520, "dur": 1,
    "args": {
      "External id": 24905,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 24905,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 24905, "pid": 5, "tid": 7, "ts": 1716454217659520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629827, "dur": 6,
    "args": {
      "External id": 24905,
      "cbid": 51, "correlation": 24905
    }
  },
  {
    "ph": "s", "id": 24905, "pid": 76337, "tid": -914061504, "ts": 1716454217629827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217659523, "dur": 265,
    "args": {
      "External id": 24906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24906,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24906, "pid": 5, "tid": 7, "ts": 1716454217659523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629834, "dur": 6,
    "args": {
      "External id": 24906,
      "cbid": 211, "correlation": 24906
    }
  },
  {
    "ph": "s", "id": 24906, "pid": 76337, "tid": -914061504, "ts": 1716454217629834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659789, "dur": 6,
    "args": {
      "External id": 24908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24908,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 24908, "pid": 5, "tid": 7, "ts": 1716454217659789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629844, "dur": 5,
    "args": {
      "External id": 24908,
      "cbid": 211, "correlation": 24908
    }
  },
  {
    "ph": "s", "id": 24908, "pid": 76337, "tid": -914061504, "ts": 1716454217629844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659796, "dur": 6,
    "args": {
      "External id": 24914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24914, "pid": 5, "tid": 7, "ts": 1716454217659796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629872, "dur": 9,
    "args": {
      "External id": 24914,
      "cbid": 211, "correlation": 24914
    }
  },
  {
    "ph": "s", "id": 24914, "pid": 76337, "tid": -914061504, "ts": 1716454217629872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217659804, "dur": 3,
    "args": {
      "External id": 24922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24922,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 24922, "pid": 5, "tid": 7, "ts": 1716454217659804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629918, "dur": 9,
    "args": {
      "External id": 24922,
      "cbid": 211, "correlation": 24922
    }
  },
  {
    "ph": "s", "id": 24922, "pid": 76337, "tid": -914061504, "ts": 1716454217629918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629992, "dur": 1,
    "args": {
      "External id": 24938,
      "cbid": 251, "correlation": 24938
    }
  },
  {
    "ph": "f", "id": 24938, "pid": 76337, "tid": -914061504, "ts": 1716454217629992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629997, "dur": 0,
    "args": {
      "External id": 24940,
      "cbid": 251, "correlation": 24940
    }
  },
  {
    "ph": "f", "id": 24940, "pid": 76337, "tid": -914061504, "ts": 1716454217629997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217659808, "dur": 12,
    "args": {
      "External id": 24941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24941,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24941, "pid": 5, "tid": 7, "ts": 1716454217659808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217629999, "dur": 13,
    "args": {
      "External id": 24941,
      "cbid": 211, "correlation": 24941
    }
  },
  {
    "ph": "s", "id": 24941, "pid": 76337, "tid": -914061504, "ts": 1716454217629999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217659822, "dur": 5,
    "args": {
      "External id": 24943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24943,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 24943, "pid": 5, "tid": 7, "ts": 1716454217659822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630014, "dur": 5,
    "args": {
      "External id": 24943,
      "cbid": 211, "correlation": 24943
    }
  },
  {
    "ph": "s", "id": 24943, "pid": 76337, "tid": -914061504, "ts": 1716454217630014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659828, "dur": 6,
    "args": {
      "External id": 24953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24953,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24953, "pid": 5, "tid": 7, "ts": 1716454217659828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630073, "dur": 12,
    "args": {
      "External id": 24953,
      "cbid": 211, "correlation": 24953
    }
  },
  {
    "ph": "s", "id": 24953, "pid": 76337, "tid": -914061504, "ts": 1716454217630073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659835, "dur": 10,
    "args": {
      "External id": 24973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24973,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 24973, "pid": 5, "tid": 7, "ts": 1716454217659835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630139, "dur": 10,
    "args": {
      "External id": 24973,
      "cbid": 211, "correlation": 24973
    }
  },
  {
    "ph": "s", "id": 24973, "pid": 76337, "tid": -914061504, "ts": 1716454217630139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217659846, "dur": 4,
    "args": {
      "External id": 24985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24985,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 24985, "pid": 5, "tid": 7, "ts": 1716454217659846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630160, "dur": 6,
    "args": {
      "External id": 24985,
      "cbid": 211, "correlation": 24985
    }
  },
  {
    "ph": "s", "id": 24985, "pid": 76337, "tid": -914061504, "ts": 1716454217630160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217659851, "dur": 7,
    "args": {
      "External id": 24988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24988,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24988, "pid": 5, "tid": 7, "ts": 1716454217659851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630178, "dur": 7,
    "args": {
      "External id": 24988,
      "cbid": 211, "correlation": 24988
    }
  },
  {
    "ph": "s", "id": 24988, "pid": 76337, "tid": -914061504, "ts": 1716454217630178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217659859, "dur": 4,
    "args": {
      "External id": 24997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 24997,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 24997, "pid": 5, "tid": 7, "ts": 1716454217659859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630218, "dur": 10,
    "args": {
      "External id": 24997,
      "cbid": 211, "correlation": 24997
    }
  },
  {
    "ph": "s", "id": 24997, "pid": 76337, "tid": -914061504, "ts": 1716454217630218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630282, "dur": 0,
    "args": {
      "External id": 25007,
      "cbid": 317, "correlation": 25007
    }
  },
  {
    "ph": "f", "id": 25007, "pid": 76337, "tid": -914061504, "ts": 1716454217630282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630283, "dur": 0,
    "args": {
      "External id": 25008,
      "cbid": 203, "correlation": 25008
    }
  },
  {
    "ph": "f", "id": 25008, "pid": 76337, "tid": -914061504, "ts": 1716454217630283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630284, "dur": 0,
    "args": {
      "External id": 25009,
      "cbid": 205, "correlation": 25009
    }
  },
  {
    "ph": "f", "id": 25009, "pid": 76337, "tid": -914061504, "ts": 1716454217630284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659865, "dur": 5,
    "args": {
      "External id": 25013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25013,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25013, "pid": 5, "tid": 7, "ts": 1716454217659865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630297, "dur": 12,
    "args": {
      "External id": 25013,
      "cbid": 211, "correlation": 25013
    }
  },
  {
    "ph": "s", "id": 25013, "pid": 76337, "tid": -914061504, "ts": 1716454217630297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217659871, "dur": 160,
    "args": {
      "External id": 25015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25015,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25015, "pid": 5, "tid": 7, "ts": 1716454217659871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630311, "dur": 5,
    "args": {
      "External id": 25015,
      "cbid": 211, "correlation": 25015
    }
  },
  {
    "ph": "s", "id": 25015, "pid": 76337, "tid": -914061504, "ts": 1716454217630311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217660033, "dur": 1,
    "args": {
      "External id": 25017,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25017,
      "bytes": 240, "memory bandwidth (GB/s)": 0.13157894736842105
    }
  },
  {
    "ph": "f", "id": 25017, "pid": 5, "tid": 7, "ts": 1716454217660033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630321, "dur": 6,
    "args": {
      "External id": 25017,
      "cbid": 51, "correlation": 25017
    }
  },
  {
    "ph": "s", "id": 25017, "pid": 76337, "tid": -914061504, "ts": 1716454217630321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217660037, "dur": 254,
    "args": {
      "External id": 25018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25018,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25018, "pid": 5, "tid": 7, "ts": 1716454217660037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630329, "dur": 7,
    "args": {
      "External id": 25018,
      "cbid": 211, "correlation": 25018
    }
  },
  {
    "ph": "s", "id": 25018, "pid": 76337, "tid": -914061504, "ts": 1716454217630329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217660293, "dur": 6,
    "args": {
      "External id": 25020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25020, "pid": 5, "tid": 7, "ts": 1716454217660293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630339, "dur": 5,
    "args": {
      "External id": 25020,
      "cbid": 211, "correlation": 25020
    }
  },
  {
    "ph": "s", "id": 25020, "pid": 76337, "tid": -914061504, "ts": 1716454217630339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217660300, "dur": 6,
    "args": {
      "External id": 25026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25026, "pid": 5, "tid": 7, "ts": 1716454217660300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630368, "dur": 8,
    "args": {
      "External id": 25026,
      "cbid": 211, "correlation": 25026
    }
  },
  {
    "ph": "s", "id": 25026, "pid": 76337, "tid": -914061504, "ts": 1716454217630368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217660308, "dur": 4,
    "args": {
      "External id": 25034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25034,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25034, "pid": 5, "tid": 7, "ts": 1716454217660308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630400, "dur": 8,
    "args": {
      "External id": 25034,
      "cbid": 211, "correlation": 25034
    }
  },
  {
    "ph": "s", "id": 25034, "pid": 76337, "tid": -914061504, "ts": 1716454217630400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217660313, "dur": 4,
    "args": {
      "External id": 25042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25042,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25042, "pid": 5, "tid": 7, "ts": 1716454217660313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630431, "dur": 9,
    "args": {
      "External id": 25042,
      "cbid": 211, "correlation": 25042
    }
  },
  {
    "ph": "s", "id": 25042, "pid": 76337, "tid": -914061504, "ts": 1716454217630431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217660319, "dur": 11,
    "args": {
      "External id": 25051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25051,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25051, "pid": 5, "tid": 7, "ts": 1716454217660319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630559, "dur": 15,
    "args": {
      "External id": 25051,
      "cbid": 211, "correlation": 25051
    }
  },
  {
    "ph": "s", "id": 25051, "pid": 76337, "tid": -914061504, "ts": 1716454217630559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217660332, "dur": 12,
    "args": {
      "External id": 25071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25071,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25071, "pid": 5, "tid": 7, "ts": 1716454217660332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630640, "dur": 13,
    "args": {
      "External id": 25071,
      "cbid": 211, "correlation": 25071
    }
  },
  {
    "ph": "s", "id": 25071, "pid": 76337, "tid": -914061504, "ts": 1716454217630640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217660345, "dur": 4,
    "args": {
      "External id": 25083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25083,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25083, "pid": 5, "tid": 7, "ts": 1716454217660345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630663, "dur": 6,
    "args": {
      "External id": 25083,
      "cbid": 211, "correlation": 25083
    }
  },
  {
    "ph": "s", "id": 25083, "pid": 76337, "tid": -914061504, "ts": 1716454217630663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217660351, "dur": 10,
    "args": {
      "External id": 25086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25086, "pid": 5, "tid": 7, "ts": 1716454217660351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630682, "dur": 7,
    "args": {
      "External id": 25086,
      "cbid": 211, "correlation": 25086
    }
  },
  {
    "ph": "s", "id": 25086, "pid": 76337, "tid": -914061504, "ts": 1716454217630682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217660362, "dur": 6,
    "args": {
      "External id": 25095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25095,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25095, "pid": 5, "tid": 7, "ts": 1716454217660362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630723, "dur": 10,
    "args": {
      "External id": 25095,
      "cbid": 211, "correlation": 25095
    }
  },
  {
    "ph": "s", "id": 25095, "pid": 76337, "tid": -914061504, "ts": 1716454217630723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630789, "dur": 0,
    "args": {
      "External id": 25105,
      "cbid": 317, "correlation": 25105
    }
  },
  {
    "ph": "f", "id": 25105, "pid": 76337, "tid": -914061504, "ts": 1716454217630789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630790, "dur": 0,
    "args": {
      "External id": 25106,
      "cbid": 203, "correlation": 25106
    }
  },
  {
    "ph": "f", "id": 25106, "pid": 76337, "tid": -914061504, "ts": 1716454217630790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630791, "dur": 0,
    "args": {
      "External id": 25107,
      "cbid": 205, "correlation": 25107
    }
  },
  {
    "ph": "f", "id": 25107, "pid": 76337, "tid": -914061504, "ts": 1716454217630791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217660370, "dur": 6,
    "args": {
      "External id": 25111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25111,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25111, "pid": 5, "tid": 7, "ts": 1716454217660370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630808, "dur": 13,
    "args": {
      "External id": 25111,
      "cbid": 211, "correlation": 25111
    }
  },
  {
    "ph": "s", "id": 25111, "pid": 76337, "tid": -914061504, "ts": 1716454217630808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217660377, "dur": 315,
    "args": {
      "External id": 25113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25113,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25113, "pid": 5, "tid": 7, "ts": 1716454217660377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630823, "dur": 6,
    "args": {
      "External id": 25113,
      "cbid": 211, "correlation": 25113
    }
  },
  {
    "ph": "s", "id": 25113, "pid": 76337, "tid": -914061504, "ts": 1716454217630823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217660695, "dur": 1,
    "args": {
      "External id": 25115,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25115,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 25115, "pid": 5, "tid": 7, "ts": 1716454217660695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630835, "dur": 7,
    "args": {
      "External id": 25115,
      "cbid": 51, "correlation": 25115
    }
  },
  {
    "ph": "s", "id": 25115, "pid": 76337, "tid": -914061504, "ts": 1716454217630835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217660698, "dur": 486,
    "args": {
      "External id": 25116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25116,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25116, "pid": 5, "tid": 7, "ts": 1716454217660698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630842, "dur": 6,
    "args": {
      "External id": 25116,
      "cbid": 211, "correlation": 25116
    }
  },
  {
    "ph": "s", "id": 25116, "pid": 76337, "tid": -914061504, "ts": 1716454217630842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661186, "dur": 5,
    "args": {
      "External id": 25118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25118, "pid": 5, "tid": 7, "ts": 1716454217661186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630853, "dur": 5,
    "args": {
      "External id": 25118,
      "cbid": 211, "correlation": 25118
    }
  },
  {
    "ph": "s", "id": 25118, "pid": 76337, "tid": -914061504, "ts": 1716454217630853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217661192, "dur": 6,
    "args": {
      "External id": 25124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25124,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25124, "pid": 5, "tid": 7, "ts": 1716454217661192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630883, "dur": 9,
    "args": {
      "External id": 25124,
      "cbid": 211, "correlation": 25124
    }
  },
  {
    "ph": "s", "id": 25124, "pid": 76337, "tid": -914061504, "ts": 1716454217630883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217661200, "dur": 3,
    "args": {
      "External id": 25132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25132,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 25132, "pid": 5, "tid": 7, "ts": 1716454217661200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217630927, "dur": 9,
    "args": {
      "External id": 25132,
      "cbid": 211, "correlation": 25132
    }
  },
  {
    "ph": "s", "id": 25132, "pid": 76337, "tid": -914061504, "ts": 1716454217630927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631011, "dur": 1,
    "args": {
      "External id": 25148,
      "cbid": 251, "correlation": 25148
    }
  },
  {
    "ph": "f", "id": 25148, "pid": 76337, "tid": -914061504, "ts": 1716454217631011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631016, "dur": 0,
    "args": {
      "External id": 25150,
      "cbid": 251, "correlation": 25150
    }
  },
  {
    "ph": "f", "id": 25150, "pid": 76337, "tid": -914061504, "ts": 1716454217631016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217661204, "dur": 11,
    "args": {
      "External id": 25151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25151,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25151, "pid": 5, "tid": 7, "ts": 1716454217661204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631018, "dur": 13,
    "args": {
      "External id": 25151,
      "cbid": 211, "correlation": 25151
    }
  },
  {
    "ph": "s", "id": 25151, "pid": 76337, "tid": -914061504, "ts": 1716454217631018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217661216, "dur": 5,
    "args": {
      "External id": 25153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25153,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25153, "pid": 5, "tid": 7, "ts": 1716454217661216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631034, "dur": 5,
    "args": {
      "External id": 25153,
      "cbid": 211, "correlation": 25153
    }
  },
  {
    "ph": "s", "id": 25153, "pid": 76337, "tid": -914061504, "ts": 1716454217631034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217661222, "dur": 6,
    "args": {
      "External id": 25163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25163,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25163, "pid": 5, "tid": 7, "ts": 1716454217661222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631092, "dur": 12,
    "args": {
      "External id": 25163,
      "cbid": 211, "correlation": 25163
    }
  },
  {
    "ph": "s", "id": 25163, "pid": 76337, "tid": -914061504, "ts": 1716454217631092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661229, "dur": 9,
    "args": {
      "External id": 25183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25183,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25183, "pid": 5, "tid": 7, "ts": 1716454217661229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631159, "dur": 11,
    "args": {
      "External id": 25183,
      "cbid": 211, "correlation": 25183
    }
  },
  {
    "ph": "s", "id": 25183, "pid": 76337, "tid": -914061504, "ts": 1716454217631159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217661240, "dur": 3,
    "args": {
      "External id": 25195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25195,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 25195, "pid": 5, "tid": 7, "ts": 1716454217661240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631179, "dur": 6,
    "args": {
      "External id": 25195,
      "cbid": 211, "correlation": 25195
    }
  },
  {
    "ph": "s", "id": 25195, "pid": 76337, "tid": -914061504, "ts": 1716454217631179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217661245, "dur": 7,
    "args": {
      "External id": 25198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25198, "pid": 5, "tid": 7, "ts": 1716454217661245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631198, "dur": 6,
    "args": {
      "External id": 25198,
      "cbid": 211, "correlation": 25198
    }
  },
  {
    "ph": "s", "id": 25198, "pid": 76337, "tid": -914061504, "ts": 1716454217631198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217661253, "dur": 4,
    "args": {
      "External id": 25207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25207,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25207, "pid": 5, "tid": 7, "ts": 1716454217661253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631237, "dur": 11,
    "args": {
      "External id": 25207,
      "cbid": 211, "correlation": 25207
    }
  },
  {
    "ph": "s", "id": 25207, "pid": 76337, "tid": -914061504, "ts": 1716454217631237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631300, "dur": 0,
    "args": {
      "External id": 25217,
      "cbid": 317, "correlation": 25217
    }
  },
  {
    "ph": "f", "id": 25217, "pid": 76337, "tid": -914061504, "ts": 1716454217631300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631301, "dur": 0,
    "args": {
      "External id": 25218,
      "cbid": 203, "correlation": 25218
    }
  },
  {
    "ph": "f", "id": 25218, "pid": 76337, "tid": -914061504, "ts": 1716454217631301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631302, "dur": 0,
    "args": {
      "External id": 25219,
      "cbid": 205, "correlation": 25219
    }
  },
  {
    "ph": "f", "id": 25219, "pid": 76337, "tid": -914061504, "ts": 1716454217631302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661258, "dur": 5,
    "args": {
      "External id": 25223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25223,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25223, "pid": 5, "tid": 7, "ts": 1716454217661258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631315, "dur": 12,
    "args": {
      "External id": 25223,
      "cbid": 211, "correlation": 25223
    }
  },
  {
    "ph": "s", "id": 25223, "pid": 76337, "tid": -914061504, "ts": 1716454217631315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661265, "dur": 159,
    "args": {
      "External id": 25225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25225,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25225, "pid": 5, "tid": 7, "ts": 1716454217661265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631330, "dur": 5,
    "args": {
      "External id": 25225,
      "cbid": 211, "correlation": 25225
    }
  },
  {
    "ph": "s", "id": 25225, "pid": 76337, "tid": -914061504, "ts": 1716454217631330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217661426, "dur": 1,
    "args": {
      "External id": 25227,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25227,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 25227, "pid": 5, "tid": 7, "ts": 1716454217661426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631341, "dur": 6,
    "args": {
      "External id": 25227,
      "cbid": 51, "correlation": 25227
    }
  },
  {
    "ph": "s", "id": 25227, "pid": 76337, "tid": -914061504, "ts": 1716454217631341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217661429, "dur": 255,
    "args": {
      "External id": 25228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25228,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25228, "pid": 5, "tid": 7, "ts": 1716454217661429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631348, "dur": 6,
    "args": {
      "External id": 25228,
      "cbid": 211, "correlation": 25228
    }
  },
  {
    "ph": "s", "id": 25228, "pid": 76337, "tid": -914061504, "ts": 1716454217631348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661685, "dur": 5,
    "args": {
      "External id": 25230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25230,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25230, "pid": 5, "tid": 7, "ts": 1716454217661685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631358, "dur": 5,
    "args": {
      "External id": 25230,
      "cbid": 211, "correlation": 25230
    }
  },
  {
    "ph": "s", "id": 25230, "pid": 76337, "tid": -914061504, "ts": 1716454217631358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217661692, "dur": 6,
    "args": {
      "External id": 25236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25236, "pid": 5, "tid": 7, "ts": 1716454217661692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631386, "dur": 8,
    "args": {
      "External id": 25236,
      "cbid": 211, "correlation": 25236
    }
  },
  {
    "ph": "s", "id": 25236, "pid": 76337, "tid": -914061504, "ts": 1716454217631386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631444, "dur": 0,
    "args": {
      "External id": 25246,
      "cbid": 317, "correlation": 25246
    }
  },
  {
    "ph": "f", "id": 25246, "pid": 76337, "tid": -914061504, "ts": 1716454217631444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631445, "dur": 0,
    "args": {
      "External id": 25247,
      "cbid": 203, "correlation": 25247
    }
  },
  {
    "ph": "f", "id": 25247, "pid": 76337, "tid": -914061504, "ts": 1716454217631445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631446, "dur": 0,
    "args": {
      "External id": 25248,
      "cbid": 205, "correlation": 25248
    }
  },
  {
    "ph": "f", "id": 25248, "pid": 76337, "tid": -914061504, "ts": 1716454217631446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661699, "dur": 7,
    "args": {
      "External id": 25252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25252,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25252, "pid": 5, "tid": 7, "ts": 1716454217661699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631461, "dur": 12,
    "args": {
      "External id": 25252,
      "cbid": 211, "correlation": 25252
    }
  },
  {
    "ph": "s", "id": 25252, "pid": 76337, "tid": -914061504, "ts": 1716454217631461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217661708, "dur": 3,
    "args": {
      "External id": 25254,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25254,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4285714285714286
    }
  },
  {
    "ph": "f", "id": 25254, "pid": 5, "tid": 7, "ts": 1716454217661708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631479, "dur": 15,
    "args": {
      "External id": 25254,
      "cbid": 51, "correlation": 25254
    }
  },
  {
    "ph": "s", "id": 25254, "pid": 76337, "tid": -914061504, "ts": 1716454217631479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217661712, "dur": 97,
    "args": {
      "External id": 25255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25255,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 25255, "pid": 5, "tid": 7, "ts": 1716454217661712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631495, "dur": 7,
    "args": {
      "External id": 25255,
      "cbid": 211, "correlation": 25255
    }
  },
  {
    "ph": "s", "id": 25255, "pid": 76337, "tid": -914061504, "ts": 1716454217631495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661811, "dur": 6,
    "args": {
      "External id": 25257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25257,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25257, "pid": 5, "tid": 7, "ts": 1716454217661811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631506, "dur": 5,
    "args": {
      "External id": 25257,
      "cbid": 211, "correlation": 25257
    }
  },
  {
    "ph": "s", "id": 25257, "pid": 76337, "tid": -914061504, "ts": 1716454217631506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217661817, "dur": 6,
    "args": {
      "External id": 25263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25263, "pid": 5, "tid": 7, "ts": 1716454217661817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631533, "dur": 9,
    "args": {
      "External id": 25263,
      "cbid": 211, "correlation": 25263
    }
  },
  {
    "ph": "s", "id": 25263, "pid": 76337, "tid": -914061504, "ts": 1716454217631533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217661825, "dur": 5,
    "args": {
      "External id": 25271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25271,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25271, "pid": 5, "tid": 7, "ts": 1716454217661825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631563, "dur": 8,
    "args": {
      "External id": 25271,
      "cbid": 211, "correlation": 25271
    }
  },
  {
    "ph": "s", "id": 25271, "pid": 76337, "tid": -914061504, "ts": 1716454217631563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217661831, "dur": 4,
    "args": {
      "External id": 25279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25279,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25279, "pid": 5, "tid": 7, "ts": 1716454217661831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631592, "dur": 8,
    "args": {
      "External id": 25279,
      "cbid": 211, "correlation": 25279
    }
  },
  {
    "ph": "s", "id": 25279, "pid": 76337, "tid": -914061504, "ts": 1716454217631592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217661837, "dur": 11,
    "args": {
      "External id": 25288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25288,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25288, "pid": 5, "tid": 7, "ts": 1716454217661837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631636, "dur": 10,
    "args": {
      "External id": 25288,
      "cbid": 211, "correlation": 25288
    }
  },
  {
    "ph": "s", "id": 25288, "pid": 76337, "tid": -914061504, "ts": 1716454217631636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661849, "dur": 12,
    "args": {
      "External id": 25308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25308,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25308, "pid": 5, "tid": 7, "ts": 1716454217661849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631706, "dur": 12,
    "args": {
      "External id": 25308,
      "cbid": 211, "correlation": 25308
    }
  },
  {
    "ph": "s", "id": 25308, "pid": 76337, "tid": -914061504, "ts": 1716454217631706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217661862, "dur": 4,
    "args": {
      "External id": 25320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25320,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25320, "pid": 5, "tid": 7, "ts": 1716454217661862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631728, "dur": 6,
    "args": {
      "External id": 25320,
      "cbid": 211, "correlation": 25320
    }
  },
  {
    "ph": "s", "id": 25320, "pid": 76337, "tid": -914061504, "ts": 1716454217631728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217661868, "dur": 10,
    "args": {
      "External id": 25323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25323, "pid": 5, "tid": 7, "ts": 1716454217661868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631746, "dur": 6,
    "args": {
      "External id": 25323,
      "cbid": 211, "correlation": 25323
    }
  },
  {
    "ph": "s", "id": 25323, "pid": 76337, "tid": -914061504, "ts": 1716454217631746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217661879, "dur": 6,
    "args": {
      "External id": 25332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25332,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25332, "pid": 5, "tid": 7, "ts": 1716454217661879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631784, "dur": 10,
    "args": {
      "External id": 25332,
      "cbid": 211, "correlation": 25332
    }
  },
  {
    "ph": "s", "id": 25332, "pid": 76337, "tid": -914061504, "ts": 1716454217631784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631854, "dur": 0,
    "args": {
      "External id": 25342,
      "cbid": 317, "correlation": 25342
    }
  },
  {
    "ph": "f", "id": 25342, "pid": 76337, "tid": -914061504, "ts": 1716454217631854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631855, "dur": 0,
    "args": {
      "External id": 25343,
      "cbid": 203, "correlation": 25343
    }
  },
  {
    "ph": "f", "id": 25343, "pid": 76337, "tid": -914061504, "ts": 1716454217631855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631856, "dur": 0,
    "args": {
      "External id": 25344,
      "cbid": 205, "correlation": 25344
    }
  },
  {
    "ph": "f", "id": 25344, "pid": 76337, "tid": -914061504, "ts": 1716454217631856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661886, "dur": 6,
    "args": {
      "External id": 25348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25348, "pid": 5, "tid": 7, "ts": 1716454217661886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631870, "dur": 12,
    "args": {
      "External id": 25348,
      "cbid": 211, "correlation": 25348
    }
  },
  {
    "ph": "s", "id": 25348, "pid": 76337, "tid": -914061504, "ts": 1716454217631870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217661894, "dur": 314,
    "args": {
      "External id": 25350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25350, "pid": 5, "tid": 7, "ts": 1716454217661894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631885, "dur": 5,
    "args": {
      "External id": 25350,
      "cbid": 211, "correlation": 25350
    }
  },
  {
    "ph": "s", "id": 25350, "pid": 76337, "tid": -914061504, "ts": 1716454217631885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217662210, "dur": 1,
    "args": {
      "External id": 25352,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25352,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 25352, "pid": 5, "tid": 7, "ts": 1716454217662210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631896, "dur": 7,
    "args": {
      "External id": 25352,
      "cbid": 51, "correlation": 25352
    }
  },
  {
    "ph": "s", "id": 25352, "pid": 76337, "tid": -914061504, "ts": 1716454217631896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217662214, "dur": 488,
    "args": {
      "External id": 25353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25353,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25353, "pid": 5, "tid": 7, "ts": 1716454217662214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631904, "dur": 6,
    "args": {
      "External id": 25353,
      "cbid": 211, "correlation": 25353
    }
  },
  {
    "ph": "s", "id": 25353, "pid": 76337, "tid": -914061504, "ts": 1716454217631904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217662703, "dur": 5,
    "args": {
      "External id": 25355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25355,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25355, "pid": 5, "tid": 7, "ts": 1716454217662703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631913, "dur": 5,
    "args": {
      "External id": 25355,
      "cbid": 211, "correlation": 25355
    }
  },
  {
    "ph": "s", "id": 25355, "pid": 76337, "tid": -914061504, "ts": 1716454217631913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217662710, "dur": 6,
    "args": {
      "External id": 25361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25361, "pid": 5, "tid": 7, "ts": 1716454217662710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631942, "dur": 9,
    "args": {
      "External id": 25361,
      "cbid": 211, "correlation": 25361
    }
  },
  {
    "ph": "s", "id": 25361, "pid": 76337, "tid": -914061504, "ts": 1716454217631942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217662717, "dur": 3,
    "args": {
      "External id": 25369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25369,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 25369, "pid": 5, "tid": 7, "ts": 1716454217662717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217631995, "dur": 10,
    "args": {
      "External id": 25369,
      "cbid": 211, "correlation": 25369
    }
  },
  {
    "ph": "s", "id": 25369, "pid": 76337, "tid": -914061504, "ts": 1716454217631995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632060, "dur": 1,
    "args": {
      "External id": 25385,
      "cbid": 251, "correlation": 25385
    }
  },
  {
    "ph": "f", "id": 25385, "pid": 76337, "tid": -914061504, "ts": 1716454217632060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632065, "dur": 0,
    "args": {
      "External id": 25387,
      "cbid": 251, "correlation": 25387
    }
  },
  {
    "ph": "f", "id": 25387, "pid": 76337, "tid": -914061504, "ts": 1716454217632065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217662721, "dur": 13,
    "args": {
      "External id": 25388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25388,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25388, "pid": 5, "tid": 7, "ts": 1716454217662721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632067, "dur": 11,
    "args": {
      "External id": 25388,
      "cbid": 211, "correlation": 25388
    }
  },
  {
    "ph": "s", "id": 25388, "pid": 76337, "tid": -914061504, "ts": 1716454217632067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217662735, "dur": 5,
    "args": {
      "External id": 25390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25390,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25390, "pid": 5, "tid": 7, "ts": 1716454217662735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632080, "dur": 6,
    "args": {
      "External id": 25390,
      "cbid": 211, "correlation": 25390
    }
  },
  {
    "ph": "s", "id": 25390, "pid": 76337, "tid": -914061504, "ts": 1716454217632080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217662741, "dur": 6,
    "args": {
      "External id": 25400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25400,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25400, "pid": 5, "tid": 7, "ts": 1716454217662741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632137, "dur": 13,
    "args": {
      "External id": 25400,
      "cbid": 211, "correlation": 25400
    }
  },
  {
    "ph": "s", "id": 25400, "pid": 76337, "tid": -914061504, "ts": 1716454217632137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217662748, "dur": 9,
    "args": {
      "External id": 25420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25420,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25420, "pid": 5, "tid": 7, "ts": 1716454217662748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632204, "dur": 11,
    "args": {
      "External id": 25420,
      "cbid": 211, "correlation": 25420
    }
  },
  {
    "ph": "s", "id": 25420, "pid": 76337, "tid": -914061504, "ts": 1716454217632204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217662759, "dur": 4,
    "args": {
      "External id": 25432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25432,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 25432, "pid": 5, "tid": 7, "ts": 1716454217662759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632224, "dur": 6,
    "args": {
      "External id": 25432,
      "cbid": 211, "correlation": 25432
    }
  },
  {
    "ph": "s", "id": 25432, "pid": 76337, "tid": -914061504, "ts": 1716454217632224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217662764, "dur": 7,
    "args": {
      "External id": 25435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25435, "pid": 5, "tid": 7, "ts": 1716454217662764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632242, "dur": 7,
    "args": {
      "External id": 25435,
      "cbid": 211, "correlation": 25435
    }
  },
  {
    "ph": "s", "id": 25435, "pid": 76337, "tid": -914061504, "ts": 1716454217632242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217662772, "dur": 4,
    "args": {
      "External id": 25444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25444,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25444, "pid": 5, "tid": 7, "ts": 1716454217662772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632282, "dur": 10,
    "args": {
      "External id": 25444,
      "cbid": 211, "correlation": 25444
    }
  },
  {
    "ph": "s", "id": 25444, "pid": 76337, "tid": -914061504, "ts": 1716454217632282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632360, "dur": 0,
    "args": {
      "External id": 25454,
      "cbid": 317, "correlation": 25454
    }
  },
  {
    "ph": "f", "id": 25454, "pid": 76337, "tid": -914061504, "ts": 1716454217632360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632361, "dur": 0,
    "args": {
      "External id": 25455,
      "cbid": 203, "correlation": 25455
    }
  },
  {
    "ph": "f", "id": 25455, "pid": 76337, "tid": -914061504, "ts": 1716454217632361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632362, "dur": 0,
    "args": {
      "External id": 25456,
      "cbid": 205, "correlation": 25456
    }
  },
  {
    "ph": "f", "id": 25456, "pid": 76337, "tid": -914061504, "ts": 1716454217632362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217662778, "dur": 5,
    "args": {
      "External id": 25460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25460,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25460, "pid": 5, "tid": 7, "ts": 1716454217662778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632375, "dur": 13,
    "args": {
      "External id": 25460,
      "cbid": 211, "correlation": 25460
    }
  },
  {
    "ph": "s", "id": 25460, "pid": 76337, "tid": -914061504, "ts": 1716454217632375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217662784, "dur": 159,
    "args": {
      "External id": 25462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25462,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25462, "pid": 5, "tid": 7, "ts": 1716454217662784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632391, "dur": 5,
    "args": {
      "External id": 25462,
      "cbid": 211, "correlation": 25462
    }
  },
  {
    "ph": "s", "id": 25462, "pid": 76337, "tid": -914061504, "ts": 1716454217632391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217662946, "dur": 1,
    "args": {
      "External id": 25464,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25464,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 25464, "pid": 5, "tid": 7, "ts": 1716454217662946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632402, "dur": 6,
    "args": {
      "External id": 25464,
      "cbid": 51, "correlation": 25464
    }
  },
  {
    "ph": "s", "id": 25464, "pid": 76337, "tid": -914061504, "ts": 1716454217632402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217662949, "dur": 254,
    "args": {
      "External id": 25465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25465,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25465, "pid": 5, "tid": 7, "ts": 1716454217662949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632410, "dur": 6,
    "args": {
      "External id": 25465,
      "cbid": 211, "correlation": 25465
    }
  },
  {
    "ph": "s", "id": 25465, "pid": 76337, "tid": -914061504, "ts": 1716454217632410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217663204, "dur": 6,
    "args": {
      "External id": 25467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25467,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25467, "pid": 5, "tid": 7, "ts": 1716454217663204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632420, "dur": 5,
    "args": {
      "External id": 25467,
      "cbid": 211, "correlation": 25467
    }
  },
  {
    "ph": "s", "id": 25467, "pid": 76337, "tid": -914061504, "ts": 1716454217632420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217663212, "dur": 6,
    "args": {
      "External id": 25473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25473, "pid": 5, "tid": 7, "ts": 1716454217663212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632448, "dur": 9,
    "args": {
      "External id": 25473,
      "cbid": 211, "correlation": 25473
    }
  },
  {
    "ph": "s", "id": 25473, "pid": 76337, "tid": -914061504, "ts": 1716454217632448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632508, "dur": 0,
    "args": {
      "External id": 25483,
      "cbid": 317, "correlation": 25483
    }
  },
  {
    "ph": "f", "id": 25483, "pid": 76337, "tid": -914061504, "ts": 1716454217632508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632508, "dur": 0,
    "args": {
      "External id": 25484,
      "cbid": 203, "correlation": 25484
    }
  },
  {
    "ph": "f", "id": 25484, "pid": 76337, "tid": -914061504, "ts": 1716454217632508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632509, "dur": 0,
    "args": {
      "External id": 25485,
      "cbid": 205, "correlation": 25485
    }
  },
  {
    "ph": "f", "id": 25485, "pid": 76337, "tid": -914061504, "ts": 1716454217632509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217663219, "dur": 7,
    "args": {
      "External id": 25489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25489,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25489, "pid": 5, "tid": 7, "ts": 1716454217663219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632521, "dur": 11,
    "args": {
      "External id": 25489,
      "cbid": 211, "correlation": 25489
    }
  },
  {
    "ph": "s", "id": 25489, "pid": 76337, "tid": -914061504, "ts": 1716454217632521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217663228, "dur": 3,
    "args": {
      "External id": 25491,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25491,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5151515151515151
    }
  },
  {
    "ph": "f", "id": 25491, "pid": 5, "tid": 7, "ts": 1716454217663228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632537, "dur": 9,
    "args": {
      "External id": 25491,
      "cbid": 51, "correlation": 25491
    }
  },
  {
    "ph": "s", "id": 25491, "pid": 76337, "tid": -914061504, "ts": 1716454217632537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217663232, "dur": 93,
    "args": {
      "External id": 25492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25492,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 25492, "pid": 5, "tid": 7, "ts": 1716454217663232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632547, "dur": 6,
    "args": {
      "External id": 25492,
      "cbid": 211, "correlation": 25492
    }
  },
  {
    "ph": "s", "id": 25492, "pid": 76337, "tid": -914061504, "ts": 1716454217632547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217663326, "dur": 5,
    "args": {
      "External id": 25494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25494,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25494, "pid": 5, "tid": 7, "ts": 1716454217663326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632558, "dur": 5,
    "args": {
      "External id": 25494,
      "cbid": 211, "correlation": 25494
    }
  },
  {
    "ph": "s", "id": 25494, "pid": 76337, "tid": -914061504, "ts": 1716454217632558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217663333, "dur": 6,
    "args": {
      "External id": 25500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25500, "pid": 5, "tid": 7, "ts": 1716454217663333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632584, "dur": 8,
    "args": {
      "External id": 25500,
      "cbid": 211, "correlation": 25500
    }
  },
  {
    "ph": "s", "id": 25500, "pid": 76337, "tid": -914061504, "ts": 1716454217632584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217663341, "dur": 5,
    "args": {
      "External id": 25508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25508,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25508, "pid": 5, "tid": 7, "ts": 1716454217663341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632614, "dur": 8,
    "args": {
      "External id": 25508,
      "cbid": 211, "correlation": 25508
    }
  },
  {
    "ph": "s", "id": 25508, "pid": 76337, "tid": -914061504, "ts": 1716454217632614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217663347, "dur": 4,
    "args": {
      "External id": 25516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25516,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25516, "pid": 5, "tid": 7, "ts": 1716454217663347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632643, "dur": 8,
    "args": {
      "External id": 25516,
      "cbid": 211, "correlation": 25516
    }
  },
  {
    "ph": "s", "id": 25516, "pid": 76337, "tid": -914061504, "ts": 1716454217632643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217663352, "dur": 11,
    "args": {
      "External id": 25525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25525,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25525, "pid": 5, "tid": 7, "ts": 1716454217663352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632686, "dur": 11,
    "args": {
      "External id": 25525,
      "cbid": 211, "correlation": 25525
    }
  },
  {
    "ph": "s", "id": 25525, "pid": 76337, "tid": -914061504, "ts": 1716454217632686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217663365, "dur": 12,
    "args": {
      "External id": 25545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25545,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25545, "pid": 5, "tid": 7, "ts": 1716454217663365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632770, "dur": 12,
    "args": {
      "External id": 25545,
      "cbid": 211, "correlation": 25545
    }
  },
  {
    "ph": "s", "id": 25545, "pid": 76337, "tid": -914061504, "ts": 1716454217632770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217663378, "dur": 4,
    "args": {
      "External id": 25557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25557,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25557, "pid": 5, "tid": 7, "ts": 1716454217663378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632792, "dur": 6,
    "args": {
      "External id": 25557,
      "cbid": 211, "correlation": 25557
    }
  },
  {
    "ph": "s", "id": 25557, "pid": 76337, "tid": -914061504, "ts": 1716454217632792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217663383, "dur": 11,
    "args": {
      "External id": 25560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25560, "pid": 5, "tid": 7, "ts": 1716454217663383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632811, "dur": 7,
    "args": {
      "External id": 25560,
      "cbid": 211, "correlation": 25560
    }
  },
  {
    "ph": "s", "id": 25560, "pid": 76337, "tid": -914061504, "ts": 1716454217632811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217663395, "dur": 6,
    "args": {
      "External id": 25569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25569,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25569, "pid": 5, "tid": 7, "ts": 1716454217663395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632849, "dur": 10,
    "args": {
      "External id": 25569,
      "cbid": 211, "correlation": 25569
    }
  },
  {
    "ph": "s", "id": 25569, "pid": 76337, "tid": -914061504, "ts": 1716454217632849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632902, "dur": 0,
    "args": {
      "External id": 25579,
      "cbid": 317, "correlation": 25579
    }
  },
  {
    "ph": "f", "id": 25579, "pid": 76337, "tid": -914061504, "ts": 1716454217632902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632903, "dur": 0,
    "args": {
      "External id": 25580,
      "cbid": 203, "correlation": 25580
    }
  },
  {
    "ph": "f", "id": 25580, "pid": 76337, "tid": -914061504, "ts": 1716454217632903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632903, "dur": 0,
    "args": {
      "External id": 25581,
      "cbid": 205, "correlation": 25581
    }
  },
  {
    "ph": "f", "id": 25581, "pid": 76337, "tid": -914061504, "ts": 1716454217632903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217663403, "dur": 6,
    "args": {
      "External id": 25585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25585, "pid": 5, "tid": 7, "ts": 1716454217663403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632917, "dur": 12,
    "args": {
      "External id": 25585,
      "cbid": 211, "correlation": 25585
    }
  },
  {
    "ph": "s", "id": 25585, "pid": 76337, "tid": -914061504, "ts": 1716454217632917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217663410, "dur": 313,
    "args": {
      "External id": 25587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25587, "pid": 5, "tid": 7, "ts": 1716454217663410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632932, "dur": 5,
    "args": {
      "External id": 25587,
      "cbid": 211, "correlation": 25587
    }
  },
  {
    "ph": "s", "id": 25587, "pid": 76337, "tid": -914061504, "ts": 1716454217632932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217663726, "dur": 1,
    "args": {
      "External id": 25589,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25589,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 25589, "pid": 5, "tid": 7, "ts": 1716454217663726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632942, "dur": 6,
    "args": {
      "External id": 25589,
      "cbid": 51, "correlation": 25589
    }
  },
  {
    "ph": "s", "id": 25589, "pid": 76337, "tid": -914061504, "ts": 1716454217632942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217663729, "dur": 488,
    "args": {
      "External id": 25590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25590,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25590, "pid": 5, "tid": 7, "ts": 1716454217663729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632950, "dur": 6,
    "args": {
      "External id": 25590,
      "cbid": 211, "correlation": 25590
    }
  },
  {
    "ph": "s", "id": 25590, "pid": 76337, "tid": -914061504, "ts": 1716454217632950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664218, "dur": 6,
    "args": {
      "External id": 25592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25592, "pid": 5, "tid": 7, "ts": 1716454217664218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632960, "dur": 5,
    "args": {
      "External id": 25592,
      "cbid": 211, "correlation": 25592
    }
  },
  {
    "ph": "s", "id": 25592, "pid": 76337, "tid": -914061504, "ts": 1716454217632960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217664225, "dur": 6,
    "args": {
      "External id": 25598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25598, "pid": 5, "tid": 7, "ts": 1716454217664225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217632997, "dur": 10,
    "args": {
      "External id": 25598,
      "cbid": 211, "correlation": 25598
    }
  },
  {
    "ph": "s", "id": 25598, "pid": 76337, "tid": -914061504, "ts": 1716454217632997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217664233, "dur": 3,
    "args": {
      "External id": 25606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25606,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 25606, "pid": 5, "tid": 7, "ts": 1716454217664233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633042, "dur": 9,
    "args": {
      "External id": 25606,
      "cbid": 211, "correlation": 25606
    }
  },
  {
    "ph": "s", "id": 25606, "pid": 76337, "tid": -914061504, "ts": 1716454217633042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633116, "dur": 1,
    "args": {
      "External id": 25622,
      "cbid": 251, "correlation": 25622
    }
  },
  {
    "ph": "f", "id": 25622, "pid": 76337, "tid": -914061504, "ts": 1716454217633116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633121, "dur": 0,
    "args": {
      "External id": 25624,
      "cbid": 251, "correlation": 25624
    }
  },
  {
    "ph": "f", "id": 25624, "pid": 76337, "tid": -914061504, "ts": 1716454217633121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217664237, "dur": 12,
    "args": {
      "External id": 25625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25625,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25625, "pid": 5, "tid": 7, "ts": 1716454217664237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633123, "dur": 13,
    "args": {
      "External id": 25625,
      "cbid": 211, "correlation": 25625
    }
  },
  {
    "ph": "s", "id": 25625, "pid": 76337, "tid": -914061504, "ts": 1716454217633123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217664250, "dur": 5,
    "args": {
      "External id": 25627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25627,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25627, "pid": 5, "tid": 7, "ts": 1716454217664250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633137, "dur": 5,
    "args": {
      "External id": 25627,
      "cbid": 211, "correlation": 25627
    }
  },
  {
    "ph": "s", "id": 25627, "pid": 76337, "tid": -914061504, "ts": 1716454217633137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217664257, "dur": 6,
    "args": {
      "External id": 25637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25637, "pid": 5, "tid": 7, "ts": 1716454217664257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633195, "dur": 12,
    "args": {
      "External id": 25637,
      "cbid": 211, "correlation": 25637
    }
  },
  {
    "ph": "s", "id": 25637, "pid": 76337, "tid": -914061504, "ts": 1716454217633195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664264, "dur": 9,
    "args": {
      "External id": 25657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25657,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25657, "pid": 5, "tid": 7, "ts": 1716454217664264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633261, "dur": 10,
    "args": {
      "External id": 25657,
      "cbid": 211, "correlation": 25657
    }
  },
  {
    "ph": "s", "id": 25657, "pid": 76337, "tid": -914061504, "ts": 1716454217633261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217664275, "dur": 3,
    "args": {
      "External id": 25669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25669,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 25669, "pid": 5, "tid": 7, "ts": 1716454217664275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633281, "dur": 6,
    "args": {
      "External id": 25669,
      "cbid": 211, "correlation": 25669
    }
  },
  {
    "ph": "s", "id": 25669, "pid": 76337, "tid": -914061504, "ts": 1716454217633281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217664279, "dur": 7,
    "args": {
      "External id": 25672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25672, "pid": 5, "tid": 7, "ts": 1716454217664279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633300, "dur": 6,
    "args": {
      "External id": 25672,
      "cbid": 211, "correlation": 25672
    }
  },
  {
    "ph": "s", "id": 25672, "pid": 76337, "tid": -914061504, "ts": 1716454217633300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217664287, "dur": 4,
    "args": {
      "External id": 25681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25681,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25681, "pid": 5, "tid": 7, "ts": 1716454217664287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633340, "dur": 10,
    "args": {
      "External id": 25681,
      "cbid": 211, "correlation": 25681
    }
  },
  {
    "ph": "s", "id": 25681, "pid": 76337, "tid": -914061504, "ts": 1716454217633340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633416, "dur": 0,
    "args": {
      "External id": 25691,
      "cbid": 317, "correlation": 25691
    }
  },
  {
    "ph": "f", "id": 25691, "pid": 76337, "tid": -914061504, "ts": 1716454217633416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633417, "dur": 0,
    "args": {
      "External id": 25692,
      "cbid": 203, "correlation": 25692
    }
  },
  {
    "ph": "f", "id": 25692, "pid": 76337, "tid": -914061504, "ts": 1716454217633417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633418, "dur": 0,
    "args": {
      "External id": 25693,
      "cbid": 205, "correlation": 25693
    }
  },
  {
    "ph": "f", "id": 25693, "pid": 76337, "tid": -914061504, "ts": 1716454217633418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664293, "dur": 5,
    "args": {
      "External id": 25697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25697,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25697, "pid": 5, "tid": 7, "ts": 1716454217664293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633431, "dur": 13,
    "args": {
      "External id": 25697,
      "cbid": 211, "correlation": 25697
    }
  },
  {
    "ph": "s", "id": 25697, "pid": 76337, "tid": -914061504, "ts": 1716454217633431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664299, "dur": 159,
    "args": {
      "External id": 25699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25699, "pid": 5, "tid": 7, "ts": 1716454217664299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633447, "dur": 5,
    "args": {
      "External id": 25699,
      "cbid": 211, "correlation": 25699
    }
  },
  {
    "ph": "s", "id": 25699, "pid": 76337, "tid": -914061504, "ts": 1716454217633447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217664461, "dur": 1,
    "args": {
      "External id": 25701,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25701,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 25701, "pid": 5, "tid": 7, "ts": 1716454217664461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633458, "dur": 6,
    "args": {
      "External id": 25701,
      "cbid": 51, "correlation": 25701
    }
  },
  {
    "ph": "s", "id": 25701, "pid": 76337, "tid": -914061504, "ts": 1716454217633458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217664464, "dur": 254,
    "args": {
      "External id": 25702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25702,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25702, "pid": 5, "tid": 7, "ts": 1716454217664464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633465, "dur": 6,
    "args": {
      "External id": 25702,
      "cbid": 211, "correlation": 25702
    }
  },
  {
    "ph": "s", "id": 25702, "pid": 76337, "tid": -914061504, "ts": 1716454217633465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664720, "dur": 6,
    "args": {
      "External id": 25704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25704, "pid": 5, "tid": 7, "ts": 1716454217664720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633475, "dur": 5,
    "args": {
      "External id": 25704,
      "cbid": 211, "correlation": 25704
    }
  },
  {
    "ph": "s", "id": 25704, "pid": 76337, "tid": -914061504, "ts": 1716454217633475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217664726, "dur": 6,
    "args": {
      "External id": 25710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25710, "pid": 5, "tid": 7, "ts": 1716454217664726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633504, "dur": 8,
    "args": {
      "External id": 25710,
      "cbid": 211, "correlation": 25710
    }
  },
  {
    "ph": "s", "id": 25710, "pid": 76337, "tid": -914061504, "ts": 1716454217633504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633562, "dur": 0,
    "args": {
      "External id": 25720,
      "cbid": 317, "correlation": 25720
    }
  },
  {
    "ph": "f", "id": 25720, "pid": 76337, "tid": -914061504, "ts": 1716454217633562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633563, "dur": 0,
    "args": {
      "External id": 25721,
      "cbid": 203, "correlation": 25721
    }
  },
  {
    "ph": "f", "id": 25721, "pid": 76337, "tid": -914061504, "ts": 1716454217633563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633563, "dur": 0,
    "args": {
      "External id": 25722,
      "cbid": 205, "correlation": 25722
    }
  },
  {
    "ph": "f", "id": 25722, "pid": 76337, "tid": -914061504, "ts": 1716454217633563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664734, "dur": 8,
    "args": {
      "External id": 25726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25726, "pid": 5, "tid": 7, "ts": 1716454217664734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633575, "dur": 12,
    "args": {
      "External id": 25726,
      "cbid": 211, "correlation": 25726
    }
  },
  {
    "ph": "s", "id": 25726, "pid": 76337, "tid": -914061504, "ts": 1716454217633575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217664743, "dur": 3,
    "args": {
      "External id": 25728,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25728,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5004688965301656
    }
  },
  {
    "ph": "f", "id": 25728, "pid": 5, "tid": 7, "ts": 1716454217664743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633592, "dur": 10,
    "args": {
      "External id": 25728,
      "cbid": 51, "correlation": 25728
    }
  },
  {
    "ph": "s", "id": 25728, "pid": 76337, "tid": -914061504, "ts": 1716454217633592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217664747, "dur": 93,
    "args": {
      "External id": 25729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25729,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 25729, "pid": 5, "tid": 7, "ts": 1716454217664747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633602, "dur": 6,
    "args": {
      "External id": 25729,
      "cbid": 211, "correlation": 25729
    }
  },
  {
    "ph": "s", "id": 25729, "pid": 76337, "tid": -914061504, "ts": 1716454217633602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664842, "dur": 5,
    "args": {
      "External id": 25731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25731,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25731, "pid": 5, "tid": 7, "ts": 1716454217664842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633612, "dur": 5,
    "args": {
      "External id": 25731,
      "cbid": 211, "correlation": 25731
    }
  },
  {
    "ph": "s", "id": 25731, "pid": 76337, "tid": -914061504, "ts": 1716454217633612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217664849, "dur": 6,
    "args": {
      "External id": 25737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25737, "pid": 5, "tid": 7, "ts": 1716454217664849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633638, "dur": 9,
    "args": {
      "External id": 25737,
      "cbid": 211, "correlation": 25737
    }
  },
  {
    "ph": "s", "id": 25737, "pid": 76337, "tid": -914061504, "ts": 1716454217633638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217664856, "dur": 5,
    "args": {
      "External id": 25745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25745,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25745, "pid": 5, "tid": 7, "ts": 1716454217664856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633668, "dur": 8,
    "args": {
      "External id": 25745,
      "cbid": 211, "correlation": 25745
    }
  },
  {
    "ph": "s", "id": 25745, "pid": 76337, "tid": -914061504, "ts": 1716454217633668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217664862, "dur": 4,
    "args": {
      "External id": 25753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25753,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 25753, "pid": 5, "tid": 7, "ts": 1716454217664862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633696, "dur": 9,
    "args": {
      "External id": 25753,
      "cbid": 211, "correlation": 25753
    }
  },
  {
    "ph": "s", "id": 25753, "pid": 76337, "tid": -914061504, "ts": 1716454217633696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454217664868, "dur": 14,
    "args": {
      "External id": 25764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25764,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25764, "pid": 5, "tid": 7, "ts": 1716454217664868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633777, "dur": 13,
    "args": {
      "External id": 25764,
      "cbid": 211, "correlation": 25764
    }
  },
  {
    "ph": "s", "id": 25764, "pid": 76337, "tid": -914061504, "ts": 1716454217633777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633833, "dur": 0,
    "args": {
      "External id": 25774,
      "cbid": 317, "correlation": 25774
    }
  },
  {
    "ph": "f", "id": 25774, "pid": 76337, "tid": -914061504, "ts": 1716454217633833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633834, "dur": 0,
    "args": {
      "External id": 25775,
      "cbid": 203, "correlation": 25775
    }
  },
  {
    "ph": "f", "id": 25775, "pid": 76337, "tid": -914061504, "ts": 1716454217633834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633834, "dur": 0,
    "args": {
      "External id": 25776,
      "cbid": 205, "correlation": 25776
    }
  },
  {
    "ph": "f", "id": 25776, "pid": 76337, "tid": -914061504, "ts": 1716454217633834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664883, "dur": 9,
    "args": {
      "External id": 25780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25780,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25780, "pid": 5, "tid": 7, "ts": 1716454217664883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633848, "dur": 12,
    "args": {
      "External id": 25780,
      "cbid": 211, "correlation": 25780
    }
  },
  {
    "ph": "s", "id": 25780, "pid": 76337, "tid": -914061504, "ts": 1716454217633848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217664893, "dur": 160,
    "args": {
      "External id": 25782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25782,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25782, "pid": 5, "tid": 7, "ts": 1716454217664893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633862, "dur": 5,
    "args": {
      "External id": 25782,
      "cbid": 211, "correlation": 25782
    }
  },
  {
    "ph": "s", "id": 25782, "pid": 76337, "tid": -914061504, "ts": 1716454217633862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217665055, "dur": 1,
    "args": {
      "External id": 25784,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25784,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 25784, "pid": 5, "tid": 7, "ts": 1716454217665055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633873, "dur": 6,
    "args": {
      "External id": 25784,
      "cbid": 51, "correlation": 25784
    }
  },
  {
    "ph": "s", "id": 25784, "pid": 76337, "tid": -914061504, "ts": 1716454217633873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217665059, "dur": 641,
    "args": {
      "External id": 25785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25785,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25785, "pid": 5, "tid": 7, "ts": 1716454217665059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633880, "dur": 6,
    "args": {
      "External id": 25785,
      "cbid": 211, "correlation": 25785
    }
  },
  {
    "ph": "s", "id": 25785, "pid": 76337, "tid": -914061504, "ts": 1716454217633880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217665701, "dur": 12,
    "args": {
      "External id": 25787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25787,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25787, "pid": 5, "tid": 7, "ts": 1716454217665701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633890, "dur": 5,
    "args": {
      "External id": 25787,
      "cbid": 211, "correlation": 25787
    }
  },
  {
    "ph": "s", "id": 25787, "pid": 76337, "tid": -914061504, "ts": 1716454217633890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217665714, "dur": 14,
    "args": {
      "External id": 25793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25793,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25793, "pid": 5, "tid": 7, "ts": 1716454217665714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217633918, "dur": 8,
    "args": {
      "External id": 25793,
      "cbid": 211, "correlation": 25793
    }
  },
  {
    "ph": "s", "id": 25793, "pid": 76337, "tid": -914061504, "ts": 1716454217633918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217665730, "dur": 29,
    "args": {
      "External id": 25802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25802,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25802, "pid": 5, "tid": 7, "ts": 1716454217665730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634050, "dur": 15,
    "args": {
      "External id": 25802,
      "cbid": 211, "correlation": 25802
    }
  },
  {
    "ph": "s", "id": 25802, "pid": 76337, "tid": -914061504, "ts": 1716454217634050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217665761, "dur": 29,
    "args": {
      "External id": 25822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25822,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25822, "pid": 5, "tid": 7, "ts": 1716454217665761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634133, "dur": 12,
    "args": {
      "External id": 25822,
      "cbid": 211, "correlation": 25822
    }
  },
  {
    "ph": "s", "id": 25822, "pid": 76337, "tid": -914061504, "ts": 1716454217634133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217665791, "dur": 4,
    "args": {
      "External id": 25834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25834,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25834, "pid": 5, "tid": 7, "ts": 1716454217665791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634164, "dur": 8,
    "args": {
      "External id": 25834,
      "cbid": 211, "correlation": 25834
    }
  },
  {
    "ph": "s", "id": 25834, "pid": 76337, "tid": -914061504, "ts": 1716454217634164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217665797, "dur": 31,
    "args": {
      "External id": 25837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25837,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25837, "pid": 5, "tid": 7, "ts": 1716454217665797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634185, "dur": 6,
    "args": {
      "External id": 25837,
      "cbid": 211, "correlation": 25837
    }
  },
  {
    "ph": "s", "id": 25837, "pid": 76337, "tid": -914061504, "ts": 1716454217634185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217665828, "dur": 21,
    "args": {
      "External id": 25846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25846,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25846, "pid": 5, "tid": 7, "ts": 1716454217665828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634227, "dur": 10,
    "args": {
      "External id": 25846,
      "cbid": 211, "correlation": 25846
    }
  },
  {
    "ph": "s", "id": 25846, "pid": 76337, "tid": -914061504, "ts": 1716454217634227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634302, "dur": 0,
    "args": {
      "External id": 25856,
      "cbid": 317, "correlation": 25856
    }
  },
  {
    "ph": "f", "id": 25856, "pid": 76337, "tid": -914061504, "ts": 1716454217634302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634303, "dur": 0,
    "args": {
      "External id": 25857,
      "cbid": 203, "correlation": 25857
    }
  },
  {
    "ph": "f", "id": 25857, "pid": 76337, "tid": -914061504, "ts": 1716454217634303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634303, "dur": 0,
    "args": {
      "External id": 25858,
      "cbid": 205, "correlation": 25858
    }
  },
  {
    "ph": "f", "id": 25858, "pid": 76337, "tid": -914061504, "ts": 1716454217634303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217665851, "dur": 21,
    "args": {
      "External id": 25862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25862,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25862, "pid": 5, "tid": 7, "ts": 1716454217665851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634332, "dur": 13,
    "args": {
      "External id": 25862,
      "cbid": 211, "correlation": 25862
    }
  },
  {
    "ph": "s", "id": 25862, "pid": 76337, "tid": -914061504, "ts": 1716454217634332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217665874, "dur": 315,
    "args": {
      "External id": 25864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25864,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25864, "pid": 5, "tid": 7, "ts": 1716454217665874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634347, "dur": 5,
    "args": {
      "External id": 25864,
      "cbid": 211, "correlation": 25864
    }
  },
  {
    "ph": "s", "id": 25864, "pid": 76337, "tid": -914061504, "ts": 1716454217634347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217666191, "dur": 1,
    "args": {
      "External id": 25866,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25866,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 25866, "pid": 5, "tid": 7, "ts": 1716454217666191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634358, "dur": 6,
    "args": {
      "External id": 25866,
      "cbid": 51, "correlation": 25866
    }
  },
  {
    "ph": "s", "id": 25866, "pid": 76337, "tid": -914061504, "ts": 1716454217634358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217666195, "dur": 1222,
    "args": {
      "External id": 25867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25867,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25867, "pid": 5, "tid": 7, "ts": 1716454217666195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634366, "dur": 7,
    "args": {
      "External id": 25867,
      "cbid": 211, "correlation": 25867
    }
  },
  {
    "ph": "s", "id": 25867, "pid": 76337, "tid": -914061504, "ts": 1716454217634366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217667419, "dur": 12,
    "args": {
      "External id": 25869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25869,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25869, "pid": 5, "tid": 7, "ts": 1716454217667419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634376, "dur": 5,
    "args": {
      "External id": 25869,
      "cbid": 211, "correlation": 25869
    }
  },
  {
    "ph": "s", "id": 25869, "pid": 76337, "tid": -914061504, "ts": 1716454217634376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217667432, "dur": 14,
    "args": {
      "External id": 25875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25875, "pid": 5, "tid": 7, "ts": 1716454217667432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634413, "dur": 11,
    "args": {
      "External id": 25875,
      "cbid": 211, "correlation": 25875
    }
  },
  {
    "ph": "s", "id": 25875, "pid": 76337, "tid": -914061504, "ts": 1716454217634413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217667448, "dur": 3,
    "args": {
      "External id": 25883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25883,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 25883, "pid": 5, "tid": 7, "ts": 1716454217667448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634462, "dur": 10,
    "args": {
      "External id": 25883,
      "cbid": 211, "correlation": 25883
    }
  },
  {
    "ph": "s", "id": 25883, "pid": 76337, "tid": -914061504, "ts": 1716454217634462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634536, "dur": 1,
    "args": {
      "External id": 25899,
      "cbid": 251, "correlation": 25899
    }
  },
  {
    "ph": "f", "id": 25899, "pid": 76337, "tid": -914061504, "ts": 1716454217634536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634541, "dur": 0,
    "args": {
      "External id": 25901,
      "cbid": 251, "correlation": 25901
    }
  },
  {
    "ph": "f", "id": 25901, "pid": 76337, "tid": -914061504, "ts": 1716454217634541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217667453, "dur": 13,
    "args": {
      "External id": 25902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25902,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25902, "pid": 5, "tid": 7, "ts": 1716454217667453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634543, "dur": 12,
    "args": {
      "External id": 25902,
      "cbid": 211, "correlation": 25902
    }
  },
  {
    "ph": "s", "id": 25902, "pid": 76337, "tid": -914061504, "ts": 1716454217634543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217667467, "dur": 5,
    "args": {
      "External id": 25904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25904,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25904, "pid": 5, "tid": 7, "ts": 1716454217667467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634557, "dur": 5,
    "args": {
      "External id": 25904,
      "cbid": 211, "correlation": 25904
    }
  },
  {
    "ph": "s", "id": 25904, "pid": 76337, "tid": -914061504, "ts": 1716454217634557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217667473, "dur": 17,
    "args": {
      "External id": 25914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25914, "pid": 5, "tid": 7, "ts": 1716454217667473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634615, "dur": 12,
    "args": {
      "External id": 25914,
      "cbid": 211, "correlation": 25914
    }
  },
  {
    "ph": "s", "id": 25914, "pid": 76337, "tid": -914061504, "ts": 1716454217634615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217667491, "dur": 17,
    "args": {
      "External id": 25934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25934,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 25934, "pid": 5, "tid": 7, "ts": 1716454217667491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634681, "dur": 11,
    "args": {
      "External id": 25934,
      "cbid": 211, "correlation": 25934
    }
  },
  {
    "ph": "s", "id": 25934, "pid": 76337, "tid": -914061504, "ts": 1716454217634681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217667510, "dur": 4,
    "args": {
      "External id": 25946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25946,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 25946, "pid": 5, "tid": 7, "ts": 1716454217667510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634703, "dur": 6,
    "args": {
      "External id": 25946,
      "cbid": 211, "correlation": 25946
    }
  },
  {
    "ph": "s", "id": 25946, "pid": 76337, "tid": -914061504, "ts": 1716454217634703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217667515, "dur": 16,
    "args": {
      "External id": 25949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25949,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25949, "pid": 5, "tid": 7, "ts": 1716454217667515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634720, "dur": 6,
    "args": {
      "External id": 25949,
      "cbid": 211, "correlation": 25949
    }
  },
  {
    "ph": "s", "id": 25949, "pid": 76337, "tid": -914061504, "ts": 1716454217634720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217667533, "dur": 11,
    "args": {
      "External id": 25958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25958,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25958, "pid": 5, "tid": 7, "ts": 1716454217667533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634761, "dur": 10,
    "args": {
      "External id": 25958,
      "cbid": 211, "correlation": 25958
    }
  },
  {
    "ph": "s", "id": 25958, "pid": 76337, "tid": -914061504, "ts": 1716454217634761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634824, "dur": 0,
    "args": {
      "External id": 25968,
      "cbid": 317, "correlation": 25968
    }
  },
  {
    "ph": "f", "id": 25968, "pid": 76337, "tid": -914061504, "ts": 1716454217634824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634825, "dur": 0,
    "args": {
      "External id": 25969,
      "cbid": 203, "correlation": 25969
    }
  },
  {
    "ph": "f", "id": 25969, "pid": 76337, "tid": -914061504, "ts": 1716454217634825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634825, "dur": 0,
    "args": {
      "External id": 25970,
      "cbid": 205, "correlation": 25970
    }
  },
  {
    "ph": "f", "id": 25970, "pid": 76337, "tid": -914061504, "ts": 1716454217634825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217667545, "dur": 11,
    "args": {
      "External id": 25974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25974,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25974, "pid": 5, "tid": 7, "ts": 1716454217667545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634853, "dur": 13,
    "args": {
      "External id": 25974,
      "cbid": 211, "correlation": 25974
    }
  },
  {
    "ph": "s", "id": 25974, "pid": 76337, "tid": -914061504, "ts": 1716454217634853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217667557, "dur": 160,
    "args": {
      "External id": 25976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25976,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25976, "pid": 5, "tid": 7, "ts": 1716454217667557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634868, "dur": 5,
    "args": {
      "External id": 25976,
      "cbid": 211, "correlation": 25976
    }
  },
  {
    "ph": "s", "id": 25976, "pid": 76337, "tid": -914061504, "ts": 1716454217634868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217667720, "dur": 1,
    "args": {
      "External id": 25978,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 25978,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 25978, "pid": 5, "tid": 7, "ts": 1716454217667720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634879, "dur": 7,
    "args": {
      "External id": 25978,
      "cbid": 51, "correlation": 25978
    }
  },
  {
    "ph": "s", "id": 25978, "pid": 76337, "tid": -914061504, "ts": 1716454217634879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217667724, "dur": 640,
    "args": {
      "External id": 25979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25979,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 25979, "pid": 5, "tid": 7, "ts": 1716454217667724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634887, "dur": 6,
    "args": {
      "External id": 25979,
      "cbid": 211, "correlation": 25979
    }
  },
  {
    "ph": "s", "id": 25979, "pid": 76337, "tid": -914061504, "ts": 1716454217634887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217668365, "dur": 13,
    "args": {
      "External id": 25981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25981, "pid": 5, "tid": 7, "ts": 1716454217668365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634897, "dur": 5,
    "args": {
      "External id": 25981,
      "cbid": 211, "correlation": 25981
    }
  },
  {
    "ph": "s", "id": 25981, "pid": 76337, "tid": -914061504, "ts": 1716454217634897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217668379, "dur": 14,
    "args": {
      "External id": 25987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 25987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 25987, "pid": 5, "tid": 7, "ts": 1716454217668379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634926, "dur": 9,
    "args": {
      "External id": 25987,
      "cbid": 211, "correlation": 25987
    }
  },
  {
    "ph": "s", "id": 25987, "pid": 76337, "tid": -914061504, "ts": 1716454217634926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634994, "dur": 0,
    "args": {
      "External id": 25997,
      "cbid": 317, "correlation": 25997
    }
  },
  {
    "ph": "f", "id": 25997, "pid": 76337, "tid": -914061504, "ts": 1716454217634994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634995, "dur": 0,
    "args": {
      "External id": 25998,
      "cbid": 203, "correlation": 25998
    }
  },
  {
    "ph": "f", "id": 25998, "pid": 76337, "tid": -914061504, "ts": 1716454217634995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217634995, "dur": 0,
    "args": {
      "External id": 25999,
      "cbid": 205, "correlation": 25999
    }
  },
  {
    "ph": "f", "id": 25999, "pid": 76337, "tid": -914061504, "ts": 1716454217634995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217668395, "dur": 21,
    "args": {
      "External id": 26003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26003,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26003, "pid": 5, "tid": 7, "ts": 1716454217668395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635019, "dur": 13,
    "args": {
      "External id": 26003,
      "cbid": 211, "correlation": 26003
    }
  },
  {
    "ph": "s", "id": 26003, "pid": 76337, "tid": -914061504, "ts": 1716454217635019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217668417, "dur": 4,
    "args": {
      "External id": 26005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26005, "pid": 5, "tid": 7, "ts": 1716454217668417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635038, "dur": 7,
    "args": {
      "External id": 26005,
      "cbid": 211, "correlation": 26005
    }
  },
  {
    "ph": "s", "id": 26005, "pid": 76337, "tid": -914061504, "ts": 1716454217635038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635049, "dur": 0,
    "args": {
      "External id": 26006,
      "cbid": 51, "correlation": 26006
    }
  },
  {
    "ph": "s", "id": 26006, "pid": 76337, "tid": -914061504, "ts": 1716454217635049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217668422, "dur": 172,
    "args": {
      "External id": 26007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26007,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 26007, "pid": 5, "tid": 7, "ts": 1716454217668422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635050, "dur": 6,
    "args": {
      "External id": 26007,
      "cbid": 211, "correlation": 26007
    }
  },
  {
    "ph": "s", "id": 26007, "pid": 76337, "tid": -914061504, "ts": 1716454217635050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217668596, "dur": 16,
    "args": {
      "External id": 26012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26012,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26012, "pid": 5, "tid": 7, "ts": 1716454217668596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635078, "dur": 9,
    "args": {
      "External id": 26012,
      "cbid": 211, "correlation": 26012
    }
  },
  {
    "ph": "s", "id": 26012, "pid": 76337, "tid": -914061504, "ts": 1716454217635078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217668613, "dur": 12,
    "args": {
      "External id": 26020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26020,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26020, "pid": 5, "tid": 7, "ts": 1716454217668613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635109, "dur": 8,
    "args": {
      "External id": 26020,
      "cbid": 211, "correlation": 26020
    }
  },
  {
    "ph": "s", "id": 26020, "pid": 76337, "tid": -914061504, "ts": 1716454217635109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217668626, "dur": 10,
    "args": {
      "External id": 26028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26028,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26028, "pid": 5, "tid": 7, "ts": 1716454217668626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635138, "dur": 9,
    "args": {
      "External id": 26028,
      "cbid": 211, "correlation": 26028
    }
  },
  {
    "ph": "s", "id": 26028, "pid": 76337, "tid": -914061504, "ts": 1716454217635138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217668637, "dur": 18,
    "args": {
      "External id": 26048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26048,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 26048, "pid": 5, "tid": 7, "ts": 1716454217668637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635221, "dur": 12,
    "args": {
      "External id": 26048,
      "cbid": 211, "correlation": 26048
    }
  },
  {
    "ph": "s", "id": 26048, "pid": 76337, "tid": -914061504, "ts": 1716454217635221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217668657, "dur": 5,
    "args": {
      "External id": 26060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26060,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 26060, "pid": 5, "tid": 7, "ts": 1716454217668657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635243, "dur": 6,
    "args": {
      "External id": 26060,
      "cbid": 211, "correlation": 26060
    }
  },
  {
    "ph": "s", "id": 26060, "pid": 76337, "tid": -914061504, "ts": 1716454217635243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217668663, "dur": 16,
    "args": {
      "External id": 26063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26063,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26063, "pid": 5, "tid": 7, "ts": 1716454217668663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635261, "dur": 6,
    "args": {
      "External id": 26063,
      "cbid": 211, "correlation": 26063
    }
  },
  {
    "ph": "s", "id": 26063, "pid": 76337, "tid": -914061504, "ts": 1716454217635261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635319, "dur": 0,
    "args": {
      "External id": 26074,
      "cbid": 317, "correlation": 26074
    }
  },
  {
    "ph": "f", "id": 26074, "pid": 76337, "tid": -914061504, "ts": 1716454217635319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635320, "dur": 0,
    "args": {
      "External id": 26075,
      "cbid": 203, "correlation": 26075
    }
  },
  {
    "ph": "f", "id": 26075, "pid": 76337, "tid": -914061504, "ts": 1716454217635320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635321, "dur": 0,
    "args": {
      "External id": 26076,
      "cbid": 205, "correlation": 26076
    }
  },
  {
    "ph": "f", "id": 26076, "pid": 76337, "tid": -914061504, "ts": 1716454217635321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217668680, "dur": 11,
    "args": {
      "External id": 26080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26080,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26080, "pid": 5, "tid": 7, "ts": 1716454217668680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635334, "dur": 12,
    "args": {
      "External id": 26080,
      "cbid": 211, "correlation": 26080
    }
  },
  {
    "ph": "s", "id": 26080, "pid": 76337, "tid": -914061504, "ts": 1716454217635334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217668692, "dur": 3,
    "args": {
      "External id": 26082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26082,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26082, "pid": 5, "tid": 7, "ts": 1716454217668692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635350, "dur": 6,
    "args": {
      "External id": 26082,
      "cbid": 211, "correlation": 26082
    }
  },
  {
    "ph": "s", "id": 26082, "pid": 76337, "tid": -914061504, "ts": 1716454217635350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635359, "dur": 0,
    "args": {
      "External id": 26083,
      "cbid": 51, "correlation": 26083
    }
  },
  {
    "ph": "s", "id": 26083, "pid": 76337, "tid": -914061504, "ts": 1716454217635359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217668697, "dur": 89,
    "args": {
      "External id": 26084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26084,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 26084, "pid": 5, "tid": 7, "ts": 1716454217668697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635360, "dur": 5,
    "args": {
      "External id": 26084,
      "cbid": 211, "correlation": 26084
    }
  },
  {
    "ph": "s", "id": 26084, "pid": 76337, "tid": -914061504, "ts": 1716454217635360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217668787, "dur": 15,
    "args": {
      "External id": 26089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26089, "pid": 5, "tid": 7, "ts": 1716454217668787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635386, "dur": 8,
    "args": {
      "External id": 26089,
      "cbid": 211, "correlation": 26089
    }
  },
  {
    "ph": "s", "id": 26089, "pid": 76337, "tid": -914061504, "ts": 1716454217635386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217668804, "dur": 82,
    "args": {
      "External id": 26098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26098,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26098, "pid": 5, "tid": 7, "ts": 1716454217668804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635469, "dur": 15,
    "args": {
      "External id": 26098,
      "cbid": 211, "correlation": 26098
    }
  },
  {
    "ph": "s", "id": 26098, "pid": 76337, "tid": -914061504, "ts": 1716454217635469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217668887, "dur": 30,
    "args": {
      "External id": 26120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26120,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26120, "pid": 5, "tid": 7, "ts": 1716454217668887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635531, "dur": 10,
    "args": {
      "External id": 26120,
      "cbid": 211, "correlation": 26120
    }
  },
  {
    "ph": "s", "id": 26120, "pid": 76337, "tid": -914061504, "ts": 1716454217635531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635644, "dur": 2,
    "args": {
      "External id": 26131,
      "cbid": 251, "correlation": 26131
    }
  },
  {
    "ph": "f", "id": 26131, "pid": 76337, "tid": -914061504, "ts": 1716454217635644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217668918, "dur": 161,
    "args": {
      "External id": 26132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26132,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26132, "pid": 5, "tid": 7, "ts": 1716454217668918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635650, "dur": 14,
    "args": {
      "External id": 26132,
      "cbid": 211, "correlation": 26132
    }
  },
  {
    "ph": "s", "id": 26132, "pid": 76337, "tid": -914061504, "ts": 1716454217635650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635733, "dur": 1,
    "args": {
      "External id": 26143,
      "cbid": 251, "correlation": 26143
    }
  },
  {
    "ph": "f", "id": 26143, "pid": 76337, "tid": -914061504, "ts": 1716454217635733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217669081, "dur": 157,
    "args": {
      "External id": 26144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26144,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26144, "pid": 5, "tid": 7, "ts": 1716454217669081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635736, "dur": 12,
    "args": {
      "External id": 26144,
      "cbid": 211, "correlation": 26144
    }
  },
  {
    "ph": "s", "id": 26144, "pid": 76337, "tid": -914061504, "ts": 1716454217635736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635802, "dur": 1,
    "args": {
      "External id": 26155,
      "cbid": 251, "correlation": 26155
    }
  },
  {
    "ph": "f", "id": 26155, "pid": 76337, "tid": -914061504, "ts": 1716454217635802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217669239, "dur": 150,
    "args": {
      "External id": 26156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26156,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26156, "pid": 5, "tid": 7, "ts": 1716454217669239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635806, "dur": 11,
    "args": {
      "External id": 26156,
      "cbid": 211, "correlation": 26156
    }
  },
  {
    "ph": "s", "id": 26156, "pid": 76337, "tid": -914061504, "ts": 1716454217635806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217669391, "dur": 329,
    "args": {
      "External id": 26181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26181,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26181, "pid": 5, "tid": 7, "ts": 1716454217669391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217635896, "dur": 14,
    "args": {
      "External id": 26181,
      "cbid": 211, "correlation": 26181
    }
  },
  {
    "ph": "s", "id": 26181, "pid": 76337, "tid": -914061504, "ts": 1716454217635896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636021, "dur": 2,
    "args": {
      "External id": 26199,
      "cbid": 251, "correlation": 26199
    }
  },
  {
    "ph": "f", "id": 26199, "pid": 76337, "tid": -914061504, "ts": 1716454217636021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217669722, "dur": 164,
    "args": {
      "External id": 26201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26201,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26201, "pid": 5, "tid": 7, "ts": 1716454217669722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636027, "dur": 14,
    "args": {
      "External id": 26201,
      "cbid": 211, "correlation": 26201
    }
  },
  {
    "ph": "s", "id": 26201, "pid": 76337, "tid": -914061504, "ts": 1716454217636027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217669887, "dur": 20,
    "args": {
      "External id": 26209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26209,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26209, "pid": 5, "tid": 7, "ts": 1716454217669887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636100, "dur": 13,
    "args": {
      "External id": 26209,
      "cbid": 211, "correlation": 26209
    }
  },
  {
    "ph": "s", "id": 26209, "pid": 76337, "tid": -914061504, "ts": 1716454217636100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217669907, "dur": 28,
    "args": {
      "External id": 26217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26217,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26217, "pid": 5, "tid": 7, "ts": 1716454217669907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636141, "dur": 8,
    "args": {
      "External id": 26217,
      "cbid": 211, "correlation": 26217
    }
  },
  {
    "ph": "s", "id": 26217, "pid": 76337, "tid": -914061504, "ts": 1716454217636141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217669936, "dur": 18,
    "args": {
      "External id": 26228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26228,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26228, "pid": 5, "tid": 7, "ts": 1716454217669936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636217, "dur": 13,
    "args": {
      "External id": 26228,
      "cbid": 211, "correlation": 26228
    }
  },
  {
    "ph": "s", "id": 26228, "pid": 76337, "tid": -914061504, "ts": 1716454217636217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217669956, "dur": 16,
    "args": {
      "External id": 26250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26250,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26250, "pid": 5, "tid": 7, "ts": 1716454217669956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636248, "dur": 8,
    "args": {
      "External id": 26250,
      "cbid": 211, "correlation": 26250
    }
  },
  {
    "ph": "s", "id": 26250, "pid": 76337, "tid": -914061504, "ts": 1716454217636248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636349, "dur": 2,
    "args": {
      "External id": 26261,
      "cbid": 251, "correlation": 26261
    }
  },
  {
    "ph": "f", "id": 26261, "pid": 76337, "tid": -914061504, "ts": 1716454217636349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217669973, "dur": 88,
    "args": {
      "External id": 26262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26262,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26262, "pid": 5, "tid": 7, "ts": 1716454217669973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636355, "dur": 15,
    "args": {
      "External id": 26262,
      "cbid": 211, "correlation": 26262
    }
  },
  {
    "ph": "s", "id": 26262, "pid": 76337, "tid": -914061504, "ts": 1716454217636355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636439, "dur": 1,
    "args": {
      "External id": 26273,
      "cbid": 251, "correlation": 26273
    }
  },
  {
    "ph": "f", "id": 26273, "pid": 76337, "tid": -914061504, "ts": 1716454217636439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636443, "dur": 0,
    "args": {
      "External id": 26274,
      "cbid": 251, "correlation": 26274
    }
  },
  {
    "ph": "f", "id": 26274, "pid": 76337, "tid": -914061504, "ts": 1716454217636443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217670062, "dur": 12,
    "args": {
      "External id": 26275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26275,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26275, "pid": 5, "tid": 7, "ts": 1716454217670062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636445, "dur": 13,
    "args": {
      "External id": 26275,
      "cbid": 211, "correlation": 26275
    }
  },
  {
    "ph": "s", "id": 26275, "pid": 76337, "tid": -914061504, "ts": 1716454217636445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217670076, "dur": 5,
    "args": {
      "External id": 26277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26277,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26277, "pid": 5, "tid": 7, "ts": 1716454217670076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636460, "dur": 7,
    "args": {
      "External id": 26277,
      "cbid": 211, "correlation": 26277
    }
  },
  {
    "ph": "s", "id": 26277, "pid": 76337, "tid": -914061504, "ts": 1716454217636460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636530, "dur": 1,
    "args": {
      "External id": 26288,
      "cbid": 251, "correlation": 26288
    }
  },
  {
    "ph": "f", "id": 26288, "pid": 76337, "tid": -914061504, "ts": 1716454217636530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636534, "dur": 0,
    "args": {
      "External id": 26289,
      "cbid": 251, "correlation": 26289
    }
  },
  {
    "ph": "f", "id": 26289, "pid": 76337, "tid": -914061504, "ts": 1716454217636534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217670083, "dur": 8,
    "args": {
      "External id": 26290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26290,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26290, "pid": 5, "tid": 7, "ts": 1716454217670083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636535, "dur": 13,
    "args": {
      "External id": 26290,
      "cbid": 211, "correlation": 26290
    }
  },
  {
    "ph": "s", "id": 26290, "pid": 76337, "tid": -914061504, "ts": 1716454217636535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217670093, "dur": 3,
    "args": {
      "External id": 26292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26292,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26292, "pid": 5, "tid": 7, "ts": 1716454217670093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636550, "dur": 6,
    "args": {
      "External id": 26292,
      "cbid": 211, "correlation": 26292
    }
  },
  {
    "ph": "s", "id": 26292, "pid": 76337, "tid": -914061504, "ts": 1716454217636550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217670097, "dur": 54,
    "args": {
      "External id": 26317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26317,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26317, "pid": 5, "tid": 7, "ts": 1716454217670097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636630, "dur": 13,
    "args": {
      "External id": 26317,
      "cbid": 211, "correlation": 26317
    }
  },
  {
    "ph": "s", "id": 26317, "pid": 76337, "tid": -914061504, "ts": 1716454217636630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636743, "dur": 2,
    "args": {
      "External id": 26335,
      "cbid": 251, "correlation": 26335
    }
  },
  {
    "ph": "f", "id": 26335, "pid": 76337, "tid": -914061504, "ts": 1716454217636743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217670152, "dur": 90,
    "args": {
      "External id": 26337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26337,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26337, "pid": 5, "tid": 7, "ts": 1716454217670152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636749, "dur": 15,
    "args": {
      "External id": 26337,
      "cbid": 211, "correlation": 26337
    }
  },
  {
    "ph": "s", "id": 26337, "pid": 76337, "tid": -914061504, "ts": 1716454217636749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217670244, "dur": 9,
    "args": {
      "External id": 26345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26345,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26345, "pid": 5, "tid": 7, "ts": 1716454217670244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636821, "dur": 14,
    "args": {
      "External id": 26345,
      "cbid": 211, "correlation": 26345
    }
  },
  {
    "ph": "s", "id": 26345, "pid": 76337, "tid": -914061504, "ts": 1716454217636821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217670254, "dur": 21,
    "args": {
      "External id": 26353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26353, "pid": 5, "tid": 7, "ts": 1716454217670254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636865, "dur": 9,
    "args": {
      "External id": 26353,
      "cbid": 211, "correlation": 26353
    }
  },
  {
    "ph": "s", "id": 26353, "pid": 76337, "tid": -914061504, "ts": 1716454217636865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217670276, "dur": 17,
    "args": {
      "External id": 26375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26375,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26375, "pid": 5, "tid": 7, "ts": 1716454217670276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217636929, "dur": 12,
    "args": {
      "External id": 26375,
      "cbid": 211, "correlation": 26375
    }
  },
  {
    "ph": "s", "id": 26375, "pid": 76337, "tid": -914061504, "ts": 1716454217636929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637030, "dur": 1,
    "args": {
      "External id": 26391,
      "cbid": 251, "correlation": 26391
    }
  },
  {
    "ph": "f", "id": 26391, "pid": 76337, "tid": -914061504, "ts": 1716454217637030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637035, "dur": 0,
    "args": {
      "External id": 26393,
      "cbid": 251, "correlation": 26393
    }
  },
  {
    "ph": "f", "id": 26393, "pid": 76337, "tid": -914061504, "ts": 1716454217637035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217670295, "dur": 496,
    "args": {
      "External id": 26394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26394,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26394, "pid": 5, "tid": 7, "ts": 1716454217670295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637037, "dur": 14,
    "args": {
      "External id": 26394,
      "cbid": 211, "correlation": 26394
    }
  },
  {
    "ph": "s", "id": 26394, "pid": 76337, "tid": -914061504, "ts": 1716454217637037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217670793, "dur": 65,
    "args": {
      "External id": 26402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26402,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26402, "pid": 5, "tid": 7, "ts": 1716454217670793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637109, "dur": 12,
    "args": {
      "External id": 26402,
      "cbid": 211, "correlation": 26402
    }
  },
  {
    "ph": "s", "id": 26402, "pid": 76337, "tid": -914061504, "ts": 1716454217637109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217670859, "dur": 66,
    "args": {
      "External id": 26410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26410,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26410, "pid": 5, "tid": 7, "ts": 1716454217670859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637142, "dur": 8,
    "args": {
      "External id": 26410,
      "cbid": 211, "correlation": 26410
    }
  },
  {
    "ph": "s", "id": 26410, "pid": 76337, "tid": -914061504, "ts": 1716454217637142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637234, "dur": 1,
    "args": {
      "External id": 26426,
      "cbid": 251, "correlation": 26426
    }
  },
  {
    "ph": "f", "id": 26426, "pid": 76337, "tid": -914061504, "ts": 1716454217637234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217670928, "dur": 1,
    "args": {
      "External id": 26428,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 26428,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 26428, "pid": 5, "tid": 7, "ts": 1716454217670928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637240, "dur": 11,
    "args": {
      "External id": 26428,
      "cbid": 51, "correlation": 26428
    }
  },
  {
    "ph": "s", "id": 26428, "pid": 76337, "tid": -914061504, "ts": 1716454217637240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217670931, "dur": 268,
    "args": {
      "External id": 26429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26429,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26429, "pid": 5, "tid": 7, "ts": 1716454217670931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637252, "dur": 12,
    "args": {
      "External id": 26429,
      "cbid": 211, "correlation": 26429
    }
  },
  {
    "ph": "s", "id": 26429, "pid": 76337, "tid": -914061504, "ts": 1716454217637252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217671201, "dur": 14,
    "args": {
      "External id": 26437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26437,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26437, "pid": 5, "tid": 7, "ts": 1716454217671201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637296, "dur": 11,
    "args": {
      "External id": 26437,
      "cbid": 211, "correlation": 26437
    }
  },
  {
    "ph": "s", "id": 26437, "pid": 76337, "tid": -914061504, "ts": 1716454217637296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217671216, "dur": 37,
    "args": {
      "External id": 26448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26448,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26448, "pid": 5, "tid": 7, "ts": 1716454217671216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637366, "dur": 12,
    "args": {
      "External id": 26448,
      "cbid": 211, "correlation": 26448
    }
  },
  {
    "ph": "s", "id": 26448, "pid": 76337, "tid": -914061504, "ts": 1716454217637366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637432, "dur": 0,
    "args": {
      "External id": 26460,
      "cbid": 317, "correlation": 26460
    }
  },
  {
    "ph": "f", "id": 26460, "pid": 76337, "tid": -914061504, "ts": 1716454217637432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637433, "dur": 0,
    "args": {
      "External id": 26461,
      "cbid": 203, "correlation": 26461
    }
  },
  {
    "ph": "f", "id": 26461, "pid": 76337, "tid": -914061504, "ts": 1716454217637433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637434, "dur": 0,
    "args": {
      "External id": 26462,
      "cbid": 205, "correlation": 26462
    }
  },
  {
    "ph": "f", "id": 26462, "pid": 76337, "tid": -914061504, "ts": 1716454217637434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217671255, "dur": 12,
    "args": {
      "External id": 26466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26466,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26466, "pid": 5, "tid": 7, "ts": 1716454217671255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637449, "dur": 12,
    "args": {
      "External id": 26466,
      "cbid": 211, "correlation": 26466
    }
  },
  {
    "ph": "s", "id": 26466, "pid": 76337, "tid": -914061504, "ts": 1716454217637449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217671268, "dur": 4,
    "args": {
      "External id": 26468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26468, "pid": 5, "tid": 7, "ts": 1716454217671268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637466, "dur": 6,
    "args": {
      "External id": 26468,
      "cbid": 211, "correlation": 26468
    }
  },
  {
    "ph": "s", "id": 26468, "pid": 76337, "tid": -914061504, "ts": 1716454217637466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637474, "dur": 0,
    "args": {
      "External id": 26469,
      "cbid": 51, "correlation": 26469
    }
  },
  {
    "ph": "s", "id": 26469, "pid": 76337, "tid": -914061504, "ts": 1716454217637474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217671273, "dur": 94,
    "args": {
      "External id": 26470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26470,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 26470, "pid": 5, "tid": 7, "ts": 1716454217671273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637475, "dur": 5,
    "args": {
      "External id": 26470,
      "cbid": 211, "correlation": 26470
    }
  },
  {
    "ph": "s", "id": 26470, "pid": 76337, "tid": -914061504, "ts": 1716454217637475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217671368, "dur": 16,
    "args": {
      "External id": 26475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26475, "pid": 5, "tid": 7, "ts": 1716454217671368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637503, "dur": 10,
    "args": {
      "External id": 26475,
      "cbid": 211, "correlation": 26475
    }
  },
  {
    "ph": "s", "id": 26475, "pid": 76337, "tid": -914061504, "ts": 1716454217637503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217671386, "dur": 12,
    "args": {
      "External id": 26483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26483,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26483, "pid": 5, "tid": 7, "ts": 1716454217671386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637535, "dur": 8,
    "args": {
      "External id": 26483,
      "cbid": 211, "correlation": 26483
    }
  },
  {
    "ph": "s", "id": 26483, "pid": 76337, "tid": -914061504, "ts": 1716454217637535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217671398, "dur": 29,
    "args": {
      "External id": 26492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26492,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26492, "pid": 5, "tid": 7, "ts": 1716454217671398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637575, "dur": 10,
    "args": {
      "External id": 26492,
      "cbid": 211, "correlation": 26492
    }
  },
  {
    "ph": "s", "id": 26492, "pid": 76337, "tid": -914061504, "ts": 1716454217637575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217671429, "dur": 31,
    "args": {
      "External id": 26512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26512,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 26512, "pid": 5, "tid": 7, "ts": 1716454217671429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637649, "dur": 11,
    "args": {
      "External id": 26512,
      "cbid": 211, "correlation": 26512
    }
  },
  {
    "ph": "s", "id": 26512, "pid": 76337, "tid": -914061504, "ts": 1716454217637649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217671461, "dur": 5,
    "args": {
      "External id": 26524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26524,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26524, "pid": 5, "tid": 7, "ts": 1716454217671461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637682, "dur": 8,
    "args": {
      "External id": 26524,
      "cbid": 211, "correlation": 26524
    }
  },
  {
    "ph": "s", "id": 26524, "pid": 76337, "tid": -914061504, "ts": 1716454217637682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217671467, "dur": 31,
    "args": {
      "External id": 26527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26527,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26527, "pid": 5, "tid": 7, "ts": 1716454217671467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637702, "dur": 7,
    "args": {
      "External id": 26527,
      "cbid": 211, "correlation": 26527
    }
  },
  {
    "ph": "s", "id": 26527, "pid": 76337, "tid": -914061504, "ts": 1716454217637702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217671499, "dur": 21,
    "args": {
      "External id": 26536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26536,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26536, "pid": 5, "tid": 7, "ts": 1716454217671499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637744, "dur": 9,
    "args": {
      "External id": 26536,
      "cbid": 211, "correlation": 26536
    }
  },
  {
    "ph": "s", "id": 26536, "pid": 76337, "tid": -914061504, "ts": 1716454217637744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637807, "dur": 0,
    "args": {
      "External id": 26546,
      "cbid": 317, "correlation": 26546
    }
  },
  {
    "ph": "f", "id": 26546, "pid": 76337, "tid": -914061504, "ts": 1716454217637807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637808, "dur": 0,
    "args": {
      "External id": 26547,
      "cbid": 203, "correlation": 26547
    }
  },
  {
    "ph": "f", "id": 26547, "pid": 76337, "tid": -914061504, "ts": 1716454217637808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637809, "dur": 0,
    "args": {
      "External id": 26548,
      "cbid": 205, "correlation": 26548
    }
  },
  {
    "ph": "f", "id": 26548, "pid": 76337, "tid": -914061504, "ts": 1716454217637809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217671521, "dur": 22,
    "args": {
      "External id": 26552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26552,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26552, "pid": 5, "tid": 7, "ts": 1716454217671521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637830, "dur": 12,
    "args": {
      "External id": 26552,
      "cbid": 211, "correlation": 26552
    }
  },
  {
    "ph": "s", "id": 26552, "pid": 76337, "tid": -914061504, "ts": 1716454217637830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217671544, "dur": 315,
    "args": {
      "External id": 26554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26554,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26554, "pid": 5, "tid": 7, "ts": 1716454217671544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637844, "dur": 5,
    "args": {
      "External id": 26554,
      "cbid": 211, "correlation": 26554
    }
  },
  {
    "ph": "s", "id": 26554, "pid": 76337, "tid": -914061504, "ts": 1716454217637844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217671861, "dur": 1,
    "args": {
      "External id": 26556,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 26556,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 26556, "pid": 5, "tid": 7, "ts": 1716454217671861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637856, "dur": 6,
    "args": {
      "External id": 26556,
      "cbid": 51, "correlation": 26556
    }
  },
  {
    "ph": "s", "id": 26556, "pid": 76337, "tid": -914061504, "ts": 1716454217637856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217671865, "dur": 1238,
    "args": {
      "External id": 26557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26557,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26557, "pid": 5, "tid": 7, "ts": 1716454217671865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637864, "dur": 6,
    "args": {
      "External id": 26557,
      "cbid": 211, "correlation": 26557
    }
  },
  {
    "ph": "s", "id": 26557, "pid": 76337, "tid": -914061504, "ts": 1716454217637864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217673104, "dur": 12,
    "args": {
      "External id": 26559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26559, "pid": 5, "tid": 7, "ts": 1716454217673104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637874, "dur": 6,
    "args": {
      "External id": 26559,
      "cbid": 211, "correlation": 26559
    }
  },
  {
    "ph": "s", "id": 26559, "pid": 76337, "tid": -914061504, "ts": 1716454217637874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217673117, "dur": 15,
    "args": {
      "External id": 26565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26565, "pid": 5, "tid": 7, "ts": 1716454217673117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637911, "dur": 10,
    "args": {
      "External id": 26565,
      "cbid": 211, "correlation": 26565
    }
  },
  {
    "ph": "s", "id": 26565, "pid": 76337, "tid": -914061504, "ts": 1716454217637911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217673133, "dur": 3,
    "args": {
      "External id": 26573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26573,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 26573, "pid": 5, "tid": 7, "ts": 1716454217673133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217637958, "dur": 10,
    "args": {
      "External id": 26573,
      "cbid": 211, "correlation": 26573
    }
  },
  {
    "ph": "s", "id": 26573, "pid": 76337, "tid": -914061504, "ts": 1716454217637958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638034, "dur": 1,
    "args": {
      "External id": 26589,
      "cbid": 251, "correlation": 26589
    }
  },
  {
    "ph": "f", "id": 26589, "pid": 76337, "tid": -914061504, "ts": 1716454217638034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638039, "dur": 0,
    "args": {
      "External id": 26591,
      "cbid": 251, "correlation": 26591
    }
  },
  {
    "ph": "f", "id": 26591, "pid": 76337, "tid": -914061504, "ts": 1716454217638039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217673138, "dur": 12,
    "args": {
      "External id": 26592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26592,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26592, "pid": 5, "tid": 7, "ts": 1716454217673138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638041, "dur": 12,
    "args": {
      "External id": 26592,
      "cbid": 211, "correlation": 26592
    }
  },
  {
    "ph": "s", "id": 26592, "pid": 76337, "tid": -914061504, "ts": 1716454217638041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217673152, "dur": 5,
    "args": {
      "External id": 26594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26594,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26594, "pid": 5, "tid": 7, "ts": 1716454217673152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638055, "dur": 5,
    "args": {
      "External id": 26594,
      "cbid": 211, "correlation": 26594
    }
  },
  {
    "ph": "s", "id": 26594, "pid": 76337, "tid": -914061504, "ts": 1716454217638055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217673158, "dur": 16,
    "args": {
      "External id": 26604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26604, "pid": 5, "tid": 7, "ts": 1716454217673158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638113, "dur": 12,
    "args": {
      "External id": 26604,
      "cbid": 211, "correlation": 26604
    }
  },
  {
    "ph": "s", "id": 26604, "pid": 76337, "tid": -914061504, "ts": 1716454217638113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217673175, "dur": 17,
    "args": {
      "External id": 26624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26624,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 26624, "pid": 5, "tid": 7, "ts": 1716454217673175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638180, "dur": 11,
    "args": {
      "External id": 26624,
      "cbid": 211, "correlation": 26624
    }
  },
  {
    "ph": "s", "id": 26624, "pid": 76337, "tid": -914061504, "ts": 1716454217638180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217673194, "dur": 4,
    "args": {
      "External id": 26636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26636,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 26636, "pid": 5, "tid": 7, "ts": 1716454217673194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638201, "dur": 6,
    "args": {
      "External id": 26636,
      "cbid": 211, "correlation": 26636
    }
  },
  {
    "ph": "s", "id": 26636, "pid": 76337, "tid": -914061504, "ts": 1716454217638201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217673199, "dur": 16,
    "args": {
      "External id": 26639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26639,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26639, "pid": 5, "tid": 7, "ts": 1716454217673199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638220, "dur": 6,
    "args": {
      "External id": 26639,
      "cbid": 211, "correlation": 26639
    }
  },
  {
    "ph": "s", "id": 26639, "pid": 76337, "tid": -914061504, "ts": 1716454217638220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217673216, "dur": 11,
    "args": {
      "External id": 26648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26648,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26648, "pid": 5, "tid": 7, "ts": 1716454217673216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638262, "dur": 10,
    "args": {
      "External id": 26648,
      "cbid": 211, "correlation": 26648
    }
  },
  {
    "ph": "s", "id": 26648, "pid": 76337, "tid": -914061504, "ts": 1716454217638262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638325, "dur": 0,
    "args": {
      "External id": 26658,
      "cbid": 317, "correlation": 26658
    }
  },
  {
    "ph": "f", "id": 26658, "pid": 76337, "tid": -914061504, "ts": 1716454217638325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638326, "dur": 0,
    "args": {
      "External id": 26659,
      "cbid": 203, "correlation": 26659
    }
  },
  {
    "ph": "f", "id": 26659, "pid": 76337, "tid": -914061504, "ts": 1716454217638326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638327, "dur": 0,
    "args": {
      "External id": 26660,
      "cbid": 205, "correlation": 26660
    }
  },
  {
    "ph": "f", "id": 26660, "pid": 76337, "tid": -914061504, "ts": 1716454217638327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217673228, "dur": 11,
    "args": {
      "External id": 26664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26664,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26664, "pid": 5, "tid": 7, "ts": 1716454217673228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638340, "dur": 13,
    "args": {
      "External id": 26664,
      "cbid": 211, "correlation": 26664
    }
  },
  {
    "ph": "s", "id": 26664, "pid": 76337, "tid": -914061504, "ts": 1716454217638340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217673240, "dur": 161,
    "args": {
      "External id": 26666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26666, "pid": 5, "tid": 7, "ts": 1716454217673240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638356, "dur": 5,
    "args": {
      "External id": 26666,
      "cbid": 211, "correlation": 26666
    }
  },
  {
    "ph": "s", "id": 26666, "pid": 76337, "tid": -914061504, "ts": 1716454217638356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217673403, "dur": 1,
    "args": {
      "External id": 26668,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 26668,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 26668, "pid": 5, "tid": 7, "ts": 1716454217673403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638367, "dur": 6,
    "args": {
      "External id": 26668,
      "cbid": 51, "correlation": 26668
    }
  },
  {
    "ph": "s", "id": 26668, "pid": 76337, "tid": -914061504, "ts": 1716454217638367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217673407, "dur": 639,
    "args": {
      "External id": 26669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26669,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26669, "pid": 5, "tid": 7, "ts": 1716454217673407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638374, "dur": 6,
    "args": {
      "External id": 26669,
      "cbid": 211, "correlation": 26669
    }
  },
  {
    "ph": "s", "id": 26669, "pid": 76337, "tid": -914061504, "ts": 1716454217638374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217674047, "dur": 12,
    "args": {
      "External id": 26671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26671, "pid": 5, "tid": 7, "ts": 1716454217674047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638384, "dur": 5,
    "args": {
      "External id": 26671,
      "cbid": 211, "correlation": 26671
    }
  },
  {
    "ph": "s", "id": 26671, "pid": 76337, "tid": -914061504, "ts": 1716454217638384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217674061, "dur": 15,
    "args": {
      "External id": 26677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26677, "pid": 5, "tid": 7, "ts": 1716454217674061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638414, "dur": 9,
    "args": {
      "External id": 26677,
      "cbid": 211, "correlation": 26677
    }
  },
  {
    "ph": "s", "id": 26677, "pid": 76337, "tid": -914061504, "ts": 1716454217638414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638472, "dur": 0,
    "args": {
      "External id": 26687,
      "cbid": 317, "correlation": 26687
    }
  },
  {
    "ph": "f", "id": 26687, "pid": 76337, "tid": -914061504, "ts": 1716454217638472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638473, "dur": 0,
    "args": {
      "External id": 26688,
      "cbid": 203, "correlation": 26688
    }
  },
  {
    "ph": "f", "id": 26688, "pid": 76337, "tid": -914061504, "ts": 1716454217638473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638474, "dur": 0,
    "args": {
      "External id": 26689,
      "cbid": 205, "correlation": 26689
    }
  },
  {
    "ph": "f", "id": 26689, "pid": 76337, "tid": -914061504, "ts": 1716454217638474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217674077, "dur": 22,
    "args": {
      "External id": 26693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26693,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26693, "pid": 5, "tid": 7, "ts": 1716454217674077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638485, "dur": 11,
    "args": {
      "External id": 26693,
      "cbid": 211, "correlation": 26693
    }
  },
  {
    "ph": "s", "id": 26693, "pid": 76337, "tid": -914061504, "ts": 1716454217638485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217674100, "dur": 4,
    "args": {
      "External id": 26695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26695,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26695, "pid": 5, "tid": 7, "ts": 1716454217674100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638501, "dur": 6,
    "args": {
      "External id": 26695,
      "cbid": 211, "correlation": 26695
    }
  },
  {
    "ph": "s", "id": 26695, "pid": 76337, "tid": -914061504, "ts": 1716454217638501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638510, "dur": 0,
    "args": {
      "External id": 26696,
      "cbid": 51, "correlation": 26696
    }
  },
  {
    "ph": "s", "id": 26696, "pid": 76337, "tid": -914061504, "ts": 1716454217638510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217674105, "dur": 167,
    "args": {
      "External id": 26697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26697,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 26697, "pid": 5, "tid": 7, "ts": 1716454217674105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638511, "dur": 5,
    "args": {
      "External id": 26697,
      "cbid": 211, "correlation": 26697
    }
  },
  {
    "ph": "s", "id": 26697, "pid": 76337, "tid": -914061504, "ts": 1716454217638511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217674274, "dur": 16,
    "args": {
      "External id": 26702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26702, "pid": 5, "tid": 7, "ts": 1716454217674274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638536, "dur": 8,
    "args": {
      "External id": 26702,
      "cbid": 211, "correlation": 26702
    }
  },
  {
    "ph": "s", "id": 26702, "pid": 76337, "tid": -914061504, "ts": 1716454217638536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217674291, "dur": 12,
    "args": {
      "External id": 26710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26710,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26710, "pid": 5, "tid": 7, "ts": 1716454217674291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638565, "dur": 8,
    "args": {
      "External id": 26710,
      "cbid": 211, "correlation": 26710
    }
  },
  {
    "ph": "s", "id": 26710, "pid": 76337, "tid": -914061504, "ts": 1716454217638565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217674304, "dur": 10,
    "args": {
      "External id": 26718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26718,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26718, "pid": 5, "tid": 7, "ts": 1716454217674304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638594, "dur": 8,
    "args": {
      "External id": 26718,
      "cbid": 211, "correlation": 26718
    }
  },
  {
    "ph": "s", "id": 26718, "pid": 76337, "tid": -914061504, "ts": 1716454217638594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217674315, "dur": 18,
    "args": {
      "External id": 26738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26738,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 26738, "pid": 5, "tid": 7, "ts": 1716454217674315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638678, "dur": 12,
    "args": {
      "External id": 26738,
      "cbid": 211, "correlation": 26738
    }
  },
  {
    "ph": "s", "id": 26738, "pid": 76337, "tid": -914061504, "ts": 1716454217638678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217674334, "dur": 4,
    "args": {
      "External id": 26750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26750,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 26750, "pid": 5, "tid": 7, "ts": 1716454217674334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638700, "dur": 6,
    "args": {
      "External id": 26750,
      "cbid": 211, "correlation": 26750
    }
  },
  {
    "ph": "s", "id": 26750, "pid": 76337, "tid": -914061504, "ts": 1716454217638700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217674339, "dur": 16,
    "args": {
      "External id": 26753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26753,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26753, "pid": 5, "tid": 7, "ts": 1716454217674339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638718, "dur": 7,
    "args": {
      "External id": 26753,
      "cbid": 211, "correlation": 26753
    }
  },
  {
    "ph": "s", "id": 26753, "pid": 76337, "tid": -914061504, "ts": 1716454217638718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638775, "dur": 0,
    "args": {
      "External id": 26764,
      "cbid": 317, "correlation": 26764
    }
  },
  {
    "ph": "f", "id": 26764, "pid": 76337, "tid": -914061504, "ts": 1716454217638775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638776, "dur": 0,
    "args": {
      "External id": 26765,
      "cbid": 203, "correlation": 26765
    }
  },
  {
    "ph": "f", "id": 26765, "pid": 76337, "tid": -914061504, "ts": 1716454217638776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638777, "dur": 0,
    "args": {
      "External id": 26766,
      "cbid": 205, "correlation": 26766
    }
  },
  {
    "ph": "f", "id": 26766, "pid": 76337, "tid": -914061504, "ts": 1716454217638777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217674357, "dur": 11,
    "args": {
      "External id": 26770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26770, "pid": 5, "tid": 7, "ts": 1716454217674357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638791, "dur": 11,
    "args": {
      "External id": 26770,
      "cbid": 211, "correlation": 26770
    }
  },
  {
    "ph": "s", "id": 26770, "pid": 76337, "tid": -914061504, "ts": 1716454217638791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217674370, "dur": 3,
    "args": {
      "External id": 26772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26772, "pid": 5, "tid": 7, "ts": 1716454217674370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638806, "dur": 6,
    "args": {
      "External id": 26772,
      "cbid": 211, "correlation": 26772
    }
  },
  {
    "ph": "s", "id": 26772, "pid": 76337, "tid": -914061504, "ts": 1716454217638806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638814, "dur": 0,
    "args": {
      "External id": 26773,
      "cbid": 51, "correlation": 26773
    }
  },
  {
    "ph": "s", "id": 26773, "pid": 76337, "tid": -914061504, "ts": 1716454217638814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217674374, "dur": 90,
    "args": {
      "External id": 26774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26774,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 26774, "pid": 5, "tid": 7, "ts": 1716454217674374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638815, "dur": 5,
    "args": {
      "External id": 26774,
      "cbid": 211, "correlation": 26774
    }
  },
  {
    "ph": "s", "id": 26774, "pid": 76337, "tid": -914061504, "ts": 1716454217638815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217674466, "dur": 16,
    "args": {
      "External id": 26779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26779, "pid": 5, "tid": 7, "ts": 1716454217674466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638842, "dur": 9,
    "args": {
      "External id": 26779,
      "cbid": 211, "correlation": 26779
    }
  },
  {
    "ph": "s", "id": 26779, "pid": 76337, "tid": -914061504, "ts": 1716454217638842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217674483, "dur": 82,
    "args": {
      "External id": 26788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26788,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26788, "pid": 5, "tid": 7, "ts": 1716454217674483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638924, "dur": 14,
    "args": {
      "External id": 26788,
      "cbid": 211, "correlation": 26788
    }
  },
  {
    "ph": "s", "id": 26788, "pid": 76337, "tid": -914061504, "ts": 1716454217638924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217674566, "dur": 30,
    "args": {
      "External id": 26810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26810,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26810, "pid": 5, "tid": 7, "ts": 1716454217674566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217638990, "dur": 11,
    "args": {
      "External id": 26810,
      "cbid": 211, "correlation": 26810
    }
  },
  {
    "ph": "s", "id": 26810, "pid": 76337, "tid": -914061504, "ts": 1716454217638990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639104, "dur": 1,
    "args": {
      "External id": 26821,
      "cbid": 251, "correlation": 26821
    }
  },
  {
    "ph": "f", "id": 26821, "pid": 76337, "tid": -914061504, "ts": 1716454217639104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217674597, "dur": 163,
    "args": {
      "External id": 26822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26822,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26822, "pid": 5, "tid": 7, "ts": 1716454217674597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639109, "dur": 14,
    "args": {
      "External id": 26822,
      "cbid": 211, "correlation": 26822
    }
  },
  {
    "ph": "s", "id": 26822, "pid": 76337, "tid": -914061504, "ts": 1716454217639109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639195, "dur": 1,
    "args": {
      "External id": 26833,
      "cbid": 251, "correlation": 26833
    }
  },
  {
    "ph": "f", "id": 26833, "pid": 76337, "tid": -914061504, "ts": 1716454217639195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217674761, "dur": 156,
    "args": {
      "External id": 26834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26834,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26834, "pid": 5, "tid": 7, "ts": 1716454217674761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639199, "dur": 13,
    "args": {
      "External id": 26834,
      "cbid": 211, "correlation": 26834
    }
  },
  {
    "ph": "s", "id": 26834, "pid": 76337, "tid": -914061504, "ts": 1716454217639199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639279, "dur": 1,
    "args": {
      "External id": 26845,
      "cbid": 251, "correlation": 26845
    }
  },
  {
    "ph": "f", "id": 26845, "pid": 76337, "tid": -914061504, "ts": 1716454217639279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217674919, "dur": 156,
    "args": {
      "External id": 26846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26846,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26846, "pid": 5, "tid": 7, "ts": 1716454217674919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639283, "dur": 12,
    "args": {
      "External id": 26846,
      "cbid": 211, "correlation": 26846
    }
  },
  {
    "ph": "s", "id": 26846, "pid": 76337, "tid": -914061504, "ts": 1716454217639283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217675076, "dur": 332,
    "args": {
      "External id": 26871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26871,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26871, "pid": 5, "tid": 7, "ts": 1716454217675076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639371, "dur": 13,
    "args": {
      "External id": 26871,
      "cbid": 211, "correlation": 26871
    }
  },
  {
    "ph": "s", "id": 26871, "pid": 76337, "tid": -914061504, "ts": 1716454217639371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639487, "dur": 1,
    "args": {
      "External id": 26889,
      "cbid": 251, "correlation": 26889
    }
  },
  {
    "ph": "f", "id": 26889, "pid": 76337, "tid": -914061504, "ts": 1716454217639487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217675409, "dur": 164,
    "args": {
      "External id": 26891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26891,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26891, "pid": 5, "tid": 7, "ts": 1716454217675409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639493, "dur": 14,
    "args": {
      "External id": 26891,
      "cbid": 211, "correlation": 26891
    }
  },
  {
    "ph": "s", "id": 26891, "pid": 76337, "tid": -914061504, "ts": 1716454217639493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217675574, "dur": 19,
    "args": {
      "External id": 26899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26899,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26899, "pid": 5, "tid": 7, "ts": 1716454217675574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639566, "dur": 13,
    "args": {
      "External id": 26899,
      "cbid": 211, "correlation": 26899
    }
  },
  {
    "ph": "s", "id": 26899, "pid": 76337, "tid": -914061504, "ts": 1716454217639566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217675595, "dur": 27,
    "args": {
      "External id": 26907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26907,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26907, "pid": 5, "tid": 7, "ts": 1716454217675595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639605, "dur": 9,
    "args": {
      "External id": 26907,
      "cbid": 211, "correlation": 26907
    }
  },
  {
    "ph": "s", "id": 26907, "pid": 76337, "tid": -914061504, "ts": 1716454217639605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217675624, "dur": 18,
    "args": {
      "External id": 26918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26918,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26918, "pid": 5, "tid": 7, "ts": 1716454217675624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639678, "dur": 13,
    "args": {
      "External id": 26918,
      "cbid": 211, "correlation": 26918
    }
  },
  {
    "ph": "s", "id": 26918, "pid": 76337, "tid": -914061504, "ts": 1716454217639678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217675643, "dur": 16,
    "args": {
      "External id": 26940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26940,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 26940, "pid": 5, "tid": 7, "ts": 1716454217675643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639709, "dur": 7,
    "args": {
      "External id": 26940,
      "cbid": 211, "correlation": 26940
    }
  },
  {
    "ph": "s", "id": 26940, "pid": 76337, "tid": -914061504, "ts": 1716454217639709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639807, "dur": 1,
    "args": {
      "External id": 26951,
      "cbid": 251, "correlation": 26951
    }
  },
  {
    "ph": "f", "id": 26951, "pid": 76337, "tid": -914061504, "ts": 1716454217639807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217675660, "dur": 88,
    "args": {
      "External id": 26952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26952,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 26952, "pid": 5, "tid": 7, "ts": 1716454217675660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639813, "dur": 14,
    "args": {
      "External id": 26952,
      "cbid": 211, "correlation": 26952
    }
  },
  {
    "ph": "s", "id": 26952, "pid": 76337, "tid": -914061504, "ts": 1716454217639813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639896, "dur": 1,
    "args": {
      "External id": 26963,
      "cbid": 251, "correlation": 26963
    }
  },
  {
    "ph": "f", "id": 26963, "pid": 76337, "tid": -914061504, "ts": 1716454217639896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639900, "dur": 0,
    "args": {
      "External id": 26964,
      "cbid": 251, "correlation": 26964
    }
  },
  {
    "ph": "f", "id": 26964, "pid": 76337, "tid": -914061504, "ts": 1716454217639900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217675750, "dur": 12,
    "args": {
      "External id": 26965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26965,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26965, "pid": 5, "tid": 7, "ts": 1716454217675750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639902, "dur": 13,
    "args": {
      "External id": 26965,
      "cbid": 211, "correlation": 26965
    }
  },
  {
    "ph": "s", "id": 26965, "pid": 76337, "tid": -914061504, "ts": 1716454217639902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217675763, "dur": 6,
    "args": {
      "External id": 26967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26967,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26967, "pid": 5, "tid": 7, "ts": 1716454217675763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639917, "dur": 6,
    "args": {
      "External id": 26967,
      "cbid": 211, "correlation": 26967
    }
  },
  {
    "ph": "s", "id": 26967, "pid": 76337, "tid": -914061504, "ts": 1716454217639917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639996, "dur": 1,
    "args": {
      "External id": 26978,
      "cbid": 251, "correlation": 26978
    }
  },
  {
    "ph": "f", "id": 26978, "pid": 76337, "tid": -914061504, "ts": 1716454217639996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217639999, "dur": 0,
    "args": {
      "External id": 26979,
      "cbid": 251, "correlation": 26979
    }
  },
  {
    "ph": "f", "id": 26979, "pid": 76337, "tid": -914061504, "ts": 1716454217639999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217675770, "dur": 8,
    "args": {
      "External id": 26980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26980,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26980, "pid": 5, "tid": 7, "ts": 1716454217675770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640001, "dur": 13,
    "args": {
      "External id": 26980,
      "cbid": 211, "correlation": 26980
    }
  },
  {
    "ph": "s", "id": 26980, "pid": 76337, "tid": -914061504, "ts": 1716454217640001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217675780, "dur": 3,
    "args": {
      "External id": 26982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 26982,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 26982, "pid": 5, "tid": 7, "ts": 1716454217675780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640015, "dur": 5,
    "args": {
      "External id": 26982,
      "cbid": 211, "correlation": 26982
    }
  },
  {
    "ph": "s", "id": 26982, "pid": 76337, "tid": -914061504, "ts": 1716454217640015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217675784, "dur": 54,
    "args": {
      "External id": 27007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27007,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27007, "pid": 5, "tid": 7, "ts": 1716454217675784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640094, "dur": 12,
    "args": {
      "External id": 27007,
      "cbid": 211, "correlation": 27007
    }
  },
  {
    "ph": "s", "id": 27007, "pid": 76337, "tid": -914061504, "ts": 1716454217640094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640204, "dur": 1,
    "args": {
      "External id": 27025,
      "cbid": 251, "correlation": 27025
    }
  },
  {
    "ph": "f", "id": 27025, "pid": 76337, "tid": -914061504, "ts": 1716454217640204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217675839, "dur": 89,
    "args": {
      "External id": 27027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27027,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27027, "pid": 5, "tid": 7, "ts": 1716454217675839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640210, "dur": 14,
    "args": {
      "External id": 27027,
      "cbid": 211, "correlation": 27027
    }
  },
  {
    "ph": "s", "id": 27027, "pid": 76337, "tid": -914061504, "ts": 1716454217640210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217675930, "dur": 10,
    "args": {
      "External id": 27035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27035,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27035, "pid": 5, "tid": 7, "ts": 1716454217675930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640282, "dur": 12,
    "args": {
      "External id": 27035,
      "cbid": 211, "correlation": 27035
    }
  },
  {
    "ph": "s", "id": 27035, "pid": 76337, "tid": -914061504, "ts": 1716454217640282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217675941, "dur": 21,
    "args": {
      "External id": 27043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27043,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27043, "pid": 5, "tid": 7, "ts": 1716454217675941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640323, "dur": 9,
    "args": {
      "External id": 27043,
      "cbid": 211, "correlation": 27043
    }
  },
  {
    "ph": "s", "id": 27043, "pid": 76337, "tid": -914061504, "ts": 1716454217640323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217675963, "dur": 17,
    "args": {
      "External id": 27065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27065,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27065, "pid": 5, "tid": 7, "ts": 1716454217675963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640375, "dur": 10,
    "args": {
      "External id": 27065,
      "cbid": 211, "correlation": 27065
    }
  },
  {
    "ph": "s", "id": 27065, "pid": 76337, "tid": -914061504, "ts": 1716454217640375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640475, "dur": 1,
    "args": {
      "External id": 27081,
      "cbid": 251, "correlation": 27081
    }
  },
  {
    "ph": "f", "id": 27081, "pid": 76337, "tid": -914061504, "ts": 1716454217640475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640480, "dur": 0,
    "args": {
      "External id": 27083,
      "cbid": 251, "correlation": 27083
    }
  },
  {
    "ph": "f", "id": 27083, "pid": 76337, "tid": -914061504, "ts": 1716454217640480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217675982, "dur": 496,
    "args": {
      "External id": 27084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27084,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27084, "pid": 5, "tid": 7, "ts": 1716454217675982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640482, "dur": 14,
    "args": {
      "External id": 27084,
      "cbid": 211, "correlation": 27084
    }
  },
  {
    "ph": "s", "id": 27084, "pid": 76337, "tid": -914061504, "ts": 1716454217640482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217676478, "dur": 64,
    "args": {
      "External id": 27092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27092, "pid": 5, "tid": 7, "ts": 1716454217676478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640550, "dur": 14,
    "args": {
      "External id": 27092,
      "cbid": 211, "correlation": 27092
    }
  },
  {
    "ph": "s", "id": 27092, "pid": 76337, "tid": -914061504, "ts": 1716454217640550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217676544, "dur": 66,
    "args": {
      "External id": 27100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27100, "pid": 5, "tid": 7, "ts": 1716454217676544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640582, "dur": 8,
    "args": {
      "External id": 27100,
      "cbid": 211, "correlation": 27100
    }
  },
  {
    "ph": "s", "id": 27100, "pid": 76337, "tid": -914061504, "ts": 1716454217640582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640662, "dur": 1,
    "args": {
      "External id": 27116,
      "cbid": 251, "correlation": 27116
    }
  },
  {
    "ph": "f", "id": 27116, "pid": 76337, "tid": -914061504, "ts": 1716454217640662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217676612, "dur": 1,
    "args": {
      "External id": 27118,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 27118,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 27118, "pid": 5, "tid": 7, "ts": 1716454217676612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640668, "dur": 10,
    "args": {
      "External id": 27118,
      "cbid": 51, "correlation": 27118
    }
  },
  {
    "ph": "s", "id": 27118, "pid": 76337, "tid": -914061504, "ts": 1716454217640668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217676615, "dur": 268,
    "args": {
      "External id": 27119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27119,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27119, "pid": 5, "tid": 7, "ts": 1716454217676615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640679, "dur": 11,
    "args": {
      "External id": 27119,
      "cbid": 211, "correlation": 27119
    }
  },
  {
    "ph": "s", "id": 27119, "pid": 76337, "tid": -914061504, "ts": 1716454217640679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217676885, "dur": 14,
    "args": {
      "External id": 27127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27127,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27127, "pid": 5, "tid": 7, "ts": 1716454217676885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640721, "dur": 10,
    "args": {
      "External id": 27127,
      "cbid": 211, "correlation": 27127
    }
  },
  {
    "ph": "s", "id": 27127, "pid": 76337, "tid": -914061504, "ts": 1716454217640721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217676900, "dur": 37,
    "args": {
      "External id": 27138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27138, "pid": 5, "tid": 7, "ts": 1716454217676900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640790, "dur": 12,
    "args": {
      "External id": 27138,
      "cbid": 211, "correlation": 27138
    }
  },
  {
    "ph": "s", "id": 27138, "pid": 76337, "tid": -914061504, "ts": 1716454217640790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640854, "dur": 0,
    "args": {
      "External id": 27150,
      "cbid": 317, "correlation": 27150
    }
  },
  {
    "ph": "f", "id": 27150, "pid": 76337, "tid": -914061504, "ts": 1716454217640854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640855, "dur": 0,
    "args": {
      "External id": 27151,
      "cbid": 203, "correlation": 27151
    }
  },
  {
    "ph": "f", "id": 27151, "pid": 76337, "tid": -914061504, "ts": 1716454217640855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640856, "dur": 0,
    "args": {
      "External id": 27152,
      "cbid": 205, "correlation": 27152
    }
  },
  {
    "ph": "f", "id": 27152, "pid": 76337, "tid": -914061504, "ts": 1716454217640856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217676938, "dur": 14,
    "args": {
      "External id": 27156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27156, "pid": 5, "tid": 7, "ts": 1716454217676938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640872, "dur": 12,
    "args": {
      "External id": 27156,
      "cbid": 211, "correlation": 27156
    }
  },
  {
    "ph": "s", "id": 27156, "pid": 76337, "tid": -914061504, "ts": 1716454217640872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217676954, "dur": 4,
    "args": {
      "External id": 27158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27158, "pid": 5, "tid": 7, "ts": 1716454217676954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640889, "dur": 6,
    "args": {
      "External id": 27158,
      "cbid": 211, "correlation": 27158
    }
  },
  {
    "ph": "s", "id": 27158, "pid": 76337, "tid": -914061504, "ts": 1716454217640889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640897, "dur": 0,
    "args": {
      "External id": 27159,
      "cbid": 51, "correlation": 27159
    }
  },
  {
    "ph": "s", "id": 27159, "pid": 76337, "tid": -914061504, "ts": 1716454217640897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217676959, "dur": 95,
    "args": {
      "External id": 27160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27160,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 27160, "pid": 5, "tid": 7, "ts": 1716454217676959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640898, "dur": 5,
    "args": {
      "External id": 27160,
      "cbid": 211, "correlation": 27160
    }
  },
  {
    "ph": "s", "id": 27160, "pid": 76337, "tid": -914061504, "ts": 1716454217640898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217677056, "dur": 16,
    "args": {
      "External id": 27165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27165, "pid": 5, "tid": 7, "ts": 1716454217677056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640926, "dur": 9,
    "args": {
      "External id": 27165,
      "cbid": 211, "correlation": 27165
    }
  },
  {
    "ph": "s", "id": 27165, "pid": 76337, "tid": -914061504, "ts": 1716454217640926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217677073, "dur": 11,
    "args": {
      "External id": 27173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27173,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27173, "pid": 5, "tid": 7, "ts": 1716454217677073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217640958, "dur": 8,
    "args": {
      "External id": 27173,
      "cbid": 211, "correlation": 27173
    }
  },
  {
    "ph": "s", "id": 27173, "pid": 76337, "tid": -914061504, "ts": 1716454217640958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217677085, "dur": 25,
    "args": {
      "External id": 27182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27182,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27182, "pid": 5, "tid": 7, "ts": 1716454217677085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641005, "dur": 11,
    "args": {
      "External id": 27182,
      "cbid": 211, "correlation": 27182
    }
  },
  {
    "ph": "s", "id": 27182, "pid": 76337, "tid": -914061504, "ts": 1716454217641005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217677112, "dur": 23,
    "args": {
      "External id": 27202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27202,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 27202, "pid": 5, "tid": 7, "ts": 1716454217677112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641090, "dur": 13,
    "args": {
      "External id": 27202,
      "cbid": 211, "correlation": 27202
    }
  },
  {
    "ph": "s", "id": 27202, "pid": 76337, "tid": -914061504, "ts": 1716454217641090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217677136, "dur": 5,
    "args": {
      "External id": 27214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27214,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 27214, "pid": 5, "tid": 7, "ts": 1716454217677136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641113, "dur": 7,
    "args": {
      "External id": 27214,
      "cbid": 211, "correlation": 27214
    }
  },
  {
    "ph": "s", "id": 27214, "pid": 76337, "tid": -914061504, "ts": 1716454217641113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217677142, "dur": 24,
    "args": {
      "External id": 27217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27217,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27217, "pid": 5, "tid": 7, "ts": 1716454217677142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641133, "dur": 7,
    "args": {
      "External id": 27217,
      "cbid": 211, "correlation": 27217
    }
  },
  {
    "ph": "s", "id": 27217, "pid": 76337, "tid": -914061504, "ts": 1716454217641133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217677168, "dur": 18,
    "args": {
      "External id": 27226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27226,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27226, "pid": 5, "tid": 7, "ts": 1716454217677168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641172, "dur": 10,
    "args": {
      "External id": 27226,
      "cbid": 211, "correlation": 27226
    }
  },
  {
    "ph": "s", "id": 27226, "pid": 76337, "tid": -914061504, "ts": 1716454217641172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641236, "dur": 0,
    "args": {
      "External id": 27236,
      "cbid": 317, "correlation": 27236
    }
  },
  {
    "ph": "f", "id": 27236, "pid": 76337, "tid": -914061504, "ts": 1716454217641236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641237, "dur": 0,
    "args": {
      "External id": 27237,
      "cbid": 203, "correlation": 27237
    }
  },
  {
    "ph": "f", "id": 27237, "pid": 76337, "tid": -914061504, "ts": 1716454217641237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641237, "dur": 0,
    "args": {
      "External id": 27238,
      "cbid": 205, "correlation": 27238
    }
  },
  {
    "ph": "f", "id": 27238, "pid": 76337, "tid": -914061504, "ts": 1716454217641237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217677187, "dur": 17,
    "args": {
      "External id": 27242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27242,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27242, "pid": 5, "tid": 7, "ts": 1716454217677187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641253, "dur": 12,
    "args": {
      "External id": 27242,
      "cbid": 211, "correlation": 27242
    }
  },
  {
    "ph": "s", "id": 27242, "pid": 76337, "tid": -914061504, "ts": 1716454217641253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217677206, "dur": 237,
    "args": {
      "External id": 27244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27244,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27244, "pid": 5, "tid": 7, "ts": 1716454217677206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641268, "dur": 5,
    "args": {
      "External id": 27244,
      "cbid": 211, "correlation": 27244
    }
  },
  {
    "ph": "s", "id": 27244, "pid": 76337, "tid": -914061504, "ts": 1716454217641268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217677445, "dur": 1,
    "args": {
      "External id": 27246,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 27246,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5
    }
  },
  {
    "ph": "f", "id": 27246, "pid": 5, "tid": 7, "ts": 1716454217677445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641279, "dur": 8,
    "args": {
      "External id": 27246,
      "cbid": 51, "correlation": 27246
    }
  },
  {
    "ph": "s", "id": 27246, "pid": 76337, "tid": -914061504, "ts": 1716454217641279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217677449, "dur": 805,
    "args": {
      "External id": 27247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27247,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27247, "pid": 5, "tid": 7, "ts": 1716454217677449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641288, "dur": 6,
    "args": {
      "External id": 27247,
      "cbid": 211, "correlation": 27247
    }
  },
  {
    "ph": "s", "id": 27247, "pid": 76337, "tid": -914061504, "ts": 1716454217641288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217678255, "dur": 13,
    "args": {
      "External id": 27249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27249,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27249, "pid": 5, "tid": 7, "ts": 1716454217678255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641299, "dur": 5,
    "args": {
      "External id": 27249,
      "cbid": 211, "correlation": 27249
    }
  },
  {
    "ph": "s", "id": 27249, "pid": 76337, "tid": -914061504, "ts": 1716454217641299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217678270, "dur": 14,
    "args": {
      "External id": 27255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27255, "pid": 5, "tid": 7, "ts": 1716454217678270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641327, "dur": 9,
    "args": {
      "External id": 27255,
      "cbid": 211, "correlation": 27255
    }
  },
  {
    "ph": "s", "id": 27255, "pid": 76337, "tid": -914061504, "ts": 1716454217641327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217678285, "dur": 3,
    "args": {
      "External id": 27263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27263,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 27263, "pid": 5, "tid": 7, "ts": 1716454217678285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641372, "dur": 10,
    "args": {
      "External id": 27263,
      "cbid": 211, "correlation": 27263
    }
  },
  {
    "ph": "s", "id": 27263, "pid": 76337, "tid": -914061504, "ts": 1716454217641372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641451, "dur": 1,
    "args": {
      "External id": 27279,
      "cbid": 251, "correlation": 27279
    }
  },
  {
    "ph": "f", "id": 27279, "pid": 76337, "tid": -914061504, "ts": 1716454217641451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641455, "dur": 0,
    "args": {
      "External id": 27281,
      "cbid": 251, "correlation": 27281
    }
  },
  {
    "ph": "f", "id": 27281, "pid": 76337, "tid": -914061504, "ts": 1716454217641455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217678290, "dur": 13,
    "args": {
      "External id": 27282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27282,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27282, "pid": 5, "tid": 7, "ts": 1716454217678290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641457, "dur": 12,
    "args": {
      "External id": 27282,
      "cbid": 211, "correlation": 27282
    }
  },
  {
    "ph": "s", "id": 27282, "pid": 76337, "tid": -914061504, "ts": 1716454217641457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217678305, "dur": 5,
    "args": {
      "External id": 27284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27284,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27284, "pid": 5, "tid": 7, "ts": 1716454217678305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641471, "dur": 5,
    "args": {
      "External id": 27284,
      "cbid": 211, "correlation": 27284
    }
  },
  {
    "ph": "s", "id": 27284, "pid": 76337, "tid": -914061504, "ts": 1716454217641471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217678311, "dur": 17,
    "args": {
      "External id": 27294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27294, "pid": 5, "tid": 7, "ts": 1716454217678311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641531, "dur": 13,
    "args": {
      "External id": 27294,
      "cbid": 211, "correlation": 27294
    }
  },
  {
    "ph": "s", "id": 27294, "pid": 76337, "tid": -914061504, "ts": 1716454217641531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217678329, "dur": 18,
    "args": {
      "External id": 27314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27314,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 27314, "pid": 5, "tid": 7, "ts": 1716454217678329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641599, "dur": 11,
    "args": {
      "External id": 27314,
      "cbid": 211, "correlation": 27314
    }
  },
  {
    "ph": "s", "id": 27314, "pid": 76337, "tid": -914061504, "ts": 1716454217641599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217678349, "dur": 4,
    "args": {
      "External id": 27326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27326,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 27326, "pid": 5, "tid": 7, "ts": 1716454217678349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641619, "dur": 6,
    "args": {
      "External id": 27326,
      "cbid": 211, "correlation": 27326
    }
  },
  {
    "ph": "s", "id": 27326, "pid": 76337, "tid": -914061504, "ts": 1716454217641619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217678354, "dur": 16,
    "args": {
      "External id": 27329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27329,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27329, "pid": 5, "tid": 7, "ts": 1716454217678354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641637, "dur": 6,
    "args": {
      "External id": 27329,
      "cbid": 211, "correlation": 27329
    }
  },
  {
    "ph": "s", "id": 27329, "pid": 76337, "tid": -914061504, "ts": 1716454217641637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217678372, "dur": 11,
    "args": {
      "External id": 27338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27338,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27338, "pid": 5, "tid": 7, "ts": 1716454217678372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641678, "dur": 10,
    "args": {
      "External id": 27338,
      "cbid": 211, "correlation": 27338
    }
  },
  {
    "ph": "s", "id": 27338, "pid": 76337, "tid": -914061504, "ts": 1716454217641678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641741, "dur": 0,
    "args": {
      "External id": 27348,
      "cbid": 317, "correlation": 27348
    }
  },
  {
    "ph": "f", "id": 27348, "pid": 76337, "tid": -914061504, "ts": 1716454217641741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641742, "dur": 0,
    "args": {
      "External id": 27349,
      "cbid": 203, "correlation": 27349
    }
  },
  {
    "ph": "f", "id": 27349, "pid": 76337, "tid": -914061504, "ts": 1716454217641742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641743, "dur": 0,
    "args": {
      "External id": 27350,
      "cbid": 205, "correlation": 27350
    }
  },
  {
    "ph": "f", "id": 27350, "pid": 76337, "tid": -914061504, "ts": 1716454217641743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217678384, "dur": 11,
    "args": {
      "External id": 27354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27354,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27354, "pid": 5, "tid": 7, "ts": 1716454217678384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641757, "dur": 13,
    "args": {
      "External id": 27354,
      "cbid": 211, "correlation": 27354
    }
  },
  {
    "ph": "s", "id": 27354, "pid": 76337, "tid": -914061504, "ts": 1716454217641757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217678396, "dur": 160,
    "args": {
      "External id": 27356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27356, "pid": 5, "tid": 7, "ts": 1716454217678396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641772, "dur": 5,
    "args": {
      "External id": 27356,
      "cbid": 211, "correlation": 27356
    }
  },
  {
    "ph": "s", "id": 27356, "pid": 76337, "tid": -914061504, "ts": 1716454217641772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217678559, "dur": 1,
    "args": {
      "External id": 27358,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 27358,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 27358, "pid": 5, "tid": 7, "ts": 1716454217678559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641783, "dur": 6,
    "args": {
      "External id": 27358,
      "cbid": 51, "correlation": 27358
    }
  },
  {
    "ph": "s", "id": 27358, "pid": 76337, "tid": -914061504, "ts": 1716454217641783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217678562, "dur": 640,
    "args": {
      "External id": 27359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27359,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27359, "pid": 5, "tid": 7, "ts": 1716454217678562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641790, "dur": 6,
    "args": {
      "External id": 27359,
      "cbid": 211, "correlation": 27359
    }
  },
  {
    "ph": "s", "id": 27359, "pid": 76337, "tid": -914061504, "ts": 1716454217641790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217679203, "dur": 13,
    "args": {
      "External id": 27361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27361,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27361, "pid": 5, "tid": 7, "ts": 1716454217679203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641800, "dur": 5,
    "args": {
      "External id": 27361,
      "cbid": 211, "correlation": 27361
    }
  },
  {
    "ph": "s", "id": 27361, "pid": 76337, "tid": -914061504, "ts": 1716454217641800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217679217, "dur": 15,
    "args": {
      "External id": 27367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27367, "pid": 5, "tid": 7, "ts": 1716454217679217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641829, "dur": 9,
    "args": {
      "External id": 27367,
      "cbid": 211, "correlation": 27367
    }
  },
  {
    "ph": "s", "id": 27367, "pid": 76337, "tid": -914061504, "ts": 1716454217641829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641887, "dur": 0,
    "args": {
      "External id": 27377,
      "cbid": 317, "correlation": 27377
    }
  },
  {
    "ph": "f", "id": 27377, "pid": 76337, "tid": -914061504, "ts": 1716454217641887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641887, "dur": 0,
    "args": {
      "External id": 27378,
      "cbid": 203, "correlation": 27378
    }
  },
  {
    "ph": "f", "id": 27378, "pid": 76337, "tid": -914061504, "ts": 1716454217641887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641888, "dur": 0,
    "args": {
      "External id": 27379,
      "cbid": 205, "correlation": 27379
    }
  },
  {
    "ph": "f", "id": 27379, "pid": 76337, "tid": -914061504, "ts": 1716454217641888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217679234, "dur": 17,
    "args": {
      "External id": 27383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27383,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27383, "pid": 5, "tid": 7, "ts": 1716454217679234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641902, "dur": 12,
    "args": {
      "External id": 27383,
      "cbid": 211, "correlation": 27383
    }
  },
  {
    "ph": "s", "id": 27383, "pid": 76337, "tid": -914061504, "ts": 1716454217641902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217679252, "dur": 4,
    "args": {
      "External id": 27385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27385,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27385, "pid": 5, "tid": 7, "ts": 1716454217679252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641919, "dur": 7,
    "args": {
      "External id": 27385,
      "cbid": 211, "correlation": 27385
    }
  },
  {
    "ph": "s", "id": 27385, "pid": 76337, "tid": -914061504, "ts": 1716454217641919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641928, "dur": 0,
    "args": {
      "External id": 27386,
      "cbid": 51, "correlation": 27386
    }
  },
  {
    "ph": "s", "id": 27386, "pid": 76337, "tid": -914061504, "ts": 1716454217641928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217679257, "dur": 129,
    "args": {
      "External id": 27387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27387,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 27387, "pid": 5, "tid": 7, "ts": 1716454217679257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641929, "dur": 5,
    "args": {
      "External id": 27387,
      "cbid": 211, "correlation": 27387
    }
  },
  {
    "ph": "s", "id": 27387, "pid": 76337, "tid": -914061504, "ts": 1716454217641929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217679387, "dur": 15,
    "args": {
      "External id": 27392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27392,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27392, "pid": 5, "tid": 7, "ts": 1716454217679387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641955, "dur": 8,
    "args": {
      "External id": 27392,
      "cbid": 211, "correlation": 27392
    }
  },
  {
    "ph": "s", "id": 27392, "pid": 76337, "tid": -914061504, "ts": 1716454217641955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217679403, "dur": 13,
    "args": {
      "External id": 27400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27400,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27400, "pid": 5, "tid": 7, "ts": 1716454217679403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217641992, "dur": 8,
    "args": {
      "External id": 27400,
      "cbid": 211, "correlation": 27400
    }
  },
  {
    "ph": "s", "id": 27400, "pid": 76337, "tid": -914061504, "ts": 1716454217641992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217679417, "dur": 10,
    "args": {
      "External id": 27408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27408,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27408, "pid": 5, "tid": 7, "ts": 1716454217679417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642021, "dur": 8,
    "args": {
      "External id": 27408,
      "cbid": 211, "correlation": 27408
    }
  },
  {
    "ph": "s", "id": 27408, "pid": 76337, "tid": -914061504, "ts": 1716454217642021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217679429, "dur": 18,
    "args": {
      "External id": 27428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27428,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 27428, "pid": 5, "tid": 7, "ts": 1716454217679429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642104, "dur": 13,
    "args": {
      "External id": 27428,
      "cbid": 211, "correlation": 27428
    }
  },
  {
    "ph": "s", "id": 27428, "pid": 76337, "tid": -914061504, "ts": 1716454217642104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217679448, "dur": 4,
    "args": {
      "External id": 27440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27440,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 27440, "pid": 5, "tid": 7, "ts": 1716454217679448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642127, "dur": 6,
    "args": {
      "External id": 27440,
      "cbid": 211, "correlation": 27440
    }
  },
  {
    "ph": "s", "id": 27440, "pid": 76337, "tid": -914061504, "ts": 1716454217642127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217679454, "dur": 16,
    "args": {
      "External id": 27443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27443, "pid": 5, "tid": 7, "ts": 1716454217679454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642146, "dur": 6,
    "args": {
      "External id": 27443,
      "cbid": 211, "correlation": 27443
    }
  },
  {
    "ph": "s", "id": 27443, "pid": 76337, "tid": -914061504, "ts": 1716454217642146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642203, "dur": 0,
    "args": {
      "External id": 27454,
      "cbid": 317, "correlation": 27454
    }
  },
  {
    "ph": "f", "id": 27454, "pid": 76337, "tid": -914061504, "ts": 1716454217642203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642204, "dur": 0,
    "args": {
      "External id": 27455,
      "cbid": 203, "correlation": 27455
    }
  },
  {
    "ph": "f", "id": 27455, "pid": 76337, "tid": -914061504, "ts": 1716454217642204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642204, "dur": 0,
    "args": {
      "External id": 27456,
      "cbid": 205, "correlation": 27456
    }
  },
  {
    "ph": "f", "id": 27456, "pid": 76337, "tid": -914061504, "ts": 1716454217642204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217679471, "dur": 11,
    "args": {
      "External id": 27460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27460,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27460, "pid": 5, "tid": 7, "ts": 1716454217679471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642217, "dur": 12,
    "args": {
      "External id": 27460,
      "cbid": 211, "correlation": 27460
    }
  },
  {
    "ph": "s", "id": 27460, "pid": 76337, "tid": -914061504, "ts": 1716454217642217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217679484, "dur": 3,
    "args": {
      "External id": 27462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27462, "pid": 5, "tid": 7, "ts": 1716454217679484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642234, "dur": 6,
    "args": {
      "External id": 27462,
      "cbid": 211, "correlation": 27462
    }
  },
  {
    "ph": "s", "id": 27462, "pid": 76337, "tid": -914061504, "ts": 1716454217642234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642243, "dur": 0,
    "args": {
      "External id": 27463,
      "cbid": 51, "correlation": 27463
    }
  },
  {
    "ph": "s", "id": 27463, "pid": 76337, "tid": -914061504, "ts": 1716454217642243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217679489, "dur": 89,
    "args": {
      "External id": 27464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27464,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 27464, "pid": 5, "tid": 7, "ts": 1716454217679489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642243, "dur": 5,
    "args": {
      "External id": 27464,
      "cbid": 211, "correlation": 27464
    }
  },
  {
    "ph": "s", "id": 27464, "pid": 76337, "tid": -914061504, "ts": 1716454217642243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217679579, "dur": 15,
    "args": {
      "External id": 27469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27469,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27469, "pid": 5, "tid": 7, "ts": 1716454217679579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642270, "dur": 9,
    "args": {
      "External id": 27469,
      "cbid": 211, "correlation": 27469
    }
  },
  {
    "ph": "s", "id": 27469, "pid": 76337, "tid": -914061504, "ts": 1716454217642270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217679595, "dur": 82,
    "args": {
      "External id": 27478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27478,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27478, "pid": 5, "tid": 7, "ts": 1716454217679595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642353, "dur": 14,
    "args": {
      "External id": 27478,
      "cbid": 211, "correlation": 27478
    }
  },
  {
    "ph": "s", "id": 27478, "pid": 76337, "tid": -914061504, "ts": 1716454217642353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217679679, "dur": 28,
    "args": {
      "External id": 27500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27500,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27500, "pid": 5, "tid": 7, "ts": 1716454217679679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642409, "dur": 10,
    "args": {
      "External id": 27500,
      "cbid": 211, "correlation": 27500
    }
  },
  {
    "ph": "s", "id": 27500, "pid": 76337, "tid": -914061504, "ts": 1716454217642409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642525, "dur": 1,
    "args": {
      "External id": 27511,
      "cbid": 251, "correlation": 27511
    }
  },
  {
    "ph": "f", "id": 27511, "pid": 76337, "tid": -914061504, "ts": 1716454217642525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217679708, "dur": 162,
    "args": {
      "External id": 27512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27512,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27512, "pid": 5, "tid": 7, "ts": 1716454217679708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642531, "dur": 15,
    "args": {
      "External id": 27512,
      "cbid": 211, "correlation": 27512
    }
  },
  {
    "ph": "s", "id": 27512, "pid": 76337, "tid": -914061504, "ts": 1716454217642531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642614, "dur": 1,
    "args": {
      "External id": 27523,
      "cbid": 251, "correlation": 27523
    }
  },
  {
    "ph": "f", "id": 27523, "pid": 76337, "tid": -914061504, "ts": 1716454217642614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217679871, "dur": 155,
    "args": {
      "External id": 27524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27524,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27524, "pid": 5, "tid": 7, "ts": 1716454217679871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642618, "dur": 12,
    "args": {
      "External id": 27524,
      "cbid": 211, "correlation": 27524
    }
  },
  {
    "ph": "s", "id": 27524, "pid": 76337, "tid": -914061504, "ts": 1716454217642618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642696, "dur": 1,
    "args": {
      "External id": 27535,
      "cbid": 251, "correlation": 27535
    }
  },
  {
    "ph": "f", "id": 27535, "pid": 76337, "tid": -914061504, "ts": 1716454217642696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217680028, "dur": 154,
    "args": {
      "External id": 27536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27536,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27536, "pid": 5, "tid": 7, "ts": 1716454217680028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642700, "dur": 13,
    "args": {
      "External id": 27536,
      "cbid": 211, "correlation": 27536
    }
  },
  {
    "ph": "s", "id": 27536, "pid": 76337, "tid": -914061504, "ts": 1716454217642700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217680183, "dur": 330,
    "args": {
      "External id": 27561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27561,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27561, "pid": 5, "tid": 7, "ts": 1716454217680183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642786, "dur": 13,
    "args": {
      "External id": 27561,
      "cbid": 211, "correlation": 27561
    }
  },
  {
    "ph": "s", "id": 27561, "pid": 76337, "tid": -914061504, "ts": 1716454217642786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642900, "dur": 1,
    "args": {
      "External id": 27579,
      "cbid": 251, "correlation": 27579
    }
  },
  {
    "ph": "f", "id": 27579, "pid": 76337, "tid": -914061504, "ts": 1716454217642900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217680515, "dur": 164,
    "args": {
      "External id": 27581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27581,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27581, "pid": 5, "tid": 7, "ts": 1716454217680515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642906, "dur": 14,
    "args": {
      "External id": 27581,
      "cbid": 211, "correlation": 27581
    }
  },
  {
    "ph": "s", "id": 27581, "pid": 76337, "tid": -914061504, "ts": 1716454217642906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217680680, "dur": 19,
    "args": {
      "External id": 27589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27589,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27589, "pid": 5, "tid": 7, "ts": 1716454217680680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217642987, "dur": 13,
    "args": {
      "External id": 27589,
      "cbid": 211, "correlation": 27589
    }
  },
  {
    "ph": "s", "id": 27589, "pid": 76337, "tid": -914061504, "ts": 1716454217642987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217680700, "dur": 27,
    "args": {
      "External id": 27597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27597,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27597, "pid": 5, "tid": 7, "ts": 1716454217680700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643027, "dur": 9,
    "args": {
      "External id": 27597,
      "cbid": 211, "correlation": 27597
    }
  },
  {
    "ph": "s", "id": 27597, "pid": 76337, "tid": -914061504, "ts": 1716454217643027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217680729, "dur": 18,
    "args": {
      "External id": 27608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27608, "pid": 5, "tid": 7, "ts": 1716454217680729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643101, "dur": 13,
    "args": {
      "External id": 27608,
      "cbid": 211, "correlation": 27608
    }
  },
  {
    "ph": "s", "id": 27608, "pid": 76337, "tid": -914061504, "ts": 1716454217643101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217680748, "dur": 16,
    "args": {
      "External id": 27630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27630,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27630, "pid": 5, "tid": 7, "ts": 1716454217680748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643132, "dur": 8,
    "args": {
      "External id": 27630,
      "cbid": 211, "correlation": 27630
    }
  },
  {
    "ph": "s", "id": 27630, "pid": 76337, "tid": -914061504, "ts": 1716454217643132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643227, "dur": 1,
    "args": {
      "External id": 27641,
      "cbid": 251, "correlation": 27641
    }
  },
  {
    "ph": "f", "id": 27641, "pid": 76337, "tid": -914061504, "ts": 1716454217643227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217680765, "dur": 88,
    "args": {
      "External id": 27642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27642,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27642, "pid": 5, "tid": 7, "ts": 1716454217680765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643232, "dur": 14,
    "args": {
      "External id": 27642,
      "cbid": 211, "correlation": 27642
    }
  },
  {
    "ph": "s", "id": 27642, "pid": 76337, "tid": -914061504, "ts": 1716454217643232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643303, "dur": 1,
    "args": {
      "External id": 27653,
      "cbid": 251, "correlation": 27653
    }
  },
  {
    "ph": "f", "id": 27653, "pid": 76337, "tid": -914061504, "ts": 1716454217643303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643307, "dur": 0,
    "args": {
      "External id": 27654,
      "cbid": 251, "correlation": 27654
    }
  },
  {
    "ph": "f", "id": 27654, "pid": 76337, "tid": -914061504, "ts": 1716454217643307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217680854, "dur": 11,
    "args": {
      "External id": 27655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27655,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27655, "pid": 5, "tid": 7, "ts": 1716454217680854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643309, "dur": 12,
    "args": {
      "External id": 27655,
      "cbid": 211, "correlation": 27655
    }
  },
  {
    "ph": "s", "id": 27655, "pid": 76337, "tid": -914061504, "ts": 1716454217643309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217680867, "dur": 5,
    "args": {
      "External id": 27657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27657,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27657, "pid": 5, "tid": 7, "ts": 1716454217680867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643323, "dur": 6,
    "args": {
      "External id": 27657,
      "cbid": 211, "correlation": 27657
    }
  },
  {
    "ph": "s", "id": 27657, "pid": 76337, "tid": -914061504, "ts": 1716454217643323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643396, "dur": 1,
    "args": {
      "External id": 27668,
      "cbid": 251, "correlation": 27668
    }
  },
  {
    "ph": "f", "id": 27668, "pid": 76337, "tid": -914061504, "ts": 1716454217643396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643401, "dur": 0,
    "args": {
      "External id": 27669,
      "cbid": 251, "correlation": 27669
    }
  },
  {
    "ph": "f", "id": 27669, "pid": 76337, "tid": -914061504, "ts": 1716454217643401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217680873, "dur": 9,
    "args": {
      "External id": 27670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27670,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27670, "pid": 5, "tid": 7, "ts": 1716454217680873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643402, "dur": 12,
    "args": {
      "External id": 27670,
      "cbid": 211, "correlation": 27670
    }
  },
  {
    "ph": "s", "id": 27670, "pid": 76337, "tid": -914061504, "ts": 1716454217643402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217680883, "dur": 3,
    "args": {
      "External id": 27672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27672,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27672, "pid": 5, "tid": 7, "ts": 1716454217680883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643416, "dur": 5,
    "args": {
      "External id": 27672,
      "cbid": 211, "correlation": 27672
    }
  },
  {
    "ph": "s", "id": 27672, "pid": 76337, "tid": -914061504, "ts": 1716454217643416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217680888, "dur": 55,
    "args": {
      "External id": 27697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27697,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27697, "pid": 5, "tid": 7, "ts": 1716454217680888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643493, "dur": 13,
    "args": {
      "External id": 27697,
      "cbid": 211, "correlation": 27697
    }
  },
  {
    "ph": "s", "id": 27697, "pid": 76337, "tid": -914061504, "ts": 1716454217643493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643604, "dur": 1,
    "args": {
      "External id": 27715,
      "cbid": 251, "correlation": 27715
    }
  },
  {
    "ph": "f", "id": 27715, "pid": 76337, "tid": -914061504, "ts": 1716454217643604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217680944, "dur": 90,
    "args": {
      "External id": 27717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27717,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27717, "pid": 5, "tid": 7, "ts": 1716454217680944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643610, "dur": 14,
    "args": {
      "External id": 27717,
      "cbid": 211, "correlation": 27717
    }
  },
  {
    "ph": "s", "id": 27717, "pid": 76337, "tid": -914061504, "ts": 1716454217643610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217681035, "dur": 10,
    "args": {
      "External id": 27725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27725,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27725, "pid": 5, "tid": 7, "ts": 1716454217681035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643683, "dur": 12,
    "args": {
      "External id": 27725,
      "cbid": 211, "correlation": 27725
    }
  },
  {
    "ph": "s", "id": 27725, "pid": 76337, "tid": -914061504, "ts": 1716454217643683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217681046, "dur": 21,
    "args": {
      "External id": 27733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27733, "pid": 5, "tid": 7, "ts": 1716454217681046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643724, "dur": 10,
    "args": {
      "External id": 27733,
      "cbid": 211, "correlation": 27733
    }
  },
  {
    "ph": "s", "id": 27733, "pid": 76337, "tid": -914061504, "ts": 1716454217643724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217681068, "dur": 17,
    "args": {
      "External id": 27755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27755,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27755, "pid": 5, "tid": 7, "ts": 1716454217681068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643776, "dur": 10,
    "args": {
      "External id": 27755,
      "cbid": 211, "correlation": 27755
    }
  },
  {
    "ph": "s", "id": 27755, "pid": 76337, "tid": -914061504, "ts": 1716454217643776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643877, "dur": 1,
    "args": {
      "External id": 27771,
      "cbid": 251, "correlation": 27771
    }
  },
  {
    "ph": "f", "id": 27771, "pid": 76337, "tid": -914061504, "ts": 1716454217643877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643882, "dur": 0,
    "args": {
      "External id": 27773,
      "cbid": 251, "correlation": 27773
    }
  },
  {
    "ph": "f", "id": 27773, "pid": 76337, "tid": -914061504, "ts": 1716454217643882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217681087, "dur": 495,
    "args": {
      "External id": 27774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27774,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27774, "pid": 5, "tid": 7, "ts": 1716454217681087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643884, "dur": 14,
    "args": {
      "External id": 27774,
      "cbid": 211, "correlation": 27774
    }
  },
  {
    "ph": "s", "id": 27774, "pid": 76337, "tid": -914061504, "ts": 1716454217643884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217681583, "dur": 64,
    "args": {
      "External id": 27782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27782, "pid": 5, "tid": 7, "ts": 1716454217681583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643952, "dur": 12,
    "args": {
      "External id": 27782,
      "cbid": 211, "correlation": 27782
    }
  },
  {
    "ph": "s", "id": 27782, "pid": 76337, "tid": -914061504, "ts": 1716454217643952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217681649, "dur": 65,
    "args": {
      "External id": 27790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27790, "pid": 5, "tid": 7, "ts": 1716454217681649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217643990, "dur": 9,
    "args": {
      "External id": 27790,
      "cbid": 211, "correlation": 27790
    }
  },
  {
    "ph": "s", "id": 27790, "pid": 76337, "tid": -914061504, "ts": 1716454217643990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644072, "dur": 1,
    "args": {
      "External id": 27806,
      "cbid": 251, "correlation": 27806
    }
  },
  {
    "ph": "f", "id": 27806, "pid": 76337, "tid": -914061504, "ts": 1716454217644072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454217681716, "dur": 1,
    "args": {
      "External id": 27808,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 27808,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 27808, "pid": 5, "tid": 7, "ts": 1716454217681716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644077, "dur": 10,
    "args": {
      "External id": 27808,
      "cbid": 51, "correlation": 27808
    }
  },
  {
    "ph": "s", "id": 27808, "pid": 76337, "tid": -914061504, "ts": 1716454217644077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217681720, "dur": 267,
    "args": {
      "External id": 27809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27809,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27809, "pid": 5, "tid": 7, "ts": 1716454217681720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644088, "dur": 11,
    "args": {
      "External id": 27809,
      "cbid": 211, "correlation": 27809
    }
  },
  {
    "ph": "s", "id": 27809, "pid": 76337, "tid": -914061504, "ts": 1716454217644088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217681988, "dur": 14,
    "args": {
      "External id": 27817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27817,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27817, "pid": 5, "tid": 7, "ts": 1716454217681988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644130, "dur": 11,
    "args": {
      "External id": 27817,
      "cbid": 211, "correlation": 27817
    }
  },
  {
    "ph": "s", "id": 27817, "pid": 76337, "tid": -914061504, "ts": 1716454217644130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217682003, "dur": 37,
    "args": {
      "External id": 27828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27828,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27828, "pid": 5, "tid": 7, "ts": 1716454217682003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644199, "dur": 12,
    "args": {
      "External id": 27828,
      "cbid": 211, "correlation": 27828
    }
  },
  {
    "ph": "s", "id": 27828, "pid": 76337, "tid": -914061504, "ts": 1716454217644199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644263, "dur": 0,
    "args": {
      "External id": 27840,
      "cbid": 317, "correlation": 27840
    }
  },
  {
    "ph": "f", "id": 27840, "pid": 76337, "tid": -914061504, "ts": 1716454217644263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644264, "dur": 0,
    "args": {
      "External id": 27841,
      "cbid": 203, "correlation": 27841
    }
  },
  {
    "ph": "f", "id": 27841, "pid": 76337, "tid": -914061504, "ts": 1716454217644264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644265, "dur": 0,
    "args": {
      "External id": 27842,
      "cbid": 205, "correlation": 27842
    }
  },
  {
    "ph": "f", "id": 27842, "pid": 76337, "tid": -914061504, "ts": 1716454217644265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217682041, "dur": 13,
    "args": {
      "External id": 27846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27846, "pid": 5, "tid": 7, "ts": 1716454217682041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644280, "dur": 13,
    "args": {
      "External id": 27846,
      "cbid": 211, "correlation": 27846
    }
  },
  {
    "ph": "s", "id": 27846, "pid": 76337, "tid": -914061504, "ts": 1716454217644280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217682055, "dur": 4,
    "args": {
      "External id": 27848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 27848, "pid": 5, "tid": 7, "ts": 1716454217682055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644298, "dur": 6,
    "args": {
      "External id": 27848,
      "cbid": 211, "correlation": 27848
    }
  },
  {
    "ph": "s", "id": 27848, "pid": 76337, "tid": -914061504, "ts": 1716454217644298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644306, "dur": 0,
    "args": {
      "External id": 27849,
      "cbid": 51, "correlation": 27849
    }
  },
  {
    "ph": "s", "id": 27849, "pid": 76337, "tid": -914061504, "ts": 1716454217644306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217682060, "dur": 96,
    "args": {
      "External id": 27850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27850,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 27850, "pid": 5, "tid": 7, "ts": 1716454217682060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644307, "dur": 5,
    "args": {
      "External id": 27850,
      "cbid": 211, "correlation": 27850
    }
  },
  {
    "ph": "s", "id": 27850, "pid": 76337, "tid": -914061504, "ts": 1716454217644307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217682158, "dur": 16,
    "args": {
      "External id": 27855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27855, "pid": 5, "tid": 7, "ts": 1716454217682158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644333, "dur": 10,
    "args": {
      "External id": 27855,
      "cbid": 211, "correlation": 27855
    }
  },
  {
    "ph": "s", "id": 27855, "pid": 76337, "tid": -914061504, "ts": 1716454217644333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217682175, "dur": 11,
    "args": {
      "External id": 27863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27863,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27863, "pid": 5, "tid": 7, "ts": 1716454217682175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644365, "dur": 8,
    "args": {
      "External id": 27863,
      "cbid": 211, "correlation": 27863
    }
  },
  {
    "ph": "s", "id": 27863, "pid": 76337, "tid": -914061504, "ts": 1716454217644365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454217682188, "dur": 55,
    "args": {
      "External id": 27874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27874,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27874, "pid": 5, "tid": 7, "ts": 1716454217682188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644430, "dur": 12,
    "args": {
      "External id": 27874,
      "cbid": 211, "correlation": 27874
    }
  },
  {
    "ph": "s", "id": 27874, "pid": 76337, "tid": -914061504, "ts": 1716454217644430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644485, "dur": 0,
    "args": {
      "External id": 27884,
      "cbid": 317, "correlation": 27884
    }
  },
  {
    "ph": "f", "id": 27884, "pid": 76337, "tid": -914061504, "ts": 1716454217644485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644486, "dur": 0,
    "args": {
      "External id": 27885,
      "cbid": 203, "correlation": 27885
    }
  },
  {
    "ph": "f", "id": 27885, "pid": 76337, "tid": -914061504, "ts": 1716454217644486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644487, "dur": 0,
    "args": {
      "External id": 27886,
      "cbid": 205, "correlation": 27886
    }
  },
  {
    "ph": "f", "id": 27886, "pid": 76337, "tid": -914061504, "ts": 1716454217644487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217682244, "dur": 38,
    "args": {
      "External id": 27890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27890,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27890, "pid": 5, "tid": 7, "ts": 1716454217682244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644504, "dur": 11,
    "args": {
      "External id": 27890,
      "cbid": 211, "correlation": 27890
    }
  },
  {
    "ph": "s", "id": 27890, "pid": 76337, "tid": -914061504, "ts": 1716454217644504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217682284, "dur": 160,
    "args": {
      "External id": 27892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27892,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27892, "pid": 5, "tid": 7, "ts": 1716454217682284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644518, "dur": 5,
    "args": {
      "External id": 27892,
      "cbid": 211, "correlation": 27892
    }
  },
  {
    "ph": "s", "id": 27892, "pid": 76337, "tid": -914061504, "ts": 1716454217644518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217682446, "dur": 1954,
    "args": {
      "External id": 27894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27894,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27894, "pid": 5, "tid": 7, "ts": 1716454217682446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644531, "dur": 8,
    "args": {
      "External id": 27894,
      "cbid": 211, "correlation": 27894
    }
  },
  {
    "ph": "s", "id": 27894, "pid": 76337, "tid": -914061504, "ts": 1716454217644531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217684401, "dur": 40,
    "args": {
      "External id": 27896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27896,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27896, "pid": 5, "tid": 7, "ts": 1716454217684401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644543, "dur": 6,
    "args": {
      "External id": 27896,
      "cbid": 211, "correlation": 27896
    }
  },
  {
    "ph": "s", "id": 27896, "pid": 76337, "tid": -914061504, "ts": 1716454217644543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217684442, "dur": 58,
    "args": {
      "External id": 27902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27902, "pid": 5, "tid": 7, "ts": 1716454217684442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644571, "dur": 8,
    "args": {
      "External id": 27902,
      "cbid": 211, "correlation": 27902
    }
  },
  {
    "ph": "s", "id": 27902, "pid": 76337, "tid": -914061504, "ts": 1716454217644571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217684502, "dur": 82,
    "args": {
      "External id": 27911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27911,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27911, "pid": 5, "tid": 7, "ts": 1716454217684502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644662, "dur": 13,
    "args": {
      "External id": 27911,
      "cbid": 211, "correlation": 27911
    }
  },
  {
    "ph": "s", "id": 27911, "pid": 76337, "tid": -914061504, "ts": 1716454217644662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217684585, "dur": 74,
    "args": {
      "External id": 27931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27931,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 27931, "pid": 5, "tid": 7, "ts": 1716454217684585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644734, "dur": 11,
    "args": {
      "External id": 27931,
      "cbid": 211, "correlation": 27931
    }
  },
  {
    "ph": "s", "id": 27931, "pid": 76337, "tid": -914061504, "ts": 1716454217644734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217684660, "dur": 5,
    "args": {
      "External id": 27943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27943,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 27943, "pid": 5, "tid": 7, "ts": 1716454217684660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644754, "dur": 7,
    "args": {
      "External id": 27943,
      "cbid": 211, "correlation": 27943
    }
  },
  {
    "ph": "s", "id": 27943, "pid": 76337, "tid": -914061504, "ts": 1716454217644754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217684666, "dur": 81,
    "args": {
      "External id": 27946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27946, "pid": 5, "tid": 7, "ts": 1716454217684666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644774, "dur": 7,
    "args": {
      "External id": 27946,
      "cbid": 211, "correlation": 27946
    }
  },
  {
    "ph": "s", "id": 27946, "pid": 76337, "tid": -914061504, "ts": 1716454217644774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217684748, "dur": 53,
    "args": {
      "External id": 27955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27955,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27955, "pid": 5, "tid": 7, "ts": 1716454217684748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644814, "dur": 10,
    "args": {
      "External id": 27955,
      "cbid": 211, "correlation": 27955
    }
  },
  {
    "ph": "s", "id": 27955, "pid": 76337, "tid": -914061504, "ts": 1716454217644814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644878, "dur": 0,
    "args": {
      "External id": 27965,
      "cbid": 317, "correlation": 27965
    }
  },
  {
    "ph": "f", "id": 27965, "pid": 76337, "tid": -914061504, "ts": 1716454217644878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644879, "dur": 0,
    "args": {
      "External id": 27966,
      "cbid": 203, "correlation": 27966
    }
  },
  {
    "ph": "f", "id": 27966, "pid": 76337, "tid": -914061504, "ts": 1716454217644879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644880, "dur": 0,
    "args": {
      "External id": 27967,
      "cbid": 205, "correlation": 27967
    }
  },
  {
    "ph": "f", "id": 27967, "pid": 76337, "tid": -914061504, "ts": 1716454217644880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217684802, "dur": 58,
    "args": {
      "External id": 27971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27971,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27971, "pid": 5, "tid": 7, "ts": 1716454217684802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644896, "dur": 12,
    "args": {
      "External id": 27971,
      "cbid": 211, "correlation": 27971
    }
  },
  {
    "ph": "s", "id": 27971, "pid": 76337, "tid": -914061504, "ts": 1716454217644896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217684861, "dur": 121,
    "args": {
      "External id": 27973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27973,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27973, "pid": 5, "tid": 7, "ts": 1716454217684861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644911, "dur": 5,
    "args": {
      "External id": 27973,
      "cbid": 211, "correlation": 27973
    }
  },
  {
    "ph": "s", "id": 27973, "pid": 76337, "tid": -914061504, "ts": 1716454217644911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217684983, "dur": 1870,
    "args": {
      "External id": 27975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27975,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 27975, "pid": 5, "tid": 7, "ts": 1716454217684983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644923, "dur": 6,
    "args": {
      "External id": 27975,
      "cbid": 211, "correlation": 27975
    }
  },
  {
    "ph": "s", "id": 27975, "pid": 76337, "tid": -914061504, "ts": 1716454217644923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217686855, "dur": 21,
    "args": {
      "External id": 27977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27977,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27977, "pid": 5, "tid": 7, "ts": 1716454217686855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644932, "dur": 5,
    "args": {
      "External id": 27977,
      "cbid": 211, "correlation": 27977
    }
  },
  {
    "ph": "s", "id": 27977, "pid": 76337, "tid": -914061504, "ts": 1716454217644932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217686877, "dur": 32,
    "args": {
      "External id": 27983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 27983, "pid": 5, "tid": 7, "ts": 1716454217686877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217644961, "dur": 9,
    "args": {
      "External id": 27983,
      "cbid": 211, "correlation": 27983
    }
  },
  {
    "ph": "s", "id": 27983, "pid": 76337, "tid": -914061504, "ts": 1716454217644961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217686911, "dur": 3,
    "args": {
      "External id": 27991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 27991,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 27991, "pid": 5, "tid": 7, "ts": 1716454217686911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645014, "dur": 11,
    "args": {
      "External id": 27991,
      "cbid": 211, "correlation": 27991
    }
  },
  {
    "ph": "s", "id": 27991, "pid": 76337, "tid": -914061504, "ts": 1716454217645014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645081, "dur": 1,
    "args": {
      "External id": 28007,
      "cbid": 251, "correlation": 28007
    }
  },
  {
    "ph": "f", "id": 28007, "pid": 76337, "tid": -914061504, "ts": 1716454217645081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645086, "dur": 0,
    "args": {
      "External id": 28009,
      "cbid": 251, "correlation": 28009
    }
  },
  {
    "ph": "f", "id": 28009, "pid": 76337, "tid": -914061504, "ts": 1716454217645086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217686916, "dur": 12,
    "args": {
      "External id": 28010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28010,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 28010, "pid": 5, "tid": 7, "ts": 1716454217686916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645088, "dur": 11,
    "args": {
      "External id": 28010,
      "cbid": 211, "correlation": 28010
    }
  },
  {
    "ph": "s", "id": 28010, "pid": 76337, "tid": -914061504, "ts": 1716454217645088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217686929, "dur": 5,
    "args": {
      "External id": 28012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28012,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 28012, "pid": 5, "tid": 7, "ts": 1716454217686929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645101, "dur": 6,
    "args": {
      "External id": 28012,
      "cbid": 211, "correlation": 28012
    }
  },
  {
    "ph": "s", "id": 28012, "pid": 76337, "tid": -914061504, "ts": 1716454217645101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217686936, "dur": 29,
    "args": {
      "External id": 28022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28022,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28022, "pid": 5, "tid": 7, "ts": 1716454217686936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645160, "dur": 12,
    "args": {
      "External id": 28022,
      "cbid": 211, "correlation": 28022
    }
  },
  {
    "ph": "s", "id": 28022, "pid": 76337, "tid": -914061504, "ts": 1716454217645160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217686966, "dur": 31,
    "args": {
      "External id": 28042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28042,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 28042, "pid": 5, "tid": 7, "ts": 1716454217686966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645225, "dur": 11,
    "args": {
      "External id": 28042,
      "cbid": 211, "correlation": 28042
    }
  },
  {
    "ph": "s", "id": 28042, "pid": 76337, "tid": -914061504, "ts": 1716454217645225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217686998, "dur": 4,
    "args": {
      "External id": 28054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28054,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 28054, "pid": 5, "tid": 7, "ts": 1716454217686998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645247, "dur": 6,
    "args": {
      "External id": 28054,
      "cbid": 211, "correlation": 28054
    }
  },
  {
    "ph": "s", "id": 28054, "pid": 76337, "tid": -914061504, "ts": 1716454217645247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217687004, "dur": 30,
    "args": {
      "External id": 28057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28057, "pid": 5, "tid": 7, "ts": 1716454217687004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645264, "dur": 6,
    "args": {
      "External id": 28057,
      "cbid": 211, "correlation": 28057
    }
  },
  {
    "ph": "s", "id": 28057, "pid": 76337, "tid": -914061504, "ts": 1716454217645264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217687035, "dur": 20,
    "args": {
      "External id": 28066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28066,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28066, "pid": 5, "tid": 7, "ts": 1716454217687035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645305, "dur": 10,
    "args": {
      "External id": 28066,
      "cbid": 211, "correlation": 28066
    }
  },
  {
    "ph": "s", "id": 28066, "pid": 76337, "tid": -914061504, "ts": 1716454217645305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645382, "dur": 0,
    "args": {
      "External id": 28076,
      "cbid": 317, "correlation": 28076
    }
  },
  {
    "ph": "f", "id": 28076, "pid": 76337, "tid": -914061504, "ts": 1716454217645382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645382, "dur": 0,
    "args": {
      "External id": 28077,
      "cbid": 203, "correlation": 28077
    }
  },
  {
    "ph": "f", "id": 28077, "pid": 76337, "tid": -914061504, "ts": 1716454217645382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645383, "dur": 0,
    "args": {
      "External id": 28078,
      "cbid": 205, "correlation": 28078
    }
  },
  {
    "ph": "f", "id": 28078, "pid": 76337, "tid": -914061504, "ts": 1716454217645383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217687057, "dur": 22,
    "args": {
      "External id": 28082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28082, "pid": 5, "tid": 7, "ts": 1716454217687057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645399, "dur": 13,
    "args": {
      "External id": 28082,
      "cbid": 211, "correlation": 28082
    }
  },
  {
    "ph": "s", "id": 28082, "pid": 76337, "tid": -914061504, "ts": 1716454217645399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217687080, "dur": 44,
    "args": {
      "External id": 28084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28084, "pid": 5, "tid": 7, "ts": 1716454217687080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645415, "dur": 5,
    "args": {
      "External id": 28084,
      "cbid": 211, "correlation": 28084
    }
  },
  {
    "ph": "s", "id": 28084, "pid": 76337, "tid": -914061504, "ts": 1716454217645415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217687125, "dur": 641,
    "args": {
      "External id": 28086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28086,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28086, "pid": 5, "tid": 7, "ts": 1716454217687125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645427, "dur": 6,
    "args": {
      "External id": 28086,
      "cbid": 211, "correlation": 28086
    }
  },
  {
    "ph": "s", "id": 28086, "pid": 76337, "tid": -914061504, "ts": 1716454217645427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217687767, "dur": 22,
    "args": {
      "External id": 28088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28088, "pid": 5, "tid": 7, "ts": 1716454217687767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645436, "dur": 5,
    "args": {
      "External id": 28088,
      "cbid": 211, "correlation": 28088
    }
  },
  {
    "ph": "s", "id": 28088, "pid": 76337, "tid": -914061504, "ts": 1716454217645436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217687791, "dur": 32,
    "args": {
      "External id": 28094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28094, "pid": 5, "tid": 7, "ts": 1716454217687791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645465, "dur": 9,
    "args": {
      "External id": 28094,
      "cbid": 211, "correlation": 28094
    }
  },
  {
    "ph": "s", "id": 28094, "pid": 76337, "tid": -914061504, "ts": 1716454217645465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645524, "dur": 0,
    "args": {
      "External id": 28104,
      "cbid": 317, "correlation": 28104
    }
  },
  {
    "ph": "f", "id": 28104, "pid": 76337, "tid": -914061504, "ts": 1716454217645524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645524, "dur": 0,
    "args": {
      "External id": 28105,
      "cbid": 203, "correlation": 28105
    }
  },
  {
    "ph": "f", "id": 28105, "pid": 76337, "tid": -914061504, "ts": 1716454217645524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645525, "dur": 0,
    "args": {
      "External id": 28106,
      "cbid": 205, "correlation": 28106
    }
  },
  {
    "ph": "f", "id": 28106, "pid": 76337, "tid": -914061504, "ts": 1716454217645525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217687824, "dur": 55,
    "args": {
      "External id": 28110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28110, "pid": 5, "tid": 7, "ts": 1716454217687824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645538, "dur": 11,
    "args": {
      "External id": 28110,
      "cbid": 211, "correlation": 28110
    }
  },
  {
    "ph": "s", "id": 28110, "pid": 76337, "tid": -914061504, "ts": 1716454217645538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217687881, "dur": 266,
    "args": {
      "External id": 28112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28112,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28112, "pid": 5, "tid": 7, "ts": 1716454217687881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645556, "dur": 8,
    "args": {
      "External id": 28112,
      "cbid": 211, "correlation": 28112
    }
  },
  {
    "ph": "s", "id": 28112, "pid": 76337, "tid": -914061504, "ts": 1716454217645556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217688148, "dur": 21,
    "args": {
      "External id": 28114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28114,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28114, "pid": 5, "tid": 7, "ts": 1716454217688148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645567, "dur": 5,
    "args": {
      "External id": 28114,
      "cbid": 211, "correlation": 28114
    }
  },
  {
    "ph": "s", "id": 28114, "pid": 76337, "tid": -914061504, "ts": 1716454217645567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217688170, "dur": 32,
    "args": {
      "External id": 28120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28120, "pid": 5, "tid": 7, "ts": 1716454217688170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645593, "dur": 8,
    "args": {
      "External id": 28120,
      "cbid": 211, "correlation": 28120
    }
  },
  {
    "ph": "s", "id": 28120, "pid": 76337, "tid": -914061504, "ts": 1716454217645593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217688203, "dur": 27,
    "args": {
      "External id": 28128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28128,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28128, "pid": 5, "tid": 7, "ts": 1716454217688203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645622, "dur": 8,
    "args": {
      "External id": 28128,
      "cbid": 211, "correlation": 28128
    }
  },
  {
    "ph": "s", "id": 28128, "pid": 76337, "tid": -914061504, "ts": 1716454217645622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217688232, "dur": 20,
    "args": {
      "External id": 28136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28136,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28136, "pid": 5, "tid": 7, "ts": 1716454217688232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645650, "dur": 9,
    "args": {
      "External id": 28136,
      "cbid": 211, "correlation": 28136
    }
  },
  {
    "ph": "s", "id": 28136, "pid": 76337, "tid": -914061504, "ts": 1716454217645650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217688253, "dur": 30,
    "args": {
      "External id": 28156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28156,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 28156, "pid": 5, "tid": 7, "ts": 1716454217688253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645733, "dur": 12,
    "args": {
      "External id": 28156,
      "cbid": 211, "correlation": 28156
    }
  },
  {
    "ph": "s", "id": 28156, "pid": 76337, "tid": -914061504, "ts": 1716454217645733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217688284, "dur": 5,
    "args": {
      "External id": 28168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28168,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 28168, "pid": 5, "tid": 7, "ts": 1716454217688284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645755, "dur": 6,
    "args": {
      "External id": 28168,
      "cbid": 211, "correlation": 28168
    }
  },
  {
    "ph": "s", "id": 28168, "pid": 76337, "tid": -914061504, "ts": 1716454217645755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217688290, "dur": 31,
    "args": {
      "External id": 28171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28171,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28171, "pid": 5, "tid": 7, "ts": 1716454217688290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645773, "dur": 6,
    "args": {
      "External id": 28171,
      "cbid": 211, "correlation": 28171
    }
  },
  {
    "ph": "s", "id": 28171, "pid": 76337, "tid": -914061504, "ts": 1716454217645773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645842, "dur": 0,
    "args": {
      "External id": 28182,
      "cbid": 317, "correlation": 28182
    }
  },
  {
    "ph": "f", "id": 28182, "pid": 76337, "tid": -914061504, "ts": 1716454217645842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645843, "dur": 0,
    "args": {
      "External id": 28183,
      "cbid": 203, "correlation": 28183
    }
  },
  {
    "ph": "f", "id": 28183, "pid": 76337, "tid": -914061504, "ts": 1716454217645843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645844, "dur": 0,
    "args": {
      "External id": 28184,
      "cbid": 205, "correlation": 28184
    }
  },
  {
    "ph": "f", "id": 28184, "pid": 76337, "tid": -914061504, "ts": 1716454217645844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217688322, "dur": 22,
    "args": {
      "External id": 28188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28188,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28188, "pid": 5, "tid": 7, "ts": 1716454217688322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645858, "dur": 13,
    "args": {
      "External id": 28188,
      "cbid": 211, "correlation": 28188
    }
  },
  {
    "ph": "s", "id": 28188, "pid": 76337, "tid": -914061504, "ts": 1716454217645858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217688345, "dur": 103,
    "args": {
      "External id": 28190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28190,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28190, "pid": 5, "tid": 7, "ts": 1716454217688345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645878, "dur": 6,
    "args": {
      "External id": 28190,
      "cbid": 211, "correlation": 28190
    }
  },
  {
    "ph": "s", "id": 28190, "pid": 76337, "tid": -914061504, "ts": 1716454217645878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217688450, "dur": 21,
    "args": {
      "External id": 28192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28192, "pid": 5, "tid": 7, "ts": 1716454217688450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645888, "dur": 5,
    "args": {
      "External id": 28192,
      "cbid": 211, "correlation": 28192
    }
  },
  {
    "ph": "s", "id": 28192, "pid": 76337, "tid": -914061504, "ts": 1716454217645888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217688472, "dur": 32,
    "args": {
      "External id": 28198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28198, "pid": 5, "tid": 7, "ts": 1716454217688472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217645927, "dur": 10,
    "args": {
      "External id": 28198,
      "cbid": 211, "correlation": 28198
    }
  },
  {
    "ph": "s", "id": 28198, "pid": 76337, "tid": -914061504, "ts": 1716454217645927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217688505, "dur": 192,
    "args": {
      "External id": 28207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28207,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28207, "pid": 5, "tid": 7, "ts": 1716454217688505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646020, "dur": 15,
    "args": {
      "External id": 28207,
      "cbid": 211, "correlation": 28207
    }
  },
  {
    "ph": "s", "id": 28207, "pid": 76337, "tid": -914061504, "ts": 1716454217646020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217688698, "dur": 64,
    "args": {
      "External id": 28229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28229,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28229, "pid": 5, "tid": 7, "ts": 1716454217688698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646097, "dur": 12,
    "args": {
      "External id": 28229,
      "cbid": 211, "correlation": 28229
    }
  },
  {
    "ph": "s", "id": 28229, "pid": 76337, "tid": -914061504, "ts": 1716454217646097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646202, "dur": 1,
    "args": {
      "External id": 28240,
      "cbid": 251, "correlation": 28240
    }
  },
  {
    "ph": "f", "id": 28240, "pid": 76337, "tid": -914061504, "ts": 1716454217646202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217688763, "dur": 153,
    "args": {
      "External id": 28241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28241,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28241, "pid": 5, "tid": 7, "ts": 1716454217688763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646208, "dur": 14,
    "args": {
      "External id": 28241,
      "cbid": 211, "correlation": 28241
    }
  },
  {
    "ph": "s", "id": 28241, "pid": 76337, "tid": -914061504, "ts": 1716454217646208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646291, "dur": 1,
    "args": {
      "External id": 28252,
      "cbid": 251, "correlation": 28252
    }
  },
  {
    "ph": "f", "id": 28252, "pid": 76337, "tid": -914061504, "ts": 1716454217646291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217688918, "dur": 146,
    "args": {
      "External id": 28253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28253,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28253, "pid": 5, "tid": 7, "ts": 1716454217688918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646295, "dur": 12,
    "args": {
      "External id": 28253,
      "cbid": 211, "correlation": 28253
    }
  },
  {
    "ph": "s", "id": 28253, "pid": 76337, "tid": -914061504, "ts": 1716454217646295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646373, "dur": 1,
    "args": {
      "External id": 28264,
      "cbid": 251, "correlation": 28264
    }
  },
  {
    "ph": "f", "id": 28264, "pid": 76337, "tid": -914061504, "ts": 1716454217646373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217689065, "dur": 142,
    "args": {
      "External id": 28265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28265,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28265, "pid": 5, "tid": 7, "ts": 1716454217689065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646377, "dur": 13,
    "args": {
      "External id": 28265,
      "cbid": 211, "correlation": 28265
    }
  },
  {
    "ph": "s", "id": 28265, "pid": 76337, "tid": -914061504, "ts": 1716454217646377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217689209, "dur": 1907,
    "args": {
      "External id": 28286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28286,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 28286, "pid": 5, "tid": 7, "ts": 1716454217689209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646464, "dur": 14,
    "args": {
      "External id": 28286,
      "cbid": 211, "correlation": 28286
    }
  },
  {
    "ph": "s", "id": 28286, "pid": 76337, "tid": -914061504, "ts": 1716454217646464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646583, "dur": 2,
    "args": {
      "External id": 28304,
      "cbid": 251, "correlation": 28304
    }
  },
  {
    "ph": "f", "id": 28304, "pid": 76337, "tid": -914061504, "ts": 1716454217646583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217691117, "dur": 145,
    "args": {
      "External id": 28306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28306,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 28306, "pid": 5, "tid": 7, "ts": 1716454217691117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646590, "dur": 14,
    "args": {
      "External id": 28306,
      "cbid": 211, "correlation": 28306
    }
  },
  {
    "ph": "s", "id": 28306, "pid": 76337, "tid": -914061504, "ts": 1716454217646590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217691264, "dur": 36,
    "args": {
      "External id": 28314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28314,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28314, "pid": 5, "tid": 7, "ts": 1716454217691264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646663, "dur": 12,
    "args": {
      "External id": 28314,
      "cbid": 211, "correlation": 28314
    }
  },
  {
    "ph": "s", "id": 28314, "pid": 76337, "tid": -914061504, "ts": 1716454217646663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217691301, "dur": 50,
    "args": {
      "External id": 28322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28322,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28322, "pid": 5, "tid": 7, "ts": 1716454217691301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646703, "dur": 10,
    "args": {
      "External id": 28322,
      "cbid": 211, "correlation": 28322
    }
  },
  {
    "ph": "s", "id": 28322, "pid": 76337, "tid": -914061504, "ts": 1716454217646703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217691352, "dur": 30,
    "args": {
      "External id": 28333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28333, "pid": 5, "tid": 7, "ts": 1716454217691352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646779, "dur": 12,
    "args": {
      "External id": 28333,
      "cbid": 211, "correlation": 28333
    }
  },
  {
    "ph": "s", "id": 28333, "pid": 76337, "tid": -914061504, "ts": 1716454217646779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217691384, "dur": 34,
    "args": {
      "External id": 28355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28355,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28355, "pid": 5, "tid": 7, "ts": 1716454217691384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646809, "dur": 7,
    "args": {
      "External id": 28355,
      "cbid": 211, "correlation": 28355
    }
  },
  {
    "ph": "s", "id": 28355, "pid": 76337, "tid": -914061504, "ts": 1716454217646809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646907, "dur": 1,
    "args": {
      "External id": 28366,
      "cbid": 251, "correlation": 28366
    }
  },
  {
    "ph": "f", "id": 28366, "pid": 76337, "tid": -914061504, "ts": 1716454217646907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217691419, "dur": 89,
    "args": {
      "External id": 28367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28367,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28367, "pid": 5, "tid": 7, "ts": 1716454217691419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217646912, "dur": 14,
    "args": {
      "External id": 28367,
      "cbid": 211, "correlation": 28367
    }
  },
  {
    "ph": "s", "id": 28367, "pid": 76337, "tid": -914061504, "ts": 1716454217646912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647006, "dur": 1,
    "args": {
      "External id": 28378,
      "cbid": 251, "correlation": 28378
    }
  },
  {
    "ph": "f", "id": 28378, "pid": 76337, "tid": -914061504, "ts": 1716454217647006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647011, "dur": 0,
    "args": {
      "External id": 28379,
      "cbid": 251, "correlation": 28379
    }
  },
  {
    "ph": "f", "id": 28379, "pid": 76337, "tid": -914061504, "ts": 1716454217647011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217691509, "dur": 11,
    "args": {
      "External id": 28380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28380,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 28380, "pid": 5, "tid": 7, "ts": 1716454217691509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647012, "dur": 14,
    "args": {
      "External id": 28380,
      "cbid": 211, "correlation": 28380
    }
  },
  {
    "ph": "s", "id": 28380, "pid": 76337, "tid": -914061504, "ts": 1716454217647012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217691521, "dur": 5,
    "args": {
      "External id": 28382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28382,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 28382, "pid": 5, "tid": 7, "ts": 1716454217691521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647028, "dur": 7,
    "args": {
      "External id": 28382,
      "cbid": 211, "correlation": 28382
    }
  },
  {
    "ph": "s", "id": 28382, "pid": 76337, "tid": -914061504, "ts": 1716454217647028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647101, "dur": 1,
    "args": {
      "External id": 28393,
      "cbid": 251, "correlation": 28393
    }
  },
  {
    "ph": "f", "id": 28393, "pid": 76337, "tid": -914061504, "ts": 1716454217647101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647105, "dur": 0,
    "args": {
      "External id": 28394,
      "cbid": 251, "correlation": 28394
    }
  },
  {
    "ph": "f", "id": 28394, "pid": 76337, "tid": -914061504, "ts": 1716454217647105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217691527, "dur": 7,
    "args": {
      "External id": 28395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28395,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 28395, "pid": 5, "tid": 7, "ts": 1716454217691527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647106, "dur": 13,
    "args": {
      "External id": 28395,
      "cbid": 211, "correlation": 28395
    }
  },
  {
    "ph": "s", "id": 28395, "pid": 76337, "tid": -914061504, "ts": 1716454217647106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217691536, "dur": 4,
    "args": {
      "External id": 28397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28397,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 28397, "pid": 5, "tid": 7, "ts": 1716454217691536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647120, "dur": 5,
    "args": {
      "External id": 28397,
      "cbid": 211, "correlation": 28397
    }
  },
  {
    "ph": "s", "id": 28397, "pid": 76337, "tid": -914061504, "ts": 1716454217647120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217691541, "dur": 90,
    "args": {
      "External id": 28418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28418,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 28418, "pid": 5, "tid": 7, "ts": 1716454217691541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647196, "dur": 12,
    "args": {
      "External id": 28418,
      "cbid": 211, "correlation": 28418
    }
  },
  {
    "ph": "s", "id": 28418, "pid": 76337, "tid": -914061504, "ts": 1716454217647196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647305, "dur": 1,
    "args": {
      "External id": 28436,
      "cbid": 251, "correlation": 28436
    }
  },
  {
    "ph": "f", "id": 28436, "pid": 76337, "tid": -914061504, "ts": 1716454217647305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217691632, "dur": 96,
    "args": {
      "External id": 28438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28438,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28438, "pid": 5, "tid": 7, "ts": 1716454217691632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647310, "dur": 14,
    "args": {
      "External id": 28438,
      "cbid": 211, "correlation": 28438
    }
  },
  {
    "ph": "s", "id": 28438, "pid": 76337, "tid": -914061504, "ts": 1716454217647310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217691730, "dur": 19,
    "args": {
      "External id": 28446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28446,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28446, "pid": 5, "tid": 7, "ts": 1716454217691730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647381, "dur": 13,
    "args": {
      "External id": 28446,
      "cbid": 211, "correlation": 28446
    }
  },
  {
    "ph": "s", "id": 28446, "pid": 76337, "tid": -914061504, "ts": 1716454217647381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217691750, "dur": 38,
    "args": {
      "External id": 28454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28454, "pid": 5, "tid": 7, "ts": 1716454217691750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647423, "dur": 9,
    "args": {
      "External id": 28454,
      "cbid": 211, "correlation": 28454
    }
  },
  {
    "ph": "s", "id": 28454, "pid": 76337, "tid": -914061504, "ts": 1716454217647423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217691789, "dur": 34,
    "args": {
      "External id": 28476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28476,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28476, "pid": 5, "tid": 7, "ts": 1716454217691789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647475, "dur": 10,
    "args": {
      "External id": 28476,
      "cbid": 211, "correlation": 28476
    }
  },
  {
    "ph": "s", "id": 28476, "pid": 76337, "tid": -914061504, "ts": 1716454217647475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647577, "dur": 1,
    "args": {
      "External id": 28492,
      "cbid": 251, "correlation": 28492
    }
  },
  {
    "ph": "f", "id": 28492, "pid": 76337, "tid": -914061504, "ts": 1716454217647577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647582, "dur": 0,
    "args": {
      "External id": 28494,
      "cbid": 251, "correlation": 28494
    }
  },
  {
    "ph": "f", "id": 28494, "pid": 76337, "tid": -914061504, "ts": 1716454217647582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217691825, "dur": 531,
    "args": {
      "External id": 28495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28495,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 28495, "pid": 5, "tid": 7, "ts": 1716454217691825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647587, "dur": 14,
    "args": {
      "External id": 28495,
      "cbid": 211, "correlation": 28495
    }
  },
  {
    "ph": "s", "id": 28495, "pid": 76337, "tid": -914061504, "ts": 1716454217647587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217692357, "dur": 123,
    "args": {
      "External id": 28503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28503,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28503, "pid": 5, "tid": 7, "ts": 1716454217692357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647657, "dur": 13,
    "args": {
      "External id": 28503,
      "cbid": 211, "correlation": 28503
    }
  },
  {
    "ph": "s", "id": 28503, "pid": 76337, "tid": -914061504, "ts": 1716454217647657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217692481, "dur": 127,
    "args": {
      "External id": 28511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28511, "pid": 5, "tid": 7, "ts": 1716454217692481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647690, "dur": 8,
    "args": {
      "External id": 28511,
      "cbid": 211, "correlation": 28511
    }
  },
  {
    "ph": "s", "id": 28511, "pid": 76337, "tid": -914061504, "ts": 1716454217647690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647780, "dur": 1,
    "args": {
      "External id": 28527,
      "cbid": 251, "correlation": 28527
    }
  },
  {
    "ph": "f", "id": 28527, "pid": 76337, "tid": -914061504, "ts": 1716454217647780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217692610, "dur": 303,
    "args": {
      "External id": 28529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28529,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28529, "pid": 5, "tid": 7, "ts": 1716454217692610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647786, "dur": 13,
    "args": {
      "External id": 28529,
      "cbid": 211, "correlation": 28529
    }
  },
  {
    "ph": "s", "id": 28529, "pid": 76337, "tid": -914061504, "ts": 1716454217647786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217692915, "dur": 28,
    "args": {
      "External id": 28537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28537,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28537, "pid": 5, "tid": 7, "ts": 1716454217692915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647830, "dur": 10,
    "args": {
      "External id": 28537,
      "cbid": 211, "correlation": 28537
    }
  },
  {
    "ph": "s", "id": 28537, "pid": 76337, "tid": -914061504, "ts": 1716454217647830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217692944, "dur": 80,
    "args": {
      "External id": 28548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28548, "pid": 5, "tid": 7, "ts": 1716454217692944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647899, "dur": 12,
    "args": {
      "External id": 28548,
      "cbid": 211, "correlation": 28548
    }
  },
  {
    "ph": "s", "id": 28548, "pid": 76337, "tid": -914061504, "ts": 1716454217647899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647966, "dur": 0,
    "args": {
      "External id": 28560,
      "cbid": 317, "correlation": 28560
    }
  },
  {
    "ph": "f", "id": 28560, "pid": 76337, "tid": -914061504, "ts": 1716454217647966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647967, "dur": 0,
    "args": {
      "External id": 28561,
      "cbid": 203, "correlation": 28561
    }
  },
  {
    "ph": "f", "id": 28561, "pid": 76337, "tid": -914061504, "ts": 1716454217647967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217647968, "dur": 0,
    "args": {
      "External id": 28562,
      "cbid": 205, "correlation": 28562
    }
  },
  {
    "ph": "f", "id": 28562, "pid": 76337, "tid": -914061504, "ts": 1716454217647968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217693024, "dur": 22,
    "args": {
      "External id": 28566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28566,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28566, "pid": 5, "tid": 7, "ts": 1716454217693024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648004, "dur": 14,
    "args": {
      "External id": 28566,
      "cbid": 211, "correlation": 28566
    }
  },
  {
    "ph": "s", "id": 28566, "pid": 76337, "tid": -914061504, "ts": 1716454217648004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217693048, "dur": 118,
    "args": {
      "External id": 28568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28568,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28568, "pid": 5, "tid": 7, "ts": 1716454217693048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648026, "dur": 7,
    "args": {
      "External id": 28568,
      "cbid": 211, "correlation": 28568
    }
  },
  {
    "ph": "s", "id": 28568, "pid": 76337, "tid": -914061504, "ts": 1716454217648026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217693167, "dur": 23,
    "args": {
      "External id": 28570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28570,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28570, "pid": 5, "tid": 7, "ts": 1716454217693167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648037, "dur": 5,
    "args": {
      "External id": 28570,
      "cbid": 211, "correlation": 28570
    }
  },
  {
    "ph": "s", "id": 28570, "pid": 76337, "tid": -914061504, "ts": 1716454217648037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217693192, "dur": 32,
    "args": {
      "External id": 28576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28576, "pid": 5, "tid": 7, "ts": 1716454217693192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648065, "dur": 9,
    "args": {
      "External id": 28576,
      "cbid": 211, "correlation": 28576
    }
  },
  {
    "ph": "s", "id": 28576, "pid": 76337, "tid": -914061504, "ts": 1716454217648065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217693225, "dur": 26,
    "args": {
      "External id": 28584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28584,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28584, "pid": 5, "tid": 7, "ts": 1716454217693225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648098, "dur": 8,
    "args": {
      "External id": 28584,
      "cbid": 211, "correlation": 28584
    }
  },
  {
    "ph": "s", "id": 28584, "pid": 76337, "tid": -914061504, "ts": 1716454217648098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217693253, "dur": 54,
    "args": {
      "External id": 28593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28593,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28593, "pid": 5, "tid": 7, "ts": 1716454217693253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648140, "dur": 10,
    "args": {
      "External id": 28593,
      "cbid": 211, "correlation": 28593
    }
  },
  {
    "ph": "s", "id": 28593, "pid": 76337, "tid": -914061504, "ts": 1716454217648140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217693308, "dur": 51,
    "args": {
      "External id": 28613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28613,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 28613, "pid": 5, "tid": 7, "ts": 1716454217693308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648214, "dur": 11,
    "args": {
      "External id": 28613,
      "cbid": 211, "correlation": 28613
    }
  },
  {
    "ph": "s", "id": 28613, "pid": 76337, "tid": -914061504, "ts": 1716454217648214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217693361, "dur": 4,
    "args": {
      "External id": 28625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28625,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 28625, "pid": 5, "tid": 7, "ts": 1716454217693361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648235, "dur": 7,
    "args": {
      "External id": 28625,
      "cbid": 211, "correlation": 28625
    }
  },
  {
    "ph": "s", "id": 28625, "pid": 76337, "tid": -914061504, "ts": 1716454217648235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217693366, "dur": 55,
    "args": {
      "External id": 28628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28628, "pid": 5, "tid": 7, "ts": 1716454217693366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648254, "dur": 7,
    "args": {
      "External id": 28628,
      "cbid": 211, "correlation": 28628
    }
  },
  {
    "ph": "s", "id": 28628, "pid": 76337, "tid": -914061504, "ts": 1716454217648254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217693423, "dur": 37,
    "args": {
      "External id": 28637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28637,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28637, "pid": 5, "tid": 7, "ts": 1716454217693423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648294, "dur": 10,
    "args": {
      "External id": 28637,
      "cbid": 211, "correlation": 28637
    }
  },
  {
    "ph": "s", "id": 28637, "pid": 76337, "tid": -914061504, "ts": 1716454217648294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648346, "dur": 0,
    "args": {
      "External id": 28647,
      "cbid": 317, "correlation": 28647
    }
  },
  {
    "ph": "f", "id": 28647, "pid": 76337, "tid": -914061504, "ts": 1716454217648346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648347, "dur": 0,
    "args": {
      "External id": 28648,
      "cbid": 203, "correlation": 28648
    }
  },
  {
    "ph": "f", "id": 28648, "pid": 76337, "tid": -914061504, "ts": 1716454217648347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648347, "dur": 0,
    "args": {
      "External id": 28649,
      "cbid": 205, "correlation": 28649
    }
  },
  {
    "ph": "f", "id": 28649, "pid": 76337, "tid": -914061504, "ts": 1716454217648347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217693461, "dur": 40,
    "args": {
      "External id": 28653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28653,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28653, "pid": 5, "tid": 7, "ts": 1716454217693461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648362, "dur": 11,
    "args": {
      "External id": 28653,
      "cbid": 211, "correlation": 28653
    }
  },
  {
    "ph": "s", "id": 28653, "pid": 76337, "tid": -914061504, "ts": 1716454217648362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217693502, "dur": 82,
    "args": {
      "External id": 28655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28655, "pid": 5, "tid": 7, "ts": 1716454217693502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648375, "dur": 5,
    "args": {
      "External id": 28655,
      "cbid": 211, "correlation": 28655
    }
  },
  {
    "ph": "s", "id": 28655, "pid": 76337, "tid": -914061504, "ts": 1716454217648375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217693585, "dur": 1264,
    "args": {
      "External id": 28657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28657,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28657, "pid": 5, "tid": 7, "ts": 1716454217693585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648387, "dur": 8,
    "args": {
      "External id": 28657,
      "cbid": 211, "correlation": 28657
    }
  },
  {
    "ph": "s", "id": 28657, "pid": 76337, "tid": -914061504, "ts": 1716454217648387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217694851, "dur": 22,
    "args": {
      "External id": 28659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28659, "pid": 5, "tid": 7, "ts": 1716454217694851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648398, "dur": 5,
    "args": {
      "External id": 28659,
      "cbid": 211, "correlation": 28659
    }
  },
  {
    "ph": "s", "id": 28659, "pid": 76337, "tid": -914061504, "ts": 1716454217648398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217694874, "dur": 34,
    "args": {
      "External id": 28665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28665, "pid": 5, "tid": 7, "ts": 1716454217694874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648426, "dur": 8,
    "args": {
      "External id": 28665,
      "cbid": 211, "correlation": 28665
    }
  },
  {
    "ph": "s", "id": 28665, "pid": 76337, "tid": -914061504, "ts": 1716454217648426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217694909, "dur": 3,
    "args": {
      "External id": 28673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28673,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 28673, "pid": 5, "tid": 7, "ts": 1716454217694909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648469, "dur": 9,
    "args": {
      "External id": 28673,
      "cbid": 211, "correlation": 28673
    }
  },
  {
    "ph": "s", "id": 28673, "pid": 76337, "tid": -914061504, "ts": 1716454217648469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648549, "dur": 1,
    "args": {
      "External id": 28689,
      "cbid": 251, "correlation": 28689
    }
  },
  {
    "ph": "f", "id": 28689, "pid": 76337, "tid": -914061504, "ts": 1716454217648549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648554, "dur": 0,
    "args": {
      "External id": 28691,
      "cbid": 251, "correlation": 28691
    }
  },
  {
    "ph": "f", "id": 28691, "pid": 76337, "tid": -914061504, "ts": 1716454217648554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217694914, "dur": 12,
    "args": {
      "External id": 28692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28692,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 28692, "pid": 5, "tid": 7, "ts": 1716454217694914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648556, "dur": 12,
    "args": {
      "External id": 28692,
      "cbid": 211, "correlation": 28692
    }
  },
  {
    "ph": "s", "id": 28692, "pid": 76337, "tid": -914061504, "ts": 1716454217648556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217694928, "dur": 5,
    "args": {
      "External id": 28694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28694,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 28694, "pid": 5, "tid": 7, "ts": 1716454217694928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648570, "dur": 5,
    "args": {
      "External id": 28694,
      "cbid": 211, "correlation": 28694
    }
  },
  {
    "ph": "s", "id": 28694, "pid": 76337, "tid": -914061504, "ts": 1716454217648570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217694934, "dur": 29,
    "args": {
      "External id": 28704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28704,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28704, "pid": 5, "tid": 7, "ts": 1716454217694934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648630, "dur": 12,
    "args": {
      "External id": 28704,
      "cbid": 211, "correlation": 28704
    }
  },
  {
    "ph": "s", "id": 28704, "pid": 76337, "tid": -914061504, "ts": 1716454217648630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217694964, "dur": 30,
    "args": {
      "External id": 28724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28724,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 28724, "pid": 5, "tid": 7, "ts": 1716454217694964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648696, "dur": 12,
    "args": {
      "External id": 28724,
      "cbid": 211, "correlation": 28724
    }
  },
  {
    "ph": "s", "id": 28724, "pid": 76337, "tid": -914061504, "ts": 1716454217648696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217694995, "dur": 5,
    "args": {
      "External id": 28736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28736,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 28736, "pid": 5, "tid": 7, "ts": 1716454217694995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648717, "dur": 6,
    "args": {
      "External id": 28736,
      "cbid": 211, "correlation": 28736
    }
  },
  {
    "ph": "s", "id": 28736, "pid": 76337, "tid": -914061504, "ts": 1716454217648717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217695001, "dur": 30,
    "args": {
      "External id": 28739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28739, "pid": 5, "tid": 7, "ts": 1716454217695001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648735, "dur": 6,
    "args": {
      "External id": 28739,
      "cbid": 211, "correlation": 28739
    }
  },
  {
    "ph": "s", "id": 28739, "pid": 76337, "tid": -914061504, "ts": 1716454217648735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217695032, "dur": 21,
    "args": {
      "External id": 28748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28748,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28748, "pid": 5, "tid": 7, "ts": 1716454217695032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648776, "dur": 10,
    "args": {
      "External id": 28748,
      "cbid": 211, "correlation": 28748
    }
  },
  {
    "ph": "s", "id": 28748, "pid": 76337, "tid": -914061504, "ts": 1716454217648776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648851, "dur": 0,
    "args": {
      "External id": 28758,
      "cbid": 317, "correlation": 28758
    }
  },
  {
    "ph": "f", "id": 28758, "pid": 76337, "tid": -914061504, "ts": 1716454217648851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648851, "dur": 0,
    "args": {
      "External id": 28759,
      "cbid": 203, "correlation": 28759
    }
  },
  {
    "ph": "f", "id": 28759, "pid": 76337, "tid": -914061504, "ts": 1716454217648851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648852, "dur": 0,
    "args": {
      "External id": 28760,
      "cbid": 205, "correlation": 28760
    }
  },
  {
    "ph": "f", "id": 28760, "pid": 76337, "tid": -914061504, "ts": 1716454217648852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217695055, "dur": 22,
    "args": {
      "External id": 28764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28764,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28764, "pid": 5, "tid": 7, "ts": 1716454217695055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648867, "dur": 13,
    "args": {
      "External id": 28764,
      "cbid": 211, "correlation": 28764
    }
  },
  {
    "ph": "s", "id": 28764, "pid": 76337, "tid": -914061504, "ts": 1716454217648867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217695078, "dur": 43,
    "args": {
      "External id": 28766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28766, "pid": 5, "tid": 7, "ts": 1716454217695078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648883, "dur": 5,
    "args": {
      "External id": 28766,
      "cbid": 211, "correlation": 28766
    }
  },
  {
    "ph": "s", "id": 28766, "pid": 76337, "tid": -914061504, "ts": 1716454217648883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217695122, "dur": 635,
    "args": {
      "External id": 28768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28768,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28768, "pid": 5, "tid": 7, "ts": 1716454217695122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648894, "dur": 6,
    "args": {
      "External id": 28768,
      "cbid": 211, "correlation": 28768
    }
  },
  {
    "ph": "s", "id": 28768, "pid": 76337, "tid": -914061504, "ts": 1716454217648894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217695758, "dur": 22,
    "args": {
      "External id": 28770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28770, "pid": 5, "tid": 7, "ts": 1716454217695758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648903, "dur": 5,
    "args": {
      "External id": 28770,
      "cbid": 211, "correlation": 28770
    }
  },
  {
    "ph": "s", "id": 28770, "pid": 76337, "tid": -914061504, "ts": 1716454217648903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217695782, "dur": 32,
    "args": {
      "External id": 28776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28776, "pid": 5, "tid": 7, "ts": 1716454217695782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648932, "dur": 8,
    "args": {
      "External id": 28776,
      "cbid": 211, "correlation": 28776
    }
  },
  {
    "ph": "s", "id": 28776, "pid": 76337, "tid": -914061504, "ts": 1716454217648932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217648999, "dur": 0,
    "args": {
      "External id": 28786,
      "cbid": 317, "correlation": 28786
    }
  },
  {
    "ph": "f", "id": 28786, "pid": 76337, "tid": -914061504, "ts": 1716454217648999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649000, "dur": 0,
    "args": {
      "External id": 28787,
      "cbid": 203, "correlation": 28787
    }
  },
  {
    "ph": "f", "id": 28787, "pid": 76337, "tid": -914061504, "ts": 1716454217649000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649000, "dur": 0,
    "args": {
      "External id": 28788,
      "cbid": 205, "correlation": 28788
    }
  },
  {
    "ph": "f", "id": 28788, "pid": 76337, "tid": -914061504, "ts": 1716454217649000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217695815, "dur": 37,
    "args": {
      "External id": 28792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28792,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28792, "pid": 5, "tid": 7, "ts": 1716454217695815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649013, "dur": 13,
    "args": {
      "External id": 28792,
      "cbid": 211, "correlation": 28792
    }
  },
  {
    "ph": "s", "id": 28792, "pid": 76337, "tid": -914061504, "ts": 1716454217649013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217695854, "dur": 187,
    "args": {
      "External id": 28794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28794,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28794, "pid": 5, "tid": 7, "ts": 1716454217695854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649031, "dur": 6,
    "args": {
      "External id": 28794,
      "cbid": 211, "correlation": 28794
    }
  },
  {
    "ph": "s", "id": 28794, "pid": 76337, "tid": -914061504, "ts": 1716454217649031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217696042, "dur": 22,
    "args": {
      "External id": 28796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28796, "pid": 5, "tid": 7, "ts": 1716454217696042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649041, "dur": 5,
    "args": {
      "External id": 28796,
      "cbid": 211, "correlation": 28796
    }
  },
  {
    "ph": "s", "id": 28796, "pid": 76337, "tid": -914061504, "ts": 1716454217649041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217696065, "dur": 32,
    "args": {
      "External id": 28802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28802,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28802, "pid": 5, "tid": 7, "ts": 1716454217696065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649067, "dur": 9,
    "args": {
      "External id": 28802,
      "cbid": 211, "correlation": 28802
    }
  },
  {
    "ph": "s", "id": 28802, "pid": 76337, "tid": -914061504, "ts": 1716454217649067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217696098, "dur": 27,
    "args": {
      "External id": 28810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28810,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28810, "pid": 5, "tid": 7, "ts": 1716454217696098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649096, "dur": 8,
    "args": {
      "External id": 28810,
      "cbid": 211, "correlation": 28810
    }
  },
  {
    "ph": "s", "id": 28810, "pid": 76337, "tid": -914061504, "ts": 1716454217649096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217696126, "dur": 20,
    "args": {
      "External id": 28818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28818,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28818, "pid": 5, "tid": 7, "ts": 1716454217696126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649126, "dur": 8,
    "args": {
      "External id": 28818,
      "cbid": 211, "correlation": 28818
    }
  },
  {
    "ph": "s", "id": 28818, "pid": 76337, "tid": -914061504, "ts": 1716454217649126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217696148, "dur": 30,
    "args": {
      "External id": 28838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28838,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 28838, "pid": 5, "tid": 7, "ts": 1716454217696148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649209, "dur": 12,
    "args": {
      "External id": 28838,
      "cbid": 211, "correlation": 28838
    }
  },
  {
    "ph": "s", "id": 28838, "pid": 76337, "tid": -914061504, "ts": 1716454217649209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217696179, "dur": 5,
    "args": {
      "External id": 28850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28850,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 28850, "pid": 5, "tid": 7, "ts": 1716454217696179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649231, "dur": 6,
    "args": {
      "External id": 28850,
      "cbid": 211, "correlation": 28850
    }
  },
  {
    "ph": "s", "id": 28850, "pid": 76337, "tid": -914061504, "ts": 1716454217649231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217696185, "dur": 30,
    "args": {
      "External id": 28853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28853, "pid": 5, "tid": 7, "ts": 1716454217696185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649249, "dur": 6,
    "args": {
      "External id": 28853,
      "cbid": 211, "correlation": 28853
    }
  },
  {
    "ph": "s", "id": 28853, "pid": 76337, "tid": -914061504, "ts": 1716454217649249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649317, "dur": 0,
    "args": {
      "External id": 28864,
      "cbid": 317, "correlation": 28864
    }
  },
  {
    "ph": "f", "id": 28864, "pid": 76337, "tid": -914061504, "ts": 1716454217649317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649318, "dur": 0,
    "args": {
      "External id": 28865,
      "cbid": 203, "correlation": 28865
    }
  },
  {
    "ph": "f", "id": 28865, "pid": 76337, "tid": -914061504, "ts": 1716454217649318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649319, "dur": 0,
    "args": {
      "External id": 28866,
      "cbid": 205, "correlation": 28866
    }
  },
  {
    "ph": "f", "id": 28866, "pid": 76337, "tid": -914061504, "ts": 1716454217649319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217696216, "dur": 23,
    "args": {
      "External id": 28870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28870, "pid": 5, "tid": 7, "ts": 1716454217696216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649332, "dur": 13,
    "args": {
      "External id": 28870,
      "cbid": 211, "correlation": 28870
    }
  },
  {
    "ph": "s", "id": 28870, "pid": 76337, "tid": -914061504, "ts": 1716454217649332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217696240, "dur": 103,
    "args": {
      "External id": 28872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28872,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28872, "pid": 5, "tid": 7, "ts": 1716454217696240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649351, "dur": 6,
    "args": {
      "External id": 28872,
      "cbid": 211, "correlation": 28872
    }
  },
  {
    "ph": "s", "id": 28872, "pid": 76337, "tid": -914061504, "ts": 1716454217649351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217696345, "dur": 21,
    "args": {
      "External id": 28874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28874,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28874, "pid": 5, "tid": 7, "ts": 1716454217696345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649361, "dur": 5,
    "args": {
      "External id": 28874,
      "cbid": 211, "correlation": 28874
    }
  },
  {
    "ph": "s", "id": 28874, "pid": 76337, "tid": -914061504, "ts": 1716454217649361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217696367, "dur": 32,
    "args": {
      "External id": 28880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28880, "pid": 5, "tid": 7, "ts": 1716454217696367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649389, "dur": 8,
    "args": {
      "External id": 28880,
      "cbid": 211, "correlation": 28880
    }
  },
  {
    "ph": "s", "id": 28880, "pid": 76337, "tid": -914061504, "ts": 1716454217649389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217696401, "dur": 193,
    "args": {
      "External id": 28889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28889,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28889, "pid": 5, "tid": 7, "ts": 1716454217696401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649472, "dur": 14,
    "args": {
      "External id": 28889,
      "cbid": 211, "correlation": 28889
    }
  },
  {
    "ph": "s", "id": 28889, "pid": 76337, "tid": -914061504, "ts": 1716454217649472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217696595, "dur": 64,
    "args": {
      "External id": 28911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28911,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28911, "pid": 5, "tid": 7, "ts": 1716454217696595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649528, "dur": 11,
    "args": {
      "External id": 28911,
      "cbid": 211, "correlation": 28911
    }
  },
  {
    "ph": "s", "id": 28911, "pid": 76337, "tid": -914061504, "ts": 1716454217649528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649629, "dur": 1,
    "args": {
      "External id": 28922,
      "cbid": 251, "correlation": 28922
    }
  },
  {
    "ph": "f", "id": 28922, "pid": 76337, "tid": -914061504, "ts": 1716454217649629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217696661, "dur": 152,
    "args": {
      "External id": 28923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28923,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28923, "pid": 5, "tid": 7, "ts": 1716454217696661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649635, "dur": 14,
    "args": {
      "External id": 28923,
      "cbid": 211, "correlation": 28923
    }
  },
  {
    "ph": "s", "id": 28923, "pid": 76337, "tid": -914061504, "ts": 1716454217649635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649718, "dur": 1,
    "args": {
      "External id": 28934,
      "cbid": 251, "correlation": 28934
    }
  },
  {
    "ph": "f", "id": 28934, "pid": 76337, "tid": -914061504, "ts": 1716454217649718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217696814, "dur": 148,
    "args": {
      "External id": 28935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28935,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28935, "pid": 5, "tid": 7, "ts": 1716454217696814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649723, "dur": 12,
    "args": {
      "External id": 28935,
      "cbid": 211, "correlation": 28935
    }
  },
  {
    "ph": "s", "id": 28935, "pid": 76337, "tid": -914061504, "ts": 1716454217649723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649808, "dur": 1,
    "args": {
      "External id": 28946,
      "cbid": 251, "correlation": 28946
    }
  },
  {
    "ph": "f", "id": 28946, "pid": 76337, "tid": -914061504, "ts": 1716454217649808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217696963, "dur": 145,
    "args": {
      "External id": 28947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28947,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 28947, "pid": 5, "tid": 7, "ts": 1716454217696963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649812, "dur": 12,
    "args": {
      "External id": 28947,
      "cbid": 211, "correlation": 28947
    }
  },
  {
    "ph": "s", "id": 28947, "pid": 76337, "tid": -914061504, "ts": 1716454217649812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217697110, "dur": 1908,
    "args": {
      "External id": 28968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28968,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 28968, "pid": 5, "tid": 7, "ts": 1716454217697110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217649895, "dur": 13,
    "args": {
      "External id": 28968,
      "cbid": 211, "correlation": 28968
    }
  },
  {
    "ph": "s", "id": 28968, "pid": 76337, "tid": -914061504, "ts": 1716454217649895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650015, "dur": 1,
    "args": {
      "External id": 28986,
      "cbid": 251, "correlation": 28986
    }
  },
  {
    "ph": "f", "id": 28986, "pid": 76337, "tid": -914061504, "ts": 1716454217650015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217699019, "dur": 145,
    "args": {
      "External id": 28988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28988,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 28988, "pid": 5, "tid": 7, "ts": 1716454217699019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650021, "dur": 14,
    "args": {
      "External id": 28988,
      "cbid": 211, "correlation": 28988
    }
  },
  {
    "ph": "s", "id": 28988, "pid": 76337, "tid": -914061504, "ts": 1716454217650021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217699166, "dur": 35,
    "args": {
      "External id": 28996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 28996,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 28996, "pid": 5, "tid": 7, "ts": 1716454217699166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650094, "dur": 12,
    "args": {
      "External id": 28996,
      "cbid": 211, "correlation": 28996
    }
  },
  {
    "ph": "s", "id": 28996, "pid": 76337, "tid": -914061504, "ts": 1716454217650094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217699202, "dur": 50,
    "args": {
      "External id": 29004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29004,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29004, "pid": 5, "tid": 7, "ts": 1716454217699202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650133, "dur": 9,
    "args": {
      "External id": 29004,
      "cbid": 211, "correlation": 29004
    }
  },
  {
    "ph": "s", "id": 29004, "pid": 76337, "tid": -914061504, "ts": 1716454217650133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217699253, "dur": 30,
    "args": {
      "External id": 29015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29015,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29015, "pid": 5, "tid": 7, "ts": 1716454217699253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650205, "dur": 12,
    "args": {
      "External id": 29015,
      "cbid": 211, "correlation": 29015
    }
  },
  {
    "ph": "s", "id": 29015, "pid": 76337, "tid": -914061504, "ts": 1716454217650205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217699285, "dur": 34,
    "args": {
      "External id": 29037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29037,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29037, "pid": 5, "tid": 7, "ts": 1716454217699285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650237, "dur": 8,
    "args": {
      "External id": 29037,
      "cbid": 211, "correlation": 29037
    }
  },
  {
    "ph": "s", "id": 29037, "pid": 76337, "tid": -914061504, "ts": 1716454217650237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650333, "dur": 1,
    "args": {
      "External id": 29048,
      "cbid": 251, "correlation": 29048
    }
  },
  {
    "ph": "f", "id": 29048, "pid": 76337, "tid": -914061504, "ts": 1716454217650333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217699320, "dur": 88,
    "args": {
      "External id": 29049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29049,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29049, "pid": 5, "tid": 7, "ts": 1716454217699320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650339, "dur": 13,
    "args": {
      "External id": 29049,
      "cbid": 211, "correlation": 29049
    }
  },
  {
    "ph": "s", "id": 29049, "pid": 76337, "tid": -914061504, "ts": 1716454217650339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650423, "dur": 1,
    "args": {
      "External id": 29060,
      "cbid": 251, "correlation": 29060
    }
  },
  {
    "ph": "f", "id": 29060, "pid": 76337, "tid": -914061504, "ts": 1716454217650423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650427, "dur": 0,
    "args": {
      "External id": 29061,
      "cbid": 251, "correlation": 29061
    }
  },
  {
    "ph": "f", "id": 29061, "pid": 76337, "tid": -914061504, "ts": 1716454217650427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217699410, "dur": 11,
    "args": {
      "External id": 29062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29062,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 29062, "pid": 5, "tid": 7, "ts": 1716454217699410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650429, "dur": 13,
    "args": {
      "External id": 29062,
      "cbid": 211, "correlation": 29062
    }
  },
  {
    "ph": "s", "id": 29062, "pid": 76337, "tid": -914061504, "ts": 1716454217650429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217699422, "dur": 5,
    "args": {
      "External id": 29064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29064,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 29064, "pid": 5, "tid": 7, "ts": 1716454217699422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650444, "dur": 6,
    "args": {
      "External id": 29064,
      "cbid": 211, "correlation": 29064
    }
  },
  {
    "ph": "s", "id": 29064, "pid": 76337, "tid": -914061504, "ts": 1716454217650444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650514, "dur": 1,
    "args": {
      "External id": 29075,
      "cbid": 251, "correlation": 29075
    }
  },
  {
    "ph": "f", "id": 29075, "pid": 76337, "tid": -914061504, "ts": 1716454217650514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650518, "dur": 0,
    "args": {
      "External id": 29076,
      "cbid": 251, "correlation": 29076
    }
  },
  {
    "ph": "f", "id": 29076, "pid": 76337, "tid": -914061504, "ts": 1716454217650518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217699428, "dur": 7,
    "args": {
      "External id": 29077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29077,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 29077, "pid": 5, "tid": 7, "ts": 1716454217699428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650520, "dur": 12,
    "args": {
      "External id": 29077,
      "cbid": 211, "correlation": 29077
    }
  },
  {
    "ph": "s", "id": 29077, "pid": 76337, "tid": -914061504, "ts": 1716454217650520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217699437, "dur": 3,
    "args": {
      "External id": 29079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29079,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 29079, "pid": 5, "tid": 7, "ts": 1716454217699437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650534, "dur": 6,
    "args": {
      "External id": 29079,
      "cbid": 211, "correlation": 29079
    }
  },
  {
    "ph": "s", "id": 29079, "pid": 76337, "tid": -914061504, "ts": 1716454217650534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217699441, "dur": 89,
    "args": {
      "External id": 29100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29100,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 29100, "pid": 5, "tid": 7, "ts": 1716454217699441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650609, "dur": 13,
    "args": {
      "External id": 29100,
      "cbid": 211, "correlation": 29100
    }
  },
  {
    "ph": "s", "id": 29100, "pid": 76337, "tid": -914061504, "ts": 1716454217650609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650717, "dur": 1,
    "args": {
      "External id": 29118,
      "cbid": 251, "correlation": 29118
    }
  },
  {
    "ph": "f", "id": 29118, "pid": 76337, "tid": -914061504, "ts": 1716454217650717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217699532, "dur": 98,
    "args": {
      "External id": 29120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29120,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29120, "pid": 5, "tid": 7, "ts": 1716454217699532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650722, "dur": 14,
    "args": {
      "External id": 29120,
      "cbid": 211, "correlation": 29120
    }
  },
  {
    "ph": "s", "id": 29120, "pid": 76337, "tid": -914061504, "ts": 1716454217650722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217699631, "dur": 19,
    "args": {
      "External id": 29128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29128,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29128, "pid": 5, "tid": 7, "ts": 1716454217699631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650793, "dur": 12,
    "args": {
      "External id": 29128,
      "cbid": 211, "correlation": 29128
    }
  },
  {
    "ph": "s", "id": 29128, "pid": 76337, "tid": -914061504, "ts": 1716454217650793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217699651, "dur": 37,
    "args": {
      "External id": 29136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29136, "pid": 5, "tid": 7, "ts": 1716454217699651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650834, "dur": 9,
    "args": {
      "External id": 29136,
      "cbid": 211, "correlation": 29136
    }
  },
  {
    "ph": "s", "id": 29136, "pid": 76337, "tid": -914061504, "ts": 1716454217650834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217699689, "dur": 34,
    "args": {
      "External id": 29158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29158,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29158, "pid": 5, "tid": 7, "ts": 1716454217699689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650885, "dur": 10,
    "args": {
      "External id": 29158,
      "cbid": 211, "correlation": 29158
    }
  },
  {
    "ph": "s", "id": 29158, "pid": 76337, "tid": -914061504, "ts": 1716454217650885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650994, "dur": 1,
    "args": {
      "External id": 29174,
      "cbid": 251, "correlation": 29174
    }
  },
  {
    "ph": "f", "id": 29174, "pid": 76337, "tid": -914061504, "ts": 1716454217650994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217650999, "dur": 0,
    "args": {
      "External id": 29176,
      "cbid": 251, "correlation": 29176
    }
  },
  {
    "ph": "f", "id": 29176, "pid": 76337, "tid": -914061504, "ts": 1716454217650999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217699724, "dur": 533,
    "args": {
      "External id": 29177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29177,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 29177, "pid": 5, "tid": 7, "ts": 1716454217699724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651003, "dur": 14,
    "args": {
      "External id": 29177,
      "cbid": 211, "correlation": 29177
    }
  },
  {
    "ph": "s", "id": 29177, "pid": 76337, "tid": -914061504, "ts": 1716454217651003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217700259, "dur": 124,
    "args": {
      "External id": 29185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29185, "pid": 5, "tid": 7, "ts": 1716454217700259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651072, "dur": 13,
    "args": {
      "External id": 29185,
      "cbid": 211, "correlation": 29185
    }
  },
  {
    "ph": "s", "id": 29185, "pid": 76337, "tid": -914061504, "ts": 1716454217651072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217700384, "dur": 126,
    "args": {
      "External id": 29193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29193, "pid": 5, "tid": 7, "ts": 1716454217700384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651104, "dur": 8,
    "args": {
      "External id": 29193,
      "cbid": 211, "correlation": 29193
    }
  },
  {
    "ph": "s", "id": 29193, "pid": 76337, "tid": -914061504, "ts": 1716454217651104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651195, "dur": 1,
    "args": {
      "External id": 29209,
      "cbid": 251, "correlation": 29209
    }
  },
  {
    "ph": "f", "id": 29209, "pid": 76337, "tid": -914061504, "ts": 1716454217651195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217700511, "dur": 306,
    "args": {
      "External id": 29211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29211,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29211, "pid": 5, "tid": 7, "ts": 1716454217700511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651201, "dur": 13,
    "args": {
      "External id": 29211,
      "cbid": 211, "correlation": 29211
    }
  },
  {
    "ph": "s", "id": 29211, "pid": 76337, "tid": -914061504, "ts": 1716454217651201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217700818, "dur": 27,
    "args": {
      "External id": 29219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29219,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29219, "pid": 5, "tid": 7, "ts": 1716454217700818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651245, "dur": 10,
    "args": {
      "External id": 29219,
      "cbid": 211, "correlation": 29219
    }
  },
  {
    "ph": "s", "id": 29219, "pid": 76337, "tid": -914061504, "ts": 1716454217651245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217700847, "dur": 80,
    "args": {
      "External id": 29230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29230, "pid": 5, "tid": 7, "ts": 1716454217700847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651312, "dur": 13,
    "args": {
      "External id": 29230,
      "cbid": 211, "correlation": 29230
    }
  },
  {
    "ph": "s", "id": 29230, "pid": 76337, "tid": -914061504, "ts": 1716454217651312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651376, "dur": 0,
    "args": {
      "External id": 29242,
      "cbid": 317, "correlation": 29242
    }
  },
  {
    "ph": "f", "id": 29242, "pid": 76337, "tid": -914061504, "ts": 1716454217651376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651377, "dur": 0,
    "args": {
      "External id": 29243,
      "cbid": 203, "correlation": 29243
    }
  },
  {
    "ph": "f", "id": 29243, "pid": 76337, "tid": -914061504, "ts": 1716454217651377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651378, "dur": 0,
    "args": {
      "External id": 29244,
      "cbid": 205, "correlation": 29244
    }
  },
  {
    "ph": "f", "id": 29244, "pid": 76337, "tid": -914061504, "ts": 1716454217651378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217700928, "dur": 24,
    "args": {
      "External id": 29248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29248,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29248, "pid": 5, "tid": 7, "ts": 1716454217700928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651394, "dur": 11,
    "args": {
      "External id": 29248,
      "cbid": 211, "correlation": 29248
    }
  },
  {
    "ph": "s", "id": 29248, "pid": 76337, "tid": -914061504, "ts": 1716454217651394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217700953, "dur": 119,
    "args": {
      "External id": 29250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29250,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29250, "pid": 5, "tid": 7, "ts": 1716454217700953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651412, "dur": 6,
    "args": {
      "External id": 29250,
      "cbid": 211, "correlation": 29250
    }
  },
  {
    "ph": "s", "id": 29250, "pid": 76337, "tid": -914061504, "ts": 1716454217651412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217701073, "dur": 22,
    "args": {
      "External id": 29252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29252,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29252, "pid": 5, "tid": 7, "ts": 1716454217701073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651422, "dur": 6,
    "args": {
      "External id": 29252,
      "cbid": 211, "correlation": 29252
    }
  },
  {
    "ph": "s", "id": 29252, "pid": 76337, "tid": -914061504, "ts": 1716454217651422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217701097, "dur": 32,
    "args": {
      "External id": 29258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29258, "pid": 5, "tid": 7, "ts": 1716454217701097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651450, "dur": 8,
    "args": {
      "External id": 29258,
      "cbid": 211, "correlation": 29258
    }
  },
  {
    "ph": "s", "id": 29258, "pid": 76337, "tid": -914061504, "ts": 1716454217651450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217701130, "dur": 26,
    "args": {
      "External id": 29266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29266,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29266, "pid": 5, "tid": 7, "ts": 1716454217701130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651482, "dur": 8,
    "args": {
      "External id": 29266,
      "cbid": 211, "correlation": 29266
    }
  },
  {
    "ph": "s", "id": 29266, "pid": 76337, "tid": -914061504, "ts": 1716454217651482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217701158, "dur": 47,
    "args": {
      "External id": 29275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29275,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29275, "pid": 5, "tid": 7, "ts": 1716454217701158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651520, "dur": 10,
    "args": {
      "External id": 29275,
      "cbid": 211, "correlation": 29275
    }
  },
  {
    "ph": "s", "id": 29275, "pid": 76337, "tid": -914061504, "ts": 1716454217651520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217701206, "dur": 43,
    "args": {
      "External id": 29295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29295,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 29295, "pid": 5, "tid": 7, "ts": 1716454217701206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651592, "dur": 12,
    "args": {
      "External id": 29295,
      "cbid": 211, "correlation": 29295
    }
  },
  {
    "ph": "s", "id": 29295, "pid": 76337, "tid": -914061504, "ts": 1716454217651592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217701250, "dur": 5,
    "args": {
      "External id": 29307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29307,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 29307, "pid": 5, "tid": 7, "ts": 1716454217701250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651614, "dur": 6,
    "args": {
      "External id": 29307,
      "cbid": 211, "correlation": 29307
    }
  },
  {
    "ph": "s", "id": 29307, "pid": 76337, "tid": -914061504, "ts": 1716454217651614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217701256, "dur": 44,
    "args": {
      "External id": 29310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29310,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29310, "pid": 5, "tid": 7, "ts": 1716454217701256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651632, "dur": 7,
    "args": {
      "External id": 29310,
      "cbid": 211, "correlation": 29310
    }
  },
  {
    "ph": "s", "id": 29310, "pid": 76337, "tid": -914061504, "ts": 1716454217651632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217701302, "dur": 29,
    "args": {
      "External id": 29319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29319,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29319, "pid": 5, "tid": 7, "ts": 1716454217701302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651672, "dur": 9,
    "args": {
      "External id": 29319,
      "cbid": 211, "correlation": 29319
    }
  },
  {
    "ph": "s", "id": 29319, "pid": 76337, "tid": -914061504, "ts": 1716454217651672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651724, "dur": 0,
    "args": {
      "External id": 29329,
      "cbid": 317, "correlation": 29329
    }
  },
  {
    "ph": "f", "id": 29329, "pid": 76337, "tid": -914061504, "ts": 1716454217651724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651725, "dur": 0,
    "args": {
      "External id": 29330,
      "cbid": 203, "correlation": 29330
    }
  },
  {
    "ph": "f", "id": 29330, "pid": 76337, "tid": -914061504, "ts": 1716454217651725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651725, "dur": 0,
    "args": {
      "External id": 29331,
      "cbid": 205, "correlation": 29331
    }
  },
  {
    "ph": "f", "id": 29331, "pid": 76337, "tid": -914061504, "ts": 1716454217651725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217701332, "dur": 30,
    "args": {
      "External id": 29335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29335,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29335, "pid": 5, "tid": 7, "ts": 1716454217701332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651740, "dur": 11,
    "args": {
      "External id": 29335,
      "cbid": 211, "correlation": 29335
    }
  },
  {
    "ph": "s", "id": 29335, "pid": 76337, "tid": -914061504, "ts": 1716454217651740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217701364, "dur": 62,
    "args": {
      "External id": 29337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29337, "pid": 5, "tid": 7, "ts": 1716454217701364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651754, "dur": 6,
    "args": {
      "External id": 29337,
      "cbid": 211, "correlation": 29337
    }
  },
  {
    "ph": "s", "id": 29337, "pid": 76337, "tid": -914061504, "ts": 1716454217651754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217701427, "dur": 955,
    "args": {
      "External id": 29339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29339,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29339, "pid": 5, "tid": 7, "ts": 1716454217701427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651766, "dur": 6,
    "args": {
      "External id": 29339,
      "cbid": 211, "correlation": 29339
    }
  },
  {
    "ph": "s", "id": 29339, "pid": 76337, "tid": -914061504, "ts": 1716454217651766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217702383, "dur": 21,
    "args": {
      "External id": 29341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29341,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29341, "pid": 5, "tid": 7, "ts": 1716454217702383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651776, "dur": 5,
    "args": {
      "External id": 29341,
      "cbid": 211, "correlation": 29341
    }
  },
  {
    "ph": "s", "id": 29341, "pid": 76337, "tid": -914061504, "ts": 1716454217651776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217702405, "dur": 32,
    "args": {
      "External id": 29347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29347, "pid": 5, "tid": 7, "ts": 1716454217702405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651803, "dur": 9,
    "args": {
      "External id": 29347,
      "cbid": 211, "correlation": 29347
    }
  },
  {
    "ph": "s", "id": 29347, "pid": 76337, "tid": -914061504, "ts": 1716454217651803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217702439, "dur": 3,
    "args": {
      "External id": 29355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29355,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 29355, "pid": 5, "tid": 7, "ts": 1716454217702439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651847, "dur": 9,
    "args": {
      "External id": 29355,
      "cbid": 211, "correlation": 29355
    }
  },
  {
    "ph": "s", "id": 29355, "pid": 76337, "tid": -914061504, "ts": 1716454217651847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651926, "dur": 1,
    "args": {
      "External id": 29371,
      "cbid": 251, "correlation": 29371
    }
  },
  {
    "ph": "f", "id": 29371, "pid": 76337, "tid": -914061504, "ts": 1716454217651926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651931, "dur": 0,
    "args": {
      "External id": 29373,
      "cbid": 251, "correlation": 29373
    }
  },
  {
    "ph": "f", "id": 29373, "pid": 76337, "tid": -914061504, "ts": 1716454217651931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217702443, "dur": 12,
    "args": {
      "External id": 29374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29374,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 29374, "pid": 5, "tid": 7, "ts": 1716454217702443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651933, "dur": 13,
    "args": {
      "External id": 29374,
      "cbid": 211, "correlation": 29374
    }
  },
  {
    "ph": "s", "id": 29374, "pid": 76337, "tid": -914061504, "ts": 1716454217651933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217702457, "dur": 5,
    "args": {
      "External id": 29376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29376,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 29376, "pid": 5, "tid": 7, "ts": 1716454217702457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217651948, "dur": 6,
    "args": {
      "External id": 29376,
      "cbid": 211, "correlation": 29376
    }
  },
  {
    "ph": "s", "id": 29376, "pid": 76337, "tid": -914061504, "ts": 1716454217651948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217702463, "dur": 28,
    "args": {
      "External id": 29386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29386, "pid": 5, "tid": 7, "ts": 1716454217702463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652017, "dur": 13,
    "args": {
      "External id": 29386,
      "cbid": 211, "correlation": 29386
    }
  },
  {
    "ph": "s", "id": 29386, "pid": 76337, "tid": -914061504, "ts": 1716454217652017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217702492, "dur": 31,
    "args": {
      "External id": 29406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29406,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 29406, "pid": 5, "tid": 7, "ts": 1716454217702492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652084, "dur": 11,
    "args": {
      "External id": 29406,
      "cbid": 211, "correlation": 29406
    }
  },
  {
    "ph": "s", "id": 29406, "pid": 76337, "tid": -914061504, "ts": 1716454217652084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217702524, "dur": 4,
    "args": {
      "External id": 29418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29418,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 29418, "pid": 5, "tid": 7, "ts": 1716454217702524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652105, "dur": 7,
    "args": {
      "External id": 29418,
      "cbid": 211, "correlation": 29418
    }
  },
  {
    "ph": "s", "id": 29418, "pid": 76337, "tid": -914061504, "ts": 1716454217652105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217702530, "dur": 30,
    "args": {
      "External id": 29421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29421, "pid": 5, "tid": 7, "ts": 1716454217702530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652124, "dur": 6,
    "args": {
      "External id": 29421,
      "cbid": 211, "correlation": 29421
    }
  },
  {
    "ph": "s", "id": 29421, "pid": 76337, "tid": -914061504, "ts": 1716454217652124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217702561, "dur": 21,
    "args": {
      "External id": 29430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29430,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29430, "pid": 5, "tid": 7, "ts": 1716454217702561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652164, "dur": 10,
    "args": {
      "External id": 29430,
      "cbid": 211, "correlation": 29430
    }
  },
  {
    "ph": "s", "id": 29430, "pid": 76337, "tid": -914061504, "ts": 1716454217652164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652241, "dur": 0,
    "args": {
      "External id": 29440,
      "cbid": 317, "correlation": 29440
    }
  },
  {
    "ph": "f", "id": 29440, "pid": 76337, "tid": -914061504, "ts": 1716454217652241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652241, "dur": 0,
    "args": {
      "External id": 29441,
      "cbid": 203, "correlation": 29441
    }
  },
  {
    "ph": "f", "id": 29441, "pid": 76337, "tid": -914061504, "ts": 1716454217652241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652242, "dur": 0,
    "args": {
      "External id": 29442,
      "cbid": 205, "correlation": 29442
    }
  },
  {
    "ph": "f", "id": 29442, "pid": 76337, "tid": -914061504, "ts": 1716454217652242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217702583, "dur": 22,
    "args": {
      "External id": 29446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29446, "pid": 5, "tid": 7, "ts": 1716454217702583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652256, "dur": 13,
    "args": {
      "External id": 29446,
      "cbid": 211, "correlation": 29446
    }
  },
  {
    "ph": "s", "id": 29446, "pid": 76337, "tid": -914061504, "ts": 1716454217652256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217702606, "dur": 43,
    "args": {
      "External id": 29448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29448,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29448, "pid": 5, "tid": 7, "ts": 1716454217702606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652272, "dur": 5,
    "args": {
      "External id": 29448,
      "cbid": 211, "correlation": 29448
    }
  },
  {
    "ph": "s", "id": 29448, "pid": 76337, "tid": -914061504, "ts": 1716454217652272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217702651, "dur": 635,
    "args": {
      "External id": 29450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29450,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29450, "pid": 5, "tid": 7, "ts": 1716454217702651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652284, "dur": 6,
    "args": {
      "External id": 29450,
      "cbid": 211, "correlation": 29450
    }
  },
  {
    "ph": "s", "id": 29450, "pid": 76337, "tid": -914061504, "ts": 1716454217652284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217703287, "dur": 22,
    "args": {
      "External id": 29452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29452, "pid": 5, "tid": 7, "ts": 1716454217703287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652293, "dur": 5,
    "args": {
      "External id": 29452,
      "cbid": 211, "correlation": 29452
    }
  },
  {
    "ph": "s", "id": 29452, "pid": 76337, "tid": -914061504, "ts": 1716454217652293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217703311, "dur": 33,
    "args": {
      "External id": 29458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29458, "pid": 5, "tid": 7, "ts": 1716454217703311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652322, "dur": 9,
    "args": {
      "External id": 29458,
      "cbid": 211, "correlation": 29458
    }
  },
  {
    "ph": "s", "id": 29458, "pid": 76337, "tid": -914061504, "ts": 1716454217652322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652381, "dur": 0,
    "args": {
      "External id": 29468,
      "cbid": 317, "correlation": 29468
    }
  },
  {
    "ph": "f", "id": 29468, "pid": 76337, "tid": -914061504, "ts": 1716454217652381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652381, "dur": 0,
    "args": {
      "External id": 29469,
      "cbid": 203, "correlation": 29469
    }
  },
  {
    "ph": "f", "id": 29469, "pid": 76337, "tid": -914061504, "ts": 1716454217652381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652382, "dur": 0,
    "args": {
      "External id": 29470,
      "cbid": 205, "correlation": 29470
    }
  },
  {
    "ph": "f", "id": 29470, "pid": 76337, "tid": -914061504, "ts": 1716454217652382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217703345, "dur": 31,
    "args": {
      "External id": 29474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29474,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29474, "pid": 5, "tid": 7, "ts": 1716454217703345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652396, "dur": 11,
    "args": {
      "External id": 29474,
      "cbid": 211, "correlation": 29474
    }
  },
  {
    "ph": "s", "id": 29474, "pid": 76337, "tid": -914061504, "ts": 1716454217652396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217703377, "dur": 149,
    "args": {
      "External id": 29476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29476,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29476, "pid": 5, "tid": 7, "ts": 1716454217703377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652413, "dur": 6,
    "args": {
      "External id": 29476,
      "cbid": 211, "correlation": 29476
    }
  },
  {
    "ph": "s", "id": 29476, "pid": 76337, "tid": -914061504, "ts": 1716454217652413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217703528, "dur": 22,
    "args": {
      "External id": 29478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29478, "pid": 5, "tid": 7, "ts": 1716454217703528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652424, "dur": 5,
    "args": {
      "External id": 29478,
      "cbid": 211, "correlation": 29478
    }
  },
  {
    "ph": "s", "id": 29478, "pid": 76337, "tid": -914061504, "ts": 1716454217652424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217703551, "dur": 32,
    "args": {
      "External id": 29484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29484, "pid": 5, "tid": 7, "ts": 1716454217703551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652450, "dur": 8,
    "args": {
      "External id": 29484,
      "cbid": 211, "correlation": 29484
    }
  },
  {
    "ph": "s", "id": 29484, "pid": 76337, "tid": -914061504, "ts": 1716454217652450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217703584, "dur": 27,
    "args": {
      "External id": 29492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29492,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29492, "pid": 5, "tid": 7, "ts": 1716454217703584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652479, "dur": 8,
    "args": {
      "External id": 29492,
      "cbid": 211, "correlation": 29492
    }
  },
  {
    "ph": "s", "id": 29492, "pid": 76337, "tid": -914061504, "ts": 1716454217652479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217703612, "dur": 20,
    "args": {
      "External id": 29500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29500,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29500, "pid": 5, "tid": 7, "ts": 1716454217703612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652507, "dur": 8,
    "args": {
      "External id": 29500,
      "cbid": 211, "correlation": 29500
    }
  },
  {
    "ph": "s", "id": 29500, "pid": 76337, "tid": -914061504, "ts": 1716454217652507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217703633, "dur": 30,
    "args": {
      "External id": 29520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29520,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 29520, "pid": 5, "tid": 7, "ts": 1716454217703633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652591, "dur": 12,
    "args": {
      "External id": 29520,
      "cbid": 211, "correlation": 29520
    }
  },
  {
    "ph": "s", "id": 29520, "pid": 76337, "tid": -914061504, "ts": 1716454217652591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217703665, "dur": 5,
    "args": {
      "External id": 29532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29532,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 29532, "pid": 5, "tid": 7, "ts": 1716454217703665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652613, "dur": 6,
    "args": {
      "External id": 29532,
      "cbid": 211, "correlation": 29532
    }
  },
  {
    "ph": "s", "id": 29532, "pid": 76337, "tid": -914061504, "ts": 1716454217652613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217703671, "dur": 31,
    "args": {
      "External id": 29535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29535, "pid": 5, "tid": 7, "ts": 1716454217703671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652631, "dur": 7,
    "args": {
      "External id": 29535,
      "cbid": 211, "correlation": 29535
    }
  },
  {
    "ph": "s", "id": 29535, "pid": 76337, "tid": -914061504, "ts": 1716454217652631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652701, "dur": 0,
    "args": {
      "External id": 29546,
      "cbid": 317, "correlation": 29546
    }
  },
  {
    "ph": "f", "id": 29546, "pid": 76337, "tid": -914061504, "ts": 1716454217652701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652702, "dur": 0,
    "args": {
      "External id": 29547,
      "cbid": 203, "correlation": 29547
    }
  },
  {
    "ph": "f", "id": 29547, "pid": 76337, "tid": -914061504, "ts": 1716454217652702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652702, "dur": 0,
    "args": {
      "External id": 29548,
      "cbid": 205, "correlation": 29548
    }
  },
  {
    "ph": "f", "id": 29548, "pid": 76337, "tid": -914061504, "ts": 1716454217652702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217703703, "dur": 23,
    "args": {
      "External id": 29552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29552,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29552, "pid": 5, "tid": 7, "ts": 1716454217703703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652716, "dur": 12,
    "args": {
      "External id": 29552,
      "cbid": 211, "correlation": 29552
    }
  },
  {
    "ph": "s", "id": 29552, "pid": 76337, "tid": -914061504, "ts": 1716454217652716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217703727, "dur": 103,
    "args": {
      "External id": 29554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29554,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29554, "pid": 5, "tid": 7, "ts": 1716454217703727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652734, "dur": 6,
    "args": {
      "External id": 29554,
      "cbid": 211, "correlation": 29554
    }
  },
  {
    "ph": "s", "id": 29554, "pid": 76337, "tid": -914061504, "ts": 1716454217652734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217703832, "dur": 21,
    "args": {
      "External id": 29556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29556,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29556, "pid": 5, "tid": 7, "ts": 1716454217703832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652744, "dur": 5,
    "args": {
      "External id": 29556,
      "cbid": 211, "correlation": 29556
    }
  },
  {
    "ph": "s", "id": 29556, "pid": 76337, "tid": -914061504, "ts": 1716454217652744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217703854, "dur": 32,
    "args": {
      "External id": 29562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29562,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29562, "pid": 5, "tid": 7, "ts": 1716454217703854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652772, "dur": 8,
    "args": {
      "External id": 29562,
      "cbid": 211, "correlation": 29562
    }
  },
  {
    "ph": "s", "id": 29562, "pid": 76337, "tid": -914061504, "ts": 1716454217652772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217703887, "dur": 195,
    "args": {
      "External id": 29571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29571,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29571, "pid": 5, "tid": 7, "ts": 1716454217703887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652855, "dur": 14,
    "args": {
      "External id": 29571,
      "cbid": 211, "correlation": 29571
    }
  },
  {
    "ph": "s", "id": 29571, "pid": 76337, "tid": -914061504, "ts": 1716454217652855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217704084, "dur": 64,
    "args": {
      "External id": 29593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29593,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29593, "pid": 5, "tid": 7, "ts": 1716454217704084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217652912, "dur": 10,
    "args": {
      "External id": 29593,
      "cbid": 211, "correlation": 29593
    }
  },
  {
    "ph": "s", "id": 29593, "pid": 76337, "tid": -914061504, "ts": 1716454217652912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653022, "dur": 1,
    "args": {
      "External id": 29604,
      "cbid": 251, "correlation": 29604
    }
  },
  {
    "ph": "f", "id": 29604, "pid": 76337, "tid": -914061504, "ts": 1716454217653022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217704150, "dur": 150,
    "args": {
      "External id": 29605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29605,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29605, "pid": 5, "tid": 7, "ts": 1716454217704150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653028, "dur": 14,
    "args": {
      "External id": 29605,
      "cbid": 211, "correlation": 29605
    }
  },
  {
    "ph": "s", "id": 29605, "pid": 76337, "tid": -914061504, "ts": 1716454217653028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653114, "dur": 1,
    "args": {
      "External id": 29616,
      "cbid": 251, "correlation": 29616
    }
  },
  {
    "ph": "f", "id": 29616, "pid": 76337, "tid": -914061504, "ts": 1716454217653114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217704301, "dur": 142,
    "args": {
      "External id": 29617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29617,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29617, "pid": 5, "tid": 7, "ts": 1716454217704301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653118, "dur": 12,
    "args": {
      "External id": 29617,
      "cbid": 211, "correlation": 29617
    }
  },
  {
    "ph": "s", "id": 29617, "pid": 76337, "tid": -914061504, "ts": 1716454217653118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653196, "dur": 1,
    "args": {
      "External id": 29628,
      "cbid": 251, "correlation": 29628
    }
  },
  {
    "ph": "f", "id": 29628, "pid": 76337, "tid": -914061504, "ts": 1716454217653196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217704445, "dur": 143,
    "args": {
      "External id": 29629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29629,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29629, "pid": 5, "tid": 7, "ts": 1716454217704445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653201, "dur": 12,
    "args": {
      "External id": 29629,
      "cbid": 211, "correlation": 29629
    }
  },
  {
    "ph": "s", "id": 29629, "pid": 76337, "tid": -914061504, "ts": 1716454217653201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217704589, "dur": 1906,
    "args": {
      "External id": 29650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29650,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 29650, "pid": 5, "tid": 7, "ts": 1716454217704589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653285, "dur": 13,
    "args": {
      "External id": 29650,
      "cbid": 211, "correlation": 29650
    }
  },
  {
    "ph": "s", "id": 29650, "pid": 76337, "tid": -914061504, "ts": 1716454217653285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653397, "dur": 1,
    "args": {
      "External id": 29668,
      "cbid": 251, "correlation": 29668
    }
  },
  {
    "ph": "f", "id": 29668, "pid": 76337, "tid": -914061504, "ts": 1716454217653397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217706496, "dur": 145,
    "args": {
      "External id": 29670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29670,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 29670, "pid": 5, "tid": 7, "ts": 1716454217706496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653403, "dur": 15,
    "args": {
      "External id": 29670,
      "cbid": 211, "correlation": 29670
    }
  },
  {
    "ph": "s", "id": 29670, "pid": 76337, "tid": -914061504, "ts": 1716454217653403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217706642, "dur": 36,
    "args": {
      "External id": 29678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29678,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29678, "pid": 5, "tid": 7, "ts": 1716454217706642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653476, "dur": 12,
    "args": {
      "External id": 29678,
      "cbid": 211, "correlation": 29678
    }
  },
  {
    "ph": "s", "id": 29678, "pid": 76337, "tid": -914061504, "ts": 1716454217653476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217706679, "dur": 51,
    "args": {
      "External id": 29686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29686,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29686, "pid": 5, "tid": 7, "ts": 1716454217706679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653515, "dur": 9,
    "args": {
      "External id": 29686,
      "cbid": 211, "correlation": 29686
    }
  },
  {
    "ph": "s", "id": 29686, "pid": 76337, "tid": -914061504, "ts": 1716454217653515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217706731, "dur": 31,
    "args": {
      "External id": 29697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29697,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29697, "pid": 5, "tid": 7, "ts": 1716454217706731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653601, "dur": 14,
    "args": {
      "External id": 29697,
      "cbid": 211, "correlation": 29697
    }
  },
  {
    "ph": "s", "id": 29697, "pid": 76337, "tid": -914061504, "ts": 1716454217653601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217706763, "dur": 34,
    "args": {
      "External id": 29719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29719,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29719, "pid": 5, "tid": 7, "ts": 1716454217706763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653634, "dur": 8,
    "args": {
      "External id": 29719,
      "cbid": 211, "correlation": 29719
    }
  },
  {
    "ph": "s", "id": 29719, "pid": 76337, "tid": -914061504, "ts": 1716454217653634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653730, "dur": 1,
    "args": {
      "External id": 29730,
      "cbid": 251, "correlation": 29730
    }
  },
  {
    "ph": "f", "id": 29730, "pid": 76337, "tid": -914061504, "ts": 1716454217653730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217706799, "dur": 88,
    "args": {
      "External id": 29731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29731,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29731, "pid": 5, "tid": 7, "ts": 1716454217706799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653736, "dur": 14,
    "args": {
      "External id": 29731,
      "cbid": 211, "correlation": 29731
    }
  },
  {
    "ph": "s", "id": 29731, "pid": 76337, "tid": -914061504, "ts": 1716454217653736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653818, "dur": 1,
    "args": {
      "External id": 29742,
      "cbid": 251, "correlation": 29742
    }
  },
  {
    "ph": "f", "id": 29742, "pid": 76337, "tid": -914061504, "ts": 1716454217653818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653823, "dur": 0,
    "args": {
      "External id": 29743,
      "cbid": 251, "correlation": 29743
    }
  },
  {
    "ph": "f", "id": 29743, "pid": 76337, "tid": -914061504, "ts": 1716454217653823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217706888, "dur": 11,
    "args": {
      "External id": 29744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29744,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 29744, "pid": 5, "tid": 7, "ts": 1716454217706888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653824, "dur": 13,
    "args": {
      "External id": 29744,
      "cbid": 211, "correlation": 29744
    }
  },
  {
    "ph": "s", "id": 29744, "pid": 76337, "tid": -914061504, "ts": 1716454217653824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217706900, "dur": 5,
    "args": {
      "External id": 29746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29746,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 29746, "pid": 5, "tid": 7, "ts": 1716454217706900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653839, "dur": 6,
    "args": {
      "External id": 29746,
      "cbid": 211, "correlation": 29746
    }
  },
  {
    "ph": "s", "id": 29746, "pid": 76337, "tid": -914061504, "ts": 1716454217653839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653910, "dur": 1,
    "args": {
      "External id": 29757,
      "cbid": 251, "correlation": 29757
    }
  },
  {
    "ph": "f", "id": 29757, "pid": 76337, "tid": -914061504, "ts": 1716454217653910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653914, "dur": 0,
    "args": {
      "External id": 29758,
      "cbid": 251, "correlation": 29758
    }
  },
  {
    "ph": "f", "id": 29758, "pid": 76337, "tid": -914061504, "ts": 1716454217653914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217706907, "dur": 7,
    "args": {
      "External id": 29759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29759,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 29759, "pid": 5, "tid": 7, "ts": 1716454217706907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653916, "dur": 13,
    "args": {
      "External id": 29759,
      "cbid": 211, "correlation": 29759
    }
  },
  {
    "ph": "s", "id": 29759, "pid": 76337, "tid": -914061504, "ts": 1716454217653916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217706915, "dur": 4,
    "args": {
      "External id": 29761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29761,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 29761, "pid": 5, "tid": 7, "ts": 1716454217706915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217653930, "dur": 6,
    "args": {
      "External id": 29761,
      "cbid": 211, "correlation": 29761
    }
  },
  {
    "ph": "s", "id": 29761, "pid": 76337, "tid": -914061504, "ts": 1716454217653930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217706920, "dur": 90,
    "args": {
      "External id": 29782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29782,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 29782, "pid": 5, "tid": 7, "ts": 1716454217706920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654014, "dur": 13,
    "args": {
      "External id": 29782,
      "cbid": 211, "correlation": 29782
    }
  },
  {
    "ph": "s", "id": 29782, "pid": 76337, "tid": -914061504, "ts": 1716454217654014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654124, "dur": 1,
    "args": {
      "External id": 29800,
      "cbid": 251, "correlation": 29800
    }
  },
  {
    "ph": "f", "id": 29800, "pid": 76337, "tid": -914061504, "ts": 1716454217654124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217707012, "dur": 98,
    "args": {
      "External id": 29802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29802,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29802, "pid": 5, "tid": 7, "ts": 1716454217707012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654130, "dur": 13,
    "args": {
      "External id": 29802,
      "cbid": 211, "correlation": 29802
    }
  },
  {
    "ph": "s", "id": 29802, "pid": 76337, "tid": -914061504, "ts": 1716454217654130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217707111, "dur": 19,
    "args": {
      "External id": 29810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29810,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29810, "pid": 5, "tid": 7, "ts": 1716454217707111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654202, "dur": 12,
    "args": {
      "External id": 29810,
      "cbid": 211, "correlation": 29810
    }
  },
  {
    "ph": "s", "id": 29810, "pid": 76337, "tid": -914061504, "ts": 1716454217654202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217707132, "dur": 38,
    "args": {
      "External id": 29818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29818, "pid": 5, "tid": 7, "ts": 1716454217707132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654242, "dur": 10,
    "args": {
      "External id": 29818,
      "cbid": 211, "correlation": 29818
    }
  },
  {
    "ph": "s", "id": 29818, "pid": 76337, "tid": -914061504, "ts": 1716454217654242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217707171, "dur": 34,
    "args": {
      "External id": 29840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29840,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29840, "pid": 5, "tid": 7, "ts": 1716454217707171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654293, "dur": 11,
    "args": {
      "External id": 29840,
      "cbid": 211, "correlation": 29840
    }
  },
  {
    "ph": "s", "id": 29840, "pid": 76337, "tid": -914061504, "ts": 1716454217654293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654396, "dur": 1,
    "args": {
      "External id": 29856,
      "cbid": 251, "correlation": 29856
    }
  },
  {
    "ph": "f", "id": 29856, "pid": 76337, "tid": -914061504, "ts": 1716454217654396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654401, "dur": 0,
    "args": {
      "External id": 29858,
      "cbid": 251, "correlation": 29858
    }
  },
  {
    "ph": "f", "id": 29858, "pid": 76337, "tid": -914061504, "ts": 1716454217654401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217707206, "dur": 529,
    "args": {
      "External id": 29859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29859,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 29859, "pid": 5, "tid": 7, "ts": 1716454217707206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654405, "dur": 14,
    "args": {
      "External id": 29859,
      "cbid": 211, "correlation": 29859
    }
  },
  {
    "ph": "s", "id": 29859, "pid": 76337, "tid": -914061504, "ts": 1716454217654405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217707737, "dur": 126,
    "args": {
      "External id": 29867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29867, "pid": 5, "tid": 7, "ts": 1716454217707737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654473, "dur": 13,
    "args": {
      "External id": 29867,
      "cbid": 211, "correlation": 29867
    }
  },
  {
    "ph": "s", "id": 29867, "pid": 76337, "tid": -914061504, "ts": 1716454217654473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217707864, "dur": 129,
    "args": {
      "External id": 29875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29875, "pid": 5, "tid": 7, "ts": 1716454217707864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654504, "dur": 9,
    "args": {
      "External id": 29875,
      "cbid": 211, "correlation": 29875
    }
  },
  {
    "ph": "s", "id": 29875, "pid": 76337, "tid": -914061504, "ts": 1716454217654504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654595, "dur": 1,
    "args": {
      "External id": 29891,
      "cbid": 251, "correlation": 29891
    }
  },
  {
    "ph": "f", "id": 29891, "pid": 76337, "tid": -914061504, "ts": 1716454217654595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217707994, "dur": 301,
    "args": {
      "External id": 29893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29893,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29893, "pid": 5, "tid": 7, "ts": 1716454217707994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654601, "dur": 14,
    "args": {
      "External id": 29893,
      "cbid": 211, "correlation": 29893
    }
  },
  {
    "ph": "s", "id": 29893, "pid": 76337, "tid": -914061504, "ts": 1716454217654601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217708296, "dur": 27,
    "args": {
      "External id": 29901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29901,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29901, "pid": 5, "tid": 7, "ts": 1716454217708296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654645, "dur": 10,
    "args": {
      "External id": 29901,
      "cbid": 211, "correlation": 29901
    }
  },
  {
    "ph": "s", "id": 29901, "pid": 76337, "tid": -914061504, "ts": 1716454217654645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217708324, "dur": 80,
    "args": {
      "External id": 29912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29912, "pid": 5, "tid": 7, "ts": 1716454217708324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654714, "dur": 12,
    "args": {
      "External id": 29912,
      "cbid": 211, "correlation": 29912
    }
  },
  {
    "ph": "s", "id": 29912, "pid": 76337, "tid": -914061504, "ts": 1716454217654714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654778, "dur": 0,
    "args": {
      "External id": 29924,
      "cbid": 317, "correlation": 29924
    }
  },
  {
    "ph": "f", "id": 29924, "pid": 76337, "tid": -914061504, "ts": 1716454217654778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654778, "dur": 0,
    "args": {
      "External id": 29925,
      "cbid": 203, "correlation": 29925
    }
  },
  {
    "ph": "f", "id": 29925, "pid": 76337, "tid": -914061504, "ts": 1716454217654778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654779, "dur": 0,
    "args": {
      "External id": 29926,
      "cbid": 205, "correlation": 29926
    }
  },
  {
    "ph": "f", "id": 29926, "pid": 76337, "tid": -914061504, "ts": 1716454217654779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217708405, "dur": 21,
    "args": {
      "External id": 29930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29930,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29930, "pid": 5, "tid": 7, "ts": 1716454217708405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654794, "dur": 12,
    "args": {
      "External id": 29930,
      "cbid": 211, "correlation": 29930
    }
  },
  {
    "ph": "s", "id": 29930, "pid": 76337, "tid": -914061504, "ts": 1716454217654794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217708428, "dur": 118,
    "args": {
      "External id": 29932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29932,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29932, "pid": 5, "tid": 7, "ts": 1716454217708428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654812, "dur": 7,
    "args": {
      "External id": 29932,
      "cbid": 211, "correlation": 29932
    }
  },
  {
    "ph": "s", "id": 29932, "pid": 76337, "tid": -914061504, "ts": 1716454217654812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217708547, "dur": 24,
    "args": {
      "External id": 29934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29934,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29934, "pid": 5, "tid": 7, "ts": 1716454217708547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654824, "dur": 5,
    "args": {
      "External id": 29934,
      "cbid": 211, "correlation": 29934
    }
  },
  {
    "ph": "s", "id": 29934, "pid": 76337, "tid": -914061504, "ts": 1716454217654824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217708572, "dur": 32,
    "args": {
      "External id": 29940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29940,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29940, "pid": 5, "tid": 7, "ts": 1716454217708572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654851, "dur": 8,
    "args": {
      "External id": 29940,
      "cbid": 211, "correlation": 29940
    }
  },
  {
    "ph": "s", "id": 29940, "pid": 76337, "tid": -914061504, "ts": 1716454217654851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217708605, "dur": 27,
    "args": {
      "External id": 29948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29948,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29948, "pid": 5, "tid": 7, "ts": 1716454217708605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654882, "dur": 9,
    "args": {
      "External id": 29948,
      "cbid": 211, "correlation": 29948
    }
  },
  {
    "ph": "s", "id": 29948, "pid": 76337, "tid": -914061504, "ts": 1716454217654882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454217708633, "dur": 99,
    "args": {
      "External id": 29959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29959,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29959, "pid": 5, "tid": 7, "ts": 1716454217708633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217654949, "dur": 12,
    "args": {
      "External id": 29959,
      "cbid": 211, "correlation": 29959
    }
  },
  {
    "ph": "s", "id": 29959, "pid": 76337, "tid": -914061504, "ts": 1716454217654949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655013, "dur": 0,
    "args": {
      "External id": 29969,
      "cbid": 317, "correlation": 29969
    }
  },
  {
    "ph": "f", "id": 29969, "pid": 76337, "tid": -914061504, "ts": 1716454217655013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655014, "dur": 0,
    "args": {
      "External id": 29970,
      "cbid": 203, "correlation": 29970
    }
  },
  {
    "ph": "f", "id": 29970, "pid": 76337, "tid": -914061504, "ts": 1716454217655014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655015, "dur": 0,
    "args": {
      "External id": 29971,
      "cbid": 205, "correlation": 29971
    }
  },
  {
    "ph": "f", "id": 29971, "pid": 76337, "tid": -914061504, "ts": 1716454217655015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217708734, "dur": 73,
    "args": {
      "External id": 29975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29975,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29975, "pid": 5, "tid": 7, "ts": 1716454217708734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655030, "dur": 12,
    "args": {
      "External id": 29975,
      "cbid": 211, "correlation": 29975
    }
  },
  {
    "ph": "s", "id": 29975, "pid": 76337, "tid": -914061504, "ts": 1716454217655030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217708808, "dur": 44,
    "args": {
      "External id": 29977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29977,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29977, "pid": 5, "tid": 7, "ts": 1716454217708808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655044, "dur": 5,
    "args": {
      "External id": 29977,
      "cbid": 211, "correlation": 29977
    }
  },
  {
    "ph": "s", "id": 29977, "pid": 76337, "tid": -914061504, "ts": 1716454217655044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217708853, "dur": 4,
    "args": {
      "External id": 29979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 29979, "pid": 5, "tid": 7, "ts": 1716454217708853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655055, "dur": 9,
    "args": {
      "External id": 29979,
      "cbid": 211, "correlation": 29979
    }
  },
  {
    "ph": "s", "id": 29979, "pid": 76337, "tid": -914061504, "ts": 1716454217655055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655068, "dur": 0,
    "args": {
      "External id": 29980,
      "cbid": 51, "correlation": 29980
    }
  },
  {
    "ph": "s", "id": 29980, "pid": 76337, "tid": -914061504, "ts": 1716454217655068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217708858, "dur": 2214,
    "args": {
      "External id": 29981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29981,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 29981, "pid": 5, "tid": 7, "ts": 1716454217708858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655069, "dur": 6,
    "args": {
      "External id": 29981,
      "cbid": 211, "correlation": 29981
    }
  },
  {
    "ph": "s", "id": 29981, "pid": 76337, "tid": -914061504, "ts": 1716454217655069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217711074, "dur": 111,
    "args": {
      "External id": 29986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29986,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29986, "pid": 5, "tid": 7, "ts": 1716454217711074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655098, "dur": 9,
    "args": {
      "External id": 29986,
      "cbid": 211, "correlation": 29986
    }
  },
  {
    "ph": "s", "id": 29986, "pid": 76337, "tid": -914061504, "ts": 1716454217655098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217711186, "dur": 162,
    "args": {
      "External id": 29995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 29995,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 29995, "pid": 5, "tid": 7, "ts": 1716454217711186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655192, "dur": 13,
    "args": {
      "External id": 29995,
      "cbid": 211, "correlation": 29995
    }
  },
  {
    "ph": "s", "id": 29995, "pid": 76337, "tid": -914061504, "ts": 1716454217655192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217711350, "dur": 128,
    "args": {
      "External id": 30015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30015,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 30015, "pid": 5, "tid": 7, "ts": 1716454217711350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655278, "dur": 14,
    "args": {
      "External id": 30015,
      "cbid": 211, "correlation": 30015
    }
  },
  {
    "ph": "s", "id": 30015, "pid": 76337, "tid": -914061504, "ts": 1716454217655278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217711478, "dur": 4,
    "args": {
      "External id": 30027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30027,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 30027, "pid": 5, "tid": 7, "ts": 1716454217711478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655301, "dur": 6,
    "args": {
      "External id": 30027,
      "cbid": 211, "correlation": 30027
    }
  },
  {
    "ph": "s", "id": 30027, "pid": 76337, "tid": -914061504, "ts": 1716454217655301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217711484, "dur": 156,
    "args": {
      "External id": 30030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30030, "pid": 5, "tid": 7, "ts": 1716454217711484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655320, "dur": 7,
    "args": {
      "External id": 30030,
      "cbid": 211, "correlation": 30030
    }
  },
  {
    "ph": "s", "id": 30030, "pid": 76337, "tid": -914061504, "ts": 1716454217655320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217711642, "dur": 101,
    "args": {
      "External id": 30039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30039,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30039, "pid": 5, "tid": 7, "ts": 1716454217711642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655361, "dur": 9,
    "args": {
      "External id": 30039,
      "cbid": 211, "correlation": 30039
    }
  },
  {
    "ph": "s", "id": 30039, "pid": 76337, "tid": -914061504, "ts": 1716454217655361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655424, "dur": 0,
    "args": {
      "External id": 30049,
      "cbid": 317, "correlation": 30049
    }
  },
  {
    "ph": "f", "id": 30049, "pid": 76337, "tid": -914061504, "ts": 1716454217655424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655425, "dur": 0,
    "args": {
      "External id": 30050,
      "cbid": 203, "correlation": 30050
    }
  },
  {
    "ph": "f", "id": 30050, "pid": 76337, "tid": -914061504, "ts": 1716454217655425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655426, "dur": 0,
    "args": {
      "External id": 30051,
      "cbid": 205, "correlation": 30051
    }
  },
  {
    "ph": "f", "id": 30051, "pid": 76337, "tid": -914061504, "ts": 1716454217655426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217711745, "dur": 111,
    "args": {
      "External id": 30055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30055,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30055, "pid": 5, "tid": 7, "ts": 1716454217711745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655442, "dur": 14,
    "args": {
      "External id": 30055,
      "cbid": 211, "correlation": 30055
    }
  },
  {
    "ph": "s", "id": 30055, "pid": 76337, "tid": -914061504, "ts": 1716454217655442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217711857, "dur": 34,
    "args": {
      "External id": 30057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30057,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30057, "pid": 5, "tid": 7, "ts": 1716454217711857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655458, "dur": 5,
    "args": {
      "External id": 30057,
      "cbid": 211, "correlation": 30057
    }
  },
  {
    "ph": "s", "id": 30057, "pid": 76337, "tid": -914061504, "ts": 1716454217655458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217711892, "dur": 3,
    "args": {
      "External id": 30059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30059, "pid": 5, "tid": 7, "ts": 1716454217711892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655469, "dur": 6,
    "args": {
      "External id": 30059,
      "cbid": 211, "correlation": 30059
    }
  },
  {
    "ph": "s", "id": 30059, "pid": 76337, "tid": -914061504, "ts": 1716454217655469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655478, "dur": 0,
    "args": {
      "External id": 30060,
      "cbid": 51, "correlation": 30060
    }
  },
  {
    "ph": "s", "id": 30060, "pid": 76337, "tid": -914061504, "ts": 1716454217655478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217711896, "dur": 1987,
    "args": {
      "External id": 30061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30061,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30061, "pid": 5, "tid": 7, "ts": 1716454217711896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655479, "dur": 6,
    "args": {
      "External id": 30061,
      "cbid": 211, "correlation": 30061
    }
  },
  {
    "ph": "s", "id": 30061, "pid": 76337, "tid": -914061504, "ts": 1716454217655479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217713884, "dur": 59,
    "args": {
      "External id": 30066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30066, "pid": 5, "tid": 7, "ts": 1716454217713884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655509, "dur": 8,
    "args": {
      "External id": 30066,
      "cbid": 211, "correlation": 30066
    }
  },
  {
    "ph": "s", "id": 30066, "pid": 76337, "tid": -914061504, "ts": 1716454217655509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217713945, "dur": 3,
    "args": {
      "External id": 30074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30074,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30074, "pid": 5, "tid": 7, "ts": 1716454217713945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655553, "dur": 10,
    "args": {
      "External id": 30074,
      "cbid": 211, "correlation": 30074
    }
  },
  {
    "ph": "s", "id": 30074, "pid": 76337, "tid": -914061504, "ts": 1716454217655553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655629, "dur": 1,
    "args": {
      "External id": 30090,
      "cbid": 251, "correlation": 30090
    }
  },
  {
    "ph": "f", "id": 30090, "pid": 76337, "tid": -914061504, "ts": 1716454217655629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655634, "dur": 0,
    "args": {
      "External id": 30092,
      "cbid": 251, "correlation": 30092
    }
  },
  {
    "ph": "f", "id": 30092, "pid": 76337, "tid": -914061504, "ts": 1716454217655634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217713949, "dur": 11,
    "args": {
      "External id": 30093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30093,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 30093, "pid": 5, "tid": 7, "ts": 1716454217713949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655636, "dur": 12,
    "args": {
      "External id": 30093,
      "cbid": 211, "correlation": 30093
    }
  },
  {
    "ph": "s", "id": 30093, "pid": 76337, "tid": -914061504, "ts": 1716454217655636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217713962, "dur": 5,
    "args": {
      "External id": 30095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30095,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 30095, "pid": 5, "tid": 7, "ts": 1716454217713962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655651, "dur": 6,
    "args": {
      "External id": 30095,
      "cbid": 211, "correlation": 30095
    }
  },
  {
    "ph": "s", "id": 30095, "pid": 76337, "tid": -914061504, "ts": 1716454217655651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217713968, "dur": 52,
    "args": {
      "External id": 30105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30105,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30105, "pid": 5, "tid": 7, "ts": 1716454217713968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655710, "dur": 12,
    "args": {
      "External id": 30105,
      "cbid": 211, "correlation": 30105
    }
  },
  {
    "ph": "s", "id": 30105, "pid": 76337, "tid": -914061504, "ts": 1716454217655710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217714022, "dur": 49,
    "args": {
      "External id": 30125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30125,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 30125, "pid": 5, "tid": 7, "ts": 1716454217714022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655777, "dur": 11,
    "args": {
      "External id": 30125,
      "cbid": 211, "correlation": 30125
    }
  },
  {
    "ph": "s", "id": 30125, "pid": 76337, "tid": -914061504, "ts": 1716454217655777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217714073, "dur": 4,
    "args": {
      "External id": 30137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30137,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30137, "pid": 5, "tid": 7, "ts": 1716454217714073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655798, "dur": 6,
    "args": {
      "External id": 30137,
      "cbid": 211, "correlation": 30137
    }
  },
  {
    "ph": "s", "id": 30137, "pid": 76337, "tid": -914061504, "ts": 1716454217655798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217714078, "dur": 55,
    "args": {
      "External id": 30140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30140, "pid": 5, "tid": 7, "ts": 1716454217714078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655815, "dur": 7,
    "args": {
      "External id": 30140,
      "cbid": 211, "correlation": 30140
    }
  },
  {
    "ph": "s", "id": 30140, "pid": 76337, "tid": -914061504, "ts": 1716454217655815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217714134, "dur": 36,
    "args": {
      "External id": 30149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30149,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30149, "pid": 5, "tid": 7, "ts": 1716454217714134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655855, "dur": 10,
    "args": {
      "External id": 30149,
      "cbid": 211, "correlation": 30149
    }
  },
  {
    "ph": "s", "id": 30149, "pid": 76337, "tid": -914061504, "ts": 1716454217655855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655919, "dur": 0,
    "args": {
      "External id": 30159,
      "cbid": 317, "correlation": 30159
    }
  },
  {
    "ph": "f", "id": 30159, "pid": 76337, "tid": -914061504, "ts": 1716454217655919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655920, "dur": 0,
    "args": {
      "External id": 30160,
      "cbid": 203, "correlation": 30160
    }
  },
  {
    "ph": "f", "id": 30160, "pid": 76337, "tid": -914061504, "ts": 1716454217655920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655921, "dur": 0,
    "args": {
      "External id": 30161,
      "cbid": 205, "correlation": 30161
    }
  },
  {
    "ph": "f", "id": 30161, "pid": 76337, "tid": -914061504, "ts": 1716454217655921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217714171, "dur": 41,
    "args": {
      "External id": 30165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30165,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30165, "pid": 5, "tid": 7, "ts": 1716454217714171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655939, "dur": 12,
    "args": {
      "External id": 30165,
      "cbid": 211, "correlation": 30165
    }
  },
  {
    "ph": "s", "id": 30165, "pid": 76337, "tid": -914061504, "ts": 1716454217655939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217714213, "dur": 14,
    "args": {
      "External id": 30167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30167,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30167, "pid": 5, "tid": 7, "ts": 1716454217714213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655953, "dur": 5,
    "args": {
      "External id": 30167,
      "cbid": 211, "correlation": 30167
    }
  },
  {
    "ph": "s", "id": 30167, "pid": 76337, "tid": -914061504, "ts": 1716454217655953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217714229, "dur": 3,
    "args": {
      "External id": 30169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30169, "pid": 5, "tid": 7, "ts": 1716454217714229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655963, "dur": 6,
    "args": {
      "External id": 30169,
      "cbid": 211, "correlation": 30169
    }
  },
  {
    "ph": "s", "id": 30169, "pid": 76337, "tid": -914061504, "ts": 1716454217655963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655972, "dur": 0,
    "args": {
      "External id": 30170,
      "cbid": 51, "correlation": 30170
    }
  },
  {
    "ph": "s", "id": 30170, "pid": 76337, "tid": -914061504, "ts": 1716454217655972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217714233, "dur": 691,
    "args": {
      "External id": 30171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30171,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30171, "pid": 5, "tid": 7, "ts": 1716454217714233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217655981, "dur": 5,
    "args": {
      "External id": 30171,
      "cbid": 211, "correlation": 30171
    }
  },
  {
    "ph": "s", "id": 30171, "pid": 76337, "tid": -914061504, "ts": 1716454217655981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217714925, "dur": 60,
    "args": {
      "External id": 30176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30176, "pid": 5, "tid": 7, "ts": 1716454217714925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656010, "dur": 9,
    "args": {
      "External id": 30176,
      "cbid": 211, "correlation": 30176
    }
  },
  {
    "ph": "s", "id": 30176, "pid": 76337, "tid": -914061504, "ts": 1716454217656010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656069, "dur": 0,
    "args": {
      "External id": 30186,
      "cbid": 317, "correlation": 30186
    }
  },
  {
    "ph": "f", "id": 30186, "pid": 76337, "tid": -914061504, "ts": 1716454217656069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656070, "dur": 0,
    "args": {
      "External id": 30187,
      "cbid": 203, "correlation": 30187
    }
  },
  {
    "ph": "f", "id": 30187, "pid": 76337, "tid": -914061504, "ts": 1716454217656070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656070, "dur": 0,
    "args": {
      "External id": 30188,
      "cbid": 205, "correlation": 30188
    }
  },
  {
    "ph": "f", "id": 30188, "pid": 76337, "tid": -914061504, "ts": 1716454217656070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217714987, "dur": 4,
    "args": {
      "External id": 30192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30192, "pid": 5, "tid": 7, "ts": 1716454217714987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656087, "dur": 11,
    "args": {
      "External id": 30192,
      "cbid": 211, "correlation": 30192
    }
  },
  {
    "ph": "s", "id": 30192, "pid": 76337, "tid": -914061504, "ts": 1716454217656087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656103, "dur": 0,
    "args": {
      "External id": 30193,
      "cbid": 51, "correlation": 30193
    }
  },
  {
    "ph": "s", "id": 30193, "pid": 76337, "tid": -914061504, "ts": 1716454217656103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217714991, "dur": 262,
    "args": {
      "External id": 30194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30194,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30194, "pid": 5, "tid": 7, "ts": 1716454217714991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656104, "dur": 6,
    "args": {
      "External id": 30194,
      "cbid": 211, "correlation": 30194
    }
  },
  {
    "ph": "s", "id": 30194, "pid": 76337, "tid": -914061504, "ts": 1716454217656104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217715255, "dur": 58,
    "args": {
      "External id": 30199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30199, "pid": 5, "tid": 7, "ts": 1716454217715255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656131, "dur": 8,
    "args": {
      "External id": 30199,
      "cbid": 211, "correlation": 30199
    }
  },
  {
    "ph": "s", "id": 30199, "pid": 76337, "tid": -914061504, "ts": 1716454217656131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217715314, "dur": 50,
    "args": {
      "External id": 30207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30207,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30207, "pid": 5, "tid": 7, "ts": 1716454217715314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656159, "dur": 8,
    "args": {
      "External id": 30207,
      "cbid": 211, "correlation": 30207
    }
  },
  {
    "ph": "s", "id": 30207, "pid": 76337, "tid": -914061504, "ts": 1716454217656159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217715366, "dur": 35,
    "args": {
      "External id": 30215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30215,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30215, "pid": 5, "tid": 7, "ts": 1716454217715366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656189, "dur": 8,
    "args": {
      "External id": 30215,
      "cbid": 211, "correlation": 30215
    }
  },
  {
    "ph": "s", "id": 30215, "pid": 76337, "tid": -914061504, "ts": 1716454217656189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217715402, "dur": 52,
    "args": {
      "External id": 30235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30235,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 30235, "pid": 5, "tid": 7, "ts": 1716454217715402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656269, "dur": 12,
    "args": {
      "External id": 30235,
      "cbid": 211, "correlation": 30235
    }
  },
  {
    "ph": "s", "id": 30235, "pid": 76337, "tid": -914061504, "ts": 1716454217656269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217715456, "dur": 4,
    "args": {
      "External id": 30247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30247,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30247, "pid": 5, "tid": 7, "ts": 1716454217715456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656291, "dur": 6,
    "args": {
      "External id": 30247,
      "cbid": 211, "correlation": 30247
    }
  },
  {
    "ph": "s", "id": 30247, "pid": 76337, "tid": -914061504, "ts": 1716454217656291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217715461, "dur": 56,
    "args": {
      "External id": 30250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30250,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30250, "pid": 5, "tid": 7, "ts": 1716454217715461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656308, "dur": 7,
    "args": {
      "External id": 30250,
      "cbid": 211, "correlation": 30250
    }
  },
  {
    "ph": "s", "id": 30250, "pid": 76337, "tid": -914061504, "ts": 1716454217656308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656377, "dur": 0,
    "args": {
      "External id": 30261,
      "cbid": 317, "correlation": 30261
    }
  },
  {
    "ph": "f", "id": 30261, "pid": 76337, "tid": -914061504, "ts": 1716454217656377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656378, "dur": 0,
    "args": {
      "External id": 30262,
      "cbid": 203, "correlation": 30262
    }
  },
  {
    "ph": "f", "id": 30262, "pid": 76337, "tid": -914061504, "ts": 1716454217656378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656379, "dur": 0,
    "args": {
      "External id": 30263,
      "cbid": 205, "correlation": 30263
    }
  },
  {
    "ph": "f", "id": 30263, "pid": 76337, "tid": -914061504, "ts": 1716454217656379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656411, "dur": 2,
    "args": {
      "External id": 30267,
      "cbid": 251, "correlation": 30267
    }
  },
  {
    "ph": "f", "id": 30267, "pid": 76337, "tid": -914061504, "ts": 1716454217656411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656414, "dur": 1,
    "args": {
      "External id": 30268,
      "cbid": 251, "correlation": 30268
    }
  },
  {
    "ph": "f", "id": 30268, "pid": 76337, "tid": -914061504, "ts": 1716454217656414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656416, "dur": 0,
    "args": {
      "External id": 30269,
      "cbid": 251, "correlation": 30269
    }
  },
  {
    "ph": "f", "id": 30269, "pid": 76337, "tid": -914061504, "ts": 1716454217656416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656417, "dur": 1,
    "args": {
      "External id": 30270,
      "cbid": 251, "correlation": 30270
    }
  },
  {
    "ph": "f", "id": 30270, "pid": 76337, "tid": -914061504, "ts": 1716454217656417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656419, "dur": 1,
    "args": {
      "External id": 30271,
      "cbid": 251, "correlation": 30271
    }
  },
  {
    "ph": "f", "id": 30271, "pid": 76337, "tid": -914061504, "ts": 1716454217656419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656421, "dur": 1,
    "args": {
      "External id": 30272,
      "cbid": 251, "correlation": 30272
    }
  },
  {
    "ph": "f", "id": 30272, "pid": 76337, "tid": -914061504, "ts": 1716454217656421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656423, "dur": 1,
    "args": {
      "External id": 30273,
      "cbid": 251, "correlation": 30273
    }
  },
  {
    "ph": "f", "id": 30273, "pid": 76337, "tid": -914061504, "ts": 1716454217656423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656425, "dur": 1,
    "args": {
      "External id": 30274,
      "cbid": 251, "correlation": 30274
    }
  },
  {
    "ph": "f", "id": 30274, "pid": 76337, "tid": -914061504, "ts": 1716454217656425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656427, "dur": 0,
    "args": {
      "External id": 30275,
      "cbid": 251, "correlation": 30275
    }
  },
  {
    "ph": "f", "id": 30275, "pid": 76337, "tid": -914061504, "ts": 1716454217656427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217715519, "dur": 114,
    "args": {
      "External id": 30276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30276,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 30276, "pid": 5, "tid": 7, "ts": 1716454217715519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656429, "dur": 14,
    "args": {
      "External id": 30276,
      "cbid": 211, "correlation": 30276
    }
  },
  {
    "ph": "s", "id": 30276, "pid": 76337, "tid": -914061504, "ts": 1716454217656429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217715634, "dur": 59,
    "args": {
      "External id": 30282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30282, "pid": 5, "tid": 7, "ts": 1716454217715634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656476, "dur": 10,
    "args": {
      "External id": 30282,
      "cbid": 211, "correlation": 30282
    }
  },
  {
    "ph": "s", "id": 30282, "pid": 76337, "tid": -914061504, "ts": 1716454217656476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217715694, "dur": 568,
    "args": {
      "External id": 30291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30291,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30291, "pid": 5, "tid": 7, "ts": 1716454217715694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656563, "dur": 14,
    "args": {
      "External id": 30291,
      "cbid": 211, "correlation": 30291
    }
  },
  {
    "ph": "s", "id": 30291, "pid": 76337, "tid": -914061504, "ts": 1716454217656563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217716263, "dur": 179,
    "args": {
      "External id": 30313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30313,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30313, "pid": 5, "tid": 7, "ts": 1716454217716263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656631, "dur": 12,
    "args": {
      "External id": 30313,
      "cbid": 211, "correlation": 30313
    }
  },
  {
    "ph": "s", "id": 30313, "pid": 76337, "tid": -914061504, "ts": 1716454217656631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656734, "dur": 1,
    "args": {
      "External id": 30324,
      "cbid": 251, "correlation": 30324
    }
  },
  {
    "ph": "f", "id": 30324, "pid": 76337, "tid": -914061504, "ts": 1716454217656734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217716443, "dur": 194,
    "args": {
      "External id": 30325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30325,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30325, "pid": 5, "tid": 7, "ts": 1716454217716443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656739, "dur": 14,
    "args": {
      "External id": 30325,
      "cbid": 211, "correlation": 30325
    }
  },
  {
    "ph": "s", "id": 30325, "pid": 76337, "tid": -914061504, "ts": 1716454217656739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656823, "dur": 1,
    "args": {
      "External id": 30336,
      "cbid": 251, "correlation": 30336
    }
  },
  {
    "ph": "f", "id": 30336, "pid": 76337, "tid": -914061504, "ts": 1716454217656823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217716639, "dur": 187,
    "args": {
      "External id": 30337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30337,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30337, "pid": 5, "tid": 7, "ts": 1716454217716639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656827, "dur": 13,
    "args": {
      "External id": 30337,
      "cbid": 211, "correlation": 30337
    }
  },
  {
    "ph": "s", "id": 30337, "pid": 76337, "tid": -914061504, "ts": 1716454217656827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656905, "dur": 1,
    "args": {
      "External id": 30348,
      "cbid": 251, "correlation": 30348
    }
  },
  {
    "ph": "f", "id": 30348, "pid": 76337, "tid": -914061504, "ts": 1716454217656905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217716827, "dur": 185,
    "args": {
      "External id": 30349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30349,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30349, "pid": 5, "tid": 7, "ts": 1716454217716827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217656910, "dur": 13,
    "args": {
      "External id": 30349,
      "cbid": 211, "correlation": 30349
    }
  },
  {
    "ph": "s", "id": 30349, "pid": 76337, "tid": -914061504, "ts": 1716454217656910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217717013, "dur": 18349,
    "args": {
      "External id": 30370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30370,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 30370, "pid": 5, "tid": 7, "ts": 1716454217717013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657003, "dur": 14,
    "args": {
      "External id": 30370,
      "cbid": 211, "correlation": 30370
    }
  },
  {
    "ph": "s", "id": 30370, "pid": 76337, "tid": -914061504, "ts": 1716454217657003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657116, "dur": 1,
    "args": {
      "External id": 30388,
      "cbid": 251, "correlation": 30388
    }
  },
  {
    "ph": "f", "id": 30388, "pid": 76337, "tid": -914061504, "ts": 1716454217657116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217735364, "dur": 201,
    "args": {
      "External id": 30390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30390,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30390, "pid": 5, "tid": 7, "ts": 1716454217735364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657122, "dur": 14,
    "args": {
      "External id": 30390,
      "cbid": 211, "correlation": 30390
    }
  },
  {
    "ph": "s", "id": 30390, "pid": 76337, "tid": -914061504, "ts": 1716454217657122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217735567, "dur": 67,
    "args": {
      "External id": 30398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30398,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30398, "pid": 5, "tid": 7, "ts": 1716454217735567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657195, "dur": 13,
    "args": {
      "External id": 30398,
      "cbid": 211, "correlation": 30398
    }
  },
  {
    "ph": "s", "id": 30398, "pid": 76337, "tid": -914061504, "ts": 1716454217657195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217735635, "dur": 96,
    "args": {
      "External id": 30406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30406,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30406, "pid": 5, "tid": 7, "ts": 1716454217735635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657235, "dur": 8,
    "args": {
      "External id": 30406,
      "cbid": 211, "correlation": 30406
    }
  },
  {
    "ph": "s", "id": 30406, "pid": 76337, "tid": -914061504, "ts": 1716454217657235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217735732, "dur": 53,
    "args": {
      "External id": 30417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30417, "pid": 5, "tid": 7, "ts": 1716454217735732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657307, "dur": 14,
    "args": {
      "External id": 30417,
      "cbid": 211, "correlation": 30417
    }
  },
  {
    "ph": "s", "id": 30417, "pid": 76337, "tid": -914061504, "ts": 1716454217657307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217735787, "dur": 91,
    "args": {
      "External id": 30439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30439,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30439, "pid": 5, "tid": 7, "ts": 1716454217735787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657340, "dur": 8,
    "args": {
      "External id": 30439,
      "cbid": 211, "correlation": 30439
    }
  },
  {
    "ph": "s", "id": 30439, "pid": 76337, "tid": -914061504, "ts": 1716454217657340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657436, "dur": 1,
    "args": {
      "External id": 30450,
      "cbid": 251, "correlation": 30450
    }
  },
  {
    "ph": "f", "id": 30450, "pid": 76337, "tid": -914061504, "ts": 1716454217657436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217735879, "dur": 103,
    "args": {
      "External id": 30451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30451,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30451, "pid": 5, "tid": 7, "ts": 1716454217735879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657441, "dur": 14,
    "args": {
      "External id": 30451,
      "cbid": 211, "correlation": 30451
    }
  },
  {
    "ph": "s", "id": 30451, "pid": 76337, "tid": -914061504, "ts": 1716454217657441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657535, "dur": 2,
    "args": {
      "External id": 30462,
      "cbid": 251, "correlation": 30462
    }
  },
  {
    "ph": "f", "id": 30462, "pid": 76337, "tid": -914061504, "ts": 1716454217657535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657540, "dur": 0,
    "args": {
      "External id": 30463,
      "cbid": 251, "correlation": 30463
    }
  },
  {
    "ph": "f", "id": 30463, "pid": 76337, "tid": -914061504, "ts": 1716454217657540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217735983, "dur": 10,
    "args": {
      "External id": 30464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30464,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 30464, "pid": 5, "tid": 7, "ts": 1716454217735983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657542, "dur": 15,
    "args": {
      "External id": 30464,
      "cbid": 211, "correlation": 30464
    }
  },
  {
    "ph": "s", "id": 30464, "pid": 76337, "tid": -914061504, "ts": 1716454217657542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217735994, "dur": 5,
    "args": {
      "External id": 30466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30466,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 30466, "pid": 5, "tid": 7, "ts": 1716454217735994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657561, "dur": 8,
    "args": {
      "External id": 30466,
      "cbid": 211, "correlation": 30466
    }
  },
  {
    "ph": "s", "id": 30466, "pid": 76337, "tid": -914061504, "ts": 1716454217657561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657638, "dur": 1,
    "args": {
      "External id": 30477,
      "cbid": 251, "correlation": 30477
    }
  },
  {
    "ph": "f", "id": 30477, "pid": 76337, "tid": -914061504, "ts": 1716454217657638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657642, "dur": 0,
    "args": {
      "External id": 30478,
      "cbid": 251, "correlation": 30478
    }
  },
  {
    "ph": "f", "id": 30478, "pid": 76337, "tid": -914061504, "ts": 1716454217657642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217736001, "dur": 6,
    "args": {
      "External id": 30479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30479,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 30479, "pid": 5, "tid": 7, "ts": 1716454217736001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657643, "dur": 13,
    "args": {
      "External id": 30479,
      "cbid": 211, "correlation": 30479
    }
  },
  {
    "ph": "s", "id": 30479, "pid": 76337, "tid": -914061504, "ts": 1716454217657643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217736008, "dur": 3,
    "args": {
      "External id": 30481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30481,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 30481, "pid": 5, "tid": 7, "ts": 1716454217736008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657658, "dur": 6,
    "args": {
      "External id": 30481,
      "cbid": 211, "correlation": 30481
    }
  },
  {
    "ph": "s", "id": 30481, "pid": 76337, "tid": -914061504, "ts": 1716454217657658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217736013, "dur": 153,
    "args": {
      "External id": 30502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30502,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 30502, "pid": 5, "tid": 7, "ts": 1716454217736013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657733, "dur": 13,
    "args": {
      "External id": 30502,
      "cbid": 211, "correlation": 30502
    }
  },
  {
    "ph": "s", "id": 30502, "pid": 76337, "tid": -914061504, "ts": 1716454217657733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657843, "dur": 2,
    "args": {
      "External id": 30520,
      "cbid": 251, "correlation": 30520
    }
  },
  {
    "ph": "f", "id": 30520, "pid": 76337, "tid": -914061504, "ts": 1716454217657843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217736167, "dur": 106,
    "args": {
      "External id": 30522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30522,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 30522, "pid": 5, "tid": 7, "ts": 1716454217736167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657849, "dur": 14,
    "args": {
      "External id": 30522,
      "cbid": 211, "correlation": 30522
    }
  },
  {
    "ph": "s", "id": 30522, "pid": 76337, "tid": -914061504, "ts": 1716454217657849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217736274, "dur": 35,
    "args": {
      "External id": 30530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30530,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30530, "pid": 5, "tid": 7, "ts": 1716454217736274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657921, "dur": 12,
    "args": {
      "External id": 30530,
      "cbid": 211, "correlation": 30530
    }
  },
  {
    "ph": "s", "id": 30530, "pid": 76337, "tid": -914061504, "ts": 1716454217657921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217736310, "dur": 68,
    "args": {
      "External id": 30538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30538, "pid": 5, "tid": 7, "ts": 1716454217736310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217657963, "dur": 9,
    "args": {
      "External id": 30538,
      "cbid": 211, "correlation": 30538
    }
  },
  {
    "ph": "s", "id": 30538, "pid": 76337, "tid": -914061504, "ts": 1716454217657963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217736380, "dur": 91,
    "args": {
      "External id": 30560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30560,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30560, "pid": 5, "tid": 7, "ts": 1716454217736380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217658024, "dur": 11,
    "args": {
      "External id": 30560,
      "cbid": 211, "correlation": 30560
    }
  },
  {
    "ph": "s", "id": 30560, "pid": 76337, "tid": -914061504, "ts": 1716454217658024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217658112, "dur": 1,
    "args": {
      "External id": 30576,
      "cbid": 251, "correlation": 30576
    }
  },
  {
    "ph": "f", "id": 30576, "pid": 76337, "tid": -914061504, "ts": 1716454217658112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217736472, "dur": 566,
    "args": {
      "External id": 30578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30578,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30578, "pid": 5, "tid": 7, "ts": 1716454217736472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217658117, "dur": 13,
    "args": {
      "External id": 30578,
      "cbid": 211, "correlation": 30578
    }
  },
  {
    "ph": "s", "id": 30578, "pid": 76337, "tid": -914061504, "ts": 1716454217658117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217737039, "dur": 242,
    "args": {
      "External id": 30586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30586,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30586, "pid": 5, "tid": 7, "ts": 1716454217737039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217658186, "dur": 12,
    "args": {
      "External id": 30586,
      "cbid": 211, "correlation": 30586
    }
  },
  {
    "ph": "s", "id": 30586, "pid": 76337, "tid": -914061504, "ts": 1716454217658186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217658218, "dur": 1,
    "args": {
      "External id": 30594,
      "cbid": 317, "correlation": 30594
    }
  },
  {
    "ph": "f", "id": 30594, "pid": 76337, "tid": -914061504, "ts": 1716454217658218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454217658220, "dur": 295342,
    "args": {
      "External id": 30595,
      "cbid": 20, "correlation": 30595
    }
  },
  {
    "ph": "f", "id": 30595, "pid": 76337, "tid": -914061504, "ts": 1716454217658220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217953625, "dur": 255,
    "args": {
      "External id": 30598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30598, "pid": 5, "tid": 7, "ts": 1716454217953625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217953590, "dur": 33,
    "args": {
      "External id": 30598,
      "cbid": 211, "correlation": 30598
    }
  },
  {
    "ph": "s", "id": 30598, "pid": 76337, "tid": -914061504, "ts": 1716454217953590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217953811, "dur": 4,
    "args": {
      "External id": 30614,
      "cbid": 251, "correlation": 30614
    }
  },
  {
    "ph": "f", "id": 30614, "pid": 76337, "tid": -914061504, "ts": 1716454217953811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217953819, "dur": 0,
    "args": {
      "External id": 30616,
      "cbid": 251, "correlation": 30616
    }
  },
  {
    "ph": "f", "id": 30616, "pid": 76337, "tid": -914061504, "ts": 1716454217953819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217953880, "dur": 359,
    "args": {
      "External id": 30617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30617,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 30617, "pid": 5, "tid": 7, "ts": 1716454217953880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217953825, "dur": 16,
    "args": {
      "External id": 30617,
      "cbid": 211, "correlation": 30617
    }
  },
  {
    "ph": "s", "id": 30617, "pid": 76337, "tid": -914061504, "ts": 1716454217953825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217954241, "dur": 50,
    "args": {
      "External id": 30625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30625,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30625, "pid": 5, "tid": 7, "ts": 1716454217954241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217953884, "dur": 12,
    "args": {
      "External id": 30625,
      "cbid": 211, "correlation": 30625
    }
  },
  {
    "ph": "s", "id": 30625, "pid": 76337, "tid": -914061504, "ts": 1716454217953884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217954292, "dur": 156,
    "args": {
      "External id": 30636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30636, "pid": 5, "tid": 7, "ts": 1716454217954292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954001, "dur": 14,
    "args": {
      "External id": 30636,
      "cbid": 211, "correlation": 30636
    }
  },
  {
    "ph": "s", "id": 30636, "pid": 76337, "tid": -914061504, "ts": 1716454217954001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954102, "dur": 1,
    "args": {
      "External id": 30648,
      "cbid": 317, "correlation": 30648
    }
  },
  {
    "ph": "f", "id": 30648, "pid": 76337, "tid": -914061504, "ts": 1716454217954102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954104, "dur": 1,
    "args": {
      "External id": 30649,
      "cbid": 203, "correlation": 30649
    }
  },
  {
    "ph": "f", "id": 30649, "pid": 76337, "tid": -914061504, "ts": 1716454217954104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954106, "dur": 0,
    "args": {
      "External id": 30650,
      "cbid": 205, "correlation": 30650
    }
  },
  {
    "ph": "f", "id": 30650, "pid": 76337, "tid": -914061504, "ts": 1716454217954106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954145, "dur": 2,
    "args": {
      "External id": 30654,
      "cbid": 251, "correlation": 30654
    }
  },
  {
    "ph": "f", "id": 30654, "pid": 76337, "tid": -914061504, "ts": 1716454217954145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954149, "dur": 0,
    "args": {
      "External id": 30655,
      "cbid": 251, "correlation": 30655
    }
  },
  {
    "ph": "f", "id": 30655, "pid": 76337, "tid": -914061504, "ts": 1716454217954149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954150, "dur": 0,
    "args": {
      "External id": 30656,
      "cbid": 251, "correlation": 30656
    }
  },
  {
    "ph": "f", "id": 30656, "pid": 76337, "tid": -914061504, "ts": 1716454217954150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954151, "dur": 0,
    "args": {
      "External id": 30657,
      "cbid": 251, "correlation": 30657
    }
  },
  {
    "ph": "f", "id": 30657, "pid": 76337, "tid": -914061504, "ts": 1716454217954151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954152, "dur": 1,
    "args": {
      "External id": 30658,
      "cbid": 251, "correlation": 30658
    }
  },
  {
    "ph": "f", "id": 30658, "pid": 76337, "tid": -914061504, "ts": 1716454217954152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954154, "dur": 1,
    "args": {
      "External id": 30659,
      "cbid": 251, "correlation": 30659
    }
  },
  {
    "ph": "f", "id": 30659, "pid": 76337, "tid": -914061504, "ts": 1716454217954154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954156, "dur": 1,
    "args": {
      "External id": 30660,
      "cbid": 251, "correlation": 30660
    }
  },
  {
    "ph": "f", "id": 30660, "pid": 76337, "tid": -914061504, "ts": 1716454217954156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954158, "dur": 1,
    "args": {
      "External id": 30661,
      "cbid": 251, "correlation": 30661
    }
  },
  {
    "ph": "f", "id": 30661, "pid": 76337, "tid": -914061504, "ts": 1716454217954158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954160, "dur": 0,
    "args": {
      "External id": 30662,
      "cbid": 251, "correlation": 30662
    }
  },
  {
    "ph": "f", "id": 30662, "pid": 76337, "tid": -914061504, "ts": 1716454217954160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217954450, "dur": 115,
    "args": {
      "External id": 30663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30663,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 30663, "pid": 5, "tid": 7, "ts": 1716454217954450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954163, "dur": 12,
    "args": {
      "External id": 30663,
      "cbid": 211, "correlation": 30663
    }
  },
  {
    "ph": "s", "id": 30663, "pid": 76337, "tid": -914061504, "ts": 1716454217954163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217954566, "dur": 59,
    "args": {
      "External id": 30669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30669, "pid": 5, "tid": 7, "ts": 1716454217954566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954202, "dur": 10,
    "args": {
      "External id": 30669,
      "cbid": 211, "correlation": 30669
    }
  },
  {
    "ph": "s", "id": 30669, "pid": 76337, "tid": -914061504, "ts": 1716454217954202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217954626, "dur": 50,
    "args": {
      "External id": 30677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30677,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30677, "pid": 5, "tid": 7, "ts": 1716454217954626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954235, "dur": 9,
    "args": {
      "External id": 30677,
      "cbid": 211, "correlation": 30677
    }
  },
  {
    "ph": "s", "id": 30677, "pid": 76337, "tid": -914061504, "ts": 1716454217954235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217954678, "dur": 97,
    "args": {
      "External id": 30686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30686,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30686, "pid": 5, "tid": 7, "ts": 1716454217954678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954293, "dur": 12,
    "args": {
      "External id": 30686,
      "cbid": 211, "correlation": 30686
    }
  },
  {
    "ph": "s", "id": 30686, "pid": 76337, "tid": -914061504, "ts": 1716454217954293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217954776, "dur": 91,
    "args": {
      "External id": 30706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30706,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 30706, "pid": 5, "tid": 7, "ts": 1716454217954776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954410, "dur": 14,
    "args": {
      "External id": 30706,
      "cbid": 211, "correlation": 30706
    }
  },
  {
    "ph": "s", "id": 30706, "pid": 76337, "tid": -914061504, "ts": 1716454217954410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217954869, "dur": 5,
    "args": {
      "External id": 30718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30718,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 30718, "pid": 5, "tid": 7, "ts": 1716454217954869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954434, "dur": 8,
    "args": {
      "External id": 30718,
      "cbid": 211, "correlation": 30718
    }
  },
  {
    "ph": "s", "id": 30718, "pid": 76337, "tid": -914061504, "ts": 1716454217954434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217954875, "dur": 108,
    "args": {
      "External id": 30721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30721,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30721, "pid": 5, "tid": 7, "ts": 1716454217954875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954457, "dur": 8,
    "args": {
      "External id": 30721,
      "cbid": 211, "correlation": 30721
    }
  },
  {
    "ph": "s", "id": 30721, "pid": 76337, "tid": -914061504, "ts": 1716454217954457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217954984, "dur": 70,
    "args": {
      "External id": 30730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30730,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30730, "pid": 5, "tid": 7, "ts": 1716454217954984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954507, "dur": 10,
    "args": {
      "External id": 30730,
      "cbid": 211, "correlation": 30730
    }
  },
  {
    "ph": "s", "id": 30730, "pid": 76337, "tid": -914061504, "ts": 1716454217954507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954574, "dur": 0,
    "args": {
      "External id": 30740,
      "cbid": 317, "correlation": 30740
    }
  },
  {
    "ph": "f", "id": 30740, "pid": 76337, "tid": -914061504, "ts": 1716454217954574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954575, "dur": 0,
    "args": {
      "External id": 30741,
      "cbid": 203, "correlation": 30741
    }
  },
  {
    "ph": "f", "id": 30741, "pid": 76337, "tid": -914061504, "ts": 1716454217954575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954576, "dur": 0,
    "args": {
      "External id": 30742,
      "cbid": 205, "correlation": 30742
    }
  },
  {
    "ph": "f", "id": 30742, "pid": 76337, "tid": -914061504, "ts": 1716454217954576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217955056, "dur": 75,
    "args": {
      "External id": 30746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30746,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30746, "pid": 5, "tid": 7, "ts": 1716454217955056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954599, "dur": 13,
    "args": {
      "External id": 30746,
      "cbid": 211, "correlation": 30746
    }
  },
  {
    "ph": "s", "id": 30746, "pid": 76337, "tid": -914061504, "ts": 1716454217954599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217955132, "dur": 24,
    "args": {
      "External id": 30748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30748,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30748, "pid": 5, "tid": 7, "ts": 1716454217955132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954615, "dur": 5,
    "args": {
      "External id": 30748,
      "cbid": 211, "correlation": 30748
    }
  },
  {
    "ph": "s", "id": 30748, "pid": 76337, "tid": -914061504, "ts": 1716454217954615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217955158, "dur": 3,
    "args": {
      "External id": 30750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30750, "pid": 5, "tid": 7, "ts": 1716454217955158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954631, "dur": 8,
    "args": {
      "External id": 30750,
      "cbid": 211, "correlation": 30750
    }
  },
  {
    "ph": "s", "id": 30750, "pid": 76337, "tid": -914061504, "ts": 1716454217954631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954645, "dur": 0,
    "args": {
      "External id": 30751,
      "cbid": 51, "correlation": 30751
    }
  },
  {
    "ph": "s", "id": 30751, "pid": 76337, "tid": -914061504, "ts": 1716454217954645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217955163, "dur": 1349,
    "args": {
      "External id": 30752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30752,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30752, "pid": 5, "tid": 7, "ts": 1716454217955163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954647, "dur": 7,
    "args": {
      "External id": 30752,
      "cbid": 211, "correlation": 30752
    }
  },
  {
    "ph": "s", "id": 30752, "pid": 76337, "tid": -914061504, "ts": 1716454217954647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217956513, "dur": 59,
    "args": {
      "External id": 30757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30757,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30757, "pid": 5, "tid": 7, "ts": 1716454217956513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954677, "dur": 8,
    "args": {
      "External id": 30757,
      "cbid": 211, "correlation": 30757
    }
  },
  {
    "ph": "s", "id": 30757, "pid": 76337, "tid": -914061504, "ts": 1716454217954677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217956573, "dur": 4,
    "args": {
      "External id": 30765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30765,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30765, "pid": 5, "tid": 7, "ts": 1716454217956573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954722, "dur": 9,
    "args": {
      "External id": 30765,
      "cbid": 211, "correlation": 30765
    }
  },
  {
    "ph": "s", "id": 30765, "pid": 76337, "tid": -914061504, "ts": 1716454217954722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954809, "dur": 2,
    "args": {
      "External id": 30781,
      "cbid": 251, "correlation": 30781
    }
  },
  {
    "ph": "f", "id": 30781, "pid": 76337, "tid": -914061504, "ts": 1716454217954809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954815, "dur": 0,
    "args": {
      "External id": 30783,
      "cbid": 251, "correlation": 30783
    }
  },
  {
    "ph": "f", "id": 30783, "pid": 76337, "tid": -914061504, "ts": 1716454217954815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217956578, "dur": 11,
    "args": {
      "External id": 30784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30784,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 30784, "pid": 5, "tid": 7, "ts": 1716454217956578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954818, "dur": 14,
    "args": {
      "External id": 30784,
      "cbid": 211, "correlation": 30784
    }
  },
  {
    "ph": "s", "id": 30784, "pid": 76337, "tid": -914061504, "ts": 1716454217954818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217956591, "dur": 5,
    "args": {
      "External id": 30786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30786,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 30786, "pid": 5, "tid": 7, "ts": 1716454217956591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954835, "dur": 7,
    "args": {
      "External id": 30786,
      "cbid": 211, "correlation": 30786
    }
  },
  {
    "ph": "s", "id": 30786, "pid": 76337, "tid": -914061504, "ts": 1716454217954835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217956597, "dur": 54,
    "args": {
      "External id": 30796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30796, "pid": 5, "tid": 7, "ts": 1716454217956597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954906, "dur": 12,
    "args": {
      "External id": 30796,
      "cbid": 211, "correlation": 30796
    }
  },
  {
    "ph": "s", "id": 30796, "pid": 76337, "tid": -914061504, "ts": 1716454217954906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217956653, "dur": 50,
    "args": {
      "External id": 30816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30816,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 30816, "pid": 5, "tid": 7, "ts": 1716454217956653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217954985, "dur": 12,
    "args": {
      "External id": 30816,
      "cbid": 211, "correlation": 30816
    }
  },
  {
    "ph": "s", "id": 30816, "pid": 76337, "tid": -914061504, "ts": 1716454217954985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217956704, "dur": 4,
    "args": {
      "External id": 30828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30828,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30828, "pid": 5, "tid": 7, "ts": 1716454217956704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955008, "dur": 6,
    "args": {
      "External id": 30828,
      "cbid": 211, "correlation": 30828
    }
  },
  {
    "ph": "s", "id": 30828, "pid": 76337, "tid": -914061504, "ts": 1716454217955008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217956709, "dur": 54,
    "args": {
      "External id": 30831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30831,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30831, "pid": 5, "tid": 7, "ts": 1716454217956709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955026, "dur": 7,
    "args": {
      "External id": 30831,
      "cbid": 211, "correlation": 30831
    }
  },
  {
    "ph": "s", "id": 30831, "pid": 76337, "tid": -914061504, "ts": 1716454217955026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217956765, "dur": 37,
    "args": {
      "External id": 30840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30840,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30840, "pid": 5, "tid": 7, "ts": 1716454217956765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955068, "dur": 9,
    "args": {
      "External id": 30840,
      "cbid": 211, "correlation": 30840
    }
  },
  {
    "ph": "s", "id": 30840, "pid": 76337, "tid": -914061504, "ts": 1716454217955068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955132, "dur": 0,
    "args": {
      "External id": 30850,
      "cbid": 317, "correlation": 30850
    }
  },
  {
    "ph": "f", "id": 30850, "pid": 76337, "tid": -914061504, "ts": 1716454217955132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955133, "dur": 0,
    "args": {
      "External id": 30851,
      "cbid": 203, "correlation": 30851
    }
  },
  {
    "ph": "f", "id": 30851, "pid": 76337, "tid": -914061504, "ts": 1716454217955133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955134, "dur": 0,
    "args": {
      "External id": 30852,
      "cbid": 205, "correlation": 30852
    }
  },
  {
    "ph": "f", "id": 30852, "pid": 76337, "tid": -914061504, "ts": 1716454217955134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217956803, "dur": 41,
    "args": {
      "External id": 30856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30856, "pid": 5, "tid": 7, "ts": 1716454217956803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955150, "dur": 13,
    "args": {
      "External id": 30856,
      "cbid": 211, "correlation": 30856
    }
  },
  {
    "ph": "s", "id": 30856, "pid": 76337, "tid": -914061504, "ts": 1716454217955150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217956845, "dur": 14,
    "args": {
      "External id": 30858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30858,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30858, "pid": 5, "tid": 7, "ts": 1716454217956845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955165, "dur": 5,
    "args": {
      "External id": 30858,
      "cbid": 211, "correlation": 30858
    }
  },
  {
    "ph": "s", "id": 30858, "pid": 76337, "tid": -914061504, "ts": 1716454217955165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217956860, "dur": 3,
    "args": {
      "External id": 30860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30860,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30860, "pid": 5, "tid": 7, "ts": 1716454217956860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955176, "dur": 5,
    "args": {
      "External id": 30860,
      "cbid": 211, "correlation": 30860
    }
  },
  {
    "ph": "s", "id": 30860, "pid": 76337, "tid": -914061504, "ts": 1716454217955176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955184, "dur": 0,
    "args": {
      "External id": 30861,
      "cbid": 51, "correlation": 30861
    }
  },
  {
    "ph": "s", "id": 30861, "pid": 76337, "tid": -914061504, "ts": 1716454217955184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217956865, "dur": 690,
    "args": {
      "External id": 30862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30862,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30862, "pid": 5, "tid": 7, "ts": 1716454217956865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955185, "dur": 5,
    "args": {
      "External id": 30862,
      "cbid": 211, "correlation": 30862
    }
  },
  {
    "ph": "s", "id": 30862, "pid": 76337, "tid": -914061504, "ts": 1716454217955185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217957556, "dur": 58,
    "args": {
      "External id": 30867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30867, "pid": 5, "tid": 7, "ts": 1716454217957556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955213, "dur": 8,
    "args": {
      "External id": 30867,
      "cbid": 211, "correlation": 30867
    }
  },
  {
    "ph": "s", "id": 30867, "pid": 76337, "tid": -914061504, "ts": 1716454217955213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955272, "dur": 0,
    "args": {
      "External id": 30877,
      "cbid": 317, "correlation": 30877
    }
  },
  {
    "ph": "f", "id": 30877, "pid": 76337, "tid": -914061504, "ts": 1716454217955272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955273, "dur": 0,
    "args": {
      "External id": 30878,
      "cbid": 203, "correlation": 30878
    }
  },
  {
    "ph": "f", "id": 30878, "pid": 76337, "tid": -914061504, "ts": 1716454217955273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955273, "dur": 0,
    "args": {
      "External id": 30879,
      "cbid": 205, "correlation": 30879
    }
  },
  {
    "ph": "f", "id": 30879, "pid": 76337, "tid": -914061504, "ts": 1716454217955273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217957616, "dur": 74,
    "args": {
      "External id": 30883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30883,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30883, "pid": 5, "tid": 7, "ts": 1716454217957616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955287, "dur": 12,
    "args": {
      "External id": 30883,
      "cbid": 211, "correlation": 30883
    }
  },
  {
    "ph": "s", "id": 30883, "pid": 76337, "tid": -914061504, "ts": 1716454217955287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217957691, "dur": 208,
    "args": {
      "External id": 30885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30885,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 30885, "pid": 5, "tid": 7, "ts": 1716454217957691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955311, "dur": 9,
    "args": {
      "External id": 30885,
      "cbid": 211, "correlation": 30885
    }
  },
  {
    "ph": "s", "id": 30885, "pid": 76337, "tid": -914061504, "ts": 1716454217955311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217957900, "dur": 40,
    "args": {
      "External id": 30887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30887,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30887, "pid": 5, "tid": 7, "ts": 1716454217957900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955327, "dur": 7,
    "args": {
      "External id": 30887,
      "cbid": 211, "correlation": 30887
    }
  },
  {
    "ph": "s", "id": 30887, "pid": 76337, "tid": -914061504, "ts": 1716454217955327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217957942, "dur": 59,
    "args": {
      "External id": 30893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30893,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30893, "pid": 5, "tid": 7, "ts": 1716454217957942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955355, "dur": 8,
    "args": {
      "External id": 30893,
      "cbid": 211, "correlation": 30893
    }
  },
  {
    "ph": "s", "id": 30893, "pid": 76337, "tid": -914061504, "ts": 1716454217955355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217958002, "dur": 50,
    "args": {
      "External id": 30901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30901,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30901, "pid": 5, "tid": 7, "ts": 1716454217958002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955384, "dur": 8,
    "args": {
      "External id": 30901,
      "cbid": 211, "correlation": 30901
    }
  },
  {
    "ph": "s", "id": 30901, "pid": 76337, "tid": -914061504, "ts": 1716454217955384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217958053, "dur": 35,
    "args": {
      "External id": 30909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30909,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30909, "pid": 5, "tid": 7, "ts": 1716454217958053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955422, "dur": 10,
    "args": {
      "External id": 30909,
      "cbid": 211, "correlation": 30909
    }
  },
  {
    "ph": "s", "id": 30909, "pid": 76337, "tid": -914061504, "ts": 1716454217955422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217958090, "dur": 52,
    "args": {
      "External id": 30929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30929,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 30929, "pid": 5, "tid": 7, "ts": 1716454217958090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955522, "dur": 13,
    "args": {
      "External id": 30929,
      "cbid": 211, "correlation": 30929
    }
  },
  {
    "ph": "s", "id": 30929, "pid": 76337, "tid": -914061504, "ts": 1716454217955522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217958143, "dur": 5,
    "args": {
      "External id": 30941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30941,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 30941, "pid": 5, "tid": 7, "ts": 1716454217958143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955554, "dur": 7,
    "args": {
      "External id": 30941,
      "cbid": 211, "correlation": 30941
    }
  },
  {
    "ph": "s", "id": 30941, "pid": 76337, "tid": -914061504, "ts": 1716454217955554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217958150, "dur": 56,
    "args": {
      "External id": 30944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30944,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30944, "pid": 5, "tid": 7, "ts": 1716454217958150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955574, "dur": 6,
    "args": {
      "External id": 30944,
      "cbid": 211, "correlation": 30944
    }
  },
  {
    "ph": "s", "id": 30944, "pid": 76337, "tid": -914061504, "ts": 1716454217955574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955643, "dur": 0,
    "args": {
      "External id": 30955,
      "cbid": 317, "correlation": 30955
    }
  },
  {
    "ph": "f", "id": 30955, "pid": 76337, "tid": -914061504, "ts": 1716454217955643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955644, "dur": 0,
    "args": {
      "External id": 30956,
      "cbid": 203, "correlation": 30956
    }
  },
  {
    "ph": "f", "id": 30956, "pid": 76337, "tid": -914061504, "ts": 1716454217955644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955645, "dur": 0,
    "args": {
      "External id": 30957,
      "cbid": 205, "correlation": 30957
    }
  },
  {
    "ph": "f", "id": 30957, "pid": 76337, "tid": -914061504, "ts": 1716454217955645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955669, "dur": 1,
    "args": {
      "External id": 30961,
      "cbid": 251, "correlation": 30961
    }
  },
  {
    "ph": "f", "id": 30961, "pid": 76337, "tid": -914061504, "ts": 1716454217955669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955671, "dur": 0,
    "args": {
      "External id": 30962,
      "cbid": 251, "correlation": 30962
    }
  },
  {
    "ph": "f", "id": 30962, "pid": 76337, "tid": -914061504, "ts": 1716454217955671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955672, "dur": 0,
    "args": {
      "External id": 30963,
      "cbid": 251, "correlation": 30963
    }
  },
  {
    "ph": "f", "id": 30963, "pid": 76337, "tid": -914061504, "ts": 1716454217955672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955673, "dur": 0,
    "args": {
      "External id": 30964,
      "cbid": 251, "correlation": 30964
    }
  },
  {
    "ph": "f", "id": 30964, "pid": 76337, "tid": -914061504, "ts": 1716454217955673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955673, "dur": 0,
    "args": {
      "External id": 30965,
      "cbid": 251, "correlation": 30965
    }
  },
  {
    "ph": "f", "id": 30965, "pid": 76337, "tid": -914061504, "ts": 1716454217955673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955674, "dur": 0,
    "args": {
      "External id": 30966,
      "cbid": 251, "correlation": 30966
    }
  },
  {
    "ph": "f", "id": 30966, "pid": 76337, "tid": -914061504, "ts": 1716454217955674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955675, "dur": 0,
    "args": {
      "External id": 30967,
      "cbid": 251, "correlation": 30967
    }
  },
  {
    "ph": "f", "id": 30967, "pid": 76337, "tid": -914061504, "ts": 1716454217955675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955676, "dur": 0,
    "args": {
      "External id": 30968,
      "cbid": 251, "correlation": 30968
    }
  },
  {
    "ph": "f", "id": 30968, "pid": 76337, "tid": -914061504, "ts": 1716454217955676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955677, "dur": 0,
    "args": {
      "External id": 30969,
      "cbid": 251, "correlation": 30969
    }
  },
  {
    "ph": "f", "id": 30969, "pid": 76337, "tid": -914061504, "ts": 1716454217955677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217958207, "dur": 113,
    "args": {
      "External id": 30970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30970,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 30970, "pid": 5, "tid": 7, "ts": 1716454217958207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955680, "dur": 13,
    "args": {
      "External id": 30970,
      "cbid": 211, "correlation": 30970
    }
  },
  {
    "ph": "s", "id": 30970, "pid": 76337, "tid": -914061504, "ts": 1716454217955680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217958322, "dur": 60,
    "args": {
      "External id": 30976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30976,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30976, "pid": 5, "tid": 7, "ts": 1716454217958322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955715, "dur": 10,
    "args": {
      "External id": 30976,
      "cbid": 211, "correlation": 30976
    }
  },
  {
    "ph": "s", "id": 30976, "pid": 76337, "tid": -914061504, "ts": 1716454217955715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217958383, "dur": 574,
    "args": {
      "External id": 30985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 30985,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 30985, "pid": 5, "tid": 7, "ts": 1716454217958383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955806, "dur": 15,
    "args": {
      "External id": 30985,
      "cbid": 211, "correlation": 30985
    }
  },
  {
    "ph": "s", "id": 30985, "pid": 76337, "tid": -914061504, "ts": 1716454217955806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217958958, "dur": 179,
    "args": {
      "External id": 31007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31007,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31007, "pid": 5, "tid": 7, "ts": 1716454217958958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217955875, "dur": 12,
    "args": {
      "External id": 31007,
      "cbid": 211, "correlation": 31007
    }
  },
  {
    "ph": "s", "id": 31007, "pid": 76337, "tid": -914061504, "ts": 1716454217955875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956010, "dur": 2,
    "args": {
      "External id": 31018,
      "cbid": 251, "correlation": 31018
    }
  },
  {
    "ph": "f", "id": 31018, "pid": 76337, "tid": -914061504, "ts": 1716454217956010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217959138, "dur": 194,
    "args": {
      "External id": 31019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31019,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31019, "pid": 5, "tid": 7, "ts": 1716454217959138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956016, "dur": 15,
    "args": {
      "External id": 31019,
      "cbid": 211, "correlation": 31019
    }
  },
  {
    "ph": "s", "id": 31019, "pid": 76337, "tid": -914061504, "ts": 1716454217956016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956105, "dur": 1,
    "args": {
      "External id": 31030,
      "cbid": 251, "correlation": 31030
    }
  },
  {
    "ph": "f", "id": 31030, "pid": 76337, "tid": -914061504, "ts": 1716454217956105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217959334, "dur": 187,
    "args": {
      "External id": 31031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31031,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31031, "pid": 5, "tid": 7, "ts": 1716454217959334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956109, "dur": 13,
    "args": {
      "External id": 31031,
      "cbid": 211, "correlation": 31031
    }
  },
  {
    "ph": "s", "id": 31031, "pid": 76337, "tid": -914061504, "ts": 1716454217956109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956187, "dur": 1,
    "args": {
      "External id": 31042,
      "cbid": 251, "correlation": 31042
    }
  },
  {
    "ph": "f", "id": 31042, "pid": 76337, "tid": -914061504, "ts": 1716454217956187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217959522, "dur": 185,
    "args": {
      "External id": 31043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31043,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31043, "pid": 5, "tid": 7, "ts": 1716454217959522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956192, "dur": 12,
    "args": {
      "External id": 31043,
      "cbid": 211, "correlation": 31043
    }
  },
  {
    "ph": "s", "id": 31043, "pid": 76337, "tid": -914061504, "ts": 1716454217956192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217959708, "dur": 18357,
    "args": {
      "External id": 31064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31064,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31064, "pid": 5, "tid": 7, "ts": 1716454217959708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956297, "dur": 15,
    "args": {
      "External id": 31064,
      "cbid": 211, "correlation": 31064
    }
  },
  {
    "ph": "s", "id": 31064, "pid": 76337, "tid": -914061504, "ts": 1716454217956297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956433, "dur": 2,
    "args": {
      "External id": 31082,
      "cbid": 251, "correlation": 31082
    }
  },
  {
    "ph": "f", "id": 31082, "pid": 76337, "tid": -914061504, "ts": 1716454217956433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217978066, "dur": 199,
    "args": {
      "External id": 31084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31084,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31084, "pid": 5, "tid": 7, "ts": 1716454217978066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956440, "dur": 14,
    "args": {
      "External id": 31084,
      "cbid": 211, "correlation": 31084
    }
  },
  {
    "ph": "s", "id": 31084, "pid": 76337, "tid": -914061504, "ts": 1716454217956440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217978267, "dur": 68,
    "args": {
      "External id": 31092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31092,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31092, "pid": 5, "tid": 7, "ts": 1716454217978267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956513, "dur": 12,
    "args": {
      "External id": 31092,
      "cbid": 211, "correlation": 31092
    }
  },
  {
    "ph": "s", "id": 31092, "pid": 76337, "tid": -914061504, "ts": 1716454217956513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217978335, "dur": 97,
    "args": {
      "External id": 31100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31100,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31100, "pid": 5, "tid": 7, "ts": 1716454217978335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956552, "dur": 8,
    "args": {
      "External id": 31100,
      "cbid": 211, "correlation": 31100
    }
  },
  {
    "ph": "s", "id": 31100, "pid": 76337, "tid": -914061504, "ts": 1716454217956552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217978434, "dur": 53,
    "args": {
      "External id": 31111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31111,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31111, "pid": 5, "tid": 7, "ts": 1716454217978434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956633, "dur": 13,
    "args": {
      "External id": 31111,
      "cbid": 211, "correlation": 31111
    }
  },
  {
    "ph": "s", "id": 31111, "pid": 76337, "tid": -914061504, "ts": 1716454217956633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217978488, "dur": 91,
    "args": {
      "External id": 31133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31133,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31133, "pid": 5, "tid": 7, "ts": 1716454217978488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956667, "dur": 8,
    "args": {
      "External id": 31133,
      "cbid": 211, "correlation": 31133
    }
  },
  {
    "ph": "s", "id": 31133, "pid": 76337, "tid": -914061504, "ts": 1716454217956667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956751, "dur": 1,
    "args": {
      "External id": 31144,
      "cbid": 251, "correlation": 31144
    }
  },
  {
    "ph": "f", "id": 31144, "pid": 76337, "tid": -914061504, "ts": 1716454217956751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217978581, "dur": 107,
    "args": {
      "External id": 31145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31145,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31145, "pid": 5, "tid": 7, "ts": 1716454217978581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956756, "dur": 13,
    "args": {
      "External id": 31145,
      "cbid": 211, "correlation": 31145
    }
  },
  {
    "ph": "s", "id": 31145, "pid": 76337, "tid": -914061504, "ts": 1716454217956756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956838, "dur": 1,
    "args": {
      "External id": 31156,
      "cbid": 251, "correlation": 31156
    }
  },
  {
    "ph": "f", "id": 31156, "pid": 76337, "tid": -914061504, "ts": 1716454217956838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956842, "dur": 0,
    "args": {
      "External id": 31157,
      "cbid": 251, "correlation": 31157
    }
  },
  {
    "ph": "f", "id": 31157, "pid": 76337, "tid": -914061504, "ts": 1716454217956842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217978689, "dur": 11,
    "args": {
      "External id": 31158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31158,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 31158, "pid": 5, "tid": 7, "ts": 1716454217978689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956844, "dur": 14,
    "args": {
      "External id": 31158,
      "cbid": 211, "correlation": 31158
    }
  },
  {
    "ph": "s", "id": 31158, "pid": 76337, "tid": -914061504, "ts": 1716454217956844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217978701, "dur": 5,
    "args": {
      "External id": 31160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31160,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 31160, "pid": 5, "tid": 7, "ts": 1716454217978701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956861, "dur": 7,
    "args": {
      "External id": 31160,
      "cbid": 211, "correlation": 31160
    }
  },
  {
    "ph": "s", "id": 31160, "pid": 76337, "tid": -914061504, "ts": 1716454217956861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956938, "dur": 1,
    "args": {
      "External id": 31171,
      "cbid": 251, "correlation": 31171
    }
  },
  {
    "ph": "f", "id": 31171, "pid": 76337, "tid": -914061504, "ts": 1716454217956938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956941, "dur": 0,
    "args": {
      "External id": 31172,
      "cbid": 251, "correlation": 31172
    }
  },
  {
    "ph": "f", "id": 31172, "pid": 76337, "tid": -914061504, "ts": 1716454217956941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217978707, "dur": 6,
    "args": {
      "External id": 31173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31173,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 31173, "pid": 5, "tid": 7, "ts": 1716454217978707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956943, "dur": 13,
    "args": {
      "External id": 31173,
      "cbid": 211, "correlation": 31173
    }
  },
  {
    "ph": "s", "id": 31173, "pid": 76337, "tid": -914061504, "ts": 1716454217956943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217978715, "dur": 3,
    "args": {
      "External id": 31175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31175,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 31175, "pid": 5, "tid": 7, "ts": 1716454217978715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217956958, "dur": 6,
    "args": {
      "External id": 31175,
      "cbid": 211, "correlation": 31175
    }
  },
  {
    "ph": "s", "id": 31175, "pid": 76337, "tid": -914061504, "ts": 1716454217956958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217978720, "dur": 155,
    "args": {
      "External id": 31196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31196,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31196, "pid": 5, "tid": 7, "ts": 1716454217978720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957040, "dur": 13,
    "args": {
      "External id": 31196,
      "cbid": 211, "correlation": 31196
    }
  },
  {
    "ph": "s", "id": 31196, "pid": 76337, "tid": -914061504, "ts": 1716454217957040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957139, "dur": 2,
    "args": {
      "External id": 31214,
      "cbid": 251, "correlation": 31214
    }
  },
  {
    "ph": "f", "id": 31214, "pid": 76337, "tid": -914061504, "ts": 1716454217957139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217978876, "dur": 106,
    "args": {
      "External id": 31216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31216,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31216, "pid": 5, "tid": 7, "ts": 1716454217978876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957145, "dur": 14,
    "args": {
      "External id": 31216,
      "cbid": 211, "correlation": 31216
    }
  },
  {
    "ph": "s", "id": 31216, "pid": 76337, "tid": -914061504, "ts": 1716454217957145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217978984, "dur": 34,
    "args": {
      "External id": 31224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31224,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31224, "pid": 5, "tid": 7, "ts": 1716454217978984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957215, "dur": 13,
    "args": {
      "External id": 31224,
      "cbid": 211, "correlation": 31224
    }
  },
  {
    "ph": "s", "id": 31224, "pid": 76337, "tid": -914061504, "ts": 1716454217957215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217979019, "dur": 66,
    "args": {
      "External id": 31232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31232,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31232, "pid": 5, "tid": 7, "ts": 1716454217979019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957256, "dur": 9,
    "args": {
      "External id": 31232,
      "cbid": 211, "correlation": 31232
    }
  },
  {
    "ph": "s", "id": 31232, "pid": 76337, "tid": -914061504, "ts": 1716454217957256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217979087, "dur": 91,
    "args": {
      "External id": 31254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31254,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31254, "pid": 5, "tid": 7, "ts": 1716454217979087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957308, "dur": 10,
    "args": {
      "External id": 31254,
      "cbid": 211, "correlation": 31254
    }
  },
  {
    "ph": "s", "id": 31254, "pid": 76337, "tid": -914061504, "ts": 1716454217957308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957418, "dur": 1,
    "args": {
      "External id": 31270,
      "cbid": 251, "correlation": 31270
    }
  },
  {
    "ph": "f", "id": 31270, "pid": 76337, "tid": -914061504, "ts": 1716454217957418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217979180, "dur": 569,
    "args": {
      "External id": 31272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31272,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31272, "pid": 5, "tid": 7, "ts": 1716454217979180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957424, "dur": 13,
    "args": {
      "External id": 31272,
      "cbid": 211, "correlation": 31272
    }
  },
  {
    "ph": "s", "id": 31272, "pid": 76337, "tid": -914061504, "ts": 1716454217957424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217979750, "dur": 239,
    "args": {
      "External id": 31280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31280, "pid": 5, "tid": 7, "ts": 1716454217979750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957509, "dur": 14,
    "args": {
      "External id": 31280,
      "cbid": 211, "correlation": 31280
    }
  },
  {
    "ph": "s", "id": 31280, "pid": 76337, "tid": -914061504, "ts": 1716454217957509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217979990, "dur": 252,
    "args": {
      "External id": 31288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31288, "pid": 5, "tid": 7, "ts": 1716454217979990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957546, "dur": 8,
    "args": {
      "External id": 31288,
      "cbid": 211, "correlation": 31288
    }
  },
  {
    "ph": "s", "id": 31288, "pid": 76337, "tid": -914061504, "ts": 1716454217957546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957641, "dur": 1,
    "args": {
      "External id": 31304,
      "cbid": 251, "correlation": 31304
    }
  },
  {
    "ph": "f", "id": 31304, "pid": 76337, "tid": -914061504, "ts": 1716454217957641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957646, "dur": 0,
    "args": {
      "External id": 31306,
      "cbid": 251, "correlation": 31306
    }
  },
  {
    "ph": "f", "id": 31306, "pid": 76337, "tid": -914061504, "ts": 1716454217957646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454217980243, "dur": 358,
    "args": {
      "External id": 31307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31307,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 31307, "pid": 5, "tid": 7, "ts": 1716454217980243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957650, "dur": 14,
    "args": {
      "External id": 31307,
      "cbid": 211, "correlation": 31307
    }
  },
  {
    "ph": "s", "id": 31307, "pid": 76337, "tid": -914061504, "ts": 1716454217957650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217980603, "dur": 50,
    "args": {
      "External id": 31315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31315,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31315, "pid": 5, "tid": 7, "ts": 1716454217980603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957694, "dur": 10,
    "args": {
      "External id": 31315,
      "cbid": 211, "correlation": 31315
    }
  },
  {
    "ph": "s", "id": 31315, "pid": 76337, "tid": -914061504, "ts": 1716454217957694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217980654, "dur": 156,
    "args": {
      "External id": 31326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31326, "pid": 5, "tid": 7, "ts": 1716454217980654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957763, "dur": 12,
    "args": {
      "External id": 31326,
      "cbid": 211, "correlation": 31326
    }
  },
  {
    "ph": "s", "id": 31326, "pid": 76337, "tid": -914061504, "ts": 1716454217957763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957827, "dur": 0,
    "args": {
      "External id": 31338,
      "cbid": 317, "correlation": 31338
    }
  },
  {
    "ph": "f", "id": 31338, "pid": 76337, "tid": -914061504, "ts": 1716454217957827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957828, "dur": 0,
    "args": {
      "External id": 31339,
      "cbid": 203, "correlation": 31339
    }
  },
  {
    "ph": "f", "id": 31339, "pid": 76337, "tid": -914061504, "ts": 1716454217957828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957829, "dur": 0,
    "args": {
      "External id": 31340,
      "cbid": 205, "correlation": 31340
    }
  },
  {
    "ph": "f", "id": 31340, "pid": 76337, "tid": -914061504, "ts": 1716454217957829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957853, "dur": 1,
    "args": {
      "External id": 31344,
      "cbid": 251, "correlation": 31344
    }
  },
  {
    "ph": "f", "id": 31344, "pid": 76337, "tid": -914061504, "ts": 1716454217957853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957855, "dur": 0,
    "args": {
      "External id": 31345,
      "cbid": 251, "correlation": 31345
    }
  },
  {
    "ph": "f", "id": 31345, "pid": 76337, "tid": -914061504, "ts": 1716454217957855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957856, "dur": 0,
    "args": {
      "External id": 31346,
      "cbid": 251, "correlation": 31346
    }
  },
  {
    "ph": "f", "id": 31346, "pid": 76337, "tid": -914061504, "ts": 1716454217957856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957856, "dur": 0,
    "args": {
      "External id": 31347,
      "cbid": 251, "correlation": 31347
    }
  },
  {
    "ph": "f", "id": 31347, "pid": 76337, "tid": -914061504, "ts": 1716454217957856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957857, "dur": 0,
    "args": {
      "External id": 31348,
      "cbid": 251, "correlation": 31348
    }
  },
  {
    "ph": "f", "id": 31348, "pid": 76337, "tid": -914061504, "ts": 1716454217957857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957858, "dur": 0,
    "args": {
      "External id": 31349,
      "cbid": 251, "correlation": 31349
    }
  },
  {
    "ph": "f", "id": 31349, "pid": 76337, "tid": -914061504, "ts": 1716454217957858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957859, "dur": 0,
    "args": {
      "External id": 31350,
      "cbid": 251, "correlation": 31350
    }
  },
  {
    "ph": "f", "id": 31350, "pid": 76337, "tid": -914061504, "ts": 1716454217957859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957860, "dur": 0,
    "args": {
      "External id": 31351,
      "cbid": 251, "correlation": 31351
    }
  },
  {
    "ph": "f", "id": 31351, "pid": 76337, "tid": -914061504, "ts": 1716454217957860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957861, "dur": 0,
    "args": {
      "External id": 31352,
      "cbid": 251, "correlation": 31352
    }
  },
  {
    "ph": "f", "id": 31352, "pid": 76337, "tid": -914061504, "ts": 1716454217957861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217980812, "dur": 115,
    "args": {
      "External id": 31353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31353,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31353, "pid": 5, "tid": 7, "ts": 1716454217980812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957863, "dur": 12,
    "args": {
      "External id": 31353,
      "cbid": 211, "correlation": 31353
    }
  },
  {
    "ph": "s", "id": 31353, "pid": 76337, "tid": -914061504, "ts": 1716454217957863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217980928, "dur": 59,
    "args": {
      "External id": 31359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31359, "pid": 5, "tid": 7, "ts": 1716454217980928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957898, "dur": 10,
    "args": {
      "External id": 31359,
      "cbid": 211, "correlation": 31359
    }
  },
  {
    "ph": "s", "id": 31359, "pid": 76337, "tid": -914061504, "ts": 1716454217957898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217980988, "dur": 50,
    "args": {
      "External id": 31367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31367,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31367, "pid": 5, "tid": 7, "ts": 1716454217980988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957930, "dur": 8,
    "args": {
      "External id": 31367,
      "cbid": 211, "correlation": 31367
    }
  },
  {
    "ph": "s", "id": 31367, "pid": 76337, "tid": -914061504, "ts": 1716454217957930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217981039, "dur": 99,
    "args": {
      "External id": 31376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31376,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31376, "pid": 5, "tid": 7, "ts": 1716454217981039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217957969, "dur": 18,
    "args": {
      "External id": 31376,
      "cbid": 211, "correlation": 31376
    }
  },
  {
    "ph": "s", "id": 31376, "pid": 76337, "tid": -914061504, "ts": 1716454217957969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217981140, "dur": 92,
    "args": {
      "External id": 31396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31396,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 31396, "pid": 5, "tid": 7, "ts": 1716454217981140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958050, "dur": 12,
    "args": {
      "External id": 31396,
      "cbid": 211, "correlation": 31396
    }
  },
  {
    "ph": "s", "id": 31396, "pid": 76337, "tid": -914061504, "ts": 1716454217958050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217981234, "dur": 5,
    "args": {
      "External id": 31408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31408,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 31408, "pid": 5, "tid": 7, "ts": 1716454217981234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958071, "dur": 7,
    "args": {
      "External id": 31408,
      "cbid": 211, "correlation": 31408
    }
  },
  {
    "ph": "s", "id": 31408, "pid": 76337, "tid": -914061504, "ts": 1716454217958071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217981240, "dur": 107,
    "args": {
      "External id": 31411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31411, "pid": 5, "tid": 7, "ts": 1716454217981240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958090, "dur": 7,
    "args": {
      "External id": 31411,
      "cbid": 211, "correlation": 31411
    }
  },
  {
    "ph": "s", "id": 31411, "pid": 76337, "tid": -914061504, "ts": 1716454217958090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217981348, "dur": 68,
    "args": {
      "External id": 31420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31420,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31420, "pid": 5, "tid": 7, "ts": 1716454217981348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958128, "dur": 10,
    "args": {
      "External id": 31420,
      "cbid": 211, "correlation": 31420
    }
  },
  {
    "ph": "s", "id": 31420, "pid": 76337, "tid": -914061504, "ts": 1716454217958128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958193, "dur": 0,
    "args": {
      "External id": 31430,
      "cbid": 317, "correlation": 31430
    }
  },
  {
    "ph": "f", "id": 31430, "pid": 76337, "tid": -914061504, "ts": 1716454217958193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958194, "dur": 0,
    "args": {
      "External id": 31431,
      "cbid": 203, "correlation": 31431
    }
  },
  {
    "ph": "f", "id": 31431, "pid": 76337, "tid": -914061504, "ts": 1716454217958194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958195, "dur": 0,
    "args": {
      "External id": 31432,
      "cbid": 205, "correlation": 31432
    }
  },
  {
    "ph": "f", "id": 31432, "pid": 76337, "tid": -914061504, "ts": 1716454217958195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217981418, "dur": 75,
    "args": {
      "External id": 31436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31436,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31436, "pid": 5, "tid": 7, "ts": 1716454217981418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958210, "dur": 12,
    "args": {
      "External id": 31436,
      "cbid": 211, "correlation": 31436
    }
  },
  {
    "ph": "s", "id": 31436, "pid": 76337, "tid": -914061504, "ts": 1716454217958210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217981495, "dur": 24,
    "args": {
      "External id": 31438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31438, "pid": 5, "tid": 7, "ts": 1716454217981495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958225, "dur": 5,
    "args": {
      "External id": 31438,
      "cbid": 211, "correlation": 31438
    }
  },
  {
    "ph": "s", "id": 31438, "pid": 76337, "tid": -914061504, "ts": 1716454217958225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217981521, "dur": 3,
    "args": {
      "External id": 31440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 31440, "pid": 5, "tid": 7, "ts": 1716454217981521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958235, "dur": 6,
    "args": {
      "External id": 31440,
      "cbid": 211, "correlation": 31440
    }
  },
  {
    "ph": "s", "id": 31440, "pid": 76337, "tid": -914061504, "ts": 1716454217958235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958244, "dur": 0,
    "args": {
      "External id": 31441,
      "cbid": 51, "correlation": 31441
    }
  },
  {
    "ph": "s", "id": 31441, "pid": 76337, "tid": -914061504, "ts": 1716454217958244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217981526, "dur": 1345,
    "args": {
      "External id": 31442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31442,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31442, "pid": 5, "tid": 7, "ts": 1716454217981526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958245, "dur": 6,
    "args": {
      "External id": 31442,
      "cbid": 211, "correlation": 31442
    }
  },
  {
    "ph": "s", "id": 31442, "pid": 76337, "tid": -914061504, "ts": 1716454217958245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217982872, "dur": 60,
    "args": {
      "External id": 31447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31447, "pid": 5, "tid": 7, "ts": 1716454217982872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958273, "dur": 8,
    "args": {
      "External id": 31447,
      "cbid": 211, "correlation": 31447
    }
  },
  {
    "ph": "s", "id": 31447, "pid": 76337, "tid": -914061504, "ts": 1716454217958273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217982933, "dur": 3,
    "args": {
      "External id": 31455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31455,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 31455, "pid": 5, "tid": 7, "ts": 1716454217982933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958318, "dur": 9,
    "args": {
      "External id": 31455,
      "cbid": 211, "correlation": 31455
    }
  },
  {
    "ph": "s", "id": 31455, "pid": 76337, "tid": -914061504, "ts": 1716454217958318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958396, "dur": 1,
    "args": {
      "External id": 31471,
      "cbid": 251, "correlation": 31471
    }
  },
  {
    "ph": "f", "id": 31471, "pid": 76337, "tid": -914061504, "ts": 1716454217958396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958402, "dur": 0,
    "args": {
      "External id": 31473,
      "cbid": 251, "correlation": 31473
    }
  },
  {
    "ph": "f", "id": 31473, "pid": 76337, "tid": -914061504, "ts": 1716454217958402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217982937, "dur": 11,
    "args": {
      "External id": 31474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31474,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 31474, "pid": 5, "tid": 7, "ts": 1716454217982937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958403, "dur": 13,
    "args": {
      "External id": 31474,
      "cbid": 211, "correlation": 31474
    }
  },
  {
    "ph": "s", "id": 31474, "pid": 76337, "tid": -914061504, "ts": 1716454217958403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454217982950, "dur": 5,
    "args": {
      "External id": 31476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31476,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 31476, "pid": 5, "tid": 7, "ts": 1716454217982950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958418, "dur": 6,
    "args": {
      "External id": 31476,
      "cbid": 211, "correlation": 31476
    }
  },
  {
    "ph": "s", "id": 31476, "pid": 76337, "tid": -914061504, "ts": 1716454217958418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217982957, "dur": 54,
    "args": {
      "External id": 31486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31486, "pid": 5, "tid": 7, "ts": 1716454217982957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958479, "dur": 12,
    "args": {
      "External id": 31486,
      "cbid": 211, "correlation": 31486
    }
  },
  {
    "ph": "s", "id": 31486, "pid": 76337, "tid": -914061504, "ts": 1716454217958479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217983011, "dur": 52,
    "args": {
      "External id": 31506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31506,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 31506, "pid": 5, "tid": 7, "ts": 1716454217983011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958546, "dur": 11,
    "args": {
      "External id": 31506,
      "cbid": 211, "correlation": 31506
    }
  },
  {
    "ph": "s", "id": 31506, "pid": 76337, "tid": -914061504, "ts": 1716454217958546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217983065, "dur": 4,
    "args": {
      "External id": 31518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31518,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 31518, "pid": 5, "tid": 7, "ts": 1716454217983065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958567, "dur": 6,
    "args": {
      "External id": 31518,
      "cbid": 211, "correlation": 31518
    }
  },
  {
    "ph": "s", "id": 31518, "pid": 76337, "tid": -914061504, "ts": 1716454217958567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217983070, "dur": 54,
    "args": {
      "External id": 31521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31521, "pid": 5, "tid": 7, "ts": 1716454217983070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958585, "dur": 6,
    "args": {
      "External id": 31521,
      "cbid": 211, "correlation": 31521
    }
  },
  {
    "ph": "s", "id": 31521, "pid": 76337, "tid": -914061504, "ts": 1716454217958585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217983125, "dur": 36,
    "args": {
      "External id": 31530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31530,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31530, "pid": 5, "tid": 7, "ts": 1716454217983125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958625, "dur": 9,
    "args": {
      "External id": 31530,
      "cbid": 211, "correlation": 31530
    }
  },
  {
    "ph": "s", "id": 31530, "pid": 76337, "tid": -914061504, "ts": 1716454217958625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958688, "dur": 0,
    "args": {
      "External id": 31540,
      "cbid": 317, "correlation": 31540
    }
  },
  {
    "ph": "f", "id": 31540, "pid": 76337, "tid": -914061504, "ts": 1716454217958688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958688, "dur": 0,
    "args": {
      "External id": 31541,
      "cbid": 203, "correlation": 31541
    }
  },
  {
    "ph": "f", "id": 31541, "pid": 76337, "tid": -914061504, "ts": 1716454217958688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958689, "dur": 0,
    "args": {
      "External id": 31542,
      "cbid": 205, "correlation": 31542
    }
  },
  {
    "ph": "f", "id": 31542, "pid": 76337, "tid": -914061504, "ts": 1716454217958689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217983163, "dur": 40,
    "args": {
      "External id": 31546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31546, "pid": 5, "tid": 7, "ts": 1716454217983163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958704, "dur": 12,
    "args": {
      "External id": 31546,
      "cbid": 211, "correlation": 31546
    }
  },
  {
    "ph": "s", "id": 31546, "pid": 76337, "tid": -914061504, "ts": 1716454217958704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217983204, "dur": 14,
    "args": {
      "External id": 31548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31548,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31548, "pid": 5, "tid": 7, "ts": 1716454217983204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958718, "dur": 6,
    "args": {
      "External id": 31548,
      "cbid": 211, "correlation": 31548
    }
  },
  {
    "ph": "s", "id": 31548, "pid": 76337, "tid": -914061504, "ts": 1716454217958718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454217983220, "dur": 3,
    "args": {
      "External id": 31550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31550,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 31550, "pid": 5, "tid": 7, "ts": 1716454217983220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958729, "dur": 5,
    "args": {
      "External id": 31550,
      "cbid": 211, "correlation": 31550
    }
  },
  {
    "ph": "s", "id": 31550, "pid": 76337, "tid": -914061504, "ts": 1716454217958729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958737, "dur": 0,
    "args": {
      "External id": 31551,
      "cbid": 51, "correlation": 31551
    }
  },
  {
    "ph": "s", "id": 31551, "pid": 76337, "tid": -914061504, "ts": 1716454217958737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454217983225, "dur": 689,
    "args": {
      "External id": 31552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31552,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31552, "pid": 5, "tid": 7, "ts": 1716454217983225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958738, "dur": 5,
    "args": {
      "External id": 31552,
      "cbid": 211, "correlation": 31552
    }
  },
  {
    "ph": "s", "id": 31552, "pid": 76337, "tid": -914061504, "ts": 1716454217958738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217983915, "dur": 59,
    "args": {
      "External id": 31557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31557, "pid": 5, "tid": 7, "ts": 1716454217983915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958765, "dur": 8,
    "args": {
      "External id": 31557,
      "cbid": 211, "correlation": 31557
    }
  },
  {
    "ph": "s", "id": 31557, "pid": 76337, "tid": -914061504, "ts": 1716454217958765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958823, "dur": 0,
    "args": {
      "External id": 31567,
      "cbid": 317, "correlation": 31567
    }
  },
  {
    "ph": "f", "id": 31567, "pid": 76337, "tid": -914061504, "ts": 1716454217958823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958824, "dur": 0,
    "args": {
      "External id": 31568,
      "cbid": 203, "correlation": 31568
    }
  },
  {
    "ph": "f", "id": 31568, "pid": 76337, "tid": -914061504, "ts": 1716454217958824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958824, "dur": 0,
    "args": {
      "External id": 31569,
      "cbid": 205, "correlation": 31569
    }
  },
  {
    "ph": "f", "id": 31569, "pid": 76337, "tid": -914061504, "ts": 1716454217958824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217983975, "dur": 74,
    "args": {
      "External id": 31573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31573,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31573, "pid": 5, "tid": 7, "ts": 1716454217983975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958836, "dur": 12,
    "args": {
      "External id": 31573,
      "cbid": 211, "correlation": 31573
    }
  },
  {
    "ph": "s", "id": 31573, "pid": 76337, "tid": -914061504, "ts": 1716454217958836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454217984050, "dur": 209,
    "args": {
      "External id": 31575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31575,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31575, "pid": 5, "tid": 7, "ts": 1716454217984050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958855, "dur": 7,
    "args": {
      "External id": 31575,
      "cbid": 211, "correlation": 31575
    }
  },
  {
    "ph": "s", "id": 31575, "pid": 76337, "tid": -914061504, "ts": 1716454217958855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454217984260, "dur": 38,
    "args": {
      "External id": 31577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31577,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31577, "pid": 5, "tid": 7, "ts": 1716454217984260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958866, "dur": 5,
    "args": {
      "External id": 31577,
      "cbid": 211, "correlation": 31577
    }
  },
  {
    "ph": "s", "id": 31577, "pid": 76337, "tid": -914061504, "ts": 1716454217958866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217984300, "dur": 58,
    "args": {
      "External id": 31583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31583,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31583, "pid": 5, "tid": 7, "ts": 1716454217984300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958892, "dur": 9,
    "args": {
      "External id": 31583,
      "cbid": 211, "correlation": 31583
    }
  },
  {
    "ph": "s", "id": 31583, "pid": 76337, "tid": -914061504, "ts": 1716454217958892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454217984360, "dur": 50,
    "args": {
      "External id": 31591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31591,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31591, "pid": 5, "tid": 7, "ts": 1716454217984360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958921, "dur": 7,
    "args": {
      "External id": 31591,
      "cbid": 211, "correlation": 31591
    }
  },
  {
    "ph": "s", "id": 31591, "pid": 76337, "tid": -914061504, "ts": 1716454217958921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454217984412, "dur": 35,
    "args": {
      "External id": 31599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31599,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31599, "pid": 5, "tid": 7, "ts": 1716454217984412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217958949, "dur": 9,
    "args": {
      "External id": 31599,
      "cbid": 211, "correlation": 31599
    }
  },
  {
    "ph": "s", "id": 31599, "pid": 76337, "tid": -914061504, "ts": 1716454217958949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217984448, "dur": 51,
    "args": {
      "External id": 31619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31619,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 31619, "pid": 5, "tid": 7, "ts": 1716454217984448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959039, "dur": 13,
    "args": {
      "External id": 31619,
      "cbid": 211, "correlation": 31619
    }
  },
  {
    "ph": "s", "id": 31619, "pid": 76337, "tid": -914061504, "ts": 1716454217959039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454217984500, "dur": 5,
    "args": {
      "External id": 31631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31631,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 31631, "pid": 5, "tid": 7, "ts": 1716454217984500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959062, "dur": 6,
    "args": {
      "External id": 31631,
      "cbid": 211, "correlation": 31631
    }
  },
  {
    "ph": "s", "id": 31631, "pid": 76337, "tid": -914061504, "ts": 1716454217959062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217984506, "dur": 55,
    "args": {
      "External id": 31634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31634, "pid": 5, "tid": 7, "ts": 1716454217984506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959080, "dur": 6,
    "args": {
      "External id": 31634,
      "cbid": 211, "correlation": 31634
    }
  },
  {
    "ph": "s", "id": 31634, "pid": 76337, "tid": -914061504, "ts": 1716454217959080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959150, "dur": 0,
    "args": {
      "External id": 31645,
      "cbid": 317, "correlation": 31645
    }
  },
  {
    "ph": "f", "id": 31645, "pid": 76337, "tid": -914061504, "ts": 1716454217959150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959151, "dur": 0,
    "args": {
      "External id": 31646,
      "cbid": 203, "correlation": 31646
    }
  },
  {
    "ph": "f", "id": 31646, "pid": 76337, "tid": -914061504, "ts": 1716454217959151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959151, "dur": 0,
    "args": {
      "External id": 31647,
      "cbid": 205, "correlation": 31647
    }
  },
  {
    "ph": "f", "id": 31647, "pid": 76337, "tid": -914061504, "ts": 1716454217959151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959175, "dur": 1,
    "args": {
      "External id": 31651,
      "cbid": 251, "correlation": 31651
    }
  },
  {
    "ph": "f", "id": 31651, "pid": 76337, "tid": -914061504, "ts": 1716454217959175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959177, "dur": 0,
    "args": {
      "External id": 31652,
      "cbid": 251, "correlation": 31652
    }
  },
  {
    "ph": "f", "id": 31652, "pid": 76337, "tid": -914061504, "ts": 1716454217959177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959177, "dur": 0,
    "args": {
      "External id": 31653,
      "cbid": 251, "correlation": 31653
    }
  },
  {
    "ph": "f", "id": 31653, "pid": 76337, "tid": -914061504, "ts": 1716454217959177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959178, "dur": 0,
    "args": {
      "External id": 31654,
      "cbid": 251, "correlation": 31654
    }
  },
  {
    "ph": "f", "id": 31654, "pid": 76337, "tid": -914061504, "ts": 1716454217959178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959179, "dur": 0,
    "args": {
      "External id": 31655,
      "cbid": 251, "correlation": 31655
    }
  },
  {
    "ph": "f", "id": 31655, "pid": 76337, "tid": -914061504, "ts": 1716454217959179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959179, "dur": 0,
    "args": {
      "External id": 31656,
      "cbid": 251, "correlation": 31656
    }
  },
  {
    "ph": "f", "id": 31656, "pid": 76337, "tid": -914061504, "ts": 1716454217959179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959180, "dur": 0,
    "args": {
      "External id": 31657,
      "cbid": 251, "correlation": 31657
    }
  },
  {
    "ph": "f", "id": 31657, "pid": 76337, "tid": -914061504, "ts": 1716454217959180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959181, "dur": 0,
    "args": {
      "External id": 31658,
      "cbid": 251, "correlation": 31658
    }
  },
  {
    "ph": "f", "id": 31658, "pid": 76337, "tid": -914061504, "ts": 1716454217959181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959182, "dur": 0,
    "args": {
      "External id": 31659,
      "cbid": 251, "correlation": 31659
    }
  },
  {
    "ph": "f", "id": 31659, "pid": 76337, "tid": -914061504, "ts": 1716454217959182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454217984563, "dur": 114,
    "args": {
      "External id": 31660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31660,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31660, "pid": 5, "tid": 7, "ts": 1716454217984563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959184, "dur": 13,
    "args": {
      "External id": 31660,
      "cbid": 211, "correlation": 31660
    }
  },
  {
    "ph": "s", "id": 31660, "pid": 76337, "tid": -914061504, "ts": 1716454217959184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454217984679, "dur": 59,
    "args": {
      "External id": 31666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31666,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31666, "pid": 5, "tid": 7, "ts": 1716454217984679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959221, "dur": 9,
    "args": {
      "External id": 31666,
      "cbid": 211, "correlation": 31666
    }
  },
  {
    "ph": "s", "id": 31666, "pid": 76337, "tid": -914061504, "ts": 1716454217959221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454217984739, "dur": 563,
    "args": {
      "External id": 31675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31675,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31675, "pid": 5, "tid": 7, "ts": 1716454217984739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959305, "dur": 14,
    "args": {
      "External id": 31675,
      "cbid": 211, "correlation": 31675
    }
  },
  {
    "ph": "s", "id": 31675, "pid": 76337, "tid": -914061504, "ts": 1716454217959305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454217985304, "dur": 179,
    "args": {
      "External id": 31697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31697,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31697, "pid": 5, "tid": 7, "ts": 1716454217985304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959362, "dur": 10,
    "args": {
      "External id": 31697,
      "cbid": 211, "correlation": 31697
    }
  },
  {
    "ph": "s", "id": 31697, "pid": 76337, "tid": -914061504, "ts": 1716454217959362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959463, "dur": 1,
    "args": {
      "External id": 31708,
      "cbid": 251, "correlation": 31708
    }
  },
  {
    "ph": "f", "id": 31708, "pid": 76337, "tid": -914061504, "ts": 1716454217959463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217985485, "dur": 193,
    "args": {
      "External id": 31709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31709,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31709, "pid": 5, "tid": 7, "ts": 1716454217985485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959468, "dur": 15,
    "args": {
      "External id": 31709,
      "cbid": 211, "correlation": 31709
    }
  },
  {
    "ph": "s", "id": 31709, "pid": 76337, "tid": -914061504, "ts": 1716454217959468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959551, "dur": 1,
    "args": {
      "External id": 31720,
      "cbid": 251, "correlation": 31720
    }
  },
  {
    "ph": "f", "id": 31720, "pid": 76337, "tid": -914061504, "ts": 1716454217959551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217985680, "dur": 187,
    "args": {
      "External id": 31721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31721,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31721, "pid": 5, "tid": 7, "ts": 1716454217985680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959556, "dur": 12,
    "args": {
      "External id": 31721,
      "cbid": 211, "correlation": 31721
    }
  },
  {
    "ph": "s", "id": 31721, "pid": 76337, "tid": -914061504, "ts": 1716454217959556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959632, "dur": 1,
    "args": {
      "External id": 31732,
      "cbid": 251, "correlation": 31732
    }
  },
  {
    "ph": "f", "id": 31732, "pid": 76337, "tid": -914061504, "ts": 1716454217959632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454217985868, "dur": 185,
    "args": {
      "External id": 31733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31733,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31733, "pid": 5, "tid": 7, "ts": 1716454217985868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959636, "dur": 12,
    "args": {
      "External id": 31733,
      "cbid": 211, "correlation": 31733
    }
  },
  {
    "ph": "s", "id": 31733, "pid": 76337, "tid": -914061504, "ts": 1716454217959636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454217986054, "dur": 18297,
    "args": {
      "External id": 31754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31754,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31754, "pid": 5, "tid": 7, "ts": 1716454217986054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959719, "dur": 12,
    "args": {
      "External id": 31754,
      "cbid": 211, "correlation": 31754
    }
  },
  {
    "ph": "s", "id": 31754, "pid": 76337, "tid": -914061504, "ts": 1716454217959719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959828, "dur": 1,
    "args": {
      "External id": 31772,
      "cbid": 251, "correlation": 31772
    }
  },
  {
    "ph": "f", "id": 31772, "pid": 76337, "tid": -914061504, "ts": 1716454217959828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218004353, "dur": 201,
    "args": {
      "External id": 31774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31774,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31774, "pid": 5, "tid": 7, "ts": 1716454218004353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959834, "dur": 14,
    "args": {
      "External id": 31774,
      "cbid": 211, "correlation": 31774
    }
  },
  {
    "ph": "s", "id": 31774, "pid": 76337, "tid": -914061504, "ts": 1716454217959834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218004555, "dur": 66,
    "args": {
      "External id": 31782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31782,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31782, "pid": 5, "tid": 7, "ts": 1716454218004555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959907, "dur": 12,
    "args": {
      "External id": 31782,
      "cbid": 211, "correlation": 31782
    }
  },
  {
    "ph": "s", "id": 31782, "pid": 76337, "tid": -914061504, "ts": 1716454217959907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218004622, "dur": 96,
    "args": {
      "External id": 31790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31790,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31790, "pid": 5, "tid": 7, "ts": 1716454218004622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217959946, "dur": 9,
    "args": {
      "External id": 31790,
      "cbid": 211, "correlation": 31790
    }
  },
  {
    "ph": "s", "id": 31790, "pid": 76337, "tid": -914061504, "ts": 1716454217959946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218004720, "dur": 54,
    "args": {
      "External id": 31801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31801, "pid": 5, "tid": 7, "ts": 1716454218004720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960028, "dur": 13,
    "args": {
      "External id": 31801,
      "cbid": 211, "correlation": 31801
    }
  },
  {
    "ph": "s", "id": 31801, "pid": 76337, "tid": -914061504, "ts": 1716454217960028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218004775, "dur": 91,
    "args": {
      "External id": 31823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31823,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31823, "pid": 5, "tid": 7, "ts": 1716454218004775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960059, "dur": 7,
    "args": {
      "External id": 31823,
      "cbid": 211, "correlation": 31823
    }
  },
  {
    "ph": "s", "id": 31823, "pid": 76337, "tid": -914061504, "ts": 1716454217960059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960155, "dur": 1,
    "args": {
      "External id": 31834,
      "cbid": 251, "correlation": 31834
    }
  },
  {
    "ph": "f", "id": 31834, "pid": 76337, "tid": -914061504, "ts": 1716454217960155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218004868, "dur": 104,
    "args": {
      "External id": 31835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31835,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31835, "pid": 5, "tid": 7, "ts": 1716454218004868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960160, "dur": 14,
    "args": {
      "External id": 31835,
      "cbid": 211, "correlation": 31835
    }
  },
  {
    "ph": "s", "id": 31835, "pid": 76337, "tid": -914061504, "ts": 1716454217960160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960246, "dur": 1,
    "args": {
      "External id": 31846,
      "cbid": 251, "correlation": 31846
    }
  },
  {
    "ph": "f", "id": 31846, "pid": 76337, "tid": -914061504, "ts": 1716454217960246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960250, "dur": 0,
    "args": {
      "External id": 31847,
      "cbid": 251, "correlation": 31847
    }
  },
  {
    "ph": "f", "id": 31847, "pid": 76337, "tid": -914061504, "ts": 1716454217960250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218004973, "dur": 10,
    "args": {
      "External id": 31848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31848,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 31848, "pid": 5, "tid": 7, "ts": 1716454218004973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960252, "dur": 13,
    "args": {
      "External id": 31848,
      "cbid": 211, "correlation": 31848
    }
  },
  {
    "ph": "s", "id": 31848, "pid": 76337, "tid": -914061504, "ts": 1716454217960252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218004984, "dur": 5,
    "args": {
      "External id": 31850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31850,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 31850, "pid": 5, "tid": 7, "ts": 1716454218004984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960267, "dur": 6,
    "args": {
      "External id": 31850,
      "cbid": 211, "correlation": 31850
    }
  },
  {
    "ph": "s", "id": 31850, "pid": 76337, "tid": -914061504, "ts": 1716454217960267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960339, "dur": 1,
    "args": {
      "External id": 31861,
      "cbid": 251, "correlation": 31861
    }
  },
  {
    "ph": "f", "id": 31861, "pid": 76337, "tid": -914061504, "ts": 1716454217960339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960343, "dur": 0,
    "args": {
      "External id": 31862,
      "cbid": 251, "correlation": 31862
    }
  },
  {
    "ph": "f", "id": 31862, "pid": 76337, "tid": -914061504, "ts": 1716454217960343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218004990, "dur": 6,
    "args": {
      "External id": 31863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31863,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 31863, "pid": 5, "tid": 7, "ts": 1716454218004990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960345, "dur": 13,
    "args": {
      "External id": 31863,
      "cbid": 211, "correlation": 31863
    }
  },
  {
    "ph": "s", "id": 31863, "pid": 76337, "tid": -914061504, "ts": 1716454217960345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218004997, "dur": 3,
    "args": {
      "External id": 31865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31865,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 31865, "pid": 5, "tid": 7, "ts": 1716454218004997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960360, "dur": 6,
    "args": {
      "External id": 31865,
      "cbid": 211, "correlation": 31865
    }
  },
  {
    "ph": "s", "id": 31865, "pid": 76337, "tid": -914061504, "ts": 1716454217960360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218005002, "dur": 154,
    "args": {
      "External id": 31886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31886,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31886, "pid": 5, "tid": 7, "ts": 1716454218005002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960434, "dur": 12,
    "args": {
      "External id": 31886,
      "cbid": 211, "correlation": 31886
    }
  },
  {
    "ph": "s", "id": 31886, "pid": 76337, "tid": -914061504, "ts": 1716454217960434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960543, "dur": 1,
    "args": {
      "External id": 31904,
      "cbid": 251, "correlation": 31904
    }
  },
  {
    "ph": "f", "id": 31904, "pid": 76337, "tid": -914061504, "ts": 1716454217960543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218005157, "dur": 107,
    "args": {
      "External id": 31906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31906,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 31906, "pid": 5, "tid": 7, "ts": 1716454218005157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960549, "dur": 14,
    "args": {
      "External id": 31906,
      "cbid": 211, "correlation": 31906
    }
  },
  {
    "ph": "s", "id": 31906, "pid": 76337, "tid": -914061504, "ts": 1716454217960549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218005266, "dur": 34,
    "args": {
      "External id": 31914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31914,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31914, "pid": 5, "tid": 7, "ts": 1716454218005266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960621, "dur": 13,
    "args": {
      "External id": 31914,
      "cbid": 211, "correlation": 31914
    }
  },
  {
    "ph": "s", "id": 31914, "pid": 76337, "tid": -914061504, "ts": 1716454217960621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218005302, "dur": 66,
    "args": {
      "External id": 31922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31922, "pid": 5, "tid": 7, "ts": 1716454218005302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960663, "dur": 9,
    "args": {
      "External id": 31922,
      "cbid": 211, "correlation": 31922
    }
  },
  {
    "ph": "s", "id": 31922, "pid": 76337, "tid": -914061504, "ts": 1716454217960663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218005369, "dur": 91,
    "args": {
      "External id": 31944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31944,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31944, "pid": 5, "tid": 7, "ts": 1716454218005369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960714, "dur": 10,
    "args": {
      "External id": 31944,
      "cbid": 211, "correlation": 31944
    }
  },
  {
    "ph": "s", "id": 31944, "pid": 76337, "tid": -914061504, "ts": 1716454217960714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960813, "dur": 1,
    "args": {
      "External id": 31960,
      "cbid": 251, "correlation": 31960
    }
  },
  {
    "ph": "f", "id": 31960, "pid": 76337, "tid": -914061504, "ts": 1716454217960813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218005462, "dur": 568,
    "args": {
      "External id": 31962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31962,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 31962, "pid": 5, "tid": 7, "ts": 1716454218005462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960819, "dur": 13,
    "args": {
      "External id": 31962,
      "cbid": 211, "correlation": 31962
    }
  },
  {
    "ph": "s", "id": 31962, "pid": 76337, "tid": -914061504, "ts": 1716454217960819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218006031, "dur": 242,
    "args": {
      "External id": 31970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31970, "pid": 5, "tid": 7, "ts": 1716454218006031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960886, "dur": 13,
    "args": {
      "External id": 31970,
      "cbid": 211, "correlation": 31970
    }
  },
  {
    "ph": "s", "id": 31970, "pid": 76337, "tid": -914061504, "ts": 1716454217960886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218006275, "dur": 252,
    "args": {
      "External id": 31978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31978,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 31978, "pid": 5, "tid": 7, "ts": 1716454218006275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217960918, "dur": 8,
    "args": {
      "External id": 31978,
      "cbid": 211, "correlation": 31978
    }
  },
  {
    "ph": "s", "id": 31978, "pid": 76337, "tid": -914061504, "ts": 1716454217960918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961021, "dur": 2,
    "args": {
      "External id": 31994,
      "cbid": 251, "correlation": 31994
    }
  },
  {
    "ph": "f", "id": 31994, "pid": 76337, "tid": -914061504, "ts": 1716454217961021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961027, "dur": 0,
    "args": {
      "External id": 31996,
      "cbid": 251, "correlation": 31996
    }
  },
  {
    "ph": "f", "id": 31996, "pid": 76337, "tid": -914061504, "ts": 1716454217961027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218006528, "dur": 356,
    "args": {
      "External id": 31997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 31997,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 31997, "pid": 5, "tid": 7, "ts": 1716454218006528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961030, "dur": 14,
    "args": {
      "External id": 31997,
      "cbid": 211, "correlation": 31997
    }
  },
  {
    "ph": "s", "id": 31997, "pid": 76337, "tid": -914061504, "ts": 1716454217961030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218006886, "dur": 50,
    "args": {
      "External id": 32005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32005,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32005, "pid": 5, "tid": 7, "ts": 1716454218006886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961075, "dur": 10,
    "args": {
      "External id": 32005,
      "cbid": 211, "correlation": 32005
    }
  },
  {
    "ph": "s", "id": 32005, "pid": 76337, "tid": -914061504, "ts": 1716454217961075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218006937, "dur": 156,
    "args": {
      "External id": 32016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32016, "pid": 5, "tid": 7, "ts": 1716454218006937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961144, "dur": 13,
    "args": {
      "External id": 32016,
      "cbid": 211, "correlation": 32016
    }
  },
  {
    "ph": "s", "id": 32016, "pid": 76337, "tid": -914061504, "ts": 1716454217961144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961209, "dur": 0,
    "args": {
      "External id": 32028,
      "cbid": 317, "correlation": 32028
    }
  },
  {
    "ph": "f", "id": 32028, "pid": 76337, "tid": -914061504, "ts": 1716454217961209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961210, "dur": 0,
    "args": {
      "External id": 32029,
      "cbid": 203, "correlation": 32029
    }
  },
  {
    "ph": "f", "id": 32029, "pid": 76337, "tid": -914061504, "ts": 1716454217961210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961211, "dur": 0,
    "args": {
      "External id": 32030,
      "cbid": 205, "correlation": 32030
    }
  },
  {
    "ph": "f", "id": 32030, "pid": 76337, "tid": -914061504, "ts": 1716454217961211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961234, "dur": 1,
    "args": {
      "External id": 32034,
      "cbid": 251, "correlation": 32034
    }
  },
  {
    "ph": "f", "id": 32034, "pid": 76337, "tid": -914061504, "ts": 1716454217961234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961236, "dur": 0,
    "args": {
      "External id": 32035,
      "cbid": 251, "correlation": 32035
    }
  },
  {
    "ph": "f", "id": 32035, "pid": 76337, "tid": -914061504, "ts": 1716454217961236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961237, "dur": 0,
    "args": {
      "External id": 32036,
      "cbid": 251, "correlation": 32036
    }
  },
  {
    "ph": "f", "id": 32036, "pid": 76337, "tid": -914061504, "ts": 1716454217961237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961237, "dur": 0,
    "args": {
      "External id": 32037,
      "cbid": 251, "correlation": 32037
    }
  },
  {
    "ph": "f", "id": 32037, "pid": 76337, "tid": -914061504, "ts": 1716454217961237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961238, "dur": 0,
    "args": {
      "External id": 32038,
      "cbid": 251, "correlation": 32038
    }
  },
  {
    "ph": "f", "id": 32038, "pid": 76337, "tid": -914061504, "ts": 1716454217961238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961239, "dur": 0,
    "args": {
      "External id": 32039,
      "cbid": 251, "correlation": 32039
    }
  },
  {
    "ph": "f", "id": 32039, "pid": 76337, "tid": -914061504, "ts": 1716454217961239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961240, "dur": 0,
    "args": {
      "External id": 32040,
      "cbid": 251, "correlation": 32040
    }
  },
  {
    "ph": "f", "id": 32040, "pid": 76337, "tid": -914061504, "ts": 1716454217961240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961240, "dur": 0,
    "args": {
      "External id": 32041,
      "cbid": 251, "correlation": 32041
    }
  },
  {
    "ph": "f", "id": 32041, "pid": 76337, "tid": -914061504, "ts": 1716454217961240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961242, "dur": 0,
    "args": {
      "External id": 32042,
      "cbid": 251, "correlation": 32042
    }
  },
  {
    "ph": "f", "id": 32042, "pid": 76337, "tid": -914061504, "ts": 1716454217961242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218007094, "dur": 115,
    "args": {
      "External id": 32043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32043,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 32043, "pid": 5, "tid": 7, "ts": 1716454218007094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961244, "dur": 13,
    "args": {
      "External id": 32043,
      "cbid": 211, "correlation": 32043
    }
  },
  {
    "ph": "s", "id": 32043, "pid": 76337, "tid": -914061504, "ts": 1716454217961244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218007211, "dur": 59,
    "args": {
      "External id": 32049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32049, "pid": 5, "tid": 7, "ts": 1716454218007211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961279, "dur": 9,
    "args": {
      "External id": 32049,
      "cbid": 211, "correlation": 32049
    }
  },
  {
    "ph": "s", "id": 32049, "pid": 76337, "tid": -914061504, "ts": 1716454217961279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218007271, "dur": 50,
    "args": {
      "External id": 32057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32057,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32057, "pid": 5, "tid": 7, "ts": 1716454218007271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961311, "dur": 8,
    "args": {
      "External id": 32057,
      "cbid": 211, "correlation": 32057
    }
  },
  {
    "ph": "s", "id": 32057, "pid": 76337, "tid": -914061504, "ts": 1716454217961311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218007322, "dur": 53,
    "args": {
      "External id": 32077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32077,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 32077, "pid": 5, "tid": 7, "ts": 1716454218007322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961396, "dur": 13,
    "args": {
      "External id": 32077,
      "cbid": 211, "correlation": 32077
    }
  },
  {
    "ph": "s", "id": 32077, "pid": 76337, "tid": -914061504, "ts": 1716454217961396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218007376, "dur": 5,
    "args": {
      "External id": 32089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32089,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32089, "pid": 5, "tid": 7, "ts": 1716454218007376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961418, "dur": 6,
    "args": {
      "External id": 32089,
      "cbid": 211, "correlation": 32089
    }
  },
  {
    "ph": "s", "id": 32089, "pid": 76337, "tid": -914061504, "ts": 1716454217961418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218007382, "dur": 57,
    "args": {
      "External id": 32092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32092, "pid": 5, "tid": 7, "ts": 1716454218007382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961436, "dur": 7,
    "args": {
      "External id": 32092,
      "cbid": 211, "correlation": 32092
    }
  },
  {
    "ph": "s", "id": 32092, "pid": 76337, "tid": -914061504, "ts": 1716454217961436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218007440, "dur": 37,
    "args": {
      "External id": 32101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32101,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32101, "pid": 5, "tid": 7, "ts": 1716454218007440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961481, "dur": 10,
    "args": {
      "External id": 32101,
      "cbid": 211, "correlation": 32101
    }
  },
  {
    "ph": "s", "id": 32101, "pid": 76337, "tid": -914061504, "ts": 1716454217961481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961557, "dur": 0,
    "args": {
      "External id": 32111,
      "cbid": 317, "correlation": 32111
    }
  },
  {
    "ph": "f", "id": 32111, "pid": 76337, "tid": -914061504, "ts": 1716454217961557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961558, "dur": 0,
    "args": {
      "External id": 32112,
      "cbid": 203, "correlation": 32112
    }
  },
  {
    "ph": "f", "id": 32112, "pid": 76337, "tid": -914061504, "ts": 1716454217961558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961559, "dur": 0,
    "args": {
      "External id": 32113,
      "cbid": 205, "correlation": 32113
    }
  },
  {
    "ph": "f", "id": 32113, "pid": 76337, "tid": -914061504, "ts": 1716454217961559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218007478, "dur": 41,
    "args": {
      "External id": 32117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32117, "pid": 5, "tid": 7, "ts": 1716454218007478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961576, "dur": 13,
    "args": {
      "External id": 32117,
      "cbid": 211, "correlation": 32117
    }
  },
  {
    "ph": "s", "id": 32117, "pid": 76337, "tid": -914061504, "ts": 1716454217961576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218007520, "dur": 3,
    "args": {
      "External id": 32119,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32119,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.413793103448276
    }
  },
  {
    "ph": "f", "id": 32119, "pid": 5, "tid": 7, "ts": 1716454218007520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961592, "dur": 24,
    "args": {
      "External id": 32119,
      "cbid": 51, "correlation": 32119
    }
  },
  {
    "ph": "s", "id": 32119, "pid": 76337, "tid": -914061504, "ts": 1716454217961592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961622, "dur": 1,
    "args": {
      "External id": 32121,
      "cbid": 200, "correlation": 32121
    }
  },
  {
    "ph": "f", "id": 32121, "pid": 76337, "tid": -914061504, "ts": 1716454217961622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961624, "dur": 0,
    "args": {
      "External id": 32122,
      "cbid": 200, "correlation": 32122
    }
  },
  {
    "ph": "f", "id": 32122, "pid": 76337, "tid": -914061504, "ts": 1716454217961624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961625, "dur": 0,
    "args": {
      "External id": 32123,
      "cbid": 200, "correlation": 32123
    }
  },
  {
    "ph": "f", "id": 32123, "pid": 76337, "tid": -914061504, "ts": 1716454217961625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961626, "dur": 0,
    "args": {
      "External id": 32124,
      "cbid": 200, "correlation": 32124
    }
  },
  {
    "ph": "f", "id": 32124, "pid": 76337, "tid": -914061504, "ts": 1716454217961626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961627, "dur": 15,
    "args": {
      "External id": 32125,
      "cbid": 15, "correlation": 32125
    }
  },
  {
    "ph": "f", "id": 32125, "pid": 76337, "tid": -914061504, "ts": 1716454217961627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961643, "dur": 1,
    "args": {
      "External id": 32126,
      "cbid": 251, "correlation": 32126
    }
  },
  {
    "ph": "f", "id": 32126, "pid": 76337, "tid": -914061504, "ts": 1716454217961643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218007524, "dur": 25,
    "args": {
      "External id": 32127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32127,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32127, "pid": 5, "tid": 7, "ts": 1716454218007524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961651, "dur": 11,
    "args": {
      "External id": 32127,
      "cbid": 211, "correlation": 32127
    }
  },
  {
    "ph": "s", "id": 32127, "pid": 76337, "tid": -914061504, "ts": 1716454217961651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218007550, "dur": 4,
    "args": {
      "External id": 32129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32129, "pid": 5, "tid": 7, "ts": 1716454218007550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961668, "dur": 6,
    "args": {
      "External id": 32129,
      "cbid": 211, "correlation": 32129
    }
  },
  {
    "ph": "s", "id": 32129, "pid": 76337, "tid": -914061504, "ts": 1716454217961668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961678, "dur": 0,
    "args": {
      "External id": 32130,
      "cbid": 51, "correlation": 32130
    }
  },
  {
    "ph": "s", "id": 32130, "pid": 76337, "tid": -914061504, "ts": 1716454217961678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218007556, "dur": 188,
    "args": {
      "External id": 32131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32131,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32131, "pid": 5, "tid": 7, "ts": 1716454218007556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961679, "dur": 8,
    "args": {
      "External id": 32131,
      "cbid": 211, "correlation": 32131
    }
  },
  {
    "ph": "s", "id": 32131, "pid": 76337, "tid": -914061504, "ts": 1716454217961679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218007745, "dur": 6,
    "args": {
      "External id": 32132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32132, "pid": 5, "tid": 7, "ts": 1716454218007745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961690, "dur": 6,
    "args": {
      "External id": 32132,
      "cbid": 211, "correlation": 32132
    }
  },
  {
    "ph": "s", "id": 32132, "pid": 76337, "tid": -914061504, "ts": 1716454217961690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218007753, "dur": 5,
    "args": {
      "External id": 32138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 32138, "pid": 5, "tid": 7, "ts": 1716454218007753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217961721, "dur": 9,
    "args": {
      "External id": 32138,
      "cbid": 211, "correlation": 32138
    }
  },
  {
    "ph": "s", "id": 32138, "pid": 76337, "tid": -914061504, "ts": 1716454217961721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218007759, "dur": 3,
    "args": {
      "External id": 32146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32146,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32146, "pid": 5, "tid": 7, "ts": 1716454218007759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963489, "dur": 15,
    "args": {
      "External id": 32146,
      "cbid": 211, "correlation": 32146
    }
  },
  {
    "ph": "s", "id": 32146, "pid": 76337, "tid": -914061504, "ts": 1716454217963489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218007763, "dur": 3,
    "args": {
      "External id": 32154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32154,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32154, "pid": 5, "tid": 7, "ts": 1716454218007763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963530, "dur": 10,
    "args": {
      "External id": 32154,
      "cbid": 211, "correlation": 32154
    }
  },
  {
    "ph": "s", "id": 32154, "pid": 76337, "tid": -914061504, "ts": 1716454217963530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218007767, "dur": 3,
    "args": {
      "External id": 32162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32162,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32162, "pid": 5, "tid": 7, "ts": 1716454218007767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963559, "dur": 8,
    "args": {
      "External id": 32162,
      "cbid": 211, "correlation": 32162
    }
  },
  {
    "ph": "s", "id": 32162, "pid": 76337, "tid": -914061504, "ts": 1716454217963559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218007778, "dur": 3,
    "args": {
      "External id": 32170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32170,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32170, "pid": 5, "tid": 7, "ts": 1716454218007778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963675, "dur": 163,
    "args": {
      "External id": 32170,
      "cbid": 211, "correlation": 32170
    }
  },
  {
    "ph": "s", "id": 32170, "pid": 76337, "tid": -914061504, "ts": 1716454217963675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceCount", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963903, "dur": 0,
    "args": {
      "External id": 32182,
      "cbid": 3, "correlation": 32182
    }
  },
  {
    "ph": "f", "id": 32182, "pid": 76337, "tid": -914061504, "ts": 1716454217963903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963905, "dur": 79,
    "args": {
      "External id": 32186,
      "cbid": 15, "correlation": 32186
    }
  },
  {
    "ph": "f", "id": 32186, "pid": 76337, "tid": -914061504, "ts": 1716454217963905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454218007793, "dur": 3,
    "args": {
      "External id": 32196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32196,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32196, "pid": 5, "tid": 7, "ts": 1716454218007793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454217963997, "dur": 56,
    "args": {
      "External id": 32196,
      "cbid": 211, "correlation": 32196
    }
  },
  {
    "ph": "s", "id": 32196, "pid": 76337, "tid": -914061504, "ts": 1716454217963997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454217964054, "dur": 0,
    "args": {
      "External id": 32197,
      "cbid": 11, "correlation": 32197
    }
  },
  {
    "ph": "f", "id": 32197, "pid": 76337, "tid": -914061504, "ts": 1716454217964054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454217964054, "dur": 0,
    "args": {
      "External id": 32198,
      "cbid": 11, "correlation": 32198
    }
  },
  {
    "ph": "f", "id": 32198, "pid": 76337, "tid": -914061504, "ts": 1716454217964054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218007800, "dur": 3,
    "args": {
      "External id": 32201,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32201,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0010869565217391304
    }
  },
  {
    "ph": "f", "id": 32201, "pid": 5, "tid": 7, "ts": 1716454218007800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454217964056, "dur": 43762,
    "args": {
      "External id": 32201,
      "cbid": 41, "correlation": 32201
    }
  },
  {
    "ph": "s", "id": 32201, "pid": 76337, "tid": -914061504, "ts": 1716454217964056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007820, "dur": 11,
    "args": {
      "External id": 32202,
      "cbid": 131, "correlation": 32202
    }
  },
  {
    "ph": "f", "id": 32202, "pid": 76337, "tid": -914061504, "ts": 1716454218007820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218007913, "dur": 2,
    "args": {
      "External id": 32226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32226, "pid": 5, "tid": 7, "ts": 1716454218007913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007869, "dur": 45,
    "args": {
      "External id": 32226,
      "cbid": 211, "correlation": 32226
    }
  },
  {
    "ph": "s", "id": 32226, "pid": 76337, "tid": -914061504, "ts": 1716454218007869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007914, "dur": 0,
    "args": {
      "External id": 32227,
      "cbid": 11, "correlation": 32227
    }
  },
  {
    "ph": "f", "id": 32227, "pid": 76337, "tid": -914061504, "ts": 1716454218007914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007915, "dur": 0,
    "args": {
      "External id": 32228,
      "cbid": 11, "correlation": 32228
    }
  },
  {
    "ph": "f", "id": 32228, "pid": 76337, "tid": -914061504, "ts": 1716454218007915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007916, "dur": 0,
    "args": {
      "External id": 32230,
      "cbid": 200, "correlation": 32230
    }
  },
  {
    "ph": "f", "id": 32230, "pid": 76337, "tid": -914061504, "ts": 1716454218007916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454218007972, "dur": 3,
    "args": {
      "External id": 32232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32232,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32232, "pid": 5, "tid": 7, "ts": 1716454218007972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007918, "dur": 62,
    "args": {
      "External id": 32232,
      "cbid": 211, "correlation": 32232
    }
  },
  {
    "ph": "s", "id": 32232, "pid": 76337, "tid": -914061504, "ts": 1716454218007918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007981, "dur": 0,
    "args": {
      "External id": 32233,
      "cbid": 11, "correlation": 32233
    }
  },
  {
    "ph": "f", "id": 32233, "pid": 76337, "tid": -914061504, "ts": 1716454218007981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218007982, "dur": 0,
    "args": {
      "External id": 32234,
      "cbid": 11, "correlation": 32234
    }
  },
  {
    "ph": "f", "id": 32234, "pid": 76337, "tid": -914061504, "ts": 1716454218007982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218008122, "dur": 2,
    "args": {
      "External id": 32243,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32243,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0035714285714285713
    }
  },
  {
    "ph": "f", "id": 32243, "pid": 5, "tid": 7, "ts": 1716454218008122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008095, "dur": 37,
    "args": {
      "External id": 32243,
      "cbid": 41, "correlation": 32243
    }
  },
  {
    "ph": "s", "id": 32243, "pid": 76337, "tid": -914061504, "ts": 1716454218008095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008133, "dur": 4,
    "args": {
      "External id": 32244,
      "cbid": 131, "correlation": 32244
    }
  },
  {
    "ph": "f", "id": 32244, "pid": 76337, "tid": -914061504, "ts": 1716454218008133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218008387, "dur": 3,
    "args": {
      "External id": 32252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32252,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32252, "pid": 5, "tid": 7, "ts": 1716454218008387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008370, "dur": 16,
    "args": {
      "External id": 32252,
      "cbid": 211, "correlation": 32252
    }
  },
  {
    "ph": "s", "id": 32252, "pid": 76337, "tid": -914061504, "ts": 1716454218008370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218008415, "dur": 4,
    "args": {
      "External id": 32261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32261,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32261, "pid": 5, "tid": 7, "ts": 1716454218008415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008406, "dur": 8,
    "args": {
      "External id": 32261,
      "cbid": 211, "correlation": 32261
    }
  },
  {
    "ph": "s", "id": 32261, "pid": 76337, "tid": -914061504, "ts": 1716454218008406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218008445, "dur": 3,
    "args": {
      "External id": 32269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32269,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32269, "pid": 5, "tid": 7, "ts": 1716454218008445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008435, "dur": 9,
    "args": {
      "External id": 32269,
      "cbid": 211, "correlation": 32269
    }
  },
  {
    "ph": "s", "id": 32269, "pid": 76337, "tid": -914061504, "ts": 1716454218008435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218008733, "dur": 1,
    "args": {
      "External id": 32279,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32279,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0021551724137931034
    }
  },
  {
    "ph": "f", "id": 32279, "pid": 5, "tid": 7, "ts": 1716454218008733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008705, "dur": 26,
    "args": {
      "External id": 32279,
      "cbid": 41, "correlation": 32279
    }
  },
  {
    "ph": "s", "id": 32279, "pid": 76337, "tid": -914061504, "ts": 1716454218008705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008732, "dur": 9,
    "args": {
      "External id": 32280,
      "cbid": 131, "correlation": 32280
    }
  },
  {
    "ph": "f", "id": 32280, "pid": 76337, "tid": -914061504, "ts": 1716454218008732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454218009014, "dur": 3,
    "args": {
      "External id": 32288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32288, "pid": 5, "tid": 7, "ts": 1716454218009014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218008995, "dur": 19,
    "args": {
      "External id": 32288,
      "cbid": 211, "correlation": 32288
    }
  },
  {
    "ph": "s", "id": 32288, "pid": 76337, "tid": -914061504, "ts": 1716454218008995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218009137, "dur": 3,
    "args": {
      "External id": 32297,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32297,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0011467889908256881
    }
  },
  {
    "ph": "f", "id": 32297, "pid": 5, "tid": 7, "ts": 1716454218009137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009111, "dur": 26,
    "args": {
      "External id": 32297,
      "cbid": 41, "correlation": 32297
    }
  },
  {
    "ph": "s", "id": 32297, "pid": 76337, "tid": -914061504, "ts": 1716454218009111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218009212, "dur": 1,
    "args": {
      "External id": 32308,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32308,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 32308, "pid": 5, "tid": 7, "ts": 1716454218009212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009184, "dur": 26,
    "args": {
      "External id": 32308,
      "cbid": 41, "correlation": 32308
    }
  },
  {
    "ph": "s", "id": 32308, "pid": 76337, "tid": -914061504, "ts": 1716454218009184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009211, "dur": 8,
    "args": {
      "External id": 32309,
      "cbid": 131, "correlation": 32309
    }
  },
  {
    "ph": "f", "id": 32309, "pid": 76337, "tid": -914061504, "ts": 1716454218009211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218009268, "dur": 3,
    "args": {
      "External id": 32317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32317,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32317, "pid": 5, "tid": 7, "ts": 1716454218009268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009254, "dur": 14,
    "args": {
      "External id": 32317,
      "cbid": 211, "correlation": 32317
    }
  },
  {
    "ph": "s", "id": 32317, "pid": 76337, "tid": -914061504, "ts": 1716454218009254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218009298, "dur": 3,
    "args": {
      "External id": 32327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32327,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32327, "pid": 5, "tid": 7, "ts": 1716454218009298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009289, "dur": 8,
    "args": {
      "External id": 32327,
      "cbid": 211, "correlation": 32327
    }
  },
  {
    "ph": "s", "id": 32327, "pid": 76337, "tid": -914061504, "ts": 1716454218009289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218009321, "dur": 3,
    "args": {
      "External id": 32336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32336,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32336, "pid": 5, "tid": 7, "ts": 1716454218009321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009312, "dur": 7,
    "args": {
      "External id": 32336,
      "cbid": 211, "correlation": 32336
    }
  },
  {
    "ph": "s", "id": 32336, "pid": 76337, "tid": -914061504, "ts": 1716454218009312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218009422, "dur": 3,
    "args": {
      "External id": 32344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32344,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32344, "pid": 5, "tid": 7, "ts": 1716454218009422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009360, "dur": 62,
    "args": {
      "External id": 32344,
      "cbid": 211, "correlation": 32344
    }
  },
  {
    "ph": "s", "id": 32344, "pid": 76337, "tid": -914061504, "ts": 1716454218009360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218009495, "dur": 3,
    "args": {
      "External id": 32352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32352,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32352, "pid": 5, "tid": 7, "ts": 1716454218009495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009483, "dur": 12,
    "args": {
      "External id": 32352,
      "cbid": 211, "correlation": 32352
    }
  },
  {
    "ph": "s", "id": 32352, "pid": 76337, "tid": -914061504, "ts": 1716454218009483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218009555, "dur": 2,
    "args": {
      "External id": 32360,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32360,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00390625
    }
  },
  {
    "ph": "f", "id": 32360, "pid": 5, "tid": 7, "ts": 1716454218009555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009538, "dur": 27,
    "args": {
      "External id": 32360,
      "cbid": 41, "correlation": 32360
    }
  },
  {
    "ph": "s", "id": 32360, "pid": 76337, "tid": -914061504, "ts": 1716454218009538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009566, "dur": 3,
    "args": {
      "External id": 32361,
      "cbid": 131, "correlation": 32361
    }
  },
  {
    "ph": "f", "id": 32361, "pid": 76337, "tid": -914061504, "ts": 1716454218009566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218009648, "dur": 1,
    "args": {
      "External id": 32371,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32371,
      "bytes": 42, "memory bandwidth (GB/s)": 0.027925531914893616
    }
  },
  {
    "ph": "f", "id": 32371, "pid": 5, "tid": 7, "ts": 1716454218009648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009634, "dur": 12,
    "args": {
      "External id": 32371,
      "cbid": 41, "correlation": 32371
    }
  },
  {
    "ph": "s", "id": 32371, "pid": 76337, "tid": -914061504, "ts": 1716454218009634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009646, "dur": 8,
    "args": {
      "External id": 32372,
      "cbid": 131, "correlation": 32372
    }
  },
  {
    "ph": "f", "id": 32372, "pid": 76337, "tid": -914061504, "ts": 1716454218009646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218009696, "dur": 1,
    "args": {
      "External id": 32381,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32381,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 32381, "pid": 5, "tid": 7, "ts": 1716454218009696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009686, "dur": 8,
    "args": {
      "External id": 32381,
      "cbid": 41, "correlation": 32381
    }
  },
  {
    "ph": "s", "id": 32381, "pid": 76337, "tid": -914061504, "ts": 1716454218009686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009695, "dur": 8,
    "args": {
      "External id": 32382,
      "cbid": 131, "correlation": 32382
    }
  },
  {
    "ph": "f", "id": 32382, "pid": 76337, "tid": -914061504, "ts": 1716454218009695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218009760, "dur": 3,
    "args": {
      "External id": 32389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32389, "pid": 5, "tid": 7, "ts": 1716454218009760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009744, "dur": 16,
    "args": {
      "External id": 32389,
      "cbid": 211, "correlation": 32389
    }
  },
  {
    "ph": "s", "id": 32389, "pid": 76337, "tid": -914061504, "ts": 1716454218009744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454218009787, "dur": 3,
    "args": {
      "External id": 32409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32409,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32409, "pid": 5, "tid": 7, "ts": 1716454218009787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009778, "dur": 8,
    "args": {
      "External id": 32409,
      "cbid": 211, "correlation": 32409
    }
  },
  {
    "ph": "s", "id": 32409, "pid": 76337, "tid": -914061504, "ts": 1716454218009778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009787, "dur": 0,
    "args": {
      "External id": 32410,
      "cbid": 11, "correlation": 32410
    }
  },
  {
    "ph": "f", "id": 32410, "pid": 76337, "tid": -914061504, "ts": 1716454218009787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009787, "dur": 0,
    "args": {
      "External id": 32411,
      "cbid": 11, "correlation": 32411
    }
  },
  {
    "ph": "f", "id": 32411, "pid": 76337, "tid": -914061504, "ts": 1716454218009787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218009813, "dur": 1,
    "args": {
      "External id": 32414,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32414,
      "bytes": 4, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 32414, "pid": 5, "tid": 7, "ts": 1716454218009813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009789, "dur": 33,
    "args": {
      "External id": 32414,
      "cbid": 41, "correlation": 32414
    }
  },
  {
    "ph": "s", "id": 32414, "pid": 76337, "tid": -914061504, "ts": 1716454218009789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009823, "dur": 3,
    "args": {
      "External id": 32415,
      "cbid": 131, "correlation": 32415
    }
  },
  {
    "ph": "f", "id": 32415, "pid": 76337, "tid": -914061504, "ts": 1716454218009823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218009858, "dur": 2,
    "args": {
      "External id": 32439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32439,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32439, "pid": 5, "tid": 7, "ts": 1716454218009858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009848, "dur": 9,
    "args": {
      "External id": 32439,
      "cbid": 211, "correlation": 32439
    }
  },
  {
    "ph": "s", "id": 32439, "pid": 76337, "tid": -914061504, "ts": 1716454218009848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009858, "dur": 0,
    "args": {
      "External id": 32440,
      "cbid": 11, "correlation": 32440
    }
  },
  {
    "ph": "f", "id": 32440, "pid": 76337, "tid": -914061504, "ts": 1716454218009858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009858, "dur": 0,
    "args": {
      "External id": 32441,
      "cbid": 11, "correlation": 32441
    }
  },
  {
    "ph": "f", "id": 32441, "pid": 76337, "tid": -914061504, "ts": 1716454218009858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009859, "dur": 0,
    "args": {
      "External id": 32443,
      "cbid": 200, "correlation": 32443
    }
  },
  {
    "ph": "f", "id": 32443, "pid": 76337, "tid": -914061504, "ts": 1716454218009859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454218009868, "dur": 3,
    "args": {
      "External id": 32445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32445,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32445, "pid": 5, "tid": 7, "ts": 1716454218009868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009861, "dur": 7,
    "args": {
      "External id": 32445,
      "cbid": 211, "correlation": 32445
    }
  },
  {
    "ph": "s", "id": 32445, "pid": 76337, "tid": -914061504, "ts": 1716454218009861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009868, "dur": 0,
    "args": {
      "External id": 32446,
      "cbid": 11, "correlation": 32446
    }
  },
  {
    "ph": "f", "id": 32446, "pid": 76337, "tid": -914061504, "ts": 1716454218009868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009869, "dur": 0,
    "args": {
      "External id": 32447,
      "cbid": 11, "correlation": 32447
    }
  },
  {
    "ph": "f", "id": 32447, "pid": 76337, "tid": -914061504, "ts": 1716454218009869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218009902, "dur": 2,
    "args": {
      "External id": 32454,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32454,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0038461538461538464
    }
  },
  {
    "ph": "f", "id": 32454, "pid": 5, "tid": 7, "ts": 1716454218009902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009888, "dur": 23,
    "args": {
      "External id": 32454,
      "cbid": 41, "correlation": 32454
    }
  },
  {
    "ph": "s", "id": 32454, "pid": 76337, "tid": -914061504, "ts": 1716454218009888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009912, "dur": 3,
    "args": {
      "External id": 32455,
      "cbid": 131, "correlation": 32455
    }
  },
  {
    "ph": "f", "id": 32455, "pid": 76337, "tid": -914061504, "ts": 1716454218009912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218009963, "dur": 1,
    "args": {
      "External id": 32465,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 32465,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 32465, "pid": 5, "tid": 7, "ts": 1716454218009963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009950, "dur": 10,
    "args": {
      "External id": 32465,
      "cbid": 41, "correlation": 32465
    }
  },
  {
    "ph": "s", "id": 32465, "pid": 76337, "tid": -914061504, "ts": 1716454218009950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218009961, "dur": 8,
    "args": {
      "External id": 32466,
      "cbid": 131, "correlation": 32466
    }
  },
  {
    "ph": "f", "id": 32466, "pid": 76337, "tid": -914061504, "ts": 1716454218009961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218010283, "dur": 4,
    "args": {
      "External id": 32473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32473,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32473, "pid": 5, "tid": 7, "ts": 1716454218010283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010034, "dur": 251,
    "args": {
      "External id": 32473,
      "cbid": 211, "correlation": 32473
    }
  },
  {
    "ph": "s", "id": 32473, "pid": 76337, "tid": -914061504, "ts": 1716454218010034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218010451, "dur": 3,
    "args": {
      "External id": 32482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32482,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32482, "pid": 5, "tid": 7, "ts": 1716454218010451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010395, "dur": 57,
    "args": {
      "External id": 32482,
      "cbid": 211, "correlation": 32482
    }
  },
  {
    "ph": "s", "id": 32482, "pid": 76337, "tid": -914061504, "ts": 1716454218010395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218010492, "dur": 3,
    "args": {
      "External id": 32490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32490,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32490, "pid": 5, "tid": 7, "ts": 1716454218010492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010481, "dur": 10,
    "args": {
      "External id": 32490,
      "cbid": 211, "correlation": 32490
    }
  },
  {
    "ph": "s", "id": 32490, "pid": 76337, "tid": -914061504, "ts": 1716454218010481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218010585, "dur": 3,
    "args": {
      "External id": 32498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32498,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32498, "pid": 5, "tid": 7, "ts": 1716454218010585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010528, "dur": 58,
    "args": {
      "External id": 32498,
      "cbid": 211, "correlation": 32498
    }
  },
  {
    "ph": "s", "id": 32498, "pid": 76337, "tid": -914061504, "ts": 1716454218010528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218010687, "dur": 3,
    "args": {
      "External id": 32506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32506,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32506, "pid": 5, "tid": 7, "ts": 1716454218010687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010628, "dur": 59,
    "args": {
      "External id": 32506,
      "cbid": 211, "correlation": 32506
    }
  },
  {
    "ph": "s", "id": 32506, "pid": 76337, "tid": -914061504, "ts": 1716454218010628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218010724, "dur": 3,
    "args": {
      "External id": 32514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32514,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32514, "pid": 5, "tid": 7, "ts": 1716454218010724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010713, "dur": 10,
    "args": {
      "External id": 32514,
      "cbid": 211, "correlation": 32514
    }
  },
  {
    "ph": "s", "id": 32514, "pid": 76337, "tid": -914061504, "ts": 1716454218010713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218010791, "dur": 3,
    "args": {
      "External id": 32522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32522,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32522, "pid": 5, "tid": 7, "ts": 1716454218010791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010742, "dur": 49,
    "args": {
      "External id": 32522,
      "cbid": 211, "correlation": 32522
    }
  },
  {
    "ph": "s", "id": 32522, "pid": 76337, "tid": -914061504, "ts": 1716454218010742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218010823, "dur": 6,
    "args": {
      "External id": 32530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32530, "pid": 5, "tid": 7, "ts": 1716454218010823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010813, "dur": 10,
    "args": {
      "External id": 32530,
      "cbid": 211, "correlation": 32530
    }
  },
  {
    "ph": "s", "id": 32530, "pid": 76337, "tid": -914061504, "ts": 1716454218010813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218010844, "dur": 5,
    "args": {
      "External id": 32538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32538, "pid": 5, "tid": 7, "ts": 1716454218010844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010835, "dur": 7,
    "args": {
      "External id": 32538,
      "cbid": 211, "correlation": 32538
    }
  },
  {
    "ph": "s", "id": 32538, "pid": 76337, "tid": -914061504, "ts": 1716454218010835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218010864, "dur": 3,
    "args": {
      "External id": 32546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32546,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32546, "pid": 5, "tid": 7, "ts": 1716454218010864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010856, "dur": 7,
    "args": {
      "External id": 32546,
      "cbid": 211, "correlation": 32546
    }
  },
  {
    "ph": "s", "id": 32546, "pid": 76337, "tid": -914061504, "ts": 1716454218010856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218010990, "dur": 3,
    "args": {
      "External id": 32554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32554,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32554, "pid": 5, "tid": 7, "ts": 1716454218010990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218010925, "dur": 65,
    "args": {
      "External id": 32554,
      "cbid": 211, "correlation": 32554
    }
  },
  {
    "ph": "s", "id": 32554, "pid": 76337, "tid": -914061504, "ts": 1716454218010925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218011027, "dur": 5,
    "args": {
      "External id": 32562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32562,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32562, "pid": 5, "tid": 7, "ts": 1716454218011027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011016, "dur": 10,
    "args": {
      "External id": 32562,
      "cbid": 211, "correlation": 32562
    }
  },
  {
    "ph": "s", "id": 32562, "pid": 76337, "tid": -914061504, "ts": 1716454218011016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218011051, "dur": 4,
    "args": {
      "External id": 32570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32570, "pid": 5, "tid": 7, "ts": 1716454218011051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011042, "dur": 8,
    "args": {
      "External id": 32570,
      "cbid": 211, "correlation": 32570
    }
  },
  {
    "ph": "s", "id": 32570, "pid": 76337, "tid": -914061504, "ts": 1716454218011042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218011072, "dur": 3,
    "args": {
      "External id": 32578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32578,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 32578, "pid": 5, "tid": 7, "ts": 1716454218011072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011063, "dur": 7,
    "args": {
      "External id": 32578,
      "cbid": 211, "correlation": 32578
    }
  },
  {
    "ph": "s", "id": 32578, "pid": 76337, "tid": -914061504, "ts": 1716454218011063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218011452, "dur": 5,
    "args": {
      "External id": 32587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32587,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32587, "pid": 5, "tid": 7, "ts": 1716454218011452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011434, "dur": 19,
    "args": {
      "External id": 32587,
      "cbid": 211, "correlation": 32587
    }
  },
  {
    "ph": "s", "id": 32587, "pid": 76337, "tid": -914061504, "ts": 1716454218011434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218011489, "dur": 5,
    "args": {
      "External id": 32596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32596,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32596, "pid": 5, "tid": 7, "ts": 1716454218011489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011479, "dur": 9,
    "args": {
      "External id": 32596,
      "cbid": 211, "correlation": 32596
    }
  },
  {
    "ph": "s", "id": 32596, "pid": 76337, "tid": -914061504, "ts": 1716454218011479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454218011630, "dur": 3,
    "args": {
      "External id": 32612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32612, "pid": 5, "tid": 7, "ts": 1716454218011630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011615, "dur": 16,
    "args": {
      "External id": 32612,
      "cbid": 211, "correlation": 32612
    }
  },
  {
    "ph": "s", "id": 32612, "pid": 76337, "tid": -914061504, "ts": 1716454218011615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218011667, "dur": 3,
    "args": {
      "External id": 32620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32620,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32620, "pid": 5, "tid": 7, "ts": 1716454218011667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011657, "dur": 9,
    "args": {
      "External id": 32620,
      "cbid": 211, "correlation": 32620
    }
  },
  {
    "ph": "s", "id": 32620, "pid": 76337, "tid": -914061504, "ts": 1716454218011657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218011699, "dur": 3,
    "args": {
      "External id": 32628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32628, "pid": 5, "tid": 7, "ts": 1716454218011699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011689, "dur": 9,
    "args": {
      "External id": 32628,
      "cbid": 211, "correlation": 32628
    }
  },
  {
    "ph": "s", "id": 32628, "pid": 76337, "tid": -914061504, "ts": 1716454218011689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218011731, "dur": 4,
    "args": {
      "External id": 32636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32636,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32636, "pid": 5, "tid": 7, "ts": 1716454218011731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011721, "dur": 9,
    "args": {
      "External id": 32636,
      "cbid": 211, "correlation": 32636
    }
  },
  {
    "ph": "s", "id": 32636, "pid": 76337, "tid": -914061504, "ts": 1716454218011721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454218011789, "dur": 4,
    "args": {
      "External id": 32648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32648,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32648, "pid": 5, "tid": 7, "ts": 1716454218011789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011776, "dur": 13,
    "args": {
      "External id": 32648,
      "cbid": 211, "correlation": 32648
    }
  },
  {
    "ph": "s", "id": 32648, "pid": 76337, "tid": -914061504, "ts": 1716454218011776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218011836, "dur": 4,
    "args": {
      "External id": 32659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32659,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32659, "pid": 5, "tid": 7, "ts": 1716454218011836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011824, "dur": 12,
    "args": {
      "External id": 32659,
      "cbid": 211, "correlation": 32659
    }
  },
  {
    "ph": "s", "id": 32659, "pid": 76337, "tid": -914061504, "ts": 1716454218011824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218011867, "dur": 3,
    "args": {
      "External id": 32667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32667, "pid": 5, "tid": 7, "ts": 1716454218011867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011857, "dur": 9,
    "args": {
      "External id": 32667,
      "cbid": 211, "correlation": 32667
    }
  },
  {
    "ph": "s", "id": 32667, "pid": 76337, "tid": -914061504, "ts": 1716454218011857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218011898, "dur": 5,
    "args": {
      "External id": 32675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32675,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32675, "pid": 5, "tid": 7, "ts": 1716454218011898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011888, "dur": 9,
    "args": {
      "External id": 32675,
      "cbid": 211, "correlation": 32675
    }
  },
  {
    "ph": "s", "id": 32675, "pid": 76337, "tid": -914061504, "ts": 1716454218011888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218011927, "dur": 5,
    "args": {
      "External id": 32683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32683,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32683, "pid": 5, "tid": 7, "ts": 1716454218011927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011918, "dur": 9,
    "args": {
      "External id": 32683,
      "cbid": 211, "correlation": 32683
    }
  },
  {
    "ph": "s", "id": 32683, "pid": 76337, "tid": -914061504, "ts": 1716454218011918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218011959, "dur": 4,
    "args": {
      "External id": 32692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32692,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32692, "pid": 5, "tid": 7, "ts": 1716454218011959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218011948, "dur": 10,
    "args": {
      "External id": 32692,
      "cbid": 211, "correlation": 32692
    }
  },
  {
    "ph": "s", "id": 32692, "pid": 76337, "tid": -914061504, "ts": 1716454218011948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218012032, "dur": 4,
    "args": {
      "External id": 32705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32705,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32705, "pid": 5, "tid": 7, "ts": 1716454218012032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012018, "dur": 15,
    "args": {
      "External id": 32705,
      "cbid": 211, "correlation": 32705
    }
  },
  {
    "ph": "s", "id": 32705, "pid": 76337, "tid": -914061504, "ts": 1716454218012018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454218012075, "dur": 8,
    "args": {
      "External id": 32715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32715,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 32715, "pid": 5, "tid": 7, "ts": 1716454218012075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012063, "dur": 12,
    "args": {
      "External id": 32715,
      "cbid": 211, "correlation": 32715
    }
  },
  {
    "ph": "s", "id": 32715, "pid": 76337, "tid": -914061504, "ts": 1716454218012063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012185, "dur": 3,
    "args": {
      "External id": 32732,
      "cbid": 251, "correlation": 32732
    }
  },
  {
    "ph": "f", "id": 32732, "pid": 76337, "tid": -914061504, "ts": 1716454218012185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454218012219, "dur": 11,
    "args": {
      "External id": 32734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32734,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 32734, "pid": 5, "tid": 7, "ts": 1716454218012219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012197, "dur": 23,
    "args": {
      "External id": 32734,
      "cbid": 211, "correlation": 32734
    }
  },
  {
    "ph": "s", "id": 32734, "pid": 76337, "tid": -914061504, "ts": 1716454218012197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218012277, "dur": 4,
    "args": {
      "External id": 32742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32742,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32742, "pid": 5, "tid": 7, "ts": 1716454218012277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012265, "dur": 11,
    "args": {
      "External id": 32742,
      "cbid": 211, "correlation": 32742
    }
  },
  {
    "ph": "s", "id": 32742, "pid": 76337, "tid": -914061504, "ts": 1716454218012265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012332, "dur": 2,
    "args": {
      "External id": 32758,
      "cbid": 251, "correlation": 32758
    }
  },
  {
    "ph": "f", "id": 32758, "pid": 76337, "tid": -914061504, "ts": 1716454218012332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012338, "dur": 0,
    "args": {
      "External id": 32760,
      "cbid": 251, "correlation": 32760
    }
  },
  {
    "ph": "f", "id": 32760, "pid": 76337, "tid": -914061504, "ts": 1716454218012338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218012353, "dur": 13,
    "args": {
      "External id": 32761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32761,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 32761, "pid": 5, "tid": 7, "ts": 1716454218012353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012340, "dur": 14,
    "args": {
      "External id": 32761,
      "cbid": 211, "correlation": 32761
    }
  },
  {
    "ph": "s", "id": 32761, "pid": 76337, "tid": -914061504, "ts": 1716454218012340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218012368, "dur": 5,
    "args": {
      "External id": 32763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32763,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 32763, "pid": 5, "tid": 7, "ts": 1716454218012368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012358, "dur": 9,
    "args": {
      "External id": 32763,
      "cbid": 211, "correlation": 32763
    }
  },
  {
    "ph": "s", "id": 32763, "pid": 76337, "tid": -914061504, "ts": 1716454218012358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012454, "dur": 1,
    "args": {
      "External id": 32773,
      "cbid": 317, "correlation": 32773
    }
  },
  {
    "ph": "f", "id": 32773, "pid": 76337, "tid": -914061504, "ts": 1716454218012454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012456, "dur": 1,
    "args": {
      "External id": 32774,
      "cbid": 203, "correlation": 32774
    }
  },
  {
    "ph": "f", "id": 32774, "pid": 76337, "tid": -914061504, "ts": 1716454218012456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012458, "dur": 1,
    "args": {
      "External id": 32775,
      "cbid": 205, "correlation": 32775
    }
  },
  {
    "ph": "f", "id": 32775, "pid": 76337, "tid": -914061504, "ts": 1716454218012458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218012503, "dur": 6,
    "args": {
      "External id": 32779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32779,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32779, "pid": 5, "tid": 7, "ts": 1716454218012503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012489, "dur": 14,
    "args": {
      "External id": 32779,
      "cbid": 211, "correlation": 32779
    }
  },
  {
    "ph": "s", "id": 32779, "pid": 76337, "tid": -914061504, "ts": 1716454218012489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218012515, "dur": 4,
    "args": {
      "External id": 32781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32781,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 32781, "pid": 5, "tid": 7, "ts": 1716454218012515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012507, "dur": 7,
    "args": {
      "External id": 32781,
      "cbid": 211, "correlation": 32781
    }
  },
  {
    "ph": "s", "id": 32781, "pid": 76337, "tid": -914061504, "ts": 1716454218012507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218012531, "dur": 3,
    "args": {
      "External id": 32783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32783, "pid": 5, "tid": 7, "ts": 1716454218012531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012523, "dur": 7,
    "args": {
      "External id": 32783,
      "cbid": 211, "correlation": 32783
    }
  },
  {
    "ph": "s", "id": 32783, "pid": 76337, "tid": -914061504, "ts": 1716454218012523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012534, "dur": 0,
    "args": {
      "External id": 32784,
      "cbid": 51, "correlation": 32784
    }
  },
  {
    "ph": "s", "id": 32784, "pid": 76337, "tid": -914061504, "ts": 1716454218012534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218012545, "dur": 83,
    "args": {
      "External id": 32785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32785,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 32785, "pid": 5, "tid": 7, "ts": 1716454218012545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012536, "dur": 7,
    "args": {
      "External id": 32785,
      "cbid": 211, "correlation": 32785
    }
  },
  {
    "ph": "s", "id": 32785, "pid": 76337, "tid": -914061504, "ts": 1716454218012536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218012629, "dur": 60,
    "args": {
      "External id": 32790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32790, "pid": 5, "tid": 7, "ts": 1716454218012629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218012572, "dur": 9,
    "args": {
      "External id": 32790,
      "cbid": 211, "correlation": 32790
    }
  },
  {
    "ph": "s", "id": 32790, "pid": 76337, "tid": -914061504, "ts": 1716454218012572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218014337, "dur": 50,
    "args": {
      "External id": 32810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32810,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 32810, "pid": 5, "tid": 7, "ts": 1716454218014337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014321, "dur": 16,
    "args": {
      "External id": 32810,
      "cbid": 211, "correlation": 32810
    }
  },
  {
    "ph": "s", "id": 32810, "pid": 76337, "tid": -914061504, "ts": 1716454218014321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218014389, "dur": 4,
    "args": {
      "External id": 32822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32822,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32822, "pid": 5, "tid": 7, "ts": 1716454218014389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014349, "dur": 8,
    "args": {
      "External id": 32822,
      "cbid": 211, "correlation": 32822
    }
  },
  {
    "ph": "s", "id": 32822, "pid": 76337, "tid": -914061504, "ts": 1716454218014349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218014395, "dur": 56,
    "args": {
      "External id": 32825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32825,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32825, "pid": 5, "tid": 7, "ts": 1716454218014395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014373, "dur": 7,
    "args": {
      "External id": 32825,
      "cbid": 211, "correlation": 32825
    }
  },
  {
    "ph": "s", "id": 32825, "pid": 76337, "tid": -914061504, "ts": 1716454218014373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218014452, "dur": 36,
    "args": {
      "External id": 32834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32834,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32834, "pid": 5, "tid": 7, "ts": 1716454218014452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014418, "dur": 10,
    "args": {
      "External id": 32834,
      "cbid": 211, "correlation": 32834
    }
  },
  {
    "ph": "s", "id": 32834, "pid": 76337, "tid": -914061504, "ts": 1716454218014418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014475, "dur": 0,
    "args": {
      "External id": 32844,
      "cbid": 317, "correlation": 32844
    }
  },
  {
    "ph": "f", "id": 32844, "pid": 76337, "tid": -914061504, "ts": 1716454218014475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014476, "dur": 0,
    "args": {
      "External id": 32845,
      "cbid": 203, "correlation": 32845
    }
  },
  {
    "ph": "f", "id": 32845, "pid": 76337, "tid": -914061504, "ts": 1716454218014476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014477, "dur": 0,
    "args": {
      "External id": 32846,
      "cbid": 205, "correlation": 32846
    }
  },
  {
    "ph": "f", "id": 32846, "pid": 76337, "tid": -914061504, "ts": 1716454218014477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218014507, "dur": 40,
    "args": {
      "External id": 32850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32850,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32850, "pid": 5, "tid": 7, "ts": 1716454218014507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014494, "dur": 12,
    "args": {
      "External id": 32850,
      "cbid": 211, "correlation": 32850
    }
  },
  {
    "ph": "s", "id": 32850, "pid": 76337, "tid": -914061504, "ts": 1716454218014494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218014549, "dur": 14,
    "args": {
      "External id": 32852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32852,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32852, "pid": 5, "tid": 7, "ts": 1716454218014549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014509, "dur": 6,
    "args": {
      "External id": 32852,
      "cbid": 211, "correlation": 32852
    }
  },
  {
    "ph": "s", "id": 32852, "pid": 76337, "tid": -914061504, "ts": 1716454218014509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218014564, "dur": 3,
    "args": {
      "External id": 32854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32854, "pid": 5, "tid": 7, "ts": 1716454218014564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014520, "dur": 6,
    "args": {
      "External id": 32854,
      "cbid": 211, "correlation": 32854
    }
  },
  {
    "ph": "s", "id": 32854, "pid": 76337, "tid": -914061504, "ts": 1716454218014520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014530, "dur": 0,
    "args": {
      "External id": 32855,
      "cbid": 51, "correlation": 32855
    }
  },
  {
    "ph": "s", "id": 32855, "pid": 76337, "tid": -914061504, "ts": 1716454218014530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218014568, "dur": 694,
    "args": {
      "External id": 32856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32856,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 32856, "pid": 5, "tid": 7, "ts": 1716454218014568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014531, "dur": 6,
    "args": {
      "External id": 32856,
      "cbid": 211, "correlation": 32856
    }
  },
  {
    "ph": "s", "id": 32856, "pid": 76337, "tid": -914061504, "ts": 1716454218014531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218015264, "dur": 59,
    "args": {
      "External id": 32861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32861, "pid": 5, "tid": 7, "ts": 1716454218015264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014560, "dur": 9,
    "args": {
      "External id": 32861,
      "cbid": 211, "correlation": 32861
    }
  },
  {
    "ph": "s", "id": 32861, "pid": 76337, "tid": -914061504, "ts": 1716454218014560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218015324, "dur": 3,
    "args": {
      "External id": 32869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32869,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32869, "pid": 5, "tid": 7, "ts": 1716454218015324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014604, "dur": 9,
    "args": {
      "External id": 32869,
      "cbid": 211, "correlation": 32869
    }
  },
  {
    "ph": "s", "id": 32869, "pid": 76337, "tid": -914061504, "ts": 1716454218014604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014670, "dur": 2,
    "args": {
      "External id": 32885,
      "cbid": 251, "correlation": 32885
    }
  },
  {
    "ph": "f", "id": 32885, "pid": 76337, "tid": -914061504, "ts": 1716454218014670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014676, "dur": 0,
    "args": {
      "External id": 32887,
      "cbid": 251, "correlation": 32887
    }
  },
  {
    "ph": "f", "id": 32887, "pid": 76337, "tid": -914061504, "ts": 1716454218014676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218015329, "dur": 9,
    "args": {
      "External id": 32888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32888,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 32888, "pid": 5, "tid": 7, "ts": 1716454218015329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014678, "dur": 12,
    "args": {
      "External id": 32888,
      "cbid": 211, "correlation": 32888
    }
  },
  {
    "ph": "s", "id": 32888, "pid": 76337, "tid": -914061504, "ts": 1716454218014678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218015339, "dur": 4,
    "args": {
      "External id": 32890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32890,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 32890, "pid": 5, "tid": 7, "ts": 1716454218015339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014692, "dur": 6,
    "args": {
      "External id": 32890,
      "cbid": 211, "correlation": 32890
    }
  },
  {
    "ph": "s", "id": 32890, "pid": 76337, "tid": -914061504, "ts": 1716454218014692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218015344, "dur": 53,
    "args": {
      "External id": 32900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32900, "pid": 5, "tid": 7, "ts": 1716454218015344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014751, "dur": 12,
    "args": {
      "External id": 32900,
      "cbid": 211, "correlation": 32900
    }
  },
  {
    "ph": "s", "id": 32900, "pid": 76337, "tid": -914061504, "ts": 1716454218014751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218015399, "dur": 51,
    "args": {
      "External id": 32920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32920,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 32920, "pid": 5, "tid": 7, "ts": 1716454218015399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014816, "dur": 11,
    "args": {
      "External id": 32920,
      "cbid": 211, "correlation": 32920
    }
  },
  {
    "ph": "s", "id": 32920, "pid": 76337, "tid": -914061504, "ts": 1716454218014816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218015452, "dur": 4,
    "args": {
      "External id": 32932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32932,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32932, "pid": 5, "tid": 7, "ts": 1716454218015452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014838, "dur": 6,
    "args": {
      "External id": 32932,
      "cbid": 211, "correlation": 32932
    }
  },
  {
    "ph": "s", "id": 32932, "pid": 76337, "tid": -914061504, "ts": 1716454218014838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218015457, "dur": 55,
    "args": {
      "External id": 32935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32935, "pid": 5, "tid": 7, "ts": 1716454218015457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014857, "dur": 6,
    "args": {
      "External id": 32935,
      "cbid": 211, "correlation": 32935
    }
  },
  {
    "ph": "s", "id": 32935, "pid": 76337, "tid": -914061504, "ts": 1716454218014857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218015513, "dur": 37,
    "args": {
      "External id": 32944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32944,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32944, "pid": 5, "tid": 7, "ts": 1716454218015513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014897, "dur": 9,
    "args": {
      "External id": 32944,
      "cbid": 211, "correlation": 32944
    }
  },
  {
    "ph": "s", "id": 32944, "pid": 76337, "tid": -914061504, "ts": 1716454218014897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014966, "dur": 0,
    "args": {
      "External id": 32954,
      "cbid": 317, "correlation": 32954
    }
  },
  {
    "ph": "f", "id": 32954, "pid": 76337, "tid": -914061504, "ts": 1716454218014966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014967, "dur": 0,
    "args": {
      "External id": 32955,
      "cbid": 203, "correlation": 32955
    }
  },
  {
    "ph": "f", "id": 32955, "pid": 76337, "tid": -914061504, "ts": 1716454218014967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014968, "dur": 0,
    "args": {
      "External id": 32956,
      "cbid": 205, "correlation": 32956
    }
  },
  {
    "ph": "f", "id": 32956, "pid": 76337, "tid": -914061504, "ts": 1716454218014968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218015551, "dur": 41,
    "args": {
      "External id": 32960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32960,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32960, "pid": 5, "tid": 7, "ts": 1716454218015551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218014990, "dur": 12,
    "args": {
      "External id": 32960,
      "cbid": 211, "correlation": 32960
    }
  },
  {
    "ph": "s", "id": 32960, "pid": 76337, "tid": -914061504, "ts": 1716454218014990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218015593, "dur": 14,
    "args": {
      "External id": 32962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32962,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32962, "pid": 5, "tid": 7, "ts": 1716454218015593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015005, "dur": 5,
    "args": {
      "External id": 32962,
      "cbid": 211, "correlation": 32962
    }
  },
  {
    "ph": "s", "id": 32962, "pid": 76337, "tid": -914061504, "ts": 1716454218015005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218015609, "dur": 3,
    "args": {
      "External id": 32964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32964,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 32964, "pid": 5, "tid": 7, "ts": 1716454218015609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015014, "dur": 5,
    "args": {
      "External id": 32964,
      "cbid": 211, "correlation": 32964
    }
  },
  {
    "ph": "s", "id": 32964, "pid": 76337, "tid": -914061504, "ts": 1716454218015014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015023, "dur": 0,
    "args": {
      "External id": 32965,
      "cbid": 51, "correlation": 32965
    }
  },
  {
    "ph": "s", "id": 32965, "pid": 76337, "tid": -914061504, "ts": 1716454218015023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218015613, "dur": 687,
    "args": {
      "External id": 32966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32966,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 32966, "pid": 5, "tid": 7, "ts": 1716454218015613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015023, "dur": 5,
    "args": {
      "External id": 32966,
      "cbid": 211, "correlation": 32966
    }
  },
  {
    "ph": "s", "id": 32966, "pid": 76337, "tid": -914061504, "ts": 1716454218015023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218016302, "dur": 59,
    "args": {
      "External id": 32971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32971, "pid": 5, "tid": 7, "ts": 1716454218016302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015052, "dur": 8,
    "args": {
      "External id": 32971,
      "cbid": 211, "correlation": 32971
    }
  },
  {
    "ph": "s", "id": 32971, "pid": 76337, "tid": -914061504, "ts": 1716454218015052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218016362, "dur": 50,
    "args": {
      "External id": 32979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32979,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32979, "pid": 5, "tid": 7, "ts": 1716454218016362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015085, "dur": 9,
    "args": {
      "External id": 32979,
      "cbid": 211, "correlation": 32979
    }
  },
  {
    "ph": "s", "id": 32979, "pid": 76337, "tid": -914061504, "ts": 1716454218015085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218016413, "dur": 35,
    "args": {
      "External id": 32987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 32987,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 32987, "pid": 5, "tid": 7, "ts": 1716454218016413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015118, "dur": 9,
    "args": {
      "External id": 32987,
      "cbid": 211, "correlation": 32987
    }
  },
  {
    "ph": "s", "id": 32987, "pid": 76337, "tid": -914061504, "ts": 1716454218015118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218016449, "dur": 51,
    "args": {
      "External id": 33007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33007,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 33007, "pid": 5, "tid": 7, "ts": 1716454218016449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015200, "dur": 12,
    "args": {
      "External id": 33007,
      "cbid": 211, "correlation": 33007
    }
  },
  {
    "ph": "s", "id": 33007, "pid": 76337, "tid": -914061504, "ts": 1716454218015200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218016501, "dur": 4,
    "args": {
      "External id": 33019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33019,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33019, "pid": 5, "tid": 7, "ts": 1716454218016501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015222, "dur": 6,
    "args": {
      "External id": 33019,
      "cbid": 211, "correlation": 33019
    }
  },
  {
    "ph": "s", "id": 33019, "pid": 76337, "tid": -914061504, "ts": 1716454218015222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218016506, "dur": 55,
    "args": {
      "External id": 33022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33022,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33022, "pid": 5, "tid": 7, "ts": 1716454218016506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015239, "dur": 7,
    "args": {
      "External id": 33022,
      "cbid": 211, "correlation": 33022
    }
  },
  {
    "ph": "s", "id": 33022, "pid": 76337, "tid": -914061504, "ts": 1716454218015239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015298, "dur": 0,
    "args": {
      "External id": 33033,
      "cbid": 317, "correlation": 33033
    }
  },
  {
    "ph": "f", "id": 33033, "pid": 76337, "tid": -914061504, "ts": 1716454218015298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015298, "dur": 0,
    "args": {
      "External id": 33034,
      "cbid": 203, "correlation": 33034
    }
  },
  {
    "ph": "f", "id": 33034, "pid": 76337, "tid": -914061504, "ts": 1716454218015298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015299, "dur": 0,
    "args": {
      "External id": 33035,
      "cbid": 205, "correlation": 33035
    }
  },
  {
    "ph": "f", "id": 33035, "pid": 76337, "tid": -914061504, "ts": 1716454218015299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015333, "dur": 2,
    "args": {
      "External id": 33039,
      "cbid": 251, "correlation": 33039
    }
  },
  {
    "ph": "f", "id": 33039, "pid": 76337, "tid": -914061504, "ts": 1716454218015333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015337, "dur": 1,
    "args": {
      "External id": 33040,
      "cbid": 251, "correlation": 33040
    }
  },
  {
    "ph": "f", "id": 33040, "pid": 76337, "tid": -914061504, "ts": 1716454218015337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015339, "dur": 1,
    "args": {
      "External id": 33041,
      "cbid": 251, "correlation": 33041
    }
  },
  {
    "ph": "f", "id": 33041, "pid": 76337, "tid": -914061504, "ts": 1716454218015339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015341, "dur": 1,
    "args": {
      "External id": 33042,
      "cbid": 251, "correlation": 33042
    }
  },
  {
    "ph": "f", "id": 33042, "pid": 76337, "tid": -914061504, "ts": 1716454218015341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015343, "dur": 1,
    "args": {
      "External id": 33043,
      "cbid": 251, "correlation": 33043
    }
  },
  {
    "ph": "f", "id": 33043, "pid": 76337, "tid": -914061504, "ts": 1716454218015343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015345, "dur": 1,
    "args": {
      "External id": 33044,
      "cbid": 251, "correlation": 33044
    }
  },
  {
    "ph": "f", "id": 33044, "pid": 76337, "tid": -914061504, "ts": 1716454218015345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015347, "dur": 1,
    "args": {
      "External id": 33045,
      "cbid": 251, "correlation": 33045
    }
  },
  {
    "ph": "f", "id": 33045, "pid": 76337, "tid": -914061504, "ts": 1716454218015347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015349, "dur": 1,
    "args": {
      "External id": 33046,
      "cbid": 251, "correlation": 33046
    }
  },
  {
    "ph": "f", "id": 33046, "pid": 76337, "tid": -914061504, "ts": 1716454218015349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015352, "dur": 0,
    "args": {
      "External id": 33047,
      "cbid": 251, "correlation": 33047
    }
  },
  {
    "ph": "f", "id": 33047, "pid": 76337, "tid": -914061504, "ts": 1716454218015352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218016563, "dur": 113,
    "args": {
      "External id": 33048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33048,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33048, "pid": 5, "tid": 7, "ts": 1716454218016563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015356, "dur": 13,
    "args": {
      "External id": 33048,
      "cbid": 211, "correlation": 33048
    }
  },
  {
    "ph": "s", "id": 33048, "pid": 76337, "tid": -914061504, "ts": 1716454218015356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218016677, "dur": 60,
    "args": {
      "External id": 33054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33054,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33054, "pid": 5, "tid": 7, "ts": 1716454218016677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015391, "dur": 9,
    "args": {
      "External id": 33054,
      "cbid": 211, "correlation": 33054
    }
  },
  {
    "ph": "s", "id": 33054, "pid": 76337, "tid": -914061504, "ts": 1716454218015391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218016738, "dur": 619,
    "args": {
      "External id": 33063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33063,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33063, "pid": 5, "tid": 7, "ts": 1716454218016738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015505, "dur": 15,
    "args": {
      "External id": 33063,
      "cbid": 211, "correlation": 33063
    }
  },
  {
    "ph": "s", "id": 33063, "pid": 76337, "tid": -914061504, "ts": 1716454218015505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218017358, "dur": 179,
    "args": {
      "External id": 33085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33085,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33085, "pid": 5, "tid": 7, "ts": 1716454218017358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015575, "dur": 11,
    "args": {
      "External id": 33085,
      "cbid": 211, "correlation": 33085
    }
  },
  {
    "ph": "s", "id": 33085, "pid": 76337, "tid": -914061504, "ts": 1716454218015575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015690, "dur": 2,
    "args": {
      "External id": 33096,
      "cbid": 251, "correlation": 33096
    }
  },
  {
    "ph": "f", "id": 33096, "pid": 76337, "tid": -914061504, "ts": 1716454218015690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218017539, "dur": 196,
    "args": {
      "External id": 33097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33097,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33097, "pid": 5, "tid": 7, "ts": 1716454218017539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015697, "dur": 14,
    "args": {
      "External id": 33097,
      "cbid": 211, "correlation": 33097
    }
  },
  {
    "ph": "s", "id": 33097, "pid": 76337, "tid": -914061504, "ts": 1716454218015697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015768, "dur": 1,
    "args": {
      "External id": 33108,
      "cbid": 251, "correlation": 33108
    }
  },
  {
    "ph": "f", "id": 33108, "pid": 76337, "tid": -914061504, "ts": 1716454218015768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218017736, "dur": 190,
    "args": {
      "External id": 33109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33109,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33109, "pid": 5, "tid": 7, "ts": 1716454218017736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015772, "dur": 11,
    "args": {
      "External id": 33109,
      "cbid": 211, "correlation": 33109
    }
  },
  {
    "ph": "s", "id": 33109, "pid": 76337, "tid": -914061504, "ts": 1716454218015772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015836, "dur": 1,
    "args": {
      "External id": 33120,
      "cbid": 251, "correlation": 33120
    }
  },
  {
    "ph": "f", "id": 33120, "pid": 76337, "tid": -914061504, "ts": 1716454218015836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218017927, "dur": 184,
    "args": {
      "External id": 33121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33121,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33121, "pid": 5, "tid": 7, "ts": 1716454218017927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015840, "dur": 11,
    "args": {
      "External id": 33121,
      "cbid": 211, "correlation": 33121
    }
  },
  {
    "ph": "s", "id": 33121, "pid": 76337, "tid": -914061504, "ts": 1716454218015840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218018113, "dur": 18418,
    "args": {
      "External id": 33142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33142,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33142, "pid": 5, "tid": 7, "ts": 1716454218018113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218015944, "dur": 20,
    "args": {
      "External id": 33142,
      "cbid": 211, "correlation": 33142
    }
  },
  {
    "ph": "s", "id": 33142, "pid": 76337, "tid": -914061504, "ts": 1716454218015944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016070, "dur": 2,
    "args": {
      "External id": 33160,
      "cbid": 251, "correlation": 33160
    }
  },
  {
    "ph": "f", "id": 33160, "pid": 76337, "tid": -914061504, "ts": 1716454218016070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218036532, "dur": 198,
    "args": {
      "External id": 33162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33162,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33162, "pid": 5, "tid": 7, "ts": 1716454218036532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016076, "dur": 14,
    "args": {
      "External id": 33162,
      "cbid": 211, "correlation": 33162
    }
  },
  {
    "ph": "s", "id": 33162, "pid": 76337, "tid": -914061504, "ts": 1716454218016076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218036732, "dur": 66,
    "args": {
      "External id": 33170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33170,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33170, "pid": 5, "tid": 7, "ts": 1716454218036732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016149, "dur": 13,
    "args": {
      "External id": 33170,
      "cbid": 211, "correlation": 33170
    }
  },
  {
    "ph": "s", "id": 33170, "pid": 76337, "tid": -914061504, "ts": 1716454218016149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218036799, "dur": 97,
    "args": {
      "External id": 33178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33178,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33178, "pid": 5, "tid": 7, "ts": 1716454218036799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016188, "dur": 9,
    "args": {
      "External id": 33178,
      "cbid": 211, "correlation": 33178
    }
  },
  {
    "ph": "s", "id": 33178, "pid": 76337, "tid": -914061504, "ts": 1716454218016188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218036897, "dur": 53,
    "args": {
      "External id": 33189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33189, "pid": 5, "tid": 7, "ts": 1716454218036897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016273, "dur": 14,
    "args": {
      "External id": 33189,
      "cbid": 211, "correlation": 33189
    }
  },
  {
    "ph": "s", "id": 33189, "pid": 76337, "tid": -914061504, "ts": 1716454218016273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218036951, "dur": 91,
    "args": {
      "External id": 33211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33211,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33211, "pid": 5, "tid": 7, "ts": 1716454218036951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016306, "dur": 8,
    "args": {
      "External id": 33211,
      "cbid": 211, "correlation": 33211
    }
  },
  {
    "ph": "s", "id": 33211, "pid": 76337, "tid": -914061504, "ts": 1716454218016306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016389, "dur": 1,
    "args": {
      "External id": 33222,
      "cbid": 251, "correlation": 33222
    }
  },
  {
    "ph": "f", "id": 33222, "pid": 76337, "tid": -914061504, "ts": 1716454218016389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218037044, "dur": 103,
    "args": {
      "External id": 33223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33223,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33223, "pid": 5, "tid": 7, "ts": 1716454218037044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016394, "dur": 13,
    "args": {
      "External id": 33223,
      "cbid": 211, "correlation": 33223
    }
  },
  {
    "ph": "s", "id": 33223, "pid": 76337, "tid": -914061504, "ts": 1716454218016394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016475, "dur": 1,
    "args": {
      "External id": 33234,
      "cbid": 251, "correlation": 33234
    }
  },
  {
    "ph": "f", "id": 33234, "pid": 76337, "tid": -914061504, "ts": 1716454218016475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016479, "dur": 0,
    "args": {
      "External id": 33235,
      "cbid": 251, "correlation": 33235
    }
  },
  {
    "ph": "f", "id": 33235, "pid": 76337, "tid": -914061504, "ts": 1716454218016479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218037148, "dur": 10,
    "args": {
      "External id": 33236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33236,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 33236, "pid": 5, "tid": 7, "ts": 1716454218037148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016481, "dur": 13,
    "args": {
      "External id": 33236,
      "cbid": 211, "correlation": 33236
    }
  },
  {
    "ph": "s", "id": 33236, "pid": 76337, "tid": -914061504, "ts": 1716454218016481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218037160, "dur": 5,
    "args": {
      "External id": 33238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33238,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 33238, "pid": 5, "tid": 7, "ts": 1716454218037160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016498, "dur": 7,
    "args": {
      "External id": 33238,
      "cbid": 211, "correlation": 33238
    }
  },
  {
    "ph": "s", "id": 33238, "pid": 76337, "tid": -914061504, "ts": 1716454218016498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016560, "dur": 1,
    "args": {
      "External id": 33249,
      "cbid": 251, "correlation": 33249
    }
  },
  {
    "ph": "f", "id": 33249, "pid": 76337, "tid": -914061504, "ts": 1716454218016560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016564, "dur": 0,
    "args": {
      "External id": 33250,
      "cbid": 251, "correlation": 33250
    }
  },
  {
    "ph": "f", "id": 33250, "pid": 76337, "tid": -914061504, "ts": 1716454218016564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218037166, "dur": 6,
    "args": {
      "External id": 33251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33251,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 33251, "pid": 5, "tid": 7, "ts": 1716454218037166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016565, "dur": 12,
    "args": {
      "External id": 33251,
      "cbid": 211, "correlation": 33251
    }
  },
  {
    "ph": "s", "id": 33251, "pid": 76337, "tid": -914061504, "ts": 1716454218016565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218037174, "dur": 3,
    "args": {
      "External id": 33253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33253,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 33253, "pid": 5, "tid": 7, "ts": 1716454218037174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016579, "dur": 6,
    "args": {
      "External id": 33253,
      "cbid": 211, "correlation": 33253
    }
  },
  {
    "ph": "s", "id": 33253, "pid": 76337, "tid": -914061504, "ts": 1716454218016579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218037178, "dur": 154,
    "args": {
      "External id": 33274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33274,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33274, "pid": 5, "tid": 7, "ts": 1716454218037178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016653, "dur": 12,
    "args": {
      "External id": 33274,
      "cbid": 211, "correlation": 33274
    }
  },
  {
    "ph": "s", "id": 33274, "pid": 76337, "tid": -914061504, "ts": 1716454218016653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016750, "dur": 2,
    "args": {
      "External id": 33292,
      "cbid": 251, "correlation": 33292
    }
  },
  {
    "ph": "f", "id": 33292, "pid": 76337, "tid": -914061504, "ts": 1716454218016750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218037334, "dur": 106,
    "args": {
      "External id": 33294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33294,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33294, "pid": 5, "tid": 7, "ts": 1716454218037334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016756, "dur": 14,
    "args": {
      "External id": 33294,
      "cbid": 211, "correlation": 33294
    }
  },
  {
    "ph": "s", "id": 33294, "pid": 76337, "tid": -914061504, "ts": 1716454218016756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218037442, "dur": 34,
    "args": {
      "External id": 33302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33302,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33302, "pid": 5, "tid": 7, "ts": 1716454218037442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016828, "dur": 12,
    "args": {
      "External id": 33302,
      "cbid": 211, "correlation": 33302
    }
  },
  {
    "ph": "s", "id": 33302, "pid": 76337, "tid": -914061504, "ts": 1716454218016828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218037477, "dur": 68,
    "args": {
      "External id": 33310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33310,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33310, "pid": 5, "tid": 7, "ts": 1716454218037477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016869, "dur": 9,
    "args": {
      "External id": 33310,
      "cbid": 211, "correlation": 33310
    }
  },
  {
    "ph": "s", "id": 33310, "pid": 76337, "tid": -914061504, "ts": 1716454218016869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218037546, "dur": 91,
    "args": {
      "External id": 33332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33332,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33332, "pid": 5, "tid": 7, "ts": 1716454218037546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218016921, "dur": 10,
    "args": {
      "External id": 33332,
      "cbid": 211, "correlation": 33332
    }
  },
  {
    "ph": "s", "id": 33332, "pid": 76337, "tid": -914061504, "ts": 1716454218016921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017022, "dur": 1,
    "args": {
      "External id": 33348,
      "cbid": 251, "correlation": 33348
    }
  },
  {
    "ph": "f", "id": 33348, "pid": 76337, "tid": -914061504, "ts": 1716454218017022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218037639, "dur": 565,
    "args": {
      "External id": 33350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33350,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33350, "pid": 5, "tid": 7, "ts": 1716454218037639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017028, "dur": 13,
    "args": {
      "External id": 33350,
      "cbid": 211, "correlation": 33350
    }
  },
  {
    "ph": "s", "id": 33350, "pid": 76337, "tid": -914061504, "ts": 1716454218017028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218038205, "dur": 241,
    "args": {
      "External id": 33358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33358, "pid": 5, "tid": 7, "ts": 1716454218038205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017107, "dur": 15,
    "args": {
      "External id": 33358,
      "cbid": 211, "correlation": 33358
    }
  },
  {
    "ph": "s", "id": 33358, "pid": 76337, "tid": -914061504, "ts": 1716454218017107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218038447, "dur": 251,
    "args": {
      "External id": 33366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33366,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33366, "pid": 5, "tid": 7, "ts": 1716454218038447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017142, "dur": 8,
    "args": {
      "External id": 33366,
      "cbid": 211, "correlation": 33366
    }
  },
  {
    "ph": "s", "id": 33366, "pid": 76337, "tid": -914061504, "ts": 1716454218017142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017226, "dur": 2,
    "args": {
      "External id": 33382,
      "cbid": 251, "correlation": 33382
    }
  },
  {
    "ph": "f", "id": 33382, "pid": 76337, "tid": -914061504, "ts": 1716454218017226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017232, "dur": 0,
    "args": {
      "External id": 33384,
      "cbid": 251, "correlation": 33384
    }
  },
  {
    "ph": "f", "id": 33384, "pid": 76337, "tid": -914061504, "ts": 1716454218017232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218038700, "dur": 358,
    "args": {
      "External id": 33385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33385,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 33385, "pid": 5, "tid": 7, "ts": 1716454218038700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017237, "dur": 14,
    "args": {
      "External id": 33385,
      "cbid": 211, "correlation": 33385
    }
  },
  {
    "ph": "s", "id": 33385, "pid": 76337, "tid": -914061504, "ts": 1716454218017237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218039059, "dur": 50,
    "args": {
      "External id": 33393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33393,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33393, "pid": 5, "tid": 7, "ts": 1716454218039059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017280, "dur": 10,
    "args": {
      "External id": 33393,
      "cbid": 211, "correlation": 33393
    }
  },
  {
    "ph": "s", "id": 33393, "pid": 76337, "tid": -914061504, "ts": 1716454218017280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218039110, "dur": 156,
    "args": {
      "External id": 33404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33404,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33404, "pid": 5, "tid": 7, "ts": 1716454218039110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017352, "dur": 12,
    "args": {
      "External id": 33404,
      "cbid": 211, "correlation": 33404
    }
  },
  {
    "ph": "s", "id": 33404, "pid": 76337, "tid": -914061504, "ts": 1716454218017352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017417, "dur": 0,
    "args": {
      "External id": 33416,
      "cbid": 317, "correlation": 33416
    }
  },
  {
    "ph": "f", "id": 33416, "pid": 76337, "tid": -914061504, "ts": 1716454218017417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017418, "dur": 0,
    "args": {
      "External id": 33417,
      "cbid": 203, "correlation": 33417
    }
  },
  {
    "ph": "f", "id": 33417, "pid": 76337, "tid": -914061504, "ts": 1716454218017418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017419, "dur": 0,
    "args": {
      "External id": 33418,
      "cbid": 205, "correlation": 33418
    }
  },
  {
    "ph": "f", "id": 33418, "pid": 76337, "tid": -914061504, "ts": 1716454218017419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017443, "dur": 1,
    "args": {
      "External id": 33422,
      "cbid": 251, "correlation": 33422
    }
  },
  {
    "ph": "f", "id": 33422, "pid": 76337, "tid": -914061504, "ts": 1716454218017443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017445, "dur": 0,
    "args": {
      "External id": 33423,
      "cbid": 251, "correlation": 33423
    }
  },
  {
    "ph": "f", "id": 33423, "pid": 76337, "tid": -914061504, "ts": 1716454218017445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017445, "dur": 0,
    "args": {
      "External id": 33424,
      "cbid": 251, "correlation": 33424
    }
  },
  {
    "ph": "f", "id": 33424, "pid": 76337, "tid": -914061504, "ts": 1716454218017445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017446, "dur": 0,
    "args": {
      "External id": 33425,
      "cbid": 251, "correlation": 33425
    }
  },
  {
    "ph": "f", "id": 33425, "pid": 76337, "tid": -914061504, "ts": 1716454218017446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017447, "dur": 0,
    "args": {
      "External id": 33426,
      "cbid": 251, "correlation": 33426
    }
  },
  {
    "ph": "f", "id": 33426, "pid": 76337, "tid": -914061504, "ts": 1716454218017447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017448, "dur": 0,
    "args": {
      "External id": 33427,
      "cbid": 251, "correlation": 33427
    }
  },
  {
    "ph": "f", "id": 33427, "pid": 76337, "tid": -914061504, "ts": 1716454218017448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017449, "dur": 0,
    "args": {
      "External id": 33428,
      "cbid": 251, "correlation": 33428
    }
  },
  {
    "ph": "f", "id": 33428, "pid": 76337, "tid": -914061504, "ts": 1716454218017449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017450, "dur": 0,
    "args": {
      "External id": 33429,
      "cbid": 251, "correlation": 33429
    }
  },
  {
    "ph": "f", "id": 33429, "pid": 76337, "tid": -914061504, "ts": 1716454218017450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017451, "dur": 0,
    "args": {
      "External id": 33430,
      "cbid": 251, "correlation": 33430
    }
  },
  {
    "ph": "f", "id": 33430, "pid": 76337, "tid": -914061504, "ts": 1716454218017451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218039267, "dur": 114,
    "args": {
      "External id": 33431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33431,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33431, "pid": 5, "tid": 7, "ts": 1716454218039267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017453, "dur": 12,
    "args": {
      "External id": 33431,
      "cbid": 211, "correlation": 33431
    }
  },
  {
    "ph": "s", "id": 33431, "pid": 76337, "tid": -914061504, "ts": 1716454218017453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218039383, "dur": 59,
    "args": {
      "External id": 33437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33437, "pid": 5, "tid": 7, "ts": 1716454218039383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017488, "dur": 9,
    "args": {
      "External id": 33437,
      "cbid": 211, "correlation": 33437
    }
  },
  {
    "ph": "s", "id": 33437, "pid": 76337, "tid": -914061504, "ts": 1716454218017488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218039443, "dur": 50,
    "args": {
      "External id": 33445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33445,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33445, "pid": 5, "tid": 7, "ts": 1716454218039443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017521, "dur": 9,
    "args": {
      "External id": 33445,
      "cbid": 211, "correlation": 33445
    }
  },
  {
    "ph": "s", "id": 33445, "pid": 76337, "tid": -914061504, "ts": 1716454218017521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218039494, "dur": 51,
    "args": {
      "External id": 33465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33465,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 33465, "pid": 5, "tid": 7, "ts": 1716454218039494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017596, "dur": 11,
    "args": {
      "External id": 33465,
      "cbid": 211, "correlation": 33465
    }
  },
  {
    "ph": "s", "id": 33465, "pid": 76337, "tid": -914061504, "ts": 1716454218017596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218039547, "dur": 5,
    "args": {
      "External id": 33477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33477,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33477, "pid": 5, "tid": 7, "ts": 1716454218039547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017616, "dur": 7,
    "args": {
      "External id": 33477,
      "cbid": 211, "correlation": 33477
    }
  },
  {
    "ph": "s", "id": 33477, "pid": 76337, "tid": -914061504, "ts": 1716454218017616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218039553, "dur": 55,
    "args": {
      "External id": 33480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33480, "pid": 5, "tid": 7, "ts": 1716454218039553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017635, "dur": 7,
    "args": {
      "External id": 33480,
      "cbid": 211, "correlation": 33480
    }
  },
  {
    "ph": "s", "id": 33480, "pid": 76337, "tid": -914061504, "ts": 1716454218017635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218039610, "dur": 37,
    "args": {
      "External id": 33489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33489,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33489, "pid": 5, "tid": 7, "ts": 1716454218039610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017674, "dur": 10,
    "args": {
      "External id": 33489,
      "cbid": 211, "correlation": 33489
    }
  },
  {
    "ph": "s", "id": 33489, "pid": 76337, "tid": -914061504, "ts": 1716454218017674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017725, "dur": 0,
    "args": {
      "External id": 33499,
      "cbid": 317, "correlation": 33499
    }
  },
  {
    "ph": "f", "id": 33499, "pid": 76337, "tid": -914061504, "ts": 1716454218017725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017726, "dur": 0,
    "args": {
      "External id": 33500,
      "cbid": 203, "correlation": 33500
    }
  },
  {
    "ph": "f", "id": 33500, "pid": 76337, "tid": -914061504, "ts": 1716454218017726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017727, "dur": 0,
    "args": {
      "External id": 33501,
      "cbid": 205, "correlation": 33501
    }
  },
  {
    "ph": "f", "id": 33501, "pid": 76337, "tid": -914061504, "ts": 1716454218017727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218039648, "dur": 41,
    "args": {
      "External id": 33505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33505,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33505, "pid": 5, "tid": 7, "ts": 1716454218039648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017742, "dur": 11,
    "args": {
      "External id": 33505,
      "cbid": 211, "correlation": 33505
    }
  },
  {
    "ph": "s", "id": 33505, "pid": 76337, "tid": -914061504, "ts": 1716454218017742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218039691, "dur": 14,
    "args": {
      "External id": 33507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33507,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33507, "pid": 5, "tid": 7, "ts": 1716454218039691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017756, "dur": 5,
    "args": {
      "External id": 33507,
      "cbid": 211, "correlation": 33507
    }
  },
  {
    "ph": "s", "id": 33507, "pid": 76337, "tid": -914061504, "ts": 1716454218017756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218039706, "dur": 4,
    "args": {
      "External id": 33509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33509, "pid": 5, "tid": 7, "ts": 1716454218039706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017766, "dur": 6,
    "args": {
      "External id": 33509,
      "cbid": 211, "correlation": 33509
    }
  },
  {
    "ph": "s", "id": 33509, "pid": 76337, "tid": -914061504, "ts": 1716454218017766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017775, "dur": 0,
    "args": {
      "External id": 33510,
      "cbid": 51, "correlation": 33510
    }
  },
  {
    "ph": "s", "id": 33510, "pid": 76337, "tid": -914061504, "ts": 1716454218017775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218039710, "dur": 696,
    "args": {
      "External id": 33511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33511,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33511, "pid": 5, "tid": 7, "ts": 1716454218039710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017775, "dur": 5,
    "args": {
      "External id": 33511,
      "cbid": 211, "correlation": 33511
    }
  },
  {
    "ph": "s", "id": 33511, "pid": 76337, "tid": -914061504, "ts": 1716454218017775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218040408, "dur": 59,
    "args": {
      "External id": 33516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33516,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33516, "pid": 5, "tid": 7, "ts": 1716454218040408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017803, "dur": 8,
    "args": {
      "External id": 33516,
      "cbid": 211, "correlation": 33516
    }
  },
  {
    "ph": "s", "id": 33516, "pid": 76337, "tid": -914061504, "ts": 1716454218017803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218040468, "dur": 3,
    "args": {
      "External id": 33524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33524,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33524, "pid": 5, "tid": 7, "ts": 1716454218040468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017846, "dur": 9,
    "args": {
      "External id": 33524,
      "cbid": 211, "correlation": 33524
    }
  },
  {
    "ph": "s", "id": 33524, "pid": 76337, "tid": -914061504, "ts": 1716454218017846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017913, "dur": 1,
    "args": {
      "External id": 33540,
      "cbid": 251, "correlation": 33540
    }
  },
  {
    "ph": "f", "id": 33540, "pid": 76337, "tid": -914061504, "ts": 1716454218017913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017918, "dur": 0,
    "args": {
      "External id": 33542,
      "cbid": 251, "correlation": 33542
    }
  },
  {
    "ph": "f", "id": 33542, "pid": 76337, "tid": -914061504, "ts": 1716454218017918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218040473, "dur": 11,
    "args": {
      "External id": 33543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33543,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 33543, "pid": 5, "tid": 7, "ts": 1716454218040473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017920, "dur": 11,
    "args": {
      "External id": 33543,
      "cbid": 211, "correlation": 33543
    }
  },
  {
    "ph": "s", "id": 33543, "pid": 76337, "tid": -914061504, "ts": 1716454218017920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218040485, "dur": 5,
    "args": {
      "External id": 33545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33545,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 33545, "pid": 5, "tid": 7, "ts": 1716454218040485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218017933, "dur": 6,
    "args": {
      "External id": 33545,
      "cbid": 211, "correlation": 33545
    }
  },
  {
    "ph": "s", "id": 33545, "pid": 76337, "tid": -914061504, "ts": 1716454218017933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218040491, "dur": 54,
    "args": {
      "External id": 33555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33555, "pid": 5, "tid": 7, "ts": 1716454218040491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018000, "dur": 12,
    "args": {
      "External id": 33555,
      "cbid": 211, "correlation": 33555
    }
  },
  {
    "ph": "s", "id": 33555, "pid": 76337, "tid": -914061504, "ts": 1716454218018000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218040547, "dur": 50,
    "args": {
      "External id": 33575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33575,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 33575, "pid": 5, "tid": 7, "ts": 1716454218040547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018067, "dur": 11,
    "args": {
      "External id": 33575,
      "cbid": 211, "correlation": 33575
    }
  },
  {
    "ph": "s", "id": 33575, "pid": 76337, "tid": -914061504, "ts": 1716454218018067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218040598, "dur": 4,
    "args": {
      "External id": 33587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33587,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33587, "pid": 5, "tid": 7, "ts": 1716454218040598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018087, "dur": 7,
    "args": {
      "External id": 33587,
      "cbid": 211, "correlation": 33587
    }
  },
  {
    "ph": "s", "id": 33587, "pid": 76337, "tid": -914061504, "ts": 1716454218018087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218040604, "dur": 55,
    "args": {
      "External id": 33590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33590, "pid": 5, "tid": 7, "ts": 1716454218040604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018107, "dur": 6,
    "args": {
      "External id": 33590,
      "cbid": 211, "correlation": 33590
    }
  },
  {
    "ph": "s", "id": 33590, "pid": 76337, "tid": -914061504, "ts": 1716454218018107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218040660, "dur": 36,
    "args": {
      "External id": 33599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33599,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33599, "pid": 5, "tid": 7, "ts": 1716454218040660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018147, "dur": 10,
    "args": {
      "External id": 33599,
      "cbid": 211, "correlation": 33599
    }
  },
  {
    "ph": "s", "id": 33599, "pid": 76337, "tid": -914061504, "ts": 1716454218018147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018210, "dur": 0,
    "args": {
      "External id": 33609,
      "cbid": 317, "correlation": 33609
    }
  },
  {
    "ph": "f", "id": 33609, "pid": 76337, "tid": -914061504, "ts": 1716454218018210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018211, "dur": 0,
    "args": {
      "External id": 33610,
      "cbid": 203, "correlation": 33610
    }
  },
  {
    "ph": "f", "id": 33610, "pid": 76337, "tid": -914061504, "ts": 1716454218018211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018212, "dur": 0,
    "args": {
      "External id": 33611,
      "cbid": 205, "correlation": 33611
    }
  },
  {
    "ph": "f", "id": 33611, "pid": 76337, "tid": -914061504, "ts": 1716454218018212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218040698, "dur": 40,
    "args": {
      "External id": 33615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33615,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33615, "pid": 5, "tid": 7, "ts": 1716454218040698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018226, "dur": 12,
    "args": {
      "External id": 33615,
      "cbid": 211, "correlation": 33615
    }
  },
  {
    "ph": "s", "id": 33615, "pid": 76337, "tid": -914061504, "ts": 1716454218018226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218040739, "dur": 14,
    "args": {
      "External id": 33617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33617,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33617, "pid": 5, "tid": 7, "ts": 1716454218040739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018240, "dur": 5,
    "args": {
      "External id": 33617,
      "cbid": 211, "correlation": 33617
    }
  },
  {
    "ph": "s", "id": 33617, "pid": 76337, "tid": -914061504, "ts": 1716454218018240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218040754, "dur": 3,
    "args": {
      "External id": 33619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33619, "pid": 5, "tid": 7, "ts": 1716454218040754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018249, "dur": 6,
    "args": {
      "External id": 33619,
      "cbid": 211, "correlation": 33619
    }
  },
  {
    "ph": "s", "id": 33619, "pid": 76337, "tid": -914061504, "ts": 1716454218018249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018258, "dur": 0,
    "args": {
      "External id": 33620,
      "cbid": 51, "correlation": 33620
    }
  },
  {
    "ph": "s", "id": 33620, "pid": 76337, "tid": -914061504, "ts": 1716454218018258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218040759, "dur": 689,
    "args": {
      "External id": 33621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33621,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33621, "pid": 5, "tid": 7, "ts": 1716454218040759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018259, "dur": 5,
    "args": {
      "External id": 33621,
      "cbid": 211, "correlation": 33621
    }
  },
  {
    "ph": "s", "id": 33621, "pid": 76337, "tid": -914061504, "ts": 1716454218018259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218041449, "dur": 59,
    "args": {
      "External id": 33626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33626, "pid": 5, "tid": 7, "ts": 1716454218041449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018286, "dur": 8,
    "args": {
      "External id": 33626,
      "cbid": 211, "correlation": 33626
    }
  },
  {
    "ph": "s", "id": 33626, "pid": 76337, "tid": -914061504, "ts": 1716454218018286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218041509, "dur": 50,
    "args": {
      "External id": 33634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33634,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33634, "pid": 5, "tid": 7, "ts": 1716454218041509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018318, "dur": 8,
    "args": {
      "External id": 33634,
      "cbid": 211, "correlation": 33634
    }
  },
  {
    "ph": "s", "id": 33634, "pid": 76337, "tid": -914061504, "ts": 1716454218018318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218041560, "dur": 35,
    "args": {
      "External id": 33642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33642,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33642, "pid": 5, "tid": 7, "ts": 1716454218041560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018346, "dur": 9,
    "args": {
      "External id": 33642,
      "cbid": 211, "correlation": 33642
    }
  },
  {
    "ph": "s", "id": 33642, "pid": 76337, "tid": -914061504, "ts": 1716454218018346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218041597, "dur": 51,
    "args": {
      "External id": 33662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33662,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 33662, "pid": 5, "tid": 7, "ts": 1716454218041597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018425, "dur": 12,
    "args": {
      "External id": 33662,
      "cbid": 211, "correlation": 33662
    }
  },
  {
    "ph": "s", "id": 33662, "pid": 76337, "tid": -914061504, "ts": 1716454218018425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218041650, "dur": 4,
    "args": {
      "External id": 33674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33674,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 33674, "pid": 5, "tid": 7, "ts": 1716454218041650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018447, "dur": 6,
    "args": {
      "External id": 33674,
      "cbid": 211, "correlation": 33674
    }
  },
  {
    "ph": "s", "id": 33674, "pid": 76337, "tid": -914061504, "ts": 1716454218018447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218041655, "dur": 54,
    "args": {
      "External id": 33677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33677, "pid": 5, "tid": 7, "ts": 1716454218041655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018465, "dur": 6,
    "args": {
      "External id": 33677,
      "cbid": 211, "correlation": 33677
    }
  },
  {
    "ph": "s", "id": 33677, "pid": 76337, "tid": -914061504, "ts": 1716454218018465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018522, "dur": 0,
    "args": {
      "External id": 33688,
      "cbid": 317, "correlation": 33688
    }
  },
  {
    "ph": "f", "id": 33688, "pid": 76337, "tid": -914061504, "ts": 1716454218018522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018522, "dur": 0,
    "args": {
      "External id": 33689,
      "cbid": 203, "correlation": 33689
    }
  },
  {
    "ph": "f", "id": 33689, "pid": 76337, "tid": -914061504, "ts": 1716454218018522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018523, "dur": 0,
    "args": {
      "External id": 33690,
      "cbid": 205, "correlation": 33690
    }
  },
  {
    "ph": "f", "id": 33690, "pid": 76337, "tid": -914061504, "ts": 1716454218018523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018545, "dur": 1,
    "args": {
      "External id": 33694,
      "cbid": 251, "correlation": 33694
    }
  },
  {
    "ph": "f", "id": 33694, "pid": 76337, "tid": -914061504, "ts": 1716454218018545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018546, "dur": 0,
    "args": {
      "External id": 33695,
      "cbid": 251, "correlation": 33695
    }
  },
  {
    "ph": "f", "id": 33695, "pid": 76337, "tid": -914061504, "ts": 1716454218018546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018547, "dur": 0,
    "args": {
      "External id": 33696,
      "cbid": 251, "correlation": 33696
    }
  },
  {
    "ph": "f", "id": 33696, "pid": 76337, "tid": -914061504, "ts": 1716454218018547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018548, "dur": 0,
    "args": {
      "External id": 33697,
      "cbid": 251, "correlation": 33697
    }
  },
  {
    "ph": "f", "id": 33697, "pid": 76337, "tid": -914061504, "ts": 1716454218018548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018549, "dur": 0,
    "args": {
      "External id": 33698,
      "cbid": 251, "correlation": 33698
    }
  },
  {
    "ph": "f", "id": 33698, "pid": 76337, "tid": -914061504, "ts": 1716454218018549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018550, "dur": 0,
    "args": {
      "External id": 33699,
      "cbid": 251, "correlation": 33699
    }
  },
  {
    "ph": "f", "id": 33699, "pid": 76337, "tid": -914061504, "ts": 1716454218018550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018550, "dur": 0,
    "args": {
      "External id": 33700,
      "cbid": 251, "correlation": 33700
    }
  },
  {
    "ph": "f", "id": 33700, "pid": 76337, "tid": -914061504, "ts": 1716454218018550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018551, "dur": 0,
    "args": {
      "External id": 33701,
      "cbid": 251, "correlation": 33701
    }
  },
  {
    "ph": "f", "id": 33701, "pid": 76337, "tid": -914061504, "ts": 1716454218018551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018552, "dur": 0,
    "args": {
      "External id": 33702,
      "cbid": 251, "correlation": 33702
    }
  },
  {
    "ph": "f", "id": 33702, "pid": 76337, "tid": -914061504, "ts": 1716454218018552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218041710, "dur": 110,
    "args": {
      "External id": 33703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33703,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33703, "pid": 5, "tid": 7, "ts": 1716454218041710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018554, "dur": 13,
    "args": {
      "External id": 33703,
      "cbid": 211, "correlation": 33703
    }
  },
  {
    "ph": "s", "id": 33703, "pid": 76337, "tid": -914061504, "ts": 1716454218018554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218041821, "dur": 60,
    "args": {
      "External id": 33709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33709, "pid": 5, "tid": 7, "ts": 1716454218041821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018590, "dur": 9,
    "args": {
      "External id": 33709,
      "cbid": 211, "correlation": 33709
    }
  },
  {
    "ph": "s", "id": 33709, "pid": 76337, "tid": -914061504, "ts": 1716454218018590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218041882, "dur": 654,
    "args": {
      "External id": 33718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33718,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33718, "pid": 5, "tid": 7, "ts": 1716454218041882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018674, "dur": 13,
    "args": {
      "External id": 33718,
      "cbid": 211, "correlation": 33718
    }
  },
  {
    "ph": "s", "id": 33718, "pid": 76337, "tid": -914061504, "ts": 1716454218018674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218042538, "dur": 179,
    "args": {
      "External id": 33740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33740,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33740, "pid": 5, "tid": 7, "ts": 1716454218042538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018730, "dur": 10,
    "args": {
      "External id": 33740,
      "cbid": 211, "correlation": 33740
    }
  },
  {
    "ph": "s", "id": 33740, "pid": 76337, "tid": -914061504, "ts": 1716454218018730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018815, "dur": 1,
    "args": {
      "External id": 33751,
      "cbid": 251, "correlation": 33751
    }
  },
  {
    "ph": "f", "id": 33751, "pid": 76337, "tid": -914061504, "ts": 1716454218018815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218042718, "dur": 193,
    "args": {
      "External id": 33752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33752,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33752, "pid": 5, "tid": 7, "ts": 1716454218042718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018820, "dur": 14,
    "args": {
      "External id": 33752,
      "cbid": 211, "correlation": 33752
    }
  },
  {
    "ph": "s", "id": 33752, "pid": 76337, "tid": -914061504, "ts": 1716454218018820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018888, "dur": 1,
    "args": {
      "External id": 33763,
      "cbid": 251, "correlation": 33763
    }
  },
  {
    "ph": "f", "id": 33763, "pid": 76337, "tid": -914061504, "ts": 1716454218018888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218042912, "dur": 186,
    "args": {
      "External id": 33764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33764,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33764, "pid": 5, "tid": 7, "ts": 1716454218042912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018892, "dur": 11,
    "args": {
      "External id": 33764,
      "cbid": 211, "correlation": 33764
    }
  },
  {
    "ph": "s", "id": 33764, "pid": 76337, "tid": -914061504, "ts": 1716454218018892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018955, "dur": 1,
    "args": {
      "External id": 33775,
      "cbid": 251, "correlation": 33775
    }
  },
  {
    "ph": "f", "id": 33775, "pid": 76337, "tid": -914061504, "ts": 1716454218018955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218043100, "dur": 186,
    "args": {
      "External id": 33776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33776,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33776, "pid": 5, "tid": 7, "ts": 1716454218043100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218018959, "dur": 11,
    "args": {
      "External id": 33776,
      "cbid": 211, "correlation": 33776
    }
  },
  {
    "ph": "s", "id": 33776, "pid": 76337, "tid": -914061504, "ts": 1716454218018959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218043287, "dur": 18374,
    "args": {
      "External id": 33797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33797,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33797, "pid": 5, "tid": 7, "ts": 1716454218043287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019049, "dur": 12,
    "args": {
      "External id": 33797,
      "cbid": 211, "correlation": 33797
    }
  },
  {
    "ph": "s", "id": 33797, "pid": 76337, "tid": -914061504, "ts": 1716454218019049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019146, "dur": 1,
    "args": {
      "External id": 33815,
      "cbid": 251, "correlation": 33815
    }
  },
  {
    "ph": "f", "id": 33815, "pid": 76337, "tid": -914061504, "ts": 1716454218019146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218061663, "dur": 203,
    "args": {
      "External id": 33817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33817,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33817, "pid": 5, "tid": 7, "ts": 1716454218061663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019152, "dur": 13,
    "args": {
      "External id": 33817,
      "cbid": 211, "correlation": 33817
    }
  },
  {
    "ph": "s", "id": 33817, "pid": 76337, "tid": -914061504, "ts": 1716454218019152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218061867, "dur": 66,
    "args": {
      "External id": 33825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33825,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33825, "pid": 5, "tid": 7, "ts": 1716454218061867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019221, "dur": 12,
    "args": {
      "External id": 33825,
      "cbid": 211, "correlation": 33825
    }
  },
  {
    "ph": "s", "id": 33825, "pid": 76337, "tid": -914061504, "ts": 1716454218019221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218061934, "dur": 97,
    "args": {
      "External id": 33833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33833,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33833, "pid": 5, "tid": 7, "ts": 1716454218061934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019261, "dur": 8,
    "args": {
      "External id": 33833,
      "cbid": 211, "correlation": 33833
    }
  },
  {
    "ph": "s", "id": 33833, "pid": 76337, "tid": -914061504, "ts": 1716454218019261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218062033, "dur": 53,
    "args": {
      "External id": 33844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33844, "pid": 5, "tid": 7, "ts": 1716454218062033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019332, "dur": 13,
    "args": {
      "External id": 33844,
      "cbid": 211, "correlation": 33844
    }
  },
  {
    "ph": "s", "id": 33844, "pid": 76337, "tid": -914061504, "ts": 1716454218019332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218062086, "dur": 91,
    "args": {
      "External id": 33866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33866,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33866, "pid": 5, "tid": 7, "ts": 1716454218062086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019363, "dur": 8,
    "args": {
      "External id": 33866,
      "cbid": 211, "correlation": 33866
    }
  },
  {
    "ph": "s", "id": 33866, "pid": 76337, "tid": -914061504, "ts": 1716454218019363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019447, "dur": 1,
    "args": {
      "External id": 33877,
      "cbid": 251, "correlation": 33877
    }
  },
  {
    "ph": "f", "id": 33877, "pid": 76337, "tid": -914061504, "ts": 1716454218019447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218062179, "dur": 105,
    "args": {
      "External id": 33878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33878,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 33878, "pid": 5, "tid": 7, "ts": 1716454218062179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019452, "dur": 12,
    "args": {
      "External id": 33878,
      "cbid": 211, "correlation": 33878
    }
  },
  {
    "ph": "s", "id": 33878, "pid": 76337, "tid": -914061504, "ts": 1716454218019452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019522, "dur": 1,
    "args": {
      "External id": 33889,
      "cbid": 251, "correlation": 33889
    }
  },
  {
    "ph": "f", "id": 33889, "pid": 76337, "tid": -914061504, "ts": 1716454218019522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019526, "dur": 0,
    "args": {
      "External id": 33890,
      "cbid": 251, "correlation": 33890
    }
  },
  {
    "ph": "f", "id": 33890, "pid": 76337, "tid": -914061504, "ts": 1716454218019526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218062285, "dur": 10,
    "args": {
      "External id": 33891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33891,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 33891, "pid": 5, "tid": 7, "ts": 1716454218062285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019528, "dur": 12,
    "args": {
      "External id": 33891,
      "cbid": 211, "correlation": 33891
    }
  },
  {
    "ph": "s", "id": 33891, "pid": 76337, "tid": -914061504, "ts": 1716454218019528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218062297, "dur": 5,
    "args": {
      "External id": 33893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33893,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 33893, "pid": 5, "tid": 7, "ts": 1716454218062297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019542, "dur": 6,
    "args": {
      "External id": 33893,
      "cbid": 211, "correlation": 33893
    }
  },
  {
    "ph": "s", "id": 33893, "pid": 76337, "tid": -914061504, "ts": 1716454218019542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019603, "dur": 1,
    "args": {
      "External id": 33904,
      "cbid": 251, "correlation": 33904
    }
  },
  {
    "ph": "f", "id": 33904, "pid": 76337, "tid": -914061504, "ts": 1716454218019603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019607, "dur": 0,
    "args": {
      "External id": 33905,
      "cbid": 251, "correlation": 33905
    }
  },
  {
    "ph": "f", "id": 33905, "pid": 76337, "tid": -914061504, "ts": 1716454218019607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218062303, "dur": 6,
    "args": {
      "External id": 33906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33906,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 33906, "pid": 5, "tid": 7, "ts": 1716454218062303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019608, "dur": 11,
    "args": {
      "External id": 33906,
      "cbid": 211, "correlation": 33906
    }
  },
  {
    "ph": "s", "id": 33906, "pid": 76337, "tid": -914061504, "ts": 1716454218019608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218062310, "dur": 3,
    "args": {
      "External id": 33908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33908,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 33908, "pid": 5, "tid": 7, "ts": 1716454218062310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019621, "dur": 5,
    "args": {
      "External id": 33908,
      "cbid": 211, "correlation": 33908
    }
  },
  {
    "ph": "s", "id": 33908, "pid": 76337, "tid": -914061504, "ts": 1716454218019621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218062315, "dur": 152,
    "args": {
      "External id": 33929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33929,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33929, "pid": 5, "tid": 7, "ts": 1716454218062315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019694, "dur": 12,
    "args": {
      "External id": 33929,
      "cbid": 211, "correlation": 33929
    }
  },
  {
    "ph": "s", "id": 33929, "pid": 76337, "tid": -914061504, "ts": 1716454218019694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019790, "dur": 1,
    "args": {
      "External id": 33947,
      "cbid": 251, "correlation": 33947
    }
  },
  {
    "ph": "f", "id": 33947, "pid": 76337, "tid": -914061504, "ts": 1716454218019790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218062468, "dur": 105,
    "args": {
      "External id": 33949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33949,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 33949, "pid": 5, "tid": 7, "ts": 1716454218062468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019796, "dur": 13,
    "args": {
      "External id": 33949,
      "cbid": 211, "correlation": 33949
    }
  },
  {
    "ph": "s", "id": 33949, "pid": 76337, "tid": -914061504, "ts": 1716454218019796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218062574, "dur": 35,
    "args": {
      "External id": 33957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33957,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33957, "pid": 5, "tid": 7, "ts": 1716454218062574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019866, "dur": 12,
    "args": {
      "External id": 33957,
      "cbid": 211, "correlation": 33957
    }
  },
  {
    "ph": "s", "id": 33957, "pid": 76337, "tid": -914061504, "ts": 1716454218019866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218062611, "dur": 68,
    "args": {
      "External id": 33965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33965, "pid": 5, "tid": 7, "ts": 1716454218062611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019905, "dur": 10,
    "args": {
      "External id": 33965,
      "cbid": 211, "correlation": 33965
    }
  },
  {
    "ph": "s", "id": 33965, "pid": 76337, "tid": -914061504, "ts": 1716454218019905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218062680, "dur": 91,
    "args": {
      "External id": 33987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 33987,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 33987, "pid": 5, "tid": 7, "ts": 1716454218062680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218019957, "dur": 10,
    "args": {
      "External id": 33987,
      "cbid": 211, "correlation": 33987
    }
  },
  {
    "ph": "s", "id": 33987, "pid": 76337, "tid": -914061504, "ts": 1716454218019957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020053, "dur": 1,
    "args": {
      "External id": 34003,
      "cbid": 251, "correlation": 34003
    }
  },
  {
    "ph": "f", "id": 34003, "pid": 76337, "tid": -914061504, "ts": 1716454218020053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218062772, "dur": 570,
    "args": {
      "External id": 34005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34005,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34005, "pid": 5, "tid": 7, "ts": 1716454218062772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020058, "dur": 14,
    "args": {
      "External id": 34005,
      "cbid": 211, "correlation": 34005
    }
  },
  {
    "ph": "s", "id": 34005, "pid": 76337, "tid": -914061504, "ts": 1716454218020058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218063344, "dur": 243,
    "args": {
      "External id": 34013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34013, "pid": 5, "tid": 7, "ts": 1716454218063344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020125, "dur": 12,
    "args": {
      "External id": 34013,
      "cbid": 211, "correlation": 34013
    }
  },
  {
    "ph": "s", "id": 34013, "pid": 76337, "tid": -914061504, "ts": 1716454218020125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218063588, "dur": 249,
    "args": {
      "External id": 34021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34021,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34021, "pid": 5, "tid": 7, "ts": 1716454218063588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020154, "dur": 8,
    "args": {
      "External id": 34021,
      "cbid": 211, "correlation": 34021
    }
  },
  {
    "ph": "s", "id": 34021, "pid": 76337, "tid": -914061504, "ts": 1716454218020154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020236, "dur": 1,
    "args": {
      "External id": 34037,
      "cbid": 251, "correlation": 34037
    }
  },
  {
    "ph": "f", "id": 34037, "pid": 76337, "tid": -914061504, "ts": 1716454218020236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020241, "dur": 0,
    "args": {
      "External id": 34039,
      "cbid": 251, "correlation": 34039
    }
  },
  {
    "ph": "f", "id": 34039, "pid": 76337, "tid": -914061504, "ts": 1716454218020241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218063838, "dur": 357,
    "args": {
      "External id": 34040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34040,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 34040, "pid": 5, "tid": 7, "ts": 1716454218063838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020244, "dur": 13,
    "args": {
      "External id": 34040,
      "cbid": 211, "correlation": 34040
    }
  },
  {
    "ph": "s", "id": 34040, "pid": 76337, "tid": -914061504, "ts": 1716454218020244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218064197, "dur": 50,
    "args": {
      "External id": 34048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34048,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34048, "pid": 5, "tid": 7, "ts": 1716454218064197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020286, "dur": 10,
    "args": {
      "External id": 34048,
      "cbid": 211, "correlation": 34048
    }
  },
  {
    "ph": "s", "id": 34048, "pid": 76337, "tid": -914061504, "ts": 1716454218020286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218064248, "dur": 156,
    "args": {
      "External id": 34059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34059, "pid": 5, "tid": 7, "ts": 1716454218064248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020353, "dur": 12,
    "args": {
      "External id": 34059,
      "cbid": 211, "correlation": 34059
    }
  },
  {
    "ph": "s", "id": 34059, "pid": 76337, "tid": -914061504, "ts": 1716454218020353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020416, "dur": 0,
    "args": {
      "External id": 34071,
      "cbid": 317, "correlation": 34071
    }
  },
  {
    "ph": "f", "id": 34071, "pid": 76337, "tid": -914061504, "ts": 1716454218020416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020417, "dur": 0,
    "args": {
      "External id": 34072,
      "cbid": 203, "correlation": 34072
    }
  },
  {
    "ph": "f", "id": 34072, "pid": 76337, "tid": -914061504, "ts": 1716454218020417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020418, "dur": 0,
    "args": {
      "External id": 34073,
      "cbid": 205, "correlation": 34073
    }
  },
  {
    "ph": "f", "id": 34073, "pid": 76337, "tid": -914061504, "ts": 1716454218020418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020441, "dur": 1,
    "args": {
      "External id": 34077,
      "cbid": 251, "correlation": 34077
    }
  },
  {
    "ph": "f", "id": 34077, "pid": 76337, "tid": -914061504, "ts": 1716454218020441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020443, "dur": 0,
    "args": {
      "External id": 34078,
      "cbid": 251, "correlation": 34078
    }
  },
  {
    "ph": "f", "id": 34078, "pid": 76337, "tid": -914061504, "ts": 1716454218020443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020443, "dur": 0,
    "args": {
      "External id": 34079,
      "cbid": 251, "correlation": 34079
    }
  },
  {
    "ph": "f", "id": 34079, "pid": 76337, "tid": -914061504, "ts": 1716454218020443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020444, "dur": 0,
    "args": {
      "External id": 34080,
      "cbid": 251, "correlation": 34080
    }
  },
  {
    "ph": "f", "id": 34080, "pid": 76337, "tid": -914061504, "ts": 1716454218020444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020445, "dur": 0,
    "args": {
      "External id": 34081,
      "cbid": 251, "correlation": 34081
    }
  },
  {
    "ph": "f", "id": 34081, "pid": 76337, "tid": -914061504, "ts": 1716454218020445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020445, "dur": 0,
    "args": {
      "External id": 34082,
      "cbid": 251, "correlation": 34082
    }
  },
  {
    "ph": "f", "id": 34082, "pid": 76337, "tid": -914061504, "ts": 1716454218020445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020446, "dur": 0,
    "args": {
      "External id": 34083,
      "cbid": 251, "correlation": 34083
    }
  },
  {
    "ph": "f", "id": 34083, "pid": 76337, "tid": -914061504, "ts": 1716454218020446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020447, "dur": 0,
    "args": {
      "External id": 34084,
      "cbid": 251, "correlation": 34084
    }
  },
  {
    "ph": "f", "id": 34084, "pid": 76337, "tid": -914061504, "ts": 1716454218020447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020448, "dur": 0,
    "args": {
      "External id": 34085,
      "cbid": 251, "correlation": 34085
    }
  },
  {
    "ph": "f", "id": 34085, "pid": 76337, "tid": -914061504, "ts": 1716454218020448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218064406, "dur": 113,
    "args": {
      "External id": 34086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34086,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 34086, "pid": 5, "tid": 7, "ts": 1716454218064406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020450, "dur": 12,
    "args": {
      "External id": 34086,
      "cbid": 211, "correlation": 34086
    }
  },
  {
    "ph": "s", "id": 34086, "pid": 76337, "tid": -914061504, "ts": 1716454218020450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218064520, "dur": 59,
    "args": {
      "External id": 34092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34092, "pid": 5, "tid": 7, "ts": 1716454218064520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020485, "dur": 9,
    "args": {
      "External id": 34092,
      "cbid": 211, "correlation": 34092
    }
  },
  {
    "ph": "s", "id": 34092, "pid": 76337, "tid": -914061504, "ts": 1716454218020485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218064581, "dur": 49,
    "args": {
      "External id": 34100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34100,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34100, "pid": 5, "tid": 7, "ts": 1716454218064581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020516, "dur": 8,
    "args": {
      "External id": 34100,
      "cbid": 211, "correlation": 34100
    }
  },
  {
    "ph": "s", "id": 34100, "pid": 76337, "tid": -914061504, "ts": 1716454218020516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020589, "dur": 0,
    "args": {
      "External id": 34110,
      "cbid": 317, "correlation": 34110
    }
  },
  {
    "ph": "f", "id": 34110, "pid": 76337, "tid": -914061504, "ts": 1716454218020589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020590, "dur": 0,
    "args": {
      "External id": 34111,
      "cbid": 203, "correlation": 34111
    }
  },
  {
    "ph": "f", "id": 34111, "pid": 76337, "tid": -914061504, "ts": 1716454218020590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020591, "dur": 0,
    "args": {
      "External id": 34112,
      "cbid": 205, "correlation": 34112
    }
  },
  {
    "ph": "f", "id": 34112, "pid": 76337, "tid": -914061504, "ts": 1716454218020591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218064631, "dur": 42,
    "args": {
      "External id": 34116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34116, "pid": 5, "tid": 7, "ts": 1716454218064631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020608, "dur": 12,
    "args": {
      "External id": 34116,
      "cbid": 211, "correlation": 34116
    }
  },
  {
    "ph": "s", "id": 34116, "pid": 76337, "tid": -914061504, "ts": 1716454218020608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218064674, "dur": 14,
    "args": {
      "External id": 34118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34118, "pid": 5, "tid": 7, "ts": 1716454218064674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020623, "dur": 6,
    "args": {
      "External id": 34118,
      "cbid": 211, "correlation": 34118
    }
  },
  {
    "ph": "s", "id": 34118, "pid": 76337, "tid": -914061504, "ts": 1716454218020623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218064690, "dur": 1,
    "args": {
      "External id": 34120,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 34120,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.7868852459016393
    }
  },
  {
    "ph": "f", "id": 34120, "pid": 5, "tid": 7, "ts": 1716454218064690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020641, "dur": 17,
    "args": {
      "External id": 34120,
      "cbid": 51, "correlation": 34120
    }
  },
  {
    "ph": "s", "id": 34120, "pid": 76337, "tid": -914061504, "ts": 1716454218020641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218064694, "dur": 359,
    "args": {
      "External id": 34121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34121,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34121, "pid": 5, "tid": 7, "ts": 1716454218064694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020660, "dur": 9,
    "args": {
      "External id": 34121,
      "cbid": 211, "correlation": 34121
    }
  },
  {
    "ph": "s", "id": 34121, "pid": 76337, "tid": -914061504, "ts": 1716454218020660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065054, "dur": 13,
    "args": {
      "External id": 34123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34123, "pid": 5, "tid": 7, "ts": 1716454218065054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020678, "dur": 7,
    "args": {
      "External id": 34123,
      "cbid": 211, "correlation": 34123
    }
  },
  {
    "ph": "s", "id": 34123, "pid": 76337, "tid": -914061504, "ts": 1716454218020678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218065069, "dur": 15,
    "args": {
      "External id": 34129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34129, "pid": 5, "tid": 7, "ts": 1716454218065069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020708, "dur": 8,
    "args": {
      "External id": 34129,
      "cbid": 211, "correlation": 34129
    }
  },
  {
    "ph": "s", "id": 34129, "pid": 76337, "tid": -914061504, "ts": 1716454218020708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065085, "dur": 19,
    "args": {
      "External id": 34149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34149,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 34149, "pid": 5, "tid": 7, "ts": 1716454218065085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020800, "dur": 13,
    "args": {
      "External id": 34149,
      "cbid": 211, "correlation": 34149
    }
  },
  {
    "ph": "s", "id": 34149, "pid": 76337, "tid": -914061504, "ts": 1716454218020800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218065105, "dur": 5,
    "args": {
      "External id": 34161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34161,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 34161, "pid": 5, "tid": 7, "ts": 1716454218065105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020822, "dur": 6,
    "args": {
      "External id": 34161,
      "cbid": 211, "correlation": 34161
    }
  },
  {
    "ph": "s", "id": 34161, "pid": 76337, "tid": -914061504, "ts": 1716454218020822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218065112, "dur": 17,
    "args": {
      "External id": 34164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34164, "pid": 5, "tid": 7, "ts": 1716454218065112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020840, "dur": 7,
    "args": {
      "External id": 34164,
      "cbid": 211, "correlation": 34164
    }
  },
  {
    "ph": "s", "id": 34164, "pid": 76337, "tid": -914061504, "ts": 1716454218020840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218065130, "dur": 11,
    "args": {
      "External id": 34173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34173,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34173, "pid": 5, "tid": 7, "ts": 1716454218065130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020879, "dur": 9,
    "args": {
      "External id": 34173,
      "cbid": 211, "correlation": 34173
    }
  },
  {
    "ph": "s", "id": 34173, "pid": 76337, "tid": -914061504, "ts": 1716454218020879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020935, "dur": 0,
    "args": {
      "External id": 34183,
      "cbid": 317, "correlation": 34183
    }
  },
  {
    "ph": "f", "id": 34183, "pid": 76337, "tid": -914061504, "ts": 1716454218020935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020936, "dur": 0,
    "args": {
      "External id": 34184,
      "cbid": 203, "correlation": 34184
    }
  },
  {
    "ph": "f", "id": 34184, "pid": 76337, "tid": -914061504, "ts": 1716454218020936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020936, "dur": 0,
    "args": {
      "External id": 34185,
      "cbid": 205, "correlation": 34185
    }
  },
  {
    "ph": "f", "id": 34185, "pid": 76337, "tid": -914061504, "ts": 1716454218020936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065143, "dur": 11,
    "args": {
      "External id": 34189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34189,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34189, "pid": 5, "tid": 7, "ts": 1716454218065143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020951, "dur": 12,
    "args": {
      "External id": 34189,
      "cbid": 211, "correlation": 34189
    }
  },
  {
    "ph": "s", "id": 34189, "pid": 76337, "tid": -914061504, "ts": 1716454218020951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065155, "dur": 24,
    "args": {
      "External id": 34191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34191,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34191, "pid": 5, "tid": 7, "ts": 1716454218065155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020966, "dur": 5,
    "args": {
      "External id": 34191,
      "cbid": 211, "correlation": 34191
    }
  },
  {
    "ph": "s", "id": 34191, "pid": 76337, "tid": -914061504, "ts": 1716454218020966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218065180, "dur": 4,
    "args": {
      "External id": 34193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 34193, "pid": 5, "tid": 7, "ts": 1716454218065180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020987, "dur": 6,
    "args": {
      "External id": 34193,
      "cbid": 211, "correlation": 34193
    }
  },
  {
    "ph": "s", "id": 34193, "pid": 76337, "tid": -914061504, "ts": 1716454218020987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020997, "dur": 0,
    "args": {
      "External id": 34194,
      "cbid": 51, "correlation": 34194
    }
  },
  {
    "ph": "s", "id": 34194, "pid": 76337, "tid": -914061504, "ts": 1716454218020997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218065185, "dur": 354,
    "args": {
      "External id": 34195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34195,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34195, "pid": 5, "tid": 7, "ts": 1716454218065185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218020998, "dur": 7,
    "args": {
      "External id": 34195,
      "cbid": 211, "correlation": 34195
    }
  },
  {
    "ph": "s", "id": 34195, "pid": 76337, "tid": -914061504, "ts": 1716454218020998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065540, "dur": 19,
    "args": {
      "External id": 34196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34196,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34196, "pid": 5, "tid": 7, "ts": 1716454218065540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021008, "dur": 6,
    "args": {
      "External id": 34196,
      "cbid": 211, "correlation": 34196
    }
  },
  {
    "ph": "s", "id": 34196, "pid": 76337, "tid": -914061504, "ts": 1716454218021008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218065560, "dur": 33,
    "args": {
      "External id": 34202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34202, "pid": 5, "tid": 7, "ts": 1716454218065560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021037, "dur": 8,
    "args": {
      "External id": 34202,
      "cbid": 211, "correlation": 34202
    }
  },
  {
    "ph": "s", "id": 34202, "pid": 76337, "tid": -914061504, "ts": 1716454218021037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218065594, "dur": 3,
    "args": {
      "External id": 34210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34210,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 34210, "pid": 5, "tid": 7, "ts": 1716454218065594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021080, "dur": 9,
    "args": {
      "External id": 34210,
      "cbid": 211, "correlation": 34210
    }
  },
  {
    "ph": "s", "id": 34210, "pid": 76337, "tid": -914061504, "ts": 1716454218021080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021147, "dur": 1,
    "args": {
      "External id": 34226,
      "cbid": 251, "correlation": 34226
    }
  },
  {
    "ph": "f", "id": 34226, "pid": 76337, "tid": -914061504, "ts": 1716454218021147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021153, "dur": 0,
    "args": {
      "External id": 34228,
      "cbid": 251, "correlation": 34228
    }
  },
  {
    "ph": "f", "id": 34228, "pid": 76337, "tid": -914061504, "ts": 1716454218021153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218065599, "dur": 12,
    "args": {
      "External id": 34229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34229,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 34229, "pid": 5, "tid": 7, "ts": 1716454218065599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021154, "dur": 11,
    "args": {
      "External id": 34229,
      "cbid": 211, "correlation": 34229
    }
  },
  {
    "ph": "s", "id": 34229, "pid": 76337, "tid": -914061504, "ts": 1716454218021154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218065612, "dur": 5,
    "args": {
      "External id": 34231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34231,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 34231, "pid": 5, "tid": 7, "ts": 1716454218065612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021168, "dur": 6,
    "args": {
      "External id": 34231,
      "cbid": 211, "correlation": 34231
    }
  },
  {
    "ph": "s", "id": 34231, "pid": 76337, "tid": -914061504, "ts": 1716454218021168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218065619, "dur": 30,
    "args": {
      "External id": 34241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34241, "pid": 5, "tid": 7, "ts": 1716454218065619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021225, "dur": 12,
    "args": {
      "External id": 34241,
      "cbid": 211, "correlation": 34241
    }
  },
  {
    "ph": "s", "id": 34241, "pid": 76337, "tid": -914061504, "ts": 1716454218021225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065650, "dur": 31,
    "args": {
      "External id": 34261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34261,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 34261, "pid": 5, "tid": 7, "ts": 1716454218065650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021293, "dur": 10,
    "args": {
      "External id": 34261,
      "cbid": 211, "correlation": 34261
    }
  },
  {
    "ph": "s", "id": 34261, "pid": 76337, "tid": -914061504, "ts": 1716454218021293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218065682, "dur": 4,
    "args": {
      "External id": 34273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34273,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 34273, "pid": 5, "tid": 7, "ts": 1716454218065682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021313, "dur": 6,
    "args": {
      "External id": 34273,
      "cbid": 211, "correlation": 34273
    }
  },
  {
    "ph": "s", "id": 34273, "pid": 76337, "tid": -914061504, "ts": 1716454218021313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218065688, "dur": 30,
    "args": {
      "External id": 34276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34276, "pid": 5, "tid": 7, "ts": 1716454218065688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021332, "dur": 6,
    "args": {
      "External id": 34276,
      "cbid": 211, "correlation": 34276
    }
  },
  {
    "ph": "s", "id": 34276, "pid": 76337, "tid": -914061504, "ts": 1716454218021332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218065718, "dur": 20,
    "args": {
      "External id": 34285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34285,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34285, "pid": 5, "tid": 7, "ts": 1716454218065718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021373, "dur": 10,
    "args": {
      "External id": 34285,
      "cbid": 211, "correlation": 34285
    }
  },
  {
    "ph": "s", "id": 34285, "pid": 76337, "tid": -914061504, "ts": 1716454218021373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021437, "dur": 0,
    "args": {
      "External id": 34295,
      "cbid": 317, "correlation": 34295
    }
  },
  {
    "ph": "f", "id": 34295, "pid": 76337, "tid": -914061504, "ts": 1716454218021437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021438, "dur": 0,
    "args": {
      "External id": 34296,
      "cbid": 203, "correlation": 34296
    }
  },
  {
    "ph": "f", "id": 34296, "pid": 76337, "tid": -914061504, "ts": 1716454218021438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021439, "dur": 0,
    "args": {
      "External id": 34297,
      "cbid": 205, "correlation": 34297
    }
  },
  {
    "ph": "f", "id": 34297, "pid": 76337, "tid": -914061504, "ts": 1716454218021439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065740, "dur": 23,
    "args": {
      "External id": 34301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34301,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34301, "pid": 5, "tid": 7, "ts": 1716454218065740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021455, "dur": 12,
    "args": {
      "External id": 34301,
      "cbid": 211, "correlation": 34301
    }
  },
  {
    "ph": "s", "id": 34301, "pid": 76337, "tid": -914061504, "ts": 1716454218021455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218065764, "dur": 43,
    "args": {
      "External id": 34303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34303, "pid": 5, "tid": 7, "ts": 1716454218065764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021470, "dur": 5,
    "args": {
      "External id": 34303,
      "cbid": 211, "correlation": 34303
    }
  },
  {
    "ph": "s", "id": 34303, "pid": 76337, "tid": -914061504, "ts": 1716454218021470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218065809, "dur": 651,
    "args": {
      "External id": 34305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34305,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34305, "pid": 5, "tid": 7, "ts": 1716454218065809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021484, "dur": 9,
    "args": {
      "External id": 34305,
      "cbid": 211, "correlation": 34305
    }
  },
  {
    "ph": "s", "id": 34305, "pid": 76337, "tid": -914061504, "ts": 1716454218021484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218066461, "dur": 20,
    "args": {
      "External id": 34307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34307,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34307, "pid": 5, "tid": 7, "ts": 1716454218066461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021496, "dur": 5,
    "args": {
      "External id": 34307,
      "cbid": 211, "correlation": 34307
    }
  },
  {
    "ph": "s", "id": 34307, "pid": 76337, "tid": -914061504, "ts": 1716454218021496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218066483, "dur": 33,
    "args": {
      "External id": 34313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34313, "pid": 5, "tid": 7, "ts": 1716454218066483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021524, "dur": 9,
    "args": {
      "External id": 34313,
      "cbid": 211, "correlation": 34313
    }
  },
  {
    "ph": "s", "id": 34313, "pid": 76337, "tid": -914061504, "ts": 1716454218021524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021583, "dur": 0,
    "args": {
      "External id": 34323,
      "cbid": 317, "correlation": 34323
    }
  },
  {
    "ph": "f", "id": 34323, "pid": 76337, "tid": -914061504, "ts": 1716454218021583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021584, "dur": 0,
    "args": {
      "External id": 34324,
      "cbid": 203, "correlation": 34324
    }
  },
  {
    "ph": "f", "id": 34324, "pid": 76337, "tid": -914061504, "ts": 1716454218021584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021585, "dur": 0,
    "args": {
      "External id": 34325,
      "cbid": 205, "correlation": 34325
    }
  },
  {
    "ph": "f", "id": 34325, "pid": 76337, "tid": -914061504, "ts": 1716454218021585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021606, "dur": 1,
    "args": {
      "External id": 34329,
      "cbid": 251, "correlation": 34329
    }
  },
  {
    "ph": "f", "id": 34329, "pid": 76337, "tid": -914061504, "ts": 1716454218021606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021608, "dur": 0,
    "args": {
      "External id": 34330,
      "cbid": 251, "correlation": 34330
    }
  },
  {
    "ph": "f", "id": 34330, "pid": 76337, "tid": -914061504, "ts": 1716454218021608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021609, "dur": 0,
    "args": {
      "External id": 34331,
      "cbid": 251, "correlation": 34331
    }
  },
  {
    "ph": "f", "id": 34331, "pid": 76337, "tid": -914061504, "ts": 1716454218021609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021609, "dur": 0,
    "args": {
      "External id": 34332,
      "cbid": 251, "correlation": 34332
    }
  },
  {
    "ph": "f", "id": 34332, "pid": 76337, "tid": -914061504, "ts": 1716454218021609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021610, "dur": 0,
    "args": {
      "External id": 34333,
      "cbid": 251, "correlation": 34333
    }
  },
  {
    "ph": "f", "id": 34333, "pid": 76337, "tid": -914061504, "ts": 1716454218021610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021611, "dur": 0,
    "args": {
      "External id": 34334,
      "cbid": 251, "correlation": 34334
    }
  },
  {
    "ph": "f", "id": 34334, "pid": 76337, "tid": -914061504, "ts": 1716454218021611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021612, "dur": 0,
    "args": {
      "External id": 34335,
      "cbid": 251, "correlation": 34335
    }
  },
  {
    "ph": "f", "id": 34335, "pid": 76337, "tid": -914061504, "ts": 1716454218021612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021612, "dur": 0,
    "args": {
      "External id": 34336,
      "cbid": 251, "correlation": 34336
    }
  },
  {
    "ph": "f", "id": 34336, "pid": 76337, "tid": -914061504, "ts": 1716454218021612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021614, "dur": 0,
    "args": {
      "External id": 34337,
      "cbid": 251, "correlation": 34337
    }
  },
  {
    "ph": "f", "id": 34337, "pid": 76337, "tid": -914061504, "ts": 1716454218021614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218066516, "dur": 50,
    "args": {
      "External id": 34338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34338,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 34338, "pid": 5, "tid": 7, "ts": 1716454218066516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021616, "dur": 12,
    "args": {
      "External id": 34338,
      "cbid": 211, "correlation": 34338
    }
  },
  {
    "ph": "s", "id": 34338, "pid": 76337, "tid": -914061504, "ts": 1716454218021616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218066568, "dur": 32,
    "args": {
      "External id": 34344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34344,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34344, "pid": 5, "tid": 7, "ts": 1716454218066568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021648, "dur": 8,
    "args": {
      "External id": 34344,
      "cbid": 211, "correlation": 34344
    }
  },
  {
    "ph": "s", "id": 34344, "pid": 76337, "tid": -914061504, "ts": 1716454218021648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218066601, "dur": 27,
    "args": {
      "External id": 34352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34352,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34352, "pid": 5, "tid": 7, "ts": 1716454218066601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021677, "dur": 8,
    "args": {
      "External id": 34352,
      "cbid": 211, "correlation": 34352
    }
  },
  {
    "ph": "s", "id": 34352, "pid": 76337, "tid": -914061504, "ts": 1716454218021677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218066629, "dur": 20,
    "args": {
      "External id": 34360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34360,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34360, "pid": 5, "tid": 7, "ts": 1716454218066629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021707, "dur": 8,
    "args": {
      "External id": 34360,
      "cbid": 211, "correlation": 34360
    }
  },
  {
    "ph": "s", "id": 34360, "pid": 76337, "tid": -914061504, "ts": 1716454218021707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218066651, "dur": 30,
    "args": {
      "External id": 34380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34380,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 34380, "pid": 5, "tid": 7, "ts": 1716454218066651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021789, "dur": 13,
    "args": {
      "External id": 34380,
      "cbid": 211, "correlation": 34380
    }
  },
  {
    "ph": "s", "id": 34380, "pid": 76337, "tid": -914061504, "ts": 1716454218021789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218066682, "dur": 4,
    "args": {
      "External id": 34392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34392,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 34392, "pid": 5, "tid": 7, "ts": 1716454218066682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021811, "dur": 6,
    "args": {
      "External id": 34392,
      "cbid": 211, "correlation": 34392
    }
  },
  {
    "ph": "s", "id": 34392, "pid": 76337, "tid": -914061504, "ts": 1716454218021811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218066687, "dur": 29,
    "args": {
      "External id": 34395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34395, "pid": 5, "tid": 7, "ts": 1716454218066687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021829, "dur": 7,
    "args": {
      "External id": 34395,
      "cbid": 211, "correlation": 34395
    }
  },
  {
    "ph": "s", "id": 34395, "pid": 76337, "tid": -914061504, "ts": 1716454218021829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021887, "dur": 0,
    "args": {
      "External id": 34406,
      "cbid": 317, "correlation": 34406
    }
  },
  {
    "ph": "f", "id": 34406, "pid": 76337, "tid": -914061504, "ts": 1716454218021887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021888, "dur": 0,
    "args": {
      "External id": 34407,
      "cbid": 203, "correlation": 34407
    }
  },
  {
    "ph": "f", "id": 34407, "pid": 76337, "tid": -914061504, "ts": 1716454218021888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021888, "dur": 0,
    "args": {
      "External id": 34408,
      "cbid": 205, "correlation": 34408
    }
  },
  {
    "ph": "f", "id": 34408, "pid": 76337, "tid": -914061504, "ts": 1716454218021888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218066717, "dur": 22,
    "args": {
      "External id": 34412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34412, "pid": 5, "tid": 7, "ts": 1716454218066717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021905, "dur": 12,
    "args": {
      "External id": 34412,
      "cbid": 211, "correlation": 34412
    }
  },
  {
    "ph": "s", "id": 34412, "pid": 76337, "tid": -914061504, "ts": 1716454218021905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218066741, "dur": 120,
    "args": {
      "External id": 34414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34414,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34414, "pid": 5, "tid": 7, "ts": 1716454218066741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021925, "dur": 8,
    "args": {
      "External id": 34414,
      "cbid": 211, "correlation": 34414
    }
  },
  {
    "ph": "s", "id": 34414, "pid": 76337, "tid": -914061504, "ts": 1716454218021925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218066862, "dur": 22,
    "args": {
      "External id": 34416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34416,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34416, "pid": 5, "tid": 7, "ts": 1716454218066862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021937, "dur": 5,
    "args": {
      "External id": 34416,
      "cbid": 211, "correlation": 34416
    }
  },
  {
    "ph": "s", "id": 34416, "pid": 76337, "tid": -914061504, "ts": 1716454218021937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218066885, "dur": 32,
    "args": {
      "External id": 34422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34422, "pid": 5, "tid": 7, "ts": 1716454218066885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218021965, "dur": 17,
    "args": {
      "External id": 34422,
      "cbid": 211, "correlation": 34422
    }
  },
  {
    "ph": "s", "id": 34422, "pid": 76337, "tid": -914061504, "ts": 1716454218021965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218066918, "dur": 202,
    "args": {
      "External id": 34431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34431,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34431, "pid": 5, "tid": 7, "ts": 1716454218066918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022057, "dur": 15,
    "args": {
      "External id": 34431,
      "cbid": 211, "correlation": 34431
    }
  },
  {
    "ph": "s", "id": 34431, "pid": 76337, "tid": -914061504, "ts": 1716454218022057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218067121, "dur": 65,
    "args": {
      "External id": 34453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34453,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34453, "pid": 5, "tid": 7, "ts": 1716454218067121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022113, "dur": 10,
    "args": {
      "External id": 34453,
      "cbid": 211, "correlation": 34453
    }
  },
  {
    "ph": "s", "id": 34453, "pid": 76337, "tid": -914061504, "ts": 1716454218022113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022202, "dur": 1,
    "args": {
      "External id": 34464,
      "cbid": 251, "correlation": 34464
    }
  },
  {
    "ph": "f", "id": 34464, "pid": 76337, "tid": -914061504, "ts": 1716454218022202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218067188, "dur": 153,
    "args": {
      "External id": 34465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34465,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34465, "pid": 5, "tid": 7, "ts": 1716454218067188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022207, "dur": 14,
    "args": {
      "External id": 34465,
      "cbid": 211, "correlation": 34465
    }
  },
  {
    "ph": "s", "id": 34465, "pid": 76337, "tid": -914061504, "ts": 1716454218022207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022278, "dur": 1,
    "args": {
      "External id": 34476,
      "cbid": 251, "correlation": 34476
    }
  },
  {
    "ph": "f", "id": 34476, "pid": 76337, "tid": -914061504, "ts": 1716454218022278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218067342, "dur": 146,
    "args": {
      "External id": 34477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34477,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34477, "pid": 5, "tid": 7, "ts": 1716454218067342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022281, "dur": 11,
    "args": {
      "External id": 34477,
      "cbid": 211, "correlation": 34477
    }
  },
  {
    "ph": "s", "id": 34477, "pid": 76337, "tid": -914061504, "ts": 1716454218022281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022346, "dur": 1,
    "args": {
      "External id": 34488,
      "cbid": 251, "correlation": 34488
    }
  },
  {
    "ph": "f", "id": 34488, "pid": 76337, "tid": -914061504, "ts": 1716454218022346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218067490, "dur": 143,
    "args": {
      "External id": 34489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34489,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34489, "pid": 5, "tid": 7, "ts": 1716454218067490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022350, "dur": 11,
    "args": {
      "External id": 34489,
      "cbid": 211, "correlation": 34489
    }
  },
  {
    "ph": "s", "id": 34489, "pid": 76337, "tid": -914061504, "ts": 1716454218022350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218067634, "dur": 1926,
    "args": {
      "External id": 34510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34510,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 34510, "pid": 5, "tid": 7, "ts": 1716454218067634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022432, "dur": 13,
    "args": {
      "External id": 34510,
      "cbid": 211, "correlation": 34510
    }
  },
  {
    "ph": "s", "id": 34510, "pid": 76337, "tid": -914061504, "ts": 1716454218022432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022533, "dur": 1,
    "args": {
      "External id": 34528,
      "cbid": 251, "correlation": 34528
    }
  },
  {
    "ph": "f", "id": 34528, "pid": 76337, "tid": -914061504, "ts": 1716454218022533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218069562, "dur": 146,
    "args": {
      "External id": 34530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34530,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 34530, "pid": 5, "tid": 7, "ts": 1716454218069562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022539, "dur": 13,
    "args": {
      "External id": 34530,
      "cbid": 211, "correlation": 34530
    }
  },
  {
    "ph": "s", "id": 34530, "pid": 76337, "tid": -914061504, "ts": 1716454218022539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218069709, "dur": 35,
    "args": {
      "External id": 34538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34538,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34538, "pid": 5, "tid": 7, "ts": 1716454218069709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022609, "dur": 12,
    "args": {
      "External id": 34538,
      "cbid": 211, "correlation": 34538
    }
  },
  {
    "ph": "s", "id": 34538, "pid": 76337, "tid": -914061504, "ts": 1716454218022609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218069745, "dur": 51,
    "args": {
      "External id": 34546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34546,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34546, "pid": 5, "tid": 7, "ts": 1716454218069745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022648, "dur": 9,
    "args": {
      "External id": 34546,
      "cbid": 211, "correlation": 34546
    }
  },
  {
    "ph": "s", "id": 34546, "pid": 76337, "tid": -914061504, "ts": 1716454218022648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218069797, "dur": 30,
    "args": {
      "External id": 34557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34557, "pid": 5, "tid": 7, "ts": 1716454218069797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022720, "dur": 13,
    "args": {
      "External id": 34557,
      "cbid": 211, "correlation": 34557
    }
  },
  {
    "ph": "s", "id": 34557, "pid": 76337, "tid": -914061504, "ts": 1716454218022720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218069829, "dur": 34,
    "args": {
      "External id": 34579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34579,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34579, "pid": 5, "tid": 7, "ts": 1716454218069829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022752, "dur": 8,
    "args": {
      "External id": 34579,
      "cbid": 211, "correlation": 34579
    }
  },
  {
    "ph": "s", "id": 34579, "pid": 76337, "tid": -914061504, "ts": 1716454218022752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022836, "dur": 1,
    "args": {
      "External id": 34590,
      "cbid": 251, "correlation": 34590
    }
  },
  {
    "ph": "f", "id": 34590, "pid": 76337, "tid": -914061504, "ts": 1716454218022836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218069864, "dur": 91,
    "args": {
      "External id": 34591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34591,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34591, "pid": 5, "tid": 7, "ts": 1716454218069864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022842, "dur": 13,
    "args": {
      "External id": 34591,
      "cbid": 211, "correlation": 34591
    }
  },
  {
    "ph": "s", "id": 34591, "pid": 76337, "tid": -914061504, "ts": 1716454218022842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022911, "dur": 1,
    "args": {
      "External id": 34602,
      "cbid": 251, "correlation": 34602
    }
  },
  {
    "ph": "f", "id": 34602, "pid": 76337, "tid": -914061504, "ts": 1716454218022911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022915, "dur": 0,
    "args": {
      "External id": 34603,
      "cbid": 251, "correlation": 34603
    }
  },
  {
    "ph": "f", "id": 34603, "pid": 76337, "tid": -914061504, "ts": 1716454218022915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218069957, "dur": 11,
    "args": {
      "External id": 34604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34604,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 34604, "pid": 5, "tid": 7, "ts": 1716454218069957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022916, "dur": 11,
    "args": {
      "External id": 34604,
      "cbid": 211, "correlation": 34604
    }
  },
  {
    "ph": "s", "id": 34604, "pid": 76337, "tid": -914061504, "ts": 1716454218022916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218069969, "dur": 5,
    "args": {
      "External id": 34606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34606,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 34606, "pid": 5, "tid": 7, "ts": 1716454218069969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022932, "dur": 7,
    "args": {
      "External id": 34606,
      "cbid": 211, "correlation": 34606
    }
  },
  {
    "ph": "s", "id": 34606, "pid": 76337, "tid": -914061504, "ts": 1716454218022932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218022998, "dur": 1,
    "args": {
      "External id": 34617,
      "cbid": 251, "correlation": 34617
    }
  },
  {
    "ph": "f", "id": 34617, "pid": 76337, "tid": -914061504, "ts": 1716454218022998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023002, "dur": 0,
    "args": {
      "External id": 34618,
      "cbid": 251, "correlation": 34618
    }
  },
  {
    "ph": "f", "id": 34618, "pid": 76337, "tid": -914061504, "ts": 1716454218023002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218069975, "dur": 7,
    "args": {
      "External id": 34619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34619,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 34619, "pid": 5, "tid": 7, "ts": 1716454218069975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023003, "dur": 13,
    "args": {
      "External id": 34619,
      "cbid": 211, "correlation": 34619
    }
  },
  {
    "ph": "s", "id": 34619, "pid": 76337, "tid": -914061504, "ts": 1716454218023003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218069984, "dur": 3,
    "args": {
      "External id": 34621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34621,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 34621, "pid": 5, "tid": 7, "ts": 1716454218069984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023018, "dur": 5,
    "args": {
      "External id": 34621,
      "cbid": 211, "correlation": 34621
    }
  },
  {
    "ph": "s", "id": 34621, "pid": 76337, "tid": -914061504, "ts": 1716454218023018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218069988, "dur": 91,
    "args": {
      "External id": 34642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34642,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 34642, "pid": 5, "tid": 7, "ts": 1716454218069988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023091, "dur": 12,
    "args": {
      "External id": 34642,
      "cbid": 211, "correlation": 34642
    }
  },
  {
    "ph": "s", "id": 34642, "pid": 76337, "tid": -914061504, "ts": 1716454218023091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023188, "dur": 1,
    "args": {
      "External id": 34660,
      "cbid": 251, "correlation": 34660
    }
  },
  {
    "ph": "f", "id": 34660, "pid": 76337, "tid": -914061504, "ts": 1716454218023188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218070080, "dur": 97,
    "args": {
      "External id": 34662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34662,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34662, "pid": 5, "tid": 7, "ts": 1716454218070080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023193, "dur": 13,
    "args": {
      "External id": 34662,
      "cbid": 211, "correlation": 34662
    }
  },
  {
    "ph": "s", "id": 34662, "pid": 76337, "tid": -914061504, "ts": 1716454218023193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218070179, "dur": 19,
    "args": {
      "External id": 34670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34670,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34670, "pid": 5, "tid": 7, "ts": 1716454218070179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023261, "dur": 13,
    "args": {
      "External id": 34670,
      "cbid": 211, "correlation": 34670
    }
  },
  {
    "ph": "s", "id": 34670, "pid": 76337, "tid": -914061504, "ts": 1716454218023261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218070199, "dur": 36,
    "args": {
      "External id": 34678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34678,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34678, "pid": 5, "tid": 7, "ts": 1716454218070199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023303, "dur": 9,
    "args": {
      "External id": 34678,
      "cbid": 211, "correlation": 34678
    }
  },
  {
    "ph": "s", "id": 34678, "pid": 76337, "tid": -914061504, "ts": 1716454218023303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218070237, "dur": 35,
    "args": {
      "External id": 34700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34700,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34700, "pid": 5, "tid": 7, "ts": 1716454218070237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023355, "dur": 10,
    "args": {
      "External id": 34700,
      "cbid": 211, "correlation": 34700
    }
  },
  {
    "ph": "s", "id": 34700, "pid": 76337, "tid": -914061504, "ts": 1716454218023355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023445, "dur": 1,
    "args": {
      "External id": 34716,
      "cbid": 251, "correlation": 34716
    }
  },
  {
    "ph": "f", "id": 34716, "pid": 76337, "tid": -914061504, "ts": 1716454218023445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023450, "dur": 0,
    "args": {
      "External id": 34718,
      "cbid": 251, "correlation": 34718
    }
  },
  {
    "ph": "f", "id": 34718, "pid": 76337, "tid": -914061504, "ts": 1716454218023450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218070273, "dur": 537,
    "args": {
      "External id": 34719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34719,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 34719, "pid": 5, "tid": 7, "ts": 1716454218070273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023453, "dur": 13,
    "args": {
      "External id": 34719,
      "cbid": 211, "correlation": 34719
    }
  },
  {
    "ph": "s", "id": 34719, "pid": 76337, "tid": -914061504, "ts": 1716454218023453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218070811, "dur": 127,
    "args": {
      "External id": 34727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34727, "pid": 5, "tid": 7, "ts": 1716454218070811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023517, "dur": 13,
    "args": {
      "External id": 34727,
      "cbid": 211, "correlation": 34727
    }
  },
  {
    "ph": "s", "id": 34727, "pid": 76337, "tid": -914061504, "ts": 1716454218023517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218070939, "dur": 128,
    "args": {
      "External id": 34735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34735, "pid": 5, "tid": 7, "ts": 1716454218070939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023549, "dur": 8,
    "args": {
      "External id": 34735,
      "cbid": 211, "correlation": 34735
    }
  },
  {
    "ph": "s", "id": 34735, "pid": 76337, "tid": -914061504, "ts": 1716454218023549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023628, "dur": 1,
    "args": {
      "External id": 34751,
      "cbid": 251, "correlation": 34751
    }
  },
  {
    "ph": "f", "id": 34751, "pid": 76337, "tid": -914061504, "ts": 1716454218023628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218071068, "dur": 304,
    "args": {
      "External id": 34753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34753,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34753, "pid": 5, "tid": 7, "ts": 1716454218071068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023634, "dur": 12,
    "args": {
      "External id": 34753,
      "cbid": 211, "correlation": 34753
    }
  },
  {
    "ph": "s", "id": 34753, "pid": 76337, "tid": -914061504, "ts": 1716454218023634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218071373, "dur": 27,
    "args": {
      "External id": 34761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34761,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34761, "pid": 5, "tid": 7, "ts": 1716454218071373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023675, "dur": 10,
    "args": {
      "External id": 34761,
      "cbid": 211, "correlation": 34761
    }
  },
  {
    "ph": "s", "id": 34761, "pid": 76337, "tid": -914061504, "ts": 1716454218023675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218071402, "dur": 81,
    "args": {
      "External id": 34772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34772, "pid": 5, "tid": 7, "ts": 1716454218071402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023744, "dur": 12,
    "args": {
      "External id": 34772,
      "cbid": 211, "correlation": 34772
    }
  },
  {
    "ph": "s", "id": 34772, "pid": 76337, "tid": -914061504, "ts": 1716454218023744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023807, "dur": 0,
    "args": {
      "External id": 34784,
      "cbid": 317, "correlation": 34784
    }
  },
  {
    "ph": "f", "id": 34784, "pid": 76337, "tid": -914061504, "ts": 1716454218023807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023808, "dur": 0,
    "args": {
      "External id": 34785,
      "cbid": 203, "correlation": 34785
    }
  },
  {
    "ph": "f", "id": 34785, "pid": 76337, "tid": -914061504, "ts": 1716454218023808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023809, "dur": 0,
    "args": {
      "External id": 34786,
      "cbid": 205, "correlation": 34786
    }
  },
  {
    "ph": "f", "id": 34786, "pid": 76337, "tid": -914061504, "ts": 1716454218023809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218071484, "dur": 23,
    "args": {
      "External id": 34790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34790, "pid": 5, "tid": 7, "ts": 1716454218071484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023823, "dur": 11,
    "args": {
      "External id": 34790,
      "cbid": 211, "correlation": 34790
    }
  },
  {
    "ph": "s", "id": 34790, "pid": 76337, "tid": -914061504, "ts": 1716454218023823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218071508, "dur": 118,
    "args": {
      "External id": 34792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34792,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34792, "pid": 5, "tid": 7, "ts": 1716454218071508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023842, "dur": 7,
    "args": {
      "External id": 34792,
      "cbid": 211, "correlation": 34792
    }
  },
  {
    "ph": "s", "id": 34792, "pid": 76337, "tid": -914061504, "ts": 1716454218023842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218071627, "dur": 22,
    "args": {
      "External id": 34794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34794,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34794, "pid": 5, "tid": 7, "ts": 1716454218071627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023853, "dur": 5,
    "args": {
      "External id": 34794,
      "cbid": 211, "correlation": 34794
    }
  },
  {
    "ph": "s", "id": 34794, "pid": 76337, "tid": -914061504, "ts": 1716454218023853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218071651, "dur": 33,
    "args": {
      "External id": 34800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34800, "pid": 5, "tid": 7, "ts": 1716454218071651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023880, "dur": 8,
    "args": {
      "External id": 34800,
      "cbid": 211, "correlation": 34800
    }
  },
  {
    "ph": "s", "id": 34800, "pid": 76337, "tid": -914061504, "ts": 1716454218023880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218071685, "dur": 27,
    "args": {
      "External id": 34808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34808,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34808, "pid": 5, "tid": 7, "ts": 1716454218071685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023911, "dur": 8,
    "args": {
      "External id": 34808,
      "cbid": 211, "correlation": 34808
    }
  },
  {
    "ph": "s", "id": 34808, "pid": 76337, "tid": -914061504, "ts": 1716454218023911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218071712, "dur": 31,
    "args": {
      "External id": 34828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34828,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 34828, "pid": 5, "tid": 7, "ts": 1716454218071712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218023991, "dur": 13,
    "args": {
      "External id": 34828,
      "cbid": 211, "correlation": 34828
    }
  },
  {
    "ph": "s", "id": 34828, "pid": 76337, "tid": -914061504, "ts": 1716454218023991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218071744, "dur": 5,
    "args": {
      "External id": 34840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34840,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 34840, "pid": 5, "tid": 7, "ts": 1716454218071744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024014, "dur": 6,
    "args": {
      "External id": 34840,
      "cbid": 211, "correlation": 34840
    }
  },
  {
    "ph": "s", "id": 34840, "pid": 76337, "tid": -914061504, "ts": 1716454218024014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218071750, "dur": 32,
    "args": {
      "External id": 34843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34843,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34843, "pid": 5, "tid": 7, "ts": 1716454218071750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024031, "dur": 6,
    "args": {
      "External id": 34843,
      "cbid": 211, "correlation": 34843
    }
  },
  {
    "ph": "s", "id": 34843, "pid": 76337, "tid": -914061504, "ts": 1716454218024031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218071783, "dur": 21,
    "args": {
      "External id": 34852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34852,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34852, "pid": 5, "tid": 7, "ts": 1716454218071783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024071, "dur": 9,
    "args": {
      "External id": 34852,
      "cbid": 211, "correlation": 34852
    }
  },
  {
    "ph": "s", "id": 34852, "pid": 76337, "tid": -914061504, "ts": 1716454218024071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024121, "dur": 0,
    "args": {
      "External id": 34862,
      "cbid": 317, "correlation": 34862
    }
  },
  {
    "ph": "f", "id": 34862, "pid": 76337, "tid": -914061504, "ts": 1716454218024121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024122, "dur": 0,
    "args": {
      "External id": 34863,
      "cbid": 203, "correlation": 34863
    }
  },
  {
    "ph": "f", "id": 34863, "pid": 76337, "tid": -914061504, "ts": 1716454218024122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024123, "dur": 0,
    "args": {
      "External id": 34864,
      "cbid": 205, "correlation": 34864
    }
  },
  {
    "ph": "f", "id": 34864, "pid": 76337, "tid": -914061504, "ts": 1716454218024123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218071806, "dur": 23,
    "args": {
      "External id": 34868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34868, "pid": 5, "tid": 7, "ts": 1716454218071806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024136, "dur": 11,
    "args": {
      "External id": 34868,
      "cbid": 211, "correlation": 34868
    }
  },
  {
    "ph": "s", "id": 34868, "pid": 76337, "tid": -914061504, "ts": 1716454218024136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218071830, "dur": 44,
    "args": {
      "External id": 34870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34870, "pid": 5, "tid": 7, "ts": 1716454218071830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024150, "dur": 6,
    "args": {
      "External id": 34870,
      "cbid": 211, "correlation": 34870
    }
  },
  {
    "ph": "s", "id": 34870, "pid": 76337, "tid": -914061504, "ts": 1716454218024150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218071876, "dur": 650,
    "args": {
      "External id": 34872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34872,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34872, "pid": 5, "tid": 7, "ts": 1716454218071876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024162, "dur": 6,
    "args": {
      "External id": 34872,
      "cbid": 211, "correlation": 34872
    }
  },
  {
    "ph": "s", "id": 34872, "pid": 76337, "tid": -914061504, "ts": 1716454218024162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218072527, "dur": 22,
    "args": {
      "External id": 34874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34874,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34874, "pid": 5, "tid": 7, "ts": 1716454218072527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024172, "dur": 5,
    "args": {
      "External id": 34874,
      "cbid": 211, "correlation": 34874
    }
  },
  {
    "ph": "s", "id": 34874, "pid": 76337, "tid": -914061504, "ts": 1716454218024172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218072550, "dur": 32,
    "args": {
      "External id": 34880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34880, "pid": 5, "tid": 7, "ts": 1716454218072550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024198, "dur": 9,
    "args": {
      "External id": 34880,
      "cbid": 211, "correlation": 34880
    }
  },
  {
    "ph": "s", "id": 34880, "pid": 76337, "tid": -914061504, "ts": 1716454218024198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218072584, "dur": 3,
    "args": {
      "External id": 34888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34888,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 34888, "pid": 5, "tid": 7, "ts": 1716454218072584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024242, "dur": 9,
    "args": {
      "External id": 34888,
      "cbid": 211, "correlation": 34888
    }
  },
  {
    "ph": "s", "id": 34888, "pid": 76337, "tid": -914061504, "ts": 1716454218024242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024306, "dur": 1,
    "args": {
      "External id": 34904,
      "cbid": 251, "correlation": 34904
    }
  },
  {
    "ph": "f", "id": 34904, "pid": 76337, "tid": -914061504, "ts": 1716454218024306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024311, "dur": 0,
    "args": {
      "External id": 34906,
      "cbid": 251, "correlation": 34906
    }
  },
  {
    "ph": "f", "id": 34906, "pid": 76337, "tid": -914061504, "ts": 1716454218024311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218072589, "dur": 12,
    "args": {
      "External id": 34907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34907,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 34907, "pid": 5, "tid": 7, "ts": 1716454218072589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024313, "dur": 11,
    "args": {
      "External id": 34907,
      "cbid": 211, "correlation": 34907
    }
  },
  {
    "ph": "s", "id": 34907, "pid": 76337, "tid": -914061504, "ts": 1716454218024313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218072602, "dur": 5,
    "args": {
      "External id": 34909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34909,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 34909, "pid": 5, "tid": 7, "ts": 1716454218072602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024326, "dur": 6,
    "args": {
      "External id": 34909,
      "cbid": 211, "correlation": 34909
    }
  },
  {
    "ph": "s", "id": 34909, "pid": 76337, "tid": -914061504, "ts": 1716454218024326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218072609, "dur": 29,
    "args": {
      "External id": 34919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34919,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34919, "pid": 5, "tid": 7, "ts": 1716454218072609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024384, "dur": 12,
    "args": {
      "External id": 34919,
      "cbid": 211, "correlation": 34919
    }
  },
  {
    "ph": "s", "id": 34919, "pid": 76337, "tid": -914061504, "ts": 1716454218024384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218072639, "dur": 31,
    "args": {
      "External id": 34939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34939,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 34939, "pid": 5, "tid": 7, "ts": 1716454218072639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024450, "dur": 10,
    "args": {
      "External id": 34939,
      "cbid": 211, "correlation": 34939
    }
  },
  {
    "ph": "s", "id": 34939, "pid": 76337, "tid": -914061504, "ts": 1716454218024450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218072671, "dur": 4,
    "args": {
      "External id": 34951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34951,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 34951, "pid": 5, "tid": 7, "ts": 1716454218072671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024470, "dur": 7,
    "args": {
      "External id": 34951,
      "cbid": 211, "correlation": 34951
    }
  },
  {
    "ph": "s", "id": 34951, "pid": 76337, "tid": -914061504, "ts": 1716454218024470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218072676, "dur": 29,
    "args": {
      "External id": 34954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34954, "pid": 5, "tid": 7, "ts": 1716454218072676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024489, "dur": 6,
    "args": {
      "External id": 34954,
      "cbid": 211, "correlation": 34954
    }
  },
  {
    "ph": "s", "id": 34954, "pid": 76337, "tid": -914061504, "ts": 1716454218024489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218072706, "dur": 20,
    "args": {
      "External id": 34963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34963,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34963, "pid": 5, "tid": 7, "ts": 1716454218072706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024529, "dur": 10,
    "args": {
      "External id": 34963,
      "cbid": 211, "correlation": 34963
    }
  },
  {
    "ph": "s", "id": 34963, "pid": 76337, "tid": -914061504, "ts": 1716454218024529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024591, "dur": 0,
    "args": {
      "External id": 34973,
      "cbid": 317, "correlation": 34973
    }
  },
  {
    "ph": "f", "id": 34973, "pid": 76337, "tid": -914061504, "ts": 1716454218024591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024592, "dur": 0,
    "args": {
      "External id": 34974,
      "cbid": 203, "correlation": 34974
    }
  },
  {
    "ph": "f", "id": 34974, "pid": 76337, "tid": -914061504, "ts": 1716454218024592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024592, "dur": 0,
    "args": {
      "External id": 34975,
      "cbid": 205, "correlation": 34975
    }
  },
  {
    "ph": "f", "id": 34975, "pid": 76337, "tid": -914061504, "ts": 1716454218024592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218072728, "dur": 24,
    "args": {
      "External id": 34979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34979,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34979, "pid": 5, "tid": 7, "ts": 1716454218072728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024606, "dur": 12,
    "args": {
      "External id": 34979,
      "cbid": 211, "correlation": 34979
    }
  },
  {
    "ph": "s", "id": 34979, "pid": 76337, "tid": -914061504, "ts": 1716454218024606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218072752, "dur": 44,
    "args": {
      "External id": 34981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34981, "pid": 5, "tid": 7, "ts": 1716454218072752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024620, "dur": 5,
    "args": {
      "External id": 34981,
      "cbid": 211, "correlation": 34981
    }
  },
  {
    "ph": "s", "id": 34981, "pid": 76337, "tid": -914061504, "ts": 1716454218024620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218072797, "dur": 641,
    "args": {
      "External id": 34983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34983,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 34983, "pid": 5, "tid": 7, "ts": 1716454218072797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024632, "dur": 6,
    "args": {
      "External id": 34983,
      "cbid": 211, "correlation": 34983
    }
  },
  {
    "ph": "s", "id": 34983, "pid": 76337, "tid": -914061504, "ts": 1716454218024632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218073440, "dur": 22,
    "args": {
      "External id": 34985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34985,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34985, "pid": 5, "tid": 7, "ts": 1716454218073440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024641, "dur": 5,
    "args": {
      "External id": 34985,
      "cbid": 211, "correlation": 34985
    }
  },
  {
    "ph": "s", "id": 34985, "pid": 76337, "tid": -914061504, "ts": 1716454218024641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218073463, "dur": 33,
    "args": {
      "External id": 34991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34991, "pid": 5, "tid": 7, "ts": 1716454218073463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024668, "dur": 8,
    "args": {
      "External id": 34991,
      "cbid": 211, "correlation": 34991
    }
  },
  {
    "ph": "s", "id": 34991, "pid": 76337, "tid": -914061504, "ts": 1716454218024668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218073497, "dur": 27,
    "args": {
      "External id": 34999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 34999,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 34999, "pid": 5, "tid": 7, "ts": 1716454218073497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024701, "dur": 8,
    "args": {
      "External id": 34999,
      "cbid": 211, "correlation": 34999
    }
  },
  {
    "ph": "s", "id": 34999, "pid": 76337, "tid": -914061504, "ts": 1716454218024701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218073524, "dur": 20,
    "args": {
      "External id": 35007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35007,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35007, "pid": 5, "tid": 7, "ts": 1716454218073524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024729, "dur": 8,
    "args": {
      "External id": 35007,
      "cbid": 211, "correlation": 35007
    }
  },
  {
    "ph": "s", "id": 35007, "pid": 76337, "tid": -914061504, "ts": 1716454218024729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218073545, "dur": 29,
    "args": {
      "External id": 35027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35027,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 35027, "pid": 5, "tid": 7, "ts": 1716454218073545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024808, "dur": 12,
    "args": {
      "External id": 35027,
      "cbid": 211, "correlation": 35027
    }
  },
  {
    "ph": "s", "id": 35027, "pid": 76337, "tid": -914061504, "ts": 1716454218024808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218073576, "dur": 4,
    "args": {
      "External id": 35039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35039,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 35039, "pid": 5, "tid": 7, "ts": 1716454218073576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024830, "dur": 6,
    "args": {
      "External id": 35039,
      "cbid": 211, "correlation": 35039
    }
  },
  {
    "ph": "s", "id": 35039, "pid": 76337, "tid": -914061504, "ts": 1716454218024830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218073581, "dur": 30,
    "args": {
      "External id": 35042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35042,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35042, "pid": 5, "tid": 7, "ts": 1716454218073581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024848, "dur": 6,
    "args": {
      "External id": 35042,
      "cbid": 211, "correlation": 35042
    }
  },
  {
    "ph": "s", "id": 35042, "pid": 76337, "tid": -914061504, "ts": 1716454218024848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024905, "dur": 0,
    "args": {
      "External id": 35053,
      "cbid": 317, "correlation": 35053
    }
  },
  {
    "ph": "f", "id": 35053, "pid": 76337, "tid": -914061504, "ts": 1716454218024905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024906, "dur": 0,
    "args": {
      "External id": 35054,
      "cbid": 203, "correlation": 35054
    }
  },
  {
    "ph": "f", "id": 35054, "pid": 76337, "tid": -914061504, "ts": 1716454218024906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024907, "dur": 0,
    "args": {
      "External id": 35055,
      "cbid": 205, "correlation": 35055
    }
  },
  {
    "ph": "f", "id": 35055, "pid": 76337, "tid": -914061504, "ts": 1716454218024907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218073612, "dur": 22,
    "args": {
      "External id": 35059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35059,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35059, "pid": 5, "tid": 7, "ts": 1716454218073612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024919, "dur": 12,
    "args": {
      "External id": 35059,
      "cbid": 211, "correlation": 35059
    }
  },
  {
    "ph": "s", "id": 35059, "pid": 76337, "tid": -914061504, "ts": 1716454218024919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218073635, "dur": 115,
    "args": {
      "External id": 35061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35061,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35061, "pid": 5, "tid": 7, "ts": 1716454218073635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024937, "dur": 7,
    "args": {
      "External id": 35061,
      "cbid": 211, "correlation": 35061
    }
  },
  {
    "ph": "s", "id": 35061, "pid": 76337, "tid": -914061504, "ts": 1716454218024937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218073751, "dur": 20,
    "args": {
      "External id": 35063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35063,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35063, "pid": 5, "tid": 7, "ts": 1716454218073751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024948, "dur": 5,
    "args": {
      "External id": 35063,
      "cbid": 211, "correlation": 35063
    }
  },
  {
    "ph": "s", "id": 35063, "pid": 76337, "tid": -914061504, "ts": 1716454218024948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218073773, "dur": 32,
    "args": {
      "External id": 35069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35069, "pid": 5, "tid": 7, "ts": 1716454218073773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218024982, "dur": 9,
    "args": {
      "External id": 35069,
      "cbid": 211, "correlation": 35069
    }
  },
  {
    "ph": "s", "id": 35069, "pid": 76337, "tid": -914061504, "ts": 1716454218024982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218073806, "dur": 204,
    "args": {
      "External id": 35078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35078,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35078, "pid": 5, "tid": 7, "ts": 1716454218073806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025065, "dur": 14,
    "args": {
      "External id": 35078,
      "cbid": 211, "correlation": 35078
    }
  },
  {
    "ph": "s", "id": 35078, "pid": 76337, "tid": -914061504, "ts": 1716454218025065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218074012, "dur": 64,
    "args": {
      "External id": 35100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35100,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35100, "pid": 5, "tid": 7, "ts": 1716454218074012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025123, "dur": 10,
    "args": {
      "External id": 35100,
      "cbid": 211, "correlation": 35100
    }
  },
  {
    "ph": "s", "id": 35100, "pid": 76337, "tid": -914061504, "ts": 1716454218025123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025208, "dur": 1,
    "args": {
      "External id": 35111,
      "cbid": 251, "correlation": 35111
    }
  },
  {
    "ph": "f", "id": 35111, "pid": 76337, "tid": -914061504, "ts": 1716454218025208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218074077, "dur": 154,
    "args": {
      "External id": 35112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35112,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35112, "pid": 5, "tid": 7, "ts": 1716454218074077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025213, "dur": 14,
    "args": {
      "External id": 35112,
      "cbid": 211, "correlation": 35112
    }
  },
  {
    "ph": "s", "id": 35112, "pid": 76337, "tid": -914061504, "ts": 1716454218025213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025283, "dur": 1,
    "args": {
      "External id": 35123,
      "cbid": 251, "correlation": 35123
    }
  },
  {
    "ph": "f", "id": 35123, "pid": 76337, "tid": -914061504, "ts": 1716454218025283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218074232, "dur": 145,
    "args": {
      "External id": 35124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35124,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35124, "pid": 5, "tid": 7, "ts": 1716454218074232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025286, "dur": 11,
    "args": {
      "External id": 35124,
      "cbid": 211, "correlation": 35124
    }
  },
  {
    "ph": "s", "id": 35124, "pid": 76337, "tid": -914061504, "ts": 1716454218025286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025351, "dur": 1,
    "args": {
      "External id": 35135,
      "cbid": 251, "correlation": 35135
    }
  },
  {
    "ph": "f", "id": 35135, "pid": 76337, "tid": -914061504, "ts": 1716454218025351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218074379, "dur": 144,
    "args": {
      "External id": 35136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35136,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35136, "pid": 5, "tid": 7, "ts": 1716454218074379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025355, "dur": 11,
    "args": {
      "External id": 35136,
      "cbid": 211, "correlation": 35136
    }
  },
  {
    "ph": "s", "id": 35136, "pid": 76337, "tid": -914061504, "ts": 1716454218025355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218074524, "dur": 1916,
    "args": {
      "External id": 35157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35157,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 35157, "pid": 5, "tid": 7, "ts": 1716454218074524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025434, "dur": 12,
    "args": {
      "External id": 35157,
      "cbid": 211, "correlation": 35157
    }
  },
  {
    "ph": "s", "id": 35157, "pid": 76337, "tid": -914061504, "ts": 1716454218025434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025529, "dur": 1,
    "args": {
      "External id": 35175,
      "cbid": 251, "correlation": 35175
    }
  },
  {
    "ph": "f", "id": 35175, "pid": 76337, "tid": -914061504, "ts": 1716454218025529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218076441, "dur": 145,
    "args": {
      "External id": 35177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35177,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 35177, "pid": 5, "tid": 7, "ts": 1716454218076441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025535, "dur": 14,
    "args": {
      "External id": 35177,
      "cbid": 211, "correlation": 35177
    }
  },
  {
    "ph": "s", "id": 35177, "pid": 76337, "tid": -914061504, "ts": 1716454218025535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218076587, "dur": 35,
    "args": {
      "External id": 35185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35185,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35185, "pid": 5, "tid": 7, "ts": 1716454218076587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025605, "dur": 12,
    "args": {
      "External id": 35185,
      "cbid": 211, "correlation": 35185
    }
  },
  {
    "ph": "s", "id": 35185, "pid": 76337, "tid": -914061504, "ts": 1716454218025605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218076624, "dur": 51,
    "args": {
      "External id": 35193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35193,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35193, "pid": 5, "tid": 7, "ts": 1716454218076624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025644, "dur": 9,
    "args": {
      "External id": 35193,
      "cbid": 211, "correlation": 35193
    }
  },
  {
    "ph": "s", "id": 35193, "pid": 76337, "tid": -914061504, "ts": 1716454218025644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218076676, "dur": 30,
    "args": {
      "External id": 35204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35204, "pid": 5, "tid": 7, "ts": 1716454218076676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025715, "dur": 12,
    "args": {
      "External id": 35204,
      "cbid": 211, "correlation": 35204
    }
  },
  {
    "ph": "s", "id": 35204, "pid": 76337, "tid": -914061504, "ts": 1716454218025715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218076707, "dur": 35,
    "args": {
      "External id": 35226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35226,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35226, "pid": 5, "tid": 7, "ts": 1716454218076707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025745, "dur": 7,
    "args": {
      "External id": 35226,
      "cbid": 211, "correlation": 35226
    }
  },
  {
    "ph": "s", "id": 35226, "pid": 76337, "tid": -914061504, "ts": 1716454218025745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025829, "dur": 1,
    "args": {
      "External id": 35237,
      "cbid": 251, "correlation": 35237
    }
  },
  {
    "ph": "f", "id": 35237, "pid": 76337, "tid": -914061504, "ts": 1716454218025829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218076743, "dur": 88,
    "args": {
      "External id": 35238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35238,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35238, "pid": 5, "tid": 7, "ts": 1716454218076743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025834, "dur": 13,
    "args": {
      "External id": 35238,
      "cbid": 211, "correlation": 35238
    }
  },
  {
    "ph": "s", "id": 35238, "pid": 76337, "tid": -914061504, "ts": 1716454218025834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025902, "dur": 1,
    "args": {
      "External id": 35249,
      "cbid": 251, "correlation": 35249
    }
  },
  {
    "ph": "f", "id": 35249, "pid": 76337, "tid": -914061504, "ts": 1716454218025902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025906, "dur": 0,
    "args": {
      "External id": 35250,
      "cbid": 251, "correlation": 35250
    }
  },
  {
    "ph": "f", "id": 35250, "pid": 76337, "tid": -914061504, "ts": 1716454218025906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218076833, "dur": 11,
    "args": {
      "External id": 35251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35251,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 35251, "pid": 5, "tid": 7, "ts": 1716454218076833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025907, "dur": 12,
    "args": {
      "External id": 35251,
      "cbid": 211, "correlation": 35251
    }
  },
  {
    "ph": "s", "id": 35251, "pid": 76337, "tid": -914061504, "ts": 1716454218025907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218076845, "dur": 5,
    "args": {
      "External id": 35253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35253,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 35253, "pid": 5, "tid": 7, "ts": 1716454218076845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025921, "dur": 6,
    "args": {
      "External id": 35253,
      "cbid": 211, "correlation": 35253
    }
  },
  {
    "ph": "s", "id": 35253, "pid": 76337, "tid": -914061504, "ts": 1716454218025921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025985, "dur": 1,
    "args": {
      "External id": 35264,
      "cbid": 251, "correlation": 35264
    }
  },
  {
    "ph": "f", "id": 35264, "pid": 76337, "tid": -914061504, "ts": 1716454218025985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025988, "dur": 0,
    "args": {
      "External id": 35265,
      "cbid": 251, "correlation": 35265
    }
  },
  {
    "ph": "f", "id": 35265, "pid": 76337, "tid": -914061504, "ts": 1716454218025988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218076851, "dur": 7,
    "args": {
      "External id": 35266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35266,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 35266, "pid": 5, "tid": 7, "ts": 1716454218076851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218025990, "dur": 12,
    "args": {
      "External id": 35266,
      "cbid": 211, "correlation": 35266
    }
  },
  {
    "ph": "s", "id": 35266, "pid": 76337, "tid": -914061504, "ts": 1716454218025990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218076859, "dur": 3,
    "args": {
      "External id": 35268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35268,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 35268, "pid": 5, "tid": 7, "ts": 1716454218076859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026004, "dur": 6,
    "args": {
      "External id": 35268,
      "cbid": 211, "correlation": 35268
    }
  },
  {
    "ph": "s", "id": 35268, "pid": 76337, "tid": -914061504, "ts": 1716454218026004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218076864, "dur": 91,
    "args": {
      "External id": 35289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35289,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 35289, "pid": 5, "tid": 7, "ts": 1716454218076864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026079, "dur": 12,
    "args": {
      "External id": 35289,
      "cbid": 211, "correlation": 35289
    }
  },
  {
    "ph": "s", "id": 35289, "pid": 76337, "tid": -914061504, "ts": 1716454218026079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026175, "dur": 1,
    "args": {
      "External id": 35307,
      "cbid": 251, "correlation": 35307
    }
  },
  {
    "ph": "f", "id": 35307, "pid": 76337, "tid": -914061504, "ts": 1716454218026175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218076956, "dur": 96,
    "args": {
      "External id": 35309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35309,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35309, "pid": 5, "tid": 7, "ts": 1716454218076956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026182, "dur": 14,
    "args": {
      "External id": 35309,
      "cbid": 211, "correlation": 35309
    }
  },
  {
    "ph": "s", "id": 35309, "pid": 76337, "tid": -914061504, "ts": 1716454218026182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218077053, "dur": 19,
    "args": {
      "External id": 35317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35317,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35317, "pid": 5, "tid": 7, "ts": 1716454218077053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026251, "dur": 12,
    "args": {
      "External id": 35317,
      "cbid": 211, "correlation": 35317
    }
  },
  {
    "ph": "s", "id": 35317, "pid": 76337, "tid": -914061504, "ts": 1716454218026251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218077073, "dur": 36,
    "args": {
      "External id": 35325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35325, "pid": 5, "tid": 7, "ts": 1716454218077073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026293, "dur": 9,
    "args": {
      "External id": 35325,
      "cbid": 211, "correlation": 35325
    }
  },
  {
    "ph": "s", "id": 35325, "pid": 76337, "tid": -914061504, "ts": 1716454218026293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218077110, "dur": 34,
    "args": {
      "External id": 35347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35347,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35347, "pid": 5, "tid": 7, "ts": 1716454218077110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026344, "dur": 10,
    "args": {
      "External id": 35347,
      "cbid": 211, "correlation": 35347
    }
  },
  {
    "ph": "s", "id": 35347, "pid": 76337, "tid": -914061504, "ts": 1716454218026344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026432, "dur": 1,
    "args": {
      "External id": 35363,
      "cbid": 251, "correlation": 35363
    }
  },
  {
    "ph": "f", "id": 35363, "pid": 76337, "tid": -914061504, "ts": 1716454218026432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026437, "dur": 0,
    "args": {
      "External id": 35365,
      "cbid": 251, "correlation": 35365
    }
  },
  {
    "ph": "f", "id": 35365, "pid": 76337, "tid": -914061504, "ts": 1716454218026437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218077146, "dur": 531,
    "args": {
      "External id": 35366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35366,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 35366, "pid": 5, "tid": 7, "ts": 1716454218077146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026440, "dur": 13,
    "args": {
      "External id": 35366,
      "cbid": 211, "correlation": 35366
    }
  },
  {
    "ph": "s", "id": 35366, "pid": 76337, "tid": -914061504, "ts": 1716454218026440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218077678, "dur": 122,
    "args": {
      "External id": 35374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35374, "pid": 5, "tid": 7, "ts": 1716454218077678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026506, "dur": 12,
    "args": {
      "External id": 35374,
      "cbid": 211, "correlation": 35374
    }
  },
  {
    "ph": "s", "id": 35374, "pid": 76337, "tid": -914061504, "ts": 1716454218026506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218077802, "dur": 128,
    "args": {
      "External id": 35382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35382, "pid": 5, "tid": 7, "ts": 1716454218077802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026535, "dur": 8,
    "args": {
      "External id": 35382,
      "cbid": 211, "correlation": 35382
    }
  },
  {
    "ph": "s", "id": 35382, "pid": 76337, "tid": -914061504, "ts": 1716454218026535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026613, "dur": 1,
    "args": {
      "External id": 35398,
      "cbid": 251, "correlation": 35398
    }
  },
  {
    "ph": "f", "id": 35398, "pid": 76337, "tid": -914061504, "ts": 1716454218026613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218077931, "dur": 302,
    "args": {
      "External id": 35400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35400,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35400, "pid": 5, "tid": 7, "ts": 1716454218077931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026618, "dur": 12,
    "args": {
      "External id": 35400,
      "cbid": 211, "correlation": 35400
    }
  },
  {
    "ph": "s", "id": 35400, "pid": 76337, "tid": -914061504, "ts": 1716454218026618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218078235, "dur": 27,
    "args": {
      "External id": 35408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35408,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35408, "pid": 5, "tid": 7, "ts": 1716454218078235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026659, "dur": 11,
    "args": {
      "External id": 35408,
      "cbid": 211, "correlation": 35408
    }
  },
  {
    "ph": "s", "id": 35408, "pid": 76337, "tid": -914061504, "ts": 1716454218026659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218078263, "dur": 81,
    "args": {
      "External id": 35419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35419, "pid": 5, "tid": 7, "ts": 1716454218078263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026727, "dur": 12,
    "args": {
      "External id": 35419,
      "cbid": 211, "correlation": 35419
    }
  },
  {
    "ph": "s", "id": 35419, "pid": 76337, "tid": -914061504, "ts": 1716454218026727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026793, "dur": 0,
    "args": {
      "External id": 35431,
      "cbid": 317, "correlation": 35431
    }
  },
  {
    "ph": "f", "id": 35431, "pid": 76337, "tid": -914061504, "ts": 1716454218026793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026793, "dur": 0,
    "args": {
      "External id": 35432,
      "cbid": 203, "correlation": 35432
    }
  },
  {
    "ph": "f", "id": 35432, "pid": 76337, "tid": -914061504, "ts": 1716454218026793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026794, "dur": 0,
    "args": {
      "External id": 35433,
      "cbid": 205, "correlation": 35433
    }
  },
  {
    "ph": "f", "id": 35433, "pid": 76337, "tid": -914061504, "ts": 1716454218026794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078345, "dur": 23,
    "args": {
      "External id": 35437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35437,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35437, "pid": 5, "tid": 7, "ts": 1716454218078345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026809, "dur": 13,
    "args": {
      "External id": 35437,
      "cbid": 211, "correlation": 35437
    }
  },
  {
    "ph": "s", "id": 35437, "pid": 76337, "tid": -914061504, "ts": 1716454218026809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218078369, "dur": 119,
    "args": {
      "External id": 35439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35439,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35439, "pid": 5, "tid": 7, "ts": 1716454218078369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026828, "dur": 6,
    "args": {
      "External id": 35439,
      "cbid": 211, "correlation": 35439
    }
  },
  {
    "ph": "s", "id": 35439, "pid": 76337, "tid": -914061504, "ts": 1716454218026828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078490, "dur": 22,
    "args": {
      "External id": 35441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35441,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35441, "pid": 5, "tid": 7, "ts": 1716454218078490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026838, "dur": 5,
    "args": {
      "External id": 35441,
      "cbid": 211, "correlation": 35441
    }
  },
  {
    "ph": "s", "id": 35441, "pid": 76337, "tid": -914061504, "ts": 1716454218026838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218078513, "dur": 32,
    "args": {
      "External id": 35447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35447, "pid": 5, "tid": 7, "ts": 1716454218078513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026866, "dur": 10,
    "args": {
      "External id": 35447,
      "cbid": 211, "correlation": 35447
    }
  },
  {
    "ph": "s", "id": 35447, "pid": 76337, "tid": -914061504, "ts": 1716454218026866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218078547, "dur": 27,
    "args": {
      "External id": 35455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35455,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35455, "pid": 5, "tid": 7, "ts": 1716454218078547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026898, "dur": 8,
    "args": {
      "External id": 35455,
      "cbid": 211, "correlation": 35455
    }
  },
  {
    "ph": "s", "id": 35455, "pid": 76337, "tid": -914061504, "ts": 1716454218026898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026970, "dur": 0,
    "args": {
      "External id": 35465,
      "cbid": 317, "correlation": 35465
    }
  },
  {
    "ph": "f", "id": 35465, "pid": 76337, "tid": -914061504, "ts": 1716454218026970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026970, "dur": 0,
    "args": {
      "External id": 35466,
      "cbid": 203, "correlation": 35466
    }
  },
  {
    "ph": "f", "id": 35466, "pid": 76337, "tid": -914061504, "ts": 1716454218026970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026971, "dur": 0,
    "args": {
      "External id": 35467,
      "cbid": 205, "correlation": 35467
    }
  },
  {
    "ph": "f", "id": 35467, "pid": 76337, "tid": -914061504, "ts": 1716454218026971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078575, "dur": 22,
    "args": {
      "External id": 35471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35471,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35471, "pid": 5, "tid": 7, "ts": 1716454218078575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218026993, "dur": 13,
    "args": {
      "External id": 35471,
      "cbid": 211, "correlation": 35471
    }
  },
  {
    "ph": "s", "id": 35471, "pid": 76337, "tid": -914061504, "ts": 1716454218026993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078599, "dur": 44,
    "args": {
      "External id": 35473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35473,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35473, "pid": 5, "tid": 7, "ts": 1716454218078599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027008, "dur": 5,
    "args": {
      "External id": 35473,
      "cbid": 211, "correlation": 35473
    }
  },
  {
    "ph": "s", "id": 35473, "pid": 76337, "tid": -914061504, "ts": 1716454218027008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218078644, "dur": 232,
    "args": {
      "External id": 35475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35475,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 35475, "pid": 5, "tid": 7, "ts": 1716454218078644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027020, "dur": 7,
    "args": {
      "External id": 35475,
      "cbid": 211, "correlation": 35475
    }
  },
  {
    "ph": "s", "id": 35475, "pid": 76337, "tid": -914061504, "ts": 1716454218027020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078877, "dur": 6,
    "args": {
      "External id": 35477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35477,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35477, "pid": 5, "tid": 7, "ts": 1716454218078877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027031, "dur": 5,
    "args": {
      "External id": 35477,
      "cbid": 211, "correlation": 35477
    }
  },
  {
    "ph": "s", "id": 35477, "pid": 76337, "tid": -914061504, "ts": 1716454218027031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218078885, "dur": 9,
    "args": {
      "External id": 35483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35483, "pid": 5, "tid": 7, "ts": 1716454218078885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027057, "dur": 8,
    "args": {
      "External id": 35483,
      "cbid": 211, "correlation": 35483
    }
  },
  {
    "ph": "s", "id": 35483, "pid": 76337, "tid": -914061504, "ts": 1716454218027057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078895, "dur": 12,
    "args": {
      "External id": 35503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35503,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 35503, "pid": 5, "tid": 7, "ts": 1716454218078895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027149, "dur": 12,
    "args": {
      "External id": 35503,
      "cbid": 211, "correlation": 35503
    }
  },
  {
    "ph": "s", "id": 35503, "pid": 76337, "tid": -914061504, "ts": 1716454218027149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218078908, "dur": 5,
    "args": {
      "External id": 35515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35515,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 35515, "pid": 5, "tid": 7, "ts": 1716454218078908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027171, "dur": 6,
    "args": {
      "External id": 35515,
      "cbid": 211, "correlation": 35515
    }
  },
  {
    "ph": "s", "id": 35515, "pid": 76337, "tid": -914061504, "ts": 1716454218027171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218078914, "dur": 12,
    "args": {
      "External id": 35518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35518,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35518, "pid": 5, "tid": 7, "ts": 1716454218078914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027190, "dur": 7,
    "args": {
      "External id": 35518,
      "cbid": 211, "correlation": 35518
    }
  },
  {
    "ph": "s", "id": 35518, "pid": 76337, "tid": -914061504, "ts": 1716454218027190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218078927, "dur": 7,
    "args": {
      "External id": 35527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35527,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35527, "pid": 5, "tid": 7, "ts": 1716454218078927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027228, "dur": 10,
    "args": {
      "External id": 35527,
      "cbid": 211, "correlation": 35527
    }
  },
  {
    "ph": "s", "id": 35527, "pid": 76337, "tid": -914061504, "ts": 1716454218027228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027280, "dur": 0,
    "args": {
      "External id": 35537,
      "cbid": 317, "correlation": 35537
    }
  },
  {
    "ph": "f", "id": 35537, "pid": 76337, "tid": -914061504, "ts": 1716454218027280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027281, "dur": 0,
    "args": {
      "External id": 35538,
      "cbid": 203, "correlation": 35538
    }
  },
  {
    "ph": "f", "id": 35538, "pid": 76337, "tid": -914061504, "ts": 1716454218027281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027282, "dur": 0,
    "args": {
      "External id": 35539,
      "cbid": 205, "correlation": 35539
    }
  },
  {
    "ph": "f", "id": 35539, "pid": 76337, "tid": -914061504, "ts": 1716454218027282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078935, "dur": 5,
    "args": {
      "External id": 35543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35543,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35543, "pid": 5, "tid": 7, "ts": 1716454218078935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027298, "dur": 12,
    "args": {
      "External id": 35543,
      "cbid": 211, "correlation": 35543
    }
  },
  {
    "ph": "s", "id": 35543, "pid": 76337, "tid": -914061504, "ts": 1716454218027298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218078941, "dur": 82,
    "args": {
      "External id": 35545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35545,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35545, "pid": 5, "tid": 7, "ts": 1716454218078941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027312, "dur": 5,
    "args": {
      "External id": 35545,
      "cbid": 211, "correlation": 35545
    }
  },
  {
    "ph": "s", "id": 35545, "pid": 76337, "tid": -914061504, "ts": 1716454218027312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218079026, "dur": 1,
    "args": {
      "External id": 35547,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 35547,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5084745762711864
    }
  },
  {
    "ph": "f", "id": 35547, "pid": 5, "tid": 7, "ts": 1716454218079026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027325, "dur": 9,
    "args": {
      "External id": 35547,
      "cbid": 51, "correlation": 35547
    }
  },
  {
    "ph": "s", "id": 35547, "pid": 76337, "tid": -914061504, "ts": 1716454218027325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218079029, "dur": 533,
    "args": {
      "External id": 35548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35548,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35548, "pid": 5, "tid": 7, "ts": 1716454218079029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027335, "dur": 8,
    "args": {
      "External id": 35548,
      "cbid": 211, "correlation": 35548
    }
  },
  {
    "ph": "s", "id": 35548, "pid": 76337, "tid": -914061504, "ts": 1716454218027335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218079564, "dur": 11,
    "args": {
      "External id": 35550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35550,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35550, "pid": 5, "tid": 7, "ts": 1716454218079564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027348, "dur": 5,
    "args": {
      "External id": 35550,
      "cbid": 211, "correlation": 35550
    }
  },
  {
    "ph": "s", "id": 35550, "pid": 76337, "tid": -914061504, "ts": 1716454218027348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218079577, "dur": 14,
    "args": {
      "External id": 35556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35556, "pid": 5, "tid": 7, "ts": 1716454218079577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027375, "dur": 8,
    "args": {
      "External id": 35556,
      "cbid": 211, "correlation": 35556
    }
  },
  {
    "ph": "s", "id": 35556, "pid": 76337, "tid": -914061504, "ts": 1716454218027375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218079592, "dur": 3,
    "args": {
      "External id": 35564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35564,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 35564, "pid": 5, "tid": 7, "ts": 1716454218079592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027418, "dur": 9,
    "args": {
      "External id": 35564,
      "cbid": 211, "correlation": 35564
    }
  },
  {
    "ph": "s", "id": 35564, "pid": 76337, "tid": -914061504, "ts": 1716454218027418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027483, "dur": 1,
    "args": {
      "External id": 35580,
      "cbid": 251, "correlation": 35580
    }
  },
  {
    "ph": "f", "id": 35580, "pid": 76337, "tid": -914061504, "ts": 1716454218027483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027489, "dur": 0,
    "args": {
      "External id": 35582,
      "cbid": 251, "correlation": 35582
    }
  },
  {
    "ph": "f", "id": 35582, "pid": 76337, "tid": -914061504, "ts": 1716454218027489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218079596, "dur": 13,
    "args": {
      "External id": 35583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35583,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35583, "pid": 5, "tid": 7, "ts": 1716454218079596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027491, "dur": 11,
    "args": {
      "External id": 35583,
      "cbid": 211, "correlation": 35583
    }
  },
  {
    "ph": "s", "id": 35583, "pid": 76337, "tid": -914061504, "ts": 1716454218027491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218079611, "dur": 5,
    "args": {
      "External id": 35585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35585,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35585, "pid": 5, "tid": 7, "ts": 1716454218079611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027504, "dur": 5,
    "args": {
      "External id": 35585,
      "cbid": 211, "correlation": 35585
    }
  },
  {
    "ph": "s", "id": 35585, "pid": 76337, "tid": -914061504, "ts": 1716454218027504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218079617, "dur": 16,
    "args": {
      "External id": 35595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35595,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35595, "pid": 5, "tid": 7, "ts": 1716454218079617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027561, "dur": 12,
    "args": {
      "External id": 35595,
      "cbid": 211, "correlation": 35595
    }
  },
  {
    "ph": "s", "id": 35595, "pid": 76337, "tid": -914061504, "ts": 1716454218027561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218079635, "dur": 17,
    "args": {
      "External id": 35615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35615,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 35615, "pid": 5, "tid": 7, "ts": 1716454218079635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027627, "dur": 11,
    "args": {
      "External id": 35615,
      "cbid": 211, "correlation": 35615
    }
  },
  {
    "ph": "s", "id": 35615, "pid": 76337, "tid": -914061504, "ts": 1716454218027627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218079653, "dur": 5,
    "args": {
      "External id": 35627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35627,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 35627, "pid": 5, "tid": 7, "ts": 1716454218079653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027648, "dur": 6,
    "args": {
      "External id": 35627,
      "cbid": 211, "correlation": 35627
    }
  },
  {
    "ph": "s", "id": 35627, "pid": 76337, "tid": -914061504, "ts": 1716454218027648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218079659, "dur": 16,
    "args": {
      "External id": 35630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35630,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35630, "pid": 5, "tid": 7, "ts": 1716454218079659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027667, "dur": 6,
    "args": {
      "External id": 35630,
      "cbid": 211, "correlation": 35630
    }
  },
  {
    "ph": "s", "id": 35630, "pid": 76337, "tid": -914061504, "ts": 1716454218027667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218079677, "dur": 11,
    "args": {
      "External id": 35639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35639,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35639, "pid": 5, "tid": 7, "ts": 1716454218079677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027708, "dur": 10,
    "args": {
      "External id": 35639,
      "cbid": 211, "correlation": 35639
    }
  },
  {
    "ph": "s", "id": 35639, "pid": 76337, "tid": -914061504, "ts": 1716454218027708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027770, "dur": 0,
    "args": {
      "External id": 35649,
      "cbid": 317, "correlation": 35649
    }
  },
  {
    "ph": "f", "id": 35649, "pid": 76337, "tid": -914061504, "ts": 1716454218027770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027771, "dur": 0,
    "args": {
      "External id": 35650,
      "cbid": 203, "correlation": 35650
    }
  },
  {
    "ph": "f", "id": 35650, "pid": 76337, "tid": -914061504, "ts": 1716454218027771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027772, "dur": 0,
    "args": {
      "External id": 35651,
      "cbid": 205, "correlation": 35651
    }
  },
  {
    "ph": "f", "id": 35651, "pid": 76337, "tid": -914061504, "ts": 1716454218027772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218079689, "dur": 11,
    "args": {
      "External id": 35655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35655, "pid": 5, "tid": 7, "ts": 1716454218079689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027789, "dur": 12,
    "args": {
      "External id": 35655,
      "cbid": 211, "correlation": 35655
    }
  },
  {
    "ph": "s", "id": 35655, "pid": 76337, "tid": -914061504, "ts": 1716454218027789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218079701, "dur": 160,
    "args": {
      "External id": 35657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35657,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35657, "pid": 5, "tid": 7, "ts": 1716454218079701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027804, "dur": 6,
    "args": {
      "External id": 35657,
      "cbid": 211, "correlation": 35657
    }
  },
  {
    "ph": "s", "id": 35657, "pid": 76337, "tid": -914061504, "ts": 1716454218027804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218079863, "dur": 1,
    "args": {
      "External id": 35659,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 35659,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 35659, "pid": 5, "tid": 7, "ts": 1716454218079863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027816, "dur": 7,
    "args": {
      "External id": 35659,
      "cbid": 51, "correlation": 35659
    }
  },
  {
    "ph": "s", "id": 35659, "pid": 76337, "tid": -914061504, "ts": 1716454218027816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218079866, "dur": 654,
    "args": {
      "External id": 35660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35660,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35660, "pid": 5, "tid": 7, "ts": 1716454218079866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027825, "dur": 6,
    "args": {
      "External id": 35660,
      "cbid": 211, "correlation": 35660
    }
  },
  {
    "ph": "s", "id": 35660, "pid": 76337, "tid": -914061504, "ts": 1716454218027825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218080522, "dur": 12,
    "args": {
      "External id": 35662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35662,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35662, "pid": 5, "tid": 7, "ts": 1716454218080522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027835, "dur": 5,
    "args": {
      "External id": 35662,
      "cbid": 211, "correlation": 35662
    }
  },
  {
    "ph": "s", "id": 35662, "pid": 76337, "tid": -914061504, "ts": 1716454218027835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218080535, "dur": 16,
    "args": {
      "External id": 35668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35668, "pid": 5, "tid": 7, "ts": 1716454218080535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027864, "dur": 8,
    "args": {
      "External id": 35668,
      "cbid": 211, "correlation": 35668
    }
  },
  {
    "ph": "s", "id": 35668, "pid": 76337, "tid": -914061504, "ts": 1716454218027864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027922, "dur": 0,
    "args": {
      "External id": 35678,
      "cbid": 317, "correlation": 35678
    }
  },
  {
    "ph": "f", "id": 35678, "pid": 76337, "tid": -914061504, "ts": 1716454218027922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027922, "dur": 0,
    "args": {
      "External id": 35679,
      "cbid": 203, "correlation": 35679
    }
  },
  {
    "ph": "f", "id": 35679, "pid": 76337, "tid": -914061504, "ts": 1716454218027922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027923, "dur": 0,
    "args": {
      "External id": 35680,
      "cbid": 205, "correlation": 35680
    }
  },
  {
    "ph": "f", "id": 35680, "pid": 76337, "tid": -914061504, "ts": 1716454218027923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218080552, "dur": 8,
    "args": {
      "External id": 35684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35684,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35684, "pid": 5, "tid": 7, "ts": 1716454218080552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027937, "dur": 12,
    "args": {
      "External id": 35684,
      "cbid": 211, "correlation": 35684
    }
  },
  {
    "ph": "s", "id": 35684, "pid": 76337, "tid": -914061504, "ts": 1716454218027937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218080562, "dur": 3,
    "args": {
      "External id": 35686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35686,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 35686, "pid": 5, "tid": 7, "ts": 1716454218080562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027954, "dur": 7,
    "args": {
      "External id": 35686,
      "cbid": 211, "correlation": 35686
    }
  },
  {
    "ph": "s", "id": 35686, "pid": 76337, "tid": -914061504, "ts": 1716454218027954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027964, "dur": 0,
    "args": {
      "External id": 35687,
      "cbid": 51, "correlation": 35687
    }
  },
  {
    "ph": "s", "id": 35687, "pid": 76337, "tid": -914061504, "ts": 1716454218027964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218080566, "dur": 56,
    "args": {
      "External id": 35688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35688,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 35688, "pid": 5, "tid": 7, "ts": 1716454218080566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218027965, "dur": 5,
    "args": {
      "External id": 35688,
      "cbid": 211, "correlation": 35688
    }
  },
  {
    "ph": "s", "id": 35688, "pid": 76337, "tid": -914061504, "ts": 1716454218027965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218080623, "dur": 13,
    "args": {
      "External id": 35693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35693,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35693, "pid": 5, "tid": 7, "ts": 1716454218080623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028000, "dur": 10,
    "args": {
      "External id": 35693,
      "cbid": 211, "correlation": 35693
    }
  },
  {
    "ph": "s", "id": 35693, "pid": 76337, "tid": -914061504, "ts": 1716454218028000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218080638, "dur": 11,
    "args": {
      "External id": 35701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35701,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35701, "pid": 5, "tid": 7, "ts": 1716454218080638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028030, "dur": 8,
    "args": {
      "External id": 35701,
      "cbid": 211, "correlation": 35701
    }
  },
  {
    "ph": "s", "id": 35701, "pid": 76337, "tid": -914061504, "ts": 1716454218028030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218080651, "dur": 10,
    "args": {
      "External id": 35709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35709,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35709, "pid": 5, "tid": 7, "ts": 1716454218080651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028059, "dur": 9,
    "args": {
      "External id": 35709,
      "cbid": 211, "correlation": 35709
    }
  },
  {
    "ph": "s", "id": 35709, "pid": 76337, "tid": -914061504, "ts": 1716454218028059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218080662, "dur": 18,
    "args": {
      "External id": 35729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35729,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 35729, "pid": 5, "tid": 7, "ts": 1716454218080662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028140, "dur": 12,
    "args": {
      "External id": 35729,
      "cbid": 211, "correlation": 35729
    }
  },
  {
    "ph": "s", "id": 35729, "pid": 76337, "tid": -914061504, "ts": 1716454218028140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218080681, "dur": 4,
    "args": {
      "External id": 35741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35741,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 35741, "pid": 5, "tid": 7, "ts": 1716454218080681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028162, "dur": 6,
    "args": {
      "External id": 35741,
      "cbid": 211, "correlation": 35741
    }
  },
  {
    "ph": "s", "id": 35741, "pid": 76337, "tid": -914061504, "ts": 1716454218028162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218080687, "dur": 17,
    "args": {
      "External id": 35744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35744, "pid": 5, "tid": 7, "ts": 1716454218080687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028181, "dur": 7,
    "args": {
      "External id": 35744,
      "cbid": 211, "correlation": 35744
    }
  },
  {
    "ph": "s", "id": 35744, "pid": 76337, "tid": -914061504, "ts": 1716454218028181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028238, "dur": 0,
    "args": {
      "External id": 35755,
      "cbid": 317, "correlation": 35755
    }
  },
  {
    "ph": "f", "id": 35755, "pid": 76337, "tid": -914061504, "ts": 1716454218028238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028239, "dur": 0,
    "args": {
      "External id": 35756,
      "cbid": 203, "correlation": 35756
    }
  },
  {
    "ph": "f", "id": 35756, "pid": 76337, "tid": -914061504, "ts": 1716454218028239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028240, "dur": 0,
    "args": {
      "External id": 35757,
      "cbid": 205, "correlation": 35757
    }
  },
  {
    "ph": "f", "id": 35757, "pid": 76337, "tid": -914061504, "ts": 1716454218028240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218080705, "dur": 11,
    "args": {
      "External id": 35761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35761, "pid": 5, "tid": 7, "ts": 1716454218080705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028256, "dur": 12,
    "args": {
      "External id": 35761,
      "cbid": 211, "correlation": 35761
    }
  },
  {
    "ph": "s", "id": 35761, "pid": 76337, "tid": -914061504, "ts": 1716454218028256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218080717, "dur": 3,
    "args": {
      "External id": 35763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 35763, "pid": 5, "tid": 7, "ts": 1716454218080717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028273, "dur": 6,
    "args": {
      "External id": 35763,
      "cbid": 211, "correlation": 35763
    }
  },
  {
    "ph": "s", "id": 35763, "pid": 76337, "tid": -914061504, "ts": 1716454218028273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028282, "dur": 0,
    "args": {
      "External id": 35764,
      "cbid": 51, "correlation": 35764
    }
  },
  {
    "ph": "s", "id": 35764, "pid": 76337, "tid": -914061504, "ts": 1716454218028282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218080722, "dur": 94,
    "args": {
      "External id": 35765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35765,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 35765, "pid": 5, "tid": 7, "ts": 1716454218080722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028283, "dur": 5,
    "args": {
      "External id": 35765,
      "cbid": 211, "correlation": 35765
    }
  },
  {
    "ph": "s", "id": 35765, "pid": 76337, "tid": -914061504, "ts": 1716454218028283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218080817, "dur": 15,
    "args": {
      "External id": 35770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35770, "pid": 5, "tid": 7, "ts": 1716454218080817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028310, "dur": 8,
    "args": {
      "External id": 35770,
      "cbid": 211, "correlation": 35770
    }
  },
  {
    "ph": "s", "id": 35770, "pid": 76337, "tid": -914061504, "ts": 1716454218028310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218080834, "dur": 83,
    "args": {
      "External id": 35779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35779,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35779, "pid": 5, "tid": 7, "ts": 1716454218080834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028391, "dur": 14,
    "args": {
      "External id": 35779,
      "cbid": 211, "correlation": 35779
    }
  },
  {
    "ph": "s", "id": 35779, "pid": 76337, "tid": -914061504, "ts": 1716454218028391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218080918, "dur": 30,
    "args": {
      "External id": 35801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35801,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35801, "pid": 5, "tid": 7, "ts": 1716454218080918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028449, "dur": 10,
    "args": {
      "External id": 35801,
      "cbid": 211, "correlation": 35801
    }
  },
  {
    "ph": "s", "id": 35801, "pid": 76337, "tid": -914061504, "ts": 1716454218028449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028538, "dur": 1,
    "args": {
      "External id": 35812,
      "cbid": 251, "correlation": 35812
    }
  },
  {
    "ph": "f", "id": 35812, "pid": 76337, "tid": -914061504, "ts": 1716454218028538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218080950, "dur": 160,
    "args": {
      "External id": 35813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35813,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35813, "pid": 5, "tid": 7, "ts": 1716454218080950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028544, "dur": 13,
    "args": {
      "External id": 35813,
      "cbid": 211, "correlation": 35813
    }
  },
  {
    "ph": "s", "id": 35813, "pid": 76337, "tid": -914061504, "ts": 1716454218028544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028614, "dur": 1,
    "args": {
      "External id": 35824,
      "cbid": 251, "correlation": 35824
    }
  },
  {
    "ph": "f", "id": 35824, "pid": 76337, "tid": -914061504, "ts": 1716454218028614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218081111, "dur": 155,
    "args": {
      "External id": 35825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35825,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35825, "pid": 5, "tid": 7, "ts": 1716454218081111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028618, "dur": 11,
    "args": {
      "External id": 35825,
      "cbid": 211, "correlation": 35825
    }
  },
  {
    "ph": "s", "id": 35825, "pid": 76337, "tid": -914061504, "ts": 1716454218028618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028681, "dur": 1,
    "args": {
      "External id": 35836,
      "cbid": 251, "correlation": 35836
    }
  },
  {
    "ph": "f", "id": 35836, "pid": 76337, "tid": -914061504, "ts": 1716454218028681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218081268, "dur": 155,
    "args": {
      "External id": 35837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35837,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35837, "pid": 5, "tid": 7, "ts": 1716454218081268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028685, "dur": 12,
    "args": {
      "External id": 35837,
      "cbid": 211, "correlation": 35837
    }
  },
  {
    "ph": "s", "id": 35837, "pid": 76337, "tid": -914061504, "ts": 1716454218028685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218081425, "dur": 330,
    "args": {
      "External id": 35862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35862,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35862, "pid": 5, "tid": 7, "ts": 1716454218081425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028771, "dur": 13,
    "args": {
      "External id": 35862,
      "cbid": 211, "correlation": 35862
    }
  },
  {
    "ph": "s", "id": 35862, "pid": 76337, "tid": -914061504, "ts": 1716454218028771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028873, "dur": 1,
    "args": {
      "External id": 35880,
      "cbid": 251, "correlation": 35880
    }
  },
  {
    "ph": "f", "id": 35880, "pid": 76337, "tid": -914061504, "ts": 1716454218028873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218081756, "dur": 162,
    "args": {
      "External id": 35882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35882,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35882, "pid": 5, "tid": 7, "ts": 1716454218081756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028879, "dur": 13,
    "args": {
      "External id": 35882,
      "cbid": 211, "correlation": 35882
    }
  },
  {
    "ph": "s", "id": 35882, "pid": 76337, "tid": -914061504, "ts": 1716454218028879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218081919, "dur": 19,
    "args": {
      "External id": 35890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35890,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35890, "pid": 5, "tid": 7, "ts": 1716454218081919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028948, "dur": 12,
    "args": {
      "External id": 35890,
      "cbid": 211, "correlation": 35890
    }
  },
  {
    "ph": "s", "id": 35890, "pid": 76337, "tid": -914061504, "ts": 1716454218028948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218081939, "dur": 28,
    "args": {
      "External id": 35898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35898,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35898, "pid": 5, "tid": 7, "ts": 1716454218081939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218028996, "dur": 10,
    "args": {
      "External id": 35898,
      "cbid": 211, "correlation": 35898
    }
  },
  {
    "ph": "s", "id": 35898, "pid": 76337, "tid": -914061504, "ts": 1716454218028996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218081969, "dur": 18,
    "args": {
      "External id": 35909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35909, "pid": 5, "tid": 7, "ts": 1716454218081969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029071, "dur": 13,
    "args": {
      "External id": 35909,
      "cbid": 211, "correlation": 35909
    }
  },
  {
    "ph": "s", "id": 35909, "pid": 76337, "tid": -914061504, "ts": 1716454218029071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218081988, "dur": 16,
    "args": {
      "External id": 35931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35931,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 35931, "pid": 5, "tid": 7, "ts": 1716454218081988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029102, "dur": 8,
    "args": {
      "External id": 35931,
      "cbid": 211, "correlation": 35931
    }
  },
  {
    "ph": "s", "id": 35931, "pid": 76337, "tid": -914061504, "ts": 1716454218029102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029187, "dur": 1,
    "args": {
      "External id": 35942,
      "cbid": 251, "correlation": 35942
    }
  },
  {
    "ph": "f", "id": 35942, "pid": 76337, "tid": -914061504, "ts": 1716454218029187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218082005, "dur": 88,
    "args": {
      "External id": 35943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35943,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 35943, "pid": 5, "tid": 7, "ts": 1716454218082005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029193, "dur": 15,
    "args": {
      "External id": 35943,
      "cbid": 211, "correlation": 35943
    }
  },
  {
    "ph": "s", "id": 35943, "pid": 76337, "tid": -914061504, "ts": 1716454218029193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029264, "dur": 1,
    "args": {
      "External id": 35954,
      "cbid": 251, "correlation": 35954
    }
  },
  {
    "ph": "f", "id": 35954, "pid": 76337, "tid": -914061504, "ts": 1716454218029264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029268, "dur": 0,
    "args": {
      "External id": 35955,
      "cbid": 251, "correlation": 35955
    }
  },
  {
    "ph": "f", "id": 35955, "pid": 76337, "tid": -914061504, "ts": 1716454218029268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218082094, "dur": 12,
    "args": {
      "External id": 35956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35956,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35956, "pid": 5, "tid": 7, "ts": 1716454218082094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029270, "dur": 12,
    "args": {
      "External id": 35956,
      "cbid": 211, "correlation": 35956
    }
  },
  {
    "ph": "s", "id": 35956, "pid": 76337, "tid": -914061504, "ts": 1716454218029270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218082108, "dur": 5,
    "args": {
      "External id": 35958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35958,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35958, "pid": 5, "tid": 7, "ts": 1716454218082108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029284, "dur": 6,
    "args": {
      "External id": 35958,
      "cbid": 211, "correlation": 35958
    }
  },
  {
    "ph": "s", "id": 35958, "pid": 76337, "tid": -914061504, "ts": 1716454218029284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029342, "dur": 1,
    "args": {
      "External id": 35969,
      "cbid": 251, "correlation": 35969
    }
  },
  {
    "ph": "f", "id": 35969, "pid": 76337, "tid": -914061504, "ts": 1716454218029342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029345, "dur": 0,
    "args": {
      "External id": 35970,
      "cbid": 251, "correlation": 35970
    }
  },
  {
    "ph": "f", "id": 35970, "pid": 76337, "tid": -914061504, "ts": 1716454218029345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218082114, "dur": 8,
    "args": {
      "External id": 35971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35971,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35971, "pid": 5, "tid": 7, "ts": 1716454218082114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029347, "dur": 12,
    "args": {
      "External id": 35971,
      "cbid": 211, "correlation": 35971
    }
  },
  {
    "ph": "s", "id": 35971, "pid": 76337, "tid": -914061504, "ts": 1716454218029347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218082123, "dur": 3,
    "args": {
      "External id": 35973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35973,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35973, "pid": 5, "tid": 7, "ts": 1716454218082123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029361, "dur": 6,
    "args": {
      "External id": 35973,
      "cbid": 211, "correlation": 35973
    }
  },
  {
    "ph": "s", "id": 35973, "pid": 76337, "tid": -914061504, "ts": 1716454218029361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218082128, "dur": 55,
    "args": {
      "External id": 35998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 35998,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 35998, "pid": 5, "tid": 7, "ts": 1716454218082128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029437, "dur": 12,
    "args": {
      "External id": 35998,
      "cbid": 211, "correlation": 35998
    }
  },
  {
    "ph": "s", "id": 35998, "pid": 76337, "tid": -914061504, "ts": 1716454218029437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029536, "dur": 2,
    "args": {
      "External id": 36016,
      "cbid": 251, "correlation": 36016
    }
  },
  {
    "ph": "f", "id": 36016, "pid": 76337, "tid": -914061504, "ts": 1716454218029536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218082184, "dur": 89,
    "args": {
      "External id": 36018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36018,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36018, "pid": 5, "tid": 7, "ts": 1716454218082184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029542, "dur": 15,
    "args": {
      "External id": 36018,
      "cbid": 211, "correlation": 36018
    }
  },
  {
    "ph": "s", "id": 36018, "pid": 76337, "tid": -914061504, "ts": 1716454218029542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218082275, "dur": 9,
    "args": {
      "External id": 36026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36026,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36026, "pid": 5, "tid": 7, "ts": 1716454218082275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029613, "dur": 13,
    "args": {
      "External id": 36026,
      "cbid": 211, "correlation": 36026
    }
  },
  {
    "ph": "s", "id": 36026, "pid": 76337, "tid": -914061504, "ts": 1716454218029613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218082286, "dur": 21,
    "args": {
      "External id": 36034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36034, "pid": 5, "tid": 7, "ts": 1716454218082286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029655, "dur": 9,
    "args": {
      "External id": 36034,
      "cbid": 211, "correlation": 36034
    }
  },
  {
    "ph": "s", "id": 36034, "pid": 76337, "tid": -914061504, "ts": 1716454218029655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218082308, "dur": 17,
    "args": {
      "External id": 36056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36056,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36056, "pid": 5, "tid": 7, "ts": 1716454218082308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029707, "dur": 10,
    "args": {
      "External id": 36056,
      "cbid": 211, "correlation": 36056
    }
  },
  {
    "ph": "s", "id": 36056, "pid": 76337, "tid": -914061504, "ts": 1716454218029707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029797, "dur": 2,
    "args": {
      "External id": 36072,
      "cbid": 251, "correlation": 36072
    }
  },
  {
    "ph": "f", "id": 36072, "pid": 76337, "tid": -914061504, "ts": 1716454218029797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029803, "dur": 0,
    "args": {
      "External id": 36074,
      "cbid": 251, "correlation": 36074
    }
  },
  {
    "ph": "f", "id": 36074, "pid": 76337, "tid": -914061504, "ts": 1716454218029803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218082326, "dur": 489,
    "args": {
      "External id": 36075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36075,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36075, "pid": 5, "tid": 7, "ts": 1716454218082326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029805, "dur": 15,
    "args": {
      "External id": 36075,
      "cbid": 211, "correlation": 36075
    }
  },
  {
    "ph": "s", "id": 36075, "pid": 76337, "tid": -914061504, "ts": 1716454218029805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218082817, "dur": 64,
    "args": {
      "External id": 36083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36083,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36083, "pid": 5, "tid": 7, "ts": 1716454218082817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029873, "dur": 13,
    "args": {
      "External id": 36083,
      "cbid": 211, "correlation": 36083
    }
  },
  {
    "ph": "s", "id": 36083, "pid": 76337, "tid": -914061504, "ts": 1716454218029873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218082882, "dur": 69,
    "args": {
      "External id": 36091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36091, "pid": 5, "tid": 7, "ts": 1716454218082882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029904, "dur": 8,
    "args": {
      "External id": 36091,
      "cbid": 211, "correlation": 36091
    }
  },
  {
    "ph": "s", "id": 36091, "pid": 76337, "tid": -914061504, "ts": 1716454218029904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029993, "dur": 1,
    "args": {
      "External id": 36107,
      "cbid": 251, "correlation": 36107
    }
  },
  {
    "ph": "f", "id": 36107, "pid": 76337, "tid": -914061504, "ts": 1716454218029993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218082953, "dur": 1,
    "args": {
      "External id": 36109,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 36109,
      "bytes": 240, "memory bandwidth (GB/s)": 0.12931034482758622
    }
  },
  {
    "ph": "f", "id": 36109, "pid": 5, "tid": 7, "ts": 1716454218082953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218029999, "dur": 11,
    "args": {
      "External id": 36109,
      "cbid": 51, "correlation": 36109
    }
  },
  {
    "ph": "s", "id": 36109, "pid": 76337, "tid": -914061504, "ts": 1716454218029999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218082957, "dur": 267,
    "args": {
      "External id": 36110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36110,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36110, "pid": 5, "tid": 7, "ts": 1716454218082957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030011, "dur": 11,
    "args": {
      "External id": 36110,
      "cbid": 211, "correlation": 36110
    }
  },
  {
    "ph": "s", "id": 36110, "pid": 76337, "tid": -914061504, "ts": 1716454218030011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218083225, "dur": 14,
    "args": {
      "External id": 36118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36118,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36118, "pid": 5, "tid": 7, "ts": 1716454218083225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030055, "dur": 10,
    "args": {
      "External id": 36118,
      "cbid": 211, "correlation": 36118
    }
  },
  {
    "ph": "s", "id": 36118, "pid": 76337, "tid": -914061504, "ts": 1716454218030055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218083240, "dur": 37,
    "args": {
      "External id": 36129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36129, "pid": 5, "tid": 7, "ts": 1716454218083240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030122, "dur": 12,
    "args": {
      "External id": 36129,
      "cbid": 211, "correlation": 36129
    }
  },
  {
    "ph": "s", "id": 36129, "pid": 76337, "tid": -914061504, "ts": 1716454218030122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030187, "dur": 0,
    "args": {
      "External id": 36141,
      "cbid": 317, "correlation": 36141
    }
  },
  {
    "ph": "f", "id": 36141, "pid": 76337, "tid": -914061504, "ts": 1716454218030187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030188, "dur": 0,
    "args": {
      "External id": 36142,
      "cbid": 203, "correlation": 36142
    }
  },
  {
    "ph": "f", "id": 36142, "pid": 76337, "tid": -914061504, "ts": 1716454218030188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030189, "dur": 0,
    "args": {
      "External id": 36143,
      "cbid": 205, "correlation": 36143
    }
  },
  {
    "ph": "f", "id": 36143, "pid": 76337, "tid": -914061504, "ts": 1716454218030189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218083278, "dur": 13,
    "args": {
      "External id": 36147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36147,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36147, "pid": 5, "tid": 7, "ts": 1716454218083278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030205, "dur": 12,
    "args": {
      "External id": 36147,
      "cbid": 211, "correlation": 36147
    }
  },
  {
    "ph": "s", "id": 36147, "pid": 76337, "tid": -914061504, "ts": 1716454218030205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218083292, "dur": 4,
    "args": {
      "External id": 36149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36149,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36149, "pid": 5, "tid": 7, "ts": 1716454218083292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030222, "dur": 6,
    "args": {
      "External id": 36149,
      "cbid": 211, "correlation": 36149
    }
  },
  {
    "ph": "s", "id": 36149, "pid": 76337, "tid": -914061504, "ts": 1716454218030222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030230, "dur": 0,
    "args": {
      "External id": 36150,
      "cbid": 51, "correlation": 36150
    }
  },
  {
    "ph": "s", "id": 36150, "pid": 76337, "tid": -914061504, "ts": 1716454218030230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218083298, "dur": 95,
    "args": {
      "External id": 36151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36151,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 36151, "pid": 5, "tid": 7, "ts": 1716454218083298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030231, "dur": 5,
    "args": {
      "External id": 36151,
      "cbid": 211, "correlation": 36151
    }
  },
  {
    "ph": "s", "id": 36151, "pid": 76337, "tid": -914061504, "ts": 1716454218030231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218083394, "dur": 16,
    "args": {
      "External id": 36156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36156,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36156, "pid": 5, "tid": 7, "ts": 1716454218083394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030258, "dur": 9,
    "args": {
      "External id": 36156,
      "cbid": 211, "correlation": 36156
    }
  },
  {
    "ph": "s", "id": 36156, "pid": 76337, "tid": -914061504, "ts": 1716454218030258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218083411, "dur": 11,
    "args": {
      "External id": 36164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36164,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36164, "pid": 5, "tid": 7, "ts": 1716454218083411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030289, "dur": 9,
    "args": {
      "External id": 36164,
      "cbid": 211, "correlation": 36164
    }
  },
  {
    "ph": "s", "id": 36164, "pid": 76337, "tid": -914061504, "ts": 1716454218030289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218083424, "dur": 18,
    "args": {
      "External id": 36184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36184,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 36184, "pid": 5, "tid": 7, "ts": 1716454218083424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030362, "dur": 12,
    "args": {
      "External id": 36184,
      "cbid": 211, "correlation": 36184
    }
  },
  {
    "ph": "s", "id": 36184, "pid": 76337, "tid": -914061504, "ts": 1716454218030362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218083443, "dur": 4,
    "args": {
      "External id": 36196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36196,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 36196, "pid": 5, "tid": 7, "ts": 1716454218083443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030383, "dur": 6,
    "args": {
      "External id": 36196,
      "cbid": 211, "correlation": 36196
    }
  },
  {
    "ph": "s", "id": 36196, "pid": 76337, "tid": -914061504, "ts": 1716454218030383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218083449, "dur": 18,
    "args": {
      "External id": 36199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36199, "pid": 5, "tid": 7, "ts": 1716454218083449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030401, "dur": 7,
    "args": {
      "External id": 36199,
      "cbid": 211, "correlation": 36199
    }
  },
  {
    "ph": "s", "id": 36199, "pid": 76337, "tid": -914061504, "ts": 1716454218030401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218083468, "dur": 12,
    "args": {
      "External id": 36208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36208,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36208, "pid": 5, "tid": 7, "ts": 1716454218083468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030440, "dur": 10,
    "args": {
      "External id": 36208,
      "cbid": 211, "correlation": 36208
    }
  },
  {
    "ph": "s", "id": 36208, "pid": 76337, "tid": -914061504, "ts": 1716454218030440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030491, "dur": 0,
    "args": {
      "External id": 36218,
      "cbid": 317, "correlation": 36218
    }
  },
  {
    "ph": "f", "id": 36218, "pid": 76337, "tid": -914061504, "ts": 1716454218030491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030492, "dur": 0,
    "args": {
      "External id": 36219,
      "cbid": 203, "correlation": 36219
    }
  },
  {
    "ph": "f", "id": 36219, "pid": 76337, "tid": -914061504, "ts": 1716454218030492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030493, "dur": 0,
    "args": {
      "External id": 36220,
      "cbid": 205, "correlation": 36220
    }
  },
  {
    "ph": "f", "id": 36220, "pid": 76337, "tid": -914061504, "ts": 1716454218030493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218083481, "dur": 11,
    "args": {
      "External id": 36224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36224, "pid": 5, "tid": 7, "ts": 1716454218083481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030506, "dur": 12,
    "args": {
      "External id": 36224,
      "cbid": 211, "correlation": 36224
    }
  },
  {
    "ph": "s", "id": 36224, "pid": 76337, "tid": -914061504, "ts": 1716454218030506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218083493, "dur": 159,
    "args": {
      "External id": 36226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36226,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36226, "pid": 5, "tid": 7, "ts": 1716454218083493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030521, "dur": 5,
    "args": {
      "External id": 36226,
      "cbid": 211, "correlation": 36226
    }
  },
  {
    "ph": "s", "id": 36226, "pid": 76337, "tid": -914061504, "ts": 1716454218030521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218083654, "dur": 1,
    "args": {
      "External id": 36228,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 36228,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 36228, "pid": 5, "tid": 7, "ts": 1716454218083654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030532, "dur": 7,
    "args": {
      "External id": 36228,
      "cbid": 51, "correlation": 36228
    }
  },
  {
    "ph": "s", "id": 36228, "pid": 76337, "tid": -914061504, "ts": 1716454218030532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218083658, "dur": 653,
    "args": {
      "External id": 36229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36229,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36229, "pid": 5, "tid": 7, "ts": 1716454218083658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030540, "dur": 6,
    "args": {
      "External id": 36229,
      "cbid": 211, "correlation": 36229
    }
  },
  {
    "ph": "s", "id": 36229, "pid": 76337, "tid": -914061504, "ts": 1716454218030540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218084312, "dur": 14,
    "args": {
      "External id": 36231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36231, "pid": 5, "tid": 7, "ts": 1716454218084312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030551, "dur": 5,
    "args": {
      "External id": 36231,
      "cbid": 211, "correlation": 36231
    }
  },
  {
    "ph": "s", "id": 36231, "pid": 76337, "tid": -914061504, "ts": 1716454218030551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218084327, "dur": 14,
    "args": {
      "External id": 36237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36237, "pid": 5, "tid": 7, "ts": 1716454218084327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030580, "dur": 8,
    "args": {
      "External id": 36237,
      "cbid": 211, "correlation": 36237
    }
  },
  {
    "ph": "s", "id": 36237, "pid": 76337, "tid": -914061504, "ts": 1716454218030580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218084342, "dur": 3,
    "args": {
      "External id": 36245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36245,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 36245, "pid": 5, "tid": 7, "ts": 1716454218084342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030623, "dur": 9,
    "args": {
      "External id": 36245,
      "cbid": 211, "correlation": 36245
    }
  },
  {
    "ph": "s", "id": 36245, "pid": 76337, "tid": -914061504, "ts": 1716454218030623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030688, "dur": 1,
    "args": {
      "External id": 36261,
      "cbid": 251, "correlation": 36261
    }
  },
  {
    "ph": "f", "id": 36261, "pid": 76337, "tid": -914061504, "ts": 1716454218030688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030692, "dur": 0,
    "args": {
      "External id": 36263,
      "cbid": 251, "correlation": 36263
    }
  },
  {
    "ph": "f", "id": 36263, "pid": 76337, "tid": -914061504, "ts": 1716454218030692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218084347, "dur": 13,
    "args": {
      "External id": 36264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36264,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36264, "pid": 5, "tid": 7, "ts": 1716454218084347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030694, "dur": 11,
    "args": {
      "External id": 36264,
      "cbid": 211, "correlation": 36264
    }
  },
  {
    "ph": "s", "id": 36264, "pid": 76337, "tid": -914061504, "ts": 1716454218030694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218084361, "dur": 5,
    "args": {
      "External id": 36266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36266,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36266, "pid": 5, "tid": 7, "ts": 1716454218084361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030707, "dur": 5,
    "args": {
      "External id": 36266,
      "cbid": 211, "correlation": 36266
    }
  },
  {
    "ph": "s", "id": 36266, "pid": 76337, "tid": -914061504, "ts": 1716454218030707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218084368, "dur": 16,
    "args": {
      "External id": 36276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36276, "pid": 5, "tid": 7, "ts": 1716454218084368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030764, "dur": 13,
    "args": {
      "External id": 36276,
      "cbid": 211, "correlation": 36276
    }
  },
  {
    "ph": "s", "id": 36276, "pid": 76337, "tid": -914061504, "ts": 1716454218030764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218084385, "dur": 17,
    "args": {
      "External id": 36296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36296,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 36296, "pid": 5, "tid": 7, "ts": 1716454218084385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030831, "dur": 11,
    "args": {
      "External id": 36296,
      "cbid": 211, "correlation": 36296
    }
  },
  {
    "ph": "s", "id": 36296, "pid": 76337, "tid": -914061504, "ts": 1716454218030831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218084404, "dur": 4,
    "args": {
      "External id": 36308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36308,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 36308, "pid": 5, "tid": 7, "ts": 1716454218084404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030852, "dur": 6,
    "args": {
      "External id": 36308,
      "cbid": 211, "correlation": 36308
    }
  },
  {
    "ph": "s", "id": 36308, "pid": 76337, "tid": -914061504, "ts": 1716454218030852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218084409, "dur": 16,
    "args": {
      "External id": 36311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36311, "pid": 5, "tid": 7, "ts": 1716454218084409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030870, "dur": 7,
    "args": {
      "External id": 36311,
      "cbid": 211, "correlation": 36311
    }
  },
  {
    "ph": "s", "id": 36311, "pid": 76337, "tid": -914061504, "ts": 1716454218030870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218084426, "dur": 10,
    "args": {
      "External id": 36320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36320,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36320, "pid": 5, "tid": 7, "ts": 1716454218084426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030910, "dur": 9,
    "args": {
      "External id": 36320,
      "cbid": 211, "correlation": 36320
    }
  },
  {
    "ph": "s", "id": 36320, "pid": 76337, "tid": -914061504, "ts": 1716454218030910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030972, "dur": 0,
    "args": {
      "External id": 36330,
      "cbid": 317, "correlation": 36330
    }
  },
  {
    "ph": "f", "id": 36330, "pid": 76337, "tid": -914061504, "ts": 1716454218030972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030972, "dur": 10,
    "args": {
      "External id": 36331,
      "cbid": 203, "correlation": 36331
    }
  },
  {
    "ph": "f", "id": 36331, "pid": 76337, "tid": -914061504, "ts": 1716454218030972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030984, "dur": 0,
    "args": {
      "External id": 36332,
      "cbid": 205, "correlation": 36332
    }
  },
  {
    "ph": "f", "id": 36332, "pid": 76337, "tid": -914061504, "ts": 1716454218030984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218084438, "dur": 11,
    "args": {
      "External id": 36336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36336, "pid": 5, "tid": 7, "ts": 1716454218084438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218030998, "dur": 13,
    "args": {
      "External id": 36336,
      "cbid": 211, "correlation": 36336
    }
  },
  {
    "ph": "s", "id": 36336, "pid": 76337, "tid": -914061504, "ts": 1716454218030998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218084450, "dur": 158,
    "args": {
      "External id": 36338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36338,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36338, "pid": 5, "tid": 7, "ts": 1716454218084450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031013, "dur": 5,
    "args": {
      "External id": 36338,
      "cbid": 211, "correlation": 36338
    }
  },
  {
    "ph": "s", "id": 36338, "pid": 76337, "tid": -914061504, "ts": 1716454218031013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218084611, "dur": 1,
    "args": {
      "External id": 36340,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 36340,
      "bytes": 960, "memory bandwidth (GB/s)": 0.600375234521576
    }
  },
  {
    "ph": "f", "id": 36340, "pid": 5, "tid": 7, "ts": 1716454218084611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031024, "dur": 7,
    "args": {
      "External id": 36340,
      "cbid": 51, "correlation": 36340
    }
  },
  {
    "ph": "s", "id": 36340, "pid": 76337, "tid": -914061504, "ts": 1716454218031024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218084614, "dur": 637,
    "args": {
      "External id": 36341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36341,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36341, "pid": 5, "tid": 7, "ts": 1716454218084614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031032, "dur": 6,
    "args": {
      "External id": 36341,
      "cbid": 211, "correlation": 36341
    }
  },
  {
    "ph": "s", "id": 36341, "pid": 76337, "tid": -914061504, "ts": 1716454218031032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218085252, "dur": 12,
    "args": {
      "External id": 36343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36343, "pid": 5, "tid": 7, "ts": 1716454218085252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031042, "dur": 5,
    "args": {
      "External id": 36343,
      "cbid": 211, "correlation": 36343
    }
  },
  {
    "ph": "s", "id": 36343, "pid": 76337, "tid": -914061504, "ts": 1716454218031042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218085265, "dur": 14,
    "args": {
      "External id": 36349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36349, "pid": 5, "tid": 7, "ts": 1716454218085265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031070, "dur": 8,
    "args": {
      "External id": 36349,
      "cbid": 211, "correlation": 36349
    }
  },
  {
    "ph": "s", "id": 36349, "pid": 76337, "tid": -914061504, "ts": 1716454218031070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218085281, "dur": 11,
    "args": {
      "External id": 36357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36357,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36357, "pid": 5, "tid": 7, "ts": 1716454218085281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031103, "dur": 8,
    "args": {
      "External id": 36357,
      "cbid": 211, "correlation": 36357
    }
  },
  {
    "ph": "s", "id": 36357, "pid": 76337, "tid": -914061504, "ts": 1716454218031103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218085294, "dur": 10,
    "args": {
      "External id": 36365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36365,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36365, "pid": 5, "tid": 7, "ts": 1716454218085294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031131, "dur": 9,
    "args": {
      "External id": 36365,
      "cbid": 211, "correlation": 36365
    }
  },
  {
    "ph": "s", "id": 36365, "pid": 76337, "tid": -914061504, "ts": 1716454218031131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218085305, "dur": 17,
    "args": {
      "External id": 36385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36385,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 36385, "pid": 5, "tid": 7, "ts": 1716454218085305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031209, "dur": 12,
    "args": {
      "External id": 36385,
      "cbid": 211, "correlation": 36385
    }
  },
  {
    "ph": "s", "id": 36385, "pid": 76337, "tid": -914061504, "ts": 1716454218031209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218085323, "dur": 4,
    "args": {
      "External id": 36397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36397,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 36397, "pid": 5, "tid": 7, "ts": 1716454218085323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031231, "dur": 6,
    "args": {
      "External id": 36397,
      "cbid": 211, "correlation": 36397
    }
  },
  {
    "ph": "s", "id": 36397, "pid": 76337, "tid": -914061504, "ts": 1716454218031231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218085329, "dur": 16,
    "args": {
      "External id": 36400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36400,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36400, "pid": 5, "tid": 7, "ts": 1716454218085329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031250, "dur": 6,
    "args": {
      "External id": 36400,
      "cbid": 211, "correlation": 36400
    }
  },
  {
    "ph": "s", "id": 36400, "pid": 76337, "tid": -914061504, "ts": 1716454218031250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031307, "dur": 0,
    "args": {
      "External id": 36411,
      "cbid": 317, "correlation": 36411
    }
  },
  {
    "ph": "f", "id": 36411, "pid": 76337, "tid": -914061504, "ts": 1716454218031307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031307, "dur": 0,
    "args": {
      "External id": 36412,
      "cbid": 203, "correlation": 36412
    }
  },
  {
    "ph": "f", "id": 36412, "pid": 76337, "tid": -914061504, "ts": 1716454218031307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031308, "dur": 0,
    "args": {
      "External id": 36413,
      "cbid": 205, "correlation": 36413
    }
  },
  {
    "ph": "f", "id": 36413, "pid": 76337, "tid": -914061504, "ts": 1716454218031308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218085346, "dur": 11,
    "args": {
      "External id": 36417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36417,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36417, "pid": 5, "tid": 7, "ts": 1716454218085346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031321, "dur": 12,
    "args": {
      "External id": 36417,
      "cbid": 211, "correlation": 36417
    }
  },
  {
    "ph": "s", "id": 36417, "pid": 76337, "tid": -914061504, "ts": 1716454218031321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218085358, "dur": 4,
    "args": {
      "External id": 36419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36419, "pid": 5, "tid": 7, "ts": 1716454218085358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031337, "dur": 5,
    "args": {
      "External id": 36419,
      "cbid": 211, "correlation": 36419
    }
  },
  {
    "ph": "s", "id": 36419, "pid": 76337, "tid": -914061504, "ts": 1716454218031337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031344, "dur": 0,
    "args": {
      "External id": 36420,
      "cbid": 51, "correlation": 36420
    }
  },
  {
    "ph": "s", "id": 36420, "pid": 76337, "tid": -914061504, "ts": 1716454218031344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218085363, "dur": 93,
    "args": {
      "External id": 36421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36421,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 36421, "pid": 5, "tid": 7, "ts": 1716454218085363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031345, "dur": 5,
    "args": {
      "External id": 36421,
      "cbid": 211, "correlation": 36421
    }
  },
  {
    "ph": "s", "id": 36421, "pid": 76337, "tid": -914061504, "ts": 1716454218031345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218085458, "dur": 15,
    "args": {
      "External id": 36426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36426, "pid": 5, "tid": 7, "ts": 1716454218085458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031373, "dur": 8,
    "args": {
      "External id": 36426,
      "cbid": 211, "correlation": 36426
    }
  },
  {
    "ph": "s", "id": 36426, "pid": 76337, "tid": -914061504, "ts": 1716454218031373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218085475, "dur": 82,
    "args": {
      "External id": 36435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36435,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36435, "pid": 5, "tid": 7, "ts": 1716454218085475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031453, "dur": 15,
    "args": {
      "External id": 36435,
      "cbid": 211, "correlation": 36435
    }
  },
  {
    "ph": "s", "id": 36435, "pid": 76337, "tid": -914061504, "ts": 1716454218031453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218085558, "dur": 30,
    "args": {
      "External id": 36457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36457,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36457, "pid": 5, "tid": 7, "ts": 1716454218085558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031511, "dur": 10,
    "args": {
      "External id": 36457,
      "cbid": 211, "correlation": 36457
    }
  },
  {
    "ph": "s", "id": 36457, "pid": 76337, "tid": -914061504, "ts": 1716454218031511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031599, "dur": 1,
    "args": {
      "External id": 36468,
      "cbid": 251, "correlation": 36468
    }
  },
  {
    "ph": "f", "id": 36468, "pid": 76337, "tid": -914061504, "ts": 1716454218031599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218085589, "dur": 159,
    "args": {
      "External id": 36469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36469,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36469, "pid": 5, "tid": 7, "ts": 1716454218085589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031605, "dur": 13,
    "args": {
      "External id": 36469,
      "cbid": 211, "correlation": 36469
    }
  },
  {
    "ph": "s", "id": 36469, "pid": 76337, "tid": -914061504, "ts": 1716454218031605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031674, "dur": 1,
    "args": {
      "External id": 36480,
      "cbid": 251, "correlation": 36480
    }
  },
  {
    "ph": "f", "id": 36480, "pid": 76337, "tid": -914061504, "ts": 1716454218031674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218085749, "dur": 156,
    "args": {
      "External id": 36481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36481,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36481, "pid": 5, "tid": 7, "ts": 1716454218085749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031677, "dur": 12,
    "args": {
      "External id": 36481,
      "cbid": 211, "correlation": 36481
    }
  },
  {
    "ph": "s", "id": 36481, "pid": 76337, "tid": -914061504, "ts": 1716454218031677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031743, "dur": 1,
    "args": {
      "External id": 36492,
      "cbid": 251, "correlation": 36492
    }
  },
  {
    "ph": "f", "id": 36492, "pid": 76337, "tid": -914061504, "ts": 1716454218031743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218085906, "dur": 156,
    "args": {
      "External id": 36493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36493,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36493, "pid": 5, "tid": 7, "ts": 1716454218085906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031746, "dur": 11,
    "args": {
      "External id": 36493,
      "cbid": 211, "correlation": 36493
    }
  },
  {
    "ph": "s", "id": 36493, "pid": 76337, "tid": -914061504, "ts": 1716454218031746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218086063, "dur": 331,
    "args": {
      "External id": 36518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36518,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36518, "pid": 5, "tid": 7, "ts": 1716454218086063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031827, "dur": 13,
    "args": {
      "External id": 36518,
      "cbid": 211, "correlation": 36518
    }
  },
  {
    "ph": "s", "id": 36518, "pid": 76337, "tid": -914061504, "ts": 1716454218031827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031926, "dur": 1,
    "args": {
      "External id": 36536,
      "cbid": 251, "correlation": 36536
    }
  },
  {
    "ph": "f", "id": 36536, "pid": 76337, "tid": -914061504, "ts": 1716454218031926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218086396, "dur": 163,
    "args": {
      "External id": 36538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36538,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36538, "pid": 5, "tid": 7, "ts": 1716454218086396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218031932, "dur": 13,
    "args": {
      "External id": 36538,
      "cbid": 211, "correlation": 36538
    }
  },
  {
    "ph": "s", "id": 36538, "pid": 76337, "tid": -914061504, "ts": 1716454218031932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218086560, "dur": 19,
    "args": {
      "External id": 36546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36546,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36546, "pid": 5, "tid": 7, "ts": 1716454218086560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032010, "dur": 12,
    "args": {
      "External id": 36546,
      "cbid": 211, "correlation": 36546
    }
  },
  {
    "ph": "s", "id": 36546, "pid": 76337, "tid": -914061504, "ts": 1716454218032010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218086581, "dur": 27,
    "args": {
      "External id": 36554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36554,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36554, "pid": 5, "tid": 7, "ts": 1716454218086581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032049, "dur": 9,
    "args": {
      "External id": 36554,
      "cbid": 211, "correlation": 36554
    }
  },
  {
    "ph": "s", "id": 36554, "pid": 76337, "tid": -914061504, "ts": 1716454218032049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218086610, "dur": 18,
    "args": {
      "External id": 36565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36565, "pid": 5, "tid": 7, "ts": 1716454218086610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032119, "dur": 12,
    "args": {
      "External id": 36565,
      "cbid": 211, "correlation": 36565
    }
  },
  {
    "ph": "s", "id": 36565, "pid": 76337, "tid": -914061504, "ts": 1716454218032119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218086629, "dur": 15,
    "args": {
      "External id": 36587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36587,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36587, "pid": 5, "tid": 7, "ts": 1716454218086629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032150, "dur": 7,
    "args": {
      "External id": 36587,
      "cbid": 211, "correlation": 36587
    }
  },
  {
    "ph": "s", "id": 36587, "pid": 76337, "tid": -914061504, "ts": 1716454218032150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032235, "dur": 1,
    "args": {
      "External id": 36598,
      "cbid": 251, "correlation": 36598
    }
  },
  {
    "ph": "f", "id": 36598, "pid": 76337, "tid": -914061504, "ts": 1716454218032235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218086645, "dur": 86,
    "args": {
      "External id": 36599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36599,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36599, "pid": 5, "tid": 7, "ts": 1716454218086645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032240, "dur": 13,
    "args": {
      "External id": 36599,
      "cbid": 211, "correlation": 36599
    }
  },
  {
    "ph": "s", "id": 36599, "pid": 76337, "tid": -914061504, "ts": 1716454218032240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032309, "dur": 1,
    "args": {
      "External id": 36610,
      "cbid": 251, "correlation": 36610
    }
  },
  {
    "ph": "f", "id": 36610, "pid": 76337, "tid": -914061504, "ts": 1716454218032309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032312, "dur": 0,
    "args": {
      "External id": 36611,
      "cbid": 251, "correlation": 36611
    }
  },
  {
    "ph": "f", "id": 36611, "pid": 76337, "tid": -914061504, "ts": 1716454218032312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218086733, "dur": 12,
    "args": {
      "External id": 36612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36612,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36612, "pid": 5, "tid": 7, "ts": 1716454218086733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032314, "dur": 12,
    "args": {
      "External id": 36612,
      "cbid": 211, "correlation": 36612
    }
  },
  {
    "ph": "s", "id": 36612, "pid": 76337, "tid": -914061504, "ts": 1716454218032314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218086746, "dur": 5,
    "args": {
      "External id": 36614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36614,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36614, "pid": 5, "tid": 7, "ts": 1716454218086746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032328, "dur": 6,
    "args": {
      "External id": 36614,
      "cbid": 211, "correlation": 36614
    }
  },
  {
    "ph": "s", "id": 36614, "pid": 76337, "tid": -914061504, "ts": 1716454218032328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032386, "dur": 1,
    "args": {
      "External id": 36625,
      "cbid": 251, "correlation": 36625
    }
  },
  {
    "ph": "f", "id": 36625, "pid": 76337, "tid": -914061504, "ts": 1716454218032386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032389, "dur": 0,
    "args": {
      "External id": 36626,
      "cbid": 251, "correlation": 36626
    }
  },
  {
    "ph": "f", "id": 36626, "pid": 76337, "tid": -914061504, "ts": 1716454218032389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218086752, "dur": 8,
    "args": {
      "External id": 36627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36627,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36627, "pid": 5, "tid": 7, "ts": 1716454218086752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032390, "dur": 11,
    "args": {
      "External id": 36627,
      "cbid": 211, "correlation": 36627
    }
  },
  {
    "ph": "s", "id": 36627, "pid": 76337, "tid": -914061504, "ts": 1716454218032390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218086762, "dur": 3,
    "args": {
      "External id": 36629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36629,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36629, "pid": 5, "tid": 7, "ts": 1716454218086762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032403, "dur": 5,
    "args": {
      "External id": 36629,
      "cbid": 211, "correlation": 36629
    }
  },
  {
    "ph": "s", "id": 36629, "pid": 76337, "tid": -914061504, "ts": 1716454218032403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218086766, "dur": 54,
    "args": {
      "External id": 36654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36654,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36654, "pid": 5, "tid": 7, "ts": 1716454218086766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032480, "dur": 13,
    "args": {
      "External id": 36654,
      "cbid": 211, "correlation": 36654
    }
  },
  {
    "ph": "s", "id": 36654, "pid": 76337, "tid": -914061504, "ts": 1716454218032480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032579, "dur": 1,
    "args": {
      "External id": 36672,
      "cbid": 251, "correlation": 36672
    }
  },
  {
    "ph": "f", "id": 36672, "pid": 76337, "tid": -914061504, "ts": 1716454218032579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218086821, "dur": 89,
    "args": {
      "External id": 36674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36674,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36674, "pid": 5, "tid": 7, "ts": 1716454218086821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032585, "dur": 14,
    "args": {
      "External id": 36674,
      "cbid": 211, "correlation": 36674
    }
  },
  {
    "ph": "s", "id": 36674, "pid": 76337, "tid": -914061504, "ts": 1716454218032585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218086911, "dur": 10,
    "args": {
      "External id": 36682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36682,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36682, "pid": 5, "tid": 7, "ts": 1716454218086911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032656, "dur": 12,
    "args": {
      "External id": 36682,
      "cbid": 211, "correlation": 36682
    }
  },
  {
    "ph": "s", "id": 36682, "pid": 76337, "tid": -914061504, "ts": 1716454218032656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218086922, "dur": 21,
    "args": {
      "External id": 36690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36690,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36690, "pid": 5, "tid": 7, "ts": 1716454218086922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032696, "dur": 10,
    "args": {
      "External id": 36690,
      "cbid": 211, "correlation": 36690
    }
  },
  {
    "ph": "s", "id": 36690, "pid": 76337, "tid": -914061504, "ts": 1716454218032696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218086944, "dur": 17,
    "args": {
      "External id": 36712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36712,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36712, "pid": 5, "tid": 7, "ts": 1716454218086944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032747, "dur": 10,
    "args": {
      "External id": 36712,
      "cbid": 211, "correlation": 36712
    }
  },
  {
    "ph": "s", "id": 36712, "pid": 76337, "tid": -914061504, "ts": 1716454218032747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032835, "dur": 1,
    "args": {
      "External id": 36728,
      "cbid": 251, "correlation": 36728
    }
  },
  {
    "ph": "f", "id": 36728, "pid": 76337, "tid": -914061504, "ts": 1716454218032835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032840, "dur": 0,
    "args": {
      "External id": 36730,
      "cbid": 251, "correlation": 36730
    }
  },
  {
    "ph": "f", "id": 36730, "pid": 76337, "tid": -914061504, "ts": 1716454218032840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218086963, "dur": 488,
    "args": {
      "External id": 36731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36731,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36731, "pid": 5, "tid": 7, "ts": 1716454218086963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032841, "dur": 13,
    "args": {
      "External id": 36731,
      "cbid": 211, "correlation": 36731
    }
  },
  {
    "ph": "s", "id": 36731, "pid": 76337, "tid": -914061504, "ts": 1716454218032841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218087452, "dur": 64,
    "args": {
      "External id": 36739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36739, "pid": 5, "tid": 7, "ts": 1716454218087452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032906, "dur": 13,
    "args": {
      "External id": 36739,
      "cbid": 211, "correlation": 36739
    }
  },
  {
    "ph": "s", "id": 36739, "pid": 76337, "tid": -914061504, "ts": 1716454218032906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218087517, "dur": 67,
    "args": {
      "External id": 36747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36747, "pid": 5, "tid": 7, "ts": 1716454218087517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218032936, "dur": 8,
    "args": {
      "External id": 36747,
      "cbid": 211, "correlation": 36747
    }
  },
  {
    "ph": "s", "id": 36747, "pid": 76337, "tid": -914061504, "ts": 1716454218032936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033023, "dur": 1,
    "args": {
      "External id": 36763,
      "cbid": 251, "correlation": 36763
    }
  },
  {
    "ph": "f", "id": 36763, "pid": 76337, "tid": -914061504, "ts": 1716454218033023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218087586, "dur": 1,
    "args": {
      "External id": 36765,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 36765,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 36765, "pid": 5, "tid": 7, "ts": 1716454218087586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033029, "dur": 10,
    "args": {
      "External id": 36765,
      "cbid": 51, "correlation": 36765
    }
  },
  {
    "ph": "s", "id": 36765, "pid": 76337, "tid": -914061504, "ts": 1716454218033029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218087590, "dur": 267,
    "args": {
      "External id": 36766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36766,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36766, "pid": 5, "tid": 7, "ts": 1716454218087590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033040, "dur": 11,
    "args": {
      "External id": 36766,
      "cbid": 211, "correlation": 36766
    }
  },
  {
    "ph": "s", "id": 36766, "pid": 76337, "tid": -914061504, "ts": 1716454218033040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218087858, "dur": 14,
    "args": {
      "External id": 36774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36774,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36774, "pid": 5, "tid": 7, "ts": 1716454218087858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033083, "dur": 10,
    "args": {
      "External id": 36774,
      "cbid": 211, "correlation": 36774
    }
  },
  {
    "ph": "s", "id": 36774, "pid": 76337, "tid": -914061504, "ts": 1716454218033083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218087873, "dur": 38,
    "args": {
      "External id": 36785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36785, "pid": 5, "tid": 7, "ts": 1716454218087873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033151, "dur": 12,
    "args": {
      "External id": 36785,
      "cbid": 211, "correlation": 36785
    }
  },
  {
    "ph": "s", "id": 36785, "pid": 76337, "tid": -914061504, "ts": 1716454218033151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033215, "dur": 0,
    "args": {
      "External id": 36797,
      "cbid": 317, "correlation": 36797
    }
  },
  {
    "ph": "f", "id": 36797, "pid": 76337, "tid": -914061504, "ts": 1716454218033215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033216, "dur": 0,
    "args": {
      "External id": 36798,
      "cbid": 203, "correlation": 36798
    }
  },
  {
    "ph": "f", "id": 36798, "pid": 76337, "tid": -914061504, "ts": 1716454218033216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033217, "dur": 0,
    "args": {
      "External id": 36799,
      "cbid": 205, "correlation": 36799
    }
  },
  {
    "ph": "f", "id": 36799, "pid": 76337, "tid": -914061504, "ts": 1716454218033217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218087912, "dur": 13,
    "args": {
      "External id": 36803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36803,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36803, "pid": 5, "tid": 7, "ts": 1716454218087912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033231, "dur": 12,
    "args": {
      "External id": 36803,
      "cbid": 211, "correlation": 36803
    }
  },
  {
    "ph": "s", "id": 36803, "pid": 76337, "tid": -914061504, "ts": 1716454218033231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218087927, "dur": 4,
    "args": {
      "External id": 36805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 36805, "pid": 5, "tid": 7, "ts": 1716454218087927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033248, "dur": 6,
    "args": {
      "External id": 36805,
      "cbid": 211, "correlation": 36805
    }
  },
  {
    "ph": "s", "id": 36805, "pid": 76337, "tid": -914061504, "ts": 1716454218033248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033258, "dur": 0,
    "args": {
      "External id": 36806,
      "cbid": 51, "correlation": 36806
    }
  },
  {
    "ph": "s", "id": 36806, "pid": 76337, "tid": -914061504, "ts": 1716454218033258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218087932, "dur": 94,
    "args": {
      "External id": 36807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36807,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 36807, "pid": 5, "tid": 7, "ts": 1716454218087932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033258, "dur": 5,
    "args": {
      "External id": 36807,
      "cbid": 211, "correlation": 36807
    }
  },
  {
    "ph": "s", "id": 36807, "pid": 76337, "tid": -914061504, "ts": 1716454218033258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218088027, "dur": 16,
    "args": {
      "External id": 36812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36812, "pid": 5, "tid": 7, "ts": 1716454218088027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033285, "dur": 9,
    "args": {
      "External id": 36812,
      "cbid": 211, "correlation": 36812
    }
  },
  {
    "ph": "s", "id": 36812, "pid": 76337, "tid": -914061504, "ts": 1716454218033285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218088044, "dur": 11,
    "args": {
      "External id": 36820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36820,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36820, "pid": 5, "tid": 7, "ts": 1716454218088044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033316, "dur": 8,
    "args": {
      "External id": 36820,
      "cbid": 211, "correlation": 36820
    }
  },
  {
    "ph": "s", "id": 36820, "pid": 76337, "tid": -914061504, "ts": 1716454218033316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033386, "dur": 0,
    "args": {
      "External id": 36830,
      "cbid": 317, "correlation": 36830
    }
  },
  {
    "ph": "f", "id": 36830, "pid": 76337, "tid": -914061504, "ts": 1716454218033386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033387, "dur": 0,
    "args": {
      "External id": 36831,
      "cbid": 203, "correlation": 36831
    }
  },
  {
    "ph": "f", "id": 36831, "pid": 76337, "tid": -914061504, "ts": 1716454218033387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033388, "dur": 0,
    "args": {
      "External id": 36832,
      "cbid": 205, "correlation": 36832
    }
  },
  {
    "ph": "f", "id": 36832, "pid": 76337, "tid": -914061504, "ts": 1716454218033388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088057, "dur": 11,
    "args": {
      "External id": 36836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36836,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36836, "pid": 5, "tid": 7, "ts": 1716454218088057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033403, "dur": 12,
    "args": {
      "External id": 36836,
      "cbid": 211, "correlation": 36836
    }
  },
  {
    "ph": "s", "id": 36836, "pid": 76337, "tid": -914061504, "ts": 1716454218033403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088068, "dur": 161,
    "args": {
      "External id": 36838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36838, "pid": 5, "tid": 7, "ts": 1716454218088068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033418, "dur": 5,
    "args": {
      "External id": 36838,
      "cbid": 211, "correlation": 36838
    }
  },
  {
    "ph": "s", "id": 36838, "pid": 76337, "tid": -914061504, "ts": 1716454218033418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218088231, "dur": 1,
    "args": {
      "External id": 36840,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 36840,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 36840, "pid": 5, "tid": 7, "ts": 1716454218088231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033429, "dur": 6,
    "args": {
      "External id": 36840,
      "cbid": 51, "correlation": 36840
    }
  },
  {
    "ph": "s", "id": 36840, "pid": 76337, "tid": -914061504, "ts": 1716454218033429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218088235, "dur": 196,
    "args": {
      "External id": 36841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36841,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 36841, "pid": 5, "tid": 7, "ts": 1716454218088235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033437, "dur": 8,
    "args": {
      "External id": 36841,
      "cbid": 211, "correlation": 36841
    }
  },
  {
    "ph": "s", "id": 36841, "pid": 76337, "tid": -914061504, "ts": 1716454218033437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088432, "dur": 6,
    "args": {
      "External id": 36843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36843,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36843, "pid": 5, "tid": 7, "ts": 1716454218088432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033449, "dur": 5,
    "args": {
      "External id": 36843,
      "cbid": 211, "correlation": 36843
    }
  },
  {
    "ph": "s", "id": 36843, "pid": 76337, "tid": -914061504, "ts": 1716454218033449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218088439, "dur": 6,
    "args": {
      "External id": 36849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 36849, "pid": 5, "tid": 7, "ts": 1716454218088439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033477, "dur": 8,
    "args": {
      "External id": 36849,
      "cbid": 211, "correlation": 36849
    }
  },
  {
    "ph": "s", "id": 36849, "pid": 76337, "tid": -914061504, "ts": 1716454218033477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088446, "dur": 11,
    "args": {
      "External id": 36869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36869,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 36869, "pid": 5, "tid": 7, "ts": 1716454218088446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033568, "dur": 12,
    "args": {
      "External id": 36869,
      "cbid": 211, "correlation": 36869
    }
  },
  {
    "ph": "s", "id": 36869, "pid": 76337, "tid": -914061504, "ts": 1716454218033568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218088458, "dur": 4,
    "args": {
      "External id": 36881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36881,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 36881, "pid": 5, "tid": 7, "ts": 1716454218088458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033590, "dur": 7,
    "args": {
      "External id": 36881,
      "cbid": 211, "correlation": 36881
    }
  },
  {
    "ph": "s", "id": 36881, "pid": 76337, "tid": -914061504, "ts": 1716454218033590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218088464, "dur": 9,
    "args": {
      "External id": 36884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36884,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 36884, "pid": 5, "tid": 7, "ts": 1716454218088464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033609, "dur": 7,
    "args": {
      "External id": 36884,
      "cbid": 211, "correlation": 36884
    }
  },
  {
    "ph": "s", "id": 36884, "pid": 76337, "tid": -914061504, "ts": 1716454218033609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218088474, "dur": 5,
    "args": {
      "External id": 36893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36893,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 36893, "pid": 5, "tid": 7, "ts": 1716454218088474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033649, "dur": 10,
    "args": {
      "External id": 36893,
      "cbid": 211, "correlation": 36893
    }
  },
  {
    "ph": "s", "id": 36893, "pid": 76337, "tid": -914061504, "ts": 1716454218033649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033701, "dur": 0,
    "args": {
      "External id": 36903,
      "cbid": 317, "correlation": 36903
    }
  },
  {
    "ph": "f", "id": 36903, "pid": 76337, "tid": -914061504, "ts": 1716454218033701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033702, "dur": 0,
    "args": {
      "External id": 36904,
      "cbid": 203, "correlation": 36904
    }
  },
  {
    "ph": "f", "id": 36904, "pid": 76337, "tid": -914061504, "ts": 1716454218033702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033702, "dur": 0,
    "args": {
      "External id": 36905,
      "cbid": 205, "correlation": 36905
    }
  },
  {
    "ph": "f", "id": 36905, "pid": 76337, "tid": -914061504, "ts": 1716454218033702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088480, "dur": 5,
    "args": {
      "External id": 36909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36909, "pid": 5, "tid": 7, "ts": 1716454218088480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033718, "dur": 11,
    "args": {
      "External id": 36909,
      "cbid": 211, "correlation": 36909
    }
  },
  {
    "ph": "s", "id": 36909, "pid": 76337, "tid": -914061504, "ts": 1716454218033718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088486, "dur": 160,
    "args": {
      "External id": 36911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36911,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36911, "pid": 5, "tid": 7, "ts": 1716454218088486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033732, "dur": 5,
    "args": {
      "External id": 36911,
      "cbid": 211, "correlation": 36911
    }
  },
  {
    "ph": "s", "id": 36911, "pid": 76337, "tid": -914061504, "ts": 1716454218033732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218088648, "dur": 1,
    "args": {
      "External id": 36913,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 36913,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1271186440677966
    }
  },
  {
    "ph": "f", "id": 36913, "pid": 5, "tid": 7, "ts": 1716454218088648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033742, "dur": 6,
    "args": {
      "External id": 36913,
      "cbid": 51, "correlation": 36913
    }
  },
  {
    "ph": "s", "id": 36913, "pid": 76337, "tid": -914061504, "ts": 1716454218033742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218088652, "dur": 263,
    "args": {
      "External id": 36914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36914,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36914, "pid": 5, "tid": 7, "ts": 1716454218088652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033750, "dur": 6,
    "args": {
      "External id": 36914,
      "cbid": 211, "correlation": 36914
    }
  },
  {
    "ph": "s", "id": 36914, "pid": 76337, "tid": -914061504, "ts": 1716454218033750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088917, "dur": 6,
    "args": {
      "External id": 36916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36916,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 36916, "pid": 5, "tid": 7, "ts": 1716454218088917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033760, "dur": 6,
    "args": {
      "External id": 36916,
      "cbid": 211, "correlation": 36916
    }
  },
  {
    "ph": "s", "id": 36916, "pid": 76337, "tid": -914061504, "ts": 1716454218033760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218088924, "dur": 6,
    "args": {
      "External id": 36922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 36922, "pid": 5, "tid": 7, "ts": 1716454218088924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033789, "dur": 8,
    "args": {
      "External id": 36922,
      "cbid": 211, "correlation": 36922
    }
  },
  {
    "ph": "s", "id": 36922, "pid": 76337, "tid": -914061504, "ts": 1716454218033789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218088932, "dur": 3,
    "args": {
      "External id": 36930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36930,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 36930, "pid": 5, "tid": 7, "ts": 1716454218088932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033832, "dur": 9,
    "args": {
      "External id": 36930,
      "cbid": 211, "correlation": 36930
    }
  },
  {
    "ph": "s", "id": 36930, "pid": 76337, "tid": -914061504, "ts": 1716454218033832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033896, "dur": 1,
    "args": {
      "External id": 36946,
      "cbid": 251, "correlation": 36946
    }
  },
  {
    "ph": "f", "id": 36946, "pid": 76337, "tid": -914061504, "ts": 1716454218033896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033902, "dur": 0,
    "args": {
      "External id": 36948,
      "cbid": 251, "correlation": 36948
    }
  },
  {
    "ph": "f", "id": 36948, "pid": 76337, "tid": -914061504, "ts": 1716454218033902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218088936, "dur": 13,
    "args": {
      "External id": 36949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36949,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36949, "pid": 5, "tid": 7, "ts": 1716454218088936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033903, "dur": 11,
    "args": {
      "External id": 36949,
      "cbid": 211, "correlation": 36949
    }
  },
  {
    "ph": "s", "id": 36949, "pid": 76337, "tid": -914061504, "ts": 1716454218033903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218088950, "dur": 5,
    "args": {
      "External id": 36951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36951,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 36951, "pid": 5, "tid": 7, "ts": 1716454218088950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033916, "dur": 6,
    "args": {
      "External id": 36951,
      "cbid": 211, "correlation": 36951
    }
  },
  {
    "ph": "s", "id": 36951, "pid": 76337, "tid": -914061504, "ts": 1716454218033916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218088956, "dur": 5,
    "args": {
      "External id": 36961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 36961, "pid": 5, "tid": 7, "ts": 1716454218088956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218033981, "dur": 12,
    "args": {
      "External id": 36961,
      "cbid": 211, "correlation": 36961
    }
  },
  {
    "ph": "s", "id": 36961, "pid": 76337, "tid": -914061504, "ts": 1716454218033981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088963, "dur": 10,
    "args": {
      "External id": 36981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36981,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 36981, "pid": 5, "tid": 7, "ts": 1716454218088963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034049, "dur": 11,
    "args": {
      "External id": 36981,
      "cbid": 211, "correlation": 36981
    }
  },
  {
    "ph": "s", "id": 36981, "pid": 76337, "tid": -914061504, "ts": 1716454218034049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218088974, "dur": 4,
    "args": {
      "External id": 36993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36993,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 36993, "pid": 5, "tid": 7, "ts": 1716454218088974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034070, "dur": 6,
    "args": {
      "External id": 36993,
      "cbid": 211, "correlation": 36993
    }
  },
  {
    "ph": "s", "id": 36993, "pid": 76337, "tid": -914061504, "ts": 1716454218034070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218088979, "dur": 7,
    "args": {
      "External id": 36996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 36996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 36996, "pid": 5, "tid": 7, "ts": 1716454218088979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034089, "dur": 6,
    "args": {
      "External id": 36996,
      "cbid": 211, "correlation": 36996
    }
  },
  {
    "ph": "s", "id": 36996, "pid": 76337, "tid": -914061504, "ts": 1716454218034089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218088987, "dur": 4,
    "args": {
      "External id": 37005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37005, "pid": 5, "tid": 7, "ts": 1716454218088987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034129, "dur": 10,
    "args": {
      "External id": 37005,
      "cbid": 211, "correlation": 37005
    }
  },
  {
    "ph": "s", "id": 37005, "pid": 76337, "tid": -914061504, "ts": 1716454218034129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034192, "dur": 0,
    "args": {
      "External id": 37015,
      "cbid": 317, "correlation": 37015
    }
  },
  {
    "ph": "f", "id": 37015, "pid": 76337, "tid": -914061504, "ts": 1716454218034192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034193, "dur": 0,
    "args": {
      "External id": 37016,
      "cbid": 203, "correlation": 37016
    }
  },
  {
    "ph": "f", "id": 37016, "pid": 76337, "tid": -914061504, "ts": 1716454218034193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034194, "dur": 0,
    "args": {
      "External id": 37017,
      "cbid": 205, "correlation": 37017
    }
  },
  {
    "ph": "f", "id": 37017, "pid": 76337, "tid": -914061504, "ts": 1716454218034194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088993, "dur": 5,
    "args": {
      "External id": 37021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37021, "pid": 5, "tid": 7, "ts": 1716454218088993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034208, "dur": 12,
    "args": {
      "External id": 37021,
      "cbid": 211, "correlation": 37021
    }
  },
  {
    "ph": "s", "id": 37021, "pid": 76337, "tid": -914061504, "ts": 1716454218034208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218088999, "dur": 157,
    "args": {
      "External id": 37023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37023, "pid": 5, "tid": 7, "ts": 1716454218088999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034222, "dur": 5,
    "args": {
      "External id": 37023,
      "cbid": 211, "correlation": 37023
    }
  },
  {
    "ph": "s", "id": 37023, "pid": 76337, "tid": -914061504, "ts": 1716454218034222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218089159, "dur": 1,
    "args": {
      "External id": 37025,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37025,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 37025, "pid": 5, "tid": 7, "ts": 1716454218089159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034232, "dur": 7,
    "args": {
      "External id": 37025,
      "cbid": 51, "correlation": 37025
    }
  },
  {
    "ph": "s", "id": 37025, "pid": 76337, "tid": -914061504, "ts": 1716454218034232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218089162, "dur": 253,
    "args": {
      "External id": 37026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37026,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37026, "pid": 5, "tid": 7, "ts": 1716454218089162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034240, "dur": 6,
    "args": {
      "External id": 37026,
      "cbid": 211, "correlation": 37026
    }
  },
  {
    "ph": "s", "id": 37026, "pid": 76337, "tid": -914061504, "ts": 1716454218034240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089417, "dur": 5,
    "args": {
      "External id": 37028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37028, "pid": 5, "tid": 7, "ts": 1716454218089417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034249, "dur": 5,
    "args": {
      "External id": 37028,
      "cbid": 211, "correlation": 37028
    }
  },
  {
    "ph": "s", "id": 37028, "pid": 76337, "tid": -914061504, "ts": 1716454218034249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218089423, "dur": 6,
    "args": {
      "External id": 37034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37034, "pid": 5, "tid": 7, "ts": 1716454218089423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034277, "dur": 8,
    "args": {
      "External id": 37034,
      "cbid": 211, "correlation": 37034
    }
  },
  {
    "ph": "s", "id": 37034, "pid": 76337, "tid": -914061504, "ts": 1716454218034277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218089431, "dur": 5,
    "args": {
      "External id": 37042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37042,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37042, "pid": 5, "tid": 7, "ts": 1716454218089431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034310, "dur": 8,
    "args": {
      "External id": 37042,
      "cbid": 211, "correlation": 37042
    }
  },
  {
    "ph": "s", "id": 37042, "pid": 76337, "tid": -914061504, "ts": 1716454218034310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218089437, "dur": 4,
    "args": {
      "External id": 37050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37050,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37050, "pid": 5, "tid": 7, "ts": 1716454218089437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034339, "dur": 9,
    "args": {
      "External id": 37050,
      "cbid": 211, "correlation": 37050
    }
  },
  {
    "ph": "s", "id": 37050, "pid": 76337, "tid": -914061504, "ts": 1716454218034339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089442, "dur": 9,
    "args": {
      "External id": 37070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37070,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 37070, "pid": 5, "tid": 7, "ts": 1716454218089442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034413, "dur": 12,
    "args": {
      "External id": 37070,
      "cbid": 211, "correlation": 37070
    }
  },
  {
    "ph": "s", "id": 37070, "pid": 76337, "tid": -914061504, "ts": 1716454218034413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218089453, "dur": 3,
    "args": {
      "External id": 37082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37082,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 37082, "pid": 5, "tid": 7, "ts": 1716454218089453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034435, "dur": 7,
    "args": {
      "External id": 37082,
      "cbid": 211, "correlation": 37082
    }
  },
  {
    "ph": "s", "id": 37082, "pid": 76337, "tid": -914061504, "ts": 1716454218034435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218089457, "dur": 6,
    "args": {
      "External id": 37085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37085, "pid": 5, "tid": 7, "ts": 1716454218089457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034453, "dur": 8,
    "args": {
      "External id": 37085,
      "cbid": 211, "correlation": 37085
    }
  },
  {
    "ph": "s", "id": 37085, "pid": 76337, "tid": -914061504, "ts": 1716454218034453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218089465, "dur": 4,
    "args": {
      "External id": 37094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37094,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37094, "pid": 5, "tid": 7, "ts": 1716454218089465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034493, "dur": 10,
    "args": {
      "External id": 37094,
      "cbid": 211, "correlation": 37094
    }
  },
  {
    "ph": "s", "id": 37094, "pid": 76337, "tid": -914061504, "ts": 1716454218034493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034544, "dur": 0,
    "args": {
      "External id": 37104,
      "cbid": 317, "correlation": 37104
    }
  },
  {
    "ph": "f", "id": 37104, "pid": 76337, "tid": -914061504, "ts": 1716454218034544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034545, "dur": 0,
    "args": {
      "External id": 37105,
      "cbid": 203, "correlation": 37105
    }
  },
  {
    "ph": "f", "id": 37105, "pid": 76337, "tid": -914061504, "ts": 1716454218034545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034546, "dur": 0,
    "args": {
      "External id": 37106,
      "cbid": 205, "correlation": 37106
    }
  },
  {
    "ph": "f", "id": 37106, "pid": 76337, "tid": -914061504, "ts": 1716454218034546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089471, "dur": 5,
    "args": {
      "External id": 37110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37110, "pid": 5, "tid": 7, "ts": 1716454218089471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034559, "dur": 12,
    "args": {
      "External id": 37110,
      "cbid": 211, "correlation": 37110
    }
  },
  {
    "ph": "s", "id": 37110, "pid": 76337, "tid": -914061504, "ts": 1716454218034559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089477, "dur": 158,
    "args": {
      "External id": 37112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37112,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37112, "pid": 5, "tid": 7, "ts": 1716454218089477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034573, "dur": 5,
    "args": {
      "External id": 37112,
      "cbid": 211, "correlation": 37112
    }
  },
  {
    "ph": "s", "id": 37112, "pid": 76337, "tid": -914061504, "ts": 1716454218034573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218089638, "dur": 1,
    "args": {
      "External id": 37114,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37114,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 37114, "pid": 5, "tid": 7, "ts": 1716454218089638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034584, "dur": 6,
    "args": {
      "External id": 37114,
      "cbid": 51, "correlation": 37114
    }
  },
  {
    "ph": "s", "id": 37114, "pid": 76337, "tid": -914061504, "ts": 1716454218034584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218089641, "dur": 252,
    "args": {
      "External id": 37115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37115,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37115, "pid": 5, "tid": 7, "ts": 1716454218089641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034591, "dur": 6,
    "args": {
      "External id": 37115,
      "cbid": 211, "correlation": 37115
    }
  },
  {
    "ph": "s", "id": 37115, "pid": 76337, "tid": -914061504, "ts": 1716454218034591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089895, "dur": 5,
    "args": {
      "External id": 37117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37117, "pid": 5, "tid": 7, "ts": 1716454218089895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034601, "dur": 6,
    "args": {
      "External id": 37117,
      "cbid": 211, "correlation": 37117
    }
  },
  {
    "ph": "s", "id": 37117, "pid": 76337, "tid": -914061504, "ts": 1716454218034601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218089901, "dur": 6,
    "args": {
      "External id": 37123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37123, "pid": 5, "tid": 7, "ts": 1716454218089901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034629, "dur": 9,
    "args": {
      "External id": 37123,
      "cbid": 211, "correlation": 37123
    }
  },
  {
    "ph": "s", "id": 37123, "pid": 76337, "tid": -914061504, "ts": 1716454218034629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218089909, "dur": 3,
    "args": {
      "External id": 37131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37131,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 37131, "pid": 5, "tid": 7, "ts": 1716454218089909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034673, "dur": 10,
    "args": {
      "External id": 37131,
      "cbid": 211, "correlation": 37131
    }
  },
  {
    "ph": "s", "id": 37131, "pid": 76337, "tid": -914061504, "ts": 1716454218034673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034736, "dur": 1,
    "args": {
      "External id": 37147,
      "cbid": 251, "correlation": 37147
    }
  },
  {
    "ph": "f", "id": 37147, "pid": 76337, "tid": -914061504, "ts": 1716454218034736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034741, "dur": 0,
    "args": {
      "External id": 37149,
      "cbid": 251, "correlation": 37149
    }
  },
  {
    "ph": "f", "id": 37149, "pid": 76337, "tid": -914061504, "ts": 1716454218034741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218089913, "dur": 10,
    "args": {
      "External id": 37150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37150,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37150, "pid": 5, "tid": 7, "ts": 1716454218089913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034743, "dur": 11,
    "args": {
      "External id": 37150,
      "cbid": 211, "correlation": 37150
    }
  },
  {
    "ph": "s", "id": 37150, "pid": 76337, "tid": -914061504, "ts": 1716454218034743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218089924, "dur": 3,
    "args": {
      "External id": 37152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37152,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37152, "pid": 5, "tid": 7, "ts": 1716454218089924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034756, "dur": 5,
    "args": {
      "External id": 37152,
      "cbid": 211, "correlation": 37152
    }
  },
  {
    "ph": "s", "id": 37152, "pid": 76337, "tid": -914061504, "ts": 1716454218034756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218089929, "dur": 5,
    "args": {
      "External id": 37162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37162, "pid": 5, "tid": 7, "ts": 1716454218089929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034811, "dur": 13,
    "args": {
      "External id": 37162,
      "cbid": 211, "correlation": 37162
    }
  },
  {
    "ph": "s", "id": 37162, "pid": 76337, "tid": -914061504, "ts": 1716454218034811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089935, "dur": 9,
    "args": {
      "External id": 37182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37182,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 37182, "pid": 5, "tid": 7, "ts": 1716454218089935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034879, "dur": 10,
    "args": {
      "External id": 37182,
      "cbid": 211, "correlation": 37182
    }
  },
  {
    "ph": "s", "id": 37182, "pid": 76337, "tid": -914061504, "ts": 1716454218034879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218089946, "dur": 4,
    "args": {
      "External id": 37194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37194,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 37194, "pid": 5, "tid": 7, "ts": 1716454218089946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034899, "dur": 6,
    "args": {
      "External id": 37194,
      "cbid": 211, "correlation": 37194
    }
  },
  {
    "ph": "s", "id": 37194, "pid": 76337, "tid": -914061504, "ts": 1716454218034899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218089951, "dur": 7,
    "args": {
      "External id": 37197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37197, "pid": 5, "tid": 7, "ts": 1716454218089951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034917, "dur": 6,
    "args": {
      "External id": 37197,
      "cbid": 211, "correlation": 37197
    }
  },
  {
    "ph": "s", "id": 37197, "pid": 76337, "tid": -914061504, "ts": 1716454218034917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218089959, "dur": 4,
    "args": {
      "External id": 37206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37206,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37206, "pid": 5, "tid": 7, "ts": 1716454218089959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218034958, "dur": 10,
    "args": {
      "External id": 37206,
      "cbid": 211, "correlation": 37206
    }
  },
  {
    "ph": "s", "id": 37206, "pid": 76337, "tid": -914061504, "ts": 1716454218034958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035029, "dur": 0,
    "args": {
      "External id": 37216,
      "cbid": 317, "correlation": 37216
    }
  },
  {
    "ph": "f", "id": 37216, "pid": 76337, "tid": -914061504, "ts": 1716454218035029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035030, "dur": 0,
    "args": {
      "External id": 37217,
      "cbid": 203, "correlation": 37217
    }
  },
  {
    "ph": "f", "id": 37217, "pid": 76337, "tid": -914061504, "ts": 1716454218035030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035031, "dur": 0,
    "args": {
      "External id": 37218,
      "cbid": 205, "correlation": 37218
    }
  },
  {
    "ph": "f", "id": 37218, "pid": 76337, "tid": -914061504, "ts": 1716454218035031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089964, "dur": 5,
    "args": {
      "External id": 37222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37222,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37222, "pid": 5, "tid": 7, "ts": 1716454218089964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035045, "dur": 12,
    "args": {
      "External id": 37222,
      "cbid": 211, "correlation": 37222
    }
  },
  {
    "ph": "s", "id": 37222, "pid": 76337, "tid": -914061504, "ts": 1716454218035045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218089971, "dur": 159,
    "args": {
      "External id": 37224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37224, "pid": 5, "tid": 7, "ts": 1716454218089971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035060, "dur": 5,
    "args": {
      "External id": 37224,
      "cbid": 211, "correlation": 37224
    }
  },
  {
    "ph": "s", "id": 37224, "pid": 76337, "tid": -914061504, "ts": 1716454218035060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218090132, "dur": 1,
    "args": {
      "External id": 37226,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37226,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 37226, "pid": 5, "tid": 7, "ts": 1716454218090132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035070, "dur": 6,
    "args": {
      "External id": 37226,
      "cbid": 51, "correlation": 37226
    }
  },
  {
    "ph": "s", "id": 37226, "pid": 76337, "tid": -914061504, "ts": 1716454218035070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218090136, "dur": 253,
    "args": {
      "External id": 37227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37227,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37227, "pid": 5, "tid": 7, "ts": 1716454218090136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035077, "dur": 7,
    "args": {
      "External id": 37227,
      "cbid": 211, "correlation": 37227
    }
  },
  {
    "ph": "s", "id": 37227, "pid": 76337, "tid": -914061504, "ts": 1716454218035077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090390, "dur": 6,
    "args": {
      "External id": 37229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37229, "pid": 5, "tid": 7, "ts": 1716454218090390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035089, "dur": 5,
    "args": {
      "External id": 37229,
      "cbid": 211, "correlation": 37229
    }
  },
  {
    "ph": "s", "id": 37229, "pid": 76337, "tid": -914061504, "ts": 1716454218035089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218090397, "dur": 6,
    "args": {
      "External id": 37235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37235,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37235, "pid": 5, "tid": 7, "ts": 1716454218090397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035117, "dur": 8,
    "args": {
      "External id": 37235,
      "cbid": 211, "correlation": 37235
    }
  },
  {
    "ph": "s", "id": 37235, "pid": 76337, "tid": -914061504, "ts": 1716454218035117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218090404, "dur": 5,
    "args": {
      "External id": 37243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37243,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37243, "pid": 5, "tid": 7, "ts": 1716454218090404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035150, "dur": 9,
    "args": {
      "External id": 37243,
      "cbid": 211, "correlation": 37243
    }
  },
  {
    "ph": "s", "id": 37243, "pid": 76337, "tid": -914061504, "ts": 1716454218035150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218090410, "dur": 4,
    "args": {
      "External id": 37251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37251,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37251, "pid": 5, "tid": 7, "ts": 1716454218090410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035179, "dur": 9,
    "args": {
      "External id": 37251,
      "cbid": 211, "correlation": 37251
    }
  },
  {
    "ph": "s", "id": 37251, "pid": 76337, "tid": -914061504, "ts": 1716454218035179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090416, "dur": 9,
    "args": {
      "External id": 37271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37271,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 37271, "pid": 5, "tid": 7, "ts": 1716454218090416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035281, "dur": 13,
    "args": {
      "External id": 37271,
      "cbid": 211, "correlation": 37271
    }
  },
  {
    "ph": "s", "id": 37271, "pid": 76337, "tid": -914061504, "ts": 1716454218035281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218090426, "dur": 4,
    "args": {
      "External id": 37283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37283,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 37283, "pid": 5, "tid": 7, "ts": 1716454218090426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035305, "dur": 6,
    "args": {
      "External id": 37283,
      "cbid": 211, "correlation": 37283
    }
  },
  {
    "ph": "s", "id": 37283, "pid": 76337, "tid": -914061504, "ts": 1716454218035305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218090431, "dur": 6,
    "args": {
      "External id": 37286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37286, "pid": 5, "tid": 7, "ts": 1716454218090431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035323, "dur": 7,
    "args": {
      "External id": 37286,
      "cbid": 211, "correlation": 37286
    }
  },
  {
    "ph": "s", "id": 37286, "pid": 76337, "tid": -914061504, "ts": 1716454218035323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218090439, "dur": 5,
    "args": {
      "External id": 37295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37295,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37295, "pid": 5, "tid": 7, "ts": 1716454218090439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035362, "dur": 9,
    "args": {
      "External id": 37295,
      "cbid": 211, "correlation": 37295
    }
  },
  {
    "ph": "s", "id": 37295, "pid": 76337, "tid": -914061504, "ts": 1716454218035362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035415, "dur": 0,
    "args": {
      "External id": 37305,
      "cbid": 317, "correlation": 37305
    }
  },
  {
    "ph": "f", "id": 37305, "pid": 76337, "tid": -914061504, "ts": 1716454218035415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035416, "dur": 0,
    "args": {
      "External id": 37306,
      "cbid": 203, "correlation": 37306
    }
  },
  {
    "ph": "f", "id": 37306, "pid": 76337, "tid": -914061504, "ts": 1716454218035416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035417, "dur": 0,
    "args": {
      "External id": 37307,
      "cbid": 205, "correlation": 37307
    }
  },
  {
    "ph": "f", "id": 37307, "pid": 76337, "tid": -914061504, "ts": 1716454218035417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090445, "dur": 5,
    "args": {
      "External id": 37311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37311,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37311, "pid": 5, "tid": 7, "ts": 1716454218090445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035430, "dur": 11,
    "args": {
      "External id": 37311,
      "cbid": 211, "correlation": 37311
    }
  },
  {
    "ph": "s", "id": 37311, "pid": 76337, "tid": -914061504, "ts": 1716454218035430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090451, "dur": 158,
    "args": {
      "External id": 37313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37313, "pid": 5, "tid": 7, "ts": 1716454218090451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035444, "dur": 5,
    "args": {
      "External id": 37313,
      "cbid": 211, "correlation": 37313
    }
  },
  {
    "ph": "s", "id": 37313, "pid": 76337, "tid": -914061504, "ts": 1716454218035444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218090611, "dur": 1,
    "args": {
      "External id": 37315,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37315,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 37315, "pid": 5, "tid": 7, "ts": 1716454218090611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035454, "dur": 7,
    "args": {
      "External id": 37315,
      "cbid": 51, "correlation": 37315
    }
  },
  {
    "ph": "s", "id": 37315, "pid": 76337, "tid": -914061504, "ts": 1716454218035454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218090615, "dur": 253,
    "args": {
      "External id": 37316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37316,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37316, "pid": 5, "tid": 7, "ts": 1716454218090615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035462, "dur": 6,
    "args": {
      "External id": 37316,
      "cbid": 211, "correlation": 37316
    }
  },
  {
    "ph": "s", "id": 37316, "pid": 76337, "tid": -914061504, "ts": 1716454218035462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090869, "dur": 5,
    "args": {
      "External id": 37318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37318, "pid": 5, "tid": 7, "ts": 1716454218090869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035472, "dur": 5,
    "args": {
      "External id": 37318,
      "cbid": 211, "correlation": 37318
    }
  },
  {
    "ph": "s", "id": 37318, "pid": 76337, "tid": -914061504, "ts": 1716454218035472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218090876, "dur": 6,
    "args": {
      "External id": 37324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37324, "pid": 5, "tid": 7, "ts": 1716454218090876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035500, "dur": 9,
    "args": {
      "External id": 37324,
      "cbid": 211, "correlation": 37324
    }
  },
  {
    "ph": "s", "id": 37324, "pid": 76337, "tid": -914061504, "ts": 1716454218035500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218090883, "dur": 3,
    "args": {
      "External id": 37332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37332,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 37332, "pid": 5, "tid": 7, "ts": 1716454218090883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035545, "dur": 9,
    "args": {
      "External id": 37332,
      "cbid": 211, "correlation": 37332
    }
  },
  {
    "ph": "s", "id": 37332, "pid": 76337, "tid": -914061504, "ts": 1716454218035545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035606, "dur": 1,
    "args": {
      "External id": 37348,
      "cbid": 251, "correlation": 37348
    }
  },
  {
    "ph": "f", "id": 37348, "pid": 76337, "tid": -914061504, "ts": 1716454218035606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035611, "dur": 0,
    "args": {
      "External id": 37350,
      "cbid": 251, "correlation": 37350
    }
  },
  {
    "ph": "f", "id": 37350, "pid": 76337, "tid": -914061504, "ts": 1716454218035611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218090888, "dur": 11,
    "args": {
      "External id": 37351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37351,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37351, "pid": 5, "tid": 7, "ts": 1716454218090888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035613, "dur": 11,
    "args": {
      "External id": 37351,
      "cbid": 211, "correlation": 37351
    }
  },
  {
    "ph": "s", "id": 37351, "pid": 76337, "tid": -914061504, "ts": 1716454218035613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218090900, "dur": 3,
    "args": {
      "External id": 37353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37353,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37353, "pid": 5, "tid": 7, "ts": 1716454218090900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035627, "dur": 5,
    "args": {
      "External id": 37353,
      "cbid": 211, "correlation": 37353
    }
  },
  {
    "ph": "s", "id": 37353, "pid": 76337, "tid": -914061504, "ts": 1716454218035627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218090905, "dur": 5,
    "args": {
      "External id": 37363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37363,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37363, "pid": 5, "tid": 7, "ts": 1716454218090905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035683, "dur": 12,
    "args": {
      "External id": 37363,
      "cbid": 211, "correlation": 37363
    }
  },
  {
    "ph": "s", "id": 37363, "pid": 76337, "tid": -914061504, "ts": 1716454218035683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090912, "dur": 9,
    "args": {
      "External id": 37383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37383,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 37383, "pid": 5, "tid": 7, "ts": 1716454218090912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035749, "dur": 11,
    "args": {
      "External id": 37383,
      "cbid": 211, "correlation": 37383
    }
  },
  {
    "ph": "s", "id": 37383, "pid": 76337, "tid": -914061504, "ts": 1716454218035749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218090922, "dur": 3,
    "args": {
      "External id": 37395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37395,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 37395, "pid": 5, "tid": 7, "ts": 1716454218090922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035769, "dur": 7,
    "args": {
      "External id": 37395,
      "cbid": 211, "correlation": 37395
    }
  },
  {
    "ph": "s", "id": 37395, "pid": 76337, "tid": -914061504, "ts": 1716454218035769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218090927, "dur": 7,
    "args": {
      "External id": 37398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37398,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37398, "pid": 5, "tid": 7, "ts": 1716454218090927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035788, "dur": 6,
    "args": {
      "External id": 37398,
      "cbid": 211, "correlation": 37398
    }
  },
  {
    "ph": "s", "id": 37398, "pid": 76337, "tid": -914061504, "ts": 1716454218035788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218090935, "dur": 4,
    "args": {
      "External id": 37407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37407,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37407, "pid": 5, "tid": 7, "ts": 1716454218090935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035829, "dur": 10,
    "args": {
      "External id": 37407,
      "cbid": 211, "correlation": 37407
    }
  },
  {
    "ph": "s", "id": 37407, "pid": 76337, "tid": -914061504, "ts": 1716454218035829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035892, "dur": 0,
    "args": {
      "External id": 37417,
      "cbid": 317, "correlation": 37417
    }
  },
  {
    "ph": "f", "id": 37417, "pid": 76337, "tid": -914061504, "ts": 1716454218035892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035893, "dur": 0,
    "args": {
      "External id": 37418,
      "cbid": 203, "correlation": 37418
    }
  },
  {
    "ph": "f", "id": 37418, "pid": 76337, "tid": -914061504, "ts": 1716454218035893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035894, "dur": 0,
    "args": {
      "External id": 37419,
      "cbid": 205, "correlation": 37419
    }
  },
  {
    "ph": "f", "id": 37419, "pid": 76337, "tid": -914061504, "ts": 1716454218035894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090941, "dur": 5,
    "args": {
      "External id": 37423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37423,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37423, "pid": 5, "tid": 7, "ts": 1716454218090941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035907, "dur": 12,
    "args": {
      "External id": 37423,
      "cbid": 211, "correlation": 37423
    }
  },
  {
    "ph": "s", "id": 37423, "pid": 76337, "tid": -914061504, "ts": 1716454218035907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218090947, "dur": 159,
    "args": {
      "External id": 37425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37425,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37425, "pid": 5, "tid": 7, "ts": 1716454218090947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035921, "dur": 5,
    "args": {
      "External id": 37425,
      "cbid": 211, "correlation": 37425
    }
  },
  {
    "ph": "s", "id": 37425, "pid": 76337, "tid": -914061504, "ts": 1716454218035921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218091108, "dur": 1,
    "args": {
      "External id": 37427,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37427,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 37427, "pid": 5, "tid": 7, "ts": 1716454218091108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035932, "dur": 6,
    "args": {
      "External id": 37427,
      "cbid": 51, "correlation": 37427
    }
  },
  {
    "ph": "s", "id": 37427, "pid": 76337, "tid": -914061504, "ts": 1716454218035932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218091112, "dur": 252,
    "args": {
      "External id": 37428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37428,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37428, "pid": 5, "tid": 7, "ts": 1716454218091112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035940, "dur": 6,
    "args": {
      "External id": 37428,
      "cbid": 211, "correlation": 37428
    }
  },
  {
    "ph": "s", "id": 37428, "pid": 76337, "tid": -914061504, "ts": 1716454218035940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091365, "dur": 5,
    "args": {
      "External id": 37430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37430,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37430, "pid": 5, "tid": 7, "ts": 1716454218091365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035950, "dur": 5,
    "args": {
      "External id": 37430,
      "cbid": 211, "correlation": 37430
    }
  },
  {
    "ph": "s", "id": 37430, "pid": 76337, "tid": -914061504, "ts": 1716454218035950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218091372, "dur": 6,
    "args": {
      "External id": 37436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37436, "pid": 5, "tid": 7, "ts": 1716454218091372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218035987, "dur": 9,
    "args": {
      "External id": 37436,
      "cbid": 211, "correlation": 37436
    }
  },
  {
    "ph": "s", "id": 37436, "pid": 76337, "tid": -914061504, "ts": 1716454218035987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218091379, "dur": 5,
    "args": {
      "External id": 37444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37444,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37444, "pid": 5, "tid": 7, "ts": 1716454218091379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036020, "dur": 8,
    "args": {
      "External id": 37444,
      "cbid": 211, "correlation": 37444
    }
  },
  {
    "ph": "s", "id": 37444, "pid": 76337, "tid": -914061504, "ts": 1716454218036020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218091385, "dur": 4,
    "args": {
      "External id": 37452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37452,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37452, "pid": 5, "tid": 7, "ts": 1716454218091385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036050, "dur": 8,
    "args": {
      "External id": 37452,
      "cbid": 211, "correlation": 37452
    }
  },
  {
    "ph": "s", "id": 37452, "pid": 76337, "tid": -914061504, "ts": 1716454218036050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091391, "dur": 9,
    "args": {
      "External id": 37472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37472,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 37472, "pid": 5, "tid": 7, "ts": 1716454218091391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036191, "dur": 13,
    "args": {
      "External id": 37472,
      "cbid": 211, "correlation": 37472
    }
  },
  {
    "ph": "s", "id": 37472, "pid": 76337, "tid": -914061504, "ts": 1716454218036191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218091401, "dur": 3,
    "args": {
      "External id": 37484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37484,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 37484, "pid": 5, "tid": 7, "ts": 1716454218091401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036215, "dur": 6,
    "args": {
      "External id": 37484,
      "cbid": 211, "correlation": 37484
    }
  },
  {
    "ph": "s", "id": 37484, "pid": 76337, "tid": -914061504, "ts": 1716454218036215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218091406, "dur": 6,
    "args": {
      "External id": 37487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37487,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37487, "pid": 5, "tid": 7, "ts": 1716454218091406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036233, "dur": 6,
    "args": {
      "External id": 37487,
      "cbid": 211, "correlation": 37487
    }
  },
  {
    "ph": "s", "id": 37487, "pid": 76337, "tid": -914061504, "ts": 1716454218036233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036291, "dur": 0,
    "args": {
      "External id": 37498,
      "cbid": 317, "correlation": 37498
    }
  },
  {
    "ph": "f", "id": 37498, "pid": 76337, "tid": -914061504, "ts": 1716454218036291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036292, "dur": 0,
    "args": {
      "External id": 37499,
      "cbid": 203, "correlation": 37499
    }
  },
  {
    "ph": "f", "id": 37499, "pid": 76337, "tid": -914061504, "ts": 1716454218036292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036293, "dur": 0,
    "args": {
      "External id": 37500,
      "cbid": 205, "correlation": 37500
    }
  },
  {
    "ph": "f", "id": 37500, "pid": 76337, "tid": -914061504, "ts": 1716454218036293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091414, "dur": 5,
    "args": {
      "External id": 37504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37504,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37504, "pid": 5, "tid": 7, "ts": 1716454218091414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036311, "dur": 12,
    "args": {
      "External id": 37504,
      "cbid": 211, "correlation": 37504
    }
  },
  {
    "ph": "s", "id": 37504, "pid": 76337, "tid": -914061504, "ts": 1716454218036311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218091420, "dur": 36,
    "args": {
      "External id": 37506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37506,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 37506, "pid": 5, "tid": 7, "ts": 1716454218091420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036330, "dur": 9,
    "args": {
      "External id": 37506,
      "cbid": 211, "correlation": 37506
    }
  },
  {
    "ph": "s", "id": 37506, "pid": 76337, "tid": -914061504, "ts": 1716454218036330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091457, "dur": 5,
    "args": {
      "External id": 37508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37508,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37508, "pid": 5, "tid": 7, "ts": 1716454218091457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036343, "dur": 6,
    "args": {
      "External id": 37508,
      "cbid": 211, "correlation": 37508
    }
  },
  {
    "ph": "s", "id": 37508, "pid": 76337, "tid": -914061504, "ts": 1716454218036343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218091464, "dur": 6,
    "args": {
      "External id": 37514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37514, "pid": 5, "tid": 7, "ts": 1716454218091464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036371, "dur": 8,
    "args": {
      "External id": 37514,
      "cbid": 211, "correlation": 37514
    }
  },
  {
    "ph": "s", "id": 37514, "pid": 76337, "tid": -914061504, "ts": 1716454218036371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218091471, "dur": 20,
    "args": {
      "External id": 37523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37523,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37523, "pid": 5, "tid": 7, "ts": 1716454218091471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036454, "dur": 14,
    "args": {
      "External id": 37523,
      "cbid": 211, "correlation": 37523
    }
  },
  {
    "ph": "s", "id": 37523, "pid": 76337, "tid": -914061504, "ts": 1716454218036454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091492, "dur": 11,
    "args": {
      "External id": 37545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37545,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 37545, "pid": 5, "tid": 7, "ts": 1716454218091492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036512, "dur": 10,
    "args": {
      "External id": 37545,
      "cbid": 211, "correlation": 37545
    }
  },
  {
    "ph": "s", "id": 37545, "pid": 76337, "tid": -914061504, "ts": 1716454218036512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036601, "dur": 2,
    "args": {
      "External id": 37556,
      "cbid": 251, "correlation": 37556
    }
  },
  {
    "ph": "f", "id": 37556, "pid": 76337, "tid": -914061504, "ts": 1716454218036601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036606, "dur": 0,
    "args": {
      "External id": 37557,
      "cbid": 251, "correlation": 37557
    }
  },
  {
    "ph": "f", "id": 37557, "pid": 76337, "tid": -914061504, "ts": 1716454218036606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218091504, "dur": 52,
    "args": {
      "External id": 37558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37558,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 37558, "pid": 5, "tid": 7, "ts": 1716454218091504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036609, "dur": 14,
    "args": {
      "External id": 37558,
      "cbid": 211, "correlation": 37558
    }
  },
  {
    "ph": "s", "id": 37558, "pid": 76337, "tid": -914061504, "ts": 1716454218036609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036681, "dur": 1,
    "args": {
      "External id": 37569,
      "cbid": 251, "correlation": 37569
    }
  },
  {
    "ph": "f", "id": 37569, "pid": 76337, "tid": -914061504, "ts": 1716454218036681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036685, "dur": 0,
    "args": {
      "External id": 37570,
      "cbid": 251, "correlation": 37570
    }
  },
  {
    "ph": "f", "id": 37570, "pid": 76337, "tid": -914061504, "ts": 1716454218036685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218091557, "dur": 51,
    "args": {
      "External id": 37571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37571,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 37571, "pid": 5, "tid": 7, "ts": 1716454218091557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036687, "dur": 12,
    "args": {
      "External id": 37571,
      "cbid": 211, "correlation": 37571
    }
  },
  {
    "ph": "s", "id": 37571, "pid": 76337, "tid": -914061504, "ts": 1716454218036687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036751, "dur": 1,
    "args": {
      "External id": 37582,
      "cbid": 251, "correlation": 37582
    }
  },
  {
    "ph": "f", "id": 37582, "pid": 76337, "tid": -914061504, "ts": 1716454218036751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036755, "dur": 0,
    "args": {
      "External id": 37583,
      "cbid": 251, "correlation": 37583
    }
  },
  {
    "ph": "f", "id": 37583, "pid": 76337, "tid": -914061504, "ts": 1716454218036755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218091610, "dur": 52,
    "args": {
      "External id": 37584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37584,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 37584, "pid": 5, "tid": 7, "ts": 1716454218091610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036756, "dur": 11,
    "args": {
      "External id": 37584,
      "cbid": 211, "correlation": 37584
    }
  },
  {
    "ph": "s", "id": 37584, "pid": 76337, "tid": -914061504, "ts": 1716454218036756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218091663, "dur": 55,
    "args": {
      "External id": 37609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37609,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37609, "pid": 5, "tid": 7, "ts": 1716454218091663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036840, "dur": 13,
    "args": {
      "External id": 37609,
      "cbid": 211, "correlation": 37609
    }
  },
  {
    "ph": "s", "id": 37609, "pid": 76337, "tid": -914061504, "ts": 1716454218036840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036939, "dur": 1,
    "args": {
      "External id": 37627,
      "cbid": 251, "correlation": 37627
    }
  },
  {
    "ph": "f", "id": 37627, "pid": 76337, "tid": -914061504, "ts": 1716454218036939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218091720, "dur": 62,
    "args": {
      "External id": 37629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37629,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 37629, "pid": 5, "tid": 7, "ts": 1716454218091720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218036945, "dur": 13,
    "args": {
      "External id": 37629,
      "cbid": 211, "correlation": 37629
    }
  },
  {
    "ph": "s", "id": 37629, "pid": 76337, "tid": -914061504, "ts": 1716454218036945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218091784, "dur": 6,
    "args": {
      "External id": 37637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37637,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37637, "pid": 5, "tid": 7, "ts": 1716454218091784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037024, "dur": 13,
    "args": {
      "External id": 37637,
      "cbid": 211, "correlation": 37637
    }
  },
  {
    "ph": "s", "id": 37637, "pid": 76337, "tid": -914061504, "ts": 1716454218037024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218091791, "dur": 7,
    "args": {
      "External id": 37645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37645,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37645, "pid": 5, "tid": 7, "ts": 1716454218091791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037063, "dur": 9,
    "args": {
      "External id": 37645,
      "cbid": 211, "correlation": 37645
    }
  },
  {
    "ph": "s", "id": 37645, "pid": 76337, "tid": -914061504, "ts": 1716454218037063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218091799, "dur": 7,
    "args": {
      "External id": 37656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37656,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37656, "pid": 5, "tid": 7, "ts": 1716454218091799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037136, "dur": 13,
    "args": {
      "External id": 37656,
      "cbid": 211, "correlation": 37656
    }
  },
  {
    "ph": "s", "id": 37656, "pid": 76337, "tid": -914061504, "ts": 1716454218037136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091808, "dur": 8,
    "args": {
      "External id": 37678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37678,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 37678, "pid": 5, "tid": 7, "ts": 1716454218091808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037169, "dur": 7,
    "args": {
      "External id": 37678,
      "cbid": 211, "correlation": 37678
    }
  },
  {
    "ph": "s", "id": 37678, "pid": 76337, "tid": -914061504, "ts": 1716454218037169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037255, "dur": 2,
    "args": {
      "External id": 37689,
      "cbid": 251, "correlation": 37689
    }
  },
  {
    "ph": "f", "id": 37689, "pid": 76337, "tid": -914061504, "ts": 1716454218037255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218091818, "dur": 1,
    "args": {
      "External id": 37690,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37690,
      "bytes": 480, "memory bandwidth (GB/s)": 0.2542372881355932
    }
  },
  {
    "ph": "f", "id": 37690, "pid": 5, "tid": 7, "ts": 1716454218091818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037261, "dur": 10,
    "args": {
      "External id": 37690,
      "cbid": 51, "correlation": 37690
    }
  },
  {
    "ph": "s", "id": 37690, "pid": 76337, "tid": -914061504, "ts": 1716454218037261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218091822, "dur": 35,
    "args": {
      "External id": 37691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37691,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 37691, "pid": 5, "tid": 7, "ts": 1716454218091822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037273, "dur": 12,
    "args": {
      "External id": 37691,
      "cbid": 211, "correlation": 37691
    }
  },
  {
    "ph": "s", "id": 37691, "pid": 76337, "tid": -914061504, "ts": 1716454218037273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037344, "dur": 1,
    "args": {
      "External id": 37702,
      "cbid": 251, "correlation": 37702
    }
  },
  {
    "ph": "f", "id": 37702, "pid": 76337, "tid": -914061504, "ts": 1716454218037344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037347, "dur": 0,
    "args": {
      "External id": 37703,
      "cbid": 251, "correlation": 37703
    }
  },
  {
    "ph": "f", "id": 37703, "pid": 76337, "tid": -914061504, "ts": 1716454218037347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218091859, "dur": 11,
    "args": {
      "External id": 37704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37704,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37704, "pid": 5, "tid": 7, "ts": 1716454218091859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037349, "dur": 12,
    "args": {
      "External id": 37704,
      "cbid": 211, "correlation": 37704
    }
  },
  {
    "ph": "s", "id": 37704, "pid": 76337, "tid": -914061504, "ts": 1716454218037349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218091871, "dur": 5,
    "args": {
      "External id": 37706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37706,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37706, "pid": 5, "tid": 7, "ts": 1716454218091871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037362, "dur": 6,
    "args": {
      "External id": 37706,
      "cbid": 211, "correlation": 37706
    }
  },
  {
    "ph": "s", "id": 37706, "pid": 76337, "tid": -914061504, "ts": 1716454218037362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037419, "dur": 1,
    "args": {
      "External id": 37717,
      "cbid": 251, "correlation": 37717
    }
  },
  {
    "ph": "f", "id": 37717, "pid": 76337, "tid": -914061504, "ts": 1716454218037419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037422, "dur": 0,
    "args": {
      "External id": 37718,
      "cbid": 251, "correlation": 37718
    }
  },
  {
    "ph": "f", "id": 37718, "pid": 76337, "tid": -914061504, "ts": 1716454218037422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218091877, "dur": 8,
    "args": {
      "External id": 37719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37719,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37719, "pid": 5, "tid": 7, "ts": 1716454218091877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037424, "dur": 12,
    "args": {
      "External id": 37719,
      "cbid": 211, "correlation": 37719
    }
  },
  {
    "ph": "s", "id": 37719, "pid": 76337, "tid": -914061504, "ts": 1716454218037424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218091887, "dur": 3,
    "args": {
      "External id": 37721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37721,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37721, "pid": 5, "tid": 7, "ts": 1716454218091887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037437, "dur": 5,
    "args": {
      "External id": 37721,
      "cbid": 211, "correlation": 37721
    }
  },
  {
    "ph": "s", "id": 37721, "pid": 76337, "tid": -914061504, "ts": 1716454218037437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218091891, "dur": 20,
    "args": {
      "External id": 37746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37746,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 37746, "pid": 5, "tid": 7, "ts": 1716454218091891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037514, "dur": 12,
    "args": {
      "External id": 37746,
      "cbid": 211, "correlation": 37746
    }
  },
  {
    "ph": "s", "id": 37746, "pid": 76337, "tid": -914061504, "ts": 1716454218037514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037614, "dur": 2,
    "args": {
      "External id": 37764,
      "cbid": 251, "correlation": 37764
    }
  },
  {
    "ph": "f", "id": 37764, "pid": 76337, "tid": -914061504, "ts": 1716454218037614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218091913, "dur": 1,
    "args": {
      "External id": 37766,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37766,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 37766, "pid": 5, "tid": 7, "ts": 1716454218091913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037620, "dur": 11,
    "args": {
      "External id": 37766,
      "cbid": 51, "correlation": 37766
    }
  },
  {
    "ph": "s", "id": 37766, "pid": 76337, "tid": -914061504, "ts": 1716454218037620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218091917, "dur": 35,
    "args": {
      "External id": 37767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37767,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 37767, "pid": 5, "tid": 7, "ts": 1716454218091917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037632, "dur": 13,
    "args": {
      "External id": 37767,
      "cbid": 211, "correlation": 37767
    }
  },
  {
    "ph": "s", "id": 37767, "pid": 76337, "tid": -914061504, "ts": 1716454218037632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218091953, "dur": 4,
    "args": {
      "External id": 37775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37775,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37775, "pid": 5, "tid": 7, "ts": 1716454218091953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037703, "dur": 12,
    "args": {
      "External id": 37775,
      "cbid": 211, "correlation": 37775
    }
  },
  {
    "ph": "s", "id": 37775, "pid": 76337, "tid": -914061504, "ts": 1716454218037703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218091959, "dur": 8,
    "args": {
      "External id": 37783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37783, "pid": 5, "tid": 7, "ts": 1716454218091959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037744, "dur": 10,
    "args": {
      "External id": 37783,
      "cbid": 211, "correlation": 37783
    }
  },
  {
    "ph": "s", "id": 37783, "pid": 76337, "tid": -914061504, "ts": 1716454218037744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218091968, "dur": 8,
    "args": {
      "External id": 37805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37805,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 37805, "pid": 5, "tid": 7, "ts": 1716454218091968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037795, "dur": 11,
    "args": {
      "External id": 37805,
      "cbid": 211, "correlation": 37805
    }
  },
  {
    "ph": "s", "id": 37805, "pid": 76337, "tid": -914061504, "ts": 1716454218037795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037886, "dur": 1,
    "args": {
      "External id": 37821,
      "cbid": 251, "correlation": 37821
    }
  },
  {
    "ph": "f", "id": 37821, "pid": 76337, "tid": -914061504, "ts": 1716454218037886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037891, "dur": 0,
    "args": {
      "External id": 37823,
      "cbid": 251, "correlation": 37823
    }
  },
  {
    "ph": "f", "id": 37823, "pid": 76337, "tid": -914061504, "ts": 1716454218037891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218091978, "dur": 188,
    "args": {
      "External id": 37824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37824,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37824, "pid": 5, "tid": 7, "ts": 1716454218091978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037893, "dur": 13,
    "args": {
      "External id": 37824,
      "cbid": 211, "correlation": 37824
    }
  },
  {
    "ph": "s", "id": 37824, "pid": 76337, "tid": -914061504, "ts": 1716454218037893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092167, "dur": 20,
    "args": {
      "External id": 37832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37832, "pid": 5, "tid": 7, "ts": 1716454218092167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037960, "dur": 21,
    "args": {
      "External id": 37832,
      "cbid": 211, "correlation": 37832
    }
  },
  {
    "ph": "s", "id": 37832, "pid": 76337, "tid": -914061504, "ts": 1716454218037960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092189, "dur": 22,
    "args": {
      "External id": 37840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37840,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37840, "pid": 5, "tid": 7, "ts": 1716454218092189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218037999, "dur": 9,
    "args": {
      "External id": 37840,
      "cbid": 211, "correlation": 37840
    }
  },
  {
    "ph": "s", "id": 37840, "pid": 76337, "tid": -914061504, "ts": 1716454218037999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038083, "dur": 1,
    "args": {
      "External id": 37856,
      "cbid": 251, "correlation": 37856
    }
  },
  {
    "ph": "f", "id": 37856, "pid": 76337, "tid": -914061504, "ts": 1716454218038083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218092213, "dur": 1,
    "args": {
      "External id": 37858,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37858,
      "bytes": 120, "memory bandwidth (GB/s)": 0.0797872340425532
    }
  },
  {
    "ph": "f", "id": 37858, "pid": 5, "tid": 7, "ts": 1716454218092213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038088, "dur": 9,
    "args": {
      "External id": 37858,
      "cbid": 51, "correlation": 37858
    }
  },
  {
    "ph": "s", "id": 37858, "pid": 76337, "tid": -914061504, "ts": 1716454218038088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218092216, "dur": 108,
    "args": {
      "External id": 37859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37859,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 37859, "pid": 5, "tid": 7, "ts": 1716454218092216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038098, "dur": 12,
    "args": {
      "External id": 37859,
      "cbid": 211, "correlation": 37859
    }
  },
  {
    "ph": "s", "id": 37859, "pid": 76337, "tid": -914061504, "ts": 1716454218038098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218092325, "dur": 5,
    "args": {
      "External id": 37867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37867,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37867, "pid": 5, "tid": 7, "ts": 1716454218092325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038141, "dur": 10,
    "args": {
      "External id": 37867,
      "cbid": 211, "correlation": 37867
    }
  },
  {
    "ph": "s", "id": 37867, "pid": 76337, "tid": -914061504, "ts": 1716454218038141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092331, "dur": 10,
    "args": {
      "External id": 37878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37878, "pid": 5, "tid": 7, "ts": 1716454218092331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038209, "dur": 12,
    "args": {
      "External id": 37878,
      "cbid": 211, "correlation": 37878
    }
  },
  {
    "ph": "s", "id": 37878, "pid": 76337, "tid": -914061504, "ts": 1716454218038209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038274, "dur": 0,
    "args": {
      "External id": 37890,
      "cbid": 317, "correlation": 37890
    }
  },
  {
    "ph": "f", "id": 37890, "pid": 76337, "tid": -914061504, "ts": 1716454218038274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038275, "dur": 0,
    "args": {
      "External id": 37891,
      "cbid": 203, "correlation": 37891
    }
  },
  {
    "ph": "f", "id": 37891, "pid": 76337, "tid": -914061504, "ts": 1716454218038275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038275, "dur": 0,
    "args": {
      "External id": 37892,
      "cbid": 205, "correlation": 37892
    }
  },
  {
    "ph": "f", "id": 37892, "pid": 76337, "tid": -914061504, "ts": 1716454218038275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092342, "dur": 5,
    "args": {
      "External id": 37896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37896,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37896, "pid": 5, "tid": 7, "ts": 1716454218092342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038291, "dur": 12,
    "args": {
      "External id": 37896,
      "cbid": 211, "correlation": 37896
    }
  },
  {
    "ph": "s", "id": 37896, "pid": 76337, "tid": -914061504, "ts": 1716454218038291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218092349, "dur": 36,
    "args": {
      "External id": 37898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37898,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 37898, "pid": 5, "tid": 7, "ts": 1716454218092349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038309, "dur": 7,
    "args": {
      "External id": 37898,
      "cbid": 211, "correlation": 37898
    }
  },
  {
    "ph": "s", "id": 37898, "pid": 76337, "tid": -914061504, "ts": 1716454218038309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092386, "dur": 6,
    "args": {
      "External id": 37900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37900,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37900, "pid": 5, "tid": 7, "ts": 1716454218092386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038320, "dur": 5,
    "args": {
      "External id": 37900,
      "cbid": 211, "correlation": 37900
    }
  },
  {
    "ph": "s", "id": 37900, "pid": 76337, "tid": -914061504, "ts": 1716454218038320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092393, "dur": 7,
    "args": {
      "External id": 37906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37906,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37906, "pid": 5, "tid": 7, "ts": 1716454218092393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038347, "dur": 8,
    "args": {
      "External id": 37906,
      "cbid": 211, "correlation": 37906
    }
  },
  {
    "ph": "s", "id": 37906, "pid": 76337, "tid": -914061504, "ts": 1716454218038347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218092402, "dur": 5,
    "args": {
      "External id": 37914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37914,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37914, "pid": 5, "tid": 7, "ts": 1716454218092402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038379, "dur": 8,
    "args": {
      "External id": 37914,
      "cbid": 211, "correlation": 37914
    }
  },
  {
    "ph": "s", "id": 37914, "pid": 76337, "tid": -914061504, "ts": 1716454218038379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092408, "dur": 10,
    "args": {
      "External id": 37934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37934,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 37934, "pid": 5, "tid": 7, "ts": 1716454218092408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038453, "dur": 11,
    "args": {
      "External id": 37934,
      "cbid": 211, "correlation": 37934
    }
  },
  {
    "ph": "s", "id": 37934, "pid": 76337, "tid": -914061504, "ts": 1716454218038453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218092419, "dur": 4,
    "args": {
      "External id": 37946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37946,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 37946, "pid": 5, "tid": 7, "ts": 1716454218092419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038474, "dur": 6,
    "args": {
      "External id": 37946,
      "cbid": 211, "correlation": 37946
    }
  },
  {
    "ph": "s", "id": 37946, "pid": 76337, "tid": -914061504, "ts": 1716454218038474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092425, "dur": 8,
    "args": {
      "External id": 37949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37949,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37949, "pid": 5, "tid": 7, "ts": 1716454218092425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038491, "dur": 7,
    "args": {
      "External id": 37949,
      "cbid": 211, "correlation": 37949
    }
  },
  {
    "ph": "s", "id": 37949, "pid": 76337, "tid": -914061504, "ts": 1716454218038491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218092434, "dur": 5,
    "args": {
      "External id": 37958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37958,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37958, "pid": 5, "tid": 7, "ts": 1716454218092434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038531, "dur": 10,
    "args": {
      "External id": 37958,
      "cbid": 211, "correlation": 37958
    }
  },
  {
    "ph": "s", "id": 37958, "pid": 76337, "tid": -914061504, "ts": 1716454218038531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038583, "dur": 0,
    "args": {
      "External id": 37968,
      "cbid": 317, "correlation": 37968
    }
  },
  {
    "ph": "f", "id": 37968, "pid": 76337, "tid": -914061504, "ts": 1716454218038583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038583, "dur": 0,
    "args": {
      "External id": 37969,
      "cbid": 203, "correlation": 37969
    }
  },
  {
    "ph": "f", "id": 37969, "pid": 76337, "tid": -914061504, "ts": 1716454218038583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038584, "dur": 0,
    "args": {
      "External id": 37970,
      "cbid": 205, "correlation": 37970
    }
  },
  {
    "ph": "f", "id": 37970, "pid": 76337, "tid": -914061504, "ts": 1716454218038584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092440, "dur": 5,
    "args": {
      "External id": 37974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37974,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37974, "pid": 5, "tid": 7, "ts": 1716454218092440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038597, "dur": 11,
    "args": {
      "External id": 37974,
      "cbid": 211, "correlation": 37974
    }
  },
  {
    "ph": "s", "id": 37974, "pid": 76337, "tid": -914061504, "ts": 1716454218038597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092446, "dur": 157,
    "args": {
      "External id": 37976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37976,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37976, "pid": 5, "tid": 7, "ts": 1716454218092446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038611, "dur": 5,
    "args": {
      "External id": 37976,
      "cbid": 211, "correlation": 37976
    }
  },
  {
    "ph": "s", "id": 37976, "pid": 76337, "tid": -914061504, "ts": 1716454218038611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218092606, "dur": 1,
    "args": {
      "External id": 37978,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 37978,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 37978, "pid": 5, "tid": 7, "ts": 1716454218092606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038622, "dur": 6,
    "args": {
      "External id": 37978,
      "cbid": 51, "correlation": 37978
    }
  },
  {
    "ph": "s", "id": 37978, "pid": 76337, "tid": -914061504, "ts": 1716454218038622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218092610, "dur": 263,
    "args": {
      "External id": 37979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37979,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 37979, "pid": 5, "tid": 7, "ts": 1716454218092610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038629, "dur": 6,
    "args": {
      "External id": 37979,
      "cbid": 211, "correlation": 37979
    }
  },
  {
    "ph": "s", "id": 37979, "pid": 76337, "tid": -914061504, "ts": 1716454218038629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092874, "dur": 5,
    "args": {
      "External id": 37981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 37981, "pid": 5, "tid": 7, "ts": 1716454218092874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038639, "dur": 5,
    "args": {
      "External id": 37981,
      "cbid": 211, "correlation": 37981
    }
  },
  {
    "ph": "s", "id": 37981, "pid": 76337, "tid": -914061504, "ts": 1716454218038639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092881, "dur": 6,
    "args": {
      "External id": 37987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 37987, "pid": 5, "tid": 7, "ts": 1716454218092881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038667, "dur": 9,
    "args": {
      "External id": 37987,
      "cbid": 211, "correlation": 37987
    }
  },
  {
    "ph": "s", "id": 37987, "pid": 76337, "tid": -914061504, "ts": 1716454218038667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218092888, "dur": 3,
    "args": {
      "External id": 37995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 37995,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 37995, "pid": 5, "tid": 7, "ts": 1716454218092888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038710, "dur": 9,
    "args": {
      "External id": 37995,
      "cbid": 211, "correlation": 37995
    }
  },
  {
    "ph": "s", "id": 37995, "pid": 76337, "tid": -914061504, "ts": 1716454218038710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038775, "dur": 1,
    "args": {
      "External id": 38011,
      "cbid": 251, "correlation": 38011
    }
  },
  {
    "ph": "f", "id": 38011, "pid": 76337, "tid": -914061504, "ts": 1716454218038775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038780, "dur": 0,
    "args": {
      "External id": 38013,
      "cbid": 251, "correlation": 38013
    }
  },
  {
    "ph": "f", "id": 38013, "pid": 76337, "tid": -914061504, "ts": 1716454218038780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218092893, "dur": 12,
    "args": {
      "External id": 38014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38014,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38014, "pid": 5, "tid": 7, "ts": 1716454218092893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038782, "dur": 12,
    "args": {
      "External id": 38014,
      "cbid": 211, "correlation": 38014
    }
  },
  {
    "ph": "s", "id": 38014, "pid": 76337, "tid": -914061504, "ts": 1716454218038782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218092906, "dur": 5,
    "args": {
      "External id": 38016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38016,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38016, "pid": 5, "tid": 7, "ts": 1716454218092906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038795, "dur": 5,
    "args": {
      "External id": 38016,
      "cbid": 211, "correlation": 38016
    }
  },
  {
    "ph": "s", "id": 38016, "pid": 76337, "tid": -914061504, "ts": 1716454218038795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092912, "dur": 5,
    "args": {
      "External id": 38026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38026, "pid": 5, "tid": 7, "ts": 1716454218092912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038853, "dur": 12,
    "args": {
      "External id": 38026,
      "cbid": 211, "correlation": 38026
    }
  },
  {
    "ph": "s", "id": 38026, "pid": 76337, "tid": -914061504, "ts": 1716454218038853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092919, "dur": 10,
    "args": {
      "External id": 38046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38046,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38046, "pid": 5, "tid": 7, "ts": 1716454218092919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038919, "dur": 11,
    "args": {
      "External id": 38046,
      "cbid": 211, "correlation": 38046
    }
  },
  {
    "ph": "s", "id": 38046, "pid": 76337, "tid": -914061504, "ts": 1716454218038919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218092930, "dur": 4,
    "args": {
      "External id": 38058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38058,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 38058, "pid": 5, "tid": 7, "ts": 1716454218092930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038939, "dur": 6,
    "args": {
      "External id": 38058,
      "cbid": 211, "correlation": 38058
    }
  },
  {
    "ph": "s", "id": 38058, "pid": 76337, "tid": -914061504, "ts": 1716454218038939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218092935, "dur": 6,
    "args": {
      "External id": 38061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38061, "pid": 5, "tid": 7, "ts": 1716454218092935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218038958, "dur": 7,
    "args": {
      "External id": 38061,
      "cbid": 211, "correlation": 38061
    }
  },
  {
    "ph": "s", "id": 38061, "pid": 76337, "tid": -914061504, "ts": 1716454218038958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218092943, "dur": 4,
    "args": {
      "External id": 38070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38070,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38070, "pid": 5, "tid": 7, "ts": 1716454218092943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039007, "dur": 11,
    "args": {
      "External id": 38070,
      "cbid": 211, "correlation": 38070
    }
  },
  {
    "ph": "s", "id": 38070, "pid": 76337, "tid": -914061504, "ts": 1716454218039007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039071, "dur": 0,
    "args": {
      "External id": 38080,
      "cbid": 317, "correlation": 38080
    }
  },
  {
    "ph": "f", "id": 38080, "pid": 76337, "tid": -914061504, "ts": 1716454218039071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039072, "dur": 0,
    "args": {
      "External id": 38081,
      "cbid": 203, "correlation": 38081
    }
  },
  {
    "ph": "f", "id": 38081, "pid": 76337, "tid": -914061504, "ts": 1716454218039072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039073, "dur": 0,
    "args": {
      "External id": 38082,
      "cbid": 205, "correlation": 38082
    }
  },
  {
    "ph": "f", "id": 38082, "pid": 76337, "tid": -914061504, "ts": 1716454218039073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092949, "dur": 5,
    "args": {
      "External id": 38086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38086,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38086, "pid": 5, "tid": 7, "ts": 1716454218092949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039087, "dur": 12,
    "args": {
      "External id": 38086,
      "cbid": 211, "correlation": 38086
    }
  },
  {
    "ph": "s", "id": 38086, "pid": 76337, "tid": -914061504, "ts": 1716454218039087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218092955, "dur": 158,
    "args": {
      "External id": 38088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38088, "pid": 5, "tid": 7, "ts": 1716454218092955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039101, "dur": 5,
    "args": {
      "External id": 38088,
      "cbid": 211, "correlation": 38088
    }
  },
  {
    "ph": "s", "id": 38088, "pid": 76337, "tid": -914061504, "ts": 1716454218039101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218093115, "dur": 1,
    "args": {
      "External id": 38090,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38090,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 38090, "pid": 5, "tid": 7, "ts": 1716454218093115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039111, "dur": 6,
    "args": {
      "External id": 38090,
      "cbid": 51, "correlation": 38090
    }
  },
  {
    "ph": "s", "id": 38090, "pid": 76337, "tid": -914061504, "ts": 1716454218039111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218093119, "dur": 252,
    "args": {
      "External id": 38091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38091,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38091, "pid": 5, "tid": 7, "ts": 1716454218093119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039119, "dur": 6,
    "args": {
      "External id": 38091,
      "cbid": 211, "correlation": 38091
    }
  },
  {
    "ph": "s", "id": 38091, "pid": 76337, "tid": -914061504, "ts": 1716454218039119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218093372, "dur": 6,
    "args": {
      "External id": 38093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38093,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38093, "pid": 5, "tid": 7, "ts": 1716454218093372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039129, "dur": 5,
    "args": {
      "External id": 38093,
      "cbid": 211, "correlation": 38093
    }
  },
  {
    "ph": "s", "id": 38093, "pid": 76337, "tid": -914061504, "ts": 1716454218039129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218093380, "dur": 6,
    "args": {
      "External id": 38099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38099, "pid": 5, "tid": 7, "ts": 1716454218093380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039158, "dur": 8,
    "args": {
      "External id": 38099,
      "cbid": 211, "correlation": 38099
    }
  },
  {
    "ph": "s", "id": 38099, "pid": 76337, "tid": -914061504, "ts": 1716454218039158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218093387, "dur": 5,
    "args": {
      "External id": 38107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38107,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38107, "pid": 5, "tid": 7, "ts": 1716454218093387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039191, "dur": 8,
    "args": {
      "External id": 38107,
      "cbid": 211, "correlation": 38107
    }
  },
  {
    "ph": "s", "id": 38107, "pid": 76337, "tid": -914061504, "ts": 1716454218039191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218093393, "dur": 4,
    "args": {
      "External id": 38115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38115,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38115, "pid": 5, "tid": 7, "ts": 1716454218093393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039219, "dur": 9,
    "args": {
      "External id": 38115,
      "cbid": 211, "correlation": 38115
    }
  },
  {
    "ph": "s", "id": 38115, "pid": 76337, "tid": -914061504, "ts": 1716454218039219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218093398, "dur": 11,
    "args": {
      "External id": 38124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38124,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38124, "pid": 5, "tid": 7, "ts": 1716454218093398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039308, "dur": 13,
    "args": {
      "External id": 38124,
      "cbid": 211, "correlation": 38124
    }
  },
  {
    "ph": "s", "id": 38124, "pid": 76337, "tid": -914061504, "ts": 1716454218039308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218093411, "dur": 12,
    "args": {
      "External id": 38144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38144,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38144, "pid": 5, "tid": 7, "ts": 1716454218093411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039376, "dur": 11,
    "args": {
      "External id": 38144,
      "cbid": 211, "correlation": 38144
    }
  },
  {
    "ph": "s", "id": 38144, "pid": 76337, "tid": -914061504, "ts": 1716454218039376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218093424, "dur": 4,
    "args": {
      "External id": 38156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38156,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38156, "pid": 5, "tid": 7, "ts": 1716454218093424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039398, "dur": 6,
    "args": {
      "External id": 38156,
      "cbid": 211, "correlation": 38156
    }
  },
  {
    "ph": "s", "id": 38156, "pid": 76337, "tid": -914061504, "ts": 1716454218039398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218093429, "dur": 10,
    "args": {
      "External id": 38159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38159,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38159, "pid": 5, "tid": 7, "ts": 1716454218093429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039416, "dur": 6,
    "args": {
      "External id": 38159,
      "cbid": 211, "correlation": 38159
    }
  },
  {
    "ph": "s", "id": 38159, "pid": 76337, "tid": -914061504, "ts": 1716454218039416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218093440, "dur": 6,
    "args": {
      "External id": 38168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38168,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38168, "pid": 5, "tid": 7, "ts": 1716454218093440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039454, "dur": 9,
    "args": {
      "External id": 38168,
      "cbid": 211, "correlation": 38168
    }
  },
  {
    "ph": "s", "id": 38168, "pid": 76337, "tid": -914061504, "ts": 1716454218039454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039506, "dur": 0,
    "args": {
      "External id": 38178,
      "cbid": 317, "correlation": 38178
    }
  },
  {
    "ph": "f", "id": 38178, "pid": 76337, "tid": -914061504, "ts": 1716454218039506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039507, "dur": 0,
    "args": {
      "External id": 38179,
      "cbid": 203, "correlation": 38179
    }
  },
  {
    "ph": "f", "id": 38179, "pid": 76337, "tid": -914061504, "ts": 1716454218039507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039508, "dur": 0,
    "args": {
      "External id": 38180,
      "cbid": 205, "correlation": 38180
    }
  },
  {
    "ph": "f", "id": 38180, "pid": 76337, "tid": -914061504, "ts": 1716454218039508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218093447, "dur": 6,
    "args": {
      "External id": 38184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38184,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38184, "pid": 5, "tid": 7, "ts": 1716454218093447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039524, "dur": 11,
    "args": {
      "External id": 38184,
      "cbid": 211, "correlation": 38184
    }
  },
  {
    "ph": "s", "id": 38184, "pid": 76337, "tid": -914061504, "ts": 1716454218039524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218093455, "dur": 312,
    "args": {
      "External id": 38186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38186, "pid": 5, "tid": 7, "ts": 1716454218093455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039538, "dur": 5,
    "args": {
      "External id": 38186,
      "cbid": 211, "correlation": 38186
    }
  },
  {
    "ph": "s", "id": 38186, "pid": 76337, "tid": -914061504, "ts": 1716454218039538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218093769, "dur": 1,
    "args": {
      "External id": 38188,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38188,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1271186440677966
    }
  },
  {
    "ph": "f", "id": 38188, "pid": 5, "tid": 7, "ts": 1716454218093769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039549, "dur": 7,
    "args": {
      "External id": 38188,
      "cbid": 51, "correlation": 38188
    }
  },
  {
    "ph": "s", "id": 38188, "pid": 76337, "tid": -914061504, "ts": 1716454218039549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218093773, "dur": 484,
    "args": {
      "External id": 38189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38189,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38189, "pid": 5, "tid": 7, "ts": 1716454218093773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039557, "dur": 7,
    "args": {
      "External id": 38189,
      "cbid": 211, "correlation": 38189
    }
  },
  {
    "ph": "s", "id": 38189, "pid": 76337, "tid": -914061504, "ts": 1716454218039557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094258, "dur": 5,
    "args": {
      "External id": 38191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38191,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38191, "pid": 5, "tid": 7, "ts": 1716454218094258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039568, "dur": 5,
    "args": {
      "External id": 38191,
      "cbid": 211, "correlation": 38191
    }
  },
  {
    "ph": "s", "id": 38191, "pid": 76337, "tid": -914061504, "ts": 1716454218039568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218094265, "dur": 6,
    "args": {
      "External id": 38197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38197, "pid": 5, "tid": 7, "ts": 1716454218094265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039596, "dur": 8,
    "args": {
      "External id": 38197,
      "cbid": 211, "correlation": 38197
    }
  },
  {
    "ph": "s", "id": 38197, "pid": 76337, "tid": -914061504, "ts": 1716454218039596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218094272, "dur": 3,
    "args": {
      "External id": 38205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38205,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 38205, "pid": 5, "tid": 7, "ts": 1716454218094272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039639, "dur": 9,
    "args": {
      "External id": 38205,
      "cbid": 211, "correlation": 38205
    }
  },
  {
    "ph": "s", "id": 38205, "pid": 76337, "tid": -914061504, "ts": 1716454218039639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039704, "dur": 1,
    "args": {
      "External id": 38221,
      "cbid": 251, "correlation": 38221
    }
  },
  {
    "ph": "f", "id": 38221, "pid": 76337, "tid": -914061504, "ts": 1716454218039704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039709, "dur": 0,
    "args": {
      "External id": 38223,
      "cbid": 251, "correlation": 38223
    }
  },
  {
    "ph": "f", "id": 38223, "pid": 76337, "tid": -914061504, "ts": 1716454218039709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218094276, "dur": 10,
    "args": {
      "External id": 38224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38224,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38224, "pid": 5, "tid": 7, "ts": 1716454218094276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039711, "dur": 12,
    "args": {
      "External id": 38224,
      "cbid": 211, "correlation": 38224
    }
  },
  {
    "ph": "s", "id": 38224, "pid": 76337, "tid": -914061504, "ts": 1716454218039711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218094288, "dur": 5,
    "args": {
      "External id": 38226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38226,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38226, "pid": 5, "tid": 7, "ts": 1716454218094288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039725, "dur": 5,
    "args": {
      "External id": 38226,
      "cbid": 211, "correlation": 38226
    }
  },
  {
    "ph": "s", "id": 38226, "pid": 76337, "tid": -914061504, "ts": 1716454218039725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218094294, "dur": 5,
    "args": {
      "External id": 38236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38236, "pid": 5, "tid": 7, "ts": 1716454218094294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039782, "dur": 12,
    "args": {
      "External id": 38236,
      "cbid": 211, "correlation": 38236
    }
  },
  {
    "ph": "s", "id": 38236, "pid": 76337, "tid": -914061504, "ts": 1716454218039782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094301, "dur": 9,
    "args": {
      "External id": 38256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38256,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38256, "pid": 5, "tid": 7, "ts": 1716454218094301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039848, "dur": 11,
    "args": {
      "External id": 38256,
      "cbid": 211, "correlation": 38256
    }
  },
  {
    "ph": "s", "id": 38256, "pid": 76337, "tid": -914061504, "ts": 1716454218039848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218094311, "dur": 3,
    "args": {
      "External id": 38268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38268,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 38268, "pid": 5, "tid": 7, "ts": 1716454218094311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039869, "dur": 6,
    "args": {
      "External id": 38268,
      "cbid": 211, "correlation": 38268
    }
  },
  {
    "ph": "s", "id": 38268, "pid": 76337, "tid": -914061504, "ts": 1716454218039869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218094316, "dur": 7,
    "args": {
      "External id": 38271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38271, "pid": 5, "tid": 7, "ts": 1716454218094316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039887, "dur": 6,
    "args": {
      "External id": 38271,
      "cbid": 211, "correlation": 38271
    }
  },
  {
    "ph": "s", "id": 38271, "pid": 76337, "tid": -914061504, "ts": 1716454218039887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218094324, "dur": 4,
    "args": {
      "External id": 38280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38280,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38280, "pid": 5, "tid": 7, "ts": 1716454218094324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039926, "dur": 10,
    "args": {
      "External id": 38280,
      "cbid": 211, "correlation": 38280
    }
  },
  {
    "ph": "s", "id": 38280, "pid": 76337, "tid": -914061504, "ts": 1716454218039926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039997, "dur": 0,
    "args": {
      "External id": 38290,
      "cbid": 317, "correlation": 38290
    }
  },
  {
    "ph": "f", "id": 38290, "pid": 76337, "tid": -914061504, "ts": 1716454218039997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039998, "dur": 0,
    "args": {
      "External id": 38291,
      "cbid": 203, "correlation": 38291
    }
  },
  {
    "ph": "f", "id": 38291, "pid": 76337, "tid": -914061504, "ts": 1716454218039998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218039999, "dur": 0,
    "args": {
      "External id": 38292,
      "cbid": 205, "correlation": 38292
    }
  },
  {
    "ph": "f", "id": 38292, "pid": 76337, "tid": -914061504, "ts": 1716454218039999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094330, "dur": 5,
    "args": {
      "External id": 38296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38296,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38296, "pid": 5, "tid": 7, "ts": 1716454218094330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040013, "dur": 12,
    "args": {
      "External id": 38296,
      "cbid": 211, "correlation": 38296
    }
  },
  {
    "ph": "s", "id": 38296, "pid": 76337, "tid": -914061504, "ts": 1716454218040013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094336, "dur": 157,
    "args": {
      "External id": 38298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38298,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38298, "pid": 5, "tid": 7, "ts": 1716454218094336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040027, "dur": 5,
    "args": {
      "External id": 38298,
      "cbid": 211, "correlation": 38298
    }
  },
  {
    "ph": "s", "id": 38298, "pid": 76337, "tid": -914061504, "ts": 1716454218040027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218094495, "dur": 1,
    "args": {
      "External id": 38300,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38300,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 38300, "pid": 5, "tid": 7, "ts": 1716454218094495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040038, "dur": 7,
    "args": {
      "External id": 38300,
      "cbid": 51, "correlation": 38300
    }
  },
  {
    "ph": "s", "id": 38300, "pid": 76337, "tid": -914061504, "ts": 1716454218040038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218094499, "dur": 252,
    "args": {
      "External id": 38301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38301,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38301, "pid": 5, "tid": 7, "ts": 1716454218094499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040046, "dur": 6,
    "args": {
      "External id": 38301,
      "cbid": 211, "correlation": 38301
    }
  },
  {
    "ph": "s", "id": 38301, "pid": 76337, "tid": -914061504, "ts": 1716454218040046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094752, "dur": 5,
    "args": {
      "External id": 38303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38303, "pid": 5, "tid": 7, "ts": 1716454218094752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040056, "dur": 5,
    "args": {
      "External id": 38303,
      "cbid": 211, "correlation": 38303
    }
  },
  {
    "ph": "s", "id": 38303, "pid": 76337, "tid": -914061504, "ts": 1716454218040056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218094759, "dur": 6,
    "args": {
      "External id": 38309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38309, "pid": 5, "tid": 7, "ts": 1716454218094759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040084, "dur": 9,
    "args": {
      "External id": 38309,
      "cbid": 211, "correlation": 38309
    }
  },
  {
    "ph": "s", "id": 38309, "pid": 76337, "tid": -914061504, "ts": 1716454218040084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040143, "dur": 0,
    "args": {
      "External id": 38319,
      "cbid": 317, "correlation": 38319
    }
  },
  {
    "ph": "f", "id": 38319, "pid": 76337, "tid": -914061504, "ts": 1716454218040143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040144, "dur": 0,
    "args": {
      "External id": 38320,
      "cbid": 203, "correlation": 38320
    }
  },
  {
    "ph": "f", "id": 38320, "pid": 76337, "tid": -914061504, "ts": 1716454218040144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040145, "dur": 0,
    "args": {
      "External id": 38321,
      "cbid": 205, "correlation": 38321
    }
  },
  {
    "ph": "f", "id": 38321, "pid": 76337, "tid": -914061504, "ts": 1716454218040145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094767, "dur": 8,
    "args": {
      "External id": 38325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38325,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38325, "pid": 5, "tid": 7, "ts": 1716454218094767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040160, "dur": 11,
    "args": {
      "External id": 38325,
      "cbid": 211, "correlation": 38325
    }
  },
  {
    "ph": "s", "id": 38325, "pid": 76337, "tid": -914061504, "ts": 1716454218040160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218094776, "dur": 3,
    "args": {
      "External id": 38327,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38327,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3043478260869565
    }
  },
  {
    "ph": "f", "id": 38327, "pid": 5, "tid": 7, "ts": 1716454218094776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040178, "dur": 15,
    "args": {
      "External id": 38327,
      "cbid": 51, "correlation": 38327
    }
  },
  {
    "ph": "s", "id": 38327, "pid": 76337, "tid": -914061504, "ts": 1716454218040178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218094780, "dur": 97,
    "args": {
      "External id": 38328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38328,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 38328, "pid": 5, "tid": 7, "ts": 1716454218094780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040194, "dur": 7,
    "args": {
      "External id": 38328,
      "cbid": 211, "correlation": 38328
    }
  },
  {
    "ph": "s", "id": 38328, "pid": 76337, "tid": -914061504, "ts": 1716454218040194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094878, "dur": 6,
    "args": {
      "External id": 38330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38330,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38330, "pid": 5, "tid": 7, "ts": 1716454218094878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040205, "dur": 5,
    "args": {
      "External id": 38330,
      "cbid": 211, "correlation": 38330
    }
  },
  {
    "ph": "s", "id": 38330, "pid": 76337, "tid": -914061504, "ts": 1716454218040205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218094885, "dur": 6,
    "args": {
      "External id": 38336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38336,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38336, "pid": 5, "tid": 7, "ts": 1716454218094885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040232, "dur": 8,
    "args": {
      "External id": 38336,
      "cbid": 211, "correlation": 38336
    }
  },
  {
    "ph": "s", "id": 38336, "pid": 76337, "tid": -914061504, "ts": 1716454218040232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218094892, "dur": 5,
    "args": {
      "External id": 38344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38344,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38344, "pid": 5, "tid": 7, "ts": 1716454218094892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040261, "dur": 7,
    "args": {
      "External id": 38344,
      "cbid": 211, "correlation": 38344
    }
  },
  {
    "ph": "s", "id": 38344, "pid": 76337, "tid": -914061504, "ts": 1716454218040261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218094898, "dur": 4,
    "args": {
      "External id": 38352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38352,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38352, "pid": 5, "tid": 7, "ts": 1716454218094898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040289, "dur": 8,
    "args": {
      "External id": 38352,
      "cbid": 211, "correlation": 38352
    }
  },
  {
    "ph": "s", "id": 38352, "pid": 76337, "tid": -914061504, "ts": 1716454218040289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218094904, "dur": 11,
    "args": {
      "External id": 38361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38361,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38361, "pid": 5, "tid": 7, "ts": 1716454218094904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040333, "dur": 10,
    "args": {
      "External id": 38361,
      "cbid": 211, "correlation": 38361
    }
  },
  {
    "ph": "s", "id": 38361, "pid": 76337, "tid": -914061504, "ts": 1716454218040333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094916, "dur": 12,
    "args": {
      "External id": 38381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38381,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38381, "pid": 5, "tid": 7, "ts": 1716454218094916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040402, "dur": 12,
    "args": {
      "External id": 38381,
      "cbid": 211, "correlation": 38381
    }
  },
  {
    "ph": "s", "id": 38381, "pid": 76337, "tid": -914061504, "ts": 1716454218040402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218094929, "dur": 4,
    "args": {
      "External id": 38393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38393,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38393, "pid": 5, "tid": 7, "ts": 1716454218094929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040424, "dur": 6,
    "args": {
      "External id": 38393,
      "cbid": 211, "correlation": 38393
    }
  },
  {
    "ph": "s", "id": 38393, "pid": 76337, "tid": -914061504, "ts": 1716454218040424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218094934, "dur": 11,
    "args": {
      "External id": 38396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38396, "pid": 5, "tid": 7, "ts": 1716454218094934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040441, "dur": 6,
    "args": {
      "External id": 38396,
      "cbid": 211, "correlation": 38396
    }
  },
  {
    "ph": "s", "id": 38396, "pid": 76337, "tid": -914061504, "ts": 1716454218040441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218094946, "dur": 6,
    "args": {
      "External id": 38405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38405,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38405, "pid": 5, "tid": 7, "ts": 1716454218094946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040479, "dur": 10,
    "args": {
      "External id": 38405,
      "cbid": 211, "correlation": 38405
    }
  },
  {
    "ph": "s", "id": 38405, "pid": 76337, "tid": -914061504, "ts": 1716454218040479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040531, "dur": 0,
    "args": {
      "External id": 38415,
      "cbid": 317, "correlation": 38415
    }
  },
  {
    "ph": "f", "id": 38415, "pid": 76337, "tid": -914061504, "ts": 1716454218040531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040532, "dur": 0,
    "args": {
      "External id": 38416,
      "cbid": 203, "correlation": 38416
    }
  },
  {
    "ph": "f", "id": 38416, "pid": 76337, "tid": -914061504, "ts": 1716454218040532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040532, "dur": 0,
    "args": {
      "External id": 38417,
      "cbid": 205, "correlation": 38417
    }
  },
  {
    "ph": "f", "id": 38417, "pid": 76337, "tid": -914061504, "ts": 1716454218040532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094953, "dur": 6,
    "args": {
      "External id": 38421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38421,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38421, "pid": 5, "tid": 7, "ts": 1716454218094953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040546, "dur": 11,
    "args": {
      "External id": 38421,
      "cbid": 211, "correlation": 38421
    }
  },
  {
    "ph": "s", "id": 38421, "pid": 76337, "tid": -914061504, "ts": 1716454218040546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218094961, "dur": 313,
    "args": {
      "External id": 38423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38423,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38423, "pid": 5, "tid": 7, "ts": 1716454218094961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040559, "dur": 6,
    "args": {
      "External id": 38423,
      "cbid": 211, "correlation": 38423
    }
  },
  {
    "ph": "s", "id": 38423, "pid": 76337, "tid": -914061504, "ts": 1716454218040559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218095276, "dur": 1,
    "args": {
      "External id": 38425,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38425,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 38425, "pid": 5, "tid": 7, "ts": 1716454218095276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040571, "dur": 6,
    "args": {
      "External id": 38425,
      "cbid": 51, "correlation": 38425
    }
  },
  {
    "ph": "s", "id": 38425, "pid": 76337, "tid": -914061504, "ts": 1716454218040571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218095280, "dur": 486,
    "args": {
      "External id": 38426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38426,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38426, "pid": 5, "tid": 7, "ts": 1716454218095280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040578, "dur": 6,
    "args": {
      "External id": 38426,
      "cbid": 211, "correlation": 38426
    }
  },
  {
    "ph": "s", "id": 38426, "pid": 76337, "tid": -914061504, "ts": 1716454218040578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218095767, "dur": 6,
    "args": {
      "External id": 38428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38428,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38428, "pid": 5, "tid": 7, "ts": 1716454218095767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040588, "dur": 5,
    "args": {
      "External id": 38428,
      "cbid": 211, "correlation": 38428
    }
  },
  {
    "ph": "s", "id": 38428, "pid": 76337, "tid": -914061504, "ts": 1716454218040588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218095774, "dur": 6,
    "args": {
      "External id": 38434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38434,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38434, "pid": 5, "tid": 7, "ts": 1716454218095774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040616, "dur": 10,
    "args": {
      "External id": 38434,
      "cbid": 211, "correlation": 38434
    }
  },
  {
    "ph": "s", "id": 38434, "pid": 76337, "tid": -914061504, "ts": 1716454218040616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218095781, "dur": 3,
    "args": {
      "External id": 38442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38442,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 38442, "pid": 5, "tid": 7, "ts": 1716454218095781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040660, "dur": 9,
    "args": {
      "External id": 38442,
      "cbid": 211, "correlation": 38442
    }
  },
  {
    "ph": "s", "id": 38442, "pid": 76337, "tid": -914061504, "ts": 1716454218040660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040724, "dur": 1,
    "args": {
      "External id": 38458,
      "cbid": 251, "correlation": 38458
    }
  },
  {
    "ph": "f", "id": 38458, "pid": 76337, "tid": -914061504, "ts": 1716454218040724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040729, "dur": 0,
    "args": {
      "External id": 38460,
      "cbid": 251, "correlation": 38460
    }
  },
  {
    "ph": "f", "id": 38460, "pid": 76337, "tid": -914061504, "ts": 1716454218040729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218095786, "dur": 12,
    "args": {
      "External id": 38461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38461,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38461, "pid": 5, "tid": 7, "ts": 1716454218095786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040731, "dur": 11,
    "args": {
      "External id": 38461,
      "cbid": 211, "correlation": 38461
    }
  },
  {
    "ph": "s", "id": 38461, "pid": 76337, "tid": -914061504, "ts": 1716454218040731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218095799, "dur": 5,
    "args": {
      "External id": 38463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38463,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38463, "pid": 5, "tid": 7, "ts": 1716454218095799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040744, "dur": 5,
    "args": {
      "External id": 38463,
      "cbid": 211, "correlation": 38463
    }
  },
  {
    "ph": "s", "id": 38463, "pid": 76337, "tid": -914061504, "ts": 1716454218040744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218095806, "dur": 6,
    "args": {
      "External id": 38473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38473, "pid": 5, "tid": 7, "ts": 1716454218095806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040801, "dur": 12,
    "args": {
      "External id": 38473,
      "cbid": 211, "correlation": 38473
    }
  },
  {
    "ph": "s", "id": 38473, "pid": 76337, "tid": -914061504, "ts": 1716454218040801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218095813, "dur": 9,
    "args": {
      "External id": 38493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38493,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38493, "pid": 5, "tid": 7, "ts": 1716454218095813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040867, "dur": 11,
    "args": {
      "External id": 38493,
      "cbid": 211, "correlation": 38493
    }
  },
  {
    "ph": "s", "id": 38493, "pid": 76337, "tid": -914061504, "ts": 1716454218040867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218095823, "dur": 4,
    "args": {
      "External id": 38505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38505,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 38505, "pid": 5, "tid": 7, "ts": 1716454218095823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040889, "dur": 6,
    "args": {
      "External id": 38505,
      "cbid": 211, "correlation": 38505
    }
  },
  {
    "ph": "s", "id": 38505, "pid": 76337, "tid": -914061504, "ts": 1716454218040889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218095828, "dur": 6,
    "args": {
      "External id": 38508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38508, "pid": 5, "tid": 7, "ts": 1716454218095828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040907, "dur": 6,
    "args": {
      "External id": 38508,
      "cbid": 211, "correlation": 38508
    }
  },
  {
    "ph": "s", "id": 38508, "pid": 76337, "tid": -914061504, "ts": 1716454218040907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218095836, "dur": 4,
    "args": {
      "External id": 38517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38517,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38517, "pid": 5, "tid": 7, "ts": 1716454218095836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218040948, "dur": 10,
    "args": {
      "External id": 38517,
      "cbid": 211, "correlation": 38517
    }
  },
  {
    "ph": "s", "id": 38517, "pid": 76337, "tid": -914061504, "ts": 1716454218040948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041019, "dur": 0,
    "args": {
      "External id": 38527,
      "cbid": 317, "correlation": 38527
    }
  },
  {
    "ph": "f", "id": 38527, "pid": 76337, "tid": -914061504, "ts": 1716454218041019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041020, "dur": 0,
    "args": {
      "External id": 38528,
      "cbid": 203, "correlation": 38528
    }
  },
  {
    "ph": "f", "id": 38528, "pid": 76337, "tid": -914061504, "ts": 1716454218041020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041020, "dur": 0,
    "args": {
      "External id": 38529,
      "cbid": 205, "correlation": 38529
    }
  },
  {
    "ph": "f", "id": 38529, "pid": 76337, "tid": -914061504, "ts": 1716454218041020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218095842, "dur": 5,
    "args": {
      "External id": 38533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38533,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38533, "pid": 5, "tid": 7, "ts": 1716454218095842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041034, "dur": 13,
    "args": {
      "External id": 38533,
      "cbid": 211, "correlation": 38533
    }
  },
  {
    "ph": "s", "id": 38533, "pid": 76337, "tid": -914061504, "ts": 1716454218041034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218095848, "dur": 157,
    "args": {
      "External id": 38535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38535,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38535, "pid": 5, "tid": 7, "ts": 1716454218095848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041050, "dur": 6,
    "args": {
      "External id": 38535,
      "cbid": 211, "correlation": 38535
    }
  },
  {
    "ph": "s", "id": 38535, "pid": 76337, "tid": -914061504, "ts": 1716454218041050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218096008, "dur": 1,
    "args": {
      "External id": 38537,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38537,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 38537, "pid": 5, "tid": 7, "ts": 1716454218096008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041061, "dur": 6,
    "args": {
      "External id": 38537,
      "cbid": 51, "correlation": 38537
    }
  },
  {
    "ph": "s", "id": 38537, "pid": 76337, "tid": -914061504, "ts": 1716454218041061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218096011, "dur": 252,
    "args": {
      "External id": 38538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38538,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38538, "pid": 5, "tid": 7, "ts": 1716454218096011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041068, "dur": 6,
    "args": {
      "External id": 38538,
      "cbid": 211, "correlation": 38538
    }
  },
  {
    "ph": "s", "id": 38538, "pid": 76337, "tid": -914061504, "ts": 1716454218041068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218096265, "dur": 6,
    "args": {
      "External id": 38540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38540,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38540, "pid": 5, "tid": 7, "ts": 1716454218096265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041078, "dur": 5,
    "args": {
      "External id": 38540,
      "cbid": 211, "correlation": 38540
    }
  },
  {
    "ph": "s", "id": 38540, "pid": 76337, "tid": -914061504, "ts": 1716454218041078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218096271, "dur": 6,
    "args": {
      "External id": 38546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38546, "pid": 5, "tid": 7, "ts": 1716454218096271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041107, "dur": 9,
    "args": {
      "External id": 38546,
      "cbid": 211, "correlation": 38546
    }
  },
  {
    "ph": "s", "id": 38546, "pid": 76337, "tid": -914061504, "ts": 1716454218041107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041165, "dur": 0,
    "args": {
      "External id": 38556,
      "cbid": 317, "correlation": 38556
    }
  },
  {
    "ph": "f", "id": 38556, "pid": 76337, "tid": -914061504, "ts": 1716454218041165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041166, "dur": 0,
    "args": {
      "External id": 38557,
      "cbid": 203, "correlation": 38557
    }
  },
  {
    "ph": "f", "id": 38557, "pid": 76337, "tid": -914061504, "ts": 1716454218041166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041167, "dur": 0,
    "args": {
      "External id": 38558,
      "cbid": 205, "correlation": 38558
    }
  },
  {
    "ph": "f", "id": 38558, "pid": 76337, "tid": -914061504, "ts": 1716454218041167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218096279, "dur": 7,
    "args": {
      "External id": 38562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38562,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38562, "pid": 5, "tid": 7, "ts": 1716454218096279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041178, "dur": 11,
    "args": {
      "External id": 38562,
      "cbid": 211, "correlation": 38562
    }
  },
  {
    "ph": "s", "id": 38562, "pid": 76337, "tid": -914061504, "ts": 1716454218041178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218096287, "dur": 3,
    "args": {
      "External id": 38564,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38564,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 38564, "pid": 5, "tid": 7, "ts": 1716454218096287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041194, "dur": 10,
    "args": {
      "External id": 38564,
      "cbid": 51, "correlation": 38564
    }
  },
  {
    "ph": "s", "id": 38564, "pid": 76337, "tid": -914061504, "ts": 1716454218041194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218096291, "dur": 95,
    "args": {
      "External id": 38565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38565,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 38565, "pid": 5, "tid": 7, "ts": 1716454218096291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041206, "dur": 6,
    "args": {
      "External id": 38565,
      "cbid": 211, "correlation": 38565
    }
  },
  {
    "ph": "s", "id": 38565, "pid": 76337, "tid": -914061504, "ts": 1716454218041206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218096388, "dur": 5,
    "args": {
      "External id": 38567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38567,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38567, "pid": 5, "tid": 7, "ts": 1716454218096388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041215, "dur": 5,
    "args": {
      "External id": 38567,
      "cbid": 211, "correlation": 38567
    }
  },
  {
    "ph": "s", "id": 38567, "pid": 76337, "tid": -914061504, "ts": 1716454218041215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218096395, "dur": 6,
    "args": {
      "External id": 38573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38573, "pid": 5, "tid": 7, "ts": 1716454218096395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041241, "dur": 9,
    "args": {
      "External id": 38573,
      "cbid": 211, "correlation": 38573
    }
  },
  {
    "ph": "s", "id": 38573, "pid": 76337, "tid": -914061504, "ts": 1716454218041241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218096402, "dur": 5,
    "args": {
      "External id": 38581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38581,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38581, "pid": 5, "tid": 7, "ts": 1716454218096402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041271, "dur": 7,
    "args": {
      "External id": 38581,
      "cbid": 211, "correlation": 38581
    }
  },
  {
    "ph": "s", "id": 38581, "pid": 76337, "tid": -914061504, "ts": 1716454218041271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218096408, "dur": 5,
    "args": {
      "External id": 38589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38589,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38589, "pid": 5, "tid": 7, "ts": 1716454218096408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041298, "dur": 9,
    "args": {
      "External id": 38589,
      "cbid": 211, "correlation": 38589
    }
  },
  {
    "ph": "s", "id": 38589, "pid": 76337, "tid": -914061504, "ts": 1716454218041298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218096414, "dur": 11,
    "args": {
      "External id": 38598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38598,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38598, "pid": 5, "tid": 7, "ts": 1716454218096414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041342, "dur": 11,
    "args": {
      "External id": 38598,
      "cbid": 211, "correlation": 38598
    }
  },
  {
    "ph": "s", "id": 38598, "pid": 76337, "tid": -914061504, "ts": 1716454218041342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218096426, "dur": 12,
    "args": {
      "External id": 38618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38618,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38618, "pid": 5, "tid": 7, "ts": 1716454218096426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041413, "dur": 11,
    "args": {
      "External id": 38618,
      "cbid": 211, "correlation": 38618
    }
  },
  {
    "ph": "s", "id": 38618, "pid": 76337, "tid": -914061504, "ts": 1716454218041413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218096439, "dur": 4,
    "args": {
      "External id": 38630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38630,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38630, "pid": 5, "tid": 7, "ts": 1716454218096439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041433, "dur": 6,
    "args": {
      "External id": 38630,
      "cbid": 211, "correlation": 38630
    }
  },
  {
    "ph": "s", "id": 38630, "pid": 76337, "tid": -914061504, "ts": 1716454218041433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218096445, "dur": 11,
    "args": {
      "External id": 38633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38633, "pid": 5, "tid": 7, "ts": 1716454218096445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041451, "dur": 7,
    "args": {
      "External id": 38633,
      "cbid": 211, "correlation": 38633
    }
  },
  {
    "ph": "s", "id": 38633, "pid": 76337, "tid": -914061504, "ts": 1716454218041451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218096457, "dur": 6,
    "args": {
      "External id": 38642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38642,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38642, "pid": 5, "tid": 7, "ts": 1716454218096457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041490, "dur": 9,
    "args": {
      "External id": 38642,
      "cbid": 211, "correlation": 38642
    }
  },
  {
    "ph": "s", "id": 38642, "pid": 76337, "tid": -914061504, "ts": 1716454218041490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041541, "dur": 0,
    "args": {
      "External id": 38652,
      "cbid": 317, "correlation": 38652
    }
  },
  {
    "ph": "f", "id": 38652, "pid": 76337, "tid": -914061504, "ts": 1716454218041541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041542, "dur": 0,
    "args": {
      "External id": 38653,
      "cbid": 203, "correlation": 38653
    }
  },
  {
    "ph": "f", "id": 38653, "pid": 76337, "tid": -914061504, "ts": 1716454218041542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041542, "dur": 0,
    "args": {
      "External id": 38654,
      "cbid": 205, "correlation": 38654
    }
  },
  {
    "ph": "f", "id": 38654, "pid": 76337, "tid": -914061504, "ts": 1716454218041542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218096464, "dur": 6,
    "args": {
      "External id": 38658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38658, "pid": 5, "tid": 7, "ts": 1716454218096464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041556, "dur": 12,
    "args": {
      "External id": 38658,
      "cbid": 211, "correlation": 38658
    }
  },
  {
    "ph": "s", "id": 38658, "pid": 76337, "tid": -914061504, "ts": 1716454218041556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218096472, "dur": 312,
    "args": {
      "External id": 38660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38660, "pid": 5, "tid": 7, "ts": 1716454218096472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041570, "dur": 5,
    "args": {
      "External id": 38660,
      "cbid": 211, "correlation": 38660
    }
  },
  {
    "ph": "s", "id": 38660, "pid": 76337, "tid": -914061504, "ts": 1716454218041570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218096786, "dur": 1,
    "args": {
      "External id": 38662,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38662,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 38662, "pid": 5, "tid": 7, "ts": 1716454218096786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041580, "dur": 6,
    "args": {
      "External id": 38662,
      "cbid": 51, "correlation": 38662
    }
  },
  {
    "ph": "s", "id": 38662, "pid": 76337, "tid": -914061504, "ts": 1716454218041580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218096789, "dur": 486,
    "args": {
      "External id": 38663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38663,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38663, "pid": 5, "tid": 7, "ts": 1716454218096789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041588, "dur": 6,
    "args": {
      "External id": 38663,
      "cbid": 211, "correlation": 38663
    }
  },
  {
    "ph": "s", "id": 38663, "pid": 76337, "tid": -914061504, "ts": 1716454218041588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097277, "dur": 5,
    "args": {
      "External id": 38665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38665, "pid": 5, "tid": 7, "ts": 1716454218097277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041597, "dur": 5,
    "args": {
      "External id": 38665,
      "cbid": 211, "correlation": 38665
    }
  },
  {
    "ph": "s", "id": 38665, "pid": 76337, "tid": -914061504, "ts": 1716454218041597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218097283, "dur": 6,
    "args": {
      "External id": 38671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38671, "pid": 5, "tid": 7, "ts": 1716454218097283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041626, "dur": 8,
    "args": {
      "External id": 38671,
      "cbid": 211, "correlation": 38671
    }
  },
  {
    "ph": "s", "id": 38671, "pid": 76337, "tid": -914061504, "ts": 1716454218041626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218097291, "dur": 3,
    "args": {
      "External id": 38679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38679,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 38679, "pid": 5, "tid": 7, "ts": 1716454218097291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041669, "dur": 10,
    "args": {
      "External id": 38679,
      "cbid": 211, "correlation": 38679
    }
  },
  {
    "ph": "s", "id": 38679, "pid": 76337, "tid": -914061504, "ts": 1716454218041669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041731, "dur": 1,
    "args": {
      "External id": 38695,
      "cbid": 251, "correlation": 38695
    }
  },
  {
    "ph": "f", "id": 38695, "pid": 76337, "tid": -914061504, "ts": 1716454218041731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041737, "dur": 0,
    "args": {
      "External id": 38697,
      "cbid": 251, "correlation": 38697
    }
  },
  {
    "ph": "f", "id": 38697, "pid": 76337, "tid": -914061504, "ts": 1716454218041737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218097295, "dur": 13,
    "args": {
      "External id": 38698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38698,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38698, "pid": 5, "tid": 7, "ts": 1716454218097295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041739, "dur": 11,
    "args": {
      "External id": 38698,
      "cbid": 211, "correlation": 38698
    }
  },
  {
    "ph": "s", "id": 38698, "pid": 76337, "tid": -914061504, "ts": 1716454218041739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218097309, "dur": 5,
    "args": {
      "External id": 38700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38700,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38700, "pid": 5, "tid": 7, "ts": 1716454218097309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041751, "dur": 5,
    "args": {
      "External id": 38700,
      "cbid": 211, "correlation": 38700
    }
  },
  {
    "ph": "s", "id": 38700, "pid": 76337, "tid": -914061504, "ts": 1716454218041751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218097315, "dur": 5,
    "args": {
      "External id": 38710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38710, "pid": 5, "tid": 7, "ts": 1716454218097315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041807, "dur": 12,
    "args": {
      "External id": 38710,
      "cbid": 211, "correlation": 38710
    }
  },
  {
    "ph": "s", "id": 38710, "pid": 76337, "tid": -914061504, "ts": 1716454218041807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097322, "dur": 9,
    "args": {
      "External id": 38730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38730,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38730, "pid": 5, "tid": 7, "ts": 1716454218097322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041873, "dur": 11,
    "args": {
      "External id": 38730,
      "cbid": 211, "correlation": 38730
    }
  },
  {
    "ph": "s", "id": 38730, "pid": 76337, "tid": -914061504, "ts": 1716454218041873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218097333, "dur": 3,
    "args": {
      "External id": 38742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38742,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 38742, "pid": 5, "tid": 7, "ts": 1716454218097333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041894, "dur": 6,
    "args": {
      "External id": 38742,
      "cbid": 211, "correlation": 38742
    }
  },
  {
    "ph": "s", "id": 38742, "pid": 76337, "tid": -914061504, "ts": 1716454218041894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218097337, "dur": 6,
    "args": {
      "External id": 38745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38745, "pid": 5, "tid": 7, "ts": 1716454218097337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041912, "dur": 6,
    "args": {
      "External id": 38745,
      "cbid": 211, "correlation": 38745
    }
  },
  {
    "ph": "s", "id": 38745, "pid": 76337, "tid": -914061504, "ts": 1716454218041912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218097345, "dur": 4,
    "args": {
      "External id": 38754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38754,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38754, "pid": 5, "tid": 7, "ts": 1716454218097345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218041952, "dur": 10,
    "args": {
      "External id": 38754,
      "cbid": 211, "correlation": 38754
    }
  },
  {
    "ph": "s", "id": 38754, "pid": 76337, "tid": -914061504, "ts": 1716454218041952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042023, "dur": 0,
    "args": {
      "External id": 38764,
      "cbid": 317, "correlation": 38764
    }
  },
  {
    "ph": "f", "id": 38764, "pid": 76337, "tid": -914061504, "ts": 1716454218042023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042024, "dur": 0,
    "args": {
      "External id": 38765,
      "cbid": 203, "correlation": 38765
    }
  },
  {
    "ph": "f", "id": 38765, "pid": 76337, "tid": -914061504, "ts": 1716454218042024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042025, "dur": 0,
    "args": {
      "External id": 38766,
      "cbid": 205, "correlation": 38766
    }
  },
  {
    "ph": "f", "id": 38766, "pid": 76337, "tid": -914061504, "ts": 1716454218042025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097351, "dur": 5,
    "args": {
      "External id": 38770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38770, "pid": 5, "tid": 7, "ts": 1716454218097351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042038, "dur": 13,
    "args": {
      "External id": 38770,
      "cbid": 211, "correlation": 38770
    }
  },
  {
    "ph": "s", "id": 38770, "pid": 76337, "tid": -914061504, "ts": 1716454218042038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097357, "dur": 158,
    "args": {
      "External id": 38772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38772, "pid": 5, "tid": 7, "ts": 1716454218097357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042054, "dur": 5,
    "args": {
      "External id": 38772,
      "cbid": 211, "correlation": 38772
    }
  },
  {
    "ph": "s", "id": 38772, "pid": 76337, "tid": -914061504, "ts": 1716454218042054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218097517, "dur": 1,
    "args": {
      "External id": 38774,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38774,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 38774, "pid": 5, "tid": 7, "ts": 1716454218097517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042065, "dur": 6,
    "args": {
      "External id": 38774,
      "cbid": 51, "correlation": 38774
    }
  },
  {
    "ph": "s", "id": 38774, "pid": 76337, "tid": -914061504, "ts": 1716454218042065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218097521, "dur": 252,
    "args": {
      "External id": 38775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38775,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38775, "pid": 5, "tid": 7, "ts": 1716454218097521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042072, "dur": 6,
    "args": {
      "External id": 38775,
      "cbid": 211, "correlation": 38775
    }
  },
  {
    "ph": "s", "id": 38775, "pid": 76337, "tid": -914061504, "ts": 1716454218042072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097774, "dur": 6,
    "args": {
      "External id": 38777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38777,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38777, "pid": 5, "tid": 7, "ts": 1716454218097774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042082, "dur": 5,
    "args": {
      "External id": 38777,
      "cbid": 211, "correlation": 38777
    }
  },
  {
    "ph": "s", "id": 38777, "pid": 76337, "tid": -914061504, "ts": 1716454218042082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218097781, "dur": 6,
    "args": {
      "External id": 38783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38783, "pid": 5, "tid": 7, "ts": 1716454218097781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042111, "dur": 8,
    "args": {
      "External id": 38783,
      "cbid": 211, "correlation": 38783
    }
  },
  {
    "ph": "s", "id": 38783, "pid": 76337, "tid": -914061504, "ts": 1716454218042111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042170, "dur": 0,
    "args": {
      "External id": 38793,
      "cbid": 317, "correlation": 38793
    }
  },
  {
    "ph": "f", "id": 38793, "pid": 76337, "tid": -914061504, "ts": 1716454218042170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042170, "dur": 0,
    "args": {
      "External id": 38794,
      "cbid": 203, "correlation": 38794
    }
  },
  {
    "ph": "f", "id": 38794, "pid": 76337, "tid": -914061504, "ts": 1716454218042170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042171, "dur": 0,
    "args": {
      "External id": 38795,
      "cbid": 205, "correlation": 38795
    }
  },
  {
    "ph": "f", "id": 38795, "pid": 76337, "tid": -914061504, "ts": 1716454218042171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097789, "dur": 8,
    "args": {
      "External id": 38799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38799,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38799, "pid": 5, "tid": 7, "ts": 1716454218097789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042183, "dur": 11,
    "args": {
      "External id": 38799,
      "cbid": 211, "correlation": 38799
    }
  },
  {
    "ph": "s", "id": 38799, "pid": 76337, "tid": -914061504, "ts": 1716454218042183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218097798, "dur": 3,
    "args": {
      "External id": 38801,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38801,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3157894736842106
    }
  },
  {
    "ph": "f", "id": 38801, "pid": 5, "tid": 7, "ts": 1716454218097798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042199, "dur": 10,
    "args": {
      "External id": 38801,
      "cbid": 51, "correlation": 38801
    }
  },
  {
    "ph": "s", "id": 38801, "pid": 76337, "tid": -914061504, "ts": 1716454218042199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218097802, "dur": 94,
    "args": {
      "External id": 38802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38802,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 38802, "pid": 5, "tid": 7, "ts": 1716454218097802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042210, "dur": 6,
    "args": {
      "External id": 38802,
      "cbid": 211, "correlation": 38802
    }
  },
  {
    "ph": "s", "id": 38802, "pid": 76337, "tid": -914061504, "ts": 1716454218042210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097898, "dur": 5,
    "args": {
      "External id": 38804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38804,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38804, "pid": 5, "tid": 7, "ts": 1716454218097898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042220, "dur": 5,
    "args": {
      "External id": 38804,
      "cbid": 211, "correlation": 38804
    }
  },
  {
    "ph": "s", "id": 38804, "pid": 76337, "tid": -914061504, "ts": 1716454218042220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218097905, "dur": 6,
    "args": {
      "External id": 38810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38810, "pid": 5, "tid": 7, "ts": 1716454218097905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042246, "dur": 9,
    "args": {
      "External id": 38810,
      "cbid": 211, "correlation": 38810
    }
  },
  {
    "ph": "s", "id": 38810, "pid": 76337, "tid": -914061504, "ts": 1716454218042246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218097912, "dur": 5,
    "args": {
      "External id": 38818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38818,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38818, "pid": 5, "tid": 7, "ts": 1716454218097912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042276, "dur": 8,
    "args": {
      "External id": 38818,
      "cbid": 211, "correlation": 38818
    }
  },
  {
    "ph": "s", "id": 38818, "pid": 76337, "tid": -914061504, "ts": 1716454218042276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218097918, "dur": 4,
    "args": {
      "External id": 38826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38826,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 38826, "pid": 5, "tid": 7, "ts": 1716454218097918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042305, "dur": 8,
    "args": {
      "External id": 38826,
      "cbid": 211, "correlation": 38826
    }
  },
  {
    "ph": "s", "id": 38826, "pid": 76337, "tid": -914061504, "ts": 1716454218042305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218097924, "dur": 14,
    "args": {
      "External id": 38837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38837,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38837, "pid": 5, "tid": 7, "ts": 1716454218097924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042385, "dur": 14,
    "args": {
      "External id": 38837,
      "cbid": 211, "correlation": 38837
    }
  },
  {
    "ph": "s", "id": 38837, "pid": 76337, "tid": -914061504, "ts": 1716454218042385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042441, "dur": 0,
    "args": {
      "External id": 38847,
      "cbid": 317, "correlation": 38847
    }
  },
  {
    "ph": "f", "id": 38847, "pid": 76337, "tid": -914061504, "ts": 1716454218042441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042442, "dur": 0,
    "args": {
      "External id": 38848,
      "cbid": 203, "correlation": 38848
    }
  },
  {
    "ph": "f", "id": 38848, "pid": 76337, "tid": -914061504, "ts": 1716454218042442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042443, "dur": 0,
    "args": {
      "External id": 38849,
      "cbid": 205, "correlation": 38849
    }
  },
  {
    "ph": "f", "id": 38849, "pid": 76337, "tid": -914061504, "ts": 1716454218042443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097939, "dur": 8,
    "args": {
      "External id": 38853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38853,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38853, "pid": 5, "tid": 7, "ts": 1716454218097939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042456, "dur": 11,
    "args": {
      "External id": 38853,
      "cbid": 211, "correlation": 38853
    }
  },
  {
    "ph": "s", "id": 38853, "pid": 76337, "tid": -914061504, "ts": 1716454218042456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218097948, "dur": 158,
    "args": {
      "External id": 38855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38855,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38855, "pid": 5, "tid": 7, "ts": 1716454218097948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042470, "dur": 5,
    "args": {
      "External id": 38855,
      "cbid": 211, "correlation": 38855
    }
  },
  {
    "ph": "s", "id": 38855, "pid": 76337, "tid": -914061504, "ts": 1716454218042470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218098109, "dur": 1,
    "args": {
      "External id": 38857,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38857,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 38857, "pid": 5, "tid": 7, "ts": 1716454218098109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042480, "dur": 6,
    "args": {
      "External id": 38857,
      "cbid": 51, "correlation": 38857
    }
  },
  {
    "ph": "s", "id": 38857, "pid": 76337, "tid": -914061504, "ts": 1716454218042480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218098112, "dur": 639,
    "args": {
      "External id": 38858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38858,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38858, "pid": 5, "tid": 7, "ts": 1716454218098112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042488, "dur": 6,
    "args": {
      "External id": 38858,
      "cbid": 211, "correlation": 38858
    }
  },
  {
    "ph": "s", "id": 38858, "pid": 76337, "tid": -914061504, "ts": 1716454218042488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218098752, "dur": 13,
    "args": {
      "External id": 38860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38860,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38860, "pid": 5, "tid": 7, "ts": 1716454218098752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042497, "dur": 5,
    "args": {
      "External id": 38860,
      "cbid": 211, "correlation": 38860
    }
  },
  {
    "ph": "s", "id": 38860, "pid": 76337, "tid": -914061504, "ts": 1716454218042497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218098766, "dur": 15,
    "args": {
      "External id": 38866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38866, "pid": 5, "tid": 7, "ts": 1716454218098766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042526, "dur": 8,
    "args": {
      "External id": 38866,
      "cbid": 211, "correlation": 38866
    }
  },
  {
    "ph": "s", "id": 38866, "pid": 76337, "tid": -914061504, "ts": 1716454218042526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218098782, "dur": 30,
    "args": {
      "External id": 38875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38875,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38875, "pid": 5, "tid": 7, "ts": 1716454218098782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042615, "dur": 13,
    "args": {
      "External id": 38875,
      "cbid": 211, "correlation": 38875
    }
  },
  {
    "ph": "s", "id": 38875, "pid": 76337, "tid": -914061504, "ts": 1716454218042615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218098813, "dur": 30,
    "args": {
      "External id": 38895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38895,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 38895, "pid": 5, "tid": 7, "ts": 1716454218098813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042683, "dur": 11,
    "args": {
      "External id": 38895,
      "cbid": 211, "correlation": 38895
    }
  },
  {
    "ph": "s", "id": 38895, "pid": 76337, "tid": -914061504, "ts": 1716454218042683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218098844, "dur": 4,
    "args": {
      "External id": 38907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38907,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38907, "pid": 5, "tid": 7, "ts": 1716454218098844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042703, "dur": 6,
    "args": {
      "External id": 38907,
      "cbid": 211, "correlation": 38907
    }
  },
  {
    "ph": "s", "id": 38907, "pid": 76337, "tid": -914061504, "ts": 1716454218042703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218098850, "dur": 29,
    "args": {
      "External id": 38910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38910,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38910, "pid": 5, "tid": 7, "ts": 1716454218098850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042722, "dur": 7,
    "args": {
      "External id": 38910,
      "cbid": 211, "correlation": 38910
    }
  },
  {
    "ph": "s", "id": 38910, "pid": 76337, "tid": -914061504, "ts": 1716454218042722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218098881, "dur": 20,
    "args": {
      "External id": 38919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38919,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38919, "pid": 5, "tid": 7, "ts": 1716454218098881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042760, "dur": 10,
    "args": {
      "External id": 38919,
      "cbid": 211, "correlation": 38919
    }
  },
  {
    "ph": "s", "id": 38919, "pid": 76337, "tid": -914061504, "ts": 1716454218042760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042811, "dur": 0,
    "args": {
      "External id": 38929,
      "cbid": 317, "correlation": 38929
    }
  },
  {
    "ph": "f", "id": 38929, "pid": 76337, "tid": -914061504, "ts": 1716454218042811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042812, "dur": 0,
    "args": {
      "External id": 38930,
      "cbid": 203, "correlation": 38930
    }
  },
  {
    "ph": "f", "id": 38930, "pid": 76337, "tid": -914061504, "ts": 1716454218042812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042813, "dur": 0,
    "args": {
      "External id": 38931,
      "cbid": 205, "correlation": 38931
    }
  },
  {
    "ph": "f", "id": 38931, "pid": 76337, "tid": -914061504, "ts": 1716454218042813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218098902, "dur": 21,
    "args": {
      "External id": 38935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38935,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38935, "pid": 5, "tid": 7, "ts": 1716454218098902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042828, "dur": 12,
    "args": {
      "External id": 38935,
      "cbid": 211, "correlation": 38935
    }
  },
  {
    "ph": "s", "id": 38935, "pid": 76337, "tid": -914061504, "ts": 1716454218042828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218098925, "dur": 312,
    "args": {
      "External id": 38937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38937,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38937, "pid": 5, "tid": 7, "ts": 1716454218098925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042842, "dur": 5,
    "args": {
      "External id": 38937,
      "cbid": 211, "correlation": 38937
    }
  },
  {
    "ph": "s", "id": 38937, "pid": 76337, "tid": -914061504, "ts": 1716454218042842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218099239, "dur": 1,
    "args": {
      "External id": 38939,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 38939,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5
    }
  },
  {
    "ph": "f", "id": 38939, "pid": 5, "tid": 7, "ts": 1716454218099239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042853, "dur": 6,
    "args": {
      "External id": 38939,
      "cbid": 51, "correlation": 38939
    }
  },
  {
    "ph": "s", "id": 38939, "pid": 76337, "tid": -914061504, "ts": 1716454218042853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218099243, "dur": 1220,
    "args": {
      "External id": 38940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38940,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38940, "pid": 5, "tid": 7, "ts": 1716454218099243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042860, "dur": 6,
    "args": {
      "External id": 38940,
      "cbid": 211, "correlation": 38940
    }
  },
  {
    "ph": "s", "id": 38940, "pid": 76337, "tid": -914061504, "ts": 1716454218042860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218100464, "dur": 12,
    "args": {
      "External id": 38942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38942,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38942, "pid": 5, "tid": 7, "ts": 1716454218100464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042870, "dur": 5,
    "args": {
      "External id": 38942,
      "cbid": 211, "correlation": 38942
    }
  },
  {
    "ph": "s", "id": 38942, "pid": 76337, "tid": -914061504, "ts": 1716454218042870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218100478, "dur": 15,
    "args": {
      "External id": 38948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38948, "pid": 5, "tid": 7, "ts": 1716454218100478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042898, "dur": 8,
    "args": {
      "External id": 38948,
      "cbid": 211, "correlation": 38948
    }
  },
  {
    "ph": "s", "id": 38948, "pid": 76337, "tid": -914061504, "ts": 1716454218042898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218100494, "dur": 3,
    "args": {
      "External id": 38956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38956,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 38956, "pid": 5, "tid": 7, "ts": 1716454218100494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218042943, "dur": 9,
    "args": {
      "External id": 38956,
      "cbid": 211, "correlation": 38956
    }
  },
  {
    "ph": "s", "id": 38956, "pid": 76337, "tid": -914061504, "ts": 1716454218042943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043014, "dur": 1,
    "args": {
      "External id": 38972,
      "cbid": 251, "correlation": 38972
    }
  },
  {
    "ph": "f", "id": 38972, "pid": 76337, "tid": -914061504, "ts": 1716454218043014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043020, "dur": 0,
    "args": {
      "External id": 38974,
      "cbid": 251, "correlation": 38974
    }
  },
  {
    "ph": "f", "id": 38974, "pid": 76337, "tid": -914061504, "ts": 1716454218043020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218100499, "dur": 13,
    "args": {
      "External id": 38975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38975,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38975, "pid": 5, "tid": 7, "ts": 1716454218100499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043022, "dur": 12,
    "args": {
      "External id": 38975,
      "cbid": 211, "correlation": 38975
    }
  },
  {
    "ph": "s", "id": 38975, "pid": 76337, "tid": -914061504, "ts": 1716454218043022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218100513, "dur": 5,
    "args": {
      "External id": 38977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38977,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 38977, "pid": 5, "tid": 7, "ts": 1716454218100513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043036, "dur": 6,
    "args": {
      "External id": 38977,
      "cbid": 211, "correlation": 38977
    }
  },
  {
    "ph": "s", "id": 38977, "pid": 76337, "tid": -914061504, "ts": 1716454218043036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218100519, "dur": 17,
    "args": {
      "External id": 38987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 38987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 38987, "pid": 5, "tid": 7, "ts": 1716454218100519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043094, "dur": 12,
    "args": {
      "External id": 38987,
      "cbid": 211, "correlation": 38987
    }
  },
  {
    "ph": "s", "id": 38987, "pid": 76337, "tid": -914061504, "ts": 1716454218043094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218100537, "dur": 17,
    "args": {
      "External id": 39007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39007,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 39007, "pid": 5, "tid": 7, "ts": 1716454218100537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043160, "dur": 11,
    "args": {
      "External id": 39007,
      "cbid": 211, "correlation": 39007
    }
  },
  {
    "ph": "s", "id": 39007, "pid": 76337, "tid": -914061504, "ts": 1716454218043160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218100556, "dur": 4,
    "args": {
      "External id": 39019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39019,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 39019, "pid": 5, "tid": 7, "ts": 1716454218100556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043180, "dur": 6,
    "args": {
      "External id": 39019,
      "cbid": 211, "correlation": 39019
    }
  },
  {
    "ph": "s", "id": 39019, "pid": 76337, "tid": -914061504, "ts": 1716454218043180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218100561, "dur": 16,
    "args": {
      "External id": 39022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39022,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39022, "pid": 5, "tid": 7, "ts": 1716454218100561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043198, "dur": 6,
    "args": {
      "External id": 39022,
      "cbid": 211, "correlation": 39022
    }
  },
  {
    "ph": "s", "id": 39022, "pid": 76337, "tid": -914061504, "ts": 1716454218043198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218100578, "dur": 11,
    "args": {
      "External id": 39031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39031,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39031, "pid": 5, "tid": 7, "ts": 1716454218100578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043239, "dur": 10,
    "args": {
      "External id": 39031,
      "cbid": 211, "correlation": 39031
    }
  },
  {
    "ph": "s", "id": 39031, "pid": 76337, "tid": -914061504, "ts": 1716454218043239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043301, "dur": 0,
    "args": {
      "External id": 39041,
      "cbid": 317, "correlation": 39041
    }
  },
  {
    "ph": "f", "id": 39041, "pid": 76337, "tid": -914061504, "ts": 1716454218043301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043302, "dur": 0,
    "args": {
      "External id": 39042,
      "cbid": 203, "correlation": 39042
    }
  },
  {
    "ph": "f", "id": 39042, "pid": 76337, "tid": -914061504, "ts": 1716454218043302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043303, "dur": 0,
    "args": {
      "External id": 39043,
      "cbid": 205, "correlation": 39043
    }
  },
  {
    "ph": "f", "id": 39043, "pid": 76337, "tid": -914061504, "ts": 1716454218043303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218100590, "dur": 11,
    "args": {
      "External id": 39047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39047,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39047, "pid": 5, "tid": 7, "ts": 1716454218100590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043317, "dur": 12,
    "args": {
      "External id": 39047,
      "cbid": 211, "correlation": 39047
    }
  },
  {
    "ph": "s", "id": 39047, "pid": 76337, "tid": -914061504, "ts": 1716454218043317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218100602, "dur": 159,
    "args": {
      "External id": 39049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39049,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39049, "pid": 5, "tid": 7, "ts": 1716454218100602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043331, "dur": 5,
    "args": {
      "External id": 39049,
      "cbid": 211, "correlation": 39049
    }
  },
  {
    "ph": "s", "id": 39049, "pid": 76337, "tid": -914061504, "ts": 1716454218043331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218100763, "dur": 1,
    "args": {
      "External id": 39051,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 39051,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 39051, "pid": 5, "tid": 7, "ts": 1716454218100763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043342, "dur": 6,
    "args": {
      "External id": 39051,
      "cbid": 51, "correlation": 39051
    }
  },
  {
    "ph": "s", "id": 39051, "pid": 76337, "tid": -914061504, "ts": 1716454218043342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218100767, "dur": 637,
    "args": {
      "External id": 39052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39052,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39052, "pid": 5, "tid": 7, "ts": 1716454218100767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043349, "dur": 7,
    "args": {
      "External id": 39052,
      "cbid": 211, "correlation": 39052
    }
  },
  {
    "ph": "s", "id": 39052, "pid": 76337, "tid": -914061504, "ts": 1716454218043349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218101405, "dur": 12,
    "args": {
      "External id": 39054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39054, "pid": 5, "tid": 7, "ts": 1716454218101405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043360, "dur": 5,
    "args": {
      "External id": 39054,
      "cbid": 211, "correlation": 39054
    }
  },
  {
    "ph": "s", "id": 39054, "pid": 76337, "tid": -914061504, "ts": 1716454218043360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218101418, "dur": 14,
    "args": {
      "External id": 39060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39060, "pid": 5, "tid": 7, "ts": 1716454218101418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043388, "dur": 9,
    "args": {
      "External id": 39060,
      "cbid": 211, "correlation": 39060
    }
  },
  {
    "ph": "s", "id": 39060, "pid": 76337, "tid": -914061504, "ts": 1716454218043388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043446, "dur": 0,
    "args": {
      "External id": 39070,
      "cbid": 317, "correlation": 39070
    }
  },
  {
    "ph": "f", "id": 39070, "pid": 76337, "tid": -914061504, "ts": 1716454218043446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043447, "dur": 0,
    "args": {
      "External id": 39071,
      "cbid": 203, "correlation": 39071
    }
  },
  {
    "ph": "f", "id": 39071, "pid": 76337, "tid": -914061504, "ts": 1716454218043447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043448, "dur": 0,
    "args": {
      "External id": 39072,
      "cbid": 205, "correlation": 39072
    }
  },
  {
    "ph": "f", "id": 39072, "pid": 76337, "tid": -914061504, "ts": 1716454218043448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218101434, "dur": 21,
    "args": {
      "External id": 39076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39076, "pid": 5, "tid": 7, "ts": 1716454218101434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043461, "dur": 12,
    "args": {
      "External id": 39076,
      "cbid": 211, "correlation": 39076
    }
  },
  {
    "ph": "s", "id": 39076, "pid": 76337, "tid": -914061504, "ts": 1716454218043461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218101456, "dur": 4,
    "args": {
      "External id": 39078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39078,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39078, "pid": 5, "tid": 7, "ts": 1716454218101456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043479, "dur": 6,
    "args": {
      "External id": 39078,
      "cbid": 211, "correlation": 39078
    }
  },
  {
    "ph": "s", "id": 39078, "pid": 76337, "tid": -914061504, "ts": 1716454218043479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043488, "dur": 0,
    "args": {
      "External id": 39079,
      "cbid": 51, "correlation": 39079
    }
  },
  {
    "ph": "s", "id": 39079, "pid": 76337, "tid": -914061504, "ts": 1716454218043488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218101461, "dur": 171,
    "args": {
      "External id": 39080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39080,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 39080, "pid": 5, "tid": 7, "ts": 1716454218101461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043489, "dur": 6,
    "args": {
      "External id": 39080,
      "cbid": 211, "correlation": 39080
    }
  },
  {
    "ph": "s", "id": 39080, "pid": 76337, "tid": -914061504, "ts": 1716454218043489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218101633, "dur": 15,
    "args": {
      "External id": 39085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39085, "pid": 5, "tid": 7, "ts": 1716454218101633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043516, "dur": 8,
    "args": {
      "External id": 39085,
      "cbid": 211, "correlation": 39085
    }
  },
  {
    "ph": "s", "id": 39085, "pid": 76337, "tid": -914061504, "ts": 1716454218043516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218101650, "dur": 12,
    "args": {
      "External id": 39093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39093,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39093, "pid": 5, "tid": 7, "ts": 1716454218101650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043544, "dur": 8,
    "args": {
      "External id": 39093,
      "cbid": 211, "correlation": 39093
    }
  },
  {
    "ph": "s", "id": 39093, "pid": 76337, "tid": -914061504, "ts": 1716454218043544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218101663, "dur": 10,
    "args": {
      "External id": 39101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39101,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39101, "pid": 5, "tid": 7, "ts": 1716454218101663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043573, "dur": 8,
    "args": {
      "External id": 39101,
      "cbid": 211, "correlation": 39101
    }
  },
  {
    "ph": "s", "id": 39101, "pid": 76337, "tid": -914061504, "ts": 1716454218043573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218101674, "dur": 18,
    "args": {
      "External id": 39121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39121,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 39121, "pid": 5, "tid": 7, "ts": 1716454218101674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043655, "dur": 13,
    "args": {
      "External id": 39121,
      "cbid": 211, "correlation": 39121
    }
  },
  {
    "ph": "s", "id": 39121, "pid": 76337, "tid": -914061504, "ts": 1716454218043655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218101693, "dur": 4,
    "args": {
      "External id": 39133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39133,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 39133, "pid": 5, "tid": 7, "ts": 1716454218101693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043678, "dur": 6,
    "args": {
      "External id": 39133,
      "cbid": 211, "correlation": 39133
    }
  },
  {
    "ph": "s", "id": 39133, "pid": 76337, "tid": -914061504, "ts": 1716454218043678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218101699, "dur": 16,
    "args": {
      "External id": 39136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39136, "pid": 5, "tid": 7, "ts": 1716454218101699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043696, "dur": 6,
    "args": {
      "External id": 39136,
      "cbid": 211, "correlation": 39136
    }
  },
  {
    "ph": "s", "id": 39136, "pid": 76337, "tid": -914061504, "ts": 1716454218043696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043752, "dur": 0,
    "args": {
      "External id": 39147,
      "cbid": 317, "correlation": 39147
    }
  },
  {
    "ph": "f", "id": 39147, "pid": 76337, "tid": -914061504, "ts": 1716454218043752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043753, "dur": 0,
    "args": {
      "External id": 39148,
      "cbid": 203, "correlation": 39148
    }
  },
  {
    "ph": "f", "id": 39148, "pid": 76337, "tid": -914061504, "ts": 1716454218043753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043754, "dur": 0,
    "args": {
      "External id": 39149,
      "cbid": 205, "correlation": 39149
    }
  },
  {
    "ph": "f", "id": 39149, "pid": 76337, "tid": -914061504, "ts": 1716454218043754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218101716, "dur": 11,
    "args": {
      "External id": 39153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39153,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39153, "pid": 5, "tid": 7, "ts": 1716454218101716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043768, "dur": 12,
    "args": {
      "External id": 39153,
      "cbid": 211, "correlation": 39153
    }
  },
  {
    "ph": "s", "id": 39153, "pid": 76337, "tid": -914061504, "ts": 1716454218043768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218101729, "dur": 3,
    "args": {
      "External id": 39155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39155, "pid": 5, "tid": 7, "ts": 1716454218101729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043784, "dur": 6,
    "args": {
      "External id": 39155,
      "cbid": 211, "correlation": 39155
    }
  },
  {
    "ph": "s", "id": 39155, "pid": 76337, "tid": -914061504, "ts": 1716454218043784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043794, "dur": 0,
    "args": {
      "External id": 39156,
      "cbid": 51, "correlation": 39156
    }
  },
  {
    "ph": "s", "id": 39156, "pid": 76337, "tid": -914061504, "ts": 1716454218043794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218101733, "dur": 88,
    "args": {
      "External id": 39157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39157,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 39157, "pid": 5, "tid": 7, "ts": 1716454218101733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043795, "dur": 5,
    "args": {
      "External id": 39157,
      "cbid": 211, "correlation": 39157
    }
  },
  {
    "ph": "s", "id": 39157, "pid": 76337, "tid": -914061504, "ts": 1716454218043795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218101822, "dur": 15,
    "args": {
      "External id": 39162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39162, "pid": 5, "tid": 7, "ts": 1716454218101822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043821, "dur": 8,
    "args": {
      "External id": 39162,
      "cbid": 211, "correlation": 39162
    }
  },
  {
    "ph": "s", "id": 39162, "pid": 76337, "tid": -914061504, "ts": 1716454218043821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218101839, "dur": 81,
    "args": {
      "External id": 39171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39171,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39171, "pid": 5, "tid": 7, "ts": 1716454218101839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043904, "dur": 14,
    "args": {
      "External id": 39171,
      "cbid": 211, "correlation": 39171
    }
  },
  {
    "ph": "s", "id": 39171, "pid": 76337, "tid": -914061504, "ts": 1716454218043904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218101921, "dur": 30,
    "args": {
      "External id": 39193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39193,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39193, "pid": 5, "tid": 7, "ts": 1716454218101921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218043962, "dur": 10,
    "args": {
      "External id": 39193,
      "cbid": 211, "correlation": 39193
    }
  },
  {
    "ph": "s", "id": 39193, "pid": 76337, "tid": -914061504, "ts": 1716454218043962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044064, "dur": 2,
    "args": {
      "External id": 39204,
      "cbid": 251, "correlation": 39204
    }
  },
  {
    "ph": "f", "id": 39204, "pid": 76337, "tid": -914061504, "ts": 1716454218044064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218101952, "dur": 140,
    "args": {
      "External id": 39205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39205,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39205, "pid": 5, "tid": 7, "ts": 1716454218101952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044070, "dur": 14,
    "args": {
      "External id": 39205,
      "cbid": 211, "correlation": 39205
    }
  },
  {
    "ph": "s", "id": 39205, "pid": 76337, "tid": -914061504, "ts": 1716454218044070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044141, "dur": 1,
    "args": {
      "External id": 39216,
      "cbid": 251, "correlation": 39216
    }
  },
  {
    "ph": "f", "id": 39216, "pid": 76337, "tid": -914061504, "ts": 1716454218044141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218102094, "dur": 153,
    "args": {
      "External id": 39217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39217,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39217, "pid": 5, "tid": 7, "ts": 1716454218102094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044145, "dur": 12,
    "args": {
      "External id": 39217,
      "cbid": 211, "correlation": 39217
    }
  },
  {
    "ph": "s", "id": 39217, "pid": 76337, "tid": -914061504, "ts": 1716454218044145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044211, "dur": 1,
    "args": {
      "External id": 39228,
      "cbid": 251, "correlation": 39228
    }
  },
  {
    "ph": "f", "id": 39228, "pid": 76337, "tid": -914061504, "ts": 1716454218044211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218102248, "dur": 154,
    "args": {
      "External id": 39229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39229,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39229, "pid": 5, "tid": 7, "ts": 1716454218102248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044215, "dur": 11,
    "args": {
      "External id": 39229,
      "cbid": 211, "correlation": 39229
    }
  },
  {
    "ph": "s", "id": 39229, "pid": 76337, "tid": -914061504, "ts": 1716454218044215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218102404, "dur": 330,
    "args": {
      "External id": 39254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39254,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39254, "pid": 5, "tid": 7, "ts": 1716454218102404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044300, "dur": 13,
    "args": {
      "External id": 39254,
      "cbid": 211, "correlation": 39254
    }
  },
  {
    "ph": "s", "id": 39254, "pid": 76337, "tid": -914061504, "ts": 1716454218044300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044401, "dur": 1,
    "args": {
      "External id": 39272,
      "cbid": 251, "correlation": 39272
    }
  },
  {
    "ph": "f", "id": 39272, "pid": 76337, "tid": -914061504, "ts": 1716454218044401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218102736, "dur": 144,
    "args": {
      "External id": 39274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39274,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39274, "pid": 5, "tid": 7, "ts": 1716454218102736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044407, "dur": 13,
    "args": {
      "External id": 39274,
      "cbid": 211, "correlation": 39274
    }
  },
  {
    "ph": "s", "id": 39274, "pid": 76337, "tid": -914061504, "ts": 1716454218044407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218102881, "dur": 19,
    "args": {
      "External id": 39282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39282,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39282, "pid": 5, "tid": 7, "ts": 1716454218102881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044477, "dur": 12,
    "args": {
      "External id": 39282,
      "cbid": 211, "correlation": 39282
    }
  },
  {
    "ph": "s", "id": 39282, "pid": 76337, "tid": -914061504, "ts": 1716454218044477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218102902, "dur": 28,
    "args": {
      "External id": 39290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39290,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39290, "pid": 5, "tid": 7, "ts": 1716454218102902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044515, "dur": 9,
    "args": {
      "External id": 39290,
      "cbid": 211, "correlation": 39290
    }
  },
  {
    "ph": "s", "id": 39290, "pid": 76337, "tid": -914061504, "ts": 1716454218044515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218102932, "dur": 18,
    "args": {
      "External id": 39301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39301, "pid": 5, "tid": 7, "ts": 1716454218102932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044591, "dur": 12,
    "args": {
      "External id": 39301,
      "cbid": 211, "correlation": 39301
    }
  },
  {
    "ph": "s", "id": 39301, "pid": 76337, "tid": -914061504, "ts": 1716454218044591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218102951, "dur": 16,
    "args": {
      "External id": 39323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39323,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39323, "pid": 5, "tid": 7, "ts": 1716454218102951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044621, "dur": 8,
    "args": {
      "External id": 39323,
      "cbid": 211, "correlation": 39323
    }
  },
  {
    "ph": "s", "id": 39323, "pid": 76337, "tid": -914061504, "ts": 1716454218044621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044707, "dur": 1,
    "args": {
      "External id": 39334,
      "cbid": 251, "correlation": 39334
    }
  },
  {
    "ph": "f", "id": 39334, "pid": 76337, "tid": -914061504, "ts": 1716454218044707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218102968, "dur": 88,
    "args": {
      "External id": 39335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39335,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39335, "pid": 5, "tid": 7, "ts": 1716454218102968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044712, "dur": 13,
    "args": {
      "External id": 39335,
      "cbid": 211, "correlation": 39335
    }
  },
  {
    "ph": "s", "id": 39335, "pid": 76337, "tid": -914061504, "ts": 1716454218044712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044783, "dur": 1,
    "args": {
      "External id": 39346,
      "cbid": 251, "correlation": 39346
    }
  },
  {
    "ph": "f", "id": 39346, "pid": 76337, "tid": -914061504, "ts": 1716454218044783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044787, "dur": 0,
    "args": {
      "External id": 39347,
      "cbid": 251, "correlation": 39347
    }
  },
  {
    "ph": "f", "id": 39347, "pid": 76337, "tid": -914061504, "ts": 1716454218044787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218103057, "dur": 12,
    "args": {
      "External id": 39348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39348,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39348, "pid": 5, "tid": 7, "ts": 1716454218103057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044789, "dur": 12,
    "args": {
      "External id": 39348,
      "cbid": 211, "correlation": 39348
    }
  },
  {
    "ph": "s", "id": 39348, "pid": 76337, "tid": -914061504, "ts": 1716454218044789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218103070, "dur": 5,
    "args": {
      "External id": 39350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39350,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39350, "pid": 5, "tid": 7, "ts": 1716454218103070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044803, "dur": 6,
    "args": {
      "External id": 39350,
      "cbid": 211, "correlation": 39350
    }
  },
  {
    "ph": "s", "id": 39350, "pid": 76337, "tid": -914061504, "ts": 1716454218044803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044860, "dur": 1,
    "args": {
      "External id": 39361,
      "cbid": 251, "correlation": 39361
    }
  },
  {
    "ph": "f", "id": 39361, "pid": 76337, "tid": -914061504, "ts": 1716454218044860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044863, "dur": 0,
    "args": {
      "External id": 39362,
      "cbid": 251, "correlation": 39362
    }
  },
  {
    "ph": "f", "id": 39362, "pid": 76337, "tid": -914061504, "ts": 1716454218044863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218103077, "dur": 8,
    "args": {
      "External id": 39363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39363,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39363, "pid": 5, "tid": 7, "ts": 1716454218103077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044865, "dur": 12,
    "args": {
      "External id": 39363,
      "cbid": 211, "correlation": 39363
    }
  },
  {
    "ph": "s", "id": 39363, "pid": 76337, "tid": -914061504, "ts": 1716454218044865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218103087, "dur": 3,
    "args": {
      "External id": 39365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39365,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39365, "pid": 5, "tid": 7, "ts": 1716454218103087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044878, "dur": 5,
    "args": {
      "External id": 39365,
      "cbid": 211, "correlation": 39365
    }
  },
  {
    "ph": "s", "id": 39365, "pid": 76337, "tid": -914061504, "ts": 1716454218044878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218103091, "dur": 54,
    "args": {
      "External id": 39390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39390,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39390, "pid": 5, "tid": 7, "ts": 1716454218103091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218044955, "dur": 12,
    "args": {
      "External id": 39390,
      "cbid": 211, "correlation": 39390
    }
  },
  {
    "ph": "s", "id": 39390, "pid": 76337, "tid": -914061504, "ts": 1716454218044955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045062, "dur": 2,
    "args": {
      "External id": 39408,
      "cbid": 251, "correlation": 39408
    }
  },
  {
    "ph": "f", "id": 39408, "pid": 76337, "tid": -914061504, "ts": 1716454218045062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218103146, "dur": 89,
    "args": {
      "External id": 39410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39410,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39410, "pid": 5, "tid": 7, "ts": 1716454218103146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045068, "dur": 14,
    "args": {
      "External id": 39410,
      "cbid": 211, "correlation": 39410
    }
  },
  {
    "ph": "s", "id": 39410, "pid": 76337, "tid": -914061504, "ts": 1716454218045068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218103236, "dur": 9,
    "args": {
      "External id": 39418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39418,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39418, "pid": 5, "tid": 7, "ts": 1716454218103236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045138, "dur": 12,
    "args": {
      "External id": 39418,
      "cbid": 211, "correlation": 39418
    }
  },
  {
    "ph": "s", "id": 39418, "pid": 76337, "tid": -914061504, "ts": 1716454218045138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218103247, "dur": 21,
    "args": {
      "External id": 39426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39426, "pid": 5, "tid": 7, "ts": 1716454218103247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045181, "dur": 9,
    "args": {
      "External id": 39426,
      "cbid": 211, "correlation": 39426
    }
  },
  {
    "ph": "s", "id": 39426, "pid": 76337, "tid": -914061504, "ts": 1716454218045181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218103270, "dur": 17,
    "args": {
      "External id": 39448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39448,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39448, "pid": 5, "tid": 7, "ts": 1716454218103270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045232, "dur": 10,
    "args": {
      "External id": 39448,
      "cbid": 211, "correlation": 39448
    }
  },
  {
    "ph": "s", "id": 39448, "pid": 76337, "tid": -914061504, "ts": 1716454218045232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045320, "dur": 1,
    "args": {
      "External id": 39464,
      "cbid": 251, "correlation": 39464
    }
  },
  {
    "ph": "f", "id": 39464, "pid": 76337, "tid": -914061504, "ts": 1716454218045320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045325, "dur": 0,
    "args": {
      "External id": 39466,
      "cbid": 251, "correlation": 39466
    }
  },
  {
    "ph": "f", "id": 39466, "pid": 76337, "tid": -914061504, "ts": 1716454218045325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218103288, "dur": 491,
    "args": {
      "External id": 39467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39467,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39467, "pid": 5, "tid": 7, "ts": 1716454218103288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045327, "dur": 13,
    "args": {
      "External id": 39467,
      "cbid": 211, "correlation": 39467
    }
  },
  {
    "ph": "s", "id": 39467, "pid": 76337, "tid": -914061504, "ts": 1716454218045327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218103781, "dur": 64,
    "args": {
      "External id": 39475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39475, "pid": 5, "tid": 7, "ts": 1716454218103781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045394, "dur": 12,
    "args": {
      "External id": 39475,
      "cbid": 211, "correlation": 39475
    }
  },
  {
    "ph": "s", "id": 39475, "pid": 76337, "tid": -914061504, "ts": 1716454218045394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218103846, "dur": 69,
    "args": {
      "External id": 39483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39483, "pid": 5, "tid": 7, "ts": 1716454218103846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045425, "dur": 8,
    "args": {
      "External id": 39483,
      "cbid": 211, "correlation": 39483
    }
  },
  {
    "ph": "s", "id": 39483, "pid": 76337, "tid": -914061504, "ts": 1716454218045425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045506, "dur": 1,
    "args": {
      "External id": 39499,
      "cbid": 251, "correlation": 39499
    }
  },
  {
    "ph": "f", "id": 39499, "pid": 76337, "tid": -914061504, "ts": 1716454218045506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218103917, "dur": 1,
    "args": {
      "External id": 39501,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 39501,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 39501, "pid": 5, "tid": 7, "ts": 1716454218103917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045511, "dur": 9,
    "args": {
      "External id": 39501,
      "cbid": 51, "correlation": 39501
    }
  },
  {
    "ph": "s", "id": 39501, "pid": 76337, "tid": -914061504, "ts": 1716454218045511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218103921, "dur": 264,
    "args": {
      "External id": 39502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39502,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39502, "pid": 5, "tid": 7, "ts": 1716454218103921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045522, "dur": 11,
    "args": {
      "External id": 39502,
      "cbid": 211, "correlation": 39502
    }
  },
  {
    "ph": "s", "id": 39502, "pid": 76337, "tid": -914061504, "ts": 1716454218045522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218104186, "dur": 14,
    "args": {
      "External id": 39510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39510,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39510, "pid": 5, "tid": 7, "ts": 1716454218104186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045565, "dur": 10,
    "args": {
      "External id": 39510,
      "cbid": 211, "correlation": 39510
    }
  },
  {
    "ph": "s", "id": 39510, "pid": 76337, "tid": -914061504, "ts": 1716454218045565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218104201, "dur": 37,
    "args": {
      "External id": 39521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39521, "pid": 5, "tid": 7, "ts": 1716454218104201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045633, "dur": 12,
    "args": {
      "External id": 39521,
      "cbid": 211, "correlation": 39521
    }
  },
  {
    "ph": "s", "id": 39521, "pid": 76337, "tid": -914061504, "ts": 1716454218045633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045696, "dur": 0,
    "args": {
      "External id": 39533,
      "cbid": 317, "correlation": 39533
    }
  },
  {
    "ph": "f", "id": 39533, "pid": 76337, "tid": -914061504, "ts": 1716454218045696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045697, "dur": 0,
    "args": {
      "External id": 39534,
      "cbid": 203, "correlation": 39534
    }
  },
  {
    "ph": "f", "id": 39534, "pid": 76337, "tid": -914061504, "ts": 1716454218045697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045698, "dur": 0,
    "args": {
      "External id": 39535,
      "cbid": 205, "correlation": 39535
    }
  },
  {
    "ph": "f", "id": 39535, "pid": 76337, "tid": -914061504, "ts": 1716454218045698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218104239, "dur": 14,
    "args": {
      "External id": 39539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39539,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39539, "pid": 5, "tid": 7, "ts": 1716454218104239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045712, "dur": 12,
    "args": {
      "External id": 39539,
      "cbid": 211, "correlation": 39539
    }
  },
  {
    "ph": "s", "id": 39539, "pid": 76337, "tid": -914061504, "ts": 1716454218045712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218104254, "dur": 4,
    "args": {
      "External id": 39541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39541, "pid": 5, "tid": 7, "ts": 1716454218104254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045729, "dur": 6,
    "args": {
      "External id": 39541,
      "cbid": 211, "correlation": 39541
    }
  },
  {
    "ph": "s", "id": 39541, "pid": 76337, "tid": -914061504, "ts": 1716454218045729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045739, "dur": 0,
    "args": {
      "External id": 39542,
      "cbid": 51, "correlation": 39542
    }
  },
  {
    "ph": "s", "id": 39542, "pid": 76337, "tid": -914061504, "ts": 1716454218045739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218104259, "dur": 93,
    "args": {
      "External id": 39543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39543,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 39543, "pid": 5, "tid": 7, "ts": 1716454218104259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045739, "dur": 5,
    "args": {
      "External id": 39543,
      "cbid": 211, "correlation": 39543
    }
  },
  {
    "ph": "s", "id": 39543, "pid": 76337, "tid": -914061504, "ts": 1716454218045739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218104354, "dur": 16,
    "args": {
      "External id": 39548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39548, "pid": 5, "tid": 7, "ts": 1716454218104354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045766, "dur": 8,
    "args": {
      "External id": 39548,
      "cbid": 211, "correlation": 39548
    }
  },
  {
    "ph": "s", "id": 39548, "pid": 76337, "tid": -914061504, "ts": 1716454218045766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218104371, "dur": 11,
    "args": {
      "External id": 39556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39556,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39556, "pid": 5, "tid": 7, "ts": 1716454218104371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045798, "dur": 8,
    "args": {
      "External id": 39556,
      "cbid": 211, "correlation": 39556
    }
  },
  {
    "ph": "s", "id": 39556, "pid": 76337, "tid": -914061504, "ts": 1716454218045798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218104384, "dur": 29,
    "args": {
      "External id": 39565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39565,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39565, "pid": 5, "tid": 7, "ts": 1716454218104384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045837, "dur": 10,
    "args": {
      "External id": 39565,
      "cbid": 211, "correlation": 39565
    }
  },
  {
    "ph": "s", "id": 39565, "pid": 76337, "tid": -914061504, "ts": 1716454218045837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218104414, "dur": 30,
    "args": {
      "External id": 39585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39585,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 39585, "pid": 5, "tid": 7, "ts": 1716454218104414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045908, "dur": 12,
    "args": {
      "External id": 39585,
      "cbid": 211, "correlation": 39585
    }
  },
  {
    "ph": "s", "id": 39585, "pid": 76337, "tid": -914061504, "ts": 1716454218045908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218104446, "dur": 5,
    "args": {
      "External id": 39597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39597,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39597, "pid": 5, "tid": 7, "ts": 1716454218104446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045930, "dur": 6,
    "args": {
      "External id": 39597,
      "cbid": 211, "correlation": 39597
    }
  },
  {
    "ph": "s", "id": 39597, "pid": 76337, "tid": -914061504, "ts": 1716454218045930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218104452, "dur": 30,
    "args": {
      "External id": 39600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39600, "pid": 5, "tid": 7, "ts": 1716454218104452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045948, "dur": 6,
    "args": {
      "External id": 39600,
      "cbid": 211, "correlation": 39600
    }
  },
  {
    "ph": "s", "id": 39600, "pid": 76337, "tid": -914061504, "ts": 1716454218045948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218104484, "dur": 22,
    "args": {
      "External id": 39609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39609,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39609, "pid": 5, "tid": 7, "ts": 1716454218104484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218045995, "dur": 10,
    "args": {
      "External id": 39609,
      "cbid": 211, "correlation": 39609
    }
  },
  {
    "ph": "s", "id": 39609, "pid": 76337, "tid": -914061504, "ts": 1716454218045995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046048, "dur": 0,
    "args": {
      "External id": 39619,
      "cbid": 317, "correlation": 39619
    }
  },
  {
    "ph": "f", "id": 39619, "pid": 76337, "tid": -914061504, "ts": 1716454218046048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046049, "dur": 0,
    "args": {
      "External id": 39620,
      "cbid": 203, "correlation": 39620
    }
  },
  {
    "ph": "f", "id": 39620, "pid": 76337, "tid": -914061504, "ts": 1716454218046049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046050, "dur": 0,
    "args": {
      "External id": 39621,
      "cbid": 205, "correlation": 39621
    }
  },
  {
    "ph": "f", "id": 39621, "pid": 76337, "tid": -914061504, "ts": 1716454218046050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218104507, "dur": 22,
    "args": {
      "External id": 39625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39625, "pid": 5, "tid": 7, "ts": 1716454218104507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046063, "dur": 11,
    "args": {
      "External id": 39625,
      "cbid": 211, "correlation": 39625
    }
  },
  {
    "ph": "s", "id": 39625, "pid": 76337, "tid": -914061504, "ts": 1716454218046063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218104530, "dur": 313,
    "args": {
      "External id": 39627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39627, "pid": 5, "tid": 7, "ts": 1716454218104530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046077, "dur": 5,
    "args": {
      "External id": 39627,
      "cbid": 211, "correlation": 39627
    }
  },
  {
    "ph": "s", "id": 39627, "pid": 76337, "tid": -914061504, "ts": 1716454218046077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218104846, "dur": 1,
    "args": {
      "External id": 39629,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 39629,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 39629, "pid": 5, "tid": 7, "ts": 1716454218104846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046089, "dur": 7,
    "args": {
      "External id": 39629,
      "cbid": 51, "correlation": 39629
    }
  },
  {
    "ph": "s", "id": 39629, "pid": 76337, "tid": -914061504, "ts": 1716454218046089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218104849, "dur": 1232,
    "args": {
      "External id": 39630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39630,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39630, "pid": 5, "tid": 7, "ts": 1716454218104849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046096, "dur": 6,
    "args": {
      "External id": 39630,
      "cbid": 211, "correlation": 39630
    }
  },
  {
    "ph": "s", "id": 39630, "pid": 76337, "tid": -914061504, "ts": 1716454218046096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218106083, "dur": 13,
    "args": {
      "External id": 39632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39632, "pid": 5, "tid": 7, "ts": 1716454218106083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046107, "dur": 5,
    "args": {
      "External id": 39632,
      "cbid": 211, "correlation": 39632
    }
  },
  {
    "ph": "s", "id": 39632, "pid": 76337, "tid": -914061504, "ts": 1716454218046107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218106097, "dur": 15,
    "args": {
      "External id": 39638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39638, "pid": 5, "tid": 7, "ts": 1716454218106097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046134, "dur": 8,
    "args": {
      "External id": 39638,
      "cbid": 211, "correlation": 39638
    }
  },
  {
    "ph": "s", "id": 39638, "pid": 76337, "tid": -914061504, "ts": 1716454218046134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218106113, "dur": 3,
    "args": {
      "External id": 39646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39646,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 39646, "pid": 5, "tid": 7, "ts": 1716454218106113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046179, "dur": 9,
    "args": {
      "External id": 39646,
      "cbid": 211, "correlation": 39646
    }
  },
  {
    "ph": "s", "id": 39646, "pid": 76337, "tid": -914061504, "ts": 1716454218046179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046244, "dur": 1,
    "args": {
      "External id": 39662,
      "cbid": 251, "correlation": 39662
    }
  },
  {
    "ph": "f", "id": 39662, "pid": 76337, "tid": -914061504, "ts": 1716454218046244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046249, "dur": 0,
    "args": {
      "External id": 39664,
      "cbid": 251, "correlation": 39664
    }
  },
  {
    "ph": "f", "id": 39664, "pid": 76337, "tid": -914061504, "ts": 1716454218046249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218106118, "dur": 13,
    "args": {
      "External id": 39665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39665,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39665, "pid": 5, "tid": 7, "ts": 1716454218106118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046251, "dur": 11,
    "args": {
      "External id": 39665,
      "cbid": 211, "correlation": 39665
    }
  },
  {
    "ph": "s", "id": 39665, "pid": 76337, "tid": -914061504, "ts": 1716454218046251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218106133, "dur": 5,
    "args": {
      "External id": 39667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39667,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39667, "pid": 5, "tid": 7, "ts": 1716454218106133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046265, "dur": 5,
    "args": {
      "External id": 39667,
      "cbid": 211, "correlation": 39667
    }
  },
  {
    "ph": "s", "id": 39667, "pid": 76337, "tid": -914061504, "ts": 1716454218046265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218106139, "dur": 17,
    "args": {
      "External id": 39677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39677, "pid": 5, "tid": 7, "ts": 1716454218106139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046321, "dur": 12,
    "args": {
      "External id": 39677,
      "cbid": 211, "correlation": 39677
    }
  },
  {
    "ph": "s", "id": 39677, "pid": 76337, "tid": -914061504, "ts": 1716454218046321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218106157, "dur": 17,
    "args": {
      "External id": 39697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39697,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 39697, "pid": 5, "tid": 7, "ts": 1716454218106157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046388, "dur": 11,
    "args": {
      "External id": 39697,
      "cbid": 211, "correlation": 39697
    }
  },
  {
    "ph": "s", "id": 39697, "pid": 76337, "tid": -914061504, "ts": 1716454218046388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218106175, "dur": 4,
    "args": {
      "External id": 39709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39709,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 39709, "pid": 5, "tid": 7, "ts": 1716454218106175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046408, "dur": 6,
    "args": {
      "External id": 39709,
      "cbid": 211, "correlation": 39709
    }
  },
  {
    "ph": "s", "id": 39709, "pid": 76337, "tid": -914061504, "ts": 1716454218046408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218106181, "dur": 16,
    "args": {
      "External id": 39712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39712, "pid": 5, "tid": 7, "ts": 1716454218106181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046426, "dur": 6,
    "args": {
      "External id": 39712,
      "cbid": 211, "correlation": 39712
    }
  },
  {
    "ph": "s", "id": 39712, "pid": 76337, "tid": -914061504, "ts": 1716454218046426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218106198, "dur": 11,
    "args": {
      "External id": 39721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39721,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39721, "pid": 5, "tid": 7, "ts": 1716454218106198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046466, "dur": 10,
    "args": {
      "External id": 39721,
      "cbid": 211, "correlation": 39721
    }
  },
  {
    "ph": "s", "id": 39721, "pid": 76337, "tid": -914061504, "ts": 1716454218046466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046528, "dur": 0,
    "args": {
      "External id": 39731,
      "cbid": 317, "correlation": 39731
    }
  },
  {
    "ph": "f", "id": 39731, "pid": 76337, "tid": -914061504, "ts": 1716454218046528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046529, "dur": 0,
    "args": {
      "External id": 39732,
      "cbid": 203, "correlation": 39732
    }
  },
  {
    "ph": "f", "id": 39732, "pid": 76337, "tid": -914061504, "ts": 1716454218046529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046530, "dur": 0,
    "args": {
      "External id": 39733,
      "cbid": 205, "correlation": 39733
    }
  },
  {
    "ph": "f", "id": 39733, "pid": 76337, "tid": -914061504, "ts": 1716454218046530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218106210, "dur": 11,
    "args": {
      "External id": 39737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39737,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39737, "pid": 5, "tid": 7, "ts": 1716454218106210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046543, "dur": 12,
    "args": {
      "External id": 39737,
      "cbid": 211, "correlation": 39737
    }
  },
  {
    "ph": "s", "id": 39737, "pid": 76337, "tid": -914061504, "ts": 1716454218046543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218106222, "dur": 159,
    "args": {
      "External id": 39739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39739, "pid": 5, "tid": 7, "ts": 1716454218106222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046557, "dur": 5,
    "args": {
      "External id": 39739,
      "cbid": 211, "correlation": 39739
    }
  },
  {
    "ph": "s", "id": 39739, "pid": 76337, "tid": -914061504, "ts": 1716454218046557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218106384, "dur": 1,
    "args": {
      "External id": 39741,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 39741,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 39741, "pid": 5, "tid": 7, "ts": 1716454218106384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046569, "dur": 6,
    "args": {
      "External id": 39741,
      "cbid": 51, "correlation": 39741
    }
  },
  {
    "ph": "s", "id": 39741, "pid": 76337, "tid": -914061504, "ts": 1716454218046569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218106388, "dur": 636,
    "args": {
      "External id": 39742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39742,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39742, "pid": 5, "tid": 7, "ts": 1716454218106388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046576, "dur": 6,
    "args": {
      "External id": 39742,
      "cbid": 211, "correlation": 39742
    }
  },
  {
    "ph": "s", "id": 39742, "pid": 76337, "tid": -914061504, "ts": 1716454218046576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218107025, "dur": 12,
    "args": {
      "External id": 39744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39744, "pid": 5, "tid": 7, "ts": 1716454218107025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046585, "dur": 5,
    "args": {
      "External id": 39744,
      "cbid": 211, "correlation": 39744
    }
  },
  {
    "ph": "s", "id": 39744, "pid": 76337, "tid": -914061504, "ts": 1716454218046585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218107038, "dur": 14,
    "args": {
      "External id": 39750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39750, "pid": 5, "tid": 7, "ts": 1716454218107038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046613, "dur": 9,
    "args": {
      "External id": 39750,
      "cbid": 211, "correlation": 39750
    }
  },
  {
    "ph": "s", "id": 39750, "pid": 76337, "tid": -914061504, "ts": 1716454218046613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046671, "dur": 0,
    "args": {
      "External id": 39760,
      "cbid": 317, "correlation": 39760
    }
  },
  {
    "ph": "f", "id": 39760, "pid": 76337, "tid": -914061504, "ts": 1716454218046671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046672, "dur": 0,
    "args": {
      "External id": 39761,
      "cbid": 203, "correlation": 39761
    }
  },
  {
    "ph": "f", "id": 39761, "pid": 76337, "tid": -914061504, "ts": 1716454218046672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046673, "dur": 0,
    "args": {
      "External id": 39762,
      "cbid": 205, "correlation": 39762
    }
  },
  {
    "ph": "f", "id": 39762, "pid": 76337, "tid": -914061504, "ts": 1716454218046673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218107054, "dur": 22,
    "args": {
      "External id": 39766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39766, "pid": 5, "tid": 7, "ts": 1716454218107054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046685, "dur": 11,
    "args": {
      "External id": 39766,
      "cbid": 211, "correlation": 39766
    }
  },
  {
    "ph": "s", "id": 39766, "pid": 76337, "tid": -914061504, "ts": 1716454218046685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218107077, "dur": 4,
    "args": {
      "External id": 39768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39768,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39768, "pid": 5, "tid": 7, "ts": 1716454218107077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046700, "dur": 6,
    "args": {
      "External id": 39768,
      "cbid": 211, "correlation": 39768
    }
  },
  {
    "ph": "s", "id": 39768, "pid": 76337, "tid": -914061504, "ts": 1716454218046700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046709, "dur": 0,
    "args": {
      "External id": 39769,
      "cbid": 51, "correlation": 39769
    }
  },
  {
    "ph": "s", "id": 39769, "pid": 76337, "tid": -914061504, "ts": 1716454218046709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218107082, "dur": 166,
    "args": {
      "External id": 39770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39770,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 39770, "pid": 5, "tid": 7, "ts": 1716454218107082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046710, "dur": 5,
    "args": {
      "External id": 39770,
      "cbid": 211, "correlation": 39770
    }
  },
  {
    "ph": "s", "id": 39770, "pid": 76337, "tid": -914061504, "ts": 1716454218046710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218107249, "dur": 16,
    "args": {
      "External id": 39775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39775, "pid": 5, "tid": 7, "ts": 1716454218107249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046735, "dur": 9,
    "args": {
      "External id": 39775,
      "cbid": 211, "correlation": 39775
    }
  },
  {
    "ph": "s", "id": 39775, "pid": 76337, "tid": -914061504, "ts": 1716454218046735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218107266, "dur": 13,
    "args": {
      "External id": 39783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39783,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39783, "pid": 5, "tid": 7, "ts": 1716454218107266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046763, "dur": 8,
    "args": {
      "External id": 39783,
      "cbid": 211, "correlation": 39783
    }
  },
  {
    "ph": "s", "id": 39783, "pid": 76337, "tid": -914061504, "ts": 1716454218046763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218107281, "dur": 10,
    "args": {
      "External id": 39791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39791,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39791, "pid": 5, "tid": 7, "ts": 1716454218107281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046791, "dur": 8,
    "args": {
      "External id": 39791,
      "cbid": 211, "correlation": 39791
    }
  },
  {
    "ph": "s", "id": 39791, "pid": 76337, "tid": -914061504, "ts": 1716454218046791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218107292, "dur": 18,
    "args": {
      "External id": 39811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39811,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 39811, "pid": 5, "tid": 7, "ts": 1716454218107292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046874, "dur": 13,
    "args": {
      "External id": 39811,
      "cbid": 211, "correlation": 39811
    }
  },
  {
    "ph": "s", "id": 39811, "pid": 76337, "tid": -914061504, "ts": 1716454218046874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218107311, "dur": 4,
    "args": {
      "External id": 39823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39823,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 39823, "pid": 5, "tid": 7, "ts": 1716454218107311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046896, "dur": 6,
    "args": {
      "External id": 39823,
      "cbid": 211, "correlation": 39823
    }
  },
  {
    "ph": "s", "id": 39823, "pid": 76337, "tid": -914061504, "ts": 1716454218046896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218107317, "dur": 16,
    "args": {
      "External id": 39826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39826,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39826, "pid": 5, "tid": 7, "ts": 1716454218107317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046914, "dur": 7,
    "args": {
      "External id": 39826,
      "cbid": 211, "correlation": 39826
    }
  },
  {
    "ph": "s", "id": 39826, "pid": 76337, "tid": -914061504, "ts": 1716454218046914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046971, "dur": 0,
    "args": {
      "External id": 39837,
      "cbid": 317, "correlation": 39837
    }
  },
  {
    "ph": "f", "id": 39837, "pid": 76337, "tid": -914061504, "ts": 1716454218046971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046972, "dur": 0,
    "args": {
      "External id": 39838,
      "cbid": 203, "correlation": 39838
    }
  },
  {
    "ph": "f", "id": 39838, "pid": 76337, "tid": -914061504, "ts": 1716454218046972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046972, "dur": 8,
    "args": {
      "External id": 39839,
      "cbid": 205, "correlation": 39839
    }
  },
  {
    "ph": "f", "id": 39839, "pid": 76337, "tid": -914061504, "ts": 1716454218046972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218107334, "dur": 12,
    "args": {
      "External id": 39843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39843,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39843, "pid": 5, "tid": 7, "ts": 1716454218107334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218046993, "dur": 12,
    "args": {
      "External id": 39843,
      "cbid": 211, "correlation": 39843
    }
  },
  {
    "ph": "s", "id": 39843, "pid": 76337, "tid": -914061504, "ts": 1716454218046993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218107347, "dur": 3,
    "args": {
      "External id": 39845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 39845, "pid": 5, "tid": 7, "ts": 1716454218107347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047010, "dur": 6,
    "args": {
      "External id": 39845,
      "cbid": 211, "correlation": 39845
    }
  },
  {
    "ph": "s", "id": 39845, "pid": 76337, "tid": -914061504, "ts": 1716454218047010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047018, "dur": 0,
    "args": {
      "External id": 39846,
      "cbid": 51, "correlation": 39846
    }
  },
  {
    "ph": "s", "id": 39846, "pid": 76337, "tid": -914061504, "ts": 1716454218047018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218107352, "dur": 89,
    "args": {
      "External id": 39847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39847,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 39847, "pid": 5, "tid": 7, "ts": 1716454218107352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047019, "dur": 5,
    "args": {
      "External id": 39847,
      "cbid": 211, "correlation": 39847
    }
  },
  {
    "ph": "s", "id": 39847, "pid": 76337, "tid": -914061504, "ts": 1716454218047019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218107442, "dur": 15,
    "args": {
      "External id": 39852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39852, "pid": 5, "tid": 7, "ts": 1716454218107442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047046, "dur": 9,
    "args": {
      "External id": 39852,
      "cbid": 211, "correlation": 39852
    }
  },
  {
    "ph": "s", "id": 39852, "pid": 76337, "tid": -914061504, "ts": 1716454218047046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218107459, "dur": 82,
    "args": {
      "External id": 39861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39861,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39861, "pid": 5, "tid": 7, "ts": 1716454218107459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047129, "dur": 14,
    "args": {
      "External id": 39861,
      "cbid": 211, "correlation": 39861
    }
  },
  {
    "ph": "s", "id": 39861, "pid": 76337, "tid": -914061504, "ts": 1716454218047129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218107542, "dur": 29,
    "args": {
      "External id": 39883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39883,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39883, "pid": 5, "tid": 7, "ts": 1716454218107542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047186, "dur": 10,
    "args": {
      "External id": 39883,
      "cbid": 211, "correlation": 39883
    }
  },
  {
    "ph": "s", "id": 39883, "pid": 76337, "tid": -914061504, "ts": 1716454218047186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047278, "dur": 1,
    "args": {
      "External id": 39894,
      "cbid": 251, "correlation": 39894
    }
  },
  {
    "ph": "f", "id": 39894, "pid": 76337, "tid": -914061504, "ts": 1716454218047278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218107573, "dur": 161,
    "args": {
      "External id": 39895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39895,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39895, "pid": 5, "tid": 7, "ts": 1716454218107573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047283, "dur": 13,
    "args": {
      "External id": 39895,
      "cbid": 211, "correlation": 39895
    }
  },
  {
    "ph": "s", "id": 39895, "pid": 76337, "tid": -914061504, "ts": 1716454218047283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047353, "dur": 1,
    "args": {
      "External id": 39906,
      "cbid": 251, "correlation": 39906
    }
  },
  {
    "ph": "f", "id": 39906, "pid": 76337, "tid": -914061504, "ts": 1716454218047353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218107735, "dur": 155,
    "args": {
      "External id": 39907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39907,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39907, "pid": 5, "tid": 7, "ts": 1716454218107735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047358, "dur": 11,
    "args": {
      "External id": 39907,
      "cbid": 211, "correlation": 39907
    }
  },
  {
    "ph": "s", "id": 39907, "pid": 76337, "tid": -914061504, "ts": 1716454218047358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047423, "dur": 1,
    "args": {
      "External id": 39918,
      "cbid": 251, "correlation": 39918
    }
  },
  {
    "ph": "f", "id": 39918, "pid": 76337, "tid": -914061504, "ts": 1716454218047423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218107891, "dur": 155,
    "args": {
      "External id": 39919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39919,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39919, "pid": 5, "tid": 7, "ts": 1716454218107891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047427, "dur": 11,
    "args": {
      "External id": 39919,
      "cbid": 211, "correlation": 39919
    }
  },
  {
    "ph": "s", "id": 39919, "pid": 76337, "tid": -914061504, "ts": 1716454218047427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218108048, "dur": 326,
    "args": {
      "External id": 39944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39944,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39944, "pid": 5, "tid": 7, "ts": 1716454218108048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047511, "dur": 13,
    "args": {
      "External id": 39944,
      "cbid": 211, "correlation": 39944
    }
  },
  {
    "ph": "s", "id": 39944, "pid": 76337, "tid": -914061504, "ts": 1716454218047511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047611, "dur": 1,
    "args": {
      "External id": 39962,
      "cbid": 251, "correlation": 39962
    }
  },
  {
    "ph": "f", "id": 39962, "pid": 76337, "tid": -914061504, "ts": 1716454218047611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218108376, "dur": 162,
    "args": {
      "External id": 39964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39964,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 39964, "pid": 5, "tid": 7, "ts": 1716454218108376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047617, "dur": 13,
    "args": {
      "External id": 39964,
      "cbid": 211, "correlation": 39964
    }
  },
  {
    "ph": "s", "id": 39964, "pid": 76337, "tid": -914061504, "ts": 1716454218047617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218108539, "dur": 19,
    "args": {
      "External id": 39972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39972,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39972, "pid": 5, "tid": 7, "ts": 1716454218108539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047688, "dur": 12,
    "args": {
      "External id": 39972,
      "cbid": 211, "correlation": 39972
    }
  },
  {
    "ph": "s", "id": 39972, "pid": 76337, "tid": -914061504, "ts": 1716454218047688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218108560, "dur": 28,
    "args": {
      "External id": 39980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39980,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39980, "pid": 5, "tid": 7, "ts": 1716454218108560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047726, "dur": 9,
    "args": {
      "External id": 39980,
      "cbid": 211, "correlation": 39980
    }
  },
  {
    "ph": "s", "id": 39980, "pid": 76337, "tid": -914061504, "ts": 1716454218047726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218108588, "dur": 18,
    "args": {
      "External id": 39991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 39991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 39991, "pid": 5, "tid": 7, "ts": 1716454218108588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047800, "dur": 12,
    "args": {
      "External id": 39991,
      "cbid": 211, "correlation": 39991
    }
  },
  {
    "ph": "s", "id": 39991, "pid": 76337, "tid": -914061504, "ts": 1716454218047800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218108608, "dur": 16,
    "args": {
      "External id": 40013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40013,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40013, "pid": 5, "tid": 7, "ts": 1716454218108608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047831, "dur": 7,
    "args": {
      "External id": 40013,
      "cbid": 211, "correlation": 40013
    }
  },
  {
    "ph": "s", "id": 40013, "pid": 76337, "tid": -914061504, "ts": 1716454218047831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047917, "dur": 1,
    "args": {
      "External id": 40024,
      "cbid": 251, "correlation": 40024
    }
  },
  {
    "ph": "f", "id": 40024, "pid": 76337, "tid": -914061504, "ts": 1716454218047917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218108625, "dur": 87,
    "args": {
      "External id": 40025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40025,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40025, "pid": 5, "tid": 7, "ts": 1716454218108625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218047922, "dur": 13,
    "args": {
      "External id": 40025,
      "cbid": 211, "correlation": 40025
    }
  },
  {
    "ph": "s", "id": 40025, "pid": 76337, "tid": -914061504, "ts": 1716454218047922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048001, "dur": 1,
    "args": {
      "External id": 40036,
      "cbid": 251, "correlation": 40036
    }
  },
  {
    "ph": "f", "id": 40036, "pid": 76337, "tid": -914061504, "ts": 1716454218048001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048005, "dur": 0,
    "args": {
      "External id": 40037,
      "cbid": 251, "correlation": 40037
    }
  },
  {
    "ph": "f", "id": 40037, "pid": 76337, "tid": -914061504, "ts": 1716454218048005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218108713, "dur": 12,
    "args": {
      "External id": 40038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40038,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40038, "pid": 5, "tid": 7, "ts": 1716454218108713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048006, "dur": 12,
    "args": {
      "External id": 40038,
      "cbid": 211, "correlation": 40038
    }
  },
  {
    "ph": "s", "id": 40038, "pid": 76337, "tid": -914061504, "ts": 1716454218048006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218108726, "dur": 5,
    "args": {
      "External id": 40040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40040,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40040, "pid": 5, "tid": 7, "ts": 1716454218108726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048020, "dur": 6,
    "args": {
      "External id": 40040,
      "cbid": 211, "correlation": 40040
    }
  },
  {
    "ph": "s", "id": 40040, "pid": 76337, "tid": -914061504, "ts": 1716454218048020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048078, "dur": 1,
    "args": {
      "External id": 40051,
      "cbid": 251, "correlation": 40051
    }
  },
  {
    "ph": "f", "id": 40051, "pid": 76337, "tid": -914061504, "ts": 1716454218048078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048081, "dur": 0,
    "args": {
      "External id": 40052,
      "cbid": 251, "correlation": 40052
    }
  },
  {
    "ph": "f", "id": 40052, "pid": 76337, "tid": -914061504, "ts": 1716454218048081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218108733, "dur": 8,
    "args": {
      "External id": 40053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40053,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40053, "pid": 5, "tid": 7, "ts": 1716454218108733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048083, "dur": 12,
    "args": {
      "External id": 40053,
      "cbid": 211, "correlation": 40053
    }
  },
  {
    "ph": "s", "id": 40053, "pid": 76337, "tid": -914061504, "ts": 1716454218048083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218108742, "dur": 3,
    "args": {
      "External id": 40055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40055,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40055, "pid": 5, "tid": 7, "ts": 1716454218108742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048097, "dur": 5,
    "args": {
      "External id": 40055,
      "cbid": 211, "correlation": 40055
    }
  },
  {
    "ph": "s", "id": 40055, "pid": 76337, "tid": -914061504, "ts": 1716454218048097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218108746, "dur": 54,
    "args": {
      "External id": 40080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40080,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40080, "pid": 5, "tid": 7, "ts": 1716454218108746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048173, "dur": 12,
    "args": {
      "External id": 40080,
      "cbid": 211, "correlation": 40080
    }
  },
  {
    "ph": "s", "id": 40080, "pid": 76337, "tid": -914061504, "ts": 1716454218048173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048273, "dur": 1,
    "args": {
      "External id": 40098,
      "cbid": 251, "correlation": 40098
    }
  },
  {
    "ph": "f", "id": 40098, "pid": 76337, "tid": -914061504, "ts": 1716454218048273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218108802, "dur": 89,
    "args": {
      "External id": 40100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40100,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40100, "pid": 5, "tid": 7, "ts": 1716454218108802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048279, "dur": 14,
    "args": {
      "External id": 40100,
      "cbid": 211, "correlation": 40100
    }
  },
  {
    "ph": "s", "id": 40100, "pid": 76337, "tid": -914061504, "ts": 1716454218048279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218108892, "dur": 10,
    "args": {
      "External id": 40108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40108,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40108, "pid": 5, "tid": 7, "ts": 1716454218108892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048348, "dur": 12,
    "args": {
      "External id": 40108,
      "cbid": 211, "correlation": 40108
    }
  },
  {
    "ph": "s", "id": 40108, "pid": 76337, "tid": -914061504, "ts": 1716454218048348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218108903, "dur": 20,
    "args": {
      "External id": 40116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40116,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40116, "pid": 5, "tid": 7, "ts": 1716454218108903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048390, "dur": 9,
    "args": {
      "External id": 40116,
      "cbid": 211, "correlation": 40116
    }
  },
  {
    "ph": "s", "id": 40116, "pid": 76337, "tid": -914061504, "ts": 1716454218048390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218108925, "dur": 17,
    "args": {
      "External id": 40138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40138,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40138, "pid": 5, "tid": 7, "ts": 1716454218108925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048442, "dur": 10,
    "args": {
      "External id": 40138,
      "cbid": 211, "correlation": 40138
    }
  },
  {
    "ph": "s", "id": 40138, "pid": 76337, "tid": -914061504, "ts": 1716454218048442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048527, "dur": 1,
    "args": {
      "External id": 40154,
      "cbid": 251, "correlation": 40154
    }
  },
  {
    "ph": "f", "id": 40154, "pid": 76337, "tid": -914061504, "ts": 1716454218048527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048532, "dur": 0,
    "args": {
      "External id": 40156,
      "cbid": 251, "correlation": 40156
    }
  },
  {
    "ph": "f", "id": 40156, "pid": 76337, "tid": -914061504, "ts": 1716454218048532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218108944, "dur": 491,
    "args": {
      "External id": 40157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40157,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40157, "pid": 5, "tid": 7, "ts": 1716454218108944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048534, "dur": 13,
    "args": {
      "External id": 40157,
      "cbid": 211, "correlation": 40157
    }
  },
  {
    "ph": "s", "id": 40157, "pid": 76337, "tid": -914061504, "ts": 1716454218048534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218109436, "dur": 65,
    "args": {
      "External id": 40165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40165, "pid": 5, "tid": 7, "ts": 1716454218109436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048600, "dur": 12,
    "args": {
      "External id": 40165,
      "cbid": 211, "correlation": 40165
    }
  },
  {
    "ph": "s", "id": 40165, "pid": 76337, "tid": -914061504, "ts": 1716454218048600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218109503, "dur": 66,
    "args": {
      "External id": 40173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40173,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40173, "pid": 5, "tid": 7, "ts": 1716454218109503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048630, "dur": 9,
    "args": {
      "External id": 40173,
      "cbid": 211, "correlation": 40173
    }
  },
  {
    "ph": "s", "id": 40173, "pid": 76337, "tid": -914061504, "ts": 1716454218048630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048712, "dur": 1,
    "args": {
      "External id": 40189,
      "cbid": 251, "correlation": 40189
    }
  },
  {
    "ph": "f", "id": 40189, "pid": 76337, "tid": -914061504, "ts": 1716454218048712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218109571, "dur": 1,
    "args": {
      "External id": 40191,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 40191,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 40191, "pid": 5, "tid": 7, "ts": 1716454218109571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048717, "dur": 9,
    "args": {
      "External id": 40191,
      "cbid": 51, "correlation": 40191
    }
  },
  {
    "ph": "s", "id": 40191, "pid": 76337, "tid": -914061504, "ts": 1716454218048717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218109575, "dur": 263,
    "args": {
      "External id": 40192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40192,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40192, "pid": 5, "tid": 7, "ts": 1716454218109575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048727, "dur": 11,
    "args": {
      "External id": 40192,
      "cbid": 211, "correlation": 40192
    }
  },
  {
    "ph": "s", "id": 40192, "pid": 76337, "tid": -914061504, "ts": 1716454218048727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218109839, "dur": 14,
    "args": {
      "External id": 40200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40200,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40200, "pid": 5, "tid": 7, "ts": 1716454218109839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048769, "dur": 10,
    "args": {
      "External id": 40200,
      "cbid": 211, "correlation": 40200
    }
  },
  {
    "ph": "s", "id": 40200, "pid": 76337, "tid": -914061504, "ts": 1716454218048769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218109854, "dur": 37,
    "args": {
      "External id": 40211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40211, "pid": 5, "tid": 7, "ts": 1716454218109854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048837, "dur": 13,
    "args": {
      "External id": 40211,
      "cbid": 211, "correlation": 40211
    }
  },
  {
    "ph": "s", "id": 40211, "pid": 76337, "tid": -914061504, "ts": 1716454218048837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048947, "dur": 0,
    "args": {
      "External id": 40223,
      "cbid": 317, "correlation": 40223
    }
  },
  {
    "ph": "f", "id": 40223, "pid": 76337, "tid": -914061504, "ts": 1716454218048947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048948, "dur": 0,
    "args": {
      "External id": 40224,
      "cbid": 203, "correlation": 40224
    }
  },
  {
    "ph": "f", "id": 40224, "pid": 76337, "tid": -914061504, "ts": 1716454218048948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048949, "dur": 0,
    "args": {
      "External id": 40225,
      "cbid": 205, "correlation": 40225
    }
  },
  {
    "ph": "f", "id": 40225, "pid": 76337, "tid": -914061504, "ts": 1716454218048949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218109893, "dur": 13,
    "args": {
      "External id": 40229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40229, "pid": 5, "tid": 7, "ts": 1716454218109893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048965, "dur": 20,
    "args": {
      "External id": 40229,
      "cbid": 211, "correlation": 40229
    }
  },
  {
    "ph": "s", "id": 40229, "pid": 76337, "tid": -914061504, "ts": 1716454218048965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218109907, "dur": 4,
    "args": {
      "External id": 40231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40231, "pid": 5, "tid": 7, "ts": 1716454218109907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048990, "dur": 6,
    "args": {
      "External id": 40231,
      "cbid": 211, "correlation": 40231
    }
  },
  {
    "ph": "s", "id": 40231, "pid": 76337, "tid": -914061504, "ts": 1716454218048990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218048999, "dur": 0,
    "args": {
      "External id": 40232,
      "cbid": 51, "correlation": 40232
    }
  },
  {
    "ph": "s", "id": 40232, "pid": 76337, "tid": -914061504, "ts": 1716454218048999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218109912, "dur": 95,
    "args": {
      "External id": 40233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40233,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 40233, "pid": 5, "tid": 7, "ts": 1716454218109912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049000, "dur": 6,
    "args": {
      "External id": 40233,
      "cbid": 211, "correlation": 40233
    }
  },
  {
    "ph": "s", "id": 40233, "pid": 76337, "tid": -914061504, "ts": 1716454218049000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218110008, "dur": 15,
    "args": {
      "External id": 40238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40238, "pid": 5, "tid": 7, "ts": 1716454218110008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049029, "dur": 9,
    "args": {
      "External id": 40238,
      "cbid": 211, "correlation": 40238
    }
  },
  {
    "ph": "s", "id": 40238, "pid": 76337, "tid": -914061504, "ts": 1716454218049029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218110025, "dur": 12,
    "args": {
      "External id": 40246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40246,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40246, "pid": 5, "tid": 7, "ts": 1716454218110025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049062, "dur": 8,
    "args": {
      "External id": 40246,
      "cbid": 211, "correlation": 40246
    }
  },
  {
    "ph": "s", "id": 40246, "pid": 76337, "tid": -914061504, "ts": 1716454218049062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218110038, "dur": 25,
    "args": {
      "External id": 40255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40255,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40255, "pid": 5, "tid": 7, "ts": 1716454218110038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049101, "dur": 11,
    "args": {
      "External id": 40255,
      "cbid": 211, "correlation": 40255
    }
  },
  {
    "ph": "s", "id": 40255, "pid": 76337, "tid": -914061504, "ts": 1716454218049101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218110065, "dur": 24,
    "args": {
      "External id": 40275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40275,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 40275, "pid": 5, "tid": 7, "ts": 1716454218110065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049174, "dur": 12,
    "args": {
      "External id": 40275,
      "cbid": 211, "correlation": 40275
    }
  },
  {
    "ph": "s", "id": 40275, "pid": 76337, "tid": -914061504, "ts": 1716454218049174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218110090, "dur": 5,
    "args": {
      "External id": 40287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40287,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 40287, "pid": 5, "tid": 7, "ts": 1716454218110090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049196, "dur": 6,
    "args": {
      "External id": 40287,
      "cbid": 211, "correlation": 40287
    }
  },
  {
    "ph": "s", "id": 40287, "pid": 76337, "tid": -914061504, "ts": 1716454218049196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218110096, "dur": 24,
    "args": {
      "External id": 40290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40290, "pid": 5, "tid": 7, "ts": 1716454218110096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049215, "dur": 7,
    "args": {
      "External id": 40290,
      "cbid": 211, "correlation": 40290
    }
  },
  {
    "ph": "s", "id": 40290, "pid": 76337, "tid": -914061504, "ts": 1716454218049215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218110122, "dur": 17,
    "args": {
      "External id": 40299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40299,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40299, "pid": 5, "tid": 7, "ts": 1716454218110122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049254, "dur": 10,
    "args": {
      "External id": 40299,
      "cbid": 211, "correlation": 40299
    }
  },
  {
    "ph": "s", "id": 40299, "pid": 76337, "tid": -914061504, "ts": 1716454218049254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049306, "dur": 0,
    "args": {
      "External id": 40309,
      "cbid": 317, "correlation": 40309
    }
  },
  {
    "ph": "f", "id": 40309, "pid": 76337, "tid": -914061504, "ts": 1716454218049306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049307, "dur": 0,
    "args": {
      "External id": 40310,
      "cbid": 203, "correlation": 40310
    }
  },
  {
    "ph": "f", "id": 40310, "pid": 76337, "tid": -914061504, "ts": 1716454218049307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049307, "dur": 0,
    "args": {
      "External id": 40311,
      "cbid": 205, "correlation": 40311
    }
  },
  {
    "ph": "f", "id": 40311, "pid": 76337, "tid": -914061504, "ts": 1716454218049307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218110140, "dur": 18,
    "args": {
      "External id": 40315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40315,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40315, "pid": 5, "tid": 7, "ts": 1716454218110140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049323, "dur": 11,
    "args": {
      "External id": 40315,
      "cbid": 211, "correlation": 40315
    }
  },
  {
    "ph": "s", "id": 40315, "pid": 76337, "tid": -914061504, "ts": 1716454218049323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218110159, "dur": 237,
    "args": {
      "External id": 40317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40317,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40317, "pid": 5, "tid": 7, "ts": 1716454218110159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049337, "dur": 5,
    "args": {
      "External id": 40317,
      "cbid": 211, "correlation": 40317
    }
  },
  {
    "ph": "s", "id": 40317, "pid": 76337, "tid": -914061504, "ts": 1716454218049337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218110398, "dur": 1,
    "args": {
      "External id": 40319,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 40319,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5084745762711864
    }
  },
  {
    "ph": "f", "id": 40319, "pid": 5, "tid": 7, "ts": 1716454218110398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049349, "dur": 7,
    "args": {
      "External id": 40319,
      "cbid": 51, "correlation": 40319
    }
  },
  {
    "ph": "s", "id": 40319, "pid": 76337, "tid": -914061504, "ts": 1716454218049349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218110402, "dur": 805,
    "args": {
      "External id": 40320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40320,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40320, "pid": 5, "tid": 7, "ts": 1716454218110402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049357, "dur": 8,
    "args": {
      "External id": 40320,
      "cbid": 211, "correlation": 40320
    }
  },
  {
    "ph": "s", "id": 40320, "pid": 76337, "tid": -914061504, "ts": 1716454218049357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218111208, "dur": 13,
    "args": {
      "External id": 40322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40322,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40322, "pid": 5, "tid": 7, "ts": 1716454218111208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049370, "dur": 6,
    "args": {
      "External id": 40322,
      "cbid": 211, "correlation": 40322
    }
  },
  {
    "ph": "s", "id": 40322, "pid": 76337, "tid": -914061504, "ts": 1716454218049370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218111222, "dur": 14,
    "args": {
      "External id": 40328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40328, "pid": 5, "tid": 7, "ts": 1716454218111222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049399, "dur": 9,
    "args": {
      "External id": 40328,
      "cbid": 211, "correlation": 40328
    }
  },
  {
    "ph": "s", "id": 40328, "pid": 76337, "tid": -914061504, "ts": 1716454218049399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218111238, "dur": 3,
    "args": {
      "External id": 40336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40336,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 40336, "pid": 5, "tid": 7, "ts": 1716454218111238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049443, "dur": 9,
    "args": {
      "External id": 40336,
      "cbid": 211, "correlation": 40336
    }
  },
  {
    "ph": "s", "id": 40336, "pid": 76337, "tid": -914061504, "ts": 1716454218049443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049510, "dur": 1,
    "args": {
      "External id": 40352,
      "cbid": 251, "correlation": 40352
    }
  },
  {
    "ph": "f", "id": 40352, "pid": 76337, "tid": -914061504, "ts": 1716454218049510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049515, "dur": 0,
    "args": {
      "External id": 40354,
      "cbid": 251, "correlation": 40354
    }
  },
  {
    "ph": "f", "id": 40354, "pid": 76337, "tid": -914061504, "ts": 1716454218049515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218111242, "dur": 13,
    "args": {
      "External id": 40355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40355,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40355, "pid": 5, "tid": 7, "ts": 1716454218111242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049517, "dur": 12,
    "args": {
      "External id": 40355,
      "cbid": 211, "correlation": 40355
    }
  },
  {
    "ph": "s", "id": 40355, "pid": 76337, "tid": -914061504, "ts": 1716454218049517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218111257, "dur": 5,
    "args": {
      "External id": 40357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40357,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40357, "pid": 5, "tid": 7, "ts": 1716454218111257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049531, "dur": 5,
    "args": {
      "External id": 40357,
      "cbid": 211, "correlation": 40357
    }
  },
  {
    "ph": "s", "id": 40357, "pid": 76337, "tid": -914061504, "ts": 1716454218049531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218111263, "dur": 16,
    "args": {
      "External id": 40367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40367, "pid": 5, "tid": 7, "ts": 1716454218111263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049589, "dur": 11,
    "args": {
      "External id": 40367,
      "cbid": 211, "correlation": 40367
    }
  },
  {
    "ph": "s", "id": 40367, "pid": 76337, "tid": -914061504, "ts": 1716454218049589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218111281, "dur": 17,
    "args": {
      "External id": 40387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40387,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 40387, "pid": 5, "tid": 7, "ts": 1716454218111281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049655, "dur": 11,
    "args": {
      "External id": 40387,
      "cbid": 211, "correlation": 40387
    }
  },
  {
    "ph": "s", "id": 40387, "pid": 76337, "tid": -914061504, "ts": 1716454218049655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218111299, "dur": 4,
    "args": {
      "External id": 40399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40399,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 40399, "pid": 5, "tid": 7, "ts": 1716454218111299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049676, "dur": 7,
    "args": {
      "External id": 40399,
      "cbid": 211, "correlation": 40399
    }
  },
  {
    "ph": "s", "id": 40399, "pid": 76337, "tid": -914061504, "ts": 1716454218049676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218111304, "dur": 16,
    "args": {
      "External id": 40402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40402,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40402, "pid": 5, "tid": 7, "ts": 1716454218111304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049694, "dur": 6,
    "args": {
      "External id": 40402,
      "cbid": 211, "correlation": 40402
    }
  },
  {
    "ph": "s", "id": 40402, "pid": 76337, "tid": -914061504, "ts": 1716454218049694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218111321, "dur": 10,
    "args": {
      "External id": 40411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40411,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40411, "pid": 5, "tid": 7, "ts": 1716454218111321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049736, "dur": 10,
    "args": {
      "External id": 40411,
      "cbid": 211, "correlation": 40411
    }
  },
  {
    "ph": "s", "id": 40411, "pid": 76337, "tid": -914061504, "ts": 1716454218049736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049798, "dur": 0,
    "args": {
      "External id": 40421,
      "cbid": 317, "correlation": 40421
    }
  },
  {
    "ph": "f", "id": 40421, "pid": 76337, "tid": -914061504, "ts": 1716454218049798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049799, "dur": 0,
    "args": {
      "External id": 40422,
      "cbid": 203, "correlation": 40422
    }
  },
  {
    "ph": "f", "id": 40422, "pid": 76337, "tid": -914061504, "ts": 1716454218049799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049799, "dur": 0,
    "args": {
      "External id": 40423,
      "cbid": 205, "correlation": 40423
    }
  },
  {
    "ph": "f", "id": 40423, "pid": 76337, "tid": -914061504, "ts": 1716454218049799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218111333, "dur": 12,
    "args": {
      "External id": 40427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40427,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40427, "pid": 5, "tid": 7, "ts": 1716454218111333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049813, "dur": 12,
    "args": {
      "External id": 40427,
      "cbid": 211, "correlation": 40427
    }
  },
  {
    "ph": "s", "id": 40427, "pid": 76337, "tid": -914061504, "ts": 1716454218049813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218111346, "dur": 157,
    "args": {
      "External id": 40429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40429,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40429, "pid": 5, "tid": 7, "ts": 1716454218111346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049827, "dur": 6,
    "args": {
      "External id": 40429,
      "cbid": 211, "correlation": 40429
    }
  },
  {
    "ph": "s", "id": 40429, "pid": 76337, "tid": -914061504, "ts": 1716454218049827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218111506, "dur": 1,
    "args": {
      "External id": 40431,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 40431,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5996252342286071
    }
  },
  {
    "ph": "f", "id": 40431, "pid": 5, "tid": 7, "ts": 1716454218111506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049839, "dur": 6,
    "args": {
      "External id": 40431,
      "cbid": 51, "correlation": 40431
    }
  },
  {
    "ph": "s", "id": 40431, "pid": 76337, "tid": -914061504, "ts": 1716454218049839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218111509, "dur": 636,
    "args": {
      "External id": 40432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40432,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40432, "pid": 5, "tid": 7, "ts": 1716454218111509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049847, "dur": 6,
    "args": {
      "External id": 40432,
      "cbid": 211, "correlation": 40432
    }
  },
  {
    "ph": "s", "id": 40432, "pid": 76337, "tid": -914061504, "ts": 1716454218049847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218112147, "dur": 11,
    "args": {
      "External id": 40434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40434,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40434, "pid": 5, "tid": 7, "ts": 1716454218112147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049856, "dur": 5,
    "args": {
      "External id": 40434,
      "cbid": 211, "correlation": 40434
    }
  },
  {
    "ph": "s", "id": 40434, "pid": 76337, "tid": -914061504, "ts": 1716454218049856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218112159, "dur": 14,
    "args": {
      "External id": 40440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40440, "pid": 5, "tid": 7, "ts": 1716454218112159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049884, "dur": 9,
    "args": {
      "External id": 40440,
      "cbid": 211, "correlation": 40440
    }
  },
  {
    "ph": "s", "id": 40440, "pid": 76337, "tid": -914061504, "ts": 1716454218049884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049943, "dur": 0,
    "args": {
      "External id": 40450,
      "cbid": 317, "correlation": 40450
    }
  },
  {
    "ph": "f", "id": 40450, "pid": 76337, "tid": -914061504, "ts": 1716454218049943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049944, "dur": 0,
    "args": {
      "External id": 40451,
      "cbid": 203, "correlation": 40451
    }
  },
  {
    "ph": "f", "id": 40451, "pid": 76337, "tid": -914061504, "ts": 1716454218049944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049945, "dur": 0,
    "args": {
      "External id": 40452,
      "cbid": 205, "correlation": 40452
    }
  },
  {
    "ph": "f", "id": 40452, "pid": 76337, "tid": -914061504, "ts": 1716454218049945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218112175, "dur": 16,
    "args": {
      "External id": 40456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40456, "pid": 5, "tid": 7, "ts": 1716454218112175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049959, "dur": 12,
    "args": {
      "External id": 40456,
      "cbid": 211, "correlation": 40456
    }
  },
  {
    "ph": "s", "id": 40456, "pid": 76337, "tid": -914061504, "ts": 1716454218049959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218112192, "dur": 4,
    "args": {
      "External id": 40458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40458, "pid": 5, "tid": 7, "ts": 1716454218112192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049983, "dur": 6,
    "args": {
      "External id": 40458,
      "cbid": 211, "correlation": 40458
    }
  },
  {
    "ph": "s", "id": 40458, "pid": 76337, "tid": -914061504, "ts": 1716454218049983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049992, "dur": 0,
    "args": {
      "External id": 40459,
      "cbid": 51, "correlation": 40459
    }
  },
  {
    "ph": "s", "id": 40459, "pid": 76337, "tid": -914061504, "ts": 1716454218049992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218112197, "dur": 127,
    "args": {
      "External id": 40460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40460,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 40460, "pid": 5, "tid": 7, "ts": 1716454218112197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218049993, "dur": 5,
    "args": {
      "External id": 40460,
      "cbid": 211, "correlation": 40460
    }
  },
  {
    "ph": "s", "id": 40460, "pid": 76337, "tid": -914061504, "ts": 1716454218049993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218112326, "dur": 15,
    "args": {
      "External id": 40465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40465,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40465, "pid": 5, "tid": 7, "ts": 1716454218112326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050020, "dur": 9,
    "args": {
      "External id": 40465,
      "cbid": 211, "correlation": 40465
    }
  },
  {
    "ph": "s", "id": 40465, "pid": 76337, "tid": -914061504, "ts": 1716454218050020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218112343, "dur": 11,
    "args": {
      "External id": 40473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40473,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40473, "pid": 5, "tid": 7, "ts": 1716454218112343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050050, "dur": 9,
    "args": {
      "External id": 40473,
      "cbid": 211, "correlation": 40473
    }
  },
  {
    "ph": "s", "id": 40473, "pid": 76337, "tid": -914061504, "ts": 1716454218050050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218112355, "dur": 10,
    "args": {
      "External id": 40481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40481,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40481, "pid": 5, "tid": 7, "ts": 1716454218112355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050079, "dur": 8,
    "args": {
      "External id": 40481,
      "cbid": 211, "correlation": 40481
    }
  },
  {
    "ph": "s", "id": 40481, "pid": 76337, "tid": -914061504, "ts": 1716454218050079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218112366, "dur": 18,
    "args": {
      "External id": 40501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40501,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 40501, "pid": 5, "tid": 7, "ts": 1716454218112366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050163, "dur": 12,
    "args": {
      "External id": 40501,
      "cbid": 211, "correlation": 40501
    }
  },
  {
    "ph": "s", "id": 40501, "pid": 76337, "tid": -914061504, "ts": 1716454218050163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218112386, "dur": 4,
    "args": {
      "External id": 40513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40513,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 40513, "pid": 5, "tid": 7, "ts": 1716454218112386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050186, "dur": 6,
    "args": {
      "External id": 40513,
      "cbid": 211, "correlation": 40513
    }
  },
  {
    "ph": "s", "id": 40513, "pid": 76337, "tid": -914061504, "ts": 1716454218050186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218112392, "dur": 17,
    "args": {
      "External id": 40516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40516,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40516, "pid": 5, "tid": 7, "ts": 1716454218112392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050204, "dur": 7,
    "args": {
      "External id": 40516,
      "cbid": 211, "correlation": 40516
    }
  },
  {
    "ph": "s", "id": 40516, "pid": 76337, "tid": -914061504, "ts": 1716454218050204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050261, "dur": 0,
    "args": {
      "External id": 40527,
      "cbid": 317, "correlation": 40527
    }
  },
  {
    "ph": "f", "id": 40527, "pid": 76337, "tid": -914061504, "ts": 1716454218050261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050262, "dur": 0,
    "args": {
      "External id": 40528,
      "cbid": 203, "correlation": 40528
    }
  },
  {
    "ph": "f", "id": 40528, "pid": 76337, "tid": -914061504, "ts": 1716454218050262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050263, "dur": 0,
    "args": {
      "External id": 40529,
      "cbid": 205, "correlation": 40529
    }
  },
  {
    "ph": "f", "id": 40529, "pid": 76337, "tid": -914061504, "ts": 1716454218050263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218112410, "dur": 11,
    "args": {
      "External id": 40533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40533,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40533, "pid": 5, "tid": 7, "ts": 1716454218112410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050276, "dur": 12,
    "args": {
      "External id": 40533,
      "cbid": 211, "correlation": 40533
    }
  },
  {
    "ph": "s", "id": 40533, "pid": 76337, "tid": -914061504, "ts": 1716454218050276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218112422, "dur": 3,
    "args": {
      "External id": 40535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40535, "pid": 5, "tid": 7, "ts": 1716454218112422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050292, "dur": 6,
    "args": {
      "External id": 40535,
      "cbid": 211, "correlation": 40535
    }
  },
  {
    "ph": "s", "id": 40535, "pid": 76337, "tid": -914061504, "ts": 1716454218050292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050300, "dur": 0,
    "args": {
      "External id": 40536,
      "cbid": 51, "correlation": 40536
    }
  },
  {
    "ph": "s", "id": 40536, "pid": 76337, "tid": -914061504, "ts": 1716454218050300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218112426, "dur": 88,
    "args": {
      "External id": 40537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40537,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 40537, "pid": 5, "tid": 7, "ts": 1716454218112426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050301, "dur": 5,
    "args": {
      "External id": 40537,
      "cbid": 211, "correlation": 40537
    }
  },
  {
    "ph": "s", "id": 40537, "pid": 76337, "tid": -914061504, "ts": 1716454218050301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218112515, "dur": 15,
    "args": {
      "External id": 40542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40542, "pid": 5, "tid": 7, "ts": 1716454218112515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050328, "dur": 8,
    "args": {
      "External id": 40542,
      "cbid": 211, "correlation": 40542
    }
  },
  {
    "ph": "s", "id": 40542, "pid": 76337, "tid": -914061504, "ts": 1716454218050328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218112532, "dur": 82,
    "args": {
      "External id": 40551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40551,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40551, "pid": 5, "tid": 7, "ts": 1716454218112532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050409, "dur": 15,
    "args": {
      "External id": 40551,
      "cbid": 211, "correlation": 40551
    }
  },
  {
    "ph": "s", "id": 40551, "pid": 76337, "tid": -914061504, "ts": 1716454218050409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218112616, "dur": 30,
    "args": {
      "External id": 40573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40573,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40573, "pid": 5, "tid": 7, "ts": 1716454218112616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050467, "dur": 10,
    "args": {
      "External id": 40573,
      "cbid": 211, "correlation": 40573
    }
  },
  {
    "ph": "s", "id": 40573, "pid": 76337, "tid": -914061504, "ts": 1716454218050467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050556, "dur": 1,
    "args": {
      "External id": 40584,
      "cbid": 251, "correlation": 40584
    }
  },
  {
    "ph": "f", "id": 40584, "pid": 76337, "tid": -914061504, "ts": 1716454218050556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218112647, "dur": 159,
    "args": {
      "External id": 40585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40585,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40585, "pid": 5, "tid": 7, "ts": 1716454218112647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050562, "dur": 14,
    "args": {
      "External id": 40585,
      "cbid": 211, "correlation": 40585
    }
  },
  {
    "ph": "s", "id": 40585, "pid": 76337, "tid": -914061504, "ts": 1716454218050562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050633, "dur": 1,
    "args": {
      "External id": 40596,
      "cbid": 251, "correlation": 40596
    }
  },
  {
    "ph": "f", "id": 40596, "pid": 76337, "tid": -914061504, "ts": 1716454218050633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218112807, "dur": 155,
    "args": {
      "External id": 40597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40597,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40597, "pid": 5, "tid": 7, "ts": 1716454218112807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050637, "dur": 11,
    "args": {
      "External id": 40597,
      "cbid": 211, "correlation": 40597
    }
  },
  {
    "ph": "s", "id": 40597, "pid": 76337, "tid": -914061504, "ts": 1716454218050637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050702, "dur": 1,
    "args": {
      "External id": 40608,
      "cbid": 251, "correlation": 40608
    }
  },
  {
    "ph": "f", "id": 40608, "pid": 76337, "tid": -914061504, "ts": 1716454218050702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218112963, "dur": 155,
    "args": {
      "External id": 40609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40609,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40609, "pid": 5, "tid": 7, "ts": 1716454218112963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050706, "dur": 11,
    "args": {
      "External id": 40609,
      "cbid": 211, "correlation": 40609
    }
  },
  {
    "ph": "s", "id": 40609, "pid": 76337, "tid": -914061504, "ts": 1716454218050706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218113120, "dur": 330,
    "args": {
      "External id": 40634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40634,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40634, "pid": 5, "tid": 7, "ts": 1716454218113120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050790, "dur": 13,
    "args": {
      "External id": 40634,
      "cbid": 211, "correlation": 40634
    }
  },
  {
    "ph": "s", "id": 40634, "pid": 76337, "tid": -914061504, "ts": 1716454218050790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050890, "dur": 1,
    "args": {
      "External id": 40652,
      "cbid": 251, "correlation": 40652
    }
  },
  {
    "ph": "f", "id": 40652, "pid": 76337, "tid": -914061504, "ts": 1716454218050890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218113451, "dur": 161,
    "args": {
      "External id": 40654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40654,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40654, "pid": 5, "tid": 7, "ts": 1716454218113451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050896, "dur": 13,
    "args": {
      "External id": 40654,
      "cbid": 211, "correlation": 40654
    }
  },
  {
    "ph": "s", "id": 40654, "pid": 76337, "tid": -914061504, "ts": 1716454218050896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218113614, "dur": 19,
    "args": {
      "External id": 40662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40662,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40662, "pid": 5, "tid": 7, "ts": 1716454218113614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218050966, "dur": 20,
    "args": {
      "External id": 40662,
      "cbid": 211, "correlation": 40662
    }
  },
  {
    "ph": "s", "id": 40662, "pid": 76337, "tid": -914061504, "ts": 1716454218050966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218113635, "dur": 28,
    "args": {
      "External id": 40670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40670,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40670, "pid": 5, "tid": 7, "ts": 1716454218113635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051013, "dur": 9,
    "args": {
      "External id": 40670,
      "cbid": 211, "correlation": 40670
    }
  },
  {
    "ph": "s", "id": 40670, "pid": 76337, "tid": -914061504, "ts": 1716454218051013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218113663, "dur": 18,
    "args": {
      "External id": 40681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40681,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40681, "pid": 5, "tid": 7, "ts": 1716454218113663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051087, "dur": 12,
    "args": {
      "External id": 40681,
      "cbid": 211, "correlation": 40681
    }
  },
  {
    "ph": "s", "id": 40681, "pid": 76337, "tid": -914061504, "ts": 1716454218051087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218113683, "dur": 15,
    "args": {
      "External id": 40703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40703,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40703, "pid": 5, "tid": 7, "ts": 1716454218113683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051119, "dur": 7,
    "args": {
      "External id": 40703,
      "cbid": 211, "correlation": 40703
    }
  },
  {
    "ph": "s", "id": 40703, "pid": 76337, "tid": -914061504, "ts": 1716454218051119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051203, "dur": 1,
    "args": {
      "External id": 40714,
      "cbid": 251, "correlation": 40714
    }
  },
  {
    "ph": "f", "id": 40714, "pid": 76337, "tid": -914061504, "ts": 1716454218051203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218113699, "dur": 87,
    "args": {
      "External id": 40715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40715,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40715, "pid": 5, "tid": 7, "ts": 1716454218113699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051208, "dur": 13,
    "args": {
      "External id": 40715,
      "cbid": 211, "correlation": 40715
    }
  },
  {
    "ph": "s", "id": 40715, "pid": 76337, "tid": -914061504, "ts": 1716454218051208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051279, "dur": 1,
    "args": {
      "External id": 40726,
      "cbid": 251, "correlation": 40726
    }
  },
  {
    "ph": "f", "id": 40726, "pid": 76337, "tid": -914061504, "ts": 1716454218051279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051282, "dur": 0,
    "args": {
      "External id": 40727,
      "cbid": 251, "correlation": 40727
    }
  },
  {
    "ph": "f", "id": 40727, "pid": 76337, "tid": -914061504, "ts": 1716454218051282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218113788, "dur": 12,
    "args": {
      "External id": 40728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40728,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40728, "pid": 5, "tid": 7, "ts": 1716454218113788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051284, "dur": 12,
    "args": {
      "External id": 40728,
      "cbid": 211, "correlation": 40728
    }
  },
  {
    "ph": "s", "id": 40728, "pid": 76337, "tid": -914061504, "ts": 1716454218051284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218113801, "dur": 5,
    "args": {
      "External id": 40730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40730,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40730, "pid": 5, "tid": 7, "ts": 1716454218113801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051298, "dur": 6,
    "args": {
      "External id": 40730,
      "cbid": 211, "correlation": 40730
    }
  },
  {
    "ph": "s", "id": 40730, "pid": 76337, "tid": -914061504, "ts": 1716454218051298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051355, "dur": 1,
    "args": {
      "External id": 40741,
      "cbid": 251, "correlation": 40741
    }
  },
  {
    "ph": "f", "id": 40741, "pid": 76337, "tid": -914061504, "ts": 1716454218051355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051358, "dur": 0,
    "args": {
      "External id": 40742,
      "cbid": 251, "correlation": 40742
    }
  },
  {
    "ph": "f", "id": 40742, "pid": 76337, "tid": -914061504, "ts": 1716454218051358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218113808, "dur": 8,
    "args": {
      "External id": 40743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40743,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40743, "pid": 5, "tid": 7, "ts": 1716454218113808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051359, "dur": 11,
    "args": {
      "External id": 40743,
      "cbid": 211, "correlation": 40743
    }
  },
  {
    "ph": "s", "id": 40743, "pid": 76337, "tid": -914061504, "ts": 1716454218051359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218113817, "dur": 3,
    "args": {
      "External id": 40745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40745,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40745, "pid": 5, "tid": 7, "ts": 1716454218113817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051372, "dur": 5,
    "args": {
      "External id": 40745,
      "cbid": 211, "correlation": 40745
    }
  },
  {
    "ph": "s", "id": 40745, "pid": 76337, "tid": -914061504, "ts": 1716454218051372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218113821, "dur": 54,
    "args": {
      "External id": 40770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40770,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40770, "pid": 5, "tid": 7, "ts": 1716454218113821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051448, "dur": 13,
    "args": {
      "External id": 40770,
      "cbid": 211, "correlation": 40770
    }
  },
  {
    "ph": "s", "id": 40770, "pid": 76337, "tid": -914061504, "ts": 1716454218051448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051546, "dur": 1,
    "args": {
      "External id": 40788,
      "cbid": 251, "correlation": 40788
    }
  },
  {
    "ph": "f", "id": 40788, "pid": 76337, "tid": -914061504, "ts": 1716454218051546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218113877, "dur": 89,
    "args": {
      "External id": 40790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40790,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40790, "pid": 5, "tid": 7, "ts": 1716454218113877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051551, "dur": 14,
    "args": {
      "External id": 40790,
      "cbid": 211, "correlation": 40790
    }
  },
  {
    "ph": "s", "id": 40790, "pid": 76337, "tid": -914061504, "ts": 1716454218051551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218113967, "dur": 10,
    "args": {
      "External id": 40798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40798,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40798, "pid": 5, "tid": 7, "ts": 1716454218113967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051622, "dur": 12,
    "args": {
      "External id": 40798,
      "cbid": 211, "correlation": 40798
    }
  },
  {
    "ph": "s", "id": 40798, "pid": 76337, "tid": -914061504, "ts": 1716454218051622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218113978, "dur": 21,
    "args": {
      "External id": 40806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40806,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40806, "pid": 5, "tid": 7, "ts": 1716454218113978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051663, "dur": 9,
    "args": {
      "External id": 40806,
      "cbid": 211, "correlation": 40806
    }
  },
  {
    "ph": "s", "id": 40806, "pid": 76337, "tid": -914061504, "ts": 1716454218051663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218114001, "dur": 17,
    "args": {
      "External id": 40828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40828,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40828, "pid": 5, "tid": 7, "ts": 1716454218114001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051713, "dur": 10,
    "args": {
      "External id": 40828,
      "cbid": 211, "correlation": 40828
    }
  },
  {
    "ph": "s", "id": 40828, "pid": 76337, "tid": -914061504, "ts": 1716454218051713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051800, "dur": 1,
    "args": {
      "External id": 40844,
      "cbid": 251, "correlation": 40844
    }
  },
  {
    "ph": "f", "id": 40844, "pid": 76337, "tid": -914061504, "ts": 1716454218051800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051804, "dur": 0,
    "args": {
      "External id": 40846,
      "cbid": 251, "correlation": 40846
    }
  },
  {
    "ph": "f", "id": 40846, "pid": 76337, "tid": -914061504, "ts": 1716454218051804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218114019, "dur": 492,
    "args": {
      "External id": 40847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40847,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40847, "pid": 5, "tid": 7, "ts": 1716454218114019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051806, "dur": 13,
    "args": {
      "External id": 40847,
      "cbid": 211, "correlation": 40847
    }
  },
  {
    "ph": "s", "id": 40847, "pid": 76337, "tid": -914061504, "ts": 1716454218051806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218114513, "dur": 64,
    "args": {
      "External id": 40855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40855, "pid": 5, "tid": 7, "ts": 1716454218114513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051872, "dur": 13,
    "args": {
      "External id": 40855,
      "cbid": 211, "correlation": 40855
    }
  },
  {
    "ph": "s", "id": 40855, "pid": 76337, "tid": -914061504, "ts": 1716454218051872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218114578, "dur": 65,
    "args": {
      "External id": 40863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40863, "pid": 5, "tid": 7, "ts": 1716454218114578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051902, "dur": 8,
    "args": {
      "External id": 40863,
      "cbid": 211, "correlation": 40863
    }
  },
  {
    "ph": "s", "id": 40863, "pid": 76337, "tid": -914061504, "ts": 1716454218051902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051990, "dur": 1,
    "args": {
      "External id": 40879,
      "cbid": 251, "correlation": 40879
    }
  },
  {
    "ph": "f", "id": 40879, "pid": 76337, "tid": -914061504, "ts": 1716454218051990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218114646, "dur": 1,
    "args": {
      "External id": 40881,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 40881,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 40881, "pid": 5, "tid": 7, "ts": 1716454218114646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218051995, "dur": 9,
    "args": {
      "External id": 40881,
      "cbid": 51, "correlation": 40881
    }
  },
  {
    "ph": "s", "id": 40881, "pid": 76337, "tid": -914061504, "ts": 1716454218051995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218114649, "dur": 264,
    "args": {
      "External id": 40882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40882,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40882, "pid": 5, "tid": 7, "ts": 1716454218114649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052006, "dur": 12,
    "args": {
      "External id": 40882,
      "cbid": 211, "correlation": 40882
    }
  },
  {
    "ph": "s", "id": 40882, "pid": 76337, "tid": -914061504, "ts": 1716454218052006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218114915, "dur": 13,
    "args": {
      "External id": 40890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40890,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40890, "pid": 5, "tid": 7, "ts": 1716454218114915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052050, "dur": 10,
    "args": {
      "External id": 40890,
      "cbid": 211, "correlation": 40890
    }
  },
  {
    "ph": "s", "id": 40890, "pid": 76337, "tid": -914061504, "ts": 1716454218052050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218114930, "dur": 37,
    "args": {
      "External id": 40901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40901, "pid": 5, "tid": 7, "ts": 1716454218114930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052118, "dur": 13,
    "args": {
      "External id": 40901,
      "cbid": 211, "correlation": 40901
    }
  },
  {
    "ph": "s", "id": 40901, "pid": 76337, "tid": -914061504, "ts": 1716454218052118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052182, "dur": 0,
    "args": {
      "External id": 40913,
      "cbid": 317, "correlation": 40913
    }
  },
  {
    "ph": "f", "id": 40913, "pid": 76337, "tid": -914061504, "ts": 1716454218052182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052183, "dur": 0,
    "args": {
      "External id": 40914,
      "cbid": 203, "correlation": 40914
    }
  },
  {
    "ph": "f", "id": 40914, "pid": 76337, "tid": -914061504, "ts": 1716454218052183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052184, "dur": 0,
    "args": {
      "External id": 40915,
      "cbid": 205, "correlation": 40915
    }
  },
  {
    "ph": "f", "id": 40915, "pid": 76337, "tid": -914061504, "ts": 1716454218052184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218114968, "dur": 12,
    "args": {
      "External id": 40919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40919,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40919, "pid": 5, "tid": 7, "ts": 1716454218114968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052200, "dur": 12,
    "args": {
      "External id": 40919,
      "cbid": 211, "correlation": 40919
    }
  },
  {
    "ph": "s", "id": 40919, "pid": 76337, "tid": -914061504, "ts": 1716454218052200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218114982, "dur": 4,
    "args": {
      "External id": 40921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 40921, "pid": 5, "tid": 7, "ts": 1716454218114982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052216, "dur": 6,
    "args": {
      "External id": 40921,
      "cbid": 211, "correlation": 40921
    }
  },
  {
    "ph": "s", "id": 40921, "pid": 76337, "tid": -914061504, "ts": 1716454218052216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052224, "dur": 0,
    "args": {
      "External id": 40922,
      "cbid": 51, "correlation": 40922
    }
  },
  {
    "ph": "s", "id": 40922, "pid": 76337, "tid": -914061504, "ts": 1716454218052224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218114987, "dur": 94,
    "args": {
      "External id": 40923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40923,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 40923, "pid": 5, "tid": 7, "ts": 1716454218114987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052225, "dur": 5,
    "args": {
      "External id": 40923,
      "cbid": 211, "correlation": 40923
    }
  },
  {
    "ph": "s", "id": 40923, "pid": 76337, "tid": -914061504, "ts": 1716454218052225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218115082, "dur": 15,
    "args": {
      "External id": 40928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40928, "pid": 5, "tid": 7, "ts": 1716454218115082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052252, "dur": 9,
    "args": {
      "External id": 40928,
      "cbid": 211, "correlation": 40928
    }
  },
  {
    "ph": "s", "id": 40928, "pid": 76337, "tid": -914061504, "ts": 1716454218052252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218115099, "dur": 12,
    "args": {
      "External id": 40936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40936,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40936, "pid": 5, "tid": 7, "ts": 1716454218115099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052284, "dur": 9,
    "args": {
      "External id": 40936,
      "cbid": 211, "correlation": 40936
    }
  },
  {
    "ph": "s", "id": 40936, "pid": 76337, "tid": -914061504, "ts": 1716454218052284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218115112, "dur": 56,
    "args": {
      "External id": 40947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40947,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40947, "pid": 5, "tid": 7, "ts": 1716454218115112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052350, "dur": 13,
    "args": {
      "External id": 40947,
      "cbid": 211, "correlation": 40947
    }
  },
  {
    "ph": "s", "id": 40947, "pid": 76337, "tid": -914061504, "ts": 1716454218052350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052407, "dur": 0,
    "args": {
      "External id": 40957,
      "cbid": 317, "correlation": 40957
    }
  },
  {
    "ph": "f", "id": 40957, "pid": 76337, "tid": -914061504, "ts": 1716454218052407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052407, "dur": 0,
    "args": {
      "External id": 40958,
      "cbid": 203, "correlation": 40958
    }
  },
  {
    "ph": "f", "id": 40958, "pid": 76337, "tid": -914061504, "ts": 1716454218052407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052408, "dur": 0,
    "args": {
      "External id": 40959,
      "cbid": 205, "correlation": 40959
    }
  },
  {
    "ph": "f", "id": 40959, "pid": 76337, "tid": -914061504, "ts": 1716454218052408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218115169, "dur": 39,
    "args": {
      "External id": 40963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40963,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40963, "pid": 5, "tid": 7, "ts": 1716454218115169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052424, "dur": 12,
    "args": {
      "External id": 40963,
      "cbid": 211, "correlation": 40963
    }
  },
  {
    "ph": "s", "id": 40963, "pid": 76337, "tid": -914061504, "ts": 1716454218052424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218115209, "dur": 159,
    "args": {
      "External id": 40965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40965,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40965, "pid": 5, "tid": 7, "ts": 1716454218115209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052439, "dur": 6,
    "args": {
      "External id": 40965,
      "cbid": 211, "correlation": 40965
    }
  },
  {
    "ph": "s", "id": 40965, "pid": 76337, "tid": -914061504, "ts": 1716454218052439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218115369, "dur": 1958,
    "args": {
      "External id": 40967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40967,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 40967, "pid": 5, "tid": 7, "ts": 1716454218115369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052452, "dur": 8,
    "args": {
      "External id": 40967,
      "cbid": 211, "correlation": 40967
    }
  },
  {
    "ph": "s", "id": 40967, "pid": 76337, "tid": -914061504, "ts": 1716454218052452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218117329, "dur": 40,
    "args": {
      "External id": 40969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40969,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40969, "pid": 5, "tid": 7, "ts": 1716454218117329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052464, "dur": 5,
    "args": {
      "External id": 40969,
      "cbid": 211, "correlation": 40969
    }
  },
  {
    "ph": "s", "id": 40969, "pid": 76337, "tid": -914061504, "ts": 1716454218052464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218117370, "dur": 59,
    "args": {
      "External id": 40975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40975,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40975, "pid": 5, "tid": 7, "ts": 1716454218117370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052492, "dur": 8,
    "args": {
      "External id": 40975,
      "cbid": 211, "correlation": 40975
    }
  },
  {
    "ph": "s", "id": 40975, "pid": 76337, "tid": -914061504, "ts": 1716454218052492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218117429, "dur": 86,
    "args": {
      "External id": 40984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 40984,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 40984, "pid": 5, "tid": 7, "ts": 1716454218117429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052582, "dur": 13,
    "args": {
      "External id": 40984,
      "cbid": 211, "correlation": 40984
    }
  },
  {
    "ph": "s", "id": 40984, "pid": 76337, "tid": -914061504, "ts": 1716454218052582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218117516, "dur": 72,
    "args": {
      "External id": 41004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41004,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 41004, "pid": 5, "tid": 7, "ts": 1716454218117516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052651, "dur": 12,
    "args": {
      "External id": 41004,
      "cbid": 211, "correlation": 41004
    }
  },
  {
    "ph": "s", "id": 41004, "pid": 76337, "tid": -914061504, "ts": 1716454218052651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218117590, "dur": 5,
    "args": {
      "External id": 41016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41016,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 41016, "pid": 5, "tid": 7, "ts": 1716454218117590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052673, "dur": 6,
    "args": {
      "External id": 41016,
      "cbid": 211, "correlation": 41016
    }
  },
  {
    "ph": "s", "id": 41016, "pid": 76337, "tid": -914061504, "ts": 1716454218052673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218117596, "dur": 82,
    "args": {
      "External id": 41019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41019, "pid": 5, "tid": 7, "ts": 1716454218117596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052691, "dur": 7,
    "args": {
      "External id": 41019,
      "cbid": 211, "correlation": 41019
    }
  },
  {
    "ph": "s", "id": 41019, "pid": 76337, "tid": -914061504, "ts": 1716454218052691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218117679, "dur": 53,
    "args": {
      "External id": 41028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41028,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41028, "pid": 5, "tid": 7, "ts": 1716454218117679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052731, "dur": 9,
    "args": {
      "External id": 41028,
      "cbid": 211, "correlation": 41028
    }
  },
  {
    "ph": "s", "id": 41028, "pid": 76337, "tid": -914061504, "ts": 1716454218052731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052782, "dur": 0,
    "args": {
      "External id": 41038,
      "cbid": 317, "correlation": 41038
    }
  },
  {
    "ph": "f", "id": 41038, "pid": 76337, "tid": -914061504, "ts": 1716454218052782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052783, "dur": 0,
    "args": {
      "External id": 41039,
      "cbid": 203, "correlation": 41039
    }
  },
  {
    "ph": "f", "id": 41039, "pid": 76337, "tid": -914061504, "ts": 1716454218052783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052784, "dur": 0,
    "args": {
      "External id": 41040,
      "cbid": 205, "correlation": 41040
    }
  },
  {
    "ph": "f", "id": 41040, "pid": 76337, "tid": -914061504, "ts": 1716454218052784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218117733, "dur": 56,
    "args": {
      "External id": 41044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41044,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41044, "pid": 5, "tid": 7, "ts": 1716454218117733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052800, "dur": 12,
    "args": {
      "External id": 41044,
      "cbid": 211, "correlation": 41044
    }
  },
  {
    "ph": "s", "id": 41044, "pid": 76337, "tid": -914061504, "ts": 1716454218052800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218117790, "dur": 120,
    "args": {
      "External id": 41046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41046,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41046, "pid": 5, "tid": 7, "ts": 1716454218117790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052815, "dur": 5,
    "args": {
      "External id": 41046,
      "cbid": 211, "correlation": 41046
    }
  },
  {
    "ph": "s", "id": 41046, "pid": 76337, "tid": -914061504, "ts": 1716454218052815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218117912, "dur": 1856,
    "args": {
      "External id": 41048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41048,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41048, "pid": 5, "tid": 7, "ts": 1716454218117912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052826, "dur": 6,
    "args": {
      "External id": 41048,
      "cbid": 211, "correlation": 41048
    }
  },
  {
    "ph": "s", "id": 41048, "pid": 76337, "tid": -914061504, "ts": 1716454218052826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218119769, "dur": 19,
    "args": {
      "External id": 41050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41050, "pid": 5, "tid": 7, "ts": 1716454218119769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052835, "dur": 5,
    "args": {
      "External id": 41050,
      "cbid": 211, "correlation": 41050
    }
  },
  {
    "ph": "s", "id": 41050, "pid": 76337, "tid": -914061504, "ts": 1716454218052835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218119789, "dur": 32,
    "args": {
      "External id": 41056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41056,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41056, "pid": 5, "tid": 7, "ts": 1716454218119789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052864, "dur": 8,
    "args": {
      "External id": 41056,
      "cbid": 211, "correlation": 41056
    }
  },
  {
    "ph": "s", "id": 41056, "pid": 76337, "tid": -914061504, "ts": 1716454218052864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218119822, "dur": 3,
    "args": {
      "External id": 41064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41064,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 41064, "pid": 5, "tid": 7, "ts": 1716454218119822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052906, "dur": 11,
    "args": {
      "External id": 41064,
      "cbid": 211, "correlation": 41064
    }
  },
  {
    "ph": "s", "id": 41064, "pid": 76337, "tid": -914061504, "ts": 1716454218052906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052985, "dur": 1,
    "args": {
      "External id": 41080,
      "cbid": 251, "correlation": 41080
    }
  },
  {
    "ph": "f", "id": 41080, "pid": 76337, "tid": -914061504, "ts": 1716454218052985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052990, "dur": 0,
    "args": {
      "External id": 41082,
      "cbid": 251, "correlation": 41082
    }
  },
  {
    "ph": "f", "id": 41082, "pid": 76337, "tid": -914061504, "ts": 1716454218052990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218119827, "dur": 12,
    "args": {
      "External id": 41083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41083,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 41083, "pid": 5, "tid": 7, "ts": 1716454218119827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218052992, "dur": 12,
    "args": {
      "External id": 41083,
      "cbid": 211, "correlation": 41083
    }
  },
  {
    "ph": "s", "id": 41083, "pid": 76337, "tid": -914061504, "ts": 1716454218052992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218119841, "dur": 5,
    "args": {
      "External id": 41085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41085,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 41085, "pid": 5, "tid": 7, "ts": 1716454218119841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053006, "dur": 6,
    "args": {
      "External id": 41085,
      "cbid": 211, "correlation": 41085
    }
  },
  {
    "ph": "s", "id": 41085, "pid": 76337, "tid": -914061504, "ts": 1716454218053006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218119847, "dur": 29,
    "args": {
      "External id": 41095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41095,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41095, "pid": 5, "tid": 7, "ts": 1716454218119847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053066, "dur": 12,
    "args": {
      "External id": 41095,
      "cbid": 211, "correlation": 41095
    }
  },
  {
    "ph": "s", "id": 41095, "pid": 76337, "tid": -914061504, "ts": 1716454218053066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218119877, "dur": 30,
    "args": {
      "External id": 41115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41115,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 41115, "pid": 5, "tid": 7, "ts": 1716454218119877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053132, "dur": 11,
    "args": {
      "External id": 41115,
      "cbid": 211, "correlation": 41115
    }
  },
  {
    "ph": "s", "id": 41115, "pid": 76337, "tid": -914061504, "ts": 1716454218053132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218119909, "dur": 4,
    "args": {
      "External id": 41127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41127,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 41127, "pid": 5, "tid": 7, "ts": 1716454218119909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053153, "dur": 6,
    "args": {
      "External id": 41127,
      "cbid": 211, "correlation": 41127
    }
  },
  {
    "ph": "s", "id": 41127, "pid": 76337, "tid": -914061504, "ts": 1716454218053153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218119914, "dur": 30,
    "args": {
      "External id": 41130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41130, "pid": 5, "tid": 7, "ts": 1716454218119914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053172, "dur": 7,
    "args": {
      "External id": 41130,
      "cbid": 211, "correlation": 41130
    }
  },
  {
    "ph": "s", "id": 41130, "pid": 76337, "tid": -914061504, "ts": 1716454218053172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218119946, "dur": 20,
    "args": {
      "External id": 41139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41139,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41139, "pid": 5, "tid": 7, "ts": 1716454218119946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053211, "dur": 10,
    "args": {
      "External id": 41139,
      "cbid": 211, "correlation": 41139
    }
  },
  {
    "ph": "s", "id": 41139, "pid": 76337, "tid": -914061504, "ts": 1716454218053211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053275, "dur": 0,
    "args": {
      "External id": 41149,
      "cbid": 317, "correlation": 41149
    }
  },
  {
    "ph": "f", "id": 41149, "pid": 76337, "tid": -914061504, "ts": 1716454218053275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053276, "dur": 0,
    "args": {
      "External id": 41150,
      "cbid": 203, "correlation": 41150
    }
  },
  {
    "ph": "f", "id": 41150, "pid": 76337, "tid": -914061504, "ts": 1716454218053276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053276, "dur": 0,
    "args": {
      "External id": 41151,
      "cbid": 205, "correlation": 41151
    }
  },
  {
    "ph": "f", "id": 41151, "pid": 76337, "tid": -914061504, "ts": 1716454218053276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218119967, "dur": 22,
    "args": {
      "External id": 41155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41155,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41155, "pid": 5, "tid": 7, "ts": 1716454218119967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053291, "dur": 12,
    "args": {
      "External id": 41155,
      "cbid": 211, "correlation": 41155
    }
  },
  {
    "ph": "s", "id": 41155, "pid": 76337, "tid": -914061504, "ts": 1716454218053291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218119990, "dur": 43,
    "args": {
      "External id": 41157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41157,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41157, "pid": 5, "tid": 7, "ts": 1716454218119990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053306, "dur": 5,
    "args": {
      "External id": 41157,
      "cbid": 211, "correlation": 41157
    }
  },
  {
    "ph": "s", "id": 41157, "pid": 76337, "tid": -914061504, "ts": 1716454218053306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218120034, "dur": 635,
    "args": {
      "External id": 41159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41159,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41159, "pid": 5, "tid": 7, "ts": 1716454218120034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053318, "dur": 6,
    "args": {
      "External id": 41159,
      "cbid": 211, "correlation": 41159
    }
  },
  {
    "ph": "s", "id": 41159, "pid": 76337, "tid": -914061504, "ts": 1716454218053318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218120670, "dur": 21,
    "args": {
      "External id": 41161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41161,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41161, "pid": 5, "tid": 7, "ts": 1716454218120670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053328, "dur": 5,
    "args": {
      "External id": 41161,
      "cbid": 211, "correlation": 41161
    }
  },
  {
    "ph": "s", "id": 41161, "pid": 76337, "tid": -914061504, "ts": 1716454218053328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218120692, "dur": 32,
    "args": {
      "External id": 41167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41167, "pid": 5, "tid": 7, "ts": 1716454218120692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053355, "dur": 8,
    "args": {
      "External id": 41167,
      "cbid": 211, "correlation": 41167
    }
  },
  {
    "ph": "s", "id": 41167, "pid": 76337, "tid": -914061504, "ts": 1716454218053355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053412, "dur": 0,
    "args": {
      "External id": 41177,
      "cbid": 317, "correlation": 41177
    }
  },
  {
    "ph": "f", "id": 41177, "pid": 76337, "tid": -914061504, "ts": 1716454218053412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053413, "dur": 0,
    "args": {
      "External id": 41178,
      "cbid": 203, "correlation": 41178
    }
  },
  {
    "ph": "f", "id": 41178, "pid": 76337, "tid": -914061504, "ts": 1716454218053413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053414, "dur": 0,
    "args": {
      "External id": 41179,
      "cbid": 205, "correlation": 41179
    }
  },
  {
    "ph": "f", "id": 41179, "pid": 76337, "tid": -914061504, "ts": 1716454218053414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218120726, "dur": 57,
    "args": {
      "External id": 41183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41183,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41183, "pid": 5, "tid": 7, "ts": 1716454218120726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053427, "dur": 12,
    "args": {
      "External id": 41183,
      "cbid": 211, "correlation": 41183
    }
  },
  {
    "ph": "s", "id": 41183, "pid": 76337, "tid": -914061504, "ts": 1716454218053427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218120784, "dur": 263,
    "args": {
      "External id": 41185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41185,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41185, "pid": 5, "tid": 7, "ts": 1716454218120784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053445, "dur": 8,
    "args": {
      "External id": 41185,
      "cbid": 211, "correlation": 41185
    }
  },
  {
    "ph": "s", "id": 41185, "pid": 76337, "tid": -914061504, "ts": 1716454218053445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218121048, "dur": 21,
    "args": {
      "External id": 41187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41187,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41187, "pid": 5, "tid": 7, "ts": 1716454218121048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053456, "dur": 5,
    "args": {
      "External id": 41187,
      "cbid": 211, "correlation": 41187
    }
  },
  {
    "ph": "s", "id": 41187, "pid": 76337, "tid": -914061504, "ts": 1716454218053456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218121070, "dur": 32,
    "args": {
      "External id": 41193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41193, "pid": 5, "tid": 7, "ts": 1716454218121070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053483, "dur": 8,
    "args": {
      "External id": 41193,
      "cbid": 211, "correlation": 41193
    }
  },
  {
    "ph": "s", "id": 41193, "pid": 76337, "tid": -914061504, "ts": 1716454218053483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218121103, "dur": 27,
    "args": {
      "External id": 41201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41201,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41201, "pid": 5, "tid": 7, "ts": 1716454218121103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053511, "dur": 8,
    "args": {
      "External id": 41201,
      "cbid": 211, "correlation": 41201
    }
  },
  {
    "ph": "s", "id": 41201, "pid": 76337, "tid": -914061504, "ts": 1716454218053511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218121132, "dur": 20,
    "args": {
      "External id": 41209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41209,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41209, "pid": 5, "tid": 7, "ts": 1716454218121132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053541, "dur": 8,
    "args": {
      "External id": 41209,
      "cbid": 211, "correlation": 41209
    }
  },
  {
    "ph": "s", "id": 41209, "pid": 76337, "tid": -914061504, "ts": 1716454218053541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218121153, "dur": 30,
    "args": {
      "External id": 41229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41229,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 41229, "pid": 5, "tid": 7, "ts": 1716454218121153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053623, "dur": 13,
    "args": {
      "External id": 41229,
      "cbid": 211, "correlation": 41229
    }
  },
  {
    "ph": "s", "id": 41229, "pid": 76337, "tid": -914061504, "ts": 1716454218053623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218121184, "dur": 4,
    "args": {
      "External id": 41241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41241,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 41241, "pid": 5, "tid": 7, "ts": 1716454218121184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053645, "dur": 6,
    "args": {
      "External id": 41241,
      "cbid": 211, "correlation": 41241
    }
  },
  {
    "ph": "s", "id": 41241, "pid": 76337, "tid": -914061504, "ts": 1716454218053645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218121190, "dur": 30,
    "args": {
      "External id": 41244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41244, "pid": 5, "tid": 7, "ts": 1716454218121190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053663, "dur": 6,
    "args": {
      "External id": 41244,
      "cbid": 211, "correlation": 41244
    }
  },
  {
    "ph": "s", "id": 41244, "pid": 76337, "tid": -914061504, "ts": 1716454218053663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053719, "dur": 0,
    "args": {
      "External id": 41255,
      "cbid": 317, "correlation": 41255
    }
  },
  {
    "ph": "f", "id": 41255, "pid": 76337, "tid": -914061504, "ts": 1716454218053719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053720, "dur": 0,
    "args": {
      "External id": 41256,
      "cbid": 203, "correlation": 41256
    }
  },
  {
    "ph": "f", "id": 41256, "pid": 76337, "tid": -914061504, "ts": 1716454218053720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053721, "dur": 0,
    "args": {
      "External id": 41257,
      "cbid": 205, "correlation": 41257
    }
  },
  {
    "ph": "f", "id": 41257, "pid": 76337, "tid": -914061504, "ts": 1716454218053721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218121221, "dur": 21,
    "args": {
      "External id": 41261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41261,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41261, "pid": 5, "tid": 7, "ts": 1716454218121221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053735, "dur": 12,
    "args": {
      "External id": 41261,
      "cbid": 211, "correlation": 41261
    }
  },
  {
    "ph": "s", "id": 41261, "pid": 76337, "tid": -914061504, "ts": 1716454218053735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218121244, "dur": 103,
    "args": {
      "External id": 41263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41263,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41263, "pid": 5, "tid": 7, "ts": 1716454218121244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053752, "dur": 6,
    "args": {
      "External id": 41263,
      "cbid": 211, "correlation": 41263
    }
  },
  {
    "ph": "s", "id": 41263, "pid": 76337, "tid": -914061504, "ts": 1716454218053752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218121348, "dur": 22,
    "args": {
      "External id": 41265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41265,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41265, "pid": 5, "tid": 7, "ts": 1716454218121348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053762, "dur": 5,
    "args": {
      "External id": 41265,
      "cbid": 211, "correlation": 41265
    }
  },
  {
    "ph": "s", "id": 41265, "pid": 76337, "tid": -914061504, "ts": 1716454218053762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218121372, "dur": 31,
    "args": {
      "External id": 41271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41271, "pid": 5, "tid": 7, "ts": 1716454218121372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053790, "dur": 8,
    "args": {
      "External id": 41271,
      "cbid": 211, "correlation": 41271
    }
  },
  {
    "ph": "s", "id": 41271, "pid": 76337, "tid": -914061504, "ts": 1716454218053790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218121404, "dur": 181,
    "args": {
      "External id": 41280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41280,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41280, "pid": 5, "tid": 7, "ts": 1716454218121404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053872, "dur": 15,
    "args": {
      "External id": 41280,
      "cbid": 211, "correlation": 41280
    }
  },
  {
    "ph": "s", "id": 41280, "pid": 76337, "tid": -914061504, "ts": 1716454218053872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218121586, "dur": 63,
    "args": {
      "External id": 41302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41302,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41302, "pid": 5, "tid": 7, "ts": 1716454218121586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218053929, "dur": 10,
    "args": {
      "External id": 41302,
      "cbid": 211, "correlation": 41302
    }
  },
  {
    "ph": "s", "id": 41302, "pid": 76337, "tid": -914061504, "ts": 1716454218053929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054026, "dur": 1,
    "args": {
      "External id": 41313,
      "cbid": 251, "correlation": 41313
    }
  },
  {
    "ph": "f", "id": 41313, "pid": 76337, "tid": -914061504, "ts": 1716454218054026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218121651, "dur": 151,
    "args": {
      "External id": 41314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41314,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41314, "pid": 5, "tid": 7, "ts": 1716454218121651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054032, "dur": 14,
    "args": {
      "External id": 41314,
      "cbid": 211, "correlation": 41314
    }
  },
  {
    "ph": "s", "id": 41314, "pid": 76337, "tid": -914061504, "ts": 1716454218054032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054103, "dur": 1,
    "args": {
      "External id": 41325,
      "cbid": 251, "correlation": 41325
    }
  },
  {
    "ph": "f", "id": 41325, "pid": 76337, "tid": -914061504, "ts": 1716454218054103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218121803, "dur": 141,
    "args": {
      "External id": 41326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41326,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41326, "pid": 5, "tid": 7, "ts": 1716454218121803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054107, "dur": 11,
    "args": {
      "External id": 41326,
      "cbid": 211, "correlation": 41326
    }
  },
  {
    "ph": "s", "id": 41326, "pid": 76337, "tid": -914061504, "ts": 1716454218054107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054172, "dur": 1,
    "args": {
      "External id": 41337,
      "cbid": 251, "correlation": 41337
    }
  },
  {
    "ph": "f", "id": 41337, "pid": 76337, "tid": -914061504, "ts": 1716454218054172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218121945, "dur": 143,
    "args": {
      "External id": 41338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41338,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41338, "pid": 5, "tid": 7, "ts": 1716454218121945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054175, "dur": 11,
    "args": {
      "External id": 41338,
      "cbid": 211, "correlation": 41338
    }
  },
  {
    "ph": "s", "id": 41338, "pid": 76337, "tid": -914061504, "ts": 1716454218054175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218122089, "dur": 1890,
    "args": {
      "External id": 41359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41359,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 41359, "pid": 5, "tid": 7, "ts": 1716454218122089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054258, "dur": 13,
    "args": {
      "External id": 41359,
      "cbid": 211, "correlation": 41359
    }
  },
  {
    "ph": "s", "id": 41359, "pid": 76337, "tid": -914061504, "ts": 1716454218054258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054358, "dur": 2,
    "args": {
      "External id": 41377,
      "cbid": 251, "correlation": 41377
    }
  },
  {
    "ph": "f", "id": 41377, "pid": 76337, "tid": -914061504, "ts": 1716454218054358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218123981, "dur": 145,
    "args": {
      "External id": 41379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41379,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 41379, "pid": 5, "tid": 7, "ts": 1716454218123981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054364, "dur": 13,
    "args": {
      "External id": 41379,
      "cbid": 211, "correlation": 41379
    }
  },
  {
    "ph": "s", "id": 41379, "pid": 76337, "tid": -914061504, "ts": 1716454218054364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218124127, "dur": 35,
    "args": {
      "External id": 41387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41387,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41387, "pid": 5, "tid": 7, "ts": 1716454218124127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054434, "dur": 12,
    "args": {
      "External id": 41387,
      "cbid": 211, "correlation": 41387
    }
  },
  {
    "ph": "s", "id": 41387, "pid": 76337, "tid": -914061504, "ts": 1716454218054434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218124163, "dur": 51,
    "args": {
      "External id": 41395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41395,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41395, "pid": 5, "tid": 7, "ts": 1716454218124163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054472, "dur": 8,
    "args": {
      "External id": 41395,
      "cbid": 211, "correlation": 41395
    }
  },
  {
    "ph": "s", "id": 41395, "pid": 76337, "tid": -914061504, "ts": 1716454218054472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218124215, "dur": 30,
    "args": {
      "External id": 41406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41406,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41406, "pid": 5, "tid": 7, "ts": 1716454218124215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054544, "dur": 13,
    "args": {
      "External id": 41406,
      "cbid": 211, "correlation": 41406
    }
  },
  {
    "ph": "s", "id": 41406, "pid": 76337, "tid": -914061504, "ts": 1716454218054544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218124246, "dur": 34,
    "args": {
      "External id": 41428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41428,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41428, "pid": 5, "tid": 7, "ts": 1716454218124246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054576, "dur": 7,
    "args": {
      "External id": 41428,
      "cbid": 211, "correlation": 41428
    }
  },
  {
    "ph": "s", "id": 41428, "pid": 76337, "tid": -914061504, "ts": 1716454218054576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054659, "dur": 1,
    "args": {
      "External id": 41439,
      "cbid": 251, "correlation": 41439
    }
  },
  {
    "ph": "f", "id": 41439, "pid": 76337, "tid": -914061504, "ts": 1716454218054659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218124282, "dur": 89,
    "args": {
      "External id": 41440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41440,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41440, "pid": 5, "tid": 7, "ts": 1716454218124282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054665, "dur": 14,
    "args": {
      "External id": 41440,
      "cbid": 211, "correlation": 41440
    }
  },
  {
    "ph": "s", "id": 41440, "pid": 76337, "tid": -914061504, "ts": 1716454218054665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054735, "dur": 1,
    "args": {
      "External id": 41451,
      "cbid": 251, "correlation": 41451
    }
  },
  {
    "ph": "f", "id": 41451, "pid": 76337, "tid": -914061504, "ts": 1716454218054735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054738, "dur": 0,
    "args": {
      "External id": 41452,
      "cbid": 251, "correlation": 41452
    }
  },
  {
    "ph": "f", "id": 41452, "pid": 76337, "tid": -914061504, "ts": 1716454218054738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218124372, "dur": 11,
    "args": {
      "External id": 41453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41453,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 41453, "pid": 5, "tid": 7, "ts": 1716454218124372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054740, "dur": 12,
    "args": {
      "External id": 41453,
      "cbid": 211, "correlation": 41453
    }
  },
  {
    "ph": "s", "id": 41453, "pid": 76337, "tid": -914061504, "ts": 1716454218054740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218124383, "dur": 5,
    "args": {
      "External id": 41455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41455,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 41455, "pid": 5, "tid": 7, "ts": 1716454218124383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054753, "dur": 6,
    "args": {
      "External id": 41455,
      "cbid": 211, "correlation": 41455
    }
  },
  {
    "ph": "s", "id": 41455, "pid": 76337, "tid": -914061504, "ts": 1716454218054753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054811, "dur": 1,
    "args": {
      "External id": 41466,
      "cbid": 251, "correlation": 41466
    }
  },
  {
    "ph": "f", "id": 41466, "pid": 76337, "tid": -914061504, "ts": 1716454218054811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054814, "dur": 0,
    "args": {
      "External id": 41467,
      "cbid": 251, "correlation": 41467
    }
  },
  {
    "ph": "f", "id": 41467, "pid": 76337, "tid": -914061504, "ts": 1716454218054814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218124390, "dur": 7,
    "args": {
      "External id": 41468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41468,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 41468, "pid": 5, "tid": 7, "ts": 1716454218124390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054815, "dur": 11,
    "args": {
      "External id": 41468,
      "cbid": 211, "correlation": 41468
    }
  },
  {
    "ph": "s", "id": 41468, "pid": 76337, "tid": -914061504, "ts": 1716454218054815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218124398, "dur": 3,
    "args": {
      "External id": 41470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41470,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 41470, "pid": 5, "tid": 7, "ts": 1716454218124398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054828, "dur": 6,
    "args": {
      "External id": 41470,
      "cbid": 211, "correlation": 41470
    }
  },
  {
    "ph": "s", "id": 41470, "pid": 76337, "tid": -914061504, "ts": 1716454218054828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218124402, "dur": 90,
    "args": {
      "External id": 41491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41491,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 41491, "pid": 5, "tid": 7, "ts": 1716454218124402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218054901, "dur": 12,
    "args": {
      "External id": 41491,
      "cbid": 211, "correlation": 41491
    }
  },
  {
    "ph": "s", "id": 41491, "pid": 76337, "tid": -914061504, "ts": 1716454218054901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055004, "dur": 1,
    "args": {
      "External id": 41509,
      "cbid": 251, "correlation": 41509
    }
  },
  {
    "ph": "f", "id": 41509, "pid": 76337, "tid": -914061504, "ts": 1716454218055004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218124494, "dur": 96,
    "args": {
      "External id": 41511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41511,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41511, "pid": 5, "tid": 7, "ts": 1716454218124494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055010, "dur": 13,
    "args": {
      "External id": 41511,
      "cbid": 211, "correlation": 41511
    }
  },
  {
    "ph": "s", "id": 41511, "pid": 76337, "tid": -914061504, "ts": 1716454218055010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218124591, "dur": 19,
    "args": {
      "External id": 41519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41519,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41519, "pid": 5, "tid": 7, "ts": 1716454218124591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055080, "dur": 12,
    "args": {
      "External id": 41519,
      "cbid": 211, "correlation": 41519
    }
  },
  {
    "ph": "s", "id": 41519, "pid": 76337, "tid": -914061504, "ts": 1716454218055080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218124611, "dur": 37,
    "args": {
      "External id": 41527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41527,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41527, "pid": 5, "tid": 7, "ts": 1716454218124611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055121, "dur": 9,
    "args": {
      "External id": 41527,
      "cbid": 211, "correlation": 41527
    }
  },
  {
    "ph": "s", "id": 41527, "pid": 76337, "tid": -914061504, "ts": 1716454218055121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218124649, "dur": 34,
    "args": {
      "External id": 41549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41549,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41549, "pid": 5, "tid": 7, "ts": 1716454218124649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055172, "dur": 10,
    "args": {
      "External id": 41549,
      "cbid": 211, "correlation": 41549
    }
  },
  {
    "ph": "s", "id": 41549, "pid": 76337, "tid": -914061504, "ts": 1716454218055172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055263, "dur": 1,
    "args": {
      "External id": 41565,
      "cbid": 251, "correlation": 41565
    }
  },
  {
    "ph": "f", "id": 41565, "pid": 76337, "tid": -914061504, "ts": 1716454218055263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055268, "dur": 0,
    "args": {
      "External id": 41567,
      "cbid": 251, "correlation": 41567
    }
  },
  {
    "ph": "f", "id": 41567, "pid": 76337, "tid": -914061504, "ts": 1716454218055268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218124685, "dur": 529,
    "args": {
      "External id": 41568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41568,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 41568, "pid": 5, "tid": 7, "ts": 1716454218124685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055273, "dur": 13,
    "args": {
      "External id": 41568,
      "cbid": 211, "correlation": 41568
    }
  },
  {
    "ph": "s", "id": 41568, "pid": 76337, "tid": -914061504, "ts": 1716454218055273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218125215, "dur": 123,
    "args": {
      "External id": 41576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41576, "pid": 5, "tid": 7, "ts": 1716454218125215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055338, "dur": 13,
    "args": {
      "External id": 41576,
      "cbid": 211, "correlation": 41576
    }
  },
  {
    "ph": "s", "id": 41576, "pid": 76337, "tid": -914061504, "ts": 1716454218055338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218125340, "dur": 127,
    "args": {
      "External id": 41584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41584,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41584, "pid": 5, "tid": 7, "ts": 1716454218125340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055369, "dur": 8,
    "args": {
      "External id": 41584,
      "cbid": 211, "correlation": 41584
    }
  },
  {
    "ph": "s", "id": 41584, "pid": 76337, "tid": -914061504, "ts": 1716454218055369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055445, "dur": 1,
    "args": {
      "External id": 41600,
      "cbid": 251, "correlation": 41600
    }
  },
  {
    "ph": "f", "id": 41600, "pid": 76337, "tid": -914061504, "ts": 1716454218055445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218125468, "dur": 302,
    "args": {
      "External id": 41602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41602,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41602, "pid": 5, "tid": 7, "ts": 1716454218125468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055451, "dur": 13,
    "args": {
      "External id": 41602,
      "cbid": 211, "correlation": 41602
    }
  },
  {
    "ph": "s", "id": 41602, "pid": 76337, "tid": -914061504, "ts": 1716454218055451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218125771, "dur": 28,
    "args": {
      "External id": 41610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41610,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41610, "pid": 5, "tid": 7, "ts": 1716454218125771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055492, "dur": 10,
    "args": {
      "External id": 41610,
      "cbid": 211, "correlation": 41610
    }
  },
  {
    "ph": "s", "id": 41610, "pid": 76337, "tid": -914061504, "ts": 1716454218055492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218125800, "dur": 79,
    "args": {
      "External id": 41621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41621, "pid": 5, "tid": 7, "ts": 1716454218125800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055560, "dur": 13,
    "args": {
      "External id": 41621,
      "cbid": 211, "correlation": 41621
    }
  },
  {
    "ph": "s", "id": 41621, "pid": 76337, "tid": -914061504, "ts": 1716454218055560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055624, "dur": 0,
    "args": {
      "External id": 41633,
      "cbid": 317, "correlation": 41633
    }
  },
  {
    "ph": "f", "id": 41633, "pid": 76337, "tid": -914061504, "ts": 1716454218055624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055625, "dur": 0,
    "args": {
      "External id": 41634,
      "cbid": 203, "correlation": 41634
    }
  },
  {
    "ph": "f", "id": 41634, "pid": 76337, "tid": -914061504, "ts": 1716454218055625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055625, "dur": 0,
    "args": {
      "External id": 41635,
      "cbid": 205, "correlation": 41635
    }
  },
  {
    "ph": "f", "id": 41635, "pid": 76337, "tid": -914061504, "ts": 1716454218055625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218125880, "dur": 22,
    "args": {
      "External id": 41639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41639,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41639, "pid": 5, "tid": 7, "ts": 1716454218125880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055640, "dur": 12,
    "args": {
      "External id": 41639,
      "cbid": 211, "correlation": 41639
    }
  },
  {
    "ph": "s", "id": 41639, "pid": 76337, "tid": -914061504, "ts": 1716454218055640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218125903, "dur": 118,
    "args": {
      "External id": 41641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41641,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41641, "pid": 5, "tid": 7, "ts": 1716454218125903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055658, "dur": 6,
    "args": {
      "External id": 41641,
      "cbid": 211, "correlation": 41641
    }
  },
  {
    "ph": "s", "id": 41641, "pid": 76337, "tid": -914061504, "ts": 1716454218055658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218126023, "dur": 23,
    "args": {
      "External id": 41643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41643,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41643, "pid": 5, "tid": 7, "ts": 1716454218126023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055668, "dur": 6,
    "args": {
      "External id": 41643,
      "cbid": 211, "correlation": 41643
    }
  },
  {
    "ph": "s", "id": 41643, "pid": 76337, "tid": -914061504, "ts": 1716454218055668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218126047, "dur": 32,
    "args": {
      "External id": 41649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41649,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41649, "pid": 5, "tid": 7, "ts": 1716454218126047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055696, "dur": 8,
    "args": {
      "External id": 41649,
      "cbid": 211, "correlation": 41649
    }
  },
  {
    "ph": "s", "id": 41649, "pid": 76337, "tid": -914061504, "ts": 1716454218055696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218126080, "dur": 27,
    "args": {
      "External id": 41657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41657,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41657, "pid": 5, "tid": 7, "ts": 1716454218126080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055728, "dur": 8,
    "args": {
      "External id": 41657,
      "cbid": 211, "correlation": 41657
    }
  },
  {
    "ph": "s", "id": 41657, "pid": 76337, "tid": -914061504, "ts": 1716454218055728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218126108, "dur": 52,
    "args": {
      "External id": 41666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41666,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41666, "pid": 5, "tid": 7, "ts": 1716454218126108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055765, "dur": 10,
    "args": {
      "External id": 41666,
      "cbid": 211, "correlation": 41666
    }
  },
  {
    "ph": "s", "id": 41666, "pid": 76337, "tid": -914061504, "ts": 1716454218055765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218126162, "dur": 50,
    "args": {
      "External id": 41686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41686,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 41686, "pid": 5, "tid": 7, "ts": 1716454218126162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055837, "dur": 12,
    "args": {
      "External id": 41686,
      "cbid": 211, "correlation": 41686
    }
  },
  {
    "ph": "s", "id": 41686, "pid": 76337, "tid": -914061504, "ts": 1716454218055837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218126213, "dur": 4,
    "args": {
      "External id": 41698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41698,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 41698, "pid": 5, "tid": 7, "ts": 1716454218126213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055859, "dur": 6,
    "args": {
      "External id": 41698,
      "cbid": 211, "correlation": 41698
    }
  },
  {
    "ph": "s", "id": 41698, "pid": 76337, "tid": -914061504, "ts": 1716454218055859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218126219, "dur": 56,
    "args": {
      "External id": 41701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41701, "pid": 5, "tid": 7, "ts": 1716454218126219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055877, "dur": 7,
    "args": {
      "External id": 41701,
      "cbid": 211, "correlation": 41701
    }
  },
  {
    "ph": "s", "id": 41701, "pid": 76337, "tid": -914061504, "ts": 1716454218055877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218126276, "dur": 37,
    "args": {
      "External id": 41710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41710,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41710, "pid": 5, "tid": 7, "ts": 1716454218126276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055917, "dur": 10,
    "args": {
      "External id": 41710,
      "cbid": 211, "correlation": 41710
    }
  },
  {
    "ph": "s", "id": 41710, "pid": 76337, "tid": -914061504, "ts": 1716454218055917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055969, "dur": 0,
    "args": {
      "External id": 41720,
      "cbid": 317, "correlation": 41720
    }
  },
  {
    "ph": "f", "id": 41720, "pid": 76337, "tid": -914061504, "ts": 1716454218055969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055970, "dur": 0,
    "args": {
      "External id": 41721,
      "cbid": 203, "correlation": 41721
    }
  },
  {
    "ph": "f", "id": 41721, "pid": 76337, "tid": -914061504, "ts": 1716454218055970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055971, "dur": 0,
    "args": {
      "External id": 41722,
      "cbid": 205, "correlation": 41722
    }
  },
  {
    "ph": "f", "id": 41722, "pid": 76337, "tid": -914061504, "ts": 1716454218055971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218126315, "dur": 39,
    "args": {
      "External id": 41726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41726, "pid": 5, "tid": 7, "ts": 1716454218126315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218055994, "dur": 12,
    "args": {
      "External id": 41726,
      "cbid": 211, "correlation": 41726
    }
  },
  {
    "ph": "s", "id": 41726, "pid": 76337, "tid": -914061504, "ts": 1716454218055994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218126355, "dur": 81,
    "args": {
      "External id": 41728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41728,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41728, "pid": 5, "tid": 7, "ts": 1716454218126355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056009, "dur": 5,
    "args": {
      "External id": 41728,
      "cbid": 211, "correlation": 41728
    }
  },
  {
    "ph": "s", "id": 41728, "pid": 76337, "tid": -914061504, "ts": 1716454218056009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218126438, "dur": 1257,
    "args": {
      "External id": 41730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41730,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41730, "pid": 5, "tid": 7, "ts": 1716454218126438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056020, "dur": 6,
    "args": {
      "External id": 41730,
      "cbid": 211, "correlation": 41730
    }
  },
  {
    "ph": "s", "id": 41730, "pid": 76337, "tid": -914061504, "ts": 1716454218056020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218127697, "dur": 21,
    "args": {
      "External id": 41732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41732, "pid": 5, "tid": 7, "ts": 1716454218127697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056029, "dur": 5,
    "args": {
      "External id": 41732,
      "cbid": 211, "correlation": 41732
    }
  },
  {
    "ph": "s", "id": 41732, "pid": 76337, "tid": -914061504, "ts": 1716454218056029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218127719, "dur": 32,
    "args": {
      "External id": 41738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41738, "pid": 5, "tid": 7, "ts": 1716454218127719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056058, "dur": 8,
    "args": {
      "External id": 41738,
      "cbid": 211, "correlation": 41738
    }
  },
  {
    "ph": "s", "id": 41738, "pid": 76337, "tid": -914061504, "ts": 1716454218056058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218127753, "dur": 3,
    "args": {
      "External id": 41746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41746,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 41746, "pid": 5, "tid": 7, "ts": 1716454218127753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056101, "dur": 10,
    "args": {
      "External id": 41746,
      "cbid": 211, "correlation": 41746
    }
  },
  {
    "ph": "s", "id": 41746, "pid": 76337, "tid": -914061504, "ts": 1716454218056101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056168, "dur": 1,
    "args": {
      "External id": 41762,
      "cbid": 251, "correlation": 41762
    }
  },
  {
    "ph": "f", "id": 41762, "pid": 76337, "tid": -914061504, "ts": 1716454218056168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056173, "dur": 0,
    "args": {
      "External id": 41764,
      "cbid": 251, "correlation": 41764
    }
  },
  {
    "ph": "f", "id": 41764, "pid": 76337, "tid": -914061504, "ts": 1716454218056173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218127757, "dur": 12,
    "args": {
      "External id": 41765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41765,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 41765, "pid": 5, "tid": 7, "ts": 1716454218127757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056175, "dur": 11,
    "args": {
      "External id": 41765,
      "cbid": 211, "correlation": 41765
    }
  },
  {
    "ph": "s", "id": 41765, "pid": 76337, "tid": -914061504, "ts": 1716454218056175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218127770, "dur": 5,
    "args": {
      "External id": 41767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41767,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 41767, "pid": 5, "tid": 7, "ts": 1716454218127770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056187, "dur": 6,
    "args": {
      "External id": 41767,
      "cbid": 211, "correlation": 41767
    }
  },
  {
    "ph": "s", "id": 41767, "pid": 76337, "tid": -914061504, "ts": 1716454218056187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218127776, "dur": 29,
    "args": {
      "External id": 41777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41777, "pid": 5, "tid": 7, "ts": 1716454218127776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056244, "dur": 13,
    "args": {
      "External id": 41777,
      "cbid": 211, "correlation": 41777
    }
  },
  {
    "ph": "s", "id": 41777, "pid": 76337, "tid": -914061504, "ts": 1716454218056244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218127807, "dur": 30,
    "args": {
      "External id": 41797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41797,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 41797, "pid": 5, "tid": 7, "ts": 1716454218127807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056311, "dur": 10,
    "args": {
      "External id": 41797,
      "cbid": 211, "correlation": 41797
    }
  },
  {
    "ph": "s", "id": 41797, "pid": 76337, "tid": -914061504, "ts": 1716454218056311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218127838, "dur": 4,
    "args": {
      "External id": 41809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41809,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 41809, "pid": 5, "tid": 7, "ts": 1716454218127838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056331, "dur": 6,
    "args": {
      "External id": 41809,
      "cbid": 211, "correlation": 41809
    }
  },
  {
    "ph": "s", "id": 41809, "pid": 76337, "tid": -914061504, "ts": 1716454218056331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218127843, "dur": 29,
    "args": {
      "External id": 41812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41812, "pid": 5, "tid": 7, "ts": 1716454218127843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056350, "dur": 7,
    "args": {
      "External id": 41812,
      "cbid": 211, "correlation": 41812
    }
  },
  {
    "ph": "s", "id": 41812, "pid": 76337, "tid": -914061504, "ts": 1716454218056350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218127873, "dur": 21,
    "args": {
      "External id": 41821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41821,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41821, "pid": 5, "tid": 7, "ts": 1716454218127873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056390, "dur": 9,
    "args": {
      "External id": 41821,
      "cbid": 211, "correlation": 41821
    }
  },
  {
    "ph": "s", "id": 41821, "pid": 76337, "tid": -914061504, "ts": 1716454218056390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056452, "dur": 0,
    "args": {
      "External id": 41831,
      "cbid": 317, "correlation": 41831
    }
  },
  {
    "ph": "f", "id": 41831, "pid": 76337, "tid": -914061504, "ts": 1716454218056452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056453, "dur": 0,
    "args": {
      "External id": 41832,
      "cbid": 203, "correlation": 41832
    }
  },
  {
    "ph": "f", "id": 41832, "pid": 76337, "tid": -914061504, "ts": 1716454218056453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056454, "dur": 0,
    "args": {
      "External id": 41833,
      "cbid": 205, "correlation": 41833
    }
  },
  {
    "ph": "f", "id": 41833, "pid": 76337, "tid": -914061504, "ts": 1716454218056454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218127895, "dur": 23,
    "args": {
      "External id": 41837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41837,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41837, "pid": 5, "tid": 7, "ts": 1716454218127895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056468, "dur": 12,
    "args": {
      "External id": 41837,
      "cbid": 211, "correlation": 41837
    }
  },
  {
    "ph": "s", "id": 41837, "pid": 76337, "tid": -914061504, "ts": 1716454218056468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218127919, "dur": 43,
    "args": {
      "External id": 41839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41839,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41839, "pid": 5, "tid": 7, "ts": 1716454218127919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056483, "dur": 5,
    "args": {
      "External id": 41839,
      "cbid": 211, "correlation": 41839
    }
  },
  {
    "ph": "s", "id": 41839, "pid": 76337, "tid": -914061504, "ts": 1716454218056483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218127963, "dur": 632,
    "args": {
      "External id": 41841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41841,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41841, "pid": 5, "tid": 7, "ts": 1716454218127963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056494, "dur": 6,
    "args": {
      "External id": 41841,
      "cbid": 211, "correlation": 41841
    }
  },
  {
    "ph": "s", "id": 41841, "pid": 76337, "tid": -914061504, "ts": 1716454218056494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218128597, "dur": 22,
    "args": {
      "External id": 41843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41843,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41843, "pid": 5, "tid": 7, "ts": 1716454218128597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056503, "dur": 5,
    "args": {
      "External id": 41843,
      "cbid": 211, "correlation": 41843
    }
  },
  {
    "ph": "s", "id": 41843, "pid": 76337, "tid": -914061504, "ts": 1716454218056503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218128620, "dur": 32,
    "args": {
      "External id": 41849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41849, "pid": 5, "tid": 7, "ts": 1716454218128620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056531, "dur": 8,
    "args": {
      "External id": 41849,
      "cbid": 211, "correlation": 41849
    }
  },
  {
    "ph": "s", "id": 41849, "pid": 76337, "tid": -914061504, "ts": 1716454218056531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056588, "dur": 0,
    "args": {
      "External id": 41859,
      "cbid": 317, "correlation": 41859
    }
  },
  {
    "ph": "f", "id": 41859, "pid": 76337, "tid": -914061504, "ts": 1716454218056588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056589, "dur": 0,
    "args": {
      "External id": 41860,
      "cbid": 203, "correlation": 41860
    }
  },
  {
    "ph": "f", "id": 41860, "pid": 76337, "tid": -914061504, "ts": 1716454218056589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056590, "dur": 0,
    "args": {
      "External id": 41861,
      "cbid": 205, "correlation": 41861
    }
  },
  {
    "ph": "f", "id": 41861, "pid": 76337, "tid": -914061504, "ts": 1716454218056590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218128653, "dur": 39,
    "args": {
      "External id": 41865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41865,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41865, "pid": 5, "tid": 7, "ts": 1716454218128653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056602, "dur": 12,
    "args": {
      "External id": 41865,
      "cbid": 211, "correlation": 41865
    }
  },
  {
    "ph": "s", "id": 41865, "pid": 76337, "tid": -914061504, "ts": 1716454218056602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218128693, "dur": 184,
    "args": {
      "External id": 41867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41867,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41867, "pid": 5, "tid": 7, "ts": 1716454218128693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056620, "dur": 6,
    "args": {
      "External id": 41867,
      "cbid": 211, "correlation": 41867
    }
  },
  {
    "ph": "s", "id": 41867, "pid": 76337, "tid": -914061504, "ts": 1716454218056620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218128878, "dur": 21,
    "args": {
      "External id": 41869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41869,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41869, "pid": 5, "tid": 7, "ts": 1716454218128878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056629, "dur": 5,
    "args": {
      "External id": 41869,
      "cbid": 211, "correlation": 41869
    }
  },
  {
    "ph": "s", "id": 41869, "pid": 76337, "tid": -914061504, "ts": 1716454218056629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218128901, "dur": 32,
    "args": {
      "External id": 41875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41875, "pid": 5, "tid": 7, "ts": 1716454218128901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056654, "dur": 8,
    "args": {
      "External id": 41875,
      "cbid": 211, "correlation": 41875
    }
  },
  {
    "ph": "s", "id": 41875, "pid": 76337, "tid": -914061504, "ts": 1716454218056654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218128934, "dur": 27,
    "args": {
      "External id": 41883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41883,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41883, "pid": 5, "tid": 7, "ts": 1716454218128934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056683, "dur": 8,
    "args": {
      "External id": 41883,
      "cbid": 211, "correlation": 41883
    }
  },
  {
    "ph": "s", "id": 41883, "pid": 76337, "tid": -914061504, "ts": 1716454218056683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218128962, "dur": 20,
    "args": {
      "External id": 41891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41891,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41891, "pid": 5, "tid": 7, "ts": 1716454218128962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056712, "dur": 9,
    "args": {
      "External id": 41891,
      "cbid": 211, "correlation": 41891
    }
  },
  {
    "ph": "s", "id": 41891, "pid": 76337, "tid": -914061504, "ts": 1716454218056712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218128983, "dur": 30,
    "args": {
      "External id": 41911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41911,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 41911, "pid": 5, "tid": 7, "ts": 1716454218128983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056794, "dur": 12,
    "args": {
      "External id": 41911,
      "cbid": 211, "correlation": 41911
    }
  },
  {
    "ph": "s", "id": 41911, "pid": 76337, "tid": -914061504, "ts": 1716454218056794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218129014, "dur": 4,
    "args": {
      "External id": 41923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41923,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 41923, "pid": 5, "tid": 7, "ts": 1716454218129014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056816, "dur": 7,
    "args": {
      "External id": 41923,
      "cbid": 211, "correlation": 41923
    }
  },
  {
    "ph": "s", "id": 41923, "pid": 76337, "tid": -914061504, "ts": 1716454218056816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218129019, "dur": 31,
    "args": {
      "External id": 41926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41926, "pid": 5, "tid": 7, "ts": 1716454218129019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056834, "dur": 6,
    "args": {
      "External id": 41926,
      "cbid": 211, "correlation": 41926
    }
  },
  {
    "ph": "s", "id": 41926, "pid": 76337, "tid": -914061504, "ts": 1716454218056834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056891, "dur": 0,
    "args": {
      "External id": 41937,
      "cbid": 317, "correlation": 41937
    }
  },
  {
    "ph": "f", "id": 41937, "pid": 76337, "tid": -914061504, "ts": 1716454218056891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056892, "dur": 0,
    "args": {
      "External id": 41938,
      "cbid": 203, "correlation": 41938
    }
  },
  {
    "ph": "f", "id": 41938, "pid": 76337, "tid": -914061504, "ts": 1716454218056892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056893, "dur": 0,
    "args": {
      "External id": 41939,
      "cbid": 205, "correlation": 41939
    }
  },
  {
    "ph": "f", "id": 41939, "pid": 76337, "tid": -914061504, "ts": 1716454218056893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218129051, "dur": 21,
    "args": {
      "External id": 41943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41943,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41943, "pid": 5, "tid": 7, "ts": 1716454218129051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056905, "dur": 12,
    "args": {
      "External id": 41943,
      "cbid": 211, "correlation": 41943
    }
  },
  {
    "ph": "s", "id": 41943, "pid": 76337, "tid": -914061504, "ts": 1716454218056905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218129073, "dur": 101,
    "args": {
      "External id": 41945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41945,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41945, "pid": 5, "tid": 7, "ts": 1716454218129073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056923, "dur": 7,
    "args": {
      "External id": 41945,
      "cbid": 211, "correlation": 41945
    }
  },
  {
    "ph": "s", "id": 41945, "pid": 76337, "tid": -914061504, "ts": 1716454218056923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218129176, "dur": 22,
    "args": {
      "External id": 41947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41947, "pid": 5, "tid": 7, "ts": 1716454218129176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056933, "dur": 5,
    "args": {
      "External id": 41947,
      "cbid": 211, "correlation": 41947
    }
  },
  {
    "ph": "s", "id": 41947, "pid": 76337, "tid": -914061504, "ts": 1716454218056933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218129200, "dur": 33,
    "args": {
      "External id": 41953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41953,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41953, "pid": 5, "tid": 7, "ts": 1716454218129200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218056960, "dur": 8,
    "args": {
      "External id": 41953,
      "cbid": 211, "correlation": 41953
    }
  },
  {
    "ph": "s", "id": 41953, "pid": 76337, "tid": -914061504, "ts": 1716454218056960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218129233, "dur": 167,
    "args": {
      "External id": 41962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41962,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41962, "pid": 5, "tid": 7, "ts": 1716454218129233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062148, "dur": 65,
    "args": {
      "External id": 41962,
      "cbid": 211, "correlation": 41962
    }
  },
  {
    "ph": "s", "id": 41962, "pid": 76337, "tid": -914061504, "ts": 1716454218062148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218129402, "dur": 63,
    "args": {
      "External id": 41984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41984,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 41984, "pid": 5, "tid": 7, "ts": 1716454218129402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062299, "dur": 13,
    "args": {
      "External id": 41984,
      "cbid": 211, "correlation": 41984
    }
  },
  {
    "ph": "s", "id": 41984, "pid": 76337, "tid": -914061504, "ts": 1716454218062299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062474, "dur": 6,
    "args": {
      "External id": 41995,
      "cbid": 251, "correlation": 41995
    }
  },
  {
    "ph": "f", "id": 41995, "pid": 76337, "tid": -914061504, "ts": 1716454218062474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218129466, "dur": 151,
    "args": {
      "External id": 41996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 41996,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 41996, "pid": 5, "tid": 7, "ts": 1716454218129466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062488, "dur": 16,
    "args": {
      "External id": 41996,
      "cbid": 211, "correlation": 41996
    }
  },
  {
    "ph": "s", "id": 41996, "pid": 76337, "tid": -914061504, "ts": 1716454218062488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062567, "dur": 1,
    "args": {
      "External id": 42007,
      "cbid": 251, "correlation": 42007
    }
  },
  {
    "ph": "f", "id": 42007, "pid": 76337, "tid": -914061504, "ts": 1716454218062567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218129619, "dur": 142,
    "args": {
      "External id": 42008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42008,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42008, "pid": 5, "tid": 7, "ts": 1716454218129619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062571, "dur": 11,
    "args": {
      "External id": 42008,
      "cbid": 211, "correlation": 42008
    }
  },
  {
    "ph": "s", "id": 42008, "pid": 76337, "tid": -914061504, "ts": 1716454218062571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062637, "dur": 1,
    "args": {
      "External id": 42019,
      "cbid": 251, "correlation": 42019
    }
  },
  {
    "ph": "f", "id": 42019, "pid": 76337, "tid": -914061504, "ts": 1716454218062637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218129762, "dur": 142,
    "args": {
      "External id": 42020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42020,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42020, "pid": 5, "tid": 7, "ts": 1716454218129762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062641, "dur": 11,
    "args": {
      "External id": 42020,
      "cbid": 211, "correlation": 42020
    }
  },
  {
    "ph": "s", "id": 42020, "pid": 76337, "tid": -914061504, "ts": 1716454218062641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218129905, "dur": 1892,
    "args": {
      "External id": 42041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42041,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 42041, "pid": 5, "tid": 7, "ts": 1716454218129905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062755, "dur": 20,
    "args": {
      "External id": 42041,
      "cbid": 211, "correlation": 42041
    }
  },
  {
    "ph": "s", "id": 42041, "pid": 76337, "tid": -914061504, "ts": 1716454218062755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062900, "dur": 2,
    "args": {
      "External id": 42059,
      "cbid": 251, "correlation": 42059
    }
  },
  {
    "ph": "f", "id": 42059, "pid": 76337, "tid": -914061504, "ts": 1716454218062900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218131798, "dur": 145,
    "args": {
      "External id": 42061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42061,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 42061, "pid": 5, "tid": 7, "ts": 1716454218131798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218062908, "dur": 15,
    "args": {
      "External id": 42061,
      "cbid": 211, "correlation": 42061
    }
  },
  {
    "ph": "s", "id": 42061, "pid": 76337, "tid": -914061504, "ts": 1716454218062908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218131945, "dur": 36,
    "args": {
      "External id": 42069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42069,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42069, "pid": 5, "tid": 7, "ts": 1716454218131945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063032, "dur": 20,
    "args": {
      "External id": 42069,
      "cbid": 211, "correlation": 42069
    }
  },
  {
    "ph": "s", "id": 42069, "pid": 76337, "tid": -914061504, "ts": 1716454218063032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218131981, "dur": 51,
    "args": {
      "External id": 42077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42077,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42077, "pid": 5, "tid": 7, "ts": 1716454218131981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063089, "dur": 10,
    "args": {
      "External id": 42077,
      "cbid": 211, "correlation": 42077
    }
  },
  {
    "ph": "s", "id": 42077, "pid": 76337, "tid": -914061504, "ts": 1716454218063089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218132034, "dur": 29,
    "args": {
      "External id": 42088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42088,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42088, "pid": 5, "tid": 7, "ts": 1716454218132034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063193, "dur": 17,
    "args": {
      "External id": 42088,
      "cbid": 211, "correlation": 42088
    }
  },
  {
    "ph": "s", "id": 42088, "pid": 76337, "tid": -914061504, "ts": 1716454218063193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218132064, "dur": 34,
    "args": {
      "External id": 42110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42110,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42110, "pid": 5, "tid": 7, "ts": 1716454218132064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063230, "dur": 8,
    "args": {
      "External id": 42110,
      "cbid": 211, "correlation": 42110
    }
  },
  {
    "ph": "s", "id": 42110, "pid": 76337, "tid": -914061504, "ts": 1716454218063230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063317, "dur": 1,
    "args": {
      "External id": 42121,
      "cbid": 251, "correlation": 42121
    }
  },
  {
    "ph": "f", "id": 42121, "pid": 76337, "tid": -914061504, "ts": 1716454218063317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218132099, "dur": 89,
    "args": {
      "External id": 42122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42122,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42122, "pid": 5, "tid": 7, "ts": 1716454218132099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063322, "dur": 12,
    "args": {
      "External id": 42122,
      "cbid": 211, "correlation": 42122
    }
  },
  {
    "ph": "s", "id": 42122, "pid": 76337, "tid": -914061504, "ts": 1716454218063322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063408, "dur": 1,
    "args": {
      "External id": 42133,
      "cbid": 251, "correlation": 42133
    }
  },
  {
    "ph": "f", "id": 42133, "pid": 76337, "tid": -914061504, "ts": 1716454218063408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063413, "dur": 0,
    "args": {
      "External id": 42134,
      "cbid": 251, "correlation": 42134
    }
  },
  {
    "ph": "f", "id": 42134, "pid": 76337, "tid": -914061504, "ts": 1716454218063413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218132190, "dur": 12,
    "args": {
      "External id": 42135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42135,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 42135, "pid": 5, "tid": 7, "ts": 1716454218132190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063416, "dur": 14,
    "args": {
      "External id": 42135,
      "cbid": 211, "correlation": 42135
    }
  },
  {
    "ph": "s", "id": 42135, "pid": 76337, "tid": -914061504, "ts": 1716454218063416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218132203, "dur": 5,
    "args": {
      "External id": 42137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42137,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 42137, "pid": 5, "tid": 7, "ts": 1716454218132203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063433, "dur": 11,
    "args": {
      "External id": 42137,
      "cbid": 211, "correlation": 42137
    }
  },
  {
    "ph": "s", "id": 42137, "pid": 76337, "tid": -914061504, "ts": 1716454218063433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063497, "dur": 1,
    "args": {
      "External id": 42148,
      "cbid": 251, "correlation": 42148
    }
  },
  {
    "ph": "f", "id": 42148, "pid": 76337, "tid": -914061504, "ts": 1716454218063497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063500, "dur": 0,
    "args": {
      "External id": 42149,
      "cbid": 251, "correlation": 42149
    }
  },
  {
    "ph": "f", "id": 42149, "pid": 76337, "tid": -914061504, "ts": 1716454218063500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218132209, "dur": 7,
    "args": {
      "External id": 42150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42150,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 42150, "pid": 5, "tid": 7, "ts": 1716454218132209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063502, "dur": 12,
    "args": {
      "External id": 42150,
      "cbid": 211, "correlation": 42150
    }
  },
  {
    "ph": "s", "id": 42150, "pid": 76337, "tid": -914061504, "ts": 1716454218063502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218132217, "dur": 3,
    "args": {
      "External id": 42152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42152,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 42152, "pid": 5, "tid": 7, "ts": 1716454218132217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063515, "dur": 5,
    "args": {
      "External id": 42152,
      "cbid": 211, "correlation": 42152
    }
  },
  {
    "ph": "s", "id": 42152, "pid": 76337, "tid": -914061504, "ts": 1716454218063515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218132221, "dur": 89,
    "args": {
      "External id": 42173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42173,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 42173, "pid": 5, "tid": 7, "ts": 1716454218132221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063588, "dur": 13,
    "args": {
      "External id": 42173,
      "cbid": 211, "correlation": 42173
    }
  },
  {
    "ph": "s", "id": 42173, "pid": 76337, "tid": -914061504, "ts": 1716454218063588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063686, "dur": 2,
    "args": {
      "External id": 42191,
      "cbid": 251, "correlation": 42191
    }
  },
  {
    "ph": "f", "id": 42191, "pid": 76337, "tid": -914061504, "ts": 1716454218063686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218132311, "dur": 98,
    "args": {
      "External id": 42193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42193,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42193, "pid": 5, "tid": 7, "ts": 1716454218132311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063693, "dur": 14,
    "args": {
      "External id": 42193,
      "cbid": 211, "correlation": 42193
    }
  },
  {
    "ph": "s", "id": 42193, "pid": 76337, "tid": -914061504, "ts": 1716454218063693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218132410, "dur": 19,
    "args": {
      "External id": 42201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42201,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42201, "pid": 5, "tid": 7, "ts": 1716454218132410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063766, "dur": 12,
    "args": {
      "External id": 42201,
      "cbid": 211, "correlation": 42201
    }
  },
  {
    "ph": "s", "id": 42201, "pid": 76337, "tid": -914061504, "ts": 1716454218063766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218132431, "dur": 38,
    "args": {
      "External id": 42209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42209, "pid": 5, "tid": 7, "ts": 1716454218132431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063810, "dur": 11,
    "args": {
      "External id": 42209,
      "cbid": 211, "correlation": 42209
    }
  },
  {
    "ph": "s", "id": 42209, "pid": 76337, "tid": -914061504, "ts": 1716454218063810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218132471, "dur": 33,
    "args": {
      "External id": 42231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42231,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42231, "pid": 5, "tid": 7, "ts": 1716454218132471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063864, "dur": 10,
    "args": {
      "External id": 42231,
      "cbid": 211, "correlation": 42231
    }
  },
  {
    "ph": "s", "id": 42231, "pid": 76337, "tid": -914061504, "ts": 1716454218063864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063959, "dur": 1,
    "args": {
      "External id": 42247,
      "cbid": 251, "correlation": 42247
    }
  },
  {
    "ph": "f", "id": 42247, "pid": 76337, "tid": -914061504, "ts": 1716454218063959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063964, "dur": 0,
    "args": {
      "External id": 42249,
      "cbid": 251, "correlation": 42249
    }
  },
  {
    "ph": "f", "id": 42249, "pid": 76337, "tid": -914061504, "ts": 1716454218063964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218132506, "dur": 527,
    "args": {
      "External id": 42250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42250,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 42250, "pid": 5, "tid": 7, "ts": 1716454218132506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218063970, "dur": 21,
    "args": {
      "External id": 42250,
      "cbid": 211, "correlation": 42250
    }
  },
  {
    "ph": "s", "id": 42250, "pid": 76337, "tid": -914061504, "ts": 1716454218063970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218133034, "dur": 123,
    "args": {
      "External id": 42258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42258, "pid": 5, "tid": 7, "ts": 1716454218133034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064072, "dur": 14,
    "args": {
      "External id": 42258,
      "cbid": 211, "correlation": 42258
    }
  },
  {
    "ph": "s", "id": 42258, "pid": 76337, "tid": -914061504, "ts": 1716454218064072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218133158, "dur": 127,
    "args": {
      "External id": 42266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42266,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42266, "pid": 5, "tid": 7, "ts": 1716454218133158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064113, "dur": 11,
    "args": {
      "External id": 42266,
      "cbid": 211, "correlation": 42266
    }
  },
  {
    "ph": "s", "id": 42266, "pid": 76337, "tid": -914061504, "ts": 1716454218064113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064198, "dur": 1,
    "args": {
      "External id": 42282,
      "cbid": 251, "correlation": 42282
    }
  },
  {
    "ph": "f", "id": 42282, "pid": 76337, "tid": -914061504, "ts": 1716454218064198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218133287, "dur": 302,
    "args": {
      "External id": 42284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42284,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42284, "pid": 5, "tid": 7, "ts": 1716454218133287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064204, "dur": 13,
    "args": {
      "External id": 42284,
      "cbid": 211, "correlation": 42284
    }
  },
  {
    "ph": "s", "id": 42284, "pid": 76337, "tid": -914061504, "ts": 1716454218064204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218133590, "dur": 27,
    "args": {
      "External id": 42292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42292,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42292, "pid": 5, "tid": 7, "ts": 1716454218133590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064247, "dur": 10,
    "args": {
      "External id": 42292,
      "cbid": 211, "correlation": 42292
    }
  },
  {
    "ph": "s", "id": 42292, "pid": 76337, "tid": -914061504, "ts": 1716454218064247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218133618, "dur": 80,
    "args": {
      "External id": 42303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42303,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42303, "pid": 5, "tid": 7, "ts": 1716454218133618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064319, "dur": 13,
    "args": {
      "External id": 42303,
      "cbid": 211, "correlation": 42303
    }
  },
  {
    "ph": "s", "id": 42303, "pid": 76337, "tid": -914061504, "ts": 1716454218064319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064421, "dur": 1,
    "args": {
      "External id": 42315,
      "cbid": 317, "correlation": 42315
    }
  },
  {
    "ph": "f", "id": 42315, "pid": 76337, "tid": -914061504, "ts": 1716454218064421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064423, "dur": 1,
    "args": {
      "External id": 42316,
      "cbid": 203, "correlation": 42316
    }
  },
  {
    "ph": "f", "id": 42316, "pid": 76337, "tid": -914061504, "ts": 1716454218064423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064426, "dur": 1,
    "args": {
      "External id": 42317,
      "cbid": 205, "correlation": 42317
    }
  },
  {
    "ph": "f", "id": 42317, "pid": 76337, "tid": -914061504, "ts": 1716454218064426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218133699, "dur": 24,
    "args": {
      "External id": 42321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42321,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42321, "pid": 5, "tid": 7, "ts": 1716454218133699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064471, "dur": 15,
    "args": {
      "External id": 42321,
      "cbid": 211, "correlation": 42321
    }
  },
  {
    "ph": "s", "id": 42321, "pid": 76337, "tid": -914061504, "ts": 1716454218064471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218133725, "dur": 117,
    "args": {
      "External id": 42323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42323,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42323, "pid": 5, "tid": 7, "ts": 1716454218133725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064507, "dur": 11,
    "args": {
      "External id": 42323,
      "cbid": 211, "correlation": 42323
    }
  },
  {
    "ph": "s", "id": 42323, "pid": 76337, "tid": -914061504, "ts": 1716454218064507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218133843, "dur": 23,
    "args": {
      "External id": 42325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42325,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42325, "pid": 5, "tid": 7, "ts": 1716454218133843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064526, "dur": 8,
    "args": {
      "External id": 42325,
      "cbid": 211, "correlation": 42325
    }
  },
  {
    "ph": "s", "id": 42325, "pid": 76337, "tid": -914061504, "ts": 1716454218064526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218133868, "dur": 32,
    "args": {
      "External id": 42331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42331,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42331, "pid": 5, "tid": 7, "ts": 1716454218133868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064562, "dur": 9,
    "args": {
      "External id": 42331,
      "cbid": 211, "correlation": 42331
    }
  },
  {
    "ph": "s", "id": 42331, "pid": 76337, "tid": -914061504, "ts": 1716454218064562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218133901, "dur": 26,
    "args": {
      "External id": 42339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42339,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42339, "pid": 5, "tid": 7, "ts": 1716454218133901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064593, "dur": 8,
    "args": {
      "External id": 42339,
      "cbid": 211, "correlation": 42339
    }
  },
  {
    "ph": "s", "id": 42339, "pid": 76337, "tid": -914061504, "ts": 1716454218064593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218133928, "dur": 44,
    "args": {
      "External id": 42348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42348,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42348, "pid": 5, "tid": 7, "ts": 1716454218133928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064640, "dur": 13,
    "args": {
      "External id": 42348,
      "cbid": 211, "correlation": 42348
    }
  },
  {
    "ph": "s", "id": 42348, "pid": 76337, "tid": -914061504, "ts": 1716454218064640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218133973, "dur": 41,
    "args": {
      "External id": 42368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42368,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 42368, "pid": 5, "tid": 7, "ts": 1716454218133973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064742, "dur": 14,
    "args": {
      "External id": 42368,
      "cbid": 211, "correlation": 42368
    }
  },
  {
    "ph": "s", "id": 42368, "pid": 76337, "tid": -914061504, "ts": 1716454218064742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218134015, "dur": 5,
    "args": {
      "External id": 42380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42380,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 42380, "pid": 5, "tid": 7, "ts": 1716454218134015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064768, "dur": 7,
    "args": {
      "External id": 42380,
      "cbid": 211, "correlation": 42380
    }
  },
  {
    "ph": "s", "id": 42380, "pid": 76337, "tid": -914061504, "ts": 1716454218064768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218134022, "dur": 43,
    "args": {
      "External id": 42383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42383, "pid": 5, "tid": 7, "ts": 1716454218134022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064792, "dur": 7,
    "args": {
      "External id": 42383,
      "cbid": 211, "correlation": 42383
    }
  },
  {
    "ph": "s", "id": 42383, "pid": 76337, "tid": -914061504, "ts": 1716454218064792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218134066, "dur": 30,
    "args": {
      "External id": 42392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42392,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42392, "pid": 5, "tid": 7, "ts": 1716454218134066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064845, "dur": 12,
    "args": {
      "External id": 42392,
      "cbid": 211, "correlation": 42392
    }
  },
  {
    "ph": "s", "id": 42392, "pid": 76337, "tid": -914061504, "ts": 1716454218064845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064902, "dur": 0,
    "args": {
      "External id": 42402,
      "cbid": 317, "correlation": 42402
    }
  },
  {
    "ph": "f", "id": 42402, "pid": 76337, "tid": -914061504, "ts": 1716454218064902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064903, "dur": 0,
    "args": {
      "External id": 42403,
      "cbid": 203, "correlation": 42403
    }
  },
  {
    "ph": "f", "id": 42403, "pid": 76337, "tid": -914061504, "ts": 1716454218064903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064903, "dur": 0,
    "args": {
      "External id": 42404,
      "cbid": 205, "correlation": 42404
    }
  },
  {
    "ph": "f", "id": 42404, "pid": 76337, "tid": -914061504, "ts": 1716454218064903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218134098, "dur": 31,
    "args": {
      "External id": 42408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42408,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42408, "pid": 5, "tid": 7, "ts": 1716454218134098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064919, "dur": 12,
    "args": {
      "External id": 42408,
      "cbid": 211, "correlation": 42408
    }
  },
  {
    "ph": "s", "id": 42408, "pid": 76337, "tid": -914061504, "ts": 1716454218064919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218134130, "dur": 62,
    "args": {
      "External id": 42410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42410, "pid": 5, "tid": 7, "ts": 1716454218134130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064934, "dur": 5,
    "args": {
      "External id": 42410,
      "cbid": 211, "correlation": 42410
    }
  },
  {
    "ph": "s", "id": 42410, "pid": 76337, "tid": -914061504, "ts": 1716454218064934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218134193, "dur": 948,
    "args": {
      "External id": 42412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42412,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42412, "pid": 5, "tid": 7, "ts": 1716454218134193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064946, "dur": 12,
    "args": {
      "External id": 42412,
      "cbid": 211, "correlation": 42412
    }
  },
  {
    "ph": "s", "id": 42412, "pid": 76337, "tid": -914061504, "ts": 1716454218064946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218135143, "dur": 21,
    "args": {
      "External id": 42414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42414,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42414, "pid": 5, "tid": 7, "ts": 1716454218135143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064962, "dur": 5,
    "args": {
      "External id": 42414,
      "cbid": 211, "correlation": 42414
    }
  },
  {
    "ph": "s", "id": 42414, "pid": 76337, "tid": -914061504, "ts": 1716454218064962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218135165, "dur": 33,
    "args": {
      "External id": 42420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42420, "pid": 5, "tid": 7, "ts": 1716454218135165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218064999, "dur": 9,
    "args": {
      "External id": 42420,
      "cbid": 211, "correlation": 42420
    }
  },
  {
    "ph": "s", "id": 42420, "pid": 76337, "tid": -914061504, "ts": 1716454218064999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218135199, "dur": 3,
    "args": {
      "External id": 42428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42428,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 42428, "pid": 5, "tid": 7, "ts": 1716454218135199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065045, "dur": 9,
    "args": {
      "External id": 42428,
      "cbid": 211, "correlation": 42428
    }
  },
  {
    "ph": "s", "id": 42428, "pid": 76337, "tid": -914061504, "ts": 1716454218065045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065113, "dur": 1,
    "args": {
      "External id": 42444,
      "cbid": 251, "correlation": 42444
    }
  },
  {
    "ph": "f", "id": 42444, "pid": 76337, "tid": -914061504, "ts": 1716454218065113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065119, "dur": 0,
    "args": {
      "External id": 42446,
      "cbid": 251, "correlation": 42446
    }
  },
  {
    "ph": "f", "id": 42446, "pid": 76337, "tid": -914061504, "ts": 1716454218065119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218135204, "dur": 12,
    "args": {
      "External id": 42447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42447,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 42447, "pid": 5, "tid": 7, "ts": 1716454218135204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065120, "dur": 11,
    "args": {
      "External id": 42447,
      "cbid": 211, "correlation": 42447
    }
  },
  {
    "ph": "s", "id": 42447, "pid": 76337, "tid": -914061504, "ts": 1716454218065120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218135217, "dur": 5,
    "args": {
      "External id": 42449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42449,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 42449, "pid": 5, "tid": 7, "ts": 1716454218135217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065135, "dur": 7,
    "args": {
      "External id": 42449,
      "cbid": 211, "correlation": 42449
    }
  },
  {
    "ph": "s", "id": 42449, "pid": 76337, "tid": -914061504, "ts": 1716454218065135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218135223, "dur": 29,
    "args": {
      "External id": 42459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42459,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42459, "pid": 5, "tid": 7, "ts": 1716454218135223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065204, "dur": 12,
    "args": {
      "External id": 42459,
      "cbid": 211, "correlation": 42459
    }
  },
  {
    "ph": "s", "id": 42459, "pid": 76337, "tid": -914061504, "ts": 1716454218065204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218135254, "dur": 31,
    "args": {
      "External id": 42479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42479,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 42479, "pid": 5, "tid": 7, "ts": 1716454218135254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065269, "dur": 11,
    "args": {
      "External id": 42479,
      "cbid": 211, "correlation": 42479
    }
  },
  {
    "ph": "s", "id": 42479, "pid": 76337, "tid": -914061504, "ts": 1716454218065269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218135286, "dur": 4,
    "args": {
      "External id": 42491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42491,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 42491, "pid": 5, "tid": 7, "ts": 1716454218135286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065294, "dur": 6,
    "args": {
      "External id": 42491,
      "cbid": 211, "correlation": 42491
    }
  },
  {
    "ph": "s", "id": 42491, "pid": 76337, "tid": -914061504, "ts": 1716454218065294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218135291, "dur": 31,
    "args": {
      "External id": 42494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42494,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42494, "pid": 5, "tid": 7, "ts": 1716454218135291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065313, "dur": 7,
    "args": {
      "External id": 42494,
      "cbid": 211, "correlation": 42494
    }
  },
  {
    "ph": "s", "id": 42494, "pid": 76337, "tid": -914061504, "ts": 1716454218065313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218135324, "dur": 20,
    "args": {
      "External id": 42503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42503,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42503, "pid": 5, "tid": 7, "ts": 1716454218135324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065354, "dur": 9,
    "args": {
      "External id": 42503,
      "cbid": 211, "correlation": 42503
    }
  },
  {
    "ph": "s", "id": 42503, "pid": 76337, "tid": -914061504, "ts": 1716454218065354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065418, "dur": 0,
    "args": {
      "External id": 42513,
      "cbid": 317, "correlation": 42513
    }
  },
  {
    "ph": "f", "id": 42513, "pid": 76337, "tid": -914061504, "ts": 1716454218065418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065419, "dur": 0,
    "args": {
      "External id": 42514,
      "cbid": 203, "correlation": 42514
    }
  },
  {
    "ph": "f", "id": 42514, "pid": 76337, "tid": -914061504, "ts": 1716454218065419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065420, "dur": 0,
    "args": {
      "External id": 42515,
      "cbid": 205, "correlation": 42515
    }
  },
  {
    "ph": "f", "id": 42515, "pid": 76337, "tid": -914061504, "ts": 1716454218065420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218135345, "dur": 22,
    "args": {
      "External id": 42519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42519,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42519, "pid": 5, "tid": 7, "ts": 1716454218135345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065436, "dur": 13,
    "args": {
      "External id": 42519,
      "cbid": 211, "correlation": 42519
    }
  },
  {
    "ph": "s", "id": 42519, "pid": 76337, "tid": -914061504, "ts": 1716454218065436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218135368, "dur": 44,
    "args": {
      "External id": 42521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42521,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42521, "pid": 5, "tid": 7, "ts": 1716454218135368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065452, "dur": 5,
    "args": {
      "External id": 42521,
      "cbid": 211, "correlation": 42521
    }
  },
  {
    "ph": "s", "id": 42521, "pid": 76337, "tid": -914061504, "ts": 1716454218065452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218135413, "dur": 632,
    "args": {
      "External id": 42523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42523,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42523, "pid": 5, "tid": 7, "ts": 1716454218135413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065464, "dur": 6,
    "args": {
      "External id": 42523,
      "cbid": 211, "correlation": 42523
    }
  },
  {
    "ph": "s", "id": 42523, "pid": 76337, "tid": -914061504, "ts": 1716454218065464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136046, "dur": 21,
    "args": {
      "External id": 42525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42525,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42525, "pid": 5, "tid": 7, "ts": 1716454218136046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065474, "dur": 5,
    "args": {
      "External id": 42525,
      "cbid": 211, "correlation": 42525
    }
  },
  {
    "ph": "s", "id": 42525, "pid": 76337, "tid": -914061504, "ts": 1716454218065474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218136068, "dur": 32,
    "args": {
      "External id": 42531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42531, "pid": 5, "tid": 7, "ts": 1716454218136068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065502, "dur": 8,
    "args": {
      "External id": 42531,
      "cbid": 211, "correlation": 42531
    }
  },
  {
    "ph": "s", "id": 42531, "pid": 76337, "tid": -914061504, "ts": 1716454218065502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065560, "dur": 0,
    "args": {
      "External id": 42541,
      "cbid": 317, "correlation": 42541
    }
  },
  {
    "ph": "f", "id": 42541, "pid": 76337, "tid": -914061504, "ts": 1716454218065560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065561, "dur": 0,
    "args": {
      "External id": 42542,
      "cbid": 203, "correlation": 42542
    }
  },
  {
    "ph": "f", "id": 42542, "pid": 76337, "tid": -914061504, "ts": 1716454218065561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065561, "dur": 0,
    "args": {
      "External id": 42543,
      "cbid": 205, "correlation": 42543
    }
  },
  {
    "ph": "f", "id": 42543, "pid": 76337, "tid": -914061504, "ts": 1716454218065561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136102, "dur": 31,
    "args": {
      "External id": 42547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42547,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42547, "pid": 5, "tid": 7, "ts": 1716454218136102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065576, "dur": 14,
    "args": {
      "External id": 42547,
      "cbid": 211, "correlation": 42547
    }
  },
  {
    "ph": "s", "id": 42547, "pid": 76337, "tid": -914061504, "ts": 1716454218065576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218136134, "dur": 151,
    "args": {
      "External id": 42549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42549,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42549, "pid": 5, "tid": 7, "ts": 1716454218136134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065597, "dur": 7,
    "args": {
      "External id": 42549,
      "cbid": 211, "correlation": 42549
    }
  },
  {
    "ph": "s", "id": 42549, "pid": 76337, "tid": -914061504, "ts": 1716454218065597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136286, "dur": 23,
    "args": {
      "External id": 42551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42551,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42551, "pid": 5, "tid": 7, "ts": 1716454218136286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065608, "dur": 5,
    "args": {
      "External id": 42551,
      "cbid": 211, "correlation": 42551
    }
  },
  {
    "ph": "s", "id": 42551, "pid": 76337, "tid": -914061504, "ts": 1716454218065608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218136310, "dur": 32,
    "args": {
      "External id": 42557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42557, "pid": 5, "tid": 7, "ts": 1716454218136310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065634, "dur": 8,
    "args": {
      "External id": 42557,
      "cbid": 211, "correlation": 42557
    }
  },
  {
    "ph": "s", "id": 42557, "pid": 76337, "tid": -914061504, "ts": 1716454218065634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218136344, "dur": 27,
    "args": {
      "External id": 42565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42565,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42565, "pid": 5, "tid": 7, "ts": 1716454218136344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065664, "dur": 7,
    "args": {
      "External id": 42565,
      "cbid": 211, "correlation": 42565
    }
  },
  {
    "ph": "s", "id": 42565, "pid": 76337, "tid": -914061504, "ts": 1716454218065664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218136372, "dur": 20,
    "args": {
      "External id": 42573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42573,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42573, "pid": 5, "tid": 7, "ts": 1716454218136372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065691, "dur": 8,
    "args": {
      "External id": 42573,
      "cbid": 211, "correlation": 42573
    }
  },
  {
    "ph": "s", "id": 42573, "pid": 76337, "tid": -914061504, "ts": 1716454218065691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136393, "dur": 30,
    "args": {
      "External id": 42593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42593,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 42593, "pid": 5, "tid": 7, "ts": 1716454218136393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065777, "dur": 12,
    "args": {
      "External id": 42593,
      "cbid": 211, "correlation": 42593
    }
  },
  {
    "ph": "s", "id": 42593, "pid": 76337, "tid": -914061504, "ts": 1716454218065777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218136425, "dur": 4,
    "args": {
      "External id": 42605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42605,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 42605, "pid": 5, "tid": 7, "ts": 1716454218136425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065799, "dur": 6,
    "args": {
      "External id": 42605,
      "cbid": 211, "correlation": 42605
    }
  },
  {
    "ph": "s", "id": 42605, "pid": 76337, "tid": -914061504, "ts": 1716454218065799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218136430, "dur": 30,
    "args": {
      "External id": 42608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42608, "pid": 5, "tid": 7, "ts": 1716454218136430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065817, "dur": 6,
    "args": {
      "External id": 42608,
      "cbid": 211, "correlation": 42608
    }
  },
  {
    "ph": "s", "id": 42608, "pid": 76337, "tid": -914061504, "ts": 1716454218065817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065877, "dur": 0,
    "args": {
      "External id": 42619,
      "cbid": 317, "correlation": 42619
    }
  },
  {
    "ph": "f", "id": 42619, "pid": 76337, "tid": -914061504, "ts": 1716454218065877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065878, "dur": 0,
    "args": {
      "External id": 42620,
      "cbid": 203, "correlation": 42620
    }
  },
  {
    "ph": "f", "id": 42620, "pid": 76337, "tid": -914061504, "ts": 1716454218065878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065879, "dur": 0,
    "args": {
      "External id": 42621,
      "cbid": 205, "correlation": 42621
    }
  },
  {
    "ph": "f", "id": 42621, "pid": 76337, "tid": -914061504, "ts": 1716454218065879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136461, "dur": 22,
    "args": {
      "External id": 42625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42625, "pid": 5, "tid": 7, "ts": 1716454218136461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065892, "dur": 12,
    "args": {
      "External id": 42625,
      "cbid": 211, "correlation": 42625
    }
  },
  {
    "ph": "s", "id": 42625, "pid": 76337, "tid": -914061504, "ts": 1716454218065892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218136484, "dur": 102,
    "args": {
      "External id": 42627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42627,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42627, "pid": 5, "tid": 7, "ts": 1716454218136484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065910, "dur": 6,
    "args": {
      "External id": 42627,
      "cbid": 211, "correlation": 42627
    }
  },
  {
    "ph": "s", "id": 42627, "pid": 76337, "tid": -914061504, "ts": 1716454218065910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136588, "dur": 23,
    "args": {
      "External id": 42629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42629,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42629, "pid": 5, "tid": 7, "ts": 1716454218136588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065920, "dur": 5,
    "args": {
      "External id": 42629,
      "cbid": 211, "correlation": 42629
    }
  },
  {
    "ph": "s", "id": 42629, "pid": 76337, "tid": -914061504, "ts": 1716454218065920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218136612, "dur": 32,
    "args": {
      "External id": 42635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42635, "pid": 5, "tid": 7, "ts": 1716454218136612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218065948, "dur": 8,
    "args": {
      "External id": 42635,
      "cbid": 211, "correlation": 42635
    }
  },
  {
    "ph": "s", "id": 42635, "pid": 76337, "tid": -914061504, "ts": 1716454218065948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218136645, "dur": 178,
    "args": {
      "External id": 42644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42644,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42644, "pid": 5, "tid": 7, "ts": 1716454218136645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066039, "dur": 14,
    "args": {
      "External id": 42644,
      "cbid": 211, "correlation": 42644
    }
  },
  {
    "ph": "s", "id": 42644, "pid": 76337, "tid": -914061504, "ts": 1716454218066039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218136824, "dur": 63,
    "args": {
      "External id": 42666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42666,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42666, "pid": 5, "tid": 7, "ts": 1716454218136824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066098, "dur": 10,
    "args": {
      "External id": 42666,
      "cbid": 211, "correlation": 42666
    }
  },
  {
    "ph": "s", "id": 42666, "pid": 76337, "tid": -914061504, "ts": 1716454218066098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066186, "dur": 1,
    "args": {
      "External id": 42677,
      "cbid": 251, "correlation": 42677
    }
  },
  {
    "ph": "f", "id": 42677, "pid": 76337, "tid": -914061504, "ts": 1716454218066186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218136888, "dur": 151,
    "args": {
      "External id": 42678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42678,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42678, "pid": 5, "tid": 7, "ts": 1716454218136888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066192, "dur": 15,
    "args": {
      "External id": 42678,
      "cbid": 211, "correlation": 42678
    }
  },
  {
    "ph": "s", "id": 42678, "pid": 76337, "tid": -914061504, "ts": 1716454218066192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066266, "dur": 1,
    "args": {
      "External id": 42689,
      "cbid": 251, "correlation": 42689
    }
  },
  {
    "ph": "f", "id": 42689, "pid": 76337, "tid": -914061504, "ts": 1716454218066266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218137040, "dur": 141,
    "args": {
      "External id": 42690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42690,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42690, "pid": 5, "tid": 7, "ts": 1716454218137040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066270, "dur": 11,
    "args": {
      "External id": 42690,
      "cbid": 211, "correlation": 42690
    }
  },
  {
    "ph": "s", "id": 42690, "pid": 76337, "tid": -914061504, "ts": 1716454218066270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066336, "dur": 1,
    "args": {
      "External id": 42701,
      "cbid": 251, "correlation": 42701
    }
  },
  {
    "ph": "f", "id": 42701, "pid": 76337, "tid": -914061504, "ts": 1716454218066336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218137183, "dur": 144,
    "args": {
      "External id": 42702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42702,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42702, "pid": 5, "tid": 7, "ts": 1716454218137183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066340, "dur": 11,
    "args": {
      "External id": 42702,
      "cbid": 211, "correlation": 42702
    }
  },
  {
    "ph": "s", "id": 42702, "pid": 76337, "tid": -914061504, "ts": 1716454218066340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218137329, "dur": 1902,
    "args": {
      "External id": 42723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42723,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 42723, "pid": 5, "tid": 7, "ts": 1716454218137329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066419, "dur": 12,
    "args": {
      "External id": 42723,
      "cbid": 211, "correlation": 42723
    }
  },
  {
    "ph": "s", "id": 42723, "pid": 76337, "tid": -914061504, "ts": 1716454218066419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066516, "dur": 1,
    "args": {
      "External id": 42741,
      "cbid": 251, "correlation": 42741
    }
  },
  {
    "ph": "f", "id": 42741, "pid": 76337, "tid": -914061504, "ts": 1716454218066516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218139232, "dur": 147,
    "args": {
      "External id": 42743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42743,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 42743, "pid": 5, "tid": 7, "ts": 1716454218139232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066522, "dur": 13,
    "args": {
      "External id": 42743,
      "cbid": 211, "correlation": 42743
    }
  },
  {
    "ph": "s", "id": 42743, "pid": 76337, "tid": -914061504, "ts": 1716454218066522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218139381, "dur": 35,
    "args": {
      "External id": 42751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42751,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42751, "pid": 5, "tid": 7, "ts": 1716454218139381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066591, "dur": 12,
    "args": {
      "External id": 42751,
      "cbid": 211, "correlation": 42751
    }
  },
  {
    "ph": "s", "id": 42751, "pid": 76337, "tid": -914061504, "ts": 1716454218066591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218139417, "dur": 51,
    "args": {
      "External id": 42759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42759,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42759, "pid": 5, "tid": 7, "ts": 1716454218139417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066630, "dur": 11,
    "args": {
      "External id": 42759,
      "cbid": 211, "correlation": 42759
    }
  },
  {
    "ph": "s", "id": 42759, "pid": 76337, "tid": -914061504, "ts": 1716454218066630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218139470, "dur": 30,
    "args": {
      "External id": 42770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42770, "pid": 5, "tid": 7, "ts": 1716454218139470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066705, "dur": 13,
    "args": {
      "External id": 42770,
      "cbid": 211, "correlation": 42770
    }
  },
  {
    "ph": "s", "id": 42770, "pid": 76337, "tid": -914061504, "ts": 1716454218066705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218139501, "dur": 34,
    "args": {
      "External id": 42792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42792,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42792, "pid": 5, "tid": 7, "ts": 1716454218139501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066737, "dur": 8,
    "args": {
      "External id": 42792,
      "cbid": 211, "correlation": 42792
    }
  },
  {
    "ph": "s", "id": 42792, "pid": 76337, "tid": -914061504, "ts": 1716454218066737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066822, "dur": 1,
    "args": {
      "External id": 42803,
      "cbid": 251, "correlation": 42803
    }
  },
  {
    "ph": "f", "id": 42803, "pid": 76337, "tid": -914061504, "ts": 1716454218066822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218139536, "dur": 89,
    "args": {
      "External id": 42804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42804,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42804, "pid": 5, "tid": 7, "ts": 1716454218139536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066828, "dur": 14,
    "args": {
      "External id": 42804,
      "cbid": 211, "correlation": 42804
    }
  },
  {
    "ph": "s", "id": 42804, "pid": 76337, "tid": -914061504, "ts": 1716454218066828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066897, "dur": 1,
    "args": {
      "External id": 42815,
      "cbid": 251, "correlation": 42815
    }
  },
  {
    "ph": "f", "id": 42815, "pid": 76337, "tid": -914061504, "ts": 1716454218066897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066901, "dur": 0,
    "args": {
      "External id": 42816,
      "cbid": 251, "correlation": 42816
    }
  },
  {
    "ph": "f", "id": 42816, "pid": 76337, "tid": -914061504, "ts": 1716454218066901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218139627, "dur": 10,
    "args": {
      "External id": 42817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42817,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 42817, "pid": 5, "tid": 7, "ts": 1716454218139627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066902, "dur": 12,
    "args": {
      "External id": 42817,
      "cbid": 211, "correlation": 42817
    }
  },
  {
    "ph": "s", "id": 42817, "pid": 76337, "tid": -914061504, "ts": 1716454218066902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218139638, "dur": 5,
    "args": {
      "External id": 42819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42819,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 42819, "pid": 5, "tid": 7, "ts": 1716454218139638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066916, "dur": 6,
    "args": {
      "External id": 42819,
      "cbid": 211, "correlation": 42819
    }
  },
  {
    "ph": "s", "id": 42819, "pid": 76337, "tid": -914061504, "ts": 1716454218066916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066972, "dur": 10,
    "args": {
      "External id": 42830,
      "cbid": 251, "correlation": 42830
    }
  },
  {
    "ph": "f", "id": 42830, "pid": 76337, "tid": -914061504, "ts": 1716454218066972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066985, "dur": 0,
    "args": {
      "External id": 42831,
      "cbid": 251, "correlation": 42831
    }
  },
  {
    "ph": "f", "id": 42831, "pid": 76337, "tid": -914061504, "ts": 1716454218066985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218139644, "dur": 7,
    "args": {
      "External id": 42832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42832,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 42832, "pid": 5, "tid": 7, "ts": 1716454218139644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218066987, "dur": 13,
    "args": {
      "External id": 42832,
      "cbid": 211, "correlation": 42832
    }
  },
  {
    "ph": "s", "id": 42832, "pid": 76337, "tid": -914061504, "ts": 1716454218066987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218139652, "dur": 3,
    "args": {
      "External id": 42834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42834,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 42834, "pid": 5, "tid": 7, "ts": 1716454218139652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067002, "dur": 8,
    "args": {
      "External id": 42834,
      "cbid": 211, "correlation": 42834
    }
  },
  {
    "ph": "s", "id": 42834, "pid": 76337, "tid": -914061504, "ts": 1716454218067002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218139656, "dur": 91,
    "args": {
      "External id": 42855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42855,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 42855, "pid": 5, "tid": 7, "ts": 1716454218139656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067079, "dur": 13,
    "args": {
      "External id": 42855,
      "cbid": 211, "correlation": 42855
    }
  },
  {
    "ph": "s", "id": 42855, "pid": 76337, "tid": -914061504, "ts": 1716454218067079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067177, "dur": 1,
    "args": {
      "External id": 42873,
      "cbid": 251, "correlation": 42873
    }
  },
  {
    "ph": "f", "id": 42873, "pid": 76337, "tid": -914061504, "ts": 1716454218067177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218139748, "dur": 96,
    "args": {
      "External id": 42875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42875,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42875, "pid": 5, "tid": 7, "ts": 1716454218139748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067183, "dur": 13,
    "args": {
      "External id": 42875,
      "cbid": 211, "correlation": 42875
    }
  },
  {
    "ph": "s", "id": 42875, "pid": 76337, "tid": -914061504, "ts": 1716454218067183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218139846, "dur": 19,
    "args": {
      "External id": 42883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42883,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42883, "pid": 5, "tid": 7, "ts": 1716454218139846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067252, "dur": 12,
    "args": {
      "External id": 42883,
      "cbid": 211, "correlation": 42883
    }
  },
  {
    "ph": "s", "id": 42883, "pid": 76337, "tid": -914061504, "ts": 1716454218067252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218139866, "dur": 38,
    "args": {
      "External id": 42891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42891, "pid": 5, "tid": 7, "ts": 1716454218139866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067293, "dur": 10,
    "args": {
      "External id": 42891,
      "cbid": 211, "correlation": 42891
    }
  },
  {
    "ph": "s", "id": 42891, "pid": 76337, "tid": -914061504, "ts": 1716454218067293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218139905, "dur": 34,
    "args": {
      "External id": 42913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42913,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42913, "pid": 5, "tid": 7, "ts": 1716454218139905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067346, "dur": 14,
    "args": {
      "External id": 42913,
      "cbid": 211, "correlation": 42913
    }
  },
  {
    "ph": "s", "id": 42913, "pid": 76337, "tid": -914061504, "ts": 1716454218067346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067439, "dur": 1,
    "args": {
      "External id": 42929,
      "cbid": 251, "correlation": 42929
    }
  },
  {
    "ph": "f", "id": 42929, "pid": 76337, "tid": -914061504, "ts": 1716454218067439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067444, "dur": 0,
    "args": {
      "External id": 42931,
      "cbid": 251, "correlation": 42931
    }
  },
  {
    "ph": "f", "id": 42931, "pid": 76337, "tid": -914061504, "ts": 1716454218067444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218139940, "dur": 533,
    "args": {
      "External id": 42932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42932,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 42932, "pid": 5, "tid": 7, "ts": 1716454218139940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067447, "dur": 14,
    "args": {
      "External id": 42932,
      "cbid": 211, "correlation": 42932
    }
  },
  {
    "ph": "s", "id": 42932, "pid": 76337, "tid": -914061504, "ts": 1716454218067447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218140475, "dur": 123,
    "args": {
      "External id": 42940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42940,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42940, "pid": 5, "tid": 7, "ts": 1716454218140475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067514, "dur": 13,
    "args": {
      "External id": 42940,
      "cbid": 211, "correlation": 42940
    }
  },
  {
    "ph": "s", "id": 42940, "pid": 76337, "tid": -914061504, "ts": 1716454218067514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218140599, "dur": 131,
    "args": {
      "External id": 42948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42948, "pid": 5, "tid": 7, "ts": 1716454218140599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067544, "dur": 8,
    "args": {
      "External id": 42948,
      "cbid": 211, "correlation": 42948
    }
  },
  {
    "ph": "s", "id": 42948, "pid": 76337, "tid": -914061504, "ts": 1716454218067544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067620, "dur": 1,
    "args": {
      "External id": 42964,
      "cbid": 251, "correlation": 42964
    }
  },
  {
    "ph": "f", "id": 42964, "pid": 76337, "tid": -914061504, "ts": 1716454218067620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218140732, "dur": 303,
    "args": {
      "External id": 42966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42966,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 42966, "pid": 5, "tid": 7, "ts": 1716454218140732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067625, "dur": 12,
    "args": {
      "External id": 42966,
      "cbid": 211, "correlation": 42966
    }
  },
  {
    "ph": "s", "id": 42966, "pid": 76337, "tid": -914061504, "ts": 1716454218067625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218141037, "dur": 27,
    "args": {
      "External id": 42974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42974,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42974, "pid": 5, "tid": 7, "ts": 1716454218141037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067668, "dur": 9,
    "args": {
      "External id": 42974,
      "cbid": 211, "correlation": 42974
    }
  },
  {
    "ph": "s", "id": 42974, "pid": 76337, "tid": -914061504, "ts": 1716454218067668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218141065, "dur": 81,
    "args": {
      "External id": 42985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 42985,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 42985, "pid": 5, "tid": 7, "ts": 1716454218141065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067734, "dur": 12,
    "args": {
      "External id": 42985,
      "cbid": 211, "correlation": 42985
    }
  },
  {
    "ph": "s", "id": 42985, "pid": 76337, "tid": -914061504, "ts": 1716454218067734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067800, "dur": 0,
    "args": {
      "External id": 42997,
      "cbid": 317, "correlation": 42997
    }
  },
  {
    "ph": "f", "id": 42997, "pid": 76337, "tid": -914061504, "ts": 1716454218067800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067801, "dur": 0,
    "args": {
      "External id": 42998,
      "cbid": 203, "correlation": 42998
    }
  },
  {
    "ph": "f", "id": 42998, "pid": 76337, "tid": -914061504, "ts": 1716454218067801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067802, "dur": 0,
    "args": {
      "External id": 42999,
      "cbid": 205, "correlation": 42999
    }
  },
  {
    "ph": "f", "id": 42999, "pid": 76337, "tid": -914061504, "ts": 1716454218067802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218141146, "dur": 23,
    "args": {
      "External id": 43003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43003,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43003, "pid": 5, "tid": 7, "ts": 1716454218141146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067817, "dur": 12,
    "args": {
      "External id": 43003,
      "cbid": 211, "correlation": 43003
    }
  },
  {
    "ph": "s", "id": 43003, "pid": 76337, "tid": -914061504, "ts": 1716454218067817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218141171, "dur": 117,
    "args": {
      "External id": 43005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43005,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43005, "pid": 5, "tid": 7, "ts": 1716454218141171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067835, "dur": 6,
    "args": {
      "External id": 43005,
      "cbid": 211, "correlation": 43005
    }
  },
  {
    "ph": "s", "id": 43005, "pid": 76337, "tid": -914061504, "ts": 1716454218067835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218141289, "dur": 24,
    "args": {
      "External id": 43007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43007,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43007, "pid": 5, "tid": 7, "ts": 1716454218141289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067846, "dur": 5,
    "args": {
      "External id": 43007,
      "cbid": 211, "correlation": 43007
    }
  },
  {
    "ph": "s", "id": 43007, "pid": 76337, "tid": -914061504, "ts": 1716454218067846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218141314, "dur": 32,
    "args": {
      "External id": 43013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43013, "pid": 5, "tid": 7, "ts": 1716454218141314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067874, "dur": 8,
    "args": {
      "External id": 43013,
      "cbid": 211, "correlation": 43013
    }
  },
  {
    "ph": "s", "id": 43013, "pid": 76337, "tid": -914061504, "ts": 1716454218067874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218141348, "dur": 27,
    "args": {
      "External id": 43021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43021,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43021, "pid": 5, "tid": 7, "ts": 1716454218141348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067905, "dur": 9,
    "args": {
      "External id": 43021,
      "cbid": 211, "correlation": 43021
    }
  },
  {
    "ph": "s", "id": 43021, "pid": 76337, "tid": -914061504, "ts": 1716454218067905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218141376, "dur": 100,
    "args": {
      "External id": 43032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43032,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43032, "pid": 5, "tid": 7, "ts": 1716454218141376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218067991, "dur": 15,
    "args": {
      "External id": 43032,
      "cbid": 211, "correlation": 43032
    }
  },
  {
    "ph": "s", "id": 43032, "pid": 76337, "tid": -914061504, "ts": 1716454218067991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068050, "dur": 0,
    "args": {
      "External id": 43042,
      "cbid": 317, "correlation": 43042
    }
  },
  {
    "ph": "f", "id": 43042, "pid": 76337, "tid": -914061504, "ts": 1716454218068050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068051, "dur": 0,
    "args": {
      "External id": 43043,
      "cbid": 203, "correlation": 43043
    }
  },
  {
    "ph": "f", "id": 43043, "pid": 76337, "tid": -914061504, "ts": 1716454218068051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068052, "dur": 0,
    "args": {
      "External id": 43044,
      "cbid": 205, "correlation": 43044
    }
  },
  {
    "ph": "f", "id": 43044, "pid": 76337, "tid": -914061504, "ts": 1716454218068052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218141477, "dur": 74,
    "args": {
      "External id": 43048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43048, "pid": 5, "tid": 7, "ts": 1716454218141477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068071, "dur": 13,
    "args": {
      "External id": 43048,
      "cbid": 211, "correlation": 43048
    }
  },
  {
    "ph": "s", "id": 43048, "pid": 76337, "tid": -914061504, "ts": 1716454218068071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218141552, "dur": 43,
    "args": {
      "External id": 43050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43050, "pid": 5, "tid": 7, "ts": 1716454218141552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068086, "dur": 5,
    "args": {
      "External id": 43050,
      "cbid": 211, "correlation": 43050
    }
  },
  {
    "ph": "s", "id": 43050, "pid": 76337, "tid": -914061504, "ts": 1716454218068086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218141597, "dur": 4,
    "args": {
      "External id": 43052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43052, "pid": 5, "tid": 7, "ts": 1716454218141597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068098, "dur": 9,
    "args": {
      "External id": 43052,
      "cbid": 211, "correlation": 43052
    }
  },
  {
    "ph": "s", "id": 43052, "pid": 76337, "tid": -914061504, "ts": 1716454218068098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068113, "dur": 0,
    "args": {
      "External id": 43053,
      "cbid": 51, "correlation": 43053
    }
  },
  {
    "ph": "s", "id": 43053, "pid": 76337, "tid": -914061504, "ts": 1716454218068113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218141602, "dur": 2214,
    "args": {
      "External id": 43054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43054,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43054, "pid": 5, "tid": 7, "ts": 1716454218141602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068114, "dur": 8,
    "args": {
      "External id": 43054,
      "cbid": 211, "correlation": 43054
    }
  },
  {
    "ph": "s", "id": 43054, "pid": 76337, "tid": -914061504, "ts": 1716454218068114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218143818, "dur": 112,
    "args": {
      "External id": 43059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43059, "pid": 5, "tid": 7, "ts": 1716454218143818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068146, "dur": 8,
    "args": {
      "External id": 43059,
      "cbid": 211, "correlation": 43059
    }
  },
  {
    "ph": "s", "id": 43059, "pid": 76337, "tid": -914061504, "ts": 1716454218068146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218143931, "dur": 164,
    "args": {
      "External id": 43068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43068,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43068, "pid": 5, "tid": 7, "ts": 1716454218143931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068251, "dur": 13,
    "args": {
      "External id": 43068,
      "cbid": 211, "correlation": 43068
    }
  },
  {
    "ph": "s", "id": 43068, "pid": 76337, "tid": -914061504, "ts": 1716454218068251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218144096, "dur": 127,
    "args": {
      "External id": 43088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43088,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 43088, "pid": 5, "tid": 7, "ts": 1716454218144096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068322, "dur": 12,
    "args": {
      "External id": 43088,
      "cbid": 211, "correlation": 43088
    }
  },
  {
    "ph": "s", "id": 43088, "pid": 76337, "tid": -914061504, "ts": 1716454218068322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218144224, "dur": 5,
    "args": {
      "External id": 43100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43100,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 43100, "pid": 5, "tid": 7, "ts": 1716454218144224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068344, "dur": 6,
    "args": {
      "External id": 43100,
      "cbid": 211, "correlation": 43100
    }
  },
  {
    "ph": "s", "id": 43100, "pid": 76337, "tid": -914061504, "ts": 1716454218068344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218144230, "dur": 159,
    "args": {
      "External id": 43103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43103,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43103, "pid": 5, "tid": 7, "ts": 1716454218144230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068363, "dur": 6,
    "args": {
      "External id": 43103,
      "cbid": 211, "correlation": 43103
    }
  },
  {
    "ph": "s", "id": 43103, "pid": 76337, "tid": -914061504, "ts": 1716454218068363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218144390, "dur": 101,
    "args": {
      "External id": 43112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43112,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43112, "pid": 5, "tid": 7, "ts": 1716454218144390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068403, "dur": 10,
    "args": {
      "External id": 43112,
      "cbid": 211, "correlation": 43112
    }
  },
  {
    "ph": "s", "id": 43112, "pid": 76337, "tid": -914061504, "ts": 1716454218068403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068460, "dur": 0,
    "args": {
      "External id": 43122,
      "cbid": 317, "correlation": 43122
    }
  },
  {
    "ph": "f", "id": 43122, "pid": 76337, "tid": -914061504, "ts": 1716454218068460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068461, "dur": 0,
    "args": {
      "External id": 43123,
      "cbid": 203, "correlation": 43123
    }
  },
  {
    "ph": "f", "id": 43123, "pid": 76337, "tid": -914061504, "ts": 1716454218068461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068461, "dur": 0,
    "args": {
      "External id": 43124,
      "cbid": 205, "correlation": 43124
    }
  },
  {
    "ph": "f", "id": 43124, "pid": 76337, "tid": -914061504, "ts": 1716454218068461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218144493, "dur": 111,
    "args": {
      "External id": 43128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43128,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43128, "pid": 5, "tid": 7, "ts": 1716454218144493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068480, "dur": 13,
    "args": {
      "External id": 43128,
      "cbid": 211, "correlation": 43128
    }
  },
  {
    "ph": "s", "id": 43128, "pid": 76337, "tid": -914061504, "ts": 1716454218068480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218144605, "dur": 33,
    "args": {
      "External id": 43130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43130, "pid": 5, "tid": 7, "ts": 1716454218144605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068495, "dur": 5,
    "args": {
      "External id": 43130,
      "cbid": 211, "correlation": 43130
    }
  },
  {
    "ph": "s", "id": 43130, "pid": 76337, "tid": -914061504, "ts": 1716454218068495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218144639, "dur": 3,
    "args": {
      "External id": 43132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43132, "pid": 5, "tid": 7, "ts": 1716454218144639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068506, "dur": 6,
    "args": {
      "External id": 43132,
      "cbid": 211, "correlation": 43132
    }
  },
  {
    "ph": "s", "id": 43132, "pid": 76337, "tid": -914061504, "ts": 1716454218068506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068516, "dur": 0,
    "args": {
      "External id": 43133,
      "cbid": 51, "correlation": 43133
    }
  },
  {
    "ph": "s", "id": 43133, "pid": 76337, "tid": -914061504, "ts": 1716454218068516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218144644, "dur": 1978,
    "args": {
      "External id": 43134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43134,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43134, "pid": 5, "tid": 7, "ts": 1716454218144644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068517, "dur": 6,
    "args": {
      "External id": 43134,
      "cbid": 211, "correlation": 43134
    }
  },
  {
    "ph": "s", "id": 43134, "pid": 76337, "tid": -914061504, "ts": 1716454218068517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218146624, "dur": 59,
    "args": {
      "External id": 43139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43139,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43139, "pid": 5, "tid": 7, "ts": 1716454218146624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068547, "dur": 8,
    "args": {
      "External id": 43139,
      "cbid": 211, "correlation": 43139
    }
  },
  {
    "ph": "s", "id": 43139, "pid": 76337, "tid": -914061504, "ts": 1716454218068547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218146684, "dur": 3,
    "args": {
      "External id": 43147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43147,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43147, "pid": 5, "tid": 7, "ts": 1716454218146684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068591, "dur": 10,
    "args": {
      "External id": 43147,
      "cbid": 211, "correlation": 43147
    }
  },
  {
    "ph": "s", "id": 43147, "pid": 76337, "tid": -914061504, "ts": 1716454218068591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068658, "dur": 1,
    "args": {
      "External id": 43163,
      "cbid": 251, "correlation": 43163
    }
  },
  {
    "ph": "f", "id": 43163, "pid": 76337, "tid": -914061504, "ts": 1716454218068658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068663, "dur": 0,
    "args": {
      "External id": 43165,
      "cbid": 251, "correlation": 43165
    }
  },
  {
    "ph": "f", "id": 43165, "pid": 76337, "tid": -914061504, "ts": 1716454218068663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218146689, "dur": 11,
    "args": {
      "External id": 43166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43166,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 43166, "pid": 5, "tid": 7, "ts": 1716454218146689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068665, "dur": 11,
    "args": {
      "External id": 43166,
      "cbid": 211, "correlation": 43166
    }
  },
  {
    "ph": "s", "id": 43166, "pid": 76337, "tid": -914061504, "ts": 1716454218068665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218146701, "dur": 5,
    "args": {
      "External id": 43168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43168,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 43168, "pid": 5, "tid": 7, "ts": 1716454218146701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068678, "dur": 5,
    "args": {
      "External id": 43168,
      "cbid": 211, "correlation": 43168
    }
  },
  {
    "ph": "s", "id": 43168, "pid": 76337, "tid": -914061504, "ts": 1716454218068678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218146707, "dur": 53,
    "args": {
      "External id": 43178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43178, "pid": 5, "tid": 7, "ts": 1716454218146707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068737, "dur": 12,
    "args": {
      "External id": 43178,
      "cbid": 211, "correlation": 43178
    }
  },
  {
    "ph": "s", "id": 43178, "pid": 76337, "tid": -914061504, "ts": 1716454218068737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218146761, "dur": 50,
    "args": {
      "External id": 43198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43198,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 43198, "pid": 5, "tid": 7, "ts": 1716454218146761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068807, "dur": 11,
    "args": {
      "External id": 43198,
      "cbid": 211, "correlation": 43198
    }
  },
  {
    "ph": "s", "id": 43198, "pid": 76337, "tid": -914061504, "ts": 1716454218068807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218146813, "dur": 4,
    "args": {
      "External id": 43210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43210,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43210, "pid": 5, "tid": 7, "ts": 1716454218146813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068828, "dur": 6,
    "args": {
      "External id": 43210,
      "cbid": 211, "correlation": 43210
    }
  },
  {
    "ph": "s", "id": 43210, "pid": 76337, "tid": -914061504, "ts": 1716454218068828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218146818, "dur": 55,
    "args": {
      "External id": 43213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43213, "pid": 5, "tid": 7, "ts": 1716454218146818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068847, "dur": 7,
    "args": {
      "External id": 43213,
      "cbid": 211, "correlation": 43213
    }
  },
  {
    "ph": "s", "id": 43213, "pid": 76337, "tid": -914061504, "ts": 1716454218068847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218146874, "dur": 36,
    "args": {
      "External id": 43222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43222,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43222, "pid": 5, "tid": 7, "ts": 1716454218146874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068888, "dur": 10,
    "args": {
      "External id": 43222,
      "cbid": 211, "correlation": 43222
    }
  },
  {
    "ph": "s", "id": 43222, "pid": 76337, "tid": -914061504, "ts": 1716454218068888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068951, "dur": 0,
    "args": {
      "External id": 43232,
      "cbid": 317, "correlation": 43232
    }
  },
  {
    "ph": "f", "id": 43232, "pid": 76337, "tid": -914061504, "ts": 1716454218068951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068952, "dur": 0,
    "args": {
      "External id": 43233,
      "cbid": 203, "correlation": 43233
    }
  },
  {
    "ph": "f", "id": 43233, "pid": 76337, "tid": -914061504, "ts": 1716454218068952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068952, "dur": 0,
    "args": {
      "External id": 43234,
      "cbid": 205, "correlation": 43234
    }
  },
  {
    "ph": "f", "id": 43234, "pid": 76337, "tid": -914061504, "ts": 1716454218068952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218146912, "dur": 39,
    "args": {
      "External id": 43238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43238,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43238, "pid": 5, "tid": 7, "ts": 1716454218146912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068971, "dur": 21,
    "args": {
      "External id": 43238,
      "cbid": 211, "correlation": 43238
    }
  },
  {
    "ph": "s", "id": 43238, "pid": 76337, "tid": -914061504, "ts": 1716454218068971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218146952, "dur": 14,
    "args": {
      "External id": 43240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43240, "pid": 5, "tid": 7, "ts": 1716454218146952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218068995, "dur": 5,
    "args": {
      "External id": 43240,
      "cbid": 211, "correlation": 43240
    }
  },
  {
    "ph": "s", "id": 43240, "pid": 76337, "tid": -914061504, "ts": 1716454218068995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218146968, "dur": 3,
    "args": {
      "External id": 43242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43242, "pid": 5, "tid": 7, "ts": 1716454218146968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069005, "dur": 5,
    "args": {
      "External id": 43242,
      "cbid": 211, "correlation": 43242
    }
  },
  {
    "ph": "s", "id": 43242, "pid": 76337, "tid": -914061504, "ts": 1716454218069005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069014, "dur": 0,
    "args": {
      "External id": 43243,
      "cbid": 51, "correlation": 43243
    }
  },
  {
    "ph": "s", "id": 43243, "pid": 76337, "tid": -914061504, "ts": 1716454218069014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218146973, "dur": 687,
    "args": {
      "External id": 43244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43244,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43244, "pid": 5, "tid": 7, "ts": 1716454218146973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069014, "dur": 6,
    "args": {
      "External id": 43244,
      "cbid": 211, "correlation": 43244
    }
  },
  {
    "ph": "s", "id": 43244, "pid": 76337, "tid": -914061504, "ts": 1716454218069014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218147661, "dur": 59,
    "args": {
      "External id": 43249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43249,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43249, "pid": 5, "tid": 7, "ts": 1716454218147661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069043, "dur": 12,
    "args": {
      "External id": 43249,
      "cbid": 211, "correlation": 43249
    }
  },
  {
    "ph": "s", "id": 43249, "pid": 76337, "tid": -914061504, "ts": 1716454218069043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069104, "dur": 0,
    "args": {
      "External id": 43259,
      "cbid": 317, "correlation": 43259
    }
  },
  {
    "ph": "f", "id": 43259, "pid": 76337, "tid": -914061504, "ts": 1716454218069104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069105, "dur": 0,
    "args": {
      "External id": 43260,
      "cbid": 203, "correlation": 43260
    }
  },
  {
    "ph": "f", "id": 43260, "pid": 76337, "tid": -914061504, "ts": 1716454218069105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069106, "dur": 0,
    "args": {
      "External id": 43261,
      "cbid": 205, "correlation": 43261
    }
  },
  {
    "ph": "f", "id": 43261, "pid": 76337, "tid": -914061504, "ts": 1716454218069106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218147721, "dur": 3,
    "args": {
      "External id": 43265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43265, "pid": 5, "tid": 7, "ts": 1716454218147721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069126, "dur": 11,
    "args": {
      "External id": 43265,
      "cbid": 211, "correlation": 43265
    }
  },
  {
    "ph": "s", "id": 43265, "pid": 76337, "tid": -914061504, "ts": 1716454218069126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069142, "dur": 0,
    "args": {
      "External id": 43266,
      "cbid": 51, "correlation": 43266
    }
  },
  {
    "ph": "s", "id": 43266, "pid": 76337, "tid": -914061504, "ts": 1716454218069142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218147725, "dur": 261,
    "args": {
      "External id": 43267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43267,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43267, "pid": 5, "tid": 7, "ts": 1716454218147725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069143, "dur": 6,
    "args": {
      "External id": 43267,
      "cbid": 211, "correlation": 43267
    }
  },
  {
    "ph": "s", "id": 43267, "pid": 76337, "tid": -914061504, "ts": 1716454218069143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218147987, "dur": 58,
    "args": {
      "External id": 43272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43272, "pid": 5, "tid": 7, "ts": 1716454218147987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069170, "dur": 9,
    "args": {
      "External id": 43272,
      "cbid": 211, "correlation": 43272
    }
  },
  {
    "ph": "s", "id": 43272, "pid": 76337, "tid": -914061504, "ts": 1716454218069170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218148047, "dur": 50,
    "args": {
      "External id": 43280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43280,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43280, "pid": 5, "tid": 7, "ts": 1716454218148047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069199, "dur": 8,
    "args": {
      "External id": 43280,
      "cbid": 211, "correlation": 43280
    }
  },
  {
    "ph": "s", "id": 43280, "pid": 76337, "tid": -914061504, "ts": 1716454218069199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218148098, "dur": 35,
    "args": {
      "External id": 43288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43288,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43288, "pid": 5, "tid": 7, "ts": 1716454218148098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069229, "dur": 9,
    "args": {
      "External id": 43288,
      "cbid": 211, "correlation": 43288
    }
  },
  {
    "ph": "s", "id": 43288, "pid": 76337, "tid": -914061504, "ts": 1716454218069229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218148134, "dur": 52,
    "args": {
      "External id": 43308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43308,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 43308, "pid": 5, "tid": 7, "ts": 1716454218148134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069311, "dur": 12,
    "args": {
      "External id": 43308,
      "cbid": 211, "correlation": 43308
    }
  },
  {
    "ph": "s", "id": 43308, "pid": 76337, "tid": -914061504, "ts": 1716454218069311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218148188, "dur": 4,
    "args": {
      "External id": 43320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43320,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43320, "pid": 5, "tid": 7, "ts": 1716454218148188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069335, "dur": 9,
    "args": {
      "External id": 43320,
      "cbid": 211, "correlation": 43320
    }
  },
  {
    "ph": "s", "id": 43320, "pid": 76337, "tid": -914061504, "ts": 1716454218069335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218148193, "dur": 55,
    "args": {
      "External id": 43323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43323, "pid": 5, "tid": 7, "ts": 1716454218148193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069356, "dur": 7,
    "args": {
      "External id": 43323,
      "cbid": 211, "correlation": 43323
    }
  },
  {
    "ph": "s", "id": 43323, "pid": 76337, "tid": -914061504, "ts": 1716454218069356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069413, "dur": 0,
    "args": {
      "External id": 43334,
      "cbid": 317, "correlation": 43334
    }
  },
  {
    "ph": "f", "id": 43334, "pid": 76337, "tid": -914061504, "ts": 1716454218069413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069414, "dur": 0,
    "args": {
      "External id": 43335,
      "cbid": 203, "correlation": 43335
    }
  },
  {
    "ph": "f", "id": 43335, "pid": 76337, "tid": -914061504, "ts": 1716454218069414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069415, "dur": 0,
    "args": {
      "External id": 43336,
      "cbid": 205, "correlation": 43336
    }
  },
  {
    "ph": "f", "id": 43336, "pid": 76337, "tid": -914061504, "ts": 1716454218069415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069449, "dur": 3,
    "args": {
      "External id": 43340,
      "cbid": 251, "correlation": 43340
    }
  },
  {
    "ph": "f", "id": 43340, "pid": 76337, "tid": -914061504, "ts": 1716454218069449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069454, "dur": 1,
    "args": {
      "External id": 43341,
      "cbid": 251, "correlation": 43341
    }
  },
  {
    "ph": "f", "id": 43341, "pid": 76337, "tid": -914061504, "ts": 1716454218069454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069455, "dur": 1,
    "args": {
      "External id": 43342,
      "cbid": 251, "correlation": 43342
    }
  },
  {
    "ph": "f", "id": 43342, "pid": 76337, "tid": -914061504, "ts": 1716454218069455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069457, "dur": 1,
    "args": {
      "External id": 43343,
      "cbid": 251, "correlation": 43343
    }
  },
  {
    "ph": "f", "id": 43343, "pid": 76337, "tid": -914061504, "ts": 1716454218069457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069459, "dur": 1,
    "args": {
      "External id": 43344,
      "cbid": 251, "correlation": 43344
    }
  },
  {
    "ph": "f", "id": 43344, "pid": 76337, "tid": -914061504, "ts": 1716454218069459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069461, "dur": 1,
    "args": {
      "External id": 43345,
      "cbid": 251, "correlation": 43345
    }
  },
  {
    "ph": "f", "id": 43345, "pid": 76337, "tid": -914061504, "ts": 1716454218069461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069463, "dur": 1,
    "args": {
      "External id": 43346,
      "cbid": 251, "correlation": 43346
    }
  },
  {
    "ph": "f", "id": 43346, "pid": 76337, "tid": -914061504, "ts": 1716454218069463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069465, "dur": 1,
    "args": {
      "External id": 43347,
      "cbid": 251, "correlation": 43347
    }
  },
  {
    "ph": "f", "id": 43347, "pid": 76337, "tid": -914061504, "ts": 1716454218069465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069467, "dur": 0,
    "args": {
      "External id": 43348,
      "cbid": 251, "correlation": 43348
    }
  },
  {
    "ph": "f", "id": 43348, "pid": 76337, "tid": -914061504, "ts": 1716454218069467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218148249, "dur": 114,
    "args": {
      "External id": 43349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43349,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 43349, "pid": 5, "tid": 7, "ts": 1716454218148249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069470, "dur": 13,
    "args": {
      "External id": 43349,
      "cbid": 211, "correlation": 43349
    }
  },
  {
    "ph": "s", "id": 43349, "pid": 76337, "tid": -914061504, "ts": 1716454218069470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218148364, "dur": 58,
    "args": {
      "External id": 43355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43355, "pid": 5, "tid": 7, "ts": 1716454218148364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069507, "dur": 9,
    "args": {
      "External id": 43355,
      "cbid": 211, "correlation": 43355
    }
  },
  {
    "ph": "s", "id": 43355, "pid": 76337, "tid": -914061504, "ts": 1716454218069507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218148423, "dur": 575,
    "args": {
      "External id": 43364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43364,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43364, "pid": 5, "tid": 7, "ts": 1716454218148423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069591, "dur": 14,
    "args": {
      "External id": 43364,
      "cbid": 211, "correlation": 43364
    }
  },
  {
    "ph": "s", "id": 43364, "pid": 76337, "tid": -914061504, "ts": 1716454218069591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218149000, "dur": 177,
    "args": {
      "External id": 43386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43386,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43386, "pid": 5, "tid": 7, "ts": 1716454218149000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069649, "dur": 10,
    "args": {
      "External id": 43386,
      "cbid": 211, "correlation": 43386
    }
  },
  {
    "ph": "s", "id": 43386, "pid": 76337, "tid": -914061504, "ts": 1716454218069649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069738, "dur": 1,
    "args": {
      "External id": 43397,
      "cbid": 251, "correlation": 43397
    }
  },
  {
    "ph": "f", "id": 43397, "pid": 76337, "tid": -914061504, "ts": 1716454218069738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218149178, "dur": 195,
    "args": {
      "External id": 43398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43398,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43398, "pid": 5, "tid": 7, "ts": 1716454218149178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069743, "dur": 17,
    "args": {
      "External id": 43398,
      "cbid": 211, "correlation": 43398
    }
  },
  {
    "ph": "s", "id": 43398, "pid": 76337, "tid": -914061504, "ts": 1716454218069743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069817, "dur": 1,
    "args": {
      "External id": 43409,
      "cbid": 251, "correlation": 43409
    }
  },
  {
    "ph": "f", "id": 43409, "pid": 76337, "tid": -914061504, "ts": 1716454218069817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218149374, "dur": 182,
    "args": {
      "External id": 43410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43410,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43410, "pid": 5, "tid": 7, "ts": 1716454218149374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069821, "dur": 12,
    "args": {
      "External id": 43410,
      "cbid": 211, "correlation": 43410
    }
  },
  {
    "ph": "s", "id": 43410, "pid": 76337, "tid": -914061504, "ts": 1716454218069821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069885, "dur": 1,
    "args": {
      "External id": 43421,
      "cbid": 251, "correlation": 43421
    }
  },
  {
    "ph": "f", "id": 43421, "pid": 76337, "tid": -914061504, "ts": 1716454218069885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218149557, "dur": 182,
    "args": {
      "External id": 43422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43422,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43422, "pid": 5, "tid": 7, "ts": 1716454218149557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069889, "dur": 12,
    "args": {
      "External id": 43422,
      "cbid": 211, "correlation": 43422
    }
  },
  {
    "ph": "s", "id": 43422, "pid": 76337, "tid": -914061504, "ts": 1716454218069889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218149741, "dur": 18285,
    "args": {
      "External id": 43443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43443,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 43443, "pid": 5, "tid": 7, "ts": 1716454218149741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218069971, "dur": 23,
    "args": {
      "External id": 43443,
      "cbid": 211, "correlation": 43443
    }
  },
  {
    "ph": "s", "id": 43443, "pid": 76337, "tid": -914061504, "ts": 1716454218069971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070081, "dur": 1,
    "args": {
      "External id": 43461,
      "cbid": 251, "correlation": 43461
    }
  },
  {
    "ph": "f", "id": 43461, "pid": 76337, "tid": -914061504, "ts": 1716454218070081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218168028, "dur": 203,
    "args": {
      "External id": 43463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43463,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43463, "pid": 5, "tid": 7, "ts": 1716454218168028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070087, "dur": 13,
    "args": {
      "External id": 43463,
      "cbid": 211, "correlation": 43463
    }
  },
  {
    "ph": "s", "id": 43463, "pid": 76337, "tid": -914061504, "ts": 1716454218070087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218168232, "dur": 66,
    "args": {
      "External id": 43471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43471,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43471, "pid": 5, "tid": 7, "ts": 1716454218168232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070157, "dur": 12,
    "args": {
      "External id": 43471,
      "cbid": 211, "correlation": 43471
    }
  },
  {
    "ph": "s", "id": 43471, "pid": 76337, "tid": -914061504, "ts": 1716454218070157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218168300, "dur": 97,
    "args": {
      "External id": 43479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43479,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43479, "pid": 5, "tid": 7, "ts": 1716454218168300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070199, "dur": 9,
    "args": {
      "External id": 43479,
      "cbid": 211, "correlation": 43479
    }
  },
  {
    "ph": "s", "id": 43479, "pid": 76337, "tid": -914061504, "ts": 1716454218070199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218168398, "dur": 55,
    "args": {
      "External id": 43490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43490, "pid": 5, "tid": 7, "ts": 1716454218168398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070271, "dur": 12,
    "args": {
      "External id": 43490,
      "cbid": 211, "correlation": 43490
    }
  },
  {
    "ph": "s", "id": 43490, "pid": 76337, "tid": -914061504, "ts": 1716454218070271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218168454, "dur": 91,
    "args": {
      "External id": 43512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43512,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43512, "pid": 5, "tid": 7, "ts": 1716454218168454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070302, "dur": 8,
    "args": {
      "External id": 43512,
      "cbid": 211, "correlation": 43512
    }
  },
  {
    "ph": "s", "id": 43512, "pid": 76337, "tid": -914061504, "ts": 1716454218070302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070386, "dur": 1,
    "args": {
      "External id": 43523,
      "cbid": 251, "correlation": 43523
    }
  },
  {
    "ph": "f", "id": 43523, "pid": 76337, "tid": -914061504, "ts": 1716454218070386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218168546, "dur": 106,
    "args": {
      "External id": 43524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43524,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43524, "pid": 5, "tid": 7, "ts": 1716454218168546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070391, "dur": 12,
    "args": {
      "External id": 43524,
      "cbid": 211, "correlation": 43524
    }
  },
  {
    "ph": "s", "id": 43524, "pid": 76337, "tid": -914061504, "ts": 1716454218070391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070468, "dur": 1,
    "args": {
      "External id": 43535,
      "cbid": 251, "correlation": 43535
    }
  },
  {
    "ph": "f", "id": 43535, "pid": 76337, "tid": -914061504, "ts": 1716454218070468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070472, "dur": 0,
    "args": {
      "External id": 43536,
      "cbid": 251, "correlation": 43536
    }
  },
  {
    "ph": "f", "id": 43536, "pid": 76337, "tid": -914061504, "ts": 1716454218070472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218168653, "dur": 10,
    "args": {
      "External id": 43537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43537,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 43537, "pid": 5, "tid": 7, "ts": 1716454218168653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070474, "dur": 14,
    "args": {
      "External id": 43537,
      "cbid": 211, "correlation": 43537
    }
  },
  {
    "ph": "s", "id": 43537, "pid": 76337, "tid": -914061504, "ts": 1716454218070474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218168665, "dur": 5,
    "args": {
      "External id": 43539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43539,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 43539, "pid": 5, "tid": 7, "ts": 1716454218168665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070492, "dur": 8,
    "args": {
      "External id": 43539,
      "cbid": 211, "correlation": 43539
    }
  },
  {
    "ph": "s", "id": 43539, "pid": 76337, "tid": -914061504, "ts": 1716454218070492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070555, "dur": 1,
    "args": {
      "External id": 43550,
      "cbid": 251, "correlation": 43550
    }
  },
  {
    "ph": "f", "id": 43550, "pid": 76337, "tid": -914061504, "ts": 1716454218070555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070558, "dur": 3,
    "args": {
      "External id": 43551,
      "cbid": 251, "correlation": 43551
    }
  },
  {
    "ph": "f", "id": 43551, "pid": 76337, "tid": -914061504, "ts": 1716454218070558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218168672, "dur": 6,
    "args": {
      "External id": 43552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43552,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 43552, "pid": 5, "tid": 7, "ts": 1716454218168672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070563, "dur": 12,
    "args": {
      "External id": 43552,
      "cbid": 211, "correlation": 43552
    }
  },
  {
    "ph": "s", "id": 43552, "pid": 76337, "tid": -914061504, "ts": 1716454218070563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218168679, "dur": 4,
    "args": {
      "External id": 43554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43554,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 43554, "pid": 5, "tid": 7, "ts": 1716454218168679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070577, "dur": 5,
    "args": {
      "External id": 43554,
      "cbid": 211, "correlation": 43554
    }
  },
  {
    "ph": "s", "id": 43554, "pid": 76337, "tid": -914061504, "ts": 1716454218070577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218168684, "dur": 153,
    "args": {
      "External id": 43575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43575,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 43575, "pid": 5, "tid": 7, "ts": 1716454218168684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070650, "dur": 12,
    "args": {
      "External id": 43575,
      "cbid": 211, "correlation": 43575
    }
  },
  {
    "ph": "s", "id": 43575, "pid": 76337, "tid": -914061504, "ts": 1716454218070650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070747, "dur": 2,
    "args": {
      "External id": 43593,
      "cbid": 251, "correlation": 43593
    }
  },
  {
    "ph": "f", "id": 43593, "pid": 76337, "tid": -914061504, "ts": 1716454218070747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218168838, "dur": 105,
    "args": {
      "External id": 43595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43595,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 43595, "pid": 5, "tid": 7, "ts": 1716454218168838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070754, "dur": 14,
    "args": {
      "External id": 43595,
      "cbid": 211, "correlation": 43595
    }
  },
  {
    "ph": "s", "id": 43595, "pid": 76337, "tid": -914061504, "ts": 1716454218070754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218168944, "dur": 35,
    "args": {
      "External id": 43603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43603,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43603, "pid": 5, "tid": 7, "ts": 1716454218168944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070823, "dur": 12,
    "args": {
      "External id": 43603,
      "cbid": 211, "correlation": 43603
    }
  },
  {
    "ph": "s", "id": 43603, "pid": 76337, "tid": -914061504, "ts": 1716454218070823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218168981, "dur": 67,
    "args": {
      "External id": 43611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43611, "pid": 5, "tid": 7, "ts": 1716454218168981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070864, "dur": 10,
    "args": {
      "External id": 43611,
      "cbid": 211, "correlation": 43611
    }
  },
  {
    "ph": "s", "id": 43611, "pid": 76337, "tid": -914061504, "ts": 1716454218070864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218169049, "dur": 91,
    "args": {
      "External id": 43633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43633,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43633, "pid": 5, "tid": 7, "ts": 1716454218169049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218070918, "dur": 11,
    "args": {
      "External id": 43633,
      "cbid": 211, "correlation": 43633
    }
  },
  {
    "ph": "s", "id": 43633, "pid": 76337, "tid": -914061504, "ts": 1716454218070918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071013, "dur": 1,
    "args": {
      "External id": 43649,
      "cbid": 251, "correlation": 43649
    }
  },
  {
    "ph": "f", "id": 43649, "pid": 76337, "tid": -914061504, "ts": 1716454218071013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218169141, "dur": 566,
    "args": {
      "External id": 43651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43651,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43651, "pid": 5, "tid": 7, "ts": 1716454218169141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071019, "dur": 14,
    "args": {
      "External id": 43651,
      "cbid": 211, "correlation": 43651
    }
  },
  {
    "ph": "s", "id": 43651, "pid": 76337, "tid": -914061504, "ts": 1716454218071019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218169708, "dur": 239,
    "args": {
      "External id": 43659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43659,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43659, "pid": 5, "tid": 7, "ts": 1716454218169708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071086, "dur": 12,
    "args": {
      "External id": 43659,
      "cbid": 211, "correlation": 43659
    }
  },
  {
    "ph": "s", "id": 43659, "pid": 76337, "tid": -914061504, "ts": 1716454218071086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218169949, "dur": 249,
    "args": {
      "External id": 43667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43667, "pid": 5, "tid": 7, "ts": 1716454218169949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071116, "dur": 8,
    "args": {
      "External id": 43667,
      "cbid": 211, "correlation": 43667
    }
  },
  {
    "ph": "s", "id": 43667, "pid": 76337, "tid": -914061504, "ts": 1716454218071116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071199, "dur": 1,
    "args": {
      "External id": 43683,
      "cbid": 251, "correlation": 43683
    }
  },
  {
    "ph": "f", "id": 43683, "pid": 76337, "tid": -914061504, "ts": 1716454218071199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071204, "dur": 0,
    "args": {
      "External id": 43685,
      "cbid": 251, "correlation": 43685
    }
  },
  {
    "ph": "f", "id": 43685, "pid": 76337, "tid": -914061504, "ts": 1716454218071204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218170199, "dur": 355,
    "args": {
      "External id": 43686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43686,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 43686, "pid": 5, "tid": 7, "ts": 1716454218170199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071207, "dur": 13,
    "args": {
      "External id": 43686,
      "cbid": 211, "correlation": 43686
    }
  },
  {
    "ph": "s", "id": 43686, "pid": 76337, "tid": -914061504, "ts": 1716454218071207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218170555, "dur": 50,
    "args": {
      "External id": 43694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43694,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43694, "pid": 5, "tid": 7, "ts": 1716454218170555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071249, "dur": 10,
    "args": {
      "External id": 43694,
      "cbid": 211, "correlation": 43694
    }
  },
  {
    "ph": "s", "id": 43694, "pid": 76337, "tid": -914061504, "ts": 1716454218071249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218170607, "dur": 157,
    "args": {
      "External id": 43705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43705, "pid": 5, "tid": 7, "ts": 1716454218170607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071319, "dur": 13,
    "args": {
      "External id": 43705,
      "cbid": 211, "correlation": 43705
    }
  },
  {
    "ph": "s", "id": 43705, "pid": 76337, "tid": -914061504, "ts": 1716454218071319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071385, "dur": 0,
    "args": {
      "External id": 43717,
      "cbid": 317, "correlation": 43717
    }
  },
  {
    "ph": "f", "id": 43717, "pid": 76337, "tid": -914061504, "ts": 1716454218071385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071386, "dur": 0,
    "args": {
      "External id": 43718,
      "cbid": 203, "correlation": 43718
    }
  },
  {
    "ph": "f", "id": 43718, "pid": 76337, "tid": -914061504, "ts": 1716454218071386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071387, "dur": 0,
    "args": {
      "External id": 43719,
      "cbid": 205, "correlation": 43719
    }
  },
  {
    "ph": "f", "id": 43719, "pid": 76337, "tid": -914061504, "ts": 1716454218071387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071446, "dur": 1,
    "args": {
      "External id": 43723,
      "cbid": 251, "correlation": 43723
    }
  },
  {
    "ph": "f", "id": 43723, "pid": 76337, "tid": -914061504, "ts": 1716454218071446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071448, "dur": 0,
    "args": {
      "External id": 43724,
      "cbid": 251, "correlation": 43724
    }
  },
  {
    "ph": "f", "id": 43724, "pid": 76337, "tid": -914061504, "ts": 1716454218071448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071449, "dur": 0,
    "args": {
      "External id": 43725,
      "cbid": 251, "correlation": 43725
    }
  },
  {
    "ph": "f", "id": 43725, "pid": 76337, "tid": -914061504, "ts": 1716454218071449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071450, "dur": 0,
    "args": {
      "External id": 43726,
      "cbid": 251, "correlation": 43726
    }
  },
  {
    "ph": "f", "id": 43726, "pid": 76337, "tid": -914061504, "ts": 1716454218071450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071451, "dur": 0,
    "args": {
      "External id": 43727,
      "cbid": 251, "correlation": 43727
    }
  },
  {
    "ph": "f", "id": 43727, "pid": 76337, "tid": -914061504, "ts": 1716454218071451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071452, "dur": 0,
    "args": {
      "External id": 43728,
      "cbid": 251, "correlation": 43728
    }
  },
  {
    "ph": "f", "id": 43728, "pid": 76337, "tid": -914061504, "ts": 1716454218071452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071453, "dur": 0,
    "args": {
      "External id": 43729,
      "cbid": 251, "correlation": 43729
    }
  },
  {
    "ph": "f", "id": 43729, "pid": 76337, "tid": -914061504, "ts": 1716454218071453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071454, "dur": 0,
    "args": {
      "External id": 43730,
      "cbid": 251, "correlation": 43730
    }
  },
  {
    "ph": "f", "id": 43730, "pid": 76337, "tid": -914061504, "ts": 1716454218071454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071455, "dur": 0,
    "args": {
      "External id": 43731,
      "cbid": 251, "correlation": 43731
    }
  },
  {
    "ph": "f", "id": 43731, "pid": 76337, "tid": -914061504, "ts": 1716454218071455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218170765, "dur": 113,
    "args": {
      "External id": 43732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43732,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 43732, "pid": 5, "tid": 7, "ts": 1716454218170765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071457, "dur": 13,
    "args": {
      "External id": 43732,
      "cbid": 211, "correlation": 43732
    }
  },
  {
    "ph": "s", "id": 43732, "pid": 76337, "tid": -914061504, "ts": 1716454218071457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218170879, "dur": 59,
    "args": {
      "External id": 43738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43738, "pid": 5, "tid": 7, "ts": 1716454218170879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071495, "dur": 9,
    "args": {
      "External id": 43738,
      "cbid": 211, "correlation": 43738
    }
  },
  {
    "ph": "s", "id": 43738, "pid": 76337, "tid": -914061504, "ts": 1716454218071495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218170940, "dur": 50,
    "args": {
      "External id": 43746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43746,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43746, "pid": 5, "tid": 7, "ts": 1716454218170940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071527, "dur": 8,
    "args": {
      "External id": 43746,
      "cbid": 211, "correlation": 43746
    }
  },
  {
    "ph": "s", "id": 43746, "pid": 76337, "tid": -914061504, "ts": 1716454218071527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218170991, "dur": 98,
    "args": {
      "External id": 43755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43755,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43755, "pid": 5, "tid": 7, "ts": 1716454218170991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071566, "dur": 10,
    "args": {
      "External id": 43755,
      "cbid": 211, "correlation": 43755
    }
  },
  {
    "ph": "s", "id": 43755, "pid": 76337, "tid": -914061504, "ts": 1716454218071566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218171090, "dur": 92,
    "args": {
      "External id": 43775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43775,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 43775, "pid": 5, "tid": 7, "ts": 1716454218171090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071638, "dur": 13,
    "args": {
      "External id": 43775,
      "cbid": 211, "correlation": 43775
    }
  },
  {
    "ph": "s", "id": 43775, "pid": 76337, "tid": -914061504, "ts": 1716454218071638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218171183, "dur": 4,
    "args": {
      "External id": 43787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43787,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 43787, "pid": 5, "tid": 7, "ts": 1716454218171183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071664, "dur": 7,
    "args": {
      "External id": 43787,
      "cbid": 211, "correlation": 43787
    }
  },
  {
    "ph": "s", "id": 43787, "pid": 76337, "tid": -914061504, "ts": 1716454218071664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218171189, "dur": 107,
    "args": {
      "External id": 43790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43790, "pid": 5, "tid": 7, "ts": 1716454218171189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071683, "dur": 7,
    "args": {
      "External id": 43790,
      "cbid": 211, "correlation": 43790
    }
  },
  {
    "ph": "s", "id": 43790, "pid": 76337, "tid": -914061504, "ts": 1716454218071683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218171297, "dur": 69,
    "args": {
      "External id": 43799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43799,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43799, "pid": 5, "tid": 7, "ts": 1716454218171297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071723, "dur": 9,
    "args": {
      "External id": 43799,
      "cbid": 211, "correlation": 43799
    }
  },
  {
    "ph": "s", "id": 43799, "pid": 76337, "tid": -914061504, "ts": 1716454218071723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071774, "dur": 0,
    "args": {
      "External id": 43809,
      "cbid": 317, "correlation": 43809
    }
  },
  {
    "ph": "f", "id": 43809, "pid": 76337, "tid": -914061504, "ts": 1716454218071774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071775, "dur": 0,
    "args": {
      "External id": 43810,
      "cbid": 203, "correlation": 43810
    }
  },
  {
    "ph": "f", "id": 43810, "pid": 76337, "tid": -914061504, "ts": 1716454218071775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071776, "dur": 0,
    "args": {
      "External id": 43811,
      "cbid": 205, "correlation": 43811
    }
  },
  {
    "ph": "f", "id": 43811, "pid": 76337, "tid": -914061504, "ts": 1716454218071776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218171368, "dur": 76,
    "args": {
      "External id": 43815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43815,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43815, "pid": 5, "tid": 7, "ts": 1716454218171368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071794, "dur": 12,
    "args": {
      "External id": 43815,
      "cbid": 211, "correlation": 43815
    }
  },
  {
    "ph": "s", "id": 43815, "pid": 76337, "tid": -914061504, "ts": 1716454218071794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218171445, "dur": 24,
    "args": {
      "External id": 43817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43817,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43817, "pid": 5, "tid": 7, "ts": 1716454218171445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071808, "dur": 6,
    "args": {
      "External id": 43817,
      "cbid": 211, "correlation": 43817
    }
  },
  {
    "ph": "s", "id": 43817, "pid": 76337, "tid": -914061504, "ts": 1716454218071808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218171470, "dur": 3,
    "args": {
      "External id": 43819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43819, "pid": 5, "tid": 7, "ts": 1716454218171470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071819, "dur": 6,
    "args": {
      "External id": 43819,
      "cbid": 211, "correlation": 43819
    }
  },
  {
    "ph": "s", "id": 43819, "pid": 76337, "tid": -914061504, "ts": 1716454218071819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071828, "dur": 0,
    "args": {
      "External id": 43820,
      "cbid": 51, "correlation": 43820
    }
  },
  {
    "ph": "s", "id": 43820, "pid": 76337, "tid": -914061504, "ts": 1716454218071828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218171475, "dur": 1352,
    "args": {
      "External id": 43821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43821,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43821, "pid": 5, "tid": 7, "ts": 1716454218171475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071828, "dur": 6,
    "args": {
      "External id": 43821,
      "cbid": 211, "correlation": 43821
    }
  },
  {
    "ph": "s", "id": 43821, "pid": 76337, "tid": -914061504, "ts": 1716454218071828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218172829, "dur": 59,
    "args": {
      "External id": 43826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43826,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43826, "pid": 5, "tid": 7, "ts": 1716454218172829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071856, "dur": 9,
    "args": {
      "External id": 43826,
      "cbid": 211, "correlation": 43826
    }
  },
  {
    "ph": "s", "id": 43826, "pid": 76337, "tid": -914061504, "ts": 1716454218071856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218172889, "dur": 3,
    "args": {
      "External id": 43834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43834,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43834, "pid": 5, "tid": 7, "ts": 1716454218172889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071900, "dur": 10,
    "args": {
      "External id": 43834,
      "cbid": 211, "correlation": 43834
    }
  },
  {
    "ph": "s", "id": 43834, "pid": 76337, "tid": -914061504, "ts": 1716454218071900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071971, "dur": 1,
    "args": {
      "External id": 43850,
      "cbid": 251, "correlation": 43850
    }
  },
  {
    "ph": "f", "id": 43850, "pid": 76337, "tid": -914061504, "ts": 1716454218071971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071984, "dur": 0,
    "args": {
      "External id": 43852,
      "cbid": 251, "correlation": 43852
    }
  },
  {
    "ph": "f", "id": 43852, "pid": 76337, "tid": -914061504, "ts": 1716454218071984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218172894, "dur": 11,
    "args": {
      "External id": 43853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43853,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 43853, "pid": 5, "tid": 7, "ts": 1716454218172894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218071986, "dur": 12,
    "args": {
      "External id": 43853,
      "cbid": 211, "correlation": 43853
    }
  },
  {
    "ph": "s", "id": 43853, "pid": 76337, "tid": -914061504, "ts": 1716454218071986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218172906, "dur": 5,
    "args": {
      "External id": 43855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43855,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 43855, "pid": 5, "tid": 7, "ts": 1716454218172906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072000, "dur": 5,
    "args": {
      "External id": 43855,
      "cbid": 211, "correlation": 43855
    }
  },
  {
    "ph": "s", "id": 43855, "pid": 76337, "tid": -914061504, "ts": 1716454218072000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218172913, "dur": 54,
    "args": {
      "External id": 43865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43865,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43865, "pid": 5, "tid": 7, "ts": 1716454218172913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072059, "dur": 13,
    "args": {
      "External id": 43865,
      "cbid": 211, "correlation": 43865
    }
  },
  {
    "ph": "s", "id": 43865, "pid": 76337, "tid": -914061504, "ts": 1716454218072059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218172968, "dur": 51,
    "args": {
      "External id": 43885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43885,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 43885, "pid": 5, "tid": 7, "ts": 1716454218172968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072127, "dur": 11,
    "args": {
      "External id": 43885,
      "cbid": 211, "correlation": 43885
    }
  },
  {
    "ph": "s", "id": 43885, "pid": 76337, "tid": -914061504, "ts": 1716454218072127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218173021, "dur": 4,
    "args": {
      "External id": 43897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43897,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43897, "pid": 5, "tid": 7, "ts": 1716454218173021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072148, "dur": 6,
    "args": {
      "External id": 43897,
      "cbid": 211, "correlation": 43897
    }
  },
  {
    "ph": "s", "id": 43897, "pid": 76337, "tid": -914061504, "ts": 1716454218072148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218173026, "dur": 53,
    "args": {
      "External id": 43900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43900, "pid": 5, "tid": 7, "ts": 1716454218173026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072166, "dur": 7,
    "args": {
      "External id": 43900,
      "cbid": 211, "correlation": 43900
    }
  },
  {
    "ph": "s", "id": 43900, "pid": 76337, "tid": -914061504, "ts": 1716454218072166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218173080, "dur": 36,
    "args": {
      "External id": 43909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43909,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43909, "pid": 5, "tid": 7, "ts": 1716454218173080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072208, "dur": 10,
    "args": {
      "External id": 43909,
      "cbid": 211, "correlation": 43909
    }
  },
  {
    "ph": "s", "id": 43909, "pid": 76337, "tid": -914061504, "ts": 1716454218072208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072274, "dur": 0,
    "args": {
      "External id": 43919,
      "cbid": 317, "correlation": 43919
    }
  },
  {
    "ph": "f", "id": 43919, "pid": 76337, "tid": -914061504, "ts": 1716454218072274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072275, "dur": 0,
    "args": {
      "External id": 43920,
      "cbid": 203, "correlation": 43920
    }
  },
  {
    "ph": "f", "id": 43920, "pid": 76337, "tid": -914061504, "ts": 1716454218072275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072276, "dur": 0,
    "args": {
      "External id": 43921,
      "cbid": 205, "correlation": 43921
    }
  },
  {
    "ph": "f", "id": 43921, "pid": 76337, "tid": -914061504, "ts": 1716454218072276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218173118, "dur": 41,
    "args": {
      "External id": 43925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43925,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43925, "pid": 5, "tid": 7, "ts": 1716454218173118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072290, "dur": 12,
    "args": {
      "External id": 43925,
      "cbid": 211, "correlation": 43925
    }
  },
  {
    "ph": "s", "id": 43925, "pid": 76337, "tid": -914061504, "ts": 1716454218072290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218173160, "dur": 14,
    "args": {
      "External id": 43927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43927,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43927, "pid": 5, "tid": 7, "ts": 1716454218173160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072305, "dur": 5,
    "args": {
      "External id": 43927,
      "cbid": 211, "correlation": 43927
    }
  },
  {
    "ph": "s", "id": 43927, "pid": 76337, "tid": -914061504, "ts": 1716454218072305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218173176, "dur": 3,
    "args": {
      "External id": 43929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43929,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 43929, "pid": 5, "tid": 7, "ts": 1716454218173176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072314, "dur": 5,
    "args": {
      "External id": 43929,
      "cbid": 211, "correlation": 43929
    }
  },
  {
    "ph": "s", "id": 43929, "pid": 76337, "tid": -914061504, "ts": 1716454218072314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072322, "dur": 0,
    "args": {
      "External id": 43930,
      "cbid": 51, "correlation": 43930
    }
  },
  {
    "ph": "s", "id": 43930, "pid": 76337, "tid": -914061504, "ts": 1716454218072322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218173180, "dur": 689,
    "args": {
      "External id": 43931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43931,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43931, "pid": 5, "tid": 7, "ts": 1716454218173180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072323, "dur": 6,
    "args": {
      "External id": 43931,
      "cbid": 211, "correlation": 43931
    }
  },
  {
    "ph": "s", "id": 43931, "pid": 76337, "tid": -914061504, "ts": 1716454218072323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218173870, "dur": 59,
    "args": {
      "External id": 43936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43936, "pid": 5, "tid": 7, "ts": 1716454218173870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072351, "dur": 8,
    "args": {
      "External id": 43936,
      "cbid": 211, "correlation": 43936
    }
  },
  {
    "ph": "s", "id": 43936, "pid": 76337, "tid": -914061504, "ts": 1716454218072351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072408, "dur": 0,
    "args": {
      "External id": 43946,
      "cbid": 317, "correlation": 43946
    }
  },
  {
    "ph": "f", "id": 43946, "pid": 76337, "tid": -914061504, "ts": 1716454218072408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072409, "dur": 0,
    "args": {
      "External id": 43947,
      "cbid": 203, "correlation": 43947
    }
  },
  {
    "ph": "f", "id": 43947, "pid": 76337, "tid": -914061504, "ts": 1716454218072409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072410, "dur": 0,
    "args": {
      "External id": 43948,
      "cbid": 205, "correlation": 43948
    }
  },
  {
    "ph": "f", "id": 43948, "pid": 76337, "tid": -914061504, "ts": 1716454218072410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218173931, "dur": 75,
    "args": {
      "External id": 43952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43952, "pid": 5, "tid": 7, "ts": 1716454218173931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072423, "dur": 13,
    "args": {
      "External id": 43952,
      "cbid": 211, "correlation": 43952
    }
  },
  {
    "ph": "s", "id": 43952, "pid": 76337, "tid": -914061504, "ts": 1716454218072423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218174007, "dur": 207,
    "args": {
      "External id": 43954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43954,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 43954, "pid": 5, "tid": 7, "ts": 1716454218174007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072443, "dur": 8,
    "args": {
      "External id": 43954,
      "cbid": 211, "correlation": 43954
    }
  },
  {
    "ph": "s", "id": 43954, "pid": 76337, "tid": -914061504, "ts": 1716454218072443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218174216, "dur": 39,
    "args": {
      "External id": 43956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43956,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43956, "pid": 5, "tid": 7, "ts": 1716454218174216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072456, "dur": 6,
    "args": {
      "External id": 43956,
      "cbid": 211, "correlation": 43956
    }
  },
  {
    "ph": "s", "id": 43956, "pid": 76337, "tid": -914061504, "ts": 1716454218072456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218174256, "dur": 58,
    "args": {
      "External id": 43962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43962, "pid": 5, "tid": 7, "ts": 1716454218174256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072483, "dur": 8,
    "args": {
      "External id": 43962,
      "cbid": 211, "correlation": 43962
    }
  },
  {
    "ph": "s", "id": 43962, "pid": 76337, "tid": -914061504, "ts": 1716454218072483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218174316, "dur": 50,
    "args": {
      "External id": 43970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43970,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43970, "pid": 5, "tid": 7, "ts": 1716454218174316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072512, "dur": 8,
    "args": {
      "External id": 43970,
      "cbid": 211, "correlation": 43970
    }
  },
  {
    "ph": "s", "id": 43970, "pid": 76337, "tid": -914061504, "ts": 1716454218072512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218174367, "dur": 35,
    "args": {
      "External id": 43978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43978,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 43978, "pid": 5, "tid": 7, "ts": 1716454218174367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072544, "dur": 8,
    "args": {
      "External id": 43978,
      "cbid": 211, "correlation": 43978
    }
  },
  {
    "ph": "s", "id": 43978, "pid": 76337, "tid": -914061504, "ts": 1716454218072544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218174403, "dur": 51,
    "args": {
      "External id": 43998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 43998,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 43998, "pid": 5, "tid": 7, "ts": 1716454218174403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072627, "dur": 13,
    "args": {
      "External id": 43998,
      "cbid": 211, "correlation": 43998
    }
  },
  {
    "ph": "s", "id": 43998, "pid": 76337, "tid": -914061504, "ts": 1716454218072627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218174456, "dur": 4,
    "args": {
      "External id": 44010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44010,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 44010, "pid": 5, "tid": 7, "ts": 1716454218174456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072650, "dur": 6,
    "args": {
      "External id": 44010,
      "cbid": 211, "correlation": 44010
    }
  },
  {
    "ph": "s", "id": 44010, "pid": 76337, "tid": -914061504, "ts": 1716454218072650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218174461, "dur": 54,
    "args": {
      "External id": 44013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44013, "pid": 5, "tid": 7, "ts": 1716454218174461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072667, "dur": 7,
    "args": {
      "External id": 44013,
      "cbid": 211, "correlation": 44013
    }
  },
  {
    "ph": "s", "id": 44013, "pid": 76337, "tid": -914061504, "ts": 1716454218072667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072725, "dur": 0,
    "args": {
      "External id": 44024,
      "cbid": 317, "correlation": 44024
    }
  },
  {
    "ph": "f", "id": 44024, "pid": 76337, "tid": -914061504, "ts": 1716454218072725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072726, "dur": 0,
    "args": {
      "External id": 44025,
      "cbid": 203, "correlation": 44025
    }
  },
  {
    "ph": "f", "id": 44025, "pid": 76337, "tid": -914061504, "ts": 1716454218072726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072727, "dur": 0,
    "args": {
      "External id": 44026,
      "cbid": 205, "correlation": 44026
    }
  },
  {
    "ph": "f", "id": 44026, "pid": 76337, "tid": -914061504, "ts": 1716454218072727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072749, "dur": 1,
    "args": {
      "External id": 44030,
      "cbid": 251, "correlation": 44030
    }
  },
  {
    "ph": "f", "id": 44030, "pid": 76337, "tid": -914061504, "ts": 1716454218072749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072751, "dur": 0,
    "args": {
      "External id": 44031,
      "cbid": 251, "correlation": 44031
    }
  },
  {
    "ph": "f", "id": 44031, "pid": 76337, "tid": -914061504, "ts": 1716454218072751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072752, "dur": 0,
    "args": {
      "External id": 44032,
      "cbid": 251, "correlation": 44032
    }
  },
  {
    "ph": "f", "id": 44032, "pid": 76337, "tid": -914061504, "ts": 1716454218072752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072753, "dur": 0,
    "args": {
      "External id": 44033,
      "cbid": 251, "correlation": 44033
    }
  },
  {
    "ph": "f", "id": 44033, "pid": 76337, "tid": -914061504, "ts": 1716454218072753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072753, "dur": 0,
    "args": {
      "External id": 44034,
      "cbid": 251, "correlation": 44034
    }
  },
  {
    "ph": "f", "id": 44034, "pid": 76337, "tid": -914061504, "ts": 1716454218072753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072754, "dur": 0,
    "args": {
      "External id": 44035,
      "cbid": 251, "correlation": 44035
    }
  },
  {
    "ph": "f", "id": 44035, "pid": 76337, "tid": -914061504, "ts": 1716454218072754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072755, "dur": 0,
    "args": {
      "External id": 44036,
      "cbid": 251, "correlation": 44036
    }
  },
  {
    "ph": "f", "id": 44036, "pid": 76337, "tid": -914061504, "ts": 1716454218072755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072756, "dur": 0,
    "args": {
      "External id": 44037,
      "cbid": 251, "correlation": 44037
    }
  },
  {
    "ph": "f", "id": 44037, "pid": 76337, "tid": -914061504, "ts": 1716454218072756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072757, "dur": 0,
    "args": {
      "External id": 44038,
      "cbid": 251, "correlation": 44038
    }
  },
  {
    "ph": "f", "id": 44038, "pid": 76337, "tid": -914061504, "ts": 1716454218072757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218174517, "dur": 111,
    "args": {
      "External id": 44039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44039,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44039, "pid": 5, "tid": 7, "ts": 1716454218174517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072759, "dur": 12,
    "args": {
      "External id": 44039,
      "cbid": 211, "correlation": 44039
    }
  },
  {
    "ph": "s", "id": 44039, "pid": 76337, "tid": -914061504, "ts": 1716454218072759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218174630, "dur": 59,
    "args": {
      "External id": 44045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44045, "pid": 5, "tid": 7, "ts": 1716454218174630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072795, "dur": 8,
    "args": {
      "External id": 44045,
      "cbid": 211, "correlation": 44045
    }
  },
  {
    "ph": "s", "id": 44045, "pid": 76337, "tid": -914061504, "ts": 1716454218072795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218174690, "dur": 572,
    "args": {
      "External id": 44054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44054,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44054, "pid": 5, "tid": 7, "ts": 1716454218174690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072878, "dur": 15,
    "args": {
      "External id": 44054,
      "cbid": 211, "correlation": 44054
    }
  },
  {
    "ph": "s", "id": 44054, "pid": 76337, "tid": -914061504, "ts": 1716454218072878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218175263, "dur": 179,
    "args": {
      "External id": 44076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44076,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44076, "pid": 5, "tid": 7, "ts": 1716454218175263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218072936, "dur": 11,
    "args": {
      "External id": 44076,
      "cbid": 211, "correlation": 44076
    }
  },
  {
    "ph": "s", "id": 44076, "pid": 76337, "tid": -914061504, "ts": 1716454218072936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073033, "dur": 1,
    "args": {
      "External id": 44087,
      "cbid": 251, "correlation": 44087
    }
  },
  {
    "ph": "f", "id": 44087, "pid": 76337, "tid": -914061504, "ts": 1716454218073033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218175443, "dur": 194,
    "args": {
      "External id": 44088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44088,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44088, "pid": 5, "tid": 7, "ts": 1716454218175443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073038, "dur": 14,
    "args": {
      "External id": 44088,
      "cbid": 211, "correlation": 44088
    }
  },
  {
    "ph": "s", "id": 44088, "pid": 76337, "tid": -914061504, "ts": 1716454218073038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073107, "dur": 1,
    "args": {
      "External id": 44099,
      "cbid": 251, "correlation": 44099
    }
  },
  {
    "ph": "f", "id": 44099, "pid": 76337, "tid": -914061504, "ts": 1716454218073107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218175639, "dur": 186,
    "args": {
      "External id": 44100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44100,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44100, "pid": 5, "tid": 7, "ts": 1716454218175639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073110, "dur": 11,
    "args": {
      "External id": 44100,
      "cbid": 211, "correlation": 44100
    }
  },
  {
    "ph": "s", "id": 44100, "pid": 76337, "tid": -914061504, "ts": 1716454218073110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073173, "dur": 1,
    "args": {
      "External id": 44111,
      "cbid": 251, "correlation": 44111
    }
  },
  {
    "ph": "f", "id": 44111, "pid": 76337, "tid": -914061504, "ts": 1716454218073173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218175826, "dur": 185,
    "args": {
      "External id": 44112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44112,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44112, "pid": 5, "tid": 7, "ts": 1716454218175826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073178, "dur": 11,
    "args": {
      "External id": 44112,
      "cbid": 211, "correlation": 44112
    }
  },
  {
    "ph": "s", "id": 44112, "pid": 76337, "tid": -914061504, "ts": 1716454218073178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218176012, "dur": 18665,
    "args": {
      "External id": 44133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44133,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44133, "pid": 5, "tid": 7, "ts": 1716454218176012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073261, "dur": 13,
    "args": {
      "External id": 44133,
      "cbid": 211, "correlation": 44133
    }
  },
  {
    "ph": "s", "id": 44133, "pid": 76337, "tid": -914061504, "ts": 1716454218073261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073359, "dur": 1,
    "args": {
      "External id": 44151,
      "cbid": 251, "correlation": 44151
    }
  },
  {
    "ph": "f", "id": 44151, "pid": 76337, "tid": -914061504, "ts": 1716454218073359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218194678, "dur": 203,
    "args": {
      "External id": 44153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44153,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44153, "pid": 5, "tid": 7, "ts": 1716454218194678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073365, "dur": 13,
    "args": {
      "External id": 44153,
      "cbid": 211, "correlation": 44153
    }
  },
  {
    "ph": "s", "id": 44153, "pid": 76337, "tid": -914061504, "ts": 1716454218073365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218194882, "dur": 67,
    "args": {
      "External id": 44161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44161,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44161, "pid": 5, "tid": 7, "ts": 1716454218194882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073435, "dur": 12,
    "args": {
      "External id": 44161,
      "cbid": 211, "correlation": 44161
    }
  },
  {
    "ph": "s", "id": 44161, "pid": 76337, "tid": -914061504, "ts": 1716454218073435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218194950, "dur": 96,
    "args": {
      "External id": 44169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44169,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44169, "pid": 5, "tid": 7, "ts": 1716454218194950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073473, "dur": 8,
    "args": {
      "External id": 44169,
      "cbid": 211, "correlation": 44169
    }
  },
  {
    "ph": "s", "id": 44169, "pid": 76337, "tid": -914061504, "ts": 1716454218073473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218195048, "dur": 55,
    "args": {
      "External id": 44180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44180, "pid": 5, "tid": 7, "ts": 1716454218195048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073544, "dur": 12,
    "args": {
      "External id": 44180,
      "cbid": 211, "correlation": 44180
    }
  },
  {
    "ph": "s", "id": 44180, "pid": 76337, "tid": -914061504, "ts": 1716454218073544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218195104, "dur": 93,
    "args": {
      "External id": 44202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44202,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44202, "pid": 5, "tid": 7, "ts": 1716454218195104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073575, "dur": 8,
    "args": {
      "External id": 44202,
      "cbid": 211, "correlation": 44202
    }
  },
  {
    "ph": "s", "id": 44202, "pid": 76337, "tid": -914061504, "ts": 1716454218073575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073662, "dur": 1,
    "args": {
      "External id": 44213,
      "cbid": 251, "correlation": 44213
    }
  },
  {
    "ph": "f", "id": 44213, "pid": 76337, "tid": -914061504, "ts": 1716454218073662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218195199, "dur": 106,
    "args": {
      "External id": 44214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44214,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44214, "pid": 5, "tid": 7, "ts": 1716454218195199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073668, "dur": 14,
    "args": {
      "External id": 44214,
      "cbid": 211, "correlation": 44214
    }
  },
  {
    "ph": "s", "id": 44214, "pid": 76337, "tid": -914061504, "ts": 1716454218073668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073739, "dur": 1,
    "args": {
      "External id": 44225,
      "cbid": 251, "correlation": 44225
    }
  },
  {
    "ph": "f", "id": 44225, "pid": 76337, "tid": -914061504, "ts": 1716454218073739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073743, "dur": 0,
    "args": {
      "External id": 44226,
      "cbid": 251, "correlation": 44226
    }
  },
  {
    "ph": "f", "id": 44226, "pid": 76337, "tid": -914061504, "ts": 1716454218073743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218195306, "dur": 11,
    "args": {
      "External id": 44227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44227,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 44227, "pid": 5, "tid": 7, "ts": 1716454218195306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073744, "dur": 12,
    "args": {
      "External id": 44227,
      "cbid": 211, "correlation": 44227
    }
  },
  {
    "ph": "s", "id": 44227, "pid": 76337, "tid": -914061504, "ts": 1716454218073744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218195318, "dur": 5,
    "args": {
      "External id": 44229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44229,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 44229, "pid": 5, "tid": 7, "ts": 1716454218195318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073759, "dur": 6,
    "args": {
      "External id": 44229,
      "cbid": 211, "correlation": 44229
    }
  },
  {
    "ph": "s", "id": 44229, "pid": 76337, "tid": -914061504, "ts": 1716454218073759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073819, "dur": 1,
    "args": {
      "External id": 44240,
      "cbid": 251, "correlation": 44240
    }
  },
  {
    "ph": "f", "id": 44240, "pid": 76337, "tid": -914061504, "ts": 1716454218073819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073822, "dur": 0,
    "args": {
      "External id": 44241,
      "cbid": 251, "correlation": 44241
    }
  },
  {
    "ph": "f", "id": 44241, "pid": 76337, "tid": -914061504, "ts": 1716454218073822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218195324, "dur": 6,
    "args": {
      "External id": 44242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44242,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 44242, "pid": 5, "tid": 7, "ts": 1716454218195324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073824, "dur": 12,
    "args": {
      "External id": 44242,
      "cbid": 211, "correlation": 44242
    }
  },
  {
    "ph": "s", "id": 44242, "pid": 76337, "tid": -914061504, "ts": 1716454218073824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218195332, "dur": 4,
    "args": {
      "External id": 44244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44244,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 44244, "pid": 5, "tid": 7, "ts": 1716454218195332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073838, "dur": 6,
    "args": {
      "External id": 44244,
      "cbid": 211, "correlation": 44244
    }
  },
  {
    "ph": "s", "id": 44244, "pid": 76337, "tid": -914061504, "ts": 1716454218073838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218195337, "dur": 157,
    "args": {
      "External id": 44265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44265,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44265, "pid": 5, "tid": 7, "ts": 1716454218195337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218073912, "dur": 12,
    "args": {
      "External id": 44265,
      "cbid": 211, "correlation": 44265
    }
  },
  {
    "ph": "s", "id": 44265, "pid": 76337, "tid": -914061504, "ts": 1716454218073912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074019, "dur": 1,
    "args": {
      "External id": 44283,
      "cbid": 251, "correlation": 44283
    }
  },
  {
    "ph": "f", "id": 44283, "pid": 76337, "tid": -914061504, "ts": 1716454218074019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218195496, "dur": 107,
    "args": {
      "External id": 44285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44285,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44285, "pid": 5, "tid": 7, "ts": 1716454218195496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074025, "dur": 14,
    "args": {
      "External id": 44285,
      "cbid": 211, "correlation": 44285
    }
  },
  {
    "ph": "s", "id": 44285, "pid": 76337, "tid": -914061504, "ts": 1716454218074025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218195604, "dur": 34,
    "args": {
      "External id": 44293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44293,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44293, "pid": 5, "tid": 7, "ts": 1716454218195604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074096, "dur": 12,
    "args": {
      "External id": 44293,
      "cbid": 211, "correlation": 44293
    }
  },
  {
    "ph": "s", "id": 44293, "pid": 76337, "tid": -914061504, "ts": 1716454218074096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218195640, "dur": 66,
    "args": {
      "External id": 44301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44301, "pid": 5, "tid": 7, "ts": 1716454218195640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074137, "dur": 9,
    "args": {
      "External id": 44301,
      "cbid": 211, "correlation": 44301
    }
  },
  {
    "ph": "s", "id": 44301, "pid": 76337, "tid": -914061504, "ts": 1716454218074137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218195708, "dur": 93,
    "args": {
      "External id": 44323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44323,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44323, "pid": 5, "tid": 7, "ts": 1716454218195708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074188, "dur": 10,
    "args": {
      "External id": 44323,
      "cbid": 211, "correlation": 44323
    }
  },
  {
    "ph": "s", "id": 44323, "pid": 76337, "tid": -914061504, "ts": 1716454218074188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074275, "dur": 1,
    "args": {
      "External id": 44339,
      "cbid": 251, "correlation": 44339
    }
  },
  {
    "ph": "f", "id": 44339, "pid": 76337, "tid": -914061504, "ts": 1716454218074275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218195802, "dur": 579,
    "args": {
      "External id": 44341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44341,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44341, "pid": 5, "tid": 7, "ts": 1716454218195802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074281, "dur": 12,
    "args": {
      "External id": 44341,
      "cbid": 211, "correlation": 44341
    }
  },
  {
    "ph": "s", "id": 44341, "pid": 76337, "tid": -914061504, "ts": 1716454218074281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218196382, "dur": 244,
    "args": {
      "External id": 44349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44349, "pid": 5, "tid": 7, "ts": 1716454218196382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074345, "dur": 13,
    "args": {
      "External id": 44349,
      "cbid": 211, "correlation": 44349
    }
  },
  {
    "ph": "s", "id": 44349, "pid": 76337, "tid": -914061504, "ts": 1716454218074345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218196628, "dur": 251,
    "args": {
      "External id": 44357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44357, "pid": 5, "tid": 7, "ts": 1716454218196628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074380, "dur": 8,
    "args": {
      "External id": 44357,
      "cbid": 211, "correlation": 44357
    }
  },
  {
    "ph": "s", "id": 44357, "pid": 76337, "tid": -914061504, "ts": 1716454218074380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074463, "dur": 1,
    "args": {
      "External id": 44373,
      "cbid": 251, "correlation": 44373
    }
  },
  {
    "ph": "f", "id": 44373, "pid": 76337, "tid": -914061504, "ts": 1716454218074463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074468, "dur": 0,
    "args": {
      "External id": 44375,
      "cbid": 251, "correlation": 44375
    }
  },
  {
    "ph": "f", "id": 44375, "pid": 76337, "tid": -914061504, "ts": 1716454218074468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218196880, "dur": 362,
    "args": {
      "External id": 44376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44376,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 44376, "pid": 5, "tid": 7, "ts": 1716454218196880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074471, "dur": 13,
    "args": {
      "External id": 44376,
      "cbid": 211, "correlation": 44376
    }
  },
  {
    "ph": "s", "id": 44376, "pid": 76337, "tid": -914061504, "ts": 1716454218074471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218197243, "dur": 50,
    "args": {
      "External id": 44384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44384,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44384, "pid": 5, "tid": 7, "ts": 1716454218197243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074513, "dur": 11,
    "args": {
      "External id": 44384,
      "cbid": 211, "correlation": 44384
    }
  },
  {
    "ph": "s", "id": 44384, "pid": 76337, "tid": -914061504, "ts": 1716454218074513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218197294, "dur": 160,
    "args": {
      "External id": 44395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44395, "pid": 5, "tid": 7, "ts": 1716454218197294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074581, "dur": 12,
    "args": {
      "External id": 44395,
      "cbid": 211, "correlation": 44395
    }
  },
  {
    "ph": "s", "id": 44395, "pid": 76337, "tid": -914061504, "ts": 1716454218074581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074646, "dur": 0,
    "args": {
      "External id": 44407,
      "cbid": 317, "correlation": 44407
    }
  },
  {
    "ph": "f", "id": 44407, "pid": 76337, "tid": -914061504, "ts": 1716454218074646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074646, "dur": 0,
    "args": {
      "External id": 44408,
      "cbid": 203, "correlation": 44408
    }
  },
  {
    "ph": "f", "id": 44408, "pid": 76337, "tid": -914061504, "ts": 1716454218074646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074647, "dur": 0,
    "args": {
      "External id": 44409,
      "cbid": 205, "correlation": 44409
    }
  },
  {
    "ph": "f", "id": 44409, "pid": 76337, "tid": -914061504, "ts": 1716454218074647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074670, "dur": 1,
    "args": {
      "External id": 44413,
      "cbid": 251, "correlation": 44413
    }
  },
  {
    "ph": "f", "id": 44413, "pid": 76337, "tid": -914061504, "ts": 1716454218074670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074672, "dur": 0,
    "args": {
      "External id": 44414,
      "cbid": 251, "correlation": 44414
    }
  },
  {
    "ph": "f", "id": 44414, "pid": 76337, "tid": -914061504, "ts": 1716454218074672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074673, "dur": 0,
    "args": {
      "External id": 44415,
      "cbid": 251, "correlation": 44415
    }
  },
  {
    "ph": "f", "id": 44415, "pid": 76337, "tid": -914061504, "ts": 1716454218074673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074674, "dur": 0,
    "args": {
      "External id": 44416,
      "cbid": 251, "correlation": 44416
    }
  },
  {
    "ph": "f", "id": 44416, "pid": 76337, "tid": -914061504, "ts": 1716454218074674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074675, "dur": 0,
    "args": {
      "External id": 44417,
      "cbid": 251, "correlation": 44417
    }
  },
  {
    "ph": "f", "id": 44417, "pid": 76337, "tid": -914061504, "ts": 1716454218074675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074676, "dur": 0,
    "args": {
      "External id": 44418,
      "cbid": 251, "correlation": 44418
    }
  },
  {
    "ph": "f", "id": 44418, "pid": 76337, "tid": -914061504, "ts": 1716454218074676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074677, "dur": 0,
    "args": {
      "External id": 44419,
      "cbid": 251, "correlation": 44419
    }
  },
  {
    "ph": "f", "id": 44419, "pid": 76337, "tid": -914061504, "ts": 1716454218074677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074677, "dur": 0,
    "args": {
      "External id": 44420,
      "cbid": 251, "correlation": 44420
    }
  },
  {
    "ph": "f", "id": 44420, "pid": 76337, "tid": -914061504, "ts": 1716454218074677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074679, "dur": 0,
    "args": {
      "External id": 44421,
      "cbid": 251, "correlation": 44421
    }
  },
  {
    "ph": "f", "id": 44421, "pid": 76337, "tid": -914061504, "ts": 1716454218074679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218197455, "dur": 114,
    "args": {
      "External id": 44422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44422,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44422, "pid": 5, "tid": 7, "ts": 1716454218197455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074681, "dur": 12,
    "args": {
      "External id": 44422,
      "cbid": 211, "correlation": 44422
    }
  },
  {
    "ph": "s", "id": 44422, "pid": 76337, "tid": -914061504, "ts": 1716454218074681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218197571, "dur": 60,
    "args": {
      "External id": 44428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44428, "pid": 5, "tid": 7, "ts": 1716454218197571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074715, "dur": 9,
    "args": {
      "External id": 44428,
      "cbid": 211, "correlation": 44428
    }
  },
  {
    "ph": "s", "id": 44428, "pid": 76337, "tid": -914061504, "ts": 1716454218074715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218197632, "dur": 50,
    "args": {
      "External id": 44436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44436,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44436, "pid": 5, "tid": 7, "ts": 1716454218197632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074750, "dur": 9,
    "args": {
      "External id": 44436,
      "cbid": 211, "correlation": 44436
    }
  },
  {
    "ph": "s", "id": 44436, "pid": 76337, "tid": -914061504, "ts": 1716454218074750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218197683, "dur": 98,
    "args": {
      "External id": 44445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44445,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44445, "pid": 5, "tid": 7, "ts": 1716454218197683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074790, "dur": 10,
    "args": {
      "External id": 44445,
      "cbid": 211, "correlation": 44445
    }
  },
  {
    "ph": "s", "id": 44445, "pid": 76337, "tid": -914061504, "ts": 1716454218074790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218197782, "dur": 92,
    "args": {
      "External id": 44465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44465,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 44465, "pid": 5, "tid": 7, "ts": 1716454218197782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074861, "dur": 11,
    "args": {
      "External id": 44465,
      "cbid": 211, "correlation": 44465
    }
  },
  {
    "ph": "s", "id": 44465, "pid": 76337, "tid": -914061504, "ts": 1716454218074861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218197876, "dur": 5,
    "args": {
      "External id": 44477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44477,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 44477, "pid": 5, "tid": 7, "ts": 1716454218197876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074883, "dur": 7,
    "args": {
      "External id": 44477,
      "cbid": 211, "correlation": 44477
    }
  },
  {
    "ph": "s", "id": 44477, "pid": 76337, "tid": -914061504, "ts": 1716454218074883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218197882, "dur": 108,
    "args": {
      "External id": 44480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44480, "pid": 5, "tid": 7, "ts": 1716454218197882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074901, "dur": 7,
    "args": {
      "External id": 44480,
      "cbid": 211, "correlation": 44480
    }
  },
  {
    "ph": "s", "id": 44480, "pid": 76337, "tid": -914061504, "ts": 1716454218074901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218197991, "dur": 69,
    "args": {
      "External id": 44489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44489,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44489, "pid": 5, "tid": 7, "ts": 1716454218197991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218074941, "dur": 9,
    "args": {
      "External id": 44489,
      "cbid": 211, "correlation": 44489
    }
  },
  {
    "ph": "s", "id": 44489, "pid": 76337, "tid": -914061504, "ts": 1716454218074941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075001, "dur": 0,
    "args": {
      "External id": 44499,
      "cbid": 317, "correlation": 44499
    }
  },
  {
    "ph": "f", "id": 44499, "pid": 76337, "tid": -914061504, "ts": 1716454218075001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075002, "dur": 0,
    "args": {
      "External id": 44500,
      "cbid": 203, "correlation": 44500
    }
  },
  {
    "ph": "f", "id": 44500, "pid": 76337, "tid": -914061504, "ts": 1716454218075002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075003, "dur": 0,
    "args": {
      "External id": 44501,
      "cbid": 205, "correlation": 44501
    }
  },
  {
    "ph": "f", "id": 44501, "pid": 76337, "tid": -914061504, "ts": 1716454218075003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218198061, "dur": 76,
    "args": {
      "External id": 44505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44505,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44505, "pid": 5, "tid": 7, "ts": 1716454218198061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075017, "dur": 15,
    "args": {
      "External id": 44505,
      "cbid": 211, "correlation": 44505
    }
  },
  {
    "ph": "s", "id": 44505, "pid": 76337, "tid": -914061504, "ts": 1716454218075017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218198138, "dur": 25,
    "args": {
      "External id": 44507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44507,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44507, "pid": 5, "tid": 7, "ts": 1716454218198138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075034, "dur": 6,
    "args": {
      "External id": 44507,
      "cbid": 211, "correlation": 44507
    }
  },
  {
    "ph": "s", "id": 44507, "pid": 76337, "tid": -914061504, "ts": 1716454218075034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218198164, "dur": 4,
    "args": {
      "External id": 44509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 44509, "pid": 5, "tid": 7, "ts": 1716454218198164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075045, "dur": 5,
    "args": {
      "External id": 44509,
      "cbid": 211, "correlation": 44509
    }
  },
  {
    "ph": "s", "id": 44509, "pid": 76337, "tid": -914061504, "ts": 1716454218075045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075053, "dur": 0,
    "args": {
      "External id": 44510,
      "cbid": 51, "correlation": 44510
    }
  },
  {
    "ph": "s", "id": 44510, "pid": 76337, "tid": -914061504, "ts": 1716454218075053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218198169, "dur": 1370,
    "args": {
      "External id": 44511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44511,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44511, "pid": 5, "tid": 7, "ts": 1716454218198169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075054, "dur": 5,
    "args": {
      "External id": 44511,
      "cbid": 211, "correlation": 44511
    }
  },
  {
    "ph": "s", "id": 44511, "pid": 76337, "tid": -914061504, "ts": 1716454218075054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218199541, "dur": 59,
    "args": {
      "External id": 44516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44516,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44516, "pid": 5, "tid": 7, "ts": 1716454218199541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075081, "dur": 8,
    "args": {
      "External id": 44516,
      "cbid": 211, "correlation": 44516
    }
  },
  {
    "ph": "s", "id": 44516, "pid": 76337, "tid": -914061504, "ts": 1716454218075081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218199602, "dur": 3,
    "args": {
      "External id": 44524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44524,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 44524, "pid": 5, "tid": 7, "ts": 1716454218199602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075125, "dur": 10,
    "args": {
      "External id": 44524,
      "cbid": 211, "correlation": 44524
    }
  },
  {
    "ph": "s", "id": 44524, "pid": 76337, "tid": -914061504, "ts": 1716454218075125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075192, "dur": 2,
    "args": {
      "External id": 44540,
      "cbid": 251, "correlation": 44540
    }
  },
  {
    "ph": "f", "id": 44540, "pid": 76337, "tid": -914061504, "ts": 1716454218075192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075198, "dur": 0,
    "args": {
      "External id": 44542,
      "cbid": 251, "correlation": 44542
    }
  },
  {
    "ph": "f", "id": 44542, "pid": 76337, "tid": -914061504, "ts": 1716454218075198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218199606, "dur": 11,
    "args": {
      "External id": 44543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44543,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 44543, "pid": 5, "tid": 7, "ts": 1716454218199606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075200, "dur": 11,
    "args": {
      "External id": 44543,
      "cbid": 211, "correlation": 44543
    }
  },
  {
    "ph": "s", "id": 44543, "pid": 76337, "tid": -914061504, "ts": 1716454218075200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218199619, "dur": 5,
    "args": {
      "External id": 44545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44545,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 44545, "pid": 5, "tid": 7, "ts": 1716454218199619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075213, "dur": 5,
    "args": {
      "External id": 44545,
      "cbid": 211, "correlation": 44545
    }
  },
  {
    "ph": "s", "id": 44545, "pid": 76337, "tid": -914061504, "ts": 1716454218075213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218199625, "dur": 55,
    "args": {
      "External id": 44555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44555, "pid": 5, "tid": 7, "ts": 1716454218199625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075271, "dur": 12,
    "args": {
      "External id": 44555,
      "cbid": 211, "correlation": 44555
    }
  },
  {
    "ph": "s", "id": 44555, "pid": 76337, "tid": -914061504, "ts": 1716454218075271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218199681, "dur": 53,
    "args": {
      "External id": 44575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44575,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 44575, "pid": 5, "tid": 7, "ts": 1716454218199681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075340, "dur": 12,
    "args": {
      "External id": 44575,
      "cbid": 211, "correlation": 44575
    }
  },
  {
    "ph": "s", "id": 44575, "pid": 76337, "tid": -914061504, "ts": 1716454218075340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218199736, "dur": 4,
    "args": {
      "External id": 44587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44587,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 44587, "pid": 5, "tid": 7, "ts": 1716454218199736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075361, "dur": 6,
    "args": {
      "External id": 44587,
      "cbid": 211, "correlation": 44587
    }
  },
  {
    "ph": "s", "id": 44587, "pid": 76337, "tid": -914061504, "ts": 1716454218075361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218199741, "dur": 55,
    "args": {
      "External id": 44590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44590, "pid": 5, "tid": 7, "ts": 1716454218199741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075380, "dur": 6,
    "args": {
      "External id": 44590,
      "cbid": 211, "correlation": 44590
    }
  },
  {
    "ph": "s", "id": 44590, "pid": 76337, "tid": -914061504, "ts": 1716454218075380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218199797, "dur": 37,
    "args": {
      "External id": 44599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44599,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44599, "pid": 5, "tid": 7, "ts": 1716454218199797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075422, "dur": 9,
    "args": {
      "External id": 44599,
      "cbid": 211, "correlation": 44599
    }
  },
  {
    "ph": "s", "id": 44599, "pid": 76337, "tid": -914061504, "ts": 1716454218075422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075484, "dur": 0,
    "args": {
      "External id": 44609,
      "cbid": 317, "correlation": 44609
    }
  },
  {
    "ph": "f", "id": 44609, "pid": 76337, "tid": -914061504, "ts": 1716454218075484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075485, "dur": 0,
    "args": {
      "External id": 44610,
      "cbid": 203, "correlation": 44610
    }
  },
  {
    "ph": "f", "id": 44610, "pid": 76337, "tid": -914061504, "ts": 1716454218075485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075485, "dur": 0,
    "args": {
      "External id": 44611,
      "cbid": 205, "correlation": 44611
    }
  },
  {
    "ph": "f", "id": 44611, "pid": 76337, "tid": -914061504, "ts": 1716454218075485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218199835, "dur": 40,
    "args": {
      "External id": 44615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44615,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44615, "pid": 5, "tid": 7, "ts": 1716454218199835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075501, "dur": 13,
    "args": {
      "External id": 44615,
      "cbid": 211, "correlation": 44615
    }
  },
  {
    "ph": "s", "id": 44615, "pid": 76337, "tid": -914061504, "ts": 1716454218075501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218199877, "dur": 14,
    "args": {
      "External id": 44617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44617,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44617, "pid": 5, "tid": 7, "ts": 1716454218199877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075516, "dur": 5,
    "args": {
      "External id": 44617,
      "cbid": 211, "correlation": 44617
    }
  },
  {
    "ph": "s", "id": 44617, "pid": 76337, "tid": -914061504, "ts": 1716454218075516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218199892, "dur": 3,
    "args": {
      "External id": 44619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 44619, "pid": 5, "tid": 7, "ts": 1716454218199892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075525, "dur": 5,
    "args": {
      "External id": 44619,
      "cbid": 211, "correlation": 44619
    }
  },
  {
    "ph": "s", "id": 44619, "pid": 76337, "tid": -914061504, "ts": 1716454218075525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075533, "dur": 0,
    "args": {
      "External id": 44620,
      "cbid": 51, "correlation": 44620
    }
  },
  {
    "ph": "s", "id": 44620, "pid": 76337, "tid": -914061504, "ts": 1716454218075533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218199897, "dur": 701,
    "args": {
      "External id": 44621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44621,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44621, "pid": 5, "tid": 7, "ts": 1716454218199897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075534, "dur": 5,
    "args": {
      "External id": 44621,
      "cbid": 211, "correlation": 44621
    }
  },
  {
    "ph": "s", "id": 44621, "pid": 76337, "tid": -914061504, "ts": 1716454218075534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218200599, "dur": 60,
    "args": {
      "External id": 44626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44626, "pid": 5, "tid": 7, "ts": 1716454218200599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075561, "dur": 8,
    "args": {
      "External id": 44626,
      "cbid": 211, "correlation": 44626
    }
  },
  {
    "ph": "s", "id": 44626, "pid": 76337, "tid": -914061504, "ts": 1716454218075561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075621, "dur": 0,
    "args": {
      "External id": 44636,
      "cbid": 317, "correlation": 44636
    }
  },
  {
    "ph": "f", "id": 44636, "pid": 76337, "tid": -914061504, "ts": 1716454218075621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075622, "dur": 0,
    "args": {
      "External id": 44637,
      "cbid": 203, "correlation": 44637
    }
  },
  {
    "ph": "f", "id": 44637, "pid": 76337, "tid": -914061504, "ts": 1716454218075622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075623, "dur": 0,
    "args": {
      "External id": 44638,
      "cbid": 205, "correlation": 44638
    }
  },
  {
    "ph": "f", "id": 44638, "pid": 76337, "tid": -914061504, "ts": 1716454218075623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218200660, "dur": 75,
    "args": {
      "External id": 44642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44642,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44642, "pid": 5, "tid": 7, "ts": 1716454218200660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075635, "dur": 12,
    "args": {
      "External id": 44642,
      "cbid": 211, "correlation": 44642
    }
  },
  {
    "ph": "s", "id": 44642, "pid": 76337, "tid": -914061504, "ts": 1716454218075635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218200737, "dur": 209,
    "args": {
      "External id": 44644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44644,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44644, "pid": 5, "tid": 7, "ts": 1716454218200737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075653, "dur": 7,
    "args": {
      "External id": 44644,
      "cbid": 211, "correlation": 44644
    }
  },
  {
    "ph": "s", "id": 44644, "pid": 76337, "tid": -914061504, "ts": 1716454218075653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218200947, "dur": 39,
    "args": {
      "External id": 44646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44646,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44646, "pid": 5, "tid": 7, "ts": 1716454218200947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075664, "dur": 5,
    "args": {
      "External id": 44646,
      "cbid": 211, "correlation": 44646
    }
  },
  {
    "ph": "s", "id": 44646, "pid": 76337, "tid": -914061504, "ts": 1716454218075664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218200987, "dur": 59,
    "args": {
      "External id": 44652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44652,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44652, "pid": 5, "tid": 7, "ts": 1716454218200987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075691, "dur": 8,
    "args": {
      "External id": 44652,
      "cbid": 211, "correlation": 44652
    }
  },
  {
    "ph": "s", "id": 44652, "pid": 76337, "tid": -914061504, "ts": 1716454218075691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218201048, "dur": 50,
    "args": {
      "External id": 44660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44660,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44660, "pid": 5, "tid": 7, "ts": 1716454218201048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075719, "dur": 8,
    "args": {
      "External id": 44660,
      "cbid": 211, "correlation": 44660
    }
  },
  {
    "ph": "s", "id": 44660, "pid": 76337, "tid": -914061504, "ts": 1716454218075719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218201099, "dur": 35,
    "args": {
      "External id": 44668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44668,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44668, "pid": 5, "tid": 7, "ts": 1716454218201099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075747, "dur": 8,
    "args": {
      "External id": 44668,
      "cbid": 211, "correlation": 44668
    }
  },
  {
    "ph": "s", "id": 44668, "pid": 76337, "tid": -914061504, "ts": 1716454218075747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218201135, "dur": 53,
    "args": {
      "External id": 44688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44688,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 44688, "pid": 5, "tid": 7, "ts": 1716454218201135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075828, "dur": 12,
    "args": {
      "External id": 44688,
      "cbid": 211, "correlation": 44688
    }
  },
  {
    "ph": "s", "id": 44688, "pid": 76337, "tid": -914061504, "ts": 1716454218075828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218201189, "dur": 4,
    "args": {
      "External id": 44700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44700,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 44700, "pid": 5, "tid": 7, "ts": 1716454218201189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075850, "dur": 6,
    "args": {
      "External id": 44700,
      "cbid": 211, "correlation": 44700
    }
  },
  {
    "ph": "s", "id": 44700, "pid": 76337, "tid": -914061504, "ts": 1716454218075850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218201195, "dur": 56,
    "args": {
      "External id": 44703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44703, "pid": 5, "tid": 7, "ts": 1716454218201195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075870, "dur": 7,
    "args": {
      "External id": 44703,
      "cbid": 211, "correlation": 44703
    }
  },
  {
    "ph": "s", "id": 44703, "pid": 76337, "tid": -914061504, "ts": 1716454218075870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075928, "dur": 0,
    "args": {
      "External id": 44714,
      "cbid": 317, "correlation": 44714
    }
  },
  {
    "ph": "f", "id": 44714, "pid": 76337, "tid": -914061504, "ts": 1716454218075928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075929, "dur": 0,
    "args": {
      "External id": 44715,
      "cbid": 203, "correlation": 44715
    }
  },
  {
    "ph": "f", "id": 44715, "pid": 76337, "tid": -914061504, "ts": 1716454218075929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075930, "dur": 0,
    "args": {
      "External id": 44716,
      "cbid": 205, "correlation": 44716
    }
  },
  {
    "ph": "f", "id": 44716, "pid": 76337, "tid": -914061504, "ts": 1716454218075930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075953, "dur": 1,
    "args": {
      "External id": 44720,
      "cbid": 251, "correlation": 44720
    }
  },
  {
    "ph": "f", "id": 44720, "pid": 76337, "tid": -914061504, "ts": 1716454218075953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075955, "dur": 0,
    "args": {
      "External id": 44721,
      "cbid": 251, "correlation": 44721
    }
  },
  {
    "ph": "f", "id": 44721, "pid": 76337, "tid": -914061504, "ts": 1716454218075955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075955, "dur": 0,
    "args": {
      "External id": 44722,
      "cbid": 251, "correlation": 44722
    }
  },
  {
    "ph": "f", "id": 44722, "pid": 76337, "tid": -914061504, "ts": 1716454218075955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075956, "dur": 0,
    "args": {
      "External id": 44723,
      "cbid": 251, "correlation": 44723
    }
  },
  {
    "ph": "f", "id": 44723, "pid": 76337, "tid": -914061504, "ts": 1716454218075956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075957, "dur": 0,
    "args": {
      "External id": 44724,
      "cbid": 251, "correlation": 44724
    }
  },
  {
    "ph": "f", "id": 44724, "pid": 76337, "tid": -914061504, "ts": 1716454218075957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075958, "dur": 0,
    "args": {
      "External id": 44725,
      "cbid": 251, "correlation": 44725
    }
  },
  {
    "ph": "f", "id": 44725, "pid": 76337, "tid": -914061504, "ts": 1716454218075958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075958, "dur": 0,
    "args": {
      "External id": 44726,
      "cbid": 251, "correlation": 44726
    }
  },
  {
    "ph": "f", "id": 44726, "pid": 76337, "tid": -914061504, "ts": 1716454218075958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075959, "dur": 0,
    "args": {
      "External id": 44727,
      "cbid": 251, "correlation": 44727
    }
  },
  {
    "ph": "f", "id": 44727, "pid": 76337, "tid": -914061504, "ts": 1716454218075959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075960, "dur": 0,
    "args": {
      "External id": 44728,
      "cbid": 251, "correlation": 44728
    }
  },
  {
    "ph": "f", "id": 44728, "pid": 76337, "tid": -914061504, "ts": 1716454218075960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218201252, "dur": 115,
    "args": {
      "External id": 44729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44729,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44729, "pid": 5, "tid": 7, "ts": 1716454218201252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218075962, "dur": 20,
    "args": {
      "External id": 44729,
      "cbid": 211, "correlation": 44729
    }
  },
  {
    "ph": "s", "id": 44729, "pid": 76337, "tid": -914061504, "ts": 1716454218075962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218201369, "dur": 60,
    "args": {
      "External id": 44735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44735, "pid": 5, "tid": 7, "ts": 1716454218201369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076006, "dur": 9,
    "args": {
      "External id": 44735,
      "cbid": 211, "correlation": 44735
    }
  },
  {
    "ph": "s", "id": 44735, "pid": 76337, "tid": -914061504, "ts": 1716454218076006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218201430, "dur": 503,
    "args": {
      "External id": 44744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44744,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44744, "pid": 5, "tid": 7, "ts": 1716454218201430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076090, "dur": 14,
    "args": {
      "External id": 44744,
      "cbid": 211, "correlation": 44744
    }
  },
  {
    "ph": "s", "id": 44744, "pid": 76337, "tid": -914061504, "ts": 1716454218076090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218201933, "dur": 182,
    "args": {
      "External id": 44766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44766,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44766, "pid": 5, "tid": 7, "ts": 1716454218201933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076148, "dur": 10,
    "args": {
      "External id": 44766,
      "cbid": 211, "correlation": 44766
    }
  },
  {
    "ph": "s", "id": 44766, "pid": 76337, "tid": -914061504, "ts": 1716454218076148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076234, "dur": 1,
    "args": {
      "External id": 44777,
      "cbid": 251, "correlation": 44777
    }
  },
  {
    "ph": "f", "id": 44777, "pid": 76337, "tid": -914061504, "ts": 1716454218076234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218202117, "dur": 199,
    "args": {
      "External id": 44778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44778,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44778, "pid": 5, "tid": 7, "ts": 1716454218202117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076239, "dur": 16,
    "args": {
      "External id": 44778,
      "cbid": 211, "correlation": 44778
    }
  },
  {
    "ph": "s", "id": 44778, "pid": 76337, "tid": -914061504, "ts": 1716454218076239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076311, "dur": 1,
    "args": {
      "External id": 44789,
      "cbid": 251, "correlation": 44789
    }
  },
  {
    "ph": "f", "id": 44789, "pid": 76337, "tid": -914061504, "ts": 1716454218076311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218202318, "dur": 192,
    "args": {
      "External id": 44790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44790,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44790, "pid": 5, "tid": 7, "ts": 1716454218202318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076315, "dur": 11,
    "args": {
      "External id": 44790,
      "cbid": 211, "correlation": 44790
    }
  },
  {
    "ph": "s", "id": 44790, "pid": 76337, "tid": -914061504, "ts": 1716454218076315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076377, "dur": 1,
    "args": {
      "External id": 44801,
      "cbid": 251, "correlation": 44801
    }
  },
  {
    "ph": "f", "id": 44801, "pid": 76337, "tid": -914061504, "ts": 1716454218076377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218202511, "dur": 189,
    "args": {
      "External id": 44802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44802,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44802, "pid": 5, "tid": 7, "ts": 1716454218202511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076381, "dur": 11,
    "args": {
      "External id": 44802,
      "cbid": 211, "correlation": 44802
    }
  },
  {
    "ph": "s", "id": 44802, "pid": 76337, "tid": -914061504, "ts": 1716454218076381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218202702, "dur": 18750,
    "args": {
      "External id": 44823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44823,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44823, "pid": 5, "tid": 7, "ts": 1716454218202702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076461, "dur": 12,
    "args": {
      "External id": 44823,
      "cbid": 211, "correlation": 44823
    }
  },
  {
    "ph": "s", "id": 44823, "pid": 76337, "tid": -914061504, "ts": 1716454218076461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076557, "dur": 1,
    "args": {
      "External id": 44841,
      "cbid": 251, "correlation": 44841
    }
  },
  {
    "ph": "f", "id": 44841, "pid": 76337, "tid": -914061504, "ts": 1716454218076557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218221453, "dur": 206,
    "args": {
      "External id": 44843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44843,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44843, "pid": 5, "tid": 7, "ts": 1716454218221453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076563, "dur": 13,
    "args": {
      "External id": 44843,
      "cbid": 211, "correlation": 44843
    }
  },
  {
    "ph": "s", "id": 44843, "pid": 76337, "tid": -914061504, "ts": 1716454218076563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218221660, "dur": 66,
    "args": {
      "External id": 44851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44851,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44851, "pid": 5, "tid": 7, "ts": 1716454218221660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076633, "dur": 12,
    "args": {
      "External id": 44851,
      "cbid": 211, "correlation": 44851
    }
  },
  {
    "ph": "s", "id": 44851, "pid": 76337, "tid": -914061504, "ts": 1716454218076633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218221727, "dur": 97,
    "args": {
      "External id": 44859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44859,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44859, "pid": 5, "tid": 7, "ts": 1716454218221727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076675, "dur": 9,
    "args": {
      "External id": 44859,
      "cbid": 211, "correlation": 44859
    }
  },
  {
    "ph": "s", "id": 44859, "pid": 76337, "tid": -914061504, "ts": 1716454218076675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218221825, "dur": 54,
    "args": {
      "External id": 44870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44870, "pid": 5, "tid": 7, "ts": 1716454218221825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076745, "dur": 14,
    "args": {
      "External id": 44870,
      "cbid": 211, "correlation": 44870
    }
  },
  {
    "ph": "s", "id": 44870, "pid": 76337, "tid": -914061504, "ts": 1716454218076745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218221881, "dur": 93,
    "args": {
      "External id": 44892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44892,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44892, "pid": 5, "tid": 7, "ts": 1716454218221881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076776, "dur": 8,
    "args": {
      "External id": 44892,
      "cbid": 211, "correlation": 44892
    }
  },
  {
    "ph": "s", "id": 44892, "pid": 76337, "tid": -914061504, "ts": 1716454218076776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076860, "dur": 1,
    "args": {
      "External id": 44903,
      "cbid": 251, "correlation": 44903
    }
  },
  {
    "ph": "f", "id": 44903, "pid": 76337, "tid": -914061504, "ts": 1716454218076860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218221975, "dur": 107,
    "args": {
      "External id": 44904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44904,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 44904, "pid": 5, "tid": 7, "ts": 1716454218221975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076866, "dur": 12,
    "args": {
      "External id": 44904,
      "cbid": 211, "correlation": 44904
    }
  },
  {
    "ph": "s", "id": 44904, "pid": 76337, "tid": -914061504, "ts": 1716454218076866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076937, "dur": 1,
    "args": {
      "External id": 44915,
      "cbid": 251, "correlation": 44915
    }
  },
  {
    "ph": "f", "id": 44915, "pid": 76337, "tid": -914061504, "ts": 1716454218076937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076940, "dur": 0,
    "args": {
      "External id": 44916,
      "cbid": 251, "correlation": 44916
    }
  },
  {
    "ph": "f", "id": 44916, "pid": 76337, "tid": -914061504, "ts": 1716454218076940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218222084, "dur": 10,
    "args": {
      "External id": 44917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44917,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 44917, "pid": 5, "tid": 7, "ts": 1716454218222084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076942, "dur": 12,
    "args": {
      "External id": 44917,
      "cbid": 211, "correlation": 44917
    }
  },
  {
    "ph": "s", "id": 44917, "pid": 76337, "tid": -914061504, "ts": 1716454218076942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218222095, "dur": 5,
    "args": {
      "External id": 44919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44919,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 44919, "pid": 5, "tid": 7, "ts": 1716454218222095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218076956, "dur": 7,
    "args": {
      "External id": 44919,
      "cbid": 211, "correlation": 44919
    }
  },
  {
    "ph": "s", "id": 44919, "pid": 76337, "tid": -914061504, "ts": 1716454218076956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077033, "dur": 1,
    "args": {
      "External id": 44930,
      "cbid": 251, "correlation": 44930
    }
  },
  {
    "ph": "f", "id": 44930, "pid": 76337, "tid": -914061504, "ts": 1716454218077033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077037, "dur": 0,
    "args": {
      "External id": 44931,
      "cbid": 251, "correlation": 44931
    }
  },
  {
    "ph": "f", "id": 44931, "pid": 76337, "tid": -914061504, "ts": 1716454218077037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218222102, "dur": 6,
    "args": {
      "External id": 44932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44932,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 44932, "pid": 5, "tid": 7, "ts": 1716454218222102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077038, "dur": 15,
    "args": {
      "External id": 44932,
      "cbid": 211, "correlation": 44932
    }
  },
  {
    "ph": "s", "id": 44932, "pid": 76337, "tid": -914061504, "ts": 1716454218077038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218222110, "dur": 4,
    "args": {
      "External id": 44934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44934,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 44934, "pid": 5, "tid": 7, "ts": 1716454218222110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077055, "dur": 6,
    "args": {
      "External id": 44934,
      "cbid": 211, "correlation": 44934
    }
  },
  {
    "ph": "s", "id": 44934, "pid": 76337, "tid": -914061504, "ts": 1716454218077055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218222114, "dur": 158,
    "args": {
      "External id": 44955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44955,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44955, "pid": 5, "tid": 7, "ts": 1716454218222114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077131, "dur": 12,
    "args": {
      "External id": 44955,
      "cbid": 211, "correlation": 44955
    }
  },
  {
    "ph": "s", "id": 44955, "pid": 76337, "tid": -914061504, "ts": 1716454218077131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077227, "dur": 1,
    "args": {
      "External id": 44973,
      "cbid": 251, "correlation": 44973
    }
  },
  {
    "ph": "f", "id": 44973, "pid": 76337, "tid": -914061504, "ts": 1716454218077227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218222273, "dur": 106,
    "args": {
      "External id": 44975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44975,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 44975, "pid": 5, "tid": 7, "ts": 1716454218222273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077233, "dur": 13,
    "args": {
      "External id": 44975,
      "cbid": 211, "correlation": 44975
    }
  },
  {
    "ph": "s", "id": 44975, "pid": 76337, "tid": -914061504, "ts": 1716454218077233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218222381, "dur": 35,
    "args": {
      "External id": 44983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44983,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44983, "pid": 5, "tid": 7, "ts": 1716454218222381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077302, "dur": 12,
    "args": {
      "External id": 44983,
      "cbid": 211, "correlation": 44983
    }
  },
  {
    "ph": "s", "id": 44983, "pid": 76337, "tid": -914061504, "ts": 1716454218077302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218222417, "dur": 68,
    "args": {
      "External id": 44991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 44991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 44991, "pid": 5, "tid": 7, "ts": 1716454218222417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077343, "dur": 9,
    "args": {
      "External id": 44991,
      "cbid": 211, "correlation": 44991
    }
  },
  {
    "ph": "s", "id": 44991, "pid": 76337, "tid": -914061504, "ts": 1716454218077343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218222487, "dur": 93,
    "args": {
      "External id": 45013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45013,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45013, "pid": 5, "tid": 7, "ts": 1716454218222487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077398, "dur": 10,
    "args": {
      "External id": 45013,
      "cbid": 211, "correlation": 45013
    }
  },
  {
    "ph": "s", "id": 45013, "pid": 76337, "tid": -914061504, "ts": 1716454218077398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077485, "dur": 1,
    "args": {
      "External id": 45029,
      "cbid": 251, "correlation": 45029
    }
  },
  {
    "ph": "f", "id": 45029, "pid": 76337, "tid": -914061504, "ts": 1716454218077485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218222581, "dur": 577,
    "args": {
      "External id": 45031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45031,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 45031, "pid": 5, "tid": 7, "ts": 1716454218222581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077491, "dur": 13,
    "args": {
      "External id": 45031,
      "cbid": 211, "correlation": 45031
    }
  },
  {
    "ph": "s", "id": 45031, "pid": 76337, "tid": -914061504, "ts": 1716454218077491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218223160, "dur": 244,
    "args": {
      "External id": 45039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45039, "pid": 5, "tid": 7, "ts": 1716454218223160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077555, "dur": 13,
    "args": {
      "External id": 45039,
      "cbid": 211, "correlation": 45039
    }
  },
  {
    "ph": "s", "id": 45039, "pid": 76337, "tid": -914061504, "ts": 1716454218077555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218223405, "dur": 254,
    "args": {
      "External id": 45047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45047, "pid": 5, "tid": 7, "ts": 1716454218223405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077585, "dur": 9,
    "args": {
      "External id": 45047,
      "cbid": 211, "correlation": 45047
    }
  },
  {
    "ph": "s", "id": 45047, "pid": 76337, "tid": -914061504, "ts": 1716454218077585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077668, "dur": 1,
    "args": {
      "External id": 45063,
      "cbid": 251, "correlation": 45063
    }
  },
  {
    "ph": "f", "id": 45063, "pid": 76337, "tid": -914061504, "ts": 1716454218077668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077674, "dur": 0,
    "args": {
      "External id": 45065,
      "cbid": 251, "correlation": 45065
    }
  },
  {
    "ph": "f", "id": 45065, "pid": 76337, "tid": -914061504, "ts": 1716454218077674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218223660, "dur": 359,
    "args": {
      "External id": 45066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45066,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45066, "pid": 5, "tid": 7, "ts": 1716454218223660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077676, "dur": 12,
    "args": {
      "External id": 45066,
      "cbid": 211, "correlation": 45066
    }
  },
  {
    "ph": "s", "id": 45066, "pid": 76337, "tid": -914061504, "ts": 1716454218077676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218224020, "dur": 50,
    "args": {
      "External id": 45074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45074,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45074, "pid": 5, "tid": 7, "ts": 1716454218224020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077718, "dur": 10,
    "args": {
      "External id": 45074,
      "cbid": 211, "correlation": 45074
    }
  },
  {
    "ph": "s", "id": 45074, "pid": 76337, "tid": -914061504, "ts": 1716454218077718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218224072, "dur": 160,
    "args": {
      "External id": 45085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45085, "pid": 5, "tid": 7, "ts": 1716454218224072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077783, "dur": 36,
    "args": {
      "External id": 45085,
      "cbid": 211, "correlation": 45085
    }
  },
  {
    "ph": "s", "id": 45085, "pid": 76337, "tid": -914061504, "ts": 1716454218077783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077873, "dur": 0,
    "args": {
      "External id": 45097,
      "cbid": 317, "correlation": 45097
    }
  },
  {
    "ph": "f", "id": 45097, "pid": 76337, "tid": -914061504, "ts": 1716454218077873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077874, "dur": 0,
    "args": {
      "External id": 45098,
      "cbid": 203, "correlation": 45098
    }
  },
  {
    "ph": "f", "id": 45098, "pid": 76337, "tid": -914061504, "ts": 1716454218077874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077875, "dur": 0,
    "args": {
      "External id": 45099,
      "cbid": 205, "correlation": 45099
    }
  },
  {
    "ph": "f", "id": 45099, "pid": 76337, "tid": -914061504, "ts": 1716454218077875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077897, "dur": 1,
    "args": {
      "External id": 45103,
      "cbid": 251, "correlation": 45103
    }
  },
  {
    "ph": "f", "id": 45103, "pid": 76337, "tid": -914061504, "ts": 1716454218077897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077899, "dur": 0,
    "args": {
      "External id": 45104,
      "cbid": 251, "correlation": 45104
    }
  },
  {
    "ph": "f", "id": 45104, "pid": 76337, "tid": -914061504, "ts": 1716454218077899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077900, "dur": 0,
    "args": {
      "External id": 45105,
      "cbid": 251, "correlation": 45105
    }
  },
  {
    "ph": "f", "id": 45105, "pid": 76337, "tid": -914061504, "ts": 1716454218077900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077901, "dur": 0,
    "args": {
      "External id": 45106,
      "cbid": 251, "correlation": 45106
    }
  },
  {
    "ph": "f", "id": 45106, "pid": 76337, "tid": -914061504, "ts": 1716454218077901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077901, "dur": 0,
    "args": {
      "External id": 45107,
      "cbid": 251, "correlation": 45107
    }
  },
  {
    "ph": "f", "id": 45107, "pid": 76337, "tid": -914061504, "ts": 1716454218077901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077902, "dur": 0,
    "args": {
      "External id": 45108,
      "cbid": 251, "correlation": 45108
    }
  },
  {
    "ph": "f", "id": 45108, "pid": 76337, "tid": -914061504, "ts": 1716454218077902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077903, "dur": 0,
    "args": {
      "External id": 45109,
      "cbid": 251, "correlation": 45109
    }
  },
  {
    "ph": "f", "id": 45109, "pid": 76337, "tid": -914061504, "ts": 1716454218077903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077904, "dur": 0,
    "args": {
      "External id": 45110,
      "cbid": 251, "correlation": 45110
    }
  },
  {
    "ph": "f", "id": 45110, "pid": 76337, "tid": -914061504, "ts": 1716454218077904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077905, "dur": 0,
    "args": {
      "External id": 45111,
      "cbid": 251, "correlation": 45111
    }
  },
  {
    "ph": "f", "id": 45111, "pid": 76337, "tid": -914061504, "ts": 1716454218077905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218224233, "dur": 116,
    "args": {
      "External id": 45112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45112,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 45112, "pid": 5, "tid": 7, "ts": 1716454218224233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077907, "dur": 33,
    "args": {
      "External id": 45112,
      "cbid": 211, "correlation": 45112
    }
  },
  {
    "ph": "s", "id": 45112, "pid": 76337, "tid": -914061504, "ts": 1716454218077907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218224350, "dur": 61,
    "args": {
      "External id": 45118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45118, "pid": 5, "tid": 7, "ts": 1716454218224350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218077964, "dur": 279,
    "args": {
      "External id": 45118,
      "cbid": 211, "correlation": 45118
    }
  },
  {
    "ph": "s", "id": 45118, "pid": 76337, "tid": -914061504, "ts": 1716454218077964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218224412, "dur": 50,
    "args": {
      "External id": 45126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45126,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45126, "pid": 5, "tid": 7, "ts": 1716454218224412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078267, "dur": 9,
    "args": {
      "External id": 45126,
      "cbid": 211, "correlation": 45126
    }
  },
  {
    "ph": "s", "id": 45126, "pid": 76337, "tid": -914061504, "ts": 1716454218078267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218224464, "dur": 53,
    "args": {
      "External id": 45146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45146,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 45146, "pid": 5, "tid": 7, "ts": 1716454218224464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078349, "dur": 12,
    "args": {
      "External id": 45146,
      "cbid": 211, "correlation": 45146
    }
  },
  {
    "ph": "s", "id": 45146, "pid": 76337, "tid": -914061504, "ts": 1716454218078349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218224518, "dur": 5,
    "args": {
      "External id": 45158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45158,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 45158, "pid": 5, "tid": 7, "ts": 1716454218224518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078371, "dur": 6,
    "args": {
      "External id": 45158,
      "cbid": 211, "correlation": 45158
    }
  },
  {
    "ph": "s", "id": 45158, "pid": 76337, "tid": -914061504, "ts": 1716454218078371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218224524, "dur": 56,
    "args": {
      "External id": 45161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45161, "pid": 5, "tid": 7, "ts": 1716454218224524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078389, "dur": 111,
    "args": {
      "External id": 45161,
      "cbid": 211, "correlation": 45161
    }
  },
  {
    "ph": "s", "id": 45161, "pid": 76337, "tid": -914061504, "ts": 1716454218078389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218224581, "dur": 39,
    "args": {
      "External id": 45170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45170,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45170, "pid": 5, "tid": 7, "ts": 1716454218224581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078539, "dur": 11,
    "args": {
      "External id": 45170,
      "cbid": 211, "correlation": 45170
    }
  },
  {
    "ph": "s", "id": 45170, "pid": 76337, "tid": -914061504, "ts": 1716454218078539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078594, "dur": 0,
    "args": {
      "External id": 45180,
      "cbid": 317, "correlation": 45180
    }
  },
  {
    "ph": "f", "id": 45180, "pid": 76337, "tid": -914061504, "ts": 1716454218078594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078595, "dur": 0,
    "args": {
      "External id": 45181,
      "cbid": 203, "correlation": 45181
    }
  },
  {
    "ph": "f", "id": 45181, "pid": 76337, "tid": -914061504, "ts": 1716454218078595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078596, "dur": 0,
    "args": {
      "External id": 45182,
      "cbid": 205, "correlation": 45182
    }
  },
  {
    "ph": "f", "id": 45182, "pid": 76337, "tid": -914061504, "ts": 1716454218078596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218224621, "dur": 42,
    "args": {
      "External id": 45186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45186, "pid": 5, "tid": 7, "ts": 1716454218224621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078613, "dur": 12,
    "args": {
      "External id": 45186,
      "cbid": 211, "correlation": 45186
    }
  },
  {
    "ph": "s", "id": 45186, "pid": 76337, "tid": -914061504, "ts": 1716454218078613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218224664, "dur": 3,
    "args": {
      "External id": 45188,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45188,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.707317073170731
    }
  },
  {
    "ph": "f", "id": 45188, "pid": 5, "tid": 7, "ts": 1716454218224664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078628, "dur": 39,
    "args": {
      "External id": 45188,
      "cbid": 51, "correlation": 45188
    }
  },
  {
    "ph": "s", "id": 45188, "pid": 76337, "tid": -914061504, "ts": 1716454218078628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078674, "dur": 2,
    "args": {
      "External id": 45190,
      "cbid": 200, "correlation": 45190
    }
  },
  {
    "ph": "f", "id": 45190, "pid": 76337, "tid": -914061504, "ts": 1716454218078674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078676, "dur": 0,
    "args": {
      "External id": 45191,
      "cbid": 200, "correlation": 45191
    }
  },
  {
    "ph": "f", "id": 45191, "pid": 76337, "tid": -914061504, "ts": 1716454218078676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078677, "dur": 0,
    "args": {
      "External id": 45192,
      "cbid": 200, "correlation": 45192
    }
  },
  {
    "ph": "f", "id": 45192, "pid": 76337, "tid": -914061504, "ts": 1716454218078677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078678, "dur": 0,
    "args": {
      "External id": 45193,
      "cbid": 200, "correlation": 45193
    }
  },
  {
    "ph": "f", "id": 45193, "pid": 76337, "tid": -914061504, "ts": 1716454218078678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078678, "dur": 3,
    "args": {
      "External id": 45194,
      "cbid": 15, "correlation": 45194
    }
  },
  {
    "ph": "f", "id": 45194, "pid": 76337, "tid": -914061504, "ts": 1716454218078678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078682, "dur": 1,
    "args": {
      "External id": 45195,
      "cbid": 251, "correlation": 45195
    }
  },
  {
    "ph": "f", "id": 45195, "pid": 76337, "tid": -914061504, "ts": 1716454218078682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218224669, "dur": 23,
    "args": {
      "External id": 45196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45196,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45196, "pid": 5, "tid": 7, "ts": 1716454218224669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078686, "dur": 11,
    "args": {
      "External id": 45196,
      "cbid": 211, "correlation": 45196
    }
  },
  {
    "ph": "s", "id": 45196, "pid": 76337, "tid": -914061504, "ts": 1716454218078686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218224694, "dur": 4,
    "args": {
      "External id": 45198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 45198, "pid": 5, "tid": 7, "ts": 1716454218224694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078702, "dur": 6,
    "args": {
      "External id": 45198,
      "cbid": 211, "correlation": 45198
    }
  },
  {
    "ph": "s", "id": 45198, "pid": 76337, "tid": -914061504, "ts": 1716454218078702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078712, "dur": 0,
    "args": {
      "External id": 45199,
      "cbid": 51, "correlation": 45199
    }
  },
  {
    "ph": "s", "id": 45199, "pid": 76337, "tid": -914061504, "ts": 1716454218078712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218224699, "dur": 189,
    "args": {
      "External id": 45200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45200,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45200, "pid": 5, "tid": 7, "ts": 1716454218224699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078713, "dur": 171,
    "args": {
      "External id": 45200,
      "cbid": 211, "correlation": 45200
    }
  },
  {
    "ph": "s", "id": 45200, "pid": 76337, "tid": -914061504, "ts": 1716454218078713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218224889, "dur": 6,
    "args": {
      "External id": 45201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45201,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45201, "pid": 5, "tid": 7, "ts": 1716454218224889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078888, "dur": 6,
    "args": {
      "External id": 45201,
      "cbid": 211, "correlation": 45201
    }
  },
  {
    "ph": "s", "id": 45201, "pid": 76337, "tid": -914061504, "ts": 1716454218078888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218224897, "dur": 5,
    "args": {
      "External id": 45207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45207,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 45207, "pid": 5, "tid": 7, "ts": 1716454218224897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218078918, "dur": 9,
    "args": {
      "External id": 45207,
      "cbid": 211, "correlation": 45207
    }
  },
  {
    "ph": "s", "id": 45207, "pid": 76337, "tid": -914061504, "ts": 1716454218078918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218224903, "dur": 3,
    "args": {
      "External id": 45215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45215,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45215, "pid": 5, "tid": 7, "ts": 1716454218224903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218080663, "dur": 18,
    "args": {
      "External id": 45215,
      "cbid": 211, "correlation": 45215
    }
  },
  {
    "ph": "s", "id": 45215, "pid": 76337, "tid": -914061504, "ts": 1716454218080663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218224907, "dur": 3,
    "args": {
      "External id": 45223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45223,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45223, "pid": 5, "tid": 7, "ts": 1716454218224907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218080713, "dur": 11,
    "args": {
      "External id": 45223,
      "cbid": 211, "correlation": 45223
    }
  },
  {
    "ph": "s", "id": 45223, "pid": 76337, "tid": -914061504, "ts": 1716454218080713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218224911, "dur": 3,
    "args": {
      "External id": 45231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45231,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45231, "pid": 5, "tid": 7, "ts": 1716454218224911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218080746, "dur": 9,
    "args": {
      "External id": 45231,
      "cbid": 211, "correlation": 45231
    }
  },
  {
    "ph": "s", "id": 45231, "pid": 76337, "tid": -914061504, "ts": 1716454218080746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218224915, "dur": 3,
    "args": {
      "External id": 45240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45240,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45240, "pid": 5, "tid": 7, "ts": 1716454218224915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218080926, "dur": 15,
    "args": {
      "External id": 45240,
      "cbid": 211, "correlation": 45240
    }
  },
  {
    "ph": "s", "id": 45240, "pid": 76337, "tid": -914061504, "ts": 1716454218080926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218224919, "dur": 3,
    "args": {
      "External id": 45249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45249,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45249, "pid": 5, "tid": 7, "ts": 1716454218224919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218080955, "dur": 7,
    "args": {
      "External id": 45249,
      "cbid": 211, "correlation": 45249
    }
  },
  {
    "ph": "s", "id": 45249, "pid": 76337, "tid": -914061504, "ts": 1716454218080955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218224923, "dur": 3,
    "args": {
      "External id": 45257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45257,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45257, "pid": 5, "tid": 7, "ts": 1716454218224923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218080991, "dur": 9,
    "args": {
      "External id": 45257,
      "cbid": 211, "correlation": 45257
    }
  },
  {
    "ph": "s", "id": 45257, "pid": 76337, "tid": -914061504, "ts": 1716454218080991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218224929, "dur": 1,
    "args": {
      "External id": 45267,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45267,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 45267, "pid": 5, "tid": 7, "ts": 1716454218224929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218081241, "dur": 40,
    "args": {
      "External id": 45267,
      "cbid": 41, "correlation": 45267
    }
  },
  {
    "ph": "s", "id": 45267, "pid": 76337, "tid": -914061504, "ts": 1716454218081241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218081283, "dur": 143662,
    "args": {
      "External id": 45268,
      "cbid": 131, "correlation": 45268
    }
  },
  {
    "ph": "f", "id": 45268, "pid": 76337, "tid": -914061504, "ts": 1716454218081283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454218225099, "dur": 3,
    "args": {
      "External id": 45276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45276, "pid": 5, "tid": 7, "ts": 1716454218225099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225069, "dur": 30,
    "args": {
      "External id": 45276,
      "cbid": 211, "correlation": 45276
    }
  },
  {
    "ph": "s", "id": 45276, "pid": 76337, "tid": -914061504, "ts": 1716454218225069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218225218, "dur": 3,
    "args": {
      "External id": 45285,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45285,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0011904761904761906
    }
  },
  {
    "ph": "f", "id": 45285, "pid": 5, "tid": 7, "ts": 1716454218225218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225186, "dur": 32,
    "args": {
      "External id": 45285,
      "cbid": 41, "correlation": 45285
    }
  },
  {
    "ph": "s", "id": 45285, "pid": 76337, "tid": -914061504, "ts": 1716454218225186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218225287, "dur": 1,
    "args": {
      "External id": 45296,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45296,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 45296, "pid": 5, "tid": 7, "ts": 1716454218225287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225272, "dur": 13,
    "args": {
      "External id": 45296,
      "cbid": 41, "correlation": 45296
    }
  },
  {
    "ph": "s", "id": 45296, "pid": 76337, "tid": -914061504, "ts": 1716454218225272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225286, "dur": 8,
    "args": {
      "External id": 45297,
      "cbid": 131, "correlation": 45297
    }
  },
  {
    "ph": "f", "id": 45297, "pid": 76337, "tid": -914061504, "ts": 1716454218225286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218225359, "dur": 1,
    "args": {
      "External id": 45306,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45306,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0010245901639344263
    }
  },
  {
    "ph": "f", "id": 45306, "pid": 5, "tid": 7, "ts": 1716454218225359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225325, "dur": 32,
    "args": {
      "External id": 45306,
      "cbid": 41, "correlation": 45306
    }
  },
  {
    "ph": "s", "id": 45306, "pid": 76337, "tid": -914061504, "ts": 1716454218225325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225358, "dur": 8,
    "args": {
      "External id": 45307,
      "cbid": 131, "correlation": 45307
    }
  },
  {
    "ph": "f", "id": 45307, "pid": 76337, "tid": -914061504, "ts": 1716454218225358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218225413, "dur": 3,
    "args": {
      "External id": 45315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45315,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45315, "pid": 5, "tid": 7, "ts": 1716454218225413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225398, "dur": 14,
    "args": {
      "External id": 45315,
      "cbid": 211, "correlation": 45315
    }
  },
  {
    "ph": "s", "id": 45315, "pid": 76337, "tid": -914061504, "ts": 1716454218225398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218225443, "dur": 3,
    "args": {
      "External id": 45325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45325,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45325, "pid": 5, "tid": 7, "ts": 1716454218225443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225434, "dur": 8,
    "args": {
      "External id": 45325,
      "cbid": 211, "correlation": 45325
    }
  },
  {
    "ph": "s", "id": 45325, "pid": 76337, "tid": -914061504, "ts": 1716454218225434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218225467, "dur": 3,
    "args": {
      "External id": 45334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45334,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45334, "pid": 5, "tid": 7, "ts": 1716454218225467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225457, "dur": 8,
    "args": {
      "External id": 45334,
      "cbid": 211, "correlation": 45334
    }
  },
  {
    "ph": "s", "id": 45334, "pid": 76337, "tid": -914061504, "ts": 1716454218225457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218225488, "dur": 3,
    "args": {
      "External id": 45342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45342,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45342, "pid": 5, "tid": 7, "ts": 1716454218225488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225480, "dur": 7,
    "args": {
      "External id": 45342,
      "cbid": 211, "correlation": 45342
    }
  },
  {
    "ph": "s", "id": 45342, "pid": 76337, "tid": -914061504, "ts": 1716454218225480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218225531, "dur": 6,
    "args": {
      "External id": 45352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45352, "pid": 5, "tid": 7, "ts": 1716454218225531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225519, "dur": 10,
    "args": {
      "External id": 45352,
      "cbid": 211, "correlation": 45352
    }
  },
  {
    "ph": "s", "id": 45352, "pid": 76337, "tid": -914061504, "ts": 1716454218225519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218225565, "dur": 3,
    "args": {
      "External id": 45360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45360,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45360, "pid": 5, "tid": 7, "ts": 1716454218225565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225555, "dur": 10,
    "args": {
      "External id": 45360,
      "cbid": 211, "correlation": 45360
    }
  },
  {
    "ph": "s", "id": 45360, "pid": 76337, "tid": -914061504, "ts": 1716454218225555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218225594, "dur": 3,
    "args": {
      "External id": 45369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45369,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45369, "pid": 5, "tid": 7, "ts": 1716454218225594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225585, "dur": 8,
    "args": {
      "External id": 45369,
      "cbid": 211, "correlation": 45369
    }
  },
  {
    "ph": "s", "id": 45369, "pid": 76337, "tid": -914061504, "ts": 1716454218225585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218225616, "dur": 3,
    "args": {
      "External id": 45378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45378,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45378, "pid": 5, "tid": 7, "ts": 1716454218225616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225608, "dur": 7,
    "args": {
      "External id": 45378,
      "cbid": 211, "correlation": 45378
    }
  },
  {
    "ph": "s", "id": 45378, "pid": 76337, "tid": -914061504, "ts": 1716454218225608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218225910, "dur": 3,
    "args": {
      "External id": 45386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45386,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45386, "pid": 5, "tid": 7, "ts": 1716454218225910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225894, "dur": 15,
    "args": {
      "External id": 45386,
      "cbid": 211, "correlation": 45386
    }
  },
  {
    "ph": "s", "id": 45386, "pid": 76337, "tid": -914061504, "ts": 1716454218225894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218225934, "dur": 3,
    "args": {
      "External id": 45394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45394,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45394, "pid": 5, "tid": 7, "ts": 1716454218225934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225925, "dur": 8,
    "args": {
      "External id": 45394,
      "cbid": 211, "correlation": 45394
    }
  },
  {
    "ph": "s", "id": 45394, "pid": 76337, "tid": -914061504, "ts": 1716454218225925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218226004, "dur": 1,
    "args": {
      "External id": 45404,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45404,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 45404, "pid": 5, "tid": 7, "ts": 1716454218226004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218225969, "dur": 33,
    "args": {
      "External id": 45404,
      "cbid": 41, "correlation": 45404
    }
  },
  {
    "ph": "s", "id": 45404, "pid": 76337, "tid": -914061504, "ts": 1716454218225969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226003, "dur": 8,
    "args": {
      "External id": 45405,
      "cbid": 131, "correlation": 45405
    }
  },
  {
    "ph": "f", "id": 45405, "pid": 76337, "tid": -914061504, "ts": 1716454218226003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454218226096, "dur": 2,
    "args": {
      "External id": 45413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45413, "pid": 5, "tid": 7, "ts": 1716454218226096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226081, "dur": 15,
    "args": {
      "External id": 45413,
      "cbid": 211, "correlation": 45413
    }
  },
  {
    "ph": "s", "id": 45413, "pid": 76337, "tid": -914061504, "ts": 1716454218226081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218226167, "dur": 3,
    "args": {
      "External id": 45422,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45422,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002631578947368421
    }
  },
  {
    "ph": "f", "id": 45422, "pid": 5, "tid": 7, "ts": 1716454218226167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226149, "dur": 18,
    "args": {
      "External id": 45422,
      "cbid": 41, "correlation": 45422
    }
  },
  {
    "ph": "s", "id": 45422, "pid": 76337, "tid": -914061504, "ts": 1716454218226149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218226358, "dur": 4,
    "args": {
      "External id": 45432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45432,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45432, "pid": 5, "tid": 7, "ts": 1716454218226358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226225, "dur": 134,
    "args": {
      "External id": 45432,
      "cbid": 211, "correlation": 45432
    }
  },
  {
    "ph": "s", "id": 45432, "pid": 76337, "tid": -914061504, "ts": 1716454218226225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218226423, "dur": 1,
    "args": {
      "External id": 45442,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45442,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 45442, "pid": 5, "tid": 7, "ts": 1716454218226423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226407, "dur": 14,
    "args": {
      "External id": 45442,
      "cbid": 41, "correlation": 45442
    }
  },
  {
    "ph": "s", "id": 45442, "pid": 76337, "tid": -914061504, "ts": 1716454218226407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226421, "dur": 8,
    "args": {
      "External id": 45443,
      "cbid": 131, "correlation": 45443
    }
  },
  {
    "ph": "f", "id": 45443, "pid": 76337, "tid": -914061504, "ts": 1716454218226421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218226486, "dur": 3,
    "args": {
      "External id": 45450,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45450,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.11764705882353
    }
  },
  {
    "ph": "f", "id": 45450, "pid": 5, "tid": 7, "ts": 1716454218226486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226465, "dur": 20,
    "args": {
      "External id": 45450,
      "cbid": 41, "correlation": 45450
    }
  },
  {
    "ph": "s", "id": 45450, "pid": 76337, "tid": -914061504, "ts": 1716454218226465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218226533, "dur": 1,
    "args": {
      "External id": 45461,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45461,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 45461, "pid": 5, "tid": 7, "ts": 1716454218226533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226520, "dur": 10,
    "args": {
      "External id": 45461,
      "cbid": 41, "correlation": 45461
    }
  },
  {
    "ph": "s", "id": 45461, "pid": 76337, "tid": -914061504, "ts": 1716454218226520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226530, "dur": 10,
    "args": {
      "External id": 45462,
      "cbid": 131, "correlation": 45462
    }
  },
  {
    "ph": "f", "id": 45462, "pid": 76337, "tid": -914061504, "ts": 1716454218226530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218226582, "dur": 3,
    "args": {
      "External id": 45470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45470,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45470, "pid": 5, "tid": 7, "ts": 1716454218226582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226569, "dur": 13,
    "args": {
      "External id": 45470,
      "cbid": 211, "correlation": 45470
    }
  },
  {
    "ph": "s", "id": 45470, "pid": 76337, "tid": -914061504, "ts": 1716454218226569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218226611, "dur": 3,
    "args": {
      "External id": 45480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45480,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45480, "pid": 5, "tid": 7, "ts": 1716454218226611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226602, "dur": 8,
    "args": {
      "External id": 45480,
      "cbid": 211, "correlation": 45480
    }
  },
  {
    "ph": "s", "id": 45480, "pid": 76337, "tid": -914061504, "ts": 1716454218226602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218226634, "dur": 3,
    "args": {
      "External id": 45489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45489,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45489, "pid": 5, "tid": 7, "ts": 1716454218226634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226625, "dur": 8,
    "args": {
      "External id": 45489,
      "cbid": 211, "correlation": 45489
    }
  },
  {
    "ph": "s", "id": 45489, "pid": 76337, "tid": -914061504, "ts": 1716454218226625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218226846, "dur": 5,
    "args": {
      "External id": 45497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45497, "pid": 5, "tid": 7, "ts": 1716454218226846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226830, "dur": 16,
    "args": {
      "External id": 45497,
      "cbid": 211, "correlation": 45497
    }
  },
  {
    "ph": "s", "id": 45497, "pid": 76337, "tid": -914061504, "ts": 1716454218226830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218226893, "dur": 3,
    "args": {
      "External id": 45506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45506,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45506, "pid": 5, "tid": 7, "ts": 1716454218226893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226882, "dur": 9,
    "args": {
      "External id": 45506,
      "cbid": 211, "correlation": 45506
    }
  },
  {
    "ph": "s", "id": 45506, "pid": 76337, "tid": -914061504, "ts": 1716454218226882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218226920, "dur": 3,
    "args": {
      "External id": 45515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45515,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45515, "pid": 5, "tid": 7, "ts": 1716454218226920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226911, "dur": 8,
    "args": {
      "External id": 45515,
      "cbid": 211, "correlation": 45515
    }
  },
  {
    "ph": "s", "id": 45515, "pid": 76337, "tid": -914061504, "ts": 1716454218226911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218226989, "dur": 3,
    "args": {
      "External id": 45523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45523,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45523, "pid": 5, "tid": 7, "ts": 1716454218226989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218226970, "dur": 18,
    "args": {
      "External id": 45523,
      "cbid": 211, "correlation": 45523
    }
  },
  {
    "ph": "s", "id": 45523, "pid": 76337, "tid": -914061504, "ts": 1716454218226970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218227058, "dur": 2,
    "args": {
      "External id": 45531,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45531,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0030120481927710845
    }
  },
  {
    "ph": "f", "id": 45531, "pid": 5, "tid": 7, "ts": 1716454218227058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227036, "dur": 33,
    "args": {
      "External id": 45531,
      "cbid": 41, "correlation": 45531
    }
  },
  {
    "ph": "s", "id": 45531, "pid": 76337, "tid": -914061504, "ts": 1716454218227036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227069, "dur": 4,
    "args": {
      "External id": 45532,
      "cbid": 131, "correlation": 45532
    }
  },
  {
    "ph": "f", "id": 45532, "pid": 76337, "tid": -914061504, "ts": 1716454218227069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218227132, "dur": 1,
    "args": {
      "External id": 45542,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45542,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 45542, "pid": 5, "tid": 7, "ts": 1716454218227132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227119, "dur": 11,
    "args": {
      "External id": 45542,
      "cbid": 41, "correlation": 45542
    }
  },
  {
    "ph": "s", "id": 45542, "pid": 76337, "tid": -914061504, "ts": 1716454218227119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227131, "dur": 7,
    "args": {
      "External id": 45543,
      "cbid": 131, "correlation": 45543
    }
  },
  {
    "ph": "f", "id": 45543, "pid": 76337, "tid": -914061504, "ts": 1716454218227131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218227195, "dur": 1,
    "args": {
      "External id": 45552,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45552,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 45552, "pid": 5, "tid": 7, "ts": 1716454218227195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227170, "dur": 23,
    "args": {
      "External id": 45552,
      "cbid": 41, "correlation": 45552
    }
  },
  {
    "ph": "s", "id": 45552, "pid": 76337, "tid": -914061504, "ts": 1716454218227170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227194, "dur": 8,
    "args": {
      "External id": 45553,
      "cbid": 131, "correlation": 45553
    }
  },
  {
    "ph": "f", "id": 45553, "pid": 76337, "tid": -914061504, "ts": 1716454218227194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218227269, "dur": 4,
    "args": {
      "External id": 45560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45560, "pid": 5, "tid": 7, "ts": 1716454218227269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227253, "dur": 17,
    "args": {
      "External id": 45560,
      "cbid": 211, "correlation": 45560
    }
  },
  {
    "ph": "s", "id": 45560, "pid": 76337, "tid": -914061504, "ts": 1716454218227253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454218227316, "dur": 4,
    "args": {
      "External id": 45580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45580,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45580, "pid": 5, "tid": 7, "ts": 1716454218227316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227295, "dur": 21,
    "args": {
      "External id": 45580,
      "cbid": 211, "correlation": 45580
    }
  },
  {
    "ph": "s", "id": 45580, "pid": 76337, "tid": -914061504, "ts": 1716454218227295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227317, "dur": 0,
    "args": {
      "External id": 45581,
      "cbid": 11, "correlation": 45581
    }
  },
  {
    "ph": "f", "id": 45581, "pid": 76337, "tid": -914061504, "ts": 1716454218227317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227318, "dur": 0,
    "args": {
      "External id": 45582,
      "cbid": 11, "correlation": 45582
    }
  },
  {
    "ph": "f", "id": 45582, "pid": 76337, "tid": -914061504, "ts": 1716454218227318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218227334, "dur": 1,
    "args": {
      "External id": 45585,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45585,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 45585, "pid": 5, "tid": 7, "ts": 1716454218227334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227319, "dur": 23,
    "args": {
      "External id": 45585,
      "cbid": 41, "correlation": 45585
    }
  },
  {
    "ph": "s", "id": 45585, "pid": 76337, "tid": -914061504, "ts": 1716454218227319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227343, "dur": 3,
    "args": {
      "External id": 45586,
      "cbid": 131, "correlation": 45586
    }
  },
  {
    "ph": "f", "id": 45586, "pid": 76337, "tid": -914061504, "ts": 1716454218227343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218227379, "dur": 3,
    "args": {
      "External id": 45610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45610,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45610, "pid": 5, "tid": 7, "ts": 1716454218227379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227368, "dur": 11,
    "args": {
      "External id": 45610,
      "cbid": 211, "correlation": 45610
    }
  },
  {
    "ph": "s", "id": 45610, "pid": 76337, "tid": -914061504, "ts": 1716454218227368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227379, "dur": 0,
    "args": {
      "External id": 45611,
      "cbid": 11, "correlation": 45611
    }
  },
  {
    "ph": "f", "id": 45611, "pid": 76337, "tid": -914061504, "ts": 1716454218227379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227380, "dur": 0,
    "args": {
      "External id": 45612,
      "cbid": 11, "correlation": 45612
    }
  },
  {
    "ph": "f", "id": 45612, "pid": 76337, "tid": -914061504, "ts": 1716454218227380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227381, "dur": 1,
    "args": {
      "External id": 45614,
      "cbid": 200, "correlation": 45614
    }
  },
  {
    "ph": "f", "id": 45614, "pid": 76337, "tid": -914061504, "ts": 1716454218227381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454218227391, "dur": 4,
    "args": {
      "External id": 45616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45616,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45616, "pid": 5, "tid": 7, "ts": 1716454218227391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227384, "dur": 7,
    "args": {
      "External id": 45616,
      "cbid": 211, "correlation": 45616
    }
  },
  {
    "ph": "s", "id": 45616, "pid": 76337, "tid": -914061504, "ts": 1716454218227384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227392, "dur": 0,
    "args": {
      "External id": 45617,
      "cbid": 11, "correlation": 45617
    }
  },
  {
    "ph": "f", "id": 45617, "pid": 76337, "tid": -914061504, "ts": 1716454218227392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227392, "dur": 0,
    "args": {
      "External id": 45618,
      "cbid": 11, "correlation": 45618
    }
  },
  {
    "ph": "f", "id": 45618, "pid": 76337, "tid": -914061504, "ts": 1716454218227392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454218227432, "dur": 1,
    "args": {
      "External id": 45625,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45625,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 45625, "pid": 5, "tid": 7, "ts": 1716454218227432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227419, "dur": 22,
    "args": {
      "External id": 45625,
      "cbid": 41, "correlation": 45625
    }
  },
  {
    "ph": "s", "id": 45625, "pid": 76337, "tid": -914061504, "ts": 1716454218227419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227441, "dur": 4,
    "args": {
      "External id": 45626,
      "cbid": 131, "correlation": 45626
    }
  },
  {
    "ph": "f", "id": 45626, "pid": 76337, "tid": -914061504, "ts": 1716454218227441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218227493, "dur": 1,
    "args": {
      "External id": 45636,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 45636,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 45636, "pid": 5, "tid": 7, "ts": 1716454218227493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227480, "dur": 9,
    "args": {
      "External id": 45636,
      "cbid": 41, "correlation": 45636
    }
  },
  {
    "ph": "s", "id": 45636, "pid": 76337, "tid": -914061504, "ts": 1716454218227480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227491, "dur": 9,
    "args": {
      "External id": 45637,
      "cbid": 131, "correlation": 45637
    }
  },
  {
    "ph": "f", "id": 45637, "pid": 76337, "tid": -914061504, "ts": 1716454218227491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218227561, "dur": 5,
    "args": {
      "External id": 45644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45644,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45644, "pid": 5, "tid": 7, "ts": 1716454218227561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227545, "dur": 16,
    "args": {
      "External id": 45644,
      "cbid": 211, "correlation": 45644
    }
  },
  {
    "ph": "s", "id": 45644, "pid": 76337, "tid": -914061504, "ts": 1716454218227545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218227630, "dur": 3,
    "args": {
      "External id": 45653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45653,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45653, "pid": 5, "tid": 7, "ts": 1716454218227630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227617, "dur": 13,
    "args": {
      "External id": 45653,
      "cbid": 211, "correlation": 45653
    }
  },
  {
    "ph": "s", "id": 45653, "pid": 76337, "tid": -914061504, "ts": 1716454218227617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218227662, "dur": 3,
    "args": {
      "External id": 45661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45661,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45661, "pid": 5, "tid": 7, "ts": 1716454218227662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227652, "dur": 9,
    "args": {
      "External id": 45661,
      "cbid": 211, "correlation": 45661
    }
  },
  {
    "ph": "s", "id": 45661, "pid": 76337, "tid": -914061504, "ts": 1716454218227652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218227698, "dur": 4,
    "args": {
      "External id": 45669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45669,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45669, "pid": 5, "tid": 7, "ts": 1716454218227698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227686, "dur": 11,
    "args": {
      "External id": 45669,
      "cbid": 211, "correlation": 45669
    }
  },
  {
    "ph": "s", "id": 45669, "pid": 76337, "tid": -914061504, "ts": 1716454218227686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218227728, "dur": 4,
    "args": {
      "External id": 45677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45677,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45677, "pid": 5, "tid": 7, "ts": 1716454218227728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227717, "dur": 10,
    "args": {
      "External id": 45677,
      "cbid": 211, "correlation": 45677
    }
  },
  {
    "ph": "s", "id": 45677, "pid": 76337, "tid": -914061504, "ts": 1716454218227717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218227754, "dur": 3,
    "args": {
      "External id": 45685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45685,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45685, "pid": 5, "tid": 7, "ts": 1716454218227754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227745, "dur": 8,
    "args": {
      "External id": 45685,
      "cbid": 211, "correlation": 45685
    }
  },
  {
    "ph": "s", "id": 45685, "pid": 76337, "tid": -914061504, "ts": 1716454218227745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218227779, "dur": 3,
    "args": {
      "External id": 45693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45693,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45693, "pid": 5, "tid": 7, "ts": 1716454218227779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227771, "dur": 8,
    "args": {
      "External id": 45693,
      "cbid": 211, "correlation": 45693
    }
  },
  {
    "ph": "s", "id": 45693, "pid": 76337, "tid": -914061504, "ts": 1716454218227771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218227800, "dur": 4,
    "args": {
      "External id": 45701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45701, "pid": 5, "tid": 7, "ts": 1716454218227800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227792, "dur": 7,
    "args": {
      "External id": 45701,
      "cbid": 211, "correlation": 45701
    }
  },
  {
    "ph": "s", "id": 45701, "pid": 76337, "tid": -914061504, "ts": 1716454218227792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218227818, "dur": 5,
    "args": {
      "External id": 45709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45709, "pid": 5, "tid": 7, "ts": 1716454218227818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227810, "dur": 7,
    "args": {
      "External id": 45709,
      "cbid": 211, "correlation": 45709
    }
  },
  {
    "ph": "s", "id": 45709, "pid": 76337, "tid": -914061504, "ts": 1716454218227810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218227837, "dur": 3,
    "args": {
      "External id": 45717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45717,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45717, "pid": 5, "tid": 7, "ts": 1716454218227837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227830, "dur": 6,
    "args": {
      "External id": 45717,
      "cbid": 211, "correlation": 45717
    }
  },
  {
    "ph": "s", "id": 45717, "pid": 76337, "tid": -914061504, "ts": 1716454218227830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218227894, "dur": 4,
    "args": {
      "External id": 45725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45725,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 45725, "pid": 5, "tid": 7, "ts": 1716454218227894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227883, "dur": 10,
    "args": {
      "External id": 45725,
      "cbid": 211, "correlation": 45725
    }
  },
  {
    "ph": "s", "id": 45725, "pid": 76337, "tid": -914061504, "ts": 1716454218227883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218227921, "dur": 4,
    "args": {
      "External id": 45733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45733, "pid": 5, "tid": 7, "ts": 1716454218227921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227912, "dur": 9,
    "args": {
      "External id": 45733,
      "cbid": 211, "correlation": 45733
    }
  },
  {
    "ph": "s", "id": 45733, "pid": 76337, "tid": -914061504, "ts": 1716454218227912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218227944, "dur": 4,
    "args": {
      "External id": 45741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45741, "pid": 5, "tid": 7, "ts": 1716454218227944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227936, "dur": 7,
    "args": {
      "External id": 45741,
      "cbid": 211, "correlation": 45741
    }
  },
  {
    "ph": "s", "id": 45741, "pid": 76337, "tid": -914061504, "ts": 1716454218227936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218227963, "dur": 3,
    "args": {
      "External id": 45749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45749,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 45749, "pid": 5, "tid": 7, "ts": 1716454218227963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218227955, "dur": 7,
    "args": {
      "External id": 45749,
      "cbid": 211, "correlation": 45749
    }
  },
  {
    "ph": "s", "id": 45749, "pid": 76337, "tid": -914061504, "ts": 1716454218227955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218228350, "dur": 5,
    "args": {
      "External id": 45758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45758,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45758, "pid": 5, "tid": 7, "ts": 1716454218228350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228333, "dur": 18,
    "args": {
      "External id": 45758,
      "cbid": 211, "correlation": 45758
    }
  },
  {
    "ph": "s", "id": 45758, "pid": 76337, "tid": -914061504, "ts": 1716454218228333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218228386, "dur": 5,
    "args": {
      "External id": 45767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45767,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45767, "pid": 5, "tid": 7, "ts": 1716454218228386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228376, "dur": 9,
    "args": {
      "External id": 45767,
      "cbid": 211, "correlation": 45767
    }
  },
  {
    "ph": "s", "id": 45767, "pid": 76337, "tid": -914061504, "ts": 1716454218228376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454218228513, "dur": 3,
    "args": {
      "External id": 45783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45783, "pid": 5, "tid": 7, "ts": 1716454218228513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228499, "dur": 14,
    "args": {
      "External id": 45783,
      "cbid": 211, "correlation": 45783
    }
  },
  {
    "ph": "s", "id": 45783, "pid": 76337, "tid": -914061504, "ts": 1716454218228499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218228548, "dur": 3,
    "args": {
      "External id": 45791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45791,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45791, "pid": 5, "tid": 7, "ts": 1716454218228548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228539, "dur": 9,
    "args": {
      "External id": 45791,
      "cbid": 211, "correlation": 45791
    }
  },
  {
    "ph": "s", "id": 45791, "pid": 76337, "tid": -914061504, "ts": 1716454218228539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218228579, "dur": 3,
    "args": {
      "External id": 45799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45799, "pid": 5, "tid": 7, "ts": 1716454218228579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228570, "dur": 11,
    "args": {
      "External id": 45799,
      "cbid": 211, "correlation": 45799
    }
  },
  {
    "ph": "s", "id": 45799, "pid": 76337, "tid": -914061504, "ts": 1716454218228570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218228614, "dur": 4,
    "args": {
      "External id": 45807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45807,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45807, "pid": 5, "tid": 7, "ts": 1716454218228614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228604, "dur": 9,
    "args": {
      "External id": 45807,
      "cbid": 211, "correlation": 45807
    }
  },
  {
    "ph": "s", "id": 45807, "pid": 76337, "tid": -914061504, "ts": 1716454218228604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454218228673, "dur": 4,
    "args": {
      "External id": 45819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45819,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45819, "pid": 5, "tid": 7, "ts": 1716454218228673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228660, "dur": 13,
    "args": {
      "External id": 45819,
      "cbid": 211, "correlation": 45819
    }
  },
  {
    "ph": "s", "id": 45819, "pid": 76337, "tid": -914061504, "ts": 1716454218228660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218228720, "dur": 4,
    "args": {
      "External id": 45830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45830, "pid": 5, "tid": 7, "ts": 1716454218228720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228708, "dur": 12,
    "args": {
      "External id": 45830,
      "cbid": 211, "correlation": 45830
    }
  },
  {
    "ph": "s", "id": 45830, "pid": 76337, "tid": -914061504, "ts": 1716454218228708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218228751, "dur": 3,
    "args": {
      "External id": 45838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45838,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45838, "pid": 5, "tid": 7, "ts": 1716454218228751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228742, "dur": 8,
    "args": {
      "External id": 45838,
      "cbid": 211, "correlation": 45838
    }
  },
  {
    "ph": "s", "id": 45838, "pid": 76337, "tid": -914061504, "ts": 1716454218228742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218228784, "dur": 5,
    "args": {
      "External id": 45846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45846,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45846, "pid": 5, "tid": 7, "ts": 1716454218228784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228774, "dur": 11,
    "args": {
      "External id": 45846,
      "cbid": 211, "correlation": 45846
    }
  },
  {
    "ph": "s", "id": 45846, "pid": 76337, "tid": -914061504, "ts": 1716454218228774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218228817, "dur": 5,
    "args": {
      "External id": 45854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45854,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45854, "pid": 5, "tid": 7, "ts": 1716454218228817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228803, "dur": 13,
    "args": {
      "External id": 45854,
      "cbid": 211, "correlation": 45854
    }
  },
  {
    "ph": "s", "id": 45854, "pid": 76337, "tid": -914061504, "ts": 1716454218228803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218228847, "dur": 4,
    "args": {
      "External id": 45863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45863,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45863, "pid": 5, "tid": 7, "ts": 1716454218228847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228837, "dur": 10,
    "args": {
      "External id": 45863,
      "cbid": 211, "correlation": 45863
    }
  },
  {
    "ph": "s", "id": 45863, "pid": 76337, "tid": -914061504, "ts": 1716454218228837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218228911, "dur": 5,
    "args": {
      "External id": 45876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45876,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45876, "pid": 5, "tid": 7, "ts": 1716454218228911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228898, "dur": 13,
    "args": {
      "External id": 45876,
      "cbid": 211, "correlation": 45876
    }
  },
  {
    "ph": "s", "id": 45876, "pid": 76337, "tid": -914061504, "ts": 1716454218228898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454218228954, "dur": 8,
    "args": {
      "External id": 45886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45886,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 45886, "pid": 5, "tid": 7, "ts": 1716454218228954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218228942, "dur": 12,
    "args": {
      "External id": 45886,
      "cbid": 211, "correlation": 45886
    }
  },
  {
    "ph": "s", "id": 45886, "pid": 76337, "tid": -914061504, "ts": 1716454218228942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229080, "dur": 4,
    "args": {
      "External id": 45903,
      "cbid": 251, "correlation": 45903
    }
  },
  {
    "ph": "f", "id": 45903, "pid": 76337, "tid": -914061504, "ts": 1716454218229080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454218229107, "dur": 12,
    "args": {
      "External id": 45905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45905,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 45905, "pid": 5, "tid": 7, "ts": 1716454218229107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229092, "dur": 16,
    "args": {
      "External id": 45905,
      "cbid": 211, "correlation": 45905
    }
  },
  {
    "ph": "s", "id": 45905, "pid": 76337, "tid": -914061504, "ts": 1716454218229092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218229162, "dur": 4,
    "args": {
      "External id": 45913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45913,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 45913, "pid": 5, "tid": 7, "ts": 1716454218229162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229151, "dur": 11,
    "args": {
      "External id": 45913,
      "cbid": 211, "correlation": 45913
    }
  },
  {
    "ph": "s", "id": 45913, "pid": 76337, "tid": -914061504, "ts": 1716454218229151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229217, "dur": 1,
    "args": {
      "External id": 45929,
      "cbid": 251, "correlation": 45929
    }
  },
  {
    "ph": "f", "id": 45929, "pid": 76337, "tid": -914061504, "ts": 1716454218229217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229223, "dur": 0,
    "args": {
      "External id": 45931,
      "cbid": 251, "correlation": 45931
    }
  },
  {
    "ph": "f", "id": 45931, "pid": 76337, "tid": -914061504, "ts": 1716454218229223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218229240, "dur": 13,
    "args": {
      "External id": 45932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45932,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 45932, "pid": 5, "tid": 7, "ts": 1716454218229240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229226, "dur": 14,
    "args": {
      "External id": 45932,
      "cbid": 211, "correlation": 45932
    }
  },
  {
    "ph": "s", "id": 45932, "pid": 76337, "tid": -914061504, "ts": 1716454218229226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218229254, "dur": 5,
    "args": {
      "External id": 45934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45934,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 45934, "pid": 5, "tid": 7, "ts": 1716454218229254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229244, "dur": 9,
    "args": {
      "External id": 45934,
      "cbid": 211, "correlation": 45934
    }
  },
  {
    "ph": "s", "id": 45934, "pid": 76337, "tid": -914061504, "ts": 1716454218229244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229343, "dur": 1,
    "args": {
      "External id": 45944,
      "cbid": 317, "correlation": 45944
    }
  },
  {
    "ph": "f", "id": 45944, "pid": 76337, "tid": -914061504, "ts": 1716454218229343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229345, "dur": 3,
    "args": {
      "External id": 45945,
      "cbid": 203, "correlation": 45945
    }
  },
  {
    "ph": "f", "id": 45945, "pid": 76337, "tid": -914061504, "ts": 1716454218229345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229349, "dur": 1,
    "args": {
      "External id": 45946,
      "cbid": 205, "correlation": 45946
    }
  },
  {
    "ph": "f", "id": 45946, "pid": 76337, "tid": -914061504, "ts": 1716454218229349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218229399, "dur": 7,
    "args": {
      "External id": 45950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45950,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45950, "pid": 5, "tid": 7, "ts": 1716454218229399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229384, "dur": 14,
    "args": {
      "External id": 45950,
      "cbid": 211, "correlation": 45950
    }
  },
  {
    "ph": "s", "id": 45950, "pid": 76337, "tid": -914061504, "ts": 1716454218229384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218229410, "dur": 4,
    "args": {
      "External id": 45952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 45952, "pid": 5, "tid": 7, "ts": 1716454218229410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229402, "dur": 6,
    "args": {
      "External id": 45952,
      "cbid": 211, "correlation": 45952
    }
  },
  {
    "ph": "s", "id": 45952, "pid": 76337, "tid": -914061504, "ts": 1716454218229402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218229427, "dur": 3,
    "args": {
      "External id": 45954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 45954, "pid": 5, "tid": 7, "ts": 1716454218229427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229418, "dur": 8,
    "args": {
      "External id": 45954,
      "cbid": 211, "correlation": 45954
    }
  },
  {
    "ph": "s", "id": 45954, "pid": 76337, "tid": -914061504, "ts": 1716454218229418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229432, "dur": 0,
    "args": {
      "External id": 45955,
      "cbid": 51, "correlation": 45955
    }
  },
  {
    "ph": "s", "id": 45955, "pid": 76337, "tid": -914061504, "ts": 1716454218229432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218229441, "dur": 85,
    "args": {
      "External id": 45956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45956,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 45956, "pid": 5, "tid": 7, "ts": 1716454218229441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229433, "dur": 6,
    "args": {
      "External id": 45956,
      "cbid": 211, "correlation": 45956
    }
  },
  {
    "ph": "s", "id": 45956, "pid": 76337, "tid": -914061504, "ts": 1716454218229433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218229528, "dur": 60,
    "args": {
      "External id": 45961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45961, "pid": 5, "tid": 7, "ts": 1716454218229528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218229467, "dur": 10,
    "args": {
      "External id": 45961,
      "cbid": 211, "correlation": 45961
    }
  },
  {
    "ph": "s", "id": 45961, "pid": 76337, "tid": -914061504, "ts": 1716454218229467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218231232, "dur": 52,
    "args": {
      "External id": 45981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45981,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 45981, "pid": 5, "tid": 7, "ts": 1716454218231232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231217, "dur": 16,
    "args": {
      "External id": 45981,
      "cbid": 211, "correlation": 45981
    }
  },
  {
    "ph": "s", "id": 45981, "pid": 76337, "tid": -914061504, "ts": 1716454218231217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218231286, "dur": 4,
    "args": {
      "External id": 45993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45993,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 45993, "pid": 5, "tid": 7, "ts": 1716454218231286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231245, "dur": 8,
    "args": {
      "External id": 45993,
      "cbid": 211, "correlation": 45993
    }
  },
  {
    "ph": "s", "id": 45993, "pid": 76337, "tid": -914061504, "ts": 1716454218231245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218231292, "dur": 59,
    "args": {
      "External id": 45996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 45996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 45996, "pid": 5, "tid": 7, "ts": 1716454218231292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231268, "dur": 8,
    "args": {
      "External id": 45996,
      "cbid": 211, "correlation": 45996
    }
  },
  {
    "ph": "s", "id": 45996, "pid": 76337, "tid": -914061504, "ts": 1716454218231268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218231352, "dur": 36,
    "args": {
      "External id": 46005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46005, "pid": 5, "tid": 7, "ts": 1716454218231352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231313, "dur": 10,
    "args": {
      "External id": 46005,
      "cbid": 211, "correlation": 46005
    }
  },
  {
    "ph": "s", "id": 46005, "pid": 76337, "tid": -914061504, "ts": 1716454218231313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231375, "dur": 0,
    "args": {
      "External id": 46015,
      "cbid": 317, "correlation": 46015
    }
  },
  {
    "ph": "f", "id": 46015, "pid": 76337, "tid": -914061504, "ts": 1716454218231375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231376, "dur": 0,
    "args": {
      "External id": 46016,
      "cbid": 203, "correlation": 46016
    }
  },
  {
    "ph": "f", "id": 46016, "pid": 76337, "tid": -914061504, "ts": 1716454218231376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231377, "dur": 0,
    "args": {
      "External id": 46017,
      "cbid": 205, "correlation": 46017
    }
  },
  {
    "ph": "f", "id": 46017, "pid": 76337, "tid": -914061504, "ts": 1716454218231377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218231409, "dur": 40,
    "args": {
      "External id": 46021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46021, "pid": 5, "tid": 7, "ts": 1716454218231409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231395, "dur": 13,
    "args": {
      "External id": 46021,
      "cbid": 211, "correlation": 46021
    }
  },
  {
    "ph": "s", "id": 46021, "pid": 76337, "tid": -914061504, "ts": 1716454218231395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218231449, "dur": 15,
    "args": {
      "External id": 46023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46023, "pid": 5, "tid": 7, "ts": 1716454218231449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231411, "dur": 6,
    "args": {
      "External id": 46023,
      "cbid": 211, "correlation": 46023
    }
  },
  {
    "ph": "s", "id": 46023, "pid": 76337, "tid": -914061504, "ts": 1716454218231411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218231465, "dur": 3,
    "args": {
      "External id": 46025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46025, "pid": 5, "tid": 7, "ts": 1716454218231465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231422, "dur": 6,
    "args": {
      "External id": 46025,
      "cbid": 211, "correlation": 46025
    }
  },
  {
    "ph": "s", "id": 46025, "pid": 76337, "tid": -914061504, "ts": 1716454218231422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231432, "dur": 0,
    "args": {
      "External id": 46026,
      "cbid": 51, "correlation": 46026
    }
  },
  {
    "ph": "s", "id": 46026, "pid": 76337, "tid": -914061504, "ts": 1716454218231432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218231470, "dur": 707,
    "args": {
      "External id": 46027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46027,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46027, "pid": 5, "tid": 7, "ts": 1716454218231470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231433, "dur": 6,
    "args": {
      "External id": 46027,
      "cbid": 211, "correlation": 46027
    }
  },
  {
    "ph": "s", "id": 46027, "pid": 76337, "tid": -914061504, "ts": 1716454218231433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218232179, "dur": 59,
    "args": {
      "External id": 46032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46032, "pid": 5, "tid": 7, "ts": 1716454218232179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231464, "dur": 9,
    "args": {
      "External id": 46032,
      "cbid": 211, "correlation": 46032
    }
  },
  {
    "ph": "s", "id": 46032, "pid": 76337, "tid": -914061504, "ts": 1716454218231464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218232239, "dur": 3,
    "args": {
      "External id": 46040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46040,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46040, "pid": 5, "tid": 7, "ts": 1716454218232239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231507, "dur": 10,
    "args": {
      "External id": 46040,
      "cbid": 211, "correlation": 46040
    }
  },
  {
    "ph": "s", "id": 46040, "pid": 76337, "tid": -914061504, "ts": 1716454218231507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231573, "dur": 2,
    "args": {
      "External id": 46056,
      "cbid": 251, "correlation": 46056
    }
  },
  {
    "ph": "f", "id": 46056, "pid": 76337, "tid": -914061504, "ts": 1716454218231573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231579, "dur": 0,
    "args": {
      "External id": 46058,
      "cbid": 251, "correlation": 46058
    }
  },
  {
    "ph": "f", "id": 46058, "pid": 76337, "tid": -914061504, "ts": 1716454218231579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218232244, "dur": 9,
    "args": {
      "External id": 46059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46059,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 46059, "pid": 5, "tid": 7, "ts": 1716454218232244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231580, "dur": 11,
    "args": {
      "External id": 46059,
      "cbid": 211, "correlation": 46059
    }
  },
  {
    "ph": "s", "id": 46059, "pid": 76337, "tid": -914061504, "ts": 1716454218231580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218232255, "dur": 4,
    "args": {
      "External id": 46061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46061,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 46061, "pid": 5, "tid": 7, "ts": 1716454218232255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231594, "dur": 6,
    "args": {
      "External id": 46061,
      "cbid": 211, "correlation": 46061
    }
  },
  {
    "ph": "s", "id": 46061, "pid": 76337, "tid": -914061504, "ts": 1716454218231594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218232260, "dur": 54,
    "args": {
      "External id": 46071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46071,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46071, "pid": 5, "tid": 7, "ts": 1716454218232260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231654, "dur": 16,
    "args": {
      "External id": 46071,
      "cbid": 211, "correlation": 46071
    }
  },
  {
    "ph": "s", "id": 46071, "pid": 76337, "tid": -914061504, "ts": 1716454218231654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218232315, "dur": 52,
    "args": {
      "External id": 46091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46091,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 46091, "pid": 5, "tid": 7, "ts": 1716454218232315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231726, "dur": 11,
    "args": {
      "External id": 46091,
      "cbid": 211, "correlation": 46091
    }
  },
  {
    "ph": "s", "id": 46091, "pid": 76337, "tid": -914061504, "ts": 1716454218231726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218232369, "dur": 4,
    "args": {
      "External id": 46103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46103,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46103, "pid": 5, "tid": 7, "ts": 1716454218232369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231747, "dur": 6,
    "args": {
      "External id": 46103,
      "cbid": 211, "correlation": 46103
    }
  },
  {
    "ph": "s", "id": 46103, "pid": 76337, "tid": -914061504, "ts": 1716454218231747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218232374, "dur": 54,
    "args": {
      "External id": 46106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46106, "pid": 5, "tid": 7, "ts": 1716454218232374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231766, "dur": 7,
    "args": {
      "External id": 46106,
      "cbid": 211, "correlation": 46106
    }
  },
  {
    "ph": "s", "id": 46106, "pid": 76337, "tid": -914061504, "ts": 1716454218231766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218232430, "dur": 37,
    "args": {
      "External id": 46115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46115,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46115, "pid": 5, "tid": 7, "ts": 1716454218232430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231807, "dur": 10,
    "args": {
      "External id": 46115,
      "cbid": 211, "correlation": 46115
    }
  },
  {
    "ph": "s", "id": 46115, "pid": 76337, "tid": -914061504, "ts": 1716454218231807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231876, "dur": 0,
    "args": {
      "External id": 46125,
      "cbid": 317, "correlation": 46125
    }
  },
  {
    "ph": "f", "id": 46125, "pid": 76337, "tid": -914061504, "ts": 1716454218231876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231877, "dur": 0,
    "args": {
      "External id": 46126,
      "cbid": 203, "correlation": 46126
    }
  },
  {
    "ph": "f", "id": 46126, "pid": 76337, "tid": -914061504, "ts": 1716454218231877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231878, "dur": 0,
    "args": {
      "External id": 46127,
      "cbid": 205, "correlation": 46127
    }
  },
  {
    "ph": "f", "id": 46127, "pid": 76337, "tid": -914061504, "ts": 1716454218231878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218232468, "dur": 40,
    "args": {
      "External id": 46131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46131,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46131, "pid": 5, "tid": 7, "ts": 1716454218232468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231892, "dur": 12,
    "args": {
      "External id": 46131,
      "cbid": 211, "correlation": 46131
    }
  },
  {
    "ph": "s", "id": 46131, "pid": 76337, "tid": -914061504, "ts": 1716454218231892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218232509, "dur": 14,
    "args": {
      "External id": 46133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46133,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46133, "pid": 5, "tid": 7, "ts": 1716454218232509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231906, "dur": 5,
    "args": {
      "External id": 46133,
      "cbid": 211, "correlation": 46133
    }
  },
  {
    "ph": "s", "id": 46133, "pid": 76337, "tid": -914061504, "ts": 1716454218231906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218232525, "dur": 3,
    "args": {
      "External id": 46135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46135, "pid": 5, "tid": 7, "ts": 1716454218232525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231915, "dur": 5,
    "args": {
      "External id": 46135,
      "cbid": 211, "correlation": 46135
    }
  },
  {
    "ph": "s", "id": 46135, "pid": 76337, "tid": -914061504, "ts": 1716454218231915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231924, "dur": 0,
    "args": {
      "External id": 46136,
      "cbid": 51, "correlation": 46136
    }
  },
  {
    "ph": "s", "id": 46136, "pid": 76337, "tid": -914061504, "ts": 1716454218231924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218232529, "dur": 699,
    "args": {
      "External id": 46137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46137,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46137, "pid": 5, "tid": 7, "ts": 1716454218232529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231925, "dur": 8,
    "args": {
      "External id": 46137,
      "cbid": 211, "correlation": 46137
    }
  },
  {
    "ph": "s", "id": 46137, "pid": 76337, "tid": -914061504, "ts": 1716454218231925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218233229, "dur": 60,
    "args": {
      "External id": 46142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46142, "pid": 5, "tid": 7, "ts": 1716454218233229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231955, "dur": 9,
    "args": {
      "External id": 46142,
      "cbid": 211, "correlation": 46142
    }
  },
  {
    "ph": "s", "id": 46142, "pid": 76337, "tid": -914061504, "ts": 1716454218231955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218233290, "dur": 50,
    "args": {
      "External id": 46150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46150,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46150, "pid": 5, "tid": 7, "ts": 1716454218233290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218231996, "dur": 9,
    "args": {
      "External id": 46150,
      "cbid": 211, "correlation": 46150
    }
  },
  {
    "ph": "s", "id": 46150, "pid": 76337, "tid": -914061504, "ts": 1716454218231996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218233342, "dur": 35,
    "args": {
      "External id": 46158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46158,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46158, "pid": 5, "tid": 7, "ts": 1716454218233342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232028, "dur": 10,
    "args": {
      "External id": 46158,
      "cbid": 211, "correlation": 46158
    }
  },
  {
    "ph": "s", "id": 46158, "pid": 76337, "tid": -914061504, "ts": 1716454218232028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218233378, "dur": 51,
    "args": {
      "External id": 46178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46178,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 46178, "pid": 5, "tid": 7, "ts": 1716454218233378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232111, "dur": 12,
    "args": {
      "External id": 46178,
      "cbid": 211, "correlation": 46178
    }
  },
  {
    "ph": "s", "id": 46178, "pid": 76337, "tid": -914061504, "ts": 1716454218232111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218233431, "dur": 4,
    "args": {
      "External id": 46190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46190,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46190, "pid": 5, "tid": 7, "ts": 1716454218233431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232133, "dur": 6,
    "args": {
      "External id": 46190,
      "cbid": 211, "correlation": 46190
    }
  },
  {
    "ph": "s", "id": 46190, "pid": 76337, "tid": -914061504, "ts": 1716454218232133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218233436, "dur": 55,
    "args": {
      "External id": 46193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46193, "pid": 5, "tid": 7, "ts": 1716454218233436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232151, "dur": 6,
    "args": {
      "External id": 46193,
      "cbid": 211, "correlation": 46193
    }
  },
  {
    "ph": "s", "id": 46193, "pid": 76337, "tid": -914061504, "ts": 1716454218232151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232207, "dur": 0,
    "args": {
      "External id": 46204,
      "cbid": 317, "correlation": 46204
    }
  },
  {
    "ph": "f", "id": 46204, "pid": 76337, "tid": -914061504, "ts": 1716454218232207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232208, "dur": 0,
    "args": {
      "External id": 46205,
      "cbid": 203, "correlation": 46205
    }
  },
  {
    "ph": "f", "id": 46205, "pid": 76337, "tid": -914061504, "ts": 1716454218232208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232209, "dur": 0,
    "args": {
      "External id": 46206,
      "cbid": 205, "correlation": 46206
    }
  },
  {
    "ph": "f", "id": 46206, "pid": 76337, "tid": -914061504, "ts": 1716454218232209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232242, "dur": 2,
    "args": {
      "External id": 46210,
      "cbid": 251, "correlation": 46210
    }
  },
  {
    "ph": "f", "id": 46210, "pid": 76337, "tid": -914061504, "ts": 1716454218232242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232245, "dur": 1,
    "args": {
      "External id": 46211,
      "cbid": 251, "correlation": 46211
    }
  },
  {
    "ph": "f", "id": 46211, "pid": 76337, "tid": -914061504, "ts": 1716454218232245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232247, "dur": 1,
    "args": {
      "External id": 46212,
      "cbid": 251, "correlation": 46212
    }
  },
  {
    "ph": "f", "id": 46212, "pid": 76337, "tid": -914061504, "ts": 1716454218232247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232250, "dur": 4,
    "args": {
      "External id": 46213,
      "cbid": 251, "correlation": 46213
    }
  },
  {
    "ph": "f", "id": 46213, "pid": 76337, "tid": -914061504, "ts": 1716454218232250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232254, "dur": 0,
    "args": {
      "External id": 46214,
      "cbid": 251, "correlation": 46214
    }
  },
  {
    "ph": "f", "id": 46214, "pid": 76337, "tid": -914061504, "ts": 1716454218232254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232256, "dur": 1,
    "args": {
      "External id": 46215,
      "cbid": 251, "correlation": 46215
    }
  },
  {
    "ph": "f", "id": 46215, "pid": 76337, "tid": -914061504, "ts": 1716454218232256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232257, "dur": 1,
    "args": {
      "External id": 46216,
      "cbid": 251, "correlation": 46216
    }
  },
  {
    "ph": "f", "id": 46216, "pid": 76337, "tid": -914061504, "ts": 1716454218232257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232259, "dur": 1,
    "args": {
      "External id": 46217,
      "cbid": 251, "correlation": 46217
    }
  },
  {
    "ph": "f", "id": 46217, "pid": 76337, "tid": -914061504, "ts": 1716454218232259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232262, "dur": 0,
    "args": {
      "External id": 46218,
      "cbid": 251, "correlation": 46218
    }
  },
  {
    "ph": "f", "id": 46218, "pid": 76337, "tid": -914061504, "ts": 1716454218232262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218233492, "dur": 115,
    "args": {
      "External id": 46219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46219,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46219, "pid": 5, "tid": 7, "ts": 1716454218233492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232265, "dur": 15,
    "args": {
      "External id": 46219,
      "cbid": 211, "correlation": 46219
    }
  },
  {
    "ph": "s", "id": 46219, "pid": 76337, "tid": -914061504, "ts": 1716454218232265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218233608, "dur": 60,
    "args": {
      "External id": 46225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46225,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46225, "pid": 5, "tid": 7, "ts": 1716454218233608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232304, "dur": 9,
    "args": {
      "External id": 46225,
      "cbid": 211, "correlation": 46225
    }
  },
  {
    "ph": "s", "id": 46225, "pid": 76337, "tid": -914061504, "ts": 1716454218232304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218233670, "dur": 584,
    "args": {
      "External id": 46234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46234,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46234, "pid": 5, "tid": 7, "ts": 1716454218233670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232395, "dur": 15,
    "args": {
      "External id": 46234,
      "cbid": 211, "correlation": 46234
    }
  },
  {
    "ph": "s", "id": 46234, "pid": 76337, "tid": -914061504, "ts": 1716454218232395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218234256, "dur": 183,
    "args": {
      "External id": 46256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46256,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46256, "pid": 5, "tid": 7, "ts": 1716454218234256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232468, "dur": 12,
    "args": {
      "External id": 46256,
      "cbid": 211, "correlation": 46256
    }
  },
  {
    "ph": "s", "id": 46256, "pid": 76337, "tid": -914061504, "ts": 1716454218232468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232584, "dur": 2,
    "args": {
      "External id": 46267,
      "cbid": 251, "correlation": 46267
    }
  },
  {
    "ph": "f", "id": 46267, "pid": 76337, "tid": -914061504, "ts": 1716454218232584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218234440, "dur": 198,
    "args": {
      "External id": 46268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46268,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46268, "pid": 5, "tid": 7, "ts": 1716454218234440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232591, "dur": 14,
    "args": {
      "External id": 46268,
      "cbid": 211, "correlation": 46268
    }
  },
  {
    "ph": "s", "id": 46268, "pid": 76337, "tid": -914061504, "ts": 1716454218232591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232663, "dur": 1,
    "args": {
      "External id": 46279,
      "cbid": 251, "correlation": 46279
    }
  },
  {
    "ph": "f", "id": 46279, "pid": 76337, "tid": -914061504, "ts": 1716454218232663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218234639, "dur": 189,
    "args": {
      "External id": 46280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46280,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46280, "pid": 5, "tid": 7, "ts": 1716454218234639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232667, "dur": 11,
    "args": {
      "External id": 46280,
      "cbid": 211, "correlation": 46280
    }
  },
  {
    "ph": "s", "id": 46280, "pid": 76337, "tid": -914061504, "ts": 1716454218232667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232734, "dur": 1,
    "args": {
      "External id": 46291,
      "cbid": 251, "correlation": 46291
    }
  },
  {
    "ph": "f", "id": 46291, "pid": 76337, "tid": -914061504, "ts": 1716454218232734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218234830, "dur": 189,
    "args": {
      "External id": 46292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46292,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46292, "pid": 5, "tid": 7, "ts": 1716454218234830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232738, "dur": 12,
    "args": {
      "External id": 46292,
      "cbid": 211, "correlation": 46292
    }
  },
  {
    "ph": "s", "id": 46292, "pid": 76337, "tid": -914061504, "ts": 1716454218232738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218235020, "dur": 18796,
    "args": {
      "External id": 46313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46313,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46313, "pid": 5, "tid": 7, "ts": 1716454218235020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232846, "dur": 14,
    "args": {
      "External id": 46313,
      "cbid": 211, "correlation": 46313
    }
  },
  {
    "ph": "s", "id": 46313, "pid": 76337, "tid": -914061504, "ts": 1716454218232846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232956, "dur": 2,
    "args": {
      "External id": 46331,
      "cbid": 251, "correlation": 46331
    }
  },
  {
    "ph": "f", "id": 46331, "pid": 76337, "tid": -914061504, "ts": 1716454218232956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218253818, "dur": 203,
    "args": {
      "External id": 46333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46333,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46333, "pid": 5, "tid": 7, "ts": 1716454218253818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218232963, "dur": 20,
    "args": {
      "External id": 46333,
      "cbid": 211, "correlation": 46333
    }
  },
  {
    "ph": "s", "id": 46333, "pid": 76337, "tid": -914061504, "ts": 1716454218232963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218254022, "dur": 66,
    "args": {
      "External id": 46341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46341,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46341, "pid": 5, "tid": 7, "ts": 1716454218254022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233044, "dur": 13,
    "args": {
      "External id": 46341,
      "cbid": 211, "correlation": 46341
    }
  },
  {
    "ph": "s", "id": 46341, "pid": 76337, "tid": -914061504, "ts": 1716454218233044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218254090, "dur": 97,
    "args": {
      "External id": 46349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46349,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46349, "pid": 5, "tid": 7, "ts": 1716454218254090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233083, "dur": 9,
    "args": {
      "External id": 46349,
      "cbid": 211, "correlation": 46349
    }
  },
  {
    "ph": "s", "id": 46349, "pid": 76337, "tid": -914061504, "ts": 1716454218233083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218254187, "dur": 55,
    "args": {
      "External id": 46360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46360, "pid": 5, "tid": 7, "ts": 1716454218254187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233167, "dur": 14,
    "args": {
      "External id": 46360,
      "cbid": 211, "correlation": 46360
    }
  },
  {
    "ph": "s", "id": 46360, "pid": 76337, "tid": -914061504, "ts": 1716454218233167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218254244, "dur": 93,
    "args": {
      "External id": 46382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46382,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46382, "pid": 5, "tid": 7, "ts": 1716454218254244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233204, "dur": 8,
    "args": {
      "External id": 46382,
      "cbid": 211, "correlation": 46382
    }
  },
  {
    "ph": "s", "id": 46382, "pid": 76337, "tid": -914061504, "ts": 1716454218233204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233288, "dur": 1,
    "args": {
      "External id": 46393,
      "cbid": 251, "correlation": 46393
    }
  },
  {
    "ph": "f", "id": 46393, "pid": 76337, "tid": -914061504, "ts": 1716454218233288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218254338, "dur": 106,
    "args": {
      "External id": 46394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46394,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46394, "pid": 5, "tid": 7, "ts": 1716454218254338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233294, "dur": 13,
    "args": {
      "External id": 46394,
      "cbid": 211, "correlation": 46394
    }
  },
  {
    "ph": "s", "id": 46394, "pid": 76337, "tid": -914061504, "ts": 1716454218233294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233378, "dur": 1,
    "args": {
      "External id": 46405,
      "cbid": 251, "correlation": 46405
    }
  },
  {
    "ph": "f", "id": 46405, "pid": 76337, "tid": -914061504, "ts": 1716454218233378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233382, "dur": 0,
    "args": {
      "External id": 46406,
      "cbid": 251, "correlation": 46406
    }
  },
  {
    "ph": "f", "id": 46406, "pid": 76337, "tid": -914061504, "ts": 1716454218233382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218254445, "dur": 10,
    "args": {
      "External id": 46407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46407,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 46407, "pid": 5, "tid": 7, "ts": 1716454218254445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233384, "dur": 14,
    "args": {
      "External id": 46407,
      "cbid": 211, "correlation": 46407
    }
  },
  {
    "ph": "s", "id": 46407, "pid": 76337, "tid": -914061504, "ts": 1716454218233384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218254457, "dur": 5,
    "args": {
      "External id": 46409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46409,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 46409, "pid": 5, "tid": 7, "ts": 1716454218254457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233401, "dur": 7,
    "args": {
      "External id": 46409,
      "cbid": 211, "correlation": 46409
    }
  },
  {
    "ph": "s", "id": 46409, "pid": 76337, "tid": -914061504, "ts": 1716454218233401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233462, "dur": 1,
    "args": {
      "External id": 46420,
      "cbid": 251, "correlation": 46420
    }
  },
  {
    "ph": "f", "id": 46420, "pid": 76337, "tid": -914061504, "ts": 1716454218233462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233465, "dur": 0,
    "args": {
      "External id": 46421,
      "cbid": 251, "correlation": 46421
    }
  },
  {
    "ph": "f", "id": 46421, "pid": 76337, "tid": -914061504, "ts": 1716454218233465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218254463, "dur": 6,
    "args": {
      "External id": 46422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46422,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 46422, "pid": 5, "tid": 7, "ts": 1716454218254463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233467, "dur": 11,
    "args": {
      "External id": 46422,
      "cbid": 211, "correlation": 46422
    }
  },
  {
    "ph": "s", "id": 46422, "pid": 76337, "tid": -914061504, "ts": 1716454218233467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218254470, "dur": 3,
    "args": {
      "External id": 46424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46424,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 46424, "pid": 5, "tid": 7, "ts": 1716454218254470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233480, "dur": 5,
    "args": {
      "External id": 46424,
      "cbid": 211, "correlation": 46424
    }
  },
  {
    "ph": "s", "id": 46424, "pid": 76337, "tid": -914061504, "ts": 1716454218233480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218254475, "dur": 155,
    "args": {
      "External id": 46445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46445,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46445, "pid": 5, "tid": 7, "ts": 1716454218254475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233555, "dur": 13,
    "args": {
      "External id": 46445,
      "cbid": 211, "correlation": 46445
    }
  },
  {
    "ph": "s", "id": 46445, "pid": 76337, "tid": -914061504, "ts": 1716454218233555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233654, "dur": 2,
    "args": {
      "External id": 46463,
      "cbid": 251, "correlation": 46463
    }
  },
  {
    "ph": "f", "id": 46463, "pid": 76337, "tid": -914061504, "ts": 1716454218233654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218254631, "dur": 109,
    "args": {
      "External id": 46465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46465,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46465, "pid": 5, "tid": 7, "ts": 1716454218254631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233661, "dur": 14,
    "args": {
      "External id": 46465,
      "cbid": 211, "correlation": 46465
    }
  },
  {
    "ph": "s", "id": 46465, "pid": 76337, "tid": -914061504, "ts": 1716454218233661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218254742, "dur": 35,
    "args": {
      "External id": 46473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46473,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46473, "pid": 5, "tid": 7, "ts": 1716454218254742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233732, "dur": 12,
    "args": {
      "External id": 46473,
      "cbid": 211, "correlation": 46473
    }
  },
  {
    "ph": "s", "id": 46473, "pid": 76337, "tid": -914061504, "ts": 1716454218233732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218254778, "dur": 67,
    "args": {
      "External id": 46481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46481,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46481, "pid": 5, "tid": 7, "ts": 1716454218254778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233774, "dur": 9,
    "args": {
      "External id": 46481,
      "cbid": 211, "correlation": 46481
    }
  },
  {
    "ph": "s", "id": 46481, "pid": 76337, "tid": -914061504, "ts": 1716454218233774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218254846, "dur": 92,
    "args": {
      "External id": 46503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46503,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46503, "pid": 5, "tid": 7, "ts": 1716454218254846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233826, "dur": 10,
    "args": {
      "External id": 46503,
      "cbid": 211, "correlation": 46503
    }
  },
  {
    "ph": "s", "id": 46503, "pid": 76337, "tid": -914061504, "ts": 1716454218233826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233913, "dur": 1,
    "args": {
      "External id": 46519,
      "cbid": 251, "correlation": 46519
    }
  },
  {
    "ph": "f", "id": 46519, "pid": 76337, "tid": -914061504, "ts": 1716454218233913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218254940, "dur": 566,
    "args": {
      "External id": 46521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46521,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46521, "pid": 5, "tid": 7, "ts": 1716454218254940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218233919, "dur": 16,
    "args": {
      "External id": 46521,
      "cbid": 211, "correlation": 46521
    }
  },
  {
    "ph": "s", "id": 46521, "pid": 76337, "tid": -914061504, "ts": 1716454218233919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218255507, "dur": 239,
    "args": {
      "External id": 46529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46529, "pid": 5, "tid": 7, "ts": 1716454218255507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234007, "dur": 15,
    "args": {
      "External id": 46529,
      "cbid": 211, "correlation": 46529
    }
  },
  {
    "ph": "s", "id": 46529, "pid": 76337, "tid": -914061504, "ts": 1716454218234007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218255748, "dur": 246,
    "args": {
      "External id": 46537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46537,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46537, "pid": 5, "tid": 7, "ts": 1716454218255748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234043, "dur": 9,
    "args": {
      "External id": 46537,
      "cbid": 211, "correlation": 46537
    }
  },
  {
    "ph": "s", "id": 46537, "pid": 76337, "tid": -914061504, "ts": 1716454218234043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234127, "dur": 2,
    "args": {
      "External id": 46553,
      "cbid": 251, "correlation": 46553
    }
  },
  {
    "ph": "f", "id": 46553, "pid": 76337, "tid": -914061504, "ts": 1716454218234127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234133, "dur": 0,
    "args": {
      "External id": 46555,
      "cbid": 251, "correlation": 46555
    }
  },
  {
    "ph": "f", "id": 46555, "pid": 76337, "tid": -914061504, "ts": 1716454218234133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218255996, "dur": 352,
    "args": {
      "External id": 46556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46556,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 46556, "pid": 5, "tid": 7, "ts": 1716454218255996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234139, "dur": 13,
    "args": {
      "External id": 46556,
      "cbid": 211, "correlation": 46556
    }
  },
  {
    "ph": "s", "id": 46556, "pid": 76337, "tid": -914061504, "ts": 1716454218234139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218256348, "dur": 51,
    "args": {
      "External id": 46564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46564,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46564, "pid": 5, "tid": 7, "ts": 1716454218256348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234182, "dur": 10,
    "args": {
      "External id": 46564,
      "cbid": 211, "correlation": 46564
    }
  },
  {
    "ph": "s", "id": 46564, "pid": 76337, "tid": -914061504, "ts": 1716454218234182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218256401, "dur": 151,
    "args": {
      "External id": 46575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46575, "pid": 5, "tid": 7, "ts": 1716454218256401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234253, "dur": 12,
    "args": {
      "External id": 46575,
      "cbid": 211, "correlation": 46575
    }
  },
  {
    "ph": "s", "id": 46575, "pid": 76337, "tid": -914061504, "ts": 1716454218234253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234317, "dur": 0,
    "args": {
      "External id": 46587,
      "cbid": 317, "correlation": 46587
    }
  },
  {
    "ph": "f", "id": 46587, "pid": 76337, "tid": -914061504, "ts": 1716454218234317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234318, "dur": 0,
    "args": {
      "External id": 46588,
      "cbid": 203, "correlation": 46588
    }
  },
  {
    "ph": "f", "id": 46588, "pid": 76337, "tid": -914061504, "ts": 1716454218234318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234319, "dur": 0,
    "args": {
      "External id": 46589,
      "cbid": 205, "correlation": 46589
    }
  },
  {
    "ph": "f", "id": 46589, "pid": 76337, "tid": -914061504, "ts": 1716454218234319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234343, "dur": 1,
    "args": {
      "External id": 46593,
      "cbid": 251, "correlation": 46593
    }
  },
  {
    "ph": "f", "id": 46593, "pid": 76337, "tid": -914061504, "ts": 1716454218234343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234345, "dur": 0,
    "args": {
      "External id": 46594,
      "cbid": 251, "correlation": 46594
    }
  },
  {
    "ph": "f", "id": 46594, "pid": 76337, "tid": -914061504, "ts": 1716454218234345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234346, "dur": 0,
    "args": {
      "External id": 46595,
      "cbid": 251, "correlation": 46595
    }
  },
  {
    "ph": "f", "id": 46595, "pid": 76337, "tid": -914061504, "ts": 1716454218234346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234347, "dur": 0,
    "args": {
      "External id": 46596,
      "cbid": 251, "correlation": 46596
    }
  },
  {
    "ph": "f", "id": 46596, "pid": 76337, "tid": -914061504, "ts": 1716454218234347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234347, "dur": 0,
    "args": {
      "External id": 46597,
      "cbid": 251, "correlation": 46597
    }
  },
  {
    "ph": "f", "id": 46597, "pid": 76337, "tid": -914061504, "ts": 1716454218234347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234348, "dur": 3,
    "args": {
      "External id": 46598,
      "cbid": 251, "correlation": 46598
    }
  },
  {
    "ph": "f", "id": 46598, "pid": 76337, "tid": -914061504, "ts": 1716454218234348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234352, "dur": 0,
    "args": {
      "External id": 46599,
      "cbid": 251, "correlation": 46599
    }
  },
  {
    "ph": "f", "id": 46599, "pid": 76337, "tid": -914061504, "ts": 1716454218234352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234353, "dur": 0,
    "args": {
      "External id": 46600,
      "cbid": 251, "correlation": 46600
    }
  },
  {
    "ph": "f", "id": 46600, "pid": 76337, "tid": -914061504, "ts": 1716454218234353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234354, "dur": 0,
    "args": {
      "External id": 46601,
      "cbid": 251, "correlation": 46601
    }
  },
  {
    "ph": "f", "id": 46601, "pid": 76337, "tid": -914061504, "ts": 1716454218234354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218256553, "dur": 112,
    "args": {
      "External id": 46602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46602,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46602, "pid": 5, "tid": 7, "ts": 1716454218256553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234357, "dur": 13,
    "args": {
      "External id": 46602,
      "cbid": 211, "correlation": 46602
    }
  },
  {
    "ph": "s", "id": 46602, "pid": 76337, "tid": -914061504, "ts": 1716454218234357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218256667, "dur": 58,
    "args": {
      "External id": 46608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46608, "pid": 5, "tid": 7, "ts": 1716454218256667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234393, "dur": 9,
    "args": {
      "External id": 46608,
      "cbid": 211, "correlation": 46608
    }
  },
  {
    "ph": "s", "id": 46608, "pid": 76337, "tid": -914061504, "ts": 1716454218234393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218256726, "dur": 51,
    "args": {
      "External id": 46616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46616,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46616, "pid": 5, "tid": 7, "ts": 1716454218256726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234425, "dur": 8,
    "args": {
      "External id": 46616,
      "cbid": 211, "correlation": 46616
    }
  },
  {
    "ph": "s", "id": 46616, "pid": 76337, "tid": -914061504, "ts": 1716454218234425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218256778, "dur": 50,
    "args": {
      "External id": 46636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46636,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 46636, "pid": 5, "tid": 7, "ts": 1716454218256778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234500, "dur": 11,
    "args": {
      "External id": 46636,
      "cbid": 211, "correlation": 46636
    }
  },
  {
    "ph": "s", "id": 46636, "pid": 76337, "tid": -914061504, "ts": 1716454218234500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218256829, "dur": 4,
    "args": {
      "External id": 46648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46648,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46648, "pid": 5, "tid": 7, "ts": 1716454218256829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234521, "dur": 7,
    "args": {
      "External id": 46648,
      "cbid": 211, "correlation": 46648
    }
  },
  {
    "ph": "s", "id": 46648, "pid": 76337, "tid": -914061504, "ts": 1716454218234521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218256835, "dur": 54,
    "args": {
      "External id": 46651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46651,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46651, "pid": 5, "tid": 7, "ts": 1716454218256835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234539, "dur": 7,
    "args": {
      "External id": 46651,
      "cbid": 211, "correlation": 46651
    }
  },
  {
    "ph": "s", "id": 46651, "pid": 76337, "tid": -914061504, "ts": 1716454218234539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218256891, "dur": 36,
    "args": {
      "External id": 46660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46660,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46660, "pid": 5, "tid": 7, "ts": 1716454218256891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234577, "dur": 9,
    "args": {
      "External id": 46660,
      "cbid": 211, "correlation": 46660
    }
  },
  {
    "ph": "s", "id": 46660, "pid": 76337, "tid": -914061504, "ts": 1716454218234577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234631, "dur": 0,
    "args": {
      "External id": 46670,
      "cbid": 317, "correlation": 46670
    }
  },
  {
    "ph": "f", "id": 46670, "pid": 76337, "tid": -914061504, "ts": 1716454218234631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234632, "dur": 0,
    "args": {
      "External id": 46671,
      "cbid": 203, "correlation": 46671
    }
  },
  {
    "ph": "f", "id": 46671, "pid": 76337, "tid": -914061504, "ts": 1716454218234632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234633, "dur": 0,
    "args": {
      "External id": 46672,
      "cbid": 205, "correlation": 46672
    }
  },
  {
    "ph": "f", "id": 46672, "pid": 76337, "tid": -914061504, "ts": 1716454218234633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218256928, "dur": 40,
    "args": {
      "External id": 46676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46676,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46676, "pid": 5, "tid": 7, "ts": 1716454218256928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234648, "dur": 12,
    "args": {
      "External id": 46676,
      "cbid": 211, "correlation": 46676
    }
  },
  {
    "ph": "s", "id": 46676, "pid": 76337, "tid": -914061504, "ts": 1716454218234648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218256970, "dur": 14,
    "args": {
      "External id": 46678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46678, "pid": 5, "tid": 7, "ts": 1716454218256970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234663, "dur": 5,
    "args": {
      "External id": 46678,
      "cbid": 211, "correlation": 46678
    }
  },
  {
    "ph": "s", "id": 46678, "pid": 76337, "tid": -914061504, "ts": 1716454218234663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218256985, "dur": 4,
    "args": {
      "External id": 46680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46680, "pid": 5, "tid": 7, "ts": 1716454218256985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234672, "dur": 6,
    "args": {
      "External id": 46680,
      "cbid": 211, "correlation": 46680
    }
  },
  {
    "ph": "s", "id": 46680, "pid": 76337, "tid": -914061504, "ts": 1716454218234672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234681, "dur": 0,
    "args": {
      "External id": 46681,
      "cbid": 51, "correlation": 46681
    }
  },
  {
    "ph": "s", "id": 46681, "pid": 76337, "tid": -914061504, "ts": 1716454218234681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218256990, "dur": 682,
    "args": {
      "External id": 46682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46682,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46682, "pid": 5, "tid": 7, "ts": 1716454218256990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234682, "dur": 5,
    "args": {
      "External id": 46682,
      "cbid": 211, "correlation": 46682
    }
  },
  {
    "ph": "s", "id": 46682, "pid": 76337, "tid": -914061504, "ts": 1716454218234682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218257673, "dur": 57,
    "args": {
      "External id": 46687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46687,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46687, "pid": 5, "tid": 7, "ts": 1716454218257673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234710, "dur": 8,
    "args": {
      "External id": 46687,
      "cbid": 211, "correlation": 46687
    }
  },
  {
    "ph": "s", "id": 46687, "pid": 76337, "tid": -914061504, "ts": 1716454218234710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218257731, "dur": 4,
    "args": {
      "External id": 46695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46695,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46695, "pid": 5, "tid": 7, "ts": 1716454218257731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234753, "dur": 10,
    "args": {
      "External id": 46695,
      "cbid": 211, "correlation": 46695
    }
  },
  {
    "ph": "s", "id": 46695, "pid": 76337, "tid": -914061504, "ts": 1716454218234753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234818, "dur": 1,
    "args": {
      "External id": 46711,
      "cbid": 251, "correlation": 46711
    }
  },
  {
    "ph": "f", "id": 46711, "pid": 76337, "tid": -914061504, "ts": 1716454218234818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234823, "dur": 0,
    "args": {
      "External id": 46713,
      "cbid": 251, "correlation": 46713
    }
  },
  {
    "ph": "f", "id": 46713, "pid": 76337, "tid": -914061504, "ts": 1716454218234823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218257736, "dur": 10,
    "args": {
      "External id": 46714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46714,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 46714, "pid": 5, "tid": 7, "ts": 1716454218257736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234825, "dur": 11,
    "args": {
      "External id": 46714,
      "cbid": 211, "correlation": 46714
    }
  },
  {
    "ph": "s", "id": 46714, "pid": 76337, "tid": -914061504, "ts": 1716454218234825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218257748, "dur": 5,
    "args": {
      "External id": 46716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46716,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 46716, "pid": 5, "tid": 7, "ts": 1716454218257748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234838, "dur": 5,
    "args": {
      "External id": 46716,
      "cbid": 211, "correlation": 46716
    }
  },
  {
    "ph": "s", "id": 46716, "pid": 76337, "tid": -914061504, "ts": 1716454218234838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218257755, "dur": 52,
    "args": {
      "External id": 46726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46726,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46726, "pid": 5, "tid": 7, "ts": 1716454218257755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234896, "dur": 12,
    "args": {
      "External id": 46726,
      "cbid": 211, "correlation": 46726
    }
  },
  {
    "ph": "s", "id": 46726, "pid": 76337, "tid": -914061504, "ts": 1716454218234896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218257807, "dur": 49,
    "args": {
      "External id": 46746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46746,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 46746, "pid": 5, "tid": 7, "ts": 1716454218257807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234967, "dur": 18,
    "args": {
      "External id": 46746,
      "cbid": 211, "correlation": 46746
    }
  },
  {
    "ph": "s", "id": 46746, "pid": 76337, "tid": -914061504, "ts": 1716454218234967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218257858, "dur": 4,
    "args": {
      "External id": 46758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46758,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46758, "pid": 5, "tid": 7, "ts": 1716454218257858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218234995, "dur": 7,
    "args": {
      "External id": 46758,
      "cbid": 211, "correlation": 46758
    }
  },
  {
    "ph": "s", "id": 46758, "pid": 76337, "tid": -914061504, "ts": 1716454218234995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218257863, "dur": 54,
    "args": {
      "External id": 46761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46761,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46761, "pid": 5, "tid": 7, "ts": 1716454218257863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235016, "dur": 7,
    "args": {
      "External id": 46761,
      "cbid": 211, "correlation": 46761
    }
  },
  {
    "ph": "s", "id": 46761, "pid": 76337, "tid": -914061504, "ts": 1716454218235016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218257918, "dur": 36,
    "args": {
      "External id": 46770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46770,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46770, "pid": 5, "tid": 7, "ts": 1716454218257918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235057, "dur": 10,
    "args": {
      "External id": 46770,
      "cbid": 211, "correlation": 46770
    }
  },
  {
    "ph": "s", "id": 46770, "pid": 76337, "tid": -914061504, "ts": 1716454218235057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235119, "dur": 0,
    "args": {
      "External id": 46780,
      "cbid": 317, "correlation": 46780
    }
  },
  {
    "ph": "f", "id": 46780, "pid": 76337, "tid": -914061504, "ts": 1716454218235119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235120, "dur": 0,
    "args": {
      "External id": 46781,
      "cbid": 203, "correlation": 46781
    }
  },
  {
    "ph": "f", "id": 46781, "pid": 76337, "tid": -914061504, "ts": 1716454218235120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235121, "dur": 0,
    "args": {
      "External id": 46782,
      "cbid": 205, "correlation": 46782
    }
  },
  {
    "ph": "f", "id": 46782, "pid": 76337, "tid": -914061504, "ts": 1716454218235121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218257955, "dur": 40,
    "args": {
      "External id": 46786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46786,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46786, "pid": 5, "tid": 7, "ts": 1716454218257955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235135, "dur": 12,
    "args": {
      "External id": 46786,
      "cbid": 211, "correlation": 46786
    }
  },
  {
    "ph": "s", "id": 46786, "pid": 76337, "tid": -914061504, "ts": 1716454218235135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218257997, "dur": 14,
    "args": {
      "External id": 46788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46788, "pid": 5, "tid": 7, "ts": 1716454218257997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235150, "dur": 5,
    "args": {
      "External id": 46788,
      "cbid": 211, "correlation": 46788
    }
  },
  {
    "ph": "s", "id": 46788, "pid": 76337, "tid": -914061504, "ts": 1716454218235150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218258012, "dur": 3,
    "args": {
      "External id": 46790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46790, "pid": 5, "tid": 7, "ts": 1716454218258012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235159, "dur": 6,
    "args": {
      "External id": 46790,
      "cbid": 211, "correlation": 46790
    }
  },
  {
    "ph": "s", "id": 46790, "pid": 76337, "tid": -914061504, "ts": 1716454218235159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235168, "dur": 0,
    "args": {
      "External id": 46791,
      "cbid": 51, "correlation": 46791
    }
  },
  {
    "ph": "s", "id": 46791, "pid": 76337, "tid": -914061504, "ts": 1716454218235168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218258016, "dur": 673,
    "args": {
      "External id": 46792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46792,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46792, "pid": 5, "tid": 7, "ts": 1716454218258016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235169, "dur": 5,
    "args": {
      "External id": 46792,
      "cbid": 211, "correlation": 46792
    }
  },
  {
    "ph": "s", "id": 46792, "pid": 76337, "tid": -914061504, "ts": 1716454218235169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218258691, "dur": 57,
    "args": {
      "External id": 46797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46797, "pid": 5, "tid": 7, "ts": 1716454218258691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235197, "dur": 11,
    "args": {
      "External id": 46797,
      "cbid": 211, "correlation": 46797
    }
  },
  {
    "ph": "s", "id": 46797, "pid": 76337, "tid": -914061504, "ts": 1716454218235197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218258749, "dur": 51,
    "args": {
      "External id": 46805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46805,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46805, "pid": 5, "tid": 7, "ts": 1716454218258749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235232, "dur": 9,
    "args": {
      "External id": 46805,
      "cbid": 211, "correlation": 46805
    }
  },
  {
    "ph": "s", "id": 46805, "pid": 76337, "tid": -914061504, "ts": 1716454218235232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218258801, "dur": 35,
    "args": {
      "External id": 46813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46813,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46813, "pid": 5, "tid": 7, "ts": 1716454218258801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235261, "dur": 8,
    "args": {
      "External id": 46813,
      "cbid": 211, "correlation": 46813
    }
  },
  {
    "ph": "s", "id": 46813, "pid": 76337, "tid": -914061504, "ts": 1716454218235261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218258838, "dur": 50,
    "args": {
      "External id": 46833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46833,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 46833, "pid": 5, "tid": 7, "ts": 1716454218258838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235342, "dur": 13,
    "args": {
      "External id": 46833,
      "cbid": 211, "correlation": 46833
    }
  },
  {
    "ph": "s", "id": 46833, "pid": 76337, "tid": -914061504, "ts": 1716454218235342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218258890, "dur": 4,
    "args": {
      "External id": 46845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46845,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 46845, "pid": 5, "tid": 7, "ts": 1716454218258890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235364, "dur": 6,
    "args": {
      "External id": 46845,
      "cbid": 211, "correlation": 46845
    }
  },
  {
    "ph": "s", "id": 46845, "pid": 76337, "tid": -914061504, "ts": 1716454218235364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218258895, "dur": 54,
    "args": {
      "External id": 46848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46848, "pid": 5, "tid": 7, "ts": 1716454218258895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235381, "dur": 7,
    "args": {
      "External id": 46848,
      "cbid": 211, "correlation": 46848
    }
  },
  {
    "ph": "s", "id": 46848, "pid": 76337, "tid": -914061504, "ts": 1716454218235381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235438, "dur": 0,
    "args": {
      "External id": 46859,
      "cbid": 317, "correlation": 46859
    }
  },
  {
    "ph": "f", "id": 46859, "pid": 76337, "tid": -914061504, "ts": 1716454218235438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235439, "dur": 0,
    "args": {
      "External id": 46860,
      "cbid": 203, "correlation": 46860
    }
  },
  {
    "ph": "f", "id": 46860, "pid": 76337, "tid": -914061504, "ts": 1716454218235439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235440, "dur": 0,
    "args": {
      "External id": 46861,
      "cbid": 205, "correlation": 46861
    }
  },
  {
    "ph": "f", "id": 46861, "pid": 76337, "tid": -914061504, "ts": 1716454218235440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235461, "dur": 1,
    "args": {
      "External id": 46865,
      "cbid": 251, "correlation": 46865
    }
  },
  {
    "ph": "f", "id": 46865, "pid": 76337, "tid": -914061504, "ts": 1716454218235461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235463, "dur": 0,
    "args": {
      "External id": 46866,
      "cbid": 251, "correlation": 46866
    }
  },
  {
    "ph": "f", "id": 46866, "pid": 76337, "tid": -914061504, "ts": 1716454218235463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235464, "dur": 0,
    "args": {
      "External id": 46867,
      "cbid": 251, "correlation": 46867
    }
  },
  {
    "ph": "f", "id": 46867, "pid": 76337, "tid": -914061504, "ts": 1716454218235464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235465, "dur": 0,
    "args": {
      "External id": 46868,
      "cbid": 251, "correlation": 46868
    }
  },
  {
    "ph": "f", "id": 46868, "pid": 76337, "tid": -914061504, "ts": 1716454218235465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235466, "dur": 0,
    "args": {
      "External id": 46869,
      "cbid": 251, "correlation": 46869
    }
  },
  {
    "ph": "f", "id": 46869, "pid": 76337, "tid": -914061504, "ts": 1716454218235466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235466, "dur": 0,
    "args": {
      "External id": 46870,
      "cbid": 251, "correlation": 46870
    }
  },
  {
    "ph": "f", "id": 46870, "pid": 76337, "tid": -914061504, "ts": 1716454218235466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235467, "dur": 0,
    "args": {
      "External id": 46871,
      "cbid": 251, "correlation": 46871
    }
  },
  {
    "ph": "f", "id": 46871, "pid": 76337, "tid": -914061504, "ts": 1716454218235467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235468, "dur": 0,
    "args": {
      "External id": 46872,
      "cbid": 251, "correlation": 46872
    }
  },
  {
    "ph": "f", "id": 46872, "pid": 76337, "tid": -914061504, "ts": 1716454218235468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235469, "dur": 0,
    "args": {
      "External id": 46873,
      "cbid": 251, "correlation": 46873
    }
  },
  {
    "ph": "f", "id": 46873, "pid": 76337, "tid": -914061504, "ts": 1716454218235469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218258950, "dur": 108,
    "args": {
      "External id": 46874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46874,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46874, "pid": 5, "tid": 7, "ts": 1716454218258950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235471, "dur": 15,
    "args": {
      "External id": 46874,
      "cbid": 211, "correlation": 46874
    }
  },
  {
    "ph": "s", "id": 46874, "pid": 76337, "tid": -914061504, "ts": 1716454218235471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218259059, "dur": 58,
    "args": {
      "External id": 46880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46880, "pid": 5, "tid": 7, "ts": 1716454218259059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235509, "dur": 9,
    "args": {
      "External id": 46880,
      "cbid": 211, "correlation": 46880
    }
  },
  {
    "ph": "s", "id": 46880, "pid": 76337, "tid": -914061504, "ts": 1716454218235509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218259118, "dur": 643,
    "args": {
      "External id": 46889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46889,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46889, "pid": 5, "tid": 7, "ts": 1716454218259118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235593, "dur": 15,
    "args": {
      "External id": 46889,
      "cbid": 211, "correlation": 46889
    }
  },
  {
    "ph": "s", "id": 46889, "pid": 76337, "tid": -914061504, "ts": 1716454218235593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218259763, "dur": 173,
    "args": {
      "External id": 46911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46911,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46911, "pid": 5, "tid": 7, "ts": 1716454218259763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235651, "dur": 10,
    "args": {
      "External id": 46911,
      "cbid": 211, "correlation": 46911
    }
  },
  {
    "ph": "s", "id": 46911, "pid": 76337, "tid": -914061504, "ts": 1716454218235651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235736, "dur": 1,
    "args": {
      "External id": 46922,
      "cbid": 251, "correlation": 46922
    }
  },
  {
    "ph": "f", "id": 46922, "pid": 76337, "tid": -914061504, "ts": 1716454218235736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218259937, "dur": 188,
    "args": {
      "External id": 46923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46923,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46923, "pid": 5, "tid": 7, "ts": 1716454218259937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235741, "dur": 13,
    "args": {
      "External id": 46923,
      "cbid": 211, "correlation": 46923
    }
  },
  {
    "ph": "s", "id": 46923, "pid": 76337, "tid": -914061504, "ts": 1716454218235741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235808, "dur": 1,
    "args": {
      "External id": 46934,
      "cbid": 251, "correlation": 46934
    }
  },
  {
    "ph": "f", "id": 46934, "pid": 76337, "tid": -914061504, "ts": 1716454218235808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218260126, "dur": 180,
    "args": {
      "External id": 46935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46935,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46935, "pid": 5, "tid": 7, "ts": 1716454218260126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235812, "dur": 12,
    "args": {
      "External id": 46935,
      "cbid": 211, "correlation": 46935
    }
  },
  {
    "ph": "s", "id": 46935, "pid": 76337, "tid": -914061504, "ts": 1716454218235812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235875, "dur": 1,
    "args": {
      "External id": 46946,
      "cbid": 251, "correlation": 46946
    }
  },
  {
    "ph": "f", "id": 46946, "pid": 76337, "tid": -914061504, "ts": 1716454218235875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218260307, "dur": 179,
    "args": {
      "External id": 46947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46947,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46947, "pid": 5, "tid": 7, "ts": 1716454218260307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235879, "dur": 14,
    "args": {
      "External id": 46947,
      "cbid": 211, "correlation": 46947
    }
  },
  {
    "ph": "s", "id": 46947, "pid": 76337, "tid": -914061504, "ts": 1716454218235879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218260487, "dur": 17734,
    "args": {
      "External id": 46968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46968,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 46968, "pid": 5, "tid": 7, "ts": 1716454218260487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218235963, "dur": 21,
    "args": {
      "External id": 46968,
      "cbid": 211, "correlation": 46968
    }
  },
  {
    "ph": "s", "id": 46968, "pid": 76337, "tid": -914061504, "ts": 1716454218235963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236073, "dur": 1,
    "args": {
      "External id": 46986,
      "cbid": 251, "correlation": 46986
    }
  },
  {
    "ph": "f", "id": 46986, "pid": 76337, "tid": -914061504, "ts": 1716454218236073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218278222, "dur": 196,
    "args": {
      "External id": 46988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46988,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 46988, "pid": 5, "tid": 7, "ts": 1716454218278222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236078, "dur": 13,
    "args": {
      "External id": 46988,
      "cbid": 211, "correlation": 46988
    }
  },
  {
    "ph": "s", "id": 46988, "pid": 76337, "tid": -914061504, "ts": 1716454218236078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218278420, "dur": 66,
    "args": {
      "External id": 46996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 46996,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 46996, "pid": 5, "tid": 7, "ts": 1716454218278420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236150, "dur": 12,
    "args": {
      "External id": 46996,
      "cbid": 211, "correlation": 46996
    }
  },
  {
    "ph": "s", "id": 46996, "pid": 76337, "tid": -914061504, "ts": 1716454218236150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218278487, "dur": 98,
    "args": {
      "External id": 47004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47004,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47004, "pid": 5, "tid": 7, "ts": 1716454218278487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236189, "dur": 9,
    "args": {
      "External id": 47004,
      "cbid": 211, "correlation": 47004
    }
  },
  {
    "ph": "s", "id": 47004, "pid": 76337, "tid": -914061504, "ts": 1716454218236189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218278586, "dur": 53,
    "args": {
      "External id": 47015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47015,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47015, "pid": 5, "tid": 7, "ts": 1716454218278586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236261, "dur": 12,
    "args": {
      "External id": 47015,
      "cbid": 211, "correlation": 47015
    }
  },
  {
    "ph": "s", "id": 47015, "pid": 76337, "tid": -914061504, "ts": 1716454218236261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218278641, "dur": 88,
    "args": {
      "External id": 47037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47037,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47037, "pid": 5, "tid": 7, "ts": 1716454218278641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236296, "dur": 8,
    "args": {
      "External id": 47037,
      "cbid": 211, "correlation": 47037
    }
  },
  {
    "ph": "s", "id": 47037, "pid": 76337, "tid": -914061504, "ts": 1716454218236296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236380, "dur": 1,
    "args": {
      "External id": 47048,
      "cbid": 251, "correlation": 47048
    }
  },
  {
    "ph": "f", "id": 47048, "pid": 76337, "tid": -914061504, "ts": 1716454218236380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218278730, "dur": 100,
    "args": {
      "External id": 47049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47049,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47049, "pid": 5, "tid": 7, "ts": 1716454218278730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236385, "dur": 13,
    "args": {
      "External id": 47049,
      "cbid": 211, "correlation": 47049
    }
  },
  {
    "ph": "s", "id": 47049, "pid": 76337, "tid": -914061504, "ts": 1716454218236385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236456, "dur": 1,
    "args": {
      "External id": 47060,
      "cbid": 251, "correlation": 47060
    }
  },
  {
    "ph": "f", "id": 47060, "pid": 76337, "tid": -914061504, "ts": 1716454218236456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236460, "dur": 0,
    "args": {
      "External id": 47061,
      "cbid": 251, "correlation": 47061
    }
  },
  {
    "ph": "f", "id": 47061, "pid": 76337, "tid": -914061504, "ts": 1716454218236460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218278832, "dur": 10,
    "args": {
      "External id": 47062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47062,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 47062, "pid": 5, "tid": 7, "ts": 1716454218278832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236461, "dur": 12,
    "args": {
      "External id": 47062,
      "cbid": 211, "correlation": 47062
    }
  },
  {
    "ph": "s", "id": 47062, "pid": 76337, "tid": -914061504, "ts": 1716454218236461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218278843, "dur": 5,
    "args": {
      "External id": 47064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47064,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 47064, "pid": 5, "tid": 7, "ts": 1716454218278843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236475, "dur": 6,
    "args": {
      "External id": 47064,
      "cbid": 211, "correlation": 47064
    }
  },
  {
    "ph": "s", "id": 47064, "pid": 76337, "tid": -914061504, "ts": 1716454218236475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236535, "dur": 1,
    "args": {
      "External id": 47075,
      "cbid": 251, "correlation": 47075
    }
  },
  {
    "ph": "f", "id": 47075, "pid": 76337, "tid": -914061504, "ts": 1716454218236535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236538, "dur": 0,
    "args": {
      "External id": 47076,
      "cbid": 251, "correlation": 47076
    }
  },
  {
    "ph": "f", "id": 47076, "pid": 76337, "tid": -914061504, "ts": 1716454218236538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218278850, "dur": 6,
    "args": {
      "External id": 47077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47077,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 47077, "pid": 5, "tid": 7, "ts": 1716454218278850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236540, "dur": 12,
    "args": {
      "External id": 47077,
      "cbid": 211, "correlation": 47077
    }
  },
  {
    "ph": "s", "id": 47077, "pid": 76337, "tid": -914061504, "ts": 1716454218236540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218278857, "dur": 3,
    "args": {
      "External id": 47079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47079,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 47079, "pid": 5, "tid": 7, "ts": 1716454218278857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236554, "dur": 5,
    "args": {
      "External id": 47079,
      "cbid": 211, "correlation": 47079
    }
  },
  {
    "ph": "s", "id": 47079, "pid": 76337, "tid": -914061504, "ts": 1716454218236554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218278862, "dur": 148,
    "args": {
      "External id": 47100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47100,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 47100, "pid": 5, "tid": 7, "ts": 1716454218278862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236626, "dur": 14,
    "args": {
      "External id": 47100,
      "cbid": 211, "correlation": 47100
    }
  },
  {
    "ph": "s", "id": 47100, "pid": 76337, "tid": -914061504, "ts": 1716454218236626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236726, "dur": 1,
    "args": {
      "External id": 47118,
      "cbid": 251, "correlation": 47118
    }
  },
  {
    "ph": "f", "id": 47118, "pid": 76337, "tid": -914061504, "ts": 1716454218236726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218279011, "dur": 104,
    "args": {
      "External id": 47120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47120,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 47120, "pid": 5, "tid": 7, "ts": 1716454218279011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236731, "dur": 13,
    "args": {
      "External id": 47120,
      "cbid": 211, "correlation": 47120
    }
  },
  {
    "ph": "s", "id": 47120, "pid": 76337, "tid": -914061504, "ts": 1716454218236731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218279116, "dur": 35,
    "args": {
      "External id": 47128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47128,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47128, "pid": 5, "tid": 7, "ts": 1716454218279116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236801, "dur": 11,
    "args": {
      "External id": 47128,
      "cbid": 211, "correlation": 47128
    }
  },
  {
    "ph": "s", "id": 47128, "pid": 76337, "tid": -914061504, "ts": 1716454218236801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218279152, "dur": 67,
    "args": {
      "External id": 47136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47136, "pid": 5, "tid": 7, "ts": 1716454218279152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236842, "dur": 9,
    "args": {
      "External id": 47136,
      "cbid": 211, "correlation": 47136
    }
  },
  {
    "ph": "s", "id": 47136, "pid": 76337, "tid": -914061504, "ts": 1716454218236842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218279221, "dur": 88,
    "args": {
      "External id": 47158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47158,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47158, "pid": 5, "tid": 7, "ts": 1716454218279221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236893, "dur": 9,
    "args": {
      "External id": 47158,
      "cbid": 211, "correlation": 47158
    }
  },
  {
    "ph": "s", "id": 47158, "pid": 76337, "tid": -914061504, "ts": 1716454218236893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236985, "dur": 1,
    "args": {
      "External id": 47174,
      "cbid": 251, "correlation": 47174
    }
  },
  {
    "ph": "f", "id": 47174, "pid": 76337, "tid": -914061504, "ts": 1716454218236985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218279310, "dur": 556,
    "args": {
      "External id": 47176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47176,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47176, "pid": 5, "tid": 7, "ts": 1716454218279310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218236991, "dur": 14,
    "args": {
      "External id": 47176,
      "cbid": 211, "correlation": 47176
    }
  },
  {
    "ph": "s", "id": 47176, "pid": 76337, "tid": -914061504, "ts": 1716454218236991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218279867, "dur": 239,
    "args": {
      "External id": 47184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47184, "pid": 5, "tid": 7, "ts": 1716454218279867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237062, "dur": 13,
    "args": {
      "External id": 47184,
      "cbid": 211, "correlation": 47184
    }
  },
  {
    "ph": "s", "id": 47184, "pid": 76337, "tid": -914061504, "ts": 1716454218237062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218280108, "dur": 247,
    "args": {
      "External id": 47192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47192, "pid": 5, "tid": 7, "ts": 1716454218280108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237092, "dur": 9,
    "args": {
      "External id": 47192,
      "cbid": 211, "correlation": 47192
    }
  },
  {
    "ph": "s", "id": 47192, "pid": 76337, "tid": -914061504, "ts": 1716454218237092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237173, "dur": 1,
    "args": {
      "External id": 47208,
      "cbid": 251, "correlation": 47208
    }
  },
  {
    "ph": "f", "id": 47208, "pid": 76337, "tid": -914061504, "ts": 1716454218237173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237178, "dur": 0,
    "args": {
      "External id": 47210,
      "cbid": 251, "correlation": 47210
    }
  },
  {
    "ph": "f", "id": 47210, "pid": 76337, "tid": -914061504, "ts": 1716454218237178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218280356, "dur": 353,
    "args": {
      "External id": 47211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47211,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 47211, "pid": 5, "tid": 7, "ts": 1716454218280356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237181, "dur": 13,
    "args": {
      "External id": 47211,
      "cbid": 211, "correlation": 47211
    }
  },
  {
    "ph": "s", "id": 47211, "pid": 76337, "tid": -914061504, "ts": 1716454218237181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218280711, "dur": 50,
    "args": {
      "External id": 47219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47219,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47219, "pid": 5, "tid": 7, "ts": 1716454218280711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237223, "dur": 10,
    "args": {
      "External id": 47219,
      "cbid": 211, "correlation": 47219
    }
  },
  {
    "ph": "s", "id": 47219, "pid": 76337, "tid": -914061504, "ts": 1716454218237223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218280763, "dur": 151,
    "args": {
      "External id": 47230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47230, "pid": 5, "tid": 7, "ts": 1716454218280763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237290, "dur": 13,
    "args": {
      "External id": 47230,
      "cbid": 211, "correlation": 47230
    }
  },
  {
    "ph": "s", "id": 47230, "pid": 76337, "tid": -914061504, "ts": 1716454218237290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237354, "dur": 0,
    "args": {
      "External id": 47242,
      "cbid": 317, "correlation": 47242
    }
  },
  {
    "ph": "f", "id": 47242, "pid": 76337, "tid": -914061504, "ts": 1716454218237354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237355, "dur": 0,
    "args": {
      "External id": 47243,
      "cbid": 203, "correlation": 47243
    }
  },
  {
    "ph": "f", "id": 47243, "pid": 76337, "tid": -914061504, "ts": 1716454218237355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237356, "dur": 0,
    "args": {
      "External id": 47244,
      "cbid": 205, "correlation": 47244
    }
  },
  {
    "ph": "f", "id": 47244, "pid": 76337, "tid": -914061504, "ts": 1716454218237356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237379, "dur": 1,
    "args": {
      "External id": 47248,
      "cbid": 251, "correlation": 47248
    }
  },
  {
    "ph": "f", "id": 47248, "pid": 76337, "tid": -914061504, "ts": 1716454218237379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237381, "dur": 0,
    "args": {
      "External id": 47249,
      "cbid": 251, "correlation": 47249
    }
  },
  {
    "ph": "f", "id": 47249, "pid": 76337, "tid": -914061504, "ts": 1716454218237381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237382, "dur": 0,
    "args": {
      "External id": 47250,
      "cbid": 251, "correlation": 47250
    }
  },
  {
    "ph": "f", "id": 47250, "pid": 76337, "tid": -914061504, "ts": 1716454218237382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237383, "dur": 0,
    "args": {
      "External id": 47251,
      "cbid": 251, "correlation": 47251
    }
  },
  {
    "ph": "f", "id": 47251, "pid": 76337, "tid": -914061504, "ts": 1716454218237383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237383, "dur": 0,
    "args": {
      "External id": 47252,
      "cbid": 251, "correlation": 47252
    }
  },
  {
    "ph": "f", "id": 47252, "pid": 76337, "tid": -914061504, "ts": 1716454218237383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237384, "dur": 0,
    "args": {
      "External id": 47253,
      "cbid": 251, "correlation": 47253
    }
  },
  {
    "ph": "f", "id": 47253, "pid": 76337, "tid": -914061504, "ts": 1716454218237384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237384, "dur": 0,
    "args": {
      "External id": 47254,
      "cbid": 251, "correlation": 47254
    }
  },
  {
    "ph": "f", "id": 47254, "pid": 76337, "tid": -914061504, "ts": 1716454218237384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237385, "dur": 0,
    "args": {
      "External id": 47255,
      "cbid": 251, "correlation": 47255
    }
  },
  {
    "ph": "f", "id": 47255, "pid": 76337, "tid": -914061504, "ts": 1716454218237385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237387, "dur": 0,
    "args": {
      "External id": 47256,
      "cbid": 251, "correlation": 47256
    }
  },
  {
    "ph": "f", "id": 47256, "pid": 76337, "tid": -914061504, "ts": 1716454218237387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218280915, "dur": 109,
    "args": {
      "External id": 47257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47257,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 47257, "pid": 5, "tid": 7, "ts": 1716454218280915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237389, "dur": 16,
    "args": {
      "External id": 47257,
      "cbid": 211, "correlation": 47257
    }
  },
  {
    "ph": "s", "id": 47257, "pid": 76337, "tid": -914061504, "ts": 1716454218237389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218281026, "dur": 59,
    "args": {
      "External id": 47263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47263, "pid": 5, "tid": 7, "ts": 1716454218281026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237427, "dur": 9,
    "args": {
      "External id": 47263,
      "cbid": 211, "correlation": 47263
    }
  },
  {
    "ph": "s", "id": 47263, "pid": 76337, "tid": -914061504, "ts": 1716454218237427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218281086, "dur": 50,
    "args": {
      "External id": 47271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47271,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47271, "pid": 5, "tid": 7, "ts": 1716454218281086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237460, "dur": 8,
    "args": {
      "External id": 47271,
      "cbid": 211, "correlation": 47271
    }
  },
  {
    "ph": "s", "id": 47271, "pid": 76337, "tid": -914061504, "ts": 1716454218237460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237533, "dur": 0,
    "args": {
      "External id": 47281,
      "cbid": 317, "correlation": 47281
    }
  },
  {
    "ph": "f", "id": 47281, "pid": 76337, "tid": -914061504, "ts": 1716454218237533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237534, "dur": 0,
    "args": {
      "External id": 47282,
      "cbid": 203, "correlation": 47282
    }
  },
  {
    "ph": "f", "id": 47282, "pid": 76337, "tid": -914061504, "ts": 1716454218237534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237535, "dur": 0,
    "args": {
      "External id": 47283,
      "cbid": 205, "correlation": 47283
    }
  },
  {
    "ph": "f", "id": 47283, "pid": 76337, "tid": -914061504, "ts": 1716454218237535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218281137, "dur": 40,
    "args": {
      "External id": 47287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47287,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47287, "pid": 5, "tid": 7, "ts": 1716454218281137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237552, "dur": 13,
    "args": {
      "External id": 47287,
      "cbid": 211, "correlation": 47287
    }
  },
  {
    "ph": "s", "id": 47287, "pid": 76337, "tid": -914061504, "ts": 1716454218237552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218281178, "dur": 13,
    "args": {
      "External id": 47289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47289,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47289, "pid": 5, "tid": 7, "ts": 1716454218281178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237567, "dur": 5,
    "args": {
      "External id": 47289,
      "cbid": 211, "correlation": 47289
    }
  },
  {
    "ph": "s", "id": 47289, "pid": 76337, "tid": -914061504, "ts": 1716454218237567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218281194, "dur": 1,
    "args": {
      "External id": 47291,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 47291,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9230769230769231
    }
  },
  {
    "ph": "f", "id": 47291, "pid": 5, "tid": 7, "ts": 1716454218281194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237585, "dur": 16,
    "args": {
      "External id": 47291,
      "cbid": 51, "correlation": 47291
    }
  },
  {
    "ph": "s", "id": 47291, "pid": 76337, "tid": -914061504, "ts": 1716454218237585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218281197, "dur": 347,
    "args": {
      "External id": 47292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47292,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47292, "pid": 5, "tid": 7, "ts": 1716454218281197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237603, "dur": 10,
    "args": {
      "External id": 47292,
      "cbid": 211, "correlation": 47292
    }
  },
  {
    "ph": "s", "id": 47292, "pid": 76337, "tid": -914061504, "ts": 1716454218237603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218281546, "dur": 13,
    "args": {
      "External id": 47294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47294,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47294, "pid": 5, "tid": 7, "ts": 1716454218281546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237621, "dur": 7,
    "args": {
      "External id": 47294,
      "cbid": 211, "correlation": 47294
    }
  },
  {
    "ph": "s", "id": 47294, "pid": 76337, "tid": -914061504, "ts": 1716454218237621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218281560, "dur": 14,
    "args": {
      "External id": 47300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47300,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47300, "pid": 5, "tid": 7, "ts": 1716454218281560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237650, "dur": 9,
    "args": {
      "External id": 47300,
      "cbid": 211, "correlation": 47300
    }
  },
  {
    "ph": "s", "id": 47300, "pid": 76337, "tid": -914061504, "ts": 1716454218237650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218281575, "dur": 18,
    "args": {
      "External id": 47320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47320,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 47320, "pid": 5, "tid": 7, "ts": 1716454218281575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237743, "dur": 15,
    "args": {
      "External id": 47320,
      "cbid": 211, "correlation": 47320
    }
  },
  {
    "ph": "s", "id": 47320, "pid": 76337, "tid": -914061504, "ts": 1716454218237743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218281595, "dur": 5,
    "args": {
      "External id": 47332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47332,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 47332, "pid": 5, "tid": 7, "ts": 1716454218281595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237770, "dur": 6,
    "args": {
      "External id": 47332,
      "cbid": 211, "correlation": 47332
    }
  },
  {
    "ph": "s", "id": 47332, "pid": 76337, "tid": -914061504, "ts": 1716454218237770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218281601, "dur": 18,
    "args": {
      "External id": 47335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47335,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47335, "pid": 5, "tid": 7, "ts": 1716454218281601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237788, "dur": 6,
    "args": {
      "External id": 47335,
      "cbid": 211, "correlation": 47335
    }
  },
  {
    "ph": "s", "id": 47335, "pid": 76337, "tid": -914061504, "ts": 1716454218237788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218281620, "dur": 11,
    "args": {
      "External id": 47344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47344,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47344, "pid": 5, "tid": 7, "ts": 1716454218281620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237828, "dur": 9,
    "args": {
      "External id": 47344,
      "cbid": 211, "correlation": 47344
    }
  },
  {
    "ph": "s", "id": 47344, "pid": 76337, "tid": -914061504, "ts": 1716454218237828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237883, "dur": 0,
    "args": {
      "External id": 47354,
      "cbid": 317, "correlation": 47354
    }
  },
  {
    "ph": "f", "id": 47354, "pid": 76337, "tid": -914061504, "ts": 1716454218237883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237883, "dur": 0,
    "args": {
      "External id": 47355,
      "cbid": 203, "correlation": 47355
    }
  },
  {
    "ph": "f", "id": 47355, "pid": 76337, "tid": -914061504, "ts": 1716454218237883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237884, "dur": 0,
    "args": {
      "External id": 47356,
      "cbid": 205, "correlation": 47356
    }
  },
  {
    "ph": "f", "id": 47356, "pid": 76337, "tid": -914061504, "ts": 1716454218237884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218281632, "dur": 11,
    "args": {
      "External id": 47360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47360,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47360, "pid": 5, "tid": 7, "ts": 1716454218281632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237899, "dur": 11,
    "args": {
      "External id": 47360,
      "cbid": 211, "correlation": 47360
    }
  },
  {
    "ph": "s", "id": 47360, "pid": 76337, "tid": -914061504, "ts": 1716454218237899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218281644, "dur": 23,
    "args": {
      "External id": 47362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47362,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47362, "pid": 5, "tid": 7, "ts": 1716454218281644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237913, "dur": 6,
    "args": {
      "External id": 47362,
      "cbid": 211, "correlation": 47362
    }
  },
  {
    "ph": "s", "id": 47362, "pid": 76337, "tid": -914061504, "ts": 1716454218237913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218281669, "dur": 3,
    "args": {
      "External id": 47364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 47364, "pid": 5, "tid": 7, "ts": 1716454218281669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237925, "dur": 6,
    "args": {
      "External id": 47364,
      "cbid": 211, "correlation": 47364
    }
  },
  {
    "ph": "s", "id": 47364, "pid": 76337, "tid": -914061504, "ts": 1716454218237925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237934, "dur": 0,
    "args": {
      "External id": 47365,
      "cbid": 51, "correlation": 47365
    }
  },
  {
    "ph": "s", "id": 47365, "pid": 76337, "tid": -914061504, "ts": 1716454218237934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218281674, "dur": 345,
    "args": {
      "External id": 47366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47366,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47366, "pid": 5, "tid": 7, "ts": 1716454218281674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237935, "dur": 7,
    "args": {
      "External id": 47366,
      "cbid": 211, "correlation": 47366
    }
  },
  {
    "ph": "s", "id": 47366, "pid": 76337, "tid": -914061504, "ts": 1716454218237935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218282020, "dur": 19,
    "args": {
      "External id": 47367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47367,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47367, "pid": 5, "tid": 7, "ts": 1716454218282020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237945, "dur": 5,
    "args": {
      "External id": 47367,
      "cbid": 211, "correlation": 47367
    }
  },
  {
    "ph": "s", "id": 47367, "pid": 76337, "tid": -914061504, "ts": 1716454218237945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218282041, "dur": 31,
    "args": {
      "External id": 47373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47373, "pid": 5, "tid": 7, "ts": 1716454218282041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218237982, "dur": 9,
    "args": {
      "External id": 47373,
      "cbid": 211, "correlation": 47373
    }
  },
  {
    "ph": "s", "id": 47373, "pid": 76337, "tid": -914061504, "ts": 1716454218237982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218282073, "dur": 3,
    "args": {
      "External id": 47381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47381,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 47381, "pid": 5, "tid": 7, "ts": 1716454218282073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238026, "dur": 12,
    "args": {
      "External id": 47381,
      "cbid": 211, "correlation": 47381
    }
  },
  {
    "ph": "s", "id": 47381, "pid": 76337, "tid": -914061504, "ts": 1716454218238026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238096, "dur": 1,
    "args": {
      "External id": 47397,
      "cbid": 251, "correlation": 47397
    }
  },
  {
    "ph": "f", "id": 47397, "pid": 76337, "tid": -914061504, "ts": 1716454218238096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238102, "dur": 0,
    "args": {
      "External id": 47399,
      "cbid": 251, "correlation": 47399
    }
  },
  {
    "ph": "f", "id": 47399, "pid": 76337, "tid": -914061504, "ts": 1716454218238102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218282078, "dur": 12,
    "args": {
      "External id": 47400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47400,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 47400, "pid": 5, "tid": 7, "ts": 1716454218282078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238103, "dur": 12,
    "args": {
      "External id": 47400,
      "cbid": 211, "correlation": 47400
    }
  },
  {
    "ph": "s", "id": 47400, "pid": 76337, "tid": -914061504, "ts": 1716454218238103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218282091, "dur": 5,
    "args": {
      "External id": 47402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47402,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 47402, "pid": 5, "tid": 7, "ts": 1716454218282091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238117, "dur": 5,
    "args": {
      "External id": 47402,
      "cbid": 211, "correlation": 47402
    }
  },
  {
    "ph": "s", "id": 47402, "pid": 76337, "tid": -914061504, "ts": 1716454218238117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218282097, "dur": 29,
    "args": {
      "External id": 47412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47412,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47412, "pid": 5, "tid": 7, "ts": 1716454218282097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238175, "dur": 13,
    "args": {
      "External id": 47412,
      "cbid": 211, "correlation": 47412
    }
  },
  {
    "ph": "s", "id": 47412, "pid": 76337, "tid": -914061504, "ts": 1716454218238175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218282128, "dur": 30,
    "args": {
      "External id": 47432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47432,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 47432, "pid": 5, "tid": 7, "ts": 1716454218282128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238244, "dur": 11,
    "args": {
      "External id": 47432,
      "cbid": 211, "correlation": 47432
    }
  },
  {
    "ph": "s", "id": 47432, "pid": 76337, "tid": -914061504, "ts": 1716454218238244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218282159, "dur": 4,
    "args": {
      "External id": 47444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47444,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 47444, "pid": 5, "tid": 7, "ts": 1716454218282159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238265, "dur": 6,
    "args": {
      "External id": 47444,
      "cbid": 211, "correlation": 47444
    }
  },
  {
    "ph": "s", "id": 47444, "pid": 76337, "tid": -914061504, "ts": 1716454218238265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218282164, "dur": 30,
    "args": {
      "External id": 47447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47447, "pid": 5, "tid": 7, "ts": 1716454218282164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238283, "dur": 7,
    "args": {
      "External id": 47447,
      "cbid": 211, "correlation": 47447
    }
  },
  {
    "ph": "s", "id": 47447, "pid": 76337, "tid": -914061504, "ts": 1716454218238283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218282195, "dur": 20,
    "args": {
      "External id": 47456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47456,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47456, "pid": 5, "tid": 7, "ts": 1716454218282195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238327, "dur": 10,
    "args": {
      "External id": 47456,
      "cbid": 211, "correlation": 47456
    }
  },
  {
    "ph": "s", "id": 47456, "pid": 76337, "tid": -914061504, "ts": 1716454218238327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238392, "dur": 0,
    "args": {
      "External id": 47466,
      "cbid": 317, "correlation": 47466
    }
  },
  {
    "ph": "f", "id": 47466, "pid": 76337, "tid": -914061504, "ts": 1716454218238392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238393, "dur": 0,
    "args": {
      "External id": 47467,
      "cbid": 203, "correlation": 47467
    }
  },
  {
    "ph": "f", "id": 47467, "pid": 76337, "tid": -914061504, "ts": 1716454218238393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238394, "dur": 0,
    "args": {
      "External id": 47468,
      "cbid": 205, "correlation": 47468
    }
  },
  {
    "ph": "f", "id": 47468, "pid": 76337, "tid": -914061504, "ts": 1716454218238394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218282216, "dur": 22,
    "args": {
      "External id": 47472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47472,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47472, "pid": 5, "tid": 7, "ts": 1716454218282216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238412, "dur": 12,
    "args": {
      "External id": 47472,
      "cbid": 211, "correlation": 47472
    }
  },
  {
    "ph": "s", "id": 47472, "pid": 76337, "tid": -914061504, "ts": 1716454218238412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218282240, "dur": 43,
    "args": {
      "External id": 47474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47474,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47474, "pid": 5, "tid": 7, "ts": 1716454218282240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238426, "dur": 5,
    "args": {
      "External id": 47474,
      "cbid": 211, "correlation": 47474
    }
  },
  {
    "ph": "s", "id": 47474, "pid": 76337, "tid": -914061504, "ts": 1716454218238426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218282284, "dur": 632,
    "args": {
      "External id": 47476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47476,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47476, "pid": 5, "tid": 7, "ts": 1716454218282284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238440, "dur": 9,
    "args": {
      "External id": 47476,
      "cbid": 211, "correlation": 47476
    }
  },
  {
    "ph": "s", "id": 47476, "pid": 76337, "tid": -914061504, "ts": 1716454218238440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218282917, "dur": 21,
    "args": {
      "External id": 47478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47478, "pid": 5, "tid": 7, "ts": 1716454218282917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238453, "dur": 5,
    "args": {
      "External id": 47478,
      "cbid": 211, "correlation": 47478
    }
  },
  {
    "ph": "s", "id": 47478, "pid": 76337, "tid": -914061504, "ts": 1716454218238453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218282939, "dur": 32,
    "args": {
      "External id": 47484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47484, "pid": 5, "tid": 7, "ts": 1716454218282939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238480, "dur": 9,
    "args": {
      "External id": 47484,
      "cbid": 211, "correlation": 47484
    }
  },
  {
    "ph": "s", "id": 47484, "pid": 76337, "tid": -914061504, "ts": 1716454218238480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238539, "dur": 0,
    "args": {
      "External id": 47494,
      "cbid": 317, "correlation": 47494
    }
  },
  {
    "ph": "f", "id": 47494, "pid": 76337, "tid": -914061504, "ts": 1716454218238539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238540, "dur": 0,
    "args": {
      "External id": 47495,
      "cbid": 203, "correlation": 47495
    }
  },
  {
    "ph": "f", "id": 47495, "pid": 76337, "tid": -914061504, "ts": 1716454218238540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238541, "dur": 0,
    "args": {
      "External id": 47496,
      "cbid": 205, "correlation": 47496
    }
  },
  {
    "ph": "f", "id": 47496, "pid": 76337, "tid": -914061504, "ts": 1716454218238541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238563, "dur": 1,
    "args": {
      "External id": 47500,
      "cbid": 251, "correlation": 47500
    }
  },
  {
    "ph": "f", "id": 47500, "pid": 76337, "tid": -914061504, "ts": 1716454218238563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238565, "dur": 0,
    "args": {
      "External id": 47501,
      "cbid": 251, "correlation": 47501
    }
  },
  {
    "ph": "f", "id": 47501, "pid": 76337, "tid": -914061504, "ts": 1716454218238565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238566, "dur": 0,
    "args": {
      "External id": 47502,
      "cbid": 251, "correlation": 47502
    }
  },
  {
    "ph": "f", "id": 47502, "pid": 76337, "tid": -914061504, "ts": 1716454218238566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238567, "dur": 0,
    "args": {
      "External id": 47503,
      "cbid": 251, "correlation": 47503
    }
  },
  {
    "ph": "f", "id": 47503, "pid": 76337, "tid": -914061504, "ts": 1716454218238567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238567, "dur": 0,
    "args": {
      "External id": 47504,
      "cbid": 251, "correlation": 47504
    }
  },
  {
    "ph": "f", "id": 47504, "pid": 76337, "tid": -914061504, "ts": 1716454218238567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238568, "dur": 0,
    "args": {
      "External id": 47505,
      "cbid": 251, "correlation": 47505
    }
  },
  {
    "ph": "f", "id": 47505, "pid": 76337, "tid": -914061504, "ts": 1716454218238568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238569, "dur": 0,
    "args": {
      "External id": 47506,
      "cbid": 251, "correlation": 47506
    }
  },
  {
    "ph": "f", "id": 47506, "pid": 76337, "tid": -914061504, "ts": 1716454218238569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238570, "dur": 0,
    "args": {
      "External id": 47507,
      "cbid": 251, "correlation": 47507
    }
  },
  {
    "ph": "f", "id": 47507, "pid": 76337, "tid": -914061504, "ts": 1716454218238570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238571, "dur": 0,
    "args": {
      "External id": 47508,
      "cbid": 251, "correlation": 47508
    }
  },
  {
    "ph": "f", "id": 47508, "pid": 76337, "tid": -914061504, "ts": 1716454218238571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218282972, "dur": 50,
    "args": {
      "External id": 47509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47509,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 47509, "pid": 5, "tid": 7, "ts": 1716454218282972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238573, "dur": 12,
    "args": {
      "External id": 47509,
      "cbid": 211, "correlation": 47509
    }
  },
  {
    "ph": "s", "id": 47509, "pid": 76337, "tid": -914061504, "ts": 1716454218238573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218283023, "dur": 31,
    "args": {
      "External id": 47515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47515,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47515, "pid": 5, "tid": 7, "ts": 1716454218283023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238606, "dur": 8,
    "args": {
      "External id": 47515,
      "cbid": 211, "correlation": 47515
    }
  },
  {
    "ph": "s", "id": 47515, "pid": 76337, "tid": -914061504, "ts": 1716454218238606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218283056, "dur": 27,
    "args": {
      "External id": 47523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47523,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47523, "pid": 5, "tid": 7, "ts": 1716454218283056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238638, "dur": 8,
    "args": {
      "External id": 47523,
      "cbid": 211, "correlation": 47523
    }
  },
  {
    "ph": "s", "id": 47523, "pid": 76337, "tid": -914061504, "ts": 1716454218238638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218283085, "dur": 19,
    "args": {
      "External id": 47531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47531,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47531, "pid": 5, "tid": 7, "ts": 1716454218283085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238668, "dur": 8,
    "args": {
      "External id": 47531,
      "cbid": 211, "correlation": 47531
    }
  },
  {
    "ph": "s", "id": 47531, "pid": 76337, "tid": -914061504, "ts": 1716454218238668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218283105, "dur": 29,
    "args": {
      "External id": 47551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47551,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 47551, "pid": 5, "tid": 7, "ts": 1716454218283105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238749, "dur": 13,
    "args": {
      "External id": 47551,
      "cbid": 211, "correlation": 47551
    }
  },
  {
    "ph": "s", "id": 47551, "pid": 76337, "tid": -914061504, "ts": 1716454218238749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218283136, "dur": 4,
    "args": {
      "External id": 47563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47563,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 47563, "pid": 5, "tid": 7, "ts": 1716454218283136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238772, "dur": 6,
    "args": {
      "External id": 47563,
      "cbid": 211, "correlation": 47563
    }
  },
  {
    "ph": "s", "id": 47563, "pid": 76337, "tid": -914061504, "ts": 1716454218238772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218283141, "dur": 29,
    "args": {
      "External id": 47566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47566, "pid": 5, "tid": 7, "ts": 1716454218283141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238790, "dur": 6,
    "args": {
      "External id": 47566,
      "cbid": 211, "correlation": 47566
    }
  },
  {
    "ph": "s", "id": 47566, "pid": 76337, "tid": -914061504, "ts": 1716454218238790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238848, "dur": 0,
    "args": {
      "External id": 47577,
      "cbid": 317, "correlation": 47577
    }
  },
  {
    "ph": "f", "id": 47577, "pid": 76337, "tid": -914061504, "ts": 1716454218238848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238849, "dur": 0,
    "args": {
      "External id": 47578,
      "cbid": 203, "correlation": 47578
    }
  },
  {
    "ph": "f", "id": 47578, "pid": 76337, "tid": -914061504, "ts": 1716454218238849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238849, "dur": 0,
    "args": {
      "External id": 47579,
      "cbid": 205, "correlation": 47579
    }
  },
  {
    "ph": "f", "id": 47579, "pid": 76337, "tid": -914061504, "ts": 1716454218238849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218283172, "dur": 22,
    "args": {
      "External id": 47583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47583,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47583, "pid": 5, "tid": 7, "ts": 1716454218283172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238865, "dur": 12,
    "args": {
      "External id": 47583,
      "cbid": 211, "correlation": 47583
    }
  },
  {
    "ph": "s", "id": 47583, "pid": 76337, "tid": -914061504, "ts": 1716454218238865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218283195, "dur": 115,
    "args": {
      "External id": 47585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47585,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47585, "pid": 5, "tid": 7, "ts": 1716454218283195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238885, "dur": 8,
    "args": {
      "External id": 47585,
      "cbid": 211, "correlation": 47585
    }
  },
  {
    "ph": "s", "id": 47585, "pid": 76337, "tid": -914061504, "ts": 1716454218238885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218283311, "dur": 22,
    "args": {
      "External id": 47587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47587, "pid": 5, "tid": 7, "ts": 1716454218283311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238896, "dur": 5,
    "args": {
      "External id": 47587,
      "cbid": 211, "correlation": 47587
    }
  },
  {
    "ph": "s", "id": 47587, "pid": 76337, "tid": -914061504, "ts": 1716454218238896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218283334, "dur": 31,
    "args": {
      "External id": 47593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47593,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47593, "pid": 5, "tid": 7, "ts": 1716454218283334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218238928, "dur": 9,
    "args": {
      "External id": 47593,
      "cbid": 211, "correlation": 47593
    }
  },
  {
    "ph": "s", "id": 47593, "pid": 76337, "tid": -914061504, "ts": 1716454218238928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218283366, "dur": 163,
    "args": {
      "External id": 47602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47602,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47602, "pid": 5, "tid": 7, "ts": 1716454218283366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239020, "dur": 15,
    "args": {
      "External id": 47602,
      "cbid": 211, "correlation": 47602
    }
  },
  {
    "ph": "s", "id": 47602, "pid": 76337, "tid": -914061504, "ts": 1716454218239020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218283531, "dur": 61,
    "args": {
      "External id": 47624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47624,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47624, "pid": 5, "tid": 7, "ts": 1716454218283531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239078, "dur": 10,
    "args": {
      "External id": 47624,
      "cbid": 211, "correlation": 47624
    }
  },
  {
    "ph": "s", "id": 47624, "pid": 76337, "tid": -914061504, "ts": 1716454218239078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239166, "dur": 1,
    "args": {
      "External id": 47635,
      "cbid": 251, "correlation": 47635
    }
  },
  {
    "ph": "f", "id": 47635, "pid": 76337, "tid": -914061504, "ts": 1716454218239166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218283594, "dur": 149,
    "args": {
      "External id": 47636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47636,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47636, "pid": 5, "tid": 7, "ts": 1716454218283594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239172, "dur": 14,
    "args": {
      "External id": 47636,
      "cbid": 211, "correlation": 47636
    }
  },
  {
    "ph": "s", "id": 47636, "pid": 76337, "tid": -914061504, "ts": 1716454218239172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239243, "dur": 1,
    "args": {
      "External id": 47647,
      "cbid": 251, "correlation": 47647
    }
  },
  {
    "ph": "f", "id": 47647, "pid": 76337, "tid": -914061504, "ts": 1716454218239243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218283744, "dur": 142,
    "args": {
      "External id": 47648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47648,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47648, "pid": 5, "tid": 7, "ts": 1716454218283744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239247, "dur": 11,
    "args": {
      "External id": 47648,
      "cbid": 211, "correlation": 47648
    }
  },
  {
    "ph": "s", "id": 47648, "pid": 76337, "tid": -914061504, "ts": 1716454218239247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239311, "dur": 1,
    "args": {
      "External id": 47659,
      "cbid": 251, "correlation": 47659
    }
  },
  {
    "ph": "f", "id": 47659, "pid": 76337, "tid": -914061504, "ts": 1716454218239311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218283888, "dur": 142,
    "args": {
      "External id": 47660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47660,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47660, "pid": 5, "tid": 7, "ts": 1716454218283888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239315, "dur": 11,
    "args": {
      "External id": 47660,
      "cbid": 211, "correlation": 47660
    }
  },
  {
    "ph": "s", "id": 47660, "pid": 76337, "tid": -914061504, "ts": 1716454218239315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218284031, "dur": 1853,
    "args": {
      "External id": 47681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47681,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 47681, "pid": 5, "tid": 7, "ts": 1716454218284031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239401, "dur": 14,
    "args": {
      "External id": 47681,
      "cbid": 211, "correlation": 47681
    }
  },
  {
    "ph": "s", "id": 47681, "pid": 76337, "tid": -914061504, "ts": 1716454218239401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239503, "dur": 1,
    "args": {
      "External id": 47699,
      "cbid": 251, "correlation": 47699
    }
  },
  {
    "ph": "f", "id": 47699, "pid": 76337, "tid": -914061504, "ts": 1716454218239503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218285885, "dur": 142,
    "args": {
      "External id": 47701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47701,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 47701, "pid": 5, "tid": 7, "ts": 1716454218285885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239509, "dur": 13,
    "args": {
      "External id": 47701,
      "cbid": 211, "correlation": 47701
    }
  },
  {
    "ph": "s", "id": 47701, "pid": 76337, "tid": -914061504, "ts": 1716454218239509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218286029, "dur": 35,
    "args": {
      "External id": 47709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47709,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47709, "pid": 5, "tid": 7, "ts": 1716454218286029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239579, "dur": 12,
    "args": {
      "External id": 47709,
      "cbid": 211, "correlation": 47709
    }
  },
  {
    "ph": "s", "id": 47709, "pid": 76337, "tid": -914061504, "ts": 1716454218239579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218286065, "dur": 51,
    "args": {
      "External id": 47717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47717,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47717, "pid": 5, "tid": 7, "ts": 1716454218286065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239618, "dur": 8,
    "args": {
      "External id": 47717,
      "cbid": 211, "correlation": 47717
    }
  },
  {
    "ph": "s", "id": 47717, "pid": 76337, "tid": -914061504, "ts": 1716454218239618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218286118, "dur": 29,
    "args": {
      "External id": 47728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47728,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47728, "pid": 5, "tid": 7, "ts": 1716454218286118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239690, "dur": 13,
    "args": {
      "External id": 47728,
      "cbid": 211, "correlation": 47728
    }
  },
  {
    "ph": "s", "id": 47728, "pid": 76337, "tid": -914061504, "ts": 1716454218239690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218286149, "dur": 33,
    "args": {
      "External id": 47750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47750,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47750, "pid": 5, "tid": 7, "ts": 1716454218286149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239725, "dur": 7,
    "args": {
      "External id": 47750,
      "cbid": 211, "correlation": 47750
    }
  },
  {
    "ph": "s", "id": 47750, "pid": 76337, "tid": -914061504, "ts": 1716454218239725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239810, "dur": 1,
    "args": {
      "External id": 47761,
      "cbid": 251, "correlation": 47761
    }
  },
  {
    "ph": "f", "id": 47761, "pid": 76337, "tid": -914061504, "ts": 1716454218239810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218286183, "dur": 75,
    "args": {
      "External id": 47762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47762,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47762, "pid": 5, "tid": 7, "ts": 1716454218286183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239815, "dur": 13,
    "args": {
      "External id": 47762,
      "cbid": 211, "correlation": 47762
    }
  },
  {
    "ph": "s", "id": 47762, "pid": 76337, "tid": -914061504, "ts": 1716454218239815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239883, "dur": 1,
    "args": {
      "External id": 47773,
      "cbid": 251, "correlation": 47773
    }
  },
  {
    "ph": "f", "id": 47773, "pid": 76337, "tid": -914061504, "ts": 1716454218239883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239887, "dur": 0,
    "args": {
      "External id": 47774,
      "cbid": 251, "correlation": 47774
    }
  },
  {
    "ph": "f", "id": 47774, "pid": 76337, "tid": -914061504, "ts": 1716454218239887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218286260, "dur": 11,
    "args": {
      "External id": 47775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47775,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 47775, "pid": 5, "tid": 7, "ts": 1716454218286260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239889, "dur": 12,
    "args": {
      "External id": 47775,
      "cbid": 211, "correlation": 47775
    }
  },
  {
    "ph": "s", "id": 47775, "pid": 76337, "tid": -914061504, "ts": 1716454218239889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218286272, "dur": 5,
    "args": {
      "External id": 47777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47777,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 47777, "pid": 5, "tid": 7, "ts": 1716454218286272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239904, "dur": 8,
    "args": {
      "External id": 47777,
      "cbid": 211, "correlation": 47777
    }
  },
  {
    "ph": "s", "id": 47777, "pid": 76337, "tid": -914061504, "ts": 1716454218239904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239963, "dur": 1,
    "args": {
      "External id": 47788,
      "cbid": 251, "correlation": 47788
    }
  },
  {
    "ph": "f", "id": 47788, "pid": 76337, "tid": -914061504, "ts": 1716454218239963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239967, "dur": 0,
    "args": {
      "External id": 47789,
      "cbid": 251, "correlation": 47789
    }
  },
  {
    "ph": "f", "id": 47789, "pid": 76337, "tid": -914061504, "ts": 1716454218239967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218286278, "dur": 7,
    "args": {
      "External id": 47790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47790,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 47790, "pid": 5, "tid": 7, "ts": 1716454218286278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239969, "dur": 19,
    "args": {
      "External id": 47790,
      "cbid": 211, "correlation": 47790
    }
  },
  {
    "ph": "s", "id": 47790, "pid": 76337, "tid": -914061504, "ts": 1716454218239969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218286286, "dur": 4,
    "args": {
      "External id": 47792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47792,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 47792, "pid": 5, "tid": 7, "ts": 1716454218286286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218239989, "dur": 5,
    "args": {
      "External id": 47792,
      "cbid": 211, "correlation": 47792
    }
  },
  {
    "ph": "s", "id": 47792, "pid": 76337, "tid": -914061504, "ts": 1716454218239989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218286291, "dur": 88,
    "args": {
      "External id": 47813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47813,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 47813, "pid": 5, "tid": 7, "ts": 1716454218286291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240063, "dur": 13,
    "args": {
      "External id": 47813,
      "cbid": 211, "correlation": 47813
    }
  },
  {
    "ph": "s", "id": 47813, "pid": 76337, "tid": -914061504, "ts": 1716454218240063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240164, "dur": 1,
    "args": {
      "External id": 47831,
      "cbid": 251, "correlation": 47831
    }
  },
  {
    "ph": "f", "id": 47831, "pid": 76337, "tid": -914061504, "ts": 1716454218240164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218286380, "dur": 93,
    "args": {
      "External id": 47833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47833,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47833, "pid": 5, "tid": 7, "ts": 1716454218286380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240170, "dur": 14,
    "args": {
      "External id": 47833,
      "cbid": 211, "correlation": 47833
    }
  },
  {
    "ph": "s", "id": 47833, "pid": 76337, "tid": -914061504, "ts": 1716454218240170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218286475, "dur": 19,
    "args": {
      "External id": 47841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47841,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47841, "pid": 5, "tid": 7, "ts": 1716454218286475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240239, "dur": 12,
    "args": {
      "External id": 47841,
      "cbid": 211, "correlation": 47841
    }
  },
  {
    "ph": "s", "id": 47841, "pid": 76337, "tid": -914061504, "ts": 1716454218240239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218286495, "dur": 36,
    "args": {
      "External id": 47849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47849, "pid": 5, "tid": 7, "ts": 1716454218286495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240282, "dur": 9,
    "args": {
      "External id": 47849,
      "cbid": 211, "correlation": 47849
    }
  },
  {
    "ph": "s", "id": 47849, "pid": 76337, "tid": -914061504, "ts": 1716454218240282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218286533, "dur": 33,
    "args": {
      "External id": 47871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47871,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47871, "pid": 5, "tid": 7, "ts": 1716454218286533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240333, "dur": 10,
    "args": {
      "External id": 47871,
      "cbid": 211, "correlation": 47871
    }
  },
  {
    "ph": "s", "id": 47871, "pid": 76337, "tid": -914061504, "ts": 1716454218240333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240422, "dur": 1,
    "args": {
      "External id": 47887,
      "cbid": 251, "correlation": 47887
    }
  },
  {
    "ph": "f", "id": 47887, "pid": 76337, "tid": -914061504, "ts": 1716454218240422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240426, "dur": 0,
    "args": {
      "External id": 47889,
      "cbid": 251, "correlation": 47889
    }
  },
  {
    "ph": "f", "id": 47889, "pid": 76337, "tid": -914061504, "ts": 1716454218240426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218286568, "dur": 522,
    "args": {
      "External id": 47890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47890,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 47890, "pid": 5, "tid": 7, "ts": 1716454218286568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240430, "dur": 14,
    "args": {
      "External id": 47890,
      "cbid": 211, "correlation": 47890
    }
  },
  {
    "ph": "s", "id": 47890, "pid": 76337, "tid": -914061504, "ts": 1716454218240430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218287092, "dur": 122,
    "args": {
      "External id": 47898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47898, "pid": 5, "tid": 7, "ts": 1716454218287092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240496, "dur": 15,
    "args": {
      "External id": 47898,
      "cbid": 211, "correlation": 47898
    }
  },
  {
    "ph": "s", "id": 47898, "pid": 76337, "tid": -914061504, "ts": 1716454218240496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218287215, "dur": 126,
    "args": {
      "External id": 47906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47906,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47906, "pid": 5, "tid": 7, "ts": 1716454218287215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240529, "dur": 9,
    "args": {
      "External id": 47906,
      "cbid": 211, "correlation": 47906
    }
  },
  {
    "ph": "s", "id": 47906, "pid": 76337, "tid": -914061504, "ts": 1716454218240529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240610, "dur": 1,
    "args": {
      "External id": 47922,
      "cbid": 251, "correlation": 47922
    }
  },
  {
    "ph": "f", "id": 47922, "pid": 76337, "tid": -914061504, "ts": 1716454218240610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218287342, "dur": 291,
    "args": {
      "External id": 47924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47924,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47924, "pid": 5, "tid": 7, "ts": 1716454218287342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240616, "dur": 12,
    "args": {
      "External id": 47924,
      "cbid": 211, "correlation": 47924
    }
  },
  {
    "ph": "s", "id": 47924, "pid": 76337, "tid": -914061504, "ts": 1716454218240616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218287634, "dur": 27,
    "args": {
      "External id": 47932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47932,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47932, "pid": 5, "tid": 7, "ts": 1716454218287634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240659, "dur": 9,
    "args": {
      "External id": 47932,
      "cbid": 211, "correlation": 47932
    }
  },
  {
    "ph": "s", "id": 47932, "pid": 76337, "tid": -914061504, "ts": 1716454218240659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218287662, "dur": 78,
    "args": {
      "External id": 47943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47943, "pid": 5, "tid": 7, "ts": 1716454218287662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240726, "dur": 12,
    "args": {
      "External id": 47943,
      "cbid": 211, "correlation": 47943
    }
  },
  {
    "ph": "s", "id": 47943, "pid": 76337, "tid": -914061504, "ts": 1716454218240726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240790, "dur": 0,
    "args": {
      "External id": 47955,
      "cbid": 317, "correlation": 47955
    }
  },
  {
    "ph": "f", "id": 47955, "pid": 76337, "tid": -914061504, "ts": 1716454218240790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240790, "dur": 0,
    "args": {
      "External id": 47956,
      "cbid": 203, "correlation": 47956
    }
  },
  {
    "ph": "f", "id": 47956, "pid": 76337, "tid": -914061504, "ts": 1716454218240790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240791, "dur": 0,
    "args": {
      "External id": 47957,
      "cbid": 205, "correlation": 47957
    }
  },
  {
    "ph": "f", "id": 47957, "pid": 76337, "tid": -914061504, "ts": 1716454218240791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218287742, "dur": 22,
    "args": {
      "External id": 47961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47961,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47961, "pid": 5, "tid": 7, "ts": 1716454218287742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240806, "dur": 12,
    "args": {
      "External id": 47961,
      "cbid": 211, "correlation": 47961
    }
  },
  {
    "ph": "s", "id": 47961, "pid": 76337, "tid": -914061504, "ts": 1716454218240806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218287765, "dur": 114,
    "args": {
      "External id": 47963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47963,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 47963, "pid": 5, "tid": 7, "ts": 1716454218287765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240826, "dur": 6,
    "args": {
      "External id": 47963,
      "cbid": 211, "correlation": 47963
    }
  },
  {
    "ph": "s", "id": 47963, "pid": 76337, "tid": -914061504, "ts": 1716454218240826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218287881, "dur": 22,
    "args": {
      "External id": 47965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47965,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47965, "pid": 5, "tid": 7, "ts": 1716454218287881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240836, "dur": 8,
    "args": {
      "External id": 47965,
      "cbid": 211, "correlation": 47965
    }
  },
  {
    "ph": "s", "id": 47965, "pid": 76337, "tid": -914061504, "ts": 1716454218240836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218287904, "dur": 31,
    "args": {
      "External id": 47971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47971, "pid": 5, "tid": 7, "ts": 1716454218287904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240867, "dur": 9,
    "args": {
      "External id": 47971,
      "cbid": 211, "correlation": 47971
    }
  },
  {
    "ph": "s", "id": 47971, "pid": 76337, "tid": -914061504, "ts": 1716454218240867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218287937, "dur": 27,
    "args": {
      "External id": 47979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47979,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 47979, "pid": 5, "tid": 7, "ts": 1716454218287937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240899, "dur": 8,
    "args": {
      "External id": 47979,
      "cbid": 211, "correlation": 47979
    }
  },
  {
    "ph": "s", "id": 47979, "pid": 76337, "tid": -914061504, "ts": 1716454218240899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218287965, "dur": 29,
    "args": {
      "External id": 47999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 47999,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 47999, "pid": 5, "tid": 7, "ts": 1716454218287965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218240971, "dur": 19,
    "args": {
      "External id": 47999,
      "cbid": 211, "correlation": 47999
    }
  },
  {
    "ph": "s", "id": 47999, "pid": 76337, "tid": -914061504, "ts": 1716454218240971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218287996, "dur": 5,
    "args": {
      "External id": 48011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48011,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 48011, "pid": 5, "tid": 7, "ts": 1716454218287996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241001, "dur": 6,
    "args": {
      "External id": 48011,
      "cbid": 211, "correlation": 48011
    }
  },
  {
    "ph": "s", "id": 48011, "pid": 76337, "tid": -914061504, "ts": 1716454218241001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218288001, "dur": 30,
    "args": {
      "External id": 48014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48014,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48014, "pid": 5, "tid": 7, "ts": 1716454218288001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241018, "dur": 7,
    "args": {
      "External id": 48014,
      "cbid": 211, "correlation": 48014
    }
  },
  {
    "ph": "s", "id": 48014, "pid": 76337, "tid": -914061504, "ts": 1716454218241018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218288033, "dur": 22,
    "args": {
      "External id": 48023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48023,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48023, "pid": 5, "tid": 7, "ts": 1716454218288033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241058, "dur": 10,
    "args": {
      "External id": 48023,
      "cbid": 211, "correlation": 48023
    }
  },
  {
    "ph": "s", "id": 48023, "pid": 76337, "tid": -914061504, "ts": 1716454218241058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241110, "dur": 0,
    "args": {
      "External id": 48033,
      "cbid": 317, "correlation": 48033
    }
  },
  {
    "ph": "f", "id": 48033, "pid": 76337, "tid": -914061504, "ts": 1716454218241110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241111, "dur": 0,
    "args": {
      "External id": 48034,
      "cbid": 203, "correlation": 48034
    }
  },
  {
    "ph": "f", "id": 48034, "pid": 76337, "tid": -914061504, "ts": 1716454218241111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241112, "dur": 0,
    "args": {
      "External id": 48035,
      "cbid": 205, "correlation": 48035
    }
  },
  {
    "ph": "f", "id": 48035, "pid": 76337, "tid": -914061504, "ts": 1716454218241112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218288056, "dur": 21,
    "args": {
      "External id": 48039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48039,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48039, "pid": 5, "tid": 7, "ts": 1716454218288056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241125, "dur": 14,
    "args": {
      "External id": 48039,
      "cbid": 211, "correlation": 48039
    }
  },
  {
    "ph": "s", "id": 48039, "pid": 76337, "tid": -914061504, "ts": 1716454218241125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218288078, "dur": 42,
    "args": {
      "External id": 48041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48041,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48041, "pid": 5, "tid": 7, "ts": 1716454218288078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241143, "dur": 5,
    "args": {
      "External id": 48041,
      "cbid": 211, "correlation": 48041
    }
  },
  {
    "ph": "s", "id": 48041, "pid": 76337, "tid": -914061504, "ts": 1716454218241143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218288121, "dur": 633,
    "args": {
      "External id": 48043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48043,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48043, "pid": 5, "tid": 7, "ts": 1716454218288121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241155, "dur": 6,
    "args": {
      "External id": 48043,
      "cbid": 211, "correlation": 48043
    }
  },
  {
    "ph": "s", "id": 48043, "pid": 76337, "tid": -914061504, "ts": 1716454218241155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218288755, "dur": 20,
    "args": {
      "External id": 48045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48045,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48045, "pid": 5, "tid": 7, "ts": 1716454218288755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241164, "dur": 5,
    "args": {
      "External id": 48045,
      "cbid": 211, "correlation": 48045
    }
  },
  {
    "ph": "s", "id": 48045, "pid": 76337, "tid": -914061504, "ts": 1716454218241164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218288776, "dur": 32,
    "args": {
      "External id": 48051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48051, "pid": 5, "tid": 7, "ts": 1716454218288776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241192, "dur": 9,
    "args": {
      "External id": 48051,
      "cbid": 211, "correlation": 48051
    }
  },
  {
    "ph": "s", "id": 48051, "pid": 76337, "tid": -914061504, "ts": 1716454218241192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218288810, "dur": 3,
    "args": {
      "External id": 48059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48059,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 48059, "pid": 5, "tid": 7, "ts": 1716454218288810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241235, "dur": 10,
    "args": {
      "External id": 48059,
      "cbid": 211, "correlation": 48059
    }
  },
  {
    "ph": "s", "id": 48059, "pid": 76337, "tid": -914061504, "ts": 1716454218241235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241300, "dur": 1,
    "args": {
      "External id": 48075,
      "cbid": 251, "correlation": 48075
    }
  },
  {
    "ph": "f", "id": 48075, "pid": 76337, "tid": -914061504, "ts": 1716454218241300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241305, "dur": 0,
    "args": {
      "External id": 48077,
      "cbid": 251, "correlation": 48077
    }
  },
  {
    "ph": "f", "id": 48077, "pid": 76337, "tid": -914061504, "ts": 1716454218241305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218288815, "dur": 12,
    "args": {
      "External id": 48078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48078,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 48078, "pid": 5, "tid": 7, "ts": 1716454218288815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241307, "dur": 11,
    "args": {
      "External id": 48078,
      "cbid": 211, "correlation": 48078
    }
  },
  {
    "ph": "s", "id": 48078, "pid": 76337, "tid": -914061504, "ts": 1716454218241307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218288828, "dur": 5,
    "args": {
      "External id": 48080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48080,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 48080, "pid": 5, "tid": 7, "ts": 1716454218288828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241320, "dur": 5,
    "args": {
      "External id": 48080,
      "cbid": 211, "correlation": 48080
    }
  },
  {
    "ph": "s", "id": 48080, "pid": 76337, "tid": -914061504, "ts": 1716454218241320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218288834, "dur": 29,
    "args": {
      "External id": 48090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48090,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48090, "pid": 5, "tid": 7, "ts": 1716454218288834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241377, "dur": 12,
    "args": {
      "External id": 48090,
      "cbid": 211, "correlation": 48090
    }
  },
  {
    "ph": "s", "id": 48090, "pid": 76337, "tid": -914061504, "ts": 1716454218241377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218288865, "dur": 29,
    "args": {
      "External id": 48110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48110,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 48110, "pid": 5, "tid": 7, "ts": 1716454218288865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241446, "dur": 11,
    "args": {
      "External id": 48110,
      "cbid": 211, "correlation": 48110
    }
  },
  {
    "ph": "s", "id": 48110, "pid": 76337, "tid": -914061504, "ts": 1716454218241446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218288895, "dur": 4,
    "args": {
      "External id": 48122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48122,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 48122, "pid": 5, "tid": 7, "ts": 1716454218288895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241466, "dur": 6,
    "args": {
      "External id": 48122,
      "cbid": 211, "correlation": 48122
    }
  },
  {
    "ph": "s", "id": 48122, "pid": 76337, "tid": -914061504, "ts": 1716454218241466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218288900, "dur": 30,
    "args": {
      "External id": 48125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48125, "pid": 5, "tid": 7, "ts": 1716454218288900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241488, "dur": 7,
    "args": {
      "External id": 48125,
      "cbid": 211, "correlation": 48125
    }
  },
  {
    "ph": "s", "id": 48125, "pid": 76337, "tid": -914061504, "ts": 1716454218241488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218288931, "dur": 20,
    "args": {
      "External id": 48134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48134,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48134, "pid": 5, "tid": 7, "ts": 1716454218288931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241529, "dur": 9,
    "args": {
      "External id": 48134,
      "cbid": 211, "correlation": 48134
    }
  },
  {
    "ph": "s", "id": 48134, "pid": 76337, "tid": -914061504, "ts": 1716454218241529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241591, "dur": 0,
    "args": {
      "External id": 48144,
      "cbid": 317, "correlation": 48144
    }
  },
  {
    "ph": "f", "id": 48144, "pid": 76337, "tid": -914061504, "ts": 1716454218241591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241592, "dur": 0,
    "args": {
      "External id": 48145,
      "cbid": 203, "correlation": 48145
    }
  },
  {
    "ph": "f", "id": 48145, "pid": 76337, "tid": -914061504, "ts": 1716454218241592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241593, "dur": 0,
    "args": {
      "External id": 48146,
      "cbid": 205, "correlation": 48146
    }
  },
  {
    "ph": "f", "id": 48146, "pid": 76337, "tid": -914061504, "ts": 1716454218241593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218288952, "dur": 22,
    "args": {
      "External id": 48150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48150,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48150, "pid": 5, "tid": 7, "ts": 1716454218288952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241607, "dur": 12,
    "args": {
      "External id": 48150,
      "cbid": 211, "correlation": 48150
    }
  },
  {
    "ph": "s", "id": 48150, "pid": 76337, "tid": -914061504, "ts": 1716454218241607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218288975, "dur": 42,
    "args": {
      "External id": 48152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48152,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48152, "pid": 5, "tid": 7, "ts": 1716454218288975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241621, "dur": 5,
    "args": {
      "External id": 48152,
      "cbid": 211, "correlation": 48152
    }
  },
  {
    "ph": "s", "id": 48152, "pid": 76337, "tid": -914061504, "ts": 1716454218241621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218289019, "dur": 627,
    "args": {
      "External id": 48154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48154,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48154, "pid": 5, "tid": 7, "ts": 1716454218289019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241632, "dur": 6,
    "args": {
      "External id": 48154,
      "cbid": 211, "correlation": 48154
    }
  },
  {
    "ph": "s", "id": 48154, "pid": 76337, "tid": -914061504, "ts": 1716454218241632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218289647, "dur": 23,
    "args": {
      "External id": 48156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48156, "pid": 5, "tid": 7, "ts": 1716454218289647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241642, "dur": 5,
    "args": {
      "External id": 48156,
      "cbid": 211, "correlation": 48156
    }
  },
  {
    "ph": "s", "id": 48156, "pid": 76337, "tid": -914061504, "ts": 1716454218241642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218289671, "dur": 31,
    "args": {
      "External id": 48162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48162, "pid": 5, "tid": 7, "ts": 1716454218289671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241670, "dur": 8,
    "args": {
      "External id": 48162,
      "cbid": 211, "correlation": 48162
    }
  },
  {
    "ph": "s", "id": 48162, "pid": 76337, "tid": -914061504, "ts": 1716454218241670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218289704, "dur": 27,
    "args": {
      "External id": 48170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48170,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48170, "pid": 5, "tid": 7, "ts": 1716454218289704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241706, "dur": 9,
    "args": {
      "External id": 48170,
      "cbid": 211, "correlation": 48170
    }
  },
  {
    "ph": "s", "id": 48170, "pid": 76337, "tid": -914061504, "ts": 1716454218241706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218289732, "dur": 20,
    "args": {
      "External id": 48178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48178,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48178, "pid": 5, "tid": 7, "ts": 1716454218289732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241736, "dur": 8,
    "args": {
      "External id": 48178,
      "cbid": 211, "correlation": 48178
    }
  },
  {
    "ph": "s", "id": 48178, "pid": 76337, "tid": -914061504, "ts": 1716454218241736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218289753, "dur": 29,
    "args": {
      "External id": 48198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48198,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 48198, "pid": 5, "tid": 7, "ts": 1716454218289753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241815, "dur": 12,
    "args": {
      "External id": 48198,
      "cbid": 211, "correlation": 48198
    }
  },
  {
    "ph": "s", "id": 48198, "pid": 76337, "tid": -914061504, "ts": 1716454218241815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218289783, "dur": 4,
    "args": {
      "External id": 48210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48210,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 48210, "pid": 5, "tid": 7, "ts": 1716454218289783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241837, "dur": 6,
    "args": {
      "External id": 48210,
      "cbid": 211, "correlation": 48210
    }
  },
  {
    "ph": "s", "id": 48210, "pid": 76337, "tid": -914061504, "ts": 1716454218241837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218289788, "dur": 29,
    "args": {
      "External id": 48213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48213, "pid": 5, "tid": 7, "ts": 1716454218289788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241855, "dur": 6,
    "args": {
      "External id": 48213,
      "cbid": 211, "correlation": 48213
    }
  },
  {
    "ph": "s", "id": 48213, "pid": 76337, "tid": -914061504, "ts": 1716454218241855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241912, "dur": 0,
    "args": {
      "External id": 48224,
      "cbid": 317, "correlation": 48224
    }
  },
  {
    "ph": "f", "id": 48224, "pid": 76337, "tid": -914061504, "ts": 1716454218241912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241913, "dur": 0,
    "args": {
      "External id": 48225,
      "cbid": 203, "correlation": 48225
    }
  },
  {
    "ph": "f", "id": 48225, "pid": 76337, "tid": -914061504, "ts": 1716454218241913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241914, "dur": 0,
    "args": {
      "External id": 48226,
      "cbid": 205, "correlation": 48226
    }
  },
  {
    "ph": "f", "id": 48226, "pid": 76337, "tid": -914061504, "ts": 1716454218241914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218289818, "dur": 22,
    "args": {
      "External id": 48230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48230,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48230, "pid": 5, "tid": 7, "ts": 1716454218289818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241927, "dur": 12,
    "args": {
      "External id": 48230,
      "cbid": 211, "correlation": 48230
    }
  },
  {
    "ph": "s", "id": 48230, "pid": 76337, "tid": -914061504, "ts": 1716454218241927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218289841, "dur": 112,
    "args": {
      "External id": 48232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48232,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48232, "pid": 5, "tid": 7, "ts": 1716454218289841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241945, "dur": 6,
    "args": {
      "External id": 48232,
      "cbid": 211, "correlation": 48232
    }
  },
  {
    "ph": "s", "id": 48232, "pid": 76337, "tid": -914061504, "ts": 1716454218241945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218289954, "dur": 21,
    "args": {
      "External id": 48234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48234, "pid": 5, "tid": 7, "ts": 1716454218289954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241954, "dur": 8,
    "args": {
      "External id": 48234,
      "cbid": 211, "correlation": 48234
    }
  },
  {
    "ph": "s", "id": 48234, "pid": 76337, "tid": -914061504, "ts": 1716454218241954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218289976, "dur": 32,
    "args": {
      "External id": 48240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48240,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48240, "pid": 5, "tid": 7, "ts": 1716454218289976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218241993, "dur": 10,
    "args": {
      "External id": 48240,
      "cbid": 211, "correlation": 48240
    }
  },
  {
    "ph": "s", "id": 48240, "pid": 76337, "tid": -914061504, "ts": 1716454218241993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218290009, "dur": 179,
    "args": {
      "External id": 48249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48249,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48249, "pid": 5, "tid": 7, "ts": 1716454218290009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242077, "dur": 14,
    "args": {
      "External id": 48249,
      "cbid": 211, "correlation": 48249
    }
  },
  {
    "ph": "s", "id": 48249, "pid": 76337, "tid": -914061504, "ts": 1716454218242077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218290190, "dur": 62,
    "args": {
      "External id": 48271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48271,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48271, "pid": 5, "tid": 7, "ts": 1716454218290190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242135, "dur": 10,
    "args": {
      "External id": 48271,
      "cbid": 211, "correlation": 48271
    }
  },
  {
    "ph": "s", "id": 48271, "pid": 76337, "tid": -914061504, "ts": 1716454218242135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242222, "dur": 1,
    "args": {
      "External id": 48282,
      "cbid": 251, "correlation": 48282
    }
  },
  {
    "ph": "f", "id": 48282, "pid": 76337, "tid": -914061504, "ts": 1716454218242222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218290253, "dur": 147,
    "args": {
      "External id": 48283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48283,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48283, "pid": 5, "tid": 7, "ts": 1716454218290253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242228, "dur": 13,
    "args": {
      "External id": 48283,
      "cbid": 211, "correlation": 48283
    }
  },
  {
    "ph": "s", "id": 48283, "pid": 76337, "tid": -914061504, "ts": 1716454218242228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242298, "dur": 1,
    "args": {
      "External id": 48294,
      "cbid": 251, "correlation": 48294
    }
  },
  {
    "ph": "f", "id": 48294, "pid": 76337, "tid": -914061504, "ts": 1716454218242298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218290401, "dur": 138,
    "args": {
      "External id": 48295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48295,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48295, "pid": 5, "tid": 7, "ts": 1716454218290401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242302, "dur": 11,
    "args": {
      "External id": 48295,
      "cbid": 211, "correlation": 48295
    }
  },
  {
    "ph": "s", "id": 48295, "pid": 76337, "tid": -914061504, "ts": 1716454218242302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242368, "dur": 1,
    "args": {
      "External id": 48306,
      "cbid": 251, "correlation": 48306
    }
  },
  {
    "ph": "f", "id": 48306, "pid": 76337, "tid": -914061504, "ts": 1716454218242368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218290541, "dur": 142,
    "args": {
      "External id": 48307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48307,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48307, "pid": 5, "tid": 7, "ts": 1716454218290541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242372, "dur": 12,
    "args": {
      "External id": 48307,
      "cbid": 211, "correlation": 48307
    }
  },
  {
    "ph": "s", "id": 48307, "pid": 76337, "tid": -914061504, "ts": 1716454218242372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218290684, "dur": 1854,
    "args": {
      "External id": 48328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48328,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 48328, "pid": 5, "tid": 7, "ts": 1716454218290684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242455, "dur": 13,
    "args": {
      "External id": 48328,
      "cbid": 211, "correlation": 48328
    }
  },
  {
    "ph": "s", "id": 48328, "pid": 76337, "tid": -914061504, "ts": 1716454218242455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242553, "dur": 1,
    "args": {
      "External id": 48346,
      "cbid": 251, "correlation": 48346
    }
  },
  {
    "ph": "f", "id": 48346, "pid": 76337, "tid": -914061504, "ts": 1716454218242553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218292539, "dur": 143,
    "args": {
      "External id": 48348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48348,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 48348, "pid": 5, "tid": 7, "ts": 1716454218292539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242559, "dur": 13,
    "args": {
      "External id": 48348,
      "cbid": 211, "correlation": 48348
    }
  },
  {
    "ph": "s", "id": 48348, "pid": 76337, "tid": -914061504, "ts": 1716454218242559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218292683, "dur": 35,
    "args": {
      "External id": 48356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48356,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48356, "pid": 5, "tid": 7, "ts": 1716454218292683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242629, "dur": 13,
    "args": {
      "External id": 48356,
      "cbid": 211, "correlation": 48356
    }
  },
  {
    "ph": "s", "id": 48356, "pid": 76337, "tid": -914061504, "ts": 1716454218242629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218292720, "dur": 50,
    "args": {
      "External id": 48364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48364,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48364, "pid": 5, "tid": 7, "ts": 1716454218292720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242668, "dur": 8,
    "args": {
      "External id": 48364,
      "cbid": 211, "correlation": 48364
    }
  },
  {
    "ph": "s", "id": 48364, "pid": 76337, "tid": -914061504, "ts": 1716454218242668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218292771, "dur": 30,
    "args": {
      "External id": 48375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48375,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48375, "pid": 5, "tid": 7, "ts": 1716454218292771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242739, "dur": 12,
    "args": {
      "External id": 48375,
      "cbid": 211, "correlation": 48375
    }
  },
  {
    "ph": "s", "id": 48375, "pid": 76337, "tid": -914061504, "ts": 1716454218242739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218292803, "dur": 33,
    "args": {
      "External id": 48397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48397,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48397, "pid": 5, "tid": 7, "ts": 1716454218292803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242773, "dur": 7,
    "args": {
      "External id": 48397,
      "cbid": 211, "correlation": 48397
    }
  },
  {
    "ph": "s", "id": 48397, "pid": 76337, "tid": -914061504, "ts": 1716454218242773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242858, "dur": 1,
    "args": {
      "External id": 48408,
      "cbid": 251, "correlation": 48408
    }
  },
  {
    "ph": "f", "id": 48408, "pid": 76337, "tid": -914061504, "ts": 1716454218242858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218292837, "dur": 86,
    "args": {
      "External id": 48409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48409,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48409, "pid": 5, "tid": 7, "ts": 1716454218292837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242863, "dur": 13,
    "args": {
      "External id": 48409,
      "cbid": 211, "correlation": 48409
    }
  },
  {
    "ph": "s", "id": 48409, "pid": 76337, "tid": -914061504, "ts": 1716454218242863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242930, "dur": 1,
    "args": {
      "External id": 48420,
      "cbid": 251, "correlation": 48420
    }
  },
  {
    "ph": "f", "id": 48420, "pid": 76337, "tid": -914061504, "ts": 1716454218242930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242934, "dur": 0,
    "args": {
      "External id": 48421,
      "cbid": 251, "correlation": 48421
    }
  },
  {
    "ph": "f", "id": 48421, "pid": 76337, "tid": -914061504, "ts": 1716454218242934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218292925, "dur": 11,
    "args": {
      "External id": 48422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48422,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 48422, "pid": 5, "tid": 7, "ts": 1716454218292925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242936, "dur": 12,
    "args": {
      "External id": 48422,
      "cbid": 211, "correlation": 48422
    }
  },
  {
    "ph": "s", "id": 48422, "pid": 76337, "tid": -914061504, "ts": 1716454218242936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218292938, "dur": 5,
    "args": {
      "External id": 48424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48424,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 48424, "pid": 5, "tid": 7, "ts": 1716454218292938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218242950, "dur": 6,
    "args": {
      "External id": 48424,
      "cbid": 211, "correlation": 48424
    }
  },
  {
    "ph": "s", "id": 48424, "pid": 76337, "tid": -914061504, "ts": 1716454218242950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243017, "dur": 1,
    "args": {
      "External id": 48435,
      "cbid": 251, "correlation": 48435
    }
  },
  {
    "ph": "f", "id": 48435, "pid": 76337, "tid": -914061504, "ts": 1716454218243017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243020, "dur": 0,
    "args": {
      "External id": 48436,
      "cbid": 251, "correlation": 48436
    }
  },
  {
    "ph": "f", "id": 48436, "pid": 76337, "tid": -914061504, "ts": 1716454218243020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218292944, "dur": 7,
    "args": {
      "External id": 48437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48437,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 48437, "pid": 5, "tid": 7, "ts": 1716454218292944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243022, "dur": 12,
    "args": {
      "External id": 48437,
      "cbid": 211, "correlation": 48437
    }
  },
  {
    "ph": "s", "id": 48437, "pid": 76337, "tid": -914061504, "ts": 1716454218243022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218292952, "dur": 3,
    "args": {
      "External id": 48439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48439,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 48439, "pid": 5, "tid": 7, "ts": 1716454218292952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243035, "dur": 5,
    "args": {
      "External id": 48439,
      "cbid": 211, "correlation": 48439
    }
  },
  {
    "ph": "s", "id": 48439, "pid": 76337, "tid": -914061504, "ts": 1716454218243035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218292957, "dur": 88,
    "args": {
      "External id": 48460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48460,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 48460, "pid": 5, "tid": 7, "ts": 1716454218292957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243110, "dur": 12,
    "args": {
      "External id": 48460,
      "cbid": 211, "correlation": 48460
    }
  },
  {
    "ph": "s", "id": 48460, "pid": 76337, "tid": -914061504, "ts": 1716454218243110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243211, "dur": 1,
    "args": {
      "External id": 48478,
      "cbid": 251, "correlation": 48478
    }
  },
  {
    "ph": "f", "id": 48478, "pid": 76337, "tid": -914061504, "ts": 1716454218243211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218293046, "dur": 95,
    "args": {
      "External id": 48480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48480,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48480, "pid": 5, "tid": 7, "ts": 1716454218293046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243218, "dur": 13,
    "args": {
      "External id": 48480,
      "cbid": 211, "correlation": 48480
    }
  },
  {
    "ph": "s", "id": 48480, "pid": 76337, "tid": -914061504, "ts": 1716454218243218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218293143, "dur": 19,
    "args": {
      "External id": 48488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48488,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48488, "pid": 5, "tid": 7, "ts": 1716454218293143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243286, "dur": 12,
    "args": {
      "External id": 48488,
      "cbid": 211, "correlation": 48488
    }
  },
  {
    "ph": "s", "id": 48488, "pid": 76337, "tid": -914061504, "ts": 1716454218243286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218293162, "dur": 36,
    "args": {
      "External id": 48496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48496, "pid": 5, "tid": 7, "ts": 1716454218293162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243327, "dur": 9,
    "args": {
      "External id": 48496,
      "cbid": 211, "correlation": 48496
    }
  },
  {
    "ph": "s", "id": 48496, "pid": 76337, "tid": -914061504, "ts": 1716454218243327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218293200, "dur": 34,
    "args": {
      "External id": 48518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48518,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48518, "pid": 5, "tid": 7, "ts": 1716454218293200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243380, "dur": 10,
    "args": {
      "External id": 48518,
      "cbid": 211, "correlation": 48518
    }
  },
  {
    "ph": "s", "id": 48518, "pid": 76337, "tid": -914061504, "ts": 1716454218243380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243468, "dur": 1,
    "args": {
      "External id": 48534,
      "cbid": 251, "correlation": 48534
    }
  },
  {
    "ph": "f", "id": 48534, "pid": 76337, "tid": -914061504, "ts": 1716454218243468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243473, "dur": 0,
    "args": {
      "External id": 48536,
      "cbid": 251, "correlation": 48536
    }
  },
  {
    "ph": "f", "id": 48536, "pid": 76337, "tid": -914061504, "ts": 1716454218243473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218293235, "dur": 522,
    "args": {
      "External id": 48537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48537,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 48537, "pid": 5, "tid": 7, "ts": 1716454218293235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243477, "dur": 13,
    "args": {
      "External id": 48537,
      "cbid": 211, "correlation": 48537
    }
  },
  {
    "ph": "s", "id": 48537, "pid": 76337, "tid": -914061504, "ts": 1716454218243477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218293758, "dur": 122,
    "args": {
      "External id": 48545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48545, "pid": 5, "tid": 7, "ts": 1716454218293758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243545, "dur": 13,
    "args": {
      "External id": 48545,
      "cbid": 211, "correlation": 48545
    }
  },
  {
    "ph": "s", "id": 48545, "pid": 76337, "tid": -914061504, "ts": 1716454218243545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218293882, "dur": 126,
    "args": {
      "External id": 48553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48553, "pid": 5, "tid": 7, "ts": 1716454218293882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243575, "dur": 9,
    "args": {
      "External id": 48553,
      "cbid": 211, "correlation": 48553
    }
  },
  {
    "ph": "s", "id": 48553, "pid": 76337, "tid": -914061504, "ts": 1716454218243575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243654, "dur": 1,
    "args": {
      "External id": 48569,
      "cbid": 251, "correlation": 48569
    }
  },
  {
    "ph": "f", "id": 48569, "pid": 76337, "tid": -914061504, "ts": 1716454218243654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218294009, "dur": 296,
    "args": {
      "External id": 48571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48571,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48571, "pid": 5, "tid": 7, "ts": 1716454218294009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243659, "dur": 12,
    "args": {
      "External id": 48571,
      "cbid": 211, "correlation": 48571
    }
  },
  {
    "ph": "s", "id": 48571, "pid": 76337, "tid": -914061504, "ts": 1716454218243659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218294306, "dur": 28,
    "args": {
      "External id": 48579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48579,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48579, "pid": 5, "tid": 7, "ts": 1716454218294306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243701, "dur": 10,
    "args": {
      "External id": 48579,
      "cbid": 211, "correlation": 48579
    }
  },
  {
    "ph": "s", "id": 48579, "pid": 76337, "tid": -914061504, "ts": 1716454218243701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218294335, "dur": 78,
    "args": {
      "External id": 48590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48590, "pid": 5, "tid": 7, "ts": 1716454218294335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243768, "dur": 12,
    "args": {
      "External id": 48590,
      "cbid": 211, "correlation": 48590
    }
  },
  {
    "ph": "s", "id": 48590, "pid": 76337, "tid": -914061504, "ts": 1716454218243768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243831, "dur": 0,
    "args": {
      "External id": 48602,
      "cbid": 317, "correlation": 48602
    }
  },
  {
    "ph": "f", "id": 48602, "pid": 76337, "tid": -914061504, "ts": 1716454218243831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243832, "dur": 0,
    "args": {
      "External id": 48603,
      "cbid": 203, "correlation": 48603
    }
  },
  {
    "ph": "f", "id": 48603, "pid": 76337, "tid": -914061504, "ts": 1716454218243832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243833, "dur": 0,
    "args": {
      "External id": 48604,
      "cbid": 205, "correlation": 48604
    }
  },
  {
    "ph": "f", "id": 48604, "pid": 76337, "tid": -914061504, "ts": 1716454218243833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294414, "dur": 23,
    "args": {
      "External id": 48608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48608,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48608, "pid": 5, "tid": 7, "ts": 1716454218294414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243848, "dur": 12,
    "args": {
      "External id": 48608,
      "cbid": 211, "correlation": 48608
    }
  },
  {
    "ph": "s", "id": 48608, "pid": 76337, "tid": -914061504, "ts": 1716454218243848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218294439, "dur": 115,
    "args": {
      "External id": 48610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48610,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48610, "pid": 5, "tid": 7, "ts": 1716454218294439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243867, "dur": 6,
    "args": {
      "External id": 48610,
      "cbid": 211, "correlation": 48610
    }
  },
  {
    "ph": "s", "id": 48610, "pid": 76337, "tid": -914061504, "ts": 1716454218243867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294555, "dur": 21,
    "args": {
      "External id": 48612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48612,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48612, "pid": 5, "tid": 7, "ts": 1716454218294555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243877, "dur": 8,
    "args": {
      "External id": 48612,
      "cbid": 211, "correlation": 48612
    }
  },
  {
    "ph": "s", "id": 48612, "pid": 76337, "tid": -914061504, "ts": 1716454218243877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218294578, "dur": 32,
    "args": {
      "External id": 48618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48618,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48618, "pid": 5, "tid": 7, "ts": 1716454218294578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243908, "dur": 9,
    "args": {
      "External id": 48618,
      "cbid": 211, "correlation": 48618
    }
  },
  {
    "ph": "s", "id": 48618, "pid": 76337, "tid": -914061504, "ts": 1716454218243908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218294611, "dur": 27,
    "args": {
      "External id": 48626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48626,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48626, "pid": 5, "tid": 7, "ts": 1716454218294611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218243941, "dur": 9,
    "args": {
      "External id": 48626,
      "cbid": 211, "correlation": 48626
    }
  },
  {
    "ph": "s", "id": 48626, "pid": 76337, "tid": -914061504, "ts": 1716454218243941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244021, "dur": 0,
    "args": {
      "External id": 48636,
      "cbid": 317, "correlation": 48636
    }
  },
  {
    "ph": "f", "id": 48636, "pid": 76337, "tid": -914061504, "ts": 1716454218244021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244022, "dur": 0,
    "args": {
      "External id": 48637,
      "cbid": 203, "correlation": 48637
    }
  },
  {
    "ph": "f", "id": 48637, "pid": 76337, "tid": -914061504, "ts": 1716454218244022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244023, "dur": 0,
    "args": {
      "External id": 48638,
      "cbid": 205, "correlation": 48638
    }
  },
  {
    "ph": "f", "id": 48638, "pid": 76337, "tid": -914061504, "ts": 1716454218244023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294640, "dur": 22,
    "args": {
      "External id": 48642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48642,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48642, "pid": 5, "tid": 7, "ts": 1716454218294640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244041, "dur": 14,
    "args": {
      "External id": 48642,
      "cbid": 211, "correlation": 48642
    }
  },
  {
    "ph": "s", "id": 48642, "pid": 76337, "tid": -914061504, "ts": 1716454218244041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294663, "dur": 43,
    "args": {
      "External id": 48644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48644,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48644, "pid": 5, "tid": 7, "ts": 1716454218294663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244057, "dur": 6,
    "args": {
      "External id": 48644,
      "cbid": 211, "correlation": 48644
    }
  },
  {
    "ph": "s", "id": 48644, "pid": 76337, "tid": -914061504, "ts": 1716454218244057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218294707, "dur": 227,
    "args": {
      "External id": 48646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48646,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 48646, "pid": 5, "tid": 7, "ts": 1716454218294707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244069, "dur": 7,
    "args": {
      "External id": 48646,
      "cbid": 211, "correlation": 48646
    }
  },
  {
    "ph": "s", "id": 48646, "pid": 76337, "tid": -914061504, "ts": 1716454218244069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294936, "dur": 6,
    "args": {
      "External id": 48648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48648,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48648, "pid": 5, "tid": 7, "ts": 1716454218294936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244080, "dur": 5,
    "args": {
      "External id": 48648,
      "cbid": 211, "correlation": 48648
    }
  },
  {
    "ph": "s", "id": 48648, "pid": 76337, "tid": -914061504, "ts": 1716454218244080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218294943, "dur": 9,
    "args": {
      "External id": 48654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48654,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48654, "pid": 5, "tid": 7, "ts": 1716454218294943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244108, "dur": 8,
    "args": {
      "External id": 48654,
      "cbid": 211, "correlation": 48654
    }
  },
  {
    "ph": "s", "id": 48654, "pid": 76337, "tid": -914061504, "ts": 1716454218244108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294954, "dur": 12,
    "args": {
      "External id": 48674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48674,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 48674, "pid": 5, "tid": 7, "ts": 1716454218294954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244200, "dur": 13,
    "args": {
      "External id": 48674,
      "cbid": 211, "correlation": 48674
    }
  },
  {
    "ph": "s", "id": 48674, "pid": 76337, "tid": -914061504, "ts": 1716454218244200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218294967, "dur": 4,
    "args": {
      "External id": 48686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48686,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 48686, "pid": 5, "tid": 7, "ts": 1716454218294967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244226, "dur": 7,
    "args": {
      "External id": 48686,
      "cbid": 211, "correlation": 48686
    }
  },
  {
    "ph": "s", "id": 48686, "pid": 76337, "tid": -914061504, "ts": 1716454218244226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218294972, "dur": 11,
    "args": {
      "External id": 48689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48689, "pid": 5, "tid": 7, "ts": 1716454218294972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244244, "dur": 7,
    "args": {
      "External id": 48689,
      "cbid": 211, "correlation": 48689
    }
  },
  {
    "ph": "s", "id": 48689, "pid": 76337, "tid": -914061504, "ts": 1716454218244244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218294984, "dur": 7,
    "args": {
      "External id": 48698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48698,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48698, "pid": 5, "tid": 7, "ts": 1716454218294984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244285, "dur": 9,
    "args": {
      "External id": 48698,
      "cbid": 211, "correlation": 48698
    }
  },
  {
    "ph": "s", "id": 48698, "pid": 76337, "tid": -914061504, "ts": 1716454218244285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244337, "dur": 0,
    "args": {
      "External id": 48708,
      "cbid": 317, "correlation": 48708
    }
  },
  {
    "ph": "f", "id": 48708, "pid": 76337, "tid": -914061504, "ts": 1716454218244337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244337, "dur": 0,
    "args": {
      "External id": 48709,
      "cbid": 203, "correlation": 48709
    }
  },
  {
    "ph": "f", "id": 48709, "pid": 76337, "tid": -914061504, "ts": 1716454218244337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244338, "dur": 0,
    "args": {
      "External id": 48710,
      "cbid": 205, "correlation": 48710
    }
  },
  {
    "ph": "f", "id": 48710, "pid": 76337, "tid": -914061504, "ts": 1716454218244338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294992, "dur": 5,
    "args": {
      "External id": 48714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48714,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48714, "pid": 5, "tid": 7, "ts": 1716454218294992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244354, "dur": 12,
    "args": {
      "External id": 48714,
      "cbid": 211, "correlation": 48714
    }
  },
  {
    "ph": "s", "id": 48714, "pid": 76337, "tid": -914061504, "ts": 1716454218244354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218294999, "dur": 80,
    "args": {
      "External id": 48716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48716,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48716, "pid": 5, "tid": 7, "ts": 1716454218294999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244369, "dur": 5,
    "args": {
      "External id": 48716,
      "cbid": 211, "correlation": 48716
    }
  },
  {
    "ph": "s", "id": 48716, "pid": 76337, "tid": -914061504, "ts": 1716454218244369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218295081, "dur": 1,
    "args": {
      "External id": 48718,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 48718,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 48718, "pid": 5, "tid": 7, "ts": 1716454218295081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244382, "dur": 8,
    "args": {
      "External id": 48718,
      "cbid": 51, "correlation": 48718
    }
  },
  {
    "ph": "s", "id": 48718, "pid": 76337, "tid": -914061504, "ts": 1716454218244382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218295084, "dur": 530,
    "args": {
      "External id": 48719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48719,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48719, "pid": 5, "tid": 7, "ts": 1716454218295084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244392, "dur": 8,
    "args": {
      "External id": 48719,
      "cbid": 211, "correlation": 48719
    }
  },
  {
    "ph": "s", "id": 48719, "pid": 76337, "tid": -914061504, "ts": 1716454218244392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218295615, "dur": 12,
    "args": {
      "External id": 48721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48721,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48721, "pid": 5, "tid": 7, "ts": 1716454218295615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244403, "dur": 5,
    "args": {
      "External id": 48721,
      "cbid": 211, "correlation": 48721
    }
  },
  {
    "ph": "s", "id": 48721, "pid": 76337, "tid": -914061504, "ts": 1716454218244403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218295629, "dur": 14,
    "args": {
      "External id": 48727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48727, "pid": 5, "tid": 7, "ts": 1716454218295629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244431, "dur": 8,
    "args": {
      "External id": 48727,
      "cbid": 211, "correlation": 48727
    }
  },
  {
    "ph": "s", "id": 48727, "pid": 76337, "tid": -914061504, "ts": 1716454218244431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218295644, "dur": 3,
    "args": {
      "External id": 48735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48735,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 48735, "pid": 5, "tid": 7, "ts": 1716454218295644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244475, "dur": 9,
    "args": {
      "External id": 48735,
      "cbid": 211, "correlation": 48735
    }
  },
  {
    "ph": "s", "id": 48735, "pid": 76337, "tid": -914061504, "ts": 1716454218244475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244543, "dur": 1,
    "args": {
      "External id": 48751,
      "cbid": 251, "correlation": 48751
    }
  },
  {
    "ph": "f", "id": 48751, "pid": 76337, "tid": -914061504, "ts": 1716454218244543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244548, "dur": 0,
    "args": {
      "External id": 48753,
      "cbid": 251, "correlation": 48753
    }
  },
  {
    "ph": "f", "id": 48753, "pid": 76337, "tid": -914061504, "ts": 1716454218244548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218295649, "dur": 13,
    "args": {
      "External id": 48754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48754,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48754, "pid": 5, "tid": 7, "ts": 1716454218295649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244550, "dur": 11,
    "args": {
      "External id": 48754,
      "cbid": 211, "correlation": 48754
    }
  },
  {
    "ph": "s", "id": 48754, "pid": 76337, "tid": -914061504, "ts": 1716454218244550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218295663, "dur": 5,
    "args": {
      "External id": 48756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48756,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48756, "pid": 5, "tid": 7, "ts": 1716454218295663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244563, "dur": 5,
    "args": {
      "External id": 48756,
      "cbid": 211, "correlation": 48756
    }
  },
  {
    "ph": "s", "id": 48756, "pid": 76337, "tid": -914061504, "ts": 1716454218244563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218295670, "dur": 16,
    "args": {
      "External id": 48766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48766,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48766, "pid": 5, "tid": 7, "ts": 1716454218295670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244620, "dur": 12,
    "args": {
      "External id": 48766,
      "cbid": 211, "correlation": 48766
    }
  },
  {
    "ph": "s", "id": 48766, "pid": 76337, "tid": -914061504, "ts": 1716454218244620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218295687, "dur": 17,
    "args": {
      "External id": 48786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48786,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 48786, "pid": 5, "tid": 7, "ts": 1716454218295687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244686, "dur": 11,
    "args": {
      "External id": 48786,
      "cbid": 211, "correlation": 48786
    }
  },
  {
    "ph": "s", "id": 48786, "pid": 76337, "tid": -914061504, "ts": 1716454218244686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218295706, "dur": 5,
    "args": {
      "External id": 48798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48798,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 48798, "pid": 5, "tid": 7, "ts": 1716454218295706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244707, "dur": 6,
    "args": {
      "External id": 48798,
      "cbid": 211, "correlation": 48798
    }
  },
  {
    "ph": "s", "id": 48798, "pid": 76337, "tid": -914061504, "ts": 1716454218244707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218295712, "dur": 16,
    "args": {
      "External id": 48801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48801, "pid": 5, "tid": 7, "ts": 1716454218295712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244725, "dur": 6,
    "args": {
      "External id": 48801,
      "cbid": 211, "correlation": 48801
    }
  },
  {
    "ph": "s", "id": 48801, "pid": 76337, "tid": -914061504, "ts": 1716454218244725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218295729, "dur": 11,
    "args": {
      "External id": 48810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48810,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48810, "pid": 5, "tid": 7, "ts": 1716454218295729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244765, "dur": 12,
    "args": {
      "External id": 48810,
      "cbid": 211, "correlation": 48810
    }
  },
  {
    "ph": "s", "id": 48810, "pid": 76337, "tid": -914061504, "ts": 1716454218244765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244831, "dur": 0,
    "args": {
      "External id": 48820,
      "cbid": 317, "correlation": 48820
    }
  },
  {
    "ph": "f", "id": 48820, "pid": 76337, "tid": -914061504, "ts": 1716454218244831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244832, "dur": 0,
    "args": {
      "External id": 48821,
      "cbid": 203, "correlation": 48821
    }
  },
  {
    "ph": "f", "id": 48821, "pid": 76337, "tid": -914061504, "ts": 1716454218244832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244832, "dur": 0,
    "args": {
      "External id": 48822,
      "cbid": 205, "correlation": 48822
    }
  },
  {
    "ph": "f", "id": 48822, "pid": 76337, "tid": -914061504, "ts": 1716454218244832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218295741, "dur": 11,
    "args": {
      "External id": 48826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48826,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48826, "pid": 5, "tid": 7, "ts": 1716454218295741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244853, "dur": 12,
    "args": {
      "External id": 48826,
      "cbid": 211, "correlation": 48826
    }
  },
  {
    "ph": "s", "id": 48826, "pid": 76337, "tid": -914061504, "ts": 1716454218244853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218295753, "dur": 158,
    "args": {
      "External id": 48828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48828,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48828, "pid": 5, "tid": 7, "ts": 1716454218295753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244868, "dur": 5,
    "args": {
      "External id": 48828,
      "cbid": 211, "correlation": 48828
    }
  },
  {
    "ph": "s", "id": 48828, "pid": 76337, "tid": -914061504, "ts": 1716454218244868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218295913, "dur": 1,
    "args": {
      "External id": 48830,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 48830,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 48830, "pid": 5, "tid": 7, "ts": 1716454218295913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244880, "dur": 7,
    "args": {
      "External id": 48830,
      "cbid": 51, "correlation": 48830
    }
  },
  {
    "ph": "s", "id": 48830, "pid": 76337, "tid": -914061504, "ts": 1716454218244880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218295917, "dur": 650,
    "args": {
      "External id": 48831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48831,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48831, "pid": 5, "tid": 7, "ts": 1716454218295917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244888, "dur": 7,
    "args": {
      "External id": 48831,
      "cbid": 211, "correlation": 48831
    }
  },
  {
    "ph": "s", "id": 48831, "pid": 76337, "tid": -914061504, "ts": 1716454218244888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218296569, "dur": 13,
    "args": {
      "External id": 48833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48833,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48833, "pid": 5, "tid": 7, "ts": 1716454218296569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244900, "dur": 5,
    "args": {
      "External id": 48833,
      "cbid": 211, "correlation": 48833
    }
  },
  {
    "ph": "s", "id": 48833, "pid": 76337, "tid": -914061504, "ts": 1716454218244900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218296583, "dur": 14,
    "args": {
      "External id": 48839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48839, "pid": 5, "tid": 7, "ts": 1716454218296583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244929, "dur": 8,
    "args": {
      "External id": 48839,
      "cbid": 211, "correlation": 48839
    }
  },
  {
    "ph": "s", "id": 48839, "pid": 76337, "tid": -914061504, "ts": 1716454218244929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244996, "dur": 0,
    "args": {
      "External id": 48849,
      "cbid": 317, "correlation": 48849
    }
  },
  {
    "ph": "f", "id": 48849, "pid": 76337, "tid": -914061504, "ts": 1716454218244996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244997, "dur": 0,
    "args": {
      "External id": 48850,
      "cbid": 203, "correlation": 48850
    }
  },
  {
    "ph": "f", "id": 48850, "pid": 76337, "tid": -914061504, "ts": 1716454218244997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218244998, "dur": 0,
    "args": {
      "External id": 48851,
      "cbid": 205, "correlation": 48851
    }
  },
  {
    "ph": "f", "id": 48851, "pid": 76337, "tid": -914061504, "ts": 1716454218244998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218296598, "dur": 8,
    "args": {
      "External id": 48855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48855,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48855, "pid": 5, "tid": 7, "ts": 1716454218296598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245019, "dur": 12,
    "args": {
      "External id": 48855,
      "cbid": 211, "correlation": 48855
    }
  },
  {
    "ph": "s", "id": 48855, "pid": 76337, "tid": -914061504, "ts": 1716454218245019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218296608, "dur": 3,
    "args": {
      "External id": 48857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 48857, "pid": 5, "tid": 7, "ts": 1716454218296608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245037, "dur": 6,
    "args": {
      "External id": 48857,
      "cbid": 211, "correlation": 48857
    }
  },
  {
    "ph": "s", "id": 48857, "pid": 76337, "tid": -914061504, "ts": 1716454218245037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245047, "dur": 0,
    "args": {
      "External id": 48858,
      "cbid": 51, "correlation": 48858
    }
  },
  {
    "ph": "s", "id": 48858, "pid": 76337, "tid": -914061504, "ts": 1716454218245047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218296612, "dur": 55,
    "args": {
      "External id": 48859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48859,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 48859, "pid": 5, "tid": 7, "ts": 1716454218296612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245048, "dur": 6,
    "args": {
      "External id": 48859,
      "cbid": 211, "correlation": 48859
    }
  },
  {
    "ph": "s", "id": 48859, "pid": 76337, "tid": -914061504, "ts": 1716454218245048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218296668, "dur": 13,
    "args": {
      "External id": 48864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48864, "pid": 5, "tid": 7, "ts": 1716454218296668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245074, "dur": 8,
    "args": {
      "External id": 48864,
      "cbid": 211, "correlation": 48864
    }
  },
  {
    "ph": "s", "id": 48864, "pid": 76337, "tid": -914061504, "ts": 1716454218245074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218296683, "dur": 12,
    "args": {
      "External id": 48872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48872,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48872, "pid": 5, "tid": 7, "ts": 1716454218296683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245106, "dur": 8,
    "args": {
      "External id": 48872,
      "cbid": 211, "correlation": 48872
    }
  },
  {
    "ph": "s", "id": 48872, "pid": 76337, "tid": -914061504, "ts": 1716454218245106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218296696, "dur": 10,
    "args": {
      "External id": 48880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48880,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48880, "pid": 5, "tid": 7, "ts": 1716454218296696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245136, "dur": 8,
    "args": {
      "External id": 48880,
      "cbid": 211, "correlation": 48880
    }
  },
  {
    "ph": "s", "id": 48880, "pid": 76337, "tid": -914061504, "ts": 1716454218245136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218296708, "dur": 18,
    "args": {
      "External id": 48900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48900,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 48900, "pid": 5, "tid": 7, "ts": 1716454218296708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245216, "dur": 13,
    "args": {
      "External id": 48900,
      "cbid": 211, "correlation": 48900
    }
  },
  {
    "ph": "s", "id": 48900, "pid": 76337, "tid": -914061504, "ts": 1716454218245216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218296727, "dur": 4,
    "args": {
      "External id": 48912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48912,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 48912, "pid": 5, "tid": 7, "ts": 1716454218296727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245238, "dur": 6,
    "args": {
      "External id": 48912,
      "cbid": 211, "correlation": 48912
    }
  },
  {
    "ph": "s", "id": 48912, "pid": 76337, "tid": -914061504, "ts": 1716454218245238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218296733, "dur": 18,
    "args": {
      "External id": 48915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48915,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48915, "pid": 5, "tid": 7, "ts": 1716454218296733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245255, "dur": 7,
    "args": {
      "External id": 48915,
      "cbid": 211, "correlation": 48915
    }
  },
  {
    "ph": "s", "id": 48915, "pid": 76337, "tid": -914061504, "ts": 1716454218245255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245312, "dur": 0,
    "args": {
      "External id": 48926,
      "cbid": 317, "correlation": 48926
    }
  },
  {
    "ph": "f", "id": 48926, "pid": 76337, "tid": -914061504, "ts": 1716454218245312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245313, "dur": 0,
    "args": {
      "External id": 48927,
      "cbid": 203, "correlation": 48927
    }
  },
  {
    "ph": "f", "id": 48927, "pid": 76337, "tid": -914061504, "ts": 1716454218245313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245314, "dur": 0,
    "args": {
      "External id": 48928,
      "cbid": 205, "correlation": 48928
    }
  },
  {
    "ph": "f", "id": 48928, "pid": 76337, "tid": -914061504, "ts": 1716454218245314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218296752, "dur": 12,
    "args": {
      "External id": 48932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48932,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48932, "pid": 5, "tid": 7, "ts": 1716454218296752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245333, "dur": 12,
    "args": {
      "External id": 48932,
      "cbid": 211, "correlation": 48932
    }
  },
  {
    "ph": "s", "id": 48932, "pid": 76337, "tid": -914061504, "ts": 1716454218245333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218296766, "dur": 3,
    "args": {
      "External id": 48934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 48934, "pid": 5, "tid": 7, "ts": 1716454218296766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245349, "dur": 6,
    "args": {
      "External id": 48934,
      "cbid": 211, "correlation": 48934
    }
  },
  {
    "ph": "s", "id": 48934, "pid": 76337, "tid": -914061504, "ts": 1716454218245349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245359, "dur": 0,
    "args": {
      "External id": 48935,
      "cbid": 51, "correlation": 48935
    }
  },
  {
    "ph": "s", "id": 48935, "pid": 76337, "tid": -914061504, "ts": 1716454218245359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218296770, "dur": 95,
    "args": {
      "External id": 48936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48936,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 48936, "pid": 5, "tid": 7, "ts": 1716454218296770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245360, "dur": 7,
    "args": {
      "External id": 48936,
      "cbid": 211, "correlation": 48936
    }
  },
  {
    "ph": "s", "id": 48936, "pid": 76337, "tid": -914061504, "ts": 1716454218245360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218296866, "dur": 15,
    "args": {
      "External id": 48941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48941, "pid": 5, "tid": 7, "ts": 1716454218296866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245390, "dur": 9,
    "args": {
      "External id": 48941,
      "cbid": 211, "correlation": 48941
    }
  },
  {
    "ph": "s", "id": 48941, "pid": 76337, "tid": -914061504, "ts": 1716454218245390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218296882, "dur": 82,
    "args": {
      "External id": 48950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48950,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48950, "pid": 5, "tid": 7, "ts": 1716454218296882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245472, "dur": 14,
    "args": {
      "External id": 48950,
      "cbid": 211, "correlation": 48950
    }
  },
  {
    "ph": "s", "id": 48950, "pid": 76337, "tid": -914061504, "ts": 1716454218245472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218296966, "dur": 29,
    "args": {
      "External id": 48972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48972,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 48972, "pid": 5, "tid": 7, "ts": 1716454218296966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245528, "dur": 10,
    "args": {
      "External id": 48972,
      "cbid": 211, "correlation": 48972
    }
  },
  {
    "ph": "s", "id": 48972, "pid": 76337, "tid": -914061504, "ts": 1716454218245528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245616, "dur": 1,
    "args": {
      "External id": 48983,
      "cbid": 251, "correlation": 48983
    }
  },
  {
    "ph": "f", "id": 48983, "pid": 76337, "tid": -914061504, "ts": 1716454218245616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218296996, "dur": 137,
    "args": {
      "External id": 48984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48984,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48984, "pid": 5, "tid": 7, "ts": 1716454218296996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245622, "dur": 14,
    "args": {
      "External id": 48984,
      "cbid": 211, "correlation": 48984
    }
  },
  {
    "ph": "s", "id": 48984, "pid": 76337, "tid": -914061504, "ts": 1716454218245622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245693, "dur": 1,
    "args": {
      "External id": 48995,
      "cbid": 251, "correlation": 48995
    }
  },
  {
    "ph": "f", "id": 48995, "pid": 76337, "tid": -914061504, "ts": 1716454218245693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218297135, "dur": 154,
    "args": {
      "External id": 48996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 48996,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 48996, "pid": 5, "tid": 7, "ts": 1716454218297135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245697, "dur": 11,
    "args": {
      "External id": 48996,
      "cbid": 211, "correlation": 48996
    }
  },
  {
    "ph": "s", "id": 48996, "pid": 76337, "tid": -914061504, "ts": 1716454218245697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245761, "dur": 1,
    "args": {
      "External id": 49007,
      "cbid": 251, "correlation": 49007
    }
  },
  {
    "ph": "f", "id": 49007, "pid": 76337, "tid": -914061504, "ts": 1716454218245761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218297290, "dur": 154,
    "args": {
      "External id": 49008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49008,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49008, "pid": 5, "tid": 7, "ts": 1716454218297290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245765, "dur": 11,
    "args": {
      "External id": 49008,
      "cbid": 211, "correlation": 49008
    }
  },
  {
    "ph": "s", "id": 49008, "pid": 76337, "tid": -914061504, "ts": 1716454218245765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218297446, "dur": 330,
    "args": {
      "External id": 49033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49033,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49033, "pid": 5, "tid": 7, "ts": 1716454218297446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245853, "dur": 15,
    "args": {
      "External id": 49033,
      "cbid": 211, "correlation": 49033
    }
  },
  {
    "ph": "s", "id": 49033, "pid": 76337, "tid": -914061504, "ts": 1716454218245853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245957, "dur": 1,
    "args": {
      "External id": 49051,
      "cbid": 251, "correlation": 49051
    }
  },
  {
    "ph": "f", "id": 49051, "pid": 76337, "tid": -914061504, "ts": 1716454218245957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218297777, "dur": 161,
    "args": {
      "External id": 49053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49053,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49053, "pid": 5, "tid": 7, "ts": 1716454218297777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218245962, "dur": 22,
    "args": {
      "External id": 49053,
      "cbid": 211, "correlation": 49053
    }
  },
  {
    "ph": "s", "id": 49053, "pid": 76337, "tid": -914061504, "ts": 1716454218245962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218297939, "dur": 19,
    "args": {
      "External id": 49061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49061,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49061, "pid": 5, "tid": 7, "ts": 1716454218297939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246043, "dur": 13,
    "args": {
      "External id": 49061,
      "cbid": 211, "correlation": 49061
    }
  },
  {
    "ph": "s", "id": 49061, "pid": 76337, "tid": -914061504, "ts": 1716454218246043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218297960, "dur": 27,
    "args": {
      "External id": 49069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49069,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49069, "pid": 5, "tid": 7, "ts": 1716454218297960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246083, "dur": 9,
    "args": {
      "External id": 49069,
      "cbid": 211, "correlation": 49069
    }
  },
  {
    "ph": "s", "id": 49069, "pid": 76337, "tid": -914061504, "ts": 1716454218246083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218297988, "dur": 18,
    "args": {
      "External id": 49080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49080,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49080, "pid": 5, "tid": 7, "ts": 1716454218297988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246155, "dur": 12,
    "args": {
      "External id": 49080,
      "cbid": 211, "correlation": 49080
    }
  },
  {
    "ph": "s", "id": 49080, "pid": 76337, "tid": -914061504, "ts": 1716454218246155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218298006, "dur": 15,
    "args": {
      "External id": 49102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49102,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49102, "pid": 5, "tid": 7, "ts": 1716454218298006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246189, "dur": 8,
    "args": {
      "External id": 49102,
      "cbid": 211, "correlation": 49102
    }
  },
  {
    "ph": "s", "id": 49102, "pid": 76337, "tid": -914061504, "ts": 1716454218246189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246275, "dur": 1,
    "args": {
      "External id": 49113,
      "cbid": 251, "correlation": 49113
    }
  },
  {
    "ph": "f", "id": 49113, "pid": 76337, "tid": -914061504, "ts": 1716454218246275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218298023, "dur": 86,
    "args": {
      "External id": 49114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49114,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49114, "pid": 5, "tid": 7, "ts": 1716454218298023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246281, "dur": 14,
    "args": {
      "External id": 49114,
      "cbid": 211, "correlation": 49114
    }
  },
  {
    "ph": "s", "id": 49114, "pid": 76337, "tid": -914061504, "ts": 1716454218246281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246351, "dur": 1,
    "args": {
      "External id": 49125,
      "cbid": 251, "correlation": 49125
    }
  },
  {
    "ph": "f", "id": 49125, "pid": 76337, "tid": -914061504, "ts": 1716454218246351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246355, "dur": 0,
    "args": {
      "External id": 49126,
      "cbid": 251, "correlation": 49126
    }
  },
  {
    "ph": "f", "id": 49126, "pid": 76337, "tid": -914061504, "ts": 1716454218246355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218298110, "dur": 13,
    "args": {
      "External id": 49127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49127,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49127, "pid": 5, "tid": 7, "ts": 1716454218298110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246357, "dur": 12,
    "args": {
      "External id": 49127,
      "cbid": 211, "correlation": 49127
    }
  },
  {
    "ph": "s", "id": 49127, "pid": 76337, "tid": -914061504, "ts": 1716454218246357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218298124, "dur": 5,
    "args": {
      "External id": 49129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49129,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49129, "pid": 5, "tid": 7, "ts": 1716454218298124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246371, "dur": 6,
    "args": {
      "External id": 49129,
      "cbid": 211, "correlation": 49129
    }
  },
  {
    "ph": "s", "id": 49129, "pid": 76337, "tid": -914061504, "ts": 1716454218246371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246428, "dur": 1,
    "args": {
      "External id": 49140,
      "cbid": 251, "correlation": 49140
    }
  },
  {
    "ph": "f", "id": 49140, "pid": 76337, "tid": -914061504, "ts": 1716454218246428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246432, "dur": 0,
    "args": {
      "External id": 49141,
      "cbid": 251, "correlation": 49141
    }
  },
  {
    "ph": "f", "id": 49141, "pid": 76337, "tid": -914061504, "ts": 1716454218246432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218298131, "dur": 8,
    "args": {
      "External id": 49142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49142,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49142, "pid": 5, "tid": 7, "ts": 1716454218298131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246433, "dur": 11,
    "args": {
      "External id": 49142,
      "cbid": 211, "correlation": 49142
    }
  },
  {
    "ph": "s", "id": 49142, "pid": 76337, "tid": -914061504, "ts": 1716454218246433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218298141, "dur": 3,
    "args": {
      "External id": 49144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49144,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49144, "pid": 5, "tid": 7, "ts": 1716454218298141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246446, "dur": 5,
    "args": {
      "External id": 49144,
      "cbid": 211, "correlation": 49144
    }
  },
  {
    "ph": "s", "id": 49144, "pid": 76337, "tid": -914061504, "ts": 1716454218246446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218298145, "dur": 53,
    "args": {
      "External id": 49169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49169,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49169, "pid": 5, "tid": 7, "ts": 1716454218298145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246523, "dur": 15,
    "args": {
      "External id": 49169,
      "cbid": 211, "correlation": 49169
    }
  },
  {
    "ph": "s", "id": 49169, "pid": 76337, "tid": -914061504, "ts": 1716454218246523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246627, "dur": 2,
    "args": {
      "External id": 49187,
      "cbid": 251, "correlation": 49187
    }
  },
  {
    "ph": "f", "id": 49187, "pid": 76337, "tid": -914061504, "ts": 1716454218246627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218298199, "dur": 89,
    "args": {
      "External id": 49189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49189,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49189, "pid": 5, "tid": 7, "ts": 1716454218298199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246634, "dur": 17,
    "args": {
      "External id": 49189,
      "cbid": 211, "correlation": 49189
    }
  },
  {
    "ph": "s", "id": 49189, "pid": 76337, "tid": -914061504, "ts": 1716454218246634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218298289, "dur": 10,
    "args": {
      "External id": 49197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49197,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49197, "pid": 5, "tid": 7, "ts": 1716454218298289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246707, "dur": 12,
    "args": {
      "External id": 49197,
      "cbid": 211, "correlation": 49197
    }
  },
  {
    "ph": "s", "id": 49197, "pid": 76337, "tid": -914061504, "ts": 1716454218246707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218298300, "dur": 20,
    "args": {
      "External id": 49205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49205, "pid": 5, "tid": 7, "ts": 1716454218298300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246749, "dur": 9,
    "args": {
      "External id": 49205,
      "cbid": 211, "correlation": 49205
    }
  },
  {
    "ph": "s", "id": 49205, "pid": 76337, "tid": -914061504, "ts": 1716454218246749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218298322, "dur": 17,
    "args": {
      "External id": 49227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49227,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49227, "pid": 5, "tid": 7, "ts": 1716454218298322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246801, "dur": 10,
    "args": {
      "External id": 49227,
      "cbid": 211, "correlation": 49227
    }
  },
  {
    "ph": "s", "id": 49227, "pid": 76337, "tid": -914061504, "ts": 1716454218246801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246890, "dur": 2,
    "args": {
      "External id": 49243,
      "cbid": 251, "correlation": 49243
    }
  },
  {
    "ph": "f", "id": 49243, "pid": 76337, "tid": -914061504, "ts": 1716454218246890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246896, "dur": 0,
    "args": {
      "External id": 49245,
      "cbid": 251, "correlation": 49245
    }
  },
  {
    "ph": "f", "id": 49245, "pid": 76337, "tid": -914061504, "ts": 1716454218246896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218298341, "dur": 490,
    "args": {
      "External id": 49246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49246,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49246, "pid": 5, "tid": 7, "ts": 1716454218298341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246899, "dur": 18,
    "args": {
      "External id": 49246,
      "cbid": 211, "correlation": 49246
    }
  },
  {
    "ph": "s", "id": 49246, "pid": 76337, "tid": -914061504, "ts": 1716454218246899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218298832, "dur": 65,
    "args": {
      "External id": 49254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49254, "pid": 5, "tid": 7, "ts": 1716454218298832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218246969, "dur": 22,
    "args": {
      "External id": 49254,
      "cbid": 211, "correlation": 49254
    }
  },
  {
    "ph": "s", "id": 49254, "pid": 76337, "tid": -914061504, "ts": 1716454218246969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218298899, "dur": 67,
    "args": {
      "External id": 49262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49262,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49262, "pid": 5, "tid": 7, "ts": 1716454218298899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247010, "dur": 9,
    "args": {
      "External id": 49262,
      "cbid": 211, "correlation": 49262
    }
  },
  {
    "ph": "s", "id": 49262, "pid": 76337, "tid": -914061504, "ts": 1716454218247010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247093, "dur": 1,
    "args": {
      "External id": 49278,
      "cbid": 251, "correlation": 49278
    }
  },
  {
    "ph": "f", "id": 49278, "pid": 76337, "tid": -914061504, "ts": 1716454218247093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218298968, "dur": 1,
    "args": {
      "External id": 49280,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 49280,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 49280, "pid": 5, "tid": 7, "ts": 1716454218298968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247099, "dur": 11,
    "args": {
      "External id": 49280,
      "cbid": 51, "correlation": 49280
    }
  },
  {
    "ph": "s", "id": 49280, "pid": 76337, "tid": -914061504, "ts": 1716454218247099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218298972, "dur": 265,
    "args": {
      "External id": 49281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49281,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49281, "pid": 5, "tid": 7, "ts": 1716454218298972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247112, "dur": 11,
    "args": {
      "External id": 49281,
      "cbid": 211, "correlation": 49281
    }
  },
  {
    "ph": "s", "id": 49281, "pid": 76337, "tid": -914061504, "ts": 1716454218247112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218299239, "dur": 14,
    "args": {
      "External id": 49289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49289,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49289, "pid": 5, "tid": 7, "ts": 1716454218299239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247154, "dur": 10,
    "args": {
      "External id": 49289,
      "cbid": 211, "correlation": 49289
    }
  },
  {
    "ph": "s", "id": 49289, "pid": 76337, "tid": -914061504, "ts": 1716454218247154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218299254, "dur": 36,
    "args": {
      "External id": 49300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49300,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49300, "pid": 5, "tid": 7, "ts": 1716454218299254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247220, "dur": 12,
    "args": {
      "External id": 49300,
      "cbid": 211, "correlation": 49300
    }
  },
  {
    "ph": "s", "id": 49300, "pid": 76337, "tid": -914061504, "ts": 1716454218247220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247285, "dur": 0,
    "args": {
      "External id": 49312,
      "cbid": 317, "correlation": 49312
    }
  },
  {
    "ph": "f", "id": 49312, "pid": 76337, "tid": -914061504, "ts": 1716454218247285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247285, "dur": 0,
    "args": {
      "External id": 49313,
      "cbid": 203, "correlation": 49313
    }
  },
  {
    "ph": "f", "id": 49313, "pid": 76337, "tid": -914061504, "ts": 1716454218247285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247286, "dur": 0,
    "args": {
      "External id": 49314,
      "cbid": 205, "correlation": 49314
    }
  },
  {
    "ph": "f", "id": 49314, "pid": 76337, "tid": -914061504, "ts": 1716454218247286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218299291, "dur": 13,
    "args": {
      "External id": 49318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49318, "pid": 5, "tid": 7, "ts": 1716454218299291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247302, "dur": 12,
    "args": {
      "External id": 49318,
      "cbid": 211, "correlation": 49318
    }
  },
  {
    "ph": "s", "id": 49318, "pid": 76337, "tid": -914061504, "ts": 1716454218247302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218299305, "dur": 4,
    "args": {
      "External id": 49320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49320, "pid": 5, "tid": 7, "ts": 1716454218299305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247322, "dur": 6,
    "args": {
      "External id": 49320,
      "cbid": 211, "correlation": 49320
    }
  },
  {
    "ph": "s", "id": 49320, "pid": 76337, "tid": -914061504, "ts": 1716454218247322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247331, "dur": 0,
    "args": {
      "External id": 49321,
      "cbid": 51, "correlation": 49321
    }
  },
  {
    "ph": "s", "id": 49321, "pid": 76337, "tid": -914061504, "ts": 1716454218247331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218299310, "dur": 94,
    "args": {
      "External id": 49322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49322,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 49322, "pid": 5, "tid": 7, "ts": 1716454218299310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247332, "dur": 5,
    "args": {
      "External id": 49322,
      "cbid": 211, "correlation": 49322
    }
  },
  {
    "ph": "s", "id": 49322, "pid": 76337, "tid": -914061504, "ts": 1716454218247332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218299405, "dur": 17,
    "args": {
      "External id": 49327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49327, "pid": 5, "tid": 7, "ts": 1716454218299405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247360, "dur": 8,
    "args": {
      "External id": 49327,
      "cbid": 211, "correlation": 49327
    }
  },
  {
    "ph": "s", "id": 49327, "pid": 76337, "tid": -914061504, "ts": 1716454218247360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218299423, "dur": 12,
    "args": {
      "External id": 49335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49335,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49335, "pid": 5, "tid": 7, "ts": 1716454218299423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247391, "dur": 9,
    "args": {
      "External id": 49335,
      "cbid": 211, "correlation": 49335
    }
  },
  {
    "ph": "s", "id": 49335, "pid": 76337, "tid": -914061504, "ts": 1716454218247391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218299436, "dur": 18,
    "args": {
      "External id": 49355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49355,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 49355, "pid": 5, "tid": 7, "ts": 1716454218299436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247465, "dur": 11,
    "args": {
      "External id": 49355,
      "cbid": 211, "correlation": 49355
    }
  },
  {
    "ph": "s", "id": 49355, "pid": 76337, "tid": -914061504, "ts": 1716454218247465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218299455, "dur": 4,
    "args": {
      "External id": 49367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49367,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 49367, "pid": 5, "tid": 7, "ts": 1716454218299455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247486, "dur": 6,
    "args": {
      "External id": 49367,
      "cbid": 211, "correlation": 49367
    }
  },
  {
    "ph": "s", "id": 49367, "pid": 76337, "tid": -914061504, "ts": 1716454218247486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218299461, "dur": 18,
    "args": {
      "External id": 49370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49370,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49370, "pid": 5, "tid": 7, "ts": 1716454218299461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247504, "dur": 7,
    "args": {
      "External id": 49370,
      "cbid": 211, "correlation": 49370
    }
  },
  {
    "ph": "s", "id": 49370, "pid": 76337, "tid": -914061504, "ts": 1716454218247504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218299481, "dur": 11,
    "args": {
      "External id": 49379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49379,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49379, "pid": 5, "tid": 7, "ts": 1716454218299481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247542, "dur": 10,
    "args": {
      "External id": 49379,
      "cbid": 211, "correlation": 49379
    }
  },
  {
    "ph": "s", "id": 49379, "pid": 76337, "tid": -914061504, "ts": 1716454218247542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247596, "dur": 0,
    "args": {
      "External id": 49389,
      "cbid": 317, "correlation": 49389
    }
  },
  {
    "ph": "f", "id": 49389, "pid": 76337, "tid": -914061504, "ts": 1716454218247596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247597, "dur": 0,
    "args": {
      "External id": 49390,
      "cbid": 203, "correlation": 49390
    }
  },
  {
    "ph": "f", "id": 49390, "pid": 76337, "tid": -914061504, "ts": 1716454218247597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247598, "dur": 0,
    "args": {
      "External id": 49391,
      "cbid": 205, "correlation": 49391
    }
  },
  {
    "ph": "f", "id": 49391, "pid": 76337, "tid": -914061504, "ts": 1716454218247598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218299493, "dur": 11,
    "args": {
      "External id": 49395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49395,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49395, "pid": 5, "tid": 7, "ts": 1716454218299493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247612, "dur": 12,
    "args": {
      "External id": 49395,
      "cbid": 211, "correlation": 49395
    }
  },
  {
    "ph": "s", "id": 49395, "pid": 76337, "tid": -914061504, "ts": 1716454218247612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218299505, "dur": 158,
    "args": {
      "External id": 49397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49397,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49397, "pid": 5, "tid": 7, "ts": 1716454218299505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247626, "dur": 5,
    "args": {
      "External id": 49397,
      "cbid": 211, "correlation": 49397
    }
  },
  {
    "ph": "s", "id": 49397, "pid": 76337, "tid": -914061504, "ts": 1716454218247626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218299665, "dur": 1,
    "args": {
      "External id": 49399,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 49399,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 49399, "pid": 5, "tid": 7, "ts": 1716454218299665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247638, "dur": 7,
    "args": {
      "External id": 49399,
      "cbid": 51, "correlation": 49399
    }
  },
  {
    "ph": "s", "id": 49399, "pid": 76337, "tid": -914061504, "ts": 1716454218247638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218299669, "dur": 651,
    "args": {
      "External id": 49400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49400,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49400, "pid": 5, "tid": 7, "ts": 1716454218299669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247645, "dur": 6,
    "args": {
      "External id": 49400,
      "cbid": 211, "correlation": 49400
    }
  },
  {
    "ph": "s", "id": 49400, "pid": 76337, "tid": -914061504, "ts": 1716454218247645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218300321, "dur": 14,
    "args": {
      "External id": 49402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49402,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49402, "pid": 5, "tid": 7, "ts": 1716454218300321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247656, "dur": 6,
    "args": {
      "External id": 49402,
      "cbid": 211, "correlation": 49402
    }
  },
  {
    "ph": "s", "id": 49402, "pid": 76337, "tid": -914061504, "ts": 1716454218247656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218300337, "dur": 14,
    "args": {
      "External id": 49408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49408, "pid": 5, "tid": 7, "ts": 1716454218300337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247685, "dur": 8,
    "args": {
      "External id": 49408,
      "cbid": 211, "correlation": 49408
    }
  },
  {
    "ph": "s", "id": 49408, "pid": 76337, "tid": -914061504, "ts": 1716454218247685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218300352, "dur": 4,
    "args": {
      "External id": 49416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49416,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 49416, "pid": 5, "tid": 7, "ts": 1716454218300352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247729, "dur": 9,
    "args": {
      "External id": 49416,
      "cbid": 211, "correlation": 49416
    }
  },
  {
    "ph": "s", "id": 49416, "pid": 76337, "tid": -914061504, "ts": 1716454218247729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247796, "dur": 1,
    "args": {
      "External id": 49432,
      "cbid": 251, "correlation": 49432
    }
  },
  {
    "ph": "f", "id": 49432, "pid": 76337, "tid": -914061504, "ts": 1716454218247796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247801, "dur": 0,
    "args": {
      "External id": 49434,
      "cbid": 251, "correlation": 49434
    }
  },
  {
    "ph": "f", "id": 49434, "pid": 76337, "tid": -914061504, "ts": 1716454218247801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218300357, "dur": 13,
    "args": {
      "External id": 49435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49435,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49435, "pid": 5, "tid": 7, "ts": 1716454218300357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247803, "dur": 11,
    "args": {
      "External id": 49435,
      "cbid": 211, "correlation": 49435
    }
  },
  {
    "ph": "s", "id": 49435, "pid": 76337, "tid": -914061504, "ts": 1716454218247803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218300371, "dur": 5,
    "args": {
      "External id": 49437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49437,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49437, "pid": 5, "tid": 7, "ts": 1716454218300371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247817, "dur": 6,
    "args": {
      "External id": 49437,
      "cbid": 211, "correlation": 49437
    }
  },
  {
    "ph": "s", "id": 49437, "pid": 76337, "tid": -914061504, "ts": 1716454218247817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218300377, "dur": 16,
    "args": {
      "External id": 49447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49447, "pid": 5, "tid": 7, "ts": 1716454218300377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247875, "dur": 15,
    "args": {
      "External id": 49447,
      "cbid": 211, "correlation": 49447
    }
  },
  {
    "ph": "s", "id": 49447, "pid": 76337, "tid": -914061504, "ts": 1716454218247875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218300395, "dur": 17,
    "args": {
      "External id": 49467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49467,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 49467, "pid": 5, "tid": 7, "ts": 1716454218300395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247944, "dur": 11,
    "args": {
      "External id": 49467,
      "cbid": 211, "correlation": 49467
    }
  },
  {
    "ph": "s", "id": 49467, "pid": 76337, "tid": -914061504, "ts": 1716454218247944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218300413, "dur": 4,
    "args": {
      "External id": 49479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49479,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 49479, "pid": 5, "tid": 7, "ts": 1716454218300413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247965, "dur": 6,
    "args": {
      "External id": 49479,
      "cbid": 211, "correlation": 49479
    }
  },
  {
    "ph": "s", "id": 49479, "pid": 76337, "tid": -914061504, "ts": 1716454218247965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218300418, "dur": 16,
    "args": {
      "External id": 49482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49482,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49482, "pid": 5, "tid": 7, "ts": 1716454218300418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218247992, "dur": 7,
    "args": {
      "External id": 49482,
      "cbid": 211, "correlation": 49482
    }
  },
  {
    "ph": "s", "id": 49482, "pid": 76337, "tid": -914061504, "ts": 1716454218247992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218300435, "dur": 10,
    "args": {
      "External id": 49491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49491,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49491, "pid": 5, "tid": 7, "ts": 1716454218300435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248035, "dur": 10,
    "args": {
      "External id": 49491,
      "cbid": 211, "correlation": 49491
    }
  },
  {
    "ph": "s", "id": 49491, "pid": 76337, "tid": -914061504, "ts": 1716454218248035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248096, "dur": 0,
    "args": {
      "External id": 49501,
      "cbid": 317, "correlation": 49501
    }
  },
  {
    "ph": "f", "id": 49501, "pid": 76337, "tid": -914061504, "ts": 1716454218248096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248097, "dur": 0,
    "args": {
      "External id": 49502,
      "cbid": 203, "correlation": 49502
    }
  },
  {
    "ph": "f", "id": 49502, "pid": 76337, "tid": -914061504, "ts": 1716454218248097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248098, "dur": 0,
    "args": {
      "External id": 49503,
      "cbid": 205, "correlation": 49503
    }
  },
  {
    "ph": "f", "id": 49503, "pid": 76337, "tid": -914061504, "ts": 1716454218248098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218300447, "dur": 11,
    "args": {
      "External id": 49507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49507,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49507, "pid": 5, "tid": 7, "ts": 1716454218300447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248111, "dur": 12,
    "args": {
      "External id": 49507,
      "cbid": 211, "correlation": 49507
    }
  },
  {
    "ph": "s", "id": 49507, "pid": 76337, "tid": -914061504, "ts": 1716454218248111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218300459, "dur": 157,
    "args": {
      "External id": 49509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49509,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49509, "pid": 5, "tid": 7, "ts": 1716454218300459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248126, "dur": 6,
    "args": {
      "External id": 49509,
      "cbid": 211, "correlation": 49509
    }
  },
  {
    "ph": "s", "id": 49509, "pid": 76337, "tid": -914061504, "ts": 1716454218248126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218300619, "dur": 1,
    "args": {
      "External id": 49511,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 49511,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 49511, "pid": 5, "tid": 7, "ts": 1716454218300619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248137, "dur": 8,
    "args": {
      "External id": 49511,
      "cbid": 51, "correlation": 49511
    }
  },
  {
    "ph": "s", "id": 49511, "pid": 76337, "tid": -914061504, "ts": 1716454218248137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218300622, "dur": 637,
    "args": {
      "External id": 49512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49512,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49512, "pid": 5, "tid": 7, "ts": 1716454218300622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248146, "dur": 6,
    "args": {
      "External id": 49512,
      "cbid": 211, "correlation": 49512
    }
  },
  {
    "ph": "s", "id": 49512, "pid": 76337, "tid": -914061504, "ts": 1716454218248146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218301260, "dur": 12,
    "args": {
      "External id": 49514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49514,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49514, "pid": 5, "tid": 7, "ts": 1716454218301260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248156, "dur": 8,
    "args": {
      "External id": 49514,
      "cbid": 211, "correlation": 49514
    }
  },
  {
    "ph": "s", "id": 49514, "pid": 76337, "tid": -914061504, "ts": 1716454218248156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218301274, "dur": 15,
    "args": {
      "External id": 49520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49520, "pid": 5, "tid": 7, "ts": 1716454218301274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248188, "dur": 9,
    "args": {
      "External id": 49520,
      "cbid": 211, "correlation": 49520
    }
  },
  {
    "ph": "s", "id": 49520, "pid": 76337, "tid": -914061504, "ts": 1716454218248188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218301290, "dur": 11,
    "args": {
      "External id": 49528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49528,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49528, "pid": 5, "tid": 7, "ts": 1716454218301290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248220, "dur": 8,
    "args": {
      "External id": 49528,
      "cbid": 211, "correlation": 49528
    }
  },
  {
    "ph": "s", "id": 49528, "pid": 76337, "tid": -914061504, "ts": 1716454218248220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218301303, "dur": 9,
    "args": {
      "External id": 49536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49536,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49536, "pid": 5, "tid": 7, "ts": 1716454218301303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248250, "dur": 8,
    "args": {
      "External id": 49536,
      "cbid": 211, "correlation": 49536
    }
  },
  {
    "ph": "s", "id": 49536, "pid": 76337, "tid": -914061504, "ts": 1716454218248250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218301313, "dur": 17,
    "args": {
      "External id": 49556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49556,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 49556, "pid": 5, "tid": 7, "ts": 1716454218301313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248329, "dur": 13,
    "args": {
      "External id": 49556,
      "cbid": 211, "correlation": 49556
    }
  },
  {
    "ph": "s", "id": 49556, "pid": 76337, "tid": -914061504, "ts": 1716454218248329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218301332, "dur": 4,
    "args": {
      "External id": 49568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49568,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 49568, "pid": 5, "tid": 7, "ts": 1716454218301332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248352, "dur": 6,
    "args": {
      "External id": 49568,
      "cbid": 211, "correlation": 49568
    }
  },
  {
    "ph": "s", "id": 49568, "pid": 76337, "tid": -914061504, "ts": 1716454218248352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218301337, "dur": 17,
    "args": {
      "External id": 49571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49571,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49571, "pid": 5, "tid": 7, "ts": 1716454218301337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248369, "dur": 6,
    "args": {
      "External id": 49571,
      "cbid": 211, "correlation": 49571
    }
  },
  {
    "ph": "s", "id": 49571, "pid": 76337, "tid": -914061504, "ts": 1716454218248369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248426, "dur": 0,
    "args": {
      "External id": 49582,
      "cbid": 317, "correlation": 49582
    }
  },
  {
    "ph": "f", "id": 49582, "pid": 76337, "tid": -914061504, "ts": 1716454218248426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248426, "dur": 0,
    "args": {
      "External id": 49583,
      "cbid": 203, "correlation": 49583
    }
  },
  {
    "ph": "f", "id": 49583, "pid": 76337, "tid": -914061504, "ts": 1716454218248426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248427, "dur": 0,
    "args": {
      "External id": 49584,
      "cbid": 205, "correlation": 49584
    }
  },
  {
    "ph": "f", "id": 49584, "pid": 76337, "tid": -914061504, "ts": 1716454218248427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218301355, "dur": 11,
    "args": {
      "External id": 49588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49588,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49588, "pid": 5, "tid": 7, "ts": 1716454218301355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248445, "dur": 12,
    "args": {
      "External id": 49588,
      "cbid": 211, "correlation": 49588
    }
  },
  {
    "ph": "s", "id": 49588, "pid": 76337, "tid": -914061504, "ts": 1716454218248445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218301367, "dur": 4,
    "args": {
      "External id": 49590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49590, "pid": 5, "tid": 7, "ts": 1716454218301367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248461, "dur": 5,
    "args": {
      "External id": 49590,
      "cbid": 211, "correlation": 49590
    }
  },
  {
    "ph": "s", "id": 49590, "pid": 76337, "tid": -914061504, "ts": 1716454218248461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248470, "dur": 0,
    "args": {
      "External id": 49591,
      "cbid": 51, "correlation": 49591
    }
  },
  {
    "ph": "s", "id": 49591, "pid": 76337, "tid": -914061504, "ts": 1716454218248470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218301372, "dur": 91,
    "args": {
      "External id": 49592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49592,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 49592, "pid": 5, "tid": 7, "ts": 1716454218301372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248470, "dur": 5,
    "args": {
      "External id": 49592,
      "cbid": 211, "correlation": 49592
    }
  },
  {
    "ph": "s", "id": 49592, "pid": 76337, "tid": -914061504, "ts": 1716454218248470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218301464, "dur": 15,
    "args": {
      "External id": 49597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49597,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49597, "pid": 5, "tid": 7, "ts": 1716454218301464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248498, "dur": 8,
    "args": {
      "External id": 49597,
      "cbid": 211, "correlation": 49597
    }
  },
  {
    "ph": "s", "id": 49597, "pid": 76337, "tid": -914061504, "ts": 1716454218248498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218301480, "dur": 81,
    "args": {
      "External id": 49606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49606,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49606, "pid": 5, "tid": 7, "ts": 1716454218301480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248581, "dur": 15,
    "args": {
      "External id": 49606,
      "cbid": 211, "correlation": 49606
    }
  },
  {
    "ph": "s", "id": 49606, "pid": 76337, "tid": -914061504, "ts": 1716454218248581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218301563, "dur": 30,
    "args": {
      "External id": 49628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49628,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49628, "pid": 5, "tid": 7, "ts": 1716454218301563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248638, "dur": 11,
    "args": {
      "External id": 49628,
      "cbid": 211, "correlation": 49628
    }
  },
  {
    "ph": "s", "id": 49628, "pid": 76337, "tid": -914061504, "ts": 1716454218248638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248728, "dur": 1,
    "args": {
      "External id": 49639,
      "cbid": 251, "correlation": 49639
    }
  },
  {
    "ph": "f", "id": 49639, "pid": 76337, "tid": -914061504, "ts": 1716454218248728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218301594, "dur": 139,
    "args": {
      "External id": 49640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49640,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49640, "pid": 5, "tid": 7, "ts": 1716454218301594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248734, "dur": 13,
    "args": {
      "External id": 49640,
      "cbid": 211, "correlation": 49640
    }
  },
  {
    "ph": "s", "id": 49640, "pid": 76337, "tid": -914061504, "ts": 1716454218248734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248802, "dur": 1,
    "args": {
      "External id": 49651,
      "cbid": 251, "correlation": 49651
    }
  },
  {
    "ph": "f", "id": 49651, "pid": 76337, "tid": -914061504, "ts": 1716454218248802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218301735, "dur": 153,
    "args": {
      "External id": 49652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49652,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49652, "pid": 5, "tid": 7, "ts": 1716454218301735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248807, "dur": 11,
    "args": {
      "External id": 49652,
      "cbid": 211, "correlation": 49652
    }
  },
  {
    "ph": "s", "id": 49652, "pid": 76337, "tid": -914061504, "ts": 1716454218248807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248874, "dur": 1,
    "args": {
      "External id": 49663,
      "cbid": 251, "correlation": 49663
    }
  },
  {
    "ph": "f", "id": 49663, "pid": 76337, "tid": -914061504, "ts": 1716454218248874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218301889, "dur": 135,
    "args": {
      "External id": 49664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49664,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49664, "pid": 5, "tid": 7, "ts": 1716454218301889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248878, "dur": 11,
    "args": {
      "External id": 49664,
      "cbid": 211, "correlation": 49664
    }
  },
  {
    "ph": "s", "id": 49664, "pid": 76337, "tid": -914061504, "ts": 1716454218248878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218302026, "dur": 328,
    "args": {
      "External id": 49689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49689,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49689, "pid": 5, "tid": 7, "ts": 1716454218302026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218248963, "dur": 21,
    "args": {
      "External id": 49689,
      "cbid": 211, "correlation": 49689
    }
  },
  {
    "ph": "s", "id": 49689, "pid": 76337, "tid": -914061504, "ts": 1716454218248963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249074, "dur": 1,
    "args": {
      "External id": 49707,
      "cbid": 251, "correlation": 49707
    }
  },
  {
    "ph": "f", "id": 49707, "pid": 76337, "tid": -914061504, "ts": 1716454218249074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218302355, "dur": 164,
    "args": {
      "External id": 49709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49709,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49709, "pid": 5, "tid": 7, "ts": 1716454218302355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249080, "dur": 13,
    "args": {
      "External id": 49709,
      "cbid": 211, "correlation": 49709
    }
  },
  {
    "ph": "s", "id": 49709, "pid": 76337, "tid": -914061504, "ts": 1716454218249080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218302521, "dur": 19,
    "args": {
      "External id": 49717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49717,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49717, "pid": 5, "tid": 7, "ts": 1716454218302521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249150, "dur": 12,
    "args": {
      "External id": 49717,
      "cbid": 211, "correlation": 49717
    }
  },
  {
    "ph": "s", "id": 49717, "pid": 76337, "tid": -914061504, "ts": 1716454218249150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218302541, "dur": 28,
    "args": {
      "External id": 49725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49725,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49725, "pid": 5, "tid": 7, "ts": 1716454218302541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249189, "dur": 8,
    "args": {
      "External id": 49725,
      "cbid": 211, "correlation": 49725
    }
  },
  {
    "ph": "s", "id": 49725, "pid": 76337, "tid": -914061504, "ts": 1716454218249189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218302570, "dur": 18,
    "args": {
      "External id": 49736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49736,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49736, "pid": 5, "tid": 7, "ts": 1716454218302570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249264, "dur": 13,
    "args": {
      "External id": 49736,
      "cbid": 211, "correlation": 49736
    }
  },
  {
    "ph": "s", "id": 49736, "pid": 76337, "tid": -914061504, "ts": 1716454218249264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218302590, "dur": 15,
    "args": {
      "External id": 49758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49758,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49758, "pid": 5, "tid": 7, "ts": 1716454218302590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249297, "dur": 7,
    "args": {
      "External id": 49758,
      "cbid": 211, "correlation": 49758
    }
  },
  {
    "ph": "s", "id": 49758, "pid": 76337, "tid": -914061504, "ts": 1716454218249297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249381, "dur": 1,
    "args": {
      "External id": 49769,
      "cbid": 251, "correlation": 49769
    }
  },
  {
    "ph": "f", "id": 49769, "pid": 76337, "tid": -914061504, "ts": 1716454218249381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218302606, "dur": 85,
    "args": {
      "External id": 49770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49770,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49770, "pid": 5, "tid": 7, "ts": 1716454218302606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249387, "dur": 14,
    "args": {
      "External id": 49770,
      "cbid": 211, "correlation": 49770
    }
  },
  {
    "ph": "s", "id": 49770, "pid": 76337, "tid": -914061504, "ts": 1716454218249387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249457, "dur": 1,
    "args": {
      "External id": 49781,
      "cbid": 251, "correlation": 49781
    }
  },
  {
    "ph": "f", "id": 49781, "pid": 76337, "tid": -914061504, "ts": 1716454218249457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249461, "dur": 0,
    "args": {
      "External id": 49782,
      "cbid": 251, "correlation": 49782
    }
  },
  {
    "ph": "f", "id": 49782, "pid": 76337, "tid": -914061504, "ts": 1716454218249461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218302693, "dur": 12,
    "args": {
      "External id": 49783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49783,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49783, "pid": 5, "tid": 7, "ts": 1716454218302693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249463, "dur": 11,
    "args": {
      "External id": 49783,
      "cbid": 211, "correlation": 49783
    }
  },
  {
    "ph": "s", "id": 49783, "pid": 76337, "tid": -914061504, "ts": 1716454218249463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218302706, "dur": 6,
    "args": {
      "External id": 49785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49785,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49785, "pid": 5, "tid": 7, "ts": 1716454218302706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249476, "dur": 6,
    "args": {
      "External id": 49785,
      "cbid": 211, "correlation": 49785
    }
  },
  {
    "ph": "s", "id": 49785, "pid": 76337, "tid": -914061504, "ts": 1716454218249476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249533, "dur": 1,
    "args": {
      "External id": 49796,
      "cbid": 251, "correlation": 49796
    }
  },
  {
    "ph": "f", "id": 49796, "pid": 76337, "tid": -914061504, "ts": 1716454218249533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249537, "dur": 0,
    "args": {
      "External id": 49797,
      "cbid": 251, "correlation": 49797
    }
  },
  {
    "ph": "f", "id": 49797, "pid": 76337, "tid": -914061504, "ts": 1716454218249537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218302713, "dur": 9,
    "args": {
      "External id": 49798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49798,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49798, "pid": 5, "tid": 7, "ts": 1716454218302713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249538, "dur": 12,
    "args": {
      "External id": 49798,
      "cbid": 211, "correlation": 49798
    }
  },
  {
    "ph": "s", "id": 49798, "pid": 76337, "tid": -914061504, "ts": 1716454218249538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218302723, "dur": 3,
    "args": {
      "External id": 49800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49800,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49800, "pid": 5, "tid": 7, "ts": 1716454218302723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249551, "dur": 6,
    "args": {
      "External id": 49800,
      "cbid": 211, "correlation": 49800
    }
  },
  {
    "ph": "s", "id": 49800, "pid": 76337, "tid": -914061504, "ts": 1716454218249551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218302728, "dur": 53,
    "args": {
      "External id": 49825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49825,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49825, "pid": 5, "tid": 7, "ts": 1716454218302728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249631, "dur": 12,
    "args": {
      "External id": 49825,
      "cbid": 211, "correlation": 49825
    }
  },
  {
    "ph": "s", "id": 49825, "pid": 76337, "tid": -914061504, "ts": 1716454218249631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249730, "dur": 1,
    "args": {
      "External id": 49843,
      "cbid": 251, "correlation": 49843
    }
  },
  {
    "ph": "f", "id": 49843, "pid": 76337, "tid": -914061504, "ts": 1716454218249730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218302782, "dur": 88,
    "args": {
      "External id": 49845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49845,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49845, "pid": 5, "tid": 7, "ts": 1716454218302782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249736, "dur": 14,
    "args": {
      "External id": 49845,
      "cbid": 211, "correlation": 49845
    }
  },
  {
    "ph": "s", "id": 49845, "pid": 76337, "tid": -914061504, "ts": 1716454218249736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218302872, "dur": 9,
    "args": {
      "External id": 49853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49853,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49853, "pid": 5, "tid": 7, "ts": 1716454218302872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249805, "dur": 12,
    "args": {
      "External id": 49853,
      "cbid": 211, "correlation": 49853
    }
  },
  {
    "ph": "s", "id": 49853, "pid": 76337, "tid": -914061504, "ts": 1716454218249805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218302883, "dur": 20,
    "args": {
      "External id": 49861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49861, "pid": 5, "tid": 7, "ts": 1716454218302883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249847, "dur": 9,
    "args": {
      "External id": 49861,
      "cbid": 211, "correlation": 49861
    }
  },
  {
    "ph": "s", "id": 49861, "pid": 76337, "tid": -914061504, "ts": 1716454218249847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218302904, "dur": 17,
    "args": {
      "External id": 49883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49883,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49883, "pid": 5, "tid": 7, "ts": 1716454218302904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249898, "dur": 10,
    "args": {
      "External id": 49883,
      "cbid": 211, "correlation": 49883
    }
  },
  {
    "ph": "s", "id": 49883, "pid": 76337, "tid": -914061504, "ts": 1716454218249898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218249996, "dur": 1,
    "args": {
      "External id": 49899,
      "cbid": 251, "correlation": 49899
    }
  },
  {
    "ph": "f", "id": 49899, "pid": 76337, "tid": -914061504, "ts": 1716454218249996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250001, "dur": 0,
    "args": {
      "External id": 49901,
      "cbid": 251, "correlation": 49901
    }
  },
  {
    "ph": "f", "id": 49901, "pid": 76337, "tid": -914061504, "ts": 1716454218250001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218302923, "dur": 486,
    "args": {
      "External id": 49902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49902,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 49902, "pid": 5, "tid": 7, "ts": 1716454218302923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250004, "dur": 14,
    "args": {
      "External id": 49902,
      "cbid": 211, "correlation": 49902
    }
  },
  {
    "ph": "s", "id": 49902, "pid": 76337, "tid": -914061504, "ts": 1716454218250004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218303411, "dur": 65,
    "args": {
      "External id": 49910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49910,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49910, "pid": 5, "tid": 7, "ts": 1716454218303411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250072, "dur": 13,
    "args": {
      "External id": 49910,
      "cbid": 211, "correlation": 49910
    }
  },
  {
    "ph": "s", "id": 49910, "pid": 76337, "tid": -914061504, "ts": 1716454218250072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218303477, "dur": 66,
    "args": {
      "External id": 49918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49918,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49918, "pid": 5, "tid": 7, "ts": 1716454218303477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250104, "dur": 8,
    "args": {
      "External id": 49918,
      "cbid": 211, "correlation": 49918
    }
  },
  {
    "ph": "s", "id": 49918, "pid": 76337, "tid": -914061504, "ts": 1716454218250104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250184, "dur": 1,
    "args": {
      "External id": 49934,
      "cbid": 251, "correlation": 49934
    }
  },
  {
    "ph": "f", "id": 49934, "pid": 76337, "tid": -914061504, "ts": 1716454218250184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218303545, "dur": 1,
    "args": {
      "External id": 49936,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 49936,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 49936, "pid": 5, "tid": 7, "ts": 1716454218303545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250189, "dur": 9,
    "args": {
      "External id": 49936,
      "cbid": 51, "correlation": 49936
    }
  },
  {
    "ph": "s", "id": 49936, "pid": 76337, "tid": -914061504, "ts": 1716454218250189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218303549, "dur": 266,
    "args": {
      "External id": 49937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49937,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49937, "pid": 5, "tid": 7, "ts": 1716454218303549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250201, "dur": 11,
    "args": {
      "External id": 49937,
      "cbid": 211, "correlation": 49937
    }
  },
  {
    "ph": "s", "id": 49937, "pid": 76337, "tid": -914061504, "ts": 1716454218250201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218303816, "dur": 14,
    "args": {
      "External id": 49945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49945,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49945, "pid": 5, "tid": 7, "ts": 1716454218303816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250243, "dur": 10,
    "args": {
      "External id": 49945,
      "cbid": 211, "correlation": 49945
    }
  },
  {
    "ph": "s", "id": 49945, "pid": 76337, "tid": -914061504, "ts": 1716454218250243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218303831, "dur": 36,
    "args": {
      "External id": 49956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49956, "pid": 5, "tid": 7, "ts": 1716454218303831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250312, "dur": 13,
    "args": {
      "External id": 49956,
      "cbid": 211, "correlation": 49956
    }
  },
  {
    "ph": "s", "id": 49956, "pid": 76337, "tid": -914061504, "ts": 1716454218250312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250379, "dur": 0,
    "args": {
      "External id": 49968,
      "cbid": 317, "correlation": 49968
    }
  },
  {
    "ph": "f", "id": 49968, "pid": 76337, "tid": -914061504, "ts": 1716454218250379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250380, "dur": 0,
    "args": {
      "External id": 49969,
      "cbid": 203, "correlation": 49969
    }
  },
  {
    "ph": "f", "id": 49969, "pid": 76337, "tid": -914061504, "ts": 1716454218250380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250380, "dur": 0,
    "args": {
      "External id": 49970,
      "cbid": 205, "correlation": 49970
    }
  },
  {
    "ph": "f", "id": 49970, "pid": 76337, "tid": -914061504, "ts": 1716454218250380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218303868, "dur": 13,
    "args": {
      "External id": 49974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49974,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49974, "pid": 5, "tid": 7, "ts": 1716454218303868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250396, "dur": 12,
    "args": {
      "External id": 49974,
      "cbid": 211, "correlation": 49974
    }
  },
  {
    "ph": "s", "id": 49974, "pid": 76337, "tid": -914061504, "ts": 1716454218250396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218303882, "dur": 4,
    "args": {
      "External id": 49976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49976,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 49976, "pid": 5, "tid": 7, "ts": 1716454218303882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250412, "dur": 6,
    "args": {
      "External id": 49976,
      "cbid": 211, "correlation": 49976
    }
  },
  {
    "ph": "s", "id": 49976, "pid": 76337, "tid": -914061504, "ts": 1716454218250412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250421, "dur": 0,
    "args": {
      "External id": 49977,
      "cbid": 51, "correlation": 49977
    }
  },
  {
    "ph": "s", "id": 49977, "pid": 76337, "tid": -914061504, "ts": 1716454218250421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218303887, "dur": 96,
    "args": {
      "External id": 49978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49978,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 49978, "pid": 5, "tid": 7, "ts": 1716454218303887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250422, "dur": 5,
    "args": {
      "External id": 49978,
      "cbid": 211, "correlation": 49978
    }
  },
  {
    "ph": "s", "id": 49978, "pid": 76337, "tid": -914061504, "ts": 1716454218250422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218303985, "dur": 16,
    "args": {
      "External id": 49983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49983, "pid": 5, "tid": 7, "ts": 1716454218303985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250449, "dur": 8,
    "args": {
      "External id": 49983,
      "cbid": 211, "correlation": 49983
    }
  },
  {
    "ph": "s", "id": 49983, "pid": 76337, "tid": -914061504, "ts": 1716454218250449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218304002, "dur": 11,
    "args": {
      "External id": 49991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 49991,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 49991, "pid": 5, "tid": 7, "ts": 1716454218304002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250480, "dur": 9,
    "args": {
      "External id": 49991,
      "cbid": 211, "correlation": 49991
    }
  },
  {
    "ph": "s", "id": 49991, "pid": 76337, "tid": -914061504, "ts": 1716454218250480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250552, "dur": 0,
    "args": {
      "External id": 50001,
      "cbid": 317, "correlation": 50001
    }
  },
  {
    "ph": "f", "id": 50001, "pid": 76337, "tid": -914061504, "ts": 1716454218250552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250553, "dur": 0,
    "args": {
      "External id": 50002,
      "cbid": 203, "correlation": 50002
    }
  },
  {
    "ph": "f", "id": 50002, "pid": 76337, "tid": -914061504, "ts": 1716454218250553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250554, "dur": 0,
    "args": {
      "External id": 50003,
      "cbid": 205, "correlation": 50003
    }
  },
  {
    "ph": "f", "id": 50003, "pid": 76337, "tid": -914061504, "ts": 1716454218250554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304015, "dur": 11,
    "args": {
      "External id": 50007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50007,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50007, "pid": 5, "tid": 7, "ts": 1716454218304015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250569, "dur": 12,
    "args": {
      "External id": 50007,
      "cbid": 211, "correlation": 50007
    }
  },
  {
    "ph": "s", "id": 50007, "pid": 76337, "tid": -914061504, "ts": 1716454218250569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304028, "dur": 157,
    "args": {
      "External id": 50009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50009,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50009, "pid": 5, "tid": 7, "ts": 1716454218304028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250583, "dur": 5,
    "args": {
      "External id": 50009,
      "cbid": 211, "correlation": 50009
    }
  },
  {
    "ph": "s", "id": 50009, "pid": 76337, "tid": -914061504, "ts": 1716454218250583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218304187, "dur": 1,
    "args": {
      "External id": 50011,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50011,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 50011, "pid": 5, "tid": 7, "ts": 1716454218304187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250595, "dur": 6,
    "args": {
      "External id": 50011,
      "cbid": 51, "correlation": 50011
    }
  },
  {
    "ph": "s", "id": 50011, "pid": 76337, "tid": -914061504, "ts": 1716454218250595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218304191, "dur": 195,
    "args": {
      "External id": 50012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50012,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 50012, "pid": 5, "tid": 7, "ts": 1716454218304191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250602, "dur": 7,
    "args": {
      "External id": 50012,
      "cbid": 211, "correlation": 50012
    }
  },
  {
    "ph": "s", "id": 50012, "pid": 76337, "tid": -914061504, "ts": 1716454218250602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304387, "dur": 6,
    "args": {
      "External id": 50014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50014, "pid": 5, "tid": 7, "ts": 1716454218304387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250614, "dur": 5,
    "args": {
      "External id": 50014,
      "cbid": 211, "correlation": 50014
    }
  },
  {
    "ph": "s", "id": 50014, "pid": 76337, "tid": -914061504, "ts": 1716454218250614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218304395, "dur": 6,
    "args": {
      "External id": 50020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50020,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50020, "pid": 5, "tid": 7, "ts": 1716454218304395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250642, "dur": 12,
    "args": {
      "External id": 50020,
      "cbid": 211, "correlation": 50020
    }
  },
  {
    "ph": "s", "id": 50020, "pid": 76337, "tid": -914061504, "ts": 1716454218250642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304402, "dur": 11,
    "args": {
      "External id": 50040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50040,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50040, "pid": 5, "tid": 7, "ts": 1716454218304402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250738, "dur": 13,
    "args": {
      "External id": 50040,
      "cbid": 211, "correlation": 50040
    }
  },
  {
    "ph": "s", "id": 50040, "pid": 76337, "tid": -914061504, "ts": 1716454218250738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218304414, "dur": 4,
    "args": {
      "External id": 50052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50052,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50052, "pid": 5, "tid": 7, "ts": 1716454218304414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250761, "dur": 6,
    "args": {
      "External id": 50052,
      "cbid": 211, "correlation": 50052
    }
  },
  {
    "ph": "s", "id": 50052, "pid": 76337, "tid": -914061504, "ts": 1716454218250761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218304420, "dur": 9,
    "args": {
      "External id": 50055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50055, "pid": 5, "tid": 7, "ts": 1716454218304420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250780, "dur": 7,
    "args": {
      "External id": 50055,
      "cbid": 211, "correlation": 50055
    }
  },
  {
    "ph": "s", "id": 50055, "pid": 76337, "tid": -914061504, "ts": 1716454218250780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218304430, "dur": 5,
    "args": {
      "External id": 50064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50064,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50064, "pid": 5, "tid": 7, "ts": 1716454218304430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250820, "dur": 10,
    "args": {
      "External id": 50064,
      "cbid": 211, "correlation": 50064
    }
  },
  {
    "ph": "s", "id": 50064, "pid": 76337, "tid": -914061504, "ts": 1716454218250820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250872, "dur": 0,
    "args": {
      "External id": 50074,
      "cbid": 317, "correlation": 50074
    }
  },
  {
    "ph": "f", "id": 50074, "pid": 76337, "tid": -914061504, "ts": 1716454218250872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250873, "dur": 0,
    "args": {
      "External id": 50075,
      "cbid": 203, "correlation": 50075
    }
  },
  {
    "ph": "f", "id": 50075, "pid": 76337, "tid": -914061504, "ts": 1716454218250873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250874, "dur": 0,
    "args": {
      "External id": 50076,
      "cbid": 205, "correlation": 50076
    }
  },
  {
    "ph": "f", "id": 50076, "pid": 76337, "tid": -914061504, "ts": 1716454218250874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304436, "dur": 5,
    "args": {
      "External id": 50080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50080,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50080, "pid": 5, "tid": 7, "ts": 1716454218304436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250893, "dur": 13,
    "args": {
      "External id": 50080,
      "cbid": 211, "correlation": 50080
    }
  },
  {
    "ph": "s", "id": 50080, "pid": 76337, "tid": -914061504, "ts": 1716454218250893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304442, "dur": 158,
    "args": {
      "External id": 50082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50082, "pid": 5, "tid": 7, "ts": 1716454218304442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250908, "dur": 5,
    "args": {
      "External id": 50082,
      "cbid": 211, "correlation": 50082
    }
  },
  {
    "ph": "s", "id": 50082, "pid": 76337, "tid": -914061504, "ts": 1716454218250908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218304602, "dur": 1,
    "args": {
      "External id": 50084,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50084,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 50084, "pid": 5, "tid": 7, "ts": 1716454218304602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250919, "dur": 6,
    "args": {
      "External id": 50084,
      "cbid": 51, "correlation": 50084
    }
  },
  {
    "ph": "s", "id": 50084, "pid": 76337, "tid": -914061504, "ts": 1716454218250919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218304606, "dur": 262,
    "args": {
      "External id": 50085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50085,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50085, "pid": 5, "tid": 7, "ts": 1716454218304606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250926, "dur": 6,
    "args": {
      "External id": 50085,
      "cbid": 211, "correlation": 50085
    }
  },
  {
    "ph": "s", "id": 50085, "pid": 76337, "tid": -914061504, "ts": 1716454218250926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304869, "dur": 6,
    "args": {
      "External id": 50087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50087,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50087, "pid": 5, "tid": 7, "ts": 1716454218304869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250939, "dur": 5,
    "args": {
      "External id": 50087,
      "cbid": 211, "correlation": 50087
    }
  },
  {
    "ph": "s", "id": 50087, "pid": 76337, "tid": -914061504, "ts": 1716454218250939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218304877, "dur": 6,
    "args": {
      "External id": 50093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50093, "pid": 5, "tid": 7, "ts": 1716454218304877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218250969, "dur": 16,
    "args": {
      "External id": 50093,
      "cbid": 211, "correlation": 50093
    }
  },
  {
    "ph": "s", "id": 50093, "pid": 76337, "tid": -914061504, "ts": 1716454218250969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218304884, "dur": 3,
    "args": {
      "External id": 50101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50101,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 50101, "pid": 5, "tid": 7, "ts": 1716454218304884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251021, "dur": 10,
    "args": {
      "External id": 50101,
      "cbid": 211, "correlation": 50101
    }
  },
  {
    "ph": "s", "id": 50101, "pid": 76337, "tid": -914061504, "ts": 1716454218251021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251087, "dur": 1,
    "args": {
      "External id": 50117,
      "cbid": 251, "correlation": 50117
    }
  },
  {
    "ph": "f", "id": 50117, "pid": 76337, "tid": -914061504, "ts": 1716454218251087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251092, "dur": 0,
    "args": {
      "External id": 50119,
      "cbid": 251, "correlation": 50119
    }
  },
  {
    "ph": "f", "id": 50119, "pid": 76337, "tid": -914061504, "ts": 1716454218251092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218304889, "dur": 13,
    "args": {
      "External id": 50120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50120,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50120, "pid": 5, "tid": 7, "ts": 1716454218304889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251094, "dur": 12,
    "args": {
      "External id": 50120,
      "cbid": 211, "correlation": 50120
    }
  },
  {
    "ph": "s", "id": 50120, "pid": 76337, "tid": -914061504, "ts": 1716454218251094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218304903, "dur": 5,
    "args": {
      "External id": 50122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50122,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50122, "pid": 5, "tid": 7, "ts": 1716454218304903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251108, "dur": 5,
    "args": {
      "External id": 50122,
      "cbid": 211, "correlation": 50122
    }
  },
  {
    "ph": "s", "id": 50122, "pid": 76337, "tid": -914061504, "ts": 1716454218251108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218304909, "dur": 5,
    "args": {
      "External id": 50132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50132, "pid": 5, "tid": 7, "ts": 1716454218304909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251165, "dur": 12,
    "args": {
      "External id": 50132,
      "cbid": 211, "correlation": 50132
    }
  },
  {
    "ph": "s", "id": 50132, "pid": 76337, "tid": -914061504, "ts": 1716454218251165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304916, "dur": 9,
    "args": {
      "External id": 50152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50152,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50152, "pid": 5, "tid": 7, "ts": 1716454218304916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251232, "dur": 10,
    "args": {
      "External id": 50152,
      "cbid": 211, "correlation": 50152
    }
  },
  {
    "ph": "s", "id": 50152, "pid": 76337, "tid": -914061504, "ts": 1716454218251232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218304926, "dur": 3,
    "args": {
      "External id": 50164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50164,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50164, "pid": 5, "tid": 7, "ts": 1716454218304926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251255, "dur": 7,
    "args": {
      "External id": 50164,
      "cbid": 211, "correlation": 50164
    }
  },
  {
    "ph": "s", "id": 50164, "pid": 76337, "tid": -914061504, "ts": 1716454218251255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218304931, "dur": 8,
    "args": {
      "External id": 50167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50167, "pid": 5, "tid": 7, "ts": 1716454218304931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251275, "dur": 6,
    "args": {
      "External id": 50167,
      "cbid": 211, "correlation": 50167
    }
  },
  {
    "ph": "s", "id": 50167, "pid": 76337, "tid": -914061504, "ts": 1716454218251275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218304941, "dur": 4,
    "args": {
      "External id": 50176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50176,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50176, "pid": 5, "tid": 7, "ts": 1716454218304941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251315, "dur": 10,
    "args": {
      "External id": 50176,
      "cbid": 211, "correlation": 50176
    }
  },
  {
    "ph": "s", "id": 50176, "pid": 76337, "tid": -914061504, "ts": 1716454218251315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251378, "dur": 0,
    "args": {
      "External id": 50186,
      "cbid": 317, "correlation": 50186
    }
  },
  {
    "ph": "f", "id": 50186, "pid": 76337, "tid": -914061504, "ts": 1716454218251378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251379, "dur": 0,
    "args": {
      "External id": 50187,
      "cbid": 203, "correlation": 50187
    }
  },
  {
    "ph": "f", "id": 50187, "pid": 76337, "tid": -914061504, "ts": 1716454218251379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251380, "dur": 0,
    "args": {
      "External id": 50188,
      "cbid": 205, "correlation": 50188
    }
  },
  {
    "ph": "f", "id": 50188, "pid": 76337, "tid": -914061504, "ts": 1716454218251380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304946, "dur": 5,
    "args": {
      "External id": 50192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50192, "pid": 5, "tid": 7, "ts": 1716454218304946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251394, "dur": 12,
    "args": {
      "External id": 50192,
      "cbid": 211, "correlation": 50192
    }
  },
  {
    "ph": "s", "id": 50192, "pid": 76337, "tid": -914061504, "ts": 1716454218251394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218304952, "dur": 157,
    "args": {
      "External id": 50194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50194,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50194, "pid": 5, "tid": 7, "ts": 1716454218304952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251408, "dur": 5,
    "args": {
      "External id": 50194,
      "cbid": 211, "correlation": 50194
    }
  },
  {
    "ph": "s", "id": 50194, "pid": 76337, "tid": -914061504, "ts": 1716454218251408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218305112, "dur": 1,
    "args": {
      "External id": 50196,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50196,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 50196, "pid": 5, "tid": 7, "ts": 1716454218305112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251419, "dur": 6,
    "args": {
      "External id": 50196,
      "cbid": 51, "correlation": 50196
    }
  },
  {
    "ph": "s", "id": 50196, "pid": 76337, "tid": -914061504, "ts": 1716454218251419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218305115, "dur": 252,
    "args": {
      "External id": 50197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50197,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50197, "pid": 5, "tid": 7, "ts": 1716454218305115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251427, "dur": 6,
    "args": {
      "External id": 50197,
      "cbid": 211, "correlation": 50197
    }
  },
  {
    "ph": "s", "id": 50197, "pid": 76337, "tid": -914061504, "ts": 1716454218251427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305368, "dur": 6,
    "args": {
      "External id": 50199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50199,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50199, "pid": 5, "tid": 7, "ts": 1716454218305368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251436, "dur": 5,
    "args": {
      "External id": 50199,
      "cbid": 211, "correlation": 50199
    }
  },
  {
    "ph": "s", "id": 50199, "pid": 76337, "tid": -914061504, "ts": 1716454218251436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218305375, "dur": 6,
    "args": {
      "External id": 50205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50205, "pid": 5, "tid": 7, "ts": 1716454218305375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251464, "dur": 9,
    "args": {
      "External id": 50205,
      "cbid": 211, "correlation": 50205
    }
  },
  {
    "ph": "s", "id": 50205, "pid": 76337, "tid": -914061504, "ts": 1716454218251464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218305383, "dur": 5,
    "args": {
      "External id": 50213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50213,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50213, "pid": 5, "tid": 7, "ts": 1716454218305383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251497, "dur": 8,
    "args": {
      "External id": 50213,
      "cbid": 211, "correlation": 50213
    }
  },
  {
    "ph": "s", "id": 50213, "pid": 76337, "tid": -914061504, "ts": 1716454218251497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218305389, "dur": 4,
    "args": {
      "External id": 50221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50221,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50221, "pid": 5, "tid": 7, "ts": 1716454218305389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251529, "dur": 9,
    "args": {
      "External id": 50221,
      "cbid": 211, "correlation": 50221
    }
  },
  {
    "ph": "s", "id": 50221, "pid": 76337, "tid": -914061504, "ts": 1716454218251529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305394, "dur": 9,
    "args": {
      "External id": 50241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50241,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50241, "pid": 5, "tid": 7, "ts": 1716454218305394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251604, "dur": 11,
    "args": {
      "External id": 50241,
      "cbid": 211, "correlation": 50241
    }
  },
  {
    "ph": "s", "id": 50241, "pid": 76337, "tid": -914061504, "ts": 1716454218251604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218305405, "dur": 4,
    "args": {
      "External id": 50253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50253,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50253, "pid": 5, "tid": 7, "ts": 1716454218305405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251625, "dur": 6,
    "args": {
      "External id": 50253,
      "cbid": 211, "correlation": 50253
    }
  },
  {
    "ph": "s", "id": 50253, "pid": 76337, "tid": -914061504, "ts": 1716454218251625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218305409, "dur": 6,
    "args": {
      "External id": 50256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50256, "pid": 5, "tid": 7, "ts": 1716454218305409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251643, "dur": 6,
    "args": {
      "External id": 50256,
      "cbid": 211, "correlation": 50256
    }
  },
  {
    "ph": "s", "id": 50256, "pid": 76337, "tid": -914061504, "ts": 1716454218251643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218305417, "dur": 4,
    "args": {
      "External id": 50265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50265,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50265, "pid": 5, "tid": 7, "ts": 1716454218305417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251681, "dur": 10,
    "args": {
      "External id": 50265,
      "cbid": 211, "correlation": 50265
    }
  },
  {
    "ph": "s", "id": 50265, "pid": 76337, "tid": -914061504, "ts": 1716454218251681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251732, "dur": 0,
    "args": {
      "External id": 50275,
      "cbid": 317, "correlation": 50275
    }
  },
  {
    "ph": "f", "id": 50275, "pid": 76337, "tid": -914061504, "ts": 1716454218251732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251733, "dur": 0,
    "args": {
      "External id": 50276,
      "cbid": 203, "correlation": 50276
    }
  },
  {
    "ph": "f", "id": 50276, "pid": 76337, "tid": -914061504, "ts": 1716454218251733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251733, "dur": 0,
    "args": {
      "External id": 50277,
      "cbid": 205, "correlation": 50277
    }
  },
  {
    "ph": "f", "id": 50277, "pid": 76337, "tid": -914061504, "ts": 1716454218251733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305423, "dur": 5,
    "args": {
      "External id": 50281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50281,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50281, "pid": 5, "tid": 7, "ts": 1716454218305423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251747, "dur": 12,
    "args": {
      "External id": 50281,
      "cbid": 211, "correlation": 50281
    }
  },
  {
    "ph": "s", "id": 50281, "pid": 76337, "tid": -914061504, "ts": 1716454218251747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305429, "dur": 158,
    "args": {
      "External id": 50283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50283,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50283, "pid": 5, "tid": 7, "ts": 1716454218305429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251761, "dur": 5,
    "args": {
      "External id": 50283,
      "cbid": 211, "correlation": 50283
    }
  },
  {
    "ph": "s", "id": 50283, "pid": 76337, "tid": -914061504, "ts": 1716454218251761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218305589, "dur": 1,
    "args": {
      "External id": 50285,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50285,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 50285, "pid": 5, "tid": 7, "ts": 1716454218305589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251771, "dur": 6,
    "args": {
      "External id": 50285,
      "cbid": 51, "correlation": 50285
    }
  },
  {
    "ph": "s", "id": 50285, "pid": 76337, "tid": -914061504, "ts": 1716454218251771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218305593, "dur": 251,
    "args": {
      "External id": 50286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50286,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50286, "pid": 5, "tid": 7, "ts": 1716454218305593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251779, "dur": 8,
    "args": {
      "External id": 50286,
      "cbid": 211, "correlation": 50286
    }
  },
  {
    "ph": "s", "id": 50286, "pid": 76337, "tid": -914061504, "ts": 1716454218251779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305845, "dur": 6,
    "args": {
      "External id": 50288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50288, "pid": 5, "tid": 7, "ts": 1716454218305845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251791, "dur": 6,
    "args": {
      "External id": 50288,
      "cbid": 211, "correlation": 50288
    }
  },
  {
    "ph": "s", "id": 50288, "pid": 76337, "tid": -914061504, "ts": 1716454218251791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218305852, "dur": 6,
    "args": {
      "External id": 50294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50294, "pid": 5, "tid": 7, "ts": 1716454218305852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251819, "dur": 8,
    "args": {
      "External id": 50294,
      "cbid": 211, "correlation": 50294
    }
  },
  {
    "ph": "s", "id": 50294, "pid": 76337, "tid": -914061504, "ts": 1716454218251819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218305859, "dur": 3,
    "args": {
      "External id": 50302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50302,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 50302, "pid": 5, "tid": 7, "ts": 1716454218305859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251864, "dur": 9,
    "args": {
      "External id": 50302,
      "cbid": 211, "correlation": 50302
    }
  },
  {
    "ph": "s", "id": 50302, "pid": 76337, "tid": -914061504, "ts": 1716454218251864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251927, "dur": 1,
    "args": {
      "External id": 50318,
      "cbid": 251, "correlation": 50318
    }
  },
  {
    "ph": "f", "id": 50318, "pid": 76337, "tid": -914061504, "ts": 1716454218251927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251933, "dur": 0,
    "args": {
      "External id": 50320,
      "cbid": 251, "correlation": 50320
    }
  },
  {
    "ph": "f", "id": 50320, "pid": 76337, "tid": -914061504, "ts": 1716454218251933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218305864, "dur": 10,
    "args": {
      "External id": 50321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50321,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50321, "pid": 5, "tid": 7, "ts": 1716454218305864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251935, "dur": 11,
    "args": {
      "External id": 50321,
      "cbid": 211, "correlation": 50321
    }
  },
  {
    "ph": "s", "id": 50321, "pid": 76337, "tid": -914061504, "ts": 1716454218251935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218305875, "dur": 3,
    "args": {
      "External id": 50323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50323,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50323, "pid": 5, "tid": 7, "ts": 1716454218305875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218251948, "dur": 5,
    "args": {
      "External id": 50323,
      "cbid": 211, "correlation": 50323
    }
  },
  {
    "ph": "s", "id": 50323, "pid": 76337, "tid": -914061504, "ts": 1716454218251948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218305880, "dur": 5,
    "args": {
      "External id": 50333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50333, "pid": 5, "tid": 7, "ts": 1716454218305880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252013, "dur": 12,
    "args": {
      "External id": 50333,
      "cbid": 211, "correlation": 50333
    }
  },
  {
    "ph": "s", "id": 50333, "pid": 76337, "tid": -914061504, "ts": 1716454218252013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305886, "dur": 9,
    "args": {
      "External id": 50353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50353,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50353, "pid": 5, "tid": 7, "ts": 1716454218305886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252079, "dur": 10,
    "args": {
      "External id": 50353,
      "cbid": 211, "correlation": 50353
    }
  },
  {
    "ph": "s", "id": 50353, "pid": 76337, "tid": -914061504, "ts": 1716454218252079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218305897, "dur": 4,
    "args": {
      "External id": 50365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50365,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50365, "pid": 5, "tid": 7, "ts": 1716454218305897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252103, "dur": 6,
    "args": {
      "External id": 50365,
      "cbid": 211, "correlation": 50365
    }
  },
  {
    "ph": "s", "id": 50365, "pid": 76337, "tid": -914061504, "ts": 1716454218252103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218305902, "dur": 6,
    "args": {
      "External id": 50368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50368,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50368, "pid": 5, "tid": 7, "ts": 1716454218305902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252122, "dur": 6,
    "args": {
      "External id": 50368,
      "cbid": 211, "correlation": 50368
    }
  },
  {
    "ph": "s", "id": 50368, "pid": 76337, "tid": -914061504, "ts": 1716454218252122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218305910, "dur": 4,
    "args": {
      "External id": 50377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50377,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50377, "pid": 5, "tid": 7, "ts": 1716454218305910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252162, "dur": 9,
    "args": {
      "External id": 50377,
      "cbid": 211, "correlation": 50377
    }
  },
  {
    "ph": "s", "id": 50377, "pid": 76337, "tid": -914061504, "ts": 1716454218252162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252224, "dur": 0,
    "args": {
      "External id": 50387,
      "cbid": 317, "correlation": 50387
    }
  },
  {
    "ph": "f", "id": 50387, "pid": 76337, "tid": -914061504, "ts": 1716454218252224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252225, "dur": 0,
    "args": {
      "External id": 50388,
      "cbid": 203, "correlation": 50388
    }
  },
  {
    "ph": "f", "id": 50388, "pid": 76337, "tid": -914061504, "ts": 1716454218252225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252226, "dur": 0,
    "args": {
      "External id": 50389,
      "cbid": 205, "correlation": 50389
    }
  },
  {
    "ph": "f", "id": 50389, "pid": 76337, "tid": -914061504, "ts": 1716454218252226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305916, "dur": 5,
    "args": {
      "External id": 50393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50393,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50393, "pid": 5, "tid": 7, "ts": 1716454218305916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252240, "dur": 12,
    "args": {
      "External id": 50393,
      "cbid": 211, "correlation": 50393
    }
  },
  {
    "ph": "s", "id": 50393, "pid": 76337, "tid": -914061504, "ts": 1716454218252240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218305922, "dur": 157,
    "args": {
      "External id": 50395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50395,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50395, "pid": 5, "tid": 7, "ts": 1716454218305922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252255, "dur": 6,
    "args": {
      "External id": 50395,
      "cbid": 211, "correlation": 50395
    }
  },
  {
    "ph": "s", "id": 50395, "pid": 76337, "tid": -914061504, "ts": 1716454218252255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218306081, "dur": 1,
    "args": {
      "External id": 50397,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50397,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 50397, "pid": 5, "tid": 7, "ts": 1716454218306081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252266, "dur": 6,
    "args": {
      "External id": 50397,
      "cbid": 51, "correlation": 50397
    }
  },
  {
    "ph": "s", "id": 50397, "pid": 76337, "tid": -914061504, "ts": 1716454218252266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218306084, "dur": 251,
    "args": {
      "External id": 50398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50398,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50398, "pid": 5, "tid": 7, "ts": 1716454218306084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252273, "dur": 6,
    "args": {
      "External id": 50398,
      "cbid": 211, "correlation": 50398
    }
  },
  {
    "ph": "s", "id": 50398, "pid": 76337, "tid": -914061504, "ts": 1716454218252273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306337, "dur": 5,
    "args": {
      "External id": 50400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50400,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50400, "pid": 5, "tid": 7, "ts": 1716454218306337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252283, "dur": 5,
    "args": {
      "External id": 50400,
      "cbid": 211, "correlation": 50400
    }
  },
  {
    "ph": "s", "id": 50400, "pid": 76337, "tid": -914061504, "ts": 1716454218252283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218306344, "dur": 6,
    "args": {
      "External id": 50406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50406,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50406, "pid": 5, "tid": 7, "ts": 1716454218306344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252313, "dur": 8,
    "args": {
      "External id": 50406,
      "cbid": 211, "correlation": 50406
    }
  },
  {
    "ph": "s", "id": 50406, "pid": 76337, "tid": -914061504, "ts": 1716454218252313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218306351, "dur": 5,
    "args": {
      "External id": 50414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50414,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50414, "pid": 5, "tid": 7, "ts": 1716454218306351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252344, "dur": 8,
    "args": {
      "External id": 50414,
      "cbid": 211, "correlation": 50414
    }
  },
  {
    "ph": "s", "id": 50414, "pid": 76337, "tid": -914061504, "ts": 1716454218252344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218306357, "dur": 4,
    "args": {
      "External id": 50422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50422,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50422, "pid": 5, "tid": 7, "ts": 1716454218306357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252377, "dur": 8,
    "args": {
      "External id": 50422,
      "cbid": 211, "correlation": 50422
    }
  },
  {
    "ph": "s", "id": 50422, "pid": 76337, "tid": -914061504, "ts": 1716454218252377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306363, "dur": 9,
    "args": {
      "External id": 50442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50442,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50442, "pid": 5, "tid": 7, "ts": 1716454218306363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252481, "dur": 12,
    "args": {
      "External id": 50442,
      "cbid": 211, "correlation": 50442
    }
  },
  {
    "ph": "s", "id": 50442, "pid": 76337, "tid": -914061504, "ts": 1716454218252481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218306373, "dur": 3,
    "args": {
      "External id": 50454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50454,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50454, "pid": 5, "tid": 7, "ts": 1716454218306373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252504, "dur": 6,
    "args": {
      "External id": 50454,
      "cbid": 211, "correlation": 50454
    }
  },
  {
    "ph": "s", "id": 50454, "pid": 76337, "tid": -914061504, "ts": 1716454218252504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218306378, "dur": 6,
    "args": {
      "External id": 50457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50457, "pid": 5, "tid": 7, "ts": 1716454218306378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252522, "dur": 6,
    "args": {
      "External id": 50457,
      "cbid": 211, "correlation": 50457
    }
  },
  {
    "ph": "s", "id": 50457, "pid": 76337, "tid": -914061504, "ts": 1716454218252522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218306386, "dur": 4,
    "args": {
      "External id": 50466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50466,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50466, "pid": 5, "tid": 7, "ts": 1716454218306386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252559, "dur": 10,
    "args": {
      "External id": 50466,
      "cbid": 211, "correlation": 50466
    }
  },
  {
    "ph": "s", "id": 50466, "pid": 76337, "tid": -914061504, "ts": 1716454218252559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252612, "dur": 0,
    "args": {
      "External id": 50476,
      "cbid": 317, "correlation": 50476
    }
  },
  {
    "ph": "f", "id": 50476, "pid": 76337, "tid": -914061504, "ts": 1716454218252612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252613, "dur": 0,
    "args": {
      "External id": 50477,
      "cbid": 203, "correlation": 50477
    }
  },
  {
    "ph": "f", "id": 50477, "pid": 76337, "tid": -914061504, "ts": 1716454218252613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252614, "dur": 0,
    "args": {
      "External id": 50478,
      "cbid": 205, "correlation": 50478
    }
  },
  {
    "ph": "f", "id": 50478, "pid": 76337, "tid": -914061504, "ts": 1716454218252614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306391, "dur": 5,
    "args": {
      "External id": 50482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50482,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50482, "pid": 5, "tid": 7, "ts": 1716454218306391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252627, "dur": 11,
    "args": {
      "External id": 50482,
      "cbid": 211, "correlation": 50482
    }
  },
  {
    "ph": "s", "id": 50482, "pid": 76337, "tid": -914061504, "ts": 1716454218252627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306398, "dur": 157,
    "args": {
      "External id": 50484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50484,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50484, "pid": 5, "tid": 7, "ts": 1716454218306398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252641, "dur": 5,
    "args": {
      "External id": 50484,
      "cbid": 211, "correlation": 50484
    }
  },
  {
    "ph": "s", "id": 50484, "pid": 76337, "tid": -914061504, "ts": 1716454218252641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218306557, "dur": 1,
    "args": {
      "External id": 50486,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50486,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 50486, "pid": 5, "tid": 7, "ts": 1716454218306557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252651, "dur": 9,
    "args": {
      "External id": 50486,
      "cbid": 51, "correlation": 50486
    }
  },
  {
    "ph": "s", "id": 50486, "pid": 76337, "tid": -914061504, "ts": 1716454218252651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218306561, "dur": 251,
    "args": {
      "External id": 50487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50487,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50487, "pid": 5, "tid": 7, "ts": 1716454218306561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252662, "dur": 6,
    "args": {
      "External id": 50487,
      "cbid": 211, "correlation": 50487
    }
  },
  {
    "ph": "s", "id": 50487, "pid": 76337, "tid": -914061504, "ts": 1716454218252662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306813, "dur": 6,
    "args": {
      "External id": 50489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50489,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50489, "pid": 5, "tid": 7, "ts": 1716454218306813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252672, "dur": 6,
    "args": {
      "External id": 50489,
      "cbid": 211, "correlation": 50489
    }
  },
  {
    "ph": "s", "id": 50489, "pid": 76337, "tid": -914061504, "ts": 1716454218252672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218306820, "dur": 6,
    "args": {
      "External id": 50495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50495,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50495, "pid": 5, "tid": 7, "ts": 1716454218306820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252700, "dur": 8,
    "args": {
      "External id": 50495,
      "cbid": 211, "correlation": 50495
    }
  },
  {
    "ph": "s", "id": 50495, "pid": 76337, "tid": -914061504, "ts": 1716454218252700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218306828, "dur": 3,
    "args": {
      "External id": 50503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50503,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 50503, "pid": 5, "tid": 7, "ts": 1716454218306828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252744, "dur": 9,
    "args": {
      "External id": 50503,
      "cbid": 211, "correlation": 50503
    }
  },
  {
    "ph": "s", "id": 50503, "pid": 76337, "tid": -914061504, "ts": 1716454218252744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252805, "dur": 1,
    "args": {
      "External id": 50519,
      "cbid": 251, "correlation": 50519
    }
  },
  {
    "ph": "f", "id": 50519, "pid": 76337, "tid": -914061504, "ts": 1716454218252805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252810, "dur": 0,
    "args": {
      "External id": 50521,
      "cbid": 251, "correlation": 50521
    }
  },
  {
    "ph": "f", "id": 50521, "pid": 76337, "tid": -914061504, "ts": 1716454218252810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218306832, "dur": 10,
    "args": {
      "External id": 50522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50522,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50522, "pid": 5, "tid": 7, "ts": 1716454218306832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252812, "dur": 11,
    "args": {
      "External id": 50522,
      "cbid": 211, "correlation": 50522
    }
  },
  {
    "ph": "s", "id": 50522, "pid": 76337, "tid": -914061504, "ts": 1716454218252812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218306844, "dur": 3,
    "args": {
      "External id": 50524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50524,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50524, "pid": 5, "tid": 7, "ts": 1716454218306844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252825, "dur": 5,
    "args": {
      "External id": 50524,
      "cbid": 211, "correlation": 50524
    }
  },
  {
    "ph": "s", "id": 50524, "pid": 76337, "tid": -914061504, "ts": 1716454218252825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218306848, "dur": 5,
    "args": {
      "External id": 50534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50534, "pid": 5, "tid": 7, "ts": 1716454218306848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252882, "dur": 13,
    "args": {
      "External id": 50534,
      "cbid": 211, "correlation": 50534
    }
  },
  {
    "ph": "s", "id": 50534, "pid": 76337, "tid": -914061504, "ts": 1716454218252882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306855, "dur": 9,
    "args": {
      "External id": 50554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50554,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50554, "pid": 5, "tid": 7, "ts": 1716454218306855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252948, "dur": 13,
    "args": {
      "External id": 50554,
      "cbid": 211, "correlation": 50554
    }
  },
  {
    "ph": "s", "id": 50554, "pid": 76337, "tid": -914061504, "ts": 1716454218252948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218306866, "dur": 4,
    "args": {
      "External id": 50566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50566,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50566, "pid": 5, "tid": 7, "ts": 1716454218306866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252971, "dur": 15,
    "args": {
      "External id": 50566,
      "cbid": 211, "correlation": 50566
    }
  },
  {
    "ph": "s", "id": 50566, "pid": 76337, "tid": -914061504, "ts": 1716454218252971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218306871, "dur": 6,
    "args": {
      "External id": 50569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50569, "pid": 5, "tid": 7, "ts": 1716454218306871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218252999, "dur": 7,
    "args": {
      "External id": 50569,
      "cbid": 211, "correlation": 50569
    }
  },
  {
    "ph": "s", "id": 50569, "pid": 76337, "tid": -914061504, "ts": 1716454218252999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218306878, "dur": 4,
    "args": {
      "External id": 50578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50578,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50578, "pid": 5, "tid": 7, "ts": 1716454218306878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253042, "dur": 10,
    "args": {
      "External id": 50578,
      "cbid": 211, "correlation": 50578
    }
  },
  {
    "ph": "s", "id": 50578, "pid": 76337, "tid": -914061504, "ts": 1716454218253042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253105, "dur": 0,
    "args": {
      "External id": 50588,
      "cbid": 317, "correlation": 50588
    }
  },
  {
    "ph": "f", "id": 50588, "pid": 76337, "tid": -914061504, "ts": 1716454218253105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253106, "dur": 0,
    "args": {
      "External id": 50589,
      "cbid": 203, "correlation": 50589
    }
  },
  {
    "ph": "f", "id": 50589, "pid": 76337, "tid": -914061504, "ts": 1716454218253106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253107, "dur": 0,
    "args": {
      "External id": 50590,
      "cbid": 205, "correlation": 50590
    }
  },
  {
    "ph": "f", "id": 50590, "pid": 76337, "tid": -914061504, "ts": 1716454218253107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306884, "dur": 5,
    "args": {
      "External id": 50594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50594,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50594, "pid": 5, "tid": 7, "ts": 1716454218306884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253120, "dur": 12,
    "args": {
      "External id": 50594,
      "cbid": 211, "correlation": 50594
    }
  },
  {
    "ph": "s", "id": 50594, "pid": 76337, "tid": -914061504, "ts": 1716454218253120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218306890, "dur": 157,
    "args": {
      "External id": 50596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50596, "pid": 5, "tid": 7, "ts": 1716454218306890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253135, "dur": 6,
    "args": {
      "External id": 50596,
      "cbid": 211, "correlation": 50596
    }
  },
  {
    "ph": "s", "id": 50596, "pid": 76337, "tid": -914061504, "ts": 1716454218253135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218307049, "dur": 1,
    "args": {
      "External id": 50598,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50598,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 50598, "pid": 5, "tid": 7, "ts": 1716454218307049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253146, "dur": 6,
    "args": {
      "External id": 50598,
      "cbid": 51, "correlation": 50598
    }
  },
  {
    "ph": "s", "id": 50598, "pid": 76337, "tid": -914061504, "ts": 1716454218253146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218307052, "dur": 250,
    "args": {
      "External id": 50599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50599,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50599, "pid": 5, "tid": 7, "ts": 1716454218307052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253153, "dur": 6,
    "args": {
      "External id": 50599,
      "cbid": 211, "correlation": 50599
    }
  },
  {
    "ph": "s", "id": 50599, "pid": 76337, "tid": -914061504, "ts": 1716454218253153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307304, "dur": 6,
    "args": {
      "External id": 50601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50601,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50601, "pid": 5, "tid": 7, "ts": 1716454218307304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253163, "dur": 5,
    "args": {
      "External id": 50601,
      "cbid": 211, "correlation": 50601
    }
  },
  {
    "ph": "s", "id": 50601, "pid": 76337, "tid": -914061504, "ts": 1716454218253163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218307311, "dur": 6,
    "args": {
      "External id": 50607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50607, "pid": 5, "tid": 7, "ts": 1716454218307311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253191, "dur": 9,
    "args": {
      "External id": 50607,
      "cbid": 211, "correlation": 50607
    }
  },
  {
    "ph": "s", "id": 50607, "pid": 76337, "tid": -914061504, "ts": 1716454218253191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218307318, "dur": 5,
    "args": {
      "External id": 50615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50615,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50615, "pid": 5, "tid": 7, "ts": 1716454218307318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253223, "dur": 11,
    "args": {
      "External id": 50615,
      "cbid": 211, "correlation": 50615
    }
  },
  {
    "ph": "s", "id": 50615, "pid": 76337, "tid": -914061504, "ts": 1716454218253223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218307325, "dur": 4,
    "args": {
      "External id": 50623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50623,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50623, "pid": 5, "tid": 7, "ts": 1716454218307325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253256, "dur": 9,
    "args": {
      "External id": 50623,
      "cbid": 211, "correlation": 50623
    }
  },
  {
    "ph": "s", "id": 50623, "pid": 76337, "tid": -914061504, "ts": 1716454218253256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307330, "dur": 9,
    "args": {
      "External id": 50643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50643,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 50643, "pid": 5, "tid": 7, "ts": 1716454218307330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253400, "dur": 14,
    "args": {
      "External id": 50643,
      "cbid": 211, "correlation": 50643
    }
  },
  {
    "ph": "s", "id": 50643, "pid": 76337, "tid": -914061504, "ts": 1716454218253400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218307341, "dur": 4,
    "args": {
      "External id": 50655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50655,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 50655, "pid": 5, "tid": 7, "ts": 1716454218307341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253423, "dur": 7,
    "args": {
      "External id": 50655,
      "cbid": 211, "correlation": 50655
    }
  },
  {
    "ph": "s", "id": 50655, "pid": 76337, "tid": -914061504, "ts": 1716454218253423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218307345, "dur": 6,
    "args": {
      "External id": 50658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50658,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50658, "pid": 5, "tid": 7, "ts": 1716454218307345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253442, "dur": 7,
    "args": {
      "External id": 50658,
      "cbid": 211, "correlation": 50658
    }
  },
  {
    "ph": "s", "id": 50658, "pid": 76337, "tid": -914061504, "ts": 1716454218253442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253502, "dur": 0,
    "args": {
      "External id": 50669,
      "cbid": 317, "correlation": 50669
    }
  },
  {
    "ph": "f", "id": 50669, "pid": 76337, "tid": -914061504, "ts": 1716454218253502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253503, "dur": 0,
    "args": {
      "External id": 50670,
      "cbid": 203, "correlation": 50670
    }
  },
  {
    "ph": "f", "id": 50670, "pid": 76337, "tid": -914061504, "ts": 1716454218253503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253504, "dur": 0,
    "args": {
      "External id": 50671,
      "cbid": 205, "correlation": 50671
    }
  },
  {
    "ph": "f", "id": 50671, "pid": 76337, "tid": -914061504, "ts": 1716454218253504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307353, "dur": 5,
    "args": {
      "External id": 50675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50675,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50675, "pid": 5, "tid": 7, "ts": 1716454218307353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253520, "dur": 12,
    "args": {
      "External id": 50675,
      "cbid": 211, "correlation": 50675
    }
  },
  {
    "ph": "s", "id": 50675, "pid": 76337, "tid": -914061504, "ts": 1716454218253520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218307359, "dur": 36,
    "args": {
      "External id": 50677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50677,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 50677, "pid": 5, "tid": 7, "ts": 1716454218307359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253539, "dur": 10,
    "args": {
      "External id": 50677,
      "cbid": 211, "correlation": 50677
    }
  },
  {
    "ph": "s", "id": 50677, "pid": 76337, "tid": -914061504, "ts": 1716454218253539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307396, "dur": 5,
    "args": {
      "External id": 50679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50679,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50679, "pid": 5, "tid": 7, "ts": 1716454218307396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253554, "dur": 5,
    "args": {
      "External id": 50679,
      "cbid": 211, "correlation": 50679
    }
  },
  {
    "ph": "s", "id": 50679, "pid": 76337, "tid": -914061504, "ts": 1716454218253554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218307402, "dur": 6,
    "args": {
      "External id": 50685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50685,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50685, "pid": 5, "tid": 7, "ts": 1716454218307402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253580, "dur": 11,
    "args": {
      "External id": 50685,
      "cbid": 211, "correlation": 50685
    }
  },
  {
    "ph": "s", "id": 50685, "pid": 76337, "tid": -914061504, "ts": 1716454218253580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218307409, "dur": 19,
    "args": {
      "External id": 50694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50694,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50694, "pid": 5, "tid": 7, "ts": 1716454218307409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253666, "dur": 14,
    "args": {
      "External id": 50694,
      "cbid": 211, "correlation": 50694
    }
  },
  {
    "ph": "s", "id": 50694, "pid": 76337, "tid": -914061504, "ts": 1716454218253666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307430, "dur": 10,
    "args": {
      "External id": 50716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50716,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 50716, "pid": 5, "tid": 7, "ts": 1716454218307430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253723, "dur": 10,
    "args": {
      "External id": 50716,
      "cbid": 211, "correlation": 50716
    }
  },
  {
    "ph": "s", "id": 50716, "pid": 76337, "tid": -914061504, "ts": 1716454218253723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253813, "dur": 2,
    "args": {
      "External id": 50727,
      "cbid": 251, "correlation": 50727
    }
  },
  {
    "ph": "f", "id": 50727, "pid": 76337, "tid": -914061504, "ts": 1716454218253813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253818, "dur": 0,
    "args": {
      "External id": 50728,
      "cbid": 251, "correlation": 50728
    }
  },
  {
    "ph": "f", "id": 50728, "pid": 76337, "tid": -914061504, "ts": 1716454218253818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218307441, "dur": 52,
    "args": {
      "External id": 50729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50729,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 50729, "pid": 5, "tid": 7, "ts": 1716454218307441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253821, "dur": 14,
    "args": {
      "External id": 50729,
      "cbid": 211, "correlation": 50729
    }
  },
  {
    "ph": "s", "id": 50729, "pid": 76337, "tid": -914061504, "ts": 1716454218253821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253892, "dur": 1,
    "args": {
      "External id": 50740,
      "cbid": 251, "correlation": 50740
    }
  },
  {
    "ph": "f", "id": 50740, "pid": 76337, "tid": -914061504, "ts": 1716454218253892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253896, "dur": 0,
    "args": {
      "External id": 50741,
      "cbid": 251, "correlation": 50741
    }
  },
  {
    "ph": "f", "id": 50741, "pid": 76337, "tid": -914061504, "ts": 1716454218253896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218307495, "dur": 52,
    "args": {
      "External id": 50742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50742,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 50742, "pid": 5, "tid": 7, "ts": 1716454218307495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253898, "dur": 12,
    "args": {
      "External id": 50742,
      "cbid": 211, "correlation": 50742
    }
  },
  {
    "ph": "s", "id": 50742, "pid": 76337, "tid": -914061504, "ts": 1716454218253898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253965, "dur": 1,
    "args": {
      "External id": 50753,
      "cbid": 251, "correlation": 50753
    }
  },
  {
    "ph": "f", "id": 50753, "pid": 76337, "tid": -914061504, "ts": 1716454218253965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253968, "dur": 0,
    "args": {
      "External id": 50754,
      "cbid": 251, "correlation": 50754
    }
  },
  {
    "ph": "f", "id": 50754, "pid": 76337, "tid": -914061504, "ts": 1716454218253968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218307548, "dur": 52,
    "args": {
      "External id": 50755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50755,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 50755, "pid": 5, "tid": 7, "ts": 1716454218307548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218253970, "dur": 21,
    "args": {
      "External id": 50755,
      "cbid": 211, "correlation": 50755
    }
  },
  {
    "ph": "s", "id": 50755, "pid": 76337, "tid": -914061504, "ts": 1716454218253970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218307601, "dur": 55,
    "args": {
      "External id": 50780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50780,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50780, "pid": 5, "tid": 7, "ts": 1716454218307601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254068, "dur": 14,
    "args": {
      "External id": 50780,
      "cbid": 211, "correlation": 50780
    }
  },
  {
    "ph": "s", "id": 50780, "pid": 76337, "tid": -914061504, "ts": 1716454218254068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254169, "dur": 1,
    "args": {
      "External id": 50798,
      "cbid": 251, "correlation": 50798
    }
  },
  {
    "ph": "f", "id": 50798, "pid": 76337, "tid": -914061504, "ts": 1716454218254169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218307657, "dur": 62,
    "args": {
      "External id": 50800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50800,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 50800, "pid": 5, "tid": 7, "ts": 1716454218307657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254175, "dur": 13,
    "args": {
      "External id": 50800,
      "cbid": 211, "correlation": 50800
    }
  },
  {
    "ph": "s", "id": 50800, "pid": 76337, "tid": -914061504, "ts": 1716454218254175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218307721, "dur": 6,
    "args": {
      "External id": 50808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50808,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50808, "pid": 5, "tid": 7, "ts": 1716454218307721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254245, "dur": 12,
    "args": {
      "External id": 50808,
      "cbid": 211, "correlation": 50808
    }
  },
  {
    "ph": "s", "id": 50808, "pid": 76337, "tid": -914061504, "ts": 1716454218254245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218307728, "dur": 7,
    "args": {
      "External id": 50816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50816,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 50816, "pid": 5, "tid": 7, "ts": 1716454218307728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254284, "dur": 9,
    "args": {
      "External id": 50816,
      "cbid": 211, "correlation": 50816
    }
  },
  {
    "ph": "s", "id": 50816, "pid": 76337, "tid": -914061504, "ts": 1716454218254284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218307736, "dur": 7,
    "args": {
      "External id": 50827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50827,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50827, "pid": 5, "tid": 7, "ts": 1716454218307736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254358, "dur": 13,
    "args": {
      "External id": 50827,
      "cbid": 211, "correlation": 50827
    }
  },
  {
    "ph": "s", "id": 50827, "pid": 76337, "tid": -914061504, "ts": 1716454218254358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307744, "dur": 8,
    "args": {
      "External id": 50849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50849,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 50849, "pid": 5, "tid": 7, "ts": 1716454218307744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254394, "dur": 8,
    "args": {
      "External id": 50849,
      "cbid": 211, "correlation": 50849
    }
  },
  {
    "ph": "s", "id": 50849, "pid": 76337, "tid": -914061504, "ts": 1716454218254394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254481, "dur": 2,
    "args": {
      "External id": 50860,
      "cbid": 251, "correlation": 50860
    }
  },
  {
    "ph": "f", "id": 50860, "pid": 76337, "tid": -914061504, "ts": 1716454218254481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218307754, "dur": 1,
    "args": {
      "External id": 50861,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50861,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 50861, "pid": 5, "tid": 7, "ts": 1716454218307754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254487, "dur": 10,
    "args": {
      "External id": 50861,
      "cbid": 51, "correlation": 50861
    }
  },
  {
    "ph": "s", "id": 50861, "pid": 76337, "tid": -914061504, "ts": 1716454218254487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218307758, "dur": 36,
    "args": {
      "External id": 50862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50862,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 50862, "pid": 5, "tid": 7, "ts": 1716454218307758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254499, "dur": 13,
    "args": {
      "External id": 50862,
      "cbid": 211, "correlation": 50862
    }
  },
  {
    "ph": "s", "id": 50862, "pid": 76337, "tid": -914061504, "ts": 1716454218254499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254571, "dur": 1,
    "args": {
      "External id": 50873,
      "cbid": 251, "correlation": 50873
    }
  },
  {
    "ph": "f", "id": 50873, "pid": 76337, "tid": -914061504, "ts": 1716454218254571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254575, "dur": 0,
    "args": {
      "External id": 50874,
      "cbid": 251, "correlation": 50874
    }
  },
  {
    "ph": "f", "id": 50874, "pid": 76337, "tid": -914061504, "ts": 1716454218254575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218307795, "dur": 12,
    "args": {
      "External id": 50875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50875,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50875, "pid": 5, "tid": 7, "ts": 1716454218307795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254576, "dur": 13,
    "args": {
      "External id": 50875,
      "cbid": 211, "correlation": 50875
    }
  },
  {
    "ph": "s", "id": 50875, "pid": 76337, "tid": -914061504, "ts": 1716454218254576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218307808, "dur": 5,
    "args": {
      "External id": 50877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50877,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50877, "pid": 5, "tid": 7, "ts": 1716454218307808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254591, "dur": 6,
    "args": {
      "External id": 50877,
      "cbid": 211, "correlation": 50877
    }
  },
  {
    "ph": "s", "id": 50877, "pid": 76337, "tid": -914061504, "ts": 1716454218254591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254649, "dur": 1,
    "args": {
      "External id": 50888,
      "cbid": 251, "correlation": 50888
    }
  },
  {
    "ph": "f", "id": 50888, "pid": 76337, "tid": -914061504, "ts": 1716454218254649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254652, "dur": 0,
    "args": {
      "External id": 50889,
      "cbid": 251, "correlation": 50889
    }
  },
  {
    "ph": "f", "id": 50889, "pid": 76337, "tid": -914061504, "ts": 1716454218254652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218307815, "dur": 8,
    "args": {
      "External id": 50890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50890,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50890, "pid": 5, "tid": 7, "ts": 1716454218307815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254654, "dur": 11,
    "args": {
      "External id": 50890,
      "cbid": 211, "correlation": 50890
    }
  },
  {
    "ph": "s", "id": 50890, "pid": 76337, "tid": -914061504, "ts": 1716454218254654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218307824, "dur": 4,
    "args": {
      "External id": 50892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50892,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50892, "pid": 5, "tid": 7, "ts": 1716454218307824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254667, "dur": 5,
    "args": {
      "External id": 50892,
      "cbid": 211, "correlation": 50892
    }
  },
  {
    "ph": "s", "id": 50892, "pid": 76337, "tid": -914061504, "ts": 1716454218254667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218307829, "dur": 19,
    "args": {
      "External id": 50917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50917,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 50917, "pid": 5, "tid": 7, "ts": 1716454218307829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254748, "dur": 13,
    "args": {
      "External id": 50917,
      "cbid": 211, "correlation": 50917
    }
  },
  {
    "ph": "s", "id": 50917, "pid": 76337, "tid": -914061504, "ts": 1716454218254748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254850, "dur": 2,
    "args": {
      "External id": 50935,
      "cbid": 251, "correlation": 50935
    }
  },
  {
    "ph": "f", "id": 50935, "pid": 76337, "tid": -914061504, "ts": 1716454218254850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218307850, "dur": 1,
    "args": {
      "External id": 50937,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 50937,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3127035830618892
    }
  },
  {
    "ph": "f", "id": 50937, "pid": 5, "tid": 7, "ts": 1716454218307850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254856, "dur": 11,
    "args": {
      "External id": 50937,
      "cbid": 51, "correlation": 50937
    }
  },
  {
    "ph": "s", "id": 50937, "pid": 76337, "tid": -914061504, "ts": 1716454218254856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218307854, "dur": 36,
    "args": {
      "External id": 50938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50938,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 50938, "pid": 5, "tid": 7, "ts": 1716454218307854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254869, "dur": 13,
    "args": {
      "External id": 50938,
      "cbid": 211, "correlation": 50938
    }
  },
  {
    "ph": "s", "id": 50938, "pid": 76337, "tid": -914061504, "ts": 1716454218254869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218307891, "dur": 4,
    "args": {
      "External id": 50946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50946,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50946, "pid": 5, "tid": 7, "ts": 1716454218307891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254939, "dur": 12,
    "args": {
      "External id": 50946,
      "cbid": 211, "correlation": 50946
    }
  },
  {
    "ph": "s", "id": 50946, "pid": 76337, "tid": -914061504, "ts": 1716454218254939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218307896, "dur": 8,
    "args": {
      "External id": 50954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 50954, "pid": 5, "tid": 7, "ts": 1716454218307896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218254990, "dur": 10,
    "args": {
      "External id": 50954,
      "cbid": 211, "correlation": 50954
    }
  },
  {
    "ph": "s", "id": 50954, "pid": 76337, "tid": -914061504, "ts": 1716454218254990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218307905, "dur": 8,
    "args": {
      "External id": 50976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50976,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 50976, "pid": 5, "tid": 7, "ts": 1716454218307905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255044, "dur": 10,
    "args": {
      "External id": 50976,
      "cbid": 211, "correlation": 50976
    }
  },
  {
    "ph": "s", "id": 50976, "pid": 76337, "tid": -914061504, "ts": 1716454218255044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255136, "dur": 1,
    "args": {
      "External id": 50992,
      "cbid": 251, "correlation": 50992
    }
  },
  {
    "ph": "f", "id": 50992, "pid": 76337, "tid": -914061504, "ts": 1716454218255136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255144, "dur": 0,
    "args": {
      "External id": 50994,
      "cbid": 251, "correlation": 50994
    }
  },
  {
    "ph": "f", "id": 50994, "pid": 76337, "tid": -914061504, "ts": 1716454218255144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218307915, "dur": 187,
    "args": {
      "External id": 50995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 50995,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 50995, "pid": 5, "tid": 7, "ts": 1716454218307915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255146, "dur": 13,
    "args": {
      "External id": 50995,
      "cbid": 211, "correlation": 50995
    }
  },
  {
    "ph": "s", "id": 50995, "pid": 76337, "tid": -914061504, "ts": 1716454218255146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308103, "dur": 20,
    "args": {
      "External id": 51003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51003, "pid": 5, "tid": 7, "ts": 1716454218308103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255214, "dur": 13,
    "args": {
      "External id": 51003,
      "cbid": 211, "correlation": 51003
    }
  },
  {
    "ph": "s", "id": 51003, "pid": 76337, "tid": -914061504, "ts": 1716454218255214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308124, "dur": 22,
    "args": {
      "External id": 51011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51011, "pid": 5, "tid": 7, "ts": 1716454218308124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255246, "dur": 8,
    "args": {
      "External id": 51011,
      "cbid": 211, "correlation": 51011
    }
  },
  {
    "ph": "s", "id": 51011, "pid": 76337, "tid": -914061504, "ts": 1716454218255246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255328, "dur": 1,
    "args": {
      "External id": 51027,
      "cbid": 251, "correlation": 51027
    }
  },
  {
    "ph": "f", "id": 51027, "pid": 76337, "tid": -914061504, "ts": 1716454218255328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218308148, "dur": 1,
    "args": {
      "External id": 51029,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51029,
      "bytes": 120, "memory bandwidth (GB/s)": 0.0797872340425532
    }
  },
  {
    "ph": "f", "id": 51029, "pid": 5, "tid": 7, "ts": 1716454218308148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255333, "dur": 8,
    "args": {
      "External id": 51029,
      "cbid": 51, "correlation": 51029
    }
  },
  {
    "ph": "s", "id": 51029, "pid": 76337, "tid": -914061504, "ts": 1716454218255333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218308151, "dur": 108,
    "args": {
      "External id": 51030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51030,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 51030, "pid": 5, "tid": 7, "ts": 1716454218308151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255343, "dur": 11,
    "args": {
      "External id": 51030,
      "cbid": 211, "correlation": 51030
    }
  },
  {
    "ph": "s", "id": 51030, "pid": 76337, "tid": -914061504, "ts": 1716454218255343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218308261, "dur": 5,
    "args": {
      "External id": 51038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51038,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51038, "pid": 5, "tid": 7, "ts": 1716454218308261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255385, "dur": 10,
    "args": {
      "External id": 51038,
      "cbid": 211, "correlation": 51038
    }
  },
  {
    "ph": "s", "id": 51038, "pid": 76337, "tid": -914061504, "ts": 1716454218255385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308268, "dur": 9,
    "args": {
      "External id": 51049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51049, "pid": 5, "tid": 7, "ts": 1716454218308268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255453, "dur": 12,
    "args": {
      "External id": 51049,
      "cbid": 211, "correlation": 51049
    }
  },
  {
    "ph": "s", "id": 51049, "pid": 76337, "tid": -914061504, "ts": 1716454218255453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255519, "dur": 0,
    "args": {
      "External id": 51061,
      "cbid": 317, "correlation": 51061
    }
  },
  {
    "ph": "f", "id": 51061, "pid": 76337, "tid": -914061504, "ts": 1716454218255519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255520, "dur": 0,
    "args": {
      "External id": 51062,
      "cbid": 203, "correlation": 51062
    }
  },
  {
    "ph": "f", "id": 51062, "pid": 76337, "tid": -914061504, "ts": 1716454218255520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255521, "dur": 0,
    "args": {
      "External id": 51063,
      "cbid": 205, "correlation": 51063
    }
  },
  {
    "ph": "f", "id": 51063, "pid": 76337, "tid": -914061504, "ts": 1716454218255521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308278, "dur": 5,
    "args": {
      "External id": 51067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51067,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51067, "pid": 5, "tid": 7, "ts": 1716454218308278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255537, "dur": 15,
    "args": {
      "External id": 51067,
      "cbid": 211, "correlation": 51067
    }
  },
  {
    "ph": "s", "id": 51067, "pid": 76337, "tid": -914061504, "ts": 1716454218255537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218308285, "dur": 36,
    "args": {
      "External id": 51069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51069,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 51069, "pid": 5, "tid": 7, "ts": 1716454218308285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255559, "dur": 8,
    "args": {
      "External id": 51069,
      "cbid": 211, "correlation": 51069
    }
  },
  {
    "ph": "s", "id": 51069, "pid": 76337, "tid": -914061504, "ts": 1716454218255559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308322, "dur": 5,
    "args": {
      "External id": 51071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51071,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51071, "pid": 5, "tid": 7, "ts": 1716454218308322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255571, "dur": 6,
    "args": {
      "External id": 51071,
      "cbid": 211, "correlation": 51071
    }
  },
  {
    "ph": "s", "id": 51071, "pid": 76337, "tid": -914061504, "ts": 1716454218255571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308329, "dur": 7,
    "args": {
      "External id": 51077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51077,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51077, "pid": 5, "tid": 7, "ts": 1716454218308329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255599, "dur": 9,
    "args": {
      "External id": 51077,
      "cbid": 211, "correlation": 51077
    }
  },
  {
    "ph": "s", "id": 51077, "pid": 76337, "tid": -914061504, "ts": 1716454218255599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218308337, "dur": 5,
    "args": {
      "External id": 51085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51085,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51085, "pid": 5, "tid": 7, "ts": 1716454218308337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255632, "dur": 8,
    "args": {
      "External id": 51085,
      "cbid": 211, "correlation": 51085
    }
  },
  {
    "ph": "s", "id": 51085, "pid": 76337, "tid": -914061504, "ts": 1716454218255632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308343, "dur": 10,
    "args": {
      "External id": 51105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51105,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51105, "pid": 5, "tid": 7, "ts": 1716454218308343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255704, "dur": 12,
    "args": {
      "External id": 51105,
      "cbid": 211, "correlation": 51105
    }
  },
  {
    "ph": "s", "id": 51105, "pid": 76337, "tid": -914061504, "ts": 1716454218255704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218308355, "dur": 4,
    "args": {
      "External id": 51117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51117,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 51117, "pid": 5, "tid": 7, "ts": 1716454218308355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255725, "dur": 7,
    "args": {
      "External id": 51117,
      "cbid": 211, "correlation": 51117
    }
  },
  {
    "ph": "s", "id": 51117, "pid": 76337, "tid": -914061504, "ts": 1716454218255725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308361, "dur": 8,
    "args": {
      "External id": 51120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51120, "pid": 5, "tid": 7, "ts": 1716454218308361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255743, "dur": 6,
    "args": {
      "External id": 51120,
      "cbid": 211, "correlation": 51120
    }
  },
  {
    "ph": "s", "id": 51120, "pid": 76337, "tid": -914061504, "ts": 1716454218255743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218308370, "dur": 5,
    "args": {
      "External id": 51129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51129,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51129, "pid": 5, "tid": 7, "ts": 1716454218308370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255782, "dur": 10,
    "args": {
      "External id": 51129,
      "cbid": 211, "correlation": 51129
    }
  },
  {
    "ph": "s", "id": 51129, "pid": 76337, "tid": -914061504, "ts": 1716454218255782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255838, "dur": 0,
    "args": {
      "External id": 51139,
      "cbid": 317, "correlation": 51139
    }
  },
  {
    "ph": "f", "id": 51139, "pid": 76337, "tid": -914061504, "ts": 1716454218255838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255838, "dur": 0,
    "args": {
      "External id": 51140,
      "cbid": 203, "correlation": 51140
    }
  },
  {
    "ph": "f", "id": 51140, "pid": 76337, "tid": -914061504, "ts": 1716454218255838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255839, "dur": 0,
    "args": {
      "External id": 51141,
      "cbid": 205, "correlation": 51141
    }
  },
  {
    "ph": "f", "id": 51141, "pid": 76337, "tid": -914061504, "ts": 1716454218255839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308377, "dur": 5,
    "args": {
      "External id": 51145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51145,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51145, "pid": 5, "tid": 7, "ts": 1716454218308377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255852, "dur": 11,
    "args": {
      "External id": 51145,
      "cbid": 211, "correlation": 51145
    }
  },
  {
    "ph": "s", "id": 51145, "pid": 76337, "tid": -914061504, "ts": 1716454218255852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308383, "dur": 157,
    "args": {
      "External id": 51147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51147,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51147, "pid": 5, "tid": 7, "ts": 1716454218308383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255866, "dur": 5,
    "args": {
      "External id": 51147,
      "cbid": 211, "correlation": 51147
    }
  },
  {
    "ph": "s", "id": 51147, "pid": 76337, "tid": -914061504, "ts": 1716454218255866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218308542, "dur": 1,
    "args": {
      "External id": 51149,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51149,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 51149, "pid": 5, "tid": 7, "ts": 1716454218308542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255877, "dur": 7,
    "args": {
      "External id": 51149,
      "cbid": 51, "correlation": 51149
    }
  },
  {
    "ph": "s", "id": 51149, "pid": 76337, "tid": -914061504, "ts": 1716454218255877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218308546, "dur": 262,
    "args": {
      "External id": 51150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51150,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51150, "pid": 5, "tid": 7, "ts": 1716454218308546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255885, "dur": 6,
    "args": {
      "External id": 51150,
      "cbid": 211, "correlation": 51150
    }
  },
  {
    "ph": "s", "id": 51150, "pid": 76337, "tid": -914061504, "ts": 1716454218255885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308809, "dur": 6,
    "args": {
      "External id": 51152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51152,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51152, "pid": 5, "tid": 7, "ts": 1716454218308809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255895, "dur": 5,
    "args": {
      "External id": 51152,
      "cbid": 211, "correlation": 51152
    }
  },
  {
    "ph": "s", "id": 51152, "pid": 76337, "tid": -914061504, "ts": 1716454218255895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308816, "dur": 6,
    "args": {
      "External id": 51158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51158, "pid": 5, "tid": 7, "ts": 1716454218308816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255922, "dur": 8,
    "args": {
      "External id": 51158,
      "cbid": 211, "correlation": 51158
    }
  },
  {
    "ph": "s", "id": 51158, "pid": 76337, "tid": -914061504, "ts": 1716454218255922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218308824, "dur": 3,
    "args": {
      "External id": 51166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51166,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 51166, "pid": 5, "tid": 7, "ts": 1716454218308824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218255966, "dur": 17,
    "args": {
      "External id": 51166,
      "cbid": 211, "correlation": 51166
    }
  },
  {
    "ph": "s", "id": 51166, "pid": 76337, "tid": -914061504, "ts": 1716454218255966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256039, "dur": 1,
    "args": {
      "External id": 51182,
      "cbid": 251, "correlation": 51182
    }
  },
  {
    "ph": "f", "id": 51182, "pid": 76337, "tid": -914061504, "ts": 1716454218256039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256044, "dur": 0,
    "args": {
      "External id": 51184,
      "cbid": 251, "correlation": 51184
    }
  },
  {
    "ph": "f", "id": 51184, "pid": 76337, "tid": -914061504, "ts": 1716454218256044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218308828, "dur": 13,
    "args": {
      "External id": 51185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51185,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51185, "pid": 5, "tid": 7, "ts": 1716454218308828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256046, "dur": 11,
    "args": {
      "External id": 51185,
      "cbid": 211, "correlation": 51185
    }
  },
  {
    "ph": "s", "id": 51185, "pid": 76337, "tid": -914061504, "ts": 1716454218256046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218308842, "dur": 5,
    "args": {
      "External id": 51187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51187,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51187, "pid": 5, "tid": 7, "ts": 1716454218308842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256059, "dur": 6,
    "args": {
      "External id": 51187,
      "cbid": 211, "correlation": 51187
    }
  },
  {
    "ph": "s", "id": 51187, "pid": 76337, "tid": -914061504, "ts": 1716454218256059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308848, "dur": 5,
    "args": {
      "External id": 51197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51197, "pid": 5, "tid": 7, "ts": 1716454218308848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256120, "dur": 12,
    "args": {
      "External id": 51197,
      "cbid": 211, "correlation": 51197
    }
  },
  {
    "ph": "s", "id": 51197, "pid": 76337, "tid": -914061504, "ts": 1716454218256120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308855, "dur": 9,
    "args": {
      "External id": 51217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51217,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51217, "pid": 5, "tid": 7, "ts": 1716454218308855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256187, "dur": 11,
    "args": {
      "External id": 51217,
      "cbid": 211, "correlation": 51217
    }
  },
  {
    "ph": "s", "id": 51217, "pid": 76337, "tid": -914061504, "ts": 1716454218256187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218308865, "dur": 3,
    "args": {
      "External id": 51229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51229,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 51229, "pid": 5, "tid": 7, "ts": 1716454218308865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256207, "dur": 6,
    "args": {
      "External id": 51229,
      "cbid": 211, "correlation": 51229
    }
  },
  {
    "ph": "s", "id": 51229, "pid": 76337, "tid": -914061504, "ts": 1716454218256207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218308870, "dur": 6,
    "args": {
      "External id": 51232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51232,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51232, "pid": 5, "tid": 7, "ts": 1716454218308870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256226, "dur": 6,
    "args": {
      "External id": 51232,
      "cbid": 211, "correlation": 51232
    }
  },
  {
    "ph": "s", "id": 51232, "pid": 76337, "tid": -914061504, "ts": 1716454218256226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218308878, "dur": 4,
    "args": {
      "External id": 51241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51241,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51241, "pid": 5, "tid": 7, "ts": 1716454218308878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256266, "dur": 10,
    "args": {
      "External id": 51241,
      "cbid": 211, "correlation": 51241
    }
  },
  {
    "ph": "s", "id": 51241, "pid": 76337, "tid": -914061504, "ts": 1716454218256266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256329, "dur": 0,
    "args": {
      "External id": 51251,
      "cbid": 317, "correlation": 51251
    }
  },
  {
    "ph": "f", "id": 51251, "pid": 76337, "tid": -914061504, "ts": 1716454218256329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256330, "dur": 0,
    "args": {
      "External id": 51252,
      "cbid": 203, "correlation": 51252
    }
  },
  {
    "ph": "f", "id": 51252, "pid": 76337, "tid": -914061504, "ts": 1716454218256330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256330, "dur": 0,
    "args": {
      "External id": 51253,
      "cbid": 205, "correlation": 51253
    }
  },
  {
    "ph": "f", "id": 51253, "pid": 76337, "tid": -914061504, "ts": 1716454218256330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308884, "dur": 4,
    "args": {
      "External id": 51257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51257,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51257, "pid": 5, "tid": 7, "ts": 1716454218308884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256345, "dur": 12,
    "args": {
      "External id": 51257,
      "cbid": 211, "correlation": 51257
    }
  },
  {
    "ph": "s", "id": 51257, "pid": 76337, "tid": -914061504, "ts": 1716454218256345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218308889, "dur": 157,
    "args": {
      "External id": 51259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51259,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51259, "pid": 5, "tid": 7, "ts": 1716454218308889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256359, "dur": 5,
    "args": {
      "External id": 51259,
      "cbid": 211, "correlation": 51259
    }
  },
  {
    "ph": "s", "id": 51259, "pid": 76337, "tid": -914061504, "ts": 1716454218256359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218309048, "dur": 1,
    "args": {
      "External id": 51261,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51261,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14714898835070508
    }
  },
  {
    "ph": "f", "id": 51261, "pid": 5, "tid": 7, "ts": 1716454218309048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256370, "dur": 6,
    "args": {
      "External id": 51261,
      "cbid": 51, "correlation": 51261
    }
  },
  {
    "ph": "s", "id": 51261, "pid": 76337, "tid": -914061504, "ts": 1716454218256370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218309052, "dur": 251,
    "args": {
      "External id": 51262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51262,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51262, "pid": 5, "tid": 7, "ts": 1716454218309052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256377, "dur": 9,
    "args": {
      "External id": 51262,
      "cbid": 211, "correlation": 51262
    }
  },
  {
    "ph": "s", "id": 51262, "pid": 76337, "tid": -914061504, "ts": 1716454218256377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218309304, "dur": 6,
    "args": {
      "External id": 51264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51264,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51264, "pid": 5, "tid": 7, "ts": 1716454218309304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256391, "dur": 6,
    "args": {
      "External id": 51264,
      "cbid": 211, "correlation": 51264
    }
  },
  {
    "ph": "s", "id": 51264, "pid": 76337, "tid": -914061504, "ts": 1716454218256391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218309311, "dur": 6,
    "args": {
      "External id": 51270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51270, "pid": 5, "tid": 7, "ts": 1716454218309311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256420, "dur": 8,
    "args": {
      "External id": 51270,
      "cbid": 211, "correlation": 51270
    }
  },
  {
    "ph": "s", "id": 51270, "pid": 76337, "tid": -914061504, "ts": 1716454218256420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218309319, "dur": 4,
    "args": {
      "External id": 51278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51278,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51278, "pid": 5, "tid": 7, "ts": 1716454218309319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256453, "dur": 8,
    "args": {
      "External id": 51278,
      "cbid": 211, "correlation": 51278
    }
  },
  {
    "ph": "s", "id": 51278, "pid": 76337, "tid": -914061504, "ts": 1716454218256453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218309324, "dur": 4,
    "args": {
      "External id": 51286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51286,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51286, "pid": 5, "tid": 7, "ts": 1716454218309324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256482, "dur": 9,
    "args": {
      "External id": 51286,
      "cbid": 211, "correlation": 51286
    }
  },
  {
    "ph": "s", "id": 51286, "pid": 76337, "tid": -914061504, "ts": 1716454218256482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218309330, "dur": 11,
    "args": {
      "External id": 51295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51295,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51295, "pid": 5, "tid": 7, "ts": 1716454218309330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256572, "dur": 13,
    "args": {
      "External id": 51295,
      "cbid": 211, "correlation": 51295
    }
  },
  {
    "ph": "s", "id": 51295, "pid": 76337, "tid": -914061504, "ts": 1716454218256572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218309343, "dur": 12,
    "args": {
      "External id": 51315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51315,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51315, "pid": 5, "tid": 7, "ts": 1716454218309343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256640, "dur": 11,
    "args": {
      "External id": 51315,
      "cbid": 211, "correlation": 51315
    }
  },
  {
    "ph": "s", "id": 51315, "pid": 76337, "tid": -914061504, "ts": 1716454218256640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218309356, "dur": 4,
    "args": {
      "External id": 51327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51327,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51327, "pid": 5, "tid": 7, "ts": 1716454218309356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256662, "dur": 6,
    "args": {
      "External id": 51327,
      "cbid": 211, "correlation": 51327
    }
  },
  {
    "ph": "s", "id": 51327, "pid": 76337, "tid": -914061504, "ts": 1716454218256662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218309361, "dur": 10,
    "args": {
      "External id": 51330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51330,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51330, "pid": 5, "tid": 7, "ts": 1716454218309361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256680, "dur": 10,
    "args": {
      "External id": 51330,
      "cbid": 211, "correlation": 51330
    }
  },
  {
    "ph": "s", "id": 51330, "pid": 76337, "tid": -914061504, "ts": 1716454218256680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218309371, "dur": 6,
    "args": {
      "External id": 51339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51339,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51339, "pid": 5, "tid": 7, "ts": 1716454218309371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256724, "dur": 10,
    "args": {
      "External id": 51339,
      "cbid": 211, "correlation": 51339
    }
  },
  {
    "ph": "s", "id": 51339, "pid": 76337, "tid": -914061504, "ts": 1716454218256724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256776, "dur": 0,
    "args": {
      "External id": 51349,
      "cbid": 317, "correlation": 51349
    }
  },
  {
    "ph": "f", "id": 51349, "pid": 76337, "tid": -914061504, "ts": 1716454218256776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256777, "dur": 0,
    "args": {
      "External id": 51350,
      "cbid": 203, "correlation": 51350
    }
  },
  {
    "ph": "f", "id": 51350, "pid": 76337, "tid": -914061504, "ts": 1716454218256777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256778, "dur": 0,
    "args": {
      "External id": 51351,
      "cbid": 205, "correlation": 51351
    }
  },
  {
    "ph": "f", "id": 51351, "pid": 76337, "tid": -914061504, "ts": 1716454218256778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218309379, "dur": 6,
    "args": {
      "External id": 51355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51355,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51355, "pid": 5, "tid": 7, "ts": 1716454218309379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256793, "dur": 11,
    "args": {
      "External id": 51355,
      "cbid": 211, "correlation": 51355
    }
  },
  {
    "ph": "s", "id": 51355, "pid": 76337, "tid": -914061504, "ts": 1716454218256793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218309386, "dur": 311,
    "args": {
      "External id": 51357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51357,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51357, "pid": 5, "tid": 7, "ts": 1716454218309386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256807, "dur": 6,
    "args": {
      "External id": 51357,
      "cbid": 211, "correlation": 51357
    }
  },
  {
    "ph": "s", "id": 51357, "pid": 76337, "tid": -914061504, "ts": 1716454218256807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218309700, "dur": 1,
    "args": {
      "External id": 51359,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51359,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 51359, "pid": 5, "tid": 7, "ts": 1716454218309700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256819, "dur": 8,
    "args": {
      "External id": 51359,
      "cbid": 51, "correlation": 51359
    }
  },
  {
    "ph": "s", "id": 51359, "pid": 76337, "tid": -914061504, "ts": 1716454218256819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218309703, "dur": 483,
    "args": {
      "External id": 51360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51360,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51360, "pid": 5, "tid": 7, "ts": 1716454218309703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256827, "dur": 6,
    "args": {
      "External id": 51360,
      "cbid": 211, "correlation": 51360
    }
  },
  {
    "ph": "s", "id": 51360, "pid": 76337, "tid": -914061504, "ts": 1716454218256827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310188, "dur": 6,
    "args": {
      "External id": 51362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51362,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51362, "pid": 5, "tid": 7, "ts": 1716454218310188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256837, "dur": 5,
    "args": {
      "External id": 51362,
      "cbid": 211, "correlation": 51362
    }
  },
  {
    "ph": "s", "id": 51362, "pid": 76337, "tid": -914061504, "ts": 1716454218256837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218310194, "dur": 6,
    "args": {
      "External id": 51368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51368,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51368, "pid": 5, "tid": 7, "ts": 1716454218310194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256867, "dur": 8,
    "args": {
      "External id": 51368,
      "cbid": 211, "correlation": 51368
    }
  },
  {
    "ph": "s", "id": 51368, "pid": 76337, "tid": -914061504, "ts": 1716454218256867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218310202, "dur": 3,
    "args": {
      "External id": 51376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51376,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 51376, "pid": 5, "tid": 7, "ts": 1716454218310202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256910, "dur": 10,
    "args": {
      "External id": 51376,
      "cbid": 211, "correlation": 51376
    }
  },
  {
    "ph": "s", "id": 51376, "pid": 76337, "tid": -914061504, "ts": 1716454218256910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256981, "dur": 1,
    "args": {
      "External id": 51392,
      "cbid": 251, "correlation": 51392
    }
  },
  {
    "ph": "f", "id": 51392, "pid": 76337, "tid": -914061504, "ts": 1716454218256981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256987, "dur": 0,
    "args": {
      "External id": 51394,
      "cbid": 251, "correlation": 51394
    }
  },
  {
    "ph": "f", "id": 51394, "pid": 76337, "tid": -914061504, "ts": 1716454218256987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218310206, "dur": 11,
    "args": {
      "External id": 51395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51395,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51395, "pid": 5, "tid": 7, "ts": 1716454218310206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218256989, "dur": 14,
    "args": {
      "External id": 51395,
      "cbid": 211, "correlation": 51395
    }
  },
  {
    "ph": "s", "id": 51395, "pid": 76337, "tid": -914061504, "ts": 1716454218256989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218310218, "dur": 4,
    "args": {
      "External id": 51397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51397,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51397, "pid": 5, "tid": 7, "ts": 1716454218310218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257005, "dur": 6,
    "args": {
      "External id": 51397,
      "cbid": 211, "correlation": 51397
    }
  },
  {
    "ph": "s", "id": 51397, "pid": 76337, "tid": -914061504, "ts": 1716454218257005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218310224, "dur": 6,
    "args": {
      "External id": 51407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51407,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51407, "pid": 5, "tid": 7, "ts": 1716454218310224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257064, "dur": 12,
    "args": {
      "External id": 51407,
      "cbid": 211, "correlation": 51407
    }
  },
  {
    "ph": "s", "id": 51407, "pid": 76337, "tid": -914061504, "ts": 1716454218257064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310231, "dur": 9,
    "args": {
      "External id": 51427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51427,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51427, "pid": 5, "tid": 7, "ts": 1716454218310231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257130, "dur": 11,
    "args": {
      "External id": 51427,
      "cbid": 211, "correlation": 51427
    }
  },
  {
    "ph": "s", "id": 51427, "pid": 76337, "tid": -914061504, "ts": 1716454218257130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218310241, "dur": 3,
    "args": {
      "External id": 51439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51439,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 51439, "pid": 5, "tid": 7, "ts": 1716454218310241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257150, "dur": 6,
    "args": {
      "External id": 51439,
      "cbid": 211, "correlation": 51439
    }
  },
  {
    "ph": "s", "id": 51439, "pid": 76337, "tid": -914061504, "ts": 1716454218257150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218310246, "dur": 6,
    "args": {
      "External id": 51442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51442, "pid": 5, "tid": 7, "ts": 1716454218310246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257169, "dur": 6,
    "args": {
      "External id": 51442,
      "cbid": 211, "correlation": 51442
    }
  },
  {
    "ph": "s", "id": 51442, "pid": 76337, "tid": -914061504, "ts": 1716454218257169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218310254, "dur": 4,
    "args": {
      "External id": 51451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51451,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51451, "pid": 5, "tid": 7, "ts": 1716454218310254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257209, "dur": 10,
    "args": {
      "External id": 51451,
      "cbid": 211, "correlation": 51451
    }
  },
  {
    "ph": "s", "id": 51451, "pid": 76337, "tid": -914061504, "ts": 1716454218257209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257271, "dur": 0,
    "args": {
      "External id": 51461,
      "cbid": 317, "correlation": 51461
    }
  },
  {
    "ph": "f", "id": 51461, "pid": 76337, "tid": -914061504, "ts": 1716454218257271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257272, "dur": 0,
    "args": {
      "External id": 51462,
      "cbid": 203, "correlation": 51462
    }
  },
  {
    "ph": "f", "id": 51462, "pid": 76337, "tid": -914061504, "ts": 1716454218257272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257273, "dur": 0,
    "args": {
      "External id": 51463,
      "cbid": 205, "correlation": 51463
    }
  },
  {
    "ph": "f", "id": 51463, "pid": 76337, "tid": -914061504, "ts": 1716454218257273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310260, "dur": 5,
    "args": {
      "External id": 51467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51467,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51467, "pid": 5, "tid": 7, "ts": 1716454218310260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257287, "dur": 15,
    "args": {
      "External id": 51467,
      "cbid": 211, "correlation": 51467
    }
  },
  {
    "ph": "s", "id": 51467, "pid": 76337, "tid": -914061504, "ts": 1716454218257287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310266, "dur": 157,
    "args": {
      "External id": 51469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51469,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51469, "pid": 5, "tid": 7, "ts": 1716454218310266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257305, "dur": 5,
    "args": {
      "External id": 51469,
      "cbid": 211, "correlation": 51469
    }
  },
  {
    "ph": "s", "id": 51469, "pid": 76337, "tid": -914061504, "ts": 1716454218257305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218310425, "dur": 1,
    "args": {
      "External id": 51471,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51471,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 51471, "pid": 5, "tid": 7, "ts": 1716454218310425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257315, "dur": 6,
    "args": {
      "External id": 51471,
      "cbid": 51, "correlation": 51471
    }
  },
  {
    "ph": "s", "id": 51471, "pid": 76337, "tid": -914061504, "ts": 1716454218257315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218310429, "dur": 251,
    "args": {
      "External id": 51472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51472,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51472, "pid": 5, "tid": 7, "ts": 1716454218310429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257323, "dur": 7,
    "args": {
      "External id": 51472,
      "cbid": 211, "correlation": 51472
    }
  },
  {
    "ph": "s", "id": 51472, "pid": 76337, "tid": -914061504, "ts": 1716454218257323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310681, "dur": 5,
    "args": {
      "External id": 51474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51474,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51474, "pid": 5, "tid": 7, "ts": 1716454218310681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257334, "dur": 5,
    "args": {
      "External id": 51474,
      "cbid": 211, "correlation": 51474
    }
  },
  {
    "ph": "s", "id": 51474, "pid": 76337, "tid": -914061504, "ts": 1716454218257334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218310687, "dur": 6,
    "args": {
      "External id": 51480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51480, "pid": 5, "tid": 7, "ts": 1716454218310687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257362, "dur": 8,
    "args": {
      "External id": 51480,
      "cbid": 211, "correlation": 51480
    }
  },
  {
    "ph": "s", "id": 51480, "pid": 76337, "tid": -914061504, "ts": 1716454218257362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257420, "dur": 0,
    "args": {
      "External id": 51490,
      "cbid": 317, "correlation": 51490
    }
  },
  {
    "ph": "f", "id": 51490, "pid": 76337, "tid": -914061504, "ts": 1716454218257420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257421, "dur": 0,
    "args": {
      "External id": 51491,
      "cbid": 203, "correlation": 51491
    }
  },
  {
    "ph": "f", "id": 51491, "pid": 76337, "tid": -914061504, "ts": 1716454218257421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257422, "dur": 0,
    "args": {
      "External id": 51492,
      "cbid": 205, "correlation": 51492
    }
  },
  {
    "ph": "f", "id": 51492, "pid": 76337, "tid": -914061504, "ts": 1716454218257422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310695, "dur": 8,
    "args": {
      "External id": 51496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51496,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51496, "pid": 5, "tid": 7, "ts": 1716454218310695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257437, "dur": 12,
    "args": {
      "External id": 51496,
      "cbid": 211, "correlation": 51496
    }
  },
  {
    "ph": "s", "id": 51496, "pid": 76337, "tid": -914061504, "ts": 1716454218257437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218310704, "dur": 3,
    "args": {
      "External id": 51498,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51498,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5463917525773196
    }
  },
  {
    "ph": "f", "id": 51498, "pid": 5, "tid": 7, "ts": 1716454218310704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257455, "dur": 14,
    "args": {
      "External id": 51498,
      "cbid": 51, "correlation": 51498
    }
  },
  {
    "ph": "s", "id": 51498, "pid": 76337, "tid": -914061504, "ts": 1716454218257455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218310708, "dur": 93,
    "args": {
      "External id": 51499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51499,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 51499, "pid": 5, "tid": 7, "ts": 1716454218310708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257471, "dur": 7,
    "args": {
      "External id": 51499,
      "cbid": 211, "correlation": 51499
    }
  },
  {
    "ph": "s", "id": 51499, "pid": 76337, "tid": -914061504, "ts": 1716454218257471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310803, "dur": 6,
    "args": {
      "External id": 51501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51501,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51501, "pid": 5, "tid": 7, "ts": 1716454218310803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257481, "dur": 5,
    "args": {
      "External id": 51501,
      "cbid": 211, "correlation": 51501
    }
  },
  {
    "ph": "s", "id": 51501, "pid": 76337, "tid": -914061504, "ts": 1716454218257481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218310810, "dur": 6,
    "args": {
      "External id": 51507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51507, "pid": 5, "tid": 7, "ts": 1716454218310810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257508, "dur": 8,
    "args": {
      "External id": 51507,
      "cbid": 211, "correlation": 51507
    }
  },
  {
    "ph": "s", "id": 51507, "pid": 76337, "tid": -914061504, "ts": 1716454218257508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218310817, "dur": 5,
    "args": {
      "External id": 51515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51515,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51515, "pid": 5, "tid": 7, "ts": 1716454218310817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257537, "dur": 8,
    "args": {
      "External id": 51515,
      "cbid": 211, "correlation": 51515
    }
  },
  {
    "ph": "s", "id": 51515, "pid": 76337, "tid": -914061504, "ts": 1716454218257537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218310823, "dur": 4,
    "args": {
      "External id": 51523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51523,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51523, "pid": 5, "tid": 7, "ts": 1716454218310823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257569, "dur": 8,
    "args": {
      "External id": 51523,
      "cbid": 211, "correlation": 51523
    }
  },
  {
    "ph": "s", "id": 51523, "pid": 76337, "tid": -914061504, "ts": 1716454218257569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218310829, "dur": 11,
    "args": {
      "External id": 51532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51532,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51532, "pid": 5, "tid": 7, "ts": 1716454218310829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257614, "dur": 10,
    "args": {
      "External id": 51532,
      "cbid": 211, "correlation": 51532
    }
  },
  {
    "ph": "s", "id": 51532, "pid": 76337, "tid": -914061504, "ts": 1716454218257614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310841, "dur": 12,
    "args": {
      "External id": 51552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51552,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51552, "pid": 5, "tid": 7, "ts": 1716454218310841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257683, "dur": 12,
    "args": {
      "External id": 51552,
      "cbid": 211, "correlation": 51552
    }
  },
  {
    "ph": "s", "id": 51552, "pid": 76337, "tid": -914061504, "ts": 1716454218257683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218310854, "dur": 4,
    "args": {
      "External id": 51564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51564,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51564, "pid": 5, "tid": 7, "ts": 1716454218310854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257704, "dur": 6,
    "args": {
      "External id": 51564,
      "cbid": 211, "correlation": 51564
    }
  },
  {
    "ph": "s", "id": 51564, "pid": 76337, "tid": -914061504, "ts": 1716454218257704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218310859, "dur": 10,
    "args": {
      "External id": 51567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51567,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51567, "pid": 5, "tid": 7, "ts": 1716454218310859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257722, "dur": 7,
    "args": {
      "External id": 51567,
      "cbid": 211, "correlation": 51567
    }
  },
  {
    "ph": "s", "id": 51567, "pid": 76337, "tid": -914061504, "ts": 1716454218257722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218310870, "dur": 6,
    "args": {
      "External id": 51576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51576,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51576, "pid": 5, "tid": 7, "ts": 1716454218310870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257761, "dur": 10,
    "args": {
      "External id": 51576,
      "cbid": 211, "correlation": 51576
    }
  },
  {
    "ph": "s", "id": 51576, "pid": 76337, "tid": -914061504, "ts": 1716454218257761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257812, "dur": 0,
    "args": {
      "External id": 51586,
      "cbid": 317, "correlation": 51586
    }
  },
  {
    "ph": "f", "id": 51586, "pid": 76337, "tid": -914061504, "ts": 1716454218257812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257813, "dur": 0,
    "args": {
      "External id": 51587,
      "cbid": 203, "correlation": 51587
    }
  },
  {
    "ph": "f", "id": 51587, "pid": 76337, "tid": -914061504, "ts": 1716454218257813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257813, "dur": 0,
    "args": {
      "External id": 51588,
      "cbid": 205, "correlation": 51588
    }
  },
  {
    "ph": "f", "id": 51588, "pid": 76337, "tid": -914061504, "ts": 1716454218257813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310877, "dur": 6,
    "args": {
      "External id": 51592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51592, "pid": 5, "tid": 7, "ts": 1716454218310877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257827, "dur": 11,
    "args": {
      "External id": 51592,
      "cbid": 211, "correlation": 51592
    }
  },
  {
    "ph": "s", "id": 51592, "pid": 76337, "tid": -914061504, "ts": 1716454218257827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218310885, "dur": 309,
    "args": {
      "External id": 51594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51594,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51594, "pid": 5, "tid": 7, "ts": 1716454218310885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257844, "dur": 6,
    "args": {
      "External id": 51594,
      "cbid": 211, "correlation": 51594
    }
  },
  {
    "ph": "s", "id": 51594, "pid": 76337, "tid": -914061504, "ts": 1716454218257844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218311196, "dur": 1,
    "args": {
      "External id": 51596,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51596,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 51596, "pid": 5, "tid": 7, "ts": 1716454218311196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257856, "dur": 6,
    "args": {
      "External id": 51596,
      "cbid": 51, "correlation": 51596
    }
  },
  {
    "ph": "s", "id": 51596, "pid": 76337, "tid": -914061504, "ts": 1716454218257856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218311200, "dur": 485,
    "args": {
      "External id": 51597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51597,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51597, "pid": 5, "tid": 7, "ts": 1716454218311200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257863, "dur": 6,
    "args": {
      "External id": 51597,
      "cbid": 211, "correlation": 51597
    }
  },
  {
    "ph": "s", "id": 51597, "pid": 76337, "tid": -914061504, "ts": 1716454218257863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218311687, "dur": 5,
    "args": {
      "External id": 51599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51599,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51599, "pid": 5, "tid": 7, "ts": 1716454218311687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257874, "dur": 5,
    "args": {
      "External id": 51599,
      "cbid": 211, "correlation": 51599
    }
  },
  {
    "ph": "s", "id": 51599, "pid": 76337, "tid": -914061504, "ts": 1716454218257874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218311693, "dur": 6,
    "args": {
      "External id": 51605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51605,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51605, "pid": 5, "tid": 7, "ts": 1716454218311693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257903, "dur": 8,
    "args": {
      "External id": 51605,
      "cbid": 211, "correlation": 51605
    }
  },
  {
    "ph": "s", "id": 51605, "pid": 76337, "tid": -914061504, "ts": 1716454218257903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218311701, "dur": 3,
    "args": {
      "External id": 51613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51613,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 51613, "pid": 5, "tid": 7, "ts": 1716454218311701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218257946, "dur": 9,
    "args": {
      "External id": 51613,
      "cbid": 211, "correlation": 51613
    }
  },
  {
    "ph": "s", "id": 51613, "pid": 76337, "tid": -914061504, "ts": 1716454218257946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258016, "dur": 1,
    "args": {
      "External id": 51629,
      "cbid": 251, "correlation": 51629
    }
  },
  {
    "ph": "f", "id": 51629, "pid": 76337, "tid": -914061504, "ts": 1716454218258016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258022, "dur": 0,
    "args": {
      "External id": 51631,
      "cbid": 251, "correlation": 51631
    }
  },
  {
    "ph": "f", "id": 51631, "pid": 76337, "tid": -914061504, "ts": 1716454218258022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218311705, "dur": 12,
    "args": {
      "External id": 51632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51632,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51632, "pid": 5, "tid": 7, "ts": 1716454218311705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258023, "dur": 12,
    "args": {
      "External id": 51632,
      "cbid": 211, "correlation": 51632
    }
  },
  {
    "ph": "s", "id": 51632, "pid": 76337, "tid": -914061504, "ts": 1716454218258023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218311719, "dur": 5,
    "args": {
      "External id": 51634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51634,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51634, "pid": 5, "tid": 7, "ts": 1716454218311719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258037, "dur": 5,
    "args": {
      "External id": 51634,
      "cbid": 211, "correlation": 51634
    }
  },
  {
    "ph": "s", "id": 51634, "pid": 76337, "tid": -914061504, "ts": 1716454218258037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218311725, "dur": 6,
    "args": {
      "External id": 51644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51644,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51644, "pid": 5, "tid": 7, "ts": 1716454218311725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258093, "dur": 13,
    "args": {
      "External id": 51644,
      "cbid": 211, "correlation": 51644
    }
  },
  {
    "ph": "s", "id": 51644, "pid": 76337, "tid": -914061504, "ts": 1716454218258093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218311732, "dur": 9,
    "args": {
      "External id": 51664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51664,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51664, "pid": 5, "tid": 7, "ts": 1716454218311732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258163, "dur": 11,
    "args": {
      "External id": 51664,
      "cbid": 211, "correlation": 51664
    }
  },
  {
    "ph": "s", "id": 51664, "pid": 76337, "tid": -914061504, "ts": 1716454218258163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218311743, "dur": 3,
    "args": {
      "External id": 51676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51676,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 51676, "pid": 5, "tid": 7, "ts": 1716454218311743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258184, "dur": 6,
    "args": {
      "External id": 51676,
      "cbid": 211, "correlation": 51676
    }
  },
  {
    "ph": "s", "id": 51676, "pid": 76337, "tid": -914061504, "ts": 1716454218258184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218311747, "dur": 6,
    "args": {
      "External id": 51679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51679,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51679, "pid": 5, "tid": 7, "ts": 1716454218311747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258203, "dur": 6,
    "args": {
      "External id": 51679,
      "cbid": 211, "correlation": 51679
    }
  },
  {
    "ph": "s", "id": 51679, "pid": 76337, "tid": -914061504, "ts": 1716454218258203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218311755, "dur": 4,
    "args": {
      "External id": 51688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51688,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51688, "pid": 5, "tid": 7, "ts": 1716454218311755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258243, "dur": 10,
    "args": {
      "External id": 51688,
      "cbid": 211, "correlation": 51688
    }
  },
  {
    "ph": "s", "id": 51688, "pid": 76337, "tid": -914061504, "ts": 1716454218258243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258307, "dur": 0,
    "args": {
      "External id": 51698,
      "cbid": 317, "correlation": 51698
    }
  },
  {
    "ph": "f", "id": 51698, "pid": 76337, "tid": -914061504, "ts": 1716454218258307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258308, "dur": 0,
    "args": {
      "External id": 51699,
      "cbid": 203, "correlation": 51699
    }
  },
  {
    "ph": "f", "id": 51699, "pid": 76337, "tid": -914061504, "ts": 1716454218258308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258309, "dur": 0,
    "args": {
      "External id": 51700,
      "cbid": 205, "correlation": 51700
    }
  },
  {
    "ph": "f", "id": 51700, "pid": 76337, "tid": -914061504, "ts": 1716454218258309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218311761, "dur": 5,
    "args": {
      "External id": 51704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51704, "pid": 5, "tid": 7, "ts": 1716454218311761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258322, "dur": 12,
    "args": {
      "External id": 51704,
      "cbid": 211, "correlation": 51704
    }
  },
  {
    "ph": "s", "id": 51704, "pid": 76337, "tid": -914061504, "ts": 1716454218258322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218311767, "dur": 157,
    "args": {
      "External id": 51706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51706, "pid": 5, "tid": 7, "ts": 1716454218311767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258336, "dur": 5,
    "args": {
      "External id": 51706,
      "cbid": 211, "correlation": 51706
    }
  },
  {
    "ph": "s", "id": 51706, "pid": 76337, "tid": -914061504, "ts": 1716454218258336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218311926, "dur": 1,
    "args": {
      "External id": 51708,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51708,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 51708, "pid": 5, "tid": 7, "ts": 1716454218311926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258347, "dur": 6,
    "args": {
      "External id": 51708,
      "cbid": 51, "correlation": 51708
    }
  },
  {
    "ph": "s", "id": 51708, "pid": 76337, "tid": -914061504, "ts": 1716454218258347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218311929, "dur": 252,
    "args": {
      "External id": 51709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51709,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51709, "pid": 5, "tid": 7, "ts": 1716454218311929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258354, "dur": 7,
    "args": {
      "External id": 51709,
      "cbid": 211, "correlation": 51709
    }
  },
  {
    "ph": "s", "id": 51709, "pid": 76337, "tid": -914061504, "ts": 1716454218258354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218312182, "dur": 6,
    "args": {
      "External id": 51711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51711,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51711, "pid": 5, "tid": 7, "ts": 1716454218312182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258364, "dur": 5,
    "args": {
      "External id": 51711,
      "cbid": 211, "correlation": 51711
    }
  },
  {
    "ph": "s", "id": 51711, "pid": 76337, "tid": -914061504, "ts": 1716454218258364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218312189, "dur": 6,
    "args": {
      "External id": 51717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51717, "pid": 5, "tid": 7, "ts": 1716454218312189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258392, "dur": 8,
    "args": {
      "External id": 51717,
      "cbid": 211, "correlation": 51717
    }
  },
  {
    "ph": "s", "id": 51717, "pid": 76337, "tid": -914061504, "ts": 1716454218258392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258454, "dur": 0,
    "args": {
      "External id": 51727,
      "cbid": 317, "correlation": 51727
    }
  },
  {
    "ph": "f", "id": 51727, "pid": 76337, "tid": -914061504, "ts": 1716454218258454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258455, "dur": 0,
    "args": {
      "External id": 51728,
      "cbid": 203, "correlation": 51728
    }
  },
  {
    "ph": "f", "id": 51728, "pid": 76337, "tid": -914061504, "ts": 1716454218258455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258455, "dur": 0,
    "args": {
      "External id": 51729,
      "cbid": 205, "correlation": 51729
    }
  },
  {
    "ph": "f", "id": 51729, "pid": 76337, "tid": -914061504, "ts": 1716454218258455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218312197, "dur": 8,
    "args": {
      "External id": 51733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51733,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51733, "pid": 5, "tid": 7, "ts": 1716454218312197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258468, "dur": 12,
    "args": {
      "External id": 51733,
      "cbid": 211, "correlation": 51733
    }
  },
  {
    "ph": "s", "id": 51733, "pid": 76337, "tid": -914061504, "ts": 1716454218258468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218312206, "dur": 3,
    "args": {
      "External id": 51735,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51735,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.530612244897959
    }
  },
  {
    "ph": "f", "id": 51735, "pid": 5, "tid": 7, "ts": 1716454218312206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258485, "dur": 10,
    "args": {
      "External id": 51735,
      "cbid": 51, "correlation": 51735
    }
  },
  {
    "ph": "s", "id": 51735, "pid": 76337, "tid": -914061504, "ts": 1716454218258485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218312210, "dur": 92,
    "args": {
      "External id": 51736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51736,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 51736, "pid": 5, "tid": 7, "ts": 1716454218312210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258495, "dur": 6,
    "args": {
      "External id": 51736,
      "cbid": 211, "correlation": 51736
    }
  },
  {
    "ph": "s", "id": 51736, "pid": 76337, "tid": -914061504, "ts": 1716454218258495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218312303, "dur": 6,
    "args": {
      "External id": 51738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51738, "pid": 5, "tid": 7, "ts": 1716454218312303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258505, "dur": 5,
    "args": {
      "External id": 51738,
      "cbid": 211, "correlation": 51738
    }
  },
  {
    "ph": "s", "id": 51738, "pid": 76337, "tid": -914061504, "ts": 1716454218258505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218312310, "dur": 6,
    "args": {
      "External id": 51744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51744, "pid": 5, "tid": 7, "ts": 1716454218312310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258532, "dur": 9,
    "args": {
      "External id": 51744,
      "cbid": 211, "correlation": 51744
    }
  },
  {
    "ph": "s", "id": 51744, "pid": 76337, "tid": -914061504, "ts": 1716454218258532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218312317, "dur": 5,
    "args": {
      "External id": 51752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51752,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51752, "pid": 5, "tid": 7, "ts": 1716454218312317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258561, "dur": 9,
    "args": {
      "External id": 51752,
      "cbid": 211, "correlation": 51752
    }
  },
  {
    "ph": "s", "id": 51752, "pid": 76337, "tid": -914061504, "ts": 1716454218258561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218312323, "dur": 4,
    "args": {
      "External id": 51760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51760,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51760, "pid": 5, "tid": 7, "ts": 1716454218312323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258590, "dur": 8,
    "args": {
      "External id": 51760,
      "cbid": 211, "correlation": 51760
    }
  },
  {
    "ph": "s", "id": 51760, "pid": 76337, "tid": -914061504, "ts": 1716454218258590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218312329, "dur": 11,
    "args": {
      "External id": 51769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51769,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51769, "pid": 5, "tid": 7, "ts": 1716454218312329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258634, "dur": 10,
    "args": {
      "External id": 51769,
      "cbid": 211, "correlation": 51769
    }
  },
  {
    "ph": "s", "id": 51769, "pid": 76337, "tid": -914061504, "ts": 1716454218258634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218312341, "dur": 12,
    "args": {
      "External id": 51789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51789,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51789, "pid": 5, "tid": 7, "ts": 1716454218312341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258706, "dur": 12,
    "args": {
      "External id": 51789,
      "cbid": 211, "correlation": 51789
    }
  },
  {
    "ph": "s", "id": 51789, "pid": 76337, "tid": -914061504, "ts": 1716454218258706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218312354, "dur": 4,
    "args": {
      "External id": 51801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51801,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51801, "pid": 5, "tid": 7, "ts": 1716454218312354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258729, "dur": 6,
    "args": {
      "External id": 51801,
      "cbid": 211, "correlation": 51801
    }
  },
  {
    "ph": "s", "id": 51801, "pid": 76337, "tid": -914061504, "ts": 1716454218258729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218312359, "dur": 10,
    "args": {
      "External id": 51804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51804,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51804, "pid": 5, "tid": 7, "ts": 1716454218312359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258747, "dur": 7,
    "args": {
      "External id": 51804,
      "cbid": 211, "correlation": 51804
    }
  },
  {
    "ph": "s", "id": 51804, "pid": 76337, "tid": -914061504, "ts": 1716454218258747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218312371, "dur": 6,
    "args": {
      "External id": 51813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51813,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51813, "pid": 5, "tid": 7, "ts": 1716454218312371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258786, "dur": 10,
    "args": {
      "External id": 51813,
      "cbid": 211, "correlation": 51813
    }
  },
  {
    "ph": "s", "id": 51813, "pid": 76337, "tid": -914061504, "ts": 1716454218258786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258839, "dur": 0,
    "args": {
      "External id": 51823,
      "cbid": 317, "correlation": 51823
    }
  },
  {
    "ph": "f", "id": 51823, "pid": 76337, "tid": -914061504, "ts": 1716454218258839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258840, "dur": 0,
    "args": {
      "External id": 51824,
      "cbid": 203, "correlation": 51824
    }
  },
  {
    "ph": "f", "id": 51824, "pid": 76337, "tid": -914061504, "ts": 1716454218258840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258841, "dur": 0,
    "args": {
      "External id": 51825,
      "cbid": 205, "correlation": 51825
    }
  },
  {
    "ph": "f", "id": 51825, "pid": 76337, "tid": -914061504, "ts": 1716454218258841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218312378, "dur": 6,
    "args": {
      "External id": 51829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51829,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51829, "pid": 5, "tid": 7, "ts": 1716454218312378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258854, "dur": 12,
    "args": {
      "External id": 51829,
      "cbid": 211, "correlation": 51829
    }
  },
  {
    "ph": "s", "id": 51829, "pid": 76337, "tid": -914061504, "ts": 1716454218258854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218312386, "dur": 309,
    "args": {
      "External id": 51831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51831,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51831, "pid": 5, "tid": 7, "ts": 1716454218312386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258868, "dur": 5,
    "args": {
      "External id": 51831,
      "cbid": 211, "correlation": 51831
    }
  },
  {
    "ph": "s", "id": 51831, "pid": 76337, "tid": -914061504, "ts": 1716454218258868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218312697, "dur": 1,
    "args": {
      "External id": 51833,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51833,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 51833, "pid": 5, "tid": 7, "ts": 1716454218312697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258879, "dur": 7,
    "args": {
      "External id": 51833,
      "cbid": 51, "correlation": 51833
    }
  },
  {
    "ph": "s", "id": 51833, "pid": 76337, "tid": -914061504, "ts": 1716454218258879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218312700, "dur": 484,
    "args": {
      "External id": 51834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51834,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51834, "pid": 5, "tid": 7, "ts": 1716454218312700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258886, "dur": 6,
    "args": {
      "External id": 51834,
      "cbid": 211, "correlation": 51834
    }
  },
  {
    "ph": "s", "id": 51834, "pid": 76337, "tid": -914061504, "ts": 1716454218258886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313186, "dur": 6,
    "args": {
      "External id": 51836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51836,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51836, "pid": 5, "tid": 7, "ts": 1716454218313186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258897, "dur": 5,
    "args": {
      "External id": 51836,
      "cbid": 211, "correlation": 51836
    }
  },
  {
    "ph": "s", "id": 51836, "pid": 76337, "tid": -914061504, "ts": 1716454218258897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218313193, "dur": 6,
    "args": {
      "External id": 51842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51842, "pid": 5, "tid": 7, "ts": 1716454218313193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258925, "dur": 8,
    "args": {
      "External id": 51842,
      "cbid": 211, "correlation": 51842
    }
  },
  {
    "ph": "s", "id": 51842, "pid": 76337, "tid": -914061504, "ts": 1716454218258925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218313200, "dur": 3,
    "args": {
      "External id": 51850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51850,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 51850, "pid": 5, "tid": 7, "ts": 1716454218313200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218258971, "dur": 19,
    "args": {
      "External id": 51850,
      "cbid": 211, "correlation": 51850
    }
  },
  {
    "ph": "s", "id": 51850, "pid": 76337, "tid": -914061504, "ts": 1716454218258971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259046, "dur": 1,
    "args": {
      "External id": 51866,
      "cbid": 251, "correlation": 51866
    }
  },
  {
    "ph": "f", "id": 51866, "pid": 76337, "tid": -914061504, "ts": 1716454218259046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259052, "dur": 0,
    "args": {
      "External id": 51868,
      "cbid": 251, "correlation": 51868
    }
  },
  {
    "ph": "f", "id": 51868, "pid": 76337, "tid": -914061504, "ts": 1716454218259052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218313204, "dur": 13,
    "args": {
      "External id": 51869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51869,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51869, "pid": 5, "tid": 7, "ts": 1716454218313204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259054, "dur": 11,
    "args": {
      "External id": 51869,
      "cbid": 211, "correlation": 51869
    }
  },
  {
    "ph": "s", "id": 51869, "pid": 76337, "tid": -914061504, "ts": 1716454218259054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218313218, "dur": 5,
    "args": {
      "External id": 51871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51871,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51871, "pid": 5, "tid": 7, "ts": 1716454218313218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259067, "dur": 5,
    "args": {
      "External id": 51871,
      "cbid": 211, "correlation": 51871
    }
  },
  {
    "ph": "s", "id": 51871, "pid": 76337, "tid": -914061504, "ts": 1716454218259067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218313225, "dur": 6,
    "args": {
      "External id": 51881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51881, "pid": 5, "tid": 7, "ts": 1716454218313225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259124, "dur": 12,
    "args": {
      "External id": 51881,
      "cbid": 211, "correlation": 51881
    }
  },
  {
    "ph": "s", "id": 51881, "pid": 76337, "tid": -914061504, "ts": 1716454218259124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313232, "dur": 9,
    "args": {
      "External id": 51901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51901,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 51901, "pid": 5, "tid": 7, "ts": 1716454218313232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259191, "dur": 12,
    "args": {
      "External id": 51901,
      "cbid": 211, "correlation": 51901
    }
  },
  {
    "ph": "s", "id": 51901, "pid": 76337, "tid": -914061504, "ts": 1716454218259191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218313242, "dur": 3,
    "args": {
      "External id": 51913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51913,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 51913, "pid": 5, "tid": 7, "ts": 1716454218313242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259212, "dur": 6,
    "args": {
      "External id": 51913,
      "cbid": 211, "correlation": 51913
    }
  },
  {
    "ph": "s", "id": 51913, "pid": 76337, "tid": -914061504, "ts": 1716454218259212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218313247, "dur": 7,
    "args": {
      "External id": 51916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51916,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51916, "pid": 5, "tid": 7, "ts": 1716454218313247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259230, "dur": 6,
    "args": {
      "External id": 51916,
      "cbid": 211, "correlation": 51916
    }
  },
  {
    "ph": "s", "id": 51916, "pid": 76337, "tid": -914061504, "ts": 1716454218259230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218313255, "dur": 4,
    "args": {
      "External id": 51925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51925,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51925, "pid": 5, "tid": 7, "ts": 1716454218313255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259274, "dur": 10,
    "args": {
      "External id": 51925,
      "cbid": 211, "correlation": 51925
    }
  },
  {
    "ph": "s", "id": 51925, "pid": 76337, "tid": -914061504, "ts": 1716454218259274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259338, "dur": 0,
    "args": {
      "External id": 51935,
      "cbid": 317, "correlation": 51935
    }
  },
  {
    "ph": "f", "id": 51935, "pid": 76337, "tid": -914061504, "ts": 1716454218259338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259338, "dur": 0,
    "args": {
      "External id": 51936,
      "cbid": 203, "correlation": 51936
    }
  },
  {
    "ph": "f", "id": 51936, "pid": 76337, "tid": -914061504, "ts": 1716454218259338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259339, "dur": 0,
    "args": {
      "External id": 51937,
      "cbid": 205, "correlation": 51937
    }
  },
  {
    "ph": "f", "id": 51937, "pid": 76337, "tid": -914061504, "ts": 1716454218259339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313261, "dur": 5,
    "args": {
      "External id": 51941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51941,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51941, "pid": 5, "tid": 7, "ts": 1716454218313261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259353, "dur": 13,
    "args": {
      "External id": 51941,
      "cbid": 211, "correlation": 51941
    }
  },
  {
    "ph": "s", "id": 51941, "pid": 76337, "tid": -914061504, "ts": 1716454218259353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313267, "dur": 157,
    "args": {
      "External id": 51943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51943,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51943, "pid": 5, "tid": 7, "ts": 1716454218313267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259368, "dur": 5,
    "args": {
      "External id": 51943,
      "cbid": 211, "correlation": 51943
    }
  },
  {
    "ph": "s", "id": 51943, "pid": 76337, "tid": -914061504, "ts": 1716454218259368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218313426, "dur": 1,
    "args": {
      "External id": 51945,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51945,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 51945, "pid": 5, "tid": 7, "ts": 1716454218313426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259378, "dur": 6,
    "args": {
      "External id": 51945,
      "cbid": 51, "correlation": 51945
    }
  },
  {
    "ph": "s", "id": 51945, "pid": 76337, "tid": -914061504, "ts": 1716454218259378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218313430, "dur": 251,
    "args": {
      "External id": 51946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51946,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 51946, "pid": 5, "tid": 7, "ts": 1716454218313430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259386, "dur": 6,
    "args": {
      "External id": 51946,
      "cbid": 211, "correlation": 51946
    }
  },
  {
    "ph": "s", "id": 51946, "pid": 76337, "tid": -914061504, "ts": 1716454218259386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313682, "dur": 6,
    "args": {
      "External id": 51948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51948, "pid": 5, "tid": 7, "ts": 1716454218313682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259395, "dur": 5,
    "args": {
      "External id": 51948,
      "cbid": 211, "correlation": 51948
    }
  },
  {
    "ph": "s", "id": 51948, "pid": 76337, "tid": -914061504, "ts": 1716454218259395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218313689, "dur": 6,
    "args": {
      "External id": 51954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51954, "pid": 5, "tid": 7, "ts": 1716454218313689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259424, "dur": 8,
    "args": {
      "External id": 51954,
      "cbid": 211, "correlation": 51954
    }
  },
  {
    "ph": "s", "id": 51954, "pid": 76337, "tid": -914061504, "ts": 1716454218259424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259483, "dur": 0,
    "args": {
      "External id": 51964,
      "cbid": 317, "correlation": 51964
    }
  },
  {
    "ph": "f", "id": 51964, "pid": 76337, "tid": -914061504, "ts": 1716454218259483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259484, "dur": 0,
    "args": {
      "External id": 51965,
      "cbid": 203, "correlation": 51965
    }
  },
  {
    "ph": "f", "id": 51965, "pid": 76337, "tid": -914061504, "ts": 1716454218259484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259484, "dur": 0,
    "args": {
      "External id": 51966,
      "cbid": 205, "correlation": 51966
    }
  },
  {
    "ph": "f", "id": 51966, "pid": 76337, "tid": -914061504, "ts": 1716454218259484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313696, "dur": 7,
    "args": {
      "External id": 51970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51970,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51970, "pid": 5, "tid": 7, "ts": 1716454218313696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259496, "dur": 11,
    "args": {
      "External id": 51970,
      "cbid": 211, "correlation": 51970
    }
  },
  {
    "ph": "s", "id": 51970, "pid": 76337, "tid": -914061504, "ts": 1716454218259496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218313705, "dur": 3,
    "args": {
      "External id": 51972,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 51972,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 51972, "pid": 5, "tid": 7, "ts": 1716454218313705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259512, "dur": 9,
    "args": {
      "External id": 51972,
      "cbid": 51, "correlation": 51972
    }
  },
  {
    "ph": "s", "id": 51972, "pid": 76337, "tid": -914061504, "ts": 1716454218259512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218313709, "dur": 92,
    "args": {
      "External id": 51973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51973,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 51973, "pid": 5, "tid": 7, "ts": 1716454218313709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259523, "dur": 6,
    "args": {
      "External id": 51973,
      "cbid": 211, "correlation": 51973
    }
  },
  {
    "ph": "s", "id": 51973, "pid": 76337, "tid": -914061504, "ts": 1716454218259523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313802, "dur": 5,
    "args": {
      "External id": 51975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51975,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 51975, "pid": 5, "tid": 7, "ts": 1716454218313802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259532, "dur": 5,
    "args": {
      "External id": 51975,
      "cbid": 211, "correlation": 51975
    }
  },
  {
    "ph": "s", "id": 51975, "pid": 76337, "tid": -914061504, "ts": 1716454218259532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218313809, "dur": 6,
    "args": {
      "External id": 51981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51981,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51981, "pid": 5, "tid": 7, "ts": 1716454218313809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259562, "dur": 10,
    "args": {
      "External id": 51981,
      "cbid": 211, "correlation": 51981
    }
  },
  {
    "ph": "s", "id": 51981, "pid": 76337, "tid": -914061504, "ts": 1716454218259562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218313817, "dur": 5,
    "args": {
      "External id": 51989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51989,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51989, "pid": 5, "tid": 7, "ts": 1716454218313817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259592, "dur": 8,
    "args": {
      "External id": 51989,
      "cbid": 211, "correlation": 51989
    }
  },
  {
    "ph": "s", "id": 51989, "pid": 76337, "tid": -914061504, "ts": 1716454218259592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218313823, "dur": 4,
    "args": {
      "External id": 51997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 51997,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 51997, "pid": 5, "tid": 7, "ts": 1716454218313823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259621, "dur": 8,
    "args": {
      "External id": 51997,
      "cbid": 211, "correlation": 51997
    }
  },
  {
    "ph": "s", "id": 51997, "pid": 76337, "tid": -914061504, "ts": 1716454218259621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218313828, "dur": 14,
    "args": {
      "External id": 52008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52008,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52008, "pid": 5, "tid": 7, "ts": 1716454218313828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259702, "dur": 14,
    "args": {
      "External id": 52008,
      "cbid": 211, "correlation": 52008
    }
  },
  {
    "ph": "s", "id": 52008, "pid": 76337, "tid": -914061504, "ts": 1716454218259702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259759, "dur": 0,
    "args": {
      "External id": 52018,
      "cbid": 317, "correlation": 52018
    }
  },
  {
    "ph": "f", "id": 52018, "pid": 76337, "tid": -914061504, "ts": 1716454218259759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259760, "dur": 0,
    "args": {
      "External id": 52019,
      "cbid": 203, "correlation": 52019
    }
  },
  {
    "ph": "f", "id": 52019, "pid": 76337, "tid": -914061504, "ts": 1716454218259760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259760, "dur": 0,
    "args": {
      "External id": 52020,
      "cbid": 205, "correlation": 52020
    }
  },
  {
    "ph": "f", "id": 52020, "pid": 76337, "tid": -914061504, "ts": 1716454218259760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313844, "dur": 8,
    "args": {
      "External id": 52024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52024,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52024, "pid": 5, "tid": 7, "ts": 1716454218313844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259775, "dur": 11,
    "args": {
      "External id": 52024,
      "cbid": 211, "correlation": 52024
    }
  },
  {
    "ph": "s", "id": 52024, "pid": 76337, "tid": -914061504, "ts": 1716454218259775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218313854, "dur": 158,
    "args": {
      "External id": 52026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52026, "pid": 5, "tid": 7, "ts": 1716454218313854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259789, "dur": 5,
    "args": {
      "External id": 52026,
      "cbid": 211, "correlation": 52026
    }
  },
  {
    "ph": "s", "id": 52026, "pid": 76337, "tid": -914061504, "ts": 1716454218259789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218314014, "dur": 1,
    "args": {
      "External id": 52028,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 52028,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 52028, "pid": 5, "tid": 7, "ts": 1716454218314014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259800, "dur": 6,
    "args": {
      "External id": 52028,
      "cbid": 51, "correlation": 52028
    }
  },
  {
    "ph": "s", "id": 52028, "pid": 76337, "tid": -914061504, "ts": 1716454218259800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218314018, "dur": 640,
    "args": {
      "External id": 52029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52029,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52029, "pid": 5, "tid": 7, "ts": 1716454218314018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259807, "dur": 6,
    "args": {
      "External id": 52029,
      "cbid": 211, "correlation": 52029
    }
  },
  {
    "ph": "s", "id": 52029, "pid": 76337, "tid": -914061504, "ts": 1716454218259807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218314659, "dur": 12,
    "args": {
      "External id": 52031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52031,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52031, "pid": 5, "tid": 7, "ts": 1716454218314659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259817, "dur": 6,
    "args": {
      "External id": 52031,
      "cbid": 211, "correlation": 52031
    }
  },
  {
    "ph": "s", "id": 52031, "pid": 76337, "tid": -914061504, "ts": 1716454218259817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218314672, "dur": 15,
    "args": {
      "External id": 52037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52037,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52037, "pid": 5, "tid": 7, "ts": 1716454218314672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259846, "dur": 8,
    "args": {
      "External id": 52037,
      "cbid": 211, "correlation": 52037
    }
  },
  {
    "ph": "s", "id": 52037, "pid": 76337, "tid": -914061504, "ts": 1716454218259846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218314688, "dur": 29,
    "args": {
      "External id": 52046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52046,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52046, "pid": 5, "tid": 7, "ts": 1716454218314688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218259939, "dur": 13,
    "args": {
      "External id": 52046,
      "cbid": 211, "correlation": 52046
    }
  },
  {
    "ph": "s", "id": 52046, "pid": 76337, "tid": -914061504, "ts": 1716454218259939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218314718, "dur": 30,
    "args": {
      "External id": 52066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52066,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 52066, "pid": 5, "tid": 7, "ts": 1716454218314718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260016, "dur": 12,
    "args": {
      "External id": 52066,
      "cbid": 211, "correlation": 52066
    }
  },
  {
    "ph": "s", "id": 52066, "pid": 76337, "tid": -914061504, "ts": 1716454218260016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218314749, "dur": 4,
    "args": {
      "External id": 52078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52078,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52078, "pid": 5, "tid": 7, "ts": 1716454218314749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260039, "dur": 6,
    "args": {
      "External id": 52078,
      "cbid": 211, "correlation": 52078
    }
  },
  {
    "ph": "s", "id": 52078, "pid": 76337, "tid": -914061504, "ts": 1716454218260039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218314755, "dur": 29,
    "args": {
      "External id": 52081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52081,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52081, "pid": 5, "tid": 7, "ts": 1716454218314755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260057, "dur": 7,
    "args": {
      "External id": 52081,
      "cbid": 211, "correlation": 52081
    }
  },
  {
    "ph": "s", "id": 52081, "pid": 76337, "tid": -914061504, "ts": 1716454218260057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218314785, "dur": 20,
    "args": {
      "External id": 52090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52090,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52090, "pid": 5, "tid": 7, "ts": 1716454218314785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260097, "dur": 9,
    "args": {
      "External id": 52090,
      "cbid": 211, "correlation": 52090
    }
  },
  {
    "ph": "s", "id": 52090, "pid": 76337, "tid": -914061504, "ts": 1716454218260097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260149, "dur": 0,
    "args": {
      "External id": 52100,
      "cbid": 317, "correlation": 52100
    }
  },
  {
    "ph": "f", "id": 52100, "pid": 76337, "tid": -914061504, "ts": 1716454218260149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260150, "dur": 0,
    "args": {
      "External id": 52101,
      "cbid": 203, "correlation": 52101
    }
  },
  {
    "ph": "f", "id": 52101, "pid": 76337, "tid": -914061504, "ts": 1716454218260150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260151, "dur": 0,
    "args": {
      "External id": 52102,
      "cbid": 205, "correlation": 52102
    }
  },
  {
    "ph": "f", "id": 52102, "pid": 76337, "tid": -914061504, "ts": 1716454218260151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218314807, "dur": 22,
    "args": {
      "External id": 52106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52106,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52106, "pid": 5, "tid": 7, "ts": 1716454218314807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260169, "dur": 12,
    "args": {
      "External id": 52106,
      "cbid": 211, "correlation": 52106
    }
  },
  {
    "ph": "s", "id": 52106, "pid": 76337, "tid": -914061504, "ts": 1716454218260169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218314830, "dur": 317,
    "args": {
      "External id": 52108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52108,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52108, "pid": 5, "tid": 7, "ts": 1716454218314830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260183, "dur": 6,
    "args": {
      "External id": 52108,
      "cbid": 211, "correlation": 52108
    }
  },
  {
    "ph": "s", "id": 52108, "pid": 76337, "tid": -914061504, "ts": 1716454218260183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218315149, "dur": 1,
    "args": {
      "External id": 52110,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 52110,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 52110, "pid": 5, "tid": 7, "ts": 1716454218315149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260195, "dur": 8,
    "args": {
      "External id": 52110,
      "cbid": 51, "correlation": 52110
    }
  },
  {
    "ph": "s", "id": 52110, "pid": 76337, "tid": -914061504, "ts": 1716454218260195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218315153, "dur": 1229,
    "args": {
      "External id": 52111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52111,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52111, "pid": 5, "tid": 7, "ts": 1716454218315153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260204, "dur": 6,
    "args": {
      "External id": 52111,
      "cbid": 211, "correlation": 52111
    }
  },
  {
    "ph": "s", "id": 52111, "pid": 76337, "tid": -914061504, "ts": 1716454218260204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218316384, "dur": 13,
    "args": {
      "External id": 52113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52113,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52113, "pid": 5, "tid": 7, "ts": 1716454218316384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260214, "dur": 5,
    "args": {
      "External id": 52113,
      "cbid": 211, "correlation": 52113
    }
  },
  {
    "ph": "s", "id": 52113, "pid": 76337, "tid": -914061504, "ts": 1716454218260214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218316398, "dur": 15,
    "args": {
      "External id": 52119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52119, "pid": 5, "tid": 7, "ts": 1716454218316398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260243, "dur": 8,
    "args": {
      "External id": 52119,
      "cbid": 211, "correlation": 52119
    }
  },
  {
    "ph": "s", "id": 52119, "pid": 76337, "tid": -914061504, "ts": 1716454218260243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218316415, "dur": 3,
    "args": {
      "External id": 52127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52127,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 52127, "pid": 5, "tid": 7, "ts": 1716454218316415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260286, "dur": 10,
    "args": {
      "External id": 52127,
      "cbid": 211, "correlation": 52127
    }
  },
  {
    "ph": "s", "id": 52127, "pid": 76337, "tid": -914061504, "ts": 1716454218260286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260351, "dur": 1,
    "args": {
      "External id": 52143,
      "cbid": 251, "correlation": 52143
    }
  },
  {
    "ph": "f", "id": 52143, "pid": 76337, "tid": -914061504, "ts": 1716454218260351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260356, "dur": 0,
    "args": {
      "External id": 52145,
      "cbid": 251, "correlation": 52145
    }
  },
  {
    "ph": "f", "id": 52145, "pid": 76337, "tid": -914061504, "ts": 1716454218260356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218316419, "dur": 12,
    "args": {
      "External id": 52146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52146,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52146, "pid": 5, "tid": 7, "ts": 1716454218316419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260358, "dur": 11,
    "args": {
      "External id": 52146,
      "cbid": 211, "correlation": 52146
    }
  },
  {
    "ph": "s", "id": 52146, "pid": 76337, "tid": -914061504, "ts": 1716454218260358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218316433, "dur": 5,
    "args": {
      "External id": 52148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52148,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52148, "pid": 5, "tid": 7, "ts": 1716454218316433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260371, "dur": 6,
    "args": {
      "External id": 52148,
      "cbid": 211, "correlation": 52148
    }
  },
  {
    "ph": "s", "id": 52148, "pid": 76337, "tid": -914061504, "ts": 1716454218260371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218316439, "dur": 17,
    "args": {
      "External id": 52158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52158, "pid": 5, "tid": 7, "ts": 1716454218316439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260427, "dur": 12,
    "args": {
      "External id": 52158,
      "cbid": 211, "correlation": 52158
    }
  },
  {
    "ph": "s", "id": 52158, "pid": 76337, "tid": -914061504, "ts": 1716454218260427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218316457, "dur": 17,
    "args": {
      "External id": 52178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52178,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 52178, "pid": 5, "tid": 7, "ts": 1716454218316457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260493, "dur": 14,
    "args": {
      "External id": 52178,
      "cbid": 211, "correlation": 52178
    }
  },
  {
    "ph": "s", "id": 52178, "pid": 76337, "tid": -914061504, "ts": 1716454218260493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218316476, "dur": 4,
    "args": {
      "External id": 52190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52190,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 52190, "pid": 5, "tid": 7, "ts": 1716454218316476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260517, "dur": 6,
    "args": {
      "External id": 52190,
      "cbid": 211, "correlation": 52190
    }
  },
  {
    "ph": "s", "id": 52190, "pid": 76337, "tid": -914061504, "ts": 1716454218260517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218316481, "dur": 16,
    "args": {
      "External id": 52193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52193, "pid": 5, "tid": 7, "ts": 1716454218316481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260535, "dur": 6,
    "args": {
      "External id": 52193,
      "cbid": 211, "correlation": 52193
    }
  },
  {
    "ph": "s", "id": 52193, "pid": 76337, "tid": -914061504, "ts": 1716454218260535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218316499, "dur": 11,
    "args": {
      "External id": 52202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52202,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52202, "pid": 5, "tid": 7, "ts": 1716454218316499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260576, "dur": 9,
    "args": {
      "External id": 52202,
      "cbid": 211, "correlation": 52202
    }
  },
  {
    "ph": "s", "id": 52202, "pid": 76337, "tid": -914061504, "ts": 1716454218260576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260638, "dur": 0,
    "args": {
      "External id": 52212,
      "cbid": 317, "correlation": 52212
    }
  },
  {
    "ph": "f", "id": 52212, "pid": 76337, "tid": -914061504, "ts": 1716454218260638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260639, "dur": 0,
    "args": {
      "External id": 52213,
      "cbid": 203, "correlation": 52213
    }
  },
  {
    "ph": "f", "id": 52213, "pid": 76337, "tid": -914061504, "ts": 1716454218260639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260640, "dur": 0,
    "args": {
      "External id": 52214,
      "cbid": 205, "correlation": 52214
    }
  },
  {
    "ph": "f", "id": 52214, "pid": 76337, "tid": -914061504, "ts": 1716454218260640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218316511, "dur": 12,
    "args": {
      "External id": 52218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52218,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52218, "pid": 5, "tid": 7, "ts": 1716454218316511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260654, "dur": 12,
    "args": {
      "External id": 52218,
      "cbid": 211, "correlation": 52218
    }
  },
  {
    "ph": "s", "id": 52218, "pid": 76337, "tid": -914061504, "ts": 1716454218260654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218316524, "dur": 160,
    "args": {
      "External id": 52220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52220, "pid": 5, "tid": 7, "ts": 1716454218316524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260668, "dur": 5,
    "args": {
      "External id": 52220,
      "cbid": 211, "correlation": 52220
    }
  },
  {
    "ph": "s", "id": 52220, "pid": 76337, "tid": -914061504, "ts": 1716454218260668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218316687, "dur": 1,
    "args": {
      "External id": 52222,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 52222,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 52222, "pid": 5, "tid": 7, "ts": 1716454218316687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260679, "dur": 6,
    "args": {
      "External id": 52222,
      "cbid": 51, "correlation": 52222
    }
  },
  {
    "ph": "s", "id": 52222, "pid": 76337, "tid": -914061504, "ts": 1716454218260679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218316690, "dur": 644,
    "args": {
      "External id": 52223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52223,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52223, "pid": 5, "tid": 7, "ts": 1716454218316690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260686, "dur": 6,
    "args": {
      "External id": 52223,
      "cbid": 211, "correlation": 52223
    }
  },
  {
    "ph": "s", "id": 52223, "pid": 76337, "tid": -914061504, "ts": 1716454218260686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218317335, "dur": 12,
    "args": {
      "External id": 52225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52225,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52225, "pid": 5, "tid": 7, "ts": 1716454218317335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260696, "dur": 5,
    "args": {
      "External id": 52225,
      "cbid": 211, "correlation": 52225
    }
  },
  {
    "ph": "s", "id": 52225, "pid": 76337, "tid": -914061504, "ts": 1716454218260696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218317349, "dur": 14,
    "args": {
      "External id": 52231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52231, "pid": 5, "tid": 7, "ts": 1716454218317349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260724, "dur": 8,
    "args": {
      "External id": 52231,
      "cbid": 211, "correlation": 52231
    }
  },
  {
    "ph": "s", "id": 52231, "pid": 76337, "tid": -914061504, "ts": 1716454218260724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260783, "dur": 3,
    "args": {
      "External id": 52241,
      "cbid": 317, "correlation": 52241
    }
  },
  {
    "ph": "f", "id": 52241, "pid": 76337, "tid": -914061504, "ts": 1716454218260783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260786, "dur": 0,
    "args": {
      "External id": 52242,
      "cbid": 203, "correlation": 52242
    }
  },
  {
    "ph": "f", "id": 52242, "pid": 76337, "tid": -914061504, "ts": 1716454218260786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260787, "dur": 0,
    "args": {
      "External id": 52243,
      "cbid": 205, "correlation": 52243
    }
  },
  {
    "ph": "f", "id": 52243, "pid": 76337, "tid": -914061504, "ts": 1716454218260787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218317365, "dur": 21,
    "args": {
      "External id": 52247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52247,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52247, "pid": 5, "tid": 7, "ts": 1716454218317365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260801, "dur": 12,
    "args": {
      "External id": 52247,
      "cbid": 211, "correlation": 52247
    }
  },
  {
    "ph": "s", "id": 52247, "pid": 76337, "tid": -914061504, "ts": 1716454218260801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218317387, "dur": 4,
    "args": {
      "External id": 52249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52249,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52249, "pid": 5, "tid": 7, "ts": 1716454218317387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260818, "dur": 6,
    "args": {
      "External id": 52249,
      "cbid": 211, "correlation": 52249
    }
  },
  {
    "ph": "s", "id": 52249, "pid": 76337, "tid": -914061504, "ts": 1716454218260818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260828, "dur": 0,
    "args": {
      "External id": 52250,
      "cbid": 51, "correlation": 52250
    }
  },
  {
    "ph": "s", "id": 52250, "pid": 76337, "tid": -914061504, "ts": 1716454218260828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218317392, "dur": 173,
    "args": {
      "External id": 52251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52251,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 52251, "pid": 5, "tid": 7, "ts": 1716454218317392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260829, "dur": 5,
    "args": {
      "External id": 52251,
      "cbid": 211, "correlation": 52251
    }
  },
  {
    "ph": "s", "id": 52251, "pid": 76337, "tid": -914061504, "ts": 1716454218260829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218317566, "dur": 16,
    "args": {
      "External id": 52256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52256, "pid": 5, "tid": 7, "ts": 1716454218317566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260854, "dur": 8,
    "args": {
      "External id": 52256,
      "cbid": 211, "correlation": 52256
    }
  },
  {
    "ph": "s", "id": 52256, "pid": 76337, "tid": -914061504, "ts": 1716454218260854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218317583, "dur": 12,
    "args": {
      "External id": 52264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52264,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52264, "pid": 5, "tid": 7, "ts": 1716454218317583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260883, "dur": 8,
    "args": {
      "External id": 52264,
      "cbid": 211, "correlation": 52264
    }
  },
  {
    "ph": "s", "id": 52264, "pid": 76337, "tid": -914061504, "ts": 1716454218260883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218317596, "dur": 10,
    "args": {
      "External id": 52272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52272,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52272, "pid": 5, "tid": 7, "ts": 1716454218317596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218260911, "dur": 9,
    "args": {
      "External id": 52272,
      "cbid": 211, "correlation": 52272
    }
  },
  {
    "ph": "s", "id": 52272, "pid": 76337, "tid": -914061504, "ts": 1716454218260911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218317608, "dur": 18,
    "args": {
      "External id": 52292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52292,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 52292, "pid": 5, "tid": 7, "ts": 1716454218317608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261001, "dur": 12,
    "args": {
      "External id": 52292,
      "cbid": 211, "correlation": 52292
    }
  },
  {
    "ph": "s", "id": 52292, "pid": 76337, "tid": -914061504, "ts": 1716454218261001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218317627, "dur": 4,
    "args": {
      "External id": 52304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52304,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 52304, "pid": 5, "tid": 7, "ts": 1716454218317627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261024, "dur": 7,
    "args": {
      "External id": 52304,
      "cbid": 211, "correlation": 52304
    }
  },
  {
    "ph": "s", "id": 52304, "pid": 76337, "tid": -914061504, "ts": 1716454218261024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218317633, "dur": 17,
    "args": {
      "External id": 52307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52307,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52307, "pid": 5, "tid": 7, "ts": 1716454218317633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261042, "dur": 9,
    "args": {
      "External id": 52307,
      "cbid": 211, "correlation": 52307
    }
  },
  {
    "ph": "s", "id": 52307, "pid": 76337, "tid": -914061504, "ts": 1716454218261042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261104, "dur": 0,
    "args": {
      "External id": 52318,
      "cbid": 317, "correlation": 52318
    }
  },
  {
    "ph": "f", "id": 52318, "pid": 76337, "tid": -914061504, "ts": 1716454218261104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261105, "dur": 0,
    "args": {
      "External id": 52319,
      "cbid": 203, "correlation": 52319
    }
  },
  {
    "ph": "f", "id": 52319, "pid": 76337, "tid": -914061504, "ts": 1716454218261105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261106, "dur": 0,
    "args": {
      "External id": 52320,
      "cbid": 205, "correlation": 52320
    }
  },
  {
    "ph": "f", "id": 52320, "pid": 76337, "tid": -914061504, "ts": 1716454218261106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218317651, "dur": 11,
    "args": {
      "External id": 52324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52324,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52324, "pid": 5, "tid": 7, "ts": 1716454218317651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261119, "dur": 12,
    "args": {
      "External id": 52324,
      "cbid": 211, "correlation": 52324
    }
  },
  {
    "ph": "s", "id": 52324, "pid": 76337, "tid": -914061504, "ts": 1716454218261119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218317664, "dur": 3,
    "args": {
      "External id": 52326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52326, "pid": 5, "tid": 7, "ts": 1716454218317664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261136, "dur": 5,
    "args": {
      "External id": 52326,
      "cbid": 211, "correlation": 52326
    }
  },
  {
    "ph": "s", "id": 52326, "pid": 76337, "tid": -914061504, "ts": 1716454218261136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261144, "dur": 0,
    "args": {
      "External id": 52327,
      "cbid": 51, "correlation": 52327
    }
  },
  {
    "ph": "s", "id": 52327, "pid": 76337, "tid": -914061504, "ts": 1716454218261144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218317668, "dur": 90,
    "args": {
      "External id": 52328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52328,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 52328, "pid": 5, "tid": 7, "ts": 1716454218317668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261144, "dur": 5,
    "args": {
      "External id": 52328,
      "cbid": 211, "correlation": 52328
    }
  },
  {
    "ph": "s", "id": 52328, "pid": 76337, "tid": -914061504, "ts": 1716454218261144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218317759, "dur": 15,
    "args": {
      "External id": 52333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52333, "pid": 5, "tid": 7, "ts": 1716454218317759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261171, "dur": 8,
    "args": {
      "External id": 52333,
      "cbid": 211, "correlation": 52333
    }
  },
  {
    "ph": "s", "id": 52333, "pid": 76337, "tid": -914061504, "ts": 1716454218261171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218317776, "dur": 82,
    "args": {
      "External id": 52342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52342,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52342, "pid": 5, "tid": 7, "ts": 1716454218317776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261252, "dur": 14,
    "args": {
      "External id": 52342,
      "cbid": 211, "correlation": 52342
    }
  },
  {
    "ph": "s", "id": 52342, "pid": 76337, "tid": -914061504, "ts": 1716454218261252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218317859, "dur": 30,
    "args": {
      "External id": 52364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52364,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52364, "pid": 5, "tid": 7, "ts": 1716454218317859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261315, "dur": 10,
    "args": {
      "External id": 52364,
      "cbid": 211, "correlation": 52364
    }
  },
  {
    "ph": "s", "id": 52364, "pid": 76337, "tid": -914061504, "ts": 1716454218261315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261408, "dur": 1,
    "args": {
      "External id": 52375,
      "cbid": 251, "correlation": 52375
    }
  },
  {
    "ph": "f", "id": 52375, "pid": 76337, "tid": -914061504, "ts": 1716454218261408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218317890, "dur": 161,
    "args": {
      "External id": 52376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52376,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52376, "pid": 5, "tid": 7, "ts": 1716454218317890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261414, "dur": 16,
    "args": {
      "External id": 52376,
      "cbid": 211, "correlation": 52376
    }
  },
  {
    "ph": "s", "id": 52376, "pid": 76337, "tid": -914061504, "ts": 1716454218261414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261488, "dur": 1,
    "args": {
      "External id": 52387,
      "cbid": 251, "correlation": 52387
    }
  },
  {
    "ph": "f", "id": 52387, "pid": 76337, "tid": -914061504, "ts": 1716454218261488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218318053, "dur": 156,
    "args": {
      "External id": 52388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52388,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52388, "pid": 5, "tid": 7, "ts": 1716454218318053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261493, "dur": 13,
    "args": {
      "External id": 52388,
      "cbid": 211, "correlation": 52388
    }
  },
  {
    "ph": "s", "id": 52388, "pid": 76337, "tid": -914061504, "ts": 1716454218261493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261560, "dur": 1,
    "args": {
      "External id": 52399,
      "cbid": 251, "correlation": 52399
    }
  },
  {
    "ph": "f", "id": 52399, "pid": 76337, "tid": -914061504, "ts": 1716454218261560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218318210, "dur": 135,
    "args": {
      "External id": 52400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52400,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52400, "pid": 5, "tid": 7, "ts": 1716454218318210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261564, "dur": 11,
    "args": {
      "External id": 52400,
      "cbid": 211, "correlation": 52400
    }
  },
  {
    "ph": "s", "id": 52400, "pid": 76337, "tid": -914061504, "ts": 1716454218261564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218318346, "dur": 331,
    "args": {
      "External id": 52425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52425,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52425, "pid": 5, "tid": 7, "ts": 1716454218318346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261652, "dur": 13,
    "args": {
      "External id": 52425,
      "cbid": 211, "correlation": 52425
    }
  },
  {
    "ph": "s", "id": 52425, "pid": 76337, "tid": -914061504, "ts": 1716454218261652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261755, "dur": 1,
    "args": {
      "External id": 52443,
      "cbid": 251, "correlation": 52443
    }
  },
  {
    "ph": "f", "id": 52443, "pid": 76337, "tid": -914061504, "ts": 1716454218261755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218318679, "dur": 166,
    "args": {
      "External id": 52445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52445,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52445, "pid": 5, "tid": 7, "ts": 1716454218318679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261761, "dur": 13,
    "args": {
      "External id": 52445,
      "cbid": 211, "correlation": 52445
    }
  },
  {
    "ph": "s", "id": 52445, "pid": 76337, "tid": -914061504, "ts": 1716454218261761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218318846, "dur": 19,
    "args": {
      "External id": 52453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52453,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52453, "pid": 5, "tid": 7, "ts": 1716454218318846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261831, "dur": 12,
    "args": {
      "External id": 52453,
      "cbid": 211, "correlation": 52453
    }
  },
  {
    "ph": "s", "id": 52453, "pid": 76337, "tid": -914061504, "ts": 1716454218261831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218318866, "dur": 27,
    "args": {
      "External id": 52461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52461,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52461, "pid": 5, "tid": 7, "ts": 1716454218318866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261874, "dur": 9,
    "args": {
      "External id": 52461,
      "cbid": 211, "correlation": 52461
    }
  },
  {
    "ph": "s", "id": 52461, "pid": 76337, "tid": -914061504, "ts": 1716454218261874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218318894, "dur": 18,
    "args": {
      "External id": 52472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52472,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52472, "pid": 5, "tid": 7, "ts": 1716454218318894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261950, "dur": 13,
    "args": {
      "External id": 52472,
      "cbid": 211, "correlation": 52472
    }
  },
  {
    "ph": "s", "id": 52472, "pid": 76337, "tid": -914061504, "ts": 1716454218261950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218318914, "dur": 16,
    "args": {
      "External id": 52494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52494,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52494, "pid": 5, "tid": 7, "ts": 1716454218318914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218261990, "dur": 9,
    "args": {
      "External id": 52494,
      "cbid": 211, "correlation": 52494
    }
  },
  {
    "ph": "s", "id": 52494, "pid": 76337, "tid": -914061504, "ts": 1716454218261990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262078, "dur": 2,
    "args": {
      "External id": 52505,
      "cbid": 251, "correlation": 52505
    }
  },
  {
    "ph": "f", "id": 52505, "pid": 76337, "tid": -914061504, "ts": 1716454218262078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218318931, "dur": 88,
    "args": {
      "External id": 52506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52506,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52506, "pid": 5, "tid": 7, "ts": 1716454218318931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262083, "dur": 13,
    "args": {
      "External id": 52506,
      "cbid": 211, "correlation": 52506
    }
  },
  {
    "ph": "s", "id": 52506, "pid": 76337, "tid": -914061504, "ts": 1716454218262083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262154, "dur": 1,
    "args": {
      "External id": 52517,
      "cbid": 251, "correlation": 52517
    }
  },
  {
    "ph": "f", "id": 52517, "pid": 76337, "tid": -914061504, "ts": 1716454218262154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262157, "dur": 0,
    "args": {
      "External id": 52518,
      "cbid": 251, "correlation": 52518
    }
  },
  {
    "ph": "f", "id": 52518, "pid": 76337, "tid": -914061504, "ts": 1716454218262157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218319021, "dur": 12,
    "args": {
      "External id": 52519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52519,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52519, "pid": 5, "tid": 7, "ts": 1716454218319021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262159, "dur": 12,
    "args": {
      "External id": 52519,
      "cbid": 211, "correlation": 52519
    }
  },
  {
    "ph": "s", "id": 52519, "pid": 76337, "tid": -914061504, "ts": 1716454218262159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218319034, "dur": 5,
    "args": {
      "External id": 52521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52521,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52521, "pid": 5, "tid": 7, "ts": 1716454218319034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262172, "dur": 6,
    "args": {
      "External id": 52521,
      "cbid": 211, "correlation": 52521
    }
  },
  {
    "ph": "s", "id": 52521, "pid": 76337, "tid": -914061504, "ts": 1716454218262172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262230, "dur": 4,
    "args": {
      "External id": 52532,
      "cbid": 251, "correlation": 52532
    }
  },
  {
    "ph": "f", "id": 52532, "pid": 76337, "tid": -914061504, "ts": 1716454218262230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262237, "dur": 0,
    "args": {
      "External id": 52533,
      "cbid": 251, "correlation": 52533
    }
  },
  {
    "ph": "f", "id": 52533, "pid": 76337, "tid": -914061504, "ts": 1716454218262237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218319041, "dur": 8,
    "args": {
      "External id": 52534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52534,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52534, "pid": 5, "tid": 7, "ts": 1716454218319041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262238, "dur": 12,
    "args": {
      "External id": 52534,
      "cbid": 211, "correlation": 52534
    }
  },
  {
    "ph": "s", "id": 52534, "pid": 76337, "tid": -914061504, "ts": 1716454218262238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218319050, "dur": 3,
    "args": {
      "External id": 52536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52536,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52536, "pid": 5, "tid": 7, "ts": 1716454218319050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262251, "dur": 5,
    "args": {
      "External id": 52536,
      "cbid": 211, "correlation": 52536
    }
  },
  {
    "ph": "s", "id": 52536, "pid": 76337, "tid": -914061504, "ts": 1716454218262251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218319055, "dur": 55,
    "args": {
      "External id": 52561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52561,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52561, "pid": 5, "tid": 7, "ts": 1716454218319055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262328, "dur": 13,
    "args": {
      "External id": 52561,
      "cbid": 211, "correlation": 52561
    }
  },
  {
    "ph": "s", "id": 52561, "pid": 76337, "tid": -914061504, "ts": 1716454218262328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262428, "dur": 2,
    "args": {
      "External id": 52579,
      "cbid": 251, "correlation": 52579
    }
  },
  {
    "ph": "f", "id": 52579, "pid": 76337, "tid": -914061504, "ts": 1716454218262428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218319111, "dur": 90,
    "args": {
      "External id": 52581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52581,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52581, "pid": 5, "tid": 7, "ts": 1716454218319111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262435, "dur": 15,
    "args": {
      "External id": 52581,
      "cbid": 211, "correlation": 52581
    }
  },
  {
    "ph": "s", "id": 52581, "pid": 76337, "tid": -914061504, "ts": 1716454218262435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218319202, "dur": 10,
    "args": {
      "External id": 52589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52589,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52589, "pid": 5, "tid": 7, "ts": 1716454218319202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262505, "dur": 12,
    "args": {
      "External id": 52589,
      "cbid": 211, "correlation": 52589
    }
  },
  {
    "ph": "s", "id": 52589, "pid": 76337, "tid": -914061504, "ts": 1716454218262505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218319213, "dur": 21,
    "args": {
      "External id": 52597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52597,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52597, "pid": 5, "tid": 7, "ts": 1716454218319213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262547, "dur": 9,
    "args": {
      "External id": 52597,
      "cbid": 211, "correlation": 52597
    }
  },
  {
    "ph": "s", "id": 52597, "pid": 76337, "tid": -914061504, "ts": 1716454218262547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218319235, "dur": 18,
    "args": {
      "External id": 52619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52619,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52619, "pid": 5, "tid": 7, "ts": 1716454218319235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262601, "dur": 11,
    "args": {
      "External id": 52619,
      "cbid": 211, "correlation": 52619
    }
  },
  {
    "ph": "s", "id": 52619, "pid": 76337, "tid": -914061504, "ts": 1716454218262601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262691, "dur": 1,
    "args": {
      "External id": 52635,
      "cbid": 251, "correlation": 52635
    }
  },
  {
    "ph": "f", "id": 52635, "pid": 76337, "tid": -914061504, "ts": 1716454218262691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262696, "dur": 0,
    "args": {
      "External id": 52637,
      "cbid": 251, "correlation": 52637
    }
  },
  {
    "ph": "f", "id": 52637, "pid": 76337, "tid": -914061504, "ts": 1716454218262696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218319255, "dur": 493,
    "args": {
      "External id": 52638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52638,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52638, "pid": 5, "tid": 7, "ts": 1716454218319255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262698, "dur": 14,
    "args": {
      "External id": 52638,
      "cbid": 211, "correlation": 52638
    }
  },
  {
    "ph": "s", "id": 52638, "pid": 76337, "tid": -914061504, "ts": 1716454218262698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218319749, "dur": 67,
    "args": {
      "External id": 52646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52646, "pid": 5, "tid": 7, "ts": 1716454218319749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262766, "dur": 12,
    "args": {
      "External id": 52646,
      "cbid": 211, "correlation": 52646
    }
  },
  {
    "ph": "s", "id": 52646, "pid": 76337, "tid": -914061504, "ts": 1716454218262766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218319817, "dur": 66,
    "args": {
      "External id": 52654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52654,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52654, "pid": 5, "tid": 7, "ts": 1716454218319817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262798, "dur": 9,
    "args": {
      "External id": 52654,
      "cbid": 211, "correlation": 52654
    }
  },
  {
    "ph": "s", "id": 52654, "pid": 76337, "tid": -914061504, "ts": 1716454218262798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262879, "dur": 1,
    "args": {
      "External id": 52670,
      "cbid": 251, "correlation": 52670
    }
  },
  {
    "ph": "f", "id": 52670, "pid": 76337, "tid": -914061504, "ts": 1716454218262879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218319885, "dur": 1,
    "args": {
      "External id": 52672,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 52672,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1563517915309446
    }
  },
  {
    "ph": "f", "id": 52672, "pid": 5, "tid": 7, "ts": 1716454218319885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262884, "dur": 10,
    "args": {
      "External id": 52672,
      "cbid": 51, "correlation": 52672
    }
  },
  {
    "ph": "s", "id": 52672, "pid": 76337, "tid": -914061504, "ts": 1716454218262884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218319888, "dur": 269,
    "args": {
      "External id": 52673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52673,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52673, "pid": 5, "tid": 7, "ts": 1716454218319888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262895, "dur": 11,
    "args": {
      "External id": 52673,
      "cbid": 211, "correlation": 52673
    }
  },
  {
    "ph": "s", "id": 52673, "pid": 76337, "tid": -914061504, "ts": 1716454218262895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218320159, "dur": 15,
    "args": {
      "External id": 52681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52681,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52681, "pid": 5, "tid": 7, "ts": 1716454218320159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218262937, "dur": 10,
    "args": {
      "External id": 52681,
      "cbid": 211, "correlation": 52681
    }
  },
  {
    "ph": "s", "id": 52681, "pid": 76337, "tid": -914061504, "ts": 1716454218262937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218320175, "dur": 37,
    "args": {
      "External id": 52692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52692, "pid": 5, "tid": 7, "ts": 1716454218320175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263017, "dur": 14,
    "args": {
      "External id": 52692,
      "cbid": 211, "correlation": 52692
    }
  },
  {
    "ph": "s", "id": 52692, "pid": 76337, "tid": -914061504, "ts": 1716454218263017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263083, "dur": 0,
    "args": {
      "External id": 52704,
      "cbid": 317, "correlation": 52704
    }
  },
  {
    "ph": "f", "id": 52704, "pid": 76337, "tid": -914061504, "ts": 1716454218263083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263084, "dur": 0,
    "args": {
      "External id": 52705,
      "cbid": 203, "correlation": 52705
    }
  },
  {
    "ph": "f", "id": 52705, "pid": 76337, "tid": -914061504, "ts": 1716454218263084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263085, "dur": 0,
    "args": {
      "External id": 52706,
      "cbid": 205, "correlation": 52706
    }
  },
  {
    "ph": "f", "id": 52706, "pid": 76337, "tid": -914061504, "ts": 1716454218263085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218320214, "dur": 14,
    "args": {
      "External id": 52710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52710,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52710, "pid": 5, "tid": 7, "ts": 1716454218320214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263099, "dur": 12,
    "args": {
      "External id": 52710,
      "cbid": 211, "correlation": 52710
    }
  },
  {
    "ph": "s", "id": 52710, "pid": 76337, "tid": -914061504, "ts": 1716454218263099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218320229, "dur": 4,
    "args": {
      "External id": 52712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52712, "pid": 5, "tid": 7, "ts": 1716454218320229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263116, "dur": 6,
    "args": {
      "External id": 52712,
      "cbid": 211, "correlation": 52712
    }
  },
  {
    "ph": "s", "id": 52712, "pid": 76337, "tid": -914061504, "ts": 1716454218263116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263125, "dur": 0,
    "args": {
      "External id": 52713,
      "cbid": 51, "correlation": 52713
    }
  },
  {
    "ph": "s", "id": 52713, "pid": 76337, "tid": -914061504, "ts": 1716454218263125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218320234, "dur": 96,
    "args": {
      "External id": 52714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52714,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 52714, "pid": 5, "tid": 7, "ts": 1716454218320234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263125, "dur": 6,
    "args": {
      "External id": 52714,
      "cbid": 211, "correlation": 52714
    }
  },
  {
    "ph": "s", "id": 52714, "pid": 76337, "tid": -914061504, "ts": 1716454218263125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218320332, "dur": 16,
    "args": {
      "External id": 52719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52719,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52719, "pid": 5, "tid": 7, "ts": 1716454218320332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263153, "dur": 9,
    "args": {
      "External id": 52719,
      "cbid": 211, "correlation": 52719
    }
  },
  {
    "ph": "s", "id": 52719, "pid": 76337, "tid": -914061504, "ts": 1716454218263153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218320350, "dur": 11,
    "args": {
      "External id": 52727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52727,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52727, "pid": 5, "tid": 7, "ts": 1716454218320350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263185, "dur": 8,
    "args": {
      "External id": 52727,
      "cbid": 211, "correlation": 52727
    }
  },
  {
    "ph": "s", "id": 52727, "pid": 76337, "tid": -914061504, "ts": 1716454218263185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218320362, "dur": 29,
    "args": {
      "External id": 52736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52736,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52736, "pid": 5, "tid": 7, "ts": 1716454218320362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263224, "dur": 11,
    "args": {
      "External id": 52736,
      "cbid": 211, "correlation": 52736
    }
  },
  {
    "ph": "s", "id": 52736, "pid": 76337, "tid": -914061504, "ts": 1716454218263224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218320393, "dur": 30,
    "args": {
      "External id": 52756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52756,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 52756, "pid": 5, "tid": 7, "ts": 1716454218320393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263299, "dur": 12,
    "args": {
      "External id": 52756,
      "cbid": 211, "correlation": 52756
    }
  },
  {
    "ph": "s", "id": 52756, "pid": 76337, "tid": -914061504, "ts": 1716454218263299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218320425, "dur": 5,
    "args": {
      "External id": 52768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52768,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52768, "pid": 5, "tid": 7, "ts": 1716454218320425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263321, "dur": 6,
    "args": {
      "External id": 52768,
      "cbid": 211, "correlation": 52768
    }
  },
  {
    "ph": "s", "id": 52768, "pid": 76337, "tid": -914061504, "ts": 1716454218263321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218320431, "dur": 31,
    "args": {
      "External id": 52771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52771, "pid": 5, "tid": 7, "ts": 1716454218320431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263339, "dur": 7,
    "args": {
      "External id": 52771,
      "cbid": 211, "correlation": 52771
    }
  },
  {
    "ph": "s", "id": 52771, "pid": 76337, "tid": -914061504, "ts": 1716454218263339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218320463, "dur": 21,
    "args": {
      "External id": 52780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52780,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52780, "pid": 5, "tid": 7, "ts": 1716454218320463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263379, "dur": 9,
    "args": {
      "External id": 52780,
      "cbid": 211, "correlation": 52780
    }
  },
  {
    "ph": "s", "id": 52780, "pid": 76337, "tid": -914061504, "ts": 1716454218263379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263432, "dur": 0,
    "args": {
      "External id": 52790,
      "cbid": 317, "correlation": 52790
    }
  },
  {
    "ph": "f", "id": 52790, "pid": 76337, "tid": -914061504, "ts": 1716454218263432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263433, "dur": 0,
    "args": {
      "External id": 52791,
      "cbid": 203, "correlation": 52791
    }
  },
  {
    "ph": "f", "id": 52791, "pid": 76337, "tid": -914061504, "ts": 1716454218263433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263433, "dur": 0,
    "args": {
      "External id": 52792,
      "cbid": 205, "correlation": 52792
    }
  },
  {
    "ph": "f", "id": 52792, "pid": 76337, "tid": -914061504, "ts": 1716454218263433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218320485, "dur": 23,
    "args": {
      "External id": 52796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52796, "pid": 5, "tid": 7, "ts": 1716454218320485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263447, "dur": 11,
    "args": {
      "External id": 52796,
      "cbid": 211, "correlation": 52796
    }
  },
  {
    "ph": "s", "id": 52796, "pid": 76337, "tid": -914061504, "ts": 1716454218263447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218320509, "dur": 317,
    "args": {
      "External id": 52798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52798,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52798, "pid": 5, "tid": 7, "ts": 1716454218320509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263461, "dur": 5,
    "args": {
      "External id": 52798,
      "cbid": 211, "correlation": 52798
    }
  },
  {
    "ph": "s", "id": 52798, "pid": 76337, "tid": -914061504, "ts": 1716454218263461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218320828, "dur": 1,
    "args": {
      "External id": 52800,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 52800,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 52800, "pid": 5, "tid": 7, "ts": 1716454218320828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263473, "dur": 6,
    "args": {
      "External id": 52800,
      "cbid": 51, "correlation": 52800
    }
  },
  {
    "ph": "s", "id": 52800, "pid": 76337, "tid": -914061504, "ts": 1716454218263473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218320832, "dur": 1250,
    "args": {
      "External id": 52801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52801,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52801, "pid": 5, "tid": 7, "ts": 1716454218320832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263480, "dur": 6,
    "args": {
      "External id": 52801,
      "cbid": 211, "correlation": 52801
    }
  },
  {
    "ph": "s", "id": 52801, "pid": 76337, "tid": -914061504, "ts": 1716454218263480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218322083, "dur": 13,
    "args": {
      "External id": 52803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52803,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52803, "pid": 5, "tid": 7, "ts": 1716454218322083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263490, "dur": 5,
    "args": {
      "External id": 52803,
      "cbid": 211, "correlation": 52803
    }
  },
  {
    "ph": "s", "id": 52803, "pid": 76337, "tid": -914061504, "ts": 1716454218263490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218322098, "dur": 15,
    "args": {
      "External id": 52809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52809, "pid": 5, "tid": 7, "ts": 1716454218322098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263518, "dur": 9,
    "args": {
      "External id": 52809,
      "cbid": 211, "correlation": 52809
    }
  },
  {
    "ph": "s", "id": 52809, "pid": 76337, "tid": -914061504, "ts": 1716454218263518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218322114, "dur": 3,
    "args": {
      "External id": 52817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52817,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 52817, "pid": 5, "tid": 7, "ts": 1716454218322114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263565, "dur": 10,
    "args": {
      "External id": 52817,
      "cbid": 211, "correlation": 52817
    }
  },
  {
    "ph": "s", "id": 52817, "pid": 76337, "tid": -914061504, "ts": 1716454218263565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263631, "dur": 1,
    "args": {
      "External id": 52833,
      "cbid": 251, "correlation": 52833
    }
  },
  {
    "ph": "f", "id": 52833, "pid": 76337, "tid": -914061504, "ts": 1716454218263631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263637, "dur": 0,
    "args": {
      "External id": 52835,
      "cbid": 251, "correlation": 52835
    }
  },
  {
    "ph": "f", "id": 52835, "pid": 76337, "tid": -914061504, "ts": 1716454218263637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218322118, "dur": 13,
    "args": {
      "External id": 52836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52836,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52836, "pid": 5, "tid": 7, "ts": 1716454218322118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263639, "dur": 12,
    "args": {
      "External id": 52836,
      "cbid": 211, "correlation": 52836
    }
  },
  {
    "ph": "s", "id": 52836, "pid": 76337, "tid": -914061504, "ts": 1716454218263639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218322133, "dur": 5,
    "args": {
      "External id": 52838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52838,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52838, "pid": 5, "tid": 7, "ts": 1716454218322133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263652, "dur": 5,
    "args": {
      "External id": 52838,
      "cbid": 211, "correlation": 52838
    }
  },
  {
    "ph": "s", "id": 52838, "pid": 76337, "tid": -914061504, "ts": 1716454218263652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218322139, "dur": 16,
    "args": {
      "External id": 52848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52848, "pid": 5, "tid": 7, "ts": 1716454218322139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263710, "dur": 12,
    "args": {
      "External id": 52848,
      "cbid": 211, "correlation": 52848
    }
  },
  {
    "ph": "s", "id": 52848, "pid": 76337, "tid": -914061504, "ts": 1716454218263710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218322156, "dur": 19,
    "args": {
      "External id": 52868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52868,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 52868, "pid": 5, "tid": 7, "ts": 1716454218322156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263777, "dur": 11,
    "args": {
      "External id": 52868,
      "cbid": 211, "correlation": 52868
    }
  },
  {
    "ph": "s", "id": 52868, "pid": 76337, "tid": -914061504, "ts": 1716454218263777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218322177, "dur": 4,
    "args": {
      "External id": 52880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52880,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 52880, "pid": 5, "tid": 7, "ts": 1716454218322177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263797, "dur": 6,
    "args": {
      "External id": 52880,
      "cbid": 211, "correlation": 52880
    }
  },
  {
    "ph": "s", "id": 52880, "pid": 76337, "tid": -914061504, "ts": 1716454218263797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218322182, "dur": 16,
    "args": {
      "External id": 52883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52883, "pid": 5, "tid": 7, "ts": 1716454218322182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263815, "dur": 7,
    "args": {
      "External id": 52883,
      "cbid": 211, "correlation": 52883
    }
  },
  {
    "ph": "s", "id": 52883, "pid": 76337, "tid": -914061504, "ts": 1716454218263815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218322200, "dur": 11,
    "args": {
      "External id": 52892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52892,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52892, "pid": 5, "tid": 7, "ts": 1716454218322200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263859, "dur": 10,
    "args": {
      "External id": 52892,
      "cbid": 211, "correlation": 52892
    }
  },
  {
    "ph": "s", "id": 52892, "pid": 76337, "tid": -914061504, "ts": 1716454218263859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263921, "dur": 0,
    "args": {
      "External id": 52902,
      "cbid": 317, "correlation": 52902
    }
  },
  {
    "ph": "f", "id": 52902, "pid": 76337, "tid": -914061504, "ts": 1716454218263921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263922, "dur": 0,
    "args": {
      "External id": 52903,
      "cbid": 203, "correlation": 52903
    }
  },
  {
    "ph": "f", "id": 52903, "pid": 76337, "tid": -914061504, "ts": 1716454218263922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263923, "dur": 0,
    "args": {
      "External id": 52904,
      "cbid": 205, "correlation": 52904
    }
  },
  {
    "ph": "f", "id": 52904, "pid": 76337, "tid": -914061504, "ts": 1716454218263923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218322212, "dur": 12,
    "args": {
      "External id": 52908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52908,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52908, "pid": 5, "tid": 7, "ts": 1716454218322212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263937, "dur": 12,
    "args": {
      "External id": 52908,
      "cbid": 211, "correlation": 52908
    }
  },
  {
    "ph": "s", "id": 52908, "pid": 76337, "tid": -914061504, "ts": 1716454218263937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218322225, "dur": 162,
    "args": {
      "External id": 52910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52910,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52910, "pid": 5, "tid": 7, "ts": 1716454218322225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263952, "dur": 5,
    "args": {
      "External id": 52910,
      "cbid": 211, "correlation": 52910
    }
  },
  {
    "ph": "s", "id": 52910, "pid": 76337, "tid": -914061504, "ts": 1716454218263952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218322389, "dur": 1,
    "args": {
      "External id": 52912,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 52912,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 52912, "pid": 5, "tid": 7, "ts": 1716454218322389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263962, "dur": 7,
    "args": {
      "External id": 52912,
      "cbid": 51, "correlation": 52912
    }
  },
  {
    "ph": "s", "id": 52912, "pid": 76337, "tid": -914061504, "ts": 1716454218263962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218322393, "dur": 644,
    "args": {
      "External id": 52913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52913,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 52913, "pid": 5, "tid": 7, "ts": 1716454218322393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263970, "dur": 13,
    "args": {
      "External id": 52913,
      "cbid": 211, "correlation": 52913
    }
  },
  {
    "ph": "s", "id": 52913, "pid": 76337, "tid": -914061504, "ts": 1716454218263970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218323039, "dur": 12,
    "args": {
      "External id": 52915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52915, "pid": 5, "tid": 7, "ts": 1716454218323039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218263988, "dur": 6,
    "args": {
      "External id": 52915,
      "cbid": 211, "correlation": 52915
    }
  },
  {
    "ph": "s", "id": 52915, "pid": 76337, "tid": -914061504, "ts": 1716454218263988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218323052, "dur": 15,
    "args": {
      "External id": 52921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52921, "pid": 5, "tid": 7, "ts": 1716454218323052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264017, "dur": 8,
    "args": {
      "External id": 52921,
      "cbid": 211, "correlation": 52921
    }
  },
  {
    "ph": "s", "id": 52921, "pid": 76337, "tid": -914061504, "ts": 1716454218264017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264076, "dur": 0,
    "args": {
      "External id": 52931,
      "cbid": 317, "correlation": 52931
    }
  },
  {
    "ph": "f", "id": 52931, "pid": 76337, "tid": -914061504, "ts": 1716454218264076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264076, "dur": 0,
    "args": {
      "External id": 52932,
      "cbid": 203, "correlation": 52932
    }
  },
  {
    "ph": "f", "id": 52932, "pid": 76337, "tid": -914061504, "ts": 1716454218264076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264077, "dur": 0,
    "args": {
      "External id": 52933,
      "cbid": 205, "correlation": 52933
    }
  },
  {
    "ph": "f", "id": 52933, "pid": 76337, "tid": -914061504, "ts": 1716454218264077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218323068, "dur": 21,
    "args": {
      "External id": 52937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52937,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52937, "pid": 5, "tid": 7, "ts": 1716454218323068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264089, "dur": 12,
    "args": {
      "External id": 52937,
      "cbid": 211, "correlation": 52937
    }
  },
  {
    "ph": "s", "id": 52937, "pid": 76337, "tid": -914061504, "ts": 1716454218264089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218323090, "dur": 4,
    "args": {
      "External id": 52939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52939,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 52939, "pid": 5, "tid": 7, "ts": 1716454218323090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264105, "dur": 6,
    "args": {
      "External id": 52939,
      "cbid": 211, "correlation": 52939
    }
  },
  {
    "ph": "s", "id": 52939, "pid": 76337, "tid": -914061504, "ts": 1716454218264105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264114, "dur": 0,
    "args": {
      "External id": 52940,
      "cbid": 51, "correlation": 52940
    }
  },
  {
    "ph": "s", "id": 52940, "pid": 76337, "tid": -914061504, "ts": 1716454218264114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218323095, "dur": 169,
    "args": {
      "External id": 52941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52941,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 52941, "pid": 5, "tid": 7, "ts": 1716454218323095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264115, "dur": 5,
    "args": {
      "External id": 52941,
      "cbid": 211, "correlation": 52941
    }
  },
  {
    "ph": "s", "id": 52941, "pid": 76337, "tid": -914061504, "ts": 1716454218264115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218323266, "dur": 15,
    "args": {
      "External id": 52946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52946, "pid": 5, "tid": 7, "ts": 1716454218323266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264140, "dur": 11,
    "args": {
      "External id": 52946,
      "cbid": 211, "correlation": 52946
    }
  },
  {
    "ph": "s", "id": 52946, "pid": 76337, "tid": -914061504, "ts": 1716454218264140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218323282, "dur": 12,
    "args": {
      "External id": 52954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52954,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52954, "pid": 5, "tid": 7, "ts": 1716454218323282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264172, "dur": 9,
    "args": {
      "External id": 52954,
      "cbid": 211, "correlation": 52954
    }
  },
  {
    "ph": "s", "id": 52954, "pid": 76337, "tid": -914061504, "ts": 1716454218264172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218323296, "dur": 11,
    "args": {
      "External id": 52962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52962,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52962, "pid": 5, "tid": 7, "ts": 1716454218323296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264201, "dur": 8,
    "args": {
      "External id": 52962,
      "cbid": 211, "correlation": 52962
    }
  },
  {
    "ph": "s", "id": 52962, "pid": 76337, "tid": -914061504, "ts": 1716454218264201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218323308, "dur": 18,
    "args": {
      "External id": 52982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52982,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 52982, "pid": 5, "tid": 7, "ts": 1716454218323308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264283, "dur": 13,
    "args": {
      "External id": 52982,
      "cbid": 211, "correlation": 52982
    }
  },
  {
    "ph": "s", "id": 52982, "pid": 76337, "tid": -914061504, "ts": 1716454218264283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218323327, "dur": 4,
    "args": {
      "External id": 52994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52994,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 52994, "pid": 5, "tid": 7, "ts": 1716454218323327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264305, "dur": 6,
    "args": {
      "External id": 52994,
      "cbid": 211, "correlation": 52994
    }
  },
  {
    "ph": "s", "id": 52994, "pid": 76337, "tid": -914061504, "ts": 1716454218264305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218323333, "dur": 16,
    "args": {
      "External id": 52997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 52997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 52997, "pid": 5, "tid": 7, "ts": 1716454218323333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264322, "dur": 6,
    "args": {
      "External id": 52997,
      "cbid": 211, "correlation": 52997
    }
  },
  {
    "ph": "s", "id": 52997, "pid": 76337, "tid": -914061504, "ts": 1716454218264322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264380, "dur": 0,
    "args": {
      "External id": 53008,
      "cbid": 317, "correlation": 53008
    }
  },
  {
    "ph": "f", "id": 53008, "pid": 76337, "tid": -914061504, "ts": 1716454218264380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264381, "dur": 0,
    "args": {
      "External id": 53009,
      "cbid": 203, "correlation": 53009
    }
  },
  {
    "ph": "f", "id": 53009, "pid": 76337, "tid": -914061504, "ts": 1716454218264381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264381, "dur": 0,
    "args": {
      "External id": 53010,
      "cbid": 205, "correlation": 53010
    }
  },
  {
    "ph": "f", "id": 53010, "pid": 76337, "tid": -914061504, "ts": 1716454218264381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218323351, "dur": 12,
    "args": {
      "External id": 53014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53014, "pid": 5, "tid": 7, "ts": 1716454218323351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264394, "dur": 12,
    "args": {
      "External id": 53014,
      "cbid": 211, "correlation": 53014
    }
  },
  {
    "ph": "s", "id": 53014, "pid": 76337, "tid": -914061504, "ts": 1716454218264394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218323364, "dur": 4,
    "args": {
      "External id": 53016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53016, "pid": 5, "tid": 7, "ts": 1716454218323364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264411, "dur": 8,
    "args": {
      "External id": 53016,
      "cbid": 211, "correlation": 53016
    }
  },
  {
    "ph": "s", "id": 53016, "pid": 76337, "tid": -914061504, "ts": 1716454218264411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264422, "dur": 0,
    "args": {
      "External id": 53017,
      "cbid": 51, "correlation": 53017
    }
  },
  {
    "ph": "s", "id": 53017, "pid": 76337, "tid": -914061504, "ts": 1716454218264422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218323369, "dur": 89,
    "args": {
      "External id": 53018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53018,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 53018, "pid": 5, "tid": 7, "ts": 1716454218323369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264423, "dur": 5,
    "args": {
      "External id": 53018,
      "cbid": 211, "correlation": 53018
    }
  },
  {
    "ph": "s", "id": 53018, "pid": 76337, "tid": -914061504, "ts": 1716454218264423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218323459, "dur": 15,
    "args": {
      "External id": 53023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53023, "pid": 5, "tid": 7, "ts": 1716454218323459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264450, "dur": 9,
    "args": {
      "External id": 53023,
      "cbid": 211, "correlation": 53023
    }
  },
  {
    "ph": "s", "id": 53023, "pid": 76337, "tid": -914061504, "ts": 1716454218264450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218323476, "dur": 83,
    "args": {
      "External id": 53032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53032,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53032, "pid": 5, "tid": 7, "ts": 1716454218323476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264531, "dur": 13,
    "args": {
      "External id": 53032,
      "cbid": 211, "correlation": 53032
    }
  },
  {
    "ph": "s", "id": 53032, "pid": 76337, "tid": -914061504, "ts": 1716454218264531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218323560, "dur": 30,
    "args": {
      "External id": 53054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53054,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53054, "pid": 5, "tid": 7, "ts": 1716454218323560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264587, "dur": 10,
    "args": {
      "External id": 53054,
      "cbid": 211, "correlation": 53054
    }
  },
  {
    "ph": "s", "id": 53054, "pid": 76337, "tid": -914061504, "ts": 1716454218264587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264675, "dur": 1,
    "args": {
      "External id": 53065,
      "cbid": 251, "correlation": 53065
    }
  },
  {
    "ph": "f", "id": 53065, "pid": 76337, "tid": -914061504, "ts": 1716454218264675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218323592, "dur": 162,
    "args": {
      "External id": 53066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53066,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53066, "pid": 5, "tid": 7, "ts": 1716454218323592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264680, "dur": 13,
    "args": {
      "External id": 53066,
      "cbid": 211, "correlation": 53066
    }
  },
  {
    "ph": "s", "id": 53066, "pid": 76337, "tid": -914061504, "ts": 1716454218264680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264751, "dur": 1,
    "args": {
      "External id": 53077,
      "cbid": 251, "correlation": 53077
    }
  },
  {
    "ph": "f", "id": 53077, "pid": 76337, "tid": -914061504, "ts": 1716454218264751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218323755, "dur": 159,
    "args": {
      "External id": 53078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53078,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53078, "pid": 5, "tid": 7, "ts": 1716454218323755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264755, "dur": 11,
    "args": {
      "External id": 53078,
      "cbid": 211, "correlation": 53078
    }
  },
  {
    "ph": "s", "id": 53078, "pid": 76337, "tid": -914061504, "ts": 1716454218264755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264821, "dur": 3,
    "args": {
      "External id": 53089,
      "cbid": 251, "correlation": 53089
    }
  },
  {
    "ph": "f", "id": 53089, "pid": 76337, "tid": -914061504, "ts": 1716454218264821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218323915, "dur": 158,
    "args": {
      "External id": 53090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53090,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53090, "pid": 5, "tid": 7, "ts": 1716454218323915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264827, "dur": 11,
    "args": {
      "External id": 53090,
      "cbid": 211, "correlation": 53090
    }
  },
  {
    "ph": "s", "id": 53090, "pid": 76337, "tid": -914061504, "ts": 1716454218264827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218324075, "dur": 333,
    "args": {
      "External id": 53115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53115,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53115, "pid": 5, "tid": 7, "ts": 1716454218324075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218264911, "dur": 14,
    "args": {
      "External id": 53115,
      "cbid": 211, "correlation": 53115
    }
  },
  {
    "ph": "s", "id": 53115, "pid": 76337, "tid": -914061504, "ts": 1716454218264911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265019, "dur": 1,
    "args": {
      "External id": 53133,
      "cbid": 251, "correlation": 53133
    }
  },
  {
    "ph": "f", "id": 53133, "pid": 76337, "tid": -914061504, "ts": 1716454218265019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218324409, "dur": 166,
    "args": {
      "External id": 53135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53135,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53135, "pid": 5, "tid": 7, "ts": 1716454218324409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265025, "dur": 14,
    "args": {
      "External id": 53135,
      "cbid": 211, "correlation": 53135
    }
  },
  {
    "ph": "s", "id": 53135, "pid": 76337, "tid": -914061504, "ts": 1716454218265025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218324576, "dur": 20,
    "args": {
      "External id": 53143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53143,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53143, "pid": 5, "tid": 7, "ts": 1716454218324576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265097, "dur": 12,
    "args": {
      "External id": 53143,
      "cbid": 211, "correlation": 53143
    }
  },
  {
    "ph": "s", "id": 53143, "pid": 76337, "tid": -914061504, "ts": 1716454218265097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218324597, "dur": 27,
    "args": {
      "External id": 53151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53151,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53151, "pid": 5, "tid": 7, "ts": 1716454218324597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265136, "dur": 8,
    "args": {
      "External id": 53151,
      "cbid": 211, "correlation": 53151
    }
  },
  {
    "ph": "s", "id": 53151, "pid": 76337, "tid": -914061504, "ts": 1716454218265136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218324626, "dur": 18,
    "args": {
      "External id": 53162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53162, "pid": 5, "tid": 7, "ts": 1716454218324626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265206, "dur": 13,
    "args": {
      "External id": 53162,
      "cbid": 211, "correlation": 53162
    }
  },
  {
    "ph": "s", "id": 53162, "pid": 76337, "tid": -914061504, "ts": 1716454218265206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218324645, "dur": 16,
    "args": {
      "External id": 53184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53184,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53184, "pid": 5, "tid": 7, "ts": 1716454218324645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265240, "dur": 7,
    "args": {
      "External id": 53184,
      "cbid": 211, "correlation": 53184
    }
  },
  {
    "ph": "s", "id": 53184, "pid": 76337, "tid": -914061504, "ts": 1716454218265240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265326, "dur": 1,
    "args": {
      "External id": 53195,
      "cbid": 251, "correlation": 53195
    }
  },
  {
    "ph": "f", "id": 53195, "pid": 76337, "tid": -914061504, "ts": 1716454218265326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218324663, "dur": 88,
    "args": {
      "External id": 53196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53196,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53196, "pid": 5, "tid": 7, "ts": 1716454218324663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265332, "dur": 13,
    "args": {
      "External id": 53196,
      "cbid": 211, "correlation": 53196
    }
  },
  {
    "ph": "s", "id": 53196, "pid": 76337, "tid": -914061504, "ts": 1716454218265332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265401, "dur": 1,
    "args": {
      "External id": 53207,
      "cbid": 251, "correlation": 53207
    }
  },
  {
    "ph": "f", "id": 53207, "pid": 76337, "tid": -914061504, "ts": 1716454218265401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265405, "dur": 0,
    "args": {
      "External id": 53208,
      "cbid": 251, "correlation": 53208
    }
  },
  {
    "ph": "f", "id": 53208, "pid": 76337, "tid": -914061504, "ts": 1716454218265405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218324752, "dur": 13,
    "args": {
      "External id": 53209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53209,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53209, "pid": 5, "tid": 7, "ts": 1716454218324752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265406, "dur": 12,
    "args": {
      "External id": 53209,
      "cbid": 211, "correlation": 53209
    }
  },
  {
    "ph": "s", "id": 53209, "pid": 76337, "tid": -914061504, "ts": 1716454218265406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218324766, "dur": 5,
    "args": {
      "External id": 53211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53211,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53211, "pid": 5, "tid": 7, "ts": 1716454218324766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265420, "dur": 6,
    "args": {
      "External id": 53211,
      "cbid": 211, "correlation": 53211
    }
  },
  {
    "ph": "s", "id": 53211, "pid": 76337, "tid": -914061504, "ts": 1716454218265420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265477, "dur": 1,
    "args": {
      "External id": 53222,
      "cbid": 251, "correlation": 53222
    }
  },
  {
    "ph": "f", "id": 53222, "pid": 76337, "tid": -914061504, "ts": 1716454218265477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265480, "dur": 0,
    "args": {
      "External id": 53223,
      "cbid": 251, "correlation": 53223
    }
  },
  {
    "ph": "f", "id": 53223, "pid": 76337, "tid": -914061504, "ts": 1716454218265480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218324772, "dur": 8,
    "args": {
      "External id": 53224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53224,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53224, "pid": 5, "tid": 7, "ts": 1716454218324772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265482, "dur": 11,
    "args": {
      "External id": 53224,
      "cbid": 211, "correlation": 53224
    }
  },
  {
    "ph": "s", "id": 53224, "pid": 76337, "tid": -914061504, "ts": 1716454218265482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218324782, "dur": 3,
    "args": {
      "External id": 53226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53226,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53226, "pid": 5, "tid": 7, "ts": 1716454218324782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265494, "dur": 6,
    "args": {
      "External id": 53226,
      "cbid": 211, "correlation": 53226
    }
  },
  {
    "ph": "s", "id": 53226, "pid": 76337, "tid": -914061504, "ts": 1716454218265494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218324787, "dur": 55,
    "args": {
      "External id": 53251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53251,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53251, "pid": 5, "tid": 7, "ts": 1716454218324787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265573, "dur": 13,
    "args": {
      "External id": 53251,
      "cbid": 211, "correlation": 53251
    }
  },
  {
    "ph": "s", "id": 53251, "pid": 76337, "tid": -914061504, "ts": 1716454218265573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265672, "dur": 1,
    "args": {
      "External id": 53269,
      "cbid": 251, "correlation": 53269
    }
  },
  {
    "ph": "f", "id": 53269, "pid": 76337, "tid": -914061504, "ts": 1716454218265672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218324843, "dur": 91,
    "args": {
      "External id": 53271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53271,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53271, "pid": 5, "tid": 7, "ts": 1716454218324843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265678, "dur": 14,
    "args": {
      "External id": 53271,
      "cbid": 211, "correlation": 53271
    }
  },
  {
    "ph": "s", "id": 53271, "pid": 76337, "tid": -914061504, "ts": 1716454218265678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218324935, "dur": 9,
    "args": {
      "External id": 53279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53279,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53279, "pid": 5, "tid": 7, "ts": 1716454218324935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265747, "dur": 12,
    "args": {
      "External id": 53279,
      "cbid": 211, "correlation": 53279
    }
  },
  {
    "ph": "s", "id": 53279, "pid": 76337, "tid": -914061504, "ts": 1716454218265747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218324945, "dur": 20,
    "args": {
      "External id": 53287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53287, "pid": 5, "tid": 7, "ts": 1716454218324945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265789, "dur": 10,
    "args": {
      "External id": 53287,
      "cbid": 211, "correlation": 53287
    }
  },
  {
    "ph": "s", "id": 53287, "pid": 76337, "tid": -914061504, "ts": 1716454218265789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218324967, "dur": 17,
    "args": {
      "External id": 53309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53309,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53309, "pid": 5, "tid": 7, "ts": 1716454218324967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265839, "dur": 10,
    "args": {
      "External id": 53309,
      "cbid": 211, "correlation": 53309
    }
  },
  {
    "ph": "s", "id": 53309, "pid": 76337, "tid": -914061504, "ts": 1716454218265839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265930, "dur": 1,
    "args": {
      "External id": 53325,
      "cbid": 251, "correlation": 53325
    }
  },
  {
    "ph": "f", "id": 53325, "pid": 76337, "tid": -914061504, "ts": 1716454218265930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265935, "dur": 0,
    "args": {
      "External id": 53327,
      "cbid": 251, "correlation": 53327
    }
  },
  {
    "ph": "f", "id": 53327, "pid": 76337, "tid": -914061504, "ts": 1716454218265935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218324986, "dur": 496,
    "args": {
      "External id": 53328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53328,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53328, "pid": 5, "tid": 7, "ts": 1716454218324986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218265937, "dur": 13,
    "args": {
      "External id": 53328,
      "cbid": 211, "correlation": 53328
    }
  },
  {
    "ph": "s", "id": 53328, "pid": 76337, "tid": -914061504, "ts": 1716454218265937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218325483, "dur": 66,
    "args": {
      "External id": 53336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53336,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53336, "pid": 5, "tid": 7, "ts": 1716454218325483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266011, "dur": 13,
    "args": {
      "External id": 53336,
      "cbid": 211, "correlation": 53336
    }
  },
  {
    "ph": "s", "id": 53336, "pid": 76337, "tid": -914061504, "ts": 1716454218266011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218325550, "dur": 66,
    "args": {
      "External id": 53344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53344,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53344, "pid": 5, "tid": 7, "ts": 1716454218325550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266042, "dur": 8,
    "args": {
      "External id": 53344,
      "cbid": 211, "correlation": 53344
    }
  },
  {
    "ph": "s", "id": 53344, "pid": 76337, "tid": -914061504, "ts": 1716454218266042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266123, "dur": 1,
    "args": {
      "External id": 53360,
      "cbid": 251, "correlation": 53360
    }
  },
  {
    "ph": "f", "id": 53360, "pid": 76337, "tid": -914061504, "ts": 1716454218266123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218325618, "dur": 1,
    "args": {
      "External id": 53362,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 53362,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 53362, "pid": 5, "tid": 7, "ts": 1716454218325618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266128, "dur": 9,
    "args": {
      "External id": 53362,
      "cbid": 51, "correlation": 53362
    }
  },
  {
    "ph": "s", "id": 53362, "pid": 76337, "tid": -914061504, "ts": 1716454218266128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218325622, "dur": 269,
    "args": {
      "External id": 53363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53363,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53363, "pid": 5, "tid": 7, "ts": 1716454218325622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266139, "dur": 12,
    "args": {
      "External id": 53363,
      "cbid": 211, "correlation": 53363
    }
  },
  {
    "ph": "s", "id": 53363, "pid": 76337, "tid": -914061504, "ts": 1716454218266139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218325892, "dur": 13,
    "args": {
      "External id": 53371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53371,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53371, "pid": 5, "tid": 7, "ts": 1716454218325892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266181, "dur": 10,
    "args": {
      "External id": 53371,
      "cbid": 211, "correlation": 53371
    }
  },
  {
    "ph": "s", "id": 53371, "pid": 76337, "tid": -914061504, "ts": 1716454218266181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218325907, "dur": 37,
    "args": {
      "External id": 53382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53382, "pid": 5, "tid": 7, "ts": 1716454218325907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266249, "dur": 12,
    "args": {
      "External id": 53382,
      "cbid": 211, "correlation": 53382
    }
  },
  {
    "ph": "s", "id": 53382, "pid": 76337, "tid": -914061504, "ts": 1716454218266249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266313, "dur": 0,
    "args": {
      "External id": 53394,
      "cbid": 317, "correlation": 53394
    }
  },
  {
    "ph": "f", "id": 53394, "pid": 76337, "tid": -914061504, "ts": 1716454218266313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266314, "dur": 0,
    "args": {
      "External id": 53395,
      "cbid": 203, "correlation": 53395
    }
  },
  {
    "ph": "f", "id": 53395, "pid": 76337, "tid": -914061504, "ts": 1716454218266314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266315, "dur": 3,
    "args": {
      "External id": 53396,
      "cbid": 205, "correlation": 53396
    }
  },
  {
    "ph": "f", "id": 53396, "pid": 76337, "tid": -914061504, "ts": 1716454218266315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218325945, "dur": 15,
    "args": {
      "External id": 53400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53400,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53400, "pid": 5, "tid": 7, "ts": 1716454218325945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266334, "dur": 12,
    "args": {
      "External id": 53400,
      "cbid": 211, "correlation": 53400
    }
  },
  {
    "ph": "s", "id": 53400, "pid": 76337, "tid": -914061504, "ts": 1716454218266334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218325962, "dur": 4,
    "args": {
      "External id": 53402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53402,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53402, "pid": 5, "tid": 7, "ts": 1716454218325962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266350, "dur": 6,
    "args": {
      "External id": 53402,
      "cbid": 211, "correlation": 53402
    }
  },
  {
    "ph": "s", "id": 53402, "pid": 76337, "tid": -914061504, "ts": 1716454218266350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266359, "dur": 0,
    "args": {
      "External id": 53403,
      "cbid": 51, "correlation": 53403
    }
  },
  {
    "ph": "s", "id": 53403, "pid": 76337, "tid": -914061504, "ts": 1716454218266359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218325967, "dur": 96,
    "args": {
      "External id": 53404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53404,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 53404, "pid": 5, "tid": 7, "ts": 1716454218325967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266359, "dur": 5,
    "args": {
      "External id": 53404,
      "cbid": 211, "correlation": 53404
    }
  },
  {
    "ph": "s", "id": 53404, "pid": 76337, "tid": -914061504, "ts": 1716454218266359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218326065, "dur": 16,
    "args": {
      "External id": 53409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53409, "pid": 5, "tid": 7, "ts": 1716454218326065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266387, "dur": 8,
    "args": {
      "External id": 53409,
      "cbid": 211, "correlation": 53409
    }
  },
  {
    "ph": "s", "id": 53409, "pid": 76337, "tid": -914061504, "ts": 1716454218266387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218326082, "dur": 12,
    "args": {
      "External id": 53417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53417,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53417, "pid": 5, "tid": 7, "ts": 1716454218326082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266419, "dur": 9,
    "args": {
      "External id": 53417,
      "cbid": 211, "correlation": 53417
    }
  },
  {
    "ph": "s", "id": 53417, "pid": 76337, "tid": -914061504, "ts": 1716454218266419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218326095, "dur": 25,
    "args": {
      "External id": 53426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53426,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53426, "pid": 5, "tid": 7, "ts": 1716454218326095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266457, "dur": 10,
    "args": {
      "External id": 53426,
      "cbid": 211, "correlation": 53426
    }
  },
  {
    "ph": "s", "id": 53426, "pid": 76337, "tid": -914061504, "ts": 1716454218266457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218326122, "dur": 24,
    "args": {
      "External id": 53446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53446,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 53446, "pid": 5, "tid": 7, "ts": 1716454218326122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266529, "dur": 11,
    "args": {
      "External id": 53446,
      "cbid": 211, "correlation": 53446
    }
  },
  {
    "ph": "s", "id": 53446, "pid": 76337, "tid": -914061504, "ts": 1716454218266529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218326147, "dur": 5,
    "args": {
      "External id": 53458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53458,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 53458, "pid": 5, "tid": 7, "ts": 1716454218326147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266550, "dur": 6,
    "args": {
      "External id": 53458,
      "cbid": 211, "correlation": 53458
    }
  },
  {
    "ph": "s", "id": 53458, "pid": 76337, "tid": -914061504, "ts": 1716454218266550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218326153, "dur": 24,
    "args": {
      "External id": 53461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53461,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53461, "pid": 5, "tid": 7, "ts": 1716454218326153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266568, "dur": 10,
    "args": {
      "External id": 53461,
      "cbid": 211, "correlation": 53461
    }
  },
  {
    "ph": "s", "id": 53461, "pid": 76337, "tid": -914061504, "ts": 1716454218266568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218326178, "dur": 17,
    "args": {
      "External id": 53470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53470,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53470, "pid": 5, "tid": 7, "ts": 1716454218326178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266612, "dur": 10,
    "args": {
      "External id": 53470,
      "cbid": 211, "correlation": 53470
    }
  },
  {
    "ph": "s", "id": 53470, "pid": 76337, "tid": -914061504, "ts": 1716454218266612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266664, "dur": 0,
    "args": {
      "External id": 53480,
      "cbid": 317, "correlation": 53480
    }
  },
  {
    "ph": "f", "id": 53480, "pid": 76337, "tid": -914061504, "ts": 1716454218266664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266665, "dur": 0,
    "args": {
      "External id": 53481,
      "cbid": 203, "correlation": 53481
    }
  },
  {
    "ph": "f", "id": 53481, "pid": 76337, "tid": -914061504, "ts": 1716454218266665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266666, "dur": 0,
    "args": {
      "External id": 53482,
      "cbid": 205, "correlation": 53482
    }
  },
  {
    "ph": "f", "id": 53482, "pid": 76337, "tid": -914061504, "ts": 1716454218266666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218326196, "dur": 17,
    "args": {
      "External id": 53486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53486,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53486, "pid": 5, "tid": 7, "ts": 1716454218326196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266681, "dur": 12,
    "args": {
      "External id": 53486,
      "cbid": 211, "correlation": 53486
    }
  },
  {
    "ph": "s", "id": 53486, "pid": 76337, "tid": -914061504, "ts": 1716454218266681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218326215, "dur": 239,
    "args": {
      "External id": 53488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53488,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53488, "pid": 5, "tid": 7, "ts": 1716454218326215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266695, "dur": 5,
    "args": {
      "External id": 53488,
      "cbid": 211, "correlation": 53488
    }
  },
  {
    "ph": "s", "id": 53488, "pid": 76337, "tid": -914061504, "ts": 1716454218266695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218326456, "dur": 1,
    "args": {
      "External id": 53490,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 53490,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 53490, "pid": 5, "tid": 7, "ts": 1716454218326456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266707, "dur": 8,
    "args": {
      "External id": 53490,
      "cbid": 51, "correlation": 53490
    }
  },
  {
    "ph": "s", "id": 53490, "pid": 76337, "tid": -914061504, "ts": 1716454218266707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218326460, "dur": 806,
    "args": {
      "External id": 53491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53491,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53491, "pid": 5, "tid": 7, "ts": 1716454218326460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266716, "dur": 6,
    "args": {
      "External id": 53491,
      "cbid": 211, "correlation": 53491
    }
  },
  {
    "ph": "s", "id": 53491, "pid": 76337, "tid": -914061504, "ts": 1716454218266716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218327267, "dur": 14,
    "args": {
      "External id": 53493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53493,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53493, "pid": 5, "tid": 7, "ts": 1716454218327267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266726, "dur": 5,
    "args": {
      "External id": 53493,
      "cbid": 211, "correlation": 53493
    }
  },
  {
    "ph": "s", "id": 53493, "pid": 76337, "tid": -914061504, "ts": 1716454218266726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218327282, "dur": 14,
    "args": {
      "External id": 53499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53499, "pid": 5, "tid": 7, "ts": 1716454218327282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266755, "dur": 8,
    "args": {
      "External id": 53499,
      "cbid": 211, "correlation": 53499
    }
  },
  {
    "ph": "s", "id": 53499, "pid": 76337, "tid": -914061504, "ts": 1716454218266755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218327298, "dur": 4,
    "args": {
      "External id": 53507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53507,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 53507, "pid": 5, "tid": 7, "ts": 1716454218327298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266799, "dur": 9,
    "args": {
      "External id": 53507,
      "cbid": 211, "correlation": 53507
    }
  },
  {
    "ph": "s", "id": 53507, "pid": 76337, "tid": -914061504, "ts": 1716454218266799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266864, "dur": 1,
    "args": {
      "External id": 53523,
      "cbid": 251, "correlation": 53523
    }
  },
  {
    "ph": "f", "id": 53523, "pid": 76337, "tid": -914061504, "ts": 1716454218266864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266869, "dur": 0,
    "args": {
      "External id": 53525,
      "cbid": 251, "correlation": 53525
    }
  },
  {
    "ph": "f", "id": 53525, "pid": 76337, "tid": -914061504, "ts": 1716454218266869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218327303, "dur": 13,
    "args": {
      "External id": 53526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53526,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53526, "pid": 5, "tid": 7, "ts": 1716454218327303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266871, "dur": 14,
    "args": {
      "External id": 53526,
      "cbid": 211, "correlation": 53526
    }
  },
  {
    "ph": "s", "id": 53526, "pid": 76337, "tid": -914061504, "ts": 1716454218266871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218327318, "dur": 5,
    "args": {
      "External id": 53528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53528,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53528, "pid": 5, "tid": 7, "ts": 1716454218327318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266887, "dur": 6,
    "args": {
      "External id": 53528,
      "cbid": 211, "correlation": 53528
    }
  },
  {
    "ph": "s", "id": 53528, "pid": 76337, "tid": -914061504, "ts": 1716454218266887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218327324, "dur": 17,
    "args": {
      "External id": 53538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53538, "pid": 5, "tid": 7, "ts": 1716454218327324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218266945, "dur": 12,
    "args": {
      "External id": 53538,
      "cbid": 211, "correlation": 53538
    }
  },
  {
    "ph": "s", "id": 53538, "pid": 76337, "tid": -914061504, "ts": 1716454218266945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218327342, "dur": 18,
    "args": {
      "External id": 53558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53558,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 53558, "pid": 5, "tid": 7, "ts": 1716454218327342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267026, "dur": 11,
    "args": {
      "External id": 53558,
      "cbid": 211, "correlation": 53558
    }
  },
  {
    "ph": "s", "id": 53558, "pid": 76337, "tid": -914061504, "ts": 1716454218267026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218327361, "dur": 4,
    "args": {
      "External id": 53570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53570,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 53570, "pid": 5, "tid": 7, "ts": 1716454218327361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267047, "dur": 6,
    "args": {
      "External id": 53570,
      "cbid": 211, "correlation": 53570
    }
  },
  {
    "ph": "s", "id": 53570, "pid": 76337, "tid": -914061504, "ts": 1716454218267047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218327367, "dur": 16,
    "args": {
      "External id": 53573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53573, "pid": 5, "tid": 7, "ts": 1716454218327367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267066, "dur": 7,
    "args": {
      "External id": 53573,
      "cbid": 211, "correlation": 53573
    }
  },
  {
    "ph": "s", "id": 53573, "pid": 76337, "tid": -914061504, "ts": 1716454218267066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218327384, "dur": 11,
    "args": {
      "External id": 53582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53582,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53582, "pid": 5, "tid": 7, "ts": 1716454218327384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267107, "dur": 10,
    "args": {
      "External id": 53582,
      "cbid": 211, "correlation": 53582
    }
  },
  {
    "ph": "s", "id": 53582, "pid": 76337, "tid": -914061504, "ts": 1716454218267107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267171, "dur": 0,
    "args": {
      "External id": 53592,
      "cbid": 317, "correlation": 53592
    }
  },
  {
    "ph": "f", "id": 53592, "pid": 76337, "tid": -914061504, "ts": 1716454218267171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267172, "dur": 0,
    "args": {
      "External id": 53593,
      "cbid": 203, "correlation": 53593
    }
  },
  {
    "ph": "f", "id": 53593, "pid": 76337, "tid": -914061504, "ts": 1716454218267172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267172, "dur": 0,
    "args": {
      "External id": 53594,
      "cbid": 205, "correlation": 53594
    }
  },
  {
    "ph": "f", "id": 53594, "pid": 76337, "tid": -914061504, "ts": 1716454218267172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218327396, "dur": 11,
    "args": {
      "External id": 53598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53598, "pid": 5, "tid": 7, "ts": 1716454218327396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267186, "dur": 14,
    "args": {
      "External id": 53598,
      "cbid": 211, "correlation": 53598
    }
  },
  {
    "ph": "s", "id": 53598, "pid": 76337, "tid": -914061504, "ts": 1716454218267186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218327408, "dur": 161,
    "args": {
      "External id": 53600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53600,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53600, "pid": 5, "tid": 7, "ts": 1716454218327408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267203, "dur": 5,
    "args": {
      "External id": 53600,
      "cbid": 211, "correlation": 53600
    }
  },
  {
    "ph": "s", "id": 53600, "pid": 76337, "tid": -914061504, "ts": 1716454218267203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218327572, "dur": 1,
    "args": {
      "External id": 53602,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 53602,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 53602, "pid": 5, "tid": 7, "ts": 1716454218327572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267214, "dur": 7,
    "args": {
      "External id": 53602,
      "cbid": 51, "correlation": 53602
    }
  },
  {
    "ph": "s", "id": 53602, "pid": 76337, "tid": -914061504, "ts": 1716454218267214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218327575, "dur": 645,
    "args": {
      "External id": 53603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53603,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53603, "pid": 5, "tid": 7, "ts": 1716454218327575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267222, "dur": 6,
    "args": {
      "External id": 53603,
      "cbid": 211, "correlation": 53603
    }
  },
  {
    "ph": "s", "id": 53603, "pid": 76337, "tid": -914061504, "ts": 1716454218267222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218328222, "dur": 13,
    "args": {
      "External id": 53605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53605,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53605, "pid": 5, "tid": 7, "ts": 1716454218328222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267232, "dur": 5,
    "args": {
      "External id": 53605,
      "cbid": 211, "correlation": 53605
    }
  },
  {
    "ph": "s", "id": 53605, "pid": 76337, "tid": -914061504, "ts": 1716454218267232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218328236, "dur": 15,
    "args": {
      "External id": 53611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53611, "pid": 5, "tid": 7, "ts": 1716454218328236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267260, "dur": 9,
    "args": {
      "External id": 53611,
      "cbid": 211, "correlation": 53611
    }
  },
  {
    "ph": "s", "id": 53611, "pid": 76337, "tid": -914061504, "ts": 1716454218267260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267320, "dur": 0,
    "args": {
      "External id": 53621,
      "cbid": 317, "correlation": 53621
    }
  },
  {
    "ph": "f", "id": 53621, "pid": 76337, "tid": -914061504, "ts": 1716454218267320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267321, "dur": 0,
    "args": {
      "External id": 53622,
      "cbid": 203, "correlation": 53622
    }
  },
  {
    "ph": "f", "id": 53622, "pid": 76337, "tid": -914061504, "ts": 1716454218267321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267321, "dur": 0,
    "args": {
      "External id": 53623,
      "cbid": 205, "correlation": 53623
    }
  },
  {
    "ph": "f", "id": 53623, "pid": 76337, "tid": -914061504, "ts": 1716454218267321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218328252, "dur": 17,
    "args": {
      "External id": 53627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53627, "pid": 5, "tid": 7, "ts": 1716454218328252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267335, "dur": 11,
    "args": {
      "External id": 53627,
      "cbid": 211, "correlation": 53627
    }
  },
  {
    "ph": "s", "id": 53627, "pid": 76337, "tid": -914061504, "ts": 1716454218267335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218328270, "dur": 4,
    "args": {
      "External id": 53629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53629, "pid": 5, "tid": 7, "ts": 1716454218328270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267351, "dur": 6,
    "args": {
      "External id": 53629,
      "cbid": 211, "correlation": 53629
    }
  },
  {
    "ph": "s", "id": 53629, "pid": 76337, "tid": -914061504, "ts": 1716454218267351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267359, "dur": 0,
    "args": {
      "External id": 53630,
      "cbid": 51, "correlation": 53630
    }
  },
  {
    "ph": "s", "id": 53630, "pid": 76337, "tid": -914061504, "ts": 1716454218267359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218328276, "dur": 132,
    "args": {
      "External id": 53631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53631,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 53631, "pid": 5, "tid": 7, "ts": 1716454218328276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267360, "dur": 5,
    "args": {
      "External id": 53631,
      "cbid": 211, "correlation": 53631
    }
  },
  {
    "ph": "s", "id": 53631, "pid": 76337, "tid": -914061504, "ts": 1716454218267360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218328409, "dur": 15,
    "args": {
      "External id": 53636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53636, "pid": 5, "tid": 7, "ts": 1716454218328409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267386, "dur": 8,
    "args": {
      "External id": 53636,
      "cbid": 211, "correlation": 53636
    }
  },
  {
    "ph": "s", "id": 53636, "pid": 76337, "tid": -914061504, "ts": 1716454218267386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218328425, "dur": 12,
    "args": {
      "External id": 53644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53644,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53644, "pid": 5, "tid": 7, "ts": 1716454218328425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267414, "dur": 9,
    "args": {
      "External id": 53644,
      "cbid": 211, "correlation": 53644
    }
  },
  {
    "ph": "s", "id": 53644, "pid": 76337, "tid": -914061504, "ts": 1716454218267414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218328438, "dur": 10,
    "args": {
      "External id": 53652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53652,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53652, "pid": 5, "tid": 7, "ts": 1716454218328438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267443, "dur": 11,
    "args": {
      "External id": 53652,
      "cbid": 211, "correlation": 53652
    }
  },
  {
    "ph": "s", "id": 53652, "pid": 76337, "tid": -914061504, "ts": 1716454218267443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218328450, "dur": 19,
    "args": {
      "External id": 53672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53672,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 53672, "pid": 5, "tid": 7, "ts": 1716454218328450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267531, "dur": 12,
    "args": {
      "External id": 53672,
      "cbid": 211, "correlation": 53672
    }
  },
  {
    "ph": "s", "id": 53672, "pid": 76337, "tid": -914061504, "ts": 1716454218267531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218328469, "dur": 4,
    "args": {
      "External id": 53684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53684,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 53684, "pid": 5, "tid": 7, "ts": 1716454218328469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267553, "dur": 6,
    "args": {
      "External id": 53684,
      "cbid": 211, "correlation": 53684
    }
  },
  {
    "ph": "s", "id": 53684, "pid": 76337, "tid": -914061504, "ts": 1716454218267553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218328475, "dur": 17,
    "args": {
      "External id": 53687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53687,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53687, "pid": 5, "tid": 7, "ts": 1716454218328475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267571, "dur": 7,
    "args": {
      "External id": 53687,
      "cbid": 211, "correlation": 53687
    }
  },
  {
    "ph": "s", "id": 53687, "pid": 76337, "tid": -914061504, "ts": 1716454218267571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267629, "dur": 0,
    "args": {
      "External id": 53698,
      "cbid": 317, "correlation": 53698
    }
  },
  {
    "ph": "f", "id": 53698, "pid": 76337, "tid": -914061504, "ts": 1716454218267629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267630, "dur": 0,
    "args": {
      "External id": 53699,
      "cbid": 203, "correlation": 53699
    }
  },
  {
    "ph": "f", "id": 53699, "pid": 76337, "tid": -914061504, "ts": 1716454218267630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267631, "dur": 0,
    "args": {
      "External id": 53700,
      "cbid": 205, "correlation": 53700
    }
  },
  {
    "ph": "f", "id": 53700, "pid": 76337, "tid": -914061504, "ts": 1716454218267631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218328493, "dur": 11,
    "args": {
      "External id": 53704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53704, "pid": 5, "tid": 7, "ts": 1716454218328493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267644, "dur": 11,
    "args": {
      "External id": 53704,
      "cbid": 211, "correlation": 53704
    }
  },
  {
    "ph": "s", "id": 53704, "pid": 76337, "tid": -914061504, "ts": 1716454218267644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218328506, "dur": 3,
    "args": {
      "External id": 53706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53706, "pid": 5, "tid": 7, "ts": 1716454218328506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267659, "dur": 6,
    "args": {
      "External id": 53706,
      "cbid": 211, "correlation": 53706
    }
  },
  {
    "ph": "s", "id": 53706, "pid": 76337, "tid": -914061504, "ts": 1716454218267659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267668, "dur": 0,
    "args": {
      "External id": 53707,
      "cbid": 51, "correlation": 53707
    }
  },
  {
    "ph": "s", "id": 53707, "pid": 76337, "tid": -914061504, "ts": 1716454218267668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218328510, "dur": 91,
    "args": {
      "External id": 53708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53708,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 53708, "pid": 5, "tid": 7, "ts": 1716454218328510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267668, "dur": 5,
    "args": {
      "External id": 53708,
      "cbid": 211, "correlation": 53708
    }
  },
  {
    "ph": "s", "id": 53708, "pid": 76337, "tid": -914061504, "ts": 1716454218267668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218328602, "dur": 16,
    "args": {
      "External id": 53713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53713, "pid": 5, "tid": 7, "ts": 1716454218328602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267695, "dur": 8,
    "args": {
      "External id": 53713,
      "cbid": 211, "correlation": 53713
    }
  },
  {
    "ph": "s", "id": 53713, "pid": 76337, "tid": -914061504, "ts": 1716454218267695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218328620, "dur": 84,
    "args": {
      "External id": 53722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53722,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53722, "pid": 5, "tid": 7, "ts": 1716454218328620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267775, "dur": 18,
    "args": {
      "External id": 53722,
      "cbid": 211, "correlation": 53722
    }
  },
  {
    "ph": "s", "id": 53722, "pid": 76337, "tid": -914061504, "ts": 1716454218267775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218328705, "dur": 29,
    "args": {
      "External id": 53744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53744,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53744, "pid": 5, "tid": 7, "ts": 1716454218328705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267836, "dur": 10,
    "args": {
      "External id": 53744,
      "cbid": 211, "correlation": 53744
    }
  },
  {
    "ph": "s", "id": 53744, "pid": 76337, "tid": -914061504, "ts": 1716454218267836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267927, "dur": 1,
    "args": {
      "External id": 53755,
      "cbid": 251, "correlation": 53755
    }
  },
  {
    "ph": "f", "id": 53755, "pid": 76337, "tid": -914061504, "ts": 1716454218267927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218328736, "dur": 167,
    "args": {
      "External id": 53756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53756,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53756, "pid": 5, "tid": 7, "ts": 1716454218328736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218267933, "dur": 13,
    "args": {
      "External id": 53756,
      "cbid": 211, "correlation": 53756
    }
  },
  {
    "ph": "s", "id": 53756, "pid": 76337, "tid": -914061504, "ts": 1716454218267933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268011, "dur": 1,
    "args": {
      "External id": 53767,
      "cbid": 251, "correlation": 53767
    }
  },
  {
    "ph": "f", "id": 53767, "pid": 76337, "tid": -914061504, "ts": 1716454218268011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218328904, "dur": 156,
    "args": {
      "External id": 53768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53768,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53768, "pid": 5, "tid": 7, "ts": 1716454218328904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268015, "dur": 12,
    "args": {
      "External id": 53768,
      "cbid": 211, "correlation": 53768
    }
  },
  {
    "ph": "s", "id": 53768, "pid": 76337, "tid": -914061504, "ts": 1716454218268015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268082, "dur": 1,
    "args": {
      "External id": 53779,
      "cbid": 251, "correlation": 53779
    }
  },
  {
    "ph": "f", "id": 53779, "pid": 76337, "tid": -914061504, "ts": 1716454218268082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218329061, "dur": 158,
    "args": {
      "External id": 53780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53780,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53780, "pid": 5, "tid": 7, "ts": 1716454218329061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268086, "dur": 11,
    "args": {
      "External id": 53780,
      "cbid": 211, "correlation": 53780
    }
  },
  {
    "ph": "s", "id": 53780, "pid": 76337, "tid": -914061504, "ts": 1716454218268086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218329220, "dur": 338,
    "args": {
      "External id": 53805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53805,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53805, "pid": 5, "tid": 7, "ts": 1716454218329220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268173, "dur": 12,
    "args": {
      "External id": 53805,
      "cbid": 211, "correlation": 53805
    }
  },
  {
    "ph": "s", "id": 53805, "pid": 76337, "tid": -914061504, "ts": 1716454218268173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268274, "dur": 1,
    "args": {
      "External id": 53823,
      "cbid": 251, "correlation": 53823
    }
  },
  {
    "ph": "f", "id": 53823, "pid": 76337, "tid": -914061504, "ts": 1716454218268274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218329559, "dur": 163,
    "args": {
      "External id": 53825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53825,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53825, "pid": 5, "tid": 7, "ts": 1716454218329559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268280, "dur": 13,
    "args": {
      "External id": 53825,
      "cbid": 211, "correlation": 53825
    }
  },
  {
    "ph": "s", "id": 53825, "pid": 76337, "tid": -914061504, "ts": 1716454218268280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218329723, "dur": 19,
    "args": {
      "External id": 53833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53833,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53833, "pid": 5, "tid": 7, "ts": 1716454218329723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268350, "dur": 12,
    "args": {
      "External id": 53833,
      "cbid": 211, "correlation": 53833
    }
  },
  {
    "ph": "s", "id": 53833, "pid": 76337, "tid": -914061504, "ts": 1716454218268350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218329743, "dur": 28,
    "args": {
      "External id": 53841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53841,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53841, "pid": 5, "tid": 7, "ts": 1716454218329743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268388, "dur": 8,
    "args": {
      "External id": 53841,
      "cbid": 211, "correlation": 53841
    }
  },
  {
    "ph": "s", "id": 53841, "pid": 76337, "tid": -914061504, "ts": 1716454218268388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218329772, "dur": 19,
    "args": {
      "External id": 53852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53852, "pid": 5, "tid": 7, "ts": 1716454218329772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268460, "dur": 13,
    "args": {
      "External id": 53852,
      "cbid": 211, "correlation": 53852
    }
  },
  {
    "ph": "s", "id": 53852, "pid": 76337, "tid": -914061504, "ts": 1716454218268460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218329792, "dur": 16,
    "args": {
      "External id": 53874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53874,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53874, "pid": 5, "tid": 7, "ts": 1716454218329792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268491, "dur": 7,
    "args": {
      "External id": 53874,
      "cbid": 211, "correlation": 53874
    }
  },
  {
    "ph": "s", "id": 53874, "pid": 76337, "tid": -914061504, "ts": 1716454218268491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268580, "dur": 1,
    "args": {
      "External id": 53885,
      "cbid": 251, "correlation": 53885
    }
  },
  {
    "ph": "f", "id": 53885, "pid": 76337, "tid": -914061504, "ts": 1716454218268580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218329810, "dur": 89,
    "args": {
      "External id": 53886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53886,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53886, "pid": 5, "tid": 7, "ts": 1716454218329810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268586, "dur": 13,
    "args": {
      "External id": 53886,
      "cbid": 211, "correlation": 53886
    }
  },
  {
    "ph": "s", "id": 53886, "pid": 76337, "tid": -914061504, "ts": 1716454218268586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268656, "dur": 1,
    "args": {
      "External id": 53897,
      "cbid": 251, "correlation": 53897
    }
  },
  {
    "ph": "f", "id": 53897, "pid": 76337, "tid": -914061504, "ts": 1716454218268656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268660, "dur": 0,
    "args": {
      "External id": 53898,
      "cbid": 251, "correlation": 53898
    }
  },
  {
    "ph": "f", "id": 53898, "pid": 76337, "tid": -914061504, "ts": 1716454218268660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218329900, "dur": 11,
    "args": {
      "External id": 53899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53899,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53899, "pid": 5, "tid": 7, "ts": 1716454218329900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268661, "dur": 12,
    "args": {
      "External id": 53899,
      "cbid": 211, "correlation": 53899
    }
  },
  {
    "ph": "s", "id": 53899, "pid": 76337, "tid": -914061504, "ts": 1716454218268661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218329913, "dur": 5,
    "args": {
      "External id": 53901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53901,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53901, "pid": 5, "tid": 7, "ts": 1716454218329913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268675, "dur": 6,
    "args": {
      "External id": 53901,
      "cbid": 211, "correlation": 53901
    }
  },
  {
    "ph": "s", "id": 53901, "pid": 76337, "tid": -914061504, "ts": 1716454218268675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268733, "dur": 1,
    "args": {
      "External id": 53912,
      "cbid": 251, "correlation": 53912
    }
  },
  {
    "ph": "f", "id": 53912, "pid": 76337, "tid": -914061504, "ts": 1716454218268733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268736, "dur": 0,
    "args": {
      "External id": 53913,
      "cbid": 251, "correlation": 53913
    }
  },
  {
    "ph": "f", "id": 53913, "pid": 76337, "tid": -914061504, "ts": 1716454218268736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218329919, "dur": 8,
    "args": {
      "External id": 53914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53914,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53914, "pid": 5, "tid": 7, "ts": 1716454218329919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268738, "dur": 11,
    "args": {
      "External id": 53914,
      "cbid": 211, "correlation": 53914
    }
  },
  {
    "ph": "s", "id": 53914, "pid": 76337, "tid": -914061504, "ts": 1716454218268738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218329929, "dur": 3,
    "args": {
      "External id": 53916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53916,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53916, "pid": 5, "tid": 7, "ts": 1716454218329929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268751, "dur": 6,
    "args": {
      "External id": 53916,
      "cbid": 211, "correlation": 53916
    }
  },
  {
    "ph": "s", "id": 53916, "pid": 76337, "tid": -914061504, "ts": 1716454218268751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218329933, "dur": 55,
    "args": {
      "External id": 53941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53941,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 53941, "pid": 5, "tid": 7, "ts": 1716454218329933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268827, "dur": 13,
    "args": {
      "External id": 53941,
      "cbid": 211, "correlation": 53941
    }
  },
  {
    "ph": "s", "id": 53941, "pid": 76337, "tid": -914061504, "ts": 1716454218268827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268928, "dur": 1,
    "args": {
      "External id": 53959,
      "cbid": 251, "correlation": 53959
    }
  },
  {
    "ph": "f", "id": 53959, "pid": 76337, "tid": -914061504, "ts": 1716454218268928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218329989, "dur": 92,
    "args": {
      "External id": 53961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53961,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 53961, "pid": 5, "tid": 7, "ts": 1716454218329989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218268934, "dur": 14,
    "args": {
      "External id": 53961,
      "cbid": 211, "correlation": 53961
    }
  },
  {
    "ph": "s", "id": 53961, "pid": 76337, "tid": -914061504, "ts": 1716454218268934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218330082, "dur": 9,
    "args": {
      "External id": 53969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53969,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53969, "pid": 5, "tid": 7, "ts": 1716454218330082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269015, "dur": 12,
    "args": {
      "External id": 53969,
      "cbid": 211, "correlation": 53969
    }
  },
  {
    "ph": "s", "id": 53969, "pid": 76337, "tid": -914061504, "ts": 1716454218269015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218330093, "dur": 22,
    "args": {
      "External id": 53977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53977, "pid": 5, "tid": 7, "ts": 1716454218330093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269057, "dur": 10,
    "args": {
      "External id": 53977,
      "cbid": 211, "correlation": 53977
    }
  },
  {
    "ph": "s", "id": 53977, "pid": 76337, "tid": -914061504, "ts": 1716454218269057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218330116, "dur": 18,
    "args": {
      "External id": 53999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 53999,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 53999, "pid": 5, "tid": 7, "ts": 1716454218330116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269109, "dur": 10,
    "args": {
      "External id": 53999,
      "cbid": 211, "correlation": 53999
    }
  },
  {
    "ph": "s", "id": 53999, "pid": 76337, "tid": -914061504, "ts": 1716454218269109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269196, "dur": 1,
    "args": {
      "External id": 54015,
      "cbid": 251, "correlation": 54015
    }
  },
  {
    "ph": "f", "id": 54015, "pid": 76337, "tid": -914061504, "ts": 1716454218269196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269201, "dur": 0,
    "args": {
      "External id": 54017,
      "cbid": 251, "correlation": 54017
    }
  },
  {
    "ph": "f", "id": 54017, "pid": 76337, "tid": -914061504, "ts": 1716454218269201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218330135, "dur": 497,
    "args": {
      "External id": 54018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54018,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54018, "pid": 5, "tid": 7, "ts": 1716454218330135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269203, "dur": 13,
    "args": {
      "External id": 54018,
      "cbid": 211, "correlation": 54018
    }
  },
  {
    "ph": "s", "id": 54018, "pid": 76337, "tid": -914061504, "ts": 1716454218269203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218330633, "dur": 67,
    "args": {
      "External id": 54026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54026, "pid": 5, "tid": 7, "ts": 1716454218330633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269269, "dur": 12,
    "args": {
      "External id": 54026,
      "cbid": 211, "correlation": 54026
    }
  },
  {
    "ph": "s", "id": 54026, "pid": 76337, "tid": -914061504, "ts": 1716454218269269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218330701, "dur": 65,
    "args": {
      "External id": 54034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54034, "pid": 5, "tid": 7, "ts": 1716454218330701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269303, "dur": 10,
    "args": {
      "External id": 54034,
      "cbid": 211, "correlation": 54034
    }
  },
  {
    "ph": "s", "id": 54034, "pid": 76337, "tid": -914061504, "ts": 1716454218269303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269385, "dur": 1,
    "args": {
      "External id": 54050,
      "cbid": 251, "correlation": 54050
    }
  },
  {
    "ph": "f", "id": 54050, "pid": 76337, "tid": -914061504, "ts": 1716454218269385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218330769, "dur": 1,
    "args": {
      "External id": 54052,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 54052,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 54052, "pid": 5, "tid": 7, "ts": 1716454218330769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269390, "dur": 9,
    "args": {
      "External id": 54052,
      "cbid": 51, "correlation": 54052
    }
  },
  {
    "ph": "s", "id": 54052, "pid": 76337, "tid": -914061504, "ts": 1716454218269390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218330773, "dur": 268,
    "args": {
      "External id": 54053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54053,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 54053, "pid": 5, "tid": 7, "ts": 1716454218330773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269400, "dur": 11,
    "args": {
      "External id": 54053,
      "cbid": 211, "correlation": 54053
    }
  },
  {
    "ph": "s", "id": 54053, "pid": 76337, "tid": -914061504, "ts": 1716454218269400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218331042, "dur": 13,
    "args": {
      "External id": 54061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54061,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54061, "pid": 5, "tid": 7, "ts": 1716454218331042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269444, "dur": 10,
    "args": {
      "External id": 54061,
      "cbid": 211, "correlation": 54061
    }
  },
  {
    "ph": "s", "id": 54061, "pid": 76337, "tid": -914061504, "ts": 1716454218269444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218331057, "dur": 37,
    "args": {
      "External id": 54072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54072, "pid": 5, "tid": 7, "ts": 1716454218331057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269512, "dur": 13,
    "args": {
      "External id": 54072,
      "cbid": 211, "correlation": 54072
    }
  },
  {
    "ph": "s", "id": 54072, "pid": 76337, "tid": -914061504, "ts": 1716454218269512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269577, "dur": 0,
    "args": {
      "External id": 54084,
      "cbid": 317, "correlation": 54084
    }
  },
  {
    "ph": "f", "id": 54084, "pid": 76337, "tid": -914061504, "ts": 1716454218269577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269578, "dur": 0,
    "args": {
      "External id": 54085,
      "cbid": 203, "correlation": 54085
    }
  },
  {
    "ph": "f", "id": 54085, "pid": 76337, "tid": -914061504, "ts": 1716454218269578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269579, "dur": 0,
    "args": {
      "External id": 54086,
      "cbid": 205, "correlation": 54086
    }
  },
  {
    "ph": "f", "id": 54086, "pid": 76337, "tid": -914061504, "ts": 1716454218269579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218331095, "dur": 13,
    "args": {
      "External id": 54090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54090, "pid": 5, "tid": 7, "ts": 1716454218331095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269594, "dur": 12,
    "args": {
      "External id": 54090,
      "cbid": 211, "correlation": 54090
    }
  },
  {
    "ph": "s", "id": 54090, "pid": 76337, "tid": -914061504, "ts": 1716454218269594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218331109, "dur": 4,
    "args": {
      "External id": 54092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 54092, "pid": 5, "tid": 7, "ts": 1716454218331109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269610, "dur": 6,
    "args": {
      "External id": 54092,
      "cbid": 211, "correlation": 54092
    }
  },
  {
    "ph": "s", "id": 54092, "pid": 76337, "tid": -914061504, "ts": 1716454218269610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269619, "dur": 0,
    "args": {
      "External id": 54093,
      "cbid": 51, "correlation": 54093
    }
  },
  {
    "ph": "s", "id": 54093, "pid": 76337, "tid": -914061504, "ts": 1716454218269619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218331114, "dur": 97,
    "args": {
      "External id": 54094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54094,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 54094, "pid": 5, "tid": 7, "ts": 1716454218331114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269620, "dur": 5,
    "args": {
      "External id": 54094,
      "cbid": 211, "correlation": 54094
    }
  },
  {
    "ph": "s", "id": 54094, "pid": 76337, "tid": -914061504, "ts": 1716454218269620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218331213, "dur": 16,
    "args": {
      "External id": 54099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54099, "pid": 5, "tid": 7, "ts": 1716454218331213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269647, "dur": 11,
    "args": {
      "External id": 54099,
      "cbid": 211, "correlation": 54099
    }
  },
  {
    "ph": "s", "id": 54099, "pid": 76337, "tid": -914061504, "ts": 1716454218269647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218331230, "dur": 12,
    "args": {
      "External id": 54107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54107,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54107, "pid": 5, "tid": 7, "ts": 1716454218331230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269682, "dur": 8,
    "args": {
      "External id": 54107,
      "cbid": 211, "correlation": 54107
    }
  },
  {
    "ph": "s", "id": 54107, "pid": 76337, "tid": -914061504, "ts": 1716454218269682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218331243, "dur": 55,
    "args": {
      "External id": 54118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54118,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54118, "pid": 5, "tid": 7, "ts": 1716454218331243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269745, "dur": 12,
    "args": {
      "External id": 54118,
      "cbid": 211, "correlation": 54118
    }
  },
  {
    "ph": "s", "id": 54118, "pid": 76337, "tid": -914061504, "ts": 1716454218269745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269801, "dur": 0,
    "args": {
      "External id": 54128,
      "cbid": 317, "correlation": 54128
    }
  },
  {
    "ph": "f", "id": 54128, "pid": 76337, "tid": -914061504, "ts": 1716454218269801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269801, "dur": 0,
    "args": {
      "External id": 54129,
      "cbid": 203, "correlation": 54129
    }
  },
  {
    "ph": "f", "id": 54129, "pid": 76337, "tid": -914061504, "ts": 1716454218269801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269802, "dur": 0,
    "args": {
      "External id": 54130,
      "cbid": 205, "correlation": 54130
    }
  },
  {
    "ph": "f", "id": 54130, "pid": 76337, "tid": -914061504, "ts": 1716454218269802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218331300, "dur": 39,
    "args": {
      "External id": 54134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54134, "pid": 5, "tid": 7, "ts": 1716454218331300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269818, "dur": 11,
    "args": {
      "External id": 54134,
      "cbid": 211, "correlation": 54134
    }
  },
  {
    "ph": "s", "id": 54134, "pid": 76337, "tid": -914061504, "ts": 1716454218269818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218331340, "dur": 162,
    "args": {
      "External id": 54136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54136,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54136, "pid": 5, "tid": 7, "ts": 1716454218331340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269832, "dur": 5,
    "args": {
      "External id": 54136,
      "cbid": 211, "correlation": 54136
    }
  },
  {
    "ph": "s", "id": 54136, "pid": 76337, "tid": -914061504, "ts": 1716454218269832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218331503, "dur": 1962,
    "args": {
      "External id": 54138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54138,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54138, "pid": 5, "tid": 7, "ts": 1716454218331503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269845, "dur": 8,
    "args": {
      "External id": 54138,
      "cbid": 211, "correlation": 54138
    }
  },
  {
    "ph": "s", "id": 54138, "pid": 76337, "tid": -914061504, "ts": 1716454218269845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218333466, "dur": 40,
    "args": {
      "External id": 54140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54140,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54140, "pid": 5, "tid": 7, "ts": 1716454218333466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269857, "dur": 5,
    "args": {
      "External id": 54140,
      "cbid": 211, "correlation": 54140
    }
  },
  {
    "ph": "s", "id": 54140, "pid": 76337, "tid": -914061504, "ts": 1716454218269857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218333508, "dur": 58,
    "args": {
      "External id": 54146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54146, "pid": 5, "tid": 7, "ts": 1716454218333508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269884, "dur": 9,
    "args": {
      "External id": 54146,
      "cbid": 211, "correlation": 54146
    }
  },
  {
    "ph": "s", "id": 54146, "pid": 76337, "tid": -914061504, "ts": 1716454218269884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218333567, "dur": 85,
    "args": {
      "External id": 54155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54155,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54155, "pid": 5, "tid": 7, "ts": 1716454218333567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218269983, "dur": 14,
    "args": {
      "External id": 54155,
      "cbid": 211, "correlation": 54155
    }
  },
  {
    "ph": "s", "id": 54155, "pid": 76337, "tid": -914061504, "ts": 1716454218269983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218333654, "dur": 70,
    "args": {
      "External id": 54175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54175,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 54175, "pid": 5, "tid": 7, "ts": 1716454218333654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270061, "dur": 12,
    "args": {
      "External id": 54175,
      "cbid": 211, "correlation": 54175
    }
  },
  {
    "ph": "s", "id": 54175, "pid": 76337, "tid": -914061504, "ts": 1716454218270061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218333725, "dur": 5,
    "args": {
      "External id": 54187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54187,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 54187, "pid": 5, "tid": 7, "ts": 1716454218333725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270083, "dur": 6,
    "args": {
      "External id": 54187,
      "cbid": 211, "correlation": 54187
    }
  },
  {
    "ph": "s", "id": 54187, "pid": 76337, "tid": -914061504, "ts": 1716454218270083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218333731, "dur": 82,
    "args": {
      "External id": 54190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54190,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54190, "pid": 5, "tid": 7, "ts": 1716454218333731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270102, "dur": 7,
    "args": {
      "External id": 54190,
      "cbid": 211, "correlation": 54190
    }
  },
  {
    "ph": "s", "id": 54190, "pid": 76337, "tid": -914061504, "ts": 1716454218270102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218333814, "dur": 53,
    "args": {
      "External id": 54199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54199,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54199, "pid": 5, "tid": 7, "ts": 1716454218333814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270143, "dur": 11,
    "args": {
      "External id": 54199,
      "cbid": 211, "correlation": 54199
    }
  },
  {
    "ph": "s", "id": 54199, "pid": 76337, "tid": -914061504, "ts": 1716454218270143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270195, "dur": 0,
    "args": {
      "External id": 54209,
      "cbid": 317, "correlation": 54209
    }
  },
  {
    "ph": "f", "id": 54209, "pid": 76337, "tid": -914061504, "ts": 1716454218270195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270196, "dur": 0,
    "args": {
      "External id": 54210,
      "cbid": 203, "correlation": 54210
    }
  },
  {
    "ph": "f", "id": 54210, "pid": 76337, "tid": -914061504, "ts": 1716454218270196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270197, "dur": 0,
    "args": {
      "External id": 54211,
      "cbid": 205, "correlation": 54211
    }
  },
  {
    "ph": "f", "id": 54211, "pid": 76337, "tid": -914061504, "ts": 1716454218270197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218333869, "dur": 57,
    "args": {
      "External id": 54215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54215,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54215, "pid": 5, "tid": 7, "ts": 1716454218333869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270213, "dur": 11,
    "args": {
      "External id": 54215,
      "cbid": 211, "correlation": 54215
    }
  },
  {
    "ph": "s", "id": 54215, "pid": 76337, "tid": -914061504, "ts": 1716454218270213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218333927, "dur": 120,
    "args": {
      "External id": 54217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54217,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54217, "pid": 5, "tid": 7, "ts": 1716454218333927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270227, "dur": 5,
    "args": {
      "External id": 54217,
      "cbid": 211, "correlation": 54217
    }
  },
  {
    "ph": "s", "id": 54217, "pid": 76337, "tid": -914061504, "ts": 1716454218270227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218334049, "dur": 1880,
    "args": {
      "External id": 54219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54219,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54219, "pid": 5, "tid": 7, "ts": 1716454218334049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270238, "dur": 6,
    "args": {
      "External id": 54219,
      "cbid": 211, "correlation": 54219
    }
  },
  {
    "ph": "s", "id": 54219, "pid": 76337, "tid": -914061504, "ts": 1716454218270238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218335930, "dur": 20,
    "args": {
      "External id": 54221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54221,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54221, "pid": 5, "tid": 7, "ts": 1716454218335930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270248, "dur": 5,
    "args": {
      "External id": 54221,
      "cbid": 211, "correlation": 54221
    }
  },
  {
    "ph": "s", "id": 54221, "pid": 76337, "tid": -914061504, "ts": 1716454218270248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218335951, "dur": 32,
    "args": {
      "External id": 54227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54227,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54227, "pid": 5, "tid": 7, "ts": 1716454218335951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270276, "dur": 11,
    "args": {
      "External id": 54227,
      "cbid": 211, "correlation": 54227
    }
  },
  {
    "ph": "s", "id": 54227, "pid": 76337, "tid": -914061504, "ts": 1716454218270276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218335985, "dur": 3,
    "args": {
      "External id": 54235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54235,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 54235, "pid": 5, "tid": 7, "ts": 1716454218335985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270324, "dur": 10,
    "args": {
      "External id": 54235,
      "cbid": 211, "correlation": 54235
    }
  },
  {
    "ph": "s", "id": 54235, "pid": 76337, "tid": -914061504, "ts": 1716454218270324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270390, "dur": 1,
    "args": {
      "External id": 54251,
      "cbid": 251, "correlation": 54251
    }
  },
  {
    "ph": "f", "id": 54251, "pid": 76337, "tid": -914061504, "ts": 1716454218270390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270396, "dur": 0,
    "args": {
      "External id": 54253,
      "cbid": 251, "correlation": 54253
    }
  },
  {
    "ph": "f", "id": 54253, "pid": 76337, "tid": -914061504, "ts": 1716454218270396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218335990, "dur": 12,
    "args": {
      "External id": 54254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54254,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 54254, "pid": 5, "tid": 7, "ts": 1716454218335990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270397, "dur": 12,
    "args": {
      "External id": 54254,
      "cbid": 211, "correlation": 54254
    }
  },
  {
    "ph": "s", "id": 54254, "pid": 76337, "tid": -914061504, "ts": 1716454218270397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218336003, "dur": 5,
    "args": {
      "External id": 54256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54256,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 54256, "pid": 5, "tid": 7, "ts": 1716454218336003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270411, "dur": 6,
    "args": {
      "External id": 54256,
      "cbid": 211, "correlation": 54256
    }
  },
  {
    "ph": "s", "id": 54256, "pid": 76337, "tid": -914061504, "ts": 1716454218270411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218336010, "dur": 29,
    "args": {
      "External id": 54266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54266,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54266, "pid": 5, "tid": 7, "ts": 1716454218336010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270470, "dur": 12,
    "args": {
      "External id": 54266,
      "cbid": 211, "correlation": 54266
    }
  },
  {
    "ph": "s", "id": 54266, "pid": 76337, "tid": -914061504, "ts": 1716454218270470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218336040, "dur": 30,
    "args": {
      "External id": 54286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54286,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 54286, "pid": 5, "tid": 7, "ts": 1716454218336040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270536, "dur": 11,
    "args": {
      "External id": 54286,
      "cbid": 211, "correlation": 54286
    }
  },
  {
    "ph": "s", "id": 54286, "pid": 76337, "tid": -914061504, "ts": 1716454218270536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218336072, "dur": 4,
    "args": {
      "External id": 54298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54298,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 54298, "pid": 5, "tid": 7, "ts": 1716454218336072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270557, "dur": 6,
    "args": {
      "External id": 54298,
      "cbid": 211, "correlation": 54298
    }
  },
  {
    "ph": "s", "id": 54298, "pid": 76337, "tid": -914061504, "ts": 1716454218270557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218336077, "dur": 30,
    "args": {
      "External id": 54301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54301, "pid": 5, "tid": 7, "ts": 1716454218336077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270575, "dur": 10,
    "args": {
      "External id": 54301,
      "cbid": 211, "correlation": 54301
    }
  },
  {
    "ph": "s", "id": 54301, "pid": 76337, "tid": -914061504, "ts": 1716454218270575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218336108, "dur": 20,
    "args": {
      "External id": 54310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54310,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54310, "pid": 5, "tid": 7, "ts": 1716454218336108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270619, "dur": 10,
    "args": {
      "External id": 54310,
      "cbid": 211, "correlation": 54310
    }
  },
  {
    "ph": "s", "id": 54310, "pid": 76337, "tid": -914061504, "ts": 1716454218270619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270682, "dur": 0,
    "args": {
      "External id": 54320,
      "cbid": 317, "correlation": 54320
    }
  },
  {
    "ph": "f", "id": 54320, "pid": 76337, "tid": -914061504, "ts": 1716454218270682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270683, "dur": 0,
    "args": {
      "External id": 54321,
      "cbid": 203, "correlation": 54321
    }
  },
  {
    "ph": "f", "id": 54321, "pid": 76337, "tid": -914061504, "ts": 1716454218270683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270684, "dur": 0,
    "args": {
      "External id": 54322,
      "cbid": 205, "correlation": 54322
    }
  },
  {
    "ph": "f", "id": 54322, "pid": 76337, "tid": -914061504, "ts": 1716454218270684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218336130, "dur": 22,
    "args": {
      "External id": 54326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54326,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54326, "pid": 5, "tid": 7, "ts": 1716454218336130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270699, "dur": 12,
    "args": {
      "External id": 54326,
      "cbid": 211, "correlation": 54326
    }
  },
  {
    "ph": "s", "id": 54326, "pid": 76337, "tid": -914061504, "ts": 1716454218270699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218336153, "dur": 43,
    "args": {
      "External id": 54328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54328,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54328, "pid": 5, "tid": 7, "ts": 1716454218336153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270714, "dur": 5,
    "args": {
      "External id": 54328,
      "cbid": 211, "correlation": 54328
    }
  },
  {
    "ph": "s", "id": 54328, "pid": 76337, "tid": -914061504, "ts": 1716454218270714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218336198, "dur": 642,
    "args": {
      "External id": 54330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54330,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54330, "pid": 5, "tid": 7, "ts": 1716454218336198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270726, "dur": 6,
    "args": {
      "External id": 54330,
      "cbid": 211, "correlation": 54330
    }
  },
  {
    "ph": "s", "id": 54330, "pid": 76337, "tid": -914061504, "ts": 1716454218270726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218336841, "dur": 21,
    "args": {
      "External id": 54332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54332,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54332, "pid": 5, "tid": 7, "ts": 1716454218336841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270735, "dur": 5,
    "args": {
      "External id": 54332,
      "cbid": 211, "correlation": 54332
    }
  },
  {
    "ph": "s", "id": 54332, "pid": 76337, "tid": -914061504, "ts": 1716454218270735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218336864, "dur": 33,
    "args": {
      "External id": 54338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54338, "pid": 5, "tid": 7, "ts": 1716454218336864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270763, "dur": 8,
    "args": {
      "External id": 54338,
      "cbid": 211, "correlation": 54338
    }
  },
  {
    "ph": "s", "id": 54338, "pid": 76337, "tid": -914061504, "ts": 1716454218270763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270821, "dur": 0,
    "args": {
      "External id": 54348,
      "cbid": 317, "correlation": 54348
    }
  },
  {
    "ph": "f", "id": 54348, "pid": 76337, "tid": -914061504, "ts": 1716454218270821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270821, "dur": 0,
    "args": {
      "External id": 54349,
      "cbid": 203, "correlation": 54349
    }
  },
  {
    "ph": "f", "id": 54349, "pid": 76337, "tid": -914061504, "ts": 1716454218270821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270822, "dur": 0,
    "args": {
      "External id": 54350,
      "cbid": 205, "correlation": 54350
    }
  },
  {
    "ph": "f", "id": 54350, "pid": 76337, "tid": -914061504, "ts": 1716454218270822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218336898, "dur": 56,
    "args": {
      "External id": 54354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54354,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54354, "pid": 5, "tid": 7, "ts": 1716454218336898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270840, "dur": 12,
    "args": {
      "External id": 54354,
      "cbid": 211, "correlation": 54354
    }
  },
  {
    "ph": "s", "id": 54354, "pid": 76337, "tid": -914061504, "ts": 1716454218270840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218336956, "dur": 269,
    "args": {
      "External id": 54356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54356,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54356, "pid": 5, "tid": 7, "ts": 1716454218336956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270860, "dur": 7,
    "args": {
      "External id": 54356,
      "cbid": 211, "correlation": 54356
    }
  },
  {
    "ph": "s", "id": 54356, "pid": 76337, "tid": -914061504, "ts": 1716454218270860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218337226, "dur": 21,
    "args": {
      "External id": 54358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54358,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54358, "pid": 5, "tid": 7, "ts": 1716454218337226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270872, "dur": 6,
    "args": {
      "External id": 54358,
      "cbid": 211, "correlation": 54358
    }
  },
  {
    "ph": "s", "id": 54358, "pid": 76337, "tid": -914061504, "ts": 1716454218270872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218337248, "dur": 32,
    "args": {
      "External id": 54364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54364, "pid": 5, "tid": 7, "ts": 1716454218337248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270900, "dur": 8,
    "args": {
      "External id": 54364,
      "cbid": 211, "correlation": 54364
    }
  },
  {
    "ph": "s", "id": 54364, "pid": 76337, "tid": -914061504, "ts": 1716454218270900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218337281, "dur": 27,
    "args": {
      "External id": 54372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54372,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54372, "pid": 5, "tid": 7, "ts": 1716454218337281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270930, "dur": 8,
    "args": {
      "External id": 54372,
      "cbid": 211, "correlation": 54372
    }
  },
  {
    "ph": "s", "id": 54372, "pid": 76337, "tid": -914061504, "ts": 1716454218270930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218337309, "dur": 20,
    "args": {
      "External id": 54380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54380,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54380, "pid": 5, "tid": 7, "ts": 1716454218337309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218270958, "dur": 8,
    "args": {
      "External id": 54380,
      "cbid": 211, "correlation": 54380
    }
  },
  {
    "ph": "s", "id": 54380, "pid": 76337, "tid": -914061504, "ts": 1716454218270958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218337331, "dur": 30,
    "args": {
      "External id": 54400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54400,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 54400, "pid": 5, "tid": 7, "ts": 1716454218337331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271049, "dur": 13,
    "args": {
      "External id": 54400,
      "cbid": 211, "correlation": 54400
    }
  },
  {
    "ph": "s", "id": 54400, "pid": 76337, "tid": -914061504, "ts": 1716454218271049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218337362, "dur": 4,
    "args": {
      "External id": 54412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54412,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 54412, "pid": 5, "tid": 7, "ts": 1716454218337362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271072, "dur": 6,
    "args": {
      "External id": 54412,
      "cbid": 211, "correlation": 54412
    }
  },
  {
    "ph": "s", "id": 54412, "pid": 76337, "tid": -914061504, "ts": 1716454218271072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218337367, "dur": 31,
    "args": {
      "External id": 54415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54415, "pid": 5, "tid": 7, "ts": 1716454218337367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271090, "dur": 6,
    "args": {
      "External id": 54415,
      "cbid": 211, "correlation": 54415
    }
  },
  {
    "ph": "s", "id": 54415, "pid": 76337, "tid": -914061504, "ts": 1716454218271090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271148, "dur": 0,
    "args": {
      "External id": 54426,
      "cbid": 317, "correlation": 54426
    }
  },
  {
    "ph": "f", "id": 54426, "pid": 76337, "tid": -914061504, "ts": 1716454218271148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271149, "dur": 0,
    "args": {
      "External id": 54427,
      "cbid": 203, "correlation": 54427
    }
  },
  {
    "ph": "f", "id": 54427, "pid": 76337, "tid": -914061504, "ts": 1716454218271149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271149, "dur": 0,
    "args": {
      "External id": 54428,
      "cbid": 205, "correlation": 54428
    }
  },
  {
    "ph": "f", "id": 54428, "pid": 76337, "tid": -914061504, "ts": 1716454218271149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218337399, "dur": 22,
    "args": {
      "External id": 54432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54432, "pid": 5, "tid": 7, "ts": 1716454218337399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271163, "dur": 14,
    "args": {
      "External id": 54432,
      "cbid": 211, "correlation": 54432
    }
  },
  {
    "ph": "s", "id": 54432, "pid": 76337, "tid": -914061504, "ts": 1716454218271163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218337423, "dur": 104,
    "args": {
      "External id": 54434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54434,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54434, "pid": 5, "tid": 7, "ts": 1716454218337423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271184, "dur": 7,
    "args": {
      "External id": 54434,
      "cbid": 211, "correlation": 54434
    }
  },
  {
    "ph": "s", "id": 54434, "pid": 76337, "tid": -914061504, "ts": 1716454218271184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218337528, "dur": 22,
    "args": {
      "External id": 54436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54436,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54436, "pid": 5, "tid": 7, "ts": 1716454218337528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271195, "dur": 5,
    "args": {
      "External id": 54436,
      "cbid": 211, "correlation": 54436
    }
  },
  {
    "ph": "s", "id": 54436, "pid": 76337, "tid": -914061504, "ts": 1716454218271195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218337551, "dur": 33,
    "args": {
      "External id": 54442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54442, "pid": 5, "tid": 7, "ts": 1716454218337551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271223, "dur": 8,
    "args": {
      "External id": 54442,
      "cbid": 211, "correlation": 54442
    }
  },
  {
    "ph": "s", "id": 54442, "pid": 76337, "tid": -914061504, "ts": 1716454218271223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218337585, "dur": 178,
    "args": {
      "External id": 54451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54451,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54451, "pid": 5, "tid": 7, "ts": 1716454218337585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271305, "dur": 14,
    "args": {
      "External id": 54451,
      "cbid": 211, "correlation": 54451
    }
  },
  {
    "ph": "s", "id": 54451, "pid": 76337, "tid": -914061504, "ts": 1716454218271305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218337764, "dur": 64,
    "args": {
      "External id": 54473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54473,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54473, "pid": 5, "tid": 7, "ts": 1716454218337764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271361, "dur": 10,
    "args": {
      "External id": 54473,
      "cbid": 211, "correlation": 54473
    }
  },
  {
    "ph": "s", "id": 54473, "pid": 76337, "tid": -914061504, "ts": 1716454218271361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271452, "dur": 1,
    "args": {
      "External id": 54484,
      "cbid": 251, "correlation": 54484
    }
  },
  {
    "ph": "f", "id": 54484, "pid": 76337, "tid": -914061504, "ts": 1716454218271452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218337830, "dur": 153,
    "args": {
      "External id": 54485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54485,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54485, "pid": 5, "tid": 7, "ts": 1716454218337830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271457, "dur": 13,
    "args": {
      "External id": 54485,
      "cbid": 211, "correlation": 54485
    }
  },
  {
    "ph": "s", "id": 54485, "pid": 76337, "tid": -914061504, "ts": 1716454218271457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271527, "dur": 1,
    "args": {
      "External id": 54496,
      "cbid": 251, "correlation": 54496
    }
  },
  {
    "ph": "f", "id": 54496, "pid": 76337, "tid": -914061504, "ts": 1716454218271527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218337984, "dur": 146,
    "args": {
      "External id": 54497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54497,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54497, "pid": 5, "tid": 7, "ts": 1716454218337984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271531, "dur": 11,
    "args": {
      "External id": 54497,
      "cbid": 211, "correlation": 54497
    }
  },
  {
    "ph": "s", "id": 54497, "pid": 76337, "tid": -914061504, "ts": 1716454218271531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271600, "dur": 1,
    "args": {
      "External id": 54508,
      "cbid": 251, "correlation": 54508
    }
  },
  {
    "ph": "f", "id": 54508, "pid": 76337, "tid": -914061504, "ts": 1716454218271600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218338132, "dur": 144,
    "args": {
      "External id": 54509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54509,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54509, "pid": 5, "tid": 7, "ts": 1716454218338132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271604, "dur": 12,
    "args": {
      "External id": 54509,
      "cbid": 211, "correlation": 54509
    }
  },
  {
    "ph": "s", "id": 54509, "pid": 76337, "tid": -914061504, "ts": 1716454218271604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218338277, "dur": 1921,
    "args": {
      "External id": 54530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54530,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 54530, "pid": 5, "tid": 7, "ts": 1716454218338277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271688, "dur": 13,
    "args": {
      "External id": 54530,
      "cbid": 211, "correlation": 54530
    }
  },
  {
    "ph": "s", "id": 54530, "pid": 76337, "tid": -914061504, "ts": 1716454218271688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271787, "dur": 1,
    "args": {
      "External id": 54548,
      "cbid": 251, "correlation": 54548
    }
  },
  {
    "ph": "f", "id": 54548, "pid": 76337, "tid": -914061504, "ts": 1716454218271787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218340200, "dur": 148,
    "args": {
      "External id": 54550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54550,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 54550, "pid": 5, "tid": 7, "ts": 1716454218340200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271792, "dur": 13,
    "args": {
      "External id": 54550,
      "cbid": 211, "correlation": 54550
    }
  },
  {
    "ph": "s", "id": 54550, "pid": 76337, "tid": -914061504, "ts": 1716454218271792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218340349, "dur": 35,
    "args": {
      "External id": 54558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54558,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54558, "pid": 5, "tid": 7, "ts": 1716454218340349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271863, "dur": 12,
    "args": {
      "External id": 54558,
      "cbid": 211, "correlation": 54558
    }
  },
  {
    "ph": "s", "id": 54558, "pid": 76337, "tid": -914061504, "ts": 1716454218271863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218340385, "dur": 50,
    "args": {
      "External id": 54566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54566,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54566, "pid": 5, "tid": 7, "ts": 1716454218340385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271903, "dur": 8,
    "args": {
      "External id": 54566,
      "cbid": 211, "correlation": 54566
    }
  },
  {
    "ph": "s", "id": 54566, "pid": 76337, "tid": -914061504, "ts": 1716454218271903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218340437, "dur": 30,
    "args": {
      "External id": 54577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54577, "pid": 5, "tid": 7, "ts": 1716454218340437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218271982, "dur": 14,
    "args": {
      "External id": 54577,
      "cbid": 211, "correlation": 54577
    }
  },
  {
    "ph": "s", "id": 54577, "pid": 76337, "tid": -914061504, "ts": 1716454218271982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218340468, "dur": 34,
    "args": {
      "External id": 54599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54599,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54599, "pid": 5, "tid": 7, "ts": 1716454218340468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272019, "dur": 8,
    "args": {
      "External id": 54599,
      "cbid": 211, "correlation": 54599
    }
  },
  {
    "ph": "s", "id": 54599, "pid": 76337, "tid": -914061504, "ts": 1716454218272019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272106, "dur": 1,
    "args": {
      "External id": 54610,
      "cbid": 251, "correlation": 54610
    }
  },
  {
    "ph": "f", "id": 54610, "pid": 76337, "tid": -914061504, "ts": 1716454218272106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218340503, "dur": 90,
    "args": {
      "External id": 54611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54611,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54611, "pid": 5, "tid": 7, "ts": 1716454218340503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272111, "dur": 12,
    "args": {
      "External id": 54611,
      "cbid": 211, "correlation": 54611
    }
  },
  {
    "ph": "s", "id": 54611, "pid": 76337, "tid": -914061504, "ts": 1716454218272111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272179, "dur": 1,
    "args": {
      "External id": 54622,
      "cbid": 251, "correlation": 54622
    }
  },
  {
    "ph": "f", "id": 54622, "pid": 76337, "tid": -914061504, "ts": 1716454218272179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272183, "dur": 0,
    "args": {
      "External id": 54623,
      "cbid": 251, "correlation": 54623
    }
  },
  {
    "ph": "f", "id": 54623, "pid": 76337, "tid": -914061504, "ts": 1716454218272183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218340595, "dur": 11,
    "args": {
      "External id": 54624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54624,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 54624, "pid": 5, "tid": 7, "ts": 1716454218340595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272185, "dur": 12,
    "args": {
      "External id": 54624,
      "cbid": 211, "correlation": 54624
    }
  },
  {
    "ph": "s", "id": 54624, "pid": 76337, "tid": -914061504, "ts": 1716454218272185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218340607, "dur": 5,
    "args": {
      "External id": 54626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54626,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 54626, "pid": 5, "tid": 7, "ts": 1716454218340607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272199, "dur": 6,
    "args": {
      "External id": 54626,
      "cbid": 211, "correlation": 54626
    }
  },
  {
    "ph": "s", "id": 54626, "pid": 76337, "tid": -914061504, "ts": 1716454218272199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272256, "dur": 1,
    "args": {
      "External id": 54637,
      "cbid": 251, "correlation": 54637
    }
  },
  {
    "ph": "f", "id": 54637, "pid": 76337, "tid": -914061504, "ts": 1716454218272256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272259, "dur": 0,
    "args": {
      "External id": 54638,
      "cbid": 251, "correlation": 54638
    }
  },
  {
    "ph": "f", "id": 54638, "pid": 76337, "tid": -914061504, "ts": 1716454218272259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218340614, "dur": 7,
    "args": {
      "External id": 54639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54639,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 54639, "pid": 5, "tid": 7, "ts": 1716454218340614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272261, "dur": 11,
    "args": {
      "External id": 54639,
      "cbid": 211, "correlation": 54639
    }
  },
  {
    "ph": "s", "id": 54639, "pid": 76337, "tid": -914061504, "ts": 1716454218272261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218340623, "dur": 4,
    "args": {
      "External id": 54641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54641,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 54641, "pid": 5, "tid": 7, "ts": 1716454218340623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272273, "dur": 5,
    "args": {
      "External id": 54641,
      "cbid": 211, "correlation": 54641
    }
  },
  {
    "ph": "s", "id": 54641, "pid": 76337, "tid": -914061504, "ts": 1716454218272273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218340628, "dur": 89,
    "args": {
      "External id": 54662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54662,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 54662, "pid": 5, "tid": 7, "ts": 1716454218340628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272349, "dur": 14,
    "args": {
      "External id": 54662,
      "cbid": 211, "correlation": 54662
    }
  },
  {
    "ph": "s", "id": 54662, "pid": 76337, "tid": -914061504, "ts": 1716454218272349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272447, "dur": 1,
    "args": {
      "External id": 54680,
      "cbid": 251, "correlation": 54680
    }
  },
  {
    "ph": "f", "id": 54680, "pid": 76337, "tid": -914061504, "ts": 1716454218272447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218340718, "dur": 97,
    "args": {
      "External id": 54682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54682,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54682, "pid": 5, "tid": 7, "ts": 1716454218340718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272453, "dur": 14,
    "args": {
      "External id": 54682,
      "cbid": 211, "correlation": 54682
    }
  },
  {
    "ph": "s", "id": 54682, "pid": 76337, "tid": -914061504, "ts": 1716454218272453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218340816, "dur": 19,
    "args": {
      "External id": 54690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54690,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54690, "pid": 5, "tid": 7, "ts": 1716454218340816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272521, "dur": 12,
    "args": {
      "External id": 54690,
      "cbid": 211, "correlation": 54690
    }
  },
  {
    "ph": "s", "id": 54690, "pid": 76337, "tid": -914061504, "ts": 1716454218272521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218340837, "dur": 37,
    "args": {
      "External id": 54698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54698, "pid": 5, "tid": 7, "ts": 1716454218340837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272562, "dur": 10,
    "args": {
      "External id": 54698,
      "cbid": 211, "correlation": 54698
    }
  },
  {
    "ph": "s", "id": 54698, "pid": 76337, "tid": -914061504, "ts": 1716454218272562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218340875, "dur": 34,
    "args": {
      "External id": 54720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54720,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54720, "pid": 5, "tid": 7, "ts": 1716454218340875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272613, "dur": 10,
    "args": {
      "External id": 54720,
      "cbid": 211, "correlation": 54720
    }
  },
  {
    "ph": "s", "id": 54720, "pid": 76337, "tid": -914061504, "ts": 1716454218272613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272702, "dur": 4,
    "args": {
      "External id": 54736,
      "cbid": 251, "correlation": 54736
    }
  },
  {
    "ph": "f", "id": 54736, "pid": 76337, "tid": -914061504, "ts": 1716454218272702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272710, "dur": 0,
    "args": {
      "External id": 54738,
      "cbid": 251, "correlation": 54738
    }
  },
  {
    "ph": "f", "id": 54738, "pid": 76337, "tid": -914061504, "ts": 1716454218272710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218340911, "dur": 536,
    "args": {
      "External id": 54739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54739,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 54739, "pid": 5, "tid": 7, "ts": 1716454218340911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272714, "dur": 14,
    "args": {
      "External id": 54739,
      "cbid": 211, "correlation": 54739
    }
  },
  {
    "ph": "s", "id": 54739, "pid": 76337, "tid": -914061504, "ts": 1716454218272714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218341448, "dur": 124,
    "args": {
      "External id": 54747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54747, "pid": 5, "tid": 7, "ts": 1716454218341448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272781, "dur": 12,
    "args": {
      "External id": 54747,
      "cbid": 211, "correlation": 54747
    }
  },
  {
    "ph": "s", "id": 54747, "pid": 76337, "tid": -914061504, "ts": 1716454218272781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218341574, "dur": 129,
    "args": {
      "External id": 54755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54755,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54755, "pid": 5, "tid": 7, "ts": 1716454218341574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272812, "dur": 8,
    "args": {
      "External id": 54755,
      "cbid": 211, "correlation": 54755
    }
  },
  {
    "ph": "s", "id": 54755, "pid": 76337, "tid": -914061504, "ts": 1716454218272812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272890, "dur": 1,
    "args": {
      "External id": 54771,
      "cbid": 251, "correlation": 54771
    }
  },
  {
    "ph": "f", "id": 54771, "pid": 76337, "tid": -914061504, "ts": 1716454218272890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218341704, "dur": 303,
    "args": {
      "External id": 54773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54773,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54773, "pid": 5, "tid": 7, "ts": 1716454218341704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272895, "dur": 12,
    "args": {
      "External id": 54773,
      "cbid": 211, "correlation": 54773
    }
  },
  {
    "ph": "s", "id": 54773, "pid": 76337, "tid": -914061504, "ts": 1716454218272895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218342008, "dur": 27,
    "args": {
      "External id": 54781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54781,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54781, "pid": 5, "tid": 7, "ts": 1716454218342008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218272938, "dur": 10,
    "args": {
      "External id": 54781,
      "cbid": 211, "correlation": 54781
    }
  },
  {
    "ph": "s", "id": 54781, "pid": 76337, "tid": -914061504, "ts": 1716454218272938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218342037, "dur": 81,
    "args": {
      "External id": 54792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54792,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54792, "pid": 5, "tid": 7, "ts": 1716454218342037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273013, "dur": 13,
    "args": {
      "External id": 54792,
      "cbid": 211, "correlation": 54792
    }
  },
  {
    "ph": "s", "id": 54792, "pid": 76337, "tid": -914061504, "ts": 1716454218273013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273077, "dur": 0,
    "args": {
      "External id": 54804,
      "cbid": 317, "correlation": 54804
    }
  },
  {
    "ph": "f", "id": 54804, "pid": 76337, "tid": -914061504, "ts": 1716454218273077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273078, "dur": 0,
    "args": {
      "External id": 54805,
      "cbid": 203, "correlation": 54805
    }
  },
  {
    "ph": "f", "id": 54805, "pid": 76337, "tid": -914061504, "ts": 1716454218273078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273079, "dur": 0,
    "args": {
      "External id": 54806,
      "cbid": 205, "correlation": 54806
    }
  },
  {
    "ph": "f", "id": 54806, "pid": 76337, "tid": -914061504, "ts": 1716454218273079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218342119, "dur": 23,
    "args": {
      "External id": 54810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54810, "pid": 5, "tid": 7, "ts": 1716454218342119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273094, "dur": 15,
    "args": {
      "External id": 54810,
      "cbid": 211, "correlation": 54810
    }
  },
  {
    "ph": "s", "id": 54810, "pid": 76337, "tid": -914061504, "ts": 1716454218273094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218342143, "dur": 119,
    "args": {
      "External id": 54812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54812,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54812, "pid": 5, "tid": 7, "ts": 1716454218342143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273115, "dur": 7,
    "args": {
      "External id": 54812,
      "cbid": 211, "correlation": 54812
    }
  },
  {
    "ph": "s", "id": 54812, "pid": 76337, "tid": -914061504, "ts": 1716454218273115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218342263, "dur": 24,
    "args": {
      "External id": 54814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54814,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54814, "pid": 5, "tid": 7, "ts": 1716454218342263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273125, "dur": 5,
    "args": {
      "External id": 54814,
      "cbid": 211, "correlation": 54814
    }
  },
  {
    "ph": "s", "id": 54814, "pid": 76337, "tid": -914061504, "ts": 1716454218273125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218342289, "dur": 33,
    "args": {
      "External id": 54820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54820, "pid": 5, "tid": 7, "ts": 1716454218342289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273154, "dur": 8,
    "args": {
      "External id": 54820,
      "cbid": 211, "correlation": 54820
    }
  },
  {
    "ph": "s", "id": 54820, "pid": 76337, "tid": -914061504, "ts": 1716454218273154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218342323, "dur": 26,
    "args": {
      "External id": 54828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54828,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54828, "pid": 5, "tid": 7, "ts": 1716454218342323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273185, "dur": 9,
    "args": {
      "External id": 54828,
      "cbid": 211, "correlation": 54828
    }
  },
  {
    "ph": "s", "id": 54828, "pid": 76337, "tid": -914061504, "ts": 1716454218273185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218342351, "dur": 53,
    "args": {
      "External id": 54837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54837,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54837, "pid": 5, "tid": 7, "ts": 1716454218342351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273223, "dur": 10,
    "args": {
      "External id": 54837,
      "cbid": 211, "correlation": 54837
    }
  },
  {
    "ph": "s", "id": 54837, "pid": 76337, "tid": -914061504, "ts": 1716454218273223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218342405, "dur": 53,
    "args": {
      "External id": 54857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54857,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 54857, "pid": 5, "tid": 7, "ts": 1716454218342405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273294, "dur": 12,
    "args": {
      "External id": 54857,
      "cbid": 211, "correlation": 54857
    }
  },
  {
    "ph": "s", "id": 54857, "pid": 76337, "tid": -914061504, "ts": 1716454218273294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218342459, "dur": 5,
    "args": {
      "External id": 54869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54869,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 54869, "pid": 5, "tid": 7, "ts": 1716454218342459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273316, "dur": 6,
    "args": {
      "External id": 54869,
      "cbid": 211, "correlation": 54869
    }
  },
  {
    "ph": "s", "id": 54869, "pid": 76337, "tid": -914061504, "ts": 1716454218273316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218342465, "dur": 57,
    "args": {
      "External id": 54872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54872,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54872, "pid": 5, "tid": 7, "ts": 1716454218342465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273334, "dur": 7,
    "args": {
      "External id": 54872,
      "cbid": 211, "correlation": 54872
    }
  },
  {
    "ph": "s", "id": 54872, "pid": 76337, "tid": -914061504, "ts": 1716454218273334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218342523, "dur": 37,
    "args": {
      "External id": 54881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54881,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54881, "pid": 5, "tid": 7, "ts": 1716454218342523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273377, "dur": 10,
    "args": {
      "External id": 54881,
      "cbid": 211, "correlation": 54881
    }
  },
  {
    "ph": "s", "id": 54881, "pid": 76337, "tid": -914061504, "ts": 1716454218273377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273431, "dur": 0,
    "args": {
      "External id": 54891,
      "cbid": 317, "correlation": 54891
    }
  },
  {
    "ph": "f", "id": 54891, "pid": 76337, "tid": -914061504, "ts": 1716454218273431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273432, "dur": 0,
    "args": {
      "External id": 54892,
      "cbid": 203, "correlation": 54892
    }
  },
  {
    "ph": "f", "id": 54892, "pid": 76337, "tid": -914061504, "ts": 1716454218273432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273432, "dur": 0,
    "args": {
      "External id": 54893,
      "cbid": 205, "correlation": 54893
    }
  },
  {
    "ph": "f", "id": 54893, "pid": 76337, "tid": -914061504, "ts": 1716454218273432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218342561, "dur": 39,
    "args": {
      "External id": 54897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54897,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54897, "pid": 5, "tid": 7, "ts": 1716454218342561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273447, "dur": 12,
    "args": {
      "External id": 54897,
      "cbid": 211, "correlation": 54897
    }
  },
  {
    "ph": "s", "id": 54897, "pid": 76337, "tid": -914061504, "ts": 1716454218273447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218342602, "dur": 83,
    "args": {
      "External id": 54899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54899,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54899, "pid": 5, "tid": 7, "ts": 1716454218342602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273462, "dur": 5,
    "args": {
      "External id": 54899,
      "cbid": 211, "correlation": 54899
    }
  },
  {
    "ph": "s", "id": 54899, "pid": 76337, "tid": -914061504, "ts": 1716454218273462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218342686, "dur": 1270,
    "args": {
      "External id": 54901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54901,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 54901, "pid": 5, "tid": 7, "ts": 1716454218342686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273473, "dur": 6,
    "args": {
      "External id": 54901,
      "cbid": 211, "correlation": 54901
    }
  },
  {
    "ph": "s", "id": 54901, "pid": 76337, "tid": -914061504, "ts": 1716454218273473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218343957, "dur": 21,
    "args": {
      "External id": 54903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54903,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54903, "pid": 5, "tid": 7, "ts": 1716454218343957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273483, "dur": 5,
    "args": {
      "External id": 54903,
      "cbid": 211, "correlation": 54903
    }
  },
  {
    "ph": "s", "id": 54903, "pid": 76337, "tid": -914061504, "ts": 1716454218273483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218343980, "dur": 33,
    "args": {
      "External id": 54909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54909, "pid": 5, "tid": 7, "ts": 1716454218343980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273511, "dur": 8,
    "args": {
      "External id": 54909,
      "cbid": 211, "correlation": 54909
    }
  },
  {
    "ph": "s", "id": 54909, "pid": 76337, "tid": -914061504, "ts": 1716454218273511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218344014, "dur": 3,
    "args": {
      "External id": 54917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54917,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 54917, "pid": 5, "tid": 7, "ts": 1716454218344014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273553, "dur": 10,
    "args": {
      "External id": 54917,
      "cbid": 211, "correlation": 54917
    }
  },
  {
    "ph": "s", "id": 54917, "pid": 76337, "tid": -914061504, "ts": 1716454218273553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273619, "dur": 1,
    "args": {
      "External id": 54933,
      "cbid": 251, "correlation": 54933
    }
  },
  {
    "ph": "f", "id": 54933, "pid": 76337, "tid": -914061504, "ts": 1716454218273619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273624, "dur": 0,
    "args": {
      "External id": 54935,
      "cbid": 251, "correlation": 54935
    }
  },
  {
    "ph": "f", "id": 54935, "pid": 76337, "tid": -914061504, "ts": 1716454218273624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218344019, "dur": 12,
    "args": {
      "External id": 54936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54936,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 54936, "pid": 5, "tid": 7, "ts": 1716454218344019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273626, "dur": 11,
    "args": {
      "External id": 54936,
      "cbid": 211, "correlation": 54936
    }
  },
  {
    "ph": "s", "id": 54936, "pid": 76337, "tid": -914061504, "ts": 1716454218273626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218344032, "dur": 5,
    "args": {
      "External id": 54938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54938,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 54938, "pid": 5, "tid": 7, "ts": 1716454218344032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273639, "dur": 8,
    "args": {
      "External id": 54938,
      "cbid": 211, "correlation": 54938
    }
  },
  {
    "ph": "s", "id": 54938, "pid": 76337, "tid": -914061504, "ts": 1716454218273639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218344038, "dur": 28,
    "args": {
      "External id": 54948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54948, "pid": 5, "tid": 7, "ts": 1716454218344038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273700, "dur": 12,
    "args": {
      "External id": 54948,
      "cbid": 211, "correlation": 54948
    }
  },
  {
    "ph": "s", "id": 54948, "pid": 76337, "tid": -914061504, "ts": 1716454218273700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218344068, "dur": 30,
    "args": {
      "External id": 54968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54968,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 54968, "pid": 5, "tid": 7, "ts": 1716454218344068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273766, "dur": 11,
    "args": {
      "External id": 54968,
      "cbid": 211, "correlation": 54968
    }
  },
  {
    "ph": "s", "id": 54968, "pid": 76337, "tid": -914061504, "ts": 1716454218273766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218344099, "dur": 4,
    "args": {
      "External id": 54980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54980,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 54980, "pid": 5, "tid": 7, "ts": 1716454218344099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273786, "dur": 6,
    "args": {
      "External id": 54980,
      "cbid": 211, "correlation": 54980
    }
  },
  {
    "ph": "s", "id": 54980, "pid": 76337, "tid": -914061504, "ts": 1716454218273786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218344104, "dur": 30,
    "args": {
      "External id": 54983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54983, "pid": 5, "tid": 7, "ts": 1716454218344104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273805, "dur": 6,
    "args": {
      "External id": 54983,
      "cbid": 211, "correlation": 54983
    }
  },
  {
    "ph": "s", "id": 54983, "pid": 76337, "tid": -914061504, "ts": 1716454218273805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218344135, "dur": 20,
    "args": {
      "External id": 54992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 54992,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 54992, "pid": 5, "tid": 7, "ts": 1716454218344135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273846, "dur": 10,
    "args": {
      "External id": 54992,
      "cbid": 211, "correlation": 54992
    }
  },
  {
    "ph": "s", "id": 54992, "pid": 76337, "tid": -914061504, "ts": 1716454218273846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273908, "dur": 0,
    "args": {
      "External id": 55002,
      "cbid": 317, "correlation": 55002
    }
  },
  {
    "ph": "f", "id": 55002, "pid": 76337, "tid": -914061504, "ts": 1716454218273908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273909, "dur": 0,
    "args": {
      "External id": 55003,
      "cbid": 203, "correlation": 55003
    }
  },
  {
    "ph": "f", "id": 55003, "pid": 76337, "tid": -914061504, "ts": 1716454218273909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273910, "dur": 0,
    "args": {
      "External id": 55004,
      "cbid": 205, "correlation": 55004
    }
  },
  {
    "ph": "f", "id": 55004, "pid": 76337, "tid": -914061504, "ts": 1716454218273910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218344156, "dur": 23,
    "args": {
      "External id": 55008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55008,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55008, "pid": 5, "tid": 7, "ts": 1716454218344156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273924, "dur": 12,
    "args": {
      "External id": 55008,
      "cbid": 211, "correlation": 55008
    }
  },
  {
    "ph": "s", "id": 55008, "pid": 76337, "tid": -914061504, "ts": 1716454218273924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218344180, "dur": 44,
    "args": {
      "External id": 55010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55010,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55010, "pid": 5, "tid": 7, "ts": 1716454218344180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273939, "dur": 8,
    "args": {
      "External id": 55010,
      "cbid": 211, "correlation": 55010
    }
  },
  {
    "ph": "s", "id": 55010, "pid": 76337, "tid": -914061504, "ts": 1716454218273939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218344226, "dur": 643,
    "args": {
      "External id": 55012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55012,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55012, "pid": 5, "tid": 7, "ts": 1716454218344226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273953, "dur": 6,
    "args": {
      "External id": 55012,
      "cbid": 211, "correlation": 55012
    }
  },
  {
    "ph": "s", "id": 55012, "pid": 76337, "tid": -914061504, "ts": 1716454218273953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218344870, "dur": 22,
    "args": {
      "External id": 55014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55014, "pid": 5, "tid": 7, "ts": 1716454218344870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273963, "dur": 5,
    "args": {
      "External id": 55014,
      "cbid": 211, "correlation": 55014
    }
  },
  {
    "ph": "s", "id": 55014, "pid": 76337, "tid": -914061504, "ts": 1716454218273963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218344893, "dur": 33,
    "args": {
      "External id": 55020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55020,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55020, "pid": 5, "tid": 7, "ts": 1716454218344893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218273999, "dur": 9,
    "args": {
      "External id": 55020,
      "cbid": 211, "correlation": 55020
    }
  },
  {
    "ph": "s", "id": 55020, "pid": 76337, "tid": -914061504, "ts": 1716454218273999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274059, "dur": 0,
    "args": {
      "External id": 55030,
      "cbid": 317, "correlation": 55030
    }
  },
  {
    "ph": "f", "id": 55030, "pid": 76337, "tid": -914061504, "ts": 1716454218274059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274060, "dur": 0,
    "args": {
      "External id": 55031,
      "cbid": 203, "correlation": 55031
    }
  },
  {
    "ph": "f", "id": 55031, "pid": 76337, "tid": -914061504, "ts": 1716454218274060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274061, "dur": 0,
    "args": {
      "External id": 55032,
      "cbid": 205, "correlation": 55032
    }
  },
  {
    "ph": "f", "id": 55032, "pid": 76337, "tid": -914061504, "ts": 1716454218274061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218344927, "dur": 38,
    "args": {
      "External id": 55036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55036,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55036, "pid": 5, "tid": 7, "ts": 1716454218344927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274073, "dur": 12,
    "args": {
      "External id": 55036,
      "cbid": 211, "correlation": 55036
    }
  },
  {
    "ph": "s", "id": 55036, "pid": 76337, "tid": -914061504, "ts": 1716454218274073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218344967, "dur": 189,
    "args": {
      "External id": 55038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55038,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55038, "pid": 5, "tid": 7, "ts": 1716454218344967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274092, "dur": 6,
    "args": {
      "External id": 55038,
      "cbid": 211, "correlation": 55038
    }
  },
  {
    "ph": "s", "id": 55038, "pid": 76337, "tid": -914061504, "ts": 1716454218274092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218345157, "dur": 21,
    "args": {
      "External id": 55040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55040,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55040, "pid": 5, "tid": 7, "ts": 1716454218345157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274101, "dur": 5,
    "args": {
      "External id": 55040,
      "cbid": 211, "correlation": 55040
    }
  },
  {
    "ph": "s", "id": 55040, "pid": 76337, "tid": -914061504, "ts": 1716454218274101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218345180, "dur": 32,
    "args": {
      "External id": 55046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55046,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55046, "pid": 5, "tid": 7, "ts": 1716454218345180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274126, "dur": 9,
    "args": {
      "External id": 55046,
      "cbid": 211, "correlation": 55046
    }
  },
  {
    "ph": "s", "id": 55046, "pid": 76337, "tid": -914061504, "ts": 1716454218274126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218345213, "dur": 27,
    "args": {
      "External id": 55054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55054,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55054, "pid": 5, "tid": 7, "ts": 1716454218345213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274155, "dur": 8,
    "args": {
      "External id": 55054,
      "cbid": 211, "correlation": 55054
    }
  },
  {
    "ph": "s", "id": 55054, "pid": 76337, "tid": -914061504, "ts": 1716454218274155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218345242, "dur": 20,
    "args": {
      "External id": 55062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55062,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55062, "pid": 5, "tid": 7, "ts": 1716454218345242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274184, "dur": 8,
    "args": {
      "External id": 55062,
      "cbid": 211, "correlation": 55062
    }
  },
  {
    "ph": "s", "id": 55062, "pid": 76337, "tid": -914061504, "ts": 1716454218274184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218345263, "dur": 30,
    "args": {
      "External id": 55082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55082,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 55082, "pid": 5, "tid": 7, "ts": 1716454218345263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274268, "dur": 13,
    "args": {
      "External id": 55082,
      "cbid": 211, "correlation": 55082
    }
  },
  {
    "ph": "s", "id": 55082, "pid": 76337, "tid": -914061504, "ts": 1716454218274268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218345294, "dur": 4,
    "args": {
      "External id": 55094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55094,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 55094, "pid": 5, "tid": 7, "ts": 1716454218345294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274292, "dur": 6,
    "args": {
      "External id": 55094,
      "cbid": 211, "correlation": 55094
    }
  },
  {
    "ph": "s", "id": 55094, "pid": 76337, "tid": -914061504, "ts": 1716454218274292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218345299, "dur": 30,
    "args": {
      "External id": 55097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55097, "pid": 5, "tid": 7, "ts": 1716454218345299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274310, "dur": 6,
    "args": {
      "External id": 55097,
      "cbid": 211, "correlation": 55097
    }
  },
  {
    "ph": "s", "id": 55097, "pid": 76337, "tid": -914061504, "ts": 1716454218274310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274367, "dur": 0,
    "args": {
      "External id": 55108,
      "cbid": 317, "correlation": 55108
    }
  },
  {
    "ph": "f", "id": 55108, "pid": 76337, "tid": -914061504, "ts": 1716454218274367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274368, "dur": 0,
    "args": {
      "External id": 55109,
      "cbid": 203, "correlation": 55109
    }
  },
  {
    "ph": "f", "id": 55109, "pid": 76337, "tid": -914061504, "ts": 1716454218274368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274369, "dur": 0,
    "args": {
      "External id": 55110,
      "cbid": 205, "correlation": 55110
    }
  },
  {
    "ph": "f", "id": 55110, "pid": 76337, "tid": -914061504, "ts": 1716454218274369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218345331, "dur": 22,
    "args": {
      "External id": 55114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55114,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55114, "pid": 5, "tid": 7, "ts": 1716454218345331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274381, "dur": 12,
    "args": {
      "External id": 55114,
      "cbid": 211, "correlation": 55114
    }
  },
  {
    "ph": "s", "id": 55114, "pid": 76337, "tid": -914061504, "ts": 1716454218274381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218345354, "dur": 104,
    "args": {
      "External id": 55116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55116,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55116, "pid": 5, "tid": 7, "ts": 1716454218345354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274400, "dur": 6,
    "args": {
      "External id": 55116,
      "cbid": 211, "correlation": 55116
    }
  },
  {
    "ph": "s", "id": 55116, "pid": 76337, "tid": -914061504, "ts": 1716454218274400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218345459, "dur": 22,
    "args": {
      "External id": 55118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55118, "pid": 5, "tid": 7, "ts": 1716454218345459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274409, "dur": 5,
    "args": {
      "External id": 55118,
      "cbid": 211, "correlation": 55118
    }
  },
  {
    "ph": "s", "id": 55118, "pid": 76337, "tid": -914061504, "ts": 1716454218274409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218345482, "dur": 32,
    "args": {
      "External id": 55124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55124,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55124, "pid": 5, "tid": 7, "ts": 1716454218345482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274436, "dur": 9,
    "args": {
      "External id": 55124,
      "cbid": 211, "correlation": 55124
    }
  },
  {
    "ph": "s", "id": 55124, "pid": 76337, "tid": -914061504, "ts": 1716454218274436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218345516, "dur": 184,
    "args": {
      "External id": 55133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55133,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55133, "pid": 5, "tid": 7, "ts": 1716454218345516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274517, "dur": 14,
    "args": {
      "External id": 55133,
      "cbid": 211, "correlation": 55133
    }
  },
  {
    "ph": "s", "id": 55133, "pid": 76337, "tid": -914061504, "ts": 1716454218274517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218345702, "dur": 64,
    "args": {
      "External id": 55155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55155,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55155, "pid": 5, "tid": 7, "ts": 1716454218345702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274577, "dur": 11,
    "args": {
      "External id": 55155,
      "cbid": 211, "correlation": 55155
    }
  },
  {
    "ph": "s", "id": 55155, "pid": 76337, "tid": -914061504, "ts": 1716454218274577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274666, "dur": 1,
    "args": {
      "External id": 55166,
      "cbid": 251, "correlation": 55166
    }
  },
  {
    "ph": "f", "id": 55166, "pid": 76337, "tid": -914061504, "ts": 1716454218274666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218345767, "dur": 151,
    "args": {
      "External id": 55167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55167,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55167, "pid": 5, "tid": 7, "ts": 1716454218345767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274672, "dur": 13,
    "args": {
      "External id": 55167,
      "cbid": 211, "correlation": 55167
    }
  },
  {
    "ph": "s", "id": 55167, "pid": 76337, "tid": -914061504, "ts": 1716454218274672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274741, "dur": 1,
    "args": {
      "External id": 55178,
      "cbid": 251, "correlation": 55178
    }
  },
  {
    "ph": "f", "id": 55178, "pid": 76337, "tid": -914061504, "ts": 1716454218274741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218345920, "dur": 146,
    "args": {
      "External id": 55179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55179,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55179, "pid": 5, "tid": 7, "ts": 1716454218345920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274745, "dur": 11,
    "args": {
      "External id": 55179,
      "cbid": 211, "correlation": 55179
    }
  },
  {
    "ph": "s", "id": 55179, "pid": 76337, "tid": -914061504, "ts": 1716454218274745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274810, "dur": 1,
    "args": {
      "External id": 55190,
      "cbid": 251, "correlation": 55190
    }
  },
  {
    "ph": "f", "id": 55190, "pid": 76337, "tid": -914061504, "ts": 1716454218274810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218346067, "dur": 144,
    "args": {
      "External id": 55191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55191,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55191, "pid": 5, "tid": 7, "ts": 1716454218346067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274815, "dur": 11,
    "args": {
      "External id": 55191,
      "cbid": 211, "correlation": 55191
    }
  },
  {
    "ph": "s", "id": 55191, "pid": 76337, "tid": -914061504, "ts": 1716454218274815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218346212, "dur": 1923,
    "args": {
      "External id": 55212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55212,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 55212, "pid": 5, "tid": 7, "ts": 1716454218346212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218274893, "dur": 13,
    "args": {
      "External id": 55212,
      "cbid": 211, "correlation": 55212
    }
  },
  {
    "ph": "s", "id": 55212, "pid": 76337, "tid": -914061504, "ts": 1716454218274893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275004, "dur": 1,
    "args": {
      "External id": 55230,
      "cbid": 251, "correlation": 55230
    }
  },
  {
    "ph": "f", "id": 55230, "pid": 76337, "tid": -914061504, "ts": 1716454218275004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218348137, "dur": 148,
    "args": {
      "External id": 55232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55232,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 55232, "pid": 5, "tid": 7, "ts": 1716454218348137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275009, "dur": 15,
    "args": {
      "External id": 55232,
      "cbid": 211, "correlation": 55232
    }
  },
  {
    "ph": "s", "id": 55232, "pid": 76337, "tid": -914061504, "ts": 1716454218275009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218348286, "dur": 36,
    "args": {
      "External id": 55240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55240,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55240, "pid": 5, "tid": 7, "ts": 1716454218348286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275082, "dur": 12,
    "args": {
      "External id": 55240,
      "cbid": 211, "correlation": 55240
    }
  },
  {
    "ph": "s", "id": 55240, "pid": 76337, "tid": -914061504, "ts": 1716454218275082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218348323, "dur": 51,
    "args": {
      "External id": 55248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55248,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55248, "pid": 5, "tid": 7, "ts": 1716454218348323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275120, "dur": 9,
    "args": {
      "External id": 55248,
      "cbid": 211, "correlation": 55248
    }
  },
  {
    "ph": "s", "id": 55248, "pid": 76337, "tid": -914061504, "ts": 1716454218275120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218348375, "dur": 30,
    "args": {
      "External id": 55259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55259,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55259, "pid": 5, "tid": 7, "ts": 1716454218348375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275191, "dur": 12,
    "args": {
      "External id": 55259,
      "cbid": 211, "correlation": 55259
    }
  },
  {
    "ph": "s", "id": 55259, "pid": 76337, "tid": -914061504, "ts": 1716454218275191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218348406, "dur": 34,
    "args": {
      "External id": 55281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55281,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55281, "pid": 5, "tid": 7, "ts": 1716454218348406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275222, "dur": 8,
    "args": {
      "External id": 55281,
      "cbid": 211, "correlation": 55281
    }
  },
  {
    "ph": "s", "id": 55281, "pid": 76337, "tid": -914061504, "ts": 1716454218275222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275307, "dur": 1,
    "args": {
      "External id": 55292,
      "cbid": 251, "correlation": 55292
    }
  },
  {
    "ph": "f", "id": 55292, "pid": 76337, "tid": -914061504, "ts": 1716454218275307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218348442, "dur": 90,
    "args": {
      "External id": 55293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55293,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55293, "pid": 5, "tid": 7, "ts": 1716454218348442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275313, "dur": 12,
    "args": {
      "External id": 55293,
      "cbid": 211, "correlation": 55293
    }
  },
  {
    "ph": "s", "id": 55293, "pid": 76337, "tid": -914061504, "ts": 1716454218275313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275385, "dur": 1,
    "args": {
      "External id": 55304,
      "cbid": 251, "correlation": 55304
    }
  },
  {
    "ph": "f", "id": 55304, "pid": 76337, "tid": -914061504, "ts": 1716454218275385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275389, "dur": 0,
    "args": {
      "External id": 55305,
      "cbid": 251, "correlation": 55305
    }
  },
  {
    "ph": "f", "id": 55305, "pid": 76337, "tid": -914061504, "ts": 1716454218275389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218348533, "dur": 11,
    "args": {
      "External id": 55306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55306,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 55306, "pid": 5, "tid": 7, "ts": 1716454218348533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275390, "dur": 12,
    "args": {
      "External id": 55306,
      "cbid": 211, "correlation": 55306
    }
  },
  {
    "ph": "s", "id": 55306, "pid": 76337, "tid": -914061504, "ts": 1716454218275390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218348545, "dur": 5,
    "args": {
      "External id": 55308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55308,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 55308, "pid": 5, "tid": 7, "ts": 1716454218348545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275404, "dur": 6,
    "args": {
      "External id": 55308,
      "cbid": 211, "correlation": 55308
    }
  },
  {
    "ph": "s", "id": 55308, "pid": 76337, "tid": -914061504, "ts": 1716454218275404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275461, "dur": 1,
    "args": {
      "External id": 55319,
      "cbid": 251, "correlation": 55319
    }
  },
  {
    "ph": "f", "id": 55319, "pid": 76337, "tid": -914061504, "ts": 1716454218275461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275464, "dur": 0,
    "args": {
      "External id": 55320,
      "cbid": 251, "correlation": 55320
    }
  },
  {
    "ph": "f", "id": 55320, "pid": 76337, "tid": -914061504, "ts": 1716454218275464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218348552, "dur": 7,
    "args": {
      "External id": 55321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55321,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 55321, "pid": 5, "tid": 7, "ts": 1716454218348552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275466, "dur": 11,
    "args": {
      "External id": 55321,
      "cbid": 211, "correlation": 55321
    }
  },
  {
    "ph": "s", "id": 55321, "pid": 76337, "tid": -914061504, "ts": 1716454218275466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218348560, "dur": 4,
    "args": {
      "External id": 55323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55323,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 55323, "pid": 5, "tid": 7, "ts": 1716454218348560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275479, "dur": 6,
    "args": {
      "External id": 55323,
      "cbid": 211, "correlation": 55323
    }
  },
  {
    "ph": "s", "id": 55323, "pid": 76337, "tid": -914061504, "ts": 1716454218275479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218348565, "dur": 91,
    "args": {
      "External id": 55344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55344,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 55344, "pid": 5, "tid": 7, "ts": 1716454218348565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275553, "dur": 12,
    "args": {
      "External id": 55344,
      "cbid": 211, "correlation": 55344
    }
  },
  {
    "ph": "s", "id": 55344, "pid": 76337, "tid": -914061504, "ts": 1716454218275553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275649, "dur": 1,
    "args": {
      "External id": 55362,
      "cbid": 251, "correlation": 55362
    }
  },
  {
    "ph": "f", "id": 55362, "pid": 76337, "tid": -914061504, "ts": 1716454218275649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218348657, "dur": 100,
    "args": {
      "External id": 55364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55364,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55364, "pid": 5, "tid": 7, "ts": 1716454218348657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275655, "dur": 13,
    "args": {
      "External id": 55364,
      "cbid": 211, "correlation": 55364
    }
  },
  {
    "ph": "s", "id": 55364, "pid": 76337, "tid": -914061504, "ts": 1716454218275655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218348759, "dur": 20,
    "args": {
      "External id": 55372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55372,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55372, "pid": 5, "tid": 7, "ts": 1716454218348759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275724, "dur": 14,
    "args": {
      "External id": 55372,
      "cbid": 211, "correlation": 55372
    }
  },
  {
    "ph": "s", "id": 55372, "pid": 76337, "tid": -914061504, "ts": 1716454218275724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218348779, "dur": 38,
    "args": {
      "External id": 55380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55380, "pid": 5, "tid": 7, "ts": 1716454218348779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275770, "dur": 10,
    "args": {
      "External id": 55380,
      "cbid": 211, "correlation": 55380
    }
  },
  {
    "ph": "s", "id": 55380, "pid": 76337, "tid": -914061504, "ts": 1716454218275770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218348819, "dur": 35,
    "args": {
      "External id": 55402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55402,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55402, "pid": 5, "tid": 7, "ts": 1716454218348819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275821, "dur": 10,
    "args": {
      "External id": 55402,
      "cbid": 211, "correlation": 55402
    }
  },
  {
    "ph": "s", "id": 55402, "pid": 76337, "tid": -914061504, "ts": 1716454218275821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275910, "dur": 1,
    "args": {
      "External id": 55418,
      "cbid": 251, "correlation": 55418
    }
  },
  {
    "ph": "f", "id": 55418, "pid": 76337, "tid": -914061504, "ts": 1716454218275910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275915, "dur": 0,
    "args": {
      "External id": 55420,
      "cbid": 251, "correlation": 55420
    }
  },
  {
    "ph": "f", "id": 55420, "pid": 76337, "tid": -914061504, "ts": 1716454218275915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218348855, "dur": 534,
    "args": {
      "External id": 55421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55421,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 55421, "pid": 5, "tid": 7, "ts": 1716454218348855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275918, "dur": 13,
    "args": {
      "External id": 55421,
      "cbid": 211, "correlation": 55421
    }
  },
  {
    "ph": "s", "id": 55421, "pid": 76337, "tid": -914061504, "ts": 1716454218275918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218349390, "dur": 126,
    "args": {
      "External id": 55429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55429, "pid": 5, "tid": 7, "ts": 1716454218349390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218275991, "dur": 13,
    "args": {
      "External id": 55429,
      "cbid": 211, "correlation": 55429
    }
  },
  {
    "ph": "s", "id": 55429, "pid": 76337, "tid": -914061504, "ts": 1716454218275991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218349517, "dur": 129,
    "args": {
      "External id": 55437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55437, "pid": 5, "tid": 7, "ts": 1716454218349517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276023, "dur": 9,
    "args": {
      "External id": 55437,
      "cbid": 211, "correlation": 55437
    }
  },
  {
    "ph": "s", "id": 55437, "pid": 76337, "tid": -914061504, "ts": 1716454218276023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276105, "dur": 1,
    "args": {
      "External id": 55453,
      "cbid": 251, "correlation": 55453
    }
  },
  {
    "ph": "f", "id": 55453, "pid": 76337, "tid": -914061504, "ts": 1716454218276105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218349647, "dur": 303,
    "args": {
      "External id": 55455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55455,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55455, "pid": 5, "tid": 7, "ts": 1716454218349647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276111, "dur": 12,
    "args": {
      "External id": 55455,
      "cbid": 211, "correlation": 55455
    }
  },
  {
    "ph": "s", "id": 55455, "pid": 76337, "tid": -914061504, "ts": 1716454218276111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218349951, "dur": 27,
    "args": {
      "External id": 55463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55463,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55463, "pid": 5, "tid": 7, "ts": 1716454218349951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276153, "dur": 10,
    "args": {
      "External id": 55463,
      "cbid": 211, "correlation": 55463
    }
  },
  {
    "ph": "s", "id": 55463, "pid": 76337, "tid": -914061504, "ts": 1716454218276153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218349979, "dur": 80,
    "args": {
      "External id": 55474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55474, "pid": 5, "tid": 7, "ts": 1716454218349979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276221, "dur": 12,
    "args": {
      "External id": 55474,
      "cbid": 211, "correlation": 55474
    }
  },
  {
    "ph": "s", "id": 55474, "pid": 76337, "tid": -914061504, "ts": 1716454218276221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276286, "dur": 0,
    "args": {
      "External id": 55486,
      "cbid": 317, "correlation": 55486
    }
  },
  {
    "ph": "f", "id": 55486, "pid": 76337, "tid": -914061504, "ts": 1716454218276286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276287, "dur": 0,
    "args": {
      "External id": 55487,
      "cbid": 203, "correlation": 55487
    }
  },
  {
    "ph": "f", "id": 55487, "pid": 76337, "tid": -914061504, "ts": 1716454218276287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276288, "dur": 0,
    "args": {
      "External id": 55488,
      "cbid": 205, "correlation": 55488
    }
  },
  {
    "ph": "f", "id": 55488, "pid": 76337, "tid": -914061504, "ts": 1716454218276288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218350061, "dur": 24,
    "args": {
      "External id": 55492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55492,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55492, "pid": 5, "tid": 7, "ts": 1716454218350061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276303, "dur": 12,
    "args": {
      "External id": 55492,
      "cbid": 211, "correlation": 55492
    }
  },
  {
    "ph": "s", "id": 55492, "pid": 76337, "tid": -914061504, "ts": 1716454218276303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218350086, "dur": 120,
    "args": {
      "External id": 55494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55494,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55494, "pid": 5, "tid": 7, "ts": 1716454218350086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276322, "dur": 7,
    "args": {
      "External id": 55494,
      "cbid": 211, "correlation": 55494
    }
  },
  {
    "ph": "s", "id": 55494, "pid": 76337, "tid": -914061504, "ts": 1716454218276322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218350207, "dur": 23,
    "args": {
      "External id": 55496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55496,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55496, "pid": 5, "tid": 7, "ts": 1716454218350207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276332, "dur": 5,
    "args": {
      "External id": 55496,
      "cbid": 211, "correlation": 55496
    }
  },
  {
    "ph": "s", "id": 55496, "pid": 76337, "tid": -914061504, "ts": 1716454218276332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218350232, "dur": 32,
    "args": {
      "External id": 55502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55502, "pid": 5, "tid": 7, "ts": 1716454218350232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276360, "dur": 8,
    "args": {
      "External id": 55502,
      "cbid": 211, "correlation": 55502
    }
  },
  {
    "ph": "s", "id": 55502, "pid": 76337, "tid": -914061504, "ts": 1716454218276360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218350265, "dur": 27,
    "args": {
      "External id": 55510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55510,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55510, "pid": 5, "tid": 7, "ts": 1716454218350265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276391, "dur": 9,
    "args": {
      "External id": 55510,
      "cbid": 211, "correlation": 55510
    }
  },
  {
    "ph": "s", "id": 55510, "pid": 76337, "tid": -914061504, "ts": 1716454218276391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218350293, "dur": 46,
    "args": {
      "External id": 55519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55519,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55519, "pid": 5, "tid": 7, "ts": 1716454218350293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276433, "dur": 10,
    "args": {
      "External id": 55519,
      "cbid": 211, "correlation": 55519
    }
  },
  {
    "ph": "s", "id": 55519, "pid": 76337, "tid": -914061504, "ts": 1716454218276433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218350340, "dur": 43,
    "args": {
      "External id": 55539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55539,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 55539, "pid": 5, "tid": 7, "ts": 1716454218350340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276505, "dur": 12,
    "args": {
      "External id": 55539,
      "cbid": 211, "correlation": 55539
    }
  },
  {
    "ph": "s", "id": 55539, "pid": 76337, "tid": -914061504, "ts": 1716454218276505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218350385, "dur": 5,
    "args": {
      "External id": 55551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55551,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 55551, "pid": 5, "tid": 7, "ts": 1716454218350385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276527, "dur": 6,
    "args": {
      "External id": 55551,
      "cbid": 211, "correlation": 55551
    }
  },
  {
    "ph": "s", "id": 55551, "pid": 76337, "tid": -914061504, "ts": 1716454218276527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218350391, "dur": 43,
    "args": {
      "External id": 55554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55554,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55554, "pid": 5, "tid": 7, "ts": 1716454218350391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276546, "dur": 7,
    "args": {
      "External id": 55554,
      "cbid": 211, "correlation": 55554
    }
  },
  {
    "ph": "s", "id": 55554, "pid": 76337, "tid": -914061504, "ts": 1716454218276546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218350435, "dur": 28,
    "args": {
      "External id": 55563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55563,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55563, "pid": 5, "tid": 7, "ts": 1716454218350435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276585, "dur": 9,
    "args": {
      "External id": 55563,
      "cbid": 211, "correlation": 55563
    }
  },
  {
    "ph": "s", "id": 55563, "pid": 76337, "tid": -914061504, "ts": 1716454218276585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276636, "dur": 0,
    "args": {
      "External id": 55573,
      "cbid": 317, "correlation": 55573
    }
  },
  {
    "ph": "f", "id": 55573, "pid": 76337, "tid": -914061504, "ts": 1716454218276636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276637, "dur": 0,
    "args": {
      "External id": 55574,
      "cbid": 203, "correlation": 55574
    }
  },
  {
    "ph": "f", "id": 55574, "pid": 76337, "tid": -914061504, "ts": 1716454218276637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276638, "dur": 0,
    "args": {
      "External id": 55575,
      "cbid": 205, "correlation": 55575
    }
  },
  {
    "ph": "f", "id": 55575, "pid": 76337, "tid": -914061504, "ts": 1716454218276638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218350465, "dur": 32,
    "args": {
      "External id": 55579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55579,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55579, "pid": 5, "tid": 7, "ts": 1716454218350465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276652, "dur": 11,
    "args": {
      "External id": 55579,
      "cbid": 211, "correlation": 55579
    }
  },
  {
    "ph": "s", "id": 55579, "pid": 76337, "tid": -914061504, "ts": 1716454218276652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218350499, "dur": 63,
    "args": {
      "External id": 55581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55581,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55581, "pid": 5, "tid": 7, "ts": 1716454218350499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276666, "dur": 6,
    "args": {
      "External id": 55581,
      "cbid": 211, "correlation": 55581
    }
  },
  {
    "ph": "s", "id": 55581, "pid": 76337, "tid": -914061504, "ts": 1716454218276666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218350563, "dur": 963,
    "args": {
      "External id": 55583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55583,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55583, "pid": 5, "tid": 7, "ts": 1716454218350563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276678, "dur": 9,
    "args": {
      "External id": 55583,
      "cbid": 211, "correlation": 55583
    }
  },
  {
    "ph": "s", "id": 55583, "pid": 76337, "tid": -914061504, "ts": 1716454218276678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218351527, "dur": 20,
    "args": {
      "External id": 55585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55585, "pid": 5, "tid": 7, "ts": 1716454218351527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276691, "dur": 5,
    "args": {
      "External id": 55585,
      "cbid": 211, "correlation": 55585
    }
  },
  {
    "ph": "s", "id": 55585, "pid": 76337, "tid": -914061504, "ts": 1716454218276691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218351549, "dur": 33,
    "args": {
      "External id": 55591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55591,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55591, "pid": 5, "tid": 7, "ts": 1716454218351549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276719, "dur": 8,
    "args": {
      "External id": 55591,
      "cbid": 211, "correlation": 55591
    }
  },
  {
    "ph": "s", "id": 55591, "pid": 76337, "tid": -914061504, "ts": 1716454218276719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218351583, "dur": 3,
    "args": {
      "External id": 55599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55599,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 55599, "pid": 5, "tid": 7, "ts": 1716454218351583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276762, "dur": 10,
    "args": {
      "External id": 55599,
      "cbid": 211, "correlation": 55599
    }
  },
  {
    "ph": "s", "id": 55599, "pid": 76337, "tid": -914061504, "ts": 1716454218276762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276829, "dur": 1,
    "args": {
      "External id": 55615,
      "cbid": 251, "correlation": 55615
    }
  },
  {
    "ph": "f", "id": 55615, "pid": 76337, "tid": -914061504, "ts": 1716454218276829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276834, "dur": 0,
    "args": {
      "External id": 55617,
      "cbid": 251, "correlation": 55617
    }
  },
  {
    "ph": "f", "id": 55617, "pid": 76337, "tid": -914061504, "ts": 1716454218276834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218351587, "dur": 12,
    "args": {
      "External id": 55618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55618,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 55618, "pid": 5, "tid": 7, "ts": 1716454218351587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276836, "dur": 11,
    "args": {
      "External id": 55618,
      "cbid": 211, "correlation": 55618
    }
  },
  {
    "ph": "s", "id": 55618, "pid": 76337, "tid": -914061504, "ts": 1716454218276836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218351601, "dur": 5,
    "args": {
      "External id": 55620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55620,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 55620, "pid": 5, "tid": 7, "ts": 1716454218351601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276849, "dur": 5,
    "args": {
      "External id": 55620,
      "cbid": 211, "correlation": 55620
    }
  },
  {
    "ph": "s", "id": 55620, "pid": 76337, "tid": -914061504, "ts": 1716454218276849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218351607, "dur": 29,
    "args": {
      "External id": 55630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55630,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55630, "pid": 5, "tid": 7, "ts": 1716454218351607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276906, "dur": 12,
    "args": {
      "External id": 55630,
      "cbid": 211, "correlation": 55630
    }
  },
  {
    "ph": "s", "id": 55630, "pid": 76337, "tid": -914061504, "ts": 1716454218276906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218351638, "dur": 31,
    "args": {
      "External id": 55650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55650,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 55650, "pid": 5, "tid": 7, "ts": 1716454218351638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218276972, "dur": 19,
    "args": {
      "External id": 55650,
      "cbid": 211, "correlation": 55650
    }
  },
  {
    "ph": "s", "id": 55650, "pid": 76337, "tid": -914061504, "ts": 1716454218276972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218351670, "dur": 4,
    "args": {
      "External id": 55662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55662,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 55662, "pid": 5, "tid": 7, "ts": 1716454218351670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277004, "dur": 7,
    "args": {
      "External id": 55662,
      "cbid": 211, "correlation": 55662
    }
  },
  {
    "ph": "s", "id": 55662, "pid": 76337, "tid": -914061504, "ts": 1716454218277004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218351675, "dur": 30,
    "args": {
      "External id": 55665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55665, "pid": 5, "tid": 7, "ts": 1716454218351675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277023, "dur": 7,
    "args": {
      "External id": 55665,
      "cbid": 211, "correlation": 55665
    }
  },
  {
    "ph": "s", "id": 55665, "pid": 76337, "tid": -914061504, "ts": 1716454218277023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218351707, "dur": 20,
    "args": {
      "External id": 55674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55674,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55674, "pid": 5, "tid": 7, "ts": 1716454218351707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277065, "dur": 10,
    "args": {
      "External id": 55674,
      "cbid": 211, "correlation": 55674
    }
  },
  {
    "ph": "s", "id": 55674, "pid": 76337, "tid": -914061504, "ts": 1716454218277065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277128, "dur": 0,
    "args": {
      "External id": 55684,
      "cbid": 317, "correlation": 55684
    }
  },
  {
    "ph": "f", "id": 55684, "pid": 76337, "tid": -914061504, "ts": 1716454218277128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277129, "dur": 0,
    "args": {
      "External id": 55685,
      "cbid": 203, "correlation": 55685
    }
  },
  {
    "ph": "f", "id": 55685, "pid": 76337, "tid": -914061504, "ts": 1716454218277129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277130, "dur": 0,
    "args": {
      "External id": 55686,
      "cbid": 205, "correlation": 55686
    }
  },
  {
    "ph": "f", "id": 55686, "pid": 76337, "tid": -914061504, "ts": 1716454218277130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218351728, "dur": 22,
    "args": {
      "External id": 55690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55690,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55690, "pid": 5, "tid": 7, "ts": 1716454218351728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277145, "dur": 12,
    "args": {
      "External id": 55690,
      "cbid": 211, "correlation": 55690
    }
  },
  {
    "ph": "s", "id": 55690, "pid": 76337, "tid": -914061504, "ts": 1716454218277145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218351752, "dur": 44,
    "args": {
      "External id": 55692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55692, "pid": 5, "tid": 7, "ts": 1716454218351752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277159, "dur": 5,
    "args": {
      "External id": 55692,
      "cbid": 211, "correlation": 55692
    }
  },
  {
    "ph": "s", "id": 55692, "pid": 76337, "tid": -914061504, "ts": 1716454218277159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218351797, "dur": 641,
    "args": {
      "External id": 55694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55694,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55694, "pid": 5, "tid": 7, "ts": 1716454218351797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277171, "dur": 6,
    "args": {
      "External id": 55694,
      "cbid": 211, "correlation": 55694
    }
  },
  {
    "ph": "s", "id": 55694, "pid": 76337, "tid": -914061504, "ts": 1716454218277171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218352440, "dur": 21,
    "args": {
      "External id": 55696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55696,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55696, "pid": 5, "tid": 7, "ts": 1716454218352440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277180, "dur": 5,
    "args": {
      "External id": 55696,
      "cbid": 211, "correlation": 55696
    }
  },
  {
    "ph": "s", "id": 55696, "pid": 76337, "tid": -914061504, "ts": 1716454218277180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218352462, "dur": 33,
    "args": {
      "External id": 55702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55702, "pid": 5, "tid": 7, "ts": 1716454218352462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277209, "dur": 8,
    "args": {
      "External id": 55702,
      "cbid": 211, "correlation": 55702
    }
  },
  {
    "ph": "s", "id": 55702, "pid": 76337, "tid": -914061504, "ts": 1716454218277209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277267, "dur": 0,
    "args": {
      "External id": 55712,
      "cbid": 317, "correlation": 55712
    }
  },
  {
    "ph": "f", "id": 55712, "pid": 76337, "tid": -914061504, "ts": 1716454218277267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277267, "dur": 0,
    "args": {
      "External id": 55713,
      "cbid": 203, "correlation": 55713
    }
  },
  {
    "ph": "f", "id": 55713, "pid": 76337, "tid": -914061504, "ts": 1716454218277267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277268, "dur": 0,
    "args": {
      "External id": 55714,
      "cbid": 205, "correlation": 55714
    }
  },
  {
    "ph": "f", "id": 55714, "pid": 76337, "tid": -914061504, "ts": 1716454218277268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218352496, "dur": 30,
    "args": {
      "External id": 55718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55718,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55718, "pid": 5, "tid": 7, "ts": 1716454218352496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277283, "dur": 15,
    "args": {
      "External id": 55718,
      "cbid": 211, "correlation": 55718
    }
  },
  {
    "ph": "s", "id": 55718, "pid": 76337, "tid": -914061504, "ts": 1716454218277283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218352527, "dur": 151,
    "args": {
      "External id": 55720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55720,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55720, "pid": 5, "tid": 7, "ts": 1716454218352527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277304, "dur": 6,
    "args": {
      "External id": 55720,
      "cbid": 211, "correlation": 55720
    }
  },
  {
    "ph": "s", "id": 55720, "pid": 76337, "tid": -914061504, "ts": 1716454218277304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218352679, "dur": 22,
    "args": {
      "External id": 55722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55722,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55722, "pid": 5, "tid": 7, "ts": 1716454218352679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277314, "dur": 5,
    "args": {
      "External id": 55722,
      "cbid": 211, "correlation": 55722
    }
  },
  {
    "ph": "s", "id": 55722, "pid": 76337, "tid": -914061504, "ts": 1716454218277314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218352703, "dur": 32,
    "args": {
      "External id": 55728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55728,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55728, "pid": 5, "tid": 7, "ts": 1716454218352703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277340, "dur": 9,
    "args": {
      "External id": 55728,
      "cbid": 211, "correlation": 55728
    }
  },
  {
    "ph": "s", "id": 55728, "pid": 76337, "tid": -914061504, "ts": 1716454218277340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218352736, "dur": 27,
    "args": {
      "External id": 55736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55736,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55736, "pid": 5, "tid": 7, "ts": 1716454218352736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277369, "dur": 8,
    "args": {
      "External id": 55736,
      "cbid": 211, "correlation": 55736
    }
  },
  {
    "ph": "s", "id": 55736, "pid": 76337, "tid": -914061504, "ts": 1716454218277369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218352765, "dur": 19,
    "args": {
      "External id": 55744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55744,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55744, "pid": 5, "tid": 7, "ts": 1716454218352765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277398, "dur": 8,
    "args": {
      "External id": 55744,
      "cbid": 211, "correlation": 55744
    }
  },
  {
    "ph": "s", "id": 55744, "pid": 76337, "tid": -914061504, "ts": 1716454218277398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218352786, "dur": 30,
    "args": {
      "External id": 55764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55764,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 55764, "pid": 5, "tid": 7, "ts": 1716454218352786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277481, "dur": 13,
    "args": {
      "External id": 55764,
      "cbid": 211, "correlation": 55764
    }
  },
  {
    "ph": "s", "id": 55764, "pid": 76337, "tid": -914061504, "ts": 1716454218277481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218352817, "dur": 4,
    "args": {
      "External id": 55776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55776,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 55776, "pid": 5, "tid": 7, "ts": 1716454218352817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277504, "dur": 6,
    "args": {
      "External id": 55776,
      "cbid": 211, "correlation": 55776
    }
  },
  {
    "ph": "s", "id": 55776, "pid": 76337, "tid": -914061504, "ts": 1716454218277504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218352823, "dur": 31,
    "args": {
      "External id": 55779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55779, "pid": 5, "tid": 7, "ts": 1716454218352823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277521, "dur": 6,
    "args": {
      "External id": 55779,
      "cbid": 211, "correlation": 55779
    }
  },
  {
    "ph": "s", "id": 55779, "pid": 76337, "tid": -914061504, "ts": 1716454218277521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277581, "dur": 0,
    "args": {
      "External id": 55790,
      "cbid": 317, "correlation": 55790
    }
  },
  {
    "ph": "f", "id": 55790, "pid": 76337, "tid": -914061504, "ts": 1716454218277581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277582, "dur": 0,
    "args": {
      "External id": 55791,
      "cbid": 203, "correlation": 55791
    }
  },
  {
    "ph": "f", "id": 55791, "pid": 76337, "tid": -914061504, "ts": 1716454218277582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277583, "dur": 0,
    "args": {
      "External id": 55792,
      "cbid": 205, "correlation": 55792
    }
  },
  {
    "ph": "f", "id": 55792, "pid": 76337, "tid": -914061504, "ts": 1716454218277583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218352855, "dur": 22,
    "args": {
      "External id": 55796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55796, "pid": 5, "tid": 7, "ts": 1716454218352855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277595, "dur": 12,
    "args": {
      "External id": 55796,
      "cbid": 211, "correlation": 55796
    }
  },
  {
    "ph": "s", "id": 55796, "pid": 76337, "tid": -914061504, "ts": 1716454218277595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218352878, "dur": 104,
    "args": {
      "External id": 55798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55798,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55798, "pid": 5, "tid": 7, "ts": 1716454218352878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277614, "dur": 6,
    "args": {
      "External id": 55798,
      "cbid": 211, "correlation": 55798
    }
  },
  {
    "ph": "s", "id": 55798, "pid": 76337, "tid": -914061504, "ts": 1716454218277614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218352983, "dur": 23,
    "args": {
      "External id": 55800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55800, "pid": 5, "tid": 7, "ts": 1716454218352983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277623, "dur": 5,
    "args": {
      "External id": 55800,
      "cbid": 211, "correlation": 55800
    }
  },
  {
    "ph": "s", "id": 55800, "pid": 76337, "tid": -914061504, "ts": 1716454218277623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218353008, "dur": 32,
    "args": {
      "External id": 55806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55806,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55806, "pid": 5, "tid": 7, "ts": 1716454218353008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277650, "dur": 8,
    "args": {
      "External id": 55806,
      "cbid": 211, "correlation": 55806
    }
  },
  {
    "ph": "s", "id": 55806, "pid": 76337, "tid": -914061504, "ts": 1716454218277650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218353041, "dur": 194,
    "args": {
      "External id": 55815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55815,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55815, "pid": 5, "tid": 7, "ts": 1716454218353041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277733, "dur": 14,
    "args": {
      "External id": 55815,
      "cbid": 211, "correlation": 55815
    }
  },
  {
    "ph": "s", "id": 55815, "pid": 76337, "tid": -914061504, "ts": 1716454218277733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218353236, "dur": 64,
    "args": {
      "External id": 55837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55837,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55837, "pid": 5, "tid": 7, "ts": 1716454218353236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277789, "dur": 10,
    "args": {
      "External id": 55837,
      "cbid": 211, "correlation": 55837
    }
  },
  {
    "ph": "s", "id": 55837, "pid": 76337, "tid": -914061504, "ts": 1716454218277789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277876, "dur": 2,
    "args": {
      "External id": 55848,
      "cbid": 251, "correlation": 55848
    }
  },
  {
    "ph": "f", "id": 55848, "pid": 76337, "tid": -914061504, "ts": 1716454218277876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218353302, "dur": 151,
    "args": {
      "External id": 55849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55849,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55849, "pid": 5, "tid": 7, "ts": 1716454218353302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277882, "dur": 13,
    "args": {
      "External id": 55849,
      "cbid": 211, "correlation": 55849
    }
  },
  {
    "ph": "s", "id": 55849, "pid": 76337, "tid": -914061504, "ts": 1716454218277882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277956, "dur": 1,
    "args": {
      "External id": 55860,
      "cbid": 251, "correlation": 55860
    }
  },
  {
    "ph": "f", "id": 55860, "pid": 76337, "tid": -914061504, "ts": 1716454218277956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218353454, "dur": 141,
    "args": {
      "External id": 55861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55861,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55861, "pid": 5, "tid": 7, "ts": 1716454218353454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218277960, "dur": 12,
    "args": {
      "External id": 55861,
      "cbid": 211, "correlation": 55861
    }
  },
  {
    "ph": "s", "id": 55861, "pid": 76337, "tid": -914061504, "ts": 1716454218277960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278036, "dur": 1,
    "args": {
      "External id": 55872,
      "cbid": 251, "correlation": 55872
    }
  },
  {
    "ph": "f", "id": 55872, "pid": 76337, "tid": -914061504, "ts": 1716454218278036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218353597, "dur": 142,
    "args": {
      "External id": 55873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55873,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55873, "pid": 5, "tid": 7, "ts": 1716454218353597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278041, "dur": 12,
    "args": {
      "External id": 55873,
      "cbid": 211, "correlation": 55873
    }
  },
  {
    "ph": "s", "id": 55873, "pid": 76337, "tid": -914061504, "ts": 1716454218278041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218353740, "dur": 1934,
    "args": {
      "External id": 55894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55894,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 55894, "pid": 5, "tid": 7, "ts": 1716454218353740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278123, "dur": 13,
    "args": {
      "External id": 55894,
      "cbid": 211, "correlation": 55894
    }
  },
  {
    "ph": "s", "id": 55894, "pid": 76337, "tid": -914061504, "ts": 1716454218278123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278221, "dur": 1,
    "args": {
      "External id": 55912,
      "cbid": 251, "correlation": 55912
    }
  },
  {
    "ph": "f", "id": 55912, "pid": 76337, "tid": -914061504, "ts": 1716454218278221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218355675, "dur": 149,
    "args": {
      "External id": 55914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55914,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 55914, "pid": 5, "tid": 7, "ts": 1716454218355675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278227, "dur": 13,
    "args": {
      "External id": 55914,
      "cbid": 211, "correlation": 55914
    }
  },
  {
    "ph": "s", "id": 55914, "pid": 76337, "tid": -914061504, "ts": 1716454218278227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218355825, "dur": 35,
    "args": {
      "External id": 55922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55922,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55922, "pid": 5, "tid": 7, "ts": 1716454218355825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278297, "dur": 12,
    "args": {
      "External id": 55922,
      "cbid": 211, "correlation": 55922
    }
  },
  {
    "ph": "s", "id": 55922, "pid": 76337, "tid": -914061504, "ts": 1716454218278297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218355862, "dur": 50,
    "args": {
      "External id": 55930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55930,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55930, "pid": 5, "tid": 7, "ts": 1716454218355862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278335, "dur": 8,
    "args": {
      "External id": 55930,
      "cbid": 211, "correlation": 55930
    }
  },
  {
    "ph": "s", "id": 55930, "pid": 76337, "tid": -914061504, "ts": 1716454218278335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218355913, "dur": 31,
    "args": {
      "External id": 55941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55941, "pid": 5, "tid": 7, "ts": 1716454218355913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278411, "dur": 13,
    "args": {
      "External id": 55941,
      "cbid": 211, "correlation": 55941
    }
  },
  {
    "ph": "s", "id": 55941, "pid": 76337, "tid": -914061504, "ts": 1716454218278411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218355946, "dur": 34,
    "args": {
      "External id": 55963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55963,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 55963, "pid": 5, "tid": 7, "ts": 1716454218355946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278441, "dur": 8,
    "args": {
      "External id": 55963,
      "cbid": 211, "correlation": 55963
    }
  },
  {
    "ph": "s", "id": 55963, "pid": 76337, "tid": -914061504, "ts": 1716454218278441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278527, "dur": 1,
    "args": {
      "External id": 55974,
      "cbid": 251, "correlation": 55974
    }
  },
  {
    "ph": "f", "id": 55974, "pid": 76337, "tid": -914061504, "ts": 1716454218278527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218355981, "dur": 90,
    "args": {
      "External id": 55975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55975,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 55975, "pid": 5, "tid": 7, "ts": 1716454218355981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278532, "dur": 12,
    "args": {
      "External id": 55975,
      "cbid": 211, "correlation": 55975
    }
  },
  {
    "ph": "s", "id": 55975, "pid": 76337, "tid": -914061504, "ts": 1716454218278532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278600, "dur": 1,
    "args": {
      "External id": 55986,
      "cbid": 251, "correlation": 55986
    }
  },
  {
    "ph": "f", "id": 55986, "pid": 76337, "tid": -914061504, "ts": 1716454218278600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278604, "dur": 0,
    "args": {
      "External id": 55987,
      "cbid": 251, "correlation": 55987
    }
  },
  {
    "ph": "f", "id": 55987, "pid": 76337, "tid": -914061504, "ts": 1716454218278604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218356073, "dur": 11,
    "args": {
      "External id": 55988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55988,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 55988, "pid": 5, "tid": 7, "ts": 1716454218356073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278606, "dur": 12,
    "args": {
      "External id": 55988,
      "cbid": 211, "correlation": 55988
    }
  },
  {
    "ph": "s", "id": 55988, "pid": 76337, "tid": -914061504, "ts": 1716454218278606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218356085, "dur": 6,
    "args": {
      "External id": 55990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 55990,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 55990, "pid": 5, "tid": 7, "ts": 1716454218356085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278620, "dur": 6,
    "args": {
      "External id": 55990,
      "cbid": 211, "correlation": 55990
    }
  },
  {
    "ph": "s", "id": 55990, "pid": 76337, "tid": -914061504, "ts": 1716454218278620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278677, "dur": 1,
    "args": {
      "External id": 56001,
      "cbid": 251, "correlation": 56001
    }
  },
  {
    "ph": "f", "id": 56001, "pid": 76337, "tid": -914061504, "ts": 1716454218278677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278681, "dur": 0,
    "args": {
      "External id": 56002,
      "cbid": 251, "correlation": 56002
    }
  },
  {
    "ph": "f", "id": 56002, "pid": 76337, "tid": -914061504, "ts": 1716454218278681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218356092, "dur": 7,
    "args": {
      "External id": 56003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56003,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 56003, "pid": 5, "tid": 7, "ts": 1716454218356092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278682, "dur": 12,
    "args": {
      "External id": 56003,
      "cbid": 211, "correlation": 56003
    }
  },
  {
    "ph": "s", "id": 56003, "pid": 76337, "tid": -914061504, "ts": 1716454218278682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218356100, "dur": 4,
    "args": {
      "External id": 56005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56005,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 56005, "pid": 5, "tid": 7, "ts": 1716454218356100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278695, "dur": 8,
    "args": {
      "External id": 56005,
      "cbid": 211, "correlation": 56005
    }
  },
  {
    "ph": "s", "id": 56005, "pid": 76337, "tid": -914061504, "ts": 1716454218278695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218356105, "dur": 92,
    "args": {
      "External id": 56026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56026,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 56026, "pid": 5, "tid": 7, "ts": 1716454218356105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278772, "dur": 13,
    "args": {
      "External id": 56026,
      "cbid": 211, "correlation": 56026
    }
  },
  {
    "ph": "s", "id": 56026, "pid": 76337, "tid": -914061504, "ts": 1716454218278772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278868, "dur": 1,
    "args": {
      "External id": 56044,
      "cbid": 251, "correlation": 56044
    }
  },
  {
    "ph": "f", "id": 56044, "pid": 76337, "tid": -914061504, "ts": 1716454218278868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218356198, "dur": 99,
    "args": {
      "External id": 56046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56046,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56046, "pid": 5, "tid": 7, "ts": 1716454218356198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278875, "dur": 13,
    "args": {
      "External id": 56046,
      "cbid": 211, "correlation": 56046
    }
  },
  {
    "ph": "s", "id": 56046, "pid": 76337, "tid": -914061504, "ts": 1716454218278875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218356299, "dur": 19,
    "args": {
      "External id": 56054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56054,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56054, "pid": 5, "tid": 7, "ts": 1716454218356299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278942, "dur": 12,
    "args": {
      "External id": 56054,
      "cbid": 211, "correlation": 56054
    }
  },
  {
    "ph": "s", "id": 56054, "pid": 76337, "tid": -914061504, "ts": 1716454218278942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218356319, "dur": 37,
    "args": {
      "External id": 56062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56062, "pid": 5, "tid": 7, "ts": 1716454218356319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218278991, "dur": 10,
    "args": {
      "External id": 56062,
      "cbid": 211, "correlation": 56062
    }
  },
  {
    "ph": "s", "id": 56062, "pid": 76337, "tid": -914061504, "ts": 1716454218278991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218356357, "dur": 34,
    "args": {
      "External id": 56084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56084,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56084, "pid": 5, "tid": 7, "ts": 1716454218356357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279044, "dur": 13,
    "args": {
      "External id": 56084,
      "cbid": 211, "correlation": 56084
    }
  },
  {
    "ph": "s", "id": 56084, "pid": 76337, "tid": -914061504, "ts": 1716454218279044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279137, "dur": 1,
    "args": {
      "External id": 56100,
      "cbid": 251, "correlation": 56100
    }
  },
  {
    "ph": "f", "id": 56100, "pid": 76337, "tid": -914061504, "ts": 1716454218279137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279142, "dur": 0,
    "args": {
      "External id": 56102,
      "cbid": 251, "correlation": 56102
    }
  },
  {
    "ph": "f", "id": 56102, "pid": 76337, "tid": -914061504, "ts": 1716454218279142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218356393, "dur": 544,
    "args": {
      "External id": 56103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56103,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 56103, "pid": 5, "tid": 7, "ts": 1716454218356393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279145, "dur": 13,
    "args": {
      "External id": 56103,
      "cbid": 211, "correlation": 56103
    }
  },
  {
    "ph": "s", "id": 56103, "pid": 76337, "tid": -914061504, "ts": 1716454218279145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218356938, "dur": 126,
    "args": {
      "External id": 56111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56111,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56111, "pid": 5, "tid": 7, "ts": 1716454218356938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279211, "dur": 12,
    "args": {
      "External id": 56111,
      "cbid": 211, "correlation": 56111
    }
  },
  {
    "ph": "s", "id": 56111, "pid": 76337, "tid": -914061504, "ts": 1716454218279211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218357065, "dur": 131,
    "args": {
      "External id": 56119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56119, "pid": 5, "tid": 7, "ts": 1716454218357065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279241, "dur": 8,
    "args": {
      "External id": 56119,
      "cbid": 211, "correlation": 56119
    }
  },
  {
    "ph": "s", "id": 56119, "pid": 76337, "tid": -914061504, "ts": 1716454218279241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279317, "dur": 1,
    "args": {
      "External id": 56135,
      "cbid": 251, "correlation": 56135
    }
  },
  {
    "ph": "f", "id": 56135, "pid": 76337, "tid": -914061504, "ts": 1716454218279317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218357197, "dur": 305,
    "args": {
      "External id": 56137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56137,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56137, "pid": 5, "tid": 7, "ts": 1716454218357197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279323, "dur": 13,
    "args": {
      "External id": 56137,
      "cbid": 211, "correlation": 56137
    }
  },
  {
    "ph": "s", "id": 56137, "pid": 76337, "tid": -914061504, "ts": 1716454218279323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218357504, "dur": 28,
    "args": {
      "External id": 56145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56145,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56145, "pid": 5, "tid": 7, "ts": 1716454218357504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279365, "dur": 9,
    "args": {
      "External id": 56145,
      "cbid": 211, "correlation": 56145
    }
  },
  {
    "ph": "s", "id": 56145, "pid": 76337, "tid": -914061504, "ts": 1716454218279365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218357533, "dur": 81,
    "args": {
      "External id": 56156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56156,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56156, "pid": 5, "tid": 7, "ts": 1716454218357533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279432, "dur": 13,
    "args": {
      "External id": 56156,
      "cbid": 211, "correlation": 56156
    }
  },
  {
    "ph": "s", "id": 56156, "pid": 76337, "tid": -914061504, "ts": 1716454218279432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279500, "dur": 0,
    "args": {
      "External id": 56168,
      "cbid": 317, "correlation": 56168
    }
  },
  {
    "ph": "f", "id": 56168, "pid": 76337, "tid": -914061504, "ts": 1716454218279500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279501, "dur": 0,
    "args": {
      "External id": 56169,
      "cbid": 203, "correlation": 56169
    }
  },
  {
    "ph": "f", "id": 56169, "pid": 76337, "tid": -914061504, "ts": 1716454218279501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279502, "dur": 0,
    "args": {
      "External id": 56170,
      "cbid": 205, "correlation": 56170
    }
  },
  {
    "ph": "f", "id": 56170, "pid": 76337, "tid": -914061504, "ts": 1716454218279502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218357615, "dur": 23,
    "args": {
      "External id": 56174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56174,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56174, "pid": 5, "tid": 7, "ts": 1716454218357615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279517, "dur": 12,
    "args": {
      "External id": 56174,
      "cbid": 211, "correlation": 56174
    }
  },
  {
    "ph": "s", "id": 56174, "pid": 76337, "tid": -914061504, "ts": 1716454218279517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218357640, "dur": 121,
    "args": {
      "External id": 56176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56176,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56176, "pid": 5, "tid": 7, "ts": 1716454218357640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279535, "dur": 7,
    "args": {
      "External id": 56176,
      "cbid": 211, "correlation": 56176
    }
  },
  {
    "ph": "s", "id": 56176, "pid": 76337, "tid": -914061504, "ts": 1716454218279535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218357762, "dur": 24,
    "args": {
      "External id": 56178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56178, "pid": 5, "tid": 7, "ts": 1716454218357762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279546, "dur": 5,
    "args": {
      "External id": 56178,
      "cbid": 211, "correlation": 56178
    }
  },
  {
    "ph": "s", "id": 56178, "pid": 76337, "tid": -914061504, "ts": 1716454218279546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218357787, "dur": 33,
    "args": {
      "External id": 56184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56184, "pid": 5, "tid": 7, "ts": 1716454218357787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279573, "dur": 8,
    "args": {
      "External id": 56184,
      "cbid": 211, "correlation": 56184
    }
  },
  {
    "ph": "s", "id": 56184, "pid": 76337, "tid": -914061504, "ts": 1716454218279573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218357821, "dur": 26,
    "args": {
      "External id": 56192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56192,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56192, "pid": 5, "tid": 7, "ts": 1716454218357821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279605, "dur": 8,
    "args": {
      "External id": 56192,
      "cbid": 211, "correlation": 56192
    }
  },
  {
    "ph": "s", "id": 56192, "pid": 76337, "tid": -914061504, "ts": 1716454218279605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218357849, "dur": 102,
    "args": {
      "External id": 56203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56203,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56203, "pid": 5, "tid": 7, "ts": 1716454218357849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279666, "dur": 12,
    "args": {
      "External id": 56203,
      "cbid": 211, "correlation": 56203
    }
  },
  {
    "ph": "s", "id": 56203, "pid": 76337, "tid": -914061504, "ts": 1716454218279666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279721, "dur": 0,
    "args": {
      "External id": 56213,
      "cbid": 317, "correlation": 56213
    }
  },
  {
    "ph": "f", "id": 56213, "pid": 76337, "tid": -914061504, "ts": 1716454218279721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279722, "dur": 0,
    "args": {
      "External id": 56214,
      "cbid": 203, "correlation": 56214
    }
  },
  {
    "ph": "f", "id": 56214, "pid": 76337, "tid": -914061504, "ts": 1716454218279722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279723, "dur": 0,
    "args": {
      "External id": 56215,
      "cbid": 205, "correlation": 56215
    }
  },
  {
    "ph": "f", "id": 56215, "pid": 76337, "tid": -914061504, "ts": 1716454218279723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218357952, "dur": 75,
    "args": {
      "External id": 56219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56219,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56219, "pid": 5, "tid": 7, "ts": 1716454218357952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279736, "dur": 11,
    "args": {
      "External id": 56219,
      "cbid": 211, "correlation": 56219
    }
  },
  {
    "ph": "s", "id": 56219, "pid": 76337, "tid": -914061504, "ts": 1716454218279736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218358029, "dur": 44,
    "args": {
      "External id": 56221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56221,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56221, "pid": 5, "tid": 7, "ts": 1716454218358029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279751, "dur": 5,
    "args": {
      "External id": 56221,
      "cbid": 211, "correlation": 56221
    }
  },
  {
    "ph": "s", "id": 56221, "pid": 76337, "tid": -914061504, "ts": 1716454218279751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218358074, "dur": 4,
    "args": {
      "External id": 56223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56223, "pid": 5, "tid": 7, "ts": 1716454218358074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279761, "dur": 9,
    "args": {
      "External id": 56223,
      "cbid": 211, "correlation": 56223
    }
  },
  {
    "ph": "s", "id": 56223, "pid": 76337, "tid": -914061504, "ts": 1716454218279761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279774, "dur": 0,
    "args": {
      "External id": 56224,
      "cbid": 51, "correlation": 56224
    }
  },
  {
    "ph": "s", "id": 56224, "pid": 76337, "tid": -914061504, "ts": 1716454218279774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218358079, "dur": 2224,
    "args": {
      "External id": 56225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56225,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56225, "pid": 5, "tid": 7, "ts": 1716454218358079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279775, "dur": 6,
    "args": {
      "External id": 56225,
      "cbid": 211, "correlation": 56225
    }
  },
  {
    "ph": "s", "id": 56225, "pid": 76337, "tid": -914061504, "ts": 1716454218279775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218360305, "dur": 113,
    "args": {
      "External id": 56230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56230, "pid": 5, "tid": 7, "ts": 1716454218360305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279803, "dur": 9,
    "args": {
      "External id": 56230,
      "cbid": 211, "correlation": 56230
    }
  },
  {
    "ph": "s", "id": 56230, "pid": 76337, "tid": -914061504, "ts": 1716454218279803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218360419, "dur": 165,
    "args": {
      "External id": 56239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56239,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56239, "pid": 5, "tid": 7, "ts": 1716454218360419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279896, "dur": 13,
    "args": {
      "External id": 56239,
      "cbid": 211, "correlation": 56239
    }
  },
  {
    "ph": "s", "id": 56239, "pid": 76337, "tid": -914061504, "ts": 1716454218279896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218360585, "dur": 130,
    "args": {
      "External id": 56259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56259,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 56259, "pid": 5, "tid": 7, "ts": 1716454218360585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279967, "dur": 19,
    "args": {
      "External id": 56259,
      "cbid": 211, "correlation": 56259
    }
  },
  {
    "ph": "s", "id": 56259, "pid": 76337, "tid": -914061504, "ts": 1716454218279967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218360717, "dur": 4,
    "args": {
      "External id": 56271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56271,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 56271, "pid": 5, "tid": 7, "ts": 1716454218360717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218279996, "dur": 7,
    "args": {
      "External id": 56271,
      "cbid": 211, "correlation": 56271
    }
  },
  {
    "ph": "s", "id": 56271, "pid": 76337, "tid": -914061504, "ts": 1716454218279996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218360722, "dur": 159,
    "args": {
      "External id": 56274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56274,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56274, "pid": 5, "tid": 7, "ts": 1716454218360722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280015, "dur": 7,
    "args": {
      "External id": 56274,
      "cbid": 211, "correlation": 56274
    }
  },
  {
    "ph": "s", "id": 56274, "pid": 76337, "tid": -914061504, "ts": 1716454218280015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218360883, "dur": 102,
    "args": {
      "External id": 56283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56283,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56283, "pid": 5, "tid": 7, "ts": 1716454218360883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280056, "dur": 9,
    "args": {
      "External id": 56283,
      "cbid": 211, "correlation": 56283
    }
  },
  {
    "ph": "s", "id": 56283, "pid": 76337, "tid": -914061504, "ts": 1716454218280056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280108, "dur": 0,
    "args": {
      "External id": 56293,
      "cbid": 317, "correlation": 56293
    }
  },
  {
    "ph": "f", "id": 56293, "pid": 76337, "tid": -914061504, "ts": 1716454218280108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280109, "dur": 3,
    "args": {
      "External id": 56294,
      "cbid": 203, "correlation": 56294
    }
  },
  {
    "ph": "f", "id": 56294, "pid": 76337, "tid": -914061504, "ts": 1716454218280109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280113, "dur": 0,
    "args": {
      "External id": 56295,
      "cbid": 205, "correlation": 56295
    }
  },
  {
    "ph": "f", "id": 56295, "pid": 76337, "tid": -914061504, "ts": 1716454218280113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218360986, "dur": 111,
    "args": {
      "External id": 56299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56299,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56299, "pid": 5, "tid": 7, "ts": 1716454218360986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280129, "dur": 12,
    "args": {
      "External id": 56299,
      "cbid": 211, "correlation": 56299
    }
  },
  {
    "ph": "s", "id": 56299, "pid": 76337, "tid": -914061504, "ts": 1716454218280129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218361098, "dur": 33,
    "args": {
      "External id": 56301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56301,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56301, "pid": 5, "tid": 7, "ts": 1716454218361098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280144, "dur": 5,
    "args": {
      "External id": 56301,
      "cbid": 211, "correlation": 56301
    }
  },
  {
    "ph": "s", "id": 56301, "pid": 76337, "tid": -914061504, "ts": 1716454218280144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218361133, "dur": 3,
    "args": {
      "External id": 56303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56303,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56303, "pid": 5, "tid": 7, "ts": 1716454218361133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280153, "dur": 6,
    "args": {
      "External id": 56303,
      "cbid": 211, "correlation": 56303
    }
  },
  {
    "ph": "s", "id": 56303, "pid": 76337, "tid": -914061504, "ts": 1716454218280153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280162, "dur": 0,
    "args": {
      "External id": 56304,
      "cbid": 51, "correlation": 56304
    }
  },
  {
    "ph": "s", "id": 56304, "pid": 76337, "tid": -914061504, "ts": 1716454218280162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218361138, "dur": 2010,
    "args": {
      "External id": 56305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56305,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56305, "pid": 5, "tid": 7, "ts": 1716454218361138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280163, "dur": 6,
    "args": {
      "External id": 56305,
      "cbid": 211, "correlation": 56305
    }
  },
  {
    "ph": "s", "id": 56305, "pid": 76337, "tid": -914061504, "ts": 1716454218280163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218363150, "dur": 59,
    "args": {
      "External id": 56310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56310,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56310, "pid": 5, "tid": 7, "ts": 1716454218363150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280192, "dur": 8,
    "args": {
      "External id": 56310,
      "cbid": 211, "correlation": 56310
    }
  },
  {
    "ph": "s", "id": 56310, "pid": 76337, "tid": -914061504, "ts": 1716454218280192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218363210, "dur": 3,
    "args": {
      "External id": 56318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56318,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56318, "pid": 5, "tid": 7, "ts": 1716454218363210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280236, "dur": 9,
    "args": {
      "External id": 56318,
      "cbid": 211, "correlation": 56318
    }
  },
  {
    "ph": "s", "id": 56318, "pid": 76337, "tid": -914061504, "ts": 1716454218280236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280303, "dur": 1,
    "args": {
      "External id": 56334,
      "cbid": 251, "correlation": 56334
    }
  },
  {
    "ph": "f", "id": 56334, "pid": 76337, "tid": -914061504, "ts": 1716454218280303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280309, "dur": 0,
    "args": {
      "External id": 56336,
      "cbid": 251, "correlation": 56336
    }
  },
  {
    "ph": "f", "id": 56336, "pid": 76337, "tid": -914061504, "ts": 1716454218280309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218363214, "dur": 11,
    "args": {
      "External id": 56337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56337,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 56337, "pid": 5, "tid": 7, "ts": 1716454218363214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280310, "dur": 11,
    "args": {
      "External id": 56337,
      "cbid": 211, "correlation": 56337
    }
  },
  {
    "ph": "s", "id": 56337, "pid": 76337, "tid": -914061504, "ts": 1716454218280310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218363227, "dur": 5,
    "args": {
      "External id": 56339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56339,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 56339, "pid": 5, "tid": 7, "ts": 1716454218363227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280324, "dur": 5,
    "args": {
      "External id": 56339,
      "cbid": 211, "correlation": 56339
    }
  },
  {
    "ph": "s", "id": 56339, "pid": 76337, "tid": -914061504, "ts": 1716454218280324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218363233, "dur": 53,
    "args": {
      "External id": 56349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56349, "pid": 5, "tid": 7, "ts": 1716454218363233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280382, "dur": 12,
    "args": {
      "External id": 56349,
      "cbid": 211, "correlation": 56349
    }
  },
  {
    "ph": "s", "id": 56349, "pid": 76337, "tid": -914061504, "ts": 1716454218280382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218363288, "dur": 51,
    "args": {
      "External id": 56369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56369,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 56369, "pid": 5, "tid": 7, "ts": 1716454218363288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280453, "dur": 11,
    "args": {
      "External id": 56369,
      "cbid": 211, "correlation": 56369
    }
  },
  {
    "ph": "s", "id": 56369, "pid": 76337, "tid": -914061504, "ts": 1716454218280453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218363340, "dur": 4,
    "args": {
      "External id": 56381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56381,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56381, "pid": 5, "tid": 7, "ts": 1716454218363340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280474, "dur": 6,
    "args": {
      "External id": 56381,
      "cbid": 211, "correlation": 56381
    }
  },
  {
    "ph": "s", "id": 56381, "pid": 76337, "tid": -914061504, "ts": 1716454218280474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218363346, "dur": 55,
    "args": {
      "External id": 56384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56384, "pid": 5, "tid": 7, "ts": 1716454218363346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280493, "dur": 7,
    "args": {
      "External id": 56384,
      "cbid": 211, "correlation": 56384
    }
  },
  {
    "ph": "s", "id": 56384, "pid": 76337, "tid": -914061504, "ts": 1716454218280493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218363402, "dur": 37,
    "args": {
      "External id": 56393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56393,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56393, "pid": 5, "tid": 7, "ts": 1716454218363402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280534, "dur": 10,
    "args": {
      "External id": 56393,
      "cbid": 211, "correlation": 56393
    }
  },
  {
    "ph": "s", "id": 56393, "pid": 76337, "tid": -914061504, "ts": 1716454218280534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280597, "dur": 0,
    "args": {
      "External id": 56403,
      "cbid": 317, "correlation": 56403
    }
  },
  {
    "ph": "f", "id": 56403, "pid": 76337, "tid": -914061504, "ts": 1716454218280597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280598, "dur": 0,
    "args": {
      "External id": 56404,
      "cbid": 203, "correlation": 56404
    }
  },
  {
    "ph": "f", "id": 56404, "pid": 76337, "tid": -914061504, "ts": 1716454218280598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280599, "dur": 0,
    "args": {
      "External id": 56405,
      "cbid": 205, "correlation": 56405
    }
  },
  {
    "ph": "f", "id": 56405, "pid": 76337, "tid": -914061504, "ts": 1716454218280599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218363440, "dur": 41,
    "args": {
      "External id": 56409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56409,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56409, "pid": 5, "tid": 7, "ts": 1716454218363440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280616, "dur": 12,
    "args": {
      "External id": 56409,
      "cbid": 211, "correlation": 56409
    }
  },
  {
    "ph": "s", "id": 56409, "pid": 76337, "tid": -914061504, "ts": 1716454218280616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218363482, "dur": 14,
    "args": {
      "External id": 56411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56411,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56411, "pid": 5, "tid": 7, "ts": 1716454218363482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280630, "dur": 5,
    "args": {
      "External id": 56411,
      "cbid": 211, "correlation": 56411
    }
  },
  {
    "ph": "s", "id": 56411, "pid": 76337, "tid": -914061504, "ts": 1716454218280630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218363498, "dur": 3,
    "args": {
      "External id": 56413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56413, "pid": 5, "tid": 7, "ts": 1716454218363498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280640, "dur": 6,
    "args": {
      "External id": 56413,
      "cbid": 211, "correlation": 56413
    }
  },
  {
    "ph": "s", "id": 56413, "pid": 76337, "tid": -914061504, "ts": 1716454218280640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280648, "dur": 0,
    "args": {
      "External id": 56414,
      "cbid": 51, "correlation": 56414
    }
  },
  {
    "ph": "s", "id": 56414, "pid": 76337, "tid": -914061504, "ts": 1716454218280648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218363503, "dur": 699,
    "args": {
      "External id": 56415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56415,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56415, "pid": 5, "tid": 7, "ts": 1716454218363503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280649, "dur": 6,
    "args": {
      "External id": 56415,
      "cbid": 211, "correlation": 56415
    }
  },
  {
    "ph": "s", "id": 56415, "pid": 76337, "tid": -914061504, "ts": 1716454218280649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218364203, "dur": 60,
    "args": {
      "External id": 56420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56420, "pid": 5, "tid": 7, "ts": 1716454218364203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280677, "dur": 11,
    "args": {
      "External id": 56420,
      "cbid": 211, "correlation": 56420
    }
  },
  {
    "ph": "s", "id": 56420, "pid": 76337, "tid": -914061504, "ts": 1716454218280677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280738, "dur": 0,
    "args": {
      "External id": 56430,
      "cbid": 317, "correlation": 56430
    }
  },
  {
    "ph": "f", "id": 56430, "pid": 76337, "tid": -914061504, "ts": 1716454218280738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280739, "dur": 0,
    "args": {
      "External id": 56431,
      "cbid": 203, "correlation": 56431
    }
  },
  {
    "ph": "f", "id": 56431, "pid": 76337, "tid": -914061504, "ts": 1716454218280739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280740, "dur": 0,
    "args": {
      "External id": 56432,
      "cbid": 205, "correlation": 56432
    }
  },
  {
    "ph": "f", "id": 56432, "pid": 76337, "tid": -914061504, "ts": 1716454218280740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218364264, "dur": 3,
    "args": {
      "External id": 56436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56436, "pid": 5, "tid": 7, "ts": 1716454218364264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280755, "dur": 13,
    "args": {
      "External id": 56436,
      "cbid": 211, "correlation": 56436
    }
  },
  {
    "ph": "s", "id": 56436, "pid": 76337, "tid": -914061504, "ts": 1716454218280755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280772, "dur": 0,
    "args": {
      "External id": 56437,
      "cbid": 51, "correlation": 56437
    }
  },
  {
    "ph": "s", "id": 56437, "pid": 76337, "tid": -914061504, "ts": 1716454218280772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218364269, "dur": 265,
    "args": {
      "External id": 56438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56438,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56438, "pid": 5, "tid": 7, "ts": 1716454218364269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280773, "dur": 7,
    "args": {
      "External id": 56438,
      "cbid": 211, "correlation": 56438
    }
  },
  {
    "ph": "s", "id": 56438, "pid": 76337, "tid": -914061504, "ts": 1716454218280773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218364535, "dur": 59,
    "args": {
      "External id": 56443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56443, "pid": 5, "tid": 7, "ts": 1716454218364535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280801, "dur": 9,
    "args": {
      "External id": 56443,
      "cbid": 211, "correlation": 56443
    }
  },
  {
    "ph": "s", "id": 56443, "pid": 76337, "tid": -914061504, "ts": 1716454218280801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218364596, "dur": 50,
    "args": {
      "External id": 56451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56451,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56451, "pid": 5, "tid": 7, "ts": 1716454218364596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280830, "dur": 8,
    "args": {
      "External id": 56451,
      "cbid": 211, "correlation": 56451
    }
  },
  {
    "ph": "s", "id": 56451, "pid": 76337, "tid": -914061504, "ts": 1716454218280830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218364647, "dur": 35,
    "args": {
      "External id": 56459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56459,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56459, "pid": 5, "tid": 7, "ts": 1716454218364647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280858, "dur": 8,
    "args": {
      "External id": 56459,
      "cbid": 211, "correlation": 56459
    }
  },
  {
    "ph": "s", "id": 56459, "pid": 76337, "tid": -914061504, "ts": 1716454218280858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218364683, "dur": 53,
    "args": {
      "External id": 56479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56479,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 56479, "pid": 5, "tid": 7, "ts": 1716454218364683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280940, "dur": 12,
    "args": {
      "External id": 56479,
      "cbid": 211, "correlation": 56479
    }
  },
  {
    "ph": "s", "id": 56479, "pid": 76337, "tid": -914061504, "ts": 1716454218280940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218364738, "dur": 4,
    "args": {
      "External id": 56491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56491,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56491, "pid": 5, "tid": 7, "ts": 1716454218364738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280961, "dur": 9,
    "args": {
      "External id": 56491,
      "cbid": 211, "correlation": 56491
    }
  },
  {
    "ph": "s", "id": 56491, "pid": 76337, "tid": -914061504, "ts": 1716454218280961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218364743, "dur": 55,
    "args": {
      "External id": 56494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56494,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56494, "pid": 5, "tid": 7, "ts": 1716454218364743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218280990, "dur": 8,
    "args": {
      "External id": 56494,
      "cbid": 211, "correlation": 56494
    }
  },
  {
    "ph": "s", "id": 56494, "pid": 76337, "tid": -914061504, "ts": 1716454218280990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281049, "dur": 0,
    "args": {
      "External id": 56505,
      "cbid": 317, "correlation": 56505
    }
  },
  {
    "ph": "f", "id": 56505, "pid": 76337, "tid": -914061504, "ts": 1716454218281049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281049, "dur": 0,
    "args": {
      "External id": 56506,
      "cbid": 203, "correlation": 56506
    }
  },
  {
    "ph": "f", "id": 56506, "pid": 76337, "tid": -914061504, "ts": 1716454218281049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281050, "dur": 0,
    "args": {
      "External id": 56507,
      "cbid": 205, "correlation": 56507
    }
  },
  {
    "ph": "f", "id": 56507, "pid": 76337, "tid": -914061504, "ts": 1716454218281050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281081, "dur": 2,
    "args": {
      "External id": 56511,
      "cbid": 251, "correlation": 56511
    }
  },
  {
    "ph": "f", "id": 56511, "pid": 76337, "tid": -914061504, "ts": 1716454218281081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281084, "dur": 1,
    "args": {
      "External id": 56512,
      "cbid": 251, "correlation": 56512
    }
  },
  {
    "ph": "f", "id": 56512, "pid": 76337, "tid": -914061504, "ts": 1716454218281084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281085, "dur": 0,
    "args": {
      "External id": 56513,
      "cbid": 251, "correlation": 56513
    }
  },
  {
    "ph": "f", "id": 56513, "pid": 76337, "tid": -914061504, "ts": 1716454218281085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281087, "dur": 1,
    "args": {
      "External id": 56514,
      "cbid": 251, "correlation": 56514
    }
  },
  {
    "ph": "f", "id": 56514, "pid": 76337, "tid": -914061504, "ts": 1716454218281087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281088, "dur": 1,
    "args": {
      "External id": 56515,
      "cbid": 251, "correlation": 56515
    }
  },
  {
    "ph": "f", "id": 56515, "pid": 76337, "tid": -914061504, "ts": 1716454218281088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281090, "dur": 1,
    "args": {
      "External id": 56516,
      "cbid": 251, "correlation": 56516
    }
  },
  {
    "ph": "f", "id": 56516, "pid": 76337, "tid": -914061504, "ts": 1716454218281090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281092, "dur": 1,
    "args": {
      "External id": 56517,
      "cbid": 251, "correlation": 56517
    }
  },
  {
    "ph": "f", "id": 56517, "pid": 76337, "tid": -914061504, "ts": 1716454218281092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281093, "dur": 1,
    "args": {
      "External id": 56518,
      "cbid": 251, "correlation": 56518
    }
  },
  {
    "ph": "f", "id": 56518, "pid": 76337, "tid": -914061504, "ts": 1716454218281093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281096, "dur": 0,
    "args": {
      "External id": 56519,
      "cbid": 251, "correlation": 56519
    }
  },
  {
    "ph": "f", "id": 56519, "pid": 76337, "tid": -914061504, "ts": 1716454218281096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218364800, "dur": 115,
    "args": {
      "External id": 56520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56520,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 56520, "pid": 5, "tid": 7, "ts": 1716454218364800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281098, "dur": 13,
    "args": {
      "External id": 56520,
      "cbid": 211, "correlation": 56520
    }
  },
  {
    "ph": "s", "id": 56520, "pid": 76337, "tid": -914061504, "ts": 1716454218281098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218364916, "dur": 60,
    "args": {
      "External id": 56526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56526, "pid": 5, "tid": 7, "ts": 1716454218364916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281134, "dur": 9,
    "args": {
      "External id": 56526,
      "cbid": 211, "correlation": 56526
    }
  },
  {
    "ph": "s", "id": 56526, "pid": 76337, "tid": -914061504, "ts": 1716454218281134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218364977, "dur": 567,
    "args": {
      "External id": 56535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56535,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56535, "pid": 5, "tid": 7, "ts": 1716454218364977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281217, "dur": 14,
    "args": {
      "External id": 56535,
      "cbid": 211, "correlation": 56535
    }
  },
  {
    "ph": "s", "id": 56535, "pid": 76337, "tid": -914061504, "ts": 1716454218281217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218365545, "dur": 182,
    "args": {
      "External id": 56557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56557,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56557, "pid": 5, "tid": 7, "ts": 1716454218365545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281274, "dur": 11,
    "args": {
      "External id": 56557,
      "cbid": 211, "correlation": 56557
    }
  },
  {
    "ph": "s", "id": 56557, "pid": 76337, "tid": -914061504, "ts": 1716454218281274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281363, "dur": 1,
    "args": {
      "External id": 56568,
      "cbid": 251, "correlation": 56568
    }
  },
  {
    "ph": "f", "id": 56568, "pid": 76337, "tid": -914061504, "ts": 1716454218281363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218365728, "dur": 197,
    "args": {
      "External id": 56569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56569,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56569, "pid": 5, "tid": 7, "ts": 1716454218365728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281368, "dur": 16,
    "args": {
      "External id": 56569,
      "cbid": 211, "correlation": 56569
    }
  },
  {
    "ph": "s", "id": 56569, "pid": 76337, "tid": -914061504, "ts": 1716454218281368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281442, "dur": 1,
    "args": {
      "External id": 56580,
      "cbid": 251, "correlation": 56580
    }
  },
  {
    "ph": "f", "id": 56580, "pid": 76337, "tid": -914061504, "ts": 1716454218281442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218365926, "dur": 191,
    "args": {
      "External id": 56581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56581,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56581, "pid": 5, "tid": 7, "ts": 1716454218365926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281446, "dur": 12,
    "args": {
      "External id": 56581,
      "cbid": 211, "correlation": 56581
    }
  },
  {
    "ph": "s", "id": 56581, "pid": 76337, "tid": -914061504, "ts": 1716454218281446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281509, "dur": 1,
    "args": {
      "External id": 56592,
      "cbid": 251, "correlation": 56592
    }
  },
  {
    "ph": "f", "id": 56592, "pid": 76337, "tid": -914061504, "ts": 1716454218281509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218366119, "dur": 187,
    "args": {
      "External id": 56593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56593,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56593, "pid": 5, "tid": 7, "ts": 1716454218366119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281513, "dur": 12,
    "args": {
      "External id": 56593,
      "cbid": 211, "correlation": 56593
    }
  },
  {
    "ph": "s", "id": 56593, "pid": 76337, "tid": -914061504, "ts": 1716454218281513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218366307, "dur": 18602,
    "args": {
      "External id": 56614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56614,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 56614, "pid": 5, "tid": 7, "ts": 1716454218366307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281596, "dur": 14,
    "args": {
      "External id": 56614,
      "cbid": 211, "correlation": 56614
    }
  },
  {
    "ph": "s", "id": 56614, "pid": 76337, "tid": -914061504, "ts": 1716454218281596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281695, "dur": 1,
    "args": {
      "External id": 56632,
      "cbid": 251, "correlation": 56632
    }
  },
  {
    "ph": "f", "id": 56632, "pid": 76337, "tid": -914061504, "ts": 1716454218281695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218384911, "dur": 206,
    "args": {
      "External id": 56634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56634,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56634, "pid": 5, "tid": 7, "ts": 1716454218384911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281700, "dur": 14,
    "args": {
      "External id": 56634,
      "cbid": 211, "correlation": 56634
    }
  },
  {
    "ph": "s", "id": 56634, "pid": 76337, "tid": -914061504, "ts": 1716454218281700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218385118, "dur": 67,
    "args": {
      "External id": 56642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56642,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56642, "pid": 5, "tid": 7, "ts": 1716454218385118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281770, "dur": 12,
    "args": {
      "External id": 56642,
      "cbid": 211, "correlation": 56642
    }
  },
  {
    "ph": "s", "id": 56642, "pid": 76337, "tid": -914061504, "ts": 1716454218281770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218385186, "dur": 97,
    "args": {
      "External id": 56650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56650,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56650, "pid": 5, "tid": 7, "ts": 1716454218385186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281813, "dur": 9,
    "args": {
      "External id": 56650,
      "cbid": 211, "correlation": 56650
    }
  },
  {
    "ph": "s", "id": 56650, "pid": 76337, "tid": -914061504, "ts": 1716454218281813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218385284, "dur": 54,
    "args": {
      "External id": 56661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56661, "pid": 5, "tid": 7, "ts": 1716454218385284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281885, "dur": 12,
    "args": {
      "External id": 56661,
      "cbid": 211, "correlation": 56661
    }
  },
  {
    "ph": "s", "id": 56661, "pid": 76337, "tid": -914061504, "ts": 1716454218281885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218385340, "dur": 92,
    "args": {
      "External id": 56683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56683,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56683, "pid": 5, "tid": 7, "ts": 1716454218385340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218281916, "dur": 7,
    "args": {
      "External id": 56683,
      "cbid": 211, "correlation": 56683
    }
  },
  {
    "ph": "s", "id": 56683, "pid": 76337, "tid": -914061504, "ts": 1716454218281916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282007, "dur": 1,
    "args": {
      "External id": 56694,
      "cbid": 251, "correlation": 56694
    }
  },
  {
    "ph": "f", "id": 56694, "pid": 76337, "tid": -914061504, "ts": 1716454218282007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218385433, "dur": 105,
    "args": {
      "External id": 56695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56695,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56695, "pid": 5, "tid": 7, "ts": 1716454218385433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282012, "dur": 13,
    "args": {
      "External id": 56695,
      "cbid": 211, "correlation": 56695
    }
  },
  {
    "ph": "s", "id": 56695, "pid": 76337, "tid": -914061504, "ts": 1716454218282012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282089, "dur": 1,
    "args": {
      "External id": 56706,
      "cbid": 251, "correlation": 56706
    }
  },
  {
    "ph": "f", "id": 56706, "pid": 76337, "tid": -914061504, "ts": 1716454218282089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282093, "dur": 0,
    "args": {
      "External id": 56707,
      "cbid": 251, "correlation": 56707
    }
  },
  {
    "ph": "f", "id": 56707, "pid": 76337, "tid": -914061504, "ts": 1716454218282093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218385539, "dur": 10,
    "args": {
      "External id": 56708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56708,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 56708, "pid": 5, "tid": 7, "ts": 1716454218385539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282095, "dur": 13,
    "args": {
      "External id": 56708,
      "cbid": 211, "correlation": 56708
    }
  },
  {
    "ph": "s", "id": 56708, "pid": 76337, "tid": -914061504, "ts": 1716454218282095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218385550, "dur": 5,
    "args": {
      "External id": 56710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56710,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 56710, "pid": 5, "tid": 7, "ts": 1716454218385550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282112, "dur": 7,
    "args": {
      "External id": 56710,
      "cbid": 211, "correlation": 56710
    }
  },
  {
    "ph": "s", "id": 56710, "pid": 76337, "tid": -914061504, "ts": 1716454218282112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282173, "dur": 1,
    "args": {
      "External id": 56721,
      "cbid": 251, "correlation": 56721
    }
  },
  {
    "ph": "f", "id": 56721, "pid": 76337, "tid": -914061504, "ts": 1716454218282173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282177, "dur": 0,
    "args": {
      "External id": 56722,
      "cbid": 251, "correlation": 56722
    }
  },
  {
    "ph": "f", "id": 56722, "pid": 76337, "tid": -914061504, "ts": 1716454218282177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218385557, "dur": 6,
    "args": {
      "External id": 56723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56723,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 56723, "pid": 5, "tid": 7, "ts": 1716454218385557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282178, "dur": 15,
    "args": {
      "External id": 56723,
      "cbid": 211, "correlation": 56723
    }
  },
  {
    "ph": "s", "id": 56723, "pid": 76337, "tid": -914061504, "ts": 1716454218282178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218385564, "dur": 3,
    "args": {
      "External id": 56725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56725,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 56725, "pid": 5, "tid": 7, "ts": 1716454218385564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282195, "dur": 6,
    "args": {
      "External id": 56725,
      "cbid": 211, "correlation": 56725
    }
  },
  {
    "ph": "s", "id": 56725, "pid": 76337, "tid": -914061504, "ts": 1716454218282195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218385569, "dur": 155,
    "args": {
      "External id": 56746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56746,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 56746, "pid": 5, "tid": 7, "ts": 1716454218385569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282268, "dur": 12,
    "args": {
      "External id": 56746,
      "cbid": 211, "correlation": 56746
    }
  },
  {
    "ph": "s", "id": 56746, "pid": 76337, "tid": -914061504, "ts": 1716454218282268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282363, "dur": 2,
    "args": {
      "External id": 56764,
      "cbid": 251, "correlation": 56764
    }
  },
  {
    "ph": "f", "id": 56764, "pid": 76337, "tid": -914061504, "ts": 1716454218282363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218385725, "dur": 106,
    "args": {
      "External id": 56766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56766,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 56766, "pid": 5, "tid": 7, "ts": 1716454218385725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282369, "dur": 13,
    "args": {
      "External id": 56766,
      "cbid": 211, "correlation": 56766
    }
  },
  {
    "ph": "s", "id": 56766, "pid": 76337, "tid": -914061504, "ts": 1716454218282369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218385833, "dur": 34,
    "args": {
      "External id": 56774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56774,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56774, "pid": 5, "tid": 7, "ts": 1716454218385833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282439, "dur": 12,
    "args": {
      "External id": 56774,
      "cbid": 211, "correlation": 56774
    }
  },
  {
    "ph": "s", "id": 56774, "pid": 76337, "tid": -914061504, "ts": 1716454218282439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218385869, "dur": 67,
    "args": {
      "External id": 56782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56782, "pid": 5, "tid": 7, "ts": 1716454218385869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282480, "dur": 9,
    "args": {
      "External id": 56782,
      "cbid": 211, "correlation": 56782
    }
  },
  {
    "ph": "s", "id": 56782, "pid": 76337, "tid": -914061504, "ts": 1716454218282480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218385937, "dur": 91,
    "args": {
      "External id": 56804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56804,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56804, "pid": 5, "tid": 7, "ts": 1716454218385937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282535, "dur": 10,
    "args": {
      "External id": 56804,
      "cbid": 211, "correlation": 56804
    }
  },
  {
    "ph": "s", "id": 56804, "pid": 76337, "tid": -914061504, "ts": 1716454218282535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282622, "dur": 1,
    "args": {
      "External id": 56820,
      "cbid": 251, "correlation": 56820
    }
  },
  {
    "ph": "f", "id": 56820, "pid": 76337, "tid": -914061504, "ts": 1716454218282622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218386030, "dur": 575,
    "args": {
      "External id": 56822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56822,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56822, "pid": 5, "tid": 7, "ts": 1716454218386030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282628, "dur": 12,
    "args": {
      "External id": 56822,
      "cbid": 211, "correlation": 56822
    }
  },
  {
    "ph": "s", "id": 56822, "pid": 76337, "tid": -914061504, "ts": 1716454218282628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218386606, "dur": 243,
    "args": {
      "External id": 56830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56830, "pid": 5, "tid": 7, "ts": 1716454218386606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282693, "dur": 12,
    "args": {
      "External id": 56830,
      "cbid": 211, "correlation": 56830
    }
  },
  {
    "ph": "s", "id": 56830, "pid": 76337, "tid": -914061504, "ts": 1716454218282693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218386850, "dur": 252,
    "args": {
      "External id": 56838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56838,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56838, "pid": 5, "tid": 7, "ts": 1716454218386850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282724, "dur": 9,
    "args": {
      "External id": 56838,
      "cbid": 211, "correlation": 56838
    }
  },
  {
    "ph": "s", "id": 56838, "pid": 76337, "tid": -914061504, "ts": 1716454218282724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282806, "dur": 1,
    "args": {
      "External id": 56854,
      "cbid": 251, "correlation": 56854
    }
  },
  {
    "ph": "f", "id": 56854, "pid": 76337, "tid": -914061504, "ts": 1716454218282806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282811, "dur": 0,
    "args": {
      "External id": 56856,
      "cbid": 251, "correlation": 56856
    }
  },
  {
    "ph": "f", "id": 56856, "pid": 76337, "tid": -914061504, "ts": 1716454218282811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218387104, "dur": 357,
    "args": {
      "External id": 56857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56857,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 56857, "pid": 5, "tid": 7, "ts": 1716454218387104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282814, "dur": 12,
    "args": {
      "External id": 56857,
      "cbid": 211, "correlation": 56857
    }
  },
  {
    "ph": "s", "id": 56857, "pid": 76337, "tid": -914061504, "ts": 1716454218282814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218387462, "dur": 50,
    "args": {
      "External id": 56865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56865,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56865, "pid": 5, "tid": 7, "ts": 1716454218387462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282856, "dur": 10,
    "args": {
      "External id": 56865,
      "cbid": 211, "correlation": 56865
    }
  },
  {
    "ph": "s", "id": 56865, "pid": 76337, "tid": -914061504, "ts": 1716454218282856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218387513, "dur": 158,
    "args": {
      "External id": 56876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56876, "pid": 5, "tid": 7, "ts": 1716454218387513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218282926, "dur": 13,
    "args": {
      "External id": 56876,
      "cbid": 211, "correlation": 56876
    }
  },
  {
    "ph": "s", "id": 56876, "pid": 76337, "tid": -914061504, "ts": 1716454218282926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283001, "dur": 0,
    "args": {
      "External id": 56888,
      "cbid": 317, "correlation": 56888
    }
  },
  {
    "ph": "f", "id": 56888, "pid": 76337, "tid": -914061504, "ts": 1716454218283001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283002, "dur": 0,
    "args": {
      "External id": 56889,
      "cbid": 203, "correlation": 56889
    }
  },
  {
    "ph": "f", "id": 56889, "pid": 76337, "tid": -914061504, "ts": 1716454218283002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283003, "dur": 0,
    "args": {
      "External id": 56890,
      "cbid": 205, "correlation": 56890
    }
  },
  {
    "ph": "f", "id": 56890, "pid": 76337, "tid": -914061504, "ts": 1716454218283003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283027, "dur": 1,
    "args": {
      "External id": 56894,
      "cbid": 251, "correlation": 56894
    }
  },
  {
    "ph": "f", "id": 56894, "pid": 76337, "tid": -914061504, "ts": 1716454218283027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283029, "dur": 0,
    "args": {
      "External id": 56895,
      "cbid": 251, "correlation": 56895
    }
  },
  {
    "ph": "f", "id": 56895, "pid": 76337, "tid": -914061504, "ts": 1716454218283029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283029, "dur": 0,
    "args": {
      "External id": 56896,
      "cbid": 251, "correlation": 56896
    }
  },
  {
    "ph": "f", "id": 56896, "pid": 76337, "tid": -914061504, "ts": 1716454218283029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283030, "dur": 0,
    "args": {
      "External id": 56897,
      "cbid": 251, "correlation": 56897
    }
  },
  {
    "ph": "f", "id": 56897, "pid": 76337, "tid": -914061504, "ts": 1716454218283030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283031, "dur": 0,
    "args": {
      "External id": 56898,
      "cbid": 251, "correlation": 56898
    }
  },
  {
    "ph": "f", "id": 56898, "pid": 76337, "tid": -914061504, "ts": 1716454218283031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283032, "dur": 0,
    "args": {
      "External id": 56899,
      "cbid": 251, "correlation": 56899
    }
  },
  {
    "ph": "f", "id": 56899, "pid": 76337, "tid": -914061504, "ts": 1716454218283032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283033, "dur": 0,
    "args": {
      "External id": 56900,
      "cbid": 251, "correlation": 56900
    }
  },
  {
    "ph": "f", "id": 56900, "pid": 76337, "tid": -914061504, "ts": 1716454218283033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283034, "dur": 0,
    "args": {
      "External id": 56901,
      "cbid": 251, "correlation": 56901
    }
  },
  {
    "ph": "f", "id": 56901, "pid": 76337, "tid": -914061504, "ts": 1716454218283034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283035, "dur": 0,
    "args": {
      "External id": 56902,
      "cbid": 251, "correlation": 56902
    }
  },
  {
    "ph": "f", "id": 56902, "pid": 76337, "tid": -914061504, "ts": 1716454218283035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218387672, "dur": 114,
    "args": {
      "External id": 56903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56903,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 56903, "pid": 5, "tid": 7, "ts": 1716454218387672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283037, "dur": 14,
    "args": {
      "External id": 56903,
      "cbid": 211, "correlation": 56903
    }
  },
  {
    "ph": "s", "id": 56903, "pid": 76337, "tid": -914061504, "ts": 1716454218283037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218387788, "dur": 60,
    "args": {
      "External id": 56909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56909, "pid": 5, "tid": 7, "ts": 1716454218387788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283073, "dur": 9,
    "args": {
      "External id": 56909,
      "cbid": 211, "correlation": 56909
    }
  },
  {
    "ph": "s", "id": 56909, "pid": 76337, "tid": -914061504, "ts": 1716454218283073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218387849, "dur": 50,
    "args": {
      "External id": 56917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56917,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56917, "pid": 5, "tid": 7, "ts": 1716454218387849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283106, "dur": 8,
    "args": {
      "External id": 56917,
      "cbid": 211, "correlation": 56917
    }
  },
  {
    "ph": "s", "id": 56917, "pid": 76337, "tid": -914061504, "ts": 1716454218283106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218387900, "dur": 98,
    "args": {
      "External id": 56926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56926,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56926, "pid": 5, "tid": 7, "ts": 1716454218387900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283144, "dur": 11,
    "args": {
      "External id": 56926,
      "cbid": 211, "correlation": 56926
    }
  },
  {
    "ph": "s", "id": 56926, "pid": 76337, "tid": -914061504, "ts": 1716454218283144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218387999, "dur": 93,
    "args": {
      "External id": 56946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56946,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 56946, "pid": 5, "tid": 7, "ts": 1716454218387999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283218, "dur": 11,
    "args": {
      "External id": 56946,
      "cbid": 211, "correlation": 56946
    }
  },
  {
    "ph": "s", "id": 56946, "pid": 76337, "tid": -914061504, "ts": 1716454218283218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218388093, "dur": 5,
    "args": {
      "External id": 56958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56958,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 56958, "pid": 5, "tid": 7, "ts": 1716454218388093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283243, "dur": 7,
    "args": {
      "External id": 56958,
      "cbid": 211, "correlation": 56958
    }
  },
  {
    "ph": "s", "id": 56958, "pid": 76337, "tid": -914061504, "ts": 1716454218283243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218388099, "dur": 109,
    "args": {
      "External id": 56961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56961, "pid": 5, "tid": 7, "ts": 1716454218388099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283262, "dur": 6,
    "args": {
      "External id": 56961,
      "cbid": 211, "correlation": 56961
    }
  },
  {
    "ph": "s", "id": 56961, "pid": 76337, "tid": -914061504, "ts": 1716454218283262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218388210, "dur": 69,
    "args": {
      "External id": 56970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56970,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56970, "pid": 5, "tid": 7, "ts": 1716454218388210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283301, "dur": 10,
    "args": {
      "External id": 56970,
      "cbid": 211, "correlation": 56970
    }
  },
  {
    "ph": "s", "id": 56970, "pid": 76337, "tid": -914061504, "ts": 1716454218283301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283354, "dur": 0,
    "args": {
      "External id": 56980,
      "cbid": 317, "correlation": 56980
    }
  },
  {
    "ph": "f", "id": 56980, "pid": 76337, "tid": -914061504, "ts": 1716454218283354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283355, "dur": 0,
    "args": {
      "External id": 56981,
      "cbid": 203, "correlation": 56981
    }
  },
  {
    "ph": "f", "id": 56981, "pid": 76337, "tid": -914061504, "ts": 1716454218283355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283356, "dur": 0,
    "args": {
      "External id": 56982,
      "cbid": 205, "correlation": 56982
    }
  },
  {
    "ph": "f", "id": 56982, "pid": 76337, "tid": -914061504, "ts": 1716454218283356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218388280, "dur": 76,
    "args": {
      "External id": 56986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56986, "pid": 5, "tid": 7, "ts": 1716454218388280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283372, "dur": 12,
    "args": {
      "External id": 56986,
      "cbid": 211, "correlation": 56986
    }
  },
  {
    "ph": "s", "id": 56986, "pid": 76337, "tid": -914061504, "ts": 1716454218283372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218388357, "dur": 24,
    "args": {
      "External id": 56988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56988, "pid": 5, "tid": 7, "ts": 1716454218388357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283386, "dur": 5,
    "args": {
      "External id": 56988,
      "cbid": 211, "correlation": 56988
    }
  },
  {
    "ph": "s", "id": 56988, "pid": 76337, "tid": -914061504, "ts": 1716454218283386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218388382, "dur": 4,
    "args": {
      "External id": 56990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 56990, "pid": 5, "tid": 7, "ts": 1716454218388382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283396, "dur": 6,
    "args": {
      "External id": 56990,
      "cbid": 211, "correlation": 56990
    }
  },
  {
    "ph": "s", "id": 56990, "pid": 76337, "tid": -914061504, "ts": 1716454218283396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283404, "dur": 0,
    "args": {
      "External id": 56991,
      "cbid": 51, "correlation": 56991
    }
  },
  {
    "ph": "s", "id": 56991, "pid": 76337, "tid": -914061504, "ts": 1716454218283404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218388388, "dur": 1358,
    "args": {
      "External id": 56992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56992,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 56992, "pid": 5, "tid": 7, "ts": 1716454218388388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283405, "dur": 5,
    "args": {
      "External id": 56992,
      "cbid": 211, "correlation": 56992
    }
  },
  {
    "ph": "s", "id": 56992, "pid": 76337, "tid": -914061504, "ts": 1716454218283405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218389747, "dur": 59,
    "args": {
      "External id": 56997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 56997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 56997, "pid": 5, "tid": 7, "ts": 1716454218389747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283433, "dur": 8,
    "args": {
      "External id": 56997,
      "cbid": 211, "correlation": 56997
    }
  },
  {
    "ph": "s", "id": 56997, "pid": 76337, "tid": -914061504, "ts": 1716454218283433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218389807, "dur": 3,
    "args": {
      "External id": 57005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57005, "pid": 5, "tid": 7, "ts": 1716454218389807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283476, "dur": 9,
    "args": {
      "External id": 57005,
      "cbid": 211, "correlation": 57005
    }
  },
  {
    "ph": "s", "id": 57005, "pid": 76337, "tid": -914061504, "ts": 1716454218283476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283544, "dur": 1,
    "args": {
      "External id": 57021,
      "cbid": 251, "correlation": 57021
    }
  },
  {
    "ph": "f", "id": 57021, "pid": 76337, "tid": -914061504, "ts": 1716454218283544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283549, "dur": 0,
    "args": {
      "External id": 57023,
      "cbid": 251, "correlation": 57023
    }
  },
  {
    "ph": "f", "id": 57023, "pid": 76337, "tid": -914061504, "ts": 1716454218283549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218389811, "dur": 11,
    "args": {
      "External id": 57024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57024,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 57024, "pid": 5, "tid": 7, "ts": 1716454218389811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283551, "dur": 12,
    "args": {
      "External id": 57024,
      "cbid": 211, "correlation": 57024
    }
  },
  {
    "ph": "s", "id": 57024, "pid": 76337, "tid": -914061504, "ts": 1716454218283551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218389824, "dur": 5,
    "args": {
      "External id": 57026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57026,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 57026, "pid": 5, "tid": 7, "ts": 1716454218389824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283564, "dur": 5,
    "args": {
      "External id": 57026,
      "cbid": 211, "correlation": 57026
    }
  },
  {
    "ph": "s", "id": 57026, "pid": 76337, "tid": -914061504, "ts": 1716454218283564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218389830, "dur": 55,
    "args": {
      "External id": 57036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57036, "pid": 5, "tid": 7, "ts": 1716454218389830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283621, "dur": 13,
    "args": {
      "External id": 57036,
      "cbid": 211, "correlation": 57036
    }
  },
  {
    "ph": "s", "id": 57036, "pid": 76337, "tid": -914061504, "ts": 1716454218283621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218389886, "dur": 51,
    "args": {
      "External id": 57056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57056,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 57056, "pid": 5, "tid": 7, "ts": 1716454218389886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283687, "dur": 11,
    "args": {
      "External id": 57056,
      "cbid": 211, "correlation": 57056
    }
  },
  {
    "ph": "s", "id": 57056, "pid": 76337, "tid": -914061504, "ts": 1716454218283687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218389939, "dur": 4,
    "args": {
      "External id": 57068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57068,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57068, "pid": 5, "tid": 7, "ts": 1716454218389939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283708, "dur": 6,
    "args": {
      "External id": 57068,
      "cbid": 211, "correlation": 57068
    }
  },
  {
    "ph": "s", "id": 57068, "pid": 76337, "tid": -914061504, "ts": 1716454218283708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218389944, "dur": 54,
    "args": {
      "External id": 57071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57071,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57071, "pid": 5, "tid": 7, "ts": 1716454218389944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283726, "dur": 6,
    "args": {
      "External id": 57071,
      "cbid": 211, "correlation": 57071
    }
  },
  {
    "ph": "s", "id": 57071, "pid": 76337, "tid": -914061504, "ts": 1716454218283726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218389999, "dur": 36,
    "args": {
      "External id": 57080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57080,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57080, "pid": 5, "tid": 7, "ts": 1716454218389999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283767, "dur": 10,
    "args": {
      "External id": 57080,
      "cbid": 211, "correlation": 57080
    }
  },
  {
    "ph": "s", "id": 57080, "pid": 76337, "tid": -914061504, "ts": 1716454218283767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283832, "dur": 0,
    "args": {
      "External id": 57090,
      "cbid": 317, "correlation": 57090
    }
  },
  {
    "ph": "f", "id": 57090, "pid": 76337, "tid": -914061504, "ts": 1716454218283832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283833, "dur": 0,
    "args": {
      "External id": 57091,
      "cbid": 203, "correlation": 57091
    }
  },
  {
    "ph": "f", "id": 57091, "pid": 76337, "tid": -914061504, "ts": 1716454218283833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283834, "dur": 0,
    "args": {
      "External id": 57092,
      "cbid": 205, "correlation": 57092
    }
  },
  {
    "ph": "f", "id": 57092, "pid": 76337, "tid": -914061504, "ts": 1716454218283834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218390037, "dur": 41,
    "args": {
      "External id": 57096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57096, "pid": 5, "tid": 7, "ts": 1716454218390037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283849, "dur": 13,
    "args": {
      "External id": 57096,
      "cbid": 211, "correlation": 57096
    }
  },
  {
    "ph": "s", "id": 57096, "pid": 76337, "tid": -914061504, "ts": 1716454218283849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218390080, "dur": 15,
    "args": {
      "External id": 57098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57098, "pid": 5, "tid": 7, "ts": 1716454218390080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283864, "dur": 5,
    "args": {
      "External id": 57098,
      "cbid": 211, "correlation": 57098
    }
  },
  {
    "ph": "s", "id": 57098, "pid": 76337, "tid": -914061504, "ts": 1716454218283864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218390095, "dur": 3,
    "args": {
      "External id": 57100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57100, "pid": 5, "tid": 7, "ts": 1716454218390095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283874, "dur": 5,
    "args": {
      "External id": 57100,
      "cbid": 211, "correlation": 57100
    }
  },
  {
    "ph": "s", "id": 57100, "pid": 76337, "tid": -914061504, "ts": 1716454218283874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283882, "dur": 0,
    "args": {
      "External id": 57101,
      "cbid": 51, "correlation": 57101
    }
  },
  {
    "ph": "s", "id": 57101, "pid": 76337, "tid": -914061504, "ts": 1716454218283882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218390100, "dur": 695,
    "args": {
      "External id": 57102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57102,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57102, "pid": 5, "tid": 7, "ts": 1716454218390100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283882, "dur": 5,
    "args": {
      "External id": 57102,
      "cbid": 211, "correlation": 57102
    }
  },
  {
    "ph": "s", "id": 57102, "pid": 76337, "tid": -914061504, "ts": 1716454218283882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218390797, "dur": 60,
    "args": {
      "External id": 57107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57107, "pid": 5, "tid": 7, "ts": 1716454218390797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283911, "dur": 8,
    "args": {
      "External id": 57107,
      "cbid": 211, "correlation": 57107
    }
  },
  {
    "ph": "s", "id": 57107, "pid": 76337, "tid": -914061504, "ts": 1716454218283911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283968, "dur": 0,
    "args": {
      "External id": 57117,
      "cbid": 317, "correlation": 57117
    }
  },
  {
    "ph": "f", "id": 57117, "pid": 76337, "tid": -914061504, "ts": 1716454218283968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283969, "dur": 0,
    "args": {
      "External id": 57118,
      "cbid": 203, "correlation": 57118
    }
  },
  {
    "ph": "f", "id": 57118, "pid": 76337, "tid": -914061504, "ts": 1716454218283969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283970, "dur": 0,
    "args": {
      "External id": 57119,
      "cbid": 205, "correlation": 57119
    }
  },
  {
    "ph": "f", "id": 57119, "pid": 76337, "tid": -914061504, "ts": 1716454218283970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218390858, "dur": 75,
    "args": {
      "External id": 57123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57123, "pid": 5, "tid": 7, "ts": 1716454218390858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218283990, "dur": 12,
    "args": {
      "External id": 57123,
      "cbid": 211, "correlation": 57123
    }
  },
  {
    "ph": "s", "id": 57123, "pid": 76337, "tid": -914061504, "ts": 1716454218283990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218390934, "dur": 208,
    "args": {
      "External id": 57125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57125,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57125, "pid": 5, "tid": 7, "ts": 1716454218390934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284009, "dur": 9,
    "args": {
      "External id": 57125,
      "cbid": 211, "correlation": 57125
    }
  },
  {
    "ph": "s", "id": 57125, "pid": 76337, "tid": -914061504, "ts": 1716454218284009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218391143, "dur": 37,
    "args": {
      "External id": 57127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57127,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57127, "pid": 5, "tid": 7, "ts": 1716454218391143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284022, "dur": 5,
    "args": {
      "External id": 57127,
      "cbid": 211, "correlation": 57127
    }
  },
  {
    "ph": "s", "id": 57127, "pid": 76337, "tid": -914061504, "ts": 1716454218284022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218391182, "dur": 60,
    "args": {
      "External id": 57133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57133, "pid": 5, "tid": 7, "ts": 1716454218391182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284049, "dur": 9,
    "args": {
      "External id": 57133,
      "cbid": 211, "correlation": 57133
    }
  },
  {
    "ph": "s", "id": 57133, "pid": 76337, "tid": -914061504, "ts": 1716454218284049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218391243, "dur": 50,
    "args": {
      "External id": 57141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57141,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57141, "pid": 5, "tid": 7, "ts": 1716454218391243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284079, "dur": 8,
    "args": {
      "External id": 57141,
      "cbid": 211, "correlation": 57141
    }
  },
  {
    "ph": "s", "id": 57141, "pid": 76337, "tid": -914061504, "ts": 1716454218284079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218391294, "dur": 35,
    "args": {
      "External id": 57149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57149,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57149, "pid": 5, "tid": 7, "ts": 1716454218391294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284110, "dur": 9,
    "args": {
      "External id": 57149,
      "cbid": 211, "correlation": 57149
    }
  },
  {
    "ph": "s", "id": 57149, "pid": 76337, "tid": -914061504, "ts": 1716454218284110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218391330, "dur": 52,
    "args": {
      "External id": 57169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57169,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 57169, "pid": 5, "tid": 7, "ts": 1716454218391330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284193, "dur": 12,
    "args": {
      "External id": 57169,
      "cbid": 211, "correlation": 57169
    }
  },
  {
    "ph": "s", "id": 57169, "pid": 76337, "tid": -914061504, "ts": 1716454218284193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218391384, "dur": 4,
    "args": {
      "External id": 57181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57181,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57181, "pid": 5, "tid": 7, "ts": 1716454218391384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284215, "dur": 6,
    "args": {
      "External id": 57181,
      "cbid": 211, "correlation": 57181
    }
  },
  {
    "ph": "s", "id": 57181, "pid": 76337, "tid": -914061504, "ts": 1716454218284215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218391390, "dur": 55,
    "args": {
      "External id": 57184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57184, "pid": 5, "tid": 7, "ts": 1716454218391390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284233, "dur": 6,
    "args": {
      "External id": 57184,
      "cbid": 211, "correlation": 57184
    }
  },
  {
    "ph": "s", "id": 57184, "pid": 76337, "tid": -914061504, "ts": 1716454218284233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284291, "dur": 0,
    "args": {
      "External id": 57195,
      "cbid": 317, "correlation": 57195
    }
  },
  {
    "ph": "f", "id": 57195, "pid": 76337, "tid": -914061504, "ts": 1716454218284291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284291, "dur": 0,
    "args": {
      "External id": 57196,
      "cbid": 203, "correlation": 57196
    }
  },
  {
    "ph": "f", "id": 57196, "pid": 76337, "tid": -914061504, "ts": 1716454218284291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284292, "dur": 0,
    "args": {
      "External id": 57197,
      "cbid": 205, "correlation": 57197
    }
  },
  {
    "ph": "f", "id": 57197, "pid": 76337, "tid": -914061504, "ts": 1716454218284292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284314, "dur": 1,
    "args": {
      "External id": 57201,
      "cbid": 251, "correlation": 57201
    }
  },
  {
    "ph": "f", "id": 57201, "pid": 76337, "tid": -914061504, "ts": 1716454218284314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284316, "dur": 0,
    "args": {
      "External id": 57202,
      "cbid": 251, "correlation": 57202
    }
  },
  {
    "ph": "f", "id": 57202, "pid": 76337, "tid": -914061504, "ts": 1716454218284316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284317, "dur": 0,
    "args": {
      "External id": 57203,
      "cbid": 251, "correlation": 57203
    }
  },
  {
    "ph": "f", "id": 57203, "pid": 76337, "tid": -914061504, "ts": 1716454218284317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284318, "dur": 0,
    "args": {
      "External id": 57204,
      "cbid": 251, "correlation": 57204
    }
  },
  {
    "ph": "f", "id": 57204, "pid": 76337, "tid": -914061504, "ts": 1716454218284318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284319, "dur": 0,
    "args": {
      "External id": 57205,
      "cbid": 251, "correlation": 57205
    }
  },
  {
    "ph": "f", "id": 57205, "pid": 76337, "tid": -914061504, "ts": 1716454218284319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284319, "dur": 0,
    "args": {
      "External id": 57206,
      "cbid": 251, "correlation": 57206
    }
  },
  {
    "ph": "f", "id": 57206, "pid": 76337, "tid": -914061504, "ts": 1716454218284319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284320, "dur": 0,
    "args": {
      "External id": 57207,
      "cbid": 251, "correlation": 57207
    }
  },
  {
    "ph": "f", "id": 57207, "pid": 76337, "tid": -914061504, "ts": 1716454218284320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284321, "dur": 0,
    "args": {
      "External id": 57208,
      "cbid": 251, "correlation": 57208
    }
  },
  {
    "ph": "f", "id": 57208, "pid": 76337, "tid": -914061504, "ts": 1716454218284321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284322, "dur": 0,
    "args": {
      "External id": 57209,
      "cbid": 251, "correlation": 57209
    }
  },
  {
    "ph": "f", "id": 57209, "pid": 76337, "tid": -914061504, "ts": 1716454218284322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218391446, "dur": 115,
    "args": {
      "External id": 57210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57210,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57210, "pid": 5, "tid": 7, "ts": 1716454218391446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284324, "dur": 13,
    "args": {
      "External id": 57210,
      "cbid": 211, "correlation": 57210
    }
  },
  {
    "ph": "s", "id": 57210, "pid": 76337, "tid": -914061504, "ts": 1716454218284324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218391562, "dur": 59,
    "args": {
      "External id": 57216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57216,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57216, "pid": 5, "tid": 7, "ts": 1716454218391562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284359, "dur": 9,
    "args": {
      "External id": 57216,
      "cbid": 211, "correlation": 57216
    }
  },
  {
    "ph": "s", "id": 57216, "pid": 76337, "tid": -914061504, "ts": 1716454218284359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218391623, "dur": 604,
    "args": {
      "External id": 57225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57225,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57225, "pid": 5, "tid": 7, "ts": 1716454218391623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284444, "dur": 14,
    "args": {
      "External id": 57225,
      "cbid": 211, "correlation": 57225
    }
  },
  {
    "ph": "s", "id": 57225, "pid": 76337, "tid": -914061504, "ts": 1716454218284444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218392228, "dur": 180,
    "args": {
      "External id": 57247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57247,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57247, "pid": 5, "tid": 7, "ts": 1716454218392228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284502, "dur": 10,
    "args": {
      "External id": 57247,
      "cbid": 211, "correlation": 57247
    }
  },
  {
    "ph": "s", "id": 57247, "pid": 76337, "tid": -914061504, "ts": 1716454218284502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284587, "dur": 1,
    "args": {
      "External id": 57258,
      "cbid": 251, "correlation": 57258
    }
  },
  {
    "ph": "f", "id": 57258, "pid": 76337, "tid": -914061504, "ts": 1716454218284587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218392409, "dur": 194,
    "args": {
      "External id": 57259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57259,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57259, "pid": 5, "tid": 7, "ts": 1716454218392409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284593, "dur": 13,
    "args": {
      "External id": 57259,
      "cbid": 211, "correlation": 57259
    }
  },
  {
    "ph": "s", "id": 57259, "pid": 76337, "tid": -914061504, "ts": 1716454218284593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284661, "dur": 1,
    "args": {
      "External id": 57270,
      "cbid": 251, "correlation": 57270
    }
  },
  {
    "ph": "f", "id": 57270, "pid": 76337, "tid": -914061504, "ts": 1716454218284661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218392604, "dur": 189,
    "args": {
      "External id": 57271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57271,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57271, "pid": 5, "tid": 7, "ts": 1716454218392604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284665, "dur": 11,
    "args": {
      "External id": 57271,
      "cbid": 211, "correlation": 57271
    }
  },
  {
    "ph": "s", "id": 57271, "pid": 76337, "tid": -914061504, "ts": 1716454218284665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284728, "dur": 1,
    "args": {
      "External id": 57282,
      "cbid": 251, "correlation": 57282
    }
  },
  {
    "ph": "f", "id": 57282, "pid": 76337, "tid": -914061504, "ts": 1716454218284728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218392795, "dur": 189,
    "args": {
      "External id": 57283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57283,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57283, "pid": 5, "tid": 7, "ts": 1716454218392795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284733, "dur": 11,
    "args": {
      "External id": 57283,
      "cbid": 211, "correlation": 57283
    }
  },
  {
    "ph": "s", "id": 57283, "pid": 76337, "tid": -914061504, "ts": 1716454218284733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218392986, "dur": 18553,
    "args": {
      "External id": 57304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57304,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57304, "pid": 5, "tid": 7, "ts": 1716454218392986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284816, "dur": 12,
    "args": {
      "External id": 57304,
      "cbid": 211, "correlation": 57304
    }
  },
  {
    "ph": "s", "id": 57304, "pid": 76337, "tid": -914061504, "ts": 1716454218284816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284912, "dur": 1,
    "args": {
      "External id": 57322,
      "cbid": 251, "correlation": 57322
    }
  },
  {
    "ph": "f", "id": 57322, "pid": 76337, "tid": -914061504, "ts": 1716454218284912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218411540, "dur": 200,
    "args": {
      "External id": 57324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57324,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57324, "pid": 5, "tid": 7, "ts": 1716454218411540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284918, "dur": 14,
    "args": {
      "External id": 57324,
      "cbid": 211, "correlation": 57324
    }
  },
  {
    "ph": "s", "id": 57324, "pid": 76337, "tid": -914061504, "ts": 1716454218284918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218411742, "dur": 66,
    "args": {
      "External id": 57332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57332,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57332, "pid": 5, "tid": 7, "ts": 1716454218411742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218284996, "dur": 13,
    "args": {
      "External id": 57332,
      "cbid": 211, "correlation": 57332
    }
  },
  {
    "ph": "s", "id": 57332, "pid": 76337, "tid": -914061504, "ts": 1716454218284996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218411809, "dur": 97,
    "args": {
      "External id": 57340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57340,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57340, "pid": 5, "tid": 7, "ts": 1716454218411809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218285037, "dur": 9,
    "args": {
      "External id": 57340,
      "cbid": 211, "correlation": 57340
    }
  },
  {
    "ph": "s", "id": 57340, "pid": 76337, "tid": -914061504, "ts": 1716454218285037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218411908, "dur": 53,
    "args": {
      "External id": 57351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57351,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57351, "pid": 5, "tid": 7, "ts": 1716454218411908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218285110, "dur": 12,
    "args": {
      "External id": 57351,
      "cbid": 211, "correlation": 57351
    }
  },
  {
    "ph": "s", "id": 57351, "pid": 76337, "tid": -914061504, "ts": 1716454218285110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218411962, "dur": 92,
    "args": {
      "External id": 57373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57373,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57373, "pid": 5, "tid": 7, "ts": 1716454218411962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218285142, "dur": 751,
    "args": {
      "External id": 57373,
      "cbid": 211, "correlation": 57373
    }
  },
  {
    "ph": "s", "id": 57373, "pid": 76337, "tid": -914061504, "ts": 1716454218285142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218285972, "dur": 7,
    "args": {
      "External id": 57384,
      "cbid": 251, "correlation": 57384
    }
  },
  {
    "ph": "f", "id": 57384, "pid": 76337, "tid": -914061504, "ts": 1716454218285972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218412056, "dur": 104,
    "args": {
      "External id": 57385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57385,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57385, "pid": 5, "tid": 7, "ts": 1716454218412056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218285984, "dur": 55,
    "args": {
      "External id": 57385,
      "cbid": 211, "correlation": 57385
    }
  },
  {
    "ph": "s", "id": 57385, "pid": 76337, "tid": -914061504, "ts": 1716454218285984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286101, "dur": 1,
    "args": {
      "External id": 57396,
      "cbid": 251, "correlation": 57396
    }
  },
  {
    "ph": "f", "id": 57396, "pid": 76337, "tid": -914061504, "ts": 1716454218286101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286105, "dur": 0,
    "args": {
      "External id": 57397,
      "cbid": 251, "correlation": 57397
    }
  },
  {
    "ph": "f", "id": 57397, "pid": 76337, "tid": -914061504, "ts": 1716454218286105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218412161, "dur": 10,
    "args": {
      "External id": 57398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57398,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 57398, "pid": 5, "tid": 7, "ts": 1716454218412161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286106, "dur": 12,
    "args": {
      "External id": 57398,
      "cbid": 211, "correlation": 57398
    }
  },
  {
    "ph": "s", "id": 57398, "pid": 76337, "tid": -914061504, "ts": 1716454218286106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218412172, "dur": 5,
    "args": {
      "External id": 57400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57400,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 57400, "pid": 5, "tid": 7, "ts": 1716454218412172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286121, "dur": 6,
    "args": {
      "External id": 57400,
      "cbid": 211, "correlation": 57400
    }
  },
  {
    "ph": "s", "id": 57400, "pid": 76337, "tid": -914061504, "ts": 1716454218286121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286181, "dur": 1,
    "args": {
      "External id": 57411,
      "cbid": 251, "correlation": 57411
    }
  },
  {
    "ph": "f", "id": 57411, "pid": 76337, "tid": -914061504, "ts": 1716454218286181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286185, "dur": 0,
    "args": {
      "External id": 57412,
      "cbid": 251, "correlation": 57412
    }
  },
  {
    "ph": "f", "id": 57412, "pid": 76337, "tid": -914061504, "ts": 1716454218286185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218412178, "dur": 6,
    "args": {
      "External id": 57413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57413,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 57413, "pid": 5, "tid": 7, "ts": 1716454218412178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286186, "dur": 12,
    "args": {
      "External id": 57413,
      "cbid": 211, "correlation": 57413
    }
  },
  {
    "ph": "s", "id": 57413, "pid": 76337, "tid": -914061504, "ts": 1716454218286186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218412186, "dur": 3,
    "args": {
      "External id": 57415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57415,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 57415, "pid": 5, "tid": 7, "ts": 1716454218412186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286200, "dur": 6,
    "args": {
      "External id": 57415,
      "cbid": 211, "correlation": 57415
    }
  },
  {
    "ph": "s", "id": 57415, "pid": 76337, "tid": -914061504, "ts": 1716454218286200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218412191, "dur": 155,
    "args": {
      "External id": 57436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57436,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57436, "pid": 5, "tid": 7, "ts": 1716454218412191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286273, "dur": 13,
    "args": {
      "External id": 57436,
      "cbid": 211, "correlation": 57436
    }
  },
  {
    "ph": "s", "id": 57436, "pid": 76337, "tid": -914061504, "ts": 1716454218286273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286372, "dur": 1,
    "args": {
      "External id": 57454,
      "cbid": 251, "correlation": 57454
    }
  },
  {
    "ph": "f", "id": 57454, "pid": 76337, "tid": -914061504, "ts": 1716454218286372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218412347, "dur": 107,
    "args": {
      "External id": 57456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57456,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57456, "pid": 5, "tid": 7, "ts": 1716454218412347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286378, "dur": 13,
    "args": {
      "External id": 57456,
      "cbid": 211, "correlation": 57456
    }
  },
  {
    "ph": "s", "id": 57456, "pid": 76337, "tid": -914061504, "ts": 1716454218286378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218412455, "dur": 34,
    "args": {
      "External id": 57464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57464,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57464, "pid": 5, "tid": 7, "ts": 1716454218412455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286448, "dur": 12,
    "args": {
      "External id": 57464,
      "cbid": 211, "correlation": 57464
    }
  },
  {
    "ph": "s", "id": 57464, "pid": 76337, "tid": -914061504, "ts": 1716454218286448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218412491, "dur": 66,
    "args": {
      "External id": 57472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57472,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57472, "pid": 5, "tid": 7, "ts": 1716454218412491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286488, "dur": 9,
    "args": {
      "External id": 57472,
      "cbid": 211, "correlation": 57472
    }
  },
  {
    "ph": "s", "id": 57472, "pid": 76337, "tid": -914061504, "ts": 1716454218286488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218412558, "dur": 92,
    "args": {
      "External id": 57494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57494,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57494, "pid": 5, "tid": 7, "ts": 1716454218412558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286539, "dur": 10,
    "args": {
      "External id": 57494,
      "cbid": 211, "correlation": 57494
    }
  },
  {
    "ph": "s", "id": 57494, "pid": 76337, "tid": -914061504, "ts": 1716454218286539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286624, "dur": 1,
    "args": {
      "External id": 57510,
      "cbid": 251, "correlation": 57510
    }
  },
  {
    "ph": "f", "id": 57510, "pid": 76337, "tid": -914061504, "ts": 1716454218286624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218412651, "dur": 570,
    "args": {
      "External id": 57512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57512,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57512, "pid": 5, "tid": 7, "ts": 1716454218412651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286630, "dur": 12,
    "args": {
      "External id": 57512,
      "cbid": 211, "correlation": 57512
    }
  },
  {
    "ph": "s", "id": 57512, "pid": 76337, "tid": -914061504, "ts": 1716454218286630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218413223, "dur": 242,
    "args": {
      "External id": 57520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57520, "pid": 5, "tid": 7, "ts": 1716454218413223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286694, "dur": 13,
    "args": {
      "External id": 57520,
      "cbid": 211, "correlation": 57520
    }
  },
  {
    "ph": "s", "id": 57520, "pid": 76337, "tid": -914061504, "ts": 1716454218286694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218413466, "dur": 251,
    "args": {
      "External id": 57528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57528,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57528, "pid": 5, "tid": 7, "ts": 1716454218413466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286725, "dur": 12,
    "args": {
      "External id": 57528,
      "cbid": 211, "correlation": 57528
    }
  },
  {
    "ph": "s", "id": 57528, "pid": 76337, "tid": -914061504, "ts": 1716454218286725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286811, "dur": 1,
    "args": {
      "External id": 57544,
      "cbid": 251, "correlation": 57544
    }
  },
  {
    "ph": "f", "id": 57544, "pid": 76337, "tid": -914061504, "ts": 1716454218286811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286816, "dur": 0,
    "args": {
      "External id": 57546,
      "cbid": 251, "correlation": 57546
    }
  },
  {
    "ph": "f", "id": 57546, "pid": 76337, "tid": -914061504, "ts": 1716454218286816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218413718, "dur": 359,
    "args": {
      "External id": 57547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57547,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 57547, "pid": 5, "tid": 7, "ts": 1716454218413718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286819, "dur": 13,
    "args": {
      "External id": 57547,
      "cbid": 211, "correlation": 57547
    }
  },
  {
    "ph": "s", "id": 57547, "pid": 76337, "tid": -914061504, "ts": 1716454218286819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218414079, "dur": 50,
    "args": {
      "External id": 57555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57555,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57555, "pid": 5, "tid": 7, "ts": 1716454218414079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286861, "dur": 10,
    "args": {
      "External id": 57555,
      "cbid": 211, "correlation": 57555
    }
  },
  {
    "ph": "s", "id": 57555, "pid": 76337, "tid": -914061504, "ts": 1716454218286861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218414130, "dur": 158,
    "args": {
      "External id": 57566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57566, "pid": 5, "tid": 7, "ts": 1716454218414130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218286928, "dur": 3156,
    "args": {
      "External id": 57566,
      "cbid": 211, "correlation": 57566
    }
  },
  {
    "ph": "s", "id": 57566, "pid": 76337, "tid": -914061504, "ts": 1716454218286928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290216, "dur": 0,
    "args": {
      "External id": 57578,
      "cbid": 317, "correlation": 57578
    }
  },
  {
    "ph": "f", "id": 57578, "pid": 76337, "tid": -914061504, "ts": 1716454218290216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290218, "dur": 1,
    "args": {
      "External id": 57579,
      "cbid": 203, "correlation": 57579
    }
  },
  {
    "ph": "f", "id": 57579, "pid": 76337, "tid": -914061504, "ts": 1716454218290218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290219, "dur": 1,
    "args": {
      "External id": 57580,
      "cbid": 205, "correlation": 57580
    }
  },
  {
    "ph": "f", "id": 57580, "pid": 76337, "tid": -914061504, "ts": 1716454218290219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290268, "dur": 2,
    "args": {
      "External id": 57584,
      "cbid": 251, "correlation": 57584
    }
  },
  {
    "ph": "f", "id": 57584, "pid": 76337, "tid": -914061504, "ts": 1716454218290268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290272, "dur": 0,
    "args": {
      "External id": 57585,
      "cbid": 251, "correlation": 57585
    }
  },
  {
    "ph": "f", "id": 57585, "pid": 76337, "tid": -914061504, "ts": 1716454218290272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290273, "dur": 0,
    "args": {
      "External id": 57586,
      "cbid": 251, "correlation": 57586
    }
  },
  {
    "ph": "f", "id": 57586, "pid": 76337, "tid": -914061504, "ts": 1716454218290273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290274, "dur": 0,
    "args": {
      "External id": 57587,
      "cbid": 251, "correlation": 57587
    }
  },
  {
    "ph": "f", "id": 57587, "pid": 76337, "tid": -914061504, "ts": 1716454218290274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290275, "dur": 1,
    "args": {
      "External id": 57588,
      "cbid": 251, "correlation": 57588
    }
  },
  {
    "ph": "f", "id": 57588, "pid": 76337, "tid": -914061504, "ts": 1716454218290275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290277, "dur": 1,
    "args": {
      "External id": 57589,
      "cbid": 251, "correlation": 57589
    }
  },
  {
    "ph": "f", "id": 57589, "pid": 76337, "tid": -914061504, "ts": 1716454218290277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290278, "dur": 0,
    "args": {
      "External id": 57590,
      "cbid": 251, "correlation": 57590
    }
  },
  {
    "ph": "f", "id": 57590, "pid": 76337, "tid": -914061504, "ts": 1716454218290278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290280, "dur": 1,
    "args": {
      "External id": 57591,
      "cbid": 251, "correlation": 57591
    }
  },
  {
    "ph": "f", "id": 57591, "pid": 76337, "tid": -914061504, "ts": 1716454218290280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290282, "dur": 0,
    "args": {
      "External id": 57592,
      "cbid": 251, "correlation": 57592
    }
  },
  {
    "ph": "f", "id": 57592, "pid": 76337, "tid": -914061504, "ts": 1716454218290282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218414289, "dur": 115,
    "args": {
      "External id": 57593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57593,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57593, "pid": 5, "tid": 7, "ts": 1716454218414289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290285, "dur": 21,
    "args": {
      "External id": 57593,
      "cbid": 211, "correlation": 57593
    }
  },
  {
    "ph": "s", "id": 57593, "pid": 76337, "tid": -914061504, "ts": 1716454218290285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218414405, "dur": 59,
    "args": {
      "External id": 57599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57599, "pid": 5, "tid": 7, "ts": 1716454218414405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290346, "dur": 12,
    "args": {
      "External id": 57599,
      "cbid": 211, "correlation": 57599
    }
  },
  {
    "ph": "s", "id": 57599, "pid": 76337, "tid": -914061504, "ts": 1716454218290346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218414466, "dur": 49,
    "args": {
      "External id": 57607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57607,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57607, "pid": 5, "tid": 7, "ts": 1716454218414466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290389, "dur": 12,
    "args": {
      "External id": 57607,
      "cbid": 211, "correlation": 57607
    }
  },
  {
    "ph": "s", "id": 57607, "pid": 76337, "tid": -914061504, "ts": 1716454218290389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218414517, "dur": 98,
    "args": {
      "External id": 57616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57616,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57616, "pid": 5, "tid": 7, "ts": 1716454218414517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290448, "dur": 13,
    "args": {
      "External id": 57616,
      "cbid": 211, "correlation": 57616
    }
  },
  {
    "ph": "s", "id": 57616, "pid": 76337, "tid": -914061504, "ts": 1716454218290448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218414616, "dur": 91,
    "args": {
      "External id": 57636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57636,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 57636, "pid": 5, "tid": 7, "ts": 1716454218414616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290557, "dur": 13,
    "args": {
      "External id": 57636,
      "cbid": 211, "correlation": 57636
    }
  },
  {
    "ph": "s", "id": 57636, "pid": 76337, "tid": -914061504, "ts": 1716454218290557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218414709, "dur": 5,
    "args": {
      "External id": 57648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57648,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 57648, "pid": 5, "tid": 7, "ts": 1716454218414709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290585, "dur": 9,
    "args": {
      "External id": 57648,
      "cbid": 211, "correlation": 57648
    }
  },
  {
    "ph": "s", "id": 57648, "pid": 76337, "tid": -914061504, "ts": 1716454218290585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218414715, "dur": 108,
    "args": {
      "External id": 57651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57651,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57651, "pid": 5, "tid": 7, "ts": 1716454218414715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290609, "dur": 8,
    "args": {
      "External id": 57651,
      "cbid": 211, "correlation": 57651
    }
  },
  {
    "ph": "s", "id": 57651, "pid": 76337, "tid": -914061504, "ts": 1716454218290609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218414824, "dur": 69,
    "args": {
      "External id": 57660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57660,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57660, "pid": 5, "tid": 7, "ts": 1716454218414824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290659, "dur": 12,
    "args": {
      "External id": 57660,
      "cbid": 211, "correlation": 57660
    }
  },
  {
    "ph": "s", "id": 57660, "pid": 76337, "tid": -914061504, "ts": 1716454218290659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290714, "dur": 0,
    "args": {
      "External id": 57670,
      "cbid": 317, "correlation": 57670
    }
  },
  {
    "ph": "f", "id": 57670, "pid": 76337, "tid": -914061504, "ts": 1716454218290714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290715, "dur": 0,
    "args": {
      "External id": 57671,
      "cbid": 203, "correlation": 57671
    }
  },
  {
    "ph": "f", "id": 57671, "pid": 76337, "tid": -914061504, "ts": 1716454218290715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290715, "dur": 0,
    "args": {
      "External id": 57672,
      "cbid": 205, "correlation": 57672
    }
  },
  {
    "ph": "f", "id": 57672, "pid": 76337, "tid": -914061504, "ts": 1716454218290715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218414894, "dur": 76,
    "args": {
      "External id": 57676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57676,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57676, "pid": 5, "tid": 7, "ts": 1716454218414894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290739, "dur": 14,
    "args": {
      "External id": 57676,
      "cbid": 211, "correlation": 57676
    }
  },
  {
    "ph": "s", "id": 57676, "pid": 76337, "tid": -914061504, "ts": 1716454218290739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218414972, "dur": 24,
    "args": {
      "External id": 57678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57678, "pid": 5, "tid": 7, "ts": 1716454218414972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290756, "dur": 5,
    "args": {
      "External id": 57678,
      "cbid": 211, "correlation": 57678
    }
  },
  {
    "ph": "s", "id": 57678, "pid": 76337, "tid": -914061504, "ts": 1716454218290756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218414997, "dur": 4,
    "args": {
      "External id": 57680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57680, "pid": 5, "tid": 7, "ts": 1716454218414997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290771, "dur": 11,
    "args": {
      "External id": 57680,
      "cbid": 211, "correlation": 57680
    }
  },
  {
    "ph": "s", "id": 57680, "pid": 76337, "tid": -914061504, "ts": 1716454218290771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290787, "dur": 0,
    "args": {
      "External id": 57681,
      "cbid": 51, "correlation": 57681
    }
  },
  {
    "ph": "s", "id": 57681, "pid": 76337, "tid": -914061504, "ts": 1716454218290787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218415002, "dur": 1357,
    "args": {
      "External id": 57682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57682,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57682, "pid": 5, "tid": 7, "ts": 1716454218415002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290789, "dur": 7,
    "args": {
      "External id": 57682,
      "cbid": 211, "correlation": 57682
    }
  },
  {
    "ph": "s", "id": 57682, "pid": 76337, "tid": -914061504, "ts": 1716454218290789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218416360, "dur": 58,
    "args": {
      "External id": 57687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57687,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57687, "pid": 5, "tid": 7, "ts": 1716454218416360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290818, "dur": 9,
    "args": {
      "External id": 57687,
      "cbid": 211, "correlation": 57687
    }
  },
  {
    "ph": "s", "id": 57687, "pid": 76337, "tid": -914061504, "ts": 1716454218290818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218416420, "dur": 4,
    "args": {
      "External id": 57695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57695,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57695, "pid": 5, "tid": 7, "ts": 1716454218416420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290862, "dur": 9,
    "args": {
      "External id": 57695,
      "cbid": 211, "correlation": 57695
    }
  },
  {
    "ph": "s", "id": 57695, "pid": 76337, "tid": -914061504, "ts": 1716454218290862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290960, "dur": 2,
    "args": {
      "External id": 57711,
      "cbid": 251, "correlation": 57711
    }
  },
  {
    "ph": "f", "id": 57711, "pid": 76337, "tid": -914061504, "ts": 1716454218290960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290966, "dur": 0,
    "args": {
      "External id": 57713,
      "cbid": 251, "correlation": 57713
    }
  },
  {
    "ph": "f", "id": 57713, "pid": 76337, "tid": -914061504, "ts": 1716454218290966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218416425, "dur": 11,
    "args": {
      "External id": 57714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57714,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 57714, "pid": 5, "tid": 7, "ts": 1716454218416425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290968, "dur": 23,
    "args": {
      "External id": 57714,
      "cbid": 211, "correlation": 57714
    }
  },
  {
    "ph": "s", "id": 57714, "pid": 76337, "tid": -914061504, "ts": 1716454218290968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218416438, "dur": 5,
    "args": {
      "External id": 57716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57716,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 57716, "pid": 5, "tid": 7, "ts": 1716454218416438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218290994, "dur": 8,
    "args": {
      "External id": 57716,
      "cbid": 211, "correlation": 57716
    }
  },
  {
    "ph": "s", "id": 57716, "pid": 76337, "tid": -914061504, "ts": 1716454218290994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218416444, "dur": 55,
    "args": {
      "External id": 57726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57726,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57726, "pid": 5, "tid": 7, "ts": 1716454218416444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291069, "dur": 13,
    "args": {
      "External id": 57726,
      "cbid": 211, "correlation": 57726
    }
  },
  {
    "ph": "s", "id": 57726, "pid": 76337, "tid": -914061504, "ts": 1716454218291069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218416500, "dur": 53,
    "args": {
      "External id": 57746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57746,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 57746, "pid": 5, "tid": 7, "ts": 1716454218416500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291140, "dur": 11,
    "args": {
      "External id": 57746,
      "cbid": 211, "correlation": 57746
    }
  },
  {
    "ph": "s", "id": 57746, "pid": 76337, "tid": -914061504, "ts": 1716454218291140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218416554, "dur": 4,
    "args": {
      "External id": 57758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57758,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57758, "pid": 5, "tid": 7, "ts": 1716454218416554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291161, "dur": 6,
    "args": {
      "External id": 57758,
      "cbid": 211, "correlation": 57758
    }
  },
  {
    "ph": "s", "id": 57758, "pid": 76337, "tid": -914061504, "ts": 1716454218291161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218416560, "dur": 56,
    "args": {
      "External id": 57761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57761,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57761, "pid": 5, "tid": 7, "ts": 1716454218416560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291184, "dur": 6,
    "args": {
      "External id": 57761,
      "cbid": 211, "correlation": 57761
    }
  },
  {
    "ph": "s", "id": 57761, "pid": 76337, "tid": -914061504, "ts": 1716454218291184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218416616, "dur": 36,
    "args": {
      "External id": 57770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57770,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57770, "pid": 5, "tid": 7, "ts": 1716454218416616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291225, "dur": 11,
    "args": {
      "External id": 57770,
      "cbid": 211, "correlation": 57770
    }
  },
  {
    "ph": "s", "id": 57770, "pid": 76337, "tid": -914061504, "ts": 1716454218291225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291294, "dur": 0,
    "args": {
      "External id": 57780,
      "cbid": 317, "correlation": 57780
    }
  },
  {
    "ph": "f", "id": 57780, "pid": 76337, "tid": -914061504, "ts": 1716454218291294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291295, "dur": 0,
    "args": {
      "External id": 57781,
      "cbid": 203, "correlation": 57781
    }
  },
  {
    "ph": "f", "id": 57781, "pid": 76337, "tid": -914061504, "ts": 1716454218291295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291296, "dur": 0,
    "args": {
      "External id": 57782,
      "cbid": 205, "correlation": 57782
    }
  },
  {
    "ph": "f", "id": 57782, "pid": 76337, "tid": -914061504, "ts": 1716454218291296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218416654, "dur": 40,
    "args": {
      "External id": 57786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57786,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57786, "pid": 5, "tid": 7, "ts": 1716454218416654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291313, "dur": 12,
    "args": {
      "External id": 57786,
      "cbid": 211, "correlation": 57786
    }
  },
  {
    "ph": "s", "id": 57786, "pid": 76337, "tid": -914061504, "ts": 1716454218291313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218416695, "dur": 15,
    "args": {
      "External id": 57788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57788, "pid": 5, "tid": 7, "ts": 1716454218416695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291327, "dur": 6,
    "args": {
      "External id": 57788,
      "cbid": 211, "correlation": 57788
    }
  },
  {
    "ph": "s", "id": 57788, "pid": 76337, "tid": -914061504, "ts": 1716454218291327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218416711, "dur": 3,
    "args": {
      "External id": 57790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57790, "pid": 5, "tid": 7, "ts": 1716454218416711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291338, "dur": 5,
    "args": {
      "External id": 57790,
      "cbid": 211, "correlation": 57790
    }
  },
  {
    "ph": "s", "id": 57790, "pid": 76337, "tid": -914061504, "ts": 1716454218291338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291346, "dur": 0,
    "args": {
      "External id": 57791,
      "cbid": 51, "correlation": 57791
    }
  },
  {
    "ph": "s", "id": 57791, "pid": 76337, "tid": -914061504, "ts": 1716454218291346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218416715, "dur": 694,
    "args": {
      "External id": 57792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57792,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57792, "pid": 5, "tid": 7, "ts": 1716454218416715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291347, "dur": 5,
    "args": {
      "External id": 57792,
      "cbid": 211, "correlation": 57792
    }
  },
  {
    "ph": "s", "id": 57792, "pid": 76337, "tid": -914061504, "ts": 1716454218291347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218417411, "dur": 59,
    "args": {
      "External id": 57797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57797, "pid": 5, "tid": 7, "ts": 1716454218417411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291375, "dur": 9,
    "args": {
      "External id": 57797,
      "cbid": 211, "correlation": 57797
    }
  },
  {
    "ph": "s", "id": 57797, "pid": 76337, "tid": -914061504, "ts": 1716454218291375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291437, "dur": 0,
    "args": {
      "External id": 57807,
      "cbid": 317, "correlation": 57807
    }
  },
  {
    "ph": "f", "id": 57807, "pid": 76337, "tid": -914061504, "ts": 1716454218291437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291437, "dur": 0,
    "args": {
      "External id": 57808,
      "cbid": 203, "correlation": 57808
    }
  },
  {
    "ph": "f", "id": 57808, "pid": 76337, "tid": -914061504, "ts": 1716454218291437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291438, "dur": 0,
    "args": {
      "External id": 57809,
      "cbid": 205, "correlation": 57809
    }
  },
  {
    "ph": "f", "id": 57809, "pid": 76337, "tid": -914061504, "ts": 1716454218291438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218417471, "dur": 76,
    "args": {
      "External id": 57813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57813, "pid": 5, "tid": 7, "ts": 1716454218417471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291452, "dur": 12,
    "args": {
      "External id": 57813,
      "cbid": 211, "correlation": 57813
    }
  },
  {
    "ph": "s", "id": 57813, "pid": 76337, "tid": -914061504, "ts": 1716454218291452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454218417549, "dur": 209,
    "args": {
      "External id": 57815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57815,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57815, "pid": 5, "tid": 7, "ts": 1716454218417549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291475, "dur": 9,
    "args": {
      "External id": 57815,
      "cbid": 211, "correlation": 57815
    }
  },
  {
    "ph": "s", "id": 57815, "pid": 76337, "tid": -914061504, "ts": 1716454218291475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218417759, "dur": 39,
    "args": {
      "External id": 57817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57817,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57817, "pid": 5, "tid": 7, "ts": 1716454218417759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291490, "dur": 6,
    "args": {
      "External id": 57817,
      "cbid": 211, "correlation": 57817
    }
  },
  {
    "ph": "s", "id": 57817, "pid": 76337, "tid": -914061504, "ts": 1716454218291490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218417800, "dur": 59,
    "args": {
      "External id": 57823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57823,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57823, "pid": 5, "tid": 7, "ts": 1716454218417800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291517, "dur": 8,
    "args": {
      "External id": 57823,
      "cbid": 211, "correlation": 57823
    }
  },
  {
    "ph": "s", "id": 57823, "pid": 76337, "tid": -914061504, "ts": 1716454218291517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218417861, "dur": 50,
    "args": {
      "External id": 57831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57831,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57831, "pid": 5, "tid": 7, "ts": 1716454218417861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291546, "dur": 8,
    "args": {
      "External id": 57831,
      "cbid": 211, "correlation": 57831
    }
  },
  {
    "ph": "s", "id": 57831, "pid": 76337, "tid": -914061504, "ts": 1716454218291546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218417912, "dur": 35,
    "args": {
      "External id": 57839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57839,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57839, "pid": 5, "tid": 7, "ts": 1716454218417912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291584, "dur": 15,
    "args": {
      "External id": 57839,
      "cbid": 211, "correlation": 57839
    }
  },
  {
    "ph": "s", "id": 57839, "pid": 76337, "tid": -914061504, "ts": 1716454218291584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218417948, "dur": 52,
    "args": {
      "External id": 57859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57859,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 57859, "pid": 5, "tid": 7, "ts": 1716454218417948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291677, "dur": 13,
    "args": {
      "External id": 57859,
      "cbid": 211, "correlation": 57859
    }
  },
  {
    "ph": "s", "id": 57859, "pid": 76337, "tid": -914061504, "ts": 1716454218291677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218418002, "dur": 4,
    "args": {
      "External id": 57871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57871,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 57871, "pid": 5, "tid": 7, "ts": 1716454218418002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291701, "dur": 6,
    "args": {
      "External id": 57871,
      "cbid": 211, "correlation": 57871
    }
  },
  {
    "ph": "s", "id": 57871, "pid": 76337, "tid": -914061504, "ts": 1716454218291701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218418007, "dur": 56,
    "args": {
      "External id": 57874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57874,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57874, "pid": 5, "tid": 7, "ts": 1716454218418007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291721, "dur": 7,
    "args": {
      "External id": 57874,
      "cbid": 211, "correlation": 57874
    }
  },
  {
    "ph": "s", "id": 57874, "pid": 76337, "tid": -914061504, "ts": 1716454218291721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291779, "dur": 0,
    "args": {
      "External id": 57885,
      "cbid": 317, "correlation": 57885
    }
  },
  {
    "ph": "f", "id": 57885, "pid": 76337, "tid": -914061504, "ts": 1716454218291779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291780, "dur": 0,
    "args": {
      "External id": 57886,
      "cbid": 203, "correlation": 57886
    }
  },
  {
    "ph": "f", "id": 57886, "pid": 76337, "tid": -914061504, "ts": 1716454218291780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291781, "dur": 0,
    "args": {
      "External id": 57887,
      "cbid": 205, "correlation": 57887
    }
  },
  {
    "ph": "f", "id": 57887, "pid": 76337, "tid": -914061504, "ts": 1716454218291781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291802, "dur": 1,
    "args": {
      "External id": 57891,
      "cbid": 251, "correlation": 57891
    }
  },
  {
    "ph": "f", "id": 57891, "pid": 76337, "tid": -914061504, "ts": 1716454218291802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291804, "dur": 0,
    "args": {
      "External id": 57892,
      "cbid": 251, "correlation": 57892
    }
  },
  {
    "ph": "f", "id": 57892, "pid": 76337, "tid": -914061504, "ts": 1716454218291804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291805, "dur": 0,
    "args": {
      "External id": 57893,
      "cbid": 251, "correlation": 57893
    }
  },
  {
    "ph": "f", "id": 57893, "pid": 76337, "tid": -914061504, "ts": 1716454218291805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291806, "dur": 0,
    "args": {
      "External id": 57894,
      "cbid": 251, "correlation": 57894
    }
  },
  {
    "ph": "f", "id": 57894, "pid": 76337, "tid": -914061504, "ts": 1716454218291806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291806, "dur": 0,
    "args": {
      "External id": 57895,
      "cbid": 251, "correlation": 57895
    }
  },
  {
    "ph": "f", "id": 57895, "pid": 76337, "tid": -914061504, "ts": 1716454218291806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291807, "dur": 0,
    "args": {
      "External id": 57896,
      "cbid": 251, "correlation": 57896
    }
  },
  {
    "ph": "f", "id": 57896, "pid": 76337, "tid": -914061504, "ts": 1716454218291807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291808, "dur": 0,
    "args": {
      "External id": 57897,
      "cbid": 251, "correlation": 57897
    }
  },
  {
    "ph": "f", "id": 57897, "pid": 76337, "tid": -914061504, "ts": 1716454218291808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291809, "dur": 0,
    "args": {
      "External id": 57898,
      "cbid": 251, "correlation": 57898
    }
  },
  {
    "ph": "f", "id": 57898, "pid": 76337, "tid": -914061504, "ts": 1716454218291809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291810, "dur": 0,
    "args": {
      "External id": 57899,
      "cbid": 251, "correlation": 57899
    }
  },
  {
    "ph": "f", "id": 57899, "pid": 76337, "tid": -914061504, "ts": 1716454218291810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218418064, "dur": 115,
    "args": {
      "External id": 57900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57900,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57900, "pid": 5, "tid": 7, "ts": 1716454218418064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291812, "dur": 12,
    "args": {
      "External id": 57900,
      "cbid": 211, "correlation": 57900
    }
  },
  {
    "ph": "s", "id": 57900, "pid": 76337, "tid": -914061504, "ts": 1716454218291812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218418180, "dur": 59,
    "args": {
      "External id": 57906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57906,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57906, "pid": 5, "tid": 7, "ts": 1716454218418180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291847, "dur": 9,
    "args": {
      "External id": 57906,
      "cbid": 211, "correlation": 57906
    }
  },
  {
    "ph": "s", "id": 57906, "pid": 76337, "tid": -914061504, "ts": 1716454218291847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218418240, "dur": 618,
    "args": {
      "External id": 57915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57915,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57915, "pid": 5, "tid": 7, "ts": 1716454218418240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218291947, "dur": 16,
    "args": {
      "External id": 57915,
      "cbid": 211, "correlation": 57915
    }
  },
  {
    "ph": "s", "id": 57915, "pid": 76337, "tid": -914061504, "ts": 1716454218291947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218418860, "dur": 180,
    "args": {
      "External id": 57937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57937,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 57937, "pid": 5, "tid": 7, "ts": 1716454218418860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292026, "dur": 47,
    "args": {
      "External id": 57937,
      "cbid": 211, "correlation": 57937
    }
  },
  {
    "ph": "s", "id": 57937, "pid": 76337, "tid": -914061504, "ts": 1716454218292026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292178, "dur": 2,
    "args": {
      "External id": 57948,
      "cbid": 251, "correlation": 57948
    }
  },
  {
    "ph": "f", "id": 57948, "pid": 76337, "tid": -914061504, "ts": 1716454218292178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218419041, "dur": 197,
    "args": {
      "External id": 57949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57949,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57949, "pid": 5, "tid": 7, "ts": 1716454218419041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292184, "dur": 17,
    "args": {
      "External id": 57949,
      "cbid": 211, "correlation": 57949
    }
  },
  {
    "ph": "s", "id": 57949, "pid": 76337, "tid": -914061504, "ts": 1716454218292184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292259, "dur": 1,
    "args": {
      "External id": 57960,
      "cbid": 251, "correlation": 57960
    }
  },
  {
    "ph": "f", "id": 57960, "pid": 76337, "tid": -914061504, "ts": 1716454218292259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218419240, "dur": 187,
    "args": {
      "External id": 57961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57961,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57961, "pid": 5, "tid": 7, "ts": 1716454218419240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292263, "dur": 12,
    "args": {
      "External id": 57961,
      "cbid": 211, "correlation": 57961
    }
  },
  {
    "ph": "s", "id": 57961, "pid": 76337, "tid": -914061504, "ts": 1716454218292263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292326, "dur": 1,
    "args": {
      "External id": 57972,
      "cbid": 251, "correlation": 57972
    }
  },
  {
    "ph": "f", "id": 57972, "pid": 76337, "tid": -914061504, "ts": 1716454218292326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218419428, "dur": 188,
    "args": {
      "External id": 57973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57973,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 57973, "pid": 5, "tid": 7, "ts": 1716454218419428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292330, "dur": 11,
    "args": {
      "External id": 57973,
      "cbid": 211, "correlation": 57973
    }
  },
  {
    "ph": "s", "id": 57973, "pid": 76337, "tid": -914061504, "ts": 1716454218292330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218419617, "dur": 18515,
    "args": {
      "External id": 57994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 57994,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 57994, "pid": 5, "tid": 7, "ts": 1716454218419617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292434, "dur": 15,
    "args": {
      "External id": 57994,
      "cbid": 211, "correlation": 57994
    }
  },
  {
    "ph": "s", "id": 57994, "pid": 76337, "tid": -914061504, "ts": 1716454218292434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292554, "dur": 2,
    "args": {
      "External id": 58012,
      "cbid": 251, "correlation": 58012
    }
  },
  {
    "ph": "f", "id": 58012, "pid": 76337, "tid": -914061504, "ts": 1716454218292554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218438133, "dur": 201,
    "args": {
      "External id": 58014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58014,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 58014, "pid": 5, "tid": 7, "ts": 1716454218438133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292560, "dur": 13,
    "args": {
      "External id": 58014,
      "cbid": 211, "correlation": 58014
    }
  },
  {
    "ph": "s", "id": 58014, "pid": 76337, "tid": -914061504, "ts": 1716454218292560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218438336, "dur": 66,
    "args": {
      "External id": 58022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58022,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58022, "pid": 5, "tid": 7, "ts": 1716454218438336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292630, "dur": 12,
    "args": {
      "External id": 58022,
      "cbid": 211, "correlation": 58022
    }
  },
  {
    "ph": "s", "id": 58022, "pid": 76337, "tid": -914061504, "ts": 1716454218292630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218438404, "dur": 97,
    "args": {
      "External id": 58030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58030,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58030, "pid": 5, "tid": 7, "ts": 1716454218438404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292674, "dur": 10,
    "args": {
      "External id": 58030,
      "cbid": 211, "correlation": 58030
    }
  },
  {
    "ph": "s", "id": 58030, "pid": 76337, "tid": -914061504, "ts": 1716454218292674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218438502, "dur": 54,
    "args": {
      "External id": 58041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58041,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58041, "pid": 5, "tid": 7, "ts": 1716454218438502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292765, "dur": 13,
    "args": {
      "External id": 58041,
      "cbid": 211, "correlation": 58041
    }
  },
  {
    "ph": "s", "id": 58041, "pid": 76337, "tid": -914061504, "ts": 1716454218292765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218438557, "dur": 91,
    "args": {
      "External id": 58063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58063,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58063, "pid": 5, "tid": 7, "ts": 1716454218438557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292800, "dur": 9,
    "args": {
      "External id": 58063,
      "cbid": 211, "correlation": 58063
    }
  },
  {
    "ph": "s", "id": 58063, "pid": 76337, "tid": -914061504, "ts": 1716454218292800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292885, "dur": 1,
    "args": {
      "External id": 58074,
      "cbid": 251, "correlation": 58074
    }
  },
  {
    "ph": "f", "id": 58074, "pid": 76337, "tid": -914061504, "ts": 1716454218292885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218438650, "dur": 104,
    "args": {
      "External id": 58075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58075,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 58075, "pid": 5, "tid": 7, "ts": 1716454218438650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292890, "dur": 13,
    "args": {
      "External id": 58075,
      "cbid": 211, "correlation": 58075
    }
  },
  {
    "ph": "s", "id": 58075, "pid": 76337, "tid": -914061504, "ts": 1716454218292890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292970, "dur": 1,
    "args": {
      "External id": 58086,
      "cbid": 251, "correlation": 58086
    }
  },
  {
    "ph": "f", "id": 58086, "pid": 76337, "tid": -914061504, "ts": 1716454218292970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292982, "dur": 0,
    "args": {
      "External id": 58087,
      "cbid": 251, "correlation": 58087
    }
  },
  {
    "ph": "f", "id": 58087, "pid": 76337, "tid": -914061504, "ts": 1716454218292982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218438755, "dur": 10,
    "args": {
      "External id": 58088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58088,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58088, "pid": 5, "tid": 7, "ts": 1716454218438755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218292984, "dur": 13,
    "args": {
      "External id": 58088,
      "cbid": 211, "correlation": 58088
    }
  },
  {
    "ph": "s", "id": 58088, "pid": 76337, "tid": -914061504, "ts": 1716454218292984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218438767, "dur": 5,
    "args": {
      "External id": 58090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58090,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 58090, "pid": 5, "tid": 7, "ts": 1716454218438767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293000, "dur": 11,
    "args": {
      "External id": 58090,
      "cbid": 211, "correlation": 58090
    }
  },
  {
    "ph": "s", "id": 58090, "pid": 76337, "tid": -914061504, "ts": 1716454218293000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293069, "dur": 1,
    "args": {
      "External id": 58101,
      "cbid": 251, "correlation": 58101
    }
  },
  {
    "ph": "f", "id": 58101, "pid": 76337, "tid": -914061504, "ts": 1716454218293069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293072, "dur": 0,
    "args": {
      "External id": 58102,
      "cbid": 251, "correlation": 58102
    }
  },
  {
    "ph": "f", "id": 58102, "pid": 76337, "tid": -914061504, "ts": 1716454218293072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218438773, "dur": 6,
    "args": {
      "External id": 58103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58103,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58103, "pid": 5, "tid": 7, "ts": 1716454218438773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293074, "dur": 14,
    "args": {
      "External id": 58103,
      "cbid": 211, "correlation": 58103
    }
  },
  {
    "ph": "s", "id": 58103, "pid": 76337, "tid": -914061504, "ts": 1716454218293074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454218438780, "dur": 3,
    "args": {
      "External id": 58105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58105,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 58105, "pid": 5, "tid": 7, "ts": 1716454218438780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293089, "dur": 6,
    "args": {
      "External id": 58105,
      "cbid": 211, "correlation": 58105
    }
  },
  {
    "ph": "s", "id": 58105, "pid": 76337, "tid": -914061504, "ts": 1716454218293089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454218438785, "dur": 155,
    "args": {
      "External id": 58126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58126,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 58126, "pid": 5, "tid": 7, "ts": 1716454218438785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293164, "dur": 12,
    "args": {
      "External id": 58126,
      "cbid": 211, "correlation": 58126
    }
  },
  {
    "ph": "s", "id": 58126, "pid": 76337, "tid": -914061504, "ts": 1716454218293164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293259, "dur": 2,
    "args": {
      "External id": 58144,
      "cbid": 251, "correlation": 58144
    }
  },
  {
    "ph": "f", "id": 58144, "pid": 76337, "tid": -914061504, "ts": 1716454218293259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218438942, "dur": 106,
    "args": {
      "External id": 58146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58146,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 58146, "pid": 5, "tid": 7, "ts": 1716454218438942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293266, "dur": 13,
    "args": {
      "External id": 58146,
      "cbid": 211, "correlation": 58146
    }
  },
  {
    "ph": "s", "id": 58146, "pid": 76337, "tid": -914061504, "ts": 1716454218293266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218439049, "dur": 35,
    "args": {
      "External id": 58154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58154,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58154, "pid": 5, "tid": 7, "ts": 1716454218439049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293336, "dur": 12,
    "args": {
      "External id": 58154,
      "cbid": 211, "correlation": 58154
    }
  },
  {
    "ph": "s", "id": 58154, "pid": 76337, "tid": -914061504, "ts": 1716454218293336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218439085, "dur": 67,
    "args": {
      "External id": 58162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58162, "pid": 5, "tid": 7, "ts": 1716454218439085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293376, "dur": 9,
    "args": {
      "External id": 58162,
      "cbid": 211, "correlation": 58162
    }
  },
  {
    "ph": "s", "id": 58162, "pid": 76337, "tid": -914061504, "ts": 1716454218293376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218439154, "dur": 92,
    "args": {
      "External id": 58184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58184,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58184, "pid": 5, "tid": 7, "ts": 1716454218439154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293431, "dur": 11,
    "args": {
      "External id": 58184,
      "cbid": 211, "correlation": 58184
    }
  },
  {
    "ph": "s", "id": 58184, "pid": 76337, "tid": -914061504, "ts": 1716454218293431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293524, "dur": 1,
    "args": {
      "External id": 58200,
      "cbid": 251, "correlation": 58200
    }
  },
  {
    "ph": "f", "id": 58200, "pid": 76337, "tid": -914061504, "ts": 1716454218293524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454218439247, "dur": 571,
    "args": {
      "External id": 58202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58202,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 58202, "pid": 5, "tid": 7, "ts": 1716454218439247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293530, "dur": 12,
    "args": {
      "External id": 58202,
      "cbid": 211, "correlation": 58202
    }
  },
  {
    "ph": "s", "id": 58202, "pid": 76337, "tid": -914061504, "ts": 1716454218293530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218439819, "dur": 241,
    "args": {
      "External id": 58210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58210, "pid": 5, "tid": 7, "ts": 1716454218439819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293612, "dur": 14,
    "args": {
      "External id": 58210,
      "cbid": 211, "correlation": 58210
    }
  },
  {
    "ph": "s", "id": 58210, "pid": 76337, "tid": -914061504, "ts": 1716454218293612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218440061, "dur": 253,
    "args": {
      "External id": 58218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58218,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58218, "pid": 5, "tid": 7, "ts": 1716454218440061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293650, "dur": 10,
    "args": {
      "External id": 58218,
      "cbid": 211, "correlation": 58218
    }
  },
  {
    "ph": "s", "id": 58218, "pid": 76337, "tid": -914061504, "ts": 1716454218293650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293735, "dur": 2,
    "args": {
      "External id": 58234,
      "cbid": 251, "correlation": 58234
    }
  },
  {
    "ph": "f", "id": 58234, "pid": 76337, "tid": -914061504, "ts": 1716454218293735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293740, "dur": 0,
    "args": {
      "External id": 58236,
      "cbid": 251, "correlation": 58236
    }
  },
  {
    "ph": "f", "id": 58236, "pid": 76337, "tid": -914061504, "ts": 1716454218293740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454218440315, "dur": 358,
    "args": {
      "External id": 58237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58237,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58237, "pid": 5, "tid": 7, "ts": 1716454218440315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293744, "dur": 13,
    "args": {
      "External id": 58237,
      "cbid": 211, "correlation": 58237
    }
  },
  {
    "ph": "s", "id": 58237, "pid": 76337, "tid": -914061504, "ts": 1716454218293744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218440675, "dur": 50,
    "args": {
      "External id": 58245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58245,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58245, "pid": 5, "tid": 7, "ts": 1716454218440675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293787, "dur": 10,
    "args": {
      "External id": 58245,
      "cbid": 211, "correlation": 58245
    }
  },
  {
    "ph": "s", "id": 58245, "pid": 76337, "tid": -914061504, "ts": 1716454218293787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218440726, "dur": 156,
    "args": {
      "External id": 58256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58256, "pid": 5, "tid": 7, "ts": 1716454218440726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293854, "dur": 39,
    "args": {
      "External id": 58256,
      "cbid": 211, "correlation": 58256
    }
  },
  {
    "ph": "s", "id": 58256, "pid": 76337, "tid": -914061504, "ts": 1716454218293854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293947, "dur": 0,
    "args": {
      "External id": 58268,
      "cbid": 317, "correlation": 58268
    }
  },
  {
    "ph": "f", "id": 58268, "pid": 76337, "tid": -914061504, "ts": 1716454218293947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293948, "dur": 0,
    "args": {
      "External id": 58269,
      "cbid": 203, "correlation": 58269
    }
  },
  {
    "ph": "f", "id": 58269, "pid": 76337, "tid": -914061504, "ts": 1716454218293948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293949, "dur": 0,
    "args": {
      "External id": 58270,
      "cbid": 205, "correlation": 58270
    }
  },
  {
    "ph": "f", "id": 58270, "pid": 76337, "tid": -914061504, "ts": 1716454218293949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293981, "dur": 1,
    "args": {
      "External id": 58274,
      "cbid": 251, "correlation": 58274
    }
  },
  {
    "ph": "f", "id": 58274, "pid": 76337, "tid": -914061504, "ts": 1716454218293981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293983, "dur": 0,
    "args": {
      "External id": 58275,
      "cbid": 251, "correlation": 58275
    }
  },
  {
    "ph": "f", "id": 58275, "pid": 76337, "tid": -914061504, "ts": 1716454218293983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293984, "dur": 0,
    "args": {
      "External id": 58276,
      "cbid": 251, "correlation": 58276
    }
  },
  {
    "ph": "f", "id": 58276, "pid": 76337, "tid": -914061504, "ts": 1716454218293984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293985, "dur": 0,
    "args": {
      "External id": 58277,
      "cbid": 251, "correlation": 58277
    }
  },
  {
    "ph": "f", "id": 58277, "pid": 76337, "tid": -914061504, "ts": 1716454218293985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293985, "dur": 0,
    "args": {
      "External id": 58278,
      "cbid": 251, "correlation": 58278
    }
  },
  {
    "ph": "f", "id": 58278, "pid": 76337, "tid": -914061504, "ts": 1716454218293985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293986, "dur": 0,
    "args": {
      "External id": 58279,
      "cbid": 251, "correlation": 58279
    }
  },
  {
    "ph": "f", "id": 58279, "pid": 76337, "tid": -914061504, "ts": 1716454218293986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293987, "dur": 0,
    "args": {
      "External id": 58280,
      "cbid": 251, "correlation": 58280
    }
  },
  {
    "ph": "f", "id": 58280, "pid": 76337, "tid": -914061504, "ts": 1716454218293987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293988, "dur": 0,
    "args": {
      "External id": 58281,
      "cbid": 251, "correlation": 58281
    }
  },
  {
    "ph": "f", "id": 58281, "pid": 76337, "tid": -914061504, "ts": 1716454218293988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293989, "dur": 0,
    "args": {
      "External id": 58282,
      "cbid": 251, "correlation": 58282
    }
  },
  {
    "ph": "f", "id": 58282, "pid": 76337, "tid": -914061504, "ts": 1716454218293989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454218440884, "dur": 113,
    "args": {
      "External id": 58283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58283,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 58283, "pid": 5, "tid": 7, "ts": 1716454218440884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218293992, "dur": 26,
    "args": {
      "External id": 58283,
      "cbid": 211, "correlation": 58283
    }
  },
  {
    "ph": "s", "id": 58283, "pid": 76337, "tid": -914061504, "ts": 1716454218293992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218440998, "dur": 59,
    "args": {
      "External id": 58289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58289, "pid": 5, "tid": 7, "ts": 1716454218440998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294042, "dur": 272,
    "args": {
      "External id": 58289,
      "cbid": 211, "correlation": 58289
    }
  },
  {
    "ph": "s", "id": 58289, "pid": 76337, "tid": -914061504, "ts": 1716454218294042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218441059, "dur": 50,
    "args": {
      "External id": 58297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58297,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58297, "pid": 5, "tid": 7, "ts": 1716454218441059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294338, "dur": 9,
    "args": {
      "External id": 58297,
      "cbid": 211, "correlation": 58297
    }
  },
  {
    "ph": "s", "id": 58297, "pid": 76337, "tid": -914061504, "ts": 1716454218294338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454218441110, "dur": 52,
    "args": {
      "External id": 58317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58317,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 58317, "pid": 5, "tid": 7, "ts": 1716454218441110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294423, "dur": 12,
    "args": {
      "External id": 58317,
      "cbid": 211, "correlation": 58317
    }
  },
  {
    "ph": "s", "id": 58317, "pid": 76337, "tid": -914061504, "ts": 1716454218294423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454218441163, "dur": 4,
    "args": {
      "External id": 58329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58329,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 58329, "pid": 5, "tid": 7, "ts": 1716454218441163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294446, "dur": 7,
    "args": {
      "External id": 58329,
      "cbid": 211, "correlation": 58329
    }
  },
  {
    "ph": "s", "id": 58329, "pid": 76337, "tid": -914061504, "ts": 1716454218294446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218441168, "dur": 56,
    "args": {
      "External id": 58332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58332,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58332, "pid": 5, "tid": 7, "ts": 1716454218441168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294464, "dur": 101,
    "args": {
      "External id": 58332,
      "cbid": 211, "correlation": 58332
    }
  },
  {
    "ph": "s", "id": 58332, "pid": 76337, "tid": -914061504, "ts": 1716454218294464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218441226, "dur": 37,
    "args": {
      "External id": 58341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58341,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58341, "pid": 5, "tid": 7, "ts": 1716454218441226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294604, "dur": 11,
    "args": {
      "External id": 58341,
      "cbid": 211, "correlation": 58341
    }
  },
  {
    "ph": "s", "id": 58341, "pid": 76337, "tid": -914061504, "ts": 1716454218294604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294661, "dur": 0,
    "args": {
      "External id": 58351,
      "cbid": 317, "correlation": 58351
    }
  },
  {
    "ph": "f", "id": 58351, "pid": 76337, "tid": -914061504, "ts": 1716454218294661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294662, "dur": 0,
    "args": {
      "External id": 58352,
      "cbid": 203, "correlation": 58352
    }
  },
  {
    "ph": "f", "id": 58352, "pid": 76337, "tid": -914061504, "ts": 1716454218294662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294663, "dur": 0,
    "args": {
      "External id": 58353,
      "cbid": 205, "correlation": 58353
    }
  },
  {
    "ph": "f", "id": 58353, "pid": 76337, "tid": -914061504, "ts": 1716454218294663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218441264, "dur": 41,
    "args": {
      "External id": 58357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58357,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58357, "pid": 5, "tid": 7, "ts": 1716454218441264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294683, "dur": 12,
    "args": {
      "External id": 58357,
      "cbid": 211, "correlation": 58357
    }
  },
  {
    "ph": "s", "id": 58357, "pid": 76337, "tid": -914061504, "ts": 1716454218294683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454218441306, "dur": 3,
    "args": {
      "External id": 58359,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58359,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.417138237671786
    }
  },
  {
    "ph": "f", "id": 58359, "pid": 5, "tid": 7, "ts": 1716454218441306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294699, "dur": 24,
    "args": {
      "External id": 58359,
      "cbid": 51, "correlation": 58359
    }
  },
  {
    "ph": "s", "id": 58359, "pid": 76337, "tid": -914061504, "ts": 1716454218294699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294728, "dur": 1,
    "args": {
      "External id": 58361,
      "cbid": 200, "correlation": 58361
    }
  },
  {
    "ph": "f", "id": 58361, "pid": 76337, "tid": -914061504, "ts": 1716454218294728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294730, "dur": 0,
    "args": {
      "External id": 58362,
      "cbid": 200, "correlation": 58362
    }
  },
  {
    "ph": "f", "id": 58362, "pid": 76337, "tid": -914061504, "ts": 1716454218294730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294731, "dur": 0,
    "args": {
      "External id": 58363,
      "cbid": 200, "correlation": 58363
    }
  },
  {
    "ph": "f", "id": 58363, "pid": 76337, "tid": -914061504, "ts": 1716454218294731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294731, "dur": 0,
    "args": {
      "External id": 58364,
      "cbid": 200, "correlation": 58364
    }
  },
  {
    "ph": "f", "id": 58364, "pid": 76337, "tid": -914061504, "ts": 1716454218294731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294732, "dur": 4,
    "args": {
      "External id": 58365,
      "cbid": 15, "correlation": 58365
    }
  },
  {
    "ph": "f", "id": 58365, "pid": 76337, "tid": -914061504, "ts": 1716454218294732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294736, "dur": 1,
    "args": {
      "External id": 58366,
      "cbid": 251, "correlation": 58366
    }
  },
  {
    "ph": "f", "id": 58366, "pid": 76337, "tid": -914061504, "ts": 1716454218294736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454218441311, "dur": 24,
    "args": {
      "External id": 58367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58367,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58367, "pid": 5, "tid": 7, "ts": 1716454218441311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294740, "dur": 8,
    "args": {
      "External id": 58367,
      "cbid": 211, "correlation": 58367
    }
  },
  {
    "ph": "s", "id": 58367, "pid": 76337, "tid": -914061504, "ts": 1716454218294740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454218441336, "dur": 4,
    "args": {
      "External id": 58369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58369,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 58369, "pid": 5, "tid": 7, "ts": 1716454218441336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294754, "dur": 7,
    "args": {
      "External id": 58369,
      "cbid": 211, "correlation": 58369
    }
  },
  {
    "ph": "s", "id": 58369, "pid": 76337, "tid": -914061504, "ts": 1716454218294754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294765, "dur": 0,
    "args": {
      "External id": 58370,
      "cbid": 51, "correlation": 58370
    }
  },
  {
    "ph": "s", "id": 58370, "pid": 76337, "tid": -914061504, "ts": 1716454218294765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454218441341, "dur": 187,
    "args": {
      "External id": 58371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58371,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58371, "pid": 5, "tid": 7, "ts": 1716454218441341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294766, "dur": 179,
    "args": {
      "External id": 58371,
      "cbid": 211, "correlation": 58371
    }
  },
  {
    "ph": "s", "id": 58371, "pid": 76337, "tid": -914061504, "ts": 1716454218294766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454218441530, "dur": 7,
    "args": {
      "External id": 58372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58372,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58372, "pid": 5, "tid": 7, "ts": 1716454218441530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294949, "dur": 6,
    "args": {
      "External id": 58372,
      "cbid": 211, "correlation": 58372
    }
  },
  {
    "ph": "s", "id": 58372, "pid": 76337, "tid": -914061504, "ts": 1716454218294949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454218441538, "dur": 5,
    "args": {
      "External id": 58378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58378,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 58378, "pid": 5, "tid": 7, "ts": 1716454218441538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218294988, "dur": 10,
    "args": {
      "External id": 58378,
      "cbid": 211, "correlation": 58378
    }
  },
  {
    "ph": "s", "id": 58378, "pid": 76337, "tid": -914061504, "ts": 1716454218294988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218441544, "dur": 3,
    "args": {
      "External id": 58386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58386,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58386, "pid": 5, "tid": 7, "ts": 1716454218441544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218296720, "dur": 15,
    "args": {
      "External id": 58386,
      "cbid": 211, "correlation": 58386
    }
  },
  {
    "ph": "s", "id": 58386, "pid": 76337, "tid": -914061504, "ts": 1716454218296720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218441548, "dur": 3,
    "args": {
      "External id": 58394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58394,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58394, "pid": 5, "tid": 7, "ts": 1716454218441548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218296760, "dur": 13,
    "args": {
      "External id": 58394,
      "cbid": 211, "correlation": 58394
    }
  },
  {
    "ph": "s", "id": 58394, "pid": 76337, "tid": -914061504, "ts": 1716454218296760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218441553, "dur": 3,
    "args": {
      "External id": 58402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58402,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58402, "pid": 5, "tid": 7, "ts": 1716454218441553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218296791, "dur": 8,
    "args": {
      "External id": 58402,
      "cbid": 211, "correlation": 58402
    }
  },
  {
    "ph": "s", "id": 58402, "pid": 76337, "tid": -914061504, "ts": 1716454218296791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218441556, "dur": 3,
    "args": {
      "External id": 58411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58411,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58411, "pid": 5, "tid": 7, "ts": 1716454218441556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218296972, "dur": 23,
    "args": {
      "External id": 58411,
      "cbid": 211, "correlation": 58411
    }
  },
  {
    "ph": "s", "id": 58411, "pid": 76337, "tid": -914061504, "ts": 1716454218296972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218441561, "dur": 3,
    "args": {
      "External id": 58420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58420,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58420, "pid": 5, "tid": 7, "ts": 1716454218441561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218297012, "dur": 8,
    "args": {
      "External id": 58420,
      "cbid": 211, "correlation": 58420
    }
  },
  {
    "ph": "s", "id": 58420, "pid": 76337, "tid": -914061504, "ts": 1716454218297012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218441565, "dur": 3,
    "args": {
      "External id": 58428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58428,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58428, "pid": 5, "tid": 7, "ts": 1716454218441565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218297038, "dur": 8,
    "args": {
      "External id": 58428,
      "cbid": 211, "correlation": 58428
    }
  },
  {
    "ph": "s", "id": 58428, "pid": 76337, "tid": -914061504, "ts": 1716454218297038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454218441569, "dur": 3,
    "args": {
      "External id": 58436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58436,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58436, "pid": 5, "tid": 7, "ts": 1716454218441569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218297293, "dur": 15,
    "args": {
      "External id": 58436,
      "cbid": 211, "correlation": 58436
    }
  },
  {
    "ph": "s", "id": 58436, "pid": 76337, "tid": -914061504, "ts": 1716454218297293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454218441574, "dur": 3,
    "args": {
      "External id": 58444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58444,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58444, "pid": 5, "tid": 7, "ts": 1716454218441574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218297324, "dur": 8,
    "args": {
      "External id": 58444,
      "cbid": 211, "correlation": 58444
    }
  },
  {
    "ph": "s", "id": 58444, "pid": 76337, "tid": -914061504, "ts": 1716454218297324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218441579, "dur": 1,
    "args": {
      "External id": 58454,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58454,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 58454, "pid": 5, "tid": 7, "ts": 1716454218441579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218297390, "dur": 40,
    "args": {
      "External id": 58454,
      "cbid": 41, "correlation": 58454
    }
  },
  {
    "ph": "s", "id": 58454, "pid": 76337, "tid": -914061504, "ts": 1716454218297390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218297432, "dur": 144164,
    "args": {
      "External id": 58455,
      "cbid": 131, "correlation": 58455
    }
  },
  {
    "ph": "f", "id": 58455, "pid": 76337, "tid": -914061504, "ts": 1716454218297432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454218441726, "dur": 3,
    "args": {
      "External id": 58463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58463, "pid": 5, "tid": 7, "ts": 1716454218441726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218441707, "dur": 22,
    "args": {
      "External id": 58463,
      "cbid": 211, "correlation": 58463
    }
  },
  {
    "ph": "s", "id": 58463, "pid": 76337, "tid": -914061504, "ts": 1716454218441707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218441813, "dur": 3,
    "args": {
      "External id": 58472,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58472,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 58472, "pid": 5, "tid": 7, "ts": 1716454218441813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218441787, "dur": 26,
    "args": {
      "External id": 58472,
      "cbid": 41, "correlation": 58472
    }
  },
  {
    "ph": "s", "id": 58472, "pid": 76337, "tid": -914061504, "ts": 1716454218441787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454218441897, "dur": 4,
    "args": {
      "External id": 58482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58482,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58482, "pid": 5, "tid": 7, "ts": 1716454218441897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454218441881, "dur": 17,
    "args": {
      "External id": 58482,
      "cbid": 211, "correlation": 58482
    }
  },
  {
    "ph": "s", "id": 58482, "pid": 76337, "tid": -914061504, "ts": 1716454218441881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218441999, "dur": 1,
    "args": {
      "External id": 58492,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58492,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 58492, "pid": 5, "tid": 7, "ts": 1716454218441999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218441939, "dur": 57,
    "args": {
      "External id": 58492,
      "cbid": 41, "correlation": 58492
    }
  },
  {
    "ph": "s", "id": 58492, "pid": 76337, "tid": -914061504, "ts": 1716454218441939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454218441997, "dur": 9,
    "args": {
      "External id": 58493,
      "cbid": 131, "correlation": 58493
    }
  },
  {
    "ph": "f", "id": 58493, "pid": 76337, "tid": -914061504, "ts": 1716454218441997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218442096, "dur": 3,
    "args": {
      "External id": 58500,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58500,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.415841584158414
    }
  },
  {
    "ph": "f", "id": 58500, "pid": 5, "tid": 7, "ts": 1716454218442096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218442047, "dur": 48,
    "args": {
      "External id": 58500,
      "cbid": 41, "correlation": 58500
    }
  },
  {
    "ph": "s", "id": 58500, "pid": 76337, "tid": -914061504, "ts": 1716454218442047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454218442414, "dur": 3,
    "args": {
      "External id": 58519,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58519,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005320917858330562
    }
  },
  {
    "ph": "f", "id": 58519, "pid": 5, "tid": 7, "ts": 1716454218442414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454218442382, "dur": 31,
    "args": {
      "External id": 58519,
      "cbid": 41, "correlation": 58519
    }
  },
  {
    "ph": "s", "id": 58519, "pid": 76337, "tid": -914061504, "ts": 1716454218442382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceCount", "pid": 76337, "tid": -914061504,
    "ts": 1716454218447342, "dur": 1,
    "args": {
      "External id": 58522,
      "cbid": 3, "correlation": 58522
    }
  },
  {
    "ph": "f", "id": 58522, "pid": 76337, "tid": -914061504, "ts": 1716454218447342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454218447345, "dur": 1116947,
    "args": {
      "External id": 58523,
      "cbid": 4, "correlation": 58523
    }
  },
  {
    "ph": "f", "id": 58523, "pid": 76337, "tid": -914061504, "ts": 1716454218447345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454219564295, "dur": 72066,
    "args": {
      "External id": 58524,
      "cbid": 4, "correlation": 58524
    }
  },
  {
    "ph": "f", "id": 58524, "pid": 76337, "tid": -914061504, "ts": 1716454219564295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454219636363, "dur": 71462,
    "args": {
      "External id": 58525,
      "cbid": 4, "correlation": 58525
    }
  },
  {
    "ph": "f", "id": 58525, "pid": 76337, "tid": -914061504, "ts": 1716454219636363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454219707827, "dur": 36547,
    "args": {
      "External id": 58526,
      "cbid": 4, "correlation": 58526
    }
  },
  {
    "ph": "f", "id": 58526, "pid": 76337, "tid": -914061504, "ts": 1716454219707827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454219744375, "dur": 3188,
    "args": {
      "External id": 58527,
      "cbid": 4, "correlation": 58527
    }
  },
  {
    "ph": "f", "id": 58527, "pid": 76337, "tid": -914061504, "ts": 1716454219744375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454219747564, "dur": 1792,
    "args": {
      "External id": 58528,
      "cbid": 4, "correlation": 58528
    }
  },
  {
    "ph": "f", "id": 58528, "pid": 76337, "tid": -914061504, "ts": 1716454219747564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaGetDeviceProperties", "pid": 76337, "tid": -914061504,
    "ts": 1716454219749357, "dur": 5312,
    "args": {
      "External id": 58529,
      "cbid": 4, "correlation": 58529
    }
  },
  {
    "ph": "f", "id": 58529, "pid": 76337, "tid": -914061504, "ts": 1716454219749357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073963, "dur": 0,
    "args": {
      "External id": 58532,
      "cbid": 200, "correlation": 58532
    }
  },
  {
    "ph": "f", "id": 58532, "pid": 76337, "tid": -914061504, "ts": 1716454222073963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073964, "dur": 0,
    "args": {
      "External id": 58533,
      "cbid": 200, "correlation": 58533
    }
  },
  {
    "ph": "f", "id": 58533, "pid": 76337, "tid": -914061504, "ts": 1716454222073964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073964, "dur": 0,
    "args": {
      "External id": 58534,
      "cbid": 200, "correlation": 58534
    }
  },
  {
    "ph": "f", "id": 58534, "pid": 76337, "tid": -914061504, "ts": 1716454222073964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073964, "dur": 0,
    "args": {
      "External id": 58535,
      "cbid": 200, "correlation": 58535
    }
  },
  {
    "ph": "f", "id": 58535, "pid": 76337, "tid": -914061504, "ts": 1716454222073964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073965, "dur": 0,
    "args": {
      "External id": 58536,
      "cbid": 200, "correlation": 58536
    }
  },
  {
    "ph": "f", "id": 58536, "pid": 76337, "tid": -914061504, "ts": 1716454222073965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073965, "dur": 0,
    "args": {
      "External id": 58537,
      "cbid": 200, "correlation": 58537
    }
  },
  {
    "ph": "f", "id": 58537, "pid": 76337, "tid": -914061504, "ts": 1716454222073965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073966, "dur": 0,
    "args": {
      "External id": 58538,
      "cbid": 200, "correlation": 58538
    }
  },
  {
    "ph": "f", "id": 58538, "pid": 76337, "tid": -914061504, "ts": 1716454222073966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073966, "dur": 0,
    "args": {
      "External id": 58539,
      "cbid": 200, "correlation": 58539
    }
  },
  {
    "ph": "f", "id": 58539, "pid": 76337, "tid": -914061504, "ts": 1716454222073966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073966, "dur": 0,
    "args": {
      "External id": 58540,
      "cbid": 200, "correlation": 58540
    }
  },
  {
    "ph": "f", "id": 58540, "pid": 76337, "tid": -914061504, "ts": 1716454222073966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073967, "dur": 0,
    "args": {
      "External id": 58541,
      "cbid": 200, "correlation": 58541
    }
  },
  {
    "ph": "f", "id": 58541, "pid": 76337, "tid": -914061504, "ts": 1716454222073967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFree", "pid": 76337, "tid": -914061504,
    "ts": 1716454222073972, "dur": 12,
    "args": {
      "External id": 58542,
      "cbid": 22, "correlation": 58542
    }
  },
  {
    "ph": "f", "id": 58542, "pid": 76337, "tid": -914061504, "ts": 1716454222073972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074016, "dur": 100,
    "args": {
      "External id": 58544,
      "cbid": 20, "correlation": 58544
    }
  },
  {
    "ph": "f", "id": 58544, "pid": 76337, "tid": -914061504, "ts": 1716454222074016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074132, "dur": 9,
    "args": {
      "External id": 58553,
      "cbid": 20, "correlation": 58553
    }
  },
  {
    "ph": "f", "id": 58553, "pid": 76337, "tid": -914061504, "ts": 1716454222074132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074149, "dur": 759,
    "args": {
      "External id": 58562,
      "cbid": 20, "correlation": 58562
    }
  },
  {
    "ph": "f", "id": 58562, "pid": 76337, "tid": -914061504, "ts": 1716454222074149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFree", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074928, "dur": 0,
    "args": {
      "External id": 58565,
      "cbid": 22, "correlation": 58565
    }
  },
  {
    "ph": "f", "id": 58565, "pid": 76337, "tid": -914061504, "ts": 1716454222074928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074933, "dur": 9,
    "args": {
      "External id": 58567,
      "cbid": 20, "correlation": 58567
    }
  },
  {
    "ph": "f", "id": 58567, "pid": 76337, "tid": -914061504, "ts": 1716454222074933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074949, "dur": 11,
    "args": {
      "External id": 58576,
      "cbid": 20, "correlation": 58576
    }
  },
  {
    "ph": "f", "id": 58576, "pid": 76337, "tid": -914061504, "ts": 1716454222074949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222074968, "dur": 766,
    "args": {
      "External id": 58585,
      "cbid": 20, "correlation": 58585
    }
  },
  {
    "ph": "f", "id": 58585, "pid": 76337, "tid": -914061504, "ts": 1716454222074968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamCreateWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222075753, "dur": 69713,
    "args": {
      "External id": 58588,
      "cbid": 198, "correlation": 58588
    }
  },
  {
    "ph": "f", "id": 58588, "pid": 76337, "tid": -914061504, "ts": 1716454222075753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFree", "pid": 76337, "tid": -914061504,
    "ts": 1716454222145492, "dur": 4,
    "args": {
      "External id": 58590,
      "cbid": 22, "correlation": 58590
    }
  },
  {
    "ph": "f", "id": 58590, "pid": 76337, "tid": -914061504, "ts": 1716454222145492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222145520, "dur": 27,
    "args": {
      "External id": 58592,
      "cbid": 20, "correlation": 58592
    }
  },
  {
    "ph": "f", "id": 58592, "pid": 76337, "tid": -914061504, "ts": 1716454222145520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222145558, "dur": 8,
    "args": {
      "External id": 58601,
      "cbid": 20, "correlation": 58601
    }
  },
  {
    "ph": "f", "id": 58601, "pid": 76337, "tid": -914061504, "ts": 1716454222145558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222145571, "dur": 742,
    "args": {
      "External id": 58610,
      "cbid": 20, "correlation": 58610
    }
  },
  {
    "ph": "f", "id": 58610, "pid": 76337, "tid": -914061504, "ts": 1716454222145571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamCreateWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222146324, "dur": 16,
    "args": {
      "External id": 58613,
      "cbid": 198, "correlation": 58613
    }
  },
  {
    "ph": "f", "id": 58613, "pid": 76337, "tid": -914061504, "ts": 1716454222146324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFree", "pid": 76337, "tid": -914061504,
    "ts": 1716454222146342, "dur": 0,
    "args": {
      "External id": 58615,
      "cbid": 22, "correlation": 58615
    }
  },
  {
    "ph": "f", "id": 58615, "pid": 76337, "tid": -914061504, "ts": 1716454222146342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222146346, "dur": 8,
    "args": {
      "External id": 58617,
      "cbid": 20, "correlation": 58617
    }
  },
  {
    "ph": "f", "id": 58617, "pid": 76337, "tid": -914061504, "ts": 1716454222146346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222146360, "dur": 6,
    "args": {
      "External id": 58626,
      "cbid": 20, "correlation": 58626
    }
  },
  {
    "ph": "f", "id": 58626, "pid": 76337, "tid": -914061504, "ts": 1716454222146360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454222146371, "dur": 530,
    "args": {
      "External id": 58635,
      "cbid": 20, "correlation": 58635
    }
  },
  {
    "ph": "f", "id": 58635, "pid": 76337, "tid": -914061504, "ts": 1716454222146371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamCreateWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222146911, "dur": 464,
    "args": {
      "External id": 58638,
      "cbid": 198, "correlation": 58638
    }
  },
  {
    "ph": "f", "id": 58638, "pid": 76337, "tid": -914061504, "ts": 1716454222146911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFree", "pid": 76337, "tid": -914061504,
    "ts": 1716454222147398, "dur": 0,
    "args": {
      "External id": 58640,
      "cbid": 22, "correlation": 58640
    }
  },
  {
    "ph": "f", "id": 58640, "pid": 76337, "tid": -914061504, "ts": 1716454222147398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222147548, "dur": 3,
    "args": {
      "External id": 58643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58643, "pid": 5, "tid": 7, "ts": 1716454222147548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222147472, "dur": 79,
    "args": {
      "External id": 58643,
      "cbid": 211, "correlation": 58643
    }
  },
  {
    "ph": "s", "id": 58643, "pid": 76337, "tid": -914061504, "ts": 1716454222147472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222147589, "dur": 6,
    "args": {
      "External id": 58645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58645,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 58645, "pid": 5, "tid": 7, "ts": 1716454222147589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222147554, "dur": 37,
    "args": {
      "External id": 58645,
      "cbid": 211, "correlation": 58645
    }
  },
  {
    "ph": "s", "id": 58645, "pid": 76337, "tid": -914061504, "ts": 1716454222147554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454222147617, "dur": 3,
    "args": {
      "External id": 58647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58647, "pid": 5, "tid": 7, "ts": 1716454222147617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222147593, "dur": 23,
    "args": {
      "External id": 58647,
      "cbid": 211, "correlation": 58647
    }
  },
  {
    "ph": "s", "id": 58647, "pid": 76337, "tid": -914061504, "ts": 1716454222147593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222257417, "dur": 3,
    "args": {
      "External id": 58655,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58655,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0024271844660194173
    }
  },
  {
    "ph": "f", "id": 58655, "pid": 5, "tid": 7, "ts": 1716454222257417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222147778, "dur": 109645,
    "args": {
      "External id": 58655,
      "cbid": 41, "correlation": 58655
    }
  },
  {
    "ph": "s", "id": 58655, "pid": 76337, "tid": -914061504, "ts": 1716454222147778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222257644, "dur": 3,
    "args": {
      "External id": 58669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58669, "pid": 5, "tid": 7, "ts": 1716454222257644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222257620, "dur": 29,
    "args": {
      "External id": 58669,
      "cbid": 211, "correlation": 58669
    }
  },
  {
    "ph": "s", "id": 58669, "pid": 76337, "tid": -914061504, "ts": 1716454222257620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222257673, "dur": 2,
    "args": {
      "External id": 58683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58683, "pid": 5, "tid": 7, "ts": 1716454222257673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222257663, "dur": 9,
    "args": {
      "External id": 58683,
      "cbid": 211, "correlation": 58683
    }
  },
  {
    "ph": "s", "id": 58683, "pid": 76337, "tid": -914061504, "ts": 1716454222257663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222257961, "dur": 5,
    "args": {
      "External id": 58690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58690,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58690, "pid": 5, "tid": 7, "ts": 1716454222257961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222257772, "dur": 192,
    "args": {
      "External id": 58690,
      "cbid": 211, "correlation": 58690
    }
  },
  {
    "ph": "s", "id": 58690, "pid": 76337, "tid": -914061504, "ts": 1716454222257772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222258053, "dur": 5,
    "args": {
      "External id": 58693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58693,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58693, "pid": 5, "tid": 7, "ts": 1716454222258053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222257971, "dur": 82,
    "args": {
      "External id": 58693,
      "cbid": 211, "correlation": 58693
    }
  },
  {
    "ph": "s", "id": 58693, "pid": 76337, "tid": -914061504, "ts": 1716454222257971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222258103, "dur": 3,
    "args": {
      "External id": 58695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58695,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58695, "pid": 5, "tid": 7, "ts": 1716454222258103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258065, "dur": 39,
    "args": {
      "External id": 58695,
      "cbid": 211, "correlation": 58695
    }
  },
  {
    "ph": "s", "id": 58695, "pid": 76337, "tid": -914061504, "ts": 1716454222258065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222258159, "dur": 3,
    "args": {
      "External id": 58698,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58698,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0024271844660194173
    }
  },
  {
    "ph": "f", "id": 58698, "pid": 5, "tid": 7, "ts": 1716454222258159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258141, "dur": 18,
    "args": {
      "External id": 58698,
      "cbid": 41, "correlation": 58698
    }
  },
  {
    "ph": "s", "id": 58698, "pid": 76337, "tid": -914061504, "ts": 1716454222258141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222258357, "dur": 4,
    "args": {
      "External id": 58714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58714,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58714, "pid": 5, "tid": 7, "ts": 1716454222258357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258259, "dur": 101,
    "args": {
      "External id": 58714,
      "cbid": 211, "correlation": 58714
    }
  },
  {
    "ph": "s", "id": 58714, "pid": 76337, "tid": -914061504, "ts": 1716454222258259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222258403, "dur": 3,
    "args": {
      "External id": 58719,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58719,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0002920560747663551
    }
  },
  {
    "ph": "f", "id": 58719, "pid": 5, "tid": 7, "ts": 1716454222258403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258368, "dur": 33,
    "args": {
      "External id": 58719,
      "cbid": 41, "correlation": 58719
    }
  },
  {
    "ph": "s", "id": 58719, "pid": 76337, "tid": -914061504, "ts": 1716454222258368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222258442, "dur": 1,
    "args": {
      "External id": 58725,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58725,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005580357142857143
    }
  },
  {
    "ph": "f", "id": 58725, "pid": 5, "tid": 7, "ts": 1716454222258442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258417, "dur": 36,
    "args": {
      "External id": 58725,
      "cbid": 41, "correlation": 58725
    }
  },
  {
    "ph": "s", "id": 58725, "pid": 76337, "tid": -914061504, "ts": 1716454222258417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258454, "dur": 10,
    "args": {
      "External id": 58726,
      "cbid": 131, "correlation": 58726
    }
  },
  {
    "ph": "f", "id": 58726, "pid": 76337, "tid": -914061504, "ts": 1716454222258454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222258582, "dur": 4,
    "args": {
      "External id": 58734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58734,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58734, "pid": 5, "tid": 7, "ts": 1716454222258582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258567, "dur": 16,
    "args": {
      "External id": 58734,
      "cbid": 211, "correlation": 58734
    }
  },
  {
    "ph": "s", "id": 58734, "pid": 76337, "tid": -914061504, "ts": 1716454222258567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222258626, "dur": 3,
    "args": {
      "External id": 58744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58744,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58744, "pid": 5, "tid": 7, "ts": 1716454222258626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258616, "dur": 9,
    "args": {
      "External id": 58744,
      "cbid": 211, "correlation": 58744
    }
  },
  {
    "ph": "s", "id": 58744, "pid": 76337, "tid": -914061504, "ts": 1716454222258616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222258662, "dur": 3,
    "args": {
      "External id": 58753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58753,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58753, "pid": 5, "tid": 7, "ts": 1716454222258662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258652, "dur": 9,
    "args": {
      "External id": 58753,
      "cbid": 211, "correlation": 58753
    }
  },
  {
    "ph": "s", "id": 58753, "pid": 76337, "tid": -914061504, "ts": 1716454222258652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222259328, "dur": 9,
    "args": {
      "External id": 58763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58763,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58763, "pid": 5, "tid": 7, "ts": 1716454222259328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222258823, "dur": 513,
    "args": {
      "External id": 58763,
      "cbid": 211, "correlation": 58763
    }
  },
  {
    "ph": "s", "id": 58763, "pid": 76337, "tid": -914061504, "ts": 1716454222258823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222259387, "dur": 3,
    "args": {
      "External id": 58771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58771,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58771, "pid": 5, "tid": 7, "ts": 1716454222259387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222259376, "dur": 11,
    "args": {
      "External id": 58771,
      "cbid": 211, "correlation": 58771
    }
  },
  {
    "ph": "s", "id": 58771, "pid": 76337, "tid": -914061504, "ts": 1716454222259376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222259576, "dur": 8,
    "args": {
      "External id": 58781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58781,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58781, "pid": 5, "tid": 7, "ts": 1716454222259576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222259432, "dur": 147,
    "args": {
      "External id": 58781,
      "cbid": 211, "correlation": 58781
    }
  },
  {
    "ph": "s", "id": 58781, "pid": 76337, "tid": -914061504, "ts": 1716454222259432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222260432, "dur": 7,
    "args": {
      "External id": 58789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58789,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58789, "pid": 5, "tid": 7, "ts": 1716454222260432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222259615, "dur": 818,
    "args": {
      "External id": 58789,
      "cbid": 211, "correlation": 58789
    }
  },
  {
    "ph": "s", "id": 58789, "pid": 76337, "tid": -914061504, "ts": 1716454222259615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222260474, "dur": 3,
    "args": {
      "External id": 58798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58798,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58798, "pid": 5, "tid": 7, "ts": 1716454222260474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222260463, "dur": 10,
    "args": {
      "External id": 58798,
      "cbid": 211, "correlation": 58798
    }
  },
  {
    "ph": "s", "id": 58798, "pid": 76337, "tid": -914061504, "ts": 1716454222260463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222260501, "dur": 5,
    "args": {
      "External id": 58807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58807,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58807, "pid": 5, "tid": 7, "ts": 1716454222260501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222260491, "dur": 8,
    "args": {
      "External id": 58807,
      "cbid": 211, "correlation": 58807
    }
  },
  {
    "ph": "s", "id": 58807, "pid": 76337, "tid": -914061504, "ts": 1716454222260491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222260555, "dur": 9,
    "args": {
      "External id": 58817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58817,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58817, "pid": 5, "tid": 7, "ts": 1716454222260555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222260543, "dur": 13,
    "args": {
      "External id": 58817,
      "cbid": 211, "correlation": 58817
    }
  },
  {
    "ph": "s", "id": 58817, "pid": 76337, "tid": -914061504, "ts": 1716454222260543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222260985, "dur": 3,
    "args": {
      "External id": 58826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58826,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58826, "pid": 5, "tid": 7, "ts": 1716454222260985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222260960, "dur": 25,
    "args": {
      "External id": 58826,
      "cbid": 211, "correlation": 58826
    }
  },
  {
    "ph": "s", "id": 58826, "pid": 76337, "tid": -914061504, "ts": 1716454222260960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222261020, "dur": 3,
    "args": {
      "External id": 58834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58834,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58834, "pid": 5, "tid": 7, "ts": 1716454222261020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261007, "dur": 13,
    "args": {
      "External id": 58834,
      "cbid": 211, "correlation": 58834
    }
  },
  {
    "ph": "s", "id": 58834, "pid": 76337, "tid": -914061504, "ts": 1716454222261007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222261103, "dur": 1,
    "args": {
      "External id": 58844,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58844,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005
    }
  },
  {
    "ph": "f", "id": 58844, "pid": 5, "tid": 7, "ts": 1716454222261103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261082, "dur": 19,
    "args": {
      "External id": 58844,
      "cbid": 41, "correlation": 58844
    }
  },
  {
    "ph": "s", "id": 58844, "pid": 76337, "tid": -914061504, "ts": 1716454222261082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261102, "dur": 9,
    "args": {
      "External id": 58845,
      "cbid": 131, "correlation": 58845
    }
  },
  {
    "ph": "f", "id": 58845, "pid": 76337, "tid": -914061504, "ts": 1716454222261102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222261227, "dur": 3,
    "args": {
      "External id": 58853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 58853, "pid": 5, "tid": 7, "ts": 1716454222261227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261211, "dur": 20,
    "args": {
      "External id": 58853,
      "cbid": 211, "correlation": 58853
    }
  },
  {
    "ph": "s", "id": 58853, "pid": 76337, "tid": -914061504, "ts": 1716454222261211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222261320, "dur": 3,
    "args": {
      "External id": 58862,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58862,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0024271844660194173
    }
  },
  {
    "ph": "f", "id": 58862, "pid": 5, "tid": 7, "ts": 1716454222261320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261300, "dur": 19,
    "args": {
      "External id": 58862,
      "cbid": 41, "correlation": 58862
    }
  },
  {
    "ph": "s", "id": 58862, "pid": 76337, "tid": -914061504, "ts": 1716454222261300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222261412, "dur": 4,
    "args": {
      "External id": 58872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58872,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 58872, "pid": 5, "tid": 7, "ts": 1716454222261412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261395, "dur": 19,
    "args": {
      "External id": 58872,
      "cbid": 211, "correlation": 58872
    }
  },
  {
    "ph": "s", "id": 58872, "pid": 76337, "tid": -914061504, "ts": 1716454222261395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222261468, "dur": 1,
    "args": {
      "External id": 58882,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58882,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 58882, "pid": 5, "tid": 7, "ts": 1716454222261468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261453, "dur": 13,
    "args": {
      "External id": 58882,
      "cbid": 41, "correlation": 58882
    }
  },
  {
    "ph": "s", "id": 58882, "pid": 76337, "tid": -914061504, "ts": 1716454222261453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261467, "dur": 8,
    "args": {
      "External id": 58883,
      "cbid": 131, "correlation": 58883
    }
  },
  {
    "ph": "f", "id": 58883, "pid": 76337, "tid": -914061504, "ts": 1716454222261467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222261545, "dur": 3,
    "args": {
      "External id": 58890,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58890,
      "bytes": 98304, "memory bandwidth (GB/s)": 28.444444444444443
    }
  },
  {
    "ph": "f", "id": 58890, "pid": 5, "tid": 7, "ts": 1716454222261545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261514, "dur": 31,
    "args": {
      "External id": 58890,
      "cbid": 41, "correlation": 58890
    }
  },
  {
    "ph": "s", "id": 58890, "pid": 76337, "tid": -914061504, "ts": 1716454222261514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222261596, "dur": 1,
    "args": {
      "External id": 58901,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58901,
      "bytes": 2, "memory bandwidth (GB/s)": 0.00125
    }
  },
  {
    "ph": "f", "id": 58901, "pid": 5, "tid": 7, "ts": 1716454222261596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261584, "dur": 10,
    "args": {
      "External id": 58901,
      "cbid": 41, "correlation": 58901
    }
  },
  {
    "ph": "s", "id": 58901, "pid": 76337, "tid": -914061504, "ts": 1716454222261584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261594, "dur": 8,
    "args": {
      "External id": 58902,
      "cbid": 131, "correlation": 58902
    }
  },
  {
    "ph": "f", "id": 58902, "pid": 76337, "tid": -914061504, "ts": 1716454222261594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222261657, "dur": 3,
    "args": {
      "External id": 58910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58910,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58910, "pid": 5, "tid": 7, "ts": 1716454222261657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261643, "dur": 15,
    "args": {
      "External id": 58910,
      "cbid": 211, "correlation": 58910
    }
  },
  {
    "ph": "s", "id": 58910, "pid": 76337, "tid": -914061504, "ts": 1716454222261643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222261688, "dur": 3,
    "args": {
      "External id": 58920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58920,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58920, "pid": 5, "tid": 7, "ts": 1716454222261688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261678, "dur": 9,
    "args": {
      "External id": 58920,
      "cbid": 211, "correlation": 58920
    }
  },
  {
    "ph": "s", "id": 58920, "pid": 76337, "tid": -914061504, "ts": 1716454222261678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222261710, "dur": 3,
    "args": {
      "External id": 58929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58929,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58929, "pid": 5, "tid": 7, "ts": 1716454222261710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261701, "dur": 10,
    "args": {
      "External id": 58929,
      "cbid": 211, "correlation": 58929
    }
  },
  {
    "ph": "s", "id": 58929, "pid": 76337, "tid": -914061504, "ts": 1716454222261701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222261786, "dur": 6,
    "args": {
      "External id": 58937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58937,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58937, "pid": 5, "tid": 7, "ts": 1716454222261786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261773, "dur": 14,
    "args": {
      "External id": 58937,
      "cbid": 211, "correlation": 58937
    }
  },
  {
    "ph": "s", "id": 58937, "pid": 76337, "tid": -914061504, "ts": 1716454222261773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222261827, "dur": 3,
    "args": {
      "External id": 58946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58946,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58946, "pid": 5, "tid": 7, "ts": 1716454222261827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261817, "dur": 9,
    "args": {
      "External id": 58946,
      "cbid": 211, "correlation": 58946
    }
  },
  {
    "ph": "s", "id": 58946, "pid": 76337, "tid": -914061504, "ts": 1716454222261817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222261850, "dur": 3,
    "args": {
      "External id": 58955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58955,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58955, "pid": 5, "tid": 7, "ts": 1716454222261850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261841, "dur": 7,
    "args": {
      "External id": 58955,
      "cbid": 211, "correlation": 58955
    }
  },
  {
    "ph": "s", "id": 58955, "pid": 76337, "tid": -914061504, "ts": 1716454222261841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222261916, "dur": 3,
    "args": {
      "External id": 58963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 58963,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 58963, "pid": 5, "tid": 7, "ts": 1716454222261916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261905, "dur": 10,
    "args": {
      "External id": 58963,
      "cbid": 211, "correlation": 58963
    }
  },
  {
    "ph": "s", "id": 58963, "pid": 76337, "tid": -914061504, "ts": 1716454222261905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222261991, "dur": 1,
    "args": {
      "External id": 58971,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58971,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00423728813559322
    }
  },
  {
    "ph": "f", "id": 58971, "pid": 5, "tid": 7, "ts": 1716454222261991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222261965, "dur": 36,
    "args": {
      "External id": 58971,
      "cbid": 41, "correlation": 58971
    }
  },
  {
    "ph": "s", "id": 58971, "pid": 76337, "tid": -914061504, "ts": 1716454222261965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262002, "dur": 4,
    "args": {
      "External id": 58972,
      "cbid": 131, "correlation": 58972
    }
  },
  {
    "ph": "f", "id": 58972, "pid": 76337, "tid": -914061504, "ts": 1716454222262002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222262065, "dur": 1,
    "args": {
      "External id": 58982,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58982,
      "bytes": 42, "memory bandwidth (GB/s)": 0.026802807913209957
    }
  },
  {
    "ph": "f", "id": 58982, "pid": 5, "tid": 7, "ts": 1716454222262065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262052, "dur": 10,
    "args": {
      "External id": 58982,
      "cbid": 41, "correlation": 58982
    }
  },
  {
    "ph": "s", "id": 58982, "pid": 76337, "tid": -914061504, "ts": 1716454222262052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262063, "dur": 8,
    "args": {
      "External id": 58983,
      "cbid": 131, "correlation": 58983
    }
  },
  {
    "ph": "f", "id": 58983, "pid": 76337, "tid": -914061504, "ts": 1716454222262063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222262121, "dur": 1,
    "args": {
      "External id": 58992,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 58992,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 58992, "pid": 5, "tid": 7, "ts": 1716454222262121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262110, "dur": 8,
    "args": {
      "External id": 58992,
      "cbid": 41, "correlation": 58992
    }
  },
  {
    "ph": "s", "id": 58992, "pid": 76337, "tid": -914061504, "ts": 1716454222262110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262119, "dur": 8,
    "args": {
      "External id": 58993,
      "cbid": 131, "correlation": 58993
    }
  },
  {
    "ph": "f", "id": 58993, "pid": 76337, "tid": -914061504, "ts": 1716454222262119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222262197, "dur": 4,
    "args": {
      "External id": 59000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59000,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59000, "pid": 5, "tid": 7, "ts": 1716454222262197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262178, "dur": 21,
    "args": {
      "External id": 59000,
      "cbid": 211, "correlation": 59000
    }
  },
  {
    "ph": "s", "id": 59000, "pid": 76337, "tid": -914061504, "ts": 1716454222262178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454222262237, "dur": 4,
    "args": {
      "External id": 59020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59020,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59020, "pid": 5, "tid": 7, "ts": 1716454222262237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262226, "dur": 12,
    "args": {
      "External id": 59020,
      "cbid": 211, "correlation": 59020
    }
  },
  {
    "ph": "s", "id": 59020, "pid": 76337, "tid": -914061504, "ts": 1716454222262226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262238, "dur": 0,
    "args": {
      "External id": 59021,
      "cbid": 11, "correlation": 59021
    }
  },
  {
    "ph": "f", "id": 59021, "pid": 76337, "tid": -914061504, "ts": 1716454222262238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262239, "dur": 0,
    "args": {
      "External id": 59022,
      "cbid": 11, "correlation": 59022
    }
  },
  {
    "ph": "f", "id": 59022, "pid": 76337, "tid": -914061504, "ts": 1716454222262239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222262254, "dur": 1,
    "args": {
      "External id": 59025,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 59025,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0022727272727272726
    }
  },
  {
    "ph": "f", "id": 59025, "pid": 5, "tid": 7, "ts": 1716454222262254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262240, "dur": 23,
    "args": {
      "External id": 59025,
      "cbid": 41, "correlation": 59025
    }
  },
  {
    "ph": "s", "id": 59025, "pid": 76337, "tid": -914061504, "ts": 1716454222262240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262264, "dur": 3,
    "args": {
      "External id": 59026,
      "cbid": 131, "correlation": 59026
    }
  },
  {
    "ph": "f", "id": 59026, "pid": 76337, "tid": -914061504, "ts": 1716454222262264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222262292, "dur": 3,
    "args": {
      "External id": 59050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59050, "pid": 5, "tid": 7, "ts": 1716454222262292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262283, "dur": 9,
    "args": {
      "External id": 59050,
      "cbid": 211, "correlation": 59050
    }
  },
  {
    "ph": "s", "id": 59050, "pid": 76337, "tid": -914061504, "ts": 1716454222262283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262293, "dur": 0,
    "args": {
      "External id": 59051,
      "cbid": 11, "correlation": 59051
    }
  },
  {
    "ph": "f", "id": 59051, "pid": 76337, "tid": -914061504, "ts": 1716454222262293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262293, "dur": 0,
    "args": {
      "External id": 59052,
      "cbid": 11, "correlation": 59052
    }
  },
  {
    "ph": "f", "id": 59052, "pid": 76337, "tid": -914061504, "ts": 1716454222262293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262295, "dur": 1,
    "args": {
      "External id": 59054,
      "cbid": 200, "correlation": 59054
    }
  },
  {
    "ph": "f", "id": 59054, "pid": 76337, "tid": -914061504, "ts": 1716454222262295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222262306, "dur": 4,
    "args": {
      "External id": 59056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59056,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59056, "pid": 5, "tid": 7, "ts": 1716454222262306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262298, "dur": 9,
    "args": {
      "External id": 59056,
      "cbid": 211, "correlation": 59056
    }
  },
  {
    "ph": "s", "id": 59056, "pid": 76337, "tid": -914061504, "ts": 1716454222262298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262307, "dur": 0,
    "args": {
      "External id": 59057,
      "cbid": 11, "correlation": 59057
    }
  },
  {
    "ph": "f", "id": 59057, "pid": 76337, "tid": -914061504, "ts": 1716454222262307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262308, "dur": 0,
    "args": {
      "External id": 59058,
      "cbid": 11, "correlation": 59058
    }
  },
  {
    "ph": "f", "id": 59058, "pid": 76337, "tid": -914061504, "ts": 1716454222262308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222262345, "dur": 1,
    "args": {
      "External id": 59065,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 59065,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004629629629629629
    }
  },
  {
    "ph": "f", "id": 59065, "pid": 5, "tid": 7, "ts": 1716454222262345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262333, "dur": 20,
    "args": {
      "External id": 59065,
      "cbid": 41, "correlation": 59065
    }
  },
  {
    "ph": "s", "id": 59065, "pid": 76337, "tid": -914061504, "ts": 1716454222262333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262354, "dur": 3,
    "args": {
      "External id": 59066,
      "cbid": 131, "correlation": 59066
    }
  },
  {
    "ph": "f", "id": 59066, "pid": 76337, "tid": -914061504, "ts": 1716454222262354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222262405, "dur": 1,
    "args": {
      "External id": 59076,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 59076,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 59076, "pid": 5, "tid": 7, "ts": 1716454222262405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262392, "dur": 10,
    "args": {
      "External id": 59076,
      "cbid": 41, "correlation": 59076
    }
  },
  {
    "ph": "s", "id": 59076, "pid": 76337, "tid": -914061504, "ts": 1716454222262392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262404, "dur": 8,
    "args": {
      "External id": 59077,
      "cbid": 131, "correlation": 59077
    }
  },
  {
    "ph": "f", "id": 59077, "pid": 76337, "tid": -914061504, "ts": 1716454222262404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222262479, "dur": 6,
    "args": {
      "External id": 59084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59084,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59084, "pid": 5, "tid": 7, "ts": 1716454222262479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262463, "dur": 16,
    "args": {
      "External id": 59084,
      "cbid": 211, "correlation": 59084
    }
  },
  {
    "ph": "s", "id": 59084, "pid": 76337, "tid": -914061504, "ts": 1716454222262463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222262549, "dur": 3,
    "args": {
      "External id": 59093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59093,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59093, "pid": 5, "tid": 7, "ts": 1716454222262549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262536, "dur": 13,
    "args": {
      "External id": 59093,
      "cbid": 211, "correlation": 59093
    }
  },
  {
    "ph": "s", "id": 59093, "pid": 76337, "tid": -914061504, "ts": 1716454222262536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222262586, "dur": 3,
    "args": {
      "External id": 59101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59101,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59101, "pid": 5, "tid": 7, "ts": 1716454222262586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262575, "dur": 10,
    "args": {
      "External id": 59101,
      "cbid": 211, "correlation": 59101
    }
  },
  {
    "ph": "s", "id": 59101, "pid": 76337, "tid": -914061504, "ts": 1716454222262575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222262617, "dur": 4,
    "args": {
      "External id": 59109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59109,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59109, "pid": 5, "tid": 7, "ts": 1716454222262617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262607, "dur": 11,
    "args": {
      "External id": 59109,
      "cbid": 211, "correlation": 59109
    }
  },
  {
    "ph": "s", "id": 59109, "pid": 76337, "tid": -914061504, "ts": 1716454222262607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222262647, "dur": 5,
    "args": {
      "External id": 59117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59117,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59117, "pid": 5, "tid": 7, "ts": 1716454222262647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262637, "dur": 9,
    "args": {
      "External id": 59117,
      "cbid": 211, "correlation": 59117
    }
  },
  {
    "ph": "s", "id": 59117, "pid": 76337, "tid": -914061504, "ts": 1716454222262637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222262672, "dur": 3,
    "args": {
      "External id": 59125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59125,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59125, "pid": 5, "tid": 7, "ts": 1716454222262672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262663, "dur": 8,
    "args": {
      "External id": 59125,
      "cbid": 211, "correlation": 59125
    }
  },
  {
    "ph": "s", "id": 59125, "pid": 76337, "tid": -914061504, "ts": 1716454222262663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222262699, "dur": 4,
    "args": {
      "External id": 59133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59133,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59133, "pid": 5, "tid": 7, "ts": 1716454222262699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262688, "dur": 10,
    "args": {
      "External id": 59133,
      "cbid": 211, "correlation": 59133
    }
  },
  {
    "ph": "s", "id": 59133, "pid": 76337, "tid": -914061504, "ts": 1716454222262688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222262720, "dur": 4,
    "args": {
      "External id": 59141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59141, "pid": 5, "tid": 7, "ts": 1716454222262720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262712, "dur": 7,
    "args": {
      "External id": 59141,
      "cbid": 211, "correlation": 59141
    }
  },
  {
    "ph": "s", "id": 59141, "pid": 76337, "tid": -914061504, "ts": 1716454222262712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222262739, "dur": 5,
    "args": {
      "External id": 59149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59149,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59149, "pid": 5, "tid": 7, "ts": 1716454222262739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262731, "dur": 9,
    "args": {
      "External id": 59149,
      "cbid": 211, "correlation": 59149
    }
  },
  {
    "ph": "s", "id": 59149, "pid": 76337, "tid": -914061504, "ts": 1716454222262731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222262762, "dur": 3,
    "args": {
      "External id": 59157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59157,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59157, "pid": 5, "tid": 7, "ts": 1716454222262762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262754, "dur": 7,
    "args": {
      "External id": 59157,
      "cbid": 211, "correlation": 59157
    }
  },
  {
    "ph": "s", "id": 59157, "pid": 76337, "tid": -914061504, "ts": 1716454222262754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222262819, "dur": 3,
    "args": {
      "External id": 59165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59165,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59165, "pid": 5, "tid": 7, "ts": 1716454222262819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262808, "dur": 10,
    "args": {
      "External id": 59165,
      "cbid": 211, "correlation": 59165
    }
  },
  {
    "ph": "s", "id": 59165, "pid": 76337, "tid": -914061504, "ts": 1716454222262808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222262844, "dur": 5,
    "args": {
      "External id": 59173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59173,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59173, "pid": 5, "tid": 7, "ts": 1716454222262844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262835, "dur": 8,
    "args": {
      "External id": 59173,
      "cbid": 211, "correlation": 59173
    }
  },
  {
    "ph": "s", "id": 59173, "pid": 76337, "tid": -914061504, "ts": 1716454222262835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222262867, "dur": 5,
    "args": {
      "External id": 59181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59181,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59181, "pid": 5, "tid": 7, "ts": 1716454222262867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262858, "dur": 8,
    "args": {
      "External id": 59181,
      "cbid": 211, "correlation": 59181
    }
  },
  {
    "ph": "s", "id": 59181, "pid": 76337, "tid": -914061504, "ts": 1716454222262858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222262885, "dur": 3,
    "args": {
      "External id": 59189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59189,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59189, "pid": 5, "tid": 7, "ts": 1716454222262885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222262877, "dur": 6,
    "args": {
      "External id": 59189,
      "cbid": 211, "correlation": 59189
    }
  },
  {
    "ph": "s", "id": 59189, "pid": 76337, "tid": -914061504, "ts": 1716454222262877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222263284, "dur": 6,
    "args": {
      "External id": 59198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59198,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59198, "pid": 5, "tid": 7, "ts": 1716454222263284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263267, "dur": 17,
    "args": {
      "External id": 59198,
      "cbid": 211, "correlation": 59198
    }
  },
  {
    "ph": "s", "id": 59198, "pid": 76337, "tid": -914061504, "ts": 1716454222263267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222263321, "dur": 6,
    "args": {
      "External id": 59207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59207,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59207, "pid": 5, "tid": 7, "ts": 1716454222263321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263310, "dur": 9,
    "args": {
      "External id": 59207,
      "cbid": 211, "correlation": 59207
    }
  },
  {
    "ph": "s", "id": 59207, "pid": 76337, "tid": -914061504, "ts": 1716454222263310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222263459, "dur": 3,
    "args": {
      "External id": 59223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59223, "pid": 5, "tid": 7, "ts": 1716454222263459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263443, "dur": 16,
    "args": {
      "External id": 59223,
      "cbid": 211, "correlation": 59223
    }
  },
  {
    "ph": "s", "id": 59223, "pid": 76337, "tid": -914061504, "ts": 1716454222263443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222263492, "dur": 3,
    "args": {
      "External id": 59231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59231, "pid": 5, "tid": 7, "ts": 1716454222263492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263483, "dur": 8,
    "args": {
      "External id": 59231,
      "cbid": 211, "correlation": 59231
    }
  },
  {
    "ph": "s", "id": 59231, "pid": 76337, "tid": -914061504, "ts": 1716454222263483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222263525, "dur": 3,
    "args": {
      "External id": 59239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59239,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59239, "pid": 5, "tid": 7, "ts": 1716454222263525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263514, "dur": 9,
    "args": {
      "External id": 59239,
      "cbid": 211, "correlation": 59239
    }
  },
  {
    "ph": "s", "id": 59239, "pid": 76337, "tid": -914061504, "ts": 1716454222263514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222263555, "dur": 4,
    "args": {
      "External id": 59247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59247,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59247, "pid": 5, "tid": 7, "ts": 1716454222263555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263546, "dur": 8,
    "args": {
      "External id": 59247,
      "cbid": 211, "correlation": 59247
    }
  },
  {
    "ph": "s", "id": 59247, "pid": 76337, "tid": -914061504, "ts": 1716454222263546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222263612, "dur": 4,
    "args": {
      "External id": 59259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59259,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59259, "pid": 5, "tid": 7, "ts": 1716454222263612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263599, "dur": 13,
    "args": {
      "External id": 59259,
      "cbid": 211, "correlation": 59259
    }
  },
  {
    "ph": "s", "id": 59259, "pid": 76337, "tid": -914061504, "ts": 1716454222263599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222263657, "dur": 4,
    "args": {
      "External id": 59270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59270, "pid": 5, "tid": 7, "ts": 1716454222263657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263646, "dur": 11,
    "args": {
      "External id": 59270,
      "cbid": 211, "correlation": 59270
    }
  },
  {
    "ph": "s", "id": 59270, "pid": 76337, "tid": -914061504, "ts": 1716454222263646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222263688, "dur": 3,
    "args": {
      "External id": 59278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59278, "pid": 5, "tid": 7, "ts": 1716454222263688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263678, "dur": 9,
    "args": {
      "External id": 59278,
      "cbid": 211, "correlation": 59278
    }
  },
  {
    "ph": "s", "id": 59278, "pid": 76337, "tid": -914061504, "ts": 1716454222263678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222263724, "dur": 6,
    "args": {
      "External id": 59286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59286,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59286, "pid": 5, "tid": 7, "ts": 1716454222263724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263714, "dur": 12,
    "args": {
      "External id": 59286,
      "cbid": 211, "correlation": 59286
    }
  },
  {
    "ph": "s", "id": 59286, "pid": 76337, "tid": -914061504, "ts": 1716454222263714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222263756, "dur": 5,
    "args": {
      "External id": 59294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59294,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59294, "pid": 5, "tid": 7, "ts": 1716454222263756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263745, "dur": 10,
    "args": {
      "External id": 59294,
      "cbid": 211, "correlation": 59294
    }
  },
  {
    "ph": "s", "id": 59294, "pid": 76337, "tid": -914061504, "ts": 1716454222263745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222263787, "dur": 4,
    "args": {
      "External id": 59303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59303,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59303, "pid": 5, "tid": 7, "ts": 1716454222263787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263776, "dur": 10,
    "args": {
      "External id": 59303,
      "cbid": 211, "correlation": 59303
    }
  },
  {
    "ph": "s", "id": 59303, "pid": 76337, "tid": -914061504, "ts": 1716454222263776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222263848, "dur": 5,
    "args": {
      "External id": 59316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59316,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59316, "pid": 5, "tid": 7, "ts": 1716454222263848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263835, "dur": 12,
    "args": {
      "External id": 59316,
      "cbid": 211, "correlation": 59316
    }
  },
  {
    "ph": "s", "id": 59316, "pid": 76337, "tid": -914061504, "ts": 1716454222263835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222263888, "dur": 5,
    "args": {
      "External id": 59326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59326,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59326, "pid": 5, "tid": 7, "ts": 1716454222263888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222263877, "dur": 10,
    "args": {
      "External id": 59326,
      "cbid": 211, "correlation": 59326
    }
  },
  {
    "ph": "s", "id": 59326, "pid": 76337, "tid": -914061504, "ts": 1716454222263877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264034, "dur": 5,
    "args": {
      "External id": 59343,
      "cbid": 251, "correlation": 59343
    }
  },
  {
    "ph": "f", "id": 59343, "pid": 76337, "tid": -914061504, "ts": 1716454222264034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454222264064, "dur": 12,
    "args": {
      "External id": 59345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59345,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59345, "pid": 5, "tid": 7, "ts": 1716454222264064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264048, "dur": 17,
    "args": {
      "External id": 59345,
      "cbid": 211, "correlation": 59345
    }
  },
  {
    "ph": "s", "id": 59345, "pid": 76337, "tid": -914061504, "ts": 1716454222264048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222264130, "dur": 4,
    "args": {
      "External id": 59353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59353,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59353, "pid": 5, "tid": 7, "ts": 1716454222264130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264117, "dur": 12,
    "args": {
      "External id": 59353,
      "cbid": 211, "correlation": 59353
    }
  },
  {
    "ph": "s", "id": 59353, "pid": 76337, "tid": -914061504, "ts": 1716454222264117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264191, "dur": 2,
    "args": {
      "External id": 59369,
      "cbid": 251, "correlation": 59369
    }
  },
  {
    "ph": "f", "id": 59369, "pid": 76337, "tid": -914061504, "ts": 1716454222264191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264197, "dur": 0,
    "args": {
      "External id": 59371,
      "cbid": 251, "correlation": 59371
    }
  },
  {
    "ph": "f", "id": 59371, "pid": 76337, "tid": -914061504, "ts": 1716454222264197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222264214, "dur": 14,
    "args": {
      "External id": 59372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59372,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59372, "pid": 5, "tid": 7, "ts": 1716454222264214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264200, "dur": 14,
    "args": {
      "External id": 59372,
      "cbid": 211, "correlation": 59372
    }
  },
  {
    "ph": "s", "id": 59372, "pid": 76337, "tid": -914061504, "ts": 1716454222264200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222264230, "dur": 5,
    "args": {
      "External id": 59374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59374,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59374, "pid": 5, "tid": 7, "ts": 1716454222264230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264219, "dur": 9,
    "args": {
      "External id": 59374,
      "cbid": 211, "correlation": 59374
    }
  },
  {
    "ph": "s", "id": 59374, "pid": 76337, "tid": -914061504, "ts": 1716454222264219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264337, "dur": 1,
    "args": {
      "External id": 59384,
      "cbid": 317, "correlation": 59384
    }
  },
  {
    "ph": "f", "id": 59384, "pid": 76337, "tid": -914061504, "ts": 1716454222264337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264339, "dur": 1,
    "args": {
      "External id": 59385,
      "cbid": 203, "correlation": 59385
    }
  },
  {
    "ph": "f", "id": 59385, "pid": 76337, "tid": -914061504, "ts": 1716454222264339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264341, "dur": 1,
    "args": {
      "External id": 59386,
      "cbid": 205, "correlation": 59386
    }
  },
  {
    "ph": "f", "id": 59386, "pid": 76337, "tid": -914061504, "ts": 1716454222264341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222264401, "dur": 7,
    "args": {
      "External id": 59390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59390,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59390, "pid": 5, "tid": 7, "ts": 1716454222264401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264384, "dur": 16,
    "args": {
      "External id": 59390,
      "cbid": 211, "correlation": 59390
    }
  },
  {
    "ph": "s", "id": 59390, "pid": 76337, "tid": -914061504, "ts": 1716454222264384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222264412, "dur": 4,
    "args": {
      "External id": 59392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59392,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 59392, "pid": 5, "tid": 7, "ts": 1716454222264412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264404, "dur": 6,
    "args": {
      "External id": 59392,
      "cbid": 211, "correlation": 59392
    }
  },
  {
    "ph": "s", "id": 59392, "pid": 76337, "tid": -914061504, "ts": 1716454222264404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222264431, "dur": 4,
    "args": {
      "External id": 59394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59394, "pid": 5, "tid": 7, "ts": 1716454222264431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264422, "dur": 8,
    "args": {
      "External id": 59394,
      "cbid": 211, "correlation": 59394
    }
  },
  {
    "ph": "s", "id": 59394, "pid": 76337, "tid": -914061504, "ts": 1716454222264422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264438, "dur": 0,
    "args": {
      "External id": 59395,
      "cbid": 51, "correlation": 59395
    }
  },
  {
    "ph": "s", "id": 59395, "pid": 76337, "tid": -914061504, "ts": 1716454222264438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222264449, "dur": 93,
    "args": {
      "External id": 59396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59396,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59396, "pid": 5, "tid": 7, "ts": 1716454222264449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264439, "dur": 8,
    "args": {
      "External id": 59396,
      "cbid": 211, "correlation": 59396
    }
  },
  {
    "ph": "s", "id": 59396, "pid": 76337, "tid": -914061504, "ts": 1716454222264439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222264544, "dur": 64,
    "args": {
      "External id": 59401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59401, "pid": 5, "tid": 7, "ts": 1716454222264544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222264477, "dur": 10,
    "args": {
      "External id": 59401,
      "cbid": 211, "correlation": 59401
    }
  },
  {
    "ph": "s", "id": 59401, "pid": 76337, "tid": -914061504, "ts": 1716454222264477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222266376, "dur": 55,
    "args": {
      "External id": 59421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59421,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 59421, "pid": 5, "tid": 7, "ts": 1716454222266376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266358, "dur": 18,
    "args": {
      "External id": 59421,
      "cbid": 211, "correlation": 59421
    }
  },
  {
    "ph": "s", "id": 59421, "pid": 76337, "tid": -914061504, "ts": 1716454222266358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222266433, "dur": 5,
    "args": {
      "External id": 59433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59433,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59433, "pid": 5, "tid": 7, "ts": 1716454222266433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266392, "dur": 9,
    "args": {
      "External id": 59433,
      "cbid": 211, "correlation": 59433
    }
  },
  {
    "ph": "s", "id": 59433, "pid": 76337, "tid": -914061504, "ts": 1716454222266392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222266439, "dur": 61,
    "args": {
      "External id": 59436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59436, "pid": 5, "tid": 7, "ts": 1716454222266439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266417, "dur": 8,
    "args": {
      "External id": 59436,
      "cbid": 211, "correlation": 59436
    }
  },
  {
    "ph": "s", "id": 59436, "pid": 76337, "tid": -914061504, "ts": 1716454222266417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222266501, "dur": 38,
    "args": {
      "External id": 59445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59445,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59445, "pid": 5, "tid": 7, "ts": 1716454222266501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266463, "dur": 10,
    "args": {
      "External id": 59445,
      "cbid": 211, "correlation": 59445
    }
  },
  {
    "ph": "s", "id": 59445, "pid": 76337, "tid": -914061504, "ts": 1716454222266463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266522, "dur": 0,
    "args": {
      "External id": 59455,
      "cbid": 317, "correlation": 59455
    }
  },
  {
    "ph": "f", "id": 59455, "pid": 76337, "tid": -914061504, "ts": 1716454222266522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266523, "dur": 0,
    "args": {
      "External id": 59456,
      "cbid": 203, "correlation": 59456
    }
  },
  {
    "ph": "f", "id": 59456, "pid": 76337, "tid": -914061504, "ts": 1716454222266523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266523, "dur": 0,
    "args": {
      "External id": 59457,
      "cbid": 205, "correlation": 59457
    }
  },
  {
    "ph": "f", "id": 59457, "pid": 76337, "tid": -914061504, "ts": 1716454222266523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222266555, "dur": 42,
    "args": {
      "External id": 59461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59461,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59461, "pid": 5, "tid": 7, "ts": 1716454222266555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266542, "dur": 12,
    "args": {
      "External id": 59461,
      "cbid": 211, "correlation": 59461
    }
  },
  {
    "ph": "s", "id": 59461, "pid": 76337, "tid": -914061504, "ts": 1716454222266542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222266598, "dur": 15,
    "args": {
      "External id": 59463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59463,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59463, "pid": 5, "tid": 7, "ts": 1716454222266598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266557, "dur": 6,
    "args": {
      "External id": 59463,
      "cbid": 211, "correlation": 59463
    }
  },
  {
    "ph": "s", "id": 59463, "pid": 76337, "tid": -914061504, "ts": 1716454222266557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222266615, "dur": 3,
    "args": {
      "External id": 59465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59465,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59465, "pid": 5, "tid": 7, "ts": 1716454222266615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266569, "dur": 6,
    "args": {
      "External id": 59465,
      "cbid": 211, "correlation": 59465
    }
  },
  {
    "ph": "s", "id": 59465, "pid": 76337, "tid": -914061504, "ts": 1716454222266569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266579, "dur": 0,
    "args": {
      "External id": 59466,
      "cbid": 51, "correlation": 59466
    }
  },
  {
    "ph": "s", "id": 59466, "pid": 76337, "tid": -914061504, "ts": 1716454222266579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222266620, "dur": 770,
    "args": {
      "External id": 59467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59467,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59467, "pid": 5, "tid": 7, "ts": 1716454222266620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266581, "dur": 6,
    "args": {
      "External id": 59467,
      "cbid": 211, "correlation": 59467
    }
  },
  {
    "ph": "s", "id": 59467, "pid": 76337, "tid": -914061504, "ts": 1716454222266581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222267391, "dur": 62,
    "args": {
      "External id": 59472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59472,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59472, "pid": 5, "tid": 7, "ts": 1716454222267391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266610, "dur": 9,
    "args": {
      "External id": 59472,
      "cbid": 211, "correlation": 59472
    }
  },
  {
    "ph": "s", "id": 59472, "pid": 76337, "tid": -914061504, "ts": 1716454222266610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222267455, "dur": 4,
    "args": {
      "External id": 59480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59480,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59480, "pid": 5, "tid": 7, "ts": 1716454222267455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266656, "dur": 9,
    "args": {
      "External id": 59480,
      "cbid": 211, "correlation": 59480
    }
  },
  {
    "ph": "s", "id": 59480, "pid": 76337, "tid": -914061504, "ts": 1716454222266656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266727, "dur": 1,
    "args": {
      "External id": 59496,
      "cbid": 251, "correlation": 59496
    }
  },
  {
    "ph": "f", "id": 59496, "pid": 76337, "tid": -914061504, "ts": 1716454222266727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266733, "dur": 0,
    "args": {
      "External id": 59498,
      "cbid": 251, "correlation": 59498
    }
  },
  {
    "ph": "f", "id": 59498, "pid": 76337, "tid": -914061504, "ts": 1716454222266733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222267460, "dur": 10,
    "args": {
      "External id": 59499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59499,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 59499, "pid": 5, "tid": 7, "ts": 1716454222267460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266735, "dur": 12,
    "args": {
      "External id": 59499,
      "cbid": 211, "correlation": 59499
    }
  },
  {
    "ph": "s", "id": 59499, "pid": 76337, "tid": -914061504, "ts": 1716454222266735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222267471, "dur": 5,
    "args": {
      "External id": 59501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59501,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 59501, "pid": 5, "tid": 7, "ts": 1716454222267471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266749, "dur": 6,
    "args": {
      "External id": 59501,
      "cbid": 211, "correlation": 59501
    }
  },
  {
    "ph": "s", "id": 59501, "pid": 76337, "tid": -914061504, "ts": 1716454222266749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222267477, "dur": 56,
    "args": {
      "External id": 59511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59511, "pid": 5, "tid": 7, "ts": 1716454222267477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266810, "dur": 12,
    "args": {
      "External id": 59511,
      "cbid": 211, "correlation": 59511
    }
  },
  {
    "ph": "s", "id": 59511, "pid": 76337, "tid": -914061504, "ts": 1716454222266810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222267535, "dur": 55,
    "args": {
      "External id": 59531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59531,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 59531, "pid": 5, "tid": 7, "ts": 1716454222267535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266878, "dur": 11,
    "args": {
      "External id": 59531,
      "cbid": 211, "correlation": 59531
    }
  },
  {
    "ph": "s", "id": 59531, "pid": 76337, "tid": -914061504, "ts": 1716454222266878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222267592, "dur": 4,
    "args": {
      "External id": 59543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59543,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59543, "pid": 5, "tid": 7, "ts": 1716454222267592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266899, "dur": 6,
    "args": {
      "External id": 59543,
      "cbid": 211, "correlation": 59543
    }
  },
  {
    "ph": "s", "id": 59543, "pid": 76337, "tid": -914061504, "ts": 1716454222266899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222267597, "dur": 59,
    "args": {
      "External id": 59546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59546, "pid": 5, "tid": 7, "ts": 1716454222267597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266918, "dur": 7,
    "args": {
      "External id": 59546,
      "cbid": 211, "correlation": 59546
    }
  },
  {
    "ph": "s", "id": 59546, "pid": 76337, "tid": -914061504, "ts": 1716454222266918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222267657, "dur": 38,
    "args": {
      "External id": 59555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59555,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59555, "pid": 5, "tid": 7, "ts": 1716454222267657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222266958, "dur": 10,
    "args": {
      "External id": 59555,
      "cbid": 211, "correlation": 59555
    }
  },
  {
    "ph": "s", "id": 59555, "pid": 76337, "tid": -914061504, "ts": 1716454222266958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267040, "dur": 0,
    "args": {
      "External id": 59565,
      "cbid": 317, "correlation": 59565
    }
  },
  {
    "ph": "f", "id": 59565, "pid": 76337, "tid": -914061504, "ts": 1716454222267040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267041, "dur": 0,
    "args": {
      "External id": 59566,
      "cbid": 203, "correlation": 59566
    }
  },
  {
    "ph": "f", "id": 59566, "pid": 76337, "tid": -914061504, "ts": 1716454222267041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267042, "dur": 0,
    "args": {
      "External id": 59567,
      "cbid": 205, "correlation": 59567
    }
  },
  {
    "ph": "f", "id": 59567, "pid": 76337, "tid": -914061504, "ts": 1716454222267042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222267697, "dur": 41,
    "args": {
      "External id": 59571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59571,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59571, "pid": 5, "tid": 7, "ts": 1716454222267697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267057, "dur": 13,
    "args": {
      "External id": 59571,
      "cbid": 211, "correlation": 59571
    }
  },
  {
    "ph": "s", "id": 59571, "pid": 76337, "tid": -914061504, "ts": 1716454222267057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222267739, "dur": 16,
    "args": {
      "External id": 59573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59573,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59573, "pid": 5, "tid": 7, "ts": 1716454222267739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267072, "dur": 5,
    "args": {
      "External id": 59573,
      "cbid": 211, "correlation": 59573
    }
  },
  {
    "ph": "s", "id": 59573, "pid": 76337, "tid": -914061504, "ts": 1716454222267072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222267756, "dur": 4,
    "args": {
      "External id": 59575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59575, "pid": 5, "tid": 7, "ts": 1716454222267756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267081, "dur": 5,
    "args": {
      "External id": 59575,
      "cbid": 211, "correlation": 59575
    }
  },
  {
    "ph": "s", "id": 59575, "pid": 76337, "tid": -914061504, "ts": 1716454222267081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267090, "dur": 0,
    "args": {
      "External id": 59576,
      "cbid": 51, "correlation": 59576
    }
  },
  {
    "ph": "s", "id": 59576, "pid": 76337, "tid": -914061504, "ts": 1716454222267090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222267761, "dur": 764,
    "args": {
      "External id": 59577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59577,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59577, "pid": 5, "tid": 7, "ts": 1716454222267761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267090, "dur": 5,
    "args": {
      "External id": 59577,
      "cbid": 211, "correlation": 59577
    }
  },
  {
    "ph": "s", "id": 59577, "pid": 76337, "tid": -914061504, "ts": 1716454222267090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222268526, "dur": 63,
    "args": {
      "External id": 59582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59582, "pid": 5, "tid": 7, "ts": 1716454222268526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267119, "dur": 9,
    "args": {
      "External id": 59582,
      "cbid": 211, "correlation": 59582
    }
  },
  {
    "ph": "s", "id": 59582, "pid": 76337, "tid": -914061504, "ts": 1716454222267119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222268591, "dur": 51,
    "args": {
      "External id": 59590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59590,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59590, "pid": 5, "tid": 7, "ts": 1716454222268591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267153, "dur": 9,
    "args": {
      "External id": 59590,
      "cbid": 211, "correlation": 59590
    }
  },
  {
    "ph": "s", "id": 59590, "pid": 76337, "tid": -914061504, "ts": 1716454222267153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222268643, "dur": 37,
    "args": {
      "External id": 59598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59598,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59598, "pid": 5, "tid": 7, "ts": 1716454222268643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267185, "dur": 10,
    "args": {
      "External id": 59598,
      "cbid": 211, "correlation": 59598
    }
  },
  {
    "ph": "s", "id": 59598, "pid": 76337, "tid": -914061504, "ts": 1716454222267185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222268681, "dur": 55,
    "args": {
      "External id": 59618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59618,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 59618, "pid": 5, "tid": 7, "ts": 1716454222268681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267268, "dur": 12,
    "args": {
      "External id": 59618,
      "cbid": 211, "correlation": 59618
    }
  },
  {
    "ph": "s", "id": 59618, "pid": 76337, "tid": -914061504, "ts": 1716454222267268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222268738, "dur": 4,
    "args": {
      "External id": 59630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59630,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 59630, "pid": 5, "tid": 7, "ts": 1716454222268738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267292, "dur": 7,
    "args": {
      "External id": 59630,
      "cbid": 211, "correlation": 59630
    }
  },
  {
    "ph": "s", "id": 59630, "pid": 76337, "tid": -914061504, "ts": 1716454222267292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222268743, "dur": 58,
    "args": {
      "External id": 59633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59633, "pid": 5, "tid": 7, "ts": 1716454222268743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267312, "dur": 7,
    "args": {
      "External id": 59633,
      "cbid": 211, "correlation": 59633
    }
  },
  {
    "ph": "s", "id": 59633, "pid": 76337, "tid": -914061504, "ts": 1716454222267312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267370, "dur": 0,
    "args": {
      "External id": 59644,
      "cbid": 317, "correlation": 59644
    }
  },
  {
    "ph": "f", "id": 59644, "pid": 76337, "tid": -914061504, "ts": 1716454222267370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267371, "dur": 0,
    "args": {
      "External id": 59645,
      "cbid": 203, "correlation": 59645
    }
  },
  {
    "ph": "f", "id": 59645, "pid": 76337, "tid": -914061504, "ts": 1716454222267371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267371, "dur": 0,
    "args": {
      "External id": 59646,
      "cbid": 205, "correlation": 59646
    }
  },
  {
    "ph": "f", "id": 59646, "pid": 76337, "tid": -914061504, "ts": 1716454222267371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267408, "dur": 3,
    "args": {
      "External id": 59650,
      "cbid": 251, "correlation": 59650
    }
  },
  {
    "ph": "f", "id": 59650, "pid": 76337, "tid": -914061504, "ts": 1716454222267408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267412, "dur": 1,
    "args": {
      "External id": 59651,
      "cbid": 251, "correlation": 59651
    }
  },
  {
    "ph": "f", "id": 59651, "pid": 76337, "tid": -914061504, "ts": 1716454222267412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267414, "dur": 2,
    "args": {
      "External id": 59652,
      "cbid": 251, "correlation": 59652
    }
  },
  {
    "ph": "f", "id": 59652, "pid": 76337, "tid": -914061504, "ts": 1716454222267414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267417, "dur": 1,
    "args": {
      "External id": 59653,
      "cbid": 251, "correlation": 59653
    }
  },
  {
    "ph": "f", "id": 59653, "pid": 76337, "tid": -914061504, "ts": 1716454222267417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267418, "dur": 1,
    "args": {
      "External id": 59654,
      "cbid": 251, "correlation": 59654
    }
  },
  {
    "ph": "f", "id": 59654, "pid": 76337, "tid": -914061504, "ts": 1716454222267418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267420, "dur": 1,
    "args": {
      "External id": 59655,
      "cbid": 251, "correlation": 59655
    }
  },
  {
    "ph": "f", "id": 59655, "pid": 76337, "tid": -914061504, "ts": 1716454222267420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267422, "dur": 1,
    "args": {
      "External id": 59656,
      "cbid": 251, "correlation": 59656
    }
  },
  {
    "ph": "f", "id": 59656, "pid": 76337, "tid": -914061504, "ts": 1716454222267422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267424, "dur": 1,
    "args": {
      "External id": 59657,
      "cbid": 251, "correlation": 59657
    }
  },
  {
    "ph": "f", "id": 59657, "pid": 76337, "tid": -914061504, "ts": 1716454222267424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267426, "dur": 0,
    "args": {
      "External id": 59658,
      "cbid": 251, "correlation": 59658
    }
  },
  {
    "ph": "f", "id": 59658, "pid": 76337, "tid": -914061504, "ts": 1716454222267426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222268802, "dur": 126,
    "args": {
      "External id": 59659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59659,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 59659, "pid": 5, "tid": 7, "ts": 1716454222268802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267431, "dur": 14,
    "args": {
      "External id": 59659,
      "cbid": 211, "correlation": 59659
    }
  },
  {
    "ph": "s", "id": 59659, "pid": 76337, "tid": -914061504, "ts": 1716454222267431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222268930, "dur": 63,
    "args": {
      "External id": 59665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59665, "pid": 5, "tid": 7, "ts": 1716454222268930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267468, "dur": 9,
    "args": {
      "External id": 59665,
      "cbid": 211, "correlation": 59665
    }
  },
  {
    "ph": "s", "id": 59665, "pid": 76337, "tid": -914061504, "ts": 1716454222267468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222268994, "dur": 656,
    "args": {
      "External id": 59674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59674,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59674, "pid": 5, "tid": 7, "ts": 1716454222268994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267569, "dur": 15,
    "args": {
      "External id": 59674,
      "cbid": 211, "correlation": 59674
    }
  },
  {
    "ph": "s", "id": 59674, "pid": 76337, "tid": -914061504, "ts": 1716454222267569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222269652, "dur": 200,
    "args": {
      "External id": 59696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59696,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59696, "pid": 5, "tid": 7, "ts": 1716454222269652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267642, "dur": 13,
    "args": {
      "External id": 59696,
      "cbid": 211, "correlation": 59696
    }
  },
  {
    "ph": "s", "id": 59696, "pid": 76337, "tid": -914061504, "ts": 1716454222267642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267761, "dur": 2,
    "args": {
      "External id": 59707,
      "cbid": 251, "correlation": 59707
    }
  },
  {
    "ph": "f", "id": 59707, "pid": 76337, "tid": -914061504, "ts": 1716454222267761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222269853, "dur": 213,
    "args": {
      "External id": 59708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59708,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59708, "pid": 5, "tid": 7, "ts": 1716454222269853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267769, "dur": 14,
    "args": {
      "External id": 59708,
      "cbid": 211, "correlation": 59708
    }
  },
  {
    "ph": "s", "id": 59708, "pid": 76337, "tid": -914061504, "ts": 1716454222267769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267843, "dur": 1,
    "args": {
      "External id": 59719,
      "cbid": 251, "correlation": 59719
    }
  },
  {
    "ph": "f", "id": 59719, "pid": 76337, "tid": -914061504, "ts": 1716454222267843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222270067, "dur": 203,
    "args": {
      "External id": 59720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59720,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59720, "pid": 5, "tid": 7, "ts": 1716454222270067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267847, "dur": 12,
    "args": {
      "External id": 59720,
      "cbid": 211, "correlation": 59720
    }
  },
  {
    "ph": "s", "id": 59720, "pid": 76337, "tid": -914061504, "ts": 1716454222267847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267913, "dur": 1,
    "args": {
      "External id": 59731,
      "cbid": 251, "correlation": 59731
    }
  },
  {
    "ph": "f", "id": 59731, "pid": 76337, "tid": -914061504, "ts": 1716454222267913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222270272, "dur": 203,
    "args": {
      "External id": 59732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59732,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59732, "pid": 5, "tid": 7, "ts": 1716454222270272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222267917, "dur": 11,
    "args": {
      "External id": 59732,
      "cbid": 211, "correlation": 59732
    }
  },
  {
    "ph": "s", "id": 59732, "pid": 76337, "tid": -914061504, "ts": 1716454222267917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222270477, "dur": 20585,
    "args": {
      "External id": 59753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59753,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 59753, "pid": 5, "tid": 7, "ts": 1716454222270477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268035, "dur": 17,
    "args": {
      "External id": 59753,
      "cbid": 211, "correlation": 59753
    }
  },
  {
    "ph": "s", "id": 59753, "pid": 76337, "tid": -914061504, "ts": 1716454222268035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268154, "dur": 2,
    "args": {
      "External id": 59771,
      "cbid": 251, "correlation": 59771
    }
  },
  {
    "ph": "f", "id": 59771, "pid": 76337, "tid": -914061504, "ts": 1716454222268154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222291063, "dur": 219,
    "args": {
      "External id": 59773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59773,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59773, "pid": 5, "tid": 7, "ts": 1716454222291063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268160, "dur": 14,
    "args": {
      "External id": 59773,
      "cbid": 211, "correlation": 59773
    }
  },
  {
    "ph": "s", "id": 59773, "pid": 76337, "tid": -914061504, "ts": 1716454222268160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222291284, "dur": 66,
    "args": {
      "External id": 59781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59781,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59781, "pid": 5, "tid": 7, "ts": 1716454222291284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268233, "dur": 15,
    "args": {
      "External id": 59781,
      "cbid": 211, "correlation": 59781
    }
  },
  {
    "ph": "s", "id": 59781, "pid": 76337, "tid": -914061504, "ts": 1716454222268233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222291352, "dur": 97,
    "args": {
      "External id": 59789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59789, "pid": 5, "tid": 7, "ts": 1716454222291352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268278, "dur": 9,
    "args": {
      "External id": 59789,
      "cbid": 211, "correlation": 59789
    }
  },
  {
    "ph": "s", "id": 59789, "pid": 76337, "tid": -914061504, "ts": 1716454222268278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222291449, "dur": 56,
    "args": {
      "External id": 59800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59800, "pid": 5, "tid": 7, "ts": 1716454222291449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268364, "dur": 14,
    "args": {
      "External id": 59800,
      "cbid": 211, "correlation": 59800
    }
  },
  {
    "ph": "s", "id": 59800, "pid": 76337, "tid": -914061504, "ts": 1716454222268364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222291507, "dur": 101,
    "args": {
      "External id": 59822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59822,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59822, "pid": 5, "tid": 7, "ts": 1716454222291507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268397, "dur": 9,
    "args": {
      "External id": 59822,
      "cbid": 211, "correlation": 59822
    }
  },
  {
    "ph": "s", "id": 59822, "pid": 76337, "tid": -914061504, "ts": 1716454222268397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268483, "dur": 1,
    "args": {
      "External id": 59833,
      "cbid": 251, "correlation": 59833
    }
  },
  {
    "ph": "f", "id": 59833, "pid": 76337, "tid": -914061504, "ts": 1716454222268483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222291610, "dur": 113,
    "args": {
      "External id": 59834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59834,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59834, "pid": 5, "tid": 7, "ts": 1716454222291610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268488, "dur": 12,
    "args": {
      "External id": 59834,
      "cbid": 211, "correlation": 59834
    }
  },
  {
    "ph": "s", "id": 59834, "pid": 76337, "tid": -914061504, "ts": 1716454222268488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268571, "dur": 1,
    "args": {
      "External id": 59845,
      "cbid": 251, "correlation": 59845
    }
  },
  {
    "ph": "f", "id": 59845, "pid": 76337, "tid": -914061504, "ts": 1716454222268571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268575, "dur": 0,
    "args": {
      "External id": 59846,
      "cbid": 251, "correlation": 59846
    }
  },
  {
    "ph": "f", "id": 59846, "pid": 76337, "tid": -914061504, "ts": 1716454222268575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222291724, "dur": 11,
    "args": {
      "External id": 59847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59847,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59847, "pid": 5, "tid": 7, "ts": 1716454222291724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268577, "dur": 14,
    "args": {
      "External id": 59847,
      "cbid": 211, "correlation": 59847
    }
  },
  {
    "ph": "s", "id": 59847, "pid": 76337, "tid": -914061504, "ts": 1716454222268577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222291736, "dur": 5,
    "args": {
      "External id": 59849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59849,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 59849, "pid": 5, "tid": 7, "ts": 1716454222291736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268595, "dur": 7,
    "args": {
      "External id": 59849,
      "cbid": 211, "correlation": 59849
    }
  },
  {
    "ph": "s", "id": 59849, "pid": 76337, "tid": -914061504, "ts": 1716454222268595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268661, "dur": 1,
    "args": {
      "External id": 59860,
      "cbid": 251, "correlation": 59860
    }
  },
  {
    "ph": "f", "id": 59860, "pid": 76337, "tid": -914061504, "ts": 1716454222268661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268664, "dur": 0,
    "args": {
      "External id": 59861,
      "cbid": 251, "correlation": 59861
    }
  },
  {
    "ph": "f", "id": 59861, "pid": 76337, "tid": -914061504, "ts": 1716454222268664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222291743, "dur": 7,
    "args": {
      "External id": 59862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59862,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 59862, "pid": 5, "tid": 7, "ts": 1716454222291743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268666, "dur": 12,
    "args": {
      "External id": 59862,
      "cbid": 211, "correlation": 59862
    }
  },
  {
    "ph": "s", "id": 59862, "pid": 76337, "tid": -914061504, "ts": 1716454222268666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222291751, "dur": 4,
    "args": {
      "External id": 59864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59864,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 59864, "pid": 5, "tid": 7, "ts": 1716454222291751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268679, "dur": 6,
    "args": {
      "External id": 59864,
      "cbid": 211, "correlation": 59864
    }
  },
  {
    "ph": "s", "id": 59864, "pid": 76337, "tid": -914061504, "ts": 1716454222268679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222291756, "dur": 170,
    "args": {
      "External id": 59885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59885,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 59885, "pid": 5, "tid": 7, "ts": 1716454222291756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268755, "dur": 12,
    "args": {
      "External id": 59885,
      "cbid": 211, "correlation": 59885
    }
  },
  {
    "ph": "s", "id": 59885, "pid": 76337, "tid": -914061504, "ts": 1716454222268755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268852, "dur": 2,
    "args": {
      "External id": 59903,
      "cbid": 251, "correlation": 59903
    }
  },
  {
    "ph": "f", "id": 59903, "pid": 76337, "tid": -914061504, "ts": 1716454222268852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222291927, "dur": 113,
    "args": {
      "External id": 59905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59905,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 59905, "pid": 5, "tid": 7, "ts": 1716454222291927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268859, "dur": 14,
    "args": {
      "External id": 59905,
      "cbid": 211, "correlation": 59905
    }
  },
  {
    "ph": "s", "id": 59905, "pid": 76337, "tid": -914061504, "ts": 1716454222268859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222292042, "dur": 35,
    "args": {
      "External id": 59913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59913,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59913, "pid": 5, "tid": 7, "ts": 1716454222292042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268930, "dur": 12,
    "args": {
      "External id": 59913,
      "cbid": 211, "correlation": 59913
    }
  },
  {
    "ph": "s", "id": 59913, "pid": 76337, "tid": -914061504, "ts": 1716454222268930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222292079, "dur": 72,
    "args": {
      "External id": 59921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59921, "pid": 5, "tid": 7, "ts": 1716454222292079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222268970, "dur": 18,
    "args": {
      "External id": 59921,
      "cbid": 211, "correlation": 59921
    }
  },
  {
    "ph": "s", "id": 59921, "pid": 76337, "tid": -914061504, "ts": 1716454222268970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222292152, "dur": 101,
    "args": {
      "External id": 59943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59943,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59943, "pid": 5, "tid": 7, "ts": 1716454222292152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269036, "dur": 11,
    "args": {
      "External id": 59943,
      "cbid": 211, "correlation": 59943
    }
  },
  {
    "ph": "s", "id": 59943, "pid": 76337, "tid": -914061504, "ts": 1716454222269036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269128, "dur": 1,
    "args": {
      "External id": 59959,
      "cbid": 251, "correlation": 59959
    }
  },
  {
    "ph": "f", "id": 59959, "pid": 76337, "tid": -914061504, "ts": 1716454222269128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222292255, "dur": 630,
    "args": {
      "External id": 59961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59961,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 59961, "pid": 5, "tid": 7, "ts": 1716454222292255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269134, "dur": 13,
    "args": {
      "External id": 59961,
      "cbid": 211, "correlation": 59961
    }
  },
  {
    "ph": "s", "id": 59961, "pid": 76337, "tid": -914061504, "ts": 1716454222269134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222292887, "dur": 262,
    "args": {
      "External id": 59969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59969,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59969, "pid": 5, "tid": 7, "ts": 1716454222292887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269215, "dur": 14,
    "args": {
      "External id": 59969,
      "cbid": 211, "correlation": 59969
    }
  },
  {
    "ph": "s", "id": 59969, "pid": 76337, "tid": -914061504, "ts": 1716454222269215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222293150, "dur": 260,
    "args": {
      "External id": 59977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 59977, "pid": 5, "tid": 7, "ts": 1716454222293150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269250, "dur": 10,
    "args": {
      "External id": 59977,
      "cbid": 211, "correlation": 59977
    }
  },
  {
    "ph": "s", "id": 59977, "pid": 76337, "tid": -914061504, "ts": 1716454222269250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269335, "dur": 2,
    "args": {
      "External id": 59993,
      "cbid": 251, "correlation": 59993
    }
  },
  {
    "ph": "f", "id": 59993, "pid": 76337, "tid": -914061504, "ts": 1716454222269335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269341, "dur": 0,
    "args": {
      "External id": 59995,
      "cbid": 251, "correlation": 59995
    }
  },
  {
    "ph": "f", "id": 59995, "pid": 76337, "tid": -914061504, "ts": 1716454222269341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222293412, "dur": 378,
    "args": {
      "External id": 59996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 59996,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 59996, "pid": 5, "tid": 7, "ts": 1716454222293412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269346, "dur": 13,
    "args": {
      "External id": 59996,
      "cbid": 211, "correlation": 59996
    }
  },
  {
    "ph": "s", "id": 59996, "pid": 76337, "tid": -914061504, "ts": 1716454222269346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222293791, "dur": 50,
    "args": {
      "External id": 60004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60004,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60004, "pid": 5, "tid": 7, "ts": 1716454222293791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269389, "dur": 10,
    "args": {
      "External id": 60004,
      "cbid": 211, "correlation": 60004
    }
  },
  {
    "ph": "s", "id": 60004, "pid": 76337, "tid": -914061504, "ts": 1716454222269389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222293843, "dur": 173,
    "args": {
      "External id": 60015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60015,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60015, "pid": 5, "tid": 7, "ts": 1716454222293843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269464, "dur": 13,
    "args": {
      "External id": 60015,
      "cbid": 211, "correlation": 60015
    }
  },
  {
    "ph": "s", "id": 60015, "pid": 76337, "tid": -914061504, "ts": 1716454222269464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269530, "dur": 0,
    "args": {
      "External id": 60027,
      "cbid": 317, "correlation": 60027
    }
  },
  {
    "ph": "f", "id": 60027, "pid": 76337, "tid": -914061504, "ts": 1716454222269530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269531, "dur": 0,
    "args": {
      "External id": 60028,
      "cbid": 203, "correlation": 60028
    }
  },
  {
    "ph": "f", "id": 60028, "pid": 76337, "tid": -914061504, "ts": 1716454222269531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269532, "dur": 0,
    "args": {
      "External id": 60029,
      "cbid": 205, "correlation": 60029
    }
  },
  {
    "ph": "f", "id": 60029, "pid": 76337, "tid": -914061504, "ts": 1716454222269532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269556, "dur": 1,
    "args": {
      "External id": 60033,
      "cbid": 251, "correlation": 60033
    }
  },
  {
    "ph": "f", "id": 60033, "pid": 76337, "tid": -914061504, "ts": 1716454222269556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269558, "dur": 0,
    "args": {
      "External id": 60034,
      "cbid": 251, "correlation": 60034
    }
  },
  {
    "ph": "f", "id": 60034, "pid": 76337, "tid": -914061504, "ts": 1716454222269558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269559, "dur": 0,
    "args": {
      "External id": 60035,
      "cbid": 251, "correlation": 60035
    }
  },
  {
    "ph": "f", "id": 60035, "pid": 76337, "tid": -914061504, "ts": 1716454222269559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269559, "dur": 0,
    "args": {
      "External id": 60036,
      "cbid": 251, "correlation": 60036
    }
  },
  {
    "ph": "f", "id": 60036, "pid": 76337, "tid": -914061504, "ts": 1716454222269559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269560, "dur": 0,
    "args": {
      "External id": 60037,
      "cbid": 251, "correlation": 60037
    }
  },
  {
    "ph": "f", "id": 60037, "pid": 76337, "tid": -914061504, "ts": 1716454222269560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269561, "dur": 0,
    "args": {
      "External id": 60038,
      "cbid": 251, "correlation": 60038
    }
  },
  {
    "ph": "f", "id": 60038, "pid": 76337, "tid": -914061504, "ts": 1716454222269561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269562, "dur": 0,
    "args": {
      "External id": 60039,
      "cbid": 251, "correlation": 60039
    }
  },
  {
    "ph": "f", "id": 60039, "pid": 76337, "tid": -914061504, "ts": 1716454222269562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269563, "dur": 0,
    "args": {
      "External id": 60040,
      "cbid": 251, "correlation": 60040
    }
  },
  {
    "ph": "f", "id": 60040, "pid": 76337, "tid": -914061504, "ts": 1716454222269563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269564, "dur": 0,
    "args": {
      "External id": 60041,
      "cbid": 251, "correlation": 60041
    }
  },
  {
    "ph": "f", "id": 60041, "pid": 76337, "tid": -914061504, "ts": 1716454222269564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222294017, "dur": 121,
    "args": {
      "External id": 60042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60042,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60042, "pid": 5, "tid": 7, "ts": 1716454222294017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269566, "dur": 12,
    "args": {
      "External id": 60042,
      "cbid": 211, "correlation": 60042
    }
  },
  {
    "ph": "s", "id": 60042, "pid": 76337, "tid": -914061504, "ts": 1716454222269566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222294140, "dur": 63,
    "args": {
      "External id": 60048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60048,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60048, "pid": 5, "tid": 7, "ts": 1716454222294140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269601, "dur": 9,
    "args": {
      "External id": 60048,
      "cbid": 211, "correlation": 60048
    }
  },
  {
    "ph": "s", "id": 60048, "pid": 76337, "tid": -914061504, "ts": 1716454222269601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222294205, "dur": 51,
    "args": {
      "External id": 60056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60056,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60056, "pid": 5, "tid": 7, "ts": 1716454222294205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269634, "dur": 8,
    "args": {
      "External id": 60056,
      "cbid": 211, "correlation": 60056
    }
  },
  {
    "ph": "s", "id": 60056, "pid": 76337, "tid": -914061504, "ts": 1716454222269634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222294257, "dur": 55,
    "args": {
      "External id": 60076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60076,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 60076, "pid": 5, "tid": 7, "ts": 1716454222294257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269706, "dur": 11,
    "args": {
      "External id": 60076,
      "cbid": 211, "correlation": 60076
    }
  },
  {
    "ph": "s", "id": 60076, "pid": 76337, "tid": -914061504, "ts": 1716454222269706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222294313, "dur": 5,
    "args": {
      "External id": 60088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60088,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60088, "pid": 5, "tid": 7, "ts": 1716454222294313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269727, "dur": 10,
    "args": {
      "External id": 60088,
      "cbid": 211, "correlation": 60088
    }
  },
  {
    "ph": "s", "id": 60088, "pid": 76337, "tid": -914061504, "ts": 1716454222269727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222294319, "dur": 60,
    "args": {
      "External id": 60091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60091, "pid": 5, "tid": 7, "ts": 1716454222294319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269749, "dur": 7,
    "args": {
      "External id": 60091,
      "cbid": 211, "correlation": 60091
    }
  },
  {
    "ph": "s", "id": 60091, "pid": 76337, "tid": -914061504, "ts": 1716454222269749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222294380, "dur": 38,
    "args": {
      "External id": 60100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60100,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60100, "pid": 5, "tid": 7, "ts": 1716454222294380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269791, "dur": 10,
    "args": {
      "External id": 60100,
      "cbid": 211, "correlation": 60100
    }
  },
  {
    "ph": "s", "id": 60100, "pid": 76337, "tid": -914061504, "ts": 1716454222269791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269843, "dur": 0,
    "args": {
      "External id": 60110,
      "cbid": 317, "correlation": 60110
    }
  },
  {
    "ph": "f", "id": 60110, "pid": 76337, "tid": -914061504, "ts": 1716454222269843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269844, "dur": 0,
    "args": {
      "External id": 60111,
      "cbid": 203, "correlation": 60111
    }
  },
  {
    "ph": "f", "id": 60111, "pid": 76337, "tid": -914061504, "ts": 1716454222269844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269845, "dur": 0,
    "args": {
      "External id": 60112,
      "cbid": 205, "correlation": 60112
    }
  },
  {
    "ph": "f", "id": 60112, "pid": 76337, "tid": -914061504, "ts": 1716454222269845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222294419, "dur": 42,
    "args": {
      "External id": 60116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60116, "pid": 5, "tid": 7, "ts": 1716454222294419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269860, "dur": 12,
    "args": {
      "External id": 60116,
      "cbid": 211, "correlation": 60116
    }
  },
  {
    "ph": "s", "id": 60116, "pid": 76337, "tid": -914061504, "ts": 1716454222269860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222294463, "dur": 16,
    "args": {
      "External id": 60118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60118, "pid": 5, "tid": 7, "ts": 1716454222294463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269874, "dur": 5,
    "args": {
      "External id": 60118,
      "cbid": 211, "correlation": 60118
    }
  },
  {
    "ph": "s", "id": 60118, "pid": 76337, "tid": -914061504, "ts": 1716454222269874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222294480, "dur": 4,
    "args": {
      "External id": 60120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60120, "pid": 5, "tid": 7, "ts": 1716454222294480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269884, "dur": 6,
    "args": {
      "External id": 60120,
      "cbid": 211, "correlation": 60120
    }
  },
  {
    "ph": "s", "id": 60120, "pid": 76337, "tid": -914061504, "ts": 1716454222269884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269893, "dur": 0,
    "args": {
      "External id": 60121,
      "cbid": 51, "correlation": 60121
    }
  },
  {
    "ph": "s", "id": 60121, "pid": 76337, "tid": -914061504, "ts": 1716454222269893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222294485, "dur": 769,
    "args": {
      "External id": 60122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60122,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60122, "pid": 5, "tid": 7, "ts": 1716454222294485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269893, "dur": 5,
    "args": {
      "External id": 60122,
      "cbid": 211, "correlation": 60122
    }
  },
  {
    "ph": "s", "id": 60122, "pid": 76337, "tid": -914061504, "ts": 1716454222269893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222295256, "dur": 63,
    "args": {
      "External id": 60127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60127, "pid": 5, "tid": 7, "ts": 1716454222295256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269921, "dur": 8,
    "args": {
      "External id": 60127,
      "cbid": 211, "correlation": 60127
    }
  },
  {
    "ph": "s", "id": 60127, "pid": 76337, "tid": -914061504, "ts": 1716454222269921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222295320, "dur": 4,
    "args": {
      "External id": 60135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60135,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60135, "pid": 5, "tid": 7, "ts": 1716454222295320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222269964, "dur": 17,
    "args": {
      "External id": 60135,
      "cbid": 211, "correlation": 60135
    }
  },
  {
    "ph": "s", "id": 60135, "pid": 76337, "tid": -914061504, "ts": 1716454222269964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270038, "dur": 1,
    "args": {
      "External id": 60151,
      "cbid": 251, "correlation": 60151
    }
  },
  {
    "ph": "f", "id": 60151, "pid": 76337, "tid": -914061504, "ts": 1716454222270038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270047, "dur": 0,
    "args": {
      "External id": 60153,
      "cbid": 251, "correlation": 60153
    }
  },
  {
    "ph": "f", "id": 60153, "pid": 76337, "tid": -914061504, "ts": 1716454222270047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222295325, "dur": 12,
    "args": {
      "External id": 60154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60154,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 60154, "pid": 5, "tid": 7, "ts": 1716454222295325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270048, "dur": 12,
    "args": {
      "External id": 60154,
      "cbid": 211, "correlation": 60154
    }
  },
  {
    "ph": "s", "id": 60154, "pid": 76337, "tid": -914061504, "ts": 1716454222270048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222295339, "dur": 6,
    "args": {
      "External id": 60156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60156,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 60156, "pid": 5, "tid": 7, "ts": 1716454222295339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270062, "dur": 5,
    "args": {
      "External id": 60156,
      "cbid": 211, "correlation": 60156
    }
  },
  {
    "ph": "s", "id": 60156, "pid": 76337, "tid": -914061504, "ts": 1716454222270062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222295346, "dur": 56,
    "args": {
      "External id": 60166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60166, "pid": 5, "tid": 7, "ts": 1716454222295346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270120, "dur": 12,
    "args": {
      "External id": 60166,
      "cbid": 211, "correlation": 60166
    }
  },
  {
    "ph": "s", "id": 60166, "pid": 76337, "tid": -914061504, "ts": 1716454222270120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222295403, "dur": 54,
    "args": {
      "External id": 60186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60186,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 60186, "pid": 5, "tid": 7, "ts": 1716454222295403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270185, "dur": 11,
    "args": {
      "External id": 60186,
      "cbid": 211, "correlation": 60186
    }
  },
  {
    "ph": "s", "id": 60186, "pid": 76337, "tid": -914061504, "ts": 1716454222270185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222295458, "dur": 4,
    "args": {
      "External id": 60198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60198,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60198, "pid": 5, "tid": 7, "ts": 1716454222295458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270206, "dur": 6,
    "args": {
      "External id": 60198,
      "cbid": 211, "correlation": 60198
    }
  },
  {
    "ph": "s", "id": 60198, "pid": 76337, "tid": -914061504, "ts": 1716454222270206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222295464, "dur": 60,
    "args": {
      "External id": 60201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60201,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60201, "pid": 5, "tid": 7, "ts": 1716454222295464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270224, "dur": 6,
    "args": {
      "External id": 60201,
      "cbid": 211, "correlation": 60201
    }
  },
  {
    "ph": "s", "id": 60201, "pid": 76337, "tid": -914061504, "ts": 1716454222270224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222295525, "dur": 38,
    "args": {
      "External id": 60210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60210,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60210, "pid": 5, "tid": 7, "ts": 1716454222295525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270265, "dur": 10,
    "args": {
      "External id": 60210,
      "cbid": 211, "correlation": 60210
    }
  },
  {
    "ph": "s", "id": 60210, "pid": 76337, "tid": -914061504, "ts": 1716454222270265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270329, "dur": 0,
    "args": {
      "External id": 60220,
      "cbid": 317, "correlation": 60220
    }
  },
  {
    "ph": "f", "id": 60220, "pid": 76337, "tid": -914061504, "ts": 1716454222270329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270330, "dur": 0,
    "args": {
      "External id": 60221,
      "cbid": 203, "correlation": 60221
    }
  },
  {
    "ph": "f", "id": 60221, "pid": 76337, "tid": -914061504, "ts": 1716454222270330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270330, "dur": 0,
    "args": {
      "External id": 60222,
      "cbid": 205, "correlation": 60222
    }
  },
  {
    "ph": "f", "id": 60222, "pid": 76337, "tid": -914061504, "ts": 1716454222270330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222295564, "dur": 40,
    "args": {
      "External id": 60226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60226,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60226, "pid": 5, "tid": 7, "ts": 1716454222295564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270347, "dur": 12,
    "args": {
      "External id": 60226,
      "cbid": 211, "correlation": 60226
    }
  },
  {
    "ph": "s", "id": 60226, "pid": 76337, "tid": -914061504, "ts": 1716454222270347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222295606, "dur": 16,
    "args": {
      "External id": 60228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60228,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60228, "pid": 5, "tid": 7, "ts": 1716454222295606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270362, "dur": 6,
    "args": {
      "External id": 60228,
      "cbid": 211, "correlation": 60228
    }
  },
  {
    "ph": "s", "id": 60228, "pid": 76337, "tid": -914061504, "ts": 1716454222270362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222295623, "dur": 4,
    "args": {
      "External id": 60230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60230, "pid": 5, "tid": 7, "ts": 1716454222295623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270372, "dur": 5,
    "args": {
      "External id": 60230,
      "cbid": 211, "correlation": 60230
    }
  },
  {
    "ph": "s", "id": 60230, "pid": 76337, "tid": -914061504, "ts": 1716454222270372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270381, "dur": 0,
    "args": {
      "External id": 60231,
      "cbid": 51, "correlation": 60231
    }
  },
  {
    "ph": "s", "id": 60231, "pid": 76337, "tid": -914061504, "ts": 1716454222270381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222295628, "dur": 764,
    "args": {
      "External id": 60232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60232,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60232, "pid": 5, "tid": 7, "ts": 1716454222295628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270381, "dur": 5,
    "args": {
      "External id": 60232,
      "cbid": 211, "correlation": 60232
    }
  },
  {
    "ph": "s", "id": 60232, "pid": 76337, "tid": -914061504, "ts": 1716454222270381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222296393, "dur": 62,
    "args": {
      "External id": 60237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60237, "pid": 5, "tid": 7, "ts": 1716454222296393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270409, "dur": 8,
    "args": {
      "External id": 60237,
      "cbid": 211, "correlation": 60237
    }
  },
  {
    "ph": "s", "id": 60237, "pid": 76337, "tid": -914061504, "ts": 1716454222270409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222296457, "dur": 50,
    "args": {
      "External id": 60245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60245,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60245, "pid": 5, "tid": 7, "ts": 1716454222296457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270441, "dur": 8,
    "args": {
      "External id": 60245,
      "cbid": 211, "correlation": 60245
    }
  },
  {
    "ph": "s", "id": 60245, "pid": 76337, "tid": -914061504, "ts": 1716454222270441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222296508, "dur": 36,
    "args": {
      "External id": 60253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60253,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60253, "pid": 5, "tid": 7, "ts": 1716454222296508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270470, "dur": 8,
    "args": {
      "External id": 60253,
      "cbid": 211, "correlation": 60253
    }
  },
  {
    "ph": "s", "id": 60253, "pid": 76337, "tid": -914061504, "ts": 1716454222270470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222296545, "dur": 55,
    "args": {
      "External id": 60273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60273,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 60273, "pid": 5, "tid": 7, "ts": 1716454222296545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270550, "dur": 12,
    "args": {
      "External id": 60273,
      "cbid": 211, "correlation": 60273
    }
  },
  {
    "ph": "s", "id": 60273, "pid": 76337, "tid": -914061504, "ts": 1716454222270550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222296602, "dur": 4,
    "args": {
      "External id": 60285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60285,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60285, "pid": 5, "tid": 7, "ts": 1716454222296602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270572, "dur": 6,
    "args": {
      "External id": 60285,
      "cbid": 211, "correlation": 60285
    }
  },
  {
    "ph": "s", "id": 60285, "pid": 76337, "tid": -914061504, "ts": 1716454222270572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222296608, "dur": 58,
    "args": {
      "External id": 60288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60288, "pid": 5, "tid": 7, "ts": 1716454222296608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270592, "dur": 7,
    "args": {
      "External id": 60288,
      "cbid": 211, "correlation": 60288
    }
  },
  {
    "ph": "s", "id": 60288, "pid": 76337, "tid": -914061504, "ts": 1716454222270592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270652, "dur": 0,
    "args": {
      "External id": 60299,
      "cbid": 317, "correlation": 60299
    }
  },
  {
    "ph": "f", "id": 60299, "pid": 76337, "tid": -914061504, "ts": 1716454222270652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270653, "dur": 0,
    "args": {
      "External id": 60300,
      "cbid": 203, "correlation": 60300
    }
  },
  {
    "ph": "f", "id": 60300, "pid": 76337, "tid": -914061504, "ts": 1716454222270653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270653, "dur": 0,
    "args": {
      "External id": 60301,
      "cbid": 205, "correlation": 60301
    }
  },
  {
    "ph": "f", "id": 60301, "pid": 76337, "tid": -914061504, "ts": 1716454222270653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270676, "dur": 1,
    "args": {
      "External id": 60305,
      "cbid": 251, "correlation": 60305
    }
  },
  {
    "ph": "f", "id": 60305, "pid": 76337, "tid": -914061504, "ts": 1716454222270676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270678, "dur": 0,
    "args": {
      "External id": 60306,
      "cbid": 251, "correlation": 60306
    }
  },
  {
    "ph": "f", "id": 60306, "pid": 76337, "tid": -914061504, "ts": 1716454222270678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270679, "dur": 0,
    "args": {
      "External id": 60307,
      "cbid": 251, "correlation": 60307
    }
  },
  {
    "ph": "f", "id": 60307, "pid": 76337, "tid": -914061504, "ts": 1716454222270679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270680, "dur": 0,
    "args": {
      "External id": 60308,
      "cbid": 251, "correlation": 60308
    }
  },
  {
    "ph": "f", "id": 60308, "pid": 76337, "tid": -914061504, "ts": 1716454222270680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270680, "dur": 0,
    "args": {
      "External id": 60309,
      "cbid": 251, "correlation": 60309
    }
  },
  {
    "ph": "f", "id": 60309, "pid": 76337, "tid": -914061504, "ts": 1716454222270680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270681, "dur": 0,
    "args": {
      "External id": 60310,
      "cbid": 251, "correlation": 60310
    }
  },
  {
    "ph": "f", "id": 60310, "pid": 76337, "tid": -914061504, "ts": 1716454222270681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270682, "dur": 0,
    "args": {
      "External id": 60311,
      "cbid": 251, "correlation": 60311
    }
  },
  {
    "ph": "f", "id": 60311, "pid": 76337, "tid": -914061504, "ts": 1716454222270682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270683, "dur": 0,
    "args": {
      "External id": 60312,
      "cbid": 251, "correlation": 60312
    }
  },
  {
    "ph": "f", "id": 60312, "pid": 76337, "tid": -914061504, "ts": 1716454222270683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270684, "dur": 0,
    "args": {
      "External id": 60313,
      "cbid": 251, "correlation": 60313
    }
  },
  {
    "ph": "f", "id": 60313, "pid": 76337, "tid": -914061504, "ts": 1716454222270684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222296667, "dur": 121,
    "args": {
      "External id": 60314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60314,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60314, "pid": 5, "tid": 7, "ts": 1716454222296667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270686, "dur": 13,
    "args": {
      "External id": 60314,
      "cbid": 211, "correlation": 60314
    }
  },
  {
    "ph": "s", "id": 60314, "pid": 76337, "tid": -914061504, "ts": 1716454222270686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222296790, "dur": 63,
    "args": {
      "External id": 60320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60320, "pid": 5, "tid": 7, "ts": 1716454222296790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270722, "dur": 9,
    "args": {
      "External id": 60320,
      "cbid": 211, "correlation": 60320
    }
  },
  {
    "ph": "s", "id": 60320, "pid": 76337, "tid": -914061504, "ts": 1716454222270722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222296855, "dur": 675,
    "args": {
      "External id": 60329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60329,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60329, "pid": 5, "tid": 7, "ts": 1716454222296855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270806, "dur": 14,
    "args": {
      "External id": 60329,
      "cbid": 211, "correlation": 60329
    }
  },
  {
    "ph": "s", "id": 60329, "pid": 76337, "tid": -914061504, "ts": 1716454222270806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222297531, "dur": 199,
    "args": {
      "External id": 60351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60351,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60351, "pid": 5, "tid": 7, "ts": 1716454222297531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270862, "dur": 10,
    "args": {
      "External id": 60351,
      "cbid": 211, "correlation": 60351
    }
  },
  {
    "ph": "s", "id": 60351, "pid": 76337, "tid": -914061504, "ts": 1716454222270862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270948, "dur": 1,
    "args": {
      "External id": 60362,
      "cbid": 251, "correlation": 60362
    }
  },
  {
    "ph": "f", "id": 60362, "pid": 76337, "tid": -914061504, "ts": 1716454222270948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222297731, "dur": 217,
    "args": {
      "External id": 60363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60363,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60363, "pid": 5, "tid": 7, "ts": 1716454222297731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222270953, "dur": 16,
    "args": {
      "External id": 60363,
      "cbid": 211, "correlation": 60363
    }
  },
  {
    "ph": "s", "id": 60363, "pid": 76337, "tid": -914061504, "ts": 1716454222270953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271034, "dur": 1,
    "args": {
      "External id": 60374,
      "cbid": 251, "correlation": 60374
    }
  },
  {
    "ph": "f", "id": 60374, "pid": 76337, "tid": -914061504, "ts": 1716454222271034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222297950, "dur": 202,
    "args": {
      "External id": 60375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60375,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60375, "pid": 5, "tid": 7, "ts": 1716454222297950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271038, "dur": 12,
    "args": {
      "External id": 60375,
      "cbid": 211, "correlation": 60375
    }
  },
  {
    "ph": "s", "id": 60375, "pid": 76337, "tid": -914061504, "ts": 1716454222271038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271102, "dur": 1,
    "args": {
      "External id": 60386,
      "cbid": 251, "correlation": 60386
    }
  },
  {
    "ph": "f", "id": 60386, "pid": 76337, "tid": -914061504, "ts": 1716454222271102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222298153, "dur": 202,
    "args": {
      "External id": 60387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60387,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60387, "pid": 5, "tid": 7, "ts": 1716454222298153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271106, "dur": 11,
    "args": {
      "External id": 60387,
      "cbid": 211, "correlation": 60387
    }
  },
  {
    "ph": "s", "id": 60387, "pid": 76337, "tid": -914061504, "ts": 1716454222271106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222298356, "dur": 20584,
    "args": {
      "External id": 60408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60408,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60408, "pid": 5, "tid": 7, "ts": 1716454222298356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271186, "dur": 12,
    "args": {
      "External id": 60408,
      "cbid": 211, "correlation": 60408
    }
  },
  {
    "ph": "s", "id": 60408, "pid": 76337, "tid": -914061504, "ts": 1716454222271186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271283, "dur": 1,
    "args": {
      "External id": 60426,
      "cbid": 251, "correlation": 60426
    }
  },
  {
    "ph": "f", "id": 60426, "pid": 76337, "tid": -914061504, "ts": 1716454222271283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222318942, "dur": 218,
    "args": {
      "External id": 60428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60428,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60428, "pid": 5, "tid": 7, "ts": 1716454222318942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271288, "dur": 14,
    "args": {
      "External id": 60428,
      "cbid": 211, "correlation": 60428
    }
  },
  {
    "ph": "s", "id": 60428, "pid": 76337, "tid": -914061504, "ts": 1716454222271288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222319161, "dur": 66,
    "args": {
      "External id": 60436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60436,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60436, "pid": 5, "tid": 7, "ts": 1716454222319161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271358, "dur": 12,
    "args": {
      "External id": 60436,
      "cbid": 211, "correlation": 60436
    }
  },
  {
    "ph": "s", "id": 60436, "pid": 76337, "tid": -914061504, "ts": 1716454222271358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222319228, "dur": 99,
    "args": {
      "External id": 60444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60444,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60444, "pid": 5, "tid": 7, "ts": 1716454222319228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271400, "dur": 10,
    "args": {
      "External id": 60444,
      "cbid": 211, "correlation": 60444
    }
  },
  {
    "ph": "s", "id": 60444, "pid": 76337, "tid": -914061504, "ts": 1716454222271400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222319328, "dur": 56,
    "args": {
      "External id": 60455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60455, "pid": 5, "tid": 7, "ts": 1716454222319328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271473, "dur": 13,
    "args": {
      "External id": 60455,
      "cbid": 211, "correlation": 60455
    }
  },
  {
    "ph": "s", "id": 60455, "pid": 76337, "tid": -914061504, "ts": 1716454222271473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222319386, "dur": 101,
    "args": {
      "External id": 60477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60477,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60477, "pid": 5, "tid": 7, "ts": 1716454222319386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271504, "dur": 9,
    "args": {
      "External id": 60477,
      "cbid": 211, "correlation": 60477
    }
  },
  {
    "ph": "s", "id": 60477, "pid": 76337, "tid": -914061504, "ts": 1716454222271504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271588, "dur": 1,
    "args": {
      "External id": 60488,
      "cbid": 251, "correlation": 60488
    }
  },
  {
    "ph": "f", "id": 60488, "pid": 76337, "tid": -914061504, "ts": 1716454222271588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222319488, "dur": 111,
    "args": {
      "External id": 60489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60489,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60489, "pid": 5, "tid": 7, "ts": 1716454222319488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271593, "dur": 12,
    "args": {
      "External id": 60489,
      "cbid": 211, "correlation": 60489
    }
  },
  {
    "ph": "s", "id": 60489, "pid": 76337, "tid": -914061504, "ts": 1716454222271593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271664, "dur": 1,
    "args": {
      "External id": 60500,
      "cbid": 251, "correlation": 60500
    }
  },
  {
    "ph": "f", "id": 60500, "pid": 76337, "tid": -914061504, "ts": 1716454222271664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271668, "dur": 0,
    "args": {
      "External id": 60501,
      "cbid": 251, "correlation": 60501
    }
  },
  {
    "ph": "f", "id": 60501, "pid": 76337, "tid": -914061504, "ts": 1716454222271668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222319600, "dur": 10,
    "args": {
      "External id": 60502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60502,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 60502, "pid": 5, "tid": 7, "ts": 1716454222319600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271670, "dur": 12,
    "args": {
      "External id": 60502,
      "cbid": 211, "correlation": 60502
    }
  },
  {
    "ph": "s", "id": 60502, "pid": 76337, "tid": -914061504, "ts": 1716454222271670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222319612, "dur": 6,
    "args": {
      "External id": 60504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60504,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 60504, "pid": 5, "tid": 7, "ts": 1716454222319612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271683, "dur": 6,
    "args": {
      "External id": 60504,
      "cbid": 211, "correlation": 60504
    }
  },
  {
    "ph": "s", "id": 60504, "pid": 76337, "tid": -914061504, "ts": 1716454222271683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271745, "dur": 1,
    "args": {
      "External id": 60515,
      "cbid": 251, "correlation": 60515
    }
  },
  {
    "ph": "f", "id": 60515, "pid": 76337, "tid": -914061504, "ts": 1716454222271745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271748, "dur": 0,
    "args": {
      "External id": 60516,
      "cbid": 251, "correlation": 60516
    }
  },
  {
    "ph": "f", "id": 60516, "pid": 76337, "tid": -914061504, "ts": 1716454222271748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222319619, "dur": 7,
    "args": {
      "External id": 60517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60517,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 60517, "pid": 5, "tid": 7, "ts": 1716454222319619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271749, "dur": 14,
    "args": {
      "External id": 60517,
      "cbid": 211, "correlation": 60517
    }
  },
  {
    "ph": "s", "id": 60517, "pid": 76337, "tid": -914061504, "ts": 1716454222271749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222319628, "dur": 4,
    "args": {
      "External id": 60519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60519,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 60519, "pid": 5, "tid": 7, "ts": 1716454222319628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271765, "dur": 6,
    "args": {
      "External id": 60519,
      "cbid": 211, "correlation": 60519
    }
  },
  {
    "ph": "s", "id": 60519, "pid": 76337, "tid": -914061504, "ts": 1716454222271765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222319633, "dur": 169,
    "args": {
      "External id": 60540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60540,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60540, "pid": 5, "tid": 7, "ts": 1716454222319633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271840, "dur": 12,
    "args": {
      "External id": 60540,
      "cbid": 211, "correlation": 60540
    }
  },
  {
    "ph": "s", "id": 60540, "pid": 76337, "tid": -914061504, "ts": 1716454222271840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271935, "dur": 1,
    "args": {
      "External id": 60558,
      "cbid": 251, "correlation": 60558
    }
  },
  {
    "ph": "f", "id": 60558, "pid": 76337, "tid": -914061504, "ts": 1716454222271935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222319804, "dur": 116,
    "args": {
      "External id": 60560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60560,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60560, "pid": 5, "tid": 7, "ts": 1716454222319804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222271941, "dur": 13,
    "args": {
      "External id": 60560,
      "cbid": 211, "correlation": 60560
    }
  },
  {
    "ph": "s", "id": 60560, "pid": 76337, "tid": -914061504, "ts": 1716454222271941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222319921, "dur": 35,
    "args": {
      "External id": 60568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60568,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60568, "pid": 5, "tid": 7, "ts": 1716454222319921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272017, "dur": 12,
    "args": {
      "External id": 60568,
      "cbid": 211, "correlation": 60568
    }
  },
  {
    "ph": "s", "id": 60568, "pid": 76337, "tid": -914061504, "ts": 1716454222272017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222319957, "dur": 70,
    "args": {
      "External id": 60576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60576, "pid": 5, "tid": 7, "ts": 1716454222319957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272059, "dur": 9,
    "args": {
      "External id": 60576,
      "cbid": 211, "correlation": 60576
    }
  },
  {
    "ph": "s", "id": 60576, "pid": 76337, "tid": -914061504, "ts": 1716454222272059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222320029, "dur": 101,
    "args": {
      "External id": 60598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60598,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60598, "pid": 5, "tid": 7, "ts": 1716454222320029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272113, "dur": 10,
    "args": {
      "External id": 60598,
      "cbid": 211, "correlation": 60598
    }
  },
  {
    "ph": "s", "id": 60598, "pid": 76337, "tid": -914061504, "ts": 1716454222272113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272201, "dur": 1,
    "args": {
      "External id": 60614,
      "cbid": 251, "correlation": 60614
    }
  },
  {
    "ph": "f", "id": 60614, "pid": 76337, "tid": -914061504, "ts": 1716454222272201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222320132, "dur": 629,
    "args": {
      "External id": 60616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60616,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60616, "pid": 5, "tid": 7, "ts": 1716454222320132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272206, "dur": 12,
    "args": {
      "External id": 60616,
      "cbid": 211, "correlation": 60616
    }
  },
  {
    "ph": "s", "id": 60616, "pid": 76337, "tid": -914061504, "ts": 1716454222272206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222320762, "dur": 262,
    "args": {
      "External id": 60624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60624, "pid": 5, "tid": 7, "ts": 1716454222320762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272271, "dur": 12,
    "args": {
      "External id": 60624,
      "cbid": 211, "correlation": 60624
    }
  },
  {
    "ph": "s", "id": 60624, "pid": 76337, "tid": -914061504, "ts": 1716454222272271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222321025, "dur": 260,
    "args": {
      "External id": 60632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60632,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60632, "pid": 5, "tid": 7, "ts": 1716454222321025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272302, "dur": 8,
    "args": {
      "External id": 60632,
      "cbid": 211, "correlation": 60632
    }
  },
  {
    "ph": "s", "id": 60632, "pid": 76337, "tid": -914061504, "ts": 1716454222272302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272382, "dur": 1,
    "args": {
      "External id": 60648,
      "cbid": 251, "correlation": 60648
    }
  },
  {
    "ph": "f", "id": 60648, "pid": 76337, "tid": -914061504, "ts": 1716454222272382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272388, "dur": 0,
    "args": {
      "External id": 60650,
      "cbid": 251, "correlation": 60650
    }
  },
  {
    "ph": "f", "id": 60650, "pid": 76337, "tid": -914061504, "ts": 1716454222272388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222321287, "dur": 379,
    "args": {
      "External id": 60651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60651,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 60651, "pid": 5, "tid": 7, "ts": 1716454222321287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272390, "dur": 13,
    "args": {
      "External id": 60651,
      "cbid": 211, "correlation": 60651
    }
  },
  {
    "ph": "s", "id": 60651, "pid": 76337, "tid": -914061504, "ts": 1716454222272390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222321667, "dur": 50,
    "args": {
      "External id": 60659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60659,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60659, "pid": 5, "tid": 7, "ts": 1716454222321667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272432, "dur": 10,
    "args": {
      "External id": 60659,
      "cbid": 211, "correlation": 60659
    }
  },
  {
    "ph": "s", "id": 60659, "pid": 76337, "tid": -914061504, "ts": 1716454222272432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222321719, "dur": 171,
    "args": {
      "External id": 60670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60670,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60670, "pid": 5, "tid": 7, "ts": 1716454222321719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272500, "dur": 15,
    "args": {
      "External id": 60670,
      "cbid": 211, "correlation": 60670
    }
  },
  {
    "ph": "s", "id": 60670, "pid": 76337, "tid": -914061504, "ts": 1716454222272500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272566, "dur": 0,
    "args": {
      "External id": 60682,
      "cbid": 317, "correlation": 60682
    }
  },
  {
    "ph": "f", "id": 60682, "pid": 76337, "tid": -914061504, "ts": 1716454222272566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272567, "dur": 0,
    "args": {
      "External id": 60683,
      "cbid": 203, "correlation": 60683
    }
  },
  {
    "ph": "f", "id": 60683, "pid": 76337, "tid": -914061504, "ts": 1716454222272567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272568, "dur": 0,
    "args": {
      "External id": 60684,
      "cbid": 205, "correlation": 60684
    }
  },
  {
    "ph": "f", "id": 60684, "pid": 76337, "tid": -914061504, "ts": 1716454222272568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272590, "dur": 1,
    "args": {
      "External id": 60688,
      "cbid": 251, "correlation": 60688
    }
  },
  {
    "ph": "f", "id": 60688, "pid": 76337, "tid": -914061504, "ts": 1716454222272590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272592, "dur": 0,
    "args": {
      "External id": 60689,
      "cbid": 251, "correlation": 60689
    }
  },
  {
    "ph": "f", "id": 60689, "pid": 76337, "tid": -914061504, "ts": 1716454222272592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272592, "dur": 0,
    "args": {
      "External id": 60690,
      "cbid": 251, "correlation": 60690
    }
  },
  {
    "ph": "f", "id": 60690, "pid": 76337, "tid": -914061504, "ts": 1716454222272592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272593, "dur": 0,
    "args": {
      "External id": 60691,
      "cbid": 251, "correlation": 60691
    }
  },
  {
    "ph": "f", "id": 60691, "pid": 76337, "tid": -914061504, "ts": 1716454222272593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272594, "dur": 0,
    "args": {
      "External id": 60692,
      "cbid": 251, "correlation": 60692
    }
  },
  {
    "ph": "f", "id": 60692, "pid": 76337, "tid": -914061504, "ts": 1716454222272594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272594, "dur": 0,
    "args": {
      "External id": 60693,
      "cbid": 251, "correlation": 60693
    }
  },
  {
    "ph": "f", "id": 60693, "pid": 76337, "tid": -914061504, "ts": 1716454222272594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272595, "dur": 0,
    "args": {
      "External id": 60694,
      "cbid": 251, "correlation": 60694
    }
  },
  {
    "ph": "f", "id": 60694, "pid": 76337, "tid": -914061504, "ts": 1716454222272595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272596, "dur": 0,
    "args": {
      "External id": 60695,
      "cbid": 251, "correlation": 60695
    }
  },
  {
    "ph": "f", "id": 60695, "pid": 76337, "tid": -914061504, "ts": 1716454222272596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272597, "dur": 0,
    "args": {
      "External id": 60696,
      "cbid": 251, "correlation": 60696
    }
  },
  {
    "ph": "f", "id": 60696, "pid": 76337, "tid": -914061504, "ts": 1716454222272597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222321892, "dur": 122,
    "args": {
      "External id": 60697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60697,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60697, "pid": 5, "tid": 7, "ts": 1716454222321892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272599, "dur": 13,
    "args": {
      "External id": 60697,
      "cbid": 211, "correlation": 60697
    }
  },
  {
    "ph": "s", "id": 60697, "pid": 76337, "tid": -914061504, "ts": 1716454222272599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222322016, "dur": 63,
    "args": {
      "External id": 60703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60703, "pid": 5, "tid": 7, "ts": 1716454222322016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272634, "dur": 8,
    "args": {
      "External id": 60703,
      "cbid": 211, "correlation": 60703
    }
  },
  {
    "ph": "s", "id": 60703, "pid": 76337, "tid": -914061504, "ts": 1716454222272634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222322080, "dur": 50,
    "args": {
      "External id": 60711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60711,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60711, "pid": 5, "tid": 7, "ts": 1716454222322080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272667, "dur": 9,
    "args": {
      "External id": 60711,
      "cbid": 211, "correlation": 60711
    }
  },
  {
    "ph": "s", "id": 60711, "pid": 76337, "tid": -914061504, "ts": 1716454222272667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272747, "dur": 0,
    "args": {
      "External id": 60721,
      "cbid": 317, "correlation": 60721
    }
  },
  {
    "ph": "f", "id": 60721, "pid": 76337, "tid": -914061504, "ts": 1716454222272747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272748, "dur": 0,
    "args": {
      "External id": 60722,
      "cbid": 203, "correlation": 60722
    }
  },
  {
    "ph": "f", "id": 60722, "pid": 76337, "tid": -914061504, "ts": 1716454222272748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272748, "dur": 0,
    "args": {
      "External id": 60723,
      "cbid": 205, "correlation": 60723
    }
  },
  {
    "ph": "f", "id": 60723, "pid": 76337, "tid": -914061504, "ts": 1716454222272748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222322131, "dur": 42,
    "args": {
      "External id": 60727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60727,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60727, "pid": 5, "tid": 7, "ts": 1716454222322131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272767, "dur": 13,
    "args": {
      "External id": 60727,
      "cbid": 211, "correlation": 60727
    }
  },
  {
    "ph": "s", "id": 60727, "pid": 76337, "tid": -914061504, "ts": 1716454222272767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222322175, "dur": 16,
    "args": {
      "External id": 60729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60729,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60729, "pid": 5, "tid": 7, "ts": 1716454222322175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272782, "dur": 5,
    "args": {
      "External id": 60729,
      "cbid": 211, "correlation": 60729
    }
  },
  {
    "ph": "s", "id": 60729, "pid": 76337, "tid": -914061504, "ts": 1716454222272782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222322193, "dur": 1,
    "args": {
      "External id": 60731,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 60731,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.8571428571428571
    }
  },
  {
    "ph": "f", "id": 60731, "pid": 5, "tid": 7, "ts": 1716454222322193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272801, "dur": 16,
    "args": {
      "External id": 60731,
      "cbid": 51, "correlation": 60731
    }
  },
  {
    "ph": "s", "id": 60731, "pid": 76337, "tid": -914061504, "ts": 1716454222272801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222322197, "dur": 395,
    "args": {
      "External id": 60732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60732,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60732, "pid": 5, "tid": 7, "ts": 1716454222322197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272844, "dur": 30,
    "args": {
      "External id": 60732,
      "cbid": 211, "correlation": 60732
    }
  },
  {
    "ph": "s", "id": 60732, "pid": 76337, "tid": -914061504, "ts": 1716454222272844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222322594, "dur": 13,
    "args": {
      "External id": 60734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60734, "pid": 5, "tid": 7, "ts": 1716454222322594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272883, "dur": 8,
    "args": {
      "External id": 60734,
      "cbid": 211, "correlation": 60734
    }
  },
  {
    "ph": "s", "id": 60734, "pid": 76337, "tid": -914061504, "ts": 1716454222272883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222322609, "dur": 16,
    "args": {
      "External id": 60740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60740, "pid": 5, "tid": 7, "ts": 1716454222322609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222272918, "dur": 9,
    "args": {
      "External id": 60740,
      "cbid": 211, "correlation": 60740
    }
  },
  {
    "ph": "s", "id": 60740, "pid": 76337, "tid": -914061504, "ts": 1716454222272918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222322627, "dur": 20,
    "args": {
      "External id": 60760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60760,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 60760, "pid": 5, "tid": 7, "ts": 1716454222322627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273021, "dur": 13,
    "args": {
      "External id": 60760,
      "cbid": 211, "correlation": 60760
    }
  },
  {
    "ph": "s", "id": 60760, "pid": 76337, "tid": -914061504, "ts": 1716454222273021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222322648, "dur": 5,
    "args": {
      "External id": 60772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60772,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60772, "pid": 5, "tid": 7, "ts": 1716454222322648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273043, "dur": 7,
    "args": {
      "External id": 60772,
      "cbid": 211, "correlation": 60772
    }
  },
  {
    "ph": "s", "id": 60772, "pid": 76337, "tid": -914061504, "ts": 1716454222273043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222322654, "dur": 19,
    "args": {
      "External id": 60775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60775, "pid": 5, "tid": 7, "ts": 1716454222322654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273063, "dur": 7,
    "args": {
      "External id": 60775,
      "cbid": 211, "correlation": 60775
    }
  },
  {
    "ph": "s", "id": 60775, "pid": 76337, "tid": -914061504, "ts": 1716454222273063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222322675, "dur": 12,
    "args": {
      "External id": 60784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60784,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60784, "pid": 5, "tid": 7, "ts": 1716454222322675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273102, "dur": 10,
    "args": {
      "External id": 60784,
      "cbid": 211, "correlation": 60784
    }
  },
  {
    "ph": "s", "id": 60784, "pid": 76337, "tid": -914061504, "ts": 1716454222273102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273157, "dur": 0,
    "args": {
      "External id": 60794,
      "cbid": 317, "correlation": 60794
    }
  },
  {
    "ph": "f", "id": 60794, "pid": 76337, "tid": -914061504, "ts": 1716454222273157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273158, "dur": 0,
    "args": {
      "External id": 60795,
      "cbid": 203, "correlation": 60795
    }
  },
  {
    "ph": "f", "id": 60795, "pid": 76337, "tid": -914061504, "ts": 1716454222273158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273159, "dur": 0,
    "args": {
      "External id": 60796,
      "cbid": 205, "correlation": 60796
    }
  },
  {
    "ph": "f", "id": 60796, "pid": 76337, "tid": -914061504, "ts": 1716454222273159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222322689, "dur": 12,
    "args": {
      "External id": 60800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60800, "pid": 5, "tid": 7, "ts": 1716454222322689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273178, "dur": 12,
    "args": {
      "External id": 60800,
      "cbid": 211, "correlation": 60800
    }
  },
  {
    "ph": "s", "id": 60800, "pid": 76337, "tid": -914061504, "ts": 1716454222273178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222322703, "dur": 26,
    "args": {
      "External id": 60802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60802, "pid": 5, "tid": 7, "ts": 1716454222322703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273192, "dur": 5,
    "args": {
      "External id": 60802,
      "cbid": 211, "correlation": 60802
    }
  },
  {
    "ph": "s", "id": 60802, "pid": 76337, "tid": -914061504, "ts": 1716454222273192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222322731, "dur": 4,
    "args": {
      "External id": 60804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60804,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 60804, "pid": 5, "tid": 7, "ts": 1716454222322731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273209, "dur": 8,
    "args": {
      "External id": 60804,
      "cbid": 211, "correlation": 60804
    }
  },
  {
    "ph": "s", "id": 60804, "pid": 76337, "tid": -914061504, "ts": 1716454222273209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273220, "dur": 0,
    "args": {
      "External id": 60805,
      "cbid": 51, "correlation": 60805
    }
  },
  {
    "ph": "s", "id": 60805, "pid": 76337, "tid": -914061504, "ts": 1716454222273220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222322736, "dur": 387,
    "args": {
      "External id": 60806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60806,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60806, "pid": 5, "tid": 7, "ts": 1716454222322736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273221, "dur": 8,
    "args": {
      "External id": 60806,
      "cbid": 211, "correlation": 60806
    }
  },
  {
    "ph": "s", "id": 60806, "pid": 76337, "tid": -914061504, "ts": 1716454222273221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222323124, "dur": 21,
    "args": {
      "External id": 60807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60807,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60807, "pid": 5, "tid": 7, "ts": 1716454222323124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273232, "dur": 5,
    "args": {
      "External id": 60807,
      "cbid": 211, "correlation": 60807
    }
  },
  {
    "ph": "s", "id": 60807, "pid": 76337, "tid": -914061504, "ts": 1716454222273232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222323147, "dur": 35,
    "args": {
      "External id": 60813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60813, "pid": 5, "tid": 7, "ts": 1716454222323147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273259, "dur": 9,
    "args": {
      "External id": 60813,
      "cbid": 211, "correlation": 60813
    }
  },
  {
    "ph": "s", "id": 60813, "pid": 76337, "tid": -914061504, "ts": 1716454222273259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222323184, "dur": 4,
    "args": {
      "External id": 60821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60821,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 60821, "pid": 5, "tid": 7, "ts": 1716454222323184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273303, "dur": 9,
    "args": {
      "External id": 60821,
      "cbid": 211, "correlation": 60821
    }
  },
  {
    "ph": "s", "id": 60821, "pid": 76337, "tid": -914061504, "ts": 1716454222273303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273369, "dur": 1,
    "args": {
      "External id": 60837,
      "cbid": 251, "correlation": 60837
    }
  },
  {
    "ph": "f", "id": 60837, "pid": 76337, "tid": -914061504, "ts": 1716454222273369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273374, "dur": 0,
    "args": {
      "External id": 60839,
      "cbid": 251, "correlation": 60839
    }
  },
  {
    "ph": "f", "id": 60839, "pid": 76337, "tid": -914061504, "ts": 1716454222273374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222323189, "dur": 13,
    "args": {
      "External id": 60840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60840,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 60840, "pid": 5, "tid": 7, "ts": 1716454222323189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273376, "dur": 12,
    "args": {
      "External id": 60840,
      "cbid": 211, "correlation": 60840
    }
  },
  {
    "ph": "s", "id": 60840, "pid": 76337, "tid": -914061504, "ts": 1716454222273376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222323203, "dur": 5,
    "args": {
      "External id": 60842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60842,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 60842, "pid": 5, "tid": 7, "ts": 1716454222323203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273390, "dur": 6,
    "args": {
      "External id": 60842,
      "cbid": 211, "correlation": 60842
    }
  },
  {
    "ph": "s", "id": 60842, "pid": 76337, "tid": -914061504, "ts": 1716454222273390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222323210, "dur": 31,
    "args": {
      "External id": 60852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60852, "pid": 5, "tid": 7, "ts": 1716454222323210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273448, "dur": 13,
    "args": {
      "External id": 60852,
      "cbid": 211, "correlation": 60852
    }
  },
  {
    "ph": "s", "id": 60852, "pid": 76337, "tid": -914061504, "ts": 1716454222273448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222323242, "dur": 32,
    "args": {
      "External id": 60872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60872,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 60872, "pid": 5, "tid": 7, "ts": 1716454222323242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273516, "dur": 15,
    "args": {
      "External id": 60872,
      "cbid": 211, "correlation": 60872
    }
  },
  {
    "ph": "s", "id": 60872, "pid": 76337, "tid": -914061504, "ts": 1716454222273516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222323276, "dur": 4,
    "args": {
      "External id": 60884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60884,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 60884, "pid": 5, "tid": 7, "ts": 1716454222323276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273541, "dur": 7,
    "args": {
      "External id": 60884,
      "cbid": 211, "correlation": 60884
    }
  },
  {
    "ph": "s", "id": 60884, "pid": 76337, "tid": -914061504, "ts": 1716454222273541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222323282, "dur": 32,
    "args": {
      "External id": 60887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60887, "pid": 5, "tid": 7, "ts": 1716454222323282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273560, "dur": 6,
    "args": {
      "External id": 60887,
      "cbid": 211, "correlation": 60887
    }
  },
  {
    "ph": "s", "id": 60887, "pid": 76337, "tid": -914061504, "ts": 1716454222273560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222323315, "dur": 22,
    "args": {
      "External id": 60896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60896,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60896, "pid": 5, "tid": 7, "ts": 1716454222323315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273601, "dur": 9,
    "args": {
      "External id": 60896,
      "cbid": 211, "correlation": 60896
    }
  },
  {
    "ph": "s", "id": 60896, "pid": 76337, "tid": -914061504, "ts": 1716454222273601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273663, "dur": 0,
    "args": {
      "External id": 60906,
      "cbid": 317, "correlation": 60906
    }
  },
  {
    "ph": "f", "id": 60906, "pid": 76337, "tid": -914061504, "ts": 1716454222273663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273664, "dur": 0,
    "args": {
      "External id": 60907,
      "cbid": 203, "correlation": 60907
    }
  },
  {
    "ph": "f", "id": 60907, "pid": 76337, "tid": -914061504, "ts": 1716454222273664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273665, "dur": 0,
    "args": {
      "External id": 60908,
      "cbid": 205, "correlation": 60908
    }
  },
  {
    "ph": "f", "id": 60908, "pid": 76337, "tid": -914061504, "ts": 1716454222273665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222323338, "dur": 22,
    "args": {
      "External id": 60912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60912, "pid": 5, "tid": 7, "ts": 1716454222323338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273687, "dur": 12,
    "args": {
      "External id": 60912,
      "cbid": 211, "correlation": 60912
    }
  },
  {
    "ph": "s", "id": 60912, "pid": 76337, "tid": -914061504, "ts": 1716454222273687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222323362, "dur": 49,
    "args": {
      "External id": 60914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60914, "pid": 5, "tid": 7, "ts": 1716454222323362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273702, "dur": 5,
    "args": {
      "External id": 60914,
      "cbid": 211, "correlation": 60914
    }
  },
  {
    "ph": "s", "id": 60914, "pid": 76337, "tid": -914061504, "ts": 1716454222273702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222323412, "dur": 718,
    "args": {
      "External id": 60916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60916,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 60916, "pid": 5, "tid": 7, "ts": 1716454222323412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273721, "dur": 13,
    "args": {
      "External id": 60916,
      "cbid": 211, "correlation": 60916
    }
  },
  {
    "ph": "s", "id": 60916, "pid": 76337, "tid": -914061504, "ts": 1716454222273721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222324131, "dur": 22,
    "args": {
      "External id": 60918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60918,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60918, "pid": 5, "tid": 7, "ts": 1716454222324131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273738, "dur": 5,
    "args": {
      "External id": 60918,
      "cbid": 211, "correlation": 60918
    }
  },
  {
    "ph": "s", "id": 60918, "pid": 76337, "tid": -914061504, "ts": 1716454222273738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222324155, "dur": 34,
    "args": {
      "External id": 60924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60924, "pid": 5, "tid": 7, "ts": 1716454222324155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273766, "dur": 8,
    "args": {
      "External id": 60924,
      "cbid": 211, "correlation": 60924
    }
  },
  {
    "ph": "s", "id": 60924, "pid": 76337, "tid": -914061504, "ts": 1716454222273766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273824, "dur": 0,
    "args": {
      "External id": 60934,
      "cbid": 317, "correlation": 60934
    }
  },
  {
    "ph": "f", "id": 60934, "pid": 76337, "tid": -914061504, "ts": 1716454222273824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273825, "dur": 0,
    "args": {
      "External id": 60935,
      "cbid": 203, "correlation": 60935
    }
  },
  {
    "ph": "f", "id": 60935, "pid": 76337, "tid": -914061504, "ts": 1716454222273825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273825, "dur": 0,
    "args": {
      "External id": 60936,
      "cbid": 205, "correlation": 60936
    }
  },
  {
    "ph": "f", "id": 60936, "pid": 76337, "tid": -914061504, "ts": 1716454222273825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273852, "dur": 1,
    "args": {
      "External id": 60940,
      "cbid": 251, "correlation": 60940
    }
  },
  {
    "ph": "f", "id": 60940, "pid": 76337, "tid": -914061504, "ts": 1716454222273852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273854, "dur": 0,
    "args": {
      "External id": 60941,
      "cbid": 251, "correlation": 60941
    }
  },
  {
    "ph": "f", "id": 60941, "pid": 76337, "tid": -914061504, "ts": 1716454222273854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273854, "dur": 0,
    "args": {
      "External id": 60942,
      "cbid": 251, "correlation": 60942
    }
  },
  {
    "ph": "f", "id": 60942, "pid": 76337, "tid": -914061504, "ts": 1716454222273854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273855, "dur": 0,
    "args": {
      "External id": 60943,
      "cbid": 251, "correlation": 60943
    }
  },
  {
    "ph": "f", "id": 60943, "pid": 76337, "tid": -914061504, "ts": 1716454222273855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273856, "dur": 0,
    "args": {
      "External id": 60944,
      "cbid": 251, "correlation": 60944
    }
  },
  {
    "ph": "f", "id": 60944, "pid": 76337, "tid": -914061504, "ts": 1716454222273856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273856, "dur": 0,
    "args": {
      "External id": 60945,
      "cbid": 251, "correlation": 60945
    }
  },
  {
    "ph": "f", "id": 60945, "pid": 76337, "tid": -914061504, "ts": 1716454222273856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273857, "dur": 0,
    "args": {
      "External id": 60946,
      "cbid": 251, "correlation": 60946
    }
  },
  {
    "ph": "f", "id": 60946, "pid": 76337, "tid": -914061504, "ts": 1716454222273857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273858, "dur": 0,
    "args": {
      "External id": 60947,
      "cbid": 251, "correlation": 60947
    }
  },
  {
    "ph": "f", "id": 60947, "pid": 76337, "tid": -914061504, "ts": 1716454222273858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273859, "dur": 0,
    "args": {
      "External id": 60948,
      "cbid": 251, "correlation": 60948
    }
  },
  {
    "ph": "f", "id": 60948, "pid": 76337, "tid": -914061504, "ts": 1716454222273859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222324190, "dur": 55,
    "args": {
      "External id": 60949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60949,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 60949, "pid": 5, "tid": 7, "ts": 1716454222324190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273861, "dur": 12,
    "args": {
      "External id": 60949,
      "cbid": 211, "correlation": 60949
    }
  },
  {
    "ph": "s", "id": 60949, "pid": 76337, "tid": -914061504, "ts": 1716454222273861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222324247, "dur": 35,
    "args": {
      "External id": 60955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60955, "pid": 5, "tid": 7, "ts": 1716454222324247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273895, "dur": 8,
    "args": {
      "External id": 60955,
      "cbid": 211, "correlation": 60955
    }
  },
  {
    "ph": "s", "id": 60955, "pid": 76337, "tid": -914061504, "ts": 1716454222273895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222324283, "dur": 27,
    "args": {
      "External id": 60963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60963,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60963, "pid": 5, "tid": 7, "ts": 1716454222324283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273924, "dur": 8,
    "args": {
      "External id": 60963,
      "cbid": 211, "correlation": 60963
    }
  },
  {
    "ph": "s", "id": 60963, "pid": 76337, "tid": -914061504, "ts": 1716454222273924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222324312, "dur": 20,
    "args": {
      "External id": 60971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60971,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 60971, "pid": 5, "tid": 7, "ts": 1716454222324312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222273954, "dur": 9,
    "args": {
      "External id": 60971,
      "cbid": 211, "correlation": 60971
    }
  },
  {
    "ph": "s", "id": 60971, "pid": 76337, "tid": -914061504, "ts": 1716454222273954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222324333, "dur": 32,
    "args": {
      "External id": 60991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 60991,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 60991, "pid": 5, "tid": 7, "ts": 1716454222324333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274047, "dur": 13,
    "args": {
      "External id": 60991,
      "cbid": 211, "correlation": 60991
    }
  },
  {
    "ph": "s", "id": 60991, "pid": 76337, "tid": -914061504, "ts": 1716454222274047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222324366, "dur": 4,
    "args": {
      "External id": 61003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61003,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 61003, "pid": 5, "tid": 7, "ts": 1716454222324366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274069, "dur": 6,
    "args": {
      "External id": 61003,
      "cbid": 211, "correlation": 61003
    }
  },
  {
    "ph": "s", "id": 61003, "pid": 76337, "tid": -914061504, "ts": 1716454222274069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222324372, "dur": 32,
    "args": {
      "External id": 61006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61006, "pid": 5, "tid": 7, "ts": 1716454222324372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274087, "dur": 7,
    "args": {
      "External id": 61006,
      "cbid": 211, "correlation": 61006
    }
  },
  {
    "ph": "s", "id": 61006, "pid": 76337, "tid": -914061504, "ts": 1716454222274087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274149, "dur": 0,
    "args": {
      "External id": 61017,
      "cbid": 317, "correlation": 61017
    }
  },
  {
    "ph": "f", "id": 61017, "pid": 76337, "tid": -914061504, "ts": 1716454222274149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274150, "dur": 0,
    "args": {
      "External id": 61018,
      "cbid": 203, "correlation": 61018
    }
  },
  {
    "ph": "f", "id": 61018, "pid": 76337, "tid": -914061504, "ts": 1716454222274150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274151, "dur": 0,
    "args": {
      "External id": 61019,
      "cbid": 205, "correlation": 61019
    }
  },
  {
    "ph": "f", "id": 61019, "pid": 76337, "tid": -914061504, "ts": 1716454222274151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222324405, "dur": 23,
    "args": {
      "External id": 61023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61023, "pid": 5, "tid": 7, "ts": 1716454222324405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274167, "dur": 12,
    "args": {
      "External id": 61023,
      "cbid": 211, "correlation": 61023
    }
  },
  {
    "ph": "s", "id": 61023, "pid": 76337, "tid": -914061504, "ts": 1716454222274167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222324429, "dur": 130,
    "args": {
      "External id": 61025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61025,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61025, "pid": 5, "tid": 7, "ts": 1716454222324429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274188, "dur": 10,
    "args": {
      "External id": 61025,
      "cbid": 211, "correlation": 61025
    }
  },
  {
    "ph": "s", "id": 61025, "pid": 76337, "tid": -914061504, "ts": 1716454222274188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222324561, "dur": 22,
    "args": {
      "External id": 61027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61027,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61027, "pid": 5, "tid": 7, "ts": 1716454222324561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274201, "dur": 5,
    "args": {
      "External id": 61027,
      "cbid": 211, "correlation": 61027
    }
  },
  {
    "ph": "s", "id": 61027, "pid": 76337, "tid": -914061504, "ts": 1716454222274201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222324584, "dur": 35,
    "args": {
      "External id": 61033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61033, "pid": 5, "tid": 7, "ts": 1716454222324584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274229, "dur": 9,
    "args": {
      "External id": 61033,
      "cbid": 211, "correlation": 61033
    }
  },
  {
    "ph": "s", "id": 61033, "pid": 76337, "tid": -914061504, "ts": 1716454222274229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222324620, "dur": 207,
    "args": {
      "External id": 61042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61042,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61042, "pid": 5, "tid": 7, "ts": 1716454222324620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274313, "dur": 14,
    "args": {
      "External id": 61042,
      "cbid": 211, "correlation": 61042
    }
  },
  {
    "ph": "s", "id": 61042, "pid": 76337, "tid": -914061504, "ts": 1716454222274313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222324828, "dur": 71,
    "args": {
      "External id": 61064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61064,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61064, "pid": 5, "tid": 7, "ts": 1716454222324828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274373, "dur": 10,
    "args": {
      "External id": 61064,
      "cbid": 211, "correlation": 61064
    }
  },
  {
    "ph": "s", "id": 61064, "pid": 76337, "tid": -914061504, "ts": 1716454222274373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274465, "dur": 1,
    "args": {
      "External id": 61075,
      "cbid": 251, "correlation": 61075
    }
  },
  {
    "ph": "f", "id": 61075, "pid": 76337, "tid": -914061504, "ts": 1716454222274465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222324901, "dur": 164,
    "args": {
      "External id": 61076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61076,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61076, "pid": 5, "tid": 7, "ts": 1716454222324901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274470, "dur": 13,
    "args": {
      "External id": 61076,
      "cbid": 211, "correlation": 61076
    }
  },
  {
    "ph": "s", "id": 61076, "pid": 76337, "tid": -914061504, "ts": 1716454222274470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274545, "dur": 1,
    "args": {
      "External id": 61087,
      "cbid": 251, "correlation": 61087
    }
  },
  {
    "ph": "f", "id": 61087, "pid": 76337, "tid": -914061504, "ts": 1716454222274545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222325066, "dur": 155,
    "args": {
      "External id": 61088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61088,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61088, "pid": 5, "tid": 7, "ts": 1716454222325066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274549, "dur": 12,
    "args": {
      "External id": 61088,
      "cbid": 211, "correlation": 61088
    }
  },
  {
    "ph": "s", "id": 61088, "pid": 76337, "tid": -914061504, "ts": 1716454222274549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274613, "dur": 1,
    "args": {
      "External id": 61099,
      "cbid": 251, "correlation": 61099
    }
  },
  {
    "ph": "f", "id": 61099, "pid": 76337, "tid": -914061504, "ts": 1716454222274613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222325223, "dur": 155,
    "args": {
      "External id": 61100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61100,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61100, "pid": 5, "tid": 7, "ts": 1716454222325223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274617, "dur": 12,
    "args": {
      "External id": 61100,
      "cbid": 211, "correlation": 61100
    }
  },
  {
    "ph": "s", "id": 61100, "pid": 76337, "tid": -914061504, "ts": 1716454222274617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222325379, "dur": 2148,
    "args": {
      "External id": 61121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61121,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 61121, "pid": 5, "tid": 7, "ts": 1716454222325379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274700, "dur": 13,
    "args": {
      "External id": 61121,
      "cbid": 211, "correlation": 61121
    }
  },
  {
    "ph": "s", "id": 61121, "pid": 76337, "tid": -914061504, "ts": 1716454222274700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274800, "dur": 1,
    "args": {
      "External id": 61139,
      "cbid": 251, "correlation": 61139
    }
  },
  {
    "ph": "f", "id": 61139, "pid": 76337, "tid": -914061504, "ts": 1716454222274800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222327529, "dur": 158,
    "args": {
      "External id": 61141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61141,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 61141, "pid": 5, "tid": 7, "ts": 1716454222327529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274806, "dur": 13,
    "args": {
      "External id": 61141,
      "cbid": 211, "correlation": 61141
    }
  },
  {
    "ph": "s", "id": 61141, "pid": 76337, "tid": -914061504, "ts": 1716454222274806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222327688, "dur": 36,
    "args": {
      "External id": 61149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61149,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61149, "pid": 5, "tid": 7, "ts": 1716454222327688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274875, "dur": 12,
    "args": {
      "External id": 61149,
      "cbid": 211, "correlation": 61149
    }
  },
  {
    "ph": "s", "id": 61149, "pid": 76337, "tid": -914061504, "ts": 1716454222274875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222327725, "dur": 51,
    "args": {
      "External id": 61157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61157,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61157, "pid": 5, "tid": 7, "ts": 1716454222327725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274915, "dur": 11,
    "args": {
      "External id": 61157,
      "cbid": 211, "correlation": 61157
    }
  },
  {
    "ph": "s", "id": 61157, "pid": 76337, "tid": -914061504, "ts": 1716454222274915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222327778, "dur": 33,
    "args": {
      "External id": 61168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61168, "pid": 5, "tid": 7, "ts": 1716454222327778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222274999, "dur": 14,
    "args": {
      "External id": 61168,
      "cbid": 211, "correlation": 61168
    }
  },
  {
    "ph": "s", "id": 61168, "pid": 76337, "tid": -914061504, "ts": 1716454222274999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222327812, "dur": 37,
    "args": {
      "External id": 61190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61190,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61190, "pid": 5, "tid": 7, "ts": 1716454222327812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275032, "dur": 7,
    "args": {
      "External id": 61190,
      "cbid": 211, "correlation": 61190
    }
  },
  {
    "ph": "s", "id": 61190, "pid": 76337, "tid": -914061504, "ts": 1716454222275032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275117, "dur": 1,
    "args": {
      "External id": 61201,
      "cbid": 251, "correlation": 61201
    }
  },
  {
    "ph": "f", "id": 61201, "pid": 76337, "tid": -914061504, "ts": 1716454222275117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222327851, "dur": 98,
    "args": {
      "External id": 61202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61202,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61202, "pid": 5, "tid": 7, "ts": 1716454222327851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275122, "dur": 14,
    "args": {
      "External id": 61202,
      "cbid": 211, "correlation": 61202
    }
  },
  {
    "ph": "s", "id": 61202, "pid": 76337, "tid": -914061504, "ts": 1716454222275122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275193, "dur": 1,
    "args": {
      "External id": 61213,
      "cbid": 251, "correlation": 61213
    }
  },
  {
    "ph": "f", "id": 61213, "pid": 76337, "tid": -914061504, "ts": 1716454222275193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275197, "dur": 0,
    "args": {
      "External id": 61214,
      "cbid": 251, "correlation": 61214
    }
  },
  {
    "ph": "f", "id": 61214, "pid": 76337, "tid": -914061504, "ts": 1716454222275197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222327950, "dur": 12,
    "args": {
      "External id": 61215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61215,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 61215, "pid": 5, "tid": 7, "ts": 1716454222327950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275198, "dur": 12,
    "args": {
      "External id": 61215,
      "cbid": 211, "correlation": 61215
    }
  },
  {
    "ph": "s", "id": 61215, "pid": 76337, "tid": -914061504, "ts": 1716454222275198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222327964, "dur": 5,
    "args": {
      "External id": 61217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61217,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 61217, "pid": 5, "tid": 7, "ts": 1716454222327964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275214, "dur": 11,
    "args": {
      "External id": 61217,
      "cbid": 211, "correlation": 61217
    }
  },
  {
    "ph": "s", "id": 61217, "pid": 76337, "tid": -914061504, "ts": 1716454222275214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275279, "dur": 1,
    "args": {
      "External id": 61228,
      "cbid": 251, "correlation": 61228
    }
  },
  {
    "ph": "f", "id": 61228, "pid": 76337, "tid": -914061504, "ts": 1716454222275279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275282, "dur": 0,
    "args": {
      "External id": 61229,
      "cbid": 251, "correlation": 61229
    }
  },
  {
    "ph": "f", "id": 61229, "pid": 76337, "tid": -914061504, "ts": 1716454222275282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222327970, "dur": 8,
    "args": {
      "External id": 61230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61230,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 61230, "pid": 5, "tid": 7, "ts": 1716454222327970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275284, "dur": 12,
    "args": {
      "External id": 61230,
      "cbid": 211, "correlation": 61230
    }
  },
  {
    "ph": "s", "id": 61230, "pid": 76337, "tid": -914061504, "ts": 1716454222275284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222327979, "dur": 3,
    "args": {
      "External id": 61232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61232,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 61232, "pid": 5, "tid": 7, "ts": 1716454222327979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275297, "dur": 8,
    "args": {
      "External id": 61232,
      "cbid": 211, "correlation": 61232
    }
  },
  {
    "ph": "s", "id": 61232, "pid": 76337, "tid": -914061504, "ts": 1716454222275297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222327984, "dur": 100,
    "args": {
      "External id": 61253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61253,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 61253, "pid": 5, "tid": 7, "ts": 1716454222327984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275374, "dur": 12,
    "args": {
      "External id": 61253,
      "cbid": 211, "correlation": 61253
    }
  },
  {
    "ph": "s", "id": 61253, "pid": 76337, "tid": -914061504, "ts": 1716454222275374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275469, "dur": 1,
    "args": {
      "External id": 61271,
      "cbid": 251, "correlation": 61271
    }
  },
  {
    "ph": "f", "id": 61271, "pid": 76337, "tid": -914061504, "ts": 1716454222275469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222328085, "dur": 109,
    "args": {
      "External id": 61273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61273,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61273, "pid": 5, "tid": 7, "ts": 1716454222328085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275474, "dur": 13,
    "args": {
      "External id": 61273,
      "cbid": 211, "correlation": 61273
    }
  },
  {
    "ph": "s", "id": 61273, "pid": 76337, "tid": -914061504, "ts": 1716454222275474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222328195, "dur": 19,
    "args": {
      "External id": 61281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61281,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61281, "pid": 5, "tid": 7, "ts": 1716454222328195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275543, "dur": 12,
    "args": {
      "External id": 61281,
      "cbid": 211, "correlation": 61281
    }
  },
  {
    "ph": "s", "id": 61281, "pid": 76337, "tid": -914061504, "ts": 1716454222275543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222328216, "dur": 39,
    "args": {
      "External id": 61289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61289, "pid": 5, "tid": 7, "ts": 1716454222328216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275585, "dur": 9,
    "args": {
      "External id": 61289,
      "cbid": 211, "correlation": 61289
    }
  },
  {
    "ph": "s", "id": 61289, "pid": 76337, "tid": -914061504, "ts": 1716454222275585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222328256, "dur": 38,
    "args": {
      "External id": 61311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61311,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61311, "pid": 5, "tid": 7, "ts": 1716454222328256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275636, "dur": 13,
    "args": {
      "External id": 61311,
      "cbid": 211, "correlation": 61311
    }
  },
  {
    "ph": "s", "id": 61311, "pid": 76337, "tid": -914061504, "ts": 1716454222275636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275729, "dur": 1,
    "args": {
      "External id": 61327,
      "cbid": 251, "correlation": 61327
    }
  },
  {
    "ph": "f", "id": 61327, "pid": 76337, "tid": -914061504, "ts": 1716454222275729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275734, "dur": 0,
    "args": {
      "External id": 61329,
      "cbid": 251, "correlation": 61329
    }
  },
  {
    "ph": "f", "id": 61329, "pid": 76337, "tid": -914061504, "ts": 1716454222275734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222328296, "dur": 589,
    "args": {
      "External id": 61330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61330,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 61330, "pid": 5, "tid": 7, "ts": 1716454222328296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275738, "dur": 13,
    "args": {
      "External id": 61330,
      "cbid": 211, "correlation": 61330
    }
  },
  {
    "ph": "s", "id": 61330, "pid": 76337, "tid": -914061504, "ts": 1716454222275738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222328886, "dur": 134,
    "args": {
      "External id": 61338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61338, "pid": 5, "tid": 7, "ts": 1716454222328886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275804, "dur": 12,
    "args": {
      "External id": 61338,
      "cbid": 211, "correlation": 61338
    }
  },
  {
    "ph": "s", "id": 61338, "pid": 76337, "tid": -914061504, "ts": 1716454222275804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222329022, "dur": 131,
    "args": {
      "External id": 61346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61346, "pid": 5, "tid": 7, "ts": 1716454222329022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275835, "dur": 8,
    "args": {
      "External id": 61346,
      "cbid": 211, "correlation": 61346
    }
  },
  {
    "ph": "s", "id": 61346, "pid": 76337, "tid": -914061504, "ts": 1716454222275835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275913, "dur": 1,
    "args": {
      "External id": 61362,
      "cbid": 251, "correlation": 61362
    }
  },
  {
    "ph": "f", "id": 61362, "pid": 76337, "tid": -914061504, "ts": 1716454222275913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222329154, "dur": 331,
    "args": {
      "External id": 61364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61364,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61364, "pid": 5, "tid": 7, "ts": 1716454222329154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275918, "dur": 12,
    "args": {
      "External id": 61364,
      "cbid": 211, "correlation": 61364
    }
  },
  {
    "ph": "s", "id": 61364, "pid": 76337, "tid": -914061504, "ts": 1716454222275918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222329486, "dur": 28,
    "args": {
      "External id": 61372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61372,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61372, "pid": 5, "tid": 7, "ts": 1716454222329486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222275961, "dur": 10,
    "args": {
      "External id": 61372,
      "cbid": 211, "correlation": 61372
    }
  },
  {
    "ph": "s", "id": 61372, "pid": 76337, "tid": -914061504, "ts": 1716454222275961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222329516, "dur": 89,
    "args": {
      "External id": 61383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61383, "pid": 5, "tid": 7, "ts": 1716454222329516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276041, "dur": 14,
    "args": {
      "External id": 61383,
      "cbid": 211, "correlation": 61383
    }
  },
  {
    "ph": "s", "id": 61383, "pid": 76337, "tid": -914061504, "ts": 1716454222276041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276109, "dur": 0,
    "args": {
      "External id": 61395,
      "cbid": 317, "correlation": 61395
    }
  },
  {
    "ph": "f", "id": 61395, "pid": 76337, "tid": -914061504, "ts": 1716454222276109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276110, "dur": 0,
    "args": {
      "External id": 61396,
      "cbid": 203, "correlation": 61396
    }
  },
  {
    "ph": "f", "id": 61396, "pid": 76337, "tid": -914061504, "ts": 1716454222276110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276111, "dur": 0,
    "args": {
      "External id": 61397,
      "cbid": 205, "correlation": 61397
    }
  },
  {
    "ph": "f", "id": 61397, "pid": 76337, "tid": -914061504, "ts": 1716454222276111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222329606, "dur": 23,
    "args": {
      "External id": 61401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61401,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61401, "pid": 5, "tid": 7, "ts": 1716454222329606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276126, "dur": 12,
    "args": {
      "External id": 61401,
      "cbid": 211, "correlation": 61401
    }
  },
  {
    "ph": "s", "id": 61401, "pid": 76337, "tid": -914061504, "ts": 1716454222276126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222329631, "dur": 131,
    "args": {
      "External id": 61403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61403,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61403, "pid": 5, "tid": 7, "ts": 1716454222329631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276145, "dur": 7,
    "args": {
      "External id": 61403,
      "cbid": 211, "correlation": 61403
    }
  },
  {
    "ph": "s", "id": 61403, "pid": 76337, "tid": -914061504, "ts": 1716454222276145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222329764, "dur": 22,
    "args": {
      "External id": 61405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61405,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61405, "pid": 5, "tid": 7, "ts": 1716454222329764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276156, "dur": 5,
    "args": {
      "External id": 61405,
      "cbid": 211, "correlation": 61405
    }
  },
  {
    "ph": "s", "id": 61405, "pid": 76337, "tid": -914061504, "ts": 1716454222276156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222329788, "dur": 35,
    "args": {
      "External id": 61411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61411, "pid": 5, "tid": 7, "ts": 1716454222329788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276184, "dur": 8,
    "args": {
      "External id": 61411,
      "cbid": 211, "correlation": 61411
    }
  },
  {
    "ph": "s", "id": 61411, "pid": 76337, "tid": -914061504, "ts": 1716454222276184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222329824, "dur": 27,
    "args": {
      "External id": 61419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61419,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61419, "pid": 5, "tid": 7, "ts": 1716454222329824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276216, "dur": 8,
    "args": {
      "External id": 61419,
      "cbid": 211, "correlation": 61419
    }
  },
  {
    "ph": "s", "id": 61419, "pid": 76337, "tid": -914061504, "ts": 1716454222276216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222329853, "dur": 32,
    "args": {
      "External id": 61439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61439,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 61439, "pid": 5, "tid": 7, "ts": 1716454222329853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276289, "dur": 12,
    "args": {
      "External id": 61439,
      "cbid": 211, "correlation": 61439
    }
  },
  {
    "ph": "s", "id": 61439, "pid": 76337, "tid": -914061504, "ts": 1716454222276289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222329887, "dur": 5,
    "args": {
      "External id": 61451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61451,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 61451, "pid": 5, "tid": 7, "ts": 1716454222329887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276312, "dur": 6,
    "args": {
      "External id": 61451,
      "cbid": 211, "correlation": 61451
    }
  },
  {
    "ph": "s", "id": 61451, "pid": 76337, "tid": -914061504, "ts": 1716454222276312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222329893, "dur": 33,
    "args": {
      "External id": 61454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61454, "pid": 5, "tid": 7, "ts": 1716454222329893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276329, "dur": 10,
    "args": {
      "External id": 61454,
      "cbid": 211, "correlation": 61454
    }
  },
  {
    "ph": "s", "id": 61454, "pid": 76337, "tid": -914061504, "ts": 1716454222276329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222329927, "dur": 22,
    "args": {
      "External id": 61463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61463,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61463, "pid": 5, "tid": 7, "ts": 1716454222329927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276373, "dur": 10,
    "args": {
      "External id": 61463,
      "cbid": 211, "correlation": 61463
    }
  },
  {
    "ph": "s", "id": 61463, "pid": 76337, "tid": -914061504, "ts": 1716454222276373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276425, "dur": 0,
    "args": {
      "External id": 61473,
      "cbid": 317, "correlation": 61473
    }
  },
  {
    "ph": "f", "id": 61473, "pid": 76337, "tid": -914061504, "ts": 1716454222276425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276426, "dur": 0,
    "args": {
      "External id": 61474,
      "cbid": 203, "correlation": 61474
    }
  },
  {
    "ph": "f", "id": 61474, "pid": 76337, "tid": -914061504, "ts": 1716454222276426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276426, "dur": 0,
    "args": {
      "External id": 61475,
      "cbid": 205, "correlation": 61475
    }
  },
  {
    "ph": "f", "id": 61475, "pid": 76337, "tid": -914061504, "ts": 1716454222276426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222329951, "dur": 24,
    "args": {
      "External id": 61479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61479,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61479, "pid": 5, "tid": 7, "ts": 1716454222329951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276440, "dur": 12,
    "args": {
      "External id": 61479,
      "cbid": 211, "correlation": 61479
    }
  },
  {
    "ph": "s", "id": 61479, "pid": 76337, "tid": -914061504, "ts": 1716454222276440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222329976, "dur": 48,
    "args": {
      "External id": 61481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61481,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61481, "pid": 5, "tid": 7, "ts": 1716454222329976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276454, "dur": 5,
    "args": {
      "External id": 61481,
      "cbid": 211, "correlation": 61481
    }
  },
  {
    "ph": "s", "id": 61481, "pid": 76337, "tid": -914061504, "ts": 1716454222276454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222330026, "dur": 717,
    "args": {
      "External id": 61483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61483,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61483, "pid": 5, "tid": 7, "ts": 1716454222330026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276466, "dur": 6,
    "args": {
      "External id": 61483,
      "cbid": 211, "correlation": 61483
    }
  },
  {
    "ph": "s", "id": 61483, "pid": 76337, "tid": -914061504, "ts": 1716454222276466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222330744, "dur": 23,
    "args": {
      "External id": 61485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61485,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61485, "pid": 5, "tid": 7, "ts": 1716454222330744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276475, "dur": 5,
    "args": {
      "External id": 61485,
      "cbid": 211, "correlation": 61485
    }
  },
  {
    "ph": "s", "id": 61485, "pid": 76337, "tid": -914061504, "ts": 1716454222276475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222330768, "dur": 35,
    "args": {
      "External id": 61491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61491, "pid": 5, "tid": 7, "ts": 1716454222330768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276502, "dur": 9,
    "args": {
      "External id": 61491,
      "cbid": 211, "correlation": 61491
    }
  },
  {
    "ph": "s", "id": 61491, "pid": 76337, "tid": -914061504, "ts": 1716454222276502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222330805, "dur": 4,
    "args": {
      "External id": 61499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61499,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 61499, "pid": 5, "tid": 7, "ts": 1716454222330805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276546, "dur": 10,
    "args": {
      "External id": 61499,
      "cbid": 211, "correlation": 61499
    }
  },
  {
    "ph": "s", "id": 61499, "pid": 76337, "tid": -914061504, "ts": 1716454222276546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276612, "dur": 1,
    "args": {
      "External id": 61515,
      "cbid": 251, "correlation": 61515
    }
  },
  {
    "ph": "f", "id": 61515, "pid": 76337, "tid": -914061504, "ts": 1716454222276612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276617, "dur": 0,
    "args": {
      "External id": 61517,
      "cbid": 251, "correlation": 61517
    }
  },
  {
    "ph": "f", "id": 61517, "pid": 76337, "tid": -914061504, "ts": 1716454222276617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222330810, "dur": 13,
    "args": {
      "External id": 61518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61518,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 61518, "pid": 5, "tid": 7, "ts": 1716454222330810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276619, "dur": 12,
    "args": {
      "External id": 61518,
      "cbid": 211, "correlation": 61518
    }
  },
  {
    "ph": "s", "id": 61518, "pid": 76337, "tid": -914061504, "ts": 1716454222276619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222330824, "dur": 6,
    "args": {
      "External id": 61520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61520,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 61520, "pid": 5, "tid": 7, "ts": 1716454222330824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276636, "dur": 5,
    "args": {
      "External id": 61520,
      "cbid": 211, "correlation": 61520
    }
  },
  {
    "ph": "s", "id": 61520, "pid": 76337, "tid": -914061504, "ts": 1716454222276636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222330831, "dur": 31,
    "args": {
      "External id": 61530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61530, "pid": 5, "tid": 7, "ts": 1716454222330831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276693, "dur": 12,
    "args": {
      "External id": 61530,
      "cbid": 211, "correlation": 61530
    }
  },
  {
    "ph": "s", "id": 61530, "pid": 76337, "tid": -914061504, "ts": 1716454222276693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222330863, "dur": 32,
    "args": {
      "External id": 61550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61550,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 61550, "pid": 5, "tid": 7, "ts": 1716454222330863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276759, "dur": 10,
    "args": {
      "External id": 61550,
      "cbid": 211, "correlation": 61550
    }
  },
  {
    "ph": "s", "id": 61550, "pid": 76337, "tid": -914061504, "ts": 1716454222276759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222330897, "dur": 4,
    "args": {
      "External id": 61562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61562,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 61562, "pid": 5, "tid": 7, "ts": 1716454222330897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276779, "dur": 7,
    "args": {
      "External id": 61562,
      "cbid": 211, "correlation": 61562
    }
  },
  {
    "ph": "s", "id": 61562, "pid": 76337, "tid": -914061504, "ts": 1716454222276779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222330902, "dur": 31,
    "args": {
      "External id": 61565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61565, "pid": 5, "tid": 7, "ts": 1716454222330902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276798, "dur": 7,
    "args": {
      "External id": 61565,
      "cbid": 211, "correlation": 61565
    }
  },
  {
    "ph": "s", "id": 61565, "pid": 76337, "tid": -914061504, "ts": 1716454222276798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222330935, "dur": 22,
    "args": {
      "External id": 61574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61574,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61574, "pid": 5, "tid": 7, "ts": 1716454222330935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276839, "dur": 10,
    "args": {
      "External id": 61574,
      "cbid": 211, "correlation": 61574
    }
  },
  {
    "ph": "s", "id": 61574, "pid": 76337, "tid": -914061504, "ts": 1716454222276839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276901, "dur": 0,
    "args": {
      "External id": 61584,
      "cbid": 317, "correlation": 61584
    }
  },
  {
    "ph": "f", "id": 61584, "pid": 76337, "tid": -914061504, "ts": 1716454222276901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276902, "dur": 0,
    "args": {
      "External id": 61585,
      "cbid": 203, "correlation": 61585
    }
  },
  {
    "ph": "f", "id": 61585, "pid": 76337, "tid": -914061504, "ts": 1716454222276902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276903, "dur": 0,
    "args": {
      "External id": 61586,
      "cbid": 205, "correlation": 61586
    }
  },
  {
    "ph": "f", "id": 61586, "pid": 76337, "tid": -914061504, "ts": 1716454222276903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222330958, "dur": 23,
    "args": {
      "External id": 61590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61590,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61590, "pid": 5, "tid": 7, "ts": 1716454222330958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276916, "dur": 12,
    "args": {
      "External id": 61590,
      "cbid": 211, "correlation": 61590
    }
  },
  {
    "ph": "s", "id": 61590, "pid": 76337, "tid": -914061504, "ts": 1716454222276916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222330982, "dur": 48,
    "args": {
      "External id": 61592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61592, "pid": 5, "tid": 7, "ts": 1716454222330982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276934, "dur": 6,
    "args": {
      "External id": 61592,
      "cbid": 211, "correlation": 61592
    }
  },
  {
    "ph": "s", "id": 61592, "pid": 76337, "tid": -914061504, "ts": 1716454222276934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222331032, "dur": 709,
    "args": {
      "External id": 61594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61594,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61594, "pid": 5, "tid": 7, "ts": 1716454222331032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276946, "dur": 6,
    "args": {
      "External id": 61594,
      "cbid": 211, "correlation": 61594
    }
  },
  {
    "ph": "s", "id": 61594, "pid": 76337, "tid": -914061504, "ts": 1716454222276946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222331743, "dur": 22,
    "args": {
      "External id": 61596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61596, "pid": 5, "tid": 7, "ts": 1716454222331743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276955, "dur": 5,
    "args": {
      "External id": 61596,
      "cbid": 211, "correlation": 61596
    }
  },
  {
    "ph": "s", "id": 61596, "pid": 76337, "tid": -914061504, "ts": 1716454222276955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222331766, "dur": 36,
    "args": {
      "External id": 61602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61602, "pid": 5, "tid": 7, "ts": 1716454222331766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222276992, "dur": 10,
    "args": {
      "External id": 61602,
      "cbid": 211, "correlation": 61602
    }
  },
  {
    "ph": "s", "id": 61602, "pid": 76337, "tid": -914061504, "ts": 1716454222276992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222331803, "dur": 27,
    "args": {
      "External id": 61610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61610,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61610, "pid": 5, "tid": 7, "ts": 1716454222331803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277026, "dur": 8,
    "args": {
      "External id": 61610,
      "cbid": 211, "correlation": 61610
    }
  },
  {
    "ph": "s", "id": 61610, "pid": 76337, "tid": -914061504, "ts": 1716454222277026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222331831, "dur": 20,
    "args": {
      "External id": 61618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61618,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61618, "pid": 5, "tid": 7, "ts": 1716454222331831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277055, "dur": 8,
    "args": {
      "External id": 61618,
      "cbid": 211, "correlation": 61618
    }
  },
  {
    "ph": "s", "id": 61618, "pid": 76337, "tid": -914061504, "ts": 1716454222277055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222331853, "dur": 33,
    "args": {
      "External id": 61638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61638,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 61638, "pid": 5, "tid": 7, "ts": 1716454222331853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277134, "dur": 13,
    "args": {
      "External id": 61638,
      "cbid": 211, "correlation": 61638
    }
  },
  {
    "ph": "s", "id": 61638, "pid": 76337, "tid": -914061504, "ts": 1716454222277134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222331887, "dur": 4,
    "args": {
      "External id": 61650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61650,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 61650, "pid": 5, "tid": 7, "ts": 1716454222331887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277157, "dur": 6,
    "args": {
      "External id": 61650,
      "cbid": 211, "correlation": 61650
    }
  },
  {
    "ph": "s", "id": 61650, "pid": 76337, "tid": -914061504, "ts": 1716454222277157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222331893, "dur": 31,
    "args": {
      "External id": 61653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61653,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61653, "pid": 5, "tid": 7, "ts": 1716454222331893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277174, "dur": 6,
    "args": {
      "External id": 61653,
      "cbid": 211, "correlation": 61653
    }
  },
  {
    "ph": "s", "id": 61653, "pid": 76337, "tid": -914061504, "ts": 1716454222277174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277236, "dur": 0,
    "args": {
      "External id": 61664,
      "cbid": 317, "correlation": 61664
    }
  },
  {
    "ph": "f", "id": 61664, "pid": 76337, "tid": -914061504, "ts": 1716454222277236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277236, "dur": 0,
    "args": {
      "External id": 61665,
      "cbid": 203, "correlation": 61665
    }
  },
  {
    "ph": "f", "id": 61665, "pid": 76337, "tid": -914061504, "ts": 1716454222277236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277237, "dur": 0,
    "args": {
      "External id": 61666,
      "cbid": 205, "correlation": 61666
    }
  },
  {
    "ph": "f", "id": 61666, "pid": 76337, "tid": -914061504, "ts": 1716454222277237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222331925, "dur": 23,
    "args": {
      "External id": 61670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61670, "pid": 5, "tid": 7, "ts": 1716454222331925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277250, "dur": 12,
    "args": {
      "External id": 61670,
      "cbid": 211, "correlation": 61670
    }
  },
  {
    "ph": "s", "id": 61670, "pid": 76337, "tid": -914061504, "ts": 1716454222277250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222331949, "dur": 127,
    "args": {
      "External id": 61672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61672,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61672, "pid": 5, "tid": 7, "ts": 1716454222331949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277268, "dur": 6,
    "args": {
      "External id": 61672,
      "cbid": 211, "correlation": 61672
    }
  },
  {
    "ph": "s", "id": 61672, "pid": 76337, "tid": -914061504, "ts": 1716454222277268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222332078, "dur": 22,
    "args": {
      "External id": 61674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61674, "pid": 5, "tid": 7, "ts": 1716454222332078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277278, "dur": 5,
    "args": {
      "External id": 61674,
      "cbid": 211, "correlation": 61674
    }
  },
  {
    "ph": "s", "id": 61674, "pid": 76337, "tid": -914061504, "ts": 1716454222277278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222332101, "dur": 35,
    "args": {
      "External id": 61680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61680, "pid": 5, "tid": 7, "ts": 1716454222332101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277306, "dur": 8,
    "args": {
      "External id": 61680,
      "cbid": 211, "correlation": 61680
    }
  },
  {
    "ph": "s", "id": 61680, "pid": 76337, "tid": -914061504, "ts": 1716454222277306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222332137, "dur": 200,
    "args": {
      "External id": 61689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61689,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61689, "pid": 5, "tid": 7, "ts": 1716454222332137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277387, "dur": 14,
    "args": {
      "External id": 61689,
      "cbid": 211, "correlation": 61689
    }
  },
  {
    "ph": "s", "id": 61689, "pid": 76337, "tid": -914061504, "ts": 1716454222277387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222332338, "dur": 71,
    "args": {
      "External id": 61711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61711,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61711, "pid": 5, "tid": 7, "ts": 1716454222332338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277446, "dur": 11,
    "args": {
      "External id": 61711,
      "cbid": 211, "correlation": 61711
    }
  },
  {
    "ph": "s", "id": 61711, "pid": 76337, "tid": -914061504, "ts": 1716454222277446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277534, "dur": 1,
    "args": {
      "External id": 61722,
      "cbid": 251, "correlation": 61722
    }
  },
  {
    "ph": "f", "id": 61722, "pid": 76337, "tid": -914061504, "ts": 1716454222277534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222332410, "dur": 166,
    "args": {
      "External id": 61723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61723,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61723, "pid": 5, "tid": 7, "ts": 1716454222332410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277540, "dur": 13,
    "args": {
      "External id": 61723,
      "cbid": 211, "correlation": 61723
    }
  },
  {
    "ph": "s", "id": 61723, "pid": 76337, "tid": -914061504, "ts": 1716454222277540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277613, "dur": 1,
    "args": {
      "External id": 61734,
      "cbid": 251, "correlation": 61734
    }
  },
  {
    "ph": "f", "id": 61734, "pid": 76337, "tid": -914061504, "ts": 1716454222277613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222332577, "dur": 158,
    "args": {
      "External id": 61735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61735,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61735, "pid": 5, "tid": 7, "ts": 1716454222332577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277618, "dur": 12,
    "args": {
      "External id": 61735,
      "cbid": 211, "correlation": 61735
    }
  },
  {
    "ph": "s", "id": 61735, "pid": 76337, "tid": -914061504, "ts": 1716454222277618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277682, "dur": 1,
    "args": {
      "External id": 61746,
      "cbid": 251, "correlation": 61746
    }
  },
  {
    "ph": "f", "id": 61746, "pid": 76337, "tid": -914061504, "ts": 1716454222277682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222332736, "dur": 156,
    "args": {
      "External id": 61747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61747,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61747, "pid": 5, "tid": 7, "ts": 1716454222332736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277686, "dur": 10,
    "args": {
      "External id": 61747,
      "cbid": 211, "correlation": 61747
    }
  },
  {
    "ph": "s", "id": 61747, "pid": 76337, "tid": -914061504, "ts": 1716454222277686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222332894, "dur": 2145,
    "args": {
      "External id": 61768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61768,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 61768, "pid": 5, "tid": 7, "ts": 1716454222332894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277765, "dur": 13,
    "args": {
      "External id": 61768,
      "cbid": 211, "correlation": 61768
    }
  },
  {
    "ph": "s", "id": 61768, "pid": 76337, "tid": -914061504, "ts": 1716454222277765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277863, "dur": 1,
    "args": {
      "External id": 61786,
      "cbid": 251, "correlation": 61786
    }
  },
  {
    "ph": "f", "id": 61786, "pid": 76337, "tid": -914061504, "ts": 1716454222277863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222335040, "dur": 159,
    "args": {
      "External id": 61788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61788,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 61788, "pid": 5, "tid": 7, "ts": 1716454222335040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277868, "dur": 14,
    "args": {
      "External id": 61788,
      "cbid": 211, "correlation": 61788
    }
  },
  {
    "ph": "s", "id": 61788, "pid": 76337, "tid": -914061504, "ts": 1716454222277868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222335201, "dur": 36,
    "args": {
      "External id": 61796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61796,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61796, "pid": 5, "tid": 7, "ts": 1716454222335201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277939, "dur": 12,
    "args": {
      "External id": 61796,
      "cbid": 211, "correlation": 61796
    }
  },
  {
    "ph": "s", "id": 61796, "pid": 76337, "tid": -914061504, "ts": 1716454222277939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222335238, "dur": 51,
    "args": {
      "External id": 61804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61804,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61804, "pid": 5, "tid": 7, "ts": 1716454222335238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222277987, "dur": 12,
    "args": {
      "External id": 61804,
      "cbid": 211, "correlation": 61804
    }
  },
  {
    "ph": "s", "id": 61804, "pid": 76337, "tid": -914061504, "ts": 1716454222277987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222335290, "dur": 32,
    "args": {
      "External id": 61815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61815, "pid": 5, "tid": 7, "ts": 1716454222335290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278064, "dur": 13,
    "args": {
      "External id": 61815,
      "cbid": 211, "correlation": 61815
    }
  },
  {
    "ph": "s", "id": 61815, "pid": 76337, "tid": -914061504, "ts": 1716454222278064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222335324, "dur": 37,
    "args": {
      "External id": 61837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61837,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61837, "pid": 5, "tid": 7, "ts": 1716454222335324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278096, "dur": 8,
    "args": {
      "External id": 61837,
      "cbid": 211, "correlation": 61837
    }
  },
  {
    "ph": "s", "id": 61837, "pid": 76337, "tid": -914061504, "ts": 1716454222278096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278180, "dur": 1,
    "args": {
      "External id": 61848,
      "cbid": 251, "correlation": 61848
    }
  },
  {
    "ph": "f", "id": 61848, "pid": 76337, "tid": -914061504, "ts": 1716454222278180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222335363, "dur": 98,
    "args": {
      "External id": 61849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61849,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61849, "pid": 5, "tid": 7, "ts": 1716454222335363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278186, "dur": 13,
    "args": {
      "External id": 61849,
      "cbid": 211, "correlation": 61849
    }
  },
  {
    "ph": "s", "id": 61849, "pid": 76337, "tid": -914061504, "ts": 1716454222278186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278254, "dur": 1,
    "args": {
      "External id": 61860,
      "cbid": 251, "correlation": 61860
    }
  },
  {
    "ph": "f", "id": 61860, "pid": 76337, "tid": -914061504, "ts": 1716454222278254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278257, "dur": 0,
    "args": {
      "External id": 61861,
      "cbid": 251, "correlation": 61861
    }
  },
  {
    "ph": "f", "id": 61861, "pid": 76337, "tid": -914061504, "ts": 1716454222278257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222335462, "dur": 12,
    "args": {
      "External id": 61862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61862,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 61862, "pid": 5, "tid": 7, "ts": 1716454222335462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278259, "dur": 12,
    "args": {
      "External id": 61862,
      "cbid": 211, "correlation": 61862
    }
  },
  {
    "ph": "s", "id": 61862, "pid": 76337, "tid": -914061504, "ts": 1716454222278259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222335476, "dur": 5,
    "args": {
      "External id": 61864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61864,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 61864, "pid": 5, "tid": 7, "ts": 1716454222335476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278273, "dur": 6,
    "args": {
      "External id": 61864,
      "cbid": 211, "correlation": 61864
    }
  },
  {
    "ph": "s", "id": 61864, "pid": 76337, "tid": -914061504, "ts": 1716454222278273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278329, "dur": 1,
    "args": {
      "External id": 61875,
      "cbid": 251, "correlation": 61875
    }
  },
  {
    "ph": "f", "id": 61875, "pid": 76337, "tid": -914061504, "ts": 1716454222278329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278332, "dur": 0,
    "args": {
      "External id": 61876,
      "cbid": 251, "correlation": 61876
    }
  },
  {
    "ph": "f", "id": 61876, "pid": 76337, "tid": -914061504, "ts": 1716454222278332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222335483, "dur": 7,
    "args": {
      "External id": 61877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61877,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 61877, "pid": 5, "tid": 7, "ts": 1716454222335483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278334, "dur": 15,
    "args": {
      "External id": 61877,
      "cbid": 211, "correlation": 61877
    }
  },
  {
    "ph": "s", "id": 61877, "pid": 76337, "tid": -914061504, "ts": 1716454222278334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222335492, "dur": 3,
    "args": {
      "External id": 61879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61879,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 61879, "pid": 5, "tid": 7, "ts": 1716454222335492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278350, "dur": 6,
    "args": {
      "External id": 61879,
      "cbid": 211, "correlation": 61879
    }
  },
  {
    "ph": "s", "id": 61879, "pid": 76337, "tid": -914061504, "ts": 1716454222278350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222335497, "dur": 99,
    "args": {
      "External id": 61900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61900,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 61900, "pid": 5, "tid": 7, "ts": 1716454222335497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278424, "dur": 12,
    "args": {
      "External id": 61900,
      "cbid": 211, "correlation": 61900
    }
  },
  {
    "ph": "s", "id": 61900, "pid": 76337, "tid": -914061504, "ts": 1716454222278424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278521, "dur": 1,
    "args": {
      "External id": 61918,
      "cbid": 251, "correlation": 61918
    }
  },
  {
    "ph": "f", "id": 61918, "pid": 76337, "tid": -914061504, "ts": 1716454222278521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222335598, "dur": 104,
    "args": {
      "External id": 61920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61920,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 61920, "pid": 5, "tid": 7, "ts": 1716454222335598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278527, "dur": 13,
    "args": {
      "External id": 61920,
      "cbid": 211, "correlation": 61920
    }
  },
  {
    "ph": "s", "id": 61920, "pid": 76337, "tid": -914061504, "ts": 1716454222278527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222335703, "dur": 19,
    "args": {
      "External id": 61928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61928,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61928, "pid": 5, "tid": 7, "ts": 1716454222335703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278595, "dur": 12,
    "args": {
      "External id": 61928,
      "cbid": 211, "correlation": 61928
    }
  },
  {
    "ph": "s", "id": 61928, "pid": 76337, "tid": -914061504, "ts": 1716454222278595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222335724, "dur": 39,
    "args": {
      "External id": 61936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61936, "pid": 5, "tid": 7, "ts": 1716454222335724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278637, "dur": 9,
    "args": {
      "External id": 61936,
      "cbid": 211, "correlation": 61936
    }
  },
  {
    "ph": "s", "id": 61936, "pid": 76337, "tid": -914061504, "ts": 1716454222278637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222335765, "dur": 37,
    "args": {
      "External id": 61958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61958,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61958, "pid": 5, "tid": 7, "ts": 1716454222335765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278687, "dur": 13,
    "args": {
      "External id": 61958,
      "cbid": 211, "correlation": 61958
    }
  },
  {
    "ph": "s", "id": 61958, "pid": 76337, "tid": -914061504, "ts": 1716454222278687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278780, "dur": 1,
    "args": {
      "External id": 61974,
      "cbid": 251, "correlation": 61974
    }
  },
  {
    "ph": "f", "id": 61974, "pid": 76337, "tid": -914061504, "ts": 1716454222278780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278785, "dur": 0,
    "args": {
      "External id": 61976,
      "cbid": 251, "correlation": 61976
    }
  },
  {
    "ph": "f", "id": 61976, "pid": 76337, "tid": -914061504, "ts": 1716454222278785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222335803, "dur": 587,
    "args": {
      "External id": 61977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61977,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 61977, "pid": 5, "tid": 7, "ts": 1716454222335803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278789, "dur": 13,
    "args": {
      "External id": 61977,
      "cbid": 211, "correlation": 61977
    }
  },
  {
    "ph": "s", "id": 61977, "pid": 76337, "tid": -914061504, "ts": 1716454222278789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222336391, "dur": 133,
    "args": {
      "External id": 61985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61985,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61985, "pid": 5, "tid": 7, "ts": 1716454222336391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278854, "dur": 13,
    "args": {
      "External id": 61985,
      "cbid": 211, "correlation": 61985
    }
  },
  {
    "ph": "s", "id": 61985, "pid": 76337, "tid": -914061504, "ts": 1716454222278854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222336526, "dur": 134,
    "args": {
      "External id": 61993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 61993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 61993, "pid": 5, "tid": 7, "ts": 1716454222336526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278885, "dur": 8,
    "args": {
      "External id": 61993,
      "cbid": 211, "correlation": 61993
    }
  },
  {
    "ph": "s", "id": 61993, "pid": 76337, "tid": -914061504, "ts": 1716454222278885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278962, "dur": 1,
    "args": {
      "External id": 62009,
      "cbid": 251, "correlation": 62009
    }
  },
  {
    "ph": "f", "id": 62009, "pid": 76337, "tid": -914061504, "ts": 1716454222278962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222336662, "dur": 332,
    "args": {
      "External id": 62011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62011,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62011, "pid": 5, "tid": 7, "ts": 1716454222336662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222278967, "dur": 21,
    "args": {
      "External id": 62011,
      "cbid": 211, "correlation": 62011
    }
  },
  {
    "ph": "s", "id": 62011, "pid": 76337, "tid": -914061504, "ts": 1716454222278967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222336995, "dur": 28,
    "args": {
      "External id": 62019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62019,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62019, "pid": 5, "tid": 7, "ts": 1716454222336995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279019, "dur": 10,
    "args": {
      "External id": 62019,
      "cbid": 211, "correlation": 62019
    }
  },
  {
    "ph": "s", "id": 62019, "pid": 76337, "tid": -914061504, "ts": 1716454222279019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222337024, "dur": 89,
    "args": {
      "External id": 62030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62030, "pid": 5, "tid": 7, "ts": 1716454222337024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279087, "dur": 16,
    "args": {
      "External id": 62030,
      "cbid": 211, "correlation": 62030
    }
  },
  {
    "ph": "s", "id": 62030, "pid": 76337, "tid": -914061504, "ts": 1716454222279087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279155, "dur": 0,
    "args": {
      "External id": 62042,
      "cbid": 317, "correlation": 62042
    }
  },
  {
    "ph": "f", "id": 62042, "pid": 76337, "tid": -914061504, "ts": 1716454222279155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279156, "dur": 0,
    "args": {
      "External id": 62043,
      "cbid": 203, "correlation": 62043
    }
  },
  {
    "ph": "f", "id": 62043, "pid": 76337, "tid": -914061504, "ts": 1716454222279156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279156, "dur": 0,
    "args": {
      "External id": 62044,
      "cbid": 205, "correlation": 62044
    }
  },
  {
    "ph": "f", "id": 62044, "pid": 76337, "tid": -914061504, "ts": 1716454222279156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337115, "dur": 24,
    "args": {
      "External id": 62048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62048, "pid": 5, "tid": 7, "ts": 1716454222337115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279171, "dur": 12,
    "args": {
      "External id": 62048,
      "cbid": 211, "correlation": 62048
    }
  },
  {
    "ph": "s", "id": 62048, "pid": 76337, "tid": -914061504, "ts": 1716454222279171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222337140, "dur": 131,
    "args": {
      "External id": 62050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62050,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62050, "pid": 5, "tid": 7, "ts": 1716454222337140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279189, "dur": 6,
    "args": {
      "External id": 62050,
      "cbid": 211, "correlation": 62050
    }
  },
  {
    "ph": "s", "id": 62050, "pid": 76337, "tid": -914061504, "ts": 1716454222279189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337273, "dur": 24,
    "args": {
      "External id": 62052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62052,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62052, "pid": 5, "tid": 7, "ts": 1716454222337273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279199, "dur": 6,
    "args": {
      "External id": 62052,
      "cbid": 211, "correlation": 62052
    }
  },
  {
    "ph": "s", "id": 62052, "pid": 76337, "tid": -914061504, "ts": 1716454222279199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222337298, "dur": 34,
    "args": {
      "External id": 62058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62058,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62058, "pid": 5, "tid": 7, "ts": 1716454222337298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279227, "dur": 8,
    "args": {
      "External id": 62058,
      "cbid": 211, "correlation": 62058
    }
  },
  {
    "ph": "s", "id": 62058, "pid": 76337, "tid": -914061504, "ts": 1716454222279227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222337334, "dur": 27,
    "args": {
      "External id": 62066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62066,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62066, "pid": 5, "tid": 7, "ts": 1716454222337334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279260, "dur": 8,
    "args": {
      "External id": 62066,
      "cbid": 211, "correlation": 62066
    }
  },
  {
    "ph": "s", "id": 62066, "pid": 76337, "tid": -914061504, "ts": 1716454222279260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279337, "dur": 0,
    "args": {
      "External id": 62076,
      "cbid": 317, "correlation": 62076
    }
  },
  {
    "ph": "f", "id": 62076, "pid": 76337, "tid": -914061504, "ts": 1716454222279337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279337, "dur": 0,
    "args": {
      "External id": 62077,
      "cbid": 203, "correlation": 62077
    }
  },
  {
    "ph": "f", "id": 62077, "pid": 76337, "tid": -914061504, "ts": 1716454222279337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279338, "dur": 0,
    "args": {
      "External id": 62078,
      "cbid": 205, "correlation": 62078
    }
  },
  {
    "ph": "f", "id": 62078, "pid": 76337, "tid": -914061504, "ts": 1716454222279338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337363, "dur": 25,
    "args": {
      "External id": 62082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62082, "pid": 5, "tid": 7, "ts": 1716454222337363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279356, "dur": 13,
    "args": {
      "External id": 62082,
      "cbid": 211, "correlation": 62082
    }
  },
  {
    "ph": "s", "id": 62082, "pid": 76337, "tid": -914061504, "ts": 1716454222279356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337389, "dur": 48,
    "args": {
      "External id": 62084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62084, "pid": 5, "tid": 7, "ts": 1716454222337389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279372, "dur": 5,
    "args": {
      "External id": 62084,
      "cbid": 211, "correlation": 62084
    }
  },
  {
    "ph": "s", "id": 62084, "pid": 76337, "tid": -914061504, "ts": 1716454222279372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222337438, "dur": 248,
    "args": {
      "External id": 62086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62086,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 62086, "pid": 5, "tid": 7, "ts": 1716454222337438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279387, "dur": 7,
    "args": {
      "External id": 62086,
      "cbid": 211, "correlation": 62086
    }
  },
  {
    "ph": "s", "id": 62086, "pid": 76337, "tid": -914061504, "ts": 1716454222279387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337688, "dur": 7,
    "args": {
      "External id": 62088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62088, "pid": 5, "tid": 7, "ts": 1716454222337688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279398, "dur": 5,
    "args": {
      "External id": 62088,
      "cbid": 211, "correlation": 62088
    }
  },
  {
    "ph": "s", "id": 62088, "pid": 76337, "tid": -914061504, "ts": 1716454222279398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222337696, "dur": 10,
    "args": {
      "External id": 62094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62094, "pid": 5, "tid": 7, "ts": 1716454222337696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279426, "dur": 8,
    "args": {
      "External id": 62094,
      "cbid": 211, "correlation": 62094
    }
  },
  {
    "ph": "s", "id": 62094, "pid": 76337, "tid": -914061504, "ts": 1716454222279426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337708, "dur": 13,
    "args": {
      "External id": 62114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62114,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 62114, "pid": 5, "tid": 7, "ts": 1716454222337708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279517, "dur": 13,
    "args": {
      "External id": 62114,
      "cbid": 211, "correlation": 62114
    }
  },
  {
    "ph": "s", "id": 62114, "pid": 76337, "tid": -914061504, "ts": 1716454222279517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222337722, "dur": 5,
    "args": {
      "External id": 62126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62126,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 62126, "pid": 5, "tid": 7, "ts": 1716454222337722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279540, "dur": 6,
    "args": {
      "External id": 62126,
      "cbid": 211, "correlation": 62126
    }
  },
  {
    "ph": "s", "id": 62126, "pid": 76337, "tid": -914061504, "ts": 1716454222279540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222337728, "dur": 14,
    "args": {
      "External id": 62129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62129, "pid": 5, "tid": 7, "ts": 1716454222337728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279558, "dur": 6,
    "args": {
      "External id": 62129,
      "cbid": 211, "correlation": 62129
    }
  },
  {
    "ph": "s", "id": 62129, "pid": 76337, "tid": -914061504, "ts": 1716454222279558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222337743, "dur": 7,
    "args": {
      "External id": 62138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62138,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62138, "pid": 5, "tid": 7, "ts": 1716454222337743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279597, "dur": 10,
    "args": {
      "External id": 62138,
      "cbid": 211, "correlation": 62138
    }
  },
  {
    "ph": "s", "id": 62138, "pid": 76337, "tid": -914061504, "ts": 1716454222279597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279649, "dur": 0,
    "args": {
      "External id": 62148,
      "cbid": 317, "correlation": 62148
    }
  },
  {
    "ph": "f", "id": 62148, "pid": 76337, "tid": -914061504, "ts": 1716454222279649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279650, "dur": 0,
    "args": {
      "External id": 62149,
      "cbid": 203, "correlation": 62149
    }
  },
  {
    "ph": "f", "id": 62149, "pid": 76337, "tid": -914061504, "ts": 1716454222279650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279651, "dur": 0,
    "args": {
      "External id": 62150,
      "cbid": 205, "correlation": 62150
    }
  },
  {
    "ph": "f", "id": 62150, "pid": 76337, "tid": -914061504, "ts": 1716454222279651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337752, "dur": 6,
    "args": {
      "External id": 62154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62154,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62154, "pid": 5, "tid": 7, "ts": 1716454222337752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279679, "dur": 12,
    "args": {
      "External id": 62154,
      "cbid": 211, "correlation": 62154
    }
  },
  {
    "ph": "s", "id": 62154, "pid": 76337, "tid": -914061504, "ts": 1716454222279679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222337760, "dur": 91,
    "args": {
      "External id": 62156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62156, "pid": 5, "tid": 7, "ts": 1716454222337760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279694, "dur": 5,
    "args": {
      "External id": 62156,
      "cbid": 211, "correlation": 62156
    }
  },
  {
    "ph": "s", "id": 62156, "pid": 76337, "tid": -914061504, "ts": 1716454222279694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222337853, "dur": 1,
    "args": {
      "External id": 62158,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 62158,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5660377358490566
    }
  },
  {
    "ph": "f", "id": 62158, "pid": 5, "tid": 7, "ts": 1716454222337853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279712, "dur": 9,
    "args": {
      "External id": 62158,
      "cbid": 51, "correlation": 62158
    }
  },
  {
    "ph": "s", "id": 62158, "pid": 76337, "tid": -914061504, "ts": 1716454222279712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222337857, "dur": 571,
    "args": {
      "External id": 62159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62159,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62159, "pid": 5, "tid": 7, "ts": 1716454222337857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279723, "dur": 11,
    "args": {
      "External id": 62159,
      "cbid": 211, "correlation": 62159
    }
  },
  {
    "ph": "s", "id": 62159, "pid": 76337, "tid": -914061504, "ts": 1716454222279723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222338430, "dur": 13,
    "args": {
      "External id": 62161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62161,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62161, "pid": 5, "tid": 7, "ts": 1716454222338430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279738, "dur": 5,
    "args": {
      "External id": 62161,
      "cbid": 211, "correlation": 62161
    }
  },
  {
    "ph": "s", "id": 62161, "pid": 76337, "tid": -914061504, "ts": 1716454222279738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222338444, "dur": 16,
    "args": {
      "External id": 62167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62167, "pid": 5, "tid": 7, "ts": 1716454222338444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279766, "dur": 8,
    "args": {
      "External id": 62167,
      "cbid": 211, "correlation": 62167
    }
  },
  {
    "ph": "s", "id": 62167, "pid": 76337, "tid": -914061504, "ts": 1716454222279766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222338461, "dur": 4,
    "args": {
      "External id": 62175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62175,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 62175, "pid": 5, "tid": 7, "ts": 1716454222338461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279810, "dur": 10,
    "args": {
      "External id": 62175,
      "cbid": 211, "correlation": 62175
    }
  },
  {
    "ph": "s", "id": 62175, "pid": 76337, "tid": -914061504, "ts": 1716454222279810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279875, "dur": 1,
    "args": {
      "External id": 62191,
      "cbid": 251, "correlation": 62191
    }
  },
  {
    "ph": "f", "id": 62191, "pid": 76337, "tid": -914061504, "ts": 1716454222279875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279880, "dur": 0,
    "args": {
      "External id": 62193,
      "cbid": 251, "correlation": 62193
    }
  },
  {
    "ph": "f", "id": 62193, "pid": 76337, "tid": -914061504, "ts": 1716454222279880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222338466, "dur": 14,
    "args": {
      "External id": 62194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62194,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62194, "pid": 5, "tid": 7, "ts": 1716454222338466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279882, "dur": 12,
    "args": {
      "External id": 62194,
      "cbid": 211, "correlation": 62194
    }
  },
  {
    "ph": "s", "id": 62194, "pid": 76337, "tid": -914061504, "ts": 1716454222279882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222338482, "dur": 6,
    "args": {
      "External id": 62196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62196,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62196, "pid": 5, "tid": 7, "ts": 1716454222338482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279896, "dur": 5,
    "args": {
      "External id": 62196,
      "cbid": 211, "correlation": 62196
    }
  },
  {
    "ph": "s", "id": 62196, "pid": 76337, "tid": -914061504, "ts": 1716454222279896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222338488, "dur": 17,
    "args": {
      "External id": 62206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62206,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62206, "pid": 5, "tid": 7, "ts": 1716454222338488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222279953, "dur": 12,
    "args": {
      "External id": 62206,
      "cbid": 211, "correlation": 62206
    }
  },
  {
    "ph": "s", "id": 62206, "pid": 76337, "tid": -914061504, "ts": 1716454222279953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222338507, "dur": 19,
    "args": {
      "External id": 62226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62226,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 62226, "pid": 5, "tid": 7, "ts": 1716454222338507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280027, "dur": 12,
    "args": {
      "External id": 62226,
      "cbid": 211, "correlation": 62226
    }
  },
  {
    "ph": "s", "id": 62226, "pid": 76337, "tid": -914061504, "ts": 1716454222280027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222338528, "dur": 5,
    "args": {
      "External id": 62238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62238,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 62238, "pid": 5, "tid": 7, "ts": 1716454222338528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280053, "dur": 6,
    "args": {
      "External id": 62238,
      "cbid": 211, "correlation": 62238
    }
  },
  {
    "ph": "s", "id": 62238, "pid": 76337, "tid": -914061504, "ts": 1716454222280053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222338534, "dur": 19,
    "args": {
      "External id": 62241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62241, "pid": 5, "tid": 7, "ts": 1716454222338534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280071, "dur": 6,
    "args": {
      "External id": 62241,
      "cbid": 211, "correlation": 62241
    }
  },
  {
    "ph": "s", "id": 62241, "pid": 76337, "tid": -914061504, "ts": 1716454222280071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222338554, "dur": 12,
    "args": {
      "External id": 62250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62250,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62250, "pid": 5, "tid": 7, "ts": 1716454222338554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280112, "dur": 10,
    "args": {
      "External id": 62250,
      "cbid": 211, "correlation": 62250
    }
  },
  {
    "ph": "s", "id": 62250, "pid": 76337, "tid": -914061504, "ts": 1716454222280112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280174, "dur": 0,
    "args": {
      "External id": 62260,
      "cbid": 317, "correlation": 62260
    }
  },
  {
    "ph": "f", "id": 62260, "pid": 76337, "tid": -914061504, "ts": 1716454222280174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280175, "dur": 0,
    "args": {
      "External id": 62261,
      "cbid": 203, "correlation": 62261
    }
  },
  {
    "ph": "f", "id": 62261, "pid": 76337, "tid": -914061504, "ts": 1716454222280175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280176, "dur": 0,
    "args": {
      "External id": 62262,
      "cbid": 205, "correlation": 62262
    }
  },
  {
    "ph": "f", "id": 62262, "pid": 76337, "tid": -914061504, "ts": 1716454222280176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222338568, "dur": 12,
    "args": {
      "External id": 62266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62266,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62266, "pid": 5, "tid": 7, "ts": 1716454222338568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280200, "dur": 12,
    "args": {
      "External id": 62266,
      "cbid": 211, "correlation": 62266
    }
  },
  {
    "ph": "s", "id": 62266, "pid": 76337, "tid": -914061504, "ts": 1716454222280200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222338581, "dur": 178,
    "args": {
      "External id": 62268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62268, "pid": 5, "tid": 7, "ts": 1716454222338581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280215, "dur": 5,
    "args": {
      "External id": 62268,
      "cbid": 211, "correlation": 62268
    }
  },
  {
    "ph": "s", "id": 62268, "pid": 76337, "tid": -914061504, "ts": 1716454222280215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222338761, "dur": 1,
    "args": {
      "External id": 62270,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 62270,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5555555555555556
    }
  },
  {
    "ph": "f", "id": 62270, "pid": 5, "tid": 7, "ts": 1716454222338761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280229, "dur": 7,
    "args": {
      "External id": 62270,
      "cbid": 51, "correlation": 62270
    }
  },
  {
    "ph": "s", "id": 62270, "pid": 76337, "tid": -914061504, "ts": 1716454222280229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222338765, "dur": 729,
    "args": {
      "External id": 62271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62271,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62271, "pid": 5, "tid": 7, "ts": 1716454222338765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280237, "dur": 7,
    "args": {
      "External id": 62271,
      "cbid": 211, "correlation": 62271
    }
  },
  {
    "ph": "s", "id": 62271, "pid": 76337, "tid": -914061504, "ts": 1716454222280237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222339496, "dur": 13,
    "args": {
      "External id": 62273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62273,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62273, "pid": 5, "tid": 7, "ts": 1716454222339496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280249, "dur": 5,
    "args": {
      "External id": 62273,
      "cbid": 211, "correlation": 62273
    }
  },
  {
    "ph": "s", "id": 62273, "pid": 76337, "tid": -914061504, "ts": 1716454222280249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222339510, "dur": 16,
    "args": {
      "External id": 62279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62279, "pid": 5, "tid": 7, "ts": 1716454222339510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280277, "dur": 8,
    "args": {
      "External id": 62279,
      "cbid": 211, "correlation": 62279
    }
  },
  {
    "ph": "s", "id": 62279, "pid": 76337, "tid": -914061504, "ts": 1716454222280277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280335, "dur": 0,
    "args": {
      "External id": 62289,
      "cbid": 317, "correlation": 62289
    }
  },
  {
    "ph": "f", "id": 62289, "pid": 76337, "tid": -914061504, "ts": 1716454222280335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280336, "dur": 0,
    "args": {
      "External id": 62290,
      "cbid": 203, "correlation": 62290
    }
  },
  {
    "ph": "f", "id": 62290, "pid": 76337, "tid": -914061504, "ts": 1716454222280336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280337, "dur": 0,
    "args": {
      "External id": 62291,
      "cbid": 205, "correlation": 62291
    }
  },
  {
    "ph": "f", "id": 62291, "pid": 76337, "tid": -914061504, "ts": 1716454222280337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222339528, "dur": 9,
    "args": {
      "External id": 62295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62295,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62295, "pid": 5, "tid": 7, "ts": 1716454222339528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280358, "dur": 12,
    "args": {
      "External id": 62295,
      "cbid": 211, "correlation": 62295
    }
  },
  {
    "ph": "s", "id": 62295, "pid": 76337, "tid": -914061504, "ts": 1716454222280358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222339539, "dur": 4,
    "args": {
      "External id": 62297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62297,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 62297, "pid": 5, "tid": 7, "ts": 1716454222339539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280375, "dur": 6,
    "args": {
      "External id": 62297,
      "cbid": 211, "correlation": 62297
    }
  },
  {
    "ph": "s", "id": 62297, "pid": 76337, "tid": -914061504, "ts": 1716454222280375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280385, "dur": 0,
    "args": {
      "External id": 62298,
      "cbid": 51, "correlation": 62298
    }
  },
  {
    "ph": "s", "id": 62298, "pid": 76337, "tid": -914061504, "ts": 1716454222280385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222339544, "dur": 62,
    "args": {
      "External id": 62299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62299,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 62299, "pid": 5, "tid": 7, "ts": 1716454222339544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280385, "dur": 5,
    "args": {
      "External id": 62299,
      "cbid": 211, "correlation": 62299
    }
  },
  {
    "ph": "s", "id": 62299, "pid": 76337, "tid": -914061504, "ts": 1716454222280385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222339608, "dur": 16,
    "args": {
      "External id": 62304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62304, "pid": 5, "tid": 7, "ts": 1716454222339608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280412, "dur": 8,
    "args": {
      "External id": 62304,
      "cbid": 211, "correlation": 62304
    }
  },
  {
    "ph": "s", "id": 62304, "pid": 76337, "tid": -914061504, "ts": 1716454222280412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222339625, "dur": 13,
    "args": {
      "External id": 62312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62312,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62312, "pid": 5, "tid": 7, "ts": 1716454222339625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280441, "dur": 8,
    "args": {
      "External id": 62312,
      "cbid": 211, "correlation": 62312
    }
  },
  {
    "ph": "s", "id": 62312, "pid": 76337, "tid": -914061504, "ts": 1716454222280441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222339639, "dur": 11,
    "args": {
      "External id": 62320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62320,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62320, "pid": 5, "tid": 7, "ts": 1716454222339639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280471, "dur": 8,
    "args": {
      "External id": 62320,
      "cbid": 211, "correlation": 62320
    }
  },
  {
    "ph": "s", "id": 62320, "pid": 76337, "tid": -914061504, "ts": 1716454222280471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222339652, "dur": 20,
    "args": {
      "External id": 62340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62340,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 62340, "pid": 5, "tid": 7, "ts": 1716454222339652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280551, "dur": 13,
    "args": {
      "External id": 62340,
      "cbid": 211, "correlation": 62340
    }
  },
  {
    "ph": "s", "id": 62340, "pid": 76337, "tid": -914061504, "ts": 1716454222280551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222339673, "dur": 5,
    "args": {
      "External id": 62352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62352,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 62352, "pid": 5, "tid": 7, "ts": 1716454222339673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280573, "dur": 6,
    "args": {
      "External id": 62352,
      "cbid": 211, "correlation": 62352
    }
  },
  {
    "ph": "s", "id": 62352, "pid": 76337, "tid": -914061504, "ts": 1716454222280573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222339679, "dur": 19,
    "args": {
      "External id": 62355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62355, "pid": 5, "tid": 7, "ts": 1716454222339679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280591, "dur": 6,
    "args": {
      "External id": 62355,
      "cbid": 211, "correlation": 62355
    }
  },
  {
    "ph": "s", "id": 62355, "pid": 76337, "tid": -914061504, "ts": 1716454222280591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280652, "dur": 0,
    "args": {
      "External id": 62366,
      "cbid": 317, "correlation": 62366
    }
  },
  {
    "ph": "f", "id": 62366, "pid": 76337, "tid": -914061504, "ts": 1716454222280652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280653, "dur": 0,
    "args": {
      "External id": 62367,
      "cbid": 203, "correlation": 62367
    }
  },
  {
    "ph": "f", "id": 62367, "pid": 76337, "tid": -914061504, "ts": 1716454222280653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280653, "dur": 0,
    "args": {
      "External id": 62368,
      "cbid": 205, "correlation": 62368
    }
  },
  {
    "ph": "f", "id": 62368, "pid": 76337, "tid": -914061504, "ts": 1716454222280653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222339700, "dur": 11,
    "args": {
      "External id": 62372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62372,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62372, "pid": 5, "tid": 7, "ts": 1716454222339700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280673, "dur": 12,
    "args": {
      "External id": 62372,
      "cbid": 211, "correlation": 62372
    }
  },
  {
    "ph": "s", "id": 62372, "pid": 76337, "tid": -914061504, "ts": 1716454222280673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222339713, "dur": 4,
    "args": {
      "External id": 62374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 62374, "pid": 5, "tid": 7, "ts": 1716454222339713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280694, "dur": 6,
    "args": {
      "External id": 62374,
      "cbid": 211, "correlation": 62374
    }
  },
  {
    "ph": "s", "id": 62374, "pid": 76337, "tid": -914061504, "ts": 1716454222280694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280703, "dur": 0,
    "args": {
      "External id": 62375,
      "cbid": 51, "correlation": 62375
    }
  },
  {
    "ph": "s", "id": 62375, "pid": 76337, "tid": -914061504, "ts": 1716454222280703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222339718, "dur": 107,
    "args": {
      "External id": 62376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62376,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 62376, "pid": 5, "tid": 7, "ts": 1716454222339718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280704, "dur": 5,
    "args": {
      "External id": 62376,
      "cbid": 211, "correlation": 62376
    }
  },
  {
    "ph": "s", "id": 62376, "pid": 76337, "tid": -914061504, "ts": 1716454222280704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222339826, "dur": 17,
    "args": {
      "External id": 62381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62381, "pid": 5, "tid": 7, "ts": 1716454222339826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280733, "dur": 8,
    "args": {
      "External id": 62381,
      "cbid": 211, "correlation": 62381
    }
  },
  {
    "ph": "s", "id": 62381, "pid": 76337, "tid": -914061504, "ts": 1716454222280733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222339844, "dur": 92,
    "args": {
      "External id": 62390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62390,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62390, "pid": 5, "tid": 7, "ts": 1716454222339844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280816, "dur": 15,
    "args": {
      "External id": 62390,
      "cbid": 211, "correlation": 62390
    }
  },
  {
    "ph": "s", "id": 62390, "pid": 76337, "tid": -914061504, "ts": 1716454222280816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222339937, "dur": 33,
    "args": {
      "External id": 62412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62412,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62412, "pid": 5, "tid": 7, "ts": 1716454222339937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280873, "dur": 10,
    "args": {
      "External id": 62412,
      "cbid": 211, "correlation": 62412
    }
  },
  {
    "ph": "s", "id": 62412, "pid": 76337, "tid": -914061504, "ts": 1716454222280873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280965, "dur": 1,
    "args": {
      "External id": 62423,
      "cbid": 251, "correlation": 62423
    }
  },
  {
    "ph": "f", "id": 62423, "pid": 76337, "tid": -914061504, "ts": 1716454222280965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222339971, "dur": 154,
    "args": {
      "External id": 62424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62424,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62424, "pid": 5, "tid": 7, "ts": 1716454222339971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222280970, "dur": 23,
    "args": {
      "External id": 62424,
      "cbid": 211, "correlation": 62424
    }
  },
  {
    "ph": "s", "id": 62424, "pid": 76337, "tid": -914061504, "ts": 1716454222280970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281056, "dur": 1,
    "args": {
      "External id": 62435,
      "cbid": 251, "correlation": 62435
    }
  },
  {
    "ph": "f", "id": 62435, "pid": 76337, "tid": -914061504, "ts": 1716454222281056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222340126, "dur": 173,
    "args": {
      "External id": 62436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62436,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62436, "pid": 5, "tid": 7, "ts": 1716454222340126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281060, "dur": 11,
    "args": {
      "External id": 62436,
      "cbid": 211, "correlation": 62436
    }
  },
  {
    "ph": "s", "id": 62436, "pid": 76337, "tid": -914061504, "ts": 1716454222281060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281125, "dur": 1,
    "args": {
      "External id": 62447,
      "cbid": 251, "correlation": 62447
    }
  },
  {
    "ph": "f", "id": 62447, "pid": 76337, "tid": -914061504, "ts": 1716454222281125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222340301, "dur": 147,
    "args": {
      "External id": 62448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62448,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62448, "pid": 5, "tid": 7, "ts": 1716454222340301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281129, "dur": 11,
    "args": {
      "External id": 62448,
      "cbid": 211, "correlation": 62448
    }
  },
  {
    "ph": "s", "id": 62448, "pid": 76337, "tid": -914061504, "ts": 1716454222281129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222340449, "dur": 364,
    "args": {
      "External id": 62473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62473,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62473, "pid": 5, "tid": 7, "ts": 1716454222340449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281214, "dur": 13,
    "args": {
      "External id": 62473,
      "cbid": 211, "correlation": 62473
    }
  },
  {
    "ph": "s", "id": 62473, "pid": 76337, "tid": -914061504, "ts": 1716454222281214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281316, "dur": 1,
    "args": {
      "External id": 62491,
      "cbid": 251, "correlation": 62491
    }
  },
  {
    "ph": "f", "id": 62491, "pid": 76337, "tid": -914061504, "ts": 1716454222281316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222340815, "dur": 183,
    "args": {
      "External id": 62493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62493,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62493, "pid": 5, "tid": 7, "ts": 1716454222340815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281322, "dur": 13,
    "args": {
      "External id": 62493,
      "cbid": 211, "correlation": 62493
    }
  },
  {
    "ph": "s", "id": 62493, "pid": 76337, "tid": -914061504, "ts": 1716454222281322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222340999, "dur": 20,
    "args": {
      "External id": 62501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62501,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62501, "pid": 5, "tid": 7, "ts": 1716454222340999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281391, "dur": 13,
    "args": {
      "External id": 62501,
      "cbid": 211, "correlation": 62501
    }
  },
  {
    "ph": "s", "id": 62501, "pid": 76337, "tid": -914061504, "ts": 1716454222281391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222341021, "dur": 28,
    "args": {
      "External id": 62509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62509,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62509, "pid": 5, "tid": 7, "ts": 1716454222341021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281431, "dur": 11,
    "args": {
      "External id": 62509,
      "cbid": 211, "correlation": 62509
    }
  },
  {
    "ph": "s", "id": 62509, "pid": 76337, "tid": -914061504, "ts": 1716454222281431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222341049, "dur": 20,
    "args": {
      "External id": 62520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62520, "pid": 5, "tid": 7, "ts": 1716454222341049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281508, "dur": 13,
    "args": {
      "External id": 62520,
      "cbid": 211, "correlation": 62520
    }
  },
  {
    "ph": "s", "id": 62520, "pid": 76337, "tid": -914061504, "ts": 1716454222281508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222341071, "dur": 18,
    "args": {
      "External id": 62542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62542,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62542, "pid": 5, "tid": 7, "ts": 1716454222341071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281539, "dur": 7,
    "args": {
      "External id": 62542,
      "cbid": 211, "correlation": 62542
    }
  },
  {
    "ph": "s", "id": 62542, "pid": 76337, "tid": -914061504, "ts": 1716454222281539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281625, "dur": 1,
    "args": {
      "External id": 62553,
      "cbid": 251, "correlation": 62553
    }
  },
  {
    "ph": "f", "id": 62553, "pid": 76337, "tid": -914061504, "ts": 1716454222281625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222341090, "dur": 97,
    "args": {
      "External id": 62554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62554,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 62554, "pid": 5, "tid": 7, "ts": 1716454222341090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281631, "dur": 14,
    "args": {
      "External id": 62554,
      "cbid": 211, "correlation": 62554
    }
  },
  {
    "ph": "s", "id": 62554, "pid": 76337, "tid": -914061504, "ts": 1716454222281631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281702, "dur": 1,
    "args": {
      "External id": 62565,
      "cbid": 251, "correlation": 62565
    }
  },
  {
    "ph": "f", "id": 62565, "pid": 76337, "tid": -914061504, "ts": 1716454222281702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281706, "dur": 0,
    "args": {
      "External id": 62566,
      "cbid": 251, "correlation": 62566
    }
  },
  {
    "ph": "f", "id": 62566, "pid": 76337, "tid": -914061504, "ts": 1716454222281706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222341189, "dur": 13,
    "args": {
      "External id": 62567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62567,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62567, "pid": 5, "tid": 7, "ts": 1716454222341189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281707, "dur": 12,
    "args": {
      "External id": 62567,
      "cbid": 211, "correlation": 62567
    }
  },
  {
    "ph": "s", "id": 62567, "pid": 76337, "tid": -914061504, "ts": 1716454222281707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222341204, "dur": 6,
    "args": {
      "External id": 62569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62569,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62569, "pid": 5, "tid": 7, "ts": 1716454222341204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281721, "dur": 6,
    "args": {
      "External id": 62569,
      "cbid": 211, "correlation": 62569
    }
  },
  {
    "ph": "s", "id": 62569, "pid": 76337, "tid": -914061504, "ts": 1716454222281721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281779, "dur": 1,
    "args": {
      "External id": 62580,
      "cbid": 251, "correlation": 62580
    }
  },
  {
    "ph": "f", "id": 62580, "pid": 76337, "tid": -914061504, "ts": 1716454222281779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281783, "dur": 0,
    "args": {
      "External id": 62581,
      "cbid": 251, "correlation": 62581
    }
  },
  {
    "ph": "f", "id": 62581, "pid": 76337, "tid": -914061504, "ts": 1716454222281783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222341211, "dur": 9,
    "args": {
      "External id": 62582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62582,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62582, "pid": 5, "tid": 7, "ts": 1716454222341211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281784, "dur": 14,
    "args": {
      "External id": 62582,
      "cbid": 211, "correlation": 62582
    }
  },
  {
    "ph": "s", "id": 62582, "pid": 76337, "tid": -914061504, "ts": 1716454222281784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222341222, "dur": 3,
    "args": {
      "External id": 62584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62584,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62584, "pid": 5, "tid": 7, "ts": 1716454222341222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281800, "dur": 6,
    "args": {
      "External id": 62584,
      "cbid": 211, "correlation": 62584
    }
  },
  {
    "ph": "s", "id": 62584, "pid": 76337, "tid": -914061504, "ts": 1716454222281800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222341227, "dur": 59,
    "args": {
      "External id": 62609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62609,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62609, "pid": 5, "tid": 7, "ts": 1716454222341227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281878, "dur": 13,
    "args": {
      "External id": 62609,
      "cbid": 211, "correlation": 62609
    }
  },
  {
    "ph": "s", "id": 62609, "pid": 76337, "tid": -914061504, "ts": 1716454222281878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281985, "dur": 2,
    "args": {
      "External id": 62627,
      "cbid": 251, "correlation": 62627
    }
  },
  {
    "ph": "f", "id": 62627, "pid": 76337, "tid": -914061504, "ts": 1716454222281985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222341287, "dur": 99,
    "args": {
      "External id": 62629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62629,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 62629, "pid": 5, "tid": 7, "ts": 1716454222341287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222281992, "dur": 14,
    "args": {
      "External id": 62629,
      "cbid": 211, "correlation": 62629
    }
  },
  {
    "ph": "s", "id": 62629, "pid": 76337, "tid": -914061504, "ts": 1716454222281992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222341388, "dur": 10,
    "args": {
      "External id": 62637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62637,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62637, "pid": 5, "tid": 7, "ts": 1716454222341388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282064, "dur": 12,
    "args": {
      "External id": 62637,
      "cbid": 211, "correlation": 62637
    }
  },
  {
    "ph": "s", "id": 62637, "pid": 76337, "tid": -914061504, "ts": 1716454222282064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222341399, "dur": 24,
    "args": {
      "External id": 62645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62645,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62645, "pid": 5, "tid": 7, "ts": 1716454222341399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282105, "dur": 9,
    "args": {
      "External id": 62645,
      "cbid": 211, "correlation": 62645
    }
  },
  {
    "ph": "s", "id": 62645, "pid": 76337, "tid": -914061504, "ts": 1716454222282105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222341425, "dur": 20,
    "args": {
      "External id": 62667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62667,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62667, "pid": 5, "tid": 7, "ts": 1716454222341425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282161, "dur": 10,
    "args": {
      "External id": 62667,
      "cbid": 211, "correlation": 62667
    }
  },
  {
    "ph": "s", "id": 62667, "pid": 76337, "tid": -914061504, "ts": 1716454222282161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282250, "dur": 2,
    "args": {
      "External id": 62683,
      "cbid": 251, "correlation": 62683
    }
  },
  {
    "ph": "f", "id": 62683, "pid": 76337, "tid": -914061504, "ts": 1716454222282250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282257, "dur": 0,
    "args": {
      "External id": 62685,
      "cbid": 251, "correlation": 62685
    }
  },
  {
    "ph": "f", "id": 62685, "pid": 76337, "tid": -914061504, "ts": 1716454222282257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222341446, "dur": 525,
    "args": {
      "External id": 62686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62686,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62686, "pid": 5, "tid": 7, "ts": 1716454222341446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282259, "dur": 15,
    "args": {
      "External id": 62686,
      "cbid": 211, "correlation": 62686
    }
  },
  {
    "ph": "s", "id": 62686, "pid": 76337, "tid": -914061504, "ts": 1716454222282259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222341972, "dur": 71,
    "args": {
      "External id": 62694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62694, "pid": 5, "tid": 7, "ts": 1716454222341972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282327, "dur": 12,
    "args": {
      "External id": 62694,
      "cbid": 211, "correlation": 62694
    }
  },
  {
    "ph": "s", "id": 62694, "pid": 76337, "tid": -914061504, "ts": 1716454222282327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222342044, "dur": 70,
    "args": {
      "External id": 62702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62702, "pid": 5, "tid": 7, "ts": 1716454222342044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282357, "dur": 9,
    "args": {
      "External id": 62702,
      "cbid": 211, "correlation": 62702
    }
  },
  {
    "ph": "s", "id": 62702, "pid": 76337, "tid": -914061504, "ts": 1716454222282357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282437, "dur": 1,
    "args": {
      "External id": 62718,
      "cbid": 251, "correlation": 62718
    }
  },
  {
    "ph": "f", "id": 62718, "pid": 76337, "tid": -914061504, "ts": 1716454222282437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222342116, "dur": 1,
    "args": {
      "External id": 62720,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 62720,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14714898835070508
    }
  },
  {
    "ph": "f", "id": 62720, "pid": 5, "tid": 7, "ts": 1716454222342116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282443, "dur": 11,
    "args": {
      "External id": 62720,
      "cbid": 51, "correlation": 62720
    }
  },
  {
    "ph": "s", "id": 62720, "pid": 76337, "tid": -914061504, "ts": 1716454222282443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222342120, "dur": 289,
    "args": {
      "External id": 62721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62721,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 62721, "pid": 5, "tid": 7, "ts": 1716454222342120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282455, "dur": 12,
    "args": {
      "External id": 62721,
      "cbid": 211, "correlation": 62721
    }
  },
  {
    "ph": "s", "id": 62721, "pid": 76337, "tid": -914061504, "ts": 1716454222282455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222342411, "dur": 15,
    "args": {
      "External id": 62729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62729,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62729, "pid": 5, "tid": 7, "ts": 1716454222342411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282498, "dur": 10,
    "args": {
      "External id": 62729,
      "cbid": 211, "correlation": 62729
    }
  },
  {
    "ph": "s", "id": 62729, "pid": 76337, "tid": -914061504, "ts": 1716454222282498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222342427, "dur": 41,
    "args": {
      "External id": 62740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62740, "pid": 5, "tid": 7, "ts": 1716454222342427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282568, "dur": 14,
    "args": {
      "External id": 62740,
      "cbid": 211, "correlation": 62740
    }
  },
  {
    "ph": "s", "id": 62740, "pid": 76337, "tid": -914061504, "ts": 1716454222282568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282635, "dur": 0,
    "args": {
      "External id": 62752,
      "cbid": 317, "correlation": 62752
    }
  },
  {
    "ph": "f", "id": 62752, "pid": 76337, "tid": -914061504, "ts": 1716454222282635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282636, "dur": 0,
    "args": {
      "External id": 62753,
      "cbid": 203, "correlation": 62753
    }
  },
  {
    "ph": "f", "id": 62753, "pid": 76337, "tid": -914061504, "ts": 1716454222282636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282637, "dur": 0,
    "args": {
      "External id": 62754,
      "cbid": 205, "correlation": 62754
    }
  },
  {
    "ph": "f", "id": 62754, "pid": 76337, "tid": -914061504, "ts": 1716454222282637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222342470, "dur": 15,
    "args": {
      "External id": 62758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62758,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62758, "pid": 5, "tid": 7, "ts": 1716454222342470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282652, "dur": 12,
    "args": {
      "External id": 62758,
      "cbid": 211, "correlation": 62758
    }
  },
  {
    "ph": "s", "id": 62758, "pid": 76337, "tid": -914061504, "ts": 1716454222282652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222342486, "dur": 4,
    "args": {
      "External id": 62760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62760,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 62760, "pid": 5, "tid": 7, "ts": 1716454222342486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282668, "dur": 6,
    "args": {
      "External id": 62760,
      "cbid": 211, "correlation": 62760
    }
  },
  {
    "ph": "s", "id": 62760, "pid": 76337, "tid": -914061504, "ts": 1716454222282668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282677, "dur": 0,
    "args": {
      "External id": 62761,
      "cbid": 51, "correlation": 62761
    }
  },
  {
    "ph": "s", "id": 62761, "pid": 76337, "tid": -914061504, "ts": 1716454222282677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222342491, "dur": 106,
    "args": {
      "External id": 62762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62762,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 62762, "pid": 5, "tid": 7, "ts": 1716454222342491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282677, "dur": 5,
    "args": {
      "External id": 62762,
      "cbid": 211, "correlation": 62762
    }
  },
  {
    "ph": "s", "id": 62762, "pid": 76337, "tid": -914061504, "ts": 1716454222282677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222342599, "dur": 17,
    "args": {
      "External id": 62767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62767, "pid": 5, "tid": 7, "ts": 1716454222342599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282705, "dur": 9,
    "args": {
      "External id": 62767,
      "cbid": 211, "correlation": 62767
    }
  },
  {
    "ph": "s", "id": 62767, "pid": 76337, "tid": -914061504, "ts": 1716454222282705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222342617, "dur": 12,
    "args": {
      "External id": 62775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62775,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62775, "pid": 5, "tid": 7, "ts": 1716454222342617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282736, "dur": 9,
    "args": {
      "External id": 62775,
      "cbid": 211, "correlation": 62775
    }
  },
  {
    "ph": "s", "id": 62775, "pid": 76337, "tid": -914061504, "ts": 1716454222282736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222342630, "dur": 19,
    "args": {
      "External id": 62795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62795,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 62795, "pid": 5, "tid": 7, "ts": 1716454222342630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282808, "dur": 12,
    "args": {
      "External id": 62795,
      "cbid": 211, "correlation": 62795
    }
  },
  {
    "ph": "s", "id": 62795, "pid": 76337, "tid": -914061504, "ts": 1716454222282808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222342651, "dur": 5,
    "args": {
      "External id": 62807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62807,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 62807, "pid": 5, "tid": 7, "ts": 1716454222342651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282834, "dur": 7,
    "args": {
      "External id": 62807,
      "cbid": 211, "correlation": 62807
    }
  },
  {
    "ph": "s", "id": 62807, "pid": 76337, "tid": -914061504, "ts": 1716454222282834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222342657, "dur": 19,
    "args": {
      "External id": 62810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62810, "pid": 5, "tid": 7, "ts": 1716454222342657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282852, "dur": 7,
    "args": {
      "External id": 62810,
      "cbid": 211, "correlation": 62810
    }
  },
  {
    "ph": "s", "id": 62810, "pid": 76337, "tid": -914061504, "ts": 1716454222282852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222342678, "dur": 12,
    "args": {
      "External id": 62819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62819,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62819, "pid": 5, "tid": 7, "ts": 1716454222342678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282891, "dur": 10,
    "args": {
      "External id": 62819,
      "cbid": 211, "correlation": 62819
    }
  },
  {
    "ph": "s", "id": 62819, "pid": 76337, "tid": -914061504, "ts": 1716454222282891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282942, "dur": 0,
    "args": {
      "External id": 62829,
      "cbid": 317, "correlation": 62829
    }
  },
  {
    "ph": "f", "id": 62829, "pid": 76337, "tid": -914061504, "ts": 1716454222282942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282943, "dur": 0,
    "args": {
      "External id": 62830,
      "cbid": 203, "correlation": 62830
    }
  },
  {
    "ph": "f", "id": 62830, "pid": 76337, "tid": -914061504, "ts": 1716454222282943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282944, "dur": 0,
    "args": {
      "External id": 62831,
      "cbid": 205, "correlation": 62831
    }
  },
  {
    "ph": "f", "id": 62831, "pid": 76337, "tid": -914061504, "ts": 1716454222282944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222342691, "dur": 12,
    "args": {
      "External id": 62835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62835,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62835, "pid": 5, "tid": 7, "ts": 1716454222342691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282957, "dur": 12,
    "args": {
      "External id": 62835,
      "cbid": 211, "correlation": 62835
    }
  },
  {
    "ph": "s", "id": 62835, "pid": 76337, "tid": -914061504, "ts": 1716454222282957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222342705, "dur": 177,
    "args": {
      "External id": 62837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62837,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62837, "pid": 5, "tid": 7, "ts": 1716454222342705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282972, "dur": 14,
    "args": {
      "External id": 62837,
      "cbid": 211, "correlation": 62837
    }
  },
  {
    "ph": "s", "id": 62837, "pid": 76337, "tid": -914061504, "ts": 1716454222282972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222342884, "dur": 1,
    "args": {
      "External id": 62839,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 62839,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5555555555555556
    }
  },
  {
    "ph": "f", "id": 62839, "pid": 5, "tid": 7, "ts": 1716454222342884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222282993, "dur": 6,
    "args": {
      "External id": 62839,
      "cbid": 51, "correlation": 62839
    }
  },
  {
    "ph": "s", "id": 62839, "pid": 76337, "tid": -914061504, "ts": 1716454222282993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222342888, "dur": 727,
    "args": {
      "External id": 62840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62840,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62840, "pid": 5, "tid": 7, "ts": 1716454222342888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283000, "dur": 7,
    "args": {
      "External id": 62840,
      "cbid": 211, "correlation": 62840
    }
  },
  {
    "ph": "s", "id": 62840, "pid": 76337, "tid": -914061504, "ts": 1716454222283000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222343617, "dur": 14,
    "args": {
      "External id": 62842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62842,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62842, "pid": 5, "tid": 7, "ts": 1716454222343617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283011, "dur": 5,
    "args": {
      "External id": 62842,
      "cbid": 211, "correlation": 62842
    }
  },
  {
    "ph": "s", "id": 62842, "pid": 76337, "tid": -914061504, "ts": 1716454222283011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222343633, "dur": 16,
    "args": {
      "External id": 62848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62848, "pid": 5, "tid": 7, "ts": 1716454222343633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283041, "dur": 8,
    "args": {
      "External id": 62848,
      "cbid": 211, "correlation": 62848
    }
  },
  {
    "ph": "s", "id": 62848, "pid": 76337, "tid": -914061504, "ts": 1716454222283041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222343650, "dur": 4,
    "args": {
      "External id": 62856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62856,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 62856, "pid": 5, "tid": 7, "ts": 1716454222343650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283085, "dur": 9,
    "args": {
      "External id": 62856,
      "cbid": 211, "correlation": 62856
    }
  },
  {
    "ph": "s", "id": 62856, "pid": 76337, "tid": -914061504, "ts": 1716454222283085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283154, "dur": 1,
    "args": {
      "External id": 62872,
      "cbid": 251, "correlation": 62872
    }
  },
  {
    "ph": "f", "id": 62872, "pid": 76337, "tid": -914061504, "ts": 1716454222283154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283159, "dur": 0,
    "args": {
      "External id": 62874,
      "cbid": 251, "correlation": 62874
    }
  },
  {
    "ph": "f", "id": 62874, "pid": 76337, "tid": -914061504, "ts": 1716454222283159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222343655, "dur": 14,
    "args": {
      "External id": 62875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62875,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62875, "pid": 5, "tid": 7, "ts": 1716454222343655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283161, "dur": 11,
    "args": {
      "External id": 62875,
      "cbid": 211, "correlation": 62875
    }
  },
  {
    "ph": "s", "id": 62875, "pid": 76337, "tid": -914061504, "ts": 1716454222283161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222343671, "dur": 6,
    "args": {
      "External id": 62877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62877,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62877, "pid": 5, "tid": 7, "ts": 1716454222343671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283174, "dur": 6,
    "args": {
      "External id": 62877,
      "cbid": 211, "correlation": 62877
    }
  },
  {
    "ph": "s", "id": 62877, "pid": 76337, "tid": -914061504, "ts": 1716454222283174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222343678, "dur": 19,
    "args": {
      "External id": 62887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62887, "pid": 5, "tid": 7, "ts": 1716454222343678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283233, "dur": 12,
    "args": {
      "External id": 62887,
      "cbid": 211, "correlation": 62887
    }
  },
  {
    "ph": "s", "id": 62887, "pid": 76337, "tid": -914061504, "ts": 1716454222283233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222343699, "dur": 20,
    "args": {
      "External id": 62907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62907,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 62907, "pid": 5, "tid": 7, "ts": 1716454222343699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283298, "dur": 11,
    "args": {
      "External id": 62907,
      "cbid": 211, "correlation": 62907
    }
  },
  {
    "ph": "s", "id": 62907, "pid": 76337, "tid": -914061504, "ts": 1716454222283298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222343720, "dur": 4,
    "args": {
      "External id": 62919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62919,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 62919, "pid": 5, "tid": 7, "ts": 1716454222343720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283319, "dur": 6,
    "args": {
      "External id": 62919,
      "cbid": 211, "correlation": 62919
    }
  },
  {
    "ph": "s", "id": 62919, "pid": 76337, "tid": -914061504, "ts": 1716454222283319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222343726, "dur": 18,
    "args": {
      "External id": 62922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62922, "pid": 5, "tid": 7, "ts": 1716454222343726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283338, "dur": 6,
    "args": {
      "External id": 62922,
      "cbid": 211, "correlation": 62922
    }
  },
  {
    "ph": "s", "id": 62922, "pid": 76337, "tid": -914061504, "ts": 1716454222283338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222343745, "dur": 12,
    "args": {
      "External id": 62931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62931,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62931, "pid": 5, "tid": 7, "ts": 1716454222343745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283378, "dur": 10,
    "args": {
      "External id": 62931,
      "cbid": 211, "correlation": 62931
    }
  },
  {
    "ph": "s", "id": 62931, "pid": 76337, "tid": -914061504, "ts": 1716454222283378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283443, "dur": 0,
    "args": {
      "External id": 62941,
      "cbid": 317, "correlation": 62941
    }
  },
  {
    "ph": "f", "id": 62941, "pid": 76337, "tid": -914061504, "ts": 1716454222283443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283444, "dur": 0,
    "args": {
      "External id": 62942,
      "cbid": 203, "correlation": 62942
    }
  },
  {
    "ph": "f", "id": 62942, "pid": 76337, "tid": -914061504, "ts": 1716454222283444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283445, "dur": 0,
    "args": {
      "External id": 62943,
      "cbid": 205, "correlation": 62943
    }
  },
  {
    "ph": "f", "id": 62943, "pid": 76337, "tid": -914061504, "ts": 1716454222283445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222343758, "dur": 12,
    "args": {
      "External id": 62947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62947, "pid": 5, "tid": 7, "ts": 1716454222343758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283459, "dur": 12,
    "args": {
      "External id": 62947,
      "cbid": 211, "correlation": 62947
    }
  },
  {
    "ph": "s", "id": 62947, "pid": 76337, "tid": -914061504, "ts": 1716454222283459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222343771, "dur": 178,
    "args": {
      "External id": 62949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62949, "pid": 5, "tid": 7, "ts": 1716454222343771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283474, "dur": 5,
    "args": {
      "External id": 62949,
      "cbid": 211, "correlation": 62949
    }
  },
  {
    "ph": "s", "id": 62949, "pid": 76337, "tid": -914061504, "ts": 1716454222283474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222343951, "dur": 1,
    "args": {
      "External id": 62951,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 62951,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5555555555555556
    }
  },
  {
    "ph": "f", "id": 62951, "pid": 5, "tid": 7, "ts": 1716454222343951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283485, "dur": 7,
    "args": {
      "External id": 62951,
      "cbid": 51, "correlation": 62951
    }
  },
  {
    "ph": "s", "id": 62951, "pid": 76337, "tid": -914061504, "ts": 1716454222283485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222343955, "dur": 712,
    "args": {
      "External id": 62952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62952,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 62952, "pid": 5, "tid": 7, "ts": 1716454222343955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283493, "dur": 6,
    "args": {
      "External id": 62952,
      "cbid": 211, "correlation": 62952
    }
  },
  {
    "ph": "s", "id": 62952, "pid": 76337, "tid": -914061504, "ts": 1716454222283493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222344669, "dur": 13,
    "args": {
      "External id": 62954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62954,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62954, "pid": 5, "tid": 7, "ts": 1716454222344669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283502, "dur": 5,
    "args": {
      "External id": 62954,
      "cbid": 211, "correlation": 62954
    }
  },
  {
    "ph": "s", "id": 62954, "pid": 76337, "tid": -914061504, "ts": 1716454222283502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222344683, "dur": 16,
    "args": {
      "External id": 62960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62960, "pid": 5, "tid": 7, "ts": 1716454222344683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283530, "dur": 9,
    "args": {
      "External id": 62960,
      "cbid": 211, "correlation": 62960
    }
  },
  {
    "ph": "s", "id": 62960, "pid": 76337, "tid": -914061504, "ts": 1716454222283530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222344700, "dur": 13,
    "args": {
      "External id": 62968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62968,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62968, "pid": 5, "tid": 7, "ts": 1716454222344700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283563, "dur": 8,
    "args": {
      "External id": 62968,
      "cbid": 211, "correlation": 62968
    }
  },
  {
    "ph": "s", "id": 62968, "pid": 76337, "tid": -914061504, "ts": 1716454222283563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222344715, "dur": 11,
    "args": {
      "External id": 62976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62976,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 62976, "pid": 5, "tid": 7, "ts": 1716454222344715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283592, "dur": 9,
    "args": {
      "External id": 62976,
      "cbid": 211, "correlation": 62976
    }
  },
  {
    "ph": "s", "id": 62976, "pid": 76337, "tid": -914061504, "ts": 1716454222283592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222344727, "dur": 19,
    "args": {
      "External id": 62996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 62996,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 62996, "pid": 5, "tid": 7, "ts": 1716454222344727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283670, "dur": 12,
    "args": {
      "External id": 62996,
      "cbid": 211, "correlation": 62996
    }
  },
  {
    "ph": "s", "id": 62996, "pid": 76337, "tid": -914061504, "ts": 1716454222283670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222344747, "dur": 4,
    "args": {
      "External id": 63008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63008,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 63008, "pid": 5, "tid": 7, "ts": 1716454222344747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283695, "dur": 7,
    "args": {
      "External id": 63008,
      "cbid": 211, "correlation": 63008
    }
  },
  {
    "ph": "s", "id": 63008, "pid": 76337, "tid": -914061504, "ts": 1716454222283695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222344753, "dur": 17,
    "args": {
      "External id": 63011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63011, "pid": 5, "tid": 7, "ts": 1716454222344753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283714, "dur": 6,
    "args": {
      "External id": 63011,
      "cbid": 211, "correlation": 63011
    }
  },
  {
    "ph": "s", "id": 63011, "pid": 76337, "tid": -914061504, "ts": 1716454222283714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283771, "dur": 0,
    "args": {
      "External id": 63022,
      "cbid": 317, "correlation": 63022
    }
  },
  {
    "ph": "f", "id": 63022, "pid": 76337, "tid": -914061504, "ts": 1716454222283771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283772, "dur": 0,
    "args": {
      "External id": 63023,
      "cbid": 203, "correlation": 63023
    }
  },
  {
    "ph": "f", "id": 63023, "pid": 76337, "tid": -914061504, "ts": 1716454222283772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283772, "dur": 0,
    "args": {
      "External id": 63024,
      "cbid": 205, "correlation": 63024
    }
  },
  {
    "ph": "f", "id": 63024, "pid": 76337, "tid": -914061504, "ts": 1716454222283772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222344772, "dur": 11,
    "args": {
      "External id": 63028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63028, "pid": 5, "tid": 7, "ts": 1716454222344772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283785, "dur": 12,
    "args": {
      "External id": 63028,
      "cbid": 211, "correlation": 63028
    }
  },
  {
    "ph": "s", "id": 63028, "pid": 76337, "tid": -914061504, "ts": 1716454222283785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222344784, "dur": 4,
    "args": {
      "External id": 63030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 63030, "pid": 5, "tid": 7, "ts": 1716454222344784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283801, "dur": 6,
    "args": {
      "External id": 63030,
      "cbid": 211, "correlation": 63030
    }
  },
  {
    "ph": "s", "id": 63030, "pid": 76337, "tid": -914061504, "ts": 1716454222283801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283810, "dur": 0,
    "args": {
      "External id": 63031,
      "cbid": 51, "correlation": 63031
    }
  },
  {
    "ph": "s", "id": 63031, "pid": 76337, "tid": -914061504, "ts": 1716454222283810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222344789, "dur": 103,
    "args": {
      "External id": 63032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63032,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 63032, "pid": 5, "tid": 7, "ts": 1716454222344789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283811, "dur": 5,
    "args": {
      "External id": 63032,
      "cbid": 211, "correlation": 63032
    }
  },
  {
    "ph": "s", "id": 63032, "pid": 76337, "tid": -914061504, "ts": 1716454222283811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222344894, "dur": 17,
    "args": {
      "External id": 63037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63037,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63037, "pid": 5, "tid": 7, "ts": 1716454222344894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283837, "dur": 8,
    "args": {
      "External id": 63037,
      "cbid": 211, "correlation": 63037
    }
  },
  {
    "ph": "s", "id": 63037, "pid": 76337, "tid": -914061504, "ts": 1716454222283837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222344912, "dur": 91,
    "args": {
      "External id": 63046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63046,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63046, "pid": 5, "tid": 7, "ts": 1716454222344912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283919, "dur": 14,
    "args": {
      "External id": 63046,
      "cbid": 211, "correlation": 63046
    }
  },
  {
    "ph": "s", "id": 63046, "pid": 76337, "tid": -914061504, "ts": 1716454222283919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222345004, "dur": 32,
    "args": {
      "External id": 63068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63068,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63068, "pid": 5, "tid": 7, "ts": 1716454222345004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222283990, "dur": 11,
    "args": {
      "External id": 63068,
      "cbid": 211, "correlation": 63068
    }
  },
  {
    "ph": "s", "id": 63068, "pid": 76337, "tid": -914061504, "ts": 1716454222283990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284083, "dur": 1,
    "args": {
      "External id": 63079,
      "cbid": 251, "correlation": 63079
    }
  },
  {
    "ph": "f", "id": 63079, "pid": 76337, "tid": -914061504, "ts": 1716454222284083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222345038, "dur": 180,
    "args": {
      "External id": 63080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63080,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63080, "pid": 5, "tid": 7, "ts": 1716454222345038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284089, "dur": 13,
    "args": {
      "External id": 63080,
      "cbid": 211, "correlation": 63080
    }
  },
  {
    "ph": "s", "id": 63080, "pid": 76337, "tid": -914061504, "ts": 1716454222284089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284159, "dur": 1,
    "args": {
      "External id": 63091,
      "cbid": 251, "correlation": 63091
    }
  },
  {
    "ph": "f", "id": 63091, "pid": 76337, "tid": -914061504, "ts": 1716454222284159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222345219, "dur": 176,
    "args": {
      "External id": 63092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63092,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63092, "pid": 5, "tid": 7, "ts": 1716454222345219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284163, "dur": 12,
    "args": {
      "External id": 63092,
      "cbid": 211, "correlation": 63092
    }
  },
  {
    "ph": "s", "id": 63092, "pid": 76337, "tid": -914061504, "ts": 1716454222284163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284227, "dur": 1,
    "args": {
      "External id": 63103,
      "cbid": 251, "correlation": 63103
    }
  },
  {
    "ph": "f", "id": 63103, "pid": 76337, "tid": -914061504, "ts": 1716454222284227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222345397, "dur": 172,
    "args": {
      "External id": 63104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63104,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63104, "pid": 5, "tid": 7, "ts": 1716454222345397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284231, "dur": 11,
    "args": {
      "External id": 63104,
      "cbid": 211, "correlation": 63104
    }
  },
  {
    "ph": "s", "id": 63104, "pid": 76337, "tid": -914061504, "ts": 1716454222284231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222345570, "dur": 361,
    "args": {
      "External id": 63129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63129,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63129, "pid": 5, "tid": 7, "ts": 1716454222345570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284313, "dur": 13,
    "args": {
      "External id": 63129,
      "cbid": 211, "correlation": 63129
    }
  },
  {
    "ph": "s", "id": 63129, "pid": 76337, "tid": -914061504, "ts": 1716454222284313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284411, "dur": 1,
    "args": {
      "External id": 63147,
      "cbid": 251, "correlation": 63147
    }
  },
  {
    "ph": "f", "id": 63147, "pid": 76337, "tid": -914061504, "ts": 1716454222284411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222345933, "dur": 184,
    "args": {
      "External id": 63149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63149,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63149, "pid": 5, "tid": 7, "ts": 1716454222345933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284417, "dur": 14,
    "args": {
      "External id": 63149,
      "cbid": 211, "correlation": 63149
    }
  },
  {
    "ph": "s", "id": 63149, "pid": 76337, "tid": -914061504, "ts": 1716454222284417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222346119, "dur": 19,
    "args": {
      "External id": 63157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63157,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63157, "pid": 5, "tid": 7, "ts": 1716454222346119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284491, "dur": 12,
    "args": {
      "External id": 63157,
      "cbid": 211, "correlation": 63157
    }
  },
  {
    "ph": "s", "id": 63157, "pid": 76337, "tid": -914061504, "ts": 1716454222284491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222346139, "dur": 28,
    "args": {
      "External id": 63165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63165,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63165, "pid": 5, "tid": 7, "ts": 1716454222346139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284530, "dur": 8,
    "args": {
      "External id": 63165,
      "cbid": 211, "correlation": 63165
    }
  },
  {
    "ph": "s", "id": 63165, "pid": 76337, "tid": -914061504, "ts": 1716454222284530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222346169, "dur": 19,
    "args": {
      "External id": 63176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63176, "pid": 5, "tid": 7, "ts": 1716454222346169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284602, "dur": 12,
    "args": {
      "External id": 63176,
      "cbid": 211, "correlation": 63176
    }
  },
  {
    "ph": "s", "id": 63176, "pid": 76337, "tid": -914061504, "ts": 1716454222284602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222346189, "dur": 17,
    "args": {
      "External id": 63198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63198,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63198, "pid": 5, "tid": 7, "ts": 1716454222346189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284633, "dur": 8,
    "args": {
      "External id": 63198,
      "cbid": 211, "correlation": 63198
    }
  },
  {
    "ph": "s", "id": 63198, "pid": 76337, "tid": -914061504, "ts": 1716454222284633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284717, "dur": 1,
    "args": {
      "External id": 63209,
      "cbid": 251, "correlation": 63209
    }
  },
  {
    "ph": "f", "id": 63209, "pid": 76337, "tid": -914061504, "ts": 1716454222284717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222346208, "dur": 96,
    "args": {
      "External id": 63210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63210,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 63210, "pid": 5, "tid": 7, "ts": 1716454222346208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284722, "dur": 13,
    "args": {
      "External id": 63210,
      "cbid": 211, "correlation": 63210
    }
  },
  {
    "ph": "s", "id": 63210, "pid": 76337, "tid": -914061504, "ts": 1716454222284722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284792, "dur": 1,
    "args": {
      "External id": 63221,
      "cbid": 251, "correlation": 63221
    }
  },
  {
    "ph": "f", "id": 63221, "pid": 76337, "tid": -914061504, "ts": 1716454222284792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284795, "dur": 0,
    "args": {
      "External id": 63222,
      "cbid": 251, "correlation": 63222
    }
  },
  {
    "ph": "f", "id": 63222, "pid": 76337, "tid": -914061504, "ts": 1716454222284795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222346306, "dur": 13,
    "args": {
      "External id": 63223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63223,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63223, "pid": 5, "tid": 7, "ts": 1716454222346306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284797, "dur": 11,
    "args": {
      "External id": 63223,
      "cbid": 211, "correlation": 63223
    }
  },
  {
    "ph": "s", "id": 63223, "pid": 76337, "tid": -914061504, "ts": 1716454222284797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222346321, "dur": 6,
    "args": {
      "External id": 63225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63225,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63225, "pid": 5, "tid": 7, "ts": 1716454222346321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284810, "dur": 9,
    "args": {
      "External id": 63225,
      "cbid": 211, "correlation": 63225
    }
  },
  {
    "ph": "s", "id": 63225, "pid": 76337, "tid": -914061504, "ts": 1716454222284810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284871, "dur": 1,
    "args": {
      "External id": 63236,
      "cbid": 251, "correlation": 63236
    }
  },
  {
    "ph": "f", "id": 63236, "pid": 76337, "tid": -914061504, "ts": 1716454222284871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284874, "dur": 0,
    "args": {
      "External id": 63237,
      "cbid": 251, "correlation": 63237
    }
  },
  {
    "ph": "f", "id": 63237, "pid": 76337, "tid": -914061504, "ts": 1716454222284874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222346328, "dur": 9,
    "args": {
      "External id": 63238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63238,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63238, "pid": 5, "tid": 7, "ts": 1716454222346328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284875, "dur": 12,
    "args": {
      "External id": 63238,
      "cbid": 211, "correlation": 63238
    }
  },
  {
    "ph": "s", "id": 63238, "pid": 76337, "tid": -914061504, "ts": 1716454222284875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222346339, "dur": 4,
    "args": {
      "External id": 63240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63240,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63240, "pid": 5, "tid": 7, "ts": 1716454222346339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284889, "dur": 5,
    "args": {
      "External id": 63240,
      "cbid": 211, "correlation": 63240
    }
  },
  {
    "ph": "s", "id": 63240, "pid": 76337, "tid": -914061504, "ts": 1716454222284889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222346344, "dur": 60,
    "args": {
      "External id": 63265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63265,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63265, "pid": 5, "tid": 7, "ts": 1716454222346344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222284966, "dur": 21,
    "args": {
      "External id": 63265,
      "cbid": 211, "correlation": 63265
    }
  },
  {
    "ph": "s", "id": 63265, "pid": 76337, "tid": -914061504, "ts": 1716454222284966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285074, "dur": 1,
    "args": {
      "External id": 63283,
      "cbid": 251, "correlation": 63283
    }
  },
  {
    "ph": "f", "id": 63283, "pid": 76337, "tid": -914061504, "ts": 1716454222285074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222346406, "dur": 99,
    "args": {
      "External id": 63285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63285,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 63285, "pid": 5, "tid": 7, "ts": 1716454222346406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285080, "dur": 13,
    "args": {
      "External id": 63285,
      "cbid": 211, "correlation": 63285
    }
  },
  {
    "ph": "s", "id": 63285, "pid": 76337, "tid": -914061504, "ts": 1716454222285080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222346507, "dur": 10,
    "args": {
      "External id": 63293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63293,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63293, "pid": 5, "tid": 7, "ts": 1716454222346507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285149, "dur": 13,
    "args": {
      "External id": 63293,
      "cbid": 211, "correlation": 63293
    }
  },
  {
    "ph": "s", "id": 63293, "pid": 76337, "tid": -914061504, "ts": 1716454222285149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222346519, "dur": 23,
    "args": {
      "External id": 63301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63301, "pid": 5, "tid": 7, "ts": 1716454222346519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285191, "dur": 12,
    "args": {
      "External id": 63301,
      "cbid": 211, "correlation": 63301
    }
  },
  {
    "ph": "s", "id": 63301, "pid": 76337, "tid": -914061504, "ts": 1716454222285191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222346543, "dur": 19,
    "args": {
      "External id": 63323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63323,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63323, "pid": 5, "tid": 7, "ts": 1716454222346543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285245, "dur": 11,
    "args": {
      "External id": 63323,
      "cbid": 211, "correlation": 63323
    }
  },
  {
    "ph": "s", "id": 63323, "pid": 76337, "tid": -914061504, "ts": 1716454222285245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285333, "dur": 1,
    "args": {
      "External id": 63339,
      "cbid": 251, "correlation": 63339
    }
  },
  {
    "ph": "f", "id": 63339, "pid": 76337, "tid": -914061504, "ts": 1716454222285333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285338, "dur": 0,
    "args": {
      "External id": 63341,
      "cbid": 251, "correlation": 63341
    }
  },
  {
    "ph": "f", "id": 63341, "pid": 76337, "tid": -914061504, "ts": 1716454222285338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222346563, "dur": 525,
    "args": {
      "External id": 63342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63342,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63342, "pid": 5, "tid": 7, "ts": 1716454222346563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285340, "dur": 12,
    "args": {
      "External id": 63342,
      "cbid": 211, "correlation": 63342
    }
  },
  {
    "ph": "s", "id": 63342, "pid": 76337, "tid": -914061504, "ts": 1716454222285340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222347089, "dur": 70,
    "args": {
      "External id": 63350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63350, "pid": 5, "tid": 7, "ts": 1716454222347089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285405, "dur": 13,
    "args": {
      "External id": 63350,
      "cbid": 211, "correlation": 63350
    }
  },
  {
    "ph": "s", "id": 63350, "pid": 76337, "tid": -914061504, "ts": 1716454222285405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222347161, "dur": 70,
    "args": {
      "External id": 63358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63358, "pid": 5, "tid": 7, "ts": 1716454222347161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285435, "dur": 8,
    "args": {
      "External id": 63358,
      "cbid": 211, "correlation": 63358
    }
  },
  {
    "ph": "s", "id": 63358, "pid": 76337, "tid": -914061504, "ts": 1716454222285435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285514, "dur": 1,
    "args": {
      "External id": 63374,
      "cbid": 251, "correlation": 63374
    }
  },
  {
    "ph": "f", "id": 63374, "pid": 76337, "tid": -914061504, "ts": 1716454222285514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222347233, "dur": 1,
    "args": {
      "External id": 63376,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63376,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14990630855715179
    }
  },
  {
    "ph": "f", "id": 63376, "pid": 5, "tid": 7, "ts": 1716454222347233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285519, "dur": 9,
    "args": {
      "External id": 63376,
      "cbid": 51, "correlation": 63376
    }
  },
  {
    "ph": "s", "id": 63376, "pid": 76337, "tid": -914061504, "ts": 1716454222285519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222347237, "dur": 294,
    "args": {
      "External id": 63377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63377,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 63377, "pid": 5, "tid": 7, "ts": 1716454222347237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285530, "dur": 14,
    "args": {
      "External id": 63377,
      "cbid": 211, "correlation": 63377
    }
  },
  {
    "ph": "s", "id": 63377, "pid": 76337, "tid": -914061504, "ts": 1716454222285530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222347532, "dur": 14,
    "args": {
      "External id": 63385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63385,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63385, "pid": 5, "tid": 7, "ts": 1716454222347532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285576, "dur": 10,
    "args": {
      "External id": 63385,
      "cbid": 211, "correlation": 63385
    }
  },
  {
    "ph": "s", "id": 63385, "pid": 76337, "tid": -914061504, "ts": 1716454222285576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222347547, "dur": 41,
    "args": {
      "External id": 63396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63396, "pid": 5, "tid": 7, "ts": 1716454222347547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285645, "dur": 12,
    "args": {
      "External id": 63396,
      "cbid": 211, "correlation": 63396
    }
  },
  {
    "ph": "s", "id": 63396, "pid": 76337, "tid": -914061504, "ts": 1716454222285645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285708, "dur": 0,
    "args": {
      "External id": 63408,
      "cbid": 317, "correlation": 63408
    }
  },
  {
    "ph": "f", "id": 63408, "pid": 76337, "tid": -914061504, "ts": 1716454222285708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285709, "dur": 0,
    "args": {
      "External id": 63409,
      "cbid": 203, "correlation": 63409
    }
  },
  {
    "ph": "f", "id": 63409, "pid": 76337, "tid": -914061504, "ts": 1716454222285709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285710, "dur": 0,
    "args": {
      "External id": 63410,
      "cbid": 205, "correlation": 63410
    }
  },
  {
    "ph": "f", "id": 63410, "pid": 76337, "tid": -914061504, "ts": 1716454222285710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222347589, "dur": 14,
    "args": {
      "External id": 63414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63414,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63414, "pid": 5, "tid": 7, "ts": 1716454222347589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285726, "dur": 12,
    "args": {
      "External id": 63414,
      "cbid": 211, "correlation": 63414
    }
  },
  {
    "ph": "s", "id": 63414, "pid": 76337, "tid": -914061504, "ts": 1716454222285726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222347605, "dur": 4,
    "args": {
      "External id": 63416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 63416, "pid": 5, "tid": 7, "ts": 1716454222347605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285742, "dur": 6,
    "args": {
      "External id": 63416,
      "cbid": 211, "correlation": 63416
    }
  },
  {
    "ph": "s", "id": 63416, "pid": 76337, "tid": -914061504, "ts": 1716454222285742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285751, "dur": 0,
    "args": {
      "External id": 63417,
      "cbid": 51, "correlation": 63417
    }
  },
  {
    "ph": "s", "id": 63417, "pid": 76337, "tid": -914061504, "ts": 1716454222285751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222347610, "dur": 107,
    "args": {
      "External id": 63418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63418,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 63418, "pid": 5, "tid": 7, "ts": 1716454222347610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285751, "dur": 5,
    "args": {
      "External id": 63418,
      "cbid": 211, "correlation": 63418
    }
  },
  {
    "ph": "s", "id": 63418, "pid": 76337, "tid": -914061504, "ts": 1716454222285751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222347718, "dur": 19,
    "args": {
      "External id": 63423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63423, "pid": 5, "tid": 7, "ts": 1716454222347718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285779, "dur": 9,
    "args": {
      "External id": 63423,
      "cbid": 211, "correlation": 63423
    }
  },
  {
    "ph": "s", "id": 63423, "pid": 76337, "tid": -914061504, "ts": 1716454222285779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222347738, "dur": 13,
    "args": {
      "External id": 63431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63431,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63431, "pid": 5, "tid": 7, "ts": 1716454222347738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285810, "dur": 8,
    "args": {
      "External id": 63431,
      "cbid": 211, "correlation": 63431
    }
  },
  {
    "ph": "s", "id": 63431, "pid": 76337, "tid": -914061504, "ts": 1716454222285810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285879, "dur": 0,
    "args": {
      "External id": 63441,
      "cbid": 317, "correlation": 63441
    }
  },
  {
    "ph": "f", "id": 63441, "pid": 76337, "tid": -914061504, "ts": 1716454222285879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285879, "dur": 0,
    "args": {
      "External id": 63442,
      "cbid": 203, "correlation": 63442
    }
  },
  {
    "ph": "f", "id": 63442, "pid": 76337, "tid": -914061504, "ts": 1716454222285879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285880, "dur": 0,
    "args": {
      "External id": 63443,
      "cbid": 205, "correlation": 63443
    }
  },
  {
    "ph": "f", "id": 63443, "pid": 76337, "tid": -914061504, "ts": 1716454222285880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222347753, "dur": 12,
    "args": {
      "External id": 63447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63447,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63447, "pid": 5, "tid": 7, "ts": 1716454222347753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285904, "dur": 14,
    "args": {
      "External id": 63447,
      "cbid": 211, "correlation": 63447
    }
  },
  {
    "ph": "s", "id": 63447, "pid": 76337, "tid": -914061504, "ts": 1716454222285904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222347766, "dur": 177,
    "args": {
      "External id": 63449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63449, "pid": 5, "tid": 7, "ts": 1716454222347766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285920, "dur": 6,
    "args": {
      "External id": 63449,
      "cbid": 211, "correlation": 63449
    }
  },
  {
    "ph": "s", "id": 63449, "pid": 76337, "tid": -914061504, "ts": 1716454222285920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222347945, "dur": 1,
    "args": {
      "External id": 63451,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63451,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5454545454545454
    }
  },
  {
    "ph": "f", "id": 63451, "pid": 5, "tid": 7, "ts": 1716454222347945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285933, "dur": 7,
    "args": {
      "External id": 63451,
      "cbid": 51, "correlation": 63451
    }
  },
  {
    "ph": "s", "id": 63451, "pid": 76337, "tid": -914061504, "ts": 1716454222285933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222347949, "dur": 218,
    "args": {
      "External id": 63452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63452,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 63452, "pid": 5, "tid": 7, "ts": 1716454222347949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285942, "dur": 8,
    "args": {
      "External id": 63452,
      "cbid": 211, "correlation": 63452
    }
  },
  {
    "ph": "s", "id": 63452, "pid": 76337, "tid": -914061504, "ts": 1716454222285942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348169, "dur": 7,
    "args": {
      "External id": 63454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63454, "pid": 5, "tid": 7, "ts": 1716454222348169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285954, "dur": 5,
    "args": {
      "External id": 63454,
      "cbid": 211, "correlation": 63454
    }
  },
  {
    "ph": "s", "id": 63454, "pid": 76337, "tid": -914061504, "ts": 1716454222285954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222348176, "dur": 7,
    "args": {
      "External id": 63460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63460, "pid": 5, "tid": 7, "ts": 1716454222348176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222285991, "dur": 9,
    "args": {
      "External id": 63460,
      "cbid": 211, "correlation": 63460
    }
  },
  {
    "ph": "s", "id": 63460, "pid": 76337, "tid": -914061504, "ts": 1716454222285991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348185, "dur": 12,
    "args": {
      "External id": 63480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63480,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 63480, "pid": 5, "tid": 7, "ts": 1716454222348185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286085, "dur": 12,
    "args": {
      "External id": 63480,
      "cbid": 211, "correlation": 63480
    }
  },
  {
    "ph": "s", "id": 63480, "pid": 76337, "tid": -914061504, "ts": 1716454222286085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222348198, "dur": 5,
    "args": {
      "External id": 63492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63492,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 63492, "pid": 5, "tid": 7, "ts": 1716454222348198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286108, "dur": 7,
    "args": {
      "External id": 63492,
      "cbid": 211, "correlation": 63492
    }
  },
  {
    "ph": "s", "id": 63492, "pid": 76337, "tid": -914061504, "ts": 1716454222286108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222348204, "dur": 9,
    "args": {
      "External id": 63495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63495,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63495, "pid": 5, "tid": 7, "ts": 1716454222348204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286127, "dur": 7,
    "args": {
      "External id": 63495,
      "cbid": 211, "correlation": 63495
    }
  },
  {
    "ph": "s", "id": 63495, "pid": 76337, "tid": -914061504, "ts": 1716454222286127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222348215, "dur": 6,
    "args": {
      "External id": 63504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63504,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63504, "pid": 5, "tid": 7, "ts": 1716454222348215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286167, "dur": 10,
    "args": {
      "External id": 63504,
      "cbid": 211, "correlation": 63504
    }
  },
  {
    "ph": "s", "id": 63504, "pid": 76337, "tid": -914061504, "ts": 1716454222286167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286223, "dur": 0,
    "args": {
      "External id": 63514,
      "cbid": 317, "correlation": 63514
    }
  },
  {
    "ph": "f", "id": 63514, "pid": 76337, "tid": -914061504, "ts": 1716454222286223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286224, "dur": 0,
    "args": {
      "External id": 63515,
      "cbid": 203, "correlation": 63515
    }
  },
  {
    "ph": "f", "id": 63515, "pid": 76337, "tid": -914061504, "ts": 1716454222286224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286225, "dur": 0,
    "args": {
      "External id": 63516,
      "cbid": 205, "correlation": 63516
    }
  },
  {
    "ph": "f", "id": 63516, "pid": 76337, "tid": -914061504, "ts": 1716454222286225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348222, "dur": 6,
    "args": {
      "External id": 63520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63520,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63520, "pid": 5, "tid": 7, "ts": 1716454222348222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286242, "dur": 12,
    "args": {
      "External id": 63520,
      "cbid": 211, "correlation": 63520
    }
  },
  {
    "ph": "s", "id": 63520, "pid": 76337, "tid": -914061504, "ts": 1716454222286242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348229, "dur": 176,
    "args": {
      "External id": 63522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63522,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63522, "pid": 5, "tid": 7, "ts": 1716454222348229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286256, "dur": 5,
    "args": {
      "External id": 63522,
      "cbid": 211, "correlation": 63522
    }
  },
  {
    "ph": "s", "id": 63522, "pid": 76337, "tid": -914061504, "ts": 1716454222286256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222348408, "dur": 1,
    "args": {
      "External id": 63524,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63524,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 63524, "pid": 5, "tid": 7, "ts": 1716454222348408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286267, "dur": 6,
    "args": {
      "External id": 63524,
      "cbid": 51, "correlation": 63524
    }
  },
  {
    "ph": "s", "id": 63524, "pid": 76337, "tid": -914061504, "ts": 1716454222286267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222348412, "dur": 296,
    "args": {
      "External id": 63525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63525,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63525, "pid": 5, "tid": 7, "ts": 1716454222348412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286275, "dur": 6,
    "args": {
      "External id": 63525,
      "cbid": 211, "correlation": 63525
    }
  },
  {
    "ph": "s", "id": 63525, "pid": 76337, "tid": -914061504, "ts": 1716454222286275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348709, "dur": 6,
    "args": {
      "External id": 63527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63527,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63527, "pid": 5, "tid": 7, "ts": 1716454222348709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286285, "dur": 5,
    "args": {
      "External id": 63527,
      "cbid": 211, "correlation": 63527
    }
  },
  {
    "ph": "s", "id": 63527, "pid": 76337, "tid": -914061504, "ts": 1716454222286285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222348717, "dur": 7,
    "args": {
      "External id": 63533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63533, "pid": 5, "tid": 7, "ts": 1716454222348717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286313, "dur": 8,
    "args": {
      "External id": 63533,
      "cbid": 211, "correlation": 63533
    }
  },
  {
    "ph": "s", "id": 63533, "pid": 76337, "tid": -914061504, "ts": 1716454222286313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222348725, "dur": 3,
    "args": {
      "External id": 63541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63541,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 63541, "pid": 5, "tid": 7, "ts": 1716454222348725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286358, "dur": 9,
    "args": {
      "External id": 63541,
      "cbid": 211, "correlation": 63541
    }
  },
  {
    "ph": "s", "id": 63541, "pid": 76337, "tid": -914061504, "ts": 1716454222286358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286423, "dur": 1,
    "args": {
      "External id": 63557,
      "cbid": 251, "correlation": 63557
    }
  },
  {
    "ph": "f", "id": 63557, "pid": 76337, "tid": -914061504, "ts": 1716454222286423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286428, "dur": 0,
    "args": {
      "External id": 63559,
      "cbid": 251, "correlation": 63559
    }
  },
  {
    "ph": "f", "id": 63559, "pid": 76337, "tid": -914061504, "ts": 1716454222286428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222348730, "dur": 14,
    "args": {
      "External id": 63560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63560,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63560, "pid": 5, "tid": 7, "ts": 1716454222348730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286430, "dur": 11,
    "args": {
      "External id": 63560,
      "cbid": 211, "correlation": 63560
    }
  },
  {
    "ph": "s", "id": 63560, "pid": 76337, "tid": -914061504, "ts": 1716454222286430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222348745, "dur": 5,
    "args": {
      "External id": 63562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63562,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63562, "pid": 5, "tid": 7, "ts": 1716454222348745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286442, "dur": 5,
    "args": {
      "External id": 63562,
      "cbid": 211, "correlation": 63562
    }
  },
  {
    "ph": "s", "id": 63562, "pid": 76337, "tid": -914061504, "ts": 1716454222286442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222348752, "dur": 6,
    "args": {
      "External id": 63572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63572,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63572, "pid": 5, "tid": 7, "ts": 1716454222348752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286499, "dur": 15,
    "args": {
      "External id": 63572,
      "cbid": 211, "correlation": 63572
    }
  },
  {
    "ph": "s", "id": 63572, "pid": 76337, "tid": -914061504, "ts": 1716454222286499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348760, "dur": 11,
    "args": {
      "External id": 63592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63592,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 63592, "pid": 5, "tid": 7, "ts": 1716454222348760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286571, "dur": 11,
    "args": {
      "External id": 63592,
      "cbid": 211, "correlation": 63592
    }
  },
  {
    "ph": "s", "id": 63592, "pid": 76337, "tid": -914061504, "ts": 1716454222286571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222348772, "dur": 4,
    "args": {
      "External id": 63604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63604,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 63604, "pid": 5, "tid": 7, "ts": 1716454222348772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286592, "dur": 6,
    "args": {
      "External id": 63604,
      "cbid": 211, "correlation": 63604
    }
  },
  {
    "ph": "s", "id": 63604, "pid": 76337, "tid": -914061504, "ts": 1716454222286592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222348777, "dur": 8,
    "args": {
      "External id": 63607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63607, "pid": 5, "tid": 7, "ts": 1716454222348777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286609, "dur": 6,
    "args": {
      "External id": 63607,
      "cbid": 211, "correlation": 63607
    }
  },
  {
    "ph": "s", "id": 63607, "pid": 76337, "tid": -914061504, "ts": 1716454222286609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222348786, "dur": 5,
    "args": {
      "External id": 63616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63616,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63616, "pid": 5, "tid": 7, "ts": 1716454222348786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286650, "dur": 10,
    "args": {
      "External id": 63616,
      "cbid": 211, "correlation": 63616
    }
  },
  {
    "ph": "s", "id": 63616, "pid": 76337, "tid": -914061504, "ts": 1716454222286650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286713, "dur": 0,
    "args": {
      "External id": 63626,
      "cbid": 317, "correlation": 63626
    }
  },
  {
    "ph": "f", "id": 63626, "pid": 76337, "tid": -914061504, "ts": 1716454222286713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286714, "dur": 0,
    "args": {
      "External id": 63627,
      "cbid": 203, "correlation": 63627
    }
  },
  {
    "ph": "f", "id": 63627, "pid": 76337, "tid": -914061504, "ts": 1716454222286714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286714, "dur": 0,
    "args": {
      "External id": 63628,
      "cbid": 205, "correlation": 63628
    }
  },
  {
    "ph": "f", "id": 63628, "pid": 76337, "tid": -914061504, "ts": 1716454222286714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348793, "dur": 6,
    "args": {
      "External id": 63632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63632, "pid": 5, "tid": 7, "ts": 1716454222348793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286728, "dur": 12,
    "args": {
      "External id": 63632,
      "cbid": 211, "correlation": 63632
    }
  },
  {
    "ph": "s", "id": 63632, "pid": 76337, "tid": -914061504, "ts": 1716454222286728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222348800, "dur": 177,
    "args": {
      "External id": 63634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63634,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63634, "pid": 5, "tid": 7, "ts": 1716454222348800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286743, "dur": 5,
    "args": {
      "External id": 63634,
      "cbid": 211, "correlation": 63634
    }
  },
  {
    "ph": "s", "id": 63634, "pid": 76337, "tid": -914061504, "ts": 1716454222286743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222348979, "dur": 1,
    "args": {
      "External id": 63636,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63636,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 63636, "pid": 5, "tid": 7, "ts": 1716454222348979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286754, "dur": 6,
    "args": {
      "External id": 63636,
      "cbid": 51, "correlation": 63636
    }
  },
  {
    "ph": "s", "id": 63636, "pid": 76337, "tid": -914061504, "ts": 1716454222286754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222348983, "dur": 285,
    "args": {
      "External id": 63637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63637,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63637, "pid": 5, "tid": 7, "ts": 1716454222348983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286761, "dur": 6,
    "args": {
      "External id": 63637,
      "cbid": 211, "correlation": 63637
    }
  },
  {
    "ph": "s", "id": 63637, "pid": 76337, "tid": -914061504, "ts": 1716454222286761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349269, "dur": 6,
    "args": {
      "External id": 63639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63639,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63639, "pid": 5, "tid": 7, "ts": 1716454222349269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286770, "dur": 9,
    "args": {
      "External id": 63639,
      "cbid": 211, "correlation": 63639
    }
  },
  {
    "ph": "s", "id": 63639, "pid": 76337, "tid": -914061504, "ts": 1716454222286770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222349277, "dur": 7,
    "args": {
      "External id": 63645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63645,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63645, "pid": 5, "tid": 7, "ts": 1716454222349277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286802, "dur": 9,
    "args": {
      "External id": 63645,
      "cbid": 211, "correlation": 63645
    }
  },
  {
    "ph": "s", "id": 63645, "pid": 76337, "tid": -914061504, "ts": 1716454222286802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222349285, "dur": 5,
    "args": {
      "External id": 63653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63653,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63653, "pid": 5, "tid": 7, "ts": 1716454222349285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286835, "dur": 9,
    "args": {
      "External id": 63653,
      "cbid": 211, "correlation": 63653
    }
  },
  {
    "ph": "s", "id": 63653, "pid": 76337, "tid": -914061504, "ts": 1716454222286835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222349292, "dur": 5,
    "args": {
      "External id": 63661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63661,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63661, "pid": 5, "tid": 7, "ts": 1716454222349292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286865, "dur": 8,
    "args": {
      "External id": 63661,
      "cbid": 211, "correlation": 63661
    }
  },
  {
    "ph": "s", "id": 63661, "pid": 76337, "tid": -914061504, "ts": 1716454222286865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349298, "dur": 10,
    "args": {
      "External id": 63681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63681,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 63681, "pid": 5, "tid": 7, "ts": 1716454222349298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286938, "dur": 12,
    "args": {
      "External id": 63681,
      "cbid": 211, "correlation": 63681
    }
  },
  {
    "ph": "s", "id": 63681, "pid": 76337, "tid": -914061504, "ts": 1716454222286938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222349310, "dur": 4,
    "args": {
      "External id": 63693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63693,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 63693, "pid": 5, "tid": 7, "ts": 1716454222349310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286960, "dur": 6,
    "args": {
      "External id": 63693,
      "cbid": 211, "correlation": 63693
    }
  },
  {
    "ph": "s", "id": 63693, "pid": 76337, "tid": -914061504, "ts": 1716454222286960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222349315, "dur": 7,
    "args": {
      "External id": 63696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63696, "pid": 5, "tid": 7, "ts": 1716454222349315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222286986, "dur": 9,
    "args": {
      "External id": 63696,
      "cbid": 211, "correlation": 63696
    }
  },
  {
    "ph": "s", "id": 63696, "pid": 76337, "tid": -914061504, "ts": 1716454222286986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222349323, "dur": 5,
    "args": {
      "External id": 63705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63705,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63705, "pid": 5, "tid": 7, "ts": 1716454222349323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287027, "dur": 9,
    "args": {
      "External id": 63705,
      "cbid": 211, "correlation": 63705
    }
  },
  {
    "ph": "s", "id": 63705, "pid": 76337, "tid": -914061504, "ts": 1716454222287027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287082, "dur": 0,
    "args": {
      "External id": 63715,
      "cbid": 317, "correlation": 63715
    }
  },
  {
    "ph": "f", "id": 63715, "pid": 76337, "tid": -914061504, "ts": 1716454222287082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287083, "dur": 0,
    "args": {
      "External id": 63716,
      "cbid": 203, "correlation": 63716
    }
  },
  {
    "ph": "f", "id": 63716, "pid": 76337, "tid": -914061504, "ts": 1716454222287083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287084, "dur": 0,
    "args": {
      "External id": 63717,
      "cbid": 205, "correlation": 63717
    }
  },
  {
    "ph": "f", "id": 63717, "pid": 76337, "tid": -914061504, "ts": 1716454222287084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349330, "dur": 5,
    "args": {
      "External id": 63721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63721,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63721, "pid": 5, "tid": 7, "ts": 1716454222349330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287097, "dur": 12,
    "args": {
      "External id": 63721,
      "cbid": 211, "correlation": 63721
    }
  },
  {
    "ph": "s", "id": 63721, "pid": 76337, "tid": -914061504, "ts": 1716454222287097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349337, "dur": 177,
    "args": {
      "External id": 63723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63723,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63723, "pid": 5, "tid": 7, "ts": 1716454222349337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287112, "dur": 5,
    "args": {
      "External id": 63723,
      "cbid": 211, "correlation": 63723
    }
  },
  {
    "ph": "s", "id": 63723, "pid": 76337, "tid": -914061504, "ts": 1716454222287112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222349515, "dur": 1,
    "args": {
      "External id": 63725,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63725,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 63725, "pid": 5, "tid": 7, "ts": 1716454222349515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287122, "dur": 6,
    "args": {
      "External id": 63725,
      "cbid": 51, "correlation": 63725
    }
  },
  {
    "ph": "s", "id": 63725, "pid": 76337, "tid": -914061504, "ts": 1716454222287122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222349519, "dur": 284,
    "args": {
      "External id": 63726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63726,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63726, "pid": 5, "tid": 7, "ts": 1716454222349519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287130, "dur": 6,
    "args": {
      "External id": 63726,
      "cbid": 211, "correlation": 63726
    }
  },
  {
    "ph": "s", "id": 63726, "pid": 76337, "tid": -914061504, "ts": 1716454222287130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349805, "dur": 6,
    "args": {
      "External id": 63728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63728,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63728, "pid": 5, "tid": 7, "ts": 1716454222349805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287140, "dur": 5,
    "args": {
      "External id": 63728,
      "cbid": 211, "correlation": 63728
    }
  },
  {
    "ph": "s", "id": 63728, "pid": 76337, "tid": -914061504, "ts": 1716454222287140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222349813, "dur": 7,
    "args": {
      "External id": 63734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63734,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63734, "pid": 5, "tid": 7, "ts": 1716454222349813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287168, "dur": 9,
    "args": {
      "External id": 63734,
      "cbid": 211, "correlation": 63734
    }
  },
  {
    "ph": "s", "id": 63734, "pid": 76337, "tid": -914061504, "ts": 1716454222287168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222349821, "dur": 3,
    "args": {
      "External id": 63742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63742,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 63742, "pid": 5, "tid": 7, "ts": 1716454222349821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287212, "dur": 10,
    "args": {
      "External id": 63742,
      "cbid": 211, "correlation": 63742
    }
  },
  {
    "ph": "s", "id": 63742, "pid": 76337, "tid": -914061504, "ts": 1716454222287212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287274, "dur": 1,
    "args": {
      "External id": 63758,
      "cbid": 251, "correlation": 63758
    }
  },
  {
    "ph": "f", "id": 63758, "pid": 76337, "tid": -914061504, "ts": 1716454222287274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287279, "dur": 0,
    "args": {
      "External id": 63760,
      "cbid": 251, "correlation": 63760
    }
  },
  {
    "ph": "f", "id": 63760, "pid": 76337, "tid": -914061504, "ts": 1716454222287279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222349825, "dur": 11,
    "args": {
      "External id": 63761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63761,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63761, "pid": 5, "tid": 7, "ts": 1716454222349825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287281, "dur": 11,
    "args": {
      "External id": 63761,
      "cbid": 211, "correlation": 63761
    }
  },
  {
    "ph": "s", "id": 63761, "pid": 76337, "tid": -914061504, "ts": 1716454222287281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222349837, "dur": 4,
    "args": {
      "External id": 63763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63763,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63763, "pid": 5, "tid": 7, "ts": 1716454222349837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287294, "dur": 5,
    "args": {
      "External id": 63763,
      "cbid": 211, "correlation": 63763
    }
  },
  {
    "ph": "s", "id": 63763, "pid": 76337, "tid": -914061504, "ts": 1716454222287294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222349843, "dur": 6,
    "args": {
      "External id": 63773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63773,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63773, "pid": 5, "tid": 7, "ts": 1716454222349843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287354, "dur": 13,
    "args": {
      "External id": 63773,
      "cbid": 211, "correlation": 63773
    }
  },
  {
    "ph": "s", "id": 63773, "pid": 76337, "tid": -914061504, "ts": 1716454222287354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349850, "dur": 11,
    "args": {
      "External id": 63793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63793,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 63793, "pid": 5, "tid": 7, "ts": 1716454222349850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287421, "dur": 11,
    "args": {
      "External id": 63793,
      "cbid": 211, "correlation": 63793
    }
  },
  {
    "ph": "s", "id": 63793, "pid": 76337, "tid": -914061504, "ts": 1716454222287421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222349862, "dur": 4,
    "args": {
      "External id": 63805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63805,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 63805, "pid": 5, "tid": 7, "ts": 1716454222349862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287441, "dur": 6,
    "args": {
      "External id": 63805,
      "cbid": 211, "correlation": 63805
    }
  },
  {
    "ph": "s", "id": 63805, "pid": 76337, "tid": -914061504, "ts": 1716454222287441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222349868, "dur": 7,
    "args": {
      "External id": 63808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63808, "pid": 5, "tid": 7, "ts": 1716454222349868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287460, "dur": 6,
    "args": {
      "External id": 63808,
      "cbid": 211, "correlation": 63808
    }
  },
  {
    "ph": "s", "id": 63808, "pid": 76337, "tid": -914061504, "ts": 1716454222287460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222349876, "dur": 5,
    "args": {
      "External id": 63817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63817,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63817, "pid": 5, "tid": 7, "ts": 1716454222349876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287499, "dur": 10,
    "args": {
      "External id": 63817,
      "cbid": 211, "correlation": 63817
    }
  },
  {
    "ph": "s", "id": 63817, "pid": 76337, "tid": -914061504, "ts": 1716454222287499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287562, "dur": 0,
    "args": {
      "External id": 63827,
      "cbid": 317, "correlation": 63827
    }
  },
  {
    "ph": "f", "id": 63827, "pid": 76337, "tid": -914061504, "ts": 1716454222287562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287563, "dur": 0,
    "args": {
      "External id": 63828,
      "cbid": 203, "correlation": 63828
    }
  },
  {
    "ph": "f", "id": 63828, "pid": 76337, "tid": -914061504, "ts": 1716454222287563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287563, "dur": 0,
    "args": {
      "External id": 63829,
      "cbid": 205, "correlation": 63829
    }
  },
  {
    "ph": "f", "id": 63829, "pid": 76337, "tid": -914061504, "ts": 1716454222287563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349883, "dur": 6,
    "args": {
      "External id": 63833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63833,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63833, "pid": 5, "tid": 7, "ts": 1716454222349883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287577, "dur": 12,
    "args": {
      "External id": 63833,
      "cbid": 211, "correlation": 63833
    }
  },
  {
    "ph": "s", "id": 63833, "pid": 76337, "tid": -914061504, "ts": 1716454222287577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222349890, "dur": 177,
    "args": {
      "External id": 63835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63835,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63835, "pid": 5, "tid": 7, "ts": 1716454222349890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287591, "dur": 5,
    "args": {
      "External id": 63835,
      "cbid": 211, "correlation": 63835
    }
  },
  {
    "ph": "s", "id": 63835, "pid": 76337, "tid": -914061504, "ts": 1716454222287591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222350069, "dur": 1,
    "args": {
      "External id": 63837,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63837,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 63837, "pid": 5, "tid": 7, "ts": 1716454222350069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287601, "dur": 6,
    "args": {
      "External id": 63837,
      "cbid": 51, "correlation": 63837
    }
  },
  {
    "ph": "s", "id": 63837, "pid": 76337, "tid": -914061504, "ts": 1716454222287601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222350073, "dur": 285,
    "args": {
      "External id": 63838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63838,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63838, "pid": 5, "tid": 7, "ts": 1716454222350073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287609, "dur": 10,
    "args": {
      "External id": 63838,
      "cbid": 211, "correlation": 63838
    }
  },
  {
    "ph": "s", "id": 63838, "pid": 76337, "tid": -914061504, "ts": 1716454222287609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350360, "dur": 6,
    "args": {
      "External id": 63840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63840,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63840, "pid": 5, "tid": 7, "ts": 1716454222350360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287622, "dur": 5,
    "args": {
      "External id": 63840,
      "cbid": 211, "correlation": 63840
    }
  },
  {
    "ph": "s", "id": 63840, "pid": 76337, "tid": -914061504, "ts": 1716454222287622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222350368, "dur": 7,
    "args": {
      "External id": 63846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63846, "pid": 5, "tid": 7, "ts": 1716454222350368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287651, "dur": 8,
    "args": {
      "External id": 63846,
      "cbid": 211, "correlation": 63846
    }
  },
  {
    "ph": "s", "id": 63846, "pid": 76337, "tid": -914061504, "ts": 1716454222287651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222350375, "dur": 5,
    "args": {
      "External id": 63854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63854,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63854, "pid": 5, "tid": 7, "ts": 1716454222350375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287684, "dur": 8,
    "args": {
      "External id": 63854,
      "cbid": 211, "correlation": 63854
    }
  },
  {
    "ph": "s", "id": 63854, "pid": 76337, "tid": -914061504, "ts": 1716454222287684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222350382, "dur": 5,
    "args": {
      "External id": 63862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63862,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63862, "pid": 5, "tid": 7, "ts": 1716454222350382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287713, "dur": 9,
    "args": {
      "External id": 63862,
      "cbid": 211, "correlation": 63862
    }
  },
  {
    "ph": "s", "id": 63862, "pid": 76337, "tid": -914061504, "ts": 1716454222287713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350388, "dur": 11,
    "args": {
      "External id": 63882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63882,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 63882, "pid": 5, "tid": 7, "ts": 1716454222350388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287817, "dur": 14,
    "args": {
      "External id": 63882,
      "cbid": 211, "correlation": 63882
    }
  },
  {
    "ph": "s", "id": 63882, "pid": 76337, "tid": -914061504, "ts": 1716454222287817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222350401, "dur": 4,
    "args": {
      "External id": 63894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63894,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 63894, "pid": 5, "tid": 7, "ts": 1716454222350401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287841, "dur": 6,
    "args": {
      "External id": 63894,
      "cbid": 211, "correlation": 63894
    }
  },
  {
    "ph": "s", "id": 63894, "pid": 76337, "tid": -914061504, "ts": 1716454222287841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222350406, "dur": 7,
    "args": {
      "External id": 63897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63897,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63897, "pid": 5, "tid": 7, "ts": 1716454222350406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287859, "dur": 7,
    "args": {
      "External id": 63897,
      "cbid": 211, "correlation": 63897
    }
  },
  {
    "ph": "s", "id": 63897, "pid": 76337, "tid": -914061504, "ts": 1716454222287859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222350414, "dur": 5,
    "args": {
      "External id": 63906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63906,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63906, "pid": 5, "tid": 7, "ts": 1716454222350414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287897, "dur": 13,
    "args": {
      "External id": 63906,
      "cbid": 211, "correlation": 63906
    }
  },
  {
    "ph": "s", "id": 63906, "pid": 76337, "tid": -914061504, "ts": 1716454222287897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287953, "dur": 0,
    "args": {
      "External id": 63916,
      "cbid": 317, "correlation": 63916
    }
  },
  {
    "ph": "f", "id": 63916, "pid": 76337, "tid": -914061504, "ts": 1716454222287953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287954, "dur": 0,
    "args": {
      "External id": 63917,
      "cbid": 203, "correlation": 63917
    }
  },
  {
    "ph": "f", "id": 63917, "pid": 76337, "tid": -914061504, "ts": 1716454222287954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287955, "dur": 0,
    "args": {
      "External id": 63918,
      "cbid": 205, "correlation": 63918
    }
  },
  {
    "ph": "f", "id": 63918, "pid": 76337, "tid": -914061504, "ts": 1716454222287955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350420, "dur": 6,
    "args": {
      "External id": 63922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63922,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63922, "pid": 5, "tid": 7, "ts": 1716454222350420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222287968, "dur": 118,
    "args": {
      "External id": 63922,
      "cbid": 211, "correlation": 63922
    }
  },
  {
    "ph": "s", "id": 63922, "pid": 76337, "tid": -914061504, "ts": 1716454222287968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350427, "dur": 177,
    "args": {
      "External id": 63924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63924,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63924, "pid": 5, "tid": 7, "ts": 1716454222350427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288090, "dur": 6,
    "args": {
      "External id": 63924,
      "cbid": 211, "correlation": 63924
    }
  },
  {
    "ph": "s", "id": 63924, "pid": 76337, "tid": -914061504, "ts": 1716454222288090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222350607, "dur": 1,
    "args": {
      "External id": 63926,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 63926,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 63926, "pid": 5, "tid": 7, "ts": 1716454222350607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288102, "dur": 6,
    "args": {
      "External id": 63926,
      "cbid": 51, "correlation": 63926
    }
  },
  {
    "ph": "s", "id": 63926, "pid": 76337, "tid": -914061504, "ts": 1716454222288102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222350610, "dur": 284,
    "args": {
      "External id": 63927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63927,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63927, "pid": 5, "tid": 7, "ts": 1716454222350610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288109, "dur": 6,
    "args": {
      "External id": 63927,
      "cbid": 211, "correlation": 63927
    }
  },
  {
    "ph": "s", "id": 63927, "pid": 76337, "tid": -914061504, "ts": 1716454222288109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350895, "dur": 6,
    "args": {
      "External id": 63929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63929,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 63929, "pid": 5, "tid": 7, "ts": 1716454222350895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288119, "dur": 5,
    "args": {
      "External id": 63929,
      "cbid": 211, "correlation": 63929
    }
  },
  {
    "ph": "s", "id": 63929, "pid": 76337, "tid": -914061504, "ts": 1716454222288119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222350903, "dur": 7,
    "args": {
      "External id": 63935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63935, "pid": 5, "tid": 7, "ts": 1716454222350903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288151, "dur": 8,
    "args": {
      "External id": 63935,
      "cbid": 211, "correlation": 63935
    }
  },
  {
    "ph": "s", "id": 63935, "pid": 76337, "tid": -914061504, "ts": 1716454222288151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222350911, "dur": 3,
    "args": {
      "External id": 63943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63943,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 63943, "pid": 5, "tid": 7, "ts": 1716454222350911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288198, "dur": 9,
    "args": {
      "External id": 63943,
      "cbid": 211, "correlation": 63943
    }
  },
  {
    "ph": "s", "id": 63943, "pid": 76337, "tid": -914061504, "ts": 1716454222288198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288260, "dur": 1,
    "args": {
      "External id": 63959,
      "cbid": 251, "correlation": 63959
    }
  },
  {
    "ph": "f", "id": 63959, "pid": 76337, "tid": -914061504, "ts": 1716454222288260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288266, "dur": 0,
    "args": {
      "External id": 63961,
      "cbid": 251, "correlation": 63961
    }
  },
  {
    "ph": "f", "id": 63961, "pid": 76337, "tid": -914061504, "ts": 1716454222288266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222350916, "dur": 12,
    "args": {
      "External id": 63962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63962,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63962, "pid": 5, "tid": 7, "ts": 1716454222350916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288268, "dur": 11,
    "args": {
      "External id": 63962,
      "cbid": 211, "correlation": 63962
    }
  },
  {
    "ph": "s", "id": 63962, "pid": 76337, "tid": -914061504, "ts": 1716454222288268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222350929, "dur": 4,
    "args": {
      "External id": 63964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63964,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 63964, "pid": 5, "tid": 7, "ts": 1716454222350929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288281, "dur": 6,
    "args": {
      "External id": 63964,
      "cbid": 211, "correlation": 63964
    }
  },
  {
    "ph": "s", "id": 63964, "pid": 76337, "tid": -914061504, "ts": 1716454222288281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222350934, "dur": 6,
    "args": {
      "External id": 63974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63974,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 63974, "pid": 5, "tid": 7, "ts": 1716454222350934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288344, "dur": 12,
    "args": {
      "External id": 63974,
      "cbid": 211, "correlation": 63974
    }
  },
  {
    "ph": "s", "id": 63974, "pid": 76337, "tid": -914061504, "ts": 1716454222288344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350942, "dur": 10,
    "args": {
      "External id": 63994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 63994,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 63994, "pid": 5, "tid": 7, "ts": 1716454222350942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288411, "dur": 11,
    "args": {
      "External id": 63994,
      "cbid": 211, "correlation": 63994
    }
  },
  {
    "ph": "s", "id": 63994, "pid": 76337, "tid": -914061504, "ts": 1716454222288411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222350953, "dur": 4,
    "args": {
      "External id": 64006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64006,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 64006, "pid": 5, "tid": 7, "ts": 1716454222350953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288432, "dur": 6,
    "args": {
      "External id": 64006,
      "cbid": 211, "correlation": 64006
    }
  },
  {
    "ph": "s", "id": 64006, "pid": 76337, "tid": -914061504, "ts": 1716454222288432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222350959, "dur": 7,
    "args": {
      "External id": 64009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64009,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64009, "pid": 5, "tid": 7, "ts": 1716454222350959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288451, "dur": 6,
    "args": {
      "External id": 64009,
      "cbid": 211, "correlation": 64009
    }
  },
  {
    "ph": "s", "id": 64009, "pid": 76337, "tid": -914061504, "ts": 1716454222288451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222350967, "dur": 5,
    "args": {
      "External id": 64018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64018,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64018, "pid": 5, "tid": 7, "ts": 1716454222350967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288491, "dur": 10,
    "args": {
      "External id": 64018,
      "cbid": 211, "correlation": 64018
    }
  },
  {
    "ph": "s", "id": 64018, "pid": 76337, "tid": -914061504, "ts": 1716454222288491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288555, "dur": 0,
    "args": {
      "External id": 64028,
      "cbid": 317, "correlation": 64028
    }
  },
  {
    "ph": "f", "id": 64028, "pid": 76337, "tid": -914061504, "ts": 1716454222288555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288556, "dur": 0,
    "args": {
      "External id": 64029,
      "cbid": 203, "correlation": 64029
    }
  },
  {
    "ph": "f", "id": 64029, "pid": 76337, "tid": -914061504, "ts": 1716454222288556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288557, "dur": 0,
    "args": {
      "External id": 64030,
      "cbid": 205, "correlation": 64030
    }
  },
  {
    "ph": "f", "id": 64030, "pid": 76337, "tid": -914061504, "ts": 1716454222288557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350974, "dur": 6,
    "args": {
      "External id": 64034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64034,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64034, "pid": 5, "tid": 7, "ts": 1716454222350974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288570, "dur": 12,
    "args": {
      "External id": 64034,
      "cbid": 211, "correlation": 64034
    }
  },
  {
    "ph": "s", "id": 64034, "pid": 76337, "tid": -914061504, "ts": 1716454222288570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222350981, "dur": 176,
    "args": {
      "External id": 64036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64036,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64036, "pid": 5, "tid": 7, "ts": 1716454222350981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288585, "dur": 5,
    "args": {
      "External id": 64036,
      "cbid": 211, "correlation": 64036
    }
  },
  {
    "ph": "s", "id": 64036, "pid": 76337, "tid": -914061504, "ts": 1716454222288585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222351159, "dur": 1,
    "args": {
      "External id": 64038,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64038,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 64038, "pid": 5, "tid": 7, "ts": 1716454222351159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288596, "dur": 10,
    "args": {
      "External id": 64038,
      "cbid": 51, "correlation": 64038
    }
  },
  {
    "ph": "s", "id": 64038, "pid": 76337, "tid": -914061504, "ts": 1716454222288596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222351163, "dur": 284,
    "args": {
      "External id": 64039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64039,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64039, "pid": 5, "tid": 7, "ts": 1716454222351163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288607, "dur": 6,
    "args": {
      "External id": 64039,
      "cbid": 211, "correlation": 64039
    }
  },
  {
    "ph": "s", "id": 64039, "pid": 76337, "tid": -914061504, "ts": 1716454222288607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222351449, "dur": 7,
    "args": {
      "External id": 64041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64041,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64041, "pid": 5, "tid": 7, "ts": 1716454222351449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288618, "dur": 5,
    "args": {
      "External id": 64041,
      "cbid": 211, "correlation": 64041
    }
  },
  {
    "ph": "s", "id": 64041, "pid": 76337, "tid": -914061504, "ts": 1716454222288618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222351457, "dur": 7,
    "args": {
      "External id": 64047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64047, "pid": 5, "tid": 7, "ts": 1716454222351457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288646, "dur": 9,
    "args": {
      "External id": 64047,
      "cbid": 211, "correlation": 64047
    }
  },
  {
    "ph": "s", "id": 64047, "pid": 76337, "tid": -914061504, "ts": 1716454222288646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222351465, "dur": 5,
    "args": {
      "External id": 64055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64055,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64055, "pid": 5, "tid": 7, "ts": 1716454222351465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288679, "dur": 8,
    "args": {
      "External id": 64055,
      "cbid": 211, "correlation": 64055
    }
  },
  {
    "ph": "s", "id": 64055, "pid": 76337, "tid": -914061504, "ts": 1716454222288679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222351472, "dur": 5,
    "args": {
      "External id": 64063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64063,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64063, "pid": 5, "tid": 7, "ts": 1716454222351472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288708, "dur": 8,
    "args": {
      "External id": 64063,
      "cbid": 211, "correlation": 64063
    }
  },
  {
    "ph": "s", "id": 64063, "pid": 76337, "tid": -914061504, "ts": 1716454222288708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222351478, "dur": 11,
    "args": {
      "External id": 64083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64083,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 64083, "pid": 5, "tid": 7, "ts": 1716454222351478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288852, "dur": 14,
    "args": {
      "External id": 64083,
      "cbid": 211, "correlation": 64083
    }
  },
  {
    "ph": "s", "id": 64083, "pid": 76337, "tid": -914061504, "ts": 1716454222288852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222351490, "dur": 4,
    "args": {
      "External id": 64095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64095,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 64095, "pid": 5, "tid": 7, "ts": 1716454222351490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288876, "dur": 6,
    "args": {
      "External id": 64095,
      "cbid": 211, "correlation": 64095
    }
  },
  {
    "ph": "s", "id": 64095, "pid": 76337, "tid": -914061504, "ts": 1716454222288876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222351496, "dur": 7,
    "args": {
      "External id": 64098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64098, "pid": 5, "tid": 7, "ts": 1716454222351496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288894, "dur": 7,
    "args": {
      "External id": 64098,
      "cbid": 211, "correlation": 64098
    }
  },
  {
    "ph": "s", "id": 64098, "pid": 76337, "tid": -914061504, "ts": 1716454222288894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288956, "dur": 0,
    "args": {
      "External id": 64109,
      "cbid": 317, "correlation": 64109
    }
  },
  {
    "ph": "f", "id": 64109, "pid": 76337, "tid": -914061504, "ts": 1716454222288956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288957, "dur": 0,
    "args": {
      "External id": 64110,
      "cbid": 203, "correlation": 64110
    }
  },
  {
    "ph": "f", "id": 64110, "pid": 76337, "tid": -914061504, "ts": 1716454222288957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288958, "dur": 0,
    "args": {
      "External id": 64111,
      "cbid": 205, "correlation": 64111
    }
  },
  {
    "ph": "f", "id": 64111, "pid": 76337, "tid": -914061504, "ts": 1716454222288958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222351504, "dur": 5,
    "args": {
      "External id": 64115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64115,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64115, "pid": 5, "tid": 7, "ts": 1716454222351504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222288989, "dur": 13,
    "args": {
      "External id": 64115,
      "cbid": 211, "correlation": 64115
    }
  },
  {
    "ph": "s", "id": 64115, "pid": 76337, "tid": -914061504, "ts": 1716454222288989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222351511, "dur": 41,
    "args": {
      "External id": 64117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64117,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 64117, "pid": 5, "tid": 7, "ts": 1716454222351511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289030, "dur": 11,
    "args": {
      "External id": 64117,
      "cbid": 211, "correlation": 64117
    }
  },
  {
    "ph": "s", "id": 64117, "pid": 76337, "tid": -914061504, "ts": 1716454222289030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222351554, "dur": 6,
    "args": {
      "External id": 64119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64119,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64119, "pid": 5, "tid": 7, "ts": 1716454222351554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289046, "dur": 5,
    "args": {
      "External id": 64119,
      "cbid": 211, "correlation": 64119
    }
  },
  {
    "ph": "s", "id": 64119, "pid": 76337, "tid": -914061504, "ts": 1716454222289046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222351561, "dur": 7,
    "args": {
      "External id": 64125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64125, "pid": 5, "tid": 7, "ts": 1716454222351561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289075, "dur": 8,
    "args": {
      "External id": 64125,
      "cbid": 211, "correlation": 64125
    }
  },
  {
    "ph": "s", "id": 64125, "pid": 76337, "tid": -914061504, "ts": 1716454222289075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222351569, "dur": 22,
    "args": {
      "External id": 64134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64134,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64134, "pid": 5, "tid": 7, "ts": 1716454222351569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289163, "dur": 15,
    "args": {
      "External id": 64134,
      "cbid": 211, "correlation": 64134
    }
  },
  {
    "ph": "s", "id": 64134, "pid": 76337, "tid": -914061504, "ts": 1716454222289163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222351592, "dur": 12,
    "args": {
      "External id": 64156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64156,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 64156, "pid": 5, "tid": 7, "ts": 1716454222351592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289221, "dur": 10,
    "args": {
      "External id": 64156,
      "cbid": 211, "correlation": 64156
    }
  },
  {
    "ph": "s", "id": 64156, "pid": 76337, "tid": -914061504, "ts": 1716454222289221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289313, "dur": 2,
    "args": {
      "External id": 64167,
      "cbid": 251, "correlation": 64167
    }
  },
  {
    "ph": "f", "id": 64167, "pid": 76337, "tid": -914061504, "ts": 1716454222289313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289318, "dur": 0,
    "args": {
      "External id": 64168,
      "cbid": 251, "correlation": 64168
    }
  },
  {
    "ph": "f", "id": 64168, "pid": 76337, "tid": -914061504, "ts": 1716454222289318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222351605, "dur": 60,
    "args": {
      "External id": 64169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64169,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 64169, "pid": 5, "tid": 7, "ts": 1716454222351605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289320, "dur": 14,
    "args": {
      "External id": 64169,
      "cbid": 211, "correlation": 64169
    }
  },
  {
    "ph": "s", "id": 64169, "pid": 76337, "tid": -914061504, "ts": 1716454222289320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289395, "dur": 1,
    "args": {
      "External id": 64180,
      "cbid": 251, "correlation": 64180
    }
  },
  {
    "ph": "f", "id": 64180, "pid": 76337, "tid": -914061504, "ts": 1716454222289395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289399, "dur": 0,
    "args": {
      "External id": 64181,
      "cbid": 251, "correlation": 64181
    }
  },
  {
    "ph": "f", "id": 64181, "pid": 76337, "tid": -914061504, "ts": 1716454222289399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222351666, "dur": 57,
    "args": {
      "External id": 64182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64182,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 64182, "pid": 5, "tid": 7, "ts": 1716454222351666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289400, "dur": 12,
    "args": {
      "External id": 64182,
      "cbid": 211, "correlation": 64182
    }
  },
  {
    "ph": "s", "id": 64182, "pid": 76337, "tid": -914061504, "ts": 1716454222289400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289466, "dur": 1,
    "args": {
      "External id": 64193,
      "cbid": 251, "correlation": 64193
    }
  },
  {
    "ph": "f", "id": 64193, "pid": 76337, "tid": -914061504, "ts": 1716454222289466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289470, "dur": 0,
    "args": {
      "External id": 64194,
      "cbid": 251, "correlation": 64194
    }
  },
  {
    "ph": "f", "id": 64194, "pid": 76337, "tid": -914061504, "ts": 1716454222289470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222351725, "dur": 58,
    "args": {
      "External id": 64195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64195,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 64195, "pid": 5, "tid": 7, "ts": 1716454222351725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289472, "dur": 11,
    "args": {
      "External id": 64195,
      "cbid": 211, "correlation": 64195
    }
  },
  {
    "ph": "s", "id": 64195, "pid": 76337, "tid": -914061504, "ts": 1716454222289472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222351785, "dur": 61,
    "args": {
      "External id": 64220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64220,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64220, "pid": 5, "tid": 7, "ts": 1716454222351785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289554, "dur": 13,
    "args": {
      "External id": 64220,
      "cbid": 211, "correlation": 64220
    }
  },
  {
    "ph": "s", "id": 64220, "pid": 76337, "tid": -914061504, "ts": 1716454222289554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289655, "dur": 1,
    "args": {
      "External id": 64238,
      "cbid": 251, "correlation": 64238
    }
  },
  {
    "ph": "f", "id": 64238, "pid": 76337, "tid": -914061504, "ts": 1716454222289655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222351847, "dur": 69,
    "args": {
      "External id": 64240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64240,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 64240, "pid": 5, "tid": 7, "ts": 1716454222351847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289661, "dur": 14,
    "args": {
      "External id": 64240,
      "cbid": 211, "correlation": 64240
    }
  },
  {
    "ph": "s", "id": 64240, "pid": 76337, "tid": -914061504, "ts": 1716454222289661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222351918, "dur": 7,
    "args": {
      "External id": 64248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64248,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64248, "pid": 5, "tid": 7, "ts": 1716454222351918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289731, "dur": 12,
    "args": {
      "External id": 64248,
      "cbid": 211, "correlation": 64248
    }
  },
  {
    "ph": "s", "id": 64248, "pid": 76337, "tid": -914061504, "ts": 1716454222289731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222351926, "dur": 8,
    "args": {
      "External id": 64256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64256,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64256, "pid": 5, "tid": 7, "ts": 1716454222351926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289769, "dur": 12,
    "args": {
      "External id": 64256,
      "cbid": 211, "correlation": 64256
    }
  },
  {
    "ph": "s", "id": 64256, "pid": 76337, "tid": -914061504, "ts": 1716454222289769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222351935, "dur": 8,
    "args": {
      "External id": 64267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64267,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64267, "pid": 5, "tid": 7, "ts": 1716454222351935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289846, "dur": 13,
    "args": {
      "External id": 64267,
      "cbid": 211, "correlation": 64267
    }
  },
  {
    "ph": "s", "id": 64267, "pid": 76337, "tid": -914061504, "ts": 1716454222289846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222351945, "dur": 9,
    "args": {
      "External id": 64289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64289,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 64289, "pid": 5, "tid": 7, "ts": 1716454222351945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289879, "dur": 8,
    "args": {
      "External id": 64289,
      "cbid": 211, "correlation": 64289
    }
  },
  {
    "ph": "s", "id": 64289, "pid": 76337, "tid": -914061504, "ts": 1716454222289879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289966, "dur": 2,
    "args": {
      "External id": 64300,
      "cbid": 251, "correlation": 64300
    }
  },
  {
    "ph": "f", "id": 64300, "pid": 76337, "tid": -914061504, "ts": 1716454222289966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222351956, "dur": 1,
    "args": {
      "External id": 64301,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64301,
      "bytes": 480, "memory bandwidth (GB/s)": 0.29411764705882354
    }
  },
  {
    "ph": "f", "id": 64301, "pid": 5, "tid": 7, "ts": 1716454222351956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289972, "dur": 18,
    "args": {
      "External id": 64301,
      "cbid": 51, "correlation": 64301
    }
  },
  {
    "ph": "s", "id": 64301, "pid": 76337, "tid": -914061504, "ts": 1716454222289972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222351960, "dur": 39,
    "args": {
      "External id": 64302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64302,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 64302, "pid": 5, "tid": 7, "ts": 1716454222351960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222289992, "dur": 16,
    "args": {
      "External id": 64302,
      "cbid": 211, "correlation": 64302
    }
  },
  {
    "ph": "s", "id": 64302, "pid": 76337, "tid": -914061504, "ts": 1716454222289992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290069, "dur": 1,
    "args": {
      "External id": 64313,
      "cbid": 251, "correlation": 64313
    }
  },
  {
    "ph": "f", "id": 64313, "pid": 76337, "tid": -914061504, "ts": 1716454222290069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290073, "dur": 0,
    "args": {
      "External id": 64314,
      "cbid": 251, "correlation": 64314
    }
  },
  {
    "ph": "f", "id": 64314, "pid": 76337, "tid": -914061504, "ts": 1716454222290073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222352001, "dur": 12,
    "args": {
      "External id": 64315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64315,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64315, "pid": 5, "tid": 7, "ts": 1716454222352001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290075, "dur": 13,
    "args": {
      "External id": 64315,
      "cbid": 211, "correlation": 64315
    }
  },
  {
    "ph": "s", "id": 64315, "pid": 76337, "tid": -914061504, "ts": 1716454222290075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222352015, "dur": 6,
    "args": {
      "External id": 64317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64317,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64317, "pid": 5, "tid": 7, "ts": 1716454222352015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290090, "dur": 6,
    "args": {
      "External id": 64317,
      "cbid": 211, "correlation": 64317
    }
  },
  {
    "ph": "s", "id": 64317, "pid": 76337, "tid": -914061504, "ts": 1716454222290090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290146, "dur": 1,
    "args": {
      "External id": 64328,
      "cbid": 251, "correlation": 64328
    }
  },
  {
    "ph": "f", "id": 64328, "pid": 76337, "tid": -914061504, "ts": 1716454222290146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290150, "dur": 0,
    "args": {
      "External id": 64329,
      "cbid": 251, "correlation": 64329
    }
  },
  {
    "ph": "f", "id": 64329, "pid": 76337, "tid": -914061504, "ts": 1716454222290150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222352022, "dur": 9,
    "args": {
      "External id": 64330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64330,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64330, "pid": 5, "tid": 7, "ts": 1716454222352022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290151, "dur": 14,
    "args": {
      "External id": 64330,
      "cbid": 211, "correlation": 64330
    }
  },
  {
    "ph": "s", "id": 64330, "pid": 76337, "tid": -914061504, "ts": 1716454222290151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222352032, "dur": 4,
    "args": {
      "External id": 64332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64332,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64332, "pid": 5, "tid": 7, "ts": 1716454222352032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290167, "dur": 5,
    "args": {
      "External id": 64332,
      "cbid": 211, "correlation": 64332
    }
  },
  {
    "ph": "s", "id": 64332, "pid": 76337, "tid": -914061504, "ts": 1716454222290167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222352037, "dur": 22,
    "args": {
      "External id": 64357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64357,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 64357, "pid": 5, "tid": 7, "ts": 1716454222352037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290246, "dur": 13,
    "args": {
      "External id": 64357,
      "cbid": 211, "correlation": 64357
    }
  },
  {
    "ph": "s", "id": 64357, "pid": 76337, "tid": -914061504, "ts": 1716454222290246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290351, "dur": 2,
    "args": {
      "External id": 64375,
      "cbid": 251, "correlation": 64375
    }
  },
  {
    "ph": "f", "id": 64375, "pid": 76337, "tid": -914061504, "ts": 1716454222290351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222352061, "dur": 1,
    "args": {
      "External id": 64377,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64377,
      "bytes": 480, "memory bandwidth (GB/s)": 0.29411764705882354
    }
  },
  {
    "ph": "f", "id": 64377, "pid": 5, "tid": 7, "ts": 1716454222352061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290357, "dur": 10,
    "args": {
      "External id": 64377,
      "cbid": 51, "correlation": 64377
    }
  },
  {
    "ph": "s", "id": 64377, "pid": 76337, "tid": -914061504, "ts": 1716454222290357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222352065, "dur": 40,
    "args": {
      "External id": 64378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64378,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 64378, "pid": 5, "tid": 7, "ts": 1716454222352065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290369, "dur": 13,
    "args": {
      "External id": 64378,
      "cbid": 211, "correlation": 64378
    }
  },
  {
    "ph": "s", "id": 64378, "pid": 76337, "tid": -914061504, "ts": 1716454222290369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222352107, "dur": 5,
    "args": {
      "External id": 64386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64386,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64386, "pid": 5, "tid": 7, "ts": 1716454222352107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290440, "dur": 12,
    "args": {
      "External id": 64386,
      "cbid": 211, "correlation": 64386
    }
  },
  {
    "ph": "s", "id": 64386, "pid": 76337, "tid": -914061504, "ts": 1716454222290440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222352113, "dur": 9,
    "args": {
      "External id": 64394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64394, "pid": 5, "tid": 7, "ts": 1716454222352113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290482, "dur": 9,
    "args": {
      "External id": 64394,
      "cbid": 211, "correlation": 64394
    }
  },
  {
    "ph": "s", "id": 64394, "pid": 76337, "tid": -914061504, "ts": 1716454222290482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222352124, "dur": 9,
    "args": {
      "External id": 64416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64416,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 64416, "pid": 5, "tid": 7, "ts": 1716454222352124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290536, "dur": 11,
    "args": {
      "External id": 64416,
      "cbid": 211, "correlation": 64416
    }
  },
  {
    "ph": "s", "id": 64416, "pid": 76337, "tid": -914061504, "ts": 1716454222290536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290629, "dur": 1,
    "args": {
      "External id": 64432,
      "cbid": 251, "correlation": 64432
    }
  },
  {
    "ph": "f", "id": 64432, "pid": 76337, "tid": -914061504, "ts": 1716454222290629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290634, "dur": 0,
    "args": {
      "External id": 64434,
      "cbid": 251, "correlation": 64434
    }
  },
  {
    "ph": "f", "id": 64434, "pid": 76337, "tid": -914061504, "ts": 1716454222290634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222352134, "dur": 199,
    "args": {
      "External id": 64435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64435,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64435, "pid": 5, "tid": 7, "ts": 1716454222352134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290636, "dur": 13,
    "args": {
      "External id": 64435,
      "cbid": 211, "correlation": 64435
    }
  },
  {
    "ph": "s", "id": 64435, "pid": 76337, "tid": -914061504, "ts": 1716454222290636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222352335, "dur": 23,
    "args": {
      "External id": 64443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64443, "pid": 5, "tid": 7, "ts": 1716454222352335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290704, "dur": 13,
    "args": {
      "External id": 64443,
      "cbid": 211, "correlation": 64443
    }
  },
  {
    "ph": "s", "id": 64443, "pid": 76337, "tid": -914061504, "ts": 1716454222290704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222352359, "dur": 23,
    "args": {
      "External id": 64451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64451,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64451, "pid": 5, "tid": 7, "ts": 1716454222352359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290736, "dur": 8,
    "args": {
      "External id": 64451,
      "cbid": 211, "correlation": 64451
    }
  },
  {
    "ph": "s", "id": 64451, "pid": 76337, "tid": -914061504, "ts": 1716454222290736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290818, "dur": 1,
    "args": {
      "External id": 64467,
      "cbid": 251, "correlation": 64467
    }
  },
  {
    "ph": "f", "id": 64467, "pid": 76337, "tid": -914061504, "ts": 1716454222290818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222352384, "dur": 1,
    "args": {
      "External id": 64469,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64469,
      "bytes": 120, "memory bandwidth (GB/s)": 0.075
    }
  },
  {
    "ph": "f", "id": 64469, "pid": 5, "tid": 7, "ts": 1716454222352384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290824, "dur": 9,
    "args": {
      "External id": 64469,
      "cbid": 51, "correlation": 64469
    }
  },
  {
    "ph": "s", "id": 64469, "pid": 76337, "tid": -914061504, "ts": 1716454222290824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222352388, "dur": 119,
    "args": {
      "External id": 64470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64470,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 64470, "pid": 5, "tid": 7, "ts": 1716454222352388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290834, "dur": 16,
    "args": {
      "External id": 64470,
      "cbid": 211, "correlation": 64470
    }
  },
  {
    "ph": "s", "id": 64470, "pid": 76337, "tid": -914061504, "ts": 1716454222290834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222352509, "dur": 6,
    "args": {
      "External id": 64478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64478,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64478, "pid": 5, "tid": 7, "ts": 1716454222352509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290881, "dur": 13,
    "args": {
      "External id": 64478,
      "cbid": 211, "correlation": 64478
    }
  },
  {
    "ph": "s", "id": 64478, "pid": 76337, "tid": -914061504, "ts": 1716454222290881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222352516, "dur": 11,
    "args": {
      "External id": 64489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64489,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64489, "pid": 5, "tid": 7, "ts": 1716454222352516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222290954, "dur": 13,
    "args": {
      "External id": 64489,
      "cbid": 211, "correlation": 64489
    }
  },
  {
    "ph": "s", "id": 64489, "pid": 76337, "tid": -914061504, "ts": 1716454222290954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291028, "dur": 0,
    "args": {
      "External id": 64501,
      "cbid": 317, "correlation": 64501
    }
  },
  {
    "ph": "f", "id": 64501, "pid": 76337, "tid": -914061504, "ts": 1716454222291028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291029, "dur": 0,
    "args": {
      "External id": 64502,
      "cbid": 203, "correlation": 64502
    }
  },
  {
    "ph": "f", "id": 64502, "pid": 76337, "tid": -914061504, "ts": 1716454222291029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291030, "dur": 0,
    "args": {
      "External id": 64503,
      "cbid": 205, "correlation": 64503
    }
  },
  {
    "ph": "f", "id": 64503, "pid": 76337, "tid": -914061504, "ts": 1716454222291030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222352528, "dur": 6,
    "args": {
      "External id": 64507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64507,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64507, "pid": 5, "tid": 7, "ts": 1716454222352528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291045, "dur": 13,
    "args": {
      "External id": 64507,
      "cbid": 211, "correlation": 64507
    }
  },
  {
    "ph": "s", "id": 64507, "pid": 76337, "tid": -914061504, "ts": 1716454222291045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222352535, "dur": 41,
    "args": {
      "External id": 64509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64509,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 64509, "pid": 5, "tid": 7, "ts": 1716454222352535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291065, "dur": 7,
    "args": {
      "External id": 64509,
      "cbid": 211, "correlation": 64509
    }
  },
  {
    "ph": "s", "id": 64509, "pid": 76337, "tid": -914061504, "ts": 1716454222291065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222352577, "dur": 6,
    "args": {
      "External id": 64511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64511,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64511, "pid": 5, "tid": 7, "ts": 1716454222352577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291075, "dur": 5,
    "args": {
      "External id": 64511,
      "cbid": 211, "correlation": 64511
    }
  },
  {
    "ph": "s", "id": 64511, "pid": 76337, "tid": -914061504, "ts": 1716454222291075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222352585, "dur": 8,
    "args": {
      "External id": 64517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64517, "pid": 5, "tid": 7, "ts": 1716454222352585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291103, "dur": 9,
    "args": {
      "External id": 64517,
      "cbid": 211, "correlation": 64517
    }
  },
  {
    "ph": "s", "id": 64517, "pid": 76337, "tid": -914061504, "ts": 1716454222291103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222352594, "dur": 5,
    "args": {
      "External id": 64525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64525,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64525, "pid": 5, "tid": 7, "ts": 1716454222352594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291135, "dur": 8,
    "args": {
      "External id": 64525,
      "cbid": 211, "correlation": 64525
    }
  },
  {
    "ph": "s", "id": 64525, "pid": 76337, "tid": -914061504, "ts": 1716454222291135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222352601, "dur": 11,
    "args": {
      "External id": 64545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64545,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 64545, "pid": 5, "tid": 7, "ts": 1716454222352601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291207, "dur": 12,
    "args": {
      "External id": 64545,
      "cbid": 211, "correlation": 64545
    }
  },
  {
    "ph": "s", "id": 64545, "pid": 76337, "tid": -914061504, "ts": 1716454222291207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222352614, "dur": 5,
    "args": {
      "External id": 64557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64557,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 64557, "pid": 5, "tid": 7, "ts": 1716454222352614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291233, "dur": 7,
    "args": {
      "External id": 64557,
      "cbid": 211, "correlation": 64557
    }
  },
  {
    "ph": "s", "id": 64557, "pid": 76337, "tid": -914061504, "ts": 1716454222291233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222352620, "dur": 9,
    "args": {
      "External id": 64560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64560, "pid": 5, "tid": 7, "ts": 1716454222352620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291251, "dur": 6,
    "args": {
      "External id": 64560,
      "cbid": 211, "correlation": 64560
    }
  },
  {
    "ph": "s", "id": 64560, "pid": 76337, "tid": -914061504, "ts": 1716454222291251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222352630, "dur": 6,
    "args": {
      "External id": 64569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64569,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64569, "pid": 5, "tid": 7, "ts": 1716454222352630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291291, "dur": 10,
    "args": {
      "External id": 64569,
      "cbid": 211, "correlation": 64569
    }
  },
  {
    "ph": "s", "id": 64569, "pid": 76337, "tid": -914061504, "ts": 1716454222291291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291343, "dur": 0,
    "args": {
      "External id": 64579,
      "cbid": 317, "correlation": 64579
    }
  },
  {
    "ph": "f", "id": 64579, "pid": 76337, "tid": -914061504, "ts": 1716454222291343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291344, "dur": 0,
    "args": {
      "External id": 64580,
      "cbid": 203, "correlation": 64580
    }
  },
  {
    "ph": "f", "id": 64580, "pid": 76337, "tid": -914061504, "ts": 1716454222291344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291344, "dur": 0,
    "args": {
      "External id": 64581,
      "cbid": 205, "correlation": 64581
    }
  },
  {
    "ph": "f", "id": 64581, "pid": 76337, "tid": -914061504, "ts": 1716454222291344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222352637, "dur": 6,
    "args": {
      "External id": 64585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64585, "pid": 5, "tid": 7, "ts": 1716454222352637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291358, "dur": 11,
    "args": {
      "External id": 64585,
      "cbid": 211, "correlation": 64585
    }
  },
  {
    "ph": "s", "id": 64585, "pid": 76337, "tid": -914061504, "ts": 1716454222291358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222352644, "dur": 177,
    "args": {
      "External id": 64587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64587, "pid": 5, "tid": 7, "ts": 1716454222352644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291372, "dur": 5,
    "args": {
      "External id": 64587,
      "cbid": 211, "correlation": 64587
    }
  },
  {
    "ph": "s", "id": 64587, "pid": 76337, "tid": -914061504, "ts": 1716454222291372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222352823, "dur": 1,
    "args": {
      "External id": 64589,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64589,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 64589, "pid": 5, "tid": 7, "ts": 1716454222352823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291382, "dur": 8,
    "args": {
      "External id": 64589,
      "cbid": 51, "correlation": 64589
    }
  },
  {
    "ph": "s", "id": 64589, "pid": 76337, "tid": -914061504, "ts": 1716454222291382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222352827, "dur": 295,
    "args": {
      "External id": 64590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64590,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64590, "pid": 5, "tid": 7, "ts": 1716454222352827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291391, "dur": 6,
    "args": {
      "External id": 64590,
      "cbid": 211, "correlation": 64590
    }
  },
  {
    "ph": "s", "id": 64590, "pid": 76337, "tid": -914061504, "ts": 1716454222291391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353124, "dur": 6,
    "args": {
      "External id": 64592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64592, "pid": 5, "tid": 7, "ts": 1716454222353124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291401, "dur": 5,
    "args": {
      "External id": 64592,
      "cbid": 211, "correlation": 64592
    }
  },
  {
    "ph": "s", "id": 64592, "pid": 76337, "tid": -914061504, "ts": 1716454222291401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222353131, "dur": 7,
    "args": {
      "External id": 64598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64598, "pid": 5, "tid": 7, "ts": 1716454222353131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291429, "dur": 9,
    "args": {
      "External id": 64598,
      "cbid": 211, "correlation": 64598
    }
  },
  {
    "ph": "s", "id": 64598, "pid": 76337, "tid": -914061504, "ts": 1716454222291429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222353139, "dur": 3,
    "args": {
      "External id": 64606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64606,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 64606, "pid": 5, "tid": 7, "ts": 1716454222353139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291473, "dur": 9,
    "args": {
      "External id": 64606,
      "cbid": 211, "correlation": 64606
    }
  },
  {
    "ph": "s", "id": 64606, "pid": 76337, "tid": -914061504, "ts": 1716454222291473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291543, "dur": 1,
    "args": {
      "External id": 64622,
      "cbid": 251, "correlation": 64622
    }
  },
  {
    "ph": "f", "id": 64622, "pid": 76337, "tid": -914061504, "ts": 1716454222291543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291548, "dur": 0,
    "args": {
      "External id": 64624,
      "cbid": 251, "correlation": 64624
    }
  },
  {
    "ph": "f", "id": 64624, "pid": 76337, "tid": -914061504, "ts": 1716454222291548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222353144, "dur": 14,
    "args": {
      "External id": 64625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64625,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64625, "pid": 5, "tid": 7, "ts": 1716454222353144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291550, "dur": 11,
    "args": {
      "External id": 64625,
      "cbid": 211, "correlation": 64625
    }
  },
  {
    "ph": "s", "id": 64625, "pid": 76337, "tid": -914061504, "ts": 1716454222291550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222353159, "dur": 5,
    "args": {
      "External id": 64627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64627,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64627, "pid": 5, "tid": 7, "ts": 1716454222353159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291563, "dur": 5,
    "args": {
      "External id": 64627,
      "cbid": 211, "correlation": 64627
    }
  },
  {
    "ph": "s", "id": 64627, "pid": 76337, "tid": -914061504, "ts": 1716454222291563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222353166, "dur": 6,
    "args": {
      "External id": 64637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64637, "pid": 5, "tid": 7, "ts": 1716454222353166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291622, "dur": 12,
    "args": {
      "External id": 64637,
      "cbid": 211, "correlation": 64637
    }
  },
  {
    "ph": "s", "id": 64637, "pid": 76337, "tid": -914061504, "ts": 1716454222291622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353173, "dur": 11,
    "args": {
      "External id": 64657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64657,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 64657, "pid": 5, "tid": 7, "ts": 1716454222353173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291687, "dur": 11,
    "args": {
      "External id": 64657,
      "cbid": 211, "correlation": 64657
    }
  },
  {
    "ph": "s", "id": 64657, "pid": 76337, "tid": -914061504, "ts": 1716454222291687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222353186, "dur": 4,
    "args": {
      "External id": 64669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64669,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 64669, "pid": 5, "tid": 7, "ts": 1716454222353186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291709, "dur": 6,
    "args": {
      "External id": 64669,
      "cbid": 211, "correlation": 64669
    }
  },
  {
    "ph": "s", "id": 64669, "pid": 76337, "tid": -914061504, "ts": 1716454222291709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222353191, "dur": 8,
    "args": {
      "External id": 64672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64672, "pid": 5, "tid": 7, "ts": 1716454222353191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291728, "dur": 7,
    "args": {
      "External id": 64672,
      "cbid": 211, "correlation": 64672
    }
  },
  {
    "ph": "s", "id": 64672, "pid": 76337, "tid": -914061504, "ts": 1716454222291728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222353200, "dur": 5,
    "args": {
      "External id": 64681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64681,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64681, "pid": 5, "tid": 7, "ts": 1716454222353200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291769, "dur": 10,
    "args": {
      "External id": 64681,
      "cbid": 211, "correlation": 64681
    }
  },
  {
    "ph": "s", "id": 64681, "pid": 76337, "tid": -914061504, "ts": 1716454222291769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291836, "dur": 0,
    "args": {
      "External id": 64691,
      "cbid": 317, "correlation": 64691
    }
  },
  {
    "ph": "f", "id": 64691, "pid": 76337, "tid": -914061504, "ts": 1716454222291836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291837, "dur": 0,
    "args": {
      "External id": 64692,
      "cbid": 203, "correlation": 64692
    }
  },
  {
    "ph": "f", "id": 64692, "pid": 76337, "tid": -914061504, "ts": 1716454222291837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291838, "dur": 0,
    "args": {
      "External id": 64693,
      "cbid": 205, "correlation": 64693
    }
  },
  {
    "ph": "f", "id": 64693, "pid": 76337, "tid": -914061504, "ts": 1716454222291838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353206, "dur": 5,
    "args": {
      "External id": 64697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64697,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64697, "pid": 5, "tid": 7, "ts": 1716454222353206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291851, "dur": 12,
    "args": {
      "External id": 64697,
      "cbid": 211, "correlation": 64697
    }
  },
  {
    "ph": "s", "id": 64697, "pid": 76337, "tid": -914061504, "ts": 1716454222291851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353213, "dur": 177,
    "args": {
      "External id": 64699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64699, "pid": 5, "tid": 7, "ts": 1716454222353213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291866, "dur": 5,
    "args": {
      "External id": 64699,
      "cbid": 211, "correlation": 64699
    }
  },
  {
    "ph": "s", "id": 64699, "pid": 76337, "tid": -914061504, "ts": 1716454222291866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222353393, "dur": 1,
    "args": {
      "External id": 64701,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64701,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 64701, "pid": 5, "tid": 7, "ts": 1716454222353393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291877, "dur": 6,
    "args": {
      "External id": 64701,
      "cbid": 51, "correlation": 64701
    }
  },
  {
    "ph": "s", "id": 64701, "pid": 76337, "tid": -914061504, "ts": 1716454222291877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222353397, "dur": 284,
    "args": {
      "External id": 64702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64702,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64702, "pid": 5, "tid": 7, "ts": 1716454222353397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291884, "dur": 6,
    "args": {
      "External id": 64702,
      "cbid": 211, "correlation": 64702
    }
  },
  {
    "ph": "s", "id": 64702, "pid": 76337, "tid": -914061504, "ts": 1716454222291884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353682, "dur": 7,
    "args": {
      "External id": 64704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64704, "pid": 5, "tid": 7, "ts": 1716454222353682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291894, "dur": 5,
    "args": {
      "External id": 64704,
      "cbid": 211, "correlation": 64704
    }
  },
  {
    "ph": "s", "id": 64704, "pid": 76337, "tid": -914061504, "ts": 1716454222291894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222353690, "dur": 7,
    "args": {
      "External id": 64710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64710, "pid": 5, "tid": 7, "ts": 1716454222353690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291922, "dur": 9,
    "args": {
      "External id": 64710,
      "cbid": 211, "correlation": 64710
    }
  },
  {
    "ph": "s", "id": 64710, "pid": 76337, "tid": -914061504, "ts": 1716454222291922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222353699, "dur": 5,
    "args": {
      "External id": 64718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64718,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64718, "pid": 5, "tid": 7, "ts": 1716454222353699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291956, "dur": 8,
    "args": {
      "External id": 64718,
      "cbid": 211, "correlation": 64718
    }
  },
  {
    "ph": "s", "id": 64718, "pid": 76337, "tid": -914061504, "ts": 1716454222291956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222353705, "dur": 5,
    "args": {
      "External id": 64726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64726,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64726, "pid": 5, "tid": 7, "ts": 1716454222353705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222291996, "dur": 9,
    "args": {
      "External id": 64726,
      "cbid": 211, "correlation": 64726
    }
  },
  {
    "ph": "s", "id": 64726, "pid": 76337, "tid": -914061504, "ts": 1716454222291996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222353712, "dur": 12,
    "args": {
      "External id": 64735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64735,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64735, "pid": 5, "tid": 7, "ts": 1716454222353712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292086, "dur": 14,
    "args": {
      "External id": 64735,
      "cbid": 211, "correlation": 64735
    }
  },
  {
    "ph": "s", "id": 64735, "pid": 76337, "tid": -914061504, "ts": 1716454222292086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353725, "dur": 14,
    "args": {
      "External id": 64755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64755,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 64755, "pid": 5, "tid": 7, "ts": 1716454222353725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292158, "dur": 11,
    "args": {
      "External id": 64755,
      "cbid": 211, "correlation": 64755
    }
  },
  {
    "ph": "s", "id": 64755, "pid": 76337, "tid": -914061504, "ts": 1716454222292158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222353740, "dur": 4,
    "args": {
      "External id": 64767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64767,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64767, "pid": 5, "tid": 7, "ts": 1716454222353740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292179, "dur": 6,
    "args": {
      "External id": 64767,
      "cbid": 211, "correlation": 64767
    }
  },
  {
    "ph": "s", "id": 64767, "pid": 76337, "tid": -914061504, "ts": 1716454222292179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222353746, "dur": 11,
    "args": {
      "External id": 64770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64770, "pid": 5, "tid": 7, "ts": 1716454222353746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292197, "dur": 7,
    "args": {
      "External id": 64770,
      "cbid": 211, "correlation": 64770
    }
  },
  {
    "ph": "s", "id": 64770, "pid": 76337, "tid": -914061504, "ts": 1716454222292197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222353758, "dur": 7,
    "args": {
      "External id": 64779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64779,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64779, "pid": 5, "tid": 7, "ts": 1716454222353758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292237, "dur": 9,
    "args": {
      "External id": 64779,
      "cbid": 211, "correlation": 64779
    }
  },
  {
    "ph": "s", "id": 64779, "pid": 76337, "tid": -914061504, "ts": 1716454222292237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292291, "dur": 0,
    "args": {
      "External id": 64789,
      "cbid": 317, "correlation": 64789
    }
  },
  {
    "ph": "f", "id": 64789, "pid": 76337, "tid": -914061504, "ts": 1716454222292291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292292, "dur": 0,
    "args": {
      "External id": 64790,
      "cbid": 203, "correlation": 64790
    }
  },
  {
    "ph": "f", "id": 64790, "pid": 76337, "tid": -914061504, "ts": 1716454222292292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292293, "dur": 0,
    "args": {
      "External id": 64791,
      "cbid": 205, "correlation": 64791
    }
  },
  {
    "ph": "f", "id": 64791, "pid": 76337, "tid": -914061504, "ts": 1716454222292293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353766, "dur": 7,
    "args": {
      "External id": 64795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64795,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64795, "pid": 5, "tid": 7, "ts": 1716454222353766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292316, "dur": 13,
    "args": {
      "External id": 64795,
      "cbid": 211, "correlation": 64795
    }
  },
  {
    "ph": "s", "id": 64795, "pid": 76337, "tid": -914061504, "ts": 1716454222292316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222353774, "dur": 349,
    "args": {
      "External id": 64797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64797,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64797, "pid": 5, "tid": 7, "ts": 1716454222353774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292331, "dur": 5,
    "args": {
      "External id": 64797,
      "cbid": 211, "correlation": 64797
    }
  },
  {
    "ph": "s", "id": 64797, "pid": 76337, "tid": -914061504, "ts": 1716454222292331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222354125, "dur": 1,
    "args": {
      "External id": 64799,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64799,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 64799, "pid": 5, "tid": 7, "ts": 1716454222354125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292342, "dur": 7,
    "args": {
      "External id": 64799,
      "cbid": 51, "correlation": 64799
    }
  },
  {
    "ph": "s", "id": 64799, "pid": 76337, "tid": -914061504, "ts": 1716454222292342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222354129, "dur": 545,
    "args": {
      "External id": 64800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64800,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64800, "pid": 5, "tid": 7, "ts": 1716454222354129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292350, "dur": 6,
    "args": {
      "External id": 64800,
      "cbid": 211, "correlation": 64800
    }
  },
  {
    "ph": "s", "id": 64800, "pid": 76337, "tid": -914061504, "ts": 1716454222292350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222354675, "dur": 6,
    "args": {
      "External id": 64802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64802, "pid": 5, "tid": 7, "ts": 1716454222354675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292360, "dur": 5,
    "args": {
      "External id": 64802,
      "cbid": 211, "correlation": 64802
    }
  },
  {
    "ph": "s", "id": 64802, "pid": 76337, "tid": -914061504, "ts": 1716454222292360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222354683, "dur": 7,
    "args": {
      "External id": 64808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64808, "pid": 5, "tid": 7, "ts": 1716454222354683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292390, "dur": 10,
    "args": {
      "External id": 64808,
      "cbid": 211, "correlation": 64808
    }
  },
  {
    "ph": "s", "id": 64808, "pid": 76337, "tid": -914061504, "ts": 1716454222292390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222354691, "dur": 3,
    "args": {
      "External id": 64816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64816,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 64816, "pid": 5, "tid": 7, "ts": 1716454222354691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292436, "dur": 9,
    "args": {
      "External id": 64816,
      "cbid": 211, "correlation": 64816
    }
  },
  {
    "ph": "s", "id": 64816, "pid": 76337, "tid": -914061504, "ts": 1716454222292436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292499, "dur": 1,
    "args": {
      "External id": 64832,
      "cbid": 251, "correlation": 64832
    }
  },
  {
    "ph": "f", "id": 64832, "pid": 76337, "tid": -914061504, "ts": 1716454222292499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292504, "dur": 0,
    "args": {
      "External id": 64834,
      "cbid": 251, "correlation": 64834
    }
  },
  {
    "ph": "f", "id": 64834, "pid": 76337, "tid": -914061504, "ts": 1716454222292504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222354695, "dur": 12,
    "args": {
      "External id": 64835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64835,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64835, "pid": 5, "tid": 7, "ts": 1716454222354695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292506, "dur": 11,
    "args": {
      "External id": 64835,
      "cbid": 211, "correlation": 64835
    }
  },
  {
    "ph": "s", "id": 64835, "pid": 76337, "tid": -914061504, "ts": 1716454222292506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222354709, "dur": 5,
    "args": {
      "External id": 64837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64837,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64837, "pid": 5, "tid": 7, "ts": 1716454222354709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292518, "dur": 6,
    "args": {
      "External id": 64837,
      "cbid": 211, "correlation": 64837
    }
  },
  {
    "ph": "s", "id": 64837, "pid": 76337, "tid": -914061504, "ts": 1716454222292518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222354715, "dur": 7,
    "args": {
      "External id": 64847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64847, "pid": 5, "tid": 7, "ts": 1716454222354715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292575, "dur": 12,
    "args": {
      "External id": 64847,
      "cbid": 211, "correlation": 64847
    }
  },
  {
    "ph": "s", "id": 64847, "pid": 76337, "tid": -914061504, "ts": 1716454222292575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222354724, "dur": 11,
    "args": {
      "External id": 64867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64867,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 64867, "pid": 5, "tid": 7, "ts": 1716454222354724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292641, "dur": 12,
    "args": {
      "External id": 64867,
      "cbid": 211, "correlation": 64867
    }
  },
  {
    "ph": "s", "id": 64867, "pid": 76337, "tid": -914061504, "ts": 1716454222292641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222354736, "dur": 4,
    "args": {
      "External id": 64879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64879,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 64879, "pid": 5, "tid": 7, "ts": 1716454222354736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292663, "dur": 6,
    "args": {
      "External id": 64879,
      "cbid": 211, "correlation": 64879
    }
  },
  {
    "ph": "s", "id": 64879, "pid": 76337, "tid": -914061504, "ts": 1716454222292663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222354741, "dur": 7,
    "args": {
      "External id": 64882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64882, "pid": 5, "tid": 7, "ts": 1716454222354741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292680, "dur": 9,
    "args": {
      "External id": 64882,
      "cbid": 211, "correlation": 64882
    }
  },
  {
    "ph": "s", "id": 64882, "pid": 76337, "tid": -914061504, "ts": 1716454222292680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222354750, "dur": 5,
    "args": {
      "External id": 64891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64891,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64891, "pid": 5, "tid": 7, "ts": 1716454222354750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292725, "dur": 10,
    "args": {
      "External id": 64891,
      "cbid": 211, "correlation": 64891
    }
  },
  {
    "ph": "s", "id": 64891, "pid": 76337, "tid": -914061504, "ts": 1716454222292725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292788, "dur": 0,
    "args": {
      "External id": 64901,
      "cbid": 317, "correlation": 64901
    }
  },
  {
    "ph": "f", "id": 64901, "pid": 76337, "tid": -914061504, "ts": 1716454222292788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292789, "dur": 0,
    "args": {
      "External id": 64902,
      "cbid": 203, "correlation": 64902
    }
  },
  {
    "ph": "f", "id": 64902, "pid": 76337, "tid": -914061504, "ts": 1716454222292789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292789, "dur": 0,
    "args": {
      "External id": 64903,
      "cbid": 205, "correlation": 64903
    }
  },
  {
    "ph": "f", "id": 64903, "pid": 76337, "tid": -914061504, "ts": 1716454222292789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222354756, "dur": 6,
    "args": {
      "External id": 64907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64907,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64907, "pid": 5, "tid": 7, "ts": 1716454222354756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292803, "dur": 12,
    "args": {
      "External id": 64907,
      "cbid": 211, "correlation": 64907
    }
  },
  {
    "ph": "s", "id": 64907, "pid": 76337, "tid": -914061504, "ts": 1716454222292803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222354763, "dur": 177,
    "args": {
      "External id": 64909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64909, "pid": 5, "tid": 7, "ts": 1716454222354763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292817, "dur": 5,
    "args": {
      "External id": 64909,
      "cbid": 211, "correlation": 64909
    }
  },
  {
    "ph": "s", "id": 64909, "pid": 76337, "tid": -914061504, "ts": 1716454222292817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222354942, "dur": 1,
    "args": {
      "External id": 64911,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64911,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 64911, "pid": 5, "tid": 7, "ts": 1716454222354942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292829, "dur": 7,
    "args": {
      "External id": 64911,
      "cbid": 51, "correlation": 64911
    }
  },
  {
    "ph": "s", "id": 64911, "pid": 76337, "tid": -914061504, "ts": 1716454222292829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222354946, "dur": 285,
    "args": {
      "External id": 64912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64912,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 64912, "pid": 5, "tid": 7, "ts": 1716454222354946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292837, "dur": 6,
    "args": {
      "External id": 64912,
      "cbid": 211, "correlation": 64912
    }
  },
  {
    "ph": "s", "id": 64912, "pid": 76337, "tid": -914061504, "ts": 1716454222292837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222355232, "dur": 6,
    "args": {
      "External id": 64914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64914, "pid": 5, "tid": 7, "ts": 1716454222355232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292846, "dur": 5,
    "args": {
      "External id": 64914,
      "cbid": 211, "correlation": 64914
    }
  },
  {
    "ph": "s", "id": 64914, "pid": 76337, "tid": -914061504, "ts": 1716454222292846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222355239, "dur": 7,
    "args": {
      "External id": 64920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64920,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64920, "pid": 5, "tid": 7, "ts": 1716454222355239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292874, "dur": 9,
    "args": {
      "External id": 64920,
      "cbid": 211, "correlation": 64920
    }
  },
  {
    "ph": "s", "id": 64920, "pid": 76337, "tid": -914061504, "ts": 1716454222292874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292934, "dur": 0,
    "args": {
      "External id": 64930,
      "cbid": 317, "correlation": 64930
    }
  },
  {
    "ph": "f", "id": 64930, "pid": 76337, "tid": -914061504, "ts": 1716454222292934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292934, "dur": 0,
    "args": {
      "External id": 64931,
      "cbid": 203, "correlation": 64931
    }
  },
  {
    "ph": "f", "id": 64931, "pid": 76337, "tid": -914061504, "ts": 1716454222292934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292935, "dur": 0,
    "args": {
      "External id": 64932,
      "cbid": 205, "correlation": 64932
    }
  },
  {
    "ph": "f", "id": 64932, "pid": 76337, "tid": -914061504, "ts": 1716454222292935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222355248, "dur": 8,
    "args": {
      "External id": 64936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64936,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64936, "pid": 5, "tid": 7, "ts": 1716454222355248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292955, "dur": 12,
    "args": {
      "External id": 64936,
      "cbid": 211, "correlation": 64936
    }
  },
  {
    "ph": "s", "id": 64936, "pid": 76337, "tid": -914061504, "ts": 1716454222292955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222355257, "dur": 3,
    "args": {
      "External id": 64938,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 64938,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3761467889908257
    }
  },
  {
    "ph": "f", "id": 64938, "pid": 5, "tid": 7, "ts": 1716454222355257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292972, "dur": 23,
    "args": {
      "External id": 64938,
      "cbid": 51, "correlation": 64938
    }
  },
  {
    "ph": "s", "id": 64938, "pid": 76337, "tid": -914061504, "ts": 1716454222292972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222355262, "dur": 103,
    "args": {
      "External id": 64939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64939,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 64939, "pid": 5, "tid": 7, "ts": 1716454222355262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222292996, "dur": 7,
    "args": {
      "External id": 64939,
      "cbid": 211, "correlation": 64939
    }
  },
  {
    "ph": "s", "id": 64939, "pid": 76337, "tid": -914061504, "ts": 1716454222292996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222355366, "dur": 6,
    "args": {
      "External id": 64941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64941,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64941, "pid": 5, "tid": 7, "ts": 1716454222355366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293008, "dur": 7,
    "args": {
      "External id": 64941,
      "cbid": 211, "correlation": 64941
    }
  },
  {
    "ph": "s", "id": 64941, "pid": 76337, "tid": -914061504, "ts": 1716454222293008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222355374, "dur": 7,
    "args": {
      "External id": 64947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64947,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64947, "pid": 5, "tid": 7, "ts": 1716454222355374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293038, "dur": 9,
    "args": {
      "External id": 64947,
      "cbid": 211, "correlation": 64947
    }
  },
  {
    "ph": "s", "id": 64947, "pid": 76337, "tid": -914061504, "ts": 1716454222293038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222355382, "dur": 5,
    "args": {
      "External id": 64955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64955,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64955, "pid": 5, "tid": 7, "ts": 1716454222355382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293069, "dur": 8,
    "args": {
      "External id": 64955,
      "cbid": 211, "correlation": 64955
    }
  },
  {
    "ph": "s", "id": 64955, "pid": 76337, "tid": -914061504, "ts": 1716454222293069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222355388, "dur": 5,
    "args": {
      "External id": 64963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64963,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 64963, "pid": 5, "tid": 7, "ts": 1716454222355388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293099, "dur": 8,
    "args": {
      "External id": 64963,
      "cbid": 211, "correlation": 64963
    }
  },
  {
    "ph": "s", "id": 64963, "pid": 76337, "tid": -914061504, "ts": 1716454222293099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222355395, "dur": 12,
    "args": {
      "External id": 64972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64972,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 64972, "pid": 5, "tid": 7, "ts": 1716454222355395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293142, "dur": 11,
    "args": {
      "External id": 64972,
      "cbid": 211, "correlation": 64972
    }
  },
  {
    "ph": "s", "id": 64972, "pid": 76337, "tid": -914061504, "ts": 1716454222293142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222355408, "dur": 14,
    "args": {
      "External id": 64992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 64992,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 64992, "pid": 5, "tid": 7, "ts": 1716454222355408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293213, "dur": 11,
    "args": {
      "External id": 64992,
      "cbid": 211, "correlation": 64992
    }
  },
  {
    "ph": "s", "id": 64992, "pid": 76337, "tid": -914061504, "ts": 1716454222293213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222355423, "dur": 4,
    "args": {
      "External id": 65004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65004,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65004, "pid": 5, "tid": 7, "ts": 1716454222355423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293234, "dur": 6,
    "args": {
      "External id": 65004,
      "cbid": 211, "correlation": 65004
    }
  },
  {
    "ph": "s", "id": 65004, "pid": 76337, "tid": -914061504, "ts": 1716454222293234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222355429, "dur": 12,
    "args": {
      "External id": 65007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65007,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65007, "pid": 5, "tid": 7, "ts": 1716454222355429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293252, "dur": 6,
    "args": {
      "External id": 65007,
      "cbid": 211, "correlation": 65007
    }
  },
  {
    "ph": "s", "id": 65007, "pid": 76337, "tid": -914061504, "ts": 1716454222293252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222355443, "dur": 7,
    "args": {
      "External id": 65016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65016,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65016, "pid": 5, "tid": 7, "ts": 1716454222355443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293293, "dur": 10,
    "args": {
      "External id": 65016,
      "cbid": 211, "correlation": 65016
    }
  },
  {
    "ph": "s", "id": 65016, "pid": 76337, "tid": -914061504, "ts": 1716454222293293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293346, "dur": 0,
    "args": {
      "External id": 65026,
      "cbid": 317, "correlation": 65026
    }
  },
  {
    "ph": "f", "id": 65026, "pid": 76337, "tid": -914061504, "ts": 1716454222293346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293347, "dur": 0,
    "args": {
      "External id": 65027,
      "cbid": 203, "correlation": 65027
    }
  },
  {
    "ph": "f", "id": 65027, "pid": 76337, "tid": -914061504, "ts": 1716454222293347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293347, "dur": 0,
    "args": {
      "External id": 65028,
      "cbid": 205, "correlation": 65028
    }
  },
  {
    "ph": "f", "id": 65028, "pid": 76337, "tid": -914061504, "ts": 1716454222293347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222355451, "dur": 7,
    "args": {
      "External id": 65032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65032,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65032, "pid": 5, "tid": 7, "ts": 1716454222355451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293361, "dur": 11,
    "args": {
      "External id": 65032,
      "cbid": 211, "correlation": 65032
    }
  },
  {
    "ph": "s", "id": 65032, "pid": 76337, "tid": -914061504, "ts": 1716454222293361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222355459, "dur": 349,
    "args": {
      "External id": 65034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65034,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65034, "pid": 5, "tid": 7, "ts": 1716454222355459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293375, "dur": 5,
    "args": {
      "External id": 65034,
      "cbid": 211, "correlation": 65034
    }
  },
  {
    "ph": "s", "id": 65034, "pid": 76337, "tid": -914061504, "ts": 1716454222293375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222355811, "dur": 1,
    "args": {
      "External id": 65036,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65036,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 65036, "pid": 5, "tid": 7, "ts": 1716454222355811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293386, "dur": 6,
    "args": {
      "External id": 65036,
      "cbid": 51, "correlation": 65036
    }
  },
  {
    "ph": "s", "id": 65036, "pid": 76337, "tid": -914061504, "ts": 1716454222293386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222355815, "dur": 546,
    "args": {
      "External id": 65037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65037,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65037, "pid": 5, "tid": 7, "ts": 1716454222355815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293393, "dur": 6,
    "args": {
      "External id": 65037,
      "cbid": 211, "correlation": 65037
    }
  },
  {
    "ph": "s", "id": 65037, "pid": 76337, "tid": -914061504, "ts": 1716454222293393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222356362, "dur": 6,
    "args": {
      "External id": 65039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65039,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65039, "pid": 5, "tid": 7, "ts": 1716454222356362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293403, "dur": 5,
    "args": {
      "External id": 65039,
      "cbid": 211, "correlation": 65039
    }
  },
  {
    "ph": "s", "id": 65039, "pid": 76337, "tid": -914061504, "ts": 1716454222293403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222356369, "dur": 7,
    "args": {
      "External id": 65045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65045, "pid": 5, "tid": 7, "ts": 1716454222356369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293431, "dur": 9,
    "args": {
      "External id": 65045,
      "cbid": 211, "correlation": 65045
    }
  },
  {
    "ph": "s", "id": 65045, "pid": 76337, "tid": -914061504, "ts": 1716454222293431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222356378, "dur": 3,
    "args": {
      "External id": 65053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65053,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 65053, "pid": 5, "tid": 7, "ts": 1716454222356378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293474, "dur": 9,
    "args": {
      "External id": 65053,
      "cbid": 211, "correlation": 65053
    }
  },
  {
    "ph": "s", "id": 65053, "pid": 76337, "tid": -914061504, "ts": 1716454222293474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293535, "dur": 1,
    "args": {
      "External id": 65069,
      "cbid": 251, "correlation": 65069
    }
  },
  {
    "ph": "f", "id": 65069, "pid": 76337, "tid": -914061504, "ts": 1716454222293535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293541, "dur": 0,
    "args": {
      "External id": 65071,
      "cbid": 251, "correlation": 65071
    }
  },
  {
    "ph": "f", "id": 65071, "pid": 76337, "tid": -914061504, "ts": 1716454222293541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222356382, "dur": 14,
    "args": {
      "External id": 65072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65072,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65072, "pid": 5, "tid": 7, "ts": 1716454222356382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293543, "dur": 11,
    "args": {
      "External id": 65072,
      "cbid": 211, "correlation": 65072
    }
  },
  {
    "ph": "s", "id": 65072, "pid": 76337, "tid": -914061504, "ts": 1716454222293543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222356397, "dur": 5,
    "args": {
      "External id": 65074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65074,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65074, "pid": 5, "tid": 7, "ts": 1716454222356397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293555, "dur": 8,
    "args": {
      "External id": 65074,
      "cbid": 211, "correlation": 65074
    }
  },
  {
    "ph": "s", "id": 65074, "pid": 76337, "tid": -914061504, "ts": 1716454222293555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222356404, "dur": 7,
    "args": {
      "External id": 65084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65084,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65084, "pid": 5, "tid": 7, "ts": 1716454222356404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293616, "dur": 12,
    "args": {
      "External id": 65084,
      "cbid": 211, "correlation": 65084
    }
  },
  {
    "ph": "s", "id": 65084, "pid": 76337, "tid": -914061504, "ts": 1716454222293616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222356412, "dur": 11,
    "args": {
      "External id": 65104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65104,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 65104, "pid": 5, "tid": 7, "ts": 1716454222356412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293681, "dur": 11,
    "args": {
      "External id": 65104,
      "cbid": 211, "correlation": 65104
    }
  },
  {
    "ph": "s", "id": 65104, "pid": 76337, "tid": -914061504, "ts": 1716454222293681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222356424, "dur": 4,
    "args": {
      "External id": 65116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65116,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 65116, "pid": 5, "tid": 7, "ts": 1716454222356424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293702, "dur": 6,
    "args": {
      "External id": 65116,
      "cbid": 211, "correlation": 65116
    }
  },
  {
    "ph": "s", "id": 65116, "pid": 76337, "tid": -914061504, "ts": 1716454222293702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222356430, "dur": 7,
    "args": {
      "External id": 65119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65119, "pid": 5, "tid": 7, "ts": 1716454222356430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293720, "dur": 6,
    "args": {
      "External id": 65119,
      "cbid": 211, "correlation": 65119
    }
  },
  {
    "ph": "s", "id": 65119, "pid": 76337, "tid": -914061504, "ts": 1716454222293720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222356438, "dur": 5,
    "args": {
      "External id": 65128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65128,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65128, "pid": 5, "tid": 7, "ts": 1716454222356438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293759, "dur": 9,
    "args": {
      "External id": 65128,
      "cbid": 211, "correlation": 65128
    }
  },
  {
    "ph": "s", "id": 65128, "pid": 76337, "tid": -914061504, "ts": 1716454222293759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293821, "dur": 0,
    "args": {
      "External id": 65138,
      "cbid": 317, "correlation": 65138
    }
  },
  {
    "ph": "f", "id": 65138, "pid": 76337, "tid": -914061504, "ts": 1716454222293821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293821, "dur": 0,
    "args": {
      "External id": 65139,
      "cbid": 203, "correlation": 65139
    }
  },
  {
    "ph": "f", "id": 65139, "pid": 76337, "tid": -914061504, "ts": 1716454222293821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293822, "dur": 0,
    "args": {
      "External id": 65140,
      "cbid": 205, "correlation": 65140
    }
  },
  {
    "ph": "f", "id": 65140, "pid": 76337, "tid": -914061504, "ts": 1716454222293822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222356445, "dur": 6,
    "args": {
      "External id": 65144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65144,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65144, "pid": 5, "tid": 7, "ts": 1716454222356445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293836, "dur": 12,
    "args": {
      "External id": 65144,
      "cbid": 211, "correlation": 65144
    }
  },
  {
    "ph": "s", "id": 65144, "pid": 76337, "tid": -914061504, "ts": 1716454222293836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222356452, "dur": 176,
    "args": {
      "External id": 65146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65146,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65146, "pid": 5, "tid": 7, "ts": 1716454222356452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293850, "dur": 8,
    "args": {
      "External id": 65146,
      "cbid": 211, "correlation": 65146
    }
  },
  {
    "ph": "s", "id": 65146, "pid": 76337, "tid": -914061504, "ts": 1716454222293850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222356630, "dur": 1,
    "args": {
      "External id": 65148,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65148,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 65148, "pid": 5, "tid": 7, "ts": 1716454222356630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293864, "dur": 6,
    "args": {
      "External id": 65148,
      "cbid": 51, "correlation": 65148
    }
  },
  {
    "ph": "s", "id": 65148, "pid": 76337, "tid": -914061504, "ts": 1716454222293864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222356634, "dur": 286,
    "args": {
      "External id": 65149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65149,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65149, "pid": 5, "tid": 7, "ts": 1716454222356634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293871, "dur": 6,
    "args": {
      "External id": 65149,
      "cbid": 211, "correlation": 65149
    }
  },
  {
    "ph": "s", "id": 65149, "pid": 76337, "tid": -914061504, "ts": 1716454222293871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222356921, "dur": 6,
    "args": {
      "External id": 65151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65151,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65151, "pid": 5, "tid": 7, "ts": 1716454222356921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293881, "dur": 5,
    "args": {
      "External id": 65151,
      "cbid": 211, "correlation": 65151
    }
  },
  {
    "ph": "s", "id": 65151, "pid": 76337, "tid": -914061504, "ts": 1716454222293881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222356929, "dur": 7,
    "args": {
      "External id": 65157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65157,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65157, "pid": 5, "tid": 7, "ts": 1716454222356929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293910, "dur": 8,
    "args": {
      "External id": 65157,
      "cbid": 211, "correlation": 65157
    }
  },
  {
    "ph": "s", "id": 65157, "pid": 76337, "tid": -914061504, "ts": 1716454222293910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293968, "dur": 0,
    "args": {
      "External id": 65167,
      "cbid": 317, "correlation": 65167
    }
  },
  {
    "ph": "f", "id": 65167, "pid": 76337, "tid": -914061504, "ts": 1716454222293968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293969, "dur": 0,
    "args": {
      "External id": 65168,
      "cbid": 203, "correlation": 65168
    }
  },
  {
    "ph": "f", "id": 65168, "pid": 76337, "tid": -914061504, "ts": 1716454222293969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293970, "dur": 0,
    "args": {
      "External id": 65169,
      "cbid": 205, "correlation": 65169
    }
  },
  {
    "ph": "f", "id": 65169, "pid": 76337, "tid": -914061504, "ts": 1716454222293970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222356937, "dur": 8,
    "args": {
      "External id": 65173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65173, "pid": 5, "tid": 7, "ts": 1716454222356937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222293991, "dur": 12,
    "args": {
      "External id": 65173,
      "cbid": 211, "correlation": 65173
    }
  },
  {
    "ph": "s", "id": 65173, "pid": 76337, "tid": -914061504, "ts": 1716454222293991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222356947, "dur": 3,
    "args": {
      "External id": 65175,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65175,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3888888888888888
    }
  },
  {
    "ph": "f", "id": 65175, "pid": 5, "tid": 7, "ts": 1716454222356947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294009, "dur": 11,
    "args": {
      "External id": 65175,
      "cbid": 51, "correlation": 65175
    }
  },
  {
    "ph": "s", "id": 65175, "pid": 76337, "tid": -914061504, "ts": 1716454222294009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222356951, "dur": 100,
    "args": {
      "External id": 65176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65176,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 65176, "pid": 5, "tid": 7, "ts": 1716454222356951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294020, "dur": 6,
    "args": {
      "External id": 65176,
      "cbid": 211, "correlation": 65176
    }
  },
  {
    "ph": "s", "id": 65176, "pid": 76337, "tid": -914061504, "ts": 1716454222294020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222357053, "dur": 6,
    "args": {
      "External id": 65178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65178, "pid": 5, "tid": 7, "ts": 1716454222357053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294030, "dur": 5,
    "args": {
      "External id": 65178,
      "cbid": 211, "correlation": 65178
    }
  },
  {
    "ph": "s", "id": 65178, "pid": 76337, "tid": -914061504, "ts": 1716454222294030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222357061, "dur": 7,
    "args": {
      "External id": 65184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65184, "pid": 5, "tid": 7, "ts": 1716454222357061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294057, "dur": 9,
    "args": {
      "External id": 65184,
      "cbid": 211, "correlation": 65184
    }
  },
  {
    "ph": "s", "id": 65184, "pid": 76337, "tid": -914061504, "ts": 1716454222294057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222357069, "dur": 5,
    "args": {
      "External id": 65192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65192,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65192, "pid": 5, "tid": 7, "ts": 1716454222357069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294087, "dur": 8,
    "args": {
      "External id": 65192,
      "cbid": 211, "correlation": 65192
    }
  },
  {
    "ph": "s", "id": 65192, "pid": 76337, "tid": -914061504, "ts": 1716454222294087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222357076, "dur": 5,
    "args": {
      "External id": 65200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65200,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65200, "pid": 5, "tid": 7, "ts": 1716454222357076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294115, "dur": 11,
    "args": {
      "External id": 65200,
      "cbid": 211, "correlation": 65200
    }
  },
  {
    "ph": "s", "id": 65200, "pid": 76337, "tid": -914061504, "ts": 1716454222294115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222357082, "dur": 12,
    "args": {
      "External id": 65209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65209,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65209, "pid": 5, "tid": 7, "ts": 1716454222357082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294162, "dur": 11,
    "args": {
      "External id": 65209,
      "cbid": 211, "correlation": 65209
    }
  },
  {
    "ph": "s", "id": 65209, "pid": 76337, "tid": -914061504, "ts": 1716454222294162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222357095, "dur": 13,
    "args": {
      "External id": 65229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65229,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 65229, "pid": 5, "tid": 7, "ts": 1716454222357095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294231, "dur": 12,
    "args": {
      "External id": 65229,
      "cbid": 211, "correlation": 65229
    }
  },
  {
    "ph": "s", "id": 65229, "pid": 76337, "tid": -914061504, "ts": 1716454222294231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222357110, "dur": 4,
    "args": {
      "External id": 65241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65241,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65241, "pid": 5, "tid": 7, "ts": 1716454222357110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294253, "dur": 6,
    "args": {
      "External id": 65241,
      "cbid": 211, "correlation": 65241
    }
  },
  {
    "ph": "s", "id": 65241, "pid": 76337, "tid": -914061504, "ts": 1716454222294253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222357115, "dur": 12,
    "args": {
      "External id": 65244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65244, "pid": 5, "tid": 7, "ts": 1716454222357115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294271, "dur": 7,
    "args": {
      "External id": 65244,
      "cbid": 211, "correlation": 65244
    }
  },
  {
    "ph": "s", "id": 65244, "pid": 76337, "tid": -914061504, "ts": 1716454222294271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222357128, "dur": 7,
    "args": {
      "External id": 65253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65253,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65253, "pid": 5, "tid": 7, "ts": 1716454222357128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294309, "dur": 9,
    "args": {
      "External id": 65253,
      "cbid": 211, "correlation": 65253
    }
  },
  {
    "ph": "s", "id": 65253, "pid": 76337, "tid": -914061504, "ts": 1716454222294309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294361, "dur": 0,
    "args": {
      "External id": 65263,
      "cbid": 317, "correlation": 65263
    }
  },
  {
    "ph": "f", "id": 65263, "pid": 76337, "tid": -914061504, "ts": 1716454222294361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294362, "dur": 0,
    "args": {
      "External id": 65264,
      "cbid": 203, "correlation": 65264
    }
  },
  {
    "ph": "f", "id": 65264, "pid": 76337, "tid": -914061504, "ts": 1716454222294362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294362, "dur": 0,
    "args": {
      "External id": 65265,
      "cbid": 205, "correlation": 65265
    }
  },
  {
    "ph": "f", "id": 65265, "pid": 76337, "tid": -914061504, "ts": 1716454222294362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222357137, "dur": 7,
    "args": {
      "External id": 65269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65269,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65269, "pid": 5, "tid": 7, "ts": 1716454222357137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294397, "dur": 12,
    "args": {
      "External id": 65269,
      "cbid": 211, "correlation": 65269
    }
  },
  {
    "ph": "s", "id": 65269, "pid": 76337, "tid": -914061504, "ts": 1716454222294397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222357145, "dur": 348,
    "args": {
      "External id": 65271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65271, "pid": 5, "tid": 7, "ts": 1716454222357145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294412, "dur": 6,
    "args": {
      "External id": 65271,
      "cbid": 211, "correlation": 65271
    }
  },
  {
    "ph": "s", "id": 65271, "pid": 76337, "tid": -914061504, "ts": 1716454222294412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222357495, "dur": 1,
    "args": {
      "External id": 65273,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65273,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 65273, "pid": 5, "tid": 7, "ts": 1716454222357495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294423, "dur": 6,
    "args": {
      "External id": 65273,
      "cbid": 51, "correlation": 65273
    }
  },
  {
    "ph": "s", "id": 65273, "pid": 76337, "tid": -914061504, "ts": 1716454222294423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222357499, "dur": 546,
    "args": {
      "External id": 65274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65274,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65274, "pid": 5, "tid": 7, "ts": 1716454222357499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294431, "dur": 7,
    "args": {
      "External id": 65274,
      "cbid": 211, "correlation": 65274
    }
  },
  {
    "ph": "s", "id": 65274, "pid": 76337, "tid": -914061504, "ts": 1716454222294431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358047, "dur": 6,
    "args": {
      "External id": 65276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65276,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65276, "pid": 5, "tid": 7, "ts": 1716454222358047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294442, "dur": 5,
    "args": {
      "External id": 65276,
      "cbid": 211, "correlation": 65276
    }
  },
  {
    "ph": "s", "id": 65276, "pid": 76337, "tid": -914061504, "ts": 1716454222294442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222358054, "dur": 7,
    "args": {
      "External id": 65282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65282, "pid": 5, "tid": 7, "ts": 1716454222358054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294471, "dur": 8,
    "args": {
      "External id": 65282,
      "cbid": 211, "correlation": 65282
    }
  },
  {
    "ph": "s", "id": 65282, "pid": 76337, "tid": -914061504, "ts": 1716454222294471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222358062, "dur": 3,
    "args": {
      "External id": 65290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65290,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 65290, "pid": 5, "tid": 7, "ts": 1716454222358062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294515, "dur": 9,
    "args": {
      "External id": 65290,
      "cbid": 211, "correlation": 65290
    }
  },
  {
    "ph": "s", "id": 65290, "pid": 76337, "tid": -914061504, "ts": 1716454222294515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294578, "dur": 1,
    "args": {
      "External id": 65306,
      "cbid": 251, "correlation": 65306
    }
  },
  {
    "ph": "f", "id": 65306, "pid": 76337, "tid": -914061504, "ts": 1716454222294578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294583, "dur": 0,
    "args": {
      "External id": 65308,
      "cbid": 251, "correlation": 65308
    }
  },
  {
    "ph": "f", "id": 65308, "pid": 76337, "tid": -914061504, "ts": 1716454222294583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222358067, "dur": 13,
    "args": {
      "External id": 65309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65309,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65309, "pid": 5, "tid": 7, "ts": 1716454222358067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294585, "dur": 12,
    "args": {
      "External id": 65309,
      "cbid": 211, "correlation": 65309
    }
  },
  {
    "ph": "s", "id": 65309, "pid": 76337, "tid": -914061504, "ts": 1716454222294585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222358082, "dur": 5,
    "args": {
      "External id": 65311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65311,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65311, "pid": 5, "tid": 7, "ts": 1716454222358082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294598, "dur": 5,
    "args": {
      "External id": 65311,
      "cbid": 211, "correlation": 65311
    }
  },
  {
    "ph": "s", "id": 65311, "pid": 76337, "tid": -914061504, "ts": 1716454222294598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222358089, "dur": 6,
    "args": {
      "External id": 65321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65321, "pid": 5, "tid": 7, "ts": 1716454222358089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294656, "dur": 12,
    "args": {
      "External id": 65321,
      "cbid": 211, "correlation": 65321
    }
  },
  {
    "ph": "s", "id": 65321, "pid": 76337, "tid": -914061504, "ts": 1716454222294656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358097, "dur": 10,
    "args": {
      "External id": 65341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65341,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 65341, "pid": 5, "tid": 7, "ts": 1716454222358097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294721, "dur": 14,
    "args": {
      "External id": 65341,
      "cbid": 211, "correlation": 65341
    }
  },
  {
    "ph": "s", "id": 65341, "pid": 76337, "tid": -914061504, "ts": 1716454222294721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222358109, "dur": 4,
    "args": {
      "External id": 65353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65353,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 65353, "pid": 5, "tid": 7, "ts": 1716454222358109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294745, "dur": 7,
    "args": {
      "External id": 65353,
      "cbid": 211, "correlation": 65353
    }
  },
  {
    "ph": "s", "id": 65353, "pid": 76337, "tid": -914061504, "ts": 1716454222294745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222358114, "dur": 8,
    "args": {
      "External id": 65356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65356,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65356, "pid": 5, "tid": 7, "ts": 1716454222358114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294764, "dur": 7,
    "args": {
      "External id": 65356,
      "cbid": 211, "correlation": 65356
    }
  },
  {
    "ph": "s", "id": 65356, "pid": 76337, "tid": -914061504, "ts": 1716454222294764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222358123, "dur": 5,
    "args": {
      "External id": 65365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65365,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65365, "pid": 5, "tid": 7, "ts": 1716454222358123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294805, "dur": 10,
    "args": {
      "External id": 65365,
      "cbid": 211, "correlation": 65365
    }
  },
  {
    "ph": "s", "id": 65365, "pid": 76337, "tid": -914061504, "ts": 1716454222294805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294868, "dur": 0,
    "args": {
      "External id": 65375,
      "cbid": 317, "correlation": 65375
    }
  },
  {
    "ph": "f", "id": 65375, "pid": 76337, "tid": -914061504, "ts": 1716454222294868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294868, "dur": 0,
    "args": {
      "External id": 65376,
      "cbid": 203, "correlation": 65376
    }
  },
  {
    "ph": "f", "id": 65376, "pid": 76337, "tid": -914061504, "ts": 1716454222294868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294869, "dur": 0,
    "args": {
      "External id": 65377,
      "cbid": 205, "correlation": 65377
    }
  },
  {
    "ph": "f", "id": 65377, "pid": 76337, "tid": -914061504, "ts": 1716454222294869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358130, "dur": 6,
    "args": {
      "External id": 65381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65381,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65381, "pid": 5, "tid": 7, "ts": 1716454222358130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294883, "dur": 12,
    "args": {
      "External id": 65381,
      "cbid": 211, "correlation": 65381
    }
  },
  {
    "ph": "s", "id": 65381, "pid": 76337, "tid": -914061504, "ts": 1716454222294883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358137, "dur": 177,
    "args": {
      "External id": 65383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65383,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65383, "pid": 5, "tid": 7, "ts": 1716454222358137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294897, "dur": 5,
    "args": {
      "External id": 65383,
      "cbid": 211, "correlation": 65383
    }
  },
  {
    "ph": "s", "id": 65383, "pid": 76337, "tid": -914061504, "ts": 1716454222294897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222358316, "dur": 1,
    "args": {
      "External id": 65385,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65385,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 65385, "pid": 5, "tid": 7, "ts": 1716454222358316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294908, "dur": 6,
    "args": {
      "External id": 65385,
      "cbid": 51, "correlation": 65385
    }
  },
  {
    "ph": "s", "id": 65385, "pid": 76337, "tid": -914061504, "ts": 1716454222294908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222358320, "dur": 285,
    "args": {
      "External id": 65386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65386,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65386, "pid": 5, "tid": 7, "ts": 1716454222358320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294916, "dur": 6,
    "args": {
      "External id": 65386,
      "cbid": 211, "correlation": 65386
    }
  },
  {
    "ph": "s", "id": 65386, "pid": 76337, "tid": -914061504, "ts": 1716454222294916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358606, "dur": 6,
    "args": {
      "External id": 65388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65388,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65388, "pid": 5, "tid": 7, "ts": 1716454222358606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294925, "dur": 5,
    "args": {
      "External id": 65388,
      "cbid": 211, "correlation": 65388
    }
  },
  {
    "ph": "s", "id": 65388, "pid": 76337, "tid": -914061504, "ts": 1716454222294925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222358614, "dur": 7,
    "args": {
      "External id": 65394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65394, "pid": 5, "tid": 7, "ts": 1716454222358614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222294953, "dur": 9,
    "args": {
      "External id": 65394,
      "cbid": 211, "correlation": 65394
    }
  },
  {
    "ph": "s", "id": 65394, "pid": 76337, "tid": -914061504, "ts": 1716454222294953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295023, "dur": 0,
    "args": {
      "External id": 65404,
      "cbid": 317, "correlation": 65404
    }
  },
  {
    "ph": "f", "id": 65404, "pid": 76337, "tid": -914061504, "ts": 1716454222295023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295024, "dur": 0,
    "args": {
      "External id": 65405,
      "cbid": 203, "correlation": 65405
    }
  },
  {
    "ph": "f", "id": 65405, "pid": 76337, "tid": -914061504, "ts": 1716454222295024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295024, "dur": 0,
    "args": {
      "External id": 65406,
      "cbid": 205, "correlation": 65406
    }
  },
  {
    "ph": "f", "id": 65406, "pid": 76337, "tid": -914061504, "ts": 1716454222295024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358622, "dur": 9,
    "args": {
      "External id": 65410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65410, "pid": 5, "tid": 7, "ts": 1716454222358622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295036, "dur": 12,
    "args": {
      "External id": 65410,
      "cbid": 211, "correlation": 65410
    }
  },
  {
    "ph": "s", "id": 65410, "pid": 76337, "tid": -914061504, "ts": 1716454222295036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222358632, "dur": 3,
    "args": {
      "External id": 65412,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65412,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4018691588785046
    }
  },
  {
    "ph": "f", "id": 65412, "pid": 5, "tid": 7, "ts": 1716454222358632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295054, "dur": 10,
    "args": {
      "External id": 65412,
      "cbid": 51, "correlation": 65412
    }
  },
  {
    "ph": "s", "id": 65412, "pid": 76337, "tid": -914061504, "ts": 1716454222295054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222358636, "dur": 101,
    "args": {
      "External id": 65413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65413,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 65413, "pid": 5, "tid": 7, "ts": 1716454222358636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295065, "dur": 6,
    "args": {
      "External id": 65413,
      "cbid": 211, "correlation": 65413
    }
  },
  {
    "ph": "s", "id": 65413, "pid": 76337, "tid": -914061504, "ts": 1716454222295065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358739, "dur": 6,
    "args": {
      "External id": 65415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65415,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65415, "pid": 5, "tid": 7, "ts": 1716454222358739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295074, "dur": 5,
    "args": {
      "External id": 65415,
      "cbid": 211, "correlation": 65415
    }
  },
  {
    "ph": "s", "id": 65415, "pid": 76337, "tid": -914061504, "ts": 1716454222295074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222358746, "dur": 7,
    "args": {
      "External id": 65421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65421, "pid": 5, "tid": 7, "ts": 1716454222358746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295101, "dur": 10,
    "args": {
      "External id": 65421,
      "cbid": 211, "correlation": 65421
    }
  },
  {
    "ph": "s", "id": 65421, "pid": 76337, "tid": -914061504, "ts": 1716454222295101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222358754, "dur": 5,
    "args": {
      "External id": 65429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65429,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65429, "pid": 5, "tid": 7, "ts": 1716454222358754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295131, "dur": 8,
    "args": {
      "External id": 65429,
      "cbid": 211, "correlation": 65429
    }
  },
  {
    "ph": "s", "id": 65429, "pid": 76337, "tid": -914061504, "ts": 1716454222295131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222358761, "dur": 5,
    "args": {
      "External id": 65437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65437,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 65437, "pid": 5, "tid": 7, "ts": 1716454222358761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295161, "dur": 8,
    "args": {
      "External id": 65437,
      "cbid": 211, "correlation": 65437
    }
  },
  {
    "ph": "s", "id": 65437, "pid": 76337, "tid": -914061504, "ts": 1716454222295161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222358767, "dur": 16,
    "args": {
      "External id": 65448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65448,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65448, "pid": 5, "tid": 7, "ts": 1716454222358767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295240, "dur": 17,
    "args": {
      "External id": 65448,
      "cbid": 211, "correlation": 65448
    }
  },
  {
    "ph": "s", "id": 65448, "pid": 76337, "tid": -914061504, "ts": 1716454222295240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295300, "dur": 0,
    "args": {
      "External id": 65458,
      "cbid": 317, "correlation": 65458
    }
  },
  {
    "ph": "f", "id": 65458, "pid": 76337, "tid": -914061504, "ts": 1716454222295300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295301, "dur": 0,
    "args": {
      "External id": 65459,
      "cbid": 203, "correlation": 65459
    }
  },
  {
    "ph": "f", "id": 65459, "pid": 76337, "tid": -914061504, "ts": 1716454222295301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295301, "dur": 0,
    "args": {
      "External id": 65460,
      "cbid": 205, "correlation": 65460
    }
  },
  {
    "ph": "f", "id": 65460, "pid": 76337, "tid": -914061504, "ts": 1716454222295301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358784, "dur": 9,
    "args": {
      "External id": 65464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65464,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65464, "pid": 5, "tid": 7, "ts": 1716454222358784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295316, "dur": 11,
    "args": {
      "External id": 65464,
      "cbid": 211, "correlation": 65464
    }
  },
  {
    "ph": "s", "id": 65464, "pid": 76337, "tid": -914061504, "ts": 1716454222295316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222358795, "dur": 177,
    "args": {
      "External id": 65466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65466,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65466, "pid": 5, "tid": 7, "ts": 1716454222358795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295330, "dur": 8,
    "args": {
      "External id": 65466,
      "cbid": 211, "correlation": 65466
    }
  },
  {
    "ph": "s", "id": 65466, "pid": 76337, "tid": -914061504, "ts": 1716454222295330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222358974, "dur": 2,
    "args": {
      "External id": 65468,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65468,
      "bytes": 960, "memory bandwidth (GB/s)": 0.47619047619047616
    }
  },
  {
    "ph": "f", "id": 65468, "pid": 5, "tid": 7, "ts": 1716454222358974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295343, "dur": 6,
    "args": {
      "External id": 65468,
      "cbid": 51, "correlation": 65468
    }
  },
  {
    "ph": "s", "id": 65468, "pid": 76337, "tid": -914061504, "ts": 1716454222295343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222358979, "dur": 712,
    "args": {
      "External id": 65469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65469,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65469, "pid": 5, "tid": 7, "ts": 1716454222358979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295351, "dur": 7,
    "args": {
      "External id": 65469,
      "cbid": 211, "correlation": 65469
    }
  },
  {
    "ph": "s", "id": 65469, "pid": 76337, "tid": -914061504, "ts": 1716454222295351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222359692, "dur": 14,
    "args": {
      "External id": 65471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65471,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65471, "pid": 5, "tid": 7, "ts": 1716454222359692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295362, "dur": 5,
    "args": {
      "External id": 65471,
      "cbid": 211, "correlation": 65471
    }
  },
  {
    "ph": "s", "id": 65471, "pid": 76337, "tid": -914061504, "ts": 1716454222295362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222359707, "dur": 16,
    "args": {
      "External id": 65477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65477, "pid": 5, "tid": 7, "ts": 1716454222359707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295390, "dur": 8,
    "args": {
      "External id": 65477,
      "cbid": 211, "correlation": 65477
    }
  },
  {
    "ph": "s", "id": 65477, "pid": 76337, "tid": -914061504, "ts": 1716454222295390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222359724, "dur": 31,
    "args": {
      "External id": 65486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65486,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65486, "pid": 5, "tid": 7, "ts": 1716454222359724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295482, "dur": 13,
    "args": {
      "External id": 65486,
      "cbid": 211, "correlation": 65486
    }
  },
  {
    "ph": "s", "id": 65486, "pid": 76337, "tid": -914061504, "ts": 1716454222295482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222359756, "dur": 33,
    "args": {
      "External id": 65506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65506,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 65506, "pid": 5, "tid": 7, "ts": 1716454222359756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295551, "dur": 11,
    "args": {
      "External id": 65506,
      "cbid": 211, "correlation": 65506
    }
  },
  {
    "ph": "s", "id": 65506, "pid": 76337, "tid": -914061504, "ts": 1716454222295551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222359791, "dur": 5,
    "args": {
      "External id": 65518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65518,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65518, "pid": 5, "tid": 7, "ts": 1716454222359791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295572, "dur": 6,
    "args": {
      "External id": 65518,
      "cbid": 211, "correlation": 65518
    }
  },
  {
    "ph": "s", "id": 65518, "pid": 76337, "tid": -914061504, "ts": 1716454222295572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222359797, "dur": 32,
    "args": {
      "External id": 65521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65521, "pid": 5, "tid": 7, "ts": 1716454222359797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295590, "dur": 6,
    "args": {
      "External id": 65521,
      "cbid": 211, "correlation": 65521
    }
  },
  {
    "ph": "s", "id": 65521, "pid": 76337, "tid": -914061504, "ts": 1716454222295590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222359831, "dur": 20,
    "args": {
      "External id": 65530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65530,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65530, "pid": 5, "tid": 7, "ts": 1716454222359831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295628, "dur": 12,
    "args": {
      "External id": 65530,
      "cbid": 211, "correlation": 65530
    }
  },
  {
    "ph": "s", "id": 65530, "pid": 76337, "tid": -914061504, "ts": 1716454222295628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295684, "dur": 0,
    "args": {
      "External id": 65540,
      "cbid": 317, "correlation": 65540
    }
  },
  {
    "ph": "f", "id": 65540, "pid": 76337, "tid": -914061504, "ts": 1716454222295684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295684, "dur": 0,
    "args": {
      "External id": 65541,
      "cbid": 203, "correlation": 65541
    }
  },
  {
    "ph": "f", "id": 65541, "pid": 76337, "tid": -914061504, "ts": 1716454222295684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295685, "dur": 0,
    "args": {
      "External id": 65542,
      "cbid": 205, "correlation": 65542
    }
  },
  {
    "ph": "f", "id": 65542, "pid": 76337, "tid": -914061504, "ts": 1716454222295685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222359853, "dur": 24,
    "args": {
      "External id": 65546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65546, "pid": 5, "tid": 7, "ts": 1716454222359853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295703, "dur": 11,
    "args": {
      "External id": 65546,
      "cbid": 211, "correlation": 65546
    }
  },
  {
    "ph": "s", "id": 65546, "pid": 76337, "tid": -914061504, "ts": 1716454222295703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222359877, "dur": 349,
    "args": {
      "External id": 65548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65548,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65548, "pid": 5, "tid": 7, "ts": 1716454222359877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295717, "dur": 6,
    "args": {
      "External id": 65548,
      "cbid": 211, "correlation": 65548
    }
  },
  {
    "ph": "s", "id": 65548, "pid": 76337, "tid": -914061504, "ts": 1716454222295717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222360229, "dur": 1,
    "args": {
      "External id": 65550,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65550,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5660377358490566
    }
  },
  {
    "ph": "f", "id": 65550, "pid": 5, "tid": 7, "ts": 1716454222360229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295729, "dur": 6,
    "args": {
      "External id": 65550,
      "cbid": 51, "correlation": 65550
    }
  },
  {
    "ph": "s", "id": 65550, "pid": 76337, "tid": -914061504, "ts": 1716454222295729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222360233, "dur": 1361,
    "args": {
      "External id": 65551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65551,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65551, "pid": 5, "tid": 7, "ts": 1716454222360233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295736, "dur": 6,
    "args": {
      "External id": 65551,
      "cbid": 211, "correlation": 65551
    }
  },
  {
    "ph": "s", "id": 65551, "pid": 76337, "tid": -914061504, "ts": 1716454222295736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222361595, "dur": 13,
    "args": {
      "External id": 65553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65553, "pid": 5, "tid": 7, "ts": 1716454222361595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295746, "dur": 5,
    "args": {
      "External id": 65553,
      "cbid": 211, "correlation": 65553
    }
  },
  {
    "ph": "s", "id": 65553, "pid": 76337, "tid": -914061504, "ts": 1716454222295746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222361610, "dur": 16,
    "args": {
      "External id": 65559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65559, "pid": 5, "tid": 7, "ts": 1716454222361610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295774, "dur": 9,
    "args": {
      "External id": 65559,
      "cbid": 211, "correlation": 65559
    }
  },
  {
    "ph": "s", "id": 65559, "pid": 76337, "tid": -914061504, "ts": 1716454222295774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222361627, "dur": 4,
    "args": {
      "External id": 65567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65567,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 65567, "pid": 5, "tid": 7, "ts": 1716454222361627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295818, "dur": 9,
    "args": {
      "External id": 65567,
      "cbid": 211, "correlation": 65567
    }
  },
  {
    "ph": "s", "id": 65567, "pid": 76337, "tid": -914061504, "ts": 1716454222295818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295882, "dur": 1,
    "args": {
      "External id": 65583,
      "cbid": 251, "correlation": 65583
    }
  },
  {
    "ph": "f", "id": 65583, "pid": 76337, "tid": -914061504, "ts": 1716454222295882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295888, "dur": 0,
    "args": {
      "External id": 65585,
      "cbid": 251, "correlation": 65585
    }
  },
  {
    "ph": "f", "id": 65585, "pid": 76337, "tid": -914061504, "ts": 1716454222295888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222361632, "dur": 13,
    "args": {
      "External id": 65586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65586,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65586, "pid": 5, "tid": 7, "ts": 1716454222361632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295890, "dur": 11,
    "args": {
      "External id": 65586,
      "cbid": 211, "correlation": 65586
    }
  },
  {
    "ph": "s", "id": 65586, "pid": 76337, "tid": -914061504, "ts": 1716454222295890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222361647, "dur": 5,
    "args": {
      "External id": 65588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65588,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65588, "pid": 5, "tid": 7, "ts": 1716454222361647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295903, "dur": 5,
    "args": {
      "External id": 65588,
      "cbid": 211, "correlation": 65588
    }
  },
  {
    "ph": "s", "id": 65588, "pid": 76337, "tid": -914061504, "ts": 1716454222295903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222361653, "dur": 18,
    "args": {
      "External id": 65598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65598, "pid": 5, "tid": 7, "ts": 1716454222361653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222295964, "dur": 20,
    "args": {
      "External id": 65598,
      "cbid": 211, "correlation": 65598
    }
  },
  {
    "ph": "s", "id": 65598, "pid": 76337, "tid": -914061504, "ts": 1716454222295964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222361673, "dur": 19,
    "args": {
      "External id": 65618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65618,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 65618, "pid": 5, "tid": 7, "ts": 1716454222361673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296039, "dur": 11,
    "args": {
      "External id": 65618,
      "cbid": 211, "correlation": 65618
    }
  },
  {
    "ph": "s", "id": 65618, "pid": 76337, "tid": -914061504, "ts": 1716454222296039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222361694, "dur": 4,
    "args": {
      "External id": 65630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65630,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 65630, "pid": 5, "tid": 7, "ts": 1716454222361694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296060, "dur": 6,
    "args": {
      "External id": 65630,
      "cbid": 211, "correlation": 65630
    }
  },
  {
    "ph": "s", "id": 65630, "pid": 76337, "tid": -914061504, "ts": 1716454222296060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222361699, "dur": 18,
    "args": {
      "External id": 65633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65633, "pid": 5, "tid": 7, "ts": 1716454222361699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296078, "dur": 6,
    "args": {
      "External id": 65633,
      "cbid": 211, "correlation": 65633
    }
  },
  {
    "ph": "s", "id": 65633, "pid": 76337, "tid": -914061504, "ts": 1716454222296078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222361719, "dur": 13,
    "args": {
      "External id": 65642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65642,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65642, "pid": 5, "tid": 7, "ts": 1716454222361719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296118, "dur": 9,
    "args": {
      "External id": 65642,
      "cbid": 211, "correlation": 65642
    }
  },
  {
    "ph": "s", "id": 65642, "pid": 76337, "tid": -914061504, "ts": 1716454222296118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296181, "dur": 0,
    "args": {
      "External id": 65652,
      "cbid": 317, "correlation": 65652
    }
  },
  {
    "ph": "f", "id": 65652, "pid": 76337, "tid": -914061504, "ts": 1716454222296181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296182, "dur": 0,
    "args": {
      "External id": 65653,
      "cbid": 203, "correlation": 65653
    }
  },
  {
    "ph": "f", "id": 65653, "pid": 76337, "tid": -914061504, "ts": 1716454222296182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296182, "dur": 0,
    "args": {
      "External id": 65654,
      "cbid": 205, "correlation": 65654
    }
  },
  {
    "ph": "f", "id": 65654, "pid": 76337, "tid": -914061504, "ts": 1716454222296182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222361733, "dur": 12,
    "args": {
      "External id": 65658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65658, "pid": 5, "tid": 7, "ts": 1716454222361733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296196, "dur": 12,
    "args": {
      "External id": 65658,
      "cbid": 211, "correlation": 65658
    }
  },
  {
    "ph": "s", "id": 65658, "pid": 76337, "tid": -914061504, "ts": 1716454222296196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222361746, "dur": 177,
    "args": {
      "External id": 65660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65660, "pid": 5, "tid": 7, "ts": 1716454222361746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296210, "dur": 5,
    "args": {
      "External id": 65660,
      "cbid": 211, "correlation": 65660
    }
  },
  {
    "ph": "s", "id": 65660, "pid": 76337, "tid": -914061504, "ts": 1716454222296210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222361926, "dur": 1,
    "args": {
      "External id": 65662,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 65662,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5660377358490566
    }
  },
  {
    "ph": "f", "id": 65662, "pid": 5, "tid": 7, "ts": 1716454222361926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296221, "dur": 10,
    "args": {
      "External id": 65662,
      "cbid": 51, "correlation": 65662
    }
  },
  {
    "ph": "s", "id": 65662, "pid": 76337, "tid": -914061504, "ts": 1716454222296221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222361930, "dur": 711,
    "args": {
      "External id": 65663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65663,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65663, "pid": 5, "tid": 7, "ts": 1716454222361930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296232, "dur": 6,
    "args": {
      "External id": 65663,
      "cbid": 211, "correlation": 65663
    }
  },
  {
    "ph": "s", "id": 65663, "pid": 76337, "tid": -914061504, "ts": 1716454222296232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222362642, "dur": 12,
    "args": {
      "External id": 65665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65665, "pid": 5, "tid": 7, "ts": 1716454222362642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296242, "dur": 6,
    "args": {
      "External id": 65665,
      "cbid": 211, "correlation": 65665
    }
  },
  {
    "ph": "s", "id": 65665, "pid": 76337, "tid": -914061504, "ts": 1716454222296242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222362656, "dur": 16,
    "args": {
      "External id": 65671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65671, "pid": 5, "tid": 7, "ts": 1716454222362656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296271, "dur": 9,
    "args": {
      "External id": 65671,
      "cbid": 211, "correlation": 65671
    }
  },
  {
    "ph": "s", "id": 65671, "pid": 76337, "tid": -914061504, "ts": 1716454222296271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296330, "dur": 0,
    "args": {
      "External id": 65681,
      "cbid": 317, "correlation": 65681
    }
  },
  {
    "ph": "f", "id": 65681, "pid": 76337, "tid": -914061504, "ts": 1716454222296330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296331, "dur": 0,
    "args": {
      "External id": 65682,
      "cbid": 203, "correlation": 65682
    }
  },
  {
    "ph": "f", "id": 65682, "pid": 76337, "tid": -914061504, "ts": 1716454222296331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296331, "dur": 0,
    "args": {
      "External id": 65683,
      "cbid": 205, "correlation": 65683
    }
  },
  {
    "ph": "f", "id": 65683, "pid": 76337, "tid": -914061504, "ts": 1716454222296331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222362674, "dur": 23,
    "args": {
      "External id": 65687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65687,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65687, "pid": 5, "tid": 7, "ts": 1716454222362674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296356, "dur": 12,
    "args": {
      "External id": 65687,
      "cbid": 211, "correlation": 65687
    }
  },
  {
    "ph": "s", "id": 65687, "pid": 76337, "tid": -914061504, "ts": 1716454222296356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222362698, "dur": 4,
    "args": {
      "External id": 65689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 65689, "pid": 5, "tid": 7, "ts": 1716454222362698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296377, "dur": 6,
    "args": {
      "External id": 65689,
      "cbid": 211, "correlation": 65689
    }
  },
  {
    "ph": "s", "id": 65689, "pid": 76337, "tid": -914061504, "ts": 1716454222296377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296387, "dur": 0,
    "args": {
      "External id": 65690,
      "cbid": 51, "correlation": 65690
    }
  },
  {
    "ph": "s", "id": 65690, "pid": 76337, "tid": -914061504, "ts": 1716454222296387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222362703, "dur": 193,
    "args": {
      "External id": 65691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65691,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 65691, "pid": 5, "tid": 7, "ts": 1716454222362703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296388, "dur": 7,
    "args": {
      "External id": 65691,
      "cbid": 211, "correlation": 65691
    }
  },
  {
    "ph": "s", "id": 65691, "pid": 76337, "tid": -914061504, "ts": 1716454222296388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222362898, "dur": 17,
    "args": {
      "External id": 65696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65696, "pid": 5, "tid": 7, "ts": 1716454222362898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296415, "dur": 9,
    "args": {
      "External id": 65696,
      "cbid": 211, "correlation": 65696
    }
  },
  {
    "ph": "s", "id": 65696, "pid": 76337, "tid": -914061504, "ts": 1716454222296415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222362917, "dur": 13,
    "args": {
      "External id": 65704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65704,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65704, "pid": 5, "tid": 7, "ts": 1716454222362917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296445, "dur": 8,
    "args": {
      "External id": 65704,
      "cbid": 211, "correlation": 65704
    }
  },
  {
    "ph": "s", "id": 65704, "pid": 76337, "tid": -914061504, "ts": 1716454222296445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222362931, "dur": 11,
    "args": {
      "External id": 65712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65712,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65712, "pid": 5, "tid": 7, "ts": 1716454222362931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296474, "dur": 8,
    "args": {
      "External id": 65712,
      "cbid": 211, "correlation": 65712
    }
  },
  {
    "ph": "s", "id": 65712, "pid": 76337, "tid": -914061504, "ts": 1716454222296474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222362944, "dur": 21,
    "args": {
      "External id": 65732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65732,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 65732, "pid": 5, "tid": 7, "ts": 1716454222362944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296561, "dur": 12,
    "args": {
      "External id": 65732,
      "cbid": 211, "correlation": 65732
    }
  },
  {
    "ph": "s", "id": 65732, "pid": 76337, "tid": -914061504, "ts": 1716454222296561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222362966, "dur": 5,
    "args": {
      "External id": 65744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65744,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 65744, "pid": 5, "tid": 7, "ts": 1716454222362966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296583, "dur": 6,
    "args": {
      "External id": 65744,
      "cbid": 211, "correlation": 65744
    }
  },
  {
    "ph": "s", "id": 65744, "pid": 76337, "tid": -914061504, "ts": 1716454222296583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222362972, "dur": 18,
    "args": {
      "External id": 65747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65747, "pid": 5, "tid": 7, "ts": 1716454222362972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296601, "dur": 6,
    "args": {
      "External id": 65747,
      "cbid": 211, "correlation": 65747
    }
  },
  {
    "ph": "s", "id": 65747, "pid": 76337, "tid": -914061504, "ts": 1716454222296601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296659, "dur": 0,
    "args": {
      "External id": 65758,
      "cbid": 317, "correlation": 65758
    }
  },
  {
    "ph": "f", "id": 65758, "pid": 76337, "tid": -914061504, "ts": 1716454222296659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296660, "dur": 0,
    "args": {
      "External id": 65759,
      "cbid": 203, "correlation": 65759
    }
  },
  {
    "ph": "f", "id": 65759, "pid": 76337, "tid": -914061504, "ts": 1716454222296660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296660, "dur": 0,
    "args": {
      "External id": 65760,
      "cbid": 205, "correlation": 65760
    }
  },
  {
    "ph": "f", "id": 65760, "pid": 76337, "tid": -914061504, "ts": 1716454222296660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222362992, "dur": 13,
    "args": {
      "External id": 65764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65764,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65764, "pid": 5, "tid": 7, "ts": 1716454222362992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296673, "dur": 11,
    "args": {
      "External id": 65764,
      "cbid": 211, "correlation": 65764
    }
  },
  {
    "ph": "s", "id": 65764, "pid": 76337, "tid": -914061504, "ts": 1716454222296673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222363006, "dur": 4,
    "args": {
      "External id": 65766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65766,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 65766, "pid": 5, "tid": 7, "ts": 1716454222363006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296689, "dur": 6,
    "args": {
      "External id": 65766,
      "cbid": 211, "correlation": 65766
    }
  },
  {
    "ph": "s", "id": 65766, "pid": 76337, "tid": -914061504, "ts": 1716454222296689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296697, "dur": 0,
    "args": {
      "External id": 65767,
      "cbid": 51, "correlation": 65767
    }
  },
  {
    "ph": "s", "id": 65767, "pid": 76337, "tid": -914061504, "ts": 1716454222296697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222363011, "dur": 100,
    "args": {
      "External id": 65768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65768,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 65768, "pid": 5, "tid": 7, "ts": 1716454222363011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296698, "dur": 6,
    "args": {
      "External id": 65768,
      "cbid": 211, "correlation": 65768
    }
  },
  {
    "ph": "s", "id": 65768, "pid": 76337, "tid": -914061504, "ts": 1716454222296698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222363112, "dur": 17,
    "args": {
      "External id": 65773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65773,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65773, "pid": 5, "tid": 7, "ts": 1716454222363112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296726, "dur": 8,
    "args": {
      "External id": 65773,
      "cbid": 211, "correlation": 65773
    }
  },
  {
    "ph": "s", "id": 65773, "pid": 76337, "tid": -914061504, "ts": 1716454222296726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222363131, "dur": 93,
    "args": {
      "External id": 65782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65782,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65782, "pid": 5, "tid": 7, "ts": 1716454222363131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296808, "dur": 15,
    "args": {
      "External id": 65782,
      "cbid": 211, "correlation": 65782
    }
  },
  {
    "ph": "s", "id": 65782, "pid": 76337, "tid": -914061504, "ts": 1716454222296808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222363225, "dur": 32,
    "args": {
      "External id": 65804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65804,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65804, "pid": 5, "tid": 7, "ts": 1716454222363225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296871, "dur": 11,
    "args": {
      "External id": 65804,
      "cbid": 211, "correlation": 65804
    }
  },
  {
    "ph": "s", "id": 65804, "pid": 76337, "tid": -914061504, "ts": 1716454222296871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296966, "dur": 1,
    "args": {
      "External id": 65815,
      "cbid": 251, "correlation": 65815
    }
  },
  {
    "ph": "f", "id": 65815, "pid": 76337, "tid": -914061504, "ts": 1716454222296966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222363258, "dur": 180,
    "args": {
      "External id": 65816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65816,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65816, "pid": 5, "tid": 7, "ts": 1716454222363258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222296972, "dur": 21,
    "args": {
      "External id": 65816,
      "cbid": 211, "correlation": 65816
    }
  },
  {
    "ph": "s", "id": 65816, "pid": 76337, "tid": -914061504, "ts": 1716454222296972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297053, "dur": 1,
    "args": {
      "External id": 65827,
      "cbid": 251, "correlation": 65827
    }
  },
  {
    "ph": "f", "id": 65827, "pid": 76337, "tid": -914061504, "ts": 1716454222297053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222363440, "dur": 173,
    "args": {
      "External id": 65828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65828,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65828, "pid": 5, "tid": 7, "ts": 1716454222363440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297057, "dur": 11,
    "args": {
      "External id": 65828,
      "cbid": 211, "correlation": 65828
    }
  },
  {
    "ph": "s", "id": 65828, "pid": 76337, "tid": -914061504, "ts": 1716454222297057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297122, "dur": 1,
    "args": {
      "External id": 65839,
      "cbid": 251, "correlation": 65839
    }
  },
  {
    "ph": "f", "id": 65839, "pid": 76337, "tid": -914061504, "ts": 1716454222297122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222363614, "dur": 170,
    "args": {
      "External id": 65840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65840,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65840, "pid": 5, "tid": 7, "ts": 1716454222363614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297126, "dur": 11,
    "args": {
      "External id": 65840,
      "cbid": 211, "correlation": 65840
    }
  },
  {
    "ph": "s", "id": 65840, "pid": 76337, "tid": -914061504, "ts": 1716454222297126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222363785, "dur": 361,
    "args": {
      "External id": 65865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65865,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65865, "pid": 5, "tid": 7, "ts": 1716454222363785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297212, "dur": 13,
    "args": {
      "External id": 65865,
      "cbid": 211, "correlation": 65865
    }
  },
  {
    "ph": "s", "id": 65865, "pid": 76337, "tid": -914061504, "ts": 1716454222297212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297318, "dur": 1,
    "args": {
      "External id": 65883,
      "cbid": 251, "correlation": 65883
    }
  },
  {
    "ph": "f", "id": 65883, "pid": 76337, "tid": -914061504, "ts": 1716454222297318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222364148, "dur": 182,
    "args": {
      "External id": 65885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65885,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65885, "pid": 5, "tid": 7, "ts": 1716454222364148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297324, "dur": 13,
    "args": {
      "External id": 65885,
      "cbid": 211, "correlation": 65885
    }
  },
  {
    "ph": "s", "id": 65885, "pid": 76337, "tid": -914061504, "ts": 1716454222297324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222364330, "dur": 19,
    "args": {
      "External id": 65893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65893,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65893, "pid": 5, "tid": 7, "ts": 1716454222364330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297408, "dur": 14,
    "args": {
      "External id": 65893,
      "cbid": 211, "correlation": 65893
    }
  },
  {
    "ph": "s", "id": 65893, "pid": 76337, "tid": -914061504, "ts": 1716454222297408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222364351, "dur": 28,
    "args": {
      "External id": 65901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65901,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65901, "pid": 5, "tid": 7, "ts": 1716454222364351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297450, "dur": 9,
    "args": {
      "External id": 65901,
      "cbid": 211, "correlation": 65901
    }
  },
  {
    "ph": "s", "id": 65901, "pid": 76337, "tid": -914061504, "ts": 1716454222297450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222364381, "dur": 19,
    "args": {
      "External id": 65912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65912, "pid": 5, "tid": 7, "ts": 1716454222364381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297526, "dur": 12,
    "args": {
      "External id": 65912,
      "cbid": 211, "correlation": 65912
    }
  },
  {
    "ph": "s", "id": 65912, "pid": 76337, "tid": -914061504, "ts": 1716454222297526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222364401, "dur": 17,
    "args": {
      "External id": 65934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65934,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 65934, "pid": 5, "tid": 7, "ts": 1716454222364401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297557, "dur": 8,
    "args": {
      "External id": 65934,
      "cbid": 211, "correlation": 65934
    }
  },
  {
    "ph": "s", "id": 65934, "pid": 76337, "tid": -914061504, "ts": 1716454222297557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297644, "dur": 1,
    "args": {
      "External id": 65945,
      "cbid": 251, "correlation": 65945
    }
  },
  {
    "ph": "f", "id": 65945, "pid": 76337, "tid": -914061504, "ts": 1716454222297644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222364420, "dur": 97,
    "args": {
      "External id": 65946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65946,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 65946, "pid": 5, "tid": 7, "ts": 1716454222364420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297649, "dur": 13,
    "args": {
      "External id": 65946,
      "cbid": 211, "correlation": 65946
    }
  },
  {
    "ph": "s", "id": 65946, "pid": 76337, "tid": -914061504, "ts": 1716454222297649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297722, "dur": 1,
    "args": {
      "External id": 65957,
      "cbid": 251, "correlation": 65957
    }
  },
  {
    "ph": "f", "id": 65957, "pid": 76337, "tid": -914061504, "ts": 1716454222297722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297726, "dur": 0,
    "args": {
      "External id": 65958,
      "cbid": 251, "correlation": 65958
    }
  },
  {
    "ph": "f", "id": 65958, "pid": 76337, "tid": -914061504, "ts": 1716454222297726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222364518, "dur": 12,
    "args": {
      "External id": 65959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65959,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65959, "pid": 5, "tid": 7, "ts": 1716454222364518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297728, "dur": 12,
    "args": {
      "External id": 65959,
      "cbid": 211, "correlation": 65959
    }
  },
  {
    "ph": "s", "id": 65959, "pid": 76337, "tid": -914061504, "ts": 1716454222297728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222364532, "dur": 6,
    "args": {
      "External id": 65961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65961,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65961, "pid": 5, "tid": 7, "ts": 1716454222364532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297742, "dur": 6,
    "args": {
      "External id": 65961,
      "cbid": 211, "correlation": 65961
    }
  },
  {
    "ph": "s", "id": 65961, "pid": 76337, "tid": -914061504, "ts": 1716454222297742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297800, "dur": 1,
    "args": {
      "External id": 65972,
      "cbid": 251, "correlation": 65972
    }
  },
  {
    "ph": "f", "id": 65972, "pid": 76337, "tid": -914061504, "ts": 1716454222297800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297803, "dur": 0,
    "args": {
      "External id": 65973,
      "cbid": 251, "correlation": 65973
    }
  },
  {
    "ph": "f", "id": 65973, "pid": 76337, "tid": -914061504, "ts": 1716454222297803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222364540, "dur": 9,
    "args": {
      "External id": 65974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65974,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65974, "pid": 5, "tid": 7, "ts": 1716454222364540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297805, "dur": 12,
    "args": {
      "External id": 65974,
      "cbid": 211, "correlation": 65974
    }
  },
  {
    "ph": "s", "id": 65974, "pid": 76337, "tid": -914061504, "ts": 1716454222297805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222364550, "dur": 4,
    "args": {
      "External id": 65976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 65976,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 65976, "pid": 5, "tid": 7, "ts": 1716454222364550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297818, "dur": 6,
    "args": {
      "External id": 65976,
      "cbid": 211, "correlation": 65976
    }
  },
  {
    "ph": "s", "id": 65976, "pid": 76337, "tid": -914061504, "ts": 1716454222297818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222364555, "dur": 60,
    "args": {
      "External id": 66001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66001,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66001, "pid": 5, "tid": 7, "ts": 1716454222364555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222297895, "dur": 13,
    "args": {
      "External id": 66001,
      "cbid": 211, "correlation": 66001
    }
  },
  {
    "ph": "s", "id": 66001, "pid": 76337, "tid": -914061504, "ts": 1716454222297895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298001, "dur": 3,
    "args": {
      "External id": 66019,
      "cbid": 251, "correlation": 66019
    }
  },
  {
    "ph": "f", "id": 66019, "pid": 76337, "tid": -914061504, "ts": 1716454222298001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222364616, "dur": 99,
    "args": {
      "External id": 66021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66021,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66021, "pid": 5, "tid": 7, "ts": 1716454222364616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298008, "dur": 14,
    "args": {
      "External id": 66021,
      "cbid": 211, "correlation": 66021
    }
  },
  {
    "ph": "s", "id": 66021, "pid": 76337, "tid": -914061504, "ts": 1716454222298008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222364716, "dur": 10,
    "args": {
      "External id": 66029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66029,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66029, "pid": 5, "tid": 7, "ts": 1716454222364716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298083, "dur": 12,
    "args": {
      "External id": 66029,
      "cbid": 211, "correlation": 66029
    }
  },
  {
    "ph": "s", "id": 66029, "pid": 76337, "tid": -914061504, "ts": 1716454222298083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222364728, "dur": 21,
    "args": {
      "External id": 66037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66037,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66037, "pid": 5, "tid": 7, "ts": 1716454222364728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298125, "dur": 9,
    "args": {
      "External id": 66037,
      "cbid": 211, "correlation": 66037
    }
  },
  {
    "ph": "s", "id": 66037, "pid": 76337, "tid": -914061504, "ts": 1716454222298125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222364750, "dur": 19,
    "args": {
      "External id": 66059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66059,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66059, "pid": 5, "tid": 7, "ts": 1716454222364750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298177, "dur": 10,
    "args": {
      "External id": 66059,
      "cbid": 211, "correlation": 66059
    }
  },
  {
    "ph": "s", "id": 66059, "pid": 76337, "tid": -914061504, "ts": 1716454222298177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298265, "dur": 1,
    "args": {
      "External id": 66075,
      "cbid": 251, "correlation": 66075
    }
  },
  {
    "ph": "f", "id": 66075, "pid": 76337, "tid": -914061504, "ts": 1716454222298265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298269, "dur": 0,
    "args": {
      "External id": 66077,
      "cbid": 251, "correlation": 66077
    }
  },
  {
    "ph": "f", "id": 66077, "pid": 76337, "tid": -914061504, "ts": 1716454222298269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222364771, "dur": 527,
    "args": {
      "External id": 66078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66078,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66078, "pid": 5, "tid": 7, "ts": 1716454222364771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298271, "dur": 13,
    "args": {
      "External id": 66078,
      "cbid": 211, "correlation": 66078
    }
  },
  {
    "ph": "s", "id": 66078, "pid": 76337, "tid": -914061504, "ts": 1716454222298271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222365300, "dur": 71,
    "args": {
      "External id": 66086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66086, "pid": 5, "tid": 7, "ts": 1716454222365300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298339, "dur": 13,
    "args": {
      "External id": 66086,
      "cbid": 211, "correlation": 66086
    }
  },
  {
    "ph": "s", "id": 66086, "pid": 76337, "tid": -914061504, "ts": 1716454222298339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222365373, "dur": 69,
    "args": {
      "External id": 66094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66094, "pid": 5, "tid": 7, "ts": 1716454222365373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298371, "dur": 8,
    "args": {
      "External id": 66094,
      "cbid": 211, "correlation": 66094
    }
  },
  {
    "ph": "s", "id": 66094, "pid": 76337, "tid": -914061504, "ts": 1716454222298371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298455, "dur": 1,
    "args": {
      "External id": 66110,
      "cbid": 251, "correlation": 66110
    }
  },
  {
    "ph": "f", "id": 66110, "pid": 76337, "tid": -914061504, "ts": 1716454222298455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222365444, "dur": 1,
    "args": {
      "External id": 66112,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 66112,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 66112, "pid": 5, "tid": 7, "ts": 1716454222365444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298460, "dur": 10,
    "args": {
      "External id": 66112,
      "cbid": 51, "correlation": 66112
    }
  },
  {
    "ph": "s", "id": 66112, "pid": 76337, "tid": -914061504, "ts": 1716454222298460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222365447, "dur": 292,
    "args": {
      "External id": 66113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66113,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66113, "pid": 5, "tid": 7, "ts": 1716454222365447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298471, "dur": 11,
    "args": {
      "External id": 66113,
      "cbid": 211, "correlation": 66113
    }
  },
  {
    "ph": "s", "id": 66113, "pid": 76337, "tid": -914061504, "ts": 1716454222298471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222365741, "dur": 14,
    "args": {
      "External id": 66121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66121,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66121, "pid": 5, "tid": 7, "ts": 1716454222365741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298514, "dur": 10,
    "args": {
      "External id": 66121,
      "cbid": 211, "correlation": 66121
    }
  },
  {
    "ph": "s", "id": 66121, "pid": 76337, "tid": -914061504, "ts": 1716454222298514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222365757, "dur": 42,
    "args": {
      "External id": 66132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66132, "pid": 5, "tid": 7, "ts": 1716454222365757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298583, "dur": 12,
    "args": {
      "External id": 66132,
      "cbid": 211, "correlation": 66132
    }
  },
  {
    "ph": "s", "id": 66132, "pid": 76337, "tid": -914061504, "ts": 1716454222298583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298647, "dur": 0,
    "args": {
      "External id": 66144,
      "cbid": 317, "correlation": 66144
    }
  },
  {
    "ph": "f", "id": 66144, "pid": 76337, "tid": -914061504, "ts": 1716454222298647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298648, "dur": 0,
    "args": {
      "External id": 66145,
      "cbid": 203, "correlation": 66145
    }
  },
  {
    "ph": "f", "id": 66145, "pid": 76337, "tid": -914061504, "ts": 1716454222298648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298649, "dur": 0,
    "args": {
      "External id": 66146,
      "cbid": 205, "correlation": 66146
    }
  },
  {
    "ph": "f", "id": 66146, "pid": 76337, "tid": -914061504, "ts": 1716454222298649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222365800, "dur": 14,
    "args": {
      "External id": 66150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66150,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66150, "pid": 5, "tid": 7, "ts": 1716454222365800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298664, "dur": 12,
    "args": {
      "External id": 66150,
      "cbid": 211, "correlation": 66150
    }
  },
  {
    "ph": "s", "id": 66150, "pid": 76337, "tid": -914061504, "ts": 1716454222298664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222365815, "dur": 4,
    "args": {
      "External id": 66152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66152,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66152, "pid": 5, "tid": 7, "ts": 1716454222365815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298681, "dur": 6,
    "args": {
      "External id": 66152,
      "cbid": 211, "correlation": 66152
    }
  },
  {
    "ph": "s", "id": 66152, "pid": 76337, "tid": -914061504, "ts": 1716454222298681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298690, "dur": 0,
    "args": {
      "External id": 66153,
      "cbid": 51, "correlation": 66153
    }
  },
  {
    "ph": "s", "id": 66153, "pid": 76337, "tid": -914061504, "ts": 1716454222298690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222365820, "dur": 106,
    "args": {
      "External id": 66154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66154,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 66154, "pid": 5, "tid": 7, "ts": 1716454222365820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298690, "dur": 5,
    "args": {
      "External id": 66154,
      "cbid": 211, "correlation": 66154
    }
  },
  {
    "ph": "s", "id": 66154, "pid": 76337, "tid": -914061504, "ts": 1716454222298690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222365928, "dur": 18,
    "args": {
      "External id": 66159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66159,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66159, "pid": 5, "tid": 7, "ts": 1716454222365928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298717, "dur": 8,
    "args": {
      "External id": 66159,
      "cbid": 211, "correlation": 66159
    }
  },
  {
    "ph": "s", "id": 66159, "pid": 76337, "tid": -914061504, "ts": 1716454222298717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222365947, "dur": 12,
    "args": {
      "External id": 66167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66167,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66167, "pid": 5, "tid": 7, "ts": 1716454222365947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298749, "dur": 11,
    "args": {
      "External id": 66167,
      "cbid": 211, "correlation": 66167
    }
  },
  {
    "ph": "s", "id": 66167, "pid": 76337, "tid": -914061504, "ts": 1716454222298749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222365960, "dur": 33,
    "args": {
      "External id": 66176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66176,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66176, "pid": 5, "tid": 7, "ts": 1716454222365960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298791, "dur": 10,
    "args": {
      "External id": 66176,
      "cbid": 211, "correlation": 66176
    }
  },
  {
    "ph": "s", "id": 66176, "pid": 76337, "tid": -914061504, "ts": 1716454222298791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222365994, "dur": 33,
    "args": {
      "External id": 66196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66196,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 66196, "pid": 5, "tid": 7, "ts": 1716454222365994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298862, "dur": 11,
    "args": {
      "External id": 66196,
      "cbid": 211, "correlation": 66196
    }
  },
  {
    "ph": "s", "id": 66196, "pid": 76337, "tid": -914061504, "ts": 1716454222298862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222366029, "dur": 6,
    "args": {
      "External id": 66208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66208,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66208, "pid": 5, "tid": 7, "ts": 1716454222366029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298883, "dur": 6,
    "args": {
      "External id": 66208,
      "cbid": 211, "correlation": 66208
    }
  },
  {
    "ph": "s", "id": 66208, "pid": 76337, "tid": -914061504, "ts": 1716454222298883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222366036, "dur": 33,
    "args": {
      "External id": 66211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66211, "pid": 5, "tid": 7, "ts": 1716454222366036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298902, "dur": 6,
    "args": {
      "External id": 66211,
      "cbid": 211, "correlation": 66211
    }
  },
  {
    "ph": "s", "id": 66211, "pid": 76337, "tid": -914061504, "ts": 1716454222298902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222366070, "dur": 22,
    "args": {
      "External id": 66220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66220,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66220, "pid": 5, "tid": 7, "ts": 1716454222366070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222298940, "dur": 10,
    "args": {
      "External id": 66220,
      "cbid": 211, "correlation": 66220
    }
  },
  {
    "ph": "s", "id": 66220, "pid": 76337, "tid": -914061504, "ts": 1716454222298940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299001, "dur": 0,
    "args": {
      "External id": 66230,
      "cbid": 317, "correlation": 66230
    }
  },
  {
    "ph": "f", "id": 66230, "pid": 76337, "tid": -914061504, "ts": 1716454222299001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299001, "dur": 0,
    "args": {
      "External id": 66231,
      "cbid": 203, "correlation": 66231
    }
  },
  {
    "ph": "f", "id": 66231, "pid": 76337, "tid": -914061504, "ts": 1716454222299001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299002, "dur": 0,
    "args": {
      "External id": 66232,
      "cbid": 205, "correlation": 66232
    }
  },
  {
    "ph": "f", "id": 66232, "pid": 76337, "tid": -914061504, "ts": 1716454222299002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222366094, "dur": 23,
    "args": {
      "External id": 66236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66236, "pid": 5, "tid": 7, "ts": 1716454222366094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299017, "dur": 12,
    "args": {
      "External id": 66236,
      "cbid": 211, "correlation": 66236
    }
  },
  {
    "ph": "s", "id": 66236, "pid": 76337, "tid": -914061504, "ts": 1716454222299017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222366118, "dur": 350,
    "args": {
      "External id": 66238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66238,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66238, "pid": 5, "tid": 7, "ts": 1716454222366118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299031, "dur": 8,
    "args": {
      "External id": 66238,
      "cbid": 211, "correlation": 66238
    }
  },
  {
    "ph": "s", "id": 66238, "pid": 76337, "tid": -914061504, "ts": 1716454222299031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222366470, "dur": 1,
    "args": {
      "External id": 66240,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 66240,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5084745762711864
    }
  },
  {
    "ph": "f", "id": 66240, "pid": 5, "tid": 7, "ts": 1716454222366470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299046, "dur": 6,
    "args": {
      "External id": 66240,
      "cbid": 51, "correlation": 66240
    }
  },
  {
    "ph": "s", "id": 66240, "pid": 76337, "tid": -914061504, "ts": 1716454222299046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222366474, "dur": 1379,
    "args": {
      "External id": 66241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66241,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66241, "pid": 5, "tid": 7, "ts": 1716454222366474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299053, "dur": 7,
    "args": {
      "External id": 66241,
      "cbid": 211, "correlation": 66241
    }
  },
  {
    "ph": "s", "id": 66241, "pid": 76337, "tid": -914061504, "ts": 1716454222299053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222367854, "dur": 14,
    "args": {
      "External id": 66243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66243,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66243, "pid": 5, "tid": 7, "ts": 1716454222367854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299065, "dur": 5,
    "args": {
      "External id": 66243,
      "cbid": 211, "correlation": 66243
    }
  },
  {
    "ph": "s", "id": 66243, "pid": 76337, "tid": -914061504, "ts": 1716454222299065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222367869, "dur": 16,
    "args": {
      "External id": 66249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66249,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66249, "pid": 5, "tid": 7, "ts": 1716454222367869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299092, "dur": 8,
    "args": {
      "External id": 66249,
      "cbid": 211, "correlation": 66249
    }
  },
  {
    "ph": "s", "id": 66249, "pid": 76337, "tid": -914061504, "ts": 1716454222299092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222367887, "dur": 4,
    "args": {
      "External id": 66257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66257,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 66257, "pid": 5, "tid": 7, "ts": 1716454222367887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299137, "dur": 9,
    "args": {
      "External id": 66257,
      "cbid": 211, "correlation": 66257
    }
  },
  {
    "ph": "s", "id": 66257, "pid": 76337, "tid": -914061504, "ts": 1716454222299137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299204, "dur": 1,
    "args": {
      "External id": 66273,
      "cbid": 251, "correlation": 66273
    }
  },
  {
    "ph": "f", "id": 66273, "pid": 76337, "tid": -914061504, "ts": 1716454222299204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299209, "dur": 0,
    "args": {
      "External id": 66275,
      "cbid": 251, "correlation": 66275
    }
  },
  {
    "ph": "f", "id": 66275, "pid": 76337, "tid": -914061504, "ts": 1716454222299209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222367892, "dur": 14,
    "args": {
      "External id": 66276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66276,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66276, "pid": 5, "tid": 7, "ts": 1716454222367892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299212, "dur": 11,
    "args": {
      "External id": 66276,
      "cbid": 211, "correlation": 66276
    }
  },
  {
    "ph": "s", "id": 66276, "pid": 76337, "tid": -914061504, "ts": 1716454222299212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222367908, "dur": 6,
    "args": {
      "External id": 66278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66278,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66278, "pid": 5, "tid": 7, "ts": 1716454222367908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299225, "dur": 5,
    "args": {
      "External id": 66278,
      "cbid": 211, "correlation": 66278
    }
  },
  {
    "ph": "s", "id": 66278, "pid": 76337, "tid": -914061504, "ts": 1716454222299225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222367915, "dur": 17,
    "args": {
      "External id": 66288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66288, "pid": 5, "tid": 7, "ts": 1716454222367915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299282, "dur": 12,
    "args": {
      "External id": 66288,
      "cbid": 211, "correlation": 66288
    }
  },
  {
    "ph": "s", "id": 66288, "pid": 76337, "tid": -914061504, "ts": 1716454222299282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222367934, "dur": 19,
    "args": {
      "External id": 66308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66308,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 66308, "pid": 5, "tid": 7, "ts": 1716454222367934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299351, "dur": 11,
    "args": {
      "External id": 66308,
      "cbid": 211, "correlation": 66308
    }
  },
  {
    "ph": "s", "id": 66308, "pid": 76337, "tid": -914061504, "ts": 1716454222299351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222367954, "dur": 4,
    "args": {
      "External id": 66320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66320,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 66320, "pid": 5, "tid": 7, "ts": 1716454222367954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299372, "dur": 7,
    "args": {
      "External id": 66320,
      "cbid": 211, "correlation": 66320
    }
  },
  {
    "ph": "s", "id": 66320, "pid": 76337, "tid": -914061504, "ts": 1716454222299372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222367960, "dur": 18,
    "args": {
      "External id": 66323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66323, "pid": 5, "tid": 7, "ts": 1716454222367960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299391, "dur": 6,
    "args": {
      "External id": 66323,
      "cbid": 211, "correlation": 66323
    }
  },
  {
    "ph": "s", "id": 66323, "pid": 76337, "tid": -914061504, "ts": 1716454222299391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222367979, "dur": 12,
    "args": {
      "External id": 66332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66332,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66332, "pid": 5, "tid": 7, "ts": 1716454222367979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299431, "dur": 10,
    "args": {
      "External id": 66332,
      "cbid": 211, "correlation": 66332
    }
  },
  {
    "ph": "s", "id": 66332, "pid": 76337, "tid": -914061504, "ts": 1716454222299431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299493, "dur": 0,
    "args": {
      "External id": 66342,
      "cbid": 317, "correlation": 66342
    }
  },
  {
    "ph": "f", "id": 66342, "pid": 76337, "tid": -914061504, "ts": 1716454222299493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299494, "dur": 0,
    "args": {
      "External id": 66343,
      "cbid": 203, "correlation": 66343
    }
  },
  {
    "ph": "f", "id": 66343, "pid": 76337, "tid": -914061504, "ts": 1716454222299494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299495, "dur": 0,
    "args": {
      "External id": 66344,
      "cbid": 205, "correlation": 66344
    }
  },
  {
    "ph": "f", "id": 66344, "pid": 76337, "tid": -914061504, "ts": 1716454222299495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222367993, "dur": 12,
    "args": {
      "External id": 66348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66348, "pid": 5, "tid": 7, "ts": 1716454222367993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299509, "dur": 12,
    "args": {
      "External id": 66348,
      "cbid": 211, "correlation": 66348
    }
  },
  {
    "ph": "s", "id": 66348, "pid": 76337, "tid": -914061504, "ts": 1716454222299509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222368006, "dur": 177,
    "args": {
      "External id": 66350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66350, "pid": 5, "tid": 7, "ts": 1716454222368006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299523, "dur": 5,
    "args": {
      "External id": 66350,
      "cbid": 211, "correlation": 66350
    }
  },
  {
    "ph": "s", "id": 66350, "pid": 76337, "tid": -914061504, "ts": 1716454222299523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222368186, "dur": 1,
    "args": {
      "External id": 66352,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 66352,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5555555555555556
    }
  },
  {
    "ph": "f", "id": 66352, "pid": 5, "tid": 7, "ts": 1716454222368186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299533, "dur": 7,
    "args": {
      "External id": 66352,
      "cbid": 51, "correlation": 66352
    }
  },
  {
    "ph": "s", "id": 66352, "pid": 76337, "tid": -914061504, "ts": 1716454222299533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222368189, "dur": 710,
    "args": {
      "External id": 66353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66353,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66353, "pid": 5, "tid": 7, "ts": 1716454222368189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299542, "dur": 6,
    "args": {
      "External id": 66353,
      "cbid": 211, "correlation": 66353
    }
  },
  {
    "ph": "s", "id": 66353, "pid": 76337, "tid": -914061504, "ts": 1716454222299542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222368901, "dur": 13,
    "args": {
      "External id": 66355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66355,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66355, "pid": 5, "tid": 7, "ts": 1716454222368901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299552, "dur": 5,
    "args": {
      "External id": 66355,
      "cbid": 211, "correlation": 66355
    }
  },
  {
    "ph": "s", "id": 66355, "pid": 76337, "tid": -914061504, "ts": 1716454222299552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222368915, "dur": 17,
    "args": {
      "External id": 66361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66361, "pid": 5, "tid": 7, "ts": 1716454222368915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299580, "dur": 9,
    "args": {
      "External id": 66361,
      "cbid": 211, "correlation": 66361
    }
  },
  {
    "ph": "s", "id": 66361, "pid": 76337, "tid": -914061504, "ts": 1716454222299580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299642, "dur": 0,
    "args": {
      "External id": 66371,
      "cbid": 317, "correlation": 66371
    }
  },
  {
    "ph": "f", "id": 66371, "pid": 76337, "tid": -914061504, "ts": 1716454222299642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299643, "dur": 0,
    "args": {
      "External id": 66372,
      "cbid": 203, "correlation": 66372
    }
  },
  {
    "ph": "f", "id": 66372, "pid": 76337, "tid": -914061504, "ts": 1716454222299643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299644, "dur": 0,
    "args": {
      "External id": 66373,
      "cbid": 205, "correlation": 66373
    }
  },
  {
    "ph": "f", "id": 66373, "pid": 76337, "tid": -914061504, "ts": 1716454222299644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222368933, "dur": 21,
    "args": {
      "External id": 66377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66377,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66377, "pid": 5, "tid": 7, "ts": 1716454222368933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299655, "dur": 12,
    "args": {
      "External id": 66377,
      "cbid": 211, "correlation": 66377
    }
  },
  {
    "ph": "s", "id": 66377, "pid": 76337, "tid": -914061504, "ts": 1716454222299655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222368956, "dur": 4,
    "args": {
      "External id": 66379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66379, "pid": 5, "tid": 7, "ts": 1716454222368956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299671, "dur": 6,
    "args": {
      "External id": 66379,
      "cbid": 211, "correlation": 66379
    }
  },
  {
    "ph": "s", "id": 66379, "pid": 76337, "tid": -914061504, "ts": 1716454222299671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299679, "dur": 0,
    "args": {
      "External id": 66380,
      "cbid": 51, "correlation": 66380
    }
  },
  {
    "ph": "s", "id": 66380, "pid": 76337, "tid": -914061504, "ts": 1716454222299679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222368962, "dur": 190,
    "args": {
      "External id": 66381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66381,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 66381, "pid": 5, "tid": 7, "ts": 1716454222368962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299680, "dur": 5,
    "args": {
      "External id": 66381,
      "cbid": 211, "correlation": 66381
    }
  },
  {
    "ph": "s", "id": 66381, "pid": 76337, "tid": -914061504, "ts": 1716454222299680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222369153, "dur": 17,
    "args": {
      "External id": 66386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66386, "pid": 5, "tid": 7, "ts": 1716454222369153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299706, "dur": 8,
    "args": {
      "External id": 66386,
      "cbid": 211, "correlation": 66386
    }
  },
  {
    "ph": "s", "id": 66386, "pid": 76337, "tid": -914061504, "ts": 1716454222299706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222369172, "dur": 14,
    "args": {
      "External id": 66394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66394,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66394, "pid": 5, "tid": 7, "ts": 1716454222369172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299734, "dur": 8,
    "args": {
      "External id": 66394,
      "cbid": 211, "correlation": 66394
    }
  },
  {
    "ph": "s", "id": 66394, "pid": 76337, "tid": -914061504, "ts": 1716454222299734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222369187, "dur": 11,
    "args": {
      "External id": 66402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66402,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66402, "pid": 5, "tid": 7, "ts": 1716454222369187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299763, "dur": 8,
    "args": {
      "External id": 66402,
      "cbid": 211, "correlation": 66402
    }
  },
  {
    "ph": "s", "id": 66402, "pid": 76337, "tid": -914061504, "ts": 1716454222299763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222369199, "dur": 20,
    "args": {
      "External id": 66422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66422,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 66422, "pid": 5, "tid": 7, "ts": 1716454222369199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299844, "dur": 13,
    "args": {
      "External id": 66422,
      "cbid": 211, "correlation": 66422
    }
  },
  {
    "ph": "s", "id": 66422, "pid": 76337, "tid": -914061504, "ts": 1716454222299844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222369221, "dur": 5,
    "args": {
      "External id": 66434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66434,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 66434, "pid": 5, "tid": 7, "ts": 1716454222369221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299867, "dur": 9,
    "args": {
      "External id": 66434,
      "cbid": 211, "correlation": 66434
    }
  },
  {
    "ph": "s", "id": 66434, "pid": 76337, "tid": -914061504, "ts": 1716454222299867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222369227, "dur": 18,
    "args": {
      "External id": 66437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66437, "pid": 5, "tid": 7, "ts": 1716454222369227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299887, "dur": 7,
    "args": {
      "External id": 66437,
      "cbid": 211, "correlation": 66437
    }
  },
  {
    "ph": "s", "id": 66437, "pid": 76337, "tid": -914061504, "ts": 1716454222299887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299957, "dur": 0,
    "args": {
      "External id": 66448,
      "cbid": 317, "correlation": 66448
    }
  },
  {
    "ph": "f", "id": 66448, "pid": 76337, "tid": -914061504, "ts": 1716454222299957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299958, "dur": 0,
    "args": {
      "External id": 66449,
      "cbid": 203, "correlation": 66449
    }
  },
  {
    "ph": "f", "id": 66449, "pid": 76337, "tid": -914061504, "ts": 1716454222299958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299959, "dur": 0,
    "args": {
      "External id": 66450,
      "cbid": 205, "correlation": 66450
    }
  },
  {
    "ph": "f", "id": 66450, "pid": 76337, "tid": -914061504, "ts": 1716454222299959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222369246, "dur": 11,
    "args": {
      "External id": 66454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66454, "pid": 5, "tid": 7, "ts": 1716454222369246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222299985, "dur": 17,
    "args": {
      "External id": 66454,
      "cbid": 211, "correlation": 66454
    }
  },
  {
    "ph": "s", "id": 66454, "pid": 76337, "tid": -914061504, "ts": 1716454222299985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222369259, "dur": 4,
    "args": {
      "External id": 66456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66456, "pid": 5, "tid": 7, "ts": 1716454222369259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300008, "dur": 10,
    "args": {
      "External id": 66456,
      "cbid": 211, "correlation": 66456
    }
  },
  {
    "ph": "s", "id": 66456, "pid": 76337, "tid": -914061504, "ts": 1716454222300008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300022, "dur": 0,
    "args": {
      "External id": 66457,
      "cbid": 51, "correlation": 66457
    }
  },
  {
    "ph": "s", "id": 66457, "pid": 76337, "tid": -914061504, "ts": 1716454222300022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222369264, "dur": 100,
    "args": {
      "External id": 66458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66458,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 66458, "pid": 5, "tid": 7, "ts": 1716454222369264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300023, "dur": 7,
    "args": {
      "External id": 66458,
      "cbid": 211, "correlation": 66458
    }
  },
  {
    "ph": "s", "id": 66458, "pid": 76337, "tid": -914061504, "ts": 1716454222300023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222369365, "dur": 18,
    "args": {
      "External id": 66463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66463, "pid": 5, "tid": 7, "ts": 1716454222369365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300060, "dur": 15,
    "args": {
      "External id": 66463,
      "cbid": 211, "correlation": 66463
    }
  },
  {
    "ph": "s", "id": 66463, "pid": 76337, "tid": -914061504, "ts": 1716454222300060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222369384, "dur": 92,
    "args": {
      "External id": 66472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66472,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66472, "pid": 5, "tid": 7, "ts": 1716454222369384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300215, "dur": 17,
    "args": {
      "External id": 66472,
      "cbid": 211, "correlation": 66472
    }
  },
  {
    "ph": "s", "id": 66472, "pid": 76337, "tid": -914061504, "ts": 1716454222300215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222369478, "dur": 33,
    "args": {
      "External id": 66494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66494,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66494, "pid": 5, "tid": 7, "ts": 1716454222369478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300281, "dur": 10,
    "args": {
      "External id": 66494,
      "cbid": 211, "correlation": 66494
    }
  },
  {
    "ph": "s", "id": 66494, "pid": 76337, "tid": -914061504, "ts": 1716454222300281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300375, "dur": 5,
    "args": {
      "External id": 66505,
      "cbid": 251, "correlation": 66505
    }
  },
  {
    "ph": "f", "id": 66505, "pid": 76337, "tid": -914061504, "ts": 1716454222300375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222369512, "dur": 178,
    "args": {
      "External id": 66506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66506,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66506, "pid": 5, "tid": 7, "ts": 1716454222369512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300384, "dur": 13,
    "args": {
      "External id": 66506,
      "cbid": 211, "correlation": 66506
    }
  },
  {
    "ph": "s", "id": 66506, "pid": 76337, "tid": -914061504, "ts": 1716454222300384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300455, "dur": 1,
    "args": {
      "External id": 66517,
      "cbid": 251, "correlation": 66517
    }
  },
  {
    "ph": "f", "id": 66517, "pid": 76337, "tid": -914061504, "ts": 1716454222300455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222369691, "dur": 175,
    "args": {
      "External id": 66518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66518,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66518, "pid": 5, "tid": 7, "ts": 1716454222369691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300459, "dur": 12,
    "args": {
      "External id": 66518,
      "cbid": 211, "correlation": 66518
    }
  },
  {
    "ph": "s", "id": 66518, "pid": 76337, "tid": -914061504, "ts": 1716454222300459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300524, "dur": 1,
    "args": {
      "External id": 66529,
      "cbid": 251, "correlation": 66529
    }
  },
  {
    "ph": "f", "id": 66529, "pid": 76337, "tid": -914061504, "ts": 1716454222300524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222369867, "dur": 171,
    "args": {
      "External id": 66530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66530,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66530, "pid": 5, "tid": 7, "ts": 1716454222369867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300528, "dur": 11,
    "args": {
      "External id": 66530,
      "cbid": 211, "correlation": 66530
    }
  },
  {
    "ph": "s", "id": 66530, "pid": 76337, "tid": -914061504, "ts": 1716454222300528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222370039, "dur": 364,
    "args": {
      "External id": 66555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66555,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66555, "pid": 5, "tid": 7, "ts": 1716454222370039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300628, "dur": 13,
    "args": {
      "External id": 66555,
      "cbid": 211, "correlation": 66555
    }
  },
  {
    "ph": "s", "id": 66555, "pid": 76337, "tid": -914061504, "ts": 1716454222300628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300735, "dur": 1,
    "args": {
      "External id": 66573,
      "cbid": 251, "correlation": 66573
    }
  },
  {
    "ph": "f", "id": 66573, "pid": 76337, "tid": -914061504, "ts": 1716454222300735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222370405, "dur": 180,
    "args": {
      "External id": 66575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66575,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66575, "pid": 5, "tid": 7, "ts": 1716454222370405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300740, "dur": 13,
    "args": {
      "External id": 66575,
      "cbid": 211, "correlation": 66575
    }
  },
  {
    "ph": "s", "id": 66575, "pid": 76337, "tid": -914061504, "ts": 1716454222300740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222370587, "dur": 20,
    "args": {
      "External id": 66583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66583,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66583, "pid": 5, "tid": 7, "ts": 1716454222370587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300813, "dur": 15,
    "args": {
      "External id": 66583,
      "cbid": 211, "correlation": 66583
    }
  },
  {
    "ph": "s", "id": 66583, "pid": 76337, "tid": -914061504, "ts": 1716454222300813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222370608, "dur": 28,
    "args": {
      "External id": 66591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66591,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66591, "pid": 5, "tid": 7, "ts": 1716454222370608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300857, "dur": 9,
    "args": {
      "External id": 66591,
      "cbid": 211, "correlation": 66591
    }
  },
  {
    "ph": "s", "id": 66591, "pid": 76337, "tid": -914061504, "ts": 1716454222300857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222370637, "dur": 20,
    "args": {
      "External id": 66602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66602, "pid": 5, "tid": 7, "ts": 1716454222370637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300931, "dur": 13,
    "args": {
      "External id": 66602,
      "cbid": 211, "correlation": 66602
    }
  },
  {
    "ph": "s", "id": 66602, "pid": 76337, "tid": -914061504, "ts": 1716454222300931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222370659, "dur": 18,
    "args": {
      "External id": 66624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66624,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66624, "pid": 5, "tid": 7, "ts": 1716454222370659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222300963, "dur": 7,
    "args": {
      "External id": 66624,
      "cbid": 211, "correlation": 66624
    }
  },
  {
    "ph": "s", "id": 66624, "pid": 76337, "tid": -914061504, "ts": 1716454222300963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301058, "dur": 2,
    "args": {
      "External id": 66635,
      "cbid": 251, "correlation": 66635
    }
  },
  {
    "ph": "f", "id": 66635, "pid": 76337, "tid": -914061504, "ts": 1716454222301058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222370678, "dur": 98,
    "args": {
      "External id": 66636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66636,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66636, "pid": 5, "tid": 7, "ts": 1716454222370678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301063, "dur": 14,
    "args": {
      "External id": 66636,
      "cbid": 211, "correlation": 66636
    }
  },
  {
    "ph": "s", "id": 66636, "pid": 76337, "tid": -914061504, "ts": 1716454222301063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301134, "dur": 1,
    "args": {
      "External id": 66647,
      "cbid": 251, "correlation": 66647
    }
  },
  {
    "ph": "f", "id": 66647, "pid": 76337, "tid": -914061504, "ts": 1716454222301134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301138, "dur": 0,
    "args": {
      "External id": 66648,
      "cbid": 251, "correlation": 66648
    }
  },
  {
    "ph": "f", "id": 66648, "pid": 76337, "tid": -914061504, "ts": 1716454222301138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222370778, "dur": 13,
    "args": {
      "External id": 66649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66649,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66649, "pid": 5, "tid": 7, "ts": 1716454222370778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301140, "dur": 13,
    "args": {
      "External id": 66649,
      "cbid": 211, "correlation": 66649
    }
  },
  {
    "ph": "s", "id": 66649, "pid": 76337, "tid": -914061504, "ts": 1716454222301140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222370792, "dur": 6,
    "args": {
      "External id": 66651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66651,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66651, "pid": 5, "tid": 7, "ts": 1716454222370792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301154, "dur": 6,
    "args": {
      "External id": 66651,
      "cbid": 211, "correlation": 66651
    }
  },
  {
    "ph": "s", "id": 66651, "pid": 76337, "tid": -914061504, "ts": 1716454222301154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301214, "dur": 1,
    "args": {
      "External id": 66662,
      "cbid": 251, "correlation": 66662
    }
  },
  {
    "ph": "f", "id": 66662, "pid": 76337, "tid": -914061504, "ts": 1716454222301214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301217, "dur": 0,
    "args": {
      "External id": 66663,
      "cbid": 251, "correlation": 66663
    }
  },
  {
    "ph": "f", "id": 66663, "pid": 76337, "tid": -914061504, "ts": 1716454222301217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222370800, "dur": 9,
    "args": {
      "External id": 66664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66664,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66664, "pid": 5, "tid": 7, "ts": 1716454222370800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301219, "dur": 11,
    "args": {
      "External id": 66664,
      "cbid": 211, "correlation": 66664
    }
  },
  {
    "ph": "s", "id": 66664, "pid": 76337, "tid": -914061504, "ts": 1716454222301219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222370811, "dur": 4,
    "args": {
      "External id": 66666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66666,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66666, "pid": 5, "tid": 7, "ts": 1716454222370811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301232, "dur": 5,
    "args": {
      "External id": 66666,
      "cbid": 211, "correlation": 66666
    }
  },
  {
    "ph": "s", "id": 66666, "pid": 76337, "tid": -914061504, "ts": 1716454222301232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222370816, "dur": 60,
    "args": {
      "External id": 66691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66691,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66691, "pid": 5, "tid": 7, "ts": 1716454222370816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301309, "dur": 12,
    "args": {
      "External id": 66691,
      "cbid": 211, "correlation": 66691
    }
  },
  {
    "ph": "s", "id": 66691, "pid": 76337, "tid": -914061504, "ts": 1716454222301309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301406, "dur": 1,
    "args": {
      "External id": 66709,
      "cbid": 251, "correlation": 66709
    }
  },
  {
    "ph": "f", "id": 66709, "pid": 76337, "tid": -914061504, "ts": 1716454222301406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222370877, "dur": 100,
    "args": {
      "External id": 66711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66711,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66711, "pid": 5, "tid": 7, "ts": 1716454222370877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301413, "dur": 14,
    "args": {
      "External id": 66711,
      "cbid": 211, "correlation": 66711
    }
  },
  {
    "ph": "s", "id": 66711, "pid": 76337, "tid": -914061504, "ts": 1716454222301413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222370978, "dur": 10,
    "args": {
      "External id": 66719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66719,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66719, "pid": 5, "tid": 7, "ts": 1716454222370978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301483, "dur": 12,
    "args": {
      "External id": 66719,
      "cbid": 211, "correlation": 66719
    }
  },
  {
    "ph": "s", "id": 66719, "pid": 76337, "tid": -914061504, "ts": 1716454222301483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222370989, "dur": 22,
    "args": {
      "External id": 66727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66727, "pid": 5, "tid": 7, "ts": 1716454222370989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301525, "dur": 9,
    "args": {
      "External id": 66727,
      "cbid": 211, "correlation": 66727
    }
  },
  {
    "ph": "s", "id": 66727, "pid": 76337, "tid": -914061504, "ts": 1716454222301525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222371013, "dur": 20,
    "args": {
      "External id": 66749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66749,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66749, "pid": 5, "tid": 7, "ts": 1716454222371013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301580, "dur": 10,
    "args": {
      "External id": 66749,
      "cbid": 211, "correlation": 66749
    }
  },
  {
    "ph": "s", "id": 66749, "pid": 76337, "tid": -914061504, "ts": 1716454222301580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301667, "dur": 1,
    "args": {
      "External id": 66765,
      "cbid": 251, "correlation": 66765
    }
  },
  {
    "ph": "f", "id": 66765, "pid": 76337, "tid": -914061504, "ts": 1716454222301667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301672, "dur": 0,
    "args": {
      "External id": 66767,
      "cbid": 251, "correlation": 66767
    }
  },
  {
    "ph": "f", "id": 66767, "pid": 76337, "tid": -914061504, "ts": 1716454222301672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222371034, "dur": 531,
    "args": {
      "External id": 66768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66768,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66768, "pid": 5, "tid": 7, "ts": 1716454222371034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301674, "dur": 13,
    "args": {
      "External id": 66768,
      "cbid": 211, "correlation": 66768
    }
  },
  {
    "ph": "s", "id": 66768, "pid": 76337, "tid": -914061504, "ts": 1716454222301674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222371566, "dur": 71,
    "args": {
      "External id": 66776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66776, "pid": 5, "tid": 7, "ts": 1716454222371566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301742, "dur": 13,
    "args": {
      "External id": 66776,
      "cbid": 211, "correlation": 66776
    }
  },
  {
    "ph": "s", "id": 66776, "pid": 76337, "tid": -914061504, "ts": 1716454222301742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222371639, "dur": 68,
    "args": {
      "External id": 66784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66784, "pid": 5, "tid": 7, "ts": 1716454222371639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301774, "dur": 8,
    "args": {
      "External id": 66784,
      "cbid": 211, "correlation": 66784
    }
  },
  {
    "ph": "s", "id": 66784, "pid": 76337, "tid": -914061504, "ts": 1716454222301774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301855, "dur": 1,
    "args": {
      "External id": 66800,
      "cbid": 251, "correlation": 66800
    }
  },
  {
    "ph": "f", "id": 66800, "pid": 76337, "tid": -914061504, "ts": 1716454222301855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222371709, "dur": 1,
    "args": {
      "External id": 66802,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 66802,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 66802, "pid": 5, "tid": 7, "ts": 1716454222371709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301860, "dur": 10,
    "args": {
      "External id": 66802,
      "cbid": 51, "correlation": 66802
    }
  },
  {
    "ph": "s", "id": 66802, "pid": 76337, "tid": -914061504, "ts": 1716454222301860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222371712, "dur": 290,
    "args": {
      "External id": 66803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66803,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66803, "pid": 5, "tid": 7, "ts": 1716454222371712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301872, "dur": 11,
    "args": {
      "External id": 66803,
      "cbid": 211, "correlation": 66803
    }
  },
  {
    "ph": "s", "id": 66803, "pid": 76337, "tid": -914061504, "ts": 1716454222301872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222372004, "dur": 15,
    "args": {
      "External id": 66811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66811,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66811, "pid": 5, "tid": 7, "ts": 1716454222372004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301914, "dur": 13,
    "args": {
      "External id": 66811,
      "cbid": 211, "correlation": 66811
    }
  },
  {
    "ph": "s", "id": 66811, "pid": 76337, "tid": -914061504, "ts": 1716454222301914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222372020, "dur": 42,
    "args": {
      "External id": 66822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66822, "pid": 5, "tid": 7, "ts": 1716454222372020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222301995, "dur": 14,
    "args": {
      "External id": 66822,
      "cbid": 211, "correlation": 66822
    }
  },
  {
    "ph": "s", "id": 66822, "pid": 76337, "tid": -914061504, "ts": 1716454222301995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302064, "dur": 0,
    "args": {
      "External id": 66834,
      "cbid": 317, "correlation": 66834
    }
  },
  {
    "ph": "f", "id": 66834, "pid": 76337, "tid": -914061504, "ts": 1716454222302064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302065, "dur": 0,
    "args": {
      "External id": 66835,
      "cbid": 203, "correlation": 66835
    }
  },
  {
    "ph": "f", "id": 66835, "pid": 76337, "tid": -914061504, "ts": 1716454222302065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302065, "dur": 0,
    "args": {
      "External id": 66836,
      "cbid": 205, "correlation": 66836
    }
  },
  {
    "ph": "f", "id": 66836, "pid": 76337, "tid": -914061504, "ts": 1716454222302065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222372064, "dur": 13,
    "args": {
      "External id": 66840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66840,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66840, "pid": 5, "tid": 7, "ts": 1716454222372064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302096, "dur": 13,
    "args": {
      "External id": 66840,
      "cbid": 211, "correlation": 66840
    }
  },
  {
    "ph": "s", "id": 66840, "pid": 76337, "tid": -914061504, "ts": 1716454222302096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222372078, "dur": 4,
    "args": {
      "External id": 66842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 66842, "pid": 5, "tid": 7, "ts": 1716454222372078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302114, "dur": 6,
    "args": {
      "External id": 66842,
      "cbid": 211, "correlation": 66842
    }
  },
  {
    "ph": "s", "id": 66842, "pid": 76337, "tid": -914061504, "ts": 1716454222302114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302123, "dur": 0,
    "args": {
      "External id": 66843,
      "cbid": 51, "correlation": 66843
    }
  },
  {
    "ph": "s", "id": 66843, "pid": 76337, "tid": -914061504, "ts": 1716454222302123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222372084, "dur": 108,
    "args": {
      "External id": 66844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66844,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 66844, "pid": 5, "tid": 7, "ts": 1716454222372084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302124, "dur": 6,
    "args": {
      "External id": 66844,
      "cbid": 211, "correlation": 66844
    }
  },
  {
    "ph": "s", "id": 66844, "pid": 76337, "tid": -914061504, "ts": 1716454222302124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222372193, "dur": 17,
    "args": {
      "External id": 66849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66849, "pid": 5, "tid": 7, "ts": 1716454222372193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302153, "dur": 9,
    "args": {
      "External id": 66849,
      "cbid": 211, "correlation": 66849
    }
  },
  {
    "ph": "s", "id": 66849, "pid": 76337, "tid": -914061504, "ts": 1716454222302153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222372212, "dur": 12,
    "args": {
      "External id": 66857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66857,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66857, "pid": 5, "tid": 7, "ts": 1716454222372212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302186, "dur": 8,
    "args": {
      "External id": 66857,
      "cbid": 211, "correlation": 66857
    }
  },
  {
    "ph": "s", "id": 66857, "pid": 76337, "tid": -914061504, "ts": 1716454222302186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222372226, "dur": 28,
    "args": {
      "External id": 66866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66866,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66866, "pid": 5, "tid": 7, "ts": 1716454222372226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302227, "dur": 10,
    "args": {
      "External id": 66866,
      "cbid": 211, "correlation": 66866
    }
  },
  {
    "ph": "s", "id": 66866, "pid": 76337, "tid": -914061504, "ts": 1716454222302227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222372255, "dur": 26,
    "args": {
      "External id": 66886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66886,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 66886, "pid": 5, "tid": 7, "ts": 1716454222372255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302304, "dur": 12,
    "args": {
      "External id": 66886,
      "cbid": 211, "correlation": 66886
    }
  },
  {
    "ph": "s", "id": 66886, "pid": 76337, "tid": -914061504, "ts": 1716454222302304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222372283, "dur": 5,
    "args": {
      "External id": 66898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66898,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 66898, "pid": 5, "tid": 7, "ts": 1716454222372283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302326, "dur": 7,
    "args": {
      "External id": 66898,
      "cbid": 211, "correlation": 66898
    }
  },
  {
    "ph": "s", "id": 66898, "pid": 76337, "tid": -914061504, "ts": 1716454222302326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222372290, "dur": 26,
    "args": {
      "External id": 66901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66901, "pid": 5, "tid": 7, "ts": 1716454222372290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302345, "dur": 7,
    "args": {
      "External id": 66901,
      "cbid": 211, "correlation": 66901
    }
  },
  {
    "ph": "s", "id": 66901, "pid": 76337, "tid": -914061504, "ts": 1716454222302345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222372317, "dur": 18,
    "args": {
      "External id": 66910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66910,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66910, "pid": 5, "tid": 7, "ts": 1716454222372317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302385, "dur": 10,
    "args": {
      "External id": 66910,
      "cbid": 211, "correlation": 66910
    }
  },
  {
    "ph": "s", "id": 66910, "pid": 76337, "tid": -914061504, "ts": 1716454222302385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302438, "dur": 0,
    "args": {
      "External id": 66920,
      "cbid": 317, "correlation": 66920
    }
  },
  {
    "ph": "f", "id": 66920, "pid": 76337, "tid": -914061504, "ts": 1716454222302438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302439, "dur": 0,
    "args": {
      "External id": 66921,
      "cbid": 203, "correlation": 66921
    }
  },
  {
    "ph": "f", "id": 66921, "pid": 76337, "tid": -914061504, "ts": 1716454222302439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302440, "dur": 0,
    "args": {
      "External id": 66922,
      "cbid": 205, "correlation": 66922
    }
  },
  {
    "ph": "f", "id": 66922, "pid": 76337, "tid": -914061504, "ts": 1716454222302440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222372336, "dur": 18,
    "args": {
      "External id": 66926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66926,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66926, "pid": 5, "tid": 7, "ts": 1716454222372336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302459, "dur": 12,
    "args": {
      "External id": 66926,
      "cbid": 211, "correlation": 66926
    }
  },
  {
    "ph": "s", "id": 66926, "pid": 76337, "tid": -914061504, "ts": 1716454222302459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222372356, "dur": 262,
    "args": {
      "External id": 66928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66928,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66928, "pid": 5, "tid": 7, "ts": 1716454222372356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302473, "dur": 6,
    "args": {
      "External id": 66928,
      "cbid": 211, "correlation": 66928
    }
  },
  {
    "ph": "s", "id": 66928, "pid": 76337, "tid": -914061504, "ts": 1716454222302473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222372620, "dur": 1,
    "args": {
      "External id": 66930,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 66930,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 66930, "pid": 5, "tid": 7, "ts": 1716454222372620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302486, "dur": 7,
    "args": {
      "External id": 66930,
      "cbid": 51, "correlation": 66930
    }
  },
  {
    "ph": "s", "id": 66930, "pid": 76337, "tid": -914061504, "ts": 1716454222302486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222372624, "dur": 889,
    "args": {
      "External id": 66931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66931,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66931, "pid": 5, "tid": 7, "ts": 1716454222372624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302495, "dur": 6,
    "args": {
      "External id": 66931,
      "cbid": 211, "correlation": 66931
    }
  },
  {
    "ph": "s", "id": 66931, "pid": 76337, "tid": -914061504, "ts": 1716454222302495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222373515, "dur": 15,
    "args": {
      "External id": 66933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66933,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66933, "pid": 5, "tid": 7, "ts": 1716454222373515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302505, "dur": 5,
    "args": {
      "External id": 66933,
      "cbid": 211, "correlation": 66933
    }
  },
  {
    "ph": "s", "id": 66933, "pid": 76337, "tid": -914061504, "ts": 1716454222302505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222373531, "dur": 16,
    "args": {
      "External id": 66939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66939,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66939, "pid": 5, "tid": 7, "ts": 1716454222373531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302535, "dur": 10,
    "args": {
      "External id": 66939,
      "cbid": 211, "correlation": 66939
    }
  },
  {
    "ph": "s", "id": 66939, "pid": 76337, "tid": -914061504, "ts": 1716454222302535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222373548, "dur": 4,
    "args": {
      "External id": 66947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66947,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 66947, "pid": 5, "tid": 7, "ts": 1716454222373548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302580, "dur": 10,
    "args": {
      "External id": 66947,
      "cbid": 211, "correlation": 66947
    }
  },
  {
    "ph": "s", "id": 66947, "pid": 76337, "tid": -914061504, "ts": 1716454222302580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302649, "dur": 1,
    "args": {
      "External id": 66963,
      "cbid": 251, "correlation": 66963
    }
  },
  {
    "ph": "f", "id": 66963, "pid": 76337, "tid": -914061504, "ts": 1716454222302649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302654, "dur": 0,
    "args": {
      "External id": 66965,
      "cbid": 251, "correlation": 66965
    }
  },
  {
    "ph": "f", "id": 66965, "pid": 76337, "tid": -914061504, "ts": 1716454222302654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222373553, "dur": 13,
    "args": {
      "External id": 66966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66966,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66966, "pid": 5, "tid": 7, "ts": 1716454222373553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302656, "dur": 11,
    "args": {
      "External id": 66966,
      "cbid": 211, "correlation": 66966
    }
  },
  {
    "ph": "s", "id": 66966, "pid": 76337, "tid": -914061504, "ts": 1716454222302656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222373568, "dur": 6,
    "args": {
      "External id": 66968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66968,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 66968, "pid": 5, "tid": 7, "ts": 1716454222373568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302669, "dur": 6,
    "args": {
      "External id": 66968,
      "cbid": 211, "correlation": 66968
    }
  },
  {
    "ph": "s", "id": 66968, "pid": 76337, "tid": -914061504, "ts": 1716454222302669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222373575, "dur": 18,
    "args": {
      "External id": 66978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66978,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 66978, "pid": 5, "tid": 7, "ts": 1716454222373575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302729, "dur": 13,
    "args": {
      "External id": 66978,
      "cbid": 211, "correlation": 66978
    }
  },
  {
    "ph": "s", "id": 66978, "pid": 76337, "tid": -914061504, "ts": 1716454222302729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222373595, "dur": 20,
    "args": {
      "External id": 66998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 66998,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 66998, "pid": 5, "tid": 7, "ts": 1716454222373595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302795, "dur": 11,
    "args": {
      "External id": 66998,
      "cbid": 211, "correlation": 66998
    }
  },
  {
    "ph": "s", "id": 66998, "pid": 76337, "tid": -914061504, "ts": 1716454222302795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222373616, "dur": 4,
    "args": {
      "External id": 67010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67010,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 67010, "pid": 5, "tid": 7, "ts": 1716454222373616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302816, "dur": 6,
    "args": {
      "External id": 67010,
      "cbid": 211, "correlation": 67010
    }
  },
  {
    "ph": "s", "id": 67010, "pid": 76337, "tid": -914061504, "ts": 1716454222302816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222373622, "dur": 18,
    "args": {
      "External id": 67013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67013, "pid": 5, "tid": 7, "ts": 1716454222373622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302834, "dur": 9,
    "args": {
      "External id": 67013,
      "cbid": 211, "correlation": 67013
    }
  },
  {
    "ph": "s", "id": 67013, "pid": 76337, "tid": -914061504, "ts": 1716454222302834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222373641, "dur": 11,
    "args": {
      "External id": 67022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67022,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67022, "pid": 5, "tid": 7, "ts": 1716454222373641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302878, "dur": 10,
    "args": {
      "External id": 67022,
      "cbid": 211, "correlation": 67022
    }
  },
  {
    "ph": "s", "id": 67022, "pid": 76337, "tid": -914061504, "ts": 1716454222302878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302940, "dur": 0,
    "args": {
      "External id": 67032,
      "cbid": 317, "correlation": 67032
    }
  },
  {
    "ph": "f", "id": 67032, "pid": 76337, "tid": -914061504, "ts": 1716454222302940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302940, "dur": 0,
    "args": {
      "External id": 67033,
      "cbid": 203, "correlation": 67033
    }
  },
  {
    "ph": "f", "id": 67033, "pid": 76337, "tid": -914061504, "ts": 1716454222302940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302941, "dur": 0,
    "args": {
      "External id": 67034,
      "cbid": 205, "correlation": 67034
    }
  },
  {
    "ph": "f", "id": 67034, "pid": 76337, "tid": -914061504, "ts": 1716454222302941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222373654, "dur": 13,
    "args": {
      "External id": 67038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67038,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67038, "pid": 5, "tid": 7, "ts": 1716454222373654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302955, "dur": 12,
    "args": {
      "External id": 67038,
      "cbid": 211, "correlation": 67038
    }
  },
  {
    "ph": "s", "id": 67038, "pid": 76337, "tid": -914061504, "ts": 1716454222302955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222373668, "dur": 178,
    "args": {
      "External id": 67040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67040,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67040, "pid": 5, "tid": 7, "ts": 1716454222373668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302970, "dur": 14,
    "args": {
      "External id": 67040,
      "cbid": 211, "correlation": 67040
    }
  },
  {
    "ph": "s", "id": 67040, "pid": 76337, "tid": -914061504, "ts": 1716454222302970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222373849, "dur": 1,
    "args": {
      "External id": 67042,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 67042,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5660377358490566
    }
  },
  {
    "ph": "f", "id": 67042, "pid": 5, "tid": 7, "ts": 1716454222373849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302989, "dur": 7,
    "args": {
      "External id": 67042,
      "cbid": 51, "correlation": 67042
    }
  },
  {
    "ph": "s", "id": 67042, "pid": 76337, "tid": -914061504, "ts": 1716454222302989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222373853, "dur": 712,
    "args": {
      "External id": 67043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67043,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67043, "pid": 5, "tid": 7, "ts": 1716454222373853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222302997, "dur": 6,
    "args": {
      "External id": 67043,
      "cbid": 211, "correlation": 67043
    }
  },
  {
    "ph": "s", "id": 67043, "pid": 76337, "tid": -914061504, "ts": 1716454222302997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222374566, "dur": 12,
    "args": {
      "External id": 67045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67045,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67045, "pid": 5, "tid": 7, "ts": 1716454222374566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303008, "dur": 6,
    "args": {
      "External id": 67045,
      "cbid": 211, "correlation": 67045
    }
  },
  {
    "ph": "s", "id": 67045, "pid": 76337, "tid": -914061504, "ts": 1716454222303008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222374579, "dur": 16,
    "args": {
      "External id": 67051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67051, "pid": 5, "tid": 7, "ts": 1716454222374579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303037, "dur": 8,
    "args": {
      "External id": 67051,
      "cbid": 211, "correlation": 67051
    }
  },
  {
    "ph": "s", "id": 67051, "pid": 76337, "tid": -914061504, "ts": 1716454222303037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303096, "dur": 0,
    "args": {
      "External id": 67061,
      "cbid": 317, "correlation": 67061
    }
  },
  {
    "ph": "f", "id": 67061, "pid": 76337, "tid": -914061504, "ts": 1716454222303096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303097, "dur": 0,
    "args": {
      "External id": 67062,
      "cbid": 203, "correlation": 67062
    }
  },
  {
    "ph": "f", "id": 67062, "pid": 76337, "tid": -914061504, "ts": 1716454222303097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303098, "dur": 0,
    "args": {
      "External id": 67063,
      "cbid": 205, "correlation": 67063
    }
  },
  {
    "ph": "f", "id": 67063, "pid": 76337, "tid": -914061504, "ts": 1716454222303098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222374597, "dur": 18,
    "args": {
      "External id": 67067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67067,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67067, "pid": 5, "tid": 7, "ts": 1716454222374597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303118, "dur": 12,
    "args": {
      "External id": 67067,
      "cbid": 211, "correlation": 67067
    }
  },
  {
    "ph": "s", "id": 67067, "pid": 76337, "tid": -914061504, "ts": 1716454222303118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222374616, "dur": 4,
    "args": {
      "External id": 67069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67069, "pid": 5, "tid": 7, "ts": 1716454222374616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303135, "dur": 6,
    "args": {
      "External id": 67069,
      "cbid": 211, "correlation": 67069
    }
  },
  {
    "ph": "s", "id": 67069, "pid": 76337, "tid": -914061504, "ts": 1716454222303135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303144, "dur": 0,
    "args": {
      "External id": 67070,
      "cbid": 51, "correlation": 67070
    }
  },
  {
    "ph": "s", "id": 67070, "pid": 76337, "tid": -914061504, "ts": 1716454222303144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222374622, "dur": 145,
    "args": {
      "External id": 67071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67071,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 67071, "pid": 5, "tid": 7, "ts": 1716454222374622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303144, "dur": 6,
    "args": {
      "External id": 67071,
      "cbid": 211, "correlation": 67071
    }
  },
  {
    "ph": "s", "id": 67071, "pid": 76337, "tid": -914061504, "ts": 1716454222303144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222374768, "dur": 17,
    "args": {
      "External id": 67076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67076,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67076, "pid": 5, "tid": 7, "ts": 1716454222374768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303173, "dur": 8,
    "args": {
      "External id": 67076,
      "cbid": 211, "correlation": 67076
    }
  },
  {
    "ph": "s", "id": 67076, "pid": 76337, "tid": -914061504, "ts": 1716454222303173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222374787, "dur": 12,
    "args": {
      "External id": 67084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67084,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67084, "pid": 5, "tid": 7, "ts": 1716454222374787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303201, "dur": 8,
    "args": {
      "External id": 67084,
      "cbid": 211, "correlation": 67084
    }
  },
  {
    "ph": "s", "id": 67084, "pid": 76337, "tid": -914061504, "ts": 1716454222303201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222374800, "dur": 11,
    "args": {
      "External id": 67092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67092,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67092, "pid": 5, "tid": 7, "ts": 1716454222374800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303230, "dur": 8,
    "args": {
      "External id": 67092,
      "cbid": 211, "correlation": 67092
    }
  },
  {
    "ph": "s", "id": 67092, "pid": 76337, "tid": -914061504, "ts": 1716454222303230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222374813, "dur": 20,
    "args": {
      "External id": 67112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67112,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 67112, "pid": 5, "tid": 7, "ts": 1716454222374813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303311, "dur": 13,
    "args": {
      "External id": 67112,
      "cbid": 211, "correlation": 67112
    }
  },
  {
    "ph": "s", "id": 67112, "pid": 76337, "tid": -914061504, "ts": 1716454222303311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222374834, "dur": 4,
    "args": {
      "External id": 67124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67124,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 67124, "pid": 5, "tid": 7, "ts": 1716454222374834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303334, "dur": 6,
    "args": {
      "External id": 67124,
      "cbid": 211, "correlation": 67124
    }
  },
  {
    "ph": "s", "id": 67124, "pid": 76337, "tid": -914061504, "ts": 1716454222303334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222374840, "dur": 18,
    "args": {
      "External id": 67127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67127, "pid": 5, "tid": 7, "ts": 1716454222374840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303351, "dur": 7,
    "args": {
      "External id": 67127,
      "cbid": 211, "correlation": 67127
    }
  },
  {
    "ph": "s", "id": 67127, "pid": 76337, "tid": -914061504, "ts": 1716454222303351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303409, "dur": 0,
    "args": {
      "External id": 67138,
      "cbid": 317, "correlation": 67138
    }
  },
  {
    "ph": "f", "id": 67138, "pid": 76337, "tid": -914061504, "ts": 1716454222303409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303409, "dur": 0,
    "args": {
      "External id": 67139,
      "cbid": 203, "correlation": 67139
    }
  },
  {
    "ph": "f", "id": 67139, "pid": 76337, "tid": -914061504, "ts": 1716454222303409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303410, "dur": 0,
    "args": {
      "External id": 67140,
      "cbid": 205, "correlation": 67140
    }
  },
  {
    "ph": "f", "id": 67140, "pid": 76337, "tid": -914061504, "ts": 1716454222303410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222374859, "dur": 12,
    "args": {
      "External id": 67144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67144,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67144, "pid": 5, "tid": 7, "ts": 1716454222374859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303423, "dur": 15,
    "args": {
      "External id": 67144,
      "cbid": 211, "correlation": 67144
    }
  },
  {
    "ph": "s", "id": 67144, "pid": 76337, "tid": -914061504, "ts": 1716454222303423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222374873, "dur": 4,
    "args": {
      "External id": 67146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67146, "pid": 5, "tid": 7, "ts": 1716454222374873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303443, "dur": 6,
    "args": {
      "External id": 67146,
      "cbid": 211, "correlation": 67146
    }
  },
  {
    "ph": "s", "id": 67146, "pid": 76337, "tid": -914061504, "ts": 1716454222303443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303451, "dur": 0,
    "args": {
      "External id": 67147,
      "cbid": 51, "correlation": 67147
    }
  },
  {
    "ph": "s", "id": 67147, "pid": 76337, "tid": -914061504, "ts": 1716454222303451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222374878, "dur": 100,
    "args": {
      "External id": 67148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67148,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 67148, "pid": 5, "tid": 7, "ts": 1716454222374878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303452, "dur": 5,
    "args": {
      "External id": 67148,
      "cbid": 211, "correlation": 67148
    }
  },
  {
    "ph": "s", "id": 67148, "pid": 76337, "tid": -914061504, "ts": 1716454222303452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222374980, "dur": 17,
    "args": {
      "External id": 67153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67153,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67153, "pid": 5, "tid": 7, "ts": 1716454222374980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303479, "dur": 9,
    "args": {
      "External id": 67153,
      "cbid": 211, "correlation": 67153
    }
  },
  {
    "ph": "s", "id": 67153, "pid": 76337, "tid": -914061504, "ts": 1716454222303479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222374998, "dur": 91,
    "args": {
      "External id": 67162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67162,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67162, "pid": 5, "tid": 7, "ts": 1716454222374998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303559, "dur": 14,
    "args": {
      "External id": 67162,
      "cbid": 211, "correlation": 67162
    }
  },
  {
    "ph": "s", "id": 67162, "pid": 76337, "tid": -914061504, "ts": 1716454222303559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222375091, "dur": 32,
    "args": {
      "External id": 67184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67184,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67184, "pid": 5, "tid": 7, "ts": 1716454222375091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303616, "dur": 10,
    "args": {
      "External id": 67184,
      "cbid": 211, "correlation": 67184
    }
  },
  {
    "ph": "s", "id": 67184, "pid": 76337, "tid": -914061504, "ts": 1716454222303616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303706, "dur": 1,
    "args": {
      "External id": 67195,
      "cbid": 251, "correlation": 67195
    }
  },
  {
    "ph": "f", "id": 67195, "pid": 76337, "tid": -914061504, "ts": 1716454222303706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222375124, "dur": 182,
    "args": {
      "External id": 67196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67196,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67196, "pid": 5, "tid": 7, "ts": 1716454222375124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303712, "dur": 13,
    "args": {
      "External id": 67196,
      "cbid": 211, "correlation": 67196
    }
  },
  {
    "ph": "s", "id": 67196, "pid": 76337, "tid": -914061504, "ts": 1716454222303712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303782, "dur": 1,
    "args": {
      "External id": 67207,
      "cbid": 251, "correlation": 67207
    }
  },
  {
    "ph": "f", "id": 67207, "pid": 76337, "tid": -914061504, "ts": 1716454222303782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222375307, "dur": 175,
    "args": {
      "External id": 67208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67208,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67208, "pid": 5, "tid": 7, "ts": 1716454222375307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303786, "dur": 12,
    "args": {
      "External id": 67208,
      "cbid": 211, "correlation": 67208
    }
  },
  {
    "ph": "s", "id": 67208, "pid": 76337, "tid": -914061504, "ts": 1716454222303786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303853, "dur": 1,
    "args": {
      "External id": 67219,
      "cbid": 251, "correlation": 67219
    }
  },
  {
    "ph": "f", "id": 67219, "pid": 76337, "tid": -914061504, "ts": 1716454222303853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222375484, "dur": 173,
    "args": {
      "External id": 67220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67220,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67220, "pid": 5, "tid": 7, "ts": 1716454222375484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303857, "dur": 11,
    "args": {
      "External id": 67220,
      "cbid": 211, "correlation": 67220
    }
  },
  {
    "ph": "s", "id": 67220, "pid": 76337, "tid": -914061504, "ts": 1716454222303857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222375659, "dur": 361,
    "args": {
      "External id": 67245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67245,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67245, "pid": 5, "tid": 7, "ts": 1716454222375659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222303941, "dur": 13,
    "args": {
      "External id": 67245,
      "cbid": 211, "correlation": 67245
    }
  },
  {
    "ph": "s", "id": 67245, "pid": 76337, "tid": -914061504, "ts": 1716454222303941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304050, "dur": 1,
    "args": {
      "External id": 67263,
      "cbid": 251, "correlation": 67263
    }
  },
  {
    "ph": "f", "id": 67263, "pid": 76337, "tid": -914061504, "ts": 1716454222304050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222376021, "dur": 183,
    "args": {
      "External id": 67265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67265,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67265, "pid": 5, "tid": 7, "ts": 1716454222376021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304056, "dur": 13,
    "args": {
      "External id": 67265,
      "cbid": 211, "correlation": 67265
    }
  },
  {
    "ph": "s", "id": 67265, "pid": 76337, "tid": -914061504, "ts": 1716454222304056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222376205, "dur": 20,
    "args": {
      "External id": 67273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67273,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67273, "pid": 5, "tid": 7, "ts": 1716454222376205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304127, "dur": 12,
    "args": {
      "External id": 67273,
      "cbid": 211, "correlation": 67273
    }
  },
  {
    "ph": "s", "id": 67273, "pid": 76337, "tid": -914061504, "ts": 1716454222304127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222376226, "dur": 28,
    "args": {
      "External id": 67281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67281,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67281, "pid": 5, "tid": 7, "ts": 1716454222376226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304165, "dur": 8,
    "args": {
      "External id": 67281,
      "cbid": 211, "correlation": 67281
    }
  },
  {
    "ph": "s", "id": 67281, "pid": 76337, "tid": -914061504, "ts": 1716454222304165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222376255, "dur": 20,
    "args": {
      "External id": 67292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67292, "pid": 5, "tid": 7, "ts": 1716454222376255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304238, "dur": 15,
    "args": {
      "External id": 67292,
      "cbid": 211, "correlation": 67292
    }
  },
  {
    "ph": "s", "id": 67292, "pid": 76337, "tid": -914061504, "ts": 1716454222304238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222376276, "dur": 18,
    "args": {
      "External id": 67314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67314,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67314, "pid": 5, "tid": 7, "ts": 1716454222376276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304274, "dur": 8,
    "args": {
      "External id": 67314,
      "cbid": 211, "correlation": 67314
    }
  },
  {
    "ph": "s", "id": 67314, "pid": 76337, "tid": -914061504, "ts": 1716454222304274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304361, "dur": 2,
    "args": {
      "External id": 67325,
      "cbid": 251, "correlation": 67325
    }
  },
  {
    "ph": "f", "id": 67325, "pid": 76337, "tid": -914061504, "ts": 1716454222304361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222376295, "dur": 97,
    "args": {
      "External id": 67326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67326,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67326, "pid": 5, "tid": 7, "ts": 1716454222376295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304366, "dur": 14,
    "args": {
      "External id": 67326,
      "cbid": 211, "correlation": 67326
    }
  },
  {
    "ph": "s", "id": 67326, "pid": 76337, "tid": -914061504, "ts": 1716454222304366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304435, "dur": 1,
    "args": {
      "External id": 67337,
      "cbid": 251, "correlation": 67337
    }
  },
  {
    "ph": "f", "id": 67337, "pid": 76337, "tid": -914061504, "ts": 1716454222304435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304438, "dur": 0,
    "args": {
      "External id": 67338,
      "cbid": 251, "correlation": 67338
    }
  },
  {
    "ph": "f", "id": 67338, "pid": 76337, "tid": -914061504, "ts": 1716454222304438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222376394, "dur": 13,
    "args": {
      "External id": 67339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67339,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67339, "pid": 5, "tid": 7, "ts": 1716454222376394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304440, "dur": 12,
    "args": {
      "External id": 67339,
      "cbid": 211, "correlation": 67339
    }
  },
  {
    "ph": "s", "id": 67339, "pid": 76337, "tid": -914061504, "ts": 1716454222304440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222376408, "dur": 6,
    "args": {
      "External id": 67341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67341,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67341, "pid": 5, "tid": 7, "ts": 1716454222376408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304453, "dur": 6,
    "args": {
      "External id": 67341,
      "cbid": 211, "correlation": 67341
    }
  },
  {
    "ph": "s", "id": 67341, "pid": 76337, "tid": -914061504, "ts": 1716454222304453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304512, "dur": 1,
    "args": {
      "External id": 67352,
      "cbid": 251, "correlation": 67352
    }
  },
  {
    "ph": "f", "id": 67352, "pid": 76337, "tid": -914061504, "ts": 1716454222304512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304515, "dur": 0,
    "args": {
      "External id": 67353,
      "cbid": 251, "correlation": 67353
    }
  },
  {
    "ph": "f", "id": 67353, "pid": 76337, "tid": -914061504, "ts": 1716454222304515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222376416, "dur": 9,
    "args": {
      "External id": 67354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67354,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67354, "pid": 5, "tid": 7, "ts": 1716454222376416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304517, "dur": 12,
    "args": {
      "External id": 67354,
      "cbid": 211, "correlation": 67354
    }
  },
  {
    "ph": "s", "id": 67354, "pid": 76337, "tid": -914061504, "ts": 1716454222304517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222376426, "dur": 4,
    "args": {
      "External id": 67356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67356,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67356, "pid": 5, "tid": 7, "ts": 1716454222376426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304531, "dur": 5,
    "args": {
      "External id": 67356,
      "cbid": 211, "correlation": 67356
    }
  },
  {
    "ph": "s", "id": 67356, "pid": 76337, "tid": -914061504, "ts": 1716454222304531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222376431, "dur": 60,
    "args": {
      "External id": 67381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67381,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67381, "pid": 5, "tid": 7, "ts": 1716454222376431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304610, "dur": 12,
    "args": {
      "External id": 67381,
      "cbid": 211, "correlation": 67381
    }
  },
  {
    "ph": "s", "id": 67381, "pid": 76337, "tid": -914061504, "ts": 1716454222304610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304709, "dur": 1,
    "args": {
      "External id": 67399,
      "cbid": 251, "correlation": 67399
    }
  },
  {
    "ph": "f", "id": 67399, "pid": 76337, "tid": -914061504, "ts": 1716454222304709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222376493, "dur": 99,
    "args": {
      "External id": 67401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67401,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67401, "pid": 5, "tid": 7, "ts": 1716454222376493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304715, "dur": 13,
    "args": {
      "External id": 67401,
      "cbid": 211, "correlation": 67401
    }
  },
  {
    "ph": "s", "id": 67401, "pid": 76337, "tid": -914061504, "ts": 1716454222304715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222376594, "dur": 10,
    "args": {
      "External id": 67409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67409,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67409, "pid": 5, "tid": 7, "ts": 1716454222376594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304783, "dur": 13,
    "args": {
      "External id": 67409,
      "cbid": 211, "correlation": 67409
    }
  },
  {
    "ph": "s", "id": 67409, "pid": 76337, "tid": -914061504, "ts": 1716454222304783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222376605, "dur": 22,
    "args": {
      "External id": 67417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67417, "pid": 5, "tid": 7, "ts": 1716454222376605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304825, "dur": 9,
    "args": {
      "External id": 67417,
      "cbid": 211, "correlation": 67417
    }
  },
  {
    "ph": "s", "id": 67417, "pid": 76337, "tid": -914061504, "ts": 1716454222304825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222376629, "dur": 19,
    "args": {
      "External id": 67439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67439,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67439, "pid": 5, "tid": 7, "ts": 1716454222376629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304876, "dur": 10,
    "args": {
      "External id": 67439,
      "cbid": 211, "correlation": 67439
    }
  },
  {
    "ph": "s", "id": 67439, "pid": 76337, "tid": -914061504, "ts": 1716454222304876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304968, "dur": 1,
    "args": {
      "External id": 67455,
      "cbid": 251, "correlation": 67455
    }
  },
  {
    "ph": "f", "id": 67455, "pid": 76337, "tid": -914061504, "ts": 1716454222304968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304972, "dur": 8,
    "args": {
      "External id": 67457,
      "cbid": 251, "correlation": 67457
    }
  },
  {
    "ph": "f", "id": 67457, "pid": 76337, "tid": -914061504, "ts": 1716454222304972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222376649, "dur": 527,
    "args": {
      "External id": 67458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67458,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67458, "pid": 5, "tid": 7, "ts": 1716454222376649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222304982, "dur": 14,
    "args": {
      "External id": 67458,
      "cbid": 211, "correlation": 67458
    }
  },
  {
    "ph": "s", "id": 67458, "pid": 76337, "tid": -914061504, "ts": 1716454222304982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222377178, "dur": 70,
    "args": {
      "External id": 67466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67466, "pid": 5, "tid": 7, "ts": 1716454222377178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305050, "dur": 13,
    "args": {
      "External id": 67466,
      "cbid": 211, "correlation": 67466
    }
  },
  {
    "ph": "s", "id": 67466, "pid": 76337, "tid": -914061504, "ts": 1716454222305050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222377250, "dur": 68,
    "args": {
      "External id": 67474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67474, "pid": 5, "tid": 7, "ts": 1716454222377250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305080, "dur": 8,
    "args": {
      "External id": 67474,
      "cbid": 211, "correlation": 67474
    }
  },
  {
    "ph": "s", "id": 67474, "pid": 76337, "tid": -914061504, "ts": 1716454222305080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305160, "dur": 1,
    "args": {
      "External id": 67490,
      "cbid": 251, "correlation": 67490
    }
  },
  {
    "ph": "f", "id": 67490, "pid": 76337, "tid": -914061504, "ts": 1716454222305160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222377320, "dur": 1,
    "args": {
      "External id": 67492,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 67492,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 67492, "pid": 5, "tid": 7, "ts": 1716454222377320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305165, "dur": 9,
    "args": {
      "External id": 67492,
      "cbid": 51, "correlation": 67492
    }
  },
  {
    "ph": "s", "id": 67492, "pid": 76337, "tid": -914061504, "ts": 1716454222305165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222377324, "dur": 291,
    "args": {
      "External id": 67493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67493,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67493, "pid": 5, "tid": 7, "ts": 1716454222377324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305176, "dur": 11,
    "args": {
      "External id": 67493,
      "cbid": 211, "correlation": 67493
    }
  },
  {
    "ph": "s", "id": 67493, "pid": 76337, "tid": -914061504, "ts": 1716454222305176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222377617, "dur": 14,
    "args": {
      "External id": 67501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67501,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67501, "pid": 5, "tid": 7, "ts": 1716454222377617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305219, "dur": 10,
    "args": {
      "External id": 67501,
      "cbid": 211, "correlation": 67501
    }
  },
  {
    "ph": "s", "id": 67501, "pid": 76337, "tid": -914061504, "ts": 1716454222305219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222377632, "dur": 41,
    "args": {
      "External id": 67512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67512,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67512, "pid": 5, "tid": 7, "ts": 1716454222377632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305287, "dur": 12,
    "args": {
      "External id": 67512,
      "cbid": 211, "correlation": 67512
    }
  },
  {
    "ph": "s", "id": 67512, "pid": 76337, "tid": -914061504, "ts": 1716454222305287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305354, "dur": 0,
    "args": {
      "External id": 67524,
      "cbid": 317, "correlation": 67524
    }
  },
  {
    "ph": "f", "id": 67524, "pid": 76337, "tid": -914061504, "ts": 1716454222305354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305355, "dur": 0,
    "args": {
      "External id": 67525,
      "cbid": 203, "correlation": 67525
    }
  },
  {
    "ph": "f", "id": 67525, "pid": 76337, "tid": -914061504, "ts": 1716454222305355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305356, "dur": 0,
    "args": {
      "External id": 67526,
      "cbid": 205, "correlation": 67526
    }
  },
  {
    "ph": "f", "id": 67526, "pid": 76337, "tid": -914061504, "ts": 1716454222305356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222377675, "dur": 13,
    "args": {
      "External id": 67530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67530,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67530, "pid": 5, "tid": 7, "ts": 1716454222377675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305381, "dur": 13,
    "args": {
      "External id": 67530,
      "cbid": 211, "correlation": 67530
    }
  },
  {
    "ph": "s", "id": 67530, "pid": 76337, "tid": -914061504, "ts": 1716454222305381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222377690, "dur": 4,
    "args": {
      "External id": 67532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67532, "pid": 5, "tid": 7, "ts": 1716454222377690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305399, "dur": 6,
    "args": {
      "External id": 67532,
      "cbid": 211, "correlation": 67532
    }
  },
  {
    "ph": "s", "id": 67532, "pid": 76337, "tid": -914061504, "ts": 1716454222305399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305409, "dur": 0,
    "args": {
      "External id": 67533,
      "cbid": 51, "correlation": 67533
    }
  },
  {
    "ph": "s", "id": 67533, "pid": 76337, "tid": -914061504, "ts": 1716454222305409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222377695, "dur": 107,
    "args": {
      "External id": 67534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67534,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 67534, "pid": 5, "tid": 7, "ts": 1716454222377695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305410, "dur": 5,
    "args": {
      "External id": 67534,
      "cbid": 211, "correlation": 67534
    }
  },
  {
    "ph": "s", "id": 67534, "pid": 76337, "tid": -914061504, "ts": 1716454222305410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222377803, "dur": 18,
    "args": {
      "External id": 67539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67539,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67539, "pid": 5, "tid": 7, "ts": 1716454222377803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305437, "dur": 8,
    "args": {
      "External id": 67539,
      "cbid": 211, "correlation": 67539
    }
  },
  {
    "ph": "s", "id": 67539, "pid": 76337, "tid": -914061504, "ts": 1716454222305437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222377823, "dur": 12,
    "args": {
      "External id": 67547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67547,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67547, "pid": 5, "tid": 7, "ts": 1716454222377823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305470, "dur": 8,
    "args": {
      "External id": 67547,
      "cbid": 211, "correlation": 67547
    }
  },
  {
    "ph": "s", "id": 67547, "pid": 76337, "tid": -914061504, "ts": 1716454222305470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222377836, "dur": 61,
    "args": {
      "External id": 67558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67558,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67558, "pid": 5, "tid": 7, "ts": 1716454222377836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305537, "dur": 12,
    "args": {
      "External id": 67558,
      "cbid": 211, "correlation": 67558
    }
  },
  {
    "ph": "s", "id": 67558, "pid": 76337, "tid": -914061504, "ts": 1716454222305537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305593, "dur": 0,
    "args": {
      "External id": 67568,
      "cbid": 317, "correlation": 67568
    }
  },
  {
    "ph": "f", "id": 67568, "pid": 76337, "tid": -914061504, "ts": 1716454222305593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305594, "dur": 0,
    "args": {
      "External id": 67569,
      "cbid": 203, "correlation": 67569
    }
  },
  {
    "ph": "f", "id": 67569, "pid": 76337, "tid": -914061504, "ts": 1716454222305594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305594, "dur": 0,
    "args": {
      "External id": 67570,
      "cbid": 205, "correlation": 67570
    }
  },
  {
    "ph": "f", "id": 67570, "pid": 76337, "tid": -914061504, "ts": 1716454222305594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222377898, "dur": 39,
    "args": {
      "External id": 67574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67574,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67574, "pid": 5, "tid": 7, "ts": 1716454222377898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305616, "dur": 12,
    "args": {
      "External id": 67574,
      "cbid": 211, "correlation": 67574
    }
  },
  {
    "ph": "s", "id": 67574, "pid": 76337, "tid": -914061504, "ts": 1716454222305616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222377938, "dur": 178,
    "args": {
      "External id": 67576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67576,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67576, "pid": 5, "tid": 7, "ts": 1716454222377938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305631, "dur": 5,
    "args": {
      "External id": 67576,
      "cbid": 211, "correlation": 67576
    }
  },
  {
    "ph": "s", "id": 67576, "pid": 76337, "tid": -914061504, "ts": 1716454222305631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222378117, "dur": 2105,
    "args": {
      "External id": 67578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67578,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67578, "pid": 5, "tid": 7, "ts": 1716454222378117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305643, "dur": 8,
    "args": {
      "External id": 67578,
      "cbid": 211, "correlation": 67578
    }
  },
  {
    "ph": "s", "id": 67578, "pid": 76337, "tid": -914061504, "ts": 1716454222305643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222380224, "dur": 40,
    "args": {
      "External id": 67580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67580,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67580, "pid": 5, "tid": 7, "ts": 1716454222380224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305655, "dur": 7,
    "args": {
      "External id": 67580,
      "cbid": 211, "correlation": 67580
    }
  },
  {
    "ph": "s", "id": 67580, "pid": 76337, "tid": -914061504, "ts": 1716454222305655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222380265, "dur": 64,
    "args": {
      "External id": 67586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67586,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67586, "pid": 5, "tid": 7, "ts": 1716454222380265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305686, "dur": 8,
    "args": {
      "External id": 67586,
      "cbid": 211, "correlation": 67586
    }
  },
  {
    "ph": "s", "id": 67586, "pid": 76337, "tid": -914061504, "ts": 1716454222305686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222380330, "dur": 89,
    "args": {
      "External id": 67595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67595,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67595, "pid": 5, "tid": 7, "ts": 1716454222380330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305777, "dur": 13,
    "args": {
      "External id": 67595,
      "cbid": 211, "correlation": 67595
    }
  },
  {
    "ph": "s", "id": 67595, "pid": 76337, "tid": -914061504, "ts": 1716454222305777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222380421, "dur": 78,
    "args": {
      "External id": 67615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67615,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 67615, "pid": 5, "tid": 7, "ts": 1716454222380421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305848, "dur": 11,
    "args": {
      "External id": 67615,
      "cbid": 211, "correlation": 67615
    }
  },
  {
    "ph": "s", "id": 67615, "pid": 76337, "tid": -914061504, "ts": 1716454222305848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222380500, "dur": 5,
    "args": {
      "External id": 67627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67627,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 67627, "pid": 5, "tid": 7, "ts": 1716454222380500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305868, "dur": 6,
    "args": {
      "External id": 67627,
      "cbid": 211, "correlation": 67627
    }
  },
  {
    "ph": "s", "id": 67627, "pid": 76337, "tid": -914061504, "ts": 1716454222305868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222380507, "dur": 87,
    "args": {
      "External id": 67630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67630,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67630, "pid": 5, "tid": 7, "ts": 1716454222380507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305887, "dur": 6,
    "args": {
      "External id": 67630,
      "cbid": 211, "correlation": 67630
    }
  },
  {
    "ph": "s", "id": 67630, "pid": 76337, "tid": -914061504, "ts": 1716454222305887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222380595, "dur": 55,
    "args": {
      "External id": 67639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67639,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67639, "pid": 5, "tid": 7, "ts": 1716454222380595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305926, "dur": 10,
    "args": {
      "External id": 67639,
      "cbid": 211, "correlation": 67639
    }
  },
  {
    "ph": "s", "id": 67639, "pid": 76337, "tid": -914061504, "ts": 1716454222305926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305985, "dur": 0,
    "args": {
      "External id": 67649,
      "cbid": 317, "correlation": 67649
    }
  },
  {
    "ph": "f", "id": 67649, "pid": 76337, "tid": -914061504, "ts": 1716454222305985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305986, "dur": 0,
    "args": {
      "External id": 67650,
      "cbid": 203, "correlation": 67650
    }
  },
  {
    "ph": "f", "id": 67650, "pid": 76337, "tid": -914061504, "ts": 1716454222305986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222305987, "dur": 0,
    "args": {
      "External id": 67651,
      "cbid": 205, "correlation": 67651
    }
  },
  {
    "ph": "f", "id": 67651, "pid": 76337, "tid": -914061504, "ts": 1716454222305987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222380651, "dur": 58,
    "args": {
      "External id": 67655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67655, "pid": 5, "tid": 7, "ts": 1716454222380651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306007, "dur": 13,
    "args": {
      "External id": 67655,
      "cbid": 211, "correlation": 67655
    }
  },
  {
    "ph": "s", "id": 67655, "pid": 76337, "tid": -914061504, "ts": 1716454222306007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222380710, "dur": 133,
    "args": {
      "External id": 67657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67657,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67657, "pid": 5, "tid": 7, "ts": 1716454222380710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306022, "dur": 5,
    "args": {
      "External id": 67657,
      "cbid": 211, "correlation": 67657
    }
  },
  {
    "ph": "s", "id": 67657, "pid": 76337, "tid": -914061504, "ts": 1716454222306022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222380845, "dur": 2078,
    "args": {
      "External id": 67659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67659,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67659, "pid": 5, "tid": 7, "ts": 1716454222380845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306034, "dur": 7,
    "args": {
      "External id": 67659,
      "cbid": 211, "correlation": 67659
    }
  },
  {
    "ph": "s", "id": 67659, "pid": 76337, "tid": -914061504, "ts": 1716454222306034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222382924, "dur": 22,
    "args": {
      "External id": 67661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67661, "pid": 5, "tid": 7, "ts": 1716454222382924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306045, "dur": 5,
    "args": {
      "External id": 67661,
      "cbid": 211, "correlation": 67661
    }
  },
  {
    "ph": "s", "id": 67661, "pid": 76337, "tid": -914061504, "ts": 1716454222306045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222382947, "dur": 36,
    "args": {
      "External id": 67667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67667, "pid": 5, "tid": 7, "ts": 1716454222382947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306073, "dur": 8,
    "args": {
      "External id": 67667,
      "cbid": 211, "correlation": 67667
    }
  },
  {
    "ph": "s", "id": 67667, "pid": 76337, "tid": -914061504, "ts": 1716454222306073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222382984, "dur": 4,
    "args": {
      "External id": 67675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67675,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 67675, "pid": 5, "tid": 7, "ts": 1716454222382984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306117, "dur": 9,
    "args": {
      "External id": 67675,
      "cbid": 211, "correlation": 67675
    }
  },
  {
    "ph": "s", "id": 67675, "pid": 76337, "tid": -914061504, "ts": 1716454222306117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306182, "dur": 1,
    "args": {
      "External id": 67691,
      "cbid": 251, "correlation": 67691
    }
  },
  {
    "ph": "f", "id": 67691, "pid": 76337, "tid": -914061504, "ts": 1716454222306182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306188, "dur": 0,
    "args": {
      "External id": 67693,
      "cbid": 251, "correlation": 67693
    }
  },
  {
    "ph": "f", "id": 67693, "pid": 76337, "tid": -914061504, "ts": 1716454222306188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222382989, "dur": 13,
    "args": {
      "External id": 67694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67694,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 67694, "pid": 5, "tid": 7, "ts": 1716454222382989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306190, "dur": 11,
    "args": {
      "External id": 67694,
      "cbid": 211, "correlation": 67694
    }
  },
  {
    "ph": "s", "id": 67694, "pid": 76337, "tid": -914061504, "ts": 1716454222306190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222383004, "dur": 5,
    "args": {
      "External id": 67696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67696,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 67696, "pid": 5, "tid": 7, "ts": 1716454222383004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306204, "dur": 5,
    "args": {
      "External id": 67696,
      "cbid": 211, "correlation": 67696
    }
  },
  {
    "ph": "s", "id": 67696, "pid": 76337, "tid": -914061504, "ts": 1716454222306204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222383011, "dur": 32,
    "args": {
      "External id": 67706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67706, "pid": 5, "tid": 7, "ts": 1716454222383011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306261, "dur": 12,
    "args": {
      "External id": 67706,
      "cbid": 211, "correlation": 67706
    }
  },
  {
    "ph": "s", "id": 67706, "pid": 76337, "tid": -914061504, "ts": 1716454222306261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222383044, "dur": 34,
    "args": {
      "External id": 67726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67726,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 67726, "pid": 5, "tid": 7, "ts": 1716454222383044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306330, "dur": 11,
    "args": {
      "External id": 67726,
      "cbid": 211, "correlation": 67726
    }
  },
  {
    "ph": "s", "id": 67726, "pid": 76337, "tid": -914061504, "ts": 1716454222306330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222383079, "dur": 4,
    "args": {
      "External id": 67738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67738,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 67738, "pid": 5, "tid": 7, "ts": 1716454222383079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306352, "dur": 6,
    "args": {
      "External id": 67738,
      "cbid": 211, "correlation": 67738
    }
  },
  {
    "ph": "s", "id": 67738, "pid": 76337, "tid": -914061504, "ts": 1716454222306352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222383085, "dur": 31,
    "args": {
      "External id": 67741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67741, "pid": 5, "tid": 7, "ts": 1716454222383085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306370, "dur": 7,
    "args": {
      "External id": 67741,
      "cbid": 211, "correlation": 67741
    }
  },
  {
    "ph": "s", "id": 67741, "pid": 76337, "tid": -914061504, "ts": 1716454222306370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222383118, "dur": 21,
    "args": {
      "External id": 67750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67750,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67750, "pid": 5, "tid": 7, "ts": 1716454222383118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306411, "dur": 10,
    "args": {
      "External id": 67750,
      "cbid": 211, "correlation": 67750
    }
  },
  {
    "ph": "s", "id": 67750, "pid": 76337, "tid": -914061504, "ts": 1716454222306411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306474, "dur": 0,
    "args": {
      "External id": 67760,
      "cbid": 317, "correlation": 67760
    }
  },
  {
    "ph": "f", "id": 67760, "pid": 76337, "tid": -914061504, "ts": 1716454222306474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306475, "dur": 0,
    "args": {
      "External id": 67761,
      "cbid": 203, "correlation": 67761
    }
  },
  {
    "ph": "f", "id": 67761, "pid": 76337, "tid": -914061504, "ts": 1716454222306475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306475, "dur": 0,
    "args": {
      "External id": 67762,
      "cbid": 205, "correlation": 67762
    }
  },
  {
    "ph": "f", "id": 67762, "pid": 76337, "tid": -914061504, "ts": 1716454222306475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222383140, "dur": 23,
    "args": {
      "External id": 67766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67766, "pid": 5, "tid": 7, "ts": 1716454222383140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306491, "dur": 13,
    "args": {
      "External id": 67766,
      "cbid": 211, "correlation": 67766
    }
  },
  {
    "ph": "s", "id": 67766, "pid": 76337, "tid": -914061504, "ts": 1716454222306491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222383165, "dur": 48,
    "args": {
      "External id": 67768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67768, "pid": 5, "tid": 7, "ts": 1716454222383165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306507, "dur": 5,
    "args": {
      "External id": 67768,
      "cbid": 211, "correlation": 67768
    }
  },
  {
    "ph": "s", "id": 67768, "pid": 76337, "tid": -914061504, "ts": 1716454222306507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222383214, "dur": 709,
    "args": {
      "External id": 67770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67770,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67770, "pid": 5, "tid": 7, "ts": 1716454222383214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306518, "dur": 6,
    "args": {
      "External id": 67770,
      "cbid": 211, "correlation": 67770
    }
  },
  {
    "ph": "s", "id": 67770, "pid": 76337, "tid": -914061504, "ts": 1716454222306518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222383925, "dur": 24,
    "args": {
      "External id": 67772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67772, "pid": 5, "tid": 7, "ts": 1716454222383925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306528, "dur": 5,
    "args": {
      "External id": 67772,
      "cbid": 211, "correlation": 67772
    }
  },
  {
    "ph": "s", "id": 67772, "pid": 76337, "tid": -914061504, "ts": 1716454222306528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222383950, "dur": 34,
    "args": {
      "External id": 67778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67778, "pid": 5, "tid": 7, "ts": 1716454222383950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306556, "dur": 8,
    "args": {
      "External id": 67778,
      "cbid": 211, "correlation": 67778
    }
  },
  {
    "ph": "s", "id": 67778, "pid": 76337, "tid": -914061504, "ts": 1716454222306556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306618, "dur": 0,
    "args": {
      "External id": 67788,
      "cbid": 317, "correlation": 67788
    }
  },
  {
    "ph": "f", "id": 67788, "pid": 76337, "tid": -914061504, "ts": 1716454222306618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306619, "dur": 0,
    "args": {
      "External id": 67789,
      "cbid": 203, "correlation": 67789
    }
  },
  {
    "ph": "f", "id": 67789, "pid": 76337, "tid": -914061504, "ts": 1716454222306619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306619, "dur": 0,
    "args": {
      "External id": 67790,
      "cbid": 205, "correlation": 67790
    }
  },
  {
    "ph": "f", "id": 67790, "pid": 76337, "tid": -914061504, "ts": 1716454222306619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222383986, "dur": 58,
    "args": {
      "External id": 67794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67794,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67794, "pid": 5, "tid": 7, "ts": 1716454222383986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306632, "dur": 12,
    "args": {
      "External id": 67794,
      "cbid": 211, "correlation": 67794
    }
  },
  {
    "ph": "s", "id": 67794, "pid": 76337, "tid": -914061504, "ts": 1716454222306632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222384045, "dur": 293,
    "args": {
      "External id": 67796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67796,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67796, "pid": 5, "tid": 7, "ts": 1716454222384045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306650, "dur": 8,
    "args": {
      "External id": 67796,
      "cbid": 211, "correlation": 67796
    }
  },
  {
    "ph": "s", "id": 67796, "pid": 76337, "tid": -914061504, "ts": 1716454222306650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222384339, "dur": 22,
    "args": {
      "External id": 67798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67798,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67798, "pid": 5, "tid": 7, "ts": 1716454222384339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306662, "dur": 5,
    "args": {
      "External id": 67798,
      "cbid": 211, "correlation": 67798
    }
  },
  {
    "ph": "s", "id": 67798, "pid": 76337, "tid": -914061504, "ts": 1716454222306662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222384363, "dur": 34,
    "args": {
      "External id": 67804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67804,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67804, "pid": 5, "tid": 7, "ts": 1716454222384363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306687, "dur": 8,
    "args": {
      "External id": 67804,
      "cbid": 211, "correlation": 67804
    }
  },
  {
    "ph": "s", "id": 67804, "pid": 76337, "tid": -914061504, "ts": 1716454222306687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222384399, "dur": 27,
    "args": {
      "External id": 67812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67812,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67812, "pid": 5, "tid": 7, "ts": 1716454222384399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306716, "dur": 8,
    "args": {
      "External id": 67812,
      "cbid": 211, "correlation": 67812
    }
  },
  {
    "ph": "s", "id": 67812, "pid": 76337, "tid": -914061504, "ts": 1716454222306716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222384428, "dur": 20,
    "args": {
      "External id": 67820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67820,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67820, "pid": 5, "tid": 7, "ts": 1716454222384428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306745, "dur": 9,
    "args": {
      "External id": 67820,
      "cbid": 211, "correlation": 67820
    }
  },
  {
    "ph": "s", "id": 67820, "pid": 76337, "tid": -914061504, "ts": 1716454222306745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222384449, "dur": 32,
    "args": {
      "External id": 67840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67840,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 67840, "pid": 5, "tid": 7, "ts": 1716454222384449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306826, "dur": 12,
    "args": {
      "External id": 67840,
      "cbid": 211, "correlation": 67840
    }
  },
  {
    "ph": "s", "id": 67840, "pid": 76337, "tid": -914061504, "ts": 1716454222306826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222384483, "dur": 5,
    "args": {
      "External id": 67852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67852,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 67852, "pid": 5, "tid": 7, "ts": 1716454222384483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306848, "dur": 9,
    "args": {
      "External id": 67852,
      "cbid": 211, "correlation": 67852
    }
  },
  {
    "ph": "s", "id": 67852, "pid": 76337, "tid": -914061504, "ts": 1716454222306848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222384489, "dur": 33,
    "args": {
      "External id": 67855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67855, "pid": 5, "tid": 7, "ts": 1716454222384489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306870, "dur": 7,
    "args": {
      "External id": 67855,
      "cbid": 211, "correlation": 67855
    }
  },
  {
    "ph": "s", "id": 67855, "pid": 76337, "tid": -914061504, "ts": 1716454222306870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306928, "dur": 0,
    "args": {
      "External id": 67866,
      "cbid": 317, "correlation": 67866
    }
  },
  {
    "ph": "f", "id": 67866, "pid": 76337, "tid": -914061504, "ts": 1716454222306928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306928, "dur": 0,
    "args": {
      "External id": 67867,
      "cbid": 203, "correlation": 67867
    }
  },
  {
    "ph": "f", "id": 67867, "pid": 76337, "tid": -914061504, "ts": 1716454222306928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306929, "dur": 0,
    "args": {
      "External id": 67868,
      "cbid": 205, "correlation": 67868
    }
  },
  {
    "ph": "f", "id": 67868, "pid": 76337, "tid": -914061504, "ts": 1716454222306929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222384523, "dur": 23,
    "args": {
      "External id": 67872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67872,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67872, "pid": 5, "tid": 7, "ts": 1716454222384523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306943, "dur": 12,
    "args": {
      "External id": 67872,
      "cbid": 211, "correlation": 67872
    }
  },
  {
    "ph": "s", "id": 67872, "pid": 76337, "tid": -914061504, "ts": 1716454222306943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222384548, "dur": 116,
    "args": {
      "External id": 67874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67874,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67874, "pid": 5, "tid": 7, "ts": 1716454222384548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306962, "dur": 6,
    "args": {
      "External id": 67874,
      "cbid": 211, "correlation": 67874
    }
  },
  {
    "ph": "s", "id": 67874, "pid": 76337, "tid": -914061504, "ts": 1716454222306962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222384665, "dur": 23,
    "args": {
      "External id": 67876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67876,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67876, "pid": 5, "tid": 7, "ts": 1716454222384665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222306972, "dur": 13,
    "args": {
      "External id": 67876,
      "cbid": 211, "correlation": 67876
    }
  },
  {
    "ph": "s", "id": 67876, "pid": 76337, "tid": -914061504, "ts": 1716454222306972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222384689, "dur": 34,
    "args": {
      "External id": 67882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67882, "pid": 5, "tid": 7, "ts": 1716454222384689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307008, "dur": 10,
    "args": {
      "External id": 67882,
      "cbid": 211, "correlation": 67882
    }
  },
  {
    "ph": "s", "id": 67882, "pid": 76337, "tid": -914061504, "ts": 1716454222307008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222384725, "dur": 211,
    "args": {
      "External id": 67891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67891,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67891, "pid": 5, "tid": 7, "ts": 1716454222384725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307091, "dur": 15,
    "args": {
      "External id": 67891,
      "cbid": 211, "correlation": 67891
    }
  },
  {
    "ph": "s", "id": 67891, "pid": 76337, "tid": -914061504, "ts": 1716454222307091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222384937, "dur": 71,
    "args": {
      "External id": 67913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67913,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67913, "pid": 5, "tid": 7, "ts": 1716454222384937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307149, "dur": 10,
    "args": {
      "External id": 67913,
      "cbid": 211, "correlation": 67913
    }
  },
  {
    "ph": "s", "id": 67913, "pid": 76337, "tid": -914061504, "ts": 1716454222307149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307243, "dur": 2,
    "args": {
      "External id": 67924,
      "cbid": 251, "correlation": 67924
    }
  },
  {
    "ph": "f", "id": 67924, "pid": 76337, "tid": -914061504, "ts": 1716454222307243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222385009, "dur": 164,
    "args": {
      "External id": 67925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67925,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67925, "pid": 5, "tid": 7, "ts": 1716454222385009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307248, "dur": 13,
    "args": {
      "External id": 67925,
      "cbid": 211, "correlation": 67925
    }
  },
  {
    "ph": "s", "id": 67925, "pid": 76337, "tid": -914061504, "ts": 1716454222307248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307320, "dur": 1,
    "args": {
      "External id": 67936,
      "cbid": 251, "correlation": 67936
    }
  },
  {
    "ph": "f", "id": 67936, "pid": 76337, "tid": -914061504, "ts": 1716454222307320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222385175, "dur": 157,
    "args": {
      "External id": 67937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67937,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67937, "pid": 5, "tid": 7, "ts": 1716454222385175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307325, "dur": 11,
    "args": {
      "External id": 67937,
      "cbid": 211, "correlation": 67937
    }
  },
  {
    "ph": "s", "id": 67937, "pid": 76337, "tid": -914061504, "ts": 1716454222307325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307389, "dur": 1,
    "args": {
      "External id": 67948,
      "cbid": 251, "correlation": 67948
    }
  },
  {
    "ph": "f", "id": 67948, "pid": 76337, "tid": -914061504, "ts": 1716454222307389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222385333, "dur": 158,
    "args": {
      "External id": 67949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67949,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 67949, "pid": 5, "tid": 7, "ts": 1716454222385333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307393, "dur": 11,
    "args": {
      "External id": 67949,
      "cbid": 211, "correlation": 67949
    }
  },
  {
    "ph": "s", "id": 67949, "pid": 76337, "tid": -914061504, "ts": 1716454222307393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222385493, "dur": 2146,
    "args": {
      "External id": 67970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67970,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 67970, "pid": 5, "tid": 7, "ts": 1716454222385493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307478, "dur": 13,
    "args": {
      "External id": 67970,
      "cbid": 211, "correlation": 67970
    }
  },
  {
    "ph": "s", "id": 67970, "pid": 76337, "tid": -914061504, "ts": 1716454222307478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307575, "dur": 2,
    "args": {
      "External id": 67988,
      "cbid": 251, "correlation": 67988
    }
  },
  {
    "ph": "f", "id": 67988, "pid": 76337, "tid": -914061504, "ts": 1716454222307575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222387640, "dur": 160,
    "args": {
      "External id": 67990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67990,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 67990, "pid": 5, "tid": 7, "ts": 1716454222387640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307581, "dur": 13,
    "args": {
      "External id": 67990,
      "cbid": 211, "correlation": 67990
    }
  },
  {
    "ph": "s", "id": 67990, "pid": 76337, "tid": -914061504, "ts": 1716454222307581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222387802, "dur": 36,
    "args": {
      "External id": 67998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 67998,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 67998, "pid": 5, "tid": 7, "ts": 1716454222387802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307651, "dur": 15,
    "args": {
      "External id": 67998,
      "cbid": 211, "correlation": 67998
    }
  },
  {
    "ph": "s", "id": 67998, "pid": 76337, "tid": -914061504, "ts": 1716454222307651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222387840, "dur": 52,
    "args": {
      "External id": 68006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68006,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68006, "pid": 5, "tid": 7, "ts": 1716454222387840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307694, "dur": 9,
    "args": {
      "External id": 68006,
      "cbid": 211, "correlation": 68006
    }
  },
  {
    "ph": "s", "id": 68006, "pid": 76337, "tid": -914061504, "ts": 1716454222307694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222387893, "dur": 32,
    "args": {
      "External id": 68017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68017, "pid": 5, "tid": 7, "ts": 1716454222387893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307766, "dur": 12,
    "args": {
      "External id": 68017,
      "cbid": 211, "correlation": 68017
    }
  },
  {
    "ph": "s", "id": 68017, "pid": 76337, "tid": -914061504, "ts": 1716454222307766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222387926, "dur": 37,
    "args": {
      "External id": 68039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68039,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68039, "pid": 5, "tid": 7, "ts": 1716454222387926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307797, "dur": 8,
    "args": {
      "External id": 68039,
      "cbid": 211, "correlation": 68039
    }
  },
  {
    "ph": "s", "id": 68039, "pid": 76337, "tid": -914061504, "ts": 1716454222307797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307882, "dur": 1,
    "args": {
      "External id": 68050,
      "cbid": 251, "correlation": 68050
    }
  },
  {
    "ph": "f", "id": 68050, "pid": 76337, "tid": -914061504, "ts": 1716454222307882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222387964, "dur": 99,
    "args": {
      "External id": 68051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68051,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68051, "pid": 5, "tid": 7, "ts": 1716454222387964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307887, "dur": 13,
    "args": {
      "External id": 68051,
      "cbid": 211, "correlation": 68051
    }
  },
  {
    "ph": "s", "id": 68051, "pid": 76337, "tid": -914061504, "ts": 1716454222307887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307958, "dur": 1,
    "args": {
      "External id": 68062,
      "cbid": 251, "correlation": 68062
    }
  },
  {
    "ph": "f", "id": 68062, "pid": 76337, "tid": -914061504, "ts": 1716454222307958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307961, "dur": 0,
    "args": {
      "External id": 68063,
      "cbid": 251, "correlation": 68063
    }
  },
  {
    "ph": "f", "id": 68063, "pid": 76337, "tid": -914061504, "ts": 1716454222307961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222388064, "dur": 11,
    "args": {
      "External id": 68064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68064,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 68064, "pid": 5, "tid": 7, "ts": 1716454222388064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307963, "dur": 21,
    "args": {
      "External id": 68064,
      "cbid": 211, "correlation": 68064
    }
  },
  {
    "ph": "s", "id": 68064, "pid": 76337, "tid": -914061504, "ts": 1716454222307963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222388077, "dur": 5,
    "args": {
      "External id": 68066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68066,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 68066, "pid": 5, "tid": 7, "ts": 1716454222388077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222307986, "dur": 6,
    "args": {
      "External id": 68066,
      "cbid": 211, "correlation": 68066
    }
  },
  {
    "ph": "s", "id": 68066, "pid": 76337, "tid": -914061504, "ts": 1716454222307986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308048, "dur": 1,
    "args": {
      "External id": 68077,
      "cbid": 251, "correlation": 68077
    }
  },
  {
    "ph": "f", "id": 68077, "pid": 76337, "tid": -914061504, "ts": 1716454222308048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308052, "dur": 0,
    "args": {
      "External id": 68078,
      "cbid": 251, "correlation": 68078
    }
  },
  {
    "ph": "f", "id": 68078, "pid": 76337, "tid": -914061504, "ts": 1716454222308052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222388084, "dur": 8,
    "args": {
      "External id": 68079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68079,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 68079, "pid": 5, "tid": 7, "ts": 1716454222388084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308054, "dur": 12,
    "args": {
      "External id": 68079,
      "cbid": 211, "correlation": 68079
    }
  },
  {
    "ph": "s", "id": 68079, "pid": 76337, "tid": -914061504, "ts": 1716454222308054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222388093, "dur": 3,
    "args": {
      "External id": 68081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68081,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 68081, "pid": 5, "tid": 7, "ts": 1716454222388093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308068, "dur": 5,
    "args": {
      "External id": 68081,
      "cbid": 211, "correlation": 68081
    }
  },
  {
    "ph": "s", "id": 68081, "pid": 76337, "tid": -914061504, "ts": 1716454222308068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222388098, "dur": 100,
    "args": {
      "External id": 68102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68102,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 68102, "pid": 5, "tid": 7, "ts": 1716454222388098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308141, "dur": 12,
    "args": {
      "External id": 68102,
      "cbid": 211, "correlation": 68102
    }
  },
  {
    "ph": "s", "id": 68102, "pid": 76337, "tid": -914061504, "ts": 1716454222308141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308250, "dur": 1,
    "args": {
      "External id": 68120,
      "cbid": 251, "correlation": 68120
    }
  },
  {
    "ph": "f", "id": 68120, "pid": 76337, "tid": -914061504, "ts": 1716454222308250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222388199, "dur": 107,
    "args": {
      "External id": 68122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68122,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68122, "pid": 5, "tid": 7, "ts": 1716454222388199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308257, "dur": 15,
    "args": {
      "External id": 68122,
      "cbid": 211, "correlation": 68122
    }
  },
  {
    "ph": "s", "id": 68122, "pid": 76337, "tid": -914061504, "ts": 1716454222308257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222388307, "dur": 20,
    "args": {
      "External id": 68130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68130,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68130, "pid": 5, "tid": 7, "ts": 1716454222388307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308329, "dur": 12,
    "args": {
      "External id": 68130,
      "cbid": 211, "correlation": 68130
    }
  },
  {
    "ph": "s", "id": 68130, "pid": 76337, "tid": -914061504, "ts": 1716454222308329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222388328, "dur": 39,
    "args": {
      "External id": 68138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68138, "pid": 5, "tid": 7, "ts": 1716454222388328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308371, "dur": 9,
    "args": {
      "External id": 68138,
      "cbid": 211, "correlation": 68138
    }
  },
  {
    "ph": "s", "id": 68138, "pid": 76337, "tid": -914061504, "ts": 1716454222308371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222388368, "dur": 37,
    "args": {
      "External id": 68160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68160,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68160, "pid": 5, "tid": 7, "ts": 1716454222388368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308426, "dur": 10,
    "args": {
      "External id": 68160,
      "cbid": 211, "correlation": 68160
    }
  },
  {
    "ph": "s", "id": 68160, "pid": 76337, "tid": -914061504, "ts": 1716454222308426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308516, "dur": 1,
    "args": {
      "External id": 68176,
      "cbid": 251, "correlation": 68176
    }
  },
  {
    "ph": "f", "id": 68176, "pid": 76337, "tid": -914061504, "ts": 1716454222308516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308522, "dur": 0,
    "args": {
      "External id": 68178,
      "cbid": 251, "correlation": 68178
    }
  },
  {
    "ph": "f", "id": 68178, "pid": 76337, "tid": -914061504, "ts": 1716454222308522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222388407, "dur": 586,
    "args": {
      "External id": 68179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68179,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 68179, "pid": 5, "tid": 7, "ts": 1716454222388407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308527, "dur": 13,
    "args": {
      "External id": 68179,
      "cbid": 211, "correlation": 68179
    }
  },
  {
    "ph": "s", "id": 68179, "pid": 76337, "tid": -914061504, "ts": 1716454222308527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222388995, "dur": 133,
    "args": {
      "External id": 68187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68187,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68187, "pid": 5, "tid": 7, "ts": 1716454222388995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308593, "dur": 13,
    "args": {
      "External id": 68187,
      "cbid": 211, "correlation": 68187
    }
  },
  {
    "ph": "s", "id": 68187, "pid": 76337, "tid": -914061504, "ts": 1716454222308593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222389130, "dur": 134,
    "args": {
      "External id": 68195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68195,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68195, "pid": 5, "tid": 7, "ts": 1716454222389130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308624, "dur": 8,
    "args": {
      "External id": 68195,
      "cbid": 211, "correlation": 68195
    }
  },
  {
    "ph": "s", "id": 68195, "pid": 76337, "tid": -914061504, "ts": 1716454222308624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308702, "dur": 1,
    "args": {
      "External id": 68211,
      "cbid": 251, "correlation": 68211
    }
  },
  {
    "ph": "f", "id": 68211, "pid": 76337, "tid": -914061504, "ts": 1716454222308702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222389266, "dur": 332,
    "args": {
      "External id": 68213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68213,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68213, "pid": 5, "tid": 7, "ts": 1716454222389266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308708, "dur": 12,
    "args": {
      "External id": 68213,
      "cbid": 211, "correlation": 68213
    }
  },
  {
    "ph": "s", "id": 68213, "pid": 76337, "tid": -914061504, "ts": 1716454222308708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222389599, "dur": 27,
    "args": {
      "External id": 68221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68221,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68221, "pid": 5, "tid": 7, "ts": 1716454222389599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308750, "dur": 10,
    "args": {
      "External id": 68221,
      "cbid": 211, "correlation": 68221
    }
  },
  {
    "ph": "s", "id": 68221, "pid": 76337, "tid": -914061504, "ts": 1716454222308750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222389628, "dur": 90,
    "args": {
      "External id": 68232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68232,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68232, "pid": 5, "tid": 7, "ts": 1716454222389628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308820, "dur": 13,
    "args": {
      "External id": 68232,
      "cbid": 211, "correlation": 68232
    }
  },
  {
    "ph": "s", "id": 68232, "pid": 76337, "tid": -914061504, "ts": 1716454222308820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308885, "dur": 0,
    "args": {
      "External id": 68244,
      "cbid": 317, "correlation": 68244
    }
  },
  {
    "ph": "f", "id": 68244, "pid": 76337, "tid": -914061504, "ts": 1716454222308885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308885, "dur": 0,
    "args": {
      "External id": 68245,
      "cbid": 203, "correlation": 68245
    }
  },
  {
    "ph": "f", "id": 68245, "pid": 76337, "tid": -914061504, "ts": 1716454222308885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308886, "dur": 0,
    "args": {
      "External id": 68246,
      "cbid": 205, "correlation": 68246
    }
  },
  {
    "ph": "f", "id": 68246, "pid": 76337, "tid": -914061504, "ts": 1716454222308886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222389720, "dur": 24,
    "args": {
      "External id": 68250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68250,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68250, "pid": 5, "tid": 7, "ts": 1716454222389720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308902, "dur": 12,
    "args": {
      "External id": 68250,
      "cbid": 211, "correlation": 68250
    }
  },
  {
    "ph": "s", "id": 68250, "pid": 76337, "tid": -914061504, "ts": 1716454222308902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222389745, "dur": 133,
    "args": {
      "External id": 68252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68252,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68252, "pid": 5, "tid": 7, "ts": 1716454222389745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308920, "dur": 6,
    "args": {
      "External id": 68252,
      "cbid": 211, "correlation": 68252
    }
  },
  {
    "ph": "s", "id": 68252, "pid": 76337, "tid": -914061504, "ts": 1716454222308920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222389879, "dur": 23,
    "args": {
      "External id": 68254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68254,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68254, "pid": 5, "tid": 7, "ts": 1716454222389879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308930, "dur": 5,
    "args": {
      "External id": 68254,
      "cbid": 211, "correlation": 68254
    }
  },
  {
    "ph": "s", "id": 68254, "pid": 76337, "tid": -914061504, "ts": 1716454222308930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222389904, "dur": 35,
    "args": {
      "External id": 68260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68260,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68260, "pid": 5, "tid": 7, "ts": 1716454222389904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222308958, "dur": 8,
    "args": {
      "External id": 68260,
      "cbid": 211, "correlation": 68260
    }
  },
  {
    "ph": "s", "id": 68260, "pid": 76337, "tid": -914061504, "ts": 1716454222308958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222389940, "dur": 27,
    "args": {
      "External id": 68268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68268,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68268, "pid": 5, "tid": 7, "ts": 1716454222389940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309000, "dur": 9,
    "args": {
      "External id": 68268,
      "cbid": 211, "correlation": 68268
    }
  },
  {
    "ph": "s", "id": 68268, "pid": 76337, "tid": -914061504, "ts": 1716454222309000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222389969, "dur": 57,
    "args": {
      "External id": 68277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68277,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68277, "pid": 5, "tid": 7, "ts": 1716454222389969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309038, "dur": 11,
    "args": {
      "External id": 68277,
      "cbid": 211, "correlation": 68277
    }
  },
  {
    "ph": "s", "id": 68277, "pid": 76337, "tid": -914061504, "ts": 1716454222309038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222390028, "dur": 58,
    "args": {
      "External id": 68297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68297,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 68297, "pid": 5, "tid": 7, "ts": 1716454222390028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309114, "dur": 12,
    "args": {
      "External id": 68297,
      "cbid": 211, "correlation": 68297
    }
  },
  {
    "ph": "s", "id": 68297, "pid": 76337, "tid": -914061504, "ts": 1716454222309114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222390087, "dur": 5,
    "args": {
      "External id": 68309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68309,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 68309, "pid": 5, "tid": 7, "ts": 1716454222390087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309136, "dur": 6,
    "args": {
      "External id": 68309,
      "cbid": 211, "correlation": 68309
    }
  },
  {
    "ph": "s", "id": 68309, "pid": 76337, "tid": -914061504, "ts": 1716454222309136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222390093, "dur": 60,
    "args": {
      "External id": 68312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68312, "pid": 5, "tid": 7, "ts": 1716454222390093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309155, "dur": 7,
    "args": {
      "External id": 68312,
      "cbid": 211, "correlation": 68312
    }
  },
  {
    "ph": "s", "id": 68312, "pid": 76337, "tid": -914061504, "ts": 1716454222309155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222390154, "dur": 38,
    "args": {
      "External id": 68321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68321,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68321, "pid": 5, "tid": 7, "ts": 1716454222390154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309194, "dur": 10,
    "args": {
      "External id": 68321,
      "cbid": 211, "correlation": 68321
    }
  },
  {
    "ph": "s", "id": 68321, "pid": 76337, "tid": -914061504, "ts": 1716454222309194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309245, "dur": 0,
    "args": {
      "External id": 68331,
      "cbid": 317, "correlation": 68331
    }
  },
  {
    "ph": "f", "id": 68331, "pid": 76337, "tid": -914061504, "ts": 1716454222309245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309246, "dur": 0,
    "args": {
      "External id": 68332,
      "cbid": 203, "correlation": 68332
    }
  },
  {
    "ph": "f", "id": 68332, "pid": 76337, "tid": -914061504, "ts": 1716454222309246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309247, "dur": 0,
    "args": {
      "External id": 68333,
      "cbid": 205, "correlation": 68333
    }
  },
  {
    "ph": "f", "id": 68333, "pid": 76337, "tid": -914061504, "ts": 1716454222309247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222390194, "dur": 40,
    "args": {
      "External id": 68337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68337, "pid": 5, "tid": 7, "ts": 1716454222390194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309262, "dur": 11,
    "args": {
      "External id": 68337,
      "cbid": 211, "correlation": 68337
    }
  },
  {
    "ph": "s", "id": 68337, "pid": 76337, "tid": -914061504, "ts": 1716454222309262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222390236, "dur": 91,
    "args": {
      "External id": 68339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68339,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68339, "pid": 5, "tid": 7, "ts": 1716454222390236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309276, "dur": 5,
    "args": {
      "External id": 68339,
      "cbid": 211, "correlation": 68339
    }
  },
  {
    "ph": "s", "id": 68339, "pid": 76337, "tid": -914061504, "ts": 1716454222309276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222390329, "dur": 1404,
    "args": {
      "External id": 68341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68341,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68341, "pid": 5, "tid": 7, "ts": 1716454222390329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309287, "dur": 6,
    "args": {
      "External id": 68341,
      "cbid": 211, "correlation": 68341
    }
  },
  {
    "ph": "s", "id": 68341, "pid": 76337, "tid": -914061504, "ts": 1716454222309287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222391734, "dur": 22,
    "args": {
      "External id": 68343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68343, "pid": 5, "tid": 7, "ts": 1716454222391734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309297, "dur": 6,
    "args": {
      "External id": 68343,
      "cbid": 211, "correlation": 68343
    }
  },
  {
    "ph": "s", "id": 68343, "pid": 76337, "tid": -914061504, "ts": 1716454222309297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222391758, "dur": 35,
    "args": {
      "External id": 68349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68349, "pid": 5, "tid": 7, "ts": 1716454222391758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309325, "dur": 8,
    "args": {
      "External id": 68349,
      "cbid": 211, "correlation": 68349
    }
  },
  {
    "ph": "s", "id": 68349, "pid": 76337, "tid": -914061504, "ts": 1716454222309325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222391795, "dur": 4,
    "args": {
      "External id": 68357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68357,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 68357, "pid": 5, "tid": 7, "ts": 1716454222391795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309368, "dur": 12,
    "args": {
      "External id": 68357,
      "cbid": 211, "correlation": 68357
    }
  },
  {
    "ph": "s", "id": 68357, "pid": 76337, "tid": -914061504, "ts": 1716454222309368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309436, "dur": 1,
    "args": {
      "External id": 68373,
      "cbid": 251, "correlation": 68373
    }
  },
  {
    "ph": "f", "id": 68373, "pid": 76337, "tid": -914061504, "ts": 1716454222309436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309441, "dur": 0,
    "args": {
      "External id": 68375,
      "cbid": 251, "correlation": 68375
    }
  },
  {
    "ph": "f", "id": 68375, "pid": 76337, "tid": -914061504, "ts": 1716454222309441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222391800, "dur": 13,
    "args": {
      "External id": 68376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68376,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 68376, "pid": 5, "tid": 7, "ts": 1716454222391800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309443, "dur": 11,
    "args": {
      "External id": 68376,
      "cbid": 211, "correlation": 68376
    }
  },
  {
    "ph": "s", "id": 68376, "pid": 76337, "tid": -914061504, "ts": 1716454222309443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222391815, "dur": 5,
    "args": {
      "External id": 68378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68378,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 68378, "pid": 5, "tid": 7, "ts": 1716454222391815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309456, "dur": 6,
    "args": {
      "External id": 68378,
      "cbid": 211, "correlation": 68378
    }
  },
  {
    "ph": "s", "id": 68378, "pid": 76337, "tid": -914061504, "ts": 1716454222309456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222391822, "dur": 31,
    "args": {
      "External id": 68388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68388,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68388, "pid": 5, "tid": 7, "ts": 1716454222391822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309514, "dur": 13,
    "args": {
      "External id": 68388,
      "cbid": 211, "correlation": 68388
    }
  },
  {
    "ph": "s", "id": 68388, "pid": 76337, "tid": -914061504, "ts": 1716454222309514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222391854, "dur": 33,
    "args": {
      "External id": 68408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68408,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 68408, "pid": 5, "tid": 7, "ts": 1716454222391854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309581, "dur": 11,
    "args": {
      "External id": 68408,
      "cbid": 211, "correlation": 68408
    }
  },
  {
    "ph": "s", "id": 68408, "pid": 76337, "tid": -914061504, "ts": 1716454222309581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222391889, "dur": 4,
    "args": {
      "External id": 68420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68420,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 68420, "pid": 5, "tid": 7, "ts": 1716454222391889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309601, "dur": 6,
    "args": {
      "External id": 68420,
      "cbid": 211, "correlation": 68420
    }
  },
  {
    "ph": "s", "id": 68420, "pid": 76337, "tid": -914061504, "ts": 1716454222309601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222391894, "dur": 31,
    "args": {
      "External id": 68423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68423, "pid": 5, "tid": 7, "ts": 1716454222391894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309618, "dur": 6,
    "args": {
      "External id": 68423,
      "cbid": 211, "correlation": 68423
    }
  },
  {
    "ph": "s", "id": 68423, "pid": 76337, "tid": -914061504, "ts": 1716454222309618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222391927, "dur": 21,
    "args": {
      "External id": 68432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68432,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68432, "pid": 5, "tid": 7, "ts": 1716454222391927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309661, "dur": 10,
    "args": {
      "External id": 68432,
      "cbid": 211, "correlation": 68432
    }
  },
  {
    "ph": "s", "id": 68432, "pid": 76337, "tid": -914061504, "ts": 1716454222309661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309724, "dur": 0,
    "args": {
      "External id": 68442,
      "cbid": 317, "correlation": 68442
    }
  },
  {
    "ph": "f", "id": 68442, "pid": 76337, "tid": -914061504, "ts": 1716454222309724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309725, "dur": 0,
    "args": {
      "External id": 68443,
      "cbid": 203, "correlation": 68443
    }
  },
  {
    "ph": "f", "id": 68443, "pid": 76337, "tid": -914061504, "ts": 1716454222309725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309726, "dur": 0,
    "args": {
      "External id": 68444,
      "cbid": 205, "correlation": 68444
    }
  },
  {
    "ph": "f", "id": 68444, "pid": 76337, "tid": -914061504, "ts": 1716454222309726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222391949, "dur": 22,
    "args": {
      "External id": 68448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68448,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68448, "pid": 5, "tid": 7, "ts": 1716454222391949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309740, "dur": 12,
    "args": {
      "External id": 68448,
      "cbid": 211, "correlation": 68448
    }
  },
  {
    "ph": "s", "id": 68448, "pid": 76337, "tid": -914061504, "ts": 1716454222309740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222391973, "dur": 48,
    "args": {
      "External id": 68450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68450,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68450, "pid": 5, "tid": 7, "ts": 1716454222391973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309755, "dur": 5,
    "args": {
      "External id": 68450,
      "cbid": 211, "correlation": 68450
    }
  },
  {
    "ph": "s", "id": 68450, "pid": 76337, "tid": -914061504, "ts": 1716454222309755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222392023, "dur": 710,
    "args": {
      "External id": 68452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68452,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68452, "pid": 5, "tid": 7, "ts": 1716454222392023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309766, "dur": 6,
    "args": {
      "External id": 68452,
      "cbid": 211, "correlation": 68452
    }
  },
  {
    "ph": "s", "id": 68452, "pid": 76337, "tid": -914061504, "ts": 1716454222309766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222392734, "dur": 23,
    "args": {
      "External id": 68454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68454, "pid": 5, "tid": 7, "ts": 1716454222392734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309775, "dur": 5,
    "args": {
      "External id": 68454,
      "cbid": 211, "correlation": 68454
    }
  },
  {
    "ph": "s", "id": 68454, "pid": 76337, "tid": -914061504, "ts": 1716454222309775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222392758, "dur": 34,
    "args": {
      "External id": 68460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68460, "pid": 5, "tid": 7, "ts": 1716454222392758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309804, "dur": 8,
    "args": {
      "External id": 68460,
      "cbid": 211, "correlation": 68460
    }
  },
  {
    "ph": "s", "id": 68460, "pid": 76337, "tid": -914061504, "ts": 1716454222309804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309863, "dur": 0,
    "args": {
      "External id": 68470,
      "cbid": 317, "correlation": 68470
    }
  },
  {
    "ph": "f", "id": 68470, "pid": 76337, "tid": -914061504, "ts": 1716454222309863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309863, "dur": 0,
    "args": {
      "External id": 68471,
      "cbid": 203, "correlation": 68471
    }
  },
  {
    "ph": "f", "id": 68471, "pid": 76337, "tid": -914061504, "ts": 1716454222309863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309864, "dur": 0,
    "args": {
      "External id": 68472,
      "cbid": 205, "correlation": 68472
    }
  },
  {
    "ph": "f", "id": 68472, "pid": 76337, "tid": -914061504, "ts": 1716454222309864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222392794, "dur": 40,
    "args": {
      "External id": 68476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68476,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68476, "pid": 5, "tid": 7, "ts": 1716454222392794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309877, "dur": 11,
    "args": {
      "External id": 68476,
      "cbid": 211, "correlation": 68476
    }
  },
  {
    "ph": "s", "id": 68476, "pid": 76337, "tid": -914061504, "ts": 1716454222309877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222392835, "dur": 209,
    "args": {
      "External id": 68478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68478,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68478, "pid": 5, "tid": 7, "ts": 1716454222392835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309894, "dur": 6,
    "args": {
      "External id": 68478,
      "cbid": 211, "correlation": 68478
    }
  },
  {
    "ph": "s", "id": 68478, "pid": 76337, "tid": -914061504, "ts": 1716454222309894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222393045, "dur": 23,
    "args": {
      "External id": 68480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68480,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68480, "pid": 5, "tid": 7, "ts": 1716454222393045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309903, "dur": 6,
    "args": {
      "External id": 68480,
      "cbid": 211, "correlation": 68480
    }
  },
  {
    "ph": "s", "id": 68480, "pid": 76337, "tid": -914061504, "ts": 1716454222309903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222393069, "dur": 34,
    "args": {
      "External id": 68486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68486, "pid": 5, "tid": 7, "ts": 1716454222393069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309929, "dur": 8,
    "args": {
      "External id": 68486,
      "cbid": 211, "correlation": 68486
    }
  },
  {
    "ph": "s", "id": 68486, "pid": 76337, "tid": -914061504, "ts": 1716454222309929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222393105, "dur": 27,
    "args": {
      "External id": 68494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68494,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68494, "pid": 5, "tid": 7, "ts": 1716454222393105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309962, "dur": 8,
    "args": {
      "External id": 68494,
      "cbid": 211, "correlation": 68494
    }
  },
  {
    "ph": "s", "id": 68494, "pid": 76337, "tid": -914061504, "ts": 1716454222309962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222393133, "dur": 20,
    "args": {
      "External id": 68502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68502,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68502, "pid": 5, "tid": 7, "ts": 1716454222393133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222309999, "dur": 9,
    "args": {
      "External id": 68502,
      "cbid": 211, "correlation": 68502
    }
  },
  {
    "ph": "s", "id": 68502, "pid": 76337, "tid": -914061504, "ts": 1716454222309999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222393155, "dur": 32,
    "args": {
      "External id": 68522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68522,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 68522, "pid": 5, "tid": 7, "ts": 1716454222393155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310082, "dur": 12,
    "args": {
      "External id": 68522,
      "cbid": 211, "correlation": 68522
    }
  },
  {
    "ph": "s", "id": 68522, "pid": 76337, "tid": -914061504, "ts": 1716454222310082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222393189, "dur": 4,
    "args": {
      "External id": 68534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68534,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 68534, "pid": 5, "tid": 7, "ts": 1716454222393189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310104, "dur": 6,
    "args": {
      "External id": 68534,
      "cbid": 211, "correlation": 68534
    }
  },
  {
    "ph": "s", "id": 68534, "pid": 76337, "tid": -914061504, "ts": 1716454222310104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222393195, "dur": 32,
    "args": {
      "External id": 68537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68537,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68537, "pid": 5, "tid": 7, "ts": 1716454222393195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310123, "dur": 6,
    "args": {
      "External id": 68537,
      "cbid": 211, "correlation": 68537
    }
  },
  {
    "ph": "s", "id": 68537, "pid": 76337, "tid": -914061504, "ts": 1716454222310123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310180, "dur": 0,
    "args": {
      "External id": 68548,
      "cbid": 317, "correlation": 68548
    }
  },
  {
    "ph": "f", "id": 68548, "pid": 76337, "tid": -914061504, "ts": 1716454222310180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310181, "dur": 0,
    "args": {
      "External id": 68549,
      "cbid": 203, "correlation": 68549
    }
  },
  {
    "ph": "f", "id": 68549, "pid": 76337, "tid": -914061504, "ts": 1716454222310181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310182, "dur": 0,
    "args": {
      "External id": 68550,
      "cbid": 205, "correlation": 68550
    }
  },
  {
    "ph": "f", "id": 68550, "pid": 76337, "tid": -914061504, "ts": 1716454222310182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222393228, "dur": 22,
    "args": {
      "External id": 68554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68554,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68554, "pid": 5, "tid": 7, "ts": 1716454222393228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310194, "dur": 11,
    "args": {
      "External id": 68554,
      "cbid": 211, "correlation": 68554
    }
  },
  {
    "ph": "s", "id": 68554, "pid": 76337, "tid": -914061504, "ts": 1716454222310194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222393251, "dur": 116,
    "args": {
      "External id": 68556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68556,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68556, "pid": 5, "tid": 7, "ts": 1716454222393251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310211, "dur": 6,
    "args": {
      "External id": 68556,
      "cbid": 211, "correlation": 68556
    }
  },
  {
    "ph": "s", "id": 68556, "pid": 76337, "tid": -914061504, "ts": 1716454222310211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222393369, "dur": 23,
    "args": {
      "External id": 68558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68558, "pid": 5, "tid": 7, "ts": 1716454222393369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310221, "dur": 8,
    "args": {
      "External id": 68558,
      "cbid": 211, "correlation": 68558
    }
  },
  {
    "ph": "s", "id": 68558, "pid": 76337, "tid": -914061504, "ts": 1716454222310221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222393394, "dur": 34,
    "args": {
      "External id": 68564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68564, "pid": 5, "tid": 7, "ts": 1716454222393394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310252, "dur": 9,
    "args": {
      "External id": 68564,
      "cbid": 211, "correlation": 68564
    }
  },
  {
    "ph": "s", "id": 68564, "pid": 76337, "tid": -914061504, "ts": 1716454222310252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222393429, "dur": 186,
    "args": {
      "External id": 68573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68573,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68573, "pid": 5, "tid": 7, "ts": 1716454222393429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310335, "dur": 14,
    "args": {
      "External id": 68573,
      "cbid": 211, "correlation": 68573
    }
  },
  {
    "ph": "s", "id": 68573, "pid": 76337, "tid": -914061504, "ts": 1716454222310335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222393617, "dur": 71,
    "args": {
      "External id": 68595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68595,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68595, "pid": 5, "tid": 7, "ts": 1716454222393617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310392, "dur": 10,
    "args": {
      "External id": 68595,
      "cbid": 211, "correlation": 68595
    }
  },
  {
    "ph": "s", "id": 68595, "pid": 76337, "tid": -914061504, "ts": 1716454222310392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310481, "dur": 1,
    "args": {
      "External id": 68606,
      "cbid": 251, "correlation": 68606
    }
  },
  {
    "ph": "f", "id": 68606, "pid": 76337, "tid": -914061504, "ts": 1716454222310481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222393689, "dur": 165,
    "args": {
      "External id": 68607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68607,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68607, "pid": 5, "tid": 7, "ts": 1716454222393689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310487, "dur": 13,
    "args": {
      "External id": 68607,
      "cbid": 211, "correlation": 68607
    }
  },
  {
    "ph": "s", "id": 68607, "pid": 76337, "tid": -914061504, "ts": 1716454222310487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310557, "dur": 1,
    "args": {
      "External id": 68618,
      "cbid": 251, "correlation": 68618
    }
  },
  {
    "ph": "f", "id": 68618, "pid": 76337, "tid": -914061504, "ts": 1716454222310557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222393855, "dur": 158,
    "args": {
      "External id": 68619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68619,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68619, "pid": 5, "tid": 7, "ts": 1716454222393855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310561, "dur": 11,
    "args": {
      "External id": 68619,
      "cbid": 211, "correlation": 68619
    }
  },
  {
    "ph": "s", "id": 68619, "pid": 76337, "tid": -914061504, "ts": 1716454222310561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310626, "dur": 1,
    "args": {
      "External id": 68630,
      "cbid": 251, "correlation": 68630
    }
  },
  {
    "ph": "f", "id": 68630, "pid": 76337, "tid": -914061504, "ts": 1716454222310626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222394015, "dur": 156,
    "args": {
      "External id": 68631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68631,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68631, "pid": 5, "tid": 7, "ts": 1716454222394015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310630, "dur": 14,
    "args": {
      "External id": 68631,
      "cbid": 211, "correlation": 68631
    }
  },
  {
    "ph": "s", "id": 68631, "pid": 76337, "tid": -914061504, "ts": 1716454222310630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222394172, "dur": 2143,
    "args": {
      "External id": 68652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68652,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 68652, "pid": 5, "tid": 7, "ts": 1716454222394172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310713, "dur": 12,
    "args": {
      "External id": 68652,
      "cbid": 211, "correlation": 68652
    }
  },
  {
    "ph": "s", "id": 68652, "pid": 76337, "tid": -914061504, "ts": 1716454222310713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310810, "dur": 1,
    "args": {
      "External id": 68670,
      "cbid": 251, "correlation": 68670
    }
  },
  {
    "ph": "f", "id": 68670, "pid": 76337, "tid": -914061504, "ts": 1716454222310810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222396317, "dur": 160,
    "args": {
      "External id": 68672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68672,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 68672, "pid": 5, "tid": 7, "ts": 1716454222396317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310815, "dur": 14,
    "args": {
      "External id": 68672,
      "cbid": 211, "correlation": 68672
    }
  },
  {
    "ph": "s", "id": 68672, "pid": 76337, "tid": -914061504, "ts": 1716454222310815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222396479, "dur": 35,
    "args": {
      "External id": 68680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68680,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68680, "pid": 5, "tid": 7, "ts": 1716454222396479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310885, "dur": 12,
    "args": {
      "External id": 68680,
      "cbid": 211, "correlation": 68680
    }
  },
  {
    "ph": "s", "id": 68680, "pid": 76337, "tid": -914061504, "ts": 1716454222310885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222396515, "dur": 51,
    "args": {
      "External id": 68688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68688,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68688, "pid": 5, "tid": 7, "ts": 1716454222396515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222310923, "dur": 9,
    "args": {
      "External id": 68688,
      "cbid": 211, "correlation": 68688
    }
  },
  {
    "ph": "s", "id": 68688, "pid": 76337, "tid": -914061504, "ts": 1716454222310923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222396568, "dur": 31,
    "args": {
      "External id": 68699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68699, "pid": 5, "tid": 7, "ts": 1716454222396568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311003, "dur": 13,
    "args": {
      "External id": 68699,
      "cbid": 211, "correlation": 68699
    }
  },
  {
    "ph": "s", "id": 68699, "pid": 76337, "tid": -914061504, "ts": 1716454222311003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222396600, "dur": 38,
    "args": {
      "External id": 68721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68721,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68721, "pid": 5, "tid": 7, "ts": 1716454222396600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311037, "dur": 9,
    "args": {
      "External id": 68721,
      "cbid": 211, "correlation": 68721
    }
  },
  {
    "ph": "s", "id": 68721, "pid": 76337, "tid": -914061504, "ts": 1716454222311037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311123, "dur": 1,
    "args": {
      "External id": 68732,
      "cbid": 251, "correlation": 68732
    }
  },
  {
    "ph": "f", "id": 68732, "pid": 76337, "tid": -914061504, "ts": 1716454222311123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222396640, "dur": 99,
    "args": {
      "External id": 68733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68733,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68733, "pid": 5, "tid": 7, "ts": 1716454222396640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311128, "dur": 13,
    "args": {
      "External id": 68733,
      "cbid": 211, "correlation": 68733
    }
  },
  {
    "ph": "s", "id": 68733, "pid": 76337, "tid": -914061504, "ts": 1716454222311128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311195, "dur": 1,
    "args": {
      "External id": 68744,
      "cbid": 251, "correlation": 68744
    }
  },
  {
    "ph": "f", "id": 68744, "pid": 76337, "tid": -914061504, "ts": 1716454222311195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311199, "dur": 0,
    "args": {
      "External id": 68745,
      "cbid": 251, "correlation": 68745
    }
  },
  {
    "ph": "f", "id": 68745, "pid": 76337, "tid": -914061504, "ts": 1716454222311199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222396740, "dur": 12,
    "args": {
      "External id": 68746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68746,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 68746, "pid": 5, "tid": 7, "ts": 1716454222396740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311201, "dur": 13,
    "args": {
      "External id": 68746,
      "cbid": 211, "correlation": 68746
    }
  },
  {
    "ph": "s", "id": 68746, "pid": 76337, "tid": -914061504, "ts": 1716454222311201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222396753, "dur": 5,
    "args": {
      "External id": 68748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68748,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 68748, "pid": 5, "tid": 7, "ts": 1716454222396753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311215, "dur": 6,
    "args": {
      "External id": 68748,
      "cbid": 211, "correlation": 68748
    }
  },
  {
    "ph": "s", "id": 68748, "pid": 76337, "tid": -914061504, "ts": 1716454222311215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311273, "dur": 1,
    "args": {
      "External id": 68759,
      "cbid": 251, "correlation": 68759
    }
  },
  {
    "ph": "f", "id": 68759, "pid": 76337, "tid": -914061504, "ts": 1716454222311273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311276, "dur": 0,
    "args": {
      "External id": 68760,
      "cbid": 251, "correlation": 68760
    }
  },
  {
    "ph": "f", "id": 68760, "pid": 76337, "tid": -914061504, "ts": 1716454222311276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222396760, "dur": 8,
    "args": {
      "External id": 68761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68761,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 68761, "pid": 5, "tid": 7, "ts": 1716454222396760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311277, "dur": 11,
    "args": {
      "External id": 68761,
      "cbid": 211, "correlation": 68761
    }
  },
  {
    "ph": "s", "id": 68761, "pid": 76337, "tid": -914061504, "ts": 1716454222311277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222396769, "dur": 3,
    "args": {
      "External id": 68763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68763,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 68763, "pid": 5, "tid": 7, "ts": 1716454222396769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311290, "dur": 6,
    "args": {
      "External id": 68763,
      "cbid": 211, "correlation": 68763
    }
  },
  {
    "ph": "s", "id": 68763, "pid": 76337, "tid": -914061504, "ts": 1716454222311290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222396774, "dur": 99,
    "args": {
      "External id": 68784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68784,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 68784, "pid": 5, "tid": 7, "ts": 1716454222396774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311363, "dur": 27,
    "args": {
      "External id": 68784,
      "cbid": 211, "correlation": 68784
    }
  },
  {
    "ph": "s", "id": 68784, "pid": 76337, "tid": -914061504, "ts": 1716454222311363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311476, "dur": 1,
    "args": {
      "External id": 68802,
      "cbid": 251, "correlation": 68802
    }
  },
  {
    "ph": "f", "id": 68802, "pid": 76337, "tid": -914061504, "ts": 1716454222311476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222396875, "dur": 103,
    "args": {
      "External id": 68804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68804,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68804, "pid": 5, "tid": 7, "ts": 1716454222396875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311482, "dur": 14,
    "args": {
      "External id": 68804,
      "cbid": 211, "correlation": 68804
    }
  },
  {
    "ph": "s", "id": 68804, "pid": 76337, "tid": -914061504, "ts": 1716454222311482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222396980, "dur": 20,
    "args": {
      "External id": 68812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68812,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68812, "pid": 5, "tid": 7, "ts": 1716454222396980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311552, "dur": 12,
    "args": {
      "External id": 68812,
      "cbid": 211, "correlation": 68812
    }
  },
  {
    "ph": "s", "id": 68812, "pid": 76337, "tid": -914061504, "ts": 1716454222311552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222397001, "dur": 39,
    "args": {
      "External id": 68820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68820, "pid": 5, "tid": 7, "ts": 1716454222397001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311593, "dur": 10,
    "args": {
      "External id": 68820,
      "cbid": 211, "correlation": 68820
    }
  },
  {
    "ph": "s", "id": 68820, "pid": 76337, "tid": -914061504, "ts": 1716454222311593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222397042, "dur": 38,
    "args": {
      "External id": 68842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68842,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68842, "pid": 5, "tid": 7, "ts": 1716454222397042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311644, "dur": 11,
    "args": {
      "External id": 68842,
      "cbid": 211, "correlation": 68842
    }
  },
  {
    "ph": "s", "id": 68842, "pid": 76337, "tid": -914061504, "ts": 1716454222311644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311734, "dur": 1,
    "args": {
      "External id": 68858,
      "cbid": 251, "correlation": 68858
    }
  },
  {
    "ph": "f", "id": 68858, "pid": 76337, "tid": -914061504, "ts": 1716454222311734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311739, "dur": 0,
    "args": {
      "External id": 68860,
      "cbid": 251, "correlation": 68860
    }
  },
  {
    "ph": "f", "id": 68860, "pid": 76337, "tid": -914061504, "ts": 1716454222311739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222397081, "dur": 588,
    "args": {
      "External id": 68861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68861,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 68861, "pid": 5, "tid": 7, "ts": 1716454222397081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311743, "dur": 13,
    "args": {
      "External id": 68861,
      "cbid": 211, "correlation": 68861
    }
  },
  {
    "ph": "s", "id": 68861, "pid": 76337, "tid": -914061504, "ts": 1716454222311743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222397671, "dur": 134,
    "args": {
      "External id": 68869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68869, "pid": 5, "tid": 7, "ts": 1716454222397671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311811, "dur": 13,
    "args": {
      "External id": 68869,
      "cbid": 211, "correlation": 68869
    }
  },
  {
    "ph": "s", "id": 68869, "pid": 76337, "tid": -914061504, "ts": 1716454222311811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222397806, "dur": 133,
    "args": {
      "External id": 68877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68877,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68877, "pid": 5, "tid": 7, "ts": 1716454222397806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311842, "dur": 8,
    "args": {
      "External id": 68877,
      "cbid": 211, "correlation": 68877
    }
  },
  {
    "ph": "s", "id": 68877, "pid": 76337, "tid": -914061504, "ts": 1716454222311842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311919, "dur": 1,
    "args": {
      "External id": 68893,
      "cbid": 251, "correlation": 68893
    }
  },
  {
    "ph": "f", "id": 68893, "pid": 76337, "tid": -914061504, "ts": 1716454222311919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222397941, "dur": 331,
    "args": {
      "External id": 68895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68895,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68895, "pid": 5, "tid": 7, "ts": 1716454222397941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311925, "dur": 13,
    "args": {
      "External id": 68895,
      "cbid": 211, "correlation": 68895
    }
  },
  {
    "ph": "s", "id": 68895, "pid": 76337, "tid": -914061504, "ts": 1716454222311925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222398274, "dur": 28,
    "args": {
      "External id": 68903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68903,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68903, "pid": 5, "tid": 7, "ts": 1716454222398274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222311967, "dur": 19,
    "args": {
      "External id": 68903,
      "cbid": 211, "correlation": 68903
    }
  },
  {
    "ph": "s", "id": 68903, "pid": 76337, "tid": -914061504, "ts": 1716454222311967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222398303, "dur": 90,
    "args": {
      "External id": 68914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68914, "pid": 5, "tid": 7, "ts": 1716454222398303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312044, "dur": 12,
    "args": {
      "External id": 68914,
      "cbid": 211, "correlation": 68914
    }
  },
  {
    "ph": "s", "id": 68914, "pid": 76337, "tid": -914061504, "ts": 1716454222312044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312109, "dur": 0,
    "args": {
      "External id": 68926,
      "cbid": 317, "correlation": 68926
    }
  },
  {
    "ph": "f", "id": 68926, "pid": 76337, "tid": -914061504, "ts": 1716454222312109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312110, "dur": 0,
    "args": {
      "External id": 68927,
      "cbid": 203, "correlation": 68927
    }
  },
  {
    "ph": "f", "id": 68927, "pid": 76337, "tid": -914061504, "ts": 1716454222312110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312111, "dur": 0,
    "args": {
      "External id": 68928,
      "cbid": 205, "correlation": 68928
    }
  },
  {
    "ph": "f", "id": 68928, "pid": 76337, "tid": -914061504, "ts": 1716454222312111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222398394, "dur": 26,
    "args": {
      "External id": 68932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68932,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68932, "pid": 5, "tid": 7, "ts": 1716454222398394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312126, "dur": 13,
    "args": {
      "External id": 68932,
      "cbid": 211, "correlation": 68932
    }
  },
  {
    "ph": "s", "id": 68932, "pid": 76337, "tid": -914061504, "ts": 1716454222312126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222398422, "dur": 132,
    "args": {
      "External id": 68934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68934,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 68934, "pid": 5, "tid": 7, "ts": 1716454222398422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312145, "dur": 6,
    "args": {
      "External id": 68934,
      "cbid": 211, "correlation": 68934
    }
  },
  {
    "ph": "s", "id": 68934, "pid": 76337, "tid": -914061504, "ts": 1716454222312145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222398556, "dur": 24,
    "args": {
      "External id": 68936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68936,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68936, "pid": 5, "tid": 7, "ts": 1716454222398556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312158, "dur": 6,
    "args": {
      "External id": 68936,
      "cbid": 211, "correlation": 68936
    }
  },
  {
    "ph": "s", "id": 68936, "pid": 76337, "tid": -914061504, "ts": 1716454222312158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222398581, "dur": 36,
    "args": {
      "External id": 68942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68942,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68942, "pid": 5, "tid": 7, "ts": 1716454222398581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312186, "dur": 10,
    "args": {
      "External id": 68942,
      "cbid": 211, "correlation": 68942
    }
  },
  {
    "ph": "s", "id": 68942, "pid": 76337, "tid": -914061504, "ts": 1716454222312186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222398618, "dur": 27,
    "args": {
      "External id": 68950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68950,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68950, "pid": 5, "tid": 7, "ts": 1716454222398618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312219, "dur": 8,
    "args": {
      "External id": 68950,
      "cbid": 211, "correlation": 68950
    }
  },
  {
    "ph": "s", "id": 68950, "pid": 76337, "tid": -914061504, "ts": 1716454222312219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222398646, "dur": 45,
    "args": {
      "External id": 68959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68959,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68959, "pid": 5, "tid": 7, "ts": 1716454222398646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312258, "dur": 11,
    "args": {
      "External id": 68959,
      "cbid": 211, "correlation": 68959
    }
  },
  {
    "ph": "s", "id": 68959, "pid": 76337, "tid": -914061504, "ts": 1716454222312258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222398692, "dur": 44,
    "args": {
      "External id": 68979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68979,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 68979, "pid": 5, "tid": 7, "ts": 1716454222398692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312330, "dur": 11,
    "args": {
      "External id": 68979,
      "cbid": 211, "correlation": 68979
    }
  },
  {
    "ph": "s", "id": 68979, "pid": 76337, "tid": -914061504, "ts": 1716454222312330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222398738, "dur": 5,
    "args": {
      "External id": 68991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68991,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 68991, "pid": 5, "tid": 7, "ts": 1716454222398738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312352, "dur": 7,
    "args": {
      "External id": 68991,
      "cbid": 211, "correlation": 68991
    }
  },
  {
    "ph": "s", "id": 68991, "pid": 76337, "tid": -914061504, "ts": 1716454222312352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222398745, "dur": 47,
    "args": {
      "External id": 68994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 68994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 68994, "pid": 5, "tid": 7, "ts": 1716454222398745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312370, "dur": 6,
    "args": {
      "External id": 68994,
      "cbid": 211, "correlation": 68994
    }
  },
  {
    "ph": "s", "id": 68994, "pid": 76337, "tid": -914061504, "ts": 1716454222312370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222398793, "dur": 30,
    "args": {
      "External id": 69003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69003,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69003, "pid": 5, "tid": 7, "ts": 1716454222398793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312409, "dur": 13,
    "args": {
      "External id": 69003,
      "cbid": 211, "correlation": 69003
    }
  },
  {
    "ph": "s", "id": 69003, "pid": 76337, "tid": -914061504, "ts": 1716454222312409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312464, "dur": 0,
    "args": {
      "External id": 69013,
      "cbid": 317, "correlation": 69013
    }
  },
  {
    "ph": "f", "id": 69013, "pid": 76337, "tid": -914061504, "ts": 1716454222312464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312465, "dur": 0,
    "args": {
      "External id": 69014,
      "cbid": 203, "correlation": 69014
    }
  },
  {
    "ph": "f", "id": 69014, "pid": 76337, "tid": -914061504, "ts": 1716454222312465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312466, "dur": 0,
    "args": {
      "External id": 69015,
      "cbid": 205, "correlation": 69015
    }
  },
  {
    "ph": "f", "id": 69015, "pid": 76337, "tid": -914061504, "ts": 1716454222312466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222398824, "dur": 30,
    "args": {
      "External id": 69019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69019,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69019, "pid": 5, "tid": 7, "ts": 1716454222398824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312482, "dur": 12,
    "args": {
      "External id": 69019,
      "cbid": 211, "correlation": 69019
    }
  },
  {
    "ph": "s", "id": 69019, "pid": 76337, "tid": -914061504, "ts": 1716454222312482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222398856, "dur": 69,
    "args": {
      "External id": 69021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69021, "pid": 5, "tid": 7, "ts": 1716454222398856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312496, "dur": 5,
    "args": {
      "External id": 69021,
      "cbid": 211, "correlation": 69021
    }
  },
  {
    "ph": "s", "id": 69021, "pid": 76337, "tid": -914061504, "ts": 1716454222312496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222398926, "dur": 1065,
    "args": {
      "External id": 69023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69023,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69023, "pid": 5, "tid": 7, "ts": 1716454222398926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312507, "dur": 6,
    "args": {
      "External id": 69023,
      "cbid": 211, "correlation": 69023
    }
  },
  {
    "ph": "s", "id": 69023, "pid": 76337, "tid": -914061504, "ts": 1716454222312507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222399992, "dur": 22,
    "args": {
      "External id": 69025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69025,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69025, "pid": 5, "tid": 7, "ts": 1716454222399992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312516, "dur": 5,
    "args": {
      "External id": 69025,
      "cbid": 211, "correlation": 69025
    }
  },
  {
    "ph": "s", "id": 69025, "pid": 76337, "tid": -914061504, "ts": 1716454222312516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222400016, "dur": 35,
    "args": {
      "External id": 69031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69031, "pid": 5, "tid": 7, "ts": 1716454222400016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312545, "dur": 8,
    "args": {
      "External id": 69031,
      "cbid": 211, "correlation": 69031
    }
  },
  {
    "ph": "s", "id": 69031, "pid": 76337, "tid": -914061504, "ts": 1716454222312545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222400052, "dur": 4,
    "args": {
      "External id": 69039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69039,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69039, "pid": 5, "tid": 7, "ts": 1716454222400052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312589, "dur": 9,
    "args": {
      "External id": 69039,
      "cbid": 211, "correlation": 69039
    }
  },
  {
    "ph": "s", "id": 69039, "pid": 76337, "tid": -914061504, "ts": 1716454222312589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312653, "dur": 1,
    "args": {
      "External id": 69055,
      "cbid": 251, "correlation": 69055
    }
  },
  {
    "ph": "f", "id": 69055, "pid": 76337, "tid": -914061504, "ts": 1716454222312653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312658, "dur": 0,
    "args": {
      "External id": 69057,
      "cbid": 251, "correlation": 69057
    }
  },
  {
    "ph": "f", "id": 69057, "pid": 76337, "tid": -914061504, "ts": 1716454222312658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222400057, "dur": 13,
    "args": {
      "External id": 69058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69058,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 69058, "pid": 5, "tid": 7, "ts": 1716454222400057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312660, "dur": 11,
    "args": {
      "External id": 69058,
      "cbid": 211, "correlation": 69058
    }
  },
  {
    "ph": "s", "id": 69058, "pid": 76337, "tid": -914061504, "ts": 1716454222312660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222400071, "dur": 5,
    "args": {
      "External id": 69060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69060,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 69060, "pid": 5, "tid": 7, "ts": 1716454222400071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312673, "dur": 5,
    "args": {
      "External id": 69060,
      "cbid": 211, "correlation": 69060
    }
  },
  {
    "ph": "s", "id": 69060, "pid": 76337, "tid": -914061504, "ts": 1716454222312673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222400078, "dur": 31,
    "args": {
      "External id": 69070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69070,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69070, "pid": 5, "tid": 7, "ts": 1716454222400078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312730, "dur": 15,
    "args": {
      "External id": 69070,
      "cbid": 211, "correlation": 69070
    }
  },
  {
    "ph": "s", "id": 69070, "pid": 76337, "tid": -914061504, "ts": 1716454222312730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222400111, "dur": 33,
    "args": {
      "External id": 69090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69090,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 69090, "pid": 5, "tid": 7, "ts": 1716454222400111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312798, "dur": 11,
    "args": {
      "External id": 69090,
      "cbid": 211, "correlation": 69090
    }
  },
  {
    "ph": "s", "id": 69090, "pid": 76337, "tid": -914061504, "ts": 1716454222312798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222400145, "dur": 4,
    "args": {
      "External id": 69102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69102,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 69102, "pid": 5, "tid": 7, "ts": 1716454222400145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312819, "dur": 6,
    "args": {
      "External id": 69102,
      "cbid": 211, "correlation": 69102
    }
  },
  {
    "ph": "s", "id": 69102, "pid": 76337, "tid": -914061504, "ts": 1716454222312819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222400151, "dur": 32,
    "args": {
      "External id": 69105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69105,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69105, "pid": 5, "tid": 7, "ts": 1716454222400151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312837, "dur": 7,
    "args": {
      "External id": 69105,
      "cbid": 211, "correlation": 69105
    }
  },
  {
    "ph": "s", "id": 69105, "pid": 76337, "tid": -914061504, "ts": 1716454222312837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222400184, "dur": 22,
    "args": {
      "External id": 69114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69114,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69114, "pid": 5, "tid": 7, "ts": 1716454222400184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312878, "dur": 10,
    "args": {
      "External id": 69114,
      "cbid": 211, "correlation": 69114
    }
  },
  {
    "ph": "s", "id": 69114, "pid": 76337, "tid": -914061504, "ts": 1716454222312878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312939, "dur": 0,
    "args": {
      "External id": 69124,
      "cbid": 317, "correlation": 69124
    }
  },
  {
    "ph": "f", "id": 69124, "pid": 76337, "tid": -914061504, "ts": 1716454222312939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312940, "dur": 0,
    "args": {
      "External id": 69125,
      "cbid": 203, "correlation": 69125
    }
  },
  {
    "ph": "f", "id": 69125, "pid": 76337, "tid": -914061504, "ts": 1716454222312940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312940, "dur": 0,
    "args": {
      "External id": 69126,
      "cbid": 205, "correlation": 69126
    }
  },
  {
    "ph": "f", "id": 69126, "pid": 76337, "tid": -914061504, "ts": 1716454222312940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222400207, "dur": 24,
    "args": {
      "External id": 69130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69130, "pid": 5, "tid": 7, "ts": 1716454222400207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312954, "dur": 12,
    "args": {
      "External id": 69130,
      "cbid": 211, "correlation": 69130
    }
  },
  {
    "ph": "s", "id": 69130, "pid": 76337, "tid": -914061504, "ts": 1716454222312954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222400233, "dur": 48,
    "args": {
      "External id": 69132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69132, "pid": 5, "tid": 7, "ts": 1716454222400233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312969, "dur": 13,
    "args": {
      "External id": 69132,
      "cbid": 211, "correlation": 69132
    }
  },
  {
    "ph": "s", "id": 69132, "pid": 76337, "tid": -914061504, "ts": 1716454222312969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222400282, "dur": 709,
    "args": {
      "External id": 69134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69134,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69134, "pid": 5, "tid": 7, "ts": 1716454222400282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312988, "dur": 6,
    "args": {
      "External id": 69134,
      "cbid": 211, "correlation": 69134
    }
  },
  {
    "ph": "s", "id": 69134, "pid": 76337, "tid": -914061504, "ts": 1716454222312988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222400993, "dur": 22,
    "args": {
      "External id": 69136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69136,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69136, "pid": 5, "tid": 7, "ts": 1716454222400993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222312998, "dur": 8,
    "args": {
      "External id": 69136,
      "cbid": 211, "correlation": 69136
    }
  },
  {
    "ph": "s", "id": 69136, "pid": 76337, "tid": -914061504, "ts": 1716454222312998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222401017, "dur": 35,
    "args": {
      "External id": 69142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69142, "pid": 5, "tid": 7, "ts": 1716454222401017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313029, "dur": 9,
    "args": {
      "External id": 69142,
      "cbid": 211, "correlation": 69142
    }
  },
  {
    "ph": "s", "id": 69142, "pid": 76337, "tid": -914061504, "ts": 1716454222313029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313088, "dur": 0,
    "args": {
      "External id": 69152,
      "cbid": 317, "correlation": 69152
    }
  },
  {
    "ph": "f", "id": 69152, "pid": 76337, "tid": -914061504, "ts": 1716454222313088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313089, "dur": 0,
    "args": {
      "External id": 69153,
      "cbid": 203, "correlation": 69153
    }
  },
  {
    "ph": "f", "id": 69153, "pid": 76337, "tid": -914061504, "ts": 1716454222313089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313090, "dur": 0,
    "args": {
      "External id": 69154,
      "cbid": 205, "correlation": 69154
    }
  },
  {
    "ph": "f", "id": 69154, "pid": 76337, "tid": -914061504, "ts": 1716454222313090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222401053, "dur": 30,
    "args": {
      "External id": 69158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69158,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69158, "pid": 5, "tid": 7, "ts": 1716454222401053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313111, "dur": 12,
    "args": {
      "External id": 69158,
      "cbid": 211, "correlation": 69158
    }
  },
  {
    "ph": "s", "id": 69158, "pid": 76337, "tid": -914061504, "ts": 1716454222313111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222401085, "dur": 167,
    "args": {
      "External id": 69160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69160,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69160, "pid": 5, "tid": 7, "ts": 1716454222401085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313134, "dur": 6,
    "args": {
      "External id": 69160,
      "cbid": 211, "correlation": 69160
    }
  },
  {
    "ph": "s", "id": 69160, "pid": 76337, "tid": -914061504, "ts": 1716454222313134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222401254, "dur": 24,
    "args": {
      "External id": 69162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69162,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69162, "pid": 5, "tid": 7, "ts": 1716454222401254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313145, "dur": 5,
    "args": {
      "External id": 69162,
      "cbid": 211, "correlation": 69162
    }
  },
  {
    "ph": "s", "id": 69162, "pid": 76337, "tid": -914061504, "ts": 1716454222313145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222401279, "dur": 34,
    "args": {
      "External id": 69168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69168, "pid": 5, "tid": 7, "ts": 1716454222401279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313170, "dur": 9,
    "args": {
      "External id": 69168,
      "cbid": 211, "correlation": 69168
    }
  },
  {
    "ph": "s", "id": 69168, "pid": 76337, "tid": -914061504, "ts": 1716454222313170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222401315, "dur": 27,
    "args": {
      "External id": 69176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69176,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69176, "pid": 5, "tid": 7, "ts": 1716454222401315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313200, "dur": 8,
    "args": {
      "External id": 69176,
      "cbid": 211, "correlation": 69176
    }
  },
  {
    "ph": "s", "id": 69176, "pid": 76337, "tid": -914061504, "ts": 1716454222313200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222401344, "dur": 20,
    "args": {
      "External id": 69184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69184,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69184, "pid": 5, "tid": 7, "ts": 1716454222401344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313228, "dur": 8,
    "args": {
      "External id": 69184,
      "cbid": 211, "correlation": 69184
    }
  },
  {
    "ph": "s", "id": 69184, "pid": 76337, "tid": -914061504, "ts": 1716454222313228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222401366, "dur": 33,
    "args": {
      "External id": 69204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69204,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 69204, "pid": 5, "tid": 7, "ts": 1716454222401366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313309, "dur": 15,
    "args": {
      "External id": 69204,
      "cbid": 211, "correlation": 69204
    }
  },
  {
    "ph": "s", "id": 69204, "pid": 76337, "tid": -914061504, "ts": 1716454222313309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222401400, "dur": 5,
    "args": {
      "External id": 69216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69216,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 69216, "pid": 5, "tid": 7, "ts": 1716454222401400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313335, "dur": 6,
    "args": {
      "External id": 69216,
      "cbid": 211, "correlation": 69216
    }
  },
  {
    "ph": "s", "id": 69216, "pid": 76337, "tid": -914061504, "ts": 1716454222313335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222401406, "dur": 33,
    "args": {
      "External id": 69219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69219, "pid": 5, "tid": 7, "ts": 1716454222401406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313352, "dur": 6,
    "args": {
      "External id": 69219,
      "cbid": 211, "correlation": 69219
    }
  },
  {
    "ph": "s", "id": 69219, "pid": 76337, "tid": -914061504, "ts": 1716454222313352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313410, "dur": 0,
    "args": {
      "External id": 69230,
      "cbid": 317, "correlation": 69230
    }
  },
  {
    "ph": "f", "id": 69230, "pid": 76337, "tid": -914061504, "ts": 1716454222313410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313410, "dur": 0,
    "args": {
      "External id": 69231,
      "cbid": 203, "correlation": 69231
    }
  },
  {
    "ph": "f", "id": 69231, "pid": 76337, "tid": -914061504, "ts": 1716454222313410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313411, "dur": 0,
    "args": {
      "External id": 69232,
      "cbid": 205, "correlation": 69232
    }
  },
  {
    "ph": "f", "id": 69232, "pid": 76337, "tid": -914061504, "ts": 1716454222313411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222401441, "dur": 22,
    "args": {
      "External id": 69236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69236, "pid": 5, "tid": 7, "ts": 1716454222401441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313424, "dur": 12,
    "args": {
      "External id": 69236,
      "cbid": 211, "correlation": 69236
    }
  },
  {
    "ph": "s", "id": 69236, "pid": 76337, "tid": -914061504, "ts": 1716454222313424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222401464, "dur": 116,
    "args": {
      "External id": 69238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69238,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69238, "pid": 5, "tid": 7, "ts": 1716454222401464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313442, "dur": 6,
    "args": {
      "External id": 69238,
      "cbid": 211, "correlation": 69238
    }
  },
  {
    "ph": "s", "id": 69238, "pid": 76337, "tid": -914061504, "ts": 1716454222313442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222401581, "dur": 25,
    "args": {
      "External id": 69240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69240, "pid": 5, "tid": 7, "ts": 1716454222401581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313451, "dur": 5,
    "args": {
      "External id": 69240,
      "cbid": 211, "correlation": 69240
    }
  },
  {
    "ph": "s", "id": 69240, "pid": 76337, "tid": -914061504, "ts": 1716454222313451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222401608, "dur": 35,
    "args": {
      "External id": 69246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69246, "pid": 5, "tid": 7, "ts": 1716454222401608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313478, "dur": 8,
    "args": {
      "External id": 69246,
      "cbid": 211, "correlation": 69246
    }
  },
  {
    "ph": "s", "id": 69246, "pid": 76337, "tid": -914061504, "ts": 1716454222313478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222401644, "dur": 206,
    "args": {
      "External id": 69255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69255,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69255, "pid": 5, "tid": 7, "ts": 1716454222401644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313558, "dur": 14,
    "args": {
      "External id": 69255,
      "cbid": 211, "correlation": 69255
    }
  },
  {
    "ph": "s", "id": 69255, "pid": 76337, "tid": -914061504, "ts": 1716454222313558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222401851, "dur": 71,
    "args": {
      "External id": 69277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69277,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69277, "pid": 5, "tid": 7, "ts": 1716454222401851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313618, "dur": 11,
    "args": {
      "External id": 69277,
      "cbid": 211, "correlation": 69277
    }
  },
  {
    "ph": "s", "id": 69277, "pid": 76337, "tid": -914061504, "ts": 1716454222313618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313707, "dur": 1,
    "args": {
      "External id": 69288,
      "cbid": 251, "correlation": 69288
    }
  },
  {
    "ph": "f", "id": 69288, "pid": 76337, "tid": -914061504, "ts": 1716454222313707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222401924, "dur": 164,
    "args": {
      "External id": 69289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69289,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69289, "pid": 5, "tid": 7, "ts": 1716454222401924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313712, "dur": 13,
    "args": {
      "External id": 69289,
      "cbid": 211, "correlation": 69289
    }
  },
  {
    "ph": "s", "id": 69289, "pid": 76337, "tid": -914061504, "ts": 1716454222313712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313782, "dur": 1,
    "args": {
      "External id": 69300,
      "cbid": 251, "correlation": 69300
    }
  },
  {
    "ph": "f", "id": 69300, "pid": 76337, "tid": -914061504, "ts": 1716454222313782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222402089, "dur": 155,
    "args": {
      "External id": 69301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69301,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69301, "pid": 5, "tid": 7, "ts": 1716454222402089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313786, "dur": 11,
    "args": {
      "External id": 69301,
      "cbid": 211, "correlation": 69301
    }
  },
  {
    "ph": "s", "id": 69301, "pid": 76337, "tid": -914061504, "ts": 1716454222313786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313850, "dur": 1,
    "args": {
      "External id": 69312,
      "cbid": 251, "correlation": 69312
    }
  },
  {
    "ph": "f", "id": 69312, "pid": 76337, "tid": -914061504, "ts": 1716454222313850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222402246, "dur": 154,
    "args": {
      "External id": 69313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69313,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69313, "pid": 5, "tid": 7, "ts": 1716454222402246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313854, "dur": 12,
    "args": {
      "External id": 69313,
      "cbid": 211, "correlation": 69313
    }
  },
  {
    "ph": "s", "id": 69313, "pid": 76337, "tid": -914061504, "ts": 1716454222313854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222402402, "dur": 2143,
    "args": {
      "External id": 69334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69334,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 69334, "pid": 5, "tid": 7, "ts": 1716454222402402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222313934, "dur": 12,
    "args": {
      "External id": 69334,
      "cbid": 211, "correlation": 69334
    }
  },
  {
    "ph": "s", "id": 69334, "pid": 76337, "tid": -914061504, "ts": 1716454222313934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314040, "dur": 4,
    "args": {
      "External id": 69352,
      "cbid": 251, "correlation": 69352
    }
  },
  {
    "ph": "f", "id": 69352, "pid": 76337, "tid": -914061504, "ts": 1716454222314040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222404546, "dur": 160,
    "args": {
      "External id": 69354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69354,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 69354, "pid": 5, "tid": 7, "ts": 1716454222404546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314049, "dur": 14,
    "args": {
      "External id": 69354,
      "cbid": 211, "correlation": 69354
    }
  },
  {
    "ph": "s", "id": 69354, "pid": 76337, "tid": -914061504, "ts": 1716454222314049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222404707, "dur": 35,
    "args": {
      "External id": 69362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69362,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69362, "pid": 5, "tid": 7, "ts": 1716454222404707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314119, "dur": 12,
    "args": {
      "External id": 69362,
      "cbid": 211, "correlation": 69362
    }
  },
  {
    "ph": "s", "id": 69362, "pid": 76337, "tid": -914061504, "ts": 1716454222314119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222404744, "dur": 51,
    "args": {
      "External id": 69370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69370,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69370, "pid": 5, "tid": 7, "ts": 1716454222404744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314159, "dur": 8,
    "args": {
      "External id": 69370,
      "cbid": 211, "correlation": 69370
    }
  },
  {
    "ph": "s", "id": 69370, "pid": 76337, "tid": -914061504, "ts": 1716454222314159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222404796, "dur": 31,
    "args": {
      "External id": 69381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69381, "pid": 5, "tid": 7, "ts": 1716454222404796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314230, "dur": 12,
    "args": {
      "External id": 69381,
      "cbid": 211, "correlation": 69381
    }
  },
  {
    "ph": "s", "id": 69381, "pid": 76337, "tid": -914061504, "ts": 1716454222314230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222404829, "dur": 37,
    "args": {
      "External id": 69403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69403,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69403, "pid": 5, "tid": 7, "ts": 1716454222404829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314260, "dur": 8,
    "args": {
      "External id": 69403,
      "cbid": 211, "correlation": 69403
    }
  },
  {
    "ph": "s", "id": 69403, "pid": 76337, "tid": -914061504, "ts": 1716454222314260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314344, "dur": 1,
    "args": {
      "External id": 69414,
      "cbid": 251, "correlation": 69414
    }
  },
  {
    "ph": "f", "id": 69414, "pid": 76337, "tid": -914061504, "ts": 1716454222314344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222404868, "dur": 96,
    "args": {
      "External id": 69415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69415,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69415, "pid": 5, "tid": 7, "ts": 1716454222404868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314350, "dur": 13,
    "args": {
      "External id": 69415,
      "cbid": 211, "correlation": 69415
    }
  },
  {
    "ph": "s", "id": 69415, "pid": 76337, "tid": -914061504, "ts": 1716454222314350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314417, "dur": 1,
    "args": {
      "External id": 69426,
      "cbid": 251, "correlation": 69426
    }
  },
  {
    "ph": "f", "id": 69426, "pid": 76337, "tid": -914061504, "ts": 1716454222314417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314420, "dur": 0,
    "args": {
      "External id": 69427,
      "cbid": 251, "correlation": 69427
    }
  },
  {
    "ph": "f", "id": 69427, "pid": 76337, "tid": -914061504, "ts": 1716454222314420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222404965, "dur": 12,
    "args": {
      "External id": 69428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69428,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 69428, "pid": 5, "tid": 7, "ts": 1716454222404965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314422, "dur": 15,
    "args": {
      "External id": 69428,
      "cbid": 211, "correlation": 69428
    }
  },
  {
    "ph": "s", "id": 69428, "pid": 76337, "tid": -914061504, "ts": 1716454222314422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222404979, "dur": 5,
    "args": {
      "External id": 69430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69430,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 69430, "pid": 5, "tid": 7, "ts": 1716454222404979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314439, "dur": 6,
    "args": {
      "External id": 69430,
      "cbid": 211, "correlation": 69430
    }
  },
  {
    "ph": "s", "id": 69430, "pid": 76337, "tid": -914061504, "ts": 1716454222314439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314496, "dur": 1,
    "args": {
      "External id": 69441,
      "cbid": 251, "correlation": 69441
    }
  },
  {
    "ph": "f", "id": 69441, "pid": 76337, "tid": -914061504, "ts": 1716454222314496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314500, "dur": 0,
    "args": {
      "External id": 69442,
      "cbid": 251, "correlation": 69442
    }
  },
  {
    "ph": "f", "id": 69442, "pid": 76337, "tid": -914061504, "ts": 1716454222314500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222404986, "dur": 8,
    "args": {
      "External id": 69443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69443,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 69443, "pid": 5, "tid": 7, "ts": 1716454222404986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314501, "dur": 12,
    "args": {
      "External id": 69443,
      "cbid": 211, "correlation": 69443
    }
  },
  {
    "ph": "s", "id": 69443, "pid": 76337, "tid": -914061504, "ts": 1716454222314501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222404995, "dur": 3,
    "args": {
      "External id": 69445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69445,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 69445, "pid": 5, "tid": 7, "ts": 1716454222404995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314515, "dur": 5,
    "args": {
      "External id": 69445,
      "cbid": 211, "correlation": 69445
    }
  },
  {
    "ph": "s", "id": 69445, "pid": 76337, "tid": -914061504, "ts": 1716454222314515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222405000, "dur": 100,
    "args": {
      "External id": 69466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69466,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 69466, "pid": 5, "tid": 7, "ts": 1716454222405000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314600, "dur": 13,
    "args": {
      "External id": 69466,
      "cbid": 211, "correlation": 69466
    }
  },
  {
    "ph": "s", "id": 69466, "pid": 76337, "tid": -914061504, "ts": 1716454222314600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314699, "dur": 1,
    "args": {
      "External id": 69484,
      "cbid": 251, "correlation": 69484
    }
  },
  {
    "ph": "f", "id": 69484, "pid": 76337, "tid": -914061504, "ts": 1716454222314699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222405101, "dur": 105,
    "args": {
      "External id": 69486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69486,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69486, "pid": 5, "tid": 7, "ts": 1716454222405101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314704, "dur": 13,
    "args": {
      "External id": 69486,
      "cbid": 211, "correlation": 69486
    }
  },
  {
    "ph": "s", "id": 69486, "pid": 76337, "tid": -914061504, "ts": 1716454222314704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222405208, "dur": 19,
    "args": {
      "External id": 69494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69494,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69494, "pid": 5, "tid": 7, "ts": 1716454222405208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314773, "dur": 12,
    "args": {
      "External id": 69494,
      "cbid": 211, "correlation": 69494
    }
  },
  {
    "ph": "s", "id": 69494, "pid": 76337, "tid": -914061504, "ts": 1716454222314773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222405228, "dur": 40,
    "args": {
      "External id": 69502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69502, "pid": 5, "tid": 7, "ts": 1716454222405228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314818, "dur": 10,
    "args": {
      "External id": 69502,
      "cbid": 211, "correlation": 69502
    }
  },
  {
    "ph": "s", "id": 69502, "pid": 76337, "tid": -914061504, "ts": 1716454222314818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222405270, "dur": 37,
    "args": {
      "External id": 69524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69524,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69524, "pid": 5, "tid": 7, "ts": 1716454222405270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314871, "dur": 10,
    "args": {
      "External id": 69524,
      "cbid": 211, "correlation": 69524
    }
  },
  {
    "ph": "s", "id": 69524, "pid": 76337, "tid": -914061504, "ts": 1716454222314871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314959, "dur": 1,
    "args": {
      "External id": 69540,
      "cbid": 251, "correlation": 69540
    }
  },
  {
    "ph": "f", "id": 69540, "pid": 76337, "tid": -914061504, "ts": 1716454222314959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314964, "dur": 0,
    "args": {
      "External id": 69542,
      "cbid": 251, "correlation": 69542
    }
  },
  {
    "ph": "f", "id": 69542, "pid": 76337, "tid": -914061504, "ts": 1716454222314964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222405308, "dur": 587,
    "args": {
      "External id": 69543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69543,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 69543, "pid": 5, "tid": 7, "ts": 1716454222405308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222314968, "dur": 20,
    "args": {
      "External id": 69543,
      "cbid": 211, "correlation": 69543
    }
  },
  {
    "ph": "s", "id": 69543, "pid": 76337, "tid": -914061504, "ts": 1716454222314968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222405897, "dur": 135,
    "args": {
      "External id": 69551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69551,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69551, "pid": 5, "tid": 7, "ts": 1716454222405897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315041, "dur": 13,
    "args": {
      "External id": 69551,
      "cbid": 211, "correlation": 69551
    }
  },
  {
    "ph": "s", "id": 69551, "pid": 76337, "tid": -914061504, "ts": 1716454222315041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222406033, "dur": 134,
    "args": {
      "External id": 69559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69559, "pid": 5, "tid": 7, "ts": 1716454222406033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315070, "dur": 8,
    "args": {
      "External id": 69559,
      "cbid": 211, "correlation": 69559
    }
  },
  {
    "ph": "s", "id": 69559, "pid": 76337, "tid": -914061504, "ts": 1716454222315070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315147, "dur": 1,
    "args": {
      "External id": 69575,
      "cbid": 251, "correlation": 69575
    }
  },
  {
    "ph": "f", "id": 69575, "pid": 76337, "tid": -914061504, "ts": 1716454222315147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222406169, "dur": 333,
    "args": {
      "External id": 69577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69577,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69577, "pid": 5, "tid": 7, "ts": 1716454222406169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315153, "dur": 16,
    "args": {
      "External id": 69577,
      "cbid": 211, "correlation": 69577
    }
  },
  {
    "ph": "s", "id": 69577, "pid": 76337, "tid": -914061504, "ts": 1716454222315153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222406504, "dur": 28,
    "args": {
      "External id": 69585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69585,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69585, "pid": 5, "tid": 7, "ts": 1716454222406504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315199, "dur": 10,
    "args": {
      "External id": 69585,
      "cbid": 211, "correlation": 69585
    }
  },
  {
    "ph": "s", "id": 69585, "pid": 76337, "tid": -914061504, "ts": 1716454222315199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222406533, "dur": 90,
    "args": {
      "External id": 69596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69596,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69596, "pid": 5, "tid": 7, "ts": 1716454222406533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315265, "dur": 13,
    "args": {
      "External id": 69596,
      "cbid": 211, "correlation": 69596
    }
  },
  {
    "ph": "s", "id": 69596, "pid": 76337, "tid": -914061504, "ts": 1716454222315265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315330, "dur": 0,
    "args": {
      "External id": 69608,
      "cbid": 317, "correlation": 69608
    }
  },
  {
    "ph": "f", "id": 69608, "pid": 76337, "tid": -914061504, "ts": 1716454222315330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315331, "dur": 0,
    "args": {
      "External id": 69609,
      "cbid": 203, "correlation": 69609
    }
  },
  {
    "ph": "f", "id": 69609, "pid": 76337, "tid": -914061504, "ts": 1716454222315331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315332, "dur": 0,
    "args": {
      "External id": 69610,
      "cbid": 205, "correlation": 69610
    }
  },
  {
    "ph": "f", "id": 69610, "pid": 76337, "tid": -914061504, "ts": 1716454222315332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222406624, "dur": 26,
    "args": {
      "External id": 69614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69614,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69614, "pid": 5, "tid": 7, "ts": 1716454222406624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315347, "dur": 12,
    "args": {
      "External id": 69614,
      "cbid": 211, "correlation": 69614
    }
  },
  {
    "ph": "s", "id": 69614, "pid": 76337, "tid": -914061504, "ts": 1716454222315347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222406651, "dur": 131,
    "args": {
      "External id": 69616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69616,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69616, "pid": 5, "tid": 7, "ts": 1716454222406651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315366, "dur": 6,
    "args": {
      "External id": 69616,
      "cbid": 211, "correlation": 69616
    }
  },
  {
    "ph": "s", "id": 69616, "pid": 76337, "tid": -914061504, "ts": 1716454222315366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222406784, "dur": 25,
    "args": {
      "External id": 69618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69618,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69618, "pid": 5, "tid": 7, "ts": 1716454222406784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315376, "dur": 5,
    "args": {
      "External id": 69618,
      "cbid": 211, "correlation": 69618
    }
  },
  {
    "ph": "s", "id": 69618, "pid": 76337, "tid": -914061504, "ts": 1716454222315376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222406811, "dur": 35,
    "args": {
      "External id": 69624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69624, "pid": 5, "tid": 7, "ts": 1716454222406811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315404, "dur": 8,
    "args": {
      "External id": 69624,
      "cbid": 211, "correlation": 69624
    }
  },
  {
    "ph": "s", "id": 69624, "pid": 76337, "tid": -914061504, "ts": 1716454222315404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222406847, "dur": 28,
    "args": {
      "External id": 69632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69632,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69632, "pid": 5, "tid": 7, "ts": 1716454222406847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315435, "dur": 8,
    "args": {
      "External id": 69632,
      "cbid": 211, "correlation": 69632
    }
  },
  {
    "ph": "s", "id": 69632, "pid": 76337, "tid": -914061504, "ts": 1716454222315435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222406876, "dur": 108,
    "args": {
      "External id": 69643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69643,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69643, "pid": 5, "tid": 7, "ts": 1716454222406876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315497, "dur": 16,
    "args": {
      "External id": 69643,
      "cbid": 211, "correlation": 69643
    }
  },
  {
    "ph": "s", "id": 69643, "pid": 76337, "tid": -914061504, "ts": 1716454222315497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315556, "dur": 0,
    "args": {
      "External id": 69653,
      "cbid": 317, "correlation": 69653
    }
  },
  {
    "ph": "f", "id": 69653, "pid": 76337, "tid": -914061504, "ts": 1716454222315556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315557, "dur": 0,
    "args": {
      "External id": 69654,
      "cbid": 203, "correlation": 69654
    }
  },
  {
    "ph": "f", "id": 69654, "pid": 76337, "tid": -914061504, "ts": 1716454222315557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315558, "dur": 0,
    "args": {
      "External id": 69655,
      "cbid": 205, "correlation": 69655
    }
  },
  {
    "ph": "f", "id": 69655, "pid": 76337, "tid": -914061504, "ts": 1716454222315558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222406986, "dur": 77,
    "args": {
      "External id": 69659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69659, "pid": 5, "tid": 7, "ts": 1716454222406986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315574, "dur": 11,
    "args": {
      "External id": 69659,
      "cbid": 211, "correlation": 69659
    }
  },
  {
    "ph": "s", "id": 69659, "pid": 76337, "tid": -914061504, "ts": 1716454222315574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222407065, "dur": 48,
    "args": {
      "External id": 69661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69661, "pid": 5, "tid": 7, "ts": 1716454222407065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315588, "dur": 5,
    "args": {
      "External id": 69661,
      "cbid": 211, "correlation": 69661
    }
  },
  {
    "ph": "s", "id": 69661, "pid": 76337, "tid": -914061504, "ts": 1716454222315588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222407114, "dur": 4,
    "args": {
      "External id": 69663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69663,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69663, "pid": 5, "tid": 7, "ts": 1716454222407114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315599, "dur": 6,
    "args": {
      "External id": 69663,
      "cbid": 211, "correlation": 69663
    }
  },
  {
    "ph": "s", "id": 69663, "pid": 76337, "tid": -914061504, "ts": 1716454222315599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315608, "dur": 0,
    "args": {
      "External id": 69664,
      "cbid": 51, "correlation": 69664
    }
  },
  {
    "ph": "s", "id": 69664, "pid": 76337, "tid": -914061504, "ts": 1716454222315608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222407120, "dur": 2401,
    "args": {
      "External id": 69665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69665,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69665, "pid": 5, "tid": 7, "ts": 1716454222407120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315609, "dur": 5,
    "args": {
      "External id": 69665,
      "cbid": 211, "correlation": 69665
    }
  },
  {
    "ph": "s", "id": 69665, "pid": 76337, "tid": -914061504, "ts": 1716454222315609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222409523, "dur": 119,
    "args": {
      "External id": 69670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69670,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69670, "pid": 5, "tid": 7, "ts": 1716454222409523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315638, "dur": 9,
    "args": {
      "External id": 69670,
      "cbid": 211, "correlation": 69670
    }
  },
  {
    "ph": "s", "id": 69670, "pid": 76337, "tid": -914061504, "ts": 1716454222315638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222409643, "dur": 173,
    "args": {
      "External id": 69679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69679,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69679, "pid": 5, "tid": 7, "ts": 1716454222409643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315731, "dur": 14,
    "args": {
      "External id": 69679,
      "cbid": 211, "correlation": 69679
    }
  },
  {
    "ph": "s", "id": 69679, "pid": 76337, "tid": -914061504, "ts": 1716454222315731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222409818, "dur": 136,
    "args": {
      "External id": 69699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69699,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 69699, "pid": 5, "tid": 7, "ts": 1716454222409818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315801, "dur": 12,
    "args": {
      "External id": 69699,
      "cbid": 211, "correlation": 69699
    }
  },
  {
    "ph": "s", "id": 69699, "pid": 76337, "tid": -914061504, "ts": 1716454222315801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222409956, "dur": 5,
    "args": {
      "External id": 69711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69711,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 69711, "pid": 5, "tid": 7, "ts": 1716454222409956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315823, "dur": 9,
    "args": {
      "External id": 69711,
      "cbid": 211, "correlation": 69711
    }
  },
  {
    "ph": "s", "id": 69711, "pid": 76337, "tid": -914061504, "ts": 1716454222315823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222409962, "dur": 171,
    "args": {
      "External id": 69714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69714,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69714, "pid": 5, "tid": 7, "ts": 1716454222409962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315845, "dur": 7,
    "args": {
      "External id": 69714,
      "cbid": 211, "correlation": 69714
    }
  },
  {
    "ph": "s", "id": 69714, "pid": 76337, "tid": -914061504, "ts": 1716454222315845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222410134, "dur": 104,
    "args": {
      "External id": 69723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69723,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69723, "pid": 5, "tid": 7, "ts": 1716454222410134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315887, "dur": 9,
    "args": {
      "External id": 69723,
      "cbid": 211, "correlation": 69723
    }
  },
  {
    "ph": "s", "id": 69723, "pid": 76337, "tid": -914061504, "ts": 1716454222315887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315939, "dur": 0,
    "args": {
      "External id": 69733,
      "cbid": 317, "correlation": 69733
    }
  },
  {
    "ph": "f", "id": 69733, "pid": 76337, "tid": -914061504, "ts": 1716454222315939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315940, "dur": 0,
    "args": {
      "External id": 69734,
      "cbid": 203, "correlation": 69734
    }
  },
  {
    "ph": "f", "id": 69734, "pid": 76337, "tid": -914061504, "ts": 1716454222315940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315941, "dur": 0,
    "args": {
      "External id": 69735,
      "cbid": 205, "correlation": 69735
    }
  },
  {
    "ph": "f", "id": 69735, "pid": 76337, "tid": -914061504, "ts": 1716454222315941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222410239, "dur": 115,
    "args": {
      "External id": 69739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69739, "pid": 5, "tid": 7, "ts": 1716454222410239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315956, "dur": 11,
    "args": {
      "External id": 69739,
      "cbid": 211, "correlation": 69739
    }
  },
  {
    "ph": "s", "id": 69739, "pid": 76337, "tid": -914061504, "ts": 1716454222315956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222410356, "dur": 36,
    "args": {
      "External id": 69741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69741,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69741, "pid": 5, "tid": 7, "ts": 1716454222410356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315970, "dur": 15,
    "args": {
      "External id": 69741,
      "cbid": 211, "correlation": 69741
    }
  },
  {
    "ph": "s", "id": 69741, "pid": 76337, "tid": -914061504, "ts": 1716454222315970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222410393, "dur": 4,
    "args": {
      "External id": 69743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69743, "pid": 5, "tid": 7, "ts": 1716454222410393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222315990, "dur": 6,
    "args": {
      "External id": 69743,
      "cbid": 211, "correlation": 69743
    }
  },
  {
    "ph": "s", "id": 69743, "pid": 76337, "tid": -914061504, "ts": 1716454222315990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316000, "dur": 0,
    "args": {
      "External id": 69744,
      "cbid": 51, "correlation": 69744
    }
  },
  {
    "ph": "s", "id": 69744, "pid": 76337, "tid": -914061504, "ts": 1716454222316000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222410399, "dur": 2205,
    "args": {
      "External id": 69745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69745,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69745, "pid": 5, "tid": 7, "ts": 1716454222410399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316001, "dur": 5,
    "args": {
      "External id": 69745,
      "cbid": 211, "correlation": 69745
    }
  },
  {
    "ph": "s", "id": 69745, "pid": 76337, "tid": -914061504, "ts": 1716454222316001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222412606, "dur": 63,
    "args": {
      "External id": 69750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69750, "pid": 5, "tid": 7, "ts": 1716454222412606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316029, "dur": 9,
    "args": {
      "External id": 69750,
      "cbid": 211, "correlation": 69750
    }
  },
  {
    "ph": "s", "id": 69750, "pid": 76337, "tid": -914061504, "ts": 1716454222316029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222412670, "dur": 4,
    "args": {
      "External id": 69758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69758,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69758, "pid": 5, "tid": 7, "ts": 1716454222412670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316073, "dur": 9,
    "args": {
      "External id": 69758,
      "cbid": 211, "correlation": 69758
    }
  },
  {
    "ph": "s", "id": 69758, "pid": 76337, "tid": -914061504, "ts": 1716454222316073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316138, "dur": 1,
    "args": {
      "External id": 69774,
      "cbid": 251, "correlation": 69774
    }
  },
  {
    "ph": "f", "id": 69774, "pid": 76337, "tid": -914061504, "ts": 1716454222316138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316144, "dur": 0,
    "args": {
      "External id": 69776,
      "cbid": 251, "correlation": 69776
    }
  },
  {
    "ph": "f", "id": 69776, "pid": 76337, "tid": -914061504, "ts": 1716454222316144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222412675, "dur": 12,
    "args": {
      "External id": 69777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69777,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 69777, "pid": 5, "tid": 7, "ts": 1716454222412675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316146, "dur": 15,
    "args": {
      "External id": 69777,
      "cbid": 211, "correlation": 69777
    }
  },
  {
    "ph": "s", "id": 69777, "pid": 76337, "tid": -914061504, "ts": 1716454222316146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222412689, "dur": 5,
    "args": {
      "External id": 69779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69779,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 69779, "pid": 5, "tid": 7, "ts": 1716454222412689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316163, "dur": 6,
    "args": {
      "External id": 69779,
      "cbid": 211, "correlation": 69779
    }
  },
  {
    "ph": "s", "id": 69779, "pid": 76337, "tid": -914061504, "ts": 1716454222316163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222412695, "dur": 57,
    "args": {
      "External id": 69789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69789,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69789, "pid": 5, "tid": 7, "ts": 1716454222412695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316221, "dur": 12,
    "args": {
      "External id": 69789,
      "cbid": 211, "correlation": 69789
    }
  },
  {
    "ph": "s", "id": 69789, "pid": 76337, "tid": -914061504, "ts": 1716454222316221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222412754, "dur": 55,
    "args": {
      "External id": 69809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69809,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 69809, "pid": 5, "tid": 7, "ts": 1716454222412754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316287, "dur": 10,
    "args": {
      "External id": 69809,
      "cbid": 211, "correlation": 69809
    }
  },
  {
    "ph": "s", "id": 69809, "pid": 76337, "tid": -914061504, "ts": 1716454222316287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222412810, "dur": 4,
    "args": {
      "External id": 69821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69821,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69821, "pid": 5, "tid": 7, "ts": 1716454222412810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316307, "dur": 6,
    "args": {
      "External id": 69821,
      "cbid": 211, "correlation": 69821
    }
  },
  {
    "ph": "s", "id": 69821, "pid": 76337, "tid": -914061504, "ts": 1716454222316307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222412816, "dur": 59,
    "args": {
      "External id": 69824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69824, "pid": 5, "tid": 7, "ts": 1716454222412816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316326, "dur": 7,
    "args": {
      "External id": 69824,
      "cbid": 211, "correlation": 69824
    }
  },
  {
    "ph": "s", "id": 69824, "pid": 76337, "tid": -914061504, "ts": 1716454222316326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222412877, "dur": 38,
    "args": {
      "External id": 69833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69833,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69833, "pid": 5, "tid": 7, "ts": 1716454222412877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316366, "dur": 10,
    "args": {
      "External id": 69833,
      "cbid": 211, "correlation": 69833
    }
  },
  {
    "ph": "s", "id": 69833, "pid": 76337, "tid": -914061504, "ts": 1716454222316366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316429, "dur": 0,
    "args": {
      "External id": 69843,
      "cbid": 317, "correlation": 69843
    }
  },
  {
    "ph": "f", "id": 69843, "pid": 76337, "tid": -914061504, "ts": 1716454222316429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316430, "dur": 0,
    "args": {
      "External id": 69844,
      "cbid": 203, "correlation": 69844
    }
  },
  {
    "ph": "f", "id": 69844, "pid": 76337, "tid": -914061504, "ts": 1716454222316430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316431, "dur": 0,
    "args": {
      "External id": 69845,
      "cbid": 205, "correlation": 69845
    }
  },
  {
    "ph": "f", "id": 69845, "pid": 76337, "tid": -914061504, "ts": 1716454222316431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222412916, "dur": 41,
    "args": {
      "External id": 69849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69849, "pid": 5, "tid": 7, "ts": 1716454222412916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316451, "dur": 12,
    "args": {
      "External id": 69849,
      "cbid": 211, "correlation": 69849
    }
  },
  {
    "ph": "s", "id": 69849, "pid": 76337, "tid": -914061504, "ts": 1716454222316451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222412958, "dur": 15,
    "args": {
      "External id": 69851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69851, "pid": 5, "tid": 7, "ts": 1716454222412958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316466, "dur": 5,
    "args": {
      "External id": 69851,
      "cbid": 211, "correlation": 69851
    }
  },
  {
    "ph": "s", "id": 69851, "pid": 76337, "tid": -914061504, "ts": 1716454222316466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222412975, "dur": 4,
    "args": {
      "External id": 69853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69853, "pid": 5, "tid": 7, "ts": 1716454222412975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316476, "dur": 5,
    "args": {
      "External id": 69853,
      "cbid": 211, "correlation": 69853
    }
  },
  {
    "ph": "s", "id": 69853, "pid": 76337, "tid": -914061504, "ts": 1716454222316476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316484, "dur": 0,
    "args": {
      "External id": 69854,
      "cbid": 51, "correlation": 69854
    }
  },
  {
    "ph": "s", "id": 69854, "pid": 76337, "tid": -914061504, "ts": 1716454222316484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222412980, "dur": 765,
    "args": {
      "External id": 69855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69855,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69855, "pid": 5, "tid": 7, "ts": 1716454222412980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316485, "dur": 5,
    "args": {
      "External id": 69855,
      "cbid": 211, "correlation": 69855
    }
  },
  {
    "ph": "s", "id": 69855, "pid": 76337, "tid": -914061504, "ts": 1716454222316485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222413747, "dur": 63,
    "args": {
      "External id": 69860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69860,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69860, "pid": 5, "tid": 7, "ts": 1716454222413747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316514, "dur": 8,
    "args": {
      "External id": 69860,
      "cbid": 211, "correlation": 69860
    }
  },
  {
    "ph": "s", "id": 69860, "pid": 76337, "tid": -914061504, "ts": 1716454222316514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316570, "dur": 0,
    "args": {
      "External id": 69870,
      "cbid": 317, "correlation": 69870
    }
  },
  {
    "ph": "f", "id": 69870, "pid": 76337, "tid": -914061504, "ts": 1716454222316570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316571, "dur": 0,
    "args": {
      "External id": 69871,
      "cbid": 203, "correlation": 69871
    }
  },
  {
    "ph": "f", "id": 69871, "pid": 76337, "tid": -914061504, "ts": 1716454222316571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316572, "dur": 0,
    "args": {
      "External id": 69872,
      "cbid": 205, "correlation": 69872
    }
  },
  {
    "ph": "f", "id": 69872, "pid": 76337, "tid": -914061504, "ts": 1716454222316572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222413811, "dur": 4,
    "args": {
      "External id": 69876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69876, "pid": 5, "tid": 7, "ts": 1716454222413811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316587, "dur": 11,
    "args": {
      "External id": 69876,
      "cbid": 211, "correlation": 69876
    }
  },
  {
    "ph": "s", "id": 69876, "pid": 76337, "tid": -914061504, "ts": 1716454222316587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316603, "dur": 0,
    "args": {
      "External id": 69877,
      "cbid": 51, "correlation": 69877
    }
  },
  {
    "ph": "s", "id": 69877, "pid": 76337, "tid": -914061504, "ts": 1716454222316603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222413816, "dur": 289,
    "args": {
      "External id": 69878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69878,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 69878, "pid": 5, "tid": 7, "ts": 1716454222413816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316604, "dur": 7,
    "args": {
      "External id": 69878,
      "cbid": 211, "correlation": 69878
    }
  },
  {
    "ph": "s", "id": 69878, "pid": 76337, "tid": -914061504, "ts": 1716454222316604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222414107, "dur": 63,
    "args": {
      "External id": 69883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69883, "pid": 5, "tid": 7, "ts": 1716454222414107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316631, "dur": 8,
    "args": {
      "External id": 69883,
      "cbid": 211, "correlation": 69883
    }
  },
  {
    "ph": "s", "id": 69883, "pid": 76337, "tid": -914061504, "ts": 1716454222316631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222414171, "dur": 50,
    "args": {
      "External id": 69891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69891,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69891, "pid": 5, "tid": 7, "ts": 1716454222414171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316659, "dur": 8,
    "args": {
      "External id": 69891,
      "cbid": 211, "correlation": 69891
    }
  },
  {
    "ph": "s", "id": 69891, "pid": 76337, "tid": -914061504, "ts": 1716454222316659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222414223, "dur": 36,
    "args": {
      "External id": 69899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69899,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69899, "pid": 5, "tid": 7, "ts": 1716454222414223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316688, "dur": 8,
    "args": {
      "External id": 69899,
      "cbid": 211, "correlation": 69899
    }
  },
  {
    "ph": "s", "id": 69899, "pid": 76337, "tid": -914061504, "ts": 1716454222316688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222414260, "dur": 55,
    "args": {
      "External id": 69919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69919,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 69919, "pid": 5, "tid": 7, "ts": 1716454222414260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316772, "dur": 12,
    "args": {
      "External id": 69919,
      "cbid": 211, "correlation": 69919
    }
  },
  {
    "ph": "s", "id": 69919, "pid": 76337, "tid": -914061504, "ts": 1716454222316772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222414316, "dur": 4,
    "args": {
      "External id": 69931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69931,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 69931, "pid": 5, "tid": 7, "ts": 1716454222414316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316795, "dur": 7,
    "args": {
      "External id": 69931,
      "cbid": 211, "correlation": 69931
    }
  },
  {
    "ph": "s", "id": 69931, "pid": 76337, "tid": -914061504, "ts": 1716454222316795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222414322, "dur": 58,
    "args": {
      "External id": 69934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69934, "pid": 5, "tid": 7, "ts": 1716454222414322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316812, "dur": 6,
    "args": {
      "External id": 69934,
      "cbid": 211, "correlation": 69934
    }
  },
  {
    "ph": "s", "id": 69934, "pid": 76337, "tid": -914061504, "ts": 1716454222316812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316869, "dur": 0,
    "args": {
      "External id": 69945,
      "cbid": 317, "correlation": 69945
    }
  },
  {
    "ph": "f", "id": 69945, "pid": 76337, "tid": -914061504, "ts": 1716454222316869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316870, "dur": 0,
    "args": {
      "External id": 69946,
      "cbid": 203, "correlation": 69946
    }
  },
  {
    "ph": "f", "id": 69946, "pid": 76337, "tid": -914061504, "ts": 1716454222316870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316871, "dur": 0,
    "args": {
      "External id": 69947,
      "cbid": 205, "correlation": 69947
    }
  },
  {
    "ph": "f", "id": 69947, "pid": 76337, "tid": -914061504, "ts": 1716454222316871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316901, "dur": 2,
    "args": {
      "External id": 69951,
      "cbid": 251, "correlation": 69951
    }
  },
  {
    "ph": "f", "id": 69951, "pid": 76337, "tid": -914061504, "ts": 1716454222316901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316904, "dur": 1,
    "args": {
      "External id": 69952,
      "cbid": 251, "correlation": 69952
    }
  },
  {
    "ph": "f", "id": 69952, "pid": 76337, "tid": -914061504, "ts": 1716454222316904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316906, "dur": 1,
    "args": {
      "External id": 69953,
      "cbid": 251, "correlation": 69953
    }
  },
  {
    "ph": "f", "id": 69953, "pid": 76337, "tid": -914061504, "ts": 1716454222316906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316908, "dur": 1,
    "args": {
      "External id": 69954,
      "cbid": 251, "correlation": 69954
    }
  },
  {
    "ph": "f", "id": 69954, "pid": 76337, "tid": -914061504, "ts": 1716454222316908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316909, "dur": 1,
    "args": {
      "External id": 69955,
      "cbid": 251, "correlation": 69955
    }
  },
  {
    "ph": "f", "id": 69955, "pid": 76337, "tid": -914061504, "ts": 1716454222316909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316911, "dur": 1,
    "args": {
      "External id": 69956,
      "cbid": 251, "correlation": 69956
    }
  },
  {
    "ph": "f", "id": 69956, "pid": 76337, "tid": -914061504, "ts": 1716454222316911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316912, "dur": 0,
    "args": {
      "External id": 69957,
      "cbid": 251, "correlation": 69957
    }
  },
  {
    "ph": "f", "id": 69957, "pid": 76337, "tid": -914061504, "ts": 1716454222316912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316914, "dur": 1,
    "args": {
      "External id": 69958,
      "cbid": 251, "correlation": 69958
    }
  },
  {
    "ph": "f", "id": 69958, "pid": 76337, "tid": -914061504, "ts": 1716454222316914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316916, "dur": 0,
    "args": {
      "External id": 69959,
      "cbid": 251, "correlation": 69959
    }
  },
  {
    "ph": "f", "id": 69959, "pid": 76337, "tid": -914061504, "ts": 1716454222316916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222414381, "dur": 123,
    "args": {
      "External id": 69960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69960,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 69960, "pid": 5, "tid": 7, "ts": 1716454222414381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316919, "dur": 12,
    "args": {
      "External id": 69960,
      "cbid": 211, "correlation": 69960
    }
  },
  {
    "ph": "s", "id": 69960, "pid": 76337, "tid": -914061504, "ts": 1716454222316919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222414506, "dur": 63,
    "args": {
      "External id": 69966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69966,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69966, "pid": 5, "tid": 7, "ts": 1716454222414506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222316954, "dur": 9,
    "args": {
      "External id": 69966,
      "cbid": 211, "correlation": 69966
    }
  },
  {
    "ph": "s", "id": 69966, "pid": 76337, "tid": -914061504, "ts": 1716454222316954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222414570, "dur": 628,
    "args": {
      "External id": 69975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69975,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69975, "pid": 5, "tid": 7, "ts": 1716454222414570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317048, "dur": 14,
    "args": {
      "External id": 69975,
      "cbid": 211, "correlation": 69975
    }
  },
  {
    "ph": "s", "id": 69975, "pid": 76337, "tid": -914061504, "ts": 1716454222317048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222415199, "dur": 199,
    "args": {
      "External id": 69997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 69997,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 69997, "pid": 5, "tid": 7, "ts": 1716454222415199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317109, "dur": 11,
    "args": {
      "External id": 69997,
      "cbid": 211, "correlation": 69997
    }
  },
  {
    "ph": "s", "id": 69997, "pid": 76337, "tid": -914061504, "ts": 1716454222317109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317200, "dur": 1,
    "args": {
      "External id": 70008,
      "cbid": 251, "correlation": 70008
    }
  },
  {
    "ph": "f", "id": 70008, "pid": 76337, "tid": -914061504, "ts": 1716454222317200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222415400, "dur": 211,
    "args": {
      "External id": 70009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70009,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70009, "pid": 5, "tid": 7, "ts": 1716454222415400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317205, "dur": 13,
    "args": {
      "External id": 70009,
      "cbid": 211, "correlation": 70009
    }
  },
  {
    "ph": "s", "id": 70009, "pid": 76337, "tid": -914061504, "ts": 1716454222317205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317274, "dur": 1,
    "args": {
      "External id": 70020,
      "cbid": 251, "correlation": 70020
    }
  },
  {
    "ph": "f", "id": 70020, "pid": 76337, "tid": -914061504, "ts": 1716454222317274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222415612, "dur": 203,
    "args": {
      "External id": 70021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70021,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70021, "pid": 5, "tid": 7, "ts": 1716454222415612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317278, "dur": 11,
    "args": {
      "External id": 70021,
      "cbid": 211, "correlation": 70021
    }
  },
  {
    "ph": "s", "id": 70021, "pid": 76337, "tid": -914061504, "ts": 1716454222317278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317340, "dur": 1,
    "args": {
      "External id": 70032,
      "cbid": 251, "correlation": 70032
    }
  },
  {
    "ph": "f", "id": 70032, "pid": 76337, "tid": -914061504, "ts": 1716454222317340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222415816, "dur": 201,
    "args": {
      "External id": 70033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70033,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70033, "pid": 5, "tid": 7, "ts": 1716454222415816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222317344, "dur": 1608,
    "args": {
      "External id": 70033,
      "cbid": 211, "correlation": 70033
    }
  },
  {
    "ph": "s", "id": 70033, "pid": 76337, "tid": -914061504, "ts": 1716454222317344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222416019, "dur": 20503,
    "args": {
      "External id": 70054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70054,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70054, "pid": 5, "tid": 7, "ts": 1716454222416019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319028, "dur": 143,
    "args": {
      "External id": 70054,
      "cbid": 211, "correlation": 70054
    }
  },
  {
    "ph": "s", "id": 70054, "pid": 76337, "tid": -914061504, "ts": 1716454222319028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319261, "dur": 1,
    "args": {
      "External id": 70072,
      "cbid": 251, "correlation": 70072
    }
  },
  {
    "ph": "f", "id": 70072, "pid": 76337, "tid": -914061504, "ts": 1716454222319261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222436524, "dur": 217,
    "args": {
      "External id": 70074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70074,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70074, "pid": 5, "tid": 7, "ts": 1716454222436524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319267, "dur": 14,
    "args": {
      "External id": 70074,
      "cbid": 211, "correlation": 70074
    }
  },
  {
    "ph": "s", "id": 70074, "pid": 76337, "tid": -914061504, "ts": 1716454222319267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222436742, "dur": 67,
    "args": {
      "External id": 70082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70082,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70082, "pid": 5, "tid": 7, "ts": 1716454222436742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319337, "dur": 12,
    "args": {
      "External id": 70082,
      "cbid": 211, "correlation": 70082
    }
  },
  {
    "ph": "s", "id": 70082, "pid": 76337, "tid": -914061504, "ts": 1716454222319337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222436810, "dur": 97,
    "args": {
      "External id": 70090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70090,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70090, "pid": 5, "tid": 7, "ts": 1716454222436810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319376, "dur": 18,
    "args": {
      "External id": 70090,
      "cbid": 211, "correlation": 70090
    }
  },
  {
    "ph": "s", "id": 70090, "pid": 76337, "tid": -914061504, "ts": 1716454222319376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222436909, "dur": 57,
    "args": {
      "External id": 70101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70101, "pid": 5, "tid": 7, "ts": 1716454222436909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319458, "dur": 40,
    "args": {
      "External id": 70101,
      "cbid": 211, "correlation": 70101
    }
  },
  {
    "ph": "s", "id": 70101, "pid": 76337, "tid": -914061504, "ts": 1716454222319458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222436967, "dur": 101,
    "args": {
      "External id": 70123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70123,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70123, "pid": 5, "tid": 7, "ts": 1716454222436967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319517, "dur": 91,
    "args": {
      "External id": 70123,
      "cbid": 211, "correlation": 70123
    }
  },
  {
    "ph": "s", "id": 70123, "pid": 76337, "tid": -914061504, "ts": 1716454222319517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319684, "dur": 1,
    "args": {
      "External id": 70134,
      "cbid": 251, "correlation": 70134
    }
  },
  {
    "ph": "f", "id": 70134, "pid": 76337, "tid": -914061504, "ts": 1716454222319684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222437069, "dur": 115,
    "args": {
      "External id": 70135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70135,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70135, "pid": 5, "tid": 7, "ts": 1716454222437069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319689, "dur": 13,
    "args": {
      "External id": 70135,
      "cbid": 211, "correlation": 70135
    }
  },
  {
    "ph": "s", "id": 70135, "pid": 76337, "tid": -914061504, "ts": 1716454222319689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319766, "dur": 1,
    "args": {
      "External id": 70146,
      "cbid": 251, "correlation": 70146
    }
  },
  {
    "ph": "f", "id": 70146, "pid": 76337, "tid": -914061504, "ts": 1716454222319766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319770, "dur": 0,
    "args": {
      "External id": 70147,
      "cbid": 251, "correlation": 70147
    }
  },
  {
    "ph": "f", "id": 70147, "pid": 76337, "tid": -914061504, "ts": 1716454222319770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222437185, "dur": 11,
    "args": {
      "External id": 70148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70148,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 70148, "pid": 5, "tid": 7, "ts": 1716454222437185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319771, "dur": 14,
    "args": {
      "External id": 70148,
      "cbid": 211, "correlation": 70148
    }
  },
  {
    "ph": "s", "id": 70148, "pid": 76337, "tid": -914061504, "ts": 1716454222319771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222437198, "dur": 6,
    "args": {
      "External id": 70150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70150,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 70150, "pid": 5, "tid": 7, "ts": 1716454222437198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319789, "dur": 8,
    "args": {
      "External id": 70150,
      "cbid": 211, "correlation": 70150
    }
  },
  {
    "ph": "s", "id": 70150, "pid": 76337, "tid": -914061504, "ts": 1716454222319789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319852, "dur": 1,
    "args": {
      "External id": 70161,
      "cbid": 251, "correlation": 70161
    }
  },
  {
    "ph": "f", "id": 70161, "pid": 76337, "tid": -914061504, "ts": 1716454222319852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319855, "dur": 0,
    "args": {
      "External id": 70162,
      "cbid": 251, "correlation": 70162
    }
  },
  {
    "ph": "f", "id": 70162, "pid": 76337, "tid": -914061504, "ts": 1716454222319855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222437205, "dur": 6,
    "args": {
      "External id": 70163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70163,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 70163, "pid": 5, "tid": 7, "ts": 1716454222437205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319857, "dur": 11,
    "args": {
      "External id": 70163,
      "cbid": 211, "correlation": 70163
    }
  },
  {
    "ph": "s", "id": 70163, "pid": 76337, "tid": -914061504, "ts": 1716454222319857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222437213, "dur": 4,
    "args": {
      "External id": 70165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70165,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 70165, "pid": 5, "tid": 7, "ts": 1716454222437213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319870, "dur": 6,
    "args": {
      "External id": 70165,
      "cbid": 211, "correlation": 70165
    }
  },
  {
    "ph": "s", "id": 70165, "pid": 76337, "tid": -914061504, "ts": 1716454222319870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222437218, "dur": 170,
    "args": {
      "External id": 70186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70186,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70186, "pid": 5, "tid": 7, "ts": 1716454222437218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222319957, "dur": 14,
    "args": {
      "External id": 70186,
      "cbid": 211, "correlation": 70186
    }
  },
  {
    "ph": "s", "id": 70186, "pid": 76337, "tid": -914061504, "ts": 1716454222319957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320066, "dur": 2,
    "args": {
      "External id": 70204,
      "cbid": 251, "correlation": 70204
    }
  },
  {
    "ph": "f", "id": 70204, "pid": 76337, "tid": -914061504, "ts": 1716454222320066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222437389, "dur": 114,
    "args": {
      "External id": 70206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70206,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70206, "pid": 5, "tid": 7, "ts": 1716454222437389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320072, "dur": 14,
    "args": {
      "External id": 70206,
      "cbid": 211, "correlation": 70206
    }
  },
  {
    "ph": "s", "id": 70206, "pid": 76337, "tid": -914061504, "ts": 1716454222320072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222437505, "dur": 35,
    "args": {
      "External id": 70214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70214,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70214, "pid": 5, "tid": 7, "ts": 1716454222437505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320143, "dur": 15,
    "args": {
      "External id": 70214,
      "cbid": 211, "correlation": 70214
    }
  },
  {
    "ph": "s", "id": 70214, "pid": 76337, "tid": -914061504, "ts": 1716454222320143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222437541, "dur": 69,
    "args": {
      "External id": 70222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70222, "pid": 5, "tid": 7, "ts": 1716454222437541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320189, "dur": 10,
    "args": {
      "External id": 70222,
      "cbid": 211, "correlation": 70222
    }
  },
  {
    "ph": "s", "id": 70222, "pid": 76337, "tid": -914061504, "ts": 1716454222320189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222437612, "dur": 101,
    "args": {
      "External id": 70244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70244,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70244, "pid": 5, "tid": 7, "ts": 1716454222437612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320240, "dur": 530,
    "args": {
      "External id": 70244,
      "cbid": 211, "correlation": 70244
    }
  },
  {
    "ph": "s", "id": 70244, "pid": 76337, "tid": -914061504, "ts": 1716454222320240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320849, "dur": 1,
    "args": {
      "External id": 70260,
      "cbid": 251, "correlation": 70260
    }
  },
  {
    "ph": "f", "id": 70260, "pid": 76337, "tid": -914061504, "ts": 1716454222320849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222437715, "dur": 627,
    "args": {
      "External id": 70262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70262,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70262, "pid": 5, "tid": 7, "ts": 1716454222437715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222320855, "dur": 181,
    "args": {
      "External id": 70262,
      "cbid": 211, "correlation": 70262
    }
  },
  {
    "ph": "s", "id": 70262, "pid": 76337, "tid": -914061504, "ts": 1716454222320855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222438344, "dur": 260,
    "args": {
      "External id": 70270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70270, "pid": 5, "tid": 7, "ts": 1716454222438344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321091, "dur": 205,
    "args": {
      "External id": 70270,
      "cbid": 211, "correlation": 70270
    }
  },
  {
    "ph": "s", "id": 70270, "pid": 76337, "tid": -914061504, "ts": 1716454222321091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222438605, "dur": 261,
    "args": {
      "External id": 70278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70278, "pid": 5, "tid": 7, "ts": 1716454222438605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321316, "dur": 358,
    "args": {
      "External id": 70278,
      "cbid": 211, "correlation": 70278
    }
  },
  {
    "ph": "s", "id": 70278, "pid": 76337, "tid": -914061504, "ts": 1716454222321316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321753, "dur": 1,
    "args": {
      "External id": 70294,
      "cbid": 251, "correlation": 70294
    }
  },
  {
    "ph": "f", "id": 70294, "pid": 76337, "tid": -914061504, "ts": 1716454222321753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321758, "dur": 0,
    "args": {
      "External id": 70296,
      "cbid": 251, "correlation": 70296
    }
  },
  {
    "ph": "f", "id": 70296, "pid": 76337, "tid": -914061504, "ts": 1716454222321758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222438867, "dur": 374,
    "args": {
      "External id": 70297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70297,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 70297, "pid": 5, "tid": 7, "ts": 1716454222438867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321761, "dur": 13,
    "args": {
      "External id": 70297,
      "cbid": 211, "correlation": 70297
    }
  },
  {
    "ph": "s", "id": 70297, "pid": 76337, "tid": -914061504, "ts": 1716454222321761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222439243, "dur": 51,
    "args": {
      "External id": 70305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70305,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70305, "pid": 5, "tid": 7, "ts": 1716454222439243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321804, "dur": 96,
    "args": {
      "External id": 70305,
      "cbid": 211, "correlation": 70305
    }
  },
  {
    "ph": "s", "id": 70305, "pid": 76337, "tid": -914061504, "ts": 1716454222321804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222439295, "dur": 173,
    "args": {
      "External id": 70316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70316, "pid": 5, "tid": 7, "ts": 1716454222439295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222321958, "dur": 67,
    "args": {
      "External id": 70316,
      "cbid": 211, "correlation": 70316
    }
  },
  {
    "ph": "s", "id": 70316, "pid": 76337, "tid": -914061504, "ts": 1716454222321958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322079, "dur": 0,
    "args": {
      "External id": 70328,
      "cbid": 317, "correlation": 70328
    }
  },
  {
    "ph": "f", "id": 70328, "pid": 76337, "tid": -914061504, "ts": 1716454222322079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322080, "dur": 0,
    "args": {
      "External id": 70329,
      "cbid": 203, "correlation": 70329
    }
  },
  {
    "ph": "f", "id": 70329, "pid": 76337, "tid": -914061504, "ts": 1716454222322080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322081, "dur": 0,
    "args": {
      "External id": 70330,
      "cbid": 205, "correlation": 70330
    }
  },
  {
    "ph": "f", "id": 70330, "pid": 76337, "tid": -914061504, "ts": 1716454222322081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322105, "dur": 1,
    "args": {
      "External id": 70334,
      "cbid": 251, "correlation": 70334
    }
  },
  {
    "ph": "f", "id": 70334, "pid": 76337, "tid": -914061504, "ts": 1716454222322105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322107, "dur": 0,
    "args": {
      "External id": 70335,
      "cbid": 251, "correlation": 70335
    }
  },
  {
    "ph": "f", "id": 70335, "pid": 76337, "tid": -914061504, "ts": 1716454222322107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322107, "dur": 0,
    "args": {
      "External id": 70336,
      "cbid": 251, "correlation": 70336
    }
  },
  {
    "ph": "f", "id": 70336, "pid": 76337, "tid": -914061504, "ts": 1716454222322107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322109, "dur": 0,
    "args": {
      "External id": 70337,
      "cbid": 251, "correlation": 70337
    }
  },
  {
    "ph": "f", "id": 70337, "pid": 76337, "tid": -914061504, "ts": 1716454222322109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322109, "dur": 1,
    "args": {
      "External id": 70338,
      "cbid": 251, "correlation": 70338
    }
  },
  {
    "ph": "f", "id": 70338, "pid": 76337, "tid": -914061504, "ts": 1716454222322109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322111, "dur": 0,
    "args": {
      "External id": 70339,
      "cbid": 251, "correlation": 70339
    }
  },
  {
    "ph": "f", "id": 70339, "pid": 76337, "tid": -914061504, "ts": 1716454222322111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322111, "dur": 0,
    "args": {
      "External id": 70340,
      "cbid": 251, "correlation": 70340
    }
  },
  {
    "ph": "f", "id": 70340, "pid": 76337, "tid": -914061504, "ts": 1716454222322111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322112, "dur": 0,
    "args": {
      "External id": 70341,
      "cbid": 251, "correlation": 70341
    }
  },
  {
    "ph": "f", "id": 70341, "pid": 76337, "tid": -914061504, "ts": 1716454222322112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322114, "dur": 0,
    "args": {
      "External id": 70342,
      "cbid": 251, "correlation": 70342
    }
  },
  {
    "ph": "f", "id": 70342, "pid": 76337, "tid": -914061504, "ts": 1716454222322114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222439470, "dur": 125,
    "args": {
      "External id": 70343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70343,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70343, "pid": 5, "tid": 7, "ts": 1716454222439470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322116, "dur": 13,
    "args": {
      "External id": 70343,
      "cbid": 211, "correlation": 70343
    }
  },
  {
    "ph": "s", "id": 70343, "pid": 76337, "tid": -914061504, "ts": 1716454222322116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222439596, "dur": 63,
    "args": {
      "External id": 70349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70349, "pid": 5, "tid": 7, "ts": 1716454222439596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322152, "dur": 9,
    "args": {
      "External id": 70349,
      "cbid": 211, "correlation": 70349
    }
  },
  {
    "ph": "s", "id": 70349, "pid": 76337, "tid": -914061504, "ts": 1716454222322152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222439661, "dur": 50,
    "args": {
      "External id": 70357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70357,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70357, "pid": 5, "tid": 7, "ts": 1716454222439661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322185, "dur": 9,
    "args": {
      "External id": 70357,
      "cbid": 211, "correlation": 70357
    }
  },
  {
    "ph": "s", "id": 70357, "pid": 76337, "tid": -914061504, "ts": 1716454222322185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222439713, "dur": 105,
    "args": {
      "External id": 70366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70366,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70366, "pid": 5, "tid": 7, "ts": 1716454222439713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322224, "dur": 13,
    "args": {
      "External id": 70366,
      "cbid": 211, "correlation": 70366
    }
  },
  {
    "ph": "s", "id": 70366, "pid": 76337, "tid": -914061504, "ts": 1716454222322224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222439819, "dur": 100,
    "args": {
      "External id": 70386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70386,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 70386, "pid": 5, "tid": 7, "ts": 1716454222439819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322300, "dur": 12,
    "args": {
      "External id": 70386,
      "cbid": 211, "correlation": 70386
    }
  },
  {
    "ph": "s", "id": 70386, "pid": 76337, "tid": -914061504, "ts": 1716454222322300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222439920, "dur": 5,
    "args": {
      "External id": 70398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70398,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 70398, "pid": 5, "tid": 7, "ts": 1716454222439920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322323, "dur": 278,
    "args": {
      "External id": 70398,
      "cbid": 211, "correlation": 70398
    }
  },
  {
    "ph": "s", "id": 70398, "pid": 76337, "tid": -914061504, "ts": 1716454222322323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222439927, "dur": 114,
    "args": {
      "External id": 70401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70401, "pid": 5, "tid": 7, "ts": 1716454222439927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322613, "dur": 7,
    "args": {
      "External id": 70401,
      "cbid": 211, "correlation": 70401
    }
  },
  {
    "ph": "s", "id": 70401, "pid": 76337, "tid": -914061504, "ts": 1716454222322613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222440042, "dur": 71,
    "args": {
      "External id": 70410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70410,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70410, "pid": 5, "tid": 7, "ts": 1716454222440042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322653, "dur": 11,
    "args": {
      "External id": 70410,
      "cbid": 211, "correlation": 70410
    }
  },
  {
    "ph": "s", "id": 70410, "pid": 76337, "tid": -914061504, "ts": 1716454222322653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322706, "dur": 0,
    "args": {
      "External id": 70420,
      "cbid": 317, "correlation": 70420
    }
  },
  {
    "ph": "f", "id": 70420, "pid": 76337, "tid": -914061504, "ts": 1716454222322706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322707, "dur": 0,
    "args": {
      "External id": 70421,
      "cbid": 203, "correlation": 70421
    }
  },
  {
    "ph": "f", "id": 70421, "pid": 76337, "tid": -914061504, "ts": 1716454222322707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322707, "dur": 0,
    "args": {
      "External id": 70422,
      "cbid": 205, "correlation": 70422
    }
  },
  {
    "ph": "f", "id": 70422, "pid": 76337, "tid": -914061504, "ts": 1716454222322707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222440114, "dur": 78,
    "args": {
      "External id": 70426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70426,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70426, "pid": 5, "tid": 7, "ts": 1716454222440114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322724, "dur": 12,
    "args": {
      "External id": 70426,
      "cbid": 211, "correlation": 70426
    }
  },
  {
    "ph": "s", "id": 70426, "pid": 76337, "tid": -914061504, "ts": 1716454222322724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222440194, "dur": 27,
    "args": {
      "External id": 70428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70428,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70428, "pid": 5, "tid": 7, "ts": 1716454222440194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322738, "dur": 5,
    "args": {
      "External id": 70428,
      "cbid": 211, "correlation": 70428
    }
  },
  {
    "ph": "s", "id": 70428, "pid": 76337, "tid": -914061504, "ts": 1716454222322738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222440222, "dur": 4,
    "args": {
      "External id": 70430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 70430, "pid": 5, "tid": 7, "ts": 1716454222440222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322748, "dur": 6,
    "args": {
      "External id": 70430,
      "cbid": 211, "correlation": 70430
    }
  },
  {
    "ph": "s", "id": 70430, "pid": 76337, "tid": -914061504, "ts": 1716454222322748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322758, "dur": 0,
    "args": {
      "External id": 70431,
      "cbid": 51, "correlation": 70431
    }
  },
  {
    "ph": "s", "id": 70431, "pid": 76337, "tid": -914061504, "ts": 1716454222322758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222440227, "dur": 1493,
    "args": {
      "External id": 70432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70432,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70432, "pid": 5, "tid": 7, "ts": 1716454222440227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322758, "dur": 8,
    "args": {
      "External id": 70432,
      "cbid": 211, "correlation": 70432
    }
  },
  {
    "ph": "s", "id": 70432, "pid": 76337, "tid": -914061504, "ts": 1716454222322758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222441722, "dur": 63,
    "args": {
      "External id": 70437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70437, "pid": 5, "tid": 7, "ts": 1716454222441722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322789, "dur": 9,
    "args": {
      "External id": 70437,
      "cbid": 211, "correlation": 70437
    }
  },
  {
    "ph": "s", "id": 70437, "pid": 76337, "tid": -914061504, "ts": 1716454222322789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222441786, "dur": 4,
    "args": {
      "External id": 70445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70445,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 70445, "pid": 5, "tid": 7, "ts": 1716454222441786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322833, "dur": 10,
    "args": {
      "External id": 70445,
      "cbid": 211, "correlation": 70445
    }
  },
  {
    "ph": "s", "id": 70445, "pid": 76337, "tid": -914061504, "ts": 1716454222322833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322898, "dur": 1,
    "args": {
      "External id": 70461,
      "cbid": 251, "correlation": 70461
    }
  },
  {
    "ph": "f", "id": 70461, "pid": 76337, "tid": -914061504, "ts": 1716454222322898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322904, "dur": 0,
    "args": {
      "External id": 70463,
      "cbid": 251, "correlation": 70463
    }
  },
  {
    "ph": "f", "id": 70463, "pid": 76337, "tid": -914061504, "ts": 1716454222322904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222441791, "dur": 12,
    "args": {
      "External id": 70464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70464,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 70464, "pid": 5, "tid": 7, "ts": 1716454222441791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322905, "dur": 11,
    "args": {
      "External id": 70464,
      "cbid": 211, "correlation": 70464
    }
  },
  {
    "ph": "s", "id": 70464, "pid": 76337, "tid": -914061504, "ts": 1716454222322905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222441804, "dur": 5,
    "args": {
      "External id": 70466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70466,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 70466, "pid": 5, "tid": 7, "ts": 1716454222441804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222322919, "dur": 213,
    "args": {
      "External id": 70466,
      "cbid": 211, "correlation": 70466
    }
  },
  {
    "ph": "s", "id": 70466, "pid": 76337, "tid": -914061504, "ts": 1716454222322919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222441811, "dur": 57,
    "args": {
      "External id": 70476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70476,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70476, "pid": 5, "tid": 7, "ts": 1716454222441811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323186, "dur": 14,
    "args": {
      "External id": 70476,
      "cbid": 211, "correlation": 70476
    }
  },
  {
    "ph": "s", "id": 70476, "pid": 76337, "tid": -914061504, "ts": 1716454222323186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222441870, "dur": 54,
    "args": {
      "External id": 70496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70496,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 70496, "pid": 5, "tid": 7, "ts": 1716454222441870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323254, "dur": 11,
    "args": {
      "External id": 70496,
      "cbid": 211, "correlation": 70496
    }
  },
  {
    "ph": "s", "id": 70496, "pid": 76337, "tid": -914061504, "ts": 1716454222323254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222441926, "dur": 4,
    "args": {
      "External id": 70508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70508,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 70508, "pid": 5, "tid": 7, "ts": 1716454222441926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323277, "dur": 7,
    "args": {
      "External id": 70508,
      "cbid": 211, "correlation": 70508
    }
  },
  {
    "ph": "s", "id": 70508, "pid": 76337, "tid": -914061504, "ts": 1716454222323277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222441931, "dur": 58,
    "args": {
      "External id": 70511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70511, "pid": 5, "tid": 7, "ts": 1716454222441931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323297, "dur": 7,
    "args": {
      "External id": 70511,
      "cbid": 211, "correlation": 70511
    }
  },
  {
    "ph": "s", "id": 70511, "pid": 76337, "tid": -914061504, "ts": 1716454222323297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222441990, "dur": 37,
    "args": {
      "External id": 70520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70520,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70520, "pid": 5, "tid": 7, "ts": 1716454222441990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323338, "dur": 10,
    "args": {
      "External id": 70520,
      "cbid": 211, "correlation": 70520
    }
  },
  {
    "ph": "s", "id": 70520, "pid": 76337, "tid": -914061504, "ts": 1716454222323338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323402, "dur": 0,
    "args": {
      "External id": 70530,
      "cbid": 317, "correlation": 70530
    }
  },
  {
    "ph": "f", "id": 70530, "pid": 76337, "tid": -914061504, "ts": 1716454222323402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323402, "dur": 0,
    "args": {
      "External id": 70531,
      "cbid": 203, "correlation": 70531
    }
  },
  {
    "ph": "f", "id": 70531, "pid": 76337, "tid": -914061504, "ts": 1716454222323402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323403, "dur": 0,
    "args": {
      "External id": 70532,
      "cbid": 205, "correlation": 70532
    }
  },
  {
    "ph": "f", "id": 70532, "pid": 76337, "tid": -914061504, "ts": 1716454222323403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222442029, "dur": 40,
    "args": {
      "External id": 70536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70536,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70536, "pid": 5, "tid": 7, "ts": 1716454222442029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323418, "dur": 12,
    "args": {
      "External id": 70536,
      "cbid": 211, "correlation": 70536
    }
  },
  {
    "ph": "s", "id": 70536, "pid": 76337, "tid": -914061504, "ts": 1716454222323418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222442071, "dur": 16,
    "args": {
      "External id": 70538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70538,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70538, "pid": 5, "tid": 7, "ts": 1716454222442071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323433, "dur": 6,
    "args": {
      "External id": 70538,
      "cbid": 211, "correlation": 70538
    }
  },
  {
    "ph": "s", "id": 70538, "pid": 76337, "tid": -914061504, "ts": 1716454222323433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222442088, "dur": 3,
    "args": {
      "External id": 70540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70540,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 70540, "pid": 5, "tid": 7, "ts": 1716454222442088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323443, "dur": 6,
    "args": {
      "External id": 70540,
      "cbid": 211, "correlation": 70540
    }
  },
  {
    "ph": "s", "id": 70540, "pid": 76337, "tid": -914061504, "ts": 1716454222323443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323451, "dur": 0,
    "args": {
      "External id": 70541,
      "cbid": 51, "correlation": 70541
    }
  },
  {
    "ph": "s", "id": 70541, "pid": 76337, "tid": -914061504, "ts": 1716454222323451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222442093, "dur": 765,
    "args": {
      "External id": 70542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70542,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70542, "pid": 5, "tid": 7, "ts": 1716454222442093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323452, "dur": 5,
    "args": {
      "External id": 70542,
      "cbid": 211, "correlation": 70542
    }
  },
  {
    "ph": "s", "id": 70542, "pid": 76337, "tid": -914061504, "ts": 1716454222323452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222442859, "dur": 62,
    "args": {
      "External id": 70547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70547,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70547, "pid": 5, "tid": 7, "ts": 1716454222442859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323480, "dur": 8,
    "args": {
      "External id": 70547,
      "cbid": 211, "correlation": 70547
    }
  },
  {
    "ph": "s", "id": 70547, "pid": 76337, "tid": -914061504, "ts": 1716454222323480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323536, "dur": 0,
    "args": {
      "External id": 70557,
      "cbid": 317, "correlation": 70557
    }
  },
  {
    "ph": "f", "id": 70557, "pid": 76337, "tid": -914061504, "ts": 1716454222323536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323537, "dur": 0,
    "args": {
      "External id": 70558,
      "cbid": 203, "correlation": 70558
    }
  },
  {
    "ph": "f", "id": 70558, "pid": 76337, "tid": -914061504, "ts": 1716454222323537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323537, "dur": 0,
    "args": {
      "External id": 70559,
      "cbid": 205, "correlation": 70559
    }
  },
  {
    "ph": "f", "id": 70559, "pid": 76337, "tid": -914061504, "ts": 1716454222323537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222442923, "dur": 75,
    "args": {
      "External id": 70563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70563,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70563, "pid": 5, "tid": 7, "ts": 1716454222442923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323551, "dur": 12,
    "args": {
      "External id": 70563,
      "cbid": 211, "correlation": 70563
    }
  },
  {
    "ph": "s", "id": 70563, "pid": 76337, "tid": -914061504, "ts": 1716454222323551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222442999, "dur": 222,
    "args": {
      "External id": 70565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70565,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70565, "pid": 5, "tid": 7, "ts": 1716454222442999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323570, "dur": 11,
    "args": {
      "External id": 70565,
      "cbid": 211, "correlation": 70565
    }
  },
  {
    "ph": "s", "id": 70565, "pid": 76337, "tid": -914061504, "ts": 1716454222323570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222443223, "dur": 41,
    "args": {
      "External id": 70567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70567,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70567, "pid": 5, "tid": 7, "ts": 1716454222443223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222323586, "dur": 553,
    "args": {
      "External id": 70567,
      "cbid": 211, "correlation": 70567
    }
  },
  {
    "ph": "s", "id": 70567, "pid": 76337, "tid": -914061504, "ts": 1716454222323586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222443265, "dur": 63,
    "args": {
      "External id": 70573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70573, "pid": 5, "tid": 7, "ts": 1716454222443265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324161, "dur": 9,
    "args": {
      "External id": 70573,
      "cbid": 211, "correlation": 70573
    }
  },
  {
    "ph": "s", "id": 70573, "pid": 76337, "tid": -914061504, "ts": 1716454222324161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222443330, "dur": 50,
    "args": {
      "External id": 70581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70581,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70581, "pid": 5, "tid": 7, "ts": 1716454222443330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324193, "dur": 8,
    "args": {
      "External id": 70581,
      "cbid": 211, "correlation": 70581
    }
  },
  {
    "ph": "s", "id": 70581, "pid": 76337, "tid": -914061504, "ts": 1716454222324193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222443381, "dur": 36,
    "args": {
      "External id": 70589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70589,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70589, "pid": 5, "tid": 7, "ts": 1716454222443381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324221, "dur": 33,
    "args": {
      "External id": 70589,
      "cbid": 211, "correlation": 70589
    }
  },
  {
    "ph": "s", "id": 70589, "pid": 76337, "tid": -914061504, "ts": 1716454222324221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222443418, "dur": 54,
    "args": {
      "External id": 70609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70609,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 70609, "pid": 5, "tid": 7, "ts": 1716454222443418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324329, "dur": 12,
    "args": {
      "External id": 70609,
      "cbid": 211, "correlation": 70609
    }
  },
  {
    "ph": "s", "id": 70609, "pid": 76337, "tid": -914061504, "ts": 1716454222324329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222443474, "dur": 5,
    "args": {
      "External id": 70621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70621,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 70621, "pid": 5, "tid": 7, "ts": 1716454222443474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324351, "dur": 6,
    "args": {
      "External id": 70621,
      "cbid": 211, "correlation": 70621
    }
  },
  {
    "ph": "s", "id": 70621, "pid": 76337, "tid": -914061504, "ts": 1716454222324351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222443480, "dur": 58,
    "args": {
      "External id": 70624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70624, "pid": 5, "tid": 7, "ts": 1716454222443480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324369, "dur": 6,
    "args": {
      "External id": 70624,
      "cbid": 211, "correlation": 70624
    }
  },
  {
    "ph": "s", "id": 70624, "pid": 76337, "tid": -914061504, "ts": 1716454222324369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324426, "dur": 0,
    "args": {
      "External id": 70635,
      "cbid": 317, "correlation": 70635
    }
  },
  {
    "ph": "f", "id": 70635, "pid": 76337, "tid": -914061504, "ts": 1716454222324426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324427, "dur": 3,
    "args": {
      "External id": 70636,
      "cbid": 203, "correlation": 70636
    }
  },
  {
    "ph": "f", "id": 70636, "pid": 76337, "tid": -914061504, "ts": 1716454222324427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324431, "dur": 0,
    "args": {
      "External id": 70637,
      "cbid": 205, "correlation": 70637
    }
  },
  {
    "ph": "f", "id": 70637, "pid": 76337, "tid": -914061504, "ts": 1716454222324431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324453, "dur": 1,
    "args": {
      "External id": 70641,
      "cbid": 251, "correlation": 70641
    }
  },
  {
    "ph": "f", "id": 70641, "pid": 76337, "tid": -914061504, "ts": 1716454222324453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324455, "dur": 0,
    "args": {
      "External id": 70642,
      "cbid": 251, "correlation": 70642
    }
  },
  {
    "ph": "f", "id": 70642, "pid": 76337, "tid": -914061504, "ts": 1716454222324455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324456, "dur": 0,
    "args": {
      "External id": 70643,
      "cbid": 251, "correlation": 70643
    }
  },
  {
    "ph": "f", "id": 70643, "pid": 76337, "tid": -914061504, "ts": 1716454222324456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324457, "dur": 0,
    "args": {
      "External id": 70644,
      "cbid": 251, "correlation": 70644
    }
  },
  {
    "ph": "f", "id": 70644, "pid": 76337, "tid": -914061504, "ts": 1716454222324457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324458, "dur": 0,
    "args": {
      "External id": 70645,
      "cbid": 251, "correlation": 70645
    }
  },
  {
    "ph": "f", "id": 70645, "pid": 76337, "tid": -914061504, "ts": 1716454222324458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324459, "dur": 0,
    "args": {
      "External id": 70646,
      "cbid": 251, "correlation": 70646
    }
  },
  {
    "ph": "f", "id": 70646, "pid": 76337, "tid": -914061504, "ts": 1716454222324459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324460, "dur": 0,
    "args": {
      "External id": 70647,
      "cbid": 251, "correlation": 70647
    }
  },
  {
    "ph": "f", "id": 70647, "pid": 76337, "tid": -914061504, "ts": 1716454222324460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324461, "dur": 0,
    "args": {
      "External id": 70648,
      "cbid": 251, "correlation": 70648
    }
  },
  {
    "ph": "f", "id": 70648, "pid": 76337, "tid": -914061504, "ts": 1716454222324461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324462, "dur": 0,
    "args": {
      "External id": 70649,
      "cbid": 251, "correlation": 70649
    }
  },
  {
    "ph": "f", "id": 70649, "pid": 76337, "tid": -914061504, "ts": 1716454222324462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222443539, "dur": 122,
    "args": {
      "External id": 70650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70650,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70650, "pid": 5, "tid": 7, "ts": 1716454222443539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324464, "dur": 13,
    "args": {
      "External id": 70650,
      "cbid": 211, "correlation": 70650
    }
  },
  {
    "ph": "s", "id": 70650, "pid": 76337, "tid": -914061504, "ts": 1716454222324464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222443663, "dur": 63,
    "args": {
      "External id": 70656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70656,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70656, "pid": 5, "tid": 7, "ts": 1716454222443663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324500, "dur": 9,
    "args": {
      "External id": 70656,
      "cbid": 211, "correlation": 70656
    }
  },
  {
    "ph": "s", "id": 70656, "pid": 76337, "tid": -914061504, "ts": 1716454222324500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222443727, "dur": 567,
    "args": {
      "External id": 70665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70665,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70665, "pid": 5, "tid": 7, "ts": 1716454222443727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324583, "dur": 14,
    "args": {
      "External id": 70665,
      "cbid": 211, "correlation": 70665
    }
  },
  {
    "ph": "s", "id": 70665, "pid": 76337, "tid": -914061504, "ts": 1716454222324583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222444296, "dur": 199,
    "args": {
      "External id": 70687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70687,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70687, "pid": 5, "tid": 7, "ts": 1716454222444296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324640, "dur": 10,
    "args": {
      "External id": 70687,
      "cbid": 211, "correlation": 70687
    }
  },
  {
    "ph": "s", "id": 70687, "pid": 76337, "tid": -914061504, "ts": 1716454222324640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324728, "dur": 1,
    "args": {
      "External id": 70698,
      "cbid": 251, "correlation": 70698
    }
  },
  {
    "ph": "f", "id": 70698, "pid": 76337, "tid": -914061504, "ts": 1716454222324728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222444496, "dur": 208,
    "args": {
      "External id": 70699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70699,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70699, "pid": 5, "tid": 7, "ts": 1716454222444496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324734, "dur": 13,
    "args": {
      "External id": 70699,
      "cbid": 211, "correlation": 70699
    }
  },
  {
    "ph": "s", "id": 70699, "pid": 76337, "tid": -914061504, "ts": 1716454222324734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324802, "dur": 1,
    "args": {
      "External id": 70710,
      "cbid": 251, "correlation": 70710
    }
  },
  {
    "ph": "f", "id": 70710, "pid": 76337, "tid": -914061504, "ts": 1716454222324802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222444705, "dur": 204,
    "args": {
      "External id": 70711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70711,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70711, "pid": 5, "tid": 7, "ts": 1716454222444705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324806, "dur": 11,
    "args": {
      "External id": 70711,
      "cbid": 211, "correlation": 70711
    }
  },
  {
    "ph": "s", "id": 70711, "pid": 76337, "tid": -914061504, "ts": 1716454222324806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324874, "dur": 1,
    "args": {
      "External id": 70722,
      "cbid": 251, "correlation": 70722
    }
  },
  {
    "ph": "f", "id": 70722, "pid": 76337, "tid": -914061504, "ts": 1716454222324874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222444911, "dur": 201,
    "args": {
      "External id": 70723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70723,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70723, "pid": 5, "tid": 7, "ts": 1716454222444911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324878, "dur": 12,
    "args": {
      "External id": 70723,
      "cbid": 211, "correlation": 70723
    }
  },
  {
    "ph": "s", "id": 70723, "pid": 76337, "tid": -914061504, "ts": 1716454222324878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222445113, "dur": 19126,
    "args": {
      "External id": 70744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70744,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70744, "pid": 5, "tid": 7, "ts": 1716454222445113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222324959, "dur": 12,
    "args": {
      "External id": 70744,
      "cbid": 211, "correlation": 70744
    }
  },
  {
    "ph": "s", "id": 70744, "pid": 76337, "tid": -914061504, "ts": 1716454222324959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222325067, "dur": 1,
    "args": {
      "External id": 70762,
      "cbid": 251, "correlation": 70762
    }
  },
  {
    "ph": "f", "id": 70762, "pid": 76337, "tid": -914061504, "ts": 1716454222325067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222464241, "dur": 205,
    "args": {
      "External id": 70764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70764,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70764, "pid": 5, "tid": 7, "ts": 1716454222464241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222325073, "dur": 13,
    "args": {
      "External id": 70764,
      "cbid": 211, "correlation": 70764
    }
  },
  {
    "ph": "s", "id": 70764, "pid": 76337, "tid": -914061504, "ts": 1716454222325073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222464447, "dur": 67,
    "args": {
      "External id": 70772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70772,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70772, "pid": 5, "tid": 7, "ts": 1716454222464447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222325142, "dur": 12,
    "args": {
      "External id": 70772,
      "cbid": 211, "correlation": 70772
    }
  },
  {
    "ph": "s", "id": 70772, "pid": 76337, "tid": -914061504, "ts": 1716454222325142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222464516, "dur": 98,
    "args": {
      "External id": 70780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70780,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70780, "pid": 5, "tid": 7, "ts": 1716454222464516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222325180, "dur": 50,
    "args": {
      "External id": 70780,
      "cbid": 211, "correlation": 70780
    }
  },
  {
    "ph": "s", "id": 70780, "pid": 76337, "tid": -914061504, "ts": 1716454222325180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222464615, "dur": 55,
    "args": {
      "External id": 70791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70791,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70791, "pid": 5, "tid": 7, "ts": 1716454222464615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222325295, "dur": 96,
    "args": {
      "External id": 70791,
      "cbid": 211, "correlation": 70791
    }
  },
  {
    "ph": "s", "id": 70791, "pid": 76337, "tid": -914061504, "ts": 1716454222325295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222464671, "dur": 94,
    "args": {
      "External id": 70813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70813,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70813, "pid": 5, "tid": 7, "ts": 1716454222464671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222325410, "dur": 2127,
    "args": {
      "External id": 70813,
      "cbid": 211, "correlation": 70813
    }
  },
  {
    "ph": "s", "id": 70813, "pid": 76337, "tid": -914061504, "ts": 1716454222325410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327616, "dur": 1,
    "args": {
      "External id": 70824,
      "cbid": 251, "correlation": 70824
    }
  },
  {
    "ph": "f", "id": 70824, "pid": 76337, "tid": -914061504, "ts": 1716454222327616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222464767, "dur": 104,
    "args": {
      "External id": 70825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70825,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70825, "pid": 5, "tid": 7, "ts": 1716454222464767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327621, "dur": 76,
    "args": {
      "External id": 70825,
      "cbid": 211, "correlation": 70825
    }
  },
  {
    "ph": "s", "id": 70825, "pid": 76337, "tid": -914061504, "ts": 1716454222327621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327756, "dur": 1,
    "args": {
      "External id": 70836,
      "cbid": 251, "correlation": 70836
    }
  },
  {
    "ph": "f", "id": 70836, "pid": 76337, "tid": -914061504, "ts": 1716454222327756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327759, "dur": 0,
    "args": {
      "External id": 70837,
      "cbid": 251, "correlation": 70837
    }
  },
  {
    "ph": "f", "id": 70837, "pid": 76337, "tid": -914061504, "ts": 1716454222327759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222464873, "dur": 10,
    "args": {
      "External id": 70838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70838,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 70838, "pid": 5, "tid": 7, "ts": 1716454222464873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327761, "dur": 12,
    "args": {
      "External id": 70838,
      "cbid": 211, "correlation": 70838
    }
  },
  {
    "ph": "s", "id": 70838, "pid": 76337, "tid": -914061504, "ts": 1716454222327761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222464884, "dur": 5,
    "args": {
      "External id": 70840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70840,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 70840, "pid": 5, "tid": 7, "ts": 1716454222464884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327775, "dur": 10,
    "args": {
      "External id": 70840,
      "cbid": 211, "correlation": 70840
    }
  },
  {
    "ph": "s", "id": 70840, "pid": 76337, "tid": -914061504, "ts": 1716454222327775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327841, "dur": 1,
    "args": {
      "External id": 70851,
      "cbid": 251, "correlation": 70851
    }
  },
  {
    "ph": "f", "id": 70851, "pid": 76337, "tid": -914061504, "ts": 1716454222327841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327844, "dur": 0,
    "args": {
      "External id": 70852,
      "cbid": 251, "correlation": 70852
    }
  },
  {
    "ph": "f", "id": 70852, "pid": 76337, "tid": -914061504, "ts": 1716454222327844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222464891, "dur": 6,
    "args": {
      "External id": 70853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70853,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 70853, "pid": 5, "tid": 7, "ts": 1716454222464891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327846, "dur": 12,
    "args": {
      "External id": 70853,
      "cbid": 211, "correlation": 70853
    }
  },
  {
    "ph": "s", "id": 70853, "pid": 76337, "tid": -914061504, "ts": 1716454222327846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222464898, "dur": 4,
    "args": {
      "External id": 70855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70855,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 70855, "pid": 5, "tid": 7, "ts": 1716454222464898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327859, "dur": 6,
    "args": {
      "External id": 70855,
      "cbid": 211, "correlation": 70855
    }
  },
  {
    "ph": "s", "id": 70855, "pid": 76337, "tid": -914061504, "ts": 1716454222327859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222464903, "dur": 159,
    "args": {
      "External id": 70876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70876,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70876, "pid": 5, "tid": 7, "ts": 1716454222464903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222327956, "dur": 14,
    "args": {
      "External id": 70876,
      "cbid": 211, "correlation": 70876
    }
  },
  {
    "ph": "s", "id": 70876, "pid": 76337, "tid": -914061504, "ts": 1716454222327956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328064, "dur": 1,
    "args": {
      "External id": 70894,
      "cbid": 251, "correlation": 70894
    }
  },
  {
    "ph": "f", "id": 70894, "pid": 76337, "tid": -914061504, "ts": 1716454222328064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222465064, "dur": 110,
    "args": {
      "External id": 70896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70896,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 70896, "pid": 5, "tid": 7, "ts": 1716454222465064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328071, "dur": 14,
    "args": {
      "External id": 70896,
      "cbid": 211, "correlation": 70896
    }
  },
  {
    "ph": "s", "id": 70896, "pid": 76337, "tid": -914061504, "ts": 1716454222328071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222465175, "dur": 35,
    "args": {
      "External id": 70904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70904,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70904, "pid": 5, "tid": 7, "ts": 1716454222465175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328140, "dur": 13,
    "args": {
      "External id": 70904,
      "cbid": 211, "correlation": 70904
    }
  },
  {
    "ph": "s", "id": 70904, "pid": 76337, "tid": -914061504, "ts": 1716454222328140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222465211, "dur": 67,
    "args": {
      "External id": 70912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70912, "pid": 5, "tid": 7, "ts": 1716454222465211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328181, "dur": 9,
    "args": {
      "External id": 70912,
      "cbid": 211, "correlation": 70912
    }
  },
  {
    "ph": "s", "id": 70912, "pid": 76337, "tid": -914061504, "ts": 1716454222328181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222465279, "dur": 94,
    "args": {
      "External id": 70934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70934,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70934, "pid": 5, "tid": 7, "ts": 1716454222465279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328233, "dur": 10,
    "args": {
      "External id": 70934,
      "cbid": 211, "correlation": 70934
    }
  },
  {
    "ph": "s", "id": 70934, "pid": 76337, "tid": -914061504, "ts": 1716454222328233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328322, "dur": 1,
    "args": {
      "External id": 70950,
      "cbid": 251, "correlation": 70950
    }
  },
  {
    "ph": "f", "id": 70950, "pid": 76337, "tid": -914061504, "ts": 1716454222328322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222465374, "dur": 583,
    "args": {
      "External id": 70952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70952,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 70952, "pid": 5, "tid": 7, "ts": 1716454222465374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328327, "dur": 13,
    "args": {
      "External id": 70952,
      "cbid": 211, "correlation": 70952
    }
  },
  {
    "ph": "s", "id": 70952, "pid": 76337, "tid": -914061504, "ts": 1716454222328327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222465959, "dur": 250,
    "args": {
      "External id": 70960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70960, "pid": 5, "tid": 7, "ts": 1716454222465959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328393, "dur": 13,
    "args": {
      "External id": 70960,
      "cbid": 211, "correlation": 70960
    }
  },
  {
    "ph": "s", "id": 70960, "pid": 76337, "tid": -914061504, "ts": 1716454222328393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222466210, "dur": 254,
    "args": {
      "External id": 70968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70968, "pid": 5, "tid": 7, "ts": 1716454222466210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328423, "dur": 8,
    "args": {
      "External id": 70968,
      "cbid": 211, "correlation": 70968
    }
  },
  {
    "ph": "s", "id": 70968, "pid": 76337, "tid": -914061504, "ts": 1716454222328423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328504, "dur": 1,
    "args": {
      "External id": 70984,
      "cbid": 251, "correlation": 70984
    }
  },
  {
    "ph": "f", "id": 70984, "pid": 76337, "tid": -914061504, "ts": 1716454222328504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328509, "dur": 0,
    "args": {
      "External id": 70986,
      "cbid": 251, "correlation": 70986
    }
  },
  {
    "ph": "f", "id": 70986, "pid": 76337, "tid": -914061504, "ts": 1716454222328509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222466465, "dur": 366,
    "args": {
      "External id": 70987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70987,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 70987, "pid": 5, "tid": 7, "ts": 1716454222466465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328511, "dur": 13,
    "args": {
      "External id": 70987,
      "cbid": 211, "correlation": 70987
    }
  },
  {
    "ph": "s", "id": 70987, "pid": 76337, "tid": -914061504, "ts": 1716454222328511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222466833, "dur": 50,
    "args": {
      "External id": 70995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 70995,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 70995, "pid": 5, "tid": 7, "ts": 1716454222466833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328554, "dur": 10,
    "args": {
      "External id": 70995,
      "cbid": 211, "correlation": 70995
    }
  },
  {
    "ph": "s", "id": 70995, "pid": 76337, "tid": -914061504, "ts": 1716454222328554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222466884, "dur": 160,
    "args": {
      "External id": 71006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71006, "pid": 5, "tid": 7, "ts": 1716454222466884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328621, "dur": 275,
    "args": {
      "External id": 71006,
      "cbid": 211, "correlation": 71006
    }
  },
  {
    "ph": "s", "id": 71006, "pid": 76337, "tid": -914061504, "ts": 1716454222328621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328949, "dur": 0,
    "args": {
      "External id": 71018,
      "cbid": 317, "correlation": 71018
    }
  },
  {
    "ph": "f", "id": 71018, "pid": 76337, "tid": -914061504, "ts": 1716454222328949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328950, "dur": 0,
    "args": {
      "External id": 71019,
      "cbid": 203, "correlation": 71019
    }
  },
  {
    "ph": "f", "id": 71019, "pid": 76337, "tid": -914061504, "ts": 1716454222328950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328951, "dur": 3,
    "args": {
      "External id": 71020,
      "cbid": 205, "correlation": 71020
    }
  },
  {
    "ph": "f", "id": 71020, "pid": 76337, "tid": -914061504, "ts": 1716454222328951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328984, "dur": 1,
    "args": {
      "External id": 71024,
      "cbid": 251, "correlation": 71024
    }
  },
  {
    "ph": "f", "id": 71024, "pid": 76337, "tid": -914061504, "ts": 1716454222328984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328985, "dur": 0,
    "args": {
      "External id": 71025,
      "cbid": 251, "correlation": 71025
    }
  },
  {
    "ph": "f", "id": 71025, "pid": 76337, "tid": -914061504, "ts": 1716454222328985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328986, "dur": 0,
    "args": {
      "External id": 71026,
      "cbid": 251, "correlation": 71026
    }
  },
  {
    "ph": "f", "id": 71026, "pid": 76337, "tid": -914061504, "ts": 1716454222328986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328987, "dur": 0,
    "args": {
      "External id": 71027,
      "cbid": 251, "correlation": 71027
    }
  },
  {
    "ph": "f", "id": 71027, "pid": 76337, "tid": -914061504, "ts": 1716454222328987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328988, "dur": 0,
    "args": {
      "External id": 71028,
      "cbid": 251, "correlation": 71028
    }
  },
  {
    "ph": "f", "id": 71028, "pid": 76337, "tid": -914061504, "ts": 1716454222328988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328989, "dur": 0,
    "args": {
      "External id": 71029,
      "cbid": 251, "correlation": 71029
    }
  },
  {
    "ph": "f", "id": 71029, "pid": 76337, "tid": -914061504, "ts": 1716454222328989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328990, "dur": 0,
    "args": {
      "External id": 71030,
      "cbid": 251, "correlation": 71030
    }
  },
  {
    "ph": "f", "id": 71030, "pid": 76337, "tid": -914061504, "ts": 1716454222328990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328990, "dur": 0,
    "args": {
      "External id": 71031,
      "cbid": 251, "correlation": 71031
    }
  },
  {
    "ph": "f", "id": 71031, "pid": 76337, "tid": -914061504, "ts": 1716454222328990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328992, "dur": 0,
    "args": {
      "External id": 71032,
      "cbid": 251, "correlation": 71032
    }
  },
  {
    "ph": "f", "id": 71032, "pid": 76337, "tid": -914061504, "ts": 1716454222328992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222467045, "dur": 116,
    "args": {
      "External id": 71033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71033,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 71033, "pid": 5, "tid": 7, "ts": 1716454222467045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222328994, "dur": 37,
    "args": {
      "External id": 71033,
      "cbid": 211, "correlation": 71033
    }
  },
  {
    "ph": "s", "id": 71033, "pid": 76337, "tid": -914061504, "ts": 1716454222328994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222467163, "dur": 60,
    "args": {
      "External id": 71039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71039, "pid": 5, "tid": 7, "ts": 1716454222467163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329055, "dur": 107,
    "args": {
      "External id": 71039,
      "cbid": 211, "correlation": 71039
    }
  },
  {
    "ph": "s", "id": 71039, "pid": 76337, "tid": -914061504, "ts": 1716454222329055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222467224, "dur": 50,
    "args": {
      "External id": 71047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71047,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71047, "pid": 5, "tid": 7, "ts": 1716454222467224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329185, "dur": 309,
    "args": {
      "External id": 71047,
      "cbid": 211, "correlation": 71047
    }
  },
  {
    "ph": "s", "id": 71047, "pid": 76337, "tid": -914061504, "ts": 1716454222329185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222467275, "dur": 99,
    "args": {
      "External id": 71056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71056,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71056, "pid": 5, "tid": 7, "ts": 1716454222467275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329524, "dur": 10,
    "args": {
      "External id": 71056,
      "cbid": 211, "correlation": 71056
    }
  },
  {
    "ph": "s", "id": 71056, "pid": 76337, "tid": -914061504, "ts": 1716454222329524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222467376, "dur": 93,
    "args": {
      "External id": 71076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71076,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 71076, "pid": 5, "tid": 7, "ts": 1716454222467376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329596, "dur": 19,
    "args": {
      "External id": 71076,
      "cbid": 211, "correlation": 71076
    }
  },
  {
    "ph": "s", "id": 71076, "pid": 76337, "tid": -914061504, "ts": 1716454222329596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222467470, "dur": 5,
    "args": {
      "External id": 71088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71088,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 71088, "pid": 5, "tid": 7, "ts": 1716454222467470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329625, "dur": 12,
    "args": {
      "External id": 71088,
      "cbid": 211, "correlation": 71088
    }
  },
  {
    "ph": "s", "id": 71088, "pid": 76337, "tid": -914061504, "ts": 1716454222329625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222467476, "dur": 110,
    "args": {
      "External id": 71091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71091, "pid": 5, "tid": 7, "ts": 1716454222467476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329649, "dur": 121,
    "args": {
      "External id": 71091,
      "cbid": 211, "correlation": 71091
    }
  },
  {
    "ph": "s", "id": 71091, "pid": 76337, "tid": -914061504, "ts": 1716454222329649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222467587, "dur": 70,
    "args": {
      "External id": 71100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71100,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71100, "pid": 5, "tid": 7, "ts": 1716454222467587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329807, "dur": 11,
    "args": {
      "External id": 71100,
      "cbid": 211, "correlation": 71100
    }
  },
  {
    "ph": "s", "id": 71100, "pid": 76337, "tid": -914061504, "ts": 1716454222329807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329861, "dur": 0,
    "args": {
      "External id": 71110,
      "cbid": 317, "correlation": 71110
    }
  },
  {
    "ph": "f", "id": 71110, "pid": 76337, "tid": -914061504, "ts": 1716454222329861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329861, "dur": 0,
    "args": {
      "External id": 71111,
      "cbid": 203, "correlation": 71111
    }
  },
  {
    "ph": "f", "id": 71111, "pid": 76337, "tid": -914061504, "ts": 1716454222329861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329862, "dur": 0,
    "args": {
      "External id": 71112,
      "cbid": 205, "correlation": 71112
    }
  },
  {
    "ph": "f", "id": 71112, "pid": 76337, "tid": -914061504, "ts": 1716454222329862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222467659, "dur": 77,
    "args": {
      "External id": 71116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71116, "pid": 5, "tid": 7, "ts": 1716454222467659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329876, "dur": 12,
    "args": {
      "External id": 71116,
      "cbid": 211, "correlation": 71116
    }
  },
  {
    "ph": "s", "id": 71116, "pid": 76337, "tid": -914061504, "ts": 1716454222329876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222467737, "dur": 25,
    "args": {
      "External id": 71118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71118, "pid": 5, "tid": 7, "ts": 1716454222467737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329891, "dur": 5,
    "args": {
      "External id": 71118,
      "cbid": 211, "correlation": 71118
    }
  },
  {
    "ph": "s", "id": 71118, "pid": 76337, "tid": -914061504, "ts": 1716454222329891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222467763, "dur": 4,
    "args": {
      "External id": 71120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71120, "pid": 5, "tid": 7, "ts": 1716454222467763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329901, "dur": 6,
    "args": {
      "External id": 71120,
      "cbid": 211, "correlation": 71120
    }
  },
  {
    "ph": "s", "id": 71120, "pid": 76337, "tid": -914061504, "ts": 1716454222329901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329909, "dur": 0,
    "args": {
      "External id": 71121,
      "cbid": 51, "correlation": 71121
    }
  },
  {
    "ph": "s", "id": 71121, "pid": 76337, "tid": -914061504, "ts": 1716454222329909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222467768, "dur": 1385,
    "args": {
      "External id": 71122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71122,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71122, "pid": 5, "tid": 7, "ts": 1716454222467768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329910, "dur": 5,
    "args": {
      "External id": 71122,
      "cbid": 211, "correlation": 71122
    }
  },
  {
    "ph": "s", "id": 71122, "pid": 76337, "tid": -914061504, "ts": 1716454222329910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222469154, "dur": 60,
    "args": {
      "External id": 71127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71127, "pid": 5, "tid": 7, "ts": 1716454222469154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329937, "dur": 10,
    "args": {
      "External id": 71127,
      "cbid": 211, "correlation": 71127
    }
  },
  {
    "ph": "s", "id": 71127, "pid": 76337, "tid": -914061504, "ts": 1716454222329937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222469215, "dur": 4,
    "args": {
      "External id": 71135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71135,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71135, "pid": 5, "tid": 7, "ts": 1716454222469215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222329989, "dur": 11,
    "args": {
      "External id": 71135,
      "cbid": 211, "correlation": 71135
    }
  },
  {
    "ph": "s", "id": 71135, "pid": 76337, "tid": -914061504, "ts": 1716454222329989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330057, "dur": 1,
    "args": {
      "External id": 71151,
      "cbid": 251, "correlation": 71151
    }
  },
  {
    "ph": "f", "id": 71151, "pid": 76337, "tid": -914061504, "ts": 1716454222330057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330063, "dur": 0,
    "args": {
      "External id": 71153,
      "cbid": 251, "correlation": 71153
    }
  },
  {
    "ph": "f", "id": 71153, "pid": 76337, "tid": -914061504, "ts": 1716454222330063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222469221, "dur": 11,
    "args": {
      "External id": 71154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71154,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 71154, "pid": 5, "tid": 7, "ts": 1716454222469221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330065, "dur": 11,
    "args": {
      "External id": 71154,
      "cbid": 211, "correlation": 71154
    }
  },
  {
    "ph": "s", "id": 71154, "pid": 76337, "tid": -914061504, "ts": 1716454222330065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222469233, "dur": 5,
    "args": {
      "External id": 71156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71156,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 71156, "pid": 5, "tid": 7, "ts": 1716454222469233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330078, "dur": 8,
    "args": {
      "External id": 71156,
      "cbid": 211, "correlation": 71156
    }
  },
  {
    "ph": "s", "id": 71156, "pid": 76337, "tid": -914061504, "ts": 1716454222330078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222469240, "dur": 55,
    "args": {
      "External id": 71166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71166, "pid": 5, "tid": 7, "ts": 1716454222469240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330139, "dur": 614,
    "args": {
      "External id": 71166,
      "cbid": 211, "correlation": 71166
    }
  },
  {
    "ph": "s", "id": 71166, "pid": 76337, "tid": -914061504, "ts": 1716454222330139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222469296, "dur": 54,
    "args": {
      "External id": 71186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71186,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 71186, "pid": 5, "tid": 7, "ts": 1716454222469296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330809, "dur": 11,
    "args": {
      "External id": 71186,
      "cbid": 211, "correlation": 71186
    }
  },
  {
    "ph": "s", "id": 71186, "pid": 76337, "tid": -914061504, "ts": 1716454222330809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222469351, "dur": 4,
    "args": {
      "External id": 71198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71198,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71198, "pid": 5, "tid": 7, "ts": 1716454222469351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330830, "dur": 6,
    "args": {
      "External id": 71198,
      "cbid": 211, "correlation": 71198
    }
  },
  {
    "ph": "s", "id": 71198, "pid": 76337, "tid": -914061504, "ts": 1716454222330830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222469356, "dur": 55,
    "args": {
      "External id": 71201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71201,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71201, "pid": 5, "tid": 7, "ts": 1716454222469356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330848, "dur": 7,
    "args": {
      "External id": 71201,
      "cbid": 211, "correlation": 71201
    }
  },
  {
    "ph": "s", "id": 71201, "pid": 76337, "tid": -914061504, "ts": 1716454222330848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222469413, "dur": 36,
    "args": {
      "External id": 71210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71210,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71210, "pid": 5, "tid": 7, "ts": 1716454222469413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330889, "dur": 10,
    "args": {
      "External id": 71210,
      "cbid": 211, "correlation": 71210
    }
  },
  {
    "ph": "s", "id": 71210, "pid": 76337, "tid": -914061504, "ts": 1716454222330889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330951, "dur": 0,
    "args": {
      "External id": 71220,
      "cbid": 317, "correlation": 71220
    }
  },
  {
    "ph": "f", "id": 71220, "pid": 76337, "tid": -914061504, "ts": 1716454222330951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330952, "dur": 0,
    "args": {
      "External id": 71221,
      "cbid": 203, "correlation": 71221
    }
  },
  {
    "ph": "f", "id": 71221, "pid": 76337, "tid": -914061504, "ts": 1716454222330952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330953, "dur": 0,
    "args": {
      "External id": 71222,
      "cbid": 205, "correlation": 71222
    }
  },
  {
    "ph": "f", "id": 71222, "pid": 76337, "tid": -914061504, "ts": 1716454222330953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222469451, "dur": 40,
    "args": {
      "External id": 71226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71226,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71226, "pid": 5, "tid": 7, "ts": 1716454222469451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330967, "dur": 20,
    "args": {
      "External id": 71226,
      "cbid": 211, "correlation": 71226
    }
  },
  {
    "ph": "s", "id": 71226, "pid": 76337, "tid": -914061504, "ts": 1716454222330967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222469492, "dur": 14,
    "args": {
      "External id": 71228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71228,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71228, "pid": 5, "tid": 7, "ts": 1716454222469492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222330989, "dur": 8,
    "args": {
      "External id": 71228,
      "cbid": 211, "correlation": 71228
    }
  },
  {
    "ph": "s", "id": 71228, "pid": 76337, "tid": -914061504, "ts": 1716454222330989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222469508, "dur": 3,
    "args": {
      "External id": 71230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71230, "pid": 5, "tid": 7, "ts": 1716454222469508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331001, "dur": 6,
    "args": {
      "External id": 71230,
      "cbid": 211, "correlation": 71230
    }
  },
  {
    "ph": "s", "id": 71230, "pid": 76337, "tid": -914061504, "ts": 1716454222331001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331010, "dur": 0,
    "args": {
      "External id": 71231,
      "cbid": 51, "correlation": 71231
    }
  },
  {
    "ph": "s", "id": 71231, "pid": 76337, "tid": -914061504, "ts": 1716454222331010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222469512, "dur": 708,
    "args": {
      "External id": 71232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71232,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71232, "pid": 5, "tid": 7, "ts": 1716454222469512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331011, "dur": 5,
    "args": {
      "External id": 71232,
      "cbid": 211, "correlation": 71232
    }
  },
  {
    "ph": "s", "id": 71232, "pid": 76337, "tid": -914061504, "ts": 1716454222331011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222470222, "dur": 60,
    "args": {
      "External id": 71237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71237, "pid": 5, "tid": 7, "ts": 1716454222470222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331039, "dur": 8,
    "args": {
      "External id": 71237,
      "cbid": 211, "correlation": 71237
    }
  },
  {
    "ph": "s", "id": 71237, "pid": 76337, "tid": -914061504, "ts": 1716454222331039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331098, "dur": 0,
    "args": {
      "External id": 71247,
      "cbid": 317, "correlation": 71247
    }
  },
  {
    "ph": "f", "id": 71247, "pid": 76337, "tid": -914061504, "ts": 1716454222331098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331099, "dur": 0,
    "args": {
      "External id": 71248,
      "cbid": 203, "correlation": 71248
    }
  },
  {
    "ph": "f", "id": 71248, "pid": 76337, "tid": -914061504, "ts": 1716454222331099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331100, "dur": 0,
    "args": {
      "External id": 71249,
      "cbid": 205, "correlation": 71249
    }
  },
  {
    "ph": "f", "id": 71249, "pid": 76337, "tid": -914061504, "ts": 1716454222331100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222470282, "dur": 76,
    "args": {
      "External id": 71253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71253,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71253, "pid": 5, "tid": 7, "ts": 1716454222470282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331112, "dur": 12,
    "args": {
      "External id": 71253,
      "cbid": 211, "correlation": 71253
    }
  },
  {
    "ph": "s", "id": 71253, "pid": 76337, "tid": -914061504, "ts": 1716454222331112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222470360, "dur": 211,
    "args": {
      "External id": 71255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71255,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71255, "pid": 5, "tid": 7, "ts": 1716454222470360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331130, "dur": 7,
    "args": {
      "External id": 71255,
      "cbid": 211, "correlation": 71255
    }
  },
  {
    "ph": "s", "id": 71255, "pid": 76337, "tid": -914061504, "ts": 1716454222331130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222470572, "dur": 39,
    "args": {
      "External id": 71257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71257,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71257, "pid": 5, "tid": 7, "ts": 1716454222470572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331141, "dur": 6,
    "args": {
      "External id": 71257,
      "cbid": 211, "correlation": 71257
    }
  },
  {
    "ph": "s", "id": 71257, "pid": 76337, "tid": -914061504, "ts": 1716454222331141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222470613, "dur": 60,
    "args": {
      "External id": 71263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71263, "pid": 5, "tid": 7, "ts": 1716454222470613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331168, "dur": 582,
    "args": {
      "External id": 71263,
      "cbid": 211, "correlation": 71263
    }
  },
  {
    "ph": "s", "id": 71263, "pid": 76337, "tid": -914061504, "ts": 1716454222331168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222470674, "dur": 50,
    "args": {
      "External id": 71271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71271,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71271, "pid": 5, "tid": 7, "ts": 1716454222470674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331771, "dur": 8,
    "args": {
      "External id": 71271,
      "cbid": 211, "correlation": 71271
    }
  },
  {
    "ph": "s", "id": 71271, "pid": 76337, "tid": -914061504, "ts": 1716454222331771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222470725, "dur": 35,
    "args": {
      "External id": 71279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71279,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71279, "pid": 5, "tid": 7, "ts": 1716454222470725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331799, "dur": 11,
    "args": {
      "External id": 71279,
      "cbid": 211, "correlation": 71279
    }
  },
  {
    "ph": "s", "id": 71279, "pid": 76337, "tid": -914061504, "ts": 1716454222331799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222470761, "dur": 53,
    "args": {
      "External id": 71299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71299,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 71299, "pid": 5, "tid": 7, "ts": 1716454222470761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331888, "dur": 13,
    "args": {
      "External id": 71299,
      "cbid": 211, "correlation": 71299
    }
  },
  {
    "ph": "s", "id": 71299, "pid": 76337, "tid": -914061504, "ts": 1716454222331888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222470816, "dur": 4,
    "args": {
      "External id": 71311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71311,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71311, "pid": 5, "tid": 7, "ts": 1716454222470816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331910, "dur": 6,
    "args": {
      "External id": 71311,
      "cbid": 211, "correlation": 71311
    }
  },
  {
    "ph": "s", "id": 71311, "pid": 76337, "tid": -914061504, "ts": 1716454222331910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222470822, "dur": 55,
    "args": {
      "External id": 71314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71314, "pid": 5, "tid": 7, "ts": 1716454222470822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331928, "dur": 6,
    "args": {
      "External id": 71314,
      "cbid": 211, "correlation": 71314
    }
  },
  {
    "ph": "s", "id": 71314, "pid": 76337, "tid": -914061504, "ts": 1716454222331928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331994, "dur": 0,
    "args": {
      "External id": 71325,
      "cbid": 317, "correlation": 71325
    }
  },
  {
    "ph": "f", "id": 71325, "pid": 76337, "tid": -914061504, "ts": 1716454222331994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331995, "dur": 0,
    "args": {
      "External id": 71326,
      "cbid": 203, "correlation": 71326
    }
  },
  {
    "ph": "f", "id": 71326, "pid": 76337, "tid": -914061504, "ts": 1716454222331995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222331996, "dur": 0,
    "args": {
      "External id": 71327,
      "cbid": 205, "correlation": 71327
    }
  },
  {
    "ph": "f", "id": 71327, "pid": 76337, "tid": -914061504, "ts": 1716454222331996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332018, "dur": 1,
    "args": {
      "External id": 71331,
      "cbid": 251, "correlation": 71331
    }
  },
  {
    "ph": "f", "id": 71331, "pid": 76337, "tid": -914061504, "ts": 1716454222332018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332020, "dur": 0,
    "args": {
      "External id": 71332,
      "cbid": 251, "correlation": 71332
    }
  },
  {
    "ph": "f", "id": 71332, "pid": 76337, "tid": -914061504, "ts": 1716454222332020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332021, "dur": 0,
    "args": {
      "External id": 71333,
      "cbid": 251, "correlation": 71333
    }
  },
  {
    "ph": "f", "id": 71333, "pid": 76337, "tid": -914061504, "ts": 1716454222332021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332021, "dur": 0,
    "args": {
      "External id": 71334,
      "cbid": 251, "correlation": 71334
    }
  },
  {
    "ph": "f", "id": 71334, "pid": 76337, "tid": -914061504, "ts": 1716454222332021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332022, "dur": 0,
    "args": {
      "External id": 71335,
      "cbid": 251, "correlation": 71335
    }
  },
  {
    "ph": "f", "id": 71335, "pid": 76337, "tid": -914061504, "ts": 1716454222332022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332023, "dur": 0,
    "args": {
      "External id": 71336,
      "cbid": 251, "correlation": 71336
    }
  },
  {
    "ph": "f", "id": 71336, "pid": 76337, "tid": -914061504, "ts": 1716454222332023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332024, "dur": 0,
    "args": {
      "External id": 71337,
      "cbid": 251, "correlation": 71337
    }
  },
  {
    "ph": "f", "id": 71337, "pid": 76337, "tid": -914061504, "ts": 1716454222332024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332024, "dur": 0,
    "args": {
      "External id": 71338,
      "cbid": 251, "correlation": 71338
    }
  },
  {
    "ph": "f", "id": 71338, "pid": 76337, "tid": -914061504, "ts": 1716454222332024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332026, "dur": 0,
    "args": {
      "External id": 71339,
      "cbid": 251, "correlation": 71339
    }
  },
  {
    "ph": "f", "id": 71339, "pid": 76337, "tid": -914061504, "ts": 1716454222332026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222470879, "dur": 116,
    "args": {
      "External id": 71340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71340,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 71340, "pid": 5, "tid": 7, "ts": 1716454222470879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332028, "dur": 13,
    "args": {
      "External id": 71340,
      "cbid": 211, "correlation": 71340
    }
  },
  {
    "ph": "s", "id": 71340, "pid": 76337, "tid": -914061504, "ts": 1716454222332028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222470996, "dur": 60,
    "args": {
      "External id": 71346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71346, "pid": 5, "tid": 7, "ts": 1716454222470996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332064, "dur": 9,
    "args": {
      "External id": 71346,
      "cbid": 211, "correlation": 71346
    }
  },
  {
    "ph": "s", "id": 71346, "pid": 76337, "tid": -914061504, "ts": 1716454222332064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222471057, "dur": 611,
    "args": {
      "External id": 71355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71355,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71355, "pid": 5, "tid": 7, "ts": 1716454222471057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332147, "dur": 14,
    "args": {
      "External id": 71355,
      "cbid": 211, "correlation": 71355
    }
  },
  {
    "ph": "s", "id": 71355, "pid": 76337, "tid": -914061504, "ts": 1716454222332147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222471669, "dur": 184,
    "args": {
      "External id": 71377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71377,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71377, "pid": 5, "tid": 7, "ts": 1716454222471669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332208, "dur": 11,
    "args": {
      "External id": 71377,
      "cbid": 211, "correlation": 71377
    }
  },
  {
    "ph": "s", "id": 71377, "pid": 76337, "tid": -914061504, "ts": 1716454222332208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332294, "dur": 1,
    "args": {
      "External id": 71388,
      "cbid": 251, "correlation": 71388
    }
  },
  {
    "ph": "f", "id": 71388, "pid": 76337, "tid": -914061504, "ts": 1716454222332294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222471855, "dur": 196,
    "args": {
      "External id": 71389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71389,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71389, "pid": 5, "tid": 7, "ts": 1716454222471855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332299, "dur": 13,
    "args": {
      "External id": 71389,
      "cbid": 211, "correlation": 71389
    }
  },
  {
    "ph": "s", "id": 71389, "pid": 76337, "tid": -914061504, "ts": 1716454222332299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332365, "dur": 1,
    "args": {
      "External id": 71400,
      "cbid": 251, "correlation": 71400
    }
  },
  {
    "ph": "f", "id": 71400, "pid": 76337, "tid": -914061504, "ts": 1716454222332365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222472052, "dur": 194,
    "args": {
      "External id": 71401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71401,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71401, "pid": 5, "tid": 7, "ts": 1716454222472052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332369, "dur": 12,
    "args": {
      "External id": 71401,
      "cbid": 211, "correlation": 71401
    }
  },
  {
    "ph": "s", "id": 71401, "pid": 76337, "tid": -914061504, "ts": 1716454222332369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332431, "dur": 1,
    "args": {
      "External id": 71412,
      "cbid": 251, "correlation": 71412
    }
  },
  {
    "ph": "f", "id": 71412, "pid": 76337, "tid": -914061504, "ts": 1716454222332431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222472247, "dur": 190,
    "args": {
      "External id": 71413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71413,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71413, "pid": 5, "tid": 7, "ts": 1716454222472247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332435, "dur": 12,
    "args": {
      "External id": 71413,
      "cbid": 211, "correlation": 71413
    }
  },
  {
    "ph": "s", "id": 71413, "pid": 76337, "tid": -914061504, "ts": 1716454222332435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222472438, "dur": 18948,
    "args": {
      "External id": 71434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71434,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 71434, "pid": 5, "tid": 7, "ts": 1716454222472438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332514, "dur": 12,
    "args": {
      "External id": 71434,
      "cbid": 211, "correlation": 71434
    }
  },
  {
    "ph": "s", "id": 71434, "pid": 76337, "tid": -914061504, "ts": 1716454222332514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332613, "dur": 1,
    "args": {
      "External id": 71452,
      "cbid": 251, "correlation": 71452
    }
  },
  {
    "ph": "f", "id": 71452, "pid": 76337, "tid": -914061504, "ts": 1716454222332613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222491387, "dur": 204,
    "args": {
      "External id": 71454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71454,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71454, "pid": 5, "tid": 7, "ts": 1716454222491387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332619, "dur": 24,
    "args": {
      "External id": 71454,
      "cbid": 211, "correlation": 71454
    }
  },
  {
    "ph": "s", "id": 71454, "pid": 76337, "tid": -914061504, "ts": 1716454222332619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222491593, "dur": 66,
    "args": {
      "External id": 71462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71462,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71462, "pid": 5, "tid": 7, "ts": 1716454222491593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332703, "dur": 42,
    "args": {
      "External id": 71462,
      "cbid": 211, "correlation": 71462
    }
  },
  {
    "ph": "s", "id": 71462, "pid": 76337, "tid": -914061504, "ts": 1716454222332703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222491660, "dur": 97,
    "args": {
      "External id": 71470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71470,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71470, "pid": 5, "tid": 7, "ts": 1716454222491660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332773, "dur": 129,
    "args": {
      "External id": 71470,
      "cbid": 211, "correlation": 71470
    }
  },
  {
    "ph": "s", "id": 71470, "pid": 76337, "tid": -914061504, "ts": 1716454222332773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222491759, "dur": 55,
    "args": {
      "External id": 71481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71481,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71481, "pid": 5, "tid": 7, "ts": 1716454222491759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222332965, "dur": 2086,
    "args": {
      "External id": 71481,
      "cbid": 211, "correlation": 71481
    }
  },
  {
    "ph": "s", "id": 71481, "pid": 76337, "tid": -914061504, "ts": 1716454222332965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222491815, "dur": 94,
    "args": {
      "External id": 71503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71503,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71503, "pid": 5, "tid": 7, "ts": 1716454222491815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335070, "dur": 138,
    "args": {
      "External id": 71503,
      "cbid": 211, "correlation": 71503
    }
  },
  {
    "ph": "s", "id": 71503, "pid": 76337, "tid": -914061504, "ts": 1716454222335070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335285, "dur": 1,
    "args": {
      "External id": 71514,
      "cbid": 251, "correlation": 71514
    }
  },
  {
    "ph": "f", "id": 71514, "pid": 76337, "tid": -914061504, "ts": 1716454222335285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222491911, "dur": 105,
    "args": {
      "External id": 71515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71515,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71515, "pid": 5, "tid": 7, "ts": 1716454222491911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335290, "dur": 14,
    "args": {
      "External id": 71515,
      "cbid": 211, "correlation": 71515
    }
  },
  {
    "ph": "s", "id": 71515, "pid": 76337, "tid": -914061504, "ts": 1716454222335290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335365, "dur": 1,
    "args": {
      "External id": 71526,
      "cbid": 251, "correlation": 71526
    }
  },
  {
    "ph": "f", "id": 71526, "pid": 76337, "tid": -914061504, "ts": 1716454222335365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335368, "dur": 0,
    "args": {
      "External id": 71527,
      "cbid": 251, "correlation": 71527
    }
  },
  {
    "ph": "f", "id": 71527, "pid": 76337, "tid": -914061504, "ts": 1716454222335368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222492017, "dur": 10,
    "args": {
      "External id": 71528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71528,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71528, "pid": 5, "tid": 7, "ts": 1716454222492017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335370, "dur": 12,
    "args": {
      "External id": 71528,
      "cbid": 211, "correlation": 71528
    }
  },
  {
    "ph": "s", "id": 71528, "pid": 76337, "tid": -914061504, "ts": 1716454222335370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222492029, "dur": 5,
    "args": {
      "External id": 71530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71530,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 71530, "pid": 5, "tid": 7, "ts": 1716454222492029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335384, "dur": 6,
    "args": {
      "External id": 71530,
      "cbid": 211, "correlation": 71530
    }
  },
  {
    "ph": "s", "id": 71530, "pid": 76337, "tid": -914061504, "ts": 1716454222335384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335446, "dur": 1,
    "args": {
      "External id": 71541,
      "cbid": 251, "correlation": 71541
    }
  },
  {
    "ph": "f", "id": 71541, "pid": 76337, "tid": -914061504, "ts": 1716454222335446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335449, "dur": 0,
    "args": {
      "External id": 71542,
      "cbid": 251, "correlation": 71542
    }
  },
  {
    "ph": "f", "id": 71542, "pid": 76337, "tid": -914061504, "ts": 1716454222335449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222492035, "dur": 6,
    "args": {
      "External id": 71543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71543,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71543, "pid": 5, "tid": 7, "ts": 1716454222492035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335451, "dur": 12,
    "args": {
      "External id": 71543,
      "cbid": 211, "correlation": 71543
    }
  },
  {
    "ph": "s", "id": 71543, "pid": 76337, "tid": -914061504, "ts": 1716454222335451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222492043, "dur": 4,
    "args": {
      "External id": 71545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71545,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 71545, "pid": 5, "tid": 7, "ts": 1716454222492043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335465, "dur": 6,
    "args": {
      "External id": 71545,
      "cbid": 211, "correlation": 71545
    }
  },
  {
    "ph": "s", "id": 71545, "pid": 76337, "tid": -914061504, "ts": 1716454222335465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222492048, "dur": 156,
    "args": {
      "External id": 71566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71566,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 71566, "pid": 5, "tid": 7, "ts": 1716454222492048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335538, "dur": 13,
    "args": {
      "External id": 71566,
      "cbid": 211, "correlation": 71566
    }
  },
  {
    "ph": "s", "id": 71566, "pid": 76337, "tid": -914061504, "ts": 1716454222335538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335634, "dur": 1,
    "args": {
      "External id": 71584,
      "cbid": 251, "correlation": 71584
    }
  },
  {
    "ph": "f", "id": 71584, "pid": 76337, "tid": -914061504, "ts": 1716454222335634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222492205, "dur": 108,
    "args": {
      "External id": 71586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71586,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 71586, "pid": 5, "tid": 7, "ts": 1716454222492205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335641, "dur": 13,
    "args": {
      "External id": 71586,
      "cbid": 211, "correlation": 71586
    }
  },
  {
    "ph": "s", "id": 71586, "pid": 76337, "tid": -914061504, "ts": 1716454222335641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222492314, "dur": 35,
    "args": {
      "External id": 71594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71594,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71594, "pid": 5, "tid": 7, "ts": 1716454222492314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335709, "dur": 15,
    "args": {
      "External id": 71594,
      "cbid": 211, "correlation": 71594
    }
  },
  {
    "ph": "s", "id": 71594, "pid": 76337, "tid": -914061504, "ts": 1716454222335709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222492350, "dur": 67,
    "args": {
      "External id": 71602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71602, "pid": 5, "tid": 7, "ts": 1716454222492350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335754, "dur": 10,
    "args": {
      "External id": 71602,
      "cbid": 211, "correlation": 71602
    }
  },
  {
    "ph": "s", "id": 71602, "pid": 76337, "tid": -914061504, "ts": 1716454222335754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222492418, "dur": 94,
    "args": {
      "External id": 71624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71624,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71624, "pid": 5, "tid": 7, "ts": 1716454222492418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335806, "dur": 10,
    "args": {
      "External id": 71624,
      "cbid": 211, "correlation": 71624
    }
  },
  {
    "ph": "s", "id": 71624, "pid": 76337, "tid": -914061504, "ts": 1716454222335806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335894, "dur": 1,
    "args": {
      "External id": 71640,
      "cbid": 251, "correlation": 71640
    }
  },
  {
    "ph": "f", "id": 71640, "pid": 76337, "tid": -914061504, "ts": 1716454222335894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222492513, "dur": 586,
    "args": {
      "External id": 71642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71642,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 71642, "pid": 5, "tid": 7, "ts": 1716454222492513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335900, "dur": 13,
    "args": {
      "External id": 71642,
      "cbid": 211, "correlation": 71642
    }
  },
  {
    "ph": "s", "id": 71642, "pid": 76337, "tid": -914061504, "ts": 1716454222335900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222493101, "dur": 244,
    "args": {
      "External id": 71650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71650, "pid": 5, "tid": 7, "ts": 1716454222493101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222335965, "dur": 21,
    "args": {
      "External id": 71650,
      "cbid": 211, "correlation": 71650
    }
  },
  {
    "ph": "s", "id": 71650, "pid": 76337, "tid": -914061504, "ts": 1716454222335965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222493346, "dur": 253,
    "args": {
      "External id": 71658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71658,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71658, "pid": 5, "tid": 7, "ts": 1716454222493346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336005, "dur": 9,
    "args": {
      "External id": 71658,
      "cbid": 211, "correlation": 71658
    }
  },
  {
    "ph": "s", "id": 71658, "pid": 76337, "tid": -914061504, "ts": 1716454222336005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336089, "dur": 4,
    "args": {
      "External id": 71674,
      "cbid": 251, "correlation": 71674
    }
  },
  {
    "ph": "f", "id": 71674, "pid": 76337, "tid": -914061504, "ts": 1716454222336089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336097, "dur": 0,
    "args": {
      "External id": 71676,
      "cbid": 251, "correlation": 71676
    }
  },
  {
    "ph": "f", "id": 71676, "pid": 76337, "tid": -914061504, "ts": 1716454222336097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222493601, "dur": 364,
    "args": {
      "External id": 71677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71677,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 71677, "pid": 5, "tid": 7, "ts": 1716454222493601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336100, "dur": 13,
    "args": {
      "External id": 71677,
      "cbid": 211, "correlation": 71677
    }
  },
  {
    "ph": "s", "id": 71677, "pid": 76337, "tid": -914061504, "ts": 1716454222336100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222493966, "dur": 50,
    "args": {
      "External id": 71685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71685,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71685, "pid": 5, "tid": 7, "ts": 1716454222493966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336143, "dur": 257,
    "args": {
      "External id": 71685,
      "cbid": 211, "correlation": 71685
    }
  },
  {
    "ph": "s", "id": 71685, "pid": 76337, "tid": -914061504, "ts": 1716454222336143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222494017, "dur": 160,
    "args": {
      "External id": 71696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71696, "pid": 5, "tid": 7, "ts": 1716454222494017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336458, "dur": 77,
    "args": {
      "External id": 71696,
      "cbid": 211, "correlation": 71696
    }
  },
  {
    "ph": "s", "id": 71696, "pid": 76337, "tid": -914061504, "ts": 1716454222336458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336589, "dur": 0,
    "args": {
      "External id": 71708,
      "cbid": 317, "correlation": 71708
    }
  },
  {
    "ph": "f", "id": 71708, "pid": 76337, "tid": -914061504, "ts": 1716454222336589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336590, "dur": 0,
    "args": {
      "External id": 71709,
      "cbid": 203, "correlation": 71709
    }
  },
  {
    "ph": "f", "id": 71709, "pid": 76337, "tid": -914061504, "ts": 1716454222336590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336591, "dur": 0,
    "args": {
      "External id": 71710,
      "cbid": 205, "correlation": 71710
    }
  },
  {
    "ph": "f", "id": 71710, "pid": 76337, "tid": -914061504, "ts": 1716454222336591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336626, "dur": 1,
    "args": {
      "External id": 71714,
      "cbid": 251, "correlation": 71714
    }
  },
  {
    "ph": "f", "id": 71714, "pid": 76337, "tid": -914061504, "ts": 1716454222336626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336628, "dur": 0,
    "args": {
      "External id": 71715,
      "cbid": 251, "correlation": 71715
    }
  },
  {
    "ph": "f", "id": 71715, "pid": 76337, "tid": -914061504, "ts": 1716454222336628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336630, "dur": 0,
    "args": {
      "External id": 71716,
      "cbid": 251, "correlation": 71716
    }
  },
  {
    "ph": "f", "id": 71716, "pid": 76337, "tid": -914061504, "ts": 1716454222336630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336631, "dur": 0,
    "args": {
      "External id": 71717,
      "cbid": 251, "correlation": 71717
    }
  },
  {
    "ph": "f", "id": 71717, "pid": 76337, "tid": -914061504, "ts": 1716454222336631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336632, "dur": 0,
    "args": {
      "External id": 71718,
      "cbid": 251, "correlation": 71718
    }
  },
  {
    "ph": "f", "id": 71718, "pid": 76337, "tid": -914061504, "ts": 1716454222336632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336632, "dur": 0,
    "args": {
      "External id": 71719,
      "cbid": 251, "correlation": 71719
    }
  },
  {
    "ph": "f", "id": 71719, "pid": 76337, "tid": -914061504, "ts": 1716454222336632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336633, "dur": 0,
    "args": {
      "External id": 71720,
      "cbid": 251, "correlation": 71720
    }
  },
  {
    "ph": "f", "id": 71720, "pid": 76337, "tid": -914061504, "ts": 1716454222336633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336634, "dur": 0,
    "args": {
      "External id": 71721,
      "cbid": 251, "correlation": 71721
    }
  },
  {
    "ph": "f", "id": 71721, "pid": 76337, "tid": -914061504, "ts": 1716454222336634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336635, "dur": 0,
    "args": {
      "External id": 71722,
      "cbid": 251, "correlation": 71722
    }
  },
  {
    "ph": "f", "id": 71722, "pid": 76337, "tid": -914061504, "ts": 1716454222336635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222494179, "dur": 116,
    "args": {
      "External id": 71723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71723,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 71723, "pid": 5, "tid": 7, "ts": 1716454222494179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336637, "dur": 34,
    "args": {
      "External id": 71723,
      "cbid": 211, "correlation": 71723
    }
  },
  {
    "ph": "s", "id": 71723, "pid": 76337, "tid": -914061504, "ts": 1716454222336637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222494296, "dur": 61,
    "args": {
      "External id": 71729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71729,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71729, "pid": 5, "tid": 7, "ts": 1716454222494296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222336696, "dur": 307,
    "args": {
      "External id": 71729,
      "cbid": 211, "correlation": 71729
    }
  },
  {
    "ph": "s", "id": 71729, "pid": 76337, "tid": -914061504, "ts": 1716454222336696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222494358, "dur": 50,
    "args": {
      "External id": 71737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71737,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71737, "pid": 5, "tid": 7, "ts": 1716454222494358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337029, "dur": 9,
    "args": {
      "External id": 71737,
      "cbid": 211, "correlation": 71737
    }
  },
  {
    "ph": "s", "id": 71737, "pid": 76337, "tid": -914061504, "ts": 1716454222337029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222494409, "dur": 52,
    "args": {
      "External id": 71757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71757,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 71757, "pid": 5, "tid": 7, "ts": 1716454222494409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337111, "dur": 12,
    "args": {
      "External id": 71757,
      "cbid": 211, "correlation": 71757
    }
  },
  {
    "ph": "s", "id": 71757, "pid": 76337, "tid": -914061504, "ts": 1716454222337111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222494463, "dur": 5,
    "args": {
      "External id": 71769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71769,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71769, "pid": 5, "tid": 7, "ts": 1716454222494463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337134, "dur": 13,
    "args": {
      "External id": 71769,
      "cbid": 211, "correlation": 71769
    }
  },
  {
    "ph": "s", "id": 71769, "pid": 76337, "tid": -914061504, "ts": 1716454222337134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222494469, "dur": 57,
    "args": {
      "External id": 71772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71772, "pid": 5, "tid": 7, "ts": 1716454222494469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337159, "dur": 120,
    "args": {
      "External id": 71772,
      "cbid": 211, "correlation": 71772
    }
  },
  {
    "ph": "s", "id": 71772, "pid": 76337, "tid": -914061504, "ts": 1716454222337159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222494527, "dur": 38,
    "args": {
      "External id": 71781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71781,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71781, "pid": 5, "tid": 7, "ts": 1716454222494527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337319, "dur": 10,
    "args": {
      "External id": 71781,
      "cbid": 211, "correlation": 71781
    }
  },
  {
    "ph": "s", "id": 71781, "pid": 76337, "tid": -914061504, "ts": 1716454222337319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337374, "dur": 0,
    "args": {
      "External id": 71791,
      "cbid": 317, "correlation": 71791
    }
  },
  {
    "ph": "f", "id": 71791, "pid": 76337, "tid": -914061504, "ts": 1716454222337374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337375, "dur": 0,
    "args": {
      "External id": 71792,
      "cbid": 203, "correlation": 71792
    }
  },
  {
    "ph": "f", "id": 71792, "pid": 76337, "tid": -914061504, "ts": 1716454222337375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337376, "dur": 0,
    "args": {
      "External id": 71793,
      "cbid": 205, "correlation": 71793
    }
  },
  {
    "ph": "f", "id": 71793, "pid": 76337, "tid": -914061504, "ts": 1716454222337376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222494567, "dur": 42,
    "args": {
      "External id": 71797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71797,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71797, "pid": 5, "tid": 7, "ts": 1716454222494567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337392, "dur": 12,
    "args": {
      "External id": 71797,
      "cbid": 211, "correlation": 71797
    }
  },
  {
    "ph": "s", "id": 71797, "pid": 76337, "tid": -914061504, "ts": 1716454222337392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222494610, "dur": 3,
    "args": {
      "External id": 71799,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71799,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.203389830508474
    }
  },
  {
    "ph": "f", "id": 71799, "pid": 5, "tid": 7, "ts": 1716454222494610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337407, "dur": 18,
    "args": {
      "External id": 71799,
      "cbid": 51, "correlation": 71799
    }
  },
  {
    "ph": "s", "id": 71799, "pid": 76337, "tid": -914061504, "ts": 1716454222337407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337431, "dur": 1,
    "args": {
      "External id": 71801,
      "cbid": 200, "correlation": 71801
    }
  },
  {
    "ph": "f", "id": 71801, "pid": 76337, "tid": -914061504, "ts": 1716454222337431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337433, "dur": 0,
    "args": {
      "External id": 71802,
      "cbid": 200, "correlation": 71802
    }
  },
  {
    "ph": "f", "id": 71802, "pid": 76337, "tid": -914061504, "ts": 1716454222337433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337434, "dur": 0,
    "args": {
      "External id": 71803,
      "cbid": 200, "correlation": 71803
    }
  },
  {
    "ph": "f", "id": 71803, "pid": 76337, "tid": -914061504, "ts": 1716454222337434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337434, "dur": 0,
    "args": {
      "External id": 71804,
      "cbid": 200, "correlation": 71804
    }
  },
  {
    "ph": "f", "id": 71804, "pid": 76337, "tid": -914061504, "ts": 1716454222337434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337435, "dur": 4,
    "args": {
      "External id": 71805,
      "cbid": 15, "correlation": 71805
    }
  },
  {
    "ph": "f", "id": 71805, "pid": 76337, "tid": -914061504, "ts": 1716454222337435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337440, "dur": 1,
    "args": {
      "External id": 71806,
      "cbid": 251, "correlation": 71806
    }
  },
  {
    "ph": "f", "id": 71806, "pid": 76337, "tid": -914061504, "ts": 1716454222337440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222494615, "dur": 23,
    "args": {
      "External id": 71807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71807,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71807, "pid": 5, "tid": 7, "ts": 1716454222494615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337443, "dur": 9,
    "args": {
      "External id": 71807,
      "cbid": 211, "correlation": 71807
    }
  },
  {
    "ph": "s", "id": 71807, "pid": 76337, "tid": -914061504, "ts": 1716454222337443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222494639, "dur": 4,
    "args": {
      "External id": 71809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 71809, "pid": 5, "tid": 7, "ts": 1716454222494639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337458, "dur": 7,
    "args": {
      "External id": 71809,
      "cbid": 211, "correlation": 71809
    }
  },
  {
    "ph": "s", "id": 71809, "pid": 76337, "tid": -914061504, "ts": 1716454222337458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337469, "dur": 0,
    "args": {
      "External id": 71810,
      "cbid": 51, "correlation": 71810
    }
  },
  {
    "ph": "s", "id": 71810, "pid": 76337, "tid": -914061504, "ts": 1716454222337469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222494645, "dur": 193,
    "args": {
      "External id": 71811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71811,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71811, "pid": 5, "tid": 7, "ts": 1716454222494645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337470, "dur": 225,
    "args": {
      "External id": 71811,
      "cbid": 211, "correlation": 71811
    }
  },
  {
    "ph": "s", "id": 71811, "pid": 76337, "tid": -914061504, "ts": 1716454222337470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222494839, "dur": 6,
    "args": {
      "External id": 71812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71812, "pid": 5, "tid": 7, "ts": 1716454222494839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337699, "dur": 9,
    "args": {
      "External id": 71812,
      "cbid": 211, "correlation": 71812
    }
  },
  {
    "ph": "s", "id": 71812, "pid": 76337, "tid": -914061504, "ts": 1716454222337699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222494847, "dur": 5,
    "args": {
      "External id": 71818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 71818, "pid": 5, "tid": 7, "ts": 1716454222494847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222337732, "dur": 10,
    "args": {
      "External id": 71818,
      "cbid": 211, "correlation": 71818
    }
  },
  {
    "ph": "s", "id": 71818, "pid": 76337, "tid": -914061504, "ts": 1716454222337732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222494853, "dur": 3,
    "args": {
      "External id": 71826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71826,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71826, "pid": 5, "tid": 7, "ts": 1716454222494853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222339486, "dur": 16,
    "args": {
      "External id": 71826,
      "cbid": 211, "correlation": 71826
    }
  },
  {
    "ph": "s", "id": 71826, "pid": 76337, "tid": -914061504, "ts": 1716454222339486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222494857, "dur": 3,
    "args": {
      "External id": 71834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71834,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71834, "pid": 5, "tid": 7, "ts": 1716454222494857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222339527, "dur": 11,
    "args": {
      "External id": 71834,
      "cbid": 211, "correlation": 71834
    }
  },
  {
    "ph": "s", "id": 71834, "pid": 76337, "tid": -914061504, "ts": 1716454222339527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222494861, "dur": 3,
    "args": {
      "External id": 71842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71842,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71842, "pid": 5, "tid": 7, "ts": 1716454222494861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222339554, "dur": 8,
    "args": {
      "External id": 71842,
      "cbid": 211, "correlation": 71842
    }
  },
  {
    "ph": "s", "id": 71842, "pid": 76337, "tid": -914061504, "ts": 1716454222339554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222494865, "dur": 3,
    "args": {
      "External id": 71851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71851,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71851, "pid": 5, "tid": 7, "ts": 1716454222494865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222339724, "dur": 14,
    "args": {
      "External id": 71851,
      "cbid": 211, "correlation": 71851
    }
  },
  {
    "ph": "s", "id": 71851, "pid": 76337, "tid": -914061504, "ts": 1716454222339724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222494869, "dur": 3,
    "args": {
      "External id": 71860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71860,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71860, "pid": 5, "tid": 7, "ts": 1716454222494869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222339754, "dur": 7,
    "args": {
      "External id": 71860,
      "cbid": 211, "correlation": 71860
    }
  },
  {
    "ph": "s", "id": 71860, "pid": 76337, "tid": -914061504, "ts": 1716454222339754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222494874, "dur": 3,
    "args": {
      "External id": 71868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71868,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71868, "pid": 5, "tid": 7, "ts": 1716454222494874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222339777, "dur": 8,
    "args": {
      "External id": 71868,
      "cbid": 211, "correlation": 71868
    }
  },
  {
    "ph": "s", "id": 71868, "pid": 76337, "tid": -914061504, "ts": 1716454222339777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222494879, "dur": 3,
    "args": {
      "External id": 71876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71876,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71876, "pid": 5, "tid": 7, "ts": 1716454222494879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222340044, "dur": 16,
    "args": {
      "External id": 71876,
      "cbid": 211, "correlation": 71876
    }
  },
  {
    "ph": "s", "id": 71876, "pid": 76337, "tid": -914061504, "ts": 1716454222340044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222494883, "dur": 3,
    "args": {
      "External id": 71884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71884,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 71884, "pid": 5, "tid": 7, "ts": 1716454222494883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222340079, "dur": 8,
    "args": {
      "External id": 71884,
      "cbid": 211, "correlation": 71884
    }
  },
  {
    "ph": "s", "id": 71884, "pid": 76337, "tid": -914061504, "ts": 1716454222340079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222494888, "dur": 1,
    "args": {
      "External id": 71894,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71894,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 71894, "pid": 5, "tid": 7, "ts": 1716454222494888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222340144, "dur": 54,
    "args": {
      "External id": 71894,
      "cbid": 41, "correlation": 71894
    }
  },
  {
    "ph": "s", "id": 71894, "pid": 76337, "tid": -914061504, "ts": 1716454222340144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222340200, "dur": 154704,
    "args": {
      "External id": 71895,
      "cbid": 131, "correlation": 71895
    }
  },
  {
    "ph": "f", "id": 71895, "pid": 76337, "tid": -914061504, "ts": 1716454222340200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222495065, "dur": 3,
    "args": {
      "External id": 71903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71903,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 71903, "pid": 5, "tid": 7, "ts": 1716454222495065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495043, "dur": 24,
    "args": {
      "External id": 71903,
      "cbid": 211, "correlation": 71903
    }
  },
  {
    "ph": "s", "id": 71903, "pid": 76337, "tid": -914061504, "ts": 1716454222495043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495153, "dur": 3,
    "args": {
      "External id": 71912,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71912,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0023816612086930635
    }
  },
  {
    "ph": "f", "id": 71912, "pid": 5, "tid": 7, "ts": 1716454222495153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495126, "dur": 28,
    "args": {
      "External id": 71912,
      "cbid": 41, "correlation": 71912
    }
  },
  {
    "ph": "s", "id": 71912, "pid": 76337, "tid": -914061504, "ts": 1716454222495126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222495238, "dur": 4,
    "args": {
      "External id": 71922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71922,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 71922, "pid": 5, "tid": 7, "ts": 1716454222495238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495223, "dur": 16,
    "args": {
      "External id": 71922,
      "cbid": 211, "correlation": 71922
    }
  },
  {
    "ph": "s", "id": 71922, "pid": 76337, "tid": -914061504, "ts": 1716454222495223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495301, "dur": 1,
    "args": {
      "External id": 71932,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71932,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 71932, "pid": 5, "tid": 7, "ts": 1716454222495301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495285, "dur": 14,
    "args": {
      "External id": 71932,
      "cbid": 41, "correlation": 71932
    }
  },
  {
    "ph": "s", "id": 71932, "pid": 76337, "tid": -914061504, "ts": 1716454222495285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495300, "dur": 8,
    "args": {
      "External id": 71933,
      "cbid": 131, "correlation": 71933
    }
  },
  {
    "ph": "f", "id": 71933, "pid": 76337, "tid": -914061504, "ts": 1716454222495300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495384, "dur": 3,
    "args": {
      "External id": 71940,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71940,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.11764705882353
    }
  },
  {
    "ph": "f", "id": 71940, "pid": 5, "tid": 7, "ts": 1716454222495384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495347, "dur": 37,
    "args": {
      "External id": 71940,
      "cbid": 41, "correlation": 71940
    }
  },
  {
    "ph": "s", "id": 71940, "pid": 76337, "tid": -914061504, "ts": 1716454222495347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495477, "dur": 3,
    "args": {
      "External id": 71959,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71959,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005263157894736842
    }
  },
  {
    "ph": "f", "id": 71959, "pid": 5, "tid": 7, "ts": 1716454222495477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495458, "dur": 18,
    "args": {
      "External id": 71959,
      "cbid": 41, "correlation": 71959
    }
  },
  {
    "ph": "s", "id": 71959, "pid": 76337, "tid": -914061504, "ts": 1716454222495458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222495516, "dur": 3,
    "args": {
      "External id": 71965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 71965, "pid": 5, "tid": 7, "ts": 1716454222495516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495504, "dur": 11,
    "args": {
      "External id": 71965,
      "cbid": 211, "correlation": 71965
    }
  },
  {
    "ph": "s", "id": 71965, "pid": 76337, "tid": -914061504, "ts": 1716454222495504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222495530, "dur": 6,
    "args": {
      "External id": 71967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71967,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 71967, "pid": 5, "tid": 7, "ts": 1716454222495530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495519, "dur": 9,
    "args": {
      "External id": 71967,
      "cbid": 211, "correlation": 71967
    }
  },
  {
    "ph": "s", "id": 71967, "pid": 76337, "tid": -914061504, "ts": 1716454222495519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454222495538, "dur": 3,
    "args": {
      "External id": 71969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71969,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 71969, "pid": 5, "tid": 7, "ts": 1716454222495538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495530, "dur": 6,
    "args": {
      "External id": 71969,
      "cbid": 211, "correlation": 71969
    }
  },
  {
    "ph": "s", "id": 71969, "pid": 76337, "tid": -914061504, "ts": 1716454222495530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495571, "dur": 2,
    "args": {
      "External id": 71977,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 71977,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002717391304347826
    }
  },
  {
    "ph": "f", "id": 71977, "pid": 5, "tid": 7, "ts": 1716454222495571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495557, "dur": 13,
    "args": {
      "External id": 71977,
      "cbid": 41, "correlation": 71977
    }
  },
  {
    "ph": "s", "id": 71977, "pid": 76337, "tid": -914061504, "ts": 1716454222495557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222495616, "dur": 3,
    "args": {
      "External id": 71991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 71991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 71991, "pid": 5, "tid": 7, "ts": 1716454222495616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495606, "dur": 11,
    "args": {
      "External id": 71991,
      "cbid": 211, "correlation": 71991
    }
  },
  {
    "ph": "s", "id": 71991, "pid": 76337, "tid": -914061504, "ts": 1716454222495606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222495636, "dur": 2,
    "args": {
      "External id": 72005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72005, "pid": 5, "tid": 7, "ts": 1716454222495636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495628, "dur": 7,
    "args": {
      "External id": 72005,
      "cbid": 211, "correlation": 72005
    }
  },
  {
    "ph": "s", "id": 72005, "pid": 76337, "tid": -914061504, "ts": 1716454222495628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222495668, "dur": 6,
    "args": {
      "External id": 72012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72012,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72012, "pid": 5, "tid": 7, "ts": 1716454222495668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495658, "dur": 11,
    "args": {
      "External id": 72012,
      "cbid": 211, "correlation": 72012
    }
  },
  {
    "ph": "s", "id": 72012, "pid": 76337, "tid": -914061504, "ts": 1716454222495658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222495678, "dur": 6,
    "args": {
      "External id": 72015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72015,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72015, "pid": 5, "tid": 7, "ts": 1716454222495678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495671, "dur": 6,
    "args": {
      "External id": 72015,
      "cbid": 211, "correlation": 72015
    }
  },
  {
    "ph": "s", "id": 72015, "pid": 76337, "tid": -914061504, "ts": 1716454222495671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222495687, "dur": 3,
    "args": {
      "External id": 72017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72017,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72017, "pid": 5, "tid": 7, "ts": 1716454222495687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495679, "dur": 6,
    "args": {
      "External id": 72017,
      "cbid": 211, "correlation": 72017
    }
  },
  {
    "ph": "s", "id": 72017, "pid": 76337, "tid": -914061504, "ts": 1716454222495679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495707, "dur": 2,
    "args": {
      "External id": 72020,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72020,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 72020, "pid": 5, "tid": 7, "ts": 1716454222495707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495694, "dur": 14,
    "args": {
      "External id": 72020,
      "cbid": 41, "correlation": 72020
    }
  },
  {
    "ph": "s", "id": 72020, "pid": 76337, "tid": -914061504, "ts": 1716454222495694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222495761, "dur": 4,
    "args": {
      "External id": 72036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72036,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72036, "pid": 5, "tid": 7, "ts": 1716454222495761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495748, "dur": 12,
    "args": {
      "External id": 72036,
      "cbid": 211, "correlation": 72036
    }
  },
  {
    "ph": "s", "id": 72036, "pid": 76337, "tid": -914061504, "ts": 1716454222495748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222495781, "dur": 3,
    "args": {
      "External id": 72041,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72041,
      "bytes": 1, "memory bandwidth (GB/s)": 0.00031259768677711783
    }
  },
  {
    "ph": "f", "id": 72041, "pid": 5, "tid": 7, "ts": 1716454222495781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495765, "dur": 14,
    "args": {
      "External id": 72041,
      "cbid": 41, "correlation": 72041
    }
  },
  {
    "ph": "s", "id": 72041, "pid": 76337, "tid": -914061504, "ts": 1716454222495765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222495809, "dur": 1,
    "args": {
      "External id": 72047,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72047,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 72047, "pid": 5, "tid": 7, "ts": 1716454222495809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495790, "dur": 28,
    "args": {
      "External id": 72047,
      "cbid": 41, "correlation": 72047
    }
  },
  {
    "ph": "s", "id": 72047, "pid": 76337, "tid": -914061504, "ts": 1716454222495790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495818, "dur": 4,
    "args": {
      "External id": 72048,
      "cbid": 131, "correlation": 72048
    }
  },
  {
    "ph": "f", "id": 72048, "pid": 76337, "tid": -914061504, "ts": 1716454222495818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222495868, "dur": 3,
    "args": {
      "External id": 72056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72056,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72056, "pid": 5, "tid": 7, "ts": 1716454222495868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495855, "dur": 12,
    "args": {
      "External id": 72056,
      "cbid": 211, "correlation": 72056
    }
  },
  {
    "ph": "s", "id": 72056, "pid": 76337, "tid": -914061504, "ts": 1716454222495855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222495897, "dur": 3,
    "args": {
      "External id": 72066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72066,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72066, "pid": 5, "tid": 7, "ts": 1716454222495897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495888, "dur": 9,
    "args": {
      "External id": 72066,
      "cbid": 211, "correlation": 72066
    }
  },
  {
    "ph": "s", "id": 72066, "pid": 76337, "tid": -914061504, "ts": 1716454222495888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222495921, "dur": 3,
    "args": {
      "External id": 72075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72075,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72075, "pid": 5, "tid": 7, "ts": 1716454222495921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222495912, "dur": 7,
    "args": {
      "External id": 72075,
      "cbid": 211, "correlation": 72075
    }
  },
  {
    "ph": "s", "id": 72075, "pid": 76337, "tid": -914061504, "ts": 1716454222495912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222496038, "dur": 12,
    "args": {
      "External id": 72085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72085,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72085, "pid": 5, "tid": 7, "ts": 1716454222496038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496024, "dur": 15,
    "args": {
      "External id": 72085,
      "cbid": 211, "correlation": 72085
    }
  },
  {
    "ph": "s", "id": 72085, "pid": 76337, "tid": -914061504, "ts": 1716454222496024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222496076, "dur": 3,
    "args": {
      "External id": 72093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72093,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72093, "pid": 5, "tid": 7, "ts": 1716454222496076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496067, "dur": 11,
    "args": {
      "External id": 72093,
      "cbid": 211, "correlation": 72093
    }
  },
  {
    "ph": "s", "id": 72093, "pid": 76337, "tid": -914061504, "ts": 1716454222496067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222496124, "dur": 12,
    "args": {
      "External id": 72103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72103,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72103, "pid": 5, "tid": 7, "ts": 1716454222496124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496113, "dur": 11,
    "args": {
      "External id": 72103,
      "cbid": 211, "correlation": 72103
    }
  },
  {
    "ph": "s", "id": 72103, "pid": 76337, "tid": -914061504, "ts": 1716454222496113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222496154, "dur": 10,
    "args": {
      "External id": 72111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72111,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72111, "pid": 5, "tid": 7, "ts": 1716454222496154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496145, "dur": 9,
    "args": {
      "External id": 72111,
      "cbid": 211, "correlation": 72111
    }
  },
  {
    "ph": "s", "id": 72111, "pid": 76337, "tid": -914061504, "ts": 1716454222496145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222496181, "dur": 3,
    "args": {
      "External id": 72120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72120, "pid": 5, "tid": 7, "ts": 1716454222496181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496172, "dur": 8,
    "args": {
      "External id": 72120,
      "cbid": 211, "correlation": 72120
    }
  },
  {
    "ph": "s", "id": 72120, "pid": 76337, "tid": -914061504, "ts": 1716454222496172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222496206, "dur": 5,
    "args": {
      "External id": 72129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72129,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72129, "pid": 5, "tid": 7, "ts": 1716454222496206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496197, "dur": 8,
    "args": {
      "External id": 72129,
      "cbid": 211, "correlation": 72129
    }
  },
  {
    "ph": "s", "id": 72129, "pid": 76337, "tid": -914061504, "ts": 1716454222496197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222496241, "dur": 8,
    "args": {
      "External id": 72139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72139,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72139, "pid": 5, "tid": 7, "ts": 1716454222496241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496231, "dur": 10,
    "args": {
      "External id": 72139,
      "cbid": 211, "correlation": 72139
    }
  },
  {
    "ph": "s", "id": 72139, "pid": 76337, "tid": -914061504, "ts": 1716454222496231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222496540, "dur": 3,
    "args": {
      "External id": 72148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72148,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72148, "pid": 5, "tid": 7, "ts": 1716454222496540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496525, "dur": 15,
    "args": {
      "External id": 72148,
      "cbid": 211, "correlation": 72148
    }
  },
  {
    "ph": "s", "id": 72148, "pid": 76337, "tid": -914061504, "ts": 1716454222496525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222496566, "dur": 3,
    "args": {
      "External id": 72156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72156,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72156, "pid": 5, "tid": 7, "ts": 1716454222496566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496557, "dur": 8,
    "args": {
      "External id": 72156,
      "cbid": 211, "correlation": 72156
    }
  },
  {
    "ph": "s", "id": 72156, "pid": 76337, "tid": -914061504, "ts": 1716454222496557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222496620, "dur": 1,
    "args": {
      "External id": 72166,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72166,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 72166, "pid": 5, "tid": 7, "ts": 1716454222496620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496605, "dur": 13,
    "args": {
      "External id": 72166,
      "cbid": 41, "correlation": 72166
    }
  },
  {
    "ph": "s", "id": 72166, "pid": 76337, "tid": -914061504, "ts": 1716454222496605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496619, "dur": 8,
    "args": {
      "External id": 72167,
      "cbid": 131, "correlation": 72167
    }
  },
  {
    "ph": "f", "id": 72167, "pid": 76337, "tid": -914061504, "ts": 1716454222496619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222496709, "dur": 2,
    "args": {
      "External id": 72175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72175, "pid": 5, "tid": 7, "ts": 1716454222496709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496694, "dur": 15,
    "args": {
      "External id": 72175,
      "cbid": 211, "correlation": 72175
    }
  },
  {
    "ph": "s", "id": 72175, "pid": 76337, "tid": -914061504, "ts": 1716454222496694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222496780, "dur": 3,
    "args": {
      "External id": 72184,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72184,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 72184, "pid": 5, "tid": 7, "ts": 1716454222496780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496762, "dur": 17,
    "args": {
      "External id": 72184,
      "cbid": 41, "correlation": 72184
    }
  },
  {
    "ph": "s", "id": 72184, "pid": 76337, "tid": -914061504, "ts": 1716454222496762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222496850, "dur": 3,
    "args": {
      "External id": 72194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72194,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72194, "pid": 5, "tid": 7, "ts": 1716454222496850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222496836, "dur": 13,
    "args": {
      "External id": 72194,
      "cbid": 211, "correlation": 72194
    }
  },
  {
    "ph": "s", "id": 72194, "pid": 76337, "tid": -914061504, "ts": 1716454222496836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224868486, "dur": 51,
    "args": {
      "External id": 217151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217151,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 217151, "pid": 5, "tid": 7, "ts": 1716454224868486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758742, "dur": 25,
    "args": {
      "External id": 217151,
      "cbid": 211, "correlation": 217151
    }
  },
  {
    "ph": "s", "id": 217151, "pid": 76337, "tid": -914061504, "ts": 1716454224758742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224868538, "dur": 4,
    "args": {
      "External id": 217163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217163,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217163, "pid": 5, "tid": 7, "ts": 1716454224868538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758785, "dur": 7,
    "args": {
      "External id": 217163,
      "cbid": 211, "correlation": 217163
    }
  },
  {
    "ph": "s", "id": 217163, "pid": 76337, "tid": -914061504, "ts": 1716454224758785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224868544, "dur": 56,
    "args": {
      "External id": 217166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217166, "pid": 5, "tid": 7, "ts": 1716454224868544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758811, "dur": 8,
    "args": {
      "External id": 217166,
      "cbid": 211, "correlation": 217166
    }
  },
  {
    "ph": "s", "id": 217166, "pid": 76337, "tid": -914061504, "ts": 1716454224758811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758913, "dur": 0,
    "args": {
      "External id": 217177,
      "cbid": 317, "correlation": 217177
    }
  },
  {
    "ph": "f", "id": 217177, "pid": 76337, "tid": -914061504, "ts": 1716454224758913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758915, "dur": 1,
    "args": {
      "External id": 217178,
      "cbid": 203, "correlation": 217178
    }
  },
  {
    "ph": "f", "id": 217178, "pid": 76337, "tid": -914061504, "ts": 1716454224758915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758916, "dur": 0,
    "args": {
      "External id": 217179,
      "cbid": 205, "correlation": 217179
    }
  },
  {
    "ph": "f", "id": 217179, "pid": 76337, "tid": -914061504, "ts": 1716454224758916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758993, "dur": 4,
    "args": {
      "External id": 217183,
      "cbid": 251, "correlation": 217183
    }
  },
  {
    "ph": "f", "id": 217183, "pid": 76337, "tid": -914061504, "ts": 1716454224758993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758998, "dur": 1,
    "args": {
      "External id": 217184,
      "cbid": 251, "correlation": 217184
    }
  },
  {
    "ph": "f", "id": 217184, "pid": 76337, "tid": -914061504, "ts": 1716454224758998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759000, "dur": 1,
    "args": {
      "External id": 217185,
      "cbid": 251, "correlation": 217185
    }
  },
  {
    "ph": "f", "id": 217185, "pid": 76337, "tid": -914061504, "ts": 1716454224759000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759001, "dur": 1,
    "args": {
      "External id": 217186,
      "cbid": 251, "correlation": 217186
    }
  },
  {
    "ph": "f", "id": 217186, "pid": 76337, "tid": -914061504, "ts": 1716454224759001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759003, "dur": 1,
    "args": {
      "External id": 217187,
      "cbid": 251, "correlation": 217187
    }
  },
  {
    "ph": "f", "id": 217187, "pid": 76337, "tid": -914061504, "ts": 1716454224759003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759005, "dur": 1,
    "args": {
      "External id": 217188,
      "cbid": 251, "correlation": 217188
    }
  },
  {
    "ph": "f", "id": 217188, "pid": 76337, "tid": -914061504, "ts": 1716454224759005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759007, "dur": 0,
    "args": {
      "External id": 217189,
      "cbid": 251, "correlation": 217189
    }
  },
  {
    "ph": "f", "id": 217189, "pid": 76337, "tid": -914061504, "ts": 1716454224759007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759008, "dur": 0,
    "args": {
      "External id": 217190,
      "cbid": 251, "correlation": 217190
    }
  },
  {
    "ph": "f", "id": 217190, "pid": 76337, "tid": -914061504, "ts": 1716454224759008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759010, "dur": 0,
    "args": {
      "External id": 217191,
      "cbid": 251, "correlation": 217191
    }
  },
  {
    "ph": "f", "id": 217191, "pid": 76337, "tid": -914061504, "ts": 1716454224759010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224868601, "dur": 112,
    "args": {
      "External id": 217192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217192,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217192, "pid": 5, "tid": 7, "ts": 1716454224868601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759014, "dur": 15,
    "args": {
      "External id": 217192,
      "cbid": 211, "correlation": 217192
    }
  },
  {
    "ph": "s", "id": 217192, "pid": 76337, "tid": -914061504, "ts": 1716454224759014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224868715, "dur": 60,
    "args": {
      "External id": 217198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217198, "pid": 5, "tid": 7, "ts": 1716454224868715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759083, "dur": 11,
    "args": {
      "External id": 217198,
      "cbid": 211, "correlation": 217198
    }
  },
  {
    "ph": "s", "id": 217198, "pid": 76337, "tid": -914061504, "ts": 1716454224759083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224868776, "dur": 462,
    "args": {
      "External id": 217207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217207,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217207, "pid": 5, "tid": 7, "ts": 1716454224868776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759214, "dur": 16,
    "args": {
      "External id": 217207,
      "cbid": 211, "correlation": 217207
    }
  },
  {
    "ph": "s", "id": 217207, "pid": 76337, "tid": -914061504, "ts": 1716454224759214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224869239, "dur": 182,
    "args": {
      "External id": 217229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217229,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217229, "pid": 5, "tid": 7, "ts": 1716454224869239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759295, "dur": 11,
    "args": {
      "External id": 217229,
      "cbid": 211, "correlation": 217229
    }
  },
  {
    "ph": "s", "id": 217229, "pid": 76337, "tid": -914061504, "ts": 1716454224759295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759439, "dur": 2,
    "args": {
      "External id": 217240,
      "cbid": 251, "correlation": 217240
    }
  },
  {
    "ph": "f", "id": 217240, "pid": 76337, "tid": -914061504, "ts": 1716454224759439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224869423, "dur": 197,
    "args": {
      "External id": 217241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217241,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217241, "pid": 5, "tid": 7, "ts": 1716454224869423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759446, "dur": 13,
    "args": {
      "External id": 217241,
      "cbid": 211, "correlation": 217241
    }
  },
  {
    "ph": "s", "id": 217241, "pid": 76337, "tid": -914061504, "ts": 1716454224759446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759520, "dur": 1,
    "args": {
      "External id": 217252,
      "cbid": 251, "correlation": 217252
    }
  },
  {
    "ph": "f", "id": 217252, "pid": 76337, "tid": -914061504, "ts": 1716454224759520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224869621, "dur": 188,
    "args": {
      "External id": 217253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217253,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217253, "pid": 5, "tid": 7, "ts": 1716454224869621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759523, "dur": 12,
    "args": {
      "External id": 217253,
      "cbid": 211, "correlation": 217253
    }
  },
  {
    "ph": "s", "id": 217253, "pid": 76337, "tid": -914061504, "ts": 1716454224759523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759589, "dur": 1,
    "args": {
      "External id": 217264,
      "cbid": 251, "correlation": 217264
    }
  },
  {
    "ph": "f", "id": 217264, "pid": 76337, "tid": -914061504, "ts": 1716454224759589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224869811, "dur": 189,
    "args": {
      "External id": 217265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217265,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217265, "pid": 5, "tid": 7, "ts": 1716454224869811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759593, "dur": 11,
    "args": {
      "External id": 217265,
      "cbid": 211, "correlation": 217265
    }
  },
  {
    "ph": "s", "id": 217265, "pid": 76337, "tid": -914061504, "ts": 1716454224759593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224870001, "dur": 18734,
    "args": {
      "External id": 217286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217286,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217286, "pid": 5, "tid": 7, "ts": 1716454224870001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759706, "dur": 16,
    "args": {
      "External id": 217286,
      "cbid": 211, "correlation": 217286
    }
  },
  {
    "ph": "s", "id": 217286, "pid": 76337, "tid": -914061504, "ts": 1716454224759706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759840, "dur": 2,
    "args": {
      "External id": 217304,
      "cbid": 251, "correlation": 217304
    }
  },
  {
    "ph": "f", "id": 217304, "pid": 76337, "tid": -914061504, "ts": 1716454224759840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224888737, "dur": 202,
    "args": {
      "External id": 217306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217306,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217306, "pid": 5, "tid": 7, "ts": 1716454224888737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759846, "dur": 14,
    "args": {
      "External id": 217306,
      "cbid": 211, "correlation": 217306
    }
  },
  {
    "ph": "s", "id": 217306, "pid": 76337, "tid": -914061504, "ts": 1716454224759846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224888940, "dur": 66,
    "args": {
      "External id": 217314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217314,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217314, "pid": 5, "tid": 7, "ts": 1716454224888940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224759949, "dur": 15,
    "args": {
      "External id": 217314,
      "cbid": 211, "correlation": 217314
    }
  },
  {
    "ph": "s", "id": 217314, "pid": 76337, "tid": -914061504, "ts": 1716454224759949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224889007, "dur": 96,
    "args": {
      "External id": 217322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217322,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217322, "pid": 5, "tid": 7, "ts": 1716454224889007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224760010, "dur": 10,
    "args": {
      "External id": 217322,
      "cbid": 211, "correlation": 217322
    }
  },
  {
    "ph": "s", "id": 217322, "pid": 76337, "tid": -914061504, "ts": 1716454224760010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224889105, "dur": 55,
    "args": {
      "External id": 217333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217333, "pid": 5, "tid": 7, "ts": 1716454224889105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224760116, "dur": 15,
    "args": {
      "External id": 217333,
      "cbid": 211, "correlation": 217333
    }
  },
  {
    "ph": "s", "id": 217333, "pid": 76337, "tid": -914061504, "ts": 1716454224760116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224889160, "dur": 92,
    "args": {
      "External id": 217355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217355,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217355, "pid": 5, "tid": 7, "ts": 1716454224889160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224760153, "dur": 1134,
    "args": {
      "External id": 217355,
      "cbid": 211, "correlation": 217355
    }
  },
  {
    "ph": "s", "id": 217355, "pid": 76337, "tid": -914061504, "ts": 1716454224760153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761371, "dur": 1,
    "args": {
      "External id": 217366,
      "cbid": 251, "correlation": 217366
    }
  },
  {
    "ph": "f", "id": 217366, "pid": 76337, "tid": -914061504, "ts": 1716454224761371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224889254, "dur": 103,
    "args": {
      "External id": 217367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217367,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217367, "pid": 5, "tid": 7, "ts": 1716454224889254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761377, "dur": 59,
    "args": {
      "External id": 217367,
      "cbid": 211, "correlation": 217367
    }
  },
  {
    "ph": "s", "id": 217367, "pid": 76337, "tid": -914061504, "ts": 1716454224761377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761509, "dur": 2,
    "args": {
      "External id": 217378,
      "cbid": 251, "correlation": 217378
    }
  },
  {
    "ph": "f", "id": 217378, "pid": 76337, "tid": -914061504, "ts": 1716454224761509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761515, "dur": 0,
    "args": {
      "External id": 217379,
      "cbid": 251, "correlation": 217379
    }
  },
  {
    "ph": "f", "id": 217379, "pid": 76337, "tid": -914061504, "ts": 1716454224761515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224889359, "dur": 10,
    "args": {
      "External id": 217380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217380,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 217380, "pid": 5, "tid": 7, "ts": 1716454224889359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761517, "dur": 14,
    "args": {
      "External id": 217380,
      "cbid": 211, "correlation": 217380
    }
  },
  {
    "ph": "s", "id": 217380, "pid": 76337, "tid": -914061504, "ts": 1716454224761517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224889370, "dur": 5,
    "args": {
      "External id": 217382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217382,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 217382, "pid": 5, "tid": 7, "ts": 1716454224889370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761534, "dur": 9,
    "args": {
      "External id": 217382,
      "cbid": 211, "correlation": 217382
    }
  },
  {
    "ph": "s", "id": 217382, "pid": 76337, "tid": -914061504, "ts": 1716454224761534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761599, "dur": 1,
    "args": {
      "External id": 217393,
      "cbid": 251, "correlation": 217393
    }
  },
  {
    "ph": "f", "id": 217393, "pid": 76337, "tid": -914061504, "ts": 1716454224761599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761603, "dur": 0,
    "args": {
      "External id": 217394,
      "cbid": 251, "correlation": 217394
    }
  },
  {
    "ph": "f", "id": 217394, "pid": 76337, "tid": -914061504, "ts": 1716454224761603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224889377, "dur": 6,
    "args": {
      "External id": 217395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217395,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 217395, "pid": 5, "tid": 7, "ts": 1716454224889377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761604, "dur": 12,
    "args": {
      "External id": 217395,
      "cbid": 211, "correlation": 217395
    }
  },
  {
    "ph": "s", "id": 217395, "pid": 76337, "tid": -914061504, "ts": 1716454224761604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224889384, "dur": 3,
    "args": {
      "External id": 217397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217397,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 217397, "pid": 5, "tid": 7, "ts": 1716454224889384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761618, "dur": 5,
    "args": {
      "External id": 217397,
      "cbid": 211, "correlation": 217397
    }
  },
  {
    "ph": "s", "id": 217397, "pid": 76337, "tid": -914061504, "ts": 1716454224761618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224889389, "dur": 157,
    "args": {
      "External id": 217418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217418,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217418, "pid": 5, "tid": 7, "ts": 1716454224889389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761694, "dur": 12,
    "args": {
      "External id": 217418,
      "cbid": 211, "correlation": 217418
    }
  },
  {
    "ph": "s", "id": 217418, "pid": 76337, "tid": -914061504, "ts": 1716454224761694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761793, "dur": 2,
    "args": {
      "External id": 217436,
      "cbid": 251, "correlation": 217436
    }
  },
  {
    "ph": "f", "id": 217436, "pid": 76337, "tid": -914061504, "ts": 1716454224761793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224889548, "dur": 109,
    "args": {
      "External id": 217438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217438,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217438, "pid": 5, "tid": 7, "ts": 1716454224889548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761799, "dur": 14,
    "args": {
      "External id": 217438,
      "cbid": 211, "correlation": 217438
    }
  },
  {
    "ph": "s", "id": 217438, "pid": 76337, "tid": -914061504, "ts": 1716454224761799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224889658, "dur": 35,
    "args": {
      "External id": 217446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217446,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217446, "pid": 5, "tid": 7, "ts": 1716454224889658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761869, "dur": 12,
    "args": {
      "External id": 217446,
      "cbid": 211, "correlation": 217446
    }
  },
  {
    "ph": "s", "id": 217446, "pid": 76337, "tid": -914061504, "ts": 1716454224761869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224889693, "dur": 66,
    "args": {
      "External id": 217454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217454, "pid": 5, "tid": 7, "ts": 1716454224889693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761910, "dur": 10,
    "args": {
      "External id": 217454,
      "cbid": 211, "correlation": 217454
    }
  },
  {
    "ph": "s", "id": 217454, "pid": 76337, "tid": -914061504, "ts": 1716454224761910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224889761, "dur": 93,
    "args": {
      "External id": 217476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217476,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217476, "pid": 5, "tid": 7, "ts": 1716454224889761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224761963, "dur": 19,
    "args": {
      "External id": 217476,
      "cbid": 211, "correlation": 217476
    }
  },
  {
    "ph": "s", "id": 217476, "pid": 76337, "tid": -914061504, "ts": 1716454224761963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762071, "dur": 1,
    "args": {
      "External id": 217492,
      "cbid": 251, "correlation": 217492
    }
  },
  {
    "ph": "f", "id": 217492, "pid": 76337, "tid": -914061504, "ts": 1716454224762071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224889855, "dur": 577,
    "args": {
      "External id": 217494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217494,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217494, "pid": 5, "tid": 7, "ts": 1716454224889855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762077, "dur": 13,
    "args": {
      "External id": 217494,
      "cbid": 211, "correlation": 217494
    }
  },
  {
    "ph": "s", "id": 217494, "pid": 76337, "tid": -914061504, "ts": 1716454224762077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224890433, "dur": 246,
    "args": {
      "External id": 217502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217502, "pid": 5, "tid": 7, "ts": 1716454224890433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762166, "dur": 14,
    "args": {
      "External id": 217502,
      "cbid": 211, "correlation": 217502
    }
  },
  {
    "ph": "s", "id": 217502, "pid": 76337, "tid": -914061504, "ts": 1716454224762166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224890680, "dur": 254,
    "args": {
      "External id": 217510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217510, "pid": 5, "tid": 7, "ts": 1716454224890680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762208, "dur": 11,
    "args": {
      "External id": 217510,
      "cbid": 211, "correlation": 217510
    }
  },
  {
    "ph": "s", "id": 217510, "pid": 76337, "tid": -914061504, "ts": 1716454224762208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762294, "dur": 2,
    "args": {
      "External id": 217526,
      "cbid": 251, "correlation": 217526
    }
  },
  {
    "ph": "f", "id": 217526, "pid": 76337, "tid": -914061504, "ts": 1716454224762294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762300, "dur": 0,
    "args": {
      "External id": 217528,
      "cbid": 251, "correlation": 217528
    }
  },
  {
    "ph": "f", "id": 217528, "pid": 76337, "tid": -914061504, "ts": 1716454224762300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224890935, "dur": 364,
    "args": {
      "External id": 217529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217529,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 217529, "pid": 5, "tid": 7, "ts": 1716454224890935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762305, "dur": 13,
    "args": {
      "External id": 217529,
      "cbid": 211, "correlation": 217529
    }
  },
  {
    "ph": "s", "id": 217529, "pid": 76337, "tid": -914061504, "ts": 1716454224762305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224891300, "dur": 50,
    "args": {
      "External id": 217537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217537,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217537, "pid": 5, "tid": 7, "ts": 1716454224891300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762347, "dur": 10,
    "args": {
      "External id": 217537,
      "cbid": 211, "correlation": 217537
    }
  },
  {
    "ph": "s", "id": 217537, "pid": 76337, "tid": -914061504, "ts": 1716454224762347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224891351, "dur": 159,
    "args": {
      "External id": 217548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217548, "pid": 5, "tid": 7, "ts": 1716454224891351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762422, "dur": 133,
    "args": {
      "External id": 217548,
      "cbid": 211, "correlation": 217548
    }
  },
  {
    "ph": "s", "id": 217548, "pid": 76337, "tid": -914061504, "ts": 1716454224762422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762629, "dur": 1,
    "args": {
      "External id": 217560,
      "cbid": 317, "correlation": 217560
    }
  },
  {
    "ph": "f", "id": 217560, "pid": 76337, "tid": -914061504, "ts": 1716454224762629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762631, "dur": 1,
    "args": {
      "External id": 217561,
      "cbid": 203, "correlation": 217561
    }
  },
  {
    "ph": "f", "id": 217561, "pid": 76337, "tid": -914061504, "ts": 1716454224762631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762633, "dur": 1,
    "args": {
      "External id": 217562,
      "cbid": 205, "correlation": 217562
    }
  },
  {
    "ph": "f", "id": 217562, "pid": 76337, "tid": -914061504, "ts": 1716454224762633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762922, "dur": 2,
    "args": {
      "External id": 217566,
      "cbid": 251, "correlation": 217566
    }
  },
  {
    "ph": "f", "id": 217566, "pid": 76337, "tid": -914061504, "ts": 1716454224762922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762925, "dur": 1,
    "args": {
      "External id": 217567,
      "cbid": 251, "correlation": 217567
    }
  },
  {
    "ph": "f", "id": 217567, "pid": 76337, "tid": -914061504, "ts": 1716454224762925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762927, "dur": 1,
    "args": {
      "External id": 217568,
      "cbid": 251, "correlation": 217568
    }
  },
  {
    "ph": "f", "id": 217568, "pid": 76337, "tid": -914061504, "ts": 1716454224762927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762929, "dur": 1,
    "args": {
      "External id": 217569,
      "cbid": 251, "correlation": 217569
    }
  },
  {
    "ph": "f", "id": 217569, "pid": 76337, "tid": -914061504, "ts": 1716454224762929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762931, "dur": 1,
    "args": {
      "External id": 217570,
      "cbid": 251, "correlation": 217570
    }
  },
  {
    "ph": "f", "id": 217570, "pid": 76337, "tid": -914061504, "ts": 1716454224762931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762933, "dur": 1,
    "args": {
      "External id": 217571,
      "cbid": 251, "correlation": 217571
    }
  },
  {
    "ph": "f", "id": 217571, "pid": 76337, "tid": -914061504, "ts": 1716454224762933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762935, "dur": 0,
    "args": {
      "External id": 217572,
      "cbid": 251, "correlation": 217572
    }
  },
  {
    "ph": "f", "id": 217572, "pid": 76337, "tid": -914061504, "ts": 1716454224762935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762936, "dur": 1,
    "args": {
      "External id": 217573,
      "cbid": 251, "correlation": 217573
    }
  },
  {
    "ph": "f", "id": 217573, "pid": 76337, "tid": -914061504, "ts": 1716454224762936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762938, "dur": 0,
    "args": {
      "External id": 217574,
      "cbid": 251, "correlation": 217574
    }
  },
  {
    "ph": "f", "id": 217574, "pid": 76337, "tid": -914061504, "ts": 1716454224762938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224891511, "dur": 117,
    "args": {
      "External id": 217575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217575,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217575, "pid": 5, "tid": 7, "ts": 1716454224891511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762940, "dur": 13,
    "args": {
      "External id": 217575,
      "cbid": 211, "correlation": 217575
    }
  },
  {
    "ph": "s", "id": 217575, "pid": 76337, "tid": -914061504, "ts": 1716454224762940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224891629, "dur": 61,
    "args": {
      "External id": 217581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217581,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217581, "pid": 5, "tid": 7, "ts": 1716454224891629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224762986, "dur": 9,
    "args": {
      "External id": 217581,
      "cbid": 211, "correlation": 217581
    }
  },
  {
    "ph": "s", "id": 217581, "pid": 76337, "tid": -914061504, "ts": 1716454224762986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224891691, "dur": 50,
    "args": {
      "External id": 217589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217589,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217589, "pid": 5, "tid": 7, "ts": 1716454224891691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763019, "dur": 98,
    "args": {
      "External id": 217589,
      "cbid": 211, "correlation": 217589
    }
  },
  {
    "ph": "s", "id": 217589, "pid": 76337, "tid": -914061504, "ts": 1716454224763019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224891742, "dur": 99,
    "args": {
      "External id": 217598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217598,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217598, "pid": 5, "tid": 7, "ts": 1716454224891742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763161, "dur": 12,
    "args": {
      "External id": 217598,
      "cbid": 211, "correlation": 217598
    }
  },
  {
    "ph": "s", "id": 217598, "pid": 76337, "tid": -914061504, "ts": 1716454224763161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224891842, "dur": 93,
    "args": {
      "External id": 217618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217618,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 217618, "pid": 5, "tid": 7, "ts": 1716454224891842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763259, "dur": 13,
    "args": {
      "External id": 217618,
      "cbid": 211, "correlation": 217618
    }
  },
  {
    "ph": "s", "id": 217618, "pid": 76337, "tid": -914061504, "ts": 1716454224763259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224891937, "dur": 5,
    "args": {
      "External id": 217630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217630,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 217630, "pid": 5, "tid": 7, "ts": 1716454224891937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763283, "dur": 7,
    "args": {
      "External id": 217630,
      "cbid": 211, "correlation": 217630
    }
  },
  {
    "ph": "s", "id": 217630, "pid": 76337, "tid": -914061504, "ts": 1716454224763283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224891943, "dur": 107,
    "args": {
      "External id": 217633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217633, "pid": 5, "tid": 7, "ts": 1716454224891943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763305, "dur": 70,
    "args": {
      "External id": 217633,
      "cbid": 211, "correlation": 217633
    }
  },
  {
    "ph": "s", "id": 217633, "pid": 76337, "tid": -914061504, "ts": 1716454224763305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224892051, "dur": 69,
    "args": {
      "External id": 217642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217642,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217642, "pid": 5, "tid": 7, "ts": 1716454224892051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763422, "dur": 12,
    "args": {
      "External id": 217642,
      "cbid": 211, "correlation": 217642
    }
  },
  {
    "ph": "s", "id": 217642, "pid": 76337, "tid": -914061504, "ts": 1716454224763422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763479, "dur": 0,
    "args": {
      "External id": 217652,
      "cbid": 317, "correlation": 217652
    }
  },
  {
    "ph": "f", "id": 217652, "pid": 76337, "tid": -914061504, "ts": 1716454224763479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763480, "dur": 0,
    "args": {
      "External id": 217653,
      "cbid": 203, "correlation": 217653
    }
  },
  {
    "ph": "f", "id": 217653, "pid": 76337, "tid": -914061504, "ts": 1716454224763480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763481, "dur": 0,
    "args": {
      "External id": 217654,
      "cbid": 205, "correlation": 217654
    }
  },
  {
    "ph": "f", "id": 217654, "pid": 76337, "tid": -914061504, "ts": 1716454224763481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224892121, "dur": 78,
    "args": {
      "External id": 217658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217658, "pid": 5, "tid": 7, "ts": 1716454224892121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763523, "dur": 15,
    "args": {
      "External id": 217658,
      "cbid": 211, "correlation": 217658
    }
  },
  {
    "ph": "s", "id": 217658, "pid": 76337, "tid": -914061504, "ts": 1716454224763523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224892201, "dur": 25,
    "args": {
      "External id": 217660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217660, "pid": 5, "tid": 7, "ts": 1716454224892201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763540, "dur": 6,
    "args": {
      "External id": 217660,
      "cbid": 211, "correlation": 217660
    }
  },
  {
    "ph": "s", "id": 217660, "pid": 76337, "tid": -914061504, "ts": 1716454224763540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224892227, "dur": 4,
    "args": {
      "External id": 217662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217662, "pid": 5, "tid": 7, "ts": 1716454224892227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763558, "dur": 8,
    "args": {
      "External id": 217662,
      "cbid": 211, "correlation": 217662
    }
  },
  {
    "ph": "s", "id": 217662, "pid": 76337, "tid": -914061504, "ts": 1716454224763558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763573, "dur": 0,
    "args": {
      "External id": 217663,
      "cbid": 51, "correlation": 217663
    }
  },
  {
    "ph": "s", "id": 217663, "pid": 76337, "tid": -914061504, "ts": 1716454224763573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224892231, "dur": 1377,
    "args": {
      "External id": 217664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217664,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217664, "pid": 5, "tid": 7, "ts": 1716454224892231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763574, "dur": 7,
    "args": {
      "External id": 217664,
      "cbid": 211, "correlation": 217664
    }
  },
  {
    "ph": "s", "id": 217664, "pid": 76337, "tid": -914061504, "ts": 1716454224763574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224893609, "dur": 60,
    "args": {
      "External id": 217669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217669, "pid": 5, "tid": 7, "ts": 1716454224893609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763604, "dur": 8,
    "args": {
      "External id": 217669,
      "cbid": 211, "correlation": 217669
    }
  },
  {
    "ph": "s", "id": 217669, "pid": 76337, "tid": -914061504, "ts": 1716454224763604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224893671, "dur": 3,
    "args": {
      "External id": 217677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217677,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217677, "pid": 5, "tid": 7, "ts": 1716454224893671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763649, "dur": 9,
    "args": {
      "External id": 217677,
      "cbid": 211, "correlation": 217677
    }
  },
  {
    "ph": "s", "id": 217677, "pid": 76337, "tid": -914061504, "ts": 1716454224763649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763721, "dur": 2,
    "args": {
      "External id": 217693,
      "cbid": 251, "correlation": 217693
    }
  },
  {
    "ph": "f", "id": 217693, "pid": 76337, "tid": -914061504, "ts": 1716454224763721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763727, "dur": 0,
    "args": {
      "External id": 217695,
      "cbid": 251, "correlation": 217695
    }
  },
  {
    "ph": "f", "id": 217695, "pid": 76337, "tid": -914061504, "ts": 1716454224763727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224893676, "dur": 11,
    "args": {
      "External id": 217696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217696,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 217696, "pid": 5, "tid": 7, "ts": 1716454224893676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763729, "dur": 13,
    "args": {
      "External id": 217696,
      "cbid": 211, "correlation": 217696
    }
  },
  {
    "ph": "s", "id": 217696, "pid": 76337, "tid": -914061504, "ts": 1716454224763729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224893688, "dur": 5,
    "args": {
      "External id": 217698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217698,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 217698, "pid": 5, "tid": 7, "ts": 1716454224893688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763747, "dur": 6,
    "args": {
      "External id": 217698,
      "cbid": 211, "correlation": 217698
    }
  },
  {
    "ph": "s", "id": 217698, "pid": 76337, "tid": -914061504, "ts": 1716454224763747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224893695, "dur": 55,
    "args": {
      "External id": 217708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217708, "pid": 5, "tid": 7, "ts": 1716454224893695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224763816, "dur": 468,
    "args": {
      "External id": 217708,
      "cbid": 211, "correlation": 217708
    }
  },
  {
    "ph": "s", "id": 217708, "pid": 76337, "tid": -914061504, "ts": 1716454224763816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224893751, "dur": 52,
    "args": {
      "External id": 217728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217728,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 217728, "pid": 5, "tid": 7, "ts": 1716454224893751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764339, "dur": 11,
    "args": {
      "External id": 217728,
      "cbid": 211, "correlation": 217728
    }
  },
  {
    "ph": "s", "id": 217728, "pid": 76337, "tid": -914061504, "ts": 1716454224764339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224893804, "dur": 4,
    "args": {
      "External id": 217740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217740,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217740, "pid": 5, "tid": 7, "ts": 1716454224893804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764360, "dur": 7,
    "args": {
      "External id": 217740,
      "cbid": 211, "correlation": 217740
    }
  },
  {
    "ph": "s", "id": 217740, "pid": 76337, "tid": -914061504, "ts": 1716454224764360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224893809, "dur": 56,
    "args": {
      "External id": 217743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217743, "pid": 5, "tid": 7, "ts": 1716454224893809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764379, "dur": 7,
    "args": {
      "External id": 217743,
      "cbid": 211, "correlation": 217743
    }
  },
  {
    "ph": "s", "id": 217743, "pid": 76337, "tid": -914061504, "ts": 1716454224764379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224893866, "dur": 37,
    "args": {
      "External id": 217752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217752,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217752, "pid": 5, "tid": 7, "ts": 1716454224893866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764421, "dur": 9,
    "args": {
      "External id": 217752,
      "cbid": 211, "correlation": 217752
    }
  },
  {
    "ph": "s", "id": 217752, "pid": 76337, "tid": -914061504, "ts": 1716454224764421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764484, "dur": 0,
    "args": {
      "External id": 217762,
      "cbid": 317, "correlation": 217762
    }
  },
  {
    "ph": "f", "id": 217762, "pid": 76337, "tid": -914061504, "ts": 1716454224764484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764485, "dur": 0,
    "args": {
      "External id": 217763,
      "cbid": 203, "correlation": 217763
    }
  },
  {
    "ph": "f", "id": 217763, "pid": 76337, "tid": -914061504, "ts": 1716454224764485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764486, "dur": 0,
    "args": {
      "External id": 217764,
      "cbid": 205, "correlation": 217764
    }
  },
  {
    "ph": "f", "id": 217764, "pid": 76337, "tid": -914061504, "ts": 1716454224764486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224893904, "dur": 40,
    "args": {
      "External id": 217768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217768, "pid": 5, "tid": 7, "ts": 1716454224893904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764512, "dur": 13,
    "args": {
      "External id": 217768,
      "cbid": 211, "correlation": 217768
    }
  },
  {
    "ph": "s", "id": 217768, "pid": 76337, "tid": -914061504, "ts": 1716454224764512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224893946, "dur": 14,
    "args": {
      "External id": 217770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217770, "pid": 5, "tid": 7, "ts": 1716454224893946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764527, "dur": 5,
    "args": {
      "External id": 217770,
      "cbid": 211, "correlation": 217770
    }
  },
  {
    "ph": "s", "id": 217770, "pid": 76337, "tid": -914061504, "ts": 1716454224764527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224893961, "dur": 3,
    "args": {
      "External id": 217772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217772, "pid": 5, "tid": 7, "ts": 1716454224893961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764538, "dur": 5,
    "args": {
      "External id": 217772,
      "cbid": 211, "correlation": 217772
    }
  },
  {
    "ph": "s", "id": 217772, "pid": 76337, "tid": -914061504, "ts": 1716454224764538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764546, "dur": 0,
    "args": {
      "External id": 217773,
      "cbid": 51, "correlation": 217773
    }
  },
  {
    "ph": "s", "id": 217773, "pid": 76337, "tid": -914061504, "ts": 1716454224764546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224893966, "dur": 699,
    "args": {
      "External id": 217774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217774,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217774, "pid": 5, "tid": 7, "ts": 1716454224893966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764547, "dur": 5,
    "args": {
      "External id": 217774,
      "cbid": 211, "correlation": 217774
    }
  },
  {
    "ph": "s", "id": 217774, "pid": 76337, "tid": -914061504, "ts": 1716454224764547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224894667, "dur": 59,
    "args": {
      "External id": 217779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217779, "pid": 5, "tid": 7, "ts": 1716454224894667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764575, "dur": 8,
    "args": {
      "External id": 217779,
      "cbid": 211, "correlation": 217779
    }
  },
  {
    "ph": "s", "id": 217779, "pid": 76337, "tid": -914061504, "ts": 1716454224764575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764633, "dur": 0,
    "args": {
      "External id": 217789,
      "cbid": 317, "correlation": 217789
    }
  },
  {
    "ph": "f", "id": 217789, "pid": 76337, "tid": -914061504, "ts": 1716454224764633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764634, "dur": 0,
    "args": {
      "External id": 217790,
      "cbid": 203, "correlation": 217790
    }
  },
  {
    "ph": "f", "id": 217790, "pid": 76337, "tid": -914061504, "ts": 1716454224764634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764634, "dur": 0,
    "args": {
      "External id": 217791,
      "cbid": 205, "correlation": 217791
    }
  },
  {
    "ph": "f", "id": 217791, "pid": 76337, "tid": -914061504, "ts": 1716454224764634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224894727, "dur": 75,
    "args": {
      "External id": 217795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217795,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217795, "pid": 5, "tid": 7, "ts": 1716454224894727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764655, "dur": 12,
    "args": {
      "External id": 217795,
      "cbid": 211, "correlation": 217795
    }
  },
  {
    "ph": "s", "id": 217795, "pid": 76337, "tid": -914061504, "ts": 1716454224764655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224894803, "dur": 210,
    "args": {
      "External id": 217797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217797,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217797, "pid": 5, "tid": 7, "ts": 1716454224894803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764681, "dur": 8,
    "args": {
      "External id": 217797,
      "cbid": 211, "correlation": 217797
    }
  },
  {
    "ph": "s", "id": 217797, "pid": 76337, "tid": -914061504, "ts": 1716454224764681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224895015, "dur": 38,
    "args": {
      "External id": 217799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217799,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217799, "pid": 5, "tid": 7, "ts": 1716454224895015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764696, "dur": 6,
    "args": {
      "External id": 217799,
      "cbid": 211, "correlation": 217799
    }
  },
  {
    "ph": "s", "id": 217799, "pid": 76337, "tid": -914061504, "ts": 1716454224764696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224895054, "dur": 60,
    "args": {
      "External id": 217805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217805, "pid": 5, "tid": 7, "ts": 1716454224895054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224764724, "dur": 482,
    "args": {
      "External id": 217805,
      "cbid": 211, "correlation": 217805
    }
  },
  {
    "ph": "s", "id": 217805, "pid": 76337, "tid": -914061504, "ts": 1716454224764724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224895116, "dur": 50,
    "args": {
      "External id": 217813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217813,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217813, "pid": 5, "tid": 7, "ts": 1716454224895116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765229, "dur": 9,
    "args": {
      "External id": 217813,
      "cbid": 211, "correlation": 217813
    }
  },
  {
    "ph": "s", "id": 217813, "pid": 76337, "tid": -914061504, "ts": 1716454224765229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224895167, "dur": 36,
    "args": {
      "External id": 217821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217821,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217821, "pid": 5, "tid": 7, "ts": 1716454224895167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765259, "dur": 9,
    "args": {
      "External id": 217821,
      "cbid": 211, "correlation": 217821
    }
  },
  {
    "ph": "s", "id": 217821, "pid": 76337, "tid": -914061504, "ts": 1716454224765259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224895204, "dur": 53,
    "args": {
      "External id": 217841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217841,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 217841, "pid": 5, "tid": 7, "ts": 1716454224895204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765346, "dur": 12,
    "args": {
      "External id": 217841,
      "cbid": 211, "correlation": 217841
    }
  },
  {
    "ph": "s", "id": 217841, "pid": 76337, "tid": -914061504, "ts": 1716454224765346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224895259, "dur": 4,
    "args": {
      "External id": 217853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217853,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217853, "pid": 5, "tid": 7, "ts": 1716454224895259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765367, "dur": 6,
    "args": {
      "External id": 217853,
      "cbid": 211, "correlation": 217853
    }
  },
  {
    "ph": "s", "id": 217853, "pid": 76337, "tid": -914061504, "ts": 1716454224765367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224895264, "dur": 56,
    "args": {
      "External id": 217856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217856, "pid": 5, "tid": 7, "ts": 1716454224895264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765386, "dur": 7,
    "args": {
      "External id": 217856,
      "cbid": 211, "correlation": 217856
    }
  },
  {
    "ph": "s", "id": 217856, "pid": 76337, "tid": -914061504, "ts": 1716454224765386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765443, "dur": 0,
    "args": {
      "External id": 217867,
      "cbid": 317, "correlation": 217867
    }
  },
  {
    "ph": "f", "id": 217867, "pid": 76337, "tid": -914061504, "ts": 1716454224765443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765444, "dur": 0,
    "args": {
      "External id": 217868,
      "cbid": 203, "correlation": 217868
    }
  },
  {
    "ph": "f", "id": 217868, "pid": 76337, "tid": -914061504, "ts": 1716454224765444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765444, "dur": 0,
    "args": {
      "External id": 217869,
      "cbid": 205, "correlation": 217869
    }
  },
  {
    "ph": "f", "id": 217869, "pid": 76337, "tid": -914061504, "ts": 1716454224765444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765467, "dur": 1,
    "args": {
      "External id": 217873,
      "cbid": 251, "correlation": 217873
    }
  },
  {
    "ph": "f", "id": 217873, "pid": 76337, "tid": -914061504, "ts": 1716454224765467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765470, "dur": 0,
    "args": {
      "External id": 217874,
      "cbid": 251, "correlation": 217874
    }
  },
  {
    "ph": "f", "id": 217874, "pid": 76337, "tid": -914061504, "ts": 1716454224765470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765471, "dur": 0,
    "args": {
      "External id": 217875,
      "cbid": 251, "correlation": 217875
    }
  },
  {
    "ph": "f", "id": 217875, "pid": 76337, "tid": -914061504, "ts": 1716454224765471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765472, "dur": 0,
    "args": {
      "External id": 217876,
      "cbid": 251, "correlation": 217876
    }
  },
  {
    "ph": "f", "id": 217876, "pid": 76337, "tid": -914061504, "ts": 1716454224765472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765473, "dur": 0,
    "args": {
      "External id": 217877,
      "cbid": 251, "correlation": 217877
    }
  },
  {
    "ph": "f", "id": 217877, "pid": 76337, "tid": -914061504, "ts": 1716454224765473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765474, "dur": 0,
    "args": {
      "External id": 217878,
      "cbid": 251, "correlation": 217878
    }
  },
  {
    "ph": "f", "id": 217878, "pid": 76337, "tid": -914061504, "ts": 1716454224765474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765475, "dur": 0,
    "args": {
      "External id": 217879,
      "cbid": 251, "correlation": 217879
    }
  },
  {
    "ph": "f", "id": 217879, "pid": 76337, "tid": -914061504, "ts": 1716454224765475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765476, "dur": 0,
    "args": {
      "External id": 217880,
      "cbid": 251, "correlation": 217880
    }
  },
  {
    "ph": "f", "id": 217880, "pid": 76337, "tid": -914061504, "ts": 1716454224765476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765477, "dur": 0,
    "args": {
      "External id": 217881,
      "cbid": 251, "correlation": 217881
    }
  },
  {
    "ph": "f", "id": 217881, "pid": 76337, "tid": -914061504, "ts": 1716454224765477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224895322, "dur": 113,
    "args": {
      "External id": 217882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217882,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217882, "pid": 5, "tid": 7, "ts": 1716454224895322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765479, "dur": 12,
    "args": {
      "External id": 217882,
      "cbid": 211, "correlation": 217882
    }
  },
  {
    "ph": "s", "id": 217882, "pid": 76337, "tid": -914061504, "ts": 1716454224765479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224895436, "dur": 60,
    "args": {
      "External id": 217888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217888, "pid": 5, "tid": 7, "ts": 1716454224895436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765514, "dur": 9,
    "args": {
      "External id": 217888,
      "cbid": 211, "correlation": 217888
    }
  },
  {
    "ph": "s", "id": 217888, "pid": 76337, "tid": -914061504, "ts": 1716454224765514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224895498, "dur": 440,
    "args": {
      "External id": 217897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217897,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217897, "pid": 5, "tid": 7, "ts": 1716454224895498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765599, "dur": 14,
    "args": {
      "External id": 217897,
      "cbid": 211, "correlation": 217897
    }
  },
  {
    "ph": "s", "id": 217897, "pid": 76337, "tid": -914061504, "ts": 1716454224765599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224895940, "dur": 182,
    "args": {
      "External id": 217919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217919,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217919, "pid": 5, "tid": 7, "ts": 1716454224895940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765657, "dur": 10,
    "args": {
      "External id": 217919,
      "cbid": 211, "correlation": 217919
    }
  },
  {
    "ph": "s", "id": 217919, "pid": 76337, "tid": -914061504, "ts": 1716454224765657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765747, "dur": 1,
    "args": {
      "External id": 217930,
      "cbid": 251, "correlation": 217930
    }
  },
  {
    "ph": "f", "id": 217930, "pid": 76337, "tid": -914061504, "ts": 1716454224765747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224896123, "dur": 196,
    "args": {
      "External id": 217931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217931,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217931, "pid": 5, "tid": 7, "ts": 1716454224896123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765752, "dur": 14,
    "args": {
      "External id": 217931,
      "cbid": 211, "correlation": 217931
    }
  },
  {
    "ph": "s", "id": 217931, "pid": 76337, "tid": -914061504, "ts": 1716454224765752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765821, "dur": 1,
    "args": {
      "External id": 217942,
      "cbid": 251, "correlation": 217942
    }
  },
  {
    "ph": "f", "id": 217942, "pid": 76337, "tid": -914061504, "ts": 1716454224765821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224896321, "dur": 186,
    "args": {
      "External id": 217943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217943,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217943, "pid": 5, "tid": 7, "ts": 1716454224896321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765826, "dur": 11,
    "args": {
      "External id": 217943,
      "cbid": 211, "correlation": 217943
    }
  },
  {
    "ph": "s", "id": 217943, "pid": 76337, "tid": -914061504, "ts": 1716454224765826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765889, "dur": 1,
    "args": {
      "External id": 217954,
      "cbid": 251, "correlation": 217954
    }
  },
  {
    "ph": "f", "id": 217954, "pid": 76337, "tid": -914061504, "ts": 1716454224765889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224896509, "dur": 190,
    "args": {
      "External id": 217955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217955,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217955, "pid": 5, "tid": 7, "ts": 1716454224896509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765893, "dur": 11,
    "args": {
      "External id": 217955,
      "cbid": 211, "correlation": 217955
    }
  },
  {
    "ph": "s", "id": 217955, "pid": 76337, "tid": -914061504, "ts": 1716454224765893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224896700, "dur": 18763,
    "args": {
      "External id": 217976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217976,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 217976, "pid": 5, "tid": 7, "ts": 1716454224896700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224765982, "dur": 13,
    "args": {
      "External id": 217976,
      "cbid": 211, "correlation": 217976
    }
  },
  {
    "ph": "s", "id": 217976, "pid": 76337, "tid": -914061504, "ts": 1716454224765982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224766082, "dur": 1,
    "args": {
      "External id": 217994,
      "cbid": 251, "correlation": 217994
    }
  },
  {
    "ph": "f", "id": 217994, "pid": 76337, "tid": -914061504, "ts": 1716454224766082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224915465, "dur": 202,
    "args": {
      "External id": 217996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217996,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217996, "pid": 5, "tid": 7, "ts": 1716454224915465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224766088, "dur": 13,
    "args": {
      "External id": 217996,
      "cbid": 211, "correlation": 217996
    }
  },
  {
    "ph": "s", "id": 217996, "pid": 76337, "tid": -914061504, "ts": 1716454224766088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224915669, "dur": 67,
    "args": {
      "External id": 218004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218004,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218004, "pid": 5, "tid": 7, "ts": 1716454224915669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224766157, "dur": 12,
    "args": {
      "External id": 218004,
      "cbid": 211, "correlation": 218004
    }
  },
  {
    "ph": "s", "id": 218004, "pid": 76337, "tid": -914061504, "ts": 1716454224766157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224915737, "dur": 97,
    "args": {
      "External id": 218012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218012,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218012, "pid": 5, "tid": 7, "ts": 1716454224915737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224766196, "dur": 100,
    "args": {
      "External id": 218012,
      "cbid": 211, "correlation": 218012
    }
  },
  {
    "ph": "s", "id": 218012, "pid": 76337, "tid": -914061504, "ts": 1716454224766196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224915835, "dur": 53,
    "args": {
      "External id": 218023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218023, "pid": 5, "tid": 7, "ts": 1716454224915835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224766361, "dur": 1911,
    "args": {
      "External id": 218023,
      "cbid": 211, "correlation": 218023
    }
  },
  {
    "ph": "s", "id": 218023, "pid": 76337, "tid": -914061504, "ts": 1716454224766361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224915889, "dur": 93,
    "args": {
      "External id": 218045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218045,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218045, "pid": 5, "tid": 7, "ts": 1716454224915889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768294, "dur": 125,
    "args": {
      "External id": 218045,
      "cbid": 211, "correlation": 218045
    }
  },
  {
    "ph": "s", "id": 218045, "pid": 76337, "tid": -914061504, "ts": 1716454224768294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768499, "dur": 1,
    "args": {
      "External id": 218056,
      "cbid": 251, "correlation": 218056
    }
  },
  {
    "ph": "f", "id": 218056, "pid": 76337, "tid": -914061504, "ts": 1716454224768499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224915983, "dur": 104,
    "args": {
      "External id": 218057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218057,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 218057, "pid": 5, "tid": 7, "ts": 1716454224915983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768504, "dur": 14,
    "args": {
      "External id": 218057,
      "cbid": 211, "correlation": 218057
    }
  },
  {
    "ph": "s", "id": 218057, "pid": 76337, "tid": -914061504, "ts": 1716454224768504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768577, "dur": 1,
    "args": {
      "External id": 218068,
      "cbid": 251, "correlation": 218068
    }
  },
  {
    "ph": "f", "id": 218068, "pid": 76337, "tid": -914061504, "ts": 1716454224768577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768582, "dur": 0,
    "args": {
      "External id": 218069,
      "cbid": 251, "correlation": 218069
    }
  },
  {
    "ph": "f", "id": 218069, "pid": 76337, "tid": -914061504, "ts": 1716454224768582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224916088, "dur": 10,
    "args": {
      "External id": 218070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218070,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218070, "pid": 5, "tid": 7, "ts": 1716454224916088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768584, "dur": 13,
    "args": {
      "External id": 218070,
      "cbid": 211, "correlation": 218070
    }
  },
  {
    "ph": "s", "id": 218070, "pid": 76337, "tid": -914061504, "ts": 1716454224768584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224916100, "dur": 5,
    "args": {
      "External id": 218072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218072,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 218072, "pid": 5, "tid": 7, "ts": 1716454224916100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768598, "dur": 6,
    "args": {
      "External id": 218072,
      "cbid": 211, "correlation": 218072
    }
  },
  {
    "ph": "s", "id": 218072, "pid": 76337, "tid": -914061504, "ts": 1716454224768598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768659, "dur": 1,
    "args": {
      "External id": 218083,
      "cbid": 251, "correlation": 218083
    }
  },
  {
    "ph": "f", "id": 218083, "pid": 76337, "tid": -914061504, "ts": 1716454224768659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768662, "dur": 0,
    "args": {
      "External id": 218084,
      "cbid": 251, "correlation": 218084
    }
  },
  {
    "ph": "f", "id": 218084, "pid": 76337, "tid": -914061504, "ts": 1716454224768662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224916106, "dur": 6,
    "args": {
      "External id": 218085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218085,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218085, "pid": 5, "tid": 7, "ts": 1716454224916106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768664, "dur": 12,
    "args": {
      "External id": 218085,
      "cbid": 211, "correlation": 218085
    }
  },
  {
    "ph": "s", "id": 218085, "pid": 76337, "tid": -914061504, "ts": 1716454224768664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224916114, "dur": 3,
    "args": {
      "External id": 218087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218087,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 218087, "pid": 5, "tid": 7, "ts": 1716454224916114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768677, "dur": 6,
    "args": {
      "External id": 218087,
      "cbid": 211, "correlation": 218087
    }
  },
  {
    "ph": "s", "id": 218087, "pid": 76337, "tid": -914061504, "ts": 1716454224768677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224916118, "dur": 156,
    "args": {
      "External id": 218108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218108,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 218108, "pid": 5, "tid": 7, "ts": 1716454224916118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768752, "dur": 12,
    "args": {
      "External id": 218108,
      "cbid": 211, "correlation": 218108
    }
  },
  {
    "ph": "s", "id": 218108, "pid": 76337, "tid": -914061504, "ts": 1716454224768752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768847, "dur": 1,
    "args": {
      "External id": 218126,
      "cbid": 251, "correlation": 218126
    }
  },
  {
    "ph": "f", "id": 218126, "pid": 76337, "tid": -914061504, "ts": 1716454224768847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224916276, "dur": 106,
    "args": {
      "External id": 218128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218128,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 218128, "pid": 5, "tid": 7, "ts": 1716454224916276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768853, "dur": 13,
    "args": {
      "External id": 218128,
      "cbid": 211, "correlation": 218128
    }
  },
  {
    "ph": "s", "id": 218128, "pid": 76337, "tid": -914061504, "ts": 1716454224768853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224916383, "dur": 34,
    "args": {
      "External id": 218136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218136,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218136, "pid": 5, "tid": 7, "ts": 1716454224916383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768922, "dur": 12,
    "args": {
      "External id": 218136,
      "cbid": 211, "correlation": 218136
    }
  },
  {
    "ph": "s", "id": 218136, "pid": 76337, "tid": -914061504, "ts": 1716454224768922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224916419, "dur": 68,
    "args": {
      "External id": 218144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218144, "pid": 5, "tid": 7, "ts": 1716454224916419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224768964, "dur": 17,
    "args": {
      "External id": 218144,
      "cbid": 211, "correlation": 218144
    }
  },
  {
    "ph": "s", "id": 218144, "pid": 76337, "tid": -914061504, "ts": 1716454224768964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224916488, "dur": 92,
    "args": {
      "External id": 218166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218166,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218166, "pid": 5, "tid": 7, "ts": 1716454224916488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769025, "dur": 10,
    "args": {
      "External id": 218166,
      "cbid": 211, "correlation": 218166
    }
  },
  {
    "ph": "s", "id": 218166, "pid": 76337, "tid": -914061504, "ts": 1716454224769025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769112, "dur": 1,
    "args": {
      "External id": 218182,
      "cbid": 251, "correlation": 218182
    }
  },
  {
    "ph": "f", "id": 218182, "pid": 76337, "tid": -914061504, "ts": 1716454224769112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224916582, "dur": 584,
    "args": {
      "External id": 218184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218184,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 218184, "pid": 5, "tid": 7, "ts": 1716454224916582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769118, "dur": 12,
    "args": {
      "External id": 218184,
      "cbid": 211, "correlation": 218184
    }
  },
  {
    "ph": "s", "id": 218184, "pid": 76337, "tid": -914061504, "ts": 1716454224769118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224917167, "dur": 246,
    "args": {
      "External id": 218192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218192, "pid": 5, "tid": 7, "ts": 1716454224917167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769186, "dur": 13,
    "args": {
      "External id": 218192,
      "cbid": 211, "correlation": 218192
    }
  },
  {
    "ph": "s", "id": 218192, "pid": 76337, "tid": -914061504, "ts": 1716454224769186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224917414, "dur": 253,
    "args": {
      "External id": 218200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218200, "pid": 5, "tid": 7, "ts": 1716454224917414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769220, "dur": 8,
    "args": {
      "External id": 218200,
      "cbid": 211, "correlation": 218200
    }
  },
  {
    "ph": "s", "id": 218200, "pid": 76337, "tid": -914061504, "ts": 1716454224769220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769303, "dur": 2,
    "args": {
      "External id": 218216,
      "cbid": 251, "correlation": 218216
    }
  },
  {
    "ph": "f", "id": 218216, "pid": 76337, "tid": -914061504, "ts": 1716454224769303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769308, "dur": 0,
    "args": {
      "External id": 218218,
      "cbid": 251, "correlation": 218218
    }
  },
  {
    "ph": "f", "id": 218218, "pid": 76337, "tid": -914061504, "ts": 1716454224769308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224917669, "dur": 360,
    "args": {
      "External id": 218219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218219,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218219, "pid": 5, "tid": 7, "ts": 1716454224917669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769311, "dur": 14,
    "args": {
      "External id": 218219,
      "cbid": 211, "correlation": 218219
    }
  },
  {
    "ph": "s", "id": 218219, "pid": 76337, "tid": -914061504, "ts": 1716454224769311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224918030, "dur": 50,
    "args": {
      "External id": 218227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218227,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218227, "pid": 5, "tid": 7, "ts": 1716454224918030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769356, "dur": 175,
    "args": {
      "External id": 218227,
      "cbid": 211, "correlation": 218227
    }
  },
  {
    "ph": "s", "id": 218227, "pid": 76337, "tid": -914061504, "ts": 1716454224769356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224918081, "dur": 159,
    "args": {
      "External id": 218238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218238, "pid": 5, "tid": 7, "ts": 1716454224918081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769590, "dur": 67,
    "args": {
      "External id": 218238,
      "cbid": 211, "correlation": 218238
    }
  },
  {
    "ph": "s", "id": 218238, "pid": 76337, "tid": -914061504, "ts": 1716454224769590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769710, "dur": 0,
    "args": {
      "External id": 218250,
      "cbid": 317, "correlation": 218250
    }
  },
  {
    "ph": "f", "id": 218250, "pid": 76337, "tid": -914061504, "ts": 1716454224769710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769711, "dur": 0,
    "args": {
      "External id": 218251,
      "cbid": 203, "correlation": 218251
    }
  },
  {
    "ph": "f", "id": 218251, "pid": 76337, "tid": -914061504, "ts": 1716454224769711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769711, "dur": 0,
    "args": {
      "External id": 218252,
      "cbid": 205, "correlation": 218252
    }
  },
  {
    "ph": "f", "id": 218252, "pid": 76337, "tid": -914061504, "ts": 1716454224769711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769737, "dur": 1,
    "args": {
      "External id": 218256,
      "cbid": 251, "correlation": 218256
    }
  },
  {
    "ph": "f", "id": 218256, "pid": 76337, "tid": -914061504, "ts": 1716454224769737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769739, "dur": 0,
    "args": {
      "External id": 218257,
      "cbid": 251, "correlation": 218257
    }
  },
  {
    "ph": "f", "id": 218257, "pid": 76337, "tid": -914061504, "ts": 1716454224769739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769740, "dur": 0,
    "args": {
      "External id": 218258,
      "cbid": 251, "correlation": 218258
    }
  },
  {
    "ph": "f", "id": 218258, "pid": 76337, "tid": -914061504, "ts": 1716454224769740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769740, "dur": 0,
    "args": {
      "External id": 218259,
      "cbid": 251, "correlation": 218259
    }
  },
  {
    "ph": "f", "id": 218259, "pid": 76337, "tid": -914061504, "ts": 1716454224769740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769741, "dur": 0,
    "args": {
      "External id": 218260,
      "cbid": 251, "correlation": 218260
    }
  },
  {
    "ph": "f", "id": 218260, "pid": 76337, "tid": -914061504, "ts": 1716454224769741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769742, "dur": 0,
    "args": {
      "External id": 218261,
      "cbid": 251, "correlation": 218261
    }
  },
  {
    "ph": "f", "id": 218261, "pid": 76337, "tid": -914061504, "ts": 1716454224769742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769742, "dur": 0,
    "args": {
      "External id": 218262,
      "cbid": 251, "correlation": 218262
    }
  },
  {
    "ph": "f", "id": 218262, "pid": 76337, "tid": -914061504, "ts": 1716454224769742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769743, "dur": 0,
    "args": {
      "External id": 218263,
      "cbid": 251, "correlation": 218263
    }
  },
  {
    "ph": "f", "id": 218263, "pid": 76337, "tid": -914061504, "ts": 1716454224769743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769744, "dur": 0,
    "args": {
      "External id": 218264,
      "cbid": 251, "correlation": 218264
    }
  },
  {
    "ph": "f", "id": 218264, "pid": 76337, "tid": -914061504, "ts": 1716454224769744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224918241, "dur": 117,
    "args": {
      "External id": 218265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218265,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 218265, "pid": 5, "tid": 7, "ts": 1716454224918241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769746, "dur": 38,
    "args": {
      "External id": 218265,
      "cbid": 211, "correlation": 218265
    }
  },
  {
    "ph": "s", "id": 218265, "pid": 76337, "tid": -914061504, "ts": 1716454224769746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224918360, "dur": 59,
    "args": {
      "External id": 218271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218271, "pid": 5, "tid": 7, "ts": 1716454224918360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224769807, "dur": 285,
    "args": {
      "External id": 218271,
      "cbid": 211, "correlation": 218271
    }
  },
  {
    "ph": "s", "id": 218271, "pid": 76337, "tid": -914061504, "ts": 1716454224769807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224918421, "dur": 49,
    "args": {
      "External id": 218279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218279,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218279, "pid": 5, "tid": 7, "ts": 1716454224918421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770116, "dur": 10,
    "args": {
      "External id": 218279,
      "cbid": 211, "correlation": 218279
    }
  },
  {
    "ph": "s", "id": 218279, "pid": 76337, "tid": -914061504, "ts": 1716454224770116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224918471, "dur": 52,
    "args": {
      "External id": 218299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218299,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 218299, "pid": 5, "tid": 7, "ts": 1716454224918471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770202, "dur": 11,
    "args": {
      "External id": 218299,
      "cbid": 211, "correlation": 218299
    }
  },
  {
    "ph": "s", "id": 218299, "pid": 76337, "tid": -914061504, "ts": 1716454224770202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224918524, "dur": 4,
    "args": {
      "External id": 218311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218311,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 218311, "pid": 5, "tid": 7, "ts": 1716454224918524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770223, "dur": 7,
    "args": {
      "External id": 218311,
      "cbid": 211, "correlation": 218311
    }
  },
  {
    "ph": "s", "id": 218311, "pid": 76337, "tid": -914061504, "ts": 1716454224770223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224918530, "dur": 57,
    "args": {
      "External id": 218314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218314, "pid": 5, "tid": 7, "ts": 1716454224918530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770242, "dur": 107,
    "args": {
      "External id": 218314,
      "cbid": 211, "correlation": 218314
    }
  },
  {
    "ph": "s", "id": 218314, "pid": 76337, "tid": -914061504, "ts": 1716454224770242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224918588, "dur": 37,
    "args": {
      "External id": 218323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218323,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218323, "pid": 5, "tid": 7, "ts": 1716454224918588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770388, "dur": 11,
    "args": {
      "External id": 218323,
      "cbid": 211, "correlation": 218323
    }
  },
  {
    "ph": "s", "id": 218323, "pid": 76337, "tid": -914061504, "ts": 1716454224770388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770443, "dur": 0,
    "args": {
      "External id": 218333,
      "cbid": 317, "correlation": 218333
    }
  },
  {
    "ph": "f", "id": 218333, "pid": 76337, "tid": -914061504, "ts": 1716454224770443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770444, "dur": 0,
    "args": {
      "External id": 218334,
      "cbid": 203, "correlation": 218334
    }
  },
  {
    "ph": "f", "id": 218334, "pid": 76337, "tid": -914061504, "ts": 1716454224770444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770445, "dur": 0,
    "args": {
      "External id": 218335,
      "cbid": 205, "correlation": 218335
    }
  },
  {
    "ph": "f", "id": 218335, "pid": 76337, "tid": -914061504, "ts": 1716454224770445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224918626, "dur": 42,
    "args": {
      "External id": 218339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218339,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218339, "pid": 5, "tid": 7, "ts": 1716454224918626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770461, "dur": 12,
    "args": {
      "External id": 218339,
      "cbid": 211, "correlation": 218339
    }
  },
  {
    "ph": "s", "id": 218339, "pid": 76337, "tid": -914061504, "ts": 1716454224770461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224918669, "dur": 3,
    "args": {
      "External id": 218341,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218341,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.413793103448276
    }
  },
  {
    "ph": "f", "id": 218341, "pid": 5, "tid": 7, "ts": 1716454224918669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770476, "dur": 26,
    "args": {
      "External id": 218341,
      "cbid": 51, "correlation": 218341
    }
  },
  {
    "ph": "s", "id": 218341, "pid": 76337, "tid": -914061504, "ts": 1716454224770476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770507, "dur": 2,
    "args": {
      "External id": 218343,
      "cbid": 200, "correlation": 218343
    }
  },
  {
    "ph": "f", "id": 218343, "pid": 76337, "tid": -914061504, "ts": 1716454224770507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770510, "dur": 0,
    "args": {
      "External id": 218344,
      "cbid": 200, "correlation": 218344
    }
  },
  {
    "ph": "f", "id": 218344, "pid": 76337, "tid": -914061504, "ts": 1716454224770510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770511, "dur": 0,
    "args": {
      "External id": 218345,
      "cbid": 200, "correlation": 218345
    }
  },
  {
    "ph": "f", "id": 218345, "pid": 76337, "tid": -914061504, "ts": 1716454224770511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770511, "dur": 0,
    "args": {
      "External id": 218346,
      "cbid": 200, "correlation": 218346
    }
  },
  {
    "ph": "f", "id": 218346, "pid": 76337, "tid": -914061504, "ts": 1716454224770511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770512, "dur": 4,
    "args": {
      "External id": 218347,
      "cbid": 15, "correlation": 218347
    }
  },
  {
    "ph": "f", "id": 218347, "pid": 76337, "tid": -914061504, "ts": 1716454224770512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770517, "dur": 1,
    "args": {
      "External id": 218348,
      "cbid": 251, "correlation": 218348
    }
  },
  {
    "ph": "f", "id": 218348, "pid": 76337, "tid": -914061504, "ts": 1716454224770517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224918674, "dur": 23,
    "args": {
      "External id": 218349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218349,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218349, "pid": 5, "tid": 7, "ts": 1716454224918674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770520, "dur": 9,
    "args": {
      "External id": 218349,
      "cbid": 211, "correlation": 218349
    }
  },
  {
    "ph": "s", "id": 218349, "pid": 76337, "tid": -914061504, "ts": 1716454224770520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224918698, "dur": 4,
    "args": {
      "External id": 218351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218351,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 218351, "pid": 5, "tid": 7, "ts": 1716454224918698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770536, "dur": 5,
    "args": {
      "External id": 218351,
      "cbid": 211, "correlation": 218351
    }
  },
  {
    "ph": "s", "id": 218351, "pid": 76337, "tid": -914061504, "ts": 1716454224770536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770546, "dur": 0,
    "args": {
      "External id": 218352,
      "cbid": 51, "correlation": 218352
    }
  },
  {
    "ph": "s", "id": 218352, "pid": 76337, "tid": -914061504, "ts": 1716454224770546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224918704, "dur": 190,
    "args": {
      "External id": 218353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218353,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218353, "pid": 5, "tid": 7, "ts": 1716454224918704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770547, "dur": 197,
    "args": {
      "External id": 218353,
      "cbid": 211, "correlation": 218353
    }
  },
  {
    "ph": "s", "id": 218353, "pid": 76337, "tid": -914061504, "ts": 1716454224770547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224918896, "dur": 6,
    "args": {
      "External id": 218354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218354,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218354, "pid": 5, "tid": 7, "ts": 1716454224918896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770747, "dur": 5,
    "args": {
      "External id": 218354,
      "cbid": 211, "correlation": 218354
    }
  },
  {
    "ph": "s", "id": 218354, "pid": 76337, "tid": -914061504, "ts": 1716454224770747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224918904, "dur": 5,
    "args": {
      "External id": 218360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 218360, "pid": 5, "tid": 7, "ts": 1716454224918904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224770777, "dur": 9,
    "args": {
      "External id": 218360,
      "cbid": 211, "correlation": 218360
    }
  },
  {
    "ph": "s", "id": 218360, "pid": 76337, "tid": -914061504, "ts": 1716454224770777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224918910, "dur": 3,
    "args": {
      "External id": 218368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218368,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218368, "pid": 5, "tid": 7, "ts": 1716454224918910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224772616, "dur": 15,
    "args": {
      "External id": 218368,
      "cbid": 211, "correlation": 218368
    }
  },
  {
    "ph": "s", "id": 218368, "pid": 76337, "tid": -914061504, "ts": 1716454224772616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224918914, "dur": 3,
    "args": {
      "External id": 218376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218376,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218376, "pid": 5, "tid": 7, "ts": 1716454224918914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224772659, "dur": 10,
    "args": {
      "External id": 218376,
      "cbid": 211, "correlation": 218376
    }
  },
  {
    "ph": "s", "id": 218376, "pid": 76337, "tid": -914061504, "ts": 1716454224772659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224918919, "dur": 3,
    "args": {
      "External id": 218384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218384,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218384, "pid": 5, "tid": 7, "ts": 1716454224918919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224772685, "dur": 8,
    "args": {
      "External id": 218384,
      "cbid": 211, "correlation": 218384
    }
  },
  {
    "ph": "s", "id": 218384, "pid": 76337, "tid": -914061504, "ts": 1716454224772685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224918922, "dur": 3,
    "args": {
      "External id": 218393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218393,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218393, "pid": 5, "tid": 7, "ts": 1716454224918922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224772859, "dur": 14,
    "args": {
      "External id": 218393,
      "cbid": 211, "correlation": 218393
    }
  },
  {
    "ph": "s", "id": 218393, "pid": 76337, "tid": -914061504, "ts": 1716454224772859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224918927, "dur": 3,
    "args": {
      "External id": 218402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218402,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218402, "pid": 5, "tid": 7, "ts": 1716454224918927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224772888, "dur": 8,
    "args": {
      "External id": 218402,
      "cbid": 211, "correlation": 218402
    }
  },
  {
    "ph": "s", "id": 218402, "pid": 76337, "tid": -914061504, "ts": 1716454224772888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224918931, "dur": 3,
    "args": {
      "External id": 218410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218410,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218410, "pid": 5, "tid": 7, "ts": 1716454224918931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224772914, "dur": 8,
    "args": {
      "External id": 218410,
      "cbid": 211, "correlation": 218410
    }
  },
  {
    "ph": "s", "id": 218410, "pid": 76337, "tid": -914061504, "ts": 1716454224772914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224918936, "dur": 3,
    "args": {
      "External id": 218418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218418,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218418, "pid": 5, "tid": 7, "ts": 1716454224918936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224773180, "dur": 16,
    "args": {
      "External id": 218418,
      "cbid": 211, "correlation": 218418
    }
  },
  {
    "ph": "s", "id": 218418, "pid": 76337, "tid": -914061504, "ts": 1716454224773180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224918940, "dur": 3,
    "args": {
      "External id": 218426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218426,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218426, "pid": 5, "tid": 7, "ts": 1716454224918940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224773212, "dur": 8,
    "args": {
      "External id": 218426,
      "cbid": 211, "correlation": 218426
    }
  },
  {
    "ph": "s", "id": 218426, "pid": 76337, "tid": -914061504, "ts": 1716454224773212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224918945, "dur": 1,
    "args": {
      "External id": 218436,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218436,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 218436, "pid": 5, "tid": 7, "ts": 1716454224918945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224773278, "dur": 37,
    "args": {
      "External id": 218436,
      "cbid": 41, "correlation": 218436
    }
  },
  {
    "ph": "s", "id": 218436, "pid": 76337, "tid": -914061504, "ts": 1716454224773278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224773316, "dur": 145646,
    "args": {
      "External id": 218437,
      "cbid": 131, "correlation": 218437
    }
  },
  {
    "ph": "f", "id": 218437, "pid": 76337, "tid": -914061504, "ts": 1716454224773316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224919185, "dur": 3,
    "args": {
      "External id": 218445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218445, "pid": 5, "tid": 7, "ts": 1716454224919185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919154, "dur": 33,
    "args": {
      "External id": 218445,
      "cbid": 211, "correlation": 218445
    }
  },
  {
    "ph": "s", "id": 218445, "pid": 76337, "tid": -914061504, "ts": 1716454224919154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919292, "dur": 3,
    "args": {
      "External id": 218454,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218454,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0021929824561403508
    }
  },
  {
    "ph": "f", "id": 218454, "pid": 5, "tid": 7, "ts": 1716454224919292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919254, "dur": 39,
    "args": {
      "External id": 218454,
      "cbid": 41, "correlation": 218454
    }
  },
  {
    "ph": "s", "id": 218454, "pid": 76337, "tid": -914061504, "ts": 1716454224919254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224919400, "dur": 4,
    "args": {
      "External id": 218464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218464,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218464, "pid": 5, "tid": 7, "ts": 1716454224919400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919384, "dur": 17,
    "args": {
      "External id": 218464,
      "cbid": 211, "correlation": 218464
    }
  },
  {
    "ph": "s", "id": 218464, "pid": 76337, "tid": -914061504, "ts": 1716454224919384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919483, "dur": 1,
    "args": {
      "External id": 218474,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218474,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 218474, "pid": 5, "tid": 7, "ts": 1716454224919483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919464, "dur": 17,
    "args": {
      "External id": 218474,
      "cbid": 41, "correlation": 218474
    }
  },
  {
    "ph": "s", "id": 218474, "pid": 76337, "tid": -914061504, "ts": 1716454224919464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919482, "dur": 8,
    "args": {
      "External id": 218475,
      "cbid": 131, "correlation": 218475
    }
  },
  {
    "ph": "f", "id": 218475, "pid": 76337, "tid": -914061504, "ts": 1716454224919482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919568, "dur": 3,
    "args": {
      "External id": 218482,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218482,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.415841584158414
    }
  },
  {
    "ph": "f", "id": 218482, "pid": 5, "tid": 7, "ts": 1716454224919568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919537, "dur": 30,
    "args": {
      "External id": 218482,
      "cbid": 41, "correlation": 218482
    }
  },
  {
    "ph": "s", "id": 218482, "pid": 76337, "tid": -914061504, "ts": 1716454224919537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919661, "dur": 3,
    "args": {
      "External id": 218501,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218501,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 218501, "pid": 5, "tid": 7, "ts": 1716454224919661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919642, "dur": 17,
    "args": {
      "External id": 218501,
      "cbid": 41, "correlation": 218501
    }
  },
  {
    "ph": "s", "id": 218501, "pid": 76337, "tid": -914061504, "ts": 1716454224919642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224919699, "dur": 3,
    "args": {
      "External id": 218507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218507, "pid": 5, "tid": 7, "ts": 1716454224919699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919688, "dur": 11,
    "args": {
      "External id": 218507,
      "cbid": 211, "correlation": 218507
    }
  },
  {
    "ph": "s", "id": 218507, "pid": 76337, "tid": -914061504, "ts": 1716454224919688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224919714, "dur": 6,
    "args": {
      "External id": 218509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218509,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 218509, "pid": 5, "tid": 7, "ts": 1716454224919714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919704, "dur": 9,
    "args": {
      "External id": 218509,
      "cbid": 211, "correlation": 218509
    }
  },
  {
    "ph": "s", "id": 218509, "pid": 76337, "tid": -914061504, "ts": 1716454224919704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454224919722, "dur": 3,
    "args": {
      "External id": 218511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218511, "pid": 5, "tid": 7, "ts": 1716454224919722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919714, "dur": 6,
    "args": {
      "External id": 218511,
      "cbid": 211, "correlation": 218511
    }
  },
  {
    "ph": "s", "id": 218511, "pid": 76337, "tid": -914061504, "ts": 1716454224919714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919756, "dur": 2,
    "args": {
      "External id": 218519,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218519,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 218519, "pid": 5, "tid": 7, "ts": 1716454224919756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919742, "dur": 13,
    "args": {
      "External id": 218519,
      "cbid": 41, "correlation": 218519
    }
  },
  {
    "ph": "s", "id": 218519, "pid": 76337, "tid": -914061504, "ts": 1716454224919742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224919811, "dur": 3,
    "args": {
      "External id": 218533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218533, "pid": 5, "tid": 7, "ts": 1716454224919811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919799, "dur": 14,
    "args": {
      "External id": 218533,
      "cbid": 211, "correlation": 218533
    }
  },
  {
    "ph": "s", "id": 218533, "pid": 76337, "tid": -914061504, "ts": 1716454224919799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224919832, "dur": 2,
    "args": {
      "External id": 218547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218547,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218547, "pid": 5, "tid": 7, "ts": 1716454224919832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919824, "dur": 7,
    "args": {
      "External id": 218547,
      "cbid": 211, "correlation": 218547
    }
  },
  {
    "ph": "s", "id": 218547, "pid": 76337, "tid": -914061504, "ts": 1716454224919824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224919869, "dur": 6,
    "args": {
      "External id": 218554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218554,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218554, "pid": 5, "tid": 7, "ts": 1716454224919869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919858, "dur": 12,
    "args": {
      "External id": 218554,
      "cbid": 211, "correlation": 218554
    }
  },
  {
    "ph": "s", "id": 218554, "pid": 76337, "tid": -914061504, "ts": 1716454224919858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224919882, "dur": 6,
    "args": {
      "External id": 218557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218557,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218557, "pid": 5, "tid": 7, "ts": 1716454224919882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919873, "dur": 7,
    "args": {
      "External id": 218557,
      "cbid": 211, "correlation": 218557
    }
  },
  {
    "ph": "s", "id": 218557, "pid": 76337, "tid": -914061504, "ts": 1716454224919873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224919890, "dur": 3,
    "args": {
      "External id": 218559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218559,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218559, "pid": 5, "tid": 7, "ts": 1716454224919890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919882, "dur": 7,
    "args": {
      "External id": 218559,
      "cbid": 211, "correlation": 218559
    }
  },
  {
    "ph": "s", "id": 218559, "pid": 76337, "tid": -914061504, "ts": 1716454224919882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919910, "dur": 2,
    "args": {
      "External id": 218562,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218562,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 218562, "pid": 5, "tid": 7, "ts": 1716454224919910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919898, "dur": 12,
    "args": {
      "External id": 218562,
      "cbid": 41, "correlation": 218562
    }
  },
  {
    "ph": "s", "id": 218562, "pid": 76337, "tid": -914061504, "ts": 1716454224919898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224919962, "dur": 4,
    "args": {
      "External id": 218578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218578,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218578, "pid": 5, "tid": 7, "ts": 1716454224919962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919949, "dur": 14,
    "args": {
      "External id": 218578,
      "cbid": 211, "correlation": 218578
    }
  },
  {
    "ph": "s", "id": 218578, "pid": 76337, "tid": -914061504, "ts": 1716454224919949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224919993, "dur": 3,
    "args": {
      "External id": 218583,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218583,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003125
    }
  },
  {
    "ph": "f", "id": 218583, "pid": 5, "tid": 7, "ts": 1716454224919993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224919968, "dur": 25,
    "args": {
      "External id": 218583,
      "cbid": 41, "correlation": 218583
    }
  },
  {
    "ph": "s", "id": 218583, "pid": 76337, "tid": -914061504, "ts": 1716454224919968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224920021, "dur": 1,
    "args": {
      "External id": 218589,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218589,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 218589, "pid": 5, "tid": 7, "ts": 1716454224920021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920003, "dur": 27,
    "args": {
      "External id": 218589,
      "cbid": 41, "correlation": 218589
    }
  },
  {
    "ph": "s", "id": 218589, "pid": 76337, "tid": -914061504, "ts": 1716454224920003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920031, "dur": 3,
    "args": {
      "External id": 218590,
      "cbid": 131, "correlation": 218590
    }
  },
  {
    "ph": "f", "id": 218590, "pid": 76337, "tid": -914061504, "ts": 1716454224920031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224920086, "dur": 3,
    "args": {
      "External id": 218598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218598,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218598, "pid": 5, "tid": 7, "ts": 1716454224920086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920072, "dur": 14,
    "args": {
      "External id": 218598,
      "cbid": 211, "correlation": 218598
    }
  },
  {
    "ph": "s", "id": 218598, "pid": 76337, "tid": -914061504, "ts": 1716454224920072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224920117, "dur": 3,
    "args": {
      "External id": 218608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218608,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218608, "pid": 5, "tid": 7, "ts": 1716454224920117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920108, "dur": 9,
    "args": {
      "External id": 218608,
      "cbid": 211, "correlation": 218608
    }
  },
  {
    "ph": "s", "id": 218608, "pid": 76337, "tid": -914061504, "ts": 1716454224920108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224920143, "dur": 3,
    "args": {
      "External id": 218617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218617,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218617, "pid": 5, "tid": 7, "ts": 1716454224920143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920134, "dur": 8,
    "args": {
      "External id": 218617,
      "cbid": 211, "correlation": 218617
    }
  },
  {
    "ph": "s", "id": 218617, "pid": 76337, "tid": -914061504, "ts": 1716454224920134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224920267, "dur": 12,
    "args": {
      "External id": 218627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218627,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218627, "pid": 5, "tid": 7, "ts": 1716454224920267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920251, "dur": 16,
    "args": {
      "External id": 218627,
      "cbid": 211, "correlation": 218627
    }
  },
  {
    "ph": "s", "id": 218627, "pid": 76337, "tid": -914061504, "ts": 1716454224920251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224920306, "dur": 3,
    "args": {
      "External id": 218635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218635,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218635, "pid": 5, "tid": 7, "ts": 1716454224920306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920297, "dur": 8,
    "args": {
      "External id": 218635,
      "cbid": 211, "correlation": 218635
    }
  },
  {
    "ph": "s", "id": 218635, "pid": 76337, "tid": -914061504, "ts": 1716454224920297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224920353, "dur": 11,
    "args": {
      "External id": 218645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218645,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218645, "pid": 5, "tid": 7, "ts": 1716454224920353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920342, "dur": 12,
    "args": {
      "External id": 218645,
      "cbid": 211, "correlation": 218645
    }
  },
  {
    "ph": "s", "id": 218645, "pid": 76337, "tid": -914061504, "ts": 1716454224920342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224920387, "dur": 11,
    "args": {
      "External id": 218653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218653,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218653, "pid": 5, "tid": 7, "ts": 1716454224920387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920377, "dur": 9,
    "args": {
      "External id": 218653,
      "cbid": 211, "correlation": 218653
    }
  },
  {
    "ph": "s", "id": 218653, "pid": 76337, "tid": -914061504, "ts": 1716454224920377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224920414, "dur": 3,
    "args": {
      "External id": 218662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218662, "pid": 5, "tid": 7, "ts": 1716454224920414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920404, "dur": 10,
    "args": {
      "External id": 218662,
      "cbid": 211, "correlation": 218662
    }
  },
  {
    "ph": "s", "id": 218662, "pid": 76337, "tid": -914061504, "ts": 1716454224920404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224920439, "dur": 5,
    "args": {
      "External id": 218671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218671,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218671, "pid": 5, "tid": 7, "ts": 1716454224920439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920430, "dur": 8,
    "args": {
      "External id": 218671,
      "cbid": 211, "correlation": 218671
    }
  },
  {
    "ph": "s", "id": 218671, "pid": 76337, "tid": -914061504, "ts": 1716454224920430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224920480, "dur": 8,
    "args": {
      "External id": 218681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218681,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218681, "pid": 5, "tid": 7, "ts": 1716454224920480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920469, "dur": 11,
    "args": {
      "External id": 218681,
      "cbid": 211, "correlation": 218681
    }
  },
  {
    "ph": "s", "id": 218681, "pid": 76337, "tid": -914061504, "ts": 1716454224920469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224920823, "dur": 3,
    "args": {
      "External id": 218690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218690,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218690, "pid": 5, "tid": 7, "ts": 1716454224920823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920807, "dur": 16,
    "args": {
      "External id": 218690,
      "cbid": 211, "correlation": 218690
    }
  },
  {
    "ph": "s", "id": 218690, "pid": 76337, "tid": -914061504, "ts": 1716454224920807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224920861, "dur": 3,
    "args": {
      "External id": 218698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218698,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218698, "pid": 5, "tid": 7, "ts": 1716454224920861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920847, "dur": 14,
    "args": {
      "External id": 218698,
      "cbid": 211, "correlation": 218698
    }
  },
  {
    "ph": "s", "id": 218698, "pid": 76337, "tid": -914061504, "ts": 1716454224920847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224920913, "dur": 1,
    "args": {
      "External id": 218708,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218708,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 218708, "pid": 5, "tid": 7, "ts": 1716454224920913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920898, "dur": 13,
    "args": {
      "External id": 218708,
      "cbid": 41, "correlation": 218708
    }
  },
  {
    "ph": "s", "id": 218708, "pid": 76337, "tid": -914061504, "ts": 1716454224920898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920912, "dur": 8,
    "args": {
      "External id": 218709,
      "cbid": 131, "correlation": 218709
    }
  },
  {
    "ph": "f", "id": 218709, "pid": 76337, "tid": -914061504, "ts": 1716454224920912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224921014, "dur": 2,
    "args": {
      "External id": 218717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218717, "pid": 5, "tid": 7, "ts": 1716454224921014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224920999, "dur": 14,
    "args": {
      "External id": 218717,
      "cbid": 211, "correlation": 218717
    }
  },
  {
    "ph": "s", "id": 218717, "pid": 76337, "tid": -914061504, "ts": 1716454224920999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224921087, "dur": 3,
    "args": {
      "External id": 218726,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218726,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002577319587628866
    }
  },
  {
    "ph": "f", "id": 218726, "pid": 5, "tid": 7, "ts": 1716454224921087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921069, "dur": 18,
    "args": {
      "External id": 218726,
      "cbid": 41, "correlation": 218726
    }
  },
  {
    "ph": "s", "id": 218726, "pid": 76337, "tid": -914061504, "ts": 1716454224921069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224921162, "dur": 3,
    "args": {
      "External id": 218736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218736,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 218736, "pid": 5, "tid": 7, "ts": 1716454224921162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921148, "dur": 13,
    "args": {
      "External id": 218736,
      "cbid": 211, "correlation": 218736
    }
  },
  {
    "ph": "s", "id": 218736, "pid": 76337, "tid": -914061504, "ts": 1716454224921148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224921215, "dur": 1,
    "args": {
      "External id": 218746,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218746,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 218746, "pid": 5, "tid": 7, "ts": 1716454224921215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921201, "dur": 12,
    "args": {
      "External id": 218746,
      "cbid": 41, "correlation": 218746
    }
  },
  {
    "ph": "s", "id": 218746, "pid": 76337, "tid": -914061504, "ts": 1716454224921201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921214, "dur": 8,
    "args": {
      "External id": 218747,
      "cbid": 131, "correlation": 218747
    }
  },
  {
    "ph": "f", "id": 218747, "pid": 76337, "tid": -914061504, "ts": 1716454224921214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224921280, "dur": 3,
    "args": {
      "External id": 218754,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218754,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.670103092783506
    }
  },
  {
    "ph": "f", "id": 218754, "pid": 5, "tid": 7, "ts": 1716454224921280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921257, "dur": 23,
    "args": {
      "External id": 218754,
      "cbid": 41, "correlation": 218754
    }
  },
  {
    "ph": "s", "id": 218754, "pid": 76337, "tid": -914061504, "ts": 1716454224921257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224921332, "dur": 1,
    "args": {
      "External id": 218765,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218765,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 218765, "pid": 5, "tid": 7, "ts": 1716454224921332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921317, "dur": 13,
    "args": {
      "External id": 218765,
      "cbid": 41, "correlation": 218765
    }
  },
  {
    "ph": "s", "id": 218765, "pid": 76337, "tid": -914061504, "ts": 1716454224921317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921331, "dur": 8,
    "args": {
      "External id": 218766,
      "cbid": 131, "correlation": 218766
    }
  },
  {
    "ph": "f", "id": 218766, "pid": 76337, "tid": -914061504, "ts": 1716454224921331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224921382, "dur": 3,
    "args": {
      "External id": 218774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218774,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218774, "pid": 5, "tid": 7, "ts": 1716454224921382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921369, "dur": 13,
    "args": {
      "External id": 218774,
      "cbid": 211, "correlation": 218774
    }
  },
  {
    "ph": "s", "id": 218774, "pid": 76337, "tid": -914061504, "ts": 1716454224921369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224921412, "dur": 3,
    "args": {
      "External id": 218784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218784,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218784, "pid": 5, "tid": 7, "ts": 1716454224921412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921403, "dur": 8,
    "args": {
      "External id": 218784,
      "cbid": 211, "correlation": 218784
    }
  },
  {
    "ph": "s", "id": 218784, "pid": 76337, "tid": -914061504, "ts": 1716454224921403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224921434, "dur": 3,
    "args": {
      "External id": 218793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218793,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218793, "pid": 5, "tid": 7, "ts": 1716454224921434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921425, "dur": 7,
    "args": {
      "External id": 218793,
      "cbid": 211, "correlation": 218793
    }
  },
  {
    "ph": "s", "id": 218793, "pid": 76337, "tid": -914061504, "ts": 1716454224921425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224921504, "dur": 6,
    "args": {
      "External id": 218801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218801, "pid": 5, "tid": 7, "ts": 1716454224921504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921489, "dur": 15,
    "args": {
      "External id": 218801,
      "cbid": 211, "correlation": 218801
    }
  },
  {
    "ph": "s", "id": 218801, "pid": 76337, "tid": -914061504, "ts": 1716454224921489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224921543, "dur": 3,
    "args": {
      "External id": 218810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218810,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218810, "pid": 5, "tid": 7, "ts": 1716454224921543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921533, "dur": 10,
    "args": {
      "External id": 218810,
      "cbid": 211, "correlation": 218810
    }
  },
  {
    "ph": "s", "id": 218810, "pid": 76337, "tid": -914061504, "ts": 1716454224921533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224921567, "dur": 3,
    "args": {
      "External id": 218819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218819,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218819, "pid": 5, "tid": 7, "ts": 1716454224921567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921558, "dur": 7,
    "args": {
      "External id": 218819,
      "cbid": 211, "correlation": 218819
    }
  },
  {
    "ph": "s", "id": 218819, "pid": 76337, "tid": -914061504, "ts": 1716454224921558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224921629, "dur": 3,
    "args": {
      "External id": 218827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218827,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 218827, "pid": 5, "tid": 7, "ts": 1716454224921629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921618, "dur": 10,
    "args": {
      "External id": 218827,
      "cbid": 211, "correlation": 218827
    }
  },
  {
    "ph": "s", "id": 218827, "pid": 76337, "tid": -914061504, "ts": 1716454224921618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224921687, "dur": 1,
    "args": {
      "External id": 218835,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218835,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 218835, "pid": 5, "tid": 7, "ts": 1716454224921687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921671, "dur": 26,
    "args": {
      "External id": 218835,
      "cbid": 41, "correlation": 218835
    }
  },
  {
    "ph": "s", "id": 218835, "pid": 76337, "tid": -914061504, "ts": 1716454224921671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921697, "dur": 3,
    "args": {
      "External id": 218836,
      "cbid": 131, "correlation": 218836
    }
  },
  {
    "ph": "f", "id": 218836, "pid": 76337, "tid": -914061504, "ts": 1716454224921697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224921758, "dur": 1,
    "args": {
      "External id": 218846,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218846,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 218846, "pid": 5, "tid": 7, "ts": 1716454224921758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921746, "dur": 10,
    "args": {
      "External id": 218846,
      "cbid": 41, "correlation": 218846
    }
  },
  {
    "ph": "s", "id": 218846, "pid": 76337, "tid": -914061504, "ts": 1716454224921746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921756, "dur": 8,
    "args": {
      "External id": 218847,
      "cbid": 131, "correlation": 218847
    }
  },
  {
    "ph": "f", "id": 218847, "pid": 76337, "tid": -914061504, "ts": 1716454224921756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224921814, "dur": 1,
    "args": {
      "External id": 218856,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218856,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 218856, "pid": 5, "tid": 7, "ts": 1716454224921814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921804, "dur": 8,
    "args": {
      "External id": 218856,
      "cbid": 41, "correlation": 218856
    }
  },
  {
    "ph": "s", "id": 218856, "pid": 76337, "tid": -914061504, "ts": 1716454224921804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921813, "dur": 8,
    "args": {
      "External id": 218857,
      "cbid": 131, "correlation": 218857
    }
  },
  {
    "ph": "f", "id": 218857, "pid": 76337, "tid": -914061504, "ts": 1716454224921813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224921886, "dur": 4,
    "args": {
      "External id": 218864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218864, "pid": 5, "tid": 7, "ts": 1716454224921886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921870, "dur": 18,
    "args": {
      "External id": 218864,
      "cbid": 211, "correlation": 218864
    }
  },
  {
    "ph": "s", "id": 218864, "pid": 76337, "tid": -914061504, "ts": 1716454224921870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454224921924, "dur": 4,
    "args": {
      "External id": 218884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218884,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218884, "pid": 5, "tid": 7, "ts": 1716454224921924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921912, "dur": 12,
    "args": {
      "External id": 218884,
      "cbid": 211, "correlation": 218884
    }
  },
  {
    "ph": "s", "id": 218884, "pid": 76337, "tid": -914061504, "ts": 1716454224921912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921926, "dur": 0,
    "args": {
      "External id": 218885,
      "cbid": 11, "correlation": 218885
    }
  },
  {
    "ph": "f", "id": 218885, "pid": 76337, "tid": -914061504, "ts": 1716454224921926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921926, "dur": 0,
    "args": {
      "External id": 218886,
      "cbid": 11, "correlation": 218886
    }
  },
  {
    "ph": "f", "id": 218886, "pid": 76337, "tid": -914061504, "ts": 1716454224921926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224921941, "dur": 1,
    "args": {
      "External id": 218889,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218889,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 218889, "pid": 5, "tid": 7, "ts": 1716454224921941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921928, "dur": 22,
    "args": {
      "External id": 218889,
      "cbid": 41, "correlation": 218889
    }
  },
  {
    "ph": "s", "id": 218889, "pid": 76337, "tid": -914061504, "ts": 1716454224921928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921951, "dur": 3,
    "args": {
      "External id": 218890,
      "cbid": 131, "correlation": 218890
    }
  },
  {
    "ph": "f", "id": 218890, "pid": 76337, "tid": -914061504, "ts": 1716454224921951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224921988, "dur": 3,
    "args": {
      "External id": 218914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218914, "pid": 5, "tid": 7, "ts": 1716454224921988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921969, "dur": 19,
    "args": {
      "External id": 218914,
      "cbid": 211, "correlation": 218914
    }
  },
  {
    "ph": "s", "id": 218914, "pid": 76337, "tid": -914061504, "ts": 1716454224921969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921989, "dur": 0,
    "args": {
      "External id": 218915,
      "cbid": 11, "correlation": 218915
    }
  },
  {
    "ph": "f", "id": 218915, "pid": 76337, "tid": -914061504, "ts": 1716454224921989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921989, "dur": 0,
    "args": {
      "External id": 218916,
      "cbid": 11, "correlation": 218916
    }
  },
  {
    "ph": "f", "id": 218916, "pid": 76337, "tid": -914061504, "ts": 1716454224921989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921991, "dur": 1,
    "args": {
      "External id": 218918,
      "cbid": 200, "correlation": 218918
    }
  },
  {
    "ph": "f", "id": 218918, "pid": 76337, "tid": -914061504, "ts": 1716454224921991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224922001, "dur": 4,
    "args": {
      "External id": 218920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218920,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218920, "pid": 5, "tid": 7, "ts": 1716454224922001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224921994, "dur": 8,
    "args": {
      "External id": 218920,
      "cbid": 211, "correlation": 218920
    }
  },
  {
    "ph": "s", "id": 218920, "pid": 76337, "tid": -914061504, "ts": 1716454224921994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922003, "dur": 0,
    "args": {
      "External id": 218921,
      "cbid": 11, "correlation": 218921
    }
  },
  {
    "ph": "f", "id": 218921, "pid": 76337, "tid": -914061504, "ts": 1716454224922003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922003, "dur": 0,
    "args": {
      "External id": 218922,
      "cbid": 11, "correlation": 218922
    }
  },
  {
    "ph": "f", "id": 218922, "pid": 76337, "tid": -914061504, "ts": 1716454224922003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224922042, "dur": 1,
    "args": {
      "External id": 218929,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218929,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 218929, "pid": 5, "tid": 7, "ts": 1716454224922042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922030, "dur": 20,
    "args": {
      "External id": 218929,
      "cbid": 41, "correlation": 218929
    }
  },
  {
    "ph": "s", "id": 218929, "pid": 76337, "tid": -914061504, "ts": 1716454224922030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922051, "dur": 3,
    "args": {
      "External id": 218930,
      "cbid": 131, "correlation": 218930
    }
  },
  {
    "ph": "f", "id": 218930, "pid": 76337, "tid": -914061504, "ts": 1716454224922051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224922101, "dur": 1,
    "args": {
      "External id": 218940,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 218940,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 218940, "pid": 5, "tid": 7, "ts": 1716454224922101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922090, "dur": 9,
    "args": {
      "External id": 218940,
      "cbid": 41, "correlation": 218940
    }
  },
  {
    "ph": "s", "id": 218940, "pid": 76337, "tid": -914061504, "ts": 1716454224922090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922099, "dur": 8,
    "args": {
      "External id": 218941,
      "cbid": 131, "correlation": 218941
    }
  },
  {
    "ph": "f", "id": 218941, "pid": 76337, "tid": -914061504, "ts": 1716454224922099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224922177, "dur": 5,
    "args": {
      "External id": 218948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218948,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218948, "pid": 5, "tid": 7, "ts": 1716454224922177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922159, "dur": 18,
    "args": {
      "External id": 218948,
      "cbid": 211, "correlation": 218948
    }
  },
  {
    "ph": "s", "id": 218948, "pid": 76337, "tid": -914061504, "ts": 1716454224922159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224922247, "dur": 3,
    "args": {
      "External id": 218957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218957,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218957, "pid": 5, "tid": 7, "ts": 1716454224922247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922234, "dur": 13,
    "args": {
      "External id": 218957,
      "cbid": 211, "correlation": 218957
    }
  },
  {
    "ph": "s", "id": 218957, "pid": 76337, "tid": -914061504, "ts": 1716454224922234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224922284, "dur": 3,
    "args": {
      "External id": 218965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218965,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218965, "pid": 5, "tid": 7, "ts": 1716454224922284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922274, "dur": 9,
    "args": {
      "External id": 218965,
      "cbid": 211, "correlation": 218965
    }
  },
  {
    "ph": "s", "id": 218965, "pid": 76337, "tid": -914061504, "ts": 1716454224922274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224922323, "dur": 4,
    "args": {
      "External id": 218973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218973,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218973, "pid": 5, "tid": 7, "ts": 1716454224922323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922306, "dur": 17,
    "args": {
      "External id": 218973,
      "cbid": 211, "correlation": 218973
    }
  },
  {
    "ph": "s", "id": 218973, "pid": 76337, "tid": -914061504, "ts": 1716454224922306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224922353, "dur": 4,
    "args": {
      "External id": 218981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218981,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218981, "pid": 5, "tid": 7, "ts": 1716454224922353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922343, "dur": 9,
    "args": {
      "External id": 218981,
      "cbid": 211, "correlation": 218981
    }
  },
  {
    "ph": "s", "id": 218981, "pid": 76337, "tid": -914061504, "ts": 1716454224922343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224922381, "dur": 3,
    "args": {
      "External id": 218989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218989,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218989, "pid": 5, "tid": 7, "ts": 1716454224922381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922372, "dur": 8,
    "args": {
      "External id": 218989,
      "cbid": 211, "correlation": 218989
    }
  },
  {
    "ph": "s", "id": 218989, "pid": 76337, "tid": -914061504, "ts": 1716454224922372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224922406, "dur": 4,
    "args": {
      "External id": 218997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 218997,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 218997, "pid": 5, "tid": 7, "ts": 1716454224922406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922396, "dur": 9,
    "args": {
      "External id": 218997,
      "cbid": 211, "correlation": 218997
    }
  },
  {
    "ph": "s", "id": 218997, "pid": 76337, "tid": -914061504, "ts": 1716454224922396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224922428, "dur": 4,
    "args": {
      "External id": 219005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219005, "pid": 5, "tid": 7, "ts": 1716454224922428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922419, "dur": 7,
    "args": {
      "External id": 219005,
      "cbid": 211, "correlation": 219005
    }
  },
  {
    "ph": "s", "id": 219005, "pid": 76337, "tid": -914061504, "ts": 1716454224922419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224922446, "dur": 4,
    "args": {
      "External id": 219013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219013, "pid": 5, "tid": 7, "ts": 1716454224922446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922438, "dur": 7,
    "args": {
      "External id": 219013,
      "cbid": 211, "correlation": 219013
    }
  },
  {
    "ph": "s", "id": 219013, "pid": 76337, "tid": -914061504, "ts": 1716454224922438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224922465, "dur": 3,
    "args": {
      "External id": 219021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219021,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219021, "pid": 5, "tid": 7, "ts": 1716454224922465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922458, "dur": 7,
    "args": {
      "External id": 219021,
      "cbid": 211, "correlation": 219021
    }
  },
  {
    "ph": "s", "id": 219021, "pid": 76337, "tid": -914061504, "ts": 1716454224922458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224922524, "dur": 3,
    "args": {
      "External id": 219029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219029,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219029, "pid": 5, "tid": 7, "ts": 1716454224922524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922513, "dur": 11,
    "args": {
      "External id": 219029,
      "cbid": 211, "correlation": 219029
    }
  },
  {
    "ph": "s", "id": 219029, "pid": 76337, "tid": -914061504, "ts": 1716454224922513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224922550, "dur": 4,
    "args": {
      "External id": 219037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219037,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219037, "pid": 5, "tid": 7, "ts": 1716454224922550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922540, "dur": 9,
    "args": {
      "External id": 219037,
      "cbid": 211, "correlation": 219037
    }
  },
  {
    "ph": "s", "id": 219037, "pid": 76337, "tid": -914061504, "ts": 1716454224922540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224922573, "dur": 4,
    "args": {
      "External id": 219045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219045, "pid": 5, "tid": 7, "ts": 1716454224922573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922564, "dur": 8,
    "args": {
      "External id": 219045,
      "cbid": 211, "correlation": 219045
    }
  },
  {
    "ph": "s", "id": 219045, "pid": 76337, "tid": -914061504, "ts": 1716454224922564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224922591, "dur": 3,
    "args": {
      "External id": 219053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219053,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219053, "pid": 5, "tid": 7, "ts": 1716454224922591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224922583, "dur": 7,
    "args": {
      "External id": 219053,
      "cbid": 211, "correlation": 219053
    }
  },
  {
    "ph": "s", "id": 219053, "pid": 76337, "tid": -914061504, "ts": 1716454224922583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224923021, "dur": 5,
    "args": {
      "External id": 219062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219062,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219062, "pid": 5, "tid": 7, "ts": 1716454224923021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923005, "dur": 17,
    "args": {
      "External id": 219062,
      "cbid": 211, "correlation": 219062
    }
  },
  {
    "ph": "s", "id": 219062, "pid": 76337, "tid": -914061504, "ts": 1716454224923005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224923060, "dur": 5,
    "args": {
      "External id": 219071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219071,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219071, "pid": 5, "tid": 7, "ts": 1716454224923060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923049, "dur": 10,
    "args": {
      "External id": 219071,
      "cbid": 211, "correlation": 219071
    }
  },
  {
    "ph": "s", "id": 219071, "pid": 76337, "tid": -914061504, "ts": 1716454224923049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224923190, "dur": 3,
    "args": {
      "External id": 219087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219087, "pid": 5, "tid": 7, "ts": 1716454224923190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923174, "dur": 18,
    "args": {
      "External id": 219087,
      "cbid": 211, "correlation": 219087
    }
  },
  {
    "ph": "s", "id": 219087, "pid": 76337, "tid": -914061504, "ts": 1716454224923174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923226, "dur": 3,
    "args": {
      "External id": 219095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219095,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219095, "pid": 5, "tid": 7, "ts": 1716454224923226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923217, "dur": 9,
    "args": {
      "External id": 219095,
      "cbid": 211, "correlation": 219095
    }
  },
  {
    "ph": "s", "id": 219095, "pid": 76337, "tid": -914061504, "ts": 1716454224923217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923264, "dur": 3,
    "args": {
      "External id": 219103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219103,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219103, "pid": 5, "tid": 7, "ts": 1716454224923264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923251, "dur": 11,
    "args": {
      "External id": 219103,
      "cbid": 211, "correlation": 219103
    }
  },
  {
    "ph": "s", "id": 219103, "pid": 76337, "tid": -914061504, "ts": 1716454224923251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923300, "dur": 4,
    "args": {
      "External id": 219111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219111,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219111, "pid": 5, "tid": 7, "ts": 1716454224923300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923287, "dur": 13,
    "args": {
      "External id": 219111,
      "cbid": 211, "correlation": 219111
    }
  },
  {
    "ph": "s", "id": 219111, "pid": 76337, "tid": -914061504, "ts": 1716454224923287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224923357, "dur": 4,
    "args": {
      "External id": 219123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219123,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219123, "pid": 5, "tid": 7, "ts": 1716454224923357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923344, "dur": 13,
    "args": {
      "External id": 219123,
      "cbid": 211, "correlation": 219123
    }
  },
  {
    "ph": "s", "id": 219123, "pid": 76337, "tid": -914061504, "ts": 1716454224923344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224923403, "dur": 4,
    "args": {
      "External id": 219134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219134, "pid": 5, "tid": 7, "ts": 1716454224923403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923391, "dur": 11,
    "args": {
      "External id": 219134,
      "cbid": 211, "correlation": 219134
    }
  },
  {
    "ph": "s", "id": 219134, "pid": 76337, "tid": -914061504, "ts": 1716454224923391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923433, "dur": 3,
    "args": {
      "External id": 219142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219142, "pid": 5, "tid": 7, "ts": 1716454224923433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923424, "dur": 9,
    "args": {
      "External id": 219142,
      "cbid": 211, "correlation": 219142
    }
  },
  {
    "ph": "s", "id": 219142, "pid": 76337, "tid": -914061504, "ts": 1716454224923424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923467, "dur": 5,
    "args": {
      "External id": 219150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219150,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219150, "pid": 5, "tid": 7, "ts": 1716454224923467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923457, "dur": 10,
    "args": {
      "External id": 219150,
      "cbid": 211, "correlation": 219150
    }
  },
  {
    "ph": "s", "id": 219150, "pid": 76337, "tid": -914061504, "ts": 1716454224923457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923495, "dur": 5,
    "args": {
      "External id": 219158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219158,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219158, "pid": 5, "tid": 7, "ts": 1716454224923495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923486, "dur": 9,
    "args": {
      "External id": 219158,
      "cbid": 211, "correlation": 219158
    }
  },
  {
    "ph": "s", "id": 219158, "pid": 76337, "tid": -914061504, "ts": 1716454224923486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224923525, "dur": 4,
    "args": {
      "External id": 219167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219167,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219167, "pid": 5, "tid": 7, "ts": 1716454224923525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923515, "dur": 9,
    "args": {
      "External id": 219167,
      "cbid": 211, "correlation": 219167
    }
  },
  {
    "ph": "s", "id": 219167, "pid": 76337, "tid": -914061504, "ts": 1716454224923515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224923586, "dur": 4,
    "args": {
      "External id": 219180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219180,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219180, "pid": 5, "tid": 7, "ts": 1716454224923586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923573, "dur": 14,
    "args": {
      "External id": 219180,
      "cbid": 211, "correlation": 219180
    }
  },
  {
    "ph": "s", "id": 219180, "pid": 76337, "tid": -914061504, "ts": 1716454224923573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224923627, "dur": 5,
    "args": {
      "External id": 219190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219190,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219190, "pid": 5, "tid": 7, "ts": 1716454224923627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923615, "dur": 11,
    "args": {
      "External id": 219190,
      "cbid": 211, "correlation": 219190
    }
  },
  {
    "ph": "s", "id": 219190, "pid": 76337, "tid": -914061504, "ts": 1716454224923615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923754, "dur": 5,
    "args": {
      "External id": 219207,
      "cbid": 251, "correlation": 219207
    }
  },
  {
    "ph": "f", "id": 219207, "pid": 76337, "tid": -914061504, "ts": 1716454224923754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454224923784, "dur": 11,
    "args": {
      "External id": 219209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219209,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219209, "pid": 5, "tid": 7, "ts": 1716454224923784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923769, "dur": 16,
    "args": {
      "External id": 219209,
      "cbid": 211, "correlation": 219209
    }
  },
  {
    "ph": "s", "id": 219209, "pid": 76337, "tid": -914061504, "ts": 1716454224923769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224923850, "dur": 3,
    "args": {
      "External id": 219217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219217,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219217, "pid": 5, "tid": 7, "ts": 1716454224923850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923837, "dur": 14,
    "args": {
      "External id": 219217,
      "cbid": 211, "correlation": 219217
    }
  },
  {
    "ph": "s", "id": 219217, "pid": 76337, "tid": -914061504, "ts": 1716454224923837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923911, "dur": 2,
    "args": {
      "External id": 219233,
      "cbid": 251, "correlation": 219233
    }
  },
  {
    "ph": "f", "id": 219233, "pid": 76337, "tid": -914061504, "ts": 1716454224923911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923917, "dur": 0,
    "args": {
      "External id": 219235,
      "cbid": 251, "correlation": 219235
    }
  },
  {
    "ph": "f", "id": 219235, "pid": 76337, "tid": -914061504, "ts": 1716454224923917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224923934, "dur": 13,
    "args": {
      "External id": 219236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219236,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219236, "pid": 5, "tid": 7, "ts": 1716454224923934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923920, "dur": 14,
    "args": {
      "External id": 219236,
      "cbid": 211, "correlation": 219236
    }
  },
  {
    "ph": "s", "id": 219236, "pid": 76337, "tid": -914061504, "ts": 1716454224923920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224923949, "dur": 5,
    "args": {
      "External id": 219238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219238,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219238, "pid": 5, "tid": 7, "ts": 1716454224923949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224923940, "dur": 8,
    "args": {
      "External id": 219238,
      "cbid": 211, "correlation": 219238
    }
  },
  {
    "ph": "s", "id": 219238, "pid": 76337, "tid": -914061504, "ts": 1716454224923940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924060, "dur": 1,
    "args": {
      "External id": 219248,
      "cbid": 317, "correlation": 219248
    }
  },
  {
    "ph": "f", "id": 219248, "pid": 76337, "tid": -914061504, "ts": 1716454224924060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924062, "dur": 1,
    "args": {
      "External id": 219249,
      "cbid": 203, "correlation": 219249
    }
  },
  {
    "ph": "f", "id": 219249, "pid": 76337, "tid": -914061504, "ts": 1716454224924062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924064, "dur": 1,
    "args": {
      "External id": 219250,
      "cbid": 205, "correlation": 219250
    }
  },
  {
    "ph": "f", "id": 219250, "pid": 76337, "tid": -914061504, "ts": 1716454224924064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224924121, "dur": 7,
    "args": {
      "External id": 219254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219254,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219254, "pid": 5, "tid": 7, "ts": 1716454224924121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924105, "dur": 15,
    "args": {
      "External id": 219254,
      "cbid": 211, "correlation": 219254
    }
  },
  {
    "ph": "s", "id": 219254, "pid": 76337, "tid": -914061504, "ts": 1716454224924105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224924131, "dur": 4,
    "args": {
      "External id": 219256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219256,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 219256, "pid": 5, "tid": 7, "ts": 1716454224924131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924124, "dur": 6,
    "args": {
      "External id": 219256,
      "cbid": 211, "correlation": 219256
    }
  },
  {
    "ph": "s", "id": 219256, "pid": 76337, "tid": -914061504, "ts": 1716454224924124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224924152, "dur": 4,
    "args": {
      "External id": 219258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219258, "pid": 5, "tid": 7, "ts": 1716454224924152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924142, "dur": 8,
    "args": {
      "External id": 219258,
      "cbid": 211, "correlation": 219258
    }
  },
  {
    "ph": "s", "id": 219258, "pid": 76337, "tid": -914061504, "ts": 1716454224924142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924157, "dur": 0,
    "args": {
      "External id": 219259,
      "cbid": 51, "correlation": 219259
    }
  },
  {
    "ph": "s", "id": 219259, "pid": 76337, "tid": -914061504, "ts": 1716454224924157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224924168, "dur": 84,
    "args": {
      "External id": 219260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219260,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219260, "pid": 5, "tid": 7, "ts": 1716454224924168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924158, "dur": 8,
    "args": {
      "External id": 219260,
      "cbid": 211, "correlation": 219260
    }
  },
  {
    "ph": "s", "id": 219260, "pid": 76337, "tid": -914061504, "ts": 1716454224924158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224924254, "dur": 60,
    "args": {
      "External id": 219265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219265, "pid": 5, "tid": 7, "ts": 1716454224924254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224924198, "dur": 10,
    "args": {
      "External id": 219265,
      "cbid": 211, "correlation": 219265
    }
  },
  {
    "ph": "s", "id": 219265, "pid": 76337, "tid": -914061504, "ts": 1716454224924198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224926047, "dur": 50,
    "args": {
      "External id": 219285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219285,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 219285, "pid": 5, "tid": 7, "ts": 1716454224926047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926028, "dur": 20,
    "args": {
      "External id": 219285,
      "cbid": 211, "correlation": 219285
    }
  },
  {
    "ph": "s", "id": 219285, "pid": 76337, "tid": -914061504, "ts": 1716454224926028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224926099, "dur": 5,
    "args": {
      "External id": 219297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219297,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219297, "pid": 5, "tid": 7, "ts": 1716454224926099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926062, "dur": 10,
    "args": {
      "External id": 219297,
      "cbid": 211, "correlation": 219297
    }
  },
  {
    "ph": "s", "id": 219297, "pid": 76337, "tid": -914061504, "ts": 1716454224926062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224926105, "dur": 57,
    "args": {
      "External id": 219300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219300,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219300, "pid": 5, "tid": 7, "ts": 1716454224926105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926092, "dur": 11,
    "args": {
      "External id": 219300,
      "cbid": 211, "correlation": 219300
    }
  },
  {
    "ph": "s", "id": 219300, "pid": 76337, "tid": -914061504, "ts": 1716454224926092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224926163, "dur": 37,
    "args": {
      "External id": 219309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219309,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219309, "pid": 5, "tid": 7, "ts": 1716454224926163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926141, "dur": 10,
    "args": {
      "External id": 219309,
      "cbid": 211, "correlation": 219309
    }
  },
  {
    "ph": "s", "id": 219309, "pid": 76337, "tid": -914061504, "ts": 1716454224926141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926199, "dur": 0,
    "args": {
      "External id": 219319,
      "cbid": 317, "correlation": 219319
    }
  },
  {
    "ph": "f", "id": 219319, "pid": 76337, "tid": -914061504, "ts": 1716454224926199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926200, "dur": 0,
    "args": {
      "External id": 219320,
      "cbid": 203, "correlation": 219320
    }
  },
  {
    "ph": "f", "id": 219320, "pid": 76337, "tid": -914061504, "ts": 1716454224926200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926201, "dur": 0,
    "args": {
      "External id": 219321,
      "cbid": 205, "correlation": 219321
    }
  },
  {
    "ph": "f", "id": 219321, "pid": 76337, "tid": -914061504, "ts": 1716454224926201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224926234, "dur": 40,
    "args": {
      "External id": 219325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219325,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219325, "pid": 5, "tid": 7, "ts": 1716454224926234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926220, "dur": 13,
    "args": {
      "External id": 219325,
      "cbid": 211, "correlation": 219325
    }
  },
  {
    "ph": "s", "id": 219325, "pid": 76337, "tid": -914061504, "ts": 1716454224926220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224926275, "dur": 14,
    "args": {
      "External id": 219327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219327,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219327, "pid": 5, "tid": 7, "ts": 1716454224926275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926236, "dur": 6,
    "args": {
      "External id": 219327,
      "cbid": 211, "correlation": 219327
    }
  },
  {
    "ph": "s", "id": 219327, "pid": 76337, "tid": -914061504, "ts": 1716454224926236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224926291, "dur": 3,
    "args": {
      "External id": 219329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219329,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219329, "pid": 5, "tid": 7, "ts": 1716454224926291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926248, "dur": 8,
    "args": {
      "External id": 219329,
      "cbid": 211, "correlation": 219329
    }
  },
  {
    "ph": "s", "id": 219329, "pid": 76337, "tid": -914061504, "ts": 1716454224926248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926260, "dur": 0,
    "args": {
      "External id": 219330,
      "cbid": 51, "correlation": 219330
    }
  },
  {
    "ph": "s", "id": 219330, "pid": 76337, "tid": -914061504, "ts": 1716454224926260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224926296, "dur": 707,
    "args": {
      "External id": 219331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219331,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219331, "pid": 5, "tid": 7, "ts": 1716454224926296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926261, "dur": 7,
    "args": {
      "External id": 219331,
      "cbid": 211, "correlation": 219331
    }
  },
  {
    "ph": "s", "id": 219331, "pid": 76337, "tid": -914061504, "ts": 1716454224926261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224927004, "dur": 59,
    "args": {
      "External id": 219336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219336,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219336, "pid": 5, "tid": 7, "ts": 1716454224927004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926292, "dur": 8,
    "args": {
      "External id": 219336,
      "cbid": 211, "correlation": 219336
    }
  },
  {
    "ph": "s", "id": 219336, "pid": 76337, "tid": -914061504, "ts": 1716454224926292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224927064, "dur": 4,
    "args": {
      "External id": 219344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219344,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219344, "pid": 5, "tid": 7, "ts": 1716454224927064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926336, "dur": 9,
    "args": {
      "External id": 219344,
      "cbid": 211, "correlation": 219344
    }
  },
  {
    "ph": "s", "id": 219344, "pid": 76337, "tid": -914061504, "ts": 1716454224926336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926401, "dur": 1,
    "args": {
      "External id": 219360,
      "cbid": 251, "correlation": 219360
    }
  },
  {
    "ph": "f", "id": 219360, "pid": 76337, "tid": -914061504, "ts": 1716454224926401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926407, "dur": 0,
    "args": {
      "External id": 219362,
      "cbid": 251, "correlation": 219362
    }
  },
  {
    "ph": "f", "id": 219362, "pid": 76337, "tid": -914061504, "ts": 1716454224926407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224927069, "dur": 9,
    "args": {
      "External id": 219363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219363,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 219363, "pid": 5, "tid": 7, "ts": 1716454224927069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926409, "dur": 12,
    "args": {
      "External id": 219363,
      "cbid": 211, "correlation": 219363
    }
  },
  {
    "ph": "s", "id": 219363, "pid": 76337, "tid": -914061504, "ts": 1716454224926409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224927079, "dur": 4,
    "args": {
      "External id": 219365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219365,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 219365, "pid": 5, "tid": 7, "ts": 1716454224927079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926422, "dur": 8,
    "args": {
      "External id": 219365,
      "cbid": 211, "correlation": 219365
    }
  },
  {
    "ph": "s", "id": 219365, "pid": 76337, "tid": -914061504, "ts": 1716454224926422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224927084, "dur": 55,
    "args": {
      "External id": 219375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219375,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219375, "pid": 5, "tid": 7, "ts": 1716454224927084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926485, "dur": 12,
    "args": {
      "External id": 219375,
      "cbid": 211, "correlation": 219375
    }
  },
  {
    "ph": "s", "id": 219375, "pid": 76337, "tid": -914061504, "ts": 1716454224926485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224927140, "dur": 52,
    "args": {
      "External id": 219395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219395,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 219395, "pid": 5, "tid": 7, "ts": 1716454224927140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926551, "dur": 10,
    "args": {
      "External id": 219395,
      "cbid": 211, "correlation": 219395
    }
  },
  {
    "ph": "s", "id": 219395, "pid": 76337, "tid": -914061504, "ts": 1716454224926551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224927194, "dur": 4,
    "args": {
      "External id": 219407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219407,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219407, "pid": 5, "tid": 7, "ts": 1716454224927194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926571, "dur": 7,
    "args": {
      "External id": 219407,
      "cbid": 211, "correlation": 219407
    }
  },
  {
    "ph": "s", "id": 219407, "pid": 76337, "tid": -914061504, "ts": 1716454224926571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224927199, "dur": 56,
    "args": {
      "External id": 219410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219410,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219410, "pid": 5, "tid": 7, "ts": 1716454224927199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926591, "dur": 6,
    "args": {
      "External id": 219410,
      "cbid": 211, "correlation": 219410
    }
  },
  {
    "ph": "s", "id": 219410, "pid": 76337, "tid": -914061504, "ts": 1716454224926591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224927256, "dur": 37,
    "args": {
      "External id": 219419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219419,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219419, "pid": 5, "tid": 7, "ts": 1716454224927256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926631, "dur": 9,
    "args": {
      "External id": 219419,
      "cbid": 211, "correlation": 219419
    }
  },
  {
    "ph": "s", "id": 219419, "pid": 76337, "tid": -914061504, "ts": 1716454224926631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926699, "dur": 0,
    "args": {
      "External id": 219429,
      "cbid": 317, "correlation": 219429
    }
  },
  {
    "ph": "f", "id": 219429, "pid": 76337, "tid": -914061504, "ts": 1716454224926699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926700, "dur": 0,
    "args": {
      "External id": 219430,
      "cbid": 203, "correlation": 219430
    }
  },
  {
    "ph": "f", "id": 219430, "pid": 76337, "tid": -914061504, "ts": 1716454224926700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926701, "dur": 0,
    "args": {
      "External id": 219431,
      "cbid": 205, "correlation": 219431
    }
  },
  {
    "ph": "f", "id": 219431, "pid": 76337, "tid": -914061504, "ts": 1716454224926701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224927294, "dur": 40,
    "args": {
      "External id": 219435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219435,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219435, "pid": 5, "tid": 7, "ts": 1716454224927294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926715, "dur": 12,
    "args": {
      "External id": 219435,
      "cbid": 211, "correlation": 219435
    }
  },
  {
    "ph": "s", "id": 219435, "pid": 76337, "tid": -914061504, "ts": 1716454224926715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224927336, "dur": 14,
    "args": {
      "External id": 219437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219437,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219437, "pid": 5, "tid": 7, "ts": 1716454224927336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926730, "dur": 5,
    "args": {
      "External id": 219437,
      "cbid": 211, "correlation": 219437
    }
  },
  {
    "ph": "s", "id": 219437, "pid": 76337, "tid": -914061504, "ts": 1716454224926730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224927351, "dur": 3,
    "args": {
      "External id": 219439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219439,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219439, "pid": 5, "tid": 7, "ts": 1716454224927351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926739, "dur": 6,
    "args": {
      "External id": 219439,
      "cbid": 211, "correlation": 219439
    }
  },
  {
    "ph": "s", "id": 219439, "pid": 76337, "tid": -914061504, "ts": 1716454224926739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926748, "dur": 0,
    "args": {
      "External id": 219440,
      "cbid": 51, "correlation": 219440
    }
  },
  {
    "ph": "s", "id": 219440, "pid": 76337, "tid": -914061504, "ts": 1716454224926748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224927356, "dur": 698,
    "args": {
      "External id": 219441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219441,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219441, "pid": 5, "tid": 7, "ts": 1716454224927356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926749, "dur": 5,
    "args": {
      "External id": 219441,
      "cbid": 211, "correlation": 219441
    }
  },
  {
    "ph": "s", "id": 219441, "pid": 76337, "tid": -914061504, "ts": 1716454224926749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224928056, "dur": 60,
    "args": {
      "External id": 219446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219446,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219446, "pid": 5, "tid": 7, "ts": 1716454224928056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926777, "dur": 8,
    "args": {
      "External id": 219446,
      "cbid": 211, "correlation": 219446
    }
  },
  {
    "ph": "s", "id": 219446, "pid": 76337, "tid": -914061504, "ts": 1716454224926777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224928117, "dur": 51,
    "args": {
      "External id": 219454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219454,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219454, "pid": 5, "tid": 7, "ts": 1716454224928117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926809, "dur": 8,
    "args": {
      "External id": 219454,
      "cbid": 211, "correlation": 219454
    }
  },
  {
    "ph": "s", "id": 219454, "pid": 76337, "tid": -914061504, "ts": 1716454224926809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224928169, "dur": 35,
    "args": {
      "External id": 219462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219462,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219462, "pid": 5, "tid": 7, "ts": 1716454224928169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926840, "dur": 9,
    "args": {
      "External id": 219462,
      "cbid": 211, "correlation": 219462
    }
  },
  {
    "ph": "s", "id": 219462, "pid": 76337, "tid": -914061504, "ts": 1716454224926840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224928205, "dur": 52,
    "args": {
      "External id": 219482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219482,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 219482, "pid": 5, "tid": 7, "ts": 1716454224928205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926921, "dur": 13,
    "args": {
      "External id": 219482,
      "cbid": 211, "correlation": 219482
    }
  },
  {
    "ph": "s", "id": 219482, "pid": 76337, "tid": -914061504, "ts": 1716454224926921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224928259, "dur": 4,
    "args": {
      "External id": 219494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219494,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219494, "pid": 5, "tid": 7, "ts": 1716454224928259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926944, "dur": 6,
    "args": {
      "External id": 219494,
      "cbid": 211, "correlation": 219494
    }
  },
  {
    "ph": "s", "id": 219494, "pid": 76337, "tid": -914061504, "ts": 1716454224926944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224928264, "dur": 57,
    "args": {
      "External id": 219497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219497, "pid": 5, "tid": 7, "ts": 1716454224928264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224926962, "dur": 6,
    "args": {
      "External id": 219497,
      "cbid": 211, "correlation": 219497
    }
  },
  {
    "ph": "s", "id": 219497, "pid": 76337, "tid": -914061504, "ts": 1716454224926962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927029, "dur": 0,
    "args": {
      "External id": 219508,
      "cbid": 317, "correlation": 219508
    }
  },
  {
    "ph": "f", "id": 219508, "pid": 76337, "tid": -914061504, "ts": 1716454224927029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927030, "dur": 0,
    "args": {
      "External id": 219509,
      "cbid": 203, "correlation": 219509
    }
  },
  {
    "ph": "f", "id": 219509, "pid": 76337, "tid": -914061504, "ts": 1716454224927030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927030, "dur": 0,
    "args": {
      "External id": 219510,
      "cbid": 205, "correlation": 219510
    }
  },
  {
    "ph": "f", "id": 219510, "pid": 76337, "tid": -914061504, "ts": 1716454224927030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927068, "dur": 3,
    "args": {
      "External id": 219514,
      "cbid": 251, "correlation": 219514
    }
  },
  {
    "ph": "f", "id": 219514, "pid": 76337, "tid": -914061504, "ts": 1716454224927068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927072, "dur": 1,
    "args": {
      "External id": 219515,
      "cbid": 251, "correlation": 219515
    }
  },
  {
    "ph": "f", "id": 219515, "pid": 76337, "tid": -914061504, "ts": 1716454224927072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927074, "dur": 1,
    "args": {
      "External id": 219516,
      "cbid": 251, "correlation": 219516
    }
  },
  {
    "ph": "f", "id": 219516, "pid": 76337, "tid": -914061504, "ts": 1716454224927074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927076, "dur": 1,
    "args": {
      "External id": 219517,
      "cbid": 251, "correlation": 219517
    }
  },
  {
    "ph": "f", "id": 219517, "pid": 76337, "tid": -914061504, "ts": 1716454224927076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927078, "dur": 1,
    "args": {
      "External id": 219518,
      "cbid": 251, "correlation": 219518
    }
  },
  {
    "ph": "f", "id": 219518, "pid": 76337, "tid": -914061504, "ts": 1716454224927078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927080, "dur": 1,
    "args": {
      "External id": 219519,
      "cbid": 251, "correlation": 219519
    }
  },
  {
    "ph": "f", "id": 219519, "pid": 76337, "tid": -914061504, "ts": 1716454224927080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927082, "dur": 1,
    "args": {
      "External id": 219520,
      "cbid": 251, "correlation": 219520
    }
  },
  {
    "ph": "f", "id": 219520, "pid": 76337, "tid": -914061504, "ts": 1716454224927082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927084, "dur": 1,
    "args": {
      "External id": 219521,
      "cbid": 251, "correlation": 219521
    }
  },
  {
    "ph": "f", "id": 219521, "pid": 76337, "tid": -914061504, "ts": 1716454224927084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927087, "dur": 0,
    "args": {
      "External id": 219522,
      "cbid": 251, "correlation": 219522
    }
  },
  {
    "ph": "f", "id": 219522, "pid": 76337, "tid": -914061504, "ts": 1716454224927087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224928322, "dur": 115,
    "args": {
      "External id": 219523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219523,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 219523, "pid": 5, "tid": 7, "ts": 1716454224928322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927091, "dur": 14,
    "args": {
      "External id": 219523,
      "cbid": 211, "correlation": 219523
    }
  },
  {
    "ph": "s", "id": 219523, "pid": 76337, "tid": -914061504, "ts": 1716454224927091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224928439, "dur": 60,
    "args": {
      "External id": 219529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219529, "pid": 5, "tid": 7, "ts": 1716454224928439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927129, "dur": 8,
    "args": {
      "External id": 219529,
      "cbid": 211, "correlation": 219529
    }
  },
  {
    "ph": "s", "id": 219529, "pid": 76337, "tid": -914061504, "ts": 1716454224927129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224928500, "dur": 498,
    "args": {
      "External id": 219538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219538,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219538, "pid": 5, "tid": 7, "ts": 1716454224928500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927224, "dur": 16,
    "args": {
      "External id": 219538,
      "cbid": 211, "correlation": 219538
    }
  },
  {
    "ph": "s", "id": 219538, "pid": 76337, "tid": -914061504, "ts": 1716454224927224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224928999, "dur": 182,
    "args": {
      "External id": 219560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219560,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219560, "pid": 5, "tid": 7, "ts": 1716454224928999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927302, "dur": 13,
    "args": {
      "External id": 219560,
      "cbid": 211, "correlation": 219560
    }
  },
  {
    "ph": "s", "id": 219560, "pid": 76337, "tid": -914061504, "ts": 1716454224927302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927420, "dur": 2,
    "args": {
      "External id": 219571,
      "cbid": 251, "correlation": 219571
    }
  },
  {
    "ph": "f", "id": 219571, "pid": 76337, "tid": -914061504, "ts": 1716454224927420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224929183, "dur": 199,
    "args": {
      "External id": 219572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219572,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219572, "pid": 5, "tid": 7, "ts": 1716454224929183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927426, "dur": 15,
    "args": {
      "External id": 219572,
      "cbid": 211, "correlation": 219572
    }
  },
  {
    "ph": "s", "id": 219572, "pid": 76337, "tid": -914061504, "ts": 1716454224927426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927501, "dur": 1,
    "args": {
      "External id": 219583,
      "cbid": 251, "correlation": 219583
    }
  },
  {
    "ph": "f", "id": 219583, "pid": 76337, "tid": -914061504, "ts": 1716454224927501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224929383, "dur": 191,
    "args": {
      "External id": 219584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219584,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219584, "pid": 5, "tid": 7, "ts": 1716454224929383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927505, "dur": 12,
    "args": {
      "External id": 219584,
      "cbid": 211, "correlation": 219584
    }
  },
  {
    "ph": "s", "id": 219584, "pid": 76337, "tid": -914061504, "ts": 1716454224927505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927572, "dur": 1,
    "args": {
      "External id": 219595,
      "cbid": 251, "correlation": 219595
    }
  },
  {
    "ph": "f", "id": 219595, "pid": 76337, "tid": -914061504, "ts": 1716454224927572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224929575, "dur": 190,
    "args": {
      "External id": 219596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219596,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219596, "pid": 5, "tid": 7, "ts": 1716454224929575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927576, "dur": 11,
    "args": {
      "External id": 219596,
      "cbid": 211, "correlation": 219596
    }
  },
  {
    "ph": "s", "id": 219596, "pid": 76337, "tid": -914061504, "ts": 1716454224927576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224929766, "dur": 18784,
    "args": {
      "External id": 219617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219617,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 219617, "pid": 5, "tid": 7, "ts": 1716454224929766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927684, "dur": 14,
    "args": {
      "External id": 219617,
      "cbid": 211, "correlation": 219617
    }
  },
  {
    "ph": "s", "id": 219617, "pid": 76337, "tid": -914061504, "ts": 1716454224927684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927798, "dur": 2,
    "args": {
      "External id": 219635,
      "cbid": 251, "correlation": 219635
    }
  },
  {
    "ph": "f", "id": 219635, "pid": 76337, "tid": -914061504, "ts": 1716454224927798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224948552, "dur": 205,
    "args": {
      "External id": 219637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219637,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219637, "pid": 5, "tid": 7, "ts": 1716454224948552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927805, "dur": 13,
    "args": {
      "External id": 219637,
      "cbid": 211, "correlation": 219637
    }
  },
  {
    "ph": "s", "id": 219637, "pid": 76337, "tid": -914061504, "ts": 1716454224927805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224948758, "dur": 66,
    "args": {
      "External id": 219645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219645,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219645, "pid": 5, "tid": 7, "ts": 1716454224948758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927877, "dur": 12,
    "args": {
      "External id": 219645,
      "cbid": 211, "correlation": 219645
    }
  },
  {
    "ph": "s", "id": 219645, "pid": 76337, "tid": -914061504, "ts": 1716454224927877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224948826, "dur": 97,
    "args": {
      "External id": 219653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219653,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219653, "pid": 5, "tid": 7, "ts": 1716454224948826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224927917, "dur": 9,
    "args": {
      "External id": 219653,
      "cbid": 211, "correlation": 219653
    }
  },
  {
    "ph": "s", "id": 219653, "pid": 76337, "tid": -914061504, "ts": 1716454224927917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224948924, "dur": 54,
    "args": {
      "External id": 219664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219664,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219664, "pid": 5, "tid": 7, "ts": 1716454224948924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928011, "dur": 15,
    "args": {
      "External id": 219664,
      "cbid": 211, "correlation": 219664
    }
  },
  {
    "ph": "s", "id": 219664, "pid": 76337, "tid": -914061504, "ts": 1716454224928011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224948979, "dur": 93,
    "args": {
      "External id": 219686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219686,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219686, "pid": 5, "tid": 7, "ts": 1716454224948979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928046, "dur": 8,
    "args": {
      "External id": 219686,
      "cbid": 211, "correlation": 219686
    }
  },
  {
    "ph": "s", "id": 219686, "pid": 76337, "tid": -914061504, "ts": 1716454224928046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928130, "dur": 1,
    "args": {
      "External id": 219697,
      "cbid": 251, "correlation": 219697
    }
  },
  {
    "ph": "f", "id": 219697, "pid": 76337, "tid": -914061504, "ts": 1716454224928130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224949073, "dur": 103,
    "args": {
      "External id": 219698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219698,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219698, "pid": 5, "tid": 7, "ts": 1716454224949073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928135, "dur": 13,
    "args": {
      "External id": 219698,
      "cbid": 211, "correlation": 219698
    }
  },
  {
    "ph": "s", "id": 219698, "pid": 76337, "tid": -914061504, "ts": 1716454224928135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928218, "dur": 1,
    "args": {
      "External id": 219709,
      "cbid": 251, "correlation": 219709
    }
  },
  {
    "ph": "f", "id": 219709, "pid": 76337, "tid": -914061504, "ts": 1716454224928218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928222, "dur": 0,
    "args": {
      "External id": 219710,
      "cbid": 251, "correlation": 219710
    }
  },
  {
    "ph": "f", "id": 219710, "pid": 76337, "tid": -914061504, "ts": 1716454224928222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224949177, "dur": 10,
    "args": {
      "External id": 219711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219711,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219711, "pid": 5, "tid": 7, "ts": 1716454224949177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928224, "dur": 13,
    "args": {
      "External id": 219711,
      "cbid": 211, "correlation": 219711
    }
  },
  {
    "ph": "s", "id": 219711, "pid": 76337, "tid": -914061504, "ts": 1716454224928224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224949188, "dur": 5,
    "args": {
      "External id": 219713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219713,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 219713, "pid": 5, "tid": 7, "ts": 1716454224949188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928241, "dur": 8,
    "args": {
      "External id": 219713,
      "cbid": 211, "correlation": 219713
    }
  },
  {
    "ph": "s", "id": 219713, "pid": 76337, "tid": -914061504, "ts": 1716454224928241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928305, "dur": 1,
    "args": {
      "External id": 219724,
      "cbid": 251, "correlation": 219724
    }
  },
  {
    "ph": "f", "id": 219724, "pid": 76337, "tid": -914061504, "ts": 1716454224928305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928308, "dur": 0,
    "args": {
      "External id": 219725,
      "cbid": 251, "correlation": 219725
    }
  },
  {
    "ph": "f", "id": 219725, "pid": 76337, "tid": -914061504, "ts": 1716454224928308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224949194, "dur": 6,
    "args": {
      "External id": 219726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219726,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 219726, "pid": 5, "tid": 7, "ts": 1716454224949194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928310, "dur": 13,
    "args": {
      "External id": 219726,
      "cbid": 211, "correlation": 219726
    }
  },
  {
    "ph": "s", "id": 219726, "pid": 76337, "tid": -914061504, "ts": 1716454224928310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224949202, "dur": 4,
    "args": {
      "External id": 219728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219728,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 219728, "pid": 5, "tid": 7, "ts": 1716454224949202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928324, "dur": 5,
    "args": {
      "External id": 219728,
      "cbid": 211, "correlation": 219728
    }
  },
  {
    "ph": "s", "id": 219728, "pid": 76337, "tid": -914061504, "ts": 1716454224928324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224949207, "dur": 155,
    "args": {
      "External id": 219749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219749,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 219749, "pid": 5, "tid": 7, "ts": 1716454224949207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928400, "dur": 12,
    "args": {
      "External id": 219749,
      "cbid": 211, "correlation": 219749
    }
  },
  {
    "ph": "s", "id": 219749, "pid": 76337, "tid": -914061504, "ts": 1716454224928400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928495, "dur": 2,
    "args": {
      "External id": 219767,
      "cbid": 251, "correlation": 219767
    }
  },
  {
    "ph": "f", "id": 219767, "pid": 76337, "tid": -914061504, "ts": 1716454224928495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224949363, "dur": 107,
    "args": {
      "External id": 219769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219769,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 219769, "pid": 5, "tid": 7, "ts": 1716454224949363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928502, "dur": 14,
    "args": {
      "External id": 219769,
      "cbid": 211, "correlation": 219769
    }
  },
  {
    "ph": "s", "id": 219769, "pid": 76337, "tid": -914061504, "ts": 1716454224928502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224949471, "dur": 34,
    "args": {
      "External id": 219777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219777,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219777, "pid": 5, "tid": 7, "ts": 1716454224949471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928573, "dur": 12,
    "args": {
      "External id": 219777,
      "cbid": 211, "correlation": 219777
    }
  },
  {
    "ph": "s", "id": 219777, "pid": 76337, "tid": -914061504, "ts": 1716454224928573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224949507, "dur": 68,
    "args": {
      "External id": 219785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219785, "pid": 5, "tid": 7, "ts": 1716454224949507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928614, "dur": 9,
    "args": {
      "External id": 219785,
      "cbid": 211, "correlation": 219785
    }
  },
  {
    "ph": "s", "id": 219785, "pid": 76337, "tid": -914061504, "ts": 1716454224928614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224949576, "dur": 93,
    "args": {
      "External id": 219807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219807,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219807, "pid": 5, "tid": 7, "ts": 1716454224949576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928666, "dur": 10,
    "args": {
      "External id": 219807,
      "cbid": 211, "correlation": 219807
    }
  },
  {
    "ph": "s", "id": 219807, "pid": 76337, "tid": -914061504, "ts": 1716454224928666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928756, "dur": 1,
    "args": {
      "External id": 219823,
      "cbid": 251, "correlation": 219823
    }
  },
  {
    "ph": "f", "id": 219823, "pid": 76337, "tid": -914061504, "ts": 1716454224928756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224949670, "dur": 584,
    "args": {
      "External id": 219825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219825,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219825, "pid": 5, "tid": 7, "ts": 1716454224949670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928761, "dur": 12,
    "args": {
      "External id": 219825,
      "cbid": 211, "correlation": 219825
    }
  },
  {
    "ph": "s", "id": 219825, "pid": 76337, "tid": -914061504, "ts": 1716454224928761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224950255, "dur": 245,
    "args": {
      "External id": 219833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219833, "pid": 5, "tid": 7, "ts": 1716454224950255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928843, "dur": 19,
    "args": {
      "External id": 219833,
      "cbid": 211, "correlation": 219833
    }
  },
  {
    "ph": "s", "id": 219833, "pid": 76337, "tid": -914061504, "ts": 1716454224928843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224950501, "dur": 249,
    "args": {
      "External id": 219841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219841,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219841, "pid": 5, "tid": 7, "ts": 1716454224950501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928885, "dur": 10,
    "args": {
      "External id": 219841,
      "cbid": 211, "correlation": 219841
    }
  },
  {
    "ph": "s", "id": 219841, "pid": 76337, "tid": -914061504, "ts": 1716454224928885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928971, "dur": 1,
    "args": {
      "External id": 219857,
      "cbid": 251, "correlation": 219857
    }
  },
  {
    "ph": "f", "id": 219857, "pid": 76337, "tid": -914061504, "ts": 1716454224928971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928983, "dur": 0,
    "args": {
      "External id": 219859,
      "cbid": 251, "correlation": 219859
    }
  },
  {
    "ph": "f", "id": 219859, "pid": 76337, "tid": -914061504, "ts": 1716454224928983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224950752, "dur": 359,
    "args": {
      "External id": 219860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219860,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 219860, "pid": 5, "tid": 7, "ts": 1716454224950752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224928988, "dur": 14,
    "args": {
      "External id": 219860,
      "cbid": 211, "correlation": 219860
    }
  },
  {
    "ph": "s", "id": 219860, "pid": 76337, "tid": -914061504, "ts": 1716454224928988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224951112, "dur": 50,
    "args": {
      "External id": 219868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219868,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219868, "pid": 5, "tid": 7, "ts": 1716454224951112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929032, "dur": 11,
    "args": {
      "External id": 219868,
      "cbid": 211, "correlation": 219868
    }
  },
  {
    "ph": "s", "id": 219868, "pid": 76337, "tid": -914061504, "ts": 1716454224929032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224951164, "dur": 157,
    "args": {
      "External id": 219879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219879,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219879, "pid": 5, "tid": 7, "ts": 1716454224951164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929105, "dur": 12,
    "args": {
      "External id": 219879,
      "cbid": 211, "correlation": 219879
    }
  },
  {
    "ph": "s", "id": 219879, "pid": 76337, "tid": -914061504, "ts": 1716454224929105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929169, "dur": 0,
    "args": {
      "External id": 219891,
      "cbid": 317, "correlation": 219891
    }
  },
  {
    "ph": "f", "id": 219891, "pid": 76337, "tid": -914061504, "ts": 1716454224929169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929170, "dur": 0,
    "args": {
      "External id": 219892,
      "cbid": 203, "correlation": 219892
    }
  },
  {
    "ph": "f", "id": 219892, "pid": 76337, "tid": -914061504, "ts": 1716454224929170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929171, "dur": 0,
    "args": {
      "External id": 219893,
      "cbid": 205, "correlation": 219893
    }
  },
  {
    "ph": "f", "id": 219893, "pid": 76337, "tid": -914061504, "ts": 1716454224929171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929198, "dur": 1,
    "args": {
      "External id": 219897,
      "cbid": 251, "correlation": 219897
    }
  },
  {
    "ph": "f", "id": 219897, "pid": 76337, "tid": -914061504, "ts": 1716454224929198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929200, "dur": 0,
    "args": {
      "External id": 219898,
      "cbid": 251, "correlation": 219898
    }
  },
  {
    "ph": "f", "id": 219898, "pid": 76337, "tid": -914061504, "ts": 1716454224929200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929200, "dur": 0,
    "args": {
      "External id": 219899,
      "cbid": 251, "correlation": 219899
    }
  },
  {
    "ph": "f", "id": 219899, "pid": 76337, "tid": -914061504, "ts": 1716454224929200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929201, "dur": 0,
    "args": {
      "External id": 219900,
      "cbid": 251, "correlation": 219900
    }
  },
  {
    "ph": "f", "id": 219900, "pid": 76337, "tid": -914061504, "ts": 1716454224929201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929202, "dur": 0,
    "args": {
      "External id": 219901,
      "cbid": 251, "correlation": 219901
    }
  },
  {
    "ph": "f", "id": 219901, "pid": 76337, "tid": -914061504, "ts": 1716454224929202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929203, "dur": 0,
    "args": {
      "External id": 219902,
      "cbid": 251, "correlation": 219902
    }
  },
  {
    "ph": "f", "id": 219902, "pid": 76337, "tid": -914061504, "ts": 1716454224929203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929204, "dur": 0,
    "args": {
      "External id": 219903,
      "cbid": 251, "correlation": 219903
    }
  },
  {
    "ph": "f", "id": 219903, "pid": 76337, "tid": -914061504, "ts": 1716454224929204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929205, "dur": 0,
    "args": {
      "External id": 219904,
      "cbid": 251, "correlation": 219904
    }
  },
  {
    "ph": "f", "id": 219904, "pid": 76337, "tid": -914061504, "ts": 1716454224929205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929206, "dur": 0,
    "args": {
      "External id": 219905,
      "cbid": 251, "correlation": 219905
    }
  },
  {
    "ph": "f", "id": 219905, "pid": 76337, "tid": -914061504, "ts": 1716454224929206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224951323, "dur": 115,
    "args": {
      "External id": 219906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219906,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 219906, "pid": 5, "tid": 7, "ts": 1716454224951323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929208, "dur": 12,
    "args": {
      "External id": 219906,
      "cbid": 211, "correlation": 219906
    }
  },
  {
    "ph": "s", "id": 219906, "pid": 76337, "tid": -914061504, "ts": 1716454224929208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224951439, "dur": 60,
    "args": {
      "External id": 219912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219912, "pid": 5, "tid": 7, "ts": 1716454224951439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929243, "dur": 9,
    "args": {
      "External id": 219912,
      "cbid": 211, "correlation": 219912
    }
  },
  {
    "ph": "s", "id": 219912, "pid": 76337, "tid": -914061504, "ts": 1716454224929243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224951500, "dur": 50,
    "args": {
      "External id": 219920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219920,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219920, "pid": 5, "tid": 7, "ts": 1716454224951500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929275, "dur": 8,
    "args": {
      "External id": 219920,
      "cbid": 211, "correlation": 219920
    }
  },
  {
    "ph": "s", "id": 219920, "pid": 76337, "tid": -914061504, "ts": 1716454224929275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224951551, "dur": 53,
    "args": {
      "External id": 219940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219940,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 219940, "pid": 5, "tid": 7, "ts": 1716454224951551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929349, "dur": 11,
    "args": {
      "External id": 219940,
      "cbid": 211, "correlation": 219940
    }
  },
  {
    "ph": "s", "id": 219940, "pid": 76337, "tid": -914061504, "ts": 1716454224929349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224951606, "dur": 5,
    "args": {
      "External id": 219952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219952,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219952, "pid": 5, "tid": 7, "ts": 1716454224951606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929370, "dur": 6,
    "args": {
      "External id": 219952,
      "cbid": 211, "correlation": 219952
    }
  },
  {
    "ph": "s", "id": 219952, "pid": 76337, "tid": -914061504, "ts": 1716454224929370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224951612, "dur": 57,
    "args": {
      "External id": 219955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219955, "pid": 5, "tid": 7, "ts": 1716454224951612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929390, "dur": 6,
    "args": {
      "External id": 219955,
      "cbid": 211, "correlation": 219955
    }
  },
  {
    "ph": "s", "id": 219955, "pid": 76337, "tid": -914061504, "ts": 1716454224929390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224951670, "dur": 37,
    "args": {
      "External id": 219964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219964,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219964, "pid": 5, "tid": 7, "ts": 1716454224951670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929428, "dur": 11,
    "args": {
      "External id": 219964,
      "cbid": 211, "correlation": 219964
    }
  },
  {
    "ph": "s", "id": 219964, "pid": 76337, "tid": -914061504, "ts": 1716454224929428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929481, "dur": 0,
    "args": {
      "External id": 219974,
      "cbid": 317, "correlation": 219974
    }
  },
  {
    "ph": "f", "id": 219974, "pid": 76337, "tid": -914061504, "ts": 1716454224929481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929482, "dur": 0,
    "args": {
      "External id": 219975,
      "cbid": 203, "correlation": 219975
    }
  },
  {
    "ph": "f", "id": 219975, "pid": 76337, "tid": -914061504, "ts": 1716454224929482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929482, "dur": 0,
    "args": {
      "External id": 219976,
      "cbid": 205, "correlation": 219976
    }
  },
  {
    "ph": "f", "id": 219976, "pid": 76337, "tid": -914061504, "ts": 1716454224929482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224951708, "dur": 41,
    "args": {
      "External id": 219980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219980, "pid": 5, "tid": 7, "ts": 1716454224951708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929500, "dur": 12,
    "args": {
      "External id": 219980,
      "cbid": 211, "correlation": 219980
    }
  },
  {
    "ph": "s", "id": 219980, "pid": 76337, "tid": -914061504, "ts": 1716454224929500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224951751, "dur": 14,
    "args": {
      "External id": 219982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219982, "pid": 5, "tid": 7, "ts": 1716454224951751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929514, "dur": 7,
    "args": {
      "External id": 219982,
      "cbid": 211, "correlation": 219982
    }
  },
  {
    "ph": "s", "id": 219982, "pid": 76337, "tid": -914061504, "ts": 1716454224929514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224951767, "dur": 4,
    "args": {
      "External id": 219984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219984, "pid": 5, "tid": 7, "ts": 1716454224951767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929525, "dur": 5,
    "args": {
      "External id": 219984,
      "cbid": 211, "correlation": 219984
    }
  },
  {
    "ph": "s", "id": 219984, "pid": 76337, "tid": -914061504, "ts": 1716454224929525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929533, "dur": 0,
    "args": {
      "External id": 219985,
      "cbid": 51, "correlation": 219985
    }
  },
  {
    "ph": "s", "id": 219985, "pid": 76337, "tid": -914061504, "ts": 1716454224929533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224951772, "dur": 705,
    "args": {
      "External id": 219986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219986,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 219986, "pid": 5, "tid": 7, "ts": 1716454224951772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929534, "dur": 6,
    "args": {
      "External id": 219986,
      "cbid": 211, "correlation": 219986
    }
  },
  {
    "ph": "s", "id": 219986, "pid": 76337, "tid": -914061504, "ts": 1716454224929534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224952478, "dur": 59,
    "args": {
      "External id": 219991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 219991, "pid": 5, "tid": 7, "ts": 1716454224952478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929562, "dur": 8,
    "args": {
      "External id": 219991,
      "cbid": 211, "correlation": 219991
    }
  },
  {
    "ph": "s", "id": 219991, "pid": 76337, "tid": -914061504, "ts": 1716454224929562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224952538, "dur": 4,
    "args": {
      "External id": 219999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 219999,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 219999, "pid": 5, "tid": 7, "ts": 1716454224952538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929605, "dur": 9,
    "args": {
      "External id": 219999,
      "cbid": 211, "correlation": 219999
    }
  },
  {
    "ph": "s", "id": 219999, "pid": 76337, "tid": -914061504, "ts": 1716454224929605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929669, "dur": 1,
    "args": {
      "External id": 220015,
      "cbid": 251, "correlation": 220015
    }
  },
  {
    "ph": "f", "id": 220015, "pid": 76337, "tid": -914061504, "ts": 1716454224929669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929674, "dur": 0,
    "args": {
      "External id": 220017,
      "cbid": 251, "correlation": 220017
    }
  },
  {
    "ph": "f", "id": 220017, "pid": 76337, "tid": -914061504, "ts": 1716454224929674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224952543, "dur": 12,
    "args": {
      "External id": 220018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220018,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 220018, "pid": 5, "tid": 7, "ts": 1716454224952543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929676, "dur": 13,
    "args": {
      "External id": 220018,
      "cbid": 211, "correlation": 220018
    }
  },
  {
    "ph": "s", "id": 220018, "pid": 76337, "tid": -914061504, "ts": 1716454224929676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224952556, "dur": 5,
    "args": {
      "External id": 220020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220020,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 220020, "pid": 5, "tid": 7, "ts": 1716454224952556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929691, "dur": 5,
    "args": {
      "External id": 220020,
      "cbid": 211, "correlation": 220020
    }
  },
  {
    "ph": "s", "id": 220020, "pid": 76337, "tid": -914061504, "ts": 1716454224929691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224952563, "dur": 55,
    "args": {
      "External id": 220030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220030, "pid": 5, "tid": 7, "ts": 1716454224952563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929748, "dur": 12,
    "args": {
      "External id": 220030,
      "cbid": 211, "correlation": 220030
    }
  },
  {
    "ph": "s", "id": 220030, "pid": 76337, "tid": -914061504, "ts": 1716454224929748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224952619, "dur": 51,
    "args": {
      "External id": 220050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220050,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 220050, "pid": 5, "tid": 7, "ts": 1716454224952619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929813, "dur": 10,
    "args": {
      "External id": 220050,
      "cbid": 211, "correlation": 220050
    }
  },
  {
    "ph": "s", "id": 220050, "pid": 76337, "tid": -914061504, "ts": 1716454224929813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224952671, "dur": 4,
    "args": {
      "External id": 220062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220062,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 220062, "pid": 5, "tid": 7, "ts": 1716454224952671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929834, "dur": 6,
    "args": {
      "External id": 220062,
      "cbid": 211, "correlation": 220062
    }
  },
  {
    "ph": "s", "id": 220062, "pid": 76337, "tid": -914061504, "ts": 1716454224929834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224952676, "dur": 55,
    "args": {
      "External id": 220065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220065, "pid": 5, "tid": 7, "ts": 1716454224952676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929852, "dur": 6,
    "args": {
      "External id": 220065,
      "cbid": 211, "correlation": 220065
    }
  },
  {
    "ph": "s", "id": 220065, "pid": 76337, "tid": -914061504, "ts": 1716454224929852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224952732, "dur": 37,
    "args": {
      "External id": 220074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220074,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220074, "pid": 5, "tid": 7, "ts": 1716454224952732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929893, "dur": 9,
    "args": {
      "External id": 220074,
      "cbid": 211, "correlation": 220074
    }
  },
  {
    "ph": "s", "id": 220074, "pid": 76337, "tid": -914061504, "ts": 1716454224929893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929954, "dur": 0,
    "args": {
      "External id": 220084,
      "cbid": 317, "correlation": 220084
    }
  },
  {
    "ph": "f", "id": 220084, "pid": 76337, "tid": -914061504, "ts": 1716454224929954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929955, "dur": 0,
    "args": {
      "External id": 220085,
      "cbid": 203, "correlation": 220085
    }
  },
  {
    "ph": "f", "id": 220085, "pid": 76337, "tid": -914061504, "ts": 1716454224929955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929956, "dur": 0,
    "args": {
      "External id": 220086,
      "cbid": 205, "correlation": 220086
    }
  },
  {
    "ph": "f", "id": 220086, "pid": 76337, "tid": -914061504, "ts": 1716454224929956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224952771, "dur": 39,
    "args": {
      "External id": 220090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220090, "pid": 5, "tid": 7, "ts": 1716454224952771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929970, "dur": 22,
    "args": {
      "External id": 220090,
      "cbid": 211, "correlation": 220090
    }
  },
  {
    "ph": "s", "id": 220090, "pid": 76337, "tid": -914061504, "ts": 1716454224929970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224952811, "dur": 14,
    "args": {
      "External id": 220092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220092,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220092, "pid": 5, "tid": 7, "ts": 1716454224952811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224929994, "dur": 5,
    "args": {
      "External id": 220092,
      "cbid": 211, "correlation": 220092
    }
  },
  {
    "ph": "s", "id": 220092, "pid": 76337, "tid": -914061504, "ts": 1716454224929994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224952827, "dur": 3,
    "args": {
      "External id": 220094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 220094, "pid": 5, "tid": 7, "ts": 1716454224952827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930003, "dur": 6,
    "args": {
      "External id": 220094,
      "cbid": 211, "correlation": 220094
    }
  },
  {
    "ph": "s", "id": 220094, "pid": 76337, "tid": -914061504, "ts": 1716454224930003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930013, "dur": 0,
    "args": {
      "External id": 220095,
      "cbid": 51, "correlation": 220095
    }
  },
  {
    "ph": "s", "id": 220095, "pid": 76337, "tid": -914061504, "ts": 1716454224930013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224952832, "dur": 699,
    "args": {
      "External id": 220096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220096,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220096, "pid": 5, "tid": 7, "ts": 1716454224952832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930013, "dur": 5,
    "args": {
      "External id": 220096,
      "cbid": 211, "correlation": 220096
    }
  },
  {
    "ph": "s", "id": 220096, "pid": 76337, "tid": -914061504, "ts": 1716454224930013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224953532, "dur": 60,
    "args": {
      "External id": 220101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220101, "pid": 5, "tid": 7, "ts": 1716454224953532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930041, "dur": 8,
    "args": {
      "External id": 220101,
      "cbid": 211, "correlation": 220101
    }
  },
  {
    "ph": "s", "id": 220101, "pid": 76337, "tid": -914061504, "ts": 1716454224930041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224953594, "dur": 50,
    "args": {
      "External id": 220109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220109,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220109, "pid": 5, "tid": 7, "ts": 1716454224953594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930073, "dur": 8,
    "args": {
      "External id": 220109,
      "cbid": 211, "correlation": 220109
    }
  },
  {
    "ph": "s", "id": 220109, "pid": 76337, "tid": -914061504, "ts": 1716454224930073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224953646, "dur": 35,
    "args": {
      "External id": 220117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220117,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220117, "pid": 5, "tid": 7, "ts": 1716454224953646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930103, "dur": 8,
    "args": {
      "External id": 220117,
      "cbid": 211, "correlation": 220117
    }
  },
  {
    "ph": "s", "id": 220117, "pid": 76337, "tid": -914061504, "ts": 1716454224930103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224953682, "dur": 51,
    "args": {
      "External id": 220137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220137,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 220137, "pid": 5, "tid": 7, "ts": 1716454224953682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930182, "dur": 12,
    "args": {
      "External id": 220137,
      "cbid": 211, "correlation": 220137
    }
  },
  {
    "ph": "s", "id": 220137, "pid": 76337, "tid": -914061504, "ts": 1716454224930182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224953734, "dur": 4,
    "args": {
      "External id": 220149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220149,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 220149, "pid": 5, "tid": 7, "ts": 1716454224953734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930204, "dur": 6,
    "args": {
      "External id": 220149,
      "cbid": 211, "correlation": 220149
    }
  },
  {
    "ph": "s", "id": 220149, "pid": 76337, "tid": -914061504, "ts": 1716454224930204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224953739, "dur": 55,
    "args": {
      "External id": 220152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220152,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220152, "pid": 5, "tid": 7, "ts": 1716454224953739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930222, "dur": 6,
    "args": {
      "External id": 220152,
      "cbid": 211, "correlation": 220152
    }
  },
  {
    "ph": "s", "id": 220152, "pid": 76337, "tid": -914061504, "ts": 1716454224930222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930278, "dur": 0,
    "args": {
      "External id": 220163,
      "cbid": 317, "correlation": 220163
    }
  },
  {
    "ph": "f", "id": 220163, "pid": 76337, "tid": -914061504, "ts": 1716454224930278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930279, "dur": 0,
    "args": {
      "External id": 220164,
      "cbid": 203, "correlation": 220164
    }
  },
  {
    "ph": "f", "id": 220164, "pid": 76337, "tid": -914061504, "ts": 1716454224930279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930280, "dur": 0,
    "args": {
      "External id": 220165,
      "cbid": 205, "correlation": 220165
    }
  },
  {
    "ph": "f", "id": 220165, "pid": 76337, "tid": -914061504, "ts": 1716454224930280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930303, "dur": 1,
    "args": {
      "External id": 220169,
      "cbid": 251, "correlation": 220169
    }
  },
  {
    "ph": "f", "id": 220169, "pid": 76337, "tid": -914061504, "ts": 1716454224930303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930305, "dur": 0,
    "args": {
      "External id": 220170,
      "cbid": 251, "correlation": 220170
    }
  },
  {
    "ph": "f", "id": 220170, "pid": 76337, "tid": -914061504, "ts": 1716454224930305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930306, "dur": 0,
    "args": {
      "External id": 220171,
      "cbid": 251, "correlation": 220171
    }
  },
  {
    "ph": "f", "id": 220171, "pid": 76337, "tid": -914061504, "ts": 1716454224930306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930307, "dur": 0,
    "args": {
      "External id": 220172,
      "cbid": 251, "correlation": 220172
    }
  },
  {
    "ph": "f", "id": 220172, "pid": 76337, "tid": -914061504, "ts": 1716454224930307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930307, "dur": 0,
    "args": {
      "External id": 220173,
      "cbid": 251, "correlation": 220173
    }
  },
  {
    "ph": "f", "id": 220173, "pid": 76337, "tid": -914061504, "ts": 1716454224930307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930308, "dur": 0,
    "args": {
      "External id": 220174,
      "cbid": 251, "correlation": 220174
    }
  },
  {
    "ph": "f", "id": 220174, "pid": 76337, "tid": -914061504, "ts": 1716454224930308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930309, "dur": 0,
    "args": {
      "External id": 220175,
      "cbid": 251, "correlation": 220175
    }
  },
  {
    "ph": "f", "id": 220175, "pid": 76337, "tid": -914061504, "ts": 1716454224930309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930309, "dur": 0,
    "args": {
      "External id": 220176,
      "cbid": 251, "correlation": 220176
    }
  },
  {
    "ph": "f", "id": 220176, "pid": 76337, "tid": -914061504, "ts": 1716454224930309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930311, "dur": 0,
    "args": {
      "External id": 220177,
      "cbid": 251, "correlation": 220177
    }
  },
  {
    "ph": "f", "id": 220177, "pid": 76337, "tid": -914061504, "ts": 1716454224930311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224953796, "dur": 111,
    "args": {
      "External id": 220178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220178,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220178, "pid": 5, "tid": 7, "ts": 1716454224953796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930313, "dur": 12,
    "args": {
      "External id": 220178,
      "cbid": 211, "correlation": 220178
    }
  },
  {
    "ph": "s", "id": 220178, "pid": 76337, "tid": -914061504, "ts": 1716454224930313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224953908, "dur": 59,
    "args": {
      "External id": 220184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220184, "pid": 5, "tid": 7, "ts": 1716454224953908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930348, "dur": 9,
    "args": {
      "External id": 220184,
      "cbid": 211, "correlation": 220184
    }
  },
  {
    "ph": "s", "id": 220184, "pid": 76337, "tid": -914061504, "ts": 1716454224930348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224953968, "dur": 466,
    "args": {
      "External id": 220193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220193,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220193, "pid": 5, "tid": 7, "ts": 1716454224953968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930430, "dur": 15,
    "args": {
      "External id": 220193,
      "cbid": 211, "correlation": 220193
    }
  },
  {
    "ph": "s", "id": 220193, "pid": 76337, "tid": -914061504, "ts": 1716454224930430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224954435, "dur": 182,
    "args": {
      "External id": 220215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220215,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220215, "pid": 5, "tid": 7, "ts": 1716454224954435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930487, "dur": 10,
    "args": {
      "External id": 220215,
      "cbid": 211, "correlation": 220215
    }
  },
  {
    "ph": "s", "id": 220215, "pid": 76337, "tid": -914061504, "ts": 1716454224930487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930572, "dur": 1,
    "args": {
      "External id": 220226,
      "cbid": 251, "correlation": 220226
    }
  },
  {
    "ph": "f", "id": 220226, "pid": 76337, "tid": -914061504, "ts": 1716454224930572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224954618, "dur": 200,
    "args": {
      "External id": 220227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220227,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220227, "pid": 5, "tid": 7, "ts": 1716454224954618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930577, "dur": 13,
    "args": {
      "External id": 220227,
      "cbid": 211, "correlation": 220227
    }
  },
  {
    "ph": "s", "id": 220227, "pid": 76337, "tid": -914061504, "ts": 1716454224930577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930644, "dur": 1,
    "args": {
      "External id": 220238,
      "cbid": 251, "correlation": 220238
    }
  },
  {
    "ph": "f", "id": 220238, "pid": 76337, "tid": -914061504, "ts": 1716454224930644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224954819, "dur": 188,
    "args": {
      "External id": 220239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220239,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220239, "pid": 5, "tid": 7, "ts": 1716454224954819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930649, "dur": 11,
    "args": {
      "External id": 220239,
      "cbid": 211, "correlation": 220239
    }
  },
  {
    "ph": "s", "id": 220239, "pid": 76337, "tid": -914061504, "ts": 1716454224930649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930712, "dur": 1,
    "args": {
      "External id": 220250,
      "cbid": 251, "correlation": 220250
    }
  },
  {
    "ph": "f", "id": 220250, "pid": 76337, "tid": -914061504, "ts": 1716454224930712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224955009, "dur": 187,
    "args": {
      "External id": 220251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220251,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220251, "pid": 5, "tid": 7, "ts": 1716454224955009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930716, "dur": 11,
    "args": {
      "External id": 220251,
      "cbid": 211, "correlation": 220251
    }
  },
  {
    "ph": "s", "id": 220251, "pid": 76337, "tid": -914061504, "ts": 1716454224930716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224955197, "dur": 18840,
    "args": {
      "External id": 220272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220272,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220272, "pid": 5, "tid": 7, "ts": 1716454224955197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930796, "dur": 12,
    "args": {
      "External id": 220272,
      "cbid": 211, "correlation": 220272
    }
  },
  {
    "ph": "s", "id": 220272, "pid": 76337, "tid": -914061504, "ts": 1716454224930796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930895, "dur": 1,
    "args": {
      "External id": 220290,
      "cbid": 251, "correlation": 220290
    }
  },
  {
    "ph": "f", "id": 220290, "pid": 76337, "tid": -914061504, "ts": 1716454224930895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224974038, "dur": 205,
    "args": {
      "External id": 220292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220292,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220292, "pid": 5, "tid": 7, "ts": 1716454224974038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930900, "dur": 12,
    "args": {
      "External id": 220292,
      "cbid": 211, "correlation": 220292
    }
  },
  {
    "ph": "s", "id": 220292, "pid": 76337, "tid": -914061504, "ts": 1716454224930900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224974244, "dur": 66,
    "args": {
      "External id": 220300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220300,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220300, "pid": 5, "tid": 7, "ts": 1716454224974244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224930969, "dur": 20,
    "args": {
      "External id": 220300,
      "cbid": 211, "correlation": 220300
    }
  },
  {
    "ph": "s", "id": 220300, "pid": 76337, "tid": -914061504, "ts": 1716454224930969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224974312, "dur": 97,
    "args": {
      "External id": 220308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220308,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220308, "pid": 5, "tid": 7, "ts": 1716454224974312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931018, "dur": 9,
    "args": {
      "External id": 220308,
      "cbid": 211, "correlation": 220308
    }
  },
  {
    "ph": "s", "id": 220308, "pid": 76337, "tid": -914061504, "ts": 1716454224931018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224974410, "dur": 54,
    "args": {
      "External id": 220319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220319,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220319, "pid": 5, "tid": 7, "ts": 1716454224974410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931090, "dur": 12,
    "args": {
      "External id": 220319,
      "cbid": 211, "correlation": 220319
    }
  },
  {
    "ph": "s", "id": 220319, "pid": 76337, "tid": -914061504, "ts": 1716454224931090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224974465, "dur": 93,
    "args": {
      "External id": 220341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220341,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220341, "pid": 5, "tid": 7, "ts": 1716454224974465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931121, "dur": 7,
    "args": {
      "External id": 220341,
      "cbid": 211, "correlation": 220341
    }
  },
  {
    "ph": "s", "id": 220341, "pid": 76337, "tid": -914061504, "ts": 1716454224931121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931204, "dur": 1,
    "args": {
      "External id": 220352,
      "cbid": 251, "correlation": 220352
    }
  },
  {
    "ph": "f", "id": 220352, "pid": 76337, "tid": -914061504, "ts": 1716454224931204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224974559, "dur": 102,
    "args": {
      "External id": 220353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220353,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220353, "pid": 5, "tid": 7, "ts": 1716454224974559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931210, "dur": 12,
    "args": {
      "External id": 220353,
      "cbid": 211, "correlation": 220353
    }
  },
  {
    "ph": "s", "id": 220353, "pid": 76337, "tid": -914061504, "ts": 1716454224931210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931280, "dur": 1,
    "args": {
      "External id": 220364,
      "cbid": 251, "correlation": 220364
    }
  },
  {
    "ph": "f", "id": 220364, "pid": 76337, "tid": -914061504, "ts": 1716454224931280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931284, "dur": 0,
    "args": {
      "External id": 220365,
      "cbid": 251, "correlation": 220365
    }
  },
  {
    "ph": "f", "id": 220365, "pid": 76337, "tid": -914061504, "ts": 1716454224931284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224974662, "dur": 10,
    "args": {
      "External id": 220366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220366,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 220366, "pid": 5, "tid": 7, "ts": 1716454224974662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931286, "dur": 12,
    "args": {
      "External id": 220366,
      "cbid": 211, "correlation": 220366
    }
  },
  {
    "ph": "s", "id": 220366, "pid": 76337, "tid": -914061504, "ts": 1716454224931286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224974674, "dur": 5,
    "args": {
      "External id": 220368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220368,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 220368, "pid": 5, "tid": 7, "ts": 1716454224974674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931300, "dur": 5,
    "args": {
      "External id": 220368,
      "cbid": 211, "correlation": 220368
    }
  },
  {
    "ph": "s", "id": 220368, "pid": 76337, "tid": -914061504, "ts": 1716454224931300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931360, "dur": 1,
    "args": {
      "External id": 220379,
      "cbid": 251, "correlation": 220379
    }
  },
  {
    "ph": "f", "id": 220379, "pid": 76337, "tid": -914061504, "ts": 1716454224931360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931363, "dur": 0,
    "args": {
      "External id": 220380,
      "cbid": 251, "correlation": 220380
    }
  },
  {
    "ph": "f", "id": 220380, "pid": 76337, "tid": -914061504, "ts": 1716454224931363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224974680, "dur": 6,
    "args": {
      "External id": 220381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220381,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 220381, "pid": 5, "tid": 7, "ts": 1716454224974680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931365, "dur": 13,
    "args": {
      "External id": 220381,
      "cbid": 211, "correlation": 220381
    }
  },
  {
    "ph": "s", "id": 220381, "pid": 76337, "tid": -914061504, "ts": 1716454224931365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224974688, "dur": 3,
    "args": {
      "External id": 220383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220383,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 220383, "pid": 5, "tid": 7, "ts": 1716454224974688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931379, "dur": 5,
    "args": {
      "External id": 220383,
      "cbid": 211, "correlation": 220383
    }
  },
  {
    "ph": "s", "id": 220383, "pid": 76337, "tid": -914061504, "ts": 1716454224931379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224974693, "dur": 158,
    "args": {
      "External id": 220404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220404,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220404, "pid": 5, "tid": 7, "ts": 1716454224974693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931452, "dur": 12,
    "args": {
      "External id": 220404,
      "cbid": 211, "correlation": 220404
    }
  },
  {
    "ph": "s", "id": 220404, "pid": 76337, "tid": -914061504, "ts": 1716454224931452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931549, "dur": 1,
    "args": {
      "External id": 220422,
      "cbid": 251, "correlation": 220422
    }
  },
  {
    "ph": "f", "id": 220422, "pid": 76337, "tid": -914061504, "ts": 1716454224931549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224974852, "dur": 107,
    "args": {
      "External id": 220424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220424,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220424, "pid": 5, "tid": 7, "ts": 1716454224974852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931555, "dur": 12,
    "args": {
      "External id": 220424,
      "cbid": 211, "correlation": 220424
    }
  },
  {
    "ph": "s", "id": 220424, "pid": 76337, "tid": -914061504, "ts": 1716454224931555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224974960, "dur": 35,
    "args": {
      "External id": 220432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220432,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220432, "pid": 5, "tid": 7, "ts": 1716454224974960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931623, "dur": 12,
    "args": {
      "External id": 220432,
      "cbid": 211, "correlation": 220432
    }
  },
  {
    "ph": "s", "id": 220432, "pid": 76337, "tid": -914061504, "ts": 1716454224931623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224974996, "dur": 67,
    "args": {
      "External id": 220440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220440, "pid": 5, "tid": 7, "ts": 1716454224974996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931662, "dur": 9,
    "args": {
      "External id": 220440,
      "cbid": 211, "correlation": 220440
    }
  },
  {
    "ph": "s", "id": 220440, "pid": 76337, "tid": -914061504, "ts": 1716454224931662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224975064, "dur": 93,
    "args": {
      "External id": 220462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220462,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220462, "pid": 5, "tid": 7, "ts": 1716454224975064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931714, "dur": 10,
    "args": {
      "External id": 220462,
      "cbid": 211, "correlation": 220462
    }
  },
  {
    "ph": "s", "id": 220462, "pid": 76337, "tid": -914061504, "ts": 1716454224931714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931803, "dur": 1,
    "args": {
      "External id": 220478,
      "cbid": 251, "correlation": 220478
    }
  },
  {
    "ph": "f", "id": 220478, "pid": 76337, "tid": -914061504, "ts": 1716454224931803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224975158, "dur": 576,
    "args": {
      "External id": 220480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220480,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220480, "pid": 5, "tid": 7, "ts": 1716454224975158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931809, "dur": 12,
    "args": {
      "External id": 220480,
      "cbid": 211, "correlation": 220480
    }
  },
  {
    "ph": "s", "id": 220480, "pid": 76337, "tid": -914061504, "ts": 1716454224931809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224975736, "dur": 245,
    "args": {
      "External id": 220488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220488, "pid": 5, "tid": 7, "ts": 1716454224975736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931873, "dur": 12,
    "args": {
      "External id": 220488,
      "cbid": 211, "correlation": 220488
    }
  },
  {
    "ph": "s", "id": 220488, "pid": 76337, "tid": -914061504, "ts": 1716454224931873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224975982, "dur": 254,
    "args": {
      "External id": 220496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220496, "pid": 5, "tid": 7, "ts": 1716454224975982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931903, "dur": 8,
    "args": {
      "External id": 220496,
      "cbid": 211, "correlation": 220496
    }
  },
  {
    "ph": "s", "id": 220496, "pid": 76337, "tid": -914061504, "ts": 1716454224931903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931994, "dur": 1,
    "args": {
      "External id": 220512,
      "cbid": 251, "correlation": 220512
    }
  },
  {
    "ph": "f", "id": 220512, "pid": 76337, "tid": -914061504, "ts": 1716454224931994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224931999, "dur": 0,
    "args": {
      "External id": 220514,
      "cbid": 251, "correlation": 220514
    }
  },
  {
    "ph": "f", "id": 220514, "pid": 76337, "tid": -914061504, "ts": 1716454224931999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224976237, "dur": 360,
    "args": {
      "External id": 220515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220515,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 220515, "pid": 5, "tid": 7, "ts": 1716454224976237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932002, "dur": 13,
    "args": {
      "External id": 220515,
      "cbid": 211, "correlation": 220515
    }
  },
  {
    "ph": "s", "id": 220515, "pid": 76337, "tid": -914061504, "ts": 1716454224932002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224976599, "dur": 51,
    "args": {
      "External id": 220523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220523,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220523, "pid": 5, "tid": 7, "ts": 1716454224976599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932046, "dur": 10,
    "args": {
      "External id": 220523,
      "cbid": 211, "correlation": 220523
    }
  },
  {
    "ph": "s", "id": 220523, "pid": 76337, "tid": -914061504, "ts": 1716454224932046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224976651, "dur": 159,
    "args": {
      "External id": 220534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220534, "pid": 5, "tid": 7, "ts": 1716454224976651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932114, "dur": 12,
    "args": {
      "External id": 220534,
      "cbid": 211, "correlation": 220534
    }
  },
  {
    "ph": "s", "id": 220534, "pid": 76337, "tid": -914061504, "ts": 1716454224932114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932178, "dur": 0,
    "args": {
      "External id": 220546,
      "cbid": 317, "correlation": 220546
    }
  },
  {
    "ph": "f", "id": 220546, "pid": 76337, "tid": -914061504, "ts": 1716454224932178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932179, "dur": 0,
    "args": {
      "External id": 220547,
      "cbid": 203, "correlation": 220547
    }
  },
  {
    "ph": "f", "id": 220547, "pid": 76337, "tid": -914061504, "ts": 1716454224932179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932180, "dur": 0,
    "args": {
      "External id": 220548,
      "cbid": 205, "correlation": 220548
    }
  },
  {
    "ph": "f", "id": 220548, "pid": 76337, "tid": -914061504, "ts": 1716454224932180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932205, "dur": 1,
    "args": {
      "External id": 220552,
      "cbid": 251, "correlation": 220552
    }
  },
  {
    "ph": "f", "id": 220552, "pid": 76337, "tid": -914061504, "ts": 1716454224932205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932207, "dur": 0,
    "args": {
      "External id": 220553,
      "cbid": 251, "correlation": 220553
    }
  },
  {
    "ph": "f", "id": 220553, "pid": 76337, "tid": -914061504, "ts": 1716454224932207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932207, "dur": 0,
    "args": {
      "External id": 220554,
      "cbid": 251, "correlation": 220554
    }
  },
  {
    "ph": "f", "id": 220554, "pid": 76337, "tid": -914061504, "ts": 1716454224932207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932208, "dur": 0,
    "args": {
      "External id": 220555,
      "cbid": 251, "correlation": 220555
    }
  },
  {
    "ph": "f", "id": 220555, "pid": 76337, "tid": -914061504, "ts": 1716454224932208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932209, "dur": 0,
    "args": {
      "External id": 220556,
      "cbid": 251, "correlation": 220556
    }
  },
  {
    "ph": "f", "id": 220556, "pid": 76337, "tid": -914061504, "ts": 1716454224932209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932209, "dur": 0,
    "args": {
      "External id": 220557,
      "cbid": 251, "correlation": 220557
    }
  },
  {
    "ph": "f", "id": 220557, "pid": 76337, "tid": -914061504, "ts": 1716454224932209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932210, "dur": 0,
    "args": {
      "External id": 220558,
      "cbid": 251, "correlation": 220558
    }
  },
  {
    "ph": "f", "id": 220558, "pid": 76337, "tid": -914061504, "ts": 1716454224932210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932211, "dur": 0,
    "args": {
      "External id": 220559,
      "cbid": 251, "correlation": 220559
    }
  },
  {
    "ph": "f", "id": 220559, "pid": 76337, "tid": -914061504, "ts": 1716454224932211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932212, "dur": 0,
    "args": {
      "External id": 220560,
      "cbid": 251, "correlation": 220560
    }
  },
  {
    "ph": "f", "id": 220560, "pid": 76337, "tid": -914061504, "ts": 1716454224932212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224976811, "dur": 115,
    "args": {
      "External id": 220561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220561,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220561, "pid": 5, "tid": 7, "ts": 1716454224976811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932214, "dur": 12,
    "args": {
      "External id": 220561,
      "cbid": 211, "correlation": 220561
    }
  },
  {
    "ph": "s", "id": 220561, "pid": 76337, "tid": -914061504, "ts": 1716454224932214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224976928, "dur": 60,
    "args": {
      "External id": 220567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220567,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220567, "pid": 5, "tid": 7, "ts": 1716454224976928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932249, "dur": 9,
    "args": {
      "External id": 220567,
      "cbid": 211, "correlation": 220567
    }
  },
  {
    "ph": "s", "id": 220567, "pid": 76337, "tid": -914061504, "ts": 1716454224932249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224976989, "dur": 49,
    "args": {
      "External id": 220575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220575,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220575, "pid": 5, "tid": 7, "ts": 1716454224976989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932280, "dur": 9,
    "args": {
      "External id": 220575,
      "cbid": 211, "correlation": 220575
    }
  },
  {
    "ph": "s", "id": 220575, "pid": 76337, "tid": -914061504, "ts": 1716454224932280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932355, "dur": 0,
    "args": {
      "External id": 220585,
      "cbid": 317, "correlation": 220585
    }
  },
  {
    "ph": "f", "id": 220585, "pid": 76337, "tid": -914061504, "ts": 1716454224932355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932356, "dur": 0,
    "args": {
      "External id": 220586,
      "cbid": 203, "correlation": 220586
    }
  },
  {
    "ph": "f", "id": 220586, "pid": 76337, "tid": -914061504, "ts": 1716454224932356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932356, "dur": 0,
    "args": {
      "External id": 220587,
      "cbid": 205, "correlation": 220587
    }
  },
  {
    "ph": "f", "id": 220587, "pid": 76337, "tid": -914061504, "ts": 1716454224932356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977040, "dur": 41,
    "args": {
      "External id": 220591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220591, "pid": 5, "tid": 7, "ts": 1716454224977040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932375, "dur": 12,
    "args": {
      "External id": 220591,
      "cbid": 211, "correlation": 220591
    }
  },
  {
    "ph": "s", "id": 220591, "pid": 76337, "tid": -914061504, "ts": 1716454224932375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977082, "dur": 14,
    "args": {
      "External id": 220593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220593,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220593, "pid": 5, "tid": 7, "ts": 1716454224977082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932390, "dur": 6,
    "args": {
      "External id": 220593,
      "cbid": 211, "correlation": 220593
    }
  },
  {
    "ph": "s", "id": 220593, "pid": 76337, "tid": -914061504, "ts": 1716454224932390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224977098, "dur": 1,
    "args": {
      "External id": 220595,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 220595,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.8888888888888888
    }
  },
  {
    "ph": "f", "id": 220595, "pid": 5, "tid": 7, "ts": 1716454224977098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932410, "dur": 17,
    "args": {
      "External id": 220595,
      "cbid": 51, "correlation": 220595
    }
  },
  {
    "ph": "s", "id": 220595, "pid": 76337, "tid": -914061504, "ts": 1716454224932410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224977102, "dur": 365,
    "args": {
      "External id": 220596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220596,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220596, "pid": 5, "tid": 7, "ts": 1716454224977102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932429, "dur": 10,
    "args": {
      "External id": 220596,
      "cbid": 211, "correlation": 220596
    }
  },
  {
    "ph": "s", "id": 220596, "pid": 76337, "tid": -914061504, "ts": 1716454224932429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977469, "dur": 14,
    "args": {
      "External id": 220598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220598, "pid": 5, "tid": 7, "ts": 1716454224977469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932447, "dur": 7,
    "args": {
      "External id": 220598,
      "cbid": 211, "correlation": 220598
    }
  },
  {
    "ph": "s", "id": 220598, "pid": 76337, "tid": -914061504, "ts": 1716454224932447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224977484, "dur": 15,
    "args": {
      "External id": 220604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220604, "pid": 5, "tid": 7, "ts": 1716454224977484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932477, "dur": 9,
    "args": {
      "External id": 220604,
      "cbid": 211, "correlation": 220604
    }
  },
  {
    "ph": "s", "id": 220604, "pid": 76337, "tid": -914061504, "ts": 1716454224932477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977500, "dur": 19,
    "args": {
      "External id": 220624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220624,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 220624, "pid": 5, "tid": 7, "ts": 1716454224977500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932570, "dur": 14,
    "args": {
      "External id": 220624,
      "cbid": 211, "correlation": 220624
    }
  },
  {
    "ph": "s", "id": 220624, "pid": 76337, "tid": -914061504, "ts": 1716454224932570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224977520, "dur": 4,
    "args": {
      "External id": 220636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220636,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 220636, "pid": 5, "tid": 7, "ts": 1716454224977520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932594, "dur": 6,
    "args": {
      "External id": 220636,
      "cbid": 211, "correlation": 220636
    }
  },
  {
    "ph": "s", "id": 220636, "pid": 76337, "tid": -914061504, "ts": 1716454224932594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224977526, "dur": 17,
    "args": {
      "External id": 220639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220639,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220639, "pid": 5, "tid": 7, "ts": 1716454224977526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932611, "dur": 6,
    "args": {
      "External id": 220639,
      "cbid": 211, "correlation": 220639
    }
  },
  {
    "ph": "s", "id": 220639, "pid": 76337, "tid": -914061504, "ts": 1716454224932611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224977545, "dur": 11,
    "args": {
      "External id": 220648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220648,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220648, "pid": 5, "tid": 7, "ts": 1716454224977545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932650, "dur": 9,
    "args": {
      "External id": 220648,
      "cbid": 211, "correlation": 220648
    }
  },
  {
    "ph": "s", "id": 220648, "pid": 76337, "tid": -914061504, "ts": 1716454224932650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932704, "dur": 0,
    "args": {
      "External id": 220658,
      "cbid": 317, "correlation": 220658
    }
  },
  {
    "ph": "f", "id": 220658, "pid": 76337, "tid": -914061504, "ts": 1716454224932704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932705, "dur": 0,
    "args": {
      "External id": 220659,
      "cbid": 203, "correlation": 220659
    }
  },
  {
    "ph": "f", "id": 220659, "pid": 76337, "tid": -914061504, "ts": 1716454224932705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932706, "dur": 0,
    "args": {
      "External id": 220660,
      "cbid": 205, "correlation": 220660
    }
  },
  {
    "ph": "f", "id": 220660, "pid": 76337, "tid": -914061504, "ts": 1716454224932706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977558, "dur": 12,
    "args": {
      "External id": 220664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220664,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220664, "pid": 5, "tid": 7, "ts": 1716454224977558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932721, "dur": 11,
    "args": {
      "External id": 220664,
      "cbid": 211, "correlation": 220664
    }
  },
  {
    "ph": "s", "id": 220664, "pid": 76337, "tid": -914061504, "ts": 1716454224932721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977570, "dur": 24,
    "args": {
      "External id": 220666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220666, "pid": 5, "tid": 7, "ts": 1716454224977570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932734, "dur": 6,
    "args": {
      "External id": 220666,
      "cbid": 211, "correlation": 220666
    }
  },
  {
    "ph": "s", "id": 220666, "pid": 76337, "tid": -914061504, "ts": 1716454224932734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224977596, "dur": 4,
    "args": {
      "External id": 220668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 220668, "pid": 5, "tid": 7, "ts": 1716454224977596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932746, "dur": 5,
    "args": {
      "External id": 220668,
      "cbid": 211, "correlation": 220668
    }
  },
  {
    "ph": "s", "id": 220668, "pid": 76337, "tid": -914061504, "ts": 1716454224932746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932755, "dur": 0,
    "args": {
      "External id": 220669,
      "cbid": 51, "correlation": 220669
    }
  },
  {
    "ph": "s", "id": 220669, "pid": 76337, "tid": -914061504, "ts": 1716454224932755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224977601, "dur": 358,
    "args": {
      "External id": 220670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220670,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220670, "pid": 5, "tid": 7, "ts": 1716454224977601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932756, "dur": 7,
    "args": {
      "External id": 220670,
      "cbid": 211, "correlation": 220670
    }
  },
  {
    "ph": "s", "id": 220670, "pid": 76337, "tid": -914061504, "ts": 1716454224932756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224977961, "dur": 21,
    "args": {
      "External id": 220671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220671, "pid": 5, "tid": 7, "ts": 1716454224977961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932765, "dur": 6,
    "args": {
      "External id": 220671,
      "cbid": 211, "correlation": 220671
    }
  },
  {
    "ph": "s", "id": 220671, "pid": 76337, "tid": -914061504, "ts": 1716454224932765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224977982, "dur": 32,
    "args": {
      "External id": 220677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220677, "pid": 5, "tid": 7, "ts": 1716454224977982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932794, "dur": 8,
    "args": {
      "External id": 220677,
      "cbid": 211, "correlation": 220677
    }
  },
  {
    "ph": "s", "id": 220677, "pid": 76337, "tid": -914061504, "ts": 1716454224932794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224978016, "dur": 3,
    "args": {
      "External id": 220685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220685,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 220685, "pid": 5, "tid": 7, "ts": 1716454224978016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932836, "dur": 9,
    "args": {
      "External id": 220685,
      "cbid": 211, "correlation": 220685
    }
  },
  {
    "ph": "s", "id": 220685, "pid": 76337, "tid": -914061504, "ts": 1716454224932836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932902, "dur": 1,
    "args": {
      "External id": 220701,
      "cbid": 251, "correlation": 220701
    }
  },
  {
    "ph": "f", "id": 220701, "pid": 76337, "tid": -914061504, "ts": 1716454224932902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932908, "dur": 0,
    "args": {
      "External id": 220703,
      "cbid": 251, "correlation": 220703
    }
  },
  {
    "ph": "f", "id": 220703, "pid": 76337, "tid": -914061504, "ts": 1716454224932908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224978021, "dur": 12,
    "args": {
      "External id": 220704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220704,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 220704, "pid": 5, "tid": 7, "ts": 1716454224978021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932909, "dur": 11,
    "args": {
      "External id": 220704,
      "cbid": 211, "correlation": 220704
    }
  },
  {
    "ph": "s", "id": 220704, "pid": 76337, "tid": -914061504, "ts": 1716454224932909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224978034, "dur": 5,
    "args": {
      "External id": 220706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220706,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 220706, "pid": 5, "tid": 7, "ts": 1716454224978034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932922, "dur": 5,
    "args": {
      "External id": 220706,
      "cbid": 211, "correlation": 220706
    }
  },
  {
    "ph": "s", "id": 220706, "pid": 76337, "tid": -914061504, "ts": 1716454224932922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224978041, "dur": 29,
    "args": {
      "External id": 220716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220716, "pid": 5, "tid": 7, "ts": 1716454224978041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224932989, "dur": 13,
    "args": {
      "External id": 220716,
      "cbid": 211, "correlation": 220716
    }
  },
  {
    "ph": "s", "id": 220716, "pid": 76337, "tid": -914061504, "ts": 1716454224932989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224978072, "dur": 31,
    "args": {
      "External id": 220736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220736,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 220736, "pid": 5, "tid": 7, "ts": 1716454224978072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933057, "dur": 11,
    "args": {
      "External id": 220736,
      "cbid": 211, "correlation": 220736
    }
  },
  {
    "ph": "s", "id": 220736, "pid": 76337, "tid": -914061504, "ts": 1716454224933057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224978104, "dur": 4,
    "args": {
      "External id": 220748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220748,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 220748, "pid": 5, "tid": 7, "ts": 1716454224978104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933079, "dur": 6,
    "args": {
      "External id": 220748,
      "cbid": 211, "correlation": 220748
    }
  },
  {
    "ph": "s", "id": 220748, "pid": 76337, "tid": -914061504, "ts": 1716454224933079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224978109, "dur": 30,
    "args": {
      "External id": 220751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220751, "pid": 5, "tid": 7, "ts": 1716454224978109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933098, "dur": 7,
    "args": {
      "External id": 220751,
      "cbid": 211, "correlation": 220751
    }
  },
  {
    "ph": "s", "id": 220751, "pid": 76337, "tid": -914061504, "ts": 1716454224933098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224978140, "dur": 21,
    "args": {
      "External id": 220760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220760,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220760, "pid": 5, "tid": 7, "ts": 1716454224978140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933141, "dur": 9,
    "args": {
      "External id": 220760,
      "cbid": 211, "correlation": 220760
    }
  },
  {
    "ph": "s", "id": 220760, "pid": 76337, "tid": -914061504, "ts": 1716454224933141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933203, "dur": 0,
    "args": {
      "External id": 220770,
      "cbid": 317, "correlation": 220770
    }
  },
  {
    "ph": "f", "id": 220770, "pid": 76337, "tid": -914061504, "ts": 1716454224933203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933204, "dur": 0,
    "args": {
      "External id": 220771,
      "cbid": 203, "correlation": 220771
    }
  },
  {
    "ph": "f", "id": 220771, "pid": 76337, "tid": -914061504, "ts": 1716454224933204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933205, "dur": 0,
    "args": {
      "External id": 220772,
      "cbid": 205, "correlation": 220772
    }
  },
  {
    "ph": "f", "id": 220772, "pid": 76337, "tid": -914061504, "ts": 1716454224933205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224978163, "dur": 23,
    "args": {
      "External id": 220776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220776, "pid": 5, "tid": 7, "ts": 1716454224978163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933222, "dur": 12,
    "args": {
      "External id": 220776,
      "cbid": 211, "correlation": 220776
    }
  },
  {
    "ph": "s", "id": 220776, "pid": 76337, "tid": -914061504, "ts": 1716454224933222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224978187, "dur": 44,
    "args": {
      "External id": 220778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220778, "pid": 5, "tid": 7, "ts": 1716454224978187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933237, "dur": 6,
    "args": {
      "External id": 220778,
      "cbid": 211, "correlation": 220778
    }
  },
  {
    "ph": "s", "id": 220778, "pid": 76337, "tid": -914061504, "ts": 1716454224933237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224978233, "dur": 657,
    "args": {
      "External id": 220780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220780,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220780, "pid": 5, "tid": 7, "ts": 1716454224978233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933251, "dur": 9,
    "args": {
      "External id": 220780,
      "cbid": 211, "correlation": 220780
    }
  },
  {
    "ph": "s", "id": 220780, "pid": 76337, "tid": -914061504, "ts": 1716454224933251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224978892, "dur": 21,
    "args": {
      "External id": 220782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220782,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220782, "pid": 5, "tid": 7, "ts": 1716454224978892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933264, "dur": 5,
    "args": {
      "External id": 220782,
      "cbid": 211, "correlation": 220782
    }
  },
  {
    "ph": "s", "id": 220782, "pid": 76337, "tid": -914061504, "ts": 1716454224933264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224978914, "dur": 33,
    "args": {
      "External id": 220788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220788, "pid": 5, "tid": 7, "ts": 1716454224978914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933292, "dur": 8,
    "args": {
      "External id": 220788,
      "cbid": 211, "correlation": 220788
    }
  },
  {
    "ph": "s", "id": 220788, "pid": 76337, "tid": -914061504, "ts": 1716454224933292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933350, "dur": 0,
    "args": {
      "External id": 220798,
      "cbid": 317, "correlation": 220798
    }
  },
  {
    "ph": "f", "id": 220798, "pid": 76337, "tid": -914061504, "ts": 1716454224933350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933351, "dur": 0,
    "args": {
      "External id": 220799,
      "cbid": 203, "correlation": 220799
    }
  },
  {
    "ph": "f", "id": 220799, "pid": 76337, "tid": -914061504, "ts": 1716454224933351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933352, "dur": 0,
    "args": {
      "External id": 220800,
      "cbid": 205, "correlation": 220800
    }
  },
  {
    "ph": "f", "id": 220800, "pid": 76337, "tid": -914061504, "ts": 1716454224933352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933374, "dur": 1,
    "args": {
      "External id": 220804,
      "cbid": 251, "correlation": 220804
    }
  },
  {
    "ph": "f", "id": 220804, "pid": 76337, "tid": -914061504, "ts": 1716454224933374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933376, "dur": 0,
    "args": {
      "External id": 220805,
      "cbid": 251, "correlation": 220805
    }
  },
  {
    "ph": "f", "id": 220805, "pid": 76337, "tid": -914061504, "ts": 1716454224933376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933376, "dur": 0,
    "args": {
      "External id": 220806,
      "cbid": 251, "correlation": 220806
    }
  },
  {
    "ph": "f", "id": 220806, "pid": 76337, "tid": -914061504, "ts": 1716454224933376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933377, "dur": 0,
    "args": {
      "External id": 220807,
      "cbid": 251, "correlation": 220807
    }
  },
  {
    "ph": "f", "id": 220807, "pid": 76337, "tid": -914061504, "ts": 1716454224933377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933378, "dur": 0,
    "args": {
      "External id": 220808,
      "cbid": 251, "correlation": 220808
    }
  },
  {
    "ph": "f", "id": 220808, "pid": 76337, "tid": -914061504, "ts": 1716454224933378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933379, "dur": 0,
    "args": {
      "External id": 220809,
      "cbid": 251, "correlation": 220809
    }
  },
  {
    "ph": "f", "id": 220809, "pid": 76337, "tid": -914061504, "ts": 1716454224933379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933379, "dur": 0,
    "args": {
      "External id": 220810,
      "cbid": 251, "correlation": 220810
    }
  },
  {
    "ph": "f", "id": 220810, "pid": 76337, "tid": -914061504, "ts": 1716454224933379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933380, "dur": 0,
    "args": {
      "External id": 220811,
      "cbid": 251, "correlation": 220811
    }
  },
  {
    "ph": "f", "id": 220811, "pid": 76337, "tid": -914061504, "ts": 1716454224933380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933381, "dur": 0,
    "args": {
      "External id": 220812,
      "cbid": 251, "correlation": 220812
    }
  },
  {
    "ph": "f", "id": 220812, "pid": 76337, "tid": -914061504, "ts": 1716454224933381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224978949, "dur": 53,
    "args": {
      "External id": 220813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220813,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220813, "pid": 5, "tid": 7, "ts": 1716454224978949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933383, "dur": 12,
    "args": {
      "External id": 220813,
      "cbid": 211, "correlation": 220813
    }
  },
  {
    "ph": "s", "id": 220813, "pid": 76337, "tid": -914061504, "ts": 1716454224933383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224979003, "dur": 33,
    "args": {
      "External id": 220819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220819, "pid": 5, "tid": 7, "ts": 1716454224979003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933416, "dur": 10,
    "args": {
      "External id": 220819,
      "cbid": 211, "correlation": 220819
    }
  },
  {
    "ph": "s", "id": 220819, "pid": 76337, "tid": -914061504, "ts": 1716454224933416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224979037, "dur": 27,
    "args": {
      "External id": 220827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220827,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220827, "pid": 5, "tid": 7, "ts": 1716454224979037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933447, "dur": 8,
    "args": {
      "External id": 220827,
      "cbid": 211, "correlation": 220827
    }
  },
  {
    "ph": "s", "id": 220827, "pid": 76337, "tid": -914061504, "ts": 1716454224933447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224979066, "dur": 19,
    "args": {
      "External id": 220835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220835,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220835, "pid": 5, "tid": 7, "ts": 1716454224979066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933476, "dur": 9,
    "args": {
      "External id": 220835,
      "cbid": 211, "correlation": 220835
    }
  },
  {
    "ph": "s", "id": 220835, "pid": 76337, "tid": -914061504, "ts": 1716454224933476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224979086, "dur": 30,
    "args": {
      "External id": 220855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220855,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 220855, "pid": 5, "tid": 7, "ts": 1716454224979086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933559, "dur": 12,
    "args": {
      "External id": 220855,
      "cbid": 211, "correlation": 220855
    }
  },
  {
    "ph": "s", "id": 220855, "pid": 76337, "tid": -914061504, "ts": 1716454224933559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224979118, "dur": 4,
    "args": {
      "External id": 220867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220867,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 220867, "pid": 5, "tid": 7, "ts": 1716454224979118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933580, "dur": 7,
    "args": {
      "External id": 220867,
      "cbid": 211, "correlation": 220867
    }
  },
  {
    "ph": "s", "id": 220867, "pid": 76337, "tid": -914061504, "ts": 1716454224933580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224979123, "dur": 31,
    "args": {
      "External id": 220870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220870, "pid": 5, "tid": 7, "ts": 1716454224979123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933599, "dur": 6,
    "args": {
      "External id": 220870,
      "cbid": 211, "correlation": 220870
    }
  },
  {
    "ph": "s", "id": 220870, "pid": 76337, "tid": -914061504, "ts": 1716454224933599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933656, "dur": 0,
    "args": {
      "External id": 220881,
      "cbid": 317, "correlation": 220881
    }
  },
  {
    "ph": "f", "id": 220881, "pid": 76337, "tid": -914061504, "ts": 1716454224933656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933657, "dur": 0,
    "args": {
      "External id": 220882,
      "cbid": 203, "correlation": 220882
    }
  },
  {
    "ph": "f", "id": 220882, "pid": 76337, "tid": -914061504, "ts": 1716454224933657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933657, "dur": 0,
    "args": {
      "External id": 220883,
      "cbid": 205, "correlation": 220883
    }
  },
  {
    "ph": "f", "id": 220883, "pid": 76337, "tid": -914061504, "ts": 1716454224933657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224979155, "dur": 22,
    "args": {
      "External id": 220887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220887,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220887, "pid": 5, "tid": 7, "ts": 1716454224979155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933672, "dur": 11,
    "args": {
      "External id": 220887,
      "cbid": 211, "correlation": 220887
    }
  },
  {
    "ph": "s", "id": 220887, "pid": 76337, "tid": -914061504, "ts": 1716454224933672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224979178, "dur": 122,
    "args": {
      "External id": 220889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220889,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220889, "pid": 5, "tid": 7, "ts": 1716454224979178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933692, "dur": 9,
    "args": {
      "External id": 220889,
      "cbid": 211, "correlation": 220889
    }
  },
  {
    "ph": "s", "id": 220889, "pid": 76337, "tid": -914061504, "ts": 1716454224933692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224979301, "dur": 24,
    "args": {
      "External id": 220891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220891,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220891, "pid": 5, "tid": 7, "ts": 1716454224979301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933704, "dur": 5,
    "args": {
      "External id": 220891,
      "cbid": 211, "correlation": 220891
    }
  },
  {
    "ph": "s", "id": 220891, "pid": 76337, "tid": -914061504, "ts": 1716454224933704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224979326, "dur": 33,
    "args": {
      "External id": 220897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220897,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220897, "pid": 5, "tid": 7, "ts": 1716454224979326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933732, "dur": 9,
    "args": {
      "External id": 220897,
      "cbid": 211, "correlation": 220897
    }
  },
  {
    "ph": "s", "id": 220897, "pid": 76337, "tid": -914061504, "ts": 1716454224933732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224979361, "dur": 177,
    "args": {
      "External id": 220906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220906,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220906, "pid": 5, "tid": 7, "ts": 1716454224979361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933814, "dur": 14,
    "args": {
      "External id": 220906,
      "cbid": 211, "correlation": 220906
    }
  },
  {
    "ph": "s", "id": 220906, "pid": 76337, "tid": -914061504, "ts": 1716454224933814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224979539, "dur": 65,
    "args": {
      "External id": 220928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220928,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 220928, "pid": 5, "tid": 7, "ts": 1716454224979539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933870, "dur": 10,
    "args": {
      "External id": 220928,
      "cbid": 211, "correlation": 220928
    }
  },
  {
    "ph": "s", "id": 220928, "pid": 76337, "tid": -914061504, "ts": 1716454224933870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933962, "dur": 1,
    "args": {
      "External id": 220939,
      "cbid": 251, "correlation": 220939
    }
  },
  {
    "ph": "f", "id": 220939, "pid": 76337, "tid": -914061504, "ts": 1716454224933962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224979606, "dur": 156,
    "args": {
      "External id": 220940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220940,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220940, "pid": 5, "tid": 7, "ts": 1716454224979606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224933967, "dur": 21,
    "args": {
      "External id": 220940,
      "cbid": 211, "correlation": 220940
    }
  },
  {
    "ph": "s", "id": 220940, "pid": 76337, "tid": -914061504, "ts": 1716454224933967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934046, "dur": 1,
    "args": {
      "External id": 220951,
      "cbid": 251, "correlation": 220951
    }
  },
  {
    "ph": "f", "id": 220951, "pid": 76337, "tid": -914061504, "ts": 1716454224934046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224979763, "dur": 147,
    "args": {
      "External id": 220952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220952,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220952, "pid": 5, "tid": 7, "ts": 1716454224979763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934050, "dur": 11,
    "args": {
      "External id": 220952,
      "cbid": 211, "correlation": 220952
    }
  },
  {
    "ph": "s", "id": 220952, "pid": 76337, "tid": -914061504, "ts": 1716454224934050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934116, "dur": 1,
    "args": {
      "External id": 220963,
      "cbid": 251, "correlation": 220963
    }
  },
  {
    "ph": "f", "id": 220963, "pid": 76337, "tid": -914061504, "ts": 1716454224934116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224979912, "dur": 145,
    "args": {
      "External id": 220964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220964,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 220964, "pid": 5, "tid": 7, "ts": 1716454224979912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934120, "dur": 10,
    "args": {
      "External id": 220964,
      "cbid": 211, "correlation": 220964
    }
  },
  {
    "ph": "s", "id": 220964, "pid": 76337, "tid": -914061504, "ts": 1716454224934120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224980058, "dur": 1955,
    "args": {
      "External id": 220985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 220985,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 220985, "pid": 5, "tid": 7, "ts": 1716454224980058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934203, "dur": 14,
    "args": {
      "External id": 220985,
      "cbid": 211, "correlation": 220985
    }
  },
  {
    "ph": "s", "id": 220985, "pid": 76337, "tid": -914061504, "ts": 1716454224934203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934303, "dur": 1,
    "args": {
      "External id": 221003,
      "cbid": 251, "correlation": 221003
    }
  },
  {
    "ph": "f", "id": 221003, "pid": 76337, "tid": -914061504, "ts": 1716454224934303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224982015, "dur": 147,
    "args": {
      "External id": 221005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221005,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 221005, "pid": 5, "tid": 7, "ts": 1716454224982015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934309, "dur": 13,
    "args": {
      "External id": 221005,
      "cbid": 211, "correlation": 221005
    }
  },
  {
    "ph": "s", "id": 221005, "pid": 76337, "tid": -914061504, "ts": 1716454224934309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224982163, "dur": 35,
    "args": {
      "External id": 221013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221013,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221013, "pid": 5, "tid": 7, "ts": 1716454224982163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934381, "dur": 13,
    "args": {
      "External id": 221013,
      "cbid": 211, "correlation": 221013
    }
  },
  {
    "ph": "s", "id": 221013, "pid": 76337, "tid": -914061504, "ts": 1716454224934381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224982200, "dur": 50,
    "args": {
      "External id": 221021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221021,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221021, "pid": 5, "tid": 7, "ts": 1716454224982200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934420, "dur": 9,
    "args": {
      "External id": 221021,
      "cbid": 211, "correlation": 221021
    }
  },
  {
    "ph": "s", "id": 221021, "pid": 76337, "tid": -914061504, "ts": 1716454224934420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224982252, "dur": 30,
    "args": {
      "External id": 221032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221032, "pid": 5, "tid": 7, "ts": 1716454224982252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934494, "dur": 12,
    "args": {
      "External id": 221032,
      "cbid": 211, "correlation": 221032
    }
  },
  {
    "ph": "s", "id": 221032, "pid": 76337, "tid": -914061504, "ts": 1716454224934494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224982283, "dur": 34,
    "args": {
      "External id": 221054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221054,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221054, "pid": 5, "tid": 7, "ts": 1716454224982283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934524, "dur": 7,
    "args": {
      "External id": 221054,
      "cbid": 211, "correlation": 221054
    }
  },
  {
    "ph": "s", "id": 221054, "pid": 76337, "tid": -914061504, "ts": 1716454224934524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934609, "dur": 1,
    "args": {
      "External id": 221065,
      "cbid": 251, "correlation": 221065
    }
  },
  {
    "ph": "f", "id": 221065, "pid": 76337, "tid": -914061504, "ts": 1716454224934609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224982318, "dur": 88,
    "args": {
      "External id": 221066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221066,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221066, "pid": 5, "tid": 7, "ts": 1716454224982318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934614, "dur": 12,
    "args": {
      "External id": 221066,
      "cbid": 211, "correlation": 221066
    }
  },
  {
    "ph": "s", "id": 221066, "pid": 76337, "tid": -914061504, "ts": 1716454224934614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934684, "dur": 1,
    "args": {
      "External id": 221077,
      "cbid": 251, "correlation": 221077
    }
  },
  {
    "ph": "f", "id": 221077, "pid": 76337, "tid": -914061504, "ts": 1716454224934684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934688, "dur": 0,
    "args": {
      "External id": 221078,
      "cbid": 251, "correlation": 221078
    }
  },
  {
    "ph": "f", "id": 221078, "pid": 76337, "tid": -914061504, "ts": 1716454224934688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224982408, "dur": 11,
    "args": {
      "External id": 221079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221079,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 221079, "pid": 5, "tid": 7, "ts": 1716454224982408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934690, "dur": 11,
    "args": {
      "External id": 221079,
      "cbid": 211, "correlation": 221079
    }
  },
  {
    "ph": "s", "id": 221079, "pid": 76337, "tid": -914061504, "ts": 1716454224934690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224982420, "dur": 5,
    "args": {
      "External id": 221081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221081,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 221081, "pid": 5, "tid": 7, "ts": 1716454224982420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934703, "dur": 8,
    "args": {
      "External id": 221081,
      "cbid": 211, "correlation": 221081
    }
  },
  {
    "ph": "s", "id": 221081, "pid": 76337, "tid": -914061504, "ts": 1716454224934703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934762, "dur": 1,
    "args": {
      "External id": 221092,
      "cbid": 251, "correlation": 221092
    }
  },
  {
    "ph": "f", "id": 221092, "pid": 76337, "tid": -914061504, "ts": 1716454224934762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934765, "dur": 0,
    "args": {
      "External id": 221093,
      "cbid": 251, "correlation": 221093
    }
  },
  {
    "ph": "f", "id": 221093, "pid": 76337, "tid": -914061504, "ts": 1716454224934765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224982426, "dur": 7,
    "args": {
      "External id": 221094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221094,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 221094, "pid": 5, "tid": 7, "ts": 1716454224982426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934767, "dur": 11,
    "args": {
      "External id": 221094,
      "cbid": 211, "correlation": 221094
    }
  },
  {
    "ph": "s", "id": 221094, "pid": 76337, "tid": -914061504, "ts": 1716454224934767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224982434, "dur": 3,
    "args": {
      "External id": 221096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 221096, "pid": 5, "tid": 7, "ts": 1716454224982434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934780, "dur": 6,
    "args": {
      "External id": 221096,
      "cbid": 211, "correlation": 221096
    }
  },
  {
    "ph": "s", "id": 221096, "pid": 76337, "tid": -914061504, "ts": 1716454224934780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224982439, "dur": 92,
    "args": {
      "External id": 221117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221117,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 221117, "pid": 5, "tid": 7, "ts": 1716454224982439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934854, "dur": 12,
    "args": {
      "External id": 221117,
      "cbid": 211, "correlation": 221117
    }
  },
  {
    "ph": "s", "id": 221117, "pid": 76337, "tid": -914061504, "ts": 1716454224934854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934950, "dur": 1,
    "args": {
      "External id": 221135,
      "cbid": 251, "correlation": 221135
    }
  },
  {
    "ph": "f", "id": 221135, "pid": 76337, "tid": -914061504, "ts": 1716454224934950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224982532, "dur": 98,
    "args": {
      "External id": 221137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221137,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221137, "pid": 5, "tid": 7, "ts": 1716454224982532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224934956, "dur": 13,
    "args": {
      "External id": 221137,
      "cbid": 211, "correlation": 221137
    }
  },
  {
    "ph": "s", "id": 221137, "pid": 76337, "tid": -914061504, "ts": 1716454224934956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224982631, "dur": 19,
    "args": {
      "External id": 221145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221145,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221145, "pid": 5, "tid": 7, "ts": 1716454224982631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935032, "dur": 13,
    "args": {
      "External id": 221145,
      "cbid": 211, "correlation": 221145
    }
  },
  {
    "ph": "s", "id": 221145, "pid": 76337, "tid": -914061504, "ts": 1716454224935032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224982652, "dur": 37,
    "args": {
      "External id": 221153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221153,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221153, "pid": 5, "tid": 7, "ts": 1716454224982652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935075, "dur": 9,
    "args": {
      "External id": 221153,
      "cbid": 211, "correlation": 221153
    }
  },
  {
    "ph": "s", "id": 221153, "pid": 76337, "tid": -914061504, "ts": 1716454224935075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224982690, "dur": 35,
    "args": {
      "External id": 221175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221175,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221175, "pid": 5, "tid": 7, "ts": 1716454224982690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935127, "dur": 11,
    "args": {
      "External id": 221175,
      "cbid": 211, "correlation": 221175
    }
  },
  {
    "ph": "s", "id": 221175, "pid": 76337, "tid": -914061504, "ts": 1716454224935127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935216, "dur": 1,
    "args": {
      "External id": 221191,
      "cbid": 251, "correlation": 221191
    }
  },
  {
    "ph": "f", "id": 221191, "pid": 76337, "tid": -914061504, "ts": 1716454224935216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935221, "dur": 0,
    "args": {
      "External id": 221193,
      "cbid": 251, "correlation": 221193
    }
  },
  {
    "ph": "f", "id": 221193, "pid": 76337, "tid": -914061504, "ts": 1716454224935221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224982726, "dur": 536,
    "args": {
      "External id": 221194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221194,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 221194, "pid": 5, "tid": 7, "ts": 1716454224982726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935224, "dur": 13,
    "args": {
      "External id": 221194,
      "cbid": 211, "correlation": 221194
    }
  },
  {
    "ph": "s", "id": 221194, "pid": 76337, "tid": -914061504, "ts": 1716454224935224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224983264, "dur": 124,
    "args": {
      "External id": 221202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221202, "pid": 5, "tid": 7, "ts": 1716454224983264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935291, "dur": 12,
    "args": {
      "External id": 221202,
      "cbid": 211, "correlation": 221202
    }
  },
  {
    "ph": "s", "id": 221202, "pid": 76337, "tid": -914061504, "ts": 1716454224935291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224983389, "dur": 126,
    "args": {
      "External id": 221210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221210, "pid": 5, "tid": 7, "ts": 1716454224983389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935322, "dur": 8,
    "args": {
      "External id": 221210,
      "cbid": 211, "correlation": 221210
    }
  },
  {
    "ph": "s", "id": 221210, "pid": 76337, "tid": -914061504, "ts": 1716454224935322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935400, "dur": 1,
    "args": {
      "External id": 221226,
      "cbid": 251, "correlation": 221226
    }
  },
  {
    "ph": "f", "id": 221226, "pid": 76337, "tid": -914061504, "ts": 1716454224935400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224983517, "dur": 300,
    "args": {
      "External id": 221228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221228,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221228, "pid": 5, "tid": 7, "ts": 1716454224983517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935406, "dur": 12,
    "args": {
      "External id": 221228,
      "cbid": 211, "correlation": 221228
    }
  },
  {
    "ph": "s", "id": 221228, "pid": 76337, "tid": -914061504, "ts": 1716454224935406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224983818, "dur": 27,
    "args": {
      "External id": 221236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221236,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221236, "pid": 5, "tid": 7, "ts": 1716454224983818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935447, "dur": 9,
    "args": {
      "External id": 221236,
      "cbid": 211, "correlation": 221236
    }
  },
  {
    "ph": "s", "id": 221236, "pid": 76337, "tid": -914061504, "ts": 1716454224935447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224983846, "dur": 80,
    "args": {
      "External id": 221247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221247, "pid": 5, "tid": 7, "ts": 1716454224983846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935515, "dur": 12,
    "args": {
      "External id": 221247,
      "cbid": 211, "correlation": 221247
    }
  },
  {
    "ph": "s", "id": 221247, "pid": 76337, "tid": -914061504, "ts": 1716454224935515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935579, "dur": 0,
    "args": {
      "External id": 221259,
      "cbid": 317, "correlation": 221259
    }
  },
  {
    "ph": "f", "id": 221259, "pid": 76337, "tid": -914061504, "ts": 1716454224935579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935580, "dur": 0,
    "args": {
      "External id": 221260,
      "cbid": 203, "correlation": 221260
    }
  },
  {
    "ph": "f", "id": 221260, "pid": 76337, "tid": -914061504, "ts": 1716454224935580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935581, "dur": 0,
    "args": {
      "External id": 221261,
      "cbid": 205, "correlation": 221261
    }
  },
  {
    "ph": "f", "id": 221261, "pid": 76337, "tid": -914061504, "ts": 1716454224935581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224983928, "dur": 22,
    "args": {
      "External id": 221265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221265,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221265, "pid": 5, "tid": 7, "ts": 1716454224983928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935599, "dur": 12,
    "args": {
      "External id": 221265,
      "cbid": 211, "correlation": 221265
    }
  },
  {
    "ph": "s", "id": 221265, "pid": 76337, "tid": -914061504, "ts": 1716454224935599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224983951, "dur": 118,
    "args": {
      "External id": 221267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221267,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221267, "pid": 5, "tid": 7, "ts": 1716454224983951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935617, "dur": 6,
    "args": {
      "External id": 221267,
      "cbid": 211, "correlation": 221267
    }
  },
  {
    "ph": "s", "id": 221267, "pid": 76337, "tid": -914061504, "ts": 1716454224935617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224984070, "dur": 23,
    "args": {
      "External id": 221269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221269,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221269, "pid": 5, "tid": 7, "ts": 1716454224984070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935627, "dur": 6,
    "args": {
      "External id": 221269,
      "cbid": 211, "correlation": 221269
    }
  },
  {
    "ph": "s", "id": 221269, "pid": 76337, "tid": -914061504, "ts": 1716454224935627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224984095, "dur": 33,
    "args": {
      "External id": 221275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221275,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221275, "pid": 5, "tid": 7, "ts": 1716454224984095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935656, "dur": 8,
    "args": {
      "External id": 221275,
      "cbid": 211, "correlation": 221275
    }
  },
  {
    "ph": "s", "id": 221275, "pid": 76337, "tid": -914061504, "ts": 1716454224935656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224984129, "dur": 27,
    "args": {
      "External id": 221283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221283,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221283, "pid": 5, "tid": 7, "ts": 1716454224984129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935688, "dur": 8,
    "args": {
      "External id": 221283,
      "cbid": 211, "correlation": 221283
    }
  },
  {
    "ph": "s", "id": 221283, "pid": 76337, "tid": -914061504, "ts": 1716454224935688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224984157, "dur": 31,
    "args": {
      "External id": 221303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221303,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 221303, "pid": 5, "tid": 7, "ts": 1716454224984157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935760, "dur": 11,
    "args": {
      "External id": 221303,
      "cbid": 211, "correlation": 221303
    }
  },
  {
    "ph": "s", "id": 221303, "pid": 76337, "tid": -914061504, "ts": 1716454224935760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224984189, "dur": 4,
    "args": {
      "External id": 221315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221315,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 221315, "pid": 5, "tid": 7, "ts": 1716454224984189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935781, "dur": 7,
    "args": {
      "External id": 221315,
      "cbid": 211, "correlation": 221315
    }
  },
  {
    "ph": "s", "id": 221315, "pid": 76337, "tid": -914061504, "ts": 1716454224935781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224984195, "dur": 30,
    "args": {
      "External id": 221318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221318, "pid": 5, "tid": 7, "ts": 1716454224984195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935799, "dur": 6,
    "args": {
      "External id": 221318,
      "cbid": 211, "correlation": 221318
    }
  },
  {
    "ph": "s", "id": 221318, "pid": 76337, "tid": -914061504, "ts": 1716454224935799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224984226, "dur": 21,
    "args": {
      "External id": 221327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221327,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221327, "pid": 5, "tid": 7, "ts": 1716454224984226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935839, "dur": 9,
    "args": {
      "External id": 221327,
      "cbid": 211, "correlation": 221327
    }
  },
  {
    "ph": "s", "id": 221327, "pid": 76337, "tid": -914061504, "ts": 1716454224935839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935890, "dur": 0,
    "args": {
      "External id": 221337,
      "cbid": 317, "correlation": 221337
    }
  },
  {
    "ph": "f", "id": 221337, "pid": 76337, "tid": -914061504, "ts": 1716454224935890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935891, "dur": 0,
    "args": {
      "External id": 221338,
      "cbid": 203, "correlation": 221338
    }
  },
  {
    "ph": "f", "id": 221338, "pid": 76337, "tid": -914061504, "ts": 1716454224935891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935891, "dur": 0,
    "args": {
      "External id": 221339,
      "cbid": 205, "correlation": 221339
    }
  },
  {
    "ph": "f", "id": 221339, "pid": 76337, "tid": -914061504, "ts": 1716454224935891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224984248, "dur": 22,
    "args": {
      "External id": 221343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221343, "pid": 5, "tid": 7, "ts": 1716454224984248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935904, "dur": 11,
    "args": {
      "External id": 221343,
      "cbid": 211, "correlation": 221343
    }
  },
  {
    "ph": "s", "id": 221343, "pid": 76337, "tid": -914061504, "ts": 1716454224935904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224984272, "dur": 43,
    "args": {
      "External id": 221345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221345,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221345, "pid": 5, "tid": 7, "ts": 1716454224984272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935917, "dur": 6,
    "args": {
      "External id": 221345,
      "cbid": 211, "correlation": 221345
    }
  },
  {
    "ph": "s", "id": 221345, "pid": 76337, "tid": -914061504, "ts": 1716454224935917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224984317, "dur": 649,
    "args": {
      "External id": 221347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221347,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221347, "pid": 5, "tid": 7, "ts": 1716454224984317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935929, "dur": 6,
    "args": {
      "External id": 221347,
      "cbid": 211, "correlation": 221347
    }
  },
  {
    "ph": "s", "id": 221347, "pid": 76337, "tid": -914061504, "ts": 1716454224935929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224984967, "dur": 22,
    "args": {
      "External id": 221349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221349, "pid": 5, "tid": 7, "ts": 1716454224984967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935939, "dur": 5,
    "args": {
      "External id": 221349,
      "cbid": 211, "correlation": 221349
    }
  },
  {
    "ph": "s", "id": 221349, "pid": 76337, "tid": -914061504, "ts": 1716454224935939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224984990, "dur": 32,
    "args": {
      "External id": 221355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221355, "pid": 5, "tid": 7, "ts": 1716454224984990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224935966, "dur": 23,
    "args": {
      "External id": 221355,
      "cbid": 211, "correlation": 221355
    }
  },
  {
    "ph": "s", "id": 221355, "pid": 76337, "tid": -914061504, "ts": 1716454224935966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224985024, "dur": 3,
    "args": {
      "External id": 221363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221363,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 221363, "pid": 5, "tid": 7, "ts": 1716454224985024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936025, "dur": 10,
    "args": {
      "External id": 221363,
      "cbid": 211, "correlation": 221363
    }
  },
  {
    "ph": "s", "id": 221363, "pid": 76337, "tid": -914061504, "ts": 1716454224936025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936091, "dur": 1,
    "args": {
      "External id": 221379,
      "cbid": 251, "correlation": 221379
    }
  },
  {
    "ph": "f", "id": 221379, "pid": 76337, "tid": -914061504, "ts": 1716454224936091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936096, "dur": 0,
    "args": {
      "External id": 221381,
      "cbid": 251, "correlation": 221381
    }
  },
  {
    "ph": "f", "id": 221381, "pid": 76337, "tid": -914061504, "ts": 1716454224936096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224985028, "dur": 12,
    "args": {
      "External id": 221382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221382,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 221382, "pid": 5, "tid": 7, "ts": 1716454224985028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936098, "dur": 12,
    "args": {
      "External id": 221382,
      "cbid": 211, "correlation": 221382
    }
  },
  {
    "ph": "s", "id": 221382, "pid": 76337, "tid": -914061504, "ts": 1716454224936098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224985042, "dur": 5,
    "args": {
      "External id": 221384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221384,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 221384, "pid": 5, "tid": 7, "ts": 1716454224985042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936111, "dur": 5,
    "args": {
      "External id": 221384,
      "cbid": 211, "correlation": 221384
    }
  },
  {
    "ph": "s", "id": 221384, "pid": 76337, "tid": -914061504, "ts": 1716454224936111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224985048, "dur": 29,
    "args": {
      "External id": 221394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221394, "pid": 5, "tid": 7, "ts": 1716454224985048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936170, "dur": 12,
    "args": {
      "External id": 221394,
      "cbid": 211, "correlation": 221394
    }
  },
  {
    "ph": "s", "id": 221394, "pid": 76337, "tid": -914061504, "ts": 1716454224936170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224985078, "dur": 31,
    "args": {
      "External id": 221414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221414,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 221414, "pid": 5, "tid": 7, "ts": 1716454224985078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936235, "dur": 11,
    "args": {
      "External id": 221414,
      "cbid": 211, "correlation": 221414
    }
  },
  {
    "ph": "s", "id": 221414, "pid": 76337, "tid": -914061504, "ts": 1716454224936235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224985111, "dur": 4,
    "args": {
      "External id": 221426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221426,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 221426, "pid": 5, "tid": 7, "ts": 1716454224985111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936256, "dur": 6,
    "args": {
      "External id": 221426,
      "cbid": 211, "correlation": 221426
    }
  },
  {
    "ph": "s", "id": 221426, "pid": 76337, "tid": -914061504, "ts": 1716454224936256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224985116, "dur": 29,
    "args": {
      "External id": 221429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221429, "pid": 5, "tid": 7, "ts": 1716454224985116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936274, "dur": 6,
    "args": {
      "External id": 221429,
      "cbid": 211, "correlation": 221429
    }
  },
  {
    "ph": "s", "id": 221429, "pid": 76337, "tid": -914061504, "ts": 1716454224936274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224985147, "dur": 20,
    "args": {
      "External id": 221438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221438,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221438, "pid": 5, "tid": 7, "ts": 1716454224985147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936314, "dur": 9,
    "args": {
      "External id": 221438,
      "cbid": 211, "correlation": 221438
    }
  },
  {
    "ph": "s", "id": 221438, "pid": 76337, "tid": -914061504, "ts": 1716454224936314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936375, "dur": 0,
    "args": {
      "External id": 221448,
      "cbid": 317, "correlation": 221448
    }
  },
  {
    "ph": "f", "id": 221448, "pid": 76337, "tid": -914061504, "ts": 1716454224936375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936376, "dur": 0,
    "args": {
      "External id": 221449,
      "cbid": 203, "correlation": 221449
    }
  },
  {
    "ph": "f", "id": 221449, "pid": 76337, "tid": -914061504, "ts": 1716454224936376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936377, "dur": 0,
    "args": {
      "External id": 221450,
      "cbid": 205, "correlation": 221450
    }
  },
  {
    "ph": "f", "id": 221450, "pid": 76337, "tid": -914061504, "ts": 1716454224936377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224985168, "dur": 23,
    "args": {
      "External id": 221454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221454, "pid": 5, "tid": 7, "ts": 1716454224985168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936391, "dur": 11,
    "args": {
      "External id": 221454,
      "cbid": 211, "correlation": 221454
    }
  },
  {
    "ph": "s", "id": 221454, "pid": 76337, "tid": -914061504, "ts": 1716454224936391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224985192, "dur": 44,
    "args": {
      "External id": 221456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221456, "pid": 5, "tid": 7, "ts": 1716454224985192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936405, "dur": 5,
    "args": {
      "External id": 221456,
      "cbid": 211, "correlation": 221456
    }
  },
  {
    "ph": "s", "id": 221456, "pid": 76337, "tid": -914061504, "ts": 1716454224936405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224985237, "dur": 638,
    "args": {
      "External id": 221458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221458,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221458, "pid": 5, "tid": 7, "ts": 1716454224985237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936416, "dur": 7,
    "args": {
      "External id": 221458,
      "cbid": 211, "correlation": 221458
    }
  },
  {
    "ph": "s", "id": 221458, "pid": 76337, "tid": -914061504, "ts": 1716454224936416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224985876, "dur": 21,
    "args": {
      "External id": 221460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221460,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221460, "pid": 5, "tid": 7, "ts": 1716454224985876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936426, "dur": 5,
    "args": {
      "External id": 221460,
      "cbid": 211, "correlation": 221460
    }
  },
  {
    "ph": "s", "id": 221460, "pid": 76337, "tid": -914061504, "ts": 1716454224936426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224985898, "dur": 33,
    "args": {
      "External id": 221466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221466, "pid": 5, "tid": 7, "ts": 1716454224985898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936454, "dur": 8,
    "args": {
      "External id": 221466,
      "cbid": 211, "correlation": 221466
    }
  },
  {
    "ph": "s", "id": 221466, "pid": 76337, "tid": -914061504, "ts": 1716454224936454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224985932, "dur": 27,
    "args": {
      "External id": 221474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221474,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221474, "pid": 5, "tid": 7, "ts": 1716454224985932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936486, "dur": 9,
    "args": {
      "External id": 221474,
      "cbid": 211, "correlation": 221474
    }
  },
  {
    "ph": "s", "id": 221474, "pid": 76337, "tid": -914061504, "ts": 1716454224936486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224985960, "dur": 20,
    "args": {
      "External id": 221482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221482,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221482, "pid": 5, "tid": 7, "ts": 1716454224985960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936514, "dur": 8,
    "args": {
      "External id": 221482,
      "cbid": 211, "correlation": 221482
    }
  },
  {
    "ph": "s", "id": 221482, "pid": 76337, "tid": -914061504, "ts": 1716454224936514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224985982, "dur": 29,
    "args": {
      "External id": 221502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221502,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 221502, "pid": 5, "tid": 7, "ts": 1716454224985982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936592, "dur": 12,
    "args": {
      "External id": 221502,
      "cbid": 211, "correlation": 221502
    }
  },
  {
    "ph": "s", "id": 221502, "pid": 76337, "tid": -914061504, "ts": 1716454224936592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224986012, "dur": 4,
    "args": {
      "External id": 221514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221514,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 221514, "pid": 5, "tid": 7, "ts": 1716454224986012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936613, "dur": 6,
    "args": {
      "External id": 221514,
      "cbid": 211, "correlation": 221514
    }
  },
  {
    "ph": "s", "id": 221514, "pid": 76337, "tid": -914061504, "ts": 1716454224936613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224986017, "dur": 29,
    "args": {
      "External id": 221517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221517, "pid": 5, "tid": 7, "ts": 1716454224986017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936631, "dur": 6,
    "args": {
      "External id": 221517,
      "cbid": 211, "correlation": 221517
    }
  },
  {
    "ph": "s", "id": 221517, "pid": 76337, "tid": -914061504, "ts": 1716454224936631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936688, "dur": 0,
    "args": {
      "External id": 221528,
      "cbid": 317, "correlation": 221528
    }
  },
  {
    "ph": "f", "id": 221528, "pid": 76337, "tid": -914061504, "ts": 1716454224936688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936688, "dur": 0,
    "args": {
      "External id": 221529,
      "cbid": 203, "correlation": 221529
    }
  },
  {
    "ph": "f", "id": 221529, "pid": 76337, "tid": -914061504, "ts": 1716454224936688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936689, "dur": 0,
    "args": {
      "External id": 221530,
      "cbid": 205, "correlation": 221530
    }
  },
  {
    "ph": "f", "id": 221530, "pid": 76337, "tid": -914061504, "ts": 1716454224936689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224986048, "dur": 22,
    "args": {
      "External id": 221534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221534,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221534, "pid": 5, "tid": 7, "ts": 1716454224986048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936703, "dur": 12,
    "args": {
      "External id": 221534,
      "cbid": 211, "correlation": 221534
    }
  },
  {
    "ph": "s", "id": 221534, "pid": 76337, "tid": -914061504, "ts": 1716454224936703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224986071, "dur": 114,
    "args": {
      "External id": 221536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221536,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221536, "pid": 5, "tid": 7, "ts": 1716454224986071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936721, "dur": 6,
    "args": {
      "External id": 221536,
      "cbid": 211, "correlation": 221536
    }
  },
  {
    "ph": "s", "id": 221536, "pid": 76337, "tid": -914061504, "ts": 1716454224936721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224986186, "dur": 21,
    "args": {
      "External id": 221538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221538,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221538, "pid": 5, "tid": 7, "ts": 1716454224986186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936730, "dur": 5,
    "args": {
      "External id": 221538,
      "cbid": 211, "correlation": 221538
    }
  },
  {
    "ph": "s", "id": 221538, "pid": 76337, "tid": -914061504, "ts": 1716454224936730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224986208, "dur": 32,
    "args": {
      "External id": 221544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221544,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221544, "pid": 5, "tid": 7, "ts": 1716454224986208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936758, "dur": 8,
    "args": {
      "External id": 221544,
      "cbid": 211, "correlation": 221544
    }
  },
  {
    "ph": "s", "id": 221544, "pid": 76337, "tid": -914061504, "ts": 1716454224936758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224986241, "dur": 187,
    "args": {
      "External id": 221553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221553,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221553, "pid": 5, "tid": 7, "ts": 1716454224986241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936839, "dur": 14,
    "args": {
      "External id": 221553,
      "cbid": 211, "correlation": 221553
    }
  },
  {
    "ph": "s", "id": 221553, "pid": 76337, "tid": -914061504, "ts": 1716454224936839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224986429, "dur": 64,
    "args": {
      "External id": 221575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221575,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221575, "pid": 5, "tid": 7, "ts": 1716454224986429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936896, "dur": 10,
    "args": {
      "External id": 221575,
      "cbid": 211, "correlation": 221575
    }
  },
  {
    "ph": "s", "id": 221575, "pid": 76337, "tid": -914061504, "ts": 1716454224936896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936989, "dur": 1,
    "args": {
      "External id": 221586,
      "cbid": 251, "correlation": 221586
    }
  },
  {
    "ph": "f", "id": 221586, "pid": 76337, "tid": -914061504, "ts": 1716454224936989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224986495, "dur": 151,
    "args": {
      "External id": 221587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221587,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221587, "pid": 5, "tid": 7, "ts": 1716454224986495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224936995, "dur": 13,
    "args": {
      "External id": 221587,
      "cbid": 211, "correlation": 221587
    }
  },
  {
    "ph": "s", "id": 221587, "pid": 76337, "tid": -914061504, "ts": 1716454224936995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937066, "dur": 1,
    "args": {
      "External id": 221598,
      "cbid": 251, "correlation": 221598
    }
  },
  {
    "ph": "f", "id": 221598, "pid": 76337, "tid": -914061504, "ts": 1716454224937066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224986647, "dur": 147,
    "args": {
      "External id": 221599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221599,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221599, "pid": 5, "tid": 7, "ts": 1716454224986647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937070, "dur": 11,
    "args": {
      "External id": 221599,
      "cbid": 211, "correlation": 221599
    }
  },
  {
    "ph": "s", "id": 221599, "pid": 76337, "tid": -914061504, "ts": 1716454224937070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937134, "dur": 1,
    "args": {
      "External id": 221610,
      "cbid": 251, "correlation": 221610
    }
  },
  {
    "ph": "f", "id": 221610, "pid": 76337, "tid": -914061504, "ts": 1716454224937134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224986796, "dur": 143,
    "args": {
      "External id": 221611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221611,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221611, "pid": 5, "tid": 7, "ts": 1716454224986796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937138, "dur": 13,
    "args": {
      "External id": 221611,
      "cbid": 211, "correlation": 221611
    }
  },
  {
    "ph": "s", "id": 221611, "pid": 76337, "tid": -914061504, "ts": 1716454224937138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224986940, "dur": 1909,
    "args": {
      "External id": 221632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221632,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 221632, "pid": 5, "tid": 7, "ts": 1716454224986940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937219, "dur": 12,
    "args": {
      "External id": 221632,
      "cbid": 211, "correlation": 221632
    }
  },
  {
    "ph": "s", "id": 221632, "pid": 76337, "tid": -914061504, "ts": 1716454224937219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937318, "dur": 1,
    "args": {
      "External id": 221650,
      "cbid": 251, "correlation": 221650
    }
  },
  {
    "ph": "f", "id": 221650, "pid": 76337, "tid": -914061504, "ts": 1716454224937318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224988851, "dur": 146,
    "args": {
      "External id": 221652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221652,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 221652, "pid": 5, "tid": 7, "ts": 1716454224988851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937323, "dur": 13,
    "args": {
      "External id": 221652,
      "cbid": 211, "correlation": 221652
    }
  },
  {
    "ph": "s", "id": 221652, "pid": 76337, "tid": -914061504, "ts": 1716454224937323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224988998, "dur": 35,
    "args": {
      "External id": 221660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221660,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221660, "pid": 5, "tid": 7, "ts": 1716454224988998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937394, "dur": 13,
    "args": {
      "External id": 221660,
      "cbid": 211, "correlation": 221660
    }
  },
  {
    "ph": "s", "id": 221660, "pid": 76337, "tid": -914061504, "ts": 1716454224937394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224989034, "dur": 50,
    "args": {
      "External id": 221668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221668,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221668, "pid": 5, "tid": 7, "ts": 1716454224989034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937434, "dur": 8,
    "args": {
      "External id": 221668,
      "cbid": 211, "correlation": 221668
    }
  },
  {
    "ph": "s", "id": 221668, "pid": 76337, "tid": -914061504, "ts": 1716454224937434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224989085, "dur": 30,
    "args": {
      "External id": 221679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221679,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221679, "pid": 5, "tid": 7, "ts": 1716454224989085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937506, "dur": 12,
    "args": {
      "External id": 221679,
      "cbid": 211, "correlation": 221679
    }
  },
  {
    "ph": "s", "id": 221679, "pid": 76337, "tid": -914061504, "ts": 1716454224937506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224989117, "dur": 34,
    "args": {
      "External id": 221701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221701,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221701, "pid": 5, "tid": 7, "ts": 1716454224989117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937536, "dur": 7,
    "args": {
      "External id": 221701,
      "cbid": 211, "correlation": 221701
    }
  },
  {
    "ph": "s", "id": 221701, "pid": 76337, "tid": -914061504, "ts": 1716454224937536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937621, "dur": 1,
    "args": {
      "External id": 221712,
      "cbid": 251, "correlation": 221712
    }
  },
  {
    "ph": "f", "id": 221712, "pid": 76337, "tid": -914061504, "ts": 1716454224937621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224989152, "dur": 87,
    "args": {
      "External id": 221713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221713,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221713, "pid": 5, "tid": 7, "ts": 1716454224989152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937626, "dur": 12,
    "args": {
      "External id": 221713,
      "cbid": 211, "correlation": 221713
    }
  },
  {
    "ph": "s", "id": 221713, "pid": 76337, "tid": -914061504, "ts": 1716454224937626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937694, "dur": 1,
    "args": {
      "External id": 221724,
      "cbid": 251, "correlation": 221724
    }
  },
  {
    "ph": "f", "id": 221724, "pid": 76337, "tid": -914061504, "ts": 1716454224937694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937698, "dur": 0,
    "args": {
      "External id": 221725,
      "cbid": 251, "correlation": 221725
    }
  },
  {
    "ph": "f", "id": 221725, "pid": 76337, "tid": -914061504, "ts": 1716454224937698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224989241, "dur": 11,
    "args": {
      "External id": 221726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221726,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 221726, "pid": 5, "tid": 7, "ts": 1716454224989241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937699, "dur": 13,
    "args": {
      "External id": 221726,
      "cbid": 211, "correlation": 221726
    }
  },
  {
    "ph": "s", "id": 221726, "pid": 76337, "tid": -914061504, "ts": 1716454224937699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224989253, "dur": 5,
    "args": {
      "External id": 221728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221728,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 221728, "pid": 5, "tid": 7, "ts": 1716454224989253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937714, "dur": 5,
    "args": {
      "External id": 221728,
      "cbid": 211, "correlation": 221728
    }
  },
  {
    "ph": "s", "id": 221728, "pid": 76337, "tid": -914061504, "ts": 1716454224937714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937771, "dur": 1,
    "args": {
      "External id": 221739,
      "cbid": 251, "correlation": 221739
    }
  },
  {
    "ph": "f", "id": 221739, "pid": 76337, "tid": -914061504, "ts": 1716454224937771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937774, "dur": 0,
    "args": {
      "External id": 221740,
      "cbid": 251, "correlation": 221740
    }
  },
  {
    "ph": "f", "id": 221740, "pid": 76337, "tid": -914061504, "ts": 1716454224937774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224989259, "dur": 7,
    "args": {
      "External id": 221741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221741,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 221741, "pid": 5, "tid": 7, "ts": 1716454224989259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937776, "dur": 11,
    "args": {
      "External id": 221741,
      "cbid": 211, "correlation": 221741
    }
  },
  {
    "ph": "s", "id": 221741, "pid": 76337, "tid": -914061504, "ts": 1716454224937776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224989267, "dur": 3,
    "args": {
      "External id": 221743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221743,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 221743, "pid": 5, "tid": 7, "ts": 1716454224989267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937788, "dur": 5,
    "args": {
      "External id": 221743,
      "cbid": 211, "correlation": 221743
    }
  },
  {
    "ph": "s", "id": 221743, "pid": 76337, "tid": -914061504, "ts": 1716454224937788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224989271, "dur": 90,
    "args": {
      "External id": 221764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221764,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 221764, "pid": 5, "tid": 7, "ts": 1716454224989271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937862, "dur": 12,
    "args": {
      "External id": 221764,
      "cbid": 211, "correlation": 221764
    }
  },
  {
    "ph": "s", "id": 221764, "pid": 76337, "tid": -914061504, "ts": 1716454224937862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937958, "dur": 1,
    "args": {
      "External id": 221782,
      "cbid": 251, "correlation": 221782
    }
  },
  {
    "ph": "f", "id": 221782, "pid": 76337, "tid": -914061504, "ts": 1716454224937958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224989363, "dur": 96,
    "args": {
      "External id": 221784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221784,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221784, "pid": 5, "tid": 7, "ts": 1716454224989363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224937963, "dur": 22,
    "args": {
      "External id": 221784,
      "cbid": 211, "correlation": 221784
    }
  },
  {
    "ph": "s", "id": 221784, "pid": 76337, "tid": -914061504, "ts": 1716454224937963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224989460, "dur": 19,
    "args": {
      "External id": 221792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221792,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221792, "pid": 5, "tid": 7, "ts": 1716454224989460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938042, "dur": 13,
    "args": {
      "External id": 221792,
      "cbid": 211, "correlation": 221792
    }
  },
  {
    "ph": "s", "id": 221792, "pid": 76337, "tid": -914061504, "ts": 1716454224938042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224989480, "dur": 37,
    "args": {
      "External id": 221800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221800, "pid": 5, "tid": 7, "ts": 1716454224989480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938085, "dur": 10,
    "args": {
      "External id": 221800,
      "cbid": 211, "correlation": 221800
    }
  },
  {
    "ph": "s", "id": 221800, "pid": 76337, "tid": -914061504, "ts": 1716454224938085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224989518, "dur": 34,
    "args": {
      "External id": 221822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221822,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221822, "pid": 5, "tid": 7, "ts": 1716454224989518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938136, "dur": 11,
    "args": {
      "External id": 221822,
      "cbid": 211, "correlation": 221822
    }
  },
  {
    "ph": "s", "id": 221822, "pid": 76337, "tid": -914061504, "ts": 1716454224938136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938224, "dur": 1,
    "args": {
      "External id": 221838,
      "cbid": 251, "correlation": 221838
    }
  },
  {
    "ph": "f", "id": 221838, "pid": 76337, "tid": -914061504, "ts": 1716454224938224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938229, "dur": 0,
    "args": {
      "External id": 221840,
      "cbid": 251, "correlation": 221840
    }
  },
  {
    "ph": "f", "id": 221840, "pid": 76337, "tid": -914061504, "ts": 1716454224938229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224989553, "dur": 531,
    "args": {
      "External id": 221841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221841,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 221841, "pid": 5, "tid": 7, "ts": 1716454224989553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938233, "dur": 13,
    "args": {
      "External id": 221841,
      "cbid": 211, "correlation": 221841
    }
  },
  {
    "ph": "s", "id": 221841, "pid": 76337, "tid": -914061504, "ts": 1716454224938233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224990086, "dur": 124,
    "args": {
      "External id": 221849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221849, "pid": 5, "tid": 7, "ts": 1716454224990086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938297, "dur": 13,
    "args": {
      "External id": 221849,
      "cbid": 211, "correlation": 221849
    }
  },
  {
    "ph": "s", "id": 221849, "pid": 76337, "tid": -914061504, "ts": 1716454224938297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224990211, "dur": 128,
    "args": {
      "External id": 221857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221857, "pid": 5, "tid": 7, "ts": 1716454224990211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938327, "dur": 8,
    "args": {
      "External id": 221857,
      "cbid": 211, "correlation": 221857
    }
  },
  {
    "ph": "s", "id": 221857, "pid": 76337, "tid": -914061504, "ts": 1716454224938327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938402, "dur": 1,
    "args": {
      "External id": 221873,
      "cbid": 251, "correlation": 221873
    }
  },
  {
    "ph": "f", "id": 221873, "pid": 76337, "tid": -914061504, "ts": 1716454224938402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224990340, "dur": 304,
    "args": {
      "External id": 221875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221875,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221875, "pid": 5, "tid": 7, "ts": 1716454224990340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938408, "dur": 12,
    "args": {
      "External id": 221875,
      "cbid": 211, "correlation": 221875
    }
  },
  {
    "ph": "s", "id": 221875, "pid": 76337, "tid": -914061504, "ts": 1716454224938408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224990645, "dur": 27,
    "args": {
      "External id": 221883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221883,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221883, "pid": 5, "tid": 7, "ts": 1716454224990645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938450, "dur": 9,
    "args": {
      "External id": 221883,
      "cbid": 211, "correlation": 221883
    }
  },
  {
    "ph": "s", "id": 221883, "pid": 76337, "tid": -914061504, "ts": 1716454224938450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224990674, "dur": 80,
    "args": {
      "External id": 221894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221894,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221894, "pid": 5, "tid": 7, "ts": 1716454224990674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938517, "dur": 12,
    "args": {
      "External id": 221894,
      "cbid": 211, "correlation": 221894
    }
  },
  {
    "ph": "s", "id": 221894, "pid": 76337, "tid": -914061504, "ts": 1716454224938517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938580, "dur": 0,
    "args": {
      "External id": 221906,
      "cbid": 317, "correlation": 221906
    }
  },
  {
    "ph": "f", "id": 221906, "pid": 76337, "tid": -914061504, "ts": 1716454224938580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938581, "dur": 0,
    "args": {
      "External id": 221907,
      "cbid": 203, "correlation": 221907
    }
  },
  {
    "ph": "f", "id": 221907, "pid": 76337, "tid": -914061504, "ts": 1716454224938581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938582, "dur": 0,
    "args": {
      "External id": 221908,
      "cbid": 205, "correlation": 221908
    }
  },
  {
    "ph": "f", "id": 221908, "pid": 76337, "tid": -914061504, "ts": 1716454224938582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224990756, "dur": 22,
    "args": {
      "External id": 221912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221912, "pid": 5, "tid": 7, "ts": 1716454224990756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938598, "dur": 12,
    "args": {
      "External id": 221912,
      "cbid": 211, "correlation": 221912
    }
  },
  {
    "ph": "s", "id": 221912, "pid": 76337, "tid": -914061504, "ts": 1716454224938598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224990779, "dur": 118,
    "args": {
      "External id": 221914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221914,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 221914, "pid": 5, "tid": 7, "ts": 1716454224990779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938616, "dur": 7,
    "args": {
      "External id": 221914,
      "cbid": 211, "correlation": 221914
    }
  },
  {
    "ph": "s", "id": 221914, "pid": 76337, "tid": -914061504, "ts": 1716454224938616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224990898, "dur": 23,
    "args": {
      "External id": 221916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221916,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221916, "pid": 5, "tid": 7, "ts": 1716454224990898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938626, "dur": 5,
    "args": {
      "External id": 221916,
      "cbid": 211, "correlation": 221916
    }
  },
  {
    "ph": "s", "id": 221916, "pid": 76337, "tid": -914061504, "ts": 1716454224938626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224990923, "dur": 33,
    "args": {
      "External id": 221922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221922, "pid": 5, "tid": 7, "ts": 1716454224990923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938654, "dur": 8,
    "args": {
      "External id": 221922,
      "cbid": 211, "correlation": 221922
    }
  },
  {
    "ph": "s", "id": 221922, "pid": 76337, "tid": -914061504, "ts": 1716454224938654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224990957, "dur": 27,
    "args": {
      "External id": 221930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221930,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221930, "pid": 5, "tid": 7, "ts": 1716454224990957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938685, "dur": 8,
    "args": {
      "External id": 221930,
      "cbid": 211, "correlation": 221930
    }
  },
  {
    "ph": "s", "id": 221930, "pid": 76337, "tid": -914061504, "ts": 1716454224938685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938756, "dur": 0,
    "args": {
      "External id": 221940,
      "cbid": 317, "correlation": 221940
    }
  },
  {
    "ph": "f", "id": 221940, "pid": 76337, "tid": -914061504, "ts": 1716454224938756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938757, "dur": 0,
    "args": {
      "External id": 221941,
      "cbid": 203, "correlation": 221941
    }
  },
  {
    "ph": "f", "id": 221941, "pid": 76337, "tid": -914061504, "ts": 1716454224938757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938758, "dur": 0,
    "args": {
      "External id": 221942,
      "cbid": 205, "correlation": 221942
    }
  },
  {
    "ph": "f", "id": 221942, "pid": 76337, "tid": -914061504, "ts": 1716454224938758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224990985, "dur": 24,
    "args": {
      "External id": 221946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221946, "pid": 5, "tid": 7, "ts": 1716454224990985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938771, "dur": 12,
    "args": {
      "External id": 221946,
      "cbid": 211, "correlation": 221946
    }
  },
  {
    "ph": "s", "id": 221946, "pid": 76337, "tid": -914061504, "ts": 1716454224938771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224991010, "dur": 43,
    "args": {
      "External id": 221948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221948, "pid": 5, "tid": 7, "ts": 1716454224991010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938786, "dur": 5,
    "args": {
      "External id": 221948,
      "cbid": 211, "correlation": 221948
    }
  },
  {
    "ph": "s", "id": 221948, "pid": 76337, "tid": -914061504, "ts": 1716454224938786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224991055, "dur": 233,
    "args": {
      "External id": 221950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221950,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 221950, "pid": 5, "tid": 7, "ts": 1716454224991055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938798, "dur": 7,
    "args": {
      "External id": 221950,
      "cbid": 211, "correlation": 221950
    }
  },
  {
    "ph": "s", "id": 221950, "pid": 76337, "tid": -914061504, "ts": 1716454224938798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224991290, "dur": 6,
    "args": {
      "External id": 221952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221952, "pid": 5, "tid": 7, "ts": 1716454224991290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938808, "dur": 5,
    "args": {
      "External id": 221952,
      "cbid": 211, "correlation": 221952
    }
  },
  {
    "ph": "s", "id": 221952, "pid": 76337, "tid": -914061504, "ts": 1716454224938808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224991297, "dur": 9,
    "args": {
      "External id": 221958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221958, "pid": 5, "tid": 7, "ts": 1716454224991297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938834, "dur": 10,
    "args": {
      "External id": 221958,
      "cbid": 211, "correlation": 221958
    }
  },
  {
    "ph": "s", "id": 221958, "pid": 76337, "tid": -914061504, "ts": 1716454224938834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224991307, "dur": 12,
    "args": {
      "External id": 221978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221978,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 221978, "pid": 5, "tid": 7, "ts": 1716454224991307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938926, "dur": 12,
    "args": {
      "External id": 221978,
      "cbid": 211, "correlation": 221978
    }
  },
  {
    "ph": "s", "id": 221978, "pid": 76337, "tid": -914061504, "ts": 1716454224938926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224991320, "dur": 4,
    "args": {
      "External id": 221990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221990,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 221990, "pid": 5, "tid": 7, "ts": 1716454224991320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938948, "dur": 6,
    "args": {
      "External id": 221990,
      "cbid": 211, "correlation": 221990
    }
  },
  {
    "ph": "s", "id": 221990, "pid": 76337, "tid": -914061504, "ts": 1716454224938948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224991325, "dur": 12,
    "args": {
      "External id": 221993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 221993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 221993, "pid": 5, "tid": 7, "ts": 1716454224991325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224938967, "dur": 15,
    "args": {
      "External id": 221993,
      "cbid": 211, "correlation": 221993
    }
  },
  {
    "ph": "s", "id": 221993, "pid": 76337, "tid": -914061504, "ts": 1716454224938967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224991339, "dur": 7,
    "args": {
      "External id": 222002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222002,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222002, "pid": 5, "tid": 7, "ts": 1716454224991339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939015, "dur": 10,
    "args": {
      "External id": 222002,
      "cbid": 211, "correlation": 222002
    }
  },
  {
    "ph": "s", "id": 222002, "pid": 76337, "tid": -914061504, "ts": 1716454224939015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939069, "dur": 0,
    "args": {
      "External id": 222012,
      "cbid": 317, "correlation": 222012
    }
  },
  {
    "ph": "f", "id": 222012, "pid": 76337, "tid": -914061504, "ts": 1716454224939069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939069, "dur": 0,
    "args": {
      "External id": 222013,
      "cbid": 203, "correlation": 222013
    }
  },
  {
    "ph": "f", "id": 222013, "pid": 76337, "tid": -914061504, "ts": 1716454224939069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939070, "dur": 0,
    "args": {
      "External id": 222014,
      "cbid": 205, "correlation": 222014
    }
  },
  {
    "ph": "f", "id": 222014, "pid": 76337, "tid": -914061504, "ts": 1716454224939070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224991347, "dur": 6,
    "args": {
      "External id": 222018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222018,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222018, "pid": 5, "tid": 7, "ts": 1716454224991347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939085, "dur": 11,
    "args": {
      "External id": 222018,
      "cbid": 211, "correlation": 222018
    }
  },
  {
    "ph": "s", "id": 222018, "pid": 76337, "tid": -914061504, "ts": 1716454224939085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224991354, "dur": 81,
    "args": {
      "External id": 222020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222020, "pid": 5, "tid": 7, "ts": 1716454224991354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939100, "dur": 5,
    "args": {
      "External id": 222020,
      "cbid": 211, "correlation": 222020
    }
  },
  {
    "ph": "s", "id": 222020, "pid": 76337, "tid": -914061504, "ts": 1716454224939100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224991437, "dur": 1,
    "args": {
      "External id": 222022,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 222022,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 222022, "pid": 5, "tid": 7, "ts": 1716454224991437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939113, "dur": 9,
    "args": {
      "External id": 222022,
      "cbid": 51, "correlation": 222022
    }
  },
  {
    "ph": "s", "id": 222022, "pid": 76337, "tid": -914061504, "ts": 1716454224939113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224991441, "dur": 537,
    "args": {
      "External id": 222023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222023,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222023, "pid": 5, "tid": 7, "ts": 1716454224991441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939123, "dur": 8,
    "args": {
      "External id": 222023,
      "cbid": 211, "correlation": 222023
    }
  },
  {
    "ph": "s", "id": 222023, "pid": 76337, "tid": -914061504, "ts": 1716454224939123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224991979, "dur": 12,
    "args": {
      "External id": 222025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222025,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222025, "pid": 5, "tid": 7, "ts": 1716454224991979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939135, "dur": 6,
    "args": {
      "External id": 222025,
      "cbid": 211, "correlation": 222025
    }
  },
  {
    "ph": "s", "id": 222025, "pid": 76337, "tid": -914061504, "ts": 1716454224939135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224991992, "dur": 14,
    "args": {
      "External id": 222031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222031, "pid": 5, "tid": 7, "ts": 1716454224991992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939164, "dur": 8,
    "args": {
      "External id": 222031,
      "cbid": 211, "correlation": 222031
    }
  },
  {
    "ph": "s", "id": 222031, "pid": 76337, "tid": -914061504, "ts": 1716454224939164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224992008, "dur": 4,
    "args": {
      "External id": 222039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222039,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 222039, "pid": 5, "tid": 7, "ts": 1716454224992008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939207, "dur": 9,
    "args": {
      "External id": 222039,
      "cbid": 211, "correlation": 222039
    }
  },
  {
    "ph": "s", "id": 222039, "pid": 76337, "tid": -914061504, "ts": 1716454224939207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939271, "dur": 1,
    "args": {
      "External id": 222055,
      "cbid": 251, "correlation": 222055
    }
  },
  {
    "ph": "f", "id": 222055, "pid": 76337, "tid": -914061504, "ts": 1716454224939271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939276, "dur": 0,
    "args": {
      "External id": 222057,
      "cbid": 251, "correlation": 222057
    }
  },
  {
    "ph": "f", "id": 222057, "pid": 76337, "tid": -914061504, "ts": 1716454224939276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224992013, "dur": 14,
    "args": {
      "External id": 222058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222058,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222058, "pid": 5, "tid": 7, "ts": 1716454224992013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939278, "dur": 11,
    "args": {
      "External id": 222058,
      "cbid": 211, "correlation": 222058
    }
  },
  {
    "ph": "s", "id": 222058, "pid": 76337, "tid": -914061504, "ts": 1716454224939278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224992028, "dur": 5,
    "args": {
      "External id": 222060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222060,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222060, "pid": 5, "tid": 7, "ts": 1716454224992028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939291, "dur": 6,
    "args": {
      "External id": 222060,
      "cbid": 211, "correlation": 222060
    }
  },
  {
    "ph": "s", "id": 222060, "pid": 76337, "tid": -914061504, "ts": 1716454224939291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224992034, "dur": 17,
    "args": {
      "External id": 222070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222070,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222070, "pid": 5, "tid": 7, "ts": 1716454224992034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939349, "dur": 12,
    "args": {
      "External id": 222070,
      "cbid": 211, "correlation": 222070
    }
  },
  {
    "ph": "s", "id": 222070, "pid": 76337, "tid": -914061504, "ts": 1716454224939349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224992052, "dur": 17,
    "args": {
      "External id": 222090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222090,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 222090, "pid": 5, "tid": 7, "ts": 1716454224992052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939414, "dur": 10,
    "args": {
      "External id": 222090,
      "cbid": 211, "correlation": 222090
    }
  },
  {
    "ph": "s", "id": 222090, "pid": 76337, "tid": -914061504, "ts": 1716454224939414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224992071, "dur": 4,
    "args": {
      "External id": 222102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222102,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 222102, "pid": 5, "tid": 7, "ts": 1716454224992071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939441, "dur": 7,
    "args": {
      "External id": 222102,
      "cbid": 211, "correlation": 222102
    }
  },
  {
    "ph": "s", "id": 222102, "pid": 76337, "tid": -914061504, "ts": 1716454224939441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224992077, "dur": 16,
    "args": {
      "External id": 222105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222105,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222105, "pid": 5, "tid": 7, "ts": 1716454224992077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939461, "dur": 7,
    "args": {
      "External id": 222105,
      "cbid": 211, "correlation": 222105
    }
  },
  {
    "ph": "s", "id": 222105, "pid": 76337, "tid": -914061504, "ts": 1716454224939461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224992094, "dur": 11,
    "args": {
      "External id": 222114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222114,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222114, "pid": 5, "tid": 7, "ts": 1716454224992094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939501, "dur": 9,
    "args": {
      "External id": 222114,
      "cbid": 211, "correlation": 222114
    }
  },
  {
    "ph": "s", "id": 222114, "pid": 76337, "tid": -914061504, "ts": 1716454224939501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939562, "dur": 0,
    "args": {
      "External id": 222124,
      "cbid": 317, "correlation": 222124
    }
  },
  {
    "ph": "f", "id": 222124, "pid": 76337, "tid": -914061504, "ts": 1716454224939562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939563, "dur": 0,
    "args": {
      "External id": 222125,
      "cbid": 203, "correlation": 222125
    }
  },
  {
    "ph": "f", "id": 222125, "pid": 76337, "tid": -914061504, "ts": 1716454224939563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939564, "dur": 0,
    "args": {
      "External id": 222126,
      "cbid": 205, "correlation": 222126
    }
  },
  {
    "ph": "f", "id": 222126, "pid": 76337, "tid": -914061504, "ts": 1716454224939564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224992106, "dur": 11,
    "args": {
      "External id": 222130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222130, "pid": 5, "tid": 7, "ts": 1716454224992106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939580, "dur": 12,
    "args": {
      "External id": 222130,
      "cbid": 211, "correlation": 222130
    }
  },
  {
    "ph": "s", "id": 222130, "pid": 76337, "tid": -914061504, "ts": 1716454224939580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224992119, "dur": 159,
    "args": {
      "External id": 222132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222132, "pid": 5, "tid": 7, "ts": 1716454224992119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939595, "dur": 6,
    "args": {
      "External id": 222132,
      "cbid": 211, "correlation": 222132
    }
  },
  {
    "ph": "s", "id": 222132, "pid": 76337, "tid": -914061504, "ts": 1716454224939595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224992280, "dur": 1,
    "args": {
      "External id": 222134,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 222134,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 222134, "pid": 5, "tid": 7, "ts": 1716454224992280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939607, "dur": 6,
    "args": {
      "External id": 222134,
      "cbid": 51, "correlation": 222134
    }
  },
  {
    "ph": "s", "id": 222134, "pid": 76337, "tid": -914061504, "ts": 1716454224939607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224992284, "dur": 656,
    "args": {
      "External id": 222135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222135,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222135, "pid": 5, "tid": 7, "ts": 1716454224992284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939615, "dur": 6,
    "args": {
      "External id": 222135,
      "cbid": 211, "correlation": 222135
    }
  },
  {
    "ph": "s", "id": 222135, "pid": 76337, "tid": -914061504, "ts": 1716454224939615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224992941, "dur": 13,
    "args": {
      "External id": 222137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222137,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222137, "pid": 5, "tid": 7, "ts": 1716454224992941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939625, "dur": 5,
    "args": {
      "External id": 222137,
      "cbid": 211, "correlation": 222137
    }
  },
  {
    "ph": "s", "id": 222137, "pid": 76337, "tid": -914061504, "ts": 1716454224939625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224992955, "dur": 14,
    "args": {
      "External id": 222143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222143,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222143, "pid": 5, "tid": 7, "ts": 1716454224992955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939654, "dur": 8,
    "args": {
      "External id": 222143,
      "cbid": 211, "correlation": 222143
    }
  },
  {
    "ph": "s", "id": 222143, "pid": 76337, "tid": -914061504, "ts": 1716454224939654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939711, "dur": 0,
    "args": {
      "External id": 222153,
      "cbid": 317, "correlation": 222153
    }
  },
  {
    "ph": "f", "id": 222153, "pid": 76337, "tid": -914061504, "ts": 1716454224939711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939712, "dur": 0,
    "args": {
      "External id": 222154,
      "cbid": 203, "correlation": 222154
    }
  },
  {
    "ph": "f", "id": 222154, "pid": 76337, "tid": -914061504, "ts": 1716454224939712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939713, "dur": 0,
    "args": {
      "External id": 222155,
      "cbid": 205, "correlation": 222155
    }
  },
  {
    "ph": "f", "id": 222155, "pid": 76337, "tid": -914061504, "ts": 1716454224939713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224992971, "dur": 8,
    "args": {
      "External id": 222159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222159,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222159, "pid": 5, "tid": 7, "ts": 1716454224992971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939727, "dur": 11,
    "args": {
      "External id": 222159,
      "cbid": 211, "correlation": 222159
    }
  },
  {
    "ph": "s", "id": 222159, "pid": 76337, "tid": -914061504, "ts": 1716454224939727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224992980, "dur": 3,
    "args": {
      "External id": 222161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222161, "pid": 5, "tid": 7, "ts": 1716454224992980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939744, "dur": 7,
    "args": {
      "External id": 222161,
      "cbid": 211, "correlation": 222161
    }
  },
  {
    "ph": "s", "id": 222161, "pid": 76337, "tid": -914061504, "ts": 1716454224939744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939755, "dur": 0,
    "args": {
      "External id": 222162,
      "cbid": 51, "correlation": 222162
    }
  },
  {
    "ph": "s", "id": 222162, "pid": 76337, "tid": -914061504, "ts": 1716454224939755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224992985, "dur": 56,
    "args": {
      "External id": 222163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222163,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 222163, "pid": 5, "tid": 7, "ts": 1716454224992985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939756, "dur": 5,
    "args": {
      "External id": 222163,
      "cbid": 211, "correlation": 222163
    }
  },
  {
    "ph": "s", "id": 222163, "pid": 76337, "tid": -914061504, "ts": 1716454224939756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224993042, "dur": 14,
    "args": {
      "External id": 222168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222168, "pid": 5, "tid": 7, "ts": 1716454224993042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939781, "dur": 8,
    "args": {
      "External id": 222168,
      "cbid": 211, "correlation": 222168
    }
  },
  {
    "ph": "s", "id": 222168, "pid": 76337, "tid": -914061504, "ts": 1716454224939781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224993057, "dur": 12,
    "args": {
      "External id": 222176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222176,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222176, "pid": 5, "tid": 7, "ts": 1716454224993057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939810, "dur": 8,
    "args": {
      "External id": 222176,
      "cbid": 211, "correlation": 222176
    }
  },
  {
    "ph": "s", "id": 222176, "pid": 76337, "tid": -914061504, "ts": 1716454224939810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224993070, "dur": 10,
    "args": {
      "External id": 222184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222184,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222184, "pid": 5, "tid": 7, "ts": 1716454224993070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939838, "dur": 9,
    "args": {
      "External id": 222184,
      "cbid": 211, "correlation": 222184
    }
  },
  {
    "ph": "s", "id": 222184, "pid": 76337, "tid": -914061504, "ts": 1716454224939838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224993082, "dur": 18,
    "args": {
      "External id": 222204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222204,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 222204, "pid": 5, "tid": 7, "ts": 1716454224993082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939919, "dur": 12,
    "args": {
      "External id": 222204,
      "cbid": 211, "correlation": 222204
    }
  },
  {
    "ph": "s", "id": 222204, "pid": 76337, "tid": -914061504, "ts": 1716454224939919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224993101, "dur": 4,
    "args": {
      "External id": 222216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222216,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 222216, "pid": 5, "tid": 7, "ts": 1716454224993101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939940, "dur": 6,
    "args": {
      "External id": 222216,
      "cbid": 211, "correlation": 222216
    }
  },
  {
    "ph": "s", "id": 222216, "pid": 76337, "tid": -914061504, "ts": 1716454224939940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224993106, "dur": 17,
    "args": {
      "External id": 222219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222219, "pid": 5, "tid": 7, "ts": 1716454224993106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224939958, "dur": 7,
    "args": {
      "External id": 222219,
      "cbid": 211, "correlation": 222219
    }
  },
  {
    "ph": "s", "id": 222219, "pid": 76337, "tid": -914061504, "ts": 1716454224939958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940024, "dur": 0,
    "args": {
      "External id": 222230,
      "cbid": 317, "correlation": 222230
    }
  },
  {
    "ph": "f", "id": 222230, "pid": 76337, "tid": -914061504, "ts": 1716454224940024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940024, "dur": 0,
    "args": {
      "External id": 222231,
      "cbid": 203, "correlation": 222231
    }
  },
  {
    "ph": "f", "id": 222231, "pid": 76337, "tid": -914061504, "ts": 1716454224940024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940025, "dur": 0,
    "args": {
      "External id": 222232,
      "cbid": 205, "correlation": 222232
    }
  },
  {
    "ph": "f", "id": 222232, "pid": 76337, "tid": -914061504, "ts": 1716454224940025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224993124, "dur": 12,
    "args": {
      "External id": 222236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222236, "pid": 5, "tid": 7, "ts": 1716454224993124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940040, "dur": 14,
    "args": {
      "External id": 222236,
      "cbid": 211, "correlation": 222236
    }
  },
  {
    "ph": "s", "id": 222236, "pid": 76337, "tid": -914061504, "ts": 1716454224940040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224993138, "dur": 3,
    "args": {
      "External id": 222238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222238, "pid": 5, "tid": 7, "ts": 1716454224993138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940059, "dur": 6,
    "args": {
      "External id": 222238,
      "cbid": 211, "correlation": 222238
    }
  },
  {
    "ph": "s", "id": 222238, "pid": 76337, "tid": -914061504, "ts": 1716454224940059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940067, "dur": 0,
    "args": {
      "External id": 222239,
      "cbid": 51, "correlation": 222239
    }
  },
  {
    "ph": "s", "id": 222239, "pid": 76337, "tid": -914061504, "ts": 1716454224940067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224993142, "dur": 95,
    "args": {
      "External id": 222240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222240,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 222240, "pid": 5, "tid": 7, "ts": 1716454224993142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940068, "dur": 5,
    "args": {
      "External id": 222240,
      "cbid": 211, "correlation": 222240
    }
  },
  {
    "ph": "s", "id": 222240, "pid": 76337, "tid": -914061504, "ts": 1716454224940068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224993238, "dur": 15,
    "args": {
      "External id": 222245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222245,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222245, "pid": 5, "tid": 7, "ts": 1716454224993238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940096, "dur": 9,
    "args": {
      "External id": 222245,
      "cbid": 211, "correlation": 222245
    }
  },
  {
    "ph": "s", "id": 222245, "pid": 76337, "tid": -914061504, "ts": 1716454224940096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224993255, "dur": 81,
    "args": {
      "External id": 222254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222254,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222254, "pid": 5, "tid": 7, "ts": 1716454224993255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940178, "dur": 14,
    "args": {
      "External id": 222254,
      "cbid": 211, "correlation": 222254
    }
  },
  {
    "ph": "s", "id": 222254, "pid": 76337, "tid": -914061504, "ts": 1716454224940178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224993337, "dur": 30,
    "args": {
      "External id": 222276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222276,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222276, "pid": 5, "tid": 7, "ts": 1716454224993337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940234, "dur": 10,
    "args": {
      "External id": 222276,
      "cbid": 211, "correlation": 222276
    }
  },
  {
    "ph": "s", "id": 222276, "pid": 76337, "tid": -914061504, "ts": 1716454224940234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940325, "dur": 1,
    "args": {
      "External id": 222287,
      "cbid": 251, "correlation": 222287
    }
  },
  {
    "ph": "f", "id": 222287, "pid": 76337, "tid": -914061504, "ts": 1716454224940325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224993368, "dur": 161,
    "args": {
      "External id": 222288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222288,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222288, "pid": 5, "tid": 7, "ts": 1716454224993368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940330, "dur": 13,
    "args": {
      "External id": 222288,
      "cbid": 211, "correlation": 222288
    }
  },
  {
    "ph": "s", "id": 222288, "pid": 76337, "tid": -914061504, "ts": 1716454224940330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940401, "dur": 1,
    "args": {
      "External id": 222299,
      "cbid": 251, "correlation": 222299
    }
  },
  {
    "ph": "f", "id": 222299, "pid": 76337, "tid": -914061504, "ts": 1716454224940401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224993531, "dur": 155,
    "args": {
      "External id": 222300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222300,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222300, "pid": 5, "tid": 7, "ts": 1716454224993531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940405, "dur": 11,
    "args": {
      "External id": 222300,
      "cbid": 211, "correlation": 222300
    }
  },
  {
    "ph": "s", "id": 222300, "pid": 76337, "tid": -914061504, "ts": 1716454224940405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940469, "dur": 1,
    "args": {
      "External id": 222311,
      "cbid": 251, "correlation": 222311
    }
  },
  {
    "ph": "f", "id": 222311, "pid": 76337, "tid": -914061504, "ts": 1716454224940469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224993687, "dur": 156,
    "args": {
      "External id": 222312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222312,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222312, "pid": 5, "tid": 7, "ts": 1716454224993687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940473, "dur": 11,
    "args": {
      "External id": 222312,
      "cbid": 211, "correlation": 222312
    }
  },
  {
    "ph": "s", "id": 222312, "pid": 76337, "tid": -914061504, "ts": 1716454224940473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224993844, "dur": 333,
    "args": {
      "External id": 222337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222337,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222337, "pid": 5, "tid": 7, "ts": 1716454224993844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940560, "dur": 13,
    "args": {
      "External id": 222337,
      "cbid": 211, "correlation": 222337
    }
  },
  {
    "ph": "s", "id": 222337, "pid": 76337, "tid": -914061504, "ts": 1716454224940560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940661, "dur": 1,
    "args": {
      "External id": 222355,
      "cbid": 251, "correlation": 222355
    }
  },
  {
    "ph": "f", "id": 222355, "pid": 76337, "tid": -914061504, "ts": 1716454224940661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224994177, "dur": 166,
    "args": {
      "External id": 222357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222357,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222357, "pid": 5, "tid": 7, "ts": 1716454224994177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940667, "dur": 13,
    "args": {
      "External id": 222357,
      "cbid": 211, "correlation": 222357
    }
  },
  {
    "ph": "s", "id": 222357, "pid": 76337, "tid": -914061504, "ts": 1716454224940667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224994345, "dur": 19,
    "args": {
      "External id": 222365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222365,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222365, "pid": 5, "tid": 7, "ts": 1716454224994345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940736, "dur": 12,
    "args": {
      "External id": 222365,
      "cbid": 211, "correlation": 222365
    }
  },
  {
    "ph": "s", "id": 222365, "pid": 76337, "tid": -914061504, "ts": 1716454224940736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224994365, "dur": 27,
    "args": {
      "External id": 222373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222373,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222373, "pid": 5, "tid": 7, "ts": 1716454224994365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940775, "dur": 8,
    "args": {
      "External id": 222373,
      "cbid": 211, "correlation": 222373
    }
  },
  {
    "ph": "s", "id": 222373, "pid": 76337, "tid": -914061504, "ts": 1716454224940775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224994394, "dur": 17,
    "args": {
      "External id": 222384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222384, "pid": 5, "tid": 7, "ts": 1716454224994394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940846, "dur": 12,
    "args": {
      "External id": 222384,
      "cbid": 211, "correlation": 222384
    }
  },
  {
    "ph": "s", "id": 222384, "pid": 76337, "tid": -914061504, "ts": 1716454224940846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224994413, "dur": 16,
    "args": {
      "External id": 222406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222406,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222406, "pid": 5, "tid": 7, "ts": 1716454224994413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940876, "dur": 7,
    "args": {
      "External id": 222406,
      "cbid": 211, "correlation": 222406
    }
  },
  {
    "ph": "s", "id": 222406, "pid": 76337, "tid": -914061504, "ts": 1716454224940876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940959, "dur": 1,
    "args": {
      "External id": 222417,
      "cbid": 251, "correlation": 222417
    }
  },
  {
    "ph": "f", "id": 222417, "pid": 76337, "tid": -914061504, "ts": 1716454224940959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224994430, "dur": 89,
    "args": {
      "External id": 222418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222418,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222418, "pid": 5, "tid": 7, "ts": 1716454224994430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224940966, "dur": 22,
    "args": {
      "External id": 222418,
      "cbid": 211, "correlation": 222418
    }
  },
  {
    "ph": "s", "id": 222418, "pid": 76337, "tid": -914061504, "ts": 1716454224940966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941047, "dur": 1,
    "args": {
      "External id": 222429,
      "cbid": 251, "correlation": 222429
    }
  },
  {
    "ph": "f", "id": 222429, "pid": 76337, "tid": -914061504, "ts": 1716454224941047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941051, "dur": 0,
    "args": {
      "External id": 222430,
      "cbid": 251, "correlation": 222430
    }
  },
  {
    "ph": "f", "id": 222430, "pid": 76337, "tid": -914061504, "ts": 1716454224941051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224994520, "dur": 12,
    "args": {
      "External id": 222431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222431,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222431, "pid": 5, "tid": 7, "ts": 1716454224994520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941052, "dur": 13,
    "args": {
      "External id": 222431,
      "cbid": 211, "correlation": 222431
    }
  },
  {
    "ph": "s", "id": 222431, "pid": 76337, "tid": -914061504, "ts": 1716454224941052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224994533, "dur": 5,
    "args": {
      "External id": 222433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222433,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222433, "pid": 5, "tid": 7, "ts": 1716454224994533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941067, "dur": 5,
    "args": {
      "External id": 222433,
      "cbid": 211, "correlation": 222433
    }
  },
  {
    "ph": "s", "id": 222433, "pid": 76337, "tid": -914061504, "ts": 1716454224941067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941124, "dur": 1,
    "args": {
      "External id": 222444,
      "cbid": 251, "correlation": 222444
    }
  },
  {
    "ph": "f", "id": 222444, "pid": 76337, "tid": -914061504, "ts": 1716454224941124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941128, "dur": 0,
    "args": {
      "External id": 222445,
      "cbid": 251, "correlation": 222445
    }
  },
  {
    "ph": "f", "id": 222445, "pid": 76337, "tid": -914061504, "ts": 1716454224941128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224994539, "dur": 8,
    "args": {
      "External id": 222446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222446,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222446, "pid": 5, "tid": 7, "ts": 1716454224994539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941129, "dur": 12,
    "args": {
      "External id": 222446,
      "cbid": 211, "correlation": 222446
    }
  },
  {
    "ph": "s", "id": 222446, "pid": 76337, "tid": -914061504, "ts": 1716454224941129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224994549, "dur": 3,
    "args": {
      "External id": 222448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222448,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222448, "pid": 5, "tid": 7, "ts": 1716454224994549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941143, "dur": 6,
    "args": {
      "External id": 222448,
      "cbid": 211, "correlation": 222448
    }
  },
  {
    "ph": "s", "id": 222448, "pid": 76337, "tid": -914061504, "ts": 1716454224941143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224994553, "dur": 53,
    "args": {
      "External id": 222473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222473,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222473, "pid": 5, "tid": 7, "ts": 1716454224994553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941220, "dur": 12,
    "args": {
      "External id": 222473,
      "cbid": 211, "correlation": 222473
    }
  },
  {
    "ph": "s", "id": 222473, "pid": 76337, "tid": -914061504, "ts": 1716454224941220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941320, "dur": 2,
    "args": {
      "External id": 222491,
      "cbid": 251, "correlation": 222491
    }
  },
  {
    "ph": "f", "id": 222491, "pid": 76337, "tid": -914061504, "ts": 1716454224941320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224994607, "dur": 89,
    "args": {
      "External id": 222493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222493,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222493, "pid": 5, "tid": 7, "ts": 1716454224994607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941327, "dur": 14,
    "args": {
      "External id": 222493,
      "cbid": 211, "correlation": 222493
    }
  },
  {
    "ph": "s", "id": 222493, "pid": 76337, "tid": -914061504, "ts": 1716454224941327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224994698, "dur": 10,
    "args": {
      "External id": 222501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222501,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222501, "pid": 5, "tid": 7, "ts": 1716454224994698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941398, "dur": 12,
    "args": {
      "External id": 222501,
      "cbid": 211, "correlation": 222501
    }
  },
  {
    "ph": "s", "id": 222501, "pid": 76337, "tid": -914061504, "ts": 1716454224941398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224994709, "dur": 20,
    "args": {
      "External id": 222509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222509, "pid": 5, "tid": 7, "ts": 1716454224994709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941439, "dur": 10,
    "args": {
      "External id": 222509,
      "cbid": 211, "correlation": 222509
    }
  },
  {
    "ph": "s", "id": 222509, "pid": 76337, "tid": -914061504, "ts": 1716454224941439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224994730, "dur": 17,
    "args": {
      "External id": 222531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222531,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222531, "pid": 5, "tid": 7, "ts": 1716454224994730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941490, "dur": 10,
    "args": {
      "External id": 222531,
      "cbid": 211, "correlation": 222531
    }
  },
  {
    "ph": "s", "id": 222531, "pid": 76337, "tid": -914061504, "ts": 1716454224941490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941580, "dur": 2,
    "args": {
      "External id": 222547,
      "cbid": 251, "correlation": 222547
    }
  },
  {
    "ph": "f", "id": 222547, "pid": 76337, "tid": -914061504, "ts": 1716454224941580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941586, "dur": 0,
    "args": {
      "External id": 222549,
      "cbid": 251, "correlation": 222549
    }
  },
  {
    "ph": "f", "id": 222549, "pid": 76337, "tid": -914061504, "ts": 1716454224941586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224994749, "dur": 490,
    "args": {
      "External id": 222550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222550,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222550, "pid": 5, "tid": 7, "ts": 1716454224994749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941589, "dur": 15,
    "args": {
      "External id": 222550,
      "cbid": 211, "correlation": 222550
    }
  },
  {
    "ph": "s", "id": 222550, "pid": 76337, "tid": -914061504, "ts": 1716454224941589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224995240, "dur": 65,
    "args": {
      "External id": 222558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222558, "pid": 5, "tid": 7, "ts": 1716454224995240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941655, "dur": 12,
    "args": {
      "External id": 222558,
      "cbid": 211, "correlation": 222558
    }
  },
  {
    "ph": "s", "id": 222558, "pid": 76337, "tid": -914061504, "ts": 1716454224941655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224995306, "dur": 68,
    "args": {
      "External id": 222566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222566, "pid": 5, "tid": 7, "ts": 1716454224995306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941685, "dur": 9,
    "args": {
      "External id": 222566,
      "cbid": 211, "correlation": 222566
    }
  },
  {
    "ph": "s", "id": 222566, "pid": 76337, "tid": -914061504, "ts": 1716454224941685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941765, "dur": 1,
    "args": {
      "External id": 222582,
      "cbid": 251, "correlation": 222582
    }
  },
  {
    "ph": "f", "id": 222582, "pid": 76337, "tid": -914061504, "ts": 1716454224941765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224995376, "dur": 1,
    "args": {
      "External id": 222584,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 222584,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 222584, "pid": 5, "tid": 7, "ts": 1716454224995376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941771, "dur": 12,
    "args": {
      "External id": 222584,
      "cbid": 51, "correlation": 222584
    }
  },
  {
    "ph": "s", "id": 222584, "pid": 76337, "tid": -914061504, "ts": 1716454224941771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224995380, "dur": 268,
    "args": {
      "External id": 222585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222585,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222585, "pid": 5, "tid": 7, "ts": 1716454224995380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941784, "dur": 12,
    "args": {
      "External id": 222585,
      "cbid": 211, "correlation": 222585
    }
  },
  {
    "ph": "s", "id": 222585, "pid": 76337, "tid": -914061504, "ts": 1716454224941784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224995649, "dur": 15,
    "args": {
      "External id": 222593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222593,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222593, "pid": 5, "tid": 7, "ts": 1716454224995649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941826, "dur": 10,
    "args": {
      "External id": 222593,
      "cbid": 211, "correlation": 222593
    }
  },
  {
    "ph": "s", "id": 222593, "pid": 76337, "tid": -914061504, "ts": 1716454224941826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224995665, "dur": 37,
    "args": {
      "External id": 222604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222604, "pid": 5, "tid": 7, "ts": 1716454224995665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941893, "dur": 12,
    "args": {
      "External id": 222604,
      "cbid": 211, "correlation": 222604
    }
  },
  {
    "ph": "s", "id": 222604, "pid": 76337, "tid": -914061504, "ts": 1716454224941893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941958, "dur": 0,
    "args": {
      "External id": 222616,
      "cbid": 317, "correlation": 222616
    }
  },
  {
    "ph": "f", "id": 222616, "pid": 76337, "tid": -914061504, "ts": 1716454224941958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941959, "dur": 0,
    "args": {
      "External id": 222617,
      "cbid": 203, "correlation": 222617
    }
  },
  {
    "ph": "f", "id": 222617, "pid": 76337, "tid": -914061504, "ts": 1716454224941959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941960, "dur": 0,
    "args": {
      "External id": 222618,
      "cbid": 205, "correlation": 222618
    }
  },
  {
    "ph": "f", "id": 222618, "pid": 76337, "tid": -914061504, "ts": 1716454224941960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224995703, "dur": 14,
    "args": {
      "External id": 222622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222622,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222622, "pid": 5, "tid": 7, "ts": 1716454224995703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224941982, "dur": 13,
    "args": {
      "External id": 222622,
      "cbid": 211, "correlation": 222622
    }
  },
  {
    "ph": "s", "id": 222622, "pid": 76337, "tid": -914061504, "ts": 1716454224941982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224995718, "dur": 4,
    "args": {
      "External id": 222624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222624, "pid": 5, "tid": 7, "ts": 1716454224995718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942000, "dur": 6,
    "args": {
      "External id": 222624,
      "cbid": 211, "correlation": 222624
    }
  },
  {
    "ph": "s", "id": 222624, "pid": 76337, "tid": -914061504, "ts": 1716454224942000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942009, "dur": 0,
    "args": {
      "External id": 222625,
      "cbid": 51, "correlation": 222625
    }
  },
  {
    "ph": "s", "id": 222625, "pid": 76337, "tid": -914061504, "ts": 1716454224942009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224995723, "dur": 96,
    "args": {
      "External id": 222626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222626,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 222626, "pid": 5, "tid": 7, "ts": 1716454224995723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942010, "dur": 5,
    "args": {
      "External id": 222626,
      "cbid": 211, "correlation": 222626
    }
  },
  {
    "ph": "s", "id": 222626, "pid": 76337, "tid": -914061504, "ts": 1716454224942010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224995820, "dur": 16,
    "args": {
      "External id": 222631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222631,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222631, "pid": 5, "tid": 7, "ts": 1716454224995820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942037, "dur": 9,
    "args": {
      "External id": 222631,
      "cbid": 211, "correlation": 222631
    }
  },
  {
    "ph": "s", "id": 222631, "pid": 76337, "tid": -914061504, "ts": 1716454224942037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224995837, "dur": 11,
    "args": {
      "External id": 222639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222639,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222639, "pid": 5, "tid": 7, "ts": 1716454224995837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942070, "dur": 8,
    "args": {
      "External id": 222639,
      "cbid": 211, "correlation": 222639
    }
  },
  {
    "ph": "s", "id": 222639, "pid": 76337, "tid": -914061504, "ts": 1716454224942070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224995849, "dur": 18,
    "args": {
      "External id": 222659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222659,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 222659, "pid": 5, "tid": 7, "ts": 1716454224995849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942142, "dur": 12,
    "args": {
      "External id": 222659,
      "cbid": 211, "correlation": 222659
    }
  },
  {
    "ph": "s", "id": 222659, "pid": 76337, "tid": -914061504, "ts": 1716454224942142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224995869, "dur": 5,
    "args": {
      "External id": 222671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222671,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 222671, "pid": 5, "tid": 7, "ts": 1716454224995869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942164, "dur": 6,
    "args": {
      "External id": 222671,
      "cbid": 211, "correlation": 222671
    }
  },
  {
    "ph": "s", "id": 222671, "pid": 76337, "tid": -914061504, "ts": 1716454224942164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224995875, "dur": 18,
    "args": {
      "External id": 222674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222674, "pid": 5, "tid": 7, "ts": 1716454224995875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942181, "dur": 6,
    "args": {
      "External id": 222674,
      "cbid": 211, "correlation": 222674
    }
  },
  {
    "ph": "s", "id": 222674, "pid": 76337, "tid": -914061504, "ts": 1716454224942181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224995894, "dur": 11,
    "args": {
      "External id": 222683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222683,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222683, "pid": 5, "tid": 7, "ts": 1716454224995894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942220, "dur": 9,
    "args": {
      "External id": 222683,
      "cbid": 211, "correlation": 222683
    }
  },
  {
    "ph": "s", "id": 222683, "pid": 76337, "tid": -914061504, "ts": 1716454224942220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942270, "dur": 0,
    "args": {
      "External id": 222693,
      "cbid": 317, "correlation": 222693
    }
  },
  {
    "ph": "f", "id": 222693, "pid": 76337, "tid": -914061504, "ts": 1716454224942270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942271, "dur": 0,
    "args": {
      "External id": 222694,
      "cbid": 203, "correlation": 222694
    }
  },
  {
    "ph": "f", "id": 222694, "pid": 76337, "tid": -914061504, "ts": 1716454224942271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942271, "dur": 0,
    "args": {
      "External id": 222695,
      "cbid": 205, "correlation": 222695
    }
  },
  {
    "ph": "f", "id": 222695, "pid": 76337, "tid": -914061504, "ts": 1716454224942271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224995907, "dur": 11,
    "args": {
      "External id": 222699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222699, "pid": 5, "tid": 7, "ts": 1716454224995907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942284, "dur": 13,
    "args": {
      "External id": 222699,
      "cbid": 211, "correlation": 222699
    }
  },
  {
    "ph": "s", "id": 222699, "pid": 76337, "tid": -914061504, "ts": 1716454224942284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224995919, "dur": 160,
    "args": {
      "External id": 222701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222701,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222701, "pid": 5, "tid": 7, "ts": 1716454224995919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942300, "dur": 5,
    "args": {
      "External id": 222701,
      "cbid": 211, "correlation": 222701
    }
  },
  {
    "ph": "s", "id": 222701, "pid": 76337, "tid": -914061504, "ts": 1716454224942300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224996081, "dur": 1,
    "args": {
      "External id": 222703,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 222703,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 222703, "pid": 5, "tid": 7, "ts": 1716454224996081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942310, "dur": 6,
    "args": {
      "External id": 222703,
      "cbid": 51, "correlation": 222703
    }
  },
  {
    "ph": "s", "id": 222703, "pid": 76337, "tid": -914061504, "ts": 1716454224942310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224996085, "dur": 655,
    "args": {
      "External id": 222704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222704,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222704, "pid": 5, "tid": 7, "ts": 1716454224996085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942318, "dur": 6,
    "args": {
      "External id": 222704,
      "cbid": 211, "correlation": 222704
    }
  },
  {
    "ph": "s", "id": 222704, "pid": 76337, "tid": -914061504, "ts": 1716454224942318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224996742, "dur": 13,
    "args": {
      "External id": 222706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222706, "pid": 5, "tid": 7, "ts": 1716454224996742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942328, "dur": 5,
    "args": {
      "External id": 222706,
      "cbid": 211, "correlation": 222706
    }
  },
  {
    "ph": "s", "id": 222706, "pid": 76337, "tid": -914061504, "ts": 1716454224942328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224996756, "dur": 15,
    "args": {
      "External id": 222712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222712, "pid": 5, "tid": 7, "ts": 1716454224996756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942356, "dur": 8,
    "args": {
      "External id": 222712,
      "cbid": 211, "correlation": 222712
    }
  },
  {
    "ph": "s", "id": 222712, "pid": 76337, "tid": -914061504, "ts": 1716454224942356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224996773, "dur": 3,
    "args": {
      "External id": 222720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222720,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 222720, "pid": 5, "tid": 7, "ts": 1716454224996773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942399, "dur": 9,
    "args": {
      "External id": 222720,
      "cbid": 211, "correlation": 222720
    }
  },
  {
    "ph": "s", "id": 222720, "pid": 76337, "tid": -914061504, "ts": 1716454224942399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942463, "dur": 1,
    "args": {
      "External id": 222736,
      "cbid": 251, "correlation": 222736
    }
  },
  {
    "ph": "f", "id": 222736, "pid": 76337, "tid": -914061504, "ts": 1716454224942463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942468, "dur": 0,
    "args": {
      "External id": 222738,
      "cbid": 251, "correlation": 222738
    }
  },
  {
    "ph": "f", "id": 222738, "pid": 76337, "tid": -914061504, "ts": 1716454224942468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224996777, "dur": 13,
    "args": {
      "External id": 222739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222739,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222739, "pid": 5, "tid": 7, "ts": 1716454224996777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942470, "dur": 11,
    "args": {
      "External id": 222739,
      "cbid": 211, "correlation": 222739
    }
  },
  {
    "ph": "s", "id": 222739, "pid": 76337, "tid": -914061504, "ts": 1716454224942470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224996792, "dur": 5,
    "args": {
      "External id": 222741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222741,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222741, "pid": 5, "tid": 7, "ts": 1716454224996792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942483, "dur": 5,
    "args": {
      "External id": 222741,
      "cbid": 211, "correlation": 222741
    }
  },
  {
    "ph": "s", "id": 222741, "pid": 76337, "tid": -914061504, "ts": 1716454224942483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224996798, "dur": 17,
    "args": {
      "External id": 222751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222751, "pid": 5, "tid": 7, "ts": 1716454224996798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942541, "dur": 12,
    "args": {
      "External id": 222751,
      "cbid": 211, "correlation": 222751
    }
  },
  {
    "ph": "s", "id": 222751, "pid": 76337, "tid": -914061504, "ts": 1716454224942541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224996816, "dur": 18,
    "args": {
      "External id": 222771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222771,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 222771, "pid": 5, "tid": 7, "ts": 1716454224996816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942606, "dur": 10,
    "args": {
      "External id": 222771,
      "cbid": 211, "correlation": 222771
    }
  },
  {
    "ph": "s", "id": 222771, "pid": 76337, "tid": -914061504, "ts": 1716454224942606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224996835, "dur": 4,
    "args": {
      "External id": 222783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222783,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 222783, "pid": 5, "tid": 7, "ts": 1716454224996835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942626, "dur": 6,
    "args": {
      "External id": 222783,
      "cbid": 211, "correlation": 222783
    }
  },
  {
    "ph": "s", "id": 222783, "pid": 76337, "tid": -914061504, "ts": 1716454224942626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224996840, "dur": 16,
    "args": {
      "External id": 222786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222786, "pid": 5, "tid": 7, "ts": 1716454224996840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942645, "dur": 6,
    "args": {
      "External id": 222786,
      "cbid": 211, "correlation": 222786
    }
  },
  {
    "ph": "s", "id": 222786, "pid": 76337, "tid": -914061504, "ts": 1716454224942645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224996858, "dur": 11,
    "args": {
      "External id": 222795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222795,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222795, "pid": 5, "tid": 7, "ts": 1716454224996858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942684, "dur": 10,
    "args": {
      "External id": 222795,
      "cbid": 211, "correlation": 222795
    }
  },
  {
    "ph": "s", "id": 222795, "pid": 76337, "tid": -914061504, "ts": 1716454224942684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942745, "dur": 0,
    "args": {
      "External id": 222805,
      "cbid": 317, "correlation": 222805
    }
  },
  {
    "ph": "f", "id": 222805, "pid": 76337, "tid": -914061504, "ts": 1716454224942745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942746, "dur": 0,
    "args": {
      "External id": 222806,
      "cbid": 203, "correlation": 222806
    }
  },
  {
    "ph": "f", "id": 222806, "pid": 76337, "tid": -914061504, "ts": 1716454224942746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942746, "dur": 0,
    "args": {
      "External id": 222807,
      "cbid": 205, "correlation": 222807
    }
  },
  {
    "ph": "f", "id": 222807, "pid": 76337, "tid": -914061504, "ts": 1716454224942746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224996870, "dur": 10,
    "args": {
      "External id": 222811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222811,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222811, "pid": 5, "tid": 7, "ts": 1716454224996870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942760, "dur": 11,
    "args": {
      "External id": 222811,
      "cbid": 211, "correlation": 222811
    }
  },
  {
    "ph": "s", "id": 222811, "pid": 76337, "tid": -914061504, "ts": 1716454224942760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224996881, "dur": 159,
    "args": {
      "External id": 222813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222813, "pid": 5, "tid": 7, "ts": 1716454224996881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942774, "dur": 5,
    "args": {
      "External id": 222813,
      "cbid": 211, "correlation": 222813
    }
  },
  {
    "ph": "s", "id": 222813, "pid": 76337, "tid": -914061504, "ts": 1716454224942774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224997043, "dur": 1,
    "args": {
      "External id": 222815,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 222815,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 222815, "pid": 5, "tid": 7, "ts": 1716454224997043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942785, "dur": 8,
    "args": {
      "External id": 222815,
      "cbid": 51, "correlation": 222815
    }
  },
  {
    "ph": "s", "id": 222815, "pid": 76337, "tid": -914061504, "ts": 1716454224942785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224997047, "dur": 641,
    "args": {
      "External id": 222816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222816,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222816, "pid": 5, "tid": 7, "ts": 1716454224997047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942795, "dur": 7,
    "args": {
      "External id": 222816,
      "cbid": 211, "correlation": 222816
    }
  },
  {
    "ph": "s", "id": 222816, "pid": 76337, "tid": -914061504, "ts": 1716454224942795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224997689, "dur": 12,
    "args": {
      "External id": 222818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222818,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222818, "pid": 5, "tid": 7, "ts": 1716454224997689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942805, "dur": 5,
    "args": {
      "External id": 222818,
      "cbid": 211, "correlation": 222818
    }
  },
  {
    "ph": "s", "id": 222818, "pid": 76337, "tid": -914061504, "ts": 1716454224942805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224997703, "dur": 14,
    "args": {
      "External id": 222824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222824, "pid": 5, "tid": 7, "ts": 1716454224997703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942833, "dur": 9,
    "args": {
      "External id": 222824,
      "cbid": 211, "correlation": 222824
    }
  },
  {
    "ph": "s", "id": 222824, "pid": 76337, "tid": -914061504, "ts": 1716454224942833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224997718, "dur": 12,
    "args": {
      "External id": 222832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222832,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222832, "pid": 5, "tid": 7, "ts": 1716454224997718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942865, "dur": 8,
    "args": {
      "External id": 222832,
      "cbid": 211, "correlation": 222832
    }
  },
  {
    "ph": "s", "id": 222832, "pid": 76337, "tid": -914061504, "ts": 1716454224942865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224997732, "dur": 10,
    "args": {
      "External id": 222840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222840,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222840, "pid": 5, "tid": 7, "ts": 1716454224997732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942895, "dur": 8,
    "args": {
      "External id": 222840,
      "cbid": 211, "correlation": 222840
    }
  },
  {
    "ph": "s", "id": 222840, "pid": 76337, "tid": -914061504, "ts": 1716454224942895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224997743, "dur": 17,
    "args": {
      "External id": 222860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222860,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 222860, "pid": 5, "tid": 7, "ts": 1716454224997743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224942972, "dur": 20,
    "args": {
      "External id": 222860,
      "cbid": 211, "correlation": 222860
    }
  },
  {
    "ph": "s", "id": 222860, "pid": 76337, "tid": -914061504, "ts": 1716454224942972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224997762, "dur": 4,
    "args": {
      "External id": 222872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222872,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 222872, "pid": 5, "tid": 7, "ts": 1716454224997762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943003, "dur": 6,
    "args": {
      "External id": 222872,
      "cbid": 211, "correlation": 222872
    }
  },
  {
    "ph": "s", "id": 222872, "pid": 76337, "tid": -914061504, "ts": 1716454224943003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224997767, "dur": 16,
    "args": {
      "External id": 222875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222875, "pid": 5, "tid": 7, "ts": 1716454224997767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943021, "dur": 6,
    "args": {
      "External id": 222875,
      "cbid": 211, "correlation": 222875
    }
  },
  {
    "ph": "s", "id": 222875, "pid": 76337, "tid": -914061504, "ts": 1716454224943021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943079, "dur": 0,
    "args": {
      "External id": 222886,
      "cbid": 317, "correlation": 222886
    }
  },
  {
    "ph": "f", "id": 222886, "pid": 76337, "tid": -914061504, "ts": 1716454224943079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943080, "dur": 0,
    "args": {
      "External id": 222887,
      "cbid": 203, "correlation": 222887
    }
  },
  {
    "ph": "f", "id": 222887, "pid": 76337, "tid": -914061504, "ts": 1716454224943080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943081, "dur": 0,
    "args": {
      "External id": 222888,
      "cbid": 205, "correlation": 222888
    }
  },
  {
    "ph": "f", "id": 222888, "pid": 76337, "tid": -914061504, "ts": 1716454224943081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224997785, "dur": 11,
    "args": {
      "External id": 222892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222892,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222892, "pid": 5, "tid": 7, "ts": 1716454224997785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943095, "dur": 13,
    "args": {
      "External id": 222892,
      "cbid": 211, "correlation": 222892
    }
  },
  {
    "ph": "s", "id": 222892, "pid": 76337, "tid": -914061504, "ts": 1716454224943095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224997798, "dur": 4,
    "args": {
      "External id": 222894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222894,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 222894, "pid": 5, "tid": 7, "ts": 1716454224997798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943112, "dur": 6,
    "args": {
      "External id": 222894,
      "cbid": 211, "correlation": 222894
    }
  },
  {
    "ph": "s", "id": 222894, "pid": 76337, "tid": -914061504, "ts": 1716454224943112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943120, "dur": 0,
    "args": {
      "External id": 222895,
      "cbid": 51, "correlation": 222895
    }
  },
  {
    "ph": "s", "id": 222895, "pid": 76337, "tid": -914061504, "ts": 1716454224943120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224997803, "dur": 94,
    "args": {
      "External id": 222896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222896,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 222896, "pid": 5, "tid": 7, "ts": 1716454224997803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943121, "dur": 5,
    "args": {
      "External id": 222896,
      "cbid": 211, "correlation": 222896
    }
  },
  {
    "ph": "s", "id": 222896, "pid": 76337, "tid": -914061504, "ts": 1716454224943121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224997898, "dur": 15,
    "args": {
      "External id": 222901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222901, "pid": 5, "tid": 7, "ts": 1716454224997898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943148, "dur": 8,
    "args": {
      "External id": 222901,
      "cbid": 211, "correlation": 222901
    }
  },
  {
    "ph": "s", "id": 222901, "pid": 76337, "tid": -914061504, "ts": 1716454224943148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224997914, "dur": 82,
    "args": {
      "External id": 222910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222910,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222910, "pid": 5, "tid": 7, "ts": 1716454224997914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943228, "dur": 14,
    "args": {
      "External id": 222910,
      "cbid": 211, "correlation": 222910
    }
  },
  {
    "ph": "s", "id": 222910, "pid": 76337, "tid": -914061504, "ts": 1716454224943228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224997997, "dur": 30,
    "args": {
      "External id": 222932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222932,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 222932, "pid": 5, "tid": 7, "ts": 1716454224997997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943283, "dur": 10,
    "args": {
      "External id": 222932,
      "cbid": 211, "correlation": 222932
    }
  },
  {
    "ph": "s", "id": 222932, "pid": 76337, "tid": -914061504, "ts": 1716454224943283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943370, "dur": 1,
    "args": {
      "External id": 222943,
      "cbid": 251, "correlation": 222943
    }
  },
  {
    "ph": "f", "id": 222943, "pid": 76337, "tid": -914061504, "ts": 1716454224943370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224998029, "dur": 162,
    "args": {
      "External id": 222944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222944,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222944, "pid": 5, "tid": 7, "ts": 1716454224998029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943376, "dur": 13,
    "args": {
      "External id": 222944,
      "cbid": 211, "correlation": 222944
    }
  },
  {
    "ph": "s", "id": 222944, "pid": 76337, "tid": -914061504, "ts": 1716454224943376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943446, "dur": 1,
    "args": {
      "External id": 222955,
      "cbid": 251, "correlation": 222955
    }
  },
  {
    "ph": "f", "id": 222955, "pid": 76337, "tid": -914061504, "ts": 1716454224943446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224998192, "dur": 155,
    "args": {
      "External id": 222956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222956,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222956, "pid": 5, "tid": 7, "ts": 1716454224998192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943450, "dur": 11,
    "args": {
      "External id": 222956,
      "cbid": 211, "correlation": 222956
    }
  },
  {
    "ph": "s", "id": 222956, "pid": 76337, "tid": -914061504, "ts": 1716454224943450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943515, "dur": 1,
    "args": {
      "External id": 222967,
      "cbid": 251, "correlation": 222967
    }
  },
  {
    "ph": "f", "id": 222967, "pid": 76337, "tid": -914061504, "ts": 1716454224943515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224998349, "dur": 135,
    "args": {
      "External id": 222968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222968,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222968, "pid": 5, "tid": 7, "ts": 1716454224998349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943519, "dur": 11,
    "args": {
      "External id": 222968,
      "cbid": 211, "correlation": 222968
    }
  },
  {
    "ph": "s", "id": 222968, "pid": 76337, "tid": -914061504, "ts": 1716454224943519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224998485, "dur": 333,
    "args": {
      "External id": 222993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 222993,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 222993, "pid": 5, "tid": 7, "ts": 1716454224998485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943603, "dur": 12,
    "args": {
      "External id": 222993,
      "cbid": 211, "correlation": 222993
    }
  },
  {
    "ph": "s", "id": 222993, "pid": 76337, "tid": -914061504, "ts": 1716454224943603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943700, "dur": 1,
    "args": {
      "External id": 223011,
      "cbid": 251, "correlation": 223011
    }
  },
  {
    "ph": "f", "id": 223011, "pid": 76337, "tid": -914061504, "ts": 1716454224943700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224998819, "dur": 164,
    "args": {
      "External id": 223013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223013,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223013, "pid": 5, "tid": 7, "ts": 1716454224998819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943707, "dur": 13,
    "args": {
      "External id": 223013,
      "cbid": 211, "correlation": 223013
    }
  },
  {
    "ph": "s", "id": 223013, "pid": 76337, "tid": -914061504, "ts": 1716454224943707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224998985, "dur": 19,
    "args": {
      "External id": 223021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223021,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223021, "pid": 5, "tid": 7, "ts": 1716454224998985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943775, "dur": 12,
    "args": {
      "External id": 223021,
      "cbid": 211, "correlation": 223021
    }
  },
  {
    "ph": "s", "id": 223021, "pid": 76337, "tid": -914061504, "ts": 1716454224943775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224999005, "dur": 28,
    "args": {
      "External id": 223029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223029,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223029, "pid": 5, "tid": 7, "ts": 1716454224999005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943814, "dur": 8,
    "args": {
      "External id": 223029,
      "cbid": 211, "correlation": 223029
    }
  },
  {
    "ph": "s", "id": 223029, "pid": 76337, "tid": -914061504, "ts": 1716454224943814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224999034, "dur": 18,
    "args": {
      "External id": 223040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223040, "pid": 5, "tid": 7, "ts": 1716454224999034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943884, "dur": 12,
    "args": {
      "External id": 223040,
      "cbid": 211, "correlation": 223040
    }
  },
  {
    "ph": "s", "id": 223040, "pid": 76337, "tid": -914061504, "ts": 1716454224943884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224999053, "dur": 16,
    "args": {
      "External id": 223062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223062,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223062, "pid": 5, "tid": 7, "ts": 1716454224999053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224943916, "dur": 7,
    "args": {
      "External id": 223062,
      "cbid": 211, "correlation": 223062
    }
  },
  {
    "ph": "s", "id": 223062, "pid": 76337, "tid": -914061504, "ts": 1716454224943916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944011, "dur": 1,
    "args": {
      "External id": 223073,
      "cbid": 251, "correlation": 223073
    }
  },
  {
    "ph": "f", "id": 223073, "pid": 76337, "tid": -914061504, "ts": 1716454224944011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224999070, "dur": 88,
    "args": {
      "External id": 223074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223074,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 223074, "pid": 5, "tid": 7, "ts": 1716454224999070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944017, "dur": 13,
    "args": {
      "External id": 223074,
      "cbid": 211, "correlation": 223074
    }
  },
  {
    "ph": "s", "id": 223074, "pid": 76337, "tid": -914061504, "ts": 1716454224944017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944085, "dur": 1,
    "args": {
      "External id": 223085,
      "cbid": 251, "correlation": 223085
    }
  },
  {
    "ph": "f", "id": 223085, "pid": 76337, "tid": -914061504, "ts": 1716454224944085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944088, "dur": 0,
    "args": {
      "External id": 223086,
      "cbid": 251, "correlation": 223086
    }
  },
  {
    "ph": "f", "id": 223086, "pid": 76337, "tid": -914061504, "ts": 1716454224944088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224999159, "dur": 12,
    "args": {
      "External id": 223087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223087,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223087, "pid": 5, "tid": 7, "ts": 1716454224999159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944090, "dur": 12,
    "args": {
      "External id": 223087,
      "cbid": 211, "correlation": 223087
    }
  },
  {
    "ph": "s", "id": 223087, "pid": 76337, "tid": -914061504, "ts": 1716454224944090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224999172, "dur": 5,
    "args": {
      "External id": 223089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223089,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223089, "pid": 5, "tid": 7, "ts": 1716454224999172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944104, "dur": 6,
    "args": {
      "External id": 223089,
      "cbid": 211, "correlation": 223089
    }
  },
  {
    "ph": "s", "id": 223089, "pid": 76337, "tid": -914061504, "ts": 1716454224944104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944163, "dur": 1,
    "args": {
      "External id": 223100,
      "cbid": 251, "correlation": 223100
    }
  },
  {
    "ph": "f", "id": 223100, "pid": 76337, "tid": -914061504, "ts": 1716454224944163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944166, "dur": 0,
    "args": {
      "External id": 223101,
      "cbid": 251, "correlation": 223101
    }
  },
  {
    "ph": "f", "id": 223101, "pid": 76337, "tid": -914061504, "ts": 1716454224944166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224999178, "dur": 8,
    "args": {
      "External id": 223102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223102,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223102, "pid": 5, "tid": 7, "ts": 1716454224999178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944168, "dur": 11,
    "args": {
      "External id": 223102,
      "cbid": 211, "correlation": 223102
    }
  },
  {
    "ph": "s", "id": 223102, "pid": 76337, "tid": -914061504, "ts": 1716454224944168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224999188, "dur": 3,
    "args": {
      "External id": 223104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223104,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223104, "pid": 5, "tid": 7, "ts": 1716454224999188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944180, "dur": 5,
    "args": {
      "External id": 223104,
      "cbid": 211, "correlation": 223104
    }
  },
  {
    "ph": "s", "id": 223104, "pid": 76337, "tid": -914061504, "ts": 1716454224944180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224999193, "dur": 55,
    "args": {
      "External id": 223129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223129,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223129, "pid": 5, "tid": 7, "ts": 1716454224999193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944255, "dur": 13,
    "args": {
      "External id": 223129,
      "cbid": 211, "correlation": 223129
    }
  },
  {
    "ph": "s", "id": 223129, "pid": 76337, "tid": -914061504, "ts": 1716454224944255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944352, "dur": 1,
    "args": {
      "External id": 223147,
      "cbid": 251, "correlation": 223147
    }
  },
  {
    "ph": "f", "id": 223147, "pid": 76337, "tid": -914061504, "ts": 1716454224944352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224999249, "dur": 89,
    "args": {
      "External id": 223149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223149,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 223149, "pid": 5, "tid": 7, "ts": 1716454224999249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944358, "dur": 15,
    "args": {
      "External id": 223149,
      "cbid": 211, "correlation": 223149
    }
  },
  {
    "ph": "s", "id": 223149, "pid": 76337, "tid": -914061504, "ts": 1716454224944358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224999340, "dur": 10,
    "args": {
      "External id": 223157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223157,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223157, "pid": 5, "tid": 7, "ts": 1716454224999340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944429, "dur": 12,
    "args": {
      "External id": 223157,
      "cbid": 211, "correlation": 223157
    }
  },
  {
    "ph": "s", "id": 223157, "pid": 76337, "tid": -914061504, "ts": 1716454224944429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224999351, "dur": 22,
    "args": {
      "External id": 223165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223165, "pid": 5, "tid": 7, "ts": 1716454224999351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944470, "dur": 9,
    "args": {
      "External id": 223165,
      "cbid": 211, "correlation": 223165
    }
  },
  {
    "ph": "s", "id": 223165, "pid": 76337, "tid": -914061504, "ts": 1716454224944470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224999374, "dur": 17,
    "args": {
      "External id": 223187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223187,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223187, "pid": 5, "tid": 7, "ts": 1716454224999374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944522, "dur": 10,
    "args": {
      "External id": 223187,
      "cbid": 211, "correlation": 223187
    }
  },
  {
    "ph": "s", "id": 223187, "pid": 76337, "tid": -914061504, "ts": 1716454224944522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944609, "dur": 1,
    "args": {
      "External id": 223203,
      "cbid": 251, "correlation": 223203
    }
  },
  {
    "ph": "f", "id": 223203, "pid": 76337, "tid": -914061504, "ts": 1716454224944609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944614, "dur": 0,
    "args": {
      "External id": 223205,
      "cbid": 251, "correlation": 223205
    }
  },
  {
    "ph": "f", "id": 223205, "pid": 76337, "tid": -914061504, "ts": 1716454224944614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224999393, "dur": 491,
    "args": {
      "External id": 223206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223206,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223206, "pid": 5, "tid": 7, "ts": 1716454224999393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944616, "dur": 12,
    "args": {
      "External id": 223206,
      "cbid": 211, "correlation": 223206
    }
  },
  {
    "ph": "s", "id": 223206, "pid": 76337, "tid": -914061504, "ts": 1716454224944616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224999886, "dur": 66,
    "args": {
      "External id": 223214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223214, "pid": 5, "tid": 7, "ts": 1716454224999886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944680, "dur": 12,
    "args": {
      "External id": 223214,
      "cbid": 211, "correlation": 223214
    }
  },
  {
    "ph": "s", "id": 223214, "pid": 76337, "tid": -914061504, "ts": 1716454224944680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224999953, "dur": 66,
    "args": {
      "External id": 223222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223222, "pid": 5, "tid": 7, "ts": 1716454224999953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944709, "dur": 9,
    "args": {
      "External id": 223222,
      "cbid": 211, "correlation": 223222
    }
  },
  {
    "ph": "s", "id": 223222, "pid": 76337, "tid": -914061504, "ts": 1716454224944709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944789, "dur": 1,
    "args": {
      "External id": 223238,
      "cbid": 251, "correlation": 223238
    }
  },
  {
    "ph": "f", "id": 223238, "pid": 76337, "tid": -914061504, "ts": 1716454224944789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225000021, "dur": 1,
    "args": {
      "External id": 223240,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223240,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 223240, "pid": 5, "tid": 7, "ts": 1716454225000021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944794, "dur": 9,
    "args": {
      "External id": 223240,
      "cbid": 51, "correlation": 223240
    }
  },
  {
    "ph": "s", "id": 223240, "pid": 76337, "tid": -914061504, "ts": 1716454224944794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225000024, "dur": 268,
    "args": {
      "External id": 223241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223241,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 223241, "pid": 5, "tid": 7, "ts": 1716454225000024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944805, "dur": 12,
    "args": {
      "External id": 223241,
      "cbid": 211, "correlation": 223241
    }
  },
  {
    "ph": "s", "id": 223241, "pid": 76337, "tid": -914061504, "ts": 1716454224944805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225000294, "dur": 13,
    "args": {
      "External id": 223249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223249,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223249, "pid": 5, "tid": 7, "ts": 1716454225000294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944847, "dur": 10,
    "args": {
      "External id": 223249,
      "cbid": 211, "correlation": 223249
    }
  },
  {
    "ph": "s", "id": 223249, "pid": 76337, "tid": -914061504, "ts": 1716454224944847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225000308, "dur": 37,
    "args": {
      "External id": 223260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223260,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223260, "pid": 5, "tid": 7, "ts": 1716454225000308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944914, "dur": 12,
    "args": {
      "External id": 223260,
      "cbid": 211, "correlation": 223260
    }
  },
  {
    "ph": "s", "id": 223260, "pid": 76337, "tid": -914061504, "ts": 1716454224944914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944986, "dur": 0,
    "args": {
      "External id": 223272,
      "cbid": 317, "correlation": 223272
    }
  },
  {
    "ph": "f", "id": 223272, "pid": 76337, "tid": -914061504, "ts": 1716454224944986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944987, "dur": 0,
    "args": {
      "External id": 223273,
      "cbid": 203, "correlation": 223273
    }
  },
  {
    "ph": "f", "id": 223273, "pid": 76337, "tid": -914061504, "ts": 1716454224944987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224944988, "dur": 0,
    "args": {
      "External id": 223274,
      "cbid": 205, "correlation": 223274
    }
  },
  {
    "ph": "f", "id": 223274, "pid": 76337, "tid": -914061504, "ts": 1716454224944988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000347, "dur": 13,
    "args": {
      "External id": 223278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223278,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223278, "pid": 5, "tid": 7, "ts": 1716454225000347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945004, "dur": 13,
    "args": {
      "External id": 223278,
      "cbid": 211, "correlation": 223278
    }
  },
  {
    "ph": "s", "id": 223278, "pid": 76337, "tid": -914061504, "ts": 1716454224945004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225000361, "dur": 4,
    "args": {
      "External id": 223280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 223280, "pid": 5, "tid": 7, "ts": 1716454225000361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945022, "dur": 7,
    "args": {
      "External id": 223280,
      "cbid": 211, "correlation": 223280
    }
  },
  {
    "ph": "s", "id": 223280, "pid": 76337, "tid": -914061504, "ts": 1716454224945022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945031, "dur": 0,
    "args": {
      "External id": 223281,
      "cbid": 51, "correlation": 223281
    }
  },
  {
    "ph": "s", "id": 223281, "pid": 76337, "tid": -914061504, "ts": 1716454224945031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225000366, "dur": 96,
    "args": {
      "External id": 223282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223282,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 223282, "pid": 5, "tid": 7, "ts": 1716454225000366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945032, "dur": 5,
    "args": {
      "External id": 223282,
      "cbid": 211, "correlation": 223282
    }
  },
  {
    "ph": "s", "id": 223282, "pid": 76337, "tid": -914061504, "ts": 1716454224945032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225000464, "dur": 16,
    "args": {
      "External id": 223287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223287, "pid": 5, "tid": 7, "ts": 1716454225000464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945059, "dur": 9,
    "args": {
      "External id": 223287,
      "cbid": 211, "correlation": 223287
    }
  },
  {
    "ph": "s", "id": 223287, "pid": 76337, "tid": -914061504, "ts": 1716454224945059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225000481, "dur": 12,
    "args": {
      "External id": 223295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223295,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223295, "pid": 5, "tid": 7, "ts": 1716454225000481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945091, "dur": 8,
    "args": {
      "External id": 223295,
      "cbid": 211, "correlation": 223295
    }
  },
  {
    "ph": "s", "id": 223295, "pid": 76337, "tid": -914061504, "ts": 1716454224945091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945160, "dur": 0,
    "args": {
      "External id": 223305,
      "cbid": 317, "correlation": 223305
    }
  },
  {
    "ph": "f", "id": 223305, "pid": 76337, "tid": -914061504, "ts": 1716454224945160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945161, "dur": 0,
    "args": {
      "External id": 223306,
      "cbid": 203, "correlation": 223306
    }
  },
  {
    "ph": "f", "id": 223306, "pid": 76337, "tid": -914061504, "ts": 1716454224945161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945162, "dur": 0,
    "args": {
      "External id": 223307,
      "cbid": 205, "correlation": 223307
    }
  },
  {
    "ph": "f", "id": 223307, "pid": 76337, "tid": -914061504, "ts": 1716454224945162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000494, "dur": 12,
    "args": {
      "External id": 223311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223311,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223311, "pid": 5, "tid": 7, "ts": 1716454225000494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945176, "dur": 12,
    "args": {
      "External id": 223311,
      "cbid": 211, "correlation": 223311
    }
  },
  {
    "ph": "s", "id": 223311, "pid": 76337, "tid": -914061504, "ts": 1716454224945176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000508, "dur": 160,
    "args": {
      "External id": 223313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223313, "pid": 5, "tid": 7, "ts": 1716454225000508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945191, "dur": 5,
    "args": {
      "External id": 223313,
      "cbid": 211, "correlation": 223313
    }
  },
  {
    "ph": "s", "id": 223313, "pid": 76337, "tid": -914061504, "ts": 1716454224945191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225000669, "dur": 1,
    "args": {
      "External id": 223315,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223315,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 223315, "pid": 5, "tid": 7, "ts": 1716454225000669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945202, "dur": 7,
    "args": {
      "External id": 223315,
      "cbid": 51, "correlation": 223315
    }
  },
  {
    "ph": "s", "id": 223315, "pid": 76337, "tid": -914061504, "ts": 1716454224945202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225000673, "dur": 197,
    "args": {
      "External id": 223316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223316,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 223316, "pid": 5, "tid": 7, "ts": 1716454225000673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945210, "dur": 8,
    "args": {
      "External id": 223316,
      "cbid": 211, "correlation": 223316
    }
  },
  {
    "ph": "s", "id": 223316, "pid": 76337, "tid": -914061504, "ts": 1716454224945210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000872, "dur": 6,
    "args": {
      "External id": 223318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223318, "pid": 5, "tid": 7, "ts": 1716454225000872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945223, "dur": 5,
    "args": {
      "External id": 223318,
      "cbid": 211, "correlation": 223318
    }
  },
  {
    "ph": "s", "id": 223318, "pid": 76337, "tid": -914061504, "ts": 1716454224945223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225000879, "dur": 6,
    "args": {
      "External id": 223324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223324, "pid": 5, "tid": 7, "ts": 1716454225000879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945250, "dur": 10,
    "args": {
      "External id": 223324,
      "cbid": 211, "correlation": 223324
    }
  },
  {
    "ph": "s", "id": 223324, "pid": 76337, "tid": -914061504, "ts": 1716454224945250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000886, "dur": 11,
    "args": {
      "External id": 223344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223344,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223344, "pid": 5, "tid": 7, "ts": 1716454225000886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945344, "dur": 12,
    "args": {
      "External id": 223344,
      "cbid": 211, "correlation": 223344
    }
  },
  {
    "ph": "s", "id": 223344, "pid": 76337, "tid": -914061504, "ts": 1716454224945344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225000898, "dur": 4,
    "args": {
      "External id": 223356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223356,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223356, "pid": 5, "tid": 7, "ts": 1716454225000898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945368, "dur": 6,
    "args": {
      "External id": 223356,
      "cbid": 211, "correlation": 223356
    }
  },
  {
    "ph": "s", "id": 223356, "pid": 76337, "tid": -914061504, "ts": 1716454224945368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225000904, "dur": 8,
    "args": {
      "External id": 223359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223359, "pid": 5, "tid": 7, "ts": 1716454225000904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945386, "dur": 6,
    "args": {
      "External id": 223359,
      "cbid": 211, "correlation": 223359
    }
  },
  {
    "ph": "s", "id": 223359, "pid": 76337, "tid": -914061504, "ts": 1716454224945386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225000913, "dur": 5,
    "args": {
      "External id": 223368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223368,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223368, "pid": 5, "tid": 7, "ts": 1716454225000913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945425, "dur": 9,
    "args": {
      "External id": 223368,
      "cbid": 211, "correlation": 223368
    }
  },
  {
    "ph": "s", "id": 223368, "pid": 76337, "tid": -914061504, "ts": 1716454224945425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945477, "dur": 0,
    "args": {
      "External id": 223378,
      "cbid": 317, "correlation": 223378
    }
  },
  {
    "ph": "f", "id": 223378, "pid": 76337, "tid": -914061504, "ts": 1716454224945477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945478, "dur": 0,
    "args": {
      "External id": 223379,
      "cbid": 203, "correlation": 223379
    }
  },
  {
    "ph": "f", "id": 223379, "pid": 76337, "tid": -914061504, "ts": 1716454224945478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945479, "dur": 0,
    "args": {
      "External id": 223380,
      "cbid": 205, "correlation": 223380
    }
  },
  {
    "ph": "f", "id": 223380, "pid": 76337, "tid": -914061504, "ts": 1716454224945479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000920, "dur": 5,
    "args": {
      "External id": 223384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223384,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223384, "pid": 5, "tid": 7, "ts": 1716454225000920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945493, "dur": 11,
    "args": {
      "External id": 223384,
      "cbid": 211, "correlation": 223384
    }
  },
  {
    "ph": "s", "id": 223384, "pid": 76337, "tid": -914061504, "ts": 1716454224945493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225000926, "dur": 159,
    "args": {
      "External id": 223386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223386,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223386, "pid": 5, "tid": 7, "ts": 1716454225000926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945507, "dur": 5,
    "args": {
      "External id": 223386,
      "cbid": 211, "correlation": 223386
    }
  },
  {
    "ph": "s", "id": 223386, "pid": 76337, "tid": -914061504, "ts": 1716454224945507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225001087, "dur": 1,
    "args": {
      "External id": 223388,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223388,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 223388, "pid": 5, "tid": 7, "ts": 1716454225001087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945519, "dur": 6,
    "args": {
      "External id": 223388,
      "cbid": 51, "correlation": 223388
    }
  },
  {
    "ph": "s", "id": 223388, "pid": 76337, "tid": -914061504, "ts": 1716454224945519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225001091, "dur": 266,
    "args": {
      "External id": 223389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223389,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223389, "pid": 5, "tid": 7, "ts": 1716454225001091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945526, "dur": 6,
    "args": {
      "External id": 223389,
      "cbid": 211, "correlation": 223389
    }
  },
  {
    "ph": "s", "id": 223389, "pid": 76337, "tid": -914061504, "ts": 1716454224945526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001358, "dur": 6,
    "args": {
      "External id": 223391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223391,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223391, "pid": 5, "tid": 7, "ts": 1716454225001358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945536, "dur": 6,
    "args": {
      "External id": 223391,
      "cbid": 211, "correlation": 223391
    }
  },
  {
    "ph": "s", "id": 223391, "pid": 76337, "tid": -914061504, "ts": 1716454224945536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225001365, "dur": 6,
    "args": {
      "External id": 223397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223397, "pid": 5, "tid": 7, "ts": 1716454225001365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945566, "dur": 8,
    "args": {
      "External id": 223397,
      "cbid": 211, "correlation": 223397
    }
  },
  {
    "ph": "s", "id": 223397, "pid": 76337, "tid": -914061504, "ts": 1716454224945566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225001372, "dur": 3,
    "args": {
      "External id": 223405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223405,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 223405, "pid": 5, "tid": 7, "ts": 1716454225001372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945607, "dur": 10,
    "args": {
      "External id": 223405,
      "cbid": 211, "correlation": 223405
    }
  },
  {
    "ph": "s", "id": 223405, "pid": 76337, "tid": -914061504, "ts": 1716454224945607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945673, "dur": 1,
    "args": {
      "External id": 223421,
      "cbid": 251, "correlation": 223421
    }
  },
  {
    "ph": "f", "id": 223421, "pid": 76337, "tid": -914061504, "ts": 1716454224945673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945678, "dur": 0,
    "args": {
      "External id": 223423,
      "cbid": 251, "correlation": 223423
    }
  },
  {
    "ph": "f", "id": 223423, "pid": 76337, "tid": -914061504, "ts": 1716454224945678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225001377, "dur": 13,
    "args": {
      "External id": 223424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223424,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223424, "pid": 5, "tid": 7, "ts": 1716454225001377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945680, "dur": 11,
    "args": {
      "External id": 223424,
      "cbid": 211, "correlation": 223424
    }
  },
  {
    "ph": "s", "id": 223424, "pid": 76337, "tid": -914061504, "ts": 1716454224945680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225001391, "dur": 5,
    "args": {
      "External id": 223426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223426,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223426, "pid": 5, "tid": 7, "ts": 1716454225001391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945693, "dur": 5,
    "args": {
      "External id": 223426,
      "cbid": 211, "correlation": 223426
    }
  },
  {
    "ph": "s", "id": 223426, "pid": 76337, "tid": -914061504, "ts": 1716454224945693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225001397, "dur": 6,
    "args": {
      "External id": 223436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223436, "pid": 5, "tid": 7, "ts": 1716454225001397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945751, "dur": 13,
    "args": {
      "External id": 223436,
      "cbid": 211, "correlation": 223436
    }
  },
  {
    "ph": "s", "id": 223436, "pid": 76337, "tid": -914061504, "ts": 1716454224945751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001404, "dur": 9,
    "args": {
      "External id": 223456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223456,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223456, "pid": 5, "tid": 7, "ts": 1716454225001404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945818, "dur": 11,
    "args": {
      "External id": 223456,
      "cbid": 211, "correlation": 223456
    }
  },
  {
    "ph": "s", "id": 223456, "pid": 76337, "tid": -914061504, "ts": 1716454224945818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225001415, "dur": 4,
    "args": {
      "External id": 223468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223468,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223468, "pid": 5, "tid": 7, "ts": 1716454225001415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945839, "dur": 6,
    "args": {
      "External id": 223468,
      "cbid": 211, "correlation": 223468
    }
  },
  {
    "ph": "s", "id": 223468, "pid": 76337, "tid": -914061504, "ts": 1716454224945839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225001420, "dur": 7,
    "args": {
      "External id": 223471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223471,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223471, "pid": 5, "tid": 7, "ts": 1716454225001420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945857, "dur": 8,
    "args": {
      "External id": 223471,
      "cbid": 211, "correlation": 223471
    }
  },
  {
    "ph": "s", "id": 223471, "pid": 76337, "tid": -914061504, "ts": 1716454224945857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225001428, "dur": 4,
    "args": {
      "External id": 223480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223480,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223480, "pid": 5, "tid": 7, "ts": 1716454225001428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945898, "dur": 9,
    "args": {
      "External id": 223480,
      "cbid": 211, "correlation": 223480
    }
  },
  {
    "ph": "s", "id": 223480, "pid": 76337, "tid": -914061504, "ts": 1716454224945898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945960, "dur": 0,
    "args": {
      "External id": 223490,
      "cbid": 317, "correlation": 223490
    }
  },
  {
    "ph": "f", "id": 223490, "pid": 76337, "tid": -914061504, "ts": 1716454224945960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945961, "dur": 0,
    "args": {
      "External id": 223491,
      "cbid": 203, "correlation": 223491
    }
  },
  {
    "ph": "f", "id": 223491, "pid": 76337, "tid": -914061504, "ts": 1716454224945961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945961, "dur": 0,
    "args": {
      "External id": 223492,
      "cbid": 205, "correlation": 223492
    }
  },
  {
    "ph": "f", "id": 223492, "pid": 76337, "tid": -914061504, "ts": 1716454224945961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001434, "dur": 5,
    "args": {
      "External id": 223496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223496,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223496, "pid": 5, "tid": 7, "ts": 1716454225001434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945983, "dur": 12,
    "args": {
      "External id": 223496,
      "cbid": 211, "correlation": 223496
    }
  },
  {
    "ph": "s", "id": 223496, "pid": 76337, "tid": -914061504, "ts": 1716454224945983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001440, "dur": 158,
    "args": {
      "External id": 223498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223498,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223498, "pid": 5, "tid": 7, "ts": 1716454225001440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224945998, "dur": 5,
    "args": {
      "External id": 223498,
      "cbid": 211, "correlation": 223498
    }
  },
  {
    "ph": "s", "id": 223498, "pid": 76337, "tid": -914061504, "ts": 1716454224945998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225001600, "dur": 1,
    "args": {
      "External id": 223500,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223500,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 223500, "pid": 5, "tid": 7, "ts": 1716454225001600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946009, "dur": 7,
    "args": {
      "External id": 223500,
      "cbid": 51, "correlation": 223500
    }
  },
  {
    "ph": "s", "id": 223500, "pid": 76337, "tid": -914061504, "ts": 1716454224946009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225001604, "dur": 255,
    "args": {
      "External id": 223501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223501,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223501, "pid": 5, "tid": 7, "ts": 1716454225001604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946017, "dur": 6,
    "args": {
      "External id": 223501,
      "cbid": 211, "correlation": 223501
    }
  },
  {
    "ph": "s", "id": 223501, "pid": 76337, "tid": -914061504, "ts": 1716454224946017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001860, "dur": 6,
    "args": {
      "External id": 223503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223503,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223503, "pid": 5, "tid": 7, "ts": 1716454225001860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946027, "dur": 5,
    "args": {
      "External id": 223503,
      "cbid": 211, "correlation": 223503
    }
  },
  {
    "ph": "s", "id": 223503, "pid": 76337, "tid": -914061504, "ts": 1716454224946027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225001867, "dur": 6,
    "args": {
      "External id": 223509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223509, "pid": 5, "tid": 7, "ts": 1716454225001867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946055, "dur": 9,
    "args": {
      "External id": 223509,
      "cbid": 211, "correlation": 223509
    }
  },
  {
    "ph": "s", "id": 223509, "pid": 76337, "tid": -914061504, "ts": 1716454224946055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225001875, "dur": 5,
    "args": {
      "External id": 223517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223517,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223517, "pid": 5, "tid": 7, "ts": 1716454225001875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946087, "dur": 8,
    "args": {
      "External id": 223517,
      "cbid": 211, "correlation": 223517
    }
  },
  {
    "ph": "s", "id": 223517, "pid": 76337, "tid": -914061504, "ts": 1716454224946087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225001881, "dur": 4,
    "args": {
      "External id": 223525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223525,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223525, "pid": 5, "tid": 7, "ts": 1716454225001881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946117, "dur": 9,
    "args": {
      "External id": 223525,
      "cbid": 211, "correlation": 223525
    }
  },
  {
    "ph": "s", "id": 223525, "pid": 76337, "tid": -914061504, "ts": 1716454224946117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001887, "dur": 9,
    "args": {
      "External id": 223545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223545,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223545, "pid": 5, "tid": 7, "ts": 1716454225001887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946191, "dur": 11,
    "args": {
      "External id": 223545,
      "cbid": 211, "correlation": 223545
    }
  },
  {
    "ph": "s", "id": 223545, "pid": 76337, "tid": -914061504, "ts": 1716454224946191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225001897, "dur": 4,
    "args": {
      "External id": 223557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223557,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223557, "pid": 5, "tid": 7, "ts": 1716454225001897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946212, "dur": 6,
    "args": {
      "External id": 223557,
      "cbid": 211, "correlation": 223557
    }
  },
  {
    "ph": "s", "id": 223557, "pid": 76337, "tid": -914061504, "ts": 1716454224946212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225001902, "dur": 6,
    "args": {
      "External id": 223560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223560, "pid": 5, "tid": 7, "ts": 1716454225001902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946230, "dur": 6,
    "args": {
      "External id": 223560,
      "cbid": 211, "correlation": 223560
    }
  },
  {
    "ph": "s", "id": 223560, "pid": 76337, "tid": -914061504, "ts": 1716454224946230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225001910, "dur": 4,
    "args": {
      "External id": 223569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223569,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223569, "pid": 5, "tid": 7, "ts": 1716454225001910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946267, "dur": 9,
    "args": {
      "External id": 223569,
      "cbid": 211, "correlation": 223569
    }
  },
  {
    "ph": "s", "id": 223569, "pid": 76337, "tid": -914061504, "ts": 1716454224946267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946317, "dur": 0,
    "args": {
      "External id": 223579,
      "cbid": 317, "correlation": 223579
    }
  },
  {
    "ph": "f", "id": 223579, "pid": 76337, "tid": -914061504, "ts": 1716454224946317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946318, "dur": 0,
    "args": {
      "External id": 223580,
      "cbid": 203, "correlation": 223580
    }
  },
  {
    "ph": "f", "id": 223580, "pid": 76337, "tid": -914061504, "ts": 1716454224946318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946318, "dur": 0,
    "args": {
      "External id": 223581,
      "cbid": 205, "correlation": 223581
    }
  },
  {
    "ph": "f", "id": 223581, "pid": 76337, "tid": -914061504, "ts": 1716454224946318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001916, "dur": 5,
    "args": {
      "External id": 223585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223585, "pid": 5, "tid": 7, "ts": 1716454225001916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946332, "dur": 11,
    "args": {
      "External id": 223585,
      "cbid": 211, "correlation": 223585
    }
  },
  {
    "ph": "s", "id": 223585, "pid": 76337, "tid": -914061504, "ts": 1716454224946332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225001922, "dur": 160,
    "args": {
      "External id": 223587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223587, "pid": 5, "tid": 7, "ts": 1716454225001922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946346, "dur": 5,
    "args": {
      "External id": 223587,
      "cbid": 211, "correlation": 223587
    }
  },
  {
    "ph": "s", "id": 223587, "pid": 76337, "tid": -914061504, "ts": 1716454224946346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225002084, "dur": 1,
    "args": {
      "External id": 223589,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223589,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 223589, "pid": 5, "tid": 7, "ts": 1716454225002084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946356, "dur": 6,
    "args": {
      "External id": 223589,
      "cbid": 51, "correlation": 223589
    }
  },
  {
    "ph": "s", "id": 223589, "pid": 76337, "tid": -914061504, "ts": 1716454224946356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225002088, "dur": 254,
    "args": {
      "External id": 223590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223590,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223590, "pid": 5, "tid": 7, "ts": 1716454225002088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946363, "dur": 6,
    "args": {
      "External id": 223590,
      "cbid": 211, "correlation": 223590
    }
  },
  {
    "ph": "s", "id": 223590, "pid": 76337, "tid": -914061504, "ts": 1716454224946363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002343, "dur": 6,
    "args": {
      "External id": 223592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223592, "pid": 5, "tid": 7, "ts": 1716454225002343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946373, "dur": 6,
    "args": {
      "External id": 223592,
      "cbid": 211, "correlation": 223592
    }
  },
  {
    "ph": "s", "id": 223592, "pid": 76337, "tid": -914061504, "ts": 1716454224946373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225002350, "dur": 6,
    "args": {
      "External id": 223598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223598, "pid": 5, "tid": 7, "ts": 1716454225002350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946402, "dur": 8,
    "args": {
      "External id": 223598,
      "cbid": 211, "correlation": 223598
    }
  },
  {
    "ph": "s", "id": 223598, "pid": 76337, "tid": -914061504, "ts": 1716454224946402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225002357, "dur": 3,
    "args": {
      "External id": 223606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223606,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 223606, "pid": 5, "tid": 7, "ts": 1716454225002357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946445, "dur": 8,
    "args": {
      "External id": 223606,
      "cbid": 211, "correlation": 223606
    }
  },
  {
    "ph": "s", "id": 223606, "pid": 76337, "tid": -914061504, "ts": 1716454224946445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946507, "dur": 1,
    "args": {
      "External id": 223622,
      "cbid": 251, "correlation": 223622
    }
  },
  {
    "ph": "f", "id": 223622, "pid": 76337, "tid": -914061504, "ts": 1716454224946507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946512, "dur": 0,
    "args": {
      "External id": 223624,
      "cbid": 251, "correlation": 223624
    }
  },
  {
    "ph": "f", "id": 223624, "pid": 76337, "tid": -914061504, "ts": 1716454224946512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225002362, "dur": 9,
    "args": {
      "External id": 223625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223625,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223625, "pid": 5, "tid": 7, "ts": 1716454225002362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946514, "dur": 11,
    "args": {
      "External id": 223625,
      "cbid": 211, "correlation": 223625
    }
  },
  {
    "ph": "s", "id": 223625, "pid": 76337, "tid": -914061504, "ts": 1716454224946514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225002373, "dur": 3,
    "args": {
      "External id": 223627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223627,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223627, "pid": 5, "tid": 7, "ts": 1716454225002373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946527, "dur": 5,
    "args": {
      "External id": 223627,
      "cbid": 211, "correlation": 223627
    }
  },
  {
    "ph": "s", "id": 223627, "pid": 76337, "tid": -914061504, "ts": 1716454224946527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225002377, "dur": 6,
    "args": {
      "External id": 223637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223637, "pid": 5, "tid": 7, "ts": 1716454225002377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946582, "dur": 12,
    "args": {
      "External id": 223637,
      "cbid": 211, "correlation": 223637
    }
  },
  {
    "ph": "s", "id": 223637, "pid": 76337, "tid": -914061504, "ts": 1716454224946582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002384, "dur": 9,
    "args": {
      "External id": 223657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223657,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223657, "pid": 5, "tid": 7, "ts": 1716454225002384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946646, "dur": 10,
    "args": {
      "External id": 223657,
      "cbid": 211, "correlation": 223657
    }
  },
  {
    "ph": "s", "id": 223657, "pid": 76337, "tid": -914061504, "ts": 1716454224946646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225002395, "dur": 4,
    "args": {
      "External id": 223669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223669,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223669, "pid": 5, "tid": 7, "ts": 1716454225002395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946666, "dur": 6,
    "args": {
      "External id": 223669,
      "cbid": 211, "correlation": 223669
    }
  },
  {
    "ph": "s", "id": 223669, "pid": 76337, "tid": -914061504, "ts": 1716454224946666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225002400, "dur": 6,
    "args": {
      "External id": 223672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223672, "pid": 5, "tid": 7, "ts": 1716454225002400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946685, "dur": 7,
    "args": {
      "External id": 223672,
      "cbid": 211, "correlation": 223672
    }
  },
  {
    "ph": "s", "id": 223672, "pid": 76337, "tid": -914061504, "ts": 1716454224946685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225002407, "dur": 4,
    "args": {
      "External id": 223681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223681,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223681, "pid": 5, "tid": 7, "ts": 1716454225002407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946726, "dur": 9,
    "args": {
      "External id": 223681,
      "cbid": 211, "correlation": 223681
    }
  },
  {
    "ph": "s", "id": 223681, "pid": 76337, "tid": -914061504, "ts": 1716454224946726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946786, "dur": 0,
    "args": {
      "External id": 223691,
      "cbid": 317, "correlation": 223691
    }
  },
  {
    "ph": "f", "id": 223691, "pid": 76337, "tid": -914061504, "ts": 1716454224946786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946787, "dur": 0,
    "args": {
      "External id": 223692,
      "cbid": 203, "correlation": 223692
    }
  },
  {
    "ph": "f", "id": 223692, "pid": 76337, "tid": -914061504, "ts": 1716454224946787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946788, "dur": 0,
    "args": {
      "External id": 223693,
      "cbid": 205, "correlation": 223693
    }
  },
  {
    "ph": "f", "id": 223693, "pid": 76337, "tid": -914061504, "ts": 1716454224946788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002413, "dur": 5,
    "args": {
      "External id": 223697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223697,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223697, "pid": 5, "tid": 7, "ts": 1716454225002413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946802, "dur": 12,
    "args": {
      "External id": 223697,
      "cbid": 211, "correlation": 223697
    }
  },
  {
    "ph": "s", "id": 223697, "pid": 76337, "tid": -914061504, "ts": 1716454224946802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002419, "dur": 160,
    "args": {
      "External id": 223699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223699, "pid": 5, "tid": 7, "ts": 1716454225002419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946816, "dur": 5,
    "args": {
      "External id": 223699,
      "cbid": 211, "correlation": 223699
    }
  },
  {
    "ph": "s", "id": 223699, "pid": 76337, "tid": -914061504, "ts": 1716454224946816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225002581, "dur": 1,
    "args": {
      "External id": 223701,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223701,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 223701, "pid": 5, "tid": 7, "ts": 1716454225002581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946827, "dur": 6,
    "args": {
      "External id": 223701,
      "cbid": 51, "correlation": 223701
    }
  },
  {
    "ph": "s", "id": 223701, "pid": 76337, "tid": -914061504, "ts": 1716454224946827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225002585, "dur": 255,
    "args": {
      "External id": 223702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223702,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223702, "pid": 5, "tid": 7, "ts": 1716454225002585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946834, "dur": 6,
    "args": {
      "External id": 223702,
      "cbid": 211, "correlation": 223702
    }
  },
  {
    "ph": "s", "id": 223702, "pid": 76337, "tid": -914061504, "ts": 1716454224946834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002841, "dur": 6,
    "args": {
      "External id": 223704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223704, "pid": 5, "tid": 7, "ts": 1716454225002841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946843, "dur": 5,
    "args": {
      "External id": 223704,
      "cbid": 211, "correlation": 223704
    }
  },
  {
    "ph": "s", "id": 223704, "pid": 76337, "tid": -914061504, "ts": 1716454224946843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225002848, "dur": 6,
    "args": {
      "External id": 223710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223710, "pid": 5, "tid": 7, "ts": 1716454225002848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946871, "dur": 9,
    "args": {
      "External id": 223710,
      "cbid": 211, "correlation": 223710
    }
  },
  {
    "ph": "s", "id": 223710, "pid": 76337, "tid": -914061504, "ts": 1716454224946871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225002855, "dur": 5,
    "args": {
      "External id": 223718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223718,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223718, "pid": 5, "tid": 7, "ts": 1716454225002855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946903, "dur": 8,
    "args": {
      "External id": 223718,
      "cbid": 211, "correlation": 223718
    }
  },
  {
    "ph": "s", "id": 223718, "pid": 76337, "tid": -914061504, "ts": 1716454224946903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225002861, "dur": 4,
    "args": {
      "External id": 223726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223726,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223726, "pid": 5, "tid": 7, "ts": 1716454225002861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224946933, "dur": 9,
    "args": {
      "External id": 223726,
      "cbid": 211, "correlation": 223726
    }
  },
  {
    "ph": "s", "id": 223726, "pid": 76337, "tid": -914061504, "ts": 1716454224946933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002867, "dur": 9,
    "args": {
      "External id": 223746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223746,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223746, "pid": 5, "tid": 7, "ts": 1716454225002867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947046, "dur": 12,
    "args": {
      "External id": 223746,
      "cbid": 211, "correlation": 223746
    }
  },
  {
    "ph": "s", "id": 223746, "pid": 76337, "tid": -914061504, "ts": 1716454224947046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225002878, "dur": 4,
    "args": {
      "External id": 223758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223758,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223758, "pid": 5, "tid": 7, "ts": 1716454225002878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947068, "dur": 6,
    "args": {
      "External id": 223758,
      "cbid": 211, "correlation": 223758
    }
  },
  {
    "ph": "s", "id": 223758, "pid": 76337, "tid": -914061504, "ts": 1716454224947068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225002883, "dur": 7,
    "args": {
      "External id": 223761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223761,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223761, "pid": 5, "tid": 7, "ts": 1716454225002883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947086, "dur": 7,
    "args": {
      "External id": 223761,
      "cbid": 211, "correlation": 223761
    }
  },
  {
    "ph": "s", "id": 223761, "pid": 76337, "tid": -914061504, "ts": 1716454224947086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225002891, "dur": 4,
    "args": {
      "External id": 223770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223770,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223770, "pid": 5, "tid": 7, "ts": 1716454225002891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947124, "dur": 9,
    "args": {
      "External id": 223770,
      "cbid": 211, "correlation": 223770
    }
  },
  {
    "ph": "s", "id": 223770, "pid": 76337, "tid": -914061504, "ts": 1716454224947124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947176, "dur": 0,
    "args": {
      "External id": 223780,
      "cbid": 317, "correlation": 223780
    }
  },
  {
    "ph": "f", "id": 223780, "pid": 76337, "tid": -914061504, "ts": 1716454224947176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947177, "dur": 0,
    "args": {
      "External id": 223781,
      "cbid": 203, "correlation": 223781
    }
  },
  {
    "ph": "f", "id": 223781, "pid": 76337, "tid": -914061504, "ts": 1716454224947177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947177, "dur": 0,
    "args": {
      "External id": 223782,
      "cbid": 205, "correlation": 223782
    }
  },
  {
    "ph": "f", "id": 223782, "pid": 76337, "tid": -914061504, "ts": 1716454224947177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002897, "dur": 5,
    "args": {
      "External id": 223786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223786,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223786, "pid": 5, "tid": 7, "ts": 1716454225002897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947190, "dur": 11,
    "args": {
      "External id": 223786,
      "cbid": 211, "correlation": 223786
    }
  },
  {
    "ph": "s", "id": 223786, "pid": 76337, "tid": -914061504, "ts": 1716454224947190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225002903, "dur": 159,
    "args": {
      "External id": 223788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223788, "pid": 5, "tid": 7, "ts": 1716454225002903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947204, "dur": 5,
    "args": {
      "External id": 223788,
      "cbid": 211, "correlation": 223788
    }
  },
  {
    "ph": "s", "id": 223788, "pid": 76337, "tid": -914061504, "ts": 1716454224947204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225003064, "dur": 1,
    "args": {
      "External id": 223790,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223790,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 223790, "pid": 5, "tid": 7, "ts": 1716454225003064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947214, "dur": 6,
    "args": {
      "External id": 223790,
      "cbid": 51, "correlation": 223790
    }
  },
  {
    "ph": "s", "id": 223790, "pid": 76337, "tid": -914061504, "ts": 1716454224947214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225003068, "dur": 254,
    "args": {
      "External id": 223791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223791,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223791, "pid": 5, "tid": 7, "ts": 1716454225003068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947221, "dur": 6,
    "args": {
      "External id": 223791,
      "cbid": 211, "correlation": 223791
    }
  },
  {
    "ph": "s", "id": 223791, "pid": 76337, "tid": -914061504, "ts": 1716454224947221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003323, "dur": 5,
    "args": {
      "External id": 223793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223793,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223793, "pid": 5, "tid": 7, "ts": 1716454225003323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947233, "dur": 6,
    "args": {
      "External id": 223793,
      "cbid": 211, "correlation": 223793
    }
  },
  {
    "ph": "s", "id": 223793, "pid": 76337, "tid": -914061504, "ts": 1716454224947233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225003330, "dur": 6,
    "args": {
      "External id": 223799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223799, "pid": 5, "tid": 7, "ts": 1716454225003330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947261, "dur": 8,
    "args": {
      "External id": 223799,
      "cbid": 211, "correlation": 223799
    }
  },
  {
    "ph": "s", "id": 223799, "pid": 76337, "tid": -914061504, "ts": 1716454224947261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225003337, "dur": 3,
    "args": {
      "External id": 223807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223807,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 223807, "pid": 5, "tid": 7, "ts": 1716454225003337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947304, "dur": 9,
    "args": {
      "External id": 223807,
      "cbid": 211, "correlation": 223807
    }
  },
  {
    "ph": "s", "id": 223807, "pid": 76337, "tid": -914061504, "ts": 1716454224947304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947365, "dur": 1,
    "args": {
      "External id": 223823,
      "cbid": 251, "correlation": 223823
    }
  },
  {
    "ph": "f", "id": 223823, "pid": 76337, "tid": -914061504, "ts": 1716454224947365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947371, "dur": 0,
    "args": {
      "External id": 223825,
      "cbid": 251, "correlation": 223825
    }
  },
  {
    "ph": "f", "id": 223825, "pid": 76337, "tid": -914061504, "ts": 1716454224947371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225003342, "dur": 10,
    "args": {
      "External id": 223826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223826,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223826, "pid": 5, "tid": 7, "ts": 1716454225003342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947373, "dur": 11,
    "args": {
      "External id": 223826,
      "cbid": 211, "correlation": 223826
    }
  },
  {
    "ph": "s", "id": 223826, "pid": 76337, "tid": -914061504, "ts": 1716454224947373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225003353, "dur": 3,
    "args": {
      "External id": 223828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223828,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223828, "pid": 5, "tid": 7, "ts": 1716454225003353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947386, "dur": 6,
    "args": {
      "External id": 223828,
      "cbid": 211, "correlation": 223828
    }
  },
  {
    "ph": "s", "id": 223828, "pid": 76337, "tid": -914061504, "ts": 1716454224947386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225003358, "dur": 6,
    "args": {
      "External id": 223838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223838,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223838, "pid": 5, "tid": 7, "ts": 1716454225003358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947442, "dur": 12,
    "args": {
      "External id": 223838,
      "cbid": 211, "correlation": 223838
    }
  },
  {
    "ph": "s", "id": 223838, "pid": 76337, "tid": -914061504, "ts": 1716454224947442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003365, "dur": 9,
    "args": {
      "External id": 223858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223858,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223858, "pid": 5, "tid": 7, "ts": 1716454225003365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947508, "dur": 10,
    "args": {
      "External id": 223858,
      "cbid": 211, "correlation": 223858
    }
  },
  {
    "ph": "s", "id": 223858, "pid": 76337, "tid": -914061504, "ts": 1716454224947508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225003376, "dur": 4,
    "args": {
      "External id": 223870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223870,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223870, "pid": 5, "tid": 7, "ts": 1716454225003376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947529, "dur": 6,
    "args": {
      "External id": 223870,
      "cbid": 211, "correlation": 223870
    }
  },
  {
    "ph": "s", "id": 223870, "pid": 76337, "tid": -914061504, "ts": 1716454224947529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225003381, "dur": 6,
    "args": {
      "External id": 223873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223873,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223873, "pid": 5, "tid": 7, "ts": 1716454225003381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947547, "dur": 7,
    "args": {
      "External id": 223873,
      "cbid": 211, "correlation": 223873
    }
  },
  {
    "ph": "s", "id": 223873, "pid": 76337, "tid": -914061504, "ts": 1716454224947547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225003388, "dur": 4,
    "args": {
      "External id": 223882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223882,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223882, "pid": 5, "tid": 7, "ts": 1716454225003388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947587, "dur": 9,
    "args": {
      "External id": 223882,
      "cbid": 211, "correlation": 223882
    }
  },
  {
    "ph": "s", "id": 223882, "pid": 76337, "tid": -914061504, "ts": 1716454224947587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947650, "dur": 0,
    "args": {
      "External id": 223892,
      "cbid": 317, "correlation": 223892
    }
  },
  {
    "ph": "f", "id": 223892, "pid": 76337, "tid": -914061504, "ts": 1716454224947650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947651, "dur": 0,
    "args": {
      "External id": 223893,
      "cbid": 203, "correlation": 223893
    }
  },
  {
    "ph": "f", "id": 223893, "pid": 76337, "tid": -914061504, "ts": 1716454224947651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947652, "dur": 0,
    "args": {
      "External id": 223894,
      "cbid": 205, "correlation": 223894
    }
  },
  {
    "ph": "f", "id": 223894, "pid": 76337, "tid": -914061504, "ts": 1716454224947652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003394, "dur": 5,
    "args": {
      "External id": 223898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223898,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223898, "pid": 5, "tid": 7, "ts": 1716454225003394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947665, "dur": 13,
    "args": {
      "External id": 223898,
      "cbid": 211, "correlation": 223898
    }
  },
  {
    "ph": "s", "id": 223898, "pid": 76337, "tid": -914061504, "ts": 1716454224947665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003400, "dur": 159,
    "args": {
      "External id": 223900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223900,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223900, "pid": 5, "tid": 7, "ts": 1716454225003400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947681, "dur": 5,
    "args": {
      "External id": 223900,
      "cbid": 211, "correlation": 223900
    }
  },
  {
    "ph": "s", "id": 223900, "pid": 76337, "tid": -914061504, "ts": 1716454224947681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225003562, "dur": 1,
    "args": {
      "External id": 223902,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 223902,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 223902, "pid": 5, "tid": 7, "ts": 1716454225003562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947691, "dur": 6,
    "args": {
      "External id": 223902,
      "cbid": 51, "correlation": 223902
    }
  },
  {
    "ph": "s", "id": 223902, "pid": 76337, "tid": -914061504, "ts": 1716454224947691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225003565, "dur": 254,
    "args": {
      "External id": 223903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223903,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 223903, "pid": 5, "tid": 7, "ts": 1716454225003565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947699, "dur": 6,
    "args": {
      "External id": 223903,
      "cbid": 211, "correlation": 223903
    }
  },
  {
    "ph": "s", "id": 223903, "pid": 76337, "tid": -914061504, "ts": 1716454224947699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003821, "dur": 6,
    "args": {
      "External id": 223905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223905,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223905, "pid": 5, "tid": 7, "ts": 1716454225003821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947708, "dur": 5,
    "args": {
      "External id": 223905,
      "cbid": 211, "correlation": 223905
    }
  },
  {
    "ph": "s", "id": 223905, "pid": 76337, "tid": -914061504, "ts": 1716454224947708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225003828, "dur": 6,
    "args": {
      "External id": 223911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223911,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223911, "pid": 5, "tid": 7, "ts": 1716454225003828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947736, "dur": 8,
    "args": {
      "External id": 223911,
      "cbid": 211, "correlation": 223911
    }
  },
  {
    "ph": "s", "id": 223911, "pid": 76337, "tid": -914061504, "ts": 1716454224947736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225003835, "dur": 5,
    "args": {
      "External id": 223919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223919,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223919, "pid": 5, "tid": 7, "ts": 1716454225003835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947768, "dur": 9,
    "args": {
      "External id": 223919,
      "cbid": 211, "correlation": 223919
    }
  },
  {
    "ph": "s", "id": 223919, "pid": 76337, "tid": -914061504, "ts": 1716454224947768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225003841, "dur": 4,
    "args": {
      "External id": 223927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223927,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223927, "pid": 5, "tid": 7, "ts": 1716454225003841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947798, "dur": 9,
    "args": {
      "External id": 223927,
      "cbid": 211, "correlation": 223927
    }
  },
  {
    "ph": "s", "id": 223927, "pid": 76337, "tid": -914061504, "ts": 1716454224947798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003847, "dur": 9,
    "args": {
      "External id": 223947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223947,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 223947, "pid": 5, "tid": 7, "ts": 1716454225003847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947938, "dur": 14,
    "args": {
      "External id": 223947,
      "cbid": 211, "correlation": 223947
    }
  },
  {
    "ph": "s", "id": 223947, "pid": 76337, "tid": -914061504, "ts": 1716454224947938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225003858, "dur": 4,
    "args": {
      "External id": 223959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223959,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 223959, "pid": 5, "tid": 7, "ts": 1716454225003858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947962, "dur": 6,
    "args": {
      "External id": 223959,
      "cbid": 211, "correlation": 223959
    }
  },
  {
    "ph": "s", "id": 223959, "pid": 76337, "tid": -914061504, "ts": 1716454224947962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225003863, "dur": 6,
    "args": {
      "External id": 223962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223962, "pid": 5, "tid": 7, "ts": 1716454225003863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224947990, "dur": 7,
    "args": {
      "External id": 223962,
      "cbid": 211, "correlation": 223962
    }
  },
  {
    "ph": "s", "id": 223962, "pid": 76337, "tid": -914061504, "ts": 1716454224947990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948050, "dur": 0,
    "args": {
      "External id": 223973,
      "cbid": 317, "correlation": 223973
    }
  },
  {
    "ph": "f", "id": 223973, "pid": 76337, "tid": -914061504, "ts": 1716454224948050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948051, "dur": 0,
    "args": {
      "External id": 223974,
      "cbid": 203, "correlation": 223974
    }
  },
  {
    "ph": "f", "id": 223974, "pid": 76337, "tid": -914061504, "ts": 1716454224948051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948051, "dur": 0,
    "args": {
      "External id": 223975,
      "cbid": 205, "correlation": 223975
    }
  },
  {
    "ph": "f", "id": 223975, "pid": 76337, "tid": -914061504, "ts": 1716454224948051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003871, "dur": 5,
    "args": {
      "External id": 223979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223979,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223979, "pid": 5, "tid": 7, "ts": 1716454225003871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948068, "dur": 12,
    "args": {
      "External id": 223979,
      "cbid": 211, "correlation": 223979
    }
  },
  {
    "ph": "s", "id": 223979, "pid": 76337, "tid": -914061504, "ts": 1716454224948068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225003877, "dur": 36,
    "args": {
      "External id": 223981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223981,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 223981, "pid": 5, "tid": 7, "ts": 1716454225003877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948087, "dur": 8,
    "args": {
      "External id": 223981,
      "cbid": 211, "correlation": 223981
    }
  },
  {
    "ph": "s", "id": 223981, "pid": 76337, "tid": -914061504, "ts": 1716454224948087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003915, "dur": 5,
    "args": {
      "External id": 223983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223983, "pid": 5, "tid": 7, "ts": 1716454225003915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948099, "dur": 5,
    "args": {
      "External id": 223983,
      "cbid": 211, "correlation": 223983
    }
  },
  {
    "ph": "s", "id": 223983, "pid": 76337, "tid": -914061504, "ts": 1716454224948099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225003921, "dur": 6,
    "args": {
      "External id": 223989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 223989, "pid": 5, "tid": 7, "ts": 1716454225003921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948126, "dur": 8,
    "args": {
      "External id": 223989,
      "cbid": 211, "correlation": 223989
    }
  },
  {
    "ph": "s", "id": 223989, "pid": 76337, "tid": -914061504, "ts": 1716454224948126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225003928, "dur": 20,
    "args": {
      "External id": 223998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 223998,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 223998, "pid": 5, "tid": 7, "ts": 1716454225003928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948209, "dur": 14,
    "args": {
      "External id": 223998,
      "cbid": 211, "correlation": 223998
    }
  },
  {
    "ph": "s", "id": 223998, "pid": 76337, "tid": -914061504, "ts": 1716454224948209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225003949, "dur": 10,
    "args": {
      "External id": 224020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224020,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 224020, "pid": 5, "tid": 7, "ts": 1716454225003949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948266, "dur": 11,
    "args": {
      "External id": 224020,
      "cbid": 211, "correlation": 224020
    }
  },
  {
    "ph": "s", "id": 224020, "pid": 76337, "tid": -914061504, "ts": 1716454224948266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948355, "dur": 2,
    "args": {
      "External id": 224031,
      "cbid": 251, "correlation": 224031
    }
  },
  {
    "ph": "f", "id": 224031, "pid": 76337, "tid": -914061504, "ts": 1716454224948355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948360, "dur": 0,
    "args": {
      "External id": 224032,
      "cbid": 251, "correlation": 224032
    }
  },
  {
    "ph": "f", "id": 224032, "pid": 76337, "tid": -914061504, "ts": 1716454224948360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225003961, "dur": 53,
    "args": {
      "External id": 224033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224033,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 224033, "pid": 5, "tid": 7, "ts": 1716454225003961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948363, "dur": 15,
    "args": {
      "External id": 224033,
      "cbid": 211, "correlation": 224033
    }
  },
  {
    "ph": "s", "id": 224033, "pid": 76337, "tid": -914061504, "ts": 1716454224948363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948435, "dur": 1,
    "args": {
      "External id": 224044,
      "cbid": 251, "correlation": 224044
    }
  },
  {
    "ph": "f", "id": 224044, "pid": 76337, "tid": -914061504, "ts": 1716454224948435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948438, "dur": 0,
    "args": {
      "External id": 224045,
      "cbid": 251, "correlation": 224045
    }
  },
  {
    "ph": "f", "id": 224045, "pid": 76337, "tid": -914061504, "ts": 1716454224948438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225004015, "dur": 51,
    "args": {
      "External id": 224046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224046,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 224046, "pid": 5, "tid": 7, "ts": 1716454225004015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948440, "dur": 12,
    "args": {
      "External id": 224046,
      "cbid": 211, "correlation": 224046
    }
  },
  {
    "ph": "s", "id": 224046, "pid": 76337, "tid": -914061504, "ts": 1716454224948440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948506, "dur": 1,
    "args": {
      "External id": 224057,
      "cbid": 251, "correlation": 224057
    }
  },
  {
    "ph": "f", "id": 224057, "pid": 76337, "tid": -914061504, "ts": 1716454224948506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948510, "dur": 0,
    "args": {
      "External id": 224058,
      "cbid": 251, "correlation": 224058
    }
  },
  {
    "ph": "f", "id": 224058, "pid": 76337, "tid": -914061504, "ts": 1716454224948510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225004067, "dur": 52,
    "args": {
      "External id": 224059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224059,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 224059, "pid": 5, "tid": 7, "ts": 1716454225004067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948512, "dur": 12,
    "args": {
      "External id": 224059,
      "cbid": 211, "correlation": 224059
    }
  },
  {
    "ph": "s", "id": 224059, "pid": 76337, "tid": -914061504, "ts": 1716454224948512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225004121, "dur": 56,
    "args": {
      "External id": 224084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224084,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224084, "pid": 5, "tid": 7, "ts": 1716454225004121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948597, "dur": 13,
    "args": {
      "External id": 224084,
      "cbid": 211, "correlation": 224084
    }
  },
  {
    "ph": "s", "id": 224084, "pid": 76337, "tid": -914061504, "ts": 1716454224948597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948697, "dur": 1,
    "args": {
      "External id": 224102,
      "cbid": 251, "correlation": 224102
    }
  },
  {
    "ph": "f", "id": 224102, "pid": 76337, "tid": -914061504, "ts": 1716454224948697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225004178, "dur": 61,
    "args": {
      "External id": 224104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224104,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 224104, "pid": 5, "tid": 7, "ts": 1716454225004178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948702, "dur": 13,
    "args": {
      "External id": 224104,
      "cbid": 211, "correlation": 224104
    }
  },
  {
    "ph": "s", "id": 224104, "pid": 76337, "tid": -914061504, "ts": 1716454224948702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225004241, "dur": 6,
    "args": {
      "External id": 224112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224112,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224112, "pid": 5, "tid": 7, "ts": 1716454225004241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948772, "dur": 13,
    "args": {
      "External id": 224112,
      "cbid": 211, "correlation": 224112
    }
  },
  {
    "ph": "s", "id": 224112, "pid": 76337, "tid": -914061504, "ts": 1716454224948772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225004248, "dur": 7,
    "args": {
      "External id": 224120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224120,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224120, "pid": 5, "tid": 7, "ts": 1716454225004248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948811, "dur": 9,
    "args": {
      "External id": 224120,
      "cbid": 211, "correlation": 224120
    }
  },
  {
    "ph": "s", "id": 224120, "pid": 76337, "tid": -914061504, "ts": 1716454224948811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004257, "dur": 8,
    "args": {
      "External id": 224131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224131, "pid": 5, "tid": 7, "ts": 1716454225004257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948885, "dur": 13,
    "args": {
      "External id": 224131,
      "cbid": 211, "correlation": 224131
    }
  },
  {
    "ph": "s", "id": 224131, "pid": 76337, "tid": -914061504, "ts": 1716454224948885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004266, "dur": 9,
    "args": {
      "External id": 224153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224153,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 224153, "pid": 5, "tid": 7, "ts": 1716454225004266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224948917, "dur": 9,
    "args": {
      "External id": 224153,
      "cbid": 211, "correlation": 224153
    }
  },
  {
    "ph": "s", "id": 224153, "pid": 76337, "tid": -914061504, "ts": 1716454224948917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949011, "dur": 3,
    "args": {
      "External id": 224164,
      "cbid": 251, "correlation": 224164
    }
  },
  {
    "ph": "f", "id": 224164, "pid": 76337, "tid": -914061504, "ts": 1716454224949011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225004277, "dur": 1,
    "args": {
      "External id": 224165,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224165,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 224165, "pid": 5, "tid": 7, "ts": 1716454225004277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949017, "dur": 11,
    "args": {
      "External id": 224165,
      "cbid": 51, "correlation": 224165
    }
  },
  {
    "ph": "s", "id": 224165, "pid": 76337, "tid": -914061504, "ts": 1716454224949017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225004280, "dur": 36,
    "args": {
      "External id": 224166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224166,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 224166, "pid": 5, "tid": 7, "ts": 1716454225004280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949031, "dur": 13,
    "args": {
      "External id": 224166,
      "cbid": 211, "correlation": 224166
    }
  },
  {
    "ph": "s", "id": 224166, "pid": 76337, "tid": -914061504, "ts": 1716454224949031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949104, "dur": 1,
    "args": {
      "External id": 224177,
      "cbid": 251, "correlation": 224177
    }
  },
  {
    "ph": "f", "id": 224177, "pid": 76337, "tid": -914061504, "ts": 1716454224949104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949108, "dur": 0,
    "args": {
      "External id": 224178,
      "cbid": 251, "correlation": 224178
    }
  },
  {
    "ph": "f", "id": 224178, "pid": 76337, "tid": -914061504, "ts": 1716454224949108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225004317, "dur": 12,
    "args": {
      "External id": 224179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224179,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224179, "pid": 5, "tid": 7, "ts": 1716454225004317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949109, "dur": 12,
    "args": {
      "External id": 224179,
      "cbid": 211, "correlation": 224179
    }
  },
  {
    "ph": "s", "id": 224179, "pid": 76337, "tid": -914061504, "ts": 1716454224949109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225004330, "dur": 5,
    "args": {
      "External id": 224181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224181,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224181, "pid": 5, "tid": 7, "ts": 1716454225004330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949123, "dur": 7,
    "args": {
      "External id": 224181,
      "cbid": 211, "correlation": 224181
    }
  },
  {
    "ph": "s", "id": 224181, "pid": 76337, "tid": -914061504, "ts": 1716454224949123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949182, "dur": 1,
    "args": {
      "External id": 224192,
      "cbid": 251, "correlation": 224192
    }
  },
  {
    "ph": "f", "id": 224192, "pid": 76337, "tid": -914061504, "ts": 1716454224949182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949185, "dur": 0,
    "args": {
      "External id": 224193,
      "cbid": 251, "correlation": 224193
    }
  },
  {
    "ph": "f", "id": 224193, "pid": 76337, "tid": -914061504, "ts": 1716454224949185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225004336, "dur": 8,
    "args": {
      "External id": 224194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224194,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224194, "pid": 5, "tid": 7, "ts": 1716454225004336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949186, "dur": 11,
    "args": {
      "External id": 224194,
      "cbid": 211, "correlation": 224194
    }
  },
  {
    "ph": "s", "id": 224194, "pid": 76337, "tid": -914061504, "ts": 1716454224949186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225004345, "dur": 4,
    "args": {
      "External id": 224196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224196,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224196, "pid": 5, "tid": 7, "ts": 1716454225004345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949199, "dur": 6,
    "args": {
      "External id": 224196,
      "cbid": 211, "correlation": 224196
    }
  },
  {
    "ph": "s", "id": 224196, "pid": 76337, "tid": -914061504, "ts": 1716454224949199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225004350, "dur": 19,
    "args": {
      "External id": 224221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224221,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 224221, "pid": 5, "tid": 7, "ts": 1716454225004350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949277, "dur": 13,
    "args": {
      "External id": 224221,
      "cbid": 211, "correlation": 224221
    }
  },
  {
    "ph": "s", "id": 224221, "pid": 76337, "tid": -914061504, "ts": 1716454224949277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949377, "dur": 2,
    "args": {
      "External id": 224239,
      "cbid": 251, "correlation": 224239
    }
  },
  {
    "ph": "f", "id": 224239, "pid": 76337, "tid": -914061504, "ts": 1716454224949377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225004372, "dur": 1,
    "args": {
      "External id": 224241,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224241,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 224241, "pid": 5, "tid": 7, "ts": 1716454225004372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949384, "dur": 10,
    "args": {
      "External id": 224241,
      "cbid": 51, "correlation": 224241
    }
  },
  {
    "ph": "s", "id": 224241, "pid": 76337, "tid": -914061504, "ts": 1716454224949384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225004375, "dur": 36,
    "args": {
      "External id": 224242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224242,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 224242, "pid": 5, "tid": 7, "ts": 1716454225004375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949395, "dur": 13,
    "args": {
      "External id": 224242,
      "cbid": 211, "correlation": 224242
    }
  },
  {
    "ph": "s", "id": 224242, "pid": 76337, "tid": -914061504, "ts": 1716454224949395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225004413, "dur": 4,
    "args": {
      "External id": 224250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224250,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224250, "pid": 5, "tid": 7, "ts": 1716454225004413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949466, "dur": 12,
    "args": {
      "External id": 224250,
      "cbid": 211, "correlation": 224250
    }
  },
  {
    "ph": "s", "id": 224250, "pid": 76337, "tid": -914061504, "ts": 1716454224949466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004418, "dur": 8,
    "args": {
      "External id": 224258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224258, "pid": 5, "tid": 7, "ts": 1716454225004418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949508, "dur": 10,
    "args": {
      "External id": 224258,
      "cbid": 211, "correlation": 224258
    }
  },
  {
    "ph": "s", "id": 224258, "pid": 76337, "tid": -914061504, "ts": 1716454224949508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004428, "dur": 8,
    "args": {
      "External id": 224280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224280,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 224280, "pid": 5, "tid": 7, "ts": 1716454225004428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949561, "dur": 10,
    "args": {
      "External id": 224280,
      "cbid": 211, "correlation": 224280
    }
  },
  {
    "ph": "s", "id": 224280, "pid": 76337, "tid": -914061504, "ts": 1716454224949561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949654, "dur": 1,
    "args": {
      "External id": 224296,
      "cbid": 251, "correlation": 224296
    }
  },
  {
    "ph": "f", "id": 224296, "pid": 76337, "tid": -914061504, "ts": 1716454224949654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949659, "dur": 0,
    "args": {
      "External id": 224298,
      "cbid": 251, "correlation": 224298
    }
  },
  {
    "ph": "f", "id": 224298, "pid": 76337, "tid": -914061504, "ts": 1716454224949659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225004437, "dur": 188,
    "args": {
      "External id": 224299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224299,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224299, "pid": 5, "tid": 7, "ts": 1716454225004437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949661, "dur": 13,
    "args": {
      "External id": 224299,
      "cbid": 211, "correlation": 224299
    }
  },
  {
    "ph": "s", "id": 224299, "pid": 76337, "tid": -914061504, "ts": 1716454224949661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004627, "dur": 21,
    "args": {
      "External id": 224307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224307,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224307, "pid": 5, "tid": 7, "ts": 1716454225004627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949728, "dur": 14,
    "args": {
      "External id": 224307,
      "cbid": 211, "correlation": 224307
    }
  },
  {
    "ph": "s", "id": 224307, "pid": 76337, "tid": -914061504, "ts": 1716454224949728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004649, "dur": 21,
    "args": {
      "External id": 224315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224315, "pid": 5, "tid": 7, "ts": 1716454225004649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949760, "dur": 8,
    "args": {
      "External id": 224315,
      "cbid": 211, "correlation": 224315
    }
  },
  {
    "ph": "s", "id": 224315, "pid": 76337, "tid": -914061504, "ts": 1716454224949760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949842, "dur": 1,
    "args": {
      "External id": 224331,
      "cbid": 251, "correlation": 224331
    }
  },
  {
    "ph": "f", "id": 224331, "pid": 76337, "tid": -914061504, "ts": 1716454224949842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225004673, "dur": 1,
    "args": {
      "External id": 224333,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224333,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 224333, "pid": 5, "tid": 7, "ts": 1716454225004673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949847, "dur": 9,
    "args": {
      "External id": 224333,
      "cbid": 51, "correlation": 224333
    }
  },
  {
    "ph": "s", "id": 224333, "pid": 76337, "tid": -914061504, "ts": 1716454224949847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225004676, "dur": 109,
    "args": {
      "External id": 224334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224334,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 224334, "pid": 5, "tid": 7, "ts": 1716454225004676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949857, "dur": 12,
    "args": {
      "External id": 224334,
      "cbid": 211, "correlation": 224334
    }
  },
  {
    "ph": "s", "id": 224334, "pid": 76337, "tid": -914061504, "ts": 1716454224949857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225004787, "dur": 5,
    "args": {
      "External id": 224342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224342,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224342, "pid": 5, "tid": 7, "ts": 1716454225004787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949901, "dur": 9,
    "args": {
      "External id": 224342,
      "cbid": 211, "correlation": 224342
    }
  },
  {
    "ph": "s", "id": 224342, "pid": 76337, "tid": -914061504, "ts": 1716454224949901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004793, "dur": 9,
    "args": {
      "External id": 224353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224353, "pid": 5, "tid": 7, "ts": 1716454225004793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224949969, "dur": 22,
    "args": {
      "External id": 224353,
      "cbid": 211, "correlation": 224353
    }
  },
  {
    "ph": "s", "id": 224353, "pid": 76337, "tid": -914061504, "ts": 1716454224949969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950046, "dur": 0,
    "args": {
      "External id": 224365,
      "cbid": 317, "correlation": 224365
    }
  },
  {
    "ph": "f", "id": 224365, "pid": 76337, "tid": -914061504, "ts": 1716454224950046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950046, "dur": 0,
    "args": {
      "External id": 224366,
      "cbid": 203, "correlation": 224366
    }
  },
  {
    "ph": "f", "id": 224366, "pid": 76337, "tid": -914061504, "ts": 1716454224950046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950047, "dur": 0,
    "args": {
      "External id": 224367,
      "cbid": 205, "correlation": 224367
    }
  },
  {
    "ph": "f", "id": 224367, "pid": 76337, "tid": -914061504, "ts": 1716454224950047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004804, "dur": 5,
    "args": {
      "External id": 224371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224371, "pid": 5, "tid": 7, "ts": 1716454225004804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950063, "dur": 12,
    "args": {
      "External id": 224371,
      "cbid": 211, "correlation": 224371
    }
  },
  {
    "ph": "s", "id": 224371, "pid": 76337, "tid": -914061504, "ts": 1716454224950063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225004811, "dur": 37,
    "args": {
      "External id": 224373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224373,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 224373, "pid": 5, "tid": 7, "ts": 1716454225004811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950082, "dur": 7,
    "args": {
      "External id": 224373,
      "cbid": 211, "correlation": 224373
    }
  },
  {
    "ph": "s", "id": 224373, "pid": 76337, "tid": -914061504, "ts": 1716454224950082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004849, "dur": 6,
    "args": {
      "External id": 224375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224375,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224375, "pid": 5, "tid": 7, "ts": 1716454225004849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950093, "dur": 5,
    "args": {
      "External id": 224375,
      "cbid": 211, "correlation": 224375
    }
  },
  {
    "ph": "s", "id": 224375, "pid": 76337, "tid": -914061504, "ts": 1716454224950093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004856, "dur": 7,
    "args": {
      "External id": 224381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224381, "pid": 5, "tid": 7, "ts": 1716454225004856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950120, "dur": 9,
    "args": {
      "External id": 224381,
      "cbid": 211, "correlation": 224381
    }
  },
  {
    "ph": "s", "id": 224381, "pid": 76337, "tid": -914061504, "ts": 1716454224950120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225004864, "dur": 5,
    "args": {
      "External id": 224389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224389,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224389, "pid": 5, "tid": 7, "ts": 1716454225004864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950152, "dur": 9,
    "args": {
      "External id": 224389,
      "cbid": 211, "correlation": 224389
    }
  },
  {
    "ph": "s", "id": 224389, "pid": 76337, "tid": -914061504, "ts": 1716454224950152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004870, "dur": 11,
    "args": {
      "External id": 224409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224409,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 224409, "pid": 5, "tid": 7, "ts": 1716454225004870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950225, "dur": 12,
    "args": {
      "External id": 224409,
      "cbid": 211, "correlation": 224409
    }
  },
  {
    "ph": "s", "id": 224409, "pid": 76337, "tid": -914061504, "ts": 1716454224950225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225004882, "dur": 4,
    "args": {
      "External id": 224421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224421,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 224421, "pid": 5, "tid": 7, "ts": 1716454225004882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950247, "dur": 6,
    "args": {
      "External id": 224421,
      "cbid": 211, "correlation": 224421
    }
  },
  {
    "ph": "s", "id": 224421, "pid": 76337, "tid": -914061504, "ts": 1716454224950247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225004888, "dur": 8,
    "args": {
      "External id": 224424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224424,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224424, "pid": 5, "tid": 7, "ts": 1716454225004888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950265, "dur": 7,
    "args": {
      "External id": 224424,
      "cbid": 211, "correlation": 224424
    }
  },
  {
    "ph": "s", "id": 224424, "pid": 76337, "tid": -914061504, "ts": 1716454224950265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225004897, "dur": 5,
    "args": {
      "External id": 224433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224433,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224433, "pid": 5, "tid": 7, "ts": 1716454225004897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950305, "dur": 10,
    "args": {
      "External id": 224433,
      "cbid": 211, "correlation": 224433
    }
  },
  {
    "ph": "s", "id": 224433, "pid": 76337, "tid": -914061504, "ts": 1716454224950305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950355, "dur": 0,
    "args": {
      "External id": 224443,
      "cbid": 317, "correlation": 224443
    }
  },
  {
    "ph": "f", "id": 224443, "pid": 76337, "tid": -914061504, "ts": 1716454224950355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950356, "dur": 0,
    "args": {
      "External id": 224444,
      "cbid": 203, "correlation": 224444
    }
  },
  {
    "ph": "f", "id": 224444, "pid": 76337, "tid": -914061504, "ts": 1716454224950356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950357, "dur": 0,
    "args": {
      "External id": 224445,
      "cbid": 205, "correlation": 224445
    }
  },
  {
    "ph": "f", "id": 224445, "pid": 76337, "tid": -914061504, "ts": 1716454224950357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004904, "dur": 5,
    "args": {
      "External id": 224449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224449, "pid": 5, "tid": 7, "ts": 1716454225004904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950370, "dur": 12,
    "args": {
      "External id": 224449,
      "cbid": 211, "correlation": 224449
    }
  },
  {
    "ph": "s", "id": 224449, "pid": 76337, "tid": -914061504, "ts": 1716454224950370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225004910, "dur": 159,
    "args": {
      "External id": 224451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224451, "pid": 5, "tid": 7, "ts": 1716454225004910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950384, "dur": 6,
    "args": {
      "External id": 224451,
      "cbid": 211, "correlation": 224451
    }
  },
  {
    "ph": "s", "id": 224451, "pid": 76337, "tid": -914061504, "ts": 1716454224950384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225005072, "dur": 1,
    "args": {
      "External id": 224453,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224453,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 224453, "pid": 5, "tid": 7, "ts": 1716454225005072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950396, "dur": 6,
    "args": {
      "External id": 224453,
      "cbid": 51, "correlation": 224453
    }
  },
  {
    "ph": "s", "id": 224453, "pid": 76337, "tid": -914061504, "ts": 1716454224950396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225005075, "dur": 265,
    "args": {
      "External id": 224454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224454,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224454, "pid": 5, "tid": 7, "ts": 1716454225005075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950403, "dur": 6,
    "args": {
      "External id": 224454,
      "cbid": 211, "correlation": 224454
    }
  },
  {
    "ph": "s", "id": 224454, "pid": 76337, "tid": -914061504, "ts": 1716454224950403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005342, "dur": 5,
    "args": {
      "External id": 224456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224456, "pid": 5, "tid": 7, "ts": 1716454225005342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950413, "dur": 5,
    "args": {
      "External id": 224456,
      "cbid": 211, "correlation": 224456
    }
  },
  {
    "ph": "s", "id": 224456, "pid": 76337, "tid": -914061504, "ts": 1716454224950413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225005349, "dur": 6,
    "args": {
      "External id": 224462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224462, "pid": 5, "tid": 7, "ts": 1716454225005349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950441, "dur": 8,
    "args": {
      "External id": 224462,
      "cbid": 211, "correlation": 224462
    }
  },
  {
    "ph": "s", "id": 224462, "pid": 76337, "tid": -914061504, "ts": 1716454224950441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225005356, "dur": 3,
    "args": {
      "External id": 224470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224470,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 224470, "pid": 5, "tid": 7, "ts": 1716454225005356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950484, "dur": 10,
    "args": {
      "External id": 224470,
      "cbid": 211, "correlation": 224470
    }
  },
  {
    "ph": "s", "id": 224470, "pid": 76337, "tid": -914061504, "ts": 1716454224950484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950551, "dur": 1,
    "args": {
      "External id": 224486,
      "cbid": 251, "correlation": 224486
    }
  },
  {
    "ph": "f", "id": 224486, "pid": 76337, "tid": -914061504, "ts": 1716454224950551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950556, "dur": 0,
    "args": {
      "External id": 224488,
      "cbid": 251, "correlation": 224488
    }
  },
  {
    "ph": "f", "id": 224488, "pid": 76337, "tid": -914061504, "ts": 1716454224950556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225005360, "dur": 12,
    "args": {
      "External id": 224489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224489,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224489, "pid": 5, "tid": 7, "ts": 1716454225005360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950558, "dur": 11,
    "args": {
      "External id": 224489,
      "cbid": 211, "correlation": 224489
    }
  },
  {
    "ph": "s", "id": 224489, "pid": 76337, "tid": -914061504, "ts": 1716454224950558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225005374, "dur": 5,
    "args": {
      "External id": 224491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224491,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224491, "pid": 5, "tid": 7, "ts": 1716454225005374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950571, "dur": 5,
    "args": {
      "External id": 224491,
      "cbid": 211, "correlation": 224491
    }
  },
  {
    "ph": "s", "id": 224491, "pid": 76337, "tid": -914061504, "ts": 1716454224950571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225005380, "dur": 6,
    "args": {
      "External id": 224501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224501, "pid": 5, "tid": 7, "ts": 1716454225005380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950629, "dur": 13,
    "args": {
      "External id": 224501,
      "cbid": 211, "correlation": 224501
    }
  },
  {
    "ph": "s", "id": 224501, "pid": 76337, "tid": -914061504, "ts": 1716454224950629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005387, "dur": 10,
    "args": {
      "External id": 224521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224521,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 224521, "pid": 5, "tid": 7, "ts": 1716454225005387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950696, "dur": 11,
    "args": {
      "External id": 224521,
      "cbid": 211, "correlation": 224521
    }
  },
  {
    "ph": "s", "id": 224521, "pid": 76337, "tid": -914061504, "ts": 1716454224950696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225005399, "dur": 4,
    "args": {
      "External id": 224533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224533,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 224533, "pid": 5, "tid": 7, "ts": 1716454225005399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950717, "dur": 6,
    "args": {
      "External id": 224533,
      "cbid": 211, "correlation": 224533
    }
  },
  {
    "ph": "s", "id": 224533, "pid": 76337, "tid": -914061504, "ts": 1716454224950717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225005404, "dur": 7,
    "args": {
      "External id": 224536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224536, "pid": 5, "tid": 7, "ts": 1716454225005404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950736, "dur": 7,
    "args": {
      "External id": 224536,
      "cbid": 211, "correlation": 224536
    }
  },
  {
    "ph": "s", "id": 224536, "pid": 76337, "tid": -914061504, "ts": 1716454224950736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225005412, "dur": 4,
    "args": {
      "External id": 224545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224545,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224545, "pid": 5, "tid": 7, "ts": 1716454225005412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950775, "dur": 10,
    "args": {
      "External id": 224545,
      "cbid": 211, "correlation": 224545
    }
  },
  {
    "ph": "s", "id": 224545, "pid": 76337, "tid": -914061504, "ts": 1716454224950775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950838, "dur": 0,
    "args": {
      "External id": 224555,
      "cbid": 317, "correlation": 224555
    }
  },
  {
    "ph": "f", "id": 224555, "pid": 76337, "tid": -914061504, "ts": 1716454224950838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950839, "dur": 0,
    "args": {
      "External id": 224556,
      "cbid": 203, "correlation": 224556
    }
  },
  {
    "ph": "f", "id": 224556, "pid": 76337, "tid": -914061504, "ts": 1716454224950839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950840, "dur": 0,
    "args": {
      "External id": 224557,
      "cbid": 205, "correlation": 224557
    }
  },
  {
    "ph": "f", "id": 224557, "pid": 76337, "tid": -914061504, "ts": 1716454224950840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005418, "dur": 5,
    "args": {
      "External id": 224561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224561,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224561, "pid": 5, "tid": 7, "ts": 1716454225005418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950853, "dur": 12,
    "args": {
      "External id": 224561,
      "cbid": 211, "correlation": 224561
    }
  },
  {
    "ph": "s", "id": 224561, "pid": 76337, "tid": -914061504, "ts": 1716454224950853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005424, "dur": 159,
    "args": {
      "External id": 224563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224563,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224563, "pid": 5, "tid": 7, "ts": 1716454225005424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950867, "dur": 5,
    "args": {
      "External id": 224563,
      "cbid": 211, "correlation": 224563
    }
  },
  {
    "ph": "s", "id": 224563, "pid": 76337, "tid": -914061504, "ts": 1716454224950867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225005585, "dur": 1,
    "args": {
      "External id": 224565,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224565,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 224565, "pid": 5, "tid": 7, "ts": 1716454225005585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950878, "dur": 6,
    "args": {
      "External id": 224565,
      "cbid": 51, "correlation": 224565
    }
  },
  {
    "ph": "s", "id": 224565, "pid": 76337, "tid": -914061504, "ts": 1716454224950878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225005589, "dur": 254,
    "args": {
      "External id": 224566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224566,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224566, "pid": 5, "tid": 7, "ts": 1716454225005589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950885, "dur": 7,
    "args": {
      "External id": 224566,
      "cbid": 211, "correlation": 224566
    }
  },
  {
    "ph": "s", "id": 224566, "pid": 76337, "tid": -914061504, "ts": 1716454224950885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005844, "dur": 6,
    "args": {
      "External id": 224568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224568,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224568, "pid": 5, "tid": 7, "ts": 1716454225005844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950896, "dur": 5,
    "args": {
      "External id": 224568,
      "cbid": 211, "correlation": 224568
    }
  },
  {
    "ph": "s", "id": 224568, "pid": 76337, "tid": -914061504, "ts": 1716454224950896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225005852, "dur": 6,
    "args": {
      "External id": 224574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224574,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224574, "pid": 5, "tid": 7, "ts": 1716454225005852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950924, "dur": 8,
    "args": {
      "External id": 224574,
      "cbid": 211, "correlation": 224574
    }
  },
  {
    "ph": "s", "id": 224574, "pid": 76337, "tid": -914061504, "ts": 1716454224950924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225005859, "dur": 5,
    "args": {
      "External id": 224582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224582,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224582, "pid": 5, "tid": 7, "ts": 1716454225005859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950957, "dur": 8,
    "args": {
      "External id": 224582,
      "cbid": 211, "correlation": 224582
    }
  },
  {
    "ph": "s", "id": 224582, "pid": 76337, "tid": -914061504, "ts": 1716454224950957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225005865, "dur": 5,
    "args": {
      "External id": 224590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224590,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224590, "pid": 5, "tid": 7, "ts": 1716454225005865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224950994, "dur": 9,
    "args": {
      "External id": 224590,
      "cbid": 211, "correlation": 224590
    }
  },
  {
    "ph": "s", "id": 224590, "pid": 76337, "tid": -914061504, "ts": 1716454224950994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225005871, "dur": 11,
    "args": {
      "External id": 224599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224599,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224599, "pid": 5, "tid": 7, "ts": 1716454225005871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951082, "dur": 14,
    "args": {
      "External id": 224599,
      "cbid": 211, "correlation": 224599
    }
  },
  {
    "ph": "s", "id": 224599, "pid": 76337, "tid": -914061504, "ts": 1716454224951082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005884, "dur": 12,
    "args": {
      "External id": 224619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224619,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 224619, "pid": 5, "tid": 7, "ts": 1716454225005884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951150, "dur": 10,
    "args": {
      "External id": 224619,
      "cbid": 211, "correlation": 224619
    }
  },
  {
    "ph": "s", "id": 224619, "pid": 76337, "tid": -914061504, "ts": 1716454224951150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225005897, "dur": 4,
    "args": {
      "External id": 224631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224631,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224631, "pid": 5, "tid": 7, "ts": 1716454225005897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951171, "dur": 6,
    "args": {
      "External id": 224631,
      "cbid": 211, "correlation": 224631
    }
  },
  {
    "ph": "s", "id": 224631, "pid": 76337, "tid": -914061504, "ts": 1716454224951171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225005902, "dur": 10,
    "args": {
      "External id": 224634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224634, "pid": 5, "tid": 7, "ts": 1716454225005902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951190, "dur": 6,
    "args": {
      "External id": 224634,
      "cbid": 211, "correlation": 224634
    }
  },
  {
    "ph": "s", "id": 224634, "pid": 76337, "tid": -914061504, "ts": 1716454224951190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225005913, "dur": 6,
    "args": {
      "External id": 224643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224643,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224643, "pid": 5, "tid": 7, "ts": 1716454225005913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951227, "dur": 10,
    "args": {
      "External id": 224643,
      "cbid": 211, "correlation": 224643
    }
  },
  {
    "ph": "s", "id": 224643, "pid": 76337, "tid": -914061504, "ts": 1716454224951227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951280, "dur": 0,
    "args": {
      "External id": 224653,
      "cbid": 317, "correlation": 224653
    }
  },
  {
    "ph": "f", "id": 224653, "pid": 76337, "tid": -914061504, "ts": 1716454224951280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951280, "dur": 0,
    "args": {
      "External id": 224654,
      "cbid": 203, "correlation": 224654
    }
  },
  {
    "ph": "f", "id": 224654, "pid": 76337, "tid": -914061504, "ts": 1716454224951280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951281, "dur": 0,
    "args": {
      "External id": 224655,
      "cbid": 205, "correlation": 224655
    }
  },
  {
    "ph": "f", "id": 224655, "pid": 76337, "tid": -914061504, "ts": 1716454224951281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005921, "dur": 6,
    "args": {
      "External id": 224659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224659, "pid": 5, "tid": 7, "ts": 1716454225005921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951297, "dur": 11,
    "args": {
      "External id": 224659,
      "cbid": 211, "correlation": 224659
    }
  },
  {
    "ph": "s", "id": 224659, "pid": 76337, "tid": -914061504, "ts": 1716454224951297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225005928, "dur": 313,
    "args": {
      "External id": 224661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224661, "pid": 5, "tid": 7, "ts": 1716454225005928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951311, "dur": 5,
    "args": {
      "External id": 224661,
      "cbid": 211, "correlation": 224661
    }
  },
  {
    "ph": "s", "id": 224661, "pid": 76337, "tid": -914061504, "ts": 1716454224951311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225006243, "dur": 1,
    "args": {
      "External id": 224663,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224663,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 224663, "pid": 5, "tid": 7, "ts": 1716454225006243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951322, "dur": 7,
    "args": {
      "External id": 224663,
      "cbid": 51, "correlation": 224663
    }
  },
  {
    "ph": "s", "id": 224663, "pid": 76337, "tid": -914061504, "ts": 1716454224951322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225006247, "dur": 486,
    "args": {
      "External id": 224664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224664,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224664, "pid": 5, "tid": 7, "ts": 1716454225006247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951330, "dur": 7,
    "args": {
      "External id": 224664,
      "cbid": 211, "correlation": 224664
    }
  },
  {
    "ph": "s", "id": 224664, "pid": 76337, "tid": -914061504, "ts": 1716454224951330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225006734, "dur": 6,
    "args": {
      "External id": 224666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224666, "pid": 5, "tid": 7, "ts": 1716454225006734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951341, "dur": 5,
    "args": {
      "External id": 224666,
      "cbid": 211, "correlation": 224666
    }
  },
  {
    "ph": "s", "id": 224666, "pid": 76337, "tid": -914061504, "ts": 1716454224951341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225006741, "dur": 6,
    "args": {
      "External id": 224672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224672, "pid": 5, "tid": 7, "ts": 1716454225006741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951369, "dur": 9,
    "args": {
      "External id": 224672,
      "cbid": 211, "correlation": 224672
    }
  },
  {
    "ph": "s", "id": 224672, "pid": 76337, "tid": -914061504, "ts": 1716454224951369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225006749, "dur": 3,
    "args": {
      "External id": 224680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224680,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 224680, "pid": 5, "tid": 7, "ts": 1716454225006749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951413, "dur": 9,
    "args": {
      "External id": 224680,
      "cbid": 211, "correlation": 224680
    }
  },
  {
    "ph": "s", "id": 224680, "pid": 76337, "tid": -914061504, "ts": 1716454224951413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951475, "dur": 1,
    "args": {
      "External id": 224696,
      "cbid": 251, "correlation": 224696
    }
  },
  {
    "ph": "f", "id": 224696, "pid": 76337, "tid": -914061504, "ts": 1716454224951475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951481, "dur": 0,
    "args": {
      "External id": 224698,
      "cbid": 251, "correlation": 224698
    }
  },
  {
    "ph": "f", "id": 224698, "pid": 76337, "tid": -914061504, "ts": 1716454224951481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225006753, "dur": 11,
    "args": {
      "External id": 224699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224699,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224699, "pid": 5, "tid": 7, "ts": 1716454225006753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951483, "dur": 11,
    "args": {
      "External id": 224699,
      "cbid": 211, "correlation": 224699
    }
  },
  {
    "ph": "s", "id": 224699, "pid": 76337, "tid": -914061504, "ts": 1716454224951483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225006765, "dur": 4,
    "args": {
      "External id": 224701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224701,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224701, "pid": 5, "tid": 7, "ts": 1716454225006765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951496, "dur": 5,
    "args": {
      "External id": 224701,
      "cbid": 211, "correlation": 224701
    }
  },
  {
    "ph": "s", "id": 224701, "pid": 76337, "tid": -914061504, "ts": 1716454224951496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225006771, "dur": 6,
    "args": {
      "External id": 224711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224711, "pid": 5, "tid": 7, "ts": 1716454225006771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951552, "dur": 12,
    "args": {
      "External id": 224711,
      "cbid": 211, "correlation": 224711
    }
  },
  {
    "ph": "s", "id": 224711, "pid": 76337, "tid": -914061504, "ts": 1716454224951552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225006778, "dur": 9,
    "args": {
      "External id": 224731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224731,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 224731, "pid": 5, "tid": 7, "ts": 1716454225006778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951618, "dur": 11,
    "args": {
      "External id": 224731,
      "cbid": 211, "correlation": 224731
    }
  },
  {
    "ph": "s", "id": 224731, "pid": 76337, "tid": -914061504, "ts": 1716454224951618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225006788, "dur": 4,
    "args": {
      "External id": 224743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224743,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 224743, "pid": 5, "tid": 7, "ts": 1716454225006788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951640, "dur": 6,
    "args": {
      "External id": 224743,
      "cbid": 211, "correlation": 224743
    }
  },
  {
    "ph": "s", "id": 224743, "pid": 76337, "tid": -914061504, "ts": 1716454224951640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225006793, "dur": 7,
    "args": {
      "External id": 224746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224746,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224746, "pid": 5, "tid": 7, "ts": 1716454225006793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951658, "dur": 6,
    "args": {
      "External id": 224746,
      "cbid": 211, "correlation": 224746
    }
  },
  {
    "ph": "s", "id": 224746, "pid": 76337, "tid": -914061504, "ts": 1716454224951658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225006802, "dur": 4,
    "args": {
      "External id": 224755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224755,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224755, "pid": 5, "tid": 7, "ts": 1716454225006802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951697, "dur": 9,
    "args": {
      "External id": 224755,
      "cbid": 211, "correlation": 224755
    }
  },
  {
    "ph": "s", "id": 224755, "pid": 76337, "tid": -914061504, "ts": 1716454224951697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951758, "dur": 0,
    "args": {
      "External id": 224765,
      "cbid": 317, "correlation": 224765
    }
  },
  {
    "ph": "f", "id": 224765, "pid": 76337, "tid": -914061504, "ts": 1716454224951758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951759, "dur": 0,
    "args": {
      "External id": 224766,
      "cbid": 203, "correlation": 224766
    }
  },
  {
    "ph": "f", "id": 224766, "pid": 76337, "tid": -914061504, "ts": 1716454224951759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951760, "dur": 0,
    "args": {
      "External id": 224767,
      "cbid": 205, "correlation": 224767
    }
  },
  {
    "ph": "f", "id": 224767, "pid": 76337, "tid": -914061504, "ts": 1716454224951760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225006807, "dur": 5,
    "args": {
      "External id": 224771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224771,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224771, "pid": 5, "tid": 7, "ts": 1716454225006807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951774, "dur": 12,
    "args": {
      "External id": 224771,
      "cbid": 211, "correlation": 224771
    }
  },
  {
    "ph": "s", "id": 224771, "pid": 76337, "tid": -914061504, "ts": 1716454224951774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225006814, "dur": 159,
    "args": {
      "External id": 224773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224773, "pid": 5, "tid": 7, "ts": 1716454225006814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951788, "dur": 6,
    "args": {
      "External id": 224773,
      "cbid": 211, "correlation": 224773
    }
  },
  {
    "ph": "s", "id": 224773, "pid": 76337, "tid": -914061504, "ts": 1716454224951788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225006975, "dur": 1,
    "args": {
      "External id": 224775,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224775,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 224775, "pid": 5, "tid": 7, "ts": 1716454225006975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951799, "dur": 6,
    "args": {
      "External id": 224775,
      "cbid": 51, "correlation": 224775
    }
  },
  {
    "ph": "s", "id": 224775, "pid": 76337, "tid": -914061504, "ts": 1716454224951799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225006979, "dur": 254,
    "args": {
      "External id": 224776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224776,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224776, "pid": 5, "tid": 7, "ts": 1716454225006979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951807, "dur": 6,
    "args": {
      "External id": 224776,
      "cbid": 211, "correlation": 224776
    }
  },
  {
    "ph": "s", "id": 224776, "pid": 76337, "tid": -914061504, "ts": 1716454224951807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225007234, "dur": 6,
    "args": {
      "External id": 224778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224778, "pid": 5, "tid": 7, "ts": 1716454225007234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951816, "dur": 5,
    "args": {
      "External id": 224778,
      "cbid": 211, "correlation": 224778
    }
  },
  {
    "ph": "s", "id": 224778, "pid": 76337, "tid": -914061504, "ts": 1716454224951816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225007241, "dur": 6,
    "args": {
      "External id": 224784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224784, "pid": 5, "tid": 7, "ts": 1716454225007241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951845, "dur": 8,
    "args": {
      "External id": 224784,
      "cbid": 211, "correlation": 224784
    }
  },
  {
    "ph": "s", "id": 224784, "pid": 76337, "tid": -914061504, "ts": 1716454224951845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951904, "dur": 0,
    "args": {
      "External id": 224794,
      "cbid": 317, "correlation": 224794
    }
  },
  {
    "ph": "f", "id": 224794, "pid": 76337, "tid": -914061504, "ts": 1716454224951904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951904, "dur": 0,
    "args": {
      "External id": 224795,
      "cbid": 203, "correlation": 224795
    }
  },
  {
    "ph": "f", "id": 224795, "pid": 76337, "tid": -914061504, "ts": 1716454224951904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951905, "dur": 0,
    "args": {
      "External id": 224796,
      "cbid": 205, "correlation": 224796
    }
  },
  {
    "ph": "f", "id": 224796, "pid": 76337, "tid": -914061504, "ts": 1716454224951905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225007248, "dur": 8,
    "args": {
      "External id": 224800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224800, "pid": 5, "tid": 7, "ts": 1716454225007248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951919, "dur": 11,
    "args": {
      "External id": 224800,
      "cbid": 211, "correlation": 224800
    }
  },
  {
    "ph": "s", "id": 224800, "pid": 76337, "tid": -914061504, "ts": 1716454224951919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225007258, "dur": 3,
    "args": {
      "External id": 224802,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224802,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.530612244897959
    }
  },
  {
    "ph": "f", "id": 224802, "pid": 5, "tid": 7, "ts": 1716454225007258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951937, "dur": 14,
    "args": {
      "External id": 224802,
      "cbid": 51, "correlation": 224802
    }
  },
  {
    "ph": "s", "id": 224802, "pid": 76337, "tid": -914061504, "ts": 1716454224951937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225007262, "dur": 94,
    "args": {
      "External id": 224803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224803,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 224803, "pid": 5, "tid": 7, "ts": 1716454225007262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951952, "dur": 7,
    "args": {
      "External id": 224803,
      "cbid": 211, "correlation": 224803
    }
  },
  {
    "ph": "s", "id": 224803, "pid": 76337, "tid": -914061504, "ts": 1716454224951952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225007357, "dur": 6,
    "args": {
      "External id": 224805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224805,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224805, "pid": 5, "tid": 7, "ts": 1716454225007357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951962, "dur": 5,
    "args": {
      "External id": 224805,
      "cbid": 211, "correlation": 224805
    }
  },
  {
    "ph": "s", "id": 224805, "pid": 76337, "tid": -914061504, "ts": 1716454224951962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225007364, "dur": 6,
    "args": {
      "External id": 224811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224811,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224811, "pid": 5, "tid": 7, "ts": 1716454225007364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224951997, "dur": 9,
    "args": {
      "External id": 224811,
      "cbid": 211, "correlation": 224811
    }
  },
  {
    "ph": "s", "id": 224811, "pid": 76337, "tid": -914061504, "ts": 1716454224951997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225007371, "dur": 5,
    "args": {
      "External id": 224819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224819,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224819, "pid": 5, "tid": 7, "ts": 1716454225007371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952028, "dur": 8,
    "args": {
      "External id": 224819,
      "cbid": 211, "correlation": 224819
    }
  },
  {
    "ph": "s", "id": 224819, "pid": 76337, "tid": -914061504, "ts": 1716454224952028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225007377, "dur": 4,
    "args": {
      "External id": 224827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224827,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224827, "pid": 5, "tid": 7, "ts": 1716454225007377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952058, "dur": 8,
    "args": {
      "External id": 224827,
      "cbid": 211, "correlation": 224827
    }
  },
  {
    "ph": "s", "id": 224827, "pid": 76337, "tid": -914061504, "ts": 1716454224952058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225007383, "dur": 11,
    "args": {
      "External id": 224836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224836,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224836, "pid": 5, "tid": 7, "ts": 1716454225007383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952101, "dur": 11,
    "args": {
      "External id": 224836,
      "cbid": 211, "correlation": 224836
    }
  },
  {
    "ph": "s", "id": 224836, "pid": 76337, "tid": -914061504, "ts": 1716454224952101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225007395, "dur": 12,
    "args": {
      "External id": 224856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224856,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 224856, "pid": 5, "tid": 7, "ts": 1716454225007395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952171, "dur": 11,
    "args": {
      "External id": 224856,
      "cbid": 211, "correlation": 224856
    }
  },
  {
    "ph": "s", "id": 224856, "pid": 76337, "tid": -914061504, "ts": 1716454224952171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225007409, "dur": 4,
    "args": {
      "External id": 224868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224868,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224868, "pid": 5, "tid": 7, "ts": 1716454225007409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952193, "dur": 6,
    "args": {
      "External id": 224868,
      "cbid": 211, "correlation": 224868
    }
  },
  {
    "ph": "s", "id": 224868, "pid": 76337, "tid": -914061504, "ts": 1716454224952193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225007414, "dur": 11,
    "args": {
      "External id": 224871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224871,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224871, "pid": 5, "tid": 7, "ts": 1716454225007414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952211, "dur": 7,
    "args": {
      "External id": 224871,
      "cbid": 211, "correlation": 224871
    }
  },
  {
    "ph": "s", "id": 224871, "pid": 76337, "tid": -914061504, "ts": 1716454224952211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225007426, "dur": 6,
    "args": {
      "External id": 224880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224880,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224880, "pid": 5, "tid": 7, "ts": 1716454225007426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952249, "dur": 9,
    "args": {
      "External id": 224880,
      "cbid": 211, "correlation": 224880
    }
  },
  {
    "ph": "s", "id": 224880, "pid": 76337, "tid": -914061504, "ts": 1716454224952249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952301, "dur": 0,
    "args": {
      "External id": 224890,
      "cbid": 317, "correlation": 224890
    }
  },
  {
    "ph": "f", "id": 224890, "pid": 76337, "tid": -914061504, "ts": 1716454224952301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952302, "dur": 0,
    "args": {
      "External id": 224891,
      "cbid": 203, "correlation": 224891
    }
  },
  {
    "ph": "f", "id": 224891, "pid": 76337, "tid": -914061504, "ts": 1716454224952302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952303, "dur": 0,
    "args": {
      "External id": 224892,
      "cbid": 205, "correlation": 224892
    }
  },
  {
    "ph": "f", "id": 224892, "pid": 76337, "tid": -914061504, "ts": 1716454224952303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225007434, "dur": 7,
    "args": {
      "External id": 224896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224896,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224896, "pid": 5, "tid": 7, "ts": 1716454225007434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952317, "dur": 11,
    "args": {
      "External id": 224896,
      "cbid": 211, "correlation": 224896
    }
  },
  {
    "ph": "s", "id": 224896, "pid": 76337, "tid": -914061504, "ts": 1716454224952317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225007442, "dur": 315,
    "args": {
      "External id": 224898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224898,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224898, "pid": 5, "tid": 7, "ts": 1716454225007442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952331, "dur": 5,
    "args": {
      "External id": 224898,
      "cbid": 211, "correlation": 224898
    }
  },
  {
    "ph": "s", "id": 224898, "pid": 76337, "tid": -914061504, "ts": 1716454224952331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225007759, "dur": 1,
    "args": {
      "External id": 224900,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 224900,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 224900, "pid": 5, "tid": 7, "ts": 1716454225007759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952341, "dur": 7,
    "args": {
      "External id": 224900,
      "cbid": 51, "correlation": 224900
    }
  },
  {
    "ph": "s", "id": 224900, "pid": 76337, "tid": -914061504, "ts": 1716454224952341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225007763, "dur": 490,
    "args": {
      "External id": 224901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224901,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224901, "pid": 5, "tid": 7, "ts": 1716454225007763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952350, "dur": 6,
    "args": {
      "External id": 224901,
      "cbid": 211, "correlation": 224901
    }
  },
  {
    "ph": "s", "id": 224901, "pid": 76337, "tid": -914061504, "ts": 1716454224952350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008254, "dur": 5,
    "args": {
      "External id": 224903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224903,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 224903, "pid": 5, "tid": 7, "ts": 1716454225008254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952359, "dur": 5,
    "args": {
      "External id": 224903,
      "cbid": 211, "correlation": 224903
    }
  },
  {
    "ph": "s", "id": 224903, "pid": 76337, "tid": -914061504, "ts": 1716454224952359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225008260, "dur": 6,
    "args": {
      "External id": 224909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224909, "pid": 5, "tid": 7, "ts": 1716454225008260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952387, "dur": 9,
    "args": {
      "External id": 224909,
      "cbid": 211, "correlation": 224909
    }
  },
  {
    "ph": "s", "id": 224909, "pid": 76337, "tid": -914061504, "ts": 1716454224952387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225008268, "dur": 3,
    "args": {
      "External id": 224917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224917,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 224917, "pid": 5, "tid": 7, "ts": 1716454225008268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952430, "dur": 9,
    "args": {
      "External id": 224917,
      "cbid": 211, "correlation": 224917
    }
  },
  {
    "ph": "s", "id": 224917, "pid": 76337, "tid": -914061504, "ts": 1716454224952430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952493, "dur": 1,
    "args": {
      "External id": 224933,
      "cbid": 251, "correlation": 224933
    }
  },
  {
    "ph": "f", "id": 224933, "pid": 76337, "tid": -914061504, "ts": 1716454224952493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952499, "dur": 0,
    "args": {
      "External id": 224935,
      "cbid": 251, "correlation": 224935
    }
  },
  {
    "ph": "f", "id": 224935, "pid": 76337, "tid": -914061504, "ts": 1716454224952499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225008272, "dur": 13,
    "args": {
      "External id": 224936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224936,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224936, "pid": 5, "tid": 7, "ts": 1716454225008272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952501, "dur": 11,
    "args": {
      "External id": 224936,
      "cbid": 211, "correlation": 224936
    }
  },
  {
    "ph": "s", "id": 224936, "pid": 76337, "tid": -914061504, "ts": 1716454224952501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225008286, "dur": 5,
    "args": {
      "External id": 224938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224938,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 224938, "pid": 5, "tid": 7, "ts": 1716454225008286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952513, "dur": 5,
    "args": {
      "External id": 224938,
      "cbid": 211, "correlation": 224938
    }
  },
  {
    "ph": "s", "id": 224938, "pid": 76337, "tid": -914061504, "ts": 1716454224952513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225008292, "dur": 6,
    "args": {
      "External id": 224948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224948, "pid": 5, "tid": 7, "ts": 1716454225008292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952571, "dur": 12,
    "args": {
      "External id": 224948,
      "cbid": 211, "correlation": 224948
    }
  },
  {
    "ph": "s", "id": 224948, "pid": 76337, "tid": -914061504, "ts": 1716454224952571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008299, "dur": 9,
    "args": {
      "External id": 224968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224968,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 224968, "pid": 5, "tid": 7, "ts": 1716454225008299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952635, "dur": 11,
    "args": {
      "External id": 224968,
      "cbid": 211, "correlation": 224968
    }
  },
  {
    "ph": "s", "id": 224968, "pid": 76337, "tid": -914061504, "ts": 1716454224952635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225008310, "dur": 4,
    "args": {
      "External id": 224980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224980,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 224980, "pid": 5, "tid": 7, "ts": 1716454225008310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952657, "dur": 6,
    "args": {
      "External id": 224980,
      "cbid": 211, "correlation": 224980
    }
  },
  {
    "ph": "s", "id": 224980, "pid": 76337, "tid": -914061504, "ts": 1716454224952657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225008315, "dur": 7,
    "args": {
      "External id": 224983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224983, "pid": 5, "tid": 7, "ts": 1716454225008315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952674, "dur": 6,
    "args": {
      "External id": 224983,
      "cbid": 211, "correlation": 224983
    }
  },
  {
    "ph": "s", "id": 224983, "pid": 76337, "tid": -914061504, "ts": 1716454224952674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225008322, "dur": 5,
    "args": {
      "External id": 224992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 224992,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 224992, "pid": 5, "tid": 7, "ts": 1716454225008322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952715, "dur": 10,
    "args": {
      "External id": 224992,
      "cbid": 211, "correlation": 224992
    }
  },
  {
    "ph": "s", "id": 224992, "pid": 76337, "tid": -914061504, "ts": 1716454224952715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952777, "dur": 0,
    "args": {
      "External id": 225002,
      "cbid": 317, "correlation": 225002
    }
  },
  {
    "ph": "f", "id": 225002, "pid": 76337, "tid": -914061504, "ts": 1716454224952777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952778, "dur": 0,
    "args": {
      "External id": 225003,
      "cbid": 203, "correlation": 225003
    }
  },
  {
    "ph": "f", "id": 225003, "pid": 76337, "tid": -914061504, "ts": 1716454224952778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952778, "dur": 0,
    "args": {
      "External id": 225004,
      "cbid": 205, "correlation": 225004
    }
  },
  {
    "ph": "f", "id": 225004, "pid": 76337, "tid": -914061504, "ts": 1716454224952778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008328, "dur": 5,
    "args": {
      "External id": 225008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225008,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225008, "pid": 5, "tid": 7, "ts": 1716454225008328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952792, "dur": 13,
    "args": {
      "External id": 225008,
      "cbid": 211, "correlation": 225008
    }
  },
  {
    "ph": "s", "id": 225008, "pid": 76337, "tid": -914061504, "ts": 1716454224952792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008334, "dur": 159,
    "args": {
      "External id": 225010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225010,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225010, "pid": 5, "tid": 7, "ts": 1716454225008334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952807, "dur": 5,
    "args": {
      "External id": 225010,
      "cbid": 211, "correlation": 225010
    }
  },
  {
    "ph": "s", "id": 225010, "pid": 76337, "tid": -914061504, "ts": 1716454224952807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225008495, "dur": 1,
    "args": {
      "External id": 225012,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225012,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 225012, "pid": 5, "tid": 7, "ts": 1716454225008495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952818, "dur": 6,
    "args": {
      "External id": 225012,
      "cbid": 51, "correlation": 225012
    }
  },
  {
    "ph": "s", "id": 225012, "pid": 76337, "tid": -914061504, "ts": 1716454224952818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225008499, "dur": 254,
    "args": {
      "External id": 225013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225013,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225013, "pid": 5, "tid": 7, "ts": 1716454225008499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952825, "dur": 6,
    "args": {
      "External id": 225013,
      "cbid": 211, "correlation": 225013
    }
  },
  {
    "ph": "s", "id": 225013, "pid": 76337, "tid": -914061504, "ts": 1716454224952825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008755, "dur": 6,
    "args": {
      "External id": 225015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225015,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225015, "pid": 5, "tid": 7, "ts": 1716454225008755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952835, "dur": 5,
    "args": {
      "External id": 225015,
      "cbid": 211, "correlation": 225015
    }
  },
  {
    "ph": "s", "id": 225015, "pid": 76337, "tid": -914061504, "ts": 1716454224952835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225008762, "dur": 6,
    "args": {
      "External id": 225021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225021,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225021, "pid": 5, "tid": 7, "ts": 1716454225008762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952864, "dur": 8,
    "args": {
      "External id": 225021,
      "cbid": 211, "correlation": 225021
    }
  },
  {
    "ph": "s", "id": 225021, "pid": 76337, "tid": -914061504, "ts": 1716454224952864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952922, "dur": 0,
    "args": {
      "External id": 225031,
      "cbid": 317, "correlation": 225031
    }
  },
  {
    "ph": "f", "id": 225031, "pid": 76337, "tid": -914061504, "ts": 1716454224952922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952922, "dur": 0,
    "args": {
      "External id": 225032,
      "cbid": 203, "correlation": 225032
    }
  },
  {
    "ph": "f", "id": 225032, "pid": 76337, "tid": -914061504, "ts": 1716454224952922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952923, "dur": 0,
    "args": {
      "External id": 225033,
      "cbid": 205, "correlation": 225033
    }
  },
  {
    "ph": "f", "id": 225033, "pid": 76337, "tid": -914061504, "ts": 1716454224952923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008769, "dur": 8,
    "args": {
      "External id": 225037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225037,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225037, "pid": 5, "tid": 7, "ts": 1716454225008769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952934, "dur": 11,
    "args": {
      "External id": 225037,
      "cbid": 211, "correlation": 225037
    }
  },
  {
    "ph": "s", "id": 225037, "pid": 76337, "tid": -914061504, "ts": 1716454224952934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225008778, "dur": 3,
    "args": {
      "External id": 225039,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225039,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.530612244897959
    }
  },
  {
    "ph": "f", "id": 225039, "pid": 5, "tid": 7, "ts": 1716454225008778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952951, "dur": 10,
    "args": {
      "External id": 225039,
      "cbid": 51, "correlation": 225039
    }
  },
  {
    "ph": "s", "id": 225039, "pid": 76337, "tid": -914061504, "ts": 1716454224952951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225008782, "dur": 94,
    "args": {
      "External id": 225040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225040,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 225040, "pid": 5, "tid": 7, "ts": 1716454225008782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952962, "dur": 6,
    "args": {
      "External id": 225040,
      "cbid": 211, "correlation": 225040
    }
  },
  {
    "ph": "s", "id": 225040, "pid": 76337, "tid": -914061504, "ts": 1716454224952962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008877, "dur": 6,
    "args": {
      "External id": 225042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225042,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225042, "pid": 5, "tid": 7, "ts": 1716454225008877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224952971, "dur": 13,
    "args": {
      "External id": 225042,
      "cbid": 211, "correlation": 225042
    }
  },
  {
    "ph": "s", "id": 225042, "pid": 76337, "tid": -914061504, "ts": 1716454224952971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225008884, "dur": 6,
    "args": {
      "External id": 225048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225048,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225048, "pid": 5, "tid": 7, "ts": 1716454225008884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953007, "dur": 9,
    "args": {
      "External id": 225048,
      "cbid": 211, "correlation": 225048
    }
  },
  {
    "ph": "s", "id": 225048, "pid": 76337, "tid": -914061504, "ts": 1716454224953007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225008892, "dur": 5,
    "args": {
      "External id": 225056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225056,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225056, "pid": 5, "tid": 7, "ts": 1716454225008892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953037, "dur": 8,
    "args": {
      "External id": 225056,
      "cbid": 211, "correlation": 225056
    }
  },
  {
    "ph": "s", "id": 225056, "pid": 76337, "tid": -914061504, "ts": 1716454224953037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225008898, "dur": 4,
    "args": {
      "External id": 225064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225064,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225064, "pid": 5, "tid": 7, "ts": 1716454225008898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953066, "dur": 8,
    "args": {
      "External id": 225064,
      "cbid": 211, "correlation": 225064
    }
  },
  {
    "ph": "s", "id": 225064, "pid": 76337, "tid": -914061504, "ts": 1716454224953066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225008904, "dur": 11,
    "args": {
      "External id": 225073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225073,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225073, "pid": 5, "tid": 7, "ts": 1716454225008904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953110, "dur": 11,
    "args": {
      "External id": 225073,
      "cbid": 211, "correlation": 225073
    }
  },
  {
    "ph": "s", "id": 225073, "pid": 76337, "tid": -914061504, "ts": 1716454224953110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008916, "dur": 12,
    "args": {
      "External id": 225093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225093,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 225093, "pid": 5, "tid": 7, "ts": 1716454225008916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953179, "dur": 11,
    "args": {
      "External id": 225093,
      "cbid": 211, "correlation": 225093
    }
  },
  {
    "ph": "s", "id": 225093, "pid": 76337, "tid": -914061504, "ts": 1716454224953179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225008929, "dur": 4,
    "args": {
      "External id": 225105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225105,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225105, "pid": 5, "tid": 7, "ts": 1716454225008929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953199, "dur": 7,
    "args": {
      "External id": 225105,
      "cbid": 211, "correlation": 225105
    }
  },
  {
    "ph": "s", "id": 225105, "pid": 76337, "tid": -914061504, "ts": 1716454224953199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225008934, "dur": 11,
    "args": {
      "External id": 225108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225108, "pid": 5, "tid": 7, "ts": 1716454225008934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953218, "dur": 6,
    "args": {
      "External id": 225108,
      "cbid": 211, "correlation": 225108
    }
  },
  {
    "ph": "s", "id": 225108, "pid": 76337, "tid": -914061504, "ts": 1716454224953218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225008946, "dur": 6,
    "args": {
      "External id": 225117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225117,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225117, "pid": 5, "tid": 7, "ts": 1716454225008946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953256, "dur": 10,
    "args": {
      "External id": 225117,
      "cbid": 211, "correlation": 225117
    }
  },
  {
    "ph": "s", "id": 225117, "pid": 76337, "tid": -914061504, "ts": 1716454224953256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953308, "dur": 0,
    "args": {
      "External id": 225127,
      "cbid": 317, "correlation": 225127
    }
  },
  {
    "ph": "f", "id": 225127, "pid": 76337, "tid": -914061504, "ts": 1716454224953308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953309, "dur": 0,
    "args": {
      "External id": 225128,
      "cbid": 203, "correlation": 225128
    }
  },
  {
    "ph": "f", "id": 225128, "pid": 76337, "tid": -914061504, "ts": 1716454224953309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953310, "dur": 0,
    "args": {
      "External id": 225129,
      "cbid": 205, "correlation": 225129
    }
  },
  {
    "ph": "f", "id": 225129, "pid": 76337, "tid": -914061504, "ts": 1716454224953310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008953, "dur": 7,
    "args": {
      "External id": 225133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225133,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225133, "pid": 5, "tid": 7, "ts": 1716454225008953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953323, "dur": 11,
    "args": {
      "External id": 225133,
      "cbid": 211, "correlation": 225133
    }
  },
  {
    "ph": "s", "id": 225133, "pid": 76337, "tid": -914061504, "ts": 1716454224953323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225008961, "dur": 314,
    "args": {
      "External id": 225135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225135,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225135, "pid": 5, "tid": 7, "ts": 1716454225008961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953336, "dur": 5,
    "args": {
      "External id": 225135,
      "cbid": 211, "correlation": 225135
    }
  },
  {
    "ph": "s", "id": 225135, "pid": 76337, "tid": -914061504, "ts": 1716454224953336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225009277, "dur": 1,
    "args": {
      "External id": 225137,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225137,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 225137, "pid": 5, "tid": 7, "ts": 1716454225009277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953346, "dur": 7,
    "args": {
      "External id": 225137,
      "cbid": 51, "correlation": 225137
    }
  },
  {
    "ph": "s", "id": 225137, "pid": 76337, "tid": -914061504, "ts": 1716454224953346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225009281, "dur": 488,
    "args": {
      "External id": 225138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225138,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225138, "pid": 5, "tid": 7, "ts": 1716454225009281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953354, "dur": 6,
    "args": {
      "External id": 225138,
      "cbid": 211, "correlation": 225138
    }
  },
  {
    "ph": "s", "id": 225138, "pid": 76337, "tid": -914061504, "ts": 1716454224953354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225009770, "dur": 5,
    "args": {
      "External id": 225140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225140,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225140, "pid": 5, "tid": 7, "ts": 1716454225009770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953365, "dur": 5,
    "args": {
      "External id": 225140,
      "cbid": 211, "correlation": 225140
    }
  },
  {
    "ph": "s", "id": 225140, "pid": 76337, "tid": -914061504, "ts": 1716454224953365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225009776, "dur": 6,
    "args": {
      "External id": 225146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225146, "pid": 5, "tid": 7, "ts": 1716454225009776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953392, "dur": 8,
    "args": {
      "External id": 225146,
      "cbid": 211, "correlation": 225146
    }
  },
  {
    "ph": "s", "id": 225146, "pid": 76337, "tid": -914061504, "ts": 1716454224953392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225009784, "dur": 3,
    "args": {
      "External id": 225154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225154,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 225154, "pid": 5, "tid": 7, "ts": 1716454225009784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953435, "dur": 9,
    "args": {
      "External id": 225154,
      "cbid": 211, "correlation": 225154
    }
  },
  {
    "ph": "s", "id": 225154, "pid": 76337, "tid": -914061504, "ts": 1716454224953435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953497, "dur": 1,
    "args": {
      "External id": 225170,
      "cbid": 251, "correlation": 225170
    }
  },
  {
    "ph": "f", "id": 225170, "pid": 76337, "tid": -914061504, "ts": 1716454224953497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953503, "dur": 0,
    "args": {
      "External id": 225172,
      "cbid": 251, "correlation": 225172
    }
  },
  {
    "ph": "f", "id": 225172, "pid": 76337, "tid": -914061504, "ts": 1716454224953503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225009788, "dur": 12,
    "args": {
      "External id": 225173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225173,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225173, "pid": 5, "tid": 7, "ts": 1716454225009788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953505, "dur": 11,
    "args": {
      "External id": 225173,
      "cbid": 211, "correlation": 225173
    }
  },
  {
    "ph": "s", "id": 225173, "pid": 76337, "tid": -914061504, "ts": 1716454224953505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225009802, "dur": 5,
    "args": {
      "External id": 225175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225175,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225175, "pid": 5, "tid": 7, "ts": 1716454225009802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953518, "dur": 5,
    "args": {
      "External id": 225175,
      "cbid": 211, "correlation": 225175
    }
  },
  {
    "ph": "s", "id": 225175, "pid": 76337, "tid": -914061504, "ts": 1716454224953518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225009808, "dur": 6,
    "args": {
      "External id": 225185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225185, "pid": 5, "tid": 7, "ts": 1716454225009808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953574, "dur": 12,
    "args": {
      "External id": 225185,
      "cbid": 211, "correlation": 225185
    }
  },
  {
    "ph": "s", "id": 225185, "pid": 76337, "tid": -914061504, "ts": 1716454224953574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225009815, "dur": 9,
    "args": {
      "External id": 225205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225205,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 225205, "pid": 5, "tid": 7, "ts": 1716454225009815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953639, "dur": 11,
    "args": {
      "External id": 225205,
      "cbid": 211, "correlation": 225205
    }
  },
  {
    "ph": "s", "id": 225205, "pid": 76337, "tid": -914061504, "ts": 1716454224953639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225009825, "dur": 4,
    "args": {
      "External id": 225217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225217,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 225217, "pid": 5, "tid": 7, "ts": 1716454225009825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953661, "dur": 6,
    "args": {
      "External id": 225217,
      "cbid": 211, "correlation": 225217
    }
  },
  {
    "ph": "s", "id": 225217, "pid": 76337, "tid": -914061504, "ts": 1716454224953661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225009830, "dur": 7,
    "args": {
      "External id": 225220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225220,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225220, "pid": 5, "tid": 7, "ts": 1716454225009830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953679, "dur": 6,
    "args": {
      "External id": 225220,
      "cbid": 211, "correlation": 225220
    }
  },
  {
    "ph": "s", "id": 225220, "pid": 76337, "tid": -914061504, "ts": 1716454224953679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225009838, "dur": 4,
    "args": {
      "External id": 225229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225229,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225229, "pid": 5, "tid": 7, "ts": 1716454225009838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953719, "dur": 10,
    "args": {
      "External id": 225229,
      "cbid": 211, "correlation": 225229
    }
  },
  {
    "ph": "s", "id": 225229, "pid": 76337, "tid": -914061504, "ts": 1716454224953719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953781, "dur": 0,
    "args": {
      "External id": 225239,
      "cbid": 317, "correlation": 225239
    }
  },
  {
    "ph": "f", "id": 225239, "pid": 76337, "tid": -914061504, "ts": 1716454224953781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953782, "dur": 0,
    "args": {
      "External id": 225240,
      "cbid": 203, "correlation": 225240
    }
  },
  {
    "ph": "f", "id": 225240, "pid": 76337, "tid": -914061504, "ts": 1716454224953782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953782, "dur": 0,
    "args": {
      "External id": 225241,
      "cbid": 205, "correlation": 225241
    }
  },
  {
    "ph": "f", "id": 225241, "pid": 76337, "tid": -914061504, "ts": 1716454224953782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225009844, "dur": 5,
    "args": {
      "External id": 225245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225245,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225245, "pid": 5, "tid": 7, "ts": 1716454225009844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953796, "dur": 12,
    "args": {
      "External id": 225245,
      "cbid": 211, "correlation": 225245
    }
  },
  {
    "ph": "s", "id": 225245, "pid": 76337, "tid": -914061504, "ts": 1716454224953796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225009850, "dur": 159,
    "args": {
      "External id": 225247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225247,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225247, "pid": 5, "tid": 7, "ts": 1716454225009850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953811, "dur": 5,
    "args": {
      "External id": 225247,
      "cbid": 211, "correlation": 225247
    }
  },
  {
    "ph": "s", "id": 225247, "pid": 76337, "tid": -914061504, "ts": 1716454224953811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225010012, "dur": 1,
    "args": {
      "External id": 225249,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225249,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 225249, "pid": 5, "tid": 7, "ts": 1716454225010012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953822, "dur": 6,
    "args": {
      "External id": 225249,
      "cbid": 51, "correlation": 225249
    }
  },
  {
    "ph": "s", "id": 225249, "pid": 76337, "tid": -914061504, "ts": 1716454224953822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225010016, "dur": 255,
    "args": {
      "External id": 225250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225250,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225250, "pid": 5, "tid": 7, "ts": 1716454225010016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953829, "dur": 6,
    "args": {
      "External id": 225250,
      "cbid": 211, "correlation": 225250
    }
  },
  {
    "ph": "s", "id": 225250, "pid": 76337, "tid": -914061504, "ts": 1716454224953829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225010272, "dur": 5,
    "args": {
      "External id": 225252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225252,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225252, "pid": 5, "tid": 7, "ts": 1716454225010272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953839, "dur": 5,
    "args": {
      "External id": 225252,
      "cbid": 211, "correlation": 225252
    }
  },
  {
    "ph": "s", "id": 225252, "pid": 76337, "tid": -914061504, "ts": 1716454224953839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225010278, "dur": 6,
    "args": {
      "External id": 225258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225258, "pid": 5, "tid": 7, "ts": 1716454225010278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953867, "dur": 8,
    "args": {
      "External id": 225258,
      "cbid": 211, "correlation": 225258
    }
  },
  {
    "ph": "s", "id": 225258, "pid": 76337, "tid": -914061504, "ts": 1716454224953867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953926, "dur": 0,
    "args": {
      "External id": 225268,
      "cbid": 317, "correlation": 225268
    }
  },
  {
    "ph": "f", "id": 225268, "pid": 76337, "tid": -914061504, "ts": 1716454224953926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953927, "dur": 0,
    "args": {
      "External id": 225269,
      "cbid": 203, "correlation": 225269
    }
  },
  {
    "ph": "f", "id": 225269, "pid": 76337, "tid": -914061504, "ts": 1716454224953927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953928, "dur": 0,
    "args": {
      "External id": 225270,
      "cbid": 205, "correlation": 225270
    }
  },
  {
    "ph": "f", "id": 225270, "pid": 76337, "tid": -914061504, "ts": 1716454224953928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225010286, "dur": 7,
    "args": {
      "External id": 225274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225274,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225274, "pid": 5, "tid": 7, "ts": 1716454225010286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953939, "dur": 12,
    "args": {
      "External id": 225274,
      "cbid": 211, "correlation": 225274
    }
  },
  {
    "ph": "s", "id": 225274, "pid": 76337, "tid": -914061504, "ts": 1716454224953939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225010294, "dur": 3,
    "args": {
      "External id": 225276,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225276,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 225276, "pid": 5, "tid": 7, "ts": 1716454225010294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953956, "dur": 10,
    "args": {
      "External id": 225276,
      "cbid": 51, "correlation": 225276
    }
  },
  {
    "ph": "s", "id": 225276, "pid": 76337, "tid": -914061504, "ts": 1716454224953956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225010298, "dur": 93,
    "args": {
      "External id": 225277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225277,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 225277, "pid": 5, "tid": 7, "ts": 1716454225010298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953967, "dur": 13,
    "args": {
      "External id": 225277,
      "cbid": 211, "correlation": 225277
    }
  },
  {
    "ph": "s", "id": 225277, "pid": 76337, "tid": -914061504, "ts": 1716454224953967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225010393, "dur": 6,
    "args": {
      "External id": 225279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225279,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225279, "pid": 5, "tid": 7, "ts": 1716454225010393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224953984, "dur": 6,
    "args": {
      "External id": 225279,
      "cbid": 211, "correlation": 225279
    }
  },
  {
    "ph": "s", "id": 225279, "pid": 76337, "tid": -914061504, "ts": 1716454224953984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225010400, "dur": 6,
    "args": {
      "External id": 225285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225285,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225285, "pid": 5, "tid": 7, "ts": 1716454225010400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954013, "dur": 9,
    "args": {
      "External id": 225285,
      "cbid": 211, "correlation": 225285
    }
  },
  {
    "ph": "s", "id": 225285, "pid": 76337, "tid": -914061504, "ts": 1716454224954013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225010408, "dur": 5,
    "args": {
      "External id": 225293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225293,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225293, "pid": 5, "tid": 7, "ts": 1716454225010408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954043, "dur": 8,
    "args": {
      "External id": 225293,
      "cbid": 211, "correlation": 225293
    }
  },
  {
    "ph": "s", "id": 225293, "pid": 76337, "tid": -914061504, "ts": 1716454224954043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225010413, "dur": 4,
    "args": {
      "External id": 225301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225301,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 225301, "pid": 5, "tid": 7, "ts": 1716454225010413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954072, "dur": 8,
    "args": {
      "External id": 225301,
      "cbid": 211, "correlation": 225301
    }
  },
  {
    "ph": "s", "id": 225301, "pid": 76337, "tid": -914061504, "ts": 1716454224954072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225010419, "dur": 14,
    "args": {
      "External id": 225312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225312,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225312, "pid": 5, "tid": 7, "ts": 1716454225010419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954154, "dur": 13,
    "args": {
      "External id": 225312,
      "cbid": 211, "correlation": 225312
    }
  },
  {
    "ph": "s", "id": 225312, "pid": 76337, "tid": -914061504, "ts": 1716454224954154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954211, "dur": 0,
    "args": {
      "External id": 225322,
      "cbid": 317, "correlation": 225322
    }
  },
  {
    "ph": "f", "id": 225322, "pid": 76337, "tid": -914061504, "ts": 1716454224954211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954212, "dur": 0,
    "args": {
      "External id": 225323,
      "cbid": 203, "correlation": 225323
    }
  },
  {
    "ph": "f", "id": 225323, "pid": 76337, "tid": -914061504, "ts": 1716454224954212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954212, "dur": 0,
    "args": {
      "External id": 225324,
      "cbid": 205, "correlation": 225324
    }
  },
  {
    "ph": "f", "id": 225324, "pid": 76337, "tid": -914061504, "ts": 1716454224954212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225010434, "dur": 8,
    "args": {
      "External id": 225328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225328,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225328, "pid": 5, "tid": 7, "ts": 1716454225010434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954227, "dur": 11,
    "args": {
      "External id": 225328,
      "cbid": 211, "correlation": 225328
    }
  },
  {
    "ph": "s", "id": 225328, "pid": 76337, "tid": -914061504, "ts": 1716454224954227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225010444, "dur": 161,
    "args": {
      "External id": 225330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225330,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225330, "pid": 5, "tid": 7, "ts": 1716454225010444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954240, "dur": 5,
    "args": {
      "External id": 225330,
      "cbid": 211, "correlation": 225330
    }
  },
  {
    "ph": "s", "id": 225330, "pid": 76337, "tid": -914061504, "ts": 1716454224954240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225010607, "dur": 1,
    "args": {
      "External id": 225332,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225332,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 225332, "pid": 5, "tid": 7, "ts": 1716454225010607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954250, "dur": 7,
    "args": {
      "External id": 225332,
      "cbid": 51, "correlation": 225332
    }
  },
  {
    "ph": "s", "id": 225332, "pid": 76337, "tid": -914061504, "ts": 1716454224954250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225010610, "dur": 644,
    "args": {
      "External id": 225333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225333,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225333, "pid": 5, "tid": 7, "ts": 1716454225010610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954259, "dur": 6,
    "args": {
      "External id": 225333,
      "cbid": 211, "correlation": 225333
    }
  },
  {
    "ph": "s", "id": 225333, "pid": 76337, "tid": -914061504, "ts": 1716454224954259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225011256, "dur": 12,
    "args": {
      "External id": 225335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225335,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225335, "pid": 5, "tid": 7, "ts": 1716454225011256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954268, "dur": 5,
    "args": {
      "External id": 225335,
      "cbid": 211, "correlation": 225335
    }
  },
  {
    "ph": "s", "id": 225335, "pid": 76337, "tid": -914061504, "ts": 1716454224954268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225011269, "dur": 14,
    "args": {
      "External id": 225341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225341,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225341, "pid": 5, "tid": 7, "ts": 1716454225011269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954296, "dur": 9,
    "args": {
      "External id": 225341,
      "cbid": 211, "correlation": 225341
    }
  },
  {
    "ph": "s", "id": 225341, "pid": 76337, "tid": -914061504, "ts": 1716454224954296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225011284, "dur": 29,
    "args": {
      "External id": 225350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225350,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225350, "pid": 5, "tid": 7, "ts": 1716454225011284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954386, "dur": 12,
    "args": {
      "External id": 225350,
      "cbid": 211, "correlation": 225350
    }
  },
  {
    "ph": "s", "id": 225350, "pid": 76337, "tid": -914061504, "ts": 1716454224954386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225011314, "dur": 30,
    "args": {
      "External id": 225370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225370,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 225370, "pid": 5, "tid": 7, "ts": 1716454225011314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954455, "dur": 12,
    "args": {
      "External id": 225370,
      "cbid": 211, "correlation": 225370
    }
  },
  {
    "ph": "s", "id": 225370, "pid": 76337, "tid": -914061504, "ts": 1716454224954455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225011346, "dur": 4,
    "args": {
      "External id": 225382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225382,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225382, "pid": 5, "tid": 7, "ts": 1716454225011346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954477, "dur": 6,
    "args": {
      "External id": 225382,
      "cbid": 211, "correlation": 225382
    }
  },
  {
    "ph": "s", "id": 225382, "pid": 76337, "tid": -914061504, "ts": 1716454224954477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225011351, "dur": 30,
    "args": {
      "External id": 225385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225385,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225385, "pid": 5, "tid": 7, "ts": 1716454225011351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954495, "dur": 7,
    "args": {
      "External id": 225385,
      "cbid": 211, "correlation": 225385
    }
  },
  {
    "ph": "s", "id": 225385, "pid": 76337, "tid": -914061504, "ts": 1716454224954495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225011383, "dur": 20,
    "args": {
      "External id": 225394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225394,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225394, "pid": 5, "tid": 7, "ts": 1716454225011383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954534, "dur": 9,
    "args": {
      "External id": 225394,
      "cbid": 211, "correlation": 225394
    }
  },
  {
    "ph": "s", "id": 225394, "pid": 76337, "tid": -914061504, "ts": 1716454224954534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954586, "dur": 0,
    "args": {
      "External id": 225404,
      "cbid": 317, "correlation": 225404
    }
  },
  {
    "ph": "f", "id": 225404, "pid": 76337, "tid": -914061504, "ts": 1716454224954586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954587, "dur": 0,
    "args": {
      "External id": 225405,
      "cbid": 203, "correlation": 225405
    }
  },
  {
    "ph": "f", "id": 225405, "pid": 76337, "tid": -914061504, "ts": 1716454224954587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954587, "dur": 0,
    "args": {
      "External id": 225406,
      "cbid": 205, "correlation": 225406
    }
  },
  {
    "ph": "f", "id": 225406, "pid": 76337, "tid": -914061504, "ts": 1716454224954587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225011404, "dur": 22,
    "args": {
      "External id": 225410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225410, "pid": 5, "tid": 7, "ts": 1716454225011404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954602, "dur": 12,
    "args": {
      "External id": 225410,
      "cbid": 211, "correlation": 225410
    }
  },
  {
    "ph": "s", "id": 225410, "pid": 76337, "tid": -914061504, "ts": 1716454224954602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225011428, "dur": 315,
    "args": {
      "External id": 225412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225412, "pid": 5, "tid": 7, "ts": 1716454225011428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954617, "dur": 5,
    "args": {
      "External id": 225412,
      "cbid": 211, "correlation": 225412
    }
  },
  {
    "ph": "s", "id": 225412, "pid": 76337, "tid": -914061504, "ts": 1716454224954617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225011745, "dur": 1,
    "args": {
      "External id": 225414,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225414,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 225414, "pid": 5, "tid": 7, "ts": 1716454225011745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954628, "dur": 7,
    "args": {
      "External id": 225414,
      "cbid": 51, "correlation": 225414
    }
  },
  {
    "ph": "s", "id": 225414, "pid": 76337, "tid": -914061504, "ts": 1716454224954628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225011749, "dur": 1227,
    "args": {
      "External id": 225415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225415,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225415, "pid": 5, "tid": 7, "ts": 1716454225011749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954636, "dur": 6,
    "args": {
      "External id": 225415,
      "cbid": 211, "correlation": 225415
    }
  },
  {
    "ph": "s", "id": 225415, "pid": 76337, "tid": -914061504, "ts": 1716454224954636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225012977, "dur": 12,
    "args": {
      "External id": 225417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225417,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225417, "pid": 5, "tid": 7, "ts": 1716454225012977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954646, "dur": 5,
    "args": {
      "External id": 225417,
      "cbid": 211, "correlation": 225417
    }
  },
  {
    "ph": "s", "id": 225417, "pid": 76337, "tid": -914061504, "ts": 1716454224954646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225012990, "dur": 15,
    "args": {
      "External id": 225423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225423, "pid": 5, "tid": 7, "ts": 1716454225012990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954674, "dur": 8,
    "args": {
      "External id": 225423,
      "cbid": 211, "correlation": 225423
    }
  },
  {
    "ph": "s", "id": 225423, "pid": 76337, "tid": -914061504, "ts": 1716454224954674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225013006, "dur": 3,
    "args": {
      "External id": 225431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225431,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 225431, "pid": 5, "tid": 7, "ts": 1716454225013006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954718, "dur": 9,
    "args": {
      "External id": 225431,
      "cbid": 211, "correlation": 225431
    }
  },
  {
    "ph": "s", "id": 225431, "pid": 76337, "tid": -914061504, "ts": 1716454224954718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954781, "dur": 1,
    "args": {
      "External id": 225447,
      "cbid": 251, "correlation": 225447
    }
  },
  {
    "ph": "f", "id": 225447, "pid": 76337, "tid": -914061504, "ts": 1716454224954781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954786, "dur": 0,
    "args": {
      "External id": 225449,
      "cbid": 251, "correlation": 225449
    }
  },
  {
    "ph": "f", "id": 225449, "pid": 76337, "tid": -914061504, "ts": 1716454224954786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225013011, "dur": 12,
    "args": {
      "External id": 225450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225450,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225450, "pid": 5, "tid": 7, "ts": 1716454225013011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954788, "dur": 11,
    "args": {
      "External id": 225450,
      "cbid": 211, "correlation": 225450
    }
  },
  {
    "ph": "s", "id": 225450, "pid": 76337, "tid": -914061504, "ts": 1716454224954788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225013024, "dur": 5,
    "args": {
      "External id": 225452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225452,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225452, "pid": 5, "tid": 7, "ts": 1716454225013024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954801, "dur": 5,
    "args": {
      "External id": 225452,
      "cbid": 211, "correlation": 225452
    }
  },
  {
    "ph": "s", "id": 225452, "pid": 76337, "tid": -914061504, "ts": 1716454224954801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225013030, "dur": 17,
    "args": {
      "External id": 225462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225462, "pid": 5, "tid": 7, "ts": 1716454225013030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954857, "dur": 13,
    "args": {
      "External id": 225462,
      "cbid": 211, "correlation": 225462
    }
  },
  {
    "ph": "s", "id": 225462, "pid": 76337, "tid": -914061504, "ts": 1716454224954857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225013049, "dur": 18,
    "args": {
      "External id": 225482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225482,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 225482, "pid": 5, "tid": 7, "ts": 1716454225013049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954924, "dur": 11,
    "args": {
      "External id": 225482,
      "cbid": 211, "correlation": 225482
    }
  },
  {
    "ph": "s", "id": 225482, "pid": 76337, "tid": -914061504, "ts": 1716454224954924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225013068, "dur": 4,
    "args": {
      "External id": 225494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225494,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 225494, "pid": 5, "tid": 7, "ts": 1716454225013068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954944, "dur": 6,
    "args": {
      "External id": 225494,
      "cbid": 211, "correlation": 225494
    }
  },
  {
    "ph": "s", "id": 225494, "pid": 76337, "tid": -914061504, "ts": 1716454224954944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225013073, "dur": 16,
    "args": {
      "External id": 225497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225497, "pid": 5, "tid": 7, "ts": 1716454225013073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224954963, "dur": 7,
    "args": {
      "External id": 225497,
      "cbid": 211, "correlation": 225497
    }
  },
  {
    "ph": "s", "id": 225497, "pid": 76337, "tid": -914061504, "ts": 1716454224954963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225013091, "dur": 11,
    "args": {
      "External id": 225506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225506,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225506, "pid": 5, "tid": 7, "ts": 1716454225013091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955012, "dur": 11,
    "args": {
      "External id": 225506,
      "cbid": 211, "correlation": 225506
    }
  },
  {
    "ph": "s", "id": 225506, "pid": 76337, "tid": -914061504, "ts": 1716454224955012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955076, "dur": 0,
    "args": {
      "External id": 225516,
      "cbid": 317, "correlation": 225516
    }
  },
  {
    "ph": "f", "id": 225516, "pid": 76337, "tid": -914061504, "ts": 1716454224955076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955077, "dur": 0,
    "args": {
      "External id": 225517,
      "cbid": 203, "correlation": 225517
    }
  },
  {
    "ph": "f", "id": 225517, "pid": 76337, "tid": -914061504, "ts": 1716454224955077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955077, "dur": 0,
    "args": {
      "External id": 225518,
      "cbid": 205, "correlation": 225518
    }
  },
  {
    "ph": "f", "id": 225518, "pid": 76337, "tid": -914061504, "ts": 1716454224955077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225013103, "dur": 11,
    "args": {
      "External id": 225522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225522,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225522, "pid": 5, "tid": 7, "ts": 1716454225013103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955090, "dur": 12,
    "args": {
      "External id": 225522,
      "cbid": 211, "correlation": 225522
    }
  },
  {
    "ph": "s", "id": 225522, "pid": 76337, "tid": -914061504, "ts": 1716454224955090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225013115, "dur": 159,
    "args": {
      "External id": 225524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225524, "pid": 5, "tid": 7, "ts": 1716454225013115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955105, "dur": 6,
    "args": {
      "External id": 225524,
      "cbid": 211, "correlation": 225524
    }
  },
  {
    "ph": "s", "id": 225524, "pid": 76337, "tid": -914061504, "ts": 1716454224955105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225013277, "dur": 1,
    "args": {
      "External id": 225526,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225526,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5172413793103449
    }
  },
  {
    "ph": "f", "id": 225526, "pid": 5, "tid": 7, "ts": 1716454225013277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955117, "dur": 6,
    "args": {
      "External id": 225526,
      "cbid": 51, "correlation": 225526
    }
  },
  {
    "ph": "s", "id": 225526, "pid": 76337, "tid": -914061504, "ts": 1716454224955117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225013281, "dur": 638,
    "args": {
      "External id": 225527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225527,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225527, "pid": 5, "tid": 7, "ts": 1716454225013281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955124, "dur": 6,
    "args": {
      "External id": 225527,
      "cbid": 211, "correlation": 225527
    }
  },
  {
    "ph": "s", "id": 225527, "pid": 76337, "tid": -914061504, "ts": 1716454224955124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225013920, "dur": 12,
    "args": {
      "External id": 225529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225529,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225529, "pid": 5, "tid": 7, "ts": 1716454225013920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955134, "dur": 5,
    "args": {
      "External id": 225529,
      "cbid": 211, "correlation": 225529
    }
  },
  {
    "ph": "s", "id": 225529, "pid": 76337, "tid": -914061504, "ts": 1716454224955134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225013933, "dur": 14,
    "args": {
      "External id": 225535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225535, "pid": 5, "tid": 7, "ts": 1716454225013933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955162, "dur": 9,
    "args": {
      "External id": 225535,
      "cbid": 211, "correlation": 225535
    }
  },
  {
    "ph": "s", "id": 225535, "pid": 76337, "tid": -914061504, "ts": 1716454224955162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955222, "dur": 0,
    "args": {
      "External id": 225545,
      "cbid": 317, "correlation": 225545
    }
  },
  {
    "ph": "f", "id": 225545, "pid": 76337, "tid": -914061504, "ts": 1716454224955222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955222, "dur": 0,
    "args": {
      "External id": 225546,
      "cbid": 203, "correlation": 225546
    }
  },
  {
    "ph": "f", "id": 225546, "pid": 76337, "tid": -914061504, "ts": 1716454224955222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955223, "dur": 0,
    "args": {
      "External id": 225547,
      "cbid": 205, "correlation": 225547
    }
  },
  {
    "ph": "f", "id": 225547, "pid": 76337, "tid": -914061504, "ts": 1716454224955223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225013949, "dur": 21,
    "args": {
      "External id": 225551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225551,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225551, "pid": 5, "tid": 7, "ts": 1716454225013949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955236, "dur": 11,
    "args": {
      "External id": 225551,
      "cbid": 211, "correlation": 225551
    }
  },
  {
    "ph": "s", "id": 225551, "pid": 76337, "tid": -914061504, "ts": 1716454224955236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225013971, "dur": 4,
    "args": {
      "External id": 225553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 225553, "pid": 5, "tid": 7, "ts": 1716454225013971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955252, "dur": 8,
    "args": {
      "External id": 225553,
      "cbid": 211, "correlation": 225553
    }
  },
  {
    "ph": "s", "id": 225553, "pid": 76337, "tid": -914061504, "ts": 1716454224955252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955264, "dur": 0,
    "args": {
      "External id": 225554,
      "cbid": 51, "correlation": 225554
    }
  },
  {
    "ph": "s", "id": 225554, "pid": 76337, "tid": -914061504, "ts": 1716454224955264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225013976, "dur": 172,
    "args": {
      "External id": 225555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225555,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 225555, "pid": 5, "tid": 7, "ts": 1716454225013976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955265, "dur": 6,
    "args": {
      "External id": 225555,
      "cbid": 211, "correlation": 225555
    }
  },
  {
    "ph": "s", "id": 225555, "pid": 76337, "tid": -914061504, "ts": 1716454224955265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225014149, "dur": 15,
    "args": {
      "External id": 225560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225560, "pid": 5, "tid": 7, "ts": 1716454225014149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955290, "dur": 8,
    "args": {
      "External id": 225560,
      "cbid": 211, "correlation": 225560
    }
  },
  {
    "ph": "s", "id": 225560, "pid": 76337, "tid": -914061504, "ts": 1716454224955290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225014166, "dur": 12,
    "args": {
      "External id": 225568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225568,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225568, "pid": 5, "tid": 7, "ts": 1716454225014166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955319, "dur": 8,
    "args": {
      "External id": 225568,
      "cbid": 211, "correlation": 225568
    }
  },
  {
    "ph": "s", "id": 225568, "pid": 76337, "tid": -914061504, "ts": 1716454224955319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225014179, "dur": 10,
    "args": {
      "External id": 225576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225576,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225576, "pid": 5, "tid": 7, "ts": 1716454225014179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955347, "dur": 8,
    "args": {
      "External id": 225576,
      "cbid": 211, "correlation": 225576
    }
  },
  {
    "ph": "s", "id": 225576, "pid": 76337, "tid": -914061504, "ts": 1716454224955347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225014190, "dur": 18,
    "args": {
      "External id": 225596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225596,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 225596, "pid": 5, "tid": 7, "ts": 1716454225014190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955428, "dur": 12,
    "args": {
      "External id": 225596,
      "cbid": 211, "correlation": 225596
    }
  },
  {
    "ph": "s", "id": 225596, "pid": 76337, "tid": -914061504, "ts": 1716454224955428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225014209, "dur": 4,
    "args": {
      "External id": 225608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225608,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 225608, "pid": 5, "tid": 7, "ts": 1716454225014209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955450, "dur": 6,
    "args": {
      "External id": 225608,
      "cbid": 211, "correlation": 225608
    }
  },
  {
    "ph": "s", "id": 225608, "pid": 76337, "tid": -914061504, "ts": 1716454224955450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225014215, "dur": 16,
    "args": {
      "External id": 225611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225611, "pid": 5, "tid": 7, "ts": 1716454225014215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955469, "dur": 7,
    "args": {
      "External id": 225611,
      "cbid": 211, "correlation": 225611
    }
  },
  {
    "ph": "s", "id": 225611, "pid": 76337, "tid": -914061504, "ts": 1716454224955469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955527, "dur": 0,
    "args": {
      "External id": 225622,
      "cbid": 317, "correlation": 225622
    }
  },
  {
    "ph": "f", "id": 225622, "pid": 76337, "tid": -914061504, "ts": 1716454224955527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955527, "dur": 0,
    "args": {
      "External id": 225623,
      "cbid": 203, "correlation": 225623
    }
  },
  {
    "ph": "f", "id": 225623, "pid": 76337, "tid": -914061504, "ts": 1716454224955527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955528, "dur": 0,
    "args": {
      "External id": 225624,
      "cbid": 205, "correlation": 225624
    }
  },
  {
    "ph": "f", "id": 225624, "pid": 76337, "tid": -914061504, "ts": 1716454224955528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225014233, "dur": 11,
    "args": {
      "External id": 225628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225628,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225628, "pid": 5, "tid": 7, "ts": 1716454225014233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955541, "dur": 12,
    "args": {
      "External id": 225628,
      "cbid": 211, "correlation": 225628
    }
  },
  {
    "ph": "s", "id": 225628, "pid": 76337, "tid": -914061504, "ts": 1716454224955541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225014245, "dur": 3,
    "args": {
      "External id": 225630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225630,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 225630, "pid": 5, "tid": 7, "ts": 1716454225014245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955557, "dur": 7,
    "args": {
      "External id": 225630,
      "cbid": 211, "correlation": 225630
    }
  },
  {
    "ph": "s", "id": 225630, "pid": 76337, "tid": -914061504, "ts": 1716454224955557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955567, "dur": 0,
    "args": {
      "External id": 225631,
      "cbid": 51, "correlation": 225631
    }
  },
  {
    "ph": "s", "id": 225631, "pid": 76337, "tid": -914061504, "ts": 1716454224955567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225014250, "dur": 89,
    "args": {
      "External id": 225632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225632,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 225632, "pid": 5, "tid": 7, "ts": 1716454225014250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955567, "dur": 5,
    "args": {
      "External id": 225632,
      "cbid": 211, "correlation": 225632
    }
  },
  {
    "ph": "s", "id": 225632, "pid": 76337, "tid": -914061504, "ts": 1716454224955567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225014341, "dur": 15,
    "args": {
      "External id": 225637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225637, "pid": 5, "tid": 7, "ts": 1716454225014341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955594, "dur": 8,
    "args": {
      "External id": 225637,
      "cbid": 211, "correlation": 225637
    }
  },
  {
    "ph": "s", "id": 225637, "pid": 76337, "tid": -914061504, "ts": 1716454224955594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225014357, "dur": 82,
    "args": {
      "External id": 225646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225646,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225646, "pid": 5, "tid": 7, "ts": 1716454225014357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955676, "dur": 14,
    "args": {
      "External id": 225646,
      "cbid": 211, "correlation": 225646
    }
  },
  {
    "ph": "s", "id": 225646, "pid": 76337, "tid": -914061504, "ts": 1716454224955676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225014440, "dur": 30,
    "args": {
      "External id": 225668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225668,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225668, "pid": 5, "tid": 7, "ts": 1716454225014440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955734, "dur": 10,
    "args": {
      "External id": 225668,
      "cbid": 211, "correlation": 225668
    }
  },
  {
    "ph": "s", "id": 225668, "pid": 76337, "tid": -914061504, "ts": 1716454224955734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955825, "dur": 2,
    "args": {
      "External id": 225679,
      "cbid": 251, "correlation": 225679
    }
  },
  {
    "ph": "f", "id": 225679, "pid": 76337, "tid": -914061504, "ts": 1716454224955825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225014472, "dur": 162,
    "args": {
      "External id": 225680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225680,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225680, "pid": 5, "tid": 7, "ts": 1716454225014472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955832, "dur": 13,
    "args": {
      "External id": 225680,
      "cbid": 211, "correlation": 225680
    }
  },
  {
    "ph": "s", "id": 225680, "pid": 76337, "tid": -914061504, "ts": 1716454224955832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955902, "dur": 1,
    "args": {
      "External id": 225691,
      "cbid": 251, "correlation": 225691
    }
  },
  {
    "ph": "f", "id": 225691, "pid": 76337, "tid": -914061504, "ts": 1716454224955902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225014635, "dur": 156,
    "args": {
      "External id": 225692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225692,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225692, "pid": 5, "tid": 7, "ts": 1716454225014635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955906, "dur": 12,
    "args": {
      "External id": 225692,
      "cbid": 211, "correlation": 225692
    }
  },
  {
    "ph": "s", "id": 225692, "pid": 76337, "tid": -914061504, "ts": 1716454224955906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955981, "dur": 1,
    "args": {
      "External id": 225703,
      "cbid": 251, "correlation": 225703
    }
  },
  {
    "ph": "f", "id": 225703, "pid": 76337, "tid": -914061504, "ts": 1716454224955981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225014792, "dur": 156,
    "args": {
      "External id": 225704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225704,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225704, "pid": 5, "tid": 7, "ts": 1716454225014792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224955985, "dur": 12,
    "args": {
      "External id": 225704,
      "cbid": 211, "correlation": 225704
    }
  },
  {
    "ph": "s", "id": 225704, "pid": 76337, "tid": -914061504, "ts": 1716454224955985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225014949, "dur": 330,
    "args": {
      "External id": 225729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225729,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225729, "pid": 5, "tid": 7, "ts": 1716454225014949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956073, "dur": 13,
    "args": {
      "External id": 225729,
      "cbid": 211, "correlation": 225729
    }
  },
  {
    "ph": "s", "id": 225729, "pid": 76337, "tid": -914061504, "ts": 1716454224956073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956175, "dur": 1,
    "args": {
      "External id": 225747,
      "cbid": 251, "correlation": 225747
    }
  },
  {
    "ph": "f", "id": 225747, "pid": 76337, "tid": -914061504, "ts": 1716454224956175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225015280, "dur": 143,
    "args": {
      "External id": 225749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225749,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225749, "pid": 5, "tid": 7, "ts": 1716454225015280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956181, "dur": 13,
    "args": {
      "External id": 225749,
      "cbid": 211, "correlation": 225749
    }
  },
  {
    "ph": "s", "id": 225749, "pid": 76337, "tid": -914061504, "ts": 1716454224956181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225015424, "dur": 19,
    "args": {
      "External id": 225757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225757,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225757, "pid": 5, "tid": 7, "ts": 1716454225015424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956251, "dur": 13,
    "args": {
      "External id": 225757,
      "cbid": 211, "correlation": 225757
    }
  },
  {
    "ph": "s", "id": 225757, "pid": 76337, "tid": -914061504, "ts": 1716454224956251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225015445, "dur": 28,
    "args": {
      "External id": 225765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225765,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225765, "pid": 5, "tid": 7, "ts": 1716454225015445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956290, "dur": 9,
    "args": {
      "External id": 225765,
      "cbid": 211, "correlation": 225765
    }
  },
  {
    "ph": "s", "id": 225765, "pid": 76337, "tid": -914061504, "ts": 1716454224956290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225015474, "dur": 18,
    "args": {
      "External id": 225776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225776, "pid": 5, "tid": 7, "ts": 1716454225015474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956365, "dur": 12,
    "args": {
      "External id": 225776,
      "cbid": 211, "correlation": 225776
    }
  },
  {
    "ph": "s", "id": 225776, "pid": 76337, "tid": -914061504, "ts": 1716454224956365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225015494, "dur": 16,
    "args": {
      "External id": 225798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225798,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225798, "pid": 5, "tid": 7, "ts": 1716454225015494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956395, "dur": 8,
    "args": {
      "External id": 225798,
      "cbid": 211, "correlation": 225798
    }
  },
  {
    "ph": "s", "id": 225798, "pid": 76337, "tid": -914061504, "ts": 1716454224956395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956482, "dur": 2,
    "args": {
      "External id": 225809,
      "cbid": 251, "correlation": 225809
    }
  },
  {
    "ph": "f", "id": 225809, "pid": 76337, "tid": -914061504, "ts": 1716454224956482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225015512, "dur": 87,
    "args": {
      "External id": 225810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225810,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 225810, "pid": 5, "tid": 7, "ts": 1716454225015512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956488, "dur": 13,
    "args": {
      "External id": 225810,
      "cbid": 211, "correlation": 225810
    }
  },
  {
    "ph": "s", "id": 225810, "pid": 76337, "tid": -914061504, "ts": 1716454224956488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956559, "dur": 1,
    "args": {
      "External id": 225821,
      "cbid": 251, "correlation": 225821
    }
  },
  {
    "ph": "f", "id": 225821, "pid": 76337, "tid": -914061504, "ts": 1716454224956559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956563, "dur": 0,
    "args": {
      "External id": 225822,
      "cbid": 251, "correlation": 225822
    }
  },
  {
    "ph": "f", "id": 225822, "pid": 76337, "tid": -914061504, "ts": 1716454224956563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225015601, "dur": 12,
    "args": {
      "External id": 225823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225823,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225823, "pid": 5, "tid": 7, "ts": 1716454225015601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956564, "dur": 12,
    "args": {
      "External id": 225823,
      "cbid": 211, "correlation": 225823
    }
  },
  {
    "ph": "s", "id": 225823, "pid": 76337, "tid": -914061504, "ts": 1716454224956564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225015614, "dur": 5,
    "args": {
      "External id": 225825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225825,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225825, "pid": 5, "tid": 7, "ts": 1716454225015614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956578, "dur": 6,
    "args": {
      "External id": 225825,
      "cbid": 211, "correlation": 225825
    }
  },
  {
    "ph": "s", "id": 225825, "pid": 76337, "tid": -914061504, "ts": 1716454224956578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956636, "dur": 1,
    "args": {
      "External id": 225836,
      "cbid": 251, "correlation": 225836
    }
  },
  {
    "ph": "f", "id": 225836, "pid": 76337, "tid": -914061504, "ts": 1716454224956636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956639, "dur": 0,
    "args": {
      "External id": 225837,
      "cbid": 251, "correlation": 225837
    }
  },
  {
    "ph": "f", "id": 225837, "pid": 76337, "tid": -914061504, "ts": 1716454224956639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225015621, "dur": 8,
    "args": {
      "External id": 225838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225838,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225838, "pid": 5, "tid": 7, "ts": 1716454225015621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956640, "dur": 12,
    "args": {
      "External id": 225838,
      "cbid": 211, "correlation": 225838
    }
  },
  {
    "ph": "s", "id": 225838, "pid": 76337, "tid": -914061504, "ts": 1716454224956640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225015630, "dur": 3,
    "args": {
      "External id": 225840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225840,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225840, "pid": 5, "tid": 7, "ts": 1716454225015630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956654, "dur": 6,
    "args": {
      "External id": 225840,
      "cbid": 211, "correlation": 225840
    }
  },
  {
    "ph": "s", "id": 225840, "pid": 76337, "tid": -914061504, "ts": 1716454224956654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225015635, "dur": 54,
    "args": {
      "External id": 225865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225865,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225865, "pid": 5, "tid": 7, "ts": 1716454225015635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956731, "dur": 12,
    "args": {
      "External id": 225865,
      "cbid": 211, "correlation": 225865
    }
  },
  {
    "ph": "s", "id": 225865, "pid": 76337, "tid": -914061504, "ts": 1716454224956731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956830, "dur": 2,
    "args": {
      "External id": 225883,
      "cbid": 251, "correlation": 225883
    }
  },
  {
    "ph": "f", "id": 225883, "pid": 76337, "tid": -914061504, "ts": 1716454224956830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225015690, "dur": 90,
    "args": {
      "External id": 225885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225885,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 225885, "pid": 5, "tid": 7, "ts": 1716454225015690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956837, "dur": 14,
    "args": {
      "External id": 225885,
      "cbid": 211, "correlation": 225885
    }
  },
  {
    "ph": "s", "id": 225885, "pid": 76337, "tid": -914061504, "ts": 1716454224956837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225015782, "dur": 9,
    "args": {
      "External id": 225893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225893,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225893, "pid": 5, "tid": 7, "ts": 1716454225015782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956906, "dur": 13,
    "args": {
      "External id": 225893,
      "cbid": 211, "correlation": 225893
    }
  },
  {
    "ph": "s", "id": 225893, "pid": 76337, "tid": -914061504, "ts": 1716454224956906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225015793, "dur": 20,
    "args": {
      "External id": 225901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225901, "pid": 5, "tid": 7, "ts": 1716454225015793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224956948, "dur": 9,
    "args": {
      "External id": 225901,
      "cbid": 211, "correlation": 225901
    }
  },
  {
    "ph": "s", "id": 225901, "pid": 76337, "tid": -914061504, "ts": 1716454224956948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225015814, "dur": 17,
    "args": {
      "External id": 225923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225923,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225923, "pid": 5, "tid": 7, "ts": 1716454225015814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957007, "dur": 11,
    "args": {
      "External id": 225923,
      "cbid": 211, "correlation": 225923
    }
  },
  {
    "ph": "s", "id": 225923, "pid": 76337, "tid": -914061504, "ts": 1716454224957007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957096, "dur": 1,
    "args": {
      "External id": 225939,
      "cbid": 251, "correlation": 225939
    }
  },
  {
    "ph": "f", "id": 225939, "pid": 76337, "tid": -914061504, "ts": 1716454224957096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957101, "dur": 0,
    "args": {
      "External id": 225941,
      "cbid": 251, "correlation": 225941
    }
  },
  {
    "ph": "f", "id": 225941, "pid": 76337, "tid": -914061504, "ts": 1716454224957101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225015833, "dur": 493,
    "args": {
      "External id": 225942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225942,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 225942, "pid": 5, "tid": 7, "ts": 1716454225015833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957103, "dur": 13,
    "args": {
      "External id": 225942,
      "cbid": 211, "correlation": 225942
    }
  },
  {
    "ph": "s", "id": 225942, "pid": 76337, "tid": -914061504, "ts": 1716454224957103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225016327, "dur": 65,
    "args": {
      "External id": 225950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225950, "pid": 5, "tid": 7, "ts": 1716454225016327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957172, "dur": 12,
    "args": {
      "External id": 225950,
      "cbid": 211, "correlation": 225950
    }
  },
  {
    "ph": "s", "id": 225950, "pid": 76337, "tid": -914061504, "ts": 1716454224957172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225016394, "dur": 65,
    "args": {
      "External id": 225958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225958, "pid": 5, "tid": 7, "ts": 1716454225016394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957204, "dur": 9,
    "args": {
      "External id": 225958,
      "cbid": 211, "correlation": 225958
    }
  },
  {
    "ph": "s", "id": 225958, "pid": 76337, "tid": -914061504, "ts": 1716454224957204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957286, "dur": 1,
    "args": {
      "External id": 225974,
      "cbid": 251, "correlation": 225974
    }
  },
  {
    "ph": "f", "id": 225974, "pid": 76337, "tid": -914061504, "ts": 1716454224957286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225016461, "dur": 1,
    "args": {
      "External id": 225976,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 225976,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 225976, "pid": 5, "tid": 7, "ts": 1716454225016461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957291, "dur": 10,
    "args": {
      "External id": 225976,
      "cbid": 51, "correlation": 225976
    }
  },
  {
    "ph": "s", "id": 225976, "pid": 76337, "tid": -914061504, "ts": 1716454224957291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225016465, "dur": 267,
    "args": {
      "External id": 225977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225977,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 225977, "pid": 5, "tid": 7, "ts": 1716454225016465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957303, "dur": 12,
    "args": {
      "External id": 225977,
      "cbid": 211, "correlation": 225977
    }
  },
  {
    "ph": "s", "id": 225977, "pid": 76337, "tid": -914061504, "ts": 1716454224957303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225016733, "dur": 14,
    "args": {
      "External id": 225985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225985,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225985, "pid": 5, "tid": 7, "ts": 1716454225016733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957346, "dur": 10,
    "args": {
      "External id": 225985,
      "cbid": 211, "correlation": 225985
    }
  },
  {
    "ph": "s", "id": 225985, "pid": 76337, "tid": -914061504, "ts": 1716454224957346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225016748, "dur": 37,
    "args": {
      "External id": 225996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 225996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 225996, "pid": 5, "tid": 7, "ts": 1716454225016748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957416, "dur": 12,
    "args": {
      "External id": 225996,
      "cbid": 211, "correlation": 225996
    }
  },
  {
    "ph": "s", "id": 225996, "pid": 76337, "tid": -914061504, "ts": 1716454224957416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957480, "dur": 0,
    "args": {
      "External id": 226008,
      "cbid": 317, "correlation": 226008
    }
  },
  {
    "ph": "f", "id": 226008, "pid": 76337, "tid": -914061504, "ts": 1716454224957480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957481, "dur": 0,
    "args": {
      "External id": 226009,
      "cbid": 203, "correlation": 226009
    }
  },
  {
    "ph": "f", "id": 226009, "pid": 76337, "tid": -914061504, "ts": 1716454224957481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957482, "dur": 0,
    "args": {
      "External id": 226010,
      "cbid": 205, "correlation": 226010
    }
  },
  {
    "ph": "f", "id": 226010, "pid": 76337, "tid": -914061504, "ts": 1716454224957482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225016787, "dur": 14,
    "args": {
      "External id": 226014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226014, "pid": 5, "tid": 7, "ts": 1716454225016787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957497, "dur": 13,
    "args": {
      "External id": 226014,
      "cbid": 211, "correlation": 226014
    }
  },
  {
    "ph": "s", "id": 226014, "pid": 76337, "tid": -914061504, "ts": 1716454224957497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225016803, "dur": 4,
    "args": {
      "External id": 226016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226016, "pid": 5, "tid": 7, "ts": 1716454225016803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957515, "dur": 6,
    "args": {
      "External id": 226016,
      "cbid": 211, "correlation": 226016
    }
  },
  {
    "ph": "s", "id": 226016, "pid": 76337, "tid": -914061504, "ts": 1716454224957515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957523, "dur": 0,
    "args": {
      "External id": 226017,
      "cbid": 51, "correlation": 226017
    }
  },
  {
    "ph": "s", "id": 226017, "pid": 76337, "tid": -914061504, "ts": 1716454224957523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225016808, "dur": 96,
    "args": {
      "External id": 226018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226018,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 226018, "pid": 5, "tid": 7, "ts": 1716454225016808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957524, "dur": 5,
    "args": {
      "External id": 226018,
      "cbid": 211, "correlation": 226018
    }
  },
  {
    "ph": "s", "id": 226018, "pid": 76337, "tid": -914061504, "ts": 1716454224957524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225016905, "dur": 16,
    "args": {
      "External id": 226023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226023, "pid": 5, "tid": 7, "ts": 1716454225016905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957551, "dur": 10,
    "args": {
      "External id": 226023,
      "cbid": 211, "correlation": 226023
    }
  },
  {
    "ph": "s", "id": 226023, "pid": 76337, "tid": -914061504, "ts": 1716454224957551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225016922, "dur": 11,
    "args": {
      "External id": 226031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226031,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226031, "pid": 5, "tid": 7, "ts": 1716454225016922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957583, "dur": 8,
    "args": {
      "External id": 226031,
      "cbid": 211, "correlation": 226031
    }
  },
  {
    "ph": "s", "id": 226031, "pid": 76337, "tid": -914061504, "ts": 1716454224957583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225016934, "dur": 29,
    "args": {
      "External id": 226040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226040,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226040, "pid": 5, "tid": 7, "ts": 1716454225016934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957623, "dur": 10,
    "args": {
      "External id": 226040,
      "cbid": 211, "correlation": 226040
    }
  },
  {
    "ph": "s", "id": 226040, "pid": 76337, "tid": -914061504, "ts": 1716454224957623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225016965, "dur": 31,
    "args": {
      "External id": 226060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226060,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 226060, "pid": 5, "tid": 7, "ts": 1716454225016965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957694, "dur": 12,
    "args": {
      "External id": 226060,
      "cbid": 211, "correlation": 226060
    }
  },
  {
    "ph": "s", "id": 226060, "pid": 76337, "tid": -914061504, "ts": 1716454224957694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225016997, "dur": 6,
    "args": {
      "External id": 226072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226072,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226072, "pid": 5, "tid": 7, "ts": 1716454225016997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957716, "dur": 6,
    "args": {
      "External id": 226072,
      "cbid": 211, "correlation": 226072
    }
  },
  {
    "ph": "s", "id": 226072, "pid": 76337, "tid": -914061504, "ts": 1716454224957716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225017004, "dur": 31,
    "args": {
      "External id": 226075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226075,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226075, "pid": 5, "tid": 7, "ts": 1716454225017004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957735, "dur": 7,
    "args": {
      "External id": 226075,
      "cbid": 211, "correlation": 226075
    }
  },
  {
    "ph": "s", "id": 226075, "pid": 76337, "tid": -914061504, "ts": 1716454224957735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225017036, "dur": 21,
    "args": {
      "External id": 226084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226084,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226084, "pid": 5, "tid": 7, "ts": 1716454225017036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957774, "dur": 10,
    "args": {
      "External id": 226084,
      "cbid": 211, "correlation": 226084
    }
  },
  {
    "ph": "s", "id": 226084, "pid": 76337, "tid": -914061504, "ts": 1716454224957774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957826, "dur": 0,
    "args": {
      "External id": 226094,
      "cbid": 317, "correlation": 226094
    }
  },
  {
    "ph": "f", "id": 226094, "pid": 76337, "tid": -914061504, "ts": 1716454224957826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957827, "dur": 0,
    "args": {
      "External id": 226095,
      "cbid": 203, "correlation": 226095
    }
  },
  {
    "ph": "f", "id": 226095, "pid": 76337, "tid": -914061504, "ts": 1716454224957827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957828, "dur": 0,
    "args": {
      "External id": 226096,
      "cbid": 205, "correlation": 226096
    }
  },
  {
    "ph": "f", "id": 226096, "pid": 76337, "tid": -914061504, "ts": 1716454224957828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225017059, "dur": 21,
    "args": {
      "External id": 226100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226100, "pid": 5, "tid": 7, "ts": 1716454225017059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957841, "dur": 11,
    "args": {
      "External id": 226100,
      "cbid": 211, "correlation": 226100
    }
  },
  {
    "ph": "s", "id": 226100, "pid": 76337, "tid": -914061504, "ts": 1716454224957841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225017081, "dur": 313,
    "args": {
      "External id": 226102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226102,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226102, "pid": 5, "tid": 7, "ts": 1716454225017081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957855, "dur": 5,
    "args": {
      "External id": 226102,
      "cbid": 211, "correlation": 226102
    }
  },
  {
    "ph": "s", "id": 226102, "pid": 76337, "tid": -914061504, "ts": 1716454224957855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225017397, "dur": 1,
    "args": {
      "External id": 226104,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 226104,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 226104, "pid": 5, "tid": 7, "ts": 1716454225017397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957866, "dur": 6,
    "args": {
      "External id": 226104,
      "cbid": 51, "correlation": 226104
    }
  },
  {
    "ph": "s", "id": 226104, "pid": 76337, "tid": -914061504, "ts": 1716454224957866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225017400, "dur": 1241,
    "args": {
      "External id": 226105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226105,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226105, "pid": 5, "tid": 7, "ts": 1716454225017400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957874, "dur": 6,
    "args": {
      "External id": 226105,
      "cbid": 211, "correlation": 226105
    }
  },
  {
    "ph": "s", "id": 226105, "pid": 76337, "tid": -914061504, "ts": 1716454224957874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225018643, "dur": 14,
    "args": {
      "External id": 226107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226107,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226107, "pid": 5, "tid": 7, "ts": 1716454225018643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957884, "dur": 5,
    "args": {
      "External id": 226107,
      "cbid": 211, "correlation": 226107
    }
  },
  {
    "ph": "s", "id": 226107, "pid": 76337, "tid": -914061504, "ts": 1716454224957884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225018658, "dur": 15,
    "args": {
      "External id": 226113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226113,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226113, "pid": 5, "tid": 7, "ts": 1716454225018658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957913, "dur": 9,
    "args": {
      "External id": 226113,
      "cbid": 211, "correlation": 226113
    }
  },
  {
    "ph": "s", "id": 226113, "pid": 76337, "tid": -914061504, "ts": 1716454224957913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225018674, "dur": 3,
    "args": {
      "External id": 226121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226121,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 226121, "pid": 5, "tid": 7, "ts": 1716454225018674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224957956, "dur": 10,
    "args": {
      "External id": 226121,
      "cbid": 211, "correlation": 226121
    }
  },
  {
    "ph": "s", "id": 226121, "pid": 76337, "tid": -914061504, "ts": 1716454224957956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958031, "dur": 1,
    "args": {
      "External id": 226137,
      "cbid": 251, "correlation": 226137
    }
  },
  {
    "ph": "f", "id": 226137, "pid": 76337, "tid": -914061504, "ts": 1716454224958031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958036, "dur": 0,
    "args": {
      "External id": 226139,
      "cbid": 251, "correlation": 226139
    }
  },
  {
    "ph": "f", "id": 226139, "pid": 76337, "tid": -914061504, "ts": 1716454224958036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225018678, "dur": 13,
    "args": {
      "External id": 226140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226140,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226140, "pid": 5, "tid": 7, "ts": 1716454225018678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958038, "dur": 12,
    "args": {
      "External id": 226140,
      "cbid": 211, "correlation": 226140
    }
  },
  {
    "ph": "s", "id": 226140, "pid": 76337, "tid": -914061504, "ts": 1716454224958038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225018693, "dur": 5,
    "args": {
      "External id": 226142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226142,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226142, "pid": 5, "tid": 7, "ts": 1716454225018693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958052, "dur": 5,
    "args": {
      "External id": 226142,
      "cbid": 211, "correlation": 226142
    }
  },
  {
    "ph": "s", "id": 226142, "pid": 76337, "tid": -914061504, "ts": 1716454224958052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225018699, "dur": 17,
    "args": {
      "External id": 226152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226152,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226152, "pid": 5, "tid": 7, "ts": 1716454225018699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958110, "dur": 12,
    "args": {
      "External id": 226152,
      "cbid": 211, "correlation": 226152
    }
  },
  {
    "ph": "s", "id": 226152, "pid": 76337, "tid": -914061504, "ts": 1716454224958110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225018717, "dur": 17,
    "args": {
      "External id": 226172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226172,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 226172, "pid": 5, "tid": 7, "ts": 1716454225018717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958176, "dur": 11,
    "args": {
      "External id": 226172,
      "cbid": 211, "correlation": 226172
    }
  },
  {
    "ph": "s", "id": 226172, "pid": 76337, "tid": -914061504, "ts": 1716454224958176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225018735, "dur": 4,
    "args": {
      "External id": 226184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226184,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 226184, "pid": 5, "tid": 7, "ts": 1716454225018735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958196, "dur": 6,
    "args": {
      "External id": 226184,
      "cbid": 211, "correlation": 226184
    }
  },
  {
    "ph": "s", "id": 226184, "pid": 76337, "tid": -914061504, "ts": 1716454224958196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225018741, "dur": 17,
    "args": {
      "External id": 226187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226187,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226187, "pid": 5, "tid": 7, "ts": 1716454225018741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958215, "dur": 7,
    "args": {
      "External id": 226187,
      "cbid": 211, "correlation": 226187
    }
  },
  {
    "ph": "s", "id": 226187, "pid": 76337, "tid": -914061504, "ts": 1716454224958215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225018759, "dur": 11,
    "args": {
      "External id": 226196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226196,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226196, "pid": 5, "tid": 7, "ts": 1716454225018759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958255, "dur": 11,
    "args": {
      "External id": 226196,
      "cbid": 211, "correlation": 226196
    }
  },
  {
    "ph": "s", "id": 226196, "pid": 76337, "tid": -914061504, "ts": 1716454224958255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958319, "dur": 0,
    "args": {
      "External id": 226206,
      "cbid": 317, "correlation": 226206
    }
  },
  {
    "ph": "f", "id": 226206, "pid": 76337, "tid": -914061504, "ts": 1716454224958319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958320, "dur": 0,
    "args": {
      "External id": 226207,
      "cbid": 203, "correlation": 226207
    }
  },
  {
    "ph": "f", "id": 226207, "pid": 76337, "tid": -914061504, "ts": 1716454224958320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958321, "dur": 0,
    "args": {
      "External id": 226208,
      "cbid": 205, "correlation": 226208
    }
  },
  {
    "ph": "f", "id": 226208, "pid": 76337, "tid": -914061504, "ts": 1716454224958321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225018771, "dur": 11,
    "args": {
      "External id": 226212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226212,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226212, "pid": 5, "tid": 7, "ts": 1716454225018771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958334, "dur": 12,
    "args": {
      "External id": 226212,
      "cbid": 211, "correlation": 226212
    }
  },
  {
    "ph": "s", "id": 226212, "pid": 76337, "tid": -914061504, "ts": 1716454224958334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225018783, "dur": 160,
    "args": {
      "External id": 226214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226214,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226214, "pid": 5, "tid": 7, "ts": 1716454225018783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958349, "dur": 5,
    "args": {
      "External id": 226214,
      "cbid": 211, "correlation": 226214
    }
  },
  {
    "ph": "s", "id": 226214, "pid": 76337, "tid": -914061504, "ts": 1716454224958349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225018945, "dur": 1,
    "args": {
      "External id": 226216,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 226216,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 226216, "pid": 5, "tid": 7, "ts": 1716454225018945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958360, "dur": 7,
    "args": {
      "External id": 226216,
      "cbid": 51, "correlation": 226216
    }
  },
  {
    "ph": "s", "id": 226216, "pid": 76337, "tid": -914061504, "ts": 1716454224958360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225018949, "dur": 639,
    "args": {
      "External id": 226217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226217,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226217, "pid": 5, "tid": 7, "ts": 1716454225018949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958368, "dur": 6,
    "args": {
      "External id": 226217,
      "cbid": 211, "correlation": 226217
    }
  },
  {
    "ph": "s", "id": 226217, "pid": 76337, "tid": -914061504, "ts": 1716454224958368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225019589, "dur": 12,
    "args": {
      "External id": 226219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226219,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226219, "pid": 5, "tid": 7, "ts": 1716454225019589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958378, "dur": 5,
    "args": {
      "External id": 226219,
      "cbid": 211, "correlation": 226219
    }
  },
  {
    "ph": "s", "id": 226219, "pid": 76337, "tid": -914061504, "ts": 1716454224958378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225019603, "dur": 14,
    "args": {
      "External id": 226225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226225,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226225, "pid": 5, "tid": 7, "ts": 1716454225019603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958406, "dur": 9,
    "args": {
      "External id": 226225,
      "cbid": 211, "correlation": 226225
    }
  },
  {
    "ph": "s", "id": 226225, "pid": 76337, "tid": -914061504, "ts": 1716454224958406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958465, "dur": 0,
    "args": {
      "External id": 226235,
      "cbid": 317, "correlation": 226235
    }
  },
  {
    "ph": "f", "id": 226235, "pid": 76337, "tid": -914061504, "ts": 1716454224958465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958466, "dur": 0,
    "args": {
      "External id": 226236,
      "cbid": 203, "correlation": 226236
    }
  },
  {
    "ph": "f", "id": 226236, "pid": 76337, "tid": -914061504, "ts": 1716454224958466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958467, "dur": 0,
    "args": {
      "External id": 226237,
      "cbid": 205, "correlation": 226237
    }
  },
  {
    "ph": "f", "id": 226237, "pid": 76337, "tid": -914061504, "ts": 1716454224958467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225019618, "dur": 21,
    "args": {
      "External id": 226241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226241,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226241, "pid": 5, "tid": 7, "ts": 1716454225019618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958478, "dur": 12,
    "args": {
      "External id": 226241,
      "cbid": 211, "correlation": 226241
    }
  },
  {
    "ph": "s", "id": 226241, "pid": 76337, "tid": -914061504, "ts": 1716454224958478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225019640, "dur": 4,
    "args": {
      "External id": 226243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226243, "pid": 5, "tid": 7, "ts": 1716454225019640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958494, "dur": 6,
    "args": {
      "External id": 226243,
      "cbid": 211, "correlation": 226243
    }
  },
  {
    "ph": "s", "id": 226243, "pid": 76337, "tid": -914061504, "ts": 1716454224958494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958502, "dur": 0,
    "args": {
      "External id": 226244,
      "cbid": 51, "correlation": 226244
    }
  },
  {
    "ph": "s", "id": 226244, "pid": 76337, "tid": -914061504, "ts": 1716454224958502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225019645, "dur": 167,
    "args": {
      "External id": 226245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226245,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 226245, "pid": 5, "tid": 7, "ts": 1716454225019645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958503, "dur": 6,
    "args": {
      "External id": 226245,
      "cbid": 211, "correlation": 226245
    }
  },
  {
    "ph": "s", "id": 226245, "pid": 76337, "tid": -914061504, "ts": 1716454224958503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225019814, "dur": 16,
    "args": {
      "External id": 226250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226250,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226250, "pid": 5, "tid": 7, "ts": 1716454225019814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958529, "dur": 8,
    "args": {
      "External id": 226250,
      "cbid": 211, "correlation": 226250
    }
  },
  {
    "ph": "s", "id": 226250, "pid": 76337, "tid": -914061504, "ts": 1716454224958529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225019831, "dur": 12,
    "args": {
      "External id": 226258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226258,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226258, "pid": 5, "tid": 7, "ts": 1716454225019831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958558, "dur": 8,
    "args": {
      "External id": 226258,
      "cbid": 211, "correlation": 226258
    }
  },
  {
    "ph": "s", "id": 226258, "pid": 76337, "tid": -914061504, "ts": 1716454224958558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225019845, "dur": 10,
    "args": {
      "External id": 226266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226266,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226266, "pid": 5, "tid": 7, "ts": 1716454225019845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958586, "dur": 8,
    "args": {
      "External id": 226266,
      "cbid": 211, "correlation": 226266
    }
  },
  {
    "ph": "s", "id": 226266, "pid": 76337, "tid": -914061504, "ts": 1716454224958586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225019856, "dur": 18,
    "args": {
      "External id": 226286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226286,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 226286, "pid": 5, "tid": 7, "ts": 1716454225019856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958669, "dur": 12,
    "args": {
      "External id": 226286,
      "cbid": 211, "correlation": 226286
    }
  },
  {
    "ph": "s", "id": 226286, "pid": 76337, "tid": -914061504, "ts": 1716454224958669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225019876, "dur": 4,
    "args": {
      "External id": 226298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226298,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 226298, "pid": 5, "tid": 7, "ts": 1716454225019876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958691, "dur": 6,
    "args": {
      "External id": 226298,
      "cbid": 211, "correlation": 226298
    }
  },
  {
    "ph": "s", "id": 226298, "pid": 76337, "tid": -914061504, "ts": 1716454224958691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225019881, "dur": 17,
    "args": {
      "External id": 226301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226301, "pid": 5, "tid": 7, "ts": 1716454225019881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958709, "dur": 7,
    "args": {
      "External id": 226301,
      "cbid": 211, "correlation": 226301
    }
  },
  {
    "ph": "s", "id": 226301, "pid": 76337, "tid": -914061504, "ts": 1716454224958709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958767, "dur": 0,
    "args": {
      "External id": 226312,
      "cbid": 317, "correlation": 226312
    }
  },
  {
    "ph": "f", "id": 226312, "pid": 76337, "tid": -914061504, "ts": 1716454224958767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958768, "dur": 0,
    "args": {
      "External id": 226313,
      "cbid": 203, "correlation": 226313
    }
  },
  {
    "ph": "f", "id": 226313, "pid": 76337, "tid": -914061504, "ts": 1716454224958768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958769, "dur": 0,
    "args": {
      "External id": 226314,
      "cbid": 205, "correlation": 226314
    }
  },
  {
    "ph": "f", "id": 226314, "pid": 76337, "tid": -914061504, "ts": 1716454224958769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225019899, "dur": 12,
    "args": {
      "External id": 226318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226318, "pid": 5, "tid": 7, "ts": 1716454225019899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958781, "dur": 12,
    "args": {
      "External id": 226318,
      "cbid": 211, "correlation": 226318
    }
  },
  {
    "ph": "s", "id": 226318, "pid": 76337, "tid": -914061504, "ts": 1716454224958781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225019912, "dur": 3,
    "args": {
      "External id": 226320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226320, "pid": 5, "tid": 7, "ts": 1716454225019912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958798, "dur": 6,
    "args": {
      "External id": 226320,
      "cbid": 211, "correlation": 226320
    }
  },
  {
    "ph": "s", "id": 226320, "pid": 76337, "tid": -914061504, "ts": 1716454224958798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958806, "dur": 0,
    "args": {
      "External id": 226321,
      "cbid": 51, "correlation": 226321
    }
  },
  {
    "ph": "s", "id": 226321, "pid": 76337, "tid": -914061504, "ts": 1716454224958806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225019917, "dur": 89,
    "args": {
      "External id": 226322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226322,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 226322, "pid": 5, "tid": 7, "ts": 1716454225019917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958807, "dur": 5,
    "args": {
      "External id": 226322,
      "cbid": 211, "correlation": 226322
    }
  },
  {
    "ph": "s", "id": 226322, "pid": 76337, "tid": -914061504, "ts": 1716454224958807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225020007, "dur": 16,
    "args": {
      "External id": 226327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226327, "pid": 5, "tid": 7, "ts": 1716454225020007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958835, "dur": 9,
    "args": {
      "External id": 226327,
      "cbid": 211, "correlation": 226327
    }
  },
  {
    "ph": "s", "id": 226327, "pid": 76337, "tid": -914061504, "ts": 1716454224958835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225020023, "dur": 82,
    "args": {
      "External id": 226336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226336,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226336, "pid": 5, "tid": 7, "ts": 1716454225020023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958915, "dur": 15,
    "args": {
      "External id": 226336,
      "cbid": 211, "correlation": 226336
    }
  },
  {
    "ph": "s", "id": 226336, "pid": 76337, "tid": -914061504, "ts": 1716454224958915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225020107, "dur": 29,
    "args": {
      "External id": 226358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226358,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226358, "pid": 5, "tid": 7, "ts": 1716454225020107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224958982, "dur": 10,
    "args": {
      "External id": 226358,
      "cbid": 211, "correlation": 226358
    }
  },
  {
    "ph": "s", "id": 226358, "pid": 76337, "tid": -914061504, "ts": 1716454224958982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959072, "dur": 2,
    "args": {
      "External id": 226369,
      "cbid": 251, "correlation": 226369
    }
  },
  {
    "ph": "f", "id": 226369, "pid": 76337, "tid": -914061504, "ts": 1716454224959072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225020138, "dur": 164,
    "args": {
      "External id": 226370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226370,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226370, "pid": 5, "tid": 7, "ts": 1716454225020138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959078, "dur": 13,
    "args": {
      "External id": 226370,
      "cbid": 211, "correlation": 226370
    }
  },
  {
    "ph": "s", "id": 226370, "pid": 76337, "tid": -914061504, "ts": 1716454224959078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959148, "dur": 1,
    "args": {
      "External id": 226381,
      "cbid": 251, "correlation": 226381
    }
  },
  {
    "ph": "f", "id": 226381, "pid": 76337, "tid": -914061504, "ts": 1716454224959148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225020303, "dur": 156,
    "args": {
      "External id": 226382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226382,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226382, "pid": 5, "tid": 7, "ts": 1716454225020303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959152, "dur": 11,
    "args": {
      "External id": 226382,
      "cbid": 211, "correlation": 226382
    }
  },
  {
    "ph": "s", "id": 226382, "pid": 76337, "tid": -914061504, "ts": 1716454224959152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959217, "dur": 1,
    "args": {
      "External id": 226393,
      "cbid": 251, "correlation": 226393
    }
  },
  {
    "ph": "f", "id": 226393, "pid": 76337, "tid": -914061504, "ts": 1716454224959217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225020460, "dur": 156,
    "args": {
      "External id": 226394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226394,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226394, "pid": 5, "tid": 7, "ts": 1716454225020460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959221, "dur": 11,
    "args": {
      "External id": 226394,
      "cbid": 211, "correlation": 226394
    }
  },
  {
    "ph": "s", "id": 226394, "pid": 76337, "tid": -914061504, "ts": 1716454224959221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225020617, "dur": 336,
    "args": {
      "External id": 226419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226419,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226419, "pid": 5, "tid": 7, "ts": 1716454225020617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959305, "dur": 13,
    "args": {
      "External id": 226419,
      "cbid": 211, "correlation": 226419
    }
  },
  {
    "ph": "s", "id": 226419, "pid": 76337, "tid": -914061504, "ts": 1716454224959305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959407, "dur": 1,
    "args": {
      "External id": 226437,
      "cbid": 251, "correlation": 226437
    }
  },
  {
    "ph": "f", "id": 226437, "pid": 76337, "tid": -914061504, "ts": 1716454224959407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225020954, "dur": 164,
    "args": {
      "External id": 226439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226439,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226439, "pid": 5, "tid": 7, "ts": 1716454225020954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959413, "dur": 14,
    "args": {
      "External id": 226439,
      "cbid": 211, "correlation": 226439
    }
  },
  {
    "ph": "s", "id": 226439, "pid": 76337, "tid": -914061504, "ts": 1716454224959413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225021119, "dur": 19,
    "args": {
      "External id": 226447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226447,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226447, "pid": 5, "tid": 7, "ts": 1716454225021119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959484, "dur": 12,
    "args": {
      "External id": 226447,
      "cbid": 211, "correlation": 226447
    }
  },
  {
    "ph": "s", "id": 226447, "pid": 76337, "tid": -914061504, "ts": 1716454224959484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225021140, "dur": 28,
    "args": {
      "External id": 226455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226455,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226455, "pid": 5, "tid": 7, "ts": 1716454225021140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959523, "dur": 9,
    "args": {
      "External id": 226455,
      "cbid": 211, "correlation": 226455
    }
  },
  {
    "ph": "s", "id": 226455, "pid": 76337, "tid": -914061504, "ts": 1716454224959523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225021169, "dur": 18,
    "args": {
      "External id": 226466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226466, "pid": 5, "tid": 7, "ts": 1716454225021169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959594, "dur": 12,
    "args": {
      "External id": 226466,
      "cbid": 211, "correlation": 226466
    }
  },
  {
    "ph": "s", "id": 226466, "pid": 76337, "tid": -914061504, "ts": 1716454224959594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225021188, "dur": 16,
    "args": {
      "External id": 226488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226488,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226488, "pid": 5, "tid": 7, "ts": 1716454225021188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959624, "dur": 8,
    "args": {
      "External id": 226488,
      "cbid": 211, "correlation": 226488
    }
  },
  {
    "ph": "s", "id": 226488, "pid": 76337, "tid": -914061504, "ts": 1716454224959624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959710, "dur": 1,
    "args": {
      "External id": 226499,
      "cbid": 251, "correlation": 226499
    }
  },
  {
    "ph": "f", "id": 226499, "pid": 76337, "tid": -914061504, "ts": 1716454224959710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225021206, "dur": 88,
    "args": {
      "External id": 226500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226500,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226500, "pid": 5, "tid": 7, "ts": 1716454225021206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959715, "dur": 13,
    "args": {
      "External id": 226500,
      "cbid": 211, "correlation": 226500
    }
  },
  {
    "ph": "s", "id": 226500, "pid": 76337, "tid": -914061504, "ts": 1716454224959715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959782, "dur": 1,
    "args": {
      "External id": 226511,
      "cbid": 251, "correlation": 226511
    }
  },
  {
    "ph": "f", "id": 226511, "pid": 76337, "tid": -914061504, "ts": 1716454224959782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959787, "dur": 0,
    "args": {
      "External id": 226512,
      "cbid": 251, "correlation": 226512
    }
  },
  {
    "ph": "f", "id": 226512, "pid": 76337, "tid": -914061504, "ts": 1716454224959787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225021295, "dur": 13,
    "args": {
      "External id": 226513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226513,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226513, "pid": 5, "tid": 7, "ts": 1716454225021295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959788, "dur": 12,
    "args": {
      "External id": 226513,
      "cbid": 211, "correlation": 226513
    }
  },
  {
    "ph": "s", "id": 226513, "pid": 76337, "tid": -914061504, "ts": 1716454224959788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225021309, "dur": 5,
    "args": {
      "External id": 226515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226515,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226515, "pid": 5, "tid": 7, "ts": 1716454225021309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959802, "dur": 6,
    "args": {
      "External id": 226515,
      "cbid": 211, "correlation": 226515
    }
  },
  {
    "ph": "s", "id": 226515, "pid": 76337, "tid": -914061504, "ts": 1716454224959802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959859, "dur": 1,
    "args": {
      "External id": 226526,
      "cbid": 251, "correlation": 226526
    }
  },
  {
    "ph": "f", "id": 226526, "pid": 76337, "tid": -914061504, "ts": 1716454224959859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959862, "dur": 0,
    "args": {
      "External id": 226527,
      "cbid": 251, "correlation": 226527
    }
  },
  {
    "ph": "f", "id": 226527, "pid": 76337, "tid": -914061504, "ts": 1716454224959862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225021315, "dur": 8,
    "args": {
      "External id": 226528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226528,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226528, "pid": 5, "tid": 7, "ts": 1716454225021315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959864, "dur": 12,
    "args": {
      "External id": 226528,
      "cbid": 211, "correlation": 226528
    }
  },
  {
    "ph": "s", "id": 226528, "pid": 76337, "tid": -914061504, "ts": 1716454224959864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225021324, "dur": 4,
    "args": {
      "External id": 226530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226530,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226530, "pid": 5, "tid": 7, "ts": 1716454225021324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959878, "dur": 5,
    "args": {
      "External id": 226530,
      "cbid": 211, "correlation": 226530
    }
  },
  {
    "ph": "s", "id": 226530, "pid": 76337, "tid": -914061504, "ts": 1716454224959878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225021329, "dur": 54,
    "args": {
      "External id": 226555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226555,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226555, "pid": 5, "tid": 7, "ts": 1716454225021329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224959954, "dur": 12,
    "args": {
      "External id": 226555,
      "cbid": 211, "correlation": 226555
    }
  },
  {
    "ph": "s", "id": 226555, "pid": 76337, "tid": -914061504, "ts": 1716454224959954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960062, "dur": 1,
    "args": {
      "External id": 226573,
      "cbid": 251, "correlation": 226573
    }
  },
  {
    "ph": "f", "id": 226573, "pid": 76337, "tid": -914061504, "ts": 1716454224960062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225021385, "dur": 89,
    "args": {
      "External id": 226575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226575,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226575, "pid": 5, "tid": 7, "ts": 1716454225021385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960067, "dur": 15,
    "args": {
      "External id": 226575,
      "cbid": 211, "correlation": 226575
    }
  },
  {
    "ph": "s", "id": 226575, "pid": 76337, "tid": -914061504, "ts": 1716454224960067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225021475, "dur": 10,
    "args": {
      "External id": 226583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226583,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226583, "pid": 5, "tid": 7, "ts": 1716454225021475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960138, "dur": 12,
    "args": {
      "External id": 226583,
      "cbid": 211, "correlation": 226583
    }
  },
  {
    "ph": "s", "id": 226583, "pid": 76337, "tid": -914061504, "ts": 1716454224960138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225021487, "dur": 21,
    "args": {
      "External id": 226591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226591,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226591, "pid": 5, "tid": 7, "ts": 1716454225021487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960179, "dur": 9,
    "args": {
      "External id": 226591,
      "cbid": 211, "correlation": 226591
    }
  },
  {
    "ph": "s", "id": 226591, "pid": 76337, "tid": -914061504, "ts": 1716454224960179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225021509, "dur": 18,
    "args": {
      "External id": 226613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226613,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226613, "pid": 5, "tid": 7, "ts": 1716454225021509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960230, "dur": 10,
    "args": {
      "External id": 226613,
      "cbid": 211, "correlation": 226613
    }
  },
  {
    "ph": "s", "id": 226613, "pid": 76337, "tid": -914061504, "ts": 1716454224960230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960316, "dur": 1,
    "args": {
      "External id": 226629,
      "cbid": 251, "correlation": 226629
    }
  },
  {
    "ph": "f", "id": 226629, "pid": 76337, "tid": -914061504, "ts": 1716454224960316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960322, "dur": 0,
    "args": {
      "External id": 226631,
      "cbid": 251, "correlation": 226631
    }
  },
  {
    "ph": "f", "id": 226631, "pid": 76337, "tid": -914061504, "ts": 1716454224960322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225021528, "dur": 493,
    "args": {
      "External id": 226632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226632,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226632, "pid": 5, "tid": 7, "ts": 1716454225021528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960324, "dur": 14,
    "args": {
      "External id": 226632,
      "cbid": 211, "correlation": 226632
    }
  },
  {
    "ph": "s", "id": 226632, "pid": 76337, "tid": -914061504, "ts": 1716454224960324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225022022, "dur": 64,
    "args": {
      "External id": 226640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226640, "pid": 5, "tid": 7, "ts": 1716454225022022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960391, "dur": 13,
    "args": {
      "External id": 226640,
      "cbid": 211, "correlation": 226640
    }
  },
  {
    "ph": "s", "id": 226640, "pid": 76337, "tid": -914061504, "ts": 1716454224960391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225022087, "dur": 65,
    "args": {
      "External id": 226648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226648, "pid": 5, "tid": 7, "ts": 1716454225022087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960422, "dur": 9,
    "args": {
      "External id": 226648,
      "cbid": 211, "correlation": 226648
    }
  },
  {
    "ph": "s", "id": 226648, "pid": 76337, "tid": -914061504, "ts": 1716454224960422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960502, "dur": 1,
    "args": {
      "External id": 226664,
      "cbid": 251, "correlation": 226664
    }
  },
  {
    "ph": "f", "id": 226664, "pid": 76337, "tid": -914061504, "ts": 1716454224960502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225022155, "dur": 1,
    "args": {
      "External id": 226666,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 226666,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 226666, "pid": 5, "tid": 7, "ts": 1716454225022155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960507, "dur": 9,
    "args": {
      "External id": 226666,
      "cbid": 51, "correlation": 226666
    }
  },
  {
    "ph": "s", "id": 226666, "pid": 76337, "tid": -914061504, "ts": 1716454224960507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225022158, "dur": 263,
    "args": {
      "External id": 226667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226667,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226667, "pid": 5, "tid": 7, "ts": 1716454225022158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960518, "dur": 12,
    "args": {
      "External id": 226667,
      "cbid": 211, "correlation": 226667
    }
  },
  {
    "ph": "s", "id": 226667, "pid": 76337, "tid": -914061504, "ts": 1716454224960518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225022423, "dur": 13,
    "args": {
      "External id": 226675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226675,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226675, "pid": 5, "tid": 7, "ts": 1716454225022423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960561, "dur": 10,
    "args": {
      "External id": 226675,
      "cbid": 211, "correlation": 226675
    }
  },
  {
    "ph": "s", "id": 226675, "pid": 76337, "tid": -914061504, "ts": 1716454224960561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225022438, "dur": 37,
    "args": {
      "External id": 226686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226686,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226686, "pid": 5, "tid": 7, "ts": 1716454225022438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960630, "dur": 12,
    "args": {
      "External id": 226686,
      "cbid": 211, "correlation": 226686
    }
  },
  {
    "ph": "s", "id": 226686, "pid": 76337, "tid": -914061504, "ts": 1716454224960630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960694, "dur": 0,
    "args": {
      "External id": 226698,
      "cbid": 317, "correlation": 226698
    }
  },
  {
    "ph": "f", "id": 226698, "pid": 76337, "tid": -914061504, "ts": 1716454224960694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960694, "dur": 0,
    "args": {
      "External id": 226699,
      "cbid": 203, "correlation": 226699
    }
  },
  {
    "ph": "f", "id": 226699, "pid": 76337, "tid": -914061504, "ts": 1716454224960694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960695, "dur": 0,
    "args": {
      "External id": 226700,
      "cbid": 205, "correlation": 226700
    }
  },
  {
    "ph": "f", "id": 226700, "pid": 76337, "tid": -914061504, "ts": 1716454224960695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225022476, "dur": 13,
    "args": {
      "External id": 226704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226704, "pid": 5, "tid": 7, "ts": 1716454225022476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960710, "dur": 12,
    "args": {
      "External id": 226704,
      "cbid": 211, "correlation": 226704
    }
  },
  {
    "ph": "s", "id": 226704, "pid": 76337, "tid": -914061504, "ts": 1716454224960710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225022490, "dur": 4,
    "args": {
      "External id": 226706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226706, "pid": 5, "tid": 7, "ts": 1716454225022490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960728, "dur": 6,
    "args": {
      "External id": 226706,
      "cbid": 211, "correlation": 226706
    }
  },
  {
    "ph": "s", "id": 226706, "pid": 76337, "tid": -914061504, "ts": 1716454224960728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960736, "dur": 0,
    "args": {
      "External id": 226707,
      "cbid": 51, "correlation": 226707
    }
  },
  {
    "ph": "s", "id": 226707, "pid": 76337, "tid": -914061504, "ts": 1716454224960736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225022495, "dur": 96,
    "args": {
      "External id": 226708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226708,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 226708, "pid": 5, "tid": 7, "ts": 1716454225022495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960737, "dur": 5,
    "args": {
      "External id": 226708,
      "cbid": 211, "correlation": 226708
    }
  },
  {
    "ph": "s", "id": 226708, "pid": 76337, "tid": -914061504, "ts": 1716454224960737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225022592, "dur": 16,
    "args": {
      "External id": 226713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226713, "pid": 5, "tid": 7, "ts": 1716454225022592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960764, "dur": 9,
    "args": {
      "External id": 226713,
      "cbid": 211, "correlation": 226713
    }
  },
  {
    "ph": "s", "id": 226713, "pid": 76337, "tid": -914061504, "ts": 1716454224960764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225022610, "dur": 11,
    "args": {
      "External id": 226721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226721,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226721, "pid": 5, "tid": 7, "ts": 1716454225022610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960796, "dur": 8,
    "args": {
      "External id": 226721,
      "cbid": 211, "correlation": 226721
    }
  },
  {
    "ph": "s", "id": 226721, "pid": 76337, "tid": -914061504, "ts": 1716454224960796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225022622, "dur": 25,
    "args": {
      "External id": 226730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226730,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226730, "pid": 5, "tid": 7, "ts": 1716454225022622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960835, "dur": 10,
    "args": {
      "External id": 226730,
      "cbid": 211, "correlation": 226730
    }
  },
  {
    "ph": "s", "id": 226730, "pid": 76337, "tid": -914061504, "ts": 1716454224960835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225022649, "dur": 24,
    "args": {
      "External id": 226750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226750,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 226750, "pid": 5, "tid": 7, "ts": 1716454225022649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960905, "dur": 11,
    "args": {
      "External id": 226750,
      "cbid": 211, "correlation": 226750
    }
  },
  {
    "ph": "s", "id": 226750, "pid": 76337, "tid": -914061504, "ts": 1716454224960905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225022674, "dur": 5,
    "args": {
      "External id": 226762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226762,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 226762, "pid": 5, "tid": 7, "ts": 1716454225022674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960926, "dur": 7,
    "args": {
      "External id": 226762,
      "cbid": 211, "correlation": 226762
    }
  },
  {
    "ph": "s", "id": 226762, "pid": 76337, "tid": -914061504, "ts": 1716454224960926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225022680, "dur": 24,
    "args": {
      "External id": 226765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226765, "pid": 5, "tid": 7, "ts": 1716454225022680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960945, "dur": 7,
    "args": {
      "External id": 226765,
      "cbid": 211, "correlation": 226765
    }
  },
  {
    "ph": "s", "id": 226765, "pid": 76337, "tid": -914061504, "ts": 1716454224960945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225022705, "dur": 17,
    "args": {
      "External id": 226774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226774,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226774, "pid": 5, "tid": 7, "ts": 1716454225022705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224960992, "dur": 10,
    "args": {
      "External id": 226774,
      "cbid": 211, "correlation": 226774
    }
  },
  {
    "ph": "s", "id": 226774, "pid": 76337, "tid": -914061504, "ts": 1716454224960992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961045, "dur": 0,
    "args": {
      "External id": 226784,
      "cbid": 317, "correlation": 226784
    }
  },
  {
    "ph": "f", "id": 226784, "pid": 76337, "tid": -914061504, "ts": 1716454224961045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961046, "dur": 0,
    "args": {
      "External id": 226785,
      "cbid": 203, "correlation": 226785
    }
  },
  {
    "ph": "f", "id": 226785, "pid": 76337, "tid": -914061504, "ts": 1716454224961046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961047, "dur": 0,
    "args": {
      "External id": 226786,
      "cbid": 205, "correlation": 226786
    }
  },
  {
    "ph": "f", "id": 226786, "pid": 76337, "tid": -914061504, "ts": 1716454224961047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225022723, "dur": 17,
    "args": {
      "External id": 226790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226790, "pid": 5, "tid": 7, "ts": 1716454225022723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961062, "dur": 11,
    "args": {
      "External id": 226790,
      "cbid": 211, "correlation": 226790
    }
  },
  {
    "ph": "s", "id": 226790, "pid": 76337, "tid": -914061504, "ts": 1716454224961062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225022741, "dur": 237,
    "args": {
      "External id": 226792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226792,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226792, "pid": 5, "tid": 7, "ts": 1716454225022741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961076, "dur": 6,
    "args": {
      "External id": 226792,
      "cbid": 211, "correlation": 226792
    }
  },
  {
    "ph": "s", "id": 226792, "pid": 76337, "tid": -914061504, "ts": 1716454224961076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225022981, "dur": 1,
    "args": {
      "External id": 226794,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 226794,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 226794, "pid": 5, "tid": 7, "ts": 1716454225022981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961087, "dur": 8,
    "args": {
      "External id": 226794,
      "cbid": 51, "correlation": 226794
    }
  },
  {
    "ph": "s", "id": 226794, "pid": 76337, "tid": -914061504, "ts": 1716454224961087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225022985, "dur": 807,
    "args": {
      "External id": 226795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226795,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226795, "pid": 5, "tid": 7, "ts": 1716454225022985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961096, "dur": 6,
    "args": {
      "External id": 226795,
      "cbid": 211, "correlation": 226795
    }
  },
  {
    "ph": "s", "id": 226795, "pid": 76337, "tid": -914061504, "ts": 1716454224961096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225023793, "dur": 13,
    "args": {
      "External id": 226797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226797,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226797, "pid": 5, "tid": 7, "ts": 1716454225023793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961107, "dur": 5,
    "args": {
      "External id": 226797,
      "cbid": 211, "correlation": 226797
    }
  },
  {
    "ph": "s", "id": 226797, "pid": 76337, "tid": -914061504, "ts": 1716454224961107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225023807, "dur": 14,
    "args": {
      "External id": 226803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226803,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226803, "pid": 5, "tid": 7, "ts": 1716454225023807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961136, "dur": 8,
    "args": {
      "External id": 226803,
      "cbid": 211, "correlation": 226803
    }
  },
  {
    "ph": "s", "id": 226803, "pid": 76337, "tid": -914061504, "ts": 1716454224961136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225023823, "dur": 3,
    "args": {
      "External id": 226811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226811,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 226811, "pid": 5, "tid": 7, "ts": 1716454225023823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961180, "dur": 9,
    "args": {
      "External id": 226811,
      "cbid": 211, "correlation": 226811
    }
  },
  {
    "ph": "s", "id": 226811, "pid": 76337, "tid": -914061504, "ts": 1716454224961180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961244, "dur": 1,
    "args": {
      "External id": 226827,
      "cbid": 251, "correlation": 226827
    }
  },
  {
    "ph": "f", "id": 226827, "pid": 76337, "tid": -914061504, "ts": 1716454224961244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961249, "dur": 0,
    "args": {
      "External id": 226829,
      "cbid": 251, "correlation": 226829
    }
  },
  {
    "ph": "f", "id": 226829, "pid": 76337, "tid": -914061504, "ts": 1716454224961249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225023827, "dur": 13,
    "args": {
      "External id": 226830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226830,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226830, "pid": 5, "tid": 7, "ts": 1716454225023827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961251, "dur": 11,
    "args": {
      "External id": 226830,
      "cbid": 211, "correlation": 226830
    }
  },
  {
    "ph": "s", "id": 226830, "pid": 76337, "tid": -914061504, "ts": 1716454224961251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225023842, "dur": 5,
    "args": {
      "External id": 226832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226832,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226832, "pid": 5, "tid": 7, "ts": 1716454225023842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961265, "dur": 5,
    "args": {
      "External id": 226832,
      "cbid": 211, "correlation": 226832
    }
  },
  {
    "ph": "s", "id": 226832, "pid": 76337, "tid": -914061504, "ts": 1716454224961265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225023848, "dur": 17,
    "args": {
      "External id": 226842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226842, "pid": 5, "tid": 7, "ts": 1716454225023848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961322, "dur": 14,
    "args": {
      "External id": 226842,
      "cbid": 211, "correlation": 226842
    }
  },
  {
    "ph": "s", "id": 226842, "pid": 76337, "tid": -914061504, "ts": 1716454224961322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225023866, "dur": 17,
    "args": {
      "External id": 226862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226862,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 226862, "pid": 5, "tid": 7, "ts": 1716454225023866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961390, "dur": 10,
    "args": {
      "External id": 226862,
      "cbid": 211, "correlation": 226862
    }
  },
  {
    "ph": "s", "id": 226862, "pid": 76337, "tid": -914061504, "ts": 1716454224961390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225023884, "dur": 4,
    "args": {
      "External id": 226874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226874,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 226874, "pid": 5, "tid": 7, "ts": 1716454225023884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961411, "dur": 6,
    "args": {
      "External id": 226874,
      "cbid": 211, "correlation": 226874
    }
  },
  {
    "ph": "s", "id": 226874, "pid": 76337, "tid": -914061504, "ts": 1716454224961411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225023890, "dur": 16,
    "args": {
      "External id": 226877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226877,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226877, "pid": 5, "tid": 7, "ts": 1716454225023890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961430, "dur": 7,
    "args": {
      "External id": 226877,
      "cbid": 211, "correlation": 226877
    }
  },
  {
    "ph": "s", "id": 226877, "pid": 76337, "tid": -914061504, "ts": 1716454224961430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225023907, "dur": 11,
    "args": {
      "External id": 226886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226886,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226886, "pid": 5, "tid": 7, "ts": 1716454225023907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961469, "dur": 10,
    "args": {
      "External id": 226886,
      "cbid": 211, "correlation": 226886
    }
  },
  {
    "ph": "s", "id": 226886, "pid": 76337, "tid": -914061504, "ts": 1716454224961469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961531, "dur": 0,
    "args": {
      "External id": 226896,
      "cbid": 317, "correlation": 226896
    }
  },
  {
    "ph": "f", "id": 226896, "pid": 76337, "tid": -914061504, "ts": 1716454224961531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961532, "dur": 0,
    "args": {
      "External id": 226897,
      "cbid": 203, "correlation": 226897
    }
  },
  {
    "ph": "f", "id": 226897, "pid": 76337, "tid": -914061504, "ts": 1716454224961532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961533, "dur": 0,
    "args": {
      "External id": 226898,
      "cbid": 205, "correlation": 226898
    }
  },
  {
    "ph": "f", "id": 226898, "pid": 76337, "tid": -914061504, "ts": 1716454224961533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225023919, "dur": 11,
    "args": {
      "External id": 226902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226902,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226902, "pid": 5, "tid": 7, "ts": 1716454225023919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961546, "dur": 12,
    "args": {
      "External id": 226902,
      "cbid": 211, "correlation": 226902
    }
  },
  {
    "ph": "s", "id": 226902, "pid": 76337, "tid": -914061504, "ts": 1716454224961546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225023931, "dur": 160,
    "args": {
      "External id": 226904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226904,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226904, "pid": 5, "tid": 7, "ts": 1716454225023931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961561, "dur": 5,
    "args": {
      "External id": 226904,
      "cbid": 211, "correlation": 226904
    }
  },
  {
    "ph": "s", "id": 226904, "pid": 76337, "tid": -914061504, "ts": 1716454224961561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225024093, "dur": 1,
    "args": {
      "External id": 226906,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 226906,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 226906, "pid": 5, "tid": 7, "ts": 1716454225024093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961571, "dur": 7,
    "args": {
      "External id": 226906,
      "cbid": 51, "correlation": 226906
    }
  },
  {
    "ph": "s", "id": 226906, "pid": 76337, "tid": -914061504, "ts": 1716454224961571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225024097, "dur": 641,
    "args": {
      "External id": 226907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226907,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 226907, "pid": 5, "tid": 7, "ts": 1716454225024097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961580, "dur": 6,
    "args": {
      "External id": 226907,
      "cbid": 211, "correlation": 226907
    }
  },
  {
    "ph": "s", "id": 226907, "pid": 76337, "tid": -914061504, "ts": 1716454224961580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225024739, "dur": 12,
    "args": {
      "External id": 226909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226909, "pid": 5, "tid": 7, "ts": 1716454225024739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961590, "dur": 5,
    "args": {
      "External id": 226909,
      "cbid": 211, "correlation": 226909
    }
  },
  {
    "ph": "s", "id": 226909, "pid": 76337, "tid": -914061504, "ts": 1716454224961590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225024753, "dur": 15,
    "args": {
      "External id": 226915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226915,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226915, "pid": 5, "tid": 7, "ts": 1716454225024753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961618, "dur": 9,
    "args": {
      "External id": 226915,
      "cbid": 211, "correlation": 226915
    }
  },
  {
    "ph": "s", "id": 226915, "pid": 76337, "tid": -914061504, "ts": 1716454224961618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961677, "dur": 0,
    "args": {
      "External id": 226925,
      "cbid": 317, "correlation": 226925
    }
  },
  {
    "ph": "f", "id": 226925, "pid": 76337, "tid": -914061504, "ts": 1716454224961677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961678, "dur": 0,
    "args": {
      "External id": 226926,
      "cbid": 203, "correlation": 226926
    }
  },
  {
    "ph": "f", "id": 226926, "pid": 76337, "tid": -914061504, "ts": 1716454224961678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961679, "dur": 0,
    "args": {
      "External id": 226927,
      "cbid": 205, "correlation": 226927
    }
  },
  {
    "ph": "f", "id": 226927, "pid": 76337, "tid": -914061504, "ts": 1716454224961679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225024769, "dur": 17,
    "args": {
      "External id": 226931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226931,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226931, "pid": 5, "tid": 7, "ts": 1716454225024769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961692, "dur": 12,
    "args": {
      "External id": 226931,
      "cbid": 211, "correlation": 226931
    }
  },
  {
    "ph": "s", "id": 226931, "pid": 76337, "tid": -914061504, "ts": 1716454224961692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225024787, "dur": 4,
    "args": {
      "External id": 226933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226933,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 226933, "pid": 5, "tid": 7, "ts": 1716454225024787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961708, "dur": 6,
    "args": {
      "External id": 226933,
      "cbid": 211, "correlation": 226933
    }
  },
  {
    "ph": "s", "id": 226933, "pid": 76337, "tid": -914061504, "ts": 1716454224961708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961717, "dur": 0,
    "args": {
      "External id": 226934,
      "cbid": 51, "correlation": 226934
    }
  },
  {
    "ph": "s", "id": 226934, "pid": 76337, "tid": -914061504, "ts": 1716454224961717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225024792, "dur": 132,
    "args": {
      "External id": 226935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226935,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 226935, "pid": 5, "tid": 7, "ts": 1716454225024792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961718, "dur": 5,
    "args": {
      "External id": 226935,
      "cbid": 211, "correlation": 226935
    }
  },
  {
    "ph": "s", "id": 226935, "pid": 76337, "tid": -914061504, "ts": 1716454224961718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225024926, "dur": 16,
    "args": {
      "External id": 226940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226940,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226940, "pid": 5, "tid": 7, "ts": 1716454225024926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961744, "dur": 9,
    "args": {
      "External id": 226940,
      "cbid": 211, "correlation": 226940
    }
  },
  {
    "ph": "s", "id": 226940, "pid": 76337, "tid": -914061504, "ts": 1716454224961744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225024942, "dur": 12,
    "args": {
      "External id": 226948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226948,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226948, "pid": 5, "tid": 7, "ts": 1716454225024942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961773, "dur": 9,
    "args": {
      "External id": 226948,
      "cbid": 211, "correlation": 226948
    }
  },
  {
    "ph": "s", "id": 226948, "pid": 76337, "tid": -914061504, "ts": 1716454224961773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225024956, "dur": 10,
    "args": {
      "External id": 226956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226956,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226956, "pid": 5, "tid": 7, "ts": 1716454225024956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961802, "dur": 8,
    "args": {
      "External id": 226956,
      "cbid": 211, "correlation": 226956
    }
  },
  {
    "ph": "s", "id": 226956, "pid": 76337, "tid": -914061504, "ts": 1716454224961802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225024967, "dur": 18,
    "args": {
      "External id": 226976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226976,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 226976, "pid": 5, "tid": 7, "ts": 1716454225024967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961885, "dur": 12,
    "args": {
      "External id": 226976,
      "cbid": 211, "correlation": 226976
    }
  },
  {
    "ph": "s", "id": 226976, "pid": 76337, "tid": -914061504, "ts": 1716454224961885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225024986, "dur": 4,
    "args": {
      "External id": 226988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226988,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 226988, "pid": 5, "tid": 7, "ts": 1716454225024986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961907, "dur": 6,
    "args": {
      "External id": 226988,
      "cbid": 211, "correlation": 226988
    }
  },
  {
    "ph": "s", "id": 226988, "pid": 76337, "tid": -914061504, "ts": 1716454224961907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225024992, "dur": 16,
    "args": {
      "External id": 226991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 226991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 226991, "pid": 5, "tid": 7, "ts": 1716454225024992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961925, "dur": 7,
    "args": {
      "External id": 226991,
      "cbid": 211, "correlation": 226991
    }
  },
  {
    "ph": "s", "id": 226991, "pid": 76337, "tid": -914061504, "ts": 1716454224961925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961990, "dur": 0,
    "args": {
      "External id": 227002,
      "cbid": 317, "correlation": 227002
    }
  },
  {
    "ph": "f", "id": 227002, "pid": 76337, "tid": -914061504, "ts": 1716454224961990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961991, "dur": 0,
    "args": {
      "External id": 227003,
      "cbid": 203, "correlation": 227003
    }
  },
  {
    "ph": "f", "id": 227003, "pid": 76337, "tid": -914061504, "ts": 1716454224961991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224961992, "dur": 0,
    "args": {
      "External id": 227004,
      "cbid": 205, "correlation": 227004
    }
  },
  {
    "ph": "f", "id": 227004, "pid": 76337, "tid": -914061504, "ts": 1716454224961992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225025009, "dur": 11,
    "args": {
      "External id": 227008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227008,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227008, "pid": 5, "tid": 7, "ts": 1716454225025009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962005, "dur": 12,
    "args": {
      "External id": 227008,
      "cbid": 211, "correlation": 227008
    }
  },
  {
    "ph": "s", "id": 227008, "pid": 76337, "tid": -914061504, "ts": 1716454224962005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225025022, "dur": 3,
    "args": {
      "External id": 227010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227010,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 227010, "pid": 5, "tid": 7, "ts": 1716454225025022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962021, "dur": 6,
    "args": {
      "External id": 227010,
      "cbid": 211, "correlation": 227010
    }
  },
  {
    "ph": "s", "id": 227010, "pid": 76337, "tid": -914061504, "ts": 1716454224962021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962030, "dur": 0,
    "args": {
      "External id": 227011,
      "cbid": 51, "correlation": 227011
    }
  },
  {
    "ph": "s", "id": 227011, "pid": 76337, "tid": -914061504, "ts": 1716454224962030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225025026, "dur": 91,
    "args": {
      "External id": 227012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227012,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 227012, "pid": 5, "tid": 7, "ts": 1716454225025026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962031, "dur": 6,
    "args": {
      "External id": 227012,
      "cbid": 211, "correlation": 227012
    }
  },
  {
    "ph": "s", "id": 227012, "pid": 76337, "tid": -914061504, "ts": 1716454224962031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225025118, "dur": 16,
    "args": {
      "External id": 227017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227017, "pid": 5, "tid": 7, "ts": 1716454225025118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962058, "dur": 8,
    "args": {
      "External id": 227017,
      "cbid": 211, "correlation": 227017
    }
  },
  {
    "ph": "s", "id": 227017, "pid": 76337, "tid": -914061504, "ts": 1716454224962058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225025135, "dur": 82,
    "args": {
      "External id": 227026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227026,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227026, "pid": 5, "tid": 7, "ts": 1716454225025135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962139, "dur": 14,
    "args": {
      "External id": 227026,
      "cbid": 211, "correlation": 227026
    }
  },
  {
    "ph": "s", "id": 227026, "pid": 76337, "tid": -914061504, "ts": 1716454224962139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225025218, "dur": 30,
    "args": {
      "External id": 227048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227048,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227048, "pid": 5, "tid": 7, "ts": 1716454225025218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962197, "dur": 10,
    "args": {
      "External id": 227048,
      "cbid": 211, "correlation": 227048
    }
  },
  {
    "ph": "s", "id": 227048, "pid": 76337, "tid": -914061504, "ts": 1716454224962197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962286, "dur": 1,
    "args": {
      "External id": 227059,
      "cbid": 251, "correlation": 227059
    }
  },
  {
    "ph": "f", "id": 227059, "pid": 76337, "tid": -914061504, "ts": 1716454224962286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225025250, "dur": 162,
    "args": {
      "External id": 227060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227060,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227060, "pid": 5, "tid": 7, "ts": 1716454225025250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962292, "dur": 13,
    "args": {
      "External id": 227060,
      "cbid": 211, "correlation": 227060
    }
  },
  {
    "ph": "s", "id": 227060, "pid": 76337, "tid": -914061504, "ts": 1716454224962292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962362, "dur": 1,
    "args": {
      "External id": 227071,
      "cbid": 251, "correlation": 227071
    }
  },
  {
    "ph": "f", "id": 227071, "pid": 76337, "tid": -914061504, "ts": 1716454224962362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225025413, "dur": 152,
    "args": {
      "External id": 227072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227072,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227072, "pid": 5, "tid": 7, "ts": 1716454225025413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962366, "dur": 11,
    "args": {
      "External id": 227072,
      "cbid": 211, "correlation": 227072
    }
  },
  {
    "ph": "s", "id": 227072, "pid": 76337, "tid": -914061504, "ts": 1716454224962366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962431, "dur": 1,
    "args": {
      "External id": 227083,
      "cbid": 251, "correlation": 227083
    }
  },
  {
    "ph": "f", "id": 227083, "pid": 76337, "tid": -914061504, "ts": 1716454224962431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225025567, "dur": 156,
    "args": {
      "External id": 227084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227084,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227084, "pid": 5, "tid": 7, "ts": 1716454225025567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962436, "dur": 11,
    "args": {
      "External id": 227084,
      "cbid": 211, "correlation": 227084
    }
  },
  {
    "ph": "s", "id": 227084, "pid": 76337, "tid": -914061504, "ts": 1716454224962436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225025724, "dur": 332,
    "args": {
      "External id": 227109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227109,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227109, "pid": 5, "tid": 7, "ts": 1716454225025724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962519, "dur": 13,
    "args": {
      "External id": 227109,
      "cbid": 211, "correlation": 227109
    }
  },
  {
    "ph": "s", "id": 227109, "pid": 76337, "tid": -914061504, "ts": 1716454224962519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962618, "dur": 1,
    "args": {
      "External id": 227127,
      "cbid": 251, "correlation": 227127
    }
  },
  {
    "ph": "f", "id": 227127, "pid": 76337, "tid": -914061504, "ts": 1716454224962618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225026057, "dur": 162,
    "args": {
      "External id": 227129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227129,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227129, "pid": 5, "tid": 7, "ts": 1716454225026057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962624, "dur": 14,
    "args": {
      "External id": 227129,
      "cbid": 211, "correlation": 227129
    }
  },
  {
    "ph": "s", "id": 227129, "pid": 76337, "tid": -914061504, "ts": 1716454224962624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225026220, "dur": 19,
    "args": {
      "External id": 227137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227137,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227137, "pid": 5, "tid": 7, "ts": 1716454225026220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962694, "dur": 12,
    "args": {
      "External id": 227137,
      "cbid": 211, "correlation": 227137
    }
  },
  {
    "ph": "s", "id": 227137, "pid": 76337, "tid": -914061504, "ts": 1716454224962694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225026241, "dur": 28,
    "args": {
      "External id": 227145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227145,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227145, "pid": 5, "tid": 7, "ts": 1716454225026241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962733, "dur": 10,
    "args": {
      "External id": 227145,
      "cbid": 211, "correlation": 227145
    }
  },
  {
    "ph": "s", "id": 227145, "pid": 76337, "tid": -914061504, "ts": 1716454224962733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225026270, "dur": 19,
    "args": {
      "External id": 227156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227156,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227156, "pid": 5, "tid": 7, "ts": 1716454225026270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962804, "dur": 13,
    "args": {
      "External id": 227156,
      "cbid": 211, "correlation": 227156
    }
  },
  {
    "ph": "s", "id": 227156, "pid": 76337, "tid": -914061504, "ts": 1716454224962804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225026290, "dur": 16,
    "args": {
      "External id": 227178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227178,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227178, "pid": 5, "tid": 7, "ts": 1716454225026290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962835, "dur": 8,
    "args": {
      "External id": 227178,
      "cbid": 211, "correlation": 227178
    }
  },
  {
    "ph": "s", "id": 227178, "pid": 76337, "tid": -914061504, "ts": 1716454224962835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962922, "dur": 1,
    "args": {
      "External id": 227189,
      "cbid": 251, "correlation": 227189
    }
  },
  {
    "ph": "f", "id": 227189, "pid": 76337, "tid": -914061504, "ts": 1716454224962922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225026307, "dur": 87,
    "args": {
      "External id": 227190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227190,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 227190, "pid": 5, "tid": 7, "ts": 1716454225026307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224962927, "dur": 14,
    "args": {
      "External id": 227190,
      "cbid": 211, "correlation": 227190
    }
  },
  {
    "ph": "s", "id": 227190, "pid": 76337, "tid": -914061504, "ts": 1716454224962927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963005, "dur": 1,
    "args": {
      "External id": 227201,
      "cbid": 251, "correlation": 227201
    }
  },
  {
    "ph": "f", "id": 227201, "pid": 76337, "tid": -914061504, "ts": 1716454224963005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963008, "dur": 0,
    "args": {
      "External id": 227202,
      "cbid": 251, "correlation": 227202
    }
  },
  {
    "ph": "f", "id": 227202, "pid": 76337, "tid": -914061504, "ts": 1716454224963008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225026395, "dur": 11,
    "args": {
      "External id": 227203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227203,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227203, "pid": 5, "tid": 7, "ts": 1716454225026395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963010, "dur": 12,
    "args": {
      "External id": 227203,
      "cbid": 211, "correlation": 227203
    }
  },
  {
    "ph": "s", "id": 227203, "pid": 76337, "tid": -914061504, "ts": 1716454224963010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225026407, "dur": 5,
    "args": {
      "External id": 227205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227205,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227205, "pid": 5, "tid": 7, "ts": 1716454225026407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963024, "dur": 6,
    "args": {
      "External id": 227205,
      "cbid": 211, "correlation": 227205
    }
  },
  {
    "ph": "s", "id": 227205, "pid": 76337, "tid": -914061504, "ts": 1716454224963024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963083, "dur": 1,
    "args": {
      "External id": 227216,
      "cbid": 251, "correlation": 227216
    }
  },
  {
    "ph": "f", "id": 227216, "pid": 76337, "tid": -914061504, "ts": 1716454224963083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963086, "dur": 0,
    "args": {
      "External id": 227217,
      "cbid": 251, "correlation": 227217
    }
  },
  {
    "ph": "f", "id": 227217, "pid": 76337, "tid": -914061504, "ts": 1716454224963086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225026414, "dur": 8,
    "args": {
      "External id": 227218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227218,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227218, "pid": 5, "tid": 7, "ts": 1716454225026414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963088, "dur": 11,
    "args": {
      "External id": 227218,
      "cbid": 211, "correlation": 227218
    }
  },
  {
    "ph": "s", "id": 227218, "pid": 76337, "tid": -914061504, "ts": 1716454224963088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225026423, "dur": 3,
    "args": {
      "External id": 227220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227220,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227220, "pid": 5, "tid": 7, "ts": 1716454225026423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963101, "dur": 6,
    "args": {
      "External id": 227220,
      "cbid": 211, "correlation": 227220
    }
  },
  {
    "ph": "s", "id": 227220, "pid": 76337, "tid": -914061504, "ts": 1716454224963101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225026428, "dur": 55,
    "args": {
      "External id": 227245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227245,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227245, "pid": 5, "tid": 7, "ts": 1716454225026428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963178, "dur": 13,
    "args": {
      "External id": 227245,
      "cbid": 211, "correlation": 227245
    }
  },
  {
    "ph": "s", "id": 227245, "pid": 76337, "tid": -914061504, "ts": 1716454224963178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963276, "dur": 1,
    "args": {
      "External id": 227263,
      "cbid": 251, "correlation": 227263
    }
  },
  {
    "ph": "f", "id": 227263, "pid": 76337, "tid": -914061504, "ts": 1716454224963276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225026484, "dur": 89,
    "args": {
      "External id": 227265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227265,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 227265, "pid": 5, "tid": 7, "ts": 1716454225026484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963282, "dur": 14,
    "args": {
      "External id": 227265,
      "cbid": 211, "correlation": 227265
    }
  },
  {
    "ph": "s", "id": 227265, "pid": 76337, "tid": -914061504, "ts": 1716454224963282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225026574, "dur": 9,
    "args": {
      "External id": 227273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227273,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227273, "pid": 5, "tid": 7, "ts": 1716454225026574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963352, "dur": 11,
    "args": {
      "External id": 227273,
      "cbid": 211, "correlation": 227273
    }
  },
  {
    "ph": "s", "id": 227273, "pid": 76337, "tid": -914061504, "ts": 1716454224963352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225026585, "dur": 21,
    "args": {
      "External id": 227281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227281, "pid": 5, "tid": 7, "ts": 1716454225026585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963393, "dur": 9,
    "args": {
      "External id": 227281,
      "cbid": 211, "correlation": 227281
    }
  },
  {
    "ph": "s", "id": 227281, "pid": 76337, "tid": -914061504, "ts": 1716454224963393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225026607, "dur": 18,
    "args": {
      "External id": 227303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227303,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227303, "pid": 5, "tid": 7, "ts": 1716454225026607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963445, "dur": 10,
    "args": {
      "External id": 227303,
      "cbid": 211, "correlation": 227303
    }
  },
  {
    "ph": "s", "id": 227303, "pid": 76337, "tid": -914061504, "ts": 1716454224963445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963532, "dur": 1,
    "args": {
      "External id": 227319,
      "cbid": 251, "correlation": 227319
    }
  },
  {
    "ph": "f", "id": 227319, "pid": 76337, "tid": -914061504, "ts": 1716454224963532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963537, "dur": 0,
    "args": {
      "External id": 227321,
      "cbid": 251, "correlation": 227321
    }
  },
  {
    "ph": "f", "id": 227321, "pid": 76337, "tid": -914061504, "ts": 1716454224963537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225026626, "dur": 492,
    "args": {
      "External id": 227322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227322,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227322, "pid": 5, "tid": 7, "ts": 1716454225026626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963539, "dur": 13,
    "args": {
      "External id": 227322,
      "cbid": 211, "correlation": 227322
    }
  },
  {
    "ph": "s", "id": 227322, "pid": 76337, "tid": -914061504, "ts": 1716454224963539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225027119, "dur": 66,
    "args": {
      "External id": 227330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227330,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227330, "pid": 5, "tid": 7, "ts": 1716454225027119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963605, "dur": 12,
    "args": {
      "External id": 227330,
      "cbid": 211, "correlation": 227330
    }
  },
  {
    "ph": "s", "id": 227330, "pid": 76337, "tid": -914061504, "ts": 1716454224963605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225027187, "dur": 66,
    "args": {
      "External id": 227338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227338, "pid": 5, "tid": 7, "ts": 1716454225027187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963636, "dur": 9,
    "args": {
      "External id": 227338,
      "cbid": 211, "correlation": 227338
    }
  },
  {
    "ph": "s", "id": 227338, "pid": 76337, "tid": -914061504, "ts": 1716454224963636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963716, "dur": 1,
    "args": {
      "External id": 227354,
      "cbid": 251, "correlation": 227354
    }
  },
  {
    "ph": "f", "id": 227354, "pid": 76337, "tid": -914061504, "ts": 1716454224963716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225027255, "dur": 1,
    "args": {
      "External id": 227356,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 227356,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 227356, "pid": 5, "tid": 7, "ts": 1716454225027255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963721, "dur": 9,
    "args": {
      "External id": 227356,
      "cbid": 51, "correlation": 227356
    }
  },
  {
    "ph": "s", "id": 227356, "pid": 76337, "tid": -914061504, "ts": 1716454224963721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225027259, "dur": 267,
    "args": {
      "External id": 227357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227357,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 227357, "pid": 5, "tid": 7, "ts": 1716454225027259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963731, "dur": 12,
    "args": {
      "External id": 227357,
      "cbid": 211, "correlation": 227357
    }
  },
  {
    "ph": "s", "id": 227357, "pid": 76337, "tid": -914061504, "ts": 1716454224963731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225027527, "dur": 13,
    "args": {
      "External id": 227365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227365,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227365, "pid": 5, "tid": 7, "ts": 1716454225027527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963773, "dur": 10,
    "args": {
      "External id": 227365,
      "cbid": 211, "correlation": 227365
    }
  },
  {
    "ph": "s", "id": 227365, "pid": 76337, "tid": -914061504, "ts": 1716454224963773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225027542, "dur": 37,
    "args": {
      "External id": 227376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227376,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227376, "pid": 5, "tid": 7, "ts": 1716454225027542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963842, "dur": 12,
    "args": {
      "External id": 227376,
      "cbid": 211, "correlation": 227376
    }
  },
  {
    "ph": "s", "id": 227376, "pid": 76337, "tid": -914061504, "ts": 1716454224963842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963905, "dur": 0,
    "args": {
      "External id": 227388,
      "cbid": 317, "correlation": 227388
    }
  },
  {
    "ph": "f", "id": 227388, "pid": 76337, "tid": -914061504, "ts": 1716454224963905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963906, "dur": 0,
    "args": {
      "External id": 227389,
      "cbid": 203, "correlation": 227389
    }
  },
  {
    "ph": "f", "id": 227389, "pid": 76337, "tid": -914061504, "ts": 1716454224963906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963906, "dur": 0,
    "args": {
      "External id": 227390,
      "cbid": 205, "correlation": 227390
    }
  },
  {
    "ph": "f", "id": 227390, "pid": 76337, "tid": -914061504, "ts": 1716454224963906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225027580, "dur": 12,
    "args": {
      "External id": 227394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227394,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227394, "pid": 5, "tid": 7, "ts": 1716454225027580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963921, "dur": 12,
    "args": {
      "External id": 227394,
      "cbid": 211, "correlation": 227394
    }
  },
  {
    "ph": "s", "id": 227394, "pid": 76337, "tid": -914061504, "ts": 1716454224963921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225027593, "dur": 3,
    "args": {
      "External id": 227396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 227396, "pid": 5, "tid": 7, "ts": 1716454225027593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963938, "dur": 7,
    "args": {
      "External id": 227396,
      "cbid": 211, "correlation": 227396
    }
  },
  {
    "ph": "s", "id": 227396, "pid": 76337, "tid": -914061504, "ts": 1716454224963938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963947, "dur": 0,
    "args": {
      "External id": 227397,
      "cbid": 51, "correlation": 227397
    }
  },
  {
    "ph": "s", "id": 227397, "pid": 76337, "tid": -914061504, "ts": 1716454224963947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225027598, "dur": 95,
    "args": {
      "External id": 227398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227398,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 227398, "pid": 5, "tid": 7, "ts": 1716454225027598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963948, "dur": 5,
    "args": {
      "External id": 227398,
      "cbid": 211, "correlation": 227398
    }
  },
  {
    "ph": "s", "id": 227398, "pid": 76337, "tid": -914061504, "ts": 1716454224963948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225027695, "dur": 16,
    "args": {
      "External id": 227403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227403,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227403, "pid": 5, "tid": 7, "ts": 1716454225027695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224963982, "dur": 10,
    "args": {
      "External id": 227403,
      "cbid": 211, "correlation": 227403
    }
  },
  {
    "ph": "s", "id": 227403, "pid": 76337, "tid": -914061504, "ts": 1716454224963982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225027712, "dur": 11,
    "args": {
      "External id": 227411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227411,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227411, "pid": 5, "tid": 7, "ts": 1716454225027712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964016, "dur": 8,
    "args": {
      "External id": 227411,
      "cbid": 211, "correlation": 227411
    }
  },
  {
    "ph": "s", "id": 227411, "pid": 76337, "tid": -914061504, "ts": 1716454224964016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225027725, "dur": 55,
    "args": {
      "External id": 227422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227422,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227422, "pid": 5, "tid": 7, "ts": 1716454225027725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964080, "dur": 12,
    "args": {
      "External id": 227422,
      "cbid": 211, "correlation": 227422
    }
  },
  {
    "ph": "s", "id": 227422, "pid": 76337, "tid": -914061504, "ts": 1716454224964080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964135, "dur": 0,
    "args": {
      "External id": 227432,
      "cbid": 317, "correlation": 227432
    }
  },
  {
    "ph": "f", "id": 227432, "pid": 76337, "tid": -914061504, "ts": 1716454224964135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964136, "dur": 0,
    "args": {
      "External id": 227433,
      "cbid": 203, "correlation": 227433
    }
  },
  {
    "ph": "f", "id": 227433, "pid": 76337, "tid": -914061504, "ts": 1716454224964136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964137, "dur": 0,
    "args": {
      "External id": 227434,
      "cbid": 205, "correlation": 227434
    }
  },
  {
    "ph": "f", "id": 227434, "pid": 76337, "tid": -914061504, "ts": 1716454224964137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225027781, "dur": 38,
    "args": {
      "External id": 227438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227438, "pid": 5, "tid": 7, "ts": 1716454225027781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964154, "dur": 11,
    "args": {
      "External id": 227438,
      "cbid": 211, "correlation": 227438
    }
  },
  {
    "ph": "s", "id": 227438, "pid": 76337, "tid": -914061504, "ts": 1716454224964154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225027821, "dur": 160,
    "args": {
      "External id": 227440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227440,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227440, "pid": 5, "tid": 7, "ts": 1716454225027821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964167, "dur": 5,
    "args": {
      "External id": 227440,
      "cbid": 211, "correlation": 227440
    }
  },
  {
    "ph": "s", "id": 227440, "pid": 76337, "tid": -914061504, "ts": 1716454224964167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225027983, "dur": 1954,
    "args": {
      "External id": 227442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227442,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227442, "pid": 5, "tid": 7, "ts": 1716454225027983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964179, "dur": 8,
    "args": {
      "External id": 227442,
      "cbid": 211, "correlation": 227442
    }
  },
  {
    "ph": "s", "id": 227442, "pid": 76337, "tid": -914061504, "ts": 1716454224964179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225029938, "dur": 38,
    "args": {
      "External id": 227444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227444, "pid": 5, "tid": 7, "ts": 1716454225029938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964191, "dur": 6,
    "args": {
      "External id": 227444,
      "cbid": 211, "correlation": 227444
    }
  },
  {
    "ph": "s", "id": 227444, "pid": 76337, "tid": -914061504, "ts": 1716454224964191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225029978, "dur": 59,
    "args": {
      "External id": 227450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227450,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227450, "pid": 5, "tid": 7, "ts": 1716454225029978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964219, "dur": 8,
    "args": {
      "External id": 227450,
      "cbid": 211, "correlation": 227450
    }
  },
  {
    "ph": "s", "id": 227450, "pid": 76337, "tid": -914061504, "ts": 1716454224964219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225030038, "dur": 84,
    "args": {
      "External id": 227459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227459,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227459, "pid": 5, "tid": 7, "ts": 1716454225030038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964308, "dur": 13,
    "args": {
      "External id": 227459,
      "cbid": 211, "correlation": 227459
    }
  },
  {
    "ph": "s", "id": 227459, "pid": 76337, "tid": -914061504, "ts": 1716454224964308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225030124, "dur": 72,
    "args": {
      "External id": 227479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227479,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 227479, "pid": 5, "tid": 7, "ts": 1716454225030124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964377, "dur": 11,
    "args": {
      "External id": 227479,
      "cbid": 211, "correlation": 227479
    }
  },
  {
    "ph": "s", "id": 227479, "pid": 76337, "tid": -914061504, "ts": 1716454224964377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225030198, "dur": 5,
    "args": {
      "External id": 227491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227491,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 227491, "pid": 5, "tid": 7, "ts": 1716454225030198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964400, "dur": 7,
    "args": {
      "External id": 227491,
      "cbid": 211, "correlation": 227491
    }
  },
  {
    "ph": "s", "id": 227491, "pid": 76337, "tid": -914061504, "ts": 1716454224964400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225030204, "dur": 82,
    "args": {
      "External id": 227494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227494,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227494, "pid": 5, "tid": 7, "ts": 1716454225030204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964418, "dur": 6,
    "args": {
      "External id": 227494,
      "cbid": 211, "correlation": 227494
    }
  },
  {
    "ph": "s", "id": 227494, "pid": 76337, "tid": -914061504, "ts": 1716454224964418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225030287, "dur": 52,
    "args": {
      "External id": 227503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227503,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227503, "pid": 5, "tid": 7, "ts": 1716454225030287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964458, "dur": 9,
    "args": {
      "External id": 227503,
      "cbid": 211, "correlation": 227503
    }
  },
  {
    "ph": "s", "id": 227503, "pid": 76337, "tid": -914061504, "ts": 1716454224964458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964509, "dur": 0,
    "args": {
      "External id": 227513,
      "cbid": 317, "correlation": 227513
    }
  },
  {
    "ph": "f", "id": 227513, "pid": 76337, "tid": -914061504, "ts": 1716454224964509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964510, "dur": 0,
    "args": {
      "External id": 227514,
      "cbid": 203, "correlation": 227514
    }
  },
  {
    "ph": "f", "id": 227514, "pid": 76337, "tid": -914061504, "ts": 1716454224964510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964511, "dur": 0,
    "args": {
      "External id": 227515,
      "cbid": 205, "correlation": 227515
    }
  },
  {
    "ph": "f", "id": 227515, "pid": 76337, "tid": -914061504, "ts": 1716454224964511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225030340, "dur": 57,
    "args": {
      "External id": 227519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227519,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227519, "pid": 5, "tid": 7, "ts": 1716454225030340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964527, "dur": 11,
    "args": {
      "External id": 227519,
      "cbid": 211, "correlation": 227519
    }
  },
  {
    "ph": "s", "id": 227519, "pid": 76337, "tid": -914061504, "ts": 1716454224964527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225030398, "dur": 120,
    "args": {
      "External id": 227521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227521,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227521, "pid": 5, "tid": 7, "ts": 1716454225030398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964541, "dur": 6,
    "args": {
      "External id": 227521,
      "cbid": 211, "correlation": 227521
    }
  },
  {
    "ph": "s", "id": 227521, "pid": 76337, "tid": -914061504, "ts": 1716454224964541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225030519, "dur": 1873,
    "args": {
      "External id": 227523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227523,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227523, "pid": 5, "tid": 7, "ts": 1716454225030519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964553, "dur": 6,
    "args": {
      "External id": 227523,
      "cbid": 211, "correlation": 227523
    }
  },
  {
    "ph": "s", "id": 227523, "pid": 76337, "tid": -914061504, "ts": 1716454224964553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225032393, "dur": 19,
    "args": {
      "External id": 227525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227525,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227525, "pid": 5, "tid": 7, "ts": 1716454225032393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964563, "dur": 5,
    "args": {
      "External id": 227525,
      "cbid": 211, "correlation": 227525
    }
  },
  {
    "ph": "s", "id": 227525, "pid": 76337, "tid": -914061504, "ts": 1716454224964563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225032413, "dur": 33,
    "args": {
      "External id": 227531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227531, "pid": 5, "tid": 7, "ts": 1716454225032413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964591, "dur": 9,
    "args": {
      "External id": 227531,
      "cbid": 211, "correlation": 227531
    }
  },
  {
    "ph": "s", "id": 227531, "pid": 76337, "tid": -914061504, "ts": 1716454224964591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225032448, "dur": 4,
    "args": {
      "External id": 227539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227539,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 227539, "pid": 5, "tid": 7, "ts": 1716454225032448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964634, "dur": 10,
    "args": {
      "External id": 227539,
      "cbid": 211, "correlation": 227539
    }
  },
  {
    "ph": "s", "id": 227539, "pid": 76337, "tid": -914061504, "ts": 1716454224964634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964701, "dur": 1,
    "args": {
      "External id": 227555,
      "cbid": 251, "correlation": 227555
    }
  },
  {
    "ph": "f", "id": 227555, "pid": 76337, "tid": -914061504, "ts": 1716454224964701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964706, "dur": 0,
    "args": {
      "External id": 227557,
      "cbid": 251, "correlation": 227557
    }
  },
  {
    "ph": "f", "id": 227557, "pid": 76337, "tid": -914061504, "ts": 1716454224964706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225032452, "dur": 12,
    "args": {
      "External id": 227558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227558,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 227558, "pid": 5, "tid": 7, "ts": 1716454225032452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964708, "dur": 11,
    "args": {
      "External id": 227558,
      "cbid": 211, "correlation": 227558
    }
  },
  {
    "ph": "s", "id": 227558, "pid": 76337, "tid": -914061504, "ts": 1716454224964708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225032466, "dur": 5,
    "args": {
      "External id": 227560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227560,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 227560, "pid": 5, "tid": 7, "ts": 1716454225032466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964721, "dur": 5,
    "args": {
      "External id": 227560,
      "cbid": 211, "correlation": 227560
    }
  },
  {
    "ph": "s", "id": 227560, "pid": 76337, "tid": -914061504, "ts": 1716454224964721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225032472, "dur": 29,
    "args": {
      "External id": 227570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227570, "pid": 5, "tid": 7, "ts": 1716454225032472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964780, "dur": 12,
    "args": {
      "External id": 227570,
      "cbid": 211, "correlation": 227570
    }
  },
  {
    "ph": "s", "id": 227570, "pid": 76337, "tid": -914061504, "ts": 1716454224964780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225032502, "dur": 30,
    "args": {
      "External id": 227590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227590,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 227590, "pid": 5, "tid": 7, "ts": 1716454225032502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964845, "dur": 11,
    "args": {
      "External id": 227590,
      "cbid": 211, "correlation": 227590
    }
  },
  {
    "ph": "s", "id": 227590, "pid": 76337, "tid": -914061504, "ts": 1716454224964845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225032533, "dur": 4,
    "args": {
      "External id": 227602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227602,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 227602, "pid": 5, "tid": 7, "ts": 1716454225032533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964865, "dur": 6,
    "args": {
      "External id": 227602,
      "cbid": 211, "correlation": 227602
    }
  },
  {
    "ph": "s", "id": 227602, "pid": 76337, "tid": -914061504, "ts": 1716454224964865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225032539, "dur": 29,
    "args": {
      "External id": 227605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227605,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227605, "pid": 5, "tid": 7, "ts": 1716454225032539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964883, "dur": 6,
    "args": {
      "External id": 227605,
      "cbid": 211, "correlation": 227605
    }
  },
  {
    "ph": "s", "id": 227605, "pid": 76337, "tid": -914061504, "ts": 1716454224964883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225032570, "dur": 20,
    "args": {
      "External id": 227614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227614,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227614, "pid": 5, "tid": 7, "ts": 1716454225032570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964923, "dur": 10,
    "args": {
      "External id": 227614,
      "cbid": 211, "correlation": 227614
    }
  },
  {
    "ph": "s", "id": 227614, "pid": 76337, "tid": -914061504, "ts": 1716454224964923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964995, "dur": 0,
    "args": {
      "External id": 227624,
      "cbid": 317, "correlation": 227624
    }
  },
  {
    "ph": "f", "id": 227624, "pid": 76337, "tid": -914061504, "ts": 1716454224964995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964996, "dur": 0,
    "args": {
      "External id": 227625,
      "cbid": 203, "correlation": 227625
    }
  },
  {
    "ph": "f", "id": 227625, "pid": 76337, "tid": -914061504, "ts": 1716454224964996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224964996, "dur": 0,
    "args": {
      "External id": 227626,
      "cbid": 205, "correlation": 227626
    }
  },
  {
    "ph": "f", "id": 227626, "pid": 76337, "tid": -914061504, "ts": 1716454224964996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225032592, "dur": 23,
    "args": {
      "External id": 227630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227630,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227630, "pid": 5, "tid": 7, "ts": 1716454225032592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965013, "dur": 12,
    "args": {
      "External id": 227630,
      "cbid": 211, "correlation": 227630
    }
  },
  {
    "ph": "s", "id": 227630, "pid": 76337, "tid": -914061504, "ts": 1716454224965013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225032616, "dur": 43,
    "args": {
      "External id": 227632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227632, "pid": 5, "tid": 7, "ts": 1716454225032616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965027, "dur": 5,
    "args": {
      "External id": 227632,
      "cbid": 211, "correlation": 227632
    }
  },
  {
    "ph": "s", "id": 227632, "pid": 76337, "tid": -914061504, "ts": 1716454224965027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225032660, "dur": 637,
    "args": {
      "External id": 227634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227634,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227634, "pid": 5, "tid": 7, "ts": 1716454225032660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965039, "dur": 6,
    "args": {
      "External id": 227634,
      "cbid": 211, "correlation": 227634
    }
  },
  {
    "ph": "s", "id": 227634, "pid": 76337, "tid": -914061504, "ts": 1716454224965039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225033298, "dur": 22,
    "args": {
      "External id": 227636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227636,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227636, "pid": 5, "tid": 7, "ts": 1716454225033298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965049, "dur": 5,
    "args": {
      "External id": 227636,
      "cbid": 211, "correlation": 227636
    }
  },
  {
    "ph": "s", "id": 227636, "pid": 76337, "tid": -914061504, "ts": 1716454224965049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225033322, "dur": 32,
    "args": {
      "External id": 227642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227642,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227642, "pid": 5, "tid": 7, "ts": 1716454225033322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965077, "dur": 9,
    "args": {
      "External id": 227642,
      "cbid": 211, "correlation": 227642
    }
  },
  {
    "ph": "s", "id": 227642, "pid": 76337, "tid": -914061504, "ts": 1716454224965077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965137, "dur": 0,
    "args": {
      "External id": 227652,
      "cbid": 317, "correlation": 227652
    }
  },
  {
    "ph": "f", "id": 227652, "pid": 76337, "tid": -914061504, "ts": 1716454224965137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965137, "dur": 0,
    "args": {
      "External id": 227653,
      "cbid": 203, "correlation": 227653
    }
  },
  {
    "ph": "f", "id": 227653, "pid": 76337, "tid": -914061504, "ts": 1716454224965137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965138, "dur": 0,
    "args": {
      "External id": 227654,
      "cbid": 205, "correlation": 227654
    }
  },
  {
    "ph": "f", "id": 227654, "pid": 76337, "tid": -914061504, "ts": 1716454224965138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225033355, "dur": 55,
    "args": {
      "External id": 227658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227658, "pid": 5, "tid": 7, "ts": 1716454225033355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965151, "dur": 12,
    "args": {
      "External id": 227658,
      "cbid": 211, "correlation": 227658
    }
  },
  {
    "ph": "s", "id": 227658, "pid": 76337, "tid": -914061504, "ts": 1716454224965151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225033412, "dur": 265,
    "args": {
      "External id": 227660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227660,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227660, "pid": 5, "tid": 7, "ts": 1716454225033412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965169, "dur": 7,
    "args": {
      "External id": 227660,
      "cbid": 211, "correlation": 227660
    }
  },
  {
    "ph": "s", "id": 227660, "pid": 76337, "tid": -914061504, "ts": 1716454224965169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225033679, "dur": 22,
    "args": {
      "External id": 227662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227662,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227662, "pid": 5, "tid": 7, "ts": 1716454225033679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965180, "dur": 5,
    "args": {
      "External id": 227662,
      "cbid": 211, "correlation": 227662
    }
  },
  {
    "ph": "s", "id": 227662, "pid": 76337, "tid": -914061504, "ts": 1716454224965180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225033702, "dur": 31,
    "args": {
      "External id": 227668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227668, "pid": 5, "tid": 7, "ts": 1716454225033702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965206, "dur": 8,
    "args": {
      "External id": 227668,
      "cbid": 211, "correlation": 227668
    }
  },
  {
    "ph": "s", "id": 227668, "pid": 76337, "tid": -914061504, "ts": 1716454224965206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225033735, "dur": 27,
    "args": {
      "External id": 227676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227676,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227676, "pid": 5, "tid": 7, "ts": 1716454225033735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965234, "dur": 8,
    "args": {
      "External id": 227676,
      "cbid": 211, "correlation": 227676
    }
  },
  {
    "ph": "s", "id": 227676, "pid": 76337, "tid": -914061504, "ts": 1716454224965234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225033763, "dur": 19,
    "args": {
      "External id": 227684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227684,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227684, "pid": 5, "tid": 7, "ts": 1716454225033763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965264, "dur": 8,
    "args": {
      "External id": 227684,
      "cbid": 211, "correlation": 227684
    }
  },
  {
    "ph": "s", "id": 227684, "pid": 76337, "tid": -914061504, "ts": 1716454224965264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225033784, "dur": 30,
    "args": {
      "External id": 227704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227704,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 227704, "pid": 5, "tid": 7, "ts": 1716454225033784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965345, "dur": 13,
    "args": {
      "External id": 227704,
      "cbid": 211, "correlation": 227704
    }
  },
  {
    "ph": "s", "id": 227704, "pid": 76337, "tid": -914061504, "ts": 1716454224965345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225033815, "dur": 4,
    "args": {
      "External id": 227716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227716,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 227716, "pid": 5, "tid": 7, "ts": 1716454225033815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965367, "dur": 6,
    "args": {
      "External id": 227716,
      "cbid": 211, "correlation": 227716
    }
  },
  {
    "ph": "s", "id": 227716, "pid": 76337, "tid": -914061504, "ts": 1716454224965367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225033820, "dur": 30,
    "args": {
      "External id": 227719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227719,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227719, "pid": 5, "tid": 7, "ts": 1716454225033820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965385, "dur": 7,
    "args": {
      "External id": 227719,
      "cbid": 211, "correlation": 227719
    }
  },
  {
    "ph": "s", "id": 227719, "pid": 76337, "tid": -914061504, "ts": 1716454224965385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965443, "dur": 0,
    "args": {
      "External id": 227730,
      "cbid": 317, "correlation": 227730
    }
  },
  {
    "ph": "f", "id": 227730, "pid": 76337, "tid": -914061504, "ts": 1716454224965443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965444, "dur": 0,
    "args": {
      "External id": 227731,
      "cbid": 203, "correlation": 227731
    }
  },
  {
    "ph": "f", "id": 227731, "pid": 76337, "tid": -914061504, "ts": 1716454224965444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965444, "dur": 0,
    "args": {
      "External id": 227732,
      "cbid": 205, "correlation": 227732
    }
  },
  {
    "ph": "f", "id": 227732, "pid": 76337, "tid": -914061504, "ts": 1716454224965444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225033852, "dur": 22,
    "args": {
      "External id": 227736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227736,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227736, "pid": 5, "tid": 7, "ts": 1716454225033852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965459, "dur": 12,
    "args": {
      "External id": 227736,
      "cbid": 211, "correlation": 227736
    }
  },
  {
    "ph": "s", "id": 227736, "pid": 76337, "tid": -914061504, "ts": 1716454224965459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225033875, "dur": 103,
    "args": {
      "External id": 227738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227738,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227738, "pid": 5, "tid": 7, "ts": 1716454225033875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965477, "dur": 6,
    "args": {
      "External id": 227738,
      "cbid": 211, "correlation": 227738
    }
  },
  {
    "ph": "s", "id": 227738, "pid": 76337, "tid": -914061504, "ts": 1716454224965477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225033979, "dur": 22,
    "args": {
      "External id": 227740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227740,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227740, "pid": 5, "tid": 7, "ts": 1716454225033979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965486, "dur": 5,
    "args": {
      "External id": 227740,
      "cbid": 211, "correlation": 227740
    }
  },
  {
    "ph": "s", "id": 227740, "pid": 76337, "tid": -914061504, "ts": 1716454224965486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225034003, "dur": 32,
    "args": {
      "External id": 227746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227746,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227746, "pid": 5, "tid": 7, "ts": 1716454225034003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965515, "dur": 9,
    "args": {
      "External id": 227746,
      "cbid": 211, "correlation": 227746
    }
  },
  {
    "ph": "s", "id": 227746, "pid": 76337, "tid": -914061504, "ts": 1716454224965515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225034036, "dur": 195,
    "args": {
      "External id": 227755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227755,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227755, "pid": 5, "tid": 7, "ts": 1716454225034036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965595, "dur": 14,
    "args": {
      "External id": 227755,
      "cbid": 211, "correlation": 227755
    }
  },
  {
    "ph": "s", "id": 227755, "pid": 76337, "tid": -914061504, "ts": 1716454224965595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225034232, "dur": 63,
    "args": {
      "External id": 227777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227777,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227777, "pid": 5, "tid": 7, "ts": 1716454225034232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965653, "dur": 10,
    "args": {
      "External id": 227777,
      "cbid": 211, "correlation": 227777
    }
  },
  {
    "ph": "s", "id": 227777, "pid": 76337, "tid": -914061504, "ts": 1716454224965653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965741, "dur": 1,
    "args": {
      "External id": 227788,
      "cbid": 251, "correlation": 227788
    }
  },
  {
    "ph": "f", "id": 227788, "pid": 76337, "tid": -914061504, "ts": 1716454224965741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225034297, "dur": 152,
    "args": {
      "External id": 227789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227789,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227789, "pid": 5, "tid": 7, "ts": 1716454225034297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965746, "dur": 15,
    "args": {
      "External id": 227789,
      "cbid": 211, "correlation": 227789
    }
  },
  {
    "ph": "s", "id": 227789, "pid": 76337, "tid": -914061504, "ts": 1716454224965746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965817, "dur": 1,
    "args": {
      "External id": 227800,
      "cbid": 251, "correlation": 227800
    }
  },
  {
    "ph": "f", "id": 227800, "pid": 76337, "tid": -914061504, "ts": 1716454224965817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225034450, "dur": 145,
    "args": {
      "External id": 227801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227801,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227801, "pid": 5, "tid": 7, "ts": 1716454225034450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965821, "dur": 11,
    "args": {
      "External id": 227801,
      "cbid": 211, "correlation": 227801
    }
  },
  {
    "ph": "s", "id": 227801, "pid": 76337, "tid": -914061504, "ts": 1716454224965821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965887, "dur": 1,
    "args": {
      "External id": 227812,
      "cbid": 251, "correlation": 227812
    }
  },
  {
    "ph": "f", "id": 227812, "pid": 76337, "tid": -914061504, "ts": 1716454224965887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225034597, "dur": 144,
    "args": {
      "External id": 227813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227813,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227813, "pid": 5, "tid": 7, "ts": 1716454225034597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965891, "dur": 11,
    "args": {
      "External id": 227813,
      "cbid": 211, "correlation": 227813
    }
  },
  {
    "ph": "s", "id": 227813, "pid": 76337, "tid": -914061504, "ts": 1716454224965891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225034742, "dur": 1907,
    "args": {
      "External id": 227834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227834,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 227834, "pid": 5, "tid": 7, "ts": 1716454225034742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224965981, "dur": 13,
    "args": {
      "External id": 227834,
      "cbid": 211, "correlation": 227834
    }
  },
  {
    "ph": "s", "id": 227834, "pid": 76337, "tid": -914061504, "ts": 1716454224965981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966082, "dur": 1,
    "args": {
      "External id": 227852,
      "cbid": 251, "correlation": 227852
    }
  },
  {
    "ph": "f", "id": 227852, "pid": 76337, "tid": -914061504, "ts": 1716454224966082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225036650, "dur": 148,
    "args": {
      "External id": 227854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227854,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 227854, "pid": 5, "tid": 7, "ts": 1716454225036650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966088, "dur": 13,
    "args": {
      "External id": 227854,
      "cbid": 211, "correlation": 227854
    }
  },
  {
    "ph": "s", "id": 227854, "pid": 76337, "tid": -914061504, "ts": 1716454224966088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225036799, "dur": 35,
    "args": {
      "External id": 227862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227862,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227862, "pid": 5, "tid": 7, "ts": 1716454225036799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966159, "dur": 12,
    "args": {
      "External id": 227862,
      "cbid": 211, "correlation": 227862
    }
  },
  {
    "ph": "s", "id": 227862, "pid": 76337, "tid": -914061504, "ts": 1716454224966159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225036836, "dur": 52,
    "args": {
      "External id": 227870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227870,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227870, "pid": 5, "tid": 7, "ts": 1716454225036836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966197, "dur": 9,
    "args": {
      "External id": 227870,
      "cbid": 211, "correlation": 227870
    }
  },
  {
    "ph": "s", "id": 227870, "pid": 76337, "tid": -914061504, "ts": 1716454224966197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225036889, "dur": 30,
    "args": {
      "External id": 227881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227881, "pid": 5, "tid": 7, "ts": 1716454225036889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966269, "dur": 12,
    "args": {
      "External id": 227881,
      "cbid": 211, "correlation": 227881
    }
  },
  {
    "ph": "s", "id": 227881, "pid": 76337, "tid": -914061504, "ts": 1716454224966269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225036921, "dur": 33,
    "args": {
      "External id": 227903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227903,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227903, "pid": 5, "tid": 7, "ts": 1716454225036921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966300, "dur": 8,
    "args": {
      "External id": 227903,
      "cbid": 211, "correlation": 227903
    }
  },
  {
    "ph": "s", "id": 227903, "pid": 76337, "tid": -914061504, "ts": 1716454224966300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966386, "dur": 1,
    "args": {
      "External id": 227914,
      "cbid": 251, "correlation": 227914
    }
  },
  {
    "ph": "f", "id": 227914, "pid": 76337, "tid": -914061504, "ts": 1716454224966386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225036955, "dur": 88,
    "args": {
      "External id": 227915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227915,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227915, "pid": 5, "tid": 7, "ts": 1716454225036955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966391, "dur": 13,
    "args": {
      "External id": 227915,
      "cbid": 211, "correlation": 227915
    }
  },
  {
    "ph": "s", "id": 227915, "pid": 76337, "tid": -914061504, "ts": 1716454224966391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966461, "dur": 1,
    "args": {
      "External id": 227926,
      "cbid": 251, "correlation": 227926
    }
  },
  {
    "ph": "f", "id": 227926, "pid": 76337, "tid": -914061504, "ts": 1716454224966461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966464, "dur": 0,
    "args": {
      "External id": 227927,
      "cbid": 251, "correlation": 227927
    }
  },
  {
    "ph": "f", "id": 227927, "pid": 76337, "tid": -914061504, "ts": 1716454224966464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225037045, "dur": 10,
    "args": {
      "External id": 227928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227928,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 227928, "pid": 5, "tid": 7, "ts": 1716454225037045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966466, "dur": 12,
    "args": {
      "External id": 227928,
      "cbid": 211, "correlation": 227928
    }
  },
  {
    "ph": "s", "id": 227928, "pid": 76337, "tid": -914061504, "ts": 1716454224966466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225037057, "dur": 5,
    "args": {
      "External id": 227930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227930,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 227930, "pid": 5, "tid": 7, "ts": 1716454225037057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966480, "dur": 6,
    "args": {
      "External id": 227930,
      "cbid": 211, "correlation": 227930
    }
  },
  {
    "ph": "s", "id": 227930, "pid": 76337, "tid": -914061504, "ts": 1716454224966480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966538, "dur": 1,
    "args": {
      "External id": 227941,
      "cbid": 251, "correlation": 227941
    }
  },
  {
    "ph": "f", "id": 227941, "pid": 76337, "tid": -914061504, "ts": 1716454224966538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966541, "dur": 0,
    "args": {
      "External id": 227942,
      "cbid": 251, "correlation": 227942
    }
  },
  {
    "ph": "f", "id": 227942, "pid": 76337, "tid": -914061504, "ts": 1716454224966541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225037063, "dur": 7,
    "args": {
      "External id": 227943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227943,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 227943, "pid": 5, "tid": 7, "ts": 1716454225037063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966542, "dur": 12,
    "args": {
      "External id": 227943,
      "cbid": 211, "correlation": 227943
    }
  },
  {
    "ph": "s", "id": 227943, "pid": 76337, "tid": -914061504, "ts": 1716454224966542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225037071, "dur": 3,
    "args": {
      "External id": 227945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227945,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 227945, "pid": 5, "tid": 7, "ts": 1716454225037071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966556, "dur": 5,
    "args": {
      "External id": 227945,
      "cbid": 211, "correlation": 227945
    }
  },
  {
    "ph": "s", "id": 227945, "pid": 76337, "tid": -914061504, "ts": 1716454224966556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225037076, "dur": 89,
    "args": {
      "External id": 227966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227966,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 227966, "pid": 5, "tid": 7, "ts": 1716454225037076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966629, "dur": 12,
    "args": {
      "External id": 227966,
      "cbid": 211, "correlation": 227966
    }
  },
  {
    "ph": "s", "id": 227966, "pid": 76337, "tid": -914061504, "ts": 1716454224966629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966726, "dur": 1,
    "args": {
      "External id": 227984,
      "cbid": 251, "correlation": 227984
    }
  },
  {
    "ph": "f", "id": 227984, "pid": 76337, "tid": -914061504, "ts": 1716454224966726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225037166, "dur": 95,
    "args": {
      "External id": 227986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227986,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 227986, "pid": 5, "tid": 7, "ts": 1716454225037166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966732, "dur": 13,
    "args": {
      "External id": 227986,
      "cbid": 211, "correlation": 227986
    }
  },
  {
    "ph": "s", "id": 227986, "pid": 76337, "tid": -914061504, "ts": 1716454224966732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225037262, "dur": 19,
    "args": {
      "External id": 227994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 227994,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 227994, "pid": 5, "tid": 7, "ts": 1716454225037262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966800, "dur": 13,
    "args": {
      "External id": 227994,
      "cbid": 211, "correlation": 227994
    }
  },
  {
    "ph": "s", "id": 227994, "pid": 76337, "tid": -914061504, "ts": 1716454224966800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225037283, "dur": 37,
    "args": {
      "External id": 228002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228002, "pid": 5, "tid": 7, "ts": 1716454225037283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966841, "dur": 9,
    "args": {
      "External id": 228002,
      "cbid": 211, "correlation": 228002
    }
  },
  {
    "ph": "s", "id": 228002, "pid": 76337, "tid": -914061504, "ts": 1716454224966841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225037321, "dur": 34,
    "args": {
      "External id": 228024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228024,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228024, "pid": 5, "tid": 7, "ts": 1716454225037321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966892, "dur": 10,
    "args": {
      "External id": 228024,
      "cbid": 211, "correlation": 228024
    }
  },
  {
    "ph": "s", "id": 228024, "pid": 76337, "tid": -914061504, "ts": 1716454224966892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966990, "dur": 1,
    "args": {
      "External id": 228040,
      "cbid": 251, "correlation": 228040
    }
  },
  {
    "ph": "f", "id": 228040, "pid": 76337, "tid": -914061504, "ts": 1716454224966990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966995, "dur": 0,
    "args": {
      "External id": 228042,
      "cbid": 251, "correlation": 228042
    }
  },
  {
    "ph": "f", "id": 228042, "pid": 76337, "tid": -914061504, "ts": 1716454224966995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225037356, "dur": 529,
    "args": {
      "External id": 228043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228043,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 228043, "pid": 5, "tid": 7, "ts": 1716454225037356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224966999, "dur": 14,
    "args": {
      "External id": 228043,
      "cbid": 211, "correlation": 228043
    }
  },
  {
    "ph": "s", "id": 228043, "pid": 76337, "tid": -914061504, "ts": 1716454224966999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225037887, "dur": 124,
    "args": {
      "External id": 228051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228051, "pid": 5, "tid": 7, "ts": 1716454225037887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967068, "dur": 13,
    "args": {
      "External id": 228051,
      "cbid": 211, "correlation": 228051
    }
  },
  {
    "ph": "s", "id": 228051, "pid": 76337, "tid": -914061504, "ts": 1716454224967068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225038012, "dur": 128,
    "args": {
      "External id": 228059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228059, "pid": 5, "tid": 7, "ts": 1716454225038012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967100, "dur": 9,
    "args": {
      "External id": 228059,
      "cbid": 211, "correlation": 228059
    }
  },
  {
    "ph": "s", "id": 228059, "pid": 76337, "tid": -914061504, "ts": 1716454224967100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967179, "dur": 1,
    "args": {
      "External id": 228075,
      "cbid": 251, "correlation": 228075
    }
  },
  {
    "ph": "f", "id": 228075, "pid": 76337, "tid": -914061504, "ts": 1716454224967179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225038141, "dur": 304,
    "args": {
      "External id": 228077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228077,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228077, "pid": 5, "tid": 7, "ts": 1716454225038141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967184, "dur": 13,
    "args": {
      "External id": 228077,
      "cbid": 211, "correlation": 228077
    }
  },
  {
    "ph": "s", "id": 228077, "pid": 76337, "tid": -914061504, "ts": 1716454224967184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225038447, "dur": 27,
    "args": {
      "External id": 228085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228085,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228085, "pid": 5, "tid": 7, "ts": 1716454225038447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967227, "dur": 10,
    "args": {
      "External id": 228085,
      "cbid": 211, "correlation": 228085
    }
  },
  {
    "ph": "s", "id": 228085, "pid": 76337, "tid": -914061504, "ts": 1716454224967227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225038475, "dur": 80,
    "args": {
      "External id": 228096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228096, "pid": 5, "tid": 7, "ts": 1716454225038475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967294, "dur": 13,
    "args": {
      "External id": 228096,
      "cbid": 211, "correlation": 228096
    }
  },
  {
    "ph": "s", "id": 228096, "pid": 76337, "tid": -914061504, "ts": 1716454224967294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967358, "dur": 0,
    "args": {
      "External id": 228108,
      "cbid": 317, "correlation": 228108
    }
  },
  {
    "ph": "f", "id": 228108, "pid": 76337, "tid": -914061504, "ts": 1716454224967358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967359, "dur": 0,
    "args": {
      "External id": 228109,
      "cbid": 203, "correlation": 228109
    }
  },
  {
    "ph": "f", "id": 228109, "pid": 76337, "tid": -914061504, "ts": 1716454224967359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967360, "dur": 0,
    "args": {
      "External id": 228110,
      "cbid": 205, "correlation": 228110
    }
  },
  {
    "ph": "f", "id": 228110, "pid": 76337, "tid": -914061504, "ts": 1716454224967360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225038557, "dur": 22,
    "args": {
      "External id": 228114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228114,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228114, "pid": 5, "tid": 7, "ts": 1716454225038557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967375, "dur": 12,
    "args": {
      "External id": 228114,
      "cbid": 211, "correlation": 228114
    }
  },
  {
    "ph": "s", "id": 228114, "pid": 76337, "tid": -914061504, "ts": 1716454224967375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225038580, "dur": 117,
    "args": {
      "External id": 228116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228116,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228116, "pid": 5, "tid": 7, "ts": 1716454225038580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967394, "dur": 8,
    "args": {
      "External id": 228116,
      "cbid": 211, "correlation": 228116
    }
  },
  {
    "ph": "s", "id": 228116, "pid": 76337, "tid": -914061504, "ts": 1716454224967394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225038699, "dur": 23,
    "args": {
      "External id": 228118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228118, "pid": 5, "tid": 7, "ts": 1716454225038699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967405, "dur": 5,
    "args": {
      "External id": 228118,
      "cbid": 211, "correlation": 228118
    }
  },
  {
    "ph": "s", "id": 228118, "pid": 76337, "tid": -914061504, "ts": 1716454224967405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225038723, "dur": 32,
    "args": {
      "External id": 228124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228124,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228124, "pid": 5, "tid": 7, "ts": 1716454225038723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967433, "dur": 8,
    "args": {
      "External id": 228124,
      "cbid": 211, "correlation": 228124
    }
  },
  {
    "ph": "s", "id": 228124, "pid": 76337, "tid": -914061504, "ts": 1716454224967433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225038757, "dur": 26,
    "args": {
      "External id": 228132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228132,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228132, "pid": 5, "tid": 7, "ts": 1716454225038757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967464, "dur": 8,
    "args": {
      "External id": 228132,
      "cbid": 211, "correlation": 228132
    }
  },
  {
    "ph": "s", "id": 228132, "pid": 76337, "tid": -914061504, "ts": 1716454224967464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225038784, "dur": 52,
    "args": {
      "External id": 228141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228141,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228141, "pid": 5, "tid": 7, "ts": 1716454225038784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967504, "dur": 10,
    "args": {
      "External id": 228141,
      "cbid": 211, "correlation": 228141
    }
  },
  {
    "ph": "s", "id": 228141, "pid": 76337, "tid": -914061504, "ts": 1716454224967504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225038837, "dur": 53,
    "args": {
      "External id": 228161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228161,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 228161, "pid": 5, "tid": 7, "ts": 1716454225038837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967574, "dur": 11,
    "args": {
      "External id": 228161,
      "cbid": 211, "correlation": 228161
    }
  },
  {
    "ph": "s", "id": 228161, "pid": 76337, "tid": -914061504, "ts": 1716454224967574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225038891, "dur": 5,
    "args": {
      "External id": 228173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228173,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 228173, "pid": 5, "tid": 7, "ts": 1716454225038891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967595, "dur": 7,
    "args": {
      "External id": 228173,
      "cbid": 211, "correlation": 228173
    }
  },
  {
    "ph": "s", "id": 228173, "pid": 76337, "tid": -914061504, "ts": 1716454224967595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225038897, "dur": 55,
    "args": {
      "External id": 228176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228176, "pid": 5, "tid": 7, "ts": 1716454225038897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967614, "dur": 6,
    "args": {
      "External id": 228176,
      "cbid": 211, "correlation": 228176
    }
  },
  {
    "ph": "s", "id": 228176, "pid": 76337, "tid": -914061504, "ts": 1716454224967614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225038953, "dur": 37,
    "args": {
      "External id": 228185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228185,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228185, "pid": 5, "tid": 7, "ts": 1716454225038953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967653, "dur": 10,
    "args": {
      "External id": 228185,
      "cbid": 211, "correlation": 228185
    }
  },
  {
    "ph": "s", "id": 228185, "pid": 76337, "tid": -914061504, "ts": 1716454224967653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967706, "dur": 0,
    "args": {
      "External id": 228195,
      "cbid": 317, "correlation": 228195
    }
  },
  {
    "ph": "f", "id": 228195, "pid": 76337, "tid": -914061504, "ts": 1716454224967706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967707, "dur": 0,
    "args": {
      "External id": 228196,
      "cbid": 203, "correlation": 228196
    }
  },
  {
    "ph": "f", "id": 228196, "pid": 76337, "tid": -914061504, "ts": 1716454224967707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967707, "dur": 0,
    "args": {
      "External id": 228197,
      "cbid": 205, "correlation": 228197
    }
  },
  {
    "ph": "f", "id": 228197, "pid": 76337, "tid": -914061504, "ts": 1716454224967707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225038991, "dur": 38,
    "args": {
      "External id": 228201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228201,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228201, "pid": 5, "tid": 7, "ts": 1716454225038991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967721, "dur": 11,
    "args": {
      "External id": 228201,
      "cbid": 211, "correlation": 228201
    }
  },
  {
    "ph": "s", "id": 228201, "pid": 76337, "tid": -914061504, "ts": 1716454224967721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225039031, "dur": 82,
    "args": {
      "External id": 228203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228203,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228203, "pid": 5, "tid": 7, "ts": 1716454225039031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967735, "dur": 5,
    "args": {
      "External id": 228203,
      "cbid": 211, "correlation": 228203
    }
  },
  {
    "ph": "s", "id": 228203, "pid": 76337, "tid": -914061504, "ts": 1716454224967735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225039114, "dur": 1265,
    "args": {
      "External id": 228205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228205,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228205, "pid": 5, "tid": 7, "ts": 1716454225039114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967746, "dur": 7,
    "args": {
      "External id": 228205,
      "cbid": 211, "correlation": 228205
    }
  },
  {
    "ph": "s", "id": 228205, "pid": 76337, "tid": -914061504, "ts": 1716454224967746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225040381, "dur": 22,
    "args": {
      "External id": 228207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228207,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228207, "pid": 5, "tid": 7, "ts": 1716454225040381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967757, "dur": 5,
    "args": {
      "External id": 228207,
      "cbid": 211, "correlation": 228207
    }
  },
  {
    "ph": "s", "id": 228207, "pid": 76337, "tid": -914061504, "ts": 1716454224967757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225040403, "dur": 33,
    "args": {
      "External id": 228213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228213, "pid": 5, "tid": 7, "ts": 1716454225040403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967784, "dur": 9,
    "args": {
      "External id": 228213,
      "cbid": 211, "correlation": 228213
    }
  },
  {
    "ph": "s", "id": 228213, "pid": 76337, "tid": -914061504, "ts": 1716454224967784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225040437, "dur": 3,
    "args": {
      "External id": 228221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228221,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 228221, "pid": 5, "tid": 7, "ts": 1716454225040437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967827, "dur": 9,
    "args": {
      "External id": 228221,
      "cbid": 211, "correlation": 228221
    }
  },
  {
    "ph": "s", "id": 228221, "pid": 76337, "tid": -914061504, "ts": 1716454224967827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967892, "dur": 1,
    "args": {
      "External id": 228237,
      "cbid": 251, "correlation": 228237
    }
  },
  {
    "ph": "f", "id": 228237, "pid": 76337, "tid": -914061504, "ts": 1716454224967892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967898, "dur": 0,
    "args": {
      "External id": 228239,
      "cbid": 251, "correlation": 228239
    }
  },
  {
    "ph": "f", "id": 228239, "pid": 76337, "tid": -914061504, "ts": 1716454224967898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225040442, "dur": 12,
    "args": {
      "External id": 228240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228240,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 228240, "pid": 5, "tid": 7, "ts": 1716454225040442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967900, "dur": 11,
    "args": {
      "External id": 228240,
      "cbid": 211, "correlation": 228240
    }
  },
  {
    "ph": "s", "id": 228240, "pid": 76337, "tid": -914061504, "ts": 1716454224967900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225040455, "dur": 5,
    "args": {
      "External id": 228242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228242,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 228242, "pid": 5, "tid": 7, "ts": 1716454225040455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967913, "dur": 5,
    "args": {
      "External id": 228242,
      "cbid": 211, "correlation": 228242
    }
  },
  {
    "ph": "s", "id": 228242, "pid": 76337, "tid": -914061504, "ts": 1716454224967913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225040462, "dur": 29,
    "args": {
      "External id": 228252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228252,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228252, "pid": 5, "tid": 7, "ts": 1716454225040462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224967971, "dur": 20,
    "args": {
      "External id": 228252,
      "cbid": 211, "correlation": 228252
    }
  },
  {
    "ph": "s", "id": 228252, "pid": 76337, "tid": -914061504, "ts": 1716454224967971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225040492, "dur": 30,
    "args": {
      "External id": 228272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228272,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 228272, "pid": 5, "tid": 7, "ts": 1716454225040492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968045, "dur": 11,
    "args": {
      "External id": 228272,
      "cbid": 211, "correlation": 228272
    }
  },
  {
    "ph": "s", "id": 228272, "pid": 76337, "tid": -914061504, "ts": 1716454224968045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225040523, "dur": 4,
    "args": {
      "External id": 228284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228284,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 228284, "pid": 5, "tid": 7, "ts": 1716454225040523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968066, "dur": 6,
    "args": {
      "External id": 228284,
      "cbid": 211, "correlation": 228284
    }
  },
  {
    "ph": "s", "id": 228284, "pid": 76337, "tid": -914061504, "ts": 1716454224968066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225040529, "dur": 30,
    "args": {
      "External id": 228287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228287, "pid": 5, "tid": 7, "ts": 1716454225040529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968084, "dur": 7,
    "args": {
      "External id": 228287,
      "cbid": 211, "correlation": 228287
    }
  },
  {
    "ph": "s", "id": 228287, "pid": 76337, "tid": -914061504, "ts": 1716454224968084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225040560, "dur": 20,
    "args": {
      "External id": 228296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228296,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228296, "pid": 5, "tid": 7, "ts": 1716454225040560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968125, "dur": 10,
    "args": {
      "External id": 228296,
      "cbid": 211, "correlation": 228296
    }
  },
  {
    "ph": "s", "id": 228296, "pid": 76337, "tid": -914061504, "ts": 1716454224968125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968187, "dur": 0,
    "args": {
      "External id": 228306,
      "cbid": 317, "correlation": 228306
    }
  },
  {
    "ph": "f", "id": 228306, "pid": 76337, "tid": -914061504, "ts": 1716454224968187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968188, "dur": 0,
    "args": {
      "External id": 228307,
      "cbid": 203, "correlation": 228307
    }
  },
  {
    "ph": "f", "id": 228307, "pid": 76337, "tid": -914061504, "ts": 1716454224968188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968188, "dur": 0,
    "args": {
      "External id": 228308,
      "cbid": 205, "correlation": 228308
    }
  },
  {
    "ph": "f", "id": 228308, "pid": 76337, "tid": -914061504, "ts": 1716454224968188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225040581, "dur": 22,
    "args": {
      "External id": 228312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228312,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228312, "pid": 5, "tid": 7, "ts": 1716454225040581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968203, "dur": 12,
    "args": {
      "External id": 228312,
      "cbid": 211, "correlation": 228312
    }
  },
  {
    "ph": "s", "id": 228312, "pid": 76337, "tid": -914061504, "ts": 1716454224968203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225040605, "dur": 43,
    "args": {
      "External id": 228314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228314,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228314, "pid": 5, "tid": 7, "ts": 1716454225040605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968217, "dur": 5,
    "args": {
      "External id": 228314,
      "cbid": 211, "correlation": 228314
    }
  },
  {
    "ph": "s", "id": 228314, "pid": 76337, "tid": -914061504, "ts": 1716454224968217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225040649, "dur": 639,
    "args": {
      "External id": 228316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228316,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228316, "pid": 5, "tid": 7, "ts": 1716454225040649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968228, "dur": 6,
    "args": {
      "External id": 228316,
      "cbid": 211, "correlation": 228316
    }
  },
  {
    "ph": "s", "id": 228316, "pid": 76337, "tid": -914061504, "ts": 1716454224968228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225041289, "dur": 22,
    "args": {
      "External id": 228318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228318, "pid": 5, "tid": 7, "ts": 1716454225041289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968237, "dur": 5,
    "args": {
      "External id": 228318,
      "cbid": 211, "correlation": 228318
    }
  },
  {
    "ph": "s", "id": 228318, "pid": 76337, "tid": -914061504, "ts": 1716454224968237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225041312, "dur": 32,
    "args": {
      "External id": 228324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228324, "pid": 5, "tid": 7, "ts": 1716454225041312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968266, "dur": 9,
    "args": {
      "External id": 228324,
      "cbid": 211, "correlation": 228324
    }
  },
  {
    "ph": "s", "id": 228324, "pid": 76337, "tid": -914061504, "ts": 1716454224968266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968325, "dur": 0,
    "args": {
      "External id": 228334,
      "cbid": 317, "correlation": 228334
    }
  },
  {
    "ph": "f", "id": 228334, "pid": 76337, "tid": -914061504, "ts": 1716454224968325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968325, "dur": 0,
    "args": {
      "External id": 228335,
      "cbid": 203, "correlation": 228335
    }
  },
  {
    "ph": "f", "id": 228335, "pid": 76337, "tid": -914061504, "ts": 1716454224968325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968326, "dur": 0,
    "args": {
      "External id": 228336,
      "cbid": 205, "correlation": 228336
    }
  },
  {
    "ph": "f", "id": 228336, "pid": 76337, "tid": -914061504, "ts": 1716454224968326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225041346, "dur": 38,
    "args": {
      "External id": 228340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228340,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228340, "pid": 5, "tid": 7, "ts": 1716454225041346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968338, "dur": 14,
    "args": {
      "External id": 228340,
      "cbid": 211, "correlation": 228340
    }
  },
  {
    "ph": "s", "id": 228340, "pid": 76337, "tid": -914061504, "ts": 1716454224968338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225041385, "dur": 186,
    "args": {
      "External id": 228342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228342,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228342, "pid": 5, "tid": 7, "ts": 1716454225041385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968358, "dur": 6,
    "args": {
      "External id": 228342,
      "cbid": 211, "correlation": 228342
    }
  },
  {
    "ph": "s", "id": 228342, "pid": 76337, "tid": -914061504, "ts": 1716454224968358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225041573, "dur": 22,
    "args": {
      "External id": 228344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228344, "pid": 5, "tid": 7, "ts": 1716454225041573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968368, "dur": 5,
    "args": {
      "External id": 228344,
      "cbid": 211, "correlation": 228344
    }
  },
  {
    "ph": "s", "id": 228344, "pid": 76337, "tid": -914061504, "ts": 1716454224968368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225041596, "dur": 32,
    "args": {
      "External id": 228350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228350, "pid": 5, "tid": 7, "ts": 1716454225041596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968393, "dur": 9,
    "args": {
      "External id": 228350,
      "cbid": 211, "correlation": 228350
    }
  },
  {
    "ph": "s", "id": 228350, "pid": 76337, "tid": -914061504, "ts": 1716454224968393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225041630, "dur": 27,
    "args": {
      "External id": 228358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228358,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228358, "pid": 5, "tid": 7, "ts": 1716454225041630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968423, "dur": 8,
    "args": {
      "External id": 228358,
      "cbid": 211, "correlation": 228358
    }
  },
  {
    "ph": "s", "id": 228358, "pid": 76337, "tid": -914061504, "ts": 1716454224968423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225041658, "dur": 20,
    "args": {
      "External id": 228366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228366,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228366, "pid": 5, "tid": 7, "ts": 1716454225041658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968452, "dur": 8,
    "args": {
      "External id": 228366,
      "cbid": 211, "correlation": 228366
    }
  },
  {
    "ph": "s", "id": 228366, "pid": 76337, "tid": -914061504, "ts": 1716454224968452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225041679, "dur": 31,
    "args": {
      "External id": 228386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228386,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 228386, "pid": 5, "tid": 7, "ts": 1716454225041679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968532, "dur": 12,
    "args": {
      "External id": 228386,
      "cbid": 211, "correlation": 228386
    }
  },
  {
    "ph": "s", "id": 228386, "pid": 76337, "tid": -914061504, "ts": 1716454224968532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225041711, "dur": 4,
    "args": {
      "External id": 228398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228398,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 228398, "pid": 5, "tid": 7, "ts": 1716454225041711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968555, "dur": 6,
    "args": {
      "External id": 228398,
      "cbid": 211, "correlation": 228398
    }
  },
  {
    "ph": "s", "id": 228398, "pid": 76337, "tid": -914061504, "ts": 1716454224968555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225041717, "dur": 30,
    "args": {
      "External id": 228401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228401, "pid": 5, "tid": 7, "ts": 1716454225041717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968572, "dur": 7,
    "args": {
      "External id": 228401,
      "cbid": 211, "correlation": 228401
    }
  },
  {
    "ph": "s", "id": 228401, "pid": 76337, "tid": -914061504, "ts": 1716454224968572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968629, "dur": 0,
    "args": {
      "External id": 228412,
      "cbid": 317, "correlation": 228412
    }
  },
  {
    "ph": "f", "id": 228412, "pid": 76337, "tid": -914061504, "ts": 1716454224968629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968630, "dur": 0,
    "args": {
      "External id": 228413,
      "cbid": 203, "correlation": 228413
    }
  },
  {
    "ph": "f", "id": 228413, "pid": 76337, "tid": -914061504, "ts": 1716454224968630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968631, "dur": 0,
    "args": {
      "External id": 228414,
      "cbid": 205, "correlation": 228414
    }
  },
  {
    "ph": "f", "id": 228414, "pid": 76337, "tid": -914061504, "ts": 1716454224968631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225041748, "dur": 23,
    "args": {
      "External id": 228418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228418,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228418, "pid": 5, "tid": 7, "ts": 1716454225041748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968644, "dur": 12,
    "args": {
      "External id": 228418,
      "cbid": 211, "correlation": 228418
    }
  },
  {
    "ph": "s", "id": 228418, "pid": 76337, "tid": -914061504, "ts": 1716454224968644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225041772, "dur": 103,
    "args": {
      "External id": 228420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228420,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228420, "pid": 5, "tid": 7, "ts": 1716454225041772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968662, "dur": 6,
    "args": {
      "External id": 228420,
      "cbid": 211, "correlation": 228420
    }
  },
  {
    "ph": "s", "id": 228420, "pid": 76337, "tid": -914061504, "ts": 1716454224968662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225041876, "dur": 22,
    "args": {
      "External id": 228422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228422, "pid": 5, "tid": 7, "ts": 1716454225041876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968672, "dur": 5,
    "args": {
      "External id": 228422,
      "cbid": 211, "correlation": 228422
    }
  },
  {
    "ph": "s", "id": 228422, "pid": 76337, "tid": -914061504, "ts": 1716454224968672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225041900, "dur": 32,
    "args": {
      "External id": 228428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228428, "pid": 5, "tid": 7, "ts": 1716454225041900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968699, "dur": 9,
    "args": {
      "External id": 228428,
      "cbid": 211, "correlation": 228428
    }
  },
  {
    "ph": "s", "id": 228428, "pid": 76337, "tid": -914061504, "ts": 1716454224968699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225041933, "dur": 194,
    "args": {
      "External id": 228437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228437,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228437, "pid": 5, "tid": 7, "ts": 1716454225041933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968779, "dur": 14,
    "args": {
      "External id": 228437,
      "cbid": 211, "correlation": 228437
    }
  },
  {
    "ph": "s", "id": 228437, "pid": 76337, "tid": -914061504, "ts": 1716454224968779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225042128, "dur": 63,
    "args": {
      "External id": 228459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228459,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228459, "pid": 5, "tid": 7, "ts": 1716454225042128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968836, "dur": 10,
    "args": {
      "External id": 228459,
      "cbid": 211, "correlation": 228459
    }
  },
  {
    "ph": "s", "id": 228459, "pid": 76337, "tid": -914061504, "ts": 1716454224968836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968925, "dur": 1,
    "args": {
      "External id": 228470,
      "cbid": 251, "correlation": 228470
    }
  },
  {
    "ph": "f", "id": 228470, "pid": 76337, "tid": -914061504, "ts": 1716454224968925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225042193, "dur": 151,
    "args": {
      "External id": 228471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228471,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228471, "pid": 5, "tid": 7, "ts": 1716454225042193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224968930, "dur": 13,
    "args": {
      "External id": 228471,
      "cbid": 211, "correlation": 228471
    }
  },
  {
    "ph": "s", "id": 228471, "pid": 76337, "tid": -914061504, "ts": 1716454224968930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969011, "dur": 1,
    "args": {
      "External id": 228482,
      "cbid": 251, "correlation": 228482
    }
  },
  {
    "ph": "f", "id": 228482, "pid": 76337, "tid": -914061504, "ts": 1716454224969011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225042345, "dur": 146,
    "args": {
      "External id": 228483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228483,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228483, "pid": 5, "tid": 7, "ts": 1716454225042345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969015, "dur": 12,
    "args": {
      "External id": 228483,
      "cbid": 211, "correlation": 228483
    }
  },
  {
    "ph": "s", "id": 228483, "pid": 76337, "tid": -914061504, "ts": 1716454224969015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969083, "dur": 1,
    "args": {
      "External id": 228494,
      "cbid": 251, "correlation": 228494
    }
  },
  {
    "ph": "f", "id": 228494, "pid": 76337, "tid": -914061504, "ts": 1716454224969083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225042493, "dur": 143,
    "args": {
      "External id": 228495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228495,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228495, "pid": 5, "tid": 7, "ts": 1716454225042493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969087, "dur": 11,
    "args": {
      "External id": 228495,
      "cbid": 211, "correlation": 228495
    }
  },
  {
    "ph": "s", "id": 228495, "pid": 76337, "tid": -914061504, "ts": 1716454224969087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225042637, "dur": 1908,
    "args": {
      "External id": 228516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228516,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 228516, "pid": 5, "tid": 7, "ts": 1716454225042637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969168, "dur": 13,
    "args": {
      "External id": 228516,
      "cbid": 211, "correlation": 228516
    }
  },
  {
    "ph": "s", "id": 228516, "pid": 76337, "tid": -914061504, "ts": 1716454224969168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969268, "dur": 1,
    "args": {
      "External id": 228534,
      "cbid": 251, "correlation": 228534
    }
  },
  {
    "ph": "f", "id": 228534, "pid": 76337, "tid": -914061504, "ts": 1716454224969268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225044547, "dur": 148,
    "args": {
      "External id": 228536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228536,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 228536, "pid": 5, "tid": 7, "ts": 1716454225044547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969274, "dur": 13,
    "args": {
      "External id": 228536,
      "cbid": 211, "correlation": 228536
    }
  },
  {
    "ph": "s", "id": 228536, "pid": 76337, "tid": -914061504, "ts": 1716454224969274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225044696, "dur": 35,
    "args": {
      "External id": 228544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228544,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228544, "pid": 5, "tid": 7, "ts": 1716454225044696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969345, "dur": 12,
    "args": {
      "External id": 228544,
      "cbid": 211, "correlation": 228544
    }
  },
  {
    "ph": "s", "id": 228544, "pid": 76337, "tid": -914061504, "ts": 1716454224969345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225044732, "dur": 51,
    "args": {
      "External id": 228552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228552,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228552, "pid": 5, "tid": 7, "ts": 1716454225044732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969384, "dur": 8,
    "args": {
      "External id": 228552,
      "cbid": 211, "correlation": 228552
    }
  },
  {
    "ph": "s", "id": 228552, "pid": 76337, "tid": -914061504, "ts": 1716454224969384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225044784, "dur": 29,
    "args": {
      "External id": 228563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228563, "pid": 5, "tid": 7, "ts": 1716454225044784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969455, "dur": 13,
    "args": {
      "External id": 228563,
      "cbid": 211, "correlation": 228563
    }
  },
  {
    "ph": "s", "id": 228563, "pid": 76337, "tid": -914061504, "ts": 1716454224969455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225044815, "dur": 34,
    "args": {
      "External id": 228585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228585,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228585, "pid": 5, "tid": 7, "ts": 1716454225044815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969487, "dur": 7,
    "args": {
      "External id": 228585,
      "cbid": 211, "correlation": 228585
    }
  },
  {
    "ph": "s", "id": 228585, "pid": 76337, "tid": -914061504, "ts": 1716454224969487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969572, "dur": 1,
    "args": {
      "External id": 228596,
      "cbid": 251, "correlation": 228596
    }
  },
  {
    "ph": "f", "id": 228596, "pid": 76337, "tid": -914061504, "ts": 1716454224969572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225044850, "dur": 89,
    "args": {
      "External id": 228597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228597,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228597, "pid": 5, "tid": 7, "ts": 1716454225044850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969577, "dur": 13,
    "args": {
      "External id": 228597,
      "cbid": 211, "correlation": 228597
    }
  },
  {
    "ph": "s", "id": 228597, "pid": 76337, "tid": -914061504, "ts": 1716454224969577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969648, "dur": 1,
    "args": {
      "External id": 228608,
      "cbid": 251, "correlation": 228608
    }
  },
  {
    "ph": "f", "id": 228608, "pid": 76337, "tid": -914061504, "ts": 1716454224969648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969652, "dur": 0,
    "args": {
      "External id": 228609,
      "cbid": 251, "correlation": 228609
    }
  },
  {
    "ph": "f", "id": 228609, "pid": 76337, "tid": -914061504, "ts": 1716454224969652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225044941, "dur": 11,
    "args": {
      "External id": 228610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228610,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 228610, "pid": 5, "tid": 7, "ts": 1716454225044941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969653, "dur": 12,
    "args": {
      "External id": 228610,
      "cbid": 211, "correlation": 228610
    }
  },
  {
    "ph": "s", "id": 228610, "pid": 76337, "tid": -914061504, "ts": 1716454224969653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225044953, "dur": 5,
    "args": {
      "External id": 228612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228612,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 228612, "pid": 5, "tid": 7, "ts": 1716454225044953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969667, "dur": 6,
    "args": {
      "External id": 228612,
      "cbid": 211, "correlation": 228612
    }
  },
  {
    "ph": "s", "id": 228612, "pid": 76337, "tid": -914061504, "ts": 1716454224969667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969725, "dur": 1,
    "args": {
      "External id": 228623,
      "cbid": 251, "correlation": 228623
    }
  },
  {
    "ph": "f", "id": 228623, "pid": 76337, "tid": -914061504, "ts": 1716454224969725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969728, "dur": 0,
    "args": {
      "External id": 228624,
      "cbid": 251, "correlation": 228624
    }
  },
  {
    "ph": "f", "id": 228624, "pid": 76337, "tid": -914061504, "ts": 1716454224969728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225044959, "dur": 7,
    "args": {
      "External id": 228625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228625,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 228625, "pid": 5, "tid": 7, "ts": 1716454225044959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969730, "dur": 12,
    "args": {
      "External id": 228625,
      "cbid": 211, "correlation": 228625
    }
  },
  {
    "ph": "s", "id": 228625, "pid": 76337, "tid": -914061504, "ts": 1716454224969730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225044968, "dur": 3,
    "args": {
      "External id": 228627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228627,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 228627, "pid": 5, "tid": 7, "ts": 1716454225044968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969743, "dur": 5,
    "args": {
      "External id": 228627,
      "cbid": 211, "correlation": 228627
    }
  },
  {
    "ph": "s", "id": 228627, "pid": 76337, "tid": -914061504, "ts": 1716454224969743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225044972, "dur": 90,
    "args": {
      "External id": 228648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228648,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 228648, "pid": 5, "tid": 7, "ts": 1716454225044972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969816, "dur": 12,
    "args": {
      "External id": 228648,
      "cbid": 211, "correlation": 228648
    }
  },
  {
    "ph": "s", "id": 228648, "pid": 76337, "tid": -914061504, "ts": 1716454224969816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969914, "dur": 1,
    "args": {
      "External id": 228666,
      "cbid": 251, "correlation": 228666
    }
  },
  {
    "ph": "f", "id": 228666, "pid": 76337, "tid": -914061504, "ts": 1716454224969914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225045064, "dur": 94,
    "args": {
      "External id": 228668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228668,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228668, "pid": 5, "tid": 7, "ts": 1716454225045064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969920, "dur": 13,
    "args": {
      "External id": 228668,
      "cbid": 211, "correlation": 228668
    }
  },
  {
    "ph": "s", "id": 228668, "pid": 76337, "tid": -914061504, "ts": 1716454224969920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225045159, "dur": 19,
    "args": {
      "External id": 228676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228676,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228676, "pid": 5, "tid": 7, "ts": 1716454225045159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224969996, "dur": 13,
    "args": {
      "External id": 228676,
      "cbid": 211, "correlation": 228676
    }
  },
  {
    "ph": "s", "id": 228676, "pid": 76337, "tid": -914061504, "ts": 1716454224969996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225045180, "dur": 38,
    "args": {
      "External id": 228684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228684, "pid": 5, "tid": 7, "ts": 1716454225045180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970039, "dur": 9,
    "args": {
      "External id": 228684,
      "cbid": 211, "correlation": 228684
    }
  },
  {
    "ph": "s", "id": 228684, "pid": 76337, "tid": -914061504, "ts": 1716454224970039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225045219, "dur": 34,
    "args": {
      "External id": 228706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228706,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228706, "pid": 5, "tid": 7, "ts": 1716454225045219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970091, "dur": 10,
    "args": {
      "External id": 228706,
      "cbid": 211, "correlation": 228706
    }
  },
  {
    "ph": "s", "id": 228706, "pid": 76337, "tid": -914061504, "ts": 1716454224970091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970182, "dur": 1,
    "args": {
      "External id": 228722,
      "cbid": 251, "correlation": 228722
    }
  },
  {
    "ph": "f", "id": 228722, "pid": 76337, "tid": -914061504, "ts": 1716454224970182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970186, "dur": 0,
    "args": {
      "External id": 228724,
      "cbid": 251, "correlation": 228724
    }
  },
  {
    "ph": "f", "id": 228724, "pid": 76337, "tid": -914061504, "ts": 1716454224970186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225045254, "dur": 529,
    "args": {
      "External id": 228725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228725,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 228725, "pid": 5, "tid": 7, "ts": 1716454225045254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970190, "dur": 12,
    "args": {
      "External id": 228725,
      "cbid": 211, "correlation": 228725
    }
  },
  {
    "ph": "s", "id": 228725, "pid": 76337, "tid": -914061504, "ts": 1716454224970190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225045785, "dur": 124,
    "args": {
      "External id": 228733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228733, "pid": 5, "tid": 7, "ts": 1716454225045785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970254, "dur": 13,
    "args": {
      "External id": 228733,
      "cbid": 211, "correlation": 228733
    }
  },
  {
    "ph": "s", "id": 228733, "pid": 76337, "tid": -914061504, "ts": 1716454224970254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225045911, "dur": 131,
    "args": {
      "External id": 228741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228741, "pid": 5, "tid": 7, "ts": 1716454225045911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970286, "dur": 8,
    "args": {
      "External id": 228741,
      "cbid": 211, "correlation": 228741
    }
  },
  {
    "ph": "s", "id": 228741, "pid": 76337, "tid": -914061504, "ts": 1716454224970286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970362, "dur": 1,
    "args": {
      "External id": 228757,
      "cbid": 251, "correlation": 228757
    }
  },
  {
    "ph": "f", "id": 228757, "pid": 76337, "tid": -914061504, "ts": 1716454224970362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225046044, "dur": 297,
    "args": {
      "External id": 228759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228759,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228759, "pid": 5, "tid": 7, "ts": 1716454225046044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970368, "dur": 12,
    "args": {
      "External id": 228759,
      "cbid": 211, "correlation": 228759
    }
  },
  {
    "ph": "s", "id": 228759, "pid": 76337, "tid": -914061504, "ts": 1716454224970368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225046342, "dur": 27,
    "args": {
      "External id": 228767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228767,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228767, "pid": 5, "tid": 7, "ts": 1716454225046342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970409, "dur": 10,
    "args": {
      "External id": 228767,
      "cbid": 211, "correlation": 228767
    }
  },
  {
    "ph": "s", "id": 228767, "pid": 76337, "tid": -914061504, "ts": 1716454224970409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225046370, "dur": 79,
    "args": {
      "External id": 228778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228778, "pid": 5, "tid": 7, "ts": 1716454225046370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970477, "dur": 12,
    "args": {
      "External id": 228778,
      "cbid": 211, "correlation": 228778
    }
  },
  {
    "ph": "s", "id": 228778, "pid": 76337, "tid": -914061504, "ts": 1716454224970477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970541, "dur": 0,
    "args": {
      "External id": 228790,
      "cbid": 317, "correlation": 228790
    }
  },
  {
    "ph": "f", "id": 228790, "pid": 76337, "tid": -914061504, "ts": 1716454224970541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970542, "dur": 0,
    "args": {
      "External id": 228791,
      "cbid": 203, "correlation": 228791
    }
  },
  {
    "ph": "f", "id": 228791, "pid": 76337, "tid": -914061504, "ts": 1716454224970542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970543, "dur": 0,
    "args": {
      "External id": 228792,
      "cbid": 205, "correlation": 228792
    }
  },
  {
    "ph": "f", "id": 228792, "pid": 76337, "tid": -914061504, "ts": 1716454224970543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225046450, "dur": 23,
    "args": {
      "External id": 228796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228796, "pid": 5, "tid": 7, "ts": 1716454225046450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970559, "dur": 12,
    "args": {
      "External id": 228796,
      "cbid": 211, "correlation": 228796
    }
  },
  {
    "ph": "s", "id": 228796, "pid": 76337, "tid": -914061504, "ts": 1716454224970559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225046474, "dur": 119,
    "args": {
      "External id": 228798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228798,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228798, "pid": 5, "tid": 7, "ts": 1716454225046474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970578, "dur": 6,
    "args": {
      "External id": 228798,
      "cbid": 211, "correlation": 228798
    }
  },
  {
    "ph": "s", "id": 228798, "pid": 76337, "tid": -914061504, "ts": 1716454224970578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225046595, "dur": 23,
    "args": {
      "External id": 228800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228800, "pid": 5, "tid": 7, "ts": 1716454225046595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970588, "dur": 5,
    "args": {
      "External id": 228800,
      "cbid": 211, "correlation": 228800
    }
  },
  {
    "ph": "s", "id": 228800, "pid": 76337, "tid": -914061504, "ts": 1716454224970588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225046620, "dur": 33,
    "args": {
      "External id": 228806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228806,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228806, "pid": 5, "tid": 7, "ts": 1716454225046620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970616, "dur": 8,
    "args": {
      "External id": 228806,
      "cbid": 211, "correlation": 228806
    }
  },
  {
    "ph": "s", "id": 228806, "pid": 76337, "tid": -914061504, "ts": 1716454224970616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225046654, "dur": 26,
    "args": {
      "External id": 228814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228814,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228814, "pid": 5, "tid": 7, "ts": 1716454225046654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970646, "dur": 9,
    "args": {
      "External id": 228814,
      "cbid": 211, "correlation": 228814
    }
  },
  {
    "ph": "s", "id": 228814, "pid": 76337, "tid": -914061504, "ts": 1716454224970646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225046681, "dur": 47,
    "args": {
      "External id": 228823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228823,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228823, "pid": 5, "tid": 7, "ts": 1716454225046681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970684, "dur": 10,
    "args": {
      "External id": 228823,
      "cbid": 211, "correlation": 228823
    }
  },
  {
    "ph": "s", "id": 228823, "pid": 76337, "tid": -914061504, "ts": 1716454224970684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225046730, "dur": 43,
    "args": {
      "External id": 228843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228843,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 228843, "pid": 5, "tid": 7, "ts": 1716454225046730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970754, "dur": 12,
    "args": {
      "External id": 228843,
      "cbid": 211, "correlation": 228843
    }
  },
  {
    "ph": "s", "id": 228843, "pid": 76337, "tid": -914061504, "ts": 1716454224970754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225046775, "dur": 5,
    "args": {
      "External id": 228855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228855,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 228855, "pid": 5, "tid": 7, "ts": 1716454225046775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970776, "dur": 6,
    "args": {
      "External id": 228855,
      "cbid": 211, "correlation": 228855
    }
  },
  {
    "ph": "s", "id": 228855, "pid": 76337, "tid": -914061504, "ts": 1716454224970776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225046781, "dur": 44,
    "args": {
      "External id": 228858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228858, "pid": 5, "tid": 7, "ts": 1716454225046781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970794, "dur": 7,
    "args": {
      "External id": 228858,
      "cbid": 211, "correlation": 228858
    }
  },
  {
    "ph": "s", "id": 228858, "pid": 76337, "tid": -914061504, "ts": 1716454224970794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225046826, "dur": 28,
    "args": {
      "External id": 228867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228867,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228867, "pid": 5, "tid": 7, "ts": 1716454225046826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970834, "dur": 9,
    "args": {
      "External id": 228867,
      "cbid": 211, "correlation": 228867
    }
  },
  {
    "ph": "s", "id": 228867, "pid": 76337, "tid": -914061504, "ts": 1716454224970834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970885, "dur": 0,
    "args": {
      "External id": 228877,
      "cbid": 317, "correlation": 228877
    }
  },
  {
    "ph": "f", "id": 228877, "pid": 76337, "tid": -914061504, "ts": 1716454224970885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970886, "dur": 0,
    "args": {
      "External id": 228878,
      "cbid": 203, "correlation": 228878
    }
  },
  {
    "ph": "f", "id": 228878, "pid": 76337, "tid": -914061504, "ts": 1716454224970886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970886, "dur": 0,
    "args": {
      "External id": 228879,
      "cbid": 205, "correlation": 228879
    }
  },
  {
    "ph": "f", "id": 228879, "pid": 76337, "tid": -914061504, "ts": 1716454224970886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225046855, "dur": 30,
    "args": {
      "External id": 228883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228883,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228883, "pid": 5, "tid": 7, "ts": 1716454225046855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970901, "dur": 12,
    "args": {
      "External id": 228883,
      "cbid": 211, "correlation": 228883
    }
  },
  {
    "ph": "s", "id": 228883, "pid": 76337, "tid": -914061504, "ts": 1716454224970901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225046886, "dur": 63,
    "args": {
      "External id": 228885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228885, "pid": 5, "tid": 7, "ts": 1716454225046886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970916, "dur": 5,
    "args": {
      "External id": 228885,
      "cbid": 211, "correlation": 228885
    }
  },
  {
    "ph": "s", "id": 228885, "pid": 76337, "tid": -914061504, "ts": 1716454224970916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225046950, "dur": 959,
    "args": {
      "External id": 228887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228887,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228887, "pid": 5, "tid": 7, "ts": 1716454225046950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970927, "dur": 6,
    "args": {
      "External id": 228887,
      "cbid": 211, "correlation": 228887
    }
  },
  {
    "ph": "s", "id": 228887, "pid": 76337, "tid": -914061504, "ts": 1716454224970927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225047911, "dur": 21,
    "args": {
      "External id": 228889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228889,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228889, "pid": 5, "tid": 7, "ts": 1716454225047911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970937, "dur": 5,
    "args": {
      "External id": 228889,
      "cbid": 211, "correlation": 228889
    }
  },
  {
    "ph": "s", "id": 228889, "pid": 76337, "tid": -914061504, "ts": 1716454224970937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225047933, "dur": 33,
    "args": {
      "External id": 228895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228895,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228895, "pid": 5, "tid": 7, "ts": 1716454225047933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224970966, "dur": 16,
    "args": {
      "External id": 228895,
      "cbid": 211, "correlation": 228895
    }
  },
  {
    "ph": "s", "id": 228895, "pid": 76337, "tid": -914061504, "ts": 1716454224970966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225047967, "dur": 3,
    "args": {
      "External id": 228903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228903,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 228903, "pid": 5, "tid": 7, "ts": 1716454225047967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971018, "dur": 10,
    "args": {
      "External id": 228903,
      "cbid": 211, "correlation": 228903
    }
  },
  {
    "ph": "s", "id": 228903, "pid": 76337, "tid": -914061504, "ts": 1716454224971018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971084, "dur": 1,
    "args": {
      "External id": 228919,
      "cbid": 251, "correlation": 228919
    }
  },
  {
    "ph": "f", "id": 228919, "pid": 76337, "tid": -914061504, "ts": 1716454224971084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971090, "dur": 0,
    "args": {
      "External id": 228921,
      "cbid": 251, "correlation": 228921
    }
  },
  {
    "ph": "f", "id": 228921, "pid": 76337, "tid": -914061504, "ts": 1716454224971090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225047972, "dur": 12,
    "args": {
      "External id": 228922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228922,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 228922, "pid": 5, "tid": 7, "ts": 1716454225047972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971091, "dur": 11,
    "args": {
      "External id": 228922,
      "cbid": 211, "correlation": 228922
    }
  },
  {
    "ph": "s", "id": 228922, "pid": 76337, "tid": -914061504, "ts": 1716454224971091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225047985, "dur": 5,
    "args": {
      "External id": 228924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228924,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 228924, "pid": 5, "tid": 7, "ts": 1716454225047985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971105, "dur": 6,
    "args": {
      "External id": 228924,
      "cbid": 211, "correlation": 228924
    }
  },
  {
    "ph": "s", "id": 228924, "pid": 76337, "tid": -914061504, "ts": 1716454224971105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225047991, "dur": 29,
    "args": {
      "External id": 228934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228934, "pid": 5, "tid": 7, "ts": 1716454225047991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971162, "dur": 12,
    "args": {
      "External id": 228934,
      "cbid": 211, "correlation": 228934
    }
  },
  {
    "ph": "s", "id": 228934, "pid": 76337, "tid": -914061504, "ts": 1716454224971162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225048022, "dur": 30,
    "args": {
      "External id": 228954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228954,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 228954, "pid": 5, "tid": 7, "ts": 1716454225048022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971227, "dur": 10,
    "args": {
      "External id": 228954,
      "cbid": 211, "correlation": 228954
    }
  },
  {
    "ph": "s", "id": 228954, "pid": 76337, "tid": -914061504, "ts": 1716454224971227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225048054, "dur": 4,
    "args": {
      "External id": 228966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228966,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 228966, "pid": 5, "tid": 7, "ts": 1716454225048054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971247, "dur": 6,
    "args": {
      "External id": 228966,
      "cbid": 211, "correlation": 228966
    }
  },
  {
    "ph": "s", "id": 228966, "pid": 76337, "tid": -914061504, "ts": 1716454224971247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225048059, "dur": 29,
    "args": {
      "External id": 228969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228969,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228969, "pid": 5, "tid": 7, "ts": 1716454225048059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971266, "dur": 7,
    "args": {
      "External id": 228969,
      "cbid": 211, "correlation": 228969
    }
  },
  {
    "ph": "s", "id": 228969, "pid": 76337, "tid": -914061504, "ts": 1716454224971266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225048090, "dur": 22,
    "args": {
      "External id": 228978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228978,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228978, "pid": 5, "tid": 7, "ts": 1716454225048090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971306, "dur": 11,
    "args": {
      "External id": 228978,
      "cbid": 211, "correlation": 228978
    }
  },
  {
    "ph": "s", "id": 228978, "pid": 76337, "tid": -914061504, "ts": 1716454224971306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971369, "dur": 0,
    "args": {
      "External id": 228988,
      "cbid": 317, "correlation": 228988
    }
  },
  {
    "ph": "f", "id": 228988, "pid": 76337, "tid": -914061504, "ts": 1716454224971369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971370, "dur": 0,
    "args": {
      "External id": 228989,
      "cbid": 203, "correlation": 228989
    }
  },
  {
    "ph": "f", "id": 228989, "pid": 76337, "tid": -914061504, "ts": 1716454224971370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971371, "dur": 0,
    "args": {
      "External id": 228990,
      "cbid": 205, "correlation": 228990
    }
  },
  {
    "ph": "f", "id": 228990, "pid": 76337, "tid": -914061504, "ts": 1716454224971371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225048113, "dur": 22,
    "args": {
      "External id": 228994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228994,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228994, "pid": 5, "tid": 7, "ts": 1716454225048113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971384, "dur": 12,
    "args": {
      "External id": 228994,
      "cbid": 211, "correlation": 228994
    }
  },
  {
    "ph": "s", "id": 228994, "pid": 76337, "tid": -914061504, "ts": 1716454224971384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225048137, "dur": 43,
    "args": {
      "External id": 228996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228996,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 228996, "pid": 5, "tid": 7, "ts": 1716454225048137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971399, "dur": 6,
    "args": {
      "External id": 228996,
      "cbid": 211, "correlation": 228996
    }
  },
  {
    "ph": "s", "id": 228996, "pid": 76337, "tid": -914061504, "ts": 1716454224971399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225048181, "dur": 638,
    "args": {
      "External id": 228998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 228998,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 228998, "pid": 5, "tid": 7, "ts": 1716454225048181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971411, "dur": 6,
    "args": {
      "External id": 228998,
      "cbid": 211, "correlation": 228998
    }
  },
  {
    "ph": "s", "id": 228998, "pid": 76337, "tid": -914061504, "ts": 1716454224971411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225048820, "dur": 23,
    "args": {
      "External id": 229000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229000,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229000, "pid": 5, "tid": 7, "ts": 1716454225048820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971421, "dur": 5,
    "args": {
      "External id": 229000,
      "cbid": 211, "correlation": 229000
    }
  },
  {
    "ph": "s", "id": 229000, "pid": 76337, "tid": -914061504, "ts": 1716454224971421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225048844, "dur": 33,
    "args": {
      "External id": 229006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229006, "pid": 5, "tid": 7, "ts": 1716454225048844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971448, "dur": 10,
    "args": {
      "External id": 229006,
      "cbid": 211, "correlation": 229006
    }
  },
  {
    "ph": "s", "id": 229006, "pid": 76337, "tid": -914061504, "ts": 1716454224971448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971506, "dur": 0,
    "args": {
      "External id": 229016,
      "cbid": 317, "correlation": 229016
    }
  },
  {
    "ph": "f", "id": 229016, "pid": 76337, "tid": -914061504, "ts": 1716454224971506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971507, "dur": 0,
    "args": {
      "External id": 229017,
      "cbid": 203, "correlation": 229017
    }
  },
  {
    "ph": "f", "id": 229017, "pid": 76337, "tid": -914061504, "ts": 1716454224971507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971508, "dur": 0,
    "args": {
      "External id": 229018,
      "cbid": 205, "correlation": 229018
    }
  },
  {
    "ph": "f", "id": 229018, "pid": 76337, "tid": -914061504, "ts": 1716454224971508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225048878, "dur": 30,
    "args": {
      "External id": 229022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229022,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229022, "pid": 5, "tid": 7, "ts": 1716454225048878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971522, "dur": 11,
    "args": {
      "External id": 229022,
      "cbid": 211, "correlation": 229022
    }
  },
  {
    "ph": "s", "id": 229022, "pid": 76337, "tid": -914061504, "ts": 1716454224971522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225048909, "dur": 149,
    "args": {
      "External id": 229024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229024,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229024, "pid": 5, "tid": 7, "ts": 1716454225048909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971540, "dur": 6,
    "args": {
      "External id": 229024,
      "cbid": 211, "correlation": 229024
    }
  },
  {
    "ph": "s", "id": 229024, "pid": 76337, "tid": -914061504, "ts": 1716454224971540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225049059, "dur": 23,
    "args": {
      "External id": 229026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229026, "pid": 5, "tid": 7, "ts": 1716454225049059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971551, "dur": 5,
    "args": {
      "External id": 229026,
      "cbid": 211, "correlation": 229026
    }
  },
  {
    "ph": "s", "id": 229026, "pid": 76337, "tid": -914061504, "ts": 1716454224971551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225049084, "dur": 32,
    "args": {
      "External id": 229032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229032, "pid": 5, "tid": 7, "ts": 1716454225049084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971577, "dur": 8,
    "args": {
      "External id": 229032,
      "cbid": 211, "correlation": 229032
    }
  },
  {
    "ph": "s", "id": 229032, "pid": 76337, "tid": -914061504, "ts": 1716454224971577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225049118, "dur": 27,
    "args": {
      "External id": 229040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229040,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229040, "pid": 5, "tid": 7, "ts": 1716454225049118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971605, "dur": 9,
    "args": {
      "External id": 229040,
      "cbid": 211, "correlation": 229040
    }
  },
  {
    "ph": "s", "id": 229040, "pid": 76337, "tid": -914061504, "ts": 1716454224971605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225049146, "dur": 20,
    "args": {
      "External id": 229048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229048,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229048, "pid": 5, "tid": 7, "ts": 1716454225049146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971634, "dur": 8,
    "args": {
      "External id": 229048,
      "cbid": 211, "correlation": 229048
    }
  },
  {
    "ph": "s", "id": 229048, "pid": 76337, "tid": -914061504, "ts": 1716454224971634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225049167, "dur": 28,
    "args": {
      "External id": 229068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229068,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 229068, "pid": 5, "tid": 7, "ts": 1716454225049167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971716, "dur": 12,
    "args": {
      "External id": 229068,
      "cbid": 211, "correlation": 229068
    }
  },
  {
    "ph": "s", "id": 229068, "pid": 76337, "tid": -914061504, "ts": 1716454224971716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225049197, "dur": 5,
    "args": {
      "External id": 229080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229080,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 229080, "pid": 5, "tid": 7, "ts": 1716454225049197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971738, "dur": 6,
    "args": {
      "External id": 229080,
      "cbid": 211, "correlation": 229080
    }
  },
  {
    "ph": "s", "id": 229080, "pid": 76337, "tid": -914061504, "ts": 1716454224971738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225049202, "dur": 29,
    "args": {
      "External id": 229083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229083,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229083, "pid": 5, "tid": 7, "ts": 1716454225049202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971755, "dur": 8,
    "args": {
      "External id": 229083,
      "cbid": 211, "correlation": 229083
    }
  },
  {
    "ph": "s", "id": 229083, "pid": 76337, "tid": -914061504, "ts": 1716454224971755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971814, "dur": 0,
    "args": {
      "External id": 229094,
      "cbid": 317, "correlation": 229094
    }
  },
  {
    "ph": "f", "id": 229094, "pid": 76337, "tid": -914061504, "ts": 1716454224971814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971815, "dur": 0,
    "args": {
      "External id": 229095,
      "cbid": 203, "correlation": 229095
    }
  },
  {
    "ph": "f", "id": 229095, "pid": 76337, "tid": -914061504, "ts": 1716454224971815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971816, "dur": 0,
    "args": {
      "External id": 229096,
      "cbid": 205, "correlation": 229096
    }
  },
  {
    "ph": "f", "id": 229096, "pid": 76337, "tid": -914061504, "ts": 1716454224971816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225049233, "dur": 21,
    "args": {
      "External id": 229100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229100, "pid": 5, "tid": 7, "ts": 1716454225049233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971829, "dur": 12,
    "args": {
      "External id": 229100,
      "cbid": 211, "correlation": 229100
    }
  },
  {
    "ph": "s", "id": 229100, "pid": 76337, "tid": -914061504, "ts": 1716454224971829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225049256, "dur": 103,
    "args": {
      "External id": 229102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229102,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229102, "pid": 5, "tid": 7, "ts": 1716454225049256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971847, "dur": 6,
    "args": {
      "External id": 229102,
      "cbid": 211, "correlation": 229102
    }
  },
  {
    "ph": "s", "id": 229102, "pid": 76337, "tid": -914061504, "ts": 1716454224971847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225049360, "dur": 22,
    "args": {
      "External id": 229104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229104,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229104, "pid": 5, "tid": 7, "ts": 1716454225049360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971857, "dur": 5,
    "args": {
      "External id": 229104,
      "cbid": 211, "correlation": 229104
    }
  },
  {
    "ph": "s", "id": 229104, "pid": 76337, "tid": -914061504, "ts": 1716454224971857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225049384, "dur": 32,
    "args": {
      "External id": 229110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229110, "pid": 5, "tid": 7, "ts": 1716454225049384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971885, "dur": 9,
    "args": {
      "External id": 229110,
      "cbid": 211, "correlation": 229110
    }
  },
  {
    "ph": "s", "id": 229110, "pid": 76337, "tid": -914061504, "ts": 1716454224971885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225049417, "dur": 180,
    "args": {
      "External id": 229119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229119,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229119, "pid": 5, "tid": 7, "ts": 1716454225049417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224971967, "dur": 21,
    "args": {
      "External id": 229119,
      "cbid": 211, "correlation": 229119
    }
  },
  {
    "ph": "s", "id": 229119, "pid": 76337, "tid": -914061504, "ts": 1716454224971967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225049598, "dur": 63,
    "args": {
      "External id": 229141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229141,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229141, "pid": 5, "tid": 7, "ts": 1716454225049598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972031, "dur": 11,
    "args": {
      "External id": 229141,
      "cbid": 211, "correlation": 229141
    }
  },
  {
    "ph": "s", "id": 229141, "pid": 76337, "tid": -914061504, "ts": 1716454224972031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972120, "dur": 1,
    "args": {
      "External id": 229152,
      "cbid": 251, "correlation": 229152
    }
  },
  {
    "ph": "f", "id": 229152, "pid": 76337, "tid": -914061504, "ts": 1716454224972120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225049663, "dur": 150,
    "args": {
      "External id": 229153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229153,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229153, "pid": 5, "tid": 7, "ts": 1716454225049663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972126, "dur": 13,
    "args": {
      "External id": 229153,
      "cbid": 211, "correlation": 229153
    }
  },
  {
    "ph": "s", "id": 229153, "pid": 76337, "tid": -914061504, "ts": 1716454224972126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972196, "dur": 1,
    "args": {
      "External id": 229164,
      "cbid": 251, "correlation": 229164
    }
  },
  {
    "ph": "f", "id": 229164, "pid": 76337, "tid": -914061504, "ts": 1716454224972196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225049814, "dur": 146,
    "args": {
      "External id": 229165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229165,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229165, "pid": 5, "tid": 7, "ts": 1716454225049814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972200, "dur": 12,
    "args": {
      "External id": 229165,
      "cbid": 211, "correlation": 229165
    }
  },
  {
    "ph": "s", "id": 229165, "pid": 76337, "tid": -914061504, "ts": 1716454224972200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972266, "dur": 1,
    "args": {
      "External id": 229176,
      "cbid": 251, "correlation": 229176
    }
  },
  {
    "ph": "f", "id": 229176, "pid": 76337, "tid": -914061504, "ts": 1716454224972266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225049961, "dur": 144,
    "args": {
      "External id": 229177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229177,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229177, "pid": 5, "tid": 7, "ts": 1716454225049961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972270, "dur": 12,
    "args": {
      "External id": 229177,
      "cbid": 211, "correlation": 229177
    }
  },
  {
    "ph": "s", "id": 229177, "pid": 76337, "tid": -914061504, "ts": 1716454224972270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225050106, "dur": 1909,
    "args": {
      "External id": 229198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229198,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 229198, "pid": 5, "tid": 7, "ts": 1716454225050106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972350, "dur": 13,
    "args": {
      "External id": 229198,
      "cbid": 211, "correlation": 229198
    }
  },
  {
    "ph": "s", "id": 229198, "pid": 76337, "tid": -914061504, "ts": 1716454224972350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972450, "dur": 1,
    "args": {
      "External id": 229216,
      "cbid": 251, "correlation": 229216
    }
  },
  {
    "ph": "f", "id": 229216, "pid": 76337, "tid": -914061504, "ts": 1716454224972450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225052017, "dur": 147,
    "args": {
      "External id": 229218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229218,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 229218, "pid": 5, "tid": 7, "ts": 1716454225052017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972456, "dur": 14,
    "args": {
      "External id": 229218,
      "cbid": 211, "correlation": 229218
    }
  },
  {
    "ph": "s", "id": 229218, "pid": 76337, "tid": -914061504, "ts": 1716454224972456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225052165, "dur": 35,
    "args": {
      "External id": 229226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229226,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229226, "pid": 5, "tid": 7, "ts": 1716454225052165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972527, "dur": 12,
    "args": {
      "External id": 229226,
      "cbid": 211, "correlation": 229226
    }
  },
  {
    "ph": "s", "id": 229226, "pid": 76337, "tid": -914061504, "ts": 1716454224972527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225052201, "dur": 51,
    "args": {
      "External id": 229234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229234,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229234, "pid": 5, "tid": 7, "ts": 1716454225052201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972564, "dur": 10,
    "args": {
      "External id": 229234,
      "cbid": 211, "correlation": 229234
    }
  },
  {
    "ph": "s", "id": 229234, "pid": 76337, "tid": -914061504, "ts": 1716454224972564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225052253, "dur": 31,
    "args": {
      "External id": 229245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229245,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229245, "pid": 5, "tid": 7, "ts": 1716454225052253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972636, "dur": 13,
    "args": {
      "External id": 229245,
      "cbid": 211, "correlation": 229245
    }
  },
  {
    "ph": "s", "id": 229245, "pid": 76337, "tid": -914061504, "ts": 1716454224972636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225052285, "dur": 33,
    "args": {
      "External id": 229267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229267,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229267, "pid": 5, "tid": 7, "ts": 1716454225052285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972667, "dur": 8,
    "args": {
      "External id": 229267,
      "cbid": 211, "correlation": 229267
    }
  },
  {
    "ph": "s", "id": 229267, "pid": 76337, "tid": -914061504, "ts": 1716454224972667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972752, "dur": 1,
    "args": {
      "External id": 229278,
      "cbid": 251, "correlation": 229278
    }
  },
  {
    "ph": "f", "id": 229278, "pid": 76337, "tid": -914061504, "ts": 1716454224972752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225052320, "dur": 89,
    "args": {
      "External id": 229279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229279,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229279, "pid": 5, "tid": 7, "ts": 1716454225052320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972757, "dur": 13,
    "args": {
      "External id": 229279,
      "cbid": 211, "correlation": 229279
    }
  },
  {
    "ph": "s", "id": 229279, "pid": 76337, "tid": -914061504, "ts": 1716454224972757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972827, "dur": 1,
    "args": {
      "External id": 229290,
      "cbid": 251, "correlation": 229290
    }
  },
  {
    "ph": "f", "id": 229290, "pid": 76337, "tid": -914061504, "ts": 1716454224972827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972831, "dur": 0,
    "args": {
      "External id": 229291,
      "cbid": 251, "correlation": 229291
    }
  },
  {
    "ph": "f", "id": 229291, "pid": 76337, "tid": -914061504, "ts": 1716454224972831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225052411, "dur": 11,
    "args": {
      "External id": 229292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229292,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 229292, "pid": 5, "tid": 7, "ts": 1716454225052411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972832, "dur": 12,
    "args": {
      "External id": 229292,
      "cbid": 211, "correlation": 229292
    }
  },
  {
    "ph": "s", "id": 229292, "pid": 76337, "tid": -914061504, "ts": 1716454224972832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225052423, "dur": 5,
    "args": {
      "External id": 229294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229294,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 229294, "pid": 5, "tid": 7, "ts": 1716454225052423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972846, "dur": 6,
    "args": {
      "External id": 229294,
      "cbid": 211, "correlation": 229294
    }
  },
  {
    "ph": "s", "id": 229294, "pid": 76337, "tid": -914061504, "ts": 1716454224972846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972903, "dur": 1,
    "args": {
      "External id": 229305,
      "cbid": 251, "correlation": 229305
    }
  },
  {
    "ph": "f", "id": 229305, "pid": 76337, "tid": -914061504, "ts": 1716454224972903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972906, "dur": 0,
    "args": {
      "External id": 229306,
      "cbid": 251, "correlation": 229306
    }
  },
  {
    "ph": "f", "id": 229306, "pid": 76337, "tid": -914061504, "ts": 1716454224972906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225052429, "dur": 7,
    "args": {
      "External id": 229307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229307,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 229307, "pid": 5, "tid": 7, "ts": 1716454225052429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972908, "dur": 11,
    "args": {
      "External id": 229307,
      "cbid": 211, "correlation": 229307
    }
  },
  {
    "ph": "s", "id": 229307, "pid": 76337, "tid": -914061504, "ts": 1716454224972908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225052438, "dur": 3,
    "args": {
      "External id": 229309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229309,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 229309, "pid": 5, "tid": 7, "ts": 1716454225052438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224972921, "dur": 7,
    "args": {
      "External id": 229309,
      "cbid": 211, "correlation": 229309
    }
  },
  {
    "ph": "s", "id": 229309, "pid": 76337, "tid": -914061504, "ts": 1716454224972921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225052442, "dur": 89,
    "args": {
      "External id": 229330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229330,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 229330, "pid": 5, "tid": 7, "ts": 1716454225052442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973005, "dur": 13,
    "args": {
      "External id": 229330,
      "cbid": 211, "correlation": 229330
    }
  },
  {
    "ph": "s", "id": 229330, "pid": 76337, "tid": -914061504, "ts": 1716454224973005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973105, "dur": 1,
    "args": {
      "External id": 229348,
      "cbid": 251, "correlation": 229348
    }
  },
  {
    "ph": "f", "id": 229348, "pid": 76337, "tid": -914061504, "ts": 1716454224973105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225052532, "dur": 98,
    "args": {
      "External id": 229350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229350,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229350, "pid": 5, "tid": 7, "ts": 1716454225052532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973111, "dur": 13,
    "args": {
      "External id": 229350,
      "cbid": 211, "correlation": 229350
    }
  },
  {
    "ph": "s", "id": 229350, "pid": 76337, "tid": -914061504, "ts": 1716454224973111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225052632, "dur": 19,
    "args": {
      "External id": 229358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229358,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229358, "pid": 5, "tid": 7, "ts": 1716454225052632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973181, "dur": 12,
    "args": {
      "External id": 229358,
      "cbid": 211, "correlation": 229358
    }
  },
  {
    "ph": "s", "id": 229358, "pid": 76337, "tid": -914061504, "ts": 1716454224973181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225052652, "dur": 38,
    "args": {
      "External id": 229366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229366,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229366, "pid": 5, "tid": 7, "ts": 1716454225052652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973223, "dur": 10,
    "args": {
      "External id": 229366,
      "cbid": 211, "correlation": 229366
    }
  },
  {
    "ph": "s", "id": 229366, "pid": 76337, "tid": -914061504, "ts": 1716454224973223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225052691, "dur": 34,
    "args": {
      "External id": 229388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229388,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229388, "pid": 5, "tid": 7, "ts": 1716454225052691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973274, "dur": 10,
    "args": {
      "External id": 229388,
      "cbid": 211, "correlation": 229388
    }
  },
  {
    "ph": "s", "id": 229388, "pid": 76337, "tid": -914061504, "ts": 1716454224973274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973362, "dur": 1,
    "args": {
      "External id": 229404,
      "cbid": 251, "correlation": 229404
    }
  },
  {
    "ph": "f", "id": 229404, "pid": 76337, "tid": -914061504, "ts": 1716454224973362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973367, "dur": 0,
    "args": {
      "External id": 229406,
      "cbid": 251, "correlation": 229406
    }
  },
  {
    "ph": "f", "id": 229406, "pid": 76337, "tid": -914061504, "ts": 1716454224973367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225052726, "dur": 530,
    "args": {
      "External id": 229407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229407,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 229407, "pid": 5, "tid": 7, "ts": 1716454225052726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973370, "dur": 14,
    "args": {
      "External id": 229407,
      "cbid": 211, "correlation": 229407
    }
  },
  {
    "ph": "s", "id": 229407, "pid": 76337, "tid": -914061504, "ts": 1716454224973370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225053257, "dur": 124,
    "args": {
      "External id": 229415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229415, "pid": 5, "tid": 7, "ts": 1716454225053257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973436, "dur": 13,
    "args": {
      "External id": 229415,
      "cbid": 211, "correlation": 229415
    }
  },
  {
    "ph": "s", "id": 229415, "pid": 76337, "tid": -914061504, "ts": 1716454224973436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225053382, "dur": 128,
    "args": {
      "External id": 229423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229423, "pid": 5, "tid": 7, "ts": 1716454225053382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973467, "dur": 8,
    "args": {
      "External id": 229423,
      "cbid": 211, "correlation": 229423
    }
  },
  {
    "ph": "s", "id": 229423, "pid": 76337, "tid": -914061504, "ts": 1716454224973467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973544, "dur": 1,
    "args": {
      "External id": 229439,
      "cbid": 251, "correlation": 229439
    }
  },
  {
    "ph": "f", "id": 229439, "pid": 76337, "tid": -914061504, "ts": 1716454224973544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225053511, "dur": 302,
    "args": {
      "External id": 229441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229441,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229441, "pid": 5, "tid": 7, "ts": 1716454225053511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973550, "dur": 12,
    "args": {
      "External id": 229441,
      "cbid": 211, "correlation": 229441
    }
  },
  {
    "ph": "s", "id": 229441, "pid": 76337, "tid": -914061504, "ts": 1716454224973550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225053814, "dur": 27,
    "args": {
      "External id": 229449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229449,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229449, "pid": 5, "tid": 7, "ts": 1716454225053814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973593, "dur": 10,
    "args": {
      "External id": 229449,
      "cbid": 211, "correlation": 229449
    }
  },
  {
    "ph": "s", "id": 229449, "pid": 76337, "tid": -914061504, "ts": 1716454224973593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225053843, "dur": 80,
    "args": {
      "External id": 229460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229460, "pid": 5, "tid": 7, "ts": 1716454225053843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973659, "dur": 12,
    "args": {
      "External id": 229460,
      "cbid": 211, "correlation": 229460
    }
  },
  {
    "ph": "s", "id": 229460, "pid": 76337, "tid": -914061504, "ts": 1716454224973659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973722, "dur": 0,
    "args": {
      "External id": 229472,
      "cbid": 317, "correlation": 229472
    }
  },
  {
    "ph": "f", "id": 229472, "pid": 76337, "tid": -914061504, "ts": 1716454224973722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973723, "dur": 0,
    "args": {
      "External id": 229473,
      "cbid": 203, "correlation": 229473
    }
  },
  {
    "ph": "f", "id": 229473, "pid": 76337, "tid": -914061504, "ts": 1716454224973723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973724, "dur": 0,
    "args": {
      "External id": 229474,
      "cbid": 205, "correlation": 229474
    }
  },
  {
    "ph": "f", "id": 229474, "pid": 76337, "tid": -914061504, "ts": 1716454224973724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225053923, "dur": 21,
    "args": {
      "External id": 229478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229478, "pid": 5, "tid": 7, "ts": 1716454225053923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973741, "dur": 12,
    "args": {
      "External id": 229478,
      "cbid": 211, "correlation": 229478
    }
  },
  {
    "ph": "s", "id": 229478, "pid": 76337, "tid": -914061504, "ts": 1716454224973741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225053946, "dur": 118,
    "args": {
      "External id": 229480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229480,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229480, "pid": 5, "tid": 7, "ts": 1716454225053946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973758, "dur": 6,
    "args": {
      "External id": 229480,
      "cbid": 211, "correlation": 229480
    }
  },
  {
    "ph": "s", "id": 229480, "pid": 76337, "tid": -914061504, "ts": 1716454224973758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225054065, "dur": 23,
    "args": {
      "External id": 229482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229482,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229482, "pid": 5, "tid": 7, "ts": 1716454225054065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973768, "dur": 5,
    "args": {
      "External id": 229482,
      "cbid": 211, "correlation": 229482
    }
  },
  {
    "ph": "s", "id": 229482, "pid": 76337, "tid": -914061504, "ts": 1716454224973768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225054089, "dur": 32,
    "args": {
      "External id": 229488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229488, "pid": 5, "tid": 7, "ts": 1716454225054089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973797, "dur": 8,
    "args": {
      "External id": 229488,
      "cbid": 211, "correlation": 229488
    }
  },
  {
    "ph": "s", "id": 229488, "pid": 76337, "tid": -914061504, "ts": 1716454224973797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225054122, "dur": 27,
    "args": {
      "External id": 229496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229496,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229496, "pid": 5, "tid": 7, "ts": 1716454225054122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973828, "dur": 9,
    "args": {
      "External id": 229496,
      "cbid": 211, "correlation": 229496
    }
  },
  {
    "ph": "s", "id": 229496, "pid": 76337, "tid": -914061504, "ts": 1716454224973828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225054150, "dur": 100,
    "args": {
      "External id": 229507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229507,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229507, "pid": 5, "tid": 7, "ts": 1716454225054150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973890, "dur": 11,
    "args": {
      "External id": 229507,
      "cbid": 211, "correlation": 229507
    }
  },
  {
    "ph": "s", "id": 229507, "pid": 76337, "tid": -914061504, "ts": 1716454224973890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973944, "dur": 0,
    "args": {
      "External id": 229517,
      "cbid": 317, "correlation": 229517
    }
  },
  {
    "ph": "f", "id": 229517, "pid": 76337, "tid": -914061504, "ts": 1716454224973944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973945, "dur": 0,
    "args": {
      "External id": 229518,
      "cbid": 203, "correlation": 229518
    }
  },
  {
    "ph": "f", "id": 229518, "pid": 76337, "tid": -914061504, "ts": 1716454224973945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973946, "dur": 0,
    "args": {
      "External id": 229519,
      "cbid": 205, "correlation": 229519
    }
  },
  {
    "ph": "f", "id": 229519, "pid": 76337, "tid": -914061504, "ts": 1716454224973946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225054251, "dur": 75,
    "args": {
      "External id": 229523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229523,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229523, "pid": 5, "tid": 7, "ts": 1716454225054251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973960, "dur": 11,
    "args": {
      "External id": 229523,
      "cbid": 211, "correlation": 229523
    }
  },
  {
    "ph": "s", "id": 229523, "pid": 76337, "tid": -914061504, "ts": 1716454224973960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225054327, "dur": 43,
    "args": {
      "External id": 229525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229525,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229525, "pid": 5, "tid": 7, "ts": 1716454225054327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973985, "dur": 6,
    "args": {
      "External id": 229525,
      "cbid": 211, "correlation": 229525
    }
  },
  {
    "ph": "s", "id": 229525, "pid": 76337, "tid": -914061504, "ts": 1716454224973985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225054371, "dur": 4,
    "args": {
      "External id": 229527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229527,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229527, "pid": 5, "tid": 7, "ts": 1716454225054371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224973996, "dur": 6,
    "args": {
      "External id": 229527,
      "cbid": 211, "correlation": 229527
    }
  },
  {
    "ph": "s", "id": 229527, "pid": 76337, "tid": -914061504, "ts": 1716454224973996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974004, "dur": 0,
    "args": {
      "External id": 229528,
      "cbid": 51, "correlation": 229528
    }
  },
  {
    "ph": "s", "id": 229528, "pid": 76337, "tid": -914061504, "ts": 1716454224974004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225054376, "dur": 2221,
    "args": {
      "External id": 229529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229529,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229529, "pid": 5, "tid": 7, "ts": 1716454225054376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974005, "dur": 5,
    "args": {
      "External id": 229529,
      "cbid": 211, "correlation": 229529
    }
  },
  {
    "ph": "s", "id": 229529, "pid": 76337, "tid": -914061504, "ts": 1716454224974005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225056599, "dur": 111,
    "args": {
      "External id": 229534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229534, "pid": 5, "tid": 7, "ts": 1716454225056599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974035, "dur": 9,
    "args": {
      "External id": 229534,
      "cbid": 211, "correlation": 229534
    }
  },
  {
    "ph": "s", "id": 229534, "pid": 76337, "tid": -914061504, "ts": 1716454224974035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225056711, "dur": 165,
    "args": {
      "External id": 229543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229543,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229543, "pid": 5, "tid": 7, "ts": 1716454225056711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974127, "dur": 13,
    "args": {
      "External id": 229543,
      "cbid": 211, "correlation": 229543
    }
  },
  {
    "ph": "s", "id": 229543, "pid": 76337, "tid": -914061504, "ts": 1716454224974127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225056878, "dur": 127,
    "args": {
      "External id": 229563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229563,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 229563, "pid": 5, "tid": 7, "ts": 1716454225056878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974196, "dur": 11,
    "args": {
      "External id": 229563,
      "cbid": 211, "correlation": 229563
    }
  },
  {
    "ph": "s", "id": 229563, "pid": 76337, "tid": -914061504, "ts": 1716454224974196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225057006, "dur": 5,
    "args": {
      "External id": 229575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229575,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 229575, "pid": 5, "tid": 7, "ts": 1716454225057006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974218, "dur": 6,
    "args": {
      "External id": 229575,
      "cbid": 211, "correlation": 229575
    }
  },
  {
    "ph": "s", "id": 229575, "pid": 76337, "tid": -914061504, "ts": 1716454224974218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225057012, "dur": 160,
    "args": {
      "External id": 229578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229578, "pid": 5, "tid": 7, "ts": 1716454225057012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974237, "dur": 7,
    "args": {
      "External id": 229578,
      "cbid": 211, "correlation": 229578
    }
  },
  {
    "ph": "s", "id": 229578, "pid": 76337, "tid": -914061504, "ts": 1716454224974237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225057173, "dur": 101,
    "args": {
      "External id": 229587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229587,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229587, "pid": 5, "tid": 7, "ts": 1716454225057173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974277, "dur": 10,
    "args": {
      "External id": 229587,
      "cbid": 211, "correlation": 229587
    }
  },
  {
    "ph": "s", "id": 229587, "pid": 76337, "tid": -914061504, "ts": 1716454224974277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974330, "dur": 0,
    "args": {
      "External id": 229597,
      "cbid": 317, "correlation": 229597
    }
  },
  {
    "ph": "f", "id": 229597, "pid": 76337, "tid": -914061504, "ts": 1716454224974330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974331, "dur": 0,
    "args": {
      "External id": 229598,
      "cbid": 203, "correlation": 229598
    }
  },
  {
    "ph": "f", "id": 229598, "pid": 76337, "tid": -914061504, "ts": 1716454224974331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974332, "dur": 0,
    "args": {
      "External id": 229599,
      "cbid": 205, "correlation": 229599
    }
  },
  {
    "ph": "f", "id": 229599, "pid": 76337, "tid": -914061504, "ts": 1716454224974332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225057276, "dur": 110,
    "args": {
      "External id": 229603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229603,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229603, "pid": 5, "tid": 7, "ts": 1716454225057276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974347, "dur": 11,
    "args": {
      "External id": 229603,
      "cbid": 211, "correlation": 229603
    }
  },
  {
    "ph": "s", "id": 229603, "pid": 76337, "tid": -914061504, "ts": 1716454224974347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225057387, "dur": 33,
    "args": {
      "External id": 229605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229605,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229605, "pid": 5, "tid": 7, "ts": 1716454225057387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974361, "dur": 5,
    "args": {
      "External id": 229605,
      "cbid": 211, "correlation": 229605
    }
  },
  {
    "ph": "s", "id": 229605, "pid": 76337, "tid": -914061504, "ts": 1716454224974361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225057422, "dur": 4,
    "args": {
      "External id": 229607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229607, "pid": 5, "tid": 7, "ts": 1716454225057422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974371, "dur": 7,
    "args": {
      "External id": 229607,
      "cbid": 211, "correlation": 229607
    }
  },
  {
    "ph": "s", "id": 229607, "pid": 76337, "tid": -914061504, "ts": 1716454224974371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974381, "dur": 0,
    "args": {
      "External id": 229608,
      "cbid": 51, "correlation": 229608
    }
  },
  {
    "ph": "s", "id": 229608, "pid": 76337, "tid": -914061504, "ts": 1716454224974381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225057426, "dur": 1982,
    "args": {
      "External id": 229609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229609,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229609, "pid": 5, "tid": 7, "ts": 1716454225057426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974382, "dur": 5,
    "args": {
      "External id": 229609,
      "cbid": 211, "correlation": 229609
    }
  },
  {
    "ph": "s", "id": 229609, "pid": 76337, "tid": -914061504, "ts": 1716454224974382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225059410, "dur": 57,
    "args": {
      "External id": 229614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229614, "pid": 5, "tid": 7, "ts": 1716454225059410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974410, "dur": 8,
    "args": {
      "External id": 229614,
      "cbid": 211, "correlation": 229614
    }
  },
  {
    "ph": "s", "id": 229614, "pid": 76337, "tid": -914061504, "ts": 1716454224974410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225059469, "dur": 3,
    "args": {
      "External id": 229622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229622,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229622, "pid": 5, "tid": 7, "ts": 1716454225059469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974453, "dur": 9,
    "args": {
      "External id": 229622,
      "cbid": 211, "correlation": 229622
    }
  },
  {
    "ph": "s", "id": 229622, "pid": 76337, "tid": -914061504, "ts": 1716454224974453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974518, "dur": 1,
    "args": {
      "External id": 229638,
      "cbid": 251, "correlation": 229638
    }
  },
  {
    "ph": "f", "id": 229638, "pid": 76337, "tid": -914061504, "ts": 1716454224974518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974523, "dur": 0,
    "args": {
      "External id": 229640,
      "cbid": 251, "correlation": 229640
    }
  },
  {
    "ph": "f", "id": 229640, "pid": 76337, "tid": -914061504, "ts": 1716454224974523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225059473, "dur": 11,
    "args": {
      "External id": 229641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229641,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 229641, "pid": 5, "tid": 7, "ts": 1716454225059473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974526, "dur": 11,
    "args": {
      "External id": 229641,
      "cbid": 211, "correlation": 229641
    }
  },
  {
    "ph": "s", "id": 229641, "pid": 76337, "tid": -914061504, "ts": 1716454224974526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225059486, "dur": 5,
    "args": {
      "External id": 229643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229643,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 229643, "pid": 5, "tid": 7, "ts": 1716454225059486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974539, "dur": 6,
    "args": {
      "External id": 229643,
      "cbid": 211, "correlation": 229643
    }
  },
  {
    "ph": "s", "id": 229643, "pid": 76337, "tid": -914061504, "ts": 1716454224974539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225059492, "dur": 54,
    "args": {
      "External id": 229653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229653,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229653, "pid": 5, "tid": 7, "ts": 1716454225059492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974597, "dur": 12,
    "args": {
      "External id": 229653,
      "cbid": 211, "correlation": 229653
    }
  },
  {
    "ph": "s", "id": 229653, "pid": 76337, "tid": -914061504, "ts": 1716454224974597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225059547, "dur": 50,
    "args": {
      "External id": 229673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229673,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 229673, "pid": 5, "tid": 7, "ts": 1716454225059547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974663, "dur": 11,
    "args": {
      "External id": 229673,
      "cbid": 211, "correlation": 229673
    }
  },
  {
    "ph": "s", "id": 229673, "pid": 76337, "tid": -914061504, "ts": 1716454224974663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225059598, "dur": 4,
    "args": {
      "External id": 229685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229685,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229685, "pid": 5, "tid": 7, "ts": 1716454225059598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974686, "dur": 6,
    "args": {
      "External id": 229685,
      "cbid": 211, "correlation": 229685
    }
  },
  {
    "ph": "s", "id": 229685, "pid": 76337, "tid": -914061504, "ts": 1716454224974686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225059603, "dur": 55,
    "args": {
      "External id": 229688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229688,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229688, "pid": 5, "tid": 7, "ts": 1716454225059603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974704, "dur": 7,
    "args": {
      "External id": 229688,
      "cbid": 211, "correlation": 229688
    }
  },
  {
    "ph": "s", "id": 229688, "pid": 76337, "tid": -914061504, "ts": 1716454224974704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225059659, "dur": 36,
    "args": {
      "External id": 229697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229697,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229697, "pid": 5, "tid": 7, "ts": 1716454225059659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974745, "dur": 10,
    "args": {
      "External id": 229697,
      "cbid": 211, "correlation": 229697
    }
  },
  {
    "ph": "s", "id": 229697, "pid": 76337, "tid": -914061504, "ts": 1716454224974745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974807, "dur": 0,
    "args": {
      "External id": 229707,
      "cbid": 317, "correlation": 229707
    }
  },
  {
    "ph": "f", "id": 229707, "pid": 76337, "tid": -914061504, "ts": 1716454224974807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974808, "dur": 0,
    "args": {
      "External id": 229708,
      "cbid": 203, "correlation": 229708
    }
  },
  {
    "ph": "f", "id": 229708, "pid": 76337, "tid": -914061504, "ts": 1716454224974808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974809, "dur": 0,
    "args": {
      "External id": 229709,
      "cbid": 205, "correlation": 229709
    }
  },
  {
    "ph": "f", "id": 229709, "pid": 76337, "tid": -914061504, "ts": 1716454224974809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225059697, "dur": 40,
    "args": {
      "External id": 229713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229713,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229713, "pid": 5, "tid": 7, "ts": 1716454225059697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974826, "dur": 13,
    "args": {
      "External id": 229713,
      "cbid": 211, "correlation": 229713
    }
  },
  {
    "ph": "s", "id": 229713, "pid": 76337, "tid": -914061504, "ts": 1716454224974826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225059738, "dur": 14,
    "args": {
      "External id": 229715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229715,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229715, "pid": 5, "tid": 7, "ts": 1716454225059738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974842, "dur": 5,
    "args": {
      "External id": 229715,
      "cbid": 211, "correlation": 229715
    }
  },
  {
    "ph": "s", "id": 229715, "pid": 76337, "tid": -914061504, "ts": 1716454224974842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225059753, "dur": 3,
    "args": {
      "External id": 229717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229717, "pid": 5, "tid": 7, "ts": 1716454225059753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974852, "dur": 6,
    "args": {
      "External id": 229717,
      "cbid": 211, "correlation": 229717
    }
  },
  {
    "ph": "s", "id": 229717, "pid": 76337, "tid": -914061504, "ts": 1716454224974852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974860, "dur": 0,
    "args": {
      "External id": 229718,
      "cbid": 51, "correlation": 229718
    }
  },
  {
    "ph": "s", "id": 229718, "pid": 76337, "tid": -914061504, "ts": 1716454224974860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225059758, "dur": 688,
    "args": {
      "External id": 229719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229719,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229719, "pid": 5, "tid": 7, "ts": 1716454225059758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974861, "dur": 5,
    "args": {
      "External id": 229719,
      "cbid": 211, "correlation": 229719
    }
  },
  {
    "ph": "s", "id": 229719, "pid": 76337, "tid": -914061504, "ts": 1716454224974861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225060447, "dur": 58,
    "args": {
      "External id": 229724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229724,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229724, "pid": 5, "tid": 7, "ts": 1716454225060447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974890, "dur": 8,
    "args": {
      "External id": 229724,
      "cbid": 211, "correlation": 229724
    }
  },
  {
    "ph": "s", "id": 229724, "pid": 76337, "tid": -914061504, "ts": 1716454224974890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974946, "dur": 0,
    "args": {
      "External id": 229734,
      "cbid": 317, "correlation": 229734
    }
  },
  {
    "ph": "f", "id": 229734, "pid": 76337, "tid": -914061504, "ts": 1716454224974946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974947, "dur": 0,
    "args": {
      "External id": 229735,
      "cbid": 203, "correlation": 229735
    }
  },
  {
    "ph": "f", "id": 229735, "pid": 76337, "tid": -914061504, "ts": 1716454224974947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974948, "dur": 0,
    "args": {
      "External id": 229736,
      "cbid": 205, "correlation": 229736
    }
  },
  {
    "ph": "f", "id": 229736, "pid": 76337, "tid": -914061504, "ts": 1716454224974948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225060506, "dur": 4,
    "args": {
      "External id": 229740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229740, "pid": 5, "tid": 7, "ts": 1716454225060506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974963, "dur": 20,
    "args": {
      "External id": 229740,
      "cbid": 211, "correlation": 229740
    }
  },
  {
    "ph": "s", "id": 229740, "pid": 76337, "tid": -914061504, "ts": 1716454224974963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974989, "dur": 0,
    "args": {
      "External id": 229741,
      "cbid": 51, "correlation": 229741
    }
  },
  {
    "ph": "s", "id": 229741, "pid": 76337, "tid": -914061504, "ts": 1716454224974989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225060511, "dur": 261,
    "args": {
      "External id": 229742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229742,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229742, "pid": 5, "tid": 7, "ts": 1716454225060511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224974989, "dur": 7,
    "args": {
      "External id": 229742,
      "cbid": 211, "correlation": 229742
    }
  },
  {
    "ph": "s", "id": 229742, "pid": 76337, "tid": -914061504, "ts": 1716454224974989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225060773, "dur": 58,
    "args": {
      "External id": 229747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229747, "pid": 5, "tid": 7, "ts": 1716454225060773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975018, "dur": 8,
    "args": {
      "External id": 229747,
      "cbid": 211, "correlation": 229747
    }
  },
  {
    "ph": "s", "id": 229747, "pid": 76337, "tid": -914061504, "ts": 1716454224975018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225060832, "dur": 50,
    "args": {
      "External id": 229755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229755,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229755, "pid": 5, "tid": 7, "ts": 1716454225060832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975046, "dur": 8,
    "args": {
      "External id": 229755,
      "cbid": 211, "correlation": 229755
    }
  },
  {
    "ph": "s", "id": 229755, "pid": 76337, "tid": -914061504, "ts": 1716454224975046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225060883, "dur": 35,
    "args": {
      "External id": 229763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229763,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229763, "pid": 5, "tid": 7, "ts": 1716454225060883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975075, "dur": 9,
    "args": {
      "External id": 229763,
      "cbid": 211, "correlation": 229763
    }
  },
  {
    "ph": "s", "id": 229763, "pid": 76337, "tid": -914061504, "ts": 1716454224975075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225060920, "dur": 51,
    "args": {
      "External id": 229783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229783,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 229783, "pid": 5, "tid": 7, "ts": 1716454225060920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975158, "dur": 13,
    "args": {
      "External id": 229783,
      "cbid": 211, "correlation": 229783
    }
  },
  {
    "ph": "s", "id": 229783, "pid": 76337, "tid": -914061504, "ts": 1716454224975158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225060972, "dur": 4,
    "args": {
      "External id": 229795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229795,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 229795, "pid": 5, "tid": 7, "ts": 1716454225060972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975181, "dur": 7,
    "args": {
      "External id": 229795,
      "cbid": 211, "correlation": 229795
    }
  },
  {
    "ph": "s", "id": 229795, "pid": 76337, "tid": -914061504, "ts": 1716454224975181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225060977, "dur": 54,
    "args": {
      "External id": 229798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229798,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229798, "pid": 5, "tid": 7, "ts": 1716454225060977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975199, "dur": 6,
    "args": {
      "External id": 229798,
      "cbid": 211, "correlation": 229798
    }
  },
  {
    "ph": "s", "id": 229798, "pid": 76337, "tid": -914061504, "ts": 1716454224975199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975256, "dur": 0,
    "args": {
      "External id": 229809,
      "cbid": 317, "correlation": 229809
    }
  },
  {
    "ph": "f", "id": 229809, "pid": 76337, "tid": -914061504, "ts": 1716454224975256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975257, "dur": 0,
    "args": {
      "External id": 229810,
      "cbid": 203, "correlation": 229810
    }
  },
  {
    "ph": "f", "id": 229810, "pid": 76337, "tid": -914061504, "ts": 1716454224975257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975258, "dur": 0,
    "args": {
      "External id": 229811,
      "cbid": 205, "correlation": 229811
    }
  },
  {
    "ph": "f", "id": 229811, "pid": 76337, "tid": -914061504, "ts": 1716454224975258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975289, "dur": 2,
    "args": {
      "External id": 229815,
      "cbid": 251, "correlation": 229815
    }
  },
  {
    "ph": "f", "id": 229815, "pid": 76337, "tid": -914061504, "ts": 1716454224975289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975292, "dur": 1,
    "args": {
      "External id": 229816,
      "cbid": 251, "correlation": 229816
    }
  },
  {
    "ph": "f", "id": 229816, "pid": 76337, "tid": -914061504, "ts": 1716454224975292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975293, "dur": 0,
    "args": {
      "External id": 229817,
      "cbid": 251, "correlation": 229817
    }
  },
  {
    "ph": "f", "id": 229817, "pid": 76337, "tid": -914061504, "ts": 1716454224975293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975294, "dur": 1,
    "args": {
      "External id": 229818,
      "cbid": 251, "correlation": 229818
    }
  },
  {
    "ph": "f", "id": 229818, "pid": 76337, "tid": -914061504, "ts": 1716454224975294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975296, "dur": 1,
    "args": {
      "External id": 229819,
      "cbid": 251, "correlation": 229819
    }
  },
  {
    "ph": "f", "id": 229819, "pid": 76337, "tid": -914061504, "ts": 1716454224975296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975298, "dur": 1,
    "args": {
      "External id": 229820,
      "cbid": 251, "correlation": 229820
    }
  },
  {
    "ph": "f", "id": 229820, "pid": 76337, "tid": -914061504, "ts": 1716454224975298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975300, "dur": 1,
    "args": {
      "External id": 229821,
      "cbid": 251, "correlation": 229821
    }
  },
  {
    "ph": "f", "id": 229821, "pid": 76337, "tid": -914061504, "ts": 1716454224975300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975301, "dur": 1,
    "args": {
      "External id": 229822,
      "cbid": 251, "correlation": 229822
    }
  },
  {
    "ph": "f", "id": 229822, "pid": 76337, "tid": -914061504, "ts": 1716454224975301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975304, "dur": 0,
    "args": {
      "External id": 229823,
      "cbid": 251, "correlation": 229823
    }
  },
  {
    "ph": "f", "id": 229823, "pid": 76337, "tid": -914061504, "ts": 1716454224975304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225061033, "dur": 111,
    "args": {
      "External id": 229824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229824,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 229824, "pid": 5, "tid": 7, "ts": 1716454225061033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975306, "dur": 13,
    "args": {
      "External id": 229824,
      "cbid": 211, "correlation": 229824
    }
  },
  {
    "ph": "s", "id": 229824, "pid": 76337, "tid": -914061504, "ts": 1716454224975306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225061145, "dur": 58,
    "args": {
      "External id": 229830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229830, "pid": 5, "tid": 7, "ts": 1716454225061145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975343, "dur": 9,
    "args": {
      "External id": 229830,
      "cbid": 211, "correlation": 229830
    }
  },
  {
    "ph": "s", "id": 229830, "pid": 76337, "tid": -914061504, "ts": 1716454224975343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225061205, "dur": 567,
    "args": {
      "External id": 229839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229839,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229839, "pid": 5, "tid": 7, "ts": 1716454225061205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975425, "dur": 14,
    "args": {
      "External id": 229839,
      "cbid": 211, "correlation": 229839
    }
  },
  {
    "ph": "s", "id": 229839, "pid": 76337, "tid": -914061504, "ts": 1716454224975425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225061774, "dur": 177,
    "args": {
      "External id": 229861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229861,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229861, "pid": 5, "tid": 7, "ts": 1716454225061774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975484, "dur": 10,
    "args": {
      "External id": 229861,
      "cbid": 211, "correlation": 229861
    }
  },
  {
    "ph": "s", "id": 229861, "pid": 76337, "tid": -914061504, "ts": 1716454224975484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975572, "dur": 1,
    "args": {
      "External id": 229872,
      "cbid": 251, "correlation": 229872
    }
  },
  {
    "ph": "f", "id": 229872, "pid": 76337, "tid": -914061504, "ts": 1716454224975572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225061953, "dur": 195,
    "args": {
      "External id": 229873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229873,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229873, "pid": 5, "tid": 7, "ts": 1716454225061953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975577, "dur": 14,
    "args": {
      "External id": 229873,
      "cbid": 211, "correlation": 229873
    }
  },
  {
    "ph": "s", "id": 229873, "pid": 76337, "tid": -914061504, "ts": 1716454224975577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975646, "dur": 1,
    "args": {
      "External id": 229884,
      "cbid": 251, "correlation": 229884
    }
  },
  {
    "ph": "f", "id": 229884, "pid": 76337, "tid": -914061504, "ts": 1716454224975646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225062148, "dur": 184,
    "args": {
      "External id": 229885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229885,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229885, "pid": 5, "tid": 7, "ts": 1716454225062148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975650, "dur": 11,
    "args": {
      "External id": 229885,
      "cbid": 211, "correlation": 229885
    }
  },
  {
    "ph": "s", "id": 229885, "pid": 76337, "tid": -914061504, "ts": 1716454224975650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975714, "dur": 1,
    "args": {
      "External id": 229896,
      "cbid": 251, "correlation": 229896
    }
  },
  {
    "ph": "f", "id": 229896, "pid": 76337, "tid": -914061504, "ts": 1716454224975714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225062334, "dur": 185,
    "args": {
      "External id": 229897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229897,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229897, "pid": 5, "tid": 7, "ts": 1716454225062334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975718, "dur": 12,
    "args": {
      "External id": 229897,
      "cbid": 211, "correlation": 229897
    }
  },
  {
    "ph": "s", "id": 229897, "pid": 76337, "tid": -914061504, "ts": 1716454224975718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225062520, "dur": 18129,
    "args": {
      "External id": 229918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229918,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 229918, "pid": 5, "tid": 7, "ts": 1716454225062520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975801, "dur": 13,
    "args": {
      "External id": 229918,
      "cbid": 211, "correlation": 229918
    }
  },
  {
    "ph": "s", "id": 229918, "pid": 76337, "tid": -914061504, "ts": 1716454224975801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975900, "dur": 1,
    "args": {
      "External id": 229936,
      "cbid": 251, "correlation": 229936
    }
  },
  {
    "ph": "f", "id": 229936, "pid": 76337, "tid": -914061504, "ts": 1716454224975900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225080651, "dur": 201,
    "args": {
      "External id": 229938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229938,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229938, "pid": 5, "tid": 7, "ts": 1716454225080651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975906, "dur": 13,
    "args": {
      "External id": 229938,
      "cbid": 211, "correlation": 229938
    }
  },
  {
    "ph": "s", "id": 229938, "pid": 76337, "tid": -914061504, "ts": 1716454224975906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225080853, "dur": 66,
    "args": {
      "External id": 229946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229946,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229946, "pid": 5, "tid": 7, "ts": 1716454225080853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224975987, "dur": 12,
    "args": {
      "External id": 229946,
      "cbid": 211, "correlation": 229946
    }
  },
  {
    "ph": "s", "id": 229946, "pid": 76337, "tid": -914061504, "ts": 1716454224975987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225080921, "dur": 97,
    "args": {
      "External id": 229954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229954,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229954, "pid": 5, "tid": 7, "ts": 1716454225080921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976027, "dur": 10,
    "args": {
      "External id": 229954,
      "cbid": 211, "correlation": 229954
    }
  },
  {
    "ph": "s", "id": 229954, "pid": 76337, "tid": -914061504, "ts": 1716454224976027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225081019, "dur": 54,
    "args": {
      "External id": 229965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229965, "pid": 5, "tid": 7, "ts": 1716454225081019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976100, "dur": 13,
    "args": {
      "External id": 229965,
      "cbid": 211, "correlation": 229965
    }
  },
  {
    "ph": "s", "id": 229965, "pid": 76337, "tid": -914061504, "ts": 1716454224976100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225081075, "dur": 90,
    "args": {
      "External id": 229987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229987,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 229987, "pid": 5, "tid": 7, "ts": 1716454225081075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976132, "dur": 8,
    "args": {
      "External id": 229987,
      "cbid": 211, "correlation": 229987
    }
  },
  {
    "ph": "s", "id": 229987, "pid": 76337, "tid": -914061504, "ts": 1716454224976132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976214, "dur": 1,
    "args": {
      "External id": 229998,
      "cbid": 251, "correlation": 229998
    }
  },
  {
    "ph": "f", "id": 229998, "pid": 76337, "tid": -914061504, "ts": 1716454224976214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225081166, "dur": 105,
    "args": {
      "External id": 229999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 229999,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 229999, "pid": 5, "tid": 7, "ts": 1716454225081166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976219, "dur": 13,
    "args": {
      "External id": 229999,
      "cbid": 211, "correlation": 229999
    }
  },
  {
    "ph": "s", "id": 229999, "pid": 76337, "tid": -914061504, "ts": 1716454224976219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976295, "dur": 1,
    "args": {
      "External id": 230010,
      "cbid": 251, "correlation": 230010
    }
  },
  {
    "ph": "f", "id": 230010, "pid": 76337, "tid": -914061504, "ts": 1716454224976295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976299, "dur": 0,
    "args": {
      "External id": 230011,
      "cbid": 251, "correlation": 230011
    }
  },
  {
    "ph": "f", "id": 230011, "pid": 76337, "tid": -914061504, "ts": 1716454224976299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225081273, "dur": 10,
    "args": {
      "External id": 230012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230012,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 230012, "pid": 5, "tid": 7, "ts": 1716454225081273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976301, "dur": 13,
    "args": {
      "External id": 230012,
      "cbid": 211, "correlation": 230012
    }
  },
  {
    "ph": "s", "id": 230012, "pid": 76337, "tid": -914061504, "ts": 1716454224976301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225081285, "dur": 5,
    "args": {
      "External id": 230014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230014,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230014, "pid": 5, "tid": 7, "ts": 1716454225081285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976317, "dur": 7,
    "args": {
      "External id": 230014,
      "cbid": 211, "correlation": 230014
    }
  },
  {
    "ph": "s", "id": 230014, "pid": 76337, "tid": -914061504, "ts": 1716454224976317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976381, "dur": 1,
    "args": {
      "External id": 230025,
      "cbid": 251, "correlation": 230025
    }
  },
  {
    "ph": "f", "id": 230025, "pid": 76337, "tid": -914061504, "ts": 1716454224976381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976384, "dur": 0,
    "args": {
      "External id": 230026,
      "cbid": 251, "correlation": 230026
    }
  },
  {
    "ph": "f", "id": 230026, "pid": 76337, "tid": -914061504, "ts": 1716454224976384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225081291, "dur": 6,
    "args": {
      "External id": 230027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230027,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 230027, "pid": 5, "tid": 7, "ts": 1716454225081291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976386, "dur": 12,
    "args": {
      "External id": 230027,
      "cbid": 211, "correlation": 230027
    }
  },
  {
    "ph": "s", "id": 230027, "pid": 76337, "tid": -914061504, "ts": 1716454224976386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225081298, "dur": 3,
    "args": {
      "External id": 230029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230029,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230029, "pid": 5, "tid": 7, "ts": 1716454225081298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976399, "dur": 5,
    "args": {
      "External id": 230029,
      "cbid": 211, "correlation": 230029
    }
  },
  {
    "ph": "s", "id": 230029, "pid": 76337, "tid": -914061504, "ts": 1716454224976399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225081303, "dur": 152,
    "args": {
      "External id": 230050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230050,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230050, "pid": 5, "tid": 7, "ts": 1716454225081303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976472, "dur": 13,
    "args": {
      "External id": 230050,
      "cbid": 211, "correlation": 230050
    }
  },
  {
    "ph": "s", "id": 230050, "pid": 76337, "tid": -914061504, "ts": 1716454224976472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976569, "dur": 1,
    "args": {
      "External id": 230068,
      "cbid": 251, "correlation": 230068
    }
  },
  {
    "ph": "f", "id": 230068, "pid": 76337, "tid": -914061504, "ts": 1716454224976569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225081456, "dur": 106,
    "args": {
      "External id": 230070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230070,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230070, "pid": 5, "tid": 7, "ts": 1716454225081456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976575, "dur": 14,
    "args": {
      "External id": 230070,
      "cbid": 211, "correlation": 230070
    }
  },
  {
    "ph": "s", "id": 230070, "pid": 76337, "tid": -914061504, "ts": 1716454224976575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225081563, "dur": 34,
    "args": {
      "External id": 230078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230078,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230078, "pid": 5, "tid": 7, "ts": 1716454225081563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976645, "dur": 12,
    "args": {
      "External id": 230078,
      "cbid": 211, "correlation": 230078
    }
  },
  {
    "ph": "s", "id": 230078, "pid": 76337, "tid": -914061504, "ts": 1716454224976645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225081599, "dur": 65,
    "args": {
      "External id": 230086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230086, "pid": 5, "tid": 7, "ts": 1716454225081599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976686, "dur": 9,
    "args": {
      "External id": 230086,
      "cbid": 211, "correlation": 230086
    }
  },
  {
    "ph": "s", "id": 230086, "pid": 76337, "tid": -914061504, "ts": 1716454224976686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225081665, "dur": 90,
    "args": {
      "External id": 230108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230108,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230108, "pid": 5, "tid": 7, "ts": 1716454225081665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976739, "dur": 11,
    "args": {
      "External id": 230108,
      "cbid": 211, "correlation": 230108
    }
  },
  {
    "ph": "s", "id": 230108, "pid": 76337, "tid": -914061504, "ts": 1716454224976739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976827, "dur": 1,
    "args": {
      "External id": 230124,
      "cbid": 251, "correlation": 230124
    }
  },
  {
    "ph": "f", "id": 230124, "pid": 76337, "tid": -914061504, "ts": 1716454224976827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225081756, "dur": 564,
    "args": {
      "External id": 230126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230126,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230126, "pid": 5, "tid": 7, "ts": 1716454225081756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976833, "dur": 13,
    "args": {
      "External id": 230126,
      "cbid": 211, "correlation": 230126
    }
  },
  {
    "ph": "s", "id": 230126, "pid": 76337, "tid": -914061504, "ts": 1716454224976833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225082322, "dur": 240,
    "args": {
      "External id": 230134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230134, "pid": 5, "tid": 7, "ts": 1716454225082322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976899, "dur": 12,
    "args": {
      "External id": 230134,
      "cbid": 211, "correlation": 230134
    }
  },
  {
    "ph": "s", "id": 230134, "pid": 76337, "tid": -914061504, "ts": 1716454224976899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225082563, "dur": 252,
    "args": {
      "External id": 230142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230142, "pid": 5, "tid": 7, "ts": 1716454225082563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224976930, "dur": 9,
    "args": {
      "External id": 230142,
      "cbid": 211, "correlation": 230142
    }
  },
  {
    "ph": "s", "id": 230142, "pid": 76337, "tid": -914061504, "ts": 1716454224976930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977024, "dur": 1,
    "args": {
      "External id": 230158,
      "cbid": 251, "correlation": 230158
    }
  },
  {
    "ph": "f", "id": 230158, "pid": 76337, "tid": -914061504, "ts": 1716454224977024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977029, "dur": 0,
    "args": {
      "External id": 230160,
      "cbid": 251, "correlation": 230160
    }
  },
  {
    "ph": "f", "id": 230160, "pid": 76337, "tid": -914061504, "ts": 1716454224977029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225082816, "dur": 353,
    "args": {
      "External id": 230161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230161,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 230161, "pid": 5, "tid": 7, "ts": 1716454225082816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977032, "dur": 14,
    "args": {
      "External id": 230161,
      "cbid": 211, "correlation": 230161
    }
  },
  {
    "ph": "s", "id": 230161, "pid": 76337, "tid": -914061504, "ts": 1716454224977032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225083171, "dur": 50,
    "args": {
      "External id": 230169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230169,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230169, "pid": 5, "tid": 7, "ts": 1716454225083171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977075, "dur": 11,
    "args": {
      "External id": 230169,
      "cbid": 211, "correlation": 230169
    }
  },
  {
    "ph": "s", "id": 230169, "pid": 76337, "tid": -914061504, "ts": 1716454224977075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225083222, "dur": 155,
    "args": {
      "External id": 230180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230180, "pid": 5, "tid": 7, "ts": 1716454225083222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977143, "dur": 12,
    "args": {
      "External id": 230180,
      "cbid": 211, "correlation": 230180
    }
  },
  {
    "ph": "s", "id": 230180, "pid": 76337, "tid": -914061504, "ts": 1716454224977143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977208, "dur": 0,
    "args": {
      "External id": 230192,
      "cbid": 317, "correlation": 230192
    }
  },
  {
    "ph": "f", "id": 230192, "pid": 76337, "tid": -914061504, "ts": 1716454224977208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977209, "dur": 0,
    "args": {
      "External id": 230193,
      "cbid": 203, "correlation": 230193
    }
  },
  {
    "ph": "f", "id": 230193, "pid": 76337, "tid": -914061504, "ts": 1716454224977209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977210, "dur": 0,
    "args": {
      "External id": 230194,
      "cbid": 205, "correlation": 230194
    }
  },
  {
    "ph": "f", "id": 230194, "pid": 76337, "tid": -914061504, "ts": 1716454224977210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977235, "dur": 1,
    "args": {
      "External id": 230198,
      "cbid": 251, "correlation": 230198
    }
  },
  {
    "ph": "f", "id": 230198, "pid": 76337, "tid": -914061504, "ts": 1716454224977235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977236, "dur": 0,
    "args": {
      "External id": 230199,
      "cbid": 251, "correlation": 230199
    }
  },
  {
    "ph": "f", "id": 230199, "pid": 76337, "tid": -914061504, "ts": 1716454224977236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977237, "dur": 0,
    "args": {
      "External id": 230200,
      "cbid": 251, "correlation": 230200
    }
  },
  {
    "ph": "f", "id": 230200, "pid": 76337, "tid": -914061504, "ts": 1716454224977237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977238, "dur": 0,
    "args": {
      "External id": 230201,
      "cbid": 251, "correlation": 230201
    }
  },
  {
    "ph": "f", "id": 230201, "pid": 76337, "tid": -914061504, "ts": 1716454224977238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977239, "dur": 1,
    "args": {
      "External id": 230202,
      "cbid": 251, "correlation": 230202
    }
  },
  {
    "ph": "f", "id": 230202, "pid": 76337, "tid": -914061504, "ts": 1716454224977239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977240, "dur": 0,
    "args": {
      "External id": 230203,
      "cbid": 251, "correlation": 230203
    }
  },
  {
    "ph": "f", "id": 230203, "pid": 76337, "tid": -914061504, "ts": 1716454224977240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977242, "dur": 0,
    "args": {
      "External id": 230204,
      "cbid": 251, "correlation": 230204
    }
  },
  {
    "ph": "f", "id": 230204, "pid": 76337, "tid": -914061504, "ts": 1716454224977242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977242, "dur": 0,
    "args": {
      "External id": 230205,
      "cbid": 251, "correlation": 230205
    }
  },
  {
    "ph": "f", "id": 230205, "pid": 76337, "tid": -914061504, "ts": 1716454224977242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977244, "dur": 0,
    "args": {
      "External id": 230206,
      "cbid": 251, "correlation": 230206
    }
  },
  {
    "ph": "f", "id": 230206, "pid": 76337, "tid": -914061504, "ts": 1716454224977244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225083378, "dur": 112,
    "args": {
      "External id": 230207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230207,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230207, "pid": 5, "tid": 7, "ts": 1716454225083378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977246, "dur": 12,
    "args": {
      "External id": 230207,
      "cbid": 211, "correlation": 230207
    }
  },
  {
    "ph": "s", "id": 230207, "pid": 76337, "tid": -914061504, "ts": 1716454224977246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225083491, "dur": 58,
    "args": {
      "External id": 230213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230213, "pid": 5, "tid": 7, "ts": 1716454225083491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977280, "dur": 9,
    "args": {
      "External id": 230213,
      "cbid": 211, "correlation": 230213
    }
  },
  {
    "ph": "s", "id": 230213, "pid": 76337, "tid": -914061504, "ts": 1716454224977280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225083551, "dur": 50,
    "args": {
      "External id": 230221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230221,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230221, "pid": 5, "tid": 7, "ts": 1716454225083551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977312, "dur": 8,
    "args": {
      "External id": 230221,
      "cbid": 211, "correlation": 230221
    }
  },
  {
    "ph": "s", "id": 230221, "pid": 76337, "tid": -914061504, "ts": 1716454224977312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225083603, "dur": 97,
    "args": {
      "External id": 230230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230230,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230230, "pid": 5, "tid": 7, "ts": 1716454225083603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977351, "dur": 10,
    "args": {
      "External id": 230230,
      "cbid": 211, "correlation": 230230
    }
  },
  {
    "ph": "s", "id": 230230, "pid": 76337, "tid": -914061504, "ts": 1716454224977351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225083701, "dur": 90,
    "args": {
      "External id": 230250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230250,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 230250, "pid": 5, "tid": 7, "ts": 1716454225083701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977422, "dur": 12,
    "args": {
      "External id": 230250,
      "cbid": 211, "correlation": 230250
    }
  },
  {
    "ph": "s", "id": 230250, "pid": 76337, "tid": -914061504, "ts": 1716454224977422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225083793, "dur": 4,
    "args": {
      "External id": 230262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230262,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230262, "pid": 5, "tid": 7, "ts": 1716454225083793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977444, "dur": 32,
    "args": {
      "External id": 230262,
      "cbid": 211, "correlation": 230262
    }
  },
  {
    "ph": "s", "id": 230262, "pid": 76337, "tid": -914061504, "ts": 1716454224977444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225083798, "dur": 107,
    "args": {
      "External id": 230265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230265, "pid": 5, "tid": 7, "ts": 1716454225083798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977489, "dur": 7,
    "args": {
      "External id": 230265,
      "cbid": 211, "correlation": 230265
    }
  },
  {
    "ph": "s", "id": 230265, "pid": 76337, "tid": -914061504, "ts": 1716454224977489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225083907, "dur": 69,
    "args": {
      "External id": 230274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230274,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230274, "pid": 5, "tid": 7, "ts": 1716454225083907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977529, "dur": 10,
    "args": {
      "External id": 230274,
      "cbid": 211, "correlation": 230274
    }
  },
  {
    "ph": "s", "id": 230274, "pid": 76337, "tid": -914061504, "ts": 1716454224977529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977581, "dur": 0,
    "args": {
      "External id": 230284,
      "cbid": 317, "correlation": 230284
    }
  },
  {
    "ph": "f", "id": 230284, "pid": 76337, "tid": -914061504, "ts": 1716454224977581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977582, "dur": 0,
    "args": {
      "External id": 230285,
      "cbid": 203, "correlation": 230285
    }
  },
  {
    "ph": "f", "id": 230285, "pid": 76337, "tid": -914061504, "ts": 1716454224977582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977583, "dur": 0,
    "args": {
      "External id": 230286,
      "cbid": 205, "correlation": 230286
    }
  },
  {
    "ph": "f", "id": 230286, "pid": 76337, "tid": -914061504, "ts": 1716454224977583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225083978, "dur": 76,
    "args": {
      "External id": 230290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230290,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230290, "pid": 5, "tid": 7, "ts": 1716454225083978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977598, "dur": 11,
    "args": {
      "External id": 230290,
      "cbid": 211, "correlation": 230290
    }
  },
  {
    "ph": "s", "id": 230290, "pid": 76337, "tid": -914061504, "ts": 1716454224977598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225084055, "dur": 23,
    "args": {
      "External id": 230292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230292,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230292, "pid": 5, "tid": 7, "ts": 1716454225084055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977611, "dur": 5,
    "args": {
      "External id": 230292,
      "cbid": 211, "correlation": 230292
    }
  },
  {
    "ph": "s", "id": 230292, "pid": 76337, "tid": -914061504, "ts": 1716454224977611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225084080, "dur": 4,
    "args": {
      "External id": 230294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230294, "pid": 5, "tid": 7, "ts": 1716454225084080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977621, "dur": 6,
    "args": {
      "External id": 230294,
      "cbid": 211, "correlation": 230294
    }
  },
  {
    "ph": "s", "id": 230294, "pid": 76337, "tid": -914061504, "ts": 1716454224977621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977630, "dur": 0,
    "args": {
      "External id": 230295,
      "cbid": 51, "correlation": 230295
    }
  },
  {
    "ph": "s", "id": 230295, "pid": 76337, "tid": -914061504, "ts": 1716454224977630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225084085, "dur": 1344,
    "args": {
      "External id": 230296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230296,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230296, "pid": 5, "tid": 7, "ts": 1716454225084085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977631, "dur": 5,
    "args": {
      "External id": 230296,
      "cbid": 211, "correlation": 230296
    }
  },
  {
    "ph": "s", "id": 230296, "pid": 76337, "tid": -914061504, "ts": 1716454224977631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225085430, "dur": 58,
    "args": {
      "External id": 230301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230301, "pid": 5, "tid": 7, "ts": 1716454225085430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977659, "dur": 9,
    "args": {
      "External id": 230301,
      "cbid": 211, "correlation": 230301
    }
  },
  {
    "ph": "s", "id": 230301, "pid": 76337, "tid": -914061504, "ts": 1716454224977659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225085490, "dur": 3,
    "args": {
      "External id": 230309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230309,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230309, "pid": 5, "tid": 7, "ts": 1716454225085490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977703, "dur": 9,
    "args": {
      "External id": 230309,
      "cbid": 211, "correlation": 230309
    }
  },
  {
    "ph": "s", "id": 230309, "pid": 76337, "tid": -914061504, "ts": 1716454224977703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977768, "dur": 1,
    "args": {
      "External id": 230325,
      "cbid": 251, "correlation": 230325
    }
  },
  {
    "ph": "f", "id": 230325, "pid": 76337, "tid": -914061504, "ts": 1716454224977768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977774, "dur": 0,
    "args": {
      "External id": 230327,
      "cbid": 251, "correlation": 230327
    }
  },
  {
    "ph": "f", "id": 230327, "pid": 76337, "tid": -914061504, "ts": 1716454224977774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225085494, "dur": 11,
    "args": {
      "External id": 230328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230328,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 230328, "pid": 5, "tid": 7, "ts": 1716454225085494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977775, "dur": 12,
    "args": {
      "External id": 230328,
      "cbid": 211, "correlation": 230328
    }
  },
  {
    "ph": "s", "id": 230328, "pid": 76337, "tid": -914061504, "ts": 1716454224977775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225085507, "dur": 5,
    "args": {
      "External id": 230330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230330,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230330, "pid": 5, "tid": 7, "ts": 1716454225085507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224977790, "dur": 178,
    "args": {
      "External id": 230330,
      "cbid": 211, "correlation": 230330
    }
  },
  {
    "ph": "s", "id": 230330, "pid": 76337, "tid": -914061504, "ts": 1716454224977790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225085513, "dur": 53,
    "args": {
      "External id": 230340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230340, "pid": 5, "tid": 7, "ts": 1716454225085513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978027, "dur": 13,
    "args": {
      "External id": 230340,
      "cbid": 211, "correlation": 230340
    }
  },
  {
    "ph": "s", "id": 230340, "pid": 76337, "tid": -914061504, "ts": 1716454224978027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225085568, "dur": 51,
    "args": {
      "External id": 230360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230360,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 230360, "pid": 5, "tid": 7, "ts": 1716454225085568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978094, "dur": 11,
    "args": {
      "External id": 230360,
      "cbid": 211, "correlation": 230360
    }
  },
  {
    "ph": "s", "id": 230360, "pid": 76337, "tid": -914061504, "ts": 1716454224978094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225085620, "dur": 4,
    "args": {
      "External id": 230372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230372,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230372, "pid": 5, "tid": 7, "ts": 1716454225085620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978115, "dur": 6,
    "args": {
      "External id": 230372,
      "cbid": 211, "correlation": 230372
    }
  },
  {
    "ph": "s", "id": 230372, "pid": 76337, "tid": -914061504, "ts": 1716454224978115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225085625, "dur": 54,
    "args": {
      "External id": 230375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230375,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230375, "pid": 5, "tid": 7, "ts": 1716454225085625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978134, "dur": 7,
    "args": {
      "External id": 230375,
      "cbid": 211, "correlation": 230375
    }
  },
  {
    "ph": "s", "id": 230375, "pid": 76337, "tid": -914061504, "ts": 1716454224978134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225085681, "dur": 36,
    "args": {
      "External id": 230384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230384,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230384, "pid": 5, "tid": 7, "ts": 1716454225085681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978175, "dur": 10,
    "args": {
      "External id": 230384,
      "cbid": 211, "correlation": 230384
    }
  },
  {
    "ph": "s", "id": 230384, "pid": 76337, "tid": -914061504, "ts": 1716454224978175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978238, "dur": 0,
    "args": {
      "External id": 230394,
      "cbid": 317, "correlation": 230394
    }
  },
  {
    "ph": "f", "id": 230394, "pid": 76337, "tid": -914061504, "ts": 1716454224978238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978238, "dur": 0,
    "args": {
      "External id": 230395,
      "cbid": 203, "correlation": 230395
    }
  },
  {
    "ph": "f", "id": 230395, "pid": 76337, "tid": -914061504, "ts": 1716454224978238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978239, "dur": 0,
    "args": {
      "External id": 230396,
      "cbid": 205, "correlation": 230396
    }
  },
  {
    "ph": "f", "id": 230396, "pid": 76337, "tid": -914061504, "ts": 1716454224978239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225085718, "dur": 41,
    "args": {
      "External id": 230400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230400,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230400, "pid": 5, "tid": 7, "ts": 1716454225085718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978253, "dur": 13,
    "args": {
      "External id": 230400,
      "cbid": 211, "correlation": 230400
    }
  },
  {
    "ph": "s", "id": 230400, "pid": 76337, "tid": -914061504, "ts": 1716454224978253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225085761, "dur": 14,
    "args": {
      "External id": 230402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230402,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230402, "pid": 5, "tid": 7, "ts": 1716454225085761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978268, "dur": 5,
    "args": {
      "External id": 230402,
      "cbid": 211, "correlation": 230402
    }
  },
  {
    "ph": "s", "id": 230402, "pid": 76337, "tid": -914061504, "ts": 1716454224978268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225085775, "dur": 3,
    "args": {
      "External id": 230404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230404,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230404, "pid": 5, "tid": 7, "ts": 1716454225085775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978278, "dur": 5,
    "args": {
      "External id": 230404,
      "cbid": 211, "correlation": 230404
    }
  },
  {
    "ph": "s", "id": 230404, "pid": 76337, "tid": -914061504, "ts": 1716454224978278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978286, "dur": 0,
    "args": {
      "External id": 230405,
      "cbid": 51, "correlation": 230405
    }
  },
  {
    "ph": "s", "id": 230405, "pid": 76337, "tid": -914061504, "ts": 1716454224978286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225085780, "dur": 687,
    "args": {
      "External id": 230406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230406,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230406, "pid": 5, "tid": 7, "ts": 1716454225085780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978287, "dur": 5,
    "args": {
      "External id": 230406,
      "cbid": 211, "correlation": 230406
    }
  },
  {
    "ph": "s", "id": 230406, "pid": 76337, "tid": -914061504, "ts": 1716454224978287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225086469, "dur": 58,
    "args": {
      "External id": 230411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230411, "pid": 5, "tid": 7, "ts": 1716454225086469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978315, "dur": 8,
    "args": {
      "External id": 230411,
      "cbid": 211, "correlation": 230411
    }
  },
  {
    "ph": "s", "id": 230411, "pid": 76337, "tid": -914061504, "ts": 1716454224978315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978372, "dur": 0,
    "args": {
      "External id": 230421,
      "cbid": 317, "correlation": 230421
    }
  },
  {
    "ph": "f", "id": 230421, "pid": 76337, "tid": -914061504, "ts": 1716454224978372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978373, "dur": 0,
    "args": {
      "External id": 230422,
      "cbid": 203, "correlation": 230422
    }
  },
  {
    "ph": "f", "id": 230422, "pid": 76337, "tid": -914061504, "ts": 1716454224978373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978374, "dur": 0,
    "args": {
      "External id": 230423,
      "cbid": 205, "correlation": 230423
    }
  },
  {
    "ph": "f", "id": 230423, "pid": 76337, "tid": -914061504, "ts": 1716454224978374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225086528, "dur": 74,
    "args": {
      "External id": 230427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230427,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230427, "pid": 5, "tid": 7, "ts": 1716454225086528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978387, "dur": 11,
    "args": {
      "External id": 230427,
      "cbid": 211, "correlation": 230427
    }
  },
  {
    "ph": "s", "id": 230427, "pid": 76337, "tid": -914061504, "ts": 1716454224978387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225086603, "dur": 204,
    "args": {
      "External id": 230429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230429,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230429, "pid": 5, "tid": 7, "ts": 1716454225086603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978407, "dur": 9,
    "args": {
      "External id": 230429,
      "cbid": 211, "correlation": 230429
    }
  },
  {
    "ph": "s", "id": 230429, "pid": 76337, "tid": -914061504, "ts": 1716454224978407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225086809, "dur": 39,
    "args": {
      "External id": 230431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230431,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230431, "pid": 5, "tid": 7, "ts": 1716454225086809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978420, "dur": 478,
    "args": {
      "External id": 230431,
      "cbid": 211, "correlation": 230431
    }
  },
  {
    "ph": "s", "id": 230431, "pid": 76337, "tid": -914061504, "ts": 1716454224978420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225086849, "dur": 59,
    "args": {
      "External id": 230437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230437, "pid": 5, "tid": 7, "ts": 1716454225086849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978919, "dur": 9,
    "args": {
      "External id": 230437,
      "cbid": 211, "correlation": 230437
    }
  },
  {
    "ph": "s", "id": 230437, "pid": 76337, "tid": -914061504, "ts": 1716454224978919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225086909, "dur": 49,
    "args": {
      "External id": 230445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230445,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230445, "pid": 5, "tid": 7, "ts": 1716454225086909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978949, "dur": 8,
    "args": {
      "External id": 230445,
      "cbid": 211, "correlation": 230445
    }
  },
  {
    "ph": "s", "id": 230445, "pid": 76337, "tid": -914061504, "ts": 1716454224978949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225086960, "dur": 36,
    "args": {
      "External id": 230453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230453,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230453, "pid": 5, "tid": 7, "ts": 1716454225086960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224978985, "dur": 26,
    "args": {
      "External id": 230453,
      "cbid": 211, "correlation": 230453
    }
  },
  {
    "ph": "s", "id": 230453, "pid": 76337, "tid": -914061504, "ts": 1716454224978985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225086997, "dur": 53,
    "args": {
      "External id": 230473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230473,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 230473, "pid": 5, "tid": 7, "ts": 1716454225086997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979087, "dur": 12,
    "args": {
      "External id": 230473,
      "cbid": 211, "correlation": 230473
    }
  },
  {
    "ph": "s", "id": 230473, "pid": 76337, "tid": -914061504, "ts": 1716454224979087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225087051, "dur": 5,
    "args": {
      "External id": 230485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230485,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230485, "pid": 5, "tid": 7, "ts": 1716454225087051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979109, "dur": 7,
    "args": {
      "External id": 230485,
      "cbid": 211, "correlation": 230485
    }
  },
  {
    "ph": "s", "id": 230485, "pid": 76337, "tid": -914061504, "ts": 1716454224979109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225087057, "dur": 54,
    "args": {
      "External id": 230488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230488, "pid": 5, "tid": 7, "ts": 1716454225087057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979127, "dur": 6,
    "args": {
      "External id": 230488,
      "cbid": 211, "correlation": 230488
    }
  },
  {
    "ph": "s", "id": 230488, "pid": 76337, "tid": -914061504, "ts": 1716454224979127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979185, "dur": 0,
    "args": {
      "External id": 230499,
      "cbid": 317, "correlation": 230499
    }
  },
  {
    "ph": "f", "id": 230499, "pid": 76337, "tid": -914061504, "ts": 1716454224979185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979186, "dur": 0,
    "args": {
      "External id": 230500,
      "cbid": 203, "correlation": 230500
    }
  },
  {
    "ph": "f", "id": 230500, "pid": 76337, "tid": -914061504, "ts": 1716454224979186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979186, "dur": 0,
    "args": {
      "External id": 230501,
      "cbid": 205, "correlation": 230501
    }
  },
  {
    "ph": "f", "id": 230501, "pid": 76337, "tid": -914061504, "ts": 1716454224979186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979208, "dur": 1,
    "args": {
      "External id": 230505,
      "cbid": 251, "correlation": 230505
    }
  },
  {
    "ph": "f", "id": 230505, "pid": 76337, "tid": -914061504, "ts": 1716454224979208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979211, "dur": 0,
    "args": {
      "External id": 230506,
      "cbid": 251, "correlation": 230506
    }
  },
  {
    "ph": "f", "id": 230506, "pid": 76337, "tid": -914061504, "ts": 1716454224979211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979212, "dur": 0,
    "args": {
      "External id": 230507,
      "cbid": 251, "correlation": 230507
    }
  },
  {
    "ph": "f", "id": 230507, "pid": 76337, "tid": -914061504, "ts": 1716454224979212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979212, "dur": 0,
    "args": {
      "External id": 230508,
      "cbid": 251, "correlation": 230508
    }
  },
  {
    "ph": "f", "id": 230508, "pid": 76337, "tid": -914061504, "ts": 1716454224979212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979213, "dur": 0,
    "args": {
      "External id": 230509,
      "cbid": 251, "correlation": 230509
    }
  },
  {
    "ph": "f", "id": 230509, "pid": 76337, "tid": -914061504, "ts": 1716454224979213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979214, "dur": 0,
    "args": {
      "External id": 230510,
      "cbid": 251, "correlation": 230510
    }
  },
  {
    "ph": "f", "id": 230510, "pid": 76337, "tid": -914061504, "ts": 1716454224979214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979215, "dur": 0,
    "args": {
      "External id": 230511,
      "cbid": 251, "correlation": 230511
    }
  },
  {
    "ph": "f", "id": 230511, "pid": 76337, "tid": -914061504, "ts": 1716454224979215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979216, "dur": 0,
    "args": {
      "External id": 230512,
      "cbid": 251, "correlation": 230512
    }
  },
  {
    "ph": "f", "id": 230512, "pid": 76337, "tid": -914061504, "ts": 1716454224979216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979217, "dur": 0,
    "args": {
      "External id": 230513,
      "cbid": 251, "correlation": 230513
    }
  },
  {
    "ph": "f", "id": 230513, "pid": 76337, "tid": -914061504, "ts": 1716454224979217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225087112, "dur": 109,
    "args": {
      "External id": 230514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230514,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230514, "pid": 5, "tid": 7, "ts": 1716454225087112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979219, "dur": 12,
    "args": {
      "External id": 230514,
      "cbid": 211, "correlation": 230514
    }
  },
  {
    "ph": "s", "id": 230514, "pid": 76337, "tid": -914061504, "ts": 1716454224979219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225087222, "dur": 59,
    "args": {
      "External id": 230520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230520, "pid": 5, "tid": 7, "ts": 1716454225087222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979254, "dur": 9,
    "args": {
      "External id": 230520,
      "cbid": 211, "correlation": 230520
    }
  },
  {
    "ph": "s", "id": 230520, "pid": 76337, "tid": -914061504, "ts": 1716454224979254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225087283, "dur": 477,
    "args": {
      "External id": 230529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230529,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230529, "pid": 5, "tid": 7, "ts": 1716454225087283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979336, "dur": 14,
    "args": {
      "External id": 230529,
      "cbid": 211, "correlation": 230529
    }
  },
  {
    "ph": "s", "id": 230529, "pid": 76337, "tid": -914061504, "ts": 1716454224979336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225087761, "dur": 177,
    "args": {
      "External id": 230551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230551,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230551, "pid": 5, "tid": 7, "ts": 1716454225087761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979394, "dur": 10,
    "args": {
      "External id": 230551,
      "cbid": 211, "correlation": 230551
    }
  },
  {
    "ph": "s", "id": 230551, "pid": 76337, "tid": -914061504, "ts": 1716454224979394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979480, "dur": 1,
    "args": {
      "External id": 230562,
      "cbid": 251, "correlation": 230562
    }
  },
  {
    "ph": "f", "id": 230562, "pid": 76337, "tid": -914061504, "ts": 1716454224979480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225087940, "dur": 195,
    "args": {
      "External id": 230563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230563,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230563, "pid": 5, "tid": 7, "ts": 1716454225087940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979485, "dur": 13,
    "args": {
      "External id": 230563,
      "cbid": 211, "correlation": 230563
    }
  },
  {
    "ph": "s", "id": 230563, "pid": 76337, "tid": -914061504, "ts": 1716454224979485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979553, "dur": 1,
    "args": {
      "External id": 230574,
      "cbid": 251, "correlation": 230574
    }
  },
  {
    "ph": "f", "id": 230574, "pid": 76337, "tid": -914061504, "ts": 1716454224979553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225088136, "dur": 185,
    "args": {
      "External id": 230575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230575,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230575, "pid": 5, "tid": 7, "ts": 1716454225088136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979557, "dur": 12,
    "args": {
      "External id": 230575,
      "cbid": 211, "correlation": 230575
    }
  },
  {
    "ph": "s", "id": 230575, "pid": 76337, "tid": -914061504, "ts": 1716454224979557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979621, "dur": 1,
    "args": {
      "External id": 230586,
      "cbid": 251, "correlation": 230586
    }
  },
  {
    "ph": "f", "id": 230586, "pid": 76337, "tid": -914061504, "ts": 1716454224979621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225088323, "dur": 186,
    "args": {
      "External id": 230587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230587,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230587, "pid": 5, "tid": 7, "ts": 1716454225088323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979625, "dur": 11,
    "args": {
      "External id": 230587,
      "cbid": 211, "correlation": 230587
    }
  },
  {
    "ph": "s", "id": 230587, "pid": 76337, "tid": -914061504, "ts": 1716454224979625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225088510, "dur": 18190,
    "args": {
      "External id": 230608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230608,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230608, "pid": 5, "tid": 7, "ts": 1716454225088510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979704, "dur": 13,
    "args": {
      "External id": 230608,
      "cbid": 211, "correlation": 230608
    }
  },
  {
    "ph": "s", "id": 230608, "pid": 76337, "tid": -914061504, "ts": 1716454224979704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979801, "dur": 1,
    "args": {
      "External id": 230626,
      "cbid": 251, "correlation": 230626
    }
  },
  {
    "ph": "f", "id": 230626, "pid": 76337, "tid": -914061504, "ts": 1716454224979801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225106701, "dur": 199,
    "args": {
      "External id": 230628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230628,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230628, "pid": 5, "tid": 7, "ts": 1716454225106701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979807, "dur": 14,
    "args": {
      "External id": 230628,
      "cbid": 211, "correlation": 230628
    }
  },
  {
    "ph": "s", "id": 230628, "pid": 76337, "tid": -914061504, "ts": 1716454224979807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225106901, "dur": 66,
    "args": {
      "External id": 230636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230636,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230636, "pid": 5, "tid": 7, "ts": 1716454225106901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979877, "dur": 12,
    "args": {
      "External id": 230636,
      "cbid": 211, "correlation": 230636
    }
  },
  {
    "ph": "s", "id": 230636, "pid": 76337, "tid": -914061504, "ts": 1716454224979877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225106969, "dur": 97,
    "args": {
      "External id": 230644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230644,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230644, "pid": 5, "tid": 7, "ts": 1716454225106969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979917, "dur": 9,
    "args": {
      "External id": 230644,
      "cbid": 211, "correlation": 230644
    }
  },
  {
    "ph": "s", "id": 230644, "pid": 76337, "tid": -914061504, "ts": 1716454224979917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225107067, "dur": 54,
    "args": {
      "External id": 230655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230655, "pid": 5, "tid": 7, "ts": 1716454225107067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224979996, "dur": 72,
    "args": {
      "External id": 230655,
      "cbid": 211, "correlation": 230655
    }
  },
  {
    "ph": "s", "id": 230655, "pid": 76337, "tid": -914061504, "ts": 1716454224979996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225107122, "dur": 90,
    "args": {
      "External id": 230677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230677,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230677, "pid": 5, "tid": 7, "ts": 1716454225107122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224980087, "dur": 1935,
    "args": {
      "External id": 230677,
      "cbid": 211, "correlation": 230677
    }
  },
  {
    "ph": "s", "id": 230677, "pid": 76337, "tid": -914061504, "ts": 1716454224980087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982099, "dur": 1,
    "args": {
      "External id": 230688,
      "cbid": 251, "correlation": 230688
    }
  },
  {
    "ph": "f", "id": 230688, "pid": 76337, "tid": -914061504, "ts": 1716454224982099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225107213, "dur": 101,
    "args": {
      "External id": 230689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230689,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230689, "pid": 5, "tid": 7, "ts": 1716454225107213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982104, "dur": 69,
    "args": {
      "External id": 230689,
      "cbid": 211, "correlation": 230689
    }
  },
  {
    "ph": "s", "id": 230689, "pid": 76337, "tid": -914061504, "ts": 1716454224982104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982232, "dur": 1,
    "args": {
      "External id": 230700,
      "cbid": 251, "correlation": 230700
    }
  },
  {
    "ph": "f", "id": 230700, "pid": 76337, "tid": -914061504, "ts": 1716454224982232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982236, "dur": 0,
    "args": {
      "External id": 230701,
      "cbid": 251, "correlation": 230701
    }
  },
  {
    "ph": "f", "id": 230701, "pid": 76337, "tid": -914061504, "ts": 1716454224982236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225107315, "dur": 11,
    "args": {
      "External id": 230702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230702,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 230702, "pid": 5, "tid": 7, "ts": 1716454225107315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982238, "dur": 12,
    "args": {
      "External id": 230702,
      "cbid": 211, "correlation": 230702
    }
  },
  {
    "ph": "s", "id": 230702, "pid": 76337, "tid": -914061504, "ts": 1716454224982238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225107327, "dur": 5,
    "args": {
      "External id": 230704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230704,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230704, "pid": 5, "tid": 7, "ts": 1716454225107327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982252, "dur": 6,
    "args": {
      "External id": 230704,
      "cbid": 211, "correlation": 230704
    }
  },
  {
    "ph": "s", "id": 230704, "pid": 76337, "tid": -914061504, "ts": 1716454224982252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982313, "dur": 1,
    "args": {
      "External id": 230715,
      "cbid": 251, "correlation": 230715
    }
  },
  {
    "ph": "f", "id": 230715, "pid": 76337, "tid": -914061504, "ts": 1716454224982313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982317, "dur": 0,
    "args": {
      "External id": 230716,
      "cbid": 251, "correlation": 230716
    }
  },
  {
    "ph": "f", "id": 230716, "pid": 76337, "tid": -914061504, "ts": 1716454224982317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225107334, "dur": 6,
    "args": {
      "External id": 230717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230717,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 230717, "pid": 5, "tid": 7, "ts": 1716454225107334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982319, "dur": 12,
    "args": {
      "External id": 230717,
      "cbid": 211, "correlation": 230717
    }
  },
  {
    "ph": "s", "id": 230717, "pid": 76337, "tid": -914061504, "ts": 1716454224982319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225107341, "dur": 3,
    "args": {
      "External id": 230719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230719,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230719, "pid": 5, "tid": 7, "ts": 1716454225107341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982332, "dur": 5,
    "args": {
      "External id": 230719,
      "cbid": 211, "correlation": 230719
    }
  },
  {
    "ph": "s", "id": 230719, "pid": 76337, "tid": -914061504, "ts": 1716454224982332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225107346, "dur": 154,
    "args": {
      "External id": 230740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230740,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230740, "pid": 5, "tid": 7, "ts": 1716454225107346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982405, "dur": 13,
    "args": {
      "External id": 230740,
      "cbid": 211, "correlation": 230740
    }
  },
  {
    "ph": "s", "id": 230740, "pid": 76337, "tid": -914061504, "ts": 1716454224982405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982503, "dur": 1,
    "args": {
      "External id": 230758,
      "cbid": 251, "correlation": 230758
    }
  },
  {
    "ph": "f", "id": 230758, "pid": 76337, "tid": -914061504, "ts": 1716454224982503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225107502, "dur": 105,
    "args": {
      "External id": 230760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230760,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230760, "pid": 5, "tid": 7, "ts": 1716454225107502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982509, "dur": 14,
    "args": {
      "External id": 230760,
      "cbid": 211, "correlation": 230760
    }
  },
  {
    "ph": "s", "id": 230760, "pid": 76337, "tid": -914061504, "ts": 1716454224982509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225107608, "dur": 35,
    "args": {
      "External id": 230768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230768,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230768, "pid": 5, "tid": 7, "ts": 1716454225107608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982579, "dur": 12,
    "args": {
      "External id": 230768,
      "cbid": 211, "correlation": 230768
    }
  },
  {
    "ph": "s", "id": 230768, "pid": 76337, "tid": -914061504, "ts": 1716454224982579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225107644, "dur": 66,
    "args": {
      "External id": 230776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230776, "pid": 5, "tid": 7, "ts": 1716454225107644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982620, "dur": 9,
    "args": {
      "External id": 230776,
      "cbid": 211, "correlation": 230776
    }
  },
  {
    "ph": "s", "id": 230776, "pid": 76337, "tid": -914061504, "ts": 1716454224982620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225107711, "dur": 90,
    "args": {
      "External id": 230798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230798,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230798, "pid": 5, "tid": 7, "ts": 1716454225107711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982672, "dur": 10,
    "args": {
      "External id": 230798,
      "cbid": 211, "correlation": 230798
    }
  },
  {
    "ph": "s", "id": 230798, "pid": 76337, "tid": -914061504, "ts": 1716454224982672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982757, "dur": 1,
    "args": {
      "External id": 230814,
      "cbid": 251, "correlation": 230814
    }
  },
  {
    "ph": "f", "id": 230814, "pid": 76337, "tid": -914061504, "ts": 1716454224982757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225107802, "dur": 568,
    "args": {
      "External id": 230816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230816,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230816, "pid": 5, "tid": 7, "ts": 1716454225107802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982763, "dur": 13,
    "args": {
      "External id": 230816,
      "cbid": 211, "correlation": 230816
    }
  },
  {
    "ph": "s", "id": 230816, "pid": 76337, "tid": -914061504, "ts": 1716454224982763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225108371, "dur": 238,
    "args": {
      "External id": 230824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230824, "pid": 5, "tid": 7, "ts": 1716454225108371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982829, "dur": 12,
    "args": {
      "External id": 230824,
      "cbid": 211, "correlation": 230824
    }
  },
  {
    "ph": "s", "id": 230824, "pid": 76337, "tid": -914061504, "ts": 1716454224982829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225108610, "dur": 248,
    "args": {
      "External id": 230832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230832, "pid": 5, "tid": 7, "ts": 1716454225108610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982859, "dur": 9,
    "args": {
      "External id": 230832,
      "cbid": 211, "correlation": 230832
    }
  },
  {
    "ph": "s", "id": 230832, "pid": 76337, "tid": -914061504, "ts": 1716454224982859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982940, "dur": 1,
    "args": {
      "External id": 230848,
      "cbid": 251, "correlation": 230848
    }
  },
  {
    "ph": "f", "id": 230848, "pid": 76337, "tid": -914061504, "ts": 1716454224982940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982945, "dur": 0,
    "args": {
      "External id": 230850,
      "cbid": 251, "correlation": 230850
    }
  },
  {
    "ph": "f", "id": 230850, "pid": 76337, "tid": -914061504, "ts": 1716454224982945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225108860, "dur": 354,
    "args": {
      "External id": 230851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230851,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 230851, "pid": 5, "tid": 7, "ts": 1716454225108860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982948, "dur": 13,
    "args": {
      "External id": 230851,
      "cbid": 211, "correlation": 230851
    }
  },
  {
    "ph": "s", "id": 230851, "pid": 76337, "tid": -914061504, "ts": 1716454224982948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225109215, "dur": 50,
    "args": {
      "External id": 230859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230859,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230859, "pid": 5, "tid": 7, "ts": 1716454225109215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224982998, "dur": 11,
    "args": {
      "External id": 230859,
      "cbid": 211, "correlation": 230859
    }
  },
  {
    "ph": "s", "id": 230859, "pid": 76337, "tid": -914061504, "ts": 1716454224982998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225109267, "dur": 154,
    "args": {
      "External id": 230870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230870, "pid": 5, "tid": 7, "ts": 1716454225109267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983066, "dur": 207,
    "args": {
      "External id": 230870,
      "cbid": 211, "correlation": 230870
    }
  },
  {
    "ph": "s", "id": 230870, "pid": 76337, "tid": -914061504, "ts": 1716454224983066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983326, "dur": 0,
    "args": {
      "External id": 230882,
      "cbid": 317, "correlation": 230882
    }
  },
  {
    "ph": "f", "id": 230882, "pid": 76337, "tid": -914061504, "ts": 1716454224983326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983327, "dur": 0,
    "args": {
      "External id": 230883,
      "cbid": 203, "correlation": 230883
    }
  },
  {
    "ph": "f", "id": 230883, "pid": 76337, "tid": -914061504, "ts": 1716454224983327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983328, "dur": 0,
    "args": {
      "External id": 230884,
      "cbid": 205, "correlation": 230884
    }
  },
  {
    "ph": "f", "id": 230884, "pid": 76337, "tid": -914061504, "ts": 1716454224983328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983353, "dur": 1,
    "args": {
      "External id": 230888,
      "cbid": 251, "correlation": 230888
    }
  },
  {
    "ph": "f", "id": 230888, "pid": 76337, "tid": -914061504, "ts": 1716454224983353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983355, "dur": 0,
    "args": {
      "External id": 230889,
      "cbid": 251, "correlation": 230889
    }
  },
  {
    "ph": "f", "id": 230889, "pid": 76337, "tid": -914061504, "ts": 1716454224983355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983355, "dur": 0,
    "args": {
      "External id": 230890,
      "cbid": 251, "correlation": 230890
    }
  },
  {
    "ph": "f", "id": 230890, "pid": 76337, "tid": -914061504, "ts": 1716454224983355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983356, "dur": 0,
    "args": {
      "External id": 230891,
      "cbid": 251, "correlation": 230891
    }
  },
  {
    "ph": "f", "id": 230891, "pid": 76337, "tid": -914061504, "ts": 1716454224983356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983357, "dur": 0,
    "args": {
      "External id": 230892,
      "cbid": 251, "correlation": 230892
    }
  },
  {
    "ph": "f", "id": 230892, "pid": 76337, "tid": -914061504, "ts": 1716454224983357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983358, "dur": 0,
    "args": {
      "External id": 230893,
      "cbid": 251, "correlation": 230893
    }
  },
  {
    "ph": "f", "id": 230893, "pid": 76337, "tid": -914061504, "ts": 1716454224983358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983358, "dur": 0,
    "args": {
      "External id": 230894,
      "cbid": 251, "correlation": 230894
    }
  },
  {
    "ph": "f", "id": 230894, "pid": 76337, "tid": -914061504, "ts": 1716454224983358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983359, "dur": 0,
    "args": {
      "External id": 230895,
      "cbid": 251, "correlation": 230895
    }
  },
  {
    "ph": "f", "id": 230895, "pid": 76337, "tid": -914061504, "ts": 1716454224983359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983360, "dur": 0,
    "args": {
      "External id": 230896,
      "cbid": 251, "correlation": 230896
    }
  },
  {
    "ph": "f", "id": 230896, "pid": 76337, "tid": -914061504, "ts": 1716454224983360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225109422, "dur": 110,
    "args": {
      "External id": 230897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230897,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 230897, "pid": 5, "tid": 7, "ts": 1716454225109422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983362, "dur": 36,
    "args": {
      "External id": 230897,
      "cbid": 211, "correlation": 230897
    }
  },
  {
    "ph": "s", "id": 230897, "pid": 76337, "tid": -914061504, "ts": 1716454224983362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225109534, "dur": 59,
    "args": {
      "External id": 230903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230903,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230903, "pid": 5, "tid": 7, "ts": 1716454225109534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983420, "dur": 104,
    "args": {
      "External id": 230903,
      "cbid": 211, "correlation": 230903
    }
  },
  {
    "ph": "s", "id": 230903, "pid": 76337, "tid": -914061504, "ts": 1716454224983420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225109594, "dur": 50,
    "args": {
      "External id": 230911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230911,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230911, "pid": 5, "tid": 7, "ts": 1716454225109594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983548, "dur": 277,
    "args": {
      "External id": 230911,
      "cbid": 211, "correlation": 230911
    }
  },
  {
    "ph": "s", "id": 230911, "pid": 76337, "tid": -914061504, "ts": 1716454224983548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225109646, "dur": 95,
    "args": {
      "External id": 230920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230920,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230920, "pid": 5, "tid": 7, "ts": 1716454225109646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983857, "dur": 10,
    "args": {
      "External id": 230920,
      "cbid": 211, "correlation": 230920
    }
  },
  {
    "ph": "s", "id": 230920, "pid": 76337, "tid": -914061504, "ts": 1716454224983857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225109742, "dur": 91,
    "args": {
      "External id": 230940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230940,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 230940, "pid": 5, "tid": 7, "ts": 1716454225109742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983927, "dur": 12,
    "args": {
      "External id": 230940,
      "cbid": 211, "correlation": 230940
    }
  },
  {
    "ph": "s", "id": 230940, "pid": 76337, "tid": -914061504, "ts": 1716454224983927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225109834, "dur": 5,
    "args": {
      "External id": 230952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230952,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 230952, "pid": 5, "tid": 7, "ts": 1716454225109834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983949, "dur": 9,
    "args": {
      "External id": 230952,
      "cbid": 211, "correlation": 230952
    }
  },
  {
    "ph": "s", "id": 230952, "pid": 76337, "tid": -914061504, "ts": 1716454224983949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225109840, "dur": 109,
    "args": {
      "External id": 230955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230955, "pid": 5, "tid": 7, "ts": 1716454225109840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224983971, "dur": 107,
    "args": {
      "External id": 230955,
      "cbid": 211, "correlation": 230955
    }
  },
  {
    "ph": "s", "id": 230955, "pid": 76337, "tid": -914061504, "ts": 1716454224983971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225109950, "dur": 68,
    "args": {
      "External id": 230964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230964,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230964, "pid": 5, "tid": 7, "ts": 1716454225109950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984112, "dur": 10,
    "args": {
      "External id": 230964,
      "cbid": 211, "correlation": 230964
    }
  },
  {
    "ph": "s", "id": 230964, "pid": 76337, "tid": -914061504, "ts": 1716454224984112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984164, "dur": 0,
    "args": {
      "External id": 230974,
      "cbid": 317, "correlation": 230974
    }
  },
  {
    "ph": "f", "id": 230974, "pid": 76337, "tid": -914061504, "ts": 1716454224984164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984164, "dur": 0,
    "args": {
      "External id": 230975,
      "cbid": 203, "correlation": 230975
    }
  },
  {
    "ph": "f", "id": 230975, "pid": 76337, "tid": -914061504, "ts": 1716454224984164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984165, "dur": 0,
    "args": {
      "External id": 230976,
      "cbid": 205, "correlation": 230976
    }
  },
  {
    "ph": "f", "id": 230976, "pid": 76337, "tid": -914061504, "ts": 1716454224984165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225110019, "dur": 75,
    "args": {
      "External id": 230980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230980, "pid": 5, "tid": 7, "ts": 1716454225110019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984179, "dur": 12,
    "args": {
      "External id": 230980,
      "cbid": 211, "correlation": 230980
    }
  },
  {
    "ph": "s", "id": 230980, "pid": 76337, "tid": -914061504, "ts": 1716454224984179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225110096, "dur": 24,
    "args": {
      "External id": 230982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230982, "pid": 5, "tid": 7, "ts": 1716454225110096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984193, "dur": 6,
    "args": {
      "External id": 230982,
      "cbid": 211, "correlation": 230982
    }
  },
  {
    "ph": "s", "id": 230982, "pid": 76337, "tid": -914061504, "ts": 1716454224984193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225110121, "dur": 4,
    "args": {
      "External id": 230984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230984, "pid": 5, "tid": 7, "ts": 1716454225110121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984204, "dur": 6,
    "args": {
      "External id": 230984,
      "cbid": 211, "correlation": 230984
    }
  },
  {
    "ph": "s", "id": 230984, "pid": 76337, "tid": -914061504, "ts": 1716454224984204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984212, "dur": 0,
    "args": {
      "External id": 230985,
      "cbid": 51, "correlation": 230985
    }
  },
  {
    "ph": "s", "id": 230985, "pid": 76337, "tid": -914061504, "ts": 1716454224984212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225110126, "dur": 1346,
    "args": {
      "External id": 230986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230986,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 230986, "pid": 5, "tid": 7, "ts": 1716454225110126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984213, "dur": 5,
    "args": {
      "External id": 230986,
      "cbid": 211, "correlation": 230986
    }
  },
  {
    "ph": "s", "id": 230986, "pid": 76337, "tid": -914061504, "ts": 1716454224984213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225111473, "dur": 59,
    "args": {
      "External id": 230991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 230991, "pid": 5, "tid": 7, "ts": 1716454225111473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984240, "dur": 10,
    "args": {
      "External id": 230991,
      "cbid": 211, "correlation": 230991
    }
  },
  {
    "ph": "s", "id": 230991, "pid": 76337, "tid": -914061504, "ts": 1716454224984240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225111533, "dur": 3,
    "args": {
      "External id": 230999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 230999,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 230999, "pid": 5, "tid": 7, "ts": 1716454225111533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984284, "dur": 10,
    "args": {
      "External id": 230999,
      "cbid": 211, "correlation": 230999
    }
  },
  {
    "ph": "s", "id": 230999, "pid": 76337, "tid": -914061504, "ts": 1716454224984284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984350, "dur": 1,
    "args": {
      "External id": 231015,
      "cbid": 251, "correlation": 231015
    }
  },
  {
    "ph": "f", "id": 231015, "pid": 76337, "tid": -914061504, "ts": 1716454224984350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984355, "dur": 0,
    "args": {
      "External id": 231017,
      "cbid": 251, "correlation": 231017
    }
  },
  {
    "ph": "f", "id": 231017, "pid": 76337, "tid": -914061504, "ts": 1716454224984355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225111538, "dur": 11,
    "args": {
      "External id": 231018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231018,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 231018, "pid": 5, "tid": 7, "ts": 1716454225111538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984357, "dur": 11,
    "args": {
      "External id": 231018,
      "cbid": 211, "correlation": 231018
    }
  },
  {
    "ph": "s", "id": 231018, "pid": 76337, "tid": -914061504, "ts": 1716454224984357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225111550, "dur": 5,
    "args": {
      "External id": 231020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231020,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 231020, "pid": 5, "tid": 7, "ts": 1716454225111550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984370, "dur": 5,
    "args": {
      "External id": 231020,
      "cbid": 211, "correlation": 231020
    }
  },
  {
    "ph": "s", "id": 231020, "pid": 76337, "tid": -914061504, "ts": 1716454224984370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225111557, "dur": 53,
    "args": {
      "External id": 231030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231030, "pid": 5, "tid": 7, "ts": 1716454225111557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224984427, "dur": 556,
    "args": {
      "External id": 231030,
      "cbid": 211, "correlation": 231030
    }
  },
  {
    "ph": "s", "id": 231030, "pid": 76337, "tid": -914061504, "ts": 1716454224984427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225111611, "dur": 50,
    "args": {
      "External id": 231050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231050,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 231050, "pid": 5, "tid": 7, "ts": 1716454225111611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985038, "dur": 11,
    "args": {
      "External id": 231050,
      "cbid": 211, "correlation": 231050
    }
  },
  {
    "ph": "s", "id": 231050, "pid": 76337, "tid": -914061504, "ts": 1716454224985038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225111662, "dur": 4,
    "args": {
      "External id": 231062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231062,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 231062, "pid": 5, "tid": 7, "ts": 1716454225111662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985059, "dur": 7,
    "args": {
      "External id": 231062,
      "cbid": 211, "correlation": 231062
    }
  },
  {
    "ph": "s", "id": 231062, "pid": 76337, "tid": -914061504, "ts": 1716454224985059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225111667, "dur": 54,
    "args": {
      "External id": 231065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231065, "pid": 5, "tid": 7, "ts": 1716454225111667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985079, "dur": 7,
    "args": {
      "External id": 231065,
      "cbid": 211, "correlation": 231065
    }
  },
  {
    "ph": "s", "id": 231065, "pid": 76337, "tid": -914061504, "ts": 1716454224985079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225111723, "dur": 36,
    "args": {
      "External id": 231074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231074,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231074, "pid": 5, "tid": 7, "ts": 1716454225111723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985120, "dur": 10,
    "args": {
      "External id": 231074,
      "cbid": 211, "correlation": 231074
    }
  },
  {
    "ph": "s", "id": 231074, "pid": 76337, "tid": -914061504, "ts": 1716454224985120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985182, "dur": 0,
    "args": {
      "External id": 231084,
      "cbid": 317, "correlation": 231084
    }
  },
  {
    "ph": "f", "id": 231084, "pid": 76337, "tid": -914061504, "ts": 1716454224985182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985183, "dur": 0,
    "args": {
      "External id": 231085,
      "cbid": 203, "correlation": 231085
    }
  },
  {
    "ph": "f", "id": 231085, "pid": 76337, "tid": -914061504, "ts": 1716454224985183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985184, "dur": 0,
    "args": {
      "External id": 231086,
      "cbid": 205, "correlation": 231086
    }
  },
  {
    "ph": "f", "id": 231086, "pid": 76337, "tid": -914061504, "ts": 1716454224985184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225111760, "dur": 41,
    "args": {
      "External id": 231090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231090, "pid": 5, "tid": 7, "ts": 1716454225111760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985198, "dur": 12,
    "args": {
      "External id": 231090,
      "cbid": 211, "correlation": 231090
    }
  },
  {
    "ph": "s", "id": 231090, "pid": 76337, "tid": -914061504, "ts": 1716454224985198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225111802, "dur": 14,
    "args": {
      "External id": 231092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231092,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231092, "pid": 5, "tid": 7, "ts": 1716454225111802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985212, "dur": 6,
    "args": {
      "External id": 231092,
      "cbid": 211, "correlation": 231092
    }
  },
  {
    "ph": "s", "id": 231092, "pid": 76337, "tid": -914061504, "ts": 1716454224985212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225111817, "dur": 3,
    "args": {
      "External id": 231094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 231094, "pid": 5, "tid": 7, "ts": 1716454225111817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985222, "dur": 5,
    "args": {
      "External id": 231094,
      "cbid": 211, "correlation": 231094
    }
  },
  {
    "ph": "s", "id": 231094, "pid": 76337, "tid": -914061504, "ts": 1716454224985222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985230, "dur": 0,
    "args": {
      "External id": 231095,
      "cbid": 51, "correlation": 231095
    }
  },
  {
    "ph": "s", "id": 231095, "pid": 76337, "tid": -914061504, "ts": 1716454224985230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225111822, "dur": 686,
    "args": {
      "External id": 231096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231096,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231096, "pid": 5, "tid": 7, "ts": 1716454225111822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985231, "dur": 5,
    "args": {
      "External id": 231096,
      "cbid": 211, "correlation": 231096
    }
  },
  {
    "ph": "s", "id": 231096, "pid": 76337, "tid": -914061504, "ts": 1716454224985231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225112509, "dur": 58,
    "args": {
      "External id": 231101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231101, "pid": 5, "tid": 7, "ts": 1716454225112509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985258, "dur": 9,
    "args": {
      "External id": 231101,
      "cbid": 211, "correlation": 231101
    }
  },
  {
    "ph": "s", "id": 231101, "pid": 76337, "tid": -914061504, "ts": 1716454224985258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985315, "dur": 0,
    "args": {
      "External id": 231111,
      "cbid": 317, "correlation": 231111
    }
  },
  {
    "ph": "f", "id": 231111, "pid": 76337, "tid": -914061504, "ts": 1716454224985315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985316, "dur": 0,
    "args": {
      "External id": 231112,
      "cbid": 203, "correlation": 231112
    }
  },
  {
    "ph": "f", "id": 231112, "pid": 76337, "tid": -914061504, "ts": 1716454224985316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985317, "dur": 0,
    "args": {
      "External id": 231113,
      "cbid": 205, "correlation": 231113
    }
  },
  {
    "ph": "f", "id": 231113, "pid": 76337, "tid": -914061504, "ts": 1716454224985317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225112569, "dur": 74,
    "args": {
      "External id": 231117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231117, "pid": 5, "tid": 7, "ts": 1716454225112569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985328, "dur": 11,
    "args": {
      "External id": 231117,
      "cbid": 211, "correlation": 231117
    }
  },
  {
    "ph": "s", "id": 231117, "pid": 76337, "tid": -914061504, "ts": 1716454224985328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225112644, "dur": 203,
    "args": {
      "External id": 231119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231119,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231119, "pid": 5, "tid": 7, "ts": 1716454225112644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985346, "dur": 7,
    "args": {
      "External id": 231119,
      "cbid": 211, "correlation": 231119
    }
  },
  {
    "ph": "s", "id": 231119, "pid": 76337, "tid": -914061504, "ts": 1716454224985346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225112849, "dur": 39,
    "args": {
      "External id": 231121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231121,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231121, "pid": 5, "tid": 7, "ts": 1716454225112849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985356, "dur": 5,
    "args": {
      "External id": 231121,
      "cbid": 211, "correlation": 231121
    }
  },
  {
    "ph": "s", "id": 231121, "pid": 76337, "tid": -914061504, "ts": 1716454224985356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225112889, "dur": 59,
    "args": {
      "External id": 231127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231127, "pid": 5, "tid": 7, "ts": 1716454225112889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985383, "dur": 500,
    "args": {
      "External id": 231127,
      "cbid": 211, "correlation": 231127
    }
  },
  {
    "ph": "s", "id": 231127, "pid": 76337, "tid": -914061504, "ts": 1716454224985383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225112949, "dur": 50,
    "args": {
      "External id": 231135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231135,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231135, "pid": 5, "tid": 7, "ts": 1716454225112949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985904, "dur": 8,
    "args": {
      "External id": 231135,
      "cbid": 211, "correlation": 231135
    }
  },
  {
    "ph": "s", "id": 231135, "pid": 76337, "tid": -914061504, "ts": 1716454224985904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225113000, "dur": 36,
    "args": {
      "External id": 231143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231143,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231143, "pid": 5, "tid": 7, "ts": 1716454225113000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224985934, "dur": 8,
    "args": {
      "External id": 231143,
      "cbid": 211, "correlation": 231143
    }
  },
  {
    "ph": "s", "id": 231143, "pid": 76337, "tid": -914061504, "ts": 1716454224985934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225113036, "dur": 52,
    "args": {
      "External id": 231163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231163,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 231163, "pid": 5, "tid": 7, "ts": 1716454225113036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986024, "dur": 14,
    "args": {
      "External id": 231163,
      "cbid": 211, "correlation": 231163
    }
  },
  {
    "ph": "s", "id": 231163, "pid": 76337, "tid": -914061504, "ts": 1716454224986024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225113090, "dur": 4,
    "args": {
      "External id": 231175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231175,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 231175, "pid": 5, "tid": 7, "ts": 1716454225113090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986047, "dur": 6,
    "args": {
      "External id": 231175,
      "cbid": 211, "correlation": 231175
    }
  },
  {
    "ph": "s", "id": 231175, "pid": 76337, "tid": -914061504, "ts": 1716454224986047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225113095, "dur": 54,
    "args": {
      "External id": 231178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231178, "pid": 5, "tid": 7, "ts": 1716454225113095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986065, "dur": 6,
    "args": {
      "External id": 231178,
      "cbid": 211, "correlation": 231178
    }
  },
  {
    "ph": "s", "id": 231178, "pid": 76337, "tid": -914061504, "ts": 1716454224986065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986122, "dur": 0,
    "args": {
      "External id": 231189,
      "cbid": 317, "correlation": 231189
    }
  },
  {
    "ph": "f", "id": 231189, "pid": 76337, "tid": -914061504, "ts": 1716454224986122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986122, "dur": 0,
    "args": {
      "External id": 231190,
      "cbid": 203, "correlation": 231190
    }
  },
  {
    "ph": "f", "id": 231190, "pid": 76337, "tid": -914061504, "ts": 1716454224986122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986123, "dur": 0,
    "args": {
      "External id": 231191,
      "cbid": 205, "correlation": 231191
    }
  },
  {
    "ph": "f", "id": 231191, "pid": 76337, "tid": -914061504, "ts": 1716454224986123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986146, "dur": 1,
    "args": {
      "External id": 231195,
      "cbid": 251, "correlation": 231195
    }
  },
  {
    "ph": "f", "id": 231195, "pid": 76337, "tid": -914061504, "ts": 1716454224986146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986148, "dur": 0,
    "args": {
      "External id": 231196,
      "cbid": 251, "correlation": 231196
    }
  },
  {
    "ph": "f", "id": 231196, "pid": 76337, "tid": -914061504, "ts": 1716454224986148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986149, "dur": 1,
    "args": {
      "External id": 231197,
      "cbid": 251, "correlation": 231197
    }
  },
  {
    "ph": "f", "id": 231197, "pid": 76337, "tid": -914061504, "ts": 1716454224986149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986150, "dur": 0,
    "args": {
      "External id": 231198,
      "cbid": 251, "correlation": 231198
    }
  },
  {
    "ph": "f", "id": 231198, "pid": 76337, "tid": -914061504, "ts": 1716454224986150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986151, "dur": 0,
    "args": {
      "External id": 231199,
      "cbid": 251, "correlation": 231199
    }
  },
  {
    "ph": "f", "id": 231199, "pid": 76337, "tid": -914061504, "ts": 1716454224986151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986151, "dur": 0,
    "args": {
      "External id": 231200,
      "cbid": 251, "correlation": 231200
    }
  },
  {
    "ph": "f", "id": 231200, "pid": 76337, "tid": -914061504, "ts": 1716454224986151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986152, "dur": 0,
    "args": {
      "External id": 231201,
      "cbid": 251, "correlation": 231201
    }
  },
  {
    "ph": "f", "id": 231201, "pid": 76337, "tid": -914061504, "ts": 1716454224986152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986153, "dur": 0,
    "args": {
      "External id": 231202,
      "cbid": 251, "correlation": 231202
    }
  },
  {
    "ph": "f", "id": 231202, "pid": 76337, "tid": -914061504, "ts": 1716454224986153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986154, "dur": 0,
    "args": {
      "External id": 231203,
      "cbid": 251, "correlation": 231203
    }
  },
  {
    "ph": "f", "id": 231203, "pid": 76337, "tid": -914061504, "ts": 1716454224986154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225113151, "dur": 111,
    "args": {
      "External id": 231204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231204,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 231204, "pid": 5, "tid": 7, "ts": 1716454225113151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986156, "dur": 12,
    "args": {
      "External id": 231204,
      "cbid": 211, "correlation": 231204
    }
  },
  {
    "ph": "s", "id": 231204, "pid": 76337, "tid": -914061504, "ts": 1716454224986156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225113263, "dur": 58,
    "args": {
      "External id": 231210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231210, "pid": 5, "tid": 7, "ts": 1716454225113263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986192, "dur": 9,
    "args": {
      "External id": 231210,
      "cbid": 211, "correlation": 231210
    }
  },
  {
    "ph": "s", "id": 231210, "pid": 76337, "tid": -914061504, "ts": 1716454224986192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225113322, "dur": 459,
    "args": {
      "External id": 231219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231219,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231219, "pid": 5, "tid": 7, "ts": 1716454225113322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986274, "dur": 14,
    "args": {
      "External id": 231219,
      "cbid": 211, "correlation": 231219
    }
  },
  {
    "ph": "s", "id": 231219, "pid": 76337, "tid": -914061504, "ts": 1716454224986274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225113783, "dur": 176,
    "args": {
      "External id": 231241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231241,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231241, "pid": 5, "tid": 7, "ts": 1716454225113783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986330, "dur": 10,
    "args": {
      "External id": 231241,
      "cbid": 211, "correlation": 231241
    }
  },
  {
    "ph": "s", "id": 231241, "pid": 76337, "tid": -914061504, "ts": 1716454224986330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986415, "dur": 1,
    "args": {
      "External id": 231252,
      "cbid": 251, "correlation": 231252
    }
  },
  {
    "ph": "f", "id": 231252, "pid": 76337, "tid": -914061504, "ts": 1716454224986415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225113961, "dur": 191,
    "args": {
      "External id": 231253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231253,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231253, "pid": 5, "tid": 7, "ts": 1716454225113961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986420, "dur": 14,
    "args": {
      "External id": 231253,
      "cbid": 211, "correlation": 231253
    }
  },
  {
    "ph": "s", "id": 231253, "pid": 76337, "tid": -914061504, "ts": 1716454224986420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986490, "dur": 1,
    "args": {
      "External id": 231264,
      "cbid": 251, "correlation": 231264
    }
  },
  {
    "ph": "f", "id": 231264, "pid": 76337, "tid": -914061504, "ts": 1716454224986490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225114153, "dur": 185,
    "args": {
      "External id": 231265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231265,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231265, "pid": 5, "tid": 7, "ts": 1716454225114153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986494, "dur": 11,
    "args": {
      "External id": 231265,
      "cbid": 211, "correlation": 231265
    }
  },
  {
    "ph": "s", "id": 231265, "pid": 76337, "tid": -914061504, "ts": 1716454224986494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986556, "dur": 1,
    "args": {
      "External id": 231276,
      "cbid": 251, "correlation": 231276
    }
  },
  {
    "ph": "f", "id": 231276, "pid": 76337, "tid": -914061504, "ts": 1716454224986556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225114339, "dur": 183,
    "args": {
      "External id": 231277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231277,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231277, "pid": 5, "tid": 7, "ts": 1716454225114339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986560, "dur": 11,
    "args": {
      "External id": 231277,
      "cbid": 211, "correlation": 231277
    }
  },
  {
    "ph": "s", "id": 231277, "pid": 76337, "tid": -914061504, "ts": 1716454224986560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225114524, "dur": 18148,
    "args": {
      "External id": 231298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231298,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 231298, "pid": 5, "tid": 7, "ts": 1716454225114524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986640, "dur": 12,
    "args": {
      "External id": 231298,
      "cbid": 211, "correlation": 231298
    }
  },
  {
    "ph": "s", "id": 231298, "pid": 76337, "tid": -914061504, "ts": 1716454224986640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986735, "dur": 1,
    "args": {
      "External id": 231316,
      "cbid": 251, "correlation": 231316
    }
  },
  {
    "ph": "f", "id": 231316, "pid": 76337, "tid": -914061504, "ts": 1716454224986735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225132673, "dur": 199,
    "args": {
      "External id": 231318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231318,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231318, "pid": 5, "tid": 7, "ts": 1716454225132673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986741, "dur": 13,
    "args": {
      "External id": 231318,
      "cbid": 211, "correlation": 231318
    }
  },
  {
    "ph": "s", "id": 231318, "pid": 76337, "tid": -914061504, "ts": 1716454224986741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225132874, "dur": 67,
    "args": {
      "External id": 231326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231326,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231326, "pid": 5, "tid": 7, "ts": 1716454225132874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986810, "dur": 12,
    "args": {
      "External id": 231326,
      "cbid": 211, "correlation": 231326
    }
  },
  {
    "ph": "s", "id": 231326, "pid": 76337, "tid": -914061504, "ts": 1716454224986810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225132942, "dur": 97,
    "args": {
      "External id": 231334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231334,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231334, "pid": 5, "tid": 7, "ts": 1716454225132942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224986849, "dur": 99,
    "args": {
      "External id": 231334,
      "cbid": 211, "correlation": 231334
    }
  },
  {
    "ph": "s", "id": 231334, "pid": 76337, "tid": -914061504, "ts": 1716454224986849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225133040, "dur": 54,
    "args": {
      "External id": 231345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231345,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231345, "pid": 5, "tid": 7, "ts": 1716454225133040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224987023, "dur": 1838,
    "args": {
      "External id": 231345,
      "cbid": 211, "correlation": 231345
    }
  },
  {
    "ph": "s", "id": 231345, "pid": 76337, "tid": -914061504, "ts": 1716454224987023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225133094, "dur": 90,
    "args": {
      "External id": 231367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231367,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231367, "pid": 5, "tid": 7, "ts": 1716454225133094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224988882, "dur": 123,
    "args": {
      "External id": 231367,
      "cbid": 211, "correlation": 231367
    }
  },
  {
    "ph": "s", "id": 231367, "pid": 76337, "tid": -914061504, "ts": 1716454224988882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989084, "dur": 1,
    "args": {
      "External id": 231378,
      "cbid": 251, "correlation": 231378
    }
  },
  {
    "ph": "f", "id": 231378, "pid": 76337, "tid": -914061504, "ts": 1716454224989084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225133185, "dur": 103,
    "args": {
      "External id": 231379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231379,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231379, "pid": 5, "tid": 7, "ts": 1716454225133185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989089, "dur": 13,
    "args": {
      "External id": 231379,
      "cbid": 211, "correlation": 231379
    }
  },
  {
    "ph": "s", "id": 231379, "pid": 76337, "tid": -914061504, "ts": 1716454224989089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989161, "dur": 1,
    "args": {
      "External id": 231390,
      "cbid": 251, "correlation": 231390
    }
  },
  {
    "ph": "f", "id": 231390, "pid": 76337, "tid": -914061504, "ts": 1716454224989161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989165, "dur": 0,
    "args": {
      "External id": 231391,
      "cbid": 251, "correlation": 231391
    }
  },
  {
    "ph": "f", "id": 231391, "pid": 76337, "tid": -914061504, "ts": 1716454224989165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225133290, "dur": 10,
    "args": {
      "External id": 231392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231392,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231392, "pid": 5, "tid": 7, "ts": 1716454225133290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989166, "dur": 12,
    "args": {
      "External id": 231392,
      "cbid": 211, "correlation": 231392
    }
  },
  {
    "ph": "s", "id": 231392, "pid": 76337, "tid": -914061504, "ts": 1716454224989166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225133301, "dur": 5,
    "args": {
      "External id": 231394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231394,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 231394, "pid": 5, "tid": 7, "ts": 1716454225133301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989181, "dur": 7,
    "args": {
      "External id": 231394,
      "cbid": 211, "correlation": 231394
    }
  },
  {
    "ph": "s", "id": 231394, "pid": 76337, "tid": -914061504, "ts": 1716454224989181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989242, "dur": 1,
    "args": {
      "External id": 231405,
      "cbid": 251, "correlation": 231405
    }
  },
  {
    "ph": "f", "id": 231405, "pid": 76337, "tid": -914061504, "ts": 1716454224989242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989245, "dur": 0,
    "args": {
      "External id": 231406,
      "cbid": 251, "correlation": 231406
    }
  },
  {
    "ph": "f", "id": 231406, "pid": 76337, "tid": -914061504, "ts": 1716454224989245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225133307, "dur": 6,
    "args": {
      "External id": 231407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231407,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231407, "pid": 5, "tid": 7, "ts": 1716454225133307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989247, "dur": 12,
    "args": {
      "External id": 231407,
      "cbid": 211, "correlation": 231407
    }
  },
  {
    "ph": "s", "id": 231407, "pid": 76337, "tid": -914061504, "ts": 1716454224989247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225133314, "dur": 3,
    "args": {
      "External id": 231409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231409,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 231409, "pid": 5, "tid": 7, "ts": 1716454225133314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989260, "dur": 6,
    "args": {
      "External id": 231409,
      "cbid": 211, "correlation": 231409
    }
  },
  {
    "ph": "s", "id": 231409, "pid": 76337, "tid": -914061504, "ts": 1716454224989260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225133319, "dur": 152,
    "args": {
      "External id": 231430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231430,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 231430, "pid": 5, "tid": 7, "ts": 1716454225133319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989334, "dur": 12,
    "args": {
      "External id": 231430,
      "cbid": 211, "correlation": 231430
    }
  },
  {
    "ph": "s", "id": 231430, "pid": 76337, "tid": -914061504, "ts": 1716454224989334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989430, "dur": 1,
    "args": {
      "External id": 231448,
      "cbid": 251, "correlation": 231448
    }
  },
  {
    "ph": "f", "id": 231448, "pid": 76337, "tid": -914061504, "ts": 1716454224989430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225133473, "dur": 106,
    "args": {
      "External id": 231450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231450,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 231450, "pid": 5, "tid": 7, "ts": 1716454225133473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989437, "dur": 13,
    "args": {
      "External id": 231450,
      "cbid": 211, "correlation": 231450
    }
  },
  {
    "ph": "s", "id": 231450, "pid": 76337, "tid": -914061504, "ts": 1716454224989437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225133580, "dur": 34,
    "args": {
      "External id": 231458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231458,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231458, "pid": 5, "tid": 7, "ts": 1716454225133580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989505, "dur": 12,
    "args": {
      "External id": 231458,
      "cbid": 211, "correlation": 231458
    }
  },
  {
    "ph": "s", "id": 231458, "pid": 76337, "tid": -914061504, "ts": 1716454224989505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225133615, "dur": 65,
    "args": {
      "External id": 231466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231466, "pid": 5, "tid": 7, "ts": 1716454225133615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989546, "dur": 9,
    "args": {
      "External id": 231466,
      "cbid": 211, "correlation": 231466
    }
  },
  {
    "ph": "s", "id": 231466, "pid": 76337, "tid": -914061504, "ts": 1716454224989546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225133682, "dur": 90,
    "args": {
      "External id": 231488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231488,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231488, "pid": 5, "tid": 7, "ts": 1716454225133682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989598, "dur": 10,
    "args": {
      "External id": 231488,
      "cbid": 211, "correlation": 231488
    }
  },
  {
    "ph": "s", "id": 231488, "pid": 76337, "tid": -914061504, "ts": 1716454224989598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989682, "dur": 1,
    "args": {
      "External id": 231504,
      "cbid": 251, "correlation": 231504
    }
  },
  {
    "ph": "f", "id": 231504, "pid": 76337, "tid": -914061504, "ts": 1716454224989682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225133773, "dur": 561,
    "args": {
      "External id": 231506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231506,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 231506, "pid": 5, "tid": 7, "ts": 1716454225133773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989689, "dur": 13,
    "args": {
      "External id": 231506,
      "cbid": 211, "correlation": 231506
    }
  },
  {
    "ph": "s", "id": 231506, "pid": 76337, "tid": -914061504, "ts": 1716454224989689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225134336, "dur": 240,
    "args": {
      "External id": 231514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231514, "pid": 5, "tid": 7, "ts": 1716454225134336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989754, "dur": 12,
    "args": {
      "External id": 231514,
      "cbid": 211, "correlation": 231514
    }
  },
  {
    "ph": "s", "id": 231514, "pid": 76337, "tid": -914061504, "ts": 1716454224989754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225134577, "dur": 252,
    "args": {
      "External id": 231522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231522, "pid": 5, "tid": 7, "ts": 1716454225134577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989785, "dur": 9,
    "args": {
      "External id": 231522,
      "cbid": 211, "correlation": 231522
    }
  },
  {
    "ph": "s", "id": 231522, "pid": 76337, "tid": -914061504, "ts": 1716454224989785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989866, "dur": 1,
    "args": {
      "External id": 231538,
      "cbid": 251, "correlation": 231538
    }
  },
  {
    "ph": "f", "id": 231538, "pid": 76337, "tid": -914061504, "ts": 1716454224989866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989871, "dur": 0,
    "args": {
      "External id": 231540,
      "cbid": 251, "correlation": 231540
    }
  },
  {
    "ph": "f", "id": 231540, "pid": 76337, "tid": -914061504, "ts": 1716454224989871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225134830, "dur": 351,
    "args": {
      "External id": 231541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231541,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231541, "pid": 5, "tid": 7, "ts": 1716454225134830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989873, "dur": 12,
    "args": {
      "External id": 231541,
      "cbid": 211, "correlation": 231541
    }
  },
  {
    "ph": "s", "id": 231541, "pid": 76337, "tid": -914061504, "ts": 1716454224989873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225135182, "dur": 50,
    "args": {
      "External id": 231549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231549,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231549, "pid": 5, "tid": 7, "ts": 1716454225135182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224989915, "dur": 180,
    "args": {
      "External id": 231549,
      "cbid": 211, "correlation": 231549
    }
  },
  {
    "ph": "s", "id": 231549, "pid": 76337, "tid": -914061504, "ts": 1716454224989915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225135234, "dur": 154,
    "args": {
      "External id": 231560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231560, "pid": 5, "tid": 7, "ts": 1716454225135234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990153, "dur": 67,
    "args": {
      "External id": 231560,
      "cbid": 211, "correlation": 231560
    }
  },
  {
    "ph": "s", "id": 231560, "pid": 76337, "tid": -914061504, "ts": 1716454224990153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990274, "dur": 0,
    "args": {
      "External id": 231572,
      "cbid": 317, "correlation": 231572
    }
  },
  {
    "ph": "f", "id": 231572, "pid": 76337, "tid": -914061504, "ts": 1716454224990274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990275, "dur": 0,
    "args": {
      "External id": 231573,
      "cbid": 203, "correlation": 231573
    }
  },
  {
    "ph": "f", "id": 231573, "pid": 76337, "tid": -914061504, "ts": 1716454224990275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990277, "dur": 0,
    "args": {
      "External id": 231574,
      "cbid": 205, "correlation": 231574
    }
  },
  {
    "ph": "f", "id": 231574, "pid": 76337, "tid": -914061504, "ts": 1716454224990277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990300, "dur": 1,
    "args": {
      "External id": 231578,
      "cbid": 251, "correlation": 231578
    }
  },
  {
    "ph": "f", "id": 231578, "pid": 76337, "tid": -914061504, "ts": 1716454224990300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990302, "dur": 0,
    "args": {
      "External id": 231579,
      "cbid": 251, "correlation": 231579
    }
  },
  {
    "ph": "f", "id": 231579, "pid": 76337, "tid": -914061504, "ts": 1716454224990302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990302, "dur": 0,
    "args": {
      "External id": 231580,
      "cbid": 251, "correlation": 231580
    }
  },
  {
    "ph": "f", "id": 231580, "pid": 76337, "tid": -914061504, "ts": 1716454224990302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990303, "dur": 0,
    "args": {
      "External id": 231581,
      "cbid": 251, "correlation": 231581
    }
  },
  {
    "ph": "f", "id": 231581, "pid": 76337, "tid": -914061504, "ts": 1716454224990303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990304, "dur": 0,
    "args": {
      "External id": 231582,
      "cbid": 251, "correlation": 231582
    }
  },
  {
    "ph": "f", "id": 231582, "pid": 76337, "tid": -914061504, "ts": 1716454224990304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990304, "dur": 0,
    "args": {
      "External id": 231583,
      "cbid": 251, "correlation": 231583
    }
  },
  {
    "ph": "f", "id": 231583, "pid": 76337, "tid": -914061504, "ts": 1716454224990304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990305, "dur": 0,
    "args": {
      "External id": 231584,
      "cbid": 251, "correlation": 231584
    }
  },
  {
    "ph": "f", "id": 231584, "pid": 76337, "tid": -914061504, "ts": 1716454224990305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990306, "dur": 0,
    "args": {
      "External id": 231585,
      "cbid": 251, "correlation": 231585
    }
  },
  {
    "ph": "f", "id": 231585, "pid": 76337, "tid": -914061504, "ts": 1716454224990306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990307, "dur": 0,
    "args": {
      "External id": 231586,
      "cbid": 251, "correlation": 231586
    }
  },
  {
    "ph": "f", "id": 231586, "pid": 76337, "tid": -914061504, "ts": 1716454224990307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225135389, "dur": 115,
    "args": {
      "External id": 231587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231587,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 231587, "pid": 5, "tid": 7, "ts": 1716454225135389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990309, "dur": 40,
    "args": {
      "External id": 231587,
      "cbid": 211, "correlation": 231587
    }
  },
  {
    "ph": "s", "id": 231587, "pid": 76337, "tid": -914061504, "ts": 1716454224990309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225135506, "dur": 59,
    "args": {
      "External id": 231593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231593,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231593, "pid": 5, "tid": 7, "ts": 1716454225135506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990372, "dur": 281,
    "args": {
      "External id": 231593,
      "cbid": 211, "correlation": 231593
    }
  },
  {
    "ph": "s", "id": 231593, "pid": 76337, "tid": -914061504, "ts": 1716454224990372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225135566, "dur": 50,
    "args": {
      "External id": 231601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231601,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231601, "pid": 5, "tid": 7, "ts": 1716454225135566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990677, "dur": 9,
    "args": {
      "External id": 231601,
      "cbid": 211, "correlation": 231601
    }
  },
  {
    "ph": "s", "id": 231601, "pid": 76337, "tid": -914061504, "ts": 1716454224990677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225135617, "dur": 51,
    "args": {
      "External id": 231621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231621,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 231621, "pid": 5, "tid": 7, "ts": 1716454225135617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990757, "dur": 13,
    "args": {
      "External id": 231621,
      "cbid": 211, "correlation": 231621
    }
  },
  {
    "ph": "s", "id": 231621, "pid": 76337, "tid": -914061504, "ts": 1716454224990757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225135669, "dur": 5,
    "args": {
      "External id": 231633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231633,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 231633, "pid": 5, "tid": 7, "ts": 1716454225135669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990780, "dur": 6,
    "args": {
      "External id": 231633,
      "cbid": 211, "correlation": 231633
    }
  },
  {
    "ph": "s", "id": 231633, "pid": 76337, "tid": -914061504, "ts": 1716454224990780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225135675, "dur": 55,
    "args": {
      "External id": 231636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231636, "pid": 5, "tid": 7, "ts": 1716454225135675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990798, "dur": 108,
    "args": {
      "External id": 231636,
      "cbid": 211, "correlation": 231636
    }
  },
  {
    "ph": "s", "id": 231636, "pid": 76337, "tid": -914061504, "ts": 1716454224990798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225135732, "dur": 37,
    "args": {
      "External id": 231645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231645,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231645, "pid": 5, "tid": 7, "ts": 1716454225135732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224990945, "dur": 10,
    "args": {
      "External id": 231645,
      "cbid": 211, "correlation": 231645
    }
  },
  {
    "ph": "s", "id": 231645, "pid": 76337, "tid": -914061504, "ts": 1716454224990945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991010, "dur": 0,
    "args": {
      "External id": 231655,
      "cbid": 317, "correlation": 231655
    }
  },
  {
    "ph": "f", "id": 231655, "pid": 76337, "tid": -914061504, "ts": 1716454224991010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991011, "dur": 0,
    "args": {
      "External id": 231656,
      "cbid": 203, "correlation": 231656
    }
  },
  {
    "ph": "f", "id": 231656, "pid": 76337, "tid": -914061504, "ts": 1716454224991011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991011, "dur": 0,
    "args": {
      "External id": 231657,
      "cbid": 205, "correlation": 231657
    }
  },
  {
    "ph": "f", "id": 231657, "pid": 76337, "tid": -914061504, "ts": 1716454224991011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225135770, "dur": 41,
    "args": {
      "External id": 231661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231661, "pid": 5, "tid": 7, "ts": 1716454225135770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991027, "dur": 13,
    "args": {
      "External id": 231661,
      "cbid": 211, "correlation": 231661
    }
  },
  {
    "ph": "s", "id": 231661, "pid": 76337, "tid": -914061504, "ts": 1716454224991027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225135813, "dur": 3,
    "args": {
      "External id": 231663,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231663,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.80327868852459
    }
  },
  {
    "ph": "f", "id": 231663, "pid": 5, "tid": 7, "ts": 1716454225135813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991044, "dur": 16,
    "args": {
      "External id": 231663,
      "cbid": 51, "correlation": 231663
    }
  },
  {
    "ph": "s", "id": 231663, "pid": 76337, "tid": -914061504, "ts": 1716454224991044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991065, "dur": 2,
    "args": {
      "External id": 231665,
      "cbid": 200, "correlation": 231665
    }
  },
  {
    "ph": "f", "id": 231665, "pid": 76337, "tid": -914061504, "ts": 1716454224991065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991068, "dur": 0,
    "args": {
      "External id": 231666,
      "cbid": 200, "correlation": 231666
    }
  },
  {
    "ph": "f", "id": 231666, "pid": 76337, "tid": -914061504, "ts": 1716454224991068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991068, "dur": 0,
    "args": {
      "External id": 231667,
      "cbid": 200, "correlation": 231667
    }
  },
  {
    "ph": "f", "id": 231667, "pid": 76337, "tid": -914061504, "ts": 1716454224991068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991069, "dur": 0,
    "args": {
      "External id": 231668,
      "cbid": 200, "correlation": 231668
    }
  },
  {
    "ph": "f", "id": 231668, "pid": 76337, "tid": -914061504, "ts": 1716454224991069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991070, "dur": 4,
    "args": {
      "External id": 231669,
      "cbid": 15, "correlation": 231669
    }
  },
  {
    "ph": "f", "id": 231669, "pid": 76337, "tid": -914061504, "ts": 1716454224991070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991075, "dur": 1,
    "args": {
      "External id": 231670,
      "cbid": 251, "correlation": 231670
    }
  },
  {
    "ph": "f", "id": 231670, "pid": 76337, "tid": -914061504, "ts": 1716454224991075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225135818, "dur": 23,
    "args": {
      "External id": 231671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231671,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231671, "pid": 5, "tid": 7, "ts": 1716454225135818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991078, "dur": 10,
    "args": {
      "External id": 231671,
      "cbid": 211, "correlation": 231671
    }
  },
  {
    "ph": "s", "id": 231671, "pid": 76337, "tid": -914061504, "ts": 1716454224991078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225135842, "dur": 4,
    "args": {
      "External id": 231673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231673,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 231673, "pid": 5, "tid": 7, "ts": 1716454225135842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991094, "dur": 6,
    "args": {
      "External id": 231673,
      "cbid": 211, "correlation": 231673
    }
  },
  {
    "ph": "s", "id": 231673, "pid": 76337, "tid": -914061504, "ts": 1716454224991094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991103, "dur": 0,
    "args": {
      "External id": 231674,
      "cbid": 51, "correlation": 231674
    }
  },
  {
    "ph": "s", "id": 231674, "pid": 76337, "tid": -914061504, "ts": 1716454224991103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225135847, "dur": 183,
    "args": {
      "External id": 231675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231675,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231675, "pid": 5, "tid": 7, "ts": 1716454225135847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991104, "dur": 194,
    "args": {
      "External id": 231675,
      "cbid": 211, "correlation": 231675
    }
  },
  {
    "ph": "s", "id": 231675, "pid": 76337, "tid": -914061504, "ts": 1716454224991104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225136032, "dur": 6,
    "args": {
      "External id": 231676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231676,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231676, "pid": 5, "tid": 7, "ts": 1716454225136032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991301, "dur": 5,
    "args": {
      "External id": 231676,
      "cbid": 211, "correlation": 231676
    }
  },
  {
    "ph": "s", "id": 231676, "pid": 76337, "tid": -914061504, "ts": 1716454224991301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225136039, "dur": 5,
    "args": {
      "External id": 231682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 231682, "pid": 5, "tid": 7, "ts": 1716454225136039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224991331, "dur": 9,
    "args": {
      "External id": 231682,
      "cbid": 211, "correlation": 231682
    }
  },
  {
    "ph": "s", "id": 231682, "pid": 76337, "tid": -914061504, "ts": 1716454224991331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225136045, "dur": 3,
    "args": {
      "External id": 231690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231690,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231690, "pid": 5, "tid": 7, "ts": 1716454225136045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993007, "dur": 14,
    "args": {
      "External id": 231690,
      "cbid": 211, "correlation": 231690
    }
  },
  {
    "ph": "s", "id": 231690, "pid": 76337, "tid": -914061504, "ts": 1716454224993007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225136049, "dur": 3,
    "args": {
      "External id": 231698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231698,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231698, "pid": 5, "tid": 7, "ts": 1716454225136049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993048, "dur": 10,
    "args": {
      "External id": 231698,
      "cbid": 211, "correlation": 231698
    }
  },
  {
    "ph": "s", "id": 231698, "pid": 76337, "tid": -914061504, "ts": 1716454224993048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225136054, "dur": 3,
    "args": {
      "External id": 231706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231706,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231706, "pid": 5, "tid": 7, "ts": 1716454225136054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993074, "dur": 9,
    "args": {
      "External id": 231706,
      "cbid": 211, "correlation": 231706
    }
  },
  {
    "ph": "s", "id": 231706, "pid": 76337, "tid": -914061504, "ts": 1716454224993074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225136057, "dur": 3,
    "args": {
      "External id": 231715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231715,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231715, "pid": 5, "tid": 7, "ts": 1716454225136057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993245, "dur": 14,
    "args": {
      "External id": 231715,
      "cbid": 211, "correlation": 231715
    }
  },
  {
    "ph": "s", "id": 231715, "pid": 76337, "tid": -914061504, "ts": 1716454224993245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225136061, "dur": 3,
    "args": {
      "External id": 231724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231724,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231724, "pid": 5, "tid": 7, "ts": 1716454225136061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993274, "dur": 7,
    "args": {
      "External id": 231724,
      "cbid": 211, "correlation": 231724
    }
  },
  {
    "ph": "s", "id": 231724, "pid": 76337, "tid": -914061504, "ts": 1716454224993274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225136065, "dur": 3,
    "args": {
      "External id": 231732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231732,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231732, "pid": 5, "tid": 7, "ts": 1716454225136065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993300, "dur": 8,
    "args": {
      "External id": 231732,
      "cbid": 211, "correlation": 231732
    }
  },
  {
    "ph": "s", "id": 231732, "pid": 76337, "tid": -914061504, "ts": 1716454224993300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225136070, "dur": 3,
    "args": {
      "External id": 231740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231740,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231740, "pid": 5, "tid": 7, "ts": 1716454225136070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993563, "dur": 15,
    "args": {
      "External id": 231740,
      "cbid": 211, "correlation": 231740
    }
  },
  {
    "ph": "s", "id": 231740, "pid": 76337, "tid": -914061504, "ts": 1716454224993563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225136074, "dur": 3,
    "args": {
      "External id": 231748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231748,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231748, "pid": 5, "tid": 7, "ts": 1716454225136074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993594, "dur": 8,
    "args": {
      "External id": 231748,
      "cbid": 211, "correlation": 231748
    }
  },
  {
    "ph": "s", "id": 231748, "pid": 76337, "tid": -914061504, "ts": 1716454224993594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225136079, "dur": 1,
    "args": {
      "External id": 231758,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231758,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 231758, "pid": 5, "tid": 7, "ts": 1716454225136079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993659, "dur": 34,
    "args": {
      "External id": 231758,
      "cbid": 41, "correlation": 231758
    }
  },
  {
    "ph": "s", "id": 231758, "pid": 76337, "tid": -914061504, "ts": 1716454224993659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224993695, "dur": 142404,
    "args": {
      "External id": 231759,
      "cbid": 131, "correlation": 231759
    }
  },
  {
    "ph": "f", "id": 231759, "pid": 76337, "tid": -914061504, "ts": 1716454224993695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225136377, "dur": 3,
    "args": {
      "External id": 231767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231767, "pid": 5, "tid": 7, "ts": 1716454225136377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136336, "dur": 45,
    "args": {
      "External id": 231767,
      "cbid": 211, "correlation": 231767
    }
  },
  {
    "ph": "s", "id": 231767, "pid": 76337, "tid": -914061504, "ts": 1716454225136336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225136495, "dur": 3,
    "args": {
      "External id": 231776,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231776,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0024271844660194173
    }
  },
  {
    "ph": "f", "id": 231776, "pid": 5, "tid": 7, "ts": 1716454225136495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136456, "dur": 41,
    "args": {
      "External id": 231776,
      "cbid": 41, "correlation": 231776
    }
  },
  {
    "ph": "s", "id": 231776, "pid": 76337, "tid": -914061504, "ts": 1716454225136456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225136610, "dur": 4,
    "args": {
      "External id": 231786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231786,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 231786, "pid": 5, "tid": 7, "ts": 1716454225136610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136593, "dur": 17,
    "args": {
      "External id": 231786,
      "cbid": 211, "correlation": 231786
    }
  },
  {
    "ph": "s", "id": 231786, "pid": 76337, "tid": -914061504, "ts": 1716454225136593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225136702, "dur": 1,
    "args": {
      "External id": 231796,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231796,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 231796, "pid": 5, "tid": 7, "ts": 1716454225136702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136682, "dur": 17,
    "args": {
      "External id": 231796,
      "cbid": 41, "correlation": 231796
    }
  },
  {
    "ph": "s", "id": 231796, "pid": 76337, "tid": -914061504, "ts": 1716454225136682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136701, "dur": 8,
    "args": {
      "External id": 231797,
      "cbid": 131, "correlation": 231797
    }
  },
  {
    "ph": "f", "id": 231797, "pid": 76337, "tid": -914061504, "ts": 1716454225136701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225136782, "dur": 3,
    "args": {
      "External id": 231804,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231804,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.03030303030303
    }
  },
  {
    "ph": "f", "id": 231804, "pid": 5, "tid": 7, "ts": 1716454225136782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136759, "dur": 23,
    "args": {
      "External id": 231804,
      "cbid": 41, "correlation": 231804
    }
  },
  {
    "ph": "s", "id": 231804, "pid": 76337, "tid": -914061504, "ts": 1716454225136759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225136881, "dur": 2,
    "args": {
      "External id": 231823,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231823,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005376344086021506
    }
  },
  {
    "ph": "f", "id": 231823, "pid": 5, "tid": 7, "ts": 1716454225136881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136860, "dur": 20,
    "args": {
      "External id": 231823,
      "cbid": 41, "correlation": 231823
    }
  },
  {
    "ph": "s", "id": 231823, "pid": 76337, "tid": -914061504, "ts": 1716454225136860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225136919, "dur": 3,
    "args": {
      "External id": 231829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231829, "pid": 5, "tid": 7, "ts": 1716454225136919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136908, "dur": 11,
    "args": {
      "External id": 231829,
      "cbid": 211, "correlation": 231829
    }
  },
  {
    "ph": "s", "id": 231829, "pid": 76337, "tid": -914061504, "ts": 1716454225136908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225136933, "dur": 6,
    "args": {
      "External id": 231831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231831,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 231831, "pid": 5, "tid": 7, "ts": 1716454225136933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136923, "dur": 9,
    "args": {
      "External id": 231831,
      "cbid": 211, "correlation": 231831
    }
  },
  {
    "ph": "s", "id": 231831, "pid": 76337, "tid": -914061504, "ts": 1716454225136923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454225136941, "dur": 3,
    "args": {
      "External id": 231833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231833, "pid": 5, "tid": 7, "ts": 1716454225136941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136934, "dur": 6,
    "args": {
      "External id": 231833,
      "cbid": 211, "correlation": 231833
    }
  },
  {
    "ph": "s", "id": 231833, "pid": 76337, "tid": -914061504, "ts": 1716454225136934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225136984, "dur": 2,
    "args": {
      "External id": 231841,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231841,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0028089887640449437
    }
  },
  {
    "ph": "f", "id": 231841, "pid": 5, "tid": 7, "ts": 1716454225136984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225136961, "dur": 23,
    "args": {
      "External id": 231841,
      "cbid": 41, "correlation": 231841
    }
  },
  {
    "ph": "s", "id": 231841, "pid": 76337, "tid": -914061504, "ts": 1716454225136961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225137039, "dur": 3,
    "args": {
      "External id": 231855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231855, "pid": 5, "tid": 7, "ts": 1716454225137039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137027, "dur": 14,
    "args": {
      "External id": 231855,
      "cbid": 211, "correlation": 231855
    }
  },
  {
    "ph": "s", "id": 231855, "pid": 76337, "tid": -914061504, "ts": 1716454225137027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225137060, "dur": 2,
    "args": {
      "External id": 231869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231869, "pid": 5, "tid": 7, "ts": 1716454225137060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137052, "dur": 7,
    "args": {
      "External id": 231869,
      "cbid": 211, "correlation": 231869
    }
  },
  {
    "ph": "s", "id": 231869, "pid": 76337, "tid": -914061504, "ts": 1716454225137052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225137097, "dur": 6,
    "args": {
      "External id": 231876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231876,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231876, "pid": 5, "tid": 7, "ts": 1716454225137097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137087, "dur": 11,
    "args": {
      "External id": 231876,
      "cbid": 211, "correlation": 231876
    }
  },
  {
    "ph": "s", "id": 231876, "pid": 76337, "tid": -914061504, "ts": 1716454225137087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225137108, "dur": 6,
    "args": {
      "External id": 231879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231879,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231879, "pid": 5, "tid": 7, "ts": 1716454225137108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137101, "dur": 7,
    "args": {
      "External id": 231879,
      "cbid": 211, "correlation": 231879
    }
  },
  {
    "ph": "s", "id": 231879, "pid": 76337, "tid": -914061504, "ts": 1716454225137101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225137117, "dur": 3,
    "args": {
      "External id": 231881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231881,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231881, "pid": 5, "tid": 7, "ts": 1716454225137117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137109, "dur": 7,
    "args": {
      "External id": 231881,
      "cbid": 211, "correlation": 231881
    }
  },
  {
    "ph": "s", "id": 231881, "pid": 76337, "tid": -914061504, "ts": 1716454225137109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225137139, "dur": 2,
    "args": {
      "External id": 231884,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231884,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002840909090909091
    }
  },
  {
    "ph": "f", "id": 231884, "pid": 5, "tid": 7, "ts": 1716454225137139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137126, "dur": 12,
    "args": {
      "External id": 231884,
      "cbid": 41, "correlation": 231884
    }
  },
  {
    "ph": "s", "id": 231884, "pid": 76337, "tid": -914061504, "ts": 1716454225137126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225137205, "dur": 4,
    "args": {
      "External id": 231900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231900,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 231900, "pid": 5, "tid": 7, "ts": 1716454225137205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137183, "dur": 23,
    "args": {
      "External id": 231900,
      "cbid": 211, "correlation": 231900
    }
  },
  {
    "ph": "s", "id": 231900, "pid": 76337, "tid": -914061504, "ts": 1716454225137183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225137228, "dur": 3,
    "args": {
      "External id": 231905,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231905,
      "bytes": 1, "memory bandwidth (GB/s)": 0.00028935185185185184
    }
  },
  {
    "ph": "f", "id": 231905, "pid": 5, "tid": 7, "ts": 1716454225137228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137212, "dur": 14,
    "args": {
      "External id": 231905,
      "cbid": 41, "correlation": 231905
    }
  },
  {
    "ph": "s", "id": 231905, "pid": 76337, "tid": -914061504, "ts": 1716454225137212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225137254, "dur": 1,
    "args": {
      "External id": 231911,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 231911,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005896226415094339
    }
  },
  {
    "ph": "f", "id": 231911, "pid": 5, "tid": 7, "ts": 1716454225137254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137236, "dur": 27,
    "args": {
      "External id": 231911,
      "cbid": 41, "correlation": 231911
    }
  },
  {
    "ph": "s", "id": 231911, "pid": 76337, "tid": -914061504, "ts": 1716454225137236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137264, "dur": 4,
    "args": {
      "External id": 231912,
      "cbid": 131, "correlation": 231912
    }
  },
  {
    "ph": "f", "id": 231912, "pid": 76337, "tid": -914061504, "ts": 1716454225137264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225137324, "dur": 3,
    "args": {
      "External id": 231920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231920,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231920, "pid": 5, "tid": 7, "ts": 1716454225137324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137310, "dur": 15,
    "args": {
      "External id": 231920,
      "cbid": 211, "correlation": 231920
    }
  },
  {
    "ph": "s", "id": 231920, "pid": 76337, "tid": -914061504, "ts": 1716454225137310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225137357, "dur": 3,
    "args": {
      "External id": 231930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231930,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231930, "pid": 5, "tid": 7, "ts": 1716454225137357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137347, "dur": 10,
    "args": {
      "External id": 231930,
      "cbid": 211, "correlation": 231930
    }
  },
  {
    "ph": "s", "id": 231930, "pid": 76337, "tid": -914061504, "ts": 1716454225137347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225137393, "dur": 3,
    "args": {
      "External id": 231939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231939,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231939, "pid": 5, "tid": 7, "ts": 1716454225137393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137377, "dur": 15,
    "args": {
      "External id": 231939,
      "cbid": 211, "correlation": 231939
    }
  },
  {
    "ph": "s", "id": 231939, "pid": 76337, "tid": -914061504, "ts": 1716454225137377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225137518, "dur": 12,
    "args": {
      "External id": 231949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231949,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231949, "pid": 5, "tid": 7, "ts": 1716454225137518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137504, "dur": 15,
    "args": {
      "External id": 231949,
      "cbid": 211, "correlation": 231949
    }
  },
  {
    "ph": "s", "id": 231949, "pid": 76337, "tid": -914061504, "ts": 1716454225137504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225137560, "dur": 3,
    "args": {
      "External id": 231957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231957,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231957, "pid": 5, "tid": 7, "ts": 1716454225137560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137550, "dur": 9,
    "args": {
      "External id": 231957,
      "cbid": 211, "correlation": 231957
    }
  },
  {
    "ph": "s", "id": 231957, "pid": 76337, "tid": -914061504, "ts": 1716454225137550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225137611, "dur": 11,
    "args": {
      "External id": 231967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231967,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231967, "pid": 5, "tid": 7, "ts": 1716454225137611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137599, "dur": 12,
    "args": {
      "External id": 231967,
      "cbid": 211, "correlation": 231967
    }
  },
  {
    "ph": "s", "id": 231967, "pid": 76337, "tid": -914061504, "ts": 1716454225137599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225137651, "dur": 10,
    "args": {
      "External id": 231975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231975,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231975, "pid": 5, "tid": 7, "ts": 1716454225137651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137638, "dur": 12,
    "args": {
      "External id": 231975,
      "cbid": 211, "correlation": 231975
    }
  },
  {
    "ph": "s", "id": 231975, "pid": 76337, "tid": -914061504, "ts": 1716454225137638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225137681, "dur": 3,
    "args": {
      "External id": 231984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231984, "pid": 5, "tid": 7, "ts": 1716454225137681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137672, "dur": 9,
    "args": {
      "External id": 231984,
      "cbid": 211, "correlation": 231984
    }
  },
  {
    "ph": "s", "id": 231984, "pid": 76337, "tid": -914061504, "ts": 1716454225137672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225137708, "dur": 5,
    "args": {
      "External id": 231993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 231993,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 231993, "pid": 5, "tid": 7, "ts": 1716454225137708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137699, "dur": 8,
    "args": {
      "External id": 231993,
      "cbid": 211, "correlation": 231993
    }
  },
  {
    "ph": "s", "id": 231993, "pid": 76337, "tid": -914061504, "ts": 1716454225137699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225137750, "dur": 8,
    "args": {
      "External id": 232003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232003,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232003, "pid": 5, "tid": 7, "ts": 1716454225137750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225137740, "dur": 11,
    "args": {
      "External id": 232003,
      "cbid": 211, "correlation": 232003
    }
  },
  {
    "ph": "s", "id": 232003, "pid": 76337, "tid": -914061504, "ts": 1716454225137740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225138116, "dur": 3,
    "args": {
      "External id": 232012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232012,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232012, "pid": 5, "tid": 7, "ts": 1716454225138116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138100, "dur": 16,
    "args": {
      "External id": 232012,
      "cbid": 211, "correlation": 232012
    }
  },
  {
    "ph": "s", "id": 232012, "pid": 76337, "tid": -914061504, "ts": 1716454225138100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225138150, "dur": 3,
    "args": {
      "External id": 232020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232020,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232020, "pid": 5, "tid": 7, "ts": 1716454225138150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138140, "dur": 10,
    "args": {
      "External id": 232020,
      "cbid": 211, "correlation": 232020
    }
  },
  {
    "ph": "s", "id": 232020, "pid": 76337, "tid": -914061504, "ts": 1716454225138140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225138203, "dur": 1,
    "args": {
      "External id": 232030,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232030,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 232030, "pid": 5, "tid": 7, "ts": 1716454225138203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138188, "dur": 12,
    "args": {
      "External id": 232030,
      "cbid": 41, "correlation": 232030
    }
  },
  {
    "ph": "s", "id": 232030, "pid": 76337, "tid": -914061504, "ts": 1716454225138188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138202, "dur": 9,
    "args": {
      "External id": 232031,
      "cbid": 131, "correlation": 232031
    }
  },
  {
    "ph": "f", "id": 232031, "pid": 76337, "tid": -914061504, "ts": 1716454225138202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225138295, "dur": 2,
    "args": {
      "External id": 232039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232039, "pid": 5, "tid": 7, "ts": 1716454225138295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138280, "dur": 14,
    "args": {
      "External id": 232039,
      "cbid": 211, "correlation": 232039
    }
  },
  {
    "ph": "s", "id": 232039, "pid": 76337, "tid": -914061504, "ts": 1716454225138280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225138369, "dur": 3,
    "args": {
      "External id": 232048,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232048,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002631578947368421
    }
  },
  {
    "ph": "f", "id": 232048, "pid": 5, "tid": 7, "ts": 1716454225138369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138351, "dur": 18,
    "args": {
      "External id": 232048,
      "cbid": 41, "correlation": 232048
    }
  },
  {
    "ph": "s", "id": 232048, "pid": 76337, "tid": -914061504, "ts": 1716454225138351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225138443, "dur": 3,
    "args": {
      "External id": 232058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232058,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232058, "pid": 5, "tid": 7, "ts": 1716454225138443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138429, "dur": 13,
    "args": {
      "External id": 232058,
      "cbid": 211, "correlation": 232058
    }
  },
  {
    "ph": "s", "id": 232058, "pid": 76337, "tid": -914061504, "ts": 1716454225138429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225138494, "dur": 1,
    "args": {
      "External id": 232068,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232068,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 232068, "pid": 5, "tid": 7, "ts": 1716454225138494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138480, "dur": 12,
    "args": {
      "External id": 232068,
      "cbid": 41, "correlation": 232068
    }
  },
  {
    "ph": "s", "id": 232068, "pid": 76337, "tid": -914061504, "ts": 1716454225138480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138493, "dur": 9,
    "args": {
      "External id": 232069,
      "cbid": 131, "correlation": 232069
    }
  },
  {
    "ph": "f", "id": 232069, "pid": 76337, "tid": -914061504, "ts": 1716454225138493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225138555, "dur": 3,
    "args": {
      "External id": 232076,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232076,
      "bytes": 98304, "memory bandwidth (GB/s)": 32
    }
  },
  {
    "ph": "f", "id": 232076, "pid": 5, "tid": 7, "ts": 1716454225138555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138536, "dur": 18,
    "args": {
      "External id": 232076,
      "cbid": 41, "correlation": 232076
    }
  },
  {
    "ph": "s", "id": 232076, "pid": 76337, "tid": -914061504, "ts": 1716454225138536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225138602, "dur": 1,
    "args": {
      "External id": 232087,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232087,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 232087, "pid": 5, "tid": 7, "ts": 1716454225138602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138590, "dur": 10,
    "args": {
      "External id": 232087,
      "cbid": 41, "correlation": 232087
    }
  },
  {
    "ph": "s", "id": 232087, "pid": 76337, "tid": -914061504, "ts": 1716454225138590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138601, "dur": 8,
    "args": {
      "External id": 232088,
      "cbid": 131, "correlation": 232088
    }
  },
  {
    "ph": "f", "id": 232088, "pid": 76337, "tid": -914061504, "ts": 1716454225138601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225138651, "dur": 3,
    "args": {
      "External id": 232096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232096,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232096, "pid": 5, "tid": 7, "ts": 1716454225138651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138637, "dur": 13,
    "args": {
      "External id": 232096,
      "cbid": 211, "correlation": 232096
    }
  },
  {
    "ph": "s", "id": 232096, "pid": 76337, "tid": -914061504, "ts": 1716454225138637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225138681, "dur": 3,
    "args": {
      "External id": 232106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232106,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232106, "pid": 5, "tid": 7, "ts": 1716454225138681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138671, "dur": 8,
    "args": {
      "External id": 232106,
      "cbid": 211, "correlation": 232106
    }
  },
  {
    "ph": "s", "id": 232106, "pid": 76337, "tid": -914061504, "ts": 1716454225138671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225138702, "dur": 3,
    "args": {
      "External id": 232115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232115,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232115, "pid": 5, "tid": 7, "ts": 1716454225138702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138694, "dur": 7,
    "args": {
      "External id": 232115,
      "cbid": 211, "correlation": 232115
    }
  },
  {
    "ph": "s", "id": 232115, "pid": 76337, "tid": -914061504, "ts": 1716454225138694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225138771, "dur": 5,
    "args": {
      "External id": 232123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232123, "pid": 5, "tid": 7, "ts": 1716454225138771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138757, "dur": 15,
    "args": {
      "External id": 232123,
      "cbid": 211, "correlation": 232123
    }
  },
  {
    "ph": "s", "id": 232123, "pid": 76337, "tid": -914061504, "ts": 1716454225138757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225138811, "dur": 3,
    "args": {
      "External id": 232132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232132,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232132, "pid": 5, "tid": 7, "ts": 1716454225138811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138801, "dur": 10,
    "args": {
      "External id": 232132,
      "cbid": 211, "correlation": 232132
    }
  },
  {
    "ph": "s", "id": 232132, "pid": 76337, "tid": -914061504, "ts": 1716454225138801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225138835, "dur": 3,
    "args": {
      "External id": 232141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232141,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232141, "pid": 5, "tid": 7, "ts": 1716454225138835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138826, "dur": 7,
    "args": {
      "External id": 232141,
      "cbid": 211, "correlation": 232141
    }
  },
  {
    "ph": "s", "id": 232141, "pid": 76337, "tid": -914061504, "ts": 1716454225138826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225138898, "dur": 3,
    "args": {
      "External id": 232149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232149,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232149, "pid": 5, "tid": 7, "ts": 1716454225138898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138887, "dur": 10,
    "args": {
      "External id": 232149,
      "cbid": 211, "correlation": 232149
    }
  },
  {
    "ph": "s", "id": 232149, "pid": 76337, "tid": -914061504, "ts": 1716454225138887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225138955, "dur": 1,
    "args": {
      "External id": 232157,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232157,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 232157, "pid": 5, "tid": 7, "ts": 1716454225138955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138940, "dur": 25,
    "args": {
      "External id": 232157,
      "cbid": 41, "correlation": 232157
    }
  },
  {
    "ph": "s", "id": 232157, "pid": 76337, "tid": -914061504, "ts": 1716454225138940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225138966, "dur": 3,
    "args": {
      "External id": 232158,
      "cbid": 131, "correlation": 232158
    }
  },
  {
    "ph": "f", "id": 232158, "pid": 76337, "tid": -914061504, "ts": 1716454225138966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225139036, "dur": 1,
    "args": {
      "External id": 232168,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232168,
      "bytes": 42, "memory bandwidth (GB/s)": 0.027925531914893616
    }
  },
  {
    "ph": "f", "id": 232168, "pid": 5, "tid": 7, "ts": 1716454225139036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139022, "dur": 11,
    "args": {
      "External id": 232168,
      "cbid": 41, "correlation": 232168
    }
  },
  {
    "ph": "s", "id": 232168, "pid": 76337, "tid": -914061504, "ts": 1716454225139022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139033, "dur": 9,
    "args": {
      "External id": 232169,
      "cbid": 131, "correlation": 232169
    }
  },
  {
    "ph": "f", "id": 232169, "pid": 76337, "tid": -914061504, "ts": 1716454225139033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225139093, "dur": 1,
    "args": {
      "External id": 232178,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232178,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 232178, "pid": 5, "tid": 7, "ts": 1716454225139093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139081, "dur": 9,
    "args": {
      "External id": 232178,
      "cbid": 41, "correlation": 232178
    }
  },
  {
    "ph": "s", "id": 232178, "pid": 76337, "tid": -914061504, "ts": 1716454225139081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139091, "dur": 8,
    "args": {
      "External id": 232179,
      "cbid": 131, "correlation": 232179
    }
  },
  {
    "ph": "f", "id": 232179, "pid": 76337, "tid": -914061504, "ts": 1716454225139091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225139170, "dur": 4,
    "args": {
      "External id": 232186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232186, "pid": 5, "tid": 7, "ts": 1716454225139170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139154, "dur": 17,
    "args": {
      "External id": 232186,
      "cbid": 211, "correlation": 232186
    }
  },
  {
    "ph": "s", "id": 232186, "pid": 76337, "tid": -914061504, "ts": 1716454225139154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454225139213, "dur": 4,
    "args": {
      "External id": 232206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232206,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232206, "pid": 5, "tid": 7, "ts": 1716454225139213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139200, "dur": 15,
    "args": {
      "External id": 232206,
      "cbid": 211, "correlation": 232206
    }
  },
  {
    "ph": "s", "id": 232206, "pid": 76337, "tid": -914061504, "ts": 1716454225139200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139216, "dur": 0,
    "args": {
      "External id": 232207,
      "cbid": 11, "correlation": 232207
    }
  },
  {
    "ph": "f", "id": 232207, "pid": 76337, "tid": -914061504, "ts": 1716454225139216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139216, "dur": 0,
    "args": {
      "External id": 232208,
      "cbid": 11, "correlation": 232208
    }
  },
  {
    "ph": "f", "id": 232208, "pid": 76337, "tid": -914061504, "ts": 1716454225139216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225139229, "dur": 1,
    "args": {
      "External id": 232211,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232211,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0024509803921568627
    }
  },
  {
    "ph": "f", "id": 232211, "pid": 5, "tid": 7, "ts": 1716454225139229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139217, "dur": 21,
    "args": {
      "External id": 232211,
      "cbid": 41, "correlation": 232211
    }
  },
  {
    "ph": "s", "id": 232211, "pid": 76337, "tid": -914061504, "ts": 1716454225139217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139239, "dur": 3,
    "args": {
      "External id": 232212,
      "cbid": 131, "correlation": 232212
    }
  },
  {
    "ph": "f", "id": 232212, "pid": 76337, "tid": -914061504, "ts": 1716454225139239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225139279, "dur": 2,
    "args": {
      "External id": 232236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232236, "pid": 5, "tid": 7, "ts": 1716454225139279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139260, "dur": 18,
    "args": {
      "External id": 232236,
      "cbid": 211, "correlation": 232236
    }
  },
  {
    "ph": "s", "id": 232236, "pid": 76337, "tid": -914061504, "ts": 1716454225139260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139279, "dur": 0,
    "args": {
      "External id": 232237,
      "cbid": 11, "correlation": 232237
    }
  },
  {
    "ph": "f", "id": 232237, "pid": 76337, "tid": -914061504, "ts": 1716454225139279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139280, "dur": 0,
    "args": {
      "External id": 232238,
      "cbid": 11, "correlation": 232238
    }
  },
  {
    "ph": "f", "id": 232238, "pid": 76337, "tid": -914061504, "ts": 1716454225139280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139281, "dur": 1,
    "args": {
      "External id": 232240,
      "cbid": 200, "correlation": 232240
    }
  },
  {
    "ph": "f", "id": 232240, "pid": 76337, "tid": -914061504, "ts": 1716454225139281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225139294, "dur": 4,
    "args": {
      "External id": 232242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232242,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232242, "pid": 5, "tid": 7, "ts": 1716454225139294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139285, "dur": 10,
    "args": {
      "External id": 232242,
      "cbid": 211, "correlation": 232242
    }
  },
  {
    "ph": "s", "id": 232242, "pid": 76337, "tid": -914061504, "ts": 1716454225139285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139295, "dur": 0,
    "args": {
      "External id": 232243,
      "cbid": 11, "correlation": 232243
    }
  },
  {
    "ph": "f", "id": 232243, "pid": 76337, "tid": -914061504, "ts": 1716454225139295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139296, "dur": 0,
    "args": {
      "External id": 232244,
      "cbid": 11, "correlation": 232244
    }
  },
  {
    "ph": "f", "id": 232244, "pid": 76337, "tid": -914061504, "ts": 1716454225139296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225139334, "dur": 1,
    "args": {
      "External id": 232251,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232251,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004901960784313725
    }
  },
  {
    "ph": "f", "id": 232251, "pid": 5, "tid": 7, "ts": 1716454225139334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139322, "dur": 22,
    "args": {
      "External id": 232251,
      "cbid": 41, "correlation": 232251
    }
  },
  {
    "ph": "s", "id": 232251, "pid": 76337, "tid": -914061504, "ts": 1716454225139322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139344, "dur": 3,
    "args": {
      "External id": 232252,
      "cbid": 131, "correlation": 232252
    }
  },
  {
    "ph": "f", "id": 232252, "pid": 76337, "tid": -914061504, "ts": 1716454225139344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225139394, "dur": 1,
    "args": {
      "External id": 232262,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 232262,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 232262, "pid": 5, "tid": 7, "ts": 1716454225139394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139382, "dur": 10,
    "args": {
      "External id": 232262,
      "cbid": 41, "correlation": 232262
    }
  },
  {
    "ph": "s", "id": 232262, "pid": 76337, "tid": -914061504, "ts": 1716454225139382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139392, "dur": 7,
    "args": {
      "External id": 232263,
      "cbid": 131, "correlation": 232263
    }
  },
  {
    "ph": "f", "id": 232263, "pid": 76337, "tid": -914061504, "ts": 1716454225139392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225139465, "dur": 5,
    "args": {
      "External id": 232270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232270,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232270, "pid": 5, "tid": 7, "ts": 1716454225139465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139449, "dur": 15,
    "args": {
      "External id": 232270,
      "cbid": 211, "correlation": 232270
    }
  },
  {
    "ph": "s", "id": 232270, "pid": 76337, "tid": -914061504, "ts": 1716454225139449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225139539, "dur": 3,
    "args": {
      "External id": 232279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232279,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232279, "pid": 5, "tid": 7, "ts": 1716454225139539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139524, "dur": 14,
    "args": {
      "External id": 232279,
      "cbid": 211, "correlation": 232279
    }
  },
  {
    "ph": "s", "id": 232279, "pid": 76337, "tid": -914061504, "ts": 1716454225139524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225139578, "dur": 3,
    "args": {
      "External id": 232287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232287,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232287, "pid": 5, "tid": 7, "ts": 1716454225139578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139565, "dur": 12,
    "args": {
      "External id": 232287,
      "cbid": 211, "correlation": 232287
    }
  },
  {
    "ph": "s", "id": 232287, "pid": 76337, "tid": -914061504, "ts": 1716454225139565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225139610, "dur": 4,
    "args": {
      "External id": 232295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232295,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232295, "pid": 5, "tid": 7, "ts": 1716454225139610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139599, "dur": 11,
    "args": {
      "External id": 232295,
      "cbid": 211, "correlation": 232295
    }
  },
  {
    "ph": "s", "id": 232295, "pid": 76337, "tid": -914061504, "ts": 1716454225139599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225139644, "dur": 4,
    "args": {
      "External id": 232303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232303,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232303, "pid": 5, "tid": 7, "ts": 1716454225139644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139631, "dur": 12,
    "args": {
      "External id": 232303,
      "cbid": 211, "correlation": 232303
    }
  },
  {
    "ph": "s", "id": 232303, "pid": 76337, "tid": -914061504, "ts": 1716454225139631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225139671, "dur": 2,
    "args": {
      "External id": 232311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232311,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232311, "pid": 5, "tid": 7, "ts": 1716454225139671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139662, "dur": 8,
    "args": {
      "External id": 232311,
      "cbid": 211, "correlation": 232311
    }
  },
  {
    "ph": "s", "id": 232311, "pid": 76337, "tid": -914061504, "ts": 1716454225139662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225139697, "dur": 3,
    "args": {
      "External id": 232319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232319,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232319, "pid": 5, "tid": 7, "ts": 1716454225139697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139688, "dur": 9,
    "args": {
      "External id": 232319,
      "cbid": 211, "correlation": 232319
    }
  },
  {
    "ph": "s", "id": 232319, "pid": 76337, "tid": -914061504, "ts": 1716454225139688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225139718, "dur": 4,
    "args": {
      "External id": 232327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232327, "pid": 5, "tid": 7, "ts": 1716454225139718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139710, "dur": 7,
    "args": {
      "External id": 232327,
      "cbid": 211, "correlation": 232327
    }
  },
  {
    "ph": "s", "id": 232327, "pid": 76337, "tid": -914061504, "ts": 1716454225139710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225139738, "dur": 5,
    "args": {
      "External id": 232335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232335,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232335, "pid": 5, "tid": 7, "ts": 1716454225139738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139730, "dur": 7,
    "args": {
      "External id": 232335,
      "cbid": 211, "correlation": 232335
    }
  },
  {
    "ph": "s", "id": 232335, "pid": 76337, "tid": -914061504, "ts": 1716454225139730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225139759, "dur": 3,
    "args": {
      "External id": 232343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232343,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232343, "pid": 5, "tid": 7, "ts": 1716454225139759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139750, "dur": 7,
    "args": {
      "External id": 232343,
      "cbid": 211, "correlation": 232343
    }
  },
  {
    "ph": "s", "id": 232343, "pid": 76337, "tid": -914061504, "ts": 1716454225139750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225139822, "dur": 3,
    "args": {
      "External id": 232351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232351,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232351, "pid": 5, "tid": 7, "ts": 1716454225139822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139809, "dur": 13,
    "args": {
      "External id": 232351,
      "cbid": 211, "correlation": 232351
    }
  },
  {
    "ph": "s", "id": 232351, "pid": 76337, "tid": -914061504, "ts": 1716454225139809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225139849, "dur": 4,
    "args": {
      "External id": 232359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232359, "pid": 5, "tid": 7, "ts": 1716454225139849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139840, "dur": 8,
    "args": {
      "External id": 232359,
      "cbid": 211, "correlation": 232359
    }
  },
  {
    "ph": "s", "id": 232359, "pid": 76337, "tid": -914061504, "ts": 1716454225139840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225139870, "dur": 4,
    "args": {
      "External id": 232367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232367, "pid": 5, "tid": 7, "ts": 1716454225139870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139862, "dur": 7,
    "args": {
      "External id": 232367,
      "cbid": 211, "correlation": 232367
    }
  },
  {
    "ph": "s", "id": 232367, "pid": 76337, "tid": -914061504, "ts": 1716454225139862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225139890, "dur": 3,
    "args": {
      "External id": 232375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232375,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 232375, "pid": 5, "tid": 7, "ts": 1716454225139890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225139882, "dur": 6,
    "args": {
      "External id": 232375,
      "cbid": 211, "correlation": 232375
    }
  },
  {
    "ph": "s", "id": 232375, "pid": 76337, "tid": -914061504, "ts": 1716454225139882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225140327, "dur": 5,
    "args": {
      "External id": 232384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232384,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232384, "pid": 5, "tid": 7, "ts": 1716454225140327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140309, "dur": 17,
    "args": {
      "External id": 232384,
      "cbid": 211, "correlation": 232384
    }
  },
  {
    "ph": "s", "id": 232384, "pid": 76337, "tid": -914061504, "ts": 1716454225140309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225140364, "dur": 5,
    "args": {
      "External id": 232393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232393,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232393, "pid": 5, "tid": 7, "ts": 1716454225140364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140353, "dur": 10,
    "args": {
      "External id": 232393,
      "cbid": 211, "correlation": 232393
    }
  },
  {
    "ph": "s", "id": 232393, "pid": 76337, "tid": -914061504, "ts": 1716454225140353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225140494, "dur": 3,
    "args": {
      "External id": 232409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232409, "pid": 5, "tid": 7, "ts": 1716454225140494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140481, "dur": 14,
    "args": {
      "External id": 232409,
      "cbid": 211, "correlation": 232409
    }
  },
  {
    "ph": "s", "id": 232409, "pid": 76337, "tid": -914061504, "ts": 1716454225140481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225140529, "dur": 3,
    "args": {
      "External id": 232417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232417, "pid": 5, "tid": 7, "ts": 1716454225140529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140520, "dur": 9,
    "args": {
      "External id": 232417,
      "cbid": 211, "correlation": 232417
    }
  },
  {
    "ph": "s", "id": 232417, "pid": 76337, "tid": -914061504, "ts": 1716454225140520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225140560, "dur": 3,
    "args": {
      "External id": 232425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232425,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232425, "pid": 5, "tid": 7, "ts": 1716454225140560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140550, "dur": 9,
    "args": {
      "External id": 232425,
      "cbid": 211, "correlation": 232425
    }
  },
  {
    "ph": "s", "id": 232425, "pid": 76337, "tid": -914061504, "ts": 1716454225140550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225140590, "dur": 3,
    "args": {
      "External id": 232433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232433,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232433, "pid": 5, "tid": 7, "ts": 1716454225140590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140580, "dur": 9,
    "args": {
      "External id": 232433,
      "cbid": 211, "correlation": 232433
    }
  },
  {
    "ph": "s", "id": 232433, "pid": 76337, "tid": -914061504, "ts": 1716454225140580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225140653, "dur": 4,
    "args": {
      "External id": 232445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232445,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232445, "pid": 5, "tid": 7, "ts": 1716454225140653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140637, "dur": 16,
    "args": {
      "External id": 232445,
      "cbid": 211, "correlation": 232445
    }
  },
  {
    "ph": "s", "id": 232445, "pid": 76337, "tid": -914061504, "ts": 1716454225140637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225140701, "dur": 4,
    "args": {
      "External id": 232456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232456, "pid": 5, "tid": 7, "ts": 1716454225140701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140688, "dur": 13,
    "args": {
      "External id": 232456,
      "cbid": 211, "correlation": 232456
    }
  },
  {
    "ph": "s", "id": 232456, "pid": 76337, "tid": -914061504, "ts": 1716454225140688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225140733, "dur": 2,
    "args": {
      "External id": 232464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232464, "pid": 5, "tid": 7, "ts": 1716454225140733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140723, "dur": 9,
    "args": {
      "External id": 232464,
      "cbid": 211, "correlation": 232464
    }
  },
  {
    "ph": "s", "id": 232464, "pid": 76337, "tid": -914061504, "ts": 1716454225140723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225140767, "dur": 5,
    "args": {
      "External id": 232472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232472,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232472, "pid": 5, "tid": 7, "ts": 1716454225140767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140757, "dur": 10,
    "args": {
      "External id": 232472,
      "cbid": 211, "correlation": 232472
    }
  },
  {
    "ph": "s", "id": 232472, "pid": 76337, "tid": -914061504, "ts": 1716454225140757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225140806, "dur": 5,
    "args": {
      "External id": 232480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232480,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232480, "pid": 5, "tid": 7, "ts": 1716454225140806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140791, "dur": 16,
    "args": {
      "External id": 232480,
      "cbid": 211, "correlation": 232480
    }
  },
  {
    "ph": "s", "id": 232480, "pid": 76337, "tid": -914061504, "ts": 1716454225140791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225140838, "dur": 4,
    "args": {
      "External id": 232489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232489,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232489, "pid": 5, "tid": 7, "ts": 1716454225140838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140828, "dur": 10,
    "args": {
      "External id": 232489,
      "cbid": 211, "correlation": 232489
    }
  },
  {
    "ph": "s", "id": 232489, "pid": 76337, "tid": -914061504, "ts": 1716454225140828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225140902, "dur": 4,
    "args": {
      "External id": 232502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232502,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232502, "pid": 5, "tid": 7, "ts": 1716454225140902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140889, "dur": 13,
    "args": {
      "External id": 232502,
      "cbid": 211, "correlation": 232502
    }
  },
  {
    "ph": "s", "id": 232502, "pid": 76337, "tid": -914061504, "ts": 1716454225140889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225140941, "dur": 5,
    "args": {
      "External id": 232512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232512,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 232512, "pid": 5, "tid": 7, "ts": 1716454225140941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225140931, "dur": 10,
    "args": {
      "External id": 232512,
      "cbid": 211, "correlation": 232512
    }
  },
  {
    "ph": "s", "id": 232512, "pid": 76337, "tid": -914061504, "ts": 1716454225140931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141083, "dur": 5,
    "args": {
      "External id": 232529,
      "cbid": 251, "correlation": 232529
    }
  },
  {
    "ph": "f", "id": 232529, "pid": 76337, "tid": -914061504, "ts": 1716454225141083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454225141114, "dur": 11,
    "args": {
      "External id": 232531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232531,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232531, "pid": 5, "tid": 7, "ts": 1716454225141114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141097, "dur": 17,
    "args": {
      "External id": 232531,
      "cbid": 211, "correlation": 232531
    }
  },
  {
    "ph": "s", "id": 232531, "pid": 76337, "tid": -914061504, "ts": 1716454225141097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225141178, "dur": 3,
    "args": {
      "External id": 232539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232539,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232539, "pid": 5, "tid": 7, "ts": 1716454225141178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141165, "dur": 13,
    "args": {
      "External id": 232539,
      "cbid": 211, "correlation": 232539
    }
  },
  {
    "ph": "s", "id": 232539, "pid": 76337, "tid": -914061504, "ts": 1716454225141165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141236, "dur": 2,
    "args": {
      "External id": 232555,
      "cbid": 251, "correlation": 232555
    }
  },
  {
    "ph": "f", "id": 232555, "pid": 76337, "tid": -914061504, "ts": 1716454225141236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141242, "dur": 0,
    "args": {
      "External id": 232557,
      "cbid": 251, "correlation": 232557
    }
  },
  {
    "ph": "f", "id": 232557, "pid": 76337, "tid": -914061504, "ts": 1716454225141242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225141258, "dur": 13,
    "args": {
      "External id": 232558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232558,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232558, "pid": 5, "tid": 7, "ts": 1716454225141258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141245, "dur": 14,
    "args": {
      "External id": 232558,
      "cbid": 211, "correlation": 232558
    }
  },
  {
    "ph": "s", "id": 232558, "pid": 76337, "tid": -914061504, "ts": 1716454225141245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225141274, "dur": 5,
    "args": {
      "External id": 232560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232560,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232560, "pid": 5, "tid": 7, "ts": 1716454225141274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141263, "dur": 10,
    "args": {
      "External id": 232560,
      "cbid": 211, "correlation": 232560
    }
  },
  {
    "ph": "s", "id": 232560, "pid": 76337, "tid": -914061504, "ts": 1716454225141263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141378, "dur": 1,
    "args": {
      "External id": 232570,
      "cbid": 317, "correlation": 232570
    }
  },
  {
    "ph": "f", "id": 232570, "pid": 76337, "tid": -914061504, "ts": 1716454225141378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141380, "dur": 1,
    "args": {
      "External id": 232571,
      "cbid": 203, "correlation": 232571
    }
  },
  {
    "ph": "f", "id": 232571, "pid": 76337, "tid": -914061504, "ts": 1716454225141380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141382, "dur": 1,
    "args": {
      "External id": 232572,
      "cbid": 205, "correlation": 232572
    }
  },
  {
    "ph": "f", "id": 232572, "pid": 76337, "tid": -914061504, "ts": 1716454225141382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225141440, "dur": 7,
    "args": {
      "External id": 232576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232576,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232576, "pid": 5, "tid": 7, "ts": 1716454225141440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141424, "dur": 15,
    "args": {
      "External id": 232576,
      "cbid": 211, "correlation": 232576
    }
  },
  {
    "ph": "s", "id": 232576, "pid": 76337, "tid": -914061504, "ts": 1716454225141424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225141450, "dur": 4,
    "args": {
      "External id": 232578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232578,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 232578, "pid": 5, "tid": 7, "ts": 1716454225141450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141443, "dur": 6,
    "args": {
      "External id": 232578,
      "cbid": 211, "correlation": 232578
    }
  },
  {
    "ph": "s", "id": 232578, "pid": 76337, "tid": -914061504, "ts": 1716454225141443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225141470, "dur": 3,
    "args": {
      "External id": 232580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232580,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232580, "pid": 5, "tid": 7, "ts": 1716454225141470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141461, "dur": 8,
    "args": {
      "External id": 232580,
      "cbid": 211, "correlation": 232580
    }
  },
  {
    "ph": "s", "id": 232580, "pid": 76337, "tid": -914061504, "ts": 1716454225141461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141476, "dur": 0,
    "args": {
      "External id": 232581,
      "cbid": 51, "correlation": 232581
    }
  },
  {
    "ph": "s", "id": 232581, "pid": 76337, "tid": -914061504, "ts": 1716454225141476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225141487, "dur": 82,
    "args": {
      "External id": 232582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232582,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232582, "pid": 5, "tid": 7, "ts": 1716454225141487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141478, "dur": 7,
    "args": {
      "External id": 232582,
      "cbid": 211, "correlation": 232582
    }
  },
  {
    "ph": "s", "id": 232582, "pid": 76337, "tid": -914061504, "ts": 1716454225141478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225141570, "dur": 59,
    "args": {
      "External id": 232587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232587,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232587, "pid": 5, "tid": 7, "ts": 1716454225141570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225141516, "dur": 14,
    "args": {
      "External id": 232587,
      "cbid": 211, "correlation": 232587
    }
  },
  {
    "ph": "s", "id": 232587, "pid": 76337, "tid": -914061504, "ts": 1716454225141516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225143338, "dur": 51,
    "args": {
      "External id": 232607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232607,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 232607, "pid": 5, "tid": 7, "ts": 1716454225143338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143322, "dur": 16,
    "args": {
      "External id": 232607,
      "cbid": 211, "correlation": 232607
    }
  },
  {
    "ph": "s", "id": 232607, "pid": 76337, "tid": -914061504, "ts": 1716454225143322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225143391, "dur": 4,
    "args": {
      "External id": 232619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232619,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232619, "pid": 5, "tid": 7, "ts": 1716454225143391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143349, "dur": 8,
    "args": {
      "External id": 232619,
      "cbid": 211, "correlation": 232619
    }
  },
  {
    "ph": "s", "id": 232619, "pid": 76337, "tid": -914061504, "ts": 1716454225143349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225143396, "dur": 56,
    "args": {
      "External id": 232622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232622,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232622, "pid": 5, "tid": 7, "ts": 1716454225143396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143377, "dur": 8,
    "args": {
      "External id": 232622,
      "cbid": 211, "correlation": 232622
    }
  },
  {
    "ph": "s", "id": 232622, "pid": 76337, "tid": -914061504, "ts": 1716454225143377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225143453, "dur": 36,
    "args": {
      "External id": 232631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232631,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232631, "pid": 5, "tid": 7, "ts": 1716454225143453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143423, "dur": 10,
    "args": {
      "External id": 232631,
      "cbid": 211, "correlation": 232631
    }
  },
  {
    "ph": "s", "id": 232631, "pid": 76337, "tid": -914061504, "ts": 1716454225143423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143479, "dur": 0,
    "args": {
      "External id": 232641,
      "cbid": 317, "correlation": 232641
    }
  },
  {
    "ph": "f", "id": 232641, "pid": 76337, "tid": -914061504, "ts": 1716454225143479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143480, "dur": 0,
    "args": {
      "External id": 232642,
      "cbid": 203, "correlation": 232642
    }
  },
  {
    "ph": "f", "id": 232642, "pid": 76337, "tid": -914061504, "ts": 1716454225143480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143481, "dur": 0,
    "args": {
      "External id": 232643,
      "cbid": 205, "correlation": 232643
    }
  },
  {
    "ph": "f", "id": 232643, "pid": 76337, "tid": -914061504, "ts": 1716454225143481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225143515, "dur": 40,
    "args": {
      "External id": 232647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232647,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232647, "pid": 5, "tid": 7, "ts": 1716454225143515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143502, "dur": 13,
    "args": {
      "External id": 232647,
      "cbid": 211, "correlation": 232647
    }
  },
  {
    "ph": "s", "id": 232647, "pid": 76337, "tid": -914061504, "ts": 1716454225143502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225143557, "dur": 14,
    "args": {
      "External id": 232649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232649,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232649, "pid": 5, "tid": 7, "ts": 1716454225143557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143518, "dur": 6,
    "args": {
      "External id": 232649,
      "cbid": 211, "correlation": 232649
    }
  },
  {
    "ph": "s", "id": 232649, "pid": 76337, "tid": -914061504, "ts": 1716454225143518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225143572, "dur": 3,
    "args": {
      "External id": 232651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232651,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232651, "pid": 5, "tid": 7, "ts": 1716454225143572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143529, "dur": 6,
    "args": {
      "External id": 232651,
      "cbid": 211, "correlation": 232651
    }
  },
  {
    "ph": "s", "id": 232651, "pid": 76337, "tid": -914061504, "ts": 1716454225143529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143539, "dur": 0,
    "args": {
      "External id": 232652,
      "cbid": 51, "correlation": 232652
    }
  },
  {
    "ph": "s", "id": 232652, "pid": 76337, "tid": -914061504, "ts": 1716454225143539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225143577, "dur": 690,
    "args": {
      "External id": 232653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232653,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232653, "pid": 5, "tid": 7, "ts": 1716454225143577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143540, "dur": 6,
    "args": {
      "External id": 232653,
      "cbid": 211, "correlation": 232653
    }
  },
  {
    "ph": "s", "id": 232653, "pid": 76337, "tid": -914061504, "ts": 1716454225143540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225144268, "dur": 58,
    "args": {
      "External id": 232658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232658,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232658, "pid": 5, "tid": 7, "ts": 1716454225144268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143571, "dur": 9,
    "args": {
      "External id": 232658,
      "cbid": 211, "correlation": 232658
    }
  },
  {
    "ph": "s", "id": 232658, "pid": 76337, "tid": -914061504, "ts": 1716454225143571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225144328, "dur": 4,
    "args": {
      "External id": 232666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232666,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232666, "pid": 5, "tid": 7, "ts": 1716454225144328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143615, "dur": 9,
    "args": {
      "External id": 232666,
      "cbid": 211, "correlation": 232666
    }
  },
  {
    "ph": "s", "id": 232666, "pid": 76337, "tid": -914061504, "ts": 1716454225143615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143681, "dur": 2,
    "args": {
      "External id": 232682,
      "cbid": 251, "correlation": 232682
    }
  },
  {
    "ph": "f", "id": 232682, "pid": 76337, "tid": -914061504, "ts": 1716454225143681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143686, "dur": 0,
    "args": {
      "External id": 232684,
      "cbid": 251, "correlation": 232684
    }
  },
  {
    "ph": "f", "id": 232684, "pid": 76337, "tid": -914061504, "ts": 1716454225143686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225144333, "dur": 8,
    "args": {
      "External id": 232685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232685,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 232685, "pid": 5, "tid": 7, "ts": 1716454225144333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143688, "dur": 12,
    "args": {
      "External id": 232685,
      "cbid": 211, "correlation": 232685
    }
  },
  {
    "ph": "s", "id": 232685, "pid": 76337, "tid": -914061504, "ts": 1716454225143688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225144343, "dur": 4,
    "args": {
      "External id": 232687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232687,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 232687, "pid": 5, "tid": 7, "ts": 1716454225144343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143702, "dur": 7,
    "args": {
      "External id": 232687,
      "cbid": 211, "correlation": 232687
    }
  },
  {
    "ph": "s", "id": 232687, "pid": 76337, "tid": -914061504, "ts": 1716454225143702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225144348, "dur": 53,
    "args": {
      "External id": 232697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232697,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232697, "pid": 5, "tid": 7, "ts": 1716454225144348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143764, "dur": 12,
    "args": {
      "External id": 232697,
      "cbid": 211, "correlation": 232697
    }
  },
  {
    "ph": "s", "id": 232697, "pid": 76337, "tid": -914061504, "ts": 1716454225143764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225144402, "dur": 50,
    "args": {
      "External id": 232717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232717,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 232717, "pid": 5, "tid": 7, "ts": 1716454225144402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143830, "dur": 11,
    "args": {
      "External id": 232717,
      "cbid": 211, "correlation": 232717
    }
  },
  {
    "ph": "s", "id": 232717, "pid": 76337, "tid": -914061504, "ts": 1716454225143830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225144454, "dur": 4,
    "args": {
      "External id": 232729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232729,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232729, "pid": 5, "tid": 7, "ts": 1716454225144454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143851, "dur": 6,
    "args": {
      "External id": 232729,
      "cbid": 211, "correlation": 232729
    }
  },
  {
    "ph": "s", "id": 232729, "pid": 76337, "tid": -914061504, "ts": 1716454225143851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225144459, "dur": 55,
    "args": {
      "External id": 232732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232732,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232732, "pid": 5, "tid": 7, "ts": 1716454225144459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143870, "dur": 6,
    "args": {
      "External id": 232732,
      "cbid": 211, "correlation": 232732
    }
  },
  {
    "ph": "s", "id": 232732, "pid": 76337, "tid": -914061504, "ts": 1716454225143870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225144515, "dur": 36,
    "args": {
      "External id": 232741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232741,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232741, "pid": 5, "tid": 7, "ts": 1716454225144515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143910, "dur": 10,
    "args": {
      "External id": 232741,
      "cbid": 211, "correlation": 232741
    }
  },
  {
    "ph": "s", "id": 232741, "pid": 76337, "tid": -914061504, "ts": 1716454225143910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143988, "dur": 0,
    "args": {
      "External id": 232751,
      "cbid": 317, "correlation": 232751
    }
  },
  {
    "ph": "f", "id": 232751, "pid": 76337, "tid": -914061504, "ts": 1716454225143988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143989, "dur": 0,
    "args": {
      "External id": 232752,
      "cbid": 203, "correlation": 232752
    }
  },
  {
    "ph": "f", "id": 232752, "pid": 76337, "tid": -914061504, "ts": 1716454225143989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225143990, "dur": 0,
    "args": {
      "External id": 232753,
      "cbid": 205, "correlation": 232753
    }
  },
  {
    "ph": "f", "id": 232753, "pid": 76337, "tid": -914061504, "ts": 1716454225143990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225144553, "dur": 39,
    "args": {
      "External id": 232757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232757,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232757, "pid": 5, "tid": 7, "ts": 1716454225144553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144004, "dur": 13,
    "args": {
      "External id": 232757,
      "cbid": 211, "correlation": 232757
    }
  },
  {
    "ph": "s", "id": 232757, "pid": 76337, "tid": -914061504, "ts": 1716454225144004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225144593, "dur": 14,
    "args": {
      "External id": 232759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232759,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232759, "pid": 5, "tid": 7, "ts": 1716454225144593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144019, "dur": 5,
    "args": {
      "External id": 232759,
      "cbid": 211, "correlation": 232759
    }
  },
  {
    "ph": "s", "id": 232759, "pid": 76337, "tid": -914061504, "ts": 1716454225144019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225144608, "dur": 3,
    "args": {
      "External id": 232761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232761,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232761, "pid": 5, "tid": 7, "ts": 1716454225144608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144029, "dur": 5,
    "args": {
      "External id": 232761,
      "cbid": 211, "correlation": 232761
    }
  },
  {
    "ph": "s", "id": 232761, "pid": 76337, "tid": -914061504, "ts": 1716454225144029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144037, "dur": 0,
    "args": {
      "External id": 232762,
      "cbid": 51, "correlation": 232762
    }
  },
  {
    "ph": "s", "id": 232762, "pid": 76337, "tid": -914061504, "ts": 1716454225144037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225144613, "dur": 683,
    "args": {
      "External id": 232763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232763,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232763, "pid": 5, "tid": 7, "ts": 1716454225144613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144038, "dur": 5,
    "args": {
      "External id": 232763,
      "cbid": 211, "correlation": 232763
    }
  },
  {
    "ph": "s", "id": 232763, "pid": 76337, "tid": -914061504, "ts": 1716454225144038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225145298, "dur": 58,
    "args": {
      "External id": 232768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232768,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232768, "pid": 5, "tid": 7, "ts": 1716454225145298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144066, "dur": 9,
    "args": {
      "External id": 232768,
      "cbid": 211, "correlation": 232768
    }
  },
  {
    "ph": "s", "id": 232768, "pid": 76337, "tid": -914061504, "ts": 1716454225144066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225145358, "dur": 50,
    "args": {
      "External id": 232776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232776,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232776, "pid": 5, "tid": 7, "ts": 1716454225145358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144099, "dur": 9,
    "args": {
      "External id": 232776,
      "cbid": 211, "correlation": 232776
    }
  },
  {
    "ph": "s", "id": 232776, "pid": 76337, "tid": -914061504, "ts": 1716454225144099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225145409, "dur": 35,
    "args": {
      "External id": 232784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232784,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232784, "pid": 5, "tid": 7, "ts": 1716454225145409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144131, "dur": 9,
    "args": {
      "External id": 232784,
      "cbid": 211, "correlation": 232784
    }
  },
  {
    "ph": "s", "id": 232784, "pid": 76337, "tid": -914061504, "ts": 1716454225144131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225145446, "dur": 51,
    "args": {
      "External id": 232804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232804,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 232804, "pid": 5, "tid": 7, "ts": 1716454225145446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144216, "dur": 12,
    "args": {
      "External id": 232804,
      "cbid": 211, "correlation": 232804
    }
  },
  {
    "ph": "s", "id": 232804, "pid": 76337, "tid": -914061504, "ts": 1716454225144216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225145498, "dur": 4,
    "args": {
      "External id": 232816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232816,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 232816, "pid": 5, "tid": 7, "ts": 1716454225145498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144238, "dur": 6,
    "args": {
      "External id": 232816,
      "cbid": 211, "correlation": 232816
    }
  },
  {
    "ph": "s", "id": 232816, "pid": 76337, "tid": -914061504, "ts": 1716454225144238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225145503, "dur": 55,
    "args": {
      "External id": 232819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232819, "pid": 5, "tid": 7, "ts": 1716454225145503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144256, "dur": 7,
    "args": {
      "External id": 232819,
      "cbid": 211, "correlation": 232819
    }
  },
  {
    "ph": "s", "id": 232819, "pid": 76337, "tid": -914061504, "ts": 1716454225144256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144313, "dur": 0,
    "args": {
      "External id": 232830,
      "cbid": 317, "correlation": 232830
    }
  },
  {
    "ph": "f", "id": 232830, "pid": 76337, "tid": -914061504, "ts": 1716454225144313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144314, "dur": 0,
    "args": {
      "External id": 232831,
      "cbid": 203, "correlation": 232831
    }
  },
  {
    "ph": "f", "id": 232831, "pid": 76337, "tid": -914061504, "ts": 1716454225144314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144315, "dur": 0,
    "args": {
      "External id": 232832,
      "cbid": 205, "correlation": 232832
    }
  },
  {
    "ph": "f", "id": 232832, "pid": 76337, "tid": -914061504, "ts": 1716454225144315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144348, "dur": 3,
    "args": {
      "External id": 232836,
      "cbid": 251, "correlation": 232836
    }
  },
  {
    "ph": "f", "id": 232836, "pid": 76337, "tid": -914061504, "ts": 1716454225144348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144352, "dur": 1,
    "args": {
      "External id": 232837,
      "cbid": 251, "correlation": 232837
    }
  },
  {
    "ph": "f", "id": 232837, "pid": 76337, "tid": -914061504, "ts": 1716454225144352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144355, "dur": 1,
    "args": {
      "External id": 232838,
      "cbid": 251, "correlation": 232838
    }
  },
  {
    "ph": "f", "id": 232838, "pid": 76337, "tid": -914061504, "ts": 1716454225144355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144357, "dur": 1,
    "args": {
      "External id": 232839,
      "cbid": 251, "correlation": 232839
    }
  },
  {
    "ph": "f", "id": 232839, "pid": 76337, "tid": -914061504, "ts": 1716454225144357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144359, "dur": 1,
    "args": {
      "External id": 232840,
      "cbid": 251, "correlation": 232840
    }
  },
  {
    "ph": "f", "id": 232840, "pid": 76337, "tid": -914061504, "ts": 1716454225144359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144361, "dur": 1,
    "args": {
      "External id": 232841,
      "cbid": 251, "correlation": 232841
    }
  },
  {
    "ph": "f", "id": 232841, "pid": 76337, "tid": -914061504, "ts": 1716454225144361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144363, "dur": 1,
    "args": {
      "External id": 232842,
      "cbid": 251, "correlation": 232842
    }
  },
  {
    "ph": "f", "id": 232842, "pid": 76337, "tid": -914061504, "ts": 1716454225144363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144365, "dur": 1,
    "args": {
      "External id": 232843,
      "cbid": 251, "correlation": 232843
    }
  },
  {
    "ph": "f", "id": 232843, "pid": 76337, "tid": -914061504, "ts": 1716454225144365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144368, "dur": 0,
    "args": {
      "External id": 232844,
      "cbid": 251, "correlation": 232844
    }
  },
  {
    "ph": "f", "id": 232844, "pid": 76337, "tid": -914061504, "ts": 1716454225144368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225145559, "dur": 113,
    "args": {
      "External id": 232845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232845,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 232845, "pid": 5, "tid": 7, "ts": 1716454225145559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144372, "dur": 14,
    "args": {
      "External id": 232845,
      "cbid": 211, "correlation": 232845
    }
  },
  {
    "ph": "s", "id": 232845, "pid": 76337, "tid": -914061504, "ts": 1716454225144372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225145674, "dur": 59,
    "args": {
      "External id": 232851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232851,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232851, "pid": 5, "tid": 7, "ts": 1716454225145674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144410, "dur": 9,
    "args": {
      "External id": 232851,
      "cbid": 211, "correlation": 232851
    }
  },
  {
    "ph": "s", "id": 232851, "pid": 76337, "tid": -914061504, "ts": 1716454225144410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225145734, "dur": 537,
    "args": {
      "External id": 232860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232860,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232860, "pid": 5, "tid": 7, "ts": 1716454225145734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144506, "dur": 16,
    "args": {
      "External id": 232860,
      "cbid": 211, "correlation": 232860
    }
  },
  {
    "ph": "s", "id": 232860, "pid": 76337, "tid": -914061504, "ts": 1716454225144506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225146272, "dur": 177,
    "args": {
      "External id": 232882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232882,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232882, "pid": 5, "tid": 7, "ts": 1716454225146272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144583, "dur": 14,
    "args": {
      "External id": 232882,
      "cbid": 211, "correlation": 232882
    }
  },
  {
    "ph": "s", "id": 232882, "pid": 76337, "tid": -914061504, "ts": 1716454225144583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144701, "dur": 2,
    "args": {
      "External id": 232893,
      "cbid": 251, "correlation": 232893
    }
  },
  {
    "ph": "f", "id": 232893, "pid": 76337, "tid": -914061504, "ts": 1716454225144701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225146450, "dur": 195,
    "args": {
      "External id": 232894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232894,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232894, "pid": 5, "tid": 7, "ts": 1716454225146450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144708, "dur": 14,
    "args": {
      "External id": 232894,
      "cbid": 211, "correlation": 232894
    }
  },
  {
    "ph": "s", "id": 232894, "pid": 76337, "tid": -914061504, "ts": 1716454225144708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144782, "dur": 1,
    "args": {
      "External id": 232905,
      "cbid": 251, "correlation": 232905
    }
  },
  {
    "ph": "f", "id": 232905, "pid": 76337, "tid": -914061504, "ts": 1716454225144782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225146646, "dur": 186,
    "args": {
      "External id": 232906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232906,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232906, "pid": 5, "tid": 7, "ts": 1716454225146646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144786, "dur": 12,
    "args": {
      "External id": 232906,
      "cbid": 211, "correlation": 232906
    }
  },
  {
    "ph": "s", "id": 232906, "pid": 76337, "tid": -914061504, "ts": 1716454225144786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144851, "dur": 1,
    "args": {
      "External id": 232917,
      "cbid": 251, "correlation": 232917
    }
  },
  {
    "ph": "f", "id": 232917, "pid": 76337, "tid": -914061504, "ts": 1716454225144851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225146833, "dur": 186,
    "args": {
      "External id": 232918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232918,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232918, "pid": 5, "tid": 7, "ts": 1716454225146833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144855, "dur": 11,
    "args": {
      "External id": 232918,
      "cbid": 211, "correlation": 232918
    }
  },
  {
    "ph": "s", "id": 232918, "pid": 76337, "tid": -914061504, "ts": 1716454225144855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225147020, "dur": 18236,
    "args": {
      "External id": 232939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232939,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 232939, "pid": 5, "tid": 7, "ts": 1716454225147020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225144967, "dur": 26,
    "args": {
      "External id": 232939,
      "cbid": 211, "correlation": 232939
    }
  },
  {
    "ph": "s", "id": 232939, "pid": 76337, "tid": -914061504, "ts": 1716454225144967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145095, "dur": 2,
    "args": {
      "External id": 232957,
      "cbid": 251, "correlation": 232957
    }
  },
  {
    "ph": "f", "id": 232957, "pid": 76337, "tid": -914061504, "ts": 1716454225145095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225165258, "dur": 198,
    "args": {
      "External id": 232959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232959,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 232959, "pid": 5, "tid": 7, "ts": 1716454225165258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145102, "dur": 13,
    "args": {
      "External id": 232959,
      "cbid": 211, "correlation": 232959
    }
  },
  {
    "ph": "s", "id": 232959, "pid": 76337, "tid": -914061504, "ts": 1716454225145102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225165457, "dur": 66,
    "args": {
      "External id": 232967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232967,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232967, "pid": 5, "tid": 7, "ts": 1716454225165457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145173, "dur": 13,
    "args": {
      "External id": 232967,
      "cbid": 211, "correlation": 232967
    }
  },
  {
    "ph": "s", "id": 232967, "pid": 76337, "tid": -914061504, "ts": 1716454225145173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225165525, "dur": 96,
    "args": {
      "External id": 232975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232975,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232975, "pid": 5, "tid": 7, "ts": 1716454225165525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145213, "dur": 10,
    "args": {
      "External id": 232975,
      "cbid": 211, "correlation": 232975
    }
  },
  {
    "ph": "s", "id": 232975, "pid": 76337, "tid": -914061504, "ts": 1716454225145213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225165622, "dur": 54,
    "args": {
      "External id": 232986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 232986,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 232986, "pid": 5, "tid": 7, "ts": 1716454225165622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145302, "dur": 14,
    "args": {
      "External id": 232986,
      "cbid": 211, "correlation": 232986
    }
  },
  {
    "ph": "s", "id": 232986, "pid": 76337, "tid": -914061504, "ts": 1716454225145302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225165677, "dur": 90,
    "args": {
      "External id": 233008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233008,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233008, "pid": 5, "tid": 7, "ts": 1716454225165677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145336, "dur": 7,
    "args": {
      "External id": 233008,
      "cbid": 211, "correlation": 233008
    }
  },
  {
    "ph": "s", "id": 233008, "pid": 76337, "tid": -914061504, "ts": 1716454225145336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145419, "dur": 1,
    "args": {
      "External id": 233019,
      "cbid": 251, "correlation": 233019
    }
  },
  {
    "ph": "f", "id": 233019, "pid": 76337, "tid": -914061504, "ts": 1716454225145419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225165768, "dur": 101,
    "args": {
      "External id": 233020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233020,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233020, "pid": 5, "tid": 7, "ts": 1716454225165768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145424, "dur": 13,
    "args": {
      "External id": 233020,
      "cbid": 211, "correlation": 233020
    }
  },
  {
    "ph": "s", "id": 233020, "pid": 76337, "tid": -914061504, "ts": 1716454225145424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145506, "dur": 2,
    "args": {
      "External id": 233031,
      "cbid": 251, "correlation": 233031
    }
  },
  {
    "ph": "f", "id": 233031, "pid": 76337, "tid": -914061504, "ts": 1716454225145506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145511, "dur": 0,
    "args": {
      "External id": 233032,
      "cbid": 251, "correlation": 233032
    }
  },
  {
    "ph": "f", "id": 233032, "pid": 76337, "tid": -914061504, "ts": 1716454225145511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225165871, "dur": 10,
    "args": {
      "External id": 233033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233033,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 233033, "pid": 5, "tid": 7, "ts": 1716454225165871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145513, "dur": 14,
    "args": {
      "External id": 233033,
      "cbid": 211, "correlation": 233033
    }
  },
  {
    "ph": "s", "id": 233033, "pid": 76337, "tid": -914061504, "ts": 1716454225145513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225165882, "dur": 5,
    "args": {
      "External id": 233035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233035,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 233035, "pid": 5, "tid": 7, "ts": 1716454225165882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145531, "dur": 7,
    "args": {
      "External id": 233035,
      "cbid": 211, "correlation": 233035
    }
  },
  {
    "ph": "s", "id": 233035, "pid": 76337, "tid": -914061504, "ts": 1716454225145531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145593, "dur": 1,
    "args": {
      "External id": 233046,
      "cbid": 251, "correlation": 233046
    }
  },
  {
    "ph": "f", "id": 233046, "pid": 76337, "tid": -914061504, "ts": 1716454225145593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145597, "dur": 0,
    "args": {
      "External id": 233047,
      "cbid": 251, "correlation": 233047
    }
  },
  {
    "ph": "f", "id": 233047, "pid": 76337, "tid": -914061504, "ts": 1716454225145597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225165888, "dur": 6,
    "args": {
      "External id": 233048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233048,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 233048, "pid": 5, "tid": 7, "ts": 1716454225165888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145598, "dur": 12,
    "args": {
      "External id": 233048,
      "cbid": 211, "correlation": 233048
    }
  },
  {
    "ph": "s", "id": 233048, "pid": 76337, "tid": -914061504, "ts": 1716454225145598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225165896, "dur": 3,
    "args": {
      "External id": 233050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233050,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 233050, "pid": 5, "tid": 7, "ts": 1716454225165896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145612, "dur": 6,
    "args": {
      "External id": 233050,
      "cbid": 211, "correlation": 233050
    }
  },
  {
    "ph": "s", "id": 233050, "pid": 76337, "tid": -914061504, "ts": 1716454225145612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225165901, "dur": 153,
    "args": {
      "External id": 233071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233071,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233071, "pid": 5, "tid": 7, "ts": 1716454225165901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145685, "dur": 12,
    "args": {
      "External id": 233071,
      "cbid": 211, "correlation": 233071
    }
  },
  {
    "ph": "s", "id": 233071, "pid": 76337, "tid": -914061504, "ts": 1716454225145685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145781, "dur": 2,
    "args": {
      "External id": 233089,
      "cbid": 251, "correlation": 233089
    }
  },
  {
    "ph": "f", "id": 233089, "pid": 76337, "tid": -914061504, "ts": 1716454225145781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225166054, "dur": 105,
    "args": {
      "External id": 233091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233091,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233091, "pid": 5, "tid": 7, "ts": 1716454225166054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145788, "dur": 15,
    "args": {
      "External id": 233091,
      "cbid": 211, "correlation": 233091
    }
  },
  {
    "ph": "s", "id": 233091, "pid": 76337, "tid": -914061504, "ts": 1716454225145788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225166161, "dur": 35,
    "args": {
      "External id": 233099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233099,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233099, "pid": 5, "tid": 7, "ts": 1716454225166161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145860, "dur": 13,
    "args": {
      "External id": 233099,
      "cbid": 211, "correlation": 233099
    }
  },
  {
    "ph": "s", "id": 233099, "pid": 76337, "tid": -914061504, "ts": 1716454225145860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225166197, "dur": 67,
    "args": {
      "External id": 233107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233107, "pid": 5, "tid": 7, "ts": 1716454225166197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145902, "dur": 9,
    "args": {
      "External id": 233107,
      "cbid": 211, "correlation": 233107
    }
  },
  {
    "ph": "s", "id": 233107, "pid": 76337, "tid": -914061504, "ts": 1716454225145902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225166265, "dur": 90,
    "args": {
      "External id": 233129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233129,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233129, "pid": 5, "tid": 7, "ts": 1716454225166265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225145952, "dur": 10,
    "args": {
      "External id": 233129,
      "cbid": 211, "correlation": 233129
    }
  },
  {
    "ph": "s", "id": 233129, "pid": 76337, "tid": -914061504, "ts": 1716454225145952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146051, "dur": 1,
    "args": {
      "External id": 233145,
      "cbid": 251, "correlation": 233145
    }
  },
  {
    "ph": "f", "id": 233145, "pid": 76337, "tid": -914061504, "ts": 1716454225146051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225166357, "dur": 567,
    "args": {
      "External id": 233147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233147,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233147, "pid": 5, "tid": 7, "ts": 1716454225166357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146057, "dur": 14,
    "args": {
      "External id": 233147,
      "cbid": 211, "correlation": 233147
    }
  },
  {
    "ph": "s", "id": 233147, "pid": 76337, "tid": -914061504, "ts": 1716454225146057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225166925, "dur": 239,
    "args": {
      "External id": 233155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233155, "pid": 5, "tid": 7, "ts": 1716454225166925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146135, "dur": 14,
    "args": {
      "External id": 233155,
      "cbid": 211, "correlation": 233155
    }
  },
  {
    "ph": "s", "id": 233155, "pid": 76337, "tid": -914061504, "ts": 1716454225146135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225167165, "dur": 249,
    "args": {
      "External id": 233163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233163,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233163, "pid": 5, "tid": 7, "ts": 1716454225167165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146170, "dur": 9,
    "args": {
      "External id": 233163,
      "cbid": 211, "correlation": 233163
    }
  },
  {
    "ph": "s", "id": 233163, "pid": 76337, "tid": -914061504, "ts": 1716454225146170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146254, "dur": 2,
    "args": {
      "External id": 233179,
      "cbid": 251, "correlation": 233179
    }
  },
  {
    "ph": "f", "id": 233179, "pid": 76337, "tid": -914061504, "ts": 1716454225146254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146260, "dur": 0,
    "args": {
      "External id": 233181,
      "cbid": 251, "correlation": 233181
    }
  },
  {
    "ph": "f", "id": 233181, "pid": 76337, "tid": -914061504, "ts": 1716454225146260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225167416, "dur": 356,
    "args": {
      "External id": 233182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233182,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 233182, "pid": 5, "tid": 7, "ts": 1716454225167416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146265, "dur": 13,
    "args": {
      "External id": 233182,
      "cbid": 211, "correlation": 233182
    }
  },
  {
    "ph": "s", "id": 233182, "pid": 76337, "tid": -914061504, "ts": 1716454225146265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225167773, "dur": 49,
    "args": {
      "External id": 233190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233190,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233190, "pid": 5, "tid": 7, "ts": 1716454225167773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146309, "dur": 10,
    "args": {
      "External id": 233190,
      "cbid": 211, "correlation": 233190
    }
  },
  {
    "ph": "s", "id": 233190, "pid": 76337, "tid": -914061504, "ts": 1716454225146309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225167824, "dur": 153,
    "args": {
      "External id": 233201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233201,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233201, "pid": 5, "tid": 7, "ts": 1716454225167824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146379, "dur": 12,
    "args": {
      "External id": 233201,
      "cbid": 211, "correlation": 233201
    }
  },
  {
    "ph": "s", "id": 233201, "pid": 76337, "tid": -914061504, "ts": 1716454225146379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146445, "dur": 0,
    "args": {
      "External id": 233213,
      "cbid": 317, "correlation": 233213
    }
  },
  {
    "ph": "f", "id": 233213, "pid": 76337, "tid": -914061504, "ts": 1716454225146445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146446, "dur": 0,
    "args": {
      "External id": 233214,
      "cbid": 203, "correlation": 233214
    }
  },
  {
    "ph": "f", "id": 233214, "pid": 76337, "tid": -914061504, "ts": 1716454225146446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146447, "dur": 0,
    "args": {
      "External id": 233215,
      "cbid": 205, "correlation": 233215
    }
  },
  {
    "ph": "f", "id": 233215, "pid": 76337, "tid": -914061504, "ts": 1716454225146447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146472, "dur": 1,
    "args": {
      "External id": 233219,
      "cbid": 251, "correlation": 233219
    }
  },
  {
    "ph": "f", "id": 233219, "pid": 76337, "tid": -914061504, "ts": 1716454225146472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146474, "dur": 0,
    "args": {
      "External id": 233220,
      "cbid": 251, "correlation": 233220
    }
  },
  {
    "ph": "f", "id": 233220, "pid": 76337, "tid": -914061504, "ts": 1716454225146474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146474, "dur": 0,
    "args": {
      "External id": 233221,
      "cbid": 251, "correlation": 233221
    }
  },
  {
    "ph": "f", "id": 233221, "pid": 76337, "tid": -914061504, "ts": 1716454225146474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146475, "dur": 0,
    "args": {
      "External id": 233222,
      "cbid": 251, "correlation": 233222
    }
  },
  {
    "ph": "f", "id": 233222, "pid": 76337, "tid": -914061504, "ts": 1716454225146475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146476, "dur": 0,
    "args": {
      "External id": 233223,
      "cbid": 251, "correlation": 233223
    }
  },
  {
    "ph": "f", "id": 233223, "pid": 76337, "tid": -914061504, "ts": 1716454225146476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146477, "dur": 0,
    "args": {
      "External id": 233224,
      "cbid": 251, "correlation": 233224
    }
  },
  {
    "ph": "f", "id": 233224, "pid": 76337, "tid": -914061504, "ts": 1716454225146477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146478, "dur": 0,
    "args": {
      "External id": 233225,
      "cbid": 251, "correlation": 233225
    }
  },
  {
    "ph": "f", "id": 233225, "pid": 76337, "tid": -914061504, "ts": 1716454225146478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146479, "dur": 0,
    "args": {
      "External id": 233226,
      "cbid": 251, "correlation": 233226
    }
  },
  {
    "ph": "f", "id": 233226, "pid": 76337, "tid": -914061504, "ts": 1716454225146479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146480, "dur": 0,
    "args": {
      "External id": 233227,
      "cbid": 251, "correlation": 233227
    }
  },
  {
    "ph": "f", "id": 233227, "pid": 76337, "tid": -914061504, "ts": 1716454225146480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225167978, "dur": 111,
    "args": {
      "External id": 233228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233228,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233228, "pid": 5, "tid": 7, "ts": 1716454225167978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146482, "dur": 12,
    "args": {
      "External id": 233228,
      "cbid": 211, "correlation": 233228
    }
  },
  {
    "ph": "s", "id": 233228, "pid": 76337, "tid": -914061504, "ts": 1716454225146482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225168091, "dur": 59,
    "args": {
      "External id": 233234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233234, "pid": 5, "tid": 7, "ts": 1716454225168091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146517, "dur": 9,
    "args": {
      "External id": 233234,
      "cbid": 211, "correlation": 233234
    }
  },
  {
    "ph": "s", "id": 233234, "pid": 76337, "tid": -914061504, "ts": 1716454225146517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225168151, "dur": 50,
    "args": {
      "External id": 233242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233242,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233242, "pid": 5, "tid": 7, "ts": 1716454225168151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146550, "dur": 9,
    "args": {
      "External id": 233242,
      "cbid": 211, "correlation": 233242
    }
  },
  {
    "ph": "s", "id": 233242, "pid": 76337, "tid": -914061504, "ts": 1716454225146550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225168203, "dur": 51,
    "args": {
      "External id": 233262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233262,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 233262, "pid": 5, "tid": 7, "ts": 1716454225168203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146623, "dur": 11,
    "args": {
      "External id": 233262,
      "cbid": 211, "correlation": 233262
    }
  },
  {
    "ph": "s", "id": 233262, "pid": 76337, "tid": -914061504, "ts": 1716454225146623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225168255, "dur": 4,
    "args": {
      "External id": 233274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233274,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233274, "pid": 5, "tid": 7, "ts": 1716454225168255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146644, "dur": 7,
    "args": {
      "External id": 233274,
      "cbid": 211, "correlation": 233274
    }
  },
  {
    "ph": "s", "id": 233274, "pid": 76337, "tid": -914061504, "ts": 1716454225146644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225168261, "dur": 56,
    "args": {
      "External id": 233277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233277, "pid": 5, "tid": 7, "ts": 1716454225168261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146663, "dur": 7,
    "args": {
      "External id": 233277,
      "cbid": 211, "correlation": 233277
    }
  },
  {
    "ph": "s", "id": 233277, "pid": 76337, "tid": -914061504, "ts": 1716454225146663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225168318, "dur": 37,
    "args": {
      "External id": 233286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233286,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233286, "pid": 5, "tid": 7, "ts": 1716454225168318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146702, "dur": 10,
    "args": {
      "External id": 233286,
      "cbid": 211, "correlation": 233286
    }
  },
  {
    "ph": "s", "id": 233286, "pid": 76337, "tid": -914061504, "ts": 1716454225146702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146754, "dur": 0,
    "args": {
      "External id": 233296,
      "cbid": 317, "correlation": 233296
    }
  },
  {
    "ph": "f", "id": 233296, "pid": 76337, "tid": -914061504, "ts": 1716454225146754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146755, "dur": 0,
    "args": {
      "External id": 233297,
      "cbid": 203, "correlation": 233297
    }
  },
  {
    "ph": "f", "id": 233297, "pid": 76337, "tid": -914061504, "ts": 1716454225146755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146756, "dur": 0,
    "args": {
      "External id": 233298,
      "cbid": 205, "correlation": 233298
    }
  },
  {
    "ph": "f", "id": 233298, "pid": 76337, "tid": -914061504, "ts": 1716454225146756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225168357, "dur": 41,
    "args": {
      "External id": 233302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233302, "pid": 5, "tid": 7, "ts": 1716454225168357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146770, "dur": 11,
    "args": {
      "External id": 233302,
      "cbid": 211, "correlation": 233302
    }
  },
  {
    "ph": "s", "id": 233302, "pid": 76337, "tid": -914061504, "ts": 1716454225146770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225168399, "dur": 14,
    "args": {
      "External id": 233304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233304, "pid": 5, "tid": 7, "ts": 1716454225168399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146784, "dur": 5,
    "args": {
      "External id": 233304,
      "cbid": 211, "correlation": 233304
    }
  },
  {
    "ph": "s", "id": 233304, "pid": 76337, "tid": -914061504, "ts": 1716454225146784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225168415, "dur": 4,
    "args": {
      "External id": 233306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233306, "pid": 5, "tid": 7, "ts": 1716454225168415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146793, "dur": 7,
    "args": {
      "External id": 233306,
      "cbid": 211, "correlation": 233306
    }
  },
  {
    "ph": "s", "id": 233306, "pid": 76337, "tid": -914061504, "ts": 1716454225146793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146803, "dur": 0,
    "args": {
      "External id": 233307,
      "cbid": 51, "correlation": 233307
    }
  },
  {
    "ph": "s", "id": 233307, "pid": 76337, "tid": -914061504, "ts": 1716454225146803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225168419, "dur": 690,
    "args": {
      "External id": 233308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233308,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233308, "pid": 5, "tid": 7, "ts": 1716454225168419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146804, "dur": 5,
    "args": {
      "External id": 233308,
      "cbid": 211, "correlation": 233308
    }
  },
  {
    "ph": "s", "id": 233308, "pid": 76337, "tid": -914061504, "ts": 1716454225146804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225169111, "dur": 59,
    "args": {
      "External id": 233313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233313, "pid": 5, "tid": 7, "ts": 1716454225169111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146831, "dur": 8,
    "args": {
      "External id": 233313,
      "cbid": 211, "correlation": 233313
    }
  },
  {
    "ph": "s", "id": 233313, "pid": 76337, "tid": -914061504, "ts": 1716454225146831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225169171, "dur": 3,
    "args": {
      "External id": 233321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233321,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233321, "pid": 5, "tid": 7, "ts": 1716454225169171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146874, "dur": 9,
    "args": {
      "External id": 233321,
      "cbid": 211, "correlation": 233321
    }
  },
  {
    "ph": "s", "id": 233321, "pid": 76337, "tid": -914061504, "ts": 1716454225146874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146939, "dur": 1,
    "args": {
      "External id": 233337,
      "cbid": 251, "correlation": 233337
    }
  },
  {
    "ph": "f", "id": 233337, "pid": 76337, "tid": -914061504, "ts": 1716454225146939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146944, "dur": 0,
    "args": {
      "External id": 233339,
      "cbid": 251, "correlation": 233339
    }
  },
  {
    "ph": "f", "id": 233339, "pid": 76337, "tid": -914061504, "ts": 1716454225146944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225169176, "dur": 11,
    "args": {
      "External id": 233340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233340,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 233340, "pid": 5, "tid": 7, "ts": 1716454225169176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146945, "dur": 12,
    "args": {
      "External id": 233340,
      "cbid": 211, "correlation": 233340
    }
  },
  {
    "ph": "s", "id": 233340, "pid": 76337, "tid": -914061504, "ts": 1716454225146945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225169188, "dur": 5,
    "args": {
      "External id": 233342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233342,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 233342, "pid": 5, "tid": 7, "ts": 1716454225169188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225146960, "dur": 6,
    "args": {
      "External id": 233342,
      "cbid": 211, "correlation": 233342
    }
  },
  {
    "ph": "s", "id": 233342, "pid": 76337, "tid": -914061504, "ts": 1716454225146960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225169195, "dur": 52,
    "args": {
      "External id": 233352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233352, "pid": 5, "tid": 7, "ts": 1716454225169195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147024, "dur": 13,
    "args": {
      "External id": 233352,
      "cbid": 211, "correlation": 233352
    }
  },
  {
    "ph": "s", "id": 233352, "pid": 76337, "tid": -914061504, "ts": 1716454225147024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225169248, "dur": 51,
    "args": {
      "External id": 233372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233372,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 233372, "pid": 5, "tid": 7, "ts": 1716454225169248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147091, "dur": 11,
    "args": {
      "External id": 233372,
      "cbid": 211, "correlation": 233372
    }
  },
  {
    "ph": "s", "id": 233372, "pid": 76337, "tid": -914061504, "ts": 1716454225147091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225169300, "dur": 4,
    "args": {
      "External id": 233384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233384,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233384, "pid": 5, "tid": 7, "ts": 1716454225169300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147112, "dur": 6,
    "args": {
      "External id": 233384,
      "cbid": 211, "correlation": 233384
    }
  },
  {
    "ph": "s", "id": 233384, "pid": 76337, "tid": -914061504, "ts": 1716454225147112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225169305, "dur": 56,
    "args": {
      "External id": 233387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233387, "pid": 5, "tid": 7, "ts": 1716454225169305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147130, "dur": 6,
    "args": {
      "External id": 233387,
      "cbid": 211, "correlation": 233387
    }
  },
  {
    "ph": "s", "id": 233387, "pid": 76337, "tid": -914061504, "ts": 1716454225147130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225169363, "dur": 36,
    "args": {
      "External id": 233396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233396,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233396, "pid": 5, "tid": 7, "ts": 1716454225169363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147171, "dur": 9,
    "args": {
      "External id": 233396,
      "cbid": 211, "correlation": 233396
    }
  },
  {
    "ph": "s", "id": 233396, "pid": 76337, "tid": -914061504, "ts": 1716454225147171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147232, "dur": 0,
    "args": {
      "External id": 233406,
      "cbid": 317, "correlation": 233406
    }
  },
  {
    "ph": "f", "id": 233406, "pid": 76337, "tid": -914061504, "ts": 1716454225147232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147232, "dur": 0,
    "args": {
      "External id": 233407,
      "cbid": 203, "correlation": 233407
    }
  },
  {
    "ph": "f", "id": 233407, "pid": 76337, "tid": -914061504, "ts": 1716454225147232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147233, "dur": 0,
    "args": {
      "External id": 233408,
      "cbid": 205, "correlation": 233408
    }
  },
  {
    "ph": "f", "id": 233408, "pid": 76337, "tid": -914061504, "ts": 1716454225147233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225169400, "dur": 40,
    "args": {
      "External id": 233412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233412, "pid": 5, "tid": 7, "ts": 1716454225169400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147248, "dur": 12,
    "args": {
      "External id": 233412,
      "cbid": 211, "correlation": 233412
    }
  },
  {
    "ph": "s", "id": 233412, "pid": 76337, "tid": -914061504, "ts": 1716454225147248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225169441, "dur": 14,
    "args": {
      "External id": 233414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233414,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233414, "pid": 5, "tid": 7, "ts": 1716454225169441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147263, "dur": 5,
    "args": {
      "External id": 233414,
      "cbid": 211, "correlation": 233414
    }
  },
  {
    "ph": "s", "id": 233414, "pid": 76337, "tid": -914061504, "ts": 1716454225147263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225169457, "dur": 3,
    "args": {
      "External id": 233416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233416, "pid": 5, "tid": 7, "ts": 1716454225169457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147272, "dur": 5,
    "args": {
      "External id": 233416,
      "cbid": 211, "correlation": 233416
    }
  },
  {
    "ph": "s", "id": 233416, "pid": 76337, "tid": -914061504, "ts": 1716454225147272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147280, "dur": 0,
    "args": {
      "External id": 233417,
      "cbid": 51, "correlation": 233417
    }
  },
  {
    "ph": "s", "id": 233417, "pid": 76337, "tid": -914061504, "ts": 1716454225147280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225169461, "dur": 683,
    "args": {
      "External id": 233418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233418,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233418, "pid": 5, "tid": 7, "ts": 1716454225169461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147281, "dur": 5,
    "args": {
      "External id": 233418,
      "cbid": 211, "correlation": 233418
    }
  },
  {
    "ph": "s", "id": 233418, "pid": 76337, "tid": -914061504, "ts": 1716454225147281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225170145, "dur": 58,
    "args": {
      "External id": 233423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233423, "pid": 5, "tid": 7, "ts": 1716454225170145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147309, "dur": 8,
    "args": {
      "External id": 233423,
      "cbid": 211, "correlation": 233423
    }
  },
  {
    "ph": "s", "id": 233423, "pid": 76337, "tid": -914061504, "ts": 1716454225147309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225170204, "dur": 50,
    "args": {
      "External id": 233431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233431,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233431, "pid": 5, "tid": 7, "ts": 1716454225170204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147341, "dur": 9,
    "args": {
      "External id": 233431,
      "cbid": 211, "correlation": 233431
    }
  },
  {
    "ph": "s", "id": 233431, "pid": 76337, "tid": -914061504, "ts": 1716454225147341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225170256, "dur": 35,
    "args": {
      "External id": 233439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233439,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233439, "pid": 5, "tid": 7, "ts": 1716454225170256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147370, "dur": 8,
    "args": {
      "External id": 233439,
      "cbid": 211, "correlation": 233439
    }
  },
  {
    "ph": "s", "id": 233439, "pid": 76337, "tid": -914061504, "ts": 1716454225147370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225170292, "dur": 50,
    "args": {
      "External id": 233459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233459,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 233459, "pid": 5, "tid": 7, "ts": 1716454225170292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147447, "dur": 12,
    "args": {
      "External id": 233459,
      "cbid": 211, "correlation": 233459
    }
  },
  {
    "ph": "s", "id": 233459, "pid": 76337, "tid": -914061504, "ts": 1716454225147447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225170344, "dur": 4,
    "args": {
      "External id": 233471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233471,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233471, "pid": 5, "tid": 7, "ts": 1716454225170344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147469, "dur": 6,
    "args": {
      "External id": 233471,
      "cbid": 211, "correlation": 233471
    }
  },
  {
    "ph": "s", "id": 233471, "pid": 76337, "tid": -914061504, "ts": 1716454225147469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225170349, "dur": 54,
    "args": {
      "External id": 233474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233474, "pid": 5, "tid": 7, "ts": 1716454225170349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147486, "dur": 6,
    "args": {
      "External id": 233474,
      "cbid": 211, "correlation": 233474
    }
  },
  {
    "ph": "s", "id": 233474, "pid": 76337, "tid": -914061504, "ts": 1716454225147486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147543, "dur": 0,
    "args": {
      "External id": 233485,
      "cbid": 317, "correlation": 233485
    }
  },
  {
    "ph": "f", "id": 233485, "pid": 76337, "tid": -914061504, "ts": 1716454225147543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147544, "dur": 0,
    "args": {
      "External id": 233486,
      "cbid": 203, "correlation": 233486
    }
  },
  {
    "ph": "f", "id": 233486, "pid": 76337, "tid": -914061504, "ts": 1716454225147544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147545, "dur": 0,
    "args": {
      "External id": 233487,
      "cbid": 205, "correlation": 233487
    }
  },
  {
    "ph": "f", "id": 233487, "pid": 76337, "tid": -914061504, "ts": 1716454225147545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147567, "dur": 1,
    "args": {
      "External id": 233491,
      "cbid": 251, "correlation": 233491
    }
  },
  {
    "ph": "f", "id": 233491, "pid": 76337, "tid": -914061504, "ts": 1716454225147567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147569, "dur": 0,
    "args": {
      "External id": 233492,
      "cbid": 251, "correlation": 233492
    }
  },
  {
    "ph": "f", "id": 233492, "pid": 76337, "tid": -914061504, "ts": 1716454225147569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147570, "dur": 0,
    "args": {
      "External id": 233493,
      "cbid": 251, "correlation": 233493
    }
  },
  {
    "ph": "f", "id": 233493, "pid": 76337, "tid": -914061504, "ts": 1716454225147570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147570, "dur": 0,
    "args": {
      "External id": 233494,
      "cbid": 251, "correlation": 233494
    }
  },
  {
    "ph": "f", "id": 233494, "pid": 76337, "tid": -914061504, "ts": 1716454225147570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147571, "dur": 0,
    "args": {
      "External id": 233495,
      "cbid": 251, "correlation": 233495
    }
  },
  {
    "ph": "f", "id": 233495, "pid": 76337, "tid": -914061504, "ts": 1716454225147571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147572, "dur": 0,
    "args": {
      "External id": 233496,
      "cbid": 251, "correlation": 233496
    }
  },
  {
    "ph": "f", "id": 233496, "pid": 76337, "tid": -914061504, "ts": 1716454225147572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147573, "dur": 0,
    "args": {
      "External id": 233497,
      "cbid": 251, "correlation": 233497
    }
  },
  {
    "ph": "f", "id": 233497, "pid": 76337, "tid": -914061504, "ts": 1716454225147573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147573, "dur": 0,
    "args": {
      "External id": 233498,
      "cbid": 251, "correlation": 233498
    }
  },
  {
    "ph": "f", "id": 233498, "pid": 76337, "tid": -914061504, "ts": 1716454225147573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147574, "dur": 0,
    "args": {
      "External id": 233499,
      "cbid": 251, "correlation": 233499
    }
  },
  {
    "ph": "f", "id": 233499, "pid": 76337, "tid": -914061504, "ts": 1716454225147574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225170404, "dur": 109,
    "args": {
      "External id": 233500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233500,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233500, "pid": 5, "tid": 7, "ts": 1716454225170404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147577, "dur": 12,
    "args": {
      "External id": 233500,
      "cbid": 211, "correlation": 233500
    }
  },
  {
    "ph": "s", "id": 233500, "pid": 76337, "tid": -914061504, "ts": 1716454225147577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225170515, "dur": 58,
    "args": {
      "External id": 233506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233506,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233506, "pid": 5, "tid": 7, "ts": 1716454225170515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147611, "dur": 8,
    "args": {
      "External id": 233506,
      "cbid": 211, "correlation": 233506
    }
  },
  {
    "ph": "s", "id": 233506, "pid": 76337, "tid": -914061504, "ts": 1716454225147611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225170574, "dur": 536,
    "args": {
      "External id": 233515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233515,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233515, "pid": 5, "tid": 7, "ts": 1716454225170574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147694, "dur": 14,
    "args": {
      "External id": 233515,
      "cbid": 211, "correlation": 233515
    }
  },
  {
    "ph": "s", "id": 233515, "pid": 76337, "tid": -914061504, "ts": 1716454225147694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225171112, "dur": 176,
    "args": {
      "External id": 233537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233537,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233537, "pid": 5, "tid": 7, "ts": 1716454225171112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147751, "dur": 10,
    "args": {
      "External id": 233537,
      "cbid": 211, "correlation": 233537
    }
  },
  {
    "ph": "s", "id": 233537, "pid": 76337, "tid": -914061504, "ts": 1716454225147751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147835, "dur": 1,
    "args": {
      "External id": 233548,
      "cbid": 251, "correlation": 233548
    }
  },
  {
    "ph": "f", "id": 233548, "pid": 76337, "tid": -914061504, "ts": 1716454225147835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225171289, "dur": 190,
    "args": {
      "External id": 233549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233549,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233549, "pid": 5, "tid": 7, "ts": 1716454225171289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147840, "dur": 14,
    "args": {
      "External id": 233549,
      "cbid": 211, "correlation": 233549
    }
  },
  {
    "ph": "s", "id": 233549, "pid": 76337, "tid": -914061504, "ts": 1716454225147840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147909, "dur": 1,
    "args": {
      "External id": 233560,
      "cbid": 251, "correlation": 233560
    }
  },
  {
    "ph": "f", "id": 233560, "pid": 76337, "tid": -914061504, "ts": 1716454225147909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225171480, "dur": 184,
    "args": {
      "External id": 233561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233561,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233561, "pid": 5, "tid": 7, "ts": 1716454225171480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147913, "dur": 12,
    "args": {
      "External id": 233561,
      "cbid": 211, "correlation": 233561
    }
  },
  {
    "ph": "s", "id": 233561, "pid": 76337, "tid": -914061504, "ts": 1716454225147913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147983, "dur": 1,
    "args": {
      "External id": 233572,
      "cbid": 251, "correlation": 233572
    }
  },
  {
    "ph": "f", "id": 233572, "pid": 76337, "tid": -914061504, "ts": 1716454225147983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225171666, "dur": 186,
    "args": {
      "External id": 233573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233573,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233573, "pid": 5, "tid": 7, "ts": 1716454225171666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225147988, "dur": 13,
    "args": {
      "External id": 233573,
      "cbid": 211, "correlation": 233573
    }
  },
  {
    "ph": "s", "id": 233573, "pid": 76337, "tid": -914061504, "ts": 1716454225147988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225171853, "dur": 18206,
    "args": {
      "External id": 233594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233594,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233594, "pid": 5, "tid": 7, "ts": 1716454225171853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148070, "dur": 12,
    "args": {
      "External id": 233594,
      "cbid": 211, "correlation": 233594
    }
  },
  {
    "ph": "s", "id": 233594, "pid": 76337, "tid": -914061504, "ts": 1716454225148070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148167, "dur": 1,
    "args": {
      "External id": 233612,
      "cbid": 251, "correlation": 233612
    }
  },
  {
    "ph": "f", "id": 233612, "pid": 76337, "tid": -914061504, "ts": 1716454225148167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225190060, "dur": 197,
    "args": {
      "External id": 233614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233614,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233614, "pid": 5, "tid": 7, "ts": 1716454225190060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148173, "dur": 13,
    "args": {
      "External id": 233614,
      "cbid": 211, "correlation": 233614
    }
  },
  {
    "ph": "s", "id": 233614, "pid": 76337, "tid": -914061504, "ts": 1716454225148173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225190259, "dur": 66,
    "args": {
      "External id": 233622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233622,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233622, "pid": 5, "tid": 7, "ts": 1716454225190259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148242, "dur": 13,
    "args": {
      "External id": 233622,
      "cbid": 211, "correlation": 233622
    }
  },
  {
    "ph": "s", "id": 233622, "pid": 76337, "tid": -914061504, "ts": 1716454225148242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225190327, "dur": 97,
    "args": {
      "External id": 233630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233630,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233630, "pid": 5, "tid": 7, "ts": 1716454225190327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148282, "dur": 8,
    "args": {
      "External id": 233630,
      "cbid": 211, "correlation": 233630
    }
  },
  {
    "ph": "s", "id": 233630, "pid": 76337, "tid": -914061504, "ts": 1716454225148282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225190424, "dur": 53,
    "args": {
      "External id": 233641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233641,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233641, "pid": 5, "tid": 7, "ts": 1716454225190424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148353, "dur": 12,
    "args": {
      "External id": 233641,
      "cbid": 211, "correlation": 233641
    }
  },
  {
    "ph": "s", "id": 233641, "pid": 76337, "tid": -914061504, "ts": 1716454225148353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225190479, "dur": 90,
    "args": {
      "External id": 233663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233663,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233663, "pid": 5, "tid": 7, "ts": 1716454225190479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148384, "dur": 8,
    "args": {
      "External id": 233663,
      "cbid": 211, "correlation": 233663
    }
  },
  {
    "ph": "s", "id": 233663, "pid": 76337, "tid": -914061504, "ts": 1716454225148384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148466, "dur": 1,
    "args": {
      "External id": 233674,
      "cbid": 251, "correlation": 233674
    }
  },
  {
    "ph": "f", "id": 233674, "pid": 76337, "tid": -914061504, "ts": 1716454225148466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225190571, "dur": 103,
    "args": {
      "External id": 233675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233675,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233675, "pid": 5, "tid": 7, "ts": 1716454225190571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148471, "dur": 13,
    "args": {
      "External id": 233675,
      "cbid": 211, "correlation": 233675
    }
  },
  {
    "ph": "s", "id": 233675, "pid": 76337, "tid": -914061504, "ts": 1716454225148471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148543, "dur": 1,
    "args": {
      "External id": 233686,
      "cbid": 251, "correlation": 233686
    }
  },
  {
    "ph": "f", "id": 233686, "pid": 76337, "tid": -914061504, "ts": 1716454225148543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148547, "dur": 0,
    "args": {
      "External id": 233687,
      "cbid": 251, "correlation": 233687
    }
  },
  {
    "ph": "f", "id": 233687, "pid": 76337, "tid": -914061504, "ts": 1716454225148547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225190675, "dur": 10,
    "args": {
      "External id": 233688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233688,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 233688, "pid": 5, "tid": 7, "ts": 1716454225190675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148549, "dur": 12,
    "args": {
      "External id": 233688,
      "cbid": 211, "correlation": 233688
    }
  },
  {
    "ph": "s", "id": 233688, "pid": 76337, "tid": -914061504, "ts": 1716454225148549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225190686, "dur": 5,
    "args": {
      "External id": 233690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233690,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 233690, "pid": 5, "tid": 7, "ts": 1716454225190686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148563, "dur": 6,
    "args": {
      "External id": 233690,
      "cbid": 211, "correlation": 233690
    }
  },
  {
    "ph": "s", "id": 233690, "pid": 76337, "tid": -914061504, "ts": 1716454225148563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148623, "dur": 1,
    "args": {
      "External id": 233701,
      "cbid": 251, "correlation": 233701
    }
  },
  {
    "ph": "f", "id": 233701, "pid": 76337, "tid": -914061504, "ts": 1716454225148623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148626, "dur": 0,
    "args": {
      "External id": 233702,
      "cbid": 251, "correlation": 233702
    }
  },
  {
    "ph": "f", "id": 233702, "pid": 76337, "tid": -914061504, "ts": 1716454225148626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225190693, "dur": 6,
    "args": {
      "External id": 233703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233703,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 233703, "pid": 5, "tid": 7, "ts": 1716454225190693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148627, "dur": 12,
    "args": {
      "External id": 233703,
      "cbid": 211, "correlation": 233703
    }
  },
  {
    "ph": "s", "id": 233703, "pid": 76337, "tid": -914061504, "ts": 1716454225148627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225190700, "dur": 3,
    "args": {
      "External id": 233705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233705,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 233705, "pid": 5, "tid": 7, "ts": 1716454225190700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148642, "dur": 6,
    "args": {
      "External id": 233705,
      "cbid": 211, "correlation": 233705
    }
  },
  {
    "ph": "s", "id": 233705, "pid": 76337, "tid": -914061504, "ts": 1716454225148642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225190705, "dur": 153,
    "args": {
      "External id": 233726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233726,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233726, "pid": 5, "tid": 7, "ts": 1716454225190705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148715, "dur": 12,
    "args": {
      "External id": 233726,
      "cbid": 211, "correlation": 233726
    }
  },
  {
    "ph": "s", "id": 233726, "pid": 76337, "tid": -914061504, "ts": 1716454225148715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148813, "dur": 1,
    "args": {
      "External id": 233744,
      "cbid": 251, "correlation": 233744
    }
  },
  {
    "ph": "f", "id": 233744, "pid": 76337, "tid": -914061504, "ts": 1716454225148813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225190860, "dur": 105,
    "args": {
      "External id": 233746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233746,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233746, "pid": 5, "tid": 7, "ts": 1716454225190860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148818, "dur": 13,
    "args": {
      "External id": 233746,
      "cbid": 211, "correlation": 233746
    }
  },
  {
    "ph": "s", "id": 233746, "pid": 76337, "tid": -914061504, "ts": 1716454225148818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225190965, "dur": 35,
    "args": {
      "External id": 233754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233754,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233754, "pid": 5, "tid": 7, "ts": 1716454225190965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148886, "dur": 12,
    "args": {
      "External id": 233754,
      "cbid": 211, "correlation": 233754
    }
  },
  {
    "ph": "s", "id": 233754, "pid": 76337, "tid": -914061504, "ts": 1716454225148886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225191002, "dur": 68,
    "args": {
      "External id": 233762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233762,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233762, "pid": 5, "tid": 7, "ts": 1716454225191002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148927, "dur": 9,
    "args": {
      "External id": 233762,
      "cbid": 211, "correlation": 233762
    }
  },
  {
    "ph": "s", "id": 233762, "pid": 76337, "tid": -914061504, "ts": 1716454225148927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225191071, "dur": 90,
    "args": {
      "External id": 233784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233784,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233784, "pid": 5, "tid": 7, "ts": 1716454225191071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225148985, "dur": 11,
    "args": {
      "External id": 233784,
      "cbid": 211, "correlation": 233784
    }
  },
  {
    "ph": "s", "id": 233784, "pid": 76337, "tid": -914061504, "ts": 1716454225148985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149072, "dur": 1,
    "args": {
      "External id": 233800,
      "cbid": 251, "correlation": 233800
    }
  },
  {
    "ph": "f", "id": 233800, "pid": 76337, "tid": -914061504, "ts": 1716454225149072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225191162, "dur": 562,
    "args": {
      "External id": 233802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233802,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233802, "pid": 5, "tid": 7, "ts": 1716454225191162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149078, "dur": 13,
    "args": {
      "External id": 233802,
      "cbid": 211, "correlation": 233802
    }
  },
  {
    "ph": "s", "id": 233802, "pid": 76337, "tid": -914061504, "ts": 1716454225149078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225191725, "dur": 241,
    "args": {
      "External id": 233810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233810, "pid": 5, "tid": 7, "ts": 1716454225191725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149143, "dur": 12,
    "args": {
      "External id": 233810,
      "cbid": 211, "correlation": 233810
    }
  },
  {
    "ph": "s", "id": 233810, "pid": 76337, "tid": -914061504, "ts": 1716454225149143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225191967, "dur": 252,
    "args": {
      "External id": 233818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233818, "pid": 5, "tid": 7, "ts": 1716454225191967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149173, "dur": 8,
    "args": {
      "External id": 233818,
      "cbid": 211, "correlation": 233818
    }
  },
  {
    "ph": "s", "id": 233818, "pid": 76337, "tid": -914061504, "ts": 1716454225149173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149254, "dur": 1,
    "args": {
      "External id": 233834,
      "cbid": 251, "correlation": 233834
    }
  },
  {
    "ph": "f", "id": 233834, "pid": 76337, "tid": -914061504, "ts": 1716454225149254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149259, "dur": 0,
    "args": {
      "External id": 233836,
      "cbid": 251, "correlation": 233836
    }
  },
  {
    "ph": "f", "id": 233836, "pid": 76337, "tid": -914061504, "ts": 1716454225149259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225192220, "dur": 356,
    "args": {
      "External id": 233837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233837,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 233837, "pid": 5, "tid": 7, "ts": 1716454225192220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149262, "dur": 13,
    "args": {
      "External id": 233837,
      "cbid": 211, "correlation": 233837
    }
  },
  {
    "ph": "s", "id": 233837, "pid": 76337, "tid": -914061504, "ts": 1716454225149262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225192577, "dur": 50,
    "args": {
      "External id": 233845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233845,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233845, "pid": 5, "tid": 7, "ts": 1716454225192577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149304, "dur": 9,
    "args": {
      "External id": 233845,
      "cbid": 211, "correlation": 233845
    }
  },
  {
    "ph": "s", "id": 233845, "pid": 76337, "tid": -914061504, "ts": 1716454225149304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225192628, "dur": 154,
    "args": {
      "External id": 233856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233856, "pid": 5, "tid": 7, "ts": 1716454225192628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149370, "dur": 13,
    "args": {
      "External id": 233856,
      "cbid": 211, "correlation": 233856
    }
  },
  {
    "ph": "s", "id": 233856, "pid": 76337, "tid": -914061504, "ts": 1716454225149370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149434, "dur": 0,
    "args": {
      "External id": 233868,
      "cbid": 317, "correlation": 233868
    }
  },
  {
    "ph": "f", "id": 233868, "pid": 76337, "tid": -914061504, "ts": 1716454225149434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149435, "dur": 0,
    "args": {
      "External id": 233869,
      "cbid": 203, "correlation": 233869
    }
  },
  {
    "ph": "f", "id": 233869, "pid": 76337, "tid": -914061504, "ts": 1716454225149435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149436, "dur": 0,
    "args": {
      "External id": 233870,
      "cbid": 205, "correlation": 233870
    }
  },
  {
    "ph": "f", "id": 233870, "pid": 76337, "tid": -914061504, "ts": 1716454225149436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149458, "dur": 1,
    "args": {
      "External id": 233874,
      "cbid": 251, "correlation": 233874
    }
  },
  {
    "ph": "f", "id": 233874, "pid": 76337, "tid": -914061504, "ts": 1716454225149458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149460, "dur": 0,
    "args": {
      "External id": 233875,
      "cbid": 251, "correlation": 233875
    }
  },
  {
    "ph": "f", "id": 233875, "pid": 76337, "tid": -914061504, "ts": 1716454225149460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149460, "dur": 0,
    "args": {
      "External id": 233876,
      "cbid": 251, "correlation": 233876
    }
  },
  {
    "ph": "f", "id": 233876, "pid": 76337, "tid": -914061504, "ts": 1716454225149460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149461, "dur": 0,
    "args": {
      "External id": 233877,
      "cbid": 251, "correlation": 233877
    }
  },
  {
    "ph": "f", "id": 233877, "pid": 76337, "tid": -914061504, "ts": 1716454225149461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149462, "dur": 0,
    "args": {
      "External id": 233878,
      "cbid": 251, "correlation": 233878
    }
  },
  {
    "ph": "f", "id": 233878, "pid": 76337, "tid": -914061504, "ts": 1716454225149462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149463, "dur": 0,
    "args": {
      "External id": 233879,
      "cbid": 251, "correlation": 233879
    }
  },
  {
    "ph": "f", "id": 233879, "pid": 76337, "tid": -914061504, "ts": 1716454225149463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149464, "dur": 0,
    "args": {
      "External id": 233880,
      "cbid": 251, "correlation": 233880
    }
  },
  {
    "ph": "f", "id": 233880, "pid": 76337, "tid": -914061504, "ts": 1716454225149464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149464, "dur": 0,
    "args": {
      "External id": 233881,
      "cbid": 251, "correlation": 233881
    }
  },
  {
    "ph": "f", "id": 233881, "pid": 76337, "tid": -914061504, "ts": 1716454225149464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149466, "dur": 0,
    "args": {
      "External id": 233882,
      "cbid": 251, "correlation": 233882
    }
  },
  {
    "ph": "f", "id": 233882, "pid": 76337, "tid": -914061504, "ts": 1716454225149466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225192784, "dur": 113,
    "args": {
      "External id": 233883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233883,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 233883, "pid": 5, "tid": 7, "ts": 1716454225192784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149468, "dur": 12,
    "args": {
      "External id": 233883,
      "cbid": 211, "correlation": 233883
    }
  },
  {
    "ph": "s", "id": 233883, "pid": 76337, "tid": -914061504, "ts": 1716454225149468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225192898, "dur": 59,
    "args": {
      "External id": 233889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233889,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233889, "pid": 5, "tid": 7, "ts": 1716454225192898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149503, "dur": 9,
    "args": {
      "External id": 233889,
      "cbid": 211, "correlation": 233889
    }
  },
  {
    "ph": "s", "id": 233889, "pid": 76337, "tid": -914061504, "ts": 1716454225149503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225192958, "dur": 49,
    "args": {
      "External id": 233897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233897,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233897, "pid": 5, "tid": 7, "ts": 1716454225192958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149536, "dur": 9,
    "args": {
      "External id": 233897,
      "cbid": 211, "correlation": 233897
    }
  },
  {
    "ph": "s", "id": 233897, "pid": 76337, "tid": -914061504, "ts": 1716454225149536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149610, "dur": 0,
    "args": {
      "External id": 233907,
      "cbid": 317, "correlation": 233907
    }
  },
  {
    "ph": "f", "id": 233907, "pid": 76337, "tid": -914061504, "ts": 1716454225149610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149611, "dur": 0,
    "args": {
      "External id": 233908,
      "cbid": 203, "correlation": 233908
    }
  },
  {
    "ph": "f", "id": 233908, "pid": 76337, "tid": -914061504, "ts": 1716454225149611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149611, "dur": 0,
    "args": {
      "External id": 233909,
      "cbid": 205, "correlation": 233909
    }
  },
  {
    "ph": "f", "id": 233909, "pid": 76337, "tid": -914061504, "ts": 1716454225149611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193009, "dur": 41,
    "args": {
      "External id": 233913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233913, "pid": 5, "tid": 7, "ts": 1716454225193009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149629, "dur": 13,
    "args": {
      "External id": 233913,
      "cbid": 211, "correlation": 233913
    }
  },
  {
    "ph": "s", "id": 233913, "pid": 76337, "tid": -914061504, "ts": 1716454225149629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193051, "dur": 14,
    "args": {
      "External id": 233915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233915, "pid": 5, "tid": 7, "ts": 1716454225193051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149645, "dur": 5,
    "args": {
      "External id": 233915,
      "cbid": 211, "correlation": 233915
    }
  },
  {
    "ph": "s", "id": 233915, "pid": 76337, "tid": -914061504, "ts": 1716454225149645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225193067, "dur": 1,
    "args": {
      "External id": 233917,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 233917,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9061946902654867
    }
  },
  {
    "ph": "f", "id": 233917, "pid": 5, "tid": 7, "ts": 1716454225193067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149663, "dur": 20,
    "args": {
      "External id": 233917,
      "cbid": 51, "correlation": 233917
    }
  },
  {
    "ph": "s", "id": 233917, "pid": 76337, "tid": -914061504, "ts": 1716454225149663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225193071, "dur": 354,
    "args": {
      "External id": 233918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233918,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233918, "pid": 5, "tid": 7, "ts": 1716454225193071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149685, "dur": 10,
    "args": {
      "External id": 233918,
      "cbid": 211, "correlation": 233918
    }
  },
  {
    "ph": "s", "id": 233918, "pid": 76337, "tid": -914061504, "ts": 1716454225149685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193426, "dur": 13,
    "args": {
      "External id": 233920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233920,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233920, "pid": 5, "tid": 7, "ts": 1716454225193426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149703, "dur": 7,
    "args": {
      "External id": 233920,
      "cbid": 211, "correlation": 233920
    }
  },
  {
    "ph": "s", "id": 233920, "pid": 76337, "tid": -914061504, "ts": 1716454225149703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225193441, "dur": 14,
    "args": {
      "External id": 233926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233926, "pid": 5, "tid": 7, "ts": 1716454225193441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149734, "dur": 8,
    "args": {
      "External id": 233926,
      "cbid": 211, "correlation": 233926
    }
  },
  {
    "ph": "s", "id": 233926, "pid": 76337, "tid": -914061504, "ts": 1716454225149734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193456, "dur": 18,
    "args": {
      "External id": 233946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233946,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 233946, "pid": 5, "tid": 7, "ts": 1716454225193456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149826, "dur": 12,
    "args": {
      "External id": 233946,
      "cbid": 211, "correlation": 233946
    }
  },
  {
    "ph": "s", "id": 233946, "pid": 76337, "tid": -914061504, "ts": 1716454225149826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225193476, "dur": 4,
    "args": {
      "External id": 233958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233958,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 233958, "pid": 5, "tid": 7, "ts": 1716454225193476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149847, "dur": 6,
    "args": {
      "External id": 233958,
      "cbid": 211, "correlation": 233958
    }
  },
  {
    "ph": "s", "id": 233958, "pid": 76337, "tid": -914061504, "ts": 1716454225149847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225193481, "dur": 17,
    "args": {
      "External id": 233961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233961, "pid": 5, "tid": 7, "ts": 1716454225193481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149865, "dur": 7,
    "args": {
      "External id": 233961,
      "cbid": 211, "correlation": 233961
    }
  },
  {
    "ph": "s", "id": 233961, "pid": 76337, "tid": -914061504, "ts": 1716454225149865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225193500, "dur": 11,
    "args": {
      "External id": 233970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233970,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233970, "pid": 5, "tid": 7, "ts": 1716454225193500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149904, "dur": 10,
    "args": {
      "External id": 233970,
      "cbid": 211, "correlation": 233970
    }
  },
  {
    "ph": "s", "id": 233970, "pid": 76337, "tid": -914061504, "ts": 1716454225149904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149959, "dur": 0,
    "args": {
      "External id": 233980,
      "cbid": 317, "correlation": 233980
    }
  },
  {
    "ph": "f", "id": 233980, "pid": 76337, "tid": -914061504, "ts": 1716454225149959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149960, "dur": 0,
    "args": {
      "External id": 233981,
      "cbid": 203, "correlation": 233981
    }
  },
  {
    "ph": "f", "id": 233981, "pid": 76337, "tid": -914061504, "ts": 1716454225149960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149960, "dur": 0,
    "args": {
      "External id": 233982,
      "cbid": 205, "correlation": 233982
    }
  },
  {
    "ph": "f", "id": 233982, "pid": 76337, "tid": -914061504, "ts": 1716454225149960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193512, "dur": 11,
    "args": {
      "External id": 233986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233986, "pid": 5, "tid": 7, "ts": 1716454225193512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149984, "dur": 12,
    "args": {
      "External id": 233986,
      "cbid": 211, "correlation": 233986
    }
  },
  {
    "ph": "s", "id": 233986, "pid": 76337, "tid": -914061504, "ts": 1716454225149984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193525, "dur": 24,
    "args": {
      "External id": 233988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233988, "pid": 5, "tid": 7, "ts": 1716454225193525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225149998, "dur": 5,
    "args": {
      "External id": 233988,
      "cbid": 211, "correlation": 233988
    }
  },
  {
    "ph": "s", "id": 233988, "pid": 76337, "tid": -914061504, "ts": 1716454225149998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225193550, "dur": 4,
    "args": {
      "External id": 233990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 233990, "pid": 5, "tid": 7, "ts": 1716454225193550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150008, "dur": 5,
    "args": {
      "External id": 233990,
      "cbid": 211, "correlation": 233990
    }
  },
  {
    "ph": "s", "id": 233990, "pid": 76337, "tid": -914061504, "ts": 1716454225150008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150017, "dur": 0,
    "args": {
      "External id": 233991,
      "cbid": 51, "correlation": 233991
    }
  },
  {
    "ph": "s", "id": 233991, "pid": 76337, "tid": -914061504, "ts": 1716454225150017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225193555, "dur": 353,
    "args": {
      "External id": 233992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233992,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 233992, "pid": 5, "tid": 7, "ts": 1716454225193555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150018, "dur": 8,
    "args": {
      "External id": 233992,
      "cbid": 211, "correlation": 233992
    }
  },
  {
    "ph": "s", "id": 233992, "pid": 76337, "tid": -914061504, "ts": 1716454225150018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225193909, "dur": 21,
    "args": {
      "External id": 233993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233993,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233993, "pid": 5, "tid": 7, "ts": 1716454225193909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150029, "dur": 5,
    "args": {
      "External id": 233993,
      "cbid": 211, "correlation": 233993
    }
  },
  {
    "ph": "s", "id": 233993, "pid": 76337, "tid": -914061504, "ts": 1716454225150029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225193932, "dur": 32,
    "args": {
      "External id": 233999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 233999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 233999, "pid": 5, "tid": 7, "ts": 1716454225193932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150057, "dur": 8,
    "args": {
      "External id": 233999,
      "cbid": 211, "correlation": 233999
    }
  },
  {
    "ph": "s", "id": 233999, "pid": 76337, "tid": -914061504, "ts": 1716454225150057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225193966, "dur": 4,
    "args": {
      "External id": 234007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234007,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 234007, "pid": 5, "tid": 7, "ts": 1716454225193966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150100, "dur": 9,
    "args": {
      "External id": 234007,
      "cbid": 211, "correlation": 234007
    }
  },
  {
    "ph": "s", "id": 234007, "pid": 76337, "tid": -914061504, "ts": 1716454225150100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150165, "dur": 2,
    "args": {
      "External id": 234023,
      "cbid": 251, "correlation": 234023
    }
  },
  {
    "ph": "f", "id": 234023, "pid": 76337, "tid": -914061504, "ts": 1716454225150165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150172, "dur": 0,
    "args": {
      "External id": 234025,
      "cbid": 251, "correlation": 234025
    }
  },
  {
    "ph": "f", "id": 234025, "pid": 76337, "tid": -914061504, "ts": 1716454225150172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225193971, "dur": 13,
    "args": {
      "External id": 234026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234026,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 234026, "pid": 5, "tid": 7, "ts": 1716454225193971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150174, "dur": 11,
    "args": {
      "External id": 234026,
      "cbid": 211, "correlation": 234026
    }
  },
  {
    "ph": "s", "id": 234026, "pid": 76337, "tid": -914061504, "ts": 1716454225150174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225193984, "dur": 5,
    "args": {
      "External id": 234028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234028,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 234028, "pid": 5, "tid": 7, "ts": 1716454225193984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150187, "dur": 5,
    "args": {
      "External id": 234028,
      "cbid": 211, "correlation": 234028
    }
  },
  {
    "ph": "s", "id": 234028, "pid": 76337, "tid": -914061504, "ts": 1716454225150187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225193991, "dur": 29,
    "args": {
      "External id": 234038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234038, "pid": 5, "tid": 7, "ts": 1716454225193991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150245, "dur": 12,
    "args": {
      "External id": 234038,
      "cbid": 211, "correlation": 234038
    }
  },
  {
    "ph": "s", "id": 234038, "pid": 76337, "tid": -914061504, "ts": 1716454225150245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225194021, "dur": 30,
    "args": {
      "External id": 234058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234058,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 234058, "pid": 5, "tid": 7, "ts": 1716454225194021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150311, "dur": 11,
    "args": {
      "External id": 234058,
      "cbid": 211, "correlation": 234058
    }
  },
  {
    "ph": "s", "id": 234058, "pid": 76337, "tid": -914061504, "ts": 1716454225150311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225194052, "dur": 4,
    "args": {
      "External id": 234070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234070,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 234070, "pid": 5, "tid": 7, "ts": 1716454225194052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150332, "dur": 6,
    "args": {
      "External id": 234070,
      "cbid": 211, "correlation": 234070
    }
  },
  {
    "ph": "s", "id": 234070, "pid": 76337, "tid": -914061504, "ts": 1716454225150332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225194057, "dur": 29,
    "args": {
      "External id": 234073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234073, "pid": 5, "tid": 7, "ts": 1716454225194057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150351, "dur": 6,
    "args": {
      "External id": 234073,
      "cbid": 211, "correlation": 234073
    }
  },
  {
    "ph": "s", "id": 234073, "pid": 76337, "tid": -914061504, "ts": 1716454225150351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225194087, "dur": 20,
    "args": {
      "External id": 234082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234082,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234082, "pid": 5, "tid": 7, "ts": 1716454225194087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150392, "dur": 10,
    "args": {
      "External id": 234082,
      "cbid": 211, "correlation": 234082
    }
  },
  {
    "ph": "s", "id": 234082, "pid": 76337, "tid": -914061504, "ts": 1716454225150392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150456, "dur": 0,
    "args": {
      "External id": 234092,
      "cbid": 317, "correlation": 234092
    }
  },
  {
    "ph": "f", "id": 234092, "pid": 76337, "tid": -914061504, "ts": 1716454225150456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150457, "dur": 0,
    "args": {
      "External id": 234093,
      "cbid": 203, "correlation": 234093
    }
  },
  {
    "ph": "f", "id": 234093, "pid": 76337, "tid": -914061504, "ts": 1716454225150457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150457, "dur": 0,
    "args": {
      "External id": 234094,
      "cbid": 205, "correlation": 234094
    }
  },
  {
    "ph": "f", "id": 234094, "pid": 76337, "tid": -914061504, "ts": 1716454225150457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225194108, "dur": 23,
    "args": {
      "External id": 234098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234098, "pid": 5, "tid": 7, "ts": 1716454225194108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150475, "dur": 12,
    "args": {
      "External id": 234098,
      "cbid": 211, "correlation": 234098
    }
  },
  {
    "ph": "s", "id": 234098, "pid": 76337, "tid": -914061504, "ts": 1716454225150475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225194132, "dur": 43,
    "args": {
      "External id": 234100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234100, "pid": 5, "tid": 7, "ts": 1716454225194132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150490, "dur": 5,
    "args": {
      "External id": 234100,
      "cbid": 211, "correlation": 234100
    }
  },
  {
    "ph": "s", "id": 234100, "pid": 76337, "tid": -914061504, "ts": 1716454225150490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225194176, "dur": 644,
    "args": {
      "External id": 234102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234102,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234102, "pid": 5, "tid": 7, "ts": 1716454225194176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150504, "dur": 11,
    "args": {
      "External id": 234102,
      "cbid": 211, "correlation": 234102
    }
  },
  {
    "ph": "s", "id": 234102, "pid": 76337, "tid": -914061504, "ts": 1716454225150504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225194822, "dur": 20,
    "args": {
      "External id": 234104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234104,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234104, "pid": 5, "tid": 7, "ts": 1716454225194822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150519, "dur": 5,
    "args": {
      "External id": 234104,
      "cbid": 211, "correlation": 234104
    }
  },
  {
    "ph": "s", "id": 234104, "pid": 76337, "tid": -914061504, "ts": 1716454225150519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225194844, "dur": 32,
    "args": {
      "External id": 234110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234110, "pid": 5, "tid": 7, "ts": 1716454225194844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150546, "dur": 8,
    "args": {
      "External id": 234110,
      "cbid": 211, "correlation": 234110
    }
  },
  {
    "ph": "s", "id": 234110, "pid": 76337, "tid": -914061504, "ts": 1716454225150546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150604, "dur": 0,
    "args": {
      "External id": 234120,
      "cbid": 317, "correlation": 234120
    }
  },
  {
    "ph": "f", "id": 234120, "pid": 76337, "tid": -914061504, "ts": 1716454225150604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150605, "dur": 0,
    "args": {
      "External id": 234121,
      "cbid": 203, "correlation": 234121
    }
  },
  {
    "ph": "f", "id": 234121, "pid": 76337, "tid": -914061504, "ts": 1716454225150605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150605, "dur": 0,
    "args": {
      "External id": 234122,
      "cbid": 205, "correlation": 234122
    }
  },
  {
    "ph": "f", "id": 234122, "pid": 76337, "tid": -914061504, "ts": 1716454225150605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150627, "dur": 1,
    "args": {
      "External id": 234126,
      "cbid": 251, "correlation": 234126
    }
  },
  {
    "ph": "f", "id": 234126, "pid": 76337, "tid": -914061504, "ts": 1716454225150627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150629, "dur": 0,
    "args": {
      "External id": 234127,
      "cbid": 251, "correlation": 234127
    }
  },
  {
    "ph": "f", "id": 234127, "pid": 76337, "tid": -914061504, "ts": 1716454225150629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150630, "dur": 0,
    "args": {
      "External id": 234128,
      "cbid": 251, "correlation": 234128
    }
  },
  {
    "ph": "f", "id": 234128, "pid": 76337, "tid": -914061504, "ts": 1716454225150630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150630, "dur": 0,
    "args": {
      "External id": 234129,
      "cbid": 251, "correlation": 234129
    }
  },
  {
    "ph": "f", "id": 234129, "pid": 76337, "tid": -914061504, "ts": 1716454225150630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150631, "dur": 0,
    "args": {
      "External id": 234130,
      "cbid": 251, "correlation": 234130
    }
  },
  {
    "ph": "f", "id": 234130, "pid": 76337, "tid": -914061504, "ts": 1716454225150631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150632, "dur": 0,
    "args": {
      "External id": 234131,
      "cbid": 251, "correlation": 234131
    }
  },
  {
    "ph": "f", "id": 234131, "pid": 76337, "tid": -914061504, "ts": 1716454225150632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150633, "dur": 0,
    "args": {
      "External id": 234132,
      "cbid": 251, "correlation": 234132
    }
  },
  {
    "ph": "f", "id": 234132, "pid": 76337, "tid": -914061504, "ts": 1716454225150633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150633, "dur": 0,
    "args": {
      "External id": 234133,
      "cbid": 251, "correlation": 234133
    }
  },
  {
    "ph": "f", "id": 234133, "pid": 76337, "tid": -914061504, "ts": 1716454225150633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150635, "dur": 0,
    "args": {
      "External id": 234134,
      "cbid": 251, "correlation": 234134
    }
  },
  {
    "ph": "f", "id": 234134, "pid": 76337, "tid": -914061504, "ts": 1716454225150635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225194877, "dur": 50,
    "args": {
      "External id": 234135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234135,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 234135, "pid": 5, "tid": 7, "ts": 1716454225194877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150637, "dur": 12,
    "args": {
      "External id": 234135,
      "cbid": 211, "correlation": 234135
    }
  },
  {
    "ph": "s", "id": 234135, "pid": 76337, "tid": -914061504, "ts": 1716454225150637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225194928, "dur": 31,
    "args": {
      "External id": 234141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234141, "pid": 5, "tid": 7, "ts": 1716454225194928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150670, "dur": 9,
    "args": {
      "External id": 234141,
      "cbid": 211, "correlation": 234141
    }
  },
  {
    "ph": "s", "id": 234141, "pid": 76337, "tid": -914061504, "ts": 1716454225150670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225194961, "dur": 27,
    "args": {
      "External id": 234149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234149,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234149, "pid": 5, "tid": 7, "ts": 1716454225194961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150700, "dur": 8,
    "args": {
      "External id": 234149,
      "cbid": 211, "correlation": 234149
    }
  },
  {
    "ph": "s", "id": 234149, "pid": 76337, "tid": -914061504, "ts": 1716454225150700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225194989, "dur": 20,
    "args": {
      "External id": 234157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234157,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234157, "pid": 5, "tid": 7, "ts": 1716454225194989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150729, "dur": 8,
    "args": {
      "External id": 234157,
      "cbid": 211, "correlation": 234157
    }
  },
  {
    "ph": "s", "id": 234157, "pid": 76337, "tid": -914061504, "ts": 1716454225150729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225195011, "dur": 29,
    "args": {
      "External id": 234177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234177,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 234177, "pid": 5, "tid": 7, "ts": 1716454225195011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150811, "dur": 13,
    "args": {
      "External id": 234177,
      "cbid": 211, "correlation": 234177
    }
  },
  {
    "ph": "s", "id": 234177, "pid": 76337, "tid": -914061504, "ts": 1716454225150811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225195041, "dur": 4,
    "args": {
      "External id": 234189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234189,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 234189, "pid": 5, "tid": 7, "ts": 1716454225195041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150833, "dur": 6,
    "args": {
      "External id": 234189,
      "cbid": 211, "correlation": 234189
    }
  },
  {
    "ph": "s", "id": 234189, "pid": 76337, "tid": -914061504, "ts": 1716454225150833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225195046, "dur": 30,
    "args": {
      "External id": 234192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234192, "pid": 5, "tid": 7, "ts": 1716454225195046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150851, "dur": 6,
    "args": {
      "External id": 234192,
      "cbid": 211, "correlation": 234192
    }
  },
  {
    "ph": "s", "id": 234192, "pid": 76337, "tid": -914061504, "ts": 1716454225150851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150908, "dur": 0,
    "args": {
      "External id": 234203,
      "cbid": 317, "correlation": 234203
    }
  },
  {
    "ph": "f", "id": 234203, "pid": 76337, "tid": -914061504, "ts": 1716454225150908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150909, "dur": 0,
    "args": {
      "External id": 234204,
      "cbid": 203, "correlation": 234204
    }
  },
  {
    "ph": "f", "id": 234204, "pid": 76337, "tid": -914061504, "ts": 1716454225150909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150910, "dur": 0,
    "args": {
      "External id": 234205,
      "cbid": 205, "correlation": 234205
    }
  },
  {
    "ph": "f", "id": 234205, "pid": 76337, "tid": -914061504, "ts": 1716454225150910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225195077, "dur": 21,
    "args": {
      "External id": 234209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234209,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234209, "pid": 5, "tid": 7, "ts": 1716454225195077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150925, "dur": 11,
    "args": {
      "External id": 234209,
      "cbid": 211, "correlation": 234209
    }
  },
  {
    "ph": "s", "id": 234209, "pid": 76337, "tid": -914061504, "ts": 1716454225150925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225195100, "dur": 118,
    "args": {
      "External id": 234211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234211,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234211, "pid": 5, "tid": 7, "ts": 1716454225195100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150944, "dur": 8,
    "args": {
      "External id": 234211,
      "cbid": 211, "correlation": 234211
    }
  },
  {
    "ph": "s", "id": 234211, "pid": 76337, "tid": -914061504, "ts": 1716454225150944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225195219, "dur": 23,
    "args": {
      "External id": 234213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234213,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234213, "pid": 5, "tid": 7, "ts": 1716454225195219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150956, "dur": 5,
    "args": {
      "External id": 234213,
      "cbid": 211, "correlation": 234213
    }
  },
  {
    "ph": "s", "id": 234213, "pid": 76337, "tid": -914061504, "ts": 1716454225150956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225195243, "dur": 32,
    "args": {
      "External id": 234219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234219, "pid": 5, "tid": 7, "ts": 1716454225195243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225150994, "dur": 9,
    "args": {
      "External id": 234219,
      "cbid": 211, "correlation": 234219
    }
  },
  {
    "ph": "s", "id": 234219, "pid": 76337, "tid": -914061504, "ts": 1716454225150994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225195277, "dur": 160,
    "args": {
      "External id": 234228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234228,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234228, "pid": 5, "tid": 7, "ts": 1716454225195277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151078, "dur": 14,
    "args": {
      "External id": 234228,
      "cbid": 211, "correlation": 234228
    }
  },
  {
    "ph": "s", "id": 234228, "pid": 76337, "tid": -914061504, "ts": 1716454225151078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225195439, "dur": 64,
    "args": {
      "External id": 234250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234250,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234250, "pid": 5, "tid": 7, "ts": 1716454225195439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151136, "dur": 11,
    "args": {
      "External id": 234250,
      "cbid": 211, "correlation": 234250
    }
  },
  {
    "ph": "s", "id": 234250, "pid": 76337, "tid": -914061504, "ts": 1716454225151136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151227, "dur": 1,
    "args": {
      "External id": 234261,
      "cbid": 251, "correlation": 234261
    }
  },
  {
    "ph": "f", "id": 234261, "pid": 76337, "tid": -914061504, "ts": 1716454225151227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225195504, "dur": 148,
    "args": {
      "External id": 234262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234262,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234262, "pid": 5, "tid": 7, "ts": 1716454225195504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151232, "dur": 13,
    "args": {
      "External id": 234262,
      "cbid": 211, "correlation": 234262
    }
  },
  {
    "ph": "s", "id": 234262, "pid": 76337, "tid": -914061504, "ts": 1716454225151232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151303, "dur": 1,
    "args": {
      "External id": 234273,
      "cbid": 251, "correlation": 234273
    }
  },
  {
    "ph": "f", "id": 234273, "pid": 76337, "tid": -914061504, "ts": 1716454225151303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225195653, "dur": 144,
    "args": {
      "External id": 234274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234274,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234274, "pid": 5, "tid": 7, "ts": 1716454225195653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151307, "dur": 12,
    "args": {
      "External id": 234274,
      "cbid": 211, "correlation": 234274
    }
  },
  {
    "ph": "s", "id": 234274, "pid": 76337, "tid": -914061504, "ts": 1716454225151307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151372, "dur": 1,
    "args": {
      "External id": 234285,
      "cbid": 251, "correlation": 234285
    }
  },
  {
    "ph": "f", "id": 234285, "pid": 76337, "tid": -914061504, "ts": 1716454225151372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225195799, "dur": 144,
    "args": {
      "External id": 234286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234286,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234286, "pid": 5, "tid": 7, "ts": 1716454225195799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151376, "dur": 11,
    "args": {
      "External id": 234286,
      "cbid": 211, "correlation": 234286
    }
  },
  {
    "ph": "s", "id": 234286, "pid": 76337, "tid": -914061504, "ts": 1716454225151376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225195944, "dur": 1903,
    "args": {
      "External id": 234307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234307,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 234307, "pid": 5, "tid": 7, "ts": 1716454225195944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151458, "dur": 20,
    "args": {
      "External id": 234307,
      "cbid": 211, "correlation": 234307
    }
  },
  {
    "ph": "s", "id": 234307, "pid": 76337, "tid": -914061504, "ts": 1716454225151458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151568, "dur": 1,
    "args": {
      "External id": 234325,
      "cbid": 251, "correlation": 234325
    }
  },
  {
    "ph": "f", "id": 234325, "pid": 76337, "tid": -914061504, "ts": 1716454225151568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225197849, "dur": 144,
    "args": {
      "External id": 234327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234327,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 234327, "pid": 5, "tid": 7, "ts": 1716454225197849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151574, "dur": 13,
    "args": {
      "External id": 234327,
      "cbid": 211, "correlation": 234327
    }
  },
  {
    "ph": "s", "id": 234327, "pid": 76337, "tid": -914061504, "ts": 1716454225151574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225197994, "dur": 36,
    "args": {
      "External id": 234335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234335,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234335, "pid": 5, "tid": 7, "ts": 1716454225197994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151645, "dur": 12,
    "args": {
      "External id": 234335,
      "cbid": 211, "correlation": 234335
    }
  },
  {
    "ph": "s", "id": 234335, "pid": 76337, "tid": -914061504, "ts": 1716454225151645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225198031, "dur": 51,
    "args": {
      "External id": 234343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234343,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234343, "pid": 5, "tid": 7, "ts": 1716454225198031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151684, "dur": 9,
    "args": {
      "External id": 234343,
      "cbid": 211, "correlation": 234343
    }
  },
  {
    "ph": "s", "id": 234343, "pid": 76337, "tid": -914061504, "ts": 1716454225151684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225198083, "dur": 30,
    "args": {
      "External id": 234354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234354, "pid": 5, "tid": 7, "ts": 1716454225198083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151757, "dur": 13,
    "args": {
      "External id": 234354,
      "cbid": 211, "correlation": 234354
    }
  },
  {
    "ph": "s", "id": 234354, "pid": 76337, "tid": -914061504, "ts": 1716454225151757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225198114, "dur": 33,
    "args": {
      "External id": 234376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234376,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234376, "pid": 5, "tid": 7, "ts": 1716454225198114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151790, "dur": 7,
    "args": {
      "External id": 234376,
      "cbid": 211, "correlation": 234376
    }
  },
  {
    "ph": "s", "id": 234376, "pid": 76337, "tid": -914061504, "ts": 1716454225151790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151876, "dur": 1,
    "args": {
      "External id": 234387,
      "cbid": 251, "correlation": 234387
    }
  },
  {
    "ph": "f", "id": 234387, "pid": 76337, "tid": -914061504, "ts": 1716454225151876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225198149, "dur": 89,
    "args": {
      "External id": 234388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234388,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234388, "pid": 5, "tid": 7, "ts": 1716454225198149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151882, "dur": 13,
    "args": {
      "External id": 234388,
      "cbid": 211, "correlation": 234388
    }
  },
  {
    "ph": "s", "id": 234388, "pid": 76337, "tid": -914061504, "ts": 1716454225151882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151952, "dur": 1,
    "args": {
      "External id": 234399,
      "cbid": 251, "correlation": 234399
    }
  },
  {
    "ph": "f", "id": 234399, "pid": 76337, "tid": -914061504, "ts": 1716454225151952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151956, "dur": 0,
    "args": {
      "External id": 234400,
      "cbid": 251, "correlation": 234400
    }
  },
  {
    "ph": "f", "id": 234400, "pid": 76337, "tid": -914061504, "ts": 1716454225151956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225198240, "dur": 11,
    "args": {
      "External id": 234401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234401,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 234401, "pid": 5, "tid": 7, "ts": 1716454225198240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151957, "dur": 12,
    "args": {
      "External id": 234401,
      "cbid": 211, "correlation": 234401
    }
  },
  {
    "ph": "s", "id": 234401, "pid": 76337, "tid": -914061504, "ts": 1716454225151957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225198252, "dur": 5,
    "args": {
      "External id": 234403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234403,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 234403, "pid": 5, "tid": 7, "ts": 1716454225198252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225151981, "dur": 7,
    "args": {
      "External id": 234403,
      "cbid": 211, "correlation": 234403
    }
  },
  {
    "ph": "s", "id": 234403, "pid": 76337, "tid": -914061504, "ts": 1716454225151981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152042, "dur": 1,
    "args": {
      "External id": 234414,
      "cbid": 251, "correlation": 234414
    }
  },
  {
    "ph": "f", "id": 234414, "pid": 76337, "tid": -914061504, "ts": 1716454225152042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152045, "dur": 0,
    "args": {
      "External id": 234415,
      "cbid": 251, "correlation": 234415
    }
  },
  {
    "ph": "f", "id": 234415, "pid": 76337, "tid": -914061504, "ts": 1716454225152045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225198258, "dur": 7,
    "args": {
      "External id": 234416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234416,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 234416, "pid": 5, "tid": 7, "ts": 1716454225198258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152046, "dur": 12,
    "args": {
      "External id": 234416,
      "cbid": 211, "correlation": 234416
    }
  },
  {
    "ph": "s", "id": 234416, "pid": 76337, "tid": -914061504, "ts": 1716454225152046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225198266, "dur": 3,
    "args": {
      "External id": 234418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234418,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 234418, "pid": 5, "tid": 7, "ts": 1716454225198266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152060, "dur": 6,
    "args": {
      "External id": 234418,
      "cbid": 211, "correlation": 234418
    }
  },
  {
    "ph": "s", "id": 234418, "pid": 76337, "tid": -914061504, "ts": 1716454225152060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225198271, "dur": 90,
    "args": {
      "External id": 234439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234439,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 234439, "pid": 5, "tid": 7, "ts": 1716454225198271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152135, "dur": 13,
    "args": {
      "External id": 234439,
      "cbid": 211, "correlation": 234439
    }
  },
  {
    "ph": "s", "id": 234439, "pid": 76337, "tid": -914061504, "ts": 1716454225152135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152233, "dur": 1,
    "args": {
      "External id": 234457,
      "cbid": 251, "correlation": 234457
    }
  },
  {
    "ph": "f", "id": 234457, "pid": 76337, "tid": -914061504, "ts": 1716454225152233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225198362, "dur": 97,
    "args": {
      "External id": 234459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234459,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234459, "pid": 5, "tid": 7, "ts": 1716454225198362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152238, "dur": 13,
    "args": {
      "External id": 234459,
      "cbid": 211, "correlation": 234459
    }
  },
  {
    "ph": "s", "id": 234459, "pid": 76337, "tid": -914061504, "ts": 1716454225152238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225198460, "dur": 19,
    "args": {
      "External id": 234467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234467,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234467, "pid": 5, "tid": 7, "ts": 1716454225198460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152308, "dur": 12,
    "args": {
      "External id": 234467,
      "cbid": 211, "correlation": 234467
    }
  },
  {
    "ph": "s", "id": 234467, "pid": 76337, "tid": -914061504, "ts": 1716454225152308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225198480, "dur": 37,
    "args": {
      "External id": 234475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234475, "pid": 5, "tid": 7, "ts": 1716454225198480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152350, "dur": 9,
    "args": {
      "External id": 234475,
      "cbid": 211, "correlation": 234475
    }
  },
  {
    "ph": "s", "id": 234475, "pid": 76337, "tid": -914061504, "ts": 1716454225152350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225198519, "dur": 33,
    "args": {
      "External id": 234497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234497,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234497, "pid": 5, "tid": 7, "ts": 1716454225198519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152401, "dur": 10,
    "args": {
      "External id": 234497,
      "cbid": 211, "correlation": 234497
    }
  },
  {
    "ph": "s", "id": 234497, "pid": 76337, "tid": -914061504, "ts": 1716454225152401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152490, "dur": 1,
    "args": {
      "External id": 234513,
      "cbid": 251, "correlation": 234513
    }
  },
  {
    "ph": "f", "id": 234513, "pid": 76337, "tid": -914061504, "ts": 1716454225152490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152495, "dur": 0,
    "args": {
      "External id": 234515,
      "cbid": 251, "correlation": 234515
    }
  },
  {
    "ph": "f", "id": 234515, "pid": 76337, "tid": -914061504, "ts": 1716454225152495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225198554, "dur": 532,
    "args": {
      "External id": 234516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234516,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 234516, "pid": 5, "tid": 7, "ts": 1716454225198554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152498, "dur": 13,
    "args": {
      "External id": 234516,
      "cbid": 211, "correlation": 234516
    }
  },
  {
    "ph": "s", "id": 234516, "pid": 76337, "tid": -914061504, "ts": 1716454225152498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225199087, "dur": 125,
    "args": {
      "External id": 234524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234524, "pid": 5, "tid": 7, "ts": 1716454225199087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152564, "dur": 12,
    "args": {
      "External id": 234524,
      "cbid": 211, "correlation": 234524
    }
  },
  {
    "ph": "s", "id": 234524, "pid": 76337, "tid": -914061504, "ts": 1716454225152564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225199214, "dur": 128,
    "args": {
      "External id": 234532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234532, "pid": 5, "tid": 7, "ts": 1716454225199214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152595, "dur": 8,
    "args": {
      "External id": 234532,
      "cbid": 211, "correlation": 234532
    }
  },
  {
    "ph": "s", "id": 234532, "pid": 76337, "tid": -914061504, "ts": 1716454225152595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152671, "dur": 1,
    "args": {
      "External id": 234548,
      "cbid": 251, "correlation": 234548
    }
  },
  {
    "ph": "f", "id": 234548, "pid": 76337, "tid": -914061504, "ts": 1716454225152671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225199343, "dur": 301,
    "args": {
      "External id": 234550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234550,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234550, "pid": 5, "tid": 7, "ts": 1716454225199343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152677, "dur": 13,
    "args": {
      "External id": 234550,
      "cbid": 211, "correlation": 234550
    }
  },
  {
    "ph": "s", "id": 234550, "pid": 76337, "tid": -914061504, "ts": 1716454225152677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225199646, "dur": 27,
    "args": {
      "External id": 234558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234558,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234558, "pid": 5, "tid": 7, "ts": 1716454225199646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152719, "dur": 10,
    "args": {
      "External id": 234558,
      "cbid": 211, "correlation": 234558
    }
  },
  {
    "ph": "s", "id": 234558, "pid": 76337, "tid": -914061504, "ts": 1716454225152719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225199674, "dur": 80,
    "args": {
      "External id": 234569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234569, "pid": 5, "tid": 7, "ts": 1716454225199674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152789, "dur": 12,
    "args": {
      "External id": 234569,
      "cbid": 211, "correlation": 234569
    }
  },
  {
    "ph": "s", "id": 234569, "pid": 76337, "tid": -914061504, "ts": 1716454225152789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152853, "dur": 0,
    "args": {
      "External id": 234581,
      "cbid": 317, "correlation": 234581
    }
  },
  {
    "ph": "f", "id": 234581, "pid": 76337, "tid": -914061504, "ts": 1716454225152853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152854, "dur": 0,
    "args": {
      "External id": 234582,
      "cbid": 203, "correlation": 234582
    }
  },
  {
    "ph": "f", "id": 234582, "pid": 76337, "tid": -914061504, "ts": 1716454225152854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152855, "dur": 0,
    "args": {
      "External id": 234583,
      "cbid": 205, "correlation": 234583
    }
  },
  {
    "ph": "f", "id": 234583, "pid": 76337, "tid": -914061504, "ts": 1716454225152855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225199755, "dur": 22,
    "args": {
      "External id": 234587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234587, "pid": 5, "tid": 7, "ts": 1716454225199755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152871, "dur": 12,
    "args": {
      "External id": 234587,
      "cbid": 211, "correlation": 234587
    }
  },
  {
    "ph": "s", "id": 234587, "pid": 76337, "tid": -914061504, "ts": 1716454225152871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225199779, "dur": 117,
    "args": {
      "External id": 234589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234589,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234589, "pid": 5, "tid": 7, "ts": 1716454225199779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152890, "dur": 6,
    "args": {
      "External id": 234589,
      "cbid": 211, "correlation": 234589
    }
  },
  {
    "ph": "s", "id": 234589, "pid": 76337, "tid": -914061504, "ts": 1716454225152890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225199898, "dur": 23,
    "args": {
      "External id": 234591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234591, "pid": 5, "tid": 7, "ts": 1716454225199898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152901, "dur": 5,
    "args": {
      "External id": 234591,
      "cbid": 211, "correlation": 234591
    }
  },
  {
    "ph": "s", "id": 234591, "pid": 76337, "tid": -914061504, "ts": 1716454225152901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225199922, "dur": 33,
    "args": {
      "External id": 234597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234597,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234597, "pid": 5, "tid": 7, "ts": 1716454225199922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152928, "dur": 10,
    "args": {
      "External id": 234597,
      "cbid": 211, "correlation": 234597
    }
  },
  {
    "ph": "s", "id": 234597, "pid": 76337, "tid": -914061504, "ts": 1716454225152928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225199956, "dur": 26,
    "args": {
      "External id": 234605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234605,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234605, "pid": 5, "tid": 7, "ts": 1716454225199956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225152960, "dur": 8,
    "args": {
      "External id": 234605,
      "cbid": 211, "correlation": 234605
    }
  },
  {
    "ph": "s", "id": 234605, "pid": 76337, "tid": -914061504, "ts": 1716454225152960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225199984, "dur": 29,
    "args": {
      "External id": 234625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234625,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 234625, "pid": 5, "tid": 7, "ts": 1716454225199984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153045, "dur": 12,
    "args": {
      "External id": 234625,
      "cbid": 211, "correlation": 234625
    }
  },
  {
    "ph": "s", "id": 234625, "pid": 76337, "tid": -914061504, "ts": 1716454225153045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225200015, "dur": 5,
    "args": {
      "External id": 234637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234637,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 234637, "pid": 5, "tid": 7, "ts": 1716454225200015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153067, "dur": 7,
    "args": {
      "External id": 234637,
      "cbid": 211, "correlation": 234637
    }
  },
  {
    "ph": "s", "id": 234637, "pid": 76337, "tid": -914061504, "ts": 1716454225153067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225200021, "dur": 30,
    "args": {
      "External id": 234640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234640, "pid": 5, "tid": 7, "ts": 1716454225200021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153087, "dur": 6,
    "args": {
      "External id": 234640,
      "cbid": 211, "correlation": 234640
    }
  },
  {
    "ph": "s", "id": 234640, "pid": 76337, "tid": -914061504, "ts": 1716454225153087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225200053, "dur": 20,
    "args": {
      "External id": 234649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234649,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234649, "pid": 5, "tid": 7, "ts": 1716454225200053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153125, "dur": 10,
    "args": {
      "External id": 234649,
      "cbid": 211, "correlation": 234649
    }
  },
  {
    "ph": "s", "id": 234649, "pid": 76337, "tid": -914061504, "ts": 1716454225153125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153175, "dur": 0,
    "args": {
      "External id": 234659,
      "cbid": 317, "correlation": 234659
    }
  },
  {
    "ph": "f", "id": 234659, "pid": 76337, "tid": -914061504, "ts": 1716454225153175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153176, "dur": 0,
    "args": {
      "External id": 234660,
      "cbid": 203, "correlation": 234660
    }
  },
  {
    "ph": "f", "id": 234660, "pid": 76337, "tid": -914061504, "ts": 1716454225153176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153177, "dur": 0,
    "args": {
      "External id": 234661,
      "cbid": 205, "correlation": 234661
    }
  },
  {
    "ph": "f", "id": 234661, "pid": 76337, "tid": -914061504, "ts": 1716454225153177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225200074, "dur": 21,
    "args": {
      "External id": 234665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234665, "pid": 5, "tid": 7, "ts": 1716454225200074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153192, "dur": 11,
    "args": {
      "External id": 234665,
      "cbid": 211, "correlation": 234665
    }
  },
  {
    "ph": "s", "id": 234665, "pid": 76337, "tid": -914061504, "ts": 1716454225153192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225200097, "dur": 43,
    "args": {
      "External id": 234667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234667,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234667, "pid": 5, "tid": 7, "ts": 1716454225200097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153206, "dur": 5,
    "args": {
      "External id": 234667,
      "cbid": 211, "correlation": 234667
    }
  },
  {
    "ph": "s", "id": 234667, "pid": 76337, "tid": -914061504, "ts": 1716454225153206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225200141, "dur": 643,
    "args": {
      "External id": 234669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234669,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234669, "pid": 5, "tid": 7, "ts": 1716454225200141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153217, "dur": 6,
    "args": {
      "External id": 234669,
      "cbid": 211, "correlation": 234669
    }
  },
  {
    "ph": "s", "id": 234669, "pid": 76337, "tid": -914061504, "ts": 1716454225153217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225200786, "dur": 22,
    "args": {
      "External id": 234671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234671, "pid": 5, "tid": 7, "ts": 1716454225200786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153227, "dur": 6,
    "args": {
      "External id": 234671,
      "cbid": 211, "correlation": 234671
    }
  },
  {
    "ph": "s", "id": 234671, "pid": 76337, "tid": -914061504, "ts": 1716454225153227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225200809, "dur": 33,
    "args": {
      "External id": 234677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234677, "pid": 5, "tid": 7, "ts": 1716454225200809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153255, "dur": 9,
    "args": {
      "External id": 234677,
      "cbid": 211, "correlation": 234677
    }
  },
  {
    "ph": "s", "id": 234677, "pid": 76337, "tid": -914061504, "ts": 1716454225153255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225200843, "dur": 3,
    "args": {
      "External id": 234685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234685,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 234685, "pid": 5, "tid": 7, "ts": 1716454225200843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153299, "dur": 9,
    "args": {
      "External id": 234685,
      "cbid": 211, "correlation": 234685
    }
  },
  {
    "ph": "s", "id": 234685, "pid": 76337, "tid": -914061504, "ts": 1716454225153299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153363, "dur": 1,
    "args": {
      "External id": 234701,
      "cbid": 251, "correlation": 234701
    }
  },
  {
    "ph": "f", "id": 234701, "pid": 76337, "tid": -914061504, "ts": 1716454225153363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153368, "dur": 0,
    "args": {
      "External id": 234703,
      "cbid": 251, "correlation": 234703
    }
  },
  {
    "ph": "f", "id": 234703, "pid": 76337, "tid": -914061504, "ts": 1716454225153368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225200848, "dur": 12,
    "args": {
      "External id": 234704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234704,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 234704, "pid": 5, "tid": 7, "ts": 1716454225200848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153370, "dur": 12,
    "args": {
      "External id": 234704,
      "cbid": 211, "correlation": 234704
    }
  },
  {
    "ph": "s", "id": 234704, "pid": 76337, "tid": -914061504, "ts": 1716454225153370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225200861, "dur": 5,
    "args": {
      "External id": 234706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234706,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 234706, "pid": 5, "tid": 7, "ts": 1716454225200861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153383, "dur": 6,
    "args": {
      "External id": 234706,
      "cbid": 211, "correlation": 234706
    }
  },
  {
    "ph": "s", "id": 234706, "pid": 76337, "tid": -914061504, "ts": 1716454225153383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225200867, "dur": 30,
    "args": {
      "External id": 234716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234716, "pid": 5, "tid": 7, "ts": 1716454225200867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153442, "dur": 11,
    "args": {
      "External id": 234716,
      "cbid": 211, "correlation": 234716
    }
  },
  {
    "ph": "s", "id": 234716, "pid": 76337, "tid": -914061504, "ts": 1716454225153442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225200898, "dur": 30,
    "args": {
      "External id": 234736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234736,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 234736, "pid": 5, "tid": 7, "ts": 1716454225200898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153506, "dur": 10,
    "args": {
      "External id": 234736,
      "cbid": 211, "correlation": 234736
    }
  },
  {
    "ph": "s", "id": 234736, "pid": 76337, "tid": -914061504, "ts": 1716454225153506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225200929, "dur": 4,
    "args": {
      "External id": 234748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234748,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 234748, "pid": 5, "tid": 7, "ts": 1716454225200929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153526, "dur": 6,
    "args": {
      "External id": 234748,
      "cbid": 211, "correlation": 234748
    }
  },
  {
    "ph": "s", "id": 234748, "pid": 76337, "tid": -914061504, "ts": 1716454225153526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225200934, "dur": 29,
    "args": {
      "External id": 234751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234751, "pid": 5, "tid": 7, "ts": 1716454225200934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153546, "dur": 7,
    "args": {
      "External id": 234751,
      "cbid": 211, "correlation": 234751
    }
  },
  {
    "ph": "s", "id": 234751, "pid": 76337, "tid": -914061504, "ts": 1716454225153546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225200964, "dur": 20,
    "args": {
      "External id": 234760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234760,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234760, "pid": 5, "tid": 7, "ts": 1716454225200964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153587, "dur": 10,
    "args": {
      "External id": 234760,
      "cbid": 211, "correlation": 234760
    }
  },
  {
    "ph": "s", "id": 234760, "pid": 76337, "tid": -914061504, "ts": 1716454225153587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153648, "dur": 0,
    "args": {
      "External id": 234770,
      "cbid": 317, "correlation": 234770
    }
  },
  {
    "ph": "f", "id": 234770, "pid": 76337, "tid": -914061504, "ts": 1716454225153648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153649, "dur": 0,
    "args": {
      "External id": 234771,
      "cbid": 203, "correlation": 234771
    }
  },
  {
    "ph": "f", "id": 234771, "pid": 76337, "tid": -914061504, "ts": 1716454225153649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153650, "dur": 0,
    "args": {
      "External id": 234772,
      "cbid": 205, "correlation": 234772
    }
  },
  {
    "ph": "f", "id": 234772, "pid": 76337, "tid": -914061504, "ts": 1716454225153650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225200986, "dur": 23,
    "args": {
      "External id": 234776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234776, "pid": 5, "tid": 7, "ts": 1716454225200986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153663, "dur": 12,
    "args": {
      "External id": 234776,
      "cbid": 211, "correlation": 234776
    }
  },
  {
    "ph": "s", "id": 234776, "pid": 76337, "tid": -914061504, "ts": 1716454225153663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225201010, "dur": 43,
    "args": {
      "External id": 234778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234778, "pid": 5, "tid": 7, "ts": 1716454225201010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153677, "dur": 6,
    "args": {
      "External id": 234778,
      "cbid": 211, "correlation": 234778
    }
  },
  {
    "ph": "s", "id": 234778, "pid": 76337, "tid": -914061504, "ts": 1716454225153677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225201054, "dur": 636,
    "args": {
      "External id": 234780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234780,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234780, "pid": 5, "tid": 7, "ts": 1716454225201054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153689, "dur": 6,
    "args": {
      "External id": 234780,
      "cbid": 211, "correlation": 234780
    }
  },
  {
    "ph": "s", "id": 234780, "pid": 76337, "tid": -914061504, "ts": 1716454225153689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225201692, "dur": 20,
    "args": {
      "External id": 234782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234782,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234782, "pid": 5, "tid": 7, "ts": 1716454225201692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153698, "dur": 5,
    "args": {
      "External id": 234782,
      "cbid": 211, "correlation": 234782
    }
  },
  {
    "ph": "s", "id": 234782, "pid": 76337, "tid": -914061504, "ts": 1716454225153698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225201713, "dur": 32,
    "args": {
      "External id": 234788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234788, "pid": 5, "tid": 7, "ts": 1716454225201713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153725, "dur": 9,
    "args": {
      "External id": 234788,
      "cbid": 211, "correlation": 234788
    }
  },
  {
    "ph": "s", "id": 234788, "pid": 76337, "tid": -914061504, "ts": 1716454225153725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225201747, "dur": 27,
    "args": {
      "External id": 234796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234796,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234796, "pid": 5, "tid": 7, "ts": 1716454225201747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153759, "dur": 8,
    "args": {
      "External id": 234796,
      "cbid": 211, "correlation": 234796
    }
  },
  {
    "ph": "s", "id": 234796, "pid": 76337, "tid": -914061504, "ts": 1716454225153759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225201775, "dur": 20,
    "args": {
      "External id": 234804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234804,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234804, "pid": 5, "tid": 7, "ts": 1716454225201775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153789, "dur": 8,
    "args": {
      "External id": 234804,
      "cbid": 211, "correlation": 234804
    }
  },
  {
    "ph": "s", "id": 234804, "pid": 76337, "tid": -914061504, "ts": 1716454225153789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225201797, "dur": 30,
    "args": {
      "External id": 234824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234824,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 234824, "pid": 5, "tid": 7, "ts": 1716454225201797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153866, "dur": 12,
    "args": {
      "External id": 234824,
      "cbid": 211, "correlation": 234824
    }
  },
  {
    "ph": "s", "id": 234824, "pid": 76337, "tid": -914061504, "ts": 1716454225153866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225201828, "dur": 4,
    "args": {
      "External id": 234836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234836,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 234836, "pid": 5, "tid": 7, "ts": 1716454225201828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153888, "dur": 7,
    "args": {
      "External id": 234836,
      "cbid": 211, "correlation": 234836
    }
  },
  {
    "ph": "s", "id": 234836, "pid": 76337, "tid": -914061504, "ts": 1716454225153888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225201833, "dur": 30,
    "args": {
      "External id": 234839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234839, "pid": 5, "tid": 7, "ts": 1716454225201833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153907, "dur": 6,
    "args": {
      "External id": 234839,
      "cbid": 211, "correlation": 234839
    }
  },
  {
    "ph": "s", "id": 234839, "pid": 76337, "tid": -914061504, "ts": 1716454225153907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153963, "dur": 0,
    "args": {
      "External id": 234850,
      "cbid": 317, "correlation": 234850
    }
  },
  {
    "ph": "f", "id": 234850, "pid": 76337, "tid": -914061504, "ts": 1716454225153963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153964, "dur": 0,
    "args": {
      "External id": 234851,
      "cbid": 203, "correlation": 234851
    }
  },
  {
    "ph": "f", "id": 234851, "pid": 76337, "tid": -914061504, "ts": 1716454225153964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153965, "dur": 0,
    "args": {
      "External id": 234852,
      "cbid": 205, "correlation": 234852
    }
  },
  {
    "ph": "f", "id": 234852, "pid": 76337, "tid": -914061504, "ts": 1716454225153965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225201864, "dur": 22,
    "args": {
      "External id": 234856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234856, "pid": 5, "tid": 7, "ts": 1716454225201864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225153986, "dur": 13,
    "args": {
      "External id": 234856,
      "cbid": 211, "correlation": 234856
    }
  },
  {
    "ph": "s", "id": 234856, "pid": 76337, "tid": -914061504, "ts": 1716454225153986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225201887, "dur": 114,
    "args": {
      "External id": 234858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234858,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234858, "pid": 5, "tid": 7, "ts": 1716454225201887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154006, "dur": 6,
    "args": {
      "External id": 234858,
      "cbid": 211, "correlation": 234858
    }
  },
  {
    "ph": "s", "id": 234858, "pid": 76337, "tid": -914061504, "ts": 1716454225154006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225202003, "dur": 21,
    "args": {
      "External id": 234860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234860,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234860, "pid": 5, "tid": 7, "ts": 1716454225202003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154015, "dur": 5,
    "args": {
      "External id": 234860,
      "cbid": 211, "correlation": 234860
    }
  },
  {
    "ph": "s", "id": 234860, "pid": 76337, "tid": -914061504, "ts": 1716454225154015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225202025, "dur": 32,
    "args": {
      "External id": 234866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234866, "pid": 5, "tid": 7, "ts": 1716454225202025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154044, "dur": 9,
    "args": {
      "External id": 234866,
      "cbid": 211, "correlation": 234866
    }
  },
  {
    "ph": "s", "id": 234866, "pid": 76337, "tid": -914061504, "ts": 1716454225154044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225202058, "dur": 174,
    "args": {
      "External id": 234875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234875,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234875, "pid": 5, "tid": 7, "ts": 1716454225202058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154126, "dur": 14,
    "args": {
      "External id": 234875,
      "cbid": 211, "correlation": 234875
    }
  },
  {
    "ph": "s", "id": 234875, "pid": 76337, "tid": -914061504, "ts": 1716454225154126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225202233, "dur": 63,
    "args": {
      "External id": 234897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234897,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234897, "pid": 5, "tid": 7, "ts": 1716454225202233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154182, "dur": 10,
    "args": {
      "External id": 234897,
      "cbid": 211, "correlation": 234897
    }
  },
  {
    "ph": "s", "id": 234897, "pid": 76337, "tid": -914061504, "ts": 1716454225154182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154270, "dur": 1,
    "args": {
      "External id": 234908,
      "cbid": 251, "correlation": 234908
    }
  },
  {
    "ph": "f", "id": 234908, "pid": 76337, "tid": -914061504, "ts": 1716454225154270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225202298, "dur": 149,
    "args": {
      "External id": 234909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234909,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234909, "pid": 5, "tid": 7, "ts": 1716454225202298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154275, "dur": 13,
    "args": {
      "External id": 234909,
      "cbid": 211, "correlation": 234909
    }
  },
  {
    "ph": "s", "id": 234909, "pid": 76337, "tid": -914061504, "ts": 1716454225154275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154346, "dur": 1,
    "args": {
      "External id": 234920,
      "cbid": 251, "correlation": 234920
    }
  },
  {
    "ph": "f", "id": 234920, "pid": 76337, "tid": -914061504, "ts": 1716454225154346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225202448, "dur": 143,
    "args": {
      "External id": 234921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234921,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234921, "pid": 5, "tid": 7, "ts": 1716454225202448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154350, "dur": 12,
    "args": {
      "External id": 234921,
      "cbid": 211, "correlation": 234921
    }
  },
  {
    "ph": "s", "id": 234921, "pid": 76337, "tid": -914061504, "ts": 1716454225154350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154415, "dur": 1,
    "args": {
      "External id": 234932,
      "cbid": 251, "correlation": 234932
    }
  },
  {
    "ph": "f", "id": 234932, "pid": 76337, "tid": -914061504, "ts": 1716454225154415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225202592, "dur": 146,
    "args": {
      "External id": 234933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234933,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 234933, "pid": 5, "tid": 7, "ts": 1716454225202592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154419, "dur": 11,
    "args": {
      "External id": 234933,
      "cbid": 211, "correlation": 234933
    }
  },
  {
    "ph": "s", "id": 234933, "pid": 76337, "tid": -914061504, "ts": 1716454225154419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225202739, "dur": 1910,
    "args": {
      "External id": 234954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234954,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 234954, "pid": 5, "tid": 7, "ts": 1716454225202739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154501, "dur": 12,
    "args": {
      "External id": 234954,
      "cbid": 211, "correlation": 234954
    }
  },
  {
    "ph": "s", "id": 234954, "pid": 76337, "tid": -914061504, "ts": 1716454225154501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154601, "dur": 1,
    "args": {
      "External id": 234972,
      "cbid": 251, "correlation": 234972
    }
  },
  {
    "ph": "f", "id": 234972, "pid": 76337, "tid": -914061504, "ts": 1716454225154601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225204650, "dur": 146,
    "args": {
      "External id": 234974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234974,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 234974, "pid": 5, "tid": 7, "ts": 1716454225204650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154607, "dur": 13,
    "args": {
      "External id": 234974,
      "cbid": 211, "correlation": 234974
    }
  },
  {
    "ph": "s", "id": 234974, "pid": 76337, "tid": -914061504, "ts": 1716454225154607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225204798, "dur": 35,
    "args": {
      "External id": 234982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234982,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234982, "pid": 5, "tid": 7, "ts": 1716454225204798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154677, "dur": 12,
    "args": {
      "External id": 234982,
      "cbid": 211, "correlation": 234982
    }
  },
  {
    "ph": "s", "id": 234982, "pid": 76337, "tid": -914061504, "ts": 1716454225154677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225204834, "dur": 50,
    "args": {
      "External id": 234990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 234990,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 234990, "pid": 5, "tid": 7, "ts": 1716454225204834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154716, "dur": 8,
    "args": {
      "External id": 234990,
      "cbid": 211, "correlation": 234990
    }
  },
  {
    "ph": "s", "id": 234990, "pid": 76337, "tid": -914061504, "ts": 1716454225154716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225204886, "dur": 30,
    "args": {
      "External id": 235001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235001,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235001, "pid": 5, "tid": 7, "ts": 1716454225204886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154787, "dur": 13,
    "args": {
      "External id": 235001,
      "cbid": 211, "correlation": 235001
    }
  },
  {
    "ph": "s", "id": 235001, "pid": 76337, "tid": -914061504, "ts": 1716454225154787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225204918, "dur": 34,
    "args": {
      "External id": 235023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235023,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235023, "pid": 5, "tid": 7, "ts": 1716454225204918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154818, "dur": 8,
    "args": {
      "External id": 235023,
      "cbid": 211, "correlation": 235023
    }
  },
  {
    "ph": "s", "id": 235023, "pid": 76337, "tid": -914061504, "ts": 1716454225154818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154903, "dur": 1,
    "args": {
      "External id": 235034,
      "cbid": 251, "correlation": 235034
    }
  },
  {
    "ph": "f", "id": 235034, "pid": 76337, "tid": -914061504, "ts": 1716454225154903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225204953, "dur": 91,
    "args": {
      "External id": 235035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235035,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235035, "pid": 5, "tid": 7, "ts": 1716454225204953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154908, "dur": 12,
    "args": {
      "External id": 235035,
      "cbid": 211, "correlation": 235035
    }
  },
  {
    "ph": "s", "id": 235035, "pid": 76337, "tid": -914061504, "ts": 1716454225154908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154984, "dur": 1,
    "args": {
      "External id": 235046,
      "cbid": 251, "correlation": 235046
    }
  },
  {
    "ph": "f", "id": 235046, "pid": 76337, "tid": -914061504, "ts": 1716454225154984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154988, "dur": 0,
    "args": {
      "External id": 235047,
      "cbid": 251, "correlation": 235047
    }
  },
  {
    "ph": "f", "id": 235047, "pid": 76337, "tid": -914061504, "ts": 1716454225154988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225205045, "dur": 11,
    "args": {
      "External id": 235048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235048,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 235048, "pid": 5, "tid": 7, "ts": 1716454225205045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225154989, "dur": 13,
    "args": {
      "External id": 235048,
      "cbid": 211, "correlation": 235048
    }
  },
  {
    "ph": "s", "id": 235048, "pid": 76337, "tid": -914061504, "ts": 1716454225154989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225205057, "dur": 5,
    "args": {
      "External id": 235050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235050,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 235050, "pid": 5, "tid": 7, "ts": 1716454225205057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155004, "dur": 6,
    "args": {
      "External id": 235050,
      "cbid": 211, "correlation": 235050
    }
  },
  {
    "ph": "s", "id": 235050, "pid": 76337, "tid": -914061504, "ts": 1716454225155004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155062, "dur": 1,
    "args": {
      "External id": 235061,
      "cbid": 251, "correlation": 235061
    }
  },
  {
    "ph": "f", "id": 235061, "pid": 76337, "tid": -914061504, "ts": 1716454225155062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155066, "dur": 0,
    "args": {
      "External id": 235062,
      "cbid": 251, "correlation": 235062
    }
  },
  {
    "ph": "f", "id": 235062, "pid": 76337, "tid": -914061504, "ts": 1716454225155066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225205063, "dur": 7,
    "args": {
      "External id": 235063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235063,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 235063, "pid": 5, "tid": 7, "ts": 1716454225205063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155067, "dur": 11,
    "args": {
      "External id": 235063,
      "cbid": 211, "correlation": 235063
    }
  },
  {
    "ph": "s", "id": 235063, "pid": 76337, "tid": -914061504, "ts": 1716454225155067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225205071, "dur": 3,
    "args": {
      "External id": 235065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235065,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 235065, "pid": 5, "tid": 7, "ts": 1716454225205071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155080, "dur": 5,
    "args": {
      "External id": 235065,
      "cbid": 211, "correlation": 235065
    }
  },
  {
    "ph": "s", "id": 235065, "pid": 76337, "tid": -914061504, "ts": 1716454225155080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225205076, "dur": 91,
    "args": {
      "External id": 235086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235086,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 235086, "pid": 5, "tid": 7, "ts": 1716454225205076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155155, "dur": 13,
    "args": {
      "External id": 235086,
      "cbid": 211, "correlation": 235086
    }
  },
  {
    "ph": "s", "id": 235086, "pid": 76337, "tid": -914061504, "ts": 1716454225155155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155251, "dur": 1,
    "args": {
      "External id": 235104,
      "cbid": 251, "correlation": 235104
    }
  },
  {
    "ph": "f", "id": 235104, "pid": 76337, "tid": -914061504, "ts": 1716454225155251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225205168, "dur": 97,
    "args": {
      "External id": 235106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235106,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235106, "pid": 5, "tid": 7, "ts": 1716454225205168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155258, "dur": 13,
    "args": {
      "External id": 235106,
      "cbid": 211, "correlation": 235106
    }
  },
  {
    "ph": "s", "id": 235106, "pid": 76337, "tid": -914061504, "ts": 1716454225155258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225205267, "dur": 19,
    "args": {
      "External id": 235114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235114,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235114, "pid": 5, "tid": 7, "ts": 1716454225205267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155326, "dur": 12,
    "args": {
      "External id": 235114,
      "cbid": 211, "correlation": 235114
    }
  },
  {
    "ph": "s", "id": 235114, "pid": 76337, "tid": -914061504, "ts": 1716454225155326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225205287, "dur": 37,
    "args": {
      "External id": 235122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235122,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235122, "pid": 5, "tid": 7, "ts": 1716454225205287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155368, "dur": 9,
    "args": {
      "External id": 235122,
      "cbid": 211, "correlation": 235122
    }
  },
  {
    "ph": "s", "id": 235122, "pid": 76337, "tid": -914061504, "ts": 1716454225155368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225205324, "dur": 33,
    "args": {
      "External id": 235144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235144,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235144, "pid": 5, "tid": 7, "ts": 1716454225205324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155418, "dur": 10,
    "args": {
      "External id": 235144,
      "cbid": 211, "correlation": 235144
    }
  },
  {
    "ph": "s", "id": 235144, "pid": 76337, "tid": -914061504, "ts": 1716454225155418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155507, "dur": 1,
    "args": {
      "External id": 235160,
      "cbid": 251, "correlation": 235160
    }
  },
  {
    "ph": "f", "id": 235160, "pid": 76337, "tid": -914061504, "ts": 1716454225155507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155512, "dur": 0,
    "args": {
      "External id": 235162,
      "cbid": 251, "correlation": 235162
    }
  },
  {
    "ph": "f", "id": 235162, "pid": 76337, "tid": -914061504, "ts": 1716454225155512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225205359, "dur": 530,
    "args": {
      "External id": 235163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235163,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235163, "pid": 5, "tid": 7, "ts": 1716454225205359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155515, "dur": 13,
    "args": {
      "External id": 235163,
      "cbid": 211, "correlation": 235163
    }
  },
  {
    "ph": "s", "id": 235163, "pid": 76337, "tid": -914061504, "ts": 1716454225155515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225205890, "dur": 124,
    "args": {
      "External id": 235171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235171,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235171, "pid": 5, "tid": 7, "ts": 1716454225205890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155580, "dur": 12,
    "args": {
      "External id": 235171,
      "cbid": 211, "correlation": 235171
    }
  },
  {
    "ph": "s", "id": 235171, "pid": 76337, "tid": -914061504, "ts": 1716454225155580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225206015, "dur": 128,
    "args": {
      "External id": 235179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235179,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235179, "pid": 5, "tid": 7, "ts": 1716454225206015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155610, "dur": 8,
    "args": {
      "External id": 235179,
      "cbid": 211, "correlation": 235179
    }
  },
  {
    "ph": "s", "id": 235179, "pid": 76337, "tid": -914061504, "ts": 1716454225155610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155687, "dur": 1,
    "args": {
      "External id": 235195,
      "cbid": 251, "correlation": 235195
    }
  },
  {
    "ph": "f", "id": 235195, "pid": 76337, "tid": -914061504, "ts": 1716454225155687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225206144, "dur": 301,
    "args": {
      "External id": 235197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235197,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235197, "pid": 5, "tid": 7, "ts": 1716454225206144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155693, "dur": 12,
    "args": {
      "External id": 235197,
      "cbid": 211, "correlation": 235197
    }
  },
  {
    "ph": "s", "id": 235197, "pid": 76337, "tid": -914061504, "ts": 1716454225155693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225206447, "dur": 27,
    "args": {
      "External id": 235205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235205,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235205, "pid": 5, "tid": 7, "ts": 1716454225206447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155734, "dur": 9,
    "args": {
      "External id": 235205,
      "cbid": 211, "correlation": 235205
    }
  },
  {
    "ph": "s", "id": 235205, "pid": 76337, "tid": -914061504, "ts": 1716454225155734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225206475, "dur": 79,
    "args": {
      "External id": 235216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235216,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235216, "pid": 5, "tid": 7, "ts": 1716454225206475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155801, "dur": 13,
    "args": {
      "External id": 235216,
      "cbid": 211, "correlation": 235216
    }
  },
  {
    "ph": "s", "id": 235216, "pid": 76337, "tid": -914061504, "ts": 1716454225155801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155865, "dur": 0,
    "args": {
      "External id": 235228,
      "cbid": 317, "correlation": 235228
    }
  },
  {
    "ph": "f", "id": 235228, "pid": 76337, "tid": -914061504, "ts": 1716454225155865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155866, "dur": 0,
    "args": {
      "External id": 235229,
      "cbid": 203, "correlation": 235229
    }
  },
  {
    "ph": "f", "id": 235229, "pid": 76337, "tid": -914061504, "ts": 1716454225155866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155866, "dur": 0,
    "args": {
      "External id": 235230,
      "cbid": 205, "correlation": 235230
    }
  },
  {
    "ph": "f", "id": 235230, "pid": 76337, "tid": -914061504, "ts": 1716454225155866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225206555, "dur": 24,
    "args": {
      "External id": 235234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235234, "pid": 5, "tid": 7, "ts": 1716454225206555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155881, "dur": 12,
    "args": {
      "External id": 235234,
      "cbid": 211, "correlation": 235234
    }
  },
  {
    "ph": "s", "id": 235234, "pid": 76337, "tid": -914061504, "ts": 1716454225155881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225206580, "dur": 118,
    "args": {
      "External id": 235236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235236,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235236, "pid": 5, "tid": 7, "ts": 1716454225206580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155900, "dur": 6,
    "args": {
      "External id": 235236,
      "cbid": 211, "correlation": 235236
    }
  },
  {
    "ph": "s", "id": 235236, "pid": 76337, "tid": -914061504, "ts": 1716454225155900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225206700, "dur": 22,
    "args": {
      "External id": 235238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235238,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235238, "pid": 5, "tid": 7, "ts": 1716454225206700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155910, "dur": 5,
    "args": {
      "External id": 235238,
      "cbid": 211, "correlation": 235238
    }
  },
  {
    "ph": "s", "id": 235238, "pid": 76337, "tid": -914061504, "ts": 1716454225155910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225206723, "dur": 32,
    "args": {
      "External id": 235244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235244, "pid": 5, "tid": 7, "ts": 1716454225206723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155937, "dur": 8,
    "args": {
      "External id": 235244,
      "cbid": 211, "correlation": 235244
    }
  },
  {
    "ph": "s", "id": 235244, "pid": 76337, "tid": -914061504, "ts": 1716454225155937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225206756, "dur": 26,
    "args": {
      "External id": 235252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235252,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235252, "pid": 5, "tid": 7, "ts": 1716454225206756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225155969, "dur": 16,
    "args": {
      "External id": 235252,
      "cbid": 211, "correlation": 235252
    }
  },
  {
    "ph": "s", "id": 235252, "pid": 76337, "tid": -914061504, "ts": 1716454225155969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156048, "dur": 0,
    "args": {
      "External id": 235262,
      "cbid": 317, "correlation": 235262
    }
  },
  {
    "ph": "f", "id": 235262, "pid": 76337, "tid": -914061504, "ts": 1716454225156048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156049, "dur": 0,
    "args": {
      "External id": 235263,
      "cbid": 203, "correlation": 235263
    }
  },
  {
    "ph": "f", "id": 235263, "pid": 76337, "tid": -914061504, "ts": 1716454225156049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156050, "dur": 0,
    "args": {
      "External id": 235264,
      "cbid": 205, "correlation": 235264
    }
  },
  {
    "ph": "f", "id": 235264, "pid": 76337, "tid": -914061504, "ts": 1716454225156050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225206784, "dur": 23,
    "args": {
      "External id": 235268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235268, "pid": 5, "tid": 7, "ts": 1716454225206784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156064, "dur": 12,
    "args": {
      "External id": 235268,
      "cbid": 211, "correlation": 235268
    }
  },
  {
    "ph": "s", "id": 235268, "pid": 76337, "tid": -914061504, "ts": 1716454225156064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225206808, "dur": 44,
    "args": {
      "External id": 235270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235270, "pid": 5, "tid": 7, "ts": 1716454225206808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156079, "dur": 5,
    "args": {
      "External id": 235270,
      "cbid": 211, "correlation": 235270
    }
  },
  {
    "ph": "s", "id": 235270, "pid": 76337, "tid": -914061504, "ts": 1716454225156079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225206853, "dur": 230,
    "args": {
      "External id": 235272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235272,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 235272, "pid": 5, "tid": 7, "ts": 1716454225206853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156091, "dur": 7,
    "args": {
      "External id": 235272,
      "cbid": 211, "correlation": 235272
    }
  },
  {
    "ph": "s", "id": 235272, "pid": 76337, "tid": -914061504, "ts": 1716454225156091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207084, "dur": 6,
    "args": {
      "External id": 235274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235274,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235274, "pid": 5, "tid": 7, "ts": 1716454225207084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156101, "dur": 5,
    "args": {
      "External id": 235274,
      "cbid": 211, "correlation": 235274
    }
  },
  {
    "ph": "s", "id": 235274, "pid": 76337, "tid": -914061504, "ts": 1716454225156101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225207092, "dur": 9,
    "args": {
      "External id": 235280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235280, "pid": 5, "tid": 7, "ts": 1716454225207092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156128, "dur": 9,
    "args": {
      "External id": 235280,
      "cbid": 211, "correlation": 235280
    }
  },
  {
    "ph": "s", "id": 235280, "pid": 76337, "tid": -914061504, "ts": 1716454225156128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207102, "dur": 12,
    "args": {
      "External id": 235300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235300,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 235300, "pid": 5, "tid": 7, "ts": 1716454225207102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156220, "dur": 12,
    "args": {
      "External id": 235300,
      "cbid": 211, "correlation": 235300
    }
  },
  {
    "ph": "s", "id": 235300, "pid": 76337, "tid": -914061504, "ts": 1716454225156220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225207115, "dur": 4,
    "args": {
      "External id": 235312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235312,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 235312, "pid": 5, "tid": 7, "ts": 1716454225207115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156243, "dur": 6,
    "args": {
      "External id": 235312,
      "cbid": 211, "correlation": 235312
    }
  },
  {
    "ph": "s", "id": 235312, "pid": 76337, "tid": -914061504, "ts": 1716454225156243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225207121, "dur": 12,
    "args": {
      "External id": 235315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235315, "pid": 5, "tid": 7, "ts": 1716454225207121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156261, "dur": 7,
    "args": {
      "External id": 235315,
      "cbid": 211, "correlation": 235315
    }
  },
  {
    "ph": "s", "id": 235315, "pid": 76337, "tid": -914061504, "ts": 1716454225156261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225207133, "dur": 7,
    "args": {
      "External id": 235324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235324,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235324, "pid": 5, "tid": 7, "ts": 1716454225207133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156301, "dur": 11,
    "args": {
      "External id": 235324,
      "cbid": 211, "correlation": 235324
    }
  },
  {
    "ph": "s", "id": 235324, "pid": 76337, "tid": -914061504, "ts": 1716454225156301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156354, "dur": 0,
    "args": {
      "External id": 235334,
      "cbid": 317, "correlation": 235334
    }
  },
  {
    "ph": "f", "id": 235334, "pid": 76337, "tid": -914061504, "ts": 1716454225156354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156354, "dur": 0,
    "args": {
      "External id": 235335,
      "cbid": 203, "correlation": 235335
    }
  },
  {
    "ph": "f", "id": 235335, "pid": 76337, "tid": -914061504, "ts": 1716454225156354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156355, "dur": 0,
    "args": {
      "External id": 235336,
      "cbid": 205, "correlation": 235336
    }
  },
  {
    "ph": "f", "id": 235336, "pid": 76337, "tid": -914061504, "ts": 1716454225156355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207141, "dur": 6,
    "args": {
      "External id": 235340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235340,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235340, "pid": 5, "tid": 7, "ts": 1716454225207141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156371, "dur": 11,
    "args": {
      "External id": 235340,
      "cbid": 211, "correlation": 235340
    }
  },
  {
    "ph": "s", "id": 235340, "pid": 76337, "tid": -914061504, "ts": 1716454225156371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207148, "dur": 82,
    "args": {
      "External id": 235342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235342,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235342, "pid": 5, "tid": 7, "ts": 1716454225207148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156385, "dur": 5,
    "args": {
      "External id": 235342,
      "cbid": 211, "correlation": 235342
    }
  },
  {
    "ph": "s", "id": 235342, "pid": 76337, "tid": -914061504, "ts": 1716454225156385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225207232, "dur": 1,
    "args": {
      "External id": 235344,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 235344,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 235344, "pid": 5, "tid": 7, "ts": 1716454225207232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156398, "dur": 9,
    "args": {
      "External id": 235344,
      "cbid": 51, "correlation": 235344
    }
  },
  {
    "ph": "s", "id": 235344, "pid": 76337, "tid": -914061504, "ts": 1716454225156398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225207236, "dur": 535,
    "args": {
      "External id": 235345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235345,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235345, "pid": 5, "tid": 7, "ts": 1716454225207236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156408, "dur": 8,
    "args": {
      "External id": 235345,
      "cbid": 211, "correlation": 235345
    }
  },
  {
    "ph": "s", "id": 235345, "pid": 76337, "tid": -914061504, "ts": 1716454225156408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207772, "dur": 12,
    "args": {
      "External id": 235347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235347,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235347, "pid": 5, "tid": 7, "ts": 1716454225207772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156420, "dur": 5,
    "args": {
      "External id": 235347,
      "cbid": 211, "correlation": 235347
    }
  },
  {
    "ph": "s", "id": 235347, "pid": 76337, "tid": -914061504, "ts": 1716454225156420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225207785, "dur": 14,
    "args": {
      "External id": 235353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235353, "pid": 5, "tid": 7, "ts": 1716454225207785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156447, "dur": 9,
    "args": {
      "External id": 235353,
      "cbid": 211, "correlation": 235353
    }
  },
  {
    "ph": "s", "id": 235353, "pid": 76337, "tid": -914061504, "ts": 1716454225156447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225207800, "dur": 4,
    "args": {
      "External id": 235361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235361,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 235361, "pid": 5, "tid": 7, "ts": 1716454225207800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156491, "dur": 9,
    "args": {
      "External id": 235361,
      "cbid": 211, "correlation": 235361
    }
  },
  {
    "ph": "s", "id": 235361, "pid": 76337, "tid": -914061504, "ts": 1716454225156491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156558, "dur": 1,
    "args": {
      "External id": 235377,
      "cbid": 251, "correlation": 235377
    }
  },
  {
    "ph": "f", "id": 235377, "pid": 76337, "tid": -914061504, "ts": 1716454225156558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156563, "dur": 0,
    "args": {
      "External id": 235379,
      "cbid": 251, "correlation": 235379
    }
  },
  {
    "ph": "f", "id": 235379, "pid": 76337, "tid": -914061504, "ts": 1716454225156563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225207805, "dur": 13,
    "args": {
      "External id": 235380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235380,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235380, "pid": 5, "tid": 7, "ts": 1716454225207805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156565, "dur": 11,
    "args": {
      "External id": 235380,
      "cbid": 211, "correlation": 235380
    }
  },
  {
    "ph": "s", "id": 235380, "pid": 76337, "tid": -914061504, "ts": 1716454225156565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225207819, "dur": 5,
    "args": {
      "External id": 235382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235382,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235382, "pid": 5, "tid": 7, "ts": 1716454225207819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156578, "dur": 6,
    "args": {
      "External id": 235382,
      "cbid": 211, "correlation": 235382
    }
  },
  {
    "ph": "s", "id": 235382, "pid": 76337, "tid": -914061504, "ts": 1716454225156578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225207825, "dur": 16,
    "args": {
      "External id": 235392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235392,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235392, "pid": 5, "tid": 7, "ts": 1716454225207825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156636, "dur": 12,
    "args": {
      "External id": 235392,
      "cbid": 211, "correlation": 235392
    }
  },
  {
    "ph": "s", "id": 235392, "pid": 76337, "tid": -914061504, "ts": 1716454225156636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207843, "dur": 17,
    "args": {
      "External id": 235412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235412,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 235412, "pid": 5, "tid": 7, "ts": 1716454225207843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156701, "dur": 11,
    "args": {
      "External id": 235412,
      "cbid": 211, "correlation": 235412
    }
  },
  {
    "ph": "s", "id": 235412, "pid": 76337, "tid": -914061504, "ts": 1716454225156701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225207861, "dur": 4,
    "args": {
      "External id": 235424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235424,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 235424, "pid": 5, "tid": 7, "ts": 1716454225207861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156723, "dur": 6,
    "args": {
      "External id": 235424,
      "cbid": 211, "correlation": 235424
    }
  },
  {
    "ph": "s", "id": 235424, "pid": 76337, "tid": -914061504, "ts": 1716454225156723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225207867, "dur": 17,
    "args": {
      "External id": 235427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235427,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235427, "pid": 5, "tid": 7, "ts": 1716454225207867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156741, "dur": 6,
    "args": {
      "External id": 235427,
      "cbid": 211, "correlation": 235427
    }
  },
  {
    "ph": "s", "id": 235427, "pid": 76337, "tid": -914061504, "ts": 1716454225156741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225207885, "dur": 11,
    "args": {
      "External id": 235436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235436,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235436, "pid": 5, "tid": 7, "ts": 1716454225207885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156781, "dur": 9,
    "args": {
      "External id": 235436,
      "cbid": 211, "correlation": 235436
    }
  },
  {
    "ph": "s", "id": 235436, "pid": 76337, "tid": -914061504, "ts": 1716454225156781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156844, "dur": 0,
    "args": {
      "External id": 235446,
      "cbid": 317, "correlation": 235446
    }
  },
  {
    "ph": "f", "id": 235446, "pid": 76337, "tid": -914061504, "ts": 1716454225156844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156845, "dur": 0,
    "args": {
      "External id": 235447,
      "cbid": 203, "correlation": 235447
    }
  },
  {
    "ph": "f", "id": 235447, "pid": 76337, "tid": -914061504, "ts": 1716454225156845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156846, "dur": 0,
    "args": {
      "External id": 235448,
      "cbid": 205, "correlation": 235448
    }
  },
  {
    "ph": "f", "id": 235448, "pid": 76337, "tid": -914061504, "ts": 1716454225156846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207897, "dur": 11,
    "args": {
      "External id": 235452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235452, "pid": 5, "tid": 7, "ts": 1716454225207897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156863, "dur": 12,
    "args": {
      "External id": 235452,
      "cbid": 211, "correlation": 235452
    }
  },
  {
    "ph": "s", "id": 235452, "pid": 76337, "tid": -914061504, "ts": 1716454225156863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225207909, "dur": 162,
    "args": {
      "External id": 235454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235454, "pid": 5, "tid": 7, "ts": 1716454225207909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156877, "dur": 5,
    "args": {
      "External id": 235454,
      "cbid": 211, "correlation": 235454
    }
  },
  {
    "ph": "s", "id": 235454, "pid": 76337, "tid": -914061504, "ts": 1716454225156877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225208073, "dur": 1,
    "args": {
      "External id": 235456,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 235456,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 235456, "pid": 5, "tid": 7, "ts": 1716454225208073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156889, "dur": 7,
    "args": {
      "External id": 235456,
      "cbid": 51, "correlation": 235456
    }
  },
  {
    "ph": "s", "id": 235456, "pid": 76337, "tid": -914061504, "ts": 1716454225156889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225208077, "dur": 656,
    "args": {
      "External id": 235457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235457,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235457, "pid": 5, "tid": 7, "ts": 1716454225208077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156898, "dur": 7,
    "args": {
      "External id": 235457,
      "cbid": 211, "correlation": 235457
    }
  },
  {
    "ph": "s", "id": 235457, "pid": 76337, "tid": -914061504, "ts": 1716454225156898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225208735, "dur": 12,
    "args": {
      "External id": 235459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235459,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235459, "pid": 5, "tid": 7, "ts": 1716454225208735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156909, "dur": 5,
    "args": {
      "External id": 235459,
      "cbid": 211, "correlation": 235459
    }
  },
  {
    "ph": "s", "id": 235459, "pid": 76337, "tid": -914061504, "ts": 1716454225156909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225208748, "dur": 15,
    "args": {
      "External id": 235465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235465,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235465, "pid": 5, "tid": 7, "ts": 1716454225208748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225156937, "dur": 9,
    "args": {
      "External id": 235465,
      "cbid": 211, "correlation": 235465
    }
  },
  {
    "ph": "s", "id": 235465, "pid": 76337, "tid": -914061504, "ts": 1716454225156937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157005, "dur": 0,
    "args": {
      "External id": 235475,
      "cbid": 317, "correlation": 235475
    }
  },
  {
    "ph": "f", "id": 235475, "pid": 76337, "tid": -914061504, "ts": 1716454225157005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157006, "dur": 0,
    "args": {
      "External id": 235476,
      "cbid": 203, "correlation": 235476
    }
  },
  {
    "ph": "f", "id": 235476, "pid": 76337, "tid": -914061504, "ts": 1716454225157006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157006, "dur": 0,
    "args": {
      "External id": 235477,
      "cbid": 205, "correlation": 235477
    }
  },
  {
    "ph": "f", "id": 235477, "pid": 76337, "tid": -914061504, "ts": 1716454225157006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225208764, "dur": 8,
    "args": {
      "External id": 235481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235481,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235481, "pid": 5, "tid": 7, "ts": 1716454225208764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157020, "dur": 12,
    "args": {
      "External id": 235481,
      "cbid": 211, "correlation": 235481
    }
  },
  {
    "ph": "s", "id": 235481, "pid": 76337, "tid": -914061504, "ts": 1716454225157020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225208774, "dur": 3,
    "args": {
      "External id": 235483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235483, "pid": 5, "tid": 7, "ts": 1716454225208774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157037, "dur": 6,
    "args": {
      "External id": 235483,
      "cbid": 211, "correlation": 235483
    }
  },
  {
    "ph": "s", "id": 235483, "pid": 76337, "tid": -914061504, "ts": 1716454225157037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157046, "dur": 0,
    "args": {
      "External id": 235484,
      "cbid": 51, "correlation": 235484
    }
  },
  {
    "ph": "s", "id": 235484, "pid": 76337, "tid": -914061504, "ts": 1716454225157046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225208778, "dur": 55,
    "args": {
      "External id": 235485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235485,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 235485, "pid": 5, "tid": 7, "ts": 1716454225208778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157047, "dur": 5,
    "args": {
      "External id": 235485,
      "cbid": 211, "correlation": 235485
    }
  },
  {
    "ph": "s", "id": 235485, "pid": 76337, "tid": -914061504, "ts": 1716454225157047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225208835, "dur": 14,
    "args": {
      "External id": 235490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235490, "pid": 5, "tid": 7, "ts": 1716454225208835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157074, "dur": 8,
    "args": {
      "External id": 235490,
      "cbid": 211, "correlation": 235490
    }
  },
  {
    "ph": "s", "id": 235490, "pid": 76337, "tid": -914061504, "ts": 1716454225157074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225208850, "dur": 12,
    "args": {
      "External id": 235498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235498,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235498, "pid": 5, "tid": 7, "ts": 1716454225208850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157104, "dur": 8,
    "args": {
      "External id": 235498,
      "cbid": 211, "correlation": 235498
    }
  },
  {
    "ph": "s", "id": 235498, "pid": 76337, "tid": -914061504, "ts": 1716454225157104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225208863, "dur": 11,
    "args": {
      "External id": 235506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235506,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235506, "pid": 5, "tid": 7, "ts": 1716454225208863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157133, "dur": 8,
    "args": {
      "External id": 235506,
      "cbid": 211, "correlation": 235506
    }
  },
  {
    "ph": "s", "id": 235506, "pid": 76337, "tid": -914061504, "ts": 1716454225157133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225208875, "dur": 18,
    "args": {
      "External id": 235526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235526,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 235526, "pid": 5, "tid": 7, "ts": 1716454225208875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157213, "dur": 12,
    "args": {
      "External id": 235526,
      "cbid": 211, "correlation": 235526
    }
  },
  {
    "ph": "s", "id": 235526, "pid": 76337, "tid": -914061504, "ts": 1716454225157213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225208894, "dur": 5,
    "args": {
      "External id": 235538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235538,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 235538, "pid": 5, "tid": 7, "ts": 1716454225208894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157235, "dur": 6,
    "args": {
      "External id": 235538,
      "cbid": 211, "correlation": 235538
    }
  },
  {
    "ph": "s", "id": 235538, "pid": 76337, "tid": -914061504, "ts": 1716454225157235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225208900, "dur": 17,
    "args": {
      "External id": 235541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235541, "pid": 5, "tid": 7, "ts": 1716454225208900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157253, "dur": 6,
    "args": {
      "External id": 235541,
      "cbid": 211, "correlation": 235541
    }
  },
  {
    "ph": "s", "id": 235541, "pid": 76337, "tid": -914061504, "ts": 1716454225157253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157310, "dur": 0,
    "args": {
      "External id": 235552,
      "cbid": 317, "correlation": 235552
    }
  },
  {
    "ph": "f", "id": 235552, "pid": 76337, "tid": -914061504, "ts": 1716454225157310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157310, "dur": 0,
    "args": {
      "External id": 235553,
      "cbid": 203, "correlation": 235553
    }
  },
  {
    "ph": "f", "id": 235553, "pid": 76337, "tid": -914061504, "ts": 1716454225157310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157311, "dur": 0,
    "args": {
      "External id": 235554,
      "cbid": 205, "correlation": 235554
    }
  },
  {
    "ph": "f", "id": 235554, "pid": 76337, "tid": -914061504, "ts": 1716454225157311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225208919, "dur": 11,
    "args": {
      "External id": 235558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235558, "pid": 5, "tid": 7, "ts": 1716454225208919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157326, "dur": 11,
    "args": {
      "External id": 235558,
      "cbid": 211, "correlation": 235558
    }
  },
  {
    "ph": "s", "id": 235558, "pid": 76337, "tid": -914061504, "ts": 1716454225157326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225208931, "dur": 3,
    "args": {
      "External id": 235560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235560,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235560, "pid": 5, "tid": 7, "ts": 1716454225208931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157343, "dur": 6,
    "args": {
      "External id": 235560,
      "cbid": 211, "correlation": 235560
    }
  },
  {
    "ph": "s", "id": 235560, "pid": 76337, "tid": -914061504, "ts": 1716454225157343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157352, "dur": 0,
    "args": {
      "External id": 235561,
      "cbid": 51, "correlation": 235561
    }
  },
  {
    "ph": "s", "id": 235561, "pid": 76337, "tid": -914061504, "ts": 1716454225157352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225208936, "dur": 95,
    "args": {
      "External id": 235562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235562,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 235562, "pid": 5, "tid": 7, "ts": 1716454225208936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157353, "dur": 5,
    "args": {
      "External id": 235562,
      "cbid": 211, "correlation": 235562
    }
  },
  {
    "ph": "s", "id": 235562, "pid": 76337, "tid": -914061504, "ts": 1716454225157353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225209032, "dur": 15,
    "args": {
      "External id": 235567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235567,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235567, "pid": 5, "tid": 7, "ts": 1716454225209032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157380, "dur": 8,
    "args": {
      "External id": 235567,
      "cbid": 211, "correlation": 235567
    }
  },
  {
    "ph": "s", "id": 235567, "pid": 76337, "tid": -914061504, "ts": 1716454225157380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225209049, "dur": 81,
    "args": {
      "External id": 235576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235576,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235576, "pid": 5, "tid": 7, "ts": 1716454225209049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157462, "dur": 14,
    "args": {
      "External id": 235576,
      "cbid": 211, "correlation": 235576
    }
  },
  {
    "ph": "s", "id": 235576, "pid": 76337, "tid": -914061504, "ts": 1716454225157462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225209131, "dur": 29,
    "args": {
      "External id": 235598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235598,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235598, "pid": 5, "tid": 7, "ts": 1716454225209131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157518, "dur": 10,
    "args": {
      "External id": 235598,
      "cbid": 211, "correlation": 235598
    }
  },
  {
    "ph": "s", "id": 235598, "pid": 76337, "tid": -914061504, "ts": 1716454225157518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157608, "dur": 1,
    "args": {
      "External id": 235609,
      "cbid": 251, "correlation": 235609
    }
  },
  {
    "ph": "f", "id": 235609, "pid": 76337, "tid": -914061504, "ts": 1716454225157608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225209161, "dur": 162,
    "args": {
      "External id": 235610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235610,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235610, "pid": 5, "tid": 7, "ts": 1716454225209161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157614, "dur": 13,
    "args": {
      "External id": 235610,
      "cbid": 211, "correlation": 235610
    }
  },
  {
    "ph": "s", "id": 235610, "pid": 76337, "tid": -914061504, "ts": 1716454225157614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157684, "dur": 1,
    "args": {
      "External id": 235621,
      "cbid": 251, "correlation": 235621
    }
  },
  {
    "ph": "f", "id": 235621, "pid": 76337, "tid": -914061504, "ts": 1716454225157684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225209324, "dur": 157,
    "args": {
      "External id": 235622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235622,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235622, "pid": 5, "tid": 7, "ts": 1716454225209324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157688, "dur": 12,
    "args": {
      "External id": 235622,
      "cbid": 211, "correlation": 235622
    }
  },
  {
    "ph": "s", "id": 235622, "pid": 76337, "tid": -914061504, "ts": 1716454225157688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157753, "dur": 1,
    "args": {
      "External id": 235633,
      "cbid": 251, "correlation": 235633
    }
  },
  {
    "ph": "f", "id": 235633, "pid": 76337, "tid": -914061504, "ts": 1716454225157753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225209482, "dur": 156,
    "args": {
      "External id": 235634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235634,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235634, "pid": 5, "tid": 7, "ts": 1716454225209482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157757, "dur": 11,
    "args": {
      "External id": 235634,
      "cbid": 211, "correlation": 235634
    }
  },
  {
    "ph": "s", "id": 235634, "pid": 76337, "tid": -914061504, "ts": 1716454225157757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225209639, "dur": 334,
    "args": {
      "External id": 235659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235659,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235659, "pid": 5, "tid": 7, "ts": 1716454225209639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157842, "dur": 13,
    "args": {
      "External id": 235659,
      "cbid": 211, "correlation": 235659
    }
  },
  {
    "ph": "s", "id": 235659, "pid": 76337, "tid": -914061504, "ts": 1716454225157842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157942, "dur": 1,
    "args": {
      "External id": 235677,
      "cbid": 251, "correlation": 235677
    }
  },
  {
    "ph": "f", "id": 235677, "pid": 76337, "tid": -914061504, "ts": 1716454225157942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225209975, "dur": 162,
    "args": {
      "External id": 235679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235679,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235679, "pid": 5, "tid": 7, "ts": 1716454225209975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225157948, "dur": 13,
    "args": {
      "External id": 235679,
      "cbid": 211, "correlation": 235679
    }
  },
  {
    "ph": "s", "id": 235679, "pid": 76337, "tid": -914061504, "ts": 1716454225157948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225210138, "dur": 19,
    "args": {
      "External id": 235687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235687,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235687, "pid": 5, "tid": 7, "ts": 1716454225210138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158027, "dur": 13,
    "args": {
      "External id": 235687,
      "cbid": 211, "correlation": 235687
    }
  },
  {
    "ph": "s", "id": 235687, "pid": 76337, "tid": -914061504, "ts": 1716454225158027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225210158, "dur": 27,
    "args": {
      "External id": 235695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235695,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235695, "pid": 5, "tid": 7, "ts": 1716454225210158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158068, "dur": 9,
    "args": {
      "External id": 235695,
      "cbid": 211, "correlation": 235695
    }
  },
  {
    "ph": "s", "id": 235695, "pid": 76337, "tid": -914061504, "ts": 1716454225158068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225210187, "dur": 19,
    "args": {
      "External id": 235706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235706, "pid": 5, "tid": 7, "ts": 1716454225210187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158139, "dur": 13,
    "args": {
      "External id": 235706,
      "cbid": 211, "correlation": 235706
    }
  },
  {
    "ph": "s", "id": 235706, "pid": 76337, "tid": -914061504, "ts": 1716454225158139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225210207, "dur": 15,
    "args": {
      "External id": 235728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235728,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235728, "pid": 5, "tid": 7, "ts": 1716454225210207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158169, "dur": 8,
    "args": {
      "External id": 235728,
      "cbid": 211, "correlation": 235728
    }
  },
  {
    "ph": "s", "id": 235728, "pid": 76337, "tid": -914061504, "ts": 1716454225158169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158255, "dur": 1,
    "args": {
      "External id": 235739,
      "cbid": 251, "correlation": 235739
    }
  },
  {
    "ph": "f", "id": 235739, "pid": 76337, "tid": -914061504, "ts": 1716454225158255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225210224, "dur": 87,
    "args": {
      "External id": 235740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235740,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235740, "pid": 5, "tid": 7, "ts": 1716454225210224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158261, "dur": 14,
    "args": {
      "External id": 235740,
      "cbid": 211, "correlation": 235740
    }
  },
  {
    "ph": "s", "id": 235740, "pid": 76337, "tid": -914061504, "ts": 1716454225158261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158333, "dur": 1,
    "args": {
      "External id": 235751,
      "cbid": 251, "correlation": 235751
    }
  },
  {
    "ph": "f", "id": 235751, "pid": 76337, "tid": -914061504, "ts": 1716454225158333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158337, "dur": 0,
    "args": {
      "External id": 235752,
      "cbid": 251, "correlation": 235752
    }
  },
  {
    "ph": "f", "id": 235752, "pid": 76337, "tid": -914061504, "ts": 1716454225158337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225210313, "dur": 12,
    "args": {
      "External id": 235753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235753,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235753, "pid": 5, "tid": 7, "ts": 1716454225210313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158339, "dur": 12,
    "args": {
      "External id": 235753,
      "cbid": 211, "correlation": 235753
    }
  },
  {
    "ph": "s", "id": 235753, "pid": 76337, "tid": -914061504, "ts": 1716454225158339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225210326, "dur": 5,
    "args": {
      "External id": 235755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235755,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235755, "pid": 5, "tid": 7, "ts": 1716454225210326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158353, "dur": 6,
    "args": {
      "External id": 235755,
      "cbid": 211, "correlation": 235755
    }
  },
  {
    "ph": "s", "id": 235755, "pid": 76337, "tid": -914061504, "ts": 1716454225158353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158410, "dur": 1,
    "args": {
      "External id": 235766,
      "cbid": 251, "correlation": 235766
    }
  },
  {
    "ph": "f", "id": 235766, "pid": 76337, "tid": -914061504, "ts": 1716454225158410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158413, "dur": 0,
    "args": {
      "External id": 235767,
      "cbid": 251, "correlation": 235767
    }
  },
  {
    "ph": "f", "id": 235767, "pid": 76337, "tid": -914061504, "ts": 1716454225158413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225210333, "dur": 8,
    "args": {
      "External id": 235768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235768,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235768, "pid": 5, "tid": 7, "ts": 1716454225210333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158415, "dur": 11,
    "args": {
      "External id": 235768,
      "cbid": 211, "correlation": 235768
    }
  },
  {
    "ph": "s", "id": 235768, "pid": 76337, "tid": -914061504, "ts": 1716454225158415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225210343, "dur": 3,
    "args": {
      "External id": 235770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235770,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235770, "pid": 5, "tid": 7, "ts": 1716454225210343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158428, "dur": 5,
    "args": {
      "External id": 235770,
      "cbid": 211, "correlation": 235770
    }
  },
  {
    "ph": "s", "id": 235770, "pid": 76337, "tid": -914061504, "ts": 1716454225158428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225210347, "dur": 54,
    "args": {
      "External id": 235795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235795,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235795, "pid": 5, "tid": 7, "ts": 1716454225210347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158505, "dur": 12,
    "args": {
      "External id": 235795,
      "cbid": 211, "correlation": 235795
    }
  },
  {
    "ph": "s", "id": 235795, "pid": 76337, "tid": -914061504, "ts": 1716454225158505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158604, "dur": 2,
    "args": {
      "External id": 235813,
      "cbid": 251, "correlation": 235813
    }
  },
  {
    "ph": "f", "id": 235813, "pid": 76337, "tid": -914061504, "ts": 1716454225158604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225210402, "dur": 89,
    "args": {
      "External id": 235815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235815,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235815, "pid": 5, "tid": 7, "ts": 1716454225210402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158610, "dur": 15,
    "args": {
      "External id": 235815,
      "cbid": 211, "correlation": 235815
    }
  },
  {
    "ph": "s", "id": 235815, "pid": 76337, "tid": -914061504, "ts": 1716454225158610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225210492, "dur": 9,
    "args": {
      "External id": 235823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235823,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235823, "pid": 5, "tid": 7, "ts": 1716454225210492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158681, "dur": 13,
    "args": {
      "External id": 235823,
      "cbid": 211, "correlation": 235823
    }
  },
  {
    "ph": "s", "id": 235823, "pid": 76337, "tid": -914061504, "ts": 1716454225158681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225210503, "dur": 21,
    "args": {
      "External id": 235831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235831,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235831, "pid": 5, "tid": 7, "ts": 1716454225210503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158722, "dur": 9,
    "args": {
      "External id": 235831,
      "cbid": 211, "correlation": 235831
    }
  },
  {
    "ph": "s", "id": 235831, "pid": 76337, "tid": -914061504, "ts": 1716454225158722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225210525, "dur": 17,
    "args": {
      "External id": 235853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235853,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235853, "pid": 5, "tid": 7, "ts": 1716454225210525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158775, "dur": 11,
    "args": {
      "External id": 235853,
      "cbid": 211, "correlation": 235853
    }
  },
  {
    "ph": "s", "id": 235853, "pid": 76337, "tid": -914061504, "ts": 1716454225158775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158865, "dur": 2,
    "args": {
      "External id": 235869,
      "cbid": 251, "correlation": 235869
    }
  },
  {
    "ph": "f", "id": 235869, "pid": 76337, "tid": -914061504, "ts": 1716454225158865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158871, "dur": 0,
    "args": {
      "External id": 235871,
      "cbid": 251, "correlation": 235871
    }
  },
  {
    "ph": "f", "id": 235871, "pid": 76337, "tid": -914061504, "ts": 1716454225158871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225210544, "dur": 489,
    "args": {
      "External id": 235872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235872,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 235872, "pid": 5, "tid": 7, "ts": 1716454225210544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158873, "dur": 16,
    "args": {
      "External id": 235872,
      "cbid": 211, "correlation": 235872
    }
  },
  {
    "ph": "s", "id": 235872, "pid": 76337, "tid": -914061504, "ts": 1716454225158873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225211035, "dur": 65,
    "args": {
      "External id": 235880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235880, "pid": 5, "tid": 7, "ts": 1716454225211035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158941, "dur": 12,
    "args": {
      "External id": 235880,
      "cbid": 211, "correlation": 235880
    }
  },
  {
    "ph": "s", "id": 235880, "pid": 76337, "tid": -914061504, "ts": 1716454225158941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225211101, "dur": 68,
    "args": {
      "External id": 235888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235888, "pid": 5, "tid": 7, "ts": 1716454225211101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225158971, "dur": 18,
    "args": {
      "External id": 235888,
      "cbid": 211, "correlation": 235888
    }
  },
  {
    "ph": "s", "id": 235888, "pid": 76337, "tid": -914061504, "ts": 1716454225158971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159061, "dur": 1,
    "args": {
      "External id": 235904,
      "cbid": 251, "correlation": 235904
    }
  },
  {
    "ph": "f", "id": 235904, "pid": 76337, "tid": -914061504, "ts": 1716454225159061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225211171, "dur": 1,
    "args": {
      "External id": 235906,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 235906,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 235906, "pid": 5, "tid": 7, "ts": 1716454225211171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159067, "dur": 12,
    "args": {
      "External id": 235906,
      "cbid": 51, "correlation": 235906
    }
  },
  {
    "ph": "s", "id": 235906, "pid": 76337, "tid": -914061504, "ts": 1716454225159067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225211175, "dur": 266,
    "args": {
      "External id": 235907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235907,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235907, "pid": 5, "tid": 7, "ts": 1716454225211175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159080, "dur": 12,
    "args": {
      "External id": 235907,
      "cbid": 211, "correlation": 235907
    }
  },
  {
    "ph": "s", "id": 235907, "pid": 76337, "tid": -914061504, "ts": 1716454225159080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225211442, "dur": 14,
    "args": {
      "External id": 235915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235915,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235915, "pid": 5, "tid": 7, "ts": 1716454225211442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159122, "dur": 11,
    "args": {
      "External id": 235915,
      "cbid": 211, "correlation": 235915
    }
  },
  {
    "ph": "s", "id": 235915, "pid": 76337, "tid": -914061504, "ts": 1716454225159122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225211457, "dur": 36,
    "args": {
      "External id": 235926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235926, "pid": 5, "tid": 7, "ts": 1716454225211457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159191, "dur": 12,
    "args": {
      "External id": 235926,
      "cbid": 211, "correlation": 235926
    }
  },
  {
    "ph": "s", "id": 235926, "pid": 76337, "tid": -914061504, "ts": 1716454225159191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159256, "dur": 0,
    "args": {
      "External id": 235938,
      "cbid": 317, "correlation": 235938
    }
  },
  {
    "ph": "f", "id": 235938, "pid": 76337, "tid": -914061504, "ts": 1716454225159256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159257, "dur": 0,
    "args": {
      "External id": 235939,
      "cbid": 203, "correlation": 235939
    }
  },
  {
    "ph": "f", "id": 235939, "pid": 76337, "tid": -914061504, "ts": 1716454225159257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159257, "dur": 0,
    "args": {
      "External id": 235940,
      "cbid": 205, "correlation": 235940
    }
  },
  {
    "ph": "f", "id": 235940, "pid": 76337, "tid": -914061504, "ts": 1716454225159257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225211494, "dur": 13,
    "args": {
      "External id": 235944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235944,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235944, "pid": 5, "tid": 7, "ts": 1716454225211494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159272, "dur": 13,
    "args": {
      "External id": 235944,
      "cbid": 211, "correlation": 235944
    }
  },
  {
    "ph": "s", "id": 235944, "pid": 76337, "tid": -914061504, "ts": 1716454225159272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225211509, "dur": 4,
    "args": {
      "External id": 235946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 235946, "pid": 5, "tid": 7, "ts": 1716454225211509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159289, "dur": 6,
    "args": {
      "External id": 235946,
      "cbid": 211, "correlation": 235946
    }
  },
  {
    "ph": "s", "id": 235946, "pid": 76337, "tid": -914061504, "ts": 1716454225159289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159298, "dur": 0,
    "args": {
      "External id": 235947,
      "cbid": 51, "correlation": 235947
    }
  },
  {
    "ph": "s", "id": 235947, "pid": 76337, "tid": -914061504, "ts": 1716454225159298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225211514, "dur": 94,
    "args": {
      "External id": 235948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235948,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 235948, "pid": 5, "tid": 7, "ts": 1716454225211514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159299, "dur": 5,
    "args": {
      "External id": 235948,
      "cbid": 211, "correlation": 235948
    }
  },
  {
    "ph": "s", "id": 235948, "pid": 76337, "tid": -914061504, "ts": 1716454225159299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225211609, "dur": 16,
    "args": {
      "External id": 235953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235953,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235953, "pid": 5, "tid": 7, "ts": 1716454225211609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159325, "dur": 10,
    "args": {
      "External id": 235953,
      "cbid": 211, "correlation": 235953
    }
  },
  {
    "ph": "s", "id": 235953, "pid": 76337, "tid": -914061504, "ts": 1716454225159325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225211626, "dur": 12,
    "args": {
      "External id": 235961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235961,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235961, "pid": 5, "tid": 7, "ts": 1716454225211626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159357, "dur": 8,
    "args": {
      "External id": 235961,
      "cbid": 211, "correlation": 235961
    }
  },
  {
    "ph": "s", "id": 235961, "pid": 76337, "tid": -914061504, "ts": 1716454225159357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225211639, "dur": 18,
    "args": {
      "External id": 235981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235981,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 235981, "pid": 5, "tid": 7, "ts": 1716454225211639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159428, "dur": 12,
    "args": {
      "External id": 235981,
      "cbid": 211, "correlation": 235981
    }
  },
  {
    "ph": "s", "id": 235981, "pid": 76337, "tid": -914061504, "ts": 1716454225159428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225211659, "dur": 5,
    "args": {
      "External id": 235993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235993,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 235993, "pid": 5, "tid": 7, "ts": 1716454225211659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159450, "dur": 6,
    "args": {
      "External id": 235993,
      "cbid": 211, "correlation": 235993
    }
  },
  {
    "ph": "s", "id": 235993, "pid": 76337, "tid": -914061504, "ts": 1716454225159450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225211665, "dur": 18,
    "args": {
      "External id": 235996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 235996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 235996, "pid": 5, "tid": 7, "ts": 1716454225211665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159468, "dur": 6,
    "args": {
      "External id": 235996,
      "cbid": 211, "correlation": 235996
    }
  },
  {
    "ph": "s", "id": 235996, "pid": 76337, "tid": -914061504, "ts": 1716454225159468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225211684, "dur": 11,
    "args": {
      "External id": 236005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236005, "pid": 5, "tid": 7, "ts": 1716454225211684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159506, "dur": 10,
    "args": {
      "External id": 236005,
      "cbid": 211, "correlation": 236005
    }
  },
  {
    "ph": "s", "id": 236005, "pid": 76337, "tid": -914061504, "ts": 1716454225159506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159558, "dur": 0,
    "args": {
      "External id": 236015,
      "cbid": 317, "correlation": 236015
    }
  },
  {
    "ph": "f", "id": 236015, "pid": 76337, "tid": -914061504, "ts": 1716454225159558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159559, "dur": 0,
    "args": {
      "External id": 236016,
      "cbid": 203, "correlation": 236016
    }
  },
  {
    "ph": "f", "id": 236016, "pid": 76337, "tid": -914061504, "ts": 1716454225159559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159559, "dur": 0,
    "args": {
      "External id": 236017,
      "cbid": 205, "correlation": 236017
    }
  },
  {
    "ph": "f", "id": 236017, "pid": 76337, "tid": -914061504, "ts": 1716454225159559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225211697, "dur": 11,
    "args": {
      "External id": 236021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236021, "pid": 5, "tid": 7, "ts": 1716454225211697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159573, "dur": 12,
    "args": {
      "External id": 236021,
      "cbid": 211, "correlation": 236021
    }
  },
  {
    "ph": "s", "id": 236021, "pid": 76337, "tid": -914061504, "ts": 1716454225159573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225211709, "dur": 159,
    "args": {
      "External id": 236023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236023, "pid": 5, "tid": 7, "ts": 1716454225211709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159587, "dur": 5,
    "args": {
      "External id": 236023,
      "cbid": 211, "correlation": 236023
    }
  },
  {
    "ph": "s", "id": 236023, "pid": 76337, "tid": -914061504, "ts": 1716454225159587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225211871, "dur": 1,
    "args": {
      "External id": 236025,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236025,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 236025, "pid": 5, "tid": 7, "ts": 1716454225211871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159598, "dur": 6,
    "args": {
      "External id": 236025,
      "cbid": 51, "correlation": 236025
    }
  },
  {
    "ph": "s", "id": 236025, "pid": 76337, "tid": -914061504, "ts": 1716454225159598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225211874, "dur": 656,
    "args": {
      "External id": 236026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236026,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236026, "pid": 5, "tid": 7, "ts": 1716454225211874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159605, "dur": 6,
    "args": {
      "External id": 236026,
      "cbid": 211, "correlation": 236026
    }
  },
  {
    "ph": "s", "id": 236026, "pid": 76337, "tid": -914061504, "ts": 1716454225159605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225212531, "dur": 13,
    "args": {
      "External id": 236028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236028, "pid": 5, "tid": 7, "ts": 1716454225212531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159616, "dur": 5,
    "args": {
      "External id": 236028,
      "cbid": 211, "correlation": 236028
    }
  },
  {
    "ph": "s", "id": 236028, "pid": 76337, "tid": -914061504, "ts": 1716454225159616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225212546, "dur": 14,
    "args": {
      "External id": 236034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236034, "pid": 5, "tid": 7, "ts": 1716454225212546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159644, "dur": 8,
    "args": {
      "External id": 236034,
      "cbid": 211, "correlation": 236034
    }
  },
  {
    "ph": "s", "id": 236034, "pid": 76337, "tid": -914061504, "ts": 1716454225159644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225212561, "dur": 3,
    "args": {
      "External id": 236042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236042,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 236042, "pid": 5, "tid": 7, "ts": 1716454225212561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159687, "dur": 9,
    "args": {
      "External id": 236042,
      "cbid": 211, "correlation": 236042
    }
  },
  {
    "ph": "s", "id": 236042, "pid": 76337, "tid": -914061504, "ts": 1716454225159687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159751, "dur": 1,
    "args": {
      "External id": 236058,
      "cbid": 251, "correlation": 236058
    }
  },
  {
    "ph": "f", "id": 236058, "pid": 76337, "tid": -914061504, "ts": 1716454225159751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159756, "dur": 0,
    "args": {
      "External id": 236060,
      "cbid": 251, "correlation": 236060
    }
  },
  {
    "ph": "f", "id": 236060, "pid": 76337, "tid": -914061504, "ts": 1716454225159756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225212566, "dur": 13,
    "args": {
      "External id": 236061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236061,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236061, "pid": 5, "tid": 7, "ts": 1716454225212566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159758, "dur": 11,
    "args": {
      "External id": 236061,
      "cbid": 211, "correlation": 236061
    }
  },
  {
    "ph": "s", "id": 236061, "pid": 76337, "tid": -914061504, "ts": 1716454225159758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225212580, "dur": 5,
    "args": {
      "External id": 236063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236063,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236063, "pid": 5, "tid": 7, "ts": 1716454225212580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159771, "dur": 6,
    "args": {
      "External id": 236063,
      "cbid": 211, "correlation": 236063
    }
  },
  {
    "ph": "s", "id": 236063, "pid": 76337, "tid": -914061504, "ts": 1716454225159771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225212587, "dur": 16,
    "args": {
      "External id": 236073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236073, "pid": 5, "tid": 7, "ts": 1716454225212587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159829, "dur": 13,
    "args": {
      "External id": 236073,
      "cbid": 211, "correlation": 236073
    }
  },
  {
    "ph": "s", "id": 236073, "pid": 76337, "tid": -914061504, "ts": 1716454225159829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225212604, "dur": 18,
    "args": {
      "External id": 236093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236093,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 236093, "pid": 5, "tid": 7, "ts": 1716454225212604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159895, "dur": 11,
    "args": {
      "External id": 236093,
      "cbid": 211, "correlation": 236093
    }
  },
  {
    "ph": "s", "id": 236093, "pid": 76337, "tid": -914061504, "ts": 1716454225159895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225212623, "dur": 4,
    "args": {
      "External id": 236105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236105,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 236105, "pid": 5, "tid": 7, "ts": 1716454225212623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159915, "dur": 6,
    "args": {
      "External id": 236105,
      "cbid": 211, "correlation": 236105
    }
  },
  {
    "ph": "s", "id": 236105, "pid": 76337, "tid": -914061504, "ts": 1716454225159915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225212629, "dur": 16,
    "args": {
      "External id": 236108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236108, "pid": 5, "tid": 7, "ts": 1716454225212629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159935, "dur": 7,
    "args": {
      "External id": 236108,
      "cbid": 211, "correlation": 236108
    }
  },
  {
    "ph": "s", "id": 236108, "pid": 76337, "tid": -914061504, "ts": 1716454225159935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225212646, "dur": 10,
    "args": {
      "External id": 236117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236117,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236117, "pid": 5, "tid": 7, "ts": 1716454225212646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225159984, "dur": 10,
    "args": {
      "External id": 236117,
      "cbid": 211, "correlation": 236117
    }
  },
  {
    "ph": "s", "id": 236117, "pid": 76337, "tid": -914061504, "ts": 1716454225159984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160048, "dur": 0,
    "args": {
      "External id": 236127,
      "cbid": 317, "correlation": 236127
    }
  },
  {
    "ph": "f", "id": 236127, "pid": 76337, "tid": -914061504, "ts": 1716454225160048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160048, "dur": 0,
    "args": {
      "External id": 236128,
      "cbid": 203, "correlation": 236128
    }
  },
  {
    "ph": "f", "id": 236128, "pid": 76337, "tid": -914061504, "ts": 1716454225160048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160049, "dur": 0,
    "args": {
      "External id": 236129,
      "cbid": 205, "correlation": 236129
    }
  },
  {
    "ph": "f", "id": 236129, "pid": 76337, "tid": -914061504, "ts": 1716454225160049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225212658, "dur": 11,
    "args": {
      "External id": 236133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236133,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236133, "pid": 5, "tid": 7, "ts": 1716454225212658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160063, "dur": 12,
    "args": {
      "External id": 236133,
      "cbid": 211, "correlation": 236133
    }
  },
  {
    "ph": "s", "id": 236133, "pid": 76337, "tid": -914061504, "ts": 1716454225160063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225212670, "dur": 159,
    "args": {
      "External id": 236135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236135,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236135, "pid": 5, "tid": 7, "ts": 1716454225212670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160077, "dur": 6,
    "args": {
      "External id": 236135,
      "cbid": 211, "correlation": 236135
    }
  },
  {
    "ph": "s", "id": 236135, "pid": 76337, "tid": -914061504, "ts": 1716454225160077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225212832, "dur": 1,
    "args": {
      "External id": 236137,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236137,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 236137, "pid": 5, "tid": 7, "ts": 1716454225212832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160089, "dur": 7,
    "args": {
      "External id": 236137,
      "cbid": 51, "correlation": 236137
    }
  },
  {
    "ph": "s", "id": 236137, "pid": 76337, "tid": -914061504, "ts": 1716454225160089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225212835, "dur": 642,
    "args": {
      "External id": 236138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236138,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236138, "pid": 5, "tid": 7, "ts": 1716454225212835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160097, "dur": 6,
    "args": {
      "External id": 236138,
      "cbid": 211, "correlation": 236138
    }
  },
  {
    "ph": "s", "id": 236138, "pid": 76337, "tid": -914061504, "ts": 1716454225160097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225213479, "dur": 13,
    "args": {
      "External id": 236140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236140,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236140, "pid": 5, "tid": 7, "ts": 1716454225213479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160107, "dur": 5,
    "args": {
      "External id": 236140,
      "cbid": 211, "correlation": 236140
    }
  },
  {
    "ph": "s", "id": 236140, "pid": 76337, "tid": -914061504, "ts": 1716454225160107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225213493, "dur": 14,
    "args": {
      "External id": 236146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236146, "pid": 5, "tid": 7, "ts": 1716454225213493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160136, "dur": 8,
    "args": {
      "External id": 236146,
      "cbid": 211, "correlation": 236146
    }
  },
  {
    "ph": "s", "id": 236146, "pid": 76337, "tid": -914061504, "ts": 1716454225160136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225213508, "dur": 11,
    "args": {
      "External id": 236154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236154,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236154, "pid": 5, "tid": 7, "ts": 1716454225213508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160168, "dur": 9,
    "args": {
      "External id": 236154,
      "cbid": 211, "correlation": 236154
    }
  },
  {
    "ph": "s", "id": 236154, "pid": 76337, "tid": -914061504, "ts": 1716454225160168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225213521, "dur": 10,
    "args": {
      "External id": 236162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236162,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236162, "pid": 5, "tid": 7, "ts": 1716454225213521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160197, "dur": 8,
    "args": {
      "External id": 236162,
      "cbid": 211, "correlation": 236162
    }
  },
  {
    "ph": "s", "id": 236162, "pid": 76337, "tid": -914061504, "ts": 1716454225160197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225213532, "dur": 18,
    "args": {
      "External id": 236182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236182,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 236182, "pid": 5, "tid": 7, "ts": 1716454225213532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160275, "dur": 12,
    "args": {
      "External id": 236182,
      "cbid": 211, "correlation": 236182
    }
  },
  {
    "ph": "s", "id": 236182, "pid": 76337, "tid": -914061504, "ts": 1716454225160275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225213552, "dur": 4,
    "args": {
      "External id": 236194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236194,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 236194, "pid": 5, "tid": 7, "ts": 1716454225213552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160297, "dur": 6,
    "args": {
      "External id": 236194,
      "cbid": 211, "correlation": 236194
    }
  },
  {
    "ph": "s", "id": 236194, "pid": 76337, "tid": -914061504, "ts": 1716454225160297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225213557, "dur": 16,
    "args": {
      "External id": 236197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236197, "pid": 5, "tid": 7, "ts": 1716454225213557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160315, "dur": 6,
    "args": {
      "External id": 236197,
      "cbid": 211, "correlation": 236197
    }
  },
  {
    "ph": "s", "id": 236197, "pid": 76337, "tid": -914061504, "ts": 1716454225160315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160372, "dur": 0,
    "args": {
      "External id": 236208,
      "cbid": 317, "correlation": 236208
    }
  },
  {
    "ph": "f", "id": 236208, "pid": 76337, "tid": -914061504, "ts": 1716454225160372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160373, "dur": 0,
    "args": {
      "External id": 236209,
      "cbid": 203, "correlation": 236209
    }
  },
  {
    "ph": "f", "id": 236209, "pid": 76337, "tid": -914061504, "ts": 1716454225160373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160374, "dur": 0,
    "args": {
      "External id": 236210,
      "cbid": 205, "correlation": 236210
    }
  },
  {
    "ph": "f", "id": 236210, "pid": 76337, "tid": -914061504, "ts": 1716454225160374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225213574, "dur": 13,
    "args": {
      "External id": 236214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236214,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236214, "pid": 5, "tid": 7, "ts": 1716454225213574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160388, "dur": 11,
    "args": {
      "External id": 236214,
      "cbid": 211, "correlation": 236214
    }
  },
  {
    "ph": "s", "id": 236214, "pid": 76337, "tid": -914061504, "ts": 1716454225160388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225213588, "dur": 4,
    "args": {
      "External id": 236216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236216,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 236216, "pid": 5, "tid": 7, "ts": 1716454225213588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160404, "dur": 6,
    "args": {
      "External id": 236216,
      "cbid": 211, "correlation": 236216
    }
  },
  {
    "ph": "s", "id": 236216, "pid": 76337, "tid": -914061504, "ts": 1716454225160404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160412, "dur": 0,
    "args": {
      "External id": 236217,
      "cbid": 51, "correlation": 236217
    }
  },
  {
    "ph": "s", "id": 236217, "pid": 76337, "tid": -914061504, "ts": 1716454225160412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225213593, "dur": 92,
    "args": {
      "External id": 236218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236218,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 236218, "pid": 5, "tid": 7, "ts": 1716454225213593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160413, "dur": 5,
    "args": {
      "External id": 236218,
      "cbid": 211, "correlation": 236218
    }
  },
  {
    "ph": "s", "id": 236218, "pid": 76337, "tid": -914061504, "ts": 1716454225160413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225213686, "dur": 15,
    "args": {
      "External id": 236223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236223, "pid": 5, "tid": 7, "ts": 1716454225213686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160440, "dur": 8,
    "args": {
      "External id": 236223,
      "cbid": 211, "correlation": 236223
    }
  },
  {
    "ph": "s", "id": 236223, "pid": 76337, "tid": -914061504, "ts": 1716454225160440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225213702, "dur": 81,
    "args": {
      "External id": 236232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236232,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236232, "pid": 5, "tid": 7, "ts": 1716454225213702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160521, "dur": 15,
    "args": {
      "External id": 236232,
      "cbid": 211, "correlation": 236232
    }
  },
  {
    "ph": "s", "id": 236232, "pid": 76337, "tid": -914061504, "ts": 1716454225160521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225213785, "dur": 30,
    "args": {
      "External id": 236254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236254,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236254, "pid": 5, "tid": 7, "ts": 1716454225213785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160578, "dur": 10,
    "args": {
      "External id": 236254,
      "cbid": 211, "correlation": 236254
    }
  },
  {
    "ph": "s", "id": 236254, "pid": 76337, "tid": -914061504, "ts": 1716454225160578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160664, "dur": 1,
    "args": {
      "External id": 236265,
      "cbid": 251, "correlation": 236265
    }
  },
  {
    "ph": "f", "id": 236265, "pid": 76337, "tid": -914061504, "ts": 1716454225160664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225213816, "dur": 161,
    "args": {
      "External id": 236266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236266,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236266, "pid": 5, "tid": 7, "ts": 1716454225213816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160670, "dur": 14,
    "args": {
      "External id": 236266,
      "cbid": 211, "correlation": 236266
    }
  },
  {
    "ph": "s", "id": 236266, "pid": 76337, "tid": -914061504, "ts": 1716454225160670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160740, "dur": 1,
    "args": {
      "External id": 236277,
      "cbid": 251, "correlation": 236277
    }
  },
  {
    "ph": "f", "id": 236277, "pid": 76337, "tid": -914061504, "ts": 1716454225160740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225213979, "dur": 154,
    "args": {
      "External id": 236278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236278,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236278, "pid": 5, "tid": 7, "ts": 1716454225213979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160744, "dur": 11,
    "args": {
      "External id": 236278,
      "cbid": 211, "correlation": 236278
    }
  },
  {
    "ph": "s", "id": 236278, "pid": 76337, "tid": -914061504, "ts": 1716454225160744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160809, "dur": 1,
    "args": {
      "External id": 236289,
      "cbid": 251, "correlation": 236289
    }
  },
  {
    "ph": "f", "id": 236289, "pid": 76337, "tid": -914061504, "ts": 1716454225160809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225214134, "dur": 157,
    "args": {
      "External id": 236290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236290,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236290, "pid": 5, "tid": 7, "ts": 1716454225214134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160813, "dur": 11,
    "args": {
      "External id": 236290,
      "cbid": 211, "correlation": 236290
    }
  },
  {
    "ph": "s", "id": 236290, "pid": 76337, "tid": -914061504, "ts": 1716454225160813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225214292, "dur": 332,
    "args": {
      "External id": 236315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236315,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236315, "pid": 5, "tid": 7, "ts": 1716454225214292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225160897, "dur": 12,
    "args": {
      "External id": 236315,
      "cbid": 211, "correlation": 236315
    }
  },
  {
    "ph": "s", "id": 236315, "pid": 76337, "tid": -914061504, "ts": 1716454225160897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161004, "dur": 1,
    "args": {
      "External id": 236333,
      "cbid": 251, "correlation": 236333
    }
  },
  {
    "ph": "f", "id": 236333, "pid": 76337, "tid": -914061504, "ts": 1716454225161004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225214626, "dur": 162,
    "args": {
      "External id": 236335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236335,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236335, "pid": 5, "tid": 7, "ts": 1716454225214626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161010, "dur": 14,
    "args": {
      "External id": 236335,
      "cbid": 211, "correlation": 236335
    }
  },
  {
    "ph": "s", "id": 236335, "pid": 76337, "tid": -914061504, "ts": 1716454225161010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225214790, "dur": 19,
    "args": {
      "External id": 236343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236343,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236343, "pid": 5, "tid": 7, "ts": 1716454225214790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161082, "dur": 12,
    "args": {
      "External id": 236343,
      "cbid": 211, "correlation": 236343
    }
  },
  {
    "ph": "s", "id": 236343, "pid": 76337, "tid": -914061504, "ts": 1716454225161082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225214810, "dur": 28,
    "args": {
      "External id": 236351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236351,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236351, "pid": 5, "tid": 7, "ts": 1716454225214810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161121, "dur": 9,
    "args": {
      "External id": 236351,
      "cbid": 211, "correlation": 236351
    }
  },
  {
    "ph": "s", "id": 236351, "pid": 76337, "tid": -914061504, "ts": 1716454225161121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225214839, "dur": 19,
    "args": {
      "External id": 236362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236362, "pid": 5, "tid": 7, "ts": 1716454225214839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161192, "dur": 12,
    "args": {
      "External id": 236362,
      "cbid": 211, "correlation": 236362
    }
  },
  {
    "ph": "s", "id": 236362, "pid": 76337, "tid": -914061504, "ts": 1716454225161192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225214860, "dur": 16,
    "args": {
      "External id": 236384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236384,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236384, "pid": 5, "tid": 7, "ts": 1716454225214860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161223, "dur": 8,
    "args": {
      "External id": 236384,
      "cbid": 211, "correlation": 236384
    }
  },
  {
    "ph": "s", "id": 236384, "pid": 76337, "tid": -914061504, "ts": 1716454225161223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161308, "dur": 1,
    "args": {
      "External id": 236395,
      "cbid": 251, "correlation": 236395
    }
  },
  {
    "ph": "f", "id": 236395, "pid": 76337, "tid": -914061504, "ts": 1716454225161308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225214877, "dur": 87,
    "args": {
      "External id": 236396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236396,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 236396, "pid": 5, "tid": 7, "ts": 1716454225214877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161313, "dur": 13,
    "args": {
      "External id": 236396,
      "cbid": 211, "correlation": 236396
    }
  },
  {
    "ph": "s", "id": 236396, "pid": 76337, "tid": -914061504, "ts": 1716454225161313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161382, "dur": 1,
    "args": {
      "External id": 236407,
      "cbid": 251, "correlation": 236407
    }
  },
  {
    "ph": "f", "id": 236407, "pid": 76337, "tid": -914061504, "ts": 1716454225161382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161386, "dur": 0,
    "args": {
      "External id": 236408,
      "cbid": 251, "correlation": 236408
    }
  },
  {
    "ph": "f", "id": 236408, "pid": 76337, "tid": -914061504, "ts": 1716454225161386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225214965, "dur": 12,
    "args": {
      "External id": 236409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236409,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236409, "pid": 5, "tid": 7, "ts": 1716454225214965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161387, "dur": 12,
    "args": {
      "External id": 236409,
      "cbid": 211, "correlation": 236409
    }
  },
  {
    "ph": "s", "id": 236409, "pid": 76337, "tid": -914061504, "ts": 1716454225161387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225214979, "dur": 6,
    "args": {
      "External id": 236411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236411,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236411, "pid": 5, "tid": 7, "ts": 1716454225214979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161401, "dur": 6,
    "args": {
      "External id": 236411,
      "cbid": 211, "correlation": 236411
    }
  },
  {
    "ph": "s", "id": 236411, "pid": 76337, "tid": -914061504, "ts": 1716454225161401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161458, "dur": 1,
    "args": {
      "External id": 236422,
      "cbid": 251, "correlation": 236422
    }
  },
  {
    "ph": "f", "id": 236422, "pid": 76337, "tid": -914061504, "ts": 1716454225161458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161462, "dur": 0,
    "args": {
      "External id": 236423,
      "cbid": 251, "correlation": 236423
    }
  },
  {
    "ph": "f", "id": 236423, "pid": 76337, "tid": -914061504, "ts": 1716454225161462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225214986, "dur": 8,
    "args": {
      "External id": 236424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236424,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236424, "pid": 5, "tid": 7, "ts": 1716454225214986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161463, "dur": 12,
    "args": {
      "External id": 236424,
      "cbid": 211, "correlation": 236424
    }
  },
  {
    "ph": "s", "id": 236424, "pid": 76337, "tid": -914061504, "ts": 1716454225161463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225214995, "dur": 3,
    "args": {
      "External id": 236426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236426,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236426, "pid": 5, "tid": 7, "ts": 1716454225214995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161477, "dur": 5,
    "args": {
      "External id": 236426,
      "cbid": 211, "correlation": 236426
    }
  },
  {
    "ph": "s", "id": 236426, "pid": 76337, "tid": -914061504, "ts": 1716454225161477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225215000, "dur": 54,
    "args": {
      "External id": 236451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236451,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236451, "pid": 5, "tid": 7, "ts": 1716454225215000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161553, "dur": 12,
    "args": {
      "External id": 236451,
      "cbid": 211, "correlation": 236451
    }
  },
  {
    "ph": "s", "id": 236451, "pid": 76337, "tid": -914061504, "ts": 1716454225161553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161651, "dur": 1,
    "args": {
      "External id": 236469,
      "cbid": 251, "correlation": 236469
    }
  },
  {
    "ph": "f", "id": 236469, "pid": 76337, "tid": -914061504, "ts": 1716454225161651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225215055, "dur": 89,
    "args": {
      "External id": 236471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236471,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 236471, "pid": 5, "tid": 7, "ts": 1716454225215055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161657, "dur": 14,
    "args": {
      "External id": 236471,
      "cbid": 211, "correlation": 236471
    }
  },
  {
    "ph": "s", "id": 236471, "pid": 76337, "tid": -914061504, "ts": 1716454225161657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225215145, "dur": 9,
    "args": {
      "External id": 236479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236479,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236479, "pid": 5, "tid": 7, "ts": 1716454225215145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161728, "dur": 12,
    "args": {
      "External id": 236479,
      "cbid": 211, "correlation": 236479
    }
  },
  {
    "ph": "s", "id": 236479, "pid": 76337, "tid": -914061504, "ts": 1716454225161728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225215156, "dur": 21,
    "args": {
      "External id": 236487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236487,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236487, "pid": 5, "tid": 7, "ts": 1716454225215156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161768, "dur": 10,
    "args": {
      "External id": 236487,
      "cbid": 211, "correlation": 236487
    }
  },
  {
    "ph": "s", "id": 236487, "pid": 76337, "tid": -914061504, "ts": 1716454225161768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225215178, "dur": 17,
    "args": {
      "External id": 236509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236509,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236509, "pid": 5, "tid": 7, "ts": 1716454225215178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161818, "dur": 10,
    "args": {
      "External id": 236509,
      "cbid": 211, "correlation": 236509
    }
  },
  {
    "ph": "s", "id": 236509, "pid": 76337, "tid": -914061504, "ts": 1716454225161818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161905, "dur": 1,
    "args": {
      "External id": 236525,
      "cbid": 251, "correlation": 236525
    }
  },
  {
    "ph": "f", "id": 236525, "pid": 76337, "tid": -914061504, "ts": 1716454225161905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161910, "dur": 0,
    "args": {
      "External id": 236527,
      "cbid": 251, "correlation": 236527
    }
  },
  {
    "ph": "f", "id": 236527, "pid": 76337, "tid": -914061504, "ts": 1716454225161910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225215196, "dur": 487,
    "args": {
      "External id": 236528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236528,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236528, "pid": 5, "tid": 7, "ts": 1716454225215196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161911, "dur": 14,
    "args": {
      "External id": 236528,
      "cbid": 211, "correlation": 236528
    }
  },
  {
    "ph": "s", "id": 236528, "pid": 76337, "tid": -914061504, "ts": 1716454225161911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225215685, "dur": 64,
    "args": {
      "External id": 236536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236536, "pid": 5, "tid": 7, "ts": 1716454225215685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225161985, "dur": 13,
    "args": {
      "External id": 236536,
      "cbid": 211, "correlation": 236536
    }
  },
  {
    "ph": "s", "id": 236536, "pid": 76337, "tid": -914061504, "ts": 1716454225161985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225215750, "dur": 67,
    "args": {
      "External id": 236544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236544,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236544, "pid": 5, "tid": 7, "ts": 1716454225215750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162016, "dur": 8,
    "args": {
      "External id": 236544,
      "cbid": 211, "correlation": 236544
    }
  },
  {
    "ph": "s", "id": 236544, "pid": 76337, "tid": -914061504, "ts": 1716454225162016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162096, "dur": 1,
    "args": {
      "External id": 236560,
      "cbid": 251, "correlation": 236560
    }
  },
  {
    "ph": "f", "id": 236560, "pid": 76337, "tid": -914061504, "ts": 1716454225162096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225215820, "dur": 1,
    "args": {
      "External id": 236562,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236562,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 236562, "pid": 5, "tid": 7, "ts": 1716454225215820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162101, "dur": 9,
    "args": {
      "External id": 236562,
      "cbid": 51, "correlation": 236562
    }
  },
  {
    "ph": "s", "id": 236562, "pid": 76337, "tid": -914061504, "ts": 1716454225162101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225215823, "dur": 268,
    "args": {
      "External id": 236563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236563,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 236563, "pid": 5, "tid": 7, "ts": 1716454225215823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162111, "dur": 11,
    "args": {
      "External id": 236563,
      "cbid": 211, "correlation": 236563
    }
  },
  {
    "ph": "s", "id": 236563, "pid": 76337, "tid": -914061504, "ts": 1716454225162111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225216093, "dur": 14,
    "args": {
      "External id": 236571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236571,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236571, "pid": 5, "tid": 7, "ts": 1716454225216093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162153, "dur": 10,
    "args": {
      "External id": 236571,
      "cbid": 211, "correlation": 236571
    }
  },
  {
    "ph": "s", "id": 236571, "pid": 76337, "tid": -914061504, "ts": 1716454225162153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225216108, "dur": 36,
    "args": {
      "External id": 236582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236582, "pid": 5, "tid": 7, "ts": 1716454225216108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162220, "dur": 13,
    "args": {
      "External id": 236582,
      "cbid": 211, "correlation": 236582
    }
  },
  {
    "ph": "s", "id": 236582, "pid": 76337, "tid": -914061504, "ts": 1716454225162220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162286, "dur": 0,
    "args": {
      "External id": 236594,
      "cbid": 317, "correlation": 236594
    }
  },
  {
    "ph": "f", "id": 236594, "pid": 76337, "tid": -914061504, "ts": 1716454225162286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162287, "dur": 0,
    "args": {
      "External id": 236595,
      "cbid": 203, "correlation": 236595
    }
  },
  {
    "ph": "f", "id": 236595, "pid": 76337, "tid": -914061504, "ts": 1716454225162287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162287, "dur": 0,
    "args": {
      "External id": 236596,
      "cbid": 205, "correlation": 236596
    }
  },
  {
    "ph": "f", "id": 236596, "pid": 76337, "tid": -914061504, "ts": 1716454225162287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216146, "dur": 13,
    "args": {
      "External id": 236600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236600,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236600, "pid": 5, "tid": 7, "ts": 1716454225216146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162302, "dur": 12,
    "args": {
      "External id": 236600,
      "cbid": 211, "correlation": 236600
    }
  },
  {
    "ph": "s", "id": 236600, "pid": 76337, "tid": -914061504, "ts": 1716454225162302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225216160, "dur": 4,
    "args": {
      "External id": 236602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 236602, "pid": 5, "tid": 7, "ts": 1716454225216160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162318, "dur": 6,
    "args": {
      "External id": 236602,
      "cbid": 211, "correlation": 236602
    }
  },
  {
    "ph": "s", "id": 236602, "pid": 76337, "tid": -914061504, "ts": 1716454225162318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162328, "dur": 0,
    "args": {
      "External id": 236603,
      "cbid": 51, "correlation": 236603
    }
  },
  {
    "ph": "s", "id": 236603, "pid": 76337, "tid": -914061504, "ts": 1716454225162328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225216165, "dur": 95,
    "args": {
      "External id": 236604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236604,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 236604, "pid": 5, "tid": 7, "ts": 1716454225216165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162328, "dur": 5,
    "args": {
      "External id": 236604,
      "cbid": 211, "correlation": 236604
    }
  },
  {
    "ph": "s", "id": 236604, "pid": 76337, "tid": -914061504, "ts": 1716454225162328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225216261, "dur": 16,
    "args": {
      "External id": 236609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236609, "pid": 5, "tid": 7, "ts": 1716454225216261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162355, "dur": 9,
    "args": {
      "External id": 236609,
      "cbid": 211, "correlation": 236609
    }
  },
  {
    "ph": "s", "id": 236609, "pid": 76337, "tid": -914061504, "ts": 1716454225162355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225216279, "dur": 12,
    "args": {
      "External id": 236617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236617,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236617, "pid": 5, "tid": 7, "ts": 1716454225216279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162387, "dur": 8,
    "args": {
      "External id": 236617,
      "cbid": 211, "correlation": 236617
    }
  },
  {
    "ph": "s", "id": 236617, "pid": 76337, "tid": -914061504, "ts": 1716454225162387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162455, "dur": 0,
    "args": {
      "External id": 236627,
      "cbid": 317, "correlation": 236627
    }
  },
  {
    "ph": "f", "id": 236627, "pid": 76337, "tid": -914061504, "ts": 1716454225162455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162456, "dur": 0,
    "args": {
      "External id": 236628,
      "cbid": 203, "correlation": 236628
    }
  },
  {
    "ph": "f", "id": 236628, "pid": 76337, "tid": -914061504, "ts": 1716454225162456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162456, "dur": 0,
    "args": {
      "External id": 236629,
      "cbid": 205, "correlation": 236629
    }
  },
  {
    "ph": "f", "id": 236629, "pid": 76337, "tid": -914061504, "ts": 1716454225162456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216292, "dur": 11,
    "args": {
      "External id": 236633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236633,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236633, "pid": 5, "tid": 7, "ts": 1716454225216292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162472, "dur": 12,
    "args": {
      "External id": 236633,
      "cbid": 211, "correlation": 236633
    }
  },
  {
    "ph": "s", "id": 236633, "pid": 76337, "tid": -914061504, "ts": 1716454225162472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216305, "dur": 161,
    "args": {
      "External id": 236635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236635,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236635, "pid": 5, "tid": 7, "ts": 1716454225216305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162486, "dur": 5,
    "args": {
      "External id": 236635,
      "cbid": 211, "correlation": 236635
    }
  },
  {
    "ph": "s", "id": 236635, "pid": 76337, "tid": -914061504, "ts": 1716454225162486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225216468, "dur": 1,
    "args": {
      "External id": 236637,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236637,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 236637, "pid": 5, "tid": 7, "ts": 1716454225216468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162497, "dur": 6,
    "args": {
      "External id": 236637,
      "cbid": 51, "correlation": 236637
    }
  },
  {
    "ph": "s", "id": 236637, "pid": 76337, "tid": -914061504, "ts": 1716454225162497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225216471, "dur": 196,
    "args": {
      "External id": 236638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236638,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 236638, "pid": 5, "tid": 7, "ts": 1716454225216471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162505, "dur": 8,
    "args": {
      "External id": 236638,
      "cbid": 211, "correlation": 236638
    }
  },
  {
    "ph": "s", "id": 236638, "pid": 76337, "tid": -914061504, "ts": 1716454225162505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216669, "dur": 6,
    "args": {
      "External id": 236640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236640,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236640, "pid": 5, "tid": 7, "ts": 1716454225216669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162517, "dur": 6,
    "args": {
      "External id": 236640,
      "cbid": 211, "correlation": 236640
    }
  },
  {
    "ph": "s", "id": 236640, "pid": 76337, "tid": -914061504, "ts": 1716454225162517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225216676, "dur": 6,
    "args": {
      "External id": 236646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236646, "pid": 5, "tid": 7, "ts": 1716454225216676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162546, "dur": 8,
    "args": {
      "External id": 236646,
      "cbid": 211, "correlation": 236646
    }
  },
  {
    "ph": "s", "id": 236646, "pid": 76337, "tid": -914061504, "ts": 1716454225162546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216683, "dur": 10,
    "args": {
      "External id": 236666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236666,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 236666, "pid": 5, "tid": 7, "ts": 1716454225216683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162638, "dur": 11,
    "args": {
      "External id": 236666,
      "cbid": 211, "correlation": 236666
    }
  },
  {
    "ph": "s", "id": 236666, "pid": 76337, "tid": -914061504, "ts": 1716454225162638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225216695, "dur": 4,
    "args": {
      "External id": 236678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236678,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 236678, "pid": 5, "tid": 7, "ts": 1716454225216695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162660, "dur": 7,
    "args": {
      "External id": 236678,
      "cbid": 211, "correlation": 236678
    }
  },
  {
    "ph": "s", "id": 236678, "pid": 76337, "tid": -914061504, "ts": 1716454225162660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225216700, "dur": 8,
    "args": {
      "External id": 236681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236681,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236681, "pid": 5, "tid": 7, "ts": 1716454225216700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162680, "dur": 6,
    "args": {
      "External id": 236681,
      "cbid": 211, "correlation": 236681
    }
  },
  {
    "ph": "s", "id": 236681, "pid": 76337, "tid": -914061504, "ts": 1716454225162680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225216710, "dur": 5,
    "args": {
      "External id": 236690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236690,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236690, "pid": 5, "tid": 7, "ts": 1716454225216710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162720, "dur": 10,
    "args": {
      "External id": 236690,
      "cbid": 211, "correlation": 236690
    }
  },
  {
    "ph": "s", "id": 236690, "pid": 76337, "tid": -914061504, "ts": 1716454225162720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162773, "dur": 0,
    "args": {
      "External id": 236700,
      "cbid": 317, "correlation": 236700
    }
  },
  {
    "ph": "f", "id": 236700, "pid": 76337, "tid": -914061504, "ts": 1716454225162773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162774, "dur": 0,
    "args": {
      "External id": 236701,
      "cbid": 203, "correlation": 236701
    }
  },
  {
    "ph": "f", "id": 236701, "pid": 76337, "tid": -914061504, "ts": 1716454225162774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162775, "dur": 0,
    "args": {
      "External id": 236702,
      "cbid": 205, "correlation": 236702
    }
  },
  {
    "ph": "f", "id": 236702, "pid": 76337, "tid": -914061504, "ts": 1716454225162775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216716, "dur": 5,
    "args": {
      "External id": 236706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236706, "pid": 5, "tid": 7, "ts": 1716454225216716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162790, "dur": 11,
    "args": {
      "External id": 236706,
      "cbid": 211, "correlation": 236706
    }
  },
  {
    "ph": "s", "id": 236706, "pid": 76337, "tid": -914061504, "ts": 1716454225162790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225216722, "dur": 159,
    "args": {
      "External id": 236708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236708,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236708, "pid": 5, "tid": 7, "ts": 1716454225216722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162803, "dur": 5,
    "args": {
      "External id": 236708,
      "cbid": 211, "correlation": 236708
    }
  },
  {
    "ph": "s", "id": 236708, "pid": 76337, "tid": -914061504, "ts": 1716454225162803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225216884, "dur": 1,
    "args": {
      "External id": 236710,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236710,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 236710, "pid": 5, "tid": 7, "ts": 1716454225216884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162814, "dur": 7,
    "args": {
      "External id": 236710,
      "cbid": 51, "correlation": 236710
    }
  },
  {
    "ph": "s", "id": 236710, "pid": 76337, "tid": -914061504, "ts": 1716454225162814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225216887, "dur": 266,
    "args": {
      "External id": 236711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236711,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236711, "pid": 5, "tid": 7, "ts": 1716454225216887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162822, "dur": 6,
    "args": {
      "External id": 236711,
      "cbid": 211, "correlation": 236711
    }
  },
  {
    "ph": "s", "id": 236711, "pid": 76337, "tid": -914061504, "ts": 1716454225162822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217155, "dur": 6,
    "args": {
      "External id": 236713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236713,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236713, "pid": 5, "tid": 7, "ts": 1716454225217155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162833, "dur": 5,
    "args": {
      "External id": 236713,
      "cbid": 211, "correlation": 236713
    }
  },
  {
    "ph": "s", "id": 236713, "pid": 76337, "tid": -914061504, "ts": 1716454225162833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225217161, "dur": 6,
    "args": {
      "External id": 236719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236719,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236719, "pid": 5, "tid": 7, "ts": 1716454225217161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162861, "dur": 9,
    "args": {
      "External id": 236719,
      "cbid": 211, "correlation": 236719
    }
  },
  {
    "ph": "s", "id": 236719, "pid": 76337, "tid": -914061504, "ts": 1716454225162861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225217169, "dur": 3,
    "args": {
      "External id": 236727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236727,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 236727, "pid": 5, "tid": 7, "ts": 1716454225217169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162905, "dur": 9,
    "args": {
      "External id": 236727,
      "cbid": 211, "correlation": 236727
    }
  },
  {
    "ph": "s", "id": 236727, "pid": 76337, "tid": -914061504, "ts": 1716454225162905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162970, "dur": 1,
    "args": {
      "External id": 236743,
      "cbid": 251, "correlation": 236743
    }
  },
  {
    "ph": "f", "id": 236743, "pid": 76337, "tid": -914061504, "ts": 1716454225162970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162983, "dur": 0,
    "args": {
      "External id": 236745,
      "cbid": 251, "correlation": 236745
    }
  },
  {
    "ph": "f", "id": 236745, "pid": 76337, "tid": -914061504, "ts": 1716454225162983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225217173, "dur": 13,
    "args": {
      "External id": 236746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236746,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236746, "pid": 5, "tid": 7, "ts": 1716454225217173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162985, "dur": 11,
    "args": {
      "External id": 236746,
      "cbid": 211, "correlation": 236746
    }
  },
  {
    "ph": "s", "id": 236746, "pid": 76337, "tid": -914061504, "ts": 1716454225162985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225217188, "dur": 5,
    "args": {
      "External id": 236748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236748,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236748, "pid": 5, "tid": 7, "ts": 1716454225217188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225162999, "dur": 6,
    "args": {
      "External id": 236748,
      "cbid": 211, "correlation": 236748
    }
  },
  {
    "ph": "s", "id": 236748, "pid": 76337, "tid": -914061504, "ts": 1716454225162999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225217194, "dur": 6,
    "args": {
      "External id": 236758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236758,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236758, "pid": 5, "tid": 7, "ts": 1716454225217194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163057, "dur": 13,
    "args": {
      "External id": 236758,
      "cbid": 211, "correlation": 236758
    }
  },
  {
    "ph": "s", "id": 236758, "pid": 76337, "tid": -914061504, "ts": 1716454225163057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217201, "dur": 10,
    "args": {
      "External id": 236778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236778,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 236778, "pid": 5, "tid": 7, "ts": 1716454225217201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163125, "dur": 10,
    "args": {
      "External id": 236778,
      "cbid": 211, "correlation": 236778
    }
  },
  {
    "ph": "s", "id": 236778, "pid": 76337, "tid": -914061504, "ts": 1716454225163125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225217212, "dur": 4,
    "args": {
      "External id": 236790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236790,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 236790, "pid": 5, "tid": 7, "ts": 1716454225217212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163145, "dur": 6,
    "args": {
      "External id": 236790,
      "cbid": 211, "correlation": 236790
    }
  },
  {
    "ph": "s", "id": 236790, "pid": 76337, "tid": -914061504, "ts": 1716454225163145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225217217, "dur": 7,
    "args": {
      "External id": 236793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236793,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236793, "pid": 5, "tid": 7, "ts": 1716454225217217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163164, "dur": 7,
    "args": {
      "External id": 236793,
      "cbid": 211, "correlation": 236793
    }
  },
  {
    "ph": "s", "id": 236793, "pid": 76337, "tid": -914061504, "ts": 1716454225163164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225217225, "dur": 4,
    "args": {
      "External id": 236802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236802,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236802, "pid": 5, "tid": 7, "ts": 1716454225217225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163203, "dur": 10,
    "args": {
      "External id": 236802,
      "cbid": 211, "correlation": 236802
    }
  },
  {
    "ph": "s", "id": 236802, "pid": 76337, "tid": -914061504, "ts": 1716454225163203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163267, "dur": 0,
    "args": {
      "External id": 236812,
      "cbid": 317, "correlation": 236812
    }
  },
  {
    "ph": "f", "id": 236812, "pid": 76337, "tid": -914061504, "ts": 1716454225163267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163268, "dur": 0,
    "args": {
      "External id": 236813,
      "cbid": 203, "correlation": 236813
    }
  },
  {
    "ph": "f", "id": 236813, "pid": 76337, "tid": -914061504, "ts": 1716454225163268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163268, "dur": 0,
    "args": {
      "External id": 236814,
      "cbid": 205, "correlation": 236814
    }
  },
  {
    "ph": "f", "id": 236814, "pid": 76337, "tid": -914061504, "ts": 1716454225163268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217231, "dur": 5,
    "args": {
      "External id": 236818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236818,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236818, "pid": 5, "tid": 7, "ts": 1716454225217231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163282, "dur": 12,
    "args": {
      "External id": 236818,
      "cbid": 211, "correlation": 236818
    }
  },
  {
    "ph": "s", "id": 236818, "pid": 76337, "tid": -914061504, "ts": 1716454225163282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217237, "dur": 159,
    "args": {
      "External id": 236820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236820,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236820, "pid": 5, "tid": 7, "ts": 1716454225217237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163296, "dur": 5,
    "args": {
      "External id": 236820,
      "cbid": 211, "correlation": 236820
    }
  },
  {
    "ph": "s", "id": 236820, "pid": 76337, "tid": -914061504, "ts": 1716454225163296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225217398, "dur": 1,
    "args": {
      "External id": 236822,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236822,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 236822, "pid": 5, "tid": 7, "ts": 1716454225217398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163307, "dur": 7,
    "args": {
      "External id": 236822,
      "cbid": 51, "correlation": 236822
    }
  },
  {
    "ph": "s", "id": 236822, "pid": 76337, "tid": -914061504, "ts": 1716454225163307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225217402, "dur": 254,
    "args": {
      "External id": 236823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236823,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236823, "pid": 5, "tid": 7, "ts": 1716454225217402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163315, "dur": 6,
    "args": {
      "External id": 236823,
      "cbid": 211, "correlation": 236823
    }
  },
  {
    "ph": "s", "id": 236823, "pid": 76337, "tid": -914061504, "ts": 1716454225163315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217658, "dur": 6,
    "args": {
      "External id": 236825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236825,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236825, "pid": 5, "tid": 7, "ts": 1716454225217658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163325, "dur": 5,
    "args": {
      "External id": 236825,
      "cbid": 211, "correlation": 236825
    }
  },
  {
    "ph": "s", "id": 236825, "pid": 76337, "tid": -914061504, "ts": 1716454225163325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225217664, "dur": 6,
    "args": {
      "External id": 236831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236831,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236831, "pid": 5, "tid": 7, "ts": 1716454225217664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163353, "dur": 9,
    "args": {
      "External id": 236831,
      "cbid": 211, "correlation": 236831
    }
  },
  {
    "ph": "s", "id": 236831, "pid": 76337, "tid": -914061504, "ts": 1716454225163353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225217672, "dur": 5,
    "args": {
      "External id": 236839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236839,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236839, "pid": 5, "tid": 7, "ts": 1716454225217672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163385, "dur": 8,
    "args": {
      "External id": 236839,
      "cbid": 211, "correlation": 236839
    }
  },
  {
    "ph": "s", "id": 236839, "pid": 76337, "tid": -914061504, "ts": 1716454225163385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225217678, "dur": 4,
    "args": {
      "External id": 236847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236847,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236847, "pid": 5, "tid": 7, "ts": 1716454225217678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163416, "dur": 8,
    "args": {
      "External id": 236847,
      "cbid": 211, "correlation": 236847
    }
  },
  {
    "ph": "s", "id": 236847, "pid": 76337, "tid": -914061504, "ts": 1716454225163416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217684, "dur": 9,
    "args": {
      "External id": 236867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236867,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 236867, "pid": 5, "tid": 7, "ts": 1716454225217684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163488, "dur": 11,
    "args": {
      "External id": 236867,
      "cbid": 211, "correlation": 236867
    }
  },
  {
    "ph": "s", "id": 236867, "pid": 76337, "tid": -914061504, "ts": 1716454225163488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225217695, "dur": 4,
    "args": {
      "External id": 236879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236879,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 236879, "pid": 5, "tid": 7, "ts": 1716454225217695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163509, "dur": 7,
    "args": {
      "External id": 236879,
      "cbid": 211, "correlation": 236879
    }
  },
  {
    "ph": "s", "id": 236879, "pid": 76337, "tid": -914061504, "ts": 1716454225163509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225217700, "dur": 6,
    "args": {
      "External id": 236882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236882, "pid": 5, "tid": 7, "ts": 1716454225217700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163527, "dur": 7,
    "args": {
      "External id": 236882,
      "cbid": 211, "correlation": 236882
    }
  },
  {
    "ph": "s", "id": 236882, "pid": 76337, "tid": -914061504, "ts": 1716454225163527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225217708, "dur": 4,
    "args": {
      "External id": 236891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236891,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236891, "pid": 5, "tid": 7, "ts": 1716454225217708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163566, "dur": 9,
    "args": {
      "External id": 236891,
      "cbid": 211, "correlation": 236891
    }
  },
  {
    "ph": "s", "id": 236891, "pid": 76337, "tid": -914061504, "ts": 1716454225163566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163616, "dur": 0,
    "args": {
      "External id": 236901,
      "cbid": 317, "correlation": 236901
    }
  },
  {
    "ph": "f", "id": 236901, "pid": 76337, "tid": -914061504, "ts": 1716454225163616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163617, "dur": 0,
    "args": {
      "External id": 236902,
      "cbid": 203, "correlation": 236902
    }
  },
  {
    "ph": "f", "id": 236902, "pid": 76337, "tid": -914061504, "ts": 1716454225163617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163618, "dur": 0,
    "args": {
      "External id": 236903,
      "cbid": 205, "correlation": 236903
    }
  },
  {
    "ph": "f", "id": 236903, "pid": 76337, "tid": -914061504, "ts": 1716454225163618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217713, "dur": 5,
    "args": {
      "External id": 236907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236907,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236907, "pid": 5, "tid": 7, "ts": 1716454225217713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163631, "dur": 12,
    "args": {
      "External id": 236907,
      "cbid": 211, "correlation": 236907
    }
  },
  {
    "ph": "s", "id": 236907, "pid": 76337, "tid": -914061504, "ts": 1716454225163631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225217720, "dur": 159,
    "args": {
      "External id": 236909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236909, "pid": 5, "tid": 7, "ts": 1716454225217720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163645, "dur": 5,
    "args": {
      "External id": 236909,
      "cbid": 211, "correlation": 236909
    }
  },
  {
    "ph": "s", "id": 236909, "pid": 76337, "tid": -914061504, "ts": 1716454225163645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225217881, "dur": 1,
    "args": {
      "External id": 236911,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 236911,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 236911, "pid": 5, "tid": 7, "ts": 1716454225217881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163655, "dur": 7,
    "args": {
      "External id": 236911,
      "cbid": 51, "correlation": 236911
    }
  },
  {
    "ph": "s", "id": 236911, "pid": 76337, "tid": -914061504, "ts": 1716454225163655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225217884, "dur": 254,
    "args": {
      "External id": 236912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236912,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236912, "pid": 5, "tid": 7, "ts": 1716454225217884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163663, "dur": 6,
    "args": {
      "External id": 236912,
      "cbid": 211, "correlation": 236912
    }
  },
  {
    "ph": "s", "id": 236912, "pid": 76337, "tid": -914061504, "ts": 1716454225163663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218140, "dur": 5,
    "args": {
      "External id": 236914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 236914, "pid": 5, "tid": 7, "ts": 1716454225218140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163672, "dur": 5,
    "args": {
      "External id": 236914,
      "cbid": 211, "correlation": 236914
    }
  },
  {
    "ph": "s", "id": 236914, "pid": 76337, "tid": -914061504, "ts": 1716454225163672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225218146, "dur": 6,
    "args": {
      "External id": 236920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236920,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236920, "pid": 5, "tid": 7, "ts": 1716454225218146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163700, "dur": 8,
    "args": {
      "External id": 236920,
      "cbid": 211, "correlation": 236920
    }
  },
  {
    "ph": "s", "id": 236920, "pid": 76337, "tid": -914061504, "ts": 1716454225163700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225218154, "dur": 3,
    "args": {
      "External id": 236928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236928,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 236928, "pid": 5, "tid": 7, "ts": 1716454225218154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163744, "dur": 9,
    "args": {
      "External id": 236928,
      "cbid": 211, "correlation": 236928
    }
  },
  {
    "ph": "s", "id": 236928, "pid": 76337, "tid": -914061504, "ts": 1716454225163744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163806, "dur": 1,
    "args": {
      "External id": 236944,
      "cbid": 251, "correlation": 236944
    }
  },
  {
    "ph": "f", "id": 236944, "pid": 76337, "tid": -914061504, "ts": 1716454225163806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163812, "dur": 0,
    "args": {
      "External id": 236946,
      "cbid": 251, "correlation": 236946
    }
  },
  {
    "ph": "f", "id": 236946, "pid": 76337, "tid": -914061504, "ts": 1716454225163812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225218158, "dur": 10,
    "args": {
      "External id": 236947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236947,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236947, "pid": 5, "tid": 7, "ts": 1716454225218158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163814, "dur": 11,
    "args": {
      "External id": 236947,
      "cbid": 211, "correlation": 236947
    }
  },
  {
    "ph": "s", "id": 236947, "pid": 76337, "tid": -914061504, "ts": 1716454225163814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225218169, "dur": 4,
    "args": {
      "External id": 236949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236949,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 236949, "pid": 5, "tid": 7, "ts": 1716454225218169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163827, "dur": 5,
    "args": {
      "External id": 236949,
      "cbid": 211, "correlation": 236949
    }
  },
  {
    "ph": "s", "id": 236949, "pid": 76337, "tid": -914061504, "ts": 1716454225163827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225218174, "dur": 6,
    "args": {
      "External id": 236959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236959, "pid": 5, "tid": 7, "ts": 1716454225218174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163882, "dur": 12,
    "args": {
      "External id": 236959,
      "cbid": 211, "correlation": 236959
    }
  },
  {
    "ph": "s", "id": 236959, "pid": 76337, "tid": -914061504, "ts": 1716454225163882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218181, "dur": 10,
    "args": {
      "External id": 236979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236979,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 236979, "pid": 5, "tid": 7, "ts": 1716454225218181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163947, "dur": 10,
    "args": {
      "External id": 236979,
      "cbid": 211, "correlation": 236979
    }
  },
  {
    "ph": "s", "id": 236979, "pid": 76337, "tid": -914061504, "ts": 1716454225163947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225218192, "dur": 4,
    "args": {
      "External id": 236991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236991,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 236991, "pid": 5, "tid": 7, "ts": 1716454225218192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163968, "dur": 14,
    "args": {
      "External id": 236991,
      "cbid": 211, "correlation": 236991
    }
  },
  {
    "ph": "s", "id": 236991, "pid": 76337, "tid": -914061504, "ts": 1716454225163968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225218197, "dur": 7,
    "args": {
      "External id": 236994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 236994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 236994, "pid": 5, "tid": 7, "ts": 1716454225218197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225163996, "dur": 7,
    "args": {
      "External id": 236994,
      "cbid": 211, "correlation": 236994
    }
  },
  {
    "ph": "s", "id": 236994, "pid": 76337, "tid": -914061504, "ts": 1716454225163996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225218205, "dur": 4,
    "args": {
      "External id": 237003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237003,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237003, "pid": 5, "tid": 7, "ts": 1716454225218205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164038, "dur": 9,
    "args": {
      "External id": 237003,
      "cbid": 211, "correlation": 237003
    }
  },
  {
    "ph": "s", "id": 237003, "pid": 76337, "tid": -914061504, "ts": 1716454225164038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164099, "dur": 0,
    "args": {
      "External id": 237013,
      "cbid": 317, "correlation": 237013
    }
  },
  {
    "ph": "f", "id": 237013, "pid": 76337, "tid": -914061504, "ts": 1716454225164099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164100, "dur": 0,
    "args": {
      "External id": 237014,
      "cbid": 203, "correlation": 237014
    }
  },
  {
    "ph": "f", "id": 237014, "pid": 76337, "tid": -914061504, "ts": 1716454225164100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164101, "dur": 0,
    "args": {
      "External id": 237015,
      "cbid": 205, "correlation": 237015
    }
  },
  {
    "ph": "f", "id": 237015, "pid": 76337, "tid": -914061504, "ts": 1716454225164101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218211, "dur": 5,
    "args": {
      "External id": 237019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237019,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237019, "pid": 5, "tid": 7, "ts": 1716454225218211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164116, "dur": 12,
    "args": {
      "External id": 237019,
      "cbid": 211, "correlation": 237019
    }
  },
  {
    "ph": "s", "id": 237019, "pid": 76337, "tid": -914061504, "ts": 1716454225164116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218217, "dur": 159,
    "args": {
      "External id": 237021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237021, "pid": 5, "tid": 7, "ts": 1716454225218217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164130, "dur": 5,
    "args": {
      "External id": 237021,
      "cbid": 211, "correlation": 237021
    }
  },
  {
    "ph": "s", "id": 237021, "pid": 76337, "tid": -914061504, "ts": 1716454225164130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225218378, "dur": 1,
    "args": {
      "External id": 237023,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237023,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 237023, "pid": 5, "tid": 7, "ts": 1716454225218378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164140, "dur": 7,
    "args": {
      "External id": 237023,
      "cbid": 51, "correlation": 237023
    }
  },
  {
    "ph": "s", "id": 237023, "pid": 76337, "tid": -914061504, "ts": 1716454225164140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225218382, "dur": 254,
    "args": {
      "External id": 237024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237024,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237024, "pid": 5, "tid": 7, "ts": 1716454225218382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164148, "dur": 6,
    "args": {
      "External id": 237024,
      "cbid": 211, "correlation": 237024
    }
  },
  {
    "ph": "s", "id": 237024, "pid": 76337, "tid": -914061504, "ts": 1716454225164148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218637, "dur": 5,
    "args": {
      "External id": 237026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237026, "pid": 5, "tid": 7, "ts": 1716454225218637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164158, "dur": 6,
    "args": {
      "External id": 237026,
      "cbid": 211, "correlation": 237026
    }
  },
  {
    "ph": "s", "id": 237026, "pid": 76337, "tid": -914061504, "ts": 1716454225164158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225218644, "dur": 6,
    "args": {
      "External id": 237032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237032, "pid": 5, "tid": 7, "ts": 1716454225218644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164187, "dur": 8,
    "args": {
      "External id": 237032,
      "cbid": 211, "correlation": 237032
    }
  },
  {
    "ph": "s", "id": 237032, "pid": 76337, "tid": -914061504, "ts": 1716454225164187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225218651, "dur": 5,
    "args": {
      "External id": 237040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237040,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237040, "pid": 5, "tid": 7, "ts": 1716454225218651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164220, "dur": 8,
    "args": {
      "External id": 237040,
      "cbid": 211, "correlation": 237040
    }
  },
  {
    "ph": "s", "id": 237040, "pid": 76337, "tid": -914061504, "ts": 1716454225164220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225218657, "dur": 4,
    "args": {
      "External id": 237048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237048,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237048, "pid": 5, "tid": 7, "ts": 1716454225218657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164250, "dur": 8,
    "args": {
      "External id": 237048,
      "cbid": 211, "correlation": 237048
    }
  },
  {
    "ph": "s", "id": 237048, "pid": 76337, "tid": -914061504, "ts": 1716454225164250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218663, "dur": 9,
    "args": {
      "External id": 237068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237068,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 237068, "pid": 5, "tid": 7, "ts": 1716454225218663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164353, "dur": 13,
    "args": {
      "External id": 237068,
      "cbid": 211, "correlation": 237068
    }
  },
  {
    "ph": "s", "id": 237068, "pid": 76337, "tid": -914061504, "ts": 1716454225164353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225218673, "dur": 4,
    "args": {
      "External id": 237080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237080,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 237080, "pid": 5, "tid": 7, "ts": 1716454225218673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164376, "dur": 6,
    "args": {
      "External id": 237080,
      "cbid": 211, "correlation": 237080
    }
  },
  {
    "ph": "s", "id": 237080, "pid": 76337, "tid": -914061504, "ts": 1716454225164376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225218678, "dur": 6,
    "args": {
      "External id": 237083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237083,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237083, "pid": 5, "tid": 7, "ts": 1716454225218678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164394, "dur": 6,
    "args": {
      "External id": 237083,
      "cbid": 211, "correlation": 237083
    }
  },
  {
    "ph": "s", "id": 237083, "pid": 76337, "tid": -914061504, "ts": 1716454225164394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225218686, "dur": 4,
    "args": {
      "External id": 237092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237092,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237092, "pid": 5, "tid": 7, "ts": 1716454225218686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164432, "dur": 9,
    "args": {
      "External id": 237092,
      "cbid": 211, "correlation": 237092
    }
  },
  {
    "ph": "s", "id": 237092, "pid": 76337, "tid": -914061504, "ts": 1716454225164432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164484, "dur": 0,
    "args": {
      "External id": 237102,
      "cbid": 317, "correlation": 237102
    }
  },
  {
    "ph": "f", "id": 237102, "pid": 76337, "tid": -914061504, "ts": 1716454225164484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164484, "dur": 0,
    "args": {
      "External id": 237103,
      "cbid": 203, "correlation": 237103
    }
  },
  {
    "ph": "f", "id": 237103, "pid": 76337, "tid": -914061504, "ts": 1716454225164484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164485, "dur": 0,
    "args": {
      "External id": 237104,
      "cbid": 205, "correlation": 237104
    }
  },
  {
    "ph": "f", "id": 237104, "pid": 76337, "tid": -914061504, "ts": 1716454225164485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218692, "dur": 5,
    "args": {
      "External id": 237108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237108,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237108, "pid": 5, "tid": 7, "ts": 1716454225218692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164498, "dur": 12,
    "args": {
      "External id": 237108,
      "cbid": 211, "correlation": 237108
    }
  },
  {
    "ph": "s", "id": 237108, "pid": 76337, "tid": -914061504, "ts": 1716454225164498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225218698, "dur": 159,
    "args": {
      "External id": 237110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237110, "pid": 5, "tid": 7, "ts": 1716454225218698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164513, "dur": 5,
    "args": {
      "External id": 237110,
      "cbid": 211, "correlation": 237110
    }
  },
  {
    "ph": "s", "id": 237110, "pid": 76337, "tid": -914061504, "ts": 1716454225164513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225218859, "dur": 1,
    "args": {
      "External id": 237112,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237112,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 237112, "pid": 5, "tid": 7, "ts": 1716454225218859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164523, "dur": 6,
    "args": {
      "External id": 237112,
      "cbid": 51, "correlation": 237112
    }
  },
  {
    "ph": "s", "id": 237112, "pid": 76337, "tid": -914061504, "ts": 1716454225164523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225218862, "dur": 254,
    "args": {
      "External id": 237113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237113,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237113, "pid": 5, "tid": 7, "ts": 1716454225218862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164530, "dur": 6,
    "args": {
      "External id": 237113,
      "cbid": 211, "correlation": 237113
    }
  },
  {
    "ph": "s", "id": 237113, "pid": 76337, "tid": -914061504, "ts": 1716454225164530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219118, "dur": 6,
    "args": {
      "External id": 237115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237115,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237115, "pid": 5, "tid": 7, "ts": 1716454225219118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164540, "dur": 5,
    "args": {
      "External id": 237115,
      "cbid": 211, "correlation": 237115
    }
  },
  {
    "ph": "s", "id": 237115, "pid": 76337, "tid": -914061504, "ts": 1716454225164540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225219125, "dur": 6,
    "args": {
      "External id": 237121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237121, "pid": 5, "tid": 7, "ts": 1716454225219125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164568, "dur": 8,
    "args": {
      "External id": 237121,
      "cbid": 211, "correlation": 237121
    }
  },
  {
    "ph": "s", "id": 237121, "pid": 76337, "tid": -914061504, "ts": 1716454225164568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225219132, "dur": 3,
    "args": {
      "External id": 237129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237129,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 237129, "pid": 5, "tid": 7, "ts": 1716454225219132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164612, "dur": 9,
    "args": {
      "External id": 237129,
      "cbid": 211, "correlation": 237129
    }
  },
  {
    "ph": "s", "id": 237129, "pid": 76337, "tid": -914061504, "ts": 1716454225164612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164674, "dur": 1,
    "args": {
      "External id": 237145,
      "cbid": 251, "correlation": 237145
    }
  },
  {
    "ph": "f", "id": 237145, "pid": 76337, "tid": -914061504, "ts": 1716454225164674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164679, "dur": 0,
    "args": {
      "External id": 237147,
      "cbid": 251, "correlation": 237147
    }
  },
  {
    "ph": "f", "id": 237147, "pid": 76337, "tid": -914061504, "ts": 1716454225164679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225219137, "dur": 10,
    "args": {
      "External id": 237148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237148,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237148, "pid": 5, "tid": 7, "ts": 1716454225219137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164681, "dur": 11,
    "args": {
      "External id": 237148,
      "cbid": 211, "correlation": 237148
    }
  },
  {
    "ph": "s", "id": 237148, "pid": 76337, "tid": -914061504, "ts": 1716454225164681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225219148, "dur": 3,
    "args": {
      "External id": 237150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237150,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237150, "pid": 5, "tid": 7, "ts": 1716454225219148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164694, "dur": 5,
    "args": {
      "External id": 237150,
      "cbid": 211, "correlation": 237150
    }
  },
  {
    "ph": "s", "id": 237150, "pid": 76337, "tid": -914061504, "ts": 1716454225164694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225219153, "dur": 5,
    "args": {
      "External id": 237160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237160,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237160, "pid": 5, "tid": 7, "ts": 1716454225219153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164750, "dur": 13,
    "args": {
      "External id": 237160,
      "cbid": 211, "correlation": 237160
    }
  },
  {
    "ph": "s", "id": 237160, "pid": 76337, "tid": -914061504, "ts": 1716454225164750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219160, "dur": 9,
    "args": {
      "External id": 237180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237180,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 237180, "pid": 5, "tid": 7, "ts": 1716454225219160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164816, "dur": 11,
    "args": {
      "External id": 237180,
      "cbid": 211, "correlation": 237180
    }
  },
  {
    "ph": "s", "id": 237180, "pid": 76337, "tid": -914061504, "ts": 1716454225164816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225219170, "dur": 4,
    "args": {
      "External id": 237192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237192,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 237192, "pid": 5, "tid": 7, "ts": 1716454225219170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164836, "dur": 6,
    "args": {
      "External id": 237192,
      "cbid": 211, "correlation": 237192
    }
  },
  {
    "ph": "s", "id": 237192, "pid": 76337, "tid": -914061504, "ts": 1716454225164836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225219175, "dur": 7,
    "args": {
      "External id": 237195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237195,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237195, "pid": 5, "tid": 7, "ts": 1716454225219175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164855, "dur": 7,
    "args": {
      "External id": 237195,
      "cbid": 211, "correlation": 237195
    }
  },
  {
    "ph": "s", "id": 237195, "pid": 76337, "tid": -914061504, "ts": 1716454225164855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225219184, "dur": 4,
    "args": {
      "External id": 237204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237204,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237204, "pid": 5, "tid": 7, "ts": 1716454225219184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164894, "dur": 9,
    "args": {
      "External id": 237204,
      "cbid": 211, "correlation": 237204
    }
  },
  {
    "ph": "s", "id": 237204, "pid": 76337, "tid": -914061504, "ts": 1716454225164894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164957, "dur": 0,
    "args": {
      "External id": 237214,
      "cbid": 317, "correlation": 237214
    }
  },
  {
    "ph": "f", "id": 237214, "pid": 76337, "tid": -914061504, "ts": 1716454225164957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164958, "dur": 0,
    "args": {
      "External id": 237215,
      "cbid": 203, "correlation": 237215
    }
  },
  {
    "ph": "f", "id": 237215, "pid": 76337, "tid": -914061504, "ts": 1716454225164958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164958, "dur": 0,
    "args": {
      "External id": 237216,
      "cbid": 205, "correlation": 237216
    }
  },
  {
    "ph": "f", "id": 237216, "pid": 76337, "tid": -914061504, "ts": 1716454225164958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219189, "dur": 5,
    "args": {
      "External id": 237220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237220, "pid": 5, "tid": 7, "ts": 1716454225219189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164972, "dur": 21,
    "args": {
      "External id": 237220,
      "cbid": 211, "correlation": 237220
    }
  },
  {
    "ph": "s", "id": 237220, "pid": 76337, "tid": -914061504, "ts": 1716454225164972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219196, "dur": 159,
    "args": {
      "External id": 237222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237222,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237222, "pid": 5, "tid": 7, "ts": 1716454225219196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225164996, "dur": 6,
    "args": {
      "External id": 237222,
      "cbid": 211, "correlation": 237222
    }
  },
  {
    "ph": "s", "id": 237222, "pid": 76337, "tid": -914061504, "ts": 1716454225164996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225219357, "dur": 1,
    "args": {
      "External id": 237224,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237224,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 237224, "pid": 5, "tid": 7, "ts": 1716454225219357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165008, "dur": 7,
    "args": {
      "External id": 237224,
      "cbid": 51, "correlation": 237224
    }
  },
  {
    "ph": "s", "id": 237224, "pid": 76337, "tid": -914061504, "ts": 1716454225165008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225219361, "dur": 254,
    "args": {
      "External id": 237225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237225,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237225, "pid": 5, "tid": 7, "ts": 1716454225219361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165016, "dur": 6,
    "args": {
      "External id": 237225,
      "cbid": 211, "correlation": 237225
    }
  },
  {
    "ph": "s", "id": 237225, "pid": 76337, "tid": -914061504, "ts": 1716454225165016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219616, "dur": 6,
    "args": {
      "External id": 237227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237227,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237227, "pid": 5, "tid": 7, "ts": 1716454225219616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165025, "dur": 5,
    "args": {
      "External id": 237227,
      "cbid": 211, "correlation": 237227
    }
  },
  {
    "ph": "s", "id": 237227, "pid": 76337, "tid": -914061504, "ts": 1716454225165025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225219623, "dur": 6,
    "args": {
      "External id": 237233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237233,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237233, "pid": 5, "tid": 7, "ts": 1716454225219623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165054, "dur": 10,
    "args": {
      "External id": 237233,
      "cbid": 211, "correlation": 237233
    }
  },
  {
    "ph": "s", "id": 237233, "pid": 76337, "tid": -914061504, "ts": 1716454225165054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225219630, "dur": 5,
    "args": {
      "External id": 237241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237241,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237241, "pid": 5, "tid": 7, "ts": 1716454225219630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165087, "dur": 8,
    "args": {
      "External id": 237241,
      "cbid": 211, "correlation": 237241
    }
  },
  {
    "ph": "s", "id": 237241, "pid": 76337, "tid": -914061504, "ts": 1716454225165087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225219636, "dur": 4,
    "args": {
      "External id": 237249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237249,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237249, "pid": 5, "tid": 7, "ts": 1716454225219636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165117, "dur": 8,
    "args": {
      "External id": 237249,
      "cbid": 211, "correlation": 237249
    }
  },
  {
    "ph": "s", "id": 237249, "pid": 76337, "tid": -914061504, "ts": 1716454225165117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219642, "dur": 9,
    "args": {
      "External id": 237269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237269,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 237269, "pid": 5, "tid": 7, "ts": 1716454225219642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165258, "dur": 13,
    "args": {
      "External id": 237269,
      "cbid": 211, "correlation": 237269
    }
  },
  {
    "ph": "s", "id": 237269, "pid": 76337, "tid": -914061504, "ts": 1716454225165258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225219653, "dur": 4,
    "args": {
      "External id": 237281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237281,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 237281, "pid": 5, "tid": 7, "ts": 1716454225219653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165281, "dur": 6,
    "args": {
      "External id": 237281,
      "cbid": 211, "correlation": 237281
    }
  },
  {
    "ph": "s", "id": 237281, "pid": 76337, "tid": -914061504, "ts": 1716454225165281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225219658, "dur": 7,
    "args": {
      "External id": 237284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237284, "pid": 5, "tid": 7, "ts": 1716454225219658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165299, "dur": 8,
    "args": {
      "External id": 237284,
      "cbid": 211, "correlation": 237284
    }
  },
  {
    "ph": "s", "id": 237284, "pid": 76337, "tid": -914061504, "ts": 1716454225165299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165359, "dur": 0,
    "args": {
      "External id": 237295,
      "cbid": 317, "correlation": 237295
    }
  },
  {
    "ph": "f", "id": 237295, "pid": 76337, "tid": -914061504, "ts": 1716454225165359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165360, "dur": 0,
    "args": {
      "External id": 237296,
      "cbid": 203, "correlation": 237296
    }
  },
  {
    "ph": "f", "id": 237296, "pid": 76337, "tid": -914061504, "ts": 1716454225165360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165360, "dur": 0,
    "args": {
      "External id": 237297,
      "cbid": 205, "correlation": 237297
    }
  },
  {
    "ph": "f", "id": 237297, "pid": 76337, "tid": -914061504, "ts": 1716454225165360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219665, "dur": 5,
    "args": {
      "External id": 237301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237301,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237301, "pid": 5, "tid": 7, "ts": 1716454225219665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165376, "dur": 12,
    "args": {
      "External id": 237301,
      "cbid": 211, "correlation": 237301
    }
  },
  {
    "ph": "s", "id": 237301, "pid": 76337, "tid": -914061504, "ts": 1716454225165376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225219672, "dur": 36,
    "args": {
      "External id": 237303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237303,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 237303, "pid": 5, "tid": 7, "ts": 1716454225219672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165396, "dur": 9,
    "args": {
      "External id": 237303,
      "cbid": 211, "correlation": 237303
    }
  },
  {
    "ph": "s", "id": 237303, "pid": 76337, "tid": -914061504, "ts": 1716454225165396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219709, "dur": 5,
    "args": {
      "External id": 237305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237305, "pid": 5, "tid": 7, "ts": 1716454225219709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165409, "dur": 5,
    "args": {
      "External id": 237305,
      "cbid": 211, "correlation": 237305
    }
  },
  {
    "ph": "s", "id": 237305, "pid": 76337, "tid": -914061504, "ts": 1716454225165409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225219715, "dur": 6,
    "args": {
      "External id": 237311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237311, "pid": 5, "tid": 7, "ts": 1716454225219715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165436, "dur": 8,
    "args": {
      "External id": 237311,
      "cbid": 211, "correlation": 237311
    }
  },
  {
    "ph": "s", "id": 237311, "pid": 76337, "tid": -914061504, "ts": 1716454225165436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225219723, "dur": 20,
    "args": {
      "External id": 237320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237320,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237320, "pid": 5, "tid": 7, "ts": 1716454225219723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165519, "dur": 14,
    "args": {
      "External id": 237320,
      "cbid": 211, "correlation": 237320
    }
  },
  {
    "ph": "s", "id": 237320, "pid": 76337, "tid": -914061504, "ts": 1716454225165519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225219744, "dur": 10,
    "args": {
      "External id": 237342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237342,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 237342, "pid": 5, "tid": 7, "ts": 1716454225219744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165577, "dur": 10,
    "args": {
      "External id": 237342,
      "cbid": 211, "correlation": 237342
    }
  },
  {
    "ph": "s", "id": 237342, "pid": 76337, "tid": -914061504, "ts": 1716454225165577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165666, "dur": 2,
    "args": {
      "External id": 237353,
      "cbid": 251, "correlation": 237353
    }
  },
  {
    "ph": "f", "id": 237353, "pid": 76337, "tid": -914061504, "ts": 1716454225165666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165670, "dur": 0,
    "args": {
      "External id": 237354,
      "cbid": 251, "correlation": 237354
    }
  },
  {
    "ph": "f", "id": 237354, "pid": 76337, "tid": -914061504, "ts": 1716454225165670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225219755, "dur": 53,
    "args": {
      "External id": 237355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237355,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 237355, "pid": 5, "tid": 7, "ts": 1716454225219755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165673, "dur": 13,
    "args": {
      "External id": 237355,
      "cbid": 211, "correlation": 237355
    }
  },
  {
    "ph": "s", "id": 237355, "pid": 76337, "tid": -914061504, "ts": 1716454225165673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165743, "dur": 1,
    "args": {
      "External id": 237366,
      "cbid": 251, "correlation": 237366
    }
  },
  {
    "ph": "f", "id": 237366, "pid": 76337, "tid": -914061504, "ts": 1716454225165743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165747, "dur": 0,
    "args": {
      "External id": 237367,
      "cbid": 251, "correlation": 237367
    }
  },
  {
    "ph": "f", "id": 237367, "pid": 76337, "tid": -914061504, "ts": 1716454225165747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225219810, "dur": 52,
    "args": {
      "External id": 237368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237368,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 237368, "pid": 5, "tid": 7, "ts": 1716454225219810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165749, "dur": 12,
    "args": {
      "External id": 237368,
      "cbid": 211, "correlation": 237368
    }
  },
  {
    "ph": "s", "id": 237368, "pid": 76337, "tid": -914061504, "ts": 1716454225165749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165815, "dur": 1,
    "args": {
      "External id": 237379,
      "cbid": 251, "correlation": 237379
    }
  },
  {
    "ph": "f", "id": 237379, "pid": 76337, "tid": -914061504, "ts": 1716454225165815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165818, "dur": 0,
    "args": {
      "External id": 237380,
      "cbid": 251, "correlation": 237380
    }
  },
  {
    "ph": "f", "id": 237380, "pid": 76337, "tid": -914061504, "ts": 1716454225165818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225219863, "dur": 53,
    "args": {
      "External id": 237381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237381,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 237381, "pid": 5, "tid": 7, "ts": 1716454225219863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165820, "dur": 11,
    "args": {
      "External id": 237381,
      "cbid": 211, "correlation": 237381
    }
  },
  {
    "ph": "s", "id": 237381, "pid": 76337, "tid": -914061504, "ts": 1716454225165820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225219917, "dur": 55,
    "args": {
      "External id": 237406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237406,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237406, "pid": 5, "tid": 7, "ts": 1716454225219917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225165907, "dur": 13,
    "args": {
      "External id": 237406,
      "cbid": 211, "correlation": 237406
    }
  },
  {
    "ph": "s", "id": 237406, "pid": 76337, "tid": -914061504, "ts": 1716454225165907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166014, "dur": 1,
    "args": {
      "External id": 237424,
      "cbid": 251, "correlation": 237424
    }
  },
  {
    "ph": "f", "id": 237424, "pid": 76337, "tid": -914061504, "ts": 1716454225166014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225219973, "dur": 62,
    "args": {
      "External id": 237426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237426,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 237426, "pid": 5, "tid": 7, "ts": 1716454225219973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166020, "dur": 13,
    "args": {
      "External id": 237426,
      "cbid": 211, "correlation": 237426
    }
  },
  {
    "ph": "s", "id": 237426, "pid": 76337, "tid": -914061504, "ts": 1716454225166020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225220036, "dur": 6,
    "args": {
      "External id": 237434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237434,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237434, "pid": 5, "tid": 7, "ts": 1716454225220036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166091, "dur": 13,
    "args": {
      "External id": 237434,
      "cbid": 211, "correlation": 237434
    }
  },
  {
    "ph": "s", "id": 237434, "pid": 76337, "tid": -914061504, "ts": 1716454225166091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225220044, "dur": 7,
    "args": {
      "External id": 237442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237442,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237442, "pid": 5, "tid": 7, "ts": 1716454225220044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166130, "dur": 8,
    "args": {
      "External id": 237442,
      "cbid": 211, "correlation": 237442
    }
  },
  {
    "ph": "s", "id": 237442, "pid": 76337, "tid": -914061504, "ts": 1716454225166130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220052, "dur": 7,
    "args": {
      "External id": 237453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237453, "pid": 5, "tid": 7, "ts": 1716454225220052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166205, "dur": 13,
    "args": {
      "External id": 237453,
      "cbid": 211, "correlation": 237453
    }
  },
  {
    "ph": "s", "id": 237453, "pid": 76337, "tid": -914061504, "ts": 1716454225166205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220061, "dur": 8,
    "args": {
      "External id": 237475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237475,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 237475, "pid": 5, "tid": 7, "ts": 1716454225220061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166238, "dur": 8,
    "args": {
      "External id": 237475,
      "cbid": 211, "correlation": 237475
    }
  },
  {
    "ph": "s", "id": 237475, "pid": 76337, "tid": -914061504, "ts": 1716454225166238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166324, "dur": 2,
    "args": {
      "External id": 237486,
      "cbid": 251, "correlation": 237486
    }
  },
  {
    "ph": "f", "id": 237486, "pid": 76337, "tid": -914061504, "ts": 1716454225166324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225220071, "dur": 1,
    "args": {
      "External id": 237487,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237487,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 237487, "pid": 5, "tid": 7, "ts": 1716454225220071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166330, "dur": 10,
    "args": {
      "External id": 237487,
      "cbid": 51, "correlation": 237487
    }
  },
  {
    "ph": "s", "id": 237487, "pid": 76337, "tid": -914061504, "ts": 1716454225166330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225220075, "dur": 35,
    "args": {
      "External id": 237488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237488,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 237488, "pid": 5, "tid": 7, "ts": 1716454225220075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166342, "dur": 13,
    "args": {
      "External id": 237488,
      "cbid": 211, "correlation": 237488
    }
  },
  {
    "ph": "s", "id": 237488, "pid": 76337, "tid": -914061504, "ts": 1716454225166342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166414, "dur": 1,
    "args": {
      "External id": 237499,
      "cbid": 251, "correlation": 237499
    }
  },
  {
    "ph": "f", "id": 237499, "pid": 76337, "tid": -914061504, "ts": 1716454225166414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166418, "dur": 0,
    "args": {
      "External id": 237500,
      "cbid": 251, "correlation": 237500
    }
  },
  {
    "ph": "f", "id": 237500, "pid": 76337, "tid": -914061504, "ts": 1716454225166418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225220112, "dur": 11,
    "args": {
      "External id": 237501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237501,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237501, "pid": 5, "tid": 7, "ts": 1716454225220112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166420, "dur": 12,
    "args": {
      "External id": 237501,
      "cbid": 211, "correlation": 237501
    }
  },
  {
    "ph": "s", "id": 237501, "pid": 76337, "tid": -914061504, "ts": 1716454225166420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225220124, "dur": 6,
    "args": {
      "External id": 237503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237503,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237503, "pid": 5, "tid": 7, "ts": 1716454225220124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166433, "dur": 6,
    "args": {
      "External id": 237503,
      "cbid": 211, "correlation": 237503
    }
  },
  {
    "ph": "s", "id": 237503, "pid": 76337, "tid": -914061504, "ts": 1716454225166433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166493, "dur": 1,
    "args": {
      "External id": 237514,
      "cbid": 251, "correlation": 237514
    }
  },
  {
    "ph": "f", "id": 237514, "pid": 76337, "tid": -914061504, "ts": 1716454225166493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166496, "dur": 0,
    "args": {
      "External id": 237515,
      "cbid": 251, "correlation": 237515
    }
  },
  {
    "ph": "f", "id": 237515, "pid": 76337, "tid": -914061504, "ts": 1716454225166496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225220131, "dur": 8,
    "args": {
      "External id": 237516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237516,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237516, "pid": 5, "tid": 7, "ts": 1716454225220131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166498, "dur": 11,
    "args": {
      "External id": 237516,
      "cbid": 211, "correlation": 237516
    }
  },
  {
    "ph": "s", "id": 237516, "pid": 76337, "tid": -914061504, "ts": 1716454225166498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225220140, "dur": 4,
    "args": {
      "External id": 237518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237518,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237518, "pid": 5, "tid": 7, "ts": 1716454225220140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166511, "dur": 5,
    "args": {
      "External id": 237518,
      "cbid": 211, "correlation": 237518
    }
  },
  {
    "ph": "s", "id": 237518, "pid": 76337, "tid": -914061504, "ts": 1716454225166511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225220145, "dur": 19,
    "args": {
      "External id": 237543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237543,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 237543, "pid": 5, "tid": 7, "ts": 1716454225220145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166589, "dur": 13,
    "args": {
      "External id": 237543,
      "cbid": 211, "correlation": 237543
    }
  },
  {
    "ph": "s", "id": 237543, "pid": 76337, "tid": -914061504, "ts": 1716454225166589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166688, "dur": 2,
    "args": {
      "External id": 237561,
      "cbid": 251, "correlation": 237561
    }
  },
  {
    "ph": "f", "id": 237561, "pid": 76337, "tid": -914061504, "ts": 1716454225166688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225220167, "dur": 1,
    "args": {
      "External id": 237563,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237563,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 237563, "pid": 5, "tid": 7, "ts": 1716454225220167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166694, "dur": 10,
    "args": {
      "External id": 237563,
      "cbid": 51, "correlation": 237563
    }
  },
  {
    "ph": "s", "id": 237563, "pid": 76337, "tid": -914061504, "ts": 1716454225166694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225220170, "dur": 35,
    "args": {
      "External id": 237564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237564,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 237564, "pid": 5, "tid": 7, "ts": 1716454225220170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166706, "dur": 12,
    "args": {
      "External id": 237564,
      "cbid": 211, "correlation": 237564
    }
  },
  {
    "ph": "s", "id": 237564, "pid": 76337, "tid": -914061504, "ts": 1716454225166706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225220207, "dur": 4,
    "args": {
      "External id": 237572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237572,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237572, "pid": 5, "tid": 7, "ts": 1716454225220207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166776, "dur": 12,
    "args": {
      "External id": 237572,
      "cbid": 211, "correlation": 237572
    }
  },
  {
    "ph": "s", "id": 237572, "pid": 76337, "tid": -914061504, "ts": 1716454225166776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220212, "dur": 8,
    "args": {
      "External id": 237580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237580,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237580, "pid": 5, "tid": 7, "ts": 1716454225220212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166818, "dur": 9,
    "args": {
      "External id": 237580,
      "cbid": 211, "correlation": 237580
    }
  },
  {
    "ph": "s", "id": 237580, "pid": 76337, "tid": -914061504, "ts": 1716454225166818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220222, "dur": 8,
    "args": {
      "External id": 237602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237602,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 237602, "pid": 5, "tid": 7, "ts": 1716454225220222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166870, "dur": 10,
    "args": {
      "External id": 237602,
      "cbid": 211, "correlation": 237602
    }
  },
  {
    "ph": "s", "id": 237602, "pid": 76337, "tid": -914061504, "ts": 1716454225166870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166961, "dur": 1,
    "args": {
      "External id": 237618,
      "cbid": 251, "correlation": 237618
    }
  },
  {
    "ph": "f", "id": 237618, "pid": 76337, "tid": -914061504, "ts": 1716454225166961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166966, "dur": 0,
    "args": {
      "External id": 237620,
      "cbid": 251, "correlation": 237620
    }
  },
  {
    "ph": "f", "id": 237620, "pid": 76337, "tid": -914061504, "ts": 1716454225166966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225220231, "dur": 188,
    "args": {
      "External id": 237621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237621,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237621, "pid": 5, "tid": 7, "ts": 1716454225220231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225166968, "dur": 21,
    "args": {
      "External id": 237621,
      "cbid": 211, "correlation": 237621
    }
  },
  {
    "ph": "s", "id": 237621, "pid": 76337, "tid": -914061504, "ts": 1716454225166968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220420, "dur": 20,
    "args": {
      "External id": 237629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237629, "pid": 5, "tid": 7, "ts": 1716454225220420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167043, "dur": 13,
    "args": {
      "External id": 237629,
      "cbid": 211, "correlation": 237629
    }
  },
  {
    "ph": "s", "id": 237629, "pid": 76337, "tid": -914061504, "ts": 1716454225167043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220442, "dur": 22,
    "args": {
      "External id": 237637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237637, "pid": 5, "tid": 7, "ts": 1716454225220442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167074, "dur": 8,
    "args": {
      "External id": 237637,
      "cbid": 211, "correlation": 237637
    }
  },
  {
    "ph": "s", "id": 237637, "pid": 76337, "tid": -914061504, "ts": 1716454225167074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167156, "dur": 1,
    "args": {
      "External id": 237653,
      "cbid": 251, "correlation": 237653
    }
  },
  {
    "ph": "f", "id": 237653, "pid": 76337, "tid": -914061504, "ts": 1716454225167156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225220466, "dur": 1,
    "args": {
      "External id": 237655,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237655,
      "bytes": 120, "memory bandwidth (GB/s)": 0.0797872340425532
    }
  },
  {
    "ph": "f", "id": 237655, "pid": 5, "tid": 7, "ts": 1716454225220466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167161, "dur": 8,
    "args": {
      "External id": 237655,
      "cbid": 51, "correlation": 237655
    }
  },
  {
    "ph": "s", "id": 237655, "pid": 76337, "tid": -914061504, "ts": 1716454225167161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225220470, "dur": 109,
    "args": {
      "External id": 237656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237656,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 237656, "pid": 5, "tid": 7, "ts": 1716454225220470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167171, "dur": 11,
    "args": {
      "External id": 237656,
      "cbid": 211, "correlation": 237656
    }
  },
  {
    "ph": "s", "id": 237656, "pid": 76337, "tid": -914061504, "ts": 1716454225167171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225220580, "dur": 5,
    "args": {
      "External id": 237664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237664,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237664, "pid": 5, "tid": 7, "ts": 1716454225220580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167214, "dur": 10,
    "args": {
      "External id": 237664,
      "cbid": 211, "correlation": 237664
    }
  },
  {
    "ph": "s", "id": 237664, "pid": 76337, "tid": -914061504, "ts": 1716454225167214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220586, "dur": 9,
    "args": {
      "External id": 237675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237675, "pid": 5, "tid": 7, "ts": 1716454225220586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167283, "dur": 13,
    "args": {
      "External id": 237675,
      "cbid": 211, "correlation": 237675
    }
  },
  {
    "ph": "s", "id": 237675, "pid": 76337, "tid": -914061504, "ts": 1716454225167283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167351, "dur": 0,
    "args": {
      "External id": 237687,
      "cbid": 317, "correlation": 237687
    }
  },
  {
    "ph": "f", "id": 237687, "pid": 76337, "tid": -914061504, "ts": 1716454225167351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167352, "dur": 0,
    "args": {
      "External id": 237688,
      "cbid": 203, "correlation": 237688
    }
  },
  {
    "ph": "f", "id": 237688, "pid": 76337, "tid": -914061504, "ts": 1716454225167352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167353, "dur": 0,
    "args": {
      "External id": 237689,
      "cbid": 205, "correlation": 237689
    }
  },
  {
    "ph": "f", "id": 237689, "pid": 76337, "tid": -914061504, "ts": 1716454225167353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220597, "dur": 5,
    "args": {
      "External id": 237693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237693,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237693, "pid": 5, "tid": 7, "ts": 1716454225220597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167368, "dur": 12,
    "args": {
      "External id": 237693,
      "cbid": 211, "correlation": 237693
    }
  },
  {
    "ph": "s", "id": 237693, "pid": 76337, "tid": -914061504, "ts": 1716454225167368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225220604, "dur": 36,
    "args": {
      "External id": 237695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237695,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 237695, "pid": 5, "tid": 7, "ts": 1716454225220604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167387, "dur": 8,
    "args": {
      "External id": 237695,
      "cbid": 211, "correlation": 237695
    }
  },
  {
    "ph": "s", "id": 237695, "pid": 76337, "tid": -914061504, "ts": 1716454225167387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220641, "dur": 6,
    "args": {
      "External id": 237697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237697,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237697, "pid": 5, "tid": 7, "ts": 1716454225220641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167398, "dur": 5,
    "args": {
      "External id": 237697,
      "cbid": 211, "correlation": 237697
    }
  },
  {
    "ph": "s", "id": 237697, "pid": 76337, "tid": -914061504, "ts": 1716454225167398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220648, "dur": 7,
    "args": {
      "External id": 237703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237703, "pid": 5, "tid": 7, "ts": 1716454225220648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167424, "dur": 9,
    "args": {
      "External id": 237703,
      "cbid": 211, "correlation": 237703
    }
  },
  {
    "ph": "s", "id": 237703, "pid": 76337, "tid": -914061504, "ts": 1716454225167424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225220657, "dur": 5,
    "args": {
      "External id": 237711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237711,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237711, "pid": 5, "tid": 7, "ts": 1716454225220657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167456, "dur": 8,
    "args": {
      "External id": 237711,
      "cbid": 211, "correlation": 237711
    }
  },
  {
    "ph": "s", "id": 237711, "pid": 76337, "tid": -914061504, "ts": 1716454225167456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220663, "dur": 10,
    "args": {
      "External id": 237731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237731,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 237731, "pid": 5, "tid": 7, "ts": 1716454225220663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167528, "dur": 12,
    "args": {
      "External id": 237731,
      "cbid": 211, "correlation": 237731
    }
  },
  {
    "ph": "s", "id": 237731, "pid": 76337, "tid": -914061504, "ts": 1716454225167528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225220674, "dur": 4,
    "args": {
      "External id": 237743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237743,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 237743, "pid": 5, "tid": 7, "ts": 1716454225220674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167550, "dur": 6,
    "args": {
      "External id": 237743,
      "cbid": 211, "correlation": 237743
    }
  },
  {
    "ph": "s", "id": 237743, "pid": 76337, "tid": -914061504, "ts": 1716454225167550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225220680, "dur": 8,
    "args": {
      "External id": 237746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237746,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237746, "pid": 5, "tid": 7, "ts": 1716454225220680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167568, "dur": 6,
    "args": {
      "External id": 237746,
      "cbid": 211, "correlation": 237746
    }
  },
  {
    "ph": "s", "id": 237746, "pid": 76337, "tid": -914061504, "ts": 1716454225167568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225220689, "dur": 5,
    "args": {
      "External id": 237755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237755,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237755, "pid": 5, "tid": 7, "ts": 1716454225220689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167607, "dur": 10,
    "args": {
      "External id": 237755,
      "cbid": 211, "correlation": 237755
    }
  },
  {
    "ph": "s", "id": 237755, "pid": 76337, "tid": -914061504, "ts": 1716454225167607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167658, "dur": 0,
    "args": {
      "External id": 237765,
      "cbid": 317, "correlation": 237765
    }
  },
  {
    "ph": "f", "id": 237765, "pid": 76337, "tid": -914061504, "ts": 1716454225167658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167659, "dur": 0,
    "args": {
      "External id": 237766,
      "cbid": 203, "correlation": 237766
    }
  },
  {
    "ph": "f", "id": 237766, "pid": 76337, "tid": -914061504, "ts": 1716454225167659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167659, "dur": 0,
    "args": {
      "External id": 237767,
      "cbid": 205, "correlation": 237767
    }
  },
  {
    "ph": "f", "id": 237767, "pid": 76337, "tid": -914061504, "ts": 1716454225167659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220695, "dur": 5,
    "args": {
      "External id": 237771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237771,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237771, "pid": 5, "tid": 7, "ts": 1716454225220695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167672, "dur": 11,
    "args": {
      "External id": 237771,
      "cbid": 211, "correlation": 237771
    }
  },
  {
    "ph": "s", "id": 237771, "pid": 76337, "tid": -914061504, "ts": 1716454225167672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225220702, "dur": 158,
    "args": {
      "External id": 237773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237773, "pid": 5, "tid": 7, "ts": 1716454225220702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167686, "dur": 6,
    "args": {
      "External id": 237773,
      "cbid": 211, "correlation": 237773
    }
  },
  {
    "ph": "s", "id": 237773, "pid": 76337, "tid": -914061504, "ts": 1716454225167686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225220862, "dur": 1,
    "args": {
      "External id": 237775,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237775,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 237775, "pid": 5, "tid": 7, "ts": 1716454225220862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167697, "dur": 6,
    "args": {
      "External id": 237775,
      "cbid": 51, "correlation": 237775
    }
  },
  {
    "ph": "s", "id": 237775, "pid": 76337, "tid": -914061504, "ts": 1716454225167697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225220866, "dur": 266,
    "args": {
      "External id": 237776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237776,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237776, "pid": 5, "tid": 7, "ts": 1716454225220866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167704, "dur": 6,
    "args": {
      "External id": 237776,
      "cbid": 211, "correlation": 237776
    }
  },
  {
    "ph": "s", "id": 237776, "pid": 76337, "tid": -914061504, "ts": 1716454225167704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221133, "dur": 6,
    "args": {
      "External id": 237778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237778, "pid": 5, "tid": 7, "ts": 1716454225221133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167714, "dur": 5,
    "args": {
      "External id": 237778,
      "cbid": 211, "correlation": 237778
    }
  },
  {
    "ph": "s", "id": 237778, "pid": 76337, "tid": -914061504, "ts": 1716454225167714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225221140, "dur": 6,
    "args": {
      "External id": 237784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237784, "pid": 5, "tid": 7, "ts": 1716454225221140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167742, "dur": 9,
    "args": {
      "External id": 237784,
      "cbid": 211, "correlation": 237784
    }
  },
  {
    "ph": "s", "id": 237784, "pid": 76337, "tid": -914061504, "ts": 1716454225167742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225221147, "dur": 3,
    "args": {
      "External id": 237792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237792,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 237792, "pid": 5, "tid": 7, "ts": 1716454225221147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167785, "dur": 10,
    "args": {
      "External id": 237792,
      "cbid": 211, "correlation": 237792
    }
  },
  {
    "ph": "s", "id": 237792, "pid": 76337, "tid": -914061504, "ts": 1716454225167785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167850, "dur": 1,
    "args": {
      "External id": 237808,
      "cbid": 251, "correlation": 237808
    }
  },
  {
    "ph": "f", "id": 237808, "pid": 76337, "tid": -914061504, "ts": 1716454225167850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167855, "dur": 0,
    "args": {
      "External id": 237810,
      "cbid": 251, "correlation": 237810
    }
  },
  {
    "ph": "f", "id": 237810, "pid": 76337, "tid": -914061504, "ts": 1716454225167855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225221152, "dur": 13,
    "args": {
      "External id": 237811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237811,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237811, "pid": 5, "tid": 7, "ts": 1716454225221152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167857, "dur": 11,
    "args": {
      "External id": 237811,
      "cbid": 211, "correlation": 237811
    }
  },
  {
    "ph": "s", "id": 237811, "pid": 76337, "tid": -914061504, "ts": 1716454225167857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225221166, "dur": 5,
    "args": {
      "External id": 237813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237813,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237813, "pid": 5, "tid": 7, "ts": 1716454225221166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167870, "dur": 5,
    "args": {
      "External id": 237813,
      "cbid": 211, "correlation": 237813
    }
  },
  {
    "ph": "s", "id": 237813, "pid": 76337, "tid": -914061504, "ts": 1716454225167870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225221172, "dur": 5,
    "args": {
      "External id": 237823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237823,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237823, "pid": 5, "tid": 7, "ts": 1716454225221172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225167928, "dur": 13,
    "args": {
      "External id": 237823,
      "cbid": 211, "correlation": 237823
    }
  },
  {
    "ph": "s", "id": 237823, "pid": 76337, "tid": -914061504, "ts": 1716454225167928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221179, "dur": 10,
    "args": {
      "External id": 237843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237843,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 237843, "pid": 5, "tid": 7, "ts": 1716454225221179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168003, "dur": 11,
    "args": {
      "External id": 237843,
      "cbid": 211, "correlation": 237843
    }
  },
  {
    "ph": "s", "id": 237843, "pid": 76337, "tid": -914061504, "ts": 1716454225168003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225221190, "dur": 4,
    "args": {
      "External id": 237855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237855,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 237855, "pid": 5, "tid": 7, "ts": 1716454225221190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168024, "dur": 6,
    "args": {
      "External id": 237855,
      "cbid": 211, "correlation": 237855
    }
  },
  {
    "ph": "s", "id": 237855, "pid": 76337, "tid": -914061504, "ts": 1716454225168024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225221195, "dur": 7,
    "args": {
      "External id": 237858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237858, "pid": 5, "tid": 7, "ts": 1716454225221195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168043, "dur": 7,
    "args": {
      "External id": 237858,
      "cbid": 211, "correlation": 237858
    }
  },
  {
    "ph": "s", "id": 237858, "pid": 76337, "tid": -914061504, "ts": 1716454225168043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225221203, "dur": 4,
    "args": {
      "External id": 237867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237867,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237867, "pid": 5, "tid": 7, "ts": 1716454225221203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168084, "dur": 10,
    "args": {
      "External id": 237867,
      "cbid": 211, "correlation": 237867
    }
  },
  {
    "ph": "s", "id": 237867, "pid": 76337, "tid": -914061504, "ts": 1716454225168084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168146, "dur": 0,
    "args": {
      "External id": 237877,
      "cbid": 317, "correlation": 237877
    }
  },
  {
    "ph": "f", "id": 237877, "pid": 76337, "tid": -914061504, "ts": 1716454225168146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168146, "dur": 0,
    "args": {
      "External id": 237878,
      "cbid": 203, "correlation": 237878
    }
  },
  {
    "ph": "f", "id": 237878, "pid": 76337, "tid": -914061504, "ts": 1716454225168146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168147, "dur": 0,
    "args": {
      "External id": 237879,
      "cbid": 205, "correlation": 237879
    }
  },
  {
    "ph": "f", "id": 237879, "pid": 76337, "tid": -914061504, "ts": 1716454225168147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221209, "dur": 5,
    "args": {
      "External id": 237883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237883,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237883, "pid": 5, "tid": 7, "ts": 1716454225221209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168161, "dur": 12,
    "args": {
      "External id": 237883,
      "cbid": 211, "correlation": 237883
    }
  },
  {
    "ph": "s", "id": 237883, "pid": 76337, "tid": -914061504, "ts": 1716454225168161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221215, "dur": 159,
    "args": {
      "External id": 237885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237885, "pid": 5, "tid": 7, "ts": 1716454225221215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168175, "dur": 5,
    "args": {
      "External id": 237885,
      "cbid": 211, "correlation": 237885
    }
  },
  {
    "ph": "s", "id": 237885, "pid": 76337, "tid": -914061504, "ts": 1716454225168175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225221377, "dur": 1,
    "args": {
      "External id": 237887,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237887,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1561483409238777
    }
  },
  {
    "ph": "f", "id": 237887, "pid": 5, "tid": 7, "ts": 1716454225221377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168186, "dur": 6,
    "args": {
      "External id": 237887,
      "cbid": 51, "correlation": 237887
    }
  },
  {
    "ph": "s", "id": 237887, "pid": 76337, "tid": -914061504, "ts": 1716454225168186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225221380, "dur": 256,
    "args": {
      "External id": 237888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237888,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237888, "pid": 5, "tid": 7, "ts": 1716454225221380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168193, "dur": 6,
    "args": {
      "External id": 237888,
      "cbid": 211, "correlation": 237888
    }
  },
  {
    "ph": "s", "id": 237888, "pid": 76337, "tid": -914061504, "ts": 1716454225168193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221637, "dur": 6,
    "args": {
      "External id": 237890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237890,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237890, "pid": 5, "tid": 7, "ts": 1716454225221637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168203, "dur": 5,
    "args": {
      "External id": 237890,
      "cbid": 211, "correlation": 237890
    }
  },
  {
    "ph": "s", "id": 237890, "pid": 76337, "tid": -914061504, "ts": 1716454225168203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225221645, "dur": 7,
    "args": {
      "External id": 237896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237896,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237896, "pid": 5, "tid": 7, "ts": 1716454225221645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168231, "dur": 9,
    "args": {
      "External id": 237896,
      "cbid": 211, "correlation": 237896
    }
  },
  {
    "ph": "s", "id": 237896, "pid": 76337, "tid": -914061504, "ts": 1716454225168231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225221653, "dur": 4,
    "args": {
      "External id": 237904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237904,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237904, "pid": 5, "tid": 7, "ts": 1716454225221653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168264, "dur": 8,
    "args": {
      "External id": 237904,
      "cbid": 211, "correlation": 237904
    }
  },
  {
    "ph": "s", "id": 237904, "pid": 76337, "tid": -914061504, "ts": 1716454225168264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225221658, "dur": 4,
    "args": {
      "External id": 237912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237912,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237912, "pid": 5, "tid": 7, "ts": 1716454225221658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168293, "dur": 8,
    "args": {
      "External id": 237912,
      "cbid": 211, "correlation": 237912
    }
  },
  {
    "ph": "s", "id": 237912, "pid": 76337, "tid": -914061504, "ts": 1716454225168293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225221664, "dur": 11,
    "args": {
      "External id": 237921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237921,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237921, "pid": 5, "tid": 7, "ts": 1716454225221664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168381, "dur": 14,
    "args": {
      "External id": 237921,
      "cbid": 211, "correlation": 237921
    }
  },
  {
    "ph": "s", "id": 237921, "pid": 76337, "tid": -914061504, "ts": 1716454225168381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221677, "dur": 12,
    "args": {
      "External id": 237941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237941,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 237941, "pid": 5, "tid": 7, "ts": 1716454225221677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168449, "dur": 11,
    "args": {
      "External id": 237941,
      "cbid": 211, "correlation": 237941
    }
  },
  {
    "ph": "s", "id": 237941, "pid": 76337, "tid": -914061504, "ts": 1716454225168449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225221690, "dur": 4,
    "args": {
      "External id": 237953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237953,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237953, "pid": 5, "tid": 7, "ts": 1716454225221690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168469, "dur": 6,
    "args": {
      "External id": 237953,
      "cbid": 211, "correlation": 237953
    }
  },
  {
    "ph": "s", "id": 237953, "pid": 76337, "tid": -914061504, "ts": 1716454225168469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225221695, "dur": 10,
    "args": {
      "External id": 237956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237956, "pid": 5, "tid": 7, "ts": 1716454225221695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168488, "dur": 6,
    "args": {
      "External id": 237956,
      "cbid": 211, "correlation": 237956
    }
  },
  {
    "ph": "s", "id": 237956, "pid": 76337, "tid": -914061504, "ts": 1716454225168488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225221706, "dur": 6,
    "args": {
      "External id": 237965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237965,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237965, "pid": 5, "tid": 7, "ts": 1716454225221706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168525, "dur": 10,
    "args": {
      "External id": 237965,
      "cbid": 211, "correlation": 237965
    }
  },
  {
    "ph": "s", "id": 237965, "pid": 76337, "tid": -914061504, "ts": 1716454225168525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168577, "dur": 0,
    "args": {
      "External id": 237975,
      "cbid": 317, "correlation": 237975
    }
  },
  {
    "ph": "f", "id": 237975, "pid": 76337, "tid": -914061504, "ts": 1716454225168577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168578, "dur": 0,
    "args": {
      "External id": 237976,
      "cbid": 203, "correlation": 237976
    }
  },
  {
    "ph": "f", "id": 237976, "pid": 76337, "tid": -914061504, "ts": 1716454225168578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168579, "dur": 0,
    "args": {
      "External id": 237977,
      "cbid": 205, "correlation": 237977
    }
  },
  {
    "ph": "f", "id": 237977, "pid": 76337, "tid": -914061504, "ts": 1716454225168579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221713, "dur": 6,
    "args": {
      "External id": 237981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237981, "pid": 5, "tid": 7, "ts": 1716454225221713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168594, "dur": 11,
    "args": {
      "External id": 237981,
      "cbid": 211, "correlation": 237981
    }
  },
  {
    "ph": "s", "id": 237981, "pid": 76337, "tid": -914061504, "ts": 1716454225168594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225221721, "dur": 314,
    "args": {
      "External id": 237983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237983, "pid": 5, "tid": 7, "ts": 1716454225221721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168608, "dur": 5,
    "args": {
      "External id": 237983,
      "cbid": 211, "correlation": 237983
    }
  },
  {
    "ph": "s", "id": 237983, "pid": 76337, "tid": -914061504, "ts": 1716454225168608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225222037, "dur": 1,
    "args": {
      "External id": 237985,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 237985,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14423076923076922
    }
  },
  {
    "ph": "f", "id": 237985, "pid": 5, "tid": 7, "ts": 1716454225222037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168619, "dur": 8,
    "args": {
      "External id": 237985,
      "cbid": 51, "correlation": 237985
    }
  },
  {
    "ph": "s", "id": 237985, "pid": 76337, "tid": -914061504, "ts": 1716454225168619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225222041, "dur": 486,
    "args": {
      "External id": 237986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237986,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 237986, "pid": 5, "tid": 7, "ts": 1716454225222041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168628, "dur": 6,
    "args": {
      "External id": 237986,
      "cbid": 211, "correlation": 237986
    }
  },
  {
    "ph": "s", "id": 237986, "pid": 76337, "tid": -914061504, "ts": 1716454225168628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225222528, "dur": 5,
    "args": {
      "External id": 237988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 237988, "pid": 5, "tid": 7, "ts": 1716454225222528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168638, "dur": 5,
    "args": {
      "External id": 237988,
      "cbid": 211, "correlation": 237988
    }
  },
  {
    "ph": "s", "id": 237988, "pid": 76337, "tid": -914061504, "ts": 1716454225168638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225222535, "dur": 6,
    "args": {
      "External id": 237994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 237994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 237994, "pid": 5, "tid": 7, "ts": 1716454225222535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168666, "dur": 9,
    "args": {
      "External id": 237994,
      "cbid": 211, "correlation": 237994
    }
  },
  {
    "ph": "s", "id": 237994, "pid": 76337, "tid": -914061504, "ts": 1716454225168666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225222542, "dur": 3,
    "args": {
      "External id": 238002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238002,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 238002, "pid": 5, "tid": 7, "ts": 1716454225222542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168710, "dur": 9,
    "args": {
      "External id": 238002,
      "cbid": 211, "correlation": 238002
    }
  },
  {
    "ph": "s", "id": 238002, "pid": 76337, "tid": -914061504, "ts": 1716454225168710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168773, "dur": 1,
    "args": {
      "External id": 238018,
      "cbid": 251, "correlation": 238018
    }
  },
  {
    "ph": "f", "id": 238018, "pid": 76337, "tid": -914061504, "ts": 1716454225168773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168778, "dur": 0,
    "args": {
      "External id": 238020,
      "cbid": 251, "correlation": 238020
    }
  },
  {
    "ph": "f", "id": 238020, "pid": 76337, "tid": -914061504, "ts": 1716454225168778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225222547, "dur": 11,
    "args": {
      "External id": 238021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238021,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238021, "pid": 5, "tid": 7, "ts": 1716454225222547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168780, "dur": 11,
    "args": {
      "External id": 238021,
      "cbid": 211, "correlation": 238021
    }
  },
  {
    "ph": "s", "id": 238021, "pid": 76337, "tid": -914061504, "ts": 1716454225168780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225222558, "dur": 4,
    "args": {
      "External id": 238023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238023,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238023, "pid": 5, "tid": 7, "ts": 1716454225222558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168793, "dur": 6,
    "args": {
      "External id": 238023,
      "cbid": 211, "correlation": 238023
    }
  },
  {
    "ph": "s", "id": 238023, "pid": 76337, "tid": -914061504, "ts": 1716454225168793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225222564, "dur": 6,
    "args": {
      "External id": 238033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238033, "pid": 5, "tid": 7, "ts": 1716454225222564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168849, "dur": 12,
    "args": {
      "External id": 238033,
      "cbid": 211, "correlation": 238033
    }
  },
  {
    "ph": "s", "id": 238033, "pid": 76337, "tid": -914061504, "ts": 1716454225168849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225222571, "dur": 9,
    "args": {
      "External id": 238053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238053,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238053, "pid": 5, "tid": 7, "ts": 1716454225222571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168915, "dur": 10,
    "args": {
      "External id": 238053,
      "cbid": 211, "correlation": 238053
    }
  },
  {
    "ph": "s", "id": 238053, "pid": 76337, "tid": -914061504, "ts": 1716454225168915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225222582, "dur": 4,
    "args": {
      "External id": 238065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238065,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 238065, "pid": 5, "tid": 7, "ts": 1716454225222582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168936, "dur": 6,
    "args": {
      "External id": 238065,
      "cbid": 211, "correlation": 238065
    }
  },
  {
    "ph": "s", "id": 238065, "pid": 76337, "tid": -914061504, "ts": 1716454225168936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225222587, "dur": 7,
    "args": {
      "External id": 238068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238068,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238068, "pid": 5, "tid": 7, "ts": 1716454225222587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225168955, "dur": 6,
    "args": {
      "External id": 238068,
      "cbid": 211, "correlation": 238068
    }
  },
  {
    "ph": "s", "id": 238068, "pid": 76337, "tid": -914061504, "ts": 1716454225168955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225222595, "dur": 4,
    "args": {
      "External id": 238077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238077,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238077, "pid": 5, "tid": 7, "ts": 1716454225222595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169004, "dur": 10,
    "args": {
      "External id": 238077,
      "cbid": 211, "correlation": 238077
    }
  },
  {
    "ph": "s", "id": 238077, "pid": 76337, "tid": -914061504, "ts": 1716454225169004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169067, "dur": 0,
    "args": {
      "External id": 238087,
      "cbid": 317, "correlation": 238087
    }
  },
  {
    "ph": "f", "id": 238087, "pid": 76337, "tid": -914061504, "ts": 1716454225169067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169068, "dur": 0,
    "args": {
      "External id": 238088,
      "cbid": 203, "correlation": 238088
    }
  },
  {
    "ph": "f", "id": 238088, "pid": 76337, "tid": -914061504, "ts": 1716454225169068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169069, "dur": 0,
    "args": {
      "External id": 238089,
      "cbid": 205, "correlation": 238089
    }
  },
  {
    "ph": "f", "id": 238089, "pid": 76337, "tid": -914061504, "ts": 1716454225169069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225222601, "dur": 5,
    "args": {
      "External id": 238093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238093,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238093, "pid": 5, "tid": 7, "ts": 1716454225222601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169083, "dur": 12,
    "args": {
      "External id": 238093,
      "cbid": 211, "correlation": 238093
    }
  },
  {
    "ph": "s", "id": 238093, "pid": 76337, "tid": -914061504, "ts": 1716454225169083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225222607, "dur": 158,
    "args": {
      "External id": 238095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238095,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238095, "pid": 5, "tid": 7, "ts": 1716454225222607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169097, "dur": 5,
    "args": {
      "External id": 238095,
      "cbid": 211, "correlation": 238095
    }
  },
  {
    "ph": "s", "id": 238095, "pid": 76337, "tid": -914061504, "ts": 1716454225169097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225222767, "dur": 1,
    "args": {
      "External id": 238097,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238097,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 238097, "pid": 5, "tid": 7, "ts": 1716454225222767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169108, "dur": 7,
    "args": {
      "External id": 238097,
      "cbid": 51, "correlation": 238097
    }
  },
  {
    "ph": "s", "id": 238097, "pid": 76337, "tid": -914061504, "ts": 1716454225169108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225222771, "dur": 254,
    "args": {
      "External id": 238098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238098,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238098, "pid": 5, "tid": 7, "ts": 1716454225222771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169115, "dur": 6,
    "args": {
      "External id": 238098,
      "cbid": 211, "correlation": 238098
    }
  },
  {
    "ph": "s", "id": 238098, "pid": 76337, "tid": -914061504, "ts": 1716454225169115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225223026, "dur": 6,
    "args": {
      "External id": 238100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238100, "pid": 5, "tid": 7, "ts": 1716454225223026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169125, "dur": 6,
    "args": {
      "External id": 238100,
      "cbid": 211, "correlation": 238100
    }
  },
  {
    "ph": "s", "id": 238100, "pid": 76337, "tid": -914061504, "ts": 1716454225169125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225223033, "dur": 6,
    "args": {
      "External id": 238106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238106, "pid": 5, "tid": 7, "ts": 1716454225223033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169153, "dur": 8,
    "args": {
      "External id": 238106,
      "cbid": 211, "correlation": 238106
    }
  },
  {
    "ph": "s", "id": 238106, "pid": 76337, "tid": -914061504, "ts": 1716454225169153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169212, "dur": 0,
    "args": {
      "External id": 238116,
      "cbid": 317, "correlation": 238116
    }
  },
  {
    "ph": "f", "id": 238116, "pid": 76337, "tid": -914061504, "ts": 1716454225169212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169213, "dur": 0,
    "args": {
      "External id": 238117,
      "cbid": 203, "correlation": 238117
    }
  },
  {
    "ph": "f", "id": 238117, "pid": 76337, "tid": -914061504, "ts": 1716454225169213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169214, "dur": 0,
    "args": {
      "External id": 238118,
      "cbid": 205, "correlation": 238118
    }
  },
  {
    "ph": "f", "id": 238118, "pid": 76337, "tid": -914061504, "ts": 1716454225169214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225223041, "dur": 7,
    "args": {
      "External id": 238122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238122, "pid": 5, "tid": 7, "ts": 1716454225223041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169227, "dur": 13,
    "args": {
      "External id": 238122,
      "cbid": 211, "correlation": 238122
    }
  },
  {
    "ph": "s", "id": 238122, "pid": 76337, "tid": -914061504, "ts": 1716454225169227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225223049, "dur": 3,
    "args": {
      "External id": 238124,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238124,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 238124, "pid": 5, "tid": 7, "ts": 1716454225223049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169246, "dur": 14,
    "args": {
      "External id": 238124,
      "cbid": 51, "correlation": 238124
    }
  },
  {
    "ph": "s", "id": 238124, "pid": 76337, "tid": -914061504, "ts": 1716454225169246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225223053, "dur": 97,
    "args": {
      "External id": 238125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238125,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 238125, "pid": 5, "tid": 7, "ts": 1716454225223053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169261, "dur": 7,
    "args": {
      "External id": 238125,
      "cbid": 211, "correlation": 238125
    }
  },
  {
    "ph": "s", "id": 238125, "pid": 76337, "tid": -914061504, "ts": 1716454225169261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225223152, "dur": 6,
    "args": {
      "External id": 238127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238127,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238127, "pid": 5, "tid": 7, "ts": 1716454225223152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169271, "dur": 5,
    "args": {
      "External id": 238127,
      "cbid": 211, "correlation": 238127
    }
  },
  {
    "ph": "s", "id": 238127, "pid": 76337, "tid": -914061504, "ts": 1716454225169271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225223158, "dur": 6,
    "args": {
      "External id": 238133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238133, "pid": 5, "tid": 7, "ts": 1716454225223158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169298, "dur": 8,
    "args": {
      "External id": 238133,
      "cbid": 211, "correlation": 238133
    }
  },
  {
    "ph": "s", "id": 238133, "pid": 76337, "tid": -914061504, "ts": 1716454225169298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225223166, "dur": 5,
    "args": {
      "External id": 238141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238141,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238141, "pid": 5, "tid": 7, "ts": 1716454225223166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169327, "dur": 8,
    "args": {
      "External id": 238141,
      "cbid": 211, "correlation": 238141
    }
  },
  {
    "ph": "s", "id": 238141, "pid": 76337, "tid": -914061504, "ts": 1716454225169327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225223172, "dur": 4,
    "args": {
      "External id": 238149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238149,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238149, "pid": 5, "tid": 7, "ts": 1716454225223172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169356, "dur": 8,
    "args": {
      "External id": 238149,
      "cbid": 211, "correlation": 238149
    }
  },
  {
    "ph": "s", "id": 238149, "pid": 76337, "tid": -914061504, "ts": 1716454225169356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225223177, "dur": 11,
    "args": {
      "External id": 238158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238158,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238158, "pid": 5, "tid": 7, "ts": 1716454225223177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169400, "dur": 10,
    "args": {
      "External id": 238158,
      "cbid": 211, "correlation": 238158
    }
  },
  {
    "ph": "s", "id": 238158, "pid": 76337, "tid": -914061504, "ts": 1716454225169400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225223190, "dur": 12,
    "args": {
      "External id": 238178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238178,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238178, "pid": 5, "tid": 7, "ts": 1716454225223190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169469, "dur": 12,
    "args": {
      "External id": 238178,
      "cbid": 211, "correlation": 238178
    }
  },
  {
    "ph": "s", "id": 238178, "pid": 76337, "tid": -914061504, "ts": 1716454225169469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225223203, "dur": 4,
    "args": {
      "External id": 238190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238190,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238190, "pid": 5, "tid": 7, "ts": 1716454225223203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169491, "dur": 6,
    "args": {
      "External id": 238190,
      "cbid": 211, "correlation": 238190
    }
  },
  {
    "ph": "s", "id": 238190, "pid": 76337, "tid": -914061504, "ts": 1716454225169491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225223208, "dur": 11,
    "args": {
      "External id": 238193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238193, "pid": 5, "tid": 7, "ts": 1716454225223208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169508, "dur": 7,
    "args": {
      "External id": 238193,
      "cbid": 211, "correlation": 238193
    }
  },
  {
    "ph": "s", "id": 238193, "pid": 76337, "tid": -914061504, "ts": 1716454225169508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225223220, "dur": 6,
    "args": {
      "External id": 238202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238202,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238202, "pid": 5, "tid": 7, "ts": 1716454225223220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169546, "dur": 9,
    "args": {
      "External id": 238202,
      "cbid": 211, "correlation": 238202
    }
  },
  {
    "ph": "s", "id": 238202, "pid": 76337, "tid": -914061504, "ts": 1716454225169546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169599, "dur": 0,
    "args": {
      "External id": 238212,
      "cbid": 317, "correlation": 238212
    }
  },
  {
    "ph": "f", "id": 238212, "pid": 76337, "tid": -914061504, "ts": 1716454225169599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169599, "dur": 0,
    "args": {
      "External id": 238213,
      "cbid": 203, "correlation": 238213
    }
  },
  {
    "ph": "f", "id": 238213, "pid": 76337, "tid": -914061504, "ts": 1716454225169599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169600, "dur": 0,
    "args": {
      "External id": 238214,
      "cbid": 205, "correlation": 238214
    }
  },
  {
    "ph": "f", "id": 238214, "pid": 76337, "tid": -914061504, "ts": 1716454225169600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225223227, "dur": 6,
    "args": {
      "External id": 238218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238218,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238218, "pid": 5, "tid": 7, "ts": 1716454225223227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169613, "dur": 11,
    "args": {
      "External id": 238218,
      "cbid": 211, "correlation": 238218
    }
  },
  {
    "ph": "s", "id": 238218, "pid": 76337, "tid": -914061504, "ts": 1716454225169613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225223235, "dur": 314,
    "args": {
      "External id": 238220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238220, "pid": 5, "tid": 7, "ts": 1716454225223235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169627, "dur": 6,
    "args": {
      "External id": 238220,
      "cbid": 211, "correlation": 238220
    }
  },
  {
    "ph": "s", "id": 238220, "pid": 76337, "tid": -914061504, "ts": 1716454225169627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225223551, "dur": 1,
    "args": {
      "External id": 238222,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238222,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 238222, "pid": 5, "tid": 7, "ts": 1716454225223551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169639, "dur": 7,
    "args": {
      "External id": 238222,
      "cbid": 51, "correlation": 238222
    }
  },
  {
    "ph": "s", "id": 238222, "pid": 76337, "tid": -914061504, "ts": 1716454225169639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225223555, "dur": 488,
    "args": {
      "External id": 238223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238223,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238223, "pid": 5, "tid": 7, "ts": 1716454225223555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169647, "dur": 6,
    "args": {
      "External id": 238223,
      "cbid": 211, "correlation": 238223
    }
  },
  {
    "ph": "s", "id": 238223, "pid": 76337, "tid": -914061504, "ts": 1716454225169647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224045, "dur": 5,
    "args": {
      "External id": 238225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238225,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238225, "pid": 5, "tid": 7, "ts": 1716454225224045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169656, "dur": 5,
    "args": {
      "External id": 238225,
      "cbid": 211, "correlation": 238225
    }
  },
  {
    "ph": "s", "id": 238225, "pid": 76337, "tid": -914061504, "ts": 1716454225169656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225224051, "dur": 6,
    "args": {
      "External id": 238231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238231, "pid": 5, "tid": 7, "ts": 1716454225224051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169686, "dur": 8,
    "args": {
      "External id": 238231,
      "cbid": 211, "correlation": 238231
    }
  },
  {
    "ph": "s", "id": 238231, "pid": 76337, "tid": -914061504, "ts": 1716454225169686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225224059, "dur": 3,
    "args": {
      "External id": 238239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238239,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 238239, "pid": 5, "tid": 7, "ts": 1716454225224059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169729, "dur": 9,
    "args": {
      "External id": 238239,
      "cbid": 211, "correlation": 238239
    }
  },
  {
    "ph": "s", "id": 238239, "pid": 76337, "tid": -914061504, "ts": 1716454225169729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169791, "dur": 1,
    "args": {
      "External id": 238255,
      "cbid": 251, "correlation": 238255
    }
  },
  {
    "ph": "f", "id": 238255, "pid": 76337, "tid": -914061504, "ts": 1716454225169791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169796, "dur": 0,
    "args": {
      "External id": 238257,
      "cbid": 251, "correlation": 238257
    }
  },
  {
    "ph": "f", "id": 238257, "pid": 76337, "tid": -914061504, "ts": 1716454225169796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225224063, "dur": 12,
    "args": {
      "External id": 238258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238258,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238258, "pid": 5, "tid": 7, "ts": 1716454225224063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169798, "dur": 11,
    "args": {
      "External id": 238258,
      "cbid": 211, "correlation": 238258
    }
  },
  {
    "ph": "s", "id": 238258, "pid": 76337, "tid": -914061504, "ts": 1716454225169798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225224077, "dur": 5,
    "args": {
      "External id": 238260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238260,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238260, "pid": 5, "tid": 7, "ts": 1716454225224077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169811, "dur": 5,
    "args": {
      "External id": 238260,
      "cbid": 211, "correlation": 238260
    }
  },
  {
    "ph": "s", "id": 238260, "pid": 76337, "tid": -914061504, "ts": 1716454225169811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225224083, "dur": 6,
    "args": {
      "External id": 238270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238270, "pid": 5, "tid": 7, "ts": 1716454225224083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169867, "dur": 14,
    "args": {
      "External id": 238270,
      "cbid": 211, "correlation": 238270
    }
  },
  {
    "ph": "s", "id": 238270, "pid": 76337, "tid": -914061504, "ts": 1716454225169867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224091, "dur": 9,
    "args": {
      "External id": 238290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238290,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238290, "pid": 5, "tid": 7, "ts": 1716454225224091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169935, "dur": 11,
    "args": {
      "External id": 238290,
      "cbid": 211, "correlation": 238290
    }
  },
  {
    "ph": "s", "id": 238290, "pid": 76337, "tid": -914061504, "ts": 1716454225169935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225224101, "dur": 3,
    "args": {
      "External id": 238302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238302,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 238302, "pid": 5, "tid": 7, "ts": 1716454225224101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169955, "dur": 6,
    "args": {
      "External id": 238302,
      "cbid": 211, "correlation": 238302
    }
  },
  {
    "ph": "s", "id": 238302, "pid": 76337, "tid": -914061504, "ts": 1716454225169955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225224106, "dur": 6,
    "args": {
      "External id": 238305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238305,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238305, "pid": 5, "tid": 7, "ts": 1716454225224106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225169983, "dur": 7,
    "args": {
      "External id": 238305,
      "cbid": 211, "correlation": 238305
    }
  },
  {
    "ph": "s", "id": 238305, "pid": 76337, "tid": -914061504, "ts": 1716454225169983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225224113, "dur": 4,
    "args": {
      "External id": 238314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238314,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238314, "pid": 5, "tid": 7, "ts": 1716454225224113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170025, "dur": 10,
    "args": {
      "External id": 238314,
      "cbid": 211, "correlation": 238314
    }
  },
  {
    "ph": "s", "id": 238314, "pid": 76337, "tid": -914061504, "ts": 1716454225170025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170088, "dur": 0,
    "args": {
      "External id": 238324,
      "cbid": 317, "correlation": 238324
    }
  },
  {
    "ph": "f", "id": 238324, "pid": 76337, "tid": -914061504, "ts": 1716454225170088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170088, "dur": 0,
    "args": {
      "External id": 238325,
      "cbid": 203, "correlation": 238325
    }
  },
  {
    "ph": "f", "id": 238325, "pid": 76337, "tid": -914061504, "ts": 1716454225170088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170089, "dur": 0,
    "args": {
      "External id": 238326,
      "cbid": 205, "correlation": 238326
    }
  },
  {
    "ph": "f", "id": 238326, "pid": 76337, "tid": -914061504, "ts": 1716454225170089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224119, "dur": 5,
    "args": {
      "External id": 238330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238330,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238330, "pid": 5, "tid": 7, "ts": 1716454225224119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170103, "dur": 12,
    "args": {
      "External id": 238330,
      "cbid": 211, "correlation": 238330
    }
  },
  {
    "ph": "s", "id": 238330, "pid": 76337, "tid": -914061504, "ts": 1716454225170103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224125, "dur": 160,
    "args": {
      "External id": 238332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238332,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238332, "pid": 5, "tid": 7, "ts": 1716454225224125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170117, "dur": 5,
    "args": {
      "External id": 238332,
      "cbid": 211, "correlation": 238332
    }
  },
  {
    "ph": "s", "id": 238332, "pid": 76337, "tid": -914061504, "ts": 1716454225170117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225224287, "dur": 1,
    "args": {
      "External id": 238334,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238334,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 238334, "pid": 5, "tid": 7, "ts": 1716454225224287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170128, "dur": 7,
    "args": {
      "External id": 238334,
      "cbid": 51, "correlation": 238334
    }
  },
  {
    "ph": "s", "id": 238334, "pid": 76337, "tid": -914061504, "ts": 1716454225170128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225224291, "dur": 255,
    "args": {
      "External id": 238335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238335,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238335, "pid": 5, "tid": 7, "ts": 1716454225224291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170136, "dur": 6,
    "args": {
      "External id": 238335,
      "cbid": 211, "correlation": 238335
    }
  },
  {
    "ph": "s", "id": 238335, "pid": 76337, "tid": -914061504, "ts": 1716454225170136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224547, "dur": 6,
    "args": {
      "External id": 238337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238337, "pid": 5, "tid": 7, "ts": 1716454225224547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170145, "dur": 5,
    "args": {
      "External id": 238337,
      "cbid": 211, "correlation": 238337
    }
  },
  {
    "ph": "s", "id": 238337, "pid": 76337, "tid": -914061504, "ts": 1716454225170145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225224554, "dur": 6,
    "args": {
      "External id": 238343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238343,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238343, "pid": 5, "tid": 7, "ts": 1716454225224554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170173, "dur": 10,
    "args": {
      "External id": 238343,
      "cbid": 211, "correlation": 238343
    }
  },
  {
    "ph": "s", "id": 238343, "pid": 76337, "tid": -914061504, "ts": 1716454225170173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170232, "dur": 0,
    "args": {
      "External id": 238353,
      "cbid": 317, "correlation": 238353
    }
  },
  {
    "ph": "f", "id": 238353, "pid": 76337, "tid": -914061504, "ts": 1716454225170232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170233, "dur": 0,
    "args": {
      "External id": 238354,
      "cbid": 203, "correlation": 238354
    }
  },
  {
    "ph": "f", "id": 238354, "pid": 76337, "tid": -914061504, "ts": 1716454225170233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170233, "dur": 0,
    "args": {
      "External id": 238355,
      "cbid": 205, "correlation": 238355
    }
  },
  {
    "ph": "f", "id": 238355, "pid": 76337, "tid": -914061504, "ts": 1716454225170233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224561, "dur": 7,
    "args": {
      "External id": 238359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238359,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238359, "pid": 5, "tid": 7, "ts": 1716454225224561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170244, "dur": 12,
    "args": {
      "External id": 238359,
      "cbid": 211, "correlation": 238359
    }
  },
  {
    "ph": "s", "id": 238359, "pid": 76337, "tid": -914061504, "ts": 1716454225170244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225224570, "dur": 3,
    "args": {
      "External id": 238361,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238361,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4150943396226414
    }
  },
  {
    "ph": "f", "id": 238361, "pid": 5, "tid": 7, "ts": 1716454225224570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170261, "dur": 9,
    "args": {
      "External id": 238361,
      "cbid": 51, "correlation": 238361
    }
  },
  {
    "ph": "s", "id": 238361, "pid": 76337, "tid": -914061504, "ts": 1716454225170261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225224574, "dur": 95,
    "args": {
      "External id": 238362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238362,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 238362, "pid": 5, "tid": 7, "ts": 1716454225224574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170272, "dur": 7,
    "args": {
      "External id": 238362,
      "cbid": 211, "correlation": 238362
    }
  },
  {
    "ph": "s", "id": 238362, "pid": 76337, "tid": -914061504, "ts": 1716454225170272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224670, "dur": 5,
    "args": {
      "External id": 238364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238364,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238364, "pid": 5, "tid": 7, "ts": 1716454225224670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170282, "dur": 5,
    "args": {
      "External id": 238364,
      "cbid": 211, "correlation": 238364
    }
  },
  {
    "ph": "s", "id": 238364, "pid": 76337, "tid": -914061504, "ts": 1716454225170282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225224677, "dur": 6,
    "args": {
      "External id": 238370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238370,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238370, "pid": 5, "tid": 7, "ts": 1716454225224677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170308, "dur": 8,
    "args": {
      "External id": 238370,
      "cbid": 211, "correlation": 238370
    }
  },
  {
    "ph": "s", "id": 238370, "pid": 76337, "tid": -914061504, "ts": 1716454225170308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225224685, "dur": 5,
    "args": {
      "External id": 238378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238378,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238378, "pid": 5, "tid": 7, "ts": 1716454225224685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170338, "dur": 8,
    "args": {
      "External id": 238378,
      "cbid": 211, "correlation": 238378
    }
  },
  {
    "ph": "s", "id": 238378, "pid": 76337, "tid": -914061504, "ts": 1716454225170338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225224691, "dur": 4,
    "args": {
      "External id": 238386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238386,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238386, "pid": 5, "tid": 7, "ts": 1716454225224691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170366, "dur": 9,
    "args": {
      "External id": 238386,
      "cbid": 211, "correlation": 238386
    }
  },
  {
    "ph": "s", "id": 238386, "pid": 76337, "tid": -914061504, "ts": 1716454225170366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225224696, "dur": 11,
    "args": {
      "External id": 238395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238395,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238395, "pid": 5, "tid": 7, "ts": 1716454225224696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170409, "dur": 10,
    "args": {
      "External id": 238395,
      "cbid": 211, "correlation": 238395
    }
  },
  {
    "ph": "s", "id": 238395, "pid": 76337, "tid": -914061504, "ts": 1716454225170409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224709, "dur": 12,
    "args": {
      "External id": 238415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238415,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238415, "pid": 5, "tid": 7, "ts": 1716454225224709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170478, "dur": 11,
    "args": {
      "External id": 238415,
      "cbid": 211, "correlation": 238415
    }
  },
  {
    "ph": "s", "id": 238415, "pid": 76337, "tid": -914061504, "ts": 1716454225170478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225224722, "dur": 4,
    "args": {
      "External id": 238427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238427,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238427, "pid": 5, "tid": 7, "ts": 1716454225224722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170499, "dur": 6,
    "args": {
      "External id": 238427,
      "cbid": 211, "correlation": 238427
    }
  },
  {
    "ph": "s", "id": 238427, "pid": 76337, "tid": -914061504, "ts": 1716454225170499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225224728, "dur": 11,
    "args": {
      "External id": 238430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238430, "pid": 5, "tid": 7, "ts": 1716454225224728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170516, "dur": 7,
    "args": {
      "External id": 238430,
      "cbid": 211, "correlation": 238430
    }
  },
  {
    "ph": "s", "id": 238430, "pid": 76337, "tid": -914061504, "ts": 1716454225170516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225224740, "dur": 6,
    "args": {
      "External id": 238439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238439,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238439, "pid": 5, "tid": 7, "ts": 1716454225224740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170555, "dur": 10,
    "args": {
      "External id": 238439,
      "cbid": 211, "correlation": 238439
    }
  },
  {
    "ph": "s", "id": 238439, "pid": 76337, "tid": -914061504, "ts": 1716454225170555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170607, "dur": 0,
    "args": {
      "External id": 238449,
      "cbid": 317, "correlation": 238449
    }
  },
  {
    "ph": "f", "id": 238449, "pid": 76337, "tid": -914061504, "ts": 1716454225170607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170608, "dur": 0,
    "args": {
      "External id": 238450,
      "cbid": 203, "correlation": 238450
    }
  },
  {
    "ph": "f", "id": 238450, "pid": 76337, "tid": -914061504, "ts": 1716454225170608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170609, "dur": 0,
    "args": {
      "External id": 238451,
      "cbid": 205, "correlation": 238451
    }
  },
  {
    "ph": "f", "id": 238451, "pid": 76337, "tid": -914061504, "ts": 1716454225170609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224747, "dur": 6,
    "args": {
      "External id": 238455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238455,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238455, "pid": 5, "tid": 7, "ts": 1716454225224747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170622, "dur": 12,
    "args": {
      "External id": 238455,
      "cbid": 211, "correlation": 238455
    }
  },
  {
    "ph": "s", "id": 238455, "pid": 76337, "tid": -914061504, "ts": 1716454225170622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225224755, "dur": 314,
    "args": {
      "External id": 238457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238457,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238457, "pid": 5, "tid": 7, "ts": 1716454225224755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170637, "dur": 5,
    "args": {
      "External id": 238457,
      "cbid": 211, "correlation": 238457
    }
  },
  {
    "ph": "s", "id": 238457, "pid": 76337, "tid": -914061504, "ts": 1716454225170637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225225071, "dur": 1,
    "args": {
      "External id": 238459,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238459,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 238459, "pid": 5, "tid": 7, "ts": 1716454225225071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170647, "dur": 6,
    "args": {
      "External id": 238459,
      "cbid": 51, "correlation": 238459
    }
  },
  {
    "ph": "s", "id": 238459, "pid": 76337, "tid": -914061504, "ts": 1716454225170647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225225075, "dur": 488,
    "args": {
      "External id": 238460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238460,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238460, "pid": 5, "tid": 7, "ts": 1716454225225075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170654, "dur": 6,
    "args": {
      "External id": 238460,
      "cbid": 211, "correlation": 238460
    }
  },
  {
    "ph": "s", "id": 238460, "pid": 76337, "tid": -914061504, "ts": 1716454225170654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225225564, "dur": 6,
    "args": {
      "External id": 238462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238462,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238462, "pid": 5, "tid": 7, "ts": 1716454225225564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170665, "dur": 5,
    "args": {
      "External id": 238462,
      "cbid": 211, "correlation": 238462
    }
  },
  {
    "ph": "s", "id": 238462, "pid": 76337, "tid": -914061504, "ts": 1716454225170665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225225571, "dur": 6,
    "args": {
      "External id": 238468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238468, "pid": 5, "tid": 7, "ts": 1716454225225571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170693, "dur": 8,
    "args": {
      "External id": 238468,
      "cbid": 211, "correlation": 238468
    }
  },
  {
    "ph": "s", "id": 238468, "pid": 76337, "tid": -914061504, "ts": 1716454225170693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225225578, "dur": 3,
    "args": {
      "External id": 238476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238476,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 238476, "pid": 5, "tid": 7, "ts": 1716454225225578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170736, "dur": 9,
    "args": {
      "External id": 238476,
      "cbid": 211, "correlation": 238476
    }
  },
  {
    "ph": "s", "id": 238476, "pid": 76337, "tid": -914061504, "ts": 1716454225170736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170798, "dur": 1,
    "args": {
      "External id": 238492,
      "cbid": 251, "correlation": 238492
    }
  },
  {
    "ph": "f", "id": 238492, "pid": 76337, "tid": -914061504, "ts": 1716454225170798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170803, "dur": 0,
    "args": {
      "External id": 238494,
      "cbid": 251, "correlation": 238494
    }
  },
  {
    "ph": "f", "id": 238494, "pid": 76337, "tid": -914061504, "ts": 1716454225170803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225225582, "dur": 12,
    "args": {
      "External id": 238495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238495,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238495, "pid": 5, "tid": 7, "ts": 1716454225225582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170805, "dur": 11,
    "args": {
      "External id": 238495,
      "cbid": 211, "correlation": 238495
    }
  },
  {
    "ph": "s", "id": 238495, "pid": 76337, "tid": -914061504, "ts": 1716454225170805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225225596, "dur": 5,
    "args": {
      "External id": 238497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238497,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238497, "pid": 5, "tid": 7, "ts": 1716454225225596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170817, "dur": 6,
    "args": {
      "External id": 238497,
      "cbid": 211, "correlation": 238497
    }
  },
  {
    "ph": "s", "id": 238497, "pid": 76337, "tid": -914061504, "ts": 1716454225170817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225225602, "dur": 6,
    "args": {
      "External id": 238507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238507, "pid": 5, "tid": 7, "ts": 1716454225225602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170874, "dur": 12,
    "args": {
      "External id": 238507,
      "cbid": 211, "correlation": 238507
    }
  },
  {
    "ph": "s", "id": 238507, "pid": 76337, "tid": -914061504, "ts": 1716454225170874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225225609, "dur": 9,
    "args": {
      "External id": 238527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238527,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238527, "pid": 5, "tid": 7, "ts": 1716454225225609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170939, "dur": 10,
    "args": {
      "External id": 238527,
      "cbid": 211, "correlation": 238527
    }
  },
  {
    "ph": "s", "id": 238527, "pid": 76337, "tid": -914061504, "ts": 1716454225170939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225225620, "dur": 4,
    "args": {
      "External id": 238539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238539,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 238539, "pid": 5, "tid": 7, "ts": 1716454225225620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170959, "dur": 6,
    "args": {
      "External id": 238539,
      "cbid": 211, "correlation": 238539
    }
  },
  {
    "ph": "s", "id": 238539, "pid": 76337, "tid": -914061504, "ts": 1716454225170959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225225625, "dur": 7,
    "args": {
      "External id": 238542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238542, "pid": 5, "tid": 7, "ts": 1716454225225625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225170986, "dur": 7,
    "args": {
      "External id": 238542,
      "cbid": 211, "correlation": 238542
    }
  },
  {
    "ph": "s", "id": 238542, "pid": 76337, "tid": -914061504, "ts": 1716454225170986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225225633, "dur": 4,
    "args": {
      "External id": 238551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238551,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238551, "pid": 5, "tid": 7, "ts": 1716454225225633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171027, "dur": 9,
    "args": {
      "External id": 238551,
      "cbid": 211, "correlation": 238551
    }
  },
  {
    "ph": "s", "id": 238551, "pid": 76337, "tid": -914061504, "ts": 1716454225171027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171089, "dur": 0,
    "args": {
      "External id": 238561,
      "cbid": 317, "correlation": 238561
    }
  },
  {
    "ph": "f", "id": 238561, "pid": 76337, "tid": -914061504, "ts": 1716454225171089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171090, "dur": 0,
    "args": {
      "External id": 238562,
      "cbid": 203, "correlation": 238562
    }
  },
  {
    "ph": "f", "id": 238562, "pid": 76337, "tid": -914061504, "ts": 1716454225171090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171090, "dur": 0,
    "args": {
      "External id": 238563,
      "cbid": 205, "correlation": 238563
    }
  },
  {
    "ph": "f", "id": 238563, "pid": 76337, "tid": -914061504, "ts": 1716454225171090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225225639, "dur": 5,
    "args": {
      "External id": 238567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238567,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238567, "pid": 5, "tid": 7, "ts": 1716454225225639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171104, "dur": 12,
    "args": {
      "External id": 238567,
      "cbid": 211, "correlation": 238567
    }
  },
  {
    "ph": "s", "id": 238567, "pid": 76337, "tid": -914061504, "ts": 1716454225171104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225225645, "dur": 159,
    "args": {
      "External id": 238569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238569,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238569, "pid": 5, "tid": 7, "ts": 1716454225225645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171118, "dur": 5,
    "args": {
      "External id": 238569,
      "cbid": 211, "correlation": 238569
    }
  },
  {
    "ph": "s", "id": 238569, "pid": 76337, "tid": -914061504, "ts": 1716454225171118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225225806, "dur": 1,
    "args": {
      "External id": 238571,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238571,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15296367112810708
    }
  },
  {
    "ph": "f", "id": 238571, "pid": 5, "tid": 7, "ts": 1716454225225806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171129, "dur": 6,
    "args": {
      "External id": 238571,
      "cbid": 51, "correlation": 238571
    }
  },
  {
    "ph": "s", "id": 238571, "pid": 76337, "tid": -914061504, "ts": 1716454225171129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225225810, "dur": 254,
    "args": {
      "External id": 238572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238572,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238572, "pid": 5, "tid": 7, "ts": 1716454225225810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171136, "dur": 6,
    "args": {
      "External id": 238572,
      "cbid": 211, "correlation": 238572
    }
  },
  {
    "ph": "s", "id": 238572, "pid": 76337, "tid": -914061504, "ts": 1716454225171136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225226065, "dur": 5,
    "args": {
      "External id": 238574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238574,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238574, "pid": 5, "tid": 7, "ts": 1716454225226065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171146, "dur": 5,
    "args": {
      "External id": 238574,
      "cbid": 211, "correlation": 238574
    }
  },
  {
    "ph": "s", "id": 238574, "pid": 76337, "tid": -914061504, "ts": 1716454225171146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225226072, "dur": 6,
    "args": {
      "External id": 238580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238580,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238580, "pid": 5, "tid": 7, "ts": 1716454225226072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171174, "dur": 9,
    "args": {
      "External id": 238580,
      "cbid": 211, "correlation": 238580
    }
  },
  {
    "ph": "s", "id": 238580, "pid": 76337, "tid": -914061504, "ts": 1716454225171174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171233, "dur": 0,
    "args": {
      "External id": 238590,
      "cbid": 317, "correlation": 238590
    }
  },
  {
    "ph": "f", "id": 238590, "pid": 76337, "tid": -914061504, "ts": 1716454225171233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171234, "dur": 0,
    "args": {
      "External id": 238591,
      "cbid": 203, "correlation": 238591
    }
  },
  {
    "ph": "f", "id": 238591, "pid": 76337, "tid": -914061504, "ts": 1716454225171234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171235, "dur": 0,
    "args": {
      "External id": 238592,
      "cbid": 205, "correlation": 238592
    }
  },
  {
    "ph": "f", "id": 238592, "pid": 76337, "tid": -914061504, "ts": 1716454225171235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225226080, "dur": 8,
    "args": {
      "External id": 238596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238596, "pid": 5, "tid": 7, "ts": 1716454225226080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171246, "dur": 11,
    "args": {
      "External id": 238596,
      "cbid": 211, "correlation": 238596
    }
  },
  {
    "ph": "s", "id": 238596, "pid": 76337, "tid": -914061504, "ts": 1716454225171246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225226088, "dur": 3,
    "args": {
      "External id": 238598,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238598,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5151515151515151
    }
  },
  {
    "ph": "f", "id": 238598, "pid": 5, "tid": 7, "ts": 1716454225226088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171262, "dur": 9,
    "args": {
      "External id": 238598,
      "cbid": 51, "correlation": 238598
    }
  },
  {
    "ph": "s", "id": 238598, "pid": 76337, "tid": -914061504, "ts": 1716454225171262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225226092, "dur": 93,
    "args": {
      "External id": 238599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238599,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 238599, "pid": 5, "tid": 7, "ts": 1716454225226092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171272, "dur": 7,
    "args": {
      "External id": 238599,
      "cbid": 211, "correlation": 238599
    }
  },
  {
    "ph": "s", "id": 238599, "pid": 76337, "tid": -914061504, "ts": 1716454225171272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225226187, "dur": 6,
    "args": {
      "External id": 238601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238601,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238601, "pid": 5, "tid": 7, "ts": 1716454225226187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171282, "dur": 5,
    "args": {
      "External id": 238601,
      "cbid": 211, "correlation": 238601
    }
  },
  {
    "ph": "s", "id": 238601, "pid": 76337, "tid": -914061504, "ts": 1716454225171282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225226194, "dur": 6,
    "args": {
      "External id": 238607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238607, "pid": 5, "tid": 7, "ts": 1716454225226194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171308, "dur": 8,
    "args": {
      "External id": 238607,
      "cbid": 211, "correlation": 238607
    }
  },
  {
    "ph": "s", "id": 238607, "pid": 76337, "tid": -914061504, "ts": 1716454225171308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225226201, "dur": 5,
    "args": {
      "External id": 238615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238615,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238615, "pid": 5, "tid": 7, "ts": 1716454225226201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171338, "dur": 8,
    "args": {
      "External id": 238615,
      "cbid": 211, "correlation": 238615
    }
  },
  {
    "ph": "s", "id": 238615, "pid": 76337, "tid": -914061504, "ts": 1716454225171338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225226207, "dur": 4,
    "args": {
      "External id": 238623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238623,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 238623, "pid": 5, "tid": 7, "ts": 1716454225226207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171366, "dur": 8,
    "args": {
      "External id": 238623,
      "cbid": 211, "correlation": 238623
    }
  },
  {
    "ph": "s", "id": 238623, "pid": 76337, "tid": -914061504, "ts": 1716454225171366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225226213, "dur": 14,
    "args": {
      "External id": 238634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238634,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238634, "pid": 5, "tid": 7, "ts": 1716454225226213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171445, "dur": 12,
    "args": {
      "External id": 238634,
      "cbid": 211, "correlation": 238634
    }
  },
  {
    "ph": "s", "id": 238634, "pid": 76337, "tid": -914061504, "ts": 1716454225171445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171500, "dur": 0,
    "args": {
      "External id": 238644,
      "cbid": 317, "correlation": 238644
    }
  },
  {
    "ph": "f", "id": 238644, "pid": 76337, "tid": -914061504, "ts": 1716454225171500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171500, "dur": 0,
    "args": {
      "External id": 238645,
      "cbid": 203, "correlation": 238645
    }
  },
  {
    "ph": "f", "id": 238645, "pid": 76337, "tid": -914061504, "ts": 1716454225171500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171501, "dur": 0,
    "args": {
      "External id": 238646,
      "cbid": 205, "correlation": 238646
    }
  },
  {
    "ph": "f", "id": 238646, "pid": 76337, "tid": -914061504, "ts": 1716454225171501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225226228, "dur": 8,
    "args": {
      "External id": 238650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238650,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238650, "pid": 5, "tid": 7, "ts": 1716454225226228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171514, "dur": 12,
    "args": {
      "External id": 238650,
      "cbid": 211, "correlation": 238650
    }
  },
  {
    "ph": "s", "id": 238650, "pid": 76337, "tid": -914061504, "ts": 1716454225171514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225226238, "dur": 160,
    "args": {
      "External id": 238652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238652,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238652, "pid": 5, "tid": 7, "ts": 1716454225226238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171529, "dur": 5,
    "args": {
      "External id": 238652,
      "cbid": 211, "correlation": 238652
    }
  },
  {
    "ph": "s", "id": 238652, "pid": 76337, "tid": -914061504, "ts": 1716454225171529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225226399, "dur": 1,
    "args": {
      "External id": 238654,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238654,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 238654, "pid": 5, "tid": 7, "ts": 1716454225226399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171539, "dur": 7,
    "args": {
      "External id": 238654,
      "cbid": 51, "correlation": 238654
    }
  },
  {
    "ph": "s", "id": 238654, "pid": 76337, "tid": -914061504, "ts": 1716454225171539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225226403, "dur": 641,
    "args": {
      "External id": 238655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238655,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238655, "pid": 5, "tid": 7, "ts": 1716454225226403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171548, "dur": 6,
    "args": {
      "External id": 238655,
      "cbid": 211, "correlation": 238655
    }
  },
  {
    "ph": "s", "id": 238655, "pid": 76337, "tid": -914061504, "ts": 1716454225171548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225227046, "dur": 12,
    "args": {
      "External id": 238657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238657,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238657, "pid": 5, "tid": 7, "ts": 1716454225227046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171558, "dur": 5,
    "args": {
      "External id": 238657,
      "cbid": 211, "correlation": 238657
    }
  },
  {
    "ph": "s", "id": 238657, "pid": 76337, "tid": -914061504, "ts": 1716454225171558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225227059, "dur": 14,
    "args": {
      "External id": 238663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238663,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238663, "pid": 5, "tid": 7, "ts": 1716454225227059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171587, "dur": 9,
    "args": {
      "External id": 238663,
      "cbid": 211, "correlation": 238663
    }
  },
  {
    "ph": "s", "id": 238663, "pid": 76337, "tid": -914061504, "ts": 1716454225171587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225227075, "dur": 31,
    "args": {
      "External id": 238672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238672,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238672, "pid": 5, "tid": 7, "ts": 1716454225227075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171678, "dur": 12,
    "args": {
      "External id": 238672,
      "cbid": 211, "correlation": 238672
    }
  },
  {
    "ph": "s", "id": 238672, "pid": 76337, "tid": -914061504, "ts": 1716454225171678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225227107, "dur": 29,
    "args": {
      "External id": 238692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238692,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238692, "pid": 5, "tid": 7, "ts": 1716454225227107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171745, "dur": 11,
    "args": {
      "External id": 238692,
      "cbid": 211, "correlation": 238692
    }
  },
  {
    "ph": "s", "id": 238692, "pid": 76337, "tid": -914061504, "ts": 1716454225171745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225227136, "dur": 4,
    "args": {
      "External id": 238704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238704,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238704, "pid": 5, "tid": 7, "ts": 1716454225227136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171765, "dur": 6,
    "args": {
      "External id": 238704,
      "cbid": 211, "correlation": 238704
    }
  },
  {
    "ph": "s", "id": 238704, "pid": 76337, "tid": -914061504, "ts": 1716454225171765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225227142, "dur": 30,
    "args": {
      "External id": 238707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238707,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238707, "pid": 5, "tid": 7, "ts": 1716454225227142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171784, "dur": 6,
    "args": {
      "External id": 238707,
      "cbid": 211, "correlation": 238707
    }
  },
  {
    "ph": "s", "id": 238707, "pid": 76337, "tid": -914061504, "ts": 1716454225171784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225227173, "dur": 21,
    "args": {
      "External id": 238716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238716,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238716, "pid": 5, "tid": 7, "ts": 1716454225227173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171822, "dur": 10,
    "args": {
      "External id": 238716,
      "cbid": 211, "correlation": 238716
    }
  },
  {
    "ph": "s", "id": 238716, "pid": 76337, "tid": -914061504, "ts": 1716454225171822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171874, "dur": 0,
    "args": {
      "External id": 238726,
      "cbid": 317, "correlation": 238726
    }
  },
  {
    "ph": "f", "id": 238726, "pid": 76337, "tid": -914061504, "ts": 1716454225171874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171875, "dur": 0,
    "args": {
      "External id": 238727,
      "cbid": 203, "correlation": 238727
    }
  },
  {
    "ph": "f", "id": 238727, "pid": 76337, "tid": -914061504, "ts": 1716454225171875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171876, "dur": 0,
    "args": {
      "External id": 238728,
      "cbid": 205, "correlation": 238728
    }
  },
  {
    "ph": "f", "id": 238728, "pid": 76337, "tid": -914061504, "ts": 1716454225171876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225227195, "dur": 22,
    "args": {
      "External id": 238732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238732, "pid": 5, "tid": 7, "ts": 1716454225227195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171891, "dur": 11,
    "args": {
      "External id": 238732,
      "cbid": 211, "correlation": 238732
    }
  },
  {
    "ph": "s", "id": 238732, "pid": 76337, "tid": -914061504, "ts": 1716454225171891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225227219, "dur": 313,
    "args": {
      "External id": 238734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238734, "pid": 5, "tid": 7, "ts": 1716454225227219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171905, "dur": 5,
    "args": {
      "External id": 238734,
      "cbid": 211, "correlation": 238734
    }
  },
  {
    "ph": "s", "id": 238734, "pid": 76337, "tid": -914061504, "ts": 1716454225171905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225227534, "dur": 1,
    "args": {
      "External id": 238736,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238736,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 238736, "pid": 5, "tid": 7, "ts": 1716454225227534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171915, "dur": 7,
    "args": {
      "External id": 238736,
      "cbid": 51, "correlation": 238736
    }
  },
  {
    "ph": "s", "id": 238736, "pid": 76337, "tid": -914061504, "ts": 1716454225171915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225227538, "dur": 1225,
    "args": {
      "External id": 238737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238737,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238737, "pid": 5, "tid": 7, "ts": 1716454225227538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171923, "dur": 6,
    "args": {
      "External id": 238737,
      "cbid": 211, "correlation": 238737
    }
  },
  {
    "ph": "s", "id": 238737, "pid": 76337, "tid": -914061504, "ts": 1716454225171923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225228764, "dur": 12,
    "args": {
      "External id": 238739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238739, "pid": 5, "tid": 7, "ts": 1716454225228764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171933, "dur": 5,
    "args": {
      "External id": 238739,
      "cbid": 211, "correlation": 238739
    }
  },
  {
    "ph": "s", "id": 238739, "pid": 76337, "tid": -914061504, "ts": 1716454225171933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225228778, "dur": 14,
    "args": {
      "External id": 238745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238745, "pid": 5, "tid": 7, "ts": 1716454225228778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225171961, "dur": 8,
    "args": {
      "External id": 238745,
      "cbid": 211, "correlation": 238745
    }
  },
  {
    "ph": "s", "id": 238745, "pid": 76337, "tid": -914061504, "ts": 1716454225171961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225228793, "dur": 3,
    "args": {
      "External id": 238753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238753,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 238753, "pid": 5, "tid": 7, "ts": 1716454225228793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172012, "dur": 10,
    "args": {
      "External id": 238753,
      "cbid": 211, "correlation": 238753
    }
  },
  {
    "ph": "s", "id": 238753, "pid": 76337, "tid": -914061504, "ts": 1716454225172012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172076, "dur": 1,
    "args": {
      "External id": 238769,
      "cbid": 251, "correlation": 238769
    }
  },
  {
    "ph": "f", "id": 238769, "pid": 76337, "tid": -914061504, "ts": 1716454225172076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172082, "dur": 0,
    "args": {
      "External id": 238771,
      "cbid": 251, "correlation": 238771
    }
  },
  {
    "ph": "f", "id": 238771, "pid": 76337, "tid": -914061504, "ts": 1716454225172082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225228798, "dur": 12,
    "args": {
      "External id": 238772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238772,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238772, "pid": 5, "tid": 7, "ts": 1716454225228798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172084, "dur": 11,
    "args": {
      "External id": 238772,
      "cbid": 211, "correlation": 238772
    }
  },
  {
    "ph": "s", "id": 238772, "pid": 76337, "tid": -914061504, "ts": 1716454225172084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225228811, "dur": 5,
    "args": {
      "External id": 238774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238774,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238774, "pid": 5, "tid": 7, "ts": 1716454225228811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172097, "dur": 5,
    "args": {
      "External id": 238774,
      "cbid": 211, "correlation": 238774
    }
  },
  {
    "ph": "s", "id": 238774, "pid": 76337, "tid": -914061504, "ts": 1716454225172097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225228817, "dur": 18,
    "args": {
      "External id": 238784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238784, "pid": 5, "tid": 7, "ts": 1716454225228817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172154, "dur": 12,
    "args": {
      "External id": 238784,
      "cbid": 211, "correlation": 238784
    }
  },
  {
    "ph": "s", "id": 238784, "pid": 76337, "tid": -914061504, "ts": 1716454225172154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225228836, "dur": 17,
    "args": {
      "External id": 238804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238804,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238804, "pid": 5, "tid": 7, "ts": 1716454225228836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172219, "dur": 12,
    "args": {
      "External id": 238804,
      "cbid": 211, "correlation": 238804
    }
  },
  {
    "ph": "s", "id": 238804, "pid": 76337, "tid": -914061504, "ts": 1716454225172219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225228854, "dur": 4,
    "args": {
      "External id": 238816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238816,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 238816, "pid": 5, "tid": 7, "ts": 1716454225228854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172241, "dur": 6,
    "args": {
      "External id": 238816,
      "cbid": 211, "correlation": 238816
    }
  },
  {
    "ph": "s", "id": 238816, "pid": 76337, "tid": -914061504, "ts": 1716454225172241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225228860, "dur": 16,
    "args": {
      "External id": 238819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238819, "pid": 5, "tid": 7, "ts": 1716454225228860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172259, "dur": 6,
    "args": {
      "External id": 238819,
      "cbid": 211, "correlation": 238819
    }
  },
  {
    "ph": "s", "id": 238819, "pid": 76337, "tid": -914061504, "ts": 1716454225172259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225228877, "dur": 11,
    "args": {
      "External id": 238828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238828,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238828, "pid": 5, "tid": 7, "ts": 1716454225228877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172299, "dur": 10,
    "args": {
      "External id": 238828,
      "cbid": 211, "correlation": 238828
    }
  },
  {
    "ph": "s", "id": 238828, "pid": 76337, "tid": -914061504, "ts": 1716454225172299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172360, "dur": 0,
    "args": {
      "External id": 238838,
      "cbid": 317, "correlation": 238838
    }
  },
  {
    "ph": "f", "id": 238838, "pid": 76337, "tid": -914061504, "ts": 1716454225172360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172361, "dur": 0,
    "args": {
      "External id": 238839,
      "cbid": 203, "correlation": 238839
    }
  },
  {
    "ph": "f", "id": 238839, "pid": 76337, "tid": -914061504, "ts": 1716454225172361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172362, "dur": 0,
    "args": {
      "External id": 238840,
      "cbid": 205, "correlation": 238840
    }
  },
  {
    "ph": "f", "id": 238840, "pid": 76337, "tid": -914061504, "ts": 1716454225172362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225228890, "dur": 11,
    "args": {
      "External id": 238844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238844,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238844, "pid": 5, "tid": 7, "ts": 1716454225228890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172375, "dur": 12,
    "args": {
      "External id": 238844,
      "cbid": 211, "correlation": 238844
    }
  },
  {
    "ph": "s", "id": 238844, "pid": 76337, "tid": -914061504, "ts": 1716454225172375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225228902, "dur": 159,
    "args": {
      "External id": 238846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238846, "pid": 5, "tid": 7, "ts": 1716454225228902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172390, "dur": 5,
    "args": {
      "External id": 238846,
      "cbid": 211, "correlation": 238846
    }
  },
  {
    "ph": "s", "id": 238846, "pid": 76337, "tid": -914061504, "ts": 1716454225172390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225229063, "dur": 1,
    "args": {
      "External id": 238848,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 238848,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 238848, "pid": 5, "tid": 7, "ts": 1716454225229063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172401, "dur": 6,
    "args": {
      "External id": 238848,
      "cbid": 51, "correlation": 238848
    }
  },
  {
    "ph": "s", "id": 238848, "pid": 76337, "tid": -914061504, "ts": 1716454225172401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225229067, "dur": 640,
    "args": {
      "External id": 238849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238849,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 238849, "pid": 5, "tid": 7, "ts": 1716454225229067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172408, "dur": 6,
    "args": {
      "External id": 238849,
      "cbid": 211, "correlation": 238849
    }
  },
  {
    "ph": "s", "id": 238849, "pid": 76337, "tid": -914061504, "ts": 1716454225172408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225229709, "dur": 12,
    "args": {
      "External id": 238851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238851, "pid": 5, "tid": 7, "ts": 1716454225229709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172418, "dur": 5,
    "args": {
      "External id": 238851,
      "cbid": 211, "correlation": 238851
    }
  },
  {
    "ph": "s", "id": 238851, "pid": 76337, "tid": -914061504, "ts": 1716454225172418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225229722, "dur": 14,
    "args": {
      "External id": 238857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238857, "pid": 5, "tid": 7, "ts": 1716454225229722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172446, "dur": 8,
    "args": {
      "External id": 238857,
      "cbid": 211, "correlation": 238857
    }
  },
  {
    "ph": "s", "id": 238857, "pid": 76337, "tid": -914061504, "ts": 1716454225172446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172505, "dur": 0,
    "args": {
      "External id": 238867,
      "cbid": 317, "correlation": 238867
    }
  },
  {
    "ph": "f", "id": 238867, "pid": 76337, "tid": -914061504, "ts": 1716454225172505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172506, "dur": 0,
    "args": {
      "External id": 238868,
      "cbid": 203, "correlation": 238868
    }
  },
  {
    "ph": "f", "id": 238868, "pid": 76337, "tid": -914061504, "ts": 1716454225172506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172506, "dur": 0,
    "args": {
      "External id": 238869,
      "cbid": 205, "correlation": 238869
    }
  },
  {
    "ph": "f", "id": 238869, "pid": 76337, "tid": -914061504, "ts": 1716454225172506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225229737, "dur": 21,
    "args": {
      "External id": 238873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238873,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238873, "pid": 5, "tid": 7, "ts": 1716454225229737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172519, "dur": 12,
    "args": {
      "External id": 238873,
      "cbid": 211, "correlation": 238873
    }
  },
  {
    "ph": "s", "id": 238873, "pid": 76337, "tid": -914061504, "ts": 1716454225172519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225229760, "dur": 4,
    "args": {
      "External id": 238875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 238875, "pid": 5, "tid": 7, "ts": 1716454225229760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172537, "dur": 6,
    "args": {
      "External id": 238875,
      "cbid": 211, "correlation": 238875
    }
  },
  {
    "ph": "s", "id": 238875, "pid": 76337, "tid": -914061504, "ts": 1716454225172537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172546, "dur": 0,
    "args": {
      "External id": 238876,
      "cbid": 51, "correlation": 238876
    }
  },
  {
    "ph": "s", "id": 238876, "pid": 76337, "tid": -914061504, "ts": 1716454225172546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225229765, "dur": 172,
    "args": {
      "External id": 238877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238877,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 238877, "pid": 5, "tid": 7, "ts": 1716454225229765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172547, "dur": 6,
    "args": {
      "External id": 238877,
      "cbid": 211, "correlation": 238877
    }
  },
  {
    "ph": "s", "id": 238877, "pid": 76337, "tid": -914061504, "ts": 1716454225172547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225229938, "dur": 15,
    "args": {
      "External id": 238882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238882, "pid": 5, "tid": 7, "ts": 1716454225229938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172573, "dur": 9,
    "args": {
      "External id": 238882,
      "cbid": 211, "correlation": 238882
    }
  },
  {
    "ph": "s", "id": 238882, "pid": 76337, "tid": -914061504, "ts": 1716454225172573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225229954, "dur": 12,
    "args": {
      "External id": 238890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238890,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238890, "pid": 5, "tid": 7, "ts": 1716454225229954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172602, "dur": 8,
    "args": {
      "External id": 238890,
      "cbid": 211, "correlation": 238890
    }
  },
  {
    "ph": "s", "id": 238890, "pid": 76337, "tid": -914061504, "ts": 1716454225172602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225229967, "dur": 10,
    "args": {
      "External id": 238898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238898,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238898, "pid": 5, "tid": 7, "ts": 1716454225229967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172632, "dur": 8,
    "args": {
      "External id": 238898,
      "cbid": 211, "correlation": 238898
    }
  },
  {
    "ph": "s", "id": 238898, "pid": 76337, "tid": -914061504, "ts": 1716454225172632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225229978, "dur": 19,
    "args": {
      "External id": 238918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238918,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 238918, "pid": 5, "tid": 7, "ts": 1716454225229978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172712, "dur": 12,
    "args": {
      "External id": 238918,
      "cbid": 211, "correlation": 238918
    }
  },
  {
    "ph": "s", "id": 238918, "pid": 76337, "tid": -914061504, "ts": 1716454225172712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225229998, "dur": 4,
    "args": {
      "External id": 238930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238930,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 238930, "pid": 5, "tid": 7, "ts": 1716454225229998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172735, "dur": 6,
    "args": {
      "External id": 238930,
      "cbid": 211, "correlation": 238930
    }
  },
  {
    "ph": "s", "id": 238930, "pid": 76337, "tid": -914061504, "ts": 1716454225172735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225230004, "dur": 17,
    "args": {
      "External id": 238933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238933,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238933, "pid": 5, "tid": 7, "ts": 1716454225230004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172753, "dur": 6,
    "args": {
      "External id": 238933,
      "cbid": 211, "correlation": 238933
    }
  },
  {
    "ph": "s", "id": 238933, "pid": 76337, "tid": -914061504, "ts": 1716454225172753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172809, "dur": 0,
    "args": {
      "External id": 238944,
      "cbid": 317, "correlation": 238944
    }
  },
  {
    "ph": "f", "id": 238944, "pid": 76337, "tid": -914061504, "ts": 1716454225172809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172810, "dur": 0,
    "args": {
      "External id": 238945,
      "cbid": 203, "correlation": 238945
    }
  },
  {
    "ph": "f", "id": 238945, "pid": 76337, "tid": -914061504, "ts": 1716454225172810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172811, "dur": 0,
    "args": {
      "External id": 238946,
      "cbid": 205, "correlation": 238946
    }
  },
  {
    "ph": "f", "id": 238946, "pid": 76337, "tid": -914061504, "ts": 1716454225172811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225230022, "dur": 11,
    "args": {
      "External id": 238950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238950,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238950, "pid": 5, "tid": 7, "ts": 1716454225230022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172824, "dur": 12,
    "args": {
      "External id": 238950,
      "cbid": 211, "correlation": 238950
    }
  },
  {
    "ph": "s", "id": 238950, "pid": 76337, "tid": -914061504, "ts": 1716454225172824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225230034, "dur": 3,
    "args": {
      "External id": 238952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238952,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 238952, "pid": 5, "tid": 7, "ts": 1716454225230034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172841, "dur": 6,
    "args": {
      "External id": 238952,
      "cbid": 211, "correlation": 238952
    }
  },
  {
    "ph": "s", "id": 238952, "pid": 76337, "tid": -914061504, "ts": 1716454225172841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172849, "dur": 0,
    "args": {
      "External id": 238953,
      "cbid": 51, "correlation": 238953
    }
  },
  {
    "ph": "s", "id": 238953, "pid": 76337, "tid": -914061504, "ts": 1716454225172849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225230039, "dur": 88,
    "args": {
      "External id": 238954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238954,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 238954, "pid": 5, "tid": 7, "ts": 1716454225230039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172850, "dur": 5,
    "args": {
      "External id": 238954,
      "cbid": 211, "correlation": 238954
    }
  },
  {
    "ph": "s", "id": 238954, "pid": 76337, "tid": -914061504, "ts": 1716454225172850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225230129, "dur": 15,
    "args": {
      "External id": 238959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238959, "pid": 5, "tid": 7, "ts": 1716454225230129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172876, "dur": 9,
    "args": {
      "External id": 238959,
      "cbid": 211, "correlation": 238959
    }
  },
  {
    "ph": "s", "id": 238959, "pid": 76337, "tid": -914061504, "ts": 1716454225172876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225230145, "dur": 82,
    "args": {
      "External id": 238968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238968,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238968, "pid": 5, "tid": 7, "ts": 1716454225230145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225172958, "dur": 14,
    "args": {
      "External id": 238968,
      "cbid": 211, "correlation": 238968
    }
  },
  {
    "ph": "s", "id": 238968, "pid": 76337, "tid": -914061504, "ts": 1716454225172958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225230228, "dur": 30,
    "args": {
      "External id": 238990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 238990,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 238990, "pid": 5, "tid": 7, "ts": 1716454225230228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173026, "dur": 11,
    "args": {
      "External id": 238990,
      "cbid": 211, "correlation": 238990
    }
  },
  {
    "ph": "s", "id": 238990, "pid": 76337, "tid": -914061504, "ts": 1716454225173026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173118, "dur": 1,
    "args": {
      "External id": 239001,
      "cbid": 251, "correlation": 239001
    }
  },
  {
    "ph": "f", "id": 239001, "pid": 76337, "tid": -914061504, "ts": 1716454225173118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225230260, "dur": 138,
    "args": {
      "External id": 239002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239002,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239002, "pid": 5, "tid": 7, "ts": 1716454225230260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173123, "dur": 14,
    "args": {
      "External id": 239002,
      "cbid": 211, "correlation": 239002
    }
  },
  {
    "ph": "s", "id": 239002, "pid": 76337, "tid": -914061504, "ts": 1716454225173123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173195, "dur": 1,
    "args": {
      "External id": 239013,
      "cbid": 251, "correlation": 239013
    }
  },
  {
    "ph": "f", "id": 239013, "pid": 76337, "tid": -914061504, "ts": 1716454225173195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225230399, "dur": 156,
    "args": {
      "External id": 239014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239014,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239014, "pid": 5, "tid": 7, "ts": 1716454225230399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173199, "dur": 11,
    "args": {
      "External id": 239014,
      "cbid": 211, "correlation": 239014
    }
  },
  {
    "ph": "s", "id": 239014, "pid": 76337, "tid": -914061504, "ts": 1716454225173199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173265, "dur": 1,
    "args": {
      "External id": 239025,
      "cbid": 251, "correlation": 239025
    }
  },
  {
    "ph": "f", "id": 239025, "pid": 76337, "tid": -914061504, "ts": 1716454225173265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225230557, "dur": 156,
    "args": {
      "External id": 239026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239026,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239026, "pid": 5, "tid": 7, "ts": 1716454225230557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173269, "dur": 12,
    "args": {
      "External id": 239026,
      "cbid": 211, "correlation": 239026
    }
  },
  {
    "ph": "s", "id": 239026, "pid": 76337, "tid": -914061504, "ts": 1716454225173269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225230714, "dur": 333,
    "args": {
      "External id": 239051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239051,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239051, "pid": 5, "tid": 7, "ts": 1716454225230714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173356, "dur": 13,
    "args": {
      "External id": 239051,
      "cbid": 211, "correlation": 239051
    }
  },
  {
    "ph": "s", "id": 239051, "pid": 76337, "tid": -914061504, "ts": 1716454225173356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173456, "dur": 1,
    "args": {
      "External id": 239069,
      "cbid": 251, "correlation": 239069
    }
  },
  {
    "ph": "f", "id": 239069, "pid": 76337, "tid": -914061504, "ts": 1716454225173456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225231048, "dur": 164,
    "args": {
      "External id": 239071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239071,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239071, "pid": 5, "tid": 7, "ts": 1716454225231048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173462, "dur": 13,
    "args": {
      "External id": 239071,
      "cbid": 211, "correlation": 239071
    }
  },
  {
    "ph": "s", "id": 239071, "pid": 76337, "tid": -914061504, "ts": 1716454225173462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225231214, "dur": 19,
    "args": {
      "External id": 239079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239079,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239079, "pid": 5, "tid": 7, "ts": 1716454225231214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173533, "dur": 12,
    "args": {
      "External id": 239079,
      "cbid": 211, "correlation": 239079
    }
  },
  {
    "ph": "s", "id": 239079, "pid": 76337, "tid": -914061504, "ts": 1716454225173533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225231234, "dur": 27,
    "args": {
      "External id": 239087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239087,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239087, "pid": 5, "tid": 7, "ts": 1716454225231234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173571, "dur": 9,
    "args": {
      "External id": 239087,
      "cbid": 211, "correlation": 239087
    }
  },
  {
    "ph": "s", "id": 239087, "pid": 76337, "tid": -914061504, "ts": 1716454225173571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225231263, "dur": 18,
    "args": {
      "External id": 239098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239098, "pid": 5, "tid": 7, "ts": 1716454225231263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173644, "dur": 13,
    "args": {
      "External id": 239098,
      "cbid": 211, "correlation": 239098
    }
  },
  {
    "ph": "s", "id": 239098, "pid": 76337, "tid": -914061504, "ts": 1716454225173644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225231282, "dur": 16,
    "args": {
      "External id": 239120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239120,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239120, "pid": 5, "tid": 7, "ts": 1716454225231282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173675, "dur": 8,
    "args": {
      "External id": 239120,
      "cbid": 211, "correlation": 239120
    }
  },
  {
    "ph": "s", "id": 239120, "pid": 76337, "tid": -914061504, "ts": 1716454225173675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173761, "dur": 1,
    "args": {
      "External id": 239131,
      "cbid": 251, "correlation": 239131
    }
  },
  {
    "ph": "f", "id": 239131, "pid": 76337, "tid": -914061504, "ts": 1716454225173761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225231300, "dur": 87,
    "args": {
      "External id": 239132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239132,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239132, "pid": 5, "tid": 7, "ts": 1716454225231300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173766, "dur": 14,
    "args": {
      "External id": 239132,
      "cbid": 211, "correlation": 239132
    }
  },
  {
    "ph": "s", "id": 239132, "pid": 76337, "tid": -914061504, "ts": 1716454225173766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173836, "dur": 1,
    "args": {
      "External id": 239143,
      "cbid": 251, "correlation": 239143
    }
  },
  {
    "ph": "f", "id": 239143, "pid": 76337, "tid": -914061504, "ts": 1716454225173836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173839, "dur": 0,
    "args": {
      "External id": 239144,
      "cbid": 251, "correlation": 239144
    }
  },
  {
    "ph": "f", "id": 239144, "pid": 76337, "tid": -914061504, "ts": 1716454225173839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225231388, "dur": 12,
    "args": {
      "External id": 239145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239145,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239145, "pid": 5, "tid": 7, "ts": 1716454225231388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173841, "dur": 12,
    "args": {
      "External id": 239145,
      "cbid": 211, "correlation": 239145
    }
  },
  {
    "ph": "s", "id": 239145, "pid": 76337, "tid": -914061504, "ts": 1716454225173841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225231401, "dur": 6,
    "args": {
      "External id": 239147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239147,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239147, "pid": 5, "tid": 7, "ts": 1716454225231401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173854, "dur": 6,
    "args": {
      "External id": 239147,
      "cbid": 211, "correlation": 239147
    }
  },
  {
    "ph": "s", "id": 239147, "pid": 76337, "tid": -914061504, "ts": 1716454225173854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173912, "dur": 1,
    "args": {
      "External id": 239158,
      "cbid": 251, "correlation": 239158
    }
  },
  {
    "ph": "f", "id": 239158, "pid": 76337, "tid": -914061504, "ts": 1716454225173912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173915, "dur": 0,
    "args": {
      "External id": 239159,
      "cbid": 251, "correlation": 239159
    }
  },
  {
    "ph": "f", "id": 239159, "pid": 76337, "tid": -914061504, "ts": 1716454225173915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225231408, "dur": 8,
    "args": {
      "External id": 239160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239160,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239160, "pid": 5, "tid": 7, "ts": 1716454225231408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173916, "dur": 12,
    "args": {
      "External id": 239160,
      "cbid": 211, "correlation": 239160
    }
  },
  {
    "ph": "s", "id": 239160, "pid": 76337, "tid": -914061504, "ts": 1716454225173916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225231418, "dur": 3,
    "args": {
      "External id": 239162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239162,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239162, "pid": 5, "tid": 7, "ts": 1716454225231418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225173930, "dur": 5,
    "args": {
      "External id": 239162,
      "cbid": 211, "correlation": 239162
    }
  },
  {
    "ph": "s", "id": 239162, "pid": 76337, "tid": -914061504, "ts": 1716454225173930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225231422, "dur": 55,
    "args": {
      "External id": 239187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239187,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239187, "pid": 5, "tid": 7, "ts": 1716454225231422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174016, "dur": 14,
    "args": {
      "External id": 239187,
      "cbid": 211, "correlation": 239187
    }
  },
  {
    "ph": "s", "id": 239187, "pid": 76337, "tid": -914061504, "ts": 1716454225174016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174116, "dur": 2,
    "args": {
      "External id": 239205,
      "cbid": 251, "correlation": 239205
    }
  },
  {
    "ph": "f", "id": 239205, "pid": 76337, "tid": -914061504, "ts": 1716454225174116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225231478, "dur": 89,
    "args": {
      "External id": 239207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239207,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239207, "pid": 5, "tid": 7, "ts": 1716454225231478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174123, "dur": 14,
    "args": {
      "External id": 239207,
      "cbid": 211, "correlation": 239207
    }
  },
  {
    "ph": "s", "id": 239207, "pid": 76337, "tid": -914061504, "ts": 1716454225174123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225231569, "dur": 9,
    "args": {
      "External id": 239215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239215,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239215, "pid": 5, "tid": 7, "ts": 1716454225231569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174191, "dur": 12,
    "args": {
      "External id": 239215,
      "cbid": 211, "correlation": 239215
    }
  },
  {
    "ph": "s", "id": 239215, "pid": 76337, "tid": -914061504, "ts": 1716454225174191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225231580, "dur": 21,
    "args": {
      "External id": 239223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239223, "pid": 5, "tid": 7, "ts": 1716454225231580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174232, "dur": 9,
    "args": {
      "External id": 239223,
      "cbid": 211, "correlation": 239223
    }
  },
  {
    "ph": "s", "id": 239223, "pid": 76337, "tid": -914061504, "ts": 1716454225174232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225231602, "dur": 17,
    "args": {
      "External id": 239245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239245,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239245, "pid": 5, "tid": 7, "ts": 1716454225231602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174284, "dur": 10,
    "args": {
      "External id": 239245,
      "cbid": 211, "correlation": 239245
    }
  },
  {
    "ph": "s", "id": 239245, "pid": 76337, "tid": -914061504, "ts": 1716454225174284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174371, "dur": 1,
    "args": {
      "External id": 239261,
      "cbid": 251, "correlation": 239261
    }
  },
  {
    "ph": "f", "id": 239261, "pid": 76337, "tid": -914061504, "ts": 1716454225174371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174376, "dur": 0,
    "args": {
      "External id": 239263,
      "cbid": 251, "correlation": 239263
    }
  },
  {
    "ph": "f", "id": 239263, "pid": 76337, "tid": -914061504, "ts": 1716454225174376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225231620, "dur": 491,
    "args": {
      "External id": 239264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239264,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239264, "pid": 5, "tid": 7, "ts": 1716454225231620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174378, "dur": 13,
    "args": {
      "External id": 239264,
      "cbid": 211, "correlation": 239264
    }
  },
  {
    "ph": "s", "id": 239264, "pid": 76337, "tid": -914061504, "ts": 1716454225174378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225232112, "dur": 66,
    "args": {
      "External id": 239272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239272, "pid": 5, "tid": 7, "ts": 1716454225232112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174444, "dur": 12,
    "args": {
      "External id": 239272,
      "cbid": 211, "correlation": 239272
    }
  },
  {
    "ph": "s", "id": 239272, "pid": 76337, "tid": -914061504, "ts": 1716454225174444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225232179, "dur": 66,
    "args": {
      "External id": 239280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239280, "pid": 5, "tid": 7, "ts": 1716454225232179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174477, "dur": 8,
    "args": {
      "External id": 239280,
      "cbid": 211, "correlation": 239280
    }
  },
  {
    "ph": "s", "id": 239280, "pid": 76337, "tid": -914061504, "ts": 1716454225174477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174556, "dur": 1,
    "args": {
      "External id": 239296,
      "cbid": 251, "correlation": 239296
    }
  },
  {
    "ph": "f", "id": 239296, "pid": 76337, "tid": -914061504, "ts": 1716454225174556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225232247, "dur": 1,
    "args": {
      "External id": 239298,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 239298,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 239298, "pid": 5, "tid": 7, "ts": 1716454225232247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174561, "dur": 10,
    "args": {
      "External id": 239298,
      "cbid": 51, "correlation": 239298
    }
  },
  {
    "ph": "s", "id": 239298, "pid": 76337, "tid": -914061504, "ts": 1716454225174561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225232251, "dur": 265,
    "args": {
      "External id": 239299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239299,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239299, "pid": 5, "tid": 7, "ts": 1716454225232251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174573, "dur": 11,
    "args": {
      "External id": 239299,
      "cbid": 211, "correlation": 239299
    }
  },
  {
    "ph": "s", "id": 239299, "pid": 76337, "tid": -914061504, "ts": 1716454225174573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225232518, "dur": 14,
    "args": {
      "External id": 239307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239307,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239307, "pid": 5, "tid": 7, "ts": 1716454225232518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174614, "dur": 11,
    "args": {
      "External id": 239307,
      "cbid": 211, "correlation": 239307
    }
  },
  {
    "ph": "s", "id": 239307, "pid": 76337, "tid": -914061504, "ts": 1716454225174614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225232533, "dur": 37,
    "args": {
      "External id": 239318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239318, "pid": 5, "tid": 7, "ts": 1716454225232533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174683, "dur": 12,
    "args": {
      "External id": 239318,
      "cbid": 211, "correlation": 239318
    }
  },
  {
    "ph": "s", "id": 239318, "pid": 76337, "tid": -914061504, "ts": 1716454225174683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174748, "dur": 0,
    "args": {
      "External id": 239330,
      "cbid": 317, "correlation": 239330
    }
  },
  {
    "ph": "f", "id": 239330, "pid": 76337, "tid": -914061504, "ts": 1716454225174748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174749, "dur": 0,
    "args": {
      "External id": 239331,
      "cbid": 203, "correlation": 239331
    }
  },
  {
    "ph": "f", "id": 239331, "pid": 76337, "tid": -914061504, "ts": 1716454225174749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174749, "dur": 0,
    "args": {
      "External id": 239332,
      "cbid": 205, "correlation": 239332
    }
  },
  {
    "ph": "f", "id": 239332, "pid": 76337, "tid": -914061504, "ts": 1716454225174749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225232571, "dur": 13,
    "args": {
      "External id": 239336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239336, "pid": 5, "tid": 7, "ts": 1716454225232571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174764, "dur": 13,
    "args": {
      "External id": 239336,
      "cbid": 211, "correlation": 239336
    }
  },
  {
    "ph": "s", "id": 239336, "pid": 76337, "tid": -914061504, "ts": 1716454225174764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225232585, "dur": 4,
    "args": {
      "External id": 239338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239338, "pid": 5, "tid": 7, "ts": 1716454225232585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174781, "dur": 6,
    "args": {
      "External id": 239338,
      "cbid": 211, "correlation": 239338
    }
  },
  {
    "ph": "s", "id": 239338, "pid": 76337, "tid": -914061504, "ts": 1716454225174781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174790, "dur": 0,
    "args": {
      "External id": 239339,
      "cbid": 51, "correlation": 239339
    }
  },
  {
    "ph": "s", "id": 239339, "pid": 76337, "tid": -914061504, "ts": 1716454225174790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225232590, "dur": 94,
    "args": {
      "External id": 239340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239340,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 239340, "pid": 5, "tid": 7, "ts": 1716454225232590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174791, "dur": 5,
    "args": {
      "External id": 239340,
      "cbid": 211, "correlation": 239340
    }
  },
  {
    "ph": "s", "id": 239340, "pid": 76337, "tid": -914061504, "ts": 1716454225174791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225232686, "dur": 16,
    "args": {
      "External id": 239345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239345,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239345, "pid": 5, "tid": 7, "ts": 1716454225232686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174819, "dur": 9,
    "args": {
      "External id": 239345,
      "cbid": 211, "correlation": 239345
    }
  },
  {
    "ph": "s", "id": 239345, "pid": 76337, "tid": -914061504, "ts": 1716454225174819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225232703, "dur": 12,
    "args": {
      "External id": 239353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239353,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239353, "pid": 5, "tid": 7, "ts": 1716454225232703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174850, "dur": 8,
    "args": {
      "External id": 239353,
      "cbid": 211, "correlation": 239353
    }
  },
  {
    "ph": "s", "id": 239353, "pid": 76337, "tid": -914061504, "ts": 1716454225174850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225232716, "dur": 30,
    "args": {
      "External id": 239362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239362,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239362, "pid": 5, "tid": 7, "ts": 1716454225232716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174890, "dur": 10,
    "args": {
      "External id": 239362,
      "cbid": 211, "correlation": 239362
    }
  },
  {
    "ph": "s", "id": 239362, "pid": 76337, "tid": -914061504, "ts": 1716454225174890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225232747, "dur": 30,
    "args": {
      "External id": 239382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239382,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 239382, "pid": 5, "tid": 7, "ts": 1716454225232747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174962, "dur": 21,
    "args": {
      "External id": 239382,
      "cbid": 211, "correlation": 239382
    }
  },
  {
    "ph": "s", "id": 239382, "pid": 76337, "tid": -914061504, "ts": 1716454225174962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225232779, "dur": 5,
    "args": {
      "External id": 239394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239394,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239394, "pid": 5, "tid": 7, "ts": 1716454225232779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225174993, "dur": 7,
    "args": {
      "External id": 239394,
      "cbid": 211, "correlation": 239394
    }
  },
  {
    "ph": "s", "id": 239394, "pid": 76337, "tid": -914061504, "ts": 1716454225174993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225232786, "dur": 32,
    "args": {
      "External id": 239397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239397, "pid": 5, "tid": 7, "ts": 1716454225232786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175012, "dur": 8,
    "args": {
      "External id": 239397,
      "cbid": 211, "correlation": 239397
    }
  },
  {
    "ph": "s", "id": 239397, "pid": 76337, "tid": -914061504, "ts": 1716454225175012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225232818, "dur": 20,
    "args": {
      "External id": 239406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239406,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239406, "pid": 5, "tid": 7, "ts": 1716454225232818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175053, "dur": 11,
    "args": {
      "External id": 239406,
      "cbid": 211, "correlation": 239406
    }
  },
  {
    "ph": "s", "id": 239406, "pid": 76337, "tid": -914061504, "ts": 1716454225175053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175106, "dur": 0,
    "args": {
      "External id": 239416,
      "cbid": 317, "correlation": 239416
    }
  },
  {
    "ph": "f", "id": 239416, "pid": 76337, "tid": -914061504, "ts": 1716454225175106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175107, "dur": 0,
    "args": {
      "External id": 239417,
      "cbid": 203, "correlation": 239417
    }
  },
  {
    "ph": "f", "id": 239417, "pid": 76337, "tid": -914061504, "ts": 1716454225175107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175107, "dur": 0,
    "args": {
      "External id": 239418,
      "cbid": 205, "correlation": 239418
    }
  },
  {
    "ph": "f", "id": 239418, "pid": 76337, "tid": -914061504, "ts": 1716454225175107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225232840, "dur": 21,
    "args": {
      "External id": 239422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239422, "pid": 5, "tid": 7, "ts": 1716454225232840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175122, "dur": 11,
    "args": {
      "External id": 239422,
      "cbid": 211, "correlation": 239422
    }
  },
  {
    "ph": "s", "id": 239422, "pid": 76337, "tid": -914061504, "ts": 1716454225175122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225232862, "dur": 313,
    "args": {
      "External id": 239424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239424,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239424, "pid": 5, "tid": 7, "ts": 1716454225232862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175135, "dur": 5,
    "args": {
      "External id": 239424,
      "cbid": 211, "correlation": 239424
    }
  },
  {
    "ph": "s", "id": 239424, "pid": 76337, "tid": -914061504, "ts": 1716454225175135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225233178, "dur": 1,
    "args": {
      "External id": 239426,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 239426,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 239426, "pid": 5, "tid": 7, "ts": 1716454225233178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175146, "dur": 7,
    "args": {
      "External id": 239426,
      "cbid": 51, "correlation": 239426
    }
  },
  {
    "ph": "s", "id": 239426, "pid": 76337, "tid": -914061504, "ts": 1716454225175146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225233181, "dur": 1245,
    "args": {
      "External id": 239427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239427,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239427, "pid": 5, "tid": 7, "ts": 1716454225233181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175154, "dur": 6,
    "args": {
      "External id": 239427,
      "cbid": 211, "correlation": 239427
    }
  },
  {
    "ph": "s", "id": 239427, "pid": 76337, "tid": -914061504, "ts": 1716454225175154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225234428, "dur": 14,
    "args": {
      "External id": 239429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239429,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239429, "pid": 5, "tid": 7, "ts": 1716454225234428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175165, "dur": 6,
    "args": {
      "External id": 239429,
      "cbid": 211, "correlation": 239429
    }
  },
  {
    "ph": "s", "id": 239429, "pid": 76337, "tid": -914061504, "ts": 1716454225175165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225234443, "dur": 15,
    "args": {
      "External id": 239435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239435, "pid": 5, "tid": 7, "ts": 1716454225234443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175194, "dur": 8,
    "args": {
      "External id": 239435,
      "cbid": 211, "correlation": 239435
    }
  },
  {
    "ph": "s", "id": 239435, "pid": 76337, "tid": -914061504, "ts": 1716454225175194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225234459, "dur": 3,
    "args": {
      "External id": 239443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239443,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 239443, "pid": 5, "tid": 7, "ts": 1716454225234459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175238, "dur": 9,
    "args": {
      "External id": 239443,
      "cbid": 211, "correlation": 239443
    }
  },
  {
    "ph": "s", "id": 239443, "pid": 76337, "tid": -914061504, "ts": 1716454225175238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175302, "dur": 1,
    "args": {
      "External id": 239459,
      "cbid": 251, "correlation": 239459
    }
  },
  {
    "ph": "f", "id": 239459, "pid": 76337, "tid": -914061504, "ts": 1716454225175302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175307, "dur": 0,
    "args": {
      "External id": 239461,
      "cbid": 251, "correlation": 239461
    }
  },
  {
    "ph": "f", "id": 239461, "pid": 76337, "tid": -914061504, "ts": 1716454225175307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225234463, "dur": 14,
    "args": {
      "External id": 239462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239462,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239462, "pid": 5, "tid": 7, "ts": 1716454225234463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175309, "dur": 12,
    "args": {
      "External id": 239462,
      "cbid": 211, "correlation": 239462
    }
  },
  {
    "ph": "s", "id": 239462, "pid": 76337, "tid": -914061504, "ts": 1716454225175309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225234478, "dur": 5,
    "args": {
      "External id": 239464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239464,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239464, "pid": 5, "tid": 7, "ts": 1716454225234478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175322, "dur": 5,
    "args": {
      "External id": 239464,
      "cbid": 211, "correlation": 239464
    }
  },
  {
    "ph": "s", "id": 239464, "pid": 76337, "tid": -914061504, "ts": 1716454225175322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225234485, "dur": 16,
    "args": {
      "External id": 239474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239474, "pid": 5, "tid": 7, "ts": 1716454225234485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175380, "dur": 12,
    "args": {
      "External id": 239474,
      "cbid": 211, "correlation": 239474
    }
  },
  {
    "ph": "s", "id": 239474, "pid": 76337, "tid": -914061504, "ts": 1716454225175380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225234502, "dur": 17,
    "args": {
      "External id": 239494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239494,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 239494, "pid": 5, "tid": 7, "ts": 1716454225234502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175446, "dur": 11,
    "args": {
      "External id": 239494,
      "cbid": 211, "correlation": 239494
    }
  },
  {
    "ph": "s", "id": 239494, "pid": 76337, "tid": -914061504, "ts": 1716454225175446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225234521, "dur": 4,
    "args": {
      "External id": 239506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239506,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 239506, "pid": 5, "tid": 7, "ts": 1716454225234521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175468, "dur": 6,
    "args": {
      "External id": 239506,
      "cbid": 211, "correlation": 239506
    }
  },
  {
    "ph": "s", "id": 239506, "pid": 76337, "tid": -914061504, "ts": 1716454225175468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225234526, "dur": 16,
    "args": {
      "External id": 239509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239509, "pid": 5, "tid": 7, "ts": 1716454225234526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175486, "dur": 6,
    "args": {
      "External id": 239509,
      "cbid": 211, "correlation": 239509
    }
  },
  {
    "ph": "s", "id": 239509, "pid": 76337, "tid": -914061504, "ts": 1716454225175486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225234543, "dur": 12,
    "args": {
      "External id": 239518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239518,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239518, "pid": 5, "tid": 7, "ts": 1716454225234543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175526, "dur": 9,
    "args": {
      "External id": 239518,
      "cbid": 211, "correlation": 239518
    }
  },
  {
    "ph": "s", "id": 239518, "pid": 76337, "tid": -914061504, "ts": 1716454225175526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175587, "dur": 0,
    "args": {
      "External id": 239528,
      "cbid": 317, "correlation": 239528
    }
  },
  {
    "ph": "f", "id": 239528, "pid": 76337, "tid": -914061504, "ts": 1716454225175587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175588, "dur": 0,
    "args": {
      "External id": 239529,
      "cbid": 203, "correlation": 239529
    }
  },
  {
    "ph": "f", "id": 239529, "pid": 76337, "tid": -914061504, "ts": 1716454225175588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175589, "dur": 0,
    "args": {
      "External id": 239530,
      "cbid": 205, "correlation": 239530
    }
  },
  {
    "ph": "f", "id": 239530, "pid": 76337, "tid": -914061504, "ts": 1716454225175589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225234556, "dur": 11,
    "args": {
      "External id": 239534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239534,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239534, "pid": 5, "tid": 7, "ts": 1716454225234556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175602, "dur": 12,
    "args": {
      "External id": 239534,
      "cbid": 211, "correlation": 239534
    }
  },
  {
    "ph": "s", "id": 239534, "pid": 76337, "tid": -914061504, "ts": 1716454225175602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225234568, "dur": 159,
    "args": {
      "External id": 239536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239536,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239536, "pid": 5, "tid": 7, "ts": 1716454225234568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175617, "dur": 6,
    "args": {
      "External id": 239536,
      "cbid": 211, "correlation": 239536
    }
  },
  {
    "ph": "s", "id": 239536, "pid": 76337, "tid": -914061504, "ts": 1716454225175617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225234729, "dur": 1,
    "args": {
      "External id": 239538,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 239538,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 239538, "pid": 5, "tid": 7, "ts": 1716454225234729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175629, "dur": 6,
    "args": {
      "External id": 239538,
      "cbid": 51, "correlation": 239538
    }
  },
  {
    "ph": "s", "id": 239538, "pid": 76337, "tid": -914061504, "ts": 1716454225175629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225234733, "dur": 643,
    "args": {
      "External id": 239539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239539,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239539, "pid": 5, "tid": 7, "ts": 1716454225234733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175636, "dur": 6,
    "args": {
      "External id": 239539,
      "cbid": 211, "correlation": 239539
    }
  },
  {
    "ph": "s", "id": 239539, "pid": 76337, "tid": -914061504, "ts": 1716454225175636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225235377, "dur": 12,
    "args": {
      "External id": 239541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239541,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239541, "pid": 5, "tid": 7, "ts": 1716454225235377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175645, "dur": 5,
    "args": {
      "External id": 239541,
      "cbid": 211, "correlation": 239541
    }
  },
  {
    "ph": "s", "id": 239541, "pid": 76337, "tid": -914061504, "ts": 1716454225175645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225235391, "dur": 14,
    "args": {
      "External id": 239547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239547,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239547, "pid": 5, "tid": 7, "ts": 1716454225235391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175674, "dur": 9,
    "args": {
      "External id": 239547,
      "cbid": 211, "correlation": 239547
    }
  },
  {
    "ph": "s", "id": 239547, "pid": 76337, "tid": -914061504, "ts": 1716454225175674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175732, "dur": 0,
    "args": {
      "External id": 239557,
      "cbid": 317, "correlation": 239557
    }
  },
  {
    "ph": "f", "id": 239557, "pid": 76337, "tid": -914061504, "ts": 1716454225175732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175733, "dur": 0,
    "args": {
      "External id": 239558,
      "cbid": 203, "correlation": 239558
    }
  },
  {
    "ph": "f", "id": 239558, "pid": 76337, "tid": -914061504, "ts": 1716454225175733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175734, "dur": 0,
    "args": {
      "External id": 239559,
      "cbid": 205, "correlation": 239559
    }
  },
  {
    "ph": "f", "id": 239559, "pid": 76337, "tid": -914061504, "ts": 1716454225175734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225235407, "dur": 22,
    "args": {
      "External id": 239563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239563,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239563, "pid": 5, "tid": 7, "ts": 1716454225235407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175745, "dur": 11,
    "args": {
      "External id": 239563,
      "cbid": 211, "correlation": 239563
    }
  },
  {
    "ph": "s", "id": 239563, "pid": 76337, "tid": -914061504, "ts": 1716454225175745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225235430, "dur": 4,
    "args": {
      "External id": 239565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239565, "pid": 5, "tid": 7, "ts": 1716454225235430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175760, "dur": 7,
    "args": {
      "External id": 239565,
      "cbid": 211, "correlation": 239565
    }
  },
  {
    "ph": "s", "id": 239565, "pid": 76337, "tid": -914061504, "ts": 1716454225175760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175770, "dur": 0,
    "args": {
      "External id": 239566,
      "cbid": 51, "correlation": 239566
    }
  },
  {
    "ph": "s", "id": 239566, "pid": 76337, "tid": -914061504, "ts": 1716454225175770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225235435, "dur": 167,
    "args": {
      "External id": 239567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239567,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 239567, "pid": 5, "tid": 7, "ts": 1716454225235435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175771, "dur": 5,
    "args": {
      "External id": 239567,
      "cbid": 211, "correlation": 239567
    }
  },
  {
    "ph": "s", "id": 239567, "pid": 76337, "tid": -914061504, "ts": 1716454225175771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225235604, "dur": 15,
    "args": {
      "External id": 239572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239572,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239572, "pid": 5, "tid": 7, "ts": 1716454225235604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175796, "dur": 8,
    "args": {
      "External id": 239572,
      "cbid": 211, "correlation": 239572
    }
  },
  {
    "ph": "s", "id": 239572, "pid": 76337, "tid": -914061504, "ts": 1716454225175796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225235621, "dur": 13,
    "args": {
      "External id": 239580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239580,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239580, "pid": 5, "tid": 7, "ts": 1716454225235621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175825, "dur": 8,
    "args": {
      "External id": 239580,
      "cbid": 211, "correlation": 239580
    }
  },
  {
    "ph": "s", "id": 239580, "pid": 76337, "tid": -914061504, "ts": 1716454225175825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225235635, "dur": 10,
    "args": {
      "External id": 239588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239588,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239588, "pid": 5, "tid": 7, "ts": 1716454225235635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175853, "dur": 8,
    "args": {
      "External id": 239588,
      "cbid": 211, "correlation": 239588
    }
  },
  {
    "ph": "s", "id": 239588, "pid": 76337, "tid": -914061504, "ts": 1716454225175853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225235646, "dur": 18,
    "args": {
      "External id": 239608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239608,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 239608, "pid": 5, "tid": 7, "ts": 1716454225235646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175934, "dur": 12,
    "args": {
      "External id": 239608,
      "cbid": 211, "correlation": 239608
    }
  },
  {
    "ph": "s", "id": 239608, "pid": 76337, "tid": -914061504, "ts": 1716454225175934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225235665, "dur": 5,
    "args": {
      "External id": 239620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239620,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 239620, "pid": 5, "tid": 7, "ts": 1716454225235665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175956, "dur": 6,
    "args": {
      "External id": 239620,
      "cbid": 211, "correlation": 239620
    }
  },
  {
    "ph": "s", "id": 239620, "pid": 76337, "tid": -914061504, "ts": 1716454225175956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225235671, "dur": 18,
    "args": {
      "External id": 239623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239623,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239623, "pid": 5, "tid": 7, "ts": 1716454225235671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225175982, "dur": 7,
    "args": {
      "External id": 239623,
      "cbid": 211, "correlation": 239623
    }
  },
  {
    "ph": "s", "id": 239623, "pid": 76337, "tid": -914061504, "ts": 1716454225175982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176042, "dur": 0,
    "args": {
      "External id": 239634,
      "cbid": 317, "correlation": 239634
    }
  },
  {
    "ph": "f", "id": 239634, "pid": 76337, "tid": -914061504, "ts": 1716454225176042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176043, "dur": 0,
    "args": {
      "External id": 239635,
      "cbid": 203, "correlation": 239635
    }
  },
  {
    "ph": "f", "id": 239635, "pid": 76337, "tid": -914061504, "ts": 1716454225176043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176044, "dur": 0,
    "args": {
      "External id": 239636,
      "cbid": 205, "correlation": 239636
    }
  },
  {
    "ph": "f", "id": 239636, "pid": 76337, "tid": -914061504, "ts": 1716454225176044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225235690, "dur": 12,
    "args": {
      "External id": 239640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239640,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239640, "pid": 5, "tid": 7, "ts": 1716454225235690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176056, "dur": 12,
    "args": {
      "External id": 239640,
      "cbid": 211, "correlation": 239640
    }
  },
  {
    "ph": "s", "id": 239640, "pid": 76337, "tid": -914061504, "ts": 1716454225176056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225235703, "dur": 4,
    "args": {
      "External id": 239642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239642,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239642, "pid": 5, "tid": 7, "ts": 1716454225235703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176073, "dur": 6,
    "args": {
      "External id": 239642,
      "cbid": 211, "correlation": 239642
    }
  },
  {
    "ph": "s", "id": 239642, "pid": 76337, "tid": -914061504, "ts": 1716454225176073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176081, "dur": 0,
    "args": {
      "External id": 239643,
      "cbid": 51, "correlation": 239643
    }
  },
  {
    "ph": "s", "id": 239643, "pid": 76337, "tid": -914061504, "ts": 1716454225176081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225235708, "dur": 89,
    "args": {
      "External id": 239644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239644,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 239644, "pid": 5, "tid": 7, "ts": 1716454225235708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176082, "dur": 5,
    "args": {
      "External id": 239644,
      "cbid": 211, "correlation": 239644
    }
  },
  {
    "ph": "s", "id": 239644, "pid": 76337, "tid": -914061504, "ts": 1716454225176082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225235799, "dur": 15,
    "args": {
      "External id": 239649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239649,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239649, "pid": 5, "tid": 7, "ts": 1716454225235799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176108, "dur": 9,
    "args": {
      "External id": 239649,
      "cbid": 211, "correlation": 239649
    }
  },
  {
    "ph": "s", "id": 239649, "pid": 76337, "tid": -914061504, "ts": 1716454225176108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225235815, "dur": 83,
    "args": {
      "External id": 239658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239658,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239658, "pid": 5, "tid": 7, "ts": 1716454225235815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176190, "dur": 14,
    "args": {
      "External id": 239658,
      "cbid": 211, "correlation": 239658
    }
  },
  {
    "ph": "s", "id": 239658, "pid": 76337, "tid": -914061504, "ts": 1716454225176190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225235900, "dur": 30,
    "args": {
      "External id": 239680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239680,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239680, "pid": 5, "tid": 7, "ts": 1716454225235900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176248, "dur": 10,
    "args": {
      "External id": 239680,
      "cbid": 211, "correlation": 239680
    }
  },
  {
    "ph": "s", "id": 239680, "pid": 76337, "tid": -914061504, "ts": 1716454225176248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176334, "dur": 1,
    "args": {
      "External id": 239691,
      "cbid": 251, "correlation": 239691
    }
  },
  {
    "ph": "f", "id": 239691, "pid": 76337, "tid": -914061504, "ts": 1716454225176334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225235930, "dur": 162,
    "args": {
      "External id": 239692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239692,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239692, "pid": 5, "tid": 7, "ts": 1716454225235930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176339, "dur": 13,
    "args": {
      "External id": 239692,
      "cbid": 211, "correlation": 239692
    }
  },
  {
    "ph": "s", "id": 239692, "pid": 76337, "tid": -914061504, "ts": 1716454225176339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176409, "dur": 1,
    "args": {
      "External id": 239703,
      "cbid": 251, "correlation": 239703
    }
  },
  {
    "ph": "f", "id": 239703, "pid": 76337, "tid": -914061504, "ts": 1716454225176409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225236094, "dur": 154,
    "args": {
      "External id": 239704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239704,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239704, "pid": 5, "tid": 7, "ts": 1716454225236094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176413, "dur": 12,
    "args": {
      "External id": 239704,
      "cbid": 211, "correlation": 239704
    }
  },
  {
    "ph": "s", "id": 239704, "pid": 76337, "tid": -914061504, "ts": 1716454225176413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176479, "dur": 1,
    "args": {
      "External id": 239715,
      "cbid": 251, "correlation": 239715
    }
  },
  {
    "ph": "f", "id": 239715, "pid": 76337, "tid": -914061504, "ts": 1716454225176479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225236249, "dur": 155,
    "args": {
      "External id": 239716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239716,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239716, "pid": 5, "tid": 7, "ts": 1716454225236249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176483, "dur": 11,
    "args": {
      "External id": 239716,
      "cbid": 211, "correlation": 239716
    }
  },
  {
    "ph": "s", "id": 239716, "pid": 76337, "tid": -914061504, "ts": 1716454225176483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225236406, "dur": 332,
    "args": {
      "External id": 239741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239741,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239741, "pid": 5, "tid": 7, "ts": 1716454225236406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176566, "dur": 13,
    "args": {
      "External id": 239741,
      "cbid": 211, "correlation": 239741
    }
  },
  {
    "ph": "s", "id": 239741, "pid": 76337, "tid": -914061504, "ts": 1716454225176566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176666, "dur": 1,
    "args": {
      "External id": 239759,
      "cbid": 251, "correlation": 239759
    }
  },
  {
    "ph": "f", "id": 239759, "pid": 76337, "tid": -914061504, "ts": 1716454225176666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225236740, "dur": 163,
    "args": {
      "External id": 239761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239761,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239761, "pid": 5, "tid": 7, "ts": 1716454225236740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176672, "dur": 13,
    "args": {
      "External id": 239761,
      "cbid": 211, "correlation": 239761
    }
  },
  {
    "ph": "s", "id": 239761, "pid": 76337, "tid": -914061504, "ts": 1716454225176672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225236904, "dur": 20,
    "args": {
      "External id": 239769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239769,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239769, "pid": 5, "tid": 7, "ts": 1716454225236904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176741, "dur": 12,
    "args": {
      "External id": 239769,
      "cbid": 211, "correlation": 239769
    }
  },
  {
    "ph": "s", "id": 239769, "pid": 76337, "tid": -914061504, "ts": 1716454225176741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225236925, "dur": 27,
    "args": {
      "External id": 239777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239777,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239777, "pid": 5, "tid": 7, "ts": 1716454225236925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176780, "dur": 8,
    "args": {
      "External id": 239777,
      "cbid": 211, "correlation": 239777
    }
  },
  {
    "ph": "s", "id": 239777, "pid": 76337, "tid": -914061504, "ts": 1716454225176780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225236953, "dur": 18,
    "args": {
      "External id": 239788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239788, "pid": 5, "tid": 7, "ts": 1716454225236953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176850, "dur": 12,
    "args": {
      "External id": 239788,
      "cbid": 211, "correlation": 239788
    }
  },
  {
    "ph": "s", "id": 239788, "pid": 76337, "tid": -914061504, "ts": 1716454225176850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225236973, "dur": 16,
    "args": {
      "External id": 239810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239810,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239810, "pid": 5, "tid": 7, "ts": 1716454225236973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176882, "dur": 8,
    "args": {
      "External id": 239810,
      "cbid": 211, "correlation": 239810
    }
  },
  {
    "ph": "s", "id": 239810, "pid": 76337, "tid": -914061504, "ts": 1716454225176882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176967, "dur": 1,
    "args": {
      "External id": 239821,
      "cbid": 251, "correlation": 239821
    }
  },
  {
    "ph": "f", "id": 239821, "pid": 76337, "tid": -914061504, "ts": 1716454225176967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225236990, "dur": 87,
    "args": {
      "External id": 239822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239822,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239822, "pid": 5, "tid": 7, "ts": 1716454225236990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225176972, "dur": 21,
    "args": {
      "External id": 239822,
      "cbid": 211, "correlation": 239822
    }
  },
  {
    "ph": "s", "id": 239822, "pid": 76337, "tid": -914061504, "ts": 1716454225176972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177050, "dur": 1,
    "args": {
      "External id": 239833,
      "cbid": 251, "correlation": 239833
    }
  },
  {
    "ph": "f", "id": 239833, "pid": 76337, "tid": -914061504, "ts": 1716454225177050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177054, "dur": 0,
    "args": {
      "External id": 239834,
      "cbid": 251, "correlation": 239834
    }
  },
  {
    "ph": "f", "id": 239834, "pid": 76337, "tid": -914061504, "ts": 1716454225177054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225237078, "dur": 12,
    "args": {
      "External id": 239835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239835,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239835, "pid": 5, "tid": 7, "ts": 1716454225237078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177056, "dur": 12,
    "args": {
      "External id": 239835,
      "cbid": 211, "correlation": 239835
    }
  },
  {
    "ph": "s", "id": 239835, "pid": 76337, "tid": -914061504, "ts": 1716454225177056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225237092, "dur": 5,
    "args": {
      "External id": 239837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239837,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239837, "pid": 5, "tid": 7, "ts": 1716454225237092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177069, "dur": 6,
    "args": {
      "External id": 239837,
      "cbid": 211, "correlation": 239837
    }
  },
  {
    "ph": "s", "id": 239837, "pid": 76337, "tid": -914061504, "ts": 1716454225177069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177128, "dur": 1,
    "args": {
      "External id": 239848,
      "cbid": 251, "correlation": 239848
    }
  },
  {
    "ph": "f", "id": 239848, "pid": 76337, "tid": -914061504, "ts": 1716454225177128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177131, "dur": 0,
    "args": {
      "External id": 239849,
      "cbid": 251, "correlation": 239849
    }
  },
  {
    "ph": "f", "id": 239849, "pid": 76337, "tid": -914061504, "ts": 1716454225177131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225237098, "dur": 9,
    "args": {
      "External id": 239850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239850,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239850, "pid": 5, "tid": 7, "ts": 1716454225237098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177132, "dur": 11,
    "args": {
      "External id": 239850,
      "cbid": 211, "correlation": 239850
    }
  },
  {
    "ph": "s", "id": 239850, "pid": 76337, "tid": -914061504, "ts": 1716454225177132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225237108, "dur": 3,
    "args": {
      "External id": 239852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239852,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239852, "pid": 5, "tid": 7, "ts": 1716454225237108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177146, "dur": 5,
    "args": {
      "External id": 239852,
      "cbid": 211, "correlation": 239852
    }
  },
  {
    "ph": "s", "id": 239852, "pid": 76337, "tid": -914061504, "ts": 1716454225177146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225237113, "dur": 54,
    "args": {
      "External id": 239877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239877,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239877, "pid": 5, "tid": 7, "ts": 1716454225237113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177223, "dur": 12,
    "args": {
      "External id": 239877,
      "cbid": 211, "correlation": 239877
    }
  },
  {
    "ph": "s", "id": 239877, "pid": 76337, "tid": -914061504, "ts": 1716454225177223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177320, "dur": 1,
    "args": {
      "External id": 239895,
      "cbid": 251, "correlation": 239895
    }
  },
  {
    "ph": "f", "id": 239895, "pid": 76337, "tid": -914061504, "ts": 1716454225177320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225237168, "dur": 89,
    "args": {
      "External id": 239897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239897,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239897, "pid": 5, "tid": 7, "ts": 1716454225237168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177327, "dur": 14,
    "args": {
      "External id": 239897,
      "cbid": 211, "correlation": 239897
    }
  },
  {
    "ph": "s", "id": 239897, "pid": 76337, "tid": -914061504, "ts": 1716454225177327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225237258, "dur": 9,
    "args": {
      "External id": 239905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239905,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239905, "pid": 5, "tid": 7, "ts": 1716454225237258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177395, "dur": 12,
    "args": {
      "External id": 239905,
      "cbid": 211, "correlation": 239905
    }
  },
  {
    "ph": "s", "id": 239905, "pid": 76337, "tid": -914061504, "ts": 1716454225177395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225237269, "dur": 19,
    "args": {
      "External id": 239913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239913,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239913, "pid": 5, "tid": 7, "ts": 1716454225237269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177436, "dur": 9,
    "args": {
      "External id": 239913,
      "cbid": 211, "correlation": 239913
    }
  },
  {
    "ph": "s", "id": 239913, "pid": 76337, "tid": -914061504, "ts": 1716454225177436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225237289, "dur": 17,
    "args": {
      "External id": 239935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239935,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239935, "pid": 5, "tid": 7, "ts": 1716454225237289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177487, "dur": 10,
    "args": {
      "External id": 239935,
      "cbid": 211, "correlation": 239935
    }
  },
  {
    "ph": "s", "id": 239935, "pid": 76337, "tid": -914061504, "ts": 1716454225177487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177574, "dur": 1,
    "args": {
      "External id": 239951,
      "cbid": 251, "correlation": 239951
    }
  },
  {
    "ph": "f", "id": 239951, "pid": 76337, "tid": -914061504, "ts": 1716454225177574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177579, "dur": 0,
    "args": {
      "External id": 239953,
      "cbid": 251, "correlation": 239953
    }
  },
  {
    "ph": "f", "id": 239953, "pid": 76337, "tid": -914061504, "ts": 1716454225177579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225237307, "dur": 493,
    "args": {
      "External id": 239954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239954,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 239954, "pid": 5, "tid": 7, "ts": 1716454225237307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177581, "dur": 12,
    "args": {
      "External id": 239954,
      "cbid": 211, "correlation": 239954
    }
  },
  {
    "ph": "s", "id": 239954, "pid": 76337, "tid": -914061504, "ts": 1716454225177581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225237802, "dur": 65,
    "args": {
      "External id": 239962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239962, "pid": 5, "tid": 7, "ts": 1716454225237802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177645, "dur": 12,
    "args": {
      "External id": 239962,
      "cbid": 211, "correlation": 239962
    }
  },
  {
    "ph": "s", "id": 239962, "pid": 76337, "tid": -914061504, "ts": 1716454225177645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225237869, "dur": 67,
    "args": {
      "External id": 239970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239970, "pid": 5, "tid": 7, "ts": 1716454225237869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177677, "dur": 8,
    "args": {
      "External id": 239970,
      "cbid": 211, "correlation": 239970
    }
  },
  {
    "ph": "s", "id": 239970, "pid": 76337, "tid": -914061504, "ts": 1716454225177677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177756, "dur": 1,
    "args": {
      "External id": 239986,
      "cbid": 251, "correlation": 239986
    }
  },
  {
    "ph": "f", "id": 239986, "pid": 76337, "tid": -914061504, "ts": 1716454225177756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225237938, "dur": 1,
    "args": {
      "External id": 239988,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 239988,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 239988, "pid": 5, "tid": 7, "ts": 1716454225237938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177761, "dur": 10,
    "args": {
      "External id": 239988,
      "cbid": 51, "correlation": 239988
    }
  },
  {
    "ph": "s", "id": 239988, "pid": 76337, "tid": -914061504, "ts": 1716454225177761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225237942, "dur": 264,
    "args": {
      "External id": 239989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239989,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 239989, "pid": 5, "tid": 7, "ts": 1716454225237942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177773, "dur": 11,
    "args": {
      "External id": 239989,
      "cbid": 211, "correlation": 239989
    }
  },
  {
    "ph": "s", "id": 239989, "pid": 76337, "tid": -914061504, "ts": 1716454225177773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225238208, "dur": 14,
    "args": {
      "External id": 239997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 239997,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 239997, "pid": 5, "tid": 7, "ts": 1716454225238208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177814, "dur": 11,
    "args": {
      "External id": 239997,
      "cbid": 211, "correlation": 239997
    }
  },
  {
    "ph": "s", "id": 239997, "pid": 76337, "tid": -914061504, "ts": 1716454225177814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225238223, "dur": 37,
    "args": {
      "External id": 240008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240008,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240008, "pid": 5, "tid": 7, "ts": 1716454225238223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177883, "dur": 12,
    "args": {
      "External id": 240008,
      "cbid": 211, "correlation": 240008
    }
  },
  {
    "ph": "s", "id": 240008, "pid": 76337, "tid": -914061504, "ts": 1716454225177883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177946, "dur": 0,
    "args": {
      "External id": 240020,
      "cbid": 317, "correlation": 240020
    }
  },
  {
    "ph": "f", "id": 240020, "pid": 76337, "tid": -914061504, "ts": 1716454225177946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177947, "dur": 0,
    "args": {
      "External id": 240021,
      "cbid": 203, "correlation": 240021
    }
  },
  {
    "ph": "f", "id": 240021, "pid": 76337, "tid": -914061504, "ts": 1716454225177947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177948, "dur": 0,
    "args": {
      "External id": 240022,
      "cbid": 205, "correlation": 240022
    }
  },
  {
    "ph": "f", "id": 240022, "pid": 76337, "tid": -914061504, "ts": 1716454225177948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225238261, "dur": 13,
    "args": {
      "External id": 240026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240026, "pid": 5, "tid": 7, "ts": 1716454225238261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177963, "dur": 20,
    "args": {
      "External id": 240026,
      "cbid": 211, "correlation": 240026
    }
  },
  {
    "ph": "s", "id": 240026, "pid": 76337, "tid": -914061504, "ts": 1716454225177963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225238276, "dur": 4,
    "args": {
      "External id": 240028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240028, "pid": 5, "tid": 7, "ts": 1716454225238276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177988, "dur": 6,
    "args": {
      "External id": 240028,
      "cbid": 211, "correlation": 240028
    }
  },
  {
    "ph": "s", "id": 240028, "pid": 76337, "tid": -914061504, "ts": 1716454225177988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177997, "dur": 0,
    "args": {
      "External id": 240029,
      "cbid": 51, "correlation": 240029
    }
  },
  {
    "ph": "s", "id": 240029, "pid": 76337, "tid": -914061504, "ts": 1716454225177997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225238282, "dur": 96,
    "args": {
      "External id": 240030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240030,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 240030, "pid": 5, "tid": 7, "ts": 1716454225238282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225177998, "dur": 5,
    "args": {
      "External id": 240030,
      "cbid": 211, "correlation": 240030
    }
  },
  {
    "ph": "s", "id": 240030, "pid": 76337, "tid": -914061504, "ts": 1716454225177998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225238379, "dur": 17,
    "args": {
      "External id": 240035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240035,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240035, "pid": 5, "tid": 7, "ts": 1716454225238379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178026, "dur": 9,
    "args": {
      "External id": 240035,
      "cbid": 211, "correlation": 240035
    }
  },
  {
    "ph": "s", "id": 240035, "pid": 76337, "tid": -914061504, "ts": 1716454225178026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225238397, "dur": 12,
    "args": {
      "External id": 240043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240043,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240043, "pid": 5, "tid": 7, "ts": 1716454225238397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178059, "dur": 8,
    "args": {
      "External id": 240043,
      "cbid": 211, "correlation": 240043
    }
  },
  {
    "ph": "s", "id": 240043, "pid": 76337, "tid": -914061504, "ts": 1716454225178059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225238411, "dur": 26,
    "args": {
      "External id": 240052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240052,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240052, "pid": 5, "tid": 7, "ts": 1716454225238411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178098, "dur": 10,
    "args": {
      "External id": 240052,
      "cbid": 211, "correlation": 240052
    }
  },
  {
    "ph": "s", "id": 240052, "pid": 76337, "tid": -914061504, "ts": 1716454225178098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225238438, "dur": 24,
    "args": {
      "External id": 240072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240072,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 240072, "pid": 5, "tid": 7, "ts": 1716454225238438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178169, "dur": 12,
    "args": {
      "External id": 240072,
      "cbid": 211, "correlation": 240072
    }
  },
  {
    "ph": "s", "id": 240072, "pid": 76337, "tid": -914061504, "ts": 1716454225178169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225238463, "dur": 5,
    "args": {
      "External id": 240084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240084,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 240084, "pid": 5, "tid": 7, "ts": 1716454225238463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178192, "dur": 6,
    "args": {
      "External id": 240084,
      "cbid": 211, "correlation": 240084
    }
  },
  {
    "ph": "s", "id": 240084, "pid": 76337, "tid": -914061504, "ts": 1716454225178192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225238468, "dur": 24,
    "args": {
      "External id": 240087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240087, "pid": 5, "tid": 7, "ts": 1716454225238468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178210, "dur": 7,
    "args": {
      "External id": 240087,
      "cbid": 211, "correlation": 240087
    }
  },
  {
    "ph": "s", "id": 240087, "pid": 76337, "tid": -914061504, "ts": 1716454225178210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225238494, "dur": 17,
    "args": {
      "External id": 240096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240096,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240096, "pid": 5, "tid": 7, "ts": 1716454225238494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178248, "dur": 10,
    "args": {
      "External id": 240096,
      "cbid": 211, "correlation": 240096
    }
  },
  {
    "ph": "s", "id": 240096, "pid": 76337, "tid": -914061504, "ts": 1716454225178248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178300, "dur": 0,
    "args": {
      "External id": 240106,
      "cbid": 317, "correlation": 240106
    }
  },
  {
    "ph": "f", "id": 240106, "pid": 76337, "tid": -914061504, "ts": 1716454225178300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178301, "dur": 0,
    "args": {
      "External id": 240107,
      "cbid": 203, "correlation": 240107
    }
  },
  {
    "ph": "f", "id": 240107, "pid": 76337, "tid": -914061504, "ts": 1716454225178301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178301, "dur": 0,
    "args": {
      "External id": 240108,
      "cbid": 205, "correlation": 240108
    }
  },
  {
    "ph": "f", "id": 240108, "pid": 76337, "tid": -914061504, "ts": 1716454225178301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225238512, "dur": 17,
    "args": {
      "External id": 240112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240112,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240112, "pid": 5, "tid": 7, "ts": 1716454225238512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178316, "dur": 12,
    "args": {
      "External id": 240112,
      "cbid": 211, "correlation": 240112
    }
  },
  {
    "ph": "s", "id": 240112, "pid": 76337, "tid": -914061504, "ts": 1716454225178316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225238531, "dur": 237,
    "args": {
      "External id": 240114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240114,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240114, "pid": 5, "tid": 7, "ts": 1716454225238531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178331, "dur": 5,
    "args": {
      "External id": 240114,
      "cbid": 211, "correlation": 240114
    }
  },
  {
    "ph": "s", "id": 240114, "pid": 76337, "tid": -914061504, "ts": 1716454225178331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225238770, "dur": 1,
    "args": {
      "External id": 240116,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 240116,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 240116, "pid": 5, "tid": 7, "ts": 1716454225238770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178342, "dur": 8,
    "args": {
      "External id": 240116,
      "cbid": 51, "correlation": 240116
    }
  },
  {
    "ph": "s", "id": 240116, "pid": 76337, "tid": -914061504, "ts": 1716454225178342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225238774, "dur": 811,
    "args": {
      "External id": 240117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240117,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240117, "pid": 5, "tid": 7, "ts": 1716454225238774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178351, "dur": 6,
    "args": {
      "External id": 240117,
      "cbid": 211, "correlation": 240117
    }
  },
  {
    "ph": "s", "id": 240117, "pid": 76337, "tid": -914061504, "ts": 1716454225178351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225239587, "dur": 13,
    "args": {
      "External id": 240119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240119,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240119, "pid": 5, "tid": 7, "ts": 1716454225239587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178361, "dur": 5,
    "args": {
      "External id": 240119,
      "cbid": 211, "correlation": 240119
    }
  },
  {
    "ph": "s", "id": 240119, "pid": 76337, "tid": -914061504, "ts": 1716454225178361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225239602, "dur": 14,
    "args": {
      "External id": 240125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240125, "pid": 5, "tid": 7, "ts": 1716454225239602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178390, "dur": 9,
    "args": {
      "External id": 240125,
      "cbid": 211, "correlation": 240125
    }
  },
  {
    "ph": "s", "id": 240125, "pid": 76337, "tid": -914061504, "ts": 1716454225178390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225239617, "dur": 3,
    "args": {
      "External id": 240133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240133,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 240133, "pid": 5, "tid": 7, "ts": 1716454225239617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178434, "dur": 9,
    "args": {
      "External id": 240133,
      "cbid": 211, "correlation": 240133
    }
  },
  {
    "ph": "s", "id": 240133, "pid": 76337, "tid": -914061504, "ts": 1716454225178434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178498, "dur": 1,
    "args": {
      "External id": 240149,
      "cbid": 251, "correlation": 240149
    }
  },
  {
    "ph": "f", "id": 240149, "pid": 76337, "tid": -914061504, "ts": 1716454225178498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178503, "dur": 0,
    "args": {
      "External id": 240151,
      "cbid": 251, "correlation": 240151
    }
  },
  {
    "ph": "f", "id": 240151, "pid": 76337, "tid": -914061504, "ts": 1716454225178503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225239622, "dur": 13,
    "args": {
      "External id": 240152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240152,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240152, "pid": 5, "tid": 7, "ts": 1716454225239622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178505, "dur": 11,
    "args": {
      "External id": 240152,
      "cbid": 211, "correlation": 240152
    }
  },
  {
    "ph": "s", "id": 240152, "pid": 76337, "tid": -914061504, "ts": 1716454225178505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225239636, "dur": 5,
    "args": {
      "External id": 240154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240154,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240154, "pid": 5, "tid": 7, "ts": 1716454225239636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178518, "dur": 6,
    "args": {
      "External id": 240154,
      "cbid": 211, "correlation": 240154
    }
  },
  {
    "ph": "s", "id": 240154, "pid": 76337, "tid": -914061504, "ts": 1716454225178518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225239642, "dur": 16,
    "args": {
      "External id": 240164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240164, "pid": 5, "tid": 7, "ts": 1716454225239642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178577, "dur": 12,
    "args": {
      "External id": 240164,
      "cbid": 211, "correlation": 240164
    }
  },
  {
    "ph": "s", "id": 240164, "pid": 76337, "tid": -914061504, "ts": 1716454225178577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225239660, "dur": 19,
    "args": {
      "External id": 240184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240184,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 240184, "pid": 5, "tid": 7, "ts": 1716454225239660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178640, "dur": 11,
    "args": {
      "External id": 240184,
      "cbid": 211, "correlation": 240184
    }
  },
  {
    "ph": "s", "id": 240184, "pid": 76337, "tid": -914061504, "ts": 1716454225178640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225239680, "dur": 4,
    "args": {
      "External id": 240196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240196,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 240196, "pid": 5, "tid": 7, "ts": 1716454225239680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178661, "dur": 6,
    "args": {
      "External id": 240196,
      "cbid": 211, "correlation": 240196
    }
  },
  {
    "ph": "s", "id": 240196, "pid": 76337, "tid": -914061504, "ts": 1716454225178661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225239685, "dur": 16,
    "args": {
      "External id": 240199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240199, "pid": 5, "tid": 7, "ts": 1716454225239685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178680, "dur": 6,
    "args": {
      "External id": 240199,
      "cbid": 211, "correlation": 240199
    }
  },
  {
    "ph": "s", "id": 240199, "pid": 76337, "tid": -914061504, "ts": 1716454225178680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225239703, "dur": 10,
    "args": {
      "External id": 240208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240208,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240208, "pid": 5, "tid": 7, "ts": 1716454225239703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178719, "dur": 11,
    "args": {
      "External id": 240208,
      "cbid": 211, "correlation": 240208
    }
  },
  {
    "ph": "s", "id": 240208, "pid": 76337, "tid": -914061504, "ts": 1716454225178719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178781, "dur": 0,
    "args": {
      "External id": 240218,
      "cbid": 317, "correlation": 240218
    }
  },
  {
    "ph": "f", "id": 240218, "pid": 76337, "tid": -914061504, "ts": 1716454225178781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178782, "dur": 0,
    "args": {
      "External id": 240219,
      "cbid": 203, "correlation": 240219
    }
  },
  {
    "ph": "f", "id": 240219, "pid": 76337, "tid": -914061504, "ts": 1716454225178782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178783, "dur": 0,
    "args": {
      "External id": 240220,
      "cbid": 205, "correlation": 240220
    }
  },
  {
    "ph": "f", "id": 240220, "pid": 76337, "tid": -914061504, "ts": 1716454225178783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225239714, "dur": 11,
    "args": {
      "External id": 240224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240224, "pid": 5, "tid": 7, "ts": 1716454225239714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178796, "dur": 12,
    "args": {
      "External id": 240224,
      "cbid": 211, "correlation": 240224
    }
  },
  {
    "ph": "s", "id": 240224, "pid": 76337, "tid": -914061504, "ts": 1716454225178796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225239727, "dur": 161,
    "args": {
      "External id": 240226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240226,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240226, "pid": 5, "tid": 7, "ts": 1716454225239727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178810, "dur": 5,
    "args": {
      "External id": 240226,
      "cbid": 211, "correlation": 240226
    }
  },
  {
    "ph": "s", "id": 240226, "pid": 76337, "tid": -914061504, "ts": 1716454225178810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225239890, "dur": 1,
    "args": {
      "External id": 240228,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 240228,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 240228, "pid": 5, "tid": 7, "ts": 1716454225239890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178821, "dur": 7,
    "args": {
      "External id": 240228,
      "cbid": 51, "correlation": 240228
    }
  },
  {
    "ph": "s", "id": 240228, "pid": 76337, "tid": -914061504, "ts": 1716454225178821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225239894, "dur": 642,
    "args": {
      "External id": 240229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240229,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240229, "pid": 5, "tid": 7, "ts": 1716454225239894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178829, "dur": 6,
    "args": {
      "External id": 240229,
      "cbid": 211, "correlation": 240229
    }
  },
  {
    "ph": "s", "id": 240229, "pid": 76337, "tid": -914061504, "ts": 1716454225178829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225240537, "dur": 12,
    "args": {
      "External id": 240231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240231, "pid": 5, "tid": 7, "ts": 1716454225240537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178839, "dur": 5,
    "args": {
      "External id": 240231,
      "cbid": 211, "correlation": 240231
    }
  },
  {
    "ph": "s", "id": 240231, "pid": 76337, "tid": -914061504, "ts": 1716454225178839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225240551, "dur": 15,
    "args": {
      "External id": 240237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240237, "pid": 5, "tid": 7, "ts": 1716454225240551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178868, "dur": 9,
    "args": {
      "External id": 240237,
      "cbid": 211, "correlation": 240237
    }
  },
  {
    "ph": "s", "id": 240237, "pid": 76337, "tid": -914061504, "ts": 1716454225178868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178926, "dur": 0,
    "args": {
      "External id": 240247,
      "cbid": 317, "correlation": 240247
    }
  },
  {
    "ph": "f", "id": 240247, "pid": 76337, "tid": -914061504, "ts": 1716454225178926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178927, "dur": 0,
    "args": {
      "External id": 240248,
      "cbid": 203, "correlation": 240248
    }
  },
  {
    "ph": "f", "id": 240248, "pid": 76337, "tid": -914061504, "ts": 1716454225178927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178928, "dur": 0,
    "args": {
      "External id": 240249,
      "cbid": 205, "correlation": 240249
    }
  },
  {
    "ph": "f", "id": 240249, "pid": 76337, "tid": -914061504, "ts": 1716454225178928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225240567, "dur": 17,
    "args": {
      "External id": 240253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240253,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240253, "pid": 5, "tid": 7, "ts": 1716454225240567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178941, "dur": 11,
    "args": {
      "External id": 240253,
      "cbid": 211, "correlation": 240253
    }
  },
  {
    "ph": "s", "id": 240253, "pid": 76337, "tid": -914061504, "ts": 1716454225178941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225240585, "dur": 4,
    "args": {
      "External id": 240255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240255, "pid": 5, "tid": 7, "ts": 1716454225240585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178957, "dur": 5,
    "args": {
      "External id": 240255,
      "cbid": 211, "correlation": 240255
    }
  },
  {
    "ph": "s", "id": 240255, "pid": 76337, "tid": -914061504, "ts": 1716454225178957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178965, "dur": 0,
    "args": {
      "External id": 240256,
      "cbid": 51, "correlation": 240256
    }
  },
  {
    "ph": "s", "id": 240256, "pid": 76337, "tid": -914061504, "ts": 1716454225178965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225240590, "dur": 131,
    "args": {
      "External id": 240257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240257,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 240257, "pid": 5, "tid": 7, "ts": 1716454225240590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225178966, "dur": 5,
    "args": {
      "External id": 240257,
      "cbid": 211, "correlation": 240257
    }
  },
  {
    "ph": "s", "id": 240257, "pid": 76337, "tid": -914061504, "ts": 1716454225178966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225240722, "dur": 15,
    "args": {
      "External id": 240262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240262,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240262, "pid": 5, "tid": 7, "ts": 1716454225240722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179000, "dur": 9,
    "args": {
      "External id": 240262,
      "cbid": 211, "correlation": 240262
    }
  },
  {
    "ph": "s", "id": 240262, "pid": 76337, "tid": -914061504, "ts": 1716454225179000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225240738, "dur": 12,
    "args": {
      "External id": 240270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240270,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240270, "pid": 5, "tid": 7, "ts": 1716454225240738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179031, "dur": 8,
    "args": {
      "External id": 240270,
      "cbid": 211, "correlation": 240270
    }
  },
  {
    "ph": "s", "id": 240270, "pid": 76337, "tid": -914061504, "ts": 1716454225179031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225240752, "dur": 10,
    "args": {
      "External id": 240278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240278,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240278, "pid": 5, "tid": 7, "ts": 1716454225240752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179059, "dur": 8,
    "args": {
      "External id": 240278,
      "cbid": 211, "correlation": 240278
    }
  },
  {
    "ph": "s", "id": 240278, "pid": 76337, "tid": -914061504, "ts": 1716454225179059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225240763, "dur": 18,
    "args": {
      "External id": 240298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240298,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 240298, "pid": 5, "tid": 7, "ts": 1716454225240763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179141, "dur": 12,
    "args": {
      "External id": 240298,
      "cbid": 211, "correlation": 240298
    }
  },
  {
    "ph": "s", "id": 240298, "pid": 76337, "tid": -914061504, "ts": 1716454225179141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225240782, "dur": 4,
    "args": {
      "External id": 240310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240310,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 240310, "pid": 5, "tid": 7, "ts": 1716454225240782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179162, "dur": 6,
    "args": {
      "External id": 240310,
      "cbid": 211, "correlation": 240310
    }
  },
  {
    "ph": "s", "id": 240310, "pid": 76337, "tid": -914061504, "ts": 1716454225179162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225240788, "dur": 16,
    "args": {
      "External id": 240313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240313, "pid": 5, "tid": 7, "ts": 1716454225240788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179181, "dur": 6,
    "args": {
      "External id": 240313,
      "cbid": 211, "correlation": 240313
    }
  },
  {
    "ph": "s", "id": 240313, "pid": 76337, "tid": -914061504, "ts": 1716454225179181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179237, "dur": 0,
    "args": {
      "External id": 240324,
      "cbid": 317, "correlation": 240324
    }
  },
  {
    "ph": "f", "id": 240324, "pid": 76337, "tid": -914061504, "ts": 1716454225179237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179238, "dur": 0,
    "args": {
      "External id": 240325,
      "cbid": 203, "correlation": 240325
    }
  },
  {
    "ph": "f", "id": 240325, "pid": 76337, "tid": -914061504, "ts": 1716454225179238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179238, "dur": 0,
    "args": {
      "External id": 240326,
      "cbid": 205, "correlation": 240326
    }
  },
  {
    "ph": "f", "id": 240326, "pid": 76337, "tid": -914061504, "ts": 1716454225179238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225240805, "dur": 11,
    "args": {
      "External id": 240330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240330,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240330, "pid": 5, "tid": 7, "ts": 1716454225240805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179251, "dur": 11,
    "args": {
      "External id": 240330,
      "cbid": 211, "correlation": 240330
    }
  },
  {
    "ph": "s", "id": 240330, "pid": 76337, "tid": -914061504, "ts": 1716454225179251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225240818, "dur": 3,
    "args": {
      "External id": 240332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240332,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240332, "pid": 5, "tid": 7, "ts": 1716454225240818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179267, "dur": 6,
    "args": {
      "External id": 240332,
      "cbid": 211, "correlation": 240332
    }
  },
  {
    "ph": "s", "id": 240332, "pid": 76337, "tid": -914061504, "ts": 1716454225179267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179276, "dur": 0,
    "args": {
      "External id": 240333,
      "cbid": 51, "correlation": 240333
    }
  },
  {
    "ph": "s", "id": 240333, "pid": 76337, "tid": -914061504, "ts": 1716454225179276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225240822, "dur": 89,
    "args": {
      "External id": 240334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240334,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 240334, "pid": 5, "tid": 7, "ts": 1716454225240822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179277, "dur": 5,
    "args": {
      "External id": 240334,
      "cbid": 211, "correlation": 240334
    }
  },
  {
    "ph": "s", "id": 240334, "pid": 76337, "tid": -914061504, "ts": 1716454225179277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225240912, "dur": 15,
    "args": {
      "External id": 240339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240339,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240339, "pid": 5, "tid": 7, "ts": 1716454225240912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179303, "dur": 9,
    "args": {
      "External id": 240339,
      "cbid": 211, "correlation": 240339
    }
  },
  {
    "ph": "s", "id": 240339, "pid": 76337, "tid": -914061504, "ts": 1716454225179303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225240929, "dur": 81,
    "args": {
      "External id": 240348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240348,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240348, "pid": 5, "tid": 7, "ts": 1716454225240929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179384, "dur": 14,
    "args": {
      "External id": 240348,
      "cbid": 211, "correlation": 240348
    }
  },
  {
    "ph": "s", "id": 240348, "pid": 76337, "tid": -914061504, "ts": 1716454225179384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225241011, "dur": 30,
    "args": {
      "External id": 240370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240370,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240370, "pid": 5, "tid": 7, "ts": 1716454225241011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179442, "dur": 10,
    "args": {
      "External id": 240370,
      "cbid": 211, "correlation": 240370
    }
  },
  {
    "ph": "s", "id": 240370, "pid": 76337, "tid": -914061504, "ts": 1716454225179442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179529, "dur": 1,
    "args": {
      "External id": 240381,
      "cbid": 251, "correlation": 240381
    }
  },
  {
    "ph": "f", "id": 240381, "pid": 76337, "tid": -914061504, "ts": 1716454225179529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225241042, "dur": 159,
    "args": {
      "External id": 240382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240382,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240382, "pid": 5, "tid": 7, "ts": 1716454225241042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179535, "dur": 13,
    "args": {
      "External id": 240382,
      "cbid": 211, "correlation": 240382
    }
  },
  {
    "ph": "s", "id": 240382, "pid": 76337, "tid": -914061504, "ts": 1716454225179535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179605, "dur": 1,
    "args": {
      "External id": 240393,
      "cbid": 251, "correlation": 240393
    }
  },
  {
    "ph": "f", "id": 240393, "pid": 76337, "tid": -914061504, "ts": 1716454225179605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225241203, "dur": 157,
    "args": {
      "External id": 240394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240394,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240394, "pid": 5, "tid": 7, "ts": 1716454225241203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179609, "dur": 12,
    "args": {
      "External id": 240394,
      "cbid": 211, "correlation": 240394
    }
  },
  {
    "ph": "s", "id": 240394, "pid": 76337, "tid": -914061504, "ts": 1716454225179609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179675, "dur": 1,
    "args": {
      "External id": 240405,
      "cbid": 251, "correlation": 240405
    }
  },
  {
    "ph": "f", "id": 240405, "pid": 76337, "tid": -914061504, "ts": 1716454225179675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225241361, "dur": 155,
    "args": {
      "External id": 240406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240406,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240406, "pid": 5, "tid": 7, "ts": 1716454225241361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179679, "dur": 11,
    "args": {
      "External id": 240406,
      "cbid": 211, "correlation": 240406
    }
  },
  {
    "ph": "s", "id": 240406, "pid": 76337, "tid": -914061504, "ts": 1716454225179679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225241518, "dur": 329,
    "args": {
      "External id": 240431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240431,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240431, "pid": 5, "tid": 7, "ts": 1716454225241518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179764, "dur": 13,
    "args": {
      "External id": 240431,
      "cbid": 211, "correlation": 240431
    }
  },
  {
    "ph": "s", "id": 240431, "pid": 76337, "tid": -914061504, "ts": 1716454225179764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179863, "dur": 1,
    "args": {
      "External id": 240449,
      "cbid": 251, "correlation": 240449
    }
  },
  {
    "ph": "f", "id": 240449, "pid": 76337, "tid": -914061504, "ts": 1716454225179863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225241847, "dur": 163,
    "args": {
      "External id": 240451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240451,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240451, "pid": 5, "tid": 7, "ts": 1716454225241847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179870, "dur": 13,
    "args": {
      "External id": 240451,
      "cbid": 211, "correlation": 240451
    }
  },
  {
    "ph": "s", "id": 240451, "pid": 76337, "tid": -914061504, "ts": 1716454225179870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225242011, "dur": 19,
    "args": {
      "External id": 240459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240459,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240459, "pid": 5, "tid": 7, "ts": 1716454225242011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179939, "dur": 12,
    "args": {
      "External id": 240459,
      "cbid": 211, "correlation": 240459
    }
  },
  {
    "ph": "s", "id": 240459, "pid": 76337, "tid": -914061504, "ts": 1716454225179939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225242032, "dur": 27,
    "args": {
      "External id": 240467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240467,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240467, "pid": 5, "tid": 7, "ts": 1716454225242032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225179986, "dur": 9,
    "args": {
      "External id": 240467,
      "cbid": 211, "correlation": 240467
    }
  },
  {
    "ph": "s", "id": 240467, "pid": 76337, "tid": -914061504, "ts": 1716454225179986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225242061, "dur": 18,
    "args": {
      "External id": 240478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240478,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240478, "pid": 5, "tid": 7, "ts": 1716454225242061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180059, "dur": 14,
    "args": {
      "External id": 240478,
      "cbid": 211, "correlation": 240478
    }
  },
  {
    "ph": "s", "id": 240478, "pid": 76337, "tid": -914061504, "ts": 1716454225180059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225242080, "dur": 15,
    "args": {
      "External id": 240500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240500,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240500, "pid": 5, "tid": 7, "ts": 1716454225242080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180091, "dur": 7,
    "args": {
      "External id": 240500,
      "cbid": 211, "correlation": 240500
    }
  },
  {
    "ph": "s", "id": 240500, "pid": 76337, "tid": -914061504, "ts": 1716454225180091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180177, "dur": 1,
    "args": {
      "External id": 240511,
      "cbid": 251, "correlation": 240511
    }
  },
  {
    "ph": "f", "id": 240511, "pid": 76337, "tid": -914061504, "ts": 1716454225180177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225242097, "dur": 87,
    "args": {
      "External id": 240512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240512,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240512, "pid": 5, "tid": 7, "ts": 1716454225242097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180182, "dur": 13,
    "args": {
      "External id": 240512,
      "cbid": 211, "correlation": 240512
    }
  },
  {
    "ph": "s", "id": 240512, "pid": 76337, "tid": -914061504, "ts": 1716454225180182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180250, "dur": 1,
    "args": {
      "External id": 240523,
      "cbid": 251, "correlation": 240523
    }
  },
  {
    "ph": "f", "id": 240523, "pid": 76337, "tid": -914061504, "ts": 1716454225180250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180254, "dur": 0,
    "args": {
      "External id": 240524,
      "cbid": 251, "correlation": 240524
    }
  },
  {
    "ph": "f", "id": 240524, "pid": 76337, "tid": -914061504, "ts": 1716454225180254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225242185, "dur": 11,
    "args": {
      "External id": 240525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240525,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240525, "pid": 5, "tid": 7, "ts": 1716454225242185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180256, "dur": 13,
    "args": {
      "External id": 240525,
      "cbid": 211, "correlation": 240525
    }
  },
  {
    "ph": "s", "id": 240525, "pid": 76337, "tid": -914061504, "ts": 1716454225180256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225242198, "dur": 6,
    "args": {
      "External id": 240527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240527,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240527, "pid": 5, "tid": 7, "ts": 1716454225242198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180270, "dur": 6,
    "args": {
      "External id": 240527,
      "cbid": 211, "correlation": 240527
    }
  },
  {
    "ph": "s", "id": 240527, "pid": 76337, "tid": -914061504, "ts": 1716454225180270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180327, "dur": 1,
    "args": {
      "External id": 240538,
      "cbid": 251, "correlation": 240538
    }
  },
  {
    "ph": "f", "id": 240538, "pid": 76337, "tid": -914061504, "ts": 1716454225180327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180331, "dur": 0,
    "args": {
      "External id": 240539,
      "cbid": 251, "correlation": 240539
    }
  },
  {
    "ph": "f", "id": 240539, "pid": 76337, "tid": -914061504, "ts": 1716454225180331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225242205, "dur": 9,
    "args": {
      "External id": 240540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240540,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240540, "pid": 5, "tid": 7, "ts": 1716454225242205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180332, "dur": 11,
    "args": {
      "External id": 240540,
      "cbid": 211, "correlation": 240540
    }
  },
  {
    "ph": "s", "id": 240540, "pid": 76337, "tid": -914061504, "ts": 1716454225180332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225242215, "dur": 3,
    "args": {
      "External id": 240542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240542,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240542, "pid": 5, "tid": 7, "ts": 1716454225242215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180345, "dur": 5,
    "args": {
      "External id": 240542,
      "cbid": 211, "correlation": 240542
    }
  },
  {
    "ph": "s", "id": 240542, "pid": 76337, "tid": -914061504, "ts": 1716454225180345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225242220, "dur": 54,
    "args": {
      "External id": 240567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240567,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240567, "pid": 5, "tid": 7, "ts": 1716454225242220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180423, "dur": 12,
    "args": {
      "External id": 240567,
      "cbid": 211, "correlation": 240567
    }
  },
  {
    "ph": "s", "id": 240567, "pid": 76337, "tid": -914061504, "ts": 1716454225180423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180522, "dur": 1,
    "args": {
      "External id": 240585,
      "cbid": 251, "correlation": 240585
    }
  },
  {
    "ph": "f", "id": 240585, "pid": 76337, "tid": -914061504, "ts": 1716454225180522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225242275, "dur": 90,
    "args": {
      "External id": 240587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240587,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240587, "pid": 5, "tid": 7, "ts": 1716454225242275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180528, "dur": 13,
    "args": {
      "External id": 240587,
      "cbid": 211, "correlation": 240587
    }
  },
  {
    "ph": "s", "id": 240587, "pid": 76337, "tid": -914061504, "ts": 1716454225180528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225242366, "dur": 9,
    "args": {
      "External id": 240595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240595,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240595, "pid": 5, "tid": 7, "ts": 1716454225242366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180598, "dur": 13,
    "args": {
      "External id": 240595,
      "cbid": 211, "correlation": 240595
    }
  },
  {
    "ph": "s", "id": 240595, "pid": 76337, "tid": -914061504, "ts": 1716454225180598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225242376, "dur": 22,
    "args": {
      "External id": 240603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240603, "pid": 5, "tid": 7, "ts": 1716454225242376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180639, "dur": 9,
    "args": {
      "External id": 240603,
      "cbid": 211, "correlation": 240603
    }
  },
  {
    "ph": "s", "id": 240603, "pid": 76337, "tid": -914061504, "ts": 1716454225180639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225242400, "dur": 17,
    "args": {
      "External id": 240625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240625,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240625, "pid": 5, "tid": 7, "ts": 1716454225242400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180690, "dur": 10,
    "args": {
      "External id": 240625,
      "cbid": 211, "correlation": 240625
    }
  },
  {
    "ph": "s", "id": 240625, "pid": 76337, "tid": -914061504, "ts": 1716454225180690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180779, "dur": 1,
    "args": {
      "External id": 240641,
      "cbid": 251, "correlation": 240641
    }
  },
  {
    "ph": "f", "id": 240641, "pid": 76337, "tid": -914061504, "ts": 1716454225180779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180784, "dur": 0,
    "args": {
      "External id": 240643,
      "cbid": 251, "correlation": 240643
    }
  },
  {
    "ph": "f", "id": 240643, "pid": 76337, "tid": -914061504, "ts": 1716454225180784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225242418, "dur": 491,
    "args": {
      "External id": 240644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240644,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240644, "pid": 5, "tid": 7, "ts": 1716454225242418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180786, "dur": 12,
    "args": {
      "External id": 240644,
      "cbid": 211, "correlation": 240644
    }
  },
  {
    "ph": "s", "id": 240644, "pid": 76337, "tid": -914061504, "ts": 1716454225180786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225242911, "dur": 66,
    "args": {
      "External id": 240652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240652,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240652, "pid": 5, "tid": 7, "ts": 1716454225242911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180850, "dur": 13,
    "args": {
      "External id": 240652,
      "cbid": 211, "correlation": 240652
    }
  },
  {
    "ph": "s", "id": 240652, "pid": 76337, "tid": -914061504, "ts": 1716454225180850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225242978, "dur": 66,
    "args": {
      "External id": 240660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240660, "pid": 5, "tid": 7, "ts": 1716454225242978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180881, "dur": 8,
    "args": {
      "External id": 240660,
      "cbid": 211, "correlation": 240660
    }
  },
  {
    "ph": "s", "id": 240660, "pid": 76337, "tid": -914061504, "ts": 1716454225180881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180961, "dur": 1,
    "args": {
      "External id": 240676,
      "cbid": 251, "correlation": 240676
    }
  },
  {
    "ph": "f", "id": 240676, "pid": 76337, "tid": -914061504, "ts": 1716454225180961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225243046, "dur": 1,
    "args": {
      "External id": 240678,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 240678,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 240678, "pid": 5, "tid": 7, "ts": 1716454225243046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180966, "dur": 16,
    "args": {
      "External id": 240678,
      "cbid": 51, "correlation": 240678
    }
  },
  {
    "ph": "s", "id": 240678, "pid": 76337, "tid": -914061504, "ts": 1716454225180966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225243049, "dur": 266,
    "args": {
      "External id": 240679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240679,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240679, "pid": 5, "tid": 7, "ts": 1716454225243049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225180984, "dur": 11,
    "args": {
      "External id": 240679,
      "cbid": 211, "correlation": 240679
    }
  },
  {
    "ph": "s", "id": 240679, "pid": 76337, "tid": -914061504, "ts": 1716454225180984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225243317, "dur": 15,
    "args": {
      "External id": 240687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240687,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240687, "pid": 5, "tid": 7, "ts": 1716454225243317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181027, "dur": 10,
    "args": {
      "External id": 240687,
      "cbid": 211, "correlation": 240687
    }
  },
  {
    "ph": "s", "id": 240687, "pid": 76337, "tid": -914061504, "ts": 1716454225181027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225243333, "dur": 37,
    "args": {
      "External id": 240698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240698, "pid": 5, "tid": 7, "ts": 1716454225243333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181095, "dur": 13,
    "args": {
      "External id": 240698,
      "cbid": 211, "correlation": 240698
    }
  },
  {
    "ph": "s", "id": 240698, "pid": 76337, "tid": -914061504, "ts": 1716454225181095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181158, "dur": 0,
    "args": {
      "External id": 240710,
      "cbid": 317, "correlation": 240710
    }
  },
  {
    "ph": "f", "id": 240710, "pid": 76337, "tid": -914061504, "ts": 1716454225181158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181159, "dur": 0,
    "args": {
      "External id": 240711,
      "cbid": 203, "correlation": 240711
    }
  },
  {
    "ph": "f", "id": 240711, "pid": 76337, "tid": -914061504, "ts": 1716454225181159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181160, "dur": 0,
    "args": {
      "External id": 240712,
      "cbid": 205, "correlation": 240712
    }
  },
  {
    "ph": "f", "id": 240712, "pid": 76337, "tid": -914061504, "ts": 1716454225181160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225243371, "dur": 13,
    "args": {
      "External id": 240716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240716,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240716, "pid": 5, "tid": 7, "ts": 1716454225243371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181175, "dur": 12,
    "args": {
      "External id": 240716,
      "cbid": 211, "correlation": 240716
    }
  },
  {
    "ph": "s", "id": 240716, "pid": 76337, "tid": -914061504, "ts": 1716454225181175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225243385, "dur": 4,
    "args": {
      "External id": 240718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 240718, "pid": 5, "tid": 7, "ts": 1716454225243385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181191, "dur": 6,
    "args": {
      "External id": 240718,
      "cbid": 211, "correlation": 240718
    }
  },
  {
    "ph": "s", "id": 240718, "pid": 76337, "tid": -914061504, "ts": 1716454225181191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181199, "dur": 0,
    "args": {
      "External id": 240719,
      "cbid": 51, "correlation": 240719
    }
  },
  {
    "ph": "s", "id": 240719, "pid": 76337, "tid": -914061504, "ts": 1716454225181199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225243390, "dur": 95,
    "args": {
      "External id": 240720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240720,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 240720, "pid": 5, "tid": 7, "ts": 1716454225243390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181200, "dur": 5,
    "args": {
      "External id": 240720,
      "cbid": 211, "correlation": 240720
    }
  },
  {
    "ph": "s", "id": 240720, "pid": 76337, "tid": -914061504, "ts": 1716454225181200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225243486, "dur": 16,
    "args": {
      "External id": 240725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240725,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240725, "pid": 5, "tid": 7, "ts": 1716454225243486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181228, "dur": 9,
    "args": {
      "External id": 240725,
      "cbid": 211, "correlation": 240725
    }
  },
  {
    "ph": "s", "id": 240725, "pid": 76337, "tid": -914061504, "ts": 1716454225181228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225243504, "dur": 12,
    "args": {
      "External id": 240733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240733,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240733, "pid": 5, "tid": 7, "ts": 1716454225243504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181261, "dur": 8,
    "args": {
      "External id": 240733,
      "cbid": 211, "correlation": 240733
    }
  },
  {
    "ph": "s", "id": 240733, "pid": 76337, "tid": -914061504, "ts": 1716454225181261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225243517, "dur": 55,
    "args": {
      "External id": 240744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240744,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240744, "pid": 5, "tid": 7, "ts": 1716454225243517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181322, "dur": 11,
    "args": {
      "External id": 240744,
      "cbid": 211, "correlation": 240744
    }
  },
  {
    "ph": "s", "id": 240744, "pid": 76337, "tid": -914061504, "ts": 1716454225181322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181377, "dur": 0,
    "args": {
      "External id": 240754,
      "cbid": 317, "correlation": 240754
    }
  },
  {
    "ph": "f", "id": 240754, "pid": 76337, "tid": -914061504, "ts": 1716454225181377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181377, "dur": 0,
    "args": {
      "External id": 240755,
      "cbid": 203, "correlation": 240755
    }
  },
  {
    "ph": "f", "id": 240755, "pid": 76337, "tid": -914061504, "ts": 1716454225181377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181378, "dur": 0,
    "args": {
      "External id": 240756,
      "cbid": 205, "correlation": 240756
    }
  },
  {
    "ph": "f", "id": 240756, "pid": 76337, "tid": -914061504, "ts": 1716454225181378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225243573, "dur": 39,
    "args": {
      "External id": 240760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240760,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240760, "pid": 5, "tid": 7, "ts": 1716454225243573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181393, "dur": 12,
    "args": {
      "External id": 240760,
      "cbid": 211, "correlation": 240760
    }
  },
  {
    "ph": "s", "id": 240760, "pid": 76337, "tid": -914061504, "ts": 1716454225181393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225243613, "dur": 160,
    "args": {
      "External id": 240762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240762,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240762, "pid": 5, "tid": 7, "ts": 1716454225243613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181408, "dur": 5,
    "args": {
      "External id": 240762,
      "cbid": 211, "correlation": 240762
    }
  },
  {
    "ph": "s", "id": 240762, "pid": 76337, "tid": -914061504, "ts": 1716454225181408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225243775, "dur": 1962,
    "args": {
      "External id": 240764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240764,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240764, "pid": 5, "tid": 7, "ts": 1716454225243775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181420, "dur": 8,
    "args": {
      "External id": 240764,
      "cbid": 211, "correlation": 240764
    }
  },
  {
    "ph": "s", "id": 240764, "pid": 76337, "tid": -914061504, "ts": 1716454225181420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225245738, "dur": 40,
    "args": {
      "External id": 240766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240766, "pid": 5, "tid": 7, "ts": 1716454225245738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181432, "dur": 5,
    "args": {
      "External id": 240766,
      "cbid": 211, "correlation": 240766
    }
  },
  {
    "ph": "s", "id": 240766, "pid": 76337, "tid": -914061504, "ts": 1716454225181432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225245780, "dur": 58,
    "args": {
      "External id": 240772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240772, "pid": 5, "tid": 7, "ts": 1716454225245780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181461, "dur": 8,
    "args": {
      "External id": 240772,
      "cbid": 211, "correlation": 240772
    }
  },
  {
    "ph": "s", "id": 240772, "pid": 76337, "tid": -914061504, "ts": 1716454225181461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225245840, "dur": 83,
    "args": {
      "External id": 240781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240781,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240781, "pid": 5, "tid": 7, "ts": 1716454225245840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181550, "dur": 14,
    "args": {
      "External id": 240781,
      "cbid": 211, "correlation": 240781
    }
  },
  {
    "ph": "s", "id": 240781, "pid": 76337, "tid": -914061504, "ts": 1716454225181550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225245924, "dur": 72,
    "args": {
      "External id": 240801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240801,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 240801, "pid": 5, "tid": 7, "ts": 1716454225245924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181621, "dur": 11,
    "args": {
      "External id": 240801,
      "cbid": 211, "correlation": 240801
    }
  },
  {
    "ph": "s", "id": 240801, "pid": 76337, "tid": -914061504, "ts": 1716454225181621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225245998, "dur": 5,
    "args": {
      "External id": 240813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240813,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 240813, "pid": 5, "tid": 7, "ts": 1716454225245998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181642, "dur": 6,
    "args": {
      "External id": 240813,
      "cbid": 211, "correlation": 240813
    }
  },
  {
    "ph": "s", "id": 240813, "pid": 76337, "tid": -914061504, "ts": 1716454225181642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225246004, "dur": 80,
    "args": {
      "External id": 240816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240816,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240816, "pid": 5, "tid": 7, "ts": 1716454225246004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181661, "dur": 7,
    "args": {
      "External id": 240816,
      "cbid": 211, "correlation": 240816
    }
  },
  {
    "ph": "s", "id": 240816, "pid": 76337, "tid": -914061504, "ts": 1716454225181661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225246086, "dur": 53,
    "args": {
      "External id": 240825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240825,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240825, "pid": 5, "tid": 7, "ts": 1716454225246086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181700, "dur": 11,
    "args": {
      "External id": 240825,
      "cbid": 211, "correlation": 240825
    }
  },
  {
    "ph": "s", "id": 240825, "pid": 76337, "tid": -914061504, "ts": 1716454225181700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181754, "dur": 0,
    "args": {
      "External id": 240835,
      "cbid": 317, "correlation": 240835
    }
  },
  {
    "ph": "f", "id": 240835, "pid": 76337, "tid": -914061504, "ts": 1716454225181754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181755, "dur": 0,
    "args": {
      "External id": 240836,
      "cbid": 203, "correlation": 240836
    }
  },
  {
    "ph": "f", "id": 240836, "pid": 76337, "tid": -914061504, "ts": 1716454225181755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181756, "dur": 0,
    "args": {
      "External id": 240837,
      "cbid": 205, "correlation": 240837
    }
  },
  {
    "ph": "f", "id": 240837, "pid": 76337, "tid": -914061504, "ts": 1716454225181756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225246140, "dur": 56,
    "args": {
      "External id": 240841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240841,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240841, "pid": 5, "tid": 7, "ts": 1716454225246140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181772, "dur": 11,
    "args": {
      "External id": 240841,
      "cbid": 211, "correlation": 240841
    }
  },
  {
    "ph": "s", "id": 240841, "pid": 76337, "tid": -914061504, "ts": 1716454225181772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225246197, "dur": 120,
    "args": {
      "External id": 240843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240843,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240843, "pid": 5, "tid": 7, "ts": 1716454225246197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181786, "dur": 5,
    "args": {
      "External id": 240843,
      "cbid": 211, "correlation": 240843
    }
  },
  {
    "ph": "s", "id": 240843, "pid": 76337, "tid": -914061504, "ts": 1716454225181786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225246319, "dur": 1875,
    "args": {
      "External id": 240845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240845,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240845, "pid": 5, "tid": 7, "ts": 1716454225246319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181796, "dur": 6,
    "args": {
      "External id": 240845,
      "cbid": 211, "correlation": 240845
    }
  },
  {
    "ph": "s", "id": 240845, "pid": 76337, "tid": -914061504, "ts": 1716454225181796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225248195, "dur": 19,
    "args": {
      "External id": 240847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240847,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240847, "pid": 5, "tid": 7, "ts": 1716454225248195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181807, "dur": 5,
    "args": {
      "External id": 240847,
      "cbid": 211, "correlation": 240847
    }
  },
  {
    "ph": "s", "id": 240847, "pid": 76337, "tid": -914061504, "ts": 1716454225181807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225248216, "dur": 33,
    "args": {
      "External id": 240853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240853, "pid": 5, "tid": 7, "ts": 1716454225248216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181834, "dur": 8,
    "args": {
      "External id": 240853,
      "cbid": 211, "correlation": 240853
    }
  },
  {
    "ph": "s", "id": 240853, "pid": 76337, "tid": -914061504, "ts": 1716454225181834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225248250, "dur": 3,
    "args": {
      "External id": 240861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240861,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 240861, "pid": 5, "tid": 7, "ts": 1716454225248250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181879, "dur": 10,
    "args": {
      "External id": 240861,
      "cbid": 211, "correlation": 240861
    }
  },
  {
    "ph": "s", "id": 240861, "pid": 76337, "tid": -914061504, "ts": 1716454225181879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181944, "dur": 1,
    "args": {
      "External id": 240877,
      "cbid": 251, "correlation": 240877
    }
  },
  {
    "ph": "f", "id": 240877, "pid": 76337, "tid": -914061504, "ts": 1716454225181944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181950, "dur": 0,
    "args": {
      "External id": 240879,
      "cbid": 251, "correlation": 240879
    }
  },
  {
    "ph": "f", "id": 240879, "pid": 76337, "tid": -914061504, "ts": 1716454225181950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225248254, "dur": 12,
    "args": {
      "External id": 240880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240880,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 240880, "pid": 5, "tid": 7, "ts": 1716454225248254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181952, "dur": 12,
    "args": {
      "External id": 240880,
      "cbid": 211, "correlation": 240880
    }
  },
  {
    "ph": "s", "id": 240880, "pid": 76337, "tid": -914061504, "ts": 1716454225181952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225248267, "dur": 5,
    "args": {
      "External id": 240882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240882,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 240882, "pid": 5, "tid": 7, "ts": 1716454225248267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225181966, "dur": 6,
    "args": {
      "External id": 240882,
      "cbid": 211, "correlation": 240882
    }
  },
  {
    "ph": "s", "id": 240882, "pid": 76337, "tid": -914061504, "ts": 1716454225181966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225248274, "dur": 29,
    "args": {
      "External id": 240892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240892, "pid": 5, "tid": 7, "ts": 1716454225248274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182034, "dur": 13,
    "args": {
      "External id": 240892,
      "cbid": 211, "correlation": 240892
    }
  },
  {
    "ph": "s", "id": 240892, "pid": 76337, "tid": -914061504, "ts": 1716454225182034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225248304, "dur": 31,
    "args": {
      "External id": 240912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240912,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 240912, "pid": 5, "tid": 7, "ts": 1716454225248304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182100, "dur": 11,
    "args": {
      "External id": 240912,
      "cbid": 211, "correlation": 240912
    }
  },
  {
    "ph": "s", "id": 240912, "pid": 76337, "tid": -914061504, "ts": 1716454225182100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225248336, "dur": 4,
    "args": {
      "External id": 240924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240924,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 240924, "pid": 5, "tid": 7, "ts": 1716454225248336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182120, "dur": 6,
    "args": {
      "External id": 240924,
      "cbid": 211, "correlation": 240924
    }
  },
  {
    "ph": "s", "id": 240924, "pid": 76337, "tid": -914061504, "ts": 1716454225182120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225248342, "dur": 29,
    "args": {
      "External id": 240927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240927,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240927, "pid": 5, "tid": 7, "ts": 1716454225248342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182138, "dur": 6,
    "args": {
      "External id": 240927,
      "cbid": 211, "correlation": 240927
    }
  },
  {
    "ph": "s", "id": 240927, "pid": 76337, "tid": -914061504, "ts": 1716454225182138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225248372, "dur": 19,
    "args": {
      "External id": 240936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240936,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240936, "pid": 5, "tid": 7, "ts": 1716454225248372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182178, "dur": 10,
    "args": {
      "External id": 240936,
      "cbid": 211, "correlation": 240936
    }
  },
  {
    "ph": "s", "id": 240936, "pid": 76337, "tid": -914061504, "ts": 1716454225182178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182240, "dur": 0,
    "args": {
      "External id": 240946,
      "cbid": 317, "correlation": 240946
    }
  },
  {
    "ph": "f", "id": 240946, "pid": 76337, "tid": -914061504, "ts": 1716454225182240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182241, "dur": 0,
    "args": {
      "External id": 240947,
      "cbid": 203, "correlation": 240947
    }
  },
  {
    "ph": "f", "id": 240947, "pid": 76337, "tid": -914061504, "ts": 1716454225182241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182242, "dur": 0,
    "args": {
      "External id": 240948,
      "cbid": 205, "correlation": 240948
    }
  },
  {
    "ph": "f", "id": 240948, "pid": 76337, "tid": -914061504, "ts": 1716454225182242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225248393, "dur": 22,
    "args": {
      "External id": 240952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240952, "pid": 5, "tid": 7, "ts": 1716454225248393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182257, "dur": 12,
    "args": {
      "External id": 240952,
      "cbid": 211, "correlation": 240952
    }
  },
  {
    "ph": "s", "id": 240952, "pid": 76337, "tid": -914061504, "ts": 1716454225182257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225248417, "dur": 43,
    "args": {
      "External id": 240954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240954,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240954, "pid": 5, "tid": 7, "ts": 1716454225248417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182271, "dur": 5,
    "args": {
      "External id": 240954,
      "cbid": 211, "correlation": 240954
    }
  },
  {
    "ph": "s", "id": 240954, "pid": 76337, "tid": -914061504, "ts": 1716454225182271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225248461, "dur": 638,
    "args": {
      "External id": 240956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240956,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240956, "pid": 5, "tid": 7, "ts": 1716454225248461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182283, "dur": 6,
    "args": {
      "External id": 240956,
      "cbid": 211, "correlation": 240956
    }
  },
  {
    "ph": "s", "id": 240956, "pid": 76337, "tid": -914061504, "ts": 1716454225182283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225249100, "dur": 20,
    "args": {
      "External id": 240958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240958,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240958, "pid": 5, "tid": 7, "ts": 1716454225249100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182292, "dur": 5,
    "args": {
      "External id": 240958,
      "cbid": 211, "correlation": 240958
    }
  },
  {
    "ph": "s", "id": 240958, "pid": 76337, "tid": -914061504, "ts": 1716454225182292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225249122, "dur": 32,
    "args": {
      "External id": 240964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240964,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240964, "pid": 5, "tid": 7, "ts": 1716454225249122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182320, "dur": 9,
    "args": {
      "External id": 240964,
      "cbid": 211, "correlation": 240964
    }
  },
  {
    "ph": "s", "id": 240964, "pid": 76337, "tid": -914061504, "ts": 1716454225182320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182377, "dur": 0,
    "args": {
      "External id": 240974,
      "cbid": 317, "correlation": 240974
    }
  },
  {
    "ph": "f", "id": 240974, "pid": 76337, "tid": -914061504, "ts": 1716454225182377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182378, "dur": 0,
    "args": {
      "External id": 240975,
      "cbid": 203, "correlation": 240975
    }
  },
  {
    "ph": "f", "id": 240975, "pid": 76337, "tid": -914061504, "ts": 1716454225182378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182379, "dur": 0,
    "args": {
      "External id": 240976,
      "cbid": 205, "correlation": 240976
    }
  },
  {
    "ph": "f", "id": 240976, "pid": 76337, "tid": -914061504, "ts": 1716454225182379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225249155, "dur": 55,
    "args": {
      "External id": 240980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240980, "pid": 5, "tid": 7, "ts": 1716454225249155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182391, "dur": 12,
    "args": {
      "External id": 240980,
      "cbid": 211, "correlation": 240980
    }
  },
  {
    "ph": "s", "id": 240980, "pid": 76337, "tid": -914061504, "ts": 1716454225182391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225249211, "dur": 265,
    "args": {
      "External id": 240982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240982,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 240982, "pid": 5, "tid": 7, "ts": 1716454225249211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182410, "dur": 7,
    "args": {
      "External id": 240982,
      "cbid": 211, "correlation": 240982
    }
  },
  {
    "ph": "s", "id": 240982, "pid": 76337, "tid": -914061504, "ts": 1716454225182410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225249478, "dur": 22,
    "args": {
      "External id": 240984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240984,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240984, "pid": 5, "tid": 7, "ts": 1716454225249478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182420, "dur": 5,
    "args": {
      "External id": 240984,
      "cbid": 211, "correlation": 240984
    }
  },
  {
    "ph": "s", "id": 240984, "pid": 76337, "tid": -914061504, "ts": 1716454225182420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225249501, "dur": 32,
    "args": {
      "External id": 240990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240990, "pid": 5, "tid": 7, "ts": 1716454225249501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182447, "dur": 9,
    "args": {
      "External id": 240990,
      "cbid": 211, "correlation": 240990
    }
  },
  {
    "ph": "s", "id": 240990, "pid": 76337, "tid": -914061504, "ts": 1716454225182447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225249534, "dur": 27,
    "args": {
      "External id": 240998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 240998,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 240998, "pid": 5, "tid": 7, "ts": 1716454225249534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182475, "dur": 8,
    "args": {
      "External id": 240998,
      "cbid": 211, "correlation": 240998
    }
  },
  {
    "ph": "s", "id": 240998, "pid": 76337, "tid": -914061504, "ts": 1716454225182475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225249562, "dur": 19,
    "args": {
      "External id": 241006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241006,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241006, "pid": 5, "tid": 7, "ts": 1716454225249562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182505, "dur": 8,
    "args": {
      "External id": 241006,
      "cbid": 211, "correlation": 241006
    }
  },
  {
    "ph": "s", "id": 241006, "pid": 76337, "tid": -914061504, "ts": 1716454225182505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225249582, "dur": 29,
    "args": {
      "External id": 241026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241026,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 241026, "pid": 5, "tid": 7, "ts": 1716454225249582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182586, "dur": 13,
    "args": {
      "External id": 241026,
      "cbid": 211, "correlation": 241026
    }
  },
  {
    "ph": "s", "id": 241026, "pid": 76337, "tid": -914061504, "ts": 1716454225182586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225249613, "dur": 5,
    "args": {
      "External id": 241038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241038,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 241038, "pid": 5, "tid": 7, "ts": 1716454225249613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182609, "dur": 6,
    "args": {
      "External id": 241038,
      "cbid": 211, "correlation": 241038
    }
  },
  {
    "ph": "s", "id": 241038, "pid": 76337, "tid": -914061504, "ts": 1716454225182609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225249619, "dur": 30,
    "args": {
      "External id": 241041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241041,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241041, "pid": 5, "tid": 7, "ts": 1716454225249619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182626, "dur": 7,
    "args": {
      "External id": 241041,
      "cbid": 211, "correlation": 241041
    }
  },
  {
    "ph": "s", "id": 241041, "pid": 76337, "tid": -914061504, "ts": 1716454225182626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182684, "dur": 0,
    "args": {
      "External id": 241052,
      "cbid": 317, "correlation": 241052
    }
  },
  {
    "ph": "f", "id": 241052, "pid": 76337, "tid": -914061504, "ts": 1716454225182684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182685, "dur": 0,
    "args": {
      "External id": 241053,
      "cbid": 203, "correlation": 241053
    }
  },
  {
    "ph": "f", "id": 241053, "pid": 76337, "tid": -914061504, "ts": 1716454225182685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182686, "dur": 0,
    "args": {
      "External id": 241054,
      "cbid": 205, "correlation": 241054
    }
  },
  {
    "ph": "f", "id": 241054, "pid": 76337, "tid": -914061504, "ts": 1716454225182686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225249650, "dur": 22,
    "args": {
      "External id": 241058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241058,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241058, "pid": 5, "tid": 7, "ts": 1716454225249650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182699, "dur": 12,
    "args": {
      "External id": 241058,
      "cbid": 211, "correlation": 241058
    }
  },
  {
    "ph": "s", "id": 241058, "pid": 76337, "tid": -914061504, "ts": 1716454225182699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225249674, "dur": 103,
    "args": {
      "External id": 241060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241060,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241060, "pid": 5, "tid": 7, "ts": 1716454225249674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182717, "dur": 6,
    "args": {
      "External id": 241060,
      "cbid": 211, "correlation": 241060
    }
  },
  {
    "ph": "s", "id": 241060, "pid": 76337, "tid": -914061504, "ts": 1716454225182717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225249778, "dur": 23,
    "args": {
      "External id": 241062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241062,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241062, "pid": 5, "tid": 7, "ts": 1716454225249778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182727, "dur": 5,
    "args": {
      "External id": 241062,
      "cbid": 211, "correlation": 241062
    }
  },
  {
    "ph": "s", "id": 241062, "pid": 76337, "tid": -914061504, "ts": 1716454225182727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225249803, "dur": 32,
    "args": {
      "External id": 241068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241068,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241068, "pid": 5, "tid": 7, "ts": 1716454225249803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182755, "dur": 9,
    "args": {
      "External id": 241068,
      "cbid": 211, "correlation": 241068
    }
  },
  {
    "ph": "s", "id": 241068, "pid": 76337, "tid": -914061504, "ts": 1716454225182755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225249836, "dur": 195,
    "args": {
      "External id": 241077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241077,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241077, "pid": 5, "tid": 7, "ts": 1716454225249836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182836, "dur": 15,
    "args": {
      "External id": 241077,
      "cbid": 211, "correlation": 241077
    }
  },
  {
    "ph": "s", "id": 241077, "pid": 76337, "tid": -914061504, "ts": 1716454225182836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225250033, "dur": 64,
    "args": {
      "External id": 241099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241099,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241099, "pid": 5, "tid": 7, "ts": 1716454225250033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182893, "dur": 11,
    "args": {
      "External id": 241099,
      "cbid": 211, "correlation": 241099
    }
  },
  {
    "ph": "s", "id": 241099, "pid": 76337, "tid": -914061504, "ts": 1716454225182893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182989, "dur": 2,
    "args": {
      "External id": 241110,
      "cbid": 251, "correlation": 241110
    }
  },
  {
    "ph": "f", "id": 241110, "pid": 76337, "tid": -914061504, "ts": 1716454225182989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225250098, "dur": 152,
    "args": {
      "External id": 241111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241111,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241111, "pid": 5, "tid": 7, "ts": 1716454225250098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225182995, "dur": 13,
    "args": {
      "External id": 241111,
      "cbid": 211, "correlation": 241111
    }
  },
  {
    "ph": "s", "id": 241111, "pid": 76337, "tid": -914061504, "ts": 1716454225182995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183066, "dur": 1,
    "args": {
      "External id": 241122,
      "cbid": 251, "correlation": 241122
    }
  },
  {
    "ph": "f", "id": 241122, "pid": 76337, "tid": -914061504, "ts": 1716454225183066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225250251, "dur": 143,
    "args": {
      "External id": 241123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241123,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241123, "pid": 5, "tid": 7, "ts": 1716454225250251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183070, "dur": 12,
    "args": {
      "External id": 241123,
      "cbid": 211, "correlation": 241123
    }
  },
  {
    "ph": "s", "id": 241123, "pid": 76337, "tid": -914061504, "ts": 1716454225183070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183135, "dur": 1,
    "args": {
      "External id": 241134,
      "cbid": 251, "correlation": 241134
    }
  },
  {
    "ph": "f", "id": 241134, "pid": 76337, "tid": -914061504, "ts": 1716454225183135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225250395, "dur": 141,
    "args": {
      "External id": 241135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241135,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241135, "pid": 5, "tid": 7, "ts": 1716454225250395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183139, "dur": 11,
    "args": {
      "External id": 241135,
      "cbid": 211, "correlation": 241135
    }
  },
  {
    "ph": "s", "id": 241135, "pid": 76337, "tid": -914061504, "ts": 1716454225183139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225250538, "dur": 1910,
    "args": {
      "External id": 241156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241156,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 241156, "pid": 5, "tid": 7, "ts": 1716454225250538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183222, "dur": 13,
    "args": {
      "External id": 241156,
      "cbid": 211, "correlation": 241156
    }
  },
  {
    "ph": "s", "id": 241156, "pid": 76337, "tid": -914061504, "ts": 1716454225183222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183320, "dur": 1,
    "args": {
      "External id": 241174,
      "cbid": 251, "correlation": 241174
    }
  },
  {
    "ph": "f", "id": 241174, "pid": 76337, "tid": -914061504, "ts": 1716454225183320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225252450, "dur": 146,
    "args": {
      "External id": 241176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241176,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 241176, "pid": 5, "tid": 7, "ts": 1716454225252450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183326, "dur": 15,
    "args": {
      "External id": 241176,
      "cbid": 211, "correlation": 241176
    }
  },
  {
    "ph": "s", "id": 241176, "pid": 76337, "tid": -914061504, "ts": 1716454225183326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225252597, "dur": 36,
    "args": {
      "External id": 241184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241184,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241184, "pid": 5, "tid": 7, "ts": 1716454225252597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183397, "dur": 12,
    "args": {
      "External id": 241184,
      "cbid": 211, "correlation": 241184
    }
  },
  {
    "ph": "s", "id": 241184, "pid": 76337, "tid": -914061504, "ts": 1716454225183397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225252635, "dur": 50,
    "args": {
      "External id": 241192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241192,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241192, "pid": 5, "tid": 7, "ts": 1716454225252635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183437, "dur": 9,
    "args": {
      "External id": 241192,
      "cbid": 211, "correlation": 241192
    }
  },
  {
    "ph": "s", "id": 241192, "pid": 76337, "tid": -914061504, "ts": 1716454225183437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225252686, "dur": 30,
    "args": {
      "External id": 241203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241203, "pid": 5, "tid": 7, "ts": 1716454225252686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183506, "dur": 12,
    "args": {
      "External id": 241203,
      "cbid": 211, "correlation": 241203
    }
  },
  {
    "ph": "s", "id": 241203, "pid": 76337, "tid": -914061504, "ts": 1716454225183506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225252718, "dur": 33,
    "args": {
      "External id": 241225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241225,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241225, "pid": 5, "tid": 7, "ts": 1716454225252718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183537, "dur": 8,
    "args": {
      "External id": 241225,
      "cbid": 211, "correlation": 241225
    }
  },
  {
    "ph": "s", "id": 241225, "pid": 76337, "tid": -914061504, "ts": 1716454225183537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183621, "dur": 1,
    "args": {
      "External id": 241236,
      "cbid": 251, "correlation": 241236
    }
  },
  {
    "ph": "f", "id": 241236, "pid": 76337, "tid": -914061504, "ts": 1716454225183621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225252752, "dur": 75,
    "args": {
      "External id": 241237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241237,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241237, "pid": 5, "tid": 7, "ts": 1716454225252752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183626, "dur": 14,
    "args": {
      "External id": 241237,
      "cbid": 211, "correlation": 241237
    }
  },
  {
    "ph": "s", "id": 241237, "pid": 76337, "tid": -914061504, "ts": 1716454225183626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183695, "dur": 1,
    "args": {
      "External id": 241248,
      "cbid": 251, "correlation": 241248
    }
  },
  {
    "ph": "f", "id": 241248, "pid": 76337, "tid": -914061504, "ts": 1716454225183695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183699, "dur": 0,
    "args": {
      "External id": 241249,
      "cbid": 251, "correlation": 241249
    }
  },
  {
    "ph": "f", "id": 241249, "pid": 76337, "tid": -914061504, "ts": 1716454225183699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225252828, "dur": 11,
    "args": {
      "External id": 241250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241250,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 241250, "pid": 5, "tid": 7, "ts": 1716454225252828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183700, "dur": 12,
    "args": {
      "External id": 241250,
      "cbid": 211, "correlation": 241250
    }
  },
  {
    "ph": "s", "id": 241250, "pid": 76337, "tid": -914061504, "ts": 1716454225183700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225252841, "dur": 5,
    "args": {
      "External id": 241252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241252,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 241252, "pid": 5, "tid": 7, "ts": 1716454225252841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183714, "dur": 6,
    "args": {
      "External id": 241252,
      "cbid": 211, "correlation": 241252
    }
  },
  {
    "ph": "s", "id": 241252, "pid": 76337, "tid": -914061504, "ts": 1716454225183714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183772, "dur": 1,
    "args": {
      "External id": 241263,
      "cbid": 251, "correlation": 241263
    }
  },
  {
    "ph": "f", "id": 241263, "pid": 76337, "tid": -914061504, "ts": 1716454225183772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183775, "dur": 0,
    "args": {
      "External id": 241264,
      "cbid": 251, "correlation": 241264
    }
  },
  {
    "ph": "f", "id": 241264, "pid": 76337, "tid": -914061504, "ts": 1716454225183775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225252847, "dur": 7,
    "args": {
      "External id": 241265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241265,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 241265, "pid": 5, "tid": 7, "ts": 1716454225252847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183777, "dur": 12,
    "args": {
      "External id": 241265,
      "cbid": 211, "correlation": 241265
    }
  },
  {
    "ph": "s", "id": 241265, "pid": 76337, "tid": -914061504, "ts": 1716454225183777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225252855, "dur": 3,
    "args": {
      "External id": 241267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241267,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 241267, "pid": 5, "tid": 7, "ts": 1716454225252855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183790, "dur": 5,
    "args": {
      "External id": 241267,
      "cbid": 211, "correlation": 241267
    }
  },
  {
    "ph": "s", "id": 241267, "pid": 76337, "tid": -914061504, "ts": 1716454225183790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225252860, "dur": 90,
    "args": {
      "External id": 241288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241288,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 241288, "pid": 5, "tid": 7, "ts": 1716454225252860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183863, "dur": 12,
    "args": {
      "External id": 241288,
      "cbid": 211, "correlation": 241288
    }
  },
  {
    "ph": "s", "id": 241288, "pid": 76337, "tid": -914061504, "ts": 1716454225183863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183959, "dur": 1,
    "args": {
      "External id": 241306,
      "cbid": 251, "correlation": 241306
    }
  },
  {
    "ph": "f", "id": 241306, "pid": 76337, "tid": -914061504, "ts": 1716454225183959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225252951, "dur": 99,
    "args": {
      "External id": 241308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241308,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241308, "pid": 5, "tid": 7, "ts": 1716454225252951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225183965, "dur": 22,
    "args": {
      "External id": 241308,
      "cbid": 211, "correlation": 241308
    }
  },
  {
    "ph": "s", "id": 241308, "pid": 76337, "tid": -914061504, "ts": 1716454225183965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225253052, "dur": 19,
    "args": {
      "External id": 241316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241316,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241316, "pid": 5, "tid": 7, "ts": 1716454225253052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184044, "dur": 13,
    "args": {
      "External id": 241316,
      "cbid": 211, "correlation": 241316
    }
  },
  {
    "ph": "s", "id": 241316, "pid": 76337, "tid": -914061504, "ts": 1716454225184044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225253072, "dur": 38,
    "args": {
      "External id": 241324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241324, "pid": 5, "tid": 7, "ts": 1716454225253072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184087, "dur": 9,
    "args": {
      "External id": 241324,
      "cbid": 211, "correlation": 241324
    }
  },
  {
    "ph": "s", "id": 241324, "pid": 76337, "tid": -914061504, "ts": 1716454225184087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225253111, "dur": 34,
    "args": {
      "External id": 241346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241346,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241346, "pid": 5, "tid": 7, "ts": 1716454225253111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184138, "dur": 10,
    "args": {
      "External id": 241346,
      "cbid": 211, "correlation": 241346
    }
  },
  {
    "ph": "s", "id": 241346, "pid": 76337, "tid": -914061504, "ts": 1716454225184138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184228, "dur": 1,
    "args": {
      "External id": 241362,
      "cbid": 251, "correlation": 241362
    }
  },
  {
    "ph": "f", "id": 241362, "pid": 76337, "tid": -914061504, "ts": 1716454225184228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184233, "dur": 0,
    "args": {
      "External id": 241364,
      "cbid": 251, "correlation": 241364
    }
  },
  {
    "ph": "f", "id": 241364, "pid": 76337, "tid": -914061504, "ts": 1716454225184233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225253146, "dur": 531,
    "args": {
      "External id": 241365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241365,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 241365, "pid": 5, "tid": 7, "ts": 1716454225253146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184238, "dur": 13,
    "args": {
      "External id": 241365,
      "cbid": 211, "correlation": 241365
    }
  },
  {
    "ph": "s", "id": 241365, "pid": 76337, "tid": -914061504, "ts": 1716454225184238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225253679, "dur": 123,
    "args": {
      "External id": 241373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241373, "pid": 5, "tid": 7, "ts": 1716454225253679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184303, "dur": 12,
    "args": {
      "External id": 241373,
      "cbid": 211, "correlation": 241373
    }
  },
  {
    "ph": "s", "id": 241373, "pid": 76337, "tid": -914061504, "ts": 1716454225184303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225253803, "dur": 129,
    "args": {
      "External id": 241381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241381, "pid": 5, "tid": 7, "ts": 1716454225253803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184334, "dur": 8,
    "args": {
      "External id": 241381,
      "cbid": 211, "correlation": 241381
    }
  },
  {
    "ph": "s", "id": 241381, "pid": 76337, "tid": -914061504, "ts": 1716454225184334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184410, "dur": 1,
    "args": {
      "External id": 241397,
      "cbid": 251, "correlation": 241397
    }
  },
  {
    "ph": "f", "id": 241397, "pid": 76337, "tid": -914061504, "ts": 1716454225184410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225253933, "dur": 303,
    "args": {
      "External id": 241399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241399,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241399, "pid": 5, "tid": 7, "ts": 1716454225253933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184415, "dur": 13,
    "args": {
      "External id": 241399,
      "cbid": 211, "correlation": 241399
    }
  },
  {
    "ph": "s", "id": 241399, "pid": 76337, "tid": -914061504, "ts": 1716454225184415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225254238, "dur": 28,
    "args": {
      "External id": 241407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241407,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241407, "pid": 5, "tid": 7, "ts": 1716454225254238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184458, "dur": 10,
    "args": {
      "External id": 241407,
      "cbid": 211, "correlation": 241407
    }
  },
  {
    "ph": "s", "id": 241407, "pid": 76337, "tid": -914061504, "ts": 1716454225184458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225254267, "dur": 80,
    "args": {
      "External id": 241418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241418, "pid": 5, "tid": 7, "ts": 1716454225254267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184525, "dur": 13,
    "args": {
      "External id": 241418,
      "cbid": 211, "correlation": 241418
    }
  },
  {
    "ph": "s", "id": 241418, "pid": 76337, "tid": -914061504, "ts": 1716454225184525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184589, "dur": 0,
    "args": {
      "External id": 241430,
      "cbid": 317, "correlation": 241430
    }
  },
  {
    "ph": "f", "id": 241430, "pid": 76337, "tid": -914061504, "ts": 1716454225184589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184590, "dur": 0,
    "args": {
      "External id": 241431,
      "cbid": 203, "correlation": 241431
    }
  },
  {
    "ph": "f", "id": 241431, "pid": 76337, "tid": -914061504, "ts": 1716454225184590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184590, "dur": 0,
    "args": {
      "External id": 241432,
      "cbid": 205, "correlation": 241432
    }
  },
  {
    "ph": "f", "id": 241432, "pid": 76337, "tid": -914061504, "ts": 1716454225184590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225254348, "dur": 23,
    "args": {
      "External id": 241436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241436,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241436, "pid": 5, "tid": 7, "ts": 1716454225254348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184605, "dur": 12,
    "args": {
      "External id": 241436,
      "cbid": 211, "correlation": 241436
    }
  },
  {
    "ph": "s", "id": 241436, "pid": 76337, "tid": -914061504, "ts": 1716454225184605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225254373, "dur": 118,
    "args": {
      "External id": 241438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241438,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241438, "pid": 5, "tid": 7, "ts": 1716454225254373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184624, "dur": 7,
    "args": {
      "External id": 241438,
      "cbid": 211, "correlation": 241438
    }
  },
  {
    "ph": "s", "id": 241438, "pid": 76337, "tid": -914061504, "ts": 1716454225184624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225254492, "dur": 23,
    "args": {
      "External id": 241440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241440,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241440, "pid": 5, "tid": 7, "ts": 1716454225254492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184635, "dur": 5,
    "args": {
      "External id": 241440,
      "cbid": 211, "correlation": 241440
    }
  },
  {
    "ph": "s", "id": 241440, "pid": 76337, "tid": -914061504, "ts": 1716454225184635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225254517, "dur": 32,
    "args": {
      "External id": 241446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241446,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241446, "pid": 5, "tid": 7, "ts": 1716454225254517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184662, "dur": 8,
    "args": {
      "External id": 241446,
      "cbid": 211, "correlation": 241446
    }
  },
  {
    "ph": "s", "id": 241446, "pid": 76337, "tid": -914061504, "ts": 1716454225184662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225254550, "dur": 27,
    "args": {
      "External id": 241454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241454,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241454, "pid": 5, "tid": 7, "ts": 1716454225254550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184694, "dur": 9,
    "args": {
      "External id": 241454,
      "cbid": 211, "correlation": 241454
    }
  },
  {
    "ph": "s", "id": 241454, "pid": 76337, "tid": -914061504, "ts": 1716454225184694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225254578, "dur": 53,
    "args": {
      "External id": 241463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241463,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241463, "pid": 5, "tid": 7, "ts": 1716454225254578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184733, "dur": 10,
    "args": {
      "External id": 241463,
      "cbid": 211, "correlation": 241463
    }
  },
  {
    "ph": "s", "id": 241463, "pid": 76337, "tid": -914061504, "ts": 1716454225184733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225254633, "dur": 52,
    "args": {
      "External id": 241483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241483,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 241483, "pid": 5, "tid": 7, "ts": 1716454225254633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184803, "dur": 11,
    "args": {
      "External id": 241483,
      "cbid": 211, "correlation": 241483
    }
  },
  {
    "ph": "s", "id": 241483, "pid": 76337, "tid": -914061504, "ts": 1716454225184803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225254686, "dur": 4,
    "args": {
      "External id": 241495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241495,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 241495, "pid": 5, "tid": 7, "ts": 1716454225254686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184824, "dur": 7,
    "args": {
      "External id": 241495,
      "cbid": 211, "correlation": 241495
    }
  },
  {
    "ph": "s", "id": 241495, "pid": 76337, "tid": -914061504, "ts": 1716454225184824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225254692, "dur": 56,
    "args": {
      "External id": 241498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241498, "pid": 5, "tid": 7, "ts": 1716454225254692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184843, "dur": 7,
    "args": {
      "External id": 241498,
      "cbid": 211, "correlation": 241498
    }
  },
  {
    "ph": "s", "id": 241498, "pid": 76337, "tid": -914061504, "ts": 1716454225184843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225254749, "dur": 38,
    "args": {
      "External id": 241507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241507,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241507, "pid": 5, "tid": 7, "ts": 1716454225254749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184882, "dur": 10,
    "args": {
      "External id": 241507,
      "cbid": 211, "correlation": 241507
    }
  },
  {
    "ph": "s", "id": 241507, "pid": 76337, "tid": -914061504, "ts": 1716454225184882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184935, "dur": 0,
    "args": {
      "External id": 241517,
      "cbid": 317, "correlation": 241517
    }
  },
  {
    "ph": "f", "id": 241517, "pid": 76337, "tid": -914061504, "ts": 1716454225184935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184936, "dur": 0,
    "args": {
      "External id": 241518,
      "cbid": 203, "correlation": 241518
    }
  },
  {
    "ph": "f", "id": 241518, "pid": 76337, "tid": -914061504, "ts": 1716454225184936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184937, "dur": 0,
    "args": {
      "External id": 241519,
      "cbid": 205, "correlation": 241519
    }
  },
  {
    "ph": "f", "id": 241519, "pid": 76337, "tid": -914061504, "ts": 1716454225184937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225254788, "dur": 39,
    "args": {
      "External id": 241523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241523,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241523, "pid": 5, "tid": 7, "ts": 1716454225254788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184951, "dur": 11,
    "args": {
      "External id": 241523,
      "cbid": 211, "correlation": 241523
    }
  },
  {
    "ph": "s", "id": 241523, "pid": 76337, "tid": -914061504, "ts": 1716454225184951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225254829, "dur": 81,
    "args": {
      "External id": 241525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241525,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241525, "pid": 5, "tid": 7, "ts": 1716454225254829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184964, "dur": 5,
    "args": {
      "External id": 241525,
      "cbid": 211, "correlation": 241525
    }
  },
  {
    "ph": "s", "id": 241525, "pid": 76337, "tid": -914061504, "ts": 1716454225184964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225254912, "dur": 1264,
    "args": {
      "External id": 241527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241527,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241527, "pid": 5, "tid": 7, "ts": 1716454225254912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184985, "dur": 7,
    "args": {
      "External id": 241527,
      "cbid": 211, "correlation": 241527
    }
  },
  {
    "ph": "s", "id": 241527, "pid": 76337, "tid": -914061504, "ts": 1716454225184985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225256177, "dur": 23,
    "args": {
      "External id": 241529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241529,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241529, "pid": 5, "tid": 7, "ts": 1716454225256177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225184995, "dur": 5,
    "args": {
      "External id": 241529,
      "cbid": 211, "correlation": 241529
    }
  },
  {
    "ph": "s", "id": 241529, "pid": 76337, "tid": -914061504, "ts": 1716454225184995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225256201, "dur": 33,
    "args": {
      "External id": 241535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241535, "pid": 5, "tid": 7, "ts": 1716454225256201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185023, "dur": 9,
    "args": {
      "External id": 241535,
      "cbid": 211, "correlation": 241535
    }
  },
  {
    "ph": "s", "id": 241535, "pid": 76337, "tid": -914061504, "ts": 1716454225185023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225256235, "dur": 4,
    "args": {
      "External id": 241543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241543,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 241543, "pid": 5, "tid": 7, "ts": 1716454225256235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185066, "dur": 10,
    "args": {
      "External id": 241543,
      "cbid": 211, "correlation": 241543
    }
  },
  {
    "ph": "s", "id": 241543, "pid": 76337, "tid": -914061504, "ts": 1716454225185066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185130, "dur": 1,
    "args": {
      "External id": 241559,
      "cbid": 251, "correlation": 241559
    }
  },
  {
    "ph": "f", "id": 241559, "pid": 76337, "tid": -914061504, "ts": 1716454225185130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185135, "dur": 0,
    "args": {
      "External id": 241561,
      "cbid": 251, "correlation": 241561
    }
  },
  {
    "ph": "f", "id": 241561, "pid": 76337, "tid": -914061504, "ts": 1716454225185135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225256240, "dur": 11,
    "args": {
      "External id": 241562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241562,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 241562, "pid": 5, "tid": 7, "ts": 1716454225256240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185137, "dur": 11,
    "args": {
      "External id": 241562,
      "cbid": 211, "correlation": 241562
    }
  },
  {
    "ph": "s", "id": 241562, "pid": 76337, "tid": -914061504, "ts": 1716454225185137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225256253, "dur": 5,
    "args": {
      "External id": 241564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241564,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 241564, "pid": 5, "tid": 7, "ts": 1716454225256253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185150, "dur": 5,
    "args": {
      "External id": 241564,
      "cbid": 211, "correlation": 241564
    }
  },
  {
    "ph": "s", "id": 241564, "pid": 76337, "tid": -914061504, "ts": 1716454225185150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225256259, "dur": 28,
    "args": {
      "External id": 241574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241574,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241574, "pid": 5, "tid": 7, "ts": 1716454225256259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185207, "dur": 12,
    "args": {
      "External id": 241574,
      "cbid": 211, "correlation": 241574
    }
  },
  {
    "ph": "s", "id": 241574, "pid": 76337, "tid": -914061504, "ts": 1716454225185207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225256289, "dur": 31,
    "args": {
      "External id": 241594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241594,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 241594, "pid": 5, "tid": 7, "ts": 1716454225256289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185271, "dur": 11,
    "args": {
      "External id": 241594,
      "cbid": 211, "correlation": 241594
    }
  },
  {
    "ph": "s", "id": 241594, "pid": 76337, "tid": -914061504, "ts": 1716454225185271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225256321, "dur": 4,
    "args": {
      "External id": 241606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241606,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 241606, "pid": 5, "tid": 7, "ts": 1716454225256321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185291, "dur": 6,
    "args": {
      "External id": 241606,
      "cbid": 211, "correlation": 241606
    }
  },
  {
    "ph": "s", "id": 241606, "pid": 76337, "tid": -914061504, "ts": 1716454225185291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225256326, "dur": 31,
    "args": {
      "External id": 241609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241609, "pid": 5, "tid": 7, "ts": 1716454225256326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185310, "dur": 7,
    "args": {
      "External id": 241609,
      "cbid": 211, "correlation": 241609
    }
  },
  {
    "ph": "s", "id": 241609, "pid": 76337, "tid": -914061504, "ts": 1716454225185310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225256358, "dur": 20,
    "args": {
      "External id": 241618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241618,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241618, "pid": 5, "tid": 7, "ts": 1716454225256358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185351, "dur": 11,
    "args": {
      "External id": 241618,
      "cbid": 211, "correlation": 241618
    }
  },
  {
    "ph": "s", "id": 241618, "pid": 76337, "tid": -914061504, "ts": 1716454225185351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185414, "dur": 0,
    "args": {
      "External id": 241628,
      "cbid": 317, "correlation": 241628
    }
  },
  {
    "ph": "f", "id": 241628, "pid": 76337, "tid": -914061504, "ts": 1716454225185414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185415, "dur": 0,
    "args": {
      "External id": 241629,
      "cbid": 203, "correlation": 241629
    }
  },
  {
    "ph": "f", "id": 241629, "pid": 76337, "tid": -914061504, "ts": 1716454225185415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185416, "dur": 0,
    "args": {
      "External id": 241630,
      "cbid": 205, "correlation": 241630
    }
  },
  {
    "ph": "f", "id": 241630, "pid": 76337, "tid": -914061504, "ts": 1716454225185416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225256380, "dur": 22,
    "args": {
      "External id": 241634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241634,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241634, "pid": 5, "tid": 7, "ts": 1716454225256380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185430, "dur": 12,
    "args": {
      "External id": 241634,
      "cbid": 211, "correlation": 241634
    }
  },
  {
    "ph": "s", "id": 241634, "pid": 76337, "tid": -914061504, "ts": 1716454225185430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225256403, "dur": 43,
    "args": {
      "External id": 241636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241636,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241636, "pid": 5, "tid": 7, "ts": 1716454225256403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185445, "dur": 5,
    "args": {
      "External id": 241636,
      "cbid": 211, "correlation": 241636
    }
  },
  {
    "ph": "s", "id": 241636, "pid": 76337, "tid": -914061504, "ts": 1716454225185445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225256448, "dur": 638,
    "args": {
      "External id": 241638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241638,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241638, "pid": 5, "tid": 7, "ts": 1716454225256448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185456, "dur": 6,
    "args": {
      "External id": 241638,
      "cbid": 211, "correlation": 241638
    }
  },
  {
    "ph": "s", "id": 241638, "pid": 76337, "tid": -914061504, "ts": 1716454225185456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257088, "dur": 24,
    "args": {
      "External id": 241640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241640,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241640, "pid": 5, "tid": 7, "ts": 1716454225257088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185465, "dur": 5,
    "args": {
      "External id": 241640,
      "cbid": 211, "correlation": 241640
    }
  },
  {
    "ph": "s", "id": 241640, "pid": 76337, "tid": -914061504, "ts": 1716454225185465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225257112, "dur": 32,
    "args": {
      "External id": 241646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241646, "pid": 5, "tid": 7, "ts": 1716454225257112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185493, "dur": 8,
    "args": {
      "External id": 241646,
      "cbid": 211, "correlation": 241646
    }
  },
  {
    "ph": "s", "id": 241646, "pid": 76337, "tid": -914061504, "ts": 1716454225185493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185551, "dur": 0,
    "args": {
      "External id": 241656,
      "cbid": 317, "correlation": 241656
    }
  },
  {
    "ph": "f", "id": 241656, "pid": 76337, "tid": -914061504, "ts": 1716454225185551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185552, "dur": 0,
    "args": {
      "External id": 241657,
      "cbid": 203, "correlation": 241657
    }
  },
  {
    "ph": "f", "id": 241657, "pid": 76337, "tid": -914061504, "ts": 1716454225185552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185552, "dur": 0,
    "args": {
      "External id": 241658,
      "cbid": 205, "correlation": 241658
    }
  },
  {
    "ph": "f", "id": 241658, "pid": 76337, "tid": -914061504, "ts": 1716454225185552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257146, "dur": 38,
    "args": {
      "External id": 241662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241662,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241662, "pid": 5, "tid": 7, "ts": 1716454225257146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185564, "dur": 12,
    "args": {
      "External id": 241662,
      "cbid": 211, "correlation": 241662
    }
  },
  {
    "ph": "s", "id": 241662, "pid": 76337, "tid": -914061504, "ts": 1716454225185564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225257186, "dur": 186,
    "args": {
      "External id": 241664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241664,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241664, "pid": 5, "tid": 7, "ts": 1716454225257186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185582, "dur": 6,
    "args": {
      "External id": 241664,
      "cbid": 211, "correlation": 241664
    }
  },
  {
    "ph": "s", "id": 241664, "pid": 76337, "tid": -914061504, "ts": 1716454225185582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257373, "dur": 21,
    "args": {
      "External id": 241666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241666, "pid": 5, "tid": 7, "ts": 1716454225257373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185591, "dur": 5,
    "args": {
      "External id": 241666,
      "cbid": 211, "correlation": 241666
    }
  },
  {
    "ph": "s", "id": 241666, "pid": 76337, "tid": -914061504, "ts": 1716454225185591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225257395, "dur": 32,
    "args": {
      "External id": 241672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241672, "pid": 5, "tid": 7, "ts": 1716454225257395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185617, "dur": 9,
    "args": {
      "External id": 241672,
      "cbid": 211, "correlation": 241672
    }
  },
  {
    "ph": "s", "id": 241672, "pid": 76337, "tid": -914061504, "ts": 1716454225185617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225257428, "dur": 27,
    "args": {
      "External id": 241680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241680,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241680, "pid": 5, "tid": 7, "ts": 1716454225257428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185646, "dur": 8,
    "args": {
      "External id": 241680,
      "cbid": 211, "correlation": 241680
    }
  },
  {
    "ph": "s", "id": 241680, "pid": 76337, "tid": -914061504, "ts": 1716454225185646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225257457, "dur": 20,
    "args": {
      "External id": 241688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241688,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241688, "pid": 5, "tid": 7, "ts": 1716454225257457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185674, "dur": 8,
    "args": {
      "External id": 241688,
      "cbid": 211, "correlation": 241688
    }
  },
  {
    "ph": "s", "id": 241688, "pid": 76337, "tid": -914061504, "ts": 1716454225185674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257478, "dur": 29,
    "args": {
      "External id": 241708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241708,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 241708, "pid": 5, "tid": 7, "ts": 1716454225257478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185755, "dur": 12,
    "args": {
      "External id": 241708,
      "cbid": 211, "correlation": 241708
    }
  },
  {
    "ph": "s", "id": 241708, "pid": 76337, "tid": -914061504, "ts": 1716454225185755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225257509, "dur": 4,
    "args": {
      "External id": 241720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241720,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 241720, "pid": 5, "tid": 7, "ts": 1716454225257509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185777, "dur": 7,
    "args": {
      "External id": 241720,
      "cbid": 211, "correlation": 241720
    }
  },
  {
    "ph": "s", "id": 241720, "pid": 76337, "tid": -914061504, "ts": 1716454225185777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225257514, "dur": 30,
    "args": {
      "External id": 241723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241723,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241723, "pid": 5, "tid": 7, "ts": 1716454225257514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185795, "dur": 6,
    "args": {
      "External id": 241723,
      "cbid": 211, "correlation": 241723
    }
  },
  {
    "ph": "s", "id": 241723, "pid": 76337, "tid": -914061504, "ts": 1716454225185795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185852, "dur": 0,
    "args": {
      "External id": 241734,
      "cbid": 317, "correlation": 241734
    }
  },
  {
    "ph": "f", "id": 241734, "pid": 76337, "tid": -914061504, "ts": 1716454225185852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185853, "dur": 0,
    "args": {
      "External id": 241735,
      "cbid": 203, "correlation": 241735
    }
  },
  {
    "ph": "f", "id": 241735, "pid": 76337, "tid": -914061504, "ts": 1716454225185853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185854, "dur": 0,
    "args": {
      "External id": 241736,
      "cbid": 205, "correlation": 241736
    }
  },
  {
    "ph": "f", "id": 241736, "pid": 76337, "tid": -914061504, "ts": 1716454225185854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257545, "dur": 23,
    "args": {
      "External id": 241740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241740,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241740, "pid": 5, "tid": 7, "ts": 1716454225257545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185866, "dur": 12,
    "args": {
      "External id": 241740,
      "cbid": 211, "correlation": 241740
    }
  },
  {
    "ph": "s", "id": 241740, "pid": 76337, "tid": -914061504, "ts": 1716454225185866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225257569, "dur": 103,
    "args": {
      "External id": 241742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241742,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241742, "pid": 5, "tid": 7, "ts": 1716454225257569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185884, "dur": 6,
    "args": {
      "External id": 241742,
      "cbid": 211, "correlation": 241742
    }
  },
  {
    "ph": "s", "id": 241742, "pid": 76337, "tid": -914061504, "ts": 1716454225185884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257674, "dur": 23,
    "args": {
      "External id": 241744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241744, "pid": 5, "tid": 7, "ts": 1716454225257674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185894, "dur": 5,
    "args": {
      "External id": 241744,
      "cbid": 211, "correlation": 241744
    }
  },
  {
    "ph": "s", "id": 241744, "pid": 76337, "tid": -914061504, "ts": 1716454225185894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225257698, "dur": 32,
    "args": {
      "External id": 241750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241750, "pid": 5, "tid": 7, "ts": 1716454225257698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225185920, "dur": 9,
    "args": {
      "External id": 241750,
      "cbid": 211, "correlation": 241750
    }
  },
  {
    "ph": "s", "id": 241750, "pid": 76337, "tid": -914061504, "ts": 1716454225185920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225257731, "dur": 184,
    "args": {
      "External id": 241759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241759,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241759, "pid": 5, "tid": 7, "ts": 1716454225257731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186010, "dur": 14,
    "args": {
      "External id": 241759,
      "cbid": 211, "correlation": 241759
    }
  },
  {
    "ph": "s", "id": 241759, "pid": 76337, "tid": -914061504, "ts": 1716454225186010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225257917, "dur": 63,
    "args": {
      "External id": 241781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241781,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241781, "pid": 5, "tid": 7, "ts": 1716454225257917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186067, "dur": 10,
    "args": {
      "External id": 241781,
      "cbid": 211, "correlation": 241781
    }
  },
  {
    "ph": "s", "id": 241781, "pid": 76337, "tid": -914061504, "ts": 1716454225186067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186154, "dur": 1,
    "args": {
      "External id": 241792,
      "cbid": 251, "correlation": 241792
    }
  },
  {
    "ph": "f", "id": 241792, "pid": 76337, "tid": -914061504, "ts": 1716454225186154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225257981, "dur": 153,
    "args": {
      "External id": 241793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241793,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241793, "pid": 5, "tid": 7, "ts": 1716454225257981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186159, "dur": 13,
    "args": {
      "External id": 241793,
      "cbid": 211, "correlation": 241793
    }
  },
  {
    "ph": "s", "id": 241793, "pid": 76337, "tid": -914061504, "ts": 1716454225186159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186229, "dur": 1,
    "args": {
      "External id": 241804,
      "cbid": 251, "correlation": 241804
    }
  },
  {
    "ph": "f", "id": 241804, "pid": 76337, "tid": -914061504, "ts": 1716454225186229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225258136, "dur": 145,
    "args": {
      "External id": 241805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241805,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241805, "pid": 5, "tid": 7, "ts": 1716454225258136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186233, "dur": 11,
    "args": {
      "External id": 241805,
      "cbid": 211, "correlation": 241805
    }
  },
  {
    "ph": "s", "id": 241805, "pid": 76337, "tid": -914061504, "ts": 1716454225186233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186297, "dur": 1,
    "args": {
      "External id": 241816,
      "cbid": 251, "correlation": 241816
    }
  },
  {
    "ph": "f", "id": 241816, "pid": 76337, "tid": -914061504, "ts": 1716454225186297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225258283, "dur": 143,
    "args": {
      "External id": 241817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241817,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241817, "pid": 5, "tid": 7, "ts": 1716454225258283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186301, "dur": 11,
    "args": {
      "External id": 241817,
      "cbid": 211, "correlation": 241817
    }
  },
  {
    "ph": "s", "id": 241817, "pid": 76337, "tid": -914061504, "ts": 1716454225186301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225258426, "dur": 1905,
    "args": {
      "External id": 241838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241838,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 241838, "pid": 5, "tid": 7, "ts": 1716454225258426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186383, "dur": 13,
    "args": {
      "External id": 241838,
      "cbid": 211, "correlation": 241838
    }
  },
  {
    "ph": "s", "id": 241838, "pid": 76337, "tid": -914061504, "ts": 1716454225186383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186480, "dur": 1,
    "args": {
      "External id": 241856,
      "cbid": 251, "correlation": 241856
    }
  },
  {
    "ph": "f", "id": 241856, "pid": 76337, "tid": -914061504, "ts": 1716454225186480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225260333, "dur": 146,
    "args": {
      "External id": 241858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241858,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 241858, "pid": 5, "tid": 7, "ts": 1716454225260333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186486, "dur": 13,
    "args": {
      "External id": 241858,
      "cbid": 211, "correlation": 241858
    }
  },
  {
    "ph": "s", "id": 241858, "pid": 76337, "tid": -914061504, "ts": 1716454225186486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225260481, "dur": 36,
    "args": {
      "External id": 241866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241866,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241866, "pid": 5, "tid": 7, "ts": 1716454225260481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186555, "dur": 12,
    "args": {
      "External id": 241866,
      "cbid": 211, "correlation": 241866
    }
  },
  {
    "ph": "s", "id": 241866, "pid": 76337, "tid": -914061504, "ts": 1716454225186555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225260518, "dur": 51,
    "args": {
      "External id": 241874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241874,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241874, "pid": 5, "tid": 7, "ts": 1716454225260518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186595, "dur": 8,
    "args": {
      "External id": 241874,
      "cbid": 211, "correlation": 241874
    }
  },
  {
    "ph": "s", "id": 241874, "pid": 76337, "tid": -914061504, "ts": 1716454225186595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225260570, "dur": 29,
    "args": {
      "External id": 241885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241885,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241885, "pid": 5, "tid": 7, "ts": 1716454225260570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186666, "dur": 13,
    "args": {
      "External id": 241885,
      "cbid": 211, "correlation": 241885
    }
  },
  {
    "ph": "s", "id": 241885, "pid": 76337, "tid": -914061504, "ts": 1716454225186666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225260600, "dur": 34,
    "args": {
      "External id": 241907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241907,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241907, "pid": 5, "tid": 7, "ts": 1716454225260600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186698, "dur": 7,
    "args": {
      "External id": 241907,
      "cbid": 211, "correlation": 241907
    }
  },
  {
    "ph": "s", "id": 241907, "pid": 76337, "tid": -914061504, "ts": 1716454225186698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186782, "dur": 1,
    "args": {
      "External id": 241918,
      "cbid": 251, "correlation": 241918
    }
  },
  {
    "ph": "f", "id": 241918, "pid": 76337, "tid": -914061504, "ts": 1716454225186782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225260635, "dur": 88,
    "args": {
      "External id": 241919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241919,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241919, "pid": 5, "tid": 7, "ts": 1716454225260635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186787, "dur": 13,
    "args": {
      "External id": 241919,
      "cbid": 211, "correlation": 241919
    }
  },
  {
    "ph": "s", "id": 241919, "pid": 76337, "tid": -914061504, "ts": 1716454225186787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186855, "dur": 1,
    "args": {
      "External id": 241930,
      "cbid": 251, "correlation": 241930
    }
  },
  {
    "ph": "f", "id": 241930, "pid": 76337, "tid": -914061504, "ts": 1716454225186855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186859, "dur": 0,
    "args": {
      "External id": 241931,
      "cbid": 251, "correlation": 241931
    }
  },
  {
    "ph": "f", "id": 241931, "pid": 76337, "tid": -914061504, "ts": 1716454225186859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225260725, "dur": 11,
    "args": {
      "External id": 241932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241932,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 241932, "pid": 5, "tid": 7, "ts": 1716454225260725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186861, "dur": 11,
    "args": {
      "External id": 241932,
      "cbid": 211, "correlation": 241932
    }
  },
  {
    "ph": "s", "id": 241932, "pid": 76337, "tid": -914061504, "ts": 1716454225186861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225260737, "dur": 5,
    "args": {
      "External id": 241934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241934,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 241934, "pid": 5, "tid": 7, "ts": 1716454225260737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186874, "dur": 6,
    "args": {
      "External id": 241934,
      "cbid": 211, "correlation": 241934
    }
  },
  {
    "ph": "s", "id": 241934, "pid": 76337, "tid": -914061504, "ts": 1716454225186874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186932, "dur": 1,
    "args": {
      "External id": 241945,
      "cbid": 251, "correlation": 241945
    }
  },
  {
    "ph": "f", "id": 241945, "pid": 76337, "tid": -914061504, "ts": 1716454225186932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186937, "dur": 0,
    "args": {
      "External id": 241946,
      "cbid": 251, "correlation": 241946
    }
  },
  {
    "ph": "f", "id": 241946, "pid": 76337, "tid": -914061504, "ts": 1716454225186937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225260743, "dur": 7,
    "args": {
      "External id": 241947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241947,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 241947, "pid": 5, "tid": 7, "ts": 1716454225260743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186938, "dur": 11,
    "args": {
      "External id": 241947,
      "cbid": 211, "correlation": 241947
    }
  },
  {
    "ph": "s", "id": 241947, "pid": 76337, "tid": -914061504, "ts": 1716454225186938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225260752, "dur": 4,
    "args": {
      "External id": 241949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241949,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 241949, "pid": 5, "tid": 7, "ts": 1716454225260752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225186951, "dur": 5,
    "args": {
      "External id": 241949,
      "cbid": 211, "correlation": 241949
    }
  },
  {
    "ph": "s", "id": 241949, "pid": 76337, "tid": -914061504, "ts": 1716454225186951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225260756, "dur": 89,
    "args": {
      "External id": 241970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241970,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 241970, "pid": 5, "tid": 7, "ts": 1716454225260756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187033, "dur": 13,
    "args": {
      "External id": 241970,
      "cbid": 211, "correlation": 241970
    }
  },
  {
    "ph": "s", "id": 241970, "pid": 76337, "tid": -914061504, "ts": 1716454225187033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187129, "dur": 1,
    "args": {
      "External id": 241988,
      "cbid": 251, "correlation": 241988
    }
  },
  {
    "ph": "f", "id": 241988, "pid": 76337, "tid": -914061504, "ts": 1716454225187129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225260847, "dur": 94,
    "args": {
      "External id": 241990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241990,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 241990, "pid": 5, "tid": 7, "ts": 1716454225260847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187136, "dur": 13,
    "args": {
      "External id": 241990,
      "cbid": 211, "correlation": 241990
    }
  },
  {
    "ph": "s", "id": 241990, "pid": 76337, "tid": -914061504, "ts": 1716454225187136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225260942, "dur": 19,
    "args": {
      "External id": 241998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 241998,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 241998, "pid": 5, "tid": 7, "ts": 1716454225260942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187204, "dur": 12,
    "args": {
      "External id": 241998,
      "cbid": 211, "correlation": 241998
    }
  },
  {
    "ph": "s", "id": 241998, "pid": 76337, "tid": -914061504, "ts": 1716454225187204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225260962, "dur": 37,
    "args": {
      "External id": 242006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242006, "pid": 5, "tid": 7, "ts": 1716454225260962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187245, "dur": 9,
    "args": {
      "External id": 242006,
      "cbid": 211, "correlation": 242006
    }
  },
  {
    "ph": "s", "id": 242006, "pid": 76337, "tid": -914061504, "ts": 1716454225187245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225261001, "dur": 34,
    "args": {
      "External id": 242028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242028,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242028, "pid": 5, "tid": 7, "ts": 1716454225261001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187296, "dur": 10,
    "args": {
      "External id": 242028,
      "cbid": 211, "correlation": 242028
    }
  },
  {
    "ph": "s", "id": 242028, "pid": 76337, "tid": -914061504, "ts": 1716454225187296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187385, "dur": 1,
    "args": {
      "External id": 242044,
      "cbid": 251, "correlation": 242044
    }
  },
  {
    "ph": "f", "id": 242044, "pid": 76337, "tid": -914061504, "ts": 1716454225187385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187390, "dur": 0,
    "args": {
      "External id": 242046,
      "cbid": 251, "correlation": 242046
    }
  },
  {
    "ph": "f", "id": 242046, "pid": 76337, "tid": -914061504, "ts": 1716454225187390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225261036, "dur": 529,
    "args": {
      "External id": 242047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242047,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 242047, "pid": 5, "tid": 7, "ts": 1716454225261036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187394, "dur": 13,
    "args": {
      "External id": 242047,
      "cbid": 211, "correlation": 242047
    }
  },
  {
    "ph": "s", "id": 242047, "pid": 76337, "tid": -914061504, "ts": 1716454225187394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225261566, "dur": 124,
    "args": {
      "External id": 242055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242055, "pid": 5, "tid": 7, "ts": 1716454225261566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187458, "dur": 12,
    "args": {
      "External id": 242055,
      "cbid": 211, "correlation": 242055
    }
  },
  {
    "ph": "s", "id": 242055, "pid": 76337, "tid": -914061504, "ts": 1716454225187458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225261692, "dur": 129,
    "args": {
      "External id": 242063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242063,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242063, "pid": 5, "tid": 7, "ts": 1716454225261692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187489, "dur": 8,
    "args": {
      "External id": 242063,
      "cbid": 211, "correlation": 242063
    }
  },
  {
    "ph": "s", "id": 242063, "pid": 76337, "tid": -914061504, "ts": 1716454225187489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187565, "dur": 1,
    "args": {
      "External id": 242079,
      "cbid": 251, "correlation": 242079
    }
  },
  {
    "ph": "f", "id": 242079, "pid": 76337, "tid": -914061504, "ts": 1716454225187565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225261822, "dur": 303,
    "args": {
      "External id": 242081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242081,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242081, "pid": 5, "tid": 7, "ts": 1716454225261822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187571, "dur": 13,
    "args": {
      "External id": 242081,
      "cbid": 211, "correlation": 242081
    }
  },
  {
    "ph": "s", "id": 242081, "pid": 76337, "tid": -914061504, "ts": 1716454225187571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225262126, "dur": 27,
    "args": {
      "External id": 242089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242089,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242089, "pid": 5, "tid": 7, "ts": 1716454225262126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187613, "dur": 10,
    "args": {
      "External id": 242089,
      "cbid": 211, "correlation": 242089
    }
  },
  {
    "ph": "s", "id": 242089, "pid": 76337, "tid": -914061504, "ts": 1716454225187613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225262154, "dur": 79,
    "args": {
      "External id": 242100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242100, "pid": 5, "tid": 7, "ts": 1716454225262154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187680, "dur": 13,
    "args": {
      "External id": 242100,
      "cbid": 211, "correlation": 242100
    }
  },
  {
    "ph": "s", "id": 242100, "pid": 76337, "tid": -914061504, "ts": 1716454225187680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187744, "dur": 0,
    "args": {
      "External id": 242112,
      "cbid": 317, "correlation": 242112
    }
  },
  {
    "ph": "f", "id": 242112, "pid": 76337, "tid": -914061504, "ts": 1716454225187744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187745, "dur": 0,
    "args": {
      "External id": 242113,
      "cbid": 203, "correlation": 242113
    }
  },
  {
    "ph": "f", "id": 242113, "pid": 76337, "tid": -914061504, "ts": 1716454225187745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187746, "dur": 0,
    "args": {
      "External id": 242114,
      "cbid": 205, "correlation": 242114
    }
  },
  {
    "ph": "f", "id": 242114, "pid": 76337, "tid": -914061504, "ts": 1716454225187746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225262235, "dur": 24,
    "args": {
      "External id": 242118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242118, "pid": 5, "tid": 7, "ts": 1716454225262235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187760, "dur": 12,
    "args": {
      "External id": 242118,
      "cbid": 211, "correlation": 242118
    }
  },
  {
    "ph": "s", "id": 242118, "pid": 76337, "tid": -914061504, "ts": 1716454225187760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225262260, "dur": 118,
    "args": {
      "External id": 242120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242120,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242120, "pid": 5, "tid": 7, "ts": 1716454225262260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187778, "dur": 7,
    "args": {
      "External id": 242120,
      "cbid": 211, "correlation": 242120
    }
  },
  {
    "ph": "s", "id": 242120, "pid": 76337, "tid": -914061504, "ts": 1716454225187778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225262380, "dur": 23,
    "args": {
      "External id": 242122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242122, "pid": 5, "tid": 7, "ts": 1716454225262380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187789, "dur": 5,
    "args": {
      "External id": 242122,
      "cbid": 211, "correlation": 242122
    }
  },
  {
    "ph": "s", "id": 242122, "pid": 76337, "tid": -914061504, "ts": 1716454225187789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225262405, "dur": 33,
    "args": {
      "External id": 242128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242128,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242128, "pid": 5, "tid": 7, "ts": 1716454225262405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187816, "dur": 8,
    "args": {
      "External id": 242128,
      "cbid": 211, "correlation": 242128
    }
  },
  {
    "ph": "s", "id": 242128, "pid": 76337, "tid": -914061504, "ts": 1716454225187816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225262438, "dur": 27,
    "args": {
      "External id": 242136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242136,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242136, "pid": 5, "tid": 7, "ts": 1716454225262438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187848, "dur": 8,
    "args": {
      "External id": 242136,
      "cbid": 211, "correlation": 242136
    }
  },
  {
    "ph": "s", "id": 242136, "pid": 76337, "tid": -914061504, "ts": 1716454225187848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225262467, "dur": 45,
    "args": {
      "External id": 242145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242145,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242145, "pid": 5, "tid": 7, "ts": 1716454225262467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187887, "dur": 10,
    "args": {
      "External id": 242145,
      "cbid": 211, "correlation": 242145
    }
  },
  {
    "ph": "s", "id": 242145, "pid": 76337, "tid": -914061504, "ts": 1716454225187887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225262513, "dur": 42,
    "args": {
      "External id": 242165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242165,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 242165, "pid": 5, "tid": 7, "ts": 1716454225262513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187960, "dur": 11,
    "args": {
      "External id": 242165,
      "cbid": 211, "correlation": 242165
    }
  },
  {
    "ph": "s", "id": 242165, "pid": 76337, "tid": -914061504, "ts": 1716454225187960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225262557, "dur": 5,
    "args": {
      "External id": 242177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242177,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 242177, "pid": 5, "tid": 7, "ts": 1716454225262557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225187990, "dur": 7,
    "args": {
      "External id": 242177,
      "cbid": 211, "correlation": 242177
    }
  },
  {
    "ph": "s", "id": 242177, "pid": 76337, "tid": -914061504, "ts": 1716454225187990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225262563, "dur": 43,
    "args": {
      "External id": 242180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242180, "pid": 5, "tid": 7, "ts": 1716454225262563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188010, "dur": 7,
    "args": {
      "External id": 242180,
      "cbid": 211, "correlation": 242180
    }
  },
  {
    "ph": "s", "id": 242180, "pid": 76337, "tid": -914061504, "ts": 1716454225188010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225262608, "dur": 28,
    "args": {
      "External id": 242189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242189,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242189, "pid": 5, "tid": 7, "ts": 1716454225262608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188050, "dur": 10,
    "args": {
      "External id": 242189,
      "cbid": 211, "correlation": 242189
    }
  },
  {
    "ph": "s", "id": 242189, "pid": 76337, "tid": -914061504, "ts": 1716454225188050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188101, "dur": 0,
    "args": {
      "External id": 242199,
      "cbid": 317, "correlation": 242199
    }
  },
  {
    "ph": "f", "id": 242199, "pid": 76337, "tid": -914061504, "ts": 1716454225188101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188102, "dur": 0,
    "args": {
      "External id": 242200,
      "cbid": 203, "correlation": 242200
    }
  },
  {
    "ph": "f", "id": 242200, "pid": 76337, "tid": -914061504, "ts": 1716454225188102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188103, "dur": 0,
    "args": {
      "External id": 242201,
      "cbid": 205, "correlation": 242201
    }
  },
  {
    "ph": "f", "id": 242201, "pid": 76337, "tid": -914061504, "ts": 1716454225188103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225262637, "dur": 31,
    "args": {
      "External id": 242205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242205,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242205, "pid": 5, "tid": 7, "ts": 1716454225262637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188118, "dur": 11,
    "args": {
      "External id": 242205,
      "cbid": 211, "correlation": 242205
    }
  },
  {
    "ph": "s", "id": 242205, "pid": 76337, "tid": -914061504, "ts": 1716454225188118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225262670, "dur": 63,
    "args": {
      "External id": 242207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242207,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242207, "pid": 5, "tid": 7, "ts": 1716454225262670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188132, "dur": 6,
    "args": {
      "External id": 242207,
      "cbid": 211, "correlation": 242207
    }
  },
  {
    "ph": "s", "id": 242207, "pid": 76337, "tid": -914061504, "ts": 1716454225188132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225262733, "dur": 957,
    "args": {
      "External id": 242209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242209,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242209, "pid": 5, "tid": 7, "ts": 1716454225262733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188144, "dur": 6,
    "args": {
      "External id": 242209,
      "cbid": 211, "correlation": 242209
    }
  },
  {
    "ph": "s", "id": 242209, "pid": 76337, "tid": -914061504, "ts": 1716454225188144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225263692, "dur": 21,
    "args": {
      "External id": 242211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242211,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242211, "pid": 5, "tid": 7, "ts": 1716454225263692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188153, "dur": 5,
    "args": {
      "External id": 242211,
      "cbid": 211, "correlation": 242211
    }
  },
  {
    "ph": "s", "id": 242211, "pid": 76337, "tid": -914061504, "ts": 1716454225188153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225263714, "dur": 32,
    "args": {
      "External id": 242217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242217,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242217, "pid": 5, "tid": 7, "ts": 1716454225263714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188181, "dur": 9,
    "args": {
      "External id": 242217,
      "cbid": 211, "correlation": 242217
    }
  },
  {
    "ph": "s", "id": 242217, "pid": 76337, "tid": -914061504, "ts": 1716454225188181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225263747, "dur": 3,
    "args": {
      "External id": 242225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242225,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 242225, "pid": 5, "tid": 7, "ts": 1716454225263747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188225, "dur": 10,
    "args": {
      "External id": 242225,
      "cbid": 211, "correlation": 242225
    }
  },
  {
    "ph": "s", "id": 242225, "pid": 76337, "tid": -914061504, "ts": 1716454225188225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188289, "dur": 1,
    "args": {
      "External id": 242241,
      "cbid": 251, "correlation": 242241
    }
  },
  {
    "ph": "f", "id": 242241, "pid": 76337, "tid": -914061504, "ts": 1716454225188289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188294, "dur": 0,
    "args": {
      "External id": 242243,
      "cbid": 251, "correlation": 242243
    }
  },
  {
    "ph": "f", "id": 242243, "pid": 76337, "tid": -914061504, "ts": 1716454225188294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225263752, "dur": 12,
    "args": {
      "External id": 242244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242244,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 242244, "pid": 5, "tid": 7, "ts": 1716454225263752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188296, "dur": 11,
    "args": {
      "External id": 242244,
      "cbid": 211, "correlation": 242244
    }
  },
  {
    "ph": "s", "id": 242244, "pid": 76337, "tid": -914061504, "ts": 1716454225188296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225263765, "dur": 5,
    "args": {
      "External id": 242246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242246,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 242246, "pid": 5, "tid": 7, "ts": 1716454225263765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188309, "dur": 6,
    "args": {
      "External id": 242246,
      "cbid": 211, "correlation": 242246
    }
  },
  {
    "ph": "s", "id": 242246, "pid": 76337, "tid": -914061504, "ts": 1716454225188309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225263771, "dur": 29,
    "args": {
      "External id": 242256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242256, "pid": 5, "tid": 7, "ts": 1716454225263771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188366, "dur": 12,
    "args": {
      "External id": 242256,
      "cbid": 211, "correlation": 242256
    }
  },
  {
    "ph": "s", "id": 242256, "pid": 76337, "tid": -914061504, "ts": 1716454225188366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225263802, "dur": 32,
    "args": {
      "External id": 242276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242276,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 242276, "pid": 5, "tid": 7, "ts": 1716454225263802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188430, "dur": 11,
    "args": {
      "External id": 242276,
      "cbid": 211, "correlation": 242276
    }
  },
  {
    "ph": "s", "id": 242276, "pid": 76337, "tid": -914061504, "ts": 1716454225188430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225263835, "dur": 4,
    "args": {
      "External id": 242288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242288,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 242288, "pid": 5, "tid": 7, "ts": 1716454225263835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188451, "dur": 6,
    "args": {
      "External id": 242288,
      "cbid": 211, "correlation": 242288
    }
  },
  {
    "ph": "s", "id": 242288, "pid": 76337, "tid": -914061504, "ts": 1716454225188451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225263840, "dur": 29,
    "args": {
      "External id": 242291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242291,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242291, "pid": 5, "tid": 7, "ts": 1716454225263840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188469, "dur": 7,
    "args": {
      "External id": 242291,
      "cbid": 211, "correlation": 242291
    }
  },
  {
    "ph": "s", "id": 242291, "pid": 76337, "tid": -914061504, "ts": 1716454225188469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225263870, "dur": 21,
    "args": {
      "External id": 242300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242300,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242300, "pid": 5, "tid": 7, "ts": 1716454225263870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188510, "dur": 10,
    "args": {
      "External id": 242300,
      "cbid": 211, "correlation": 242300
    }
  },
  {
    "ph": "s", "id": 242300, "pid": 76337, "tid": -914061504, "ts": 1716454225188510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188571, "dur": 0,
    "args": {
      "External id": 242310,
      "cbid": 317, "correlation": 242310
    }
  },
  {
    "ph": "f", "id": 242310, "pid": 76337, "tid": -914061504, "ts": 1716454225188571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188572, "dur": 0,
    "args": {
      "External id": 242311,
      "cbid": 203, "correlation": 242311
    }
  },
  {
    "ph": "f", "id": 242311, "pid": 76337, "tid": -914061504, "ts": 1716454225188572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188573, "dur": 0,
    "args": {
      "External id": 242312,
      "cbid": 205, "correlation": 242312
    }
  },
  {
    "ph": "f", "id": 242312, "pid": 76337, "tid": -914061504, "ts": 1716454225188573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225263893, "dur": 22,
    "args": {
      "External id": 242316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242316,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242316, "pid": 5, "tid": 7, "ts": 1716454225263893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188587, "dur": 12,
    "args": {
      "External id": 242316,
      "cbid": 211, "correlation": 242316
    }
  },
  {
    "ph": "s", "id": 242316, "pid": 76337, "tid": -914061504, "ts": 1716454225188587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225263916, "dur": 44,
    "args": {
      "External id": 242318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242318, "pid": 5, "tid": 7, "ts": 1716454225263916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188601, "dur": 5,
    "args": {
      "External id": 242318,
      "cbid": 211, "correlation": 242318
    }
  },
  {
    "ph": "s", "id": 242318, "pid": 76337, "tid": -914061504, "ts": 1716454225188601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225263961, "dur": 638,
    "args": {
      "External id": 242320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242320,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242320, "pid": 5, "tid": 7, "ts": 1716454225263961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188612, "dur": 6,
    "args": {
      "External id": 242320,
      "cbid": 211, "correlation": 242320
    }
  },
  {
    "ph": "s", "id": 242320, "pid": 76337, "tid": -914061504, "ts": 1716454225188612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225264600, "dur": 21,
    "args": {
      "External id": 242322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242322,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242322, "pid": 5, "tid": 7, "ts": 1716454225264600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188622, "dur": 5,
    "args": {
      "External id": 242322,
      "cbid": 211, "correlation": 242322
    }
  },
  {
    "ph": "s", "id": 242322, "pid": 76337, "tid": -914061504, "ts": 1716454225188622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225264623, "dur": 33,
    "args": {
      "External id": 242328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242328, "pid": 5, "tid": 7, "ts": 1716454225264623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188650, "dur": 8,
    "args": {
      "External id": 242328,
      "cbid": 211, "correlation": 242328
    }
  },
  {
    "ph": "s", "id": 242328, "pid": 76337, "tid": -914061504, "ts": 1716454225188650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188707, "dur": 0,
    "args": {
      "External id": 242338,
      "cbid": 317, "correlation": 242338
    }
  },
  {
    "ph": "f", "id": 242338, "pid": 76337, "tid": -914061504, "ts": 1716454225188707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188708, "dur": 0,
    "args": {
      "External id": 242339,
      "cbid": 203, "correlation": 242339
    }
  },
  {
    "ph": "f", "id": 242339, "pid": 76337, "tid": -914061504, "ts": 1716454225188708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188709, "dur": 0,
    "args": {
      "External id": 242340,
      "cbid": 205, "correlation": 242340
    }
  },
  {
    "ph": "f", "id": 242340, "pid": 76337, "tid": -914061504, "ts": 1716454225188709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225264657, "dur": 29,
    "args": {
      "External id": 242344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242344, "pid": 5, "tid": 7, "ts": 1716454225264657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188722, "dur": 12,
    "args": {
      "External id": 242344,
      "cbid": 211, "correlation": 242344
    }
  },
  {
    "ph": "s", "id": 242344, "pid": 76337, "tid": -914061504, "ts": 1716454225188722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225264688, "dur": 149,
    "args": {
      "External id": 242346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242346,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242346, "pid": 5, "tid": 7, "ts": 1716454225264688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188740, "dur": 6,
    "args": {
      "External id": 242346,
      "cbid": 211, "correlation": 242346
    }
  },
  {
    "ph": "s", "id": 242346, "pid": 76337, "tid": -914061504, "ts": 1716454225188740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225264838, "dur": 22,
    "args": {
      "External id": 242348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242348, "pid": 5, "tid": 7, "ts": 1716454225264838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188750, "dur": 5,
    "args": {
      "External id": 242348,
      "cbid": 211, "correlation": 242348
    }
  },
  {
    "ph": "s", "id": 242348, "pid": 76337, "tid": -914061504, "ts": 1716454225188750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225264862, "dur": 32,
    "args": {
      "External id": 242354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242354, "pid": 5, "tid": 7, "ts": 1716454225264862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188775, "dur": 9,
    "args": {
      "External id": 242354,
      "cbid": 211, "correlation": 242354
    }
  },
  {
    "ph": "s", "id": 242354, "pid": 76337, "tid": -914061504, "ts": 1716454225188775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225264895, "dur": 27,
    "args": {
      "External id": 242362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242362,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242362, "pid": 5, "tid": 7, "ts": 1716454225264895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188804, "dur": 7,
    "args": {
      "External id": 242362,
      "cbid": 211, "correlation": 242362
    }
  },
  {
    "ph": "s", "id": 242362, "pid": 76337, "tid": -914061504, "ts": 1716454225188804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225264924, "dur": 19,
    "args": {
      "External id": 242370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242370,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242370, "pid": 5, "tid": 7, "ts": 1716454225264924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188833, "dur": 8,
    "args": {
      "External id": 242370,
      "cbid": 211, "correlation": 242370
    }
  },
  {
    "ph": "s", "id": 242370, "pid": 76337, "tid": -914061504, "ts": 1716454225188833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225264944, "dur": 30,
    "args": {
      "External id": 242390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242390,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 242390, "pid": 5, "tid": 7, "ts": 1716454225264944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188914, "dur": 12,
    "args": {
      "External id": 242390,
      "cbid": 211, "correlation": 242390
    }
  },
  {
    "ph": "s", "id": 242390, "pid": 76337, "tid": -914061504, "ts": 1716454225188914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225264976, "dur": 4,
    "args": {
      "External id": 242402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242402,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 242402, "pid": 5, "tid": 7, "ts": 1716454225264976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188938, "dur": 6,
    "args": {
      "External id": 242402,
      "cbid": 211, "correlation": 242402
    }
  },
  {
    "ph": "s", "id": 242402, "pid": 76337, "tid": -914061504, "ts": 1716454225188938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225264981, "dur": 30,
    "args": {
      "External id": 242405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242405,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242405, "pid": 5, "tid": 7, "ts": 1716454225264981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225188955, "dur": 7,
    "args": {
      "External id": 242405,
      "cbid": 211, "correlation": 242405
    }
  },
  {
    "ph": "s", "id": 242405, "pid": 76337, "tid": -914061504, "ts": 1716454225188955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189025, "dur": 0,
    "args": {
      "External id": 242416,
      "cbid": 317, "correlation": 242416
    }
  },
  {
    "ph": "f", "id": 242416, "pid": 76337, "tid": -914061504, "ts": 1716454225189025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189026, "dur": 0,
    "args": {
      "External id": 242417,
      "cbid": 203, "correlation": 242417
    }
  },
  {
    "ph": "f", "id": 242417, "pid": 76337, "tid": -914061504, "ts": 1716454225189026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189026, "dur": 0,
    "args": {
      "External id": 242418,
      "cbid": 205, "correlation": 242418
    }
  },
  {
    "ph": "f", "id": 242418, "pid": 76337, "tid": -914061504, "ts": 1716454225189026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225265013, "dur": 22,
    "args": {
      "External id": 242422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242422, "pid": 5, "tid": 7, "ts": 1716454225265013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189040, "dur": 12,
    "args": {
      "External id": 242422,
      "cbid": 211, "correlation": 242422
    }
  },
  {
    "ph": "s", "id": 242422, "pid": 76337, "tid": -914061504, "ts": 1716454225189040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225265036, "dur": 102,
    "args": {
      "External id": 242424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242424,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242424, "pid": 5, "tid": 7, "ts": 1716454225265036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189059, "dur": 6,
    "args": {
      "External id": 242424,
      "cbid": 211, "correlation": 242424
    }
  },
  {
    "ph": "s", "id": 242424, "pid": 76337, "tid": -914061504, "ts": 1716454225189059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225265140, "dur": 22,
    "args": {
      "External id": 242426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242426,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242426, "pid": 5, "tid": 7, "ts": 1716454225265140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189068, "dur": 5,
    "args": {
      "External id": 242426,
      "cbid": 211, "correlation": 242426
    }
  },
  {
    "ph": "s", "id": 242426, "pid": 76337, "tid": -914061504, "ts": 1716454225189068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225265163, "dur": 32,
    "args": {
      "External id": 242432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242432, "pid": 5, "tid": 7, "ts": 1716454225265163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189096, "dur": 8,
    "args": {
      "External id": 242432,
      "cbid": 211, "correlation": 242432
    }
  },
  {
    "ph": "s", "id": 242432, "pid": 76337, "tid": -914061504, "ts": 1716454225189096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225265197, "dur": 174,
    "args": {
      "External id": 242441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242441,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242441, "pid": 5, "tid": 7, "ts": 1716454225265197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189178, "dur": 15,
    "args": {
      "External id": 242441,
      "cbid": 211, "correlation": 242441
    }
  },
  {
    "ph": "s", "id": 242441, "pid": 76337, "tid": -914061504, "ts": 1716454225189178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225265372, "dur": 63,
    "args": {
      "External id": 242463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242463,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242463, "pid": 5, "tid": 7, "ts": 1716454225265372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189235, "dur": 10,
    "args": {
      "External id": 242463,
      "cbid": 211, "correlation": 242463
    }
  },
  {
    "ph": "s", "id": 242463, "pid": 76337, "tid": -914061504, "ts": 1716454225189235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189320, "dur": 1,
    "args": {
      "External id": 242474,
      "cbid": 251, "correlation": 242474
    }
  },
  {
    "ph": "f", "id": 242474, "pid": 76337, "tid": -914061504, "ts": 1716454225189320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225265437, "dur": 150,
    "args": {
      "External id": 242475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242475,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242475, "pid": 5, "tid": 7, "ts": 1716454225265437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189325, "dur": 14,
    "args": {
      "External id": 242475,
      "cbid": 211, "correlation": 242475
    }
  },
  {
    "ph": "s", "id": 242475, "pid": 76337, "tid": -914061504, "ts": 1716454225189325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189395, "dur": 1,
    "args": {
      "External id": 242486,
      "cbid": 251, "correlation": 242486
    }
  },
  {
    "ph": "f", "id": 242486, "pid": 76337, "tid": -914061504, "ts": 1716454225189395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225265588, "dur": 143,
    "args": {
      "External id": 242487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242487,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242487, "pid": 5, "tid": 7, "ts": 1716454225265588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189399, "dur": 11,
    "args": {
      "External id": 242487,
      "cbid": 211, "correlation": 242487
    }
  },
  {
    "ph": "s", "id": 242487, "pid": 76337, "tid": -914061504, "ts": 1716454225189399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189463, "dur": 1,
    "args": {
      "External id": 242498,
      "cbid": 251, "correlation": 242498
    }
  },
  {
    "ph": "f", "id": 242498, "pid": 76337, "tid": -914061504, "ts": 1716454225189463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225265733, "dur": 144,
    "args": {
      "External id": 242499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242499,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242499, "pid": 5, "tid": 7, "ts": 1716454225265733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189467, "dur": 11,
    "args": {
      "External id": 242499,
      "cbid": 211, "correlation": 242499
    }
  },
  {
    "ph": "s", "id": 242499, "pid": 76337, "tid": -914061504, "ts": 1716454225189467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225265878, "dur": 1905,
    "args": {
      "External id": 242520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242520,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 242520, "pid": 5, "tid": 7, "ts": 1716454225265878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189546, "dur": 13,
    "args": {
      "External id": 242520,
      "cbid": 211, "correlation": 242520
    }
  },
  {
    "ph": "s", "id": 242520, "pid": 76337, "tid": -914061504, "ts": 1716454225189546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189645, "dur": 1,
    "args": {
      "External id": 242538,
      "cbid": 251, "correlation": 242538
    }
  },
  {
    "ph": "f", "id": 242538, "pid": 76337, "tid": -914061504, "ts": 1716454225189645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225267785, "dur": 148,
    "args": {
      "External id": 242540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242540,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 242540, "pid": 5, "tid": 7, "ts": 1716454225267785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189650, "dur": 13,
    "args": {
      "External id": 242540,
      "cbid": 211, "correlation": 242540
    }
  },
  {
    "ph": "s", "id": 242540, "pid": 76337, "tid": -914061504, "ts": 1716454225189650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225267934, "dur": 35,
    "args": {
      "External id": 242548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242548,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242548, "pid": 5, "tid": 7, "ts": 1716454225267934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189719, "dur": 13,
    "args": {
      "External id": 242548,
      "cbid": 211, "correlation": 242548
    }
  },
  {
    "ph": "s", "id": 242548, "pid": 76337, "tid": -914061504, "ts": 1716454225189719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225267971, "dur": 50,
    "args": {
      "External id": 242556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242556,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242556, "pid": 5, "tid": 7, "ts": 1716454225267971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189759, "dur": 8,
    "args": {
      "External id": 242556,
      "cbid": 211, "correlation": 242556
    }
  },
  {
    "ph": "s", "id": 242556, "pid": 76337, "tid": -914061504, "ts": 1716454225189759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225268022, "dur": 31,
    "args": {
      "External id": 242567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242567,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242567, "pid": 5, "tid": 7, "ts": 1716454225268022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189830, "dur": 12,
    "args": {
      "External id": 242567,
      "cbid": 211, "correlation": 242567
    }
  },
  {
    "ph": "s", "id": 242567, "pid": 76337, "tid": -914061504, "ts": 1716454225189830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225268054, "dur": 34,
    "args": {
      "External id": 242589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242589,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242589, "pid": 5, "tid": 7, "ts": 1716454225268054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189860, "dur": 7,
    "args": {
      "External id": 242589,
      "cbid": 211, "correlation": 242589
    }
  },
  {
    "ph": "s", "id": 242589, "pid": 76337, "tid": -914061504, "ts": 1716454225189860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189945, "dur": 1,
    "args": {
      "External id": 242600,
      "cbid": 251, "correlation": 242600
    }
  },
  {
    "ph": "f", "id": 242600, "pid": 76337, "tid": -914061504, "ts": 1716454225189945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225268089, "dur": 88,
    "args": {
      "External id": 242601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242601,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242601, "pid": 5, "tid": 7, "ts": 1716454225268089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225189950, "dur": 12,
    "args": {
      "External id": 242601,
      "cbid": 211, "correlation": 242601
    }
  },
  {
    "ph": "s", "id": 242601, "pid": 76337, "tid": -914061504, "ts": 1716454225189950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190030, "dur": 1,
    "args": {
      "External id": 242612,
      "cbid": 251, "correlation": 242612
    }
  },
  {
    "ph": "f", "id": 242612, "pid": 76337, "tid": -914061504, "ts": 1716454225190030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190033, "dur": 0,
    "args": {
      "External id": 242613,
      "cbid": 251, "correlation": 242613
    }
  },
  {
    "ph": "f", "id": 242613, "pid": 76337, "tid": -914061504, "ts": 1716454225190033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225268179, "dur": 11,
    "args": {
      "External id": 242614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242614,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 242614, "pid": 5, "tid": 7, "ts": 1716454225268179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190035, "dur": 13,
    "args": {
      "External id": 242614,
      "cbid": 211, "correlation": 242614
    }
  },
  {
    "ph": "s", "id": 242614, "pid": 76337, "tid": -914061504, "ts": 1716454225190035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225268191, "dur": 5,
    "args": {
      "External id": 242616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242616,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 242616, "pid": 5, "tid": 7, "ts": 1716454225268191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190049, "dur": 6,
    "args": {
      "External id": 242616,
      "cbid": 211, "correlation": 242616
    }
  },
  {
    "ph": "s", "id": 242616, "pid": 76337, "tid": -914061504, "ts": 1716454225190049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190106, "dur": 1,
    "args": {
      "External id": 242627,
      "cbid": 251, "correlation": 242627
    }
  },
  {
    "ph": "f", "id": 242627, "pid": 76337, "tid": -914061504, "ts": 1716454225190106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190110, "dur": 0,
    "args": {
      "External id": 242628,
      "cbid": 251, "correlation": 242628
    }
  },
  {
    "ph": "f", "id": 242628, "pid": 76337, "tid": -914061504, "ts": 1716454225190110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225268197, "dur": 7,
    "args": {
      "External id": 242629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242629,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 242629, "pid": 5, "tid": 7, "ts": 1716454225268197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190111, "dur": 11,
    "args": {
      "External id": 242629,
      "cbid": 211, "correlation": 242629
    }
  },
  {
    "ph": "s", "id": 242629, "pid": 76337, "tid": -914061504, "ts": 1716454225190111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225268206, "dur": 4,
    "args": {
      "External id": 242631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242631,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 242631, "pid": 5, "tid": 7, "ts": 1716454225268206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190124, "dur": 6,
    "args": {
      "External id": 242631,
      "cbid": 211, "correlation": 242631
    }
  },
  {
    "ph": "s", "id": 242631, "pid": 76337, "tid": -914061504, "ts": 1716454225190124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225268211, "dur": 90,
    "args": {
      "External id": 242652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242652,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 242652, "pid": 5, "tid": 7, "ts": 1716454225268211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190198, "dur": 12,
    "args": {
      "External id": 242652,
      "cbid": 211, "correlation": 242652
    }
  },
  {
    "ph": "s", "id": 242652, "pid": 76337, "tid": -914061504, "ts": 1716454225190198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190296, "dur": 1,
    "args": {
      "External id": 242670,
      "cbid": 251, "correlation": 242670
    }
  },
  {
    "ph": "f", "id": 242670, "pid": 76337, "tid": -914061504, "ts": 1716454225190296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225268302, "dur": 96,
    "args": {
      "External id": 242672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242672,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242672, "pid": 5, "tid": 7, "ts": 1716454225268302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190302, "dur": 13,
    "args": {
      "External id": 242672,
      "cbid": 211, "correlation": 242672
    }
  },
  {
    "ph": "s", "id": 242672, "pid": 76337, "tid": -914061504, "ts": 1716454225190302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225268400, "dur": 20,
    "args": {
      "External id": 242680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242680,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242680, "pid": 5, "tid": 7, "ts": 1716454225268400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190370, "dur": 13,
    "args": {
      "External id": 242680,
      "cbid": 211, "correlation": 242680
    }
  },
  {
    "ph": "s", "id": 242680, "pid": 76337, "tid": -914061504, "ts": 1716454225190370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225268421, "dur": 37,
    "args": {
      "External id": 242688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242688,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242688, "pid": 5, "tid": 7, "ts": 1716454225268421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190411, "dur": 9,
    "args": {
      "External id": 242688,
      "cbid": 211, "correlation": 242688
    }
  },
  {
    "ph": "s", "id": 242688, "pid": 76337, "tid": -914061504, "ts": 1716454225190411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225268459, "dur": 34,
    "args": {
      "External id": 242710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242710,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242710, "pid": 5, "tid": 7, "ts": 1716454225268459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190462, "dur": 11,
    "args": {
      "External id": 242710,
      "cbid": 211, "correlation": 242710
    }
  },
  {
    "ph": "s", "id": 242710, "pid": 76337, "tid": -914061504, "ts": 1716454225190462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190551, "dur": 1,
    "args": {
      "External id": 242726,
      "cbid": 251, "correlation": 242726
    }
  },
  {
    "ph": "f", "id": 242726, "pid": 76337, "tid": -914061504, "ts": 1716454225190551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190556, "dur": 0,
    "args": {
      "External id": 242728,
      "cbid": 251, "correlation": 242728
    }
  },
  {
    "ph": "f", "id": 242728, "pid": 76337, "tid": -914061504, "ts": 1716454225190556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225268494, "dur": 529,
    "args": {
      "External id": 242729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242729,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 242729, "pid": 5, "tid": 7, "ts": 1716454225268494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190559, "dur": 13,
    "args": {
      "External id": 242729,
      "cbid": 211, "correlation": 242729
    }
  },
  {
    "ph": "s", "id": 242729, "pid": 76337, "tid": -914061504, "ts": 1716454225190559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225269024, "dur": 124,
    "args": {
      "External id": 242737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242737, "pid": 5, "tid": 7, "ts": 1716454225269024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190625, "dur": 12,
    "args": {
      "External id": 242737,
      "cbid": 211, "correlation": 242737
    }
  },
  {
    "ph": "s", "id": 242737, "pid": 76337, "tid": -914061504, "ts": 1716454225190625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225269149, "dur": 128,
    "args": {
      "External id": 242745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242745, "pid": 5, "tid": 7, "ts": 1716454225269149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190654, "dur": 8,
    "args": {
      "External id": 242745,
      "cbid": 211, "correlation": 242745
    }
  },
  {
    "ph": "s", "id": 242745, "pid": 76337, "tid": -914061504, "ts": 1716454225190654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190730, "dur": 1,
    "args": {
      "External id": 242761,
      "cbid": 251, "correlation": 242761
    }
  },
  {
    "ph": "f", "id": 242761, "pid": 76337, "tid": -914061504, "ts": 1716454225190730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225269279, "dur": 303,
    "args": {
      "External id": 242763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242763,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242763, "pid": 5, "tid": 7, "ts": 1716454225269279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190736, "dur": 12,
    "args": {
      "External id": 242763,
      "cbid": 211, "correlation": 242763
    }
  },
  {
    "ph": "s", "id": 242763, "pid": 76337, "tid": -914061504, "ts": 1716454225190736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225269584, "dur": 27,
    "args": {
      "External id": 242771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242771,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242771, "pid": 5, "tid": 7, "ts": 1716454225269584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190778, "dur": 10,
    "args": {
      "External id": 242771,
      "cbid": 211, "correlation": 242771
    }
  },
  {
    "ph": "s", "id": 242771, "pid": 76337, "tid": -914061504, "ts": 1716454225190778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225269612, "dur": 80,
    "args": {
      "External id": 242782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242782, "pid": 5, "tid": 7, "ts": 1716454225269612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190845, "dur": 12,
    "args": {
      "External id": 242782,
      "cbid": 211, "correlation": 242782
    }
  },
  {
    "ph": "s", "id": 242782, "pid": 76337, "tid": -914061504, "ts": 1716454225190845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190908, "dur": 0,
    "args": {
      "External id": 242794,
      "cbid": 317, "correlation": 242794
    }
  },
  {
    "ph": "f", "id": 242794, "pid": 76337, "tid": -914061504, "ts": 1716454225190908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190909, "dur": 0,
    "args": {
      "External id": 242795,
      "cbid": 203, "correlation": 242795
    }
  },
  {
    "ph": "f", "id": 242795, "pid": 76337, "tid": -914061504, "ts": 1716454225190909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190910, "dur": 0,
    "args": {
      "External id": 242796,
      "cbid": 205, "correlation": 242796
    }
  },
  {
    "ph": "f", "id": 242796, "pid": 76337, "tid": -914061504, "ts": 1716454225190910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225269693, "dur": 23,
    "args": {
      "External id": 242800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242800, "pid": 5, "tid": 7, "ts": 1716454225269693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190925, "dur": 12,
    "args": {
      "External id": 242800,
      "cbid": 211, "correlation": 242800
    }
  },
  {
    "ph": "s", "id": 242800, "pid": 76337, "tid": -914061504, "ts": 1716454225190925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225269717, "dur": 118,
    "args": {
      "External id": 242802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242802,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242802, "pid": 5, "tid": 7, "ts": 1716454225269717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190943, "dur": 6,
    "args": {
      "External id": 242802,
      "cbid": 211, "correlation": 242802
    }
  },
  {
    "ph": "s", "id": 242802, "pid": 76337, "tid": -914061504, "ts": 1716454225190943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225269837, "dur": 23,
    "args": {
      "External id": 242804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242804,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242804, "pid": 5, "tid": 7, "ts": 1716454225269837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190953, "dur": 5,
    "args": {
      "External id": 242804,
      "cbid": 211, "correlation": 242804
    }
  },
  {
    "ph": "s", "id": 242804, "pid": 76337, "tid": -914061504, "ts": 1716454225190953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225269861, "dur": 33,
    "args": {
      "External id": 242810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242810, "pid": 5, "tid": 7, "ts": 1716454225269861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225190989, "dur": 9,
    "args": {
      "External id": 242810,
      "cbid": 211, "correlation": 242810
    }
  },
  {
    "ph": "s", "id": 242810, "pid": 76337, "tid": -914061504, "ts": 1716454225190989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225269895, "dur": 27,
    "args": {
      "External id": 242818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242818,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242818, "pid": 5, "tid": 7, "ts": 1716454225269895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191022, "dur": 8,
    "args": {
      "External id": 242818,
      "cbid": 211, "correlation": 242818
    }
  },
  {
    "ph": "s", "id": 242818, "pid": 76337, "tid": -914061504, "ts": 1716454225191022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225269923, "dur": 100,
    "args": {
      "External id": 242829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242829,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242829, "pid": 5, "tid": 7, "ts": 1716454225269923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191083, "dur": 11,
    "args": {
      "External id": 242829,
      "cbid": 211, "correlation": 242829
    }
  },
  {
    "ph": "s", "id": 242829, "pid": 76337, "tid": -914061504, "ts": 1716454225191083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191137, "dur": 0,
    "args": {
      "External id": 242839,
      "cbid": 317, "correlation": 242839
    }
  },
  {
    "ph": "f", "id": 242839, "pid": 76337, "tid": -914061504, "ts": 1716454225191137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191138, "dur": 0,
    "args": {
      "External id": 242840,
      "cbid": 203, "correlation": 242840
    }
  },
  {
    "ph": "f", "id": 242840, "pid": 76337, "tid": -914061504, "ts": 1716454225191138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191138, "dur": 0,
    "args": {
      "External id": 242841,
      "cbid": 205, "correlation": 242841
    }
  },
  {
    "ph": "f", "id": 242841, "pid": 76337, "tid": -914061504, "ts": 1716454225191138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225270024, "dur": 75,
    "args": {
      "External id": 242845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242845,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242845, "pid": 5, "tid": 7, "ts": 1716454225270024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191153, "dur": 12,
    "args": {
      "External id": 242845,
      "cbid": 211, "correlation": 242845
    }
  },
  {
    "ph": "s", "id": 242845, "pid": 76337, "tid": -914061504, "ts": 1716454225191153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225270101, "dur": 43,
    "args": {
      "External id": 242847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242847,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242847, "pid": 5, "tid": 7, "ts": 1716454225270101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191167, "dur": 6,
    "args": {
      "External id": 242847,
      "cbid": 211, "correlation": 242847
    }
  },
  {
    "ph": "s", "id": 242847, "pid": 76337, "tid": -914061504, "ts": 1716454225191167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225270145, "dur": 4,
    "args": {
      "External id": 242849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 242849, "pid": 5, "tid": 7, "ts": 1716454225270145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191178, "dur": 6,
    "args": {
      "External id": 242849,
      "cbid": 211, "correlation": 242849
    }
  },
  {
    "ph": "s", "id": 242849, "pid": 76337, "tid": -914061504, "ts": 1716454225191178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191187, "dur": 0,
    "args": {
      "External id": 242850,
      "cbid": 51, "correlation": 242850
    }
  },
  {
    "ph": "s", "id": 242850, "pid": 76337, "tid": -914061504, "ts": 1716454225191187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225270150, "dur": 2225,
    "args": {
      "External id": 242851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242851,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242851, "pid": 5, "tid": 7, "ts": 1716454225270150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191188, "dur": 5,
    "args": {
      "External id": 242851,
      "cbid": 211, "correlation": 242851
    }
  },
  {
    "ph": "s", "id": 242851, "pid": 76337, "tid": -914061504, "ts": 1716454225191188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225272376, "dur": 111,
    "args": {
      "External id": 242856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242856, "pid": 5, "tid": 7, "ts": 1716454225272376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191215, "dur": 9,
    "args": {
      "External id": 242856,
      "cbid": 211, "correlation": 242856
    }
  },
  {
    "ph": "s", "id": 242856, "pid": 76337, "tid": -914061504, "ts": 1716454225191215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225272489, "dur": 162,
    "args": {
      "External id": 242865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242865,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242865, "pid": 5, "tid": 7, "ts": 1716454225272489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191305, "dur": 14,
    "args": {
      "External id": 242865,
      "cbid": 211, "correlation": 242865
    }
  },
  {
    "ph": "s", "id": 242865, "pid": 76337, "tid": -914061504, "ts": 1716454225191305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225272652, "dur": 125,
    "args": {
      "External id": 242885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242885,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 242885, "pid": 5, "tid": 7, "ts": 1716454225272652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191375, "dur": 11,
    "args": {
      "External id": 242885,
      "cbid": 211, "correlation": 242885
    }
  },
  {
    "ph": "s", "id": 242885, "pid": 76337, "tid": -914061504, "ts": 1716454225191375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225272778, "dur": 5,
    "args": {
      "External id": 242897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242897,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 242897, "pid": 5, "tid": 7, "ts": 1716454225272778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191396, "dur": 6,
    "args": {
      "External id": 242897,
      "cbid": 211, "correlation": 242897
    }
  },
  {
    "ph": "s", "id": 242897, "pid": 76337, "tid": -914061504, "ts": 1716454225191396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225272784, "dur": 156,
    "args": {
      "External id": 242900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242900, "pid": 5, "tid": 7, "ts": 1716454225272784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191415, "dur": 8,
    "args": {
      "External id": 242900,
      "cbid": 211, "correlation": 242900
    }
  },
  {
    "ph": "s", "id": 242900, "pid": 76337, "tid": -914061504, "ts": 1716454225191415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225272942, "dur": 101,
    "args": {
      "External id": 242909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242909,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242909, "pid": 5, "tid": 7, "ts": 1716454225272942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191456, "dur": 11,
    "args": {
      "External id": 242909,
      "cbid": 211, "correlation": 242909
    }
  },
  {
    "ph": "s", "id": 242909, "pid": 76337, "tid": -914061504, "ts": 1716454225191456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191508, "dur": 0,
    "args": {
      "External id": 242919,
      "cbid": 317, "correlation": 242919
    }
  },
  {
    "ph": "f", "id": 242919, "pid": 76337, "tid": -914061504, "ts": 1716454225191508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191509, "dur": 0,
    "args": {
      "External id": 242920,
      "cbid": 203, "correlation": 242920
    }
  },
  {
    "ph": "f", "id": 242920, "pid": 76337, "tid": -914061504, "ts": 1716454225191509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191510, "dur": 0,
    "args": {
      "External id": 242921,
      "cbid": 205, "correlation": 242921
    }
  },
  {
    "ph": "f", "id": 242921, "pid": 76337, "tid": -914061504, "ts": 1716454225191510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225273044, "dur": 112,
    "args": {
      "External id": 242925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242925,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242925, "pid": 5, "tid": 7, "ts": 1716454225273044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191526, "dur": 11,
    "args": {
      "External id": 242925,
      "cbid": 211, "correlation": 242925
    }
  },
  {
    "ph": "s", "id": 242925, "pid": 76337, "tid": -914061504, "ts": 1716454225191526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225273157, "dur": 33,
    "args": {
      "External id": 242927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242927,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242927, "pid": 5, "tid": 7, "ts": 1716454225273157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191540, "dur": 5,
    "args": {
      "External id": 242927,
      "cbid": 211, "correlation": 242927
    }
  },
  {
    "ph": "s", "id": 242927, "pid": 76337, "tid": -914061504, "ts": 1716454225191540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225273191, "dur": 3,
    "args": {
      "External id": 242929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242929,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 242929, "pid": 5, "tid": 7, "ts": 1716454225273191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191549, "dur": 5,
    "args": {
      "External id": 242929,
      "cbid": 211, "correlation": 242929
    }
  },
  {
    "ph": "s", "id": 242929, "pid": 76337, "tid": -914061504, "ts": 1716454225191549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191558, "dur": 0,
    "args": {
      "External id": 242930,
      "cbid": 51, "correlation": 242930
    }
  },
  {
    "ph": "s", "id": 242930, "pid": 76337, "tid": -914061504, "ts": 1716454225191558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225273196, "dur": 1987,
    "args": {
      "External id": 242931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242931,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 242931, "pid": 5, "tid": 7, "ts": 1716454225273196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191559, "dur": 6,
    "args": {
      "External id": 242931,
      "cbid": 211, "correlation": 242931
    }
  },
  {
    "ph": "s", "id": 242931, "pid": 76337, "tid": -914061504, "ts": 1716454225191559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225275184, "dur": 58,
    "args": {
      "External id": 242936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242936, "pid": 5, "tid": 7, "ts": 1716454225275184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191588, "dur": 9,
    "args": {
      "External id": 242936,
      "cbid": 211, "correlation": 242936
    }
  },
  {
    "ph": "s", "id": 242936, "pid": 76337, "tid": -914061504, "ts": 1716454225191588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225275244, "dur": 3,
    "args": {
      "External id": 242944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242944,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 242944, "pid": 5, "tid": 7, "ts": 1716454225275244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191631, "dur": 9,
    "args": {
      "External id": 242944,
      "cbid": 211, "correlation": 242944
    }
  },
  {
    "ph": "s", "id": 242944, "pid": 76337, "tid": -914061504, "ts": 1716454225191631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191696, "dur": 1,
    "args": {
      "External id": 242960,
      "cbid": 251, "correlation": 242960
    }
  },
  {
    "ph": "f", "id": 242960, "pid": 76337, "tid": -914061504, "ts": 1716454225191696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191701, "dur": 0,
    "args": {
      "External id": 242962,
      "cbid": 251, "correlation": 242962
    }
  },
  {
    "ph": "f", "id": 242962, "pid": 76337, "tid": -914061504, "ts": 1716454225191701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225275248, "dur": 11,
    "args": {
      "External id": 242963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242963,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 242963, "pid": 5, "tid": 7, "ts": 1716454225275248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191703, "dur": 12,
    "args": {
      "External id": 242963,
      "cbid": 211, "correlation": 242963
    }
  },
  {
    "ph": "s", "id": 242963, "pid": 76337, "tid": -914061504, "ts": 1716454225191703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225275260, "dur": 5,
    "args": {
      "External id": 242965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242965,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 242965, "pid": 5, "tid": 7, "ts": 1716454225275260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191717, "dur": 5,
    "args": {
      "External id": 242965,
      "cbid": 211, "correlation": 242965
    }
  },
  {
    "ph": "s", "id": 242965, "pid": 76337, "tid": -914061504, "ts": 1716454225191717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225275267, "dur": 52,
    "args": {
      "External id": 242975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242975,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 242975, "pid": 5, "tid": 7, "ts": 1716454225275267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191775, "dur": 12,
    "args": {
      "External id": 242975,
      "cbid": 211, "correlation": 242975
    }
  },
  {
    "ph": "s", "id": 242975, "pid": 76337, "tid": -914061504, "ts": 1716454225191775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225275321, "dur": 51,
    "args": {
      "External id": 242995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 242995,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 242995, "pid": 5, "tid": 7, "ts": 1716454225275321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191841, "dur": 10,
    "args": {
      "External id": 242995,
      "cbid": 211, "correlation": 242995
    }
  },
  {
    "ph": "s", "id": 242995, "pid": 76337, "tid": -914061504, "ts": 1716454225191841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225275373, "dur": 4,
    "args": {
      "External id": 243007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243007,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243007, "pid": 5, "tid": 7, "ts": 1716454225275373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191861, "dur": 7,
    "args": {
      "External id": 243007,
      "cbid": 211, "correlation": 243007
    }
  },
  {
    "ph": "s", "id": 243007, "pid": 76337, "tid": -914061504, "ts": 1716454225191861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225275378, "dur": 55,
    "args": {
      "External id": 243010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243010,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243010, "pid": 5, "tid": 7, "ts": 1716454225275378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191879, "dur": 6,
    "args": {
      "External id": 243010,
      "cbid": 211, "correlation": 243010
    }
  },
  {
    "ph": "s", "id": 243010, "pid": 76337, "tid": -914061504, "ts": 1716454225191879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225275435, "dur": 37,
    "args": {
      "External id": 243019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243019,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243019, "pid": 5, "tid": 7, "ts": 1716454225275435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191920, "dur": 10,
    "args": {
      "External id": 243019,
      "cbid": 211, "correlation": 243019
    }
  },
  {
    "ph": "s", "id": 243019, "pid": 76337, "tid": -914061504, "ts": 1716454225191920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191991, "dur": 0,
    "args": {
      "External id": 243029,
      "cbid": 317, "correlation": 243029
    }
  },
  {
    "ph": "f", "id": 243029, "pid": 76337, "tid": -914061504, "ts": 1716454225191991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191992, "dur": 0,
    "args": {
      "External id": 243030,
      "cbid": 203, "correlation": 243030
    }
  },
  {
    "ph": "f", "id": 243030, "pid": 76337, "tid": -914061504, "ts": 1716454225191992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225191993, "dur": 0,
    "args": {
      "External id": 243031,
      "cbid": 205, "correlation": 243031
    }
  },
  {
    "ph": "f", "id": 243031, "pid": 76337, "tid": -914061504, "ts": 1716454225191993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225275472, "dur": 40,
    "args": {
      "External id": 243035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243035,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243035, "pid": 5, "tid": 7, "ts": 1716454225275472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192008, "dur": 13,
    "args": {
      "External id": 243035,
      "cbid": 211, "correlation": 243035
    }
  },
  {
    "ph": "s", "id": 243035, "pid": 76337, "tid": -914061504, "ts": 1716454225192008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225275514, "dur": 14,
    "args": {
      "External id": 243037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243037,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243037, "pid": 5, "tid": 7, "ts": 1716454225275514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192024, "dur": 5,
    "args": {
      "External id": 243037,
      "cbid": 211, "correlation": 243037
    }
  },
  {
    "ph": "s", "id": 243037, "pid": 76337, "tid": -914061504, "ts": 1716454225192024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225275529, "dur": 3,
    "args": {
      "External id": 243039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243039, "pid": 5, "tid": 7, "ts": 1716454225275529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192034, "dur": 5,
    "args": {
      "External id": 243039,
      "cbid": 211, "correlation": 243039
    }
  },
  {
    "ph": "s", "id": 243039, "pid": 76337, "tid": -914061504, "ts": 1716454225192034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192042, "dur": 0,
    "args": {
      "External id": 243040,
      "cbid": 51, "correlation": 243040
    }
  },
  {
    "ph": "s", "id": 243040, "pid": 76337, "tid": -914061504, "ts": 1716454225192042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225275533, "dur": 688,
    "args": {
      "External id": 243041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243041,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243041, "pid": 5, "tid": 7, "ts": 1716454225275533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192043, "dur": 5,
    "args": {
      "External id": 243041,
      "cbid": 211, "correlation": 243041
    }
  },
  {
    "ph": "s", "id": 243041, "pid": 76337, "tid": -914061504, "ts": 1716454225192043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225276222, "dur": 58,
    "args": {
      "External id": 243046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243046,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243046, "pid": 5, "tid": 7, "ts": 1716454225276222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192071, "dur": 8,
    "args": {
      "External id": 243046,
      "cbid": 211, "correlation": 243046
    }
  },
  {
    "ph": "s", "id": 243046, "pid": 76337, "tid": -914061504, "ts": 1716454225192071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192128, "dur": 0,
    "args": {
      "External id": 243056,
      "cbid": 317, "correlation": 243056
    }
  },
  {
    "ph": "f", "id": 243056, "pid": 76337, "tid": -914061504, "ts": 1716454225192128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192128, "dur": 0,
    "args": {
      "External id": 243057,
      "cbid": 203, "correlation": 243057
    }
  },
  {
    "ph": "f", "id": 243057, "pid": 76337, "tid": -914061504, "ts": 1716454225192128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192129, "dur": 0,
    "args": {
      "External id": 243058,
      "cbid": 205, "correlation": 243058
    }
  },
  {
    "ph": "f", "id": 243058, "pid": 76337, "tid": -914061504, "ts": 1716454225192129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225276282, "dur": 3,
    "args": {
      "External id": 243062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243062, "pid": 5, "tid": 7, "ts": 1716454225276282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192144, "dur": 11,
    "args": {
      "External id": 243062,
      "cbid": 211, "correlation": 243062
    }
  },
  {
    "ph": "s", "id": 243062, "pid": 76337, "tid": -914061504, "ts": 1716454225192144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192160, "dur": 0,
    "args": {
      "External id": 243063,
      "cbid": 51, "correlation": 243063
    }
  },
  {
    "ph": "s", "id": 243063, "pid": 76337, "tid": -914061504, "ts": 1716454225192160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225276286, "dur": 261,
    "args": {
      "External id": 243064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243064,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243064, "pid": 5, "tid": 7, "ts": 1716454225276286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192161, "dur": 7,
    "args": {
      "External id": 243064,
      "cbid": 211, "correlation": 243064
    }
  },
  {
    "ph": "s", "id": 243064, "pid": 76337, "tid": -914061504, "ts": 1716454225192161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225276548, "dur": 59,
    "args": {
      "External id": 243069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243069, "pid": 5, "tid": 7, "ts": 1716454225276548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192188, "dur": 8,
    "args": {
      "External id": 243069,
      "cbid": 211, "correlation": 243069
    }
  },
  {
    "ph": "s", "id": 243069, "pid": 76337, "tid": -914061504, "ts": 1716454225192188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225276609, "dur": 51,
    "args": {
      "External id": 243077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243077,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243077, "pid": 5, "tid": 7, "ts": 1716454225276609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192217, "dur": 8,
    "args": {
      "External id": 243077,
      "cbid": 211, "correlation": 243077
    }
  },
  {
    "ph": "s", "id": 243077, "pid": 76337, "tid": -914061504, "ts": 1716454225192217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225276661, "dur": 35,
    "args": {
      "External id": 243085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243085,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243085, "pid": 5, "tid": 7, "ts": 1716454225276661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192245, "dur": 8,
    "args": {
      "External id": 243085,
      "cbid": 211, "correlation": 243085
    }
  },
  {
    "ph": "s", "id": 243085, "pid": 76337, "tid": -914061504, "ts": 1716454225192245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225276697, "dur": 52,
    "args": {
      "External id": 243105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243105,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 243105, "pid": 5, "tid": 7, "ts": 1716454225276697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192327, "dur": 12,
    "args": {
      "External id": 243105,
      "cbid": 211, "correlation": 243105
    }
  },
  {
    "ph": "s", "id": 243105, "pid": 76337, "tid": -914061504, "ts": 1716454225192327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225276751, "dur": 4,
    "args": {
      "External id": 243117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243117,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243117, "pid": 5, "tid": 7, "ts": 1716454225276751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192349, "dur": 6,
    "args": {
      "External id": 243117,
      "cbid": 211, "correlation": 243117
    }
  },
  {
    "ph": "s", "id": 243117, "pid": 76337, "tid": -914061504, "ts": 1716454225192349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225276756, "dur": 53,
    "args": {
      "External id": 243120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243120, "pid": 5, "tid": 7, "ts": 1716454225276756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192368, "dur": 7,
    "args": {
      "External id": 243120,
      "cbid": 211, "correlation": 243120
    }
  },
  {
    "ph": "s", "id": 243120, "pid": 76337, "tid": -914061504, "ts": 1716454225192368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192425, "dur": 0,
    "args": {
      "External id": 243131,
      "cbid": 317, "correlation": 243131
    }
  },
  {
    "ph": "f", "id": 243131, "pid": 76337, "tid": -914061504, "ts": 1716454225192425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192426, "dur": 0,
    "args": {
      "External id": 243132,
      "cbid": 203, "correlation": 243132
    }
  },
  {
    "ph": "f", "id": 243132, "pid": 76337, "tid": -914061504, "ts": 1716454225192426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192426, "dur": 0,
    "args": {
      "External id": 243133,
      "cbid": 205, "correlation": 243133
    }
  },
  {
    "ph": "f", "id": 243133, "pid": 76337, "tid": -914061504, "ts": 1716454225192426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192455, "dur": 2,
    "args": {
      "External id": 243137,
      "cbid": 251, "correlation": 243137
    }
  },
  {
    "ph": "f", "id": 243137, "pid": 76337, "tid": -914061504, "ts": 1716454225192455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192458, "dur": 1,
    "args": {
      "External id": 243138,
      "cbid": 251, "correlation": 243138
    }
  },
  {
    "ph": "f", "id": 243138, "pid": 76337, "tid": -914061504, "ts": 1716454225192458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192460, "dur": 0,
    "args": {
      "External id": 243139,
      "cbid": 251, "correlation": 243139
    }
  },
  {
    "ph": "f", "id": 243139, "pid": 76337, "tid": -914061504, "ts": 1716454225192460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192461, "dur": 1,
    "args": {
      "External id": 243140,
      "cbid": 251, "correlation": 243140
    }
  },
  {
    "ph": "f", "id": 243140, "pid": 76337, "tid": -914061504, "ts": 1716454225192461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192463, "dur": 1,
    "args": {
      "External id": 243141,
      "cbid": 251, "correlation": 243141
    }
  },
  {
    "ph": "f", "id": 243141, "pid": 76337, "tid": -914061504, "ts": 1716454225192463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192464, "dur": 1,
    "args": {
      "External id": 243142,
      "cbid": 251, "correlation": 243142
    }
  },
  {
    "ph": "f", "id": 243142, "pid": 76337, "tid": -914061504, "ts": 1716454225192464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192466, "dur": 1,
    "args": {
      "External id": 243143,
      "cbid": 251, "correlation": 243143
    }
  },
  {
    "ph": "f", "id": 243143, "pid": 76337, "tid": -914061504, "ts": 1716454225192466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192467, "dur": 1,
    "args": {
      "External id": 243144,
      "cbid": 251, "correlation": 243144
    }
  },
  {
    "ph": "f", "id": 243144, "pid": 76337, "tid": -914061504, "ts": 1716454225192467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192470, "dur": 0,
    "args": {
      "External id": 243145,
      "cbid": 251, "correlation": 243145
    }
  },
  {
    "ph": "f", "id": 243145, "pid": 76337, "tid": -914061504, "ts": 1716454225192470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225276810, "dur": 112,
    "args": {
      "External id": 243146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243146,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243146, "pid": 5, "tid": 7, "ts": 1716454225276810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192472, "dur": 12,
    "args": {
      "External id": 243146,
      "cbid": 211, "correlation": 243146
    }
  },
  {
    "ph": "s", "id": 243146, "pid": 76337, "tid": -914061504, "ts": 1716454225192472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225276924, "dur": 59,
    "args": {
      "External id": 243152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243152,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243152, "pid": 5, "tid": 7, "ts": 1716454225276924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192507, "dur": 9,
    "args": {
      "External id": 243152,
      "cbid": 211, "correlation": 243152
    }
  },
  {
    "ph": "s", "id": 243152, "pid": 76337, "tid": -914061504, "ts": 1716454225192507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225276984, "dur": 587,
    "args": {
      "External id": 243161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243161,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243161, "pid": 5, "tid": 7, "ts": 1716454225276984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192589, "dur": 14,
    "args": {
      "External id": 243161,
      "cbid": 211, "correlation": 243161
    }
  },
  {
    "ph": "s", "id": 243161, "pid": 76337, "tid": -914061504, "ts": 1716454225192589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225277572, "dur": 176,
    "args": {
      "External id": 243183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243183,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243183, "pid": 5, "tid": 7, "ts": 1716454225277572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192645, "dur": 11,
    "args": {
      "External id": 243183,
      "cbid": 211, "correlation": 243183
    }
  },
  {
    "ph": "s", "id": 243183, "pid": 76337, "tid": -914061504, "ts": 1716454225192645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192733, "dur": 1,
    "args": {
      "External id": 243194,
      "cbid": 251, "correlation": 243194
    }
  },
  {
    "ph": "f", "id": 243194, "pid": 76337, "tid": -914061504, "ts": 1716454225192733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225277750, "dur": 192,
    "args": {
      "External id": 243195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243195,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243195, "pid": 5, "tid": 7, "ts": 1716454225277750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192738, "dur": 13,
    "args": {
      "External id": 243195,
      "cbid": 211, "correlation": 243195
    }
  },
  {
    "ph": "s", "id": 243195, "pid": 76337, "tid": -914061504, "ts": 1716454225192738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192807, "dur": 1,
    "args": {
      "External id": 243206,
      "cbid": 251, "correlation": 243206
    }
  },
  {
    "ph": "f", "id": 243206, "pid": 76337, "tid": -914061504, "ts": 1716454225192807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225277943, "dur": 186,
    "args": {
      "External id": 243207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243207,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243207, "pid": 5, "tid": 7, "ts": 1716454225277943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192811, "dur": 11,
    "args": {
      "External id": 243207,
      "cbid": 211, "correlation": 243207
    }
  },
  {
    "ph": "s", "id": 243207, "pid": 76337, "tid": -914061504, "ts": 1716454225192811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192874, "dur": 1,
    "args": {
      "External id": 243218,
      "cbid": 251, "correlation": 243218
    }
  },
  {
    "ph": "f", "id": 243218, "pid": 76337, "tid": -914061504, "ts": 1716454225192874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225278130, "dur": 185,
    "args": {
      "External id": 243219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243219,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243219, "pid": 5, "tid": 7, "ts": 1716454225278130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192878, "dur": 12,
    "args": {
      "External id": 243219,
      "cbid": 211, "correlation": 243219
    }
  },
  {
    "ph": "s", "id": 243219, "pid": 76337, "tid": -914061504, "ts": 1716454225192878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225278316, "dur": 18119,
    "args": {
      "External id": 243240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243240,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243240, "pid": 5, "tid": 7, "ts": 1716454225278316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225192962, "dur": 23,
    "args": {
      "External id": 243240,
      "cbid": 211, "correlation": 243240
    }
  },
  {
    "ph": "s", "id": 243240, "pid": 76337, "tid": -914061504, "ts": 1716454225192962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193072, "dur": 1,
    "args": {
      "External id": 243258,
      "cbid": 251, "correlation": 243258
    }
  },
  {
    "ph": "f", "id": 243258, "pid": 76337, "tid": -914061504, "ts": 1716454225193072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225296436, "dur": 199,
    "args": {
      "External id": 243260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243260,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243260, "pid": 5, "tid": 7, "ts": 1716454225296436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193077, "dur": 13,
    "args": {
      "External id": 243260,
      "cbid": 211, "correlation": 243260
    }
  },
  {
    "ph": "s", "id": 243260, "pid": 76337, "tid": -914061504, "ts": 1716454225193077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225296636, "dur": 67,
    "args": {
      "External id": 243268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243268,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243268, "pid": 5, "tid": 7, "ts": 1716454225296636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193147, "dur": 13,
    "args": {
      "External id": 243268,
      "cbid": 211, "correlation": 243268
    }
  },
  {
    "ph": "s", "id": 243268, "pid": 76337, "tid": -914061504, "ts": 1716454225193147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225296705, "dur": 98,
    "args": {
      "External id": 243276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243276,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243276, "pid": 5, "tid": 7, "ts": 1716454225296705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193187, "dur": 8,
    "args": {
      "External id": 243276,
      "cbid": 211, "correlation": 243276
    }
  },
  {
    "ph": "s", "id": 243276, "pid": 76337, "tid": -914061504, "ts": 1716454225193187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225296804, "dur": 54,
    "args": {
      "External id": 243287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243287, "pid": 5, "tid": 7, "ts": 1716454225296804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193259, "dur": 12,
    "args": {
      "External id": 243287,
      "cbid": 211, "correlation": 243287
    }
  },
  {
    "ph": "s", "id": 243287, "pid": 76337, "tid": -914061504, "ts": 1716454225193259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225296859, "dur": 90,
    "args": {
      "External id": 243309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243309,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243309, "pid": 5, "tid": 7, "ts": 1716454225296859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193290, "dur": 8,
    "args": {
      "External id": 243309,
      "cbid": 211, "correlation": 243309
    }
  },
  {
    "ph": "s", "id": 243309, "pid": 76337, "tid": -914061504, "ts": 1716454225193290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193371, "dur": 1,
    "args": {
      "External id": 243320,
      "cbid": 251, "correlation": 243320
    }
  },
  {
    "ph": "f", "id": 243320, "pid": 76337, "tid": -914061504, "ts": 1716454225193371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225296950, "dur": 101,
    "args": {
      "External id": 243321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243321,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243321, "pid": 5, "tid": 7, "ts": 1716454225296950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193376, "dur": 13,
    "args": {
      "External id": 243321,
      "cbid": 211, "correlation": 243321
    }
  },
  {
    "ph": "s", "id": 243321, "pid": 76337, "tid": -914061504, "ts": 1716454225193376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193450, "dur": 1,
    "args": {
      "External id": 243332,
      "cbid": 251, "correlation": 243332
    }
  },
  {
    "ph": "f", "id": 243332, "pid": 76337, "tid": -914061504, "ts": 1716454225193450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193454, "dur": 0,
    "args": {
      "External id": 243333,
      "cbid": 251, "correlation": 243333
    }
  },
  {
    "ph": "f", "id": 243333, "pid": 76337, "tid": -914061504, "ts": 1716454225193454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225297053, "dur": 10,
    "args": {
      "External id": 243334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243334,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 243334, "pid": 5, "tid": 7, "ts": 1716454225297053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193456, "dur": 13,
    "args": {
      "External id": 243334,
      "cbid": 211, "correlation": 243334
    }
  },
  {
    "ph": "s", "id": 243334, "pid": 76337, "tid": -914061504, "ts": 1716454225193456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225297064, "dur": 5,
    "args": {
      "External id": 243336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243336,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 243336, "pid": 5, "tid": 7, "ts": 1716454225297064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193472, "dur": 7,
    "args": {
      "External id": 243336,
      "cbid": 211, "correlation": 243336
    }
  },
  {
    "ph": "s", "id": 243336, "pid": 76337, "tid": -914061504, "ts": 1716454225193472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193533, "dur": 1,
    "args": {
      "External id": 243347,
      "cbid": 251, "correlation": 243347
    }
  },
  {
    "ph": "f", "id": 243347, "pid": 76337, "tid": -914061504, "ts": 1716454225193533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193536, "dur": 0,
    "args": {
      "External id": 243348,
      "cbid": 251, "correlation": 243348
    }
  },
  {
    "ph": "f", "id": 243348, "pid": 76337, "tid": -914061504, "ts": 1716454225193536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225297071, "dur": 6,
    "args": {
      "External id": 243349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243349,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 243349, "pid": 5, "tid": 7, "ts": 1716454225297071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193538, "dur": 12,
    "args": {
      "External id": 243349,
      "cbid": 211, "correlation": 243349
    }
  },
  {
    "ph": "s", "id": 243349, "pid": 76337, "tid": -914061504, "ts": 1716454225193538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225297078, "dur": 3,
    "args": {
      "External id": 243351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243351,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 243351, "pid": 5, "tid": 7, "ts": 1716454225297078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193552, "dur": 5,
    "args": {
      "External id": 243351,
      "cbid": 211, "correlation": 243351
    }
  },
  {
    "ph": "s", "id": 243351, "pid": 76337, "tid": -914061504, "ts": 1716454225193552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225297083, "dur": 153,
    "args": {
      "External id": 243372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243372,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243372, "pid": 5, "tid": 7, "ts": 1716454225297083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193625, "dur": 12,
    "args": {
      "External id": 243372,
      "cbid": 211, "correlation": 243372
    }
  },
  {
    "ph": "s", "id": 243372, "pid": 76337, "tid": -914061504, "ts": 1716454225193625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193721, "dur": 1,
    "args": {
      "External id": 243390,
      "cbid": 251, "correlation": 243390
    }
  },
  {
    "ph": "f", "id": 243390, "pid": 76337, "tid": -914061504, "ts": 1716454225193721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225297237, "dur": 103,
    "args": {
      "External id": 243392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243392,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243392, "pid": 5, "tid": 7, "ts": 1716454225297237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193727, "dur": 13,
    "args": {
      "External id": 243392,
      "cbid": 211, "correlation": 243392
    }
  },
  {
    "ph": "s", "id": 243392, "pid": 76337, "tid": -914061504, "ts": 1716454225193727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225297341, "dur": 35,
    "args": {
      "External id": 243400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243400,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243400, "pid": 5, "tid": 7, "ts": 1716454225297341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193796, "dur": 13,
    "args": {
      "External id": 243400,
      "cbid": 211, "correlation": 243400
    }
  },
  {
    "ph": "s", "id": 243400, "pid": 76337, "tid": -914061504, "ts": 1716454225193796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225297377, "dur": 66,
    "args": {
      "External id": 243408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243408, "pid": 5, "tid": 7, "ts": 1716454225297377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193837, "dur": 10,
    "args": {
      "External id": 243408,
      "cbid": 211, "correlation": 243408
    }
  },
  {
    "ph": "s", "id": 243408, "pid": 76337, "tid": -914061504, "ts": 1716454225193837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225297444, "dur": 90,
    "args": {
      "External id": 243430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243430,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243430, "pid": 5, "tid": 7, "ts": 1716454225297444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193888, "dur": 11,
    "args": {
      "External id": 243430,
      "cbid": 211, "correlation": 243430
    }
  },
  {
    "ph": "s", "id": 243430, "pid": 76337, "tid": -914061504, "ts": 1716454225193888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193982, "dur": 1,
    "args": {
      "External id": 243446,
      "cbid": 251, "correlation": 243446
    }
  },
  {
    "ph": "f", "id": 243446, "pid": 76337, "tid": -914061504, "ts": 1716454225193982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225297536, "dur": 562,
    "args": {
      "External id": 243448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243448,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243448, "pid": 5, "tid": 7, "ts": 1716454225297536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225193987, "dur": 14,
    "args": {
      "External id": 243448,
      "cbid": 211, "correlation": 243448
    }
  },
  {
    "ph": "s", "id": 243448, "pid": 76337, "tid": -914061504, "ts": 1716454225193987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225298099, "dur": 241,
    "args": {
      "External id": 243456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243456, "pid": 5, "tid": 7, "ts": 1716454225298099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194055, "dur": 13,
    "args": {
      "External id": 243456,
      "cbid": 211, "correlation": 243456
    }
  },
  {
    "ph": "s", "id": 243456, "pid": 76337, "tid": -914061504, "ts": 1716454225194055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225298341, "dur": 252,
    "args": {
      "External id": 243464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243464, "pid": 5, "tid": 7, "ts": 1716454225298341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194086, "dur": 8,
    "args": {
      "External id": 243464,
      "cbid": 211, "correlation": 243464
    }
  },
  {
    "ph": "s", "id": 243464, "pid": 76337, "tid": -914061504, "ts": 1716454225194086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194168, "dur": 1,
    "args": {
      "External id": 243480,
      "cbid": 251, "correlation": 243480
    }
  },
  {
    "ph": "f", "id": 243480, "pid": 76337, "tid": -914061504, "ts": 1716454225194168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194173, "dur": 0,
    "args": {
      "External id": 243482,
      "cbid": 251, "correlation": 243482
    }
  },
  {
    "ph": "f", "id": 243482, "pid": 76337, "tid": -914061504, "ts": 1716454225194173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225298594, "dur": 352,
    "args": {
      "External id": 243483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243483,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 243483, "pid": 5, "tid": 7, "ts": 1716454225298594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194176, "dur": 12,
    "args": {
      "External id": 243483,
      "cbid": 211, "correlation": 243483
    }
  },
  {
    "ph": "s", "id": 243483, "pid": 76337, "tid": -914061504, "ts": 1716454225194176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225298947, "dur": 50,
    "args": {
      "External id": 243491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243491,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243491, "pid": 5, "tid": 7, "ts": 1716454225298947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194219, "dur": 10,
    "args": {
      "External id": 243491,
      "cbid": 211, "correlation": 243491
    }
  },
  {
    "ph": "s", "id": 243491, "pid": 76337, "tid": -914061504, "ts": 1716454225194219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225298998, "dur": 154,
    "args": {
      "External id": 243502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243502, "pid": 5, "tid": 7, "ts": 1716454225298998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194286, "dur": 13,
    "args": {
      "External id": 243502,
      "cbid": 211, "correlation": 243502
    }
  },
  {
    "ph": "s", "id": 243502, "pid": 76337, "tid": -914061504, "ts": 1716454225194286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194351, "dur": 0,
    "args": {
      "External id": 243514,
      "cbid": 317, "correlation": 243514
    }
  },
  {
    "ph": "f", "id": 243514, "pid": 76337, "tid": -914061504, "ts": 1716454225194351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194352, "dur": 0,
    "args": {
      "External id": 243515,
      "cbid": 203, "correlation": 243515
    }
  },
  {
    "ph": "f", "id": 243515, "pid": 76337, "tid": -914061504, "ts": 1716454225194352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194353, "dur": 0,
    "args": {
      "External id": 243516,
      "cbid": 205, "correlation": 243516
    }
  },
  {
    "ph": "f", "id": 243516, "pid": 76337, "tid": -914061504, "ts": 1716454225194353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194375, "dur": 1,
    "args": {
      "External id": 243520,
      "cbid": 251, "correlation": 243520
    }
  },
  {
    "ph": "f", "id": 243520, "pid": 76337, "tid": -914061504, "ts": 1716454225194375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194377, "dur": 0,
    "args": {
      "External id": 243521,
      "cbid": 251, "correlation": 243521
    }
  },
  {
    "ph": "f", "id": 243521, "pid": 76337, "tid": -914061504, "ts": 1716454225194377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194378, "dur": 0,
    "args": {
      "External id": 243522,
      "cbid": 251, "correlation": 243522
    }
  },
  {
    "ph": "f", "id": 243522, "pid": 76337, "tid": -914061504, "ts": 1716454225194378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194379, "dur": 0,
    "args": {
      "External id": 243523,
      "cbid": 251, "correlation": 243523
    }
  },
  {
    "ph": "f", "id": 243523, "pid": 76337, "tid": -914061504, "ts": 1716454225194379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194380, "dur": 0,
    "args": {
      "External id": 243524,
      "cbid": 251, "correlation": 243524
    }
  },
  {
    "ph": "f", "id": 243524, "pid": 76337, "tid": -914061504, "ts": 1716454225194380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194381, "dur": 0,
    "args": {
      "External id": 243525,
      "cbid": 251, "correlation": 243525
    }
  },
  {
    "ph": "f", "id": 243525, "pid": 76337, "tid": -914061504, "ts": 1716454225194381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194382, "dur": 0,
    "args": {
      "External id": 243526,
      "cbid": 251, "correlation": 243526
    }
  },
  {
    "ph": "f", "id": 243526, "pid": 76337, "tid": -914061504, "ts": 1716454225194382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194383, "dur": 0,
    "args": {
      "External id": 243527,
      "cbid": 251, "correlation": 243527
    }
  },
  {
    "ph": "f", "id": 243527, "pid": 76337, "tid": -914061504, "ts": 1716454225194383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194384, "dur": 0,
    "args": {
      "External id": 243528,
      "cbid": 251, "correlation": 243528
    }
  },
  {
    "ph": "f", "id": 243528, "pid": 76337, "tid": -914061504, "ts": 1716454225194384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225299154, "dur": 113,
    "args": {
      "External id": 243529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243529,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243529, "pid": 5, "tid": 7, "ts": 1716454225299154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194386, "dur": 13,
    "args": {
      "External id": 243529,
      "cbid": 211, "correlation": 243529
    }
  },
  {
    "ph": "s", "id": 243529, "pid": 76337, "tid": -914061504, "ts": 1716454225194386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225299268, "dur": 58,
    "args": {
      "External id": 243535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243535, "pid": 5, "tid": 7, "ts": 1716454225299268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194421, "dur": 9,
    "args": {
      "External id": 243535,
      "cbid": 211, "correlation": 243535
    }
  },
  {
    "ph": "s", "id": 243535, "pid": 76337, "tid": -914061504, "ts": 1716454225194421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225299328, "dur": 50,
    "args": {
      "External id": 243543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243543,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243543, "pid": 5, "tid": 7, "ts": 1716454225299328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194453, "dur": 8,
    "args": {
      "External id": 243543,
      "cbid": 211, "correlation": 243543
    }
  },
  {
    "ph": "s", "id": 243543, "pid": 76337, "tid": -914061504, "ts": 1716454225194453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225299379, "dur": 97,
    "args": {
      "External id": 243552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243552,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243552, "pid": 5, "tid": 7, "ts": 1716454225299379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194492, "dur": 10,
    "args": {
      "External id": 243552,
      "cbid": 211, "correlation": 243552
    }
  },
  {
    "ph": "s", "id": 243552, "pid": 76337, "tid": -914061504, "ts": 1716454225194492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225299477, "dur": 91,
    "args": {
      "External id": 243572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243572,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 243572, "pid": 5, "tid": 7, "ts": 1716454225299477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194563, "dur": 11,
    "args": {
      "External id": 243572,
      "cbid": 211, "correlation": 243572
    }
  },
  {
    "ph": "s", "id": 243572, "pid": 76337, "tid": -914061504, "ts": 1716454225194563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225299570, "dur": 4,
    "args": {
      "External id": 243584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243584,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 243584, "pid": 5, "tid": 7, "ts": 1716454225299570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194584, "dur": 6,
    "args": {
      "External id": 243584,
      "cbid": 211, "correlation": 243584
    }
  },
  {
    "ph": "s", "id": 243584, "pid": 76337, "tid": -914061504, "ts": 1716454225194584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225299576, "dur": 108,
    "args": {
      "External id": 243587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243587,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243587, "pid": 5, "tid": 7, "ts": 1716454225299576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194603, "dur": 7,
    "args": {
      "External id": 243587,
      "cbid": 211, "correlation": 243587
    }
  },
  {
    "ph": "s", "id": 243587, "pid": 76337, "tid": -914061504, "ts": 1716454225194603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225299685, "dur": 69,
    "args": {
      "External id": 243596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243596,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243596, "pid": 5, "tid": 7, "ts": 1716454225299685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194641, "dur": 10,
    "args": {
      "External id": 243596,
      "cbid": 211, "correlation": 243596
    }
  },
  {
    "ph": "s", "id": 243596, "pid": 76337, "tid": -914061504, "ts": 1716454225194641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194692, "dur": 0,
    "args": {
      "External id": 243606,
      "cbid": 317, "correlation": 243606
    }
  },
  {
    "ph": "f", "id": 243606, "pid": 76337, "tid": -914061504, "ts": 1716454225194692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194693, "dur": 0,
    "args": {
      "External id": 243607,
      "cbid": 203, "correlation": 243607
    }
  },
  {
    "ph": "f", "id": 243607, "pid": 76337, "tid": -914061504, "ts": 1716454225194693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194693, "dur": 0,
    "args": {
      "External id": 243608,
      "cbid": 205, "correlation": 243608
    }
  },
  {
    "ph": "f", "id": 243608, "pid": 76337, "tid": -914061504, "ts": 1716454225194693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225299755, "dur": 76,
    "args": {
      "External id": 243612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243612,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243612, "pid": 5, "tid": 7, "ts": 1716454225299755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194709, "dur": 12,
    "args": {
      "External id": 243612,
      "cbid": 211, "correlation": 243612
    }
  },
  {
    "ph": "s", "id": 243612, "pid": 76337, "tid": -914061504, "ts": 1716454225194709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225299832, "dur": 23,
    "args": {
      "External id": 243614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243614,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243614, "pid": 5, "tid": 7, "ts": 1716454225299832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194723, "dur": 5,
    "args": {
      "External id": 243614,
      "cbid": 211, "correlation": 243614
    }
  },
  {
    "ph": "s", "id": 243614, "pid": 76337, "tid": -914061504, "ts": 1716454225194723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225299857, "dur": 4,
    "args": {
      "External id": 243616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243616,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243616, "pid": 5, "tid": 7, "ts": 1716454225299857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194733, "dur": 6,
    "args": {
      "External id": 243616,
      "cbid": 211, "correlation": 243616
    }
  },
  {
    "ph": "s", "id": 243616, "pid": 76337, "tid": -914061504, "ts": 1716454225194733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194742, "dur": 0,
    "args": {
      "External id": 243617,
      "cbid": 51, "correlation": 243617
    }
  },
  {
    "ph": "s", "id": 243617, "pid": 76337, "tid": -914061504, "ts": 1716454225194742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225299862, "dur": 1349,
    "args": {
      "External id": 243618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243618,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243618, "pid": 5, "tid": 7, "ts": 1716454225299862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194742, "dur": 5,
    "args": {
      "External id": 243618,
      "cbid": 211, "correlation": 243618
    }
  },
  {
    "ph": "s", "id": 243618, "pid": 76337, "tid": -914061504, "ts": 1716454225194742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225301212, "dur": 58,
    "args": {
      "External id": 243623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243623,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243623, "pid": 5, "tid": 7, "ts": 1716454225301212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194770, "dur": 9,
    "args": {
      "External id": 243623,
      "cbid": 211, "correlation": 243623
    }
  },
  {
    "ph": "s", "id": 243623, "pid": 76337, "tid": -914061504, "ts": 1716454225194770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225301271, "dur": 4,
    "args": {
      "External id": 243631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243631,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243631, "pid": 5, "tid": 7, "ts": 1716454225301271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194814, "dur": 9,
    "args": {
      "External id": 243631,
      "cbid": 211, "correlation": 243631
    }
  },
  {
    "ph": "s", "id": 243631, "pid": 76337, "tid": -914061504, "ts": 1716454225194814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194879, "dur": 1,
    "args": {
      "External id": 243647,
      "cbid": 251, "correlation": 243647
    }
  },
  {
    "ph": "f", "id": 243647, "pid": 76337, "tid": -914061504, "ts": 1716454225194879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194884, "dur": 0,
    "args": {
      "External id": 243649,
      "cbid": 251, "correlation": 243649
    }
  },
  {
    "ph": "f", "id": 243649, "pid": 76337, "tid": -914061504, "ts": 1716454225194884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225301276, "dur": 11,
    "args": {
      "External id": 243650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243650,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 243650, "pid": 5, "tid": 7, "ts": 1716454225301276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194886, "dur": 13,
    "args": {
      "External id": 243650,
      "cbid": 211, "correlation": 243650
    }
  },
  {
    "ph": "s", "id": 243650, "pid": 76337, "tid": -914061504, "ts": 1716454225194886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225301288, "dur": 5,
    "args": {
      "External id": 243652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243652,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 243652, "pid": 5, "tid": 7, "ts": 1716454225301288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194901, "dur": 5,
    "args": {
      "External id": 243652,
      "cbid": 211, "correlation": 243652
    }
  },
  {
    "ph": "s", "id": 243652, "pid": 76337, "tid": -914061504, "ts": 1716454225194901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225301294, "dur": 53,
    "args": {
      "External id": 243662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243662, "pid": 5, "tid": 7, "ts": 1716454225301294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225194958, "dur": 12,
    "args": {
      "External id": 243662,
      "cbid": 211, "correlation": 243662
    }
  },
  {
    "ph": "s", "id": 243662, "pid": 76337, "tid": -914061504, "ts": 1716454225194958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225301349, "dur": 51,
    "args": {
      "External id": 243682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243682,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 243682, "pid": 5, "tid": 7, "ts": 1716454225301349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195032, "dur": 12,
    "args": {
      "External id": 243682,
      "cbid": 211, "correlation": 243682
    }
  },
  {
    "ph": "s", "id": 243682, "pid": 76337, "tid": -914061504, "ts": 1716454225195032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225301401, "dur": 4,
    "args": {
      "External id": 243694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243694,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243694, "pid": 5, "tid": 7, "ts": 1716454225301401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195054, "dur": 6,
    "args": {
      "External id": 243694,
      "cbid": 211, "correlation": 243694
    }
  },
  {
    "ph": "s", "id": 243694, "pid": 76337, "tid": -914061504, "ts": 1716454225195054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225301406, "dur": 55,
    "args": {
      "External id": 243697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243697,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243697, "pid": 5, "tid": 7, "ts": 1716454225301406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195072, "dur": 6,
    "args": {
      "External id": 243697,
      "cbid": 211, "correlation": 243697
    }
  },
  {
    "ph": "s", "id": 243697, "pid": 76337, "tid": -914061504, "ts": 1716454225195072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225301463, "dur": 37,
    "args": {
      "External id": 243706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243706,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243706, "pid": 5, "tid": 7, "ts": 1716454225301463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195113, "dur": 9,
    "args": {
      "External id": 243706,
      "cbid": 211, "correlation": 243706
    }
  },
  {
    "ph": "s", "id": 243706, "pid": 76337, "tid": -914061504, "ts": 1716454225195113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195174, "dur": 0,
    "args": {
      "External id": 243716,
      "cbid": 317, "correlation": 243716
    }
  },
  {
    "ph": "f", "id": 243716, "pid": 76337, "tid": -914061504, "ts": 1716454225195174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195175, "dur": 0,
    "args": {
      "External id": 243717,
      "cbid": 203, "correlation": 243717
    }
  },
  {
    "ph": "f", "id": 243717, "pid": 76337, "tid": -914061504, "ts": 1716454225195175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195176, "dur": 0,
    "args": {
      "External id": 243718,
      "cbid": 205, "correlation": 243718
    }
  },
  {
    "ph": "f", "id": 243718, "pid": 76337, "tid": -914061504, "ts": 1716454225195176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225301501, "dur": 39,
    "args": {
      "External id": 243722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243722,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243722, "pid": 5, "tid": 7, "ts": 1716454225301501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195191, "dur": 12,
    "args": {
      "External id": 243722,
      "cbid": 211, "correlation": 243722
    }
  },
  {
    "ph": "s", "id": 243722, "pid": 76337, "tid": -914061504, "ts": 1716454225195191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225301542, "dur": 14,
    "args": {
      "External id": 243724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243724,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243724, "pid": 5, "tid": 7, "ts": 1716454225301542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195205, "dur": 5,
    "args": {
      "External id": 243724,
      "cbid": 211, "correlation": 243724
    }
  },
  {
    "ph": "s", "id": 243724, "pid": 76337, "tid": -914061504, "ts": 1716454225195205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225301557, "dur": 3,
    "args": {
      "External id": 243726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243726,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243726, "pid": 5, "tid": 7, "ts": 1716454225301557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195214, "dur": 5,
    "args": {
      "External id": 243726,
      "cbid": 211, "correlation": 243726
    }
  },
  {
    "ph": "s", "id": 243726, "pid": 76337, "tid": -914061504, "ts": 1716454225195214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195222, "dur": 0,
    "args": {
      "External id": 243727,
      "cbid": 51, "correlation": 243727
    }
  },
  {
    "ph": "s", "id": 243727, "pid": 76337, "tid": -914061504, "ts": 1716454225195222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225301561, "dur": 687,
    "args": {
      "External id": 243728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243728,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243728, "pid": 5, "tid": 7, "ts": 1716454225301561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195223, "dur": 5,
    "args": {
      "External id": 243728,
      "cbid": 211, "correlation": 243728
    }
  },
  {
    "ph": "s", "id": 243728, "pid": 76337, "tid": -914061504, "ts": 1716454225195223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225302249, "dur": 58,
    "args": {
      "External id": 243733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243733, "pid": 5, "tid": 7, "ts": 1716454225302249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195251, "dur": 8,
    "args": {
      "External id": 243733,
      "cbid": 211, "correlation": 243733
    }
  },
  {
    "ph": "s", "id": 243733, "pid": 76337, "tid": -914061504, "ts": 1716454225195251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195308, "dur": 0,
    "args": {
      "External id": 243743,
      "cbid": 317, "correlation": 243743
    }
  },
  {
    "ph": "f", "id": 243743, "pid": 76337, "tid": -914061504, "ts": 1716454225195308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195309, "dur": 0,
    "args": {
      "External id": 243744,
      "cbid": 203, "correlation": 243744
    }
  },
  {
    "ph": "f", "id": 243744, "pid": 76337, "tid": -914061504, "ts": 1716454225195309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195310, "dur": 0,
    "args": {
      "External id": 243745,
      "cbid": 205, "correlation": 243745
    }
  },
  {
    "ph": "f", "id": 243745, "pid": 76337, "tid": -914061504, "ts": 1716454225195310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225302309, "dur": 75,
    "args": {
      "External id": 243749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243749,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243749, "pid": 5, "tid": 7, "ts": 1716454225302309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195322, "dur": 12,
    "args": {
      "External id": 243749,
      "cbid": 211, "correlation": 243749
    }
  },
  {
    "ph": "s", "id": 243749, "pid": 76337, "tid": -914061504, "ts": 1716454225195322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225302385, "dur": 204,
    "args": {
      "External id": 243751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243751,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243751, "pid": 5, "tid": 7, "ts": 1716454225302385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195342, "dur": 8,
    "args": {
      "External id": 243751,
      "cbid": 211, "correlation": 243751
    }
  },
  {
    "ph": "s", "id": 243751, "pid": 76337, "tid": -914061504, "ts": 1716454225195342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225302591, "dur": 37,
    "args": {
      "External id": 243753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243753,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243753, "pid": 5, "tid": 7, "ts": 1716454225302591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195354, "dur": 6,
    "args": {
      "External id": 243753,
      "cbid": 211, "correlation": 243753
    }
  },
  {
    "ph": "s", "id": 243753, "pid": 76337, "tid": -914061504, "ts": 1716454225195354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225302629, "dur": 59,
    "args": {
      "External id": 243759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243759, "pid": 5, "tid": 7, "ts": 1716454225302629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195381, "dur": 9,
    "args": {
      "External id": 243759,
      "cbid": 211, "correlation": 243759
    }
  },
  {
    "ph": "s", "id": 243759, "pid": 76337, "tid": -914061504, "ts": 1716454225195381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225302689, "dur": 50,
    "args": {
      "External id": 243767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243767,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243767, "pid": 5, "tid": 7, "ts": 1716454225302689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195410, "dur": 8,
    "args": {
      "External id": 243767,
      "cbid": 211, "correlation": 243767
    }
  },
  {
    "ph": "s", "id": 243767, "pid": 76337, "tid": -914061504, "ts": 1716454225195410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225302741, "dur": 36,
    "args": {
      "External id": 243775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243775,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243775, "pid": 5, "tid": 7, "ts": 1716454225302741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195439, "dur": 8,
    "args": {
      "External id": 243775,
      "cbid": 211, "correlation": 243775
    }
  },
  {
    "ph": "s", "id": 243775, "pid": 76337, "tid": -914061504, "ts": 1716454225195439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225302778, "dur": 50,
    "args": {
      "External id": 243795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243795,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 243795, "pid": 5, "tid": 7, "ts": 1716454225302778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195522, "dur": 13,
    "args": {
      "External id": 243795,
      "cbid": 211, "correlation": 243795
    }
  },
  {
    "ph": "s", "id": 243795, "pid": 76337, "tid": -914061504, "ts": 1716454225195522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225302830, "dur": 4,
    "args": {
      "External id": 243807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243807,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 243807, "pid": 5, "tid": 7, "ts": 1716454225302830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195545, "dur": 6,
    "args": {
      "External id": 243807,
      "cbid": 211, "correlation": 243807
    }
  },
  {
    "ph": "s", "id": 243807, "pid": 76337, "tid": -914061504, "ts": 1716454225195545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225302835, "dur": 56,
    "args": {
      "External id": 243810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243810, "pid": 5, "tid": 7, "ts": 1716454225302835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195562, "dur": 6,
    "args": {
      "External id": 243810,
      "cbid": 211, "correlation": 243810
    }
  },
  {
    "ph": "s", "id": 243810, "pid": 76337, "tid": -914061504, "ts": 1716454225195562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195618, "dur": 0,
    "args": {
      "External id": 243821,
      "cbid": 317, "correlation": 243821
    }
  },
  {
    "ph": "f", "id": 243821, "pid": 76337, "tid": -914061504, "ts": 1716454225195618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195619, "dur": 0,
    "args": {
      "External id": 243822,
      "cbid": 203, "correlation": 243822
    }
  },
  {
    "ph": "f", "id": 243822, "pid": 76337, "tid": -914061504, "ts": 1716454225195619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195620, "dur": 0,
    "args": {
      "External id": 243823,
      "cbid": 205, "correlation": 243823
    }
  },
  {
    "ph": "f", "id": 243823, "pid": 76337, "tid": -914061504, "ts": 1716454225195620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195642, "dur": 1,
    "args": {
      "External id": 243827,
      "cbid": 251, "correlation": 243827
    }
  },
  {
    "ph": "f", "id": 243827, "pid": 76337, "tid": -914061504, "ts": 1716454225195642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195644, "dur": 0,
    "args": {
      "External id": 243828,
      "cbid": 251, "correlation": 243828
    }
  },
  {
    "ph": "f", "id": 243828, "pid": 76337, "tid": -914061504, "ts": 1716454225195644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195645, "dur": 0,
    "args": {
      "External id": 243829,
      "cbid": 251, "correlation": 243829
    }
  },
  {
    "ph": "f", "id": 243829, "pid": 76337, "tid": -914061504, "ts": 1716454225195645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195645, "dur": 0,
    "args": {
      "External id": 243830,
      "cbid": 251, "correlation": 243830
    }
  },
  {
    "ph": "f", "id": 243830, "pid": 76337, "tid": -914061504, "ts": 1716454225195645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195646, "dur": 0,
    "args": {
      "External id": 243831,
      "cbid": 251, "correlation": 243831
    }
  },
  {
    "ph": "f", "id": 243831, "pid": 76337, "tid": -914061504, "ts": 1716454225195646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195647, "dur": 0,
    "args": {
      "External id": 243832,
      "cbid": 251, "correlation": 243832
    }
  },
  {
    "ph": "f", "id": 243832, "pid": 76337, "tid": -914061504, "ts": 1716454225195647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195648, "dur": 0,
    "args": {
      "External id": 243833,
      "cbid": 251, "correlation": 243833
    }
  },
  {
    "ph": "f", "id": 243833, "pid": 76337, "tid": -914061504, "ts": 1716454225195648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195648, "dur": 0,
    "args": {
      "External id": 243834,
      "cbid": 251, "correlation": 243834
    }
  },
  {
    "ph": "f", "id": 243834, "pid": 76337, "tid": -914061504, "ts": 1716454225195648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195650, "dur": 0,
    "args": {
      "External id": 243835,
      "cbid": 251, "correlation": 243835
    }
  },
  {
    "ph": "f", "id": 243835, "pid": 76337, "tid": -914061504, "ts": 1716454225195650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225302892, "dur": 112,
    "args": {
      "External id": 243836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243836,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243836, "pid": 5, "tid": 7, "ts": 1716454225302892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195652, "dur": 12,
    "args": {
      "External id": 243836,
      "cbid": 211, "correlation": 243836
    }
  },
  {
    "ph": "s", "id": 243836, "pid": 76337, "tid": -914061504, "ts": 1716454225195652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225303006, "dur": 59,
    "args": {
      "External id": 243842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243842, "pid": 5, "tid": 7, "ts": 1716454225303006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195688, "dur": 8,
    "args": {
      "External id": 243842,
      "cbid": 211, "correlation": 243842
    }
  },
  {
    "ph": "s", "id": 243842, "pid": 76337, "tid": -914061504, "ts": 1716454225195688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225303066, "dur": 617,
    "args": {
      "External id": 243851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243851,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243851, "pid": 5, "tid": 7, "ts": 1716454225303066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195768, "dur": 15,
    "args": {
      "External id": 243851,
      "cbid": 211, "correlation": 243851
    }
  },
  {
    "ph": "s", "id": 243851, "pid": 76337, "tid": -914061504, "ts": 1716454225195768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225303684, "dur": 176,
    "args": {
      "External id": 243873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243873,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243873, "pid": 5, "tid": 7, "ts": 1716454225303684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195824, "dur": 11,
    "args": {
      "External id": 243873,
      "cbid": 211, "correlation": 243873
    }
  },
  {
    "ph": "s", "id": 243873, "pid": 76337, "tid": -914061504, "ts": 1716454225195824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195910, "dur": 1,
    "args": {
      "External id": 243884,
      "cbid": 251, "correlation": 243884
    }
  },
  {
    "ph": "f", "id": 243884, "pid": 76337, "tid": -914061504, "ts": 1716454225195910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225303862, "dur": 194,
    "args": {
      "External id": 243885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243885,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243885, "pid": 5, "tid": 7, "ts": 1716454225303862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195915, "dur": 13,
    "args": {
      "External id": 243885,
      "cbid": 211, "correlation": 243885
    }
  },
  {
    "ph": "s", "id": 243885, "pid": 76337, "tid": -914061504, "ts": 1716454225195915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195992, "dur": 1,
    "args": {
      "External id": 243896,
      "cbid": 251, "correlation": 243896
    }
  },
  {
    "ph": "f", "id": 243896, "pid": 76337, "tid": -914061504, "ts": 1716454225195992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225304058, "dur": 183,
    "args": {
      "External id": 243897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243897,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243897, "pid": 5, "tid": 7, "ts": 1716454225304058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225195997, "dur": 12,
    "args": {
      "External id": 243897,
      "cbid": 211, "correlation": 243897
    }
  },
  {
    "ph": "s", "id": 243897, "pid": 76337, "tid": -914061504, "ts": 1716454225195997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196062, "dur": 1,
    "args": {
      "External id": 243908,
      "cbid": 251, "correlation": 243908
    }
  },
  {
    "ph": "f", "id": 243908, "pid": 76337, "tid": -914061504, "ts": 1716454225196062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225304241, "dur": 185,
    "args": {
      "External id": 243909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243909,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243909, "pid": 5, "tid": 7, "ts": 1716454225304241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196066, "dur": 11,
    "args": {
      "External id": 243909,
      "cbid": 211, "correlation": 243909
    }
  },
  {
    "ph": "s", "id": 243909, "pid": 76337, "tid": -914061504, "ts": 1716454225196066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225304428, "dur": 18246,
    "args": {
      "External id": 243930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243930,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 243930, "pid": 5, "tid": 7, "ts": 1716454225304428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196147, "dur": 13,
    "args": {
      "External id": 243930,
      "cbid": 211, "correlation": 243930
    }
  },
  {
    "ph": "s", "id": 243930, "pid": 76337, "tid": -914061504, "ts": 1716454225196147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196244, "dur": 1,
    "args": {
      "External id": 243948,
      "cbid": 251, "correlation": 243948
    }
  },
  {
    "ph": "f", "id": 243948, "pid": 76337, "tid": -914061504, "ts": 1716454225196244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225322676, "dur": 199,
    "args": {
      "External id": 243950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243950,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 243950, "pid": 5, "tid": 7, "ts": 1716454225322676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196249, "dur": 13,
    "args": {
      "External id": 243950,
      "cbid": 211, "correlation": 243950
    }
  },
  {
    "ph": "s", "id": 243950, "pid": 76337, "tid": -914061504, "ts": 1716454225196249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225322876, "dur": 66,
    "args": {
      "External id": 243958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243958,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243958, "pid": 5, "tid": 7, "ts": 1716454225322876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196319, "dur": 13,
    "args": {
      "External id": 243958,
      "cbid": 211, "correlation": 243958
    }
  },
  {
    "ph": "s", "id": 243958, "pid": 76337, "tid": -914061504, "ts": 1716454225196319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225322943, "dur": 97,
    "args": {
      "External id": 243966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243966,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243966, "pid": 5, "tid": 7, "ts": 1716454225322943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196358, "dur": 9,
    "args": {
      "External id": 243966,
      "cbid": 211, "correlation": 243966
    }
  },
  {
    "ph": "s", "id": 243966, "pid": 76337, "tid": -914061504, "ts": 1716454225196358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225323042, "dur": 54,
    "args": {
      "External id": 243977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243977, "pid": 5, "tid": 7, "ts": 1716454225323042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196429, "dur": 12,
    "args": {
      "External id": 243977,
      "cbid": 211, "correlation": 243977
    }
  },
  {
    "ph": "s", "id": 243977, "pid": 76337, "tid": -914061504, "ts": 1716454225196429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225323096, "dur": 90,
    "args": {
      "External id": 243999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 243999,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 243999, "pid": 5, "tid": 7, "ts": 1716454225323096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225196459, "dur": 1398,
    "args": {
      "External id": 243999,
      "cbid": 211, "correlation": 243999
    }
  },
  {
    "ph": "s", "id": 243999, "pid": 76337, "tid": -914061504, "ts": 1716454225196459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225197934, "dur": 1,
    "args": {
      "External id": 244010,
      "cbid": 251, "correlation": 244010
    }
  },
  {
    "ph": "f", "id": 244010, "pid": 76337, "tid": -914061504, "ts": 1716454225197934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225323187, "dur": 103,
    "args": {
      "External id": 244011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244011,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244011, "pid": 5, "tid": 7, "ts": 1716454225323187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225197939, "dur": 65,
    "args": {
      "External id": 244011,
      "cbid": 211, "correlation": 244011
    }
  },
  {
    "ph": "s", "id": 244011, "pid": 76337, "tid": -914061504, "ts": 1716454225197939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198065, "dur": 1,
    "args": {
      "External id": 244022,
      "cbid": 251, "correlation": 244022
    }
  },
  {
    "ph": "f", "id": 244022, "pid": 76337, "tid": -914061504, "ts": 1716454225198065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198069, "dur": 0,
    "args": {
      "External id": 244023,
      "cbid": 251, "correlation": 244023
    }
  },
  {
    "ph": "f", "id": 244023, "pid": 76337, "tid": -914061504, "ts": 1716454225198069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225323291, "dur": 10,
    "args": {
      "External id": 244024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244024,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 244024, "pid": 5, "tid": 7, "ts": 1716454225323291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198070, "dur": 12,
    "args": {
      "External id": 244024,
      "cbid": 211, "correlation": 244024
    }
  },
  {
    "ph": "s", "id": 244024, "pid": 76337, "tid": -914061504, "ts": 1716454225198070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225323302, "dur": 5,
    "args": {
      "External id": 244026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244026,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 244026, "pid": 5, "tid": 7, "ts": 1716454225323302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198084, "dur": 6,
    "args": {
      "External id": 244026,
      "cbid": 211, "correlation": 244026
    }
  },
  {
    "ph": "s", "id": 244026, "pid": 76337, "tid": -914061504, "ts": 1716454225198084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198145, "dur": 1,
    "args": {
      "External id": 244037,
      "cbid": 251, "correlation": 244037
    }
  },
  {
    "ph": "f", "id": 244037, "pid": 76337, "tid": -914061504, "ts": 1716454225198145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198148, "dur": 0,
    "args": {
      "External id": 244038,
      "cbid": 251, "correlation": 244038
    }
  },
  {
    "ph": "f", "id": 244038, "pid": 76337, "tid": -914061504, "ts": 1716454225198148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225323309, "dur": 6,
    "args": {
      "External id": 244039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244039,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 244039, "pid": 5, "tid": 7, "ts": 1716454225323309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198149, "dur": 13,
    "args": {
      "External id": 244039,
      "cbid": 211, "correlation": 244039
    }
  },
  {
    "ph": "s", "id": 244039, "pid": 76337, "tid": -914061504, "ts": 1716454225198149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225323316, "dur": 3,
    "args": {
      "External id": 244041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244041,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 244041, "pid": 5, "tid": 7, "ts": 1716454225323316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198164, "dur": 5,
    "args": {
      "External id": 244041,
      "cbid": 211, "correlation": 244041
    }
  },
  {
    "ph": "s", "id": 244041, "pid": 76337, "tid": -914061504, "ts": 1716454225198164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225323321, "dur": 154,
    "args": {
      "External id": 244062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244062,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244062, "pid": 5, "tid": 7, "ts": 1716454225323321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198238, "dur": 12,
    "args": {
      "External id": 244062,
      "cbid": 211, "correlation": 244062
    }
  },
  {
    "ph": "s", "id": 244062, "pid": 76337, "tid": -914061504, "ts": 1716454225198238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198334, "dur": 1,
    "args": {
      "External id": 244080,
      "cbid": 251, "correlation": 244080
    }
  },
  {
    "ph": "f", "id": 244080, "pid": 76337, "tid": -914061504, "ts": 1716454225198334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225323477, "dur": 103,
    "args": {
      "External id": 244082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244082,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244082, "pid": 5, "tid": 7, "ts": 1716454225323477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198340, "dur": 14,
    "args": {
      "External id": 244082,
      "cbid": 211, "correlation": 244082
    }
  },
  {
    "ph": "s", "id": 244082, "pid": 76337, "tid": -914061504, "ts": 1716454225198340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225323581, "dur": 35,
    "args": {
      "External id": 244090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244090,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244090, "pid": 5, "tid": 7, "ts": 1716454225323581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198409, "dur": 12,
    "args": {
      "External id": 244090,
      "cbid": 211, "correlation": 244090
    }
  },
  {
    "ph": "s", "id": 244090, "pid": 76337, "tid": -914061504, "ts": 1716454225198409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225323618, "dur": 67,
    "args": {
      "External id": 244098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244098, "pid": 5, "tid": 7, "ts": 1716454225323618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198450, "dur": 9,
    "args": {
      "External id": 244098,
      "cbid": 211, "correlation": 244098
    }
  },
  {
    "ph": "s", "id": 244098, "pid": 76337, "tid": -914061504, "ts": 1716454225198450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225323686, "dur": 89,
    "args": {
      "External id": 244120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244120,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244120, "pid": 5, "tid": 7, "ts": 1716454225323686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198500, "dur": 10,
    "args": {
      "External id": 244120,
      "cbid": 211, "correlation": 244120
    }
  },
  {
    "ph": "s", "id": 244120, "pid": 76337, "tid": -914061504, "ts": 1716454225198500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198585, "dur": 1,
    "args": {
      "External id": 244136,
      "cbid": 251, "correlation": 244136
    }
  },
  {
    "ph": "f", "id": 244136, "pid": 76337, "tid": -914061504, "ts": 1716454225198585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225323776, "dur": 566,
    "args": {
      "External id": 244138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244138,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244138, "pid": 5, "tid": 7, "ts": 1716454225323776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198590, "dur": 14,
    "args": {
      "External id": 244138,
      "cbid": 211, "correlation": 244138
    }
  },
  {
    "ph": "s", "id": 244138, "pid": 76337, "tid": -914061504, "ts": 1716454225198590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225324343, "dur": 238,
    "args": {
      "External id": 244146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244146, "pid": 5, "tid": 7, "ts": 1716454225324343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198657, "dur": 13,
    "args": {
      "External id": 244146,
      "cbid": 211, "correlation": 244146
    }
  },
  {
    "ph": "s", "id": 244146, "pid": 76337, "tid": -914061504, "ts": 1716454225198657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225324583, "dur": 251,
    "args": {
      "External id": 244154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244154,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244154, "pid": 5, "tid": 7, "ts": 1716454225324583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198687, "dur": 9,
    "args": {
      "External id": 244154,
      "cbid": 211, "correlation": 244154
    }
  },
  {
    "ph": "s", "id": 244154, "pid": 76337, "tid": -914061504, "ts": 1716454225198687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198768, "dur": 1,
    "args": {
      "External id": 244170,
      "cbid": 251, "correlation": 244170
    }
  },
  {
    "ph": "f", "id": 244170, "pid": 76337, "tid": -914061504, "ts": 1716454225198768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198774, "dur": 0,
    "args": {
      "External id": 244172,
      "cbid": 251, "correlation": 244172
    }
  },
  {
    "ph": "f", "id": 244172, "pid": 76337, "tid": -914061504, "ts": 1716454225198774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225324836, "dur": 354,
    "args": {
      "External id": 244173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244173,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 244173, "pid": 5, "tid": 7, "ts": 1716454225324836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198776, "dur": 12,
    "args": {
      "External id": 244173,
      "cbid": 211, "correlation": 244173
    }
  },
  {
    "ph": "s", "id": 244173, "pid": 76337, "tid": -914061504, "ts": 1716454225198776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225325191, "dur": 51,
    "args": {
      "External id": 244181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244181,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244181, "pid": 5, "tid": 7, "ts": 1716454225325191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198818, "dur": 10,
    "args": {
      "External id": 244181,
      "cbid": 211, "correlation": 244181
    }
  },
  {
    "ph": "s", "id": 244181, "pid": 76337, "tid": -914061504, "ts": 1716454225198818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225325243, "dur": 155,
    "args": {
      "External id": 244192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244192, "pid": 5, "tid": 7, "ts": 1716454225325243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225198884, "dur": 213,
    "args": {
      "External id": 244192,
      "cbid": 211, "correlation": 244192
    }
  },
  {
    "ph": "s", "id": 244192, "pid": 76337, "tid": -914061504, "ts": 1716454225198884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199150, "dur": 0,
    "args": {
      "External id": 244204,
      "cbid": 317, "correlation": 244204
    }
  },
  {
    "ph": "f", "id": 244204, "pid": 76337, "tid": -914061504, "ts": 1716454225199150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199151, "dur": 0,
    "args": {
      "External id": 244205,
      "cbid": 203, "correlation": 244205
    }
  },
  {
    "ph": "f", "id": 244205, "pid": 76337, "tid": -914061504, "ts": 1716454225199151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199152, "dur": 0,
    "args": {
      "External id": 244206,
      "cbid": 205, "correlation": 244206
    }
  },
  {
    "ph": "f", "id": 244206, "pid": 76337, "tid": -914061504, "ts": 1716454225199152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199173, "dur": 1,
    "args": {
      "External id": 244210,
      "cbid": 251, "correlation": 244210
    }
  },
  {
    "ph": "f", "id": 244210, "pid": 76337, "tid": -914061504, "ts": 1716454225199173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199176, "dur": 0,
    "args": {
      "External id": 244211,
      "cbid": 251, "correlation": 244211
    }
  },
  {
    "ph": "f", "id": 244211, "pid": 76337, "tid": -914061504, "ts": 1716454225199176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199177, "dur": 0,
    "args": {
      "External id": 244212,
      "cbid": 251, "correlation": 244212
    }
  },
  {
    "ph": "f", "id": 244212, "pid": 76337, "tid": -914061504, "ts": 1716454225199177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199178, "dur": 0,
    "args": {
      "External id": 244213,
      "cbid": 251, "correlation": 244213
    }
  },
  {
    "ph": "f", "id": 244213, "pid": 76337, "tid": -914061504, "ts": 1716454225199178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199178, "dur": 0,
    "args": {
      "External id": 244214,
      "cbid": 251, "correlation": 244214
    }
  },
  {
    "ph": "f", "id": 244214, "pid": 76337, "tid": -914061504, "ts": 1716454225199178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199179, "dur": 0,
    "args": {
      "External id": 244215,
      "cbid": 251, "correlation": 244215
    }
  },
  {
    "ph": "f", "id": 244215, "pid": 76337, "tid": -914061504, "ts": 1716454225199179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199180, "dur": 0,
    "args": {
      "External id": 244216,
      "cbid": 251, "correlation": 244216
    }
  },
  {
    "ph": "f", "id": 244216, "pid": 76337, "tid": -914061504, "ts": 1716454225199180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199181, "dur": 0,
    "args": {
      "External id": 244217,
      "cbid": 251, "correlation": 244217
    }
  },
  {
    "ph": "f", "id": 244217, "pid": 76337, "tid": -914061504, "ts": 1716454225199181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199182, "dur": 0,
    "args": {
      "External id": 244218,
      "cbid": 251, "correlation": 244218
    }
  },
  {
    "ph": "f", "id": 244218, "pid": 76337, "tid": -914061504, "ts": 1716454225199182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225325400, "dur": 113,
    "args": {
      "External id": 244219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244219,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244219, "pid": 5, "tid": 7, "ts": 1716454225325400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199184, "dur": 39,
    "args": {
      "External id": 244219,
      "cbid": 211, "correlation": 244219
    }
  },
  {
    "ph": "s", "id": 244219, "pid": 76337, "tid": -914061504, "ts": 1716454225199184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225325514, "dur": 58,
    "args": {
      "External id": 244225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244225,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244225, "pid": 5, "tid": 7, "ts": 1716454225325514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199247, "dur": 104,
    "args": {
      "External id": 244225,
      "cbid": 211, "correlation": 244225
    }
  },
  {
    "ph": "s", "id": 244225, "pid": 76337, "tid": -914061504, "ts": 1716454225199247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225325574, "dur": 51,
    "args": {
      "External id": 244233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244233,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244233, "pid": 5, "tid": 7, "ts": 1716454225325574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199373, "dur": 280,
    "args": {
      "External id": 244233,
      "cbid": 211, "correlation": 244233
    }
  },
  {
    "ph": "s", "id": 244233, "pid": 76337, "tid": -914061504, "ts": 1716454225199373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225325626, "dur": 98,
    "args": {
      "External id": 244242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244242,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244242, "pid": 5, "tid": 7, "ts": 1716454225325626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199684, "dur": 10,
    "args": {
      "External id": 244242,
      "cbid": 211, "correlation": 244242
    }
  },
  {
    "ph": "s", "id": 244242, "pid": 76337, "tid": -914061504, "ts": 1716454225199684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225325725, "dur": 90,
    "args": {
      "External id": 244262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244262,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 244262, "pid": 5, "tid": 7, "ts": 1716454225325725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199756, "dur": 12,
    "args": {
      "External id": 244262,
      "cbid": 211, "correlation": 244262
    }
  },
  {
    "ph": "s", "id": 244262, "pid": 76337, "tid": -914061504, "ts": 1716454225199756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225325816, "dur": 5,
    "args": {
      "External id": 244274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244274,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 244274, "pid": 5, "tid": 7, "ts": 1716454225325816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199777, "dur": 10,
    "args": {
      "External id": 244274,
      "cbid": 211, "correlation": 244274
    }
  },
  {
    "ph": "s", "id": 244274, "pid": 76337, "tid": -914061504, "ts": 1716454225199777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225325822, "dur": 106,
    "args": {
      "External id": 244277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244277, "pid": 5, "tid": 7, "ts": 1716454225325822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199799, "dur": 105,
    "args": {
      "External id": 244277,
      "cbid": 211, "correlation": 244277
    }
  },
  {
    "ph": "s", "id": 244277, "pid": 76337, "tid": -914061504, "ts": 1716454225199799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225325930, "dur": 69,
    "args": {
      "External id": 244286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244286,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244286, "pid": 5, "tid": 7, "ts": 1716454225325930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199937, "dur": 11,
    "args": {
      "External id": 244286,
      "cbid": 211, "correlation": 244286
    }
  },
  {
    "ph": "s", "id": 244286, "pid": 76337, "tid": -914061504, "ts": 1716454225199937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199998, "dur": 0,
    "args": {
      "External id": 244296,
      "cbid": 317, "correlation": 244296
    }
  },
  {
    "ph": "f", "id": 244296, "pid": 76337, "tid": -914061504, "ts": 1716454225199998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199999, "dur": 0,
    "args": {
      "External id": 244297,
      "cbid": 203, "correlation": 244297
    }
  },
  {
    "ph": "f", "id": 244297, "pid": 76337, "tid": -914061504, "ts": 1716454225199999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225199999, "dur": 0,
    "args": {
      "External id": 244298,
      "cbid": 205, "correlation": 244298
    }
  },
  {
    "ph": "f", "id": 244298, "pid": 76337, "tid": -914061504, "ts": 1716454225199999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225326000, "dur": 75,
    "args": {
      "External id": 244302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244302, "pid": 5, "tid": 7, "ts": 1716454225326000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200013, "dur": 12,
    "args": {
      "External id": 244302,
      "cbid": 211, "correlation": 244302
    }
  },
  {
    "ph": "s", "id": 244302, "pid": 76337, "tid": -914061504, "ts": 1716454225200013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225326077, "dur": 24,
    "args": {
      "External id": 244304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244304, "pid": 5, "tid": 7, "ts": 1716454225326077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200028, "dur": 5,
    "args": {
      "External id": 244304,
      "cbid": 211, "correlation": 244304
    }
  },
  {
    "ph": "s", "id": 244304, "pid": 76337, "tid": -914061504, "ts": 1716454225200028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225326102, "dur": 3,
    "args": {
      "External id": 244306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244306, "pid": 5, "tid": 7, "ts": 1716454225326102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200038, "dur": 7,
    "args": {
      "External id": 244306,
      "cbid": 211, "correlation": 244306
    }
  },
  {
    "ph": "s", "id": 244306, "pid": 76337, "tid": -914061504, "ts": 1716454225200038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200049, "dur": 0,
    "args": {
      "External id": 244307,
      "cbid": 51, "correlation": 244307
    }
  },
  {
    "ph": "s", "id": 244307, "pid": 76337, "tid": -914061504, "ts": 1716454225200049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225326107, "dur": 1346,
    "args": {
      "External id": 244308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244308,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244308, "pid": 5, "tid": 7, "ts": 1716454225326107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200049, "dur": 5,
    "args": {
      "External id": 244308,
      "cbid": 211, "correlation": 244308
    }
  },
  {
    "ph": "s", "id": 244308, "pid": 76337, "tid": -914061504, "ts": 1716454225200049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225327454, "dur": 58,
    "args": {
      "External id": 244313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244313, "pid": 5, "tid": 7, "ts": 1716454225327454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200076, "dur": 9,
    "args": {
      "External id": 244313,
      "cbid": 211, "correlation": 244313
    }
  },
  {
    "ph": "s", "id": 244313, "pid": 76337, "tid": -914061504, "ts": 1716454225200076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225327513, "dur": 4,
    "args": {
      "External id": 244321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244321,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244321, "pid": 5, "tid": 7, "ts": 1716454225327513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200120, "dur": 9,
    "args": {
      "External id": 244321,
      "cbid": 211, "correlation": 244321
    }
  },
  {
    "ph": "s", "id": 244321, "pid": 76337, "tid": -914061504, "ts": 1716454225200120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200184, "dur": 1,
    "args": {
      "External id": 244337,
      "cbid": 251, "correlation": 244337
    }
  },
  {
    "ph": "f", "id": 244337, "pid": 76337, "tid": -914061504, "ts": 1716454225200184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200190, "dur": 0,
    "args": {
      "External id": 244339,
      "cbid": 251, "correlation": 244339
    }
  },
  {
    "ph": "f", "id": 244339, "pid": 76337, "tid": -914061504, "ts": 1716454225200190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225327518, "dur": 11,
    "args": {
      "External id": 244340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244340,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 244340, "pid": 5, "tid": 7, "ts": 1716454225327518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200193, "dur": 12,
    "args": {
      "External id": 244340,
      "cbid": 211, "correlation": 244340
    }
  },
  {
    "ph": "s", "id": 244340, "pid": 76337, "tid": -914061504, "ts": 1716454225200193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225327530, "dur": 5,
    "args": {
      "External id": 244342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244342,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 244342, "pid": 5, "tid": 7, "ts": 1716454225327530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200206, "dur": 5,
    "args": {
      "External id": 244342,
      "cbid": 211, "correlation": 244342
    }
  },
  {
    "ph": "s", "id": 244342, "pid": 76337, "tid": -914061504, "ts": 1716454225200206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225327536, "dur": 53,
    "args": {
      "External id": 244352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244352, "pid": 5, "tid": 7, "ts": 1716454225327536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200263, "dur": 532,
    "args": {
      "External id": 244352,
      "cbid": 211, "correlation": 244352
    }
  },
  {
    "ph": "s", "id": 244352, "pid": 76337, "tid": -914061504, "ts": 1716454225200263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225327590, "dur": 51,
    "args": {
      "External id": 244372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244372,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 244372, "pid": 5, "tid": 7, "ts": 1716454225327590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200850, "dur": 12,
    "args": {
      "External id": 244372,
      "cbid": 211, "correlation": 244372
    }
  },
  {
    "ph": "s", "id": 244372, "pid": 76337, "tid": -914061504, "ts": 1716454225200850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225327643, "dur": 4,
    "args": {
      "External id": 244384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244384,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244384, "pid": 5, "tid": 7, "ts": 1716454225327643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200871, "dur": 6,
    "args": {
      "External id": 244384,
      "cbid": 211, "correlation": 244384
    }
  },
  {
    "ph": "s", "id": 244384, "pid": 76337, "tid": -914061504, "ts": 1716454225200871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225327648, "dur": 54,
    "args": {
      "External id": 244387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244387, "pid": 5, "tid": 7, "ts": 1716454225327648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200890, "dur": 6,
    "args": {
      "External id": 244387,
      "cbid": 211, "correlation": 244387
    }
  },
  {
    "ph": "s", "id": 244387, "pid": 76337, "tid": -914061504, "ts": 1716454225200890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225327704, "dur": 37,
    "args": {
      "External id": 244396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244396,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244396, "pid": 5, "tid": 7, "ts": 1716454225327704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225200931, "dur": 10,
    "args": {
      "External id": 244396,
      "cbid": 211, "correlation": 244396
    }
  },
  {
    "ph": "s", "id": 244396, "pid": 76337, "tid": -914061504, "ts": 1716454225200931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201002, "dur": 0,
    "args": {
      "External id": 244406,
      "cbid": 317, "correlation": 244406
    }
  },
  {
    "ph": "f", "id": 244406, "pid": 76337, "tid": -914061504, "ts": 1716454225201002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201002, "dur": 0,
    "args": {
      "External id": 244407,
      "cbid": 203, "correlation": 244407
    }
  },
  {
    "ph": "f", "id": 244407, "pid": 76337, "tid": -914061504, "ts": 1716454225201002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201003, "dur": 0,
    "args": {
      "External id": 244408,
      "cbid": 205, "correlation": 244408
    }
  },
  {
    "ph": "f", "id": 244408, "pid": 76337, "tid": -914061504, "ts": 1716454225201003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225327742, "dur": 40,
    "args": {
      "External id": 244412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244412, "pid": 5, "tid": 7, "ts": 1716454225327742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201018, "dur": 12,
    "args": {
      "External id": 244412,
      "cbid": 211, "correlation": 244412
    }
  },
  {
    "ph": "s", "id": 244412, "pid": 76337, "tid": -914061504, "ts": 1716454225201018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225327783, "dur": 14,
    "args": {
      "External id": 244414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244414,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244414, "pid": 5, "tid": 7, "ts": 1716454225327783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201033, "dur": 5,
    "args": {
      "External id": 244414,
      "cbid": 211, "correlation": 244414
    }
  },
  {
    "ph": "s", "id": 244414, "pid": 76337, "tid": -914061504, "ts": 1716454225201033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225327798, "dur": 3,
    "args": {
      "External id": 244416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244416, "pid": 5, "tid": 7, "ts": 1716454225327798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201042, "dur": 6,
    "args": {
      "External id": 244416,
      "cbid": 211, "correlation": 244416
    }
  },
  {
    "ph": "s", "id": 244416, "pid": 76337, "tid": -914061504, "ts": 1716454225201042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201050, "dur": 0,
    "args": {
      "External id": 244417,
      "cbid": 51, "correlation": 244417
    }
  },
  {
    "ph": "s", "id": 244417, "pid": 76337, "tid": -914061504, "ts": 1716454225201050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225327803, "dur": 686,
    "args": {
      "External id": 244418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244418,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244418, "pid": 5, "tid": 7, "ts": 1716454225327803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201051, "dur": 5,
    "args": {
      "External id": 244418,
      "cbid": 211, "correlation": 244418
    }
  },
  {
    "ph": "s", "id": 244418, "pid": 76337, "tid": -914061504, "ts": 1716454225201051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225328490, "dur": 57,
    "args": {
      "External id": 244423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244423, "pid": 5, "tid": 7, "ts": 1716454225328490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201079, "dur": 9,
    "args": {
      "External id": 244423,
      "cbid": 211, "correlation": 244423
    }
  },
  {
    "ph": "s", "id": 244423, "pid": 76337, "tid": -914061504, "ts": 1716454225201079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201137, "dur": 0,
    "args": {
      "External id": 244433,
      "cbid": 317, "correlation": 244433
    }
  },
  {
    "ph": "f", "id": 244433, "pid": 76337, "tid": -914061504, "ts": 1716454225201137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201138, "dur": 0,
    "args": {
      "External id": 244434,
      "cbid": 203, "correlation": 244434
    }
  },
  {
    "ph": "f", "id": 244434, "pid": 76337, "tid": -914061504, "ts": 1716454225201138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201138, "dur": 0,
    "args": {
      "External id": 244435,
      "cbid": 205, "correlation": 244435
    }
  },
  {
    "ph": "f", "id": 244435, "pid": 76337, "tid": -914061504, "ts": 1716454225201138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225328549, "dur": 75,
    "args": {
      "External id": 244439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244439,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244439, "pid": 5, "tid": 7, "ts": 1716454225328549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201150, "dur": 12,
    "args": {
      "External id": 244439,
      "cbid": 211, "correlation": 244439
    }
  },
  {
    "ph": "s", "id": 244439, "pid": 76337, "tid": -914061504, "ts": 1716454225201150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225328625, "dur": 206,
    "args": {
      "External id": 244441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244441,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244441, "pid": 5, "tid": 7, "ts": 1716454225328625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201169, "dur": 6,
    "args": {
      "External id": 244441,
      "cbid": 211, "correlation": 244441
    }
  },
  {
    "ph": "s", "id": 244441, "pid": 76337, "tid": -914061504, "ts": 1716454225201169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225328832, "dur": 39,
    "args": {
      "External id": 244443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244443,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244443, "pid": 5, "tid": 7, "ts": 1716454225328832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201179, "dur": 6,
    "args": {
      "External id": 244443,
      "cbid": 211, "correlation": 244443
    }
  },
  {
    "ph": "s", "id": 244443, "pid": 76337, "tid": -914061504, "ts": 1716454225201179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225328873, "dur": 58,
    "args": {
      "External id": 244449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244449, "pid": 5, "tid": 7, "ts": 1716454225328873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201205, "dur": 493,
    "args": {
      "External id": 244449,
      "cbid": 211, "correlation": 244449
    }
  },
  {
    "ph": "s", "id": 244449, "pid": 76337, "tid": -914061504, "ts": 1716454225201205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225328933, "dur": 51,
    "args": {
      "External id": 244457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244457,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244457, "pid": 5, "tid": 7, "ts": 1716454225328933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201721, "dur": 8,
    "args": {
      "External id": 244457,
      "cbid": 211, "correlation": 244457
    }
  },
  {
    "ph": "s", "id": 244457, "pid": 76337, "tid": -914061504, "ts": 1716454225201721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225328985, "dur": 35,
    "args": {
      "External id": 244465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244465,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244465, "pid": 5, "tid": 7, "ts": 1716454225328985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201750, "dur": 8,
    "args": {
      "External id": 244465,
      "cbid": 211, "correlation": 244465
    }
  },
  {
    "ph": "s", "id": 244465, "pid": 76337, "tid": -914061504, "ts": 1716454225201750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225329021, "dur": 50,
    "args": {
      "External id": 244485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244485,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 244485, "pid": 5, "tid": 7, "ts": 1716454225329021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201831, "dur": 12,
    "args": {
      "External id": 244485,
      "cbid": 211, "correlation": 244485
    }
  },
  {
    "ph": "s", "id": 244485, "pid": 76337, "tid": -914061504, "ts": 1716454225201831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225329073, "dur": 4,
    "args": {
      "External id": 244497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244497,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244497, "pid": 5, "tid": 7, "ts": 1716454225329073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201854, "dur": 6,
    "args": {
      "External id": 244497,
      "cbid": 211, "correlation": 244497
    }
  },
  {
    "ph": "s", "id": 244497, "pid": 76337, "tid": -914061504, "ts": 1716454225201854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225329078, "dur": 55,
    "args": {
      "External id": 244500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244500, "pid": 5, "tid": 7, "ts": 1716454225329078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201872, "dur": 6,
    "args": {
      "External id": 244500,
      "cbid": 211, "correlation": 244500
    }
  },
  {
    "ph": "s", "id": 244500, "pid": 76337, "tid": -914061504, "ts": 1716454225201872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201930, "dur": 0,
    "args": {
      "External id": 244511,
      "cbid": 317, "correlation": 244511
    }
  },
  {
    "ph": "f", "id": 244511, "pid": 76337, "tid": -914061504, "ts": 1716454225201930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201930, "dur": 0,
    "args": {
      "External id": 244512,
      "cbid": 203, "correlation": 244512
    }
  },
  {
    "ph": "f", "id": 244512, "pid": 76337, "tid": -914061504, "ts": 1716454225201930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201931, "dur": 0,
    "args": {
      "External id": 244513,
      "cbid": 205, "correlation": 244513
    }
  },
  {
    "ph": "f", "id": 244513, "pid": 76337, "tid": -914061504, "ts": 1716454225201931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201952, "dur": 1,
    "args": {
      "External id": 244517,
      "cbid": 251, "correlation": 244517
    }
  },
  {
    "ph": "f", "id": 244517, "pid": 76337, "tid": -914061504, "ts": 1716454225201952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201954, "dur": 0,
    "args": {
      "External id": 244518,
      "cbid": 251, "correlation": 244518
    }
  },
  {
    "ph": "f", "id": 244518, "pid": 76337, "tid": -914061504, "ts": 1716454225201954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201955, "dur": 0,
    "args": {
      "External id": 244519,
      "cbid": 251, "correlation": 244519
    }
  },
  {
    "ph": "f", "id": 244519, "pid": 76337, "tid": -914061504, "ts": 1716454225201955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201956, "dur": 0,
    "args": {
      "External id": 244520,
      "cbid": 251, "correlation": 244520
    }
  },
  {
    "ph": "f", "id": 244520, "pid": 76337, "tid": -914061504, "ts": 1716454225201956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201957, "dur": 0,
    "args": {
      "External id": 244521,
      "cbid": 251, "correlation": 244521
    }
  },
  {
    "ph": "f", "id": 244521, "pid": 76337, "tid": -914061504, "ts": 1716454225201957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201957, "dur": 0,
    "args": {
      "External id": 244522,
      "cbid": 251, "correlation": 244522
    }
  },
  {
    "ph": "f", "id": 244522, "pid": 76337, "tid": -914061504, "ts": 1716454225201957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201958, "dur": 0,
    "args": {
      "External id": 244523,
      "cbid": 251, "correlation": 244523
    }
  },
  {
    "ph": "f", "id": 244523, "pid": 76337, "tid": -914061504, "ts": 1716454225201958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201959, "dur": 0,
    "args": {
      "External id": 244524,
      "cbid": 251, "correlation": 244524
    }
  },
  {
    "ph": "f", "id": 244524, "pid": 76337, "tid": -914061504, "ts": 1716454225201959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201960, "dur": 0,
    "args": {
      "External id": 244525,
      "cbid": 251, "correlation": 244525
    }
  },
  {
    "ph": "f", "id": 244525, "pid": 76337, "tid": -914061504, "ts": 1716454225201960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225329134, "dur": 111,
    "args": {
      "External id": 244526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244526,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244526, "pid": 5, "tid": 7, "ts": 1716454225329134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225201962, "dur": 21,
    "args": {
      "External id": 244526,
      "cbid": 211, "correlation": 244526
    }
  },
  {
    "ph": "s", "id": 244526, "pid": 76337, "tid": -914061504, "ts": 1716454225201962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225329247, "dur": 59,
    "args": {
      "External id": 244532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244532, "pid": 5, "tid": 7, "ts": 1716454225329247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202007, "dur": 10,
    "args": {
      "External id": 244532,
      "cbid": 211, "correlation": 244532
    }
  },
  {
    "ph": "s", "id": 244532, "pid": 76337, "tid": -914061504, "ts": 1716454225202007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225329307, "dur": 629,
    "args": {
      "External id": 244541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244541,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244541, "pid": 5, "tid": 7, "ts": 1716454225329307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202090, "dur": 13,
    "args": {
      "External id": 244541,
      "cbid": 211, "correlation": 244541
    }
  },
  {
    "ph": "s", "id": 244541, "pid": 76337, "tid": -914061504, "ts": 1716454225202090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225329937, "dur": 176,
    "args": {
      "External id": 244563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244563,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244563, "pid": 5, "tid": 7, "ts": 1716454225329937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202148, "dur": 10,
    "args": {
      "External id": 244563,
      "cbid": 211, "correlation": 244563
    }
  },
  {
    "ph": "s", "id": 244563, "pid": 76337, "tid": -914061504, "ts": 1716454225202148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202233, "dur": 1,
    "args": {
      "External id": 244574,
      "cbid": 251, "correlation": 244574
    }
  },
  {
    "ph": "f", "id": 244574, "pid": 76337, "tid": -914061504, "ts": 1716454225202233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225330115, "dur": 193,
    "args": {
      "External id": 244575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244575,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244575, "pid": 5, "tid": 7, "ts": 1716454225330115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202238, "dur": 13,
    "args": {
      "External id": 244575,
      "cbid": 211, "correlation": 244575
    }
  },
  {
    "ph": "s", "id": 244575, "pid": 76337, "tid": -914061504, "ts": 1716454225202238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202306, "dur": 1,
    "args": {
      "External id": 244586,
      "cbid": 251, "correlation": 244586
    }
  },
  {
    "ph": "f", "id": 244586, "pid": 76337, "tid": -914061504, "ts": 1716454225202306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225330309, "dur": 182,
    "args": {
      "External id": 244587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244587,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244587, "pid": 5, "tid": 7, "ts": 1716454225330309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202310, "dur": 12,
    "args": {
      "External id": 244587,
      "cbid": 211, "correlation": 244587
    }
  },
  {
    "ph": "s", "id": 244587, "pid": 76337, "tid": -914061504, "ts": 1716454225202310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202373, "dur": 1,
    "args": {
      "External id": 244598,
      "cbid": 251, "correlation": 244598
    }
  },
  {
    "ph": "f", "id": 244598, "pid": 76337, "tid": -914061504, "ts": 1716454225202373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225330493, "dur": 182,
    "args": {
      "External id": 244599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244599,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244599, "pid": 5, "tid": 7, "ts": 1716454225330493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202376, "dur": 11,
    "args": {
      "External id": 244599,
      "cbid": 211, "correlation": 244599
    }
  },
  {
    "ph": "s", "id": 244599, "pid": 76337, "tid": -914061504, "ts": 1716454225202376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225330676, "dur": 18930,
    "args": {
      "External id": 244620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244620,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244620, "pid": 5, "tid": 7, "ts": 1716454225330676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202456, "dur": 13,
    "args": {
      "External id": 244620,
      "cbid": 211, "correlation": 244620
    }
  },
  {
    "ph": "s", "id": 244620, "pid": 76337, "tid": -914061504, "ts": 1716454225202456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202552, "dur": 1,
    "args": {
      "External id": 244638,
      "cbid": 251, "correlation": 244638
    }
  },
  {
    "ph": "f", "id": 244638, "pid": 76337, "tid": -914061504, "ts": 1716454225202552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225349607, "dur": 204,
    "args": {
      "External id": 244640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244640,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244640, "pid": 5, "tid": 7, "ts": 1716454225349607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202558, "dur": 13,
    "args": {
      "External id": 244640,
      "cbid": 211, "correlation": 244640
    }
  },
  {
    "ph": "s", "id": 244640, "pid": 76337, "tid": -914061504, "ts": 1716454225202558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225349813, "dur": 67,
    "args": {
      "External id": 244648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244648,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244648, "pid": 5, "tid": 7, "ts": 1716454225349813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202628, "dur": 12,
    "args": {
      "External id": 244648,
      "cbid": 211, "correlation": 244648
    }
  },
  {
    "ph": "s", "id": 244648, "pid": 76337, "tid": -914061504, "ts": 1716454225202628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225349881, "dur": 96,
    "args": {
      "External id": 244656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244656,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244656, "pid": 5, "tid": 7, "ts": 1716454225349881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202667, "dur": 80,
    "args": {
      "External id": 244656,
      "cbid": 211, "correlation": 244656
    }
  },
  {
    "ph": "s", "id": 244656, "pid": 76337, "tid": -914061504, "ts": 1716454225202667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225349978, "dur": 55,
    "args": {
      "External id": 244667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244667, "pid": 5, "tid": 7, "ts": 1716454225349978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225202810, "dur": 1850,
    "args": {
      "External id": 244667,
      "cbid": 211, "correlation": 244667
    }
  },
  {
    "ph": "s", "id": 244667, "pid": 76337, "tid": -914061504, "ts": 1716454225202810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225350034, "dur": 94,
    "args": {
      "External id": 244689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244689,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244689, "pid": 5, "tid": 7, "ts": 1716454225350034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204679, "dur": 126,
    "args": {
      "External id": 244689,
      "cbid": 211, "correlation": 244689
    }
  },
  {
    "ph": "s", "id": 244689, "pid": 76337, "tid": -914061504, "ts": 1716454225204679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204881, "dur": 1,
    "args": {
      "External id": 244700,
      "cbid": 251, "correlation": 244700
    }
  },
  {
    "ph": "f", "id": 244700, "pid": 76337, "tid": -914061504, "ts": 1716454225204881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225350129, "dur": 105,
    "args": {
      "External id": 244701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244701,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244701, "pid": 5, "tid": 7, "ts": 1716454225350129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204886, "dur": 13,
    "args": {
      "External id": 244701,
      "cbid": 211, "correlation": 244701
    }
  },
  {
    "ph": "s", "id": 244701, "pid": 76337, "tid": -914061504, "ts": 1716454225204886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204957, "dur": 1,
    "args": {
      "External id": 244712,
      "cbid": 251, "correlation": 244712
    }
  },
  {
    "ph": "f", "id": 244712, "pid": 76337, "tid": -914061504, "ts": 1716454225204957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204960, "dur": 0,
    "args": {
      "External id": 244713,
      "cbid": 251, "correlation": 244713
    }
  },
  {
    "ph": "f", "id": 244713, "pid": 76337, "tid": -914061504, "ts": 1716454225204960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225350236, "dur": 10,
    "args": {
      "External id": 244714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244714,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 244714, "pid": 5, "tid": 7, "ts": 1716454225350236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204962, "dur": 19,
    "args": {
      "External id": 244714,
      "cbid": 211, "correlation": 244714
    }
  },
  {
    "ph": "s", "id": 244714, "pid": 76337, "tid": -914061504, "ts": 1716454225204962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225350247, "dur": 5,
    "args": {
      "External id": 244716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244716,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 244716, "pid": 5, "tid": 7, "ts": 1716454225350247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225204983, "dur": 6,
    "args": {
      "External id": 244716,
      "cbid": 211, "correlation": 244716
    }
  },
  {
    "ph": "s", "id": 244716, "pid": 76337, "tid": -914061504, "ts": 1716454225204983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205045, "dur": 1,
    "args": {
      "External id": 244727,
      "cbid": 251, "correlation": 244727
    }
  },
  {
    "ph": "f", "id": 244727, "pid": 76337, "tid": -914061504, "ts": 1716454225205045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205049, "dur": 0,
    "args": {
      "External id": 244728,
      "cbid": 251, "correlation": 244728
    }
  },
  {
    "ph": "f", "id": 244728, "pid": 76337, "tid": -914061504, "ts": 1716454225205049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225350254, "dur": 6,
    "args": {
      "External id": 244729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244729,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 244729, "pid": 5, "tid": 7, "ts": 1716454225350254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205051, "dur": 12,
    "args": {
      "External id": 244729,
      "cbid": 211, "correlation": 244729
    }
  },
  {
    "ph": "s", "id": 244729, "pid": 76337, "tid": -914061504, "ts": 1716454225205051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225350261, "dur": 3,
    "args": {
      "External id": 244731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244731,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 244731, "pid": 5, "tid": 7, "ts": 1716454225350261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205064, "dur": 6,
    "args": {
      "External id": 244731,
      "cbid": 211, "correlation": 244731
    }
  },
  {
    "ph": "s", "id": 244731, "pid": 76337, "tid": -914061504, "ts": 1716454225205064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225350266, "dur": 158,
    "args": {
      "External id": 244752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244752,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244752, "pid": 5, "tid": 7, "ts": 1716454225350266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205137, "dur": 13,
    "args": {
      "External id": 244752,
      "cbid": 211, "correlation": 244752
    }
  },
  {
    "ph": "s", "id": 244752, "pid": 76337, "tid": -914061504, "ts": 1716454225205137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205233, "dur": 1,
    "args": {
      "External id": 244770,
      "cbid": 251, "correlation": 244770
    }
  },
  {
    "ph": "f", "id": 244770, "pid": 76337, "tid": -914061504, "ts": 1716454225205233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225350426, "dur": 108,
    "args": {
      "External id": 244772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244772,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244772, "pid": 5, "tid": 7, "ts": 1716454225350426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205239, "dur": 14,
    "args": {
      "External id": 244772,
      "cbid": 211, "correlation": 244772
    }
  },
  {
    "ph": "s", "id": 244772, "pid": 76337, "tid": -914061504, "ts": 1716454225205239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225350535, "dur": 35,
    "args": {
      "External id": 244780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244780,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244780, "pid": 5, "tid": 7, "ts": 1716454225350535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205309, "dur": 12,
    "args": {
      "External id": 244780,
      "cbid": 211, "correlation": 244780
    }
  },
  {
    "ph": "s", "id": 244780, "pid": 76337, "tid": -914061504, "ts": 1716454225205309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225350571, "dur": 69,
    "args": {
      "External id": 244788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244788, "pid": 5, "tid": 7, "ts": 1716454225350571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205351, "dur": 9,
    "args": {
      "External id": 244788,
      "cbid": 211, "correlation": 244788
    }
  },
  {
    "ph": "s", "id": 244788, "pid": 76337, "tid": -914061504, "ts": 1716454225205351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225350641, "dur": 93,
    "args": {
      "External id": 244810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244810,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244810, "pid": 5, "tid": 7, "ts": 1716454225350641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205401, "dur": 10,
    "args": {
      "External id": 244810,
      "cbid": 211, "correlation": 244810
    }
  },
  {
    "ph": "s", "id": 244810, "pid": 76337, "tid": -914061504, "ts": 1716454225205401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205486, "dur": 1,
    "args": {
      "External id": 244826,
      "cbid": 251, "correlation": 244826
    }
  },
  {
    "ph": "f", "id": 244826, "pid": 76337, "tid": -914061504, "ts": 1716454225205486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225350736, "dur": 584,
    "args": {
      "External id": 244828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244828,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 244828, "pid": 5, "tid": 7, "ts": 1716454225350736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205492, "dur": 13,
    "args": {
      "External id": 244828,
      "cbid": 211, "correlation": 244828
    }
  },
  {
    "ph": "s", "id": 244828, "pid": 76337, "tid": -914061504, "ts": 1716454225205492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225351321, "dur": 246,
    "args": {
      "External id": 244836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244836,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244836, "pid": 5, "tid": 7, "ts": 1716454225351321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205558, "dur": 13,
    "args": {
      "External id": 244836,
      "cbid": 211, "correlation": 244836
    }
  },
  {
    "ph": "s", "id": 244836, "pid": 76337, "tid": -914061504, "ts": 1716454225205558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225351569, "dur": 252,
    "args": {
      "External id": 244844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244844, "pid": 5, "tid": 7, "ts": 1716454225351569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205588, "dur": 9,
    "args": {
      "External id": 244844,
      "cbid": 211, "correlation": 244844
    }
  },
  {
    "ph": "s", "id": 244844, "pid": 76337, "tid": -914061504, "ts": 1716454225205588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205669, "dur": 1,
    "args": {
      "External id": 244860,
      "cbid": 251, "correlation": 244860
    }
  },
  {
    "ph": "f", "id": 244860, "pid": 76337, "tid": -914061504, "ts": 1716454225205669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205674, "dur": 0,
    "args": {
      "External id": 244862,
      "cbid": 251, "correlation": 244862
    }
  },
  {
    "ph": "f", "id": 244862, "pid": 76337, "tid": -914061504, "ts": 1716454225205674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225351823, "dur": 361,
    "args": {
      "External id": 244863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244863,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 244863, "pid": 5, "tid": 7, "ts": 1716454225351823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205677, "dur": 13,
    "args": {
      "External id": 244863,
      "cbid": 211, "correlation": 244863
    }
  },
  {
    "ph": "s", "id": 244863, "pid": 76337, "tid": -914061504, "ts": 1716454225205677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225352185, "dur": 50,
    "args": {
      "External id": 244871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244871,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244871, "pid": 5, "tid": 7, "ts": 1716454225352185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205719, "dur": 179,
    "args": {
      "External id": 244871,
      "cbid": 211, "correlation": 244871
    }
  },
  {
    "ph": "s", "id": 244871, "pid": 76337, "tid": -914061504, "ts": 1716454225205719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225352236, "dur": 161,
    "args": {
      "External id": 244882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244882, "pid": 5, "tid": 7, "ts": 1716454225352236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225205955, "dur": 70,
    "args": {
      "External id": 244882,
      "cbid": 211, "correlation": 244882
    }
  },
  {
    "ph": "s", "id": 244882, "pid": 76337, "tid": -914061504, "ts": 1716454225205955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206079, "dur": 0,
    "args": {
      "External id": 244894,
      "cbid": 317, "correlation": 244894
    }
  },
  {
    "ph": "f", "id": 244894, "pid": 76337, "tid": -914061504, "ts": 1716454225206079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206080, "dur": 0,
    "args": {
      "External id": 244895,
      "cbid": 203, "correlation": 244895
    }
  },
  {
    "ph": "f", "id": 244895, "pid": 76337, "tid": -914061504, "ts": 1716454225206080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206080, "dur": 0,
    "args": {
      "External id": 244896,
      "cbid": 205, "correlation": 244896
    }
  },
  {
    "ph": "f", "id": 244896, "pid": 76337, "tid": -914061504, "ts": 1716454225206080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206103, "dur": 1,
    "args": {
      "External id": 244900,
      "cbid": 251, "correlation": 244900
    }
  },
  {
    "ph": "f", "id": 244900, "pid": 76337, "tid": -914061504, "ts": 1716454225206103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206104, "dur": 0,
    "args": {
      "External id": 244901,
      "cbid": 251, "correlation": 244901
    }
  },
  {
    "ph": "f", "id": 244901, "pid": 76337, "tid": -914061504, "ts": 1716454225206104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206105, "dur": 0,
    "args": {
      "External id": 244902,
      "cbid": 251, "correlation": 244902
    }
  },
  {
    "ph": "f", "id": 244902, "pid": 76337, "tid": -914061504, "ts": 1716454225206105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206106, "dur": 0,
    "args": {
      "External id": 244903,
      "cbid": 251, "correlation": 244903
    }
  },
  {
    "ph": "f", "id": 244903, "pid": 76337, "tid": -914061504, "ts": 1716454225206106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206107, "dur": 0,
    "args": {
      "External id": 244904,
      "cbid": 251, "correlation": 244904
    }
  },
  {
    "ph": "f", "id": 244904, "pid": 76337, "tid": -914061504, "ts": 1716454225206107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206107, "dur": 0,
    "args": {
      "External id": 244905,
      "cbid": 251, "correlation": 244905
    }
  },
  {
    "ph": "f", "id": 244905, "pid": 76337, "tid": -914061504, "ts": 1716454225206107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206108, "dur": 0,
    "args": {
      "External id": 244906,
      "cbid": 251, "correlation": 244906
    }
  },
  {
    "ph": "f", "id": 244906, "pid": 76337, "tid": -914061504, "ts": 1716454225206108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206109, "dur": 0,
    "args": {
      "External id": 244907,
      "cbid": 251, "correlation": 244907
    }
  },
  {
    "ph": "f", "id": 244907, "pid": 76337, "tid": -914061504, "ts": 1716454225206109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206110, "dur": 0,
    "args": {
      "External id": 244908,
      "cbid": 251, "correlation": 244908
    }
  },
  {
    "ph": "f", "id": 244908, "pid": 76337, "tid": -914061504, "ts": 1716454225206110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225352398, "dur": 117,
    "args": {
      "External id": 244909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244909,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 244909, "pid": 5, "tid": 7, "ts": 1716454225352398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206112, "dur": 41,
    "args": {
      "External id": 244909,
      "cbid": 211, "correlation": 244909
    }
  },
  {
    "ph": "s", "id": 244909, "pid": 76337, "tid": -914061504, "ts": 1716454225206112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225352517, "dur": 60,
    "args": {
      "External id": 244915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244915,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244915, "pid": 5, "tid": 7, "ts": 1716454225352517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206176, "dur": 279,
    "args": {
      "External id": 244915,
      "cbid": 211, "correlation": 244915
    }
  },
  {
    "ph": "s", "id": 244915, "pid": 76337, "tid": -914061504, "ts": 1716454225206176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225352578, "dur": 50,
    "args": {
      "External id": 244923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244923,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244923, "pid": 5, "tid": 7, "ts": 1716454225352578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206479, "dur": 9,
    "args": {
      "External id": 244923,
      "cbid": 211, "correlation": 244923
    }
  },
  {
    "ph": "s", "id": 244923, "pid": 76337, "tid": -914061504, "ts": 1716454225206479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225352629, "dur": 52,
    "args": {
      "External id": 244943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244943,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 244943, "pid": 5, "tid": 7, "ts": 1716454225352629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206557, "dur": 12,
    "args": {
      "External id": 244943,
      "cbid": 211, "correlation": 244943
    }
  },
  {
    "ph": "s", "id": 244943, "pid": 76337, "tid": -914061504, "ts": 1716454225206557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225352683, "dur": 5,
    "args": {
      "External id": 244955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244955,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244955, "pid": 5, "tid": 7, "ts": 1716454225352683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206579, "dur": 10,
    "args": {
      "External id": 244955,
      "cbid": 211, "correlation": 244955
    }
  },
  {
    "ph": "s", "id": 244955, "pid": 76337, "tid": -914061504, "ts": 1716454225206579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225352689, "dur": 56,
    "args": {
      "External id": 244958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244958, "pid": 5, "tid": 7, "ts": 1716454225352689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206601, "dur": 105,
    "args": {
      "External id": 244958,
      "cbid": 211, "correlation": 244958
    }
  },
  {
    "ph": "s", "id": 244958, "pid": 76337, "tid": -914061504, "ts": 1716454225206601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225352746, "dur": 38,
    "args": {
      "External id": 244967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244967,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244967, "pid": 5, "tid": 7, "ts": 1716454225352746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206745, "dur": 10,
    "args": {
      "External id": 244967,
      "cbid": 211, "correlation": 244967
    }
  },
  {
    "ph": "s", "id": 244967, "pid": 76337, "tid": -914061504, "ts": 1716454225206745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206800, "dur": 0,
    "args": {
      "External id": 244977,
      "cbid": 317, "correlation": 244977
    }
  },
  {
    "ph": "f", "id": 244977, "pid": 76337, "tid": -914061504, "ts": 1716454225206800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206801, "dur": 0,
    "args": {
      "External id": 244978,
      "cbid": 203, "correlation": 244978
    }
  },
  {
    "ph": "f", "id": 244978, "pid": 76337, "tid": -914061504, "ts": 1716454225206801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206802, "dur": 0,
    "args": {
      "External id": 244979,
      "cbid": 205, "correlation": 244979
    }
  },
  {
    "ph": "f", "id": 244979, "pid": 76337, "tid": -914061504, "ts": 1716454225206802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225352785, "dur": 41,
    "args": {
      "External id": 244983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244983, "pid": 5, "tid": 7, "ts": 1716454225352785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206819, "dur": 12,
    "args": {
      "External id": 244983,
      "cbid": 211, "correlation": 244983
    }
  },
  {
    "ph": "s", "id": 244983, "pid": 76337, "tid": -914061504, "ts": 1716454225206819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225352828, "dur": 3,
    "args": {
      "External id": 244985,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 244985,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.203389830508474
    }
  },
  {
    "ph": "f", "id": 244985, "pid": 5, "tid": 7, "ts": 1716454225352828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206834, "dur": 17,
    "args": {
      "External id": 244985,
      "cbid": 51, "correlation": 244985
    }
  },
  {
    "ph": "s", "id": 244985, "pid": 76337, "tid": -914061504, "ts": 1716454225206834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206857, "dur": 1,
    "args": {
      "External id": 244987,
      "cbid": 200, "correlation": 244987
    }
  },
  {
    "ph": "f", "id": 244987, "pid": 76337, "tid": -914061504, "ts": 1716454225206857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206859, "dur": 0,
    "args": {
      "External id": 244988,
      "cbid": 200, "correlation": 244988
    }
  },
  {
    "ph": "f", "id": 244988, "pid": 76337, "tid": -914061504, "ts": 1716454225206859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206859, "dur": 0,
    "args": {
      "External id": 244989,
      "cbid": 200, "correlation": 244989
    }
  },
  {
    "ph": "f", "id": 244989, "pid": 76337, "tid": -914061504, "ts": 1716454225206859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206860, "dur": 0,
    "args": {
      "External id": 244990,
      "cbid": 200, "correlation": 244990
    }
  },
  {
    "ph": "f", "id": 244990, "pid": 76337, "tid": -914061504, "ts": 1716454225206860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206861, "dur": 4,
    "args": {
      "External id": 244991,
      "cbid": 15, "correlation": 244991
    }
  },
  {
    "ph": "f", "id": 244991, "pid": 76337, "tid": -914061504, "ts": 1716454225206861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206867, "dur": 1,
    "args": {
      "External id": 244992,
      "cbid": 251, "correlation": 244992
    }
  },
  {
    "ph": "f", "id": 244992, "pid": 76337, "tid": -914061504, "ts": 1716454225206867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225352832, "dur": 25,
    "args": {
      "External id": 244993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244993,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244993, "pid": 5, "tid": 7, "ts": 1716454225352832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206870, "dur": 8,
    "args": {
      "External id": 244993,
      "cbid": 211, "correlation": 244993
    }
  },
  {
    "ph": "s", "id": 244993, "pid": 76337, "tid": -914061504, "ts": 1716454225206870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225352859, "dur": 4,
    "args": {
      "External id": 244995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244995,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 244995, "pid": 5, "tid": 7, "ts": 1716454225352859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206883, "dur": 6,
    "args": {
      "External id": 244995,
      "cbid": 211, "correlation": 244995
    }
  },
  {
    "ph": "s", "id": 244995, "pid": 76337, "tid": -914061504, "ts": 1716454225206883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206892, "dur": 0,
    "args": {
      "External id": 244996,
      "cbid": 51, "correlation": 244996
    }
  },
  {
    "ph": "s", "id": 244996, "pid": 76337, "tid": -914061504, "ts": 1716454225206892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225352864, "dur": 193,
    "args": {
      "External id": 244997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244997,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 244997, "pid": 5, "tid": 7, "ts": 1716454225352864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225206893, "dur": 199,
    "args": {
      "External id": 244997,
      "cbid": 211, "correlation": 244997
    }
  },
  {
    "ph": "s", "id": 244997, "pid": 76337, "tid": -914061504, "ts": 1716454225206893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225353058, "dur": 6,
    "args": {
      "External id": 244998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 244998,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 244998, "pid": 5, "tid": 7, "ts": 1716454225353058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225207096, "dur": 6,
    "args": {
      "External id": 244998,
      "cbid": 211, "correlation": 244998
    }
  },
  {
    "ph": "s", "id": 244998, "pid": 76337, "tid": -914061504, "ts": 1716454225207096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225353066, "dur": 5,
    "args": {
      "External id": 245004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 245004, "pid": 5, "tid": 7, "ts": 1716454225353066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225207125, "dur": 10,
    "args": {
      "External id": 245004,
      "cbid": 211, "correlation": 245004
    }
  },
  {
    "ph": "s", "id": 245004, "pid": 76337, "tid": -914061504, "ts": 1716454225207125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225353072, "dur": 3,
    "args": {
      "External id": 245012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245012,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245012, "pid": 5, "tid": 7, "ts": 1716454225353072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225208784, "dur": 16,
    "args": {
      "External id": 245012,
      "cbid": 211, "correlation": 245012
    }
  },
  {
    "ph": "s", "id": 245012, "pid": 76337, "tid": -914061504, "ts": 1716454225208784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225353076, "dur": 3,
    "args": {
      "External id": 245020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245020,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245020, "pid": 5, "tid": 7, "ts": 1716454225353076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225208825, "dur": 11,
    "args": {
      "External id": 245020,
      "cbid": 211, "correlation": 245020
    }
  },
  {
    "ph": "s", "id": 245020, "pid": 76337, "tid": -914061504, "ts": 1716454225208825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225353080, "dur": 3,
    "args": {
      "External id": 245028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245028,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245028, "pid": 5, "tid": 7, "ts": 1716454225353080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225208853, "dur": 8,
    "args": {
      "External id": 245028,
      "cbid": 211, "correlation": 245028
    }
  },
  {
    "ph": "s", "id": 245028, "pid": 76337, "tid": -914061504, "ts": 1716454225208853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225353084, "dur": 3,
    "args": {
      "External id": 245037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245037,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245037, "pid": 5, "tid": 7, "ts": 1716454225353084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209033, "dur": 14,
    "args": {
      "External id": 245037,
      "cbid": 211, "correlation": 245037
    }
  },
  {
    "ph": "s", "id": 245037, "pid": 76337, "tid": -914061504, "ts": 1716454225209033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225353088, "dur": 3,
    "args": {
      "External id": 245046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245046,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245046, "pid": 5, "tid": 7, "ts": 1716454225353088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209062, "dur": 7,
    "args": {
      "External id": 245046,
      "cbid": 211, "correlation": 245046
    }
  },
  {
    "ph": "s", "id": 245046, "pid": 76337, "tid": -914061504, "ts": 1716454225209062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225353092, "dur": 3,
    "args": {
      "External id": 245054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245054,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245054, "pid": 5, "tid": 7, "ts": 1716454225353092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209088, "dur": 8,
    "args": {
      "External id": 245054,
      "cbid": 211, "correlation": 245054
    }
  },
  {
    "ph": "s", "id": 245054, "pid": 76337, "tid": -914061504, "ts": 1716454225209088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225353097, "dur": 3,
    "args": {
      "External id": 245062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245062,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245062, "pid": 5, "tid": 7, "ts": 1716454225353097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209346, "dur": 15,
    "args": {
      "External id": 245062,
      "cbid": 211, "correlation": 245062
    }
  },
  {
    "ph": "s", "id": 245062, "pid": 76337, "tid": -914061504, "ts": 1716454225209346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225353101, "dur": 3,
    "args": {
      "External id": 245070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245070,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245070, "pid": 5, "tid": 7, "ts": 1716454225353101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209377, "dur": 7,
    "args": {
      "External id": 245070,
      "cbid": 211, "correlation": 245070
    }
  },
  {
    "ph": "s", "id": 245070, "pid": 76337, "tid": -914061504, "ts": 1716454225209377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353106, "dur": 1,
    "args": {
      "External id": 245080,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245080,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 245080, "pid": 5, "tid": 7, "ts": 1716454225353106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209441, "dur": 35,
    "args": {
      "External id": 245080,
      "cbid": 41, "correlation": 245080
    }
  },
  {
    "ph": "s", "id": 245080, "pid": 76337, "tid": -914061504, "ts": 1716454225209441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225209477, "dur": 143645,
    "args": {
      "External id": 245081,
      "cbid": 131, "correlation": 245081
    }
  },
  {
    "ph": "f", "id": 245081, "pid": 76337, "tid": -914061504, "ts": 1716454225209477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225353278, "dur": 3,
    "args": {
      "External id": 245089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245089, "pid": 5, "tid": 7, "ts": 1716454225353278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353255, "dur": 25,
    "args": {
      "External id": 245089,
      "cbid": 211, "correlation": 245089
    }
  },
  {
    "ph": "s", "id": 245089, "pid": 76337, "tid": -914061504, "ts": 1716454225353255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353371, "dur": 3,
    "args": {
      "External id": 245098,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245098,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002380952380952381
    }
  },
  {
    "ph": "f", "id": 245098, "pid": 5, "tid": 7, "ts": 1716454225353371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353342, "dur": 30,
    "args": {
      "External id": 245098,
      "cbid": 41, "correlation": 245098
    }
  },
  {
    "ph": "s", "id": 245098, "pid": 76337, "tid": -914061504, "ts": 1716454225353342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225353466, "dur": 4,
    "args": {
      "External id": 245108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245108,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245108, "pid": 5, "tid": 7, "ts": 1716454225353466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353451, "dur": 17,
    "args": {
      "External id": 245108,
      "cbid": 211, "correlation": 245108
    }
  },
  {
    "ph": "s", "id": 245108, "pid": 76337, "tid": -914061504, "ts": 1716454225353451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353538, "dur": 1,
    "args": {
      "External id": 245118,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245118,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 245118, "pid": 5, "tid": 7, "ts": 1716454225353538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353520, "dur": 16,
    "args": {
      "External id": 245118,
      "cbid": 41, "correlation": 245118
    }
  },
  {
    "ph": "s", "id": 245118, "pid": 76337, "tid": -914061504, "ts": 1716454225353520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353537, "dur": 8,
    "args": {
      "External id": 245119,
      "cbid": 131, "correlation": 245119
    }
  },
  {
    "ph": "f", "id": 245119, "pid": 76337, "tid": -914061504, "ts": 1716454225353537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353603, "dur": 3,
    "args": {
      "External id": 245126,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245126,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.11764705882353
    }
  },
  {
    "ph": "f", "id": 245126, "pid": 5, "tid": 7, "ts": 1716454225353603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353583, "dur": 19,
    "args": {
      "External id": 245126,
      "cbid": 41, "correlation": 245126
    }
  },
  {
    "ph": "s", "id": 245126, "pid": 76337, "tid": -914061504, "ts": 1716454225353583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353691, "dur": 3,
    "args": {
      "External id": 245145,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245145,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005263157894736842
    }
  },
  {
    "ph": "f", "id": 245145, "pid": 5, "tid": 7, "ts": 1716454225353691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353671, "dur": 19,
    "args": {
      "External id": 245145,
      "cbid": 41, "correlation": 245145
    }
  },
  {
    "ph": "s", "id": 245145, "pid": 76337, "tid": -914061504, "ts": 1716454225353671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225353730, "dur": 3,
    "args": {
      "External id": 245151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245151, "pid": 5, "tid": 7, "ts": 1716454225353730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353719, "dur": 11,
    "args": {
      "External id": 245151,
      "cbid": 211, "correlation": 245151
    }
  },
  {
    "ph": "s", "id": 245151, "pid": 76337, "tid": -914061504, "ts": 1716454225353719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225353745, "dur": 6,
    "args": {
      "External id": 245153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245153,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 245153, "pid": 5, "tid": 7, "ts": 1716454225353745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353734, "dur": 10,
    "args": {
      "External id": 245153,
      "cbid": 211, "correlation": 245153
    }
  },
  {
    "ph": "s", "id": 245153, "pid": 76337, "tid": -914061504, "ts": 1716454225353734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454225353753, "dur": 3,
    "args": {
      "External id": 245155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245155, "pid": 5, "tid": 7, "ts": 1716454225353753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353745, "dur": 7,
    "args": {
      "External id": 245155,
      "cbid": 211, "correlation": 245155
    }
  },
  {
    "ph": "s", "id": 245155, "pid": 76337, "tid": -914061504, "ts": 1716454225353745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353787, "dur": 2,
    "args": {
      "External id": 245163,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245163,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 245163, "pid": 5, "tid": 7, "ts": 1716454225353787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353772, "dur": 13,
    "args": {
      "External id": 245163,
      "cbid": 41, "correlation": 245163
    }
  },
  {
    "ph": "s", "id": 245163, "pid": 76337, "tid": -914061504, "ts": 1716454225353772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225353835, "dur": 3,
    "args": {
      "External id": 245177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245177, "pid": 5, "tid": 7, "ts": 1716454225353835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353823, "dur": 13,
    "args": {
      "External id": 245177,
      "cbid": 211, "correlation": 245177
    }
  },
  {
    "ph": "s", "id": 245177, "pid": 76337, "tid": -914061504, "ts": 1716454225353823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225353855, "dur": 2,
    "args": {
      "External id": 245191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245191, "pid": 5, "tid": 7, "ts": 1716454225353855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353847, "dur": 6,
    "args": {
      "External id": 245191,
      "cbid": 211, "correlation": 245191
    }
  },
  {
    "ph": "s", "id": 245191, "pid": 76337, "tid": -914061504, "ts": 1716454225353847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225353888, "dur": 6,
    "args": {
      "External id": 245198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245198,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245198, "pid": 5, "tid": 7, "ts": 1716454225353888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353877, "dur": 12,
    "args": {
      "External id": 245198,
      "cbid": 211, "correlation": 245198
    }
  },
  {
    "ph": "s", "id": 245198, "pid": 76337, "tid": -914061504, "ts": 1716454225353877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225353898, "dur": 6,
    "args": {
      "External id": 245201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245201,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245201, "pid": 5, "tid": 7, "ts": 1716454225353898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353891, "dur": 6,
    "args": {
      "External id": 245201,
      "cbid": 211, "correlation": 245201
    }
  },
  {
    "ph": "s", "id": 245201, "pid": 76337, "tid": -914061504, "ts": 1716454225353891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225353907, "dur": 3,
    "args": {
      "External id": 245203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245203,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245203, "pid": 5, "tid": 7, "ts": 1716454225353907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353899, "dur": 6,
    "args": {
      "External id": 245203,
      "cbid": 211, "correlation": 245203
    }
  },
  {
    "ph": "s", "id": 245203, "pid": 76337, "tid": -914061504, "ts": 1716454225353899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225353927, "dur": 2,
    "args": {
      "External id": 245206,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245206,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 245206, "pid": 5, "tid": 7, "ts": 1716454225353927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353914, "dur": 11,
    "args": {
      "External id": 245206,
      "cbid": 41, "correlation": 245206
    }
  },
  {
    "ph": "s", "id": 245206, "pid": 76337, "tid": -914061504, "ts": 1716454225353914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225353989, "dur": 4,
    "args": {
      "External id": 245222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245222,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245222, "pid": 5, "tid": 7, "ts": 1716454225353989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353966, "dur": 23,
    "args": {
      "External id": 245222,
      "cbid": 211, "correlation": 245222
    }
  },
  {
    "ph": "s", "id": 245222, "pid": 76337, "tid": -914061504, "ts": 1716454225353966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225354010, "dur": 3,
    "args": {
      "External id": 245227,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245227,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003094059405940594
    }
  },
  {
    "ph": "f", "id": 245227, "pid": 5, "tid": 7, "ts": 1716454225354010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225353995, "dur": 15,
    "args": {
      "External id": 245227,
      "cbid": 41, "correlation": 245227
    }
  },
  {
    "ph": "s", "id": 245227, "pid": 76337, "tid": -914061504, "ts": 1716454225353995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225354038, "dur": 1,
    "args": {
      "External id": 245233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245233,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 245233, "pid": 5, "tid": 7, "ts": 1716454225354038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354019, "dur": 27,
    "args": {
      "External id": 245233,
      "cbid": 41, "correlation": 245233
    }
  },
  {
    "ph": "s", "id": 245233, "pid": 76337, "tid": -914061504, "ts": 1716454225354019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354047, "dur": 3,
    "args": {
      "External id": 245234,
      "cbid": 131, "correlation": 245234
    }
  },
  {
    "ph": "f", "id": 245234, "pid": 76337, "tid": -914061504, "ts": 1716454225354047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225354098, "dur": 3,
    "args": {
      "External id": 245242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245242,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245242, "pid": 5, "tid": 7, "ts": 1716454225354098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354086, "dur": 12,
    "args": {
      "External id": 245242,
      "cbid": 211, "correlation": 245242
    }
  },
  {
    "ph": "s", "id": 245242, "pid": 76337, "tid": -914061504, "ts": 1716454225354086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225354133, "dur": 3,
    "args": {
      "External id": 245252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245252,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245252, "pid": 5, "tid": 7, "ts": 1716454225354133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354119, "dur": 13,
    "args": {
      "External id": 245252,
      "cbid": 211, "correlation": 245252
    }
  },
  {
    "ph": "s", "id": 245252, "pid": 76337, "tid": -914061504, "ts": 1716454225354119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225354157, "dur": 3,
    "args": {
      "External id": 245261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245261,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245261, "pid": 5, "tid": 7, "ts": 1716454225354157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354148, "dur": 9,
    "args": {
      "External id": 245261,
      "cbid": 211, "correlation": 245261
    }
  },
  {
    "ph": "s", "id": 245261, "pid": 76337, "tid": -914061504, "ts": 1716454225354148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225354269, "dur": 12,
    "args": {
      "External id": 245271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245271,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245271, "pid": 5, "tid": 7, "ts": 1716454225354269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354255, "dur": 15,
    "args": {
      "External id": 245271,
      "cbid": 211, "correlation": 245271
    }
  },
  {
    "ph": "s", "id": 245271, "pid": 76337, "tid": -914061504, "ts": 1716454225354255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225354309, "dur": 3,
    "args": {
      "External id": 245279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245279,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245279, "pid": 5, "tid": 7, "ts": 1716454225354309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354300, "dur": 8,
    "args": {
      "External id": 245279,
      "cbid": 211, "correlation": 245279
    }
  },
  {
    "ph": "s", "id": 245279, "pid": 76337, "tid": -914061504, "ts": 1716454225354300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225354353, "dur": 12,
    "args": {
      "External id": 245289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245289,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245289, "pid": 5, "tid": 7, "ts": 1716454225354353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354342, "dur": 11,
    "args": {
      "External id": 245289,
      "cbid": 211, "correlation": 245289
    }
  },
  {
    "ph": "s", "id": 245289, "pid": 76337, "tid": -914061504, "ts": 1716454225354342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225354384, "dur": 10,
    "args": {
      "External id": 245297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245297,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245297, "pid": 5, "tid": 7, "ts": 1716454225354384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354374, "dur": 9,
    "args": {
      "External id": 245297,
      "cbid": 211, "correlation": 245297
    }
  },
  {
    "ph": "s", "id": 245297, "pid": 76337, "tid": -914061504, "ts": 1716454225354374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225354411, "dur": 3,
    "args": {
      "External id": 245306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245306, "pid": 5, "tid": 7, "ts": 1716454225354411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354401, "dur": 10,
    "args": {
      "External id": 245306,
      "cbid": 211, "correlation": 245306
    }
  },
  {
    "ph": "s", "id": 245306, "pid": 76337, "tid": -914061504, "ts": 1716454225354401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225354435, "dur": 5,
    "args": {
      "External id": 245315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245315,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245315, "pid": 5, "tid": 7, "ts": 1716454225354435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354427, "dur": 8,
    "args": {
      "External id": 245315,
      "cbid": 211, "correlation": 245315
    }
  },
  {
    "ph": "s", "id": 245315, "pid": 76337, "tid": -914061504, "ts": 1716454225354427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225354475, "dur": 8,
    "args": {
      "External id": 245325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245325,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245325, "pid": 5, "tid": 7, "ts": 1716454225354475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354464, "dur": 11,
    "args": {
      "External id": 245325,
      "cbid": 211, "correlation": 245325
    }
  },
  {
    "ph": "s", "id": 245325, "pid": 76337, "tid": -914061504, "ts": 1716454225354464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225354795, "dur": 3,
    "args": {
      "External id": 245334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245334,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245334, "pid": 5, "tid": 7, "ts": 1716454225354795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354779, "dur": 15,
    "args": {
      "External id": 245334,
      "cbid": 211, "correlation": 245334
    }
  },
  {
    "ph": "s", "id": 245334, "pid": 76337, "tid": -914061504, "ts": 1716454225354779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225354822, "dur": 3,
    "args": {
      "External id": 245342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245342,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245342, "pid": 5, "tid": 7, "ts": 1716454225354822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354813, "dur": 8,
    "args": {
      "External id": 245342,
      "cbid": 211, "correlation": 245342
    }
  },
  {
    "ph": "s", "id": 245342, "pid": 76337, "tid": -914061504, "ts": 1716454225354813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225354875, "dur": 1,
    "args": {
      "External id": 245352,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245352,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 245352, "pid": 5, "tid": 7, "ts": 1716454225354875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354860, "dur": 13,
    "args": {
      "External id": 245352,
      "cbid": 41, "correlation": 245352
    }
  },
  {
    "ph": "s", "id": 245352, "pid": 76337, "tid": -914061504, "ts": 1716454225354860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354874, "dur": 8,
    "args": {
      "External id": 245353,
      "cbid": 131, "correlation": 245353
    }
  },
  {
    "ph": "f", "id": 245353, "pid": 76337, "tid": -914061504, "ts": 1716454225354874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225354967, "dur": 3,
    "args": {
      "External id": 245361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245361, "pid": 5, "tid": 7, "ts": 1716454225354967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225354951, "dur": 16,
    "args": {
      "External id": 245361,
      "cbid": 211, "correlation": 245361
    }
  },
  {
    "ph": "s", "id": 245361, "pid": 76337, "tid": -914061504, "ts": 1716454225354951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225355049, "dur": 3,
    "args": {
      "External id": 245370,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245370,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002577319587628866
    }
  },
  {
    "ph": "f", "id": 245370, "pid": 5, "tid": 7, "ts": 1716454225355049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355031, "dur": 18,
    "args": {
      "External id": 245370,
      "cbid": 41, "correlation": 245370
    }
  },
  {
    "ph": "s", "id": 245370, "pid": 76337, "tid": -914061504, "ts": 1716454225355031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225355122, "dur": 3,
    "args": {
      "External id": 245380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245380,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245380, "pid": 5, "tid": 7, "ts": 1716454225355122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355108, "dur": 14,
    "args": {
      "External id": 245380,
      "cbid": 211, "correlation": 245380
    }
  },
  {
    "ph": "s", "id": 245380, "pid": 76337, "tid": -914061504, "ts": 1716454225355108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225355175, "dur": 1,
    "args": {
      "External id": 245390,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245390,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005211726384364821
    }
  },
  {
    "ph": "f", "id": 245390, "pid": 5, "tid": 7, "ts": 1716454225355175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355161, "dur": 12,
    "args": {
      "External id": 245390,
      "cbid": 41, "correlation": 245390
    }
  },
  {
    "ph": "s", "id": 245390, "pid": 76337, "tid": -914061504, "ts": 1716454225355161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355174, "dur": 8,
    "args": {
      "External id": 245391,
      "cbid": 131, "correlation": 245391
    }
  },
  {
    "ph": "f", "id": 245391, "pid": 76337, "tid": -914061504, "ts": 1716454225355174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225355247, "dur": 3,
    "args": {
      "External id": 245398,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245398,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 245398, "pid": 5, "tid": 7, "ts": 1716454225355247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355217, "dur": 29,
    "args": {
      "External id": 245398,
      "cbid": 41, "correlation": 245398
    }
  },
  {
    "ph": "s", "id": 245398, "pid": 76337, "tid": -914061504, "ts": 1716454225355217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225355296, "dur": 1,
    "args": {
      "External id": 245409,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245409,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 245409, "pid": 5, "tid": 7, "ts": 1716454225355296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355284, "dur": 10,
    "args": {
      "External id": 245409,
      "cbid": 41, "correlation": 245409
    }
  },
  {
    "ph": "s", "id": 245409, "pid": 76337, "tid": -914061504, "ts": 1716454225355284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355294, "dur": 8,
    "args": {
      "External id": 245410,
      "cbid": 131, "correlation": 245410
    }
  },
  {
    "ph": "f", "id": 245410, "pid": 76337, "tid": -914061504, "ts": 1716454225355294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225355345, "dur": 3,
    "args": {
      "External id": 245418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245418,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245418, "pid": 5, "tid": 7, "ts": 1716454225355345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355332, "dur": 13,
    "args": {
      "External id": 245418,
      "cbid": 211, "correlation": 245418
    }
  },
  {
    "ph": "s", "id": 245418, "pid": 76337, "tid": -914061504, "ts": 1716454225355332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225355374, "dur": 3,
    "args": {
      "External id": 245428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245428,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245428, "pid": 5, "tid": 7, "ts": 1716454225355374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355365, "dur": 8,
    "args": {
      "External id": 245428,
      "cbid": 211, "correlation": 245428
    }
  },
  {
    "ph": "s", "id": 245428, "pid": 76337, "tid": -914061504, "ts": 1716454225355365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225355396, "dur": 3,
    "args": {
      "External id": 245437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245437,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245437, "pid": 5, "tid": 7, "ts": 1716454225355396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355388, "dur": 7,
    "args": {
      "External id": 245437,
      "cbid": 211, "correlation": 245437
    }
  },
  {
    "ph": "s", "id": 245437, "pid": 76337, "tid": -914061504, "ts": 1716454225355388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225355465, "dur": 6,
    "args": {
      "External id": 245445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245445, "pid": 5, "tid": 7, "ts": 1716454225355465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355452, "dur": 13,
    "args": {
      "External id": 245445,
      "cbid": 211, "correlation": 245445
    }
  },
  {
    "ph": "s", "id": 245445, "pid": 76337, "tid": -914061504, "ts": 1716454225355452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225355504, "dur": 3,
    "args": {
      "External id": 245454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245454,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245454, "pid": 5, "tid": 7, "ts": 1716454225355504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355493, "dur": 10,
    "args": {
      "External id": 245454,
      "cbid": 211, "correlation": 245454
    }
  },
  {
    "ph": "s", "id": 245454, "pid": 76337, "tid": -914061504, "ts": 1716454225355493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225355527, "dur": 3,
    "args": {
      "External id": 245463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245463,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245463, "pid": 5, "tid": 7, "ts": 1716454225355527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355518, "dur": 7,
    "args": {
      "External id": 245463,
      "cbid": 211, "correlation": 245463
    }
  },
  {
    "ph": "s", "id": 245463, "pid": 76337, "tid": -914061504, "ts": 1716454225355518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225355588, "dur": 3,
    "args": {
      "External id": 245471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245471,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245471, "pid": 5, "tid": 7, "ts": 1716454225355588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355577, "dur": 10,
    "args": {
      "External id": 245471,
      "cbid": 211, "correlation": 245471
    }
  },
  {
    "ph": "s", "id": 245471, "pid": 76337, "tid": -914061504, "ts": 1716454225355577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225355647, "dur": 2,
    "args": {
      "External id": 245479,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245479,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00390625
    }
  },
  {
    "ph": "f", "id": 245479, "pid": 5, "tid": 7, "ts": 1716454225355647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355631, "dur": 26,
    "args": {
      "External id": 245479,
      "cbid": 41, "correlation": 245479
    }
  },
  {
    "ph": "s", "id": 245479, "pid": 76337, "tid": -914061504, "ts": 1716454225355631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355658, "dur": 3,
    "args": {
      "External id": 245480,
      "cbid": 131, "correlation": 245480
    }
  },
  {
    "ph": "f", "id": 245480, "pid": 76337, "tid": -914061504, "ts": 1716454225355658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225355718, "dur": 1,
    "args": {
      "External id": 245490,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245490,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 245490, "pid": 5, "tid": 7, "ts": 1716454225355718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355706, "dur": 10,
    "args": {
      "External id": 245490,
      "cbid": 41, "correlation": 245490
    }
  },
  {
    "ph": "s", "id": 245490, "pid": 76337, "tid": -914061504, "ts": 1716454225355706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355717, "dur": 8,
    "args": {
      "External id": 245491,
      "cbid": 131, "correlation": 245491
    }
  },
  {
    "ph": "f", "id": 245491, "pid": 76337, "tid": -914061504, "ts": 1716454225355717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225355773, "dur": 1,
    "args": {
      "External id": 245500,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245500,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 245500, "pid": 5, "tid": 7, "ts": 1716454225355773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355762, "dur": 8,
    "args": {
      "External id": 245500,
      "cbid": 41, "correlation": 245500
    }
  },
  {
    "ph": "s", "id": 245500, "pid": 76337, "tid": -914061504, "ts": 1716454225355762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355771, "dur": 8,
    "args": {
      "External id": 245501,
      "cbid": 131, "correlation": 245501
    }
  },
  {
    "ph": "f", "id": 245501, "pid": 76337, "tid": -914061504, "ts": 1716454225355771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225355847, "dur": 4,
    "args": {
      "External id": 245508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245508, "pid": 5, "tid": 7, "ts": 1716454225355847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355828, "dur": 19,
    "args": {
      "External id": 245508,
      "cbid": 211, "correlation": 245508
    }
  },
  {
    "ph": "s", "id": 245508, "pid": 76337, "tid": -914061504, "ts": 1716454225355828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454225355884, "dur": 4,
    "args": {
      "External id": 245528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245528,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245528, "pid": 5, "tid": 7, "ts": 1716454225355884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355872, "dur": 12,
    "args": {
      "External id": 245528,
      "cbid": 211, "correlation": 245528
    }
  },
  {
    "ph": "s", "id": 245528, "pid": 76337, "tid": -914061504, "ts": 1716454225355872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355885, "dur": 0,
    "args": {
      "External id": 245529,
      "cbid": 11, "correlation": 245529
    }
  },
  {
    "ph": "f", "id": 245529, "pid": 76337, "tid": -914061504, "ts": 1716454225355885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355886, "dur": 0,
    "args": {
      "External id": 245530,
      "cbid": 11, "correlation": 245530
    }
  },
  {
    "ph": "f", "id": 245530, "pid": 76337, "tid": -914061504, "ts": 1716454225355886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225355900, "dur": 1,
    "args": {
      "External id": 245533,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245533,
      "bytes": 4, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 245533, "pid": 5, "tid": 7, "ts": 1716454225355900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355887, "dur": 22,
    "args": {
      "External id": 245533,
      "cbid": 41, "correlation": 245533
    }
  },
  {
    "ph": "s", "id": 245533, "pid": 76337, "tid": -914061504, "ts": 1716454225355887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355910, "dur": 3,
    "args": {
      "External id": 245534,
      "cbid": 131, "correlation": 245534
    }
  },
  {
    "ph": "f", "id": 245534, "pid": 76337, "tid": -914061504, "ts": 1716454225355910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225355938, "dur": 3,
    "args": {
      "External id": 245558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245558, "pid": 5, "tid": 7, "ts": 1716454225355938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355928, "dur": 10,
    "args": {
      "External id": 245558,
      "cbid": 211, "correlation": 245558
    }
  },
  {
    "ph": "s", "id": 245558, "pid": 76337, "tid": -914061504, "ts": 1716454225355928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355939, "dur": 0,
    "args": {
      "External id": 245559,
      "cbid": 11, "correlation": 245559
    }
  },
  {
    "ph": "f", "id": 245559, "pid": 76337, "tid": -914061504, "ts": 1716454225355939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355939, "dur": 0,
    "args": {
      "External id": 245560,
      "cbid": 11, "correlation": 245560
    }
  },
  {
    "ph": "f", "id": 245560, "pid": 76337, "tid": -914061504, "ts": 1716454225355939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355941, "dur": 1,
    "args": {
      "External id": 245562,
      "cbid": 200, "correlation": 245562
    }
  },
  {
    "ph": "f", "id": 245562, "pid": 76337, "tid": -914061504, "ts": 1716454225355941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225355952, "dur": 4,
    "args": {
      "External id": 245564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245564,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245564, "pid": 5, "tid": 7, "ts": 1716454225355952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355944, "dur": 8,
    "args": {
      "External id": 245564,
      "cbid": 211, "correlation": 245564
    }
  },
  {
    "ph": "s", "id": 245564, "pid": 76337, "tid": -914061504, "ts": 1716454225355944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355953, "dur": 0,
    "args": {
      "External id": 245565,
      "cbid": 11, "correlation": 245565
    }
  },
  {
    "ph": "f", "id": 245565, "pid": 76337, "tid": -914061504, "ts": 1716454225355953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355953, "dur": 0,
    "args": {
      "External id": 245566,
      "cbid": 11, "correlation": 245566
    }
  },
  {
    "ph": "f", "id": 245566, "pid": 76337, "tid": -914061504, "ts": 1716454225355953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225355999, "dur": 1,
    "args": {
      "External id": 245573,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245573,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 245573, "pid": 5, "tid": 7, "ts": 1716454225355999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225355987, "dur": 20,
    "args": {
      "External id": 245573,
      "cbid": 41, "correlation": 245573
    }
  },
  {
    "ph": "s", "id": 245573, "pid": 76337, "tid": -914061504, "ts": 1716454225355987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356008, "dur": 3,
    "args": {
      "External id": 245574,
      "cbid": 131, "correlation": 245574
    }
  },
  {
    "ph": "f", "id": 245574, "pid": 76337, "tid": -914061504, "ts": 1716454225356008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225356059, "dur": 1,
    "args": {
      "External id": 245584,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 245584,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 245584, "pid": 5, "tid": 7, "ts": 1716454225356059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356048, "dur": 9,
    "args": {
      "External id": 245584,
      "cbid": 41, "correlation": 245584
    }
  },
  {
    "ph": "s", "id": 245584, "pid": 76337, "tid": -914061504, "ts": 1716454225356048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356058, "dur": 8,
    "args": {
      "External id": 245585,
      "cbid": 131, "correlation": 245585
    }
  },
  {
    "ph": "f", "id": 245585, "pid": 76337, "tid": -914061504, "ts": 1716454225356058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225356129, "dur": 5,
    "args": {
      "External id": 245592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245592,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245592, "pid": 5, "tid": 7, "ts": 1716454225356129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356114, "dur": 16,
    "args": {
      "External id": 245592,
      "cbid": 211, "correlation": 245592
    }
  },
  {
    "ph": "s", "id": 245592, "pid": 76337, "tid": -914061504, "ts": 1716454225356114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225356197, "dur": 3,
    "args": {
      "External id": 245601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245601,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245601, "pid": 5, "tid": 7, "ts": 1716454225356197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356185, "dur": 12,
    "args": {
      "External id": 245601,
      "cbid": 211, "correlation": 245601
    }
  },
  {
    "ph": "s", "id": 245601, "pid": 76337, "tid": -914061504, "ts": 1716454225356185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225356234, "dur": 3,
    "args": {
      "External id": 245609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245609,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245609, "pid": 5, "tid": 7, "ts": 1716454225356234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356223, "dur": 10,
    "args": {
      "External id": 245609,
      "cbid": 211, "correlation": 245609
    }
  },
  {
    "ph": "s", "id": 245609, "pid": 76337, "tid": -914061504, "ts": 1716454225356223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225356265, "dur": 4,
    "args": {
      "External id": 245617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245617,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245617, "pid": 5, "tid": 7, "ts": 1716454225356265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356254, "dur": 11,
    "args": {
      "External id": 245617,
      "cbid": 211, "correlation": 245617
    }
  },
  {
    "ph": "s", "id": 245617, "pid": 76337, "tid": -914061504, "ts": 1716454225356254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225356295, "dur": 4,
    "args": {
      "External id": 245625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245625,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245625, "pid": 5, "tid": 7, "ts": 1716454225356295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356284, "dur": 9,
    "args": {
      "External id": 245625,
      "cbid": 211, "correlation": 245625
    }
  },
  {
    "ph": "s", "id": 245625, "pid": 76337, "tid": -914061504, "ts": 1716454225356284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225356321, "dur": 3,
    "args": {
      "External id": 245633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245633,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245633, "pid": 5, "tid": 7, "ts": 1716454225356321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356312, "dur": 8,
    "args": {
      "External id": 245633,
      "cbid": 211, "correlation": 245633
    }
  },
  {
    "ph": "s", "id": 245633, "pid": 76337, "tid": -914061504, "ts": 1716454225356312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225356346, "dur": 3,
    "args": {
      "External id": 245641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245641,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245641, "pid": 5, "tid": 7, "ts": 1716454225356346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356337, "dur": 8,
    "args": {
      "External id": 245641,
      "cbid": 211, "correlation": 245641
    }
  },
  {
    "ph": "s", "id": 245641, "pid": 76337, "tid": -914061504, "ts": 1716454225356337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225356368, "dur": 4,
    "args": {
      "External id": 245649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245649,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245649, "pid": 5, "tid": 7, "ts": 1716454225356368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356359, "dur": 7,
    "args": {
      "External id": 245649,
      "cbid": 211, "correlation": 245649
    }
  },
  {
    "ph": "s", "id": 245649, "pid": 76337, "tid": -914061504, "ts": 1716454225356359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225356388, "dur": 4,
    "args": {
      "External id": 245657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245657,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245657, "pid": 5, "tid": 7, "ts": 1716454225356388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356379, "dur": 7,
    "args": {
      "External id": 245657,
      "cbid": 211, "correlation": 245657
    }
  },
  {
    "ph": "s", "id": 245657, "pid": 76337, "tid": -914061504, "ts": 1716454225356379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225356406, "dur": 3,
    "args": {
      "External id": 245665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245665,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245665, "pid": 5, "tid": 7, "ts": 1716454225356406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356398, "dur": 7,
    "args": {
      "External id": 245665,
      "cbid": 211, "correlation": 245665
    }
  },
  {
    "ph": "s", "id": 245665, "pid": 76337, "tid": -914061504, "ts": 1716454225356398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225356465, "dur": 3,
    "args": {
      "External id": 245673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245673,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 245673, "pid": 5, "tid": 7, "ts": 1716454225356465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356454, "dur": 10,
    "args": {
      "External id": 245673,
      "cbid": 211, "correlation": 245673
    }
  },
  {
    "ph": "s", "id": 245673, "pid": 76337, "tid": -914061504, "ts": 1716454225356454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225356492, "dur": 4,
    "args": {
      "External id": 245681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245681,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245681, "pid": 5, "tid": 7, "ts": 1716454225356492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356482, "dur": 8,
    "args": {
      "External id": 245681,
      "cbid": 211, "correlation": 245681
    }
  },
  {
    "ph": "s", "id": 245681, "pid": 76337, "tid": -914061504, "ts": 1716454225356482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225356515, "dur": 4,
    "args": {
      "External id": 245689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245689, "pid": 5, "tid": 7, "ts": 1716454225356515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356506, "dur": 7,
    "args": {
      "External id": 245689,
      "cbid": 211, "correlation": 245689
    }
  },
  {
    "ph": "s", "id": 245689, "pid": 76337, "tid": -914061504, "ts": 1716454225356506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225356534, "dur": 3,
    "args": {
      "External id": 245697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245697,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 245697, "pid": 5, "tid": 7, "ts": 1716454225356534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356527, "dur": 6,
    "args": {
      "External id": 245697,
      "cbid": 211, "correlation": 245697
    }
  },
  {
    "ph": "s", "id": 245697, "pid": 76337, "tid": -914061504, "ts": 1716454225356527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225356952, "dur": 6,
    "args": {
      "External id": 245706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245706,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245706, "pid": 5, "tid": 7, "ts": 1716454225356952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356936, "dur": 17,
    "args": {
      "External id": 245706,
      "cbid": 211, "correlation": 245706
    }
  },
  {
    "ph": "s", "id": 245706, "pid": 76337, "tid": -914061504, "ts": 1716454225356936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225356997, "dur": 5,
    "args": {
      "External id": 245715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245715,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245715, "pid": 5, "tid": 7, "ts": 1716454225356997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225356987, "dur": 10,
    "args": {
      "External id": 245715,
      "cbid": 211, "correlation": 245715
    }
  },
  {
    "ph": "s", "id": 245715, "pid": 76337, "tid": -914061504, "ts": 1716454225356987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225357125, "dur": 3,
    "args": {
      "External id": 245731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245731, "pid": 5, "tid": 7, "ts": 1716454225357125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357111, "dur": 14,
    "args": {
      "External id": 245731,
      "cbid": 211, "correlation": 245731
    }
  },
  {
    "ph": "s", "id": 245731, "pid": 76337, "tid": -914061504, "ts": 1716454225357111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357159, "dur": 3,
    "args": {
      "External id": 245739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245739, "pid": 5, "tid": 7, "ts": 1716454225357159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357150, "dur": 8,
    "args": {
      "External id": 245739,
      "cbid": 211, "correlation": 245739
    }
  },
  {
    "ph": "s", "id": 245739, "pid": 76337, "tid": -914061504, "ts": 1716454225357150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357190, "dur": 3,
    "args": {
      "External id": 245747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245747, "pid": 5, "tid": 7, "ts": 1716454225357190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357180, "dur": 9,
    "args": {
      "External id": 245747,
      "cbid": 211, "correlation": 245747
    }
  },
  {
    "ph": "s", "id": 245747, "pid": 76337, "tid": -914061504, "ts": 1716454225357180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357221, "dur": 4,
    "args": {
      "External id": 245755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245755,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245755, "pid": 5, "tid": 7, "ts": 1716454225357221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357212, "dur": 8,
    "args": {
      "External id": 245755,
      "cbid": 211, "correlation": 245755
    }
  },
  {
    "ph": "s", "id": 245755, "pid": 76337, "tid": -914061504, "ts": 1716454225357212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225357277, "dur": 4,
    "args": {
      "External id": 245767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245767,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245767, "pid": 5, "tid": 7, "ts": 1716454225357277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357264, "dur": 13,
    "args": {
      "External id": 245767,
      "cbid": 211, "correlation": 245767
    }
  },
  {
    "ph": "s", "id": 245767, "pid": 76337, "tid": -914061504, "ts": 1716454225357264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225357322, "dur": 4,
    "args": {
      "External id": 245778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245778, "pid": 5, "tid": 7, "ts": 1716454225357322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357311, "dur": 11,
    "args": {
      "External id": 245778,
      "cbid": 211, "correlation": 245778
    }
  },
  {
    "ph": "s", "id": 245778, "pid": 76337, "tid": -914061504, "ts": 1716454225357311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357354, "dur": 3,
    "args": {
      "External id": 245786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245786, "pid": 5, "tid": 7, "ts": 1716454225357354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357345, "dur": 9,
    "args": {
      "External id": 245786,
      "cbid": 211, "correlation": 245786
    }
  },
  {
    "ph": "s", "id": 245786, "pid": 76337, "tid": -914061504, "ts": 1716454225357345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357386, "dur": 5,
    "args": {
      "External id": 245794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245794,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245794, "pid": 5, "tid": 7, "ts": 1716454225357386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357377, "dur": 10,
    "args": {
      "External id": 245794,
      "cbid": 211, "correlation": 245794
    }
  },
  {
    "ph": "s", "id": 245794, "pid": 76337, "tid": -914061504, "ts": 1716454225357377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357417, "dur": 5,
    "args": {
      "External id": 245802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245802,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245802, "pid": 5, "tid": 7, "ts": 1716454225357417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357406, "dur": 9,
    "args": {
      "External id": 245802,
      "cbid": 211, "correlation": 245802
    }
  },
  {
    "ph": "s", "id": 245802, "pid": 76337, "tid": -914061504, "ts": 1716454225357406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225357447, "dur": 4,
    "args": {
      "External id": 245811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245811,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245811, "pid": 5, "tid": 7, "ts": 1716454225357447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357437, "dur": 10,
    "args": {
      "External id": 245811,
      "cbid": 211, "correlation": 245811
    }
  },
  {
    "ph": "s", "id": 245811, "pid": 76337, "tid": -914061504, "ts": 1716454225357437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225357508, "dur": 4,
    "args": {
      "External id": 245824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245824,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245824, "pid": 5, "tid": 7, "ts": 1716454225357508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357495, "dur": 13,
    "args": {
      "External id": 245824,
      "cbid": 211, "correlation": 245824
    }
  },
  {
    "ph": "s", "id": 245824, "pid": 76337, "tid": -914061504, "ts": 1716454225357495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225357548, "dur": 5,
    "args": {
      "External id": 245834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245834,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 245834, "pid": 5, "tid": 7, "ts": 1716454225357548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357538, "dur": 10,
    "args": {
      "External id": 245834,
      "cbid": 211, "correlation": 245834
    }
  },
  {
    "ph": "s", "id": 245834, "pid": 76337, "tid": -914061504, "ts": 1716454225357538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357672, "dur": 4,
    "args": {
      "External id": 245851,
      "cbid": 251, "correlation": 245851
    }
  },
  {
    "ph": "f", "id": 245851, "pid": 76337, "tid": -914061504, "ts": 1716454225357672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454225357699, "dur": 11,
    "args": {
      "External id": 245853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245853,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 245853, "pid": 5, "tid": 7, "ts": 1716454225357699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357684, "dur": 16,
    "args": {
      "External id": 245853,
      "cbid": 211, "correlation": 245853
    }
  },
  {
    "ph": "s", "id": 245853, "pid": 76337, "tid": -914061504, "ts": 1716454225357684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225357759, "dur": 4,
    "args": {
      "External id": 245861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245861,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 245861, "pid": 5, "tid": 7, "ts": 1716454225357759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357747, "dur": 12,
    "args": {
      "External id": 245861,
      "cbid": 211, "correlation": 245861
    }
  },
  {
    "ph": "s", "id": 245861, "pid": 76337, "tid": -914061504, "ts": 1716454225357747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357816, "dur": 2,
    "args": {
      "External id": 245877,
      "cbid": 251, "correlation": 245877
    }
  },
  {
    "ph": "f", "id": 245877, "pid": 76337, "tid": -914061504, "ts": 1716454225357816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357823, "dur": 0,
    "args": {
      "External id": 245879,
      "cbid": 251, "correlation": 245879
    }
  },
  {
    "ph": "f", "id": 245879, "pid": 76337, "tid": -914061504, "ts": 1716454225357823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225357839, "dur": 14,
    "args": {
      "External id": 245880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245880,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 245880, "pid": 5, "tid": 7, "ts": 1716454225357839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357826, "dur": 14,
    "args": {
      "External id": 245880,
      "cbid": 211, "correlation": 245880
    }
  },
  {
    "ph": "s", "id": 245880, "pid": 76337, "tid": -914061504, "ts": 1716454225357826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225357855, "dur": 5,
    "args": {
      "External id": 245882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245882,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 245882, "pid": 5, "tid": 7, "ts": 1716454225357855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357843, "dur": 9,
    "args": {
      "External id": 245882,
      "cbid": 211, "correlation": 245882
    }
  },
  {
    "ph": "s", "id": 245882, "pid": 76337, "tid": -914061504, "ts": 1716454225357843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357954, "dur": 1,
    "args": {
      "External id": 245892,
      "cbid": 317, "correlation": 245892
    }
  },
  {
    "ph": "f", "id": 245892, "pid": 76337, "tid": -914061504, "ts": 1716454225357954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357956, "dur": 1,
    "args": {
      "External id": 245893,
      "cbid": 203, "correlation": 245893
    }
  },
  {
    "ph": "f", "id": 245893, "pid": 76337, "tid": -914061504, "ts": 1716454225357956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225357958, "dur": 1,
    "args": {
      "External id": 245894,
      "cbid": 205, "correlation": 245894
    }
  },
  {
    "ph": "f", "id": 245894, "pid": 76337, "tid": -914061504, "ts": 1716454225357958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225358020, "dur": 7,
    "args": {
      "External id": 245898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245898,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245898, "pid": 5, "tid": 7, "ts": 1716454225358020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225358005, "dur": 15,
    "args": {
      "External id": 245898,
      "cbid": 211, "correlation": 245898
    }
  },
  {
    "ph": "s", "id": 245898, "pid": 76337, "tid": -914061504, "ts": 1716454225358005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225358030, "dur": 4,
    "args": {
      "External id": 245900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245900,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 245900, "pid": 5, "tid": 7, "ts": 1716454225358030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225358023, "dur": 6,
    "args": {
      "External id": 245900,
      "cbid": 211, "correlation": 245900
    }
  },
  {
    "ph": "s", "id": 245900, "pid": 76337, "tid": -914061504, "ts": 1716454225358023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225358049, "dur": 3,
    "args": {
      "External id": 245902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 245902, "pid": 5, "tid": 7, "ts": 1716454225358049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225358040, "dur": 8,
    "args": {
      "External id": 245902,
      "cbid": 211, "correlation": 245902
    }
  },
  {
    "ph": "s", "id": 245902, "pid": 76337, "tid": -914061504, "ts": 1716454225358040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225358055, "dur": 0,
    "args": {
      "External id": 245903,
      "cbid": 51, "correlation": 245903
    }
  },
  {
    "ph": "s", "id": 245903, "pid": 76337, "tid": -914061504, "ts": 1716454225358055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225358066, "dur": 87,
    "args": {
      "External id": 245904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245904,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 245904, "pid": 5, "tid": 7, "ts": 1716454225358066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225358056, "dur": 7,
    "args": {
      "External id": 245904,
      "cbid": 211, "correlation": 245904
    }
  },
  {
    "ph": "s", "id": 245904, "pid": 76337, "tid": -914061504, "ts": 1716454225358056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225358154, "dur": 60,
    "args": {
      "External id": 245909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245909, "pid": 5, "tid": 7, "ts": 1716454225358154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225358093, "dur": 11,
    "args": {
      "External id": 245909,
      "cbid": 211, "correlation": 245909
    }
  },
  {
    "ph": "s", "id": 245909, "pid": 76337, "tid": -914061504, "ts": 1716454225358093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225364511, "dur": 51,
    "args": {
      "External id": 245929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245929,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 245929, "pid": 5, "tid": 7, "ts": 1716454225364511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364474, "dur": 38,
    "args": {
      "External id": 245929,
      "cbid": 211, "correlation": 245929
    }
  },
  {
    "ph": "s", "id": 245929, "pid": 76337, "tid": -914061504, "ts": 1716454225364474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225364564, "dur": 5,
    "args": {
      "External id": 245941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245941,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 245941, "pid": 5, "tid": 7, "ts": 1716454225364564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364531, "dur": 11,
    "args": {
      "External id": 245941,
      "cbid": 211, "correlation": 245941
    }
  },
  {
    "ph": "s", "id": 245941, "pid": 76337, "tid": -914061504, "ts": 1716454225364531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225364576, "dur": 57,
    "args": {
      "External id": 245944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245944,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245944, "pid": 5, "tid": 7, "ts": 1716454225364576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364567, "dur": 8,
    "args": {
      "External id": 245944,
      "cbid": 211, "correlation": 245944
    }
  },
  {
    "ph": "s", "id": 245944, "pid": 76337, "tid": -914061504, "ts": 1716454225364567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225364642, "dur": 36,
    "args": {
      "External id": 245953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245953,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245953, "pid": 5, "tid": 7, "ts": 1716454225364642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364629, "dur": 13,
    "args": {
      "External id": 245953,
      "cbid": 211, "correlation": 245953
    }
  },
  {
    "ph": "s", "id": 245953, "pid": 76337, "tid": -914061504, "ts": 1716454225364629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364728, "dur": 1,
    "args": {
      "External id": 245963,
      "cbid": 317, "correlation": 245963
    }
  },
  {
    "ph": "f", "id": 245963, "pid": 76337, "tid": -914061504, "ts": 1716454225364728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364730, "dur": 1,
    "args": {
      "External id": 245964,
      "cbid": 203, "correlation": 245964
    }
  },
  {
    "ph": "f", "id": 245964, "pid": 76337, "tid": -914061504, "ts": 1716454225364730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364732, "dur": 1,
    "args": {
      "External id": 245965,
      "cbid": 205, "correlation": 245965
    }
  },
  {
    "ph": "f", "id": 245965, "pid": 76337, "tid": -914061504, "ts": 1716454225364732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225364782, "dur": 40,
    "args": {
      "External id": 245969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245969,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245969, "pid": 5, "tid": 7, "ts": 1716454225364782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364767, "dur": 14,
    "args": {
      "External id": 245969,
      "cbid": 211, "correlation": 245969
    }
  },
  {
    "ph": "s", "id": 245969, "pid": 76337, "tid": -914061504, "ts": 1716454225364767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225364823, "dur": 14,
    "args": {
      "External id": 245971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245971,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245971, "pid": 5, "tid": 7, "ts": 1716454225364823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364784, "dur": 6,
    "args": {
      "External id": 245971,
      "cbid": 211, "correlation": 245971
    }
  },
  {
    "ph": "s", "id": 245971, "pid": 76337, "tid": -914061504, "ts": 1716454225364784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225364839, "dur": 3,
    "args": {
      "External id": 245973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245973,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 245973, "pid": 5, "tid": 7, "ts": 1716454225364839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364800, "dur": 9,
    "args": {
      "External id": 245973,
      "cbid": 211, "correlation": 245973
    }
  },
  {
    "ph": "s", "id": 245973, "pid": 76337, "tid": -914061504, "ts": 1716454225364800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364817, "dur": 0,
    "args": {
      "External id": 245974,
      "cbid": 51, "correlation": 245974
    }
  },
  {
    "ph": "s", "id": 245974, "pid": 76337, "tid": -914061504, "ts": 1716454225364817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225364844, "dur": 712,
    "args": {
      "External id": 245975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245975,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 245975, "pid": 5, "tid": 7, "ts": 1716454225364844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364819, "dur": 8,
    "args": {
      "External id": 245975,
      "cbid": 211, "correlation": 245975
    }
  },
  {
    "ph": "s", "id": 245975, "pid": 76337, "tid": -914061504, "ts": 1716454225364819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225365557, "dur": 59,
    "args": {
      "External id": 245980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245980,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 245980, "pid": 5, "tid": 7, "ts": 1716454225365557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364863, "dur": 11,
    "args": {
      "External id": 245980,
      "cbid": 211, "correlation": 245980
    }
  },
  {
    "ph": "s", "id": 245980, "pid": 76337, "tid": -914061504, "ts": 1716454225364863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225365618, "dur": 3,
    "args": {
      "External id": 245988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 245988,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 245988, "pid": 5, "tid": 7, "ts": 1716454225365618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225364911, "dur": 10,
    "args": {
      "External id": 245988,
      "cbid": 211, "correlation": 245988
    }
  },
  {
    "ph": "s", "id": 245988, "pid": 76337, "tid": -914061504, "ts": 1716454225364911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365037, "dur": 4,
    "args": {
      "External id": 246004,
      "cbid": 251, "correlation": 246004
    }
  },
  {
    "ph": "f", "id": 246004, "pid": 76337, "tid": -914061504, "ts": 1716454225365037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365047, "dur": 0,
    "args": {
      "External id": 246006,
      "cbid": 251, "correlation": 246006
    }
  },
  {
    "ph": "f", "id": 246006, "pid": 76337, "tid": -914061504, "ts": 1716454225365047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225365623, "dur": 9,
    "args": {
      "External id": 246007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246007,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 246007, "pid": 5, "tid": 7, "ts": 1716454225365623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365051, "dur": 14,
    "args": {
      "External id": 246007,
      "cbid": 211, "correlation": 246007
    }
  },
  {
    "ph": "s", "id": 246007, "pid": 76337, "tid": -914061504, "ts": 1716454225365051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225365632, "dur": 4,
    "args": {
      "External id": 246009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246009,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 246009, "pid": 5, "tid": 7, "ts": 1716454225365632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365070, "dur": 9,
    "args": {
      "External id": 246009,
      "cbid": 211, "correlation": 246009
    }
  },
  {
    "ph": "s", "id": 246009, "pid": 76337, "tid": -914061504, "ts": 1716454225365070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225365638, "dur": 55,
    "args": {
      "External id": 246019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246019, "pid": 5, "tid": 7, "ts": 1716454225365638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365158, "dur": 12,
    "args": {
      "External id": 246019,
      "cbid": 211, "correlation": 246019
    }
  },
  {
    "ph": "s", "id": 246019, "pid": 76337, "tid": -914061504, "ts": 1716454225365158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225365694, "dur": 54,
    "args": {
      "External id": 246039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246039,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 246039, "pid": 5, "tid": 7, "ts": 1716454225365694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365227, "dur": 11,
    "args": {
      "External id": 246039,
      "cbid": 211, "correlation": 246039
    }
  },
  {
    "ph": "s", "id": 246039, "pid": 76337, "tid": -914061504, "ts": 1716454225365227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225365749, "dur": 4,
    "args": {
      "External id": 246051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246051,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246051, "pid": 5, "tid": 7, "ts": 1716454225365749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365250, "dur": 6,
    "args": {
      "External id": 246051,
      "cbid": 211, "correlation": 246051
    }
  },
  {
    "ph": "s", "id": 246051, "pid": 76337, "tid": -914061504, "ts": 1716454225365250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225365754, "dur": 56,
    "args": {
      "External id": 246054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246054,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246054, "pid": 5, "tid": 7, "ts": 1716454225365754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365268, "dur": 7,
    "args": {
      "External id": 246054,
      "cbid": 211, "correlation": 246054
    }
  },
  {
    "ph": "s", "id": 246054, "pid": 76337, "tid": -914061504, "ts": 1716454225365268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225365811, "dur": 36,
    "args": {
      "External id": 246063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246063,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246063, "pid": 5, "tid": 7, "ts": 1716454225365811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365308, "dur": 9,
    "args": {
      "External id": 246063,
      "cbid": 211, "correlation": 246063
    }
  },
  {
    "ph": "s", "id": 246063, "pid": 76337, "tid": -914061504, "ts": 1716454225365308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365386, "dur": 0,
    "args": {
      "External id": 246073,
      "cbid": 317, "correlation": 246073
    }
  },
  {
    "ph": "f", "id": 246073, "pid": 76337, "tid": -914061504, "ts": 1716454225365386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365386, "dur": 0,
    "args": {
      "External id": 246074,
      "cbid": 203, "correlation": 246074
    }
  },
  {
    "ph": "f", "id": 246074, "pid": 76337, "tid": -914061504, "ts": 1716454225365386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365387, "dur": 0,
    "args": {
      "External id": 246075,
      "cbid": 205, "correlation": 246075
    }
  },
  {
    "ph": "f", "id": 246075, "pid": 76337, "tid": -914061504, "ts": 1716454225365387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225365849, "dur": 40,
    "args": {
      "External id": 246079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246079,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246079, "pid": 5, "tid": 7, "ts": 1716454225365849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365402, "dur": 12,
    "args": {
      "External id": 246079,
      "cbid": 211, "correlation": 246079
    }
  },
  {
    "ph": "s", "id": 246079, "pid": 76337, "tid": -914061504, "ts": 1716454225365402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225365890, "dur": 15,
    "args": {
      "External id": 246081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246081,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246081, "pid": 5, "tid": 7, "ts": 1716454225365890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365417, "dur": 5,
    "args": {
      "External id": 246081,
      "cbid": 211, "correlation": 246081
    }
  },
  {
    "ph": "s", "id": 246081, "pid": 76337, "tid": -914061504, "ts": 1716454225365417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225365906, "dur": 4,
    "args": {
      "External id": 246083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246083,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246083, "pid": 5, "tid": 7, "ts": 1716454225365906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365426, "dur": 6,
    "args": {
      "External id": 246083,
      "cbid": 211, "correlation": 246083
    }
  },
  {
    "ph": "s", "id": 246083, "pid": 76337, "tid": -914061504, "ts": 1716454225365426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365435, "dur": 0,
    "args": {
      "External id": 246084,
      "cbid": 51, "correlation": 246084
    }
  },
  {
    "ph": "s", "id": 246084, "pid": 76337, "tid": -914061504, "ts": 1716454225365435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225365911, "dur": 707,
    "args": {
      "External id": 246085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246085,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246085, "pid": 5, "tid": 7, "ts": 1716454225365911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365436, "dur": 6,
    "args": {
      "External id": 246085,
      "cbid": 211, "correlation": 246085
    }
  },
  {
    "ph": "s", "id": 246085, "pid": 76337, "tid": -914061504, "ts": 1716454225365436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225366619, "dur": 59,
    "args": {
      "External id": 246090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246090,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246090, "pid": 5, "tid": 7, "ts": 1716454225366619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365464, "dur": 8,
    "args": {
      "External id": 246090,
      "cbid": 211, "correlation": 246090
    }
  },
  {
    "ph": "s", "id": 246090, "pid": 76337, "tid": -914061504, "ts": 1716454225365464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225366680, "dur": 50,
    "args": {
      "External id": 246098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246098,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246098, "pid": 5, "tid": 7, "ts": 1716454225366680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365497, "dur": 9,
    "args": {
      "External id": 246098,
      "cbid": 211, "correlation": 246098
    }
  },
  {
    "ph": "s", "id": 246098, "pid": 76337, "tid": -914061504, "ts": 1716454225365497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225366731, "dur": 35,
    "args": {
      "External id": 246106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246106,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246106, "pid": 5, "tid": 7, "ts": 1716454225366731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365541, "dur": 11,
    "args": {
      "External id": 246106,
      "cbid": 211, "correlation": 246106
    }
  },
  {
    "ph": "s", "id": 246106, "pid": 76337, "tid": -914061504, "ts": 1716454225365541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225366767, "dur": 52,
    "args": {
      "External id": 246126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246126,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 246126, "pid": 5, "tid": 7, "ts": 1716454225366767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365632, "dur": 13,
    "args": {
      "External id": 246126,
      "cbid": 211, "correlation": 246126
    }
  },
  {
    "ph": "s", "id": 246126, "pid": 76337, "tid": -914061504, "ts": 1716454225365632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225366821, "dur": 4,
    "args": {
      "External id": 246138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246138,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246138, "pid": 5, "tid": 7, "ts": 1716454225366821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365657, "dur": 7,
    "args": {
      "External id": 246138,
      "cbid": 211, "correlation": 246138
    }
  },
  {
    "ph": "s", "id": 246138, "pid": 76337, "tid": -914061504, "ts": 1716454225365657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225366826, "dur": 56,
    "args": {
      "External id": 246141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246141, "pid": 5, "tid": 7, "ts": 1716454225366826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365675, "dur": 7,
    "args": {
      "External id": 246141,
      "cbid": 211, "correlation": 246141
    }
  },
  {
    "ph": "s", "id": 246141, "pid": 76337, "tid": -914061504, "ts": 1716454225365675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365734, "dur": 0,
    "args": {
      "External id": 246152,
      "cbid": 317, "correlation": 246152
    }
  },
  {
    "ph": "f", "id": 246152, "pid": 76337, "tid": -914061504, "ts": 1716454225365734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365735, "dur": 0,
    "args": {
      "External id": 246153,
      "cbid": 203, "correlation": 246153
    }
  },
  {
    "ph": "f", "id": 246153, "pid": 76337, "tid": -914061504, "ts": 1716454225365735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365736, "dur": 0,
    "args": {
      "External id": 246154,
      "cbid": 205, "correlation": 246154
    }
  },
  {
    "ph": "f", "id": 246154, "pid": 76337, "tid": -914061504, "ts": 1716454225365736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365771, "dur": 3,
    "args": {
      "External id": 246158,
      "cbid": 251, "correlation": 246158
    }
  },
  {
    "ph": "f", "id": 246158, "pid": 76337, "tid": -914061504, "ts": 1716454225365771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365775, "dur": 1,
    "args": {
      "External id": 246159,
      "cbid": 251, "correlation": 246159
    }
  },
  {
    "ph": "f", "id": 246159, "pid": 76337, "tid": -914061504, "ts": 1716454225365775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365778, "dur": 1,
    "args": {
      "External id": 246160,
      "cbid": 251, "correlation": 246160
    }
  },
  {
    "ph": "f", "id": 246160, "pid": 76337, "tid": -914061504, "ts": 1716454225365778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365780, "dur": 1,
    "args": {
      "External id": 246161,
      "cbid": 251, "correlation": 246161
    }
  },
  {
    "ph": "f", "id": 246161, "pid": 76337, "tid": -914061504, "ts": 1716454225365780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365782, "dur": 1,
    "args": {
      "External id": 246162,
      "cbid": 251, "correlation": 246162
    }
  },
  {
    "ph": "f", "id": 246162, "pid": 76337, "tid": -914061504, "ts": 1716454225365782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365784, "dur": 1,
    "args": {
      "External id": 246163,
      "cbid": 251, "correlation": 246163
    }
  },
  {
    "ph": "f", "id": 246163, "pid": 76337, "tid": -914061504, "ts": 1716454225365784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365786, "dur": 1,
    "args": {
      "External id": 246164,
      "cbid": 251, "correlation": 246164
    }
  },
  {
    "ph": "f", "id": 246164, "pid": 76337, "tid": -914061504, "ts": 1716454225365786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365788, "dur": 1,
    "args": {
      "External id": 246165,
      "cbid": 251, "correlation": 246165
    }
  },
  {
    "ph": "f", "id": 246165, "pid": 76337, "tid": -914061504, "ts": 1716454225365788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365791, "dur": 0,
    "args": {
      "External id": 246166,
      "cbid": 251, "correlation": 246166
    }
  },
  {
    "ph": "f", "id": 246166, "pid": 76337, "tid": -914061504, "ts": 1716454225365791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225366883, "dur": 117,
    "args": {
      "External id": 246167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246167,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246167, "pid": 5, "tid": 7, "ts": 1716454225366883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365795, "dur": 15,
    "args": {
      "External id": 246167,
      "cbid": 211, "correlation": 246167
    }
  },
  {
    "ph": "s", "id": 246167, "pid": 76337, "tid": -914061504, "ts": 1716454225365795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225367001, "dur": 61,
    "args": {
      "External id": 246173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246173,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246173, "pid": 5, "tid": 7, "ts": 1716454225367001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365835, "dur": 9,
    "args": {
      "External id": 246173,
      "cbid": 211, "correlation": 246173
    }
  },
  {
    "ph": "s", "id": 246173, "pid": 76337, "tid": -914061504, "ts": 1716454225365835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225367064, "dur": 562,
    "args": {
      "External id": 246182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246182,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246182, "pid": 5, "tid": 7, "ts": 1716454225367064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225365951, "dur": 18,
    "args": {
      "External id": 246182,
      "cbid": 211, "correlation": 246182
    }
  },
  {
    "ph": "s", "id": 246182, "pid": 76337, "tid": -914061504, "ts": 1716454225365951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225367627, "dur": 185,
    "args": {
      "External id": 246204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246204,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246204, "pid": 5, "tid": 7, "ts": 1716454225367627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366044, "dur": 13,
    "args": {
      "External id": 246204,
      "cbid": 211, "correlation": 246204
    }
  },
  {
    "ph": "s", "id": 246204, "pid": 76337, "tid": -914061504, "ts": 1716454225366044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366168, "dur": 2,
    "args": {
      "External id": 246215,
      "cbid": 251, "correlation": 246215
    }
  },
  {
    "ph": "f", "id": 246215, "pid": 76337, "tid": -914061504, "ts": 1716454225366168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225367813, "dur": 199,
    "args": {
      "External id": 246216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246216,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246216, "pid": 5, "tid": 7, "ts": 1716454225367813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366175, "dur": 14,
    "args": {
      "External id": 246216,
      "cbid": 211, "correlation": 246216
    }
  },
  {
    "ph": "s", "id": 246216, "pid": 76337, "tid": -914061504, "ts": 1716454225366175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366254, "dur": 1,
    "args": {
      "External id": 246227,
      "cbid": 251, "correlation": 246227
    }
  },
  {
    "ph": "f", "id": 246227, "pid": 76337, "tid": -914061504, "ts": 1716454225366254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225368013, "dur": 190,
    "args": {
      "External id": 246228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246228,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246228, "pid": 5, "tid": 7, "ts": 1716454225368013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366258, "dur": 12,
    "args": {
      "External id": 246228,
      "cbid": 211, "correlation": 246228
    }
  },
  {
    "ph": "s", "id": 246228, "pid": 76337, "tid": -914061504, "ts": 1716454225366258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366323, "dur": 1,
    "args": {
      "External id": 246239,
      "cbid": 251, "correlation": 246239
    }
  },
  {
    "ph": "f", "id": 246239, "pid": 76337, "tid": -914061504, "ts": 1716454225366323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225368205, "dur": 191,
    "args": {
      "External id": 246240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246240,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246240, "pid": 5, "tid": 7, "ts": 1716454225368205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366327, "dur": 11,
    "args": {
      "External id": 246240,
      "cbid": 211, "correlation": 246240
    }
  },
  {
    "ph": "s", "id": 246240, "pid": 76337, "tid": -914061504, "ts": 1716454225366327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225368397, "dur": 18981,
    "args": {
      "External id": 246261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246261,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246261, "pid": 5, "tid": 7, "ts": 1716454225368397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366441, "dur": 17,
    "args": {
      "External id": 246261,
      "cbid": 211, "correlation": 246261
    }
  },
  {
    "ph": "s", "id": 246261, "pid": 76337, "tid": -914061504, "ts": 1716454225366441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366571, "dur": 2,
    "args": {
      "External id": 246279,
      "cbid": 251, "correlation": 246279
    }
  },
  {
    "ph": "f", "id": 246279, "pid": 76337, "tid": -914061504, "ts": 1716454225366571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225387380, "dur": 207,
    "args": {
      "External id": 246281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246281,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246281, "pid": 5, "tid": 7, "ts": 1716454225387380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366578, "dur": 14,
    "args": {
      "External id": 246281,
      "cbid": 211, "correlation": 246281
    }
  },
  {
    "ph": "s", "id": 246281, "pid": 76337, "tid": -914061504, "ts": 1716454225366578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225387588, "dur": 66,
    "args": {
      "External id": 246289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246289,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246289, "pid": 5, "tid": 7, "ts": 1716454225387588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366654, "dur": 12,
    "args": {
      "External id": 246289,
      "cbid": 211, "correlation": 246289
    }
  },
  {
    "ph": "s", "id": 246289, "pid": 76337, "tid": -914061504, "ts": 1716454225366654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225387656, "dur": 96,
    "args": {
      "External id": 246297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246297,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246297, "pid": 5, "tid": 7, "ts": 1716454225387656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366695, "dur": 9,
    "args": {
      "External id": 246297,
      "cbid": 211, "correlation": 246297
    }
  },
  {
    "ph": "s", "id": 246297, "pid": 76337, "tid": -914061504, "ts": 1716454225366695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225387753, "dur": 55,
    "args": {
      "External id": 246308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246308,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246308, "pid": 5, "tid": 7, "ts": 1716454225387753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366796, "dur": 16,
    "args": {
      "External id": 246308,
      "cbid": 211, "correlation": 246308
    }
  },
  {
    "ph": "s", "id": 246308, "pid": 76337, "tid": -914061504, "ts": 1716454225366796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225387810, "dur": 94,
    "args": {
      "External id": 246330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246330,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246330, "pid": 5, "tid": 7, "ts": 1716454225387810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366834, "dur": 10,
    "args": {
      "External id": 246330,
      "cbid": 211, "correlation": 246330
    }
  },
  {
    "ph": "s", "id": 246330, "pid": 76337, "tid": -914061504, "ts": 1716454225366834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366921, "dur": 1,
    "args": {
      "External id": 246341,
      "cbid": 251, "correlation": 246341
    }
  },
  {
    "ph": "f", "id": 246341, "pid": 76337, "tid": -914061504, "ts": 1716454225366921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225387905, "dur": 109,
    "args": {
      "External id": 246342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246342,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246342, "pid": 5, "tid": 7, "ts": 1716454225387905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225366927, "dur": 13,
    "args": {
      "External id": 246342,
      "cbid": 211, "correlation": 246342
    }
  },
  {
    "ph": "s", "id": 246342, "pid": 76337, "tid": -914061504, "ts": 1716454225366927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367021, "dur": 1,
    "args": {
      "External id": 246353,
      "cbid": 251, "correlation": 246353
    }
  },
  {
    "ph": "f", "id": 246353, "pid": 76337, "tid": -914061504, "ts": 1716454225367021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367025, "dur": 0,
    "args": {
      "External id": 246354,
      "cbid": 251, "correlation": 246354
    }
  },
  {
    "ph": "f", "id": 246354, "pid": 76337, "tid": -914061504, "ts": 1716454225367025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225388015, "dur": 10,
    "args": {
      "External id": 246355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246355,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 246355, "pid": 5, "tid": 7, "ts": 1716454225388015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367027, "dur": 15,
    "args": {
      "External id": 246355,
      "cbid": 211, "correlation": 246355
    }
  },
  {
    "ph": "s", "id": 246355, "pid": 76337, "tid": -914061504, "ts": 1716454225367027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225388026, "dur": 5,
    "args": {
      "External id": 246357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246357,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 246357, "pid": 5, "tid": 7, "ts": 1716454225388026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367046, "dur": 8,
    "args": {
      "External id": 246357,
      "cbid": 211, "correlation": 246357
    }
  },
  {
    "ph": "s", "id": 246357, "pid": 76337, "tid": -914061504, "ts": 1716454225367046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367110, "dur": 1,
    "args": {
      "External id": 246368,
      "cbid": 251, "correlation": 246368
    }
  },
  {
    "ph": "f", "id": 246368, "pid": 76337, "tid": -914061504, "ts": 1716454225367110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367113, "dur": 0,
    "args": {
      "External id": 246369,
      "cbid": 251, "correlation": 246369
    }
  },
  {
    "ph": "f", "id": 246369, "pid": 76337, "tid": -914061504, "ts": 1716454225367113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225388033, "dur": 6,
    "args": {
      "External id": 246370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246370,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 246370, "pid": 5, "tid": 7, "ts": 1716454225388033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367115, "dur": 12,
    "args": {
      "External id": 246370,
      "cbid": 211, "correlation": 246370
    }
  },
  {
    "ph": "s", "id": 246370, "pid": 76337, "tid": -914061504, "ts": 1716454225367115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225388040, "dur": 4,
    "args": {
      "External id": 246372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246372,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 246372, "pid": 5, "tid": 7, "ts": 1716454225388040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367128, "dur": 5,
    "args": {
      "External id": 246372,
      "cbid": 211, "correlation": 246372
    }
  },
  {
    "ph": "s", "id": 246372, "pid": 76337, "tid": -914061504, "ts": 1716454225367128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225388045, "dur": 159,
    "args": {
      "External id": 246393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246393,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246393, "pid": 5, "tid": 7, "ts": 1716454225388045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367202, "dur": 12,
    "args": {
      "External id": 246393,
      "cbid": 211, "correlation": 246393
    }
  },
  {
    "ph": "s", "id": 246393, "pid": 76337, "tid": -914061504, "ts": 1716454225367202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367300, "dur": 2,
    "args": {
      "External id": 246411,
      "cbid": 251, "correlation": 246411
    }
  },
  {
    "ph": "f", "id": 246411, "pid": 76337, "tid": -914061504, "ts": 1716454225367300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225388205, "dur": 110,
    "args": {
      "External id": 246413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246413,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246413, "pid": 5, "tid": 7, "ts": 1716454225388205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367306, "dur": 14,
    "args": {
      "External id": 246413,
      "cbid": 211, "correlation": 246413
    }
  },
  {
    "ph": "s", "id": 246413, "pid": 76337, "tid": -914061504, "ts": 1716454225367306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225388317, "dur": 35,
    "args": {
      "External id": 246421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246421,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246421, "pid": 5, "tid": 7, "ts": 1716454225388317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367377, "dur": 13,
    "args": {
      "External id": 246421,
      "cbid": 211, "correlation": 246421
    }
  },
  {
    "ph": "s", "id": 246421, "pid": 76337, "tid": -914061504, "ts": 1716454225367377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225388353, "dur": 69,
    "args": {
      "External id": 246429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246429, "pid": 5, "tid": 7, "ts": 1716454225388353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367418, "dur": 9,
    "args": {
      "External id": 246429,
      "cbid": 211, "correlation": 246429
    }
  },
  {
    "ph": "s", "id": 246429, "pid": 76337, "tid": -914061504, "ts": 1716454225367418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225388424, "dur": 94,
    "args": {
      "External id": 246451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246451,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246451, "pid": 5, "tid": 7, "ts": 1716454225388424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367476, "dur": 11,
    "args": {
      "External id": 246451,
      "cbid": 211, "correlation": 246451
    }
  },
  {
    "ph": "s", "id": 246451, "pid": 76337, "tid": -914061504, "ts": 1716454225367476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367571, "dur": 1,
    "args": {
      "External id": 246467,
      "cbid": 251, "correlation": 246467
    }
  },
  {
    "ph": "f", "id": 246467, "pid": 76337, "tid": -914061504, "ts": 1716454225367571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225388519, "dur": 586,
    "args": {
      "External id": 246469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246469,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246469, "pid": 5, "tid": 7, "ts": 1716454225388519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367578, "dur": 13,
    "args": {
      "External id": 246469,
      "cbid": 211, "correlation": 246469
    }
  },
  {
    "ph": "s", "id": 246469, "pid": 76337, "tid": -914061504, "ts": 1716454225367578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225389106, "dur": 248,
    "args": {
      "External id": 246477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246477, "pid": 5, "tid": 7, "ts": 1716454225389106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367663, "dur": 14,
    "args": {
      "External id": 246477,
      "cbid": 211, "correlation": 246477
    }
  },
  {
    "ph": "s", "id": 246477, "pid": 76337, "tid": -914061504, "ts": 1716454225367663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225389355, "dur": 251,
    "args": {
      "External id": 246485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246485,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246485, "pid": 5, "tid": 7, "ts": 1716454225389355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367706, "dur": 11,
    "args": {
      "External id": 246485,
      "cbid": 211, "correlation": 246485
    }
  },
  {
    "ph": "s", "id": 246485, "pid": 76337, "tid": -914061504, "ts": 1716454225367706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367797, "dur": 2,
    "args": {
      "External id": 246501,
      "cbid": 251, "correlation": 246501
    }
  },
  {
    "ph": "f", "id": 246501, "pid": 76337, "tid": -914061504, "ts": 1716454225367797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367802, "dur": 0,
    "args": {
      "External id": 246503,
      "cbid": 251, "correlation": 246503
    }
  },
  {
    "ph": "f", "id": 246503, "pid": 76337, "tid": -914061504, "ts": 1716454225367802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225389608, "dur": 365,
    "args": {
      "External id": 246504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246504,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 246504, "pid": 5, "tid": 7, "ts": 1716454225389608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367806, "dur": 14,
    "args": {
      "External id": 246504,
      "cbid": 211, "correlation": 246504
    }
  },
  {
    "ph": "s", "id": 246504, "pid": 76337, "tid": -914061504, "ts": 1716454225367806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225389974, "dur": 50,
    "args": {
      "External id": 246512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246512,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246512, "pid": 5, "tid": 7, "ts": 1716454225389974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367850, "dur": 10,
    "args": {
      "External id": 246512,
      "cbid": 211, "correlation": 246512
    }
  },
  {
    "ph": "s", "id": 246512, "pid": 76337, "tid": -914061504, "ts": 1716454225367850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225390025, "dur": 161,
    "args": {
      "External id": 246523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246523,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246523, "pid": 5, "tid": 7, "ts": 1716454225390025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225367923, "dur": 13,
    "args": {
      "External id": 246523,
      "cbid": 211, "correlation": 246523
    }
  },
  {
    "ph": "s", "id": 246523, "pid": 76337, "tid": -914061504, "ts": 1716454225367923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368005, "dur": 0,
    "args": {
      "External id": 246535,
      "cbid": 317, "correlation": 246535
    }
  },
  {
    "ph": "f", "id": 246535, "pid": 76337, "tid": -914061504, "ts": 1716454225368005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368005, "dur": 0,
    "args": {
      "External id": 246536,
      "cbid": 203, "correlation": 246536
    }
  },
  {
    "ph": "f", "id": 246536, "pid": 76337, "tid": -914061504, "ts": 1716454225368005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368006, "dur": 0,
    "args": {
      "External id": 246537,
      "cbid": 205, "correlation": 246537
    }
  },
  {
    "ph": "f", "id": 246537, "pid": 76337, "tid": -914061504, "ts": 1716454225368006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368030, "dur": 1,
    "args": {
      "External id": 246541,
      "cbid": 251, "correlation": 246541
    }
  },
  {
    "ph": "f", "id": 246541, "pid": 76337, "tid": -914061504, "ts": 1716454225368030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368032, "dur": 0,
    "args": {
      "External id": 246542,
      "cbid": 251, "correlation": 246542
    }
  },
  {
    "ph": "f", "id": 246542, "pid": 76337, "tid": -914061504, "ts": 1716454225368032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368033, "dur": 0,
    "args": {
      "External id": 246543,
      "cbid": 251, "correlation": 246543
    }
  },
  {
    "ph": "f", "id": 246543, "pid": 76337, "tid": -914061504, "ts": 1716454225368033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368034, "dur": 0,
    "args": {
      "External id": 246544,
      "cbid": 251, "correlation": 246544
    }
  },
  {
    "ph": "f", "id": 246544, "pid": 76337, "tid": -914061504, "ts": 1716454225368034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368035, "dur": 0,
    "args": {
      "External id": 246545,
      "cbid": 251, "correlation": 246545
    }
  },
  {
    "ph": "f", "id": 246545, "pid": 76337, "tid": -914061504, "ts": 1716454225368035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368035, "dur": 0,
    "args": {
      "External id": 246546,
      "cbid": 251, "correlation": 246546
    }
  },
  {
    "ph": "f", "id": 246546, "pid": 76337, "tid": -914061504, "ts": 1716454225368035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368037, "dur": 0,
    "args": {
      "External id": 246547,
      "cbid": 251, "correlation": 246547
    }
  },
  {
    "ph": "f", "id": 246547, "pid": 76337, "tid": -914061504, "ts": 1716454225368037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368037, "dur": 0,
    "args": {
      "External id": 246548,
      "cbid": 251, "correlation": 246548
    }
  },
  {
    "ph": "f", "id": 246548, "pid": 76337, "tid": -914061504, "ts": 1716454225368037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368039, "dur": 0,
    "args": {
      "External id": 246549,
      "cbid": 251, "correlation": 246549
    }
  },
  {
    "ph": "f", "id": 246549, "pid": 76337, "tid": -914061504, "ts": 1716454225368039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225390188, "dur": 116,
    "args": {
      "External id": 246550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246550,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246550, "pid": 5, "tid": 7, "ts": 1716454225390188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368041, "dur": 13,
    "args": {
      "External id": 246550,
      "cbid": 211, "correlation": 246550
    }
  },
  {
    "ph": "s", "id": 246550, "pid": 76337, "tid": -914061504, "ts": 1716454225368041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225390305, "dur": 60,
    "args": {
      "External id": 246556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246556, "pid": 5, "tid": 7, "ts": 1716454225390305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368077, "dur": 10,
    "args": {
      "External id": 246556,
      "cbid": 211, "correlation": 246556
    }
  },
  {
    "ph": "s", "id": 246556, "pid": 76337, "tid": -914061504, "ts": 1716454225368077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225390367, "dur": 49,
    "args": {
      "External id": 246564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246564,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246564, "pid": 5, "tid": 7, "ts": 1716454225390367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368110, "dur": 9,
    "args": {
      "External id": 246564,
      "cbid": 211, "correlation": 246564
    }
  },
  {
    "ph": "s", "id": 246564, "pid": 76337, "tid": -914061504, "ts": 1716454225368110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225390417, "dur": 53,
    "args": {
      "External id": 246584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246584,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 246584, "pid": 5, "tid": 7, "ts": 1716454225390417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368193, "dur": 12,
    "args": {
      "External id": 246584,
      "cbid": 211, "correlation": 246584
    }
  },
  {
    "ph": "s", "id": 246584, "pid": 76337, "tid": -914061504, "ts": 1716454225368193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225390472, "dur": 5,
    "args": {
      "External id": 246596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246596,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246596, "pid": 5, "tid": 7, "ts": 1716454225390472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368214, "dur": 6,
    "args": {
      "External id": 246596,
      "cbid": 211, "correlation": 246596
    }
  },
  {
    "ph": "s", "id": 246596, "pid": 76337, "tid": -914061504, "ts": 1716454225368214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225390477, "dur": 57,
    "args": {
      "External id": 246599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246599, "pid": 5, "tid": 7, "ts": 1716454225390477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368233, "dur": 6,
    "args": {
      "External id": 246599,
      "cbid": 211, "correlation": 246599
    }
  },
  {
    "ph": "s", "id": 246599, "pid": 76337, "tid": -914061504, "ts": 1716454225368233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225390536, "dur": 37,
    "args": {
      "External id": 246608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246608,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246608, "pid": 5, "tid": 7, "ts": 1716454225390536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368273, "dur": 11,
    "args": {
      "External id": 246608,
      "cbid": 211, "correlation": 246608
    }
  },
  {
    "ph": "s", "id": 246608, "pid": 76337, "tid": -914061504, "ts": 1716454225368273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368331, "dur": 0,
    "args": {
      "External id": 246618,
      "cbid": 317, "correlation": 246618
    }
  },
  {
    "ph": "f", "id": 246618, "pid": 76337, "tid": -914061504, "ts": 1716454225368331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368332, "dur": 0,
    "args": {
      "External id": 246619,
      "cbid": 203, "correlation": 246619
    }
  },
  {
    "ph": "f", "id": 246619, "pid": 76337, "tid": -914061504, "ts": 1716454225368332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368333, "dur": 0,
    "args": {
      "External id": 246620,
      "cbid": 205, "correlation": 246620
    }
  },
  {
    "ph": "f", "id": 246620, "pid": 76337, "tid": -914061504, "ts": 1716454225368333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225390574, "dur": 40,
    "args": {
      "External id": 246624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246624, "pid": 5, "tid": 7, "ts": 1716454225390574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368347, "dur": 11,
    "args": {
      "External id": 246624,
      "cbid": 211, "correlation": 246624
    }
  },
  {
    "ph": "s", "id": 246624, "pid": 76337, "tid": -914061504, "ts": 1716454225368347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225390616, "dur": 15,
    "args": {
      "External id": 246626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246626, "pid": 5, "tid": 7, "ts": 1716454225390616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368361, "dur": 5,
    "args": {
      "External id": 246626,
      "cbid": 211, "correlation": 246626
    }
  },
  {
    "ph": "s", "id": 246626, "pid": 76337, "tid": -914061504, "ts": 1716454225368361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225390632, "dur": 4,
    "args": {
      "External id": 246628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246628, "pid": 5, "tid": 7, "ts": 1716454225390632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368371, "dur": 6,
    "args": {
      "External id": 246628,
      "cbid": 211, "correlation": 246628
    }
  },
  {
    "ph": "s", "id": 246628, "pid": 76337, "tid": -914061504, "ts": 1716454225368371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368380, "dur": 0,
    "args": {
      "External id": 246629,
      "cbid": 51, "correlation": 246629
    }
  },
  {
    "ph": "s", "id": 246629, "pid": 76337, "tid": -914061504, "ts": 1716454225368380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225390637, "dur": 711,
    "args": {
      "External id": 246630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246630,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246630, "pid": 5, "tid": 7, "ts": 1716454225390637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368380, "dur": 5,
    "args": {
      "External id": 246630,
      "cbid": 211, "correlation": 246630
    }
  },
  {
    "ph": "s", "id": 246630, "pid": 76337, "tid": -914061504, "ts": 1716454225368380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225391349, "dur": 60,
    "args": {
      "External id": 246635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246635, "pid": 5, "tid": 7, "ts": 1716454225391349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368408, "dur": 8,
    "args": {
      "External id": 246635,
      "cbid": 211, "correlation": 246635
    }
  },
  {
    "ph": "s", "id": 246635, "pid": 76337, "tid": -914061504, "ts": 1716454225368408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225391410, "dur": 3,
    "args": {
      "External id": 246643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246643,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246643, "pid": 5, "tid": 7, "ts": 1716454225391410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368451, "dur": 9,
    "args": {
      "External id": 246643,
      "cbid": 211, "correlation": 246643
    }
  },
  {
    "ph": "s", "id": 246643, "pid": 76337, "tid": -914061504, "ts": 1716454225368451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368522, "dur": 2,
    "args": {
      "External id": 246659,
      "cbid": 251, "correlation": 246659
    }
  },
  {
    "ph": "f", "id": 246659, "pid": 76337, "tid": -914061504, "ts": 1716454225368522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368528, "dur": 0,
    "args": {
      "External id": 246661,
      "cbid": 251, "correlation": 246661
    }
  },
  {
    "ph": "f", "id": 246661, "pid": 76337, "tid": -914061504, "ts": 1716454225368528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225391415, "dur": 12,
    "args": {
      "External id": 246662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246662,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 246662, "pid": 5, "tid": 7, "ts": 1716454225391415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368530, "dur": 12,
    "args": {
      "External id": 246662,
      "cbid": 211, "correlation": 246662
    }
  },
  {
    "ph": "s", "id": 246662, "pid": 76337, "tid": -914061504, "ts": 1716454225368530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225391428, "dur": 5,
    "args": {
      "External id": 246664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246664,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 246664, "pid": 5, "tid": 7, "ts": 1716454225391428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368544, "dur": 5,
    "args": {
      "External id": 246664,
      "cbid": 211, "correlation": 246664
    }
  },
  {
    "ph": "s", "id": 246664, "pid": 76337, "tid": -914061504, "ts": 1716454225368544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225391435, "dur": 55,
    "args": {
      "External id": 246674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246674, "pid": 5, "tid": 7, "ts": 1716454225391435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368602, "dur": 12,
    "args": {
      "External id": 246674,
      "cbid": 211, "correlation": 246674
    }
  },
  {
    "ph": "s", "id": 246674, "pid": 76337, "tid": -914061504, "ts": 1716454225368602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225391491, "dur": 54,
    "args": {
      "External id": 246694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246694,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 246694, "pid": 5, "tid": 7, "ts": 1716454225391491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368673, "dur": 12,
    "args": {
      "External id": 246694,
      "cbid": 211, "correlation": 246694
    }
  },
  {
    "ph": "s", "id": 246694, "pid": 76337, "tid": -914061504, "ts": 1716454225368673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225391546, "dur": 4,
    "args": {
      "External id": 246706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246706,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246706, "pid": 5, "tid": 7, "ts": 1716454225391546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368694, "dur": 6,
    "args": {
      "External id": 246706,
      "cbid": 211, "correlation": 246706
    }
  },
  {
    "ph": "s", "id": 246706, "pid": 76337, "tid": -914061504, "ts": 1716454225368694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225391551, "dur": 56,
    "args": {
      "External id": 246709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246709, "pid": 5, "tid": 7, "ts": 1716454225391551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368712, "dur": 6,
    "args": {
      "External id": 246709,
      "cbid": 211, "correlation": 246709
    }
  },
  {
    "ph": "s", "id": 246709, "pid": 76337, "tid": -914061504, "ts": 1716454225368712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225391609, "dur": 37,
    "args": {
      "External id": 246718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246718,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246718, "pid": 5, "tid": 7, "ts": 1716454225391609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368752, "dur": 10,
    "args": {
      "External id": 246718,
      "cbid": 211, "correlation": 246718
    }
  },
  {
    "ph": "s", "id": 246718, "pid": 76337, "tid": -914061504, "ts": 1716454225368752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368820, "dur": 0,
    "args": {
      "External id": 246728,
      "cbid": 317, "correlation": 246728
    }
  },
  {
    "ph": "f", "id": 246728, "pid": 76337, "tid": -914061504, "ts": 1716454225368820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368821, "dur": 0,
    "args": {
      "External id": 246729,
      "cbid": 203, "correlation": 246729
    }
  },
  {
    "ph": "f", "id": 246729, "pid": 76337, "tid": -914061504, "ts": 1716454225368821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368821, "dur": 0,
    "args": {
      "External id": 246730,
      "cbid": 205, "correlation": 246730
    }
  },
  {
    "ph": "f", "id": 246730, "pid": 76337, "tid": -914061504, "ts": 1716454225368821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225391647, "dur": 39,
    "args": {
      "External id": 246734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246734, "pid": 5, "tid": 7, "ts": 1716454225391647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368835, "dur": 12,
    "args": {
      "External id": 246734,
      "cbid": 211, "correlation": 246734
    }
  },
  {
    "ph": "s", "id": 246734, "pid": 76337, "tid": -914061504, "ts": 1716454225368835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225391687, "dur": 14,
    "args": {
      "External id": 246736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246736,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246736, "pid": 5, "tid": 7, "ts": 1716454225391687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368850, "dur": 5,
    "args": {
      "External id": 246736,
      "cbid": 211, "correlation": 246736
    }
  },
  {
    "ph": "s", "id": 246736, "pid": 76337, "tid": -914061504, "ts": 1716454225368850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225391703, "dur": 3,
    "args": {
      "External id": 246738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246738, "pid": 5, "tid": 7, "ts": 1716454225391703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368859, "dur": 5,
    "args": {
      "External id": 246738,
      "cbid": 211, "correlation": 246738
    }
  },
  {
    "ph": "s", "id": 246738, "pid": 76337, "tid": -914061504, "ts": 1716454225368859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368867, "dur": 0,
    "args": {
      "External id": 246739,
      "cbid": 51, "correlation": 246739
    }
  },
  {
    "ph": "s", "id": 246739, "pid": 76337, "tid": -914061504, "ts": 1716454225368867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225391708, "dur": 707,
    "args": {
      "External id": 246740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246740,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246740, "pid": 5, "tid": 7, "ts": 1716454225391708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368868, "dur": 5,
    "args": {
      "External id": 246740,
      "cbid": 211, "correlation": 246740
    }
  },
  {
    "ph": "s", "id": 246740, "pid": 76337, "tid": -914061504, "ts": 1716454225368868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225392416, "dur": 60,
    "args": {
      "External id": 246745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246745, "pid": 5, "tid": 7, "ts": 1716454225392416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368896, "dur": 8,
    "args": {
      "External id": 246745,
      "cbid": 211, "correlation": 246745
    }
  },
  {
    "ph": "s", "id": 246745, "pid": 76337, "tid": -914061504, "ts": 1716454225368896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225392477, "dur": 50,
    "args": {
      "External id": 246753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246753,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246753, "pid": 5, "tid": 7, "ts": 1716454225392477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368928, "dur": 9,
    "args": {
      "External id": 246753,
      "cbid": 211, "correlation": 246753
    }
  },
  {
    "ph": "s", "id": 246753, "pid": 76337, "tid": -914061504, "ts": 1716454225368928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225392528, "dur": 35,
    "args": {
      "External id": 246761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246761,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246761, "pid": 5, "tid": 7, "ts": 1716454225392528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225368958, "dur": 8,
    "args": {
      "External id": 246761,
      "cbid": 211, "correlation": 246761
    }
  },
  {
    "ph": "s", "id": 246761, "pid": 76337, "tid": -914061504, "ts": 1716454225368958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225392565, "dur": 51,
    "args": {
      "External id": 246781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246781,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 246781, "pid": 5, "tid": 7, "ts": 1716454225392565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369049, "dur": 13,
    "args": {
      "External id": 246781,
      "cbid": 211, "correlation": 246781
    }
  },
  {
    "ph": "s", "id": 246781, "pid": 76337, "tid": -914061504, "ts": 1716454225369049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225392618, "dur": 4,
    "args": {
      "External id": 246793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246793,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 246793, "pid": 5, "tid": 7, "ts": 1716454225392618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369072, "dur": 6,
    "args": {
      "External id": 246793,
      "cbid": 211, "correlation": 246793
    }
  },
  {
    "ph": "s", "id": 246793, "pid": 76337, "tid": -914061504, "ts": 1716454225369072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225392623, "dur": 57,
    "args": {
      "External id": 246796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246796, "pid": 5, "tid": 7, "ts": 1716454225392623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369091, "dur": 6,
    "args": {
      "External id": 246796,
      "cbid": 211, "correlation": 246796
    }
  },
  {
    "ph": "s", "id": 246796, "pid": 76337, "tid": -914061504, "ts": 1716454225369091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369148, "dur": 0,
    "args": {
      "External id": 246807,
      "cbid": 317, "correlation": 246807
    }
  },
  {
    "ph": "f", "id": 246807, "pid": 76337, "tid": -914061504, "ts": 1716454225369148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369149, "dur": 0,
    "args": {
      "External id": 246808,
      "cbid": 203, "correlation": 246808
    }
  },
  {
    "ph": "f", "id": 246808, "pid": 76337, "tid": -914061504, "ts": 1716454225369149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369149, "dur": 0,
    "args": {
      "External id": 246809,
      "cbid": 205, "correlation": 246809
    }
  },
  {
    "ph": "f", "id": 246809, "pid": 76337, "tid": -914061504, "ts": 1716454225369149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369171, "dur": 1,
    "args": {
      "External id": 246813,
      "cbid": 251, "correlation": 246813
    }
  },
  {
    "ph": "f", "id": 246813, "pid": 76337, "tid": -914061504, "ts": 1716454225369171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369173, "dur": 0,
    "args": {
      "External id": 246814,
      "cbid": 251, "correlation": 246814
    }
  },
  {
    "ph": "f", "id": 246814, "pid": 76337, "tid": -914061504, "ts": 1716454225369173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369174, "dur": 0,
    "args": {
      "External id": 246815,
      "cbid": 251, "correlation": 246815
    }
  },
  {
    "ph": "f", "id": 246815, "pid": 76337, "tid": -914061504, "ts": 1716454225369174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369174, "dur": 0,
    "args": {
      "External id": 246816,
      "cbid": 251, "correlation": 246816
    }
  },
  {
    "ph": "f", "id": 246816, "pid": 76337, "tid": -914061504, "ts": 1716454225369174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369175, "dur": 0,
    "args": {
      "External id": 246817,
      "cbid": 251, "correlation": 246817
    }
  },
  {
    "ph": "f", "id": 246817, "pid": 76337, "tid": -914061504, "ts": 1716454225369175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369176, "dur": 0,
    "args": {
      "External id": 246818,
      "cbid": 251, "correlation": 246818
    }
  },
  {
    "ph": "f", "id": 246818, "pid": 76337, "tid": -914061504, "ts": 1716454225369176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369177, "dur": 0,
    "args": {
      "External id": 246819,
      "cbid": 251, "correlation": 246819
    }
  },
  {
    "ph": "f", "id": 246819, "pid": 76337, "tid": -914061504, "ts": 1716454225369177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369177, "dur": 0,
    "args": {
      "External id": 246820,
      "cbid": 251, "correlation": 246820
    }
  },
  {
    "ph": "f", "id": 246820, "pid": 76337, "tid": -914061504, "ts": 1716454225369177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369179, "dur": 0,
    "args": {
      "External id": 246821,
      "cbid": 251, "correlation": 246821
    }
  },
  {
    "ph": "f", "id": 246821, "pid": 76337, "tid": -914061504, "ts": 1716454225369179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225392681, "dur": 113,
    "args": {
      "External id": 246822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246822,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246822, "pid": 5, "tid": 7, "ts": 1716454225392681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369181, "dur": 13,
    "args": {
      "External id": 246822,
      "cbid": 211, "correlation": 246822
    }
  },
  {
    "ph": "s", "id": 246822, "pid": 76337, "tid": -914061504, "ts": 1716454225369181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225392796, "dur": 60,
    "args": {
      "External id": 246828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246828,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246828, "pid": 5, "tid": 7, "ts": 1716454225392796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369216, "dur": 9,
    "args": {
      "External id": 246828,
      "cbid": 211, "correlation": 246828
    }
  },
  {
    "ph": "s", "id": 246828, "pid": 76337, "tid": -914061504, "ts": 1716454225369216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225392858, "dur": 569,
    "args": {
      "External id": 246837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246837,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246837, "pid": 5, "tid": 7, "ts": 1716454225392858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369300, "dur": 13,
    "args": {
      "External id": 246837,
      "cbid": 211, "correlation": 246837
    }
  },
  {
    "ph": "s", "id": 246837, "pid": 76337, "tid": -914061504, "ts": 1716454225369300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225393428, "dur": 184,
    "args": {
      "External id": 246859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246859,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246859, "pid": 5, "tid": 7, "ts": 1716454225393428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369357, "dur": 10,
    "args": {
      "External id": 246859,
      "cbid": 211, "correlation": 246859
    }
  },
  {
    "ph": "s", "id": 246859, "pid": 76337, "tid": -914061504, "ts": 1716454225369357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369443, "dur": 1,
    "args": {
      "External id": 246870,
      "cbid": 251, "correlation": 246870
    }
  },
  {
    "ph": "f", "id": 246870, "pid": 76337, "tid": -914061504, "ts": 1716454225369443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225393614, "dur": 197,
    "args": {
      "External id": 246871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246871,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246871, "pid": 5, "tid": 7, "ts": 1716454225393614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369449, "dur": 13,
    "args": {
      "External id": 246871,
      "cbid": 211, "correlation": 246871
    }
  },
  {
    "ph": "s", "id": 246871, "pid": 76337, "tid": -914061504, "ts": 1716454225369449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369517, "dur": 1,
    "args": {
      "External id": 246882,
      "cbid": 251, "correlation": 246882
    }
  },
  {
    "ph": "f", "id": 246882, "pid": 76337, "tid": -914061504, "ts": 1716454225369517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225393812, "dur": 191,
    "args": {
      "External id": 246883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246883,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246883, "pid": 5, "tid": 7, "ts": 1716454225393812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369521, "dur": 12,
    "args": {
      "External id": 246883,
      "cbid": 211, "correlation": 246883
    }
  },
  {
    "ph": "s", "id": 246883, "pid": 76337, "tid": -914061504, "ts": 1716454225369521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369585, "dur": 1,
    "args": {
      "External id": 246894,
      "cbid": 251, "correlation": 246894
    }
  },
  {
    "ph": "f", "id": 246894, "pid": 76337, "tid": -914061504, "ts": 1716454225369585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225394004, "dur": 190,
    "args": {
      "External id": 246895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246895,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246895, "pid": 5, "tid": 7, "ts": 1716454225394004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369589, "dur": 11,
    "args": {
      "External id": 246895,
      "cbid": 211, "correlation": 246895
    }
  },
  {
    "ph": "s", "id": 246895, "pid": 76337, "tid": -914061504, "ts": 1716454225369589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225394195, "dur": 19016,
    "args": {
      "External id": 246916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246916,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 246916, "pid": 5, "tid": 7, "ts": 1716454225394195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369669, "dur": 13,
    "args": {
      "External id": 246916,
      "cbid": 211, "correlation": 246916
    }
  },
  {
    "ph": "s", "id": 246916, "pid": 76337, "tid": -914061504, "ts": 1716454225369669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369770, "dur": 1,
    "args": {
      "External id": 246934,
      "cbid": 251, "correlation": 246934
    }
  },
  {
    "ph": "f", "id": 246934, "pid": 76337, "tid": -914061504, "ts": 1716454225369770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225413213, "dur": 205,
    "args": {
      "External id": 246936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246936,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246936, "pid": 5, "tid": 7, "ts": 1716454225413213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369776, "dur": 14,
    "args": {
      "External id": 246936,
      "cbid": 211, "correlation": 246936
    }
  },
  {
    "ph": "s", "id": 246936, "pid": 76337, "tid": -914061504, "ts": 1716454225369776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225413419, "dur": 66,
    "args": {
      "External id": 246944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246944,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246944, "pid": 5, "tid": 7, "ts": 1716454225413419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369848, "dur": 12,
    "args": {
      "External id": 246944,
      "cbid": 211, "correlation": 246944
    }
  },
  {
    "ph": "s", "id": 246944, "pid": 76337, "tid": -914061504, "ts": 1716454225369848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225413487, "dur": 98,
    "args": {
      "External id": 246952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246952,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246952, "pid": 5, "tid": 7, "ts": 1716454225413487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369887, "dur": 9,
    "args": {
      "External id": 246952,
      "cbid": 211, "correlation": 246952
    }
  },
  {
    "ph": "s", "id": 246952, "pid": 76337, "tid": -914061504, "ts": 1716454225369887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225413585, "dur": 56,
    "args": {
      "External id": 246963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246963,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246963, "pid": 5, "tid": 7, "ts": 1716454225413585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225369961, "dur": 21,
    "args": {
      "External id": 246963,
      "cbid": 211, "correlation": 246963
    }
  },
  {
    "ph": "s", "id": 246963, "pid": 76337, "tid": -914061504, "ts": 1716454225369961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225413643, "dur": 94,
    "args": {
      "External id": 246985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246985,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 246985, "pid": 5, "tid": 7, "ts": 1716454225413643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370002, "dur": 8,
    "args": {
      "External id": 246985,
      "cbid": 211, "correlation": 246985
    }
  },
  {
    "ph": "s", "id": 246985, "pid": 76337, "tid": -914061504, "ts": 1716454225370002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370092, "dur": 1,
    "args": {
      "External id": 246996,
      "cbid": 251, "correlation": 246996
    }
  },
  {
    "ph": "f", "id": 246996, "pid": 76337, "tid": -914061504, "ts": 1716454225370092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225413738, "dur": 106,
    "args": {
      "External id": 246997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 246997,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 246997, "pid": 5, "tid": 7, "ts": 1716454225413738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370097, "dur": 14,
    "args": {
      "External id": 246997,
      "cbid": 211, "correlation": 246997
    }
  },
  {
    "ph": "s", "id": 246997, "pid": 76337, "tid": -914061504, "ts": 1716454225370097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370171, "dur": 1,
    "args": {
      "External id": 247008,
      "cbid": 251, "correlation": 247008
    }
  },
  {
    "ph": "f", "id": 247008, "pid": 76337, "tid": -914061504, "ts": 1716454225370171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370174, "dur": 0,
    "args": {
      "External id": 247009,
      "cbid": 251, "correlation": 247009
    }
  },
  {
    "ph": "f", "id": 247009, "pid": 76337, "tid": -914061504, "ts": 1716454225370174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225413845, "dur": 11,
    "args": {
      "External id": 247010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247010,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 247010, "pid": 5, "tid": 7, "ts": 1716454225413845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370176, "dur": 13,
    "args": {
      "External id": 247010,
      "cbid": 211, "correlation": 247010
    }
  },
  {
    "ph": "s", "id": 247010, "pid": 76337, "tid": -914061504, "ts": 1716454225370176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225413857, "dur": 5,
    "args": {
      "External id": 247012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247012,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 247012, "pid": 5, "tid": 7, "ts": 1716454225413857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370190, "dur": 6,
    "args": {
      "External id": 247012,
      "cbid": 211, "correlation": 247012
    }
  },
  {
    "ph": "s", "id": 247012, "pid": 76337, "tid": -914061504, "ts": 1716454225370190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370254, "dur": 1,
    "args": {
      "External id": 247023,
      "cbid": 251, "correlation": 247023
    }
  },
  {
    "ph": "f", "id": 247023, "pid": 76337, "tid": -914061504, "ts": 1716454225370254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370258, "dur": 0,
    "args": {
      "External id": 247024,
      "cbid": 251, "correlation": 247024
    }
  },
  {
    "ph": "f", "id": 247024, "pid": 76337, "tid": -914061504, "ts": 1716454225370258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225413864, "dur": 6,
    "args": {
      "External id": 247025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247025,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 247025, "pid": 5, "tid": 7, "ts": 1716454225413864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370259, "dur": 12,
    "args": {
      "External id": 247025,
      "cbid": 211, "correlation": 247025
    }
  },
  {
    "ph": "s", "id": 247025, "pid": 76337, "tid": -914061504, "ts": 1716454225370259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225413872, "dur": 3,
    "args": {
      "External id": 247027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247027,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 247027, "pid": 5, "tid": 7, "ts": 1716454225413872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370273, "dur": 5,
    "args": {
      "External id": 247027,
      "cbid": 211, "correlation": 247027
    }
  },
  {
    "ph": "s", "id": 247027, "pid": 76337, "tid": -914061504, "ts": 1716454225370273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225413876, "dur": 158,
    "args": {
      "External id": 247048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247048,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 247048, "pid": 5, "tid": 7, "ts": 1716454225413876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370347, "dur": 12,
    "args": {
      "External id": 247048,
      "cbid": 211, "correlation": 247048
    }
  },
  {
    "ph": "s", "id": 247048, "pid": 76337, "tid": -914061504, "ts": 1716454225370347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370448, "dur": 1,
    "args": {
      "External id": 247066,
      "cbid": 251, "correlation": 247066
    }
  },
  {
    "ph": "f", "id": 247066, "pid": 76337, "tid": -914061504, "ts": 1716454225370448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225414036, "dur": 109,
    "args": {
      "External id": 247068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247068,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 247068, "pid": 5, "tid": 7, "ts": 1716454225414036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370454, "dur": 13,
    "args": {
      "External id": 247068,
      "cbid": 211, "correlation": 247068
    }
  },
  {
    "ph": "s", "id": 247068, "pid": 76337, "tid": -914061504, "ts": 1716454225370454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225414146, "dur": 34,
    "args": {
      "External id": 247076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247076,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247076, "pid": 5, "tid": 7, "ts": 1716454225414146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370523, "dur": 12,
    "args": {
      "External id": 247076,
      "cbid": 211, "correlation": 247076
    }
  },
  {
    "ph": "s", "id": 247076, "pid": 76337, "tid": -914061504, "ts": 1716454225370523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225414182, "dur": 68,
    "args": {
      "External id": 247084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247084,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247084, "pid": 5, "tid": 7, "ts": 1716454225414182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370564, "dur": 9,
    "args": {
      "External id": 247084,
      "cbid": 211, "correlation": 247084
    }
  },
  {
    "ph": "s", "id": 247084, "pid": 76337, "tid": -914061504, "ts": 1716454225370564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225414251, "dur": 94,
    "args": {
      "External id": 247106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247106,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247106, "pid": 5, "tid": 7, "ts": 1716454225414251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370620, "dur": 10,
    "args": {
      "External id": 247106,
      "cbid": 211, "correlation": 247106
    }
  },
  {
    "ph": "s", "id": 247106, "pid": 76337, "tid": -914061504, "ts": 1716454225370620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370708, "dur": 1,
    "args": {
      "External id": 247122,
      "cbid": 251, "correlation": 247122
    }
  },
  {
    "ph": "f", "id": 247122, "pid": 76337, "tid": -914061504, "ts": 1716454225370708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225414346, "dur": 585,
    "args": {
      "External id": 247124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247124,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247124, "pid": 5, "tid": 7, "ts": 1716454225414346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370714, "dur": 12,
    "args": {
      "External id": 247124,
      "cbid": 211, "correlation": 247124
    }
  },
  {
    "ph": "s", "id": 247124, "pid": 76337, "tid": -914061504, "ts": 1716454225370714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225414932, "dur": 248,
    "args": {
      "External id": 247132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247132, "pid": 5, "tid": 7, "ts": 1716454225414932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370778, "dur": 13,
    "args": {
      "External id": 247132,
      "cbid": 211, "correlation": 247132
    }
  },
  {
    "ph": "s", "id": 247132, "pid": 76337, "tid": -914061504, "ts": 1716454225370778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225415181, "dur": 254,
    "args": {
      "External id": 247140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247140, "pid": 5, "tid": 7, "ts": 1716454225415181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370809, "dur": 8,
    "args": {
      "External id": 247140,
      "cbid": 211, "correlation": 247140
    }
  },
  {
    "ph": "s", "id": 247140, "pid": 76337, "tid": -914061504, "ts": 1716454225370809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370892, "dur": 1,
    "args": {
      "External id": 247156,
      "cbid": 251, "correlation": 247156
    }
  },
  {
    "ph": "f", "id": 247156, "pid": 76337, "tid": -914061504, "ts": 1716454225370892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370897, "dur": 0,
    "args": {
      "External id": 247158,
      "cbid": 251, "correlation": 247158
    }
  },
  {
    "ph": "f", "id": 247158, "pid": 76337, "tid": -914061504, "ts": 1716454225370897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225415437, "dur": 363,
    "args": {
      "External id": 247159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247159,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 247159, "pid": 5, "tid": 7, "ts": 1716454225415437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370899, "dur": 13,
    "args": {
      "External id": 247159,
      "cbid": 211, "correlation": 247159
    }
  },
  {
    "ph": "s", "id": 247159, "pid": 76337, "tid": -914061504, "ts": 1716454225370899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225415801, "dur": 50,
    "args": {
      "External id": 247167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247167,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247167, "pid": 5, "tid": 7, "ts": 1716454225415801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225370941, "dur": 9,
    "args": {
      "External id": 247167,
      "cbid": 211, "correlation": 247167
    }
  },
  {
    "ph": "s", "id": 247167, "pid": 76337, "tid": -914061504, "ts": 1716454225370941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225415852, "dur": 160,
    "args": {
      "External id": 247178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247178, "pid": 5, "tid": 7, "ts": 1716454225415852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371016, "dur": 13,
    "args": {
      "External id": 247178,
      "cbid": 211, "correlation": 247178
    }
  },
  {
    "ph": "s", "id": 247178, "pid": 76337, "tid": -914061504, "ts": 1716454225371016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371081, "dur": 0,
    "args": {
      "External id": 247190,
      "cbid": 317, "correlation": 247190
    }
  },
  {
    "ph": "f", "id": 247190, "pid": 76337, "tid": -914061504, "ts": 1716454225371081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371082, "dur": 0,
    "args": {
      "External id": 247191,
      "cbid": 203, "correlation": 247191
    }
  },
  {
    "ph": "f", "id": 247191, "pid": 76337, "tid": -914061504, "ts": 1716454225371082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371082, "dur": 0,
    "args": {
      "External id": 247192,
      "cbid": 205, "correlation": 247192
    }
  },
  {
    "ph": "f", "id": 247192, "pid": 76337, "tid": -914061504, "ts": 1716454225371082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371105, "dur": 1,
    "args": {
      "External id": 247196,
      "cbid": 251, "correlation": 247196
    }
  },
  {
    "ph": "f", "id": 247196, "pid": 76337, "tid": -914061504, "ts": 1716454225371105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371107, "dur": 0,
    "args": {
      "External id": 247197,
      "cbid": 251, "correlation": 247197
    }
  },
  {
    "ph": "f", "id": 247197, "pid": 76337, "tid": -914061504, "ts": 1716454225371107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371108, "dur": 0,
    "args": {
      "External id": 247198,
      "cbid": 251, "correlation": 247198
    }
  },
  {
    "ph": "f", "id": 247198, "pid": 76337, "tid": -914061504, "ts": 1716454225371108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371109, "dur": 0,
    "args": {
      "External id": 247199,
      "cbid": 251, "correlation": 247199
    }
  },
  {
    "ph": "f", "id": 247199, "pid": 76337, "tid": -914061504, "ts": 1716454225371109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371110, "dur": 0,
    "args": {
      "External id": 247200,
      "cbid": 251, "correlation": 247200
    }
  },
  {
    "ph": "f", "id": 247200, "pid": 76337, "tid": -914061504, "ts": 1716454225371110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371110, "dur": 0,
    "args": {
      "External id": 247201,
      "cbid": 251, "correlation": 247201
    }
  },
  {
    "ph": "f", "id": 247201, "pid": 76337, "tid": -914061504, "ts": 1716454225371110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371111, "dur": 0,
    "args": {
      "External id": 247202,
      "cbid": 251, "correlation": 247202
    }
  },
  {
    "ph": "f", "id": 247202, "pid": 76337, "tid": -914061504, "ts": 1716454225371111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371112, "dur": 0,
    "args": {
      "External id": 247203,
      "cbid": 251, "correlation": 247203
    }
  },
  {
    "ph": "f", "id": 247203, "pid": 76337, "tid": -914061504, "ts": 1716454225371112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371113, "dur": 0,
    "args": {
      "External id": 247204,
      "cbid": 251, "correlation": 247204
    }
  },
  {
    "ph": "f", "id": 247204, "pid": 76337, "tid": -914061504, "ts": 1716454225371113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225416014, "dur": 117,
    "args": {
      "External id": 247205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247205,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 247205, "pid": 5, "tid": 7, "ts": 1716454225416014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371115, "dur": 12,
    "args": {
      "External id": 247205,
      "cbid": 211, "correlation": 247205
    }
  },
  {
    "ph": "s", "id": 247205, "pid": 76337, "tid": -914061504, "ts": 1716454225371115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225416132, "dur": 60,
    "args": {
      "External id": 247211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247211, "pid": 5, "tid": 7, "ts": 1716454225416132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371150, "dur": 8,
    "args": {
      "External id": 247211,
      "cbid": 211, "correlation": 247211
    }
  },
  {
    "ph": "s", "id": 247211, "pid": 76337, "tid": -914061504, "ts": 1716454225371150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225416193, "dur": 50,
    "args": {
      "External id": 247219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247219,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247219, "pid": 5, "tid": 7, "ts": 1716454225416193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371181, "dur": 9,
    "args": {
      "External id": 247219,
      "cbid": 211, "correlation": 247219
    }
  },
  {
    "ph": "s", "id": 247219, "pid": 76337, "tid": -914061504, "ts": 1716454225371181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371256, "dur": 0,
    "args": {
      "External id": 247229,
      "cbid": 317, "correlation": 247229
    }
  },
  {
    "ph": "f", "id": 247229, "pid": 76337, "tid": -914061504, "ts": 1716454225371256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371256, "dur": 0,
    "args": {
      "External id": 247230,
      "cbid": 203, "correlation": 247230
    }
  },
  {
    "ph": "f", "id": 247230, "pid": 76337, "tid": -914061504, "ts": 1716454225371256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371257, "dur": 0,
    "args": {
      "External id": 247231,
      "cbid": 205, "correlation": 247231
    }
  },
  {
    "ph": "f", "id": 247231, "pid": 76337, "tid": -914061504, "ts": 1716454225371257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225416245, "dur": 40,
    "args": {
      "External id": 247235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247235, "pid": 5, "tid": 7, "ts": 1716454225416245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371274, "dur": 13,
    "args": {
      "External id": 247235,
      "cbid": 211, "correlation": 247235
    }
  },
  {
    "ph": "s", "id": 247235, "pid": 76337, "tid": -914061504, "ts": 1716454225371274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225416286, "dur": 15,
    "args": {
      "External id": 247237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247237, "pid": 5, "tid": 7, "ts": 1716454225416286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371290, "dur": 5,
    "args": {
      "External id": 247237,
      "cbid": 211, "correlation": 247237
    }
  },
  {
    "ph": "s", "id": 247237, "pid": 76337, "tid": -914061504, "ts": 1716454225371290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225416303, "dur": 1,
    "args": {
      "External id": 247239,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 247239,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 247239, "pid": 5, "tid": 7, "ts": 1716454225416303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371309, "dur": 19,
    "args": {
      "External id": 247239,
      "cbid": 51, "correlation": 247239
    }
  },
  {
    "ph": "s", "id": 247239, "pid": 76337, "tid": -914061504, "ts": 1716454225371309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225416307, "dur": 367,
    "args": {
      "External id": 247240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247240,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247240, "pid": 5, "tid": 7, "ts": 1716454225416307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371330, "dur": 10,
    "args": {
      "External id": 247240,
      "cbid": 211, "correlation": 247240
    }
  },
  {
    "ph": "s", "id": 247240, "pid": 76337, "tid": -914061504, "ts": 1716454225371330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225416676, "dur": 14,
    "args": {
      "External id": 247242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247242,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247242, "pid": 5, "tid": 7, "ts": 1716454225416676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371353, "dur": 7,
    "args": {
      "External id": 247242,
      "cbid": 211, "correlation": 247242
    }
  },
  {
    "ph": "s", "id": 247242, "pid": 76337, "tid": -914061504, "ts": 1716454225371353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225416691, "dur": 15,
    "args": {
      "External id": 247248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247248, "pid": 5, "tid": 7, "ts": 1716454225416691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371384, "dur": 9,
    "args": {
      "External id": 247248,
      "cbid": 211, "correlation": 247248
    }
  },
  {
    "ph": "s", "id": 247248, "pid": 76337, "tid": -914061504, "ts": 1716454225371384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225416707, "dur": 20,
    "args": {
      "External id": 247268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247268,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 247268, "pid": 5, "tid": 7, "ts": 1716454225416707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371480, "dur": 13,
    "args": {
      "External id": 247268,
      "cbid": 211, "correlation": 247268
    }
  },
  {
    "ph": "s", "id": 247268, "pid": 76337, "tid": -914061504, "ts": 1716454225371480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225416728, "dur": 5,
    "args": {
      "External id": 247280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247280,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 247280, "pid": 5, "tid": 7, "ts": 1716454225416728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371504, "dur": 6,
    "args": {
      "External id": 247280,
      "cbid": 211, "correlation": 247280
    }
  },
  {
    "ph": "s", "id": 247280, "pid": 76337, "tid": -914061504, "ts": 1716454225371504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225416734, "dur": 18,
    "args": {
      "External id": 247283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247283,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247283, "pid": 5, "tid": 7, "ts": 1716454225416734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371522, "dur": 7,
    "args": {
      "External id": 247283,
      "cbid": 211, "correlation": 247283
    }
  },
  {
    "ph": "s", "id": 247283, "pid": 76337, "tid": -914061504, "ts": 1716454225371522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225416753, "dur": 12,
    "args": {
      "External id": 247292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247292,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247292, "pid": 5, "tid": 7, "ts": 1716454225416753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371563, "dur": 9,
    "args": {
      "External id": 247292,
      "cbid": 211, "correlation": 247292
    }
  },
  {
    "ph": "s", "id": 247292, "pid": 76337, "tid": -914061504, "ts": 1716454225371563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371622, "dur": 0,
    "args": {
      "External id": 247302,
      "cbid": 317, "correlation": 247302
    }
  },
  {
    "ph": "f", "id": 247302, "pid": 76337, "tid": -914061504, "ts": 1716454225371622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371623, "dur": 0,
    "args": {
      "External id": 247303,
      "cbid": 203, "correlation": 247303
    }
  },
  {
    "ph": "f", "id": 247303, "pid": 76337, "tid": -914061504, "ts": 1716454225371623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371624, "dur": 0,
    "args": {
      "External id": 247304,
      "cbid": 205, "correlation": 247304
    }
  },
  {
    "ph": "f", "id": 247304, "pid": 76337, "tid": -914061504, "ts": 1716454225371624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225416766, "dur": 11,
    "args": {
      "External id": 247308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247308,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247308, "pid": 5, "tid": 7, "ts": 1716454225416766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371640, "dur": 11,
    "args": {
      "External id": 247308,
      "cbid": 211, "correlation": 247308
    }
  },
  {
    "ph": "s", "id": 247308, "pid": 76337, "tid": -914061504, "ts": 1716454225371640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225416779, "dur": 25,
    "args": {
      "External id": 247310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247310, "pid": 5, "tid": 7, "ts": 1716454225416779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371653, "dur": 5,
    "args": {
      "External id": 247310,
      "cbid": 211, "correlation": 247310
    }
  },
  {
    "ph": "s", "id": 247310, "pid": 76337, "tid": -914061504, "ts": 1716454225371653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225416805, "dur": 4,
    "args": {
      "External id": 247312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 247312, "pid": 5, "tid": 7, "ts": 1716454225416805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371664, "dur": 5,
    "args": {
      "External id": 247312,
      "cbid": 211, "correlation": 247312
    }
  },
  {
    "ph": "s", "id": 247312, "pid": 76337, "tid": -914061504, "ts": 1716454225371664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371675, "dur": 0,
    "args": {
      "External id": 247313,
      "cbid": 51, "correlation": 247313
    }
  },
  {
    "ph": "s", "id": 247313, "pid": 76337, "tid": -914061504, "ts": 1716454225371675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225416810, "dur": 363,
    "args": {
      "External id": 247314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247314,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247314, "pid": 5, "tid": 7, "ts": 1716454225416810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371677, "dur": 9,
    "args": {
      "External id": 247314,
      "cbid": 211, "correlation": 247314
    }
  },
  {
    "ph": "s", "id": 247314, "pid": 76337, "tid": -914061504, "ts": 1716454225371677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225417174, "dur": 20,
    "args": {
      "External id": 247315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247315,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247315, "pid": 5, "tid": 7, "ts": 1716454225417174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371688, "dur": 5,
    "args": {
      "External id": 247315,
      "cbid": 211, "correlation": 247315
    }
  },
  {
    "ph": "s", "id": 247315, "pid": 76337, "tid": -914061504, "ts": 1716454225371688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225417196, "dur": 33,
    "args": {
      "External id": 247321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247321, "pid": 5, "tid": 7, "ts": 1716454225417196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371716, "dur": 8,
    "args": {
      "External id": 247321,
      "cbid": 211, "correlation": 247321
    }
  },
  {
    "ph": "s", "id": 247321, "pid": 76337, "tid": -914061504, "ts": 1716454225371716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225417230, "dur": 4,
    "args": {
      "External id": 247329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247329,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 247329, "pid": 5, "tid": 7, "ts": 1716454225417230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371758, "dur": 9,
    "args": {
      "External id": 247329,
      "cbid": 211, "correlation": 247329
    }
  },
  {
    "ph": "s", "id": 247329, "pid": 76337, "tid": -914061504, "ts": 1716454225371758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371825, "dur": 2,
    "args": {
      "External id": 247345,
      "cbid": 251, "correlation": 247345
    }
  },
  {
    "ph": "f", "id": 247345, "pid": 76337, "tid": -914061504, "ts": 1716454225371825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371832, "dur": 0,
    "args": {
      "External id": 247347,
      "cbid": 251, "correlation": 247347
    }
  },
  {
    "ph": "f", "id": 247347, "pid": 76337, "tid": -914061504, "ts": 1716454225371832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225417235, "dur": 12,
    "args": {
      "External id": 247348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247348,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 247348, "pid": 5, "tid": 7, "ts": 1716454225417235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371833, "dur": 11,
    "args": {
      "External id": 247348,
      "cbid": 211, "correlation": 247348
    }
  },
  {
    "ph": "s", "id": 247348, "pid": 76337, "tid": -914061504, "ts": 1716454225371833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225417248, "dur": 5,
    "args": {
      "External id": 247350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247350,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 247350, "pid": 5, "tid": 7, "ts": 1716454225417248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371846, "dur": 5,
    "args": {
      "External id": 247350,
      "cbid": 211, "correlation": 247350
    }
  },
  {
    "ph": "s", "id": 247350, "pid": 76337, "tid": -914061504, "ts": 1716454225371846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225417255, "dur": 30,
    "args": {
      "External id": 247360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247360, "pid": 5, "tid": 7, "ts": 1716454225417255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371904, "dur": 12,
    "args": {
      "External id": 247360,
      "cbid": 211, "correlation": 247360
    }
  },
  {
    "ph": "s", "id": 247360, "pid": 76337, "tid": -914061504, "ts": 1716454225371904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225417286, "dur": 31,
    "args": {
      "External id": 247380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247380,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 247380, "pid": 5, "tid": 7, "ts": 1716454225417286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225371971, "dur": 21,
    "args": {
      "External id": 247380,
      "cbid": 211, "correlation": 247380
    }
  },
  {
    "ph": "s", "id": 247380, "pid": 76337, "tid": -914061504, "ts": 1716454225371971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225417318, "dur": 4,
    "args": {
      "External id": 247392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247392,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 247392, "pid": 5, "tid": 7, "ts": 1716454225417318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372003, "dur": 6,
    "args": {
      "External id": 247392,
      "cbid": 211, "correlation": 247392
    }
  },
  {
    "ph": "s", "id": 247392, "pid": 76337, "tid": -914061504, "ts": 1716454225372003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225417324, "dur": 31,
    "args": {
      "External id": 247395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247395, "pid": 5, "tid": 7, "ts": 1716454225417324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372022, "dur": 7,
    "args": {
      "External id": 247395,
      "cbid": 211, "correlation": 247395
    }
  },
  {
    "ph": "s", "id": 247395, "pid": 76337, "tid": -914061504, "ts": 1716454225372022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225417356, "dur": 21,
    "args": {
      "External id": 247404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247404,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247404, "pid": 5, "tid": 7, "ts": 1716454225417356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372064, "dur": 9,
    "args": {
      "External id": 247404,
      "cbid": 211, "correlation": 247404
    }
  },
  {
    "ph": "s", "id": 247404, "pid": 76337, "tid": -914061504, "ts": 1716454225372064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372127, "dur": 0,
    "args": {
      "External id": 247414,
      "cbid": 317, "correlation": 247414
    }
  },
  {
    "ph": "f", "id": 247414, "pid": 76337, "tid": -914061504, "ts": 1716454225372127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372128, "dur": 0,
    "args": {
      "External id": 247415,
      "cbid": 203, "correlation": 247415
    }
  },
  {
    "ph": "f", "id": 247415, "pid": 76337, "tid": -914061504, "ts": 1716454225372128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372129, "dur": 0,
    "args": {
      "External id": 247416,
      "cbid": 205, "correlation": 247416
    }
  },
  {
    "ph": "f", "id": 247416, "pid": 76337, "tid": -914061504, "ts": 1716454225372129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225417378, "dur": 23,
    "args": {
      "External id": 247420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247420, "pid": 5, "tid": 7, "ts": 1716454225417378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372151, "dur": 12,
    "args": {
      "External id": 247420,
      "cbid": 211, "correlation": 247420
    }
  },
  {
    "ph": "s", "id": 247420, "pid": 76337, "tid": -914061504, "ts": 1716454225372151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225417403, "dur": 44,
    "args": {
      "External id": 247422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247422, "pid": 5, "tid": 7, "ts": 1716454225417403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372165, "dur": 5,
    "args": {
      "External id": 247422,
      "cbid": 211, "correlation": 247422
    }
  },
  {
    "ph": "s", "id": 247422, "pid": 76337, "tid": -914061504, "ts": 1716454225372165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225417448, "dur": 667,
    "args": {
      "External id": 247424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247424,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247424, "pid": 5, "tid": 7, "ts": 1716454225417448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372180, "dur": 9,
    "args": {
      "External id": 247424,
      "cbid": 211, "correlation": 247424
    }
  },
  {
    "ph": "s", "id": 247424, "pid": 76337, "tid": -914061504, "ts": 1716454225372180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225418117, "dur": 22,
    "args": {
      "External id": 247426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247426,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247426, "pid": 5, "tid": 7, "ts": 1716454225418117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372192, "dur": 5,
    "args": {
      "External id": 247426,
      "cbid": 211, "correlation": 247426
    }
  },
  {
    "ph": "s", "id": 247426, "pid": 76337, "tid": -914061504, "ts": 1716454225372192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225418140, "dur": 34,
    "args": {
      "External id": 247432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247432, "pid": 5, "tid": 7, "ts": 1716454225418140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372220, "dur": 9,
    "args": {
      "External id": 247432,
      "cbid": 211, "correlation": 247432
    }
  },
  {
    "ph": "s", "id": 247432, "pid": 76337, "tid": -914061504, "ts": 1716454225372220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372283, "dur": 0,
    "args": {
      "External id": 247442,
      "cbid": 317, "correlation": 247442
    }
  },
  {
    "ph": "f", "id": 247442, "pid": 76337, "tid": -914061504, "ts": 1716454225372283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372284, "dur": 0,
    "args": {
      "External id": 247443,
      "cbid": 203, "correlation": 247443
    }
  },
  {
    "ph": "f", "id": 247443, "pid": 76337, "tid": -914061504, "ts": 1716454225372284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372285, "dur": 0,
    "args": {
      "External id": 247444,
      "cbid": 205, "correlation": 247444
    }
  },
  {
    "ph": "f", "id": 247444, "pid": 76337, "tid": -914061504, "ts": 1716454225372285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372307, "dur": 1,
    "args": {
      "External id": 247448,
      "cbid": 251, "correlation": 247448
    }
  },
  {
    "ph": "f", "id": 247448, "pid": 76337, "tid": -914061504, "ts": 1716454225372307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372309, "dur": 0,
    "args": {
      "External id": 247449,
      "cbid": 251, "correlation": 247449
    }
  },
  {
    "ph": "f", "id": 247449, "pid": 76337, "tid": -914061504, "ts": 1716454225372309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372310, "dur": 0,
    "args": {
      "External id": 247450,
      "cbid": 251, "correlation": 247450
    }
  },
  {
    "ph": "f", "id": 247450, "pid": 76337, "tid": -914061504, "ts": 1716454225372310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372311, "dur": 0,
    "args": {
      "External id": 247451,
      "cbid": 251, "correlation": 247451
    }
  },
  {
    "ph": "f", "id": 247451, "pid": 76337, "tid": -914061504, "ts": 1716454225372311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372311, "dur": 0,
    "args": {
      "External id": 247452,
      "cbid": 251, "correlation": 247452
    }
  },
  {
    "ph": "f", "id": 247452, "pid": 76337, "tid": -914061504, "ts": 1716454225372311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372312, "dur": 0,
    "args": {
      "External id": 247453,
      "cbid": 251, "correlation": 247453
    }
  },
  {
    "ph": "f", "id": 247453, "pid": 76337, "tid": -914061504, "ts": 1716454225372312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372313, "dur": 0,
    "args": {
      "External id": 247454,
      "cbid": 251, "correlation": 247454
    }
  },
  {
    "ph": "f", "id": 247454, "pid": 76337, "tid": -914061504, "ts": 1716454225372313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372314, "dur": 0,
    "args": {
      "External id": 247455,
      "cbid": 251, "correlation": 247455
    }
  },
  {
    "ph": "f", "id": 247455, "pid": 76337, "tid": -914061504, "ts": 1716454225372314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372315, "dur": 0,
    "args": {
      "External id": 247456,
      "cbid": 251, "correlation": 247456
    }
  },
  {
    "ph": "f", "id": 247456, "pid": 76337, "tid": -914061504, "ts": 1716454225372315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225418175, "dur": 52,
    "args": {
      "External id": 247457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247457,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 247457, "pid": 5, "tid": 7, "ts": 1716454225418175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372317, "dur": 13,
    "args": {
      "External id": 247457,
      "cbid": 211, "correlation": 247457
    }
  },
  {
    "ph": "s", "id": 247457, "pid": 76337, "tid": -914061504, "ts": 1716454225372317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225418228, "dur": 33,
    "args": {
      "External id": 247463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247463, "pid": 5, "tid": 7, "ts": 1716454225418228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372350, "dur": 8,
    "args": {
      "External id": 247463,
      "cbid": 211, "correlation": 247463
    }
  },
  {
    "ph": "s", "id": 247463, "pid": 76337, "tid": -914061504, "ts": 1716454225372350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225418262, "dur": 27,
    "args": {
      "External id": 247471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247471,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247471, "pid": 5, "tid": 7, "ts": 1716454225418262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372381, "dur": 8,
    "args": {
      "External id": 247471,
      "cbid": 211, "correlation": 247471
    }
  },
  {
    "ph": "s", "id": 247471, "pid": 76337, "tid": -914061504, "ts": 1716454225372381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225418290, "dur": 20,
    "args": {
      "External id": 247479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247479,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247479, "pid": 5, "tid": 7, "ts": 1716454225418290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372410, "dur": 8,
    "args": {
      "External id": 247479,
      "cbid": 211, "correlation": 247479
    }
  },
  {
    "ph": "s", "id": 247479, "pid": 76337, "tid": -914061504, "ts": 1716454225372410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225418311, "dur": 30,
    "args": {
      "External id": 247499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247499,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 247499, "pid": 5, "tid": 7, "ts": 1716454225418311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372497, "dur": 12,
    "args": {
      "External id": 247499,
      "cbid": 211, "correlation": 247499
    }
  },
  {
    "ph": "s", "id": 247499, "pid": 76337, "tid": -914061504, "ts": 1716454225372497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225418343, "dur": 4,
    "args": {
      "External id": 247511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247511,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 247511, "pid": 5, "tid": 7, "ts": 1716454225418343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372519, "dur": 6,
    "args": {
      "External id": 247511,
      "cbid": 211, "correlation": 247511
    }
  },
  {
    "ph": "s", "id": 247511, "pid": 76337, "tid": -914061504, "ts": 1716454225372519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225418349, "dur": 30,
    "args": {
      "External id": 247514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247514, "pid": 5, "tid": 7, "ts": 1716454225418349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372537, "dur": 6,
    "args": {
      "External id": 247514,
      "cbid": 211, "correlation": 247514
    }
  },
  {
    "ph": "s", "id": 247514, "pid": 76337, "tid": -914061504, "ts": 1716454225372537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372595, "dur": 0,
    "args": {
      "External id": 247525,
      "cbid": 317, "correlation": 247525
    }
  },
  {
    "ph": "f", "id": 247525, "pid": 76337, "tid": -914061504, "ts": 1716454225372595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372595, "dur": 0,
    "args": {
      "External id": 247526,
      "cbid": 203, "correlation": 247526
    }
  },
  {
    "ph": "f", "id": 247526, "pid": 76337, "tid": -914061504, "ts": 1716454225372595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372596, "dur": 0,
    "args": {
      "External id": 247527,
      "cbid": 205, "correlation": 247527
    }
  },
  {
    "ph": "f", "id": 247527, "pid": 76337, "tid": -914061504, "ts": 1716454225372596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225418380, "dur": 22,
    "args": {
      "External id": 247531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247531,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247531, "pid": 5, "tid": 7, "ts": 1716454225418380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372611, "dur": 12,
    "args": {
      "External id": 247531,
      "cbid": 211, "correlation": 247531
    }
  },
  {
    "ph": "s", "id": 247531, "pid": 76337, "tid": -914061504, "ts": 1716454225372611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225418404, "dur": 123,
    "args": {
      "External id": 247533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247533,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247533, "pid": 5, "tid": 7, "ts": 1716454225418404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372632, "dur": 8,
    "args": {
      "External id": 247533,
      "cbid": 211, "correlation": 247533
    }
  },
  {
    "ph": "s", "id": 247533, "pid": 76337, "tid": -914061504, "ts": 1716454225372632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225418528, "dur": 23,
    "args": {
      "External id": 247535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247535,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247535, "pid": 5, "tid": 7, "ts": 1716454225418528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372644, "dur": 5,
    "args": {
      "External id": 247535,
      "cbid": 211, "correlation": 247535
    }
  },
  {
    "ph": "s", "id": 247535, "pid": 76337, "tid": -914061504, "ts": 1716454225372644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225418552, "dur": 33,
    "args": {
      "External id": 247541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247541, "pid": 5, "tid": 7, "ts": 1716454225418552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372671, "dur": 9,
    "args": {
      "External id": 247541,
      "cbid": 211, "correlation": 247541
    }
  },
  {
    "ph": "s", "id": 247541, "pid": 76337, "tid": -914061504, "ts": 1716454225372671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225418587, "dur": 202,
    "args": {
      "External id": 247550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247550,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247550, "pid": 5, "tid": 7, "ts": 1716454225418587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372755, "dur": 14,
    "args": {
      "External id": 247550,
      "cbid": 211, "correlation": 247550
    }
  },
  {
    "ph": "s", "id": 247550, "pid": 76337, "tid": -914061504, "ts": 1716454225372755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225418790, "dur": 66,
    "args": {
      "External id": 247572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247572,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247572, "pid": 5, "tid": 7, "ts": 1716454225418790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372817, "dur": 11,
    "args": {
      "External id": 247572,
      "cbid": 211, "correlation": 247572
    }
  },
  {
    "ph": "s", "id": 247572, "pid": 76337, "tid": -914061504, "ts": 1716454225372817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372911, "dur": 1,
    "args": {
      "External id": 247583,
      "cbid": 251, "correlation": 247583
    }
  },
  {
    "ph": "f", "id": 247583, "pid": 76337, "tid": -914061504, "ts": 1716454225372911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225418858, "dur": 158,
    "args": {
      "External id": 247584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247584,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247584, "pid": 5, "tid": 7, "ts": 1716454225418858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372916, "dur": 13,
    "args": {
      "External id": 247584,
      "cbid": 211, "correlation": 247584
    }
  },
  {
    "ph": "s", "id": 247584, "pid": 76337, "tid": -914061504, "ts": 1716454225372916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372993, "dur": 1,
    "args": {
      "External id": 247595,
      "cbid": 251, "correlation": 247595
    }
  },
  {
    "ph": "f", "id": 247595, "pid": 76337, "tid": -914061504, "ts": 1716454225372993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225419017, "dur": 150,
    "args": {
      "External id": 247596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247596,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247596, "pid": 5, "tid": 7, "ts": 1716454225419017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225372997, "dur": 12,
    "args": {
      "External id": 247596,
      "cbid": 211, "correlation": 247596
    }
  },
  {
    "ph": "s", "id": 247596, "pid": 76337, "tid": -914061504, "ts": 1716454225372997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373066, "dur": 1,
    "args": {
      "External id": 247607,
      "cbid": 251, "correlation": 247607
    }
  },
  {
    "ph": "f", "id": 247607, "pid": 76337, "tid": -914061504, "ts": 1716454225373066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225419169, "dur": 147,
    "args": {
      "External id": 247608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247608,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247608, "pid": 5, "tid": 7, "ts": 1716454225419169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373070, "dur": 12,
    "args": {
      "External id": 247608,
      "cbid": 211, "correlation": 247608
    }
  },
  {
    "ph": "s", "id": 247608, "pid": 76337, "tid": -914061504, "ts": 1716454225373070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225419317, "dur": 1987,
    "args": {
      "External id": 247629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247629,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 247629, "pid": 5, "tid": 7, "ts": 1716454225419317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373155, "dur": 13,
    "args": {
      "External id": 247629,
      "cbid": 211, "correlation": 247629
    }
  },
  {
    "ph": "s", "id": 247629, "pid": 76337, "tid": -914061504, "ts": 1716454225373155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373261, "dur": 1,
    "args": {
      "External id": 247647,
      "cbid": 251, "correlation": 247647
    }
  },
  {
    "ph": "f", "id": 247647, "pid": 76337, "tid": -914061504, "ts": 1716454225373261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225421305, "dur": 150,
    "args": {
      "External id": 247649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247649,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 247649, "pid": 5, "tid": 7, "ts": 1716454225421305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373267, "dur": 14,
    "args": {
      "External id": 247649,
      "cbid": 211, "correlation": 247649
    }
  },
  {
    "ph": "s", "id": 247649, "pid": 76337, "tid": -914061504, "ts": 1716454225373267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225421457, "dur": 36,
    "args": {
      "External id": 247657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247657,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247657, "pid": 5, "tid": 7, "ts": 1716454225421457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373338, "dur": 12,
    "args": {
      "External id": 247657,
      "cbid": 211, "correlation": 247657
    }
  },
  {
    "ph": "s", "id": 247657, "pid": 76337, "tid": -914061504, "ts": 1716454225373338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225421494, "dur": 50,
    "args": {
      "External id": 247665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247665,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247665, "pid": 5, "tid": 7, "ts": 1716454225421494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373376, "dur": 9,
    "args": {
      "External id": 247665,
      "cbid": 211, "correlation": 247665
    }
  },
  {
    "ph": "s", "id": 247665, "pid": 76337, "tid": -914061504, "ts": 1716454225373376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225421545, "dur": 31,
    "args": {
      "External id": 247676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247676, "pid": 5, "tid": 7, "ts": 1716454225421545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373450, "dur": 13,
    "args": {
      "External id": 247676,
      "cbid": 211, "correlation": 247676
    }
  },
  {
    "ph": "s", "id": 247676, "pid": 76337, "tid": -914061504, "ts": 1716454225373450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225421577, "dur": 35,
    "args": {
      "External id": 247698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247698,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247698, "pid": 5, "tid": 7, "ts": 1716454225421577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373481, "dur": 7,
    "args": {
      "External id": 247698,
      "cbid": 211, "correlation": 247698
    }
  },
  {
    "ph": "s", "id": 247698, "pid": 76337, "tid": -914061504, "ts": 1716454225373481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373570, "dur": 1,
    "args": {
      "External id": 247709,
      "cbid": 251, "correlation": 247709
    }
  },
  {
    "ph": "f", "id": 247709, "pid": 76337, "tid": -914061504, "ts": 1716454225373570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225421614, "dur": 91,
    "args": {
      "External id": 247710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247710,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247710, "pid": 5, "tid": 7, "ts": 1716454225421614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373575, "dur": 14,
    "args": {
      "External id": 247710,
      "cbid": 211, "correlation": 247710
    }
  },
  {
    "ph": "s", "id": 247710, "pid": 76337, "tid": -914061504, "ts": 1716454225373575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373648, "dur": 1,
    "args": {
      "External id": 247721,
      "cbid": 251, "correlation": 247721
    }
  },
  {
    "ph": "f", "id": 247721, "pid": 76337, "tid": -914061504, "ts": 1716454225373648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373651, "dur": 0,
    "args": {
      "External id": 247722,
      "cbid": 251, "correlation": 247722
    }
  },
  {
    "ph": "f", "id": 247722, "pid": 76337, "tid": -914061504, "ts": 1716454225373651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225421707, "dur": 11,
    "args": {
      "External id": 247723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247723,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 247723, "pid": 5, "tid": 7, "ts": 1716454225421707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373653, "dur": 12,
    "args": {
      "External id": 247723,
      "cbid": 211, "correlation": 247723
    }
  },
  {
    "ph": "s", "id": 247723, "pid": 76337, "tid": -914061504, "ts": 1716454225373653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225421720, "dur": 5,
    "args": {
      "External id": 247725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247725,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 247725, "pid": 5, "tid": 7, "ts": 1716454225421720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373669, "dur": 7,
    "args": {
      "External id": 247725,
      "cbid": 211, "correlation": 247725
    }
  },
  {
    "ph": "s", "id": 247725, "pid": 76337, "tid": -914061504, "ts": 1716454225373669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373728, "dur": 1,
    "args": {
      "External id": 247736,
      "cbid": 251, "correlation": 247736
    }
  },
  {
    "ph": "f", "id": 247736, "pid": 76337, "tid": -914061504, "ts": 1716454225373728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373732, "dur": 0,
    "args": {
      "External id": 247737,
      "cbid": 251, "correlation": 247737
    }
  },
  {
    "ph": "f", "id": 247737, "pid": 76337, "tid": -914061504, "ts": 1716454225373732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225421726, "dur": 7,
    "args": {
      "External id": 247738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247738,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 247738, "pid": 5, "tid": 7, "ts": 1716454225421726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373734, "dur": 12,
    "args": {
      "External id": 247738,
      "cbid": 211, "correlation": 247738
    }
  },
  {
    "ph": "s", "id": 247738, "pid": 76337, "tid": -914061504, "ts": 1716454225373734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225421734, "dur": 4,
    "args": {
      "External id": 247740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247740,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 247740, "pid": 5, "tid": 7, "ts": 1716454225421734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373747, "dur": 5,
    "args": {
      "External id": 247740,
      "cbid": 211, "correlation": 247740
    }
  },
  {
    "ph": "s", "id": 247740, "pid": 76337, "tid": -914061504, "ts": 1716454225373747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225421739, "dur": 93,
    "args": {
      "External id": 247761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247761,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 247761, "pid": 5, "tid": 7, "ts": 1716454225421739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373820, "dur": 12,
    "args": {
      "External id": 247761,
      "cbid": 211, "correlation": 247761
    }
  },
  {
    "ph": "s", "id": 247761, "pid": 76337, "tid": -914061504, "ts": 1716454225373820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373916, "dur": 1,
    "args": {
      "External id": 247779,
      "cbid": 251, "correlation": 247779
    }
  },
  {
    "ph": "f", "id": 247779, "pid": 76337, "tid": -914061504, "ts": 1716454225373916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225421834, "dur": 100,
    "args": {
      "External id": 247781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247781,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247781, "pid": 5, "tid": 7, "ts": 1716454225421834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373921, "dur": 14,
    "args": {
      "External id": 247781,
      "cbid": 211, "correlation": 247781
    }
  },
  {
    "ph": "s", "id": 247781, "pid": 76337, "tid": -914061504, "ts": 1716454225373921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225421935, "dur": 19,
    "args": {
      "External id": 247789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247789,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247789, "pid": 5, "tid": 7, "ts": 1716454225421935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225373999, "dur": 12,
    "args": {
      "External id": 247789,
      "cbid": 211, "correlation": 247789
    }
  },
  {
    "ph": "s", "id": 247789, "pid": 76337, "tid": -914061504, "ts": 1716454225373999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225421955, "dur": 38,
    "args": {
      "External id": 247797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247797, "pid": 5, "tid": 7, "ts": 1716454225421955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374042, "dur": 9,
    "args": {
      "External id": 247797,
      "cbid": 211, "correlation": 247797
    }
  },
  {
    "ph": "s", "id": 247797, "pid": 76337, "tid": -914061504, "ts": 1716454225374042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225421994, "dur": 35,
    "args": {
      "External id": 247819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247819,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247819, "pid": 5, "tid": 7, "ts": 1716454225421994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374095, "dur": 10,
    "args": {
      "External id": 247819,
      "cbid": 211, "correlation": 247819
    }
  },
  {
    "ph": "s", "id": 247819, "pid": 76337, "tid": -914061504, "ts": 1716454225374095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374184, "dur": 1,
    "args": {
      "External id": 247835,
      "cbid": 251, "correlation": 247835
    }
  },
  {
    "ph": "f", "id": 247835, "pid": 76337, "tid": -914061504, "ts": 1716454225374184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374190, "dur": 0,
    "args": {
      "External id": 247837,
      "cbid": 251, "correlation": 247837
    }
  },
  {
    "ph": "f", "id": 247837, "pid": 76337, "tid": -914061504, "ts": 1716454225374190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225422031, "dur": 551,
    "args": {
      "External id": 247838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247838,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 247838, "pid": 5, "tid": 7, "ts": 1716454225422031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374193, "dur": 12,
    "args": {
      "External id": 247838,
      "cbid": 211, "correlation": 247838
    }
  },
  {
    "ph": "s", "id": 247838, "pid": 76337, "tid": -914061504, "ts": 1716454225374193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225422583, "dur": 128,
    "args": {
      "External id": 247846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247846, "pid": 5, "tid": 7, "ts": 1716454225422583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374258, "dur": 12,
    "args": {
      "External id": 247846,
      "cbid": 211, "correlation": 247846
    }
  },
  {
    "ph": "s", "id": 247846, "pid": 76337, "tid": -914061504, "ts": 1716454225374258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225422713, "dur": 129,
    "args": {
      "External id": 247854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247854, "pid": 5, "tid": 7, "ts": 1716454225422713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374290, "dur": 8,
    "args": {
      "External id": 247854,
      "cbid": 211, "correlation": 247854
    }
  },
  {
    "ph": "s", "id": 247854, "pid": 76337, "tid": -914061504, "ts": 1716454225374290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374368, "dur": 1,
    "args": {
      "External id": 247870,
      "cbid": 251, "correlation": 247870
    }
  },
  {
    "ph": "f", "id": 247870, "pid": 76337, "tid": -914061504, "ts": 1716454225374368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225422843, "dur": 308,
    "args": {
      "External id": 247872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247872,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247872, "pid": 5, "tid": 7, "ts": 1716454225422843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374374, "dur": 13,
    "args": {
      "External id": 247872,
      "cbid": 211, "correlation": 247872
    }
  },
  {
    "ph": "s", "id": 247872, "pid": 76337, "tid": -914061504, "ts": 1716454225374374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225423152, "dur": 28,
    "args": {
      "External id": 247880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247880,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247880, "pid": 5, "tid": 7, "ts": 1716454225423152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374416, "dur": 10,
    "args": {
      "External id": 247880,
      "cbid": 211, "correlation": 247880
    }
  },
  {
    "ph": "s", "id": 247880, "pid": 76337, "tid": -914061504, "ts": 1716454225374416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225423182, "dur": 83,
    "args": {
      "External id": 247891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247891, "pid": 5, "tid": 7, "ts": 1716454225423182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374486, "dur": 12,
    "args": {
      "External id": 247891,
      "cbid": 211, "correlation": 247891
    }
  },
  {
    "ph": "s", "id": 247891, "pid": 76337, "tid": -914061504, "ts": 1716454225374486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374549, "dur": 0,
    "args": {
      "External id": 247903,
      "cbid": 317, "correlation": 247903
    }
  },
  {
    "ph": "f", "id": 247903, "pid": 76337, "tid": -914061504, "ts": 1716454225374549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374550, "dur": 0,
    "args": {
      "External id": 247904,
      "cbid": 203, "correlation": 247904
    }
  },
  {
    "ph": "f", "id": 247904, "pid": 76337, "tid": -914061504, "ts": 1716454225374550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374551, "dur": 0,
    "args": {
      "External id": 247905,
      "cbid": 205, "correlation": 247905
    }
  },
  {
    "ph": "f", "id": 247905, "pid": 76337, "tid": -914061504, "ts": 1716454225374551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225423266, "dur": 23,
    "args": {
      "External id": 247909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247909, "pid": 5, "tid": 7, "ts": 1716454225423266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374566, "dur": 12,
    "args": {
      "External id": 247909,
      "cbid": 211, "correlation": 247909
    }
  },
  {
    "ph": "s", "id": 247909, "pid": 76337, "tid": -914061504, "ts": 1716454225374566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225423290, "dur": 121,
    "args": {
      "External id": 247911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247911,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247911, "pid": 5, "tid": 7, "ts": 1716454225423290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374585, "dur": 7,
    "args": {
      "External id": 247911,
      "cbid": 211, "correlation": 247911
    }
  },
  {
    "ph": "s", "id": 247911, "pid": 76337, "tid": -914061504, "ts": 1716454225374585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225423413, "dur": 23,
    "args": {
      "External id": 247913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247913, "pid": 5, "tid": 7, "ts": 1716454225423413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374596, "dur": 5,
    "args": {
      "External id": 247913,
      "cbid": 211, "correlation": 247913
    }
  },
  {
    "ph": "s", "id": 247913, "pid": 76337, "tid": -914061504, "ts": 1716454225374596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225423437, "dur": 33,
    "args": {
      "External id": 247919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247919,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247919, "pid": 5, "tid": 7, "ts": 1716454225423437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374623, "dur": 9,
    "args": {
      "External id": 247919,
      "cbid": 211, "correlation": 247919
    }
  },
  {
    "ph": "s", "id": 247919, "pid": 76337, "tid": -914061504, "ts": 1716454225374623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225423472, "dur": 27,
    "args": {
      "External id": 247927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247927,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247927, "pid": 5, "tid": 7, "ts": 1716454225423472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374655, "dur": 8,
    "args": {
      "External id": 247927,
      "cbid": 211, "correlation": 247927
    }
  },
  {
    "ph": "s", "id": 247927, "pid": 76337, "tid": -914061504, "ts": 1716454225374655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225423500, "dur": 31,
    "args": {
      "External id": 247947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247947,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 247947, "pid": 5, "tid": 7, "ts": 1716454225423500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374726, "dur": 12,
    "args": {
      "External id": 247947,
      "cbid": 211, "correlation": 247947
    }
  },
  {
    "ph": "s", "id": 247947, "pid": 76337, "tid": -914061504, "ts": 1716454225374726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225423533, "dur": 5,
    "args": {
      "External id": 247959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247959,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 247959, "pid": 5, "tid": 7, "ts": 1716454225423533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374748, "dur": 6,
    "args": {
      "External id": 247959,
      "cbid": 211, "correlation": 247959
    }
  },
  {
    "ph": "s", "id": 247959, "pid": 76337, "tid": -914061504, "ts": 1716454225374748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225423539, "dur": 31,
    "args": {
      "External id": 247962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247962, "pid": 5, "tid": 7, "ts": 1716454225423539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374766, "dur": 7,
    "args": {
      "External id": 247962,
      "cbid": 211, "correlation": 247962
    }
  },
  {
    "ph": "s", "id": 247962, "pid": 76337, "tid": -914061504, "ts": 1716454225374766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225423571, "dur": 21,
    "args": {
      "External id": 247971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247971,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247971, "pid": 5, "tid": 7, "ts": 1716454225423571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374805, "dur": 9,
    "args": {
      "External id": 247971,
      "cbid": 211, "correlation": 247971
    }
  },
  {
    "ph": "s", "id": 247971, "pid": 76337, "tid": -914061504, "ts": 1716454225374805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374857, "dur": 0,
    "args": {
      "External id": 247981,
      "cbid": 317, "correlation": 247981
    }
  },
  {
    "ph": "f", "id": 247981, "pid": 76337, "tid": -914061504, "ts": 1716454225374857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374857, "dur": 0,
    "args": {
      "External id": 247982,
      "cbid": 203, "correlation": 247982
    }
  },
  {
    "ph": "f", "id": 247982, "pid": 76337, "tid": -914061504, "ts": 1716454225374857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374858, "dur": 0,
    "args": {
      "External id": 247983,
      "cbid": 205, "correlation": 247983
    }
  },
  {
    "ph": "f", "id": 247983, "pid": 76337, "tid": -914061504, "ts": 1716454225374858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225423594, "dur": 21,
    "args": {
      "External id": 247987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247987, "pid": 5, "tid": 7, "ts": 1716454225423594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374875, "dur": 12,
    "args": {
      "External id": 247987,
      "cbid": 211, "correlation": 247987
    }
  },
  {
    "ph": "s", "id": 247987, "pid": 76337, "tid": -914061504, "ts": 1716454225374875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225423617, "dur": 44,
    "args": {
      "External id": 247989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247989,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247989, "pid": 5, "tid": 7, "ts": 1716454225423617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374890, "dur": 5,
    "args": {
      "External id": 247989,
      "cbid": 211, "correlation": 247989
    }
  },
  {
    "ph": "s", "id": 247989, "pid": 76337, "tid": -914061504, "ts": 1716454225374890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225423662, "dur": 667,
    "args": {
      "External id": 247991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247991,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 247991, "pid": 5, "tid": 7, "ts": 1716454225423662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374901, "dur": 6,
    "args": {
      "External id": 247991,
      "cbid": 211, "correlation": 247991
    }
  },
  {
    "ph": "s", "id": 247991, "pid": 76337, "tid": -914061504, "ts": 1716454225374901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225424331, "dur": 23,
    "args": {
      "External id": 247993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247993,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247993, "pid": 5, "tid": 7, "ts": 1716454225424331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374911, "dur": 5,
    "args": {
      "External id": 247993,
      "cbid": 211, "correlation": 247993
    }
  },
  {
    "ph": "s", "id": 247993, "pid": 76337, "tid": -914061504, "ts": 1716454225374911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225424355, "dur": 34,
    "args": {
      "External id": 247999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 247999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 247999, "pid": 5, "tid": 7, "ts": 1716454225424355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374938, "dur": 8,
    "args": {
      "External id": 247999,
      "cbid": 211, "correlation": 247999
    }
  },
  {
    "ph": "s", "id": 247999, "pid": 76337, "tid": -914061504, "ts": 1716454225374938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225424390, "dur": 3,
    "args": {
      "External id": 248007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248007,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 248007, "pid": 5, "tid": 7, "ts": 1716454225424390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225374990, "dur": 10,
    "args": {
      "External id": 248007,
      "cbid": 211, "correlation": 248007
    }
  },
  {
    "ph": "s", "id": 248007, "pid": 76337, "tid": -914061504, "ts": 1716454225374990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375060, "dur": 1,
    "args": {
      "External id": 248023,
      "cbid": 251, "correlation": 248023
    }
  },
  {
    "ph": "f", "id": 248023, "pid": 76337, "tid": -914061504, "ts": 1716454225375060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375065, "dur": 0,
    "args": {
      "External id": 248025,
      "cbid": 251, "correlation": 248025
    }
  },
  {
    "ph": "f", "id": 248025, "pid": 76337, "tid": -914061504, "ts": 1716454225375065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225424395, "dur": 13,
    "args": {
      "External id": 248026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248026,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 248026, "pid": 5, "tid": 7, "ts": 1716454225424395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375067, "dur": 12,
    "args": {
      "External id": 248026,
      "cbid": 211, "correlation": 248026
    }
  },
  {
    "ph": "s", "id": 248026, "pid": 76337, "tid": -914061504, "ts": 1716454225375067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225424409, "dur": 5,
    "args": {
      "External id": 248028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248028,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 248028, "pid": 5, "tid": 7, "ts": 1716454225424409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375080, "dur": 6,
    "args": {
      "External id": 248028,
      "cbid": 211, "correlation": 248028
    }
  },
  {
    "ph": "s", "id": 248028, "pid": 76337, "tid": -914061504, "ts": 1716454225375080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225424415, "dur": 29,
    "args": {
      "External id": 248038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248038, "pid": 5, "tid": 7, "ts": 1716454225424415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375140, "dur": 12,
    "args": {
      "External id": 248038,
      "cbid": 211, "correlation": 248038
    }
  },
  {
    "ph": "s", "id": 248038, "pid": 76337, "tid": -914061504, "ts": 1716454225375140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225424446, "dur": 31,
    "args": {
      "External id": 248058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248058,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 248058, "pid": 5, "tid": 7, "ts": 1716454225424446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375205, "dur": 10,
    "args": {
      "External id": 248058,
      "cbid": 211, "correlation": 248058
    }
  },
  {
    "ph": "s", "id": 248058, "pid": 76337, "tid": -914061504, "ts": 1716454225375205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225424478, "dur": 5,
    "args": {
      "External id": 248070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248070,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 248070, "pid": 5, "tid": 7, "ts": 1716454225424478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375225, "dur": 6,
    "args": {
      "External id": 248070,
      "cbid": 211, "correlation": 248070
    }
  },
  {
    "ph": "s", "id": 248070, "pid": 76337, "tid": -914061504, "ts": 1716454225375225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225424484, "dur": 31,
    "args": {
      "External id": 248073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248073, "pid": 5, "tid": 7, "ts": 1716454225424484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375245, "dur": 6,
    "args": {
      "External id": 248073,
      "cbid": 211, "correlation": 248073
    }
  },
  {
    "ph": "s", "id": 248073, "pid": 76337, "tid": -914061504, "ts": 1716454225375245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225424516, "dur": 21,
    "args": {
      "External id": 248082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248082,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248082, "pid": 5, "tid": 7, "ts": 1716454225424516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375286, "dur": 10,
    "args": {
      "External id": 248082,
      "cbid": 211, "correlation": 248082
    }
  },
  {
    "ph": "s", "id": 248082, "pid": 76337, "tid": -914061504, "ts": 1716454225375286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375348, "dur": 0,
    "args": {
      "External id": 248092,
      "cbid": 317, "correlation": 248092
    }
  },
  {
    "ph": "f", "id": 248092, "pid": 76337, "tid": -914061504, "ts": 1716454225375348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375349, "dur": 0,
    "args": {
      "External id": 248093,
      "cbid": 203, "correlation": 248093
    }
  },
  {
    "ph": "f", "id": 248093, "pid": 76337, "tid": -914061504, "ts": 1716454225375349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375350, "dur": 0,
    "args": {
      "External id": 248094,
      "cbid": 205, "correlation": 248094
    }
  },
  {
    "ph": "f", "id": 248094, "pid": 76337, "tid": -914061504, "ts": 1716454225375350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225424539, "dur": 22,
    "args": {
      "External id": 248098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248098, "pid": 5, "tid": 7, "ts": 1716454225424539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375366, "dur": 12,
    "args": {
      "External id": 248098,
      "cbid": 211, "correlation": 248098
    }
  },
  {
    "ph": "s", "id": 248098, "pid": 76337, "tid": -914061504, "ts": 1716454225375366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225424562, "dur": 44,
    "args": {
      "External id": 248100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248100, "pid": 5, "tid": 7, "ts": 1716454225424562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375380, "dur": 5,
    "args": {
      "External id": 248100,
      "cbid": 211, "correlation": 248100
    }
  },
  {
    "ph": "s", "id": 248100, "pid": 76337, "tid": -914061504, "ts": 1716454225375380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225424608, "dur": 655,
    "args": {
      "External id": 248102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248102,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248102, "pid": 5, "tid": 7, "ts": 1716454225424608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375392, "dur": 6,
    "args": {
      "External id": 248102,
      "cbid": 211, "correlation": 248102
    }
  },
  {
    "ph": "s", "id": 248102, "pid": 76337, "tid": -914061504, "ts": 1716454225375392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225425264, "dur": 21,
    "args": {
      "External id": 248104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248104,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248104, "pid": 5, "tid": 7, "ts": 1716454225425264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375401, "dur": 5,
    "args": {
      "External id": 248104,
      "cbid": 211, "correlation": 248104
    }
  },
  {
    "ph": "s", "id": 248104, "pid": 76337, "tid": -914061504, "ts": 1716454225375401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225425286, "dur": 33,
    "args": {
      "External id": 248110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248110, "pid": 5, "tid": 7, "ts": 1716454225425286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375428, "dur": 9,
    "args": {
      "External id": 248110,
      "cbid": 211, "correlation": 248110
    }
  },
  {
    "ph": "s", "id": 248110, "pid": 76337, "tid": -914061504, "ts": 1716454225375428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225425321, "dur": 27,
    "args": {
      "External id": 248118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248118,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248118, "pid": 5, "tid": 7, "ts": 1716454225425321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375461, "dur": 8,
    "args": {
      "External id": 248118,
      "cbid": 211, "correlation": 248118
    }
  },
  {
    "ph": "s", "id": 248118, "pid": 76337, "tid": -914061504, "ts": 1716454225375461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225425349, "dur": 19,
    "args": {
      "External id": 248126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248126,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248126, "pid": 5, "tid": 7, "ts": 1716454225425349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375492, "dur": 8,
    "args": {
      "External id": 248126,
      "cbid": 211, "correlation": 248126
    }
  },
  {
    "ph": "s", "id": 248126, "pid": 76337, "tid": -914061504, "ts": 1716454225375492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225425369, "dur": 31,
    "args": {
      "External id": 248146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248146,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 248146, "pid": 5, "tid": 7, "ts": 1716454225425369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375570, "dur": 12,
    "args": {
      "External id": 248146,
      "cbid": 211, "correlation": 248146
    }
  },
  {
    "ph": "s", "id": 248146, "pid": 76337, "tid": -914061504, "ts": 1716454225375570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225425401, "dur": 4,
    "args": {
      "External id": 248158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248158,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 248158, "pid": 5, "tid": 7, "ts": 1716454225425401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375593, "dur": 6,
    "args": {
      "External id": 248158,
      "cbid": 211, "correlation": 248158
    }
  },
  {
    "ph": "s", "id": 248158, "pid": 76337, "tid": -914061504, "ts": 1716454225375593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225425407, "dur": 30,
    "args": {
      "External id": 248161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248161, "pid": 5, "tid": 7, "ts": 1716454225425407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375611, "dur": 6,
    "args": {
      "External id": 248161,
      "cbid": 211, "correlation": 248161
    }
  },
  {
    "ph": "s", "id": 248161, "pid": 76337, "tid": -914061504, "ts": 1716454225375611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375668, "dur": 0,
    "args": {
      "External id": 248172,
      "cbid": 317, "correlation": 248172
    }
  },
  {
    "ph": "f", "id": 248172, "pid": 76337, "tid": -914061504, "ts": 1716454225375668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375669, "dur": 0,
    "args": {
      "External id": 248173,
      "cbid": 203, "correlation": 248173
    }
  },
  {
    "ph": "f", "id": 248173, "pid": 76337, "tid": -914061504, "ts": 1716454225375669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375670, "dur": 0,
    "args": {
      "External id": 248174,
      "cbid": 205, "correlation": 248174
    }
  },
  {
    "ph": "f", "id": 248174, "pid": 76337, "tid": -914061504, "ts": 1716454225375670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225425439, "dur": 22,
    "args": {
      "External id": 248178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248178, "pid": 5, "tid": 7, "ts": 1716454225425439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375683, "dur": 12,
    "args": {
      "External id": 248178,
      "cbid": 211, "correlation": 248178
    }
  },
  {
    "ph": "s", "id": 248178, "pid": 76337, "tid": -914061504, "ts": 1716454225375683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225425462, "dur": 118,
    "args": {
      "External id": 248180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248180,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248180, "pid": 5, "tid": 7, "ts": 1716454225425462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375701, "dur": 6,
    "args": {
      "External id": 248180,
      "cbid": 211, "correlation": 248180
    }
  },
  {
    "ph": "s", "id": 248180, "pid": 76337, "tid": -914061504, "ts": 1716454225375701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225425581, "dur": 21,
    "args": {
      "External id": 248182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248182,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248182, "pid": 5, "tid": 7, "ts": 1716454225425581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375711, "dur": 5,
    "args": {
      "External id": 248182,
      "cbid": 211, "correlation": 248182
    }
  },
  {
    "ph": "s", "id": 248182, "pid": 76337, "tid": -914061504, "ts": 1716454225375711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225425604, "dur": 33,
    "args": {
      "External id": 248188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248188,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248188, "pid": 5, "tid": 7, "ts": 1716454225425604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375742, "dur": 9,
    "args": {
      "External id": 248188,
      "cbid": 211, "correlation": 248188
    }
  },
  {
    "ph": "s", "id": 248188, "pid": 76337, "tid": -914061504, "ts": 1716454225375742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225425638, "dur": 189,
    "args": {
      "External id": 248197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248197,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248197, "pid": 5, "tid": 7, "ts": 1716454225425638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375825, "dur": 14,
    "args": {
      "External id": 248197,
      "cbid": 211, "correlation": 248197
    }
  },
  {
    "ph": "s", "id": 248197, "pid": 76337, "tid": -914061504, "ts": 1716454225375825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225425828, "dur": 67,
    "args": {
      "External id": 248219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248219,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248219, "pid": 5, "tid": 7, "ts": 1716454225425828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375882, "dur": 10,
    "args": {
      "External id": 248219,
      "cbid": 211, "correlation": 248219
    }
  },
  {
    "ph": "s", "id": 248219, "pid": 76337, "tid": -914061504, "ts": 1716454225375882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375969, "dur": 1,
    "args": {
      "External id": 248230,
      "cbid": 251, "correlation": 248230
    }
  },
  {
    "ph": "f", "id": 248230, "pid": 76337, "tid": -914061504, "ts": 1716454225375969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225425896, "dur": 157,
    "args": {
      "External id": 248231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248231,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248231, "pid": 5, "tid": 7, "ts": 1716454225425896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225375983, "dur": 14,
    "args": {
      "External id": 248231,
      "cbid": 211, "correlation": 248231
    }
  },
  {
    "ph": "s", "id": 248231, "pid": 76337, "tid": -914061504, "ts": 1716454225375983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376055, "dur": 1,
    "args": {
      "External id": 248242,
      "cbid": 251, "correlation": 248242
    }
  },
  {
    "ph": "f", "id": 248242, "pid": 76337, "tid": -914061504, "ts": 1716454225376055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225426054, "dur": 150,
    "args": {
      "External id": 248243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248243,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248243, "pid": 5, "tid": 7, "ts": 1716454225426054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376060, "dur": 11,
    "args": {
      "External id": 248243,
      "cbid": 211, "correlation": 248243
    }
  },
  {
    "ph": "s", "id": 248243, "pid": 76337, "tid": -914061504, "ts": 1716454225376060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376129, "dur": 1,
    "args": {
      "External id": 248254,
      "cbid": 251, "correlation": 248254
    }
  },
  {
    "ph": "f", "id": 248254, "pid": 76337, "tid": -914061504, "ts": 1716454225376129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225426205, "dur": 149,
    "args": {
      "External id": 248255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248255,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248255, "pid": 5, "tid": 7, "ts": 1716454225426205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376133, "dur": 12,
    "args": {
      "External id": 248255,
      "cbid": 211, "correlation": 248255
    }
  },
  {
    "ph": "s", "id": 248255, "pid": 76337, "tid": -914061504, "ts": 1716454225376133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225426356, "dur": 1989,
    "args": {
      "External id": 248276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248276,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 248276, "pid": 5, "tid": 7, "ts": 1716454225426356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376215, "dur": 13,
    "args": {
      "External id": 248276,
      "cbid": 211, "correlation": 248276
    }
  },
  {
    "ph": "s", "id": 248276, "pid": 76337, "tid": -914061504, "ts": 1716454225376215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376314, "dur": 1,
    "args": {
      "External id": 248294,
      "cbid": 251, "correlation": 248294
    }
  },
  {
    "ph": "f", "id": 248294, "pid": 76337, "tid": -914061504, "ts": 1716454225376314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225428346, "dur": 150,
    "args": {
      "External id": 248296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248296,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 248296, "pid": 5, "tid": 7, "ts": 1716454225428346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376319, "dur": 13,
    "args": {
      "External id": 248296,
      "cbid": 211, "correlation": 248296
    }
  },
  {
    "ph": "s", "id": 248296, "pid": 76337, "tid": -914061504, "ts": 1716454225376319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225428498, "dur": 35,
    "args": {
      "External id": 248304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248304,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248304, "pid": 5, "tid": 7, "ts": 1716454225428498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376389, "dur": 13,
    "args": {
      "External id": 248304,
      "cbid": 211, "correlation": 248304
    }
  },
  {
    "ph": "s", "id": 248304, "pid": 76337, "tid": -914061504, "ts": 1716454225376389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225428534, "dur": 50,
    "args": {
      "External id": 248312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248312,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248312, "pid": 5, "tid": 7, "ts": 1716454225428534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376428, "dur": 8,
    "args": {
      "External id": 248312,
      "cbid": 211, "correlation": 248312
    }
  },
  {
    "ph": "s", "id": 248312, "pid": 76337, "tid": -914061504, "ts": 1716454225376428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225428586, "dur": 31,
    "args": {
      "External id": 248323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248323, "pid": 5, "tid": 7, "ts": 1716454225428586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376505, "dur": 13,
    "args": {
      "External id": 248323,
      "cbid": 211, "correlation": 248323
    }
  },
  {
    "ph": "s", "id": 248323, "pid": 76337, "tid": -914061504, "ts": 1716454225376505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225428619, "dur": 36,
    "args": {
      "External id": 248345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248345,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248345, "pid": 5, "tid": 7, "ts": 1716454225428619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376538, "dur": 9,
    "args": {
      "External id": 248345,
      "cbid": 211, "correlation": 248345
    }
  },
  {
    "ph": "s", "id": 248345, "pid": 76337, "tid": -914061504, "ts": 1716454225376538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376626, "dur": 1,
    "args": {
      "External id": 248356,
      "cbid": 251, "correlation": 248356
    }
  },
  {
    "ph": "f", "id": 248356, "pid": 76337, "tid": -914061504, "ts": 1716454225376626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225428656, "dur": 92,
    "args": {
      "External id": 248357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248357,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248357, "pid": 5, "tid": 7, "ts": 1716454225428656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376631, "dur": 14,
    "args": {
      "External id": 248357,
      "cbid": 211, "correlation": 248357
    }
  },
  {
    "ph": "s", "id": 248357, "pid": 76337, "tid": -914061504, "ts": 1716454225376631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376704, "dur": 1,
    "args": {
      "External id": 248368,
      "cbid": 251, "correlation": 248368
    }
  },
  {
    "ph": "f", "id": 248368, "pid": 76337, "tid": -914061504, "ts": 1716454225376704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376708, "dur": 0,
    "args": {
      "External id": 248369,
      "cbid": 251, "correlation": 248369
    }
  },
  {
    "ph": "f", "id": 248369, "pid": 76337, "tid": -914061504, "ts": 1716454225376708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225428749, "dur": 11,
    "args": {
      "External id": 248370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248370,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 248370, "pid": 5, "tid": 7, "ts": 1716454225428749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376709, "dur": 12,
    "args": {
      "External id": 248370,
      "cbid": 211, "correlation": 248370
    }
  },
  {
    "ph": "s", "id": 248370, "pid": 76337, "tid": -914061504, "ts": 1716454225376709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225428762, "dur": 5,
    "args": {
      "External id": 248372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248372,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 248372, "pid": 5, "tid": 7, "ts": 1716454225428762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376723, "dur": 6,
    "args": {
      "External id": 248372,
      "cbid": 211, "correlation": 248372
    }
  },
  {
    "ph": "s", "id": 248372, "pid": 76337, "tid": -914061504, "ts": 1716454225376723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376781, "dur": 1,
    "args": {
      "External id": 248383,
      "cbid": 251, "correlation": 248383
    }
  },
  {
    "ph": "f", "id": 248383, "pid": 76337, "tid": -914061504, "ts": 1716454225376781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376785, "dur": 0,
    "args": {
      "External id": 248384,
      "cbid": 251, "correlation": 248384
    }
  },
  {
    "ph": "f", "id": 248384, "pid": 76337, "tid": -914061504, "ts": 1716454225376785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225428768, "dur": 7,
    "args": {
      "External id": 248385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248385,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 248385, "pid": 5, "tid": 7, "ts": 1716454225428768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376786, "dur": 12,
    "args": {
      "External id": 248385,
      "cbid": 211, "correlation": 248385
    }
  },
  {
    "ph": "s", "id": 248385, "pid": 76337, "tid": -914061504, "ts": 1716454225376786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225428776, "dur": 3,
    "args": {
      "External id": 248387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248387,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 248387, "pid": 5, "tid": 7, "ts": 1716454225428776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376800, "dur": 6,
    "args": {
      "External id": 248387,
      "cbid": 211, "correlation": 248387
    }
  },
  {
    "ph": "s", "id": 248387, "pid": 76337, "tid": -914061504, "ts": 1716454225376800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225428781, "dur": 93,
    "args": {
      "External id": 248408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248408,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 248408, "pid": 5, "tid": 7, "ts": 1716454225428781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376874, "dur": 12,
    "args": {
      "External id": 248408,
      "cbid": 211, "correlation": 248408
    }
  },
  {
    "ph": "s", "id": 248408, "pid": 76337, "tid": -914061504, "ts": 1716454225376874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376970, "dur": 1,
    "args": {
      "External id": 248426,
      "cbid": 251, "correlation": 248426
    }
  },
  {
    "ph": "f", "id": 248426, "pid": 76337, "tid": -914061504, "ts": 1716454225376970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225428875, "dur": 100,
    "args": {
      "External id": 248428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248428,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248428, "pid": 5, "tid": 7, "ts": 1716454225428875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225376983, "dur": 14,
    "args": {
      "External id": 248428,
      "cbid": 211, "correlation": 248428
    }
  },
  {
    "ph": "s", "id": 248428, "pid": 76337, "tid": -914061504, "ts": 1716454225376983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225428976, "dur": 19,
    "args": {
      "External id": 248436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248436,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248436, "pid": 5, "tid": 7, "ts": 1716454225428976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377054, "dur": 12,
    "args": {
      "External id": 248436,
      "cbid": 211, "correlation": 248436
    }
  },
  {
    "ph": "s", "id": 248436, "pid": 76337, "tid": -914061504, "ts": 1716454225377054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225428997, "dur": 38,
    "args": {
      "External id": 248444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248444,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248444, "pid": 5, "tid": 7, "ts": 1716454225428997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377096, "dur": 9,
    "args": {
      "External id": 248444,
      "cbid": 211, "correlation": 248444
    }
  },
  {
    "ph": "s", "id": 248444, "pid": 76337, "tid": -914061504, "ts": 1716454225377096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225429036, "dur": 35,
    "args": {
      "External id": 248466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248466,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248466, "pid": 5, "tid": 7, "ts": 1716454225429036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377149, "dur": 11,
    "args": {
      "External id": 248466,
      "cbid": 211, "correlation": 248466
    }
  },
  {
    "ph": "s", "id": 248466, "pid": 76337, "tid": -914061504, "ts": 1716454225377149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377240, "dur": 1,
    "args": {
      "External id": 248482,
      "cbid": 251, "correlation": 248482
    }
  },
  {
    "ph": "f", "id": 248482, "pid": 76337, "tid": -914061504, "ts": 1716454225377240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377246, "dur": 0,
    "args": {
      "External id": 248484,
      "cbid": 251, "correlation": 248484
    }
  },
  {
    "ph": "f", "id": 248484, "pid": 76337, "tid": -914061504, "ts": 1716454225377246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225429072, "dur": 549,
    "args": {
      "External id": 248485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248485,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 248485, "pid": 5, "tid": 7, "ts": 1716454225429072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377250, "dur": 13,
    "args": {
      "External id": 248485,
      "cbid": 211, "correlation": 248485
    }
  },
  {
    "ph": "s", "id": 248485, "pid": 76337, "tid": -914061504, "ts": 1716454225377250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225429622, "dur": 128,
    "args": {
      "External id": 248493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248493,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248493, "pid": 5, "tid": 7, "ts": 1716454225429622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377315, "dur": 13,
    "args": {
      "External id": 248493,
      "cbid": 211, "correlation": 248493
    }
  },
  {
    "ph": "s", "id": 248493, "pid": 76337, "tid": -914061504, "ts": 1716454225377315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225429752, "dur": 128,
    "args": {
      "External id": 248501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248501, "pid": 5, "tid": 7, "ts": 1716454225429752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377346, "dur": 8,
    "args": {
      "External id": 248501,
      "cbid": 211, "correlation": 248501
    }
  },
  {
    "ph": "s", "id": 248501, "pid": 76337, "tid": -914061504, "ts": 1716454225377346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377422, "dur": 1,
    "args": {
      "External id": 248517,
      "cbid": 251, "correlation": 248517
    }
  },
  {
    "ph": "f", "id": 248517, "pid": 76337, "tid": -914061504, "ts": 1716454225377422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225429881, "dur": 311,
    "args": {
      "External id": 248519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248519,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248519, "pid": 5, "tid": 7, "ts": 1716454225429881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377427, "dur": 12,
    "args": {
      "External id": 248519,
      "cbid": 211, "correlation": 248519
    }
  },
  {
    "ph": "s", "id": 248519, "pid": 76337, "tid": -914061504, "ts": 1716454225377427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225430194, "dur": 27,
    "args": {
      "External id": 248527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248527,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248527, "pid": 5, "tid": 7, "ts": 1716454225430194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377469, "dur": 9,
    "args": {
      "External id": 248527,
      "cbid": 211, "correlation": 248527
    }
  },
  {
    "ph": "s", "id": 248527, "pid": 76337, "tid": -914061504, "ts": 1716454225377469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225430222, "dur": 83,
    "args": {
      "External id": 248538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248538, "pid": 5, "tid": 7, "ts": 1716454225430222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377536, "dur": 13,
    "args": {
      "External id": 248538,
      "cbid": 211, "correlation": 248538
    }
  },
  {
    "ph": "s", "id": 248538, "pid": 76337, "tid": -914061504, "ts": 1716454225377536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377599, "dur": 0,
    "args": {
      "External id": 248550,
      "cbid": 317, "correlation": 248550
    }
  },
  {
    "ph": "f", "id": 248550, "pid": 76337, "tid": -914061504, "ts": 1716454225377599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377600, "dur": 0,
    "args": {
      "External id": 248551,
      "cbid": 203, "correlation": 248551
    }
  },
  {
    "ph": "f", "id": 248551, "pid": 76337, "tid": -914061504, "ts": 1716454225377600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377601, "dur": 0,
    "args": {
      "External id": 248552,
      "cbid": 205, "correlation": 248552
    }
  },
  {
    "ph": "f", "id": 248552, "pid": 76337, "tid": -914061504, "ts": 1716454225377601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430306, "dur": 23,
    "args": {
      "External id": 248556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248556,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248556, "pid": 5, "tid": 7, "ts": 1716454225430306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377615, "dur": 12,
    "args": {
      "External id": 248556,
      "cbid": 211, "correlation": 248556
    }
  },
  {
    "ph": "s", "id": 248556, "pid": 76337, "tid": -914061504, "ts": 1716454225377615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225430330, "dur": 124,
    "args": {
      "External id": 248558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248558,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248558, "pid": 5, "tid": 7, "ts": 1716454225430330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377633, "dur": 6,
    "args": {
      "External id": 248558,
      "cbid": 211, "correlation": 248558
    }
  },
  {
    "ph": "s", "id": 248558, "pid": 76337, "tid": -914061504, "ts": 1716454225377633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430455, "dur": 22,
    "args": {
      "External id": 248560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248560,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248560, "pid": 5, "tid": 7, "ts": 1716454225430455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377643, "dur": 6,
    "args": {
      "External id": 248560,
      "cbid": 211, "correlation": 248560
    }
  },
  {
    "ph": "s", "id": 248560, "pid": 76337, "tid": -914061504, "ts": 1716454225377643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225430478, "dur": 32,
    "args": {
      "External id": 248566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248566, "pid": 5, "tid": 7, "ts": 1716454225430478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377671, "dur": 9,
    "args": {
      "External id": 248566,
      "cbid": 211, "correlation": 248566
    }
  },
  {
    "ph": "s", "id": 248566, "pid": 76337, "tid": -914061504, "ts": 1716454225377671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225430512, "dur": 27,
    "args": {
      "External id": 248574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248574,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248574, "pid": 5, "tid": 7, "ts": 1716454225430512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377704, "dur": 8,
    "args": {
      "External id": 248574,
      "cbid": 211, "correlation": 248574
    }
  },
  {
    "ph": "s", "id": 248574, "pid": 76337, "tid": -914061504, "ts": 1716454225377704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377775, "dur": 0,
    "args": {
      "External id": 248584,
      "cbid": 317, "correlation": 248584
    }
  },
  {
    "ph": "f", "id": 248584, "pid": 76337, "tid": -914061504, "ts": 1716454225377775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377776, "dur": 0,
    "args": {
      "External id": 248585,
      "cbid": 203, "correlation": 248585
    }
  },
  {
    "ph": "f", "id": 248585, "pid": 76337, "tid": -914061504, "ts": 1716454225377776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377777, "dur": 0,
    "args": {
      "External id": 248586,
      "cbid": 205, "correlation": 248586
    }
  },
  {
    "ph": "f", "id": 248586, "pid": 76337, "tid": -914061504, "ts": 1716454225377777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430541, "dur": 22,
    "args": {
      "External id": 248590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248590,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248590, "pid": 5, "tid": 7, "ts": 1716454225430541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377792, "dur": 13,
    "args": {
      "External id": 248590,
      "cbid": 211, "correlation": 248590
    }
  },
  {
    "ph": "s", "id": 248590, "pid": 76337, "tid": -914061504, "ts": 1716454225377792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430564, "dur": 45,
    "args": {
      "External id": 248592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248592, "pid": 5, "tid": 7, "ts": 1716454225430564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377807, "dur": 5,
    "args": {
      "External id": 248592,
      "cbid": 211, "correlation": 248592
    }
  },
  {
    "ph": "s", "id": 248592, "pid": 76337, "tid": -914061504, "ts": 1716454225377807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225430610, "dur": 236,
    "args": {
      "External id": 248594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248594,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 248594, "pid": 5, "tid": 7, "ts": 1716454225430610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377819, "dur": 6,
    "args": {
      "External id": 248594,
      "cbid": 211, "correlation": 248594
    }
  },
  {
    "ph": "s", "id": 248594, "pid": 76337, "tid": -914061504, "ts": 1716454225377819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430848, "dur": 7,
    "args": {
      "External id": 248596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248596, "pid": 5, "tid": 7, "ts": 1716454225430848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377829, "dur": 5,
    "args": {
      "External id": 248596,
      "cbid": 211, "correlation": 248596
    }
  },
  {
    "ph": "s", "id": 248596, "pid": 76337, "tid": -914061504, "ts": 1716454225377829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225430856, "dur": 9,
    "args": {
      "External id": 248602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248602, "pid": 5, "tid": 7, "ts": 1716454225430856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377856, "dur": 8,
    "args": {
      "External id": 248602,
      "cbid": 211, "correlation": 248602
    }
  },
  {
    "ph": "s", "id": 248602, "pid": 76337, "tid": -914061504, "ts": 1716454225377856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430867, "dur": 12,
    "args": {
      "External id": 248622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248622,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 248622, "pid": 5, "tid": 7, "ts": 1716454225430867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377948, "dur": 12,
    "args": {
      "External id": 248622,
      "cbid": 211, "correlation": 248622
    }
  },
  {
    "ph": "s", "id": 248622, "pid": 76337, "tid": -914061504, "ts": 1716454225377948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225430880, "dur": 5,
    "args": {
      "External id": 248634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248634,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 248634, "pid": 5, "tid": 7, "ts": 1716454225430880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377970, "dur": 15,
    "args": {
      "External id": 248634,
      "cbid": 211, "correlation": 248634
    }
  },
  {
    "ph": "s", "id": 248634, "pid": 76337, "tid": -914061504, "ts": 1716454225377970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225430887, "dur": 12,
    "args": {
      "External id": 248637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248637, "pid": 5, "tid": 7, "ts": 1716454225430887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225377998, "dur": 7,
    "args": {
      "External id": 248637,
      "cbid": 211, "correlation": 248637
    }
  },
  {
    "ph": "s", "id": 248637, "pid": 76337, "tid": -914061504, "ts": 1716454225377998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225430900, "dur": 7,
    "args": {
      "External id": 248646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248646,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248646, "pid": 5, "tid": 7, "ts": 1716454225430900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378038, "dur": 10,
    "args": {
      "External id": 248646,
      "cbid": 211, "correlation": 248646
    }
  },
  {
    "ph": "s", "id": 248646, "pid": 76337, "tid": -914061504, "ts": 1716454225378038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378090, "dur": 0,
    "args": {
      "External id": 248656,
      "cbid": 317, "correlation": 248656
    }
  },
  {
    "ph": "f", "id": 248656, "pid": 76337, "tid": -914061504, "ts": 1716454225378090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378091, "dur": 0,
    "args": {
      "External id": 248657,
      "cbid": 203, "correlation": 248657
    }
  },
  {
    "ph": "f", "id": 248657, "pid": 76337, "tid": -914061504, "ts": 1716454225378091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378091, "dur": 0,
    "args": {
      "External id": 248658,
      "cbid": 205, "correlation": 248658
    }
  },
  {
    "ph": "f", "id": 248658, "pid": 76337, "tid": -914061504, "ts": 1716454225378091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430909, "dur": 6,
    "args": {
      "External id": 248662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248662,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248662, "pid": 5, "tid": 7, "ts": 1716454225430909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378108, "dur": 11,
    "args": {
      "External id": 248662,
      "cbid": 211, "correlation": 248662
    }
  },
  {
    "ph": "s", "id": 248662, "pid": 76337, "tid": -914061504, "ts": 1716454225378108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225430916, "dur": 85,
    "args": {
      "External id": 248664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248664,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248664, "pid": 5, "tid": 7, "ts": 1716454225430916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378122, "dur": 5,
    "args": {
      "External id": 248664,
      "cbid": 211, "correlation": 248664
    }
  },
  {
    "ph": "s", "id": 248664, "pid": 76337, "tid": -914061504, "ts": 1716454225378122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225431003, "dur": 1,
    "args": {
      "External id": 248666,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 248666,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 248666, "pid": 5, "tid": 7, "ts": 1716454225431003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378135, "dur": 9,
    "args": {
      "External id": 248666,
      "cbid": 51, "correlation": 248666
    }
  },
  {
    "ph": "s", "id": 248666, "pid": 76337, "tid": -914061504, "ts": 1716454225378135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225431007, "dur": 544,
    "args": {
      "External id": 248667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248667,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248667, "pid": 5, "tid": 7, "ts": 1716454225431007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378145, "dur": 8,
    "args": {
      "External id": 248667,
      "cbid": 211, "correlation": 248667
    }
  },
  {
    "ph": "s", "id": 248667, "pid": 76337, "tid": -914061504, "ts": 1716454225378145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225431553, "dur": 12,
    "args": {
      "External id": 248669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248669,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248669, "pid": 5, "tid": 7, "ts": 1716454225431553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378158, "dur": 5,
    "args": {
      "External id": 248669,
      "cbid": 211, "correlation": 248669
    }
  },
  {
    "ph": "s", "id": 248669, "pid": 76337, "tid": -914061504, "ts": 1716454225378158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225431566, "dur": 15,
    "args": {
      "External id": 248675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248675, "pid": 5, "tid": 7, "ts": 1716454225431566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378185, "dur": 9,
    "args": {
      "External id": 248675,
      "cbid": 211, "correlation": 248675
    }
  },
  {
    "ph": "s", "id": 248675, "pid": 76337, "tid": -914061504, "ts": 1716454225378185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225431582, "dur": 3,
    "args": {
      "External id": 248683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248683,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 248683, "pid": 5, "tid": 7, "ts": 1716454225431582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378229, "dur": 9,
    "args": {
      "External id": 248683,
      "cbid": 211, "correlation": 248683
    }
  },
  {
    "ph": "s", "id": 248683, "pid": 76337, "tid": -914061504, "ts": 1716454225378229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378296, "dur": 1,
    "args": {
      "External id": 248699,
      "cbid": 251, "correlation": 248699
    }
  },
  {
    "ph": "f", "id": 248699, "pid": 76337, "tid": -914061504, "ts": 1716454225378296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378302, "dur": 0,
    "args": {
      "External id": 248701,
      "cbid": 251, "correlation": 248701
    }
  },
  {
    "ph": "f", "id": 248701, "pid": 76337, "tid": -914061504, "ts": 1716454225378302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225431587, "dur": 13,
    "args": {
      "External id": 248702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248702,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248702, "pid": 5, "tid": 7, "ts": 1716454225431587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378303, "dur": 11,
    "args": {
      "External id": 248702,
      "cbid": 211, "correlation": 248702
    }
  },
  {
    "ph": "s", "id": 248702, "pid": 76337, "tid": -914061504, "ts": 1716454225378303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225431601, "dur": 5,
    "args": {
      "External id": 248704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248704,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248704, "pid": 5, "tid": 7, "ts": 1716454225431601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378316, "dur": 6,
    "args": {
      "External id": 248704,
      "cbid": 211, "correlation": 248704
    }
  },
  {
    "ph": "s", "id": 248704, "pid": 76337, "tid": -914061504, "ts": 1716454225378316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225431608, "dur": 17,
    "args": {
      "External id": 248714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248714,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248714, "pid": 5, "tid": 7, "ts": 1716454225431608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378374, "dur": 12,
    "args": {
      "External id": 248714,
      "cbid": 211, "correlation": 248714
    }
  },
  {
    "ph": "s", "id": 248714, "pid": 76337, "tid": -914061504, "ts": 1716454225378374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225431626, "dur": 18,
    "args": {
      "External id": 248734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248734,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 248734, "pid": 5, "tid": 7, "ts": 1716454225431626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378440, "dur": 11,
    "args": {
      "External id": 248734,
      "cbid": 211, "correlation": 248734
    }
  },
  {
    "ph": "s", "id": 248734, "pid": 76337, "tid": -914061504, "ts": 1716454225378440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225431646, "dur": 5,
    "args": {
      "External id": 248746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248746,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 248746, "pid": 5, "tid": 7, "ts": 1716454225431646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378464, "dur": 7,
    "args": {
      "External id": 248746,
      "cbid": 211, "correlation": 248746
    }
  },
  {
    "ph": "s", "id": 248746, "pid": 76337, "tid": -914061504, "ts": 1716454225378464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225431652, "dur": 17,
    "args": {
      "External id": 248749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248749,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248749, "pid": 5, "tid": 7, "ts": 1716454225431652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378484, "dur": 6,
    "args": {
      "External id": 248749,
      "cbid": 211, "correlation": 248749
    }
  },
  {
    "ph": "s", "id": 248749, "pid": 76337, "tid": -914061504, "ts": 1716454225378484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225431670, "dur": 11,
    "args": {
      "External id": 248758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248758,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248758, "pid": 5, "tid": 7, "ts": 1716454225431670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378524, "dur": 10,
    "args": {
      "External id": 248758,
      "cbid": 211, "correlation": 248758
    }
  },
  {
    "ph": "s", "id": 248758, "pid": 76337, "tid": -914061504, "ts": 1716454225378524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378587, "dur": 0,
    "args": {
      "External id": 248768,
      "cbid": 317, "correlation": 248768
    }
  },
  {
    "ph": "f", "id": 248768, "pid": 76337, "tid": -914061504, "ts": 1716454225378587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378588, "dur": 0,
    "args": {
      "External id": 248769,
      "cbid": 203, "correlation": 248769
    }
  },
  {
    "ph": "f", "id": 248769, "pid": 76337, "tid": -914061504, "ts": 1716454225378588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378589, "dur": 0,
    "args": {
      "External id": 248770,
      "cbid": 205, "correlation": 248770
    }
  },
  {
    "ph": "f", "id": 248770, "pid": 76337, "tid": -914061504, "ts": 1716454225378589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225431683, "dur": 11,
    "args": {
      "External id": 248774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248774,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248774, "pid": 5, "tid": 7, "ts": 1716454225431683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378607, "dur": 12,
    "args": {
      "External id": 248774,
      "cbid": 211, "correlation": 248774
    }
  },
  {
    "ph": "s", "id": 248774, "pid": 76337, "tid": -914061504, "ts": 1716454225378607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225431695, "dur": 166,
    "args": {
      "External id": 248776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248776, "pid": 5, "tid": 7, "ts": 1716454225431695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378621, "dur": 5,
    "args": {
      "External id": 248776,
      "cbid": 211, "correlation": 248776
    }
  },
  {
    "ph": "s", "id": 248776, "pid": 76337, "tid": -914061504, "ts": 1716454225378621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225431864, "dur": 1,
    "args": {
      "External id": 248778,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 248778,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 248778, "pid": 5, "tid": 7, "ts": 1716454225431864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378633, "dur": 6,
    "args": {
      "External id": 248778,
      "cbid": 51, "correlation": 248778
    }
  },
  {
    "ph": "s", "id": 248778, "pid": 76337, "tid": -914061504, "ts": 1716454225378633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225431867, "dur": 675,
    "args": {
      "External id": 248779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248779,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248779, "pid": 5, "tid": 7, "ts": 1716454225431867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378641, "dur": 7,
    "args": {
      "External id": 248779,
      "cbid": 211, "correlation": 248779
    }
  },
  {
    "ph": "s", "id": 248779, "pid": 76337, "tid": -914061504, "ts": 1716454225378641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225432544, "dur": 12,
    "args": {
      "External id": 248781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248781,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248781, "pid": 5, "tid": 7, "ts": 1716454225432544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378652, "dur": 5,
    "args": {
      "External id": 248781,
      "cbid": 211, "correlation": 248781
    }
  },
  {
    "ph": "s", "id": 248781, "pid": 76337, "tid": -914061504, "ts": 1716454225378652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225432558, "dur": 15,
    "args": {
      "External id": 248787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248787,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248787, "pid": 5, "tid": 7, "ts": 1716454225432558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378681, "dur": 9,
    "args": {
      "External id": 248787,
      "cbid": 211, "correlation": 248787
    }
  },
  {
    "ph": "s", "id": 248787, "pid": 76337, "tid": -914061504, "ts": 1716454225378681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378738, "dur": 0,
    "args": {
      "External id": 248797,
      "cbid": 317, "correlation": 248797
    }
  },
  {
    "ph": "f", "id": 248797, "pid": 76337, "tid": -914061504, "ts": 1716454225378738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378739, "dur": 0,
    "args": {
      "External id": 248798,
      "cbid": 203, "correlation": 248798
    }
  },
  {
    "ph": "f", "id": 248798, "pid": 76337, "tid": -914061504, "ts": 1716454225378739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378740, "dur": 0,
    "args": {
      "External id": 248799,
      "cbid": 205, "correlation": 248799
    }
  },
  {
    "ph": "f", "id": 248799, "pid": 76337, "tid": -914061504, "ts": 1716454225378740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225432574, "dur": 9,
    "args": {
      "External id": 248803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248803,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248803, "pid": 5, "tid": 7, "ts": 1716454225432574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378755, "dur": 11,
    "args": {
      "External id": 248803,
      "cbid": 211, "correlation": 248803
    }
  },
  {
    "ph": "s", "id": 248803, "pid": 76337, "tid": -914061504, "ts": 1716454225378755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225432584, "dur": 3,
    "args": {
      "External id": 248805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 248805, "pid": 5, "tid": 7, "ts": 1716454225432584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378772, "dur": 6,
    "args": {
      "External id": 248805,
      "cbid": 211, "correlation": 248805
    }
  },
  {
    "ph": "s", "id": 248805, "pid": 76337, "tid": -914061504, "ts": 1716454225378772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378781, "dur": 0,
    "args": {
      "External id": 248806,
      "cbid": 51, "correlation": 248806
    }
  },
  {
    "ph": "s", "id": 248806, "pid": 76337, "tid": -914061504, "ts": 1716454225378781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225432589, "dur": 59,
    "args": {
      "External id": 248807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248807,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 248807, "pid": 5, "tid": 7, "ts": 1716454225432589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378782, "dur": 5,
    "args": {
      "External id": 248807,
      "cbid": 211, "correlation": 248807
    }
  },
  {
    "ph": "s", "id": 248807, "pid": 76337, "tid": -914061504, "ts": 1716454225378782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225432649, "dur": 14,
    "args": {
      "External id": 248812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248812, "pid": 5, "tid": 7, "ts": 1716454225432649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378808, "dur": 8,
    "args": {
      "External id": 248812,
      "cbid": 211, "correlation": 248812
    }
  },
  {
    "ph": "s", "id": 248812, "pid": 76337, "tid": -914061504, "ts": 1716454225378808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225432665, "dur": 13,
    "args": {
      "External id": 248820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248820,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248820, "pid": 5, "tid": 7, "ts": 1716454225432665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378837, "dur": 8,
    "args": {
      "External id": 248820,
      "cbid": 211, "correlation": 248820
    }
  },
  {
    "ph": "s", "id": 248820, "pid": 76337, "tid": -914061504, "ts": 1716454225378837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225432679, "dur": 10,
    "args": {
      "External id": 248828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248828,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248828, "pid": 5, "tid": 7, "ts": 1716454225432679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378866, "dur": 8,
    "args": {
      "External id": 248828,
      "cbid": 211, "correlation": 248828
    }
  },
  {
    "ph": "s", "id": 248828, "pid": 76337, "tid": -914061504, "ts": 1716454225378866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225432691, "dur": 18,
    "args": {
      "External id": 248848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248848,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 248848, "pid": 5, "tid": 7, "ts": 1716454225432691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378945, "dur": 13,
    "args": {
      "External id": 248848,
      "cbid": 211, "correlation": 248848
    }
  },
  {
    "ph": "s", "id": 248848, "pid": 76337, "tid": -914061504, "ts": 1716454225378945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225432710, "dur": 5,
    "args": {
      "External id": 248860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248860,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 248860, "pid": 5, "tid": 7, "ts": 1716454225432710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378968, "dur": 18,
    "args": {
      "External id": 248860,
      "cbid": 211, "correlation": 248860
    }
  },
  {
    "ph": "s", "id": 248860, "pid": 76337, "tid": -914061504, "ts": 1716454225378968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225432716, "dur": 17,
    "args": {
      "External id": 248863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248863, "pid": 5, "tid": 7, "ts": 1716454225432716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225378998, "dur": 8,
    "args": {
      "External id": 248863,
      "cbid": 211, "correlation": 248863
    }
  },
  {
    "ph": "s", "id": 248863, "pid": 76337, "tid": -914061504, "ts": 1716454225378998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379059, "dur": 0,
    "args": {
      "External id": 248874,
      "cbid": 317, "correlation": 248874
    }
  },
  {
    "ph": "f", "id": 248874, "pid": 76337, "tid": -914061504, "ts": 1716454225379059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379060, "dur": 0,
    "args": {
      "External id": 248875,
      "cbid": 203, "correlation": 248875
    }
  },
  {
    "ph": "f", "id": 248875, "pid": 76337, "tid": -914061504, "ts": 1716454225379060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379060, "dur": 0,
    "args": {
      "External id": 248876,
      "cbid": 205, "correlation": 248876
    }
  },
  {
    "ph": "f", "id": 248876, "pid": 76337, "tid": -914061504, "ts": 1716454225379060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225432734, "dur": 11,
    "args": {
      "External id": 248880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248880, "pid": 5, "tid": 7, "ts": 1716454225432734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379076, "dur": 12,
    "args": {
      "External id": 248880,
      "cbid": 211, "correlation": 248880
    }
  },
  {
    "ph": "s", "id": 248880, "pid": 76337, "tid": -914061504, "ts": 1716454225379076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225432747, "dur": 3,
    "args": {
      "External id": 248882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 248882, "pid": 5, "tid": 7, "ts": 1716454225432747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379093, "dur": 6,
    "args": {
      "External id": 248882,
      "cbid": 211, "correlation": 248882
    }
  },
  {
    "ph": "s", "id": 248882, "pid": 76337, "tid": -914061504, "ts": 1716454225379093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379103, "dur": 0,
    "args": {
      "External id": 248883,
      "cbid": 51, "correlation": 248883
    }
  },
  {
    "ph": "s", "id": 248883, "pid": 76337, "tid": -914061504, "ts": 1716454225379103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225432752, "dur": 100,
    "args": {
      "External id": 248884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248884,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 248884, "pid": 5, "tid": 7, "ts": 1716454225432752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379104, "dur": 6,
    "args": {
      "External id": 248884,
      "cbid": 211, "correlation": 248884
    }
  },
  {
    "ph": "s", "id": 248884, "pid": 76337, "tid": -914061504, "ts": 1716454225379104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225432853, "dur": 16,
    "args": {
      "External id": 248889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248889,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248889, "pid": 5, "tid": 7, "ts": 1716454225432853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379132, "dur": 9,
    "args": {
      "External id": 248889,
      "cbid": 211, "correlation": 248889
    }
  },
  {
    "ph": "s", "id": 248889, "pid": 76337, "tid": -914061504, "ts": 1716454225379132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225432871, "dur": 84,
    "args": {
      "External id": 248898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248898,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248898, "pid": 5, "tid": 7, "ts": 1716454225432871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379215, "dur": 14,
    "args": {
      "External id": 248898,
      "cbid": 211, "correlation": 248898
    }
  },
  {
    "ph": "s", "id": 248898, "pid": 76337, "tid": -914061504, "ts": 1716454225379215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225432956, "dur": 31,
    "args": {
      "External id": 248920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248920,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 248920, "pid": 5, "tid": 7, "ts": 1716454225432956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379272, "dur": 11,
    "args": {
      "External id": 248920,
      "cbid": 211, "correlation": 248920
    }
  },
  {
    "ph": "s", "id": 248920, "pid": 76337, "tid": -914061504, "ts": 1716454225379272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379362, "dur": 1,
    "args": {
      "External id": 248931,
      "cbid": 251, "correlation": 248931
    }
  },
  {
    "ph": "f", "id": 248931, "pid": 76337, "tid": -914061504, "ts": 1716454225379362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225432989, "dur": 145,
    "args": {
      "External id": 248932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248932,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248932, "pid": 5, "tid": 7, "ts": 1716454225432989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379368, "dur": 13,
    "args": {
      "External id": 248932,
      "cbid": 211, "correlation": 248932
    }
  },
  {
    "ph": "s", "id": 248932, "pid": 76337, "tid": -914061504, "ts": 1716454225379368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379438, "dur": 1,
    "args": {
      "External id": 248943,
      "cbid": 251, "correlation": 248943
    }
  },
  {
    "ph": "f", "id": 248943, "pid": 76337, "tid": -914061504, "ts": 1716454225379438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225433135, "dur": 159,
    "args": {
      "External id": 248944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248944,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248944, "pid": 5, "tid": 7, "ts": 1716454225433135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379442, "dur": 12,
    "args": {
      "External id": 248944,
      "cbid": 211, "correlation": 248944
    }
  },
  {
    "ph": "s", "id": 248944, "pid": 76337, "tid": -914061504, "ts": 1716454225379442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379508, "dur": 1,
    "args": {
      "External id": 248955,
      "cbid": 251, "correlation": 248955
    }
  },
  {
    "ph": "f", "id": 248955, "pid": 76337, "tid": -914061504, "ts": 1716454225379508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225433295, "dur": 161,
    "args": {
      "External id": 248956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248956,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248956, "pid": 5, "tid": 7, "ts": 1716454225433295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379512, "dur": 11,
    "args": {
      "External id": 248956,
      "cbid": 211, "correlation": 248956
    }
  },
  {
    "ph": "s", "id": 248956, "pid": 76337, "tid": -914061504, "ts": 1716454225379512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225433458, "dur": 343,
    "args": {
      "External id": 248981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 248981,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 248981, "pid": 5, "tid": 7, "ts": 1716454225433458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379602, "dur": 13,
    "args": {
      "External id": 248981,
      "cbid": 211, "correlation": 248981
    }
  },
  {
    "ph": "s", "id": 248981, "pid": 76337, "tid": -914061504, "ts": 1716454225379602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379701, "dur": 1,
    "args": {
      "External id": 248999,
      "cbid": 251, "correlation": 248999
    }
  },
  {
    "ph": "f", "id": 248999, "pid": 76337, "tid": -914061504, "ts": 1716454225379701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225433802, "dur": 169,
    "args": {
      "External id": 249001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249001,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249001, "pid": 5, "tid": 7, "ts": 1716454225433802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379707, "dur": 13,
    "args": {
      "External id": 249001,
      "cbid": 211, "correlation": 249001
    }
  },
  {
    "ph": "s", "id": 249001, "pid": 76337, "tid": -914061504, "ts": 1716454225379707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225433973, "dur": 20,
    "args": {
      "External id": 249009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249009,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249009, "pid": 5, "tid": 7, "ts": 1716454225433973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379776, "dur": 12,
    "args": {
      "External id": 249009,
      "cbid": 211, "correlation": 249009
    }
  },
  {
    "ph": "s", "id": 249009, "pid": 76337, "tid": -914061504, "ts": 1716454225379776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225433995, "dur": 28,
    "args": {
      "External id": 249017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249017,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249017, "pid": 5, "tid": 7, "ts": 1716454225433995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379815, "dur": 8,
    "args": {
      "External id": 249017,
      "cbid": 211, "correlation": 249017
    }
  },
  {
    "ph": "s", "id": 249017, "pid": 76337, "tid": -914061504, "ts": 1716454225379815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225434024, "dur": 18,
    "args": {
      "External id": 249028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249028, "pid": 5, "tid": 7, "ts": 1716454225434024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379892, "dur": 13,
    "args": {
      "External id": 249028,
      "cbid": 211, "correlation": 249028
    }
  },
  {
    "ph": "s", "id": 249028, "pid": 76337, "tid": -914061504, "ts": 1716454225379892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225434043, "dur": 17,
    "args": {
      "External id": 249050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249050,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249050, "pid": 5, "tid": 7, "ts": 1716454225434043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225379923, "dur": 8,
    "args": {
      "External id": 249050,
      "cbid": 211, "correlation": 249050
    }
  },
  {
    "ph": "s", "id": 249050, "pid": 76337, "tid": -914061504, "ts": 1716454225379923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380017, "dur": 2,
    "args": {
      "External id": 249061,
      "cbid": 251, "correlation": 249061
    }
  },
  {
    "ph": "f", "id": 249061, "pid": 76337, "tid": -914061504, "ts": 1716454225380017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225434061, "dur": 91,
    "args": {
      "External id": 249062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249062,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249062, "pid": 5, "tid": 7, "ts": 1716454225434061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380023, "dur": 15,
    "args": {
      "External id": 249062,
      "cbid": 211, "correlation": 249062
    }
  },
  {
    "ph": "s", "id": 249062, "pid": 76337, "tid": -914061504, "ts": 1716454225380023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380095, "dur": 1,
    "args": {
      "External id": 249073,
      "cbid": 251, "correlation": 249073
    }
  },
  {
    "ph": "f", "id": 249073, "pid": 76337, "tid": -914061504, "ts": 1716454225380095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380099, "dur": 0,
    "args": {
      "External id": 249074,
      "cbid": 251, "correlation": 249074
    }
  },
  {
    "ph": "f", "id": 249074, "pid": 76337, "tid": -914061504, "ts": 1716454225380099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225434153, "dur": 12,
    "args": {
      "External id": 249075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249075,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249075, "pid": 5, "tid": 7, "ts": 1716454225434153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380101, "dur": 12,
    "args": {
      "External id": 249075,
      "cbid": 211, "correlation": 249075
    }
  },
  {
    "ph": "s", "id": 249075, "pid": 76337, "tid": -914061504, "ts": 1716454225380101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225434166, "dur": 6,
    "args": {
      "External id": 249077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249077,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249077, "pid": 5, "tid": 7, "ts": 1716454225434166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380114, "dur": 6,
    "args": {
      "External id": 249077,
      "cbid": 211, "correlation": 249077
    }
  },
  {
    "ph": "s", "id": 249077, "pid": 76337, "tid": -914061504, "ts": 1716454225380114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380172, "dur": 1,
    "args": {
      "External id": 249088,
      "cbid": 251, "correlation": 249088
    }
  },
  {
    "ph": "f", "id": 249088, "pid": 76337, "tid": -914061504, "ts": 1716454225380172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380175, "dur": 0,
    "args": {
      "External id": 249089,
      "cbid": 251, "correlation": 249089
    }
  },
  {
    "ph": "f", "id": 249089, "pid": 76337, "tid": -914061504, "ts": 1716454225380175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225434174, "dur": 9,
    "args": {
      "External id": 249090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249090,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249090, "pid": 5, "tid": 7, "ts": 1716454225434174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380176, "dur": 12,
    "args": {
      "External id": 249090,
      "cbid": 211, "correlation": 249090
    }
  },
  {
    "ph": "s", "id": 249090, "pid": 76337, "tid": -914061504, "ts": 1716454225380176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225434184, "dur": 3,
    "args": {
      "External id": 249092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249092,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249092, "pid": 5, "tid": 7, "ts": 1716454225434184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380190, "dur": 5,
    "args": {
      "External id": 249092,
      "cbid": 211, "correlation": 249092
    }
  },
  {
    "ph": "s", "id": 249092, "pid": 76337, "tid": -914061504, "ts": 1716454225380190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225434189, "dur": 56,
    "args": {
      "External id": 249117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249117,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249117, "pid": 5, "tid": 7, "ts": 1716454225434189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380267, "dur": 13,
    "args": {
      "External id": 249117,
      "cbid": 211, "correlation": 249117
    }
  },
  {
    "ph": "s", "id": 249117, "pid": 76337, "tid": -914061504, "ts": 1716454225380267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380367, "dur": 2,
    "args": {
      "External id": 249135,
      "cbid": 251, "correlation": 249135
    }
  },
  {
    "ph": "f", "id": 249135, "pid": 76337, "tid": -914061504, "ts": 1716454225380367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225434246, "dur": 93,
    "args": {
      "External id": 249137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249137,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249137, "pid": 5, "tid": 7, "ts": 1716454225434246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380374, "dur": 15,
    "args": {
      "External id": 249137,
      "cbid": 211, "correlation": 249137
    }
  },
  {
    "ph": "s", "id": 249137, "pid": 76337, "tid": -914061504, "ts": 1716454225380374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225434340, "dur": 10,
    "args": {
      "External id": 249145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249145,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249145, "pid": 5, "tid": 7, "ts": 1716454225434340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380444, "dur": 12,
    "args": {
      "External id": 249145,
      "cbid": 211, "correlation": 249145
    }
  },
  {
    "ph": "s", "id": 249145, "pid": 76337, "tid": -914061504, "ts": 1716454225380444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225434351, "dur": 20,
    "args": {
      "External id": 249153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249153,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249153, "pid": 5, "tid": 7, "ts": 1716454225434351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380487, "dur": 9,
    "args": {
      "External id": 249153,
      "cbid": 211, "correlation": 249153
    }
  },
  {
    "ph": "s", "id": 249153, "pid": 76337, "tid": -914061504, "ts": 1716454225380487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225434373, "dur": 18,
    "args": {
      "External id": 249175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249175,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249175, "pid": 5, "tid": 7, "ts": 1716454225434373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380538, "dur": 10,
    "args": {
      "External id": 249175,
      "cbid": 211, "correlation": 249175
    }
  },
  {
    "ph": "s", "id": 249175, "pid": 76337, "tid": -914061504, "ts": 1716454225380538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380628, "dur": 2,
    "args": {
      "External id": 249191,
      "cbid": 251, "correlation": 249191
    }
  },
  {
    "ph": "f", "id": 249191, "pid": 76337, "tid": -914061504, "ts": 1716454225380628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380635, "dur": 0,
    "args": {
      "External id": 249193,
      "cbid": 251, "correlation": 249193
    }
  },
  {
    "ph": "f", "id": 249193, "pid": 76337, "tid": -914061504, "ts": 1716454225380635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225434392, "dur": 498,
    "args": {
      "External id": 249194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249194,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249194, "pid": 5, "tid": 7, "ts": 1716454225434392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380637, "dur": 15,
    "args": {
      "External id": 249194,
      "cbid": 211, "correlation": 249194
    }
  },
  {
    "ph": "s", "id": 249194, "pid": 76337, "tid": -914061504, "ts": 1716454225380637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225434892, "dur": 68,
    "args": {
      "External id": 249202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249202, "pid": 5, "tid": 7, "ts": 1716454225434892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380705, "dur": 12,
    "args": {
      "External id": 249202,
      "cbid": 211, "correlation": 249202
    }
  },
  {
    "ph": "s", "id": 249202, "pid": 76337, "tid": -914061504, "ts": 1716454225380705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225434961, "dur": 69,
    "args": {
      "External id": 249210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249210, "pid": 5, "tid": 7, "ts": 1716454225434961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380735, "dur": 8,
    "args": {
      "External id": 249210,
      "cbid": 211, "correlation": 249210
    }
  },
  {
    "ph": "s", "id": 249210, "pid": 76337, "tid": -914061504, "ts": 1716454225380735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380815, "dur": 1,
    "args": {
      "External id": 249226,
      "cbid": 251, "correlation": 249226
    }
  },
  {
    "ph": "f", "id": 249226, "pid": 76337, "tid": -914061504, "ts": 1716454225380815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225435032, "dur": 1,
    "args": {
      "External id": 249228,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 249228,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 249228, "pid": 5, "tid": 7, "ts": 1716454225435032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380821, "dur": 13,
    "args": {
      "External id": 249228,
      "cbid": 51, "correlation": 249228
    }
  },
  {
    "ph": "s", "id": 249228, "pid": 76337, "tid": -914061504, "ts": 1716454225380821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225435036, "dur": 275,
    "args": {
      "External id": 249229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249229,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249229, "pid": 5, "tid": 7, "ts": 1716454225435036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380835, "dur": 11,
    "args": {
      "External id": 249229,
      "cbid": 211, "correlation": 249229
    }
  },
  {
    "ph": "s", "id": 249229, "pid": 76337, "tid": -914061504, "ts": 1716454225380835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225435312, "dur": 13,
    "args": {
      "External id": 249237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249237,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249237, "pid": 5, "tid": 7, "ts": 1716454225435312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380876, "dur": 11,
    "args": {
      "External id": 249237,
      "cbid": 211, "correlation": 249237
    }
  },
  {
    "ph": "s", "id": 249237, "pid": 76337, "tid": -914061504, "ts": 1716454225380876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225435327, "dur": 38,
    "args": {
      "External id": 249248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249248, "pid": 5, "tid": 7, "ts": 1716454225435327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225380945, "dur": 12,
    "args": {
      "External id": 249248,
      "cbid": 211, "correlation": 249248
    }
  },
  {
    "ph": "s", "id": 249248, "pid": 76337, "tid": -914061504, "ts": 1716454225380945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381017, "dur": 0,
    "args": {
      "External id": 249260,
      "cbid": 317, "correlation": 249260
    }
  },
  {
    "ph": "f", "id": 249260, "pid": 76337, "tid": -914061504, "ts": 1716454225381017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381018, "dur": 0,
    "args": {
      "External id": 249261,
      "cbid": 203, "correlation": 249261
    }
  },
  {
    "ph": "f", "id": 249261, "pid": 76337, "tid": -914061504, "ts": 1716454225381018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381019, "dur": 0,
    "args": {
      "External id": 249262,
      "cbid": 205, "correlation": 249262
    }
  },
  {
    "ph": "f", "id": 249262, "pid": 76337, "tid": -914061504, "ts": 1716454225381019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225435366, "dur": 12,
    "args": {
      "External id": 249266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249266,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249266, "pid": 5, "tid": 7, "ts": 1716454225435366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381035, "dur": 12,
    "args": {
      "External id": 249266,
      "cbid": 211, "correlation": 249266
    }
  },
  {
    "ph": "s", "id": 249266, "pid": 76337, "tid": -914061504, "ts": 1716454225381035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225435380, "dur": 4,
    "args": {
      "External id": 249268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249268,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249268, "pid": 5, "tid": 7, "ts": 1716454225435380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381052, "dur": 6,
    "args": {
      "External id": 249268,
      "cbid": 211, "correlation": 249268
    }
  },
  {
    "ph": "s", "id": 249268, "pid": 76337, "tid": -914061504, "ts": 1716454225381052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381060, "dur": 0,
    "args": {
      "External id": 249269,
      "cbid": 51, "correlation": 249269
    }
  },
  {
    "ph": "s", "id": 249269, "pid": 76337, "tid": -914061504, "ts": 1716454225381060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225435385, "dur": 100,
    "args": {
      "External id": 249270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249270,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 249270, "pid": 5, "tid": 7, "ts": 1716454225435385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381061, "dur": 5,
    "args": {
      "External id": 249270,
      "cbid": 211, "correlation": 249270
    }
  },
  {
    "ph": "s", "id": 249270, "pid": 76337, "tid": -914061504, "ts": 1716454225381061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225435487, "dur": 16,
    "args": {
      "External id": 249275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249275,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249275, "pid": 5, "tid": 7, "ts": 1716454225435487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381089, "dur": 9,
    "args": {
      "External id": 249275,
      "cbid": 211, "correlation": 249275
    }
  },
  {
    "ph": "s", "id": 249275, "pid": 76337, "tid": -914061504, "ts": 1716454225381089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225435504, "dur": 11,
    "args": {
      "External id": 249283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249283,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249283, "pid": 5, "tid": 7, "ts": 1716454225435504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381121, "dur": 8,
    "args": {
      "External id": 249283,
      "cbid": 211, "correlation": 249283
    }
  },
  {
    "ph": "s", "id": 249283, "pid": 76337, "tid": -914061504, "ts": 1716454225381121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225435517, "dur": 18,
    "args": {
      "External id": 249303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249303,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 249303, "pid": 5, "tid": 7, "ts": 1716454225435517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381192, "dur": 11,
    "args": {
      "External id": 249303,
      "cbid": 211, "correlation": 249303
    }
  },
  {
    "ph": "s", "id": 249303, "pid": 76337, "tid": -914061504, "ts": 1716454225381192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225435536, "dur": 5,
    "args": {
      "External id": 249315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249315,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 249315, "pid": 5, "tid": 7, "ts": 1716454225435536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381214, "dur": 6,
    "args": {
      "External id": 249315,
      "cbid": 211, "correlation": 249315
    }
  },
  {
    "ph": "s", "id": 249315, "pid": 76337, "tid": -914061504, "ts": 1716454225381214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225435543, "dur": 18,
    "args": {
      "External id": 249318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249318, "pid": 5, "tid": 7, "ts": 1716454225435543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381232, "dur": 7,
    "args": {
      "External id": 249318,
      "cbid": 211, "correlation": 249318
    }
  },
  {
    "ph": "s", "id": 249318, "pid": 76337, "tid": -914061504, "ts": 1716454225381232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225435562, "dur": 12,
    "args": {
      "External id": 249327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249327,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249327, "pid": 5, "tid": 7, "ts": 1716454225435562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381270, "dur": 10,
    "args": {
      "External id": 249327,
      "cbid": 211, "correlation": 249327
    }
  },
  {
    "ph": "s", "id": 249327, "pid": 76337, "tid": -914061504, "ts": 1716454225381270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381321, "dur": 0,
    "args": {
      "External id": 249337,
      "cbid": 317, "correlation": 249337
    }
  },
  {
    "ph": "f", "id": 249337, "pid": 76337, "tid": -914061504, "ts": 1716454225381321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381322, "dur": 0,
    "args": {
      "External id": 249338,
      "cbid": 203, "correlation": 249338
    }
  },
  {
    "ph": "f", "id": 249338, "pid": 76337, "tid": -914061504, "ts": 1716454225381322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381323, "dur": 0,
    "args": {
      "External id": 249339,
      "cbid": 205, "correlation": 249339
    }
  },
  {
    "ph": "f", "id": 249339, "pid": 76337, "tid": -914061504, "ts": 1716454225381323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225435575, "dur": 12,
    "args": {
      "External id": 249343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249343, "pid": 5, "tid": 7, "ts": 1716454225435575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381340, "dur": 12,
    "args": {
      "External id": 249343,
      "cbid": 211, "correlation": 249343
    }
  },
  {
    "ph": "s", "id": 249343, "pid": 76337, "tid": -914061504, "ts": 1716454225381340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225435588, "dur": 164,
    "args": {
      "External id": 249345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249345,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249345, "pid": 5, "tid": 7, "ts": 1716454225435588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381354, "dur": 5,
    "args": {
      "External id": 249345,
      "cbid": 211, "correlation": 249345
    }
  },
  {
    "ph": "s", "id": 249345, "pid": 76337, "tid": -914061504, "ts": 1716454225381354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225435755, "dur": 1,
    "args": {
      "External id": 249347,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 249347,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 249347, "pid": 5, "tid": 7, "ts": 1716454225435755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381366, "dur": 7,
    "args": {
      "External id": 249347,
      "cbid": 51, "correlation": 249347
    }
  },
  {
    "ph": "s", "id": 249347, "pid": 76337, "tid": -914061504, "ts": 1716454225381366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225435758, "dur": 674,
    "args": {
      "External id": 249348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249348,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249348, "pid": 5, "tid": 7, "ts": 1716454225435758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381373, "dur": 7,
    "args": {
      "External id": 249348,
      "cbid": 211, "correlation": 249348
    }
  },
  {
    "ph": "s", "id": 249348, "pid": 76337, "tid": -914061504, "ts": 1716454225381373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225436434, "dur": 13,
    "args": {
      "External id": 249350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249350, "pid": 5, "tid": 7, "ts": 1716454225436434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381385, "dur": 5,
    "args": {
      "External id": 249350,
      "cbid": 211, "correlation": 249350
    }
  },
  {
    "ph": "s", "id": 249350, "pid": 76337, "tid": -914061504, "ts": 1716454225381385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225436449, "dur": 15,
    "args": {
      "External id": 249356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249356,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249356, "pid": 5, "tid": 7, "ts": 1716454225436449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381413, "dur": 8,
    "args": {
      "External id": 249356,
      "cbid": 211, "correlation": 249356
    }
  },
  {
    "ph": "s", "id": 249356, "pid": 76337, "tid": -914061504, "ts": 1716454225381413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225436465, "dur": 3,
    "args": {
      "External id": 249364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249364,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 249364, "pid": 5, "tid": 7, "ts": 1716454225436465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381456, "dur": 9,
    "args": {
      "External id": 249364,
      "cbid": 211, "correlation": 249364
    }
  },
  {
    "ph": "s", "id": 249364, "pid": 76337, "tid": -914061504, "ts": 1716454225381456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381523, "dur": 1,
    "args": {
      "External id": 249380,
      "cbid": 251, "correlation": 249380
    }
  },
  {
    "ph": "f", "id": 249380, "pid": 76337, "tid": -914061504, "ts": 1716454225381523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381530, "dur": 0,
    "args": {
      "External id": 249382,
      "cbid": 251, "correlation": 249382
    }
  },
  {
    "ph": "f", "id": 249382, "pid": 76337, "tid": -914061504, "ts": 1716454225381530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225436470, "dur": 14,
    "args": {
      "External id": 249383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249383,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249383, "pid": 5, "tid": 7, "ts": 1716454225436470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381531, "dur": 11,
    "args": {
      "External id": 249383,
      "cbid": 211, "correlation": 249383
    }
  },
  {
    "ph": "s", "id": 249383, "pid": 76337, "tid": -914061504, "ts": 1716454225381531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225436485, "dur": 5,
    "args": {
      "External id": 249385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249385,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249385, "pid": 5, "tid": 7, "ts": 1716454225436485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381545, "dur": 6,
    "args": {
      "External id": 249385,
      "cbid": 211, "correlation": 249385
    }
  },
  {
    "ph": "s", "id": 249385, "pid": 76337, "tid": -914061504, "ts": 1716454225381545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225436492, "dur": 17,
    "args": {
      "External id": 249395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249395, "pid": 5, "tid": 7, "ts": 1716454225436492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381603, "dur": 12,
    "args": {
      "External id": 249395,
      "cbid": 211, "correlation": 249395
    }
  },
  {
    "ph": "s", "id": 249395, "pid": 76337, "tid": -914061504, "ts": 1716454225381603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225436510, "dur": 18,
    "args": {
      "External id": 249415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249415,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 249415, "pid": 5, "tid": 7, "ts": 1716454225436510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381668, "dur": 11,
    "args": {
      "External id": 249415,
      "cbid": 211, "correlation": 249415
    }
  },
  {
    "ph": "s", "id": 249415, "pid": 76337, "tid": -914061504, "ts": 1716454225381668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225436529, "dur": 4,
    "args": {
      "External id": 249427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249427,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 249427, "pid": 5, "tid": 7, "ts": 1716454225436529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381690, "dur": 6,
    "args": {
      "External id": 249427,
      "cbid": 211, "correlation": 249427
    }
  },
  {
    "ph": "s", "id": 249427, "pid": 76337, "tid": -914061504, "ts": 1716454225381690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225436535, "dur": 17,
    "args": {
      "External id": 249430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249430, "pid": 5, "tid": 7, "ts": 1716454225436535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381708, "dur": 6,
    "args": {
      "External id": 249430,
      "cbid": 211, "correlation": 249430
    }
  },
  {
    "ph": "s", "id": 249430, "pid": 76337, "tid": -914061504, "ts": 1716454225381708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225436553, "dur": 11,
    "args": {
      "External id": 249439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249439,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249439, "pid": 5, "tid": 7, "ts": 1716454225436553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381748, "dur": 9,
    "args": {
      "External id": 249439,
      "cbid": 211, "correlation": 249439
    }
  },
  {
    "ph": "s", "id": 249439, "pid": 76337, "tid": -914061504, "ts": 1716454225381748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381809, "dur": 0,
    "args": {
      "External id": 249449,
      "cbid": 317, "correlation": 249449
    }
  },
  {
    "ph": "f", "id": 249449, "pid": 76337, "tid": -914061504, "ts": 1716454225381809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381811, "dur": 0,
    "args": {
      "External id": 249450,
      "cbid": 203, "correlation": 249450
    }
  },
  {
    "ph": "f", "id": 249450, "pid": 76337, "tid": -914061504, "ts": 1716454225381811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381811, "dur": 0,
    "args": {
      "External id": 249451,
      "cbid": 205, "correlation": 249451
    }
  },
  {
    "ph": "f", "id": 249451, "pid": 76337, "tid": -914061504, "ts": 1716454225381811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225436565, "dur": 11,
    "args": {
      "External id": 249455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249455,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249455, "pid": 5, "tid": 7, "ts": 1716454225436565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381825, "dur": 13,
    "args": {
      "External id": 249455,
      "cbid": 211, "correlation": 249455
    }
  },
  {
    "ph": "s", "id": 249455, "pid": 76337, "tid": -914061504, "ts": 1716454225381825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225436578, "dur": 166,
    "args": {
      "External id": 249457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249457,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249457, "pid": 5, "tid": 7, "ts": 1716454225436578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381840, "dur": 5,
    "args": {
      "External id": 249457,
      "cbid": 211, "correlation": 249457
    }
  },
  {
    "ph": "s", "id": 249457, "pid": 76337, "tid": -914061504, "ts": 1716454225381840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225436746, "dur": 1,
    "args": {
      "External id": 249459,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 249459,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 249459, "pid": 5, "tid": 7, "ts": 1716454225436746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381851, "dur": 6,
    "args": {
      "External id": 249459,
      "cbid": 51, "correlation": 249459
    }
  },
  {
    "ph": "s", "id": 249459, "pid": 76337, "tid": -914061504, "ts": 1716454225381851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225436750, "dur": 660,
    "args": {
      "External id": 249460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249460,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249460, "pid": 5, "tid": 7, "ts": 1716454225436750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381858, "dur": 6,
    "args": {
      "External id": 249460,
      "cbid": 211, "correlation": 249460
    }
  },
  {
    "ph": "s", "id": 249460, "pid": 76337, "tid": -914061504, "ts": 1716454225381858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225437411, "dur": 13,
    "args": {
      "External id": 249462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249462,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249462, "pid": 5, "tid": 7, "ts": 1716454225437411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381869, "dur": 5,
    "args": {
      "External id": 249462,
      "cbid": 211, "correlation": 249462
    }
  },
  {
    "ph": "s", "id": 249462, "pid": 76337, "tid": -914061504, "ts": 1716454225381869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225437425, "dur": 15,
    "args": {
      "External id": 249468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249468, "pid": 5, "tid": 7, "ts": 1716454225437425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381901, "dur": 9,
    "args": {
      "External id": 249468,
      "cbid": 211, "correlation": 249468
    }
  },
  {
    "ph": "s", "id": 249468, "pid": 76337, "tid": -914061504, "ts": 1716454225381901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225437441, "dur": 12,
    "args": {
      "External id": 249476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249476,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249476, "pid": 5, "tid": 7, "ts": 1716454225437441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381935, "dur": 8,
    "args": {
      "External id": 249476,
      "cbid": 211, "correlation": 249476
    }
  },
  {
    "ph": "s", "id": 249476, "pid": 76337, "tid": -914061504, "ts": 1716454225381935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225437454, "dur": 11,
    "args": {
      "External id": 249484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249484,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249484, "pid": 5, "tid": 7, "ts": 1716454225437454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225381964, "dur": 16,
    "args": {
      "External id": 249484,
      "cbid": 211, "correlation": 249484
    }
  },
  {
    "ph": "s", "id": 249484, "pid": 76337, "tid": -914061504, "ts": 1716454225381964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225437466, "dur": 18,
    "args": {
      "External id": 249504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249504,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 249504, "pid": 5, "tid": 7, "ts": 1716454225437466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382051, "dur": 13,
    "args": {
      "External id": 249504,
      "cbid": 211, "correlation": 249504
    }
  },
  {
    "ph": "s", "id": 249504, "pid": 76337, "tid": -914061504, "ts": 1716454225382051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225437486, "dur": 4,
    "args": {
      "External id": 249516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249516,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 249516, "pid": 5, "tid": 7, "ts": 1716454225437486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382073, "dur": 7,
    "args": {
      "External id": 249516,
      "cbid": 211, "correlation": 249516
    }
  },
  {
    "ph": "s", "id": 249516, "pid": 76337, "tid": -914061504, "ts": 1716454225382073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225437491, "dur": 17,
    "args": {
      "External id": 249519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249519,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249519, "pid": 5, "tid": 7, "ts": 1716454225437491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382092, "dur": 6,
    "args": {
      "External id": 249519,
      "cbid": 211, "correlation": 249519
    }
  },
  {
    "ph": "s", "id": 249519, "pid": 76337, "tid": -914061504, "ts": 1716454225382092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382148, "dur": 0,
    "args": {
      "External id": 249530,
      "cbid": 317, "correlation": 249530
    }
  },
  {
    "ph": "f", "id": 249530, "pid": 76337, "tid": -914061504, "ts": 1716454225382148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382149, "dur": 0,
    "args": {
      "External id": 249531,
      "cbid": 203, "correlation": 249531
    }
  },
  {
    "ph": "f", "id": 249531, "pid": 76337, "tid": -914061504, "ts": 1716454225382149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382150, "dur": 0,
    "args": {
      "External id": 249532,
      "cbid": 205, "correlation": 249532
    }
  },
  {
    "ph": "f", "id": 249532, "pid": 76337, "tid": -914061504, "ts": 1716454225382150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225437510, "dur": 10,
    "args": {
      "External id": 249536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249536,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249536, "pid": 5, "tid": 7, "ts": 1716454225437510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382163, "dur": 12,
    "args": {
      "External id": 249536,
      "cbid": 211, "correlation": 249536
    }
  },
  {
    "ph": "s", "id": 249536, "pid": 76337, "tid": -914061504, "ts": 1716454225382163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225437522, "dur": 4,
    "args": {
      "External id": 249538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249538, "pid": 5, "tid": 7, "ts": 1716454225437522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382179, "dur": 6,
    "args": {
      "External id": 249538,
      "cbid": 211, "correlation": 249538
    }
  },
  {
    "ph": "s", "id": 249538, "pid": 76337, "tid": -914061504, "ts": 1716454225382179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382189, "dur": 0,
    "args": {
      "External id": 249539,
      "cbid": 51, "correlation": 249539
    }
  },
  {
    "ph": "s", "id": 249539, "pid": 76337, "tid": -914061504, "ts": 1716454225382189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225437527, "dur": 96,
    "args": {
      "External id": 249540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249540,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 249540, "pid": 5, "tid": 7, "ts": 1716454225437527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382189, "dur": 5,
    "args": {
      "External id": 249540,
      "cbid": 211, "correlation": 249540
    }
  },
  {
    "ph": "s", "id": 249540, "pid": 76337, "tid": -914061504, "ts": 1716454225382189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225437625, "dur": 15,
    "args": {
      "External id": 249545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249545, "pid": 5, "tid": 7, "ts": 1716454225437625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382220, "dur": 9,
    "args": {
      "External id": 249545,
      "cbid": 211, "correlation": 249545
    }
  },
  {
    "ph": "s", "id": 249545, "pid": 76337, "tid": -914061504, "ts": 1716454225382220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225437642, "dur": 85,
    "args": {
      "External id": 249554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249554,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249554, "pid": 5, "tid": 7, "ts": 1716454225437642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382303, "dur": 14,
    "args": {
      "External id": 249554,
      "cbid": 211, "correlation": 249554
    }
  },
  {
    "ph": "s", "id": 249554, "pid": 76337, "tid": -914061504, "ts": 1716454225382303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225437728, "dur": 30,
    "args": {
      "External id": 249576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249576,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249576, "pid": 5, "tid": 7, "ts": 1716454225437728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382360, "dur": 10,
    "args": {
      "External id": 249576,
      "cbid": 211, "correlation": 249576
    }
  },
  {
    "ph": "s", "id": 249576, "pid": 76337, "tid": -914061504, "ts": 1716454225382360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382447, "dur": 1,
    "args": {
      "External id": 249587,
      "cbid": 251, "correlation": 249587
    }
  },
  {
    "ph": "f", "id": 249587, "pid": 76337, "tid": -914061504, "ts": 1716454225382447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225437760, "dur": 167,
    "args": {
      "External id": 249588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249588,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249588, "pid": 5, "tid": 7, "ts": 1716454225437760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382452, "dur": 13,
    "args": {
      "External id": 249588,
      "cbid": 211, "correlation": 249588
    }
  },
  {
    "ph": "s", "id": 249588, "pid": 76337, "tid": -914061504, "ts": 1716454225382452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382521, "dur": 1,
    "args": {
      "External id": 249599,
      "cbid": 251, "correlation": 249599
    }
  },
  {
    "ph": "f", "id": 249599, "pid": 76337, "tid": -914061504, "ts": 1716454225382521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225437928, "dur": 160,
    "args": {
      "External id": 249600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249600,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249600, "pid": 5, "tid": 7, "ts": 1716454225437928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382525, "dur": 12,
    "args": {
      "External id": 249600,
      "cbid": 211, "correlation": 249600
    }
  },
  {
    "ph": "s", "id": 249600, "pid": 76337, "tid": -914061504, "ts": 1716454225382525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382591, "dur": 1,
    "args": {
      "External id": 249611,
      "cbid": 251, "correlation": 249611
    }
  },
  {
    "ph": "f", "id": 249611, "pid": 76337, "tid": -914061504, "ts": 1716454225382591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225438090, "dur": 162,
    "args": {
      "External id": 249612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249612,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249612, "pid": 5, "tid": 7, "ts": 1716454225438090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382595, "dur": 11,
    "args": {
      "External id": 249612,
      "cbid": 211, "correlation": 249612
    }
  },
  {
    "ph": "s", "id": 249612, "pid": 76337, "tid": -914061504, "ts": 1716454225382595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225438254, "dur": 345,
    "args": {
      "External id": 249637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249637,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249637, "pid": 5, "tid": 7, "ts": 1716454225438254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382679, "dur": 14,
    "args": {
      "External id": 249637,
      "cbid": 211, "correlation": 249637
    }
  },
  {
    "ph": "s", "id": 249637, "pid": 76337, "tid": -914061504, "ts": 1716454225382679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382778, "dur": 1,
    "args": {
      "External id": 249655,
      "cbid": 251, "correlation": 249655
    }
  },
  {
    "ph": "f", "id": 249655, "pid": 76337, "tid": -914061504, "ts": 1716454225382778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225438600, "dur": 168,
    "args": {
      "External id": 249657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249657,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249657, "pid": 5, "tid": 7, "ts": 1716454225438600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382784, "dur": 14,
    "args": {
      "External id": 249657,
      "cbid": 211, "correlation": 249657
    }
  },
  {
    "ph": "s", "id": 249657, "pid": 76337, "tid": -914061504, "ts": 1716454225382784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225438769, "dur": 19,
    "args": {
      "External id": 249665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249665,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249665, "pid": 5, "tid": 7, "ts": 1716454225438769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382854, "dur": 12,
    "args": {
      "External id": 249665,
      "cbid": 211, "correlation": 249665
    }
  },
  {
    "ph": "s", "id": 249665, "pid": 76337, "tid": -914061504, "ts": 1716454225382854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225438790, "dur": 28,
    "args": {
      "External id": 249673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249673,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249673, "pid": 5, "tid": 7, "ts": 1716454225438790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382893, "dur": 8,
    "args": {
      "External id": 249673,
      "cbid": 211, "correlation": 249673
    }
  },
  {
    "ph": "s", "id": 249673, "pid": 76337, "tid": -914061504, "ts": 1716454225382893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225438819, "dur": 18,
    "args": {
      "External id": 249684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249684, "pid": 5, "tid": 7, "ts": 1716454225438819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225382963, "dur": 20,
    "args": {
      "External id": 249684,
      "cbid": 211, "correlation": 249684
    }
  },
  {
    "ph": "s", "id": 249684, "pid": 76337, "tid": -914061504, "ts": 1716454225382963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225438839, "dur": 17,
    "args": {
      "External id": 249706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249706,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249706, "pid": 5, "tid": 7, "ts": 1716454225438839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383003, "dur": 8,
    "args": {
      "External id": 249706,
      "cbid": 211, "correlation": 249706
    }
  },
  {
    "ph": "s", "id": 249706, "pid": 76337, "tid": -914061504, "ts": 1716454225383003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383091, "dur": 1,
    "args": {
      "External id": 249717,
      "cbid": 251, "correlation": 249717
    }
  },
  {
    "ph": "f", "id": 249717, "pid": 76337, "tid": -914061504, "ts": 1716454225383091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225438857, "dur": 91,
    "args": {
      "External id": 249718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249718,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249718, "pid": 5, "tid": 7, "ts": 1716454225438857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383096, "dur": 13,
    "args": {
      "External id": 249718,
      "cbid": 211, "correlation": 249718
    }
  },
  {
    "ph": "s", "id": 249718, "pid": 76337, "tid": -914061504, "ts": 1716454225383096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383164, "dur": 1,
    "args": {
      "External id": 249729,
      "cbid": 251, "correlation": 249729
    }
  },
  {
    "ph": "f", "id": 249729, "pid": 76337, "tid": -914061504, "ts": 1716454225383164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383168, "dur": 0,
    "args": {
      "External id": 249730,
      "cbid": 251, "correlation": 249730
    }
  },
  {
    "ph": "f", "id": 249730, "pid": 76337, "tid": -914061504, "ts": 1716454225383168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225438949, "dur": 13,
    "args": {
      "External id": 249731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249731,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249731, "pid": 5, "tid": 7, "ts": 1716454225438949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383169, "dur": 12,
    "args": {
      "External id": 249731,
      "cbid": 211, "correlation": 249731
    }
  },
  {
    "ph": "s", "id": 249731, "pid": 76337, "tid": -914061504, "ts": 1716454225383169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225438963, "dur": 6,
    "args": {
      "External id": 249733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249733,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249733, "pid": 5, "tid": 7, "ts": 1716454225438963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383183, "dur": 6,
    "args": {
      "External id": 249733,
      "cbid": 211, "correlation": 249733
    }
  },
  {
    "ph": "s", "id": 249733, "pid": 76337, "tid": -914061504, "ts": 1716454225383183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383241, "dur": 1,
    "args": {
      "External id": 249744,
      "cbid": 251, "correlation": 249744
    }
  },
  {
    "ph": "f", "id": 249744, "pid": 76337, "tid": -914061504, "ts": 1716454225383241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383245, "dur": 0,
    "args": {
      "External id": 249745,
      "cbid": 251, "correlation": 249745
    }
  },
  {
    "ph": "f", "id": 249745, "pid": 76337, "tid": -914061504, "ts": 1716454225383245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225438970, "dur": 9,
    "args": {
      "External id": 249746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249746,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249746, "pid": 5, "tid": 7, "ts": 1716454225438970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383247, "dur": 12,
    "args": {
      "External id": 249746,
      "cbid": 211, "correlation": 249746
    }
  },
  {
    "ph": "s", "id": 249746, "pid": 76337, "tid": -914061504, "ts": 1716454225383247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225438980, "dur": 4,
    "args": {
      "External id": 249748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249748,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249748, "pid": 5, "tid": 7, "ts": 1716454225438980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383260, "dur": 5,
    "args": {
      "External id": 249748,
      "cbid": 211, "correlation": 249748
    }
  },
  {
    "ph": "s", "id": 249748, "pid": 76337, "tid": -914061504, "ts": 1716454225383260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225438985, "dur": 55,
    "args": {
      "External id": 249773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249773,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249773, "pid": 5, "tid": 7, "ts": 1716454225438985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383337, "dur": 13,
    "args": {
      "External id": 249773,
      "cbid": 211, "correlation": 249773
    }
  },
  {
    "ph": "s", "id": 249773, "pid": 76337, "tid": -914061504, "ts": 1716454225383337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383435, "dur": 1,
    "args": {
      "External id": 249791,
      "cbid": 251, "correlation": 249791
    }
  },
  {
    "ph": "f", "id": 249791, "pid": 76337, "tid": -914061504, "ts": 1716454225383435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225439041, "dur": 92,
    "args": {
      "External id": 249793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249793,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249793, "pid": 5, "tid": 7, "ts": 1716454225439041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383441, "dur": 14,
    "args": {
      "External id": 249793,
      "cbid": 211, "correlation": 249793
    }
  },
  {
    "ph": "s", "id": 249793, "pid": 76337, "tid": -914061504, "ts": 1716454225383441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225439135, "dur": 9,
    "args": {
      "External id": 249801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249801,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249801, "pid": 5, "tid": 7, "ts": 1716454225439135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383511, "dur": 12,
    "args": {
      "External id": 249801,
      "cbid": 211, "correlation": 249801
    }
  },
  {
    "ph": "s", "id": 249801, "pid": 76337, "tid": -914061504, "ts": 1716454225383511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225439146, "dur": 21,
    "args": {
      "External id": 249809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249809, "pid": 5, "tid": 7, "ts": 1716454225439146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383552, "dur": 9,
    "args": {
      "External id": 249809,
      "cbid": 211, "correlation": 249809
    }
  },
  {
    "ph": "s", "id": 249809, "pid": 76337, "tid": -914061504, "ts": 1716454225383552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225439168, "dur": 17,
    "args": {
      "External id": 249831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249831,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249831, "pid": 5, "tid": 7, "ts": 1716454225439168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383604, "dur": 10,
    "args": {
      "External id": 249831,
      "cbid": 211, "correlation": 249831
    }
  },
  {
    "ph": "s", "id": 249831, "pid": 76337, "tid": -914061504, "ts": 1716454225383604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383691, "dur": 1,
    "args": {
      "External id": 249847,
      "cbid": 251, "correlation": 249847
    }
  },
  {
    "ph": "f", "id": 249847, "pid": 76337, "tid": -914061504, "ts": 1716454225383691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383696, "dur": 0,
    "args": {
      "External id": 249849,
      "cbid": 251, "correlation": 249849
    }
  },
  {
    "ph": "f", "id": 249849, "pid": 76337, "tid": -914061504, "ts": 1716454225383696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225439187, "dur": 499,
    "args": {
      "External id": 249850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249850,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 249850, "pid": 5, "tid": 7, "ts": 1716454225439187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383698, "dur": 13,
    "args": {
      "External id": 249850,
      "cbid": 211, "correlation": 249850
    }
  },
  {
    "ph": "s", "id": 249850, "pid": 76337, "tid": -914061504, "ts": 1716454225383698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225439687, "dur": 66,
    "args": {
      "External id": 249858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249858, "pid": 5, "tid": 7, "ts": 1716454225439687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383763, "dur": 12,
    "args": {
      "External id": 249858,
      "cbid": 211, "correlation": 249858
    }
  },
  {
    "ph": "s", "id": 249858, "pid": 76337, "tid": -914061504, "ts": 1716454225383763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225439754, "dur": 68,
    "args": {
      "External id": 249866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249866, "pid": 5, "tid": 7, "ts": 1716454225439754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383793, "dur": 9,
    "args": {
      "External id": 249866,
      "cbid": 211, "correlation": 249866
    }
  },
  {
    "ph": "s", "id": 249866, "pid": 76337, "tid": -914061504, "ts": 1716454225383793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383872, "dur": 1,
    "args": {
      "External id": 249882,
      "cbid": 251, "correlation": 249882
    }
  },
  {
    "ph": "f", "id": 249882, "pid": 76337, "tid": -914061504, "ts": 1716454225383872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225439824, "dur": 1,
    "args": {
      "External id": 249884,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 249884,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 249884, "pid": 5, "tid": 7, "ts": 1716454225439824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383877, "dur": 9,
    "args": {
      "External id": 249884,
      "cbid": 51, "correlation": 249884
    }
  },
  {
    "ph": "s", "id": 249884, "pid": 76337, "tid": -914061504, "ts": 1716454225383877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225439828, "dur": 275,
    "args": {
      "External id": 249885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249885,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249885, "pid": 5, "tid": 7, "ts": 1716454225439828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383888, "dur": 12,
    "args": {
      "External id": 249885,
      "cbid": 211, "correlation": 249885
    }
  },
  {
    "ph": "s", "id": 249885, "pid": 76337, "tid": -914061504, "ts": 1716454225383888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225440105, "dur": 14,
    "args": {
      "External id": 249893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249893,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249893, "pid": 5, "tid": 7, "ts": 1716454225440105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225383930, "dur": 10,
    "args": {
      "External id": 249893,
      "cbid": 211, "correlation": 249893
    }
  },
  {
    "ph": "s", "id": 249893, "pid": 76337, "tid": -914061504, "ts": 1716454225383930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225440120, "dur": 39,
    "args": {
      "External id": 249904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249904,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249904, "pid": 5, "tid": 7, "ts": 1716454225440120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384007, "dur": 13,
    "args": {
      "External id": 249904,
      "cbid": 211, "correlation": 249904
    }
  },
  {
    "ph": "s", "id": 249904, "pid": 76337, "tid": -914061504, "ts": 1716454225384007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384072, "dur": 0,
    "args": {
      "External id": 249916,
      "cbid": 317, "correlation": 249916
    }
  },
  {
    "ph": "f", "id": 249916, "pid": 76337, "tid": -914061504, "ts": 1716454225384072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384072, "dur": 0,
    "args": {
      "External id": 249917,
      "cbid": 203, "correlation": 249917
    }
  },
  {
    "ph": "f", "id": 249917, "pid": 76337, "tid": -914061504, "ts": 1716454225384072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384073, "dur": 0,
    "args": {
      "External id": 249918,
      "cbid": 205, "correlation": 249918
    }
  },
  {
    "ph": "f", "id": 249918, "pid": 76337, "tid": -914061504, "ts": 1716454225384073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440160, "dur": 12,
    "args": {
      "External id": 249922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249922,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249922, "pid": 5, "tid": 7, "ts": 1716454225440160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384089, "dur": 13,
    "args": {
      "External id": 249922,
      "cbid": 211, "correlation": 249922
    }
  },
  {
    "ph": "s", "id": 249922, "pid": 76337, "tid": -914061504, "ts": 1716454225384089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225440174, "dur": 4,
    "args": {
      "External id": 249924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 249924, "pid": 5, "tid": 7, "ts": 1716454225440174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384107, "dur": 6,
    "args": {
      "External id": 249924,
      "cbid": 211, "correlation": 249924
    }
  },
  {
    "ph": "s", "id": 249924, "pid": 76337, "tid": -914061504, "ts": 1716454225384107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384115, "dur": 0,
    "args": {
      "External id": 249925,
      "cbid": 51, "correlation": 249925
    }
  },
  {
    "ph": "s", "id": 249925, "pid": 76337, "tid": -914061504, "ts": 1716454225384115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225440179, "dur": 99,
    "args": {
      "External id": 249926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249926,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 249926, "pid": 5, "tid": 7, "ts": 1716454225440179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384116, "dur": 5,
    "args": {
      "External id": 249926,
      "cbid": 211, "correlation": 249926
    }
  },
  {
    "ph": "s", "id": 249926, "pid": 76337, "tid": -914061504, "ts": 1716454225384116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225440280, "dur": 17,
    "args": {
      "External id": 249931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249931, "pid": 5, "tid": 7, "ts": 1716454225440280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384143, "dur": 9,
    "args": {
      "External id": 249931,
      "cbid": 211, "correlation": 249931
    }
  },
  {
    "ph": "s", "id": 249931, "pid": 76337, "tid": -914061504, "ts": 1716454225384143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225440298, "dur": 11,
    "args": {
      "External id": 249939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249939,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249939, "pid": 5, "tid": 7, "ts": 1716454225440298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384175, "dur": 8,
    "args": {
      "External id": 249939,
      "cbid": 211, "correlation": 249939
    }
  },
  {
    "ph": "s", "id": 249939, "pid": 76337, "tid": -914061504, "ts": 1716454225384175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384244, "dur": 0,
    "args": {
      "External id": 249949,
      "cbid": 317, "correlation": 249949
    }
  },
  {
    "ph": "f", "id": 249949, "pid": 76337, "tid": -914061504, "ts": 1716454225384244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384244, "dur": 0,
    "args": {
      "External id": 249950,
      "cbid": 203, "correlation": 249950
    }
  },
  {
    "ph": "f", "id": 249950, "pid": 76337, "tid": -914061504, "ts": 1716454225384244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384245, "dur": 0,
    "args": {
      "External id": 249951,
      "cbid": 205, "correlation": 249951
    }
  },
  {
    "ph": "f", "id": 249951, "pid": 76337, "tid": -914061504, "ts": 1716454225384245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440310, "dur": 12,
    "args": {
      "External id": 249955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249955,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249955, "pid": 5, "tid": 7, "ts": 1716454225440310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384261, "dur": 12,
    "args": {
      "External id": 249955,
      "cbid": 211, "correlation": 249955
    }
  },
  {
    "ph": "s", "id": 249955, "pid": 76337, "tid": -914061504, "ts": 1716454225384261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440324, "dur": 164,
    "args": {
      "External id": 249957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249957,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249957, "pid": 5, "tid": 7, "ts": 1716454225440324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384275, "dur": 5,
    "args": {
      "External id": 249957,
      "cbid": 211, "correlation": 249957
    }
  },
  {
    "ph": "s", "id": 249957, "pid": 76337, "tid": -914061504, "ts": 1716454225384275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225440491, "dur": 1,
    "args": {
      "External id": 249959,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 249959,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 249959, "pid": 5, "tid": 7, "ts": 1716454225440491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384286, "dur": 6,
    "args": {
      "External id": 249959,
      "cbid": 51, "correlation": 249959
    }
  },
  {
    "ph": "s", "id": 249959, "pid": 76337, "tid": -914061504, "ts": 1716454225384286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225440495, "dur": 202,
    "args": {
      "External id": 249960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249960,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 249960, "pid": 5, "tid": 7, "ts": 1716454225440495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384294, "dur": 9,
    "args": {
      "External id": 249960,
      "cbid": 211, "correlation": 249960
    }
  },
  {
    "ph": "s", "id": 249960, "pid": 76337, "tid": -914061504, "ts": 1716454225384294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440698, "dur": 6,
    "args": {
      "External id": 249962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249962,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 249962, "pid": 5, "tid": 7, "ts": 1716454225440698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384307, "dur": 5,
    "args": {
      "External id": 249962,
      "cbid": 211, "correlation": 249962
    }
  },
  {
    "ph": "s", "id": 249962, "pid": 76337, "tid": -914061504, "ts": 1716454225384307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225440705, "dur": 6,
    "args": {
      "External id": 249968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 249968, "pid": 5, "tid": 7, "ts": 1716454225440705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384338, "dur": 9,
    "args": {
      "External id": 249968,
      "cbid": 211, "correlation": 249968
    }
  },
  {
    "ph": "s", "id": 249968, "pid": 76337, "tid": -914061504, "ts": 1716454225384338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440713, "dur": 11,
    "args": {
      "External id": 249988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 249988,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 249988, "pid": 5, "tid": 7, "ts": 1716454225440713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384433, "dur": 12,
    "args": {
      "External id": 249988,
      "cbid": 211, "correlation": 249988
    }
  },
  {
    "ph": "s", "id": 249988, "pid": 76337, "tid": -914061504, "ts": 1716454225384433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225440725, "dur": 4,
    "args": {
      "External id": 250000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250000,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250000, "pid": 5, "tid": 7, "ts": 1716454225440725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384460, "dur": 7,
    "args": {
      "External id": 250000,
      "cbid": 211, "correlation": 250000
    }
  },
  {
    "ph": "s", "id": 250000, "pid": 76337, "tid": -914061504, "ts": 1716454225384460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225440731, "dur": 8,
    "args": {
      "External id": 250003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250003, "pid": 5, "tid": 7, "ts": 1716454225440731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384480, "dur": 6,
    "args": {
      "External id": 250003,
      "cbid": 211, "correlation": 250003
    }
  },
  {
    "ph": "s", "id": 250003, "pid": 76337, "tid": -914061504, "ts": 1716454225384480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225440741, "dur": 6,
    "args": {
      "External id": 250012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250012,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250012, "pid": 5, "tid": 7, "ts": 1716454225440741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384520, "dur": 10,
    "args": {
      "External id": 250012,
      "cbid": 211, "correlation": 250012
    }
  },
  {
    "ph": "s", "id": 250012, "pid": 76337, "tid": -914061504, "ts": 1716454225384520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384572, "dur": 0,
    "args": {
      "External id": 250022,
      "cbid": 317, "correlation": 250022
    }
  },
  {
    "ph": "f", "id": 250022, "pid": 76337, "tid": -914061504, "ts": 1716454225384572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384572, "dur": 0,
    "args": {
      "External id": 250023,
      "cbid": 203, "correlation": 250023
    }
  },
  {
    "ph": "f", "id": 250023, "pid": 76337, "tid": -914061504, "ts": 1716454225384572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384573, "dur": 0,
    "args": {
      "External id": 250024,
      "cbid": 205, "correlation": 250024
    }
  },
  {
    "ph": "f", "id": 250024, "pid": 76337, "tid": -914061504, "ts": 1716454225384573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440748, "dur": 5,
    "args": {
      "External id": 250028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250028, "pid": 5, "tid": 7, "ts": 1716454225440748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384588, "dur": 12,
    "args": {
      "External id": 250028,
      "cbid": 211, "correlation": 250028
    }
  },
  {
    "ph": "s", "id": 250028, "pid": 76337, "tid": -914061504, "ts": 1716454225384588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225440754, "dur": 167,
    "args": {
      "External id": 250030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250030,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250030, "pid": 5, "tid": 7, "ts": 1716454225440754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384602, "dur": 5,
    "args": {
      "External id": 250030,
      "cbid": 211, "correlation": 250030
    }
  },
  {
    "ph": "s", "id": 250030, "pid": 76337, "tid": -914061504, "ts": 1716454225384602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225440923, "dur": 1,
    "args": {
      "External id": 250032,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250032,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 250032, "pid": 5, "tid": 7, "ts": 1716454225440923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384613, "dur": 6,
    "args": {
      "External id": 250032,
      "cbid": 51, "correlation": 250032
    }
  },
  {
    "ph": "s", "id": 250032, "pid": 76337, "tid": -914061504, "ts": 1716454225384613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225440927, "dur": 275,
    "args": {
      "External id": 250033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250033,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250033, "pid": 5, "tid": 7, "ts": 1716454225440927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384621, "dur": 6,
    "args": {
      "External id": 250033,
      "cbid": 211, "correlation": 250033
    }
  },
  {
    "ph": "s", "id": 250033, "pid": 76337, "tid": -914061504, "ts": 1716454225384621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441203, "dur": 6,
    "args": {
      "External id": 250035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250035,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250035, "pid": 5, "tid": 7, "ts": 1716454225441203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384631, "dur": 5,
    "args": {
      "External id": 250035,
      "cbid": 211, "correlation": 250035
    }
  },
  {
    "ph": "s", "id": 250035, "pid": 76337, "tid": -914061504, "ts": 1716454225384631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225441210, "dur": 6,
    "args": {
      "External id": 250041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250041,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250041, "pid": 5, "tid": 7, "ts": 1716454225441210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384659, "dur": 8,
    "args": {
      "External id": 250041,
      "cbid": 211, "correlation": 250041
    }
  },
  {
    "ph": "s", "id": 250041, "pid": 76337, "tid": -914061504, "ts": 1716454225384659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225441218, "dur": 3,
    "args": {
      "External id": 250049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250049,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 250049, "pid": 5, "tid": 7, "ts": 1716454225441218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384703, "dur": 9,
    "args": {
      "External id": 250049,
      "cbid": 211, "correlation": 250049
    }
  },
  {
    "ph": "s", "id": 250049, "pid": 76337, "tid": -914061504, "ts": 1716454225384703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384766, "dur": 1,
    "args": {
      "External id": 250065,
      "cbid": 251, "correlation": 250065
    }
  },
  {
    "ph": "f", "id": 250065, "pid": 76337, "tid": -914061504, "ts": 1716454225384766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384771, "dur": 0,
    "args": {
      "External id": 250067,
      "cbid": 251, "correlation": 250067
    }
  },
  {
    "ph": "f", "id": 250067, "pid": 76337, "tid": -914061504, "ts": 1716454225384771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225441222, "dur": 14,
    "args": {
      "External id": 250068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250068,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250068, "pid": 5, "tid": 7, "ts": 1716454225441222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384773, "dur": 11,
    "args": {
      "External id": 250068,
      "cbid": 211, "correlation": 250068
    }
  },
  {
    "ph": "s", "id": 250068, "pid": 76337, "tid": -914061504, "ts": 1716454225384773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225441237, "dur": 5,
    "args": {
      "External id": 250070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250070,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250070, "pid": 5, "tid": 7, "ts": 1716454225441237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384786, "dur": 5,
    "args": {
      "External id": 250070,
      "cbid": 211, "correlation": 250070
    }
  },
  {
    "ph": "s", "id": 250070, "pid": 76337, "tid": -914061504, "ts": 1716454225384786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225441243, "dur": 6,
    "args": {
      "External id": 250080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250080,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250080, "pid": 5, "tid": 7, "ts": 1716454225441243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384842, "dur": 13,
    "args": {
      "External id": 250080,
      "cbid": 211, "correlation": 250080
    }
  },
  {
    "ph": "s", "id": 250080, "pid": 76337, "tid": -914061504, "ts": 1716454225384842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441250, "dur": 10,
    "args": {
      "External id": 250100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250100,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 250100, "pid": 5, "tid": 7, "ts": 1716454225441250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384910, "dur": 11,
    "args": {
      "External id": 250100,
      "cbid": 211, "correlation": 250100
    }
  },
  {
    "ph": "s", "id": 250100, "pid": 76337, "tid": -914061504, "ts": 1716454225384910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225441262, "dur": 4,
    "args": {
      "External id": 250112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250112,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250112, "pid": 5, "tid": 7, "ts": 1716454225441262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384930, "dur": 6,
    "args": {
      "External id": 250112,
      "cbid": 211, "correlation": 250112
    }
  },
  {
    "ph": "s", "id": 250112, "pid": 76337, "tid": -914061504, "ts": 1716454225384930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225441267, "dur": 7,
    "args": {
      "External id": 250115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250115,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250115, "pid": 5, "tid": 7, "ts": 1716454225441267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384949, "dur": 7,
    "args": {
      "External id": 250115,
      "cbid": 211, "correlation": 250115
    }
  },
  {
    "ph": "s", "id": 250115, "pid": 76337, "tid": -914061504, "ts": 1716454225384949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225441275, "dur": 5,
    "args": {
      "External id": 250124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250124,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250124, "pid": 5, "tid": 7, "ts": 1716454225441275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225384998, "dur": 11,
    "args": {
      "External id": 250124,
      "cbid": 211, "correlation": 250124
    }
  },
  {
    "ph": "s", "id": 250124, "pid": 76337, "tid": -914061504, "ts": 1716454225384998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385063, "dur": 0,
    "args": {
      "External id": 250134,
      "cbid": 317, "correlation": 250134
    }
  },
  {
    "ph": "f", "id": 250134, "pid": 76337, "tid": -914061504, "ts": 1716454225385063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385064, "dur": 0,
    "args": {
      "External id": 250135,
      "cbid": 203, "correlation": 250135
    }
  },
  {
    "ph": "f", "id": 250135, "pid": 76337, "tid": -914061504, "ts": 1716454225385064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385064, "dur": 0,
    "args": {
      "External id": 250136,
      "cbid": 205, "correlation": 250136
    }
  },
  {
    "ph": "f", "id": 250136, "pid": 76337, "tid": -914061504, "ts": 1716454225385064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441281, "dur": 5,
    "args": {
      "External id": 250140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250140,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250140, "pid": 5, "tid": 7, "ts": 1716454225441281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385078, "dur": 12,
    "args": {
      "External id": 250140,
      "cbid": 211, "correlation": 250140
    }
  },
  {
    "ph": "s", "id": 250140, "pid": 76337, "tid": -914061504, "ts": 1716454225385078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441287, "dur": 165,
    "args": {
      "External id": 250142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250142,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250142, "pid": 5, "tid": 7, "ts": 1716454225441287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385092, "dur": 5,
    "args": {
      "External id": 250142,
      "cbid": 211, "correlation": 250142
    }
  },
  {
    "ph": "s", "id": 250142, "pid": 76337, "tid": -914061504, "ts": 1716454225385092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225441454, "dur": 1,
    "args": {
      "External id": 250144,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250144,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 250144, "pid": 5, "tid": 7, "ts": 1716454225441454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385103, "dur": 7,
    "args": {
      "External id": 250144,
      "cbid": 51, "correlation": 250144
    }
  },
  {
    "ph": "s", "id": 250144, "pid": 76337, "tid": -914061504, "ts": 1716454225385103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225441458, "dur": 264,
    "args": {
      "External id": 250145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250145,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250145, "pid": 5, "tid": 7, "ts": 1716454225441458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385111, "dur": 6,
    "args": {
      "External id": 250145,
      "cbid": 211, "correlation": 250145
    }
  },
  {
    "ph": "s", "id": 250145, "pid": 76337, "tid": -914061504, "ts": 1716454225385111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441724, "dur": 6,
    "args": {
      "External id": 250147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250147,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250147, "pid": 5, "tid": 7, "ts": 1716454225441724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385121, "dur": 5,
    "args": {
      "External id": 250147,
      "cbid": 211, "correlation": 250147
    }
  },
  {
    "ph": "s", "id": 250147, "pid": 76337, "tid": -914061504, "ts": 1716454225385121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225441731, "dur": 6,
    "args": {
      "External id": 250153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250153,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250153, "pid": 5, "tid": 7, "ts": 1716454225441731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385149, "dur": 10,
    "args": {
      "External id": 250153,
      "cbid": 211, "correlation": 250153
    }
  },
  {
    "ph": "s", "id": 250153, "pid": 76337, "tid": -914061504, "ts": 1716454225385149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225441739, "dur": 5,
    "args": {
      "External id": 250161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250161,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250161, "pid": 5, "tid": 7, "ts": 1716454225441739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385182, "dur": 8,
    "args": {
      "External id": 250161,
      "cbid": 211, "correlation": 250161
    }
  },
  {
    "ph": "s", "id": 250161, "pid": 76337, "tid": -914061504, "ts": 1716454225385182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225441745, "dur": 5,
    "args": {
      "External id": 250169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250169,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250169, "pid": 5, "tid": 7, "ts": 1716454225441745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385212, "dur": 8,
    "args": {
      "External id": 250169,
      "cbid": 211, "correlation": 250169
    }
  },
  {
    "ph": "s", "id": 250169, "pid": 76337, "tid": -914061504, "ts": 1716454225385212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441751, "dur": 10,
    "args": {
      "External id": 250189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250189,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 250189, "pid": 5, "tid": 7, "ts": 1716454225441751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385285, "dur": 12,
    "args": {
      "External id": 250189,
      "cbid": 211, "correlation": 250189
    }
  },
  {
    "ph": "s", "id": 250189, "pid": 76337, "tid": -914061504, "ts": 1716454225385285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225441762, "dur": 4,
    "args": {
      "External id": 250201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250201,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250201, "pid": 5, "tid": 7, "ts": 1716454225441762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385306, "dur": 7,
    "args": {
      "External id": 250201,
      "cbid": 211, "correlation": 250201
    }
  },
  {
    "ph": "s", "id": 250201, "pid": 76337, "tid": -914061504, "ts": 1716454225385306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225441767, "dur": 7,
    "args": {
      "External id": 250204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250204, "pid": 5, "tid": 7, "ts": 1716454225441767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385325, "dur": 6,
    "args": {
      "External id": 250204,
      "cbid": 211, "correlation": 250204
    }
  },
  {
    "ph": "s", "id": 250204, "pid": 76337, "tid": -914061504, "ts": 1716454225385325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225441775, "dur": 5,
    "args": {
      "External id": 250213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250213,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250213, "pid": 5, "tid": 7, "ts": 1716454225441775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385363, "dur": 9,
    "args": {
      "External id": 250213,
      "cbid": 211, "correlation": 250213
    }
  },
  {
    "ph": "s", "id": 250213, "pid": 76337, "tid": -914061504, "ts": 1716454225385363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385413, "dur": 0,
    "args": {
      "External id": 250223,
      "cbid": 317, "correlation": 250223
    }
  },
  {
    "ph": "f", "id": 250223, "pid": 76337, "tid": -914061504, "ts": 1716454225385413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385414, "dur": 0,
    "args": {
      "External id": 250224,
      "cbid": 203, "correlation": 250224
    }
  },
  {
    "ph": "f", "id": 250224, "pid": 76337, "tid": -914061504, "ts": 1716454225385414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385415, "dur": 0,
    "args": {
      "External id": 250225,
      "cbid": 205, "correlation": 250225
    }
  },
  {
    "ph": "f", "id": 250225, "pid": 76337, "tid": -914061504, "ts": 1716454225385415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441781, "dur": 5,
    "args": {
      "External id": 250229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250229, "pid": 5, "tid": 7, "ts": 1716454225441781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385427, "dur": 12,
    "args": {
      "External id": 250229,
      "cbid": 211, "correlation": 250229
    }
  },
  {
    "ph": "s", "id": 250229, "pid": 76337, "tid": -914061504, "ts": 1716454225385427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225441788, "dur": 165,
    "args": {
      "External id": 250231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250231, "pid": 5, "tid": 7, "ts": 1716454225441788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385441, "dur": 5,
    "args": {
      "External id": 250231,
      "cbid": 211, "correlation": 250231
    }
  },
  {
    "ph": "s", "id": 250231, "pid": 76337, "tid": -914061504, "ts": 1716454225385441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225441955, "dur": 1,
    "args": {
      "External id": 250233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250233,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 250233, "pid": 5, "tid": 7, "ts": 1716454225441955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385452, "dur": 6,
    "args": {
      "External id": 250233,
      "cbid": 51, "correlation": 250233
    }
  },
  {
    "ph": "s", "id": 250233, "pid": 76337, "tid": -914061504, "ts": 1716454225385452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225441958, "dur": 263,
    "args": {
      "External id": 250234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250234,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250234, "pid": 5, "tid": 7, "ts": 1716454225441958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385459, "dur": 7,
    "args": {
      "External id": 250234,
      "cbid": 211, "correlation": 250234
    }
  },
  {
    "ph": "s", "id": 250234, "pid": 76337, "tid": -914061504, "ts": 1716454225385459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442223, "dur": 6,
    "args": {
      "External id": 250236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250236, "pid": 5, "tid": 7, "ts": 1716454225442223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385469, "dur": 5,
    "args": {
      "External id": 250236,
      "cbid": 211, "correlation": 250236
    }
  },
  {
    "ph": "s", "id": 250236, "pid": 76337, "tid": -914061504, "ts": 1716454225385469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225442230, "dur": 6,
    "args": {
      "External id": 250242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250242, "pid": 5, "tid": 7, "ts": 1716454225442230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385497, "dur": 8,
    "args": {
      "External id": 250242,
      "cbid": 211, "correlation": 250242
    }
  },
  {
    "ph": "s", "id": 250242, "pid": 76337, "tid": -914061504, "ts": 1716454225385497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225442237, "dur": 3,
    "args": {
      "External id": 250250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250250,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 250250, "pid": 5, "tid": 7, "ts": 1716454225442237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385540, "dur": 10,
    "args": {
      "External id": 250250,
      "cbid": 211, "correlation": 250250
    }
  },
  {
    "ph": "s", "id": 250250, "pid": 76337, "tid": -914061504, "ts": 1716454225385540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385603, "dur": 1,
    "args": {
      "External id": 250266,
      "cbid": 251, "correlation": 250266
    }
  },
  {
    "ph": "f", "id": 250266, "pid": 76337, "tid": -914061504, "ts": 1716454225385603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385608, "dur": 0,
    "args": {
      "External id": 250268,
      "cbid": 251, "correlation": 250268
    }
  },
  {
    "ph": "f", "id": 250268, "pid": 76337, "tid": -914061504, "ts": 1716454225385608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225442242, "dur": 10,
    "args": {
      "External id": 250269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250269,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250269, "pid": 5, "tid": 7, "ts": 1716454225442242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385611, "dur": 11,
    "args": {
      "External id": 250269,
      "cbid": 211, "correlation": 250269
    }
  },
  {
    "ph": "s", "id": 250269, "pid": 76337, "tid": -914061504, "ts": 1716454225385611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225442253, "dur": 4,
    "args": {
      "External id": 250271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250271,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250271, "pid": 5, "tid": 7, "ts": 1716454225442253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385624, "dur": 5,
    "args": {
      "External id": 250271,
      "cbid": 211, "correlation": 250271
    }
  },
  {
    "ph": "s", "id": 250271, "pid": 76337, "tid": -914061504, "ts": 1716454225385624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225442258, "dur": 6,
    "args": {
      "External id": 250281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250281, "pid": 5, "tid": 7, "ts": 1716454225442258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385679, "dur": 12,
    "args": {
      "External id": 250281,
      "cbid": 211, "correlation": 250281
    }
  },
  {
    "ph": "s", "id": 250281, "pid": 76337, "tid": -914061504, "ts": 1716454225385679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442265, "dur": 10,
    "args": {
      "External id": 250301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250301,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 250301, "pid": 5, "tid": 7, "ts": 1716454225442265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385744, "dur": 10,
    "args": {
      "External id": 250301,
      "cbid": 211, "correlation": 250301
    }
  },
  {
    "ph": "s", "id": 250301, "pid": 76337, "tid": -914061504, "ts": 1716454225385744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225442276, "dur": 4,
    "args": {
      "External id": 250313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250313,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250313, "pid": 5, "tid": 7, "ts": 1716454225442276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385765, "dur": 6,
    "args": {
      "External id": 250313,
      "cbid": 211, "correlation": 250313
    }
  },
  {
    "ph": "s", "id": 250313, "pid": 76337, "tid": -914061504, "ts": 1716454225385765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225442282, "dur": 7,
    "args": {
      "External id": 250316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250316, "pid": 5, "tid": 7, "ts": 1716454225442282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385783, "dur": 6,
    "args": {
      "External id": 250316,
      "cbid": 211, "correlation": 250316
    }
  },
  {
    "ph": "s", "id": 250316, "pid": 76337, "tid": -914061504, "ts": 1716454225385783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225442290, "dur": 5,
    "args": {
      "External id": 250325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250325,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250325, "pid": 5, "tid": 7, "ts": 1716454225442290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385823, "dur": 9,
    "args": {
      "External id": 250325,
      "cbid": 211, "correlation": 250325
    }
  },
  {
    "ph": "s", "id": 250325, "pid": 76337, "tid": -914061504, "ts": 1716454225385823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385884, "dur": 0,
    "args": {
      "External id": 250335,
      "cbid": 317, "correlation": 250335
    }
  },
  {
    "ph": "f", "id": 250335, "pid": 76337, "tid": -914061504, "ts": 1716454225385884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385884, "dur": 0,
    "args": {
      "External id": 250336,
      "cbid": 203, "correlation": 250336
    }
  },
  {
    "ph": "f", "id": 250336, "pid": 76337, "tid": -914061504, "ts": 1716454225385884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385885, "dur": 0,
    "args": {
      "External id": 250337,
      "cbid": 205, "correlation": 250337
    }
  },
  {
    "ph": "f", "id": 250337, "pid": 76337, "tid": -914061504, "ts": 1716454225385885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442296, "dur": 5,
    "args": {
      "External id": 250341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250341,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250341, "pid": 5, "tid": 7, "ts": 1716454225442296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385899, "dur": 12,
    "args": {
      "External id": 250341,
      "cbid": 211, "correlation": 250341
    }
  },
  {
    "ph": "s", "id": 250341, "pid": 76337, "tid": -914061504, "ts": 1716454225385899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442302, "dur": 165,
    "args": {
      "External id": 250343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250343, "pid": 5, "tid": 7, "ts": 1716454225442302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385913, "dur": 6,
    "args": {
      "External id": 250343,
      "cbid": 211, "correlation": 250343
    }
  },
  {
    "ph": "s", "id": 250343, "pid": 76337, "tid": -914061504, "ts": 1716454225385913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225442469, "dur": 1,
    "args": {
      "External id": 250345,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250345,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 250345, "pid": 5, "tid": 7, "ts": 1716454225442469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385925, "dur": 6,
    "args": {
      "External id": 250345,
      "cbid": 51, "correlation": 250345
    }
  },
  {
    "ph": "s", "id": 250345, "pid": 76337, "tid": -914061504, "ts": 1716454225385925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225442473, "dur": 263,
    "args": {
      "External id": 250346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250346,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250346, "pid": 5, "tid": 7, "ts": 1716454225442473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385932, "dur": 6,
    "args": {
      "External id": 250346,
      "cbid": 211, "correlation": 250346
    }
  },
  {
    "ph": "s", "id": 250346, "pid": 76337, "tid": -914061504, "ts": 1716454225385932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442738, "dur": 6,
    "args": {
      "External id": 250348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250348, "pid": 5, "tid": 7, "ts": 1716454225442738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385943, "dur": 5,
    "args": {
      "External id": 250348,
      "cbid": 211, "correlation": 250348
    }
  },
  {
    "ph": "s", "id": 250348, "pid": 76337, "tid": -914061504, "ts": 1716454225385943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225442745, "dur": 6,
    "args": {
      "External id": 250354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250354, "pid": 5, "tid": 7, "ts": 1716454225442745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225385972, "dur": 18,
    "args": {
      "External id": 250354,
      "cbid": 211, "correlation": 250354
    }
  },
  {
    "ph": "s", "id": 250354, "pid": 76337, "tid": -914061504, "ts": 1716454225385972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225442753, "dur": 5,
    "args": {
      "External id": 250362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250362,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250362, "pid": 5, "tid": 7, "ts": 1716454225442753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386014, "dur": 9,
    "args": {
      "External id": 250362,
      "cbid": 211, "correlation": 250362
    }
  },
  {
    "ph": "s", "id": 250362, "pid": 76337, "tid": -914061504, "ts": 1716454225386014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225442759, "dur": 5,
    "args": {
      "External id": 250370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250370,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250370, "pid": 5, "tid": 7, "ts": 1716454225442759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386044, "dur": 8,
    "args": {
      "External id": 250370,
      "cbid": 211, "correlation": 250370
    }
  },
  {
    "ph": "s", "id": 250370, "pid": 76337, "tid": -914061504, "ts": 1716454225386044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442765, "dur": 10,
    "args": {
      "External id": 250390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250390,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 250390, "pid": 5, "tid": 7, "ts": 1716454225442765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386146, "dur": 12,
    "args": {
      "External id": 250390,
      "cbid": 211, "correlation": 250390
    }
  },
  {
    "ph": "s", "id": 250390, "pid": 76337, "tid": -914061504, "ts": 1716454225386146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225442776, "dur": 4,
    "args": {
      "External id": 250402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250402,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250402, "pid": 5, "tid": 7, "ts": 1716454225442776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386168, "dur": 7,
    "args": {
      "External id": 250402,
      "cbid": 211, "correlation": 250402
    }
  },
  {
    "ph": "s", "id": 250402, "pid": 76337, "tid": -914061504, "ts": 1716454225386168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225442781, "dur": 7,
    "args": {
      "External id": 250405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250405,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250405, "pid": 5, "tid": 7, "ts": 1716454225442781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386186, "dur": 7,
    "args": {
      "External id": 250405,
      "cbid": 211, "correlation": 250405
    }
  },
  {
    "ph": "s", "id": 250405, "pid": 76337, "tid": -914061504, "ts": 1716454225386186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225442789, "dur": 5,
    "args": {
      "External id": 250414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250414,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250414, "pid": 5, "tid": 7, "ts": 1716454225442789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386225, "dur": 9,
    "args": {
      "External id": 250414,
      "cbid": 211, "correlation": 250414
    }
  },
  {
    "ph": "s", "id": 250414, "pid": 76337, "tid": -914061504, "ts": 1716454225386225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386282, "dur": 0,
    "args": {
      "External id": 250424,
      "cbid": 317, "correlation": 250424
    }
  },
  {
    "ph": "f", "id": 250424, "pid": 76337, "tid": -914061504, "ts": 1716454225386282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386283, "dur": 0,
    "args": {
      "External id": 250425,
      "cbid": 203, "correlation": 250425
    }
  },
  {
    "ph": "f", "id": 250425, "pid": 76337, "tid": -914061504, "ts": 1716454225386283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386283, "dur": 0,
    "args": {
      "External id": 250426,
      "cbid": 205, "correlation": 250426
    }
  },
  {
    "ph": "f", "id": 250426, "pid": 76337, "tid": -914061504, "ts": 1716454225386283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442795, "dur": 5,
    "args": {
      "External id": 250430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250430,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250430, "pid": 5, "tid": 7, "ts": 1716454225442795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386296, "dur": 12,
    "args": {
      "External id": 250430,
      "cbid": 211, "correlation": 250430
    }
  },
  {
    "ph": "s", "id": 250430, "pid": 76337, "tid": -914061504, "ts": 1716454225386296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225442802, "dur": 165,
    "args": {
      "External id": 250432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250432, "pid": 5, "tid": 7, "ts": 1716454225442802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386310, "dur": 5,
    "args": {
      "External id": 250432,
      "cbid": 211, "correlation": 250432
    }
  },
  {
    "ph": "s", "id": 250432, "pid": 76337, "tid": -914061504, "ts": 1716454225386310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225442969, "dur": 1,
    "args": {
      "External id": 250434,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250434,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 250434, "pid": 5, "tid": 7, "ts": 1716454225442969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386322, "dur": 6,
    "args": {
      "External id": 250434,
      "cbid": 51, "correlation": 250434
    }
  },
  {
    "ph": "s", "id": 250434, "pid": 76337, "tid": -914061504, "ts": 1716454225386322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225442972, "dur": 263,
    "args": {
      "External id": 250435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250435,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250435, "pid": 5, "tid": 7, "ts": 1716454225442972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386329, "dur": 6,
    "args": {
      "External id": 250435,
      "cbid": 211, "correlation": 250435
    }
  },
  {
    "ph": "s", "id": 250435, "pid": 76337, "tid": -914061504, "ts": 1716454225386329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443237, "dur": 6,
    "args": {
      "External id": 250437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250437,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250437, "pid": 5, "tid": 7, "ts": 1716454225443237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386339, "dur": 5,
    "args": {
      "External id": 250437,
      "cbid": 211, "correlation": 250437
    }
  },
  {
    "ph": "s", "id": 250437, "pid": 76337, "tid": -914061504, "ts": 1716454225386339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225443244, "dur": 7,
    "args": {
      "External id": 250443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250443, "pid": 5, "tid": 7, "ts": 1716454225443244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386367, "dur": 9,
    "args": {
      "External id": 250443,
      "cbid": 211, "correlation": 250443
    }
  },
  {
    "ph": "s", "id": 250443, "pid": 76337, "tid": -914061504, "ts": 1716454225386367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225443252, "dur": 3,
    "args": {
      "External id": 250451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250451,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 250451, "pid": 5, "tid": 7, "ts": 1716454225443252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386411, "dur": 10,
    "args": {
      "External id": 250451,
      "cbid": 211, "correlation": 250451
    }
  },
  {
    "ph": "s", "id": 250451, "pid": 76337, "tid": -914061504, "ts": 1716454225386411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386475, "dur": 1,
    "args": {
      "External id": 250467,
      "cbid": 251, "correlation": 250467
    }
  },
  {
    "ph": "f", "id": 250467, "pid": 76337, "tid": -914061504, "ts": 1716454225386475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386480, "dur": 0,
    "args": {
      "External id": 250469,
      "cbid": 251, "correlation": 250469
    }
  },
  {
    "ph": "f", "id": 250469, "pid": 76337, "tid": -914061504, "ts": 1716454225386480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225443256, "dur": 10,
    "args": {
      "External id": 250470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250470,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250470, "pid": 5, "tid": 7, "ts": 1716454225443256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386482, "dur": 11,
    "args": {
      "External id": 250470,
      "cbid": 211, "correlation": 250470
    }
  },
  {
    "ph": "s", "id": 250470, "pid": 76337, "tid": -914061504, "ts": 1716454225386482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225443268, "dur": 4,
    "args": {
      "External id": 250472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250472,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250472, "pid": 5, "tid": 7, "ts": 1716454225443268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386494, "dur": 5,
    "args": {
      "External id": 250472,
      "cbid": 211, "correlation": 250472
    }
  },
  {
    "ph": "s", "id": 250472, "pid": 76337, "tid": -914061504, "ts": 1716454225386494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225443273, "dur": 6,
    "args": {
      "External id": 250482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250482,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250482, "pid": 5, "tid": 7, "ts": 1716454225443273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386550, "dur": 12,
    "args": {
      "External id": 250482,
      "cbid": 211, "correlation": 250482
    }
  },
  {
    "ph": "s", "id": 250482, "pid": 76337, "tid": -914061504, "ts": 1716454225386550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443280, "dur": 10,
    "args": {
      "External id": 250502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250502,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 250502, "pid": 5, "tid": 7, "ts": 1716454225443280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386615, "dur": 11,
    "args": {
      "External id": 250502,
      "cbid": 211, "correlation": 250502
    }
  },
  {
    "ph": "s", "id": 250502, "pid": 76337, "tid": -914061504, "ts": 1716454225386615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225443291, "dur": 4,
    "args": {
      "External id": 250514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250514,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250514, "pid": 5, "tid": 7, "ts": 1716454225443291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386636, "dur": 6,
    "args": {
      "External id": 250514,
      "cbid": 211, "correlation": 250514
    }
  },
  {
    "ph": "s", "id": 250514, "pid": 76337, "tid": -914061504, "ts": 1716454225386636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225443296, "dur": 7,
    "args": {
      "External id": 250517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250517, "pid": 5, "tid": 7, "ts": 1716454225443296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386655, "dur": 6,
    "args": {
      "External id": 250517,
      "cbid": 211, "correlation": 250517
    }
  },
  {
    "ph": "s", "id": 250517, "pid": 76337, "tid": -914061504, "ts": 1716454225386655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225443304, "dur": 5,
    "args": {
      "External id": 250526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250526,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250526, "pid": 5, "tid": 7, "ts": 1716454225443304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386695, "dur": 10,
    "args": {
      "External id": 250526,
      "cbid": 211, "correlation": 250526
    }
  },
  {
    "ph": "s", "id": 250526, "pid": 76337, "tid": -914061504, "ts": 1716454225386695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386758, "dur": 0,
    "args": {
      "External id": 250536,
      "cbid": 317, "correlation": 250536
    }
  },
  {
    "ph": "f", "id": 250536, "pid": 76337, "tid": -914061504, "ts": 1716454225386758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386759, "dur": 0,
    "args": {
      "External id": 250537,
      "cbid": 203, "correlation": 250537
    }
  },
  {
    "ph": "f", "id": 250537, "pid": 76337, "tid": -914061504, "ts": 1716454225386759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386760, "dur": 0,
    "args": {
      "External id": 250538,
      "cbid": 205, "correlation": 250538
    }
  },
  {
    "ph": "f", "id": 250538, "pid": 76337, "tid": -914061504, "ts": 1716454225386760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443310, "dur": 5,
    "args": {
      "External id": 250542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250542, "pid": 5, "tid": 7, "ts": 1716454225443310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386774, "dur": 12,
    "args": {
      "External id": 250542,
      "cbid": 211, "correlation": 250542
    }
  },
  {
    "ph": "s", "id": 250542, "pid": 76337, "tid": -914061504, "ts": 1716454225386774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443316, "dur": 166,
    "args": {
      "External id": 250544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250544,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250544, "pid": 5, "tid": 7, "ts": 1716454225443316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386788, "dur": 5,
    "args": {
      "External id": 250544,
      "cbid": 211, "correlation": 250544
    }
  },
  {
    "ph": "s", "id": 250544, "pid": 76337, "tid": -914061504, "ts": 1716454225386788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225443485, "dur": 1,
    "args": {
      "External id": 250546,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250546,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 250546, "pid": 5, "tid": 7, "ts": 1716454225443485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386799, "dur": 6,
    "args": {
      "External id": 250546,
      "cbid": 51, "correlation": 250546
    }
  },
  {
    "ph": "s", "id": 250546, "pid": 76337, "tid": -914061504, "ts": 1716454225386799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225443488, "dur": 263,
    "args": {
      "External id": 250547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250547,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250547, "pid": 5, "tid": 7, "ts": 1716454225443488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386806, "dur": 6,
    "args": {
      "External id": 250547,
      "cbid": 211, "correlation": 250547
    }
  },
  {
    "ph": "s", "id": 250547, "pid": 76337, "tid": -914061504, "ts": 1716454225386806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443753, "dur": 6,
    "args": {
      "External id": 250549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250549,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250549, "pid": 5, "tid": 7, "ts": 1716454225443753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386816, "dur": 5,
    "args": {
      "External id": 250549,
      "cbid": 211, "correlation": 250549
    }
  },
  {
    "ph": "s", "id": 250549, "pid": 76337, "tid": -914061504, "ts": 1716454225386816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225443761, "dur": 6,
    "args": {
      "External id": 250555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250555, "pid": 5, "tid": 7, "ts": 1716454225443761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386844, "dur": 9,
    "args": {
      "External id": 250555,
      "cbid": 211, "correlation": 250555
    }
  },
  {
    "ph": "s", "id": 250555, "pid": 76337, "tid": -914061504, "ts": 1716454225386844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225443768, "dur": 5,
    "args": {
      "External id": 250563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250563,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250563, "pid": 5, "tid": 7, "ts": 1716454225443768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386877, "dur": 8,
    "args": {
      "External id": 250563,
      "cbid": 211, "correlation": 250563
    }
  },
  {
    "ph": "s", "id": 250563, "pid": 76337, "tid": -914061504, "ts": 1716454225386877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225443775, "dur": 5,
    "args": {
      "External id": 250571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250571,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250571, "pid": 5, "tid": 7, "ts": 1716454225443775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225386905, "dur": 8,
    "args": {
      "External id": 250571,
      "cbid": 211, "correlation": 250571
    }
  },
  {
    "ph": "s", "id": 250571, "pid": 76337, "tid": -914061504, "ts": 1716454225386905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443781, "dur": 10,
    "args": {
      "External id": 250591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250591,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 250591, "pid": 5, "tid": 7, "ts": 1716454225443781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387059, "dur": 14,
    "args": {
      "External id": 250591,
      "cbid": 211, "correlation": 250591
    }
  },
  {
    "ph": "s", "id": 250591, "pid": 76337, "tid": -914061504, "ts": 1716454225387059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225443792, "dur": 4,
    "args": {
      "External id": 250603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250603,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 250603, "pid": 5, "tid": 7, "ts": 1716454225443792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387084, "dur": 6,
    "args": {
      "External id": 250603,
      "cbid": 211, "correlation": 250603
    }
  },
  {
    "ph": "s", "id": 250603, "pid": 76337, "tid": -914061504, "ts": 1716454225387084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225443797, "dur": 7,
    "args": {
      "External id": 250606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250606,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250606, "pid": 5, "tid": 7, "ts": 1716454225443797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387102, "dur": 7,
    "args": {
      "External id": 250606,
      "cbid": 211, "correlation": 250606
    }
  },
  {
    "ph": "s", "id": 250606, "pid": 76337, "tid": -914061504, "ts": 1716454225387102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387162, "dur": 0,
    "args": {
      "External id": 250617,
      "cbid": 317, "correlation": 250617
    }
  },
  {
    "ph": "f", "id": 250617, "pid": 76337, "tid": -914061504, "ts": 1716454225387162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387163, "dur": 0,
    "args": {
      "External id": 250618,
      "cbid": 203, "correlation": 250618
    }
  },
  {
    "ph": "f", "id": 250618, "pid": 76337, "tid": -914061504, "ts": 1716454225387163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387164, "dur": 0,
    "args": {
      "External id": 250619,
      "cbid": 205, "correlation": 250619
    }
  },
  {
    "ph": "f", "id": 250619, "pid": 76337, "tid": -914061504, "ts": 1716454225387164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443805, "dur": 5,
    "args": {
      "External id": 250623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250623,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250623, "pid": 5, "tid": 7, "ts": 1716454225443805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387186, "dur": 12,
    "args": {
      "External id": 250623,
      "cbid": 211, "correlation": 250623
    }
  },
  {
    "ph": "s", "id": 250623, "pid": 76337, "tid": -914061504, "ts": 1716454225387186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225443811, "dur": 38,
    "args": {
      "External id": 250625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250625,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 250625, "pid": 5, "tid": 7, "ts": 1716454225443811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387205, "dur": 9,
    "args": {
      "External id": 250625,
      "cbid": 211, "correlation": 250625
    }
  },
  {
    "ph": "s", "id": 250625, "pid": 76337, "tid": -914061504, "ts": 1716454225387205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443851, "dur": 5,
    "args": {
      "External id": 250627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250627, "pid": 5, "tid": 7, "ts": 1716454225443851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387218, "dur": 5,
    "args": {
      "External id": 250627,
      "cbid": 211, "correlation": 250627
    }
  },
  {
    "ph": "s", "id": 250627, "pid": 76337, "tid": -914061504, "ts": 1716454225387218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225443858, "dur": 6,
    "args": {
      "External id": 250633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250633, "pid": 5, "tid": 7, "ts": 1716454225443858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387245, "dur": 8,
    "args": {
      "External id": 250633,
      "cbid": 211, "correlation": 250633
    }
  },
  {
    "ph": "s", "id": 250633, "pid": 76337, "tid": -914061504, "ts": 1716454225387245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225443865, "dur": 21,
    "args": {
      "External id": 250642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250642,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250642, "pid": 5, "tid": 7, "ts": 1716454225443865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387329, "dur": 14,
    "args": {
      "External id": 250642,
      "cbid": 211, "correlation": 250642
    }
  },
  {
    "ph": "s", "id": 250642, "pid": 76337, "tid": -914061504, "ts": 1716454225387329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225443887, "dur": 11,
    "args": {
      "External id": 250664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250664,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 250664, "pid": 5, "tid": 7, "ts": 1716454225443887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387385, "dur": 10,
    "args": {
      "External id": 250664,
      "cbid": 211, "correlation": 250664
    }
  },
  {
    "ph": "s", "id": 250664, "pid": 76337, "tid": -914061504, "ts": 1716454225387385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387476, "dur": 2,
    "args": {
      "External id": 250675,
      "cbid": 251, "correlation": 250675
    }
  },
  {
    "ph": "f", "id": 250675, "pid": 76337, "tid": -914061504, "ts": 1716454225387476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387481, "dur": 0,
    "args": {
      "External id": 250676,
      "cbid": 251, "correlation": 250676
    }
  },
  {
    "ph": "f", "id": 250676, "pid": 76337, "tid": -914061504, "ts": 1716454225387481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225443900, "dur": 55,
    "args": {
      "External id": 250677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250677,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 250677, "pid": 5, "tid": 7, "ts": 1716454225443900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387484, "dur": 15,
    "args": {
      "External id": 250677,
      "cbid": 211, "correlation": 250677
    }
  },
  {
    "ph": "s", "id": 250677, "pid": 76337, "tid": -914061504, "ts": 1716454225387484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387555, "dur": 1,
    "args": {
      "External id": 250688,
      "cbid": 251, "correlation": 250688
    }
  },
  {
    "ph": "f", "id": 250688, "pid": 76337, "tid": -914061504, "ts": 1716454225387555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387559, "dur": 0,
    "args": {
      "External id": 250689,
      "cbid": 251, "correlation": 250689
    }
  },
  {
    "ph": "f", "id": 250689, "pid": 76337, "tid": -914061504, "ts": 1716454225387559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225443956, "dur": 53,
    "args": {
      "External id": 250690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250690,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 250690, "pid": 5, "tid": 7, "ts": 1716454225443956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387561, "dur": 13,
    "args": {
      "External id": 250690,
      "cbid": 211, "correlation": 250690
    }
  },
  {
    "ph": "s", "id": 250690, "pid": 76337, "tid": -914061504, "ts": 1716454225387561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387632, "dur": 1,
    "args": {
      "External id": 250701,
      "cbid": 251, "correlation": 250701
    }
  },
  {
    "ph": "f", "id": 250701, "pid": 76337, "tid": -914061504, "ts": 1716454225387632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387635, "dur": 0,
    "args": {
      "External id": 250702,
      "cbid": 251, "correlation": 250702
    }
  },
  {
    "ph": "f", "id": 250702, "pid": 76337, "tid": -914061504, "ts": 1716454225387635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225444011, "dur": 54,
    "args": {
      "External id": 250703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250703,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 250703, "pid": 5, "tid": 7, "ts": 1716454225444011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387637, "dur": 11,
    "args": {
      "External id": 250703,
      "cbid": 211, "correlation": 250703
    }
  },
  {
    "ph": "s", "id": 250703, "pid": 76337, "tid": -914061504, "ts": 1716454225387637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225444066, "dur": 57,
    "args": {
      "External id": 250728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250728,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250728, "pid": 5, "tid": 7, "ts": 1716454225444066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387722, "dur": 14,
    "args": {
      "External id": 250728,
      "cbid": 211, "correlation": 250728
    }
  },
  {
    "ph": "s", "id": 250728, "pid": 76337, "tid": -914061504, "ts": 1716454225387722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387821, "dur": 1,
    "args": {
      "External id": 250746,
      "cbid": 251, "correlation": 250746
    }
  },
  {
    "ph": "f", "id": 250746, "pid": 76337, "tid": -914061504, "ts": 1716454225387821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225444124, "dur": 66,
    "args": {
      "External id": 250748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250748,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 250748, "pid": 5, "tid": 7, "ts": 1716454225444124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387827, "dur": 13,
    "args": {
      "External id": 250748,
      "cbid": 211, "correlation": 250748
    }
  },
  {
    "ph": "s", "id": 250748, "pid": 76337, "tid": -914061504, "ts": 1716454225387827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225444191, "dur": 6,
    "args": {
      "External id": 250756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250756,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250756, "pid": 5, "tid": 7, "ts": 1716454225444191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387900, "dur": 12,
    "args": {
      "External id": 250756,
      "cbid": 211, "correlation": 250756
    }
  },
  {
    "ph": "s", "id": 250756, "pid": 76337, "tid": -914061504, "ts": 1716454225387900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225444199, "dur": 7,
    "args": {
      "External id": 250764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250764,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250764, "pid": 5, "tid": 7, "ts": 1716454225444199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225387941, "dur": 9,
    "args": {
      "External id": 250764,
      "cbid": 211, "correlation": 250764
    }
  },
  {
    "ph": "s", "id": 250764, "pid": 76337, "tid": -914061504, "ts": 1716454225387941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444207, "dur": 8,
    "args": {
      "External id": 250775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250775, "pid": 5, "tid": 7, "ts": 1716454225444207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388023, "dur": 14,
    "args": {
      "External id": 250775,
      "cbid": 211, "correlation": 250775
    }
  },
  {
    "ph": "s", "id": 250775, "pid": 76337, "tid": -914061504, "ts": 1716454225388023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444217, "dur": 10,
    "args": {
      "External id": 250797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250797,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 250797, "pid": 5, "tid": 7, "ts": 1716454225444217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388057, "dur": 8,
    "args": {
      "External id": 250797,
      "cbid": 211, "correlation": 250797
    }
  },
  {
    "ph": "s", "id": 250797, "pid": 76337, "tid": -914061504, "ts": 1716454225388057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388144, "dur": 2,
    "args": {
      "External id": 250808,
      "cbid": 251, "correlation": 250808
    }
  },
  {
    "ph": "f", "id": 250808, "pid": 76337, "tid": -914061504, "ts": 1716454225388144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225444228, "dur": 1,
    "args": {
      "External id": 250809,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250809,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3
    }
  },
  {
    "ph": "f", "id": 250809, "pid": 5, "tid": 7, "ts": 1716454225444228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388149, "dur": 12,
    "args": {
      "External id": 250809,
      "cbid": 51, "correlation": 250809
    }
  },
  {
    "ph": "s", "id": 250809, "pid": 76337, "tid": -914061504, "ts": 1716454225388149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225444232, "dur": 37,
    "args": {
      "External id": 250810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250810,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 250810, "pid": 5, "tid": 7, "ts": 1716454225444232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388164, "dur": 13,
    "args": {
      "External id": 250810,
      "cbid": 211, "correlation": 250810
    }
  },
  {
    "ph": "s", "id": 250810, "pid": 76337, "tid": -914061504, "ts": 1716454225388164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388237, "dur": 1,
    "args": {
      "External id": 250821,
      "cbid": 251, "correlation": 250821
    }
  },
  {
    "ph": "f", "id": 250821, "pid": 76337, "tid": -914061504, "ts": 1716454225388237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388241, "dur": 0,
    "args": {
      "External id": 250822,
      "cbid": 251, "correlation": 250822
    }
  },
  {
    "ph": "f", "id": 250822, "pid": 76337, "tid": -914061504, "ts": 1716454225388241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225444270, "dur": 12,
    "args": {
      "External id": 250823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250823,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250823, "pid": 5, "tid": 7, "ts": 1716454225444270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388242, "dur": 12,
    "args": {
      "External id": 250823,
      "cbid": 211, "correlation": 250823
    }
  },
  {
    "ph": "s", "id": 250823, "pid": 76337, "tid": -914061504, "ts": 1716454225388242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225444284, "dur": 5,
    "args": {
      "External id": 250825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250825,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250825, "pid": 5, "tid": 7, "ts": 1716454225444284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388256, "dur": 6,
    "args": {
      "External id": 250825,
      "cbid": 211, "correlation": 250825
    }
  },
  {
    "ph": "s", "id": 250825, "pid": 76337, "tid": -914061504, "ts": 1716454225388256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388314, "dur": 1,
    "args": {
      "External id": 250836,
      "cbid": 251, "correlation": 250836
    }
  },
  {
    "ph": "f", "id": 250836, "pid": 76337, "tid": -914061504, "ts": 1716454225388314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388317, "dur": 0,
    "args": {
      "External id": 250837,
      "cbid": 251, "correlation": 250837
    }
  },
  {
    "ph": "f", "id": 250837, "pid": 76337, "tid": -914061504, "ts": 1716454225388317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225444290, "dur": 8,
    "args": {
      "External id": 250838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250838,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250838, "pid": 5, "tid": 7, "ts": 1716454225444290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388319, "dur": 13,
    "args": {
      "External id": 250838,
      "cbid": 211, "correlation": 250838
    }
  },
  {
    "ph": "s", "id": 250838, "pid": 76337, "tid": -914061504, "ts": 1716454225388319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225444300, "dur": 4,
    "args": {
      "External id": 250840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250840,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250840, "pid": 5, "tid": 7, "ts": 1716454225444300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388333, "dur": 5,
    "args": {
      "External id": 250840,
      "cbid": 211, "correlation": 250840
    }
  },
  {
    "ph": "s", "id": 250840, "pid": 76337, "tid": -914061504, "ts": 1716454225388333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225444305, "dur": 20,
    "args": {
      "External id": 250865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250865,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 250865, "pid": 5, "tid": 7, "ts": 1716454225444305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388411, "dur": 12,
    "args": {
      "External id": 250865,
      "cbid": 211, "correlation": 250865
    }
  },
  {
    "ph": "s", "id": 250865, "pid": 76337, "tid": -914061504, "ts": 1716454225388411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388511, "dur": 2,
    "args": {
      "External id": 250883,
      "cbid": 251, "correlation": 250883
    }
  },
  {
    "ph": "f", "id": 250883, "pid": 76337, "tid": -914061504, "ts": 1716454225388511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225444327, "dur": 1,
    "args": {
      "External id": 250885,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250885,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 250885, "pid": 5, "tid": 7, "ts": 1716454225444327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388517, "dur": 11,
    "args": {
      "External id": 250885,
      "cbid": 51, "correlation": 250885
    }
  },
  {
    "ph": "s", "id": 250885, "pid": 76337, "tid": -914061504, "ts": 1716454225388517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225444330, "dur": 38,
    "args": {
      "External id": 250886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250886,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 250886, "pid": 5, "tid": 7, "ts": 1716454225444330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388530, "dur": 12,
    "args": {
      "External id": 250886,
      "cbid": 211, "correlation": 250886
    }
  },
  {
    "ph": "s", "id": 250886, "pid": 76337, "tid": -914061504, "ts": 1716454225388530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225444370, "dur": 4,
    "args": {
      "External id": 250894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250894,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250894, "pid": 5, "tid": 7, "ts": 1716454225444370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388600, "dur": 12,
    "args": {
      "External id": 250894,
      "cbid": 211, "correlation": 250894
    }
  },
  {
    "ph": "s", "id": 250894, "pid": 76337, "tid": -914061504, "ts": 1716454225388600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444375, "dur": 8,
    "args": {
      "External id": 250902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250902, "pid": 5, "tid": 7, "ts": 1716454225444375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388642, "dur": 9,
    "args": {
      "External id": 250902,
      "cbid": 211, "correlation": 250902
    }
  },
  {
    "ph": "s", "id": 250902, "pid": 76337, "tid": -914061504, "ts": 1716454225388642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444385, "dur": 8,
    "args": {
      "External id": 250924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250924,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 250924, "pid": 5, "tid": 7, "ts": 1716454225444385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388693, "dur": 10,
    "args": {
      "External id": 250924,
      "cbid": 211, "correlation": 250924
    }
  },
  {
    "ph": "s", "id": 250924, "pid": 76337, "tid": -914061504, "ts": 1716454225388693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388786, "dur": 1,
    "args": {
      "External id": 250940,
      "cbid": 251, "correlation": 250940
    }
  },
  {
    "ph": "f", "id": 250940, "pid": 76337, "tid": -914061504, "ts": 1716454225388786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388791, "dur": 0,
    "args": {
      "External id": 250942,
      "cbid": 251, "correlation": 250942
    }
  },
  {
    "ph": "f", "id": 250942, "pid": 76337, "tid": -914061504, "ts": 1716454225388791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225444395, "dur": 192,
    "args": {
      "External id": 250943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250943,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 250943, "pid": 5, "tid": 7, "ts": 1716454225444395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388793, "dur": 13,
    "args": {
      "External id": 250943,
      "cbid": 211, "correlation": 250943
    }
  },
  {
    "ph": "s", "id": 250943, "pid": 76337, "tid": -914061504, "ts": 1716454225388793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444589, "dur": 21,
    "args": {
      "External id": 250951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250951,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250951, "pid": 5, "tid": 7, "ts": 1716454225444589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388860, "dur": 12,
    "args": {
      "External id": 250951,
      "cbid": 211, "correlation": 250951
    }
  },
  {
    "ph": "s", "id": 250951, "pid": 76337, "tid": -914061504, "ts": 1716454225388860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444611, "dur": 22,
    "args": {
      "External id": 250959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 250959, "pid": 5, "tid": 7, "ts": 1716454225444611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388891, "dur": 8,
    "args": {
      "External id": 250959,
      "cbid": 211, "correlation": 250959
    }
  },
  {
    "ph": "s", "id": 250959, "pid": 76337, "tid": -914061504, "ts": 1716454225388891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388971, "dur": 8,
    "args": {
      "External id": 250975,
      "cbid": 251, "correlation": 250975
    }
  },
  {
    "ph": "f", "id": 250975, "pid": 76337, "tid": -914061504, "ts": 1716454225388971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225444635, "dur": 1,
    "args": {
      "External id": 250977,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 250977,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 250977, "pid": 5, "tid": 7, "ts": 1716454225444635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388984, "dur": 9,
    "args": {
      "External id": 250977,
      "cbid": 51, "correlation": 250977
    }
  },
  {
    "ph": "s", "id": 250977, "pid": 76337, "tid": -914061504, "ts": 1716454225388984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225444639, "dur": 111,
    "args": {
      "External id": 250978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250978,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 250978, "pid": 5, "tid": 7, "ts": 1716454225444639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225388994, "dur": 12,
    "args": {
      "External id": 250978,
      "cbid": 211, "correlation": 250978
    }
  },
  {
    "ph": "s", "id": 250978, "pid": 76337, "tid": -914061504, "ts": 1716454225388994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225444752, "dur": 5,
    "args": {
      "External id": 250986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250986,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250986, "pid": 5, "tid": 7, "ts": 1716454225444752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389038, "dur": 10,
    "args": {
      "External id": 250986,
      "cbid": 211, "correlation": 250986
    }
  },
  {
    "ph": "s", "id": 250986, "pid": 76337, "tid": -914061504, "ts": 1716454225389038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444759, "dur": 10,
    "args": {
      "External id": 250997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 250997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 250997, "pid": 5, "tid": 7, "ts": 1716454225444759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389107, "dur": 12,
    "args": {
      "External id": 250997,
      "cbid": 211, "correlation": 250997
    }
  },
  {
    "ph": "s", "id": 250997, "pid": 76337, "tid": -914061504, "ts": 1716454225389107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389170, "dur": 0,
    "args": {
      "External id": 251009,
      "cbid": 317, "correlation": 251009
    }
  },
  {
    "ph": "f", "id": 251009, "pid": 76337, "tid": -914061504, "ts": 1716454225389170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389171, "dur": 0,
    "args": {
      "External id": 251010,
      "cbid": 203, "correlation": 251010
    }
  },
  {
    "ph": "f", "id": 251010, "pid": 76337, "tid": -914061504, "ts": 1716454225389171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389172, "dur": 0,
    "args": {
      "External id": 251011,
      "cbid": 205, "correlation": 251011
    }
  },
  {
    "ph": "f", "id": 251011, "pid": 76337, "tid": -914061504, "ts": 1716454225389172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444770, "dur": 6,
    "args": {
      "External id": 251015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251015,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251015, "pid": 5, "tid": 7, "ts": 1716454225444770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389188, "dur": 12,
    "args": {
      "External id": 251015,
      "cbid": 211, "correlation": 251015
    }
  },
  {
    "ph": "s", "id": 251015, "pid": 76337, "tid": -914061504, "ts": 1716454225389188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225444777, "dur": 38,
    "args": {
      "External id": 251017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251017,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 251017, "pid": 5, "tid": 7, "ts": 1716454225444777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389206, "dur": 7,
    "args": {
      "External id": 251017,
      "cbid": 211, "correlation": 251017
    }
  },
  {
    "ph": "s", "id": 251017, "pid": 76337, "tid": -914061504, "ts": 1716454225389206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444816, "dur": 6,
    "args": {
      "External id": 251019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251019,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251019, "pid": 5, "tid": 7, "ts": 1716454225444816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389217, "dur": 5,
    "args": {
      "External id": 251019,
      "cbid": 211, "correlation": 251019
    }
  },
  {
    "ph": "s", "id": 251019, "pid": 76337, "tid": -914061504, "ts": 1716454225389217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444824, "dur": 7,
    "args": {
      "External id": 251025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251025, "pid": 5, "tid": 7, "ts": 1716454225444824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389244, "dur": 8,
    "args": {
      "External id": 251025,
      "cbid": 211, "correlation": 251025
    }
  },
  {
    "ph": "s", "id": 251025, "pid": 76337, "tid": -914061504, "ts": 1716454225389244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225444832, "dur": 5,
    "args": {
      "External id": 251033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251033,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251033, "pid": 5, "tid": 7, "ts": 1716454225444832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389275, "dur": 9,
    "args": {
      "External id": 251033,
      "cbid": 211, "correlation": 251033
    }
  },
  {
    "ph": "s", "id": 251033, "pid": 76337, "tid": -914061504, "ts": 1716454225389275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444839, "dur": 11,
    "args": {
      "External id": 251053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251053,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251053, "pid": 5, "tid": 7, "ts": 1716454225444839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389348, "dur": 11,
    "args": {
      "External id": 251053,
      "cbid": 211, "correlation": 251053
    }
  },
  {
    "ph": "s", "id": 251053, "pid": 76337, "tid": -914061504, "ts": 1716454225389348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225444850, "dur": 4,
    "args": {
      "External id": 251065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251065,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 251065, "pid": 5, "tid": 7, "ts": 1716454225444850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389369, "dur": 6,
    "args": {
      "External id": 251065,
      "cbid": 211, "correlation": 251065
    }
  },
  {
    "ph": "s", "id": 251065, "pid": 76337, "tid": -914061504, "ts": 1716454225389369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225444856, "dur": 8,
    "args": {
      "External id": 251068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251068,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251068, "pid": 5, "tid": 7, "ts": 1716454225444856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389388, "dur": 6,
    "args": {
      "External id": 251068,
      "cbid": 211, "correlation": 251068
    }
  },
  {
    "ph": "s", "id": 251068, "pid": 76337, "tid": -914061504, "ts": 1716454225389388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225444866, "dur": 5,
    "args": {
      "External id": 251077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251077,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251077, "pid": 5, "tid": 7, "ts": 1716454225444866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389425, "dur": 11,
    "args": {
      "External id": 251077,
      "cbid": 211, "correlation": 251077
    }
  },
  {
    "ph": "s", "id": 251077, "pid": 76337, "tid": -914061504, "ts": 1716454225389425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389477, "dur": 0,
    "args": {
      "External id": 251087,
      "cbid": 317, "correlation": 251087
    }
  },
  {
    "ph": "f", "id": 251087, "pid": 76337, "tid": -914061504, "ts": 1716454225389477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389478, "dur": 0,
    "args": {
      "External id": 251088,
      "cbid": 203, "correlation": 251088
    }
  },
  {
    "ph": "f", "id": 251088, "pid": 76337, "tid": -914061504, "ts": 1716454225389478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389479, "dur": 0,
    "args": {
      "External id": 251089,
      "cbid": 205, "correlation": 251089
    }
  },
  {
    "ph": "f", "id": 251089, "pid": 76337, "tid": -914061504, "ts": 1716454225389479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444872, "dur": 5,
    "args": {
      "External id": 251093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251093,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251093, "pid": 5, "tid": 7, "ts": 1716454225444872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389492, "dur": 11,
    "args": {
      "External id": 251093,
      "cbid": 211, "correlation": 251093
    }
  },
  {
    "ph": "s", "id": 251093, "pid": 76337, "tid": -914061504, "ts": 1716454225389492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225444879, "dur": 165,
    "args": {
      "External id": 251095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251095,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251095, "pid": 5, "tid": 7, "ts": 1716454225444879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389506, "dur": 5,
    "args": {
      "External id": 251095,
      "cbid": 211, "correlation": 251095
    }
  },
  {
    "ph": "s", "id": 251095, "pid": 76337, "tid": -914061504, "ts": 1716454225389506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225445046, "dur": 1,
    "args": {
      "External id": 251097,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251097,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15315890236119975
    }
  },
  {
    "ph": "f", "id": 251097, "pid": 5, "tid": 7, "ts": 1716454225445046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389516, "dur": 6,
    "args": {
      "External id": 251097,
      "cbid": 51, "correlation": 251097
    }
  },
  {
    "ph": "s", "id": 251097, "pid": 76337, "tid": -914061504, "ts": 1716454225389516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225445050, "dur": 275,
    "args": {
      "External id": 251098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251098,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251098, "pid": 5, "tid": 7, "ts": 1716454225445050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389523, "dur": 7,
    "args": {
      "External id": 251098,
      "cbid": 211, "correlation": 251098
    }
  },
  {
    "ph": "s", "id": 251098, "pid": 76337, "tid": -914061504, "ts": 1716454225389523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445326, "dur": 6,
    "args": {
      "External id": 251100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251100, "pid": 5, "tid": 7, "ts": 1716454225445326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389534, "dur": 5,
    "args": {
      "External id": 251100,
      "cbid": 211, "correlation": 251100
    }
  },
  {
    "ph": "s", "id": 251100, "pid": 76337, "tid": -914061504, "ts": 1716454225389534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225445333, "dur": 6,
    "args": {
      "External id": 251106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251106, "pid": 5, "tid": 7, "ts": 1716454225445333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389561, "dur": 8,
    "args": {
      "External id": 251106,
      "cbid": 211, "correlation": 251106
    }
  },
  {
    "ph": "s", "id": 251106, "pid": 76337, "tid": -914061504, "ts": 1716454225389561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225445341, "dur": 3,
    "args": {
      "External id": 251114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251114,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 251114, "pid": 5, "tid": 7, "ts": 1716454225445341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389605, "dur": 9,
    "args": {
      "External id": 251114,
      "cbid": 211, "correlation": 251114
    }
  },
  {
    "ph": "s", "id": 251114, "pid": 76337, "tid": -914061504, "ts": 1716454225389605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389669, "dur": 1,
    "args": {
      "External id": 251130,
      "cbid": 251, "correlation": 251130
    }
  },
  {
    "ph": "f", "id": 251130, "pid": 76337, "tid": -914061504, "ts": 1716454225389669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389674, "dur": 0,
    "args": {
      "External id": 251132,
      "cbid": 251, "correlation": 251132
    }
  },
  {
    "ph": "f", "id": 251132, "pid": 76337, "tid": -914061504, "ts": 1716454225389674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225445345, "dur": 13,
    "args": {
      "External id": 251133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251133,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251133, "pid": 5, "tid": 7, "ts": 1716454225445345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389676, "dur": 11,
    "args": {
      "External id": 251133,
      "cbid": 211, "correlation": 251133
    }
  },
  {
    "ph": "s", "id": 251133, "pid": 76337, "tid": -914061504, "ts": 1716454225389676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225445359, "dur": 5,
    "args": {
      "External id": 251135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251135,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251135, "pid": 5, "tid": 7, "ts": 1716454225445359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389689, "dur": 5,
    "args": {
      "External id": 251135,
      "cbid": 211, "correlation": 251135
    }
  },
  {
    "ph": "s", "id": 251135, "pid": 76337, "tid": -914061504, "ts": 1716454225389689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225445366, "dur": 6,
    "args": {
      "External id": 251145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251145,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251145, "pid": 5, "tid": 7, "ts": 1716454225445366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389746, "dur": 12,
    "args": {
      "External id": 251145,
      "cbid": 211, "correlation": 251145
    }
  },
  {
    "ph": "s", "id": 251145, "pid": 76337, "tid": -914061504, "ts": 1716454225389746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445373, "dur": 10,
    "args": {
      "External id": 251165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251165,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251165, "pid": 5, "tid": 7, "ts": 1716454225445373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389812, "dur": 10,
    "args": {
      "External id": 251165,
      "cbid": 211, "correlation": 251165
    }
  },
  {
    "ph": "s", "id": 251165, "pid": 76337, "tid": -914061504, "ts": 1716454225389812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225445384, "dur": 4,
    "args": {
      "External id": 251177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251177,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 251177, "pid": 5, "tid": 7, "ts": 1716454225445384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389832, "dur": 7,
    "args": {
      "External id": 251177,
      "cbid": 211, "correlation": 251177
    }
  },
  {
    "ph": "s", "id": 251177, "pid": 76337, "tid": -914061504, "ts": 1716454225389832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225445389, "dur": 7,
    "args": {
      "External id": 251180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251180, "pid": 5, "tid": 7, "ts": 1716454225445389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389851, "dur": 6,
    "args": {
      "External id": 251180,
      "cbid": 211, "correlation": 251180
    }
  },
  {
    "ph": "s", "id": 251180, "pid": 76337, "tid": -914061504, "ts": 1716454225389851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225445397, "dur": 5,
    "args": {
      "External id": 251189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251189,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251189, "pid": 5, "tid": 7, "ts": 1716454225445397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389892, "dur": 9,
    "args": {
      "External id": 251189,
      "cbid": 211, "correlation": 251189
    }
  },
  {
    "ph": "s", "id": 251189, "pid": 76337, "tid": -914061504, "ts": 1716454225389892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389953, "dur": 0,
    "args": {
      "External id": 251199,
      "cbid": 317, "correlation": 251199
    }
  },
  {
    "ph": "f", "id": 251199, "pid": 76337, "tid": -914061504, "ts": 1716454225389953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389953, "dur": 0,
    "args": {
      "External id": 251200,
      "cbid": 203, "correlation": 251200
    }
  },
  {
    "ph": "f", "id": 251200, "pid": 76337, "tid": -914061504, "ts": 1716454225389953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389954, "dur": 0,
    "args": {
      "External id": 251201,
      "cbid": 205, "correlation": 251201
    }
  },
  {
    "ph": "f", "id": 251201, "pid": 76337, "tid": -914061504, "ts": 1716454225389954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445403, "dur": 5,
    "args": {
      "External id": 251205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251205,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251205, "pid": 5, "tid": 7, "ts": 1716454225445403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389968, "dur": 21,
    "args": {
      "External id": 251205,
      "cbid": 211, "correlation": 251205
    }
  },
  {
    "ph": "s", "id": 251205, "pid": 76337, "tid": -914061504, "ts": 1716454225389968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445409, "dur": 166,
    "args": {
      "External id": 251207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251207,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251207, "pid": 5, "tid": 7, "ts": 1716454225445409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225389991, "dur": 5,
    "args": {
      "External id": 251207,
      "cbid": 211, "correlation": 251207
    }
  },
  {
    "ph": "s", "id": 251207, "pid": 76337, "tid": -914061504, "ts": 1716454225389991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225445578, "dur": 1,
    "args": {
      "External id": 251209,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251209,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 251209, "pid": 5, "tid": 7, "ts": 1716454225445578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390002, "dur": 6,
    "args": {
      "External id": 251209,
      "cbid": 51, "correlation": 251209
    }
  },
  {
    "ph": "s", "id": 251209, "pid": 76337, "tid": -914061504, "ts": 1716454225390002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225445581, "dur": 264,
    "args": {
      "External id": 251210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251210,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251210, "pid": 5, "tid": 7, "ts": 1716454225445581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390009, "dur": 6,
    "args": {
      "External id": 251210,
      "cbid": 211, "correlation": 251210
    }
  },
  {
    "ph": "s", "id": 251210, "pid": 76337, "tid": -914061504, "ts": 1716454225390009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445847, "dur": 6,
    "args": {
      "External id": 251212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251212,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251212, "pid": 5, "tid": 7, "ts": 1716454225445847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390019, "dur": 5,
    "args": {
      "External id": 251212,
      "cbid": 211, "correlation": 251212
    }
  },
  {
    "ph": "s", "id": 251212, "pid": 76337, "tid": -914061504, "ts": 1716454225390019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225445854, "dur": 6,
    "args": {
      "External id": 251218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251218,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251218, "pid": 5, "tid": 7, "ts": 1716454225445854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390048, "dur": 9,
    "args": {
      "External id": 251218,
      "cbid": 211, "correlation": 251218
    }
  },
  {
    "ph": "s", "id": 251218, "pid": 76337, "tid": -914061504, "ts": 1716454225390048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225445862, "dur": 5,
    "args": {
      "External id": 251226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251226,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251226, "pid": 5, "tid": 7, "ts": 1716454225445862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390081, "dur": 9,
    "args": {
      "External id": 251226,
      "cbid": 211, "correlation": 251226
    }
  },
  {
    "ph": "s", "id": 251226, "pid": 76337, "tid": -914061504, "ts": 1716454225390081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225445867, "dur": 5,
    "args": {
      "External id": 251234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251234,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251234, "pid": 5, "tid": 7, "ts": 1716454225445867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390111, "dur": 8,
    "args": {
      "External id": 251234,
      "cbid": 211, "correlation": 251234
    }
  },
  {
    "ph": "s", "id": 251234, "pid": 76337, "tid": -914061504, "ts": 1716454225390111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225445873, "dur": 12,
    "args": {
      "External id": 251243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251243,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251243, "pid": 5, "tid": 7, "ts": 1716454225445873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390199, "dur": 13,
    "args": {
      "External id": 251243,
      "cbid": 211, "correlation": 251243
    }
  },
  {
    "ph": "s", "id": 251243, "pid": 76337, "tid": -914061504, "ts": 1716454225390199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445886, "dur": 12,
    "args": {
      "External id": 251263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251263,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251263, "pid": 5, "tid": 7, "ts": 1716454225445886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390265, "dur": 11,
    "args": {
      "External id": 251263,
      "cbid": 211, "correlation": 251263
    }
  },
  {
    "ph": "s", "id": 251263, "pid": 76337, "tid": -914061504, "ts": 1716454225390265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225445900, "dur": 4,
    "args": {
      "External id": 251275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251275,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251275, "pid": 5, "tid": 7, "ts": 1716454225445900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390290, "dur": 7,
    "args": {
      "External id": 251275,
      "cbid": 211, "correlation": 251275
    }
  },
  {
    "ph": "s", "id": 251275, "pid": 76337, "tid": -914061504, "ts": 1716454225390290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225445906, "dur": 10,
    "args": {
      "External id": 251278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251278, "pid": 5, "tid": 7, "ts": 1716454225445906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390309, "dur": 6,
    "args": {
      "External id": 251278,
      "cbid": 211, "correlation": 251278
    }
  },
  {
    "ph": "s", "id": 251278, "pid": 76337, "tid": -914061504, "ts": 1716454225390309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225445917, "dur": 6,
    "args": {
      "External id": 251287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251287,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251287, "pid": 5, "tid": 7, "ts": 1716454225445917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390349, "dur": 10,
    "args": {
      "External id": 251287,
      "cbid": 211, "correlation": 251287
    }
  },
  {
    "ph": "s", "id": 251287, "pid": 76337, "tid": -914061504, "ts": 1716454225390349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390405, "dur": 0,
    "args": {
      "External id": 251297,
      "cbid": 317, "correlation": 251297
    }
  },
  {
    "ph": "f", "id": 251297, "pid": 76337, "tid": -914061504, "ts": 1716454225390405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390406, "dur": 0,
    "args": {
      "External id": 251298,
      "cbid": 203, "correlation": 251298
    }
  },
  {
    "ph": "f", "id": 251298, "pid": 76337, "tid": -914061504, "ts": 1716454225390406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390407, "dur": 0,
    "args": {
      "External id": 251299,
      "cbid": 205, "correlation": 251299
    }
  },
  {
    "ph": "f", "id": 251299, "pid": 76337, "tid": -914061504, "ts": 1716454225390407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445925, "dur": 7,
    "args": {
      "External id": 251303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251303, "pid": 5, "tid": 7, "ts": 1716454225445925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390422, "dur": 12,
    "args": {
      "External id": 251303,
      "cbid": 211, "correlation": 251303
    }
  },
  {
    "ph": "s", "id": 251303, "pid": 76337, "tid": -914061504, "ts": 1716454225390422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225445933, "dur": 325,
    "args": {
      "External id": 251305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251305, "pid": 5, "tid": 7, "ts": 1716454225445933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390437, "dur": 6,
    "args": {
      "External id": 251305,
      "cbid": 211, "correlation": 251305
    }
  },
  {
    "ph": "s", "id": 251305, "pid": 76337, "tid": -914061504, "ts": 1716454225390437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225446260, "dur": 1,
    "args": {
      "External id": 251307,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251307,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1561483409238777
    }
  },
  {
    "ph": "f", "id": 251307, "pid": 5, "tid": 7, "ts": 1716454225446260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390449, "dur": 7,
    "args": {
      "External id": 251307,
      "cbid": 51, "correlation": 251307
    }
  },
  {
    "ph": "s", "id": 251307, "pid": 76337, "tid": -914061504, "ts": 1716454225390449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225446264, "dur": 503,
    "args": {
      "External id": 251308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251308,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251308, "pid": 5, "tid": 7, "ts": 1716454225446264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390457, "dur": 6,
    "args": {
      "External id": 251308,
      "cbid": 211, "correlation": 251308
    }
  },
  {
    "ph": "s", "id": 251308, "pid": 76337, "tid": -914061504, "ts": 1716454225390457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225446768, "dur": 5,
    "args": {
      "External id": 251310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251310, "pid": 5, "tid": 7, "ts": 1716454225446768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390468, "dur": 5,
    "args": {
      "External id": 251310,
      "cbid": 211, "correlation": 251310
    }
  },
  {
    "ph": "s", "id": 251310, "pid": 76337, "tid": -914061504, "ts": 1716454225390468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225446775, "dur": 6,
    "args": {
      "External id": 251316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251316, "pid": 5, "tid": 7, "ts": 1716454225446775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390498, "dur": 8,
    "args": {
      "External id": 251316,
      "cbid": 211, "correlation": 251316
    }
  },
  {
    "ph": "s", "id": 251316, "pid": 76337, "tid": -914061504, "ts": 1716454225390498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225446783, "dur": 3,
    "args": {
      "External id": 251324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251324,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 251324, "pid": 5, "tid": 7, "ts": 1716454225446783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390541, "dur": 9,
    "args": {
      "External id": 251324,
      "cbid": 211, "correlation": 251324
    }
  },
  {
    "ph": "s", "id": 251324, "pid": 76337, "tid": -914061504, "ts": 1716454225390541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390604, "dur": 1,
    "args": {
      "External id": 251340,
      "cbid": 251, "correlation": 251340
    }
  },
  {
    "ph": "f", "id": 251340, "pid": 76337, "tid": -914061504, "ts": 1716454225390604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390609, "dur": 0,
    "args": {
      "External id": 251342,
      "cbid": 251, "correlation": 251342
    }
  },
  {
    "ph": "f", "id": 251342, "pid": 76337, "tid": -914061504, "ts": 1716454225390609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225446787, "dur": 11,
    "args": {
      "External id": 251343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251343,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251343, "pid": 5, "tid": 7, "ts": 1716454225446787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390612, "dur": 11,
    "args": {
      "External id": 251343,
      "cbid": 211, "correlation": 251343
    }
  },
  {
    "ph": "s", "id": 251343, "pid": 76337, "tid": -914061504, "ts": 1716454225390612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225446800, "dur": 5,
    "args": {
      "External id": 251345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251345,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251345, "pid": 5, "tid": 7, "ts": 1716454225446800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390624, "dur": 5,
    "args": {
      "External id": 251345,
      "cbid": 211, "correlation": 251345
    }
  },
  {
    "ph": "s", "id": 251345, "pid": 76337, "tid": -914061504, "ts": 1716454225390624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225446806, "dur": 6,
    "args": {
      "External id": 251355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251355, "pid": 5, "tid": 7, "ts": 1716454225446806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390680, "dur": 13,
    "args": {
      "External id": 251355,
      "cbid": 211, "correlation": 251355
    }
  },
  {
    "ph": "s", "id": 251355, "pid": 76337, "tid": -914061504, "ts": 1716454225390680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225446813, "dur": 10,
    "args": {
      "External id": 251375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251375,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251375, "pid": 5, "tid": 7, "ts": 1716454225446813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390748, "dur": 10,
    "args": {
      "External id": 251375,
      "cbid": 211, "correlation": 251375
    }
  },
  {
    "ph": "s", "id": 251375, "pid": 76337, "tid": -914061504, "ts": 1716454225390748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225446824, "dur": 4,
    "args": {
      "External id": 251387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251387,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 251387, "pid": 5, "tid": 7, "ts": 1716454225446824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390769, "dur": 6,
    "args": {
      "External id": 251387,
      "cbid": 211, "correlation": 251387
    }
  },
  {
    "ph": "s", "id": 251387, "pid": 76337, "tid": -914061504, "ts": 1716454225390769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225446829, "dur": 7,
    "args": {
      "External id": 251390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251390,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251390, "pid": 5, "tid": 7, "ts": 1716454225446829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390787, "dur": 7,
    "args": {
      "External id": 251390,
      "cbid": 211, "correlation": 251390
    }
  },
  {
    "ph": "s", "id": 251390, "pid": 76337, "tid": -914061504, "ts": 1716454225390787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225446837, "dur": 5,
    "args": {
      "External id": 251399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251399,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251399, "pid": 5, "tid": 7, "ts": 1716454225446837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390827, "dur": 10,
    "args": {
      "External id": 251399,
      "cbid": 211, "correlation": 251399
    }
  },
  {
    "ph": "s", "id": 251399, "pid": 76337, "tid": -914061504, "ts": 1716454225390827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390890, "dur": 0,
    "args": {
      "External id": 251409,
      "cbid": 317, "correlation": 251409
    }
  },
  {
    "ph": "f", "id": 251409, "pid": 76337, "tid": -914061504, "ts": 1716454225390890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390891, "dur": 0,
    "args": {
      "External id": 251410,
      "cbid": 203, "correlation": 251410
    }
  },
  {
    "ph": "f", "id": 251410, "pid": 76337, "tid": -914061504, "ts": 1716454225390891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390892, "dur": 0,
    "args": {
      "External id": 251411,
      "cbid": 205, "correlation": 251411
    }
  },
  {
    "ph": "f", "id": 251411, "pid": 76337, "tid": -914061504, "ts": 1716454225390892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225446843, "dur": 5,
    "args": {
      "External id": 251415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251415,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251415, "pid": 5, "tid": 7, "ts": 1716454225446843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390907, "dur": 12,
    "args": {
      "External id": 251415,
      "cbid": 211, "correlation": 251415
    }
  },
  {
    "ph": "s", "id": 251415, "pid": 76337, "tid": -914061504, "ts": 1716454225390907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225446850, "dur": 166,
    "args": {
      "External id": 251417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251417,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251417, "pid": 5, "tid": 7, "ts": 1716454225446850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390921, "dur": 5,
    "args": {
      "External id": 251417,
      "cbid": 211, "correlation": 251417
    }
  },
  {
    "ph": "s", "id": 251417, "pid": 76337, "tid": -914061504, "ts": 1716454225390921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225447018, "dur": 1,
    "args": {
      "External id": 251419,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251419,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 251419, "pid": 5, "tid": 7, "ts": 1716454225447018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390932, "dur": 6,
    "args": {
      "External id": 251419,
      "cbid": 51, "correlation": 251419
    }
  },
  {
    "ph": "s", "id": 251419, "pid": 76337, "tid": -914061504, "ts": 1716454225390932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225447022, "dur": 263,
    "args": {
      "External id": 251420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251420,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251420, "pid": 5, "tid": 7, "ts": 1716454225447022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390939, "dur": 6,
    "args": {
      "External id": 251420,
      "cbid": 211, "correlation": 251420
    }
  },
  {
    "ph": "s", "id": 251420, "pid": 76337, "tid": -914061504, "ts": 1716454225390939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225447286, "dur": 6,
    "args": {
      "External id": 251422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251422, "pid": 5, "tid": 7, "ts": 1716454225447286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390949, "dur": 5,
    "args": {
      "External id": 251422,
      "cbid": 211, "correlation": 251422
    }
  },
  {
    "ph": "s", "id": 251422, "pid": 76337, "tid": -914061504, "ts": 1716454225390949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225447293, "dur": 6,
    "args": {
      "External id": 251428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251428, "pid": 5, "tid": 7, "ts": 1716454225447293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225390985, "dur": 9,
    "args": {
      "External id": 251428,
      "cbid": 211, "correlation": 251428
    }
  },
  {
    "ph": "s", "id": 251428, "pid": 76337, "tid": -914061504, "ts": 1716454225390985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391044, "dur": 0,
    "args": {
      "External id": 251438,
      "cbid": 317, "correlation": 251438
    }
  },
  {
    "ph": "f", "id": 251438, "pid": 76337, "tid": -914061504, "ts": 1716454225391044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391045, "dur": 0,
    "args": {
      "External id": 251439,
      "cbid": 203, "correlation": 251439
    }
  },
  {
    "ph": "f", "id": 251439, "pid": 76337, "tid": -914061504, "ts": 1716454225391045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391046, "dur": 0,
    "args": {
      "External id": 251440,
      "cbid": 205, "correlation": 251440
    }
  },
  {
    "ph": "f", "id": 251440, "pid": 76337, "tid": -914061504, "ts": 1716454225391046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225447300, "dur": 8,
    "args": {
      "External id": 251444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251444, "pid": 5, "tid": 7, "ts": 1716454225447300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391061, "dur": 12,
    "args": {
      "External id": 251444,
      "cbid": 211, "correlation": 251444
    }
  },
  {
    "ph": "s", "id": 251444, "pid": 76337, "tid": -914061504, "ts": 1716454225391061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225447309, "dur": 3,
    "args": {
      "External id": 251446,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251446,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 251446, "pid": 5, "tid": 7, "ts": 1716454225447309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391079, "dur": 15,
    "args": {
      "External id": 251446,
      "cbid": 51, "correlation": 251446
    }
  },
  {
    "ph": "s", "id": 251446, "pid": 76337, "tid": -914061504, "ts": 1716454225391079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225447314, "dur": 97,
    "args": {
      "External id": 251447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251447,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 251447, "pid": 5, "tid": 7, "ts": 1716454225447314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391095, "dur": 7,
    "args": {
      "External id": 251447,
      "cbid": 211, "correlation": 251447
    }
  },
  {
    "ph": "s", "id": 251447, "pid": 76337, "tid": -914061504, "ts": 1716454225391095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225447412, "dur": 6,
    "args": {
      "External id": 251449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251449, "pid": 5, "tid": 7, "ts": 1716454225447412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391106, "dur": 5,
    "args": {
      "External id": 251449,
      "cbid": 211, "correlation": 251449
    }
  },
  {
    "ph": "s", "id": 251449, "pid": 76337, "tid": -914061504, "ts": 1716454225391106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225447419, "dur": 6,
    "args": {
      "External id": 251455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251455, "pid": 5, "tid": 7, "ts": 1716454225447419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391133, "dur": 8,
    "args": {
      "External id": 251455,
      "cbid": 211, "correlation": 251455
    }
  },
  {
    "ph": "s", "id": 251455, "pid": 76337, "tid": -914061504, "ts": 1716454225391133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225447427, "dur": 5,
    "args": {
      "External id": 251463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251463,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251463, "pid": 5, "tid": 7, "ts": 1716454225447427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391162, "dur": 8,
    "args": {
      "External id": 251463,
      "cbid": 211, "correlation": 251463
    }
  },
  {
    "ph": "s", "id": 251463, "pid": 76337, "tid": -914061504, "ts": 1716454225391162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225447433, "dur": 4,
    "args": {
      "External id": 251471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251471,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251471, "pid": 5, "tid": 7, "ts": 1716454225447433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391190, "dur": 9,
    "args": {
      "External id": 251471,
      "cbid": 211, "correlation": 251471
    }
  },
  {
    "ph": "s", "id": 251471, "pid": 76337, "tid": -914061504, "ts": 1716454225391190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225447438, "dur": 11,
    "args": {
      "External id": 251480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251480,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251480, "pid": 5, "tid": 7, "ts": 1716454225447438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391235, "dur": 10,
    "args": {
      "External id": 251480,
      "cbid": 211, "correlation": 251480
    }
  },
  {
    "ph": "s", "id": 251480, "pid": 76337, "tid": -914061504, "ts": 1716454225391235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225447451, "dur": 12,
    "args": {
      "External id": 251500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251500,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251500, "pid": 5, "tid": 7, "ts": 1716454225447451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391305, "dur": 11,
    "args": {
      "External id": 251500,
      "cbid": 211, "correlation": 251500
    }
  },
  {
    "ph": "s", "id": 251500, "pid": 76337, "tid": -914061504, "ts": 1716454225391305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225447465, "dur": 4,
    "args": {
      "External id": 251512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251512,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251512, "pid": 5, "tid": 7, "ts": 1716454225447465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391329, "dur": 6,
    "args": {
      "External id": 251512,
      "cbid": 211, "correlation": 251512
    }
  },
  {
    "ph": "s", "id": 251512, "pid": 76337, "tid": -914061504, "ts": 1716454225391329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225447470, "dur": 11,
    "args": {
      "External id": 251515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251515,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251515, "pid": 5, "tid": 7, "ts": 1716454225447470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391347, "dur": 8,
    "args": {
      "External id": 251515,
      "cbid": 211, "correlation": 251515
    }
  },
  {
    "ph": "s", "id": 251515, "pid": 76337, "tid": -914061504, "ts": 1716454225391347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225447482, "dur": 7,
    "args": {
      "External id": 251524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251524,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251524, "pid": 5, "tid": 7, "ts": 1716454225447482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391387, "dur": 10,
    "args": {
      "External id": 251524,
      "cbid": 211, "correlation": 251524
    }
  },
  {
    "ph": "s", "id": 251524, "pid": 76337, "tid": -914061504, "ts": 1716454225391387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391440, "dur": 0,
    "args": {
      "External id": 251534,
      "cbid": 317, "correlation": 251534
    }
  },
  {
    "ph": "f", "id": 251534, "pid": 76337, "tid": -914061504, "ts": 1716454225391440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391441, "dur": 0,
    "args": {
      "External id": 251535,
      "cbid": 203, "correlation": 251535
    }
  },
  {
    "ph": "f", "id": 251535, "pid": 76337, "tid": -914061504, "ts": 1716454225391441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391441, "dur": 0,
    "args": {
      "External id": 251536,
      "cbid": 205, "correlation": 251536
    }
  },
  {
    "ph": "f", "id": 251536, "pid": 76337, "tid": -914061504, "ts": 1716454225391441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225447491, "dur": 7,
    "args": {
      "External id": 251540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251540,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251540, "pid": 5, "tid": 7, "ts": 1716454225447491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391456, "dur": 12,
    "args": {
      "External id": 251540,
      "cbid": 211, "correlation": 251540
    }
  },
  {
    "ph": "s", "id": 251540, "pid": 76337, "tid": -914061504, "ts": 1716454225391456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225447499, "dur": 326,
    "args": {
      "External id": 251542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251542, "pid": 5, "tid": 7, "ts": 1716454225447499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391471, "dur": 5,
    "args": {
      "External id": 251542,
      "cbid": 211, "correlation": 251542
    }
  },
  {
    "ph": "s", "id": 251542, "pid": 76337, "tid": -914061504, "ts": 1716454225391471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225447826, "dur": 1,
    "args": {
      "External id": 251544,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251544,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 251544, "pid": 5, "tid": 7, "ts": 1716454225447826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391481, "dur": 6,
    "args": {
      "External id": 251544,
      "cbid": 51, "correlation": 251544
    }
  },
  {
    "ph": "s", "id": 251544, "pid": 76337, "tid": -914061504, "ts": 1716454225391481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225447830, "dur": 505,
    "args": {
      "External id": 251545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251545,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251545, "pid": 5, "tid": 7, "ts": 1716454225447830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391488, "dur": 6,
    "args": {
      "External id": 251545,
      "cbid": 211, "correlation": 251545
    }
  },
  {
    "ph": "s", "id": 251545, "pid": 76337, "tid": -914061504, "ts": 1716454225391488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448337, "dur": 6,
    "args": {
      "External id": 251547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251547,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251547, "pid": 5, "tid": 7, "ts": 1716454225448337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391498, "dur": 6,
    "args": {
      "External id": 251547,
      "cbid": 211, "correlation": 251547
    }
  },
  {
    "ph": "s", "id": 251547, "pid": 76337, "tid": -914061504, "ts": 1716454225391498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225448344, "dur": 6,
    "args": {
      "External id": 251553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251553, "pid": 5, "tid": 7, "ts": 1716454225448344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391527, "dur": 9,
    "args": {
      "External id": 251553,
      "cbid": 211, "correlation": 251553
    }
  },
  {
    "ph": "s", "id": 251553, "pid": 76337, "tid": -914061504, "ts": 1716454225391527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225448352, "dur": 3,
    "args": {
      "External id": 251561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251561,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 251561, "pid": 5, "tid": 7, "ts": 1716454225448352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391571, "dur": 10,
    "args": {
      "External id": 251561,
      "cbid": 211, "correlation": 251561
    }
  },
  {
    "ph": "s", "id": 251561, "pid": 76337, "tid": -914061504, "ts": 1716454225391571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391635, "dur": 1,
    "args": {
      "External id": 251577,
      "cbid": 251, "correlation": 251577
    }
  },
  {
    "ph": "f", "id": 251577, "pid": 76337, "tid": -914061504, "ts": 1716454225391635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391641, "dur": 0,
    "args": {
      "External id": 251579,
      "cbid": 251, "correlation": 251579
    }
  },
  {
    "ph": "f", "id": 251579, "pid": 76337, "tid": -914061504, "ts": 1716454225391641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225448356, "dur": 13,
    "args": {
      "External id": 251580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251580,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251580, "pid": 5, "tid": 7, "ts": 1716454225448356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391643, "dur": 12,
    "args": {
      "External id": 251580,
      "cbid": 211, "correlation": 251580
    }
  },
  {
    "ph": "s", "id": 251580, "pid": 76337, "tid": -914061504, "ts": 1716454225391643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225448371, "dur": 5,
    "args": {
      "External id": 251582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251582,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251582, "pid": 5, "tid": 7, "ts": 1716454225448371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391657, "dur": 5,
    "args": {
      "External id": 251582,
      "cbid": 211, "correlation": 251582
    }
  },
  {
    "ph": "s", "id": 251582, "pid": 76337, "tid": -914061504, "ts": 1716454225391657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225448377, "dur": 6,
    "args": {
      "External id": 251592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251592, "pid": 5, "tid": 7, "ts": 1716454225448377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391714, "dur": 13,
    "args": {
      "External id": 251592,
      "cbid": 211, "correlation": 251592
    }
  },
  {
    "ph": "s", "id": 251592, "pid": 76337, "tid": -914061504, "ts": 1716454225391714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448384, "dur": 10,
    "args": {
      "External id": 251612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251612,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251612, "pid": 5, "tid": 7, "ts": 1716454225448384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391780, "dur": 11,
    "args": {
      "External id": 251612,
      "cbid": 211, "correlation": 251612
    }
  },
  {
    "ph": "s", "id": 251612, "pid": 76337, "tid": -914061504, "ts": 1716454225391780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225448395, "dur": 4,
    "args": {
      "External id": 251624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251624,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 251624, "pid": 5, "tid": 7, "ts": 1716454225448395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391801, "dur": 6,
    "args": {
      "External id": 251624,
      "cbid": 211, "correlation": 251624
    }
  },
  {
    "ph": "s", "id": 251624, "pid": 76337, "tid": -914061504, "ts": 1716454225391801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225448400, "dur": 7,
    "args": {
      "External id": 251627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251627,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251627, "pid": 5, "tid": 7, "ts": 1716454225448400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391820, "dur": 6,
    "args": {
      "External id": 251627,
      "cbid": 211, "correlation": 251627
    }
  },
  {
    "ph": "s", "id": 251627, "pid": 76337, "tid": -914061504, "ts": 1716454225391820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225448408, "dur": 5,
    "args": {
      "External id": 251636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251636,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251636, "pid": 5, "tid": 7, "ts": 1716454225448408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391861, "dur": 9,
    "args": {
      "External id": 251636,
      "cbid": 211, "correlation": 251636
    }
  },
  {
    "ph": "s", "id": 251636, "pid": 76337, "tid": -914061504, "ts": 1716454225391861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391923, "dur": 0,
    "args": {
      "External id": 251646,
      "cbid": 317, "correlation": 251646
    }
  },
  {
    "ph": "f", "id": 251646, "pid": 76337, "tid": -914061504, "ts": 1716454225391923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391924, "dur": 0,
    "args": {
      "External id": 251647,
      "cbid": 203, "correlation": 251647
    }
  },
  {
    "ph": "f", "id": 251647, "pid": 76337, "tid": -914061504, "ts": 1716454225391924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391925, "dur": 0,
    "args": {
      "External id": 251648,
      "cbid": 205, "correlation": 251648
    }
  },
  {
    "ph": "f", "id": 251648, "pid": 76337, "tid": -914061504, "ts": 1716454225391925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448415, "dur": 5,
    "args": {
      "External id": 251652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251652,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251652, "pid": 5, "tid": 7, "ts": 1716454225448415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391938, "dur": 12,
    "args": {
      "External id": 251652,
      "cbid": 211, "correlation": 251652
    }
  },
  {
    "ph": "s", "id": 251652, "pid": 76337, "tid": -914061504, "ts": 1716454225391938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448421, "dur": 165,
    "args": {
      "External id": 251654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251654,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251654, "pid": 5, "tid": 7, "ts": 1716454225448421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391953, "dur": 6,
    "args": {
      "External id": 251654,
      "cbid": 211, "correlation": 251654
    }
  },
  {
    "ph": "s", "id": 251654, "pid": 76337, "tid": -914061504, "ts": 1716454225391953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225448588, "dur": 1,
    "args": {
      "External id": 251656,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251656,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 251656, "pid": 5, "tid": 7, "ts": 1716454225448588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391964, "dur": 7,
    "args": {
      "External id": 251656,
      "cbid": 51, "correlation": 251656
    }
  },
  {
    "ph": "s", "id": 251656, "pid": 76337, "tid": -914061504, "ts": 1716454225391964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225448592, "dur": 264,
    "args": {
      "External id": 251657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251657,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251657, "pid": 5, "tid": 7, "ts": 1716454225448592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391972, "dur": 14,
    "args": {
      "External id": 251657,
      "cbid": 211, "correlation": 251657
    }
  },
  {
    "ph": "s", "id": 251657, "pid": 76337, "tid": -914061504, "ts": 1716454225391972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448857, "dur": 6,
    "args": {
      "External id": 251659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251659, "pid": 5, "tid": 7, "ts": 1716454225448857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225391989, "dur": 5,
    "args": {
      "External id": 251659,
      "cbid": 211, "correlation": 251659
    }
  },
  {
    "ph": "s", "id": 251659, "pid": 76337, "tid": -914061504, "ts": 1716454225391989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225448864, "dur": 6,
    "args": {
      "External id": 251665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251665, "pid": 5, "tid": 7, "ts": 1716454225448864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392019, "dur": 8,
    "args": {
      "External id": 251665,
      "cbid": 211, "correlation": 251665
    }
  },
  {
    "ph": "s", "id": 251665, "pid": 76337, "tid": -914061504, "ts": 1716454225392019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392079, "dur": 0,
    "args": {
      "External id": 251675,
      "cbid": 317, "correlation": 251675
    }
  },
  {
    "ph": "f", "id": 251675, "pid": 76337, "tid": -914061504, "ts": 1716454225392079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392079, "dur": 0,
    "args": {
      "External id": 251676,
      "cbid": 203, "correlation": 251676
    }
  },
  {
    "ph": "f", "id": 251676, "pid": 76337, "tid": -914061504, "ts": 1716454225392079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392080, "dur": 0,
    "args": {
      "External id": 251677,
      "cbid": 205, "correlation": 251677
    }
  },
  {
    "ph": "f", "id": 251677, "pid": 76337, "tid": -914061504, "ts": 1716454225392080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448871, "dur": 8,
    "args": {
      "External id": 251681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251681,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251681, "pid": 5, "tid": 7, "ts": 1716454225448871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392092, "dur": 12,
    "args": {
      "External id": 251681,
      "cbid": 211, "correlation": 251681
    }
  },
  {
    "ph": "s", "id": 251681, "pid": 76337, "tid": -914061504, "ts": 1716454225392092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225448880, "dur": 3,
    "args": {
      "External id": 251683,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251683,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 251683, "pid": 5, "tid": 7, "ts": 1716454225448880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392109, "dur": 9,
    "args": {
      "External id": 251683,
      "cbid": 51, "correlation": 251683
    }
  },
  {
    "ph": "s", "id": 251683, "pid": 76337, "tid": -914061504, "ts": 1716454225392109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225448884, "dur": 98,
    "args": {
      "External id": 251684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251684,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 251684, "pid": 5, "tid": 7, "ts": 1716454225448884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392119, "dur": 6,
    "args": {
      "External id": 251684,
      "cbid": 211, "correlation": 251684
    }
  },
  {
    "ph": "s", "id": 251684, "pid": 76337, "tid": -914061504, "ts": 1716454225392119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225448984, "dur": 6,
    "args": {
      "External id": 251686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251686,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251686, "pid": 5, "tid": 7, "ts": 1716454225448984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392129, "dur": 5,
    "args": {
      "External id": 251686,
      "cbid": 211, "correlation": 251686
    }
  },
  {
    "ph": "s", "id": 251686, "pid": 76337, "tid": -914061504, "ts": 1716454225392129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225448991, "dur": 6,
    "args": {
      "External id": 251692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251692, "pid": 5, "tid": 7, "ts": 1716454225448991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392156, "dur": 9,
    "args": {
      "External id": 251692,
      "cbid": 211, "correlation": 251692
    }
  },
  {
    "ph": "s", "id": 251692, "pid": 76337, "tid": -914061504, "ts": 1716454225392156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225448998, "dur": 5,
    "args": {
      "External id": 251700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251700,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251700, "pid": 5, "tid": 7, "ts": 1716454225448998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392186, "dur": 8,
    "args": {
      "External id": 251700,
      "cbid": 211, "correlation": 251700
    }
  },
  {
    "ph": "s", "id": 251700, "pid": 76337, "tid": -914061504, "ts": 1716454225392186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225449004, "dur": 5,
    "args": {
      "External id": 251708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251708,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251708, "pid": 5, "tid": 7, "ts": 1716454225449004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392215, "dur": 8,
    "args": {
      "External id": 251708,
      "cbid": 211, "correlation": 251708
    }
  },
  {
    "ph": "s", "id": 251708, "pid": 76337, "tid": -914061504, "ts": 1716454225392215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225449010, "dur": 11,
    "args": {
      "External id": 251717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251717,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251717, "pid": 5, "tid": 7, "ts": 1716454225449010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392259, "dur": 10,
    "args": {
      "External id": 251717,
      "cbid": 211, "correlation": 251717
    }
  },
  {
    "ph": "s", "id": 251717, "pid": 76337, "tid": -914061504, "ts": 1716454225392259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449023, "dur": 12,
    "args": {
      "External id": 251737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251737,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251737, "pid": 5, "tid": 7, "ts": 1716454225449023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392332, "dur": 12,
    "args": {
      "External id": 251737,
      "cbid": 211, "correlation": 251737
    }
  },
  {
    "ph": "s", "id": 251737, "pid": 76337, "tid": -914061504, "ts": 1716454225392332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225449037, "dur": 4,
    "args": {
      "External id": 251749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251749,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251749, "pid": 5, "tid": 7, "ts": 1716454225449037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392354, "dur": 7,
    "args": {
      "External id": 251749,
      "cbid": 211, "correlation": 251749
    }
  },
  {
    "ph": "s", "id": 251749, "pid": 76337, "tid": -914061504, "ts": 1716454225392354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225449042, "dur": 11,
    "args": {
      "External id": 251752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251752, "pid": 5, "tid": 7, "ts": 1716454225449042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392372, "dur": 6,
    "args": {
      "External id": 251752,
      "cbid": 211, "correlation": 251752
    }
  },
  {
    "ph": "s", "id": 251752, "pid": 76337, "tid": -914061504, "ts": 1716454225392372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225449054, "dur": 7,
    "args": {
      "External id": 251761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251761,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251761, "pid": 5, "tid": 7, "ts": 1716454225449054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392411, "dur": 10,
    "args": {
      "External id": 251761,
      "cbid": 211, "correlation": 251761
    }
  },
  {
    "ph": "s", "id": 251761, "pid": 76337, "tid": -914061504, "ts": 1716454225392411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392463, "dur": 0,
    "args": {
      "External id": 251771,
      "cbid": 317, "correlation": 251771
    }
  },
  {
    "ph": "f", "id": 251771, "pid": 76337, "tid": -914061504, "ts": 1716454225392463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392464, "dur": 0,
    "args": {
      "External id": 251772,
      "cbid": 203, "correlation": 251772
    }
  },
  {
    "ph": "f", "id": 251772, "pid": 76337, "tid": -914061504, "ts": 1716454225392464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392465, "dur": 0,
    "args": {
      "External id": 251773,
      "cbid": 205, "correlation": 251773
    }
  },
  {
    "ph": "f", "id": 251773, "pid": 76337, "tid": -914061504, "ts": 1716454225392465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449062, "dur": 7,
    "args": {
      "External id": 251777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251777,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251777, "pid": 5, "tid": 7, "ts": 1716454225449062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392478, "dur": 11,
    "args": {
      "External id": 251777,
      "cbid": 211, "correlation": 251777
    }
  },
  {
    "ph": "s", "id": 251777, "pid": 76337, "tid": -914061504, "ts": 1716454225392478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449070, "dur": 325,
    "args": {
      "External id": 251779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251779,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251779, "pid": 5, "tid": 7, "ts": 1716454225449070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392492, "dur": 5,
    "args": {
      "External id": 251779,
      "cbid": 211, "correlation": 251779
    }
  },
  {
    "ph": "s", "id": 251779, "pid": 76337, "tid": -914061504, "ts": 1716454225392492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225449398, "dur": 1,
    "args": {
      "External id": 251781,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251781,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 251781, "pid": 5, "tid": 7, "ts": 1716454225449398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392503, "dur": 7,
    "args": {
      "External id": 251781,
      "cbid": 51, "correlation": 251781
    }
  },
  {
    "ph": "s", "id": 251781, "pid": 76337, "tid": -914061504, "ts": 1716454225392503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225449401, "dur": 505,
    "args": {
      "External id": 251782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251782,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251782, "pid": 5, "tid": 7, "ts": 1716454225449401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392511, "dur": 6,
    "args": {
      "External id": 251782,
      "cbid": 211, "correlation": 251782
    }
  },
  {
    "ph": "s", "id": 251782, "pid": 76337, "tid": -914061504, "ts": 1716454225392511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449908, "dur": 6,
    "args": {
      "External id": 251784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251784,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251784, "pid": 5, "tid": 7, "ts": 1716454225449908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392520, "dur": 5,
    "args": {
      "External id": 251784,
      "cbid": 211, "correlation": 251784
    }
  },
  {
    "ph": "s", "id": 251784, "pid": 76337, "tid": -914061504, "ts": 1716454225392520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225449915, "dur": 6,
    "args": {
      "External id": 251790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251790, "pid": 5, "tid": 7, "ts": 1716454225449915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392548, "dur": 10,
    "args": {
      "External id": 251790,
      "cbid": 211, "correlation": 251790
    }
  },
  {
    "ph": "s", "id": 251790, "pid": 76337, "tid": -914061504, "ts": 1716454225392548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225449923, "dur": 3,
    "args": {
      "External id": 251798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251798,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 251798, "pid": 5, "tid": 7, "ts": 1716454225449923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392592, "dur": 9,
    "args": {
      "External id": 251798,
      "cbid": 211, "correlation": 251798
    }
  },
  {
    "ph": "s", "id": 251798, "pid": 76337, "tid": -914061504, "ts": 1716454225392592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392655, "dur": 1,
    "args": {
      "External id": 251814,
      "cbid": 251, "correlation": 251814
    }
  },
  {
    "ph": "f", "id": 251814, "pid": 76337, "tid": -914061504, "ts": 1716454225392655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392661, "dur": 0,
    "args": {
      "External id": 251816,
      "cbid": 251, "correlation": 251816
    }
  },
  {
    "ph": "f", "id": 251816, "pid": 76337, "tid": -914061504, "ts": 1716454225392661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225449927, "dur": 13,
    "args": {
      "External id": 251817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251817,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251817, "pid": 5, "tid": 7, "ts": 1716454225449927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392663, "dur": 11,
    "args": {
      "External id": 251817,
      "cbid": 211, "correlation": 251817
    }
  },
  {
    "ph": "s", "id": 251817, "pid": 76337, "tid": -914061504, "ts": 1716454225392663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225449942, "dur": 5,
    "args": {
      "External id": 251819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251819,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251819, "pid": 5, "tid": 7, "ts": 1716454225449942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392676, "dur": 5,
    "args": {
      "External id": 251819,
      "cbid": 211, "correlation": 251819
    }
  },
  {
    "ph": "s", "id": 251819, "pid": 76337, "tid": -914061504, "ts": 1716454225392676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225449948, "dur": 6,
    "args": {
      "External id": 251829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251829, "pid": 5, "tid": 7, "ts": 1716454225449948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392732, "dur": 12,
    "args": {
      "External id": 251829,
      "cbid": 211, "correlation": 251829
    }
  },
  {
    "ph": "s", "id": 251829, "pid": 76337, "tid": -914061504, "ts": 1716454225392732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449955, "dur": 10,
    "args": {
      "External id": 251849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251849,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 251849, "pid": 5, "tid": 7, "ts": 1716454225449955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392798, "dur": 11,
    "args": {
      "External id": 251849,
      "cbid": 211, "correlation": 251849
    }
  },
  {
    "ph": "s", "id": 251849, "pid": 76337, "tid": -914061504, "ts": 1716454225392798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225449966, "dur": 3,
    "args": {
      "External id": 251861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251861,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 251861, "pid": 5, "tid": 7, "ts": 1716454225449966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392819, "dur": 6,
    "args": {
      "External id": 251861,
      "cbid": 211, "correlation": 251861
    }
  },
  {
    "ph": "s", "id": 251861, "pid": 76337, "tid": -914061504, "ts": 1716454225392819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225449971, "dur": 7,
    "args": {
      "External id": 251864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251864, "pid": 5, "tid": 7, "ts": 1716454225449971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392838, "dur": 6,
    "args": {
      "External id": 251864,
      "cbid": 211, "correlation": 251864
    }
  },
  {
    "ph": "s", "id": 251864, "pid": 76337, "tid": -914061504, "ts": 1716454225392838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225449979, "dur": 5,
    "args": {
      "External id": 251873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251873,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251873, "pid": 5, "tid": 7, "ts": 1716454225449979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392877, "dur": 10,
    "args": {
      "External id": 251873,
      "cbid": 211, "correlation": 251873
    }
  },
  {
    "ph": "s", "id": 251873, "pid": 76337, "tid": -914061504, "ts": 1716454225392877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392940, "dur": 0,
    "args": {
      "External id": 251883,
      "cbid": 317, "correlation": 251883
    }
  },
  {
    "ph": "f", "id": 251883, "pid": 76337, "tid": -914061504, "ts": 1716454225392940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392941, "dur": 0,
    "args": {
      "External id": 251884,
      "cbid": 203, "correlation": 251884
    }
  },
  {
    "ph": "f", "id": 251884, "pid": 76337, "tid": -914061504, "ts": 1716454225392941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392941, "dur": 0,
    "args": {
      "External id": 251885,
      "cbid": 205, "correlation": 251885
    }
  },
  {
    "ph": "f", "id": 251885, "pid": 76337, "tid": -914061504, "ts": 1716454225392941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449985, "dur": 5,
    "args": {
      "External id": 251889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251889,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251889, "pid": 5, "tid": 7, "ts": 1716454225449985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392955, "dur": 13,
    "args": {
      "External id": 251889,
      "cbid": 211, "correlation": 251889
    }
  },
  {
    "ph": "s", "id": 251889, "pid": 76337, "tid": -914061504, "ts": 1716454225392955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225449991, "dur": 165,
    "args": {
      "External id": 251891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251891,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251891, "pid": 5, "tid": 7, "ts": 1716454225449991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392970, "dur": 13,
    "args": {
      "External id": 251891,
      "cbid": 211, "correlation": 251891
    }
  },
  {
    "ph": "s", "id": 251891, "pid": 76337, "tid": -914061504, "ts": 1716454225392970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225450159, "dur": 1,
    "args": {
      "External id": 251893,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251893,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 251893, "pid": 5, "tid": 7, "ts": 1716454225450159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392988, "dur": 6,
    "args": {
      "External id": 251893,
      "cbid": 51, "correlation": 251893
    }
  },
  {
    "ph": "s", "id": 251893, "pid": 76337, "tid": -914061504, "ts": 1716454225392988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225450162, "dur": 264,
    "args": {
      "External id": 251894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251894,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251894, "pid": 5, "tid": 7, "ts": 1716454225450162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225392995, "dur": 6,
    "args": {
      "External id": 251894,
      "cbid": 211, "correlation": 251894
    }
  },
  {
    "ph": "s", "id": 251894, "pid": 76337, "tid": -914061504, "ts": 1716454225392995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225450428, "dur": 6,
    "args": {
      "External id": 251896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251896,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251896, "pid": 5, "tid": 7, "ts": 1716454225450428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393007, "dur": 5,
    "args": {
      "External id": 251896,
      "cbid": 211, "correlation": 251896
    }
  },
  {
    "ph": "s", "id": 251896, "pid": 76337, "tid": -914061504, "ts": 1716454225393007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225450435, "dur": 6,
    "args": {
      "External id": 251902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251902, "pid": 5, "tid": 7, "ts": 1716454225450435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393036, "dur": 9,
    "args": {
      "External id": 251902,
      "cbid": 211, "correlation": 251902
    }
  },
  {
    "ph": "s", "id": 251902, "pid": 76337, "tid": -914061504, "ts": 1716454225393036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393095, "dur": 0,
    "args": {
      "External id": 251912,
      "cbid": 317, "correlation": 251912
    }
  },
  {
    "ph": "f", "id": 251912, "pid": 76337, "tid": -914061504, "ts": 1716454225393095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393096, "dur": 0,
    "args": {
      "External id": 251913,
      "cbid": 203, "correlation": 251913
    }
  },
  {
    "ph": "f", "id": 251913, "pid": 76337, "tid": -914061504, "ts": 1716454225393096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393097, "dur": 0,
    "args": {
      "External id": 251914,
      "cbid": 205, "correlation": 251914
    }
  },
  {
    "ph": "f", "id": 251914, "pid": 76337, "tid": -914061504, "ts": 1716454225393097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225450442, "dur": 8,
    "args": {
      "External id": 251918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251918,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251918, "pid": 5, "tid": 7, "ts": 1716454225450442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393109, "dur": 12,
    "args": {
      "External id": 251918,
      "cbid": 211, "correlation": 251918
    }
  },
  {
    "ph": "s", "id": 251918, "pid": 76337, "tid": -914061504, "ts": 1716454225393109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225450451, "dur": 3,
    "args": {
      "External id": 251920,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251920,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 251920, "pid": 5, "tid": 7, "ts": 1716454225450451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393126, "dur": 9,
    "args": {
      "External id": 251920,
      "cbid": 51, "correlation": 251920
    }
  },
  {
    "ph": "s", "id": 251920, "pid": 76337, "tid": -914061504, "ts": 1716454225393126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225450456, "dur": 97,
    "args": {
      "External id": 251921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251921,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 251921, "pid": 5, "tid": 7, "ts": 1716454225450456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393136, "dur": 6,
    "args": {
      "External id": 251921,
      "cbid": 211, "correlation": 251921
    }
  },
  {
    "ph": "s", "id": 251921, "pid": 76337, "tid": -914061504, "ts": 1716454225393136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225450554, "dur": 6,
    "args": {
      "External id": 251923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251923,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251923, "pid": 5, "tid": 7, "ts": 1716454225450554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393145, "dur": 5,
    "args": {
      "External id": 251923,
      "cbid": 211, "correlation": 251923
    }
  },
  {
    "ph": "s", "id": 251923, "pid": 76337, "tid": -914061504, "ts": 1716454225393145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225450561, "dur": 6,
    "args": {
      "External id": 251929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251929,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251929, "pid": 5, "tid": 7, "ts": 1716454225450561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393173, "dur": 8,
    "args": {
      "External id": 251929,
      "cbid": 211, "correlation": 251929
    }
  },
  {
    "ph": "s", "id": 251929, "pid": 76337, "tid": -914061504, "ts": 1716454225393173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225450569, "dur": 5,
    "args": {
      "External id": 251937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251937,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251937, "pid": 5, "tid": 7, "ts": 1716454225450569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393201, "dur": 7,
    "args": {
      "External id": 251937,
      "cbid": 211, "correlation": 251937
    }
  },
  {
    "ph": "s", "id": 251937, "pid": 76337, "tid": -914061504, "ts": 1716454225393201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225450575, "dur": 4,
    "args": {
      "External id": 251945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251945,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 251945, "pid": 5, "tid": 7, "ts": 1716454225450575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393231, "dur": 8,
    "args": {
      "External id": 251945,
      "cbid": 211, "correlation": 251945
    }
  },
  {
    "ph": "s", "id": 251945, "pid": 76337, "tid": -914061504, "ts": 1716454225393231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225450581, "dur": 14,
    "args": {
      "External id": 251956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251956,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251956, "pid": 5, "tid": 7, "ts": 1716454225450581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393313, "dur": 14,
    "args": {
      "External id": 251956,
      "cbid": 211, "correlation": 251956
    }
  },
  {
    "ph": "s", "id": 251956, "pid": 76337, "tid": -914061504, "ts": 1716454225393313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393370, "dur": 0,
    "args": {
      "External id": 251966,
      "cbid": 317, "correlation": 251966
    }
  },
  {
    "ph": "f", "id": 251966, "pid": 76337, "tid": -914061504, "ts": 1716454225393370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393371, "dur": 0,
    "args": {
      "External id": 251967,
      "cbid": 203, "correlation": 251967
    }
  },
  {
    "ph": "f", "id": 251967, "pid": 76337, "tid": -914061504, "ts": 1716454225393371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393372, "dur": 0,
    "args": {
      "External id": 251968,
      "cbid": 205, "correlation": 251968
    }
  },
  {
    "ph": "f", "id": 251968, "pid": 76337, "tid": -914061504, "ts": 1716454225393372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225450596, "dur": 9,
    "args": {
      "External id": 251972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251972,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251972, "pid": 5, "tid": 7, "ts": 1716454225450596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393385, "dur": 11,
    "args": {
      "External id": 251972,
      "cbid": 211, "correlation": 251972
    }
  },
  {
    "ph": "s", "id": 251972, "pid": 76337, "tid": -914061504, "ts": 1716454225393385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225450607, "dur": 165,
    "args": {
      "External id": 251974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251974,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251974, "pid": 5, "tid": 7, "ts": 1716454225450607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393399, "dur": 5,
    "args": {
      "External id": 251974,
      "cbid": 211, "correlation": 251974
    }
  },
  {
    "ph": "s", "id": 251974, "pid": 76337, "tid": -914061504, "ts": 1716454225393399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225450774, "dur": 1,
    "args": {
      "External id": 251976,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 251976,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 251976, "pid": 5, "tid": 7, "ts": 1716454225450774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393410, "dur": 7,
    "args": {
      "External id": 251976,
      "cbid": 51, "correlation": 251976
    }
  },
  {
    "ph": "s", "id": 251976, "pid": 76337, "tid": -914061504, "ts": 1716454225393410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225450777, "dur": 660,
    "args": {
      "External id": 251977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251977,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 251977, "pid": 5, "tid": 7, "ts": 1716454225450777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393418, "dur": 6,
    "args": {
      "External id": 251977,
      "cbid": 211, "correlation": 251977
    }
  },
  {
    "ph": "s", "id": 251977, "pid": 76337, "tid": -914061504, "ts": 1716454225393418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225451439, "dur": 12,
    "args": {
      "External id": 251979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251979,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251979, "pid": 5, "tid": 7, "ts": 1716454225451439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393427, "dur": 5,
    "args": {
      "External id": 251979,
      "cbid": 211, "correlation": 251979
    }
  },
  {
    "ph": "s", "id": 251979, "pid": 76337, "tid": -914061504, "ts": 1716454225393427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225451452, "dur": 15,
    "args": {
      "External id": 251985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251985,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251985, "pid": 5, "tid": 7, "ts": 1716454225451452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393455, "dur": 9,
    "args": {
      "External id": 251985,
      "cbid": 211, "correlation": 251985
    }
  },
  {
    "ph": "s", "id": 251985, "pid": 76337, "tid": -914061504, "ts": 1716454225393455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225451468, "dur": 31,
    "args": {
      "External id": 251994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 251994,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 251994, "pid": 5, "tid": 7, "ts": 1716454225451468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393546, "dur": 12,
    "args": {
      "External id": 251994,
      "cbid": 211, "correlation": 251994
    }
  },
  {
    "ph": "s", "id": 251994, "pid": 76337, "tid": -914061504, "ts": 1716454225393546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225451501, "dur": 31,
    "args": {
      "External id": 252014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252014,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 252014, "pid": 5, "tid": 7, "ts": 1716454225451501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393612, "dur": 12,
    "args": {
      "External id": 252014,
      "cbid": 211, "correlation": 252014
    }
  },
  {
    "ph": "s", "id": 252014, "pid": 76337, "tid": -914061504, "ts": 1716454225393612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225451533, "dur": 4,
    "args": {
      "External id": 252026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252026,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252026, "pid": 5, "tid": 7, "ts": 1716454225451533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393634, "dur": 6,
    "args": {
      "External id": 252026,
      "cbid": 211, "correlation": 252026
    }
  },
  {
    "ph": "s", "id": 252026, "pid": 76337, "tid": -914061504, "ts": 1716454225393634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225451539, "dur": 29,
    "args": {
      "External id": 252029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252029,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252029, "pid": 5, "tid": 7, "ts": 1716454225451539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393652, "dur": 7,
    "args": {
      "External id": 252029,
      "cbid": 211, "correlation": 252029
    }
  },
  {
    "ph": "s", "id": 252029, "pid": 76337, "tid": -914061504, "ts": 1716454225393652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225451570, "dur": 21,
    "args": {
      "External id": 252038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252038,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252038, "pid": 5, "tid": 7, "ts": 1716454225451570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393690, "dur": 9,
    "args": {
      "External id": 252038,
      "cbid": 211, "correlation": 252038
    }
  },
  {
    "ph": "s", "id": 252038, "pid": 76337, "tid": -914061504, "ts": 1716454225393690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393743, "dur": 0,
    "args": {
      "External id": 252048,
      "cbid": 317, "correlation": 252048
    }
  },
  {
    "ph": "f", "id": 252048, "pid": 76337, "tid": -914061504, "ts": 1716454225393743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393744, "dur": 0,
    "args": {
      "External id": 252049,
      "cbid": 203, "correlation": 252049
    }
  },
  {
    "ph": "f", "id": 252049, "pid": 76337, "tid": -914061504, "ts": 1716454225393744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393745, "dur": 0,
    "args": {
      "External id": 252050,
      "cbid": 205, "correlation": 252050
    }
  },
  {
    "ph": "f", "id": 252050, "pid": 76337, "tid": -914061504, "ts": 1716454225393745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225451592, "dur": 22,
    "args": {
      "External id": 252054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252054, "pid": 5, "tid": 7, "ts": 1716454225451592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393760, "dur": 12,
    "args": {
      "External id": 252054,
      "cbid": 211, "correlation": 252054
    }
  },
  {
    "ph": "s", "id": 252054, "pid": 76337, "tid": -914061504, "ts": 1716454225393760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225451615, "dur": 327,
    "args": {
      "External id": 252056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252056, "pid": 5, "tid": 7, "ts": 1716454225451615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393775, "dur": 5,
    "args": {
      "External id": 252056,
      "cbid": 211, "correlation": 252056
    }
  },
  {
    "ph": "s", "id": 252056, "pid": 76337, "tid": -914061504, "ts": 1716454225393775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225451945, "dur": 1,
    "args": {
      "External id": 252058,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 252058,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 252058, "pid": 5, "tid": 7, "ts": 1716454225451945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393786, "dur": 6,
    "args": {
      "External id": 252058,
      "cbid": 51, "correlation": 252058
    }
  },
  {
    "ph": "s", "id": 252058, "pid": 76337, "tid": -914061504, "ts": 1716454225393786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225451948, "dur": 1261,
    "args": {
      "External id": 252059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252059,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252059, "pid": 5, "tid": 7, "ts": 1716454225451948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393793, "dur": 6,
    "args": {
      "External id": 252059,
      "cbid": 211, "correlation": 252059
    }
  },
  {
    "ph": "s", "id": 252059, "pid": 76337, "tid": -914061504, "ts": 1716454225393793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225453211, "dur": 14,
    "args": {
      "External id": 252061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252061,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252061, "pid": 5, "tid": 7, "ts": 1716454225453211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393804, "dur": 5,
    "args": {
      "External id": 252061,
      "cbid": 211, "correlation": 252061
    }
  },
  {
    "ph": "s", "id": 252061, "pid": 76337, "tid": -914061504, "ts": 1716454225393804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225453226, "dur": 15,
    "args": {
      "External id": 252067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252067, "pid": 5, "tid": 7, "ts": 1716454225453226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393832, "dur": 8,
    "args": {
      "External id": 252067,
      "cbid": 211, "correlation": 252067
    }
  },
  {
    "ph": "s", "id": 252067, "pid": 76337, "tid": -914061504, "ts": 1716454225393832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225453242, "dur": 3,
    "args": {
      "External id": 252075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252075,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 252075, "pid": 5, "tid": 7, "ts": 1716454225453242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393876, "dur": 9,
    "args": {
      "External id": 252075,
      "cbid": 211, "correlation": 252075
    }
  },
  {
    "ph": "s", "id": 252075, "pid": 76337, "tid": -914061504, "ts": 1716454225393876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393940, "dur": 1,
    "args": {
      "External id": 252091,
      "cbid": 251, "correlation": 252091
    }
  },
  {
    "ph": "f", "id": 252091, "pid": 76337, "tid": -914061504, "ts": 1716454225393940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393945, "dur": 0,
    "args": {
      "External id": 252093,
      "cbid": 251, "correlation": 252093
    }
  },
  {
    "ph": "f", "id": 252093, "pid": 76337, "tid": -914061504, "ts": 1716454225393945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225453247, "dur": 13,
    "args": {
      "External id": 252094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252094,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252094, "pid": 5, "tid": 7, "ts": 1716454225453247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393948, "dur": 11,
    "args": {
      "External id": 252094,
      "cbid": 211, "correlation": 252094
    }
  },
  {
    "ph": "s", "id": 252094, "pid": 76337, "tid": -914061504, "ts": 1716454225393948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225453261, "dur": 5,
    "args": {
      "External id": 252096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252096, "pid": 5, "tid": 7, "ts": 1716454225453261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225393960, "dur": 5,
    "args": {
      "External id": 252096,
      "cbid": 211, "correlation": 252096
    }
  },
  {
    "ph": "s", "id": 252096, "pid": 76337, "tid": -914061504, "ts": 1716454225393960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225453267, "dur": 17,
    "args": {
      "External id": 252106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252106, "pid": 5, "tid": 7, "ts": 1716454225453267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394027, "dur": 12,
    "args": {
      "External id": 252106,
      "cbid": 211, "correlation": 252106
    }
  },
  {
    "ph": "s", "id": 252106, "pid": 76337, "tid": -914061504, "ts": 1716454225394027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225453286, "dur": 18,
    "args": {
      "External id": 252126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252126,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 252126, "pid": 5, "tid": 7, "ts": 1716454225453286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394092, "dur": 11,
    "args": {
      "External id": 252126,
      "cbid": 211, "correlation": 252126
    }
  },
  {
    "ph": "s", "id": 252126, "pid": 76337, "tid": -914061504, "ts": 1716454225394092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225453305, "dur": 4,
    "args": {
      "External id": 252138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252138,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 252138, "pid": 5, "tid": 7, "ts": 1716454225453305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394113, "dur": 6,
    "args": {
      "External id": 252138,
      "cbid": 211, "correlation": 252138
    }
  },
  {
    "ph": "s", "id": 252138, "pid": 76337, "tid": -914061504, "ts": 1716454225394113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225453310, "dur": 17,
    "args": {
      "External id": 252141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252141, "pid": 5, "tid": 7, "ts": 1716454225453310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394132, "dur": 6,
    "args": {
      "External id": 252141,
      "cbid": 211, "correlation": 252141
    }
  },
  {
    "ph": "s", "id": 252141, "pid": 76337, "tid": -914061504, "ts": 1716454225394132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225453328, "dur": 11,
    "args": {
      "External id": 252150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252150,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252150, "pid": 5, "tid": 7, "ts": 1716454225453328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394173, "dur": 10,
    "args": {
      "External id": 252150,
      "cbid": 211, "correlation": 252150
    }
  },
  {
    "ph": "s", "id": 252150, "pid": 76337, "tid": -914061504, "ts": 1716454225394173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394235, "dur": 0,
    "args": {
      "External id": 252160,
      "cbid": 317, "correlation": 252160
    }
  },
  {
    "ph": "f", "id": 252160, "pid": 76337, "tid": -914061504, "ts": 1716454225394235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394236, "dur": 0,
    "args": {
      "External id": 252161,
      "cbid": 203, "correlation": 252161
    }
  },
  {
    "ph": "f", "id": 252161, "pid": 76337, "tid": -914061504, "ts": 1716454225394236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394236, "dur": 0,
    "args": {
      "External id": 252162,
      "cbid": 205, "correlation": 252162
    }
  },
  {
    "ph": "f", "id": 252162, "pid": 76337, "tid": -914061504, "ts": 1716454225394236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225453340, "dur": 11,
    "args": {
      "External id": 252166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252166,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252166, "pid": 5, "tid": 7, "ts": 1716454225453340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394250, "dur": 12,
    "args": {
      "External id": 252166,
      "cbid": 211, "correlation": 252166
    }
  },
  {
    "ph": "s", "id": 252166, "pid": 76337, "tid": -914061504, "ts": 1716454225394250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225453353, "dur": 165,
    "args": {
      "External id": 252168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252168, "pid": 5, "tid": 7, "ts": 1716454225453353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394264, "dur": 6,
    "args": {
      "External id": 252168,
      "cbid": 211, "correlation": 252168
    }
  },
  {
    "ph": "s", "id": 252168, "pid": 76337, "tid": -914061504, "ts": 1716454225394264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225453521, "dur": 1,
    "args": {
      "External id": 252170,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 252170,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5357142857142857
    }
  },
  {
    "ph": "f", "id": 252170, "pid": 5, "tid": 7, "ts": 1716454225453521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394276, "dur": 6,
    "args": {
      "External id": 252170,
      "cbid": 51, "correlation": 252170
    }
  },
  {
    "ph": "s", "id": 252170, "pid": 76337, "tid": -914061504, "ts": 1716454225394276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225453525, "dur": 660,
    "args": {
      "External id": 252171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252171,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252171, "pid": 5, "tid": 7, "ts": 1716454225453525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394283, "dur": 6,
    "args": {
      "External id": 252171,
      "cbid": 211, "correlation": 252171
    }
  },
  {
    "ph": "s", "id": 252171, "pid": 76337, "tid": -914061504, "ts": 1716454225394283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225454186, "dur": 12,
    "args": {
      "External id": 252173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252173, "pid": 5, "tid": 7, "ts": 1716454225454186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394293, "dur": 5,
    "args": {
      "External id": 252173,
      "cbid": 211, "correlation": 252173
    }
  },
  {
    "ph": "s", "id": 252173, "pid": 76337, "tid": -914061504, "ts": 1716454225394293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225454199, "dur": 15,
    "args": {
      "External id": 252179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252179,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252179, "pid": 5, "tid": 7, "ts": 1716454225454199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394322, "dur": 8,
    "args": {
      "External id": 252179,
      "cbid": 211, "correlation": 252179
    }
  },
  {
    "ph": "s", "id": 252179, "pid": 76337, "tid": -914061504, "ts": 1716454225394322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394380, "dur": 0,
    "args": {
      "External id": 252189,
      "cbid": 317, "correlation": 252189
    }
  },
  {
    "ph": "f", "id": 252189, "pid": 76337, "tid": -914061504, "ts": 1716454225394380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394381, "dur": 0,
    "args": {
      "External id": 252190,
      "cbid": 203, "correlation": 252190
    }
  },
  {
    "ph": "f", "id": 252190, "pid": 76337, "tid": -914061504, "ts": 1716454225394381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394382, "dur": 0,
    "args": {
      "External id": 252191,
      "cbid": 205, "correlation": 252191
    }
  },
  {
    "ph": "f", "id": 252191, "pid": 76337, "tid": -914061504, "ts": 1716454225394382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225454215, "dur": 21,
    "args": {
      "External id": 252195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252195,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252195, "pid": 5, "tid": 7, "ts": 1716454225454215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394395, "dur": 11,
    "args": {
      "External id": 252195,
      "cbid": 211, "correlation": 252195
    }
  },
  {
    "ph": "s", "id": 252195, "pid": 76337, "tid": -914061504, "ts": 1716454225394395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225454237, "dur": 4,
    "args": {
      "External id": 252197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252197, "pid": 5, "tid": 7, "ts": 1716454225454237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394411, "dur": 7,
    "args": {
      "External id": 252197,
      "cbid": 211, "correlation": 252197
    }
  },
  {
    "ph": "s", "id": 252197, "pid": 76337, "tid": -914061504, "ts": 1716454225394411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394422, "dur": 0,
    "args": {
      "External id": 252198,
      "cbid": 51, "correlation": 252198
    }
  },
  {
    "ph": "s", "id": 252198, "pid": 76337, "tid": -914061504, "ts": 1716454225394422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225454243, "dur": 180,
    "args": {
      "External id": 252199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252199,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 252199, "pid": 5, "tid": 7, "ts": 1716454225454243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394423, "dur": 5,
    "args": {
      "External id": 252199,
      "cbid": 211, "correlation": 252199
    }
  },
  {
    "ph": "s", "id": 252199, "pid": 76337, "tid": -914061504, "ts": 1716454225394423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225454425, "dur": 16,
    "args": {
      "External id": 252204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252204, "pid": 5, "tid": 7, "ts": 1716454225454425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394449, "dur": 8,
    "args": {
      "External id": 252204,
      "cbid": 211, "correlation": 252204
    }
  },
  {
    "ph": "s", "id": 252204, "pid": 76337, "tid": -914061504, "ts": 1716454225394449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225454442, "dur": 12,
    "args": {
      "External id": 252212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252212,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252212, "pid": 5, "tid": 7, "ts": 1716454225454442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394479, "dur": 8,
    "args": {
      "External id": 252212,
      "cbid": 211, "correlation": 252212
    }
  },
  {
    "ph": "s", "id": 252212, "pid": 76337, "tid": -914061504, "ts": 1716454225394479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225454455, "dur": 10,
    "args": {
      "External id": 252220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252220,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252220, "pid": 5, "tid": 7, "ts": 1716454225454455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394507, "dur": 9,
    "args": {
      "External id": 252220,
      "cbid": 211, "correlation": 252220
    }
  },
  {
    "ph": "s", "id": 252220, "pid": 76337, "tid": -914061504, "ts": 1716454225394507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225454466, "dur": 19,
    "args": {
      "External id": 252240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252240,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 252240, "pid": 5, "tid": 7, "ts": 1716454225454466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394589, "dur": 13,
    "args": {
      "External id": 252240,
      "cbid": 211, "correlation": 252240
    }
  },
  {
    "ph": "s", "id": 252240, "pid": 76337, "tid": -914061504, "ts": 1716454225394589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225454487, "dur": 5,
    "args": {
      "External id": 252252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252252,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 252252, "pid": 5, "tid": 7, "ts": 1716454225454487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394612, "dur": 6,
    "args": {
      "External id": 252252,
      "cbid": 211, "correlation": 252252
    }
  },
  {
    "ph": "s", "id": 252252, "pid": 76337, "tid": -914061504, "ts": 1716454225394612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225454493, "dur": 17,
    "args": {
      "External id": 252255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252255, "pid": 5, "tid": 7, "ts": 1716454225454493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394629, "dur": 6,
    "args": {
      "External id": 252255,
      "cbid": 211, "correlation": 252255
    }
  },
  {
    "ph": "s", "id": 252255, "pid": 76337, "tid": -914061504, "ts": 1716454225394629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394686, "dur": 0,
    "args": {
      "External id": 252266,
      "cbid": 317, "correlation": 252266
    }
  },
  {
    "ph": "f", "id": 252266, "pid": 76337, "tid": -914061504, "ts": 1716454225394686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394687, "dur": 0,
    "args": {
      "External id": 252267,
      "cbid": 203, "correlation": 252267
    }
  },
  {
    "ph": "f", "id": 252267, "pid": 76337, "tid": -914061504, "ts": 1716454225394687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394688, "dur": 0,
    "args": {
      "External id": 252268,
      "cbid": 205, "correlation": 252268
    }
  },
  {
    "ph": "f", "id": 252268, "pid": 76337, "tid": -914061504, "ts": 1716454225394688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225454511, "dur": 11,
    "args": {
      "External id": 252272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252272,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252272, "pid": 5, "tid": 7, "ts": 1716454225454511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394701, "dur": 11,
    "args": {
      "External id": 252272,
      "cbid": 211, "correlation": 252272
    }
  },
  {
    "ph": "s", "id": 252272, "pid": 76337, "tid": -914061504, "ts": 1716454225394701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225454524, "dur": 4,
    "args": {
      "External id": 252274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252274,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252274, "pid": 5, "tid": 7, "ts": 1716454225454524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394717, "dur": 7,
    "args": {
      "External id": 252274,
      "cbid": 211, "correlation": 252274
    }
  },
  {
    "ph": "s", "id": 252274, "pid": 76337, "tid": -914061504, "ts": 1716454225394717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394726, "dur": 0,
    "args": {
      "External id": 252275,
      "cbid": 51, "correlation": 252275
    }
  },
  {
    "ph": "s", "id": 252275, "pid": 76337, "tid": -914061504, "ts": 1716454225394726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225454529, "dur": 93,
    "args": {
      "External id": 252276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252276,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 252276, "pid": 5, "tid": 7, "ts": 1716454225454529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394727, "dur": 5,
    "args": {
      "External id": 252276,
      "cbid": 211, "correlation": 252276
    }
  },
  {
    "ph": "s", "id": 252276, "pid": 76337, "tid": -914061504, "ts": 1716454225394727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225454623, "dur": 16,
    "args": {
      "External id": 252281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252281, "pid": 5, "tid": 7, "ts": 1716454225454623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394753, "dur": 9,
    "args": {
      "External id": 252281,
      "cbid": 211, "correlation": 252281
    }
  },
  {
    "ph": "s", "id": 252281, "pid": 76337, "tid": -914061504, "ts": 1716454225394753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225454641, "dur": 85,
    "args": {
      "External id": 252290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252290,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252290, "pid": 5, "tid": 7, "ts": 1716454225454641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394836, "dur": 14,
    "args": {
      "External id": 252290,
      "cbid": 211, "correlation": 252290
    }
  },
  {
    "ph": "s", "id": 252290, "pid": 76337, "tid": -914061504, "ts": 1716454225394836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225454727, "dur": 30,
    "args": {
      "External id": 252312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252312,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252312, "pid": 5, "tid": 7, "ts": 1716454225454727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225394895, "dur": 10,
    "args": {
      "External id": 252312,
      "cbid": 211, "correlation": 252312
    }
  },
  {
    "ph": "s", "id": 252312, "pid": 76337, "tid": -914061504, "ts": 1716454225394895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395000, "dur": 2,
    "args": {
      "External id": 252323,
      "cbid": 251, "correlation": 252323
    }
  },
  {
    "ph": "f", "id": 252323, "pid": 76337, "tid": -914061504, "ts": 1716454225395000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225454759, "dur": 169,
    "args": {
      "External id": 252324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252324,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252324, "pid": 5, "tid": 7, "ts": 1716454225454759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395006, "dur": 14,
    "args": {
      "External id": 252324,
      "cbid": 211, "correlation": 252324
    }
  },
  {
    "ph": "s", "id": 252324, "pid": 76337, "tid": -914061504, "ts": 1716454225395006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395080, "dur": 1,
    "args": {
      "External id": 252335,
      "cbid": 251, "correlation": 252335
    }
  },
  {
    "ph": "f", "id": 252335, "pid": 76337, "tid": -914061504, "ts": 1716454225395080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225454929, "dur": 160,
    "args": {
      "External id": 252336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252336,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252336, "pid": 5, "tid": 7, "ts": 1716454225454929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395085, "dur": 11,
    "args": {
      "External id": 252336,
      "cbid": 211, "correlation": 252336
    }
  },
  {
    "ph": "s", "id": 252336, "pid": 76337, "tid": -914061504, "ts": 1716454225395085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395150, "dur": 1,
    "args": {
      "External id": 252347,
      "cbid": 251, "correlation": 252347
    }
  },
  {
    "ph": "f", "id": 252347, "pid": 76337, "tid": -914061504, "ts": 1716454225395150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225455090, "dur": 158,
    "args": {
      "External id": 252348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252348,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252348, "pid": 5, "tid": 7, "ts": 1716454225455090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395154, "dur": 11,
    "args": {
      "External id": 252348,
      "cbid": 211, "correlation": 252348
    }
  },
  {
    "ph": "s", "id": 252348, "pid": 76337, "tid": -914061504, "ts": 1716454225395154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225455249, "dur": 347,
    "args": {
      "External id": 252373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252373,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252373, "pid": 5, "tid": 7, "ts": 1716454225455249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395242, "dur": 13,
    "args": {
      "External id": 252373,
      "cbid": 211, "correlation": 252373
    }
  },
  {
    "ph": "s", "id": 252373, "pid": 76337, "tid": -914061504, "ts": 1716454225395242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395342, "dur": 1,
    "args": {
      "External id": 252391,
      "cbid": 251, "correlation": 252391
    }
  },
  {
    "ph": "f", "id": 252391, "pid": 76337, "tid": -914061504, "ts": 1716454225395342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225455598, "dur": 170,
    "args": {
      "External id": 252393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252393,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252393, "pid": 5, "tid": 7, "ts": 1716454225455598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395348, "dur": 13,
    "args": {
      "External id": 252393,
      "cbid": 211, "correlation": 252393
    }
  },
  {
    "ph": "s", "id": 252393, "pid": 76337, "tid": -914061504, "ts": 1716454225395348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225455769, "dur": 20,
    "args": {
      "External id": 252401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252401,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252401, "pid": 5, "tid": 7, "ts": 1716454225455769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395418, "dur": 12,
    "args": {
      "External id": 252401,
      "cbid": 211, "correlation": 252401
    }
  },
  {
    "ph": "s", "id": 252401, "pid": 76337, "tid": -914061504, "ts": 1716454225395418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225455790, "dur": 28,
    "args": {
      "External id": 252409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252409,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252409, "pid": 5, "tid": 7, "ts": 1716454225455790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395457, "dur": 9,
    "args": {
      "External id": 252409,
      "cbid": 211, "correlation": 252409
    }
  },
  {
    "ph": "s", "id": 252409, "pid": 76337, "tid": -914061504, "ts": 1716454225395457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225455819, "dur": 19,
    "args": {
      "External id": 252420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252420, "pid": 5, "tid": 7, "ts": 1716454225455819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395530, "dur": 12,
    "args": {
      "External id": 252420,
      "cbid": 211, "correlation": 252420
    }
  },
  {
    "ph": "s", "id": 252420, "pid": 76337, "tid": -914061504, "ts": 1716454225395530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225455839, "dur": 16,
    "args": {
      "External id": 252442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252442,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252442, "pid": 5, "tid": 7, "ts": 1716454225455839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395560, "dur": 9,
    "args": {
      "External id": 252442,
      "cbid": 211, "correlation": 252442
    }
  },
  {
    "ph": "s", "id": 252442, "pid": 76337, "tid": -914061504, "ts": 1716454225395560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395646, "dur": 1,
    "args": {
      "External id": 252453,
      "cbid": 251, "correlation": 252453
    }
  },
  {
    "ph": "f", "id": 252453, "pid": 76337, "tid": -914061504, "ts": 1716454225395646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225455857, "dur": 91,
    "args": {
      "External id": 252454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252454,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252454, "pid": 5, "tid": 7, "ts": 1716454225455857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395652, "dur": 13,
    "args": {
      "External id": 252454,
      "cbid": 211, "correlation": 252454
    }
  },
  {
    "ph": "s", "id": 252454, "pid": 76337, "tid": -914061504, "ts": 1716454225395652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395722, "dur": 1,
    "args": {
      "External id": 252465,
      "cbid": 251, "correlation": 252465
    }
  },
  {
    "ph": "f", "id": 252465, "pid": 76337, "tid": -914061504, "ts": 1716454225395722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395726, "dur": 0,
    "args": {
      "External id": 252466,
      "cbid": 251, "correlation": 252466
    }
  },
  {
    "ph": "f", "id": 252466, "pid": 76337, "tid": -914061504, "ts": 1716454225395726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225455949, "dur": 12,
    "args": {
      "External id": 252467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252467,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252467, "pid": 5, "tid": 7, "ts": 1716454225455949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395727, "dur": 12,
    "args": {
      "External id": 252467,
      "cbid": 211, "correlation": 252467
    }
  },
  {
    "ph": "s", "id": 252467, "pid": 76337, "tid": -914061504, "ts": 1716454225395727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225455963, "dur": 6,
    "args": {
      "External id": 252469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252469,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252469, "pid": 5, "tid": 7, "ts": 1716454225455963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395741, "dur": 6,
    "args": {
      "External id": 252469,
      "cbid": 211, "correlation": 252469
    }
  },
  {
    "ph": "s", "id": 252469, "pid": 76337, "tid": -914061504, "ts": 1716454225395741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395799, "dur": 1,
    "args": {
      "External id": 252480,
      "cbid": 251, "correlation": 252480
    }
  },
  {
    "ph": "f", "id": 252480, "pid": 76337, "tid": -914061504, "ts": 1716454225395799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395802, "dur": 0,
    "args": {
      "External id": 252481,
      "cbid": 251, "correlation": 252481
    }
  },
  {
    "ph": "f", "id": 252481, "pid": 76337, "tid": -914061504, "ts": 1716454225395802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225455970, "dur": 9,
    "args": {
      "External id": 252482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252482,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252482, "pid": 5, "tid": 7, "ts": 1716454225455970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395803, "dur": 12,
    "args": {
      "External id": 252482,
      "cbid": 211, "correlation": 252482
    }
  },
  {
    "ph": "s", "id": 252482, "pid": 76337, "tid": -914061504, "ts": 1716454225395803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225455980, "dur": 4,
    "args": {
      "External id": 252484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252484,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252484, "pid": 5, "tid": 7, "ts": 1716454225455980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395817, "dur": 5,
    "args": {
      "External id": 252484,
      "cbid": 211, "correlation": 252484
    }
  },
  {
    "ph": "s", "id": 252484, "pid": 76337, "tid": -914061504, "ts": 1716454225395817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225455984, "dur": 56,
    "args": {
      "External id": 252509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252509,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252509, "pid": 5, "tid": 7, "ts": 1716454225455984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395894, "dur": 12,
    "args": {
      "External id": 252509,
      "cbid": 211, "correlation": 252509
    }
  },
  {
    "ph": "s", "id": 252509, "pid": 76337, "tid": -914061504, "ts": 1716454225395894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225395999, "dur": 2,
    "args": {
      "External id": 252527,
      "cbid": 251, "correlation": 252527
    }
  },
  {
    "ph": "f", "id": 252527, "pid": 76337, "tid": -914061504, "ts": 1716454225395999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225456042, "dur": 92,
    "args": {
      "External id": 252529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252529,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252529, "pid": 5, "tid": 7, "ts": 1716454225456042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396005, "dur": 14,
    "args": {
      "External id": 252529,
      "cbid": 211, "correlation": 252529
    }
  },
  {
    "ph": "s", "id": 252529, "pid": 76337, "tid": -914061504, "ts": 1716454225396005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225456136, "dur": 10,
    "args": {
      "External id": 252537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252537,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252537, "pid": 5, "tid": 7, "ts": 1716454225456136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396076, "dur": 12,
    "args": {
      "External id": 252537,
      "cbid": 211, "correlation": 252537
    }
  },
  {
    "ph": "s", "id": 252537, "pid": 76337, "tid": -914061504, "ts": 1716454225396076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225456147, "dur": 21,
    "args": {
      "External id": 252545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252545, "pid": 5, "tid": 7, "ts": 1716454225456147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396118, "dur": 9,
    "args": {
      "External id": 252545,
      "cbid": 211, "correlation": 252545
    }
  },
  {
    "ph": "s", "id": 252545, "pid": 76337, "tid": -914061504, "ts": 1716454225396118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225456169, "dur": 17,
    "args": {
      "External id": 252567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252567,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252567, "pid": 5, "tid": 7, "ts": 1716454225456169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396172, "dur": 10,
    "args": {
      "External id": 252567,
      "cbid": 211, "correlation": 252567
    }
  },
  {
    "ph": "s", "id": 252567, "pid": 76337, "tid": -914061504, "ts": 1716454225396172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396261, "dur": 1,
    "args": {
      "External id": 252583,
      "cbid": 251, "correlation": 252583
    }
  },
  {
    "ph": "f", "id": 252583, "pid": 76337, "tid": -914061504, "ts": 1716454225396261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396266, "dur": 0,
    "args": {
      "External id": 252585,
      "cbid": 251, "correlation": 252585
    }
  },
  {
    "ph": "f", "id": 252585, "pid": 76337, "tid": -914061504, "ts": 1716454225396266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225456188, "dur": 504,
    "args": {
      "External id": 252586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252586,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252586, "pid": 5, "tid": 7, "ts": 1716454225456188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396268, "dur": 13,
    "args": {
      "External id": 252586,
      "cbid": 211, "correlation": 252586
    }
  },
  {
    "ph": "s", "id": 252586, "pid": 76337, "tid": -914061504, "ts": 1716454225396268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225456694, "dur": 66,
    "args": {
      "External id": 252594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252594,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252594, "pid": 5, "tid": 7, "ts": 1716454225456694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396336, "dur": 13,
    "args": {
      "External id": 252594,
      "cbid": 211, "correlation": 252594
    }
  },
  {
    "ph": "s", "id": 252594, "pid": 76337, "tid": -914061504, "ts": 1716454225396336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225456761, "dur": 67,
    "args": {
      "External id": 252602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252602, "pid": 5, "tid": 7, "ts": 1716454225456761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396369, "dur": 8,
    "args": {
      "External id": 252602,
      "cbid": 211, "correlation": 252602
    }
  },
  {
    "ph": "s", "id": 252602, "pid": 76337, "tid": -914061504, "ts": 1716454225396369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396447, "dur": 1,
    "args": {
      "External id": 252618,
      "cbid": 251, "correlation": 252618
    }
  },
  {
    "ph": "f", "id": 252618, "pid": 76337, "tid": -914061504, "ts": 1716454225396447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225456831, "dur": 1,
    "args": {
      "External id": 252620,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 252620,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 252620, "pid": 5, "tid": 7, "ts": 1716454225456831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396452, "dur": 11,
    "args": {
      "External id": 252620,
      "cbid": 51, "correlation": 252620
    }
  },
  {
    "ph": "s", "id": 252620, "pid": 76337, "tid": -914061504, "ts": 1716454225396452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225456834, "dur": 274,
    "args": {
      "External id": 252621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252621,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252621, "pid": 5, "tid": 7, "ts": 1716454225456834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396465, "dur": 11,
    "args": {
      "External id": 252621,
      "cbid": 211, "correlation": 252621
    }
  },
  {
    "ph": "s", "id": 252621, "pid": 76337, "tid": -914061504, "ts": 1716454225396465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225457110, "dur": 14,
    "args": {
      "External id": 252629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252629,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252629, "pid": 5, "tid": 7, "ts": 1716454225457110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396507, "dur": 11,
    "args": {
      "External id": 252629,
      "cbid": 211, "correlation": 252629
    }
  },
  {
    "ph": "s", "id": 252629, "pid": 76337, "tid": -914061504, "ts": 1716454225396507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225457125, "dur": 39,
    "args": {
      "External id": 252640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252640, "pid": 5, "tid": 7, "ts": 1716454225457125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396577, "dur": 12,
    "args": {
      "External id": 252640,
      "cbid": 211, "correlation": 252640
    }
  },
  {
    "ph": "s", "id": 252640, "pid": 76337, "tid": -914061504, "ts": 1716454225396577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396641, "dur": 0,
    "args": {
      "External id": 252652,
      "cbid": 317, "correlation": 252652
    }
  },
  {
    "ph": "f", "id": 252652, "pid": 76337, "tid": -914061504, "ts": 1716454225396641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396642, "dur": 0,
    "args": {
      "External id": 252653,
      "cbid": 203, "correlation": 252653
    }
  },
  {
    "ph": "f", "id": 252653, "pid": 76337, "tid": -914061504, "ts": 1716454225396642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396642, "dur": 0,
    "args": {
      "External id": 252654,
      "cbid": 205, "correlation": 252654
    }
  },
  {
    "ph": "f", "id": 252654, "pid": 76337, "tid": -914061504, "ts": 1716454225396642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225457165, "dur": 13,
    "args": {
      "External id": 252658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252658, "pid": 5, "tid": 7, "ts": 1716454225457165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396657, "dur": 13,
    "args": {
      "External id": 252658,
      "cbid": 211, "correlation": 252658
    }
  },
  {
    "ph": "s", "id": 252658, "pid": 76337, "tid": -914061504, "ts": 1716454225396657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225457179, "dur": 4,
    "args": {
      "External id": 252660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252660, "pid": 5, "tid": 7, "ts": 1716454225457179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396675, "dur": 6,
    "args": {
      "External id": 252660,
      "cbid": 211, "correlation": 252660
    }
  },
  {
    "ph": "s", "id": 252660, "pid": 76337, "tid": -914061504, "ts": 1716454225396675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396684, "dur": 0,
    "args": {
      "External id": 252661,
      "cbid": 51, "correlation": 252661
    }
  },
  {
    "ph": "s", "id": 252661, "pid": 76337, "tid": -914061504, "ts": 1716454225396684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225457184, "dur": 98,
    "args": {
      "External id": 252662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252662,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 252662, "pid": 5, "tid": 7, "ts": 1716454225457184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396685, "dur": 5,
    "args": {
      "External id": 252662,
      "cbid": 211, "correlation": 252662
    }
  },
  {
    "ph": "s", "id": 252662, "pid": 76337, "tid": -914061504, "ts": 1716454225396685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225457284, "dur": 17,
    "args": {
      "External id": 252667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252667, "pid": 5, "tid": 7, "ts": 1716454225457284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396712, "dur": 9,
    "args": {
      "External id": 252667,
      "cbid": 211, "correlation": 252667
    }
  },
  {
    "ph": "s", "id": 252667, "pid": 76337, "tid": -914061504, "ts": 1716454225396712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225457302, "dur": 13,
    "args": {
      "External id": 252675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252675,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252675, "pid": 5, "tid": 7, "ts": 1716454225457302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396745, "dur": 8,
    "args": {
      "External id": 252675,
      "cbid": 211, "correlation": 252675
    }
  },
  {
    "ph": "s", "id": 252675, "pid": 76337, "tid": -914061504, "ts": 1716454225396745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225457316, "dur": 31,
    "args": {
      "External id": 252684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252684,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252684, "pid": 5, "tid": 7, "ts": 1716454225457316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396783, "dur": 10,
    "args": {
      "External id": 252684,
      "cbid": 211, "correlation": 252684
    }
  },
  {
    "ph": "s", "id": 252684, "pid": 76337, "tid": -914061504, "ts": 1716454225396783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225457348, "dur": 32,
    "args": {
      "External id": 252704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252704,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 252704, "pid": 5, "tid": 7, "ts": 1716454225457348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396855, "dur": 13,
    "args": {
      "External id": 252704,
      "cbid": 211, "correlation": 252704
    }
  },
  {
    "ph": "s", "id": 252704, "pid": 76337, "tid": -914061504, "ts": 1716454225396855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225457381, "dur": 5,
    "args": {
      "External id": 252716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252716,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252716, "pid": 5, "tid": 7, "ts": 1716454225457381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396877, "dur": 6,
    "args": {
      "External id": 252716,
      "cbid": 211, "correlation": 252716
    }
  },
  {
    "ph": "s", "id": 252716, "pid": 76337, "tid": -914061504, "ts": 1716454225396877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225457387, "dur": 32,
    "args": {
      "External id": 252719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252719,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252719, "pid": 5, "tid": 7, "ts": 1716454225457387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396895, "dur": 6,
    "args": {
      "External id": 252719,
      "cbid": 211, "correlation": 252719
    }
  },
  {
    "ph": "s", "id": 252719, "pid": 76337, "tid": -914061504, "ts": 1716454225396895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225457420, "dur": 22,
    "args": {
      "External id": 252728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252728,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252728, "pid": 5, "tid": 7, "ts": 1716454225457420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396935, "dur": 10,
    "args": {
      "External id": 252728,
      "cbid": 211, "correlation": 252728
    }
  },
  {
    "ph": "s", "id": 252728, "pid": 76337, "tid": -914061504, "ts": 1716454225396935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396994, "dur": 0,
    "args": {
      "External id": 252738,
      "cbid": 317, "correlation": 252738
    }
  },
  {
    "ph": "f", "id": 252738, "pid": 76337, "tid": -914061504, "ts": 1716454225396994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396995, "dur": 0,
    "args": {
      "External id": 252739,
      "cbid": 203, "correlation": 252739
    }
  },
  {
    "ph": "f", "id": 252739, "pid": 76337, "tid": -914061504, "ts": 1716454225396995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225396995, "dur": 0,
    "args": {
      "External id": 252740,
      "cbid": 205, "correlation": 252740
    }
  },
  {
    "ph": "f", "id": 252740, "pid": 76337, "tid": -914061504, "ts": 1716454225396995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225457443, "dur": 22,
    "args": {
      "External id": 252744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252744, "pid": 5, "tid": 7, "ts": 1716454225457443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397009, "dur": 13,
    "args": {
      "External id": 252744,
      "cbid": 211, "correlation": 252744
    }
  },
  {
    "ph": "s", "id": 252744, "pid": 76337, "tid": -914061504, "ts": 1716454225397009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225457467, "dur": 326,
    "args": {
      "External id": 252746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252746,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252746, "pid": 5, "tid": 7, "ts": 1716454225457467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397024, "dur": 5,
    "args": {
      "External id": 252746,
      "cbid": 211, "correlation": 252746
    }
  },
  {
    "ph": "s", "id": 252746, "pid": 76337, "tid": -914061504, "ts": 1716454225397024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225457794, "dur": 1,
    "args": {
      "External id": 252748,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 252748,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 252748, "pid": 5, "tid": 7, "ts": 1716454225457794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397035, "dur": 7,
    "args": {
      "External id": 252748,
      "cbid": 51, "correlation": 252748
    }
  },
  {
    "ph": "s", "id": 252748, "pid": 76337, "tid": -914061504, "ts": 1716454225397035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225457798, "dur": 1278,
    "args": {
      "External id": 252749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252749,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252749, "pid": 5, "tid": 7, "ts": 1716454225457798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397043, "dur": 6,
    "args": {
      "External id": 252749,
      "cbid": 211, "correlation": 252749
    }
  },
  {
    "ph": "s", "id": 252749, "pid": 76337, "tid": -914061504, "ts": 1716454225397043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225459078, "dur": 14,
    "args": {
      "External id": 252751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252751,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252751, "pid": 5, "tid": 7, "ts": 1716454225459078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397054, "dur": 5,
    "args": {
      "External id": 252751,
      "cbid": 211, "correlation": 252751
    }
  },
  {
    "ph": "s", "id": 252751, "pid": 76337, "tid": -914061504, "ts": 1716454225397054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225459093, "dur": 15,
    "args": {
      "External id": 252757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252757,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252757, "pid": 5, "tid": 7, "ts": 1716454225459093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397082, "dur": 8,
    "args": {
      "External id": 252757,
      "cbid": 211, "correlation": 252757
    }
  },
  {
    "ph": "s", "id": 252757, "pid": 76337, "tid": -914061504, "ts": 1716454225397082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225459109, "dur": 3,
    "args": {
      "External id": 252765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252765,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 252765, "pid": 5, "tid": 7, "ts": 1716454225459109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397127, "dur": 9,
    "args": {
      "External id": 252765,
      "cbid": 211, "correlation": 252765
    }
  },
  {
    "ph": "s", "id": 252765, "pid": 76337, "tid": -914061504, "ts": 1716454225397127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397191, "dur": 1,
    "args": {
      "External id": 252781,
      "cbid": 251, "correlation": 252781
    }
  },
  {
    "ph": "f", "id": 252781, "pid": 76337, "tid": -914061504, "ts": 1716454225397191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397196, "dur": 0,
    "args": {
      "External id": 252783,
      "cbid": 251, "correlation": 252783
    }
  },
  {
    "ph": "f", "id": 252783, "pid": 76337, "tid": -914061504, "ts": 1716454225397196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225459114, "dur": 13,
    "args": {
      "External id": 252784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252784,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252784, "pid": 5, "tid": 7, "ts": 1716454225459114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397198, "dur": 11,
    "args": {
      "External id": 252784,
      "cbid": 211, "correlation": 252784
    }
  },
  {
    "ph": "s", "id": 252784, "pid": 76337, "tid": -914061504, "ts": 1716454225397198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225459129, "dur": 5,
    "args": {
      "External id": 252786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252786,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252786, "pid": 5, "tid": 7, "ts": 1716454225459129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397211, "dur": 6,
    "args": {
      "External id": 252786,
      "cbid": 211, "correlation": 252786
    }
  },
  {
    "ph": "s", "id": 252786, "pid": 76337, "tid": -914061504, "ts": 1716454225397211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225459135, "dur": 16,
    "args": {
      "External id": 252796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252796, "pid": 5, "tid": 7, "ts": 1716454225459135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397269, "dur": 13,
    "args": {
      "External id": 252796,
      "cbid": 211, "correlation": 252796
    }
  },
  {
    "ph": "s", "id": 252796, "pid": 76337, "tid": -914061504, "ts": 1716454225397269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225459153, "dur": 17,
    "args": {
      "External id": 252816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252816,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 252816, "pid": 5, "tid": 7, "ts": 1716454225459153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397334, "dur": 10,
    "args": {
      "External id": 252816,
      "cbid": 211, "correlation": 252816
    }
  },
  {
    "ph": "s", "id": 252816, "pid": 76337, "tid": -914061504, "ts": 1716454225397334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225459172, "dur": 4,
    "args": {
      "External id": 252828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252828,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 252828, "pid": 5, "tid": 7, "ts": 1716454225459172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397355, "dur": 6,
    "args": {
      "External id": 252828,
      "cbid": 211, "correlation": 252828
    }
  },
  {
    "ph": "s", "id": 252828, "pid": 76337, "tid": -914061504, "ts": 1716454225397355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225459178, "dur": 17,
    "args": {
      "External id": 252831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252831,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252831, "pid": 5, "tid": 7, "ts": 1716454225459178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397374, "dur": 6,
    "args": {
      "External id": 252831,
      "cbid": 211, "correlation": 252831
    }
  },
  {
    "ph": "s", "id": 252831, "pid": 76337, "tid": -914061504, "ts": 1716454225397374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225459196, "dur": 13,
    "args": {
      "External id": 252840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252840,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252840, "pid": 5, "tid": 7, "ts": 1716454225459196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397414, "dur": 11,
    "args": {
      "External id": 252840,
      "cbid": 211, "correlation": 252840
    }
  },
  {
    "ph": "s", "id": 252840, "pid": 76337, "tid": -914061504, "ts": 1716454225397414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397477, "dur": 0,
    "args": {
      "External id": 252850,
      "cbid": 317, "correlation": 252850
    }
  },
  {
    "ph": "f", "id": 252850, "pid": 76337, "tid": -914061504, "ts": 1716454225397477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397478, "dur": 0,
    "args": {
      "External id": 252851,
      "cbid": 203, "correlation": 252851
    }
  },
  {
    "ph": "f", "id": 252851, "pid": 76337, "tid": -914061504, "ts": 1716454225397478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397479, "dur": 0,
    "args": {
      "External id": 252852,
      "cbid": 205, "correlation": 252852
    }
  },
  {
    "ph": "f", "id": 252852, "pid": 76337, "tid": -914061504, "ts": 1716454225397479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225459211, "dur": 12,
    "args": {
      "External id": 252856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252856, "pid": 5, "tid": 7, "ts": 1716454225459211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397493, "dur": 12,
    "args": {
      "External id": 252856,
      "cbid": 211, "correlation": 252856
    }
  },
  {
    "ph": "s", "id": 252856, "pid": 76337, "tid": -914061504, "ts": 1716454225397493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225459224, "dur": 164,
    "args": {
      "External id": 252858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252858,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252858, "pid": 5, "tid": 7, "ts": 1716454225459224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397507, "dur": 5,
    "args": {
      "External id": 252858,
      "cbid": 211, "correlation": 252858
    }
  },
  {
    "ph": "s", "id": 252858, "pid": 76337, "tid": -914061504, "ts": 1716454225397507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225459391, "dur": 1,
    "args": {
      "External id": 252860,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 252860,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 252860, "pid": 5, "tid": 7, "ts": 1716454225459391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397518, "dur": 7,
    "args": {
      "External id": 252860,
      "cbid": 51, "correlation": 252860
    }
  },
  {
    "ph": "s", "id": 252860, "pid": 76337, "tid": -914061504, "ts": 1716454225397518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225459394, "dur": 660,
    "args": {
      "External id": 252861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252861,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 252861, "pid": 5, "tid": 7, "ts": 1716454225459394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397525, "dur": 6,
    "args": {
      "External id": 252861,
      "cbid": 211, "correlation": 252861
    }
  },
  {
    "ph": "s", "id": 252861, "pid": 76337, "tid": -914061504, "ts": 1716454225397525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225460056, "dur": 12,
    "args": {
      "External id": 252863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252863,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252863, "pid": 5, "tid": 7, "ts": 1716454225460056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397536, "dur": 5,
    "args": {
      "External id": 252863,
      "cbid": 211, "correlation": 252863
    }
  },
  {
    "ph": "s", "id": 252863, "pid": 76337, "tid": -914061504, "ts": 1716454225397536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225460069, "dur": 15,
    "args": {
      "External id": 252869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252869, "pid": 5, "tid": 7, "ts": 1716454225460069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397564, "dur": 10,
    "args": {
      "External id": 252869,
      "cbid": 211, "correlation": 252869
    }
  },
  {
    "ph": "s", "id": 252869, "pid": 76337, "tid": -914061504, "ts": 1716454225397564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397623, "dur": 0,
    "args": {
      "External id": 252879,
      "cbid": 317, "correlation": 252879
    }
  },
  {
    "ph": "f", "id": 252879, "pid": 76337, "tid": -914061504, "ts": 1716454225397623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397624, "dur": 0,
    "args": {
      "External id": 252880,
      "cbid": 203, "correlation": 252880
    }
  },
  {
    "ph": "f", "id": 252880, "pid": 76337, "tid": -914061504, "ts": 1716454225397624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397625, "dur": 0,
    "args": {
      "External id": 252881,
      "cbid": 205, "correlation": 252881
    }
  },
  {
    "ph": "f", "id": 252881, "pid": 76337, "tid": -914061504, "ts": 1716454225397625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225460085, "dur": 21,
    "args": {
      "External id": 252885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252885, "pid": 5, "tid": 7, "ts": 1716454225460085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397636, "dur": 11,
    "args": {
      "External id": 252885,
      "cbid": 211, "correlation": 252885
    }
  },
  {
    "ph": "s", "id": 252885, "pid": 76337, "tid": -914061504, "ts": 1716454225397636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225460108, "dur": 4,
    "args": {
      "External id": 252887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252887, "pid": 5, "tid": 7, "ts": 1716454225460108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397651, "dur": 6,
    "args": {
      "External id": 252887,
      "cbid": 211, "correlation": 252887
    }
  },
  {
    "ph": "s", "id": 252887, "pid": 76337, "tid": -914061504, "ts": 1716454225397651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397660, "dur": 0,
    "args": {
      "External id": 252888,
      "cbid": 51, "correlation": 252888
    }
  },
  {
    "ph": "s", "id": 252888, "pid": 76337, "tid": -914061504, "ts": 1716454225397660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225460114, "dur": 176,
    "args": {
      "External id": 252889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252889,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 252889, "pid": 5, "tid": 7, "ts": 1716454225460114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397661, "dur": 5,
    "args": {
      "External id": 252889,
      "cbid": 211, "correlation": 252889
    }
  },
  {
    "ph": "s", "id": 252889, "pid": 76337, "tid": -914061504, "ts": 1716454225397661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225460291, "dur": 16,
    "args": {
      "External id": 252894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252894,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252894, "pid": 5, "tid": 7, "ts": 1716454225460291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397686, "dur": 8,
    "args": {
      "External id": 252894,
      "cbid": 211, "correlation": 252894
    }
  },
  {
    "ph": "s", "id": 252894, "pid": 76337, "tid": -914061504, "ts": 1716454225397686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225460309, "dur": 13,
    "args": {
      "External id": 252902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252902,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252902, "pid": 5, "tid": 7, "ts": 1716454225460309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397714, "dur": 9,
    "args": {
      "External id": 252902,
      "cbid": 211, "correlation": 252902
    }
  },
  {
    "ph": "s", "id": 252902, "pid": 76337, "tid": -914061504, "ts": 1716454225397714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225460323, "dur": 10,
    "args": {
      "External id": 252910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252910,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252910, "pid": 5, "tid": 7, "ts": 1716454225460323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397743, "dur": 8,
    "args": {
      "External id": 252910,
      "cbid": 211, "correlation": 252910
    }
  },
  {
    "ph": "s", "id": 252910, "pid": 76337, "tid": -914061504, "ts": 1716454225397743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225460334, "dur": 19,
    "args": {
      "External id": 252930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252930,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 252930, "pid": 5, "tid": 7, "ts": 1716454225460334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397825, "dur": 13,
    "args": {
      "External id": 252930,
      "cbid": 211, "correlation": 252930
    }
  },
  {
    "ph": "s", "id": 252930, "pid": 76337, "tid": -914061504, "ts": 1716454225397825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225460355, "dur": 4,
    "args": {
      "External id": 252942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252942,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 252942, "pid": 5, "tid": 7, "ts": 1716454225460355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397848, "dur": 6,
    "args": {
      "External id": 252942,
      "cbid": 211, "correlation": 252942
    }
  },
  {
    "ph": "s", "id": 252942, "pid": 76337, "tid": -914061504, "ts": 1716454225397848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225460361, "dur": 17,
    "args": {
      "External id": 252945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252945, "pid": 5, "tid": 7, "ts": 1716454225460361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397866, "dur": 7,
    "args": {
      "External id": 252945,
      "cbid": 211, "correlation": 252945
    }
  },
  {
    "ph": "s", "id": 252945, "pid": 76337, "tid": -914061504, "ts": 1716454225397866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397923, "dur": 0,
    "args": {
      "External id": 252956,
      "cbid": 317, "correlation": 252956
    }
  },
  {
    "ph": "f", "id": 252956, "pid": 76337, "tid": -914061504, "ts": 1716454225397923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397924, "dur": 0,
    "args": {
      "External id": 252957,
      "cbid": 203, "correlation": 252957
    }
  },
  {
    "ph": "f", "id": 252957, "pid": 76337, "tid": -914061504, "ts": 1716454225397924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397925, "dur": 0,
    "args": {
      "External id": 252958,
      "cbid": 205, "correlation": 252958
    }
  },
  {
    "ph": "f", "id": 252958, "pid": 76337, "tid": -914061504, "ts": 1716454225397925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225460379, "dur": 12,
    "args": {
      "External id": 252962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252962,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252962, "pid": 5, "tid": 7, "ts": 1716454225460379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397939, "dur": 11,
    "args": {
      "External id": 252962,
      "cbid": 211, "correlation": 252962
    }
  },
  {
    "ph": "s", "id": 252962, "pid": 76337, "tid": -914061504, "ts": 1716454225397939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225460393, "dur": 4,
    "args": {
      "External id": 252964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252964,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 252964, "pid": 5, "tid": 7, "ts": 1716454225460393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397954, "dur": 6,
    "args": {
      "External id": 252964,
      "cbid": 211, "correlation": 252964
    }
  },
  {
    "ph": "s", "id": 252964, "pid": 76337, "tid": -914061504, "ts": 1716454225397954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397963, "dur": 0,
    "args": {
      "External id": 252965,
      "cbid": 51, "correlation": 252965
    }
  },
  {
    "ph": "s", "id": 252965, "pid": 76337, "tid": -914061504, "ts": 1716454225397963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225460398, "dur": 93,
    "args": {
      "External id": 252966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252966,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 252966, "pid": 5, "tid": 7, "ts": 1716454225460398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397964, "dur": 5,
    "args": {
      "External id": 252966,
      "cbid": 211, "correlation": 252966
    }
  },
  {
    "ph": "s", "id": 252966, "pid": 76337, "tid": -914061504, "ts": 1716454225397964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225460492, "dur": 16,
    "args": {
      "External id": 252971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252971, "pid": 5, "tid": 7, "ts": 1716454225460492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225397998, "dur": 10,
    "args": {
      "External id": 252971,
      "cbid": 211, "correlation": 252971
    }
  },
  {
    "ph": "s", "id": 252971, "pid": 76337, "tid": -914061504, "ts": 1716454225397998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225460509, "dur": 85,
    "args": {
      "External id": 252980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 252980,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 252980, "pid": 5, "tid": 7, "ts": 1716454225460509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398081, "dur": 14,
    "args": {
      "External id": 252980,
      "cbid": 211, "correlation": 252980
    }
  },
  {
    "ph": "s", "id": 252980, "pid": 76337, "tid": -914061504, "ts": 1716454225398081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225460595, "dur": 31,
    "args": {
      "External id": 253002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253002,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253002, "pid": 5, "tid": 7, "ts": 1716454225460595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398137, "dur": 10,
    "args": {
      "External id": 253002,
      "cbid": 211, "correlation": 253002
    }
  },
  {
    "ph": "s", "id": 253002, "pid": 76337, "tid": -914061504, "ts": 1716454225398137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398225, "dur": 1,
    "args": {
      "External id": 253013,
      "cbid": 251, "correlation": 253013
    }
  },
  {
    "ph": "f", "id": 253013, "pid": 76337, "tid": -914061504, "ts": 1716454225398225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225460627, "dur": 168,
    "args": {
      "External id": 253014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253014,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253014, "pid": 5, "tid": 7, "ts": 1716454225460627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398231, "dur": 13,
    "args": {
      "External id": 253014,
      "cbid": 211, "correlation": 253014
    }
  },
  {
    "ph": "s", "id": 253014, "pid": 76337, "tid": -914061504, "ts": 1716454225398231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398301, "dur": 1,
    "args": {
      "External id": 253025,
      "cbid": 251, "correlation": 253025
    }
  },
  {
    "ph": "f", "id": 253025, "pid": 76337, "tid": -914061504, "ts": 1716454225398301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225460796, "dur": 160,
    "args": {
      "External id": 253026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253026,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253026, "pid": 5, "tid": 7, "ts": 1716454225460796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398305, "dur": 11,
    "args": {
      "External id": 253026,
      "cbid": 211, "correlation": 253026
    }
  },
  {
    "ph": "s", "id": 253026, "pid": 76337, "tid": -914061504, "ts": 1716454225398305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398371, "dur": 1,
    "args": {
      "External id": 253037,
      "cbid": 251, "correlation": 253037
    }
  },
  {
    "ph": "f", "id": 253037, "pid": 76337, "tid": -914061504, "ts": 1716454225398371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225460957, "dur": 161,
    "args": {
      "External id": 253038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253038,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253038, "pid": 5, "tid": 7, "ts": 1716454225460957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398375, "dur": 12,
    "args": {
      "External id": 253038,
      "cbid": 211, "correlation": 253038
    }
  },
  {
    "ph": "s", "id": 253038, "pid": 76337, "tid": -914061504, "ts": 1716454225398375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225461120, "dur": 341,
    "args": {
      "External id": 253063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253063,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253063, "pid": 5, "tid": 7, "ts": 1716454225461120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398460, "dur": 13,
    "args": {
      "External id": 253063,
      "cbid": 211, "correlation": 253063
    }
  },
  {
    "ph": "s", "id": 253063, "pid": 76337, "tid": -914061504, "ts": 1716454225398460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398559, "dur": 1,
    "args": {
      "External id": 253081,
      "cbid": 251, "correlation": 253081
    }
  },
  {
    "ph": "f", "id": 253081, "pid": 76337, "tid": -914061504, "ts": 1716454225398559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225461462, "dur": 169,
    "args": {
      "External id": 253083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253083,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253083, "pid": 5, "tid": 7, "ts": 1716454225461462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398565, "dur": 14,
    "args": {
      "External id": 253083,
      "cbid": 211, "correlation": 253083
    }
  },
  {
    "ph": "s", "id": 253083, "pid": 76337, "tid": -914061504, "ts": 1716454225398565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225461632, "dur": 19,
    "args": {
      "External id": 253091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253091,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253091, "pid": 5, "tid": 7, "ts": 1716454225461632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398635, "dur": 12,
    "args": {
      "External id": 253091,
      "cbid": 211, "correlation": 253091
    }
  },
  {
    "ph": "s", "id": 253091, "pid": 76337, "tid": -914061504, "ts": 1716454225398635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225461653, "dur": 28,
    "args": {
      "External id": 253099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253099,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253099, "pid": 5, "tid": 7, "ts": 1716454225461653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398674, "dur": 9,
    "args": {
      "External id": 253099,
      "cbid": 211, "correlation": 253099
    }
  },
  {
    "ph": "s", "id": 253099, "pid": 76337, "tid": -914061504, "ts": 1716454225398674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225461682, "dur": 19,
    "args": {
      "External id": 253110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253110, "pid": 5, "tid": 7, "ts": 1716454225461682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398746, "dur": 12,
    "args": {
      "External id": 253110,
      "cbid": 211, "correlation": 253110
    }
  },
  {
    "ph": "s", "id": 253110, "pid": 76337, "tid": -914061504, "ts": 1716454225398746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225461702, "dur": 17,
    "args": {
      "External id": 253132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253132,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253132, "pid": 5, "tid": 7, "ts": 1716454225461702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398777, "dur": 8,
    "args": {
      "External id": 253132,
      "cbid": 211, "correlation": 253132
    }
  },
  {
    "ph": "s", "id": 253132, "pid": 76337, "tid": -914061504, "ts": 1716454225398777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398862, "dur": 1,
    "args": {
      "External id": 253143,
      "cbid": 251, "correlation": 253143
    }
  },
  {
    "ph": "f", "id": 253143, "pid": 76337, "tid": -914061504, "ts": 1716454225398862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225461720, "dur": 91,
    "args": {
      "External id": 253144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253144,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253144, "pid": 5, "tid": 7, "ts": 1716454225461720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398867, "dur": 13,
    "args": {
      "External id": 253144,
      "cbid": 211, "correlation": 253144
    }
  },
  {
    "ph": "s", "id": 253144, "pid": 76337, "tid": -914061504, "ts": 1716454225398867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398938, "dur": 1,
    "args": {
      "External id": 253155,
      "cbid": 251, "correlation": 253155
    }
  },
  {
    "ph": "f", "id": 253155, "pid": 76337, "tid": -914061504, "ts": 1716454225398938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398942, "dur": 0,
    "args": {
      "External id": 253156,
      "cbid": 251, "correlation": 253156
    }
  },
  {
    "ph": "f", "id": 253156, "pid": 76337, "tid": -914061504, "ts": 1716454225398942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225461812, "dur": 13,
    "args": {
      "External id": 253157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253157,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253157, "pid": 5, "tid": 7, "ts": 1716454225461812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398944, "dur": 12,
    "args": {
      "External id": 253157,
      "cbid": 211, "correlation": 253157
    }
  },
  {
    "ph": "s", "id": 253157, "pid": 76337, "tid": -914061504, "ts": 1716454225398944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225461826, "dur": 6,
    "args": {
      "External id": 253159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253159,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253159, "pid": 5, "tid": 7, "ts": 1716454225461826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225398958, "dur": 6,
    "args": {
      "External id": 253159,
      "cbid": 211, "correlation": 253159
    }
  },
  {
    "ph": "s", "id": 253159, "pid": 76337, "tid": -914061504, "ts": 1716454225398958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399024, "dur": 1,
    "args": {
      "External id": 253170,
      "cbid": 251, "correlation": 253170
    }
  },
  {
    "ph": "f", "id": 253170, "pid": 76337, "tid": -914061504, "ts": 1716454225399024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399028, "dur": 0,
    "args": {
      "External id": 253171,
      "cbid": 251, "correlation": 253171
    }
  },
  {
    "ph": "f", "id": 253171, "pid": 76337, "tid": -914061504, "ts": 1716454225399028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225461833, "dur": 8,
    "args": {
      "External id": 253172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253172,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253172, "pid": 5, "tid": 7, "ts": 1716454225461833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399029, "dur": 13,
    "args": {
      "External id": 253172,
      "cbid": 211, "correlation": 253172
    }
  },
  {
    "ph": "s", "id": 253172, "pid": 76337, "tid": -914061504, "ts": 1716454225399029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225461843, "dur": 4,
    "args": {
      "External id": 253174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253174,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253174, "pid": 5, "tid": 7, "ts": 1716454225461843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399044, "dur": 5,
    "args": {
      "External id": 253174,
      "cbid": 211, "correlation": 253174
    }
  },
  {
    "ph": "s", "id": 253174, "pid": 76337, "tid": -914061504, "ts": 1716454225399044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225461848, "dur": 56,
    "args": {
      "External id": 253199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253199,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253199, "pid": 5, "tid": 7, "ts": 1716454225461848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399122, "dur": 13,
    "args": {
      "External id": 253199,
      "cbid": 211, "correlation": 253199
    }
  },
  {
    "ph": "s", "id": 253199, "pid": 76337, "tid": -914061504, "ts": 1716454225399122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399221, "dur": 1,
    "args": {
      "External id": 253217,
      "cbid": 251, "correlation": 253217
    }
  },
  {
    "ph": "f", "id": 253217, "pid": 76337, "tid": -914061504, "ts": 1716454225399221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225461905, "dur": 93,
    "args": {
      "External id": 253219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253219,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253219, "pid": 5, "tid": 7, "ts": 1716454225461905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399228, "dur": 14,
    "args": {
      "External id": 253219,
      "cbid": 211, "correlation": 253219
    }
  },
  {
    "ph": "s", "id": 253219, "pid": 76337, "tid": -914061504, "ts": 1716454225399228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225462000, "dur": 9,
    "args": {
      "External id": 253227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253227,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253227, "pid": 5, "tid": 7, "ts": 1716454225462000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399297, "dur": 12,
    "args": {
      "External id": 253227,
      "cbid": 211, "correlation": 253227
    }
  },
  {
    "ph": "s", "id": 253227, "pid": 76337, "tid": -914061504, "ts": 1716454225399297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225462010, "dur": 21,
    "args": {
      "External id": 253235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253235,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253235, "pid": 5, "tid": 7, "ts": 1716454225462010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399338, "dur": 9,
    "args": {
      "External id": 253235,
      "cbid": 211, "correlation": 253235
    }
  },
  {
    "ph": "s", "id": 253235, "pid": 76337, "tid": -914061504, "ts": 1716454225399338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225462033, "dur": 18,
    "args": {
      "External id": 253257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253257,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253257, "pid": 5, "tid": 7, "ts": 1716454225462033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399390, "dur": 10,
    "args": {
      "External id": 253257,
      "cbid": 211, "correlation": 253257
    }
  },
  {
    "ph": "s", "id": 253257, "pid": 76337, "tid": -914061504, "ts": 1716454225399390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399476, "dur": 1,
    "args": {
      "External id": 253273,
      "cbid": 251, "correlation": 253273
    }
  },
  {
    "ph": "f", "id": 253273, "pid": 76337, "tid": -914061504, "ts": 1716454225399476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399481, "dur": 0,
    "args": {
      "External id": 253275,
      "cbid": 251, "correlation": 253275
    }
  },
  {
    "ph": "f", "id": 253275, "pid": 76337, "tid": -914061504, "ts": 1716454225399481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225462052, "dur": 502,
    "args": {
      "External id": 253276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253276,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253276, "pid": 5, "tid": 7, "ts": 1716454225462052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399483, "dur": 14,
    "args": {
      "External id": 253276,
      "cbid": 211, "correlation": 253276
    }
  },
  {
    "ph": "s", "id": 253276, "pid": 76337, "tid": -914061504, "ts": 1716454225399483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225462555, "dur": 66,
    "args": {
      "External id": 253284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253284, "pid": 5, "tid": 7, "ts": 1716454225462555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399549, "dur": 12,
    "args": {
      "External id": 253284,
      "cbid": 211, "correlation": 253284
    }
  },
  {
    "ph": "s", "id": 253284, "pid": 76337, "tid": -914061504, "ts": 1716454225399549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225462622, "dur": 67,
    "args": {
      "External id": 253292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253292, "pid": 5, "tid": 7, "ts": 1716454225462622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399580, "dur": 9,
    "args": {
      "External id": 253292,
      "cbid": 211, "correlation": 253292
    }
  },
  {
    "ph": "s", "id": 253292, "pid": 76337, "tid": -914061504, "ts": 1716454225399580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399661, "dur": 1,
    "args": {
      "External id": 253308,
      "cbid": 251, "correlation": 253308
    }
  },
  {
    "ph": "f", "id": 253308, "pid": 76337, "tid": -914061504, "ts": 1716454225399661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225462691, "dur": 1,
    "args": {
      "External id": 253310,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 253310,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 253310, "pid": 5, "tid": 7, "ts": 1716454225462691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399666, "dur": 9,
    "args": {
      "External id": 253310,
      "cbid": 51, "correlation": 253310
    }
  },
  {
    "ph": "s", "id": 253310, "pid": 76337, "tid": -914061504, "ts": 1716454225399666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225462695, "dur": 273,
    "args": {
      "External id": 253311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253311,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253311, "pid": 5, "tid": 7, "ts": 1716454225462695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399677, "dur": 12,
    "args": {
      "External id": 253311,
      "cbid": 211, "correlation": 253311
    }
  },
  {
    "ph": "s", "id": 253311, "pid": 76337, "tid": -914061504, "ts": 1716454225399677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225462969, "dur": 14,
    "args": {
      "External id": 253319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253319,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253319, "pid": 5, "tid": 7, "ts": 1716454225462969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399719, "dur": 10,
    "args": {
      "External id": 253319,
      "cbid": 211, "correlation": 253319
    }
  },
  {
    "ph": "s", "id": 253319, "pid": 76337, "tid": -914061504, "ts": 1716454225399719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225462984, "dur": 39,
    "args": {
      "External id": 253330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253330,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253330, "pid": 5, "tid": 7, "ts": 1716454225462984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399787, "dur": 12,
    "args": {
      "External id": 253330,
      "cbid": 211, "correlation": 253330
    }
  },
  {
    "ph": "s", "id": 253330, "pid": 76337, "tid": -914061504, "ts": 1716454225399787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399851, "dur": 0,
    "args": {
      "External id": 253342,
      "cbid": 317, "correlation": 253342
    }
  },
  {
    "ph": "f", "id": 253342, "pid": 76337, "tid": -914061504, "ts": 1716454225399851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399852, "dur": 0,
    "args": {
      "External id": 253343,
      "cbid": 203, "correlation": 253343
    }
  },
  {
    "ph": "f", "id": 253343, "pid": 76337, "tid": -914061504, "ts": 1716454225399852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399852, "dur": 0,
    "args": {
      "External id": 253344,
      "cbid": 205, "correlation": 253344
    }
  },
  {
    "ph": "f", "id": 253344, "pid": 76337, "tid": -914061504, "ts": 1716454225399852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225463025, "dur": 13,
    "args": {
      "External id": 253348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253348, "pid": 5, "tid": 7, "ts": 1716454225463025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399866, "dur": 12,
    "args": {
      "External id": 253348,
      "cbid": 211, "correlation": 253348
    }
  },
  {
    "ph": "s", "id": 253348, "pid": 76337, "tid": -914061504, "ts": 1716454225399866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225463039, "dur": 4,
    "args": {
      "External id": 253350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253350, "pid": 5, "tid": 7, "ts": 1716454225463039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399882, "dur": 6,
    "args": {
      "External id": 253350,
      "cbid": 211, "correlation": 253350
    }
  },
  {
    "ph": "s", "id": 253350, "pid": 76337, "tid": -914061504, "ts": 1716454225399882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399892, "dur": 0,
    "args": {
      "External id": 253351,
      "cbid": 51, "correlation": 253351
    }
  },
  {
    "ph": "s", "id": 253351, "pid": 76337, "tid": -914061504, "ts": 1716454225399892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225463045, "dur": 98,
    "args": {
      "External id": 253352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253352,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 253352, "pid": 5, "tid": 7, "ts": 1716454225463045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399893, "dur": 5,
    "args": {
      "External id": 253352,
      "cbid": 211, "correlation": 253352
    }
  },
  {
    "ph": "s", "id": 253352, "pid": 76337, "tid": -914061504, "ts": 1716454225399893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225463144, "dur": 17,
    "args": {
      "External id": 253357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253357, "pid": 5, "tid": 7, "ts": 1716454225463144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399920, "dur": 9,
    "args": {
      "External id": 253357,
      "cbid": 211, "correlation": 253357
    }
  },
  {
    "ph": "s", "id": 253357, "pid": 76337, "tid": -914061504, "ts": 1716454225399920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225463163, "dur": 12,
    "args": {
      "External id": 253365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253365,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253365, "pid": 5, "tid": 7, "ts": 1716454225463163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225399953, "dur": 8,
    "args": {
      "External id": 253365,
      "cbid": 211, "correlation": 253365
    }
  },
  {
    "ph": "s", "id": 253365, "pid": 76337, "tid": -914061504, "ts": 1716454225399953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225463176, "dur": 27,
    "args": {
      "External id": 253374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253374,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253374, "pid": 5, "tid": 7, "ts": 1716454225463176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400000, "dur": 11,
    "args": {
      "External id": 253374,
      "cbid": 211, "correlation": 253374
    }
  },
  {
    "ph": "s", "id": 253374, "pid": 76337, "tid": -914061504, "ts": 1716454225400000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225463204, "dur": 25,
    "args": {
      "External id": 253394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253394,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 253394, "pid": 5, "tid": 7, "ts": 1716454225463204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400073, "dur": 11,
    "args": {
      "External id": 253394,
      "cbid": 211, "correlation": 253394
    }
  },
  {
    "ph": "s", "id": 253394, "pid": 76337, "tid": -914061504, "ts": 1716454225400073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225463231, "dur": 5,
    "args": {
      "External id": 253406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253406,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 253406, "pid": 5, "tid": 7, "ts": 1716454225463231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400095, "dur": 7,
    "args": {
      "External id": 253406,
      "cbid": 211, "correlation": 253406
    }
  },
  {
    "ph": "s", "id": 253406, "pid": 76337, "tid": -914061504, "ts": 1716454225400095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225463237, "dur": 24,
    "args": {
      "External id": 253409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253409, "pid": 5, "tid": 7, "ts": 1716454225463237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400114, "dur": 6,
    "args": {
      "External id": 253409,
      "cbid": 211, "correlation": 253409
    }
  },
  {
    "ph": "s", "id": 253409, "pid": 76337, "tid": -914061504, "ts": 1716454225400114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225463262, "dur": 17,
    "args": {
      "External id": 253418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253418,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253418, "pid": 5, "tid": 7, "ts": 1716454225463262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400153, "dur": 10,
    "args": {
      "External id": 253418,
      "cbid": 211, "correlation": 253418
    }
  },
  {
    "ph": "s", "id": 253418, "pid": 76337, "tid": -914061504, "ts": 1716454225400153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400204, "dur": 0,
    "args": {
      "External id": 253428,
      "cbid": 317, "correlation": 253428
    }
  },
  {
    "ph": "f", "id": 253428, "pid": 76337, "tid": -914061504, "ts": 1716454225400204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400205, "dur": 0,
    "args": {
      "External id": 253429,
      "cbid": 203, "correlation": 253429
    }
  },
  {
    "ph": "f", "id": 253429, "pid": 76337, "tid": -914061504, "ts": 1716454225400205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400206, "dur": 0,
    "args": {
      "External id": 253430,
      "cbid": 205, "correlation": 253430
    }
  },
  {
    "ph": "f", "id": 253430, "pid": 76337, "tid": -914061504, "ts": 1716454225400206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225463281, "dur": 17,
    "args": {
      "External id": 253434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253434,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253434, "pid": 5, "tid": 7, "ts": 1716454225463281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400220, "dur": 11,
    "args": {
      "External id": 253434,
      "cbid": 211, "correlation": 253434
    }
  },
  {
    "ph": "s", "id": 253434, "pid": 76337, "tid": -914061504, "ts": 1716454225400220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225463300, "dur": 246,
    "args": {
      "External id": 253436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253436,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253436, "pid": 5, "tid": 7, "ts": 1716454225463300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400234, "dur": 5,
    "args": {
      "External id": 253436,
      "cbid": 211, "correlation": 253436
    }
  },
  {
    "ph": "s", "id": 253436, "pid": 76337, "tid": -914061504, "ts": 1716454225400234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225463548, "dur": 1,
    "args": {
      "External id": 253438,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 253438,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 253438, "pid": 5, "tid": 7, "ts": 1716454225463548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400246, "dur": 8,
    "args": {
      "External id": 253438,
      "cbid": 51, "correlation": 253438
    }
  },
  {
    "ph": "s", "id": 253438, "pid": 76337, "tid": -914061504, "ts": 1716454225400246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225463551, "dur": 823,
    "args": {
      "External id": 253439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253439,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253439, "pid": 5, "tid": 7, "ts": 1716454225463551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400255, "dur": 6,
    "args": {
      "External id": 253439,
      "cbid": 211, "correlation": 253439
    }
  },
  {
    "ph": "s", "id": 253439, "pid": 76337, "tid": -914061504, "ts": 1716454225400255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225464376, "dur": 14,
    "args": {
      "External id": 253441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253441,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253441, "pid": 5, "tid": 7, "ts": 1716454225464376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400266, "dur": 5,
    "args": {
      "External id": 253441,
      "cbid": 211, "correlation": 253441
    }
  },
  {
    "ph": "s", "id": 253441, "pid": 76337, "tid": -914061504, "ts": 1716454225400266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225464391, "dur": 15,
    "args": {
      "External id": 253447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253447, "pid": 5, "tid": 7, "ts": 1716454225464391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400294, "dur": 8,
    "args": {
      "External id": 253447,
      "cbid": 211, "correlation": 253447
    }
  },
  {
    "ph": "s", "id": 253447, "pid": 76337, "tid": -914061504, "ts": 1716454225400294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225464407, "dur": 3,
    "args": {
      "External id": 253455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253455,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 253455, "pid": 5, "tid": 7, "ts": 1716454225464407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400337, "dur": 10,
    "args": {
      "External id": 253455,
      "cbid": 211, "correlation": 253455
    }
  },
  {
    "ph": "s", "id": 253455, "pid": 76337, "tid": -914061504, "ts": 1716454225400337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400401, "dur": 1,
    "args": {
      "External id": 253471,
      "cbid": 251, "correlation": 253471
    }
  },
  {
    "ph": "f", "id": 253471, "pid": 76337, "tid": -914061504, "ts": 1716454225400401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400407, "dur": 0,
    "args": {
      "External id": 253473,
      "cbid": 251, "correlation": 253473
    }
  },
  {
    "ph": "f", "id": 253473, "pid": 76337, "tid": -914061504, "ts": 1716454225400407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225464412, "dur": 13,
    "args": {
      "External id": 253474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253474,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253474, "pid": 5, "tid": 7, "ts": 1716454225464412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400409, "dur": 11,
    "args": {
      "External id": 253474,
      "cbid": 211, "correlation": 253474
    }
  },
  {
    "ph": "s", "id": 253474, "pid": 76337, "tid": -914061504, "ts": 1716454225400409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225464427, "dur": 5,
    "args": {
      "External id": 253476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253476,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253476, "pid": 5, "tid": 7, "ts": 1716454225464427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400421, "dur": 5,
    "args": {
      "External id": 253476,
      "cbid": 211, "correlation": 253476
    }
  },
  {
    "ph": "s", "id": 253476, "pid": 76337, "tid": -914061504, "ts": 1716454225400421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225464433, "dur": 17,
    "args": {
      "External id": 253486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253486, "pid": 5, "tid": 7, "ts": 1716454225464433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400479, "dur": 13,
    "args": {
      "External id": 253486,
      "cbid": 211, "correlation": 253486
    }
  },
  {
    "ph": "s", "id": 253486, "pid": 76337, "tid": -914061504, "ts": 1716454225400479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225464452, "dur": 18,
    "args": {
      "External id": 253506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253506,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 253506, "pid": 5, "tid": 7, "ts": 1716454225464452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400545, "dur": 10,
    "args": {
      "External id": 253506,
      "cbid": 211, "correlation": 253506
    }
  },
  {
    "ph": "s", "id": 253506, "pid": 76337, "tid": -914061504, "ts": 1716454225400545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225464471, "dur": 4,
    "args": {
      "External id": 253518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253518,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 253518, "pid": 5, "tid": 7, "ts": 1716454225464471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400565, "dur": 6,
    "args": {
      "External id": 253518,
      "cbid": 211, "correlation": 253518
    }
  },
  {
    "ph": "s", "id": 253518, "pid": 76337, "tid": -914061504, "ts": 1716454225400565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225464476, "dur": 17,
    "args": {
      "External id": 253521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253521, "pid": 5, "tid": 7, "ts": 1716454225464476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400584, "dur": 7,
    "args": {
      "External id": 253521,
      "cbid": 211, "correlation": 253521
    }
  },
  {
    "ph": "s", "id": 253521, "pid": 76337, "tid": -914061504, "ts": 1716454225400584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225464494, "dur": 11,
    "args": {
      "External id": 253530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253530,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253530, "pid": 5, "tid": 7, "ts": 1716454225464494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400625, "dur": 10,
    "args": {
      "External id": 253530,
      "cbid": 211, "correlation": 253530
    }
  },
  {
    "ph": "s", "id": 253530, "pid": 76337, "tid": -914061504, "ts": 1716454225400625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400687, "dur": 0,
    "args": {
      "External id": 253540,
      "cbid": 317, "correlation": 253540
    }
  },
  {
    "ph": "f", "id": 253540, "pid": 76337, "tid": -914061504, "ts": 1716454225400687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400688, "dur": 0,
    "args": {
      "External id": 253541,
      "cbid": 203, "correlation": 253541
    }
  },
  {
    "ph": "f", "id": 253541, "pid": 76337, "tid": -914061504, "ts": 1716454225400688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400689, "dur": 0,
    "args": {
      "External id": 253542,
      "cbid": 205, "correlation": 253542
    }
  },
  {
    "ph": "f", "id": 253542, "pid": 76337, "tid": -914061504, "ts": 1716454225400689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225464507, "dur": 11,
    "args": {
      "External id": 253546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253546, "pid": 5, "tid": 7, "ts": 1716454225464507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400704, "dur": 12,
    "args": {
      "External id": 253546,
      "cbid": 211, "correlation": 253546
    }
  },
  {
    "ph": "s", "id": 253546, "pid": 76337, "tid": -914061504, "ts": 1716454225400704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225464519, "dur": 166,
    "args": {
      "External id": 253548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253548,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253548, "pid": 5, "tid": 7, "ts": 1716454225464519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400718, "dur": 5,
    "args": {
      "External id": 253548,
      "cbid": 211, "correlation": 253548
    }
  },
  {
    "ph": "s", "id": 253548, "pid": 76337, "tid": -914061504, "ts": 1716454225400718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225464687, "dur": 1,
    "args": {
      "External id": 253550,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 253550,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 253550, "pid": 5, "tid": 7, "ts": 1716454225464687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400729, "dur": 6,
    "args": {
      "External id": 253550,
      "cbid": 51, "correlation": 253550
    }
  },
  {
    "ph": "s", "id": 253550, "pid": 76337, "tid": -914061504, "ts": 1716454225400729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225464691, "dur": 659,
    "args": {
      "External id": 253551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253551,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253551, "pid": 5, "tid": 7, "ts": 1716454225464691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400736, "dur": 6,
    "args": {
      "External id": 253551,
      "cbid": 211, "correlation": 253551
    }
  },
  {
    "ph": "s", "id": 253551, "pid": 76337, "tid": -914061504, "ts": 1716454225400736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225465351, "dur": 13,
    "args": {
      "External id": 253553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253553, "pid": 5, "tid": 7, "ts": 1716454225465351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400747, "dur": 5,
    "args": {
      "External id": 253553,
      "cbid": 211, "correlation": 253553
    }
  },
  {
    "ph": "s", "id": 253553, "pid": 76337, "tid": -914061504, "ts": 1716454225400747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225465366, "dur": 15,
    "args": {
      "External id": 253559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253559, "pid": 5, "tid": 7, "ts": 1716454225465366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400775, "dur": 9,
    "args": {
      "External id": 253559,
      "cbid": 211, "correlation": 253559
    }
  },
  {
    "ph": "s", "id": 253559, "pid": 76337, "tid": -914061504, "ts": 1716454225400775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400833, "dur": 0,
    "args": {
      "External id": 253569,
      "cbid": 317, "correlation": 253569
    }
  },
  {
    "ph": "f", "id": 253569, "pid": 76337, "tid": -914061504, "ts": 1716454225400833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400834, "dur": 0,
    "args": {
      "External id": 253570,
      "cbid": 203, "correlation": 253570
    }
  },
  {
    "ph": "f", "id": 253570, "pid": 76337, "tid": -914061504, "ts": 1716454225400834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400835, "dur": 0,
    "args": {
      "External id": 253571,
      "cbid": 205, "correlation": 253571
    }
  },
  {
    "ph": "f", "id": 253571, "pid": 76337, "tid": -914061504, "ts": 1716454225400835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225465383, "dur": 18,
    "args": {
      "External id": 253575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253575,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253575, "pid": 5, "tid": 7, "ts": 1716454225465383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400849, "dur": 12,
    "args": {
      "External id": 253575,
      "cbid": 211, "correlation": 253575
    }
  },
  {
    "ph": "s", "id": 253575, "pid": 76337, "tid": -914061504, "ts": 1716454225400849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225465402, "dur": 4,
    "args": {
      "External id": 253577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253577, "pid": 5, "tid": 7, "ts": 1716454225465402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400865, "dur": 6,
    "args": {
      "External id": 253577,
      "cbid": 211, "correlation": 253577
    }
  },
  {
    "ph": "s", "id": 253577, "pid": 76337, "tid": -914061504, "ts": 1716454225400865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400874, "dur": 0,
    "args": {
      "External id": 253578,
      "cbid": 51, "correlation": 253578
    }
  },
  {
    "ph": "s", "id": 253578, "pid": 76337, "tid": -914061504, "ts": 1716454225400874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225465407, "dur": 135,
    "args": {
      "External id": 253579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253579,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 253579, "pid": 5, "tid": 7, "ts": 1716454225465407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400875, "dur": 5,
    "args": {
      "External id": 253579,
      "cbid": 211, "correlation": 253579
    }
  },
  {
    "ph": "s", "id": 253579, "pid": 76337, "tid": -914061504, "ts": 1716454225400875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225465544, "dur": 16,
    "args": {
      "External id": 253584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253584,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253584, "pid": 5, "tid": 7, "ts": 1716454225465544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400901, "dur": 8,
    "args": {
      "External id": 253584,
      "cbid": 211, "correlation": 253584
    }
  },
  {
    "ph": "s", "id": 253584, "pid": 76337, "tid": -914061504, "ts": 1716454225400901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225465561, "dur": 12,
    "args": {
      "External id": 253592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253592,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253592, "pid": 5, "tid": 7, "ts": 1716454225465561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400929, "dur": 8,
    "args": {
      "External id": 253592,
      "cbid": 211, "correlation": 253592
    }
  },
  {
    "ph": "s", "id": 253592, "pid": 76337, "tid": -914061504, "ts": 1716454225400929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225465575, "dur": 11,
    "args": {
      "External id": 253600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253600,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253600, "pid": 5, "tid": 7, "ts": 1716454225465575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225400958, "dur": 8,
    "args": {
      "External id": 253600,
      "cbid": 211, "correlation": 253600
    }
  },
  {
    "ph": "s", "id": 253600, "pid": 76337, "tid": -914061504, "ts": 1716454225400958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225465586, "dur": 19,
    "args": {
      "External id": 253620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253620,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 253620, "pid": 5, "tid": 7, "ts": 1716454225465586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401051, "dur": 13,
    "args": {
      "External id": 253620,
      "cbid": 211, "correlation": 253620
    }
  },
  {
    "ph": "s", "id": 253620, "pid": 76337, "tid": -914061504, "ts": 1716454225401051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225465607, "dur": 4,
    "args": {
      "External id": 253632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253632,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 253632, "pid": 5, "tid": 7, "ts": 1716454225465607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401073, "dur": 6,
    "args": {
      "External id": 253632,
      "cbid": 211, "correlation": 253632
    }
  },
  {
    "ph": "s", "id": 253632, "pid": 76337, "tid": -914061504, "ts": 1716454225401073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225465612, "dur": 17,
    "args": {
      "External id": 253635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253635, "pid": 5, "tid": 7, "ts": 1716454225465612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401091, "dur": 7,
    "args": {
      "External id": 253635,
      "cbid": 211, "correlation": 253635
    }
  },
  {
    "ph": "s", "id": 253635, "pid": 76337, "tid": -914061504, "ts": 1716454225401091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401150, "dur": 0,
    "args": {
      "External id": 253646,
      "cbid": 317, "correlation": 253646
    }
  },
  {
    "ph": "f", "id": 253646, "pid": 76337, "tid": -914061504, "ts": 1716454225401150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401151, "dur": 0,
    "args": {
      "External id": 253647,
      "cbid": 203, "correlation": 253647
    }
  },
  {
    "ph": "f", "id": 253647, "pid": 76337, "tid": -914061504, "ts": 1716454225401151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401152, "dur": 0,
    "args": {
      "External id": 253648,
      "cbid": 205, "correlation": 253648
    }
  },
  {
    "ph": "f", "id": 253648, "pid": 76337, "tid": -914061504, "ts": 1716454225401152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225465631, "dur": 12,
    "args": {
      "External id": 253652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253652,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253652, "pid": 5, "tid": 7, "ts": 1716454225465631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401165, "dur": 12,
    "args": {
      "External id": 253652,
      "cbid": 211, "correlation": 253652
    }
  },
  {
    "ph": "s", "id": 253652, "pid": 76337, "tid": -914061504, "ts": 1716454225401165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225465644, "dur": 3,
    "args": {
      "External id": 253654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253654,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253654, "pid": 5, "tid": 7, "ts": 1716454225465644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401181, "dur": 5,
    "args": {
      "External id": 253654,
      "cbid": 211, "correlation": 253654
    }
  },
  {
    "ph": "s", "id": 253654, "pid": 76337, "tid": -914061504, "ts": 1716454225401181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401189, "dur": 0,
    "args": {
      "External id": 253655,
      "cbid": 51, "correlation": 253655
    }
  },
  {
    "ph": "s", "id": 253655, "pid": 76337, "tid": -914061504, "ts": 1716454225401189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225465649, "dur": 92,
    "args": {
      "External id": 253656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253656,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 253656, "pid": 5, "tid": 7, "ts": 1716454225465649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401190, "dur": 5,
    "args": {
      "External id": 253656,
      "cbid": 211, "correlation": 253656
    }
  },
  {
    "ph": "s", "id": 253656, "pid": 76337, "tid": -914061504, "ts": 1716454225401190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225465742, "dur": 16,
    "args": {
      "External id": 253661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253661, "pid": 5, "tid": 7, "ts": 1716454225465742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401217, "dur": 8,
    "args": {
      "External id": 253661,
      "cbid": 211, "correlation": 253661
    }
  },
  {
    "ph": "s", "id": 253661, "pid": 76337, "tid": -914061504, "ts": 1716454225401217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225465759, "dur": 86,
    "args": {
      "External id": 253670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253670,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253670, "pid": 5, "tid": 7, "ts": 1716454225465759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401298, "dur": 13,
    "args": {
      "External id": 253670,
      "cbid": 211, "correlation": 253670
    }
  },
  {
    "ph": "s", "id": 253670, "pid": 76337, "tid": -914061504, "ts": 1716454225401298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225465846, "dur": 31,
    "args": {
      "External id": 253692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253692,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253692, "pid": 5, "tid": 7, "ts": 1716454225465846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401356, "dur": 10,
    "args": {
      "External id": 253692,
      "cbid": 211, "correlation": 253692
    }
  },
  {
    "ph": "s", "id": 253692, "pid": 76337, "tid": -914061504, "ts": 1716454225401356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401444, "dur": 2,
    "args": {
      "External id": 253703,
      "cbid": 251, "correlation": 253703
    }
  },
  {
    "ph": "f", "id": 253703, "pid": 76337, "tid": -914061504, "ts": 1716454225401444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225465879, "dur": 171,
    "args": {
      "External id": 253704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253704,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253704, "pid": 5, "tid": 7, "ts": 1716454225465879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401450, "dur": 13,
    "args": {
      "External id": 253704,
      "cbid": 211, "correlation": 253704
    }
  },
  {
    "ph": "s", "id": 253704, "pid": 76337, "tid": -914061504, "ts": 1716454225401450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401519, "dur": 1,
    "args": {
      "External id": 253715,
      "cbid": 251, "correlation": 253715
    }
  },
  {
    "ph": "f", "id": 253715, "pid": 76337, "tid": -914061504, "ts": 1716454225401519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225466051, "dur": 160,
    "args": {
      "External id": 253716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253716,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253716, "pid": 5, "tid": 7, "ts": 1716454225466051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401523, "dur": 11,
    "args": {
      "External id": 253716,
      "cbid": 211, "correlation": 253716
    }
  },
  {
    "ph": "s", "id": 253716, "pid": 76337, "tid": -914061504, "ts": 1716454225401523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401587, "dur": 1,
    "args": {
      "External id": 253727,
      "cbid": 251, "correlation": 253727
    }
  },
  {
    "ph": "f", "id": 253727, "pid": 76337, "tid": -914061504, "ts": 1716454225401587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225466213, "dur": 160,
    "args": {
      "External id": 253728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253728,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253728, "pid": 5, "tid": 7, "ts": 1716454225466213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401590, "dur": 12,
    "args": {
      "External id": 253728,
      "cbid": 211, "correlation": 253728
    }
  },
  {
    "ph": "s", "id": 253728, "pid": 76337, "tid": -914061504, "ts": 1716454225401590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225466374, "dur": 342,
    "args": {
      "External id": 253753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253753,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253753, "pid": 5, "tid": 7, "ts": 1716454225466374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401675, "dur": 12,
    "args": {
      "External id": 253753,
      "cbid": 211, "correlation": 253753
    }
  },
  {
    "ph": "s", "id": 253753, "pid": 76337, "tid": -914061504, "ts": 1716454225401675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401774, "dur": 1,
    "args": {
      "External id": 253771,
      "cbid": 251, "correlation": 253771
    }
  },
  {
    "ph": "f", "id": 253771, "pid": 76337, "tid": -914061504, "ts": 1716454225401774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225466718, "dur": 170,
    "args": {
      "External id": 253773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253773,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253773, "pid": 5, "tid": 7, "ts": 1716454225466718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401779, "dur": 13,
    "args": {
      "External id": 253773,
      "cbid": 211, "correlation": 253773
    }
  },
  {
    "ph": "s", "id": 253773, "pid": 76337, "tid": -914061504, "ts": 1716454225401779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225466889, "dur": 19,
    "args": {
      "External id": 253781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253781,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253781, "pid": 5, "tid": 7, "ts": 1716454225466889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401849, "dur": 12,
    "args": {
      "External id": 253781,
      "cbid": 211, "correlation": 253781
    }
  },
  {
    "ph": "s", "id": 253781, "pid": 76337, "tid": -914061504, "ts": 1716454225401849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225466910, "dur": 28,
    "args": {
      "External id": 253789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253789, "pid": 5, "tid": 7, "ts": 1716454225466910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401887, "dur": 9,
    "args": {
      "External id": 253789,
      "cbid": 211, "correlation": 253789
    }
  },
  {
    "ph": "s", "id": 253789, "pid": 76337, "tid": -914061504, "ts": 1716454225401887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225466939, "dur": 18,
    "args": {
      "External id": 253800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253800, "pid": 5, "tid": 7, "ts": 1716454225466939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401959, "dur": 13,
    "args": {
      "External id": 253800,
      "cbid": 211, "correlation": 253800
    }
  },
  {
    "ph": "s", "id": 253800, "pid": 76337, "tid": -914061504, "ts": 1716454225401959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225466958, "dur": 17,
    "args": {
      "External id": 253822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253822,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253822, "pid": 5, "tid": 7, "ts": 1716454225466958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225401999, "dur": 9,
    "args": {
      "External id": 253822,
      "cbid": 211, "correlation": 253822
    }
  },
  {
    "ph": "s", "id": 253822, "pid": 76337, "tid": -914061504, "ts": 1716454225401999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402086, "dur": 1,
    "args": {
      "External id": 253833,
      "cbid": 251, "correlation": 253833
    }
  },
  {
    "ph": "f", "id": 253833, "pid": 76337, "tid": -914061504, "ts": 1716454225402086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225466976, "dur": 91,
    "args": {
      "External id": 253834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253834,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253834, "pid": 5, "tid": 7, "ts": 1716454225466976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402091, "dur": 14,
    "args": {
      "External id": 253834,
      "cbid": 211, "correlation": 253834
    }
  },
  {
    "ph": "s", "id": 253834, "pid": 76337, "tid": -914061504, "ts": 1716454225402091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402161, "dur": 1,
    "args": {
      "External id": 253845,
      "cbid": 251, "correlation": 253845
    }
  },
  {
    "ph": "f", "id": 253845, "pid": 76337, "tid": -914061504, "ts": 1716454225402161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402164, "dur": 0,
    "args": {
      "External id": 253846,
      "cbid": 251, "correlation": 253846
    }
  },
  {
    "ph": "f", "id": 253846, "pid": 76337, "tid": -914061504, "ts": 1716454225402164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225467069, "dur": 12,
    "args": {
      "External id": 253847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253847,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253847, "pid": 5, "tid": 7, "ts": 1716454225467069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402166, "dur": 12,
    "args": {
      "External id": 253847,
      "cbid": 211, "correlation": 253847
    }
  },
  {
    "ph": "s", "id": 253847, "pid": 76337, "tid": -914061504, "ts": 1716454225402166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225467082, "dur": 6,
    "args": {
      "External id": 253849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253849,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253849, "pid": 5, "tid": 7, "ts": 1716454225467082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402179, "dur": 6,
    "args": {
      "External id": 253849,
      "cbid": 211, "correlation": 253849
    }
  },
  {
    "ph": "s", "id": 253849, "pid": 76337, "tid": -914061504, "ts": 1716454225402179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402238, "dur": 1,
    "args": {
      "External id": 253860,
      "cbid": 251, "correlation": 253860
    }
  },
  {
    "ph": "f", "id": 253860, "pid": 76337, "tid": -914061504, "ts": 1716454225402238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402241, "dur": 0,
    "args": {
      "External id": 253861,
      "cbid": 251, "correlation": 253861
    }
  },
  {
    "ph": "f", "id": 253861, "pid": 76337, "tid": -914061504, "ts": 1716454225402241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225467089, "dur": 9,
    "args": {
      "External id": 253862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253862,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253862, "pid": 5, "tid": 7, "ts": 1716454225467089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402243, "dur": 12,
    "args": {
      "External id": 253862,
      "cbid": 211, "correlation": 253862
    }
  },
  {
    "ph": "s", "id": 253862, "pid": 76337, "tid": -914061504, "ts": 1716454225402243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225467099, "dur": 4,
    "args": {
      "External id": 253864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253864,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253864, "pid": 5, "tid": 7, "ts": 1716454225467099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402257, "dur": 5,
    "args": {
      "External id": 253864,
      "cbid": 211, "correlation": 253864
    }
  },
  {
    "ph": "s", "id": 253864, "pid": 76337, "tid": -914061504, "ts": 1716454225402257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225467104, "dur": 57,
    "args": {
      "External id": 253889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253889,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253889, "pid": 5, "tid": 7, "ts": 1716454225467104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402333, "dur": 12,
    "args": {
      "External id": 253889,
      "cbid": 211, "correlation": 253889
    }
  },
  {
    "ph": "s", "id": 253889, "pid": 76337, "tid": -914061504, "ts": 1716454225402333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402433, "dur": 1,
    "args": {
      "External id": 253907,
      "cbid": 251, "correlation": 253907
    }
  },
  {
    "ph": "f", "id": 253907, "pid": 76337, "tid": -914061504, "ts": 1716454225402433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225467162, "dur": 94,
    "args": {
      "External id": 253909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253909,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 253909, "pid": 5, "tid": 7, "ts": 1716454225467162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402439, "dur": 15,
    "args": {
      "External id": 253909,
      "cbid": 211, "correlation": 253909
    }
  },
  {
    "ph": "s", "id": 253909, "pid": 76337, "tid": -914061504, "ts": 1716454225402439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225467257, "dur": 10,
    "args": {
      "External id": 253917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253917,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253917, "pid": 5, "tid": 7, "ts": 1716454225467257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402509, "dur": 12,
    "args": {
      "External id": 253917,
      "cbid": 211, "correlation": 253917
    }
  },
  {
    "ph": "s", "id": 253917, "pid": 76337, "tid": -914061504, "ts": 1716454225402509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225467268, "dur": 21,
    "args": {
      "External id": 253925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253925, "pid": 5, "tid": 7, "ts": 1716454225467268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402551, "dur": 10,
    "args": {
      "External id": 253925,
      "cbid": 211, "correlation": 253925
    }
  },
  {
    "ph": "s", "id": 253925, "pid": 76337, "tid": -914061504, "ts": 1716454225402551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225467291, "dur": 18,
    "args": {
      "External id": 253947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253947,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253947, "pid": 5, "tid": 7, "ts": 1716454225467291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402602, "dur": 10,
    "args": {
      "External id": 253947,
      "cbid": 211, "correlation": 253947
    }
  },
  {
    "ph": "s", "id": 253947, "pid": 76337, "tid": -914061504, "ts": 1716454225402602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402689, "dur": 1,
    "args": {
      "External id": 253963,
      "cbid": 251, "correlation": 253963
    }
  },
  {
    "ph": "f", "id": 253963, "pid": 76337, "tid": -914061504, "ts": 1716454225402689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402694, "dur": 0,
    "args": {
      "External id": 253965,
      "cbid": 251, "correlation": 253965
    }
  },
  {
    "ph": "f", "id": 253965, "pid": 76337, "tid": -914061504, "ts": 1716454225402694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225467310, "dur": 501,
    "args": {
      "External id": 253966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253966,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 253966, "pid": 5, "tid": 7, "ts": 1716454225467310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402696, "dur": 13,
    "args": {
      "External id": 253966,
      "cbid": 211, "correlation": 253966
    }
  },
  {
    "ph": "s", "id": 253966, "pid": 76337, "tid": -914061504, "ts": 1716454225402696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225467812, "dur": 67,
    "args": {
      "External id": 253974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253974,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253974, "pid": 5, "tid": 7, "ts": 1716454225467812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402762, "dur": 13,
    "args": {
      "External id": 253974,
      "cbid": 211, "correlation": 253974
    }
  },
  {
    "ph": "s", "id": 253974, "pid": 76337, "tid": -914061504, "ts": 1716454225402762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225467880, "dur": 66,
    "args": {
      "External id": 253982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 253982,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 253982, "pid": 5, "tid": 7, "ts": 1716454225467880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402793, "dur": 9,
    "args": {
      "External id": 253982,
      "cbid": 211, "correlation": 253982
    }
  },
  {
    "ph": "s", "id": 253982, "pid": 76337, "tid": -914061504, "ts": 1716454225402793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402872, "dur": 1,
    "args": {
      "External id": 253998,
      "cbid": 251, "correlation": 253998
    }
  },
  {
    "ph": "f", "id": 253998, "pid": 76337, "tid": -914061504, "ts": 1716454225402872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225467949, "dur": 1,
    "args": {
      "External id": 254000,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 254000,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 254000, "pid": 5, "tid": 7, "ts": 1716454225467949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402876, "dur": 9,
    "args": {
      "External id": 254000,
      "cbid": 51, "correlation": 254000
    }
  },
  {
    "ph": "s", "id": 254000, "pid": 76337, "tid": -914061504, "ts": 1716454225402876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225467952, "dur": 275,
    "args": {
      "External id": 254001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254001,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 254001, "pid": 5, "tid": 7, "ts": 1716454225467952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402887, "dur": 11,
    "args": {
      "External id": 254001,
      "cbid": 211, "correlation": 254001
    }
  },
  {
    "ph": "s", "id": 254001, "pid": 76337, "tid": -914061504, "ts": 1716454225402887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225468229, "dur": 14,
    "args": {
      "External id": 254009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254009,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254009, "pid": 5, "tid": 7, "ts": 1716454225468229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225402929, "dur": 10,
    "args": {
      "External id": 254009,
      "cbid": 211, "correlation": 254009
    }
  },
  {
    "ph": "s", "id": 254009, "pid": 76337, "tid": -914061504, "ts": 1716454225402929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225468244, "dur": 38,
    "args": {
      "External id": 254020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254020,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254020, "pid": 5, "tid": 7, "ts": 1716454225468244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403005, "dur": 13,
    "args": {
      "External id": 254020,
      "cbid": 211, "correlation": 254020
    }
  },
  {
    "ph": "s", "id": 254020, "pid": 76337, "tid": -914061504, "ts": 1716454225403005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403070, "dur": 0,
    "args": {
      "External id": 254032,
      "cbid": 317, "correlation": 254032
    }
  },
  {
    "ph": "f", "id": 254032, "pid": 76337, "tid": -914061504, "ts": 1716454225403070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403071, "dur": 0,
    "args": {
      "External id": 254033,
      "cbid": 203, "correlation": 254033
    }
  },
  {
    "ph": "f", "id": 254033, "pid": 76337, "tid": -914061504, "ts": 1716454225403071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403072, "dur": 0,
    "args": {
      "External id": 254034,
      "cbid": 205, "correlation": 254034
    }
  },
  {
    "ph": "f", "id": 254034, "pid": 76337, "tid": -914061504, "ts": 1716454225403072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225468283, "dur": 13,
    "args": {
      "External id": 254038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254038,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254038, "pid": 5, "tid": 7, "ts": 1716454225468283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403087, "dur": 12,
    "args": {
      "External id": 254038,
      "cbid": 211, "correlation": 254038
    }
  },
  {
    "ph": "s", "id": 254038, "pid": 76337, "tid": -914061504, "ts": 1716454225403087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225468298, "dur": 4,
    "args": {
      "External id": 254040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 254040, "pid": 5, "tid": 7, "ts": 1716454225468298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403104, "dur": 6,
    "args": {
      "External id": 254040,
      "cbid": 211, "correlation": 254040
    }
  },
  {
    "ph": "s", "id": 254040, "pid": 76337, "tid": -914061504, "ts": 1716454225403104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403113, "dur": 0,
    "args": {
      "External id": 254041,
      "cbid": 51, "correlation": 254041
    }
  },
  {
    "ph": "s", "id": 254041, "pid": 76337, "tid": -914061504, "ts": 1716454225403113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225468304, "dur": 99,
    "args": {
      "External id": 254042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254042,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 254042, "pid": 5, "tid": 7, "ts": 1716454225468304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403114, "dur": 5,
    "args": {
      "External id": 254042,
      "cbid": 211, "correlation": 254042
    }
  },
  {
    "ph": "s", "id": 254042, "pid": 76337, "tid": -914061504, "ts": 1716454225403114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225468404, "dur": 16,
    "args": {
      "External id": 254047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254047, "pid": 5, "tid": 7, "ts": 1716454225468404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403141, "dur": 9,
    "args": {
      "External id": 254047,
      "cbid": 211, "correlation": 254047
    }
  },
  {
    "ph": "s", "id": 254047, "pid": 76337, "tid": -914061504, "ts": 1716454225403141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225468422, "dur": 12,
    "args": {
      "External id": 254055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254055,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254055, "pid": 5, "tid": 7, "ts": 1716454225468422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403173, "dur": 8,
    "args": {
      "External id": 254055,
      "cbid": 211, "correlation": 254055
    }
  },
  {
    "ph": "s", "id": 254055, "pid": 76337, "tid": -914061504, "ts": 1716454225403173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225468435, "dur": 58,
    "args": {
      "External id": 254066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254066,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254066, "pid": 5, "tid": 7, "ts": 1716454225468435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403234, "dur": 11,
    "args": {
      "External id": 254066,
      "cbid": 211, "correlation": 254066
    }
  },
  {
    "ph": "s", "id": 254066, "pid": 76337, "tid": -914061504, "ts": 1716454225403234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403289, "dur": 0,
    "args": {
      "External id": 254076,
      "cbid": 317, "correlation": 254076
    }
  },
  {
    "ph": "f", "id": 254076, "pid": 76337, "tid": -914061504, "ts": 1716454225403289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403290, "dur": 0,
    "args": {
      "External id": 254077,
      "cbid": 203, "correlation": 254077
    }
  },
  {
    "ph": "f", "id": 254077, "pid": 76337, "tid": -914061504, "ts": 1716454225403290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403291, "dur": 0,
    "args": {
      "External id": 254078,
      "cbid": 205, "correlation": 254078
    }
  },
  {
    "ph": "f", "id": 254078, "pid": 76337, "tid": -914061504, "ts": 1716454225403291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225468494, "dur": 39,
    "args": {
      "External id": 254082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254082, "pid": 5, "tid": 7, "ts": 1716454225468494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403307, "dur": 11,
    "args": {
      "External id": 254082,
      "cbid": 211, "correlation": 254082
    }
  },
  {
    "ph": "s", "id": 254082, "pid": 76337, "tid": -914061504, "ts": 1716454225403307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225468535, "dur": 166,
    "args": {
      "External id": 254084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254084, "pid": 5, "tid": 7, "ts": 1716454225468535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403321, "dur": 5,
    "args": {
      "External id": 254084,
      "cbid": 211, "correlation": 254084
    }
  },
  {
    "ph": "s", "id": 254084, "pid": 76337, "tid": -914061504, "ts": 1716454225403321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225468702, "dur": 1968,
    "args": {
      "External id": 254086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254086,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254086, "pid": 5, "tid": 7, "ts": 1716454225468702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403332, "dur": 7,
    "args": {
      "External id": 254086,
      "cbid": 211, "correlation": 254086
    }
  },
  {
    "ph": "s", "id": 254086, "pid": 76337, "tid": -914061504, "ts": 1716454225403332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225470671, "dur": 38,
    "args": {
      "External id": 254088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254088, "pid": 5, "tid": 7, "ts": 1716454225470671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403344, "dur": 5,
    "args": {
      "External id": 254088,
      "cbid": 211, "correlation": 254088
    }
  },
  {
    "ph": "s", "id": 254088, "pid": 76337, "tid": -914061504, "ts": 1716454225403344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225470711, "dur": 60,
    "args": {
      "External id": 254094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254094, "pid": 5, "tid": 7, "ts": 1716454225470711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403371, "dur": 9,
    "args": {
      "External id": 254094,
      "cbid": 211, "correlation": 254094
    }
  },
  {
    "ph": "s", "id": 254094, "pid": 76337, "tid": -914061504, "ts": 1716454225403371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225470771, "dur": 85,
    "args": {
      "External id": 254103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254103,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254103, "pid": 5, "tid": 7, "ts": 1716454225470771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403461, "dur": 13,
    "args": {
      "External id": 254103,
      "cbid": 211, "correlation": 254103
    }
  },
  {
    "ph": "s", "id": 254103, "pid": 76337, "tid": -914061504, "ts": 1716454225403461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225470858, "dur": 74,
    "args": {
      "External id": 254123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254123,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 254123, "pid": 5, "tid": 7, "ts": 1716454225470858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403530, "dur": 11,
    "args": {
      "External id": 254123,
      "cbid": 211, "correlation": 254123
    }
  },
  {
    "ph": "s", "id": 254123, "pid": 76337, "tid": -914061504, "ts": 1716454225403530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225470933, "dur": 5,
    "args": {
      "External id": 254135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254135,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 254135, "pid": 5, "tid": 7, "ts": 1716454225470933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403551, "dur": 6,
    "args": {
      "External id": 254135,
      "cbid": 211, "correlation": 254135
    }
  },
  {
    "ph": "s", "id": 254135, "pid": 76337, "tid": -914061504, "ts": 1716454225403551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225470939, "dur": 82,
    "args": {
      "External id": 254138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254138, "pid": 5, "tid": 7, "ts": 1716454225470939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403570, "dur": 7,
    "args": {
      "External id": 254138,
      "cbid": 211, "correlation": 254138
    }
  },
  {
    "ph": "s", "id": 254138, "pid": 76337, "tid": -914061504, "ts": 1716454225403570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225471022, "dur": 53,
    "args": {
      "External id": 254147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254147,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254147, "pid": 5, "tid": 7, "ts": 1716454225471022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403610, "dur": 10,
    "args": {
      "External id": 254147,
      "cbid": 211, "correlation": 254147
    }
  },
  {
    "ph": "s", "id": 254147, "pid": 76337, "tid": -914061504, "ts": 1716454225403610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403662, "dur": 0,
    "args": {
      "External id": 254157,
      "cbid": 317, "correlation": 254157
    }
  },
  {
    "ph": "f", "id": 254157, "pid": 76337, "tid": -914061504, "ts": 1716454225403662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403663, "dur": 0,
    "args": {
      "External id": 254158,
      "cbid": 203, "correlation": 254158
    }
  },
  {
    "ph": "f", "id": 254158, "pid": 76337, "tid": -914061504, "ts": 1716454225403663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403664, "dur": 0,
    "args": {
      "External id": 254159,
      "cbid": 205, "correlation": 254159
    }
  },
  {
    "ph": "f", "id": 254159, "pid": 76337, "tid": -914061504, "ts": 1716454225403664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225471076, "dur": 56,
    "args": {
      "External id": 254163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254163,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254163, "pid": 5, "tid": 7, "ts": 1716454225471076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403681, "dur": 11,
    "args": {
      "External id": 254163,
      "cbid": 211, "correlation": 254163
    }
  },
  {
    "ph": "s", "id": 254163, "pid": 76337, "tid": -914061504, "ts": 1716454225403681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225471134, "dur": 125,
    "args": {
      "External id": 254165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254165,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254165, "pid": 5, "tid": 7, "ts": 1716454225471134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403694, "dur": 5,
    "args": {
      "External id": 254165,
      "cbid": 211, "correlation": 254165
    }
  },
  {
    "ph": "s", "id": 254165, "pid": 76337, "tid": -914061504, "ts": 1716454225403694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225471260, "dur": 1922,
    "args": {
      "External id": 254167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254167,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254167, "pid": 5, "tid": 7, "ts": 1716454225471260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403706, "dur": 6,
    "args": {
      "External id": 254167,
      "cbid": 211, "correlation": 254167
    }
  },
  {
    "ph": "s", "id": 254167, "pid": 76337, "tid": -914061504, "ts": 1716454225403706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225473183, "dur": 19,
    "args": {
      "External id": 254169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254169,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254169, "pid": 5, "tid": 7, "ts": 1716454225473183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403716, "dur": 5,
    "args": {
      "External id": 254169,
      "cbid": 211, "correlation": 254169
    }
  },
  {
    "ph": "s", "id": 254169, "pid": 76337, "tid": -914061504, "ts": 1716454225403716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225473204, "dur": 33,
    "args": {
      "External id": 254175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254175, "pid": 5, "tid": 7, "ts": 1716454225473204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403743, "dur": 9,
    "args": {
      "External id": 254175,
      "cbid": 211, "correlation": 254175
    }
  },
  {
    "ph": "s", "id": 254175, "pid": 76337, "tid": -914061504, "ts": 1716454225403743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225473238, "dur": 3,
    "args": {
      "External id": 254183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254183,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 254183, "pid": 5, "tid": 7, "ts": 1716454225473238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403787, "dur": 9,
    "args": {
      "External id": 254183,
      "cbid": 211, "correlation": 254183
    }
  },
  {
    "ph": "s", "id": 254183, "pid": 76337, "tid": -914061504, "ts": 1716454225403787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403853, "dur": 1,
    "args": {
      "External id": 254199,
      "cbid": 251, "correlation": 254199
    }
  },
  {
    "ph": "f", "id": 254199, "pid": 76337, "tid": -914061504, "ts": 1716454225403853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403858, "dur": 0,
    "args": {
      "External id": 254201,
      "cbid": 251, "correlation": 254201
    }
  },
  {
    "ph": "f", "id": 254201, "pid": 76337, "tid": -914061504, "ts": 1716454225403858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225473243, "dur": 12,
    "args": {
      "External id": 254202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254202,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 254202, "pid": 5, "tid": 7, "ts": 1716454225473243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403860, "dur": 11,
    "args": {
      "External id": 254202,
      "cbid": 211, "correlation": 254202
    }
  },
  {
    "ph": "s", "id": 254202, "pid": 76337, "tid": -914061504, "ts": 1716454225403860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225473257, "dur": 5,
    "args": {
      "External id": 254204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254204,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 254204, "pid": 5, "tid": 7, "ts": 1716454225473257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403874, "dur": 5,
    "args": {
      "External id": 254204,
      "cbid": 211, "correlation": 254204
    }
  },
  {
    "ph": "s", "id": 254204, "pid": 76337, "tid": -914061504, "ts": 1716454225403874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225473263, "dur": 30,
    "args": {
      "External id": 254214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254214, "pid": 5, "tid": 7, "ts": 1716454225473263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225403932, "dur": 12,
    "args": {
      "External id": 254214,
      "cbid": 211, "correlation": 254214
    }
  },
  {
    "ph": "s", "id": 254214, "pid": 76337, "tid": -914061504, "ts": 1716454225403932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225473295, "dur": 32,
    "args": {
      "External id": 254234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254234,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 254234, "pid": 5, "tid": 7, "ts": 1716454225473295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404007, "dur": 11,
    "args": {
      "External id": 254234,
      "cbid": 211, "correlation": 254234
    }
  },
  {
    "ph": "s", "id": 254234, "pid": 76337, "tid": -914061504, "ts": 1716454225404007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225473328, "dur": 5,
    "args": {
      "External id": 254246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254246,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 254246, "pid": 5, "tid": 7, "ts": 1716454225473328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404028, "dur": 6,
    "args": {
      "External id": 254246,
      "cbid": 211, "correlation": 254246
    }
  },
  {
    "ph": "s", "id": 254246, "pid": 76337, "tid": -914061504, "ts": 1716454225404028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225473334, "dur": 30,
    "args": {
      "External id": 254249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254249,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254249, "pid": 5, "tid": 7, "ts": 1716454225473334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404046, "dur": 7,
    "args": {
      "External id": 254249,
      "cbid": 211, "correlation": 254249
    }
  },
  {
    "ph": "s", "id": 254249, "pid": 76337, "tid": -914061504, "ts": 1716454225404046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225473365, "dur": 20,
    "args": {
      "External id": 254258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254258,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254258, "pid": 5, "tid": 7, "ts": 1716454225473365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404086, "dur": 9,
    "args": {
      "External id": 254258,
      "cbid": 211, "correlation": 254258
    }
  },
  {
    "ph": "s", "id": 254258, "pid": 76337, "tid": -914061504, "ts": 1716454225404086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404147, "dur": 0,
    "args": {
      "External id": 254268,
      "cbid": 317, "correlation": 254268
    }
  },
  {
    "ph": "f", "id": 254268, "pid": 76337, "tid": -914061504, "ts": 1716454225404147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404148, "dur": 0,
    "args": {
      "External id": 254269,
      "cbid": 203, "correlation": 254269
    }
  },
  {
    "ph": "f", "id": 254269, "pid": 76337, "tid": -914061504, "ts": 1716454225404148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404149, "dur": 0,
    "args": {
      "External id": 254270,
      "cbid": 205, "correlation": 254270
    }
  },
  {
    "ph": "f", "id": 254270, "pid": 76337, "tid": -914061504, "ts": 1716454225404149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225473386, "dur": 23,
    "args": {
      "External id": 254274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254274,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254274, "pid": 5, "tid": 7, "ts": 1716454225473386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404164, "dur": 12,
    "args": {
      "External id": 254274,
      "cbid": 211, "correlation": 254274
    }
  },
  {
    "ph": "s", "id": 254274, "pid": 76337, "tid": -914061504, "ts": 1716454225404164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225473411, "dur": 45,
    "args": {
      "External id": 254276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254276,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254276, "pid": 5, "tid": 7, "ts": 1716454225473411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404179, "dur": 5,
    "args": {
      "External id": 254276,
      "cbid": 211, "correlation": 254276
    }
  },
  {
    "ph": "s", "id": 254276, "pid": 76337, "tid": -914061504, "ts": 1716454225404179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225473458, "dur": 658,
    "args": {
      "External id": 254278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254278,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254278, "pid": 5, "tid": 7, "ts": 1716454225473458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404190, "dur": 6,
    "args": {
      "External id": 254278,
      "cbid": 211, "correlation": 254278
    }
  },
  {
    "ph": "s", "id": 254278, "pid": 76337, "tid": -914061504, "ts": 1716454225404190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225474117, "dur": 23,
    "args": {
      "External id": 254280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254280,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254280, "pid": 5, "tid": 7, "ts": 1716454225474117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404201, "dur": 6,
    "args": {
      "External id": 254280,
      "cbid": 211, "correlation": 254280
    }
  },
  {
    "ph": "s", "id": 254280, "pid": 76337, "tid": -914061504, "ts": 1716454225404201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225474141, "dur": 33,
    "args": {
      "External id": 254286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254286, "pid": 5, "tid": 7, "ts": 1716454225474141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404229, "dur": 9,
    "args": {
      "External id": 254286,
      "cbid": 211, "correlation": 254286
    }
  },
  {
    "ph": "s", "id": 254286, "pid": 76337, "tid": -914061504, "ts": 1716454225404229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404288, "dur": 0,
    "args": {
      "External id": 254296,
      "cbid": 317, "correlation": 254296
    }
  },
  {
    "ph": "f", "id": 254296, "pid": 76337, "tid": -914061504, "ts": 1716454225404288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404289, "dur": 0,
    "args": {
      "External id": 254297,
      "cbid": 203, "correlation": 254297
    }
  },
  {
    "ph": "f", "id": 254297, "pid": 76337, "tid": -914061504, "ts": 1716454225404289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404289, "dur": 0,
    "args": {
      "External id": 254298,
      "cbid": 205, "correlation": 254298
    }
  },
  {
    "ph": "f", "id": 254298, "pid": 76337, "tid": -914061504, "ts": 1716454225404289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225474175, "dur": 56,
    "args": {
      "External id": 254302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254302, "pid": 5, "tid": 7, "ts": 1716454225474175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404302, "dur": 12,
    "args": {
      "External id": 254302,
      "cbid": 211, "correlation": 254302
    }
  },
  {
    "ph": "s", "id": 254302, "pid": 76337, "tid": -914061504, "ts": 1716454225404302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225474233, "dur": 275,
    "args": {
      "External id": 254304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254304,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254304, "pid": 5, "tid": 7, "ts": 1716454225474233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404320, "dur": 7,
    "args": {
      "External id": 254304,
      "cbid": 211, "correlation": 254304
    }
  },
  {
    "ph": "s", "id": 254304, "pid": 76337, "tid": -914061504, "ts": 1716454225404320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225474509, "dur": 22,
    "args": {
      "External id": 254306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254306,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254306, "pid": 5, "tid": 7, "ts": 1716454225474509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404331, "dur": 5,
    "args": {
      "External id": 254306,
      "cbid": 211, "correlation": 254306
    }
  },
  {
    "ph": "s", "id": 254306, "pid": 76337, "tid": -914061504, "ts": 1716454225404331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225474533, "dur": 32,
    "args": {
      "External id": 254312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254312, "pid": 5, "tid": 7, "ts": 1716454225474533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404357, "dur": 8,
    "args": {
      "External id": 254312,
      "cbid": 211, "correlation": 254312
    }
  },
  {
    "ph": "s", "id": 254312, "pid": 76337, "tid": -914061504, "ts": 1716454225404357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225474567, "dur": 27,
    "args": {
      "External id": 254320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254320,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254320, "pid": 5, "tid": 7, "ts": 1716454225474567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404386, "dur": 8,
    "args": {
      "External id": 254320,
      "cbid": 211, "correlation": 254320
    }
  },
  {
    "ph": "s", "id": 254320, "pid": 76337, "tid": -914061504, "ts": 1716454225404386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225474596, "dur": 20,
    "args": {
      "External id": 254328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254328,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254328, "pid": 5, "tid": 7, "ts": 1716454225474596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404415, "dur": 8,
    "args": {
      "External id": 254328,
      "cbid": 211, "correlation": 254328
    }
  },
  {
    "ph": "s", "id": 254328, "pid": 76337, "tid": -914061504, "ts": 1716454225404415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225474617, "dur": 30,
    "args": {
      "External id": 254348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254348,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 254348, "pid": 5, "tid": 7, "ts": 1716454225474617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404495, "dur": 13,
    "args": {
      "External id": 254348,
      "cbid": 211, "correlation": 254348
    }
  },
  {
    "ph": "s", "id": 254348, "pid": 76337, "tid": -914061504, "ts": 1716454225404495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225474649, "dur": 5,
    "args": {
      "External id": 254360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254360,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 254360, "pid": 5, "tid": 7, "ts": 1716454225474649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404518, "dur": 6,
    "args": {
      "External id": 254360,
      "cbid": 211, "correlation": 254360
    }
  },
  {
    "ph": "s", "id": 254360, "pid": 76337, "tid": -914061504, "ts": 1716454225404518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225474655, "dur": 32,
    "args": {
      "External id": 254363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254363,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254363, "pid": 5, "tid": 7, "ts": 1716454225474655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404535, "dur": 6,
    "args": {
      "External id": 254363,
      "cbid": 211, "correlation": 254363
    }
  },
  {
    "ph": "s", "id": 254363, "pid": 76337, "tid": -914061504, "ts": 1716454225404535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404591, "dur": 0,
    "args": {
      "External id": 254374,
      "cbid": 317, "correlation": 254374
    }
  },
  {
    "ph": "f", "id": 254374, "pid": 76337, "tid": -914061504, "ts": 1716454225404591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404592, "dur": 0,
    "args": {
      "External id": 254375,
      "cbid": 203, "correlation": 254375
    }
  },
  {
    "ph": "f", "id": 254375, "pid": 76337, "tid": -914061504, "ts": 1716454225404592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404593, "dur": 0,
    "args": {
      "External id": 254376,
      "cbid": 205, "correlation": 254376
    }
  },
  {
    "ph": "f", "id": 254376, "pid": 76337, "tid": -914061504, "ts": 1716454225404593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225474689, "dur": 22,
    "args": {
      "External id": 254380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254380,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254380, "pid": 5, "tid": 7, "ts": 1716454225474689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404607, "dur": 11,
    "args": {
      "External id": 254380,
      "cbid": 211, "correlation": 254380
    }
  },
  {
    "ph": "s", "id": 254380, "pid": 76337, "tid": -914061504, "ts": 1716454225404607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225474712, "dur": 107,
    "args": {
      "External id": 254382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254382,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254382, "pid": 5, "tid": 7, "ts": 1716454225474712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404624, "dur": 7,
    "args": {
      "External id": 254382,
      "cbid": 211, "correlation": 254382
    }
  },
  {
    "ph": "s", "id": 254382, "pid": 76337, "tid": -914061504, "ts": 1716454225404624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225474820, "dur": 23,
    "args": {
      "External id": 254384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254384,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254384, "pid": 5, "tid": 7, "ts": 1716454225474820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404634, "dur": 5,
    "args": {
      "External id": 254384,
      "cbid": 211, "correlation": 254384
    }
  },
  {
    "ph": "s", "id": 254384, "pid": 76337, "tid": -914061504, "ts": 1716454225404634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225474844, "dur": 33,
    "args": {
      "External id": 254390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254390,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254390, "pid": 5, "tid": 7, "ts": 1716454225474844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404662, "dur": 9,
    "args": {
      "External id": 254390,
      "cbid": 211, "correlation": 254390
    }
  },
  {
    "ph": "s", "id": 254390, "pid": 76337, "tid": -914061504, "ts": 1716454225404662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225474879, "dur": 199,
    "args": {
      "External id": 254399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254399,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254399, "pid": 5, "tid": 7, "ts": 1716454225474879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404742, "dur": 14,
    "args": {
      "External id": 254399,
      "cbid": 211, "correlation": 254399
    }
  },
  {
    "ph": "s", "id": 254399, "pid": 76337, "tid": -914061504, "ts": 1716454225404742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225475079, "dur": 66,
    "args": {
      "External id": 254421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254421,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254421, "pid": 5, "tid": 7, "ts": 1716454225475079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404800, "dur": 10,
    "args": {
      "External id": 254421,
      "cbid": 211, "correlation": 254421
    }
  },
  {
    "ph": "s", "id": 254421, "pid": 76337, "tid": -914061504, "ts": 1716454225404800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404890, "dur": 1,
    "args": {
      "External id": 254432,
      "cbid": 251, "correlation": 254432
    }
  },
  {
    "ph": "f", "id": 254432, "pid": 76337, "tid": -914061504, "ts": 1716454225404890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225475147, "dur": 157,
    "args": {
      "External id": 254433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254433,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254433, "pid": 5, "tid": 7, "ts": 1716454225475147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404895, "dur": 14,
    "args": {
      "External id": 254433,
      "cbid": 211, "correlation": 254433
    }
  },
  {
    "ph": "s", "id": 254433, "pid": 76337, "tid": -914061504, "ts": 1716454225404895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404966, "dur": 1,
    "args": {
      "External id": 254444,
      "cbid": 251, "correlation": 254444
    }
  },
  {
    "ph": "f", "id": 254444, "pid": 76337, "tid": -914061504, "ts": 1716454225404966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225475305, "dur": 149,
    "args": {
      "External id": 254445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254445,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254445, "pid": 5, "tid": 7, "ts": 1716454225475305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225404970, "dur": 19,
    "args": {
      "External id": 254445,
      "cbid": 211, "correlation": 254445
    }
  },
  {
    "ph": "s", "id": 254445, "pid": 76337, "tid": -914061504, "ts": 1716454225404970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405045, "dur": 1,
    "args": {
      "External id": 254456,
      "cbid": 251, "correlation": 254456
    }
  },
  {
    "ph": "f", "id": 254456, "pid": 76337, "tid": -914061504, "ts": 1716454225405045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225475455, "dur": 146,
    "args": {
      "External id": 254457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254457,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254457, "pid": 5, "tid": 7, "ts": 1716454225475455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405049, "dur": 12,
    "args": {
      "External id": 254457,
      "cbid": 211, "correlation": 254457
    }
  },
  {
    "ph": "s", "id": 254457, "pid": 76337, "tid": -914061504, "ts": 1716454225405049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225475602, "dur": 1984,
    "args": {
      "External id": 254478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254478,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 254478, "pid": 5, "tid": 7, "ts": 1716454225475602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405131, "dur": 13,
    "args": {
      "External id": 254478,
      "cbid": 211, "correlation": 254478
    }
  },
  {
    "ph": "s", "id": 254478, "pid": 76337, "tid": -914061504, "ts": 1716454225405131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405229, "dur": 1,
    "args": {
      "External id": 254496,
      "cbid": 251, "correlation": 254496
    }
  },
  {
    "ph": "f", "id": 254496, "pid": 76337, "tid": -914061504, "ts": 1716454225405229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225477588, "dur": 153,
    "args": {
      "External id": 254498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254498,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 254498, "pid": 5, "tid": 7, "ts": 1716454225477588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405235, "dur": 13,
    "args": {
      "External id": 254498,
      "cbid": 211, "correlation": 254498
    }
  },
  {
    "ph": "s", "id": 254498, "pid": 76337, "tid": -914061504, "ts": 1716454225405235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225477742, "dur": 35,
    "args": {
      "External id": 254506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254506,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254506, "pid": 5, "tid": 7, "ts": 1716454225477742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405305, "dur": 12,
    "args": {
      "External id": 254506,
      "cbid": 211, "correlation": 254506
    }
  },
  {
    "ph": "s", "id": 254506, "pid": 76337, "tid": -914061504, "ts": 1716454225405305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225477779, "dur": 50,
    "args": {
      "External id": 254514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254514,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254514, "pid": 5, "tid": 7, "ts": 1716454225477779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405344, "dur": 9,
    "args": {
      "External id": 254514,
      "cbid": 211, "correlation": 254514
    }
  },
  {
    "ph": "s", "id": 254514, "pid": 76337, "tid": -914061504, "ts": 1716454225405344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225477830, "dur": 31,
    "args": {
      "External id": 254525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254525,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254525, "pid": 5, "tid": 7, "ts": 1716454225477830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405416, "dur": 12,
    "args": {
      "External id": 254525,
      "cbid": 211, "correlation": 254525
    }
  },
  {
    "ph": "s", "id": 254525, "pid": 76337, "tid": -914061504, "ts": 1716454225405416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225477862, "dur": 35,
    "args": {
      "External id": 254547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254547,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254547, "pid": 5, "tid": 7, "ts": 1716454225477862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405446, "dur": 8,
    "args": {
      "External id": 254547,
      "cbid": 211, "correlation": 254547
    }
  },
  {
    "ph": "s", "id": 254547, "pid": 76337, "tid": -914061504, "ts": 1716454225405446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405531, "dur": 1,
    "args": {
      "External id": 254558,
      "cbid": 251, "correlation": 254558
    }
  },
  {
    "ph": "f", "id": 254558, "pid": 76337, "tid": -914061504, "ts": 1716454225405531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225477899, "dur": 94,
    "args": {
      "External id": 254559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254559,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254559, "pid": 5, "tid": 7, "ts": 1716454225477899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405536, "dur": 13,
    "args": {
      "External id": 254559,
      "cbid": 211, "correlation": 254559
    }
  },
  {
    "ph": "s", "id": 254559, "pid": 76337, "tid": -914061504, "ts": 1716454225405536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405606, "dur": 1,
    "args": {
      "External id": 254570,
      "cbid": 251, "correlation": 254570
    }
  },
  {
    "ph": "f", "id": 254570, "pid": 76337, "tid": -914061504, "ts": 1716454225405606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405610, "dur": 0,
    "args": {
      "External id": 254571,
      "cbid": 251, "correlation": 254571
    }
  },
  {
    "ph": "f", "id": 254571, "pid": 76337, "tid": -914061504, "ts": 1716454225405610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225477994, "dur": 11,
    "args": {
      "External id": 254572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254572,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 254572, "pid": 5, "tid": 7, "ts": 1716454225477994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405612, "dur": 12,
    "args": {
      "External id": 254572,
      "cbid": 211, "correlation": 254572
    }
  },
  {
    "ph": "s", "id": 254572, "pid": 76337, "tid": -914061504, "ts": 1716454225405612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225478007, "dur": 5,
    "args": {
      "External id": 254574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254574,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 254574, "pid": 5, "tid": 7, "ts": 1716454225478007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405625, "dur": 6,
    "args": {
      "External id": 254574,
      "cbid": 211, "correlation": 254574
    }
  },
  {
    "ph": "s", "id": 254574, "pid": 76337, "tid": -914061504, "ts": 1716454225405625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405682, "dur": 1,
    "args": {
      "External id": 254585,
      "cbid": 251, "correlation": 254585
    }
  },
  {
    "ph": "f", "id": 254585, "pid": 76337, "tid": -914061504, "ts": 1716454225405682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405685, "dur": 0,
    "args": {
      "External id": 254586,
      "cbid": 251, "correlation": 254586
    }
  },
  {
    "ph": "f", "id": 254586, "pid": 76337, "tid": -914061504, "ts": 1716454225405685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225478014, "dur": 8,
    "args": {
      "External id": 254587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254587,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 254587, "pid": 5, "tid": 7, "ts": 1716454225478014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405686, "dur": 11,
    "args": {
      "External id": 254587,
      "cbid": 211, "correlation": 254587
    }
  },
  {
    "ph": "s", "id": 254587, "pid": 76337, "tid": -914061504, "ts": 1716454225405686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225478022, "dur": 4,
    "args": {
      "External id": 254589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254589,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 254589, "pid": 5, "tid": 7, "ts": 1716454225478022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405699, "dur": 6,
    "args": {
      "External id": 254589,
      "cbid": 211, "correlation": 254589
    }
  },
  {
    "ph": "s", "id": 254589, "pid": 76337, "tid": -914061504, "ts": 1716454225405699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225478028, "dur": 94,
    "args": {
      "External id": 254610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254610,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 254610, "pid": 5, "tid": 7, "ts": 1716454225478028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405773, "dur": 13,
    "args": {
      "External id": 254610,
      "cbid": 211, "correlation": 254610
    }
  },
  {
    "ph": "s", "id": 254610, "pid": 76337, "tid": -914061504, "ts": 1716454225405773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405869, "dur": 1,
    "args": {
      "External id": 254628,
      "cbid": 251, "correlation": 254628
    }
  },
  {
    "ph": "f", "id": 254628, "pid": 76337, "tid": -914061504, "ts": 1716454225405869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225478123, "dur": 100,
    "args": {
      "External id": 254630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254630,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254630, "pid": 5, "tid": 7, "ts": 1716454225478123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405874, "dur": 13,
    "args": {
      "External id": 254630,
      "cbid": 211, "correlation": 254630
    }
  },
  {
    "ph": "s", "id": 254630, "pid": 76337, "tid": -914061504, "ts": 1716454225405874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225478225, "dur": 19,
    "args": {
      "External id": 254638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254638,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254638, "pid": 5, "tid": 7, "ts": 1716454225478225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405943, "dur": 12,
    "args": {
      "External id": 254638,
      "cbid": 211, "correlation": 254638
    }
  },
  {
    "ph": "s", "id": 254638, "pid": 76337, "tid": -914061504, "ts": 1716454225405943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225478245, "dur": 38,
    "args": {
      "External id": 254646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254646, "pid": 5, "tid": 7, "ts": 1716454225478245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225405993, "dur": 10,
    "args": {
      "External id": 254646,
      "cbid": 211, "correlation": 254646
    }
  },
  {
    "ph": "s", "id": 254646, "pid": 76337, "tid": -914061504, "ts": 1716454225405993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225478284, "dur": 35,
    "args": {
      "External id": 254668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254668,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254668, "pid": 5, "tid": 7, "ts": 1716454225478284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406045, "dur": 10,
    "args": {
      "External id": 254668,
      "cbid": 211, "correlation": 254668
    }
  },
  {
    "ph": "s", "id": 254668, "pid": 76337, "tid": -914061504, "ts": 1716454225406045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406134, "dur": 1,
    "args": {
      "External id": 254684,
      "cbid": 251, "correlation": 254684
    }
  },
  {
    "ph": "f", "id": 254684, "pid": 76337, "tid": -914061504, "ts": 1716454225406134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406139, "dur": 0,
    "args": {
      "External id": 254686,
      "cbid": 251, "correlation": 254686
    }
  },
  {
    "ph": "f", "id": 254686, "pid": 76337, "tid": -914061504, "ts": 1716454225406139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225478321, "dur": 550,
    "args": {
      "External id": 254687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254687,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 254687, "pid": 5, "tid": 7, "ts": 1716454225478321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406144, "dur": 13,
    "args": {
      "External id": 254687,
      "cbid": 211, "correlation": 254687
    }
  },
  {
    "ph": "s", "id": 254687, "pid": 76337, "tid": -914061504, "ts": 1716454225406144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225478872, "dur": 126,
    "args": {
      "External id": 254695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254695,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254695, "pid": 5, "tid": 7, "ts": 1716454225478872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406211, "dur": 12,
    "args": {
      "External id": 254695,
      "cbid": 211, "correlation": 254695
    }
  },
  {
    "ph": "s", "id": 254695, "pid": 76337, "tid": -914061504, "ts": 1716454225406211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225479000, "dur": 130,
    "args": {
      "External id": 254703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254703, "pid": 5, "tid": 7, "ts": 1716454225479000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406241, "dur": 8,
    "args": {
      "External id": 254703,
      "cbid": 211, "correlation": 254703
    }
  },
  {
    "ph": "s", "id": 254703, "pid": 76337, "tid": -914061504, "ts": 1716454225406241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406318, "dur": 1,
    "args": {
      "External id": 254719,
      "cbid": 251, "correlation": 254719
    }
  },
  {
    "ph": "f", "id": 254719, "pid": 76337, "tid": -914061504, "ts": 1716454225406318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225479131, "dur": 310,
    "args": {
      "External id": 254721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254721,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254721, "pid": 5, "tid": 7, "ts": 1716454225479131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406324, "dur": 12,
    "args": {
      "External id": 254721,
      "cbid": 211, "correlation": 254721
    }
  },
  {
    "ph": "s", "id": 254721, "pid": 76337, "tid": -914061504, "ts": 1716454225406324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225479442, "dur": 27,
    "args": {
      "External id": 254729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254729,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254729, "pid": 5, "tid": 7, "ts": 1716454225479442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406366, "dur": 9,
    "args": {
      "External id": 254729,
      "cbid": 211, "correlation": 254729
    }
  },
  {
    "ph": "s", "id": 254729, "pid": 76337, "tid": -914061504, "ts": 1716454225406366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225479470, "dur": 83,
    "args": {
      "External id": 254740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254740, "pid": 5, "tid": 7, "ts": 1716454225479470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406433, "dur": 12,
    "args": {
      "External id": 254740,
      "cbid": 211, "correlation": 254740
    }
  },
  {
    "ph": "s", "id": 254740, "pid": 76337, "tid": -914061504, "ts": 1716454225406433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406497, "dur": 0,
    "args": {
      "External id": 254752,
      "cbid": 317, "correlation": 254752
    }
  },
  {
    "ph": "f", "id": 254752, "pid": 76337, "tid": -914061504, "ts": 1716454225406497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406498, "dur": 0,
    "args": {
      "External id": 254753,
      "cbid": 203, "correlation": 254753
    }
  },
  {
    "ph": "f", "id": 254753, "pid": 76337, "tid": -914061504, "ts": 1716454225406498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406498, "dur": 0,
    "args": {
      "External id": 254754,
      "cbid": 205, "correlation": 254754
    }
  },
  {
    "ph": "f", "id": 254754, "pid": 76337, "tid": -914061504, "ts": 1716454225406498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225479554, "dur": 23,
    "args": {
      "External id": 254758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254758,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254758, "pid": 5, "tid": 7, "ts": 1716454225479554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406515, "dur": 12,
    "args": {
      "External id": 254758,
      "cbid": 211, "correlation": 254758
    }
  },
  {
    "ph": "s", "id": 254758, "pid": 76337, "tid": -914061504, "ts": 1716454225406515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225479579, "dur": 123,
    "args": {
      "External id": 254760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254760,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254760, "pid": 5, "tid": 7, "ts": 1716454225479579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406533, "dur": 6,
    "args": {
      "External id": 254760,
      "cbid": 211, "correlation": 254760
    }
  },
  {
    "ph": "s", "id": 254760, "pid": 76337, "tid": -914061504, "ts": 1716454225406533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225479704, "dur": 23,
    "args": {
      "External id": 254762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254762,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254762, "pid": 5, "tid": 7, "ts": 1716454225479704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406543, "dur": 5,
    "args": {
      "External id": 254762,
      "cbid": 211, "correlation": 254762
    }
  },
  {
    "ph": "s", "id": 254762, "pid": 76337, "tid": -914061504, "ts": 1716454225406543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225479728, "dur": 33,
    "args": {
      "External id": 254768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254768,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254768, "pid": 5, "tid": 7, "ts": 1716454225479728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406571, "dur": 8,
    "args": {
      "External id": 254768,
      "cbid": 211, "correlation": 254768
    }
  },
  {
    "ph": "s", "id": 254768, "pid": 76337, "tid": -914061504, "ts": 1716454225406571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225479763, "dur": 27,
    "args": {
      "External id": 254776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254776,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254776, "pid": 5, "tid": 7, "ts": 1716454225479763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406603, "dur": 9,
    "args": {
      "External id": 254776,
      "cbid": 211, "correlation": 254776
    }
  },
  {
    "ph": "s", "id": 254776, "pid": 76337, "tid": -914061504, "ts": 1716454225406603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225479791, "dur": 54,
    "args": {
      "External id": 254785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254785,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254785, "pid": 5, "tid": 7, "ts": 1716454225479791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406641, "dur": 10,
    "args": {
      "External id": 254785,
      "cbid": 211, "correlation": 254785
    }
  },
  {
    "ph": "s", "id": 254785, "pid": 76337, "tid": -914061504, "ts": 1716454225406641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225479847, "dur": 52,
    "args": {
      "External id": 254805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254805,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 254805, "pid": 5, "tid": 7, "ts": 1716454225479847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406713, "dur": 11,
    "args": {
      "External id": 254805,
      "cbid": 211, "correlation": 254805
    }
  },
  {
    "ph": "s", "id": 254805, "pid": 76337, "tid": -914061504, "ts": 1716454225406713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225479900, "dur": 5,
    "args": {
      "External id": 254817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254817,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 254817, "pid": 5, "tid": 7, "ts": 1716454225479900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406734, "dur": 6,
    "args": {
      "External id": 254817,
      "cbid": 211, "correlation": 254817
    }
  },
  {
    "ph": "s", "id": 254817, "pid": 76337, "tid": -914061504, "ts": 1716454225406734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225479906, "dur": 58,
    "args": {
      "External id": 254820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254820, "pid": 5, "tid": 7, "ts": 1716454225479906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406752, "dur": 7,
    "args": {
      "External id": 254820,
      "cbid": 211, "correlation": 254820
    }
  },
  {
    "ph": "s", "id": 254820, "pid": 76337, "tid": -914061504, "ts": 1716454225406752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225479966, "dur": 37,
    "args": {
      "External id": 254829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254829,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254829, "pid": 5, "tid": 7, "ts": 1716454225479966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406792, "dur": 9,
    "args": {
      "External id": 254829,
      "cbid": 211, "correlation": 254829
    }
  },
  {
    "ph": "s", "id": 254829, "pid": 76337, "tid": -914061504, "ts": 1716454225406792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406843, "dur": 0,
    "args": {
      "External id": 254839,
      "cbid": 317, "correlation": 254839
    }
  },
  {
    "ph": "f", "id": 254839, "pid": 76337, "tid": -914061504, "ts": 1716454225406843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406844, "dur": 0,
    "args": {
      "External id": 254840,
      "cbid": 203, "correlation": 254840
    }
  },
  {
    "ph": "f", "id": 254840, "pid": 76337, "tid": -914061504, "ts": 1716454225406844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406845, "dur": 0,
    "args": {
      "External id": 254841,
      "cbid": 205, "correlation": 254841
    }
  },
  {
    "ph": "f", "id": 254841, "pid": 76337, "tid": -914061504, "ts": 1716454225406845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225480004, "dur": 39,
    "args": {
      "External id": 254845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254845,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254845, "pid": 5, "tid": 7, "ts": 1716454225480004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406860, "dur": 11,
    "args": {
      "External id": 254845,
      "cbid": 211, "correlation": 254845
    }
  },
  {
    "ph": "s", "id": 254845, "pid": 76337, "tid": -914061504, "ts": 1716454225406860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225480045, "dur": 84,
    "args": {
      "External id": 254847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254847,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254847, "pid": 5, "tid": 7, "ts": 1716454225480045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406873, "dur": 5,
    "args": {
      "External id": 254847,
      "cbid": 211, "correlation": 254847
    }
  },
  {
    "ph": "s", "id": 254847, "pid": 76337, "tid": -914061504, "ts": 1716454225406873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225480131, "dur": 1299,
    "args": {
      "External id": 254849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254849,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254849, "pid": 5, "tid": 7, "ts": 1716454225480131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406884, "dur": 6,
    "args": {
      "External id": 254849,
      "cbid": 211, "correlation": 254849
    }
  },
  {
    "ph": "s", "id": 254849, "pid": 76337, "tid": -914061504, "ts": 1716454225406884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225481431, "dur": 22,
    "args": {
      "External id": 254851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254851, "pid": 5, "tid": 7, "ts": 1716454225481431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406894, "dur": 5,
    "args": {
      "External id": 254851,
      "cbid": 211, "correlation": 254851
    }
  },
  {
    "ph": "s", "id": 254851, "pid": 76337, "tid": -914061504, "ts": 1716454225406894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225481454, "dur": 33,
    "args": {
      "External id": 254857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254857, "pid": 5, "tid": 7, "ts": 1716454225481454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406922, "dur": 8,
    "args": {
      "External id": 254857,
      "cbid": 211, "correlation": 254857
    }
  },
  {
    "ph": "s", "id": 254857, "pid": 76337, "tid": -914061504, "ts": 1716454225406922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225481489, "dur": 3,
    "args": {
      "External id": 254865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254865,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 254865, "pid": 5, "tid": 7, "ts": 1716454225481489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225406965, "dur": 16,
    "args": {
      "External id": 254865,
      "cbid": 211, "correlation": 254865
    }
  },
  {
    "ph": "s", "id": 254865, "pid": 76337, "tid": -914061504, "ts": 1716454225406965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407036, "dur": 1,
    "args": {
      "External id": 254881,
      "cbid": 251, "correlation": 254881
    }
  },
  {
    "ph": "f", "id": 254881, "pid": 76337, "tid": -914061504, "ts": 1716454225407036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407041, "dur": 0,
    "args": {
      "External id": 254883,
      "cbid": 251, "correlation": 254883
    }
  },
  {
    "ph": "f", "id": 254883, "pid": 76337, "tid": -914061504, "ts": 1716454225407041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225481493, "dur": 12,
    "args": {
      "External id": 254884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254884,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 254884, "pid": 5, "tid": 7, "ts": 1716454225481493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407043, "dur": 12,
    "args": {
      "External id": 254884,
      "cbid": 211, "correlation": 254884
    }
  },
  {
    "ph": "s", "id": 254884, "pid": 76337, "tid": -914061504, "ts": 1716454225407043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225481507, "dur": 5,
    "args": {
      "External id": 254886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254886,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 254886, "pid": 5, "tid": 7, "ts": 1716454225481507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407056, "dur": 6,
    "args": {
      "External id": 254886,
      "cbid": 211, "correlation": 254886
    }
  },
  {
    "ph": "s", "id": 254886, "pid": 76337, "tid": -914061504, "ts": 1716454225407056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225481513, "dur": 29,
    "args": {
      "External id": 254896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254896,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254896, "pid": 5, "tid": 7, "ts": 1716454225481513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407115, "dur": 12,
    "args": {
      "External id": 254896,
      "cbid": 211, "correlation": 254896
    }
  },
  {
    "ph": "s", "id": 254896, "pid": 76337, "tid": -914061504, "ts": 1716454225407115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225481543, "dur": 31,
    "args": {
      "External id": 254916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254916,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 254916, "pid": 5, "tid": 7, "ts": 1716454225481543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407179, "dur": 10,
    "args": {
      "External id": 254916,
      "cbid": 211, "correlation": 254916
    }
  },
  {
    "ph": "s", "id": 254916, "pid": 76337, "tid": -914061504, "ts": 1716454225407179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225481576, "dur": 5,
    "args": {
      "External id": 254928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254928,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 254928, "pid": 5, "tid": 7, "ts": 1716454225481576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407199, "dur": 6,
    "args": {
      "External id": 254928,
      "cbid": 211, "correlation": 254928
    }
  },
  {
    "ph": "s", "id": 254928, "pid": 76337, "tid": -914061504, "ts": 1716454225407199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225481582, "dur": 30,
    "args": {
      "External id": 254931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254931, "pid": 5, "tid": 7, "ts": 1716454225481582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407218, "dur": 9,
    "args": {
      "External id": 254931,
      "cbid": 211, "correlation": 254931
    }
  },
  {
    "ph": "s", "id": 254931, "pid": 76337, "tid": -914061504, "ts": 1716454225407218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225481614, "dur": 20,
    "args": {
      "External id": 254940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254940,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254940, "pid": 5, "tid": 7, "ts": 1716454225481614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407261, "dur": 10,
    "args": {
      "External id": 254940,
      "cbid": 211, "correlation": 254940
    }
  },
  {
    "ph": "s", "id": 254940, "pid": 76337, "tid": -914061504, "ts": 1716454225407261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407324, "dur": 0,
    "args": {
      "External id": 254950,
      "cbid": 317, "correlation": 254950
    }
  },
  {
    "ph": "f", "id": 254950, "pid": 76337, "tid": -914061504, "ts": 1716454225407324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407325, "dur": 0,
    "args": {
      "External id": 254951,
      "cbid": 203, "correlation": 254951
    }
  },
  {
    "ph": "f", "id": 254951, "pid": 76337, "tid": -914061504, "ts": 1716454225407325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407326, "dur": 0,
    "args": {
      "External id": 254952,
      "cbid": 205, "correlation": 254952
    }
  },
  {
    "ph": "f", "id": 254952, "pid": 76337, "tid": -914061504, "ts": 1716454225407326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225481635, "dur": 23,
    "args": {
      "External id": 254956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254956,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254956, "pid": 5, "tid": 7, "ts": 1716454225481635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407340, "dur": 11,
    "args": {
      "External id": 254956,
      "cbid": 211, "correlation": 254956
    }
  },
  {
    "ph": "s", "id": 254956, "pid": 76337, "tid": -914061504, "ts": 1716454225407340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225481660, "dur": 45,
    "args": {
      "External id": 254958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254958,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254958, "pid": 5, "tid": 7, "ts": 1716454225481660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407353, "dur": 6,
    "args": {
      "External id": 254958,
      "cbid": 211, "correlation": 254958
    }
  },
  {
    "ph": "s", "id": 254958, "pid": 76337, "tid": -914061504, "ts": 1716454225407353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225481706, "dur": 658,
    "args": {
      "External id": 254960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254960,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254960, "pid": 5, "tid": 7, "ts": 1716454225481706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407365, "dur": 6,
    "args": {
      "External id": 254960,
      "cbid": 211, "correlation": 254960
    }
  },
  {
    "ph": "s", "id": 254960, "pid": 76337, "tid": -914061504, "ts": 1716454225407365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225482365, "dur": 22,
    "args": {
      "External id": 254962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254962,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254962, "pid": 5, "tid": 7, "ts": 1716454225482365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407375, "dur": 5,
    "args": {
      "External id": 254962,
      "cbid": 211, "correlation": 254962
    }
  },
  {
    "ph": "s", "id": 254962, "pid": 76337, "tid": -914061504, "ts": 1716454225407375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225482389, "dur": 34,
    "args": {
      "External id": 254968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254968, "pid": 5, "tid": 7, "ts": 1716454225482389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407403, "dur": 9,
    "args": {
      "External id": 254968,
      "cbid": 211, "correlation": 254968
    }
  },
  {
    "ph": "s", "id": 254968, "pid": 76337, "tid": -914061504, "ts": 1716454225407403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407462, "dur": 0,
    "args": {
      "External id": 254978,
      "cbid": 317, "correlation": 254978
    }
  },
  {
    "ph": "f", "id": 254978, "pid": 76337, "tid": -914061504, "ts": 1716454225407462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407462, "dur": 0,
    "args": {
      "External id": 254979,
      "cbid": 203, "correlation": 254979
    }
  },
  {
    "ph": "f", "id": 254979, "pid": 76337, "tid": -914061504, "ts": 1716454225407462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407463, "dur": 0,
    "args": {
      "External id": 254980,
      "cbid": 205, "correlation": 254980
    }
  },
  {
    "ph": "f", "id": 254980, "pid": 76337, "tid": -914061504, "ts": 1716454225407463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225482424, "dur": 40,
    "args": {
      "External id": 254984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254984,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254984, "pid": 5, "tid": 7, "ts": 1716454225482424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407475, "dur": 11,
    "args": {
      "External id": 254984,
      "cbid": 211, "correlation": 254984
    }
  },
  {
    "ph": "s", "id": 254984, "pid": 76337, "tid": -914061504, "ts": 1716454225407475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225482465, "dur": 193,
    "args": {
      "External id": 254986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254986,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 254986, "pid": 5, "tid": 7, "ts": 1716454225482465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407492, "dur": 6,
    "args": {
      "External id": 254986,
      "cbid": 211, "correlation": 254986
    }
  },
  {
    "ph": "s", "id": 254986, "pid": 76337, "tid": -914061504, "ts": 1716454225407492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225482660, "dur": 23,
    "args": {
      "External id": 254988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254988, "pid": 5, "tid": 7, "ts": 1716454225482660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407502, "dur": 5,
    "args": {
      "External id": 254988,
      "cbid": 211, "correlation": 254988
    }
  },
  {
    "ph": "s", "id": 254988, "pid": 76337, "tid": -914061504, "ts": 1716454225407502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225482685, "dur": 33,
    "args": {
      "External id": 254994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 254994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 254994, "pid": 5, "tid": 7, "ts": 1716454225482685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407528, "dur": 8,
    "args": {
      "External id": 254994,
      "cbid": 211, "correlation": 254994
    }
  },
  {
    "ph": "s", "id": 254994, "pid": 76337, "tid": -914061504, "ts": 1716454225407528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225482719, "dur": 27,
    "args": {
      "External id": 255002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255002,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255002, "pid": 5, "tid": 7, "ts": 1716454225482719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407557, "dur": 8,
    "args": {
      "External id": 255002,
      "cbid": 211, "correlation": 255002
    }
  },
  {
    "ph": "s", "id": 255002, "pid": 76337, "tid": -914061504, "ts": 1716454225407557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225482748, "dur": 20,
    "args": {
      "External id": 255010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255010,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255010, "pid": 5, "tid": 7, "ts": 1716454225482748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407586, "dur": 8,
    "args": {
      "External id": 255010,
      "cbid": 211, "correlation": 255010
    }
  },
  {
    "ph": "s", "id": 255010, "pid": 76337, "tid": -914061504, "ts": 1716454225407586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225482769, "dur": 31,
    "args": {
      "External id": 255030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255030,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 255030, "pid": 5, "tid": 7, "ts": 1716454225482769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407669, "dur": 12,
    "args": {
      "External id": 255030,
      "cbid": 211, "correlation": 255030
    }
  },
  {
    "ph": "s", "id": 255030, "pid": 76337, "tid": -914061504, "ts": 1716454225407669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225482801, "dur": 5,
    "args": {
      "External id": 255042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255042,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 255042, "pid": 5, "tid": 7, "ts": 1716454225482801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407690, "dur": 6,
    "args": {
      "External id": 255042,
      "cbid": 211, "correlation": 255042
    }
  },
  {
    "ph": "s", "id": 255042, "pid": 76337, "tid": -914061504, "ts": 1716454225407690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225482808, "dur": 30,
    "args": {
      "External id": 255045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255045, "pid": 5, "tid": 7, "ts": 1716454225482808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407708, "dur": 7,
    "args": {
      "External id": 255045,
      "cbid": 211, "correlation": 255045
    }
  },
  {
    "ph": "s", "id": 255045, "pid": 76337, "tid": -914061504, "ts": 1716454225407708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407766, "dur": 0,
    "args": {
      "External id": 255056,
      "cbid": 317, "correlation": 255056
    }
  },
  {
    "ph": "f", "id": 255056, "pid": 76337, "tid": -914061504, "ts": 1716454225407766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407767, "dur": 0,
    "args": {
      "External id": 255057,
      "cbid": 203, "correlation": 255057
    }
  },
  {
    "ph": "f", "id": 255057, "pid": 76337, "tid": -914061504, "ts": 1716454225407767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407767, "dur": 0,
    "args": {
      "External id": 255058,
      "cbid": 205, "correlation": 255058
    }
  },
  {
    "ph": "f", "id": 255058, "pid": 76337, "tid": -914061504, "ts": 1716454225407767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225482840, "dur": 21,
    "args": {
      "External id": 255062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255062,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255062, "pid": 5, "tid": 7, "ts": 1716454225482840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407780, "dur": 11,
    "args": {
      "External id": 255062,
      "cbid": 211, "correlation": 255062
    }
  },
  {
    "ph": "s", "id": 255062, "pid": 76337, "tid": -914061504, "ts": 1716454225407780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225482862, "dur": 107,
    "args": {
      "External id": 255064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255064,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255064, "pid": 5, "tid": 7, "ts": 1716454225482862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407797, "dur": 6,
    "args": {
      "External id": 255064,
      "cbid": 211, "correlation": 255064
    }
  },
  {
    "ph": "s", "id": 255064, "pid": 76337, "tid": -914061504, "ts": 1716454225407797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225482971, "dur": 22,
    "args": {
      "External id": 255066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255066,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255066, "pid": 5, "tid": 7, "ts": 1716454225482971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407807, "dur": 6,
    "args": {
      "External id": 255066,
      "cbid": 211, "correlation": 255066
    }
  },
  {
    "ph": "s", "id": 255066, "pid": 76337, "tid": -914061504, "ts": 1716454225407807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225482995, "dur": 34,
    "args": {
      "External id": 255072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255072, "pid": 5, "tid": 7, "ts": 1716454225482995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407835, "dur": 9,
    "args": {
      "External id": 255072,
      "cbid": 211, "correlation": 255072
    }
  },
  {
    "ph": "s", "id": 255072, "pid": 76337, "tid": -914061504, "ts": 1716454225407835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225483030, "dur": 200,
    "args": {
      "External id": 255081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255081,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255081, "pid": 5, "tid": 7, "ts": 1716454225483030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407918, "dur": 14,
    "args": {
      "External id": 255081,
      "cbid": 211, "correlation": 255081
    }
  },
  {
    "ph": "s", "id": 255081, "pid": 76337, "tid": -914061504, "ts": 1716454225407918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225483231, "dur": 66,
    "args": {
      "External id": 255103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255103,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255103, "pid": 5, "tid": 7, "ts": 1716454225483231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225407982, "dur": 11,
    "args": {
      "External id": 255103,
      "cbid": 211, "correlation": 255103
    }
  },
  {
    "ph": "s", "id": 255103, "pid": 76337, "tid": -914061504, "ts": 1716454225407982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408072, "dur": 1,
    "args": {
      "External id": 255114,
      "cbid": 251, "correlation": 255114
    }
  },
  {
    "ph": "f", "id": 255114, "pid": 76337, "tid": -914061504, "ts": 1716454225408072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225483299, "dur": 158,
    "args": {
      "External id": 255115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255115,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255115, "pid": 5, "tid": 7, "ts": 1716454225483299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408078, "dur": 13,
    "args": {
      "External id": 255115,
      "cbid": 211, "correlation": 255115
    }
  },
  {
    "ph": "s", "id": 255115, "pid": 76337, "tid": -914061504, "ts": 1716454225408078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408148, "dur": 1,
    "args": {
      "External id": 255126,
      "cbid": 251, "correlation": 255126
    }
  },
  {
    "ph": "f", "id": 255126, "pid": 76337, "tid": -914061504, "ts": 1716454225408148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225483459, "dur": 151,
    "args": {
      "External id": 255127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255127,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255127, "pid": 5, "tid": 7, "ts": 1716454225483459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408152, "dur": 13,
    "args": {
      "External id": 255127,
      "cbid": 211, "correlation": 255127
    }
  },
  {
    "ph": "s", "id": 255127, "pid": 76337, "tid": -914061504, "ts": 1716454225408152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408219, "dur": 1,
    "args": {
      "External id": 255138,
      "cbid": 251, "correlation": 255138
    }
  },
  {
    "ph": "f", "id": 255138, "pid": 76337, "tid": -914061504, "ts": 1716454225408219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225483611, "dur": 147,
    "args": {
      "External id": 255139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255139,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255139, "pid": 5, "tid": 7, "ts": 1716454225483611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408223, "dur": 11,
    "args": {
      "External id": 255139,
      "cbid": 211, "correlation": 255139
    }
  },
  {
    "ph": "s", "id": 255139, "pid": 76337, "tid": -914061504, "ts": 1716454225408223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225483759, "dur": 1982,
    "args": {
      "External id": 255160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255160,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 255160, "pid": 5, "tid": 7, "ts": 1716454225483759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408303, "dur": 13,
    "args": {
      "External id": 255160,
      "cbid": 211, "correlation": 255160
    }
  },
  {
    "ph": "s", "id": 255160, "pid": 76337, "tid": -914061504, "ts": 1716454225408303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408402, "dur": 1,
    "args": {
      "External id": 255178,
      "cbid": 251, "correlation": 255178
    }
  },
  {
    "ph": "f", "id": 255178, "pid": 76337, "tid": -914061504, "ts": 1716454225408402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225485742, "dur": 149,
    "args": {
      "External id": 255180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255180,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 255180, "pid": 5, "tid": 7, "ts": 1716454225485742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408408, "dur": 14,
    "args": {
      "External id": 255180,
      "cbid": 211, "correlation": 255180
    }
  },
  {
    "ph": "s", "id": 255180, "pid": 76337, "tid": -914061504, "ts": 1716454225408408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225485893, "dur": 35,
    "args": {
      "External id": 255188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255188,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255188, "pid": 5, "tid": 7, "ts": 1716454225485893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408480, "dur": 12,
    "args": {
      "External id": 255188,
      "cbid": 211, "correlation": 255188
    }
  },
  {
    "ph": "s", "id": 255188, "pid": 76337, "tid": -914061504, "ts": 1716454225408480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225485929, "dur": 51,
    "args": {
      "External id": 255196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255196,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255196, "pid": 5, "tid": 7, "ts": 1716454225485929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408519, "dur": 8,
    "args": {
      "External id": 255196,
      "cbid": 211, "correlation": 255196
    }
  },
  {
    "ph": "s", "id": 255196, "pid": 76337, "tid": -914061504, "ts": 1716454225408519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225485982, "dur": 31,
    "args": {
      "External id": 255207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255207,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255207, "pid": 5, "tid": 7, "ts": 1716454225485982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408590, "dur": 13,
    "args": {
      "External id": 255207,
      "cbid": 211, "correlation": 255207
    }
  },
  {
    "ph": "s", "id": 255207, "pid": 76337, "tid": -914061504, "ts": 1716454225408590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225486013, "dur": 35,
    "args": {
      "External id": 255229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255229,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255229, "pid": 5, "tid": 7, "ts": 1716454225486013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408622, "dur": 8,
    "args": {
      "External id": 255229,
      "cbid": 211, "correlation": 255229
    }
  },
  {
    "ph": "s", "id": 255229, "pid": 76337, "tid": -914061504, "ts": 1716454225408622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408707, "dur": 1,
    "args": {
      "External id": 255240,
      "cbid": 251, "correlation": 255240
    }
  },
  {
    "ph": "f", "id": 255240, "pid": 76337, "tid": -914061504, "ts": 1716454225408707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225486050, "dur": 92,
    "args": {
      "External id": 255241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255241,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255241, "pid": 5, "tid": 7, "ts": 1716454225486050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408712, "dur": 13,
    "args": {
      "External id": 255241,
      "cbid": 211, "correlation": 255241
    }
  },
  {
    "ph": "s", "id": 255241, "pid": 76337, "tid": -914061504, "ts": 1716454225408712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408782, "dur": 1,
    "args": {
      "External id": 255252,
      "cbid": 251, "correlation": 255252
    }
  },
  {
    "ph": "f", "id": 255252, "pid": 76337, "tid": -914061504, "ts": 1716454225408782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408786, "dur": 0,
    "args": {
      "External id": 255253,
      "cbid": 251, "correlation": 255253
    }
  },
  {
    "ph": "f", "id": 255253, "pid": 76337, "tid": -914061504, "ts": 1716454225408786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225486143, "dur": 11,
    "args": {
      "External id": 255254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255254,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 255254, "pid": 5, "tid": 7, "ts": 1716454225486143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408788, "dur": 12,
    "args": {
      "External id": 255254,
      "cbid": 211, "correlation": 255254
    }
  },
  {
    "ph": "s", "id": 255254, "pid": 76337, "tid": -914061504, "ts": 1716454225408788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225486155, "dur": 5,
    "args": {
      "External id": 255256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255256,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 255256, "pid": 5, "tid": 7, "ts": 1716454225486155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408801, "dur": 5,
    "args": {
      "External id": 255256,
      "cbid": 211, "correlation": 255256
    }
  },
  {
    "ph": "s", "id": 255256, "pid": 76337, "tid": -914061504, "ts": 1716454225408801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408858, "dur": 1,
    "args": {
      "External id": 255267,
      "cbid": 251, "correlation": 255267
    }
  },
  {
    "ph": "f", "id": 255267, "pid": 76337, "tid": -914061504, "ts": 1716454225408858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408861, "dur": 0,
    "args": {
      "External id": 255268,
      "cbid": 251, "correlation": 255268
    }
  },
  {
    "ph": "f", "id": 255268, "pid": 76337, "tid": -914061504, "ts": 1716454225408861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225486162, "dur": 8,
    "args": {
      "External id": 255269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255269,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 255269, "pid": 5, "tid": 7, "ts": 1716454225486162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408862, "dur": 12,
    "args": {
      "External id": 255269,
      "cbid": 211, "correlation": 255269
    }
  },
  {
    "ph": "s", "id": 255269, "pid": 76337, "tid": -914061504, "ts": 1716454225408862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225486171, "dur": 4,
    "args": {
      "External id": 255271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255271,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 255271, "pid": 5, "tid": 7, "ts": 1716454225486171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408876, "dur": 6,
    "args": {
      "External id": 255271,
      "cbid": 211, "correlation": 255271
    }
  },
  {
    "ph": "s", "id": 255271, "pid": 76337, "tid": -914061504, "ts": 1716454225408876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225486176, "dur": 94,
    "args": {
      "External id": 255292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255292,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 255292, "pid": 5, "tid": 7, "ts": 1716454225486176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225408949, "dur": 12,
    "args": {
      "External id": 255292,
      "cbid": 211, "correlation": 255292
    }
  },
  {
    "ph": "s", "id": 255292, "pid": 76337, "tid": -914061504, "ts": 1716454225408949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409054, "dur": 1,
    "args": {
      "External id": 255310,
      "cbid": 251, "correlation": 255310
    }
  },
  {
    "ph": "f", "id": 255310, "pid": 76337, "tid": -914061504, "ts": 1716454225409054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225486272, "dur": 97,
    "args": {
      "External id": 255312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255312,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255312, "pid": 5, "tid": 7, "ts": 1716454225486272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409060, "dur": 14,
    "args": {
      "External id": 255312,
      "cbid": 211, "correlation": 255312
    }
  },
  {
    "ph": "s", "id": 255312, "pid": 76337, "tid": -914061504, "ts": 1716454225409060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225486370, "dur": 19,
    "args": {
      "External id": 255320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255320,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255320, "pid": 5, "tid": 7, "ts": 1716454225486370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409129, "dur": 12,
    "args": {
      "External id": 255320,
      "cbid": 211, "correlation": 255320
    }
  },
  {
    "ph": "s", "id": 255320, "pid": 76337, "tid": -914061504, "ts": 1716454225409129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225486390, "dur": 38,
    "args": {
      "External id": 255328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255328, "pid": 5, "tid": 7, "ts": 1716454225486390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409171, "dur": 9,
    "args": {
      "External id": 255328,
      "cbid": 211, "correlation": 255328
    }
  },
  {
    "ph": "s", "id": 255328, "pid": 76337, "tid": -914061504, "ts": 1716454225409171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225486429, "dur": 36,
    "args": {
      "External id": 255350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255350,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255350, "pid": 5, "tid": 7, "ts": 1716454225486429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409221, "dur": 11,
    "args": {
      "External id": 255350,
      "cbid": 211, "correlation": 255350
    }
  },
  {
    "ph": "s", "id": 255350, "pid": 76337, "tid": -914061504, "ts": 1716454225409221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409309, "dur": 1,
    "args": {
      "External id": 255366,
      "cbid": 251, "correlation": 255366
    }
  },
  {
    "ph": "f", "id": 255366, "pid": 76337, "tid": -914061504, "ts": 1716454225409309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409313, "dur": 0,
    "args": {
      "External id": 255368,
      "cbid": 251, "correlation": 255368
    }
  },
  {
    "ph": "f", "id": 255368, "pid": 76337, "tid": -914061504, "ts": 1716454225409313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225486466, "dur": 549,
    "args": {
      "External id": 255369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255369,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 255369, "pid": 5, "tid": 7, "ts": 1716454225486466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409317, "dur": 13,
    "args": {
      "External id": 255369,
      "cbid": 211, "correlation": 255369
    }
  },
  {
    "ph": "s", "id": 255369, "pid": 76337, "tid": -914061504, "ts": 1716454225409317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225487016, "dur": 127,
    "args": {
      "External id": 255377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255377,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255377, "pid": 5, "tid": 7, "ts": 1716454225487016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409382, "dur": 12,
    "args": {
      "External id": 255377,
      "cbid": 211, "correlation": 255377
    }
  },
  {
    "ph": "s", "id": 255377, "pid": 76337, "tid": -914061504, "ts": 1716454225409382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225487145, "dur": 130,
    "args": {
      "External id": 255385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255385,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255385, "pid": 5, "tid": 7, "ts": 1716454225487145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409412, "dur": 9,
    "args": {
      "External id": 255385,
      "cbid": 211, "correlation": 255385
    }
  },
  {
    "ph": "s", "id": 255385, "pid": 76337, "tid": -914061504, "ts": 1716454225409412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409490, "dur": 1,
    "args": {
      "External id": 255401,
      "cbid": 251, "correlation": 255401
    }
  },
  {
    "ph": "f", "id": 255401, "pid": 76337, "tid": -914061504, "ts": 1716454225409490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225487277, "dur": 313,
    "args": {
      "External id": 255403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255403,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255403, "pid": 5, "tid": 7, "ts": 1716454225487277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409496, "dur": 12,
    "args": {
      "External id": 255403,
      "cbid": 211, "correlation": 255403
    }
  },
  {
    "ph": "s", "id": 255403, "pid": 76337, "tid": -914061504, "ts": 1716454225409496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225487591, "dur": 28,
    "args": {
      "External id": 255411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255411,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255411, "pid": 5, "tid": 7, "ts": 1716454225487591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409538, "dur": 10,
    "args": {
      "External id": 255411,
      "cbid": 211, "correlation": 255411
    }
  },
  {
    "ph": "s", "id": 255411, "pid": 76337, "tid": -914061504, "ts": 1716454225409538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225487620, "dur": 83,
    "args": {
      "External id": 255422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255422, "pid": 5, "tid": 7, "ts": 1716454225487620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409605, "dur": 12,
    "args": {
      "External id": 255422,
      "cbid": 211, "correlation": 255422
    }
  },
  {
    "ph": "s", "id": 255422, "pid": 76337, "tid": -914061504, "ts": 1716454225409605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409669, "dur": 0,
    "args": {
      "External id": 255434,
      "cbid": 317, "correlation": 255434
    }
  },
  {
    "ph": "f", "id": 255434, "pid": 76337, "tid": -914061504, "ts": 1716454225409669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409670, "dur": 0,
    "args": {
      "External id": 255435,
      "cbid": 203, "correlation": 255435
    }
  },
  {
    "ph": "f", "id": 255435, "pid": 76337, "tid": -914061504, "ts": 1716454225409670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409672, "dur": 0,
    "args": {
      "External id": 255436,
      "cbid": 205, "correlation": 255436
    }
  },
  {
    "ph": "f", "id": 255436, "pid": 76337, "tid": -914061504, "ts": 1716454225409672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225487704, "dur": 24,
    "args": {
      "External id": 255440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255440,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255440, "pid": 5, "tid": 7, "ts": 1716454225487704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409687, "dur": 12,
    "args": {
      "External id": 255440,
      "cbid": 211, "correlation": 255440
    }
  },
  {
    "ph": "s", "id": 255440, "pid": 76337, "tid": -914061504, "ts": 1716454225409687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225487730, "dur": 123,
    "args": {
      "External id": 255442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255442,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255442, "pid": 5, "tid": 7, "ts": 1716454225487730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409706, "dur": 6,
    "args": {
      "External id": 255442,
      "cbid": 211, "correlation": 255442
    }
  },
  {
    "ph": "s", "id": 255442, "pid": 76337, "tid": -914061504, "ts": 1716454225409706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225487854, "dur": 26,
    "args": {
      "External id": 255444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255444, "pid": 5, "tid": 7, "ts": 1716454225487854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409716, "dur": 5,
    "args": {
      "External id": 255444,
      "cbid": 211, "correlation": 255444
    }
  },
  {
    "ph": "s", "id": 255444, "pid": 76337, "tid": -914061504, "ts": 1716454225409716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225487880, "dur": 34,
    "args": {
      "External id": 255450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255450,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255450, "pid": 5, "tid": 7, "ts": 1716454225487880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409743, "dur": 8,
    "args": {
      "External id": 255450,
      "cbid": 211, "correlation": 255450
    }
  },
  {
    "ph": "s", "id": 255450, "pid": 76337, "tid": -914061504, "ts": 1716454225409743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225487916, "dur": 26,
    "args": {
      "External id": 255458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255458,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255458, "pid": 5, "tid": 7, "ts": 1716454225487916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409776, "dur": 8,
    "args": {
      "External id": 255458,
      "cbid": 211, "correlation": 255458
    }
  },
  {
    "ph": "s", "id": 255458, "pid": 76337, "tid": -914061504, "ts": 1716454225409776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225487943, "dur": 48,
    "args": {
      "External id": 255467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255467,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255467, "pid": 5, "tid": 7, "ts": 1716454225487943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409813, "dur": 11,
    "args": {
      "External id": 255467,
      "cbid": 211, "correlation": 255467
    }
  },
  {
    "ph": "s", "id": 255467, "pid": 76337, "tid": -914061504, "ts": 1716454225409813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225487993, "dur": 42,
    "args": {
      "External id": 255487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255487,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 255487, "pid": 5, "tid": 7, "ts": 1716454225487993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409885, "dur": 12,
    "args": {
      "External id": 255487,
      "cbid": 211, "correlation": 255487
    }
  },
  {
    "ph": "s", "id": 255487, "pid": 76337, "tid": -914061504, "ts": 1716454225409885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225488036, "dur": 4,
    "args": {
      "External id": 255499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255499,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 255499, "pid": 5, "tid": 7, "ts": 1716454225488036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409906, "dur": 6,
    "args": {
      "External id": 255499,
      "cbid": 211, "correlation": 255499
    }
  },
  {
    "ph": "s", "id": 255499, "pid": 76337, "tid": -914061504, "ts": 1716454225409906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225488042, "dur": 43,
    "args": {
      "External id": 255502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255502, "pid": 5, "tid": 7, "ts": 1716454225488042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409926, "dur": 7,
    "args": {
      "External id": 255502,
      "cbid": 211, "correlation": 255502
    }
  },
  {
    "ph": "s", "id": 255502, "pid": 76337, "tid": -914061504, "ts": 1716454225409926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225488087, "dur": 29,
    "args": {
      "External id": 255511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255511,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255511, "pid": 5, "tid": 7, "ts": 1716454225488087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225409964, "dur": 18,
    "args": {
      "External id": 255511,
      "cbid": 211, "correlation": 255511
    }
  },
  {
    "ph": "s", "id": 255511, "pid": 76337, "tid": -914061504, "ts": 1716454225409964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410025, "dur": 0,
    "args": {
      "External id": 255521,
      "cbid": 317, "correlation": 255521
    }
  },
  {
    "ph": "f", "id": 255521, "pid": 76337, "tid": -914061504, "ts": 1716454225410025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410026, "dur": 0,
    "args": {
      "External id": 255522,
      "cbid": 203, "correlation": 255522
    }
  },
  {
    "ph": "f", "id": 255522, "pid": 76337, "tid": -914061504, "ts": 1716454225410026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410027, "dur": 0,
    "args": {
      "External id": 255523,
      "cbid": 205, "correlation": 255523
    }
  },
  {
    "ph": "f", "id": 255523, "pid": 76337, "tid": -914061504, "ts": 1716454225410027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225488117, "dur": 31,
    "args": {
      "External id": 255527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255527,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255527, "pid": 5, "tid": 7, "ts": 1716454225488117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410041, "dur": 12,
    "args": {
      "External id": 255527,
      "cbid": 211, "correlation": 255527
    }
  },
  {
    "ph": "s", "id": 255527, "pid": 76337, "tid": -914061504, "ts": 1716454225410041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225488149, "dur": 64,
    "args": {
      "External id": 255529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255529,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255529, "pid": 5, "tid": 7, "ts": 1716454225488149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410055, "dur": 5,
    "args": {
      "External id": 255529,
      "cbid": 211, "correlation": 255529
    }
  },
  {
    "ph": "s", "id": 255529, "pid": 76337, "tid": -914061504, "ts": 1716454225410055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225488215, "dur": 980,
    "args": {
      "External id": 255531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255531,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255531, "pid": 5, "tid": 7, "ts": 1716454225488215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410067, "dur": 6,
    "args": {
      "External id": 255531,
      "cbid": 211, "correlation": 255531
    }
  },
  {
    "ph": "s", "id": 255531, "pid": 76337, "tid": -914061504, "ts": 1716454225410067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225489196, "dur": 22,
    "args": {
      "External id": 255533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255533,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255533, "pid": 5, "tid": 7, "ts": 1716454225489196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410077, "dur": 5,
    "args": {
      "External id": 255533,
      "cbid": 211, "correlation": 255533
    }
  },
  {
    "ph": "s", "id": 255533, "pid": 76337, "tid": -914061504, "ts": 1716454225410077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225489219, "dur": 33,
    "args": {
      "External id": 255539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255539,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255539, "pid": 5, "tid": 7, "ts": 1716454225489219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410105, "dur": 9,
    "args": {
      "External id": 255539,
      "cbid": 211, "correlation": 255539
    }
  },
  {
    "ph": "s", "id": 255539, "pid": 76337, "tid": -914061504, "ts": 1716454225410105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225489254, "dur": 3,
    "args": {
      "External id": 255547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255547,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 255547, "pid": 5, "tid": 7, "ts": 1716454225489254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410149, "dur": 9,
    "args": {
      "External id": 255547,
      "cbid": 211, "correlation": 255547
    }
  },
  {
    "ph": "s", "id": 255547, "pid": 76337, "tid": -914061504, "ts": 1716454225410149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410216, "dur": 1,
    "args": {
      "External id": 255563,
      "cbid": 251, "correlation": 255563
    }
  },
  {
    "ph": "f", "id": 255563, "pid": 76337, "tid": -914061504, "ts": 1716454225410216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410222, "dur": 0,
    "args": {
      "External id": 255565,
      "cbid": 251, "correlation": 255565
    }
  },
  {
    "ph": "f", "id": 255565, "pid": 76337, "tid": -914061504, "ts": 1716454225410222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225489258, "dur": 12,
    "args": {
      "External id": 255566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255566,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 255566, "pid": 5, "tid": 7, "ts": 1716454225489258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410224, "dur": 11,
    "args": {
      "External id": 255566,
      "cbid": 211, "correlation": 255566
    }
  },
  {
    "ph": "s", "id": 255566, "pid": 76337, "tid": -914061504, "ts": 1716454225410224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225489272, "dur": 5,
    "args": {
      "External id": 255568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255568,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 255568, "pid": 5, "tid": 7, "ts": 1716454225489272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410237, "dur": 5,
    "args": {
      "External id": 255568,
      "cbid": 211, "correlation": 255568
    }
  },
  {
    "ph": "s", "id": 255568, "pid": 76337, "tid": -914061504, "ts": 1716454225410237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225489279, "dur": 29,
    "args": {
      "External id": 255578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255578, "pid": 5, "tid": 7, "ts": 1716454225489279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410294, "dur": 12,
    "args": {
      "External id": 255578,
      "cbid": 211, "correlation": 255578
    }
  },
  {
    "ph": "s", "id": 255578, "pid": 76337, "tid": -914061504, "ts": 1716454225410294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225489309, "dur": 32,
    "args": {
      "External id": 255598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255598,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 255598, "pid": 5, "tid": 7, "ts": 1716454225489309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410359, "dur": 11,
    "args": {
      "External id": 255598,
      "cbid": 211, "correlation": 255598
    }
  },
  {
    "ph": "s", "id": 255598, "pid": 76337, "tid": -914061504, "ts": 1716454225410359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225489343, "dur": 4,
    "args": {
      "External id": 255610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255610,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 255610, "pid": 5, "tid": 7, "ts": 1716454225489343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410380, "dur": 6,
    "args": {
      "External id": 255610,
      "cbid": 211, "correlation": 255610
    }
  },
  {
    "ph": "s", "id": 255610, "pid": 76337, "tid": -914061504, "ts": 1716454225410380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225489349, "dur": 30,
    "args": {
      "External id": 255613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255613,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255613, "pid": 5, "tid": 7, "ts": 1716454225489349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410399, "dur": 6,
    "args": {
      "External id": 255613,
      "cbid": 211, "correlation": 255613
    }
  },
  {
    "ph": "s", "id": 255613, "pid": 76337, "tid": -914061504, "ts": 1716454225410399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225489380, "dur": 20,
    "args": {
      "External id": 255622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255622,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255622, "pid": 5, "tid": 7, "ts": 1716454225489380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410439, "dur": 10,
    "args": {
      "External id": 255622,
      "cbid": 211, "correlation": 255622
    }
  },
  {
    "ph": "s", "id": 255622, "pid": 76337, "tid": -914061504, "ts": 1716454225410439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410501, "dur": 0,
    "args": {
      "External id": 255632,
      "cbid": 317, "correlation": 255632
    }
  },
  {
    "ph": "f", "id": 255632, "pid": 76337, "tid": -914061504, "ts": 1716454225410501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410502, "dur": 0,
    "args": {
      "External id": 255633,
      "cbid": 203, "correlation": 255633
    }
  },
  {
    "ph": "f", "id": 255633, "pid": 76337, "tid": -914061504, "ts": 1716454225410502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410502, "dur": 0,
    "args": {
      "External id": 255634,
      "cbid": 205, "correlation": 255634
    }
  },
  {
    "ph": "f", "id": 255634, "pid": 76337, "tid": -914061504, "ts": 1716454225410502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225489402, "dur": 23,
    "args": {
      "External id": 255638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255638,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255638, "pid": 5, "tid": 7, "ts": 1716454225489402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410516, "dur": 12,
    "args": {
      "External id": 255638,
      "cbid": 211, "correlation": 255638
    }
  },
  {
    "ph": "s", "id": 255638, "pid": 76337, "tid": -914061504, "ts": 1716454225410516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225489426, "dur": 45,
    "args": {
      "External id": 255640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255640,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255640, "pid": 5, "tid": 7, "ts": 1716454225489426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410531, "dur": 5,
    "args": {
      "External id": 255640,
      "cbid": 211, "correlation": 255640
    }
  },
  {
    "ph": "s", "id": 255640, "pid": 76337, "tid": -914061504, "ts": 1716454225410531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225489473, "dur": 654,
    "args": {
      "External id": 255642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255642,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255642, "pid": 5, "tid": 7, "ts": 1716454225489473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410542, "dur": 5,
    "args": {
      "External id": 255642,
      "cbid": 211, "correlation": 255642
    }
  },
  {
    "ph": "s", "id": 255642, "pid": 76337, "tid": -914061504, "ts": 1716454225410542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490128, "dur": 22,
    "args": {
      "External id": 255644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255644,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255644, "pid": 5, "tid": 7, "ts": 1716454225490128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410551, "dur": 5,
    "args": {
      "External id": 255644,
      "cbid": 211, "correlation": 255644
    }
  },
  {
    "ph": "s", "id": 255644, "pid": 76337, "tid": -914061504, "ts": 1716454225410551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225490152, "dur": 34,
    "args": {
      "External id": 255650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255650, "pid": 5, "tid": 7, "ts": 1716454225490152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410579, "dur": 8,
    "args": {
      "External id": 255650,
      "cbid": 211, "correlation": 255650
    }
  },
  {
    "ph": "s", "id": 255650, "pid": 76337, "tid": -914061504, "ts": 1716454225410579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410636, "dur": 0,
    "args": {
      "External id": 255660,
      "cbid": 317, "correlation": 255660
    }
  },
  {
    "ph": "f", "id": 255660, "pid": 76337, "tid": -914061504, "ts": 1716454225410636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410637, "dur": 0,
    "args": {
      "External id": 255661,
      "cbid": 203, "correlation": 255661
    }
  },
  {
    "ph": "f", "id": 255661, "pid": 76337, "tid": -914061504, "ts": 1716454225410637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410638, "dur": 0,
    "args": {
      "External id": 255662,
      "cbid": 205, "correlation": 255662
    }
  },
  {
    "ph": "f", "id": 255662, "pid": 76337, "tid": -914061504, "ts": 1716454225410638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490187, "dur": 30,
    "args": {
      "External id": 255666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255666, "pid": 5, "tid": 7, "ts": 1716454225490187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410651, "dur": 12,
    "args": {
      "External id": 255666,
      "cbid": 211, "correlation": 255666
    }
  },
  {
    "ph": "s", "id": 255666, "pid": 76337, "tid": -914061504, "ts": 1716454225410651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225490218, "dur": 155,
    "args": {
      "External id": 255668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255668,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255668, "pid": 5, "tid": 7, "ts": 1716454225490218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410669, "dur": 6,
    "args": {
      "External id": 255668,
      "cbid": 211, "correlation": 255668
    }
  },
  {
    "ph": "s", "id": 255668, "pid": 76337, "tid": -914061504, "ts": 1716454225410669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490374, "dur": 23,
    "args": {
      "External id": 255670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255670, "pid": 5, "tid": 7, "ts": 1716454225490374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410679, "dur": 5,
    "args": {
      "External id": 255670,
      "cbid": 211, "correlation": 255670
    }
  },
  {
    "ph": "s", "id": 255670, "pid": 76337, "tid": -914061504, "ts": 1716454225410679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225490398, "dur": 33,
    "args": {
      "External id": 255676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255676, "pid": 5, "tid": 7, "ts": 1716454225490398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410705, "dur": 8,
    "args": {
      "External id": 255676,
      "cbid": 211, "correlation": 255676
    }
  },
  {
    "ph": "s", "id": 255676, "pid": 76337, "tid": -914061504, "ts": 1716454225410705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225490432, "dur": 27,
    "args": {
      "External id": 255684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255684,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255684, "pid": 5, "tid": 7, "ts": 1716454225490432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410734, "dur": 7,
    "args": {
      "External id": 255684,
      "cbid": 211, "correlation": 255684
    }
  },
  {
    "ph": "s", "id": 255684, "pid": 76337, "tid": -914061504, "ts": 1716454225410734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225490461, "dur": 20,
    "args": {
      "External id": 255692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255692,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255692, "pid": 5, "tid": 7, "ts": 1716454225490461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410762, "dur": 8,
    "args": {
      "External id": 255692,
      "cbid": 211, "correlation": 255692
    }
  },
  {
    "ph": "s", "id": 255692, "pid": 76337, "tid": -914061504, "ts": 1716454225410762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490481, "dur": 31,
    "args": {
      "External id": 255712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255712,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 255712, "pid": 5, "tid": 7, "ts": 1716454225490481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410843, "dur": 13,
    "args": {
      "External id": 255712,
      "cbid": 211, "correlation": 255712
    }
  },
  {
    "ph": "s", "id": 255712, "pid": 76337, "tid": -914061504, "ts": 1716454225410843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225490514, "dur": 5,
    "args": {
      "External id": 255724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255724,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 255724, "pid": 5, "tid": 7, "ts": 1716454225490514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410865, "dur": 6,
    "args": {
      "External id": 255724,
      "cbid": 211, "correlation": 255724
    }
  },
  {
    "ph": "s", "id": 255724, "pid": 76337, "tid": -914061504, "ts": 1716454225410865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225490520, "dur": 31,
    "args": {
      "External id": 255727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255727, "pid": 5, "tid": 7, "ts": 1716454225490520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410883, "dur": 6,
    "args": {
      "External id": 255727,
      "cbid": 211, "correlation": 255727
    }
  },
  {
    "ph": "s", "id": 255727, "pid": 76337, "tid": -914061504, "ts": 1716454225410883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410940, "dur": 0,
    "args": {
      "External id": 255738,
      "cbid": 317, "correlation": 255738
    }
  },
  {
    "ph": "f", "id": 255738, "pid": 76337, "tid": -914061504, "ts": 1716454225410940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410940, "dur": 0,
    "args": {
      "External id": 255739,
      "cbid": 203, "correlation": 255739
    }
  },
  {
    "ph": "f", "id": 255739, "pid": 76337, "tid": -914061504, "ts": 1716454225410940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410941, "dur": 0,
    "args": {
      "External id": 255740,
      "cbid": 205, "correlation": 255740
    }
  },
  {
    "ph": "f", "id": 255740, "pid": 76337, "tid": -914061504, "ts": 1716454225410941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490553, "dur": 24,
    "args": {
      "External id": 255744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255744, "pid": 5, "tid": 7, "ts": 1716454225490553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410954, "dur": 12,
    "args": {
      "External id": 255744,
      "cbid": 211, "correlation": 255744
    }
  },
  {
    "ph": "s", "id": 255744, "pid": 76337, "tid": -914061504, "ts": 1716454225410954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225490578, "dur": 107,
    "args": {
      "External id": 255746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255746,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255746, "pid": 5, "tid": 7, "ts": 1716454225490578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410972, "dur": 15,
    "args": {
      "External id": 255746,
      "cbid": 211, "correlation": 255746
    }
  },
  {
    "ph": "s", "id": 255746, "pid": 76337, "tid": -914061504, "ts": 1716454225410972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490686, "dur": 23,
    "args": {
      "External id": 255748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255748,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255748, "pid": 5, "tid": 7, "ts": 1716454225490686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225410990, "dur": 5,
    "args": {
      "External id": 255748,
      "cbid": 211, "correlation": 255748
    }
  },
  {
    "ph": "s", "id": 255748, "pid": 76337, "tid": -914061504, "ts": 1716454225410990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225490711, "dur": 33,
    "args": {
      "External id": 255754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255754, "pid": 5, "tid": 7, "ts": 1716454225490711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411018, "dur": 8,
    "args": {
      "External id": 255754,
      "cbid": 211, "correlation": 255754
    }
  },
  {
    "ph": "s", "id": 255754, "pid": 76337, "tid": -914061504, "ts": 1716454225411018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225490746, "dur": 204,
    "args": {
      "External id": 255763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255763,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255763, "pid": 5, "tid": 7, "ts": 1716454225490746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411099, "dur": 14,
    "args": {
      "External id": 255763,
      "cbid": 211, "correlation": 255763
    }
  },
  {
    "ph": "s", "id": 255763, "pid": 76337, "tid": -914061504, "ts": 1716454225411099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225490951, "dur": 67,
    "args": {
      "External id": 255785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255785,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255785, "pid": 5, "tid": 7, "ts": 1716454225490951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411156, "dur": 10,
    "args": {
      "External id": 255785,
      "cbid": 211, "correlation": 255785
    }
  },
  {
    "ph": "s", "id": 255785, "pid": 76337, "tid": -914061504, "ts": 1716454225411156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411244, "dur": 1,
    "args": {
      "External id": 255796,
      "cbid": 251, "correlation": 255796
    }
  },
  {
    "ph": "f", "id": 255796, "pid": 76337, "tid": -914061504, "ts": 1716454225411244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225491019, "dur": 151,
    "args": {
      "External id": 255797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255797,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255797, "pid": 5, "tid": 7, "ts": 1716454225491019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411249, "dur": 13,
    "args": {
      "External id": 255797,
      "cbid": 211, "correlation": 255797
    }
  },
  {
    "ph": "s", "id": 255797, "pid": 76337, "tid": -914061504, "ts": 1716454225411249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411318, "dur": 1,
    "args": {
      "External id": 255808,
      "cbid": 251, "correlation": 255808
    }
  },
  {
    "ph": "f", "id": 255808, "pid": 76337, "tid": -914061504, "ts": 1716454225411318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225491171, "dur": 145,
    "args": {
      "External id": 255809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255809,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255809, "pid": 5, "tid": 7, "ts": 1716454225491171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411322, "dur": 12,
    "args": {
      "External id": 255809,
      "cbid": 211, "correlation": 255809
    }
  },
  {
    "ph": "s", "id": 255809, "pid": 76337, "tid": -914061504, "ts": 1716454225411322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411389, "dur": 1,
    "args": {
      "External id": 255820,
      "cbid": 251, "correlation": 255820
    }
  },
  {
    "ph": "f", "id": 255820, "pid": 76337, "tid": -914061504, "ts": 1716454225411389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225491318, "dur": 146,
    "args": {
      "External id": 255821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255821,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255821, "pid": 5, "tid": 7, "ts": 1716454225491318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411393, "dur": 11,
    "args": {
      "External id": 255821,
      "cbid": 211, "correlation": 255821
    }
  },
  {
    "ph": "s", "id": 255821, "pid": 76337, "tid": -914061504, "ts": 1716454225411393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225491465, "dur": 1988,
    "args": {
      "External id": 255842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255842,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 255842, "pid": 5, "tid": 7, "ts": 1716454225491465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411473, "dur": 13,
    "args": {
      "External id": 255842,
      "cbid": 211, "correlation": 255842
    }
  },
  {
    "ph": "s", "id": 255842, "pid": 76337, "tid": -914061504, "ts": 1716454225411473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411571, "dur": 1,
    "args": {
      "External id": 255860,
      "cbid": 251, "correlation": 255860
    }
  },
  {
    "ph": "f", "id": 255860, "pid": 76337, "tid": -914061504, "ts": 1716454225411571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225493454, "dur": 151,
    "args": {
      "External id": 255862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255862,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 255862, "pid": 5, "tid": 7, "ts": 1716454225493454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411577, "dur": 14,
    "args": {
      "External id": 255862,
      "cbid": 211, "correlation": 255862
    }
  },
  {
    "ph": "s", "id": 255862, "pid": 76337, "tid": -914061504, "ts": 1716454225411577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225493606, "dur": 36,
    "args": {
      "External id": 255870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255870,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255870, "pid": 5, "tid": 7, "ts": 1716454225493606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411647, "dur": 12,
    "args": {
      "External id": 255870,
      "cbid": 211, "correlation": 255870
    }
  },
  {
    "ph": "s", "id": 255870, "pid": 76337, "tid": -914061504, "ts": 1716454225411647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225493644, "dur": 50,
    "args": {
      "External id": 255878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255878,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255878, "pid": 5, "tid": 7, "ts": 1716454225493644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411686, "dur": 8,
    "args": {
      "External id": 255878,
      "cbid": 211, "correlation": 255878
    }
  },
  {
    "ph": "s", "id": 255878, "pid": 76337, "tid": -914061504, "ts": 1716454225411686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225493695, "dur": 31,
    "args": {
      "External id": 255889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255889,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255889, "pid": 5, "tid": 7, "ts": 1716454225493695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411756, "dur": 13,
    "args": {
      "External id": 255889,
      "cbid": 211, "correlation": 255889
    }
  },
  {
    "ph": "s", "id": 255889, "pid": 76337, "tid": -914061504, "ts": 1716454225411756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225493728, "dur": 35,
    "args": {
      "External id": 255911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255911,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 255911, "pid": 5, "tid": 7, "ts": 1716454225493728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411788, "dur": 8,
    "args": {
      "External id": 255911,
      "cbid": 211, "correlation": 255911
    }
  },
  {
    "ph": "s", "id": 255911, "pid": 76337, "tid": -914061504, "ts": 1716454225411788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411872, "dur": 1,
    "args": {
      "External id": 255922,
      "cbid": 251, "correlation": 255922
    }
  },
  {
    "ph": "f", "id": 255922, "pid": 76337, "tid": -914061504, "ts": 1716454225411872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225493764, "dur": 77,
    "args": {
      "External id": 255923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255923,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255923, "pid": 5, "tid": 7, "ts": 1716454225493764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411877, "dur": 14,
    "args": {
      "External id": 255923,
      "cbid": 211, "correlation": 255923
    }
  },
  {
    "ph": "s", "id": 255923, "pid": 76337, "tid": -914061504, "ts": 1716454225411877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411947, "dur": 1,
    "args": {
      "External id": 255934,
      "cbid": 251, "correlation": 255934
    }
  },
  {
    "ph": "f", "id": 255934, "pid": 76337, "tid": -914061504, "ts": 1716454225411947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411950, "dur": 0,
    "args": {
      "External id": 255935,
      "cbid": 251, "correlation": 255935
    }
  },
  {
    "ph": "f", "id": 255935, "pid": 76337, "tid": -914061504, "ts": 1716454225411950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225493843, "dur": 12,
    "args": {
      "External id": 255936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255936,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 255936, "pid": 5, "tid": 7, "ts": 1716454225493843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411952, "dur": 12,
    "args": {
      "External id": 255936,
      "cbid": 211, "correlation": 255936
    }
  },
  {
    "ph": "s", "id": 255936, "pid": 76337, "tid": -914061504, "ts": 1716454225411952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225493856, "dur": 5,
    "args": {
      "External id": 255938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255938,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 255938, "pid": 5, "tid": 7, "ts": 1716454225493856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225411965, "dur": 6,
    "args": {
      "External id": 255938,
      "cbid": 211, "correlation": 255938
    }
  },
  {
    "ph": "s", "id": 255938, "pid": 76337, "tid": -914061504, "ts": 1716454225411965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412032, "dur": 1,
    "args": {
      "External id": 255949,
      "cbid": 251, "correlation": 255949
    }
  },
  {
    "ph": "f", "id": 255949, "pid": 76337, "tid": -914061504, "ts": 1716454225412032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412036, "dur": 0,
    "args": {
      "External id": 255950,
      "cbid": 251, "correlation": 255950
    }
  },
  {
    "ph": "f", "id": 255950, "pid": 76337, "tid": -914061504, "ts": 1716454225412036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225493862, "dur": 7,
    "args": {
      "External id": 255951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255951,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 255951, "pid": 5, "tid": 7, "ts": 1716454225493862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412038, "dur": 12,
    "args": {
      "External id": 255951,
      "cbid": 211, "correlation": 255951
    }
  },
  {
    "ph": "s", "id": 255951, "pid": 76337, "tid": -914061504, "ts": 1716454225412038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225493871, "dur": 3,
    "args": {
      "External id": 255953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255953,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 255953, "pid": 5, "tid": 7, "ts": 1716454225493871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412051, "dur": 5,
    "args": {
      "External id": 255953,
      "cbid": 211, "correlation": 255953
    }
  },
  {
    "ph": "s", "id": 255953, "pid": 76337, "tid": -914061504, "ts": 1716454225412051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225493875, "dur": 93,
    "args": {
      "External id": 255974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255974,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 255974, "pid": 5, "tid": 7, "ts": 1716454225493875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412125, "dur": 14,
    "args": {
      "External id": 255974,
      "cbid": 211, "correlation": 255974
    }
  },
  {
    "ph": "s", "id": 255974, "pid": 76337, "tid": -914061504, "ts": 1716454225412125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412221, "dur": 1,
    "args": {
      "External id": 255992,
      "cbid": 251, "correlation": 255992
    }
  },
  {
    "ph": "f", "id": 255992, "pid": 76337, "tid": -914061504, "ts": 1716454225412221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225493970, "dur": 97,
    "args": {
      "External id": 255994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 255994,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 255994, "pid": 5, "tid": 7, "ts": 1716454225493970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412227, "dur": 14,
    "args": {
      "External id": 255994,
      "cbid": 211, "correlation": 255994
    }
  },
  {
    "ph": "s", "id": 255994, "pid": 76337, "tid": -914061504, "ts": 1716454225412227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225494068, "dur": 20,
    "args": {
      "External id": 256002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256002,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256002, "pid": 5, "tid": 7, "ts": 1716454225494068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412297, "dur": 12,
    "args": {
      "External id": 256002,
      "cbid": 211, "correlation": 256002
    }
  },
  {
    "ph": "s", "id": 256002, "pid": 76337, "tid": -914061504, "ts": 1716454225412297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225494089, "dur": 38,
    "args": {
      "External id": 256010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256010,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256010, "pid": 5, "tid": 7, "ts": 1716454225494089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412338, "dur": 9,
    "args": {
      "External id": 256010,
      "cbid": 211, "correlation": 256010
    }
  },
  {
    "ph": "s", "id": 256010, "pid": 76337, "tid": -914061504, "ts": 1716454225412338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225494129, "dur": 36,
    "args": {
      "External id": 256032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256032,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256032, "pid": 5, "tid": 7, "ts": 1716454225494129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412389, "dur": 10,
    "args": {
      "External id": 256032,
      "cbid": 211, "correlation": 256032
    }
  },
  {
    "ph": "s", "id": 256032, "pid": 76337, "tid": -914061504, "ts": 1716454225412389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412477, "dur": 1,
    "args": {
      "External id": 256048,
      "cbid": 251, "correlation": 256048
    }
  },
  {
    "ph": "f", "id": 256048, "pid": 76337, "tid": -914061504, "ts": 1716454225412477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412482, "dur": 0,
    "args": {
      "External id": 256050,
      "cbid": 251, "correlation": 256050
    }
  },
  {
    "ph": "f", "id": 256050, "pid": 76337, "tid": -914061504, "ts": 1716454225412482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225494166, "dur": 550,
    "args": {
      "External id": 256051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256051,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 256051, "pid": 5, "tid": 7, "ts": 1716454225494166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412486, "dur": 13,
    "args": {
      "External id": 256051,
      "cbid": 211, "correlation": 256051
    }
  },
  {
    "ph": "s", "id": 256051, "pid": 76337, "tid": -914061504, "ts": 1716454225412486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225494717, "dur": 128,
    "args": {
      "External id": 256059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256059, "pid": 5, "tid": 7, "ts": 1716454225494717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412551, "dur": 12,
    "args": {
      "External id": 256059,
      "cbid": 211, "correlation": 256059
    }
  },
  {
    "ph": "s", "id": 256059, "pid": 76337, "tid": -914061504, "ts": 1716454225412551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225494846, "dur": 128,
    "args": {
      "External id": 256067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256067, "pid": 5, "tid": 7, "ts": 1716454225494846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412580, "dur": 9,
    "args": {
      "External id": 256067,
      "cbid": 211, "correlation": 256067
    }
  },
  {
    "ph": "s", "id": 256067, "pid": 76337, "tid": -914061504, "ts": 1716454225412580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412657, "dur": 1,
    "args": {
      "External id": 256083,
      "cbid": 251, "correlation": 256083
    }
  },
  {
    "ph": "f", "id": 256083, "pid": 76337, "tid": -914061504, "ts": 1716454225412657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225494976, "dur": 315,
    "args": {
      "External id": 256085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256085,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256085, "pid": 5, "tid": 7, "ts": 1716454225494976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412663, "dur": 12,
    "args": {
      "External id": 256085,
      "cbid": 211, "correlation": 256085
    }
  },
  {
    "ph": "s", "id": 256085, "pid": 76337, "tid": -914061504, "ts": 1716454225412663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225495292, "dur": 27,
    "args": {
      "External id": 256093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256093,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256093, "pid": 5, "tid": 7, "ts": 1716454225495292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412705, "dur": 10,
    "args": {
      "External id": 256093,
      "cbid": 211, "correlation": 256093
    }
  },
  {
    "ph": "s", "id": 256093, "pid": 76337, "tid": -914061504, "ts": 1716454225412705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225495321, "dur": 83,
    "args": {
      "External id": 256104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256104, "pid": 5, "tid": 7, "ts": 1716454225495321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412772, "dur": 12,
    "args": {
      "External id": 256104,
      "cbid": 211, "correlation": 256104
    }
  },
  {
    "ph": "s", "id": 256104, "pid": 76337, "tid": -914061504, "ts": 1716454225412772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412835, "dur": 0,
    "args": {
      "External id": 256116,
      "cbid": 317, "correlation": 256116
    }
  },
  {
    "ph": "f", "id": 256116, "pid": 76337, "tid": -914061504, "ts": 1716454225412835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412836, "dur": 0,
    "args": {
      "External id": 256117,
      "cbid": 203, "correlation": 256117
    }
  },
  {
    "ph": "f", "id": 256117, "pid": 76337, "tid": -914061504, "ts": 1716454225412836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412837, "dur": 0,
    "args": {
      "External id": 256118,
      "cbid": 205, "correlation": 256118
    }
  },
  {
    "ph": "f", "id": 256118, "pid": 76337, "tid": -914061504, "ts": 1716454225412837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225495405, "dur": 23,
    "args": {
      "External id": 256122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256122, "pid": 5, "tid": 7, "ts": 1716454225495405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412851, "dur": 12,
    "args": {
      "External id": 256122,
      "cbid": 211, "correlation": 256122
    }
  },
  {
    "ph": "s", "id": 256122, "pid": 76337, "tid": -914061504, "ts": 1716454225412851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225495429, "dur": 122,
    "args": {
      "External id": 256124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256124,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256124, "pid": 5, "tid": 7, "ts": 1716454225495429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412870, "dur": 6,
    "args": {
      "External id": 256124,
      "cbid": 211, "correlation": 256124
    }
  },
  {
    "ph": "s", "id": 256124, "pid": 76337, "tid": -914061504, "ts": 1716454225412870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225495552, "dur": 24,
    "args": {
      "External id": 256126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256126,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256126, "pid": 5, "tid": 7, "ts": 1716454225495552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412880, "dur": 5,
    "args": {
      "External id": 256126,
      "cbid": 211, "correlation": 256126
    }
  },
  {
    "ph": "s", "id": 256126, "pid": 76337, "tid": -914061504, "ts": 1716454225412880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225495578, "dur": 33,
    "args": {
      "External id": 256132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256132, "pid": 5, "tid": 7, "ts": 1716454225495578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412908, "dur": 9,
    "args": {
      "External id": 256132,
      "cbid": 211, "correlation": 256132
    }
  },
  {
    "ph": "s", "id": 256132, "pid": 76337, "tid": -914061504, "ts": 1716454225412908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225495613, "dur": 27,
    "args": {
      "External id": 256140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256140,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256140, "pid": 5, "tid": 7, "ts": 1716454225495613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225412940, "dur": 8,
    "args": {
      "External id": 256140,
      "cbid": 211, "correlation": 256140
    }
  },
  {
    "ph": "s", "id": 256140, "pid": 76337, "tid": -914061504, "ts": 1716454225412940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225495641, "dur": 103,
    "args": {
      "External id": 256151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256151,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256151, "pid": 5, "tid": 7, "ts": 1716454225495641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413009, "dur": 12,
    "args": {
      "External id": 256151,
      "cbid": 211, "correlation": 256151
    }
  },
  {
    "ph": "s", "id": 256151, "pid": 76337, "tid": -914061504, "ts": 1716454225413009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413065, "dur": 0,
    "args": {
      "External id": 256161,
      "cbid": 317, "correlation": 256161
    }
  },
  {
    "ph": "f", "id": 256161, "pid": 76337, "tid": -914061504, "ts": 1716454225413065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413066, "dur": 0,
    "args": {
      "External id": 256162,
      "cbid": 203, "correlation": 256162
    }
  },
  {
    "ph": "f", "id": 256162, "pid": 76337, "tid": -914061504, "ts": 1716454225413066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413066, "dur": 0,
    "args": {
      "External id": 256163,
      "cbid": 205, "correlation": 256163
    }
  },
  {
    "ph": "f", "id": 256163, "pid": 76337, "tid": -914061504, "ts": 1716454225413066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225495745, "dur": 75,
    "args": {
      "External id": 256167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256167,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256167, "pid": 5, "tid": 7, "ts": 1716454225495745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413080, "dur": 12,
    "args": {
      "External id": 256167,
      "cbid": 211, "correlation": 256167
    }
  },
  {
    "ph": "s", "id": 256167, "pid": 76337, "tid": -914061504, "ts": 1716454225413080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225495821, "dur": 45,
    "args": {
      "External id": 256169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256169,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256169, "pid": 5, "tid": 7, "ts": 1716454225495821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413095, "dur": 5,
    "args": {
      "External id": 256169,
      "cbid": 211, "correlation": 256169
    }
  },
  {
    "ph": "s", "id": 256169, "pid": 76337, "tid": -914061504, "ts": 1716454225413095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225495867, "dur": 4,
    "args": {
      "External id": 256171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256171,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256171, "pid": 5, "tid": 7, "ts": 1716454225495867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413104, "dur": 6,
    "args": {
      "External id": 256171,
      "cbid": 211, "correlation": 256171
    }
  },
  {
    "ph": "s", "id": 256171, "pid": 76337, "tid": -914061504, "ts": 1716454225413104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413113, "dur": 0,
    "args": {
      "External id": 256172,
      "cbid": 51, "correlation": 256172
    }
  },
  {
    "ph": "s", "id": 256172, "pid": 76337, "tid": -914061504, "ts": 1716454225413113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225495872, "dur": 2239,
    "args": {
      "External id": 256173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256173,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256173, "pid": 5, "tid": 7, "ts": 1716454225495872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413114, "dur": 5,
    "args": {
      "External id": 256173,
      "cbid": 211, "correlation": 256173
    }
  },
  {
    "ph": "s", "id": 256173, "pid": 76337, "tid": -914061504, "ts": 1716454225413114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225498112, "dur": 114,
    "args": {
      "External id": 256178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256178, "pid": 5, "tid": 7, "ts": 1716454225498112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413143, "dur": 9,
    "args": {
      "External id": 256178,
      "cbid": 211, "correlation": 256178
    }
  },
  {
    "ph": "s", "id": 256178, "pid": 76337, "tid": -914061504, "ts": 1716454225413143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225498227, "dur": 167,
    "args": {
      "External id": 256187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256187,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256187, "pid": 5, "tid": 7, "ts": 1716454225498227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413233, "dur": 14,
    "args": {
      "External id": 256187,
      "cbid": 211, "correlation": 256187
    }
  },
  {
    "ph": "s", "id": 256187, "pid": 76337, "tid": -914061504, "ts": 1716454225413233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225498395, "dur": 128,
    "args": {
      "External id": 256207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256207,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 256207, "pid": 5, "tid": 7, "ts": 1716454225498395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413303, "dur": 11,
    "args": {
      "External id": 256207,
      "cbid": 211, "correlation": 256207
    }
  },
  {
    "ph": "s", "id": 256207, "pid": 76337, "tid": -914061504, "ts": 1716454225413303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225498525, "dur": 5,
    "args": {
      "External id": 256219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256219,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 256219, "pid": 5, "tid": 7, "ts": 1716454225498525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413324, "dur": 6,
    "args": {
      "External id": 256219,
      "cbid": 211, "correlation": 256219
    }
  },
  {
    "ph": "s", "id": 256219, "pid": 76337, "tid": -914061504, "ts": 1716454225413324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225498531, "dur": 161,
    "args": {
      "External id": 256222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256222, "pid": 5, "tid": 7, "ts": 1716454225498531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413344, "dur": 7,
    "args": {
      "External id": 256222,
      "cbid": 211, "correlation": 256222
    }
  },
  {
    "ph": "s", "id": 256222, "pid": 76337, "tid": -914061504, "ts": 1716454225413344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225498694, "dur": 102,
    "args": {
      "External id": 256231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256231,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256231, "pid": 5, "tid": 7, "ts": 1716454225498694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413383, "dur": 11,
    "args": {
      "External id": 256231,
      "cbid": 211, "correlation": 256231
    }
  },
  {
    "ph": "s", "id": 256231, "pid": 76337, "tid": -914061504, "ts": 1716454225413383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413436, "dur": 0,
    "args": {
      "External id": 256241,
      "cbid": 317, "correlation": 256241
    }
  },
  {
    "ph": "f", "id": 256241, "pid": 76337, "tid": -914061504, "ts": 1716454225413436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413437, "dur": 0,
    "args": {
      "External id": 256242,
      "cbid": 203, "correlation": 256242
    }
  },
  {
    "ph": "f", "id": 256242, "pid": 76337, "tid": -914061504, "ts": 1716454225413437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413438, "dur": 0,
    "args": {
      "External id": 256243,
      "cbid": 205, "correlation": 256243
    }
  },
  {
    "ph": "f", "id": 256243, "pid": 76337, "tid": -914061504, "ts": 1716454225413438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225498796, "dur": 110,
    "args": {
      "External id": 256247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256247,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256247, "pid": 5, "tid": 7, "ts": 1716454225498796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413455, "dur": 11,
    "args": {
      "External id": 256247,
      "cbid": 211, "correlation": 256247
    }
  },
  {
    "ph": "s", "id": 256247, "pid": 76337, "tid": -914061504, "ts": 1716454225413455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225498908, "dur": 34,
    "args": {
      "External id": 256249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256249,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256249, "pid": 5, "tid": 7, "ts": 1716454225498908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413468, "dur": 5,
    "args": {
      "External id": 256249,
      "cbid": 211, "correlation": 256249
    }
  },
  {
    "ph": "s", "id": 256249, "pid": 76337, "tid": -914061504, "ts": 1716454225413468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225498944, "dur": 4,
    "args": {
      "External id": 256251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256251,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256251, "pid": 5, "tid": 7, "ts": 1716454225498944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413478, "dur": 6,
    "args": {
      "External id": 256251,
      "cbid": 211, "correlation": 256251
    }
  },
  {
    "ph": "s", "id": 256251, "pid": 76337, "tid": -914061504, "ts": 1716454225413478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413488, "dur": 0,
    "args": {
      "External id": 256252,
      "cbid": 51, "correlation": 256252
    }
  },
  {
    "ph": "s", "id": 256252, "pid": 76337, "tid": -914061504, "ts": 1716454225413488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225498949, "dur": 2049,
    "args": {
      "External id": 256253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256253,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256253, "pid": 5, "tid": 7, "ts": 1716454225498949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413488, "dur": 5,
    "args": {
      "External id": 256253,
      "cbid": 211, "correlation": 256253
    }
  },
  {
    "ph": "s", "id": 256253, "pid": 76337, "tid": -914061504, "ts": 1716454225413488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225500998, "dur": 60,
    "args": {
      "External id": 256258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256258, "pid": 5, "tid": 7, "ts": 1716454225500998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413517, "dur": 9,
    "args": {
      "External id": 256258,
      "cbid": 211, "correlation": 256258
    }
  },
  {
    "ph": "s", "id": 256258, "pid": 76337, "tid": -914061504, "ts": 1716454225413517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225501059, "dur": 4,
    "args": {
      "External id": 256266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256266,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256266, "pid": 5, "tid": 7, "ts": 1716454225501059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413560, "dur": 9,
    "args": {
      "External id": 256266,
      "cbid": 211, "correlation": 256266
    }
  },
  {
    "ph": "s", "id": 256266, "pid": 76337, "tid": -914061504, "ts": 1716454225413560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413627, "dur": 2,
    "args": {
      "External id": 256282,
      "cbid": 251, "correlation": 256282
    }
  },
  {
    "ph": "f", "id": 256282, "pid": 76337, "tid": -914061504, "ts": 1716454225413627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413633, "dur": 0,
    "args": {
      "External id": 256284,
      "cbid": 251, "correlation": 256284
    }
  },
  {
    "ph": "f", "id": 256284, "pid": 76337, "tid": -914061504, "ts": 1716454225413633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225501064, "dur": 12,
    "args": {
      "External id": 256285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256285,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 256285, "pid": 5, "tid": 7, "ts": 1716454225501064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413635, "dur": 11,
    "args": {
      "External id": 256285,
      "cbid": 211, "correlation": 256285
    }
  },
  {
    "ph": "s", "id": 256285, "pid": 76337, "tid": -914061504, "ts": 1716454225413635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225501077, "dur": 5,
    "args": {
      "External id": 256287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256287,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 256287, "pid": 5, "tid": 7, "ts": 1716454225501077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413649, "dur": 6,
    "args": {
      "External id": 256287,
      "cbid": 211, "correlation": 256287
    }
  },
  {
    "ph": "s", "id": 256287, "pid": 76337, "tid": -914061504, "ts": 1716454225413649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225501084, "dur": 54,
    "args": {
      "External id": 256297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256297,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256297, "pid": 5, "tid": 7, "ts": 1716454225501084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413706, "dur": 12,
    "args": {
      "External id": 256297,
      "cbid": 211, "correlation": 256297
    }
  },
  {
    "ph": "s", "id": 256297, "pid": 76337, "tid": -914061504, "ts": 1716454225413706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225501139, "dur": 53,
    "args": {
      "External id": 256317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256317,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 256317, "pid": 5, "tid": 7, "ts": 1716454225501139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413772, "dur": 11,
    "args": {
      "External id": 256317,
      "cbid": 211, "correlation": 256317
    }
  },
  {
    "ph": "s", "id": 256317, "pid": 76337, "tid": -914061504, "ts": 1716454225413772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225501194, "dur": 4,
    "args": {
      "External id": 256329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256329,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256329, "pid": 5, "tid": 7, "ts": 1716454225501194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413793, "dur": 6,
    "args": {
      "External id": 256329,
      "cbid": 211, "correlation": 256329
    }
  },
  {
    "ph": "s", "id": 256329, "pid": 76337, "tid": -914061504, "ts": 1716454225413793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225501199, "dur": 56,
    "args": {
      "External id": 256332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256332,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256332, "pid": 5, "tid": 7, "ts": 1716454225501199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413811, "dur": 6,
    "args": {
      "External id": 256332,
      "cbid": 211, "correlation": 256332
    }
  },
  {
    "ph": "s", "id": 256332, "pid": 76337, "tid": -914061504, "ts": 1716454225413811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225501256, "dur": 36,
    "args": {
      "External id": 256341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256341,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256341, "pid": 5, "tid": 7, "ts": 1716454225501256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413851, "dur": 10,
    "args": {
      "External id": 256341,
      "cbid": 211, "correlation": 256341
    }
  },
  {
    "ph": "s", "id": 256341, "pid": 76337, "tid": -914061504, "ts": 1716454225413851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413914, "dur": 0,
    "args": {
      "External id": 256351,
      "cbid": 317, "correlation": 256351
    }
  },
  {
    "ph": "f", "id": 256351, "pid": 76337, "tid": -914061504, "ts": 1716454225413914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413915, "dur": 0,
    "args": {
      "External id": 256352,
      "cbid": 203, "correlation": 256352
    }
  },
  {
    "ph": "f", "id": 256352, "pid": 76337, "tid": -914061504, "ts": 1716454225413915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413915, "dur": 0,
    "args": {
      "External id": 256353,
      "cbid": 205, "correlation": 256353
    }
  },
  {
    "ph": "f", "id": 256353, "pid": 76337, "tid": -914061504, "ts": 1716454225413915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225501294, "dur": 41,
    "args": {
      "External id": 256357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256357,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256357, "pid": 5, "tid": 7, "ts": 1716454225501294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413932, "dur": 13,
    "args": {
      "External id": 256357,
      "cbid": 211, "correlation": 256357
    }
  },
  {
    "ph": "s", "id": 256357, "pid": 76337, "tid": -914061504, "ts": 1716454225413932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225501336, "dur": 14,
    "args": {
      "External id": 256359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256359,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256359, "pid": 5, "tid": 7, "ts": 1716454225501336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413947, "dur": 5,
    "args": {
      "External id": 256359,
      "cbid": 211, "correlation": 256359
    }
  },
  {
    "ph": "s", "id": 256359, "pid": 76337, "tid": -914061504, "ts": 1716454225413947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225501351, "dur": 3,
    "args": {
      "External id": 256361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256361, "pid": 5, "tid": 7, "ts": 1716454225501351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413957, "dur": 5,
    "args": {
      "External id": 256361,
      "cbid": 211, "correlation": 256361
    }
  },
  {
    "ph": "s", "id": 256361, "pid": 76337, "tid": -914061504, "ts": 1716454225413957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413965, "dur": 0,
    "args": {
      "External id": 256362,
      "cbid": 51, "correlation": 256362
    }
  },
  {
    "ph": "s", "id": 256362, "pid": 76337, "tid": -914061504, "ts": 1716454225413965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225501356, "dur": 713,
    "args": {
      "External id": 256363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256363,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256363, "pid": 5, "tid": 7, "ts": 1716454225501356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225413965, "dur": 5,
    "args": {
      "External id": 256363,
      "cbid": 211, "correlation": 256363
    }
  },
  {
    "ph": "s", "id": 256363, "pid": 76337, "tid": -914061504, "ts": 1716454225413965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225502071, "dur": 60,
    "args": {
      "External id": 256368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256368,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256368, "pid": 5, "tid": 7, "ts": 1716454225502071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414002, "dur": 9,
    "args": {
      "External id": 256368,
      "cbid": 211, "correlation": 256368
    }
  },
  {
    "ph": "s", "id": 256368, "pid": 76337, "tid": -914061504, "ts": 1716454225414002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414061, "dur": 0,
    "args": {
      "External id": 256378,
      "cbid": 317, "correlation": 256378
    }
  },
  {
    "ph": "f", "id": 256378, "pid": 76337, "tid": -914061504, "ts": 1716454225414061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414062, "dur": 0,
    "args": {
      "External id": 256379,
      "cbid": 203, "correlation": 256379
    }
  },
  {
    "ph": "f", "id": 256379, "pid": 76337, "tid": -914061504, "ts": 1716454225414062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414062, "dur": 0,
    "args": {
      "External id": 256380,
      "cbid": 205, "correlation": 256380
    }
  },
  {
    "ph": "f", "id": 256380, "pid": 76337, "tid": -914061504, "ts": 1716454225414062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225502132, "dur": 4,
    "args": {
      "External id": 256384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256384, "pid": 5, "tid": 7, "ts": 1716454225502132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414078, "dur": 12,
    "args": {
      "External id": 256384,
      "cbid": 211, "correlation": 256384
    }
  },
  {
    "ph": "s", "id": 256384, "pid": 76337, "tid": -914061504, "ts": 1716454225414078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414095, "dur": 0,
    "args": {
      "External id": 256385,
      "cbid": 51, "correlation": 256385
    }
  },
  {
    "ph": "s", "id": 256385, "pid": 76337, "tid": -914061504, "ts": 1716454225414095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225502137, "dur": 270,
    "args": {
      "External id": 256386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256386,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256386, "pid": 5, "tid": 7, "ts": 1716454225502137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414096, "dur": 7,
    "args": {
      "External id": 256386,
      "cbid": 211, "correlation": 256386
    }
  },
  {
    "ph": "s", "id": 256386, "pid": 76337, "tid": -914061504, "ts": 1716454225414096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225502409, "dur": 61,
    "args": {
      "External id": 256391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256391, "pid": 5, "tid": 7, "ts": 1716454225502409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414123, "dur": 8,
    "args": {
      "External id": 256391,
      "cbid": 211, "correlation": 256391
    }
  },
  {
    "ph": "s", "id": 256391, "pid": 76337, "tid": -914061504, "ts": 1716454225414123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225502471, "dur": 50,
    "args": {
      "External id": 256399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256399,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256399, "pid": 5, "tid": 7, "ts": 1716454225502471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414152, "dur": 8,
    "args": {
      "External id": 256399,
      "cbid": 211, "correlation": 256399
    }
  },
  {
    "ph": "s", "id": 256399, "pid": 76337, "tid": -914061504, "ts": 1716454225414152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225502522, "dur": 36,
    "args": {
      "External id": 256407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256407,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256407, "pid": 5, "tid": 7, "ts": 1716454225502522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414181, "dur": 9,
    "args": {
      "External id": 256407,
      "cbid": 211, "correlation": 256407
    }
  },
  {
    "ph": "s", "id": 256407, "pid": 76337, "tid": -914061504, "ts": 1716454225414181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225502559, "dur": 53,
    "args": {
      "External id": 256427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256427,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 256427, "pid": 5, "tid": 7, "ts": 1716454225502559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414262, "dur": 12,
    "args": {
      "External id": 256427,
      "cbid": 211, "correlation": 256427
    }
  },
  {
    "ph": "s", "id": 256427, "pid": 76337, "tid": -914061504, "ts": 1716454225414262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225502613, "dur": 4,
    "args": {
      "External id": 256439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256439,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256439, "pid": 5, "tid": 7, "ts": 1716454225502613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414284, "dur": 7,
    "args": {
      "External id": 256439,
      "cbid": 211, "correlation": 256439
    }
  },
  {
    "ph": "s", "id": 256439, "pid": 76337, "tid": -914061504, "ts": 1716454225414284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225502618, "dur": 55,
    "args": {
      "External id": 256442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256442, "pid": 5, "tid": 7, "ts": 1716454225502618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414302, "dur": 6,
    "args": {
      "External id": 256442,
      "cbid": 211, "correlation": 256442
    }
  },
  {
    "ph": "s", "id": 256442, "pid": 76337, "tid": -914061504, "ts": 1716454225414302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414358, "dur": 0,
    "args": {
      "External id": 256453,
      "cbid": 317, "correlation": 256453
    }
  },
  {
    "ph": "f", "id": 256453, "pid": 76337, "tid": -914061504, "ts": 1716454225414358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414359, "dur": 0,
    "args": {
      "External id": 256454,
      "cbid": 203, "correlation": 256454
    }
  },
  {
    "ph": "f", "id": 256454, "pid": 76337, "tid": -914061504, "ts": 1716454225414359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414360, "dur": 0,
    "args": {
      "External id": 256455,
      "cbid": 205, "correlation": 256455
    }
  },
  {
    "ph": "f", "id": 256455, "pid": 76337, "tid": -914061504, "ts": 1716454225414360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414388, "dur": 1,
    "args": {
      "External id": 256459,
      "cbid": 251, "correlation": 256459
    }
  },
  {
    "ph": "f", "id": 256459, "pid": 76337, "tid": -914061504, "ts": 1716454225414388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414391, "dur": 0,
    "args": {
      "External id": 256460,
      "cbid": 251, "correlation": 256460
    }
  },
  {
    "ph": "f", "id": 256460, "pid": 76337, "tid": -914061504, "ts": 1716454225414391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414392, "dur": 0,
    "args": {
      "External id": 256461,
      "cbid": 251, "correlation": 256461
    }
  },
  {
    "ph": "f", "id": 256461, "pid": 76337, "tid": -914061504, "ts": 1716454225414392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414393, "dur": 0,
    "args": {
      "External id": 256462,
      "cbid": 251, "correlation": 256462
    }
  },
  {
    "ph": "f", "id": 256462, "pid": 76337, "tid": -914061504, "ts": 1716454225414393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414395, "dur": 1,
    "args": {
      "External id": 256463,
      "cbid": 251, "correlation": 256463
    }
  },
  {
    "ph": "f", "id": 256463, "pid": 76337, "tid": -914061504, "ts": 1716454225414395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414397, "dur": 1,
    "args": {
      "External id": 256464,
      "cbid": 251, "correlation": 256464
    }
  },
  {
    "ph": "f", "id": 256464, "pid": 76337, "tid": -914061504, "ts": 1716454225414397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414399, "dur": 1,
    "args": {
      "External id": 256465,
      "cbid": 251, "correlation": 256465
    }
  },
  {
    "ph": "f", "id": 256465, "pid": 76337, "tid": -914061504, "ts": 1716454225414399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414400, "dur": 1,
    "args": {
      "External id": 256466,
      "cbid": 251, "correlation": 256466
    }
  },
  {
    "ph": "f", "id": 256466, "pid": 76337, "tid": -914061504, "ts": 1716454225414400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414402, "dur": 0,
    "args": {
      "External id": 256467,
      "cbid": 251, "correlation": 256467
    }
  },
  {
    "ph": "f", "id": 256467, "pid": 76337, "tid": -914061504, "ts": 1716454225414402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225502675, "dur": 114,
    "args": {
      "External id": 256468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256468,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 256468, "pid": 5, "tid": 7, "ts": 1716454225502675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414405, "dur": 12,
    "args": {
      "External id": 256468,
      "cbid": 211, "correlation": 256468
    }
  },
  {
    "ph": "s", "id": 256468, "pid": 76337, "tid": -914061504, "ts": 1716454225414405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225502790, "dur": 60,
    "args": {
      "External id": 256474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256474, "pid": 5, "tid": 7, "ts": 1716454225502790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414441, "dur": 9,
    "args": {
      "External id": 256474,
      "cbid": 211, "correlation": 256474
    }
  },
  {
    "ph": "s", "id": 256474, "pid": 76337, "tid": -914061504, "ts": 1716454225414441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225502852, "dur": 499,
    "args": {
      "External id": 256483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256483,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256483, "pid": 5, "tid": 7, "ts": 1716454225502852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414522, "dur": 14,
    "args": {
      "External id": 256483,
      "cbid": 211, "correlation": 256483
    }
  },
  {
    "ph": "s", "id": 256483, "pid": 76337, "tid": -914061504, "ts": 1716454225414522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225503352, "dur": 184,
    "args": {
      "External id": 256505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256505,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256505, "pid": 5, "tid": 7, "ts": 1716454225503352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414579, "dur": 10,
    "args": {
      "External id": 256505,
      "cbid": 211, "correlation": 256505
    }
  },
  {
    "ph": "s", "id": 256505, "pid": 76337, "tid": -914061504, "ts": 1716454225414579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414667, "dur": 1,
    "args": {
      "External id": 256516,
      "cbid": 251, "correlation": 256516
    }
  },
  {
    "ph": "f", "id": 256516, "pid": 76337, "tid": -914061504, "ts": 1716454225414667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225503537, "dur": 198,
    "args": {
      "External id": 256517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256517,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256517, "pid": 5, "tid": 7, "ts": 1716454225503537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414672, "dur": 14,
    "args": {
      "External id": 256517,
      "cbid": 211, "correlation": 256517
    }
  },
  {
    "ph": "s", "id": 256517, "pid": 76337, "tid": -914061504, "ts": 1716454225414672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414742, "dur": 1,
    "args": {
      "External id": 256528,
      "cbid": 251, "correlation": 256528
    }
  },
  {
    "ph": "f", "id": 256528, "pid": 76337, "tid": -914061504, "ts": 1716454225414742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225503737, "dur": 189,
    "args": {
      "External id": 256529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256529,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256529, "pid": 5, "tid": 7, "ts": 1716454225503737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414746, "dur": 11,
    "args": {
      "External id": 256529,
      "cbid": 211, "correlation": 256529
    }
  },
  {
    "ph": "s", "id": 256529, "pid": 76337, "tid": -914061504, "ts": 1716454225414746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414817, "dur": 1,
    "args": {
      "External id": 256540,
      "cbid": 251, "correlation": 256540
    }
  },
  {
    "ph": "f", "id": 256540, "pid": 76337, "tid": -914061504, "ts": 1716454225414817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225503928, "dur": 189,
    "args": {
      "External id": 256541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256541,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256541, "pid": 5, "tid": 7, "ts": 1716454225503928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414821, "dur": 13,
    "args": {
      "External id": 256541,
      "cbid": 211, "correlation": 256541
    }
  },
  {
    "ph": "s", "id": 256541, "pid": 76337, "tid": -914061504, "ts": 1716454225414821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225504118, "dur": 18959,
    "args": {
      "External id": 256562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256562,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 256562, "pid": 5, "tid": 7, "ts": 1716454225504118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225414904, "dur": 14,
    "args": {
      "External id": 256562,
      "cbid": 211, "correlation": 256562
    }
  },
  {
    "ph": "s", "id": 256562, "pid": 76337, "tid": -914061504, "ts": 1716454225414904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415008, "dur": 1,
    "args": {
      "External id": 256580,
      "cbid": 251, "correlation": 256580
    }
  },
  {
    "ph": "f", "id": 256580, "pid": 76337, "tid": -914061504, "ts": 1716454225415008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225523078, "dur": 204,
    "args": {
      "External id": 256582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256582,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256582, "pid": 5, "tid": 7, "ts": 1716454225523078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415014, "dur": 14,
    "args": {
      "External id": 256582,
      "cbid": 211, "correlation": 256582
    }
  },
  {
    "ph": "s", "id": 256582, "pid": 76337, "tid": -914061504, "ts": 1716454225415014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225523284, "dur": 67,
    "args": {
      "External id": 256590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256590,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256590, "pid": 5, "tid": 7, "ts": 1716454225523284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415086, "dur": 12,
    "args": {
      "External id": 256590,
      "cbid": 211, "correlation": 256590
    }
  },
  {
    "ph": "s", "id": 256590, "pid": 76337, "tid": -914061504, "ts": 1716454225415086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225523352, "dur": 97,
    "args": {
      "External id": 256598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256598,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256598, "pid": 5, "tid": 7, "ts": 1716454225523352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415124, "dur": 9,
    "args": {
      "External id": 256598,
      "cbid": 211, "correlation": 256598
    }
  },
  {
    "ph": "s", "id": 256598, "pid": 76337, "tid": -914061504, "ts": 1716454225415124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225523450, "dur": 55,
    "args": {
      "External id": 256609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256609, "pid": 5, "tid": 7, "ts": 1716454225523450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415197, "dur": 12,
    "args": {
      "External id": 256609,
      "cbid": 211, "correlation": 256609
    }
  },
  {
    "ph": "s", "id": 256609, "pid": 76337, "tid": -914061504, "ts": 1716454225415197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225523506, "dur": 94,
    "args": {
      "External id": 256631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256631,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256631, "pid": 5, "tid": 7, "ts": 1716454225523506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415229, "dur": 8,
    "args": {
      "External id": 256631,
      "cbid": 211, "correlation": 256631
    }
  },
  {
    "ph": "s", "id": 256631, "pid": 76337, "tid": -914061504, "ts": 1716454225415229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415312, "dur": 1,
    "args": {
      "External id": 256642,
      "cbid": 251, "correlation": 256642
    }
  },
  {
    "ph": "f", "id": 256642, "pid": 76337, "tid": -914061504, "ts": 1716454225415312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225523602, "dur": 108,
    "args": {
      "External id": 256643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256643,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256643, "pid": 5, "tid": 7, "ts": 1716454225523602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415317, "dur": 13,
    "args": {
      "External id": 256643,
      "cbid": 211, "correlation": 256643
    }
  },
  {
    "ph": "s", "id": 256643, "pid": 76337, "tid": -914061504, "ts": 1716454225415317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415391, "dur": 1,
    "args": {
      "External id": 256654,
      "cbid": 251, "correlation": 256654
    }
  },
  {
    "ph": "f", "id": 256654, "pid": 76337, "tid": -914061504, "ts": 1716454225415391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415395, "dur": 0,
    "args": {
      "External id": 256655,
      "cbid": 251, "correlation": 256655
    }
  },
  {
    "ph": "f", "id": 256655, "pid": 76337, "tid": -914061504, "ts": 1716454225415395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225523711, "dur": 10,
    "args": {
      "External id": 256656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256656,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 256656, "pid": 5, "tid": 7, "ts": 1716454225523711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415397, "dur": 13,
    "args": {
      "External id": 256656,
      "cbid": 211, "correlation": 256656
    }
  },
  {
    "ph": "s", "id": 256656, "pid": 76337, "tid": -914061504, "ts": 1716454225415397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225523723, "dur": 5,
    "args": {
      "External id": 256658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256658,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 256658, "pid": 5, "tid": 7, "ts": 1716454225523723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415413, "dur": 7,
    "args": {
      "External id": 256658,
      "cbid": 211, "correlation": 256658
    }
  },
  {
    "ph": "s", "id": 256658, "pid": 76337, "tid": -914061504, "ts": 1716454225415413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415475, "dur": 1,
    "args": {
      "External id": 256669,
      "cbid": 251, "correlation": 256669
    }
  },
  {
    "ph": "f", "id": 256669, "pid": 76337, "tid": -914061504, "ts": 1716454225415475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415479, "dur": 0,
    "args": {
      "External id": 256670,
      "cbid": 251, "correlation": 256670
    }
  },
  {
    "ph": "f", "id": 256670, "pid": 76337, "tid": -914061504, "ts": 1716454225415479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225523729, "dur": 6,
    "args": {
      "External id": 256671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256671,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 256671, "pid": 5, "tid": 7, "ts": 1716454225523729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415481, "dur": 11,
    "args": {
      "External id": 256671,
      "cbid": 211, "correlation": 256671
    }
  },
  {
    "ph": "s", "id": 256671, "pid": 76337, "tid": -914061504, "ts": 1716454225415481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225523736, "dur": 4,
    "args": {
      "External id": 256673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256673,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 256673, "pid": 5, "tid": 7, "ts": 1716454225523736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415494, "dur": 5,
    "args": {
      "External id": 256673,
      "cbid": 211, "correlation": 256673
    }
  },
  {
    "ph": "s", "id": 256673, "pid": 76337, "tid": -914061504, "ts": 1716454225415494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225523741, "dur": 157,
    "args": {
      "External id": 256694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256694,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 256694, "pid": 5, "tid": 7, "ts": 1716454225523741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415567, "dur": 13,
    "args": {
      "External id": 256694,
      "cbid": 211, "correlation": 256694
    }
  },
  {
    "ph": "s", "id": 256694, "pid": 76337, "tid": -914061504, "ts": 1716454225415567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415662, "dur": 1,
    "args": {
      "External id": 256712,
      "cbid": 251, "correlation": 256712
    }
  },
  {
    "ph": "f", "id": 256712, "pid": 76337, "tid": -914061504, "ts": 1716454225415662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225523900, "dur": 108,
    "args": {
      "External id": 256714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256714,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 256714, "pid": 5, "tid": 7, "ts": 1716454225523900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415668, "dur": 14,
    "args": {
      "External id": 256714,
      "cbid": 211, "correlation": 256714
    }
  },
  {
    "ph": "s", "id": 256714, "pid": 76337, "tid": -914061504, "ts": 1716454225415668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225524009, "dur": 35,
    "args": {
      "External id": 256722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256722,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256722, "pid": 5, "tid": 7, "ts": 1716454225524009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415737, "dur": 11,
    "args": {
      "External id": 256722,
      "cbid": 211, "correlation": 256722
    }
  },
  {
    "ph": "s", "id": 256722, "pid": 76337, "tid": -914061504, "ts": 1716454225415737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225524045, "dur": 68,
    "args": {
      "External id": 256730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256730, "pid": 5, "tid": 7, "ts": 1716454225524045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415777, "dur": 9,
    "args": {
      "External id": 256730,
      "cbid": 211, "correlation": 256730
    }
  },
  {
    "ph": "s", "id": 256730, "pid": 76337, "tid": -914061504, "ts": 1716454225415777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225524115, "dur": 94,
    "args": {
      "External id": 256752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256752,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256752, "pid": 5, "tid": 7, "ts": 1716454225524115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415828, "dur": 11,
    "args": {
      "External id": 256752,
      "cbid": 211, "correlation": 256752
    }
  },
  {
    "ph": "s", "id": 256752, "pid": 76337, "tid": -914061504, "ts": 1716454225415828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415914, "dur": 1,
    "args": {
      "External id": 256768,
      "cbid": 251, "correlation": 256768
    }
  },
  {
    "ph": "f", "id": 256768, "pid": 76337, "tid": -914061504, "ts": 1716454225415914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225524210, "dur": 583,
    "args": {
      "External id": 256770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256770,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256770, "pid": 5, "tid": 7, "ts": 1716454225524210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415920, "dur": 13,
    "args": {
      "External id": 256770,
      "cbid": 211, "correlation": 256770
    }
  },
  {
    "ph": "s", "id": 256770, "pid": 76337, "tid": -914061504, "ts": 1716454225415920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225524794, "dur": 247,
    "args": {
      "External id": 256778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256778, "pid": 5, "tid": 7, "ts": 1716454225524794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225415992, "dur": 12,
    "args": {
      "External id": 256778,
      "cbid": 211, "correlation": 256778
    }
  },
  {
    "ph": "s", "id": 256778, "pid": 76337, "tid": -914061504, "ts": 1716454225415992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225525043, "dur": 256,
    "args": {
      "External id": 256786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256786, "pid": 5, "tid": 7, "ts": 1716454225525043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416023, "dur": 9,
    "args": {
      "External id": 256786,
      "cbid": 211, "correlation": 256786
    }
  },
  {
    "ph": "s", "id": 256786, "pid": 76337, "tid": -914061504, "ts": 1716454225416023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416105, "dur": 1,
    "args": {
      "External id": 256802,
      "cbid": 251, "correlation": 256802
    }
  },
  {
    "ph": "f", "id": 256802, "pid": 76337, "tid": -914061504, "ts": 1716454225416105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416110, "dur": 0,
    "args": {
      "External id": 256804,
      "cbid": 251, "correlation": 256804
    }
  },
  {
    "ph": "f", "id": 256804, "pid": 76337, "tid": -914061504, "ts": 1716454225416110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225525300, "dur": 362,
    "args": {
      "External id": 256805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256805,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 256805, "pid": 5, "tid": 7, "ts": 1716454225525300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416113, "dur": 13,
    "args": {
      "External id": 256805,
      "cbid": 211, "correlation": 256805
    }
  },
  {
    "ph": "s", "id": 256805, "pid": 76337, "tid": -914061504, "ts": 1716454225416113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225525663, "dur": 50,
    "args": {
      "External id": 256813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256813,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256813, "pid": 5, "tid": 7, "ts": 1716454225525663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416155, "dur": 10,
    "args": {
      "External id": 256813,
      "cbid": 211, "correlation": 256813
    }
  },
  {
    "ph": "s", "id": 256813, "pid": 76337, "tid": -914061504, "ts": 1716454225416155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225525714, "dur": 161,
    "args": {
      "External id": 256824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256824, "pid": 5, "tid": 7, "ts": 1716454225525714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416222, "dur": 13,
    "args": {
      "External id": 256824,
      "cbid": 211, "correlation": 256824
    }
  },
  {
    "ph": "s", "id": 256824, "pid": 76337, "tid": -914061504, "ts": 1716454225416222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416287, "dur": 0,
    "args": {
      "External id": 256836,
      "cbid": 317, "correlation": 256836
    }
  },
  {
    "ph": "f", "id": 256836, "pid": 76337, "tid": -914061504, "ts": 1716454225416287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416287, "dur": 0,
    "args": {
      "External id": 256837,
      "cbid": 203, "correlation": 256837
    }
  },
  {
    "ph": "f", "id": 256837, "pid": 76337, "tid": -914061504, "ts": 1716454225416287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416288, "dur": 0,
    "args": {
      "External id": 256838,
      "cbid": 205, "correlation": 256838
    }
  },
  {
    "ph": "f", "id": 256838, "pid": 76337, "tid": -914061504, "ts": 1716454225416288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416312, "dur": 1,
    "args": {
      "External id": 256842,
      "cbid": 251, "correlation": 256842
    }
  },
  {
    "ph": "f", "id": 256842, "pid": 76337, "tid": -914061504, "ts": 1716454225416312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416314, "dur": 0,
    "args": {
      "External id": 256843,
      "cbid": 251, "correlation": 256843
    }
  },
  {
    "ph": "f", "id": 256843, "pid": 76337, "tid": -914061504, "ts": 1716454225416314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416314, "dur": 0,
    "args": {
      "External id": 256844,
      "cbid": 251, "correlation": 256844
    }
  },
  {
    "ph": "f", "id": 256844, "pid": 76337, "tid": -914061504, "ts": 1716454225416314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416315, "dur": 0,
    "args": {
      "External id": 256845,
      "cbid": 251, "correlation": 256845
    }
  },
  {
    "ph": "f", "id": 256845, "pid": 76337, "tid": -914061504, "ts": 1716454225416315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416316, "dur": 0,
    "args": {
      "External id": 256846,
      "cbid": 251, "correlation": 256846
    }
  },
  {
    "ph": "f", "id": 256846, "pid": 76337, "tid": -914061504, "ts": 1716454225416316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416317, "dur": 0,
    "args": {
      "External id": 256847,
      "cbid": 251, "correlation": 256847
    }
  },
  {
    "ph": "f", "id": 256847, "pid": 76337, "tid": -914061504, "ts": 1716454225416317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416318, "dur": 0,
    "args": {
      "External id": 256848,
      "cbid": 251, "correlation": 256848
    }
  },
  {
    "ph": "f", "id": 256848, "pid": 76337, "tid": -914061504, "ts": 1716454225416318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416318, "dur": 0,
    "args": {
      "External id": 256849,
      "cbid": 251, "correlation": 256849
    }
  },
  {
    "ph": "f", "id": 256849, "pid": 76337, "tid": -914061504, "ts": 1716454225416318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416320, "dur": 0,
    "args": {
      "External id": 256850,
      "cbid": 251, "correlation": 256850
    }
  },
  {
    "ph": "f", "id": 256850, "pid": 76337, "tid": -914061504, "ts": 1716454225416320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225525877, "dur": 116,
    "args": {
      "External id": 256851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256851,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 256851, "pid": 5, "tid": 7, "ts": 1716454225525877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416322, "dur": 13,
    "args": {
      "External id": 256851,
      "cbid": 211, "correlation": 256851
    }
  },
  {
    "ph": "s", "id": 256851, "pid": 76337, "tid": -914061504, "ts": 1716454225416322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225525994, "dur": 61,
    "args": {
      "External id": 256857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256857, "pid": 5, "tid": 7, "ts": 1716454225525994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416357, "dur": 9,
    "args": {
      "External id": 256857,
      "cbid": 211, "correlation": 256857
    }
  },
  {
    "ph": "s", "id": 256857, "pid": 76337, "tid": -914061504, "ts": 1716454225416357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225526056, "dur": 50,
    "args": {
      "External id": 256865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256865,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256865, "pid": 5, "tid": 7, "ts": 1716454225526056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416388, "dur": 8,
    "args": {
      "External id": 256865,
      "cbid": 211, "correlation": 256865
    }
  },
  {
    "ph": "s", "id": 256865, "pid": 76337, "tid": -914061504, "ts": 1716454225416388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225526108, "dur": 97,
    "args": {
      "External id": 256874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256874,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256874, "pid": 5, "tid": 7, "ts": 1716454225526108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416426, "dur": 11,
    "args": {
      "External id": 256874,
      "cbid": 211, "correlation": 256874
    }
  },
  {
    "ph": "s", "id": 256874, "pid": 76337, "tid": -914061504, "ts": 1716454225416426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225526207, "dur": 93,
    "args": {
      "External id": 256894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256894,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 256894, "pid": 5, "tid": 7, "ts": 1716454225526207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416497, "dur": 11,
    "args": {
      "External id": 256894,
      "cbid": 211, "correlation": 256894
    }
  },
  {
    "ph": "s", "id": 256894, "pid": 76337, "tid": -914061504, "ts": 1716454225416497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225526301, "dur": 5,
    "args": {
      "External id": 256906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256906,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 256906, "pid": 5, "tid": 7, "ts": 1716454225526301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416518, "dur": 165,
    "args": {
      "External id": 256906,
      "cbid": 211, "correlation": 256906
    }
  },
  {
    "ph": "s", "id": 256906, "pid": 76337, "tid": -914061504, "ts": 1716454225416518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225526307, "dur": 110,
    "args": {
      "External id": 256909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256909, "pid": 5, "tid": 7, "ts": 1716454225526307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416696, "dur": 7,
    "args": {
      "External id": 256909,
      "cbid": 211, "correlation": 256909
    }
  },
  {
    "ph": "s", "id": 256909, "pid": 76337, "tid": -914061504, "ts": 1716454225416696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225526419, "dur": 69,
    "args": {
      "External id": 256918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256918,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256918, "pid": 5, "tid": 7, "ts": 1716454225526419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416736, "dur": 11,
    "args": {
      "External id": 256918,
      "cbid": 211, "correlation": 256918
    }
  },
  {
    "ph": "s", "id": 256918, "pid": 76337, "tid": -914061504, "ts": 1716454225416736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416787, "dur": 0,
    "args": {
      "External id": 256928,
      "cbid": 317, "correlation": 256928
    }
  },
  {
    "ph": "f", "id": 256928, "pid": 76337, "tid": -914061504, "ts": 1716454225416787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416788, "dur": 0,
    "args": {
      "External id": 256929,
      "cbid": 203, "correlation": 256929
    }
  },
  {
    "ph": "f", "id": 256929, "pid": 76337, "tid": -914061504, "ts": 1716454225416788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416789, "dur": 0,
    "args": {
      "External id": 256930,
      "cbid": 205, "correlation": 256930
    }
  },
  {
    "ph": "f", "id": 256930, "pid": 76337, "tid": -914061504, "ts": 1716454225416789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225526489, "dur": 76,
    "args": {
      "External id": 256934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256934,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256934, "pid": 5, "tid": 7, "ts": 1716454225526489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416804, "dur": 11,
    "args": {
      "External id": 256934,
      "cbid": 211, "correlation": 256934
    }
  },
  {
    "ph": "s", "id": 256934, "pid": 76337, "tid": -914061504, "ts": 1716454225416804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225526567, "dur": 24,
    "args": {
      "External id": 256936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256936,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256936, "pid": 5, "tid": 7, "ts": 1716454225526567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416818, "dur": 5,
    "args": {
      "External id": 256936,
      "cbid": 211, "correlation": 256936
    }
  },
  {
    "ph": "s", "id": 256936, "pid": 76337, "tid": -914061504, "ts": 1716454225416818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225526592, "dur": 4,
    "args": {
      "External id": 256938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256938,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256938, "pid": 5, "tid": 7, "ts": 1716454225526592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416828, "dur": 6,
    "args": {
      "External id": 256938,
      "cbid": 211, "correlation": 256938
    }
  },
  {
    "ph": "s", "id": 256938, "pid": 76337, "tid": -914061504, "ts": 1716454225416828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416838, "dur": 0,
    "args": {
      "External id": 256939,
      "cbid": 51, "correlation": 256939
    }
  },
  {
    "ph": "s", "id": 256939, "pid": 76337, "tid": -914061504, "ts": 1716454225416838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225526598, "dur": 1388,
    "args": {
      "External id": 256940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256940,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 256940, "pid": 5, "tid": 7, "ts": 1716454225526598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416839, "dur": 6,
    "args": {
      "External id": 256940,
      "cbid": 211, "correlation": 256940
    }
  },
  {
    "ph": "s", "id": 256940, "pid": 76337, "tid": -914061504, "ts": 1716454225416839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225527987, "dur": 59,
    "args": {
      "External id": 256945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256945, "pid": 5, "tid": 7, "ts": 1716454225527987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416867, "dur": 8,
    "args": {
      "External id": 256945,
      "cbid": 211, "correlation": 256945
    }
  },
  {
    "ph": "s", "id": 256945, "pid": 76337, "tid": -914061504, "ts": 1716454225416867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225528048, "dur": 3,
    "args": {
      "External id": 256953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256953,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 256953, "pid": 5, "tid": 7, "ts": 1716454225528048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416910, "dur": 9,
    "args": {
      "External id": 256953,
      "cbid": 211, "correlation": 256953
    }
  },
  {
    "ph": "s", "id": 256953, "pid": 76337, "tid": -914061504, "ts": 1716454225416910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416982, "dur": 1,
    "args": {
      "External id": 256969,
      "cbid": 251, "correlation": 256969
    }
  },
  {
    "ph": "f", "id": 256969, "pid": 76337, "tid": -914061504, "ts": 1716454225416982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416988, "dur": 0,
    "args": {
      "External id": 256971,
      "cbid": 251, "correlation": 256971
    }
  },
  {
    "ph": "f", "id": 256971, "pid": 76337, "tid": -914061504, "ts": 1716454225416988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225528053, "dur": 11,
    "args": {
      "External id": 256972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256972,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 256972, "pid": 5, "tid": 7, "ts": 1716454225528053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225416990, "dur": 12,
    "args": {
      "External id": 256972,
      "cbid": 211, "correlation": 256972
    }
  },
  {
    "ph": "s", "id": 256972, "pid": 76337, "tid": -914061504, "ts": 1716454225416990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225528065, "dur": 5,
    "args": {
      "External id": 256974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256974,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 256974, "pid": 5, "tid": 7, "ts": 1716454225528065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417004, "dur": 176,
    "args": {
      "External id": 256974,
      "cbid": 211, "correlation": 256974
    }
  },
  {
    "ph": "s", "id": 256974, "pid": 76337, "tid": -914061504, "ts": 1716454225417004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225528071, "dur": 55,
    "args": {
      "External id": 256984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 256984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 256984, "pid": 5, "tid": 7, "ts": 1716454225528071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417234, "dur": 12,
    "args": {
      "External id": 256984,
      "cbid": 211, "correlation": 256984
    }
  },
  {
    "ph": "s", "id": 256984, "pid": 76337, "tid": -914061504, "ts": 1716454225417234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225528128, "dur": 54,
    "args": {
      "External id": 257004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257004,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 257004, "pid": 5, "tid": 7, "ts": 1716454225528128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417300, "dur": 11,
    "args": {
      "External id": 257004,
      "cbid": 211, "correlation": 257004
    }
  },
  {
    "ph": "s", "id": 257004, "pid": 76337, "tid": -914061504, "ts": 1716454225417300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225528183, "dur": 4,
    "args": {
      "External id": 257016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257016,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257016, "pid": 5, "tid": 7, "ts": 1716454225528183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417321, "dur": 6,
    "args": {
      "External id": 257016,
      "cbid": 211, "correlation": 257016
    }
  },
  {
    "ph": "s", "id": 257016, "pid": 76337, "tid": -914061504, "ts": 1716454225417321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225528189, "dur": 56,
    "args": {
      "External id": 257019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257019, "pid": 5, "tid": 7, "ts": 1716454225528189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417340, "dur": 7,
    "args": {
      "External id": 257019,
      "cbid": 211, "correlation": 257019
    }
  },
  {
    "ph": "s", "id": 257019, "pid": 76337, "tid": -914061504, "ts": 1716454225417340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225528246, "dur": 37,
    "args": {
      "External id": 257028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257028,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257028, "pid": 5, "tid": 7, "ts": 1716454225528246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417380, "dur": 9,
    "args": {
      "External id": 257028,
      "cbid": 211, "correlation": 257028
    }
  },
  {
    "ph": "s", "id": 257028, "pid": 76337, "tid": -914061504, "ts": 1716454225417380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417441, "dur": 0,
    "args": {
      "External id": 257038,
      "cbid": 317, "correlation": 257038
    }
  },
  {
    "ph": "f", "id": 257038, "pid": 76337, "tid": -914061504, "ts": 1716454225417441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417442, "dur": 0,
    "args": {
      "External id": 257039,
      "cbid": 203, "correlation": 257039
    }
  },
  {
    "ph": "f", "id": 257039, "pid": 76337, "tid": -914061504, "ts": 1716454225417442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417443, "dur": 0,
    "args": {
      "External id": 257040,
      "cbid": 205, "correlation": 257040
    }
  },
  {
    "ph": "f", "id": 257040, "pid": 76337, "tid": -914061504, "ts": 1716454225417443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225528284, "dur": 41,
    "args": {
      "External id": 257044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257044,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257044, "pid": 5, "tid": 7, "ts": 1716454225528284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417457, "dur": 12,
    "args": {
      "External id": 257044,
      "cbid": 211, "correlation": 257044
    }
  },
  {
    "ph": "s", "id": 257044, "pid": 76337, "tid": -914061504, "ts": 1716454225417457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225528326, "dur": 15,
    "args": {
      "External id": 257046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257046,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257046, "pid": 5, "tid": 7, "ts": 1716454225528326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417472, "dur": 5,
    "args": {
      "External id": 257046,
      "cbid": 211, "correlation": 257046
    }
  },
  {
    "ph": "s", "id": 257046, "pid": 76337, "tid": -914061504, "ts": 1716454225417472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225528342, "dur": 3,
    "args": {
      "External id": 257048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257048,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257048, "pid": 5, "tid": 7, "ts": 1716454225528342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417481, "dur": 5,
    "args": {
      "External id": 257048,
      "cbid": 211, "correlation": 257048
    }
  },
  {
    "ph": "s", "id": 257048, "pid": 76337, "tid": -914061504, "ts": 1716454225417481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417489, "dur": 0,
    "args": {
      "External id": 257049,
      "cbid": 51, "correlation": 257049
    }
  },
  {
    "ph": "s", "id": 257049, "pid": 76337, "tid": -914061504, "ts": 1716454225417489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225528346, "dur": 709,
    "args": {
      "External id": 257050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257050,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257050, "pid": 5, "tid": 7, "ts": 1716454225528346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417490, "dur": 5,
    "args": {
      "External id": 257050,
      "cbid": 211, "correlation": 257050
    }
  },
  {
    "ph": "s", "id": 257050, "pid": 76337, "tid": -914061504, "ts": 1716454225417490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225529057, "dur": 60,
    "args": {
      "External id": 257055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257055, "pid": 5, "tid": 7, "ts": 1716454225529057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417517, "dur": 9,
    "args": {
      "External id": 257055,
      "cbid": 211, "correlation": 257055
    }
  },
  {
    "ph": "s", "id": 257055, "pid": 76337, "tid": -914061504, "ts": 1716454225417517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417575, "dur": 0,
    "args": {
      "External id": 257065,
      "cbid": 317, "correlation": 257065
    }
  },
  {
    "ph": "f", "id": 257065, "pid": 76337, "tid": -914061504, "ts": 1716454225417575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417576, "dur": 0,
    "args": {
      "External id": 257066,
      "cbid": 203, "correlation": 257066
    }
  },
  {
    "ph": "f", "id": 257066, "pid": 76337, "tid": -914061504, "ts": 1716454225417576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417577, "dur": 0,
    "args": {
      "External id": 257067,
      "cbid": 205, "correlation": 257067
    }
  },
  {
    "ph": "f", "id": 257067, "pid": 76337, "tid": -914061504, "ts": 1716454225417577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225529118, "dur": 75,
    "args": {
      "External id": 257071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257071,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257071, "pid": 5, "tid": 7, "ts": 1716454225529118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417589, "dur": 11,
    "args": {
      "External id": 257071,
      "cbid": 211, "correlation": 257071
    }
  },
  {
    "ph": "s", "id": 257071, "pid": 76337, "tid": -914061504, "ts": 1716454225417589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225529195, "dur": 211,
    "args": {
      "External id": 257073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257073,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257073, "pid": 5, "tid": 7, "ts": 1716454225529195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417608, "dur": 9,
    "args": {
      "External id": 257073,
      "cbid": 211, "correlation": 257073
    }
  },
  {
    "ph": "s", "id": 257073, "pid": 76337, "tid": -914061504, "ts": 1716454225417608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225529407, "dur": 38,
    "args": {
      "External id": 257075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257075,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257075, "pid": 5, "tid": 7, "ts": 1716454225529407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225417622, "dur": 502,
    "args": {
      "External id": 257075,
      "cbid": 211, "correlation": 257075
    }
  },
  {
    "ph": "s", "id": 257075, "pid": 76337, "tid": -914061504, "ts": 1716454225417622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225529447, "dur": 61,
    "args": {
      "External id": 257081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257081,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257081, "pid": 5, "tid": 7, "ts": 1716454225529447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418147, "dur": 9,
    "args": {
      "External id": 257081,
      "cbid": 211, "correlation": 257081
    }
  },
  {
    "ph": "s", "id": 257081, "pid": 76337, "tid": -914061504, "ts": 1716454225418147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225529509, "dur": 50,
    "args": {
      "External id": 257089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257089,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257089, "pid": 5, "tid": 7, "ts": 1716454225529509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418177, "dur": 9,
    "args": {
      "External id": 257089,
      "cbid": 211, "correlation": 257089
    }
  },
  {
    "ph": "s", "id": 257089, "pid": 76337, "tid": -914061504, "ts": 1716454225418177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225529560, "dur": 35,
    "args": {
      "External id": 257097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257097,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257097, "pid": 5, "tid": 7, "ts": 1716454225529560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418207, "dur": 29,
    "args": {
      "External id": 257097,
      "cbid": 211, "correlation": 257097
    }
  },
  {
    "ph": "s", "id": 257097, "pid": 76337, "tid": -914061504, "ts": 1716454225418207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225529597, "dur": 53,
    "args": {
      "External id": 257117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257117,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 257117, "pid": 5, "tid": 7, "ts": 1716454225529597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418310, "dur": 12,
    "args": {
      "External id": 257117,
      "cbid": 211, "correlation": 257117
    }
  },
  {
    "ph": "s", "id": 257117, "pid": 76337, "tid": -914061504, "ts": 1716454225418310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225529651, "dur": 4,
    "args": {
      "External id": 257129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257129,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257129, "pid": 5, "tid": 7, "ts": 1716454225529651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418332, "dur": 7,
    "args": {
      "External id": 257129,
      "cbid": 211, "correlation": 257129
    }
  },
  {
    "ph": "s", "id": 257129, "pid": 76337, "tid": -914061504, "ts": 1716454225418332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225529657, "dur": 57,
    "args": {
      "External id": 257132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257132, "pid": 5, "tid": 7, "ts": 1716454225529657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418350, "dur": 6,
    "args": {
      "External id": 257132,
      "cbid": 211, "correlation": 257132
    }
  },
  {
    "ph": "s", "id": 257132, "pid": 76337, "tid": -914061504, "ts": 1716454225418350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418408, "dur": 0,
    "args": {
      "External id": 257143,
      "cbid": 317, "correlation": 257143
    }
  },
  {
    "ph": "f", "id": 257143, "pid": 76337, "tid": -914061504, "ts": 1716454225418408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418409, "dur": 0,
    "args": {
      "External id": 257144,
      "cbid": 203, "correlation": 257144
    }
  },
  {
    "ph": "f", "id": 257144, "pid": 76337, "tid": -914061504, "ts": 1716454225418409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418410, "dur": 0,
    "args": {
      "External id": 257145,
      "cbid": 205, "correlation": 257145
    }
  },
  {
    "ph": "f", "id": 257145, "pid": 76337, "tid": -914061504, "ts": 1716454225418410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418431, "dur": 1,
    "args": {
      "External id": 257149,
      "cbid": 251, "correlation": 257149
    }
  },
  {
    "ph": "f", "id": 257149, "pid": 76337, "tid": -914061504, "ts": 1716454225418431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418433, "dur": 0,
    "args": {
      "External id": 257150,
      "cbid": 251, "correlation": 257150
    }
  },
  {
    "ph": "f", "id": 257150, "pid": 76337, "tid": -914061504, "ts": 1716454225418433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418434, "dur": 0,
    "args": {
      "External id": 257151,
      "cbid": 251, "correlation": 257151
    }
  },
  {
    "ph": "f", "id": 257151, "pid": 76337, "tid": -914061504, "ts": 1716454225418434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418435, "dur": 0,
    "args": {
      "External id": 257152,
      "cbid": 251, "correlation": 257152
    }
  },
  {
    "ph": "f", "id": 257152, "pid": 76337, "tid": -914061504, "ts": 1716454225418435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418436, "dur": 0,
    "args": {
      "External id": 257153,
      "cbid": 251, "correlation": 257153
    }
  },
  {
    "ph": "f", "id": 257153, "pid": 76337, "tid": -914061504, "ts": 1716454225418436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418437, "dur": 0,
    "args": {
      "External id": 257154,
      "cbid": 251, "correlation": 257154
    }
  },
  {
    "ph": "f", "id": 257154, "pid": 76337, "tid": -914061504, "ts": 1716454225418437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418438, "dur": 0,
    "args": {
      "External id": 257155,
      "cbid": 251, "correlation": 257155
    }
  },
  {
    "ph": "f", "id": 257155, "pid": 76337, "tid": -914061504, "ts": 1716454225418438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418438, "dur": 0,
    "args": {
      "External id": 257156,
      "cbid": 251, "correlation": 257156
    }
  },
  {
    "ph": "f", "id": 257156, "pid": 76337, "tid": -914061504, "ts": 1716454225418438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418440, "dur": 0,
    "args": {
      "External id": 257157,
      "cbid": 251, "correlation": 257157
    }
  },
  {
    "ph": "f", "id": 257157, "pid": 76337, "tid": -914061504, "ts": 1716454225418440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225529715, "dur": 113,
    "args": {
      "External id": 257158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257158,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257158, "pid": 5, "tid": 7, "ts": 1716454225529715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418442, "dur": 12,
    "args": {
      "External id": 257158,
      "cbid": 211, "correlation": 257158
    }
  },
  {
    "ph": "s", "id": 257158, "pid": 76337, "tid": -914061504, "ts": 1716454225418442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225529830, "dur": 60,
    "args": {
      "External id": 257164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257164, "pid": 5, "tid": 7, "ts": 1716454225529830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418477, "dur": 9,
    "args": {
      "External id": 257164,
      "cbid": 211, "correlation": 257164
    }
  },
  {
    "ph": "s", "id": 257164, "pid": 76337, "tid": -914061504, "ts": 1716454225418477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225529891, "dur": 530,
    "args": {
      "External id": 257173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257173,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257173, "pid": 5, "tid": 7, "ts": 1716454225529891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418559, "dur": 13,
    "args": {
      "External id": 257173,
      "cbid": 211, "correlation": 257173
    }
  },
  {
    "ph": "s", "id": 257173, "pid": 76337, "tid": -914061504, "ts": 1716454225418559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225530422, "dur": 184,
    "args": {
      "External id": 257195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257195,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257195, "pid": 5, "tid": 7, "ts": 1716454225530422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418616, "dur": 10,
    "args": {
      "External id": 257195,
      "cbid": 211, "correlation": 257195
    }
  },
  {
    "ph": "s", "id": 257195, "pid": 76337, "tid": -914061504, "ts": 1716454225418616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418703, "dur": 1,
    "args": {
      "External id": 257206,
      "cbid": 251, "correlation": 257206
    }
  },
  {
    "ph": "f", "id": 257206, "pid": 76337, "tid": -914061504, "ts": 1716454225418703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225530608, "dur": 199,
    "args": {
      "External id": 257207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257207,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257207, "pid": 5, "tid": 7, "ts": 1716454225530608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418707, "dur": 13,
    "args": {
      "External id": 257207,
      "cbid": 211, "correlation": 257207
    }
  },
  {
    "ph": "s", "id": 257207, "pid": 76337, "tid": -914061504, "ts": 1716454225418707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418775, "dur": 1,
    "args": {
      "External id": 257218,
      "cbid": 251, "correlation": 257218
    }
  },
  {
    "ph": "f", "id": 257218, "pid": 76337, "tid": -914061504, "ts": 1716454225418775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225530808, "dur": 189,
    "args": {
      "External id": 257219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257219,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257219, "pid": 5, "tid": 7, "ts": 1716454225530808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418778, "dur": 12,
    "args": {
      "External id": 257219,
      "cbid": 211, "correlation": 257219
    }
  },
  {
    "ph": "s", "id": 257219, "pid": 76337, "tid": -914061504, "ts": 1716454225418778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418843, "dur": 1,
    "args": {
      "External id": 257230,
      "cbid": 251, "correlation": 257230
    }
  },
  {
    "ph": "f", "id": 257230, "pid": 76337, "tid": -914061504, "ts": 1716454225418843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225530999, "dur": 191,
    "args": {
      "External id": 257231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257231,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257231, "pid": 5, "tid": 7, "ts": 1716454225530999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418848, "dur": 11,
    "args": {
      "External id": 257231,
      "cbid": 211, "correlation": 257231
    }
  },
  {
    "ph": "s", "id": 257231, "pid": 76337, "tid": -914061504, "ts": 1716454225418848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225531191, "dur": 18997,
    "args": {
      "External id": 257252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257252,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257252, "pid": 5, "tid": 7, "ts": 1716454225531191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225418928, "dur": 12,
    "args": {
      "External id": 257252,
      "cbid": 211, "correlation": 257252
    }
  },
  {
    "ph": "s", "id": 257252, "pid": 76337, "tid": -914061504, "ts": 1716454225418928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225419031, "dur": 1,
    "args": {
      "External id": 257270,
      "cbid": 251, "correlation": 257270
    }
  },
  {
    "ph": "f", "id": 257270, "pid": 76337, "tid": -914061504, "ts": 1716454225419031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225550189, "dur": 204,
    "args": {
      "External id": 257272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257272,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257272, "pid": 5, "tid": 7, "ts": 1716454225550189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225419037, "dur": 15,
    "args": {
      "External id": 257272,
      "cbid": 211, "correlation": 257272
    }
  },
  {
    "ph": "s", "id": 257272, "pid": 76337, "tid": -914061504, "ts": 1716454225419037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225550395, "dur": 66,
    "args": {
      "External id": 257280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257280,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257280, "pid": 5, "tid": 7, "ts": 1716454225550395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225419109, "dur": 12,
    "args": {
      "External id": 257280,
      "cbid": 211, "correlation": 257280
    }
  },
  {
    "ph": "s", "id": 257280, "pid": 76337, "tid": -914061504, "ts": 1716454225419109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225550462, "dur": 96,
    "args": {
      "External id": 257288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257288,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257288, "pid": 5, "tid": 7, "ts": 1716454225550462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225419148, "dur": 28,
    "args": {
      "External id": 257288,
      "cbid": 211, "correlation": 257288
    }
  },
  {
    "ph": "s", "id": 257288, "pid": 76337, "tid": -914061504, "ts": 1716454225419148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225550560, "dur": 54,
    "args": {
      "External id": 257299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257299, "pid": 5, "tid": 7, "ts": 1716454225550560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225419239, "dur": 87,
    "args": {
      "External id": 257299,
      "cbid": 211, "correlation": 257299
    }
  },
  {
    "ph": "s", "id": 257299, "pid": 76337, "tid": -914061504, "ts": 1716454225419239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225550615, "dur": 93,
    "args": {
      "External id": 257321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257321,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257321, "pid": 5, "tid": 7, "ts": 1716454225550615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225419346, "dur": 1968,
    "args": {
      "External id": 257321,
      "cbid": 211, "correlation": 257321
    }
  },
  {
    "ph": "s", "id": 257321, "pid": 76337, "tid": -914061504, "ts": 1716454225419346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421391, "dur": 1,
    "args": {
      "External id": 257332,
      "cbid": 251, "correlation": 257332
    }
  },
  {
    "ph": "f", "id": 257332, "pid": 76337, "tid": -914061504, "ts": 1716454225421391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225550710, "dur": 105,
    "args": {
      "External id": 257333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257333,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257333, "pid": 5, "tid": 7, "ts": 1716454225550710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421396, "dur": 71,
    "args": {
      "External id": 257333,
      "cbid": 211, "correlation": 257333
    }
  },
  {
    "ph": "s", "id": 257333, "pid": 76337, "tid": -914061504, "ts": 1716454225421396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421526, "dur": 1,
    "args": {
      "External id": 257344,
      "cbid": 251, "correlation": 257344
    }
  },
  {
    "ph": "f", "id": 257344, "pid": 76337, "tid": -914061504, "ts": 1716454225421526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421530, "dur": 0,
    "args": {
      "External id": 257345,
      "cbid": 251, "correlation": 257345
    }
  },
  {
    "ph": "f", "id": 257345, "pid": 76337, "tid": -914061504, "ts": 1716454225421530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225550817, "dur": 10,
    "args": {
      "External id": 257346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257346,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 257346, "pid": 5, "tid": 7, "ts": 1716454225550817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421531, "dur": 12,
    "args": {
      "External id": 257346,
      "cbid": 211, "correlation": 257346
    }
  },
  {
    "ph": "s", "id": 257346, "pid": 76337, "tid": -914061504, "ts": 1716454225421531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225550828, "dur": 5,
    "args": {
      "External id": 257348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257348,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 257348, "pid": 5, "tid": 7, "ts": 1716454225550828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421545, "dur": 6,
    "args": {
      "External id": 257348,
      "cbid": 211, "correlation": 257348
    }
  },
  {
    "ph": "s", "id": 257348, "pid": 76337, "tid": -914061504, "ts": 1716454225421545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421606, "dur": 1,
    "args": {
      "External id": 257359,
      "cbid": 251, "correlation": 257359
    }
  },
  {
    "ph": "f", "id": 257359, "pid": 76337, "tid": -914061504, "ts": 1716454225421606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421610, "dur": 0,
    "args": {
      "External id": 257360,
      "cbid": 251, "correlation": 257360
    }
  },
  {
    "ph": "f", "id": 257360, "pid": 76337, "tid": -914061504, "ts": 1716454225421610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225550835, "dur": 6,
    "args": {
      "External id": 257361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257361,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 257361, "pid": 5, "tid": 7, "ts": 1716454225550835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421611, "dur": 12,
    "args": {
      "External id": 257361,
      "cbid": 211, "correlation": 257361
    }
  },
  {
    "ph": "s", "id": 257361, "pid": 76337, "tid": -914061504, "ts": 1716454225421611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225550842, "dur": 3,
    "args": {
      "External id": 257363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257363,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 257363, "pid": 5, "tid": 7, "ts": 1716454225550842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421625, "dur": 5,
    "args": {
      "External id": 257363,
      "cbid": 211, "correlation": 257363
    }
  },
  {
    "ph": "s", "id": 257363, "pid": 76337, "tid": -914061504, "ts": 1716454225421625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225550847, "dur": 157,
    "args": {
      "External id": 257384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257384,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257384, "pid": 5, "tid": 7, "ts": 1716454225550847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421698, "dur": 18,
    "args": {
      "External id": 257384,
      "cbid": 211, "correlation": 257384
    }
  },
  {
    "ph": "s", "id": 257384, "pid": 76337, "tid": -914061504, "ts": 1716454225421698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421800, "dur": 1,
    "args": {
      "External id": 257402,
      "cbid": 251, "correlation": 257402
    }
  },
  {
    "ph": "f", "id": 257402, "pid": 76337, "tid": -914061504, "ts": 1716454225421800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225551005, "dur": 108,
    "args": {
      "External id": 257404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257404,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257404, "pid": 5, "tid": 7, "ts": 1716454225551005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421806, "dur": 14,
    "args": {
      "External id": 257404,
      "cbid": 211, "correlation": 257404
    }
  },
  {
    "ph": "s", "id": 257404, "pid": 76337, "tid": -914061504, "ts": 1716454225421806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225551115, "dur": 35,
    "args": {
      "External id": 257412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257412,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257412, "pid": 5, "tid": 7, "ts": 1716454225551115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421876, "dur": 12,
    "args": {
      "External id": 257412,
      "cbid": 211, "correlation": 257412
    }
  },
  {
    "ph": "s", "id": 257412, "pid": 76337, "tid": -914061504, "ts": 1716454225421876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225551151, "dur": 68,
    "args": {
      "External id": 257420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257420, "pid": 5, "tid": 7, "ts": 1716454225551151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421917, "dur": 9,
    "args": {
      "External id": 257420,
      "cbid": 211, "correlation": 257420
    }
  },
  {
    "ph": "s", "id": 257420, "pid": 76337, "tid": -914061504, "ts": 1716454225421917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225551220, "dur": 93,
    "args": {
      "External id": 257442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257442,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257442, "pid": 5, "tid": 7, "ts": 1716454225551220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225421968, "dur": 17,
    "args": {
      "External id": 257442,
      "cbid": 211, "correlation": 257442
    }
  },
  {
    "ph": "s", "id": 257442, "pid": 76337, "tid": -914061504, "ts": 1716454225421968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422061, "dur": 1,
    "args": {
      "External id": 257458,
      "cbid": 251, "correlation": 257458
    }
  },
  {
    "ph": "f", "id": 257458, "pid": 76337, "tid": -914061504, "ts": 1716454225422061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225551315, "dur": 582,
    "args": {
      "External id": 257460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257460,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257460, "pid": 5, "tid": 7, "ts": 1716454225551315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422068, "dur": 14,
    "args": {
      "External id": 257460,
      "cbid": 211, "correlation": 257460
    }
  },
  {
    "ph": "s", "id": 257460, "pid": 76337, "tid": -914061504, "ts": 1716454225422068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225551898, "dur": 245,
    "args": {
      "External id": 257468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257468, "pid": 5, "tid": 7, "ts": 1716454225551898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422134, "dur": 12,
    "args": {
      "External id": 257468,
      "cbid": 211, "correlation": 257468
    }
  },
  {
    "ph": "s", "id": 257468, "pid": 76337, "tid": -914061504, "ts": 1716454225422134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225552145, "dur": 255,
    "args": {
      "External id": 257476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257476,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257476, "pid": 5, "tid": 7, "ts": 1716454225552145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422163, "dur": 9,
    "args": {
      "External id": 257476,
      "cbid": 211, "correlation": 257476
    }
  },
  {
    "ph": "s", "id": 257476, "pid": 76337, "tid": -914061504, "ts": 1716454225422163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422243, "dur": 1,
    "args": {
      "External id": 257492,
      "cbid": 251, "correlation": 257492
    }
  },
  {
    "ph": "f", "id": 257492, "pid": 76337, "tid": -914061504, "ts": 1716454225422243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422248, "dur": 0,
    "args": {
      "External id": 257494,
      "cbid": 251, "correlation": 257494
    }
  },
  {
    "ph": "f", "id": 257494, "pid": 76337, "tid": -914061504, "ts": 1716454225422248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225552402, "dur": 360,
    "args": {
      "External id": 257495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257495,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 257495, "pid": 5, "tid": 7, "ts": 1716454225552402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422251, "dur": 13,
    "args": {
      "External id": 257495,
      "cbid": 211, "correlation": 257495
    }
  },
  {
    "ph": "s", "id": 257495, "pid": 76337, "tid": -914061504, "ts": 1716454225422251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225552763, "dur": 50,
    "args": {
      "External id": 257503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257503,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257503, "pid": 5, "tid": 7, "ts": 1716454225552763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422294, "dur": 10,
    "args": {
      "External id": 257503,
      "cbid": 211, "correlation": 257503
    }
  },
  {
    "ph": "s", "id": 257503, "pid": 76337, "tid": -914061504, "ts": 1716454225422294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225552815, "dur": 161,
    "args": {
      "External id": 257514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257514, "pid": 5, "tid": 7, "ts": 1716454225552815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422361, "dur": 232,
    "args": {
      "External id": 257514,
      "cbid": 211, "correlation": 257514
    }
  },
  {
    "ph": "s", "id": 257514, "pid": 76337, "tid": -914061504, "ts": 1716454225422361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422646, "dur": 0,
    "args": {
      "External id": 257526,
      "cbid": 317, "correlation": 257526
    }
  },
  {
    "ph": "f", "id": 257526, "pid": 76337, "tid": -914061504, "ts": 1716454225422646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422647, "dur": 0,
    "args": {
      "External id": 257527,
      "cbid": 203, "correlation": 257527
    }
  },
  {
    "ph": "f", "id": 257527, "pid": 76337, "tid": -914061504, "ts": 1716454225422647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422648, "dur": 0,
    "args": {
      "External id": 257528,
      "cbid": 205, "correlation": 257528
    }
  },
  {
    "ph": "f", "id": 257528, "pid": 76337, "tid": -914061504, "ts": 1716454225422648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422671, "dur": 1,
    "args": {
      "External id": 257532,
      "cbid": 251, "correlation": 257532
    }
  },
  {
    "ph": "f", "id": 257532, "pid": 76337, "tid": -914061504, "ts": 1716454225422671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422673, "dur": 0,
    "args": {
      "External id": 257533,
      "cbid": 251, "correlation": 257533
    }
  },
  {
    "ph": "f", "id": 257533, "pid": 76337, "tid": -914061504, "ts": 1716454225422673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422673, "dur": 0,
    "args": {
      "External id": 257534,
      "cbid": 251, "correlation": 257534
    }
  },
  {
    "ph": "f", "id": 257534, "pid": 76337, "tid": -914061504, "ts": 1716454225422673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422674, "dur": 0,
    "args": {
      "External id": 257535,
      "cbid": 251, "correlation": 257535
    }
  },
  {
    "ph": "f", "id": 257535, "pid": 76337, "tid": -914061504, "ts": 1716454225422674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422675, "dur": 0,
    "args": {
      "External id": 257536,
      "cbid": 251, "correlation": 257536
    }
  },
  {
    "ph": "f", "id": 257536, "pid": 76337, "tid": -914061504, "ts": 1716454225422675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422675, "dur": 0,
    "args": {
      "External id": 257537,
      "cbid": 251, "correlation": 257537
    }
  },
  {
    "ph": "f", "id": 257537, "pid": 76337, "tid": -914061504, "ts": 1716454225422675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422676, "dur": 0,
    "args": {
      "External id": 257538,
      "cbid": 251, "correlation": 257538
    }
  },
  {
    "ph": "f", "id": 257538, "pid": 76337, "tid": -914061504, "ts": 1716454225422676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422677, "dur": 0,
    "args": {
      "External id": 257539,
      "cbid": 251, "correlation": 257539
    }
  },
  {
    "ph": "f", "id": 257539, "pid": 76337, "tid": -914061504, "ts": 1716454225422677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422678, "dur": 0,
    "args": {
      "External id": 257540,
      "cbid": 251, "correlation": 257540
    }
  },
  {
    "ph": "f", "id": 257540, "pid": 76337, "tid": -914061504, "ts": 1716454225422678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225552977, "dur": 116,
    "args": {
      "External id": 257541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257541,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257541, "pid": 5, "tid": 7, "ts": 1716454225552977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422680, "dur": 42,
    "args": {
      "External id": 257541,
      "cbid": 211, "correlation": 257541
    }
  },
  {
    "ph": "s", "id": 257541, "pid": 76337, "tid": -914061504, "ts": 1716454225422680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225553094, "dur": 60,
    "args": {
      "External id": 257547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257547,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257547, "pid": 5, "tid": 7, "ts": 1716454225553094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422745, "dur": 106,
    "args": {
      "External id": 257547,
      "cbid": 211, "correlation": 257547
    }
  },
  {
    "ph": "s", "id": 257547, "pid": 76337, "tid": -914061504, "ts": 1716454225422745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225553155, "dur": 51,
    "args": {
      "External id": 257555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257555,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257555, "pid": 5, "tid": 7, "ts": 1716454225553155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225422875, "dur": 285,
    "args": {
      "External id": 257555,
      "cbid": 211, "correlation": 257555
    }
  },
  {
    "ph": "s", "id": 257555, "pid": 76337, "tid": -914061504, "ts": 1716454225422875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225553207, "dur": 100,
    "args": {
      "External id": 257564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257564,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257564, "pid": 5, "tid": 7, "ts": 1716454225553207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423192, "dur": 11,
    "args": {
      "External id": 257564,
      "cbid": 211, "correlation": 257564
    }
  },
  {
    "ph": "s", "id": 257564, "pid": 76337, "tid": -914061504, "ts": 1716454225423192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225553309, "dur": 93,
    "args": {
      "External id": 257584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257584,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 257584, "pid": 5, "tid": 7, "ts": 1716454225553309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423263, "dur": 12,
    "args": {
      "External id": 257584,
      "cbid": 211, "correlation": 257584
    }
  },
  {
    "ph": "s", "id": 257584, "pid": 76337, "tid": -914061504, "ts": 1716454225423263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225553403, "dur": 5,
    "args": {
      "External id": 257596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257596,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 257596, "pid": 5, "tid": 7, "ts": 1716454225553403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423284, "dur": 13,
    "args": {
      "External id": 257596,
      "cbid": 211, "correlation": 257596
    }
  },
  {
    "ph": "s", "id": 257596, "pid": 76337, "tid": -914061504, "ts": 1716454225423284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225553409, "dur": 108,
    "args": {
      "External id": 257599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257599, "pid": 5, "tid": 7, "ts": 1716454225553409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423309, "dur": 110,
    "args": {
      "External id": 257599,
      "cbid": 211, "correlation": 257599
    }
  },
  {
    "ph": "s", "id": 257599, "pid": 76337, "tid": -914061504, "ts": 1716454225423309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225553519, "dur": 69,
    "args": {
      "External id": 257608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257608,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257608, "pid": 5, "tid": 7, "ts": 1716454225553519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423452, "dur": 10,
    "args": {
      "External id": 257608,
      "cbid": 211, "correlation": 257608
    }
  },
  {
    "ph": "s", "id": 257608, "pid": 76337, "tid": -914061504, "ts": 1716454225423452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423504, "dur": 0,
    "args": {
      "External id": 257618,
      "cbid": 317, "correlation": 257618
    }
  },
  {
    "ph": "f", "id": 257618, "pid": 76337, "tid": -914061504, "ts": 1716454225423504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423505, "dur": 0,
    "args": {
      "External id": 257619,
      "cbid": 203, "correlation": 257619
    }
  },
  {
    "ph": "f", "id": 257619, "pid": 76337, "tid": -914061504, "ts": 1716454225423505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423505, "dur": 0,
    "args": {
      "External id": 257620,
      "cbid": 205, "correlation": 257620
    }
  },
  {
    "ph": "f", "id": 257620, "pid": 76337, "tid": -914061504, "ts": 1716454225423505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225553589, "dur": 76,
    "args": {
      "External id": 257624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257624, "pid": 5, "tid": 7, "ts": 1716454225553589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423520, "dur": 11,
    "args": {
      "External id": 257624,
      "cbid": 211, "correlation": 257624
    }
  },
  {
    "ph": "s", "id": 257624, "pid": 76337, "tid": -914061504, "ts": 1716454225423520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225553666, "dur": 24,
    "args": {
      "External id": 257626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257626, "pid": 5, "tid": 7, "ts": 1716454225553666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423534, "dur": 5,
    "args": {
      "External id": 257626,
      "cbid": 211, "correlation": 257626
    }
  },
  {
    "ph": "s", "id": 257626, "pid": 76337, "tid": -914061504, "ts": 1716454225423534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225553692, "dur": 4,
    "args": {
      "External id": 257628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257628, "pid": 5, "tid": 7, "ts": 1716454225553692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423544, "dur": 6,
    "args": {
      "External id": 257628,
      "cbid": 211, "correlation": 257628
    }
  },
  {
    "ph": "s", "id": 257628, "pid": 76337, "tid": -914061504, "ts": 1716454225423544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423553, "dur": 0,
    "args": {
      "External id": 257629,
      "cbid": 51, "correlation": 257629
    }
  },
  {
    "ph": "s", "id": 257629, "pid": 76337, "tid": -914061504, "ts": 1716454225423553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225553697, "dur": 1383,
    "args": {
      "External id": 257630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257630,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257630, "pid": 5, "tid": 7, "ts": 1716454225553697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423554, "dur": 5,
    "args": {
      "External id": 257630,
      "cbid": 211, "correlation": 257630
    }
  },
  {
    "ph": "s", "id": 257630, "pid": 76337, "tid": -914061504, "ts": 1716454225423554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225555082, "dur": 60,
    "args": {
      "External id": 257635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257635, "pid": 5, "tid": 7, "ts": 1716454225555082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423581, "dur": 9,
    "args": {
      "External id": 257635,
      "cbid": 211, "correlation": 257635
    }
  },
  {
    "ph": "s", "id": 257635, "pid": 76337, "tid": -914061504, "ts": 1716454225423581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225555143, "dur": 4,
    "args": {
      "External id": 257643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257643,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257643, "pid": 5, "tid": 7, "ts": 1716454225555143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423624, "dur": 9,
    "args": {
      "External id": 257643,
      "cbid": 211, "correlation": 257643
    }
  },
  {
    "ph": "s", "id": 257643, "pid": 76337, "tid": -914061504, "ts": 1716454225423624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423690, "dur": 1,
    "args": {
      "External id": 257659,
      "cbid": 251, "correlation": 257659
    }
  },
  {
    "ph": "f", "id": 257659, "pid": 76337, "tid": -914061504, "ts": 1716454225423690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423695, "dur": 0,
    "args": {
      "External id": 257661,
      "cbid": 251, "correlation": 257661
    }
  },
  {
    "ph": "f", "id": 257661, "pid": 76337, "tid": -914061504, "ts": 1716454225423695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225555148, "dur": 12,
    "args": {
      "External id": 257662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257662,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 257662, "pid": 5, "tid": 7, "ts": 1716454225555148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423697, "dur": 11,
    "args": {
      "External id": 257662,
      "cbid": 211, "correlation": 257662
    }
  },
  {
    "ph": "s", "id": 257662, "pid": 76337, "tid": -914061504, "ts": 1716454225423697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225555161, "dur": 5,
    "args": {
      "External id": 257664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257664,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 257664, "pid": 5, "tid": 7, "ts": 1716454225555161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423711, "dur": 6,
    "args": {
      "External id": 257664,
      "cbid": 211, "correlation": 257664
    }
  },
  {
    "ph": "s", "id": 257664, "pid": 76337, "tid": -914061504, "ts": 1716454225423711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225555168, "dur": 55,
    "args": {
      "External id": 257674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257674, "pid": 5, "tid": 7, "ts": 1716454225555168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225423769, "dur": 572,
    "args": {
      "External id": 257674,
      "cbid": 211, "correlation": 257674
    }
  },
  {
    "ph": "s", "id": 257674, "pid": 76337, "tid": -914061504, "ts": 1716454225423769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225555224, "dur": 51,
    "args": {
      "External id": 257694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257694,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 257694, "pid": 5, "tid": 7, "ts": 1716454225555224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424397, "dur": 11,
    "args": {
      "External id": 257694,
      "cbid": 211, "correlation": 257694
    }
  },
  {
    "ph": "s", "id": 257694, "pid": 76337, "tid": -914061504, "ts": 1716454225424397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225555276, "dur": 4,
    "args": {
      "External id": 257706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257706,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257706, "pid": 5, "tid": 7, "ts": 1716454225555276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424417, "dur": 6,
    "args": {
      "External id": 257706,
      "cbid": 211, "correlation": 257706
    }
  },
  {
    "ph": "s", "id": 257706, "pid": 76337, "tid": -914061504, "ts": 1716454225424417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225555282, "dur": 56,
    "args": {
      "External id": 257709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257709, "pid": 5, "tid": 7, "ts": 1716454225555282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424436, "dur": 6,
    "args": {
      "External id": 257709,
      "cbid": 211, "correlation": 257709
    }
  },
  {
    "ph": "s", "id": 257709, "pid": 76337, "tid": -914061504, "ts": 1716454225424436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225555339, "dur": 37,
    "args": {
      "External id": 257718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257718,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257718, "pid": 5, "tid": 7, "ts": 1716454225555339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424475, "dur": 10,
    "args": {
      "External id": 257718,
      "cbid": 211, "correlation": 257718
    }
  },
  {
    "ph": "s", "id": 257718, "pid": 76337, "tid": -914061504, "ts": 1716454225424475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424538, "dur": 0,
    "args": {
      "External id": 257728,
      "cbid": 317, "correlation": 257728
    }
  },
  {
    "ph": "f", "id": 257728, "pid": 76337, "tid": -914061504, "ts": 1716454225424538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424538, "dur": 0,
    "args": {
      "External id": 257729,
      "cbid": 203, "correlation": 257729
    }
  },
  {
    "ph": "f", "id": 257729, "pid": 76337, "tid": -914061504, "ts": 1716454225424538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424539, "dur": 0,
    "args": {
      "External id": 257730,
      "cbid": 205, "correlation": 257730
    }
  },
  {
    "ph": "f", "id": 257730, "pid": 76337, "tid": -914061504, "ts": 1716454225424539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225555377, "dur": 40,
    "args": {
      "External id": 257734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257734, "pid": 5, "tid": 7, "ts": 1716454225555377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424553, "dur": 12,
    "args": {
      "External id": 257734,
      "cbid": 211, "correlation": 257734
    }
  },
  {
    "ph": "s", "id": 257734, "pid": 76337, "tid": -914061504, "ts": 1716454225424553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225555418, "dur": 14,
    "args": {
      "External id": 257736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257736,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257736, "pid": 5, "tid": 7, "ts": 1716454225555418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424567, "dur": 5,
    "args": {
      "External id": 257736,
      "cbid": 211, "correlation": 257736
    }
  },
  {
    "ph": "s", "id": 257736, "pid": 76337, "tid": -914061504, "ts": 1716454225424567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225555434, "dur": 3,
    "args": {
      "External id": 257738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257738, "pid": 5, "tid": 7, "ts": 1716454225555434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424576, "dur": 6,
    "args": {
      "External id": 257738,
      "cbid": 211, "correlation": 257738
    }
  },
  {
    "ph": "s", "id": 257738, "pid": 76337, "tid": -914061504, "ts": 1716454225424576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424586, "dur": 0,
    "args": {
      "External id": 257739,
      "cbid": 51, "correlation": 257739
    }
  },
  {
    "ph": "s", "id": 257739, "pid": 76337, "tid": -914061504, "ts": 1716454225424586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225555439, "dur": 708,
    "args": {
      "External id": 257740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257740,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257740, "pid": 5, "tid": 7, "ts": 1716454225555439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424587, "dur": 5,
    "args": {
      "External id": 257740,
      "cbid": 211, "correlation": 257740
    }
  },
  {
    "ph": "s", "id": 257740, "pid": 76337, "tid": -914061504, "ts": 1716454225424587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225556148, "dur": 60,
    "args": {
      "External id": 257745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257745, "pid": 5, "tid": 7, "ts": 1716454225556148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424614, "dur": 9,
    "args": {
      "External id": 257745,
      "cbid": 211, "correlation": 257745
    }
  },
  {
    "ph": "s", "id": 257745, "pid": 76337, "tid": -914061504, "ts": 1716454225424614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424671, "dur": 0,
    "args": {
      "External id": 257755,
      "cbid": 317, "correlation": 257755
    }
  },
  {
    "ph": "f", "id": 257755, "pid": 76337, "tid": -914061504, "ts": 1716454225424671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424672, "dur": 0,
    "args": {
      "External id": 257756,
      "cbid": 203, "correlation": 257756
    }
  },
  {
    "ph": "f", "id": 257756, "pid": 76337, "tid": -914061504, "ts": 1716454225424672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424673, "dur": 0,
    "args": {
      "External id": 257757,
      "cbid": 205, "correlation": 257757
    }
  },
  {
    "ph": "f", "id": 257757, "pid": 76337, "tid": -914061504, "ts": 1716454225424673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225556209, "dur": 74,
    "args": {
      "External id": 257761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257761, "pid": 5, "tid": 7, "ts": 1716454225556209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424685, "dur": 12,
    "args": {
      "External id": 257761,
      "cbid": 211, "correlation": 257761
    }
  },
  {
    "ph": "s", "id": 257761, "pid": 76337, "tid": -914061504, "ts": 1716454225424685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225556285, "dur": 211,
    "args": {
      "External id": 257763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257763,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257763, "pid": 5, "tid": 7, "ts": 1716454225556285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424703, "dur": 7,
    "args": {
      "External id": 257763,
      "cbid": 211, "correlation": 257763
    }
  },
  {
    "ph": "s", "id": 257763, "pid": 76337, "tid": -914061504, "ts": 1716454225424703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225556497, "dur": 39,
    "args": {
      "External id": 257765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257765,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257765, "pid": 5, "tid": 7, "ts": 1716454225556497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424714, "dur": 5,
    "args": {
      "External id": 257765,
      "cbid": 211, "correlation": 257765
    }
  },
  {
    "ph": "s", "id": 257765, "pid": 76337, "tid": -914061504, "ts": 1716454225424714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225556537, "dur": 60,
    "args": {
      "External id": 257771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257771, "pid": 5, "tid": 7, "ts": 1716454225556537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225424740, "dur": 532,
    "args": {
      "External id": 257771,
      "cbid": 211, "correlation": 257771
    }
  },
  {
    "ph": "s", "id": 257771, "pid": 76337, "tid": -914061504, "ts": 1716454225424740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225556599, "dur": 51,
    "args": {
      "External id": 257779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257779,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257779, "pid": 5, "tid": 7, "ts": 1716454225556599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425294, "dur": 9,
    "args": {
      "External id": 257779,
      "cbid": 211, "correlation": 257779
    }
  },
  {
    "ph": "s", "id": 257779, "pid": 76337, "tid": -914061504, "ts": 1716454225425294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225556650, "dur": 35,
    "args": {
      "External id": 257787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257787,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257787, "pid": 5, "tid": 7, "ts": 1716454225556650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425324, "dur": 8,
    "args": {
      "External id": 257787,
      "cbid": 211, "correlation": 257787
    }
  },
  {
    "ph": "s", "id": 257787, "pid": 76337, "tid": -914061504, "ts": 1716454225425324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225556686, "dur": 52,
    "args": {
      "External id": 257807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257807,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 257807, "pid": 5, "tid": 7, "ts": 1716454225556686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425407, "dur": 12,
    "args": {
      "External id": 257807,
      "cbid": 211, "correlation": 257807
    }
  },
  {
    "ph": "s", "id": 257807, "pid": 76337, "tid": -914061504, "ts": 1716454225425407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225556740, "dur": 4,
    "args": {
      "External id": 257819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257819,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 257819, "pid": 5, "tid": 7, "ts": 1716454225556740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425429, "dur": 6,
    "args": {
      "External id": 257819,
      "cbid": 211, "correlation": 257819
    }
  },
  {
    "ph": "s", "id": 257819, "pid": 76337, "tid": -914061504, "ts": 1716454225425429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225556746, "dur": 56,
    "args": {
      "External id": 257822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257822, "pid": 5, "tid": 7, "ts": 1716454225556746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425447, "dur": 8,
    "args": {
      "External id": 257822,
      "cbid": 211, "correlation": 257822
    }
  },
  {
    "ph": "s", "id": 257822, "pid": 76337, "tid": -914061504, "ts": 1716454225425447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425505, "dur": 0,
    "args": {
      "External id": 257833,
      "cbid": 317, "correlation": 257833
    }
  },
  {
    "ph": "f", "id": 257833, "pid": 76337, "tid": -914061504, "ts": 1716454225425505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425505, "dur": 0,
    "args": {
      "External id": 257834,
      "cbid": 203, "correlation": 257834
    }
  },
  {
    "ph": "f", "id": 257834, "pid": 76337, "tid": -914061504, "ts": 1716454225425505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425506, "dur": 0,
    "args": {
      "External id": 257835,
      "cbid": 205, "correlation": 257835
    }
  },
  {
    "ph": "f", "id": 257835, "pid": 76337, "tid": -914061504, "ts": 1716454225425506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425528, "dur": 1,
    "args": {
      "External id": 257839,
      "cbid": 251, "correlation": 257839
    }
  },
  {
    "ph": "f", "id": 257839, "pid": 76337, "tid": -914061504, "ts": 1716454225425528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425530, "dur": 0,
    "args": {
      "External id": 257840,
      "cbid": 251, "correlation": 257840
    }
  },
  {
    "ph": "f", "id": 257840, "pid": 76337, "tid": -914061504, "ts": 1716454225425530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425531, "dur": 0,
    "args": {
      "External id": 257841,
      "cbid": 251, "correlation": 257841
    }
  },
  {
    "ph": "f", "id": 257841, "pid": 76337, "tid": -914061504, "ts": 1716454225425531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425531, "dur": 0,
    "args": {
      "External id": 257842,
      "cbid": 251, "correlation": 257842
    }
  },
  {
    "ph": "f", "id": 257842, "pid": 76337, "tid": -914061504, "ts": 1716454225425531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425532, "dur": 0,
    "args": {
      "External id": 257843,
      "cbid": 251, "correlation": 257843
    }
  },
  {
    "ph": "f", "id": 257843, "pid": 76337, "tid": -914061504, "ts": 1716454225425532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425533, "dur": 0,
    "args": {
      "External id": 257844,
      "cbid": 251, "correlation": 257844
    }
  },
  {
    "ph": "f", "id": 257844, "pid": 76337, "tid": -914061504, "ts": 1716454225425533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425534, "dur": 0,
    "args": {
      "External id": 257845,
      "cbid": 251, "correlation": 257845
    }
  },
  {
    "ph": "f", "id": 257845, "pid": 76337, "tid": -914061504, "ts": 1716454225425534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425535, "dur": 0,
    "args": {
      "External id": 257846,
      "cbid": 251, "correlation": 257846
    }
  },
  {
    "ph": "f", "id": 257846, "pid": 76337, "tid": -914061504, "ts": 1716454225425535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425536, "dur": 0,
    "args": {
      "External id": 257847,
      "cbid": 251, "correlation": 257847
    }
  },
  {
    "ph": "f", "id": 257847, "pid": 76337, "tid": -914061504, "ts": 1716454225425536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225556802, "dur": 116,
    "args": {
      "External id": 257848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257848,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257848, "pid": 5, "tid": 7, "ts": 1716454225556802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425538, "dur": 13,
    "args": {
      "External id": 257848,
      "cbid": 211, "correlation": 257848
    }
  },
  {
    "ph": "s", "id": 257848, "pid": 76337, "tid": -914061504, "ts": 1716454225425538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225556920, "dur": 61,
    "args": {
      "External id": 257854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257854, "pid": 5, "tid": 7, "ts": 1716454225556920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425574, "dur": 9,
    "args": {
      "External id": 257854,
      "cbid": 211, "correlation": 257854
    }
  },
  {
    "ph": "s", "id": 257854, "pid": 76337, "tid": -914061504, "ts": 1716454225425574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225556982, "dur": 617,
    "args": {
      "External id": 257863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257863,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257863, "pid": 5, "tid": 7, "ts": 1716454225556982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425657, "dur": 13,
    "args": {
      "External id": 257863,
      "cbid": 211, "correlation": 257863
    }
  },
  {
    "ph": "s", "id": 257863, "pid": 76337, "tid": -914061504, "ts": 1716454225425657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225557601, "dur": 185,
    "args": {
      "External id": 257885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257885,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257885, "pid": 5, "tid": 7, "ts": 1716454225557601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425714, "dur": 10,
    "args": {
      "External id": 257885,
      "cbid": 211, "correlation": 257885
    }
  },
  {
    "ph": "s", "id": 257885, "pid": 76337, "tid": -914061504, "ts": 1716454225425714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425800, "dur": 1,
    "args": {
      "External id": 257896,
      "cbid": 251, "correlation": 257896
    }
  },
  {
    "ph": "f", "id": 257896, "pid": 76337, "tid": -914061504, "ts": 1716454225425800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225557787, "dur": 196,
    "args": {
      "External id": 257897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257897,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257897, "pid": 5, "tid": 7, "ts": 1716454225557787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425805, "dur": 12,
    "args": {
      "External id": 257897,
      "cbid": 211, "correlation": 257897
    }
  },
  {
    "ph": "s", "id": 257897, "pid": 76337, "tid": -914061504, "ts": 1716454225425805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425879, "dur": 1,
    "args": {
      "External id": 257908,
      "cbid": 251, "correlation": 257908
    }
  },
  {
    "ph": "f", "id": 257908, "pid": 76337, "tid": -914061504, "ts": 1716454225425879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225557984, "dur": 190,
    "args": {
      "External id": 257909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257909,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257909, "pid": 5, "tid": 7, "ts": 1716454225557984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425883, "dur": 11,
    "args": {
      "External id": 257909,
      "cbid": 211, "correlation": 257909
    }
  },
  {
    "ph": "s", "id": 257909, "pid": 76337, "tid": -914061504, "ts": 1716454225425883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425945, "dur": 1,
    "args": {
      "External id": 257920,
      "cbid": 251, "correlation": 257920
    }
  },
  {
    "ph": "f", "id": 257920, "pid": 76337, "tid": -914061504, "ts": 1716454225425945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225558176, "dur": 189,
    "args": {
      "External id": 257921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257921,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257921, "pid": 5, "tid": 7, "ts": 1716454225558176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225425950, "dur": 12,
    "args": {
      "External id": 257921,
      "cbid": 211, "correlation": 257921
    }
  },
  {
    "ph": "s", "id": 257921, "pid": 76337, "tid": -914061504, "ts": 1716454225425950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225558366, "dur": 18987,
    "args": {
      "External id": 257942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257942,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 257942, "pid": 5, "tid": 7, "ts": 1716454225558366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225426039, "dur": 13,
    "args": {
      "External id": 257942,
      "cbid": 211, "correlation": 257942
    }
  },
  {
    "ph": "s", "id": 257942, "pid": 76337, "tid": -914061504, "ts": 1716454225426039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225426139, "dur": 1,
    "args": {
      "External id": 257960,
      "cbid": 251, "correlation": 257960
    }
  },
  {
    "ph": "f", "id": 257960, "pid": 76337, "tid": -914061504, "ts": 1716454225426139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225577354, "dur": 206,
    "args": {
      "External id": 257962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257962,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 257962, "pid": 5, "tid": 7, "ts": 1716454225577354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225426145, "dur": 14,
    "args": {
      "External id": 257962,
      "cbid": 211, "correlation": 257962
    }
  },
  {
    "ph": "s", "id": 257962, "pid": 76337, "tid": -914061504, "ts": 1716454225426145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225577561, "dur": 66,
    "args": {
      "External id": 257970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257970,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257970, "pid": 5, "tid": 7, "ts": 1716454225577561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225426216, "dur": 12,
    "args": {
      "External id": 257970,
      "cbid": 211, "correlation": 257970
    }
  },
  {
    "ph": "s", "id": 257970, "pid": 76337, "tid": -914061504, "ts": 1716454225426216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225577629, "dur": 96,
    "args": {
      "External id": 257978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257978,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257978, "pid": 5, "tid": 7, "ts": 1716454225577629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225426255, "dur": 109,
    "args": {
      "External id": 257978,
      "cbid": 211, "correlation": 257978
    }
  },
  {
    "ph": "s", "id": 257978, "pid": 76337, "tid": -914061504, "ts": 1716454225426255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225577727, "dur": 54,
    "args": {
      "External id": 257989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 257989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 257989, "pid": 5, "tid": 7, "ts": 1716454225577727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225426428, "dur": 1928,
    "args": {
      "External id": 257989,
      "cbid": 211, "correlation": 257989
    }
  },
  {
    "ph": "s", "id": 257989, "pid": 76337, "tid": -914061504, "ts": 1716454225426428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225577782, "dur": 94,
    "args": {
      "External id": 258011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258011,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258011, "pid": 5, "tid": 7, "ts": 1716454225577782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428377, "dur": 128,
    "args": {
      "External id": 258011,
      "cbid": 211, "correlation": 258011
    }
  },
  {
    "ph": "s", "id": 258011, "pid": 76337, "tid": -914061504, "ts": 1716454225428377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428582, "dur": 1,
    "args": {
      "External id": 258022,
      "cbid": 251, "correlation": 258022
    }
  },
  {
    "ph": "f", "id": 258022, "pid": 76337, "tid": -914061504, "ts": 1716454225428582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225577877, "dur": 105,
    "args": {
      "External id": 258023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258023,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 258023, "pid": 5, "tid": 7, "ts": 1716454225577877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428588, "dur": 13,
    "args": {
      "External id": 258023,
      "cbid": 211, "correlation": 258023
    }
  },
  {
    "ph": "s", "id": 258023, "pid": 76337, "tid": -914061504, "ts": 1716454225428588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428659, "dur": 1,
    "args": {
      "External id": 258034,
      "cbid": 251, "correlation": 258034
    }
  },
  {
    "ph": "f", "id": 258034, "pid": 76337, "tid": -914061504, "ts": 1716454225428659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428663, "dur": 0,
    "args": {
      "External id": 258035,
      "cbid": 251, "correlation": 258035
    }
  },
  {
    "ph": "f", "id": 258035, "pid": 76337, "tid": -914061504, "ts": 1716454225428663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225577983, "dur": 10,
    "args": {
      "External id": 258036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258036,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258036, "pid": 5, "tid": 7, "ts": 1716454225577983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428665, "dur": 13,
    "args": {
      "External id": 258036,
      "cbid": 211, "correlation": 258036
    }
  },
  {
    "ph": "s", "id": 258036, "pid": 76337, "tid": -914061504, "ts": 1716454225428665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225577995, "dur": 5,
    "args": {
      "External id": 258038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258038,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 258038, "pid": 5, "tid": 7, "ts": 1716454225577995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428680, "dur": 6,
    "args": {
      "External id": 258038,
      "cbid": 211, "correlation": 258038
    }
  },
  {
    "ph": "s", "id": 258038, "pid": 76337, "tid": -914061504, "ts": 1716454225428680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428741, "dur": 1,
    "args": {
      "External id": 258049,
      "cbid": 251, "correlation": 258049
    }
  },
  {
    "ph": "f", "id": 258049, "pid": 76337, "tid": -914061504, "ts": 1716454225428741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428745, "dur": 0,
    "args": {
      "External id": 258050,
      "cbid": 251, "correlation": 258050
    }
  },
  {
    "ph": "f", "id": 258050, "pid": 76337, "tid": -914061504, "ts": 1716454225428745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225578001, "dur": 6,
    "args": {
      "External id": 258051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258051,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258051, "pid": 5, "tid": 7, "ts": 1716454225578001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428746, "dur": 12,
    "args": {
      "External id": 258051,
      "cbid": 211, "correlation": 258051
    }
  },
  {
    "ph": "s", "id": 258051, "pid": 76337, "tid": -914061504, "ts": 1716454225428746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225578009, "dur": 4,
    "args": {
      "External id": 258053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258053,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 258053, "pid": 5, "tid": 7, "ts": 1716454225578009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428759, "dur": 6,
    "args": {
      "External id": 258053,
      "cbid": 211, "correlation": 258053
    }
  },
  {
    "ph": "s", "id": 258053, "pid": 76337, "tid": -914061504, "ts": 1716454225428759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225578014, "dur": 158,
    "args": {
      "External id": 258074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258074,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 258074, "pid": 5, "tid": 7, "ts": 1716454225578014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428834, "dur": 12,
    "args": {
      "External id": 258074,
      "cbid": 211, "correlation": 258074
    }
  },
  {
    "ph": "s", "id": 258074, "pid": 76337, "tid": -914061504, "ts": 1716454225428834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428932, "dur": 1,
    "args": {
      "External id": 258092,
      "cbid": 251, "correlation": 258092
    }
  },
  {
    "ph": "f", "id": 258092, "pid": 76337, "tid": -914061504, "ts": 1716454225428932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225578173, "dur": 108,
    "args": {
      "External id": 258094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258094,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 258094, "pid": 5, "tid": 7, "ts": 1716454225578173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225428938, "dur": 13,
    "args": {
      "External id": 258094,
      "cbid": 211, "correlation": 258094
    }
  },
  {
    "ph": "s", "id": 258094, "pid": 76337, "tid": -914061504, "ts": 1716454225428938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225578282, "dur": 35,
    "args": {
      "External id": 258102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258102,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258102, "pid": 5, "tid": 7, "ts": 1716454225578282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429014, "dur": 13,
    "args": {
      "External id": 258102,
      "cbid": 211, "correlation": 258102
    }
  },
  {
    "ph": "s", "id": 258102, "pid": 76337, "tid": -914061504, "ts": 1716454225429014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225578319, "dur": 67,
    "args": {
      "External id": 258110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258110, "pid": 5, "tid": 7, "ts": 1716454225578319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429056, "dur": 9,
    "args": {
      "External id": 258110,
      "cbid": 211, "correlation": 258110
    }
  },
  {
    "ph": "s", "id": 258110, "pid": 76337, "tid": -914061504, "ts": 1716454225429056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225578388, "dur": 94,
    "args": {
      "External id": 258132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258132,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258132, "pid": 5, "tid": 7, "ts": 1716454225578388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429107, "dur": 10,
    "args": {
      "External id": 258132,
      "cbid": 211, "correlation": 258132
    }
  },
  {
    "ph": "s", "id": 258132, "pid": 76337, "tid": -914061504, "ts": 1716454225429107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429193, "dur": 1,
    "args": {
      "External id": 258148,
      "cbid": 251, "correlation": 258148
    }
  },
  {
    "ph": "f", "id": 258148, "pid": 76337, "tid": -914061504, "ts": 1716454225429193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225578483, "dur": 582,
    "args": {
      "External id": 258150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258150,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 258150, "pid": 5, "tid": 7, "ts": 1716454225578483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429199, "dur": 12,
    "args": {
      "External id": 258150,
      "cbid": 211, "correlation": 258150
    }
  },
  {
    "ph": "s", "id": 258150, "pid": 76337, "tid": -914061504, "ts": 1716454225429199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225579067, "dur": 247,
    "args": {
      "External id": 258158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258158, "pid": 5, "tid": 7, "ts": 1716454225579067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429263, "dur": 13,
    "args": {
      "External id": 258158,
      "cbid": 211, "correlation": 258158
    }
  },
  {
    "ph": "s", "id": 258158, "pid": 76337, "tid": -914061504, "ts": 1716454225429263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225579316, "dur": 252,
    "args": {
      "External id": 258166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258166, "pid": 5, "tid": 7, "ts": 1716454225579316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429293, "dur": 8,
    "args": {
      "External id": 258166,
      "cbid": 211, "correlation": 258166
    }
  },
  {
    "ph": "s", "id": 258166, "pid": 76337, "tid": -914061504, "ts": 1716454225429293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429373, "dur": 1,
    "args": {
      "External id": 258182,
      "cbid": 251, "correlation": 258182
    }
  },
  {
    "ph": "f", "id": 258182, "pid": 76337, "tid": -914061504, "ts": 1716454225429373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429379, "dur": 0,
    "args": {
      "External id": 258184,
      "cbid": 251, "correlation": 258184
    }
  },
  {
    "ph": "f", "id": 258184, "pid": 76337, "tid": -914061504, "ts": 1716454225429379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225579570, "dur": 361,
    "args": {
      "External id": 258185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258185,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258185, "pid": 5, "tid": 7, "ts": 1716454225579570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429382, "dur": 13,
    "args": {
      "External id": 258185,
      "cbid": 211, "correlation": 258185
    }
  },
  {
    "ph": "s", "id": 258185, "pid": 76337, "tid": -914061504, "ts": 1716454225429382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225579932, "dur": 50,
    "args": {
      "External id": 258193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258193,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258193, "pid": 5, "tid": 7, "ts": 1716454225579932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429423, "dur": 207,
    "args": {
      "External id": 258193,
      "cbid": 211, "correlation": 258193
    }
  },
  {
    "ph": "s", "id": 258193, "pid": 76337, "tid": -914061504, "ts": 1716454225429423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225579983, "dur": 159,
    "args": {
      "External id": 258204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258204, "pid": 5, "tid": 7, "ts": 1716454225579983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429688, "dur": 73,
    "args": {
      "External id": 258204,
      "cbid": 211, "correlation": 258204
    }
  },
  {
    "ph": "s", "id": 258204, "pid": 76337, "tid": -914061504, "ts": 1716454225429688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429815, "dur": 0,
    "args": {
      "External id": 258216,
      "cbid": 317, "correlation": 258216
    }
  },
  {
    "ph": "f", "id": 258216, "pid": 76337, "tid": -914061504, "ts": 1716454225429815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429816, "dur": 0,
    "args": {
      "External id": 258217,
      "cbid": 203, "correlation": 258217
    }
  },
  {
    "ph": "f", "id": 258217, "pid": 76337, "tid": -914061504, "ts": 1716454225429816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429816, "dur": 0,
    "args": {
      "External id": 258218,
      "cbid": 205, "correlation": 258218
    }
  },
  {
    "ph": "f", "id": 258218, "pid": 76337, "tid": -914061504, "ts": 1716454225429816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429839, "dur": 1,
    "args": {
      "External id": 258222,
      "cbid": 251, "correlation": 258222
    }
  },
  {
    "ph": "f", "id": 258222, "pid": 76337, "tid": -914061504, "ts": 1716454225429839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429841, "dur": 0,
    "args": {
      "External id": 258223,
      "cbid": 251, "correlation": 258223
    }
  },
  {
    "ph": "f", "id": 258223, "pid": 76337, "tid": -914061504, "ts": 1716454225429841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429841, "dur": 0,
    "args": {
      "External id": 258224,
      "cbid": 251, "correlation": 258224
    }
  },
  {
    "ph": "f", "id": 258224, "pid": 76337, "tid": -914061504, "ts": 1716454225429841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429842, "dur": 0,
    "args": {
      "External id": 258225,
      "cbid": 251, "correlation": 258225
    }
  },
  {
    "ph": "f", "id": 258225, "pid": 76337, "tid": -914061504, "ts": 1716454225429842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429843, "dur": 0,
    "args": {
      "External id": 258226,
      "cbid": 251, "correlation": 258226
    }
  },
  {
    "ph": "f", "id": 258226, "pid": 76337, "tid": -914061504, "ts": 1716454225429843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429843, "dur": 0,
    "args": {
      "External id": 258227,
      "cbid": 251, "correlation": 258227
    }
  },
  {
    "ph": "f", "id": 258227, "pid": 76337, "tid": -914061504, "ts": 1716454225429843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429844, "dur": 0,
    "args": {
      "External id": 258228,
      "cbid": 251, "correlation": 258228
    }
  },
  {
    "ph": "f", "id": 258228, "pid": 76337, "tid": -914061504, "ts": 1716454225429844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429845, "dur": 0,
    "args": {
      "External id": 258229,
      "cbid": 251, "correlation": 258229
    }
  },
  {
    "ph": "f", "id": 258229, "pid": 76337, "tid": -914061504, "ts": 1716454225429845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429846, "dur": 0,
    "args": {
      "External id": 258230,
      "cbid": 251, "correlation": 258230
    }
  },
  {
    "ph": "f", "id": 258230, "pid": 76337, "tid": -914061504, "ts": 1716454225429846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225580144, "dur": 116,
    "args": {
      "External id": 258231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258231,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 258231, "pid": 5, "tid": 7, "ts": 1716454225580144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429848, "dur": 42,
    "args": {
      "External id": 258231,
      "cbid": 211, "correlation": 258231
    }
  },
  {
    "ph": "s", "id": 258231, "pid": 76337, "tid": -914061504, "ts": 1716454225429848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225580261, "dur": 60,
    "args": {
      "External id": 258237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258237, "pid": 5, "tid": 7, "ts": 1716454225580261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225429914, "dur": 288,
    "args": {
      "External id": 258237,
      "cbid": 211, "correlation": 258237
    }
  },
  {
    "ph": "s", "id": 258237, "pid": 76337, "tid": -914061504, "ts": 1716454225429914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225580322, "dur": 49,
    "args": {
      "External id": 258245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258245,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258245, "pid": 5, "tid": 7, "ts": 1716454225580322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430226, "dur": 9,
    "args": {
      "External id": 258245,
      "cbid": 211, "correlation": 258245
    }
  },
  {
    "ph": "s", "id": 258245, "pid": 76337, "tid": -914061504, "ts": 1716454225430226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225580373, "dur": 52,
    "args": {
      "External id": 258265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258265,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 258265, "pid": 5, "tid": 7, "ts": 1716454225580373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430306, "dur": 12,
    "args": {
      "External id": 258265,
      "cbid": 211, "correlation": 258265
    }
  },
  {
    "ph": "s", "id": 258265, "pid": 76337, "tid": -914061504, "ts": 1716454225430306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225580427, "dur": 5,
    "args": {
      "External id": 258277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258277,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 258277, "pid": 5, "tid": 7, "ts": 1716454225580427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430328, "dur": 10,
    "args": {
      "External id": 258277,
      "cbid": 211, "correlation": 258277
    }
  },
  {
    "ph": "s", "id": 258277, "pid": 76337, "tid": -914061504, "ts": 1716454225430328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225580433, "dur": 55,
    "args": {
      "External id": 258280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258280, "pid": 5, "tid": 7, "ts": 1716454225580433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430351, "dur": 111,
    "args": {
      "External id": 258280,
      "cbid": 211, "correlation": 258280
    }
  },
  {
    "ph": "s", "id": 258280, "pid": 76337, "tid": -914061504, "ts": 1716454225430351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225580490, "dur": 37,
    "args": {
      "External id": 258289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258289,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258289, "pid": 5, "tid": 7, "ts": 1716454225580490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430499, "dur": 11,
    "args": {
      "External id": 258289,
      "cbid": 211, "correlation": 258289
    }
  },
  {
    "ph": "s", "id": 258289, "pid": 76337, "tid": -914061504, "ts": 1716454225430499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430556, "dur": 0,
    "args": {
      "External id": 258299,
      "cbid": 317, "correlation": 258299
    }
  },
  {
    "ph": "f", "id": 258299, "pid": 76337, "tid": -914061504, "ts": 1716454225430556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430557, "dur": 0,
    "args": {
      "External id": 258300,
      "cbid": 203, "correlation": 258300
    }
  },
  {
    "ph": "f", "id": 258300, "pid": 76337, "tid": -914061504, "ts": 1716454225430557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430558, "dur": 0,
    "args": {
      "External id": 258301,
      "cbid": 205, "correlation": 258301
    }
  },
  {
    "ph": "f", "id": 258301, "pid": 76337, "tid": -914061504, "ts": 1716454225430558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225580529, "dur": 41,
    "args": {
      "External id": 258305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258305, "pid": 5, "tid": 7, "ts": 1716454225580529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430574, "dur": 12,
    "args": {
      "External id": 258305,
      "cbid": 211, "correlation": 258305
    }
  },
  {
    "ph": "s", "id": 258305, "pid": 76337, "tid": -914061504, "ts": 1716454225430574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225580571, "dur": 3,
    "args": {
      "External id": 258307,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258307,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.003125814014066
    }
  },
  {
    "ph": "f", "id": 258307, "pid": 5, "tid": 7, "ts": 1716454225580571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430589, "dur": 18,
    "args": {
      "External id": 258307,
      "cbid": 51, "correlation": 258307
    }
  },
  {
    "ph": "s", "id": 258307, "pid": 76337, "tid": -914061504, "ts": 1716454225430589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430612, "dur": 1,
    "args": {
      "External id": 258309,
      "cbid": 200, "correlation": 258309
    }
  },
  {
    "ph": "f", "id": 258309, "pid": 76337, "tid": -914061504, "ts": 1716454225430612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430614, "dur": 0,
    "args": {
      "External id": 258310,
      "cbid": 200, "correlation": 258310
    }
  },
  {
    "ph": "f", "id": 258310, "pid": 76337, "tid": -914061504, "ts": 1716454225430614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430615, "dur": 0,
    "args": {
      "External id": 258311,
      "cbid": 200, "correlation": 258311
    }
  },
  {
    "ph": "f", "id": 258311, "pid": 76337, "tid": -914061504, "ts": 1716454225430615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430615, "dur": 0,
    "args": {
      "External id": 258312,
      "cbid": 200, "correlation": 258312
    }
  },
  {
    "ph": "f", "id": 258312, "pid": 76337, "tid": -914061504, "ts": 1716454225430615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430616, "dur": 3,
    "args": {
      "External id": 258313,
      "cbid": 15, "correlation": 258313
    }
  },
  {
    "ph": "f", "id": 258313, "pid": 76337, "tid": -914061504, "ts": 1716454225430616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430621, "dur": 1,
    "args": {
      "External id": 258314,
      "cbid": 251, "correlation": 258314
    }
  },
  {
    "ph": "f", "id": 258314, "pid": 76337, "tid": -914061504, "ts": 1716454225430621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225580576, "dur": 24,
    "args": {
      "External id": 258315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258315,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258315, "pid": 5, "tid": 7, "ts": 1716454225580576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430624, "dur": 8,
    "args": {
      "External id": 258315,
      "cbid": 211, "correlation": 258315
    }
  },
  {
    "ph": "s", "id": 258315, "pid": 76337, "tid": -914061504, "ts": 1716454225430624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225580601, "dur": 4,
    "args": {
      "External id": 258317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258317,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 258317, "pid": 5, "tid": 7, "ts": 1716454225580601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430638, "dur": 6,
    "args": {
      "External id": 258317,
      "cbid": 211, "correlation": 258317
    }
  },
  {
    "ph": "s", "id": 258317, "pid": 76337, "tid": -914061504, "ts": 1716454225430638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430648, "dur": 0,
    "args": {
      "External id": 258318,
      "cbid": 51, "correlation": 258318
    }
  },
  {
    "ph": "s", "id": 258318, "pid": 76337, "tid": -914061504, "ts": 1716454225430648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225580606, "dur": 192,
    "args": {
      "External id": 258319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258319,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258319, "pid": 5, "tid": 7, "ts": 1716454225580606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430649, "dur": 207,
    "args": {
      "External id": 258319,
      "cbid": 211, "correlation": 258319
    }
  },
  {
    "ph": "s", "id": 258319, "pid": 76337, "tid": -914061504, "ts": 1716454225430649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225580800, "dur": 6,
    "args": {
      "External id": 258320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258320,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258320, "pid": 5, "tid": 7, "ts": 1716454225580800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430859, "dur": 6,
    "args": {
      "External id": 258320,
      "cbid": 211, "correlation": 258320
    }
  },
  {
    "ph": "s", "id": 258320, "pid": 76337, "tid": -914061504, "ts": 1716454225430859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225580808, "dur": 5,
    "args": {
      "External id": 258326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 258326, "pid": 5, "tid": 7, "ts": 1716454225580808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225430889, "dur": 9,
    "args": {
      "External id": 258326,
      "cbid": 211, "correlation": 258326
    }
  },
  {
    "ph": "s", "id": 258326, "pid": 76337, "tid": -914061504, "ts": 1716454225430889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225580814, "dur": 3,
    "args": {
      "External id": 258334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258334,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258334, "pid": 5, "tid": 7, "ts": 1716454225580814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225432560, "dur": 15,
    "args": {
      "External id": 258334,
      "cbid": 211, "correlation": 258334
    }
  },
  {
    "ph": "s", "id": 258334, "pid": 76337, "tid": -914061504, "ts": 1716454225432560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225580818, "dur": 3,
    "args": {
      "External id": 258342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258342,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258342, "pid": 5, "tid": 7, "ts": 1716454225580818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225432601, "dur": 10,
    "args": {
      "External id": 258342,
      "cbid": 211, "correlation": 258342
    }
  },
  {
    "ph": "s", "id": 258342, "pid": 76337, "tid": -914061504, "ts": 1716454225432601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225580823, "dur": 3,
    "args": {
      "External id": 258350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258350,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258350, "pid": 5, "tid": 7, "ts": 1716454225580823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225432628, "dur": 9,
    "args": {
      "External id": 258350,
      "cbid": 211, "correlation": 258350
    }
  },
  {
    "ph": "s", "id": 258350, "pid": 76337, "tid": -914061504, "ts": 1716454225432628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225580827, "dur": 3,
    "args": {
      "External id": 258359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258359,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258359, "pid": 5, "tid": 7, "ts": 1716454225580827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225432810, "dur": 14,
    "args": {
      "External id": 258359,
      "cbid": 211, "correlation": 258359
    }
  },
  {
    "ph": "s", "id": 258359, "pid": 76337, "tid": -914061504, "ts": 1716454225432810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225580831, "dur": 3,
    "args": {
      "External id": 258368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258368,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258368, "pid": 5, "tid": 7, "ts": 1716454225580831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225432839, "dur": 7,
    "args": {
      "External id": 258368,
      "cbid": 211, "correlation": 258368
    }
  },
  {
    "ph": "s", "id": 258368, "pid": 76337, "tid": -914061504, "ts": 1716454225432839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225580836, "dur": 3,
    "args": {
      "External id": 258376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258376,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258376, "pid": 5, "tid": 7, "ts": 1716454225580836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225432865, "dur": 8,
    "args": {
      "External id": 258376,
      "cbid": 211, "correlation": 258376
    }
  },
  {
    "ph": "s", "id": 258376, "pid": 76337, "tid": -914061504, "ts": 1716454225432865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225580840, "dur": 3,
    "args": {
      "External id": 258384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258384,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258384, "pid": 5, "tid": 7, "ts": 1716454225580840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225433135, "dur": 15,
    "args": {
      "External id": 258384,
      "cbid": 211, "correlation": 258384
    }
  },
  {
    "ph": "s", "id": 258384, "pid": 76337, "tid": -914061504, "ts": 1716454225433135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225580844, "dur": 3,
    "args": {
      "External id": 258392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258392,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258392, "pid": 5, "tid": 7, "ts": 1716454225580844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225433165, "dur": 7,
    "args": {
      "External id": 258392,
      "cbid": 211, "correlation": 258392
    }
  },
  {
    "ph": "s", "id": 258392, "pid": 76337, "tid": -914061504, "ts": 1716454225433165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225580849, "dur": 1,
    "args": {
      "External id": 258402,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258402,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 258402, "pid": 5, "tid": 7, "ts": 1716454225580849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225433233, "dur": 42,
    "args": {
      "External id": 258402,
      "cbid": 41, "correlation": 258402
    }
  },
  {
    "ph": "s", "id": 258402, "pid": 76337, "tid": -914061504, "ts": 1716454225433233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225433276, "dur": 147589,
    "args": {
      "External id": 258403,
      "cbid": 131, "correlation": 258403
    }
  },
  {
    "ph": "f", "id": 258403, "pid": 76337, "tid": -914061504, "ts": 1716454225433276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225581026, "dur": 3,
    "args": {
      "External id": 258411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258411, "pid": 5, "tid": 7, "ts": 1716454225581026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581001, "dur": 28,
    "args": {
      "External id": 258411,
      "cbid": 211, "correlation": 258411
    }
  },
  {
    "ph": "s", "id": 258411, "pid": 76337, "tid": -914061504, "ts": 1716454225581001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581127, "dur": 3,
    "args": {
      "External id": 258420,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258420,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 258420, "pid": 5, "tid": 7, "ts": 1716454225581127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581092, "dur": 35,
    "args": {
      "External id": 258420,
      "cbid": 41, "correlation": 258420
    }
  },
  {
    "ph": "s", "id": 258420, "pid": 76337, "tid": -914061504, "ts": 1716454225581092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225581221, "dur": 4,
    "args": {
      "External id": 258430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258430,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258430, "pid": 5, "tid": 7, "ts": 1716454225581221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581205, "dur": 18,
    "args": {
      "External id": 258430,
      "cbid": 211, "correlation": 258430
    }
  },
  {
    "ph": "s", "id": 258430, "pid": 76337, "tid": -914061504, "ts": 1716454225581205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581296, "dur": 1,
    "args": {
      "External id": 258440,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258440,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 258440, "pid": 5, "tid": 7, "ts": 1716454225581296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581275, "dur": 19,
    "args": {
      "External id": 258440,
      "cbid": 41, "correlation": 258440
    }
  },
  {
    "ph": "s", "id": 258440, "pid": 76337, "tid": -914061504, "ts": 1716454225581275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581295, "dur": 8,
    "args": {
      "External id": 258441,
      "cbid": 131, "correlation": 258441
    }
  },
  {
    "ph": "f", "id": 258441, "pid": 76337, "tid": -914061504, "ts": 1716454225581295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581376, "dur": 3,
    "args": {
      "External id": 258448,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258448,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.11764705882353
    }
  },
  {
    "ph": "f", "id": 258448, "pid": 5, "tid": 7, "ts": 1716454225581376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581346, "dur": 30,
    "args": {
      "External id": 258448,
      "cbid": 41, "correlation": 258448
    }
  },
  {
    "ph": "s", "id": 258448, "pid": 76337, "tid": -914061504, "ts": 1716454225581346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581470, "dur": 3,
    "args": {
      "External id": 258467,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258467,
      "bytes": 16, "memory bandwidth (GB/s)": 0.0052648897663705166
    }
  },
  {
    "ph": "f", "id": 258467, "pid": 5, "tid": 7, "ts": 1716454225581470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581445, "dur": 24,
    "args": {
      "External id": 258467,
      "cbid": 41, "correlation": 258467
    }
  },
  {
    "ph": "s", "id": 258467, "pid": 76337, "tid": -914061504, "ts": 1716454225581445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225581508, "dur": 3,
    "args": {
      "External id": 258473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258473, "pid": 5, "tid": 7, "ts": 1716454225581508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581496, "dur": 13,
    "args": {
      "External id": 258473,
      "cbid": 211, "correlation": 258473
    }
  },
  {
    "ph": "s", "id": 258473, "pid": 76337, "tid": -914061504, "ts": 1716454225581496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225581523, "dur": 6,
    "args": {
      "External id": 258475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258475,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 258475, "pid": 5, "tid": 7, "ts": 1716454225581523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581513, "dur": 9,
    "args": {
      "External id": 258475,
      "cbid": 211, "correlation": 258475
    }
  },
  {
    "ph": "s", "id": 258475, "pid": 76337, "tid": -914061504, "ts": 1716454225581513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454225581531, "dur": 3,
    "args": {
      "External id": 258477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258477, "pid": 5, "tid": 7, "ts": 1716454225581531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581524, "dur": 6,
    "args": {
      "External id": 258477,
      "cbid": 211, "correlation": 258477
    }
  },
  {
    "ph": "s", "id": 258477, "pid": 76337, "tid": -914061504, "ts": 1716454225581524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581570, "dur": 2,
    "args": {
      "External id": 258485,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258485,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002717391304347826
    }
  },
  {
    "ph": "f", "id": 258485, "pid": 5, "tid": 7, "ts": 1716454225581570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581552, "dur": 17,
    "args": {
      "External id": 258485,
      "cbid": 41, "correlation": 258485
    }
  },
  {
    "ph": "s", "id": 258485, "pid": 76337, "tid": -914061504, "ts": 1716454225581552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225581618, "dur": 3,
    "args": {
      "External id": 258499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258499, "pid": 5, "tid": 7, "ts": 1716454225581618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581607, "dur": 12,
    "args": {
      "External id": 258499,
      "cbid": 211, "correlation": 258499
    }
  },
  {
    "ph": "s", "id": 258499, "pid": 76337, "tid": -914061504, "ts": 1716454225581607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225581638, "dur": 2,
    "args": {
      "External id": 258513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258513,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258513, "pid": 5, "tid": 7, "ts": 1716454225581638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581631, "dur": 6,
    "args": {
      "External id": 258513,
      "cbid": 211, "correlation": 258513
    }
  },
  {
    "ph": "s", "id": 258513, "pid": 76337, "tid": -914061504, "ts": 1716454225581631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225581673, "dur": 6,
    "args": {
      "External id": 258520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258520,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258520, "pid": 5, "tid": 7, "ts": 1716454225581673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581662, "dur": 11,
    "args": {
      "External id": 258520,
      "cbid": 211, "correlation": 258520
    }
  },
  {
    "ph": "s", "id": 258520, "pid": 76337, "tid": -914061504, "ts": 1716454225581662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225581684, "dur": 6,
    "args": {
      "External id": 258523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258523,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258523, "pid": 5, "tid": 7, "ts": 1716454225581684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581676, "dur": 7,
    "args": {
      "External id": 258523,
      "cbid": 211, "correlation": 258523
    }
  },
  {
    "ph": "s", "id": 258523, "pid": 76337, "tid": -914061504, "ts": 1716454225581676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225581692, "dur": 3,
    "args": {
      "External id": 258525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258525,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258525, "pid": 5, "tid": 7, "ts": 1716454225581692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581685, "dur": 7,
    "args": {
      "External id": 258525,
      "cbid": 211, "correlation": 258525
    }
  },
  {
    "ph": "s", "id": 258525, "pid": 76337, "tid": -914061504, "ts": 1716454225581685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581714, "dur": 2,
    "args": {
      "External id": 258528,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258528,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 258528, "pid": 5, "tid": 7, "ts": 1716454225581714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581701, "dur": 12,
    "args": {
      "External id": 258528,
      "cbid": 41, "correlation": 258528
    }
  },
  {
    "ph": "s", "id": 258528, "pid": 76337, "tid": -914061504, "ts": 1716454225581701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225581767, "dur": 4,
    "args": {
      "External id": 258544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258544,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258544, "pid": 5, "tid": 7, "ts": 1716454225581767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581755, "dur": 12,
    "args": {
      "External id": 258544,
      "cbid": 211, "correlation": 258544
    }
  },
  {
    "ph": "s", "id": 258544, "pid": 76337, "tid": -914061504, "ts": 1716454225581755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225581790, "dur": 3,
    "args": {
      "External id": 258549,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258549,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003094059405940594
    }
  },
  {
    "ph": "f", "id": 258549, "pid": 5, "tid": 7, "ts": 1716454225581790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581772, "dur": 17,
    "args": {
      "External id": 258549,
      "cbid": 41, "correlation": 258549
    }
  },
  {
    "ph": "s", "id": 258549, "pid": 76337, "tid": -914061504, "ts": 1716454225581772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225581819, "dur": 1,
    "args": {
      "External id": 258555,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258555,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 258555, "pid": 5, "tid": 7, "ts": 1716454225581819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581800, "dur": 28,
    "args": {
      "External id": 258555,
      "cbid": 41, "correlation": 258555
    }
  },
  {
    "ph": "s", "id": 258555, "pid": 76337, "tid": -914061504, "ts": 1716454225581800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581829, "dur": 4,
    "args": {
      "External id": 258556,
      "cbid": 131, "correlation": 258556
    }
  },
  {
    "ph": "f", "id": 258556, "pid": 76337, "tid": -914061504, "ts": 1716454225581829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225581881, "dur": 3,
    "args": {
      "External id": 258564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258564,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258564, "pid": 5, "tid": 7, "ts": 1716454225581881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581868, "dur": 13,
    "args": {
      "External id": 258564,
      "cbid": 211, "correlation": 258564
    }
  },
  {
    "ph": "s", "id": 258564, "pid": 76337, "tid": -914061504, "ts": 1716454225581868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225581912, "dur": 3,
    "args": {
      "External id": 258574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258574,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258574, "pid": 5, "tid": 7, "ts": 1716454225581912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581903, "dur": 8,
    "args": {
      "External id": 258574,
      "cbid": 211, "correlation": 258574
    }
  },
  {
    "ph": "s", "id": 258574, "pid": 76337, "tid": -914061504, "ts": 1716454225581903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225581936, "dur": 3,
    "args": {
      "External id": 258583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258583,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258583, "pid": 5, "tid": 7, "ts": 1716454225581936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225581927, "dur": 8,
    "args": {
      "External id": 258583,
      "cbid": 211, "correlation": 258583
    }
  },
  {
    "ph": "s", "id": 258583, "pid": 76337, "tid": -914061504, "ts": 1716454225581927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225582059, "dur": 12,
    "args": {
      "External id": 258593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258593,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258593, "pid": 5, "tid": 7, "ts": 1716454225582059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582045, "dur": 15,
    "args": {
      "External id": 258593,
      "cbid": 211, "correlation": 258593
    }
  },
  {
    "ph": "s", "id": 258593, "pid": 76337, "tid": -914061504, "ts": 1716454225582045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225582099, "dur": 3,
    "args": {
      "External id": 258601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258601,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258601, "pid": 5, "tid": 7, "ts": 1716454225582099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582090, "dur": 8,
    "args": {
      "External id": 258601,
      "cbid": 211, "correlation": 258601
    }
  },
  {
    "ph": "s", "id": 258601, "pid": 76337, "tid": -914061504, "ts": 1716454225582090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225582144, "dur": 12,
    "args": {
      "External id": 258611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258611,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258611, "pid": 5, "tid": 7, "ts": 1716454225582144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582132, "dur": 12,
    "args": {
      "External id": 258611,
      "cbid": 211, "correlation": 258611
    }
  },
  {
    "ph": "s", "id": 258611, "pid": 76337, "tid": -914061504, "ts": 1716454225582132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225582175, "dur": 10,
    "args": {
      "External id": 258619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258619,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258619, "pid": 5, "tid": 7, "ts": 1716454225582175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582166, "dur": 9,
    "args": {
      "External id": 258619,
      "cbid": 211, "correlation": 258619
    }
  },
  {
    "ph": "s", "id": 258619, "pid": 76337, "tid": -914061504, "ts": 1716454225582166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225582206, "dur": 3,
    "args": {
      "External id": 258628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258628, "pid": 5, "tid": 7, "ts": 1716454225582206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582194, "dur": 12,
    "args": {
      "External id": 258628,
      "cbid": 211, "correlation": 258628
    }
  },
  {
    "ph": "s", "id": 258628, "pid": 76337, "tid": -914061504, "ts": 1716454225582194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225582232, "dur": 5,
    "args": {
      "External id": 258637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258637,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258637, "pid": 5, "tid": 7, "ts": 1716454225582232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582222, "dur": 8,
    "args": {
      "External id": 258637,
      "cbid": 211, "correlation": 258637
    }
  },
  {
    "ph": "s", "id": 258637, "pid": 76337, "tid": -914061504, "ts": 1716454225582222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225582271, "dur": 8,
    "args": {
      "External id": 258647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258647,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258647, "pid": 5, "tid": 7, "ts": 1716454225582271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582260, "dur": 10,
    "args": {
      "External id": 258647,
      "cbid": 211, "correlation": 258647
    }
  },
  {
    "ph": "s", "id": 258647, "pid": 76337, "tid": -914061504, "ts": 1716454225582260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225582586, "dur": 3,
    "args": {
      "External id": 258656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258656,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258656, "pid": 5, "tid": 7, "ts": 1716454225582586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582570, "dur": 16,
    "args": {
      "External id": 258656,
      "cbid": 211, "correlation": 258656
    }
  },
  {
    "ph": "s", "id": 258656, "pid": 76337, "tid": -914061504, "ts": 1716454225582570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225582613, "dur": 3,
    "args": {
      "External id": 258664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258664,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258664, "pid": 5, "tid": 7, "ts": 1716454225582613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582604, "dur": 8,
    "args": {
      "External id": 258664,
      "cbid": 211, "correlation": 258664
    }
  },
  {
    "ph": "s", "id": 258664, "pid": 76337, "tid": -914061504, "ts": 1716454225582604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225582664, "dur": 1,
    "args": {
      "External id": 258674,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258674,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 258674, "pid": 5, "tid": 7, "ts": 1716454225582664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582649, "dur": 13,
    "args": {
      "External id": 258674,
      "cbid": 41, "correlation": 258674
    }
  },
  {
    "ph": "s", "id": 258674, "pid": 76337, "tid": -914061504, "ts": 1716454225582649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582663, "dur": 8,
    "args": {
      "External id": 258675,
      "cbid": 131, "correlation": 258675
    }
  },
  {
    "ph": "f", "id": 258675, "pid": 76337, "tid": -914061504, "ts": 1716454225582663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225582754, "dur": 2,
    "args": {
      "External id": 258683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258683, "pid": 5, "tid": 7, "ts": 1716454225582754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582740, "dur": 14,
    "args": {
      "External id": 258683,
      "cbid": 211, "correlation": 258683
    }
  },
  {
    "ph": "s", "id": 258683, "pid": 76337, "tid": -914061504, "ts": 1716454225582740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225582826, "dur": 3,
    "args": {
      "External id": 258692,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258692,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002551020408163265
    }
  },
  {
    "ph": "f", "id": 258692, "pid": 5, "tid": 7, "ts": 1716454225582826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582808, "dur": 18,
    "args": {
      "External id": 258692,
      "cbid": 41, "correlation": 258692
    }
  },
  {
    "ph": "s", "id": 258692, "pid": 76337, "tid": -914061504, "ts": 1716454225582808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225582898, "dur": 3,
    "args": {
      "External id": 258702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258702,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 258702, "pid": 5, "tid": 7, "ts": 1716454225582898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582883, "dur": 15,
    "args": {
      "External id": 258702,
      "cbid": 211, "correlation": 258702
    }
  },
  {
    "ph": "s", "id": 258702, "pid": 76337, "tid": -914061504, "ts": 1716454225582883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225582950, "dur": 1,
    "args": {
      "External id": 258712,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258712,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 258712, "pid": 5, "tid": 7, "ts": 1716454225582950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582936, "dur": 12,
    "args": {
      "External id": 258712,
      "cbid": 41, "correlation": 258712
    }
  },
  {
    "ph": "s", "id": 258712, "pid": 76337, "tid": -914061504, "ts": 1716454225582936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225582949, "dur": 8,
    "args": {
      "External id": 258713,
      "cbid": 131, "correlation": 258713
    }
  },
  {
    "ph": "f", "id": 258713, "pid": 76337, "tid": -914061504, "ts": 1716454225582949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225583025, "dur": 3,
    "args": {
      "External id": 258720,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258720,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.03030303030303
    }
  },
  {
    "ph": "f", "id": 258720, "pid": 5, "tid": 7, "ts": 1716454225583025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583001, "dur": 24,
    "args": {
      "External id": 258720,
      "cbid": 41, "correlation": 258720
    }
  },
  {
    "ph": "s", "id": 258720, "pid": 76337, "tid": -914061504, "ts": 1716454225583001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225583073, "dur": 1,
    "args": {
      "External id": 258731,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258731,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 258731, "pid": 5, "tid": 7, "ts": 1716454225583073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583062, "dur": 10,
    "args": {
      "External id": 258731,
      "cbid": 41, "correlation": 258731
    }
  },
  {
    "ph": "s", "id": 258731, "pid": 76337, "tid": -914061504, "ts": 1716454225583062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583072, "dur": 8,
    "args": {
      "External id": 258732,
      "cbid": 131, "correlation": 258732
    }
  },
  {
    "ph": "f", "id": 258732, "pid": 76337, "tid": -914061504, "ts": 1716454225583072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225583123, "dur": 3,
    "args": {
      "External id": 258740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258740,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258740, "pid": 5, "tid": 7, "ts": 1716454225583123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583110, "dur": 13,
    "args": {
      "External id": 258740,
      "cbid": 211, "correlation": 258740
    }
  },
  {
    "ph": "s", "id": 258740, "pid": 76337, "tid": -914061504, "ts": 1716454225583110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225583153, "dur": 3,
    "args": {
      "External id": 258750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258750,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258750, "pid": 5, "tid": 7, "ts": 1716454225583153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583144, "dur": 8,
    "args": {
      "External id": 258750,
      "cbid": 211, "correlation": 258750
    }
  },
  {
    "ph": "s", "id": 258750, "pid": 76337, "tid": -914061504, "ts": 1716454225583144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225583175, "dur": 3,
    "args": {
      "External id": 258759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258759,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258759, "pid": 5, "tid": 7, "ts": 1716454225583175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583166, "dur": 7,
    "args": {
      "External id": 258759,
      "cbid": 211, "correlation": 258759
    }
  },
  {
    "ph": "s", "id": 258759, "pid": 76337, "tid": -914061504, "ts": 1716454225583166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225583243, "dur": 5,
    "args": {
      "External id": 258767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258767, "pid": 5, "tid": 7, "ts": 1716454225583243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583230, "dur": 15,
    "args": {
      "External id": 258767,
      "cbid": 211, "correlation": 258767
    }
  },
  {
    "ph": "s", "id": 258767, "pid": 76337, "tid": -914061504, "ts": 1716454225583230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225583283, "dur": 3,
    "args": {
      "External id": 258776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258776,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258776, "pid": 5, "tid": 7, "ts": 1716454225583283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583274, "dur": 9,
    "args": {
      "External id": 258776,
      "cbid": 211, "correlation": 258776
    }
  },
  {
    "ph": "s", "id": 258776, "pid": 76337, "tid": -914061504, "ts": 1716454225583274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225583307, "dur": 3,
    "args": {
      "External id": 258785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258785,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258785, "pid": 5, "tid": 7, "ts": 1716454225583307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583299, "dur": 7,
    "args": {
      "External id": 258785,
      "cbid": 211, "correlation": 258785
    }
  },
  {
    "ph": "s", "id": 258785, "pid": 76337, "tid": -914061504, "ts": 1716454225583299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225583368, "dur": 3,
    "args": {
      "External id": 258793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258793,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258793, "pid": 5, "tid": 7, "ts": 1716454225583368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583357, "dur": 10,
    "args": {
      "External id": 258793,
      "cbid": 211, "correlation": 258793
    }
  },
  {
    "ph": "s", "id": 258793, "pid": 76337, "tid": -914061504, "ts": 1716454225583357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225583425, "dur": 1,
    "args": {
      "External id": 258801,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258801,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0043859649122807015
    }
  },
  {
    "ph": "f", "id": 258801, "pid": 5, "tid": 7, "ts": 1716454225583425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583410, "dur": 25,
    "args": {
      "External id": 258801,
      "cbid": 41, "correlation": 258801
    }
  },
  {
    "ph": "s", "id": 258801, "pid": 76337, "tid": -914061504, "ts": 1716454225583410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583436, "dur": 4,
    "args": {
      "External id": 258802,
      "cbid": 131, "correlation": 258802
    }
  },
  {
    "ph": "f", "id": 258802, "pid": 76337, "tid": -914061504, "ts": 1716454225583436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225583497, "dur": 1,
    "args": {
      "External id": 258812,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258812,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 258812, "pid": 5, "tid": 7, "ts": 1716454225583497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583484, "dur": 11,
    "args": {
      "External id": 258812,
      "cbid": 41, "correlation": 258812
    }
  },
  {
    "ph": "s", "id": 258812, "pid": 76337, "tid": -914061504, "ts": 1716454225583484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583496, "dur": 8,
    "args": {
      "External id": 258813,
      "cbid": 131, "correlation": 258813
    }
  },
  {
    "ph": "f", "id": 258813, "pid": 76337, "tid": -914061504, "ts": 1716454225583496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225583552, "dur": 1,
    "args": {
      "External id": 258822,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258822,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 258822, "pid": 5, "tid": 7, "ts": 1716454225583552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583542, "dur": 8,
    "args": {
      "External id": 258822,
      "cbid": 41, "correlation": 258822
    }
  },
  {
    "ph": "s", "id": 258822, "pid": 76337, "tid": -914061504, "ts": 1716454225583542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583550, "dur": 8,
    "args": {
      "External id": 258823,
      "cbid": 131, "correlation": 258823
    }
  },
  {
    "ph": "f", "id": 258823, "pid": 76337, "tid": -914061504, "ts": 1716454225583550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225583625, "dur": 4,
    "args": {
      "External id": 258830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258830, "pid": 5, "tid": 7, "ts": 1716454225583625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583609, "dur": 17,
    "args": {
      "External id": 258830,
      "cbid": 211, "correlation": 258830
    }
  },
  {
    "ph": "s", "id": 258830, "pid": 76337, "tid": -914061504, "ts": 1716454225583609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454225583662, "dur": 4,
    "args": {
      "External id": 258850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258850,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258850, "pid": 5, "tid": 7, "ts": 1716454225583662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583652, "dur": 11,
    "args": {
      "External id": 258850,
      "cbid": 211, "correlation": 258850
    }
  },
  {
    "ph": "s", "id": 258850, "pid": 76337, "tid": -914061504, "ts": 1716454225583652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583663, "dur": 0,
    "args": {
      "External id": 258851,
      "cbid": 11, "correlation": 258851
    }
  },
  {
    "ph": "f", "id": 258851, "pid": 76337, "tid": -914061504, "ts": 1716454225583663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583664, "dur": 0,
    "args": {
      "External id": 258852,
      "cbid": 11, "correlation": 258852
    }
  },
  {
    "ph": "f", "id": 258852, "pid": 76337, "tid": -914061504, "ts": 1716454225583664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225583677, "dur": 1,
    "args": {
      "External id": 258855,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258855,
      "bytes": 4, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 258855, "pid": 5, "tid": 7, "ts": 1716454225583677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583665, "dur": 21,
    "args": {
      "External id": 258855,
      "cbid": 41, "correlation": 258855
    }
  },
  {
    "ph": "s", "id": 258855, "pid": 76337, "tid": -914061504, "ts": 1716454225583665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583687, "dur": 3,
    "args": {
      "External id": 258856,
      "cbid": 131, "correlation": 258856
    }
  },
  {
    "ph": "f", "id": 258856, "pid": 76337, "tid": -914061504, "ts": 1716454225583687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225583714, "dur": 3,
    "args": {
      "External id": 258880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258880, "pid": 5, "tid": 7, "ts": 1716454225583714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583705, "dur": 9,
    "args": {
      "External id": 258880,
      "cbid": 211, "correlation": 258880
    }
  },
  {
    "ph": "s", "id": 258880, "pid": 76337, "tid": -914061504, "ts": 1716454225583705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583715, "dur": 0,
    "args": {
      "External id": 258881,
      "cbid": 11, "correlation": 258881
    }
  },
  {
    "ph": "f", "id": 258881, "pid": 76337, "tid": -914061504, "ts": 1716454225583715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583715, "dur": 0,
    "args": {
      "External id": 258882,
      "cbid": 11, "correlation": 258882
    }
  },
  {
    "ph": "f", "id": 258882, "pid": 76337, "tid": -914061504, "ts": 1716454225583715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583717, "dur": 1,
    "args": {
      "External id": 258884,
      "cbid": 200, "correlation": 258884
    }
  },
  {
    "ph": "f", "id": 258884, "pid": 76337, "tid": -914061504, "ts": 1716454225583717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225583728, "dur": 4,
    "args": {
      "External id": 258886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258886,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258886, "pid": 5, "tid": 7, "ts": 1716454225583728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583720, "dur": 9,
    "args": {
      "External id": 258886,
      "cbid": 211, "correlation": 258886
    }
  },
  {
    "ph": "s", "id": 258886, "pid": 76337, "tid": -914061504, "ts": 1716454225583720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583730, "dur": 0,
    "args": {
      "External id": 258887,
      "cbid": 11, "correlation": 258887
    }
  },
  {
    "ph": "f", "id": 258887, "pid": 76337, "tid": -914061504, "ts": 1716454225583730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583730, "dur": 0,
    "args": {
      "External id": 258888,
      "cbid": 11, "correlation": 258888
    }
  },
  {
    "ph": "f", "id": 258888, "pid": 76337, "tid": -914061504, "ts": 1716454225583730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225583768, "dur": 1,
    "args": {
      "External id": 258895,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258895,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 258895, "pid": 5, "tid": 7, "ts": 1716454225583768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583756, "dur": 20,
    "args": {
      "External id": 258895,
      "cbid": 41, "correlation": 258895
    }
  },
  {
    "ph": "s", "id": 258895, "pid": 76337, "tid": -914061504, "ts": 1716454225583756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583777, "dur": 3,
    "args": {
      "External id": 258896,
      "cbid": 131, "correlation": 258896
    }
  },
  {
    "ph": "f", "id": 258896, "pid": 76337, "tid": -914061504, "ts": 1716454225583777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225583826, "dur": 1,
    "args": {
      "External id": 258906,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 258906,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 258906, "pid": 5, "tid": 7, "ts": 1716454225583826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583815, "dur": 9,
    "args": {
      "External id": 258906,
      "cbid": 41, "correlation": 258906
    }
  },
  {
    "ph": "s", "id": 258906, "pid": 76337, "tid": -914061504, "ts": 1716454225583815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583825, "dur": 8,
    "args": {
      "External id": 258907,
      "cbid": 131, "correlation": 258907
    }
  },
  {
    "ph": "f", "id": 258907, "pid": 76337, "tid": -914061504, "ts": 1716454225583825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225583895, "dur": 5,
    "args": {
      "External id": 258914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258914,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258914, "pid": 5, "tid": 7, "ts": 1716454225583895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583880, "dur": 16,
    "args": {
      "External id": 258914,
      "cbid": 211, "correlation": 258914
    }
  },
  {
    "ph": "s", "id": 258914, "pid": 76337, "tid": -914061504, "ts": 1716454225583880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225583966, "dur": 3,
    "args": {
      "External id": 258923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258923,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258923, "pid": 5, "tid": 7, "ts": 1716454225583966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583953, "dur": 13,
    "args": {
      "External id": 258923,
      "cbid": 211, "correlation": 258923
    }
  },
  {
    "ph": "s", "id": 258923, "pid": 76337, "tid": -914061504, "ts": 1716454225583953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584010, "dur": 3,
    "args": {
      "External id": 258931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258931,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258931, "pid": 5, "tid": 7, "ts": 1716454225584010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225583999, "dur": 11,
    "args": {
      "External id": 258931,
      "cbid": 211, "correlation": 258931
    }
  },
  {
    "ph": "s", "id": 258931, "pid": 76337, "tid": -914061504, "ts": 1716454225583999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584045, "dur": 4,
    "args": {
      "External id": 258939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258939,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258939, "pid": 5, "tid": 7, "ts": 1716454225584045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584034, "dur": 11,
    "args": {
      "External id": 258939,
      "cbid": 211, "correlation": 258939
    }
  },
  {
    "ph": "s", "id": 258939, "pid": 76337, "tid": -914061504, "ts": 1716454225584034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584073, "dur": 4,
    "args": {
      "External id": 258947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258947,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258947, "pid": 5, "tid": 7, "ts": 1716454225584073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584063, "dur": 10,
    "args": {
      "External id": 258947,
      "cbid": 211, "correlation": 258947
    }
  },
  {
    "ph": "s", "id": 258947, "pid": 76337, "tid": -914061504, "ts": 1716454225584063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584100, "dur": 3,
    "args": {
      "External id": 258955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258955,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258955, "pid": 5, "tid": 7, "ts": 1716454225584100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584091, "dur": 8,
    "args": {
      "External id": 258955,
      "cbid": 211, "correlation": 258955
    }
  },
  {
    "ph": "s", "id": 258955, "pid": 76337, "tid": -914061504, "ts": 1716454225584091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225584124, "dur": 3,
    "args": {
      "External id": 258963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258963,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 258963, "pid": 5, "tid": 7, "ts": 1716454225584124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584115, "dur": 9,
    "args": {
      "External id": 258963,
      "cbid": 211, "correlation": 258963
    }
  },
  {
    "ph": "s", "id": 258963, "pid": 76337, "tid": -914061504, "ts": 1716454225584115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225584147, "dur": 4,
    "args": {
      "External id": 258971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258971, "pid": 5, "tid": 7, "ts": 1716454225584147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584139, "dur": 7,
    "args": {
      "External id": 258971,
      "cbid": 211, "correlation": 258971
    }
  },
  {
    "ph": "s", "id": 258971, "pid": 76337, "tid": -914061504, "ts": 1716454225584139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225584165, "dur": 4,
    "args": {
      "External id": 258979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258979, "pid": 5, "tid": 7, "ts": 1716454225584165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584157, "dur": 7,
    "args": {
      "External id": 258979,
      "cbid": 211, "correlation": 258979
    }
  },
  {
    "ph": "s", "id": 258979, "pid": 76337, "tid": -914061504, "ts": 1716454225584157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225584185, "dur": 3,
    "args": {
      "External id": 258987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258987,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 258987, "pid": 5, "tid": 7, "ts": 1716454225584185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584177, "dur": 7,
    "args": {
      "External id": 258987,
      "cbid": 211, "correlation": 258987
    }
  },
  {
    "ph": "s", "id": 258987, "pid": 76337, "tid": -914061504, "ts": 1716454225584177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584243, "dur": 3,
    "args": {
      "External id": 258995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 258995,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 258995, "pid": 5, "tid": 7, "ts": 1716454225584243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584233, "dur": 10,
    "args": {
      "External id": 258995,
      "cbid": 211, "correlation": 258995
    }
  },
  {
    "ph": "s", "id": 258995, "pid": 76337, "tid": -914061504, "ts": 1716454225584233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225584269, "dur": 4,
    "args": {
      "External id": 259003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 259003, "pid": 5, "tid": 7, "ts": 1716454225584269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584260, "dur": 8,
    "args": {
      "External id": 259003,
      "cbid": 211, "correlation": 259003
    }
  },
  {
    "ph": "s", "id": 259003, "pid": 76337, "tid": -914061504, "ts": 1716454225584260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225584293, "dur": 4,
    "args": {
      "External id": 259011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 259011, "pid": 5, "tid": 7, "ts": 1716454225584293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584284, "dur": 8,
    "args": {
      "External id": 259011,
      "cbid": 211, "correlation": 259011
    }
  },
  {
    "ph": "s", "id": 259011, "pid": 76337, "tid": -914061504, "ts": 1716454225584284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225584312, "dur": 3,
    "args": {
      "External id": 259019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259019,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 259019, "pid": 5, "tid": 7, "ts": 1716454225584312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584304, "dur": 7,
    "args": {
      "External id": 259019,
      "cbid": 211, "correlation": 259019
    }
  },
  {
    "ph": "s", "id": 259019, "pid": 76337, "tid": -914061504, "ts": 1716454225584304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225584721, "dur": 5,
    "args": {
      "External id": 259028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259028,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259028, "pid": 5, "tid": 7, "ts": 1716454225584721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584704, "dur": 18,
    "args": {
      "External id": 259028,
      "cbid": 211, "correlation": 259028
    }
  },
  {
    "ph": "s", "id": 259028, "pid": 76337, "tid": -914061504, "ts": 1716454225584704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225584758, "dur": 5,
    "args": {
      "External id": 259037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259037,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259037, "pid": 5, "tid": 7, "ts": 1716454225584758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584748, "dur": 9,
    "args": {
      "External id": 259037,
      "cbid": 211, "correlation": 259037
    }
  },
  {
    "ph": "s", "id": 259037, "pid": 76337, "tid": -914061504, "ts": 1716454225584748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225584887, "dur": 3,
    "args": {
      "External id": 259053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259053,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259053, "pid": 5, "tid": 7, "ts": 1716454225584887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584874, "dur": 14,
    "args": {
      "External id": 259053,
      "cbid": 211, "correlation": 259053
    }
  },
  {
    "ph": "s", "id": 259053, "pid": 76337, "tid": -914061504, "ts": 1716454225584874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584922, "dur": 3,
    "args": {
      "External id": 259061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259061, "pid": 5, "tid": 7, "ts": 1716454225584922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584911, "dur": 10,
    "args": {
      "External id": 259061,
      "cbid": 211, "correlation": 259061
    }
  },
  {
    "ph": "s", "id": 259061, "pid": 76337, "tid": -914061504, "ts": 1716454225584911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584953, "dur": 3,
    "args": {
      "External id": 259069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259069, "pid": 5, "tid": 7, "ts": 1716454225584953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584943, "dur": 8,
    "args": {
      "External id": 259069,
      "cbid": 211, "correlation": 259069
    }
  },
  {
    "ph": "s", "id": 259069, "pid": 76337, "tid": -914061504, "ts": 1716454225584943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225584993, "dur": 4,
    "args": {
      "External id": 259077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259077,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259077, "pid": 5, "tid": 7, "ts": 1716454225584993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225584983, "dur": 10,
    "args": {
      "External id": 259077,
      "cbid": 211, "correlation": 259077
    }
  },
  {
    "ph": "s", "id": 259077, "pid": 76337, "tid": -914061504, "ts": 1716454225584983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225585050, "dur": 4,
    "args": {
      "External id": 259089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259089,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259089, "pid": 5, "tid": 7, "ts": 1716454225585050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585037, "dur": 13,
    "args": {
      "External id": 259089,
      "cbid": 211, "correlation": 259089
    }
  },
  {
    "ph": "s", "id": 259089, "pid": 76337, "tid": -914061504, "ts": 1716454225585037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225585096, "dur": 4,
    "args": {
      "External id": 259100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259100, "pid": 5, "tid": 7, "ts": 1716454225585096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585084, "dur": 12,
    "args": {
      "External id": 259100,
      "cbid": 211, "correlation": 259100
    }
  },
  {
    "ph": "s", "id": 259100, "pid": 76337, "tid": -914061504, "ts": 1716454225585084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225585128, "dur": 3,
    "args": {
      "External id": 259108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259108, "pid": 5, "tid": 7, "ts": 1716454225585128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585118, "dur": 8,
    "args": {
      "External id": 259108,
      "cbid": 211, "correlation": 259108
    }
  },
  {
    "ph": "s", "id": 259108, "pid": 76337, "tid": -914061504, "ts": 1716454225585118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225585161, "dur": 5,
    "args": {
      "External id": 259116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259116,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259116, "pid": 5, "tid": 7, "ts": 1716454225585161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585150, "dur": 11,
    "args": {
      "External id": 259116,
      "cbid": 211, "correlation": 259116
    }
  },
  {
    "ph": "s", "id": 259116, "pid": 76337, "tid": -914061504, "ts": 1716454225585150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225585190, "dur": 5,
    "args": {
      "External id": 259124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259124,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259124, "pid": 5, "tid": 7, "ts": 1716454225585190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585181, "dur": 9,
    "args": {
      "External id": 259124,
      "cbid": 211, "correlation": 259124
    }
  },
  {
    "ph": "s", "id": 259124, "pid": 76337, "tid": -914061504, "ts": 1716454225585181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225585221, "dur": 4,
    "args": {
      "External id": 259133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259133,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259133, "pid": 5, "tid": 7, "ts": 1716454225585221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585211, "dur": 9,
    "args": {
      "External id": 259133,
      "cbid": 211, "correlation": 259133
    }
  },
  {
    "ph": "s", "id": 259133, "pid": 76337, "tid": -914061504, "ts": 1716454225585211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225585281, "dur": 4,
    "args": {
      "External id": 259146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259146,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259146, "pid": 5, "tid": 7, "ts": 1716454225585281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585269, "dur": 13,
    "args": {
      "External id": 259146,
      "cbid": 211, "correlation": 259146
    }
  },
  {
    "ph": "s", "id": 259146, "pid": 76337, "tid": -914061504, "ts": 1716454225585269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225585322, "dur": 5,
    "args": {
      "External id": 259156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259156,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259156, "pid": 5, "tid": 7, "ts": 1716454225585322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585311, "dur": 11,
    "args": {
      "External id": 259156,
      "cbid": 211, "correlation": 259156
    }
  },
  {
    "ph": "s", "id": 259156, "pid": 76337, "tid": -914061504, "ts": 1716454225585311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585449, "dur": 4,
    "args": {
      "External id": 259173,
      "cbid": 251, "correlation": 259173
    }
  },
  {
    "ph": "f", "id": 259173, "pid": 76337, "tid": -914061504, "ts": 1716454225585449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454225585478, "dur": 11,
    "args": {
      "External id": 259175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259175,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259175, "pid": 5, "tid": 7, "ts": 1716454225585478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585463, "dur": 16,
    "args": {
      "External id": 259175,
      "cbid": 211, "correlation": 259175
    }
  },
  {
    "ph": "s", "id": 259175, "pid": 76337, "tid": -914061504, "ts": 1716454225585463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225585535, "dur": 4,
    "args": {
      "External id": 259183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259183,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259183, "pid": 5, "tid": 7, "ts": 1716454225585535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585524, "dur": 11,
    "args": {
      "External id": 259183,
      "cbid": 211, "correlation": 259183
    }
  },
  {
    "ph": "s", "id": 259183, "pid": 76337, "tid": -914061504, "ts": 1716454225585524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585597, "dur": 2,
    "args": {
      "External id": 259199,
      "cbid": 251, "correlation": 259199
    }
  },
  {
    "ph": "f", "id": 259199, "pid": 76337, "tid": -914061504, "ts": 1716454225585597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585603, "dur": 0,
    "args": {
      "External id": 259201,
      "cbid": 251, "correlation": 259201
    }
  },
  {
    "ph": "f", "id": 259201, "pid": 76337, "tid": -914061504, "ts": 1716454225585603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225585619, "dur": 13,
    "args": {
      "External id": 259202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259202,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259202, "pid": 5, "tid": 7, "ts": 1716454225585619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585605, "dur": 14,
    "args": {
      "External id": 259202,
      "cbid": 211, "correlation": 259202
    }
  },
  {
    "ph": "s", "id": 259202, "pid": 76337, "tid": -914061504, "ts": 1716454225585605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225585634, "dur": 6,
    "args": {
      "External id": 259204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259204,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259204, "pid": 5, "tid": 7, "ts": 1716454225585634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585623, "dur": 9,
    "args": {
      "External id": 259204,
      "cbid": 211, "correlation": 259204
    }
  },
  {
    "ph": "s", "id": 259204, "pid": 76337, "tid": -914061504, "ts": 1716454225585623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585733, "dur": 1,
    "args": {
      "External id": 259214,
      "cbid": 317, "correlation": 259214
    }
  },
  {
    "ph": "f", "id": 259214, "pid": 76337, "tid": -914061504, "ts": 1716454225585733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585735, "dur": 1,
    "args": {
      "External id": 259215,
      "cbid": 203, "correlation": 259215
    }
  },
  {
    "ph": "f", "id": 259215, "pid": 76337, "tid": -914061504, "ts": 1716454225585735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585737, "dur": 1,
    "args": {
      "External id": 259216,
      "cbid": 205, "correlation": 259216
    }
  },
  {
    "ph": "f", "id": 259216, "pid": 76337, "tid": -914061504, "ts": 1716454225585737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225585792, "dur": 7,
    "args": {
      "External id": 259220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259220, "pid": 5, "tid": 7, "ts": 1716454225585792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585777, "dur": 14,
    "args": {
      "External id": 259220,
      "cbid": 211, "correlation": 259220
    }
  },
  {
    "ph": "s", "id": 259220, "pid": 76337, "tid": -914061504, "ts": 1716454225585777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225585802, "dur": 4,
    "args": {
      "External id": 259222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259222,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 259222, "pid": 5, "tid": 7, "ts": 1716454225585802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585795, "dur": 6,
    "args": {
      "External id": 259222,
      "cbid": 211, "correlation": 259222
    }
  },
  {
    "ph": "s", "id": 259222, "pid": 76337, "tid": -914061504, "ts": 1716454225585795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225585823, "dur": 3,
    "args": {
      "External id": 259224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259224,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259224, "pid": 5, "tid": 7, "ts": 1716454225585823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585813, "dur": 8,
    "args": {
      "External id": 259224,
      "cbid": 211, "correlation": 259224
    }
  },
  {
    "ph": "s", "id": 259224, "pid": 76337, "tid": -914061504, "ts": 1716454225585813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585828, "dur": 0,
    "args": {
      "External id": 259225,
      "cbid": 51, "correlation": 259225
    }
  },
  {
    "ph": "s", "id": 259225, "pid": 76337, "tid": -914061504, "ts": 1716454225585828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225585838, "dur": 86,
    "args": {
      "External id": 259226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259226,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259226, "pid": 5, "tid": 7, "ts": 1716454225585838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585829, "dur": 7,
    "args": {
      "External id": 259226,
      "cbid": 211, "correlation": 259226
    }
  },
  {
    "ph": "s", "id": 259226, "pid": 76337, "tid": -914061504, "ts": 1716454225585829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225585925, "dur": 61,
    "args": {
      "External id": 259231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259231, "pid": 5, "tid": 7, "ts": 1716454225585925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225585865, "dur": 11,
    "args": {
      "External id": 259231,
      "cbid": 211, "correlation": 259231
    }
  },
  {
    "ph": "s", "id": 259231, "pid": 76337, "tid": -914061504, "ts": 1716454225585865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225587673, "dur": 52,
    "args": {
      "External id": 259251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259251,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 259251, "pid": 5, "tid": 7, "ts": 1716454225587673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587657, "dur": 17,
    "args": {
      "External id": 259251,
      "cbid": 211, "correlation": 259251
    }
  },
  {
    "ph": "s", "id": 259251, "pid": 76337, "tid": -914061504, "ts": 1716454225587657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225587727, "dur": 5,
    "args": {
      "External id": 259263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259263,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259263, "pid": 5, "tid": 7, "ts": 1716454225587727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587686, "dur": 9,
    "args": {
      "External id": 259263,
      "cbid": 211, "correlation": 259263
    }
  },
  {
    "ph": "s", "id": 259263, "pid": 76337, "tid": -914061504, "ts": 1716454225587686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225587733, "dur": 58,
    "args": {
      "External id": 259266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259266,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259266, "pid": 5, "tid": 7, "ts": 1716454225587733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587710, "dur": 7,
    "args": {
      "External id": 259266,
      "cbid": 211, "correlation": 259266
    }
  },
  {
    "ph": "s", "id": 259266, "pid": 76337, "tid": -914061504, "ts": 1716454225587710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225587792, "dur": 37,
    "args": {
      "External id": 259275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259275,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259275, "pid": 5, "tid": 7, "ts": 1716454225587792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587755, "dur": 10,
    "args": {
      "External id": 259275,
      "cbid": 211, "correlation": 259275
    }
  },
  {
    "ph": "s", "id": 259275, "pid": 76337, "tid": -914061504, "ts": 1716454225587755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587813, "dur": 0,
    "args": {
      "External id": 259285,
      "cbid": 317, "correlation": 259285
    }
  },
  {
    "ph": "f", "id": 259285, "pid": 76337, "tid": -914061504, "ts": 1716454225587813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587814, "dur": 0,
    "args": {
      "External id": 259286,
      "cbid": 203, "correlation": 259286
    }
  },
  {
    "ph": "f", "id": 259286, "pid": 76337, "tid": -914061504, "ts": 1716454225587814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587815, "dur": 0,
    "args": {
      "External id": 259287,
      "cbid": 205, "correlation": 259287
    }
  },
  {
    "ph": "f", "id": 259287, "pid": 76337, "tid": -914061504, "ts": 1716454225587815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225587846, "dur": 41,
    "args": {
      "External id": 259291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259291,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259291, "pid": 5, "tid": 7, "ts": 1716454225587846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587834, "dur": 12,
    "args": {
      "External id": 259291,
      "cbid": 211, "correlation": 259291
    }
  },
  {
    "ph": "s", "id": 259291, "pid": 76337, "tid": -914061504, "ts": 1716454225587834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225587888, "dur": 15,
    "args": {
      "External id": 259293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259293,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259293, "pid": 5, "tid": 7, "ts": 1716454225587888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587848, "dur": 6,
    "args": {
      "External id": 259293,
      "cbid": 211, "correlation": 259293
    }
  },
  {
    "ph": "s", "id": 259293, "pid": 76337, "tid": -914061504, "ts": 1716454225587848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225587904, "dur": 3,
    "args": {
      "External id": 259295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259295,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259295, "pid": 5, "tid": 7, "ts": 1716454225587904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587860, "dur": 6,
    "args": {
      "External id": 259295,
      "cbid": 211, "correlation": 259295
    }
  },
  {
    "ph": "s", "id": 259295, "pid": 76337, "tid": -914061504, "ts": 1716454225587860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587871, "dur": 0,
    "args": {
      "External id": 259296,
      "cbid": 51, "correlation": 259296
    }
  },
  {
    "ph": "s", "id": 259296, "pid": 76337, "tid": -914061504, "ts": 1716454225587871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225587908, "dur": 712,
    "args": {
      "External id": 259297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259297,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259297, "pid": 5, "tid": 7, "ts": 1716454225587908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587872, "dur": 7,
    "args": {
      "External id": 259297,
      "cbid": 211, "correlation": 259297
    }
  },
  {
    "ph": "s", "id": 259297, "pid": 76337, "tid": -914061504, "ts": 1716454225587872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225588622, "dur": 60,
    "args": {
      "External id": 259302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259302,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259302, "pid": 5, "tid": 7, "ts": 1716454225588622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587902, "dur": 9,
    "args": {
      "External id": 259302,
      "cbid": 211, "correlation": 259302
    }
  },
  {
    "ph": "s", "id": 259302, "pid": 76337, "tid": -914061504, "ts": 1716454225587902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225588683, "dur": 3,
    "args": {
      "External id": 259310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259310,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259310, "pid": 5, "tid": 7, "ts": 1716454225588683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225587946, "dur": 9,
    "args": {
      "External id": 259310,
      "cbid": 211, "correlation": 259310
    }
  },
  {
    "ph": "s", "id": 259310, "pid": 76337, "tid": -914061504, "ts": 1716454225587946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588020, "dur": 2,
    "args": {
      "External id": 259326,
      "cbid": 251, "correlation": 259326
    }
  },
  {
    "ph": "f", "id": 259326, "pid": 76337, "tid": -914061504, "ts": 1716454225588020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588026, "dur": 0,
    "args": {
      "External id": 259328,
      "cbid": 251, "correlation": 259328
    }
  },
  {
    "ph": "f", "id": 259328, "pid": 76337, "tid": -914061504, "ts": 1716454225588026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225588688, "dur": 9,
    "args": {
      "External id": 259329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259329,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 259329, "pid": 5, "tid": 7, "ts": 1716454225588688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588028, "dur": 12,
    "args": {
      "External id": 259329,
      "cbid": 211, "correlation": 259329
    }
  },
  {
    "ph": "s", "id": 259329, "pid": 76337, "tid": -914061504, "ts": 1716454225588028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225588698, "dur": 4,
    "args": {
      "External id": 259331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259331,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 259331, "pid": 5, "tid": 7, "ts": 1716454225588698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588042, "dur": 6,
    "args": {
      "External id": 259331,
      "cbid": 211, "correlation": 259331
    }
  },
  {
    "ph": "s", "id": 259331, "pid": 76337, "tid": -914061504, "ts": 1716454225588042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225588704, "dur": 54,
    "args": {
      "External id": 259341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259341,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259341, "pid": 5, "tid": 7, "ts": 1716454225588704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588102, "dur": 12,
    "args": {
      "External id": 259341,
      "cbid": 211, "correlation": 259341
    }
  },
  {
    "ph": "s", "id": 259341, "pid": 76337, "tid": -914061504, "ts": 1716454225588102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225588759, "dur": 54,
    "args": {
      "External id": 259361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259361,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 259361, "pid": 5, "tid": 7, "ts": 1716454225588759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588169, "dur": 11,
    "args": {
      "External id": 259361,
      "cbid": 211, "correlation": 259361
    }
  },
  {
    "ph": "s", "id": 259361, "pid": 76337, "tid": -914061504, "ts": 1716454225588169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225588815, "dur": 4,
    "args": {
      "External id": 259373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259373,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259373, "pid": 5, "tid": 7, "ts": 1716454225588815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588190, "dur": 6,
    "args": {
      "External id": 259373,
      "cbid": 211, "correlation": 259373
    }
  },
  {
    "ph": "s", "id": 259373, "pid": 76337, "tid": -914061504, "ts": 1716454225588190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225588820, "dur": 57,
    "args": {
      "External id": 259376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259376,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259376, "pid": 5, "tid": 7, "ts": 1716454225588820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588209, "dur": 6,
    "args": {
      "External id": 259376,
      "cbid": 211, "correlation": 259376
    }
  },
  {
    "ph": "s", "id": 259376, "pid": 76337, "tid": -914061504, "ts": 1716454225588209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225588878, "dur": 36,
    "args": {
      "External id": 259385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259385,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259385, "pid": 5, "tid": 7, "ts": 1716454225588878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588250, "dur": 9,
    "args": {
      "External id": 259385,
      "cbid": 211, "correlation": 259385
    }
  },
  {
    "ph": "s", "id": 259385, "pid": 76337, "tid": -914061504, "ts": 1716454225588250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588319, "dur": 0,
    "args": {
      "External id": 259395,
      "cbid": 317, "correlation": 259395
    }
  },
  {
    "ph": "f", "id": 259395, "pid": 76337, "tid": -914061504, "ts": 1716454225588319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588319, "dur": 0,
    "args": {
      "External id": 259396,
      "cbid": 203, "correlation": 259396
    }
  },
  {
    "ph": "f", "id": 259396, "pid": 76337, "tid": -914061504, "ts": 1716454225588319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588320, "dur": 0,
    "args": {
      "External id": 259397,
      "cbid": 205, "correlation": 259397
    }
  },
  {
    "ph": "f", "id": 259397, "pid": 76337, "tid": -914061504, "ts": 1716454225588320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225588916, "dur": 40,
    "args": {
      "External id": 259401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259401,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259401, "pid": 5, "tid": 7, "ts": 1716454225588916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588335, "dur": 12,
    "args": {
      "External id": 259401,
      "cbid": 211, "correlation": 259401
    }
  },
  {
    "ph": "s", "id": 259401, "pid": 76337, "tid": -914061504, "ts": 1716454225588335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225588957, "dur": 15,
    "args": {
      "External id": 259403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259403,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259403, "pid": 5, "tid": 7, "ts": 1716454225588957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588350, "dur": 5,
    "args": {
      "External id": 259403,
      "cbid": 211, "correlation": 259403
    }
  },
  {
    "ph": "s", "id": 259403, "pid": 76337, "tid": -914061504, "ts": 1716454225588350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225588973, "dur": 4,
    "args": {
      "External id": 259405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259405,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259405, "pid": 5, "tid": 7, "ts": 1716454225588973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588359, "dur": 6,
    "args": {
      "External id": 259405,
      "cbid": 211, "correlation": 259405
    }
  },
  {
    "ph": "s", "id": 259405, "pid": 76337, "tid": -914061504, "ts": 1716454225588359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588368, "dur": 0,
    "args": {
      "External id": 259406,
      "cbid": 51, "correlation": 259406
    }
  },
  {
    "ph": "s", "id": 259406, "pid": 76337, "tid": -914061504, "ts": 1716454225588368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225588978, "dur": 709,
    "args": {
      "External id": 259407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259407,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259407, "pid": 5, "tid": 7, "ts": 1716454225588978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588369, "dur": 6,
    "args": {
      "External id": 259407,
      "cbid": 211, "correlation": 259407
    }
  },
  {
    "ph": "s", "id": 259407, "pid": 76337, "tid": -914061504, "ts": 1716454225588369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225589689, "dur": 60,
    "args": {
      "External id": 259412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259412,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259412, "pid": 5, "tid": 7, "ts": 1716454225589689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588397, "dur": 8,
    "args": {
      "External id": 259412,
      "cbid": 211, "correlation": 259412
    }
  },
  {
    "ph": "s", "id": 259412, "pid": 76337, "tid": -914061504, "ts": 1716454225588397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225589751, "dur": 50,
    "args": {
      "External id": 259420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259420,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259420, "pid": 5, "tid": 7, "ts": 1716454225589751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588430, "dur": 8,
    "args": {
      "External id": 259420,
      "cbid": 211, "correlation": 259420
    }
  },
  {
    "ph": "s", "id": 259420, "pid": 76337, "tid": -914061504, "ts": 1716454225588430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225589802, "dur": 36,
    "args": {
      "External id": 259428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259428,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259428, "pid": 5, "tid": 7, "ts": 1716454225589802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588460, "dur": 9,
    "args": {
      "External id": 259428,
      "cbid": 211, "correlation": 259428
    }
  },
  {
    "ph": "s", "id": 259428, "pid": 76337, "tid": -914061504, "ts": 1716454225588460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225589839, "dur": 51,
    "args": {
      "External id": 259448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259448,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 259448, "pid": 5, "tid": 7, "ts": 1716454225589839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588542, "dur": 12,
    "args": {
      "External id": 259448,
      "cbid": 211, "correlation": 259448
    }
  },
  {
    "ph": "s", "id": 259448, "pid": 76337, "tid": -914061504, "ts": 1716454225588542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225589891, "dur": 4,
    "args": {
      "External id": 259460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259460,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259460, "pid": 5, "tid": 7, "ts": 1716454225589891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588563, "dur": 6,
    "args": {
      "External id": 259460,
      "cbid": 211, "correlation": 259460
    }
  },
  {
    "ph": "s", "id": 259460, "pid": 76337, "tid": -914061504, "ts": 1716454225588563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225589896, "dur": 56,
    "args": {
      "External id": 259463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259463, "pid": 5, "tid": 7, "ts": 1716454225589896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588581, "dur": 6,
    "args": {
      "External id": 259463,
      "cbid": 211, "correlation": 259463
    }
  },
  {
    "ph": "s", "id": 259463, "pid": 76337, "tid": -914061504, "ts": 1716454225588581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588639, "dur": 0,
    "args": {
      "External id": 259474,
      "cbid": 317, "correlation": 259474
    }
  },
  {
    "ph": "f", "id": 259474, "pid": 76337, "tid": -914061504, "ts": 1716454225588639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588639, "dur": 0,
    "args": {
      "External id": 259475,
      "cbid": 203, "correlation": 259475
    }
  },
  {
    "ph": "f", "id": 259475, "pid": 76337, "tid": -914061504, "ts": 1716454225588639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588640, "dur": 0,
    "args": {
      "External id": 259476,
      "cbid": 205, "correlation": 259476
    }
  },
  {
    "ph": "f", "id": 259476, "pid": 76337, "tid": -914061504, "ts": 1716454225588640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588676, "dur": 2,
    "args": {
      "External id": 259480,
      "cbid": 251, "correlation": 259480
    }
  },
  {
    "ph": "f", "id": 259480, "pid": 76337, "tid": -914061504, "ts": 1716454225588676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588679, "dur": 1,
    "args": {
      "External id": 259481,
      "cbid": 251, "correlation": 259481
    }
  },
  {
    "ph": "f", "id": 259481, "pid": 76337, "tid": -914061504, "ts": 1716454225588679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588681, "dur": 1,
    "args": {
      "External id": 259482,
      "cbid": 251, "correlation": 259482
    }
  },
  {
    "ph": "f", "id": 259482, "pid": 76337, "tid": -914061504, "ts": 1716454225588681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588683, "dur": 1,
    "args": {
      "External id": 259483,
      "cbid": 251, "correlation": 259483
    }
  },
  {
    "ph": "f", "id": 259483, "pid": 76337, "tid": -914061504, "ts": 1716454225588683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588685, "dur": 1,
    "args": {
      "External id": 259484,
      "cbid": 251, "correlation": 259484
    }
  },
  {
    "ph": "f", "id": 259484, "pid": 76337, "tid": -914061504, "ts": 1716454225588685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588687, "dur": 1,
    "args": {
      "External id": 259485,
      "cbid": 251, "correlation": 259485
    }
  },
  {
    "ph": "f", "id": 259485, "pid": 76337, "tid": -914061504, "ts": 1716454225588687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588689, "dur": 1,
    "args": {
      "External id": 259486,
      "cbid": 251, "correlation": 259486
    }
  },
  {
    "ph": "f", "id": 259486, "pid": 76337, "tid": -914061504, "ts": 1716454225588689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588691, "dur": 1,
    "args": {
      "External id": 259487,
      "cbid": 251, "correlation": 259487
    }
  },
  {
    "ph": "f", "id": 259487, "pid": 76337, "tid": -914061504, "ts": 1716454225588691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588693, "dur": 0,
    "args": {
      "External id": 259488,
      "cbid": 251, "correlation": 259488
    }
  },
  {
    "ph": "f", "id": 259488, "pid": 76337, "tid": -914061504, "ts": 1716454225588693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225589954, "dur": 115,
    "args": {
      "External id": 259489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259489,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 259489, "pid": 5, "tid": 7, "ts": 1716454225589954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588698, "dur": 14,
    "args": {
      "External id": 259489,
      "cbid": 211, "correlation": 259489
    }
  },
  {
    "ph": "s", "id": 259489, "pid": 76337, "tid": -914061504, "ts": 1716454225588698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225590070, "dur": 59,
    "args": {
      "External id": 259495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259495,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259495, "pid": 5, "tid": 7, "ts": 1716454225590070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588736, "dur": 9,
    "args": {
      "External id": 259495,
      "cbid": 211, "correlation": 259495
    }
  },
  {
    "ph": "s", "id": 259495, "pid": 76337, "tid": -914061504, "ts": 1716454225588736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225590131, "dur": 578,
    "args": {
      "External id": 259504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259504,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259504, "pid": 5, "tid": 7, "ts": 1716454225590131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588829, "dur": 16,
    "args": {
      "External id": 259504,
      "cbid": 211, "correlation": 259504
    }
  },
  {
    "ph": "s", "id": 259504, "pid": 76337, "tid": -914061504, "ts": 1716454225588829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225590710, "dur": 184,
    "args": {
      "External id": 259526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259526,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259526, "pid": 5, "tid": 7, "ts": 1716454225590710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225588904, "dur": 14,
    "args": {
      "External id": 259526,
      "cbid": 211, "correlation": 259526
    }
  },
  {
    "ph": "s", "id": 259526, "pid": 76337, "tid": -914061504, "ts": 1716454225588904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589027, "dur": 2,
    "args": {
      "External id": 259537,
      "cbid": 251, "correlation": 259537
    }
  },
  {
    "ph": "f", "id": 259537, "pid": 76337, "tid": -914061504, "ts": 1716454225589027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225590896, "dur": 199,
    "args": {
      "External id": 259538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259538,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259538, "pid": 5, "tid": 7, "ts": 1716454225590896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589034, "dur": 15,
    "args": {
      "External id": 259538,
      "cbid": 211, "correlation": 259538
    }
  },
  {
    "ph": "s", "id": 259538, "pid": 76337, "tid": -914061504, "ts": 1716454225589034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589110, "dur": 1,
    "args": {
      "External id": 259549,
      "cbid": 251, "correlation": 259549
    }
  },
  {
    "ph": "f", "id": 259549, "pid": 76337, "tid": -914061504, "ts": 1716454225589110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225591096, "dur": 192,
    "args": {
      "External id": 259550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259550,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259550, "pid": 5, "tid": 7, "ts": 1716454225591096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589114, "dur": 12,
    "args": {
      "External id": 259550,
      "cbid": 211, "correlation": 259550
    }
  },
  {
    "ph": "s", "id": 259550, "pid": 76337, "tid": -914061504, "ts": 1716454225589114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589179, "dur": 1,
    "args": {
      "External id": 259561,
      "cbid": 251, "correlation": 259561
    }
  },
  {
    "ph": "f", "id": 259561, "pid": 76337, "tid": -914061504, "ts": 1716454225589179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225591290, "dur": 191,
    "args": {
      "External id": 259562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259562,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259562, "pid": 5, "tid": 7, "ts": 1716454225591290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589183, "dur": 11,
    "args": {
      "External id": 259562,
      "cbid": 211, "correlation": 259562
    }
  },
  {
    "ph": "s", "id": 259562, "pid": 76337, "tid": -914061504, "ts": 1716454225589183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225591483, "dur": 19052,
    "args": {
      "External id": 259583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259583,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 259583, "pid": 5, "tid": 7, "ts": 1716454225591483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589290, "dur": 16,
    "args": {
      "External id": 259583,
      "cbid": 211, "correlation": 259583
    }
  },
  {
    "ph": "s", "id": 259583, "pid": 76337, "tid": -914061504, "ts": 1716454225589290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589406, "dur": 2,
    "args": {
      "External id": 259601,
      "cbid": 251, "correlation": 259601
    }
  },
  {
    "ph": "f", "id": 259601, "pid": 76337, "tid": -914061504, "ts": 1716454225589406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225610536, "dur": 204,
    "args": {
      "External id": 259603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259603,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259603, "pid": 5, "tid": 7, "ts": 1716454225610536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589413, "dur": 14,
    "args": {
      "External id": 259603,
      "cbid": 211, "correlation": 259603
    }
  },
  {
    "ph": "s", "id": 259603, "pid": 76337, "tid": -914061504, "ts": 1716454225589413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225610742, "dur": 66,
    "args": {
      "External id": 259611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259611,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259611, "pid": 5, "tid": 7, "ts": 1716454225610742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589487, "dur": 13,
    "args": {
      "External id": 259611,
      "cbid": 211, "correlation": 259611
    }
  },
  {
    "ph": "s", "id": 259611, "pid": 76337, "tid": -914061504, "ts": 1716454225589487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225610809, "dur": 97,
    "args": {
      "External id": 259619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259619,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259619, "pid": 5, "tid": 7, "ts": 1716454225610809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589527, "dur": 8,
    "args": {
      "External id": 259619,
      "cbid": 211, "correlation": 259619
    }
  },
  {
    "ph": "s", "id": 259619, "pid": 76337, "tid": -914061504, "ts": 1716454225589527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225610908, "dur": 55,
    "args": {
      "External id": 259630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259630,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259630, "pid": 5, "tid": 7, "ts": 1716454225610908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589613, "dur": 13,
    "args": {
      "External id": 259630,
      "cbid": 211, "correlation": 259630
    }
  },
  {
    "ph": "s", "id": 259630, "pid": 76337, "tid": -914061504, "ts": 1716454225589613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225610964, "dur": 94,
    "args": {
      "External id": 259652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259652,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259652, "pid": 5, "tid": 7, "ts": 1716454225610964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589646, "dur": 8,
    "args": {
      "External id": 259652,
      "cbid": 211, "correlation": 259652
    }
  },
  {
    "ph": "s", "id": 259652, "pid": 76337, "tid": -914061504, "ts": 1716454225589646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589729, "dur": 1,
    "args": {
      "External id": 259663,
      "cbid": 251, "correlation": 259663
    }
  },
  {
    "ph": "f", "id": 259663, "pid": 76337, "tid": -914061504, "ts": 1716454225589729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225611059, "dur": 107,
    "args": {
      "External id": 259664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259664,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259664, "pid": 5, "tid": 7, "ts": 1716454225611059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589734, "dur": 13,
    "args": {
      "External id": 259664,
      "cbid": 211, "correlation": 259664
    }
  },
  {
    "ph": "s", "id": 259664, "pid": 76337, "tid": -914061504, "ts": 1716454225589734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589817, "dur": 2,
    "args": {
      "External id": 259675,
      "cbid": 251, "correlation": 259675
    }
  },
  {
    "ph": "f", "id": 259675, "pid": 76337, "tid": -914061504, "ts": 1716454225589817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589822, "dur": 0,
    "args": {
      "External id": 259676,
      "cbid": 251, "correlation": 259676
    }
  },
  {
    "ph": "f", "id": 259676, "pid": 76337, "tid": -914061504, "ts": 1716454225589822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225611167, "dur": 10,
    "args": {
      "External id": 259677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259677,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 259677, "pid": 5, "tid": 7, "ts": 1716454225611167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589824, "dur": 14,
    "args": {
      "External id": 259677,
      "cbid": 211, "correlation": 259677
    }
  },
  {
    "ph": "s", "id": 259677, "pid": 76337, "tid": -914061504, "ts": 1716454225589824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225611179, "dur": 5,
    "args": {
      "External id": 259679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259679,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 259679, "pid": 5, "tid": 7, "ts": 1716454225611179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589842, "dur": 8,
    "args": {
      "External id": 259679,
      "cbid": 211, "correlation": 259679
    }
  },
  {
    "ph": "s", "id": 259679, "pid": 76337, "tid": -914061504, "ts": 1716454225589842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589905, "dur": 1,
    "args": {
      "External id": 259690,
      "cbid": 251, "correlation": 259690
    }
  },
  {
    "ph": "f", "id": 259690, "pid": 76337, "tid": -914061504, "ts": 1716454225589905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589909, "dur": 0,
    "args": {
      "External id": 259691,
      "cbid": 251, "correlation": 259691
    }
  },
  {
    "ph": "f", "id": 259691, "pid": 76337, "tid": -914061504, "ts": 1716454225589909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225611185, "dur": 6,
    "args": {
      "External id": 259692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259692,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 259692, "pid": 5, "tid": 7, "ts": 1716454225611185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589911, "dur": 12,
    "args": {
      "External id": 259692,
      "cbid": 211, "correlation": 259692
    }
  },
  {
    "ph": "s", "id": 259692, "pid": 76337, "tid": -914061504, "ts": 1716454225589911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225611192, "dur": 4,
    "args": {
      "External id": 259694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259694,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 259694, "pid": 5, "tid": 7, "ts": 1716454225611192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225589924, "dur": 5,
    "args": {
      "External id": 259694,
      "cbid": 211, "correlation": 259694
    }
  },
  {
    "ph": "s", "id": 259694, "pid": 76337, "tid": -914061504, "ts": 1716454225589924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225611197, "dur": 157,
    "args": {
      "External id": 259715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259715,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 259715, "pid": 5, "tid": 7, "ts": 1716454225611197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590007, "dur": 12,
    "args": {
      "External id": 259715,
      "cbid": 211, "correlation": 259715
    }
  },
  {
    "ph": "s", "id": 259715, "pid": 76337, "tid": -914061504, "ts": 1716454225590007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590107, "dur": 2,
    "args": {
      "External id": 259733,
      "cbid": 251, "correlation": 259733
    }
  },
  {
    "ph": "f", "id": 259733, "pid": 76337, "tid": -914061504, "ts": 1716454225590107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225611356, "dur": 109,
    "args": {
      "External id": 259735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259735,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 259735, "pid": 5, "tid": 7, "ts": 1716454225611356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590114, "dur": 14,
    "args": {
      "External id": 259735,
      "cbid": 211, "correlation": 259735
    }
  },
  {
    "ph": "s", "id": 259735, "pid": 76337, "tid": -914061504, "ts": 1716454225590114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225611466, "dur": 35,
    "args": {
      "External id": 259743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259743,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259743, "pid": 5, "tid": 7, "ts": 1716454225611466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590185, "dur": 12,
    "args": {
      "External id": 259743,
      "cbid": 211, "correlation": 259743
    }
  },
  {
    "ph": "s", "id": 259743, "pid": 76337, "tid": -914061504, "ts": 1716454225590185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225611502, "dur": 69,
    "args": {
      "External id": 259751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259751, "pid": 5, "tid": 7, "ts": 1716454225611502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590227, "dur": 9,
    "args": {
      "External id": 259751,
      "cbid": 211, "correlation": 259751
    }
  },
  {
    "ph": "s", "id": 259751, "pid": 76337, "tid": -914061504, "ts": 1716454225590227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225611572, "dur": 93,
    "args": {
      "External id": 259773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259773,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259773, "pid": 5, "tid": 7, "ts": 1716454225611572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590279, "dur": 10,
    "args": {
      "External id": 259773,
      "cbid": 211, "correlation": 259773
    }
  },
  {
    "ph": "s", "id": 259773, "pid": 76337, "tid": -914061504, "ts": 1716454225590279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590368, "dur": 1,
    "args": {
      "External id": 259789,
      "cbid": 251, "correlation": 259789
    }
  },
  {
    "ph": "f", "id": 259789, "pid": 76337, "tid": -914061504, "ts": 1716454225590368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225611667, "dur": 583,
    "args": {
      "External id": 259791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259791,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259791, "pid": 5, "tid": 7, "ts": 1716454225611667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590374, "dur": 12,
    "args": {
      "External id": 259791,
      "cbid": 211, "correlation": 259791
    }
  },
  {
    "ph": "s", "id": 259791, "pid": 76337, "tid": -914061504, "ts": 1716454225590374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225612251, "dur": 248,
    "args": {
      "External id": 259799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259799, "pid": 5, "tid": 7, "ts": 1716454225612251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590452, "dur": 15,
    "args": {
      "External id": 259799,
      "cbid": 211, "correlation": 259799
    }
  },
  {
    "ph": "s", "id": 259799, "pid": 76337, "tid": -914061504, "ts": 1716454225590452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225612500, "dur": 252,
    "args": {
      "External id": 259807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259807,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259807, "pid": 5, "tid": 7, "ts": 1716454225612500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590488, "dur": 10,
    "args": {
      "External id": 259807,
      "cbid": 211, "correlation": 259807
    }
  },
  {
    "ph": "s", "id": 259807, "pid": 76337, "tid": -914061504, "ts": 1716454225590488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590572, "dur": 2,
    "args": {
      "External id": 259823,
      "cbid": 251, "correlation": 259823
    }
  },
  {
    "ph": "f", "id": 259823, "pid": 76337, "tid": -914061504, "ts": 1716454225590572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590577, "dur": 0,
    "args": {
      "External id": 259825,
      "cbid": 251, "correlation": 259825
    }
  },
  {
    "ph": "f", "id": 259825, "pid": 76337, "tid": -914061504, "ts": 1716454225590577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225612754, "dur": 361,
    "args": {
      "External id": 259826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259826,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 259826, "pid": 5, "tid": 7, "ts": 1716454225612754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590582, "dur": 14,
    "args": {
      "External id": 259826,
      "cbid": 211, "correlation": 259826
    }
  },
  {
    "ph": "s", "id": 259826, "pid": 76337, "tid": -914061504, "ts": 1716454225590582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225613116, "dur": 50,
    "args": {
      "External id": 259834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259834,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259834, "pid": 5, "tid": 7, "ts": 1716454225613116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590624, "dur": 10,
    "args": {
      "External id": 259834,
      "cbid": 211, "correlation": 259834
    }
  },
  {
    "ph": "s", "id": 259834, "pid": 76337, "tid": -914061504, "ts": 1716454225590624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225613167, "dur": 160,
    "args": {
      "External id": 259845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259845, "pid": 5, "tid": 7, "ts": 1716454225613167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590696, "dur": 13,
    "args": {
      "External id": 259845,
      "cbid": 211, "correlation": 259845
    }
  },
  {
    "ph": "s", "id": 259845, "pid": 76337, "tid": -914061504, "ts": 1716454225590696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590762, "dur": 0,
    "args": {
      "External id": 259857,
      "cbid": 317, "correlation": 259857
    }
  },
  {
    "ph": "f", "id": 259857, "pid": 76337, "tid": -914061504, "ts": 1716454225590762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590763, "dur": 0,
    "args": {
      "External id": 259858,
      "cbid": 203, "correlation": 259858
    }
  },
  {
    "ph": "f", "id": 259858, "pid": 76337, "tid": -914061504, "ts": 1716454225590763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590763, "dur": 0,
    "args": {
      "External id": 259859,
      "cbid": 205, "correlation": 259859
    }
  },
  {
    "ph": "f", "id": 259859, "pid": 76337, "tid": -914061504, "ts": 1716454225590763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590787, "dur": 1,
    "args": {
      "External id": 259863,
      "cbid": 251, "correlation": 259863
    }
  },
  {
    "ph": "f", "id": 259863, "pid": 76337, "tid": -914061504, "ts": 1716454225590787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590789, "dur": 0,
    "args": {
      "External id": 259864,
      "cbid": 251, "correlation": 259864
    }
  },
  {
    "ph": "f", "id": 259864, "pid": 76337, "tid": -914061504, "ts": 1716454225590789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590790, "dur": 0,
    "args": {
      "External id": 259865,
      "cbid": 251, "correlation": 259865
    }
  },
  {
    "ph": "f", "id": 259865, "pid": 76337, "tid": -914061504, "ts": 1716454225590790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590791, "dur": 0,
    "args": {
      "External id": 259866,
      "cbid": 251, "correlation": 259866
    }
  },
  {
    "ph": "f", "id": 259866, "pid": 76337, "tid": -914061504, "ts": 1716454225590791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590792, "dur": 0,
    "args": {
      "External id": 259867,
      "cbid": 251, "correlation": 259867
    }
  },
  {
    "ph": "f", "id": 259867, "pid": 76337, "tid": -914061504, "ts": 1716454225590792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590793, "dur": 0,
    "args": {
      "External id": 259868,
      "cbid": 251, "correlation": 259868
    }
  },
  {
    "ph": "f", "id": 259868, "pid": 76337, "tid": -914061504, "ts": 1716454225590793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590794, "dur": 0,
    "args": {
      "External id": 259869,
      "cbid": 251, "correlation": 259869
    }
  },
  {
    "ph": "f", "id": 259869, "pid": 76337, "tid": -914061504, "ts": 1716454225590794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590795, "dur": 0,
    "args": {
      "External id": 259870,
      "cbid": 251, "correlation": 259870
    }
  },
  {
    "ph": "f", "id": 259870, "pid": 76337, "tid": -914061504, "ts": 1716454225590795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590796, "dur": 0,
    "args": {
      "External id": 259871,
      "cbid": 251, "correlation": 259871
    }
  },
  {
    "ph": "f", "id": 259871, "pid": 76337, "tid": -914061504, "ts": 1716454225590796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225613329, "dur": 117,
    "args": {
      "External id": 259872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259872,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 259872, "pid": 5, "tid": 7, "ts": 1716454225613329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590799, "dur": 12,
    "args": {
      "External id": 259872,
      "cbid": 211, "correlation": 259872
    }
  },
  {
    "ph": "s", "id": 259872, "pid": 76337, "tid": -914061504, "ts": 1716454225590799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225613448, "dur": 61,
    "args": {
      "External id": 259878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259878, "pid": 5, "tid": 7, "ts": 1716454225613448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590834, "dur": 9,
    "args": {
      "External id": 259878,
      "cbid": 211, "correlation": 259878
    }
  },
  {
    "ph": "s", "id": 259878, "pid": 76337, "tid": -914061504, "ts": 1716454225590834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225613510, "dur": 50,
    "args": {
      "External id": 259886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259886,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259886, "pid": 5, "tid": 7, "ts": 1716454225613510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590867, "dur": 8,
    "args": {
      "External id": 259886,
      "cbid": 211, "correlation": 259886
    }
  },
  {
    "ph": "s", "id": 259886, "pid": 76337, "tid": -914061504, "ts": 1716454225590867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225613561, "dur": 53,
    "args": {
      "External id": 259906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259906,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 259906, "pid": 5, "tid": 7, "ts": 1716454225613561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590939, "dur": 12,
    "args": {
      "External id": 259906,
      "cbid": 211, "correlation": 259906
    }
  },
  {
    "ph": "s", "id": 259906, "pid": 76337, "tid": -914061504, "ts": 1716454225590939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225613615, "dur": 5,
    "args": {
      "External id": 259918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259918,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259918, "pid": 5, "tid": 7, "ts": 1716454225613615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590962, "dur": 6,
    "args": {
      "External id": 259918,
      "cbid": 211, "correlation": 259918
    }
  },
  {
    "ph": "s", "id": 259918, "pid": 76337, "tid": -914061504, "ts": 1716454225590962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225613621, "dur": 57,
    "args": {
      "External id": 259921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259921, "pid": 5, "tid": 7, "ts": 1716454225613621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225590988, "dur": 7,
    "args": {
      "External id": 259921,
      "cbid": 211, "correlation": 259921
    }
  },
  {
    "ph": "s", "id": 259921, "pid": 76337, "tid": -914061504, "ts": 1716454225590988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225613679, "dur": 37,
    "args": {
      "External id": 259930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259930,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259930, "pid": 5, "tid": 7, "ts": 1716454225613679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591030, "dur": 10,
    "args": {
      "External id": 259930,
      "cbid": 211, "correlation": 259930
    }
  },
  {
    "ph": "s", "id": 259930, "pid": 76337, "tid": -914061504, "ts": 1716454225591030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591083, "dur": 0,
    "args": {
      "External id": 259940,
      "cbid": 317, "correlation": 259940
    }
  },
  {
    "ph": "f", "id": 259940, "pid": 76337, "tid": -914061504, "ts": 1716454225591083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591084, "dur": 0,
    "args": {
      "External id": 259941,
      "cbid": 203, "correlation": 259941
    }
  },
  {
    "ph": "f", "id": 259941, "pid": 76337, "tid": -914061504, "ts": 1716454225591084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591085, "dur": 0,
    "args": {
      "External id": 259942,
      "cbid": 205, "correlation": 259942
    }
  },
  {
    "ph": "f", "id": 259942, "pid": 76337, "tid": -914061504, "ts": 1716454225591085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225613717, "dur": 41,
    "args": {
      "External id": 259946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259946, "pid": 5, "tid": 7, "ts": 1716454225613717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591099, "dur": 12,
    "args": {
      "External id": 259946,
      "cbid": 211, "correlation": 259946
    }
  },
  {
    "ph": "s", "id": 259946, "pid": 76337, "tid": -914061504, "ts": 1716454225591099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225613759, "dur": 15,
    "args": {
      "External id": 259948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259948, "pid": 5, "tid": 7, "ts": 1716454225613759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591114, "dur": 5,
    "args": {
      "External id": 259948,
      "cbid": 211, "correlation": 259948
    }
  },
  {
    "ph": "s", "id": 259948, "pid": 76337, "tid": -914061504, "ts": 1716454225591114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225613775, "dur": 4,
    "args": {
      "External id": 259950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259950, "pid": 5, "tid": 7, "ts": 1716454225613775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591124, "dur": 5,
    "args": {
      "External id": 259950,
      "cbid": 211, "correlation": 259950
    }
  },
  {
    "ph": "s", "id": 259950, "pid": 76337, "tid": -914061504, "ts": 1716454225591124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591133, "dur": 0,
    "args": {
      "External id": 259951,
      "cbid": 51, "correlation": 259951
    }
  },
  {
    "ph": "s", "id": 259951, "pid": 76337, "tid": -914061504, "ts": 1716454225591133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225613780, "dur": 713,
    "args": {
      "External id": 259952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259952,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 259952, "pid": 5, "tid": 7, "ts": 1716454225613780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591134, "dur": 5,
    "args": {
      "External id": 259952,
      "cbid": 211, "correlation": 259952
    }
  },
  {
    "ph": "s", "id": 259952, "pid": 76337, "tid": -914061504, "ts": 1716454225591134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225614495, "dur": 60,
    "args": {
      "External id": 259957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259957, "pid": 5, "tid": 7, "ts": 1716454225614495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591161, "dur": 9,
    "args": {
      "External id": 259957,
      "cbid": 211, "correlation": 259957
    }
  },
  {
    "ph": "s", "id": 259957, "pid": 76337, "tid": -914061504, "ts": 1716454225591161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225614556, "dur": 3,
    "args": {
      "External id": 259965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259965,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 259965, "pid": 5, "tid": 7, "ts": 1716454225614556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591205, "dur": 9,
    "args": {
      "External id": 259965,
      "cbid": 211, "correlation": 259965
    }
  },
  {
    "ph": "s", "id": 259965, "pid": 76337, "tid": -914061504, "ts": 1716454225591205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591271, "dur": 1,
    "args": {
      "External id": 259981,
      "cbid": 251, "correlation": 259981
    }
  },
  {
    "ph": "f", "id": 259981, "pid": 76337, "tid": -914061504, "ts": 1716454225591271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591276, "dur": 0,
    "args": {
      "External id": 259983,
      "cbid": 251, "correlation": 259983
    }
  },
  {
    "ph": "f", "id": 259983, "pid": 76337, "tid": -914061504, "ts": 1716454225591276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225614560, "dur": 12,
    "args": {
      "External id": 259984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259984,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 259984, "pid": 5, "tid": 7, "ts": 1716454225614560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591278, "dur": 11,
    "args": {
      "External id": 259984,
      "cbid": 211, "correlation": 259984
    }
  },
  {
    "ph": "s", "id": 259984, "pid": 76337, "tid": -914061504, "ts": 1716454225591278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225614574, "dur": 5,
    "args": {
      "External id": 259986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259986,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 259986, "pid": 5, "tid": 7, "ts": 1716454225614574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591291, "dur": 6,
    "args": {
      "External id": 259986,
      "cbid": 211, "correlation": 259986
    }
  },
  {
    "ph": "s", "id": 259986, "pid": 76337, "tid": -914061504, "ts": 1716454225591291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225614580, "dur": 54,
    "args": {
      "External id": 259996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 259996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 259996, "pid": 5, "tid": 7, "ts": 1716454225614580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591349, "dur": 13,
    "args": {
      "External id": 259996,
      "cbid": 211, "correlation": 259996
    }
  },
  {
    "ph": "s", "id": 259996, "pid": 76337, "tid": -914061504, "ts": 1716454225591349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225614636, "dur": 51,
    "args": {
      "External id": 260016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260016,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 260016, "pid": 5, "tid": 7, "ts": 1716454225614636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591415, "dur": 11,
    "args": {
      "External id": 260016,
      "cbid": 211, "correlation": 260016
    }
  },
  {
    "ph": "s", "id": 260016, "pid": 76337, "tid": -914061504, "ts": 1716454225591415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225614688, "dur": 4,
    "args": {
      "External id": 260028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260028,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 260028, "pid": 5, "tid": 7, "ts": 1716454225614688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591435, "dur": 6,
    "args": {
      "External id": 260028,
      "cbid": 211, "correlation": 260028
    }
  },
  {
    "ph": "s", "id": 260028, "pid": 76337, "tid": -914061504, "ts": 1716454225591435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225614694, "dur": 55,
    "args": {
      "External id": 260031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260031, "pid": 5, "tid": 7, "ts": 1716454225614694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591455, "dur": 6,
    "args": {
      "External id": 260031,
      "cbid": 211, "correlation": 260031
    }
  },
  {
    "ph": "s", "id": 260031, "pid": 76337, "tid": -914061504, "ts": 1716454225591455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225614750, "dur": 36,
    "args": {
      "External id": 260040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260040,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260040, "pid": 5, "tid": 7, "ts": 1716454225614750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591495, "dur": 11,
    "args": {
      "External id": 260040,
      "cbid": 211, "correlation": 260040
    }
  },
  {
    "ph": "s", "id": 260040, "pid": 76337, "tid": -914061504, "ts": 1716454225591495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591558, "dur": 0,
    "args": {
      "External id": 260050,
      "cbid": 317, "correlation": 260050
    }
  },
  {
    "ph": "f", "id": 260050, "pid": 76337, "tid": -914061504, "ts": 1716454225591558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591559, "dur": 0,
    "args": {
      "External id": 260051,
      "cbid": 203, "correlation": 260051
    }
  },
  {
    "ph": "f", "id": 260051, "pid": 76337, "tid": -914061504, "ts": 1716454225591559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591560, "dur": 0,
    "args": {
      "External id": 260052,
      "cbid": 205, "correlation": 260052
    }
  },
  {
    "ph": "f", "id": 260052, "pid": 76337, "tid": -914061504, "ts": 1716454225591560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225614788, "dur": 40,
    "args": {
      "External id": 260056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260056, "pid": 5, "tid": 7, "ts": 1716454225614788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591573, "dur": 12,
    "args": {
      "External id": 260056,
      "cbid": 211, "correlation": 260056
    }
  },
  {
    "ph": "s", "id": 260056, "pid": 76337, "tid": -914061504, "ts": 1716454225591573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225614830, "dur": 15,
    "args": {
      "External id": 260058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260058,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260058, "pid": 5, "tid": 7, "ts": 1716454225614830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591587, "dur": 5,
    "args": {
      "External id": 260058,
      "cbid": 211, "correlation": 260058
    }
  },
  {
    "ph": "s", "id": 260058, "pid": 76337, "tid": -914061504, "ts": 1716454225591587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225614845, "dur": 3,
    "args": {
      "External id": 260060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 260060, "pid": 5, "tid": 7, "ts": 1716454225614845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591597, "dur": 6,
    "args": {
      "External id": 260060,
      "cbid": 211, "correlation": 260060
    }
  },
  {
    "ph": "s", "id": 260060, "pid": 76337, "tid": -914061504, "ts": 1716454225591597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591607, "dur": 0,
    "args": {
      "External id": 260061,
      "cbid": 51, "correlation": 260061
    }
  },
  {
    "ph": "s", "id": 260061, "pid": 76337, "tid": -914061504, "ts": 1716454225591607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225614850, "dur": 706,
    "args": {
      "External id": 260062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260062,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260062, "pid": 5, "tid": 7, "ts": 1716454225614850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591607, "dur": 5,
    "args": {
      "External id": 260062,
      "cbid": 211, "correlation": 260062
    }
  },
  {
    "ph": "s", "id": 260062, "pid": 76337, "tid": -914061504, "ts": 1716454225591607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225615557, "dur": 60,
    "args": {
      "External id": 260067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260067, "pid": 5, "tid": 7, "ts": 1716454225615557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591634, "dur": 9,
    "args": {
      "External id": 260067,
      "cbid": 211, "correlation": 260067
    }
  },
  {
    "ph": "s", "id": 260067, "pid": 76337, "tid": -914061504, "ts": 1716454225591634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225615618, "dur": 50,
    "args": {
      "External id": 260075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260075,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260075, "pid": 5, "tid": 7, "ts": 1716454225615618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591666, "dur": 8,
    "args": {
      "External id": 260075,
      "cbid": 211, "correlation": 260075
    }
  },
  {
    "ph": "s", "id": 260075, "pid": 76337, "tid": -914061504, "ts": 1716454225591666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225615670, "dur": 36,
    "args": {
      "External id": 260083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260083,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260083, "pid": 5, "tid": 7, "ts": 1716454225615670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591695, "dur": 9,
    "args": {
      "External id": 260083,
      "cbid": 211, "correlation": 260083
    }
  },
  {
    "ph": "s", "id": 260083, "pid": 76337, "tid": -914061504, "ts": 1716454225591695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225615707, "dur": 51,
    "args": {
      "External id": 260103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260103,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 260103, "pid": 5, "tid": 7, "ts": 1716454225615707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591773, "dur": 12,
    "args": {
      "External id": 260103,
      "cbid": 211, "correlation": 260103
    }
  },
  {
    "ph": "s", "id": 260103, "pid": 76337, "tid": -914061504, "ts": 1716454225591773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225615759, "dur": 4,
    "args": {
      "External id": 260115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260115,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 260115, "pid": 5, "tid": 7, "ts": 1716454225615759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591795, "dur": 6,
    "args": {
      "External id": 260115,
      "cbid": 211, "correlation": 260115
    }
  },
  {
    "ph": "s", "id": 260115, "pid": 76337, "tid": -914061504, "ts": 1716454225591795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225615764, "dur": 56,
    "args": {
      "External id": 260118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260118, "pid": 5, "tid": 7, "ts": 1716454225615764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591813, "dur": 6,
    "args": {
      "External id": 260118,
      "cbid": 211, "correlation": 260118
    }
  },
  {
    "ph": "s", "id": 260118, "pid": 76337, "tid": -914061504, "ts": 1716454225591813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591869, "dur": 0,
    "args": {
      "External id": 260129,
      "cbid": 317, "correlation": 260129
    }
  },
  {
    "ph": "f", "id": 260129, "pid": 76337, "tid": -914061504, "ts": 1716454225591869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591870, "dur": 0,
    "args": {
      "External id": 260130,
      "cbid": 203, "correlation": 260130
    }
  },
  {
    "ph": "f", "id": 260130, "pid": 76337, "tid": -914061504, "ts": 1716454225591870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591871, "dur": 0,
    "args": {
      "External id": 260131,
      "cbid": 205, "correlation": 260131
    }
  },
  {
    "ph": "f", "id": 260131, "pid": 76337, "tid": -914061504, "ts": 1716454225591871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591892, "dur": 1,
    "args": {
      "External id": 260135,
      "cbid": 251, "correlation": 260135
    }
  },
  {
    "ph": "f", "id": 260135, "pid": 76337, "tid": -914061504, "ts": 1716454225591892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591894, "dur": 0,
    "args": {
      "External id": 260136,
      "cbid": 251, "correlation": 260136
    }
  },
  {
    "ph": "f", "id": 260136, "pid": 76337, "tid": -914061504, "ts": 1716454225591894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591895, "dur": 0,
    "args": {
      "External id": 260137,
      "cbid": 251, "correlation": 260137
    }
  },
  {
    "ph": "f", "id": 260137, "pid": 76337, "tid": -914061504, "ts": 1716454225591895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591895, "dur": 0,
    "args": {
      "External id": 260138,
      "cbid": 251, "correlation": 260138
    }
  },
  {
    "ph": "f", "id": 260138, "pid": 76337, "tid": -914061504, "ts": 1716454225591895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591896, "dur": 0,
    "args": {
      "External id": 260139,
      "cbid": 251, "correlation": 260139
    }
  },
  {
    "ph": "f", "id": 260139, "pid": 76337, "tid": -914061504, "ts": 1716454225591896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591897, "dur": 0,
    "args": {
      "External id": 260140,
      "cbid": 251, "correlation": 260140
    }
  },
  {
    "ph": "f", "id": 260140, "pid": 76337, "tid": -914061504, "ts": 1716454225591897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591898, "dur": 0,
    "args": {
      "External id": 260141,
      "cbid": 251, "correlation": 260141
    }
  },
  {
    "ph": "f", "id": 260141, "pid": 76337, "tid": -914061504, "ts": 1716454225591898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591898, "dur": 0,
    "args": {
      "External id": 260142,
      "cbid": 251, "correlation": 260142
    }
  },
  {
    "ph": "f", "id": 260142, "pid": 76337, "tid": -914061504, "ts": 1716454225591898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591900, "dur": 0,
    "args": {
      "External id": 260143,
      "cbid": 251, "correlation": 260143
    }
  },
  {
    "ph": "f", "id": 260143, "pid": 76337, "tid": -914061504, "ts": 1716454225591900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225615821, "dur": 111,
    "args": {
      "External id": 260144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260144,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260144, "pid": 5, "tid": 7, "ts": 1716454225615821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591902, "dur": 13,
    "args": {
      "External id": 260144,
      "cbid": 211, "correlation": 260144
    }
  },
  {
    "ph": "s", "id": 260144, "pid": 76337, "tid": -914061504, "ts": 1716454225591902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225615934, "dur": 60,
    "args": {
      "External id": 260150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260150,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260150, "pid": 5, "tid": 7, "ts": 1716454225615934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225591937, "dur": 9,
    "args": {
      "External id": 260150,
      "cbid": 211, "correlation": 260150
    }
  },
  {
    "ph": "s", "id": 260150, "pid": 76337, "tid": -914061504, "ts": 1716454225591937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225615995, "dur": 570,
    "args": {
      "External id": 260159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260159,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260159, "pid": 5, "tid": 7, "ts": 1716454225615995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592029, "dur": 14,
    "args": {
      "External id": 260159,
      "cbid": 211, "correlation": 260159
    }
  },
  {
    "ph": "s", "id": 260159, "pid": 76337, "tid": -914061504, "ts": 1716454225592029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225616566, "dur": 184,
    "args": {
      "External id": 260181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260181,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260181, "pid": 5, "tid": 7, "ts": 1716454225616566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592087, "dur": 10,
    "args": {
      "External id": 260181,
      "cbid": 211, "correlation": 260181
    }
  },
  {
    "ph": "s", "id": 260181, "pid": 76337, "tid": -914061504, "ts": 1716454225592087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592173, "dur": 1,
    "args": {
      "External id": 260192,
      "cbid": 251, "correlation": 260192
    }
  },
  {
    "ph": "f", "id": 260192, "pid": 76337, "tid": -914061504, "ts": 1716454225592173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225616752, "dur": 196,
    "args": {
      "External id": 260193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260193,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260193, "pid": 5, "tid": 7, "ts": 1716454225616752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592178, "dur": 13,
    "args": {
      "External id": 260193,
      "cbid": 211, "correlation": 260193
    }
  },
  {
    "ph": "s", "id": 260193, "pid": 76337, "tid": -914061504, "ts": 1716454225592178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592246, "dur": 1,
    "args": {
      "External id": 260204,
      "cbid": 251, "correlation": 260204
    }
  },
  {
    "ph": "f", "id": 260204, "pid": 76337, "tid": -914061504, "ts": 1716454225592246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225616950, "dur": 189,
    "args": {
      "External id": 260205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260205,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260205, "pid": 5, "tid": 7, "ts": 1716454225616950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592250, "dur": 12,
    "args": {
      "External id": 260205,
      "cbid": 211, "correlation": 260205
    }
  },
  {
    "ph": "s", "id": 260205, "pid": 76337, "tid": -914061504, "ts": 1716454225592250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592313, "dur": 1,
    "args": {
      "External id": 260216,
      "cbid": 251, "correlation": 260216
    }
  },
  {
    "ph": "f", "id": 260216, "pid": 76337, "tid": -914061504, "ts": 1716454225592313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225617140, "dur": 190,
    "args": {
      "External id": 260217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260217,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260217, "pid": 5, "tid": 7, "ts": 1716454225617140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592317, "dur": 11,
    "args": {
      "External id": 260217,
      "cbid": 211, "correlation": 260217
    }
  },
  {
    "ph": "s", "id": 260217, "pid": 76337, "tid": -914061504, "ts": 1716454225592317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225617331, "dur": 18971,
    "args": {
      "External id": 260238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260238,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260238, "pid": 5, "tid": 7, "ts": 1716454225617331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592397, "dur": 13,
    "args": {
      "External id": 260238,
      "cbid": 211, "correlation": 260238
    }
  },
  {
    "ph": "s", "id": 260238, "pid": 76337, "tid": -914061504, "ts": 1716454225592397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592496, "dur": 1,
    "args": {
      "External id": 260256,
      "cbid": 251, "correlation": 260256
    }
  },
  {
    "ph": "f", "id": 260256, "pid": 76337, "tid": -914061504, "ts": 1716454225592496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225636303, "dur": 207,
    "args": {
      "External id": 260258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260258,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260258, "pid": 5, "tid": 7, "ts": 1716454225636303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592502, "dur": 15,
    "args": {
      "External id": 260258,
      "cbid": 211, "correlation": 260258
    }
  },
  {
    "ph": "s", "id": 260258, "pid": 76337, "tid": -914061504, "ts": 1716454225592502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225636511, "dur": 65,
    "args": {
      "External id": 260266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260266,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260266, "pid": 5, "tid": 7, "ts": 1716454225636511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592574, "dur": 13,
    "args": {
      "External id": 260266,
      "cbid": 211, "correlation": 260266
    }
  },
  {
    "ph": "s", "id": 260266, "pid": 76337, "tid": -914061504, "ts": 1716454225592574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225636578, "dur": 96,
    "args": {
      "External id": 260274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260274,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260274, "pid": 5, "tid": 7, "ts": 1716454225636578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592615, "dur": 9,
    "args": {
      "External id": 260274,
      "cbid": 211, "correlation": 260274
    }
  },
  {
    "ph": "s", "id": 260274, "pid": 76337, "tid": -914061504, "ts": 1716454225592615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225636676, "dur": 56,
    "args": {
      "External id": 260285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260285,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260285, "pid": 5, "tid": 7, "ts": 1716454225636676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592686, "dur": 12,
    "args": {
      "External id": 260285,
      "cbid": 211, "correlation": 260285
    }
  },
  {
    "ph": "s", "id": 260285, "pid": 76337, "tid": -914061504, "ts": 1716454225592686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225636732, "dur": 94,
    "args": {
      "External id": 260307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260307,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260307, "pid": 5, "tid": 7, "ts": 1716454225636732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592718, "dur": 8,
    "args": {
      "External id": 260307,
      "cbid": 211, "correlation": 260307
    }
  },
  {
    "ph": "s", "id": 260307, "pid": 76337, "tid": -914061504, "ts": 1716454225592718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592803, "dur": 1,
    "args": {
      "External id": 260318,
      "cbid": 251, "correlation": 260318
    }
  },
  {
    "ph": "f", "id": 260318, "pid": 76337, "tid": -914061504, "ts": 1716454225592803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225636828, "dur": 106,
    "args": {
      "External id": 260319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260319,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260319, "pid": 5, "tid": 7, "ts": 1716454225636828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592808, "dur": 13,
    "args": {
      "External id": 260319,
      "cbid": 211, "correlation": 260319
    }
  },
  {
    "ph": "s", "id": 260319, "pid": 76337, "tid": -914061504, "ts": 1716454225592808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592881, "dur": 1,
    "args": {
      "External id": 260330,
      "cbid": 251, "correlation": 260330
    }
  },
  {
    "ph": "f", "id": 260330, "pid": 76337, "tid": -914061504, "ts": 1716454225592881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592884, "dur": 0,
    "args": {
      "External id": 260331,
      "cbid": 251, "correlation": 260331
    }
  },
  {
    "ph": "f", "id": 260331, "pid": 76337, "tid": -914061504, "ts": 1716454225592884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225636935, "dur": 10,
    "args": {
      "External id": 260332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260332,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 260332, "pid": 5, "tid": 7, "ts": 1716454225636935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592886, "dur": 12,
    "args": {
      "External id": 260332,
      "cbid": 211, "correlation": 260332
    }
  },
  {
    "ph": "s", "id": 260332, "pid": 76337, "tid": -914061504, "ts": 1716454225592886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225636947, "dur": 5,
    "args": {
      "External id": 260334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260334,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 260334, "pid": 5, "tid": 7, "ts": 1716454225636947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592900, "dur": 7,
    "args": {
      "External id": 260334,
      "cbid": 211, "correlation": 260334
    }
  },
  {
    "ph": "s", "id": 260334, "pid": 76337, "tid": -914061504, "ts": 1716454225592900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592961, "dur": 1,
    "args": {
      "External id": 260345,
      "cbid": 251, "correlation": 260345
    }
  },
  {
    "ph": "f", "id": 260345, "pid": 76337, "tid": -914061504, "ts": 1716454225592961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592965, "dur": 0,
    "args": {
      "External id": 260346,
      "cbid": 251, "correlation": 260346
    }
  },
  {
    "ph": "f", "id": 260346, "pid": 76337, "tid": -914061504, "ts": 1716454225592965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225636953, "dur": 6,
    "args": {
      "External id": 260347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260347,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 260347, "pid": 5, "tid": 7, "ts": 1716454225636953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592966, "dur": 20,
    "args": {
      "External id": 260347,
      "cbid": 211, "correlation": 260347
    }
  },
  {
    "ph": "s", "id": 260347, "pid": 76337, "tid": -914061504, "ts": 1716454225592966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225636961, "dur": 4,
    "args": {
      "External id": 260349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260349,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 260349, "pid": 5, "tid": 7, "ts": 1716454225636961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225592988, "dur": 6,
    "args": {
      "External id": 260349,
      "cbid": 211, "correlation": 260349
    }
  },
  {
    "ph": "s", "id": 260349, "pid": 76337, "tid": -914061504, "ts": 1716454225592988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225636966, "dur": 157,
    "args": {
      "External id": 260370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260370,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260370, "pid": 5, "tid": 7, "ts": 1716454225636966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593063, "dur": 13,
    "args": {
      "External id": 260370,
      "cbid": 211, "correlation": 260370
    }
  },
  {
    "ph": "s", "id": 260370, "pid": 76337, "tid": -914061504, "ts": 1716454225593063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593160, "dur": 1,
    "args": {
      "External id": 260388,
      "cbid": 251, "correlation": 260388
    }
  },
  {
    "ph": "f", "id": 260388, "pid": 76337, "tid": -914061504, "ts": 1716454225593160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225637125, "dur": 108,
    "args": {
      "External id": 260390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260390,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260390, "pid": 5, "tid": 7, "ts": 1716454225637125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593166, "dur": 14,
    "args": {
      "External id": 260390,
      "cbid": 211, "correlation": 260390
    }
  },
  {
    "ph": "s", "id": 260390, "pid": 76337, "tid": -914061504, "ts": 1716454225593166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225637234, "dur": 35,
    "args": {
      "External id": 260398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260398,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260398, "pid": 5, "tid": 7, "ts": 1716454225637234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593235, "dur": 12,
    "args": {
      "External id": 260398,
      "cbid": 211, "correlation": 260398
    }
  },
  {
    "ph": "s", "id": 260398, "pid": 76337, "tid": -914061504, "ts": 1716454225593235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225637271, "dur": 67,
    "args": {
      "External id": 260406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260406,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260406, "pid": 5, "tid": 7, "ts": 1716454225637271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593276, "dur": 9,
    "args": {
      "External id": 260406,
      "cbid": 211, "correlation": 260406
    }
  },
  {
    "ph": "s", "id": 260406, "pid": 76337, "tid": -914061504, "ts": 1716454225593276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225637339, "dur": 94,
    "args": {
      "External id": 260428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260428,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260428, "pid": 5, "tid": 7, "ts": 1716454225637339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593327, "dur": 10,
    "args": {
      "External id": 260428,
      "cbid": 211, "correlation": 260428
    }
  },
  {
    "ph": "s", "id": 260428, "pid": 76337, "tid": -914061504, "ts": 1716454225593327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593412, "dur": 1,
    "args": {
      "External id": 260444,
      "cbid": 251, "correlation": 260444
    }
  },
  {
    "ph": "f", "id": 260444, "pid": 76337, "tid": -914061504, "ts": 1716454225593412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225637434, "dur": 583,
    "args": {
      "External id": 260446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260446,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260446, "pid": 5, "tid": 7, "ts": 1716454225637434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593418, "dur": 13,
    "args": {
      "External id": 260446,
      "cbid": 211, "correlation": 260446
    }
  },
  {
    "ph": "s", "id": 260446, "pid": 76337, "tid": -914061504, "ts": 1716454225593418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225638019, "dur": 247,
    "args": {
      "External id": 260454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260454, "pid": 5, "tid": 7, "ts": 1716454225638019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593484, "dur": 12,
    "args": {
      "External id": 260454,
      "cbid": 211, "correlation": 260454
    }
  },
  {
    "ph": "s", "id": 260454, "pid": 76337, "tid": -914061504, "ts": 1716454225593484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225638267, "dur": 254,
    "args": {
      "External id": 260462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260462, "pid": 5, "tid": 7, "ts": 1716454225638267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593514, "dur": 9,
    "args": {
      "External id": 260462,
      "cbid": 211, "correlation": 260462
    }
  },
  {
    "ph": "s", "id": 260462, "pid": 76337, "tid": -914061504, "ts": 1716454225593514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593596, "dur": 1,
    "args": {
      "External id": 260478,
      "cbid": 251, "correlation": 260478
    }
  },
  {
    "ph": "f", "id": 260478, "pid": 76337, "tid": -914061504, "ts": 1716454225593596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593601, "dur": 0,
    "args": {
      "External id": 260480,
      "cbid": 251, "correlation": 260480
    }
  },
  {
    "ph": "f", "id": 260480, "pid": 76337, "tid": -914061504, "ts": 1716454225593601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225638523, "dur": 361,
    "args": {
      "External id": 260481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260481,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 260481, "pid": 5, "tid": 7, "ts": 1716454225638523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593604, "dur": 12,
    "args": {
      "External id": 260481,
      "cbid": 211, "correlation": 260481
    }
  },
  {
    "ph": "s", "id": 260481, "pid": 76337, "tid": -914061504, "ts": 1716454225593604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225638885, "dur": 50,
    "args": {
      "External id": 260489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260489,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260489, "pid": 5, "tid": 7, "ts": 1716454225638885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593645, "dur": 10,
    "args": {
      "External id": 260489,
      "cbid": 211, "correlation": 260489
    }
  },
  {
    "ph": "s", "id": 260489, "pid": 76337, "tid": -914061504, "ts": 1716454225593645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225638936, "dur": 160,
    "args": {
      "External id": 260500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260500, "pid": 5, "tid": 7, "ts": 1716454225638936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593711, "dur": 12,
    "args": {
      "External id": 260500,
      "cbid": 211, "correlation": 260500
    }
  },
  {
    "ph": "s", "id": 260500, "pid": 76337, "tid": -914061504, "ts": 1716454225593711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593776, "dur": 0,
    "args": {
      "External id": 260512,
      "cbid": 317, "correlation": 260512
    }
  },
  {
    "ph": "f", "id": 260512, "pid": 76337, "tid": -914061504, "ts": 1716454225593776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593777, "dur": 0,
    "args": {
      "External id": 260513,
      "cbid": 203, "correlation": 260513
    }
  },
  {
    "ph": "f", "id": 260513, "pid": 76337, "tid": -914061504, "ts": 1716454225593777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593778, "dur": 0,
    "args": {
      "External id": 260514,
      "cbid": 205, "correlation": 260514
    }
  },
  {
    "ph": "f", "id": 260514, "pid": 76337, "tid": -914061504, "ts": 1716454225593778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593800, "dur": 1,
    "args": {
      "External id": 260518,
      "cbid": 251, "correlation": 260518
    }
  },
  {
    "ph": "f", "id": 260518, "pid": 76337, "tid": -914061504, "ts": 1716454225593800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593802, "dur": 0,
    "args": {
      "External id": 260519,
      "cbid": 251, "correlation": 260519
    }
  },
  {
    "ph": "f", "id": 260519, "pid": 76337, "tid": -914061504, "ts": 1716454225593802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593803, "dur": 0,
    "args": {
      "External id": 260520,
      "cbid": 251, "correlation": 260520
    }
  },
  {
    "ph": "f", "id": 260520, "pid": 76337, "tid": -914061504, "ts": 1716454225593803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593803, "dur": 0,
    "args": {
      "External id": 260521,
      "cbid": 251, "correlation": 260521
    }
  },
  {
    "ph": "f", "id": 260521, "pid": 76337, "tid": -914061504, "ts": 1716454225593803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593804, "dur": 0,
    "args": {
      "External id": 260522,
      "cbid": 251, "correlation": 260522
    }
  },
  {
    "ph": "f", "id": 260522, "pid": 76337, "tid": -914061504, "ts": 1716454225593804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593805, "dur": 0,
    "args": {
      "External id": 260523,
      "cbid": 251, "correlation": 260523
    }
  },
  {
    "ph": "f", "id": 260523, "pid": 76337, "tid": -914061504, "ts": 1716454225593805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593805, "dur": 0,
    "args": {
      "External id": 260524,
      "cbid": 251, "correlation": 260524
    }
  },
  {
    "ph": "f", "id": 260524, "pid": 76337, "tid": -914061504, "ts": 1716454225593805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593806, "dur": 0,
    "args": {
      "External id": 260525,
      "cbid": 251, "correlation": 260525
    }
  },
  {
    "ph": "f", "id": 260525, "pid": 76337, "tid": -914061504, "ts": 1716454225593806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593807, "dur": 0,
    "args": {
      "External id": 260526,
      "cbid": 251, "correlation": 260526
    }
  },
  {
    "ph": "f", "id": 260526, "pid": 76337, "tid": -914061504, "ts": 1716454225593807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225639098, "dur": 116,
    "args": {
      "External id": 260527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260527,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260527, "pid": 5, "tid": 7, "ts": 1716454225639098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593809, "dur": 12,
    "args": {
      "External id": 260527,
      "cbid": 211, "correlation": 260527
    }
  },
  {
    "ph": "s", "id": 260527, "pid": 76337, "tid": -914061504, "ts": 1716454225593809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225639215, "dur": 60,
    "args": {
      "External id": 260533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260533, "pid": 5, "tid": 7, "ts": 1716454225639215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593844, "dur": 9,
    "args": {
      "External id": 260533,
      "cbid": 211, "correlation": 260533
    }
  },
  {
    "ph": "s", "id": 260533, "pid": 76337, "tid": -914061504, "ts": 1716454225593844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225639276, "dur": 49,
    "args": {
      "External id": 260541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260541,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260541, "pid": 5, "tid": 7, "ts": 1716454225639276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593876, "dur": 8,
    "args": {
      "External id": 260541,
      "cbid": 211, "correlation": 260541
    }
  },
  {
    "ph": "s", "id": 260541, "pid": 76337, "tid": -914061504, "ts": 1716454225593876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593949, "dur": 0,
    "args": {
      "External id": 260551,
      "cbid": 317, "correlation": 260551
    }
  },
  {
    "ph": "f", "id": 260551, "pid": 76337, "tid": -914061504, "ts": 1716454225593949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593950, "dur": 0,
    "args": {
      "External id": 260552,
      "cbid": 203, "correlation": 260552
    }
  },
  {
    "ph": "f", "id": 260552, "pid": 76337, "tid": -914061504, "ts": 1716454225593950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593950, "dur": 0,
    "args": {
      "External id": 260553,
      "cbid": 205, "correlation": 260553
    }
  },
  {
    "ph": "f", "id": 260553, "pid": 76337, "tid": -914061504, "ts": 1716454225593950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225639327, "dur": 41,
    "args": {
      "External id": 260557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260557,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260557, "pid": 5, "tid": 7, "ts": 1716454225639327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593967, "dur": 20,
    "args": {
      "External id": 260557,
      "cbid": 211, "correlation": 260557
    }
  },
  {
    "ph": "s", "id": 260557, "pid": 76337, "tid": -914061504, "ts": 1716454225593967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225639369, "dur": 14,
    "args": {
      "External id": 260559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260559, "pid": 5, "tid": 7, "ts": 1716454225639369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225593990, "dur": 5,
    "args": {
      "External id": 260559,
      "cbid": 211, "correlation": 260559
    }
  },
  {
    "ph": "s", "id": 260559, "pid": 76337, "tid": -914061504, "ts": 1716454225593990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225639386, "dur": 1,
    "args": {
      "External id": 260561,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 260561,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 260561, "pid": 5, "tid": 7, "ts": 1716454225639386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594009, "dur": 17,
    "args": {
      "External id": 260561,
      "cbid": 51, "correlation": 260561
    }
  },
  {
    "ph": "s", "id": 260561, "pid": 76337, "tid": -914061504, "ts": 1716454225594009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225639390, "dur": 369,
    "args": {
      "External id": 260562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260562,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260562, "pid": 5, "tid": 7, "ts": 1716454225639390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594028, "dur": 9,
    "args": {
      "External id": 260562,
      "cbid": 211, "correlation": 260562
    }
  },
  {
    "ph": "s", "id": 260562, "pid": 76337, "tid": -914061504, "ts": 1716454225594028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225639760, "dur": 15,
    "args": {
      "External id": 260564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260564,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260564, "pid": 5, "tid": 7, "ts": 1716454225639760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594045, "dur": 7,
    "args": {
      "External id": 260564,
      "cbid": 211, "correlation": 260564
    }
  },
  {
    "ph": "s", "id": 260564, "pid": 76337, "tid": -914061504, "ts": 1716454225594045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225639777, "dur": 15,
    "args": {
      "External id": 260570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260570, "pid": 5, "tid": 7, "ts": 1716454225639777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594076, "dur": 9,
    "args": {
      "External id": 260570,
      "cbid": 211, "correlation": 260570
    }
  },
  {
    "ph": "s", "id": 260570, "pid": 76337, "tid": -914061504, "ts": 1716454225594076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225639793, "dur": 19,
    "args": {
      "External id": 260590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260590,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 260590, "pid": 5, "tid": 7, "ts": 1716454225639793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594166, "dur": 13,
    "args": {
      "External id": 260590,
      "cbid": 211, "correlation": 260590
    }
  },
  {
    "ph": "s", "id": 260590, "pid": 76337, "tid": -914061504, "ts": 1716454225594166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225639813, "dur": 4,
    "args": {
      "External id": 260602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260602,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 260602, "pid": 5, "tid": 7, "ts": 1716454225639813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594190, "dur": 6,
    "args": {
      "External id": 260602,
      "cbid": 211, "correlation": 260602
    }
  },
  {
    "ph": "s", "id": 260602, "pid": 76337, "tid": -914061504, "ts": 1716454225594190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225639819, "dur": 18,
    "args": {
      "External id": 260605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260605,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260605, "pid": 5, "tid": 7, "ts": 1716454225639819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594207, "dur": 6,
    "args": {
      "External id": 260605,
      "cbid": 211, "correlation": 260605
    }
  },
  {
    "ph": "s", "id": 260605, "pid": 76337, "tid": -914061504, "ts": 1716454225594207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225639838, "dur": 12,
    "args": {
      "External id": 260614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260614,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260614, "pid": 5, "tid": 7, "ts": 1716454225639838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594247, "dur": 9,
    "args": {
      "External id": 260614,
      "cbid": 211, "correlation": 260614
    }
  },
  {
    "ph": "s", "id": 260614, "pid": 76337, "tid": -914061504, "ts": 1716454225594247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594301, "dur": 0,
    "args": {
      "External id": 260624,
      "cbid": 317, "correlation": 260624
    }
  },
  {
    "ph": "f", "id": 260624, "pid": 76337, "tid": -914061504, "ts": 1716454225594301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594302, "dur": 0,
    "args": {
      "External id": 260625,
      "cbid": 203, "correlation": 260625
    }
  },
  {
    "ph": "f", "id": 260625, "pid": 76337, "tid": -914061504, "ts": 1716454225594302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594302, "dur": 0,
    "args": {
      "External id": 260626,
      "cbid": 205, "correlation": 260626
    }
  },
  {
    "ph": "f", "id": 260626, "pid": 76337, "tid": -914061504, "ts": 1716454225594302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225639852, "dur": 11,
    "args": {
      "External id": 260630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260630,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260630, "pid": 5, "tid": 7, "ts": 1716454225639852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594317, "dur": 13,
    "args": {
      "External id": 260630,
      "cbid": 211, "correlation": 260630
    }
  },
  {
    "ph": "s", "id": 260630, "pid": 76337, "tid": -914061504, "ts": 1716454225594317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225639864, "dur": 24,
    "args": {
      "External id": 260632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260632, "pid": 5, "tid": 7, "ts": 1716454225639864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594332, "dur": 5,
    "args": {
      "External id": 260632,
      "cbid": 211, "correlation": 260632
    }
  },
  {
    "ph": "s", "id": 260632, "pid": 76337, "tid": -914061504, "ts": 1716454225594332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225639890, "dur": 4,
    "args": {
      "External id": 260634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 260634, "pid": 5, "tid": 7, "ts": 1716454225639890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594343, "dur": 6,
    "args": {
      "External id": 260634,
      "cbid": 211, "correlation": 260634
    }
  },
  {
    "ph": "s", "id": 260634, "pid": 76337, "tid": -914061504, "ts": 1716454225594343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594352, "dur": 0,
    "args": {
      "External id": 260635,
      "cbid": 51, "correlation": 260635
    }
  },
  {
    "ph": "s", "id": 260635, "pid": 76337, "tid": -914061504, "ts": 1716454225594352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225639895, "dur": 361,
    "args": {
      "External id": 260636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260636,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260636, "pid": 5, "tid": 7, "ts": 1716454225639895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594353, "dur": 8,
    "args": {
      "External id": 260636,
      "cbid": 211, "correlation": 260636
    }
  },
  {
    "ph": "s", "id": 260636, "pid": 76337, "tid": -914061504, "ts": 1716454225594353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225640258, "dur": 20,
    "args": {
      "External id": 260637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260637,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260637, "pid": 5, "tid": 7, "ts": 1716454225640258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594364, "dur": 5,
    "args": {
      "External id": 260637,
      "cbid": 211, "correlation": 260637
    }
  },
  {
    "ph": "s", "id": 260637, "pid": 76337, "tid": -914061504, "ts": 1716454225594364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225640279, "dur": 34,
    "args": {
      "External id": 260643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260643, "pid": 5, "tid": 7, "ts": 1716454225640279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594392, "dur": 9,
    "args": {
      "External id": 260643,
      "cbid": 211, "correlation": 260643
    }
  },
  {
    "ph": "s", "id": 260643, "pid": 76337, "tid": -914061504, "ts": 1716454225594392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225640314, "dur": 3,
    "args": {
      "External id": 260651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260651,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 260651, "pid": 5, "tid": 7, "ts": 1716454225640314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594435, "dur": 9,
    "args": {
      "External id": 260651,
      "cbid": 211, "correlation": 260651
    }
  },
  {
    "ph": "s", "id": 260651, "pid": 76337, "tid": -914061504, "ts": 1716454225594435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594502, "dur": 1,
    "args": {
      "External id": 260667,
      "cbid": 251, "correlation": 260667
    }
  },
  {
    "ph": "f", "id": 260667, "pid": 76337, "tid": -914061504, "ts": 1716454225594502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594507, "dur": 0,
    "args": {
      "External id": 260669,
      "cbid": 251, "correlation": 260669
    }
  },
  {
    "ph": "f", "id": 260669, "pid": 76337, "tid": -914061504, "ts": 1716454225594507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225640319, "dur": 13,
    "args": {
      "External id": 260670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260670,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 260670, "pid": 5, "tid": 7, "ts": 1716454225640319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594509, "dur": 11,
    "args": {
      "External id": 260670,
      "cbid": 211, "correlation": 260670
    }
  },
  {
    "ph": "s", "id": 260670, "pid": 76337, "tid": -914061504, "ts": 1716454225594509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225640333, "dur": 5,
    "args": {
      "External id": 260672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260672,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 260672, "pid": 5, "tid": 7, "ts": 1716454225640333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594523, "dur": 6,
    "args": {
      "External id": 260672,
      "cbid": 211, "correlation": 260672
    }
  },
  {
    "ph": "s", "id": 260672, "pid": 76337, "tid": -914061504, "ts": 1716454225594523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225640339, "dur": 31,
    "args": {
      "External id": 260682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260682, "pid": 5, "tid": 7, "ts": 1716454225640339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594582, "dur": 12,
    "args": {
      "External id": 260682,
      "cbid": 211, "correlation": 260682
    }
  },
  {
    "ph": "s", "id": 260682, "pid": 76337, "tid": -914061504, "ts": 1716454225594582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225640371, "dur": 30,
    "args": {
      "External id": 260702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260702,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 260702, "pid": 5, "tid": 7, "ts": 1716454225640371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594647, "dur": 10,
    "args": {
      "External id": 260702,
      "cbid": 211, "correlation": 260702
    }
  },
  {
    "ph": "s", "id": 260702, "pid": 76337, "tid": -914061504, "ts": 1716454225594647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225640403, "dur": 4,
    "args": {
      "External id": 260714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260714,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 260714, "pid": 5, "tid": 7, "ts": 1716454225640403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594667, "dur": 6,
    "args": {
      "External id": 260714,
      "cbid": 211, "correlation": 260714
    }
  },
  {
    "ph": "s", "id": 260714, "pid": 76337, "tid": -914061504, "ts": 1716454225594667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225640408, "dur": 30,
    "args": {
      "External id": 260717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260717, "pid": 5, "tid": 7, "ts": 1716454225640408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594687, "dur": 6,
    "args": {
      "External id": 260717,
      "cbid": 211, "correlation": 260717
    }
  },
  {
    "ph": "s", "id": 260717, "pid": 76337, "tid": -914061504, "ts": 1716454225594687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225640439, "dur": 23,
    "args": {
      "External id": 260726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260726,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260726, "pid": 5, "tid": 7, "ts": 1716454225640439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594727, "dur": 11,
    "args": {
      "External id": 260726,
      "cbid": 211, "correlation": 260726
    }
  },
  {
    "ph": "s", "id": 260726, "pid": 76337, "tid": -914061504, "ts": 1716454225594727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594792, "dur": 0,
    "args": {
      "External id": 260736,
      "cbid": 317, "correlation": 260736
    }
  },
  {
    "ph": "f", "id": 260736, "pid": 76337, "tid": -914061504, "ts": 1716454225594792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594793, "dur": 0,
    "args": {
      "External id": 260737,
      "cbid": 203, "correlation": 260737
    }
  },
  {
    "ph": "f", "id": 260737, "pid": 76337, "tid": -914061504, "ts": 1716454225594793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594793, "dur": 0,
    "args": {
      "External id": 260738,
      "cbid": 205, "correlation": 260738
    }
  },
  {
    "ph": "f", "id": 260738, "pid": 76337, "tid": -914061504, "ts": 1716454225594793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225640463, "dur": 22,
    "args": {
      "External id": 260742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260742, "pid": 5, "tid": 7, "ts": 1716454225640463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594811, "dur": 12,
    "args": {
      "External id": 260742,
      "cbid": 211, "correlation": 260742
    }
  },
  {
    "ph": "s", "id": 260742, "pid": 76337, "tid": -914061504, "ts": 1716454225594811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225640487, "dur": 44,
    "args": {
      "External id": 260744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260744, "pid": 5, "tid": 7, "ts": 1716454225640487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594825, "dur": 5,
    "args": {
      "External id": 260744,
      "cbid": 211, "correlation": 260744
    }
  },
  {
    "ph": "s", "id": 260744, "pid": 76337, "tid": -914061504, "ts": 1716454225594825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225640532, "dur": 666,
    "args": {
      "External id": 260746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260746,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260746, "pid": 5, "tid": 7, "ts": 1716454225640532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594840, "dur": 10,
    "args": {
      "External id": 260746,
      "cbid": 211, "correlation": 260746
    }
  },
  {
    "ph": "s", "id": 260746, "pid": 76337, "tid": -914061504, "ts": 1716454225594840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225641199, "dur": 21,
    "args": {
      "External id": 260748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260748,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260748, "pid": 5, "tid": 7, "ts": 1716454225641199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594854, "dur": 5,
    "args": {
      "External id": 260748,
      "cbid": 211, "correlation": 260748
    }
  },
  {
    "ph": "s", "id": 260748, "pid": 76337, "tid": -914061504, "ts": 1716454225594854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225641222, "dur": 33,
    "args": {
      "External id": 260754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260754, "pid": 5, "tid": 7, "ts": 1716454225641222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594882, "dur": 9,
    "args": {
      "External id": 260754,
      "cbid": 211, "correlation": 260754
    }
  },
  {
    "ph": "s", "id": 260754, "pid": 76337, "tid": -914061504, "ts": 1716454225594882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594941, "dur": 0,
    "args": {
      "External id": 260764,
      "cbid": 317, "correlation": 260764
    }
  },
  {
    "ph": "f", "id": 260764, "pid": 76337, "tid": -914061504, "ts": 1716454225594941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594942, "dur": 0,
    "args": {
      "External id": 260765,
      "cbid": 203, "correlation": 260765
    }
  },
  {
    "ph": "f", "id": 260765, "pid": 76337, "tid": -914061504, "ts": 1716454225594942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594943, "dur": 0,
    "args": {
      "External id": 260766,
      "cbid": 205, "correlation": 260766
    }
  },
  {
    "ph": "f", "id": 260766, "pid": 76337, "tid": -914061504, "ts": 1716454225594943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594965, "dur": 1,
    "args": {
      "External id": 260770,
      "cbid": 251, "correlation": 260770
    }
  },
  {
    "ph": "f", "id": 260770, "pid": 76337, "tid": -914061504, "ts": 1716454225594965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594967, "dur": 0,
    "args": {
      "External id": 260771,
      "cbid": 251, "correlation": 260771
    }
  },
  {
    "ph": "f", "id": 260771, "pid": 76337, "tid": -914061504, "ts": 1716454225594967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594967, "dur": 0,
    "args": {
      "External id": 260772,
      "cbid": 251, "correlation": 260772
    }
  },
  {
    "ph": "f", "id": 260772, "pid": 76337, "tid": -914061504, "ts": 1716454225594967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594968, "dur": 0,
    "args": {
      "External id": 260773,
      "cbid": 251, "correlation": 260773
    }
  },
  {
    "ph": "f", "id": 260773, "pid": 76337, "tid": -914061504, "ts": 1716454225594968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594969, "dur": 0,
    "args": {
      "External id": 260774,
      "cbid": 251, "correlation": 260774
    }
  },
  {
    "ph": "f", "id": 260774, "pid": 76337, "tid": -914061504, "ts": 1716454225594969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594969, "dur": 0,
    "args": {
      "External id": 260775,
      "cbid": 251, "correlation": 260775
    }
  },
  {
    "ph": "f", "id": 260775, "pid": 76337, "tid": -914061504, "ts": 1716454225594969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594970, "dur": 0,
    "args": {
      "External id": 260776,
      "cbid": 251, "correlation": 260776
    }
  },
  {
    "ph": "f", "id": 260776, "pid": 76337, "tid": -914061504, "ts": 1716454225594970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594971, "dur": 0,
    "args": {
      "External id": 260777,
      "cbid": 251, "correlation": 260777
    }
  },
  {
    "ph": "f", "id": 260777, "pid": 76337, "tid": -914061504, "ts": 1716454225594971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594972, "dur": 0,
    "args": {
      "External id": 260778,
      "cbid": 251, "correlation": 260778
    }
  },
  {
    "ph": "f", "id": 260778, "pid": 76337, "tid": -914061504, "ts": 1716454225594972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225641256, "dur": 52,
    "args": {
      "External id": 260779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260779,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260779, "pid": 5, "tid": 7, "ts": 1716454225641256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225594981, "dur": 14,
    "args": {
      "External id": 260779,
      "cbid": 211, "correlation": 260779
    }
  },
  {
    "ph": "s", "id": 260779, "pid": 76337, "tid": -914061504, "ts": 1716454225594981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225641310, "dur": 32,
    "args": {
      "External id": 260785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260785, "pid": 5, "tid": 7, "ts": 1716454225641310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595016, "dur": 8,
    "args": {
      "External id": 260785,
      "cbid": 211, "correlation": 260785
    }
  },
  {
    "ph": "s", "id": 260785, "pid": 76337, "tid": -914061504, "ts": 1716454225595016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225641343, "dur": 27,
    "args": {
      "External id": 260793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260793,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260793, "pid": 5, "tid": 7, "ts": 1716454225641343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595046, "dur": 7,
    "args": {
      "External id": 260793,
      "cbid": 211, "correlation": 260793
    }
  },
  {
    "ph": "s", "id": 260793, "pid": 76337, "tid": -914061504, "ts": 1716454225595046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225641372, "dur": 20,
    "args": {
      "External id": 260801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260801,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260801, "pid": 5, "tid": 7, "ts": 1716454225641372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595074, "dur": 8,
    "args": {
      "External id": 260801,
      "cbid": 211, "correlation": 260801
    }
  },
  {
    "ph": "s", "id": 260801, "pid": 76337, "tid": -914061504, "ts": 1716454225595074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225641393, "dur": 30,
    "args": {
      "External id": 260821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260821,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 260821, "pid": 5, "tid": 7, "ts": 1716454225641393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595156, "dur": 12,
    "args": {
      "External id": 260821,
      "cbid": 211, "correlation": 260821
    }
  },
  {
    "ph": "s", "id": 260821, "pid": 76337, "tid": -914061504, "ts": 1716454225595156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225641425, "dur": 4,
    "args": {
      "External id": 260833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260833,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 260833, "pid": 5, "tid": 7, "ts": 1716454225641425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595177, "dur": 6,
    "args": {
      "External id": 260833,
      "cbid": 211, "correlation": 260833
    }
  },
  {
    "ph": "s", "id": 260833, "pid": 76337, "tid": -914061504, "ts": 1716454225595177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225641430, "dur": 30,
    "args": {
      "External id": 260836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260836,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260836, "pid": 5, "tid": 7, "ts": 1716454225641430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595196, "dur": 7,
    "args": {
      "External id": 260836,
      "cbid": 211, "correlation": 260836
    }
  },
  {
    "ph": "s", "id": 260836, "pid": 76337, "tid": -914061504, "ts": 1716454225595196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595253, "dur": 0,
    "args": {
      "External id": 260847,
      "cbid": 317, "correlation": 260847
    }
  },
  {
    "ph": "f", "id": 260847, "pid": 76337, "tid": -914061504, "ts": 1716454225595253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595254, "dur": 0,
    "args": {
      "External id": 260848,
      "cbid": 203, "correlation": 260848
    }
  },
  {
    "ph": "f", "id": 260848, "pid": 76337, "tid": -914061504, "ts": 1716454225595254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595255, "dur": 0,
    "args": {
      "External id": 260849,
      "cbid": 205, "correlation": 260849
    }
  },
  {
    "ph": "f", "id": 260849, "pid": 76337, "tid": -914061504, "ts": 1716454225595255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225641461, "dur": 22,
    "args": {
      "External id": 260853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260853,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260853, "pid": 5, "tid": 7, "ts": 1716454225641461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595269, "dur": 11,
    "args": {
      "External id": 260853,
      "cbid": 211, "correlation": 260853
    }
  },
  {
    "ph": "s", "id": 260853, "pid": 76337, "tid": -914061504, "ts": 1716454225595269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225641484, "dur": 123,
    "args": {
      "External id": 260855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260855,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260855, "pid": 5, "tid": 7, "ts": 1716454225641484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595289, "dur": 9,
    "args": {
      "External id": 260855,
      "cbid": 211, "correlation": 260855
    }
  },
  {
    "ph": "s", "id": 260855, "pid": 76337, "tid": -914061504, "ts": 1716454225595289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225641608, "dur": 22,
    "args": {
      "External id": 260857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260857,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260857, "pid": 5, "tid": 7, "ts": 1716454225641608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595302, "dur": 5,
    "args": {
      "External id": 260857,
      "cbid": 211, "correlation": 260857
    }
  },
  {
    "ph": "s", "id": 260857, "pid": 76337, "tid": -914061504, "ts": 1716454225595302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225641632, "dur": 33,
    "args": {
      "External id": 260863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260863, "pid": 5, "tid": 7, "ts": 1716454225641632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595329, "dur": 9,
    "args": {
      "External id": 260863,
      "cbid": 211, "correlation": 260863
    }
  },
  {
    "ph": "s", "id": 260863, "pid": 76337, "tid": -914061504, "ts": 1716454225595329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225641666, "dur": 189,
    "args": {
      "External id": 260872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260872,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260872, "pid": 5, "tid": 7, "ts": 1716454225641666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595411, "dur": 14,
    "args": {
      "External id": 260872,
      "cbid": 211, "correlation": 260872
    }
  },
  {
    "ph": "s", "id": 260872, "pid": 76337, "tid": -914061504, "ts": 1716454225595411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225641857, "dur": 66,
    "args": {
      "External id": 260894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260894,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260894, "pid": 5, "tid": 7, "ts": 1716454225641857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595469, "dur": 10,
    "args": {
      "External id": 260894,
      "cbid": 211, "correlation": 260894
    }
  },
  {
    "ph": "s", "id": 260894, "pid": 76337, "tid": -914061504, "ts": 1716454225595469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595559, "dur": 1,
    "args": {
      "External id": 260905,
      "cbid": 251, "correlation": 260905
    }
  },
  {
    "ph": "f", "id": 260905, "pid": 76337, "tid": -914061504, "ts": 1716454225595559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225641924, "dur": 156,
    "args": {
      "External id": 260906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260906,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260906, "pid": 5, "tid": 7, "ts": 1716454225641924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595565, "dur": 13,
    "args": {
      "External id": 260906,
      "cbid": 211, "correlation": 260906
    }
  },
  {
    "ph": "s", "id": 260906, "pid": 76337, "tid": -914061504, "ts": 1716454225595565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595635, "dur": 1,
    "args": {
      "External id": 260917,
      "cbid": 251, "correlation": 260917
    }
  },
  {
    "ph": "f", "id": 260917, "pid": 76337, "tid": -914061504, "ts": 1716454225595635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225642082, "dur": 147,
    "args": {
      "External id": 260918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260918,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260918, "pid": 5, "tid": 7, "ts": 1716454225642082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595639, "dur": 11,
    "args": {
      "External id": 260918,
      "cbid": 211, "correlation": 260918
    }
  },
  {
    "ph": "s", "id": 260918, "pid": 76337, "tid": -914061504, "ts": 1716454225595639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595704, "dur": 1,
    "args": {
      "External id": 260929,
      "cbid": 251, "correlation": 260929
    }
  },
  {
    "ph": "f", "id": 260929, "pid": 76337, "tid": -914061504, "ts": 1716454225595704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225642230, "dur": 146,
    "args": {
      "External id": 260930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260930,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 260930, "pid": 5, "tid": 7, "ts": 1716454225642230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595708, "dur": 11,
    "args": {
      "External id": 260930,
      "cbid": 211, "correlation": 260930
    }
  },
  {
    "ph": "s", "id": 260930, "pid": 76337, "tid": -914061504, "ts": 1716454225595708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225642377, "dur": 1988,
    "args": {
      "External id": 260951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260951,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 260951, "pid": 5, "tid": 7, "ts": 1716454225642377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595793, "dur": 13,
    "args": {
      "External id": 260951,
      "cbid": 211, "correlation": 260951
    }
  },
  {
    "ph": "s", "id": 260951, "pid": 76337, "tid": -914061504, "ts": 1716454225595793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595895, "dur": 1,
    "args": {
      "External id": 260969,
      "cbid": 251, "correlation": 260969
    }
  },
  {
    "ph": "f", "id": 260969, "pid": 76337, "tid": -914061504, "ts": 1716454225595895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225644367, "dur": 148,
    "args": {
      "External id": 260971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260971,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 260971, "pid": 5, "tid": 7, "ts": 1716454225644367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595900, "dur": 13,
    "args": {
      "External id": 260971,
      "cbid": 211, "correlation": 260971
    }
  },
  {
    "ph": "s", "id": 260971, "pid": 76337, "tid": -914061504, "ts": 1716454225595900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225644516, "dur": 36,
    "args": {
      "External id": 260979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260979,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260979, "pid": 5, "tid": 7, "ts": 1716454225644516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225595970, "dur": 20,
    "args": {
      "External id": 260979,
      "cbid": 211, "correlation": 260979
    }
  },
  {
    "ph": "s", "id": 260979, "pid": 76337, "tid": -914061504, "ts": 1716454225595970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225644553, "dur": 50,
    "args": {
      "External id": 260987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260987,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260987, "pid": 5, "tid": 7, "ts": 1716454225644553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596018, "dur": 9,
    "args": {
      "External id": 260987,
      "cbid": 211, "correlation": 260987
    }
  },
  {
    "ph": "s", "id": 260987, "pid": 76337, "tid": -914061504, "ts": 1716454225596018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225644605, "dur": 31,
    "args": {
      "External id": 260998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 260998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 260998, "pid": 5, "tid": 7, "ts": 1716454225644605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596092, "dur": 13,
    "args": {
      "External id": 260998,
      "cbid": 211, "correlation": 260998
    }
  },
  {
    "ph": "s", "id": 260998, "pid": 76337, "tid": -914061504, "ts": 1716454225596092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225644637, "dur": 35,
    "args": {
      "External id": 261020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261020,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261020, "pid": 5, "tid": 7, "ts": 1716454225644637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596123, "dur": 7,
    "args": {
      "External id": 261020,
      "cbid": 211, "correlation": 261020
    }
  },
  {
    "ph": "s", "id": 261020, "pid": 76337, "tid": -914061504, "ts": 1716454225596123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596209, "dur": 1,
    "args": {
      "External id": 261031,
      "cbid": 251, "correlation": 261031
    }
  },
  {
    "ph": "f", "id": 261031, "pid": 76337, "tid": -914061504, "ts": 1716454225596209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225644674, "dur": 91,
    "args": {
      "External id": 261032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261032,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261032, "pid": 5, "tid": 7, "ts": 1716454225644674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596214, "dur": 13,
    "args": {
      "External id": 261032,
      "cbid": 211, "correlation": 261032
    }
  },
  {
    "ph": "s", "id": 261032, "pid": 76337, "tid": -914061504, "ts": 1716454225596214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596285, "dur": 1,
    "args": {
      "External id": 261043,
      "cbid": 251, "correlation": 261043
    }
  },
  {
    "ph": "f", "id": 261043, "pid": 76337, "tid": -914061504, "ts": 1716454225596285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596289, "dur": 0,
    "args": {
      "External id": 261044,
      "cbid": 251, "correlation": 261044
    }
  },
  {
    "ph": "f", "id": 261044, "pid": 76337, "tid": -914061504, "ts": 1716454225596289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225644766, "dur": 12,
    "args": {
      "External id": 261045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261045,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 261045, "pid": 5, "tid": 7, "ts": 1716454225644766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596291, "dur": 12,
    "args": {
      "External id": 261045,
      "cbid": 211, "correlation": 261045
    }
  },
  {
    "ph": "s", "id": 261045, "pid": 76337, "tid": -914061504, "ts": 1716454225596291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225644779, "dur": 5,
    "args": {
      "External id": 261047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261047,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 261047, "pid": 5, "tid": 7, "ts": 1716454225644779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596308, "dur": 7,
    "args": {
      "External id": 261047,
      "cbid": 211, "correlation": 261047
    }
  },
  {
    "ph": "s", "id": 261047, "pid": 76337, "tid": -914061504, "ts": 1716454225596308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596366, "dur": 1,
    "args": {
      "External id": 261058,
      "cbid": 251, "correlation": 261058
    }
  },
  {
    "ph": "f", "id": 261058, "pid": 76337, "tid": -914061504, "ts": 1716454225596366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596370, "dur": 0,
    "args": {
      "External id": 261059,
      "cbid": 251, "correlation": 261059
    }
  },
  {
    "ph": "f", "id": 261059, "pid": 76337, "tid": -914061504, "ts": 1716454225596370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225644786, "dur": 7,
    "args": {
      "External id": 261060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261060,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 261060, "pid": 5, "tid": 7, "ts": 1716454225644786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596371, "dur": 11,
    "args": {
      "External id": 261060,
      "cbid": 211, "correlation": 261060
    }
  },
  {
    "ph": "s", "id": 261060, "pid": 76337, "tid": -914061504, "ts": 1716454225596371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225644794, "dur": 3,
    "args": {
      "External id": 261062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261062,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 261062, "pid": 5, "tid": 7, "ts": 1716454225644794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596384, "dur": 6,
    "args": {
      "External id": 261062,
      "cbid": 211, "correlation": 261062
    }
  },
  {
    "ph": "s", "id": 261062, "pid": 76337, "tid": -914061504, "ts": 1716454225596384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225644799, "dur": 93,
    "args": {
      "External id": 261083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261083,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 261083, "pid": 5, "tid": 7, "ts": 1716454225644799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596458, "dur": 12,
    "args": {
      "External id": 261083,
      "cbid": 211, "correlation": 261083
    }
  },
  {
    "ph": "s", "id": 261083, "pid": 76337, "tid": -914061504, "ts": 1716454225596458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596554, "dur": 1,
    "args": {
      "External id": 261101,
      "cbid": 251, "correlation": 261101
    }
  },
  {
    "ph": "f", "id": 261101, "pid": 76337, "tid": -914061504, "ts": 1716454225596554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225644893, "dur": 100,
    "args": {
      "External id": 261103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261103,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261103, "pid": 5, "tid": 7, "ts": 1716454225644893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596560, "dur": 13,
    "args": {
      "External id": 261103,
      "cbid": 211, "correlation": 261103
    }
  },
  {
    "ph": "s", "id": 261103, "pid": 76337, "tid": -914061504, "ts": 1716454225596560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225644995, "dur": 19,
    "args": {
      "External id": 261111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261111,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261111, "pid": 5, "tid": 7, "ts": 1716454225644995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596629, "dur": 12,
    "args": {
      "External id": 261111,
      "cbid": 211, "correlation": 261111
    }
  },
  {
    "ph": "s", "id": 261111, "pid": 76337, "tid": -914061504, "ts": 1716454225596629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225645015, "dur": 37,
    "args": {
      "External id": 261119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261119, "pid": 5, "tid": 7, "ts": 1716454225645015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596671, "dur": 9,
    "args": {
      "External id": 261119,
      "cbid": 211, "correlation": 261119
    }
  },
  {
    "ph": "s", "id": 261119, "pid": 76337, "tid": -914061504, "ts": 1716454225596671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225645053, "dur": 35,
    "args": {
      "External id": 261141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261141,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261141, "pid": 5, "tid": 7, "ts": 1716454225645053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596723, "dur": 10,
    "args": {
      "External id": 261141,
      "cbid": 211, "correlation": 261141
    }
  },
  {
    "ph": "s", "id": 261141, "pid": 76337, "tid": -914061504, "ts": 1716454225596723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596812, "dur": 1,
    "args": {
      "External id": 261157,
      "cbid": 251, "correlation": 261157
    }
  },
  {
    "ph": "f", "id": 261157, "pid": 76337, "tid": -914061504, "ts": 1716454225596812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596817, "dur": 0,
    "args": {
      "External id": 261159,
      "cbid": 251, "correlation": 261159
    }
  },
  {
    "ph": "f", "id": 261159, "pid": 76337, "tid": -914061504, "ts": 1716454225596817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225645090, "dur": 549,
    "args": {
      "External id": 261160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261160,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 261160, "pid": 5, "tid": 7, "ts": 1716454225645090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596820, "dur": 12,
    "args": {
      "External id": 261160,
      "cbid": 211, "correlation": 261160
    }
  },
  {
    "ph": "s", "id": 261160, "pid": 76337, "tid": -914061504, "ts": 1716454225596820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225645641, "dur": 127,
    "args": {
      "External id": 261168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261168, "pid": 5, "tid": 7, "ts": 1716454225645641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596885, "dur": 13,
    "args": {
      "External id": 261168,
      "cbid": 211, "correlation": 261168
    }
  },
  {
    "ph": "s", "id": 261168, "pid": 76337, "tid": -914061504, "ts": 1716454225596885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225645769, "dur": 129,
    "args": {
      "External id": 261176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261176, "pid": 5, "tid": 7, "ts": 1716454225645769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225596917, "dur": 8,
    "args": {
      "External id": 261176,
      "cbid": 211, "correlation": 261176
    }
  },
  {
    "ph": "s", "id": 261176, "pid": 76337, "tid": -914061504, "ts": 1716454225596917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597004, "dur": 1,
    "args": {
      "External id": 261192,
      "cbid": 251, "correlation": 261192
    }
  },
  {
    "ph": "f", "id": 261192, "pid": 76337, "tid": -914061504, "ts": 1716454225597004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225645899, "dur": 311,
    "args": {
      "External id": 261194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261194,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261194, "pid": 5, "tid": 7, "ts": 1716454225645899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597009, "dur": 13,
    "args": {
      "External id": 261194,
      "cbid": 211, "correlation": 261194
    }
  },
  {
    "ph": "s", "id": 261194, "pid": 76337, "tid": -914061504, "ts": 1716454225597009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225646212, "dur": 27,
    "args": {
      "External id": 261202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261202,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261202, "pid": 5, "tid": 7, "ts": 1716454225646212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597054, "dur": 9,
    "args": {
      "External id": 261202,
      "cbid": 211, "correlation": 261202
    }
  },
  {
    "ph": "s", "id": 261202, "pid": 76337, "tid": -914061504, "ts": 1716454225597054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225646240, "dur": 83,
    "args": {
      "External id": 261213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261213, "pid": 5, "tid": 7, "ts": 1716454225646240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597121, "dur": 12,
    "args": {
      "External id": 261213,
      "cbid": 211, "correlation": 261213
    }
  },
  {
    "ph": "s", "id": 261213, "pid": 76337, "tid": -914061504, "ts": 1716454225597121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597185, "dur": 0,
    "args": {
      "External id": 261225,
      "cbid": 317, "correlation": 261225
    }
  },
  {
    "ph": "f", "id": 261225, "pid": 76337, "tid": -914061504, "ts": 1716454225597185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597186, "dur": 0,
    "args": {
      "External id": 261226,
      "cbid": 203, "correlation": 261226
    }
  },
  {
    "ph": "f", "id": 261226, "pid": 76337, "tid": -914061504, "ts": 1716454225597186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597187, "dur": 0,
    "args": {
      "External id": 261227,
      "cbid": 205, "correlation": 261227
    }
  },
  {
    "ph": "f", "id": 261227, "pid": 76337, "tid": -914061504, "ts": 1716454225597187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225646325, "dur": 22,
    "args": {
      "External id": 261231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261231, "pid": 5, "tid": 7, "ts": 1716454225646325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597203, "dur": 12,
    "args": {
      "External id": 261231,
      "cbid": 211, "correlation": 261231
    }
  },
  {
    "ph": "s", "id": 261231, "pid": 76337, "tid": -914061504, "ts": 1716454225597203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225646349, "dur": 123,
    "args": {
      "External id": 261233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261233,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261233, "pid": 5, "tid": 7, "ts": 1716454225646349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597221, "dur": 6,
    "args": {
      "External id": 261233,
      "cbid": 211, "correlation": 261233
    }
  },
  {
    "ph": "s", "id": 261233, "pid": 76337, "tid": -914061504, "ts": 1716454225597221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225646473, "dur": 22,
    "args": {
      "External id": 261235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261235, "pid": 5, "tid": 7, "ts": 1716454225646473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597231, "dur": 5,
    "args": {
      "External id": 261235,
      "cbid": 211, "correlation": 261235
    }
  },
  {
    "ph": "s", "id": 261235, "pid": 76337, "tid": -914061504, "ts": 1716454225597231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225646496, "dur": 33,
    "args": {
      "External id": 261241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261241, "pid": 5, "tid": 7, "ts": 1716454225646496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597260, "dur": 8,
    "args": {
      "External id": 261241,
      "cbid": 211, "correlation": 261241
    }
  },
  {
    "ph": "s", "id": 261241, "pid": 76337, "tid": -914061504, "ts": 1716454225597260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225646530, "dur": 27,
    "args": {
      "External id": 261249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261249,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261249, "pid": 5, "tid": 7, "ts": 1716454225646530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597290, "dur": 9,
    "args": {
      "External id": 261249,
      "cbid": 211, "correlation": 261249
    }
  },
  {
    "ph": "s", "id": 261249, "pid": 76337, "tid": -914061504, "ts": 1716454225597290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225646558, "dur": 31,
    "args": {
      "External id": 261269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261269,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 261269, "pid": 5, "tid": 7, "ts": 1716454225646558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597363, "dur": 11,
    "args": {
      "External id": 261269,
      "cbid": 211, "correlation": 261269
    }
  },
  {
    "ph": "s", "id": 261269, "pid": 76337, "tid": -914061504, "ts": 1716454225597363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225646591, "dur": 5,
    "args": {
      "External id": 261281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261281,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 261281, "pid": 5, "tid": 7, "ts": 1716454225646591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597384, "dur": 6,
    "args": {
      "External id": 261281,
      "cbid": 211, "correlation": 261281
    }
  },
  {
    "ph": "s", "id": 261281, "pid": 76337, "tid": -914061504, "ts": 1716454225597384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225646597, "dur": 31,
    "args": {
      "External id": 261284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261284, "pid": 5, "tid": 7, "ts": 1716454225646597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597402, "dur": 6,
    "args": {
      "External id": 261284,
      "cbid": 211, "correlation": 261284
    }
  },
  {
    "ph": "s", "id": 261284, "pid": 76337, "tid": -914061504, "ts": 1716454225597402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225646629, "dur": 21,
    "args": {
      "External id": 261293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261293,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261293, "pid": 5, "tid": 7, "ts": 1716454225646629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597440, "dur": 10,
    "args": {
      "External id": 261293,
      "cbid": 211, "correlation": 261293
    }
  },
  {
    "ph": "s", "id": 261293, "pid": 76337, "tid": -914061504, "ts": 1716454225597440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597491, "dur": 0,
    "args": {
      "External id": 261303,
      "cbid": 317, "correlation": 261303
    }
  },
  {
    "ph": "f", "id": 261303, "pid": 76337, "tid": -914061504, "ts": 1716454225597491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597492, "dur": 0,
    "args": {
      "External id": 261304,
      "cbid": 203, "correlation": 261304
    }
  },
  {
    "ph": "f", "id": 261304, "pid": 76337, "tid": -914061504, "ts": 1716454225597492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597492, "dur": 0,
    "args": {
      "External id": 261305,
      "cbid": 205, "correlation": 261305
    }
  },
  {
    "ph": "f", "id": 261305, "pid": 76337, "tid": -914061504, "ts": 1716454225597492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225646652, "dur": 22,
    "args": {
      "External id": 261309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261309, "pid": 5, "tid": 7, "ts": 1716454225646652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597506, "dur": 11,
    "args": {
      "External id": 261309,
      "cbid": 211, "correlation": 261309
    }
  },
  {
    "ph": "s", "id": 261309, "pid": 76337, "tid": -914061504, "ts": 1716454225597506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225646674, "dur": 44,
    "args": {
      "External id": 261311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261311,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261311, "pid": 5, "tid": 7, "ts": 1716454225646674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597520, "dur": 5,
    "args": {
      "External id": 261311,
      "cbid": 211, "correlation": 261311
    }
  },
  {
    "ph": "s", "id": 261311, "pid": 76337, "tid": -914061504, "ts": 1716454225597520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225646720, "dur": 663,
    "args": {
      "External id": 261313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261313,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261313, "pid": 5, "tid": 7, "ts": 1716454225646720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597531, "dur": 6,
    "args": {
      "External id": 261313,
      "cbid": 211, "correlation": 261313
    }
  },
  {
    "ph": "s", "id": 261313, "pid": 76337, "tid": -914061504, "ts": 1716454225597531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225647385, "dur": 22,
    "args": {
      "External id": 261315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261315,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261315, "pid": 5, "tid": 7, "ts": 1716454225647385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597541, "dur": 5,
    "args": {
      "External id": 261315,
      "cbid": 211, "correlation": 261315
    }
  },
  {
    "ph": "s", "id": 261315, "pid": 76337, "tid": -914061504, "ts": 1716454225597541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225647408, "dur": 34,
    "args": {
      "External id": 261321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261321, "pid": 5, "tid": 7, "ts": 1716454225647408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597568, "dur": 8,
    "args": {
      "External id": 261321,
      "cbid": 211, "correlation": 261321
    }
  },
  {
    "ph": "s", "id": 261321, "pid": 76337, "tid": -914061504, "ts": 1716454225597568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225647444, "dur": 3,
    "args": {
      "External id": 261329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261329,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 261329, "pid": 5, "tid": 7, "ts": 1716454225647444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597611, "dur": 9,
    "args": {
      "External id": 261329,
      "cbid": 211, "correlation": 261329
    }
  },
  {
    "ph": "s", "id": 261329, "pid": 76337, "tid": -914061504, "ts": 1716454225597611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597675, "dur": 1,
    "args": {
      "External id": 261345,
      "cbid": 251, "correlation": 261345
    }
  },
  {
    "ph": "f", "id": 261345, "pid": 76337, "tid": -914061504, "ts": 1716454225597675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597680, "dur": 0,
    "args": {
      "External id": 261347,
      "cbid": 251, "correlation": 261347
    }
  },
  {
    "ph": "f", "id": 261347, "pid": 76337, "tid": -914061504, "ts": 1716454225597680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225647448, "dur": 13,
    "args": {
      "External id": 261348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261348,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 261348, "pid": 5, "tid": 7, "ts": 1716454225647448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597682, "dur": 11,
    "args": {
      "External id": 261348,
      "cbid": 211, "correlation": 261348
    }
  },
  {
    "ph": "s", "id": 261348, "pid": 76337, "tid": -914061504, "ts": 1716454225597682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225647462, "dur": 5,
    "args": {
      "External id": 261350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261350,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 261350, "pid": 5, "tid": 7, "ts": 1716454225647462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597695, "dur": 6,
    "args": {
      "External id": 261350,
      "cbid": 211, "correlation": 261350
    }
  },
  {
    "ph": "s", "id": 261350, "pid": 76337, "tid": -914061504, "ts": 1716454225597695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225647469, "dur": 30,
    "args": {
      "External id": 261360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261360, "pid": 5, "tid": 7, "ts": 1716454225647469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597753, "dur": 12,
    "args": {
      "External id": 261360,
      "cbid": 211, "correlation": 261360
    }
  },
  {
    "ph": "s", "id": 261360, "pid": 76337, "tid": -914061504, "ts": 1716454225597753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225647500, "dur": 32,
    "args": {
      "External id": 261380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261380,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 261380, "pid": 5, "tid": 7, "ts": 1716454225647500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597818, "dur": 11,
    "args": {
      "External id": 261380,
      "cbid": 211, "correlation": 261380
    }
  },
  {
    "ph": "s", "id": 261380, "pid": 76337, "tid": -914061504, "ts": 1716454225597818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225647534, "dur": 4,
    "args": {
      "External id": 261392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261392,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 261392, "pid": 5, "tid": 7, "ts": 1716454225647534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597839, "dur": 6,
    "args": {
      "External id": 261392,
      "cbid": 211, "correlation": 261392
    }
  },
  {
    "ph": "s", "id": 261392, "pid": 76337, "tid": -914061504, "ts": 1716454225597839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225647539, "dur": 30,
    "args": {
      "External id": 261395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261395, "pid": 5, "tid": 7, "ts": 1716454225647539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597858, "dur": 6,
    "args": {
      "External id": 261395,
      "cbid": 211, "correlation": 261395
    }
  },
  {
    "ph": "s", "id": 261395, "pid": 76337, "tid": -914061504, "ts": 1716454225597858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225647571, "dur": 21,
    "args": {
      "External id": 261404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261404,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261404, "pid": 5, "tid": 7, "ts": 1716454225647571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597898, "dur": 11,
    "args": {
      "External id": 261404,
      "cbid": 211, "correlation": 261404
    }
  },
  {
    "ph": "s", "id": 261404, "pid": 76337, "tid": -914061504, "ts": 1716454225597898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597960, "dur": 0,
    "args": {
      "External id": 261414,
      "cbid": 317, "correlation": 261414
    }
  },
  {
    "ph": "f", "id": 261414, "pid": 76337, "tid": -914061504, "ts": 1716454225597960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597961, "dur": 0,
    "args": {
      "External id": 261415,
      "cbid": 203, "correlation": 261415
    }
  },
  {
    "ph": "f", "id": 261415, "pid": 76337, "tid": -914061504, "ts": 1716454225597961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597962, "dur": 0,
    "args": {
      "External id": 261416,
      "cbid": 205, "correlation": 261416
    }
  },
  {
    "ph": "f", "id": 261416, "pid": 76337, "tid": -914061504, "ts": 1716454225597962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225647593, "dur": 24,
    "args": {
      "External id": 261420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261420, "pid": 5, "tid": 7, "ts": 1716454225647593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597983, "dur": 12,
    "args": {
      "External id": 261420,
      "cbid": 211, "correlation": 261420
    }
  },
  {
    "ph": "s", "id": 261420, "pid": 76337, "tid": -914061504, "ts": 1716454225597983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225647619, "dur": 44,
    "args": {
      "External id": 261422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261422, "pid": 5, "tid": 7, "ts": 1716454225647619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225597998, "dur": 5,
    "args": {
      "External id": 261422,
      "cbid": 211, "correlation": 261422
    }
  },
  {
    "ph": "s", "id": 261422, "pid": 76337, "tid": -914061504, "ts": 1716454225597998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225647664, "dur": 656,
    "args": {
      "External id": 261424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261424,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261424, "pid": 5, "tid": 7, "ts": 1716454225647664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598010, "dur": 6,
    "args": {
      "External id": 261424,
      "cbid": 211, "correlation": 261424
    }
  },
  {
    "ph": "s", "id": 261424, "pid": 76337, "tid": -914061504, "ts": 1716454225598010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225648322, "dur": 22,
    "args": {
      "External id": 261426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261426,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261426, "pid": 5, "tid": 7, "ts": 1716454225648322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598019, "dur": 5,
    "args": {
      "External id": 261426,
      "cbid": 211, "correlation": 261426
    }
  },
  {
    "ph": "s", "id": 261426, "pid": 76337, "tid": -914061504, "ts": 1716454225598019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225648346, "dur": 34,
    "args": {
      "External id": 261432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261432, "pid": 5, "tid": 7, "ts": 1716454225648346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598047, "dur": 9,
    "args": {
      "External id": 261432,
      "cbid": 211, "correlation": 261432
    }
  },
  {
    "ph": "s", "id": 261432, "pid": 76337, "tid": -914061504, "ts": 1716454225598047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225648381, "dur": 26,
    "args": {
      "External id": 261440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261440,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261440, "pid": 5, "tid": 7, "ts": 1716454225648381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598080, "dur": 8,
    "args": {
      "External id": 261440,
      "cbid": 211, "correlation": 261440
    }
  },
  {
    "ph": "s", "id": 261440, "pid": 76337, "tid": -914061504, "ts": 1716454225598080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225648409, "dur": 20,
    "args": {
      "External id": 261448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261448,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261448, "pid": 5, "tid": 7, "ts": 1716454225648409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598110, "dur": 8,
    "args": {
      "External id": 261448,
      "cbid": 211, "correlation": 261448
    }
  },
  {
    "ph": "s", "id": 261448, "pid": 76337, "tid": -914061504, "ts": 1716454225598110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225648430, "dur": 29,
    "args": {
      "External id": 261468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261468,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 261468, "pid": 5, "tid": 7, "ts": 1716454225648430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598187, "dur": 12,
    "args": {
      "External id": 261468,
      "cbid": 211, "correlation": 261468
    }
  },
  {
    "ph": "s", "id": 261468, "pid": 76337, "tid": -914061504, "ts": 1716454225598187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225648461, "dur": 4,
    "args": {
      "External id": 261480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261480,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 261480, "pid": 5, "tid": 7, "ts": 1716454225648461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598209, "dur": 6,
    "args": {
      "External id": 261480,
      "cbid": 211, "correlation": 261480
    }
  },
  {
    "ph": "s", "id": 261480, "pid": 76337, "tid": -914061504, "ts": 1716454225598209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225648466, "dur": 30,
    "args": {
      "External id": 261483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261483, "pid": 5, "tid": 7, "ts": 1716454225648466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598227, "dur": 7,
    "args": {
      "External id": 261483,
      "cbid": 211, "correlation": 261483
    }
  },
  {
    "ph": "s", "id": 261483, "pid": 76337, "tid": -914061504, "ts": 1716454225598227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598285, "dur": 0,
    "args": {
      "External id": 261494,
      "cbid": 317, "correlation": 261494
    }
  },
  {
    "ph": "f", "id": 261494, "pid": 76337, "tid": -914061504, "ts": 1716454225598285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598286, "dur": 0,
    "args": {
      "External id": 261495,
      "cbid": 203, "correlation": 261495
    }
  },
  {
    "ph": "f", "id": 261495, "pid": 76337, "tid": -914061504, "ts": 1716454225598286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598287, "dur": 0,
    "args": {
      "External id": 261496,
      "cbid": 205, "correlation": 261496
    }
  },
  {
    "ph": "f", "id": 261496, "pid": 76337, "tid": -914061504, "ts": 1716454225598287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225648497, "dur": 22,
    "args": {
      "External id": 261500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261500,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261500, "pid": 5, "tid": 7, "ts": 1716454225648497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598300, "dur": 12,
    "args": {
      "External id": 261500,
      "cbid": 211, "correlation": 261500
    }
  },
  {
    "ph": "s", "id": 261500, "pid": 76337, "tid": -914061504, "ts": 1716454225598300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225648521, "dur": 117,
    "args": {
      "External id": 261502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261502,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261502, "pid": 5, "tid": 7, "ts": 1716454225648521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598319, "dur": 6,
    "args": {
      "External id": 261502,
      "cbid": 211, "correlation": 261502
    }
  },
  {
    "ph": "s", "id": 261502, "pid": 76337, "tid": -914061504, "ts": 1716454225598319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225648640, "dur": 22,
    "args": {
      "External id": 261504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261504,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261504, "pid": 5, "tid": 7, "ts": 1716454225648640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598328, "dur": 5,
    "args": {
      "External id": 261504,
      "cbid": 211, "correlation": 261504
    }
  },
  {
    "ph": "s", "id": 261504, "pid": 76337, "tid": -914061504, "ts": 1716454225598328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225648664, "dur": 33,
    "args": {
      "External id": 261510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261510, "pid": 5, "tid": 7, "ts": 1716454225648664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598356, "dur": 9,
    "args": {
      "External id": 261510,
      "cbid": 211, "correlation": 261510
    }
  },
  {
    "ph": "s", "id": 261510, "pid": 76337, "tid": -914061504, "ts": 1716454225598356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225648698, "dur": 189,
    "args": {
      "External id": 261519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261519,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261519, "pid": 5, "tid": 7, "ts": 1716454225648698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598437, "dur": 14,
    "args": {
      "External id": 261519,
      "cbid": 211, "correlation": 261519
    }
  },
  {
    "ph": "s", "id": 261519, "pid": 76337, "tid": -914061504, "ts": 1716454225598437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225648888, "dur": 66,
    "args": {
      "External id": 261541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261541,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261541, "pid": 5, "tid": 7, "ts": 1716454225648888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598493, "dur": 10,
    "args": {
      "External id": 261541,
      "cbid": 211, "correlation": 261541
    }
  },
  {
    "ph": "s", "id": 261541, "pid": 76337, "tid": -914061504, "ts": 1716454225598493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598580, "dur": 2,
    "args": {
      "External id": 261552,
      "cbid": 251, "correlation": 261552
    }
  },
  {
    "ph": "f", "id": 261552, "pid": 76337, "tid": -914061504, "ts": 1716454225598580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225648955, "dur": 155,
    "args": {
      "External id": 261553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261553,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261553, "pid": 5, "tid": 7, "ts": 1716454225648955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598585, "dur": 13,
    "args": {
      "External id": 261553,
      "cbid": 211, "correlation": 261553
    }
  },
  {
    "ph": "s", "id": 261553, "pid": 76337, "tid": -914061504, "ts": 1716454225598585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598656, "dur": 1,
    "args": {
      "External id": 261564,
      "cbid": 251, "correlation": 261564
    }
  },
  {
    "ph": "f", "id": 261564, "pid": 76337, "tid": -914061504, "ts": 1716454225598656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225649112, "dur": 146,
    "args": {
      "External id": 261565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261565,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261565, "pid": 5, "tid": 7, "ts": 1716454225649112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598660, "dur": 12,
    "args": {
      "External id": 261565,
      "cbid": 211, "correlation": 261565
    }
  },
  {
    "ph": "s", "id": 261565, "pid": 76337, "tid": -914061504, "ts": 1716454225598660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598724, "dur": 1,
    "args": {
      "External id": 261576,
      "cbid": 251, "correlation": 261576
    }
  },
  {
    "ph": "f", "id": 261576, "pid": 76337, "tid": -914061504, "ts": 1716454225598724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225649259, "dur": 147,
    "args": {
      "External id": 261577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261577,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261577, "pid": 5, "tid": 7, "ts": 1716454225649259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598729, "dur": 11,
    "args": {
      "External id": 261577,
      "cbid": 211, "correlation": 261577
    }
  },
  {
    "ph": "s", "id": 261577, "pid": 76337, "tid": -914061504, "ts": 1716454225598729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225649407, "dur": 1987,
    "args": {
      "External id": 261598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261598,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 261598, "pid": 5, "tid": 7, "ts": 1716454225649407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598808, "dur": 13,
    "args": {
      "External id": 261598,
      "cbid": 211, "correlation": 261598
    }
  },
  {
    "ph": "s", "id": 261598, "pid": 76337, "tid": -914061504, "ts": 1716454225598808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598906, "dur": 1,
    "args": {
      "External id": 261616,
      "cbid": 251, "correlation": 261616
    }
  },
  {
    "ph": "f", "id": 261616, "pid": 76337, "tid": -914061504, "ts": 1716454225598906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225651396, "dur": 150,
    "args": {
      "External id": 261618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261618,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 261618, "pid": 5, "tid": 7, "ts": 1716454225651396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598912, "dur": 14,
    "args": {
      "External id": 261618,
      "cbid": 211, "correlation": 261618
    }
  },
  {
    "ph": "s", "id": 261618, "pid": 76337, "tid": -914061504, "ts": 1716454225598912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225651547, "dur": 36,
    "args": {
      "External id": 261626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261626,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261626, "pid": 5, "tid": 7, "ts": 1716454225651547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225598991, "dur": 12,
    "args": {
      "External id": 261626,
      "cbid": 211, "correlation": 261626
    }
  },
  {
    "ph": "s", "id": 261626, "pid": 76337, "tid": -914061504, "ts": 1716454225598991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225651584, "dur": 50,
    "args": {
      "External id": 261634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261634,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261634, "pid": 5, "tid": 7, "ts": 1716454225651584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599031, "dur": 8,
    "args": {
      "External id": 261634,
      "cbid": 211, "correlation": 261634
    }
  },
  {
    "ph": "s", "id": 261634, "pid": 76337, "tid": -914061504, "ts": 1716454225599031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225651635, "dur": 32,
    "args": {
      "External id": 261645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261645,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261645, "pid": 5, "tid": 7, "ts": 1716454225651635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599102, "dur": 13,
    "args": {
      "External id": 261645,
      "cbid": 211, "correlation": 261645
    }
  },
  {
    "ph": "s", "id": 261645, "pid": 76337, "tid": -914061504, "ts": 1716454225599102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225651668, "dur": 35,
    "args": {
      "External id": 261667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261667,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261667, "pid": 5, "tid": 7, "ts": 1716454225651668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599133, "dur": 7,
    "args": {
      "External id": 261667,
      "cbid": 211, "correlation": 261667
    }
  },
  {
    "ph": "s", "id": 261667, "pid": 76337, "tid": -914061504, "ts": 1716454225599133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599217, "dur": 1,
    "args": {
      "External id": 261678,
      "cbid": 251, "correlation": 261678
    }
  },
  {
    "ph": "f", "id": 261678, "pid": 76337, "tid": -914061504, "ts": 1716454225599217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225651704, "dur": 91,
    "args": {
      "External id": 261679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261679,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261679, "pid": 5, "tid": 7, "ts": 1716454225651704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599223, "dur": 13,
    "args": {
      "External id": 261679,
      "cbid": 211, "correlation": 261679
    }
  },
  {
    "ph": "s", "id": 261679, "pid": 76337, "tid": -914061504, "ts": 1716454225599223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599292, "dur": 1,
    "args": {
      "External id": 261690,
      "cbid": 251, "correlation": 261690
    }
  },
  {
    "ph": "f", "id": 261690, "pid": 76337, "tid": -914061504, "ts": 1716454225599292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599295, "dur": 0,
    "args": {
      "External id": 261691,
      "cbid": 251, "correlation": 261691
    }
  },
  {
    "ph": "f", "id": 261691, "pid": 76337, "tid": -914061504, "ts": 1716454225599295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225651797, "dur": 12,
    "args": {
      "External id": 261692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261692,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 261692, "pid": 5, "tid": 7, "ts": 1716454225651797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599297, "dur": 12,
    "args": {
      "External id": 261692,
      "cbid": 211, "correlation": 261692
    }
  },
  {
    "ph": "s", "id": 261692, "pid": 76337, "tid": -914061504, "ts": 1716454225599297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225651810, "dur": 5,
    "args": {
      "External id": 261694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261694,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 261694, "pid": 5, "tid": 7, "ts": 1716454225651810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599311, "dur": 6,
    "args": {
      "External id": 261694,
      "cbid": 211, "correlation": 261694
    }
  },
  {
    "ph": "s", "id": 261694, "pid": 76337, "tid": -914061504, "ts": 1716454225599311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599368, "dur": 1,
    "args": {
      "External id": 261705,
      "cbid": 251, "correlation": 261705
    }
  },
  {
    "ph": "f", "id": 261705, "pid": 76337, "tid": -914061504, "ts": 1716454225599368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599371, "dur": 0,
    "args": {
      "External id": 261706,
      "cbid": 251, "correlation": 261706
    }
  },
  {
    "ph": "f", "id": 261706, "pid": 76337, "tid": -914061504, "ts": 1716454225599371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225651816, "dur": 7,
    "args": {
      "External id": 261707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261707,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 261707, "pid": 5, "tid": 7, "ts": 1716454225651816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599373, "dur": 12,
    "args": {
      "External id": 261707,
      "cbid": 211, "correlation": 261707
    }
  },
  {
    "ph": "s", "id": 261707, "pid": 76337, "tid": -914061504, "ts": 1716454225599373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225651825, "dur": 3,
    "args": {
      "External id": 261709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261709,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 261709, "pid": 5, "tid": 7, "ts": 1716454225651825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599387, "dur": 5,
    "args": {
      "External id": 261709,
      "cbid": 211, "correlation": 261709
    }
  },
  {
    "ph": "s", "id": 261709, "pid": 76337, "tid": -914061504, "ts": 1716454225599387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225651829, "dur": 93,
    "args": {
      "External id": 261730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261730,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 261730, "pid": 5, "tid": 7, "ts": 1716454225651829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599459, "dur": 13,
    "args": {
      "External id": 261730,
      "cbid": 211, "correlation": 261730
    }
  },
  {
    "ph": "s", "id": 261730, "pid": 76337, "tid": -914061504, "ts": 1716454225599459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599557, "dur": 1,
    "args": {
      "External id": 261748,
      "cbid": 251, "correlation": 261748
    }
  },
  {
    "ph": "f", "id": 261748, "pid": 76337, "tid": -914061504, "ts": 1716454225599557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225651924, "dur": 99,
    "args": {
      "External id": 261750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261750,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261750, "pid": 5, "tid": 7, "ts": 1716454225651924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599562, "dur": 14,
    "args": {
      "External id": 261750,
      "cbid": 211, "correlation": 261750
    }
  },
  {
    "ph": "s", "id": 261750, "pid": 76337, "tid": -914061504, "ts": 1716454225599562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225652024, "dur": 19,
    "args": {
      "External id": 261758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261758,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261758, "pid": 5, "tid": 7, "ts": 1716454225652024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599631, "dur": 12,
    "args": {
      "External id": 261758,
      "cbid": 211, "correlation": 261758
    }
  },
  {
    "ph": "s", "id": 261758, "pid": 76337, "tid": -914061504, "ts": 1716454225599631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225652045, "dur": 36,
    "args": {
      "External id": 261766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261766,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261766, "pid": 5, "tid": 7, "ts": 1716454225652045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599672, "dur": 10,
    "args": {
      "External id": 261766,
      "cbid": 211, "correlation": 261766
    }
  },
  {
    "ph": "s", "id": 261766, "pid": 76337, "tid": -914061504, "ts": 1716454225599672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225652082, "dur": 35,
    "args": {
      "External id": 261788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261788,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261788, "pid": 5, "tid": 7, "ts": 1716454225652082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599723, "dur": 11,
    "args": {
      "External id": 261788,
      "cbid": 211, "correlation": 261788
    }
  },
  {
    "ph": "s", "id": 261788, "pid": 76337, "tid": -914061504, "ts": 1716454225599723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599811, "dur": 1,
    "args": {
      "External id": 261804,
      "cbid": 251, "correlation": 261804
    }
  },
  {
    "ph": "f", "id": 261804, "pid": 76337, "tid": -914061504, "ts": 1716454225599811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599816, "dur": 0,
    "args": {
      "External id": 261806,
      "cbid": 251, "correlation": 261806
    }
  },
  {
    "ph": "f", "id": 261806, "pid": 76337, "tid": -914061504, "ts": 1716454225599816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225652118, "dur": 548,
    "args": {
      "External id": 261807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261807,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 261807, "pid": 5, "tid": 7, "ts": 1716454225652118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599820, "dur": 13,
    "args": {
      "External id": 261807,
      "cbid": 211, "correlation": 261807
    }
  },
  {
    "ph": "s", "id": 261807, "pid": 76337, "tid": -914061504, "ts": 1716454225599820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225652667, "dur": 127,
    "args": {
      "External id": 261815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261815, "pid": 5, "tid": 7, "ts": 1716454225652667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599885, "dur": 13,
    "args": {
      "External id": 261815,
      "cbid": 211, "correlation": 261815
    }
  },
  {
    "ph": "s", "id": 261815, "pid": 76337, "tid": -914061504, "ts": 1716454225599885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225652795, "dur": 128,
    "args": {
      "External id": 261823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261823,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261823, "pid": 5, "tid": 7, "ts": 1716454225652795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225599915, "dur": 8,
    "args": {
      "External id": 261823,
      "cbid": 211, "correlation": 261823
    }
  },
  {
    "ph": "s", "id": 261823, "pid": 76337, "tid": -914061504, "ts": 1716454225599915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600000, "dur": 1,
    "args": {
      "External id": 261839,
      "cbid": 251, "correlation": 261839
    }
  },
  {
    "ph": "f", "id": 261839, "pid": 76337, "tid": -914061504, "ts": 1716454225600000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225652925, "dur": 309,
    "args": {
      "External id": 261841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261841,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261841, "pid": 5, "tid": 7, "ts": 1716454225652925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600006, "dur": 13,
    "args": {
      "External id": 261841,
      "cbid": 211, "correlation": 261841
    }
  },
  {
    "ph": "s", "id": 261841, "pid": 76337, "tid": -914061504, "ts": 1716454225600006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225653235, "dur": 27,
    "args": {
      "External id": 261849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261849,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261849, "pid": 5, "tid": 7, "ts": 1716454225653235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600050, "dur": 10,
    "args": {
      "External id": 261849,
      "cbid": 211, "correlation": 261849
    }
  },
  {
    "ph": "s", "id": 261849, "pid": 76337, "tid": -914061504, "ts": 1716454225600050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225653264, "dur": 84,
    "args": {
      "External id": 261860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261860,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261860, "pid": 5, "tid": 7, "ts": 1716454225653264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600118, "dur": 12,
    "args": {
      "External id": 261860,
      "cbid": 211, "correlation": 261860
    }
  },
  {
    "ph": "s", "id": 261860, "pid": 76337, "tid": -914061504, "ts": 1716454225600118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600181, "dur": 0,
    "args": {
      "External id": 261872,
      "cbid": 317, "correlation": 261872
    }
  },
  {
    "ph": "f", "id": 261872, "pid": 76337, "tid": -914061504, "ts": 1716454225600181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600182, "dur": 0,
    "args": {
      "External id": 261873,
      "cbid": 203, "correlation": 261873
    }
  },
  {
    "ph": "f", "id": 261873, "pid": 76337, "tid": -914061504, "ts": 1716454225600182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600183, "dur": 0,
    "args": {
      "External id": 261874,
      "cbid": 205, "correlation": 261874
    }
  },
  {
    "ph": "f", "id": 261874, "pid": 76337, "tid": -914061504, "ts": 1716454225600183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653349, "dur": 24,
    "args": {
      "External id": 261878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261878,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261878, "pid": 5, "tid": 7, "ts": 1716454225653349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600198, "dur": 12,
    "args": {
      "External id": 261878,
      "cbid": 211, "correlation": 261878
    }
  },
  {
    "ph": "s", "id": 261878, "pid": 76337, "tid": -914061504, "ts": 1716454225600198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225653374, "dur": 123,
    "args": {
      "External id": 261880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261880,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261880, "pid": 5, "tid": 7, "ts": 1716454225653374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600216, "dur": 6,
    "args": {
      "External id": 261880,
      "cbid": 211, "correlation": 261880
    }
  },
  {
    "ph": "s", "id": 261880, "pid": 76337, "tid": -914061504, "ts": 1716454225600216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653498, "dur": 22,
    "args": {
      "External id": 261882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261882,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261882, "pid": 5, "tid": 7, "ts": 1716454225653498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600226, "dur": 5,
    "args": {
      "External id": 261882,
      "cbid": 211, "correlation": 261882
    }
  },
  {
    "ph": "s", "id": 261882, "pid": 76337, "tid": -914061504, "ts": 1716454225600226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225653521, "dur": 32,
    "args": {
      "External id": 261888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261888, "pid": 5, "tid": 7, "ts": 1716454225653521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600254, "dur": 8,
    "args": {
      "External id": 261888,
      "cbid": 211, "correlation": 261888
    }
  },
  {
    "ph": "s", "id": 261888, "pid": 76337, "tid": -914061504, "ts": 1716454225600254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225653554, "dur": 26,
    "args": {
      "External id": 261896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261896,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261896, "pid": 5, "tid": 7, "ts": 1716454225653554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600286, "dur": 9,
    "args": {
      "External id": 261896,
      "cbid": 211, "correlation": 261896
    }
  },
  {
    "ph": "s", "id": 261896, "pid": 76337, "tid": -914061504, "ts": 1716454225600286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600357, "dur": 0,
    "args": {
      "External id": 261906,
      "cbid": 317, "correlation": 261906
    }
  },
  {
    "ph": "f", "id": 261906, "pid": 76337, "tid": -914061504, "ts": 1716454225600357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600357, "dur": 0,
    "args": {
      "External id": 261907,
      "cbid": 203, "correlation": 261907
    }
  },
  {
    "ph": "f", "id": 261907, "pid": 76337, "tid": -914061504, "ts": 1716454225600357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600358, "dur": 0,
    "args": {
      "External id": 261908,
      "cbid": 205, "correlation": 261908
    }
  },
  {
    "ph": "f", "id": 261908, "pid": 76337, "tid": -914061504, "ts": 1716454225600358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653582, "dur": 23,
    "args": {
      "External id": 261912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261912, "pid": 5, "tid": 7, "ts": 1716454225653582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600372, "dur": 12,
    "args": {
      "External id": 261912,
      "cbid": 211, "correlation": 261912
    }
  },
  {
    "ph": "s", "id": 261912, "pid": 76337, "tid": -914061504, "ts": 1716454225600372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653607, "dur": 45,
    "args": {
      "External id": 261914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261914, "pid": 5, "tid": 7, "ts": 1716454225653607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600387, "dur": 6,
    "args": {
      "External id": 261914,
      "cbid": 211, "correlation": 261914
    }
  },
  {
    "ph": "s", "id": 261914, "pid": 76337, "tid": -914061504, "ts": 1716454225600387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225653653, "dur": 237,
    "args": {
      "External id": 261916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261916,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 261916, "pid": 5, "tid": 7, "ts": 1716454225653653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600400, "dur": 6,
    "args": {
      "External id": 261916,
      "cbid": 211, "correlation": 261916
    }
  },
  {
    "ph": "s", "id": 261916, "pid": 76337, "tid": -914061504, "ts": 1716454225600400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653892, "dur": 7,
    "args": {
      "External id": 261918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261918,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261918, "pid": 5, "tid": 7, "ts": 1716454225653892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600409, "dur": 5,
    "args": {
      "External id": 261918,
      "cbid": 211, "correlation": 261918
    }
  },
  {
    "ph": "s", "id": 261918, "pid": 76337, "tid": -914061504, "ts": 1716454225600409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225653899, "dur": 9,
    "args": {
      "External id": 261924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261924, "pid": 5, "tid": 7, "ts": 1716454225653899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600435, "dur": 10,
    "args": {
      "External id": 261924,
      "cbid": 211, "correlation": 261924
    }
  },
  {
    "ph": "s", "id": 261924, "pid": 76337, "tid": -914061504, "ts": 1716454225600435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653910, "dur": 12,
    "args": {
      "External id": 261944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261944,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 261944, "pid": 5, "tid": 7, "ts": 1716454225653910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600527, "dur": 13,
    "args": {
      "External id": 261944,
      "cbid": 211, "correlation": 261944
    }
  },
  {
    "ph": "s", "id": 261944, "pid": 76337, "tid": -914061504, "ts": 1716454225600527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225653923, "dur": 4,
    "args": {
      "External id": 261956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261956,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 261956, "pid": 5, "tid": 7, "ts": 1716454225653923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600550, "dur": 6,
    "args": {
      "External id": 261956,
      "cbid": 211, "correlation": 261956
    }
  },
  {
    "ph": "s", "id": 261956, "pid": 76337, "tid": -914061504, "ts": 1716454225600550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225653929, "dur": 13,
    "args": {
      "External id": 261959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261959, "pid": 5, "tid": 7, "ts": 1716454225653929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600568, "dur": 6,
    "args": {
      "External id": 261959,
      "cbid": 211, "correlation": 261959
    }
  },
  {
    "ph": "s", "id": 261959, "pid": 76337, "tid": -914061504, "ts": 1716454225600568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225653943, "dur": 7,
    "args": {
      "External id": 261968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261968,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261968, "pid": 5, "tid": 7, "ts": 1716454225653943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600607, "dur": 10,
    "args": {
      "External id": 261968,
      "cbid": 211, "correlation": 261968
    }
  },
  {
    "ph": "s", "id": 261968, "pid": 76337, "tid": -914061504, "ts": 1716454225600607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600659, "dur": 0,
    "args": {
      "External id": 261978,
      "cbid": 317, "correlation": 261978
    }
  },
  {
    "ph": "f", "id": 261978, "pid": 76337, "tid": -914061504, "ts": 1716454225600659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600660, "dur": 0,
    "args": {
      "External id": 261979,
      "cbid": 203, "correlation": 261979
    }
  },
  {
    "ph": "f", "id": 261979, "pid": 76337, "tid": -914061504, "ts": 1716454225600660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600661, "dur": 0,
    "args": {
      "External id": 261980,
      "cbid": 205, "correlation": 261980
    }
  },
  {
    "ph": "f", "id": 261980, "pid": 76337, "tid": -914061504, "ts": 1716454225600661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653951, "dur": 6,
    "args": {
      "External id": 261984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261984,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261984, "pid": 5, "tid": 7, "ts": 1716454225653951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600676, "dur": 11,
    "args": {
      "External id": 261984,
      "cbid": 211, "correlation": 261984
    }
  },
  {
    "ph": "s", "id": 261984, "pid": 76337, "tid": -914061504, "ts": 1716454225600676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225653958, "dur": 85,
    "args": {
      "External id": 261986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261986, "pid": 5, "tid": 7, "ts": 1716454225653958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600691, "dur": 6,
    "args": {
      "External id": 261986,
      "cbid": 211, "correlation": 261986
    }
  },
  {
    "ph": "s", "id": 261986, "pid": 76337, "tid": -914061504, "ts": 1716454225600691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225654046, "dur": 1,
    "args": {
      "External id": 261988,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 261988,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 261988, "pid": 5, "tid": 7, "ts": 1716454225654046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600705, "dur": 9,
    "args": {
      "External id": 261988,
      "cbid": 51, "correlation": 261988
    }
  },
  {
    "ph": "s", "id": 261988, "pid": 76337, "tid": -914061504, "ts": 1716454225600705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225654049, "dur": 546,
    "args": {
      "External id": 261989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261989,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 261989, "pid": 5, "tid": 7, "ts": 1716454225654049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600715, "dur": 8,
    "args": {
      "External id": 261989,
      "cbid": 211, "correlation": 261989
    }
  },
  {
    "ph": "s", "id": 261989, "pid": 76337, "tid": -914061504, "ts": 1716454225600715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225654597, "dur": 12,
    "args": {
      "External id": 261991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261991,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261991, "pid": 5, "tid": 7, "ts": 1716454225654597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600727, "dur": 5,
    "args": {
      "External id": 261991,
      "cbid": 211, "correlation": 261991
    }
  },
  {
    "ph": "s", "id": 261991, "pid": 76337, "tid": -914061504, "ts": 1716454225600727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225654610, "dur": 15,
    "args": {
      "External id": 261997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 261997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 261997, "pid": 5, "tid": 7, "ts": 1716454225654610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600755, "dur": 9,
    "args": {
      "External id": 261997,
      "cbid": 211, "correlation": 261997
    }
  },
  {
    "ph": "s", "id": 261997, "pid": 76337, "tid": -914061504, "ts": 1716454225600755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225654626, "dur": 3,
    "args": {
      "External id": 262005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 262005, "pid": 5, "tid": 7, "ts": 1716454225654626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600799, "dur": 9,
    "args": {
      "External id": 262005,
      "cbid": 211, "correlation": 262005
    }
  },
  {
    "ph": "s", "id": 262005, "pid": 76337, "tid": -914061504, "ts": 1716454225600799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600864, "dur": 1,
    "args": {
      "External id": 262021,
      "cbid": 251, "correlation": 262021
    }
  },
  {
    "ph": "f", "id": 262021, "pid": 76337, "tid": -914061504, "ts": 1716454225600864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600870, "dur": 0,
    "args": {
      "External id": 262023,
      "cbid": 251, "correlation": 262023
    }
  },
  {
    "ph": "f", "id": 262023, "pid": 76337, "tid": -914061504, "ts": 1716454225600870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225654631, "dur": 13,
    "args": {
      "External id": 262024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262024,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262024, "pid": 5, "tid": 7, "ts": 1716454225654631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600871, "dur": 11,
    "args": {
      "External id": 262024,
      "cbid": 211, "correlation": 262024
    }
  },
  {
    "ph": "s", "id": 262024, "pid": 76337, "tid": -914061504, "ts": 1716454225600871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225654646, "dur": 5,
    "args": {
      "External id": 262026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262026,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262026, "pid": 5, "tid": 7, "ts": 1716454225654646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600885, "dur": 5,
    "args": {
      "External id": 262026,
      "cbid": 211, "correlation": 262026
    }
  },
  {
    "ph": "s", "id": 262026, "pid": 76337, "tid": -914061504, "ts": 1716454225600885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225654652, "dur": 17,
    "args": {
      "External id": 262036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262036, "pid": 5, "tid": 7, "ts": 1716454225654652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225600942, "dur": 13,
    "args": {
      "External id": 262036,
      "cbid": 211, "correlation": 262036
    }
  },
  {
    "ph": "s", "id": 262036, "pid": 76337, "tid": -914061504, "ts": 1716454225600942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225654671, "dur": 18,
    "args": {
      "External id": 262056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262056,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 262056, "pid": 5, "tid": 7, "ts": 1716454225654671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601015, "dur": 11,
    "args": {
      "External id": 262056,
      "cbid": 211, "correlation": 262056
    }
  },
  {
    "ph": "s", "id": 262056, "pid": 76337, "tid": -914061504, "ts": 1716454225601015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225654690, "dur": 5,
    "args": {
      "External id": 262068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262068,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 262068, "pid": 5, "tid": 7, "ts": 1716454225654690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601036, "dur": 7,
    "args": {
      "External id": 262068,
      "cbid": 211, "correlation": 262068
    }
  },
  {
    "ph": "s", "id": 262068, "pid": 76337, "tid": -914061504, "ts": 1716454225601036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225654697, "dur": 17,
    "args": {
      "External id": 262071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262071,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262071, "pid": 5, "tid": 7, "ts": 1716454225654697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601056, "dur": 7,
    "args": {
      "External id": 262071,
      "cbid": 211, "correlation": 262071
    }
  },
  {
    "ph": "s", "id": 262071, "pid": 76337, "tid": -914061504, "ts": 1716454225601056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225654714, "dur": 11,
    "args": {
      "External id": 262080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262080,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262080, "pid": 5, "tid": 7, "ts": 1716454225654714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601098, "dur": 10,
    "args": {
      "External id": 262080,
      "cbid": 211, "correlation": 262080
    }
  },
  {
    "ph": "s", "id": 262080, "pid": 76337, "tid": -914061504, "ts": 1716454225601098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601161, "dur": 0,
    "args": {
      "External id": 262090,
      "cbid": 317, "correlation": 262090
    }
  },
  {
    "ph": "f", "id": 262090, "pid": 76337, "tid": -914061504, "ts": 1716454225601161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601162, "dur": 0,
    "args": {
      "External id": 262091,
      "cbid": 203, "correlation": 262091
    }
  },
  {
    "ph": "f", "id": 262091, "pid": 76337, "tid": -914061504, "ts": 1716454225601162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601162, "dur": 0,
    "args": {
      "External id": 262092,
      "cbid": 205, "correlation": 262092
    }
  },
  {
    "ph": "f", "id": 262092, "pid": 76337, "tid": -914061504, "ts": 1716454225601162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225654727, "dur": 11,
    "args": {
      "External id": 262096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262096, "pid": 5, "tid": 7, "ts": 1716454225654727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601180, "dur": 12,
    "args": {
      "External id": 262096,
      "cbid": 211, "correlation": 262096
    }
  },
  {
    "ph": "s", "id": 262096, "pid": 76337, "tid": -914061504, "ts": 1716454225601180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225654739, "dur": 165,
    "args": {
      "External id": 262098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262098, "pid": 5, "tid": 7, "ts": 1716454225654739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601194, "dur": 6,
    "args": {
      "External id": 262098,
      "cbid": 211, "correlation": 262098
    }
  },
  {
    "ph": "s", "id": 262098, "pid": 76337, "tid": -914061504, "ts": 1716454225601194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225654907, "dur": 1,
    "args": {
      "External id": 262100,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 262100,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 262100, "pid": 5, "tid": 7, "ts": 1716454225654907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601207, "dur": 7,
    "args": {
      "External id": 262100,
      "cbid": 51, "correlation": 262100
    }
  },
  {
    "ph": "s", "id": 262100, "pid": 76337, "tid": -914061504, "ts": 1716454225601207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225654911, "dur": 674,
    "args": {
      "External id": 262101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262101,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262101, "pid": 5, "tid": 7, "ts": 1716454225654911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601215, "dur": 6,
    "args": {
      "External id": 262101,
      "cbid": 211, "correlation": 262101
    }
  },
  {
    "ph": "s", "id": 262101, "pid": 76337, "tid": -914061504, "ts": 1716454225601215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225655586, "dur": 13,
    "args": {
      "External id": 262103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262103,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262103, "pid": 5, "tid": 7, "ts": 1716454225655586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601226, "dur": 5,
    "args": {
      "External id": 262103,
      "cbid": 211, "correlation": 262103
    }
  },
  {
    "ph": "s", "id": 262103, "pid": 76337, "tid": -914061504, "ts": 1716454225601226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225655600, "dur": 15,
    "args": {
      "External id": 262109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262109,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262109, "pid": 5, "tid": 7, "ts": 1716454225655600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601255, "dur": 8,
    "args": {
      "External id": 262109,
      "cbid": 211, "correlation": 262109
    }
  },
  {
    "ph": "s", "id": 262109, "pid": 76337, "tid": -914061504, "ts": 1716454225601255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601313, "dur": 0,
    "args": {
      "External id": 262119,
      "cbid": 317, "correlation": 262119
    }
  },
  {
    "ph": "f", "id": 262119, "pid": 76337, "tid": -914061504, "ts": 1716454225601313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601314, "dur": 0,
    "args": {
      "External id": 262120,
      "cbid": 203, "correlation": 262120
    }
  },
  {
    "ph": "f", "id": 262120, "pid": 76337, "tid": -914061504, "ts": 1716454225601314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601314, "dur": 0,
    "args": {
      "External id": 262121,
      "cbid": 205, "correlation": 262121
    }
  },
  {
    "ph": "f", "id": 262121, "pid": 76337, "tid": -914061504, "ts": 1716454225601314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225655617, "dur": 8,
    "args": {
      "External id": 262125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262125,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262125, "pid": 5, "tid": 7, "ts": 1716454225655617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601329, "dur": 11,
    "args": {
      "External id": 262125,
      "cbid": 211, "correlation": 262125
    }
  },
  {
    "ph": "s", "id": 262125, "pid": 76337, "tid": -914061504, "ts": 1716454225601329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225655626, "dur": 3,
    "args": {
      "External id": 262127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262127, "pid": 5, "tid": 7, "ts": 1716454225655626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601347, "dur": 6,
    "args": {
      "External id": 262127,
      "cbid": 211, "correlation": 262127
    }
  },
  {
    "ph": "s", "id": 262127, "pid": 76337, "tid": -914061504, "ts": 1716454225601347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601356, "dur": 0,
    "args": {
      "External id": 262128,
      "cbid": 51, "correlation": 262128
    }
  },
  {
    "ph": "s", "id": 262128, "pid": 76337, "tid": -914061504, "ts": 1716454225601356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225655631, "dur": 60,
    "args": {
      "External id": 262129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262129,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 262129, "pid": 5, "tid": 7, "ts": 1716454225655631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601357, "dur": 5,
    "args": {
      "External id": 262129,
      "cbid": 211, "correlation": 262129
    }
  },
  {
    "ph": "s", "id": 262129, "pid": 76337, "tid": -914061504, "ts": 1716454225601357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225655692, "dur": 14,
    "args": {
      "External id": 262134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262134, "pid": 5, "tid": 7, "ts": 1716454225655692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601383, "dur": 9,
    "args": {
      "External id": 262134,
      "cbid": 211, "correlation": 262134
    }
  },
  {
    "ph": "s", "id": 262134, "pid": 76337, "tid": -914061504, "ts": 1716454225601383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225655708, "dur": 12,
    "args": {
      "External id": 262142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262142,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262142, "pid": 5, "tid": 7, "ts": 1716454225655708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601412, "dur": 8,
    "args": {
      "External id": 262142,
      "cbid": 211, "correlation": 262142
    }
  },
  {
    "ph": "s", "id": 262142, "pid": 76337, "tid": -914061504, "ts": 1716454225601412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225655721, "dur": 11,
    "args": {
      "External id": 262150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262150,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262150, "pid": 5, "tid": 7, "ts": 1716454225655721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601440, "dur": 9,
    "args": {
      "External id": 262150,
      "cbid": 211, "correlation": 262150
    }
  },
  {
    "ph": "s", "id": 262150, "pid": 76337, "tid": -914061504, "ts": 1716454225601440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225655733, "dur": 19,
    "args": {
      "External id": 262170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262170,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 262170, "pid": 5, "tid": 7, "ts": 1716454225655733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601520, "dur": 12,
    "args": {
      "External id": 262170,
      "cbid": 211, "correlation": 262170
    }
  },
  {
    "ph": "s", "id": 262170, "pid": 76337, "tid": -914061504, "ts": 1716454225601520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225655754, "dur": 4,
    "args": {
      "External id": 262182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262182,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 262182, "pid": 5, "tid": 7, "ts": 1716454225655754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601543, "dur": 7,
    "args": {
      "External id": 262182,
      "cbid": 211, "correlation": 262182
    }
  },
  {
    "ph": "s", "id": 262182, "pid": 76337, "tid": -914061504, "ts": 1716454225601543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225655760, "dur": 17,
    "args": {
      "External id": 262185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262185, "pid": 5, "tid": 7, "ts": 1716454225655760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601561, "dur": 7,
    "args": {
      "External id": 262185,
      "cbid": 211, "correlation": 262185
    }
  },
  {
    "ph": "s", "id": 262185, "pid": 76337, "tid": -914061504, "ts": 1716454225601561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601617, "dur": 0,
    "args": {
      "External id": 262196,
      "cbid": 317, "correlation": 262196
    }
  },
  {
    "ph": "f", "id": 262196, "pid": 76337, "tid": -914061504, "ts": 1716454225601617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601618, "dur": 0,
    "args": {
      "External id": 262197,
      "cbid": 203, "correlation": 262197
    }
  },
  {
    "ph": "f", "id": 262197, "pid": 76337, "tid": -914061504, "ts": 1716454225601618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601619, "dur": 0,
    "args": {
      "External id": 262198,
      "cbid": 205, "correlation": 262198
    }
  },
  {
    "ph": "f", "id": 262198, "pid": 76337, "tid": -914061504, "ts": 1716454225601619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225655778, "dur": 12,
    "args": {
      "External id": 262202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262202, "pid": 5, "tid": 7, "ts": 1716454225655778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601634, "dur": 12,
    "args": {
      "External id": 262202,
      "cbid": 211, "correlation": 262202
    }
  },
  {
    "ph": "s", "id": 262202, "pid": 76337, "tid": -914061504, "ts": 1716454225601634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225655791, "dur": 3,
    "args": {
      "External id": 262204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262204, "pid": 5, "tid": 7, "ts": 1716454225655791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601652, "dur": 5,
    "args": {
      "External id": 262204,
      "cbid": 211, "correlation": 262204
    }
  },
  {
    "ph": "s", "id": 262204, "pid": 76337, "tid": -914061504, "ts": 1716454225601652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601661, "dur": 0,
    "args": {
      "External id": 262205,
      "cbid": 51, "correlation": 262205
    }
  },
  {
    "ph": "s", "id": 262205, "pid": 76337, "tid": -914061504, "ts": 1716454225601661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225655796, "dur": 99,
    "args": {
      "External id": 262206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262206,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 262206, "pid": 5, "tid": 7, "ts": 1716454225655796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601662, "dur": 5,
    "args": {
      "External id": 262206,
      "cbid": 211, "correlation": 262206
    }
  },
  {
    "ph": "s", "id": 262206, "pid": 76337, "tid": -914061504, "ts": 1716454225601662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225655897, "dur": 16,
    "args": {
      "External id": 262211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262211, "pid": 5, "tid": 7, "ts": 1716454225655897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601689, "dur": 9,
    "args": {
      "External id": 262211,
      "cbid": 211, "correlation": 262211
    }
  },
  {
    "ph": "s", "id": 262211, "pid": 76337, "tid": -914061504, "ts": 1716454225601689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225655914, "dur": 86,
    "args": {
      "External id": 262220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262220,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262220, "pid": 5, "tid": 7, "ts": 1716454225655914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601770, "dur": 14,
    "args": {
      "External id": 262220,
      "cbid": 211, "correlation": 262220
    }
  },
  {
    "ph": "s", "id": 262220, "pid": 76337, "tid": -914061504, "ts": 1716454225601770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225656001, "dur": 30,
    "args": {
      "External id": 262242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262242,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262242, "pid": 5, "tid": 7, "ts": 1716454225656001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601827, "dur": 10,
    "args": {
      "External id": 262242,
      "cbid": 211, "correlation": 262242
    }
  },
  {
    "ph": "s", "id": 262242, "pid": 76337, "tid": -914061504, "ts": 1716454225601827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601915, "dur": 1,
    "args": {
      "External id": 262253,
      "cbid": 251, "correlation": 262253
    }
  },
  {
    "ph": "f", "id": 262253, "pid": 76337, "tid": -914061504, "ts": 1716454225601915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225656033, "dur": 166,
    "args": {
      "External id": 262254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262254,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262254, "pid": 5, "tid": 7, "ts": 1716454225656033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601921, "dur": 13,
    "args": {
      "External id": 262254,
      "cbid": 211, "correlation": 262254
    }
  },
  {
    "ph": "s", "id": 262254, "pid": 76337, "tid": -914061504, "ts": 1716454225601921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225601998, "dur": 1,
    "args": {
      "External id": 262265,
      "cbid": 251, "correlation": 262265
    }
  },
  {
    "ph": "f", "id": 262265, "pid": 76337, "tid": -914061504, "ts": 1716454225601998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225656200, "dur": 161,
    "args": {
      "External id": 262266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262266,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262266, "pid": 5, "tid": 7, "ts": 1716454225656200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602003, "dur": 12,
    "args": {
      "External id": 262266,
      "cbid": 211, "correlation": 262266
    }
  },
  {
    "ph": "s", "id": 262266, "pid": 76337, "tid": -914061504, "ts": 1716454225602003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602070, "dur": 1,
    "args": {
      "External id": 262277,
      "cbid": 251, "correlation": 262277
    }
  },
  {
    "ph": "f", "id": 262277, "pid": 76337, "tid": -914061504, "ts": 1716454225602070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225656363, "dur": 161,
    "args": {
      "External id": 262278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262278,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262278, "pid": 5, "tid": 7, "ts": 1716454225656363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602074, "dur": 11,
    "args": {
      "External id": 262278,
      "cbid": 211, "correlation": 262278
    }
  },
  {
    "ph": "s", "id": 262278, "pid": 76337, "tid": -914061504, "ts": 1716454225602074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225656525, "dur": 342,
    "args": {
      "External id": 262303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262303,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262303, "pid": 5, "tid": 7, "ts": 1716454225656525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602161, "dur": 13,
    "args": {
      "External id": 262303,
      "cbid": 211, "correlation": 262303
    }
  },
  {
    "ph": "s", "id": 262303, "pid": 76337, "tid": -914061504, "ts": 1716454225602161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602262, "dur": 1,
    "args": {
      "External id": 262321,
      "cbid": 251, "correlation": 262321
    }
  },
  {
    "ph": "f", "id": 262321, "pid": 76337, "tid": -914061504, "ts": 1716454225602262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225656869, "dur": 169,
    "args": {
      "External id": 262323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262323,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262323, "pid": 5, "tid": 7, "ts": 1716454225656869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602267, "dur": 13,
    "args": {
      "External id": 262323,
      "cbid": 211, "correlation": 262323
    }
  },
  {
    "ph": "s", "id": 262323, "pid": 76337, "tid": -914061504, "ts": 1716454225602267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225657039, "dur": 20,
    "args": {
      "External id": 262331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262331,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262331, "pid": 5, "tid": 7, "ts": 1716454225657039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602335, "dur": 12,
    "args": {
      "External id": 262331,
      "cbid": 211, "correlation": 262331
    }
  },
  {
    "ph": "s", "id": 262331, "pid": 76337, "tid": -914061504, "ts": 1716454225602335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225657060, "dur": 28,
    "args": {
      "External id": 262339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262339,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262339, "pid": 5, "tid": 7, "ts": 1716454225657060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602374, "dur": 9,
    "args": {
      "External id": 262339,
      "cbid": 211, "correlation": 262339
    }
  },
  {
    "ph": "s", "id": 262339, "pid": 76337, "tid": -914061504, "ts": 1716454225602374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225657089, "dur": 18,
    "args": {
      "External id": 262350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262350, "pid": 5, "tid": 7, "ts": 1716454225657089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602446, "dur": 13,
    "args": {
      "External id": 262350,
      "cbid": 211, "correlation": 262350
    }
  },
  {
    "ph": "s", "id": 262350, "pid": 76337, "tid": -914061504, "ts": 1716454225602446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225657108, "dur": 17,
    "args": {
      "External id": 262372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262372,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262372, "pid": 5, "tid": 7, "ts": 1716454225657108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602477, "dur": 7,
    "args": {
      "External id": 262372,
      "cbid": 211, "correlation": 262372
    }
  },
  {
    "ph": "s", "id": 262372, "pid": 76337, "tid": -914061504, "ts": 1716454225602477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602563, "dur": 1,
    "args": {
      "External id": 262383,
      "cbid": 251, "correlation": 262383
    }
  },
  {
    "ph": "f", "id": 262383, "pid": 76337, "tid": -914061504, "ts": 1716454225602563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225657126, "dur": 91,
    "args": {
      "External id": 262384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262384,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262384, "pid": 5, "tid": 7, "ts": 1716454225657126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602569, "dur": 14,
    "args": {
      "External id": 262384,
      "cbid": 211, "correlation": 262384
    }
  },
  {
    "ph": "s", "id": 262384, "pid": 76337, "tid": -914061504, "ts": 1716454225602569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602639, "dur": 1,
    "args": {
      "External id": 262395,
      "cbid": 251, "correlation": 262395
    }
  },
  {
    "ph": "f", "id": 262395, "pid": 76337, "tid": -914061504, "ts": 1716454225602639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602643, "dur": 0,
    "args": {
      "External id": 262396,
      "cbid": 251, "correlation": 262396
    }
  },
  {
    "ph": "f", "id": 262396, "pid": 76337, "tid": -914061504, "ts": 1716454225602643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225657218, "dur": 13,
    "args": {
      "External id": 262397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262397,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262397, "pid": 5, "tid": 7, "ts": 1716454225657218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602644, "dur": 13,
    "args": {
      "External id": 262397,
      "cbid": 211, "correlation": 262397
    }
  },
  {
    "ph": "s", "id": 262397, "pid": 76337, "tid": -914061504, "ts": 1716454225602644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225657232, "dur": 6,
    "args": {
      "External id": 262399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262399,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262399, "pid": 5, "tid": 7, "ts": 1716454225657232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602659, "dur": 6,
    "args": {
      "External id": 262399,
      "cbid": 211, "correlation": 262399
    }
  },
  {
    "ph": "s", "id": 262399, "pid": 76337, "tid": -914061504, "ts": 1716454225602659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602717, "dur": 1,
    "args": {
      "External id": 262410,
      "cbid": 251, "correlation": 262410
    }
  },
  {
    "ph": "f", "id": 262410, "pid": 76337, "tid": -914061504, "ts": 1716454225602717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602720, "dur": 0,
    "args": {
      "External id": 262411,
      "cbid": 251, "correlation": 262411
    }
  },
  {
    "ph": "f", "id": 262411, "pid": 76337, "tid": -914061504, "ts": 1716454225602720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225657240, "dur": 8,
    "args": {
      "External id": 262412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262412,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262412, "pid": 5, "tid": 7, "ts": 1716454225657240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602722, "dur": 12,
    "args": {
      "External id": 262412,
      "cbid": 211, "correlation": 262412
    }
  },
  {
    "ph": "s", "id": 262412, "pid": 76337, "tid": -914061504, "ts": 1716454225602722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225657249, "dur": 3,
    "args": {
      "External id": 262414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262414,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262414, "pid": 5, "tid": 7, "ts": 1716454225657249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602735, "dur": 5,
    "args": {
      "External id": 262414,
      "cbid": 211, "correlation": 262414
    }
  },
  {
    "ph": "s", "id": 262414, "pid": 76337, "tid": -914061504, "ts": 1716454225602735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225657254, "dur": 56,
    "args": {
      "External id": 262439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262439,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262439, "pid": 5, "tid": 7, "ts": 1716454225657254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602812, "dur": 12,
    "args": {
      "External id": 262439,
      "cbid": 211, "correlation": 262439
    }
  },
  {
    "ph": "s", "id": 262439, "pid": 76337, "tid": -914061504, "ts": 1716454225602812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602911, "dur": 2,
    "args": {
      "External id": 262457,
      "cbid": 251, "correlation": 262457
    }
  },
  {
    "ph": "f", "id": 262457, "pid": 76337, "tid": -914061504, "ts": 1716454225602911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225657312, "dur": 93,
    "args": {
      "External id": 262459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262459,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262459, "pid": 5, "tid": 7, "ts": 1716454225657312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602917, "dur": 14,
    "args": {
      "External id": 262459,
      "cbid": 211, "correlation": 262459
    }
  },
  {
    "ph": "s", "id": 262459, "pid": 76337, "tid": -914061504, "ts": 1716454225602917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225657407, "dur": 9,
    "args": {
      "External id": 262467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262467,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262467, "pid": 5, "tid": 7, "ts": 1716454225657407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225602996, "dur": 12,
    "args": {
      "External id": 262467,
      "cbid": 211, "correlation": 262467
    }
  },
  {
    "ph": "s", "id": 262467, "pid": 76337, "tid": -914061504, "ts": 1716454225602996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225657417, "dur": 22,
    "args": {
      "External id": 262475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262475, "pid": 5, "tid": 7, "ts": 1716454225657417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603038, "dur": 9,
    "args": {
      "External id": 262475,
      "cbid": 211, "correlation": 262475
    }
  },
  {
    "ph": "s", "id": 262475, "pid": 76337, "tid": -914061504, "ts": 1716454225603038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225657440, "dur": 17,
    "args": {
      "External id": 262497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262497,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262497, "pid": 5, "tid": 7, "ts": 1716454225657440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603090, "dur": 10,
    "args": {
      "External id": 262497,
      "cbid": 211, "correlation": 262497
    }
  },
  {
    "ph": "s", "id": 262497, "pid": 76337, "tid": -914061504, "ts": 1716454225603090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603181, "dur": 2,
    "args": {
      "External id": 262513,
      "cbid": 251, "correlation": 262513
    }
  },
  {
    "ph": "f", "id": 262513, "pid": 76337, "tid": -914061504, "ts": 1716454225603181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603187, "dur": 0,
    "args": {
      "External id": 262515,
      "cbid": 251, "correlation": 262515
    }
  },
  {
    "ph": "f", "id": 262515, "pid": 76337, "tid": -914061504, "ts": 1716454225603187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225657459, "dur": 498,
    "args": {
      "External id": 262516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262516,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262516, "pid": 5, "tid": 7, "ts": 1716454225657459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603189, "dur": 14,
    "args": {
      "External id": 262516,
      "cbid": 211, "correlation": 262516
    }
  },
  {
    "ph": "s", "id": 262516, "pid": 76337, "tid": -914061504, "ts": 1716454225603189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225657959, "dur": 67,
    "args": {
      "External id": 262524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262524, "pid": 5, "tid": 7, "ts": 1716454225657959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603256, "dur": 13,
    "args": {
      "External id": 262524,
      "cbid": 211, "correlation": 262524
    }
  },
  {
    "ph": "s", "id": 262524, "pid": 76337, "tid": -914061504, "ts": 1716454225603256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225658027, "dur": 69,
    "args": {
      "External id": 262532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262532, "pid": 5, "tid": 7, "ts": 1716454225658027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603287, "dur": 8,
    "args": {
      "External id": 262532,
      "cbid": 211, "correlation": 262532
    }
  },
  {
    "ph": "s", "id": 262532, "pid": 76337, "tid": -914061504, "ts": 1716454225603287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603370, "dur": 1,
    "args": {
      "External id": 262548,
      "cbid": 251, "correlation": 262548
    }
  },
  {
    "ph": "f", "id": 262548, "pid": 76337, "tid": -914061504, "ts": 1716454225603370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225658099, "dur": 1,
    "args": {
      "External id": 262550,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 262550,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 262550, "pid": 5, "tid": 7, "ts": 1716454225658099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603375, "dur": 11,
    "args": {
      "External id": 262550,
      "cbid": 51, "correlation": 262550
    }
  },
  {
    "ph": "s", "id": 262550, "pid": 76337, "tid": -914061504, "ts": 1716454225603375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225658103, "dur": 274,
    "args": {
      "External id": 262551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262551,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262551, "pid": 5, "tid": 7, "ts": 1716454225658103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603388, "dur": 11,
    "args": {
      "External id": 262551,
      "cbid": 211, "correlation": 262551
    }
  },
  {
    "ph": "s", "id": 262551, "pid": 76337, "tid": -914061504, "ts": 1716454225603388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225658378, "dur": 14,
    "args": {
      "External id": 262559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262559,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262559, "pid": 5, "tid": 7, "ts": 1716454225658378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603431, "dur": 10,
    "args": {
      "External id": 262559,
      "cbid": 211, "correlation": 262559
    }
  },
  {
    "ph": "s", "id": 262559, "pid": 76337, "tid": -914061504, "ts": 1716454225603431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225658393, "dur": 39,
    "args": {
      "External id": 262570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262570, "pid": 5, "tid": 7, "ts": 1716454225658393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603498, "dur": 12,
    "args": {
      "External id": 262570,
      "cbid": 211, "correlation": 262570
    }
  },
  {
    "ph": "s", "id": 262570, "pid": 76337, "tid": -914061504, "ts": 1716454225603498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603562, "dur": 0,
    "args": {
      "External id": 262582,
      "cbid": 317, "correlation": 262582
    }
  },
  {
    "ph": "f", "id": 262582, "pid": 76337, "tid": -914061504, "ts": 1716454225603562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603563, "dur": 0,
    "args": {
      "External id": 262583,
      "cbid": 203, "correlation": 262583
    }
  },
  {
    "ph": "f", "id": 262583, "pid": 76337, "tid": -914061504, "ts": 1716454225603563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603565, "dur": 0,
    "args": {
      "External id": 262584,
      "cbid": 205, "correlation": 262584
    }
  },
  {
    "ph": "f", "id": 262584, "pid": 76337, "tid": -914061504, "ts": 1716454225603565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225658434, "dur": 13,
    "args": {
      "External id": 262588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262588,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262588, "pid": 5, "tid": 7, "ts": 1716454225658434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603580, "dur": 12,
    "args": {
      "External id": 262588,
      "cbid": 211, "correlation": 262588
    }
  },
  {
    "ph": "s", "id": 262588, "pid": 76337, "tid": -914061504, "ts": 1716454225603580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225658448, "dur": 4,
    "args": {
      "External id": 262590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262590, "pid": 5, "tid": 7, "ts": 1716454225658448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603596, "dur": 5,
    "args": {
      "External id": 262590,
      "cbid": 211, "correlation": 262590
    }
  },
  {
    "ph": "s", "id": 262590, "pid": 76337, "tid": -914061504, "ts": 1716454225603596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603604, "dur": 0,
    "args": {
      "External id": 262591,
      "cbid": 51, "correlation": 262591
    }
  },
  {
    "ph": "s", "id": 262591, "pid": 76337, "tid": -914061504, "ts": 1716454225603604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225658453, "dur": 99,
    "args": {
      "External id": 262592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262592,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 262592, "pid": 5, "tid": 7, "ts": 1716454225658453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603605, "dur": 5,
    "args": {
      "External id": 262592,
      "cbid": 211, "correlation": 262592
    }
  },
  {
    "ph": "s", "id": 262592, "pid": 76337, "tid": -914061504, "ts": 1716454225603605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225658553, "dur": 17,
    "args": {
      "External id": 262597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262597,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262597, "pid": 5, "tid": 7, "ts": 1716454225658553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603633, "dur": 9,
    "args": {
      "External id": 262597,
      "cbid": 211, "correlation": 262597
    }
  },
  {
    "ph": "s", "id": 262597, "pid": 76337, "tid": -914061504, "ts": 1716454225603633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225658572, "dur": 12,
    "args": {
      "External id": 262605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262605,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262605, "pid": 5, "tid": 7, "ts": 1716454225658572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603664, "dur": 9,
    "args": {
      "External id": 262605,
      "cbid": 211, "correlation": 262605
    }
  },
  {
    "ph": "s", "id": 262605, "pid": 76337, "tid": -914061504, "ts": 1716454225603664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225658585, "dur": 19,
    "args": {
      "External id": 262625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262625,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 262625, "pid": 5, "tid": 7, "ts": 1716454225658585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603737, "dur": 11,
    "args": {
      "External id": 262625,
      "cbid": 211, "correlation": 262625
    }
  },
  {
    "ph": "s", "id": 262625, "pid": 76337, "tid": -914061504, "ts": 1716454225603737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225658605, "dur": 5,
    "args": {
      "External id": 262637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262637,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 262637, "pid": 5, "tid": 7, "ts": 1716454225658605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603759, "dur": 6,
    "args": {
      "External id": 262637,
      "cbid": 211, "correlation": 262637
    }
  },
  {
    "ph": "s", "id": 262637, "pid": 76337, "tid": -914061504, "ts": 1716454225603759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225658611, "dur": 18,
    "args": {
      "External id": 262640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262640, "pid": 5, "tid": 7, "ts": 1716454225658611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603777, "dur": 7,
    "args": {
      "External id": 262640,
      "cbid": 211, "correlation": 262640
    }
  },
  {
    "ph": "s", "id": 262640, "pid": 76337, "tid": -914061504, "ts": 1716454225603777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225658630, "dur": 12,
    "args": {
      "External id": 262649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262649,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262649, "pid": 5, "tid": 7, "ts": 1716454225658630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603814, "dur": 11,
    "args": {
      "External id": 262649,
      "cbid": 211, "correlation": 262649
    }
  },
  {
    "ph": "s", "id": 262649, "pid": 76337, "tid": -914061504, "ts": 1716454225603814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603866, "dur": 0,
    "args": {
      "External id": 262659,
      "cbid": 317, "correlation": 262659
    }
  },
  {
    "ph": "f", "id": 262659, "pid": 76337, "tid": -914061504, "ts": 1716454225603866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603866, "dur": 0,
    "args": {
      "External id": 262660,
      "cbid": 203, "correlation": 262660
    }
  },
  {
    "ph": "f", "id": 262660, "pid": 76337, "tid": -914061504, "ts": 1716454225603866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603867, "dur": 0,
    "args": {
      "External id": 262661,
      "cbid": 205, "correlation": 262661
    }
  },
  {
    "ph": "f", "id": 262661, "pid": 76337, "tid": -914061504, "ts": 1716454225603867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225658643, "dur": 11,
    "args": {
      "External id": 262665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262665, "pid": 5, "tid": 7, "ts": 1716454225658643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603881, "dur": 11,
    "args": {
      "External id": 262665,
      "cbid": 211, "correlation": 262665
    }
  },
  {
    "ph": "s", "id": 262665, "pid": 76337, "tid": -914061504, "ts": 1716454225603881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225658656, "dur": 165,
    "args": {
      "External id": 262667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262667,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262667, "pid": 5, "tid": 7, "ts": 1716454225658656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603895, "dur": 5,
    "args": {
      "External id": 262667,
      "cbid": 211, "correlation": 262667
    }
  },
  {
    "ph": "s", "id": 262667, "pid": 76337, "tid": -914061504, "ts": 1716454225603895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225658823, "dur": 1,
    "args": {
      "External id": 262669,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 262669,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 262669, "pid": 5, "tid": 7, "ts": 1716454225658823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603906, "dur": 7,
    "args": {
      "External id": 262669,
      "cbid": 51, "correlation": 262669
    }
  },
  {
    "ph": "s", "id": 262669, "pid": 76337, "tid": -914061504, "ts": 1716454225603906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225658827, "dur": 674,
    "args": {
      "External id": 262670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262670,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262670, "pid": 5, "tid": 7, "ts": 1716454225658827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603914, "dur": 7,
    "args": {
      "External id": 262670,
      "cbid": 211, "correlation": 262670
    }
  },
  {
    "ph": "s", "id": 262670, "pid": 76337, "tid": -914061504, "ts": 1716454225603914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225659502, "dur": 14,
    "args": {
      "External id": 262672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262672, "pid": 5, "tid": 7, "ts": 1716454225659502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603925, "dur": 5,
    "args": {
      "External id": 262672,
      "cbid": 211, "correlation": 262672
    }
  },
  {
    "ph": "s", "id": 262672, "pid": 76337, "tid": -914061504, "ts": 1716454225603925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225659518, "dur": 15,
    "args": {
      "External id": 262678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262678,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262678, "pid": 5, "tid": 7, "ts": 1716454225659518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225603953, "dur": 8,
    "args": {
      "External id": 262678,
      "cbid": 211, "correlation": 262678
    }
  },
  {
    "ph": "s", "id": 262678, "pid": 76337, "tid": -914061504, "ts": 1716454225603953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225659534, "dur": 3,
    "args": {
      "External id": 262686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262686,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 262686, "pid": 5, "tid": 7, "ts": 1716454225659534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604005, "dur": 10,
    "args": {
      "External id": 262686,
      "cbid": 211, "correlation": 262686
    }
  },
  {
    "ph": "s", "id": 262686, "pid": 76337, "tid": -914061504, "ts": 1716454225604005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604069, "dur": 1,
    "args": {
      "External id": 262702,
      "cbid": 251, "correlation": 262702
    }
  },
  {
    "ph": "f", "id": 262702, "pid": 76337, "tid": -914061504, "ts": 1716454225604069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604075, "dur": 0,
    "args": {
      "External id": 262704,
      "cbid": 251, "correlation": 262704
    }
  },
  {
    "ph": "f", "id": 262704, "pid": 76337, "tid": -914061504, "ts": 1716454225604075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225659538, "dur": 13,
    "args": {
      "External id": 262705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262705,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262705, "pid": 5, "tid": 7, "ts": 1716454225659538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604077, "dur": 12,
    "args": {
      "External id": 262705,
      "cbid": 211, "correlation": 262705
    }
  },
  {
    "ph": "s", "id": 262705, "pid": 76337, "tid": -914061504, "ts": 1716454225604077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225659553, "dur": 5,
    "args": {
      "External id": 262707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262707,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262707, "pid": 5, "tid": 7, "ts": 1716454225659553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604091, "dur": 5,
    "args": {
      "External id": 262707,
      "cbid": 211, "correlation": 262707
    }
  },
  {
    "ph": "s", "id": 262707, "pid": 76337, "tid": -914061504, "ts": 1716454225604091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225659560, "dur": 17,
    "args": {
      "External id": 262717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262717, "pid": 5, "tid": 7, "ts": 1716454225659560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604148, "dur": 11,
    "args": {
      "External id": 262717,
      "cbid": 211, "correlation": 262717
    }
  },
  {
    "ph": "s", "id": 262717, "pid": 76337, "tid": -914061504, "ts": 1716454225604148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225659578, "dur": 18,
    "args": {
      "External id": 262737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262737,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 262737, "pid": 5, "tid": 7, "ts": 1716454225659578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604213, "dur": 11,
    "args": {
      "External id": 262737,
      "cbid": 211, "correlation": 262737
    }
  },
  {
    "ph": "s", "id": 262737, "pid": 76337, "tid": -914061504, "ts": 1716454225604213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225659598, "dur": 4,
    "args": {
      "External id": 262749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262749,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 262749, "pid": 5, "tid": 7, "ts": 1716454225659598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604234, "dur": 6,
    "args": {
      "External id": 262749,
      "cbid": 211, "correlation": 262749
    }
  },
  {
    "ph": "s", "id": 262749, "pid": 76337, "tid": -914061504, "ts": 1716454225604234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225659603, "dur": 17,
    "args": {
      "External id": 262752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262752, "pid": 5, "tid": 7, "ts": 1716454225659603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604252, "dur": 6,
    "args": {
      "External id": 262752,
      "cbid": 211, "correlation": 262752
    }
  },
  {
    "ph": "s", "id": 262752, "pid": 76337, "tid": -914061504, "ts": 1716454225604252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225659621, "dur": 11,
    "args": {
      "External id": 262761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262761,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262761, "pid": 5, "tid": 7, "ts": 1716454225659621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604292, "dur": 10,
    "args": {
      "External id": 262761,
      "cbid": 211, "correlation": 262761
    }
  },
  {
    "ph": "s", "id": 262761, "pid": 76337, "tid": -914061504, "ts": 1716454225604292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604353, "dur": 0,
    "args": {
      "External id": 262771,
      "cbid": 317, "correlation": 262771
    }
  },
  {
    "ph": "f", "id": 262771, "pid": 76337, "tid": -914061504, "ts": 1716454225604353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604354, "dur": 0,
    "args": {
      "External id": 262772,
      "cbid": 203, "correlation": 262772
    }
  },
  {
    "ph": "f", "id": 262772, "pid": 76337, "tid": -914061504, "ts": 1716454225604354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604355, "dur": 0,
    "args": {
      "External id": 262773,
      "cbid": 205, "correlation": 262773
    }
  },
  {
    "ph": "f", "id": 262773, "pid": 76337, "tid": -914061504, "ts": 1716454225604355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225659633, "dur": 11,
    "args": {
      "External id": 262777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262777,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262777, "pid": 5, "tid": 7, "ts": 1716454225659633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604368, "dur": 12,
    "args": {
      "External id": 262777,
      "cbid": 211, "correlation": 262777
    }
  },
  {
    "ph": "s", "id": 262777, "pid": 76337, "tid": -914061504, "ts": 1716454225604368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225659646, "dur": 166,
    "args": {
      "External id": 262779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262779,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262779, "pid": 5, "tid": 7, "ts": 1716454225659646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604383, "dur": 6,
    "args": {
      "External id": 262779,
      "cbid": 211, "correlation": 262779
    }
  },
  {
    "ph": "s", "id": 262779, "pid": 76337, "tid": -914061504, "ts": 1716454225604383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225659814, "dur": 1,
    "args": {
      "External id": 262781,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 262781,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 262781, "pid": 5, "tid": 7, "ts": 1716454225659814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604395, "dur": 6,
    "args": {
      "External id": 262781,
      "cbid": 51, "correlation": 262781
    }
  },
  {
    "ph": "s", "id": 262781, "pid": 76337, "tid": -914061504, "ts": 1716454225604395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225659818, "dur": 660,
    "args": {
      "External id": 262782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262782,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262782, "pid": 5, "tid": 7, "ts": 1716454225659818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604402, "dur": 6,
    "args": {
      "External id": 262782,
      "cbid": 211, "correlation": 262782
    }
  },
  {
    "ph": "s", "id": 262782, "pid": 76337, "tid": -914061504, "ts": 1716454225604402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225660479, "dur": 13,
    "args": {
      "External id": 262784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262784,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262784, "pid": 5, "tid": 7, "ts": 1716454225660479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604414, "dur": 5,
    "args": {
      "External id": 262784,
      "cbid": 211, "correlation": 262784
    }
  },
  {
    "ph": "s", "id": 262784, "pid": 76337, "tid": -914061504, "ts": 1716454225604414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225660493, "dur": 15,
    "args": {
      "External id": 262790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262790, "pid": 5, "tid": 7, "ts": 1716454225660493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604442, "dur": 8,
    "args": {
      "External id": 262790,
      "cbid": 211, "correlation": 262790
    }
  },
  {
    "ph": "s", "id": 262790, "pid": 76337, "tid": -914061504, "ts": 1716454225604442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225660510, "dur": 12,
    "args": {
      "External id": 262798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262798,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262798, "pid": 5, "tid": 7, "ts": 1716454225660510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604474, "dur": 9,
    "args": {
      "External id": 262798,
      "cbid": 211, "correlation": 262798
    }
  },
  {
    "ph": "s", "id": 262798, "pid": 76337, "tid": -914061504, "ts": 1716454225604474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225660523, "dur": 10,
    "args": {
      "External id": 262806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262806,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262806, "pid": 5, "tid": 7, "ts": 1716454225660523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604503, "dur": 8,
    "args": {
      "External id": 262806,
      "cbid": 211, "correlation": 262806
    }
  },
  {
    "ph": "s", "id": 262806, "pid": 76337, "tid": -914061504, "ts": 1716454225604503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225660534, "dur": 19,
    "args": {
      "External id": 262826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262826,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 262826, "pid": 5, "tid": 7, "ts": 1716454225660534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604581, "dur": 13,
    "args": {
      "External id": 262826,
      "cbid": 211, "correlation": 262826
    }
  },
  {
    "ph": "s", "id": 262826, "pid": 76337, "tid": -914061504, "ts": 1716454225604581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225660555, "dur": 4,
    "args": {
      "External id": 262838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262838,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 262838, "pid": 5, "tid": 7, "ts": 1716454225660555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604603, "dur": 6,
    "args": {
      "External id": 262838,
      "cbid": 211, "correlation": 262838
    }
  },
  {
    "ph": "s", "id": 262838, "pid": 76337, "tid": -914061504, "ts": 1716454225604603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225660560, "dur": 16,
    "args": {
      "External id": 262841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262841,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262841, "pid": 5, "tid": 7, "ts": 1716454225660560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604621, "dur": 8,
    "args": {
      "External id": 262841,
      "cbid": 211, "correlation": 262841
    }
  },
  {
    "ph": "s", "id": 262841, "pid": 76337, "tid": -914061504, "ts": 1716454225604621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604678, "dur": 0,
    "args": {
      "External id": 262852,
      "cbid": 317, "correlation": 262852
    }
  },
  {
    "ph": "f", "id": 262852, "pid": 76337, "tid": -914061504, "ts": 1716454225604678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604679, "dur": 0,
    "args": {
      "External id": 262853,
      "cbid": 203, "correlation": 262853
    }
  },
  {
    "ph": "f", "id": 262853, "pid": 76337, "tid": -914061504, "ts": 1716454225604679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604680, "dur": 0,
    "args": {
      "External id": 262854,
      "cbid": 205, "correlation": 262854
    }
  },
  {
    "ph": "f", "id": 262854, "pid": 76337, "tid": -914061504, "ts": 1716454225604680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225660578, "dur": 11,
    "args": {
      "External id": 262858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262858,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262858, "pid": 5, "tid": 7, "ts": 1716454225660578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604694, "dur": 12,
    "args": {
      "External id": 262858,
      "cbid": 211, "correlation": 262858
    }
  },
  {
    "ph": "s", "id": 262858, "pid": 76337, "tid": -914061504, "ts": 1716454225604694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225660590, "dur": 4,
    "args": {
      "External id": 262860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262860,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 262860, "pid": 5, "tid": 7, "ts": 1716454225660590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604710, "dur": 6,
    "args": {
      "External id": 262860,
      "cbid": 211, "correlation": 262860
    }
  },
  {
    "ph": "s", "id": 262860, "pid": 76337, "tid": -914061504, "ts": 1716454225604710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604719, "dur": 0,
    "args": {
      "External id": 262861,
      "cbid": 51, "correlation": 262861
    }
  },
  {
    "ph": "s", "id": 262861, "pid": 76337, "tid": -914061504, "ts": 1716454225604719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225660596, "dur": 96,
    "args": {
      "External id": 262862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262862,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 262862, "pid": 5, "tid": 7, "ts": 1716454225660596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604719, "dur": 5,
    "args": {
      "External id": 262862,
      "cbid": 211, "correlation": 262862
    }
  },
  {
    "ph": "s", "id": 262862, "pid": 76337, "tid": -914061504, "ts": 1716454225604719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225660693, "dur": 16,
    "args": {
      "External id": 262867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262867, "pid": 5, "tid": 7, "ts": 1716454225660693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604746, "dur": 9,
    "args": {
      "External id": 262867,
      "cbid": 211, "correlation": 262867
    }
  },
  {
    "ph": "s", "id": 262867, "pid": 76337, "tid": -914061504, "ts": 1716454225604746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225660711, "dur": 86,
    "args": {
      "External id": 262876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262876,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262876, "pid": 5, "tid": 7, "ts": 1716454225660711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604827, "dur": 15,
    "args": {
      "External id": 262876,
      "cbid": 211, "correlation": 262876
    }
  },
  {
    "ph": "s", "id": 262876, "pid": 76337, "tid": -914061504, "ts": 1716454225604827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225660798, "dur": 31,
    "args": {
      "External id": 262898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262898,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262898, "pid": 5, "tid": 7, "ts": 1716454225660798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604884, "dur": 11,
    "args": {
      "External id": 262898,
      "cbid": 211, "correlation": 262898
    }
  },
  {
    "ph": "s", "id": 262898, "pid": 76337, "tid": -914061504, "ts": 1716454225604884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604971, "dur": 8,
    "args": {
      "External id": 262909,
      "cbid": 251, "correlation": 262909
    }
  },
  {
    "ph": "f", "id": 262909, "pid": 76337, "tid": -914061504, "ts": 1716454225604971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225660829, "dur": 168,
    "args": {
      "External id": 262910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262910,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262910, "pid": 5, "tid": 7, "ts": 1716454225660829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225604984, "dur": 15,
    "args": {
      "External id": 262910,
      "cbid": 211, "correlation": 262910
    }
  },
  {
    "ph": "s", "id": 262910, "pid": 76337, "tid": -914061504, "ts": 1716454225604984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605057, "dur": 1,
    "args": {
      "External id": 262921,
      "cbid": 251, "correlation": 262921
    }
  },
  {
    "ph": "f", "id": 262921, "pid": 76337, "tid": -914061504, "ts": 1716454225605057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225660999, "dur": 160,
    "args": {
      "External id": 262922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262922,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262922, "pid": 5, "tid": 7, "ts": 1716454225660999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605061, "dur": 11,
    "args": {
      "External id": 262922,
      "cbid": 211, "correlation": 262922
    }
  },
  {
    "ph": "s", "id": 262922, "pid": 76337, "tid": -914061504, "ts": 1716454225605061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605125, "dur": 1,
    "args": {
      "External id": 262933,
      "cbid": 251, "correlation": 262933
    }
  },
  {
    "ph": "f", "id": 262933, "pid": 76337, "tid": -914061504, "ts": 1716454225605125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225661161, "dur": 161,
    "args": {
      "External id": 262934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262934,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262934, "pid": 5, "tid": 7, "ts": 1716454225661161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605129, "dur": 11,
    "args": {
      "External id": 262934,
      "cbid": 211, "correlation": 262934
    }
  },
  {
    "ph": "s", "id": 262934, "pid": 76337, "tid": -914061504, "ts": 1716454225605129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225661323, "dur": 343,
    "args": {
      "External id": 262959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262959,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262959, "pid": 5, "tid": 7, "ts": 1716454225661323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605214, "dur": 12,
    "args": {
      "External id": 262959,
      "cbid": 211, "correlation": 262959
    }
  },
  {
    "ph": "s", "id": 262959, "pid": 76337, "tid": -914061504, "ts": 1716454225605214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605313, "dur": 1,
    "args": {
      "External id": 262977,
      "cbid": 251, "correlation": 262977
    }
  },
  {
    "ph": "f", "id": 262977, "pid": 76337, "tid": -914061504, "ts": 1716454225605313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225661667, "dur": 145,
    "args": {
      "External id": 262979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262979,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 262979, "pid": 5, "tid": 7, "ts": 1716454225661667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605319, "dur": 13,
    "args": {
      "External id": 262979,
      "cbid": 211, "correlation": 262979
    }
  },
  {
    "ph": "s", "id": 262979, "pid": 76337, "tid": -914061504, "ts": 1716454225605319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225661814, "dur": 19,
    "args": {
      "External id": 262987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262987,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262987, "pid": 5, "tid": 7, "ts": 1716454225661814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605388, "dur": 13,
    "args": {
      "External id": 262987,
      "cbid": 211, "correlation": 262987
    }
  },
  {
    "ph": "s", "id": 262987, "pid": 76337, "tid": -914061504, "ts": 1716454225605388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225661834, "dur": 27,
    "args": {
      "External id": 262995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 262995,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 262995, "pid": 5, "tid": 7, "ts": 1716454225661834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605427, "dur": 8,
    "args": {
      "External id": 262995,
      "cbid": 211, "correlation": 262995
    }
  },
  {
    "ph": "s", "id": 262995, "pid": 76337, "tid": -914061504, "ts": 1716454225605427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225661863, "dur": 19,
    "args": {
      "External id": 263006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263006, "pid": 5, "tid": 7, "ts": 1716454225661863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605498, "dur": 12,
    "args": {
      "External id": 263006,
      "cbid": 211, "correlation": 263006
    }
  },
  {
    "ph": "s", "id": 263006, "pid": 76337, "tid": -914061504, "ts": 1716454225605498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225661883, "dur": 16,
    "args": {
      "External id": 263028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263028,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263028, "pid": 5, "tid": 7, "ts": 1716454225661883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605529, "dur": 7,
    "args": {
      "External id": 263028,
      "cbid": 211, "correlation": 263028
    }
  },
  {
    "ph": "s", "id": 263028, "pid": 76337, "tid": -914061504, "ts": 1716454225605529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605613, "dur": 1,
    "args": {
      "External id": 263039,
      "cbid": 251, "correlation": 263039
    }
  },
  {
    "ph": "f", "id": 263039, "pid": 76337, "tid": -914061504, "ts": 1716454225605613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225661901, "dur": 92,
    "args": {
      "External id": 263040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263040,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 263040, "pid": 5, "tid": 7, "ts": 1716454225661901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605619, "dur": 13,
    "args": {
      "External id": 263040,
      "cbid": 211, "correlation": 263040
    }
  },
  {
    "ph": "s", "id": 263040, "pid": 76337, "tid": -914061504, "ts": 1716454225605619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605688, "dur": 1,
    "args": {
      "External id": 263051,
      "cbid": 251, "correlation": 263051
    }
  },
  {
    "ph": "f", "id": 263051, "pid": 76337, "tid": -914061504, "ts": 1716454225605688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605692, "dur": 0,
    "args": {
      "External id": 263052,
      "cbid": 251, "correlation": 263052
    }
  },
  {
    "ph": "f", "id": 263052, "pid": 76337, "tid": -914061504, "ts": 1716454225605692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225661994, "dur": 12,
    "args": {
      "External id": 263053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263053,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263053, "pid": 5, "tid": 7, "ts": 1716454225661994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605694, "dur": 13,
    "args": {
      "External id": 263053,
      "cbid": 211, "correlation": 263053
    }
  },
  {
    "ph": "s", "id": 263053, "pid": 76337, "tid": -914061504, "ts": 1716454225605694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225662007, "dur": 6,
    "args": {
      "External id": 263055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263055,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263055, "pid": 5, "tid": 7, "ts": 1716454225662007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605708, "dur": 6,
    "args": {
      "External id": 263055,
      "cbid": 211, "correlation": 263055
    }
  },
  {
    "ph": "s", "id": 263055, "pid": 76337, "tid": -914061504, "ts": 1716454225605708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605766, "dur": 1,
    "args": {
      "External id": 263066,
      "cbid": 251, "correlation": 263066
    }
  },
  {
    "ph": "f", "id": 263066, "pid": 76337, "tid": -914061504, "ts": 1716454225605766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605769, "dur": 0,
    "args": {
      "External id": 263067,
      "cbid": 251, "correlation": 263067
    }
  },
  {
    "ph": "f", "id": 263067, "pid": 76337, "tid": -914061504, "ts": 1716454225605769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225662014, "dur": 8,
    "args": {
      "External id": 263068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263068,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263068, "pid": 5, "tid": 7, "ts": 1716454225662014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605771, "dur": 12,
    "args": {
      "External id": 263068,
      "cbid": 211, "correlation": 263068
    }
  },
  {
    "ph": "s", "id": 263068, "pid": 76337, "tid": -914061504, "ts": 1716454225605771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225662024, "dur": 3,
    "args": {
      "External id": 263070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263070,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263070, "pid": 5, "tid": 7, "ts": 1716454225662024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605784, "dur": 5,
    "args": {
      "External id": 263070,
      "cbid": 211, "correlation": 263070
    }
  },
  {
    "ph": "s", "id": 263070, "pid": 76337, "tid": -914061504, "ts": 1716454225605784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225662028, "dur": 55,
    "args": {
      "External id": 263095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263095,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263095, "pid": 5, "tid": 7, "ts": 1716454225662028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605861, "dur": 12,
    "args": {
      "External id": 263095,
      "cbid": 211, "correlation": 263095
    }
  },
  {
    "ph": "s", "id": 263095, "pid": 76337, "tid": -914061504, "ts": 1716454225605861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605959, "dur": 1,
    "args": {
      "External id": 263113,
      "cbid": 251, "correlation": 263113
    }
  },
  {
    "ph": "f", "id": 263113, "pid": 76337, "tid": -914061504, "ts": 1716454225605959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225662084, "dur": 92,
    "args": {
      "External id": 263115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263115,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 263115, "pid": 5, "tid": 7, "ts": 1716454225662084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225605965, "dur": 21,
    "args": {
      "External id": 263115,
      "cbid": 211, "correlation": 263115
    }
  },
  {
    "ph": "s", "id": 263115, "pid": 76337, "tid": -914061504, "ts": 1716454225605965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225662178, "dur": 10,
    "args": {
      "External id": 263123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263123,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263123, "pid": 5, "tid": 7, "ts": 1716454225662178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606044, "dur": 13,
    "args": {
      "External id": 263123,
      "cbid": 211, "correlation": 263123
    }
  },
  {
    "ph": "s", "id": 263123, "pid": 76337, "tid": -914061504, "ts": 1716454225606044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225662189, "dur": 21,
    "args": {
      "External id": 263131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263131, "pid": 5, "tid": 7, "ts": 1716454225662189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606086, "dur": 9,
    "args": {
      "External id": 263131,
      "cbid": 211, "correlation": 263131
    }
  },
  {
    "ph": "s", "id": 263131, "pid": 76337, "tid": -914061504, "ts": 1716454225606086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225662211, "dur": 18,
    "args": {
      "External id": 263153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263153,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263153, "pid": 5, "tid": 7, "ts": 1716454225662211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606138, "dur": 10,
    "args": {
      "External id": 263153,
      "cbid": 211, "correlation": 263153
    }
  },
  {
    "ph": "s", "id": 263153, "pid": 76337, "tid": -914061504, "ts": 1716454225606138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606225, "dur": 1,
    "args": {
      "External id": 263169,
      "cbid": 251, "correlation": 263169
    }
  },
  {
    "ph": "f", "id": 263169, "pid": 76337, "tid": -914061504, "ts": 1716454225606225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606230, "dur": 0,
    "args": {
      "External id": 263171,
      "cbid": 251, "correlation": 263171
    }
  },
  {
    "ph": "f", "id": 263171, "pid": 76337, "tid": -914061504, "ts": 1716454225606230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225662231, "dur": 498,
    "args": {
      "External id": 263172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263172,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263172, "pid": 5, "tid": 7, "ts": 1716454225662231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606232, "dur": 13,
    "args": {
      "External id": 263172,
      "cbid": 211, "correlation": 263172
    }
  },
  {
    "ph": "s", "id": 263172, "pid": 76337, "tid": -914061504, "ts": 1716454225606232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225662730, "dur": 66,
    "args": {
      "External id": 263180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263180, "pid": 5, "tid": 7, "ts": 1716454225662730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606297, "dur": 13,
    "args": {
      "External id": 263180,
      "cbid": 211, "correlation": 263180
    }
  },
  {
    "ph": "s", "id": 263180, "pid": 76337, "tid": -914061504, "ts": 1716454225606297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225662797, "dur": 68,
    "args": {
      "External id": 263188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263188,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263188, "pid": 5, "tid": 7, "ts": 1716454225662797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606327, "dur": 8,
    "args": {
      "External id": 263188,
      "cbid": 211, "correlation": 263188
    }
  },
  {
    "ph": "s", "id": 263188, "pid": 76337, "tid": -914061504, "ts": 1716454225606327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606406, "dur": 1,
    "args": {
      "External id": 263204,
      "cbid": 251, "correlation": 263204
    }
  },
  {
    "ph": "f", "id": 263204, "pid": 76337, "tid": -914061504, "ts": 1716454225606406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225662868, "dur": 1,
    "args": {
      "External id": 263206,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263206,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 263206, "pid": 5, "tid": 7, "ts": 1716454225662868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606411, "dur": 9,
    "args": {
      "External id": 263206,
      "cbid": 51, "correlation": 263206
    }
  },
  {
    "ph": "s", "id": 263206, "pid": 76337, "tid": -914061504, "ts": 1716454225606411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225662871, "dur": 275,
    "args": {
      "External id": 263207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263207,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 263207, "pid": 5, "tid": 7, "ts": 1716454225662871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606422, "dur": 11,
    "args": {
      "External id": 263207,
      "cbid": 211, "correlation": 263207
    }
  },
  {
    "ph": "s", "id": 263207, "pid": 76337, "tid": -914061504, "ts": 1716454225606422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225663148, "dur": 13,
    "args": {
      "External id": 263215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263215,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263215, "pid": 5, "tid": 7, "ts": 1716454225663148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606465, "dur": 10,
    "args": {
      "External id": 263215,
      "cbid": 211, "correlation": 263215
    }
  },
  {
    "ph": "s", "id": 263215, "pid": 76337, "tid": -914061504, "ts": 1716454225606465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225663163, "dur": 38,
    "args": {
      "External id": 263226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263226, "pid": 5, "tid": 7, "ts": 1716454225663163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606532, "dur": 12,
    "args": {
      "External id": 263226,
      "cbid": 211, "correlation": 263226
    }
  },
  {
    "ph": "s", "id": 263226, "pid": 76337, "tid": -914061504, "ts": 1716454225606532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606596, "dur": 0,
    "args": {
      "External id": 263238,
      "cbid": 317, "correlation": 263238
    }
  },
  {
    "ph": "f", "id": 263238, "pid": 76337, "tid": -914061504, "ts": 1716454225606596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606597, "dur": 0,
    "args": {
      "External id": 263239,
      "cbid": 203, "correlation": 263239
    }
  },
  {
    "ph": "f", "id": 263239, "pid": 76337, "tid": -914061504, "ts": 1716454225606597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606598, "dur": 0,
    "args": {
      "External id": 263240,
      "cbid": 205, "correlation": 263240
    }
  },
  {
    "ph": "f", "id": 263240, "pid": 76337, "tid": -914061504, "ts": 1716454225606598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663202, "dur": 13,
    "args": {
      "External id": 263244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263244,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263244, "pid": 5, "tid": 7, "ts": 1716454225663202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606613, "dur": 12,
    "args": {
      "External id": 263244,
      "cbid": 211, "correlation": 263244
    }
  },
  {
    "ph": "s", "id": 263244, "pid": 76337, "tid": -914061504, "ts": 1716454225606613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225663216, "dur": 4,
    "args": {
      "External id": 263246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 263246, "pid": 5, "tid": 7, "ts": 1716454225663216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606630, "dur": 6,
    "args": {
      "External id": 263246,
      "cbid": 211, "correlation": 263246
    }
  },
  {
    "ph": "s", "id": 263246, "pid": 76337, "tid": -914061504, "ts": 1716454225606630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606638, "dur": 0,
    "args": {
      "External id": 263247,
      "cbid": 51, "correlation": 263247
    }
  },
  {
    "ph": "s", "id": 263247, "pid": 76337, "tid": -914061504, "ts": 1716454225606638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225663222, "dur": 99,
    "args": {
      "External id": 263248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263248,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 263248, "pid": 5, "tid": 7, "ts": 1716454225663222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606639, "dur": 5,
    "args": {
      "External id": 263248,
      "cbid": 211, "correlation": 263248
    }
  },
  {
    "ph": "s", "id": 263248, "pid": 76337, "tid": -914061504, "ts": 1716454225606639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225663322, "dur": 16,
    "args": {
      "External id": 263253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263253, "pid": 5, "tid": 7, "ts": 1716454225663322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606666, "dur": 8,
    "args": {
      "External id": 263253,
      "cbid": 211, "correlation": 263253
    }
  },
  {
    "ph": "s", "id": 263253, "pid": 76337, "tid": -914061504, "ts": 1716454225606666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225663340, "dur": 11,
    "args": {
      "External id": 263261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263261,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263261, "pid": 5, "tid": 7, "ts": 1716454225663340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606697, "dur": 9,
    "args": {
      "External id": 263261,
      "cbid": 211, "correlation": 263261
    }
  },
  {
    "ph": "s", "id": 263261, "pid": 76337, "tid": -914061504, "ts": 1716454225606697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606766, "dur": 0,
    "args": {
      "External id": 263271,
      "cbid": 317, "correlation": 263271
    }
  },
  {
    "ph": "f", "id": 263271, "pid": 76337, "tid": -914061504, "ts": 1716454225606766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606767, "dur": 0,
    "args": {
      "External id": 263272,
      "cbid": 203, "correlation": 263272
    }
  },
  {
    "ph": "f", "id": 263272, "pid": 76337, "tid": -914061504, "ts": 1716454225606767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606768, "dur": 0,
    "args": {
      "External id": 263273,
      "cbid": 205, "correlation": 263273
    }
  },
  {
    "ph": "f", "id": 263273, "pid": 76337, "tid": -914061504, "ts": 1716454225606768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663353, "dur": 13,
    "args": {
      "External id": 263277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263277,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263277, "pid": 5, "tid": 7, "ts": 1716454225663353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606782, "dur": 12,
    "args": {
      "External id": 263277,
      "cbid": 211, "correlation": 263277
    }
  },
  {
    "ph": "s", "id": 263277, "pid": 76337, "tid": -914061504, "ts": 1716454225606782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663366, "dur": 166,
    "args": {
      "External id": 263279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263279,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263279, "pid": 5, "tid": 7, "ts": 1716454225663366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606797, "dur": 6,
    "args": {
      "External id": 263279,
      "cbid": 211, "correlation": 263279
    }
  },
  {
    "ph": "s", "id": 263279, "pid": 76337, "tid": -914061504, "ts": 1716454225606797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225663535, "dur": 1,
    "args": {
      "External id": 263281,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263281,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 263281, "pid": 5, "tid": 7, "ts": 1716454225663535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606809, "dur": 6,
    "args": {
      "External id": 263281,
      "cbid": 51, "correlation": 263281
    }
  },
  {
    "ph": "s", "id": 263281, "pid": 76337, "tid": -914061504, "ts": 1716454225606809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225663538, "dur": 202,
    "args": {
      "External id": 263282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263282,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 263282, "pid": 5, "tid": 7, "ts": 1716454225663538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606817, "dur": 8,
    "args": {
      "External id": 263282,
      "cbid": 211, "correlation": 263282
    }
  },
  {
    "ph": "s", "id": 263282, "pid": 76337, "tid": -914061504, "ts": 1716454225606817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663742, "dur": 6,
    "args": {
      "External id": 263284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263284,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263284, "pid": 5, "tid": 7, "ts": 1716454225663742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606829, "dur": 5,
    "args": {
      "External id": 263284,
      "cbid": 211, "correlation": 263284
    }
  },
  {
    "ph": "s", "id": 263284, "pid": 76337, "tid": -914061504, "ts": 1716454225606829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225663749, "dur": 6,
    "args": {
      "External id": 263290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263290, "pid": 5, "tid": 7, "ts": 1716454225663749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606858, "dur": 8,
    "args": {
      "External id": 263290,
      "cbid": 211, "correlation": 263290
    }
  },
  {
    "ph": "s", "id": 263290, "pid": 76337, "tid": -914061504, "ts": 1716454225606858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663757, "dur": 11,
    "args": {
      "External id": 263310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263310,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263310, "pid": 5, "tid": 7, "ts": 1716454225663757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606948, "dur": 13,
    "args": {
      "External id": 263310,
      "cbid": 211, "correlation": 263310
    }
  },
  {
    "ph": "s", "id": 263310, "pid": 76337, "tid": -914061504, "ts": 1716454225606948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225663769, "dur": 5,
    "args": {
      "External id": 263322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263322,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263322, "pid": 5, "tid": 7, "ts": 1716454225663769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225606972, "dur": 15,
    "args": {
      "External id": 263322,
      "cbid": 211, "correlation": 263322
    }
  },
  {
    "ph": "s", "id": 263322, "pid": 76337, "tid": -914061504, "ts": 1716454225606972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225663775, "dur": 8,
    "args": {
      "External id": 263325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263325, "pid": 5, "tid": 7, "ts": 1716454225663775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607000, "dur": 8,
    "args": {
      "External id": 263325,
      "cbid": 211, "correlation": 263325
    }
  },
  {
    "ph": "s", "id": 263325, "pid": 76337, "tid": -914061504, "ts": 1716454225607000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225663785, "dur": 5,
    "args": {
      "External id": 263334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263334,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263334, "pid": 5, "tid": 7, "ts": 1716454225663785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607041, "dur": 10,
    "args": {
      "External id": 263334,
      "cbid": 211, "correlation": 263334
    }
  },
  {
    "ph": "s", "id": 263334, "pid": 76337, "tid": -914061504, "ts": 1716454225607041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607094, "dur": 0,
    "args": {
      "External id": 263344,
      "cbid": 317, "correlation": 263344
    }
  },
  {
    "ph": "f", "id": 263344, "pid": 76337, "tid": -914061504, "ts": 1716454225607094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607095, "dur": 0,
    "args": {
      "External id": 263345,
      "cbid": 203, "correlation": 263345
    }
  },
  {
    "ph": "f", "id": 263345, "pid": 76337, "tid": -914061504, "ts": 1716454225607095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607096, "dur": 0,
    "args": {
      "External id": 263346,
      "cbid": 205, "correlation": 263346
    }
  },
  {
    "ph": "f", "id": 263346, "pid": 76337, "tid": -914061504, "ts": 1716454225607096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663792, "dur": 5,
    "args": {
      "External id": 263350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263350, "pid": 5, "tid": 7, "ts": 1716454225663792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607110, "dur": 12,
    "args": {
      "External id": 263350,
      "cbid": 211, "correlation": 263350
    }
  },
  {
    "ph": "s", "id": 263350, "pid": 76337, "tid": -914061504, "ts": 1716454225607110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225663798, "dur": 165,
    "args": {
      "External id": 263352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263352,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263352, "pid": 5, "tid": 7, "ts": 1716454225663798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607125, "dur": 5,
    "args": {
      "External id": 263352,
      "cbid": 211, "correlation": 263352
    }
  },
  {
    "ph": "s", "id": 263352, "pid": 76337, "tid": -914061504, "ts": 1716454225607125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225663965, "dur": 1,
    "args": {
      "External id": 263354,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263354,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 263354, "pid": 5, "tid": 7, "ts": 1716454225663965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607135, "dur": 6,
    "args": {
      "External id": 263354,
      "cbid": 51, "correlation": 263354
    }
  },
  {
    "ph": "s", "id": 263354, "pid": 76337, "tid": -914061504, "ts": 1716454225607135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225663969, "dur": 276,
    "args": {
      "External id": 263355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263355,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263355, "pid": 5, "tid": 7, "ts": 1716454225663969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607143, "dur": 6,
    "args": {
      "External id": 263355,
      "cbid": 211, "correlation": 263355
    }
  },
  {
    "ph": "s", "id": 263355, "pid": 76337, "tid": -914061504, "ts": 1716454225607143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664246, "dur": 6,
    "args": {
      "External id": 263357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263357,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263357, "pid": 5, "tid": 7, "ts": 1716454225664246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607154, "dur": 6,
    "args": {
      "External id": 263357,
      "cbid": 211, "correlation": 263357
    }
  },
  {
    "ph": "s", "id": 263357, "pid": 76337, "tid": -914061504, "ts": 1716454225607154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225664253, "dur": 6,
    "args": {
      "External id": 263363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263363,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263363, "pid": 5, "tid": 7, "ts": 1716454225664253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607183, "dur": 8,
    "args": {
      "External id": 263363,
      "cbid": 211, "correlation": 263363
    }
  },
  {
    "ph": "s", "id": 263363, "pid": 76337, "tid": -914061504, "ts": 1716454225607183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225664261, "dur": 3,
    "args": {
      "External id": 263371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263371,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 263371, "pid": 5, "tid": 7, "ts": 1716454225664261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607227, "dur": 9,
    "args": {
      "External id": 263371,
      "cbid": 211, "correlation": 263371
    }
  },
  {
    "ph": "s", "id": 263371, "pid": 76337, "tid": -914061504, "ts": 1716454225607227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607290, "dur": 1,
    "args": {
      "External id": 263387,
      "cbid": 251, "correlation": 263387
    }
  },
  {
    "ph": "f", "id": 263387, "pid": 76337, "tid": -914061504, "ts": 1716454225607290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607295, "dur": 0,
    "args": {
      "External id": 263389,
      "cbid": 251, "correlation": 263389
    }
  },
  {
    "ph": "f", "id": 263389, "pid": 76337, "tid": -914061504, "ts": 1716454225607295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225664266, "dur": 13,
    "args": {
      "External id": 263390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263390,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263390, "pid": 5, "tid": 7, "ts": 1716454225664266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607297, "dur": 13,
    "args": {
      "External id": 263390,
      "cbid": 211, "correlation": 263390
    }
  },
  {
    "ph": "s", "id": 263390, "pid": 76337, "tid": -914061504, "ts": 1716454225607297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225664280, "dur": 5,
    "args": {
      "External id": 263392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263392,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263392, "pid": 5, "tid": 7, "ts": 1716454225664280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607312, "dur": 5,
    "args": {
      "External id": 263392,
      "cbid": 211, "correlation": 263392
    }
  },
  {
    "ph": "s", "id": 263392, "pid": 76337, "tid": -914061504, "ts": 1716454225607312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225664287, "dur": 6,
    "args": {
      "External id": 263402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263402,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263402, "pid": 5, "tid": 7, "ts": 1716454225664287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607370, "dur": 12,
    "args": {
      "External id": 263402,
      "cbid": 211, "correlation": 263402
    }
  },
  {
    "ph": "s", "id": 263402, "pid": 76337, "tid": -914061504, "ts": 1716454225607370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664294, "dur": 10,
    "args": {
      "External id": 263422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263422,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263422, "pid": 5, "tid": 7, "ts": 1716454225664294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607436, "dur": 10,
    "args": {
      "External id": 263422,
      "cbid": 211, "correlation": 263422
    }
  },
  {
    "ph": "s", "id": 263422, "pid": 76337, "tid": -914061504, "ts": 1716454225607436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225664305, "dur": 4,
    "args": {
      "External id": 263434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263434,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263434, "pid": 5, "tid": 7, "ts": 1716454225664305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607456, "dur": 7,
    "args": {
      "External id": 263434,
      "cbid": 211, "correlation": 263434
    }
  },
  {
    "ph": "s", "id": 263434, "pid": 76337, "tid": -914061504, "ts": 1716454225607456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225664310, "dur": 7,
    "args": {
      "External id": 263437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263437, "pid": 5, "tid": 7, "ts": 1716454225664310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607475, "dur": 6,
    "args": {
      "External id": 263437,
      "cbid": 211, "correlation": 263437
    }
  },
  {
    "ph": "s", "id": 263437, "pid": 76337, "tid": -914061504, "ts": 1716454225607475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225664319, "dur": 5,
    "args": {
      "External id": 263446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263446,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263446, "pid": 5, "tid": 7, "ts": 1716454225664319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607515, "dur": 9,
    "args": {
      "External id": 263446,
      "cbid": 211, "correlation": 263446
    }
  },
  {
    "ph": "s", "id": 263446, "pid": 76337, "tid": -914061504, "ts": 1716454225607515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607577, "dur": 0,
    "args": {
      "External id": 263456,
      "cbid": 317, "correlation": 263456
    }
  },
  {
    "ph": "f", "id": 263456, "pid": 76337, "tid": -914061504, "ts": 1716454225607577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607578, "dur": 0,
    "args": {
      "External id": 263457,
      "cbid": 203, "correlation": 263457
    }
  },
  {
    "ph": "f", "id": 263457, "pid": 76337, "tid": -914061504, "ts": 1716454225607578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607579, "dur": 0,
    "args": {
      "External id": 263458,
      "cbid": 205, "correlation": 263458
    }
  },
  {
    "ph": "f", "id": 263458, "pid": 76337, "tid": -914061504, "ts": 1716454225607579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664325, "dur": 5,
    "args": {
      "External id": 263462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263462,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263462, "pid": 5, "tid": 7, "ts": 1716454225664325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607593, "dur": 12,
    "args": {
      "External id": 263462,
      "cbid": 211, "correlation": 263462
    }
  },
  {
    "ph": "s", "id": 263462, "pid": 76337, "tid": -914061504, "ts": 1716454225607593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664331, "dur": 165,
    "args": {
      "External id": 263464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263464,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263464, "pid": 5, "tid": 7, "ts": 1716454225664331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607607, "dur": 5,
    "args": {
      "External id": 263464,
      "cbid": 211, "correlation": 263464
    }
  },
  {
    "ph": "s", "id": 263464, "pid": 76337, "tid": -914061504, "ts": 1716454225607607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225664498, "dur": 1,
    "args": {
      "External id": 263466,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263466,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 263466, "pid": 5, "tid": 7, "ts": 1716454225664498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607619, "dur": 6,
    "args": {
      "External id": 263466,
      "cbid": 51, "correlation": 263466
    }
  },
  {
    "ph": "s", "id": 263466, "pid": 76337, "tid": -914061504, "ts": 1716454225607619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225664502, "dur": 264,
    "args": {
      "External id": 263467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263467,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263467, "pid": 5, "tid": 7, "ts": 1716454225664502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607626, "dur": 6,
    "args": {
      "External id": 263467,
      "cbid": 211, "correlation": 263467
    }
  },
  {
    "ph": "s", "id": 263467, "pid": 76337, "tid": -914061504, "ts": 1716454225607626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664767, "dur": 6,
    "args": {
      "External id": 263469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263469,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263469, "pid": 5, "tid": 7, "ts": 1716454225664767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607636, "dur": 5,
    "args": {
      "External id": 263469,
      "cbid": 211, "correlation": 263469
    }
  },
  {
    "ph": "s", "id": 263469, "pid": 76337, "tid": -914061504, "ts": 1716454225607636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225664774, "dur": 6,
    "args": {
      "External id": 263475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263475, "pid": 5, "tid": 7, "ts": 1716454225664774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607665, "dur": 9,
    "args": {
      "External id": 263475,
      "cbid": 211, "correlation": 263475
    }
  },
  {
    "ph": "s", "id": 263475, "pid": 76337, "tid": -914061504, "ts": 1716454225607665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225664782, "dur": 5,
    "args": {
      "External id": 263483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263483,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263483, "pid": 5, "tid": 7, "ts": 1716454225664782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607697, "dur": 8,
    "args": {
      "External id": 263483,
      "cbid": 211, "correlation": 263483
    }
  },
  {
    "ph": "s", "id": 263483, "pid": 76337, "tid": -914061504, "ts": 1716454225607697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225664788, "dur": 5,
    "args": {
      "External id": 263491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263491,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263491, "pid": 5, "tid": 7, "ts": 1716454225664788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607727, "dur": 8,
    "args": {
      "External id": 263491,
      "cbid": 211, "correlation": 263491
    }
  },
  {
    "ph": "s", "id": 263491, "pid": 76337, "tid": -914061504, "ts": 1716454225607727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664794, "dur": 10,
    "args": {
      "External id": 263511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263511,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263511, "pid": 5, "tid": 7, "ts": 1716454225664794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607800, "dur": 12,
    "args": {
      "External id": 263511,
      "cbid": 211, "correlation": 263511
    }
  },
  {
    "ph": "s", "id": 263511, "pid": 76337, "tid": -914061504, "ts": 1716454225607800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225664805, "dur": 4,
    "args": {
      "External id": 263523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263523,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263523, "pid": 5, "tid": 7, "ts": 1716454225664805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607823, "dur": 6,
    "args": {
      "External id": 263523,
      "cbid": 211, "correlation": 263523
    }
  },
  {
    "ph": "s", "id": 263523, "pid": 76337, "tid": -914061504, "ts": 1716454225607823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225664810, "dur": 7,
    "args": {
      "External id": 263526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263526, "pid": 5, "tid": 7, "ts": 1716454225664810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607840, "dur": 6,
    "args": {
      "External id": 263526,
      "cbid": 211, "correlation": 263526
    }
  },
  {
    "ph": "s", "id": 263526, "pid": 76337, "tid": -914061504, "ts": 1716454225607840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225664818, "dur": 5,
    "args": {
      "External id": 263535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263535,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263535, "pid": 5, "tid": 7, "ts": 1716454225664818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607878, "dur": 9,
    "args": {
      "External id": 263535,
      "cbid": 211, "correlation": 263535
    }
  },
  {
    "ph": "s", "id": 263535, "pid": 76337, "tid": -914061504, "ts": 1716454225607878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607929, "dur": 0,
    "args": {
      "External id": 263545,
      "cbid": 317, "correlation": 263545
    }
  },
  {
    "ph": "f", "id": 263545, "pid": 76337, "tid": -914061504, "ts": 1716454225607929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607930, "dur": 0,
    "args": {
      "External id": 263546,
      "cbid": 203, "correlation": 263546
    }
  },
  {
    "ph": "f", "id": 263546, "pid": 76337, "tid": -914061504, "ts": 1716454225607930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607931, "dur": 0,
    "args": {
      "External id": 263547,
      "cbid": 205, "correlation": 263547
    }
  },
  {
    "ph": "f", "id": 263547, "pid": 76337, "tid": -914061504, "ts": 1716454225607931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664824, "dur": 5,
    "args": {
      "External id": 263551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263551,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263551, "pid": 5, "tid": 7, "ts": 1716454225664824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607944, "dur": 11,
    "args": {
      "External id": 263551,
      "cbid": 211, "correlation": 263551
    }
  },
  {
    "ph": "s", "id": 263551, "pid": 76337, "tid": -914061504, "ts": 1716454225607944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225664831, "dur": 165,
    "args": {
      "External id": 263553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263553, "pid": 5, "tid": 7, "ts": 1716454225664831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607958, "dur": 5,
    "args": {
      "External id": 263553,
      "cbid": 211, "correlation": 263553
    }
  },
  {
    "ph": "s", "id": 263553, "pid": 76337, "tid": -914061504, "ts": 1716454225607958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225664997, "dur": 1,
    "args": {
      "External id": 263555,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263555,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14150943396226415
    }
  },
  {
    "ph": "f", "id": 263555, "pid": 5, "tid": 7, "ts": 1716454225664997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607969, "dur": 15,
    "args": {
      "External id": 263555,
      "cbid": 51, "correlation": 263555
    }
  },
  {
    "ph": "s", "id": 263555, "pid": 76337, "tid": -914061504, "ts": 1716454225607969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225665001, "dur": 263,
    "args": {
      "External id": 263556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263556,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263556, "pid": 5, "tid": 7, "ts": 1716454225665001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607984, "dur": 6,
    "args": {
      "External id": 263556,
      "cbid": 211, "correlation": 263556
    }
  },
  {
    "ph": "s", "id": 263556, "pid": 76337, "tid": -914061504, "ts": 1716454225607984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665266, "dur": 6,
    "args": {
      "External id": 263558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263558, "pid": 5, "tid": 7, "ts": 1716454225665266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225607994, "dur": 5,
    "args": {
      "External id": 263558,
      "cbid": 211, "correlation": 263558
    }
  },
  {
    "ph": "s", "id": 263558, "pid": 76337, "tid": -914061504, "ts": 1716454225607994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225665273, "dur": 6,
    "args": {
      "External id": 263564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263564, "pid": 5, "tid": 7, "ts": 1716454225665273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608023, "dur": 9,
    "args": {
      "External id": 263564,
      "cbid": 211, "correlation": 263564
    }
  },
  {
    "ph": "s", "id": 263564, "pid": 76337, "tid": -914061504, "ts": 1716454225608023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225665281, "dur": 3,
    "args": {
      "External id": 263572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263572,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 263572, "pid": 5, "tid": 7, "ts": 1716454225665281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608067, "dur": 10,
    "args": {
      "External id": 263572,
      "cbid": 211, "correlation": 263572
    }
  },
  {
    "ph": "s", "id": 263572, "pid": 76337, "tid": -914061504, "ts": 1716454225608067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608130, "dur": 1,
    "args": {
      "External id": 263588,
      "cbid": 251, "correlation": 263588
    }
  },
  {
    "ph": "f", "id": 263588, "pid": 76337, "tid": -914061504, "ts": 1716454225608130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608135, "dur": 0,
    "args": {
      "External id": 263590,
      "cbid": 251, "correlation": 263590
    }
  },
  {
    "ph": "f", "id": 263590, "pid": 76337, "tid": -914061504, "ts": 1716454225608135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225665285, "dur": 11,
    "args": {
      "External id": 263591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263591,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263591, "pid": 5, "tid": 7, "ts": 1716454225665285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608137, "dur": 11,
    "args": {
      "External id": 263591,
      "cbid": 211, "correlation": 263591
    }
  },
  {
    "ph": "s", "id": 263591, "pid": 76337, "tid": -914061504, "ts": 1716454225608137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225665297, "dur": 4,
    "args": {
      "External id": 263593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263593,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263593, "pid": 5, "tid": 7, "ts": 1716454225665297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608149, "dur": 5,
    "args": {
      "External id": 263593,
      "cbid": 211, "correlation": 263593
    }
  },
  {
    "ph": "s", "id": 263593, "pid": 76337, "tid": -914061504, "ts": 1716454225608149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225665302, "dur": 6,
    "args": {
      "External id": 263603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263603, "pid": 5, "tid": 7, "ts": 1716454225665302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608205, "dur": 12,
    "args": {
      "External id": 263603,
      "cbid": 211, "correlation": 263603
    }
  },
  {
    "ph": "s", "id": 263603, "pid": 76337, "tid": -914061504, "ts": 1716454225608205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665309, "dur": 10,
    "args": {
      "External id": 263623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263623,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263623, "pid": 5, "tid": 7, "ts": 1716454225665309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608269, "dur": 11,
    "args": {
      "External id": 263623,
      "cbid": 211, "correlation": 263623
    }
  },
  {
    "ph": "s", "id": 263623, "pid": 76337, "tid": -914061504, "ts": 1716454225608269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225665320, "dur": 4,
    "args": {
      "External id": 263635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263635,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263635, "pid": 5, "tid": 7, "ts": 1716454225665320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608291, "dur": 6,
    "args": {
      "External id": 263635,
      "cbid": 211, "correlation": 263635
    }
  },
  {
    "ph": "s", "id": 263635, "pid": 76337, "tid": -914061504, "ts": 1716454225608291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225665325, "dur": 7,
    "args": {
      "External id": 263638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263638, "pid": 5, "tid": 7, "ts": 1716454225665325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608309, "dur": 6,
    "args": {
      "External id": 263638,
      "cbid": 211, "correlation": 263638
    }
  },
  {
    "ph": "s", "id": 263638, "pid": 76337, "tid": -914061504, "ts": 1716454225608309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225665333, "dur": 5,
    "args": {
      "External id": 263647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263647,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263647, "pid": 5, "tid": 7, "ts": 1716454225665333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608349, "dur": 9,
    "args": {
      "External id": 263647,
      "cbid": 211, "correlation": 263647
    }
  },
  {
    "ph": "s", "id": 263647, "pid": 76337, "tid": -914061504, "ts": 1716454225608349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608409, "dur": 0,
    "args": {
      "External id": 263657,
      "cbid": 317, "correlation": 263657
    }
  },
  {
    "ph": "f", "id": 263657, "pid": 76337, "tid": -914061504, "ts": 1716454225608409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608410, "dur": 0,
    "args": {
      "External id": 263658,
      "cbid": 203, "correlation": 263658
    }
  },
  {
    "ph": "f", "id": 263658, "pid": 76337, "tid": -914061504, "ts": 1716454225608410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608411, "dur": 0,
    "args": {
      "External id": 263659,
      "cbid": 205, "correlation": 263659
    }
  },
  {
    "ph": "f", "id": 263659, "pid": 76337, "tid": -914061504, "ts": 1716454225608411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665339, "dur": 5,
    "args": {
      "External id": 263663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263663,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263663, "pid": 5, "tid": 7, "ts": 1716454225665339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608425, "dur": 13,
    "args": {
      "External id": 263663,
      "cbid": 211, "correlation": 263663
    }
  },
  {
    "ph": "s", "id": 263663, "pid": 76337, "tid": -914061504, "ts": 1716454225608425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665346, "dur": 165,
    "args": {
      "External id": 263665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263665, "pid": 5, "tid": 7, "ts": 1716454225665346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608440, "dur": 5,
    "args": {
      "External id": 263665,
      "cbid": 211, "correlation": 263665
    }
  },
  {
    "ph": "s", "id": 263665, "pid": 76337, "tid": -914061504, "ts": 1716454225608440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225665513, "dur": 1,
    "args": {
      "External id": 263667,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263667,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 263667, "pid": 5, "tid": 7, "ts": 1716454225665513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608450, "dur": 6,
    "args": {
      "External id": 263667,
      "cbid": 51, "correlation": 263667
    }
  },
  {
    "ph": "s", "id": 263667, "pid": 76337, "tid": -914061504, "ts": 1716454225608450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225665517, "dur": 264,
    "args": {
      "External id": 263668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263668,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263668, "pid": 5, "tid": 7, "ts": 1716454225665517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608458, "dur": 6,
    "args": {
      "External id": 263668,
      "cbid": 211, "correlation": 263668
    }
  },
  {
    "ph": "s", "id": 263668, "pid": 76337, "tid": -914061504, "ts": 1716454225608458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665782, "dur": 6,
    "args": {
      "External id": 263670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263670, "pid": 5, "tid": 7, "ts": 1716454225665782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608467, "dur": 5,
    "args": {
      "External id": 263670,
      "cbid": 211, "correlation": 263670
    }
  },
  {
    "ph": "s", "id": 263670, "pid": 76337, "tid": -914061504, "ts": 1716454225608467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225665790, "dur": 6,
    "args": {
      "External id": 263676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263676, "pid": 5, "tid": 7, "ts": 1716454225665790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608496, "dur": 9,
    "args": {
      "External id": 263676,
      "cbid": 211, "correlation": 263676
    }
  },
  {
    "ph": "s", "id": 263676, "pid": 76337, "tid": -914061504, "ts": 1716454225608496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225665797, "dur": 5,
    "args": {
      "External id": 263684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263684,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263684, "pid": 5, "tid": 7, "ts": 1716454225665797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608530, "dur": 9,
    "args": {
      "External id": 263684,
      "cbid": 211, "correlation": 263684
    }
  },
  {
    "ph": "s", "id": 263684, "pid": 76337, "tid": -914061504, "ts": 1716454225608530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225665804, "dur": 4,
    "args": {
      "External id": 263692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263692,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263692, "pid": 5, "tid": 7, "ts": 1716454225665804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608559, "dur": 8,
    "args": {
      "External id": 263692,
      "cbid": 211, "correlation": 263692
    }
  },
  {
    "ph": "s", "id": 263692, "pid": 76337, "tid": -914061504, "ts": 1716454225608559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665809, "dur": 10,
    "args": {
      "External id": 263712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263712,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263712, "pid": 5, "tid": 7, "ts": 1716454225665809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608660, "dur": 12,
    "args": {
      "External id": 263712,
      "cbid": 211, "correlation": 263712
    }
  },
  {
    "ph": "s", "id": 263712, "pid": 76337, "tid": -914061504, "ts": 1716454225608660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225665821, "dur": 4,
    "args": {
      "External id": 263724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263724,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263724, "pid": 5, "tid": 7, "ts": 1716454225665821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608683, "dur": 7,
    "args": {
      "External id": 263724,
      "cbid": 211, "correlation": 263724
    }
  },
  {
    "ph": "s", "id": 263724, "pid": 76337, "tid": -914061504, "ts": 1716454225608683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225665826, "dur": 7,
    "args": {
      "External id": 263727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263727, "pid": 5, "tid": 7, "ts": 1716454225665826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608701, "dur": 6,
    "args": {
      "External id": 263727,
      "cbid": 211, "correlation": 263727
    }
  },
  {
    "ph": "s", "id": 263727, "pid": 76337, "tid": -914061504, "ts": 1716454225608701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225665834, "dur": 5,
    "args": {
      "External id": 263736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263736,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263736, "pid": 5, "tid": 7, "ts": 1716454225665834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608740, "dur": 9,
    "args": {
      "External id": 263736,
      "cbid": 211, "correlation": 263736
    }
  },
  {
    "ph": "s", "id": 263736, "pid": 76337, "tid": -914061504, "ts": 1716454225608740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608793, "dur": 0,
    "args": {
      "External id": 263746,
      "cbid": 317, "correlation": 263746
    }
  },
  {
    "ph": "f", "id": 263746, "pid": 76337, "tid": -914061504, "ts": 1716454225608793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608793, "dur": 0,
    "args": {
      "External id": 263747,
      "cbid": 203, "correlation": 263747
    }
  },
  {
    "ph": "f", "id": 263747, "pid": 76337, "tid": -914061504, "ts": 1716454225608793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608794, "dur": 0,
    "args": {
      "External id": 263748,
      "cbid": 205, "correlation": 263748
    }
  },
  {
    "ph": "f", "id": 263748, "pid": 76337, "tid": -914061504, "ts": 1716454225608794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665840, "dur": 5,
    "args": {
      "External id": 263752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263752,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263752, "pid": 5, "tid": 7, "ts": 1716454225665840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608807, "dur": 11,
    "args": {
      "External id": 263752,
      "cbid": 211, "correlation": 263752
    }
  },
  {
    "ph": "s", "id": 263752, "pid": 76337, "tid": -914061504, "ts": 1716454225608807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225665846, "dur": 165,
    "args": {
      "External id": 263754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263754,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263754, "pid": 5, "tid": 7, "ts": 1716454225665846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608821, "dur": 5,
    "args": {
      "External id": 263754,
      "cbid": 211, "correlation": 263754
    }
  },
  {
    "ph": "s", "id": 263754, "pid": 76337, "tid": -914061504, "ts": 1716454225608821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225666014, "dur": 1,
    "args": {
      "External id": 263756,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263756,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 263756, "pid": 5, "tid": 7, "ts": 1716454225666014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608831, "dur": 7,
    "args": {
      "External id": 263756,
      "cbid": 51, "correlation": 263756
    }
  },
  {
    "ph": "s", "id": 263756, "pid": 76337, "tid": -914061504, "ts": 1716454225608831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225666017, "dur": 263,
    "args": {
      "External id": 263757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263757,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263757, "pid": 5, "tid": 7, "ts": 1716454225666017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608839, "dur": 6,
    "args": {
      "External id": 263757,
      "cbid": 211, "correlation": 263757
    }
  },
  {
    "ph": "s", "id": 263757, "pid": 76337, "tid": -914061504, "ts": 1716454225608839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666282, "dur": 5,
    "args": {
      "External id": 263759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263759,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263759, "pid": 5, "tid": 7, "ts": 1716454225666282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608848, "dur": 5,
    "args": {
      "External id": 263759,
      "cbid": 211, "correlation": 263759
    }
  },
  {
    "ph": "s", "id": 263759, "pid": 76337, "tid": -914061504, "ts": 1716454225608848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225666289, "dur": 6,
    "args": {
      "External id": 263765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263765, "pid": 5, "tid": 7, "ts": 1716454225666289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608876, "dur": 9,
    "args": {
      "External id": 263765,
      "cbid": 211, "correlation": 263765
    }
  },
  {
    "ph": "s", "id": 263765, "pid": 76337, "tid": -914061504, "ts": 1716454225608876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225666297, "dur": 3,
    "args": {
      "External id": 263773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263773,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 263773, "pid": 5, "tid": 7, "ts": 1716454225666297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608919, "dur": 9,
    "args": {
      "External id": 263773,
      "cbid": 211, "correlation": 263773
    }
  },
  {
    "ph": "s", "id": 263773, "pid": 76337, "tid": -914061504, "ts": 1716454225608919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608988, "dur": 1,
    "args": {
      "External id": 263789,
      "cbid": 251, "correlation": 263789
    }
  },
  {
    "ph": "f", "id": 263789, "pid": 76337, "tid": -914061504, "ts": 1716454225608988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608994, "dur": 0,
    "args": {
      "External id": 263791,
      "cbid": 251, "correlation": 263791
    }
  },
  {
    "ph": "f", "id": 263791, "pid": 76337, "tid": -914061504, "ts": 1716454225608994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225666301, "dur": 10,
    "args": {
      "External id": 263792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263792,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263792, "pid": 5, "tid": 7, "ts": 1716454225666301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225608996, "dur": 12,
    "args": {
      "External id": 263792,
      "cbid": 211, "correlation": 263792
    }
  },
  {
    "ph": "s", "id": 263792, "pid": 76337, "tid": -914061504, "ts": 1716454225608996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225666313, "dur": 4,
    "args": {
      "External id": 263794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263794,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263794, "pid": 5, "tid": 7, "ts": 1716454225666313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609010, "dur": 6,
    "args": {
      "External id": 263794,
      "cbid": 211, "correlation": 263794
    }
  },
  {
    "ph": "s", "id": 263794, "pid": 76337, "tid": -914061504, "ts": 1716454225609010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225666318, "dur": 6,
    "args": {
      "External id": 263804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263804,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263804, "pid": 5, "tid": 7, "ts": 1716454225666318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609067, "dur": 12,
    "args": {
      "External id": 263804,
      "cbid": 211, "correlation": 263804
    }
  },
  {
    "ph": "s", "id": 263804, "pid": 76337, "tid": -914061504, "ts": 1716454225609067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666325, "dur": 10,
    "args": {
      "External id": 263824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263824,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263824, "pid": 5, "tid": 7, "ts": 1716454225666325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609134, "dur": 11,
    "args": {
      "External id": 263824,
      "cbid": 211, "correlation": 263824
    }
  },
  {
    "ph": "s", "id": 263824, "pid": 76337, "tid": -914061504, "ts": 1716454225609134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225666336, "dur": 4,
    "args": {
      "External id": 263836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263836,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263836, "pid": 5, "tid": 7, "ts": 1716454225666336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609155, "dur": 6,
    "args": {
      "External id": 263836,
      "cbid": 211, "correlation": 263836
    }
  },
  {
    "ph": "s", "id": 263836, "pid": 76337, "tid": -914061504, "ts": 1716454225609155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225666341, "dur": 7,
    "args": {
      "External id": 263839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263839, "pid": 5, "tid": 7, "ts": 1716454225666341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609174, "dur": 6,
    "args": {
      "External id": 263839,
      "cbid": 211, "correlation": 263839
    }
  },
  {
    "ph": "s", "id": 263839, "pid": 76337, "tid": -914061504, "ts": 1716454225609174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225666349, "dur": 5,
    "args": {
      "External id": 263848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263848,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263848, "pid": 5, "tid": 7, "ts": 1716454225666349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609213, "dur": 9,
    "args": {
      "External id": 263848,
      "cbid": 211, "correlation": 263848
    }
  },
  {
    "ph": "s", "id": 263848, "pid": 76337, "tid": -914061504, "ts": 1716454225609213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609275, "dur": 0,
    "args": {
      "External id": 263858,
      "cbid": 317, "correlation": 263858
    }
  },
  {
    "ph": "f", "id": 263858, "pid": 76337, "tid": -914061504, "ts": 1716454225609275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609276, "dur": 0,
    "args": {
      "External id": 263859,
      "cbid": 203, "correlation": 263859
    }
  },
  {
    "ph": "f", "id": 263859, "pid": 76337, "tid": -914061504, "ts": 1716454225609276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609277, "dur": 0,
    "args": {
      "External id": 263860,
      "cbid": 205, "correlation": 263860
    }
  },
  {
    "ph": "f", "id": 263860, "pid": 76337, "tid": -914061504, "ts": 1716454225609277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666355, "dur": 5,
    "args": {
      "External id": 263864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263864,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263864, "pid": 5, "tid": 7, "ts": 1716454225666355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609291, "dur": 13,
    "args": {
      "External id": 263864,
      "cbid": 211, "correlation": 263864
    }
  },
  {
    "ph": "s", "id": 263864, "pid": 76337, "tid": -914061504, "ts": 1716454225609291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666362, "dur": 165,
    "args": {
      "External id": 263866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263866,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263866, "pid": 5, "tid": 7, "ts": 1716454225666362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609306, "dur": 5,
    "args": {
      "External id": 263866,
      "cbid": 211, "correlation": 263866
    }
  },
  {
    "ph": "s", "id": 263866, "pid": 76337, "tid": -914061504, "ts": 1716454225609306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225666529, "dur": 1,
    "args": {
      "External id": 263868,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 263868,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 263868, "pid": 5, "tid": 7, "ts": 1716454225666529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609316, "dur": 7,
    "args": {
      "External id": 263868,
      "cbid": 51, "correlation": 263868
    }
  },
  {
    "ph": "s", "id": 263868, "pid": 76337, "tid": -914061504, "ts": 1716454225609316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225666533, "dur": 264,
    "args": {
      "External id": 263869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263869,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 263869, "pid": 5, "tid": 7, "ts": 1716454225666533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609324, "dur": 6,
    "args": {
      "External id": 263869,
      "cbid": 211, "correlation": 263869
    }
  },
  {
    "ph": "s", "id": 263869, "pid": 76337, "tid": -914061504, "ts": 1716454225609324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666798, "dur": 6,
    "args": {
      "External id": 263871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263871,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263871, "pid": 5, "tid": 7, "ts": 1716454225666798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609334, "dur": 5,
    "args": {
      "External id": 263871,
      "cbid": 211, "correlation": 263871
    }
  },
  {
    "ph": "s", "id": 263871, "pid": 76337, "tid": -914061504, "ts": 1716454225609334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225666806, "dur": 6,
    "args": {
      "External id": 263877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263877,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263877, "pid": 5, "tid": 7, "ts": 1716454225666806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609363, "dur": 9,
    "args": {
      "External id": 263877,
      "cbid": 211, "correlation": 263877
    }
  },
  {
    "ph": "s", "id": 263877, "pid": 76337, "tid": -914061504, "ts": 1716454225609363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225666813, "dur": 5,
    "args": {
      "External id": 263885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263885,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263885, "pid": 5, "tid": 7, "ts": 1716454225666813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609396, "dur": 9,
    "args": {
      "External id": 263885,
      "cbid": 211, "correlation": 263885
    }
  },
  {
    "ph": "s", "id": 263885, "pid": 76337, "tid": -914061504, "ts": 1716454225609396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225666820, "dur": 5,
    "args": {
      "External id": 263893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263893,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263893, "pid": 5, "tid": 7, "ts": 1716454225666820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609425, "dur": 8,
    "args": {
      "External id": 263893,
      "cbid": 211, "correlation": 263893
    }
  },
  {
    "ph": "s", "id": 263893, "pid": 76337, "tid": -914061504, "ts": 1716454225609425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666826, "dur": 10,
    "args": {
      "External id": 263913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263913,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 263913, "pid": 5, "tid": 7, "ts": 1716454225666826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609565, "dur": 13,
    "args": {
      "External id": 263913,
      "cbid": 211, "correlation": 263913
    }
  },
  {
    "ph": "s", "id": 263913, "pid": 76337, "tid": -914061504, "ts": 1716454225609565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225666837, "dur": 4,
    "args": {
      "External id": 263925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263925,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 263925, "pid": 5, "tid": 7, "ts": 1716454225666837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609588, "dur": 7,
    "args": {
      "External id": 263925,
      "cbid": 211, "correlation": 263925
    }
  },
  {
    "ph": "s", "id": 263925, "pid": 76337, "tid": -914061504, "ts": 1716454225609588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225666842, "dur": 7,
    "args": {
      "External id": 263928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263928, "pid": 5, "tid": 7, "ts": 1716454225666842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609607, "dur": 8,
    "args": {
      "External id": 263928,
      "cbid": 211, "correlation": 263928
    }
  },
  {
    "ph": "s", "id": 263928, "pid": 76337, "tid": -914061504, "ts": 1716454225609607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609667, "dur": 0,
    "args": {
      "External id": 263939,
      "cbid": 317, "correlation": 263939
    }
  },
  {
    "ph": "f", "id": 263939, "pid": 76337, "tid": -914061504, "ts": 1716454225609667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609668, "dur": 0,
    "args": {
      "External id": 263940,
      "cbid": 203, "correlation": 263940
    }
  },
  {
    "ph": "f", "id": 263940, "pid": 76337, "tid": -914061504, "ts": 1716454225609668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609668, "dur": 0,
    "args": {
      "External id": 263941,
      "cbid": 205, "correlation": 263941
    }
  },
  {
    "ph": "f", "id": 263941, "pid": 76337, "tid": -914061504, "ts": 1716454225609668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666850, "dur": 5,
    "args": {
      "External id": 263945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263945,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263945, "pid": 5, "tid": 7, "ts": 1716454225666850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609685, "dur": 13,
    "args": {
      "External id": 263945,
      "cbid": 211, "correlation": 263945
    }
  },
  {
    "ph": "s", "id": 263945, "pid": 76337, "tid": -914061504, "ts": 1716454225609685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225666856, "dur": 38,
    "args": {
      "External id": 263947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263947,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 263947, "pid": 5, "tid": 7, "ts": 1716454225666856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609705, "dur": 9,
    "args": {
      "External id": 263947,
      "cbid": 211, "correlation": 263947
    }
  },
  {
    "ph": "s", "id": 263947, "pid": 76337, "tid": -914061504, "ts": 1716454225609705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666895, "dur": 5,
    "args": {
      "External id": 263949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263949, "pid": 5, "tid": 7, "ts": 1716454225666895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609717, "dur": 5,
    "args": {
      "External id": 263949,
      "cbid": 211, "correlation": 263949
    }
  },
  {
    "ph": "s", "id": 263949, "pid": 76337, "tid": -914061504, "ts": 1716454225609717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225666902, "dur": 6,
    "args": {
      "External id": 263955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 263955, "pid": 5, "tid": 7, "ts": 1716454225666902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609745, "dur": 9,
    "args": {
      "External id": 263955,
      "cbid": 211, "correlation": 263955
    }
  },
  {
    "ph": "s", "id": 263955, "pid": 76337, "tid": -914061504, "ts": 1716454225609745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225666909, "dur": 21,
    "args": {
      "External id": 263964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263964,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 263964, "pid": 5, "tid": 7, "ts": 1716454225666909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609828, "dur": 14,
    "args": {
      "External id": 263964,
      "cbid": 211, "correlation": 263964
    }
  },
  {
    "ph": "s", "id": 263964, "pid": 76337, "tid": -914061504, "ts": 1716454225609828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225666931, "dur": 11,
    "args": {
      "External id": 263986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263986,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 263986, "pid": 5, "tid": 7, "ts": 1716454225666931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609885, "dur": 10,
    "args": {
      "External id": 263986,
      "cbid": 211, "correlation": 263986
    }
  },
  {
    "ph": "s", "id": 263986, "pid": 76337, "tid": -914061504, "ts": 1716454225609885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609984, "dur": 2,
    "args": {
      "External id": 263997,
      "cbid": 251, "correlation": 263997
    }
  },
  {
    "ph": "f", "id": 263997, "pid": 76337, "tid": -914061504, "ts": 1716454225609984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609990, "dur": 0,
    "args": {
      "External id": 263998,
      "cbid": 251, "correlation": 263998
    }
  },
  {
    "ph": "f", "id": 263998, "pid": 76337, "tid": -914061504, "ts": 1716454225609990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225666943, "dur": 55,
    "args": {
      "External id": 263999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 263999,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 263999, "pid": 5, "tid": 7, "ts": 1716454225666943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225609993, "dur": 15,
    "args": {
      "External id": 263999,
      "cbid": 211, "correlation": 263999
    }
  },
  {
    "ph": "s", "id": 263999, "pid": 76337, "tid": -914061504, "ts": 1716454225609993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610067, "dur": 1,
    "args": {
      "External id": 264010,
      "cbid": 251, "correlation": 264010
    }
  },
  {
    "ph": "f", "id": 264010, "pid": 76337, "tid": -914061504, "ts": 1716454225610067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610070, "dur": 0,
    "args": {
      "External id": 264011,
      "cbid": 251, "correlation": 264011
    }
  },
  {
    "ph": "f", "id": 264011, "pid": 76337, "tid": -914061504, "ts": 1716454225610070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225666999, "dur": 54,
    "args": {
      "External id": 264012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264012,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 264012, "pid": 5, "tid": 7, "ts": 1716454225666999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610072, "dur": 11,
    "args": {
      "External id": 264012,
      "cbid": 211, "correlation": 264012
    }
  },
  {
    "ph": "s", "id": 264012, "pid": 76337, "tid": -914061504, "ts": 1716454225610072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610136, "dur": 1,
    "args": {
      "External id": 264023,
      "cbid": 251, "correlation": 264023
    }
  },
  {
    "ph": "f", "id": 264023, "pid": 76337, "tid": -914061504, "ts": 1716454225610136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610140, "dur": 0,
    "args": {
      "External id": 264024,
      "cbid": 251, "correlation": 264024
    }
  },
  {
    "ph": "f", "id": 264024, "pid": 76337, "tid": -914061504, "ts": 1716454225610140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225667055, "dur": 54,
    "args": {
      "External id": 264025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264025,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 264025, "pid": 5, "tid": 7, "ts": 1716454225667055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610142, "dur": 12,
    "args": {
      "External id": 264025,
      "cbid": 211, "correlation": 264025
    }
  },
  {
    "ph": "s", "id": 264025, "pid": 76337, "tid": -914061504, "ts": 1716454225610142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225667110, "dur": 58,
    "args": {
      "External id": 264050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264050,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264050, "pid": 5, "tid": 7, "ts": 1716454225667110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610227, "dur": 13,
    "args": {
      "External id": 264050,
      "cbid": 211, "correlation": 264050
    }
  },
  {
    "ph": "s", "id": 264050, "pid": 76337, "tid": -914061504, "ts": 1716454225610227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610326, "dur": 1,
    "args": {
      "External id": 264068,
      "cbid": 251, "correlation": 264068
    }
  },
  {
    "ph": "f", "id": 264068, "pid": 76337, "tid": -914061504, "ts": 1716454225610326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225667169, "dur": 64,
    "args": {
      "External id": 264070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264070,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 264070, "pid": 5, "tid": 7, "ts": 1716454225667169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610332, "dur": 13,
    "args": {
      "External id": 264070,
      "cbid": 211, "correlation": 264070
    }
  },
  {
    "ph": "s", "id": 264070, "pid": 76337, "tid": -914061504, "ts": 1716454225610332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225667234, "dur": 6,
    "args": {
      "External id": 264078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264078,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264078, "pid": 5, "tid": 7, "ts": 1716454225667234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610402, "dur": 12,
    "args": {
      "External id": 264078,
      "cbid": 211, "correlation": 264078
    }
  },
  {
    "ph": "s", "id": 264078, "pid": 76337, "tid": -914061504, "ts": 1716454225610402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225667242, "dur": 7,
    "args": {
      "External id": 264086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264086,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264086, "pid": 5, "tid": 7, "ts": 1716454225667242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610440, "dur": 9,
    "args": {
      "External id": 264086,
      "cbid": 211, "correlation": 264086
    }
  },
  {
    "ph": "s", "id": 264086, "pid": 76337, "tid": -914061504, "ts": 1716454225610440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667250, "dur": 8,
    "args": {
      "External id": 264097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264097, "pid": 5, "tid": 7, "ts": 1716454225667250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610513, "dur": 13,
    "args": {
      "External id": 264097,
      "cbid": 211, "correlation": 264097
    }
  },
  {
    "ph": "s", "id": 264097, "pid": 76337, "tid": -914061504, "ts": 1716454225610513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667259, "dur": 9,
    "args": {
      "External id": 264119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264119,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 264119, "pid": 5, "tid": 7, "ts": 1716454225667259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610545, "dur": 8,
    "args": {
      "External id": 264119,
      "cbid": 211, "correlation": 264119
    }
  },
  {
    "ph": "s", "id": 264119, "pid": 76337, "tid": -914061504, "ts": 1716454225610545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610630, "dur": 2,
    "args": {
      "External id": 264130,
      "cbid": 251, "correlation": 264130
    }
  },
  {
    "ph": "f", "id": 264130, "pid": 76337, "tid": -914061504, "ts": 1716454225610630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225667270, "dur": 1,
    "args": {
      "External id": 264131,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264131,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 264131, "pid": 5, "tid": 7, "ts": 1716454225667270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610636, "dur": 11,
    "args": {
      "External id": 264131,
      "cbid": 51, "correlation": 264131
    }
  },
  {
    "ph": "s", "id": 264131, "pid": 76337, "tid": -914061504, "ts": 1716454225610636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225667274, "dur": 37,
    "args": {
      "External id": 264132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264132,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 264132, "pid": 5, "tid": 7, "ts": 1716454225667274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610648, "dur": 12,
    "args": {
      "External id": 264132,
      "cbid": 211, "correlation": 264132
    }
  },
  {
    "ph": "s", "id": 264132, "pid": 76337, "tid": -914061504, "ts": 1716454225610648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610719, "dur": 1,
    "args": {
      "External id": 264143,
      "cbid": 251, "correlation": 264143
    }
  },
  {
    "ph": "f", "id": 264143, "pid": 76337, "tid": -914061504, "ts": 1716454225610719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610723, "dur": 0,
    "args": {
      "External id": 264144,
      "cbid": 251, "correlation": 264144
    }
  },
  {
    "ph": "f", "id": 264144, "pid": 76337, "tid": -914061504, "ts": 1716454225610723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225667312, "dur": 12,
    "args": {
      "External id": 264145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264145,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264145, "pid": 5, "tid": 7, "ts": 1716454225667312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610725, "dur": 12,
    "args": {
      "External id": 264145,
      "cbid": 211, "correlation": 264145
    }
  },
  {
    "ph": "s", "id": 264145, "pid": 76337, "tid": -914061504, "ts": 1716454225610725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225667325, "dur": 5,
    "args": {
      "External id": 264147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264147,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264147, "pid": 5, "tid": 7, "ts": 1716454225667325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610738, "dur": 5,
    "args": {
      "External id": 264147,
      "cbid": 211, "correlation": 264147
    }
  },
  {
    "ph": "s", "id": 264147, "pid": 76337, "tid": -914061504, "ts": 1716454225610738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610795, "dur": 1,
    "args": {
      "External id": 264158,
      "cbid": 251, "correlation": 264158
    }
  },
  {
    "ph": "f", "id": 264158, "pid": 76337, "tid": -914061504, "ts": 1716454225610795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610799, "dur": 0,
    "args": {
      "External id": 264159,
      "cbid": 251, "correlation": 264159
    }
  },
  {
    "ph": "f", "id": 264159, "pid": 76337, "tid": -914061504, "ts": 1716454225610799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225667332, "dur": 9,
    "args": {
      "External id": 264160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264160,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264160, "pid": 5, "tid": 7, "ts": 1716454225667332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610800, "dur": 12,
    "args": {
      "External id": 264160,
      "cbid": 211, "correlation": 264160
    }
  },
  {
    "ph": "s", "id": 264160, "pid": 76337, "tid": -914061504, "ts": 1716454225610800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225667342, "dur": 3,
    "args": {
      "External id": 264162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264162,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264162, "pid": 5, "tid": 7, "ts": 1716454225667342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610814, "dur": 5,
    "args": {
      "External id": 264162,
      "cbid": 211, "correlation": 264162
    }
  },
  {
    "ph": "s", "id": 264162, "pid": 76337, "tid": -914061504, "ts": 1716454225610814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225667346, "dur": 20,
    "args": {
      "External id": 264187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264187,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 264187, "pid": 5, "tid": 7, "ts": 1716454225667346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610891, "dur": 13,
    "args": {
      "External id": 264187,
      "cbid": 211, "correlation": 264187
    }
  },
  {
    "ph": "s", "id": 264187, "pid": 76337, "tid": -914061504, "ts": 1716454225610891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225610999, "dur": 2,
    "args": {
      "External id": 264205,
      "cbid": 251, "correlation": 264205
    }
  },
  {
    "ph": "f", "id": 264205, "pid": 76337, "tid": -914061504, "ts": 1716454225610999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225667369, "dur": 1,
    "args": {
      "External id": 264207,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264207,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 264207, "pid": 5, "tid": 7, "ts": 1716454225667369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611006, "dur": 11,
    "args": {
      "External id": 264207,
      "cbid": 51, "correlation": 264207
    }
  },
  {
    "ph": "s", "id": 264207, "pid": 76337, "tid": -914061504, "ts": 1716454225611006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225667372, "dur": 37,
    "args": {
      "External id": 264208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264208,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 264208, "pid": 5, "tid": 7, "ts": 1716454225667372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611018, "dur": 12,
    "args": {
      "External id": 264208,
      "cbid": 211, "correlation": 264208
    }
  },
  {
    "ph": "s", "id": 264208, "pid": 76337, "tid": -914061504, "ts": 1716454225611018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225667411, "dur": 4,
    "args": {
      "External id": 264216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264216,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264216, "pid": 5, "tid": 7, "ts": 1716454225667411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611090, "dur": 12,
    "args": {
      "External id": 264216,
      "cbid": 211, "correlation": 264216
    }
  },
  {
    "ph": "s", "id": 264216, "pid": 76337, "tid": -914061504, "ts": 1716454225611090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667416, "dur": 8,
    "args": {
      "External id": 264224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264224,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264224, "pid": 5, "tid": 7, "ts": 1716454225667416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611131, "dur": 9,
    "args": {
      "External id": 264224,
      "cbid": 211, "correlation": 264224
    }
  },
  {
    "ph": "s", "id": 264224, "pid": 76337, "tid": -914061504, "ts": 1716454225611131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667426, "dur": 8,
    "args": {
      "External id": 264246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264246,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 264246, "pid": 5, "tid": 7, "ts": 1716454225667426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611184, "dur": 10,
    "args": {
      "External id": 264246,
      "cbid": 211, "correlation": 264246
    }
  },
  {
    "ph": "s", "id": 264246, "pid": 76337, "tid": -914061504, "ts": 1716454225611184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611275, "dur": 1,
    "args": {
      "External id": 264262,
      "cbid": 251, "correlation": 264262
    }
  },
  {
    "ph": "f", "id": 264262, "pid": 76337, "tid": -914061504, "ts": 1716454225611275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611280, "dur": 0,
    "args": {
      "External id": 264264,
      "cbid": 251, "correlation": 264264
    }
  },
  {
    "ph": "f", "id": 264264, "pid": 76337, "tid": -914061504, "ts": 1716454225611280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225667436, "dur": 189,
    "args": {
      "External id": 264265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264265,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264265, "pid": 5, "tid": 7, "ts": 1716454225667436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611282, "dur": 13,
    "args": {
      "External id": 264265,
      "cbid": 211, "correlation": 264265
    }
  },
  {
    "ph": "s", "id": 264265, "pid": 76337, "tid": -914061504, "ts": 1716454225611282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667626, "dur": 21,
    "args": {
      "External id": 264273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264273,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264273, "pid": 5, "tid": 7, "ts": 1716454225667626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611347, "dur": 13,
    "args": {
      "External id": 264273,
      "cbid": 211, "correlation": 264273
    }
  },
  {
    "ph": "s", "id": 264273, "pid": 76337, "tid": -914061504, "ts": 1716454225611347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667649, "dur": 21,
    "args": {
      "External id": 264281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264281, "pid": 5, "tid": 7, "ts": 1716454225667649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611379, "dur": 8,
    "args": {
      "External id": 264281,
      "cbid": 211, "correlation": 264281
    }
  },
  {
    "ph": "s", "id": 264281, "pid": 76337, "tid": -914061504, "ts": 1716454225611379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611459, "dur": 1,
    "args": {
      "External id": 264297,
      "cbid": 251, "correlation": 264297
    }
  },
  {
    "ph": "f", "id": 264297, "pid": 76337, "tid": -914061504, "ts": 1716454225611459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225667672, "dur": 1,
    "args": {
      "External id": 264299,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264299,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 264299, "pid": 5, "tid": 7, "ts": 1716454225667672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611464, "dur": 9,
    "args": {
      "External id": 264299,
      "cbid": 51, "correlation": 264299
    }
  },
  {
    "ph": "s", "id": 264299, "pid": 76337, "tid": -914061504, "ts": 1716454225611464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225667676, "dur": 111,
    "args": {
      "External id": 264300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264300,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 264300, "pid": 5, "tid": 7, "ts": 1716454225667676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611475, "dur": 11,
    "args": {
      "External id": 264300,
      "cbid": 211, "correlation": 264300
    }
  },
  {
    "ph": "s", "id": 264300, "pid": 76337, "tid": -914061504, "ts": 1716454225611475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225667789, "dur": 6,
    "args": {
      "External id": 264308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264308,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264308, "pid": 5, "tid": 7, "ts": 1716454225667789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611517, "dur": 10,
    "args": {
      "External id": 264308,
      "cbid": 211, "correlation": 264308
    }
  },
  {
    "ph": "s", "id": 264308, "pid": 76337, "tid": -914061504, "ts": 1716454225611517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667796, "dur": 10,
    "args": {
      "External id": 264319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264319,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264319, "pid": 5, "tid": 7, "ts": 1716454225667796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611585, "dur": 12,
    "args": {
      "External id": 264319,
      "cbid": 211, "correlation": 264319
    }
  },
  {
    "ph": "s", "id": 264319, "pid": 76337, "tid": -914061504, "ts": 1716454225611585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611650, "dur": 0,
    "args": {
      "External id": 264331,
      "cbid": 317, "correlation": 264331
    }
  },
  {
    "ph": "f", "id": 264331, "pid": 76337, "tid": -914061504, "ts": 1716454225611650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611651, "dur": 0,
    "args": {
      "External id": 264332,
      "cbid": 203, "correlation": 264332
    }
  },
  {
    "ph": "f", "id": 264332, "pid": 76337, "tid": -914061504, "ts": 1716454225611651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611651, "dur": 0,
    "args": {
      "External id": 264333,
      "cbid": 205, "correlation": 264333
    }
  },
  {
    "ph": "f", "id": 264333, "pid": 76337, "tid": -914061504, "ts": 1716454225611651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667807, "dur": 5,
    "args": {
      "External id": 264337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264337, "pid": 5, "tid": 7, "ts": 1716454225667807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611667, "dur": 12,
    "args": {
      "External id": 264337,
      "cbid": 211, "correlation": 264337
    }
  },
  {
    "ph": "s", "id": 264337, "pid": 76337, "tid": -914061504, "ts": 1716454225611667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225667814, "dur": 38,
    "args": {
      "External id": 264339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264339,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 264339, "pid": 5, "tid": 7, "ts": 1716454225667814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611686, "dur": 7,
    "args": {
      "External id": 264339,
      "cbid": 211, "correlation": 264339
    }
  },
  {
    "ph": "s", "id": 264339, "pid": 76337, "tid": -914061504, "ts": 1716454225611686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667853, "dur": 6,
    "args": {
      "External id": 264341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264341,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264341, "pid": 5, "tid": 7, "ts": 1716454225667853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611697, "dur": 5,
    "args": {
      "External id": 264341,
      "cbid": 211, "correlation": 264341
    }
  },
  {
    "ph": "s", "id": 264341, "pid": 76337, "tid": -914061504, "ts": 1716454225611697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667860, "dur": 7,
    "args": {
      "External id": 264347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264347, "pid": 5, "tid": 7, "ts": 1716454225667860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611724, "dur": 8,
    "args": {
      "External id": 264347,
      "cbid": 211, "correlation": 264347
    }
  },
  {
    "ph": "s", "id": 264347, "pid": 76337, "tid": -914061504, "ts": 1716454225611724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225667868, "dur": 5,
    "args": {
      "External id": 264355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264355,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264355, "pid": 5, "tid": 7, "ts": 1716454225667868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611755, "dur": 9,
    "args": {
      "External id": 264355,
      "cbid": 211, "correlation": 264355
    }
  },
  {
    "ph": "s", "id": 264355, "pid": 76337, "tid": -914061504, "ts": 1716454225611755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667875, "dur": 11,
    "args": {
      "External id": 264375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264375,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 264375, "pid": 5, "tid": 7, "ts": 1716454225667875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611827, "dur": 11,
    "args": {
      "External id": 264375,
      "cbid": 211, "correlation": 264375
    }
  },
  {
    "ph": "s", "id": 264375, "pid": 76337, "tid": -914061504, "ts": 1716454225611827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225667888, "dur": 5,
    "args": {
      "External id": 264387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264387,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 264387, "pid": 5, "tid": 7, "ts": 1716454225667888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611848, "dur": 6,
    "args": {
      "External id": 264387,
      "cbid": 211, "correlation": 264387
    }
  },
  {
    "ph": "s", "id": 264387, "pid": 76337, "tid": -914061504, "ts": 1716454225611848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225667894, "dur": 8,
    "args": {
      "External id": 264390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264390,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264390, "pid": 5, "tid": 7, "ts": 1716454225667894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611866, "dur": 7,
    "args": {
      "External id": 264390,
      "cbid": 211, "correlation": 264390
    }
  },
  {
    "ph": "s", "id": 264390, "pid": 76337, "tid": -914061504, "ts": 1716454225611866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225667903, "dur": 5,
    "args": {
      "External id": 264399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264399,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264399, "pid": 5, "tid": 7, "ts": 1716454225667903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611903, "dur": 10,
    "args": {
      "External id": 264399,
      "cbid": 211, "correlation": 264399
    }
  },
  {
    "ph": "s", "id": 264399, "pid": 76337, "tid": -914061504, "ts": 1716454225611903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611954, "dur": 0,
    "args": {
      "External id": 264409,
      "cbid": 317, "correlation": 264409
    }
  },
  {
    "ph": "f", "id": 264409, "pid": 76337, "tid": -914061504, "ts": 1716454225611954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611955, "dur": 0,
    "args": {
      "External id": 264410,
      "cbid": 203, "correlation": 264410
    }
  },
  {
    "ph": "f", "id": 264410, "pid": 76337, "tid": -914061504, "ts": 1716454225611955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611956, "dur": 0,
    "args": {
      "External id": 264411,
      "cbid": 205, "correlation": 264411
    }
  },
  {
    "ph": "f", "id": 264411, "pid": 76337, "tid": -914061504, "ts": 1716454225611956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667910, "dur": 5,
    "args": {
      "External id": 264415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264415,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264415, "pid": 5, "tid": 7, "ts": 1716454225667910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611969, "dur": 20,
    "args": {
      "External id": 264415,
      "cbid": 211, "correlation": 264415
    }
  },
  {
    "ph": "s", "id": 264415, "pid": 76337, "tid": -914061504, "ts": 1716454225611969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225667916, "dur": 165,
    "args": {
      "External id": 264417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264417,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264417, "pid": 5, "tid": 7, "ts": 1716454225667916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225611992, "dur": 5,
    "args": {
      "External id": 264417,
      "cbid": 211, "correlation": 264417
    }
  },
  {
    "ph": "s", "id": 264417, "pid": 76337, "tid": -914061504, "ts": 1716454225611992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225668083, "dur": 1,
    "args": {
      "External id": 264419,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264419,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 264419, "pid": 5, "tid": 7, "ts": 1716454225668083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612003, "dur": 7,
    "args": {
      "External id": 264419,
      "cbid": 51, "correlation": 264419
    }
  },
  {
    "ph": "s", "id": 264419, "pid": 76337, "tid": -914061504, "ts": 1716454225612003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225668087, "dur": 275,
    "args": {
      "External id": 264420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264420,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264420, "pid": 5, "tid": 7, "ts": 1716454225668087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612012, "dur": 7,
    "args": {
      "External id": 264420,
      "cbid": 211, "correlation": 264420
    }
  },
  {
    "ph": "s", "id": 264420, "pid": 76337, "tid": -914061504, "ts": 1716454225612012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668363, "dur": 6,
    "args": {
      "External id": 264422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264422, "pid": 5, "tid": 7, "ts": 1716454225668363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612023, "dur": 5,
    "args": {
      "External id": 264422,
      "cbid": 211, "correlation": 264422
    }
  },
  {
    "ph": "s", "id": 264422, "pid": 76337, "tid": -914061504, "ts": 1716454225612023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225668370, "dur": 6,
    "args": {
      "External id": 264428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264428, "pid": 5, "tid": 7, "ts": 1716454225668370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612050, "dur": 8,
    "args": {
      "External id": 264428,
      "cbid": 211, "correlation": 264428
    }
  },
  {
    "ph": "s", "id": 264428, "pid": 76337, "tid": -914061504, "ts": 1716454225612050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225668378, "dur": 3,
    "args": {
      "External id": 264436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264436,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 264436, "pid": 5, "tid": 7, "ts": 1716454225668378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612095, "dur": 10,
    "args": {
      "External id": 264436,
      "cbid": 211, "correlation": 264436
    }
  },
  {
    "ph": "s", "id": 264436, "pid": 76337, "tid": -914061504, "ts": 1716454225612095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612163, "dur": 1,
    "args": {
      "External id": 264452,
      "cbid": 251, "correlation": 264452
    }
  },
  {
    "ph": "f", "id": 264452, "pid": 76337, "tid": -914061504, "ts": 1716454225612163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612169, "dur": 0,
    "args": {
      "External id": 264454,
      "cbid": 251, "correlation": 264454
    }
  },
  {
    "ph": "f", "id": 264454, "pid": 76337, "tid": -914061504, "ts": 1716454225612169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225668382, "dur": 13,
    "args": {
      "External id": 264455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264455,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264455, "pid": 5, "tid": 7, "ts": 1716454225668382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612171, "dur": 11,
    "args": {
      "External id": 264455,
      "cbid": 211, "correlation": 264455
    }
  },
  {
    "ph": "s", "id": 264455, "pid": 76337, "tid": -914061504, "ts": 1716454225612171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225668396, "dur": 5,
    "args": {
      "External id": 264457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264457,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264457, "pid": 5, "tid": 7, "ts": 1716454225668396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612183, "dur": 5,
    "args": {
      "External id": 264457,
      "cbid": 211, "correlation": 264457
    }
  },
  {
    "ph": "s", "id": 264457, "pid": 76337, "tid": -914061504, "ts": 1716454225612183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225668403, "dur": 6,
    "args": {
      "External id": 264467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264467,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264467, "pid": 5, "tid": 7, "ts": 1716454225668403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612241, "dur": 12,
    "args": {
      "External id": 264467,
      "cbid": 211, "correlation": 264467
    }
  },
  {
    "ph": "s", "id": 264467, "pid": 76337, "tid": -914061504, "ts": 1716454225612241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668410, "dur": 10,
    "args": {
      "External id": 264487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264487,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 264487, "pid": 5, "tid": 7, "ts": 1716454225668410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612306, "dur": 11,
    "args": {
      "External id": 264487,
      "cbid": 211, "correlation": 264487
    }
  },
  {
    "ph": "s", "id": 264487, "pid": 76337, "tid": -914061504, "ts": 1716454225612306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225668421, "dur": 4,
    "args": {
      "External id": 264499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264499,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 264499, "pid": 5, "tid": 7, "ts": 1716454225668421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612327, "dur": 6,
    "args": {
      "External id": 264499,
      "cbid": 211, "correlation": 264499
    }
  },
  {
    "ph": "s", "id": 264499, "pid": 76337, "tid": -914061504, "ts": 1716454225612327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225668427, "dur": 7,
    "args": {
      "External id": 264502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264502, "pid": 5, "tid": 7, "ts": 1716454225668427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612345, "dur": 6,
    "args": {
      "External id": 264502,
      "cbid": 211, "correlation": 264502
    }
  },
  {
    "ph": "s", "id": 264502, "pid": 76337, "tid": -914061504, "ts": 1716454225612345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225668435, "dur": 5,
    "args": {
      "External id": 264511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264511,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264511, "pid": 5, "tid": 7, "ts": 1716454225668435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612385, "dur": 10,
    "args": {
      "External id": 264511,
      "cbid": 211, "correlation": 264511
    }
  },
  {
    "ph": "s", "id": 264511, "pid": 76337, "tid": -914061504, "ts": 1716454225612385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612448, "dur": 0,
    "args": {
      "External id": 264521,
      "cbid": 317, "correlation": 264521
    }
  },
  {
    "ph": "f", "id": 264521, "pid": 76337, "tid": -914061504, "ts": 1716454225612448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612448, "dur": 0,
    "args": {
      "External id": 264522,
      "cbid": 203, "correlation": 264522
    }
  },
  {
    "ph": "f", "id": 264522, "pid": 76337, "tid": -914061504, "ts": 1716454225612448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612449, "dur": 0,
    "args": {
      "External id": 264523,
      "cbid": 205, "correlation": 264523
    }
  },
  {
    "ph": "f", "id": 264523, "pid": 76337, "tid": -914061504, "ts": 1716454225612449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668441, "dur": 5,
    "args": {
      "External id": 264527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264527,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264527, "pid": 5, "tid": 7, "ts": 1716454225668441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612463, "dur": 12,
    "args": {
      "External id": 264527,
      "cbid": 211, "correlation": 264527
    }
  },
  {
    "ph": "s", "id": 264527, "pid": 76337, "tid": -914061504, "ts": 1716454225612463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668447, "dur": 165,
    "args": {
      "External id": 264529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264529,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264529, "pid": 5, "tid": 7, "ts": 1716454225668447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612478, "dur": 5,
    "args": {
      "External id": 264529,
      "cbid": 211, "correlation": 264529
    }
  },
  {
    "ph": "s", "id": 264529, "pid": 76337, "tid": -914061504, "ts": 1716454225612478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225668615, "dur": 1,
    "args": {
      "External id": 264531,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264531,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 264531, "pid": 5, "tid": 7, "ts": 1716454225668615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612488, "dur": 6,
    "args": {
      "External id": 264531,
      "cbid": 51, "correlation": 264531
    }
  },
  {
    "ph": "s", "id": 264531, "pid": 76337, "tid": -914061504, "ts": 1716454225612488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225668619, "dur": 264,
    "args": {
      "External id": 264532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264532,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264532, "pid": 5, "tid": 7, "ts": 1716454225668619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612495, "dur": 6,
    "args": {
      "External id": 264532,
      "cbid": 211, "correlation": 264532
    }
  },
  {
    "ph": "s", "id": 264532, "pid": 76337, "tid": -914061504, "ts": 1716454225612495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668884, "dur": 6,
    "args": {
      "External id": 264534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264534,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264534, "pid": 5, "tid": 7, "ts": 1716454225668884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612505, "dur": 5,
    "args": {
      "External id": 264534,
      "cbid": 211, "correlation": 264534
    }
  },
  {
    "ph": "s", "id": 264534, "pid": 76337, "tid": -914061504, "ts": 1716454225612505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225668891, "dur": 6,
    "args": {
      "External id": 264540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264540,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264540, "pid": 5, "tid": 7, "ts": 1716454225668891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612534, "dur": 8,
    "args": {
      "External id": 264540,
      "cbid": 211, "correlation": 264540
    }
  },
  {
    "ph": "s", "id": 264540, "pid": 76337, "tid": -914061504, "ts": 1716454225612534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225668899, "dur": 5,
    "args": {
      "External id": 264548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264548,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264548, "pid": 5, "tid": 7, "ts": 1716454225668899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612565, "dur": 9,
    "args": {
      "External id": 264548,
      "cbid": 211, "correlation": 264548
    }
  },
  {
    "ph": "s", "id": 264548, "pid": 76337, "tid": -914061504, "ts": 1716454225612565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225668905, "dur": 5,
    "args": {
      "External id": 264556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264556,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264556, "pid": 5, "tid": 7, "ts": 1716454225668905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612595, "dur": 8,
    "args": {
      "External id": 264556,
      "cbid": 211, "correlation": 264556
    }
  },
  {
    "ph": "s", "id": 264556, "pid": 76337, "tid": -914061504, "ts": 1716454225612595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225668910, "dur": 12,
    "args": {
      "External id": 264565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264565,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264565, "pid": 5, "tid": 7, "ts": 1716454225668910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612683, "dur": 13,
    "args": {
      "External id": 264565,
      "cbid": 211, "correlation": 264565
    }
  },
  {
    "ph": "s", "id": 264565, "pid": 76337, "tid": -914061504, "ts": 1716454225612683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668923, "dur": 12,
    "args": {
      "External id": 264585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264585,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 264585, "pid": 5, "tid": 7, "ts": 1716454225668923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612750, "dur": 11,
    "args": {
      "External id": 264585,
      "cbid": 211, "correlation": 264585
    }
  },
  {
    "ph": "s", "id": 264585, "pid": 76337, "tid": -914061504, "ts": 1716454225612750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225668937, "dur": 4,
    "args": {
      "External id": 264597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264597,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264597, "pid": 5, "tid": 7, "ts": 1716454225668937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612771, "dur": 7,
    "args": {
      "External id": 264597,
      "cbid": 211, "correlation": 264597
    }
  },
  {
    "ph": "s", "id": 264597, "pid": 76337, "tid": -914061504, "ts": 1716454225612771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225668942, "dur": 10,
    "args": {
      "External id": 264600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264600, "pid": 5, "tid": 7, "ts": 1716454225668942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612790, "dur": 6,
    "args": {
      "External id": 264600,
      "cbid": 211, "correlation": 264600
    }
  },
  {
    "ph": "s", "id": 264600, "pid": 76337, "tid": -914061504, "ts": 1716454225612790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225668954, "dur": 6,
    "args": {
      "External id": 264609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264609,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264609, "pid": 5, "tid": 7, "ts": 1716454225668954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612829, "dur": 9,
    "args": {
      "External id": 264609,
      "cbid": 211, "correlation": 264609
    }
  },
  {
    "ph": "s", "id": 264609, "pid": 76337, "tid": -914061504, "ts": 1716454225612829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612881, "dur": 0,
    "args": {
      "External id": 264619,
      "cbid": 317, "correlation": 264619
    }
  },
  {
    "ph": "f", "id": 264619, "pid": 76337, "tid": -914061504, "ts": 1716454225612881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612882, "dur": 0,
    "args": {
      "External id": 264620,
      "cbid": 203, "correlation": 264620
    }
  },
  {
    "ph": "f", "id": 264620, "pid": 76337, "tid": -914061504, "ts": 1716454225612882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612883, "dur": 0,
    "args": {
      "External id": 264621,
      "cbid": 205, "correlation": 264621
    }
  },
  {
    "ph": "f", "id": 264621, "pid": 76337, "tid": -914061504, "ts": 1716454225612883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668961, "dur": 7,
    "args": {
      "External id": 264625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264625, "pid": 5, "tid": 7, "ts": 1716454225668961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612899, "dur": 11,
    "args": {
      "External id": 264625,
      "cbid": 211, "correlation": 264625
    }
  },
  {
    "ph": "s", "id": 264625, "pid": 76337, "tid": -914061504, "ts": 1716454225612899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225668969, "dur": 325,
    "args": {
      "External id": 264627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264627, "pid": 5, "tid": 7, "ts": 1716454225668969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612912, "dur": 5,
    "args": {
      "External id": 264627,
      "cbid": 211, "correlation": 264627
    }
  },
  {
    "ph": "s", "id": 264627, "pid": 76337, "tid": -914061504, "ts": 1716454225612912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225669297, "dur": 1,
    "args": {
      "External id": 264629,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264629,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 264629, "pid": 5, "tid": 7, "ts": 1716454225669297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612923, "dur": 8,
    "args": {
      "External id": 264629,
      "cbid": 51, "correlation": 264629
    }
  },
  {
    "ph": "s", "id": 264629, "pid": 76337, "tid": -914061504, "ts": 1716454225612923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225669300, "dur": 504,
    "args": {
      "External id": 264630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264630,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264630, "pid": 5, "tid": 7, "ts": 1716454225669300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612932, "dur": 6,
    "args": {
      "External id": 264630,
      "cbid": 211, "correlation": 264630
    }
  },
  {
    "ph": "s", "id": 264630, "pid": 76337, "tid": -914061504, "ts": 1716454225612932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225669806, "dur": 6,
    "args": {
      "External id": 264632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264632, "pid": 5, "tid": 7, "ts": 1716454225669806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612942, "dur": 5,
    "args": {
      "External id": 264632,
      "cbid": 211, "correlation": 264632
    }
  },
  {
    "ph": "s", "id": 264632, "pid": 76337, "tid": -914061504, "ts": 1716454225612942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225669813, "dur": 6,
    "args": {
      "External id": 264638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264638, "pid": 5, "tid": 7, "ts": 1716454225669813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225612970, "dur": 18,
    "args": {
      "External id": 264638,
      "cbid": 211, "correlation": 264638
    }
  },
  {
    "ph": "s", "id": 264638, "pid": 76337, "tid": -914061504, "ts": 1716454225612970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225669821, "dur": 3,
    "args": {
      "External id": 264646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264646,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 264646, "pid": 5, "tid": 7, "ts": 1716454225669821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613023, "dur": 10,
    "args": {
      "External id": 264646,
      "cbid": 211, "correlation": 264646
    }
  },
  {
    "ph": "s", "id": 264646, "pid": 76337, "tid": -914061504, "ts": 1716454225613023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613086, "dur": 1,
    "args": {
      "External id": 264662,
      "cbid": 251, "correlation": 264662
    }
  },
  {
    "ph": "f", "id": 264662, "pid": 76337, "tid": -914061504, "ts": 1716454225613086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613091, "dur": 0,
    "args": {
      "External id": 264664,
      "cbid": 251, "correlation": 264664
    }
  },
  {
    "ph": "f", "id": 264664, "pid": 76337, "tid": -914061504, "ts": 1716454225613091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225669825, "dur": 11,
    "args": {
      "External id": 264665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264665,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264665, "pid": 5, "tid": 7, "ts": 1716454225669825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613093, "dur": 11,
    "args": {
      "External id": 264665,
      "cbid": 211, "correlation": 264665
    }
  },
  {
    "ph": "s", "id": 264665, "pid": 76337, "tid": -914061504, "ts": 1716454225613093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225669838, "dur": 5,
    "args": {
      "External id": 264667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264667,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264667, "pid": 5, "tid": 7, "ts": 1716454225669838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613106, "dur": 5,
    "args": {
      "External id": 264667,
      "cbid": 211, "correlation": 264667
    }
  },
  {
    "ph": "s", "id": 264667, "pid": 76337, "tid": -914061504, "ts": 1716454225613106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225669844, "dur": 6,
    "args": {
      "External id": 264677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264677, "pid": 5, "tid": 7, "ts": 1716454225669844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613163, "dur": 13,
    "args": {
      "External id": 264677,
      "cbid": 211, "correlation": 264677
    }
  },
  {
    "ph": "s", "id": 264677, "pid": 76337, "tid": -914061504, "ts": 1716454225613163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225669851, "dur": 10,
    "args": {
      "External id": 264697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264697,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 264697, "pid": 5, "tid": 7, "ts": 1716454225669851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613229, "dur": 11,
    "args": {
      "External id": 264697,
      "cbid": 211, "correlation": 264697
    }
  },
  {
    "ph": "s", "id": 264697, "pid": 76337, "tid": -914061504, "ts": 1716454225613229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225669862, "dur": 4,
    "args": {
      "External id": 264709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264709,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 264709, "pid": 5, "tid": 7, "ts": 1716454225669862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613250, "dur": 6,
    "args": {
      "External id": 264709,
      "cbid": 211, "correlation": 264709
    }
  },
  {
    "ph": "s", "id": 264709, "pid": 76337, "tid": -914061504, "ts": 1716454225613250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225669867, "dur": 7,
    "args": {
      "External id": 264712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264712, "pid": 5, "tid": 7, "ts": 1716454225669867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613268, "dur": 7,
    "args": {
      "External id": 264712,
      "cbid": 211, "correlation": 264712
    }
  },
  {
    "ph": "s", "id": 264712, "pid": 76337, "tid": -914061504, "ts": 1716454225613268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225669875, "dur": 5,
    "args": {
      "External id": 264721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264721,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264721, "pid": 5, "tid": 7, "ts": 1716454225669875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613309, "dur": 10,
    "args": {
      "External id": 264721,
      "cbid": 211, "correlation": 264721
    }
  },
  {
    "ph": "s", "id": 264721, "pid": 76337, "tid": -914061504, "ts": 1716454225613309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613371, "dur": 0,
    "args": {
      "External id": 264731,
      "cbid": 317, "correlation": 264731
    }
  },
  {
    "ph": "f", "id": 264731, "pid": 76337, "tid": -914061504, "ts": 1716454225613371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613372, "dur": 0,
    "args": {
      "External id": 264732,
      "cbid": 203, "correlation": 264732
    }
  },
  {
    "ph": "f", "id": 264732, "pid": 76337, "tid": -914061504, "ts": 1716454225613372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613372, "dur": 0,
    "args": {
      "External id": 264733,
      "cbid": 205, "correlation": 264733
    }
  },
  {
    "ph": "f", "id": 264733, "pid": 76337, "tid": -914061504, "ts": 1716454225613372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225669881, "dur": 5,
    "args": {
      "External id": 264737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264737,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264737, "pid": 5, "tid": 7, "ts": 1716454225669881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613387, "dur": 12,
    "args": {
      "External id": 264737,
      "cbid": 211, "correlation": 264737
    }
  },
  {
    "ph": "s", "id": 264737, "pid": 76337, "tid": -914061504, "ts": 1716454225613387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225669887, "dur": 165,
    "args": {
      "External id": 264739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264739, "pid": 5, "tid": 7, "ts": 1716454225669887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613401, "dur": 5,
    "args": {
      "External id": 264739,
      "cbid": 211, "correlation": 264739
    }
  },
  {
    "ph": "s", "id": 264739, "pid": 76337, "tid": -914061504, "ts": 1716454225613401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225670054, "dur": 1,
    "args": {
      "External id": 264741,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264741,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 264741, "pid": 5, "tid": 7, "ts": 1716454225670054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613411, "dur": 6,
    "args": {
      "External id": 264741,
      "cbid": 51, "correlation": 264741
    }
  },
  {
    "ph": "s", "id": 264741, "pid": 76337, "tid": -914061504, "ts": 1716454225613411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225670058, "dur": 263,
    "args": {
      "External id": 264742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264742,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264742, "pid": 5, "tid": 7, "ts": 1716454225670058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613419, "dur": 6,
    "args": {
      "External id": 264742,
      "cbid": 211, "correlation": 264742
    }
  },
  {
    "ph": "s", "id": 264742, "pid": 76337, "tid": -914061504, "ts": 1716454225613419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225670322, "dur": 6,
    "args": {
      "External id": 264744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264744, "pid": 5, "tid": 7, "ts": 1716454225670322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613429, "dur": 6,
    "args": {
      "External id": 264744,
      "cbid": 211, "correlation": 264744
    }
  },
  {
    "ph": "s", "id": 264744, "pid": 76337, "tid": -914061504, "ts": 1716454225613429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225670330, "dur": 6,
    "args": {
      "External id": 264750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264750, "pid": 5, "tid": 7, "ts": 1716454225670330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613457, "dur": 9,
    "args": {
      "External id": 264750,
      "cbid": 211, "correlation": 264750
    }
  },
  {
    "ph": "s", "id": 264750, "pid": 76337, "tid": -914061504, "ts": 1716454225613457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613517, "dur": 0,
    "args": {
      "External id": 264760,
      "cbid": 317, "correlation": 264760
    }
  },
  {
    "ph": "f", "id": 264760, "pid": 76337, "tid": -914061504, "ts": 1716454225613517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613517, "dur": 0,
    "args": {
      "External id": 264761,
      "cbid": 203, "correlation": 264761
    }
  },
  {
    "ph": "f", "id": 264761, "pid": 76337, "tid": -914061504, "ts": 1716454225613517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613518, "dur": 0,
    "args": {
      "External id": 264762,
      "cbid": 205, "correlation": 264762
    }
  },
  {
    "ph": "f", "id": 264762, "pid": 76337, "tid": -914061504, "ts": 1716454225613518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225670337, "dur": 8,
    "args": {
      "External id": 264766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264766, "pid": 5, "tid": 7, "ts": 1716454225670337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613533, "dur": 12,
    "args": {
      "External id": 264766,
      "cbid": 211, "correlation": 264766
    }
  },
  {
    "ph": "s", "id": 264766, "pid": 76337, "tid": -914061504, "ts": 1716454225613533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225670347, "dur": 3,
    "args": {
      "External id": 264768,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264768,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 264768, "pid": 5, "tid": 7, "ts": 1716454225670347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613551, "dur": 15,
    "args": {
      "External id": 264768,
      "cbid": 51, "correlation": 264768
    }
  },
  {
    "ph": "s", "id": 264768, "pid": 76337, "tid": -914061504, "ts": 1716454225613551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225670351, "dur": 97,
    "args": {
      "External id": 264769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264769,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 264769, "pid": 5, "tid": 7, "ts": 1716454225670351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613567, "dur": 7,
    "args": {
      "External id": 264769,
      "cbid": 211, "correlation": 264769
    }
  },
  {
    "ph": "s", "id": 264769, "pid": 76337, "tid": -914061504, "ts": 1716454225613567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225670449, "dur": 6,
    "args": {
      "External id": 264771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264771,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264771, "pid": 5, "tid": 7, "ts": 1716454225670449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613578, "dur": 6,
    "args": {
      "External id": 264771,
      "cbid": 211, "correlation": 264771
    }
  },
  {
    "ph": "s", "id": 264771, "pid": 76337, "tid": -914061504, "ts": 1716454225613578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225670456, "dur": 6,
    "args": {
      "External id": 264777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264777, "pid": 5, "tid": 7, "ts": 1716454225670456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613605, "dur": 8,
    "args": {
      "External id": 264777,
      "cbid": 211, "correlation": 264777
    }
  },
  {
    "ph": "s", "id": 264777, "pid": 76337, "tid": -914061504, "ts": 1716454225613605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225670464, "dur": 5,
    "args": {
      "External id": 264785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264785,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264785, "pid": 5, "tid": 7, "ts": 1716454225670464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613635, "dur": 8,
    "args": {
      "External id": 264785,
      "cbid": 211, "correlation": 264785
    }
  },
  {
    "ph": "s", "id": 264785, "pid": 76337, "tid": -914061504, "ts": 1716454225613635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225670470, "dur": 5,
    "args": {
      "External id": 264793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264793,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264793, "pid": 5, "tid": 7, "ts": 1716454225670470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613663, "dur": 8,
    "args": {
      "External id": 264793,
      "cbid": 211, "correlation": 264793
    }
  },
  {
    "ph": "s", "id": 264793, "pid": 76337, "tid": -914061504, "ts": 1716454225613663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225670476, "dur": 11,
    "args": {
      "External id": 264802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264802,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264802, "pid": 5, "tid": 7, "ts": 1716454225670476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613707, "dur": 10,
    "args": {
      "External id": 264802,
      "cbid": 211, "correlation": 264802
    }
  },
  {
    "ph": "s", "id": 264802, "pid": 76337, "tid": -914061504, "ts": 1716454225613707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225670489, "dur": 12,
    "args": {
      "External id": 264822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264822,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 264822, "pid": 5, "tid": 7, "ts": 1716454225670489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613776, "dur": 12,
    "args": {
      "External id": 264822,
      "cbid": 211, "correlation": 264822
    }
  },
  {
    "ph": "s", "id": 264822, "pid": 76337, "tid": -914061504, "ts": 1716454225613776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225670503, "dur": 4,
    "args": {
      "External id": 264834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264834,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264834, "pid": 5, "tid": 7, "ts": 1716454225670503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613797, "dur": 6,
    "args": {
      "External id": 264834,
      "cbid": 211, "correlation": 264834
    }
  },
  {
    "ph": "s", "id": 264834, "pid": 76337, "tid": -914061504, "ts": 1716454225613797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225670508, "dur": 11,
    "args": {
      "External id": 264837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264837,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264837, "pid": 5, "tid": 7, "ts": 1716454225670508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613815, "dur": 6,
    "args": {
      "External id": 264837,
      "cbid": 211, "correlation": 264837
    }
  },
  {
    "ph": "s", "id": 264837, "pid": 76337, "tid": -914061504, "ts": 1716454225613815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225670521, "dur": 6,
    "args": {
      "External id": 264846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264846,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264846, "pid": 5, "tid": 7, "ts": 1716454225670521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613853, "dur": 9,
    "args": {
      "External id": 264846,
      "cbid": 211, "correlation": 264846
    }
  },
  {
    "ph": "s", "id": 264846, "pid": 76337, "tid": -914061504, "ts": 1716454225613853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613905, "dur": 0,
    "args": {
      "External id": 264856,
      "cbid": 317, "correlation": 264856
    }
  },
  {
    "ph": "f", "id": 264856, "pid": 76337, "tid": -914061504, "ts": 1716454225613905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613906, "dur": 0,
    "args": {
      "External id": 264857,
      "cbid": 203, "correlation": 264857
    }
  },
  {
    "ph": "f", "id": 264857, "pid": 76337, "tid": -914061504, "ts": 1716454225613906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613906, "dur": 0,
    "args": {
      "External id": 264858,
      "cbid": 205, "correlation": 264858
    }
  },
  {
    "ph": "f", "id": 264858, "pid": 76337, "tid": -914061504, "ts": 1716454225613906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225670528, "dur": 7,
    "args": {
      "External id": 264862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264862,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264862, "pid": 5, "tid": 7, "ts": 1716454225670528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613919, "dur": 11,
    "args": {
      "External id": 264862,
      "cbid": 211, "correlation": 264862
    }
  },
  {
    "ph": "s", "id": 264862, "pid": 76337, "tid": -914061504, "ts": 1716454225613919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225670536, "dur": 325,
    "args": {
      "External id": 264864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264864,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264864, "pid": 5, "tid": 7, "ts": 1716454225670536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613933, "dur": 5,
    "args": {
      "External id": 264864,
      "cbid": 211, "correlation": 264864
    }
  },
  {
    "ph": "s", "id": 264864, "pid": 76337, "tid": -914061504, "ts": 1716454225613933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225670863, "dur": 1,
    "args": {
      "External id": 264866,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264866,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 264866, "pid": 5, "tid": 7, "ts": 1716454225670863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613944, "dur": 6,
    "args": {
      "External id": 264866,
      "cbid": 51, "correlation": 264866
    }
  },
  {
    "ph": "s", "id": 264866, "pid": 76337, "tid": -914061504, "ts": 1716454225613944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225670867, "dur": 504,
    "args": {
      "External id": 264867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264867,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264867, "pid": 5, "tid": 7, "ts": 1716454225670867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613951, "dur": 6,
    "args": {
      "External id": 264867,
      "cbid": 211, "correlation": 264867
    }
  },
  {
    "ph": "s", "id": 264867, "pid": 76337, "tid": -914061504, "ts": 1716454225613951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225671372, "dur": 5,
    "args": {
      "External id": 264869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264869,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264869, "pid": 5, "tid": 7, "ts": 1716454225671372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613961, "dur": 5,
    "args": {
      "External id": 264869,
      "cbid": 211, "correlation": 264869
    }
  },
  {
    "ph": "s", "id": 264869, "pid": 76337, "tid": -914061504, "ts": 1716454225613961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225671379, "dur": 6,
    "args": {
      "External id": 264875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264875, "pid": 5, "tid": 7, "ts": 1716454225671379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225613997, "dur": 9,
    "args": {
      "External id": 264875,
      "cbid": 211, "correlation": 264875
    }
  },
  {
    "ph": "s", "id": 264875, "pid": 76337, "tid": -914061504, "ts": 1716454225613997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225671387, "dur": 3,
    "args": {
      "External id": 264883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264883,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 264883, "pid": 5, "tid": 7, "ts": 1716454225671387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614043, "dur": 9,
    "args": {
      "External id": 264883,
      "cbid": 211, "correlation": 264883
    }
  },
  {
    "ph": "s", "id": 264883, "pid": 76337, "tid": -914061504, "ts": 1716454225614043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614104, "dur": 1,
    "args": {
      "External id": 264899,
      "cbid": 251, "correlation": 264899
    }
  },
  {
    "ph": "f", "id": 264899, "pid": 76337, "tid": -914061504, "ts": 1716454225614104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614109, "dur": 0,
    "args": {
      "External id": 264901,
      "cbid": 251, "correlation": 264901
    }
  },
  {
    "ph": "f", "id": 264901, "pid": 76337, "tid": -914061504, "ts": 1716454225614109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225671391, "dur": 13,
    "args": {
      "External id": 264902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264902,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264902, "pid": 5, "tid": 7, "ts": 1716454225671391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614111, "dur": 11,
    "args": {
      "External id": 264902,
      "cbid": 211, "correlation": 264902
    }
  },
  {
    "ph": "s", "id": 264902, "pid": 76337, "tid": -914061504, "ts": 1716454225614111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225671405, "dur": 5,
    "args": {
      "External id": 264904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264904,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264904, "pid": 5, "tid": 7, "ts": 1716454225671405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614123, "dur": 5,
    "args": {
      "External id": 264904,
      "cbid": 211, "correlation": 264904
    }
  },
  {
    "ph": "s", "id": 264904, "pid": 76337, "tid": -914061504, "ts": 1716454225614123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225671412, "dur": 6,
    "args": {
      "External id": 264914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264914, "pid": 5, "tid": 7, "ts": 1716454225671412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614180, "dur": 13,
    "args": {
      "External id": 264914,
      "cbid": 211, "correlation": 264914
    }
  },
  {
    "ph": "s", "id": 264914, "pid": 76337, "tid": -914061504, "ts": 1716454225614180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225671419, "dur": 10,
    "args": {
      "External id": 264934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264934,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 264934, "pid": 5, "tid": 7, "ts": 1716454225671419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614246, "dur": 11,
    "args": {
      "External id": 264934,
      "cbid": 211, "correlation": 264934
    }
  },
  {
    "ph": "s", "id": 264934, "pid": 76337, "tid": -914061504, "ts": 1716454225614246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225671430, "dur": 4,
    "args": {
      "External id": 264946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264946,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 264946, "pid": 5, "tid": 7, "ts": 1716454225671430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614266, "dur": 6,
    "args": {
      "External id": 264946,
      "cbid": 211, "correlation": 264946
    }
  },
  {
    "ph": "s", "id": 264946, "pid": 76337, "tid": -914061504, "ts": 1716454225614266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225671435, "dur": 7,
    "args": {
      "External id": 264949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264949,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264949, "pid": 5, "tid": 7, "ts": 1716454225671435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614285, "dur": 6,
    "args": {
      "External id": 264949,
      "cbid": 211, "correlation": 264949
    }
  },
  {
    "ph": "s", "id": 264949, "pid": 76337, "tid": -914061504, "ts": 1716454225614285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225671443, "dur": 5,
    "args": {
      "External id": 264958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264958,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264958, "pid": 5, "tid": 7, "ts": 1716454225671443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614325, "dur": 10,
    "args": {
      "External id": 264958,
      "cbid": 211, "correlation": 264958
    }
  },
  {
    "ph": "s", "id": 264958, "pid": 76337, "tid": -914061504, "ts": 1716454225614325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614388, "dur": 0,
    "args": {
      "External id": 264968,
      "cbid": 317, "correlation": 264968
    }
  },
  {
    "ph": "f", "id": 264968, "pid": 76337, "tid": -914061504, "ts": 1716454225614388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614389, "dur": 0,
    "args": {
      "External id": 264969,
      "cbid": 203, "correlation": 264969
    }
  },
  {
    "ph": "f", "id": 264969, "pid": 76337, "tid": -914061504, "ts": 1716454225614389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614389, "dur": 0,
    "args": {
      "External id": 264970,
      "cbid": 205, "correlation": 264970
    }
  },
  {
    "ph": "f", "id": 264970, "pid": 76337, "tid": -914061504, "ts": 1716454225614389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225671449, "dur": 5,
    "args": {
      "External id": 264974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264974,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264974, "pid": 5, "tid": 7, "ts": 1716454225671449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614404, "dur": 12,
    "args": {
      "External id": 264974,
      "cbid": 211, "correlation": 264974
    }
  },
  {
    "ph": "s", "id": 264974, "pid": 76337, "tid": -914061504, "ts": 1716454225614404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225671455, "dur": 165,
    "args": {
      "External id": 264976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264976,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264976, "pid": 5, "tid": 7, "ts": 1716454225671455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614418, "dur": 5,
    "args": {
      "External id": 264976,
      "cbid": 211, "correlation": 264976
    }
  },
  {
    "ph": "s", "id": 264976, "pid": 76337, "tid": -914061504, "ts": 1716454225614418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225671622, "dur": 1,
    "args": {
      "External id": 264978,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 264978,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 264978, "pid": 5, "tid": 7, "ts": 1716454225671622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614429, "dur": 6,
    "args": {
      "External id": 264978,
      "cbid": 51, "correlation": 264978
    }
  },
  {
    "ph": "s", "id": 264978, "pid": 76337, "tid": -914061504, "ts": 1716454225614429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225671626, "dur": 264,
    "args": {
      "External id": 264979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264979,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 264979, "pid": 5, "tid": 7, "ts": 1716454225671626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614436, "dur": 6,
    "args": {
      "External id": 264979,
      "cbid": 211, "correlation": 264979
    }
  },
  {
    "ph": "s", "id": 264979, "pid": 76337, "tid": -914061504, "ts": 1716454225614436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225671891, "dur": 6,
    "args": {
      "External id": 264981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 264981, "pid": 5, "tid": 7, "ts": 1716454225671891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614445, "dur": 6,
    "args": {
      "External id": 264981,
      "cbid": 211, "correlation": 264981
    }
  },
  {
    "ph": "s", "id": 264981, "pid": 76337, "tid": -914061504, "ts": 1716454225614445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225671898, "dur": 6,
    "args": {
      "External id": 264987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 264987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 264987, "pid": 5, "tid": 7, "ts": 1716454225671898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614474, "dur": 9,
    "args": {
      "External id": 264987,
      "cbid": 211, "correlation": 264987
    }
  },
  {
    "ph": "s", "id": 264987, "pid": 76337, "tid": -914061504, "ts": 1716454225614474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614533, "dur": 0,
    "args": {
      "External id": 264997,
      "cbid": 317, "correlation": 264997
    }
  },
  {
    "ph": "f", "id": 264997, "pid": 76337, "tid": -914061504, "ts": 1716454225614533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614534, "dur": 0,
    "args": {
      "External id": 264998,
      "cbid": 203, "correlation": 264998
    }
  },
  {
    "ph": "f", "id": 264998, "pid": 76337, "tid": -914061504, "ts": 1716454225614534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614535, "dur": 0,
    "args": {
      "External id": 264999,
      "cbid": 205, "correlation": 264999
    }
  },
  {
    "ph": "f", "id": 264999, "pid": 76337, "tid": -914061504, "ts": 1716454225614535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225671906, "dur": 8,
    "args": {
      "External id": 265003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265003,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265003, "pid": 5, "tid": 7, "ts": 1716454225671906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614547, "dur": 12,
    "args": {
      "External id": 265003,
      "cbid": 211, "correlation": 265003
    }
  },
  {
    "ph": "s", "id": 265003, "pid": 76337, "tid": -914061504, "ts": 1716454225614547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225671915, "dur": 3,
    "args": {
      "External id": 265005,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265005,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 265005, "pid": 5, "tid": 7, "ts": 1716454225671915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614564, "dur": 10,
    "args": {
      "External id": 265005,
      "cbid": 51, "correlation": 265005
    }
  },
  {
    "ph": "s", "id": 265005, "pid": 76337, "tid": -914061504, "ts": 1716454225614564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225671919, "dur": 96,
    "args": {
      "External id": 265006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265006,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 265006, "pid": 5, "tid": 7, "ts": 1716454225671919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614575, "dur": 6,
    "args": {
      "External id": 265006,
      "cbid": 211, "correlation": 265006
    }
  },
  {
    "ph": "s", "id": 265006, "pid": 76337, "tid": -914061504, "ts": 1716454225614575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225672017, "dur": 6,
    "args": {
      "External id": 265008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265008,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265008, "pid": 5, "tid": 7, "ts": 1716454225672017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614584, "dur": 5,
    "args": {
      "External id": 265008,
      "cbid": 211, "correlation": 265008
    }
  },
  {
    "ph": "s", "id": 265008, "pid": 76337, "tid": -914061504, "ts": 1716454225614584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225672024, "dur": 6,
    "args": {
      "External id": 265014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265014,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265014, "pid": 5, "tid": 7, "ts": 1716454225672024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614612, "dur": 8,
    "args": {
      "External id": 265014,
      "cbid": 211, "correlation": 265014
    }
  },
  {
    "ph": "s", "id": 265014, "pid": 76337, "tid": -914061504, "ts": 1716454225614612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225672031, "dur": 5,
    "args": {
      "External id": 265022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265022,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265022, "pid": 5, "tid": 7, "ts": 1716454225672031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614641, "dur": 9,
    "args": {
      "External id": 265022,
      "cbid": 211, "correlation": 265022
    }
  },
  {
    "ph": "s", "id": 265022, "pid": 76337, "tid": -914061504, "ts": 1716454225614641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225672038, "dur": 5,
    "args": {
      "External id": 265030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265030,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265030, "pid": 5, "tid": 7, "ts": 1716454225672038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614670, "dur": 8,
    "args": {
      "External id": 265030,
      "cbid": 211, "correlation": 265030
    }
  },
  {
    "ph": "s", "id": 265030, "pid": 76337, "tid": -914061504, "ts": 1716454225614670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225672044, "dur": 11,
    "args": {
      "External id": 265039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265039,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265039, "pid": 5, "tid": 7, "ts": 1716454225672044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614713, "dur": 10,
    "args": {
      "External id": 265039,
      "cbid": 211, "correlation": 265039
    }
  },
  {
    "ph": "s", "id": 265039, "pid": 76337, "tid": -914061504, "ts": 1716454225614713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225672056, "dur": 13,
    "args": {
      "External id": 265059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265059,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 265059, "pid": 5, "tid": 7, "ts": 1716454225672056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614783, "dur": 11,
    "args": {
      "External id": 265059,
      "cbid": 211, "correlation": 265059
    }
  },
  {
    "ph": "s", "id": 265059, "pid": 76337, "tid": -914061504, "ts": 1716454225614783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225672070, "dur": 4,
    "args": {
      "External id": 265071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265071,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265071, "pid": 5, "tid": 7, "ts": 1716454225672070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614805, "dur": 7,
    "args": {
      "External id": 265071,
      "cbid": 211, "correlation": 265071
    }
  },
  {
    "ph": "s", "id": 265071, "pid": 76337, "tid": -914061504, "ts": 1716454225614805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225672075, "dur": 11,
    "args": {
      "External id": 265074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265074, "pid": 5, "tid": 7, "ts": 1716454225672075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614823, "dur": 7,
    "args": {
      "External id": 265074,
      "cbid": 211, "correlation": 265074
    }
  },
  {
    "ph": "s", "id": 265074, "pid": 76337, "tid": -914061504, "ts": 1716454225614823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225672088, "dur": 6,
    "args": {
      "External id": 265083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265083,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265083, "pid": 5, "tid": 7, "ts": 1716454225672088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614863, "dur": 9,
    "args": {
      "External id": 265083,
      "cbid": 211, "correlation": 265083
    }
  },
  {
    "ph": "s", "id": 265083, "pid": 76337, "tid": -914061504, "ts": 1716454225614863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614915, "dur": 0,
    "args": {
      "External id": 265093,
      "cbid": 317, "correlation": 265093
    }
  },
  {
    "ph": "f", "id": 265093, "pid": 76337, "tid": -914061504, "ts": 1716454225614915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614916, "dur": 0,
    "args": {
      "External id": 265094,
      "cbid": 203, "correlation": 265094
    }
  },
  {
    "ph": "f", "id": 265094, "pid": 76337, "tid": -914061504, "ts": 1716454225614916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614917, "dur": 0,
    "args": {
      "External id": 265095,
      "cbid": 205, "correlation": 265095
    }
  },
  {
    "ph": "f", "id": 265095, "pid": 76337, "tid": -914061504, "ts": 1716454225614917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225672095, "dur": 7,
    "args": {
      "External id": 265099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265099,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265099, "pid": 5, "tid": 7, "ts": 1716454225672095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614930, "dur": 11,
    "args": {
      "External id": 265099,
      "cbid": 211, "correlation": 265099
    }
  },
  {
    "ph": "s", "id": 265099, "pid": 76337, "tid": -914061504, "ts": 1716454225614930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225672103, "dur": 326,
    "args": {
      "External id": 265101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265101,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265101, "pid": 5, "tid": 7, "ts": 1716454225672103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614943, "dur": 5,
    "args": {
      "External id": 265101,
      "cbid": 211, "correlation": 265101
    }
  },
  {
    "ph": "s", "id": 265101, "pid": 76337, "tid": -914061504, "ts": 1716454225614943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225672431, "dur": 1,
    "args": {
      "External id": 265103,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265103,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 265103, "pid": 5, "tid": 7, "ts": 1716454225672431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614954, "dur": 7,
    "args": {
      "External id": 265103,
      "cbid": 51, "correlation": 265103
    }
  },
  {
    "ph": "s", "id": 265103, "pid": 76337, "tid": -914061504, "ts": 1716454225614954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225672435, "dur": 504,
    "args": {
      "External id": 265104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265104,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265104, "pid": 5, "tid": 7, "ts": 1716454225672435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614962, "dur": 6,
    "args": {
      "External id": 265104,
      "cbid": 211, "correlation": 265104
    }
  },
  {
    "ph": "s", "id": 265104, "pid": 76337, "tid": -914061504, "ts": 1716454225614962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225672940, "dur": 6,
    "args": {
      "External id": 265106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265106,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265106, "pid": 5, "tid": 7, "ts": 1716454225672940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225614972, "dur": 14,
    "args": {
      "External id": 265106,
      "cbid": 211, "correlation": 265106
    }
  },
  {
    "ph": "s", "id": 265106, "pid": 76337, "tid": -914061504, "ts": 1716454225614972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225672948, "dur": 6,
    "args": {
      "External id": 265112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265112, "pid": 5, "tid": 7, "ts": 1716454225672948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615010, "dur": 9,
    "args": {
      "External id": 265112,
      "cbid": 211, "correlation": 265112
    }
  },
  {
    "ph": "s", "id": 265112, "pid": 76337, "tid": -914061504, "ts": 1716454225615010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225672955, "dur": 3,
    "args": {
      "External id": 265120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265120,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 265120, "pid": 5, "tid": 7, "ts": 1716454225672955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615054, "dur": 11,
    "args": {
      "External id": 265120,
      "cbid": 211, "correlation": 265120
    }
  },
  {
    "ph": "s", "id": 265120, "pid": 76337, "tid": -914061504, "ts": 1716454225615054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615117, "dur": 1,
    "args": {
      "External id": 265136,
      "cbid": 251, "correlation": 265136
    }
  },
  {
    "ph": "f", "id": 265136, "pid": 76337, "tid": -914061504, "ts": 1716454225615117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615122, "dur": 0,
    "args": {
      "External id": 265138,
      "cbid": 251, "correlation": 265138
    }
  },
  {
    "ph": "f", "id": 265138, "pid": 76337, "tid": -914061504, "ts": 1716454225615122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225672960, "dur": 13,
    "args": {
      "External id": 265139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265139,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265139, "pid": 5, "tid": 7, "ts": 1716454225672960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615124, "dur": 11,
    "args": {
      "External id": 265139,
      "cbid": 211, "correlation": 265139
    }
  },
  {
    "ph": "s", "id": 265139, "pid": 76337, "tid": -914061504, "ts": 1716454225615124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225672974, "dur": 5,
    "args": {
      "External id": 265141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265141,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265141, "pid": 5, "tid": 7, "ts": 1716454225672974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615137, "dur": 5,
    "args": {
      "External id": 265141,
      "cbid": 211, "correlation": 265141
    }
  },
  {
    "ph": "s", "id": 265141, "pid": 76337, "tid": -914061504, "ts": 1716454225615137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225672980, "dur": 6,
    "args": {
      "External id": 265151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265151, "pid": 5, "tid": 7, "ts": 1716454225672980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615194, "dur": 12,
    "args": {
      "External id": 265151,
      "cbid": 211, "correlation": 265151
    }
  },
  {
    "ph": "s", "id": 265151, "pid": 76337, "tid": -914061504, "ts": 1716454225615194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225672988, "dur": 10,
    "args": {
      "External id": 265171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265171,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 265171, "pid": 5, "tid": 7, "ts": 1716454225672988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615259, "dur": 11,
    "args": {
      "External id": 265171,
      "cbid": 211, "correlation": 265171
    }
  },
  {
    "ph": "s", "id": 265171, "pid": 76337, "tid": -914061504, "ts": 1716454225615259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225672999, "dur": 4,
    "args": {
      "External id": 265183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265183,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 265183, "pid": 5, "tid": 7, "ts": 1716454225672999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615279, "dur": 6,
    "args": {
      "External id": 265183,
      "cbid": 211, "correlation": 265183
    }
  },
  {
    "ph": "s", "id": 265183, "pid": 76337, "tid": -914061504, "ts": 1716454225615279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225673004, "dur": 7,
    "args": {
      "External id": 265186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265186, "pid": 5, "tid": 7, "ts": 1716454225673004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615298, "dur": 6,
    "args": {
      "External id": 265186,
      "cbid": 211, "correlation": 265186
    }
  },
  {
    "ph": "s", "id": 265186, "pid": 76337, "tid": -914061504, "ts": 1716454225615298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225673012, "dur": 5,
    "args": {
      "External id": 265195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265195,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265195, "pid": 5, "tid": 7, "ts": 1716454225673012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615337, "dur": 10,
    "args": {
      "External id": 265195,
      "cbid": 211, "correlation": 265195
    }
  },
  {
    "ph": "s", "id": 265195, "pid": 76337, "tid": -914061504, "ts": 1716454225615337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615400, "dur": 0,
    "args": {
      "External id": 265205,
      "cbid": 317, "correlation": 265205
    }
  },
  {
    "ph": "f", "id": 265205, "pid": 76337, "tid": -914061504, "ts": 1716454225615400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615401, "dur": 0,
    "args": {
      "External id": 265206,
      "cbid": 203, "correlation": 265206
    }
  },
  {
    "ph": "f", "id": 265206, "pid": 76337, "tid": -914061504, "ts": 1716454225615401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615401, "dur": 0,
    "args": {
      "External id": 265207,
      "cbid": 205, "correlation": 265207
    }
  },
  {
    "ph": "f", "id": 265207, "pid": 76337, "tid": -914061504, "ts": 1716454225615401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673018, "dur": 5,
    "args": {
      "External id": 265211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265211,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265211, "pid": 5, "tid": 7, "ts": 1716454225673018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615416, "dur": 12,
    "args": {
      "External id": 265211,
      "cbid": 211, "correlation": 265211
    }
  },
  {
    "ph": "s", "id": 265211, "pid": 76337, "tid": -914061504, "ts": 1716454225615416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673025, "dur": 165,
    "args": {
      "External id": 265213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265213,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265213, "pid": 5, "tid": 7, "ts": 1716454225673025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615430, "dur": 5,
    "args": {
      "External id": 265213,
      "cbid": 211, "correlation": 265213
    }
  },
  {
    "ph": "s", "id": 265213, "pid": 76337, "tid": -914061504, "ts": 1716454225615430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225673192, "dur": 1,
    "args": {
      "External id": 265215,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265215,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 265215, "pid": 5, "tid": 7, "ts": 1716454225673192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615441, "dur": 6,
    "args": {
      "External id": 265215,
      "cbid": 51, "correlation": 265215
    }
  },
  {
    "ph": "s", "id": 265215, "pid": 76337, "tid": -914061504, "ts": 1716454225615441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225673195, "dur": 264,
    "args": {
      "External id": 265216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265216,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265216, "pid": 5, "tid": 7, "ts": 1716454225673195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615448, "dur": 6,
    "args": {
      "External id": 265216,
      "cbid": 211, "correlation": 265216
    }
  },
  {
    "ph": "s", "id": 265216, "pid": 76337, "tid": -914061504, "ts": 1716454225615448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673460, "dur": 6,
    "args": {
      "External id": 265218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265218,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265218, "pid": 5, "tid": 7, "ts": 1716454225673460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615457, "dur": 6,
    "args": {
      "External id": 265218,
      "cbid": 211, "correlation": 265218
    }
  },
  {
    "ph": "s", "id": 265218, "pid": 76337, "tid": -914061504, "ts": 1716454225615457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225673468, "dur": 6,
    "args": {
      "External id": 265224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265224,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265224, "pid": 5, "tid": 7, "ts": 1716454225673468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615486, "dur": 8,
    "args": {
      "External id": 265224,
      "cbid": 211, "correlation": 265224
    }
  },
  {
    "ph": "s", "id": 265224, "pid": 76337, "tid": -914061504, "ts": 1716454225615486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615544, "dur": 0,
    "args": {
      "External id": 265234,
      "cbid": 317, "correlation": 265234
    }
  },
  {
    "ph": "f", "id": 265234, "pid": 76337, "tid": -914061504, "ts": 1716454225615544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615545, "dur": 0,
    "args": {
      "External id": 265235,
      "cbid": 203, "correlation": 265235
    }
  },
  {
    "ph": "f", "id": 265235, "pid": 76337, "tid": -914061504, "ts": 1716454225615545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615545, "dur": 0,
    "args": {
      "External id": 265236,
      "cbid": 205, "correlation": 265236
    }
  },
  {
    "ph": "f", "id": 265236, "pid": 76337, "tid": -914061504, "ts": 1716454225615545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673475, "dur": 8,
    "args": {
      "External id": 265240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265240, "pid": 5, "tid": 7, "ts": 1716454225673475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615557, "dur": 12,
    "args": {
      "External id": 265240,
      "cbid": 211, "correlation": 265240
    }
  },
  {
    "ph": "s", "id": 265240, "pid": 76337, "tid": -914061504, "ts": 1716454225615557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225673484, "dur": 3,
    "args": {
      "External id": 265242,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265242,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 265242, "pid": 5, "tid": 7, "ts": 1716454225673484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615574, "dur": 9,
    "args": {
      "External id": 265242,
      "cbid": 51, "correlation": 265242
    }
  },
  {
    "ph": "s", "id": 265242, "pid": 76337, "tid": -914061504, "ts": 1716454225615574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225673488, "dur": 95,
    "args": {
      "External id": 265243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265243,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 265243, "pid": 5, "tid": 7, "ts": 1716454225673488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615584, "dur": 6,
    "args": {
      "External id": 265243,
      "cbid": 211, "correlation": 265243
    }
  },
  {
    "ph": "s", "id": 265243, "pid": 76337, "tid": -914061504, "ts": 1716454225615584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673585, "dur": 5,
    "args": {
      "External id": 265245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265245,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265245, "pid": 5, "tid": 7, "ts": 1716454225673585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615594, "dur": 5,
    "args": {
      "External id": 265245,
      "cbid": 211, "correlation": 265245
    }
  },
  {
    "ph": "s", "id": 265245, "pid": 76337, "tid": -914061504, "ts": 1716454225615594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225673591, "dur": 6,
    "args": {
      "External id": 265251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265251,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265251, "pid": 5, "tid": 7, "ts": 1716454225673591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615621, "dur": 9,
    "args": {
      "External id": 265251,
      "cbid": 211, "correlation": 265251
    }
  },
  {
    "ph": "s", "id": 265251, "pid": 76337, "tid": -914061504, "ts": 1716454225615621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225673599, "dur": 5,
    "args": {
      "External id": 265259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265259,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265259, "pid": 5, "tid": 7, "ts": 1716454225673599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615650, "dur": 8,
    "args": {
      "External id": 265259,
      "cbid": 211, "correlation": 265259
    }
  },
  {
    "ph": "s", "id": 265259, "pid": 76337, "tid": -914061504, "ts": 1716454225615650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225673605, "dur": 4,
    "args": {
      "External id": 265267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265267,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 265267, "pid": 5, "tid": 7, "ts": 1716454225673605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615679, "dur": 8,
    "args": {
      "External id": 265267,
      "cbid": 211, "correlation": 265267
    }
  },
  {
    "ph": "s", "id": 265267, "pid": 76337, "tid": -914061504, "ts": 1716454225615679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225673611, "dur": 14,
    "args": {
      "External id": 265278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265278,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265278, "pid": 5, "tid": 7, "ts": 1716454225673611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615760, "dur": 13,
    "args": {
      "External id": 265278,
      "cbid": 211, "correlation": 265278
    }
  },
  {
    "ph": "s", "id": 265278, "pid": 76337, "tid": -914061504, "ts": 1716454225615760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615817, "dur": 0,
    "args": {
      "External id": 265288,
      "cbid": 317, "correlation": 265288
    }
  },
  {
    "ph": "f", "id": 265288, "pid": 76337, "tid": -914061504, "ts": 1716454225615817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615818, "dur": 0,
    "args": {
      "External id": 265289,
      "cbid": 203, "correlation": 265289
    }
  },
  {
    "ph": "f", "id": 265289, "pid": 76337, "tid": -914061504, "ts": 1716454225615818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615819, "dur": 0,
    "args": {
      "External id": 265290,
      "cbid": 205, "correlation": 265290
    }
  },
  {
    "ph": "f", "id": 265290, "pid": 76337, "tid": -914061504, "ts": 1716454225615819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673626, "dur": 9,
    "args": {
      "External id": 265294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265294,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265294, "pid": 5, "tid": 7, "ts": 1716454225673626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615832, "dur": 11,
    "args": {
      "External id": 265294,
      "cbid": 211, "correlation": 265294
    }
  },
  {
    "ph": "s", "id": 265294, "pid": 76337, "tid": -914061504, "ts": 1716454225615832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225673637, "dur": 166,
    "args": {
      "External id": 265296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265296,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265296, "pid": 5, "tid": 7, "ts": 1716454225673637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615846, "dur": 5,
    "args": {
      "External id": 265296,
      "cbid": 211, "correlation": 265296
    }
  },
  {
    "ph": "s", "id": 265296, "pid": 76337, "tid": -914061504, "ts": 1716454225615846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225673805, "dur": 1,
    "args": {
      "External id": 265298,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265298,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 265298, "pid": 5, "tid": 7, "ts": 1716454225673805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615857, "dur": 7,
    "args": {
      "External id": 265298,
      "cbid": 51, "correlation": 265298
    }
  },
  {
    "ph": "s", "id": 265298, "pid": 76337, "tid": -914061504, "ts": 1716454225615857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225673809, "dur": 659,
    "args": {
      "External id": 265299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265299,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265299, "pid": 5, "tid": 7, "ts": 1716454225673809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615865, "dur": 6,
    "args": {
      "External id": 265299,
      "cbid": 211, "correlation": 265299
    }
  },
  {
    "ph": "s", "id": 265299, "pid": 76337, "tid": -914061504, "ts": 1716454225615865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225674469, "dur": 12,
    "args": {
      "External id": 265301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265301,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265301, "pid": 5, "tid": 7, "ts": 1716454225674469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615875, "dur": 5,
    "args": {
      "External id": 265301,
      "cbid": 211, "correlation": 265301
    }
  },
  {
    "ph": "s", "id": 265301, "pid": 76337, "tid": -914061504, "ts": 1716454225615875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225674482, "dur": 15,
    "args": {
      "External id": 265307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265307,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265307, "pid": 5, "tid": 7, "ts": 1716454225674482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225615904, "dur": 10,
    "args": {
      "External id": 265307,
      "cbid": 211, "correlation": 265307
    }
  },
  {
    "ph": "s", "id": 265307, "pid": 76337, "tid": -914061504, "ts": 1716454225615904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225674498, "dur": 31,
    "args": {
      "External id": 265316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265316,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265316, "pid": 5, "tid": 7, "ts": 1716454225674498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616003, "dur": 14,
    "args": {
      "External id": 265316,
      "cbid": 211, "correlation": 265316
    }
  },
  {
    "ph": "s", "id": 265316, "pid": 76337, "tid": -914061504, "ts": 1716454225616003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225674531, "dur": 30,
    "args": {
      "External id": 265336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265336,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 265336, "pid": 5, "tid": 7, "ts": 1716454225674531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616074, "dur": 11,
    "args": {
      "External id": 265336,
      "cbid": 211, "correlation": 265336
    }
  },
  {
    "ph": "s", "id": 265336, "pid": 76337, "tid": -914061504, "ts": 1716454225616074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225674562, "dur": 4,
    "args": {
      "External id": 265348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265348,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265348, "pid": 5, "tid": 7, "ts": 1716454225674562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616094, "dur": 6,
    "args": {
      "External id": 265348,
      "cbid": 211, "correlation": 265348
    }
  },
  {
    "ph": "s", "id": 265348, "pid": 76337, "tid": -914061504, "ts": 1716454225616094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225674568, "dur": 30,
    "args": {
      "External id": 265351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265351,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265351, "pid": 5, "tid": 7, "ts": 1716454225674568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616113, "dur": 7,
    "args": {
      "External id": 265351,
      "cbid": 211, "correlation": 265351
    }
  },
  {
    "ph": "s", "id": 265351, "pid": 76337, "tid": -914061504, "ts": 1716454225616113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225674599, "dur": 21,
    "args": {
      "External id": 265360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265360,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265360, "pid": 5, "tid": 7, "ts": 1716454225674599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616152, "dur": 9,
    "args": {
      "External id": 265360,
      "cbid": 211, "correlation": 265360
    }
  },
  {
    "ph": "s", "id": 265360, "pid": 76337, "tid": -914061504, "ts": 1716454225616152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616204, "dur": 0,
    "args": {
      "External id": 265370,
      "cbid": 317, "correlation": 265370
    }
  },
  {
    "ph": "f", "id": 265370, "pid": 76337, "tid": -914061504, "ts": 1716454225616204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616205, "dur": 0,
    "args": {
      "External id": 265371,
      "cbid": 203, "correlation": 265371
    }
  },
  {
    "ph": "f", "id": 265371, "pid": 76337, "tid": -914061504, "ts": 1716454225616205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616205, "dur": 0,
    "args": {
      "External id": 265372,
      "cbid": 205, "correlation": 265372
    }
  },
  {
    "ph": "f", "id": 265372, "pid": 76337, "tid": -914061504, "ts": 1716454225616205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225674621, "dur": 22,
    "args": {
      "External id": 265376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265376,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265376, "pid": 5, "tid": 7, "ts": 1716454225674621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616221, "dur": 12,
    "args": {
      "External id": 265376,
      "cbid": 211, "correlation": 265376
    }
  },
  {
    "ph": "s", "id": 265376, "pid": 76337, "tid": -914061504, "ts": 1716454225616221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225674644, "dur": 325,
    "args": {
      "External id": 265378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265378, "pid": 5, "tid": 7, "ts": 1716454225674644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616235, "dur": 5,
    "args": {
      "External id": 265378,
      "cbid": 211, "correlation": 265378
    }
  },
  {
    "ph": "s", "id": 265378, "pid": 76337, "tid": -914061504, "ts": 1716454225616235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225674972, "dur": 1,
    "args": {
      "External id": 265380,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265380,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 265380, "pid": 5, "tid": 7, "ts": 1716454225674972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616246, "dur": 6,
    "args": {
      "External id": 265380,
      "cbid": 51, "correlation": 265380
    }
  },
  {
    "ph": "s", "id": 265380, "pid": 76337, "tid": -914061504, "ts": 1716454225616246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225674975, "dur": 1262,
    "args": {
      "External id": 265381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265381,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265381, "pid": 5, "tid": 7, "ts": 1716454225674975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616253, "dur": 6,
    "args": {
      "External id": 265381,
      "cbid": 211, "correlation": 265381
    }
  },
  {
    "ph": "s", "id": 265381, "pid": 76337, "tid": -914061504, "ts": 1716454225616253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225676239, "dur": 12,
    "args": {
      "External id": 265383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265383,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265383, "pid": 5, "tid": 7, "ts": 1716454225676239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616262, "dur": 5,
    "args": {
      "External id": 265383,
      "cbid": 211, "correlation": 265383
    }
  },
  {
    "ph": "s", "id": 265383, "pid": 76337, "tid": -914061504, "ts": 1716454225616262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225676253, "dur": 15,
    "args": {
      "External id": 265389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265389, "pid": 5, "tid": 7, "ts": 1716454225676253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616291, "dur": 9,
    "args": {
      "External id": 265389,
      "cbid": 211, "correlation": 265389
    }
  },
  {
    "ph": "s", "id": 265389, "pid": 76337, "tid": -914061504, "ts": 1716454225616291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225676269, "dur": 3,
    "args": {
      "External id": 265397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265397,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 265397, "pid": 5, "tid": 7, "ts": 1716454225676269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616335, "dur": 9,
    "args": {
      "External id": 265397,
      "cbid": 211, "correlation": 265397
    }
  },
  {
    "ph": "s", "id": 265397, "pid": 76337, "tid": -914061504, "ts": 1716454225616335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616399, "dur": 1,
    "args": {
      "External id": 265413,
      "cbid": 251, "correlation": 265413
    }
  },
  {
    "ph": "f", "id": 265413, "pid": 76337, "tid": -914061504, "ts": 1716454225616399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616405, "dur": 0,
    "args": {
      "External id": 265415,
      "cbid": 251, "correlation": 265415
    }
  },
  {
    "ph": "f", "id": 265415, "pid": 76337, "tid": -914061504, "ts": 1716454225616405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225676273, "dur": 12,
    "args": {
      "External id": 265416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265416,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265416, "pid": 5, "tid": 7, "ts": 1716454225676273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616406, "dur": 11,
    "args": {
      "External id": 265416,
      "cbid": 211, "correlation": 265416
    }
  },
  {
    "ph": "s", "id": 265416, "pid": 76337, "tid": -914061504, "ts": 1716454225616406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225676287, "dur": 5,
    "args": {
      "External id": 265418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265418,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265418, "pid": 5, "tid": 7, "ts": 1716454225676287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616420, "dur": 6,
    "args": {
      "External id": 265418,
      "cbid": 211, "correlation": 265418
    }
  },
  {
    "ph": "s", "id": 265418, "pid": 76337, "tid": -914061504, "ts": 1716454225616420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225676293, "dur": 18,
    "args": {
      "External id": 265428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265428, "pid": 5, "tid": 7, "ts": 1716454225676293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616478, "dur": 12,
    "args": {
      "External id": 265428,
      "cbid": 211, "correlation": 265428
    }
  },
  {
    "ph": "s", "id": 265428, "pid": 76337, "tid": -914061504, "ts": 1716454225616478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225676313, "dur": 18,
    "args": {
      "External id": 265448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265448,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 265448, "pid": 5, "tid": 7, "ts": 1716454225676313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616543, "dur": 10,
    "args": {
      "External id": 265448,
      "cbid": 211, "correlation": 265448
    }
  },
  {
    "ph": "s", "id": 265448, "pid": 76337, "tid": -914061504, "ts": 1716454225616543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225676331, "dur": 4,
    "args": {
      "External id": 265460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265460,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 265460, "pid": 5, "tid": 7, "ts": 1716454225676331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616563, "dur": 7,
    "args": {
      "External id": 265460,
      "cbid": 211, "correlation": 265460
    }
  },
  {
    "ph": "s", "id": 265460, "pid": 76337, "tid": -914061504, "ts": 1716454225616563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225676337, "dur": 16,
    "args": {
      "External id": 265463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265463, "pid": 5, "tid": 7, "ts": 1716454225676337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616582, "dur": 6,
    "args": {
      "External id": 265463,
      "cbid": 211, "correlation": 265463
    }
  },
  {
    "ph": "s", "id": 265463, "pid": 76337, "tid": -914061504, "ts": 1716454225616582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225676355, "dur": 11,
    "args": {
      "External id": 265472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265472,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265472, "pid": 5, "tid": 7, "ts": 1716454225676355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616621, "dur": 10,
    "args": {
      "External id": 265472,
      "cbid": 211, "correlation": 265472
    }
  },
  {
    "ph": "s", "id": 265472, "pid": 76337, "tid": -914061504, "ts": 1716454225616621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616683, "dur": 0,
    "args": {
      "External id": 265482,
      "cbid": 317, "correlation": 265482
    }
  },
  {
    "ph": "f", "id": 265482, "pid": 76337, "tid": -914061504, "ts": 1716454225616683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616684, "dur": 0,
    "args": {
      "External id": 265483,
      "cbid": 203, "correlation": 265483
    }
  },
  {
    "ph": "f", "id": 265483, "pid": 76337, "tid": -914061504, "ts": 1716454225616684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616685, "dur": 0,
    "args": {
      "External id": 265484,
      "cbid": 205, "correlation": 265484
    }
  },
  {
    "ph": "f", "id": 265484, "pid": 76337, "tid": -914061504, "ts": 1716454225616685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225676367, "dur": 11,
    "args": {
      "External id": 265488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265488,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265488, "pid": 5, "tid": 7, "ts": 1716454225676367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616698, "dur": 12,
    "args": {
      "External id": 265488,
      "cbid": 211, "correlation": 265488
    }
  },
  {
    "ph": "s", "id": 265488, "pid": 76337, "tid": -914061504, "ts": 1716454225616698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225676379, "dur": 166,
    "args": {
      "External id": 265490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265490, "pid": 5, "tid": 7, "ts": 1716454225676379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616712, "dur": 5,
    "args": {
      "External id": 265490,
      "cbid": 211, "correlation": 265490
    }
  },
  {
    "ph": "s", "id": 265490, "pid": 76337, "tid": -914061504, "ts": 1716454225616712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225676548, "dur": 1,
    "args": {
      "External id": 265492,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265492,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 265492, "pid": 5, "tid": 7, "ts": 1716454225676548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616724, "dur": 8,
    "args": {
      "External id": 265492,
      "cbid": 51, "correlation": 265492
    }
  },
  {
    "ph": "s", "id": 265492, "pid": 76337, "tid": -914061504, "ts": 1716454225616724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225676552, "dur": 659,
    "args": {
      "External id": 265493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265493,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265493, "pid": 5, "tid": 7, "ts": 1716454225676552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616732, "dur": 6,
    "args": {
      "External id": 265493,
      "cbid": 211, "correlation": 265493
    }
  },
  {
    "ph": "s", "id": 265493, "pid": 76337, "tid": -914061504, "ts": 1716454225616732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225677212, "dur": 14,
    "args": {
      "External id": 265495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265495,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265495, "pid": 5, "tid": 7, "ts": 1716454225677212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616742, "dur": 5,
    "args": {
      "External id": 265495,
      "cbid": 211, "correlation": 265495
    }
  },
  {
    "ph": "s", "id": 265495, "pid": 76337, "tid": -914061504, "ts": 1716454225616742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225677228, "dur": 15,
    "args": {
      "External id": 265501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265501, "pid": 5, "tid": 7, "ts": 1716454225677228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616770, "dur": 10,
    "args": {
      "External id": 265501,
      "cbid": 211, "correlation": 265501
    }
  },
  {
    "ph": "s", "id": 265501, "pid": 76337, "tid": -914061504, "ts": 1716454225616770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616828, "dur": 0,
    "args": {
      "External id": 265511,
      "cbid": 317, "correlation": 265511
    }
  },
  {
    "ph": "f", "id": 265511, "pid": 76337, "tid": -914061504, "ts": 1716454225616828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616829, "dur": 0,
    "args": {
      "External id": 265512,
      "cbid": 203, "correlation": 265512
    }
  },
  {
    "ph": "f", "id": 265512, "pid": 76337, "tid": -914061504, "ts": 1716454225616829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616830, "dur": 0,
    "args": {
      "External id": 265513,
      "cbid": 205, "correlation": 265513
    }
  },
  {
    "ph": "f", "id": 265513, "pid": 76337, "tid": -914061504, "ts": 1716454225616830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225677244, "dur": 21,
    "args": {
      "External id": 265517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265517,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265517, "pid": 5, "tid": 7, "ts": 1716454225677244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616844, "dur": 11,
    "args": {
      "External id": 265517,
      "cbid": 211, "correlation": 265517
    }
  },
  {
    "ph": "s", "id": 265517, "pid": 76337, "tid": -914061504, "ts": 1716454225616844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225677267, "dur": 4,
    "args": {
      "External id": 265519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265519,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 265519, "pid": 5, "tid": 7, "ts": 1716454225677267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616860, "dur": 6,
    "args": {
      "External id": 265519,
      "cbid": 211, "correlation": 265519
    }
  },
  {
    "ph": "s", "id": 265519, "pid": 76337, "tid": -914061504, "ts": 1716454225616860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616869, "dur": 0,
    "args": {
      "External id": 265520,
      "cbid": 51, "correlation": 265520
    }
  },
  {
    "ph": "s", "id": 265520, "pid": 76337, "tid": -914061504, "ts": 1716454225616869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225677272, "dur": 177,
    "args": {
      "External id": 265521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265521,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 265521, "pid": 5, "tid": 7, "ts": 1716454225677272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616870, "dur": 5,
    "args": {
      "External id": 265521,
      "cbid": 211, "correlation": 265521
    }
  },
  {
    "ph": "s", "id": 265521, "pid": 76337, "tid": -914061504, "ts": 1716454225616870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225677450, "dur": 16,
    "args": {
      "External id": 265526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265526, "pid": 5, "tid": 7, "ts": 1716454225677450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616897, "dur": 8,
    "args": {
      "External id": 265526,
      "cbid": 211, "correlation": 265526
    }
  },
  {
    "ph": "s", "id": 265526, "pid": 76337, "tid": -914061504, "ts": 1716454225616897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225677468, "dur": 12,
    "args": {
      "External id": 265534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265534,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265534, "pid": 5, "tid": 7, "ts": 1716454225677468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616925, "dur": 8,
    "args": {
      "External id": 265534,
      "cbid": 211, "correlation": 265534
    }
  },
  {
    "ph": "s", "id": 265534, "pid": 76337, "tid": -914061504, "ts": 1716454225616925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225677481, "dur": 10,
    "args": {
      "External id": 265542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265542,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265542, "pid": 5, "tid": 7, "ts": 1716454225677481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225616954, "dur": 8,
    "args": {
      "External id": 265542,
      "cbid": 211, "correlation": 265542
    }
  },
  {
    "ph": "s", "id": 265542, "pid": 76337, "tid": -914061504, "ts": 1716454225616954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225677492, "dur": 18,
    "args": {
      "External id": 265562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265562,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 265562, "pid": 5, "tid": 7, "ts": 1716454225677492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617045, "dur": 13,
    "args": {
      "External id": 265562,
      "cbid": 211, "correlation": 265562
    }
  },
  {
    "ph": "s", "id": 265562, "pid": 76337, "tid": -914061504, "ts": 1716454225617045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225677512, "dur": 5,
    "args": {
      "External id": 265574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265574,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 265574, "pid": 5, "tid": 7, "ts": 1716454225677512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617068, "dur": 6,
    "args": {
      "External id": 265574,
      "cbid": 211, "correlation": 265574
    }
  },
  {
    "ph": "s", "id": 265574, "pid": 76337, "tid": -914061504, "ts": 1716454225617068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225677518, "dur": 17,
    "args": {
      "External id": 265577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265577, "pid": 5, "tid": 7, "ts": 1716454225677518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617086, "dur": 7,
    "args": {
      "External id": 265577,
      "cbid": 211, "correlation": 265577
    }
  },
  {
    "ph": "s", "id": 265577, "pid": 76337, "tid": -914061504, "ts": 1716454225617086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617144, "dur": 0,
    "args": {
      "External id": 265588,
      "cbid": 317, "correlation": 265588
    }
  },
  {
    "ph": "f", "id": 265588, "pid": 76337, "tid": -914061504, "ts": 1716454225617144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617145, "dur": 0,
    "args": {
      "External id": 265589,
      "cbid": 203, "correlation": 265589
    }
  },
  {
    "ph": "f", "id": 265589, "pid": 76337, "tid": -914061504, "ts": 1716454225617145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617146, "dur": 0,
    "args": {
      "External id": 265590,
      "cbid": 205, "correlation": 265590
    }
  },
  {
    "ph": "f", "id": 265590, "pid": 76337, "tid": -914061504, "ts": 1716454225617146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225677536, "dur": 11,
    "args": {
      "External id": 265594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265594,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265594, "pid": 5, "tid": 7, "ts": 1716454225677536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617159, "dur": 12,
    "args": {
      "External id": 265594,
      "cbid": 211, "correlation": 265594
    }
  },
  {
    "ph": "s", "id": 265594, "pid": 76337, "tid": -914061504, "ts": 1716454225617159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225677549, "dur": 3,
    "args": {
      "External id": 265596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265596,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 265596, "pid": 5, "tid": 7, "ts": 1716454225677549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617175, "dur": 7,
    "args": {
      "External id": 265596,
      "cbid": 211, "correlation": 265596
    }
  },
  {
    "ph": "s", "id": 265596, "pid": 76337, "tid": -914061504, "ts": 1716454225617175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617184, "dur": 0,
    "args": {
      "External id": 265597,
      "cbid": 51, "correlation": 265597
    }
  },
  {
    "ph": "s", "id": 265597, "pid": 76337, "tid": -914061504, "ts": 1716454225617184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225677554, "dur": 92,
    "args": {
      "External id": 265598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265598,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 265598, "pid": 5, "tid": 7, "ts": 1716454225677554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617185, "dur": 5,
    "args": {
      "External id": 265598,
      "cbid": 211, "correlation": 265598
    }
  },
  {
    "ph": "s", "id": 265598, "pid": 76337, "tid": -914061504, "ts": 1716454225617185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225677648, "dur": 16,
    "args": {
      "External id": 265603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265603, "pid": 5, "tid": 7, "ts": 1716454225677648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617212, "dur": 8,
    "args": {
      "External id": 265603,
      "cbid": 211, "correlation": 265603
    }
  },
  {
    "ph": "s", "id": 265603, "pid": 76337, "tid": -914061504, "ts": 1716454225617212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225677665, "dur": 84,
    "args": {
      "External id": 265612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265612,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265612, "pid": 5, "tid": 7, "ts": 1716454225677665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617294, "dur": 15,
    "args": {
      "External id": 265612,
      "cbid": 211, "correlation": 265612
    }
  },
  {
    "ph": "s", "id": 265612, "pid": 76337, "tid": -914061504, "ts": 1716454225617294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225677751, "dur": 30,
    "args": {
      "External id": 265634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265634,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265634, "pid": 5, "tid": 7, "ts": 1716454225677751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617353, "dur": 10,
    "args": {
      "External id": 265634,
      "cbid": 211, "correlation": 265634
    }
  },
  {
    "ph": "s", "id": 265634, "pid": 76337, "tid": -914061504, "ts": 1716454225617353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617445, "dur": 2,
    "args": {
      "External id": 265645,
      "cbid": 251, "correlation": 265645
    }
  },
  {
    "ph": "f", "id": 265645, "pid": 76337, "tid": -914061504, "ts": 1716454225617445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225677782, "dur": 165,
    "args": {
      "External id": 265646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265646,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265646, "pid": 5, "tid": 7, "ts": 1716454225677782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617451, "dur": 13,
    "args": {
      "External id": 265646,
      "cbid": 211, "correlation": 265646
    }
  },
  {
    "ph": "s", "id": 265646, "pid": 76337, "tid": -914061504, "ts": 1716454225617451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617522, "dur": 1,
    "args": {
      "External id": 265657,
      "cbid": 251, "correlation": 265657
    }
  },
  {
    "ph": "f", "id": 265657, "pid": 76337, "tid": -914061504, "ts": 1716454225617522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225677948, "dur": 162,
    "args": {
      "External id": 265658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265658,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265658, "pid": 5, "tid": 7, "ts": 1716454225677948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617526, "dur": 12,
    "args": {
      "External id": 265658,
      "cbid": 211, "correlation": 265658
    }
  },
  {
    "ph": "s", "id": 265658, "pid": 76337, "tid": -914061504, "ts": 1716454225617526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617593, "dur": 1,
    "args": {
      "External id": 265669,
      "cbid": 251, "correlation": 265669
    }
  },
  {
    "ph": "f", "id": 265669, "pid": 76337, "tid": -914061504, "ts": 1716454225617593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225678111, "dur": 160,
    "args": {
      "External id": 265670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265670,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265670, "pid": 5, "tid": 7, "ts": 1716454225678111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617597, "dur": 11,
    "args": {
      "External id": 265670,
      "cbid": 211, "correlation": 265670
    }
  },
  {
    "ph": "s", "id": 265670, "pid": 76337, "tid": -914061504, "ts": 1716454225617597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225678273, "dur": 341,
    "args": {
      "External id": 265695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265695,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265695, "pid": 5, "tid": 7, "ts": 1716454225678273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617684, "dur": 13,
    "args": {
      "External id": 265695,
      "cbid": 211, "correlation": 265695
    }
  },
  {
    "ph": "s", "id": 265695, "pid": 76337, "tid": -914061504, "ts": 1716454225617684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617785, "dur": 1,
    "args": {
      "External id": 265713,
      "cbid": 251, "correlation": 265713
    }
  },
  {
    "ph": "f", "id": 265713, "pid": 76337, "tid": -914061504, "ts": 1716454225617785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225678615, "dur": 170,
    "args": {
      "External id": 265715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265715,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265715, "pid": 5, "tid": 7, "ts": 1716454225678615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617791, "dur": 13,
    "args": {
      "External id": 265715,
      "cbid": 211, "correlation": 265715
    }
  },
  {
    "ph": "s", "id": 265715, "pid": 76337, "tid": -914061504, "ts": 1716454225617791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225678787, "dur": 19,
    "args": {
      "External id": 265723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265723,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265723, "pid": 5, "tid": 7, "ts": 1716454225678787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617860, "dur": 12,
    "args": {
      "External id": 265723,
      "cbid": 211, "correlation": 265723
    }
  },
  {
    "ph": "s", "id": 265723, "pid": 76337, "tid": -914061504, "ts": 1716454225617860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225678807, "dur": 27,
    "args": {
      "External id": 265731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265731,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265731, "pid": 5, "tid": 7, "ts": 1716454225678807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617899, "dur": 9,
    "args": {
      "External id": 265731,
      "cbid": 211, "correlation": 265731
    }
  },
  {
    "ph": "s", "id": 265731, "pid": 76337, "tid": -914061504, "ts": 1716454225617899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225678836, "dur": 19,
    "args": {
      "External id": 265742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265742, "pid": 5, "tid": 7, "ts": 1716454225678836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225617972, "dur": 21,
    "args": {
      "External id": 265742,
      "cbid": 211, "correlation": 265742
    }
  },
  {
    "ph": "s", "id": 265742, "pid": 76337, "tid": -914061504, "ts": 1716454225617972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225678856, "dur": 16,
    "args": {
      "External id": 265764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265764,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265764, "pid": 5, "tid": 7, "ts": 1716454225678856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618011, "dur": 8,
    "args": {
      "External id": 265764,
      "cbid": 211, "correlation": 265764
    }
  },
  {
    "ph": "s", "id": 265764, "pid": 76337, "tid": -914061504, "ts": 1716454225618011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618098, "dur": 1,
    "args": {
      "External id": 265775,
      "cbid": 251, "correlation": 265775
    }
  },
  {
    "ph": "f", "id": 265775, "pid": 76337, "tid": -914061504, "ts": 1716454225618098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225678873, "dur": 90,
    "args": {
      "External id": 265776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265776,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 265776, "pid": 5, "tid": 7, "ts": 1716454225678873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618103, "dur": 13,
    "args": {
      "External id": 265776,
      "cbid": 211, "correlation": 265776
    }
  },
  {
    "ph": "s", "id": 265776, "pid": 76337, "tid": -914061504, "ts": 1716454225618103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618172, "dur": 1,
    "args": {
      "External id": 265787,
      "cbid": 251, "correlation": 265787
    }
  },
  {
    "ph": "f", "id": 265787, "pid": 76337, "tid": -914061504, "ts": 1716454225618172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618175, "dur": 0,
    "args": {
      "External id": 265788,
      "cbid": 251, "correlation": 265788
    }
  },
  {
    "ph": "f", "id": 265788, "pid": 76337, "tid": -914061504, "ts": 1716454225618175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225678965, "dur": 13,
    "args": {
      "External id": 265789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265789,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265789, "pid": 5, "tid": 7, "ts": 1716454225678965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618177, "dur": 12,
    "args": {
      "External id": 265789,
      "cbid": 211, "correlation": 265789
    }
  },
  {
    "ph": "s", "id": 265789, "pid": 76337, "tid": -914061504, "ts": 1716454225618177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225678979, "dur": 6,
    "args": {
      "External id": 265791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265791,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265791, "pid": 5, "tid": 7, "ts": 1716454225678979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618190, "dur": 6,
    "args": {
      "External id": 265791,
      "cbid": 211, "correlation": 265791
    }
  },
  {
    "ph": "s", "id": 265791, "pid": 76337, "tid": -914061504, "ts": 1716454225618190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618247, "dur": 1,
    "args": {
      "External id": 265802,
      "cbid": 251, "correlation": 265802
    }
  },
  {
    "ph": "f", "id": 265802, "pid": 76337, "tid": -914061504, "ts": 1716454225618247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618250, "dur": 0,
    "args": {
      "External id": 265803,
      "cbid": 251, "correlation": 265803
    }
  },
  {
    "ph": "f", "id": 265803, "pid": 76337, "tid": -914061504, "ts": 1716454225618250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225678986, "dur": 8,
    "args": {
      "External id": 265804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265804,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265804, "pid": 5, "tid": 7, "ts": 1716454225678986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618252, "dur": 11,
    "args": {
      "External id": 265804,
      "cbid": 211, "correlation": 265804
    }
  },
  {
    "ph": "s", "id": 265804, "pid": 76337, "tid": -914061504, "ts": 1716454225618252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225678995, "dur": 3,
    "args": {
      "External id": 265806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265806,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265806, "pid": 5, "tid": 7, "ts": 1716454225678995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618265, "dur": 5,
    "args": {
      "External id": 265806,
      "cbid": 211, "correlation": 265806
    }
  },
  {
    "ph": "s", "id": 265806, "pid": 76337, "tid": -914061504, "ts": 1716454225618265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225679000, "dur": 57,
    "args": {
      "External id": 265831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265831,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265831, "pid": 5, "tid": 7, "ts": 1716454225679000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618342, "dur": 13,
    "args": {
      "External id": 265831,
      "cbid": 211, "correlation": 265831
    }
  },
  {
    "ph": "s", "id": 265831, "pid": 76337, "tid": -914061504, "ts": 1716454225618342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618441, "dur": 2,
    "args": {
      "External id": 265849,
      "cbid": 251, "correlation": 265849
    }
  },
  {
    "ph": "f", "id": 265849, "pid": 76337, "tid": -914061504, "ts": 1716454225618441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225679058, "dur": 92,
    "args": {
      "External id": 265851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265851,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 265851, "pid": 5, "tid": 7, "ts": 1716454225679058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618447, "dur": 13,
    "args": {
      "External id": 265851,
      "cbid": 211, "correlation": 265851
    }
  },
  {
    "ph": "s", "id": 265851, "pid": 76337, "tid": -914061504, "ts": 1716454225618447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225679152, "dur": 10,
    "args": {
      "External id": 265859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265859,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265859, "pid": 5, "tid": 7, "ts": 1716454225679152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618516, "dur": 12,
    "args": {
      "External id": 265859,
      "cbid": 211, "correlation": 265859
    }
  },
  {
    "ph": "s", "id": 265859, "pid": 76337, "tid": -914061504, "ts": 1716454225618516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225679162, "dur": 20,
    "args": {
      "External id": 265867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265867, "pid": 5, "tid": 7, "ts": 1716454225679162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618557, "dur": 9,
    "args": {
      "External id": 265867,
      "cbid": 211, "correlation": 265867
    }
  },
  {
    "ph": "s", "id": 265867, "pid": 76337, "tid": -914061504, "ts": 1716454225618557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225679184, "dur": 17,
    "args": {
      "External id": 265889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265889,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265889, "pid": 5, "tid": 7, "ts": 1716454225679184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618608, "dur": 10,
    "args": {
      "External id": 265889,
      "cbid": 211, "correlation": 265889
    }
  },
  {
    "ph": "s", "id": 265889, "pid": 76337, "tid": -914061504, "ts": 1716454225618608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618694, "dur": 1,
    "args": {
      "External id": 265905,
      "cbid": 251, "correlation": 265905
    }
  },
  {
    "ph": "f", "id": 265905, "pid": 76337, "tid": -914061504, "ts": 1716454225618694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618700, "dur": 0,
    "args": {
      "External id": 265907,
      "cbid": 251, "correlation": 265907
    }
  },
  {
    "ph": "f", "id": 265907, "pid": 76337, "tid": -914061504, "ts": 1716454225618700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225679203, "dur": 498,
    "args": {
      "External id": 265908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265908,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 265908, "pid": 5, "tid": 7, "ts": 1716454225679203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618702, "dur": 13,
    "args": {
      "External id": 265908,
      "cbid": 211, "correlation": 265908
    }
  },
  {
    "ph": "s", "id": 265908, "pid": 76337, "tid": -914061504, "ts": 1716454225618702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225679703, "dur": 67,
    "args": {
      "External id": 265916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265916,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265916, "pid": 5, "tid": 7, "ts": 1716454225679703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618768, "dur": 12,
    "args": {
      "External id": 265916,
      "cbid": 211, "correlation": 265916
    }
  },
  {
    "ph": "s", "id": 265916, "pid": 76337, "tid": -914061504, "ts": 1716454225618768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225679770, "dur": 68,
    "args": {
      "External id": 265924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265924, "pid": 5, "tid": 7, "ts": 1716454225679770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618801, "dur": 8,
    "args": {
      "External id": 265924,
      "cbid": 211, "correlation": 265924
    }
  },
  {
    "ph": "s", "id": 265924, "pid": 76337, "tid": -914061504, "ts": 1716454225618801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618880, "dur": 1,
    "args": {
      "External id": 265940,
      "cbid": 251, "correlation": 265940
    }
  },
  {
    "ph": "f", "id": 265940, "pid": 76337, "tid": -914061504, "ts": 1716454225618880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225679841, "dur": 1,
    "args": {
      "External id": 265942,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 265942,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 265942, "pid": 5, "tid": 7, "ts": 1716454225679841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618885, "dur": 10,
    "args": {
      "External id": 265942,
      "cbid": 51, "correlation": 265942
    }
  },
  {
    "ph": "s", "id": 265942, "pid": 76337, "tid": -914061504, "ts": 1716454225618885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225679844, "dur": 274,
    "args": {
      "External id": 265943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265943,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 265943, "pid": 5, "tid": 7, "ts": 1716454225679844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618897, "dur": 11,
    "args": {
      "External id": 265943,
      "cbid": 211, "correlation": 265943
    }
  },
  {
    "ph": "s", "id": 265943, "pid": 76337, "tid": -914061504, "ts": 1716454225618897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225680120, "dur": 14,
    "args": {
      "External id": 265951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265951,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265951, "pid": 5, "tid": 7, "ts": 1716454225680120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225618939, "dur": 11,
    "args": {
      "External id": 265951,
      "cbid": 211, "correlation": 265951
    }
  },
  {
    "ph": "s", "id": 265951, "pid": 76337, "tid": -914061504, "ts": 1716454225618939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225680136, "dur": 39,
    "args": {
      "External id": 265962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265962, "pid": 5, "tid": 7, "ts": 1716454225680136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619015, "dur": 13,
    "args": {
      "External id": 265962,
      "cbid": 211, "correlation": 265962
    }
  },
  {
    "ph": "s", "id": 265962, "pid": 76337, "tid": -914061504, "ts": 1716454225619015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619081, "dur": 0,
    "args": {
      "External id": 265974,
      "cbid": 317, "correlation": 265974
    }
  },
  {
    "ph": "f", "id": 265974, "pid": 76337, "tid": -914061504, "ts": 1716454225619081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619082, "dur": 0,
    "args": {
      "External id": 265975,
      "cbid": 203, "correlation": 265975
    }
  },
  {
    "ph": "f", "id": 265975, "pid": 76337, "tid": -914061504, "ts": 1716454225619082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619082, "dur": 0,
    "args": {
      "External id": 265976,
      "cbid": 205, "correlation": 265976
    }
  },
  {
    "ph": "f", "id": 265976, "pid": 76337, "tid": -914061504, "ts": 1716454225619082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225680177, "dur": 14,
    "args": {
      "External id": 265980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265980, "pid": 5, "tid": 7, "ts": 1716454225680177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619098, "dur": 12,
    "args": {
      "External id": 265980,
      "cbid": 211, "correlation": 265980
    }
  },
  {
    "ph": "s", "id": 265980, "pid": 76337, "tid": -914061504, "ts": 1716454225619098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225680192, "dur": 4,
    "args": {
      "External id": 265982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265982,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 265982, "pid": 5, "tid": 7, "ts": 1716454225680192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619115, "dur": 5,
    "args": {
      "External id": 265982,
      "cbid": 211, "correlation": 265982
    }
  },
  {
    "ph": "s", "id": 265982, "pid": 76337, "tid": -914061504, "ts": 1716454225619115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619123, "dur": 0,
    "args": {
      "External id": 265983,
      "cbid": 51, "correlation": 265983
    }
  },
  {
    "ph": "s", "id": 265983, "pid": 76337, "tid": -914061504, "ts": 1716454225619123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225680197, "dur": 98,
    "args": {
      "External id": 265984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265984,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 265984, "pid": 5, "tid": 7, "ts": 1716454225680197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619124, "dur": 5,
    "args": {
      "External id": 265984,
      "cbid": 211, "correlation": 265984
    }
  },
  {
    "ph": "s", "id": 265984, "pid": 76337, "tid": -914061504, "ts": 1716454225619124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225680296, "dur": 17,
    "args": {
      "External id": 265989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265989, "pid": 5, "tid": 7, "ts": 1716454225680296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619151, "dur": 9,
    "args": {
      "External id": 265989,
      "cbid": 211, "correlation": 265989
    }
  },
  {
    "ph": "s", "id": 265989, "pid": 76337, "tid": -914061504, "ts": 1716454225619151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225680315, "dur": 12,
    "args": {
      "External id": 265997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 265997,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 265997, "pid": 5, "tid": 7, "ts": 1716454225680315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619183, "dur": 8,
    "args": {
      "External id": 265997,
      "cbid": 211, "correlation": 265997
    }
  },
  {
    "ph": "s", "id": 265997, "pid": 76337, "tid": -914061504, "ts": 1716454225619183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225680329, "dur": 31,
    "args": {
      "External id": 266006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266006,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266006, "pid": 5, "tid": 7, "ts": 1716454225680329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619222, "dur": 10,
    "args": {
      "External id": 266006,
      "cbid": 211, "correlation": 266006
    }
  },
  {
    "ph": "s", "id": 266006, "pid": 76337, "tid": -914061504, "ts": 1716454225619222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225680361, "dur": 31,
    "args": {
      "External id": 266026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266026,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 266026, "pid": 5, "tid": 7, "ts": 1716454225680361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619292, "dur": 12,
    "args": {
      "External id": 266026,
      "cbid": 211, "correlation": 266026
    }
  },
  {
    "ph": "s", "id": 266026, "pid": 76337, "tid": -914061504, "ts": 1716454225619292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225680394, "dur": 5,
    "args": {
      "External id": 266038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266038,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266038, "pid": 5, "tid": 7, "ts": 1716454225680394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619314, "dur": 6,
    "args": {
      "External id": 266038,
      "cbid": 211, "correlation": 266038
    }
  },
  {
    "ph": "s", "id": 266038, "pid": 76337, "tid": -914061504, "ts": 1716454225619314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225680400, "dur": 32,
    "args": {
      "External id": 266041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266041,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266041, "pid": 5, "tid": 7, "ts": 1716454225680400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619331, "dur": 7,
    "args": {
      "External id": 266041,
      "cbid": 211, "correlation": 266041
    }
  },
  {
    "ph": "s", "id": 266041, "pid": 76337, "tid": -914061504, "ts": 1716454225619331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225680433, "dur": 21,
    "args": {
      "External id": 266050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266050,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266050, "pid": 5, "tid": 7, "ts": 1716454225680433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619371, "dur": 9,
    "args": {
      "External id": 266050,
      "cbid": 211, "correlation": 266050
    }
  },
  {
    "ph": "s", "id": 266050, "pid": 76337, "tid": -914061504, "ts": 1716454225619371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619423, "dur": 0,
    "args": {
      "External id": 266060,
      "cbid": 317, "correlation": 266060
    }
  },
  {
    "ph": "f", "id": 266060, "pid": 76337, "tid": -914061504, "ts": 1716454225619423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619423, "dur": 0,
    "args": {
      "External id": 266061,
      "cbid": 203, "correlation": 266061
    }
  },
  {
    "ph": "f", "id": 266061, "pid": 76337, "tid": -914061504, "ts": 1716454225619423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619424, "dur": 0,
    "args": {
      "External id": 266062,
      "cbid": 205, "correlation": 266062
    }
  },
  {
    "ph": "f", "id": 266062, "pid": 76337, "tid": -914061504, "ts": 1716454225619424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225680455, "dur": 23,
    "args": {
      "External id": 266066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266066,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266066, "pid": 5, "tid": 7, "ts": 1716454225680455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619437, "dur": 12,
    "args": {
      "External id": 266066,
      "cbid": 211, "correlation": 266066
    }
  },
  {
    "ph": "s", "id": 266066, "pid": 76337, "tid": -914061504, "ts": 1716454225619437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225680479, "dur": 326,
    "args": {
      "External id": 266068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266068,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266068, "pid": 5, "tid": 7, "ts": 1716454225680479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619452, "dur": 5,
    "args": {
      "External id": 266068,
      "cbid": 211, "correlation": 266068
    }
  },
  {
    "ph": "s", "id": 266068, "pid": 76337, "tid": -914061504, "ts": 1716454225619452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225680807, "dur": 1,
    "args": {
      "External id": 266070,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 266070,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 266070, "pid": 5, "tid": 7, "ts": 1716454225680807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619463, "dur": 6,
    "args": {
      "External id": 266070,
      "cbid": 51, "correlation": 266070
    }
  },
  {
    "ph": "s", "id": 266070, "pid": 76337, "tid": -914061504, "ts": 1716454225619463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225680811, "dur": 1277,
    "args": {
      "External id": 266071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266071,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266071, "pid": 5, "tid": 7, "ts": 1716454225680811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619470, "dur": 6,
    "args": {
      "External id": 266071,
      "cbid": 211, "correlation": 266071
    }
  },
  {
    "ph": "s", "id": 266071, "pid": 76337, "tid": -914061504, "ts": 1716454225619470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225682089, "dur": 13,
    "args": {
      "External id": 266073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266073,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266073, "pid": 5, "tid": 7, "ts": 1716454225682089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619480, "dur": 5,
    "args": {
      "External id": 266073,
      "cbid": 211, "correlation": 266073
    }
  },
  {
    "ph": "s", "id": 266073, "pid": 76337, "tid": -914061504, "ts": 1716454225619480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225682104, "dur": 15,
    "args": {
      "External id": 266079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266079,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266079, "pid": 5, "tid": 7, "ts": 1716454225682104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619509, "dur": 8,
    "args": {
      "External id": 266079,
      "cbid": 211, "correlation": 266079
    }
  },
  {
    "ph": "s", "id": 266079, "pid": 76337, "tid": -914061504, "ts": 1716454225619509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225682120, "dur": 3,
    "args": {
      "External id": 266087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266087,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 266087, "pid": 5, "tid": 7, "ts": 1716454225682120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619553, "dur": 9,
    "args": {
      "External id": 266087,
      "cbid": 211, "correlation": 266087
    }
  },
  {
    "ph": "s", "id": 266087, "pid": 76337, "tid": -914061504, "ts": 1716454225619553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619617, "dur": 1,
    "args": {
      "External id": 266103,
      "cbid": 251, "correlation": 266103
    }
  },
  {
    "ph": "f", "id": 266103, "pid": 76337, "tid": -914061504, "ts": 1716454225619617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619622, "dur": 0,
    "args": {
      "External id": 266105,
      "cbid": 251, "correlation": 266105
    }
  },
  {
    "ph": "f", "id": 266105, "pid": 76337, "tid": -914061504, "ts": 1716454225619622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225682125, "dur": 13,
    "args": {
      "External id": 266106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266106,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266106, "pid": 5, "tid": 7, "ts": 1716454225682125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619624, "dur": 11,
    "args": {
      "External id": 266106,
      "cbid": 211, "correlation": 266106
    }
  },
  {
    "ph": "s", "id": 266106, "pid": 76337, "tid": -914061504, "ts": 1716454225619624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225682140, "dur": 6,
    "args": {
      "External id": 266108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266108,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266108, "pid": 5, "tid": 7, "ts": 1716454225682140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619637, "dur": 5,
    "args": {
      "External id": 266108,
      "cbid": 211, "correlation": 266108
    }
  },
  {
    "ph": "s", "id": 266108, "pid": 76337, "tid": -914061504, "ts": 1716454225619637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225682147, "dur": 18,
    "args": {
      "External id": 266118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266118, "pid": 5, "tid": 7, "ts": 1716454225682147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619694, "dur": 13,
    "args": {
      "External id": 266118,
      "cbid": 211, "correlation": 266118
    }
  },
  {
    "ph": "s", "id": 266118, "pid": 76337, "tid": -914061504, "ts": 1716454225619694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225682166, "dur": 19,
    "args": {
      "External id": 266138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266138,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 266138, "pid": 5, "tid": 7, "ts": 1716454225682166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619761, "dur": 10,
    "args": {
      "External id": 266138,
      "cbid": 211, "correlation": 266138
    }
  },
  {
    "ph": "s", "id": 266138, "pid": 76337, "tid": -914061504, "ts": 1716454225619761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225682186, "dur": 4,
    "args": {
      "External id": 266150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266150,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 266150, "pid": 5, "tid": 7, "ts": 1716454225682186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619782, "dur": 6,
    "args": {
      "External id": 266150,
      "cbid": 211, "correlation": 266150
    }
  },
  {
    "ph": "s", "id": 266150, "pid": 76337, "tid": -914061504, "ts": 1716454225619782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225682191, "dur": 17,
    "args": {
      "External id": 266153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266153,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266153, "pid": 5, "tid": 7, "ts": 1716454225682191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619800, "dur": 7,
    "args": {
      "External id": 266153,
      "cbid": 211, "correlation": 266153
    }
  },
  {
    "ph": "s", "id": 266153, "pid": 76337, "tid": -914061504, "ts": 1716454225619800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225682209, "dur": 12,
    "args": {
      "External id": 266162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266162,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266162, "pid": 5, "tid": 7, "ts": 1716454225682209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619840, "dur": 9,
    "args": {
      "External id": 266162,
      "cbid": 211, "correlation": 266162
    }
  },
  {
    "ph": "s", "id": 266162, "pid": 76337, "tid": -914061504, "ts": 1716454225619840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619901, "dur": 0,
    "args": {
      "External id": 266172,
      "cbid": 317, "correlation": 266172
    }
  },
  {
    "ph": "f", "id": 266172, "pid": 76337, "tid": -914061504, "ts": 1716454225619901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619902, "dur": 0,
    "args": {
      "External id": 266173,
      "cbid": 203, "correlation": 266173
    }
  },
  {
    "ph": "f", "id": 266173, "pid": 76337, "tid": -914061504, "ts": 1716454225619902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619903, "dur": 0,
    "args": {
      "External id": 266174,
      "cbid": 205, "correlation": 266174
    }
  },
  {
    "ph": "f", "id": 266174, "pid": 76337, "tid": -914061504, "ts": 1716454225619903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225682223, "dur": 11,
    "args": {
      "External id": 266178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266178, "pid": 5, "tid": 7, "ts": 1716454225682223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619917, "dur": 12,
    "args": {
      "External id": 266178,
      "cbid": 211, "correlation": 266178
    }
  },
  {
    "ph": "s", "id": 266178, "pid": 76337, "tid": -914061504, "ts": 1716454225619917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225682235, "dur": 166,
    "args": {
      "External id": 266180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266180,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266180, "pid": 5, "tid": 7, "ts": 1716454225682235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619931, "dur": 5,
    "args": {
      "External id": 266180,
      "cbid": 211, "correlation": 266180
    }
  },
  {
    "ph": "s", "id": 266180, "pid": 76337, "tid": -914061504, "ts": 1716454225619931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225682403, "dur": 1,
    "args": {
      "External id": 266182,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 266182,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 266182, "pid": 5, "tid": 7, "ts": 1716454225682403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619942, "dur": 6,
    "args": {
      "External id": 266182,
      "cbid": 51, "correlation": 266182
    }
  },
  {
    "ph": "s", "id": 266182, "pid": 76337, "tid": -914061504, "ts": 1716454225619942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225682407, "dur": 659,
    "args": {
      "External id": 266183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266183,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266183, "pid": 5, "tid": 7, "ts": 1716454225682407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619949, "dur": 6,
    "args": {
      "External id": 266183,
      "cbid": 211, "correlation": 266183
    }
  },
  {
    "ph": "s", "id": 266183, "pid": 76337, "tid": -914061504, "ts": 1716454225619949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225683067, "dur": 12,
    "args": {
      "External id": 266185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266185,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266185, "pid": 5, "tid": 7, "ts": 1716454225683067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619960, "dur": 5,
    "args": {
      "External id": 266185,
      "cbid": 211, "correlation": 266185
    }
  },
  {
    "ph": "s", "id": 266185, "pid": 76337, "tid": -914061504, "ts": 1716454225619960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225683080, "dur": 15,
    "args": {
      "External id": 266191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266191, "pid": 5, "tid": 7, "ts": 1716454225683080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225619996, "dur": 10,
    "args": {
      "External id": 266191,
      "cbid": 211, "correlation": 266191
    }
  },
  {
    "ph": "s", "id": 266191, "pid": 76337, "tid": -914061504, "ts": 1716454225619996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620056, "dur": 0,
    "args": {
      "External id": 266201,
      "cbid": 317, "correlation": 266201
    }
  },
  {
    "ph": "f", "id": 266201, "pid": 76337, "tid": -914061504, "ts": 1716454225620056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620057, "dur": 0,
    "args": {
      "External id": 266202,
      "cbid": 203, "correlation": 266202
    }
  },
  {
    "ph": "f", "id": 266202, "pid": 76337, "tid": -914061504, "ts": 1716454225620057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620058, "dur": 0,
    "args": {
      "External id": 266203,
      "cbid": 205, "correlation": 266203
    }
  },
  {
    "ph": "f", "id": 266203, "pid": 76337, "tid": -914061504, "ts": 1716454225620058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225683096, "dur": 21,
    "args": {
      "External id": 266207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266207,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266207, "pid": 5, "tid": 7, "ts": 1716454225683096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620071, "dur": 11,
    "args": {
      "External id": 266207,
      "cbid": 211, "correlation": 266207
    }
  },
  {
    "ph": "s", "id": 266207, "pid": 76337, "tid": -914061504, "ts": 1716454225620071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225683119, "dur": 4,
    "args": {
      "External id": 266209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266209, "pid": 5, "tid": 7, "ts": 1716454225683119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620086, "dur": 6,
    "args": {
      "External id": 266209,
      "cbid": 211, "correlation": 266209
    }
  },
  {
    "ph": "s", "id": 266209, "pid": 76337, "tid": -914061504, "ts": 1716454225620086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620094, "dur": 0,
    "args": {
      "External id": 266210,
      "cbid": 51, "correlation": 266210
    }
  },
  {
    "ph": "s", "id": 266210, "pid": 76337, "tid": -914061504, "ts": 1716454225620094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225683124, "dur": 175,
    "args": {
      "External id": 266211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266211,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 266211, "pid": 5, "tid": 7, "ts": 1716454225683124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620095, "dur": 5,
    "args": {
      "External id": 266211,
      "cbid": 211, "correlation": 266211
    }
  },
  {
    "ph": "s", "id": 266211, "pid": 76337, "tid": -914061504, "ts": 1716454225620095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225683301, "dur": 16,
    "args": {
      "External id": 266216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266216,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266216, "pid": 5, "tid": 7, "ts": 1716454225683301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620121, "dur": 8,
    "args": {
      "External id": 266216,
      "cbid": 211, "correlation": 266216
    }
  },
  {
    "ph": "s", "id": 266216, "pid": 76337, "tid": -914061504, "ts": 1716454225620121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225683318, "dur": 13,
    "args": {
      "External id": 266224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266224,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266224, "pid": 5, "tid": 7, "ts": 1716454225683318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620149, "dur": 8,
    "args": {
      "External id": 266224,
      "cbid": 211, "correlation": 266224
    }
  },
  {
    "ph": "s", "id": 266224, "pid": 76337, "tid": -914061504, "ts": 1716454225620149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225683332, "dur": 11,
    "args": {
      "External id": 266232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266232,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266232, "pid": 5, "tid": 7, "ts": 1716454225683332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620179, "dur": 8,
    "args": {
      "External id": 266232,
      "cbid": 211, "correlation": 266232
    }
  },
  {
    "ph": "s", "id": 266232, "pid": 76337, "tid": -914061504, "ts": 1716454225620179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225683344, "dur": 19,
    "args": {
      "External id": 266252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266252,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 266252, "pid": 5, "tid": 7, "ts": 1716454225683344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620260, "dur": 13,
    "args": {
      "External id": 266252,
      "cbid": 211, "correlation": 266252
    }
  },
  {
    "ph": "s", "id": 266252, "pid": 76337, "tid": -914061504, "ts": 1716454225620260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225683365, "dur": 4,
    "args": {
      "External id": 266264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266264,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 266264, "pid": 5, "tid": 7, "ts": 1716454225683365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620282, "dur": 6,
    "args": {
      "External id": 266264,
      "cbid": 211, "correlation": 266264
    }
  },
  {
    "ph": "s", "id": 266264, "pid": 76337, "tid": -914061504, "ts": 1716454225620282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225683371, "dur": 18,
    "args": {
      "External id": 266267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266267,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266267, "pid": 5, "tid": 7, "ts": 1716454225683371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620299, "dur": 6,
    "args": {
      "External id": 266267,
      "cbid": 211, "correlation": 266267
    }
  },
  {
    "ph": "s", "id": 266267, "pid": 76337, "tid": -914061504, "ts": 1716454225620299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620356, "dur": 0,
    "args": {
      "External id": 266278,
      "cbid": 317, "correlation": 266278
    }
  },
  {
    "ph": "f", "id": 266278, "pid": 76337, "tid": -914061504, "ts": 1716454225620356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620357, "dur": 0,
    "args": {
      "External id": 266279,
      "cbid": 203, "correlation": 266279
    }
  },
  {
    "ph": "f", "id": 266279, "pid": 76337, "tid": -914061504, "ts": 1716454225620357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620358, "dur": 0,
    "args": {
      "External id": 266280,
      "cbid": 205, "correlation": 266280
    }
  },
  {
    "ph": "f", "id": 266280, "pid": 76337, "tid": -914061504, "ts": 1716454225620358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225683390, "dur": 12,
    "args": {
      "External id": 266284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266284,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266284, "pid": 5, "tid": 7, "ts": 1716454225683390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620371, "dur": 11,
    "args": {
      "External id": 266284,
      "cbid": 211, "correlation": 266284
    }
  },
  {
    "ph": "s", "id": 266284, "pid": 76337, "tid": -914061504, "ts": 1716454225620371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225683402, "dur": 3,
    "args": {
      "External id": 266286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266286, "pid": 5, "tid": 7, "ts": 1716454225683402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620387, "dur": 5,
    "args": {
      "External id": 266286,
      "cbid": 211, "correlation": 266286
    }
  },
  {
    "ph": "s", "id": 266286, "pid": 76337, "tid": -914061504, "ts": 1716454225620387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620395, "dur": 0,
    "args": {
      "External id": 266287,
      "cbid": 51, "correlation": 266287
    }
  },
  {
    "ph": "s", "id": 266287, "pid": 76337, "tid": -914061504, "ts": 1716454225620395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225683407, "dur": 93,
    "args": {
      "External id": 266288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266288,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 266288, "pid": 5, "tid": 7, "ts": 1716454225683407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620396, "dur": 5,
    "args": {
      "External id": 266288,
      "cbid": 211, "correlation": 266288
    }
  },
  {
    "ph": "s", "id": 266288, "pid": 76337, "tid": -914061504, "ts": 1716454225620396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225683502, "dur": 16,
    "args": {
      "External id": 266293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266293,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266293, "pid": 5, "tid": 7, "ts": 1716454225683502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620423, "dur": 8,
    "args": {
      "External id": 266293,
      "cbid": 211, "correlation": 266293
    }
  },
  {
    "ph": "s", "id": 266293, "pid": 76337, "tid": -914061504, "ts": 1716454225620423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225683519, "dur": 85,
    "args": {
      "External id": 266302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266302,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266302, "pid": 5, "tid": 7, "ts": 1716454225683519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620502, "dur": 15,
    "args": {
      "External id": 266302,
      "cbid": 211, "correlation": 266302
    }
  },
  {
    "ph": "s", "id": 266302, "pid": 76337, "tid": -914061504, "ts": 1716454225620502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225683605, "dur": 31,
    "args": {
      "External id": 266324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266324,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266324, "pid": 5, "tid": 7, "ts": 1716454225683605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620559, "dur": 11,
    "args": {
      "External id": 266324,
      "cbid": 211, "correlation": 266324
    }
  },
  {
    "ph": "s", "id": 266324, "pid": 76337, "tid": -914061504, "ts": 1716454225620559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620647, "dur": 1,
    "args": {
      "External id": 266335,
      "cbid": 251, "correlation": 266335
    }
  },
  {
    "ph": "f", "id": 266335, "pid": 76337, "tid": -914061504, "ts": 1716454225620647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225683637, "dur": 165,
    "args": {
      "External id": 266336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266336,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266336, "pid": 5, "tid": 7, "ts": 1716454225683637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620652, "dur": 14,
    "args": {
      "External id": 266336,
      "cbid": 211, "correlation": 266336
    }
  },
  {
    "ph": "s", "id": 266336, "pid": 76337, "tid": -914061504, "ts": 1716454225620652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620723, "dur": 1,
    "args": {
      "External id": 266347,
      "cbid": 251, "correlation": 266347
    }
  },
  {
    "ph": "f", "id": 266347, "pid": 76337, "tid": -914061504, "ts": 1716454225620723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225683804, "dur": 162,
    "args": {
      "External id": 266348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266348,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266348, "pid": 5, "tid": 7, "ts": 1716454225683804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620727, "dur": 11,
    "args": {
      "External id": 266348,
      "cbid": 211, "correlation": 266348
    }
  },
  {
    "ph": "s", "id": 266348, "pid": 76337, "tid": -914061504, "ts": 1716454225620727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620791, "dur": 1,
    "args": {
      "External id": 266359,
      "cbid": 251, "correlation": 266359
    }
  },
  {
    "ph": "f", "id": 266359, "pid": 76337, "tid": -914061504, "ts": 1716454225620791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225683967, "dur": 162,
    "args": {
      "External id": 266360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266360,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266360, "pid": 5, "tid": 7, "ts": 1716454225683967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620795, "dur": 11,
    "args": {
      "External id": 266360,
      "cbid": 211, "correlation": 266360
    }
  },
  {
    "ph": "s", "id": 266360, "pid": 76337, "tid": -914061504, "ts": 1716454225620795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225684131, "dur": 339,
    "args": {
      "External id": 266385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266385,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266385, "pid": 5, "tid": 7, "ts": 1716454225684131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620879, "dur": 13,
    "args": {
      "External id": 266385,
      "cbid": 211, "correlation": 266385
    }
  },
  {
    "ph": "s", "id": 266385, "pid": 76337, "tid": -914061504, "ts": 1716454225620879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620985, "dur": 1,
    "args": {
      "External id": 266403,
      "cbid": 251, "correlation": 266403
    }
  },
  {
    "ph": "f", "id": 266403, "pid": 76337, "tid": -914061504, "ts": 1716454225620985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225684471, "dur": 169,
    "args": {
      "External id": 266405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266405,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266405, "pid": 5, "tid": 7, "ts": 1716454225684471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225620991, "dur": 14,
    "args": {
      "External id": 266405,
      "cbid": 211, "correlation": 266405
    }
  },
  {
    "ph": "s", "id": 266405, "pid": 76337, "tid": -914061504, "ts": 1716454225620991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225684641, "dur": 20,
    "args": {
      "External id": 266413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266413,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266413, "pid": 5, "tid": 7, "ts": 1716454225684641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621061, "dur": 13,
    "args": {
      "External id": 266413,
      "cbid": 211, "correlation": 266413
    }
  },
  {
    "ph": "s", "id": 266413, "pid": 76337, "tid": -914061504, "ts": 1716454225621061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225684662, "dur": 28,
    "args": {
      "External id": 266421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266421,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266421, "pid": 5, "tid": 7, "ts": 1716454225684662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621100, "dur": 8,
    "args": {
      "External id": 266421,
      "cbid": 211, "correlation": 266421
    }
  },
  {
    "ph": "s", "id": 266421, "pid": 76337, "tid": -914061504, "ts": 1716454225621100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225684691, "dur": 18,
    "args": {
      "External id": 266432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266432, "pid": 5, "tid": 7, "ts": 1716454225684691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621171, "dur": 12,
    "args": {
      "External id": 266432,
      "cbid": 211, "correlation": 266432
    }
  },
  {
    "ph": "s", "id": 266432, "pid": 76337, "tid": -914061504, "ts": 1716454225621171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225684711, "dur": 16,
    "args": {
      "External id": 266454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266454,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266454, "pid": 5, "tid": 7, "ts": 1716454225684711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621202, "dur": 7,
    "args": {
      "External id": 266454,
      "cbid": 211, "correlation": 266454
    }
  },
  {
    "ph": "s", "id": 266454, "pid": 76337, "tid": -914061504, "ts": 1716454225621202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621286, "dur": 1,
    "args": {
      "External id": 266465,
      "cbid": 251, "correlation": 266465
    }
  },
  {
    "ph": "f", "id": 266465, "pid": 76337, "tid": -914061504, "ts": 1716454225621286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225684728, "dur": 90,
    "args": {
      "External id": 266466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266466,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266466, "pid": 5, "tid": 7, "ts": 1716454225684728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621291, "dur": 13,
    "args": {
      "External id": 266466,
      "cbid": 211, "correlation": 266466
    }
  },
  {
    "ph": "s", "id": 266466, "pid": 76337, "tid": -914061504, "ts": 1716454225621291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621360, "dur": 1,
    "args": {
      "External id": 266477,
      "cbid": 251, "correlation": 266477
    }
  },
  {
    "ph": "f", "id": 266477, "pid": 76337, "tid": -914061504, "ts": 1716454225621360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621364, "dur": 0,
    "args": {
      "External id": 266478,
      "cbid": 251, "correlation": 266478
    }
  },
  {
    "ph": "f", "id": 266478, "pid": 76337, "tid": -914061504, "ts": 1716454225621364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225684820, "dur": 12,
    "args": {
      "External id": 266479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266479,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266479, "pid": 5, "tid": 7, "ts": 1716454225684820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621366, "dur": 12,
    "args": {
      "External id": 266479,
      "cbid": 211, "correlation": 266479
    }
  },
  {
    "ph": "s", "id": 266479, "pid": 76337, "tid": -914061504, "ts": 1716454225621366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225684833, "dur": 6,
    "args": {
      "External id": 266481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266481,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266481, "pid": 5, "tid": 7, "ts": 1716454225684833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621380, "dur": 6,
    "args": {
      "External id": 266481,
      "cbid": 211, "correlation": 266481
    }
  },
  {
    "ph": "s", "id": 266481, "pid": 76337, "tid": -914061504, "ts": 1716454225621380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621437, "dur": 1,
    "args": {
      "External id": 266492,
      "cbid": 251, "correlation": 266492
    }
  },
  {
    "ph": "f", "id": 266492, "pid": 76337, "tid": -914061504, "ts": 1716454225621437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621441, "dur": 0,
    "args": {
      "External id": 266493,
      "cbid": 251, "correlation": 266493
    }
  },
  {
    "ph": "f", "id": 266493, "pid": 76337, "tid": -914061504, "ts": 1716454225621441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225684840, "dur": 8,
    "args": {
      "External id": 266494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266494,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266494, "pid": 5, "tid": 7, "ts": 1716454225684840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621442, "dur": 12,
    "args": {
      "External id": 266494,
      "cbid": 211, "correlation": 266494
    }
  },
  {
    "ph": "s", "id": 266494, "pid": 76337, "tid": -914061504, "ts": 1716454225621442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225684850, "dur": 3,
    "args": {
      "External id": 266496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266496,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266496, "pid": 5, "tid": 7, "ts": 1716454225684850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621455, "dur": 5,
    "args": {
      "External id": 266496,
      "cbid": 211, "correlation": 266496
    }
  },
  {
    "ph": "s", "id": 266496, "pid": 76337, "tid": -914061504, "ts": 1716454225621455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225684854, "dur": 56,
    "args": {
      "External id": 266521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266521,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266521, "pid": 5, "tid": 7, "ts": 1716454225684854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621531, "dur": 12,
    "args": {
      "External id": 266521,
      "cbid": 211, "correlation": 266521
    }
  },
  {
    "ph": "s", "id": 266521, "pid": 76337, "tid": -914061504, "ts": 1716454225621531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621630, "dur": 1,
    "args": {
      "External id": 266539,
      "cbid": 251, "correlation": 266539
    }
  },
  {
    "ph": "f", "id": 266539, "pid": 76337, "tid": -914061504, "ts": 1716454225621630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225684912, "dur": 92,
    "args": {
      "External id": 266541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266541,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266541, "pid": 5, "tid": 7, "ts": 1716454225684912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621636, "dur": 13,
    "args": {
      "External id": 266541,
      "cbid": 211, "correlation": 266541
    }
  },
  {
    "ph": "s", "id": 266541, "pid": 76337, "tid": -914061504, "ts": 1716454225621636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225685005, "dur": 10,
    "args": {
      "External id": 266549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266549,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266549, "pid": 5, "tid": 7, "ts": 1716454225685005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621704, "dur": 12,
    "args": {
      "External id": 266549,
      "cbid": 211, "correlation": 266549
    }
  },
  {
    "ph": "s", "id": 266549, "pid": 76337, "tid": -914061504, "ts": 1716454225621704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225685016, "dur": 20,
    "args": {
      "External id": 266557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266557, "pid": 5, "tid": 7, "ts": 1716454225685016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621746, "dur": 9,
    "args": {
      "External id": 266557,
      "cbid": 211, "correlation": 266557
    }
  },
  {
    "ph": "s", "id": 266557, "pid": 76337, "tid": -914061504, "ts": 1716454225621746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225685038, "dur": 18,
    "args": {
      "External id": 266579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266579,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266579, "pid": 5, "tid": 7, "ts": 1716454225685038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621797, "dur": 10,
    "args": {
      "External id": 266579,
      "cbid": 211, "correlation": 266579
    }
  },
  {
    "ph": "s", "id": 266579, "pid": 76337, "tid": -914061504, "ts": 1716454225621797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621883, "dur": 1,
    "args": {
      "External id": 266595,
      "cbid": 251, "correlation": 266595
    }
  },
  {
    "ph": "f", "id": 266595, "pid": 76337, "tid": -914061504, "ts": 1716454225621883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621888, "dur": 0,
    "args": {
      "External id": 266597,
      "cbid": 251, "correlation": 266597
    }
  },
  {
    "ph": "f", "id": 266597, "pid": 76337, "tid": -914061504, "ts": 1716454225621888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225685057, "dur": 498,
    "args": {
      "External id": 266598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266598,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266598, "pid": 5, "tid": 7, "ts": 1716454225685057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621890, "dur": 12,
    "args": {
      "External id": 266598,
      "cbid": 211, "correlation": 266598
    }
  },
  {
    "ph": "s", "id": 266598, "pid": 76337, "tid": -914061504, "ts": 1716454225621890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225685557, "dur": 67,
    "args": {
      "External id": 266606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266606,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266606, "pid": 5, "tid": 7, "ts": 1716454225685557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621954, "dur": 12,
    "args": {
      "External id": 266606,
      "cbid": 211, "correlation": 266606
    }
  },
  {
    "ph": "s", "id": 266606, "pid": 76337, "tid": -914061504, "ts": 1716454225621954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225685625, "dur": 65,
    "args": {
      "External id": 266614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266614, "pid": 5, "tid": 7, "ts": 1716454225685625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225621993, "dur": 9,
    "args": {
      "External id": 266614,
      "cbid": 211, "correlation": 266614
    }
  },
  {
    "ph": "s", "id": 266614, "pid": 76337, "tid": -914061504, "ts": 1716454225621993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622077, "dur": 1,
    "args": {
      "External id": 266630,
      "cbid": 251, "correlation": 266630
    }
  },
  {
    "ph": "f", "id": 266630, "pid": 76337, "tid": -914061504, "ts": 1716454225622077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225685693, "dur": 1,
    "args": {
      "External id": 266632,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 266632,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 266632, "pid": 5, "tid": 7, "ts": 1716454225685693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622082, "dur": 9,
    "args": {
      "External id": 266632,
      "cbid": 51, "correlation": 266632
    }
  },
  {
    "ph": "s", "id": 266632, "pid": 76337, "tid": -914061504, "ts": 1716454225622082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225685697, "dur": 275,
    "args": {
      "External id": 266633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266633,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266633, "pid": 5, "tid": 7, "ts": 1716454225685697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622092, "dur": 11,
    "args": {
      "External id": 266633,
      "cbid": 211, "correlation": 266633
    }
  },
  {
    "ph": "s", "id": 266633, "pid": 76337, "tid": -914061504, "ts": 1716454225622092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225685973, "dur": 14,
    "args": {
      "External id": 266641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266641,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266641, "pid": 5, "tid": 7, "ts": 1716454225685973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622135, "dur": 10,
    "args": {
      "External id": 266641,
      "cbid": 211, "correlation": 266641
    }
  },
  {
    "ph": "s", "id": 266641, "pid": 76337, "tid": -914061504, "ts": 1716454225622135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225685989, "dur": 39,
    "args": {
      "External id": 266652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266652,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266652, "pid": 5, "tid": 7, "ts": 1716454225685989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622203, "dur": 13,
    "args": {
      "External id": 266652,
      "cbid": 211, "correlation": 266652
    }
  },
  {
    "ph": "s", "id": 266652, "pid": 76337, "tid": -914061504, "ts": 1716454225622203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622267, "dur": 0,
    "args": {
      "External id": 266664,
      "cbid": 317, "correlation": 266664
    }
  },
  {
    "ph": "f", "id": 266664, "pid": 76337, "tid": -914061504, "ts": 1716454225622267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622268, "dur": 0,
    "args": {
      "External id": 266665,
      "cbid": 203, "correlation": 266665
    }
  },
  {
    "ph": "f", "id": 266665, "pid": 76337, "tid": -914061504, "ts": 1716454225622268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622269, "dur": 0,
    "args": {
      "External id": 266666,
      "cbid": 205, "correlation": 266666
    }
  },
  {
    "ph": "f", "id": 266666, "pid": 76337, "tid": -914061504, "ts": 1716454225622269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225686029, "dur": 12,
    "args": {
      "External id": 266670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266670, "pid": 5, "tid": 7, "ts": 1716454225686029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622284, "dur": 12,
    "args": {
      "External id": 266670,
      "cbid": 211, "correlation": 266670
    }
  },
  {
    "ph": "s", "id": 266670, "pid": 76337, "tid": -914061504, "ts": 1716454225622284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225686042, "dur": 4,
    "args": {
      "External id": 266672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266672, "pid": 5, "tid": 7, "ts": 1716454225686042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622301, "dur": 6,
    "args": {
      "External id": 266672,
      "cbid": 211, "correlation": 266672
    }
  },
  {
    "ph": "s", "id": 266672, "pid": 76337, "tid": -914061504, "ts": 1716454225622301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622309, "dur": 0,
    "args": {
      "External id": 266673,
      "cbid": 51, "correlation": 266673
    }
  },
  {
    "ph": "s", "id": 266673, "pid": 76337, "tid": -914061504, "ts": 1716454225622309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225686048, "dur": 98,
    "args": {
      "External id": 266674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266674,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 266674, "pid": 5, "tid": 7, "ts": 1716454225686048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622310, "dur": 5,
    "args": {
      "External id": 266674,
      "cbid": 211, "correlation": 266674
    }
  },
  {
    "ph": "s", "id": 266674, "pid": 76337, "tid": -914061504, "ts": 1716454225622310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225686147, "dur": 17,
    "args": {
      "External id": 266679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266679,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266679, "pid": 5, "tid": 7, "ts": 1716454225686147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622338, "dur": 9,
    "args": {
      "External id": 266679,
      "cbid": 211, "correlation": 266679
    }
  },
  {
    "ph": "s", "id": 266679, "pid": 76337, "tid": -914061504, "ts": 1716454225622338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225686165, "dur": 12,
    "args": {
      "External id": 266687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266687,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266687, "pid": 5, "tid": 7, "ts": 1716454225686165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622370, "dur": 9,
    "args": {
      "External id": 266687,
      "cbid": 211, "correlation": 266687
    }
  },
  {
    "ph": "s", "id": 266687, "pid": 76337, "tid": -914061504, "ts": 1716454225622370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225686178, "dur": 25,
    "args": {
      "External id": 266696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266696,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266696, "pid": 5, "tid": 7, "ts": 1716454225686178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622409, "dur": 10,
    "args": {
      "External id": 266696,
      "cbid": 211, "correlation": 266696
    }
  },
  {
    "ph": "s", "id": 266696, "pid": 76337, "tid": -914061504, "ts": 1716454225622409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225686204, "dur": 25,
    "args": {
      "External id": 266716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266716,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 266716, "pid": 5, "tid": 7, "ts": 1716454225686204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622480, "dur": 12,
    "args": {
      "External id": 266716,
      "cbid": 211, "correlation": 266716
    }
  },
  {
    "ph": "s", "id": 266716, "pid": 76337, "tid": -914061504, "ts": 1716454225622480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225686230, "dur": 5,
    "args": {
      "External id": 266728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266728,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 266728, "pid": 5, "tid": 7, "ts": 1716454225686230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622501, "dur": 7,
    "args": {
      "External id": 266728,
      "cbid": 211, "correlation": 266728
    }
  },
  {
    "ph": "s", "id": 266728, "pid": 76337, "tid": -914061504, "ts": 1716454225622501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225686236, "dur": 25,
    "args": {
      "External id": 266731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266731, "pid": 5, "tid": 7, "ts": 1716454225686236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622520, "dur": 7,
    "args": {
      "External id": 266731,
      "cbid": 211, "correlation": 266731
    }
  },
  {
    "ph": "s", "id": 266731, "pid": 76337, "tid": -914061504, "ts": 1716454225622520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225686263, "dur": 17,
    "args": {
      "External id": 266740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266740,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266740, "pid": 5, "tid": 7, "ts": 1716454225686263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622558, "dur": 9,
    "args": {
      "External id": 266740,
      "cbid": 211, "correlation": 266740
    }
  },
  {
    "ph": "s", "id": 266740, "pid": 76337, "tid": -914061504, "ts": 1716454225622558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622610, "dur": 0,
    "args": {
      "External id": 266750,
      "cbid": 317, "correlation": 266750
    }
  },
  {
    "ph": "f", "id": 266750, "pid": 76337, "tid": -914061504, "ts": 1716454225622610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622611, "dur": 0,
    "args": {
      "External id": 266751,
      "cbid": 203, "correlation": 266751
    }
  },
  {
    "ph": "f", "id": 266751, "pid": 76337, "tid": -914061504, "ts": 1716454225622611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622612, "dur": 0,
    "args": {
      "External id": 266752,
      "cbid": 205, "correlation": 266752
    }
  },
  {
    "ph": "f", "id": 266752, "pid": 76337, "tid": -914061504, "ts": 1716454225622612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225686281, "dur": 17,
    "args": {
      "External id": 266756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266756,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266756, "pid": 5, "tid": 7, "ts": 1716454225686281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622626, "dur": 13,
    "args": {
      "External id": 266756,
      "cbid": 211, "correlation": 266756
    }
  },
  {
    "ph": "s", "id": 266756, "pid": 76337, "tid": -914061504, "ts": 1716454225622626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225686299, "dur": 246,
    "args": {
      "External id": 266758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266758,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266758, "pid": 5, "tid": 7, "ts": 1716454225686299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622641, "dur": 5,
    "args": {
      "External id": 266758,
      "cbid": 211, "correlation": 266758
    }
  },
  {
    "ph": "s", "id": 266758, "pid": 76337, "tid": -914061504, "ts": 1716454225622641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225686548, "dur": 1,
    "args": {
      "External id": 266760,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 266760,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 266760, "pid": 5, "tid": 7, "ts": 1716454225686548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622653, "dur": 8,
    "args": {
      "External id": 266760,
      "cbid": 51, "correlation": 266760
    }
  },
  {
    "ph": "s", "id": 266760, "pid": 76337, "tid": -914061504, "ts": 1716454225622653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225686551, "dur": 822,
    "args": {
      "External id": 266761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266761,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266761, "pid": 5, "tid": 7, "ts": 1716454225686551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622661, "dur": 6,
    "args": {
      "External id": 266761,
      "cbid": 211, "correlation": 266761
    }
  },
  {
    "ph": "s", "id": 266761, "pid": 76337, "tid": -914061504, "ts": 1716454225622661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225687375, "dur": 13,
    "args": {
      "External id": 266763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266763,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266763, "pid": 5, "tid": 7, "ts": 1716454225687375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622671, "dur": 5,
    "args": {
      "External id": 266763,
      "cbid": 211, "correlation": 266763
    }
  },
  {
    "ph": "s", "id": 266763, "pid": 76337, "tid": -914061504, "ts": 1716454225622671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225687390, "dur": 15,
    "args": {
      "External id": 266769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266769,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266769, "pid": 5, "tid": 7, "ts": 1716454225687390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622700, "dur": 8,
    "args": {
      "External id": 266769,
      "cbid": 211, "correlation": 266769
    }
  },
  {
    "ph": "s", "id": 266769, "pid": 76337, "tid": -914061504, "ts": 1716454225622700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225687406, "dur": 3,
    "args": {
      "External id": 266777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266777,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 266777, "pid": 5, "tid": 7, "ts": 1716454225687406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622744, "dur": 9,
    "args": {
      "External id": 266777,
      "cbid": 211, "correlation": 266777
    }
  },
  {
    "ph": "s", "id": 266777, "pid": 76337, "tid": -914061504, "ts": 1716454225622744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622808, "dur": 1,
    "args": {
      "External id": 266793,
      "cbid": 251, "correlation": 266793
    }
  },
  {
    "ph": "f", "id": 266793, "pid": 76337, "tid": -914061504, "ts": 1716454225622808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622813, "dur": 0,
    "args": {
      "External id": 266795,
      "cbid": 251, "correlation": 266795
    }
  },
  {
    "ph": "f", "id": 266795, "pid": 76337, "tid": -914061504, "ts": 1716454225622813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225687410, "dur": 13,
    "args": {
      "External id": 266796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266796,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266796, "pid": 5, "tid": 7, "ts": 1716454225687410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622815, "dur": 11,
    "args": {
      "External id": 266796,
      "cbid": 211, "correlation": 266796
    }
  },
  {
    "ph": "s", "id": 266796, "pid": 76337, "tid": -914061504, "ts": 1716454225622815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225687425, "dur": 5,
    "args": {
      "External id": 266798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266798,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266798, "pid": 5, "tid": 7, "ts": 1716454225687425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622828, "dur": 6,
    "args": {
      "External id": 266798,
      "cbid": 211, "correlation": 266798
    }
  },
  {
    "ph": "s", "id": 266798, "pid": 76337, "tid": -914061504, "ts": 1716454225622828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225687431, "dur": 17,
    "args": {
      "External id": 266808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266808, "pid": 5, "tid": 7, "ts": 1716454225687431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622887, "dur": 13,
    "args": {
      "External id": 266808,
      "cbid": 211, "correlation": 266808
    }
  },
  {
    "ph": "s", "id": 266808, "pid": 76337, "tid": -914061504, "ts": 1716454225622887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225687450, "dur": 18,
    "args": {
      "External id": 266828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266828,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 266828, "pid": 5, "tid": 7, "ts": 1716454225687450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622952, "dur": 10,
    "args": {
      "External id": 266828,
      "cbid": 211, "correlation": 266828
    }
  },
  {
    "ph": "s", "id": 266828, "pid": 76337, "tid": -914061504, "ts": 1716454225622952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225687469, "dur": 4,
    "args": {
      "External id": 266840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266840,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 266840, "pid": 5, "tid": 7, "ts": 1716454225687469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622972, "dur": 15,
    "args": {
      "External id": 266840,
      "cbid": 211, "correlation": 266840
    }
  },
  {
    "ph": "s", "id": 266840, "pid": 76337, "tid": -914061504, "ts": 1716454225622972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225687475, "dur": 17,
    "args": {
      "External id": 266843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266843,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266843, "pid": 5, "tid": 7, "ts": 1716454225687475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225622999, "dur": 7,
    "args": {
      "External id": 266843,
      "cbid": 211, "correlation": 266843
    }
  },
  {
    "ph": "s", "id": 266843, "pid": 76337, "tid": -914061504, "ts": 1716454225622999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225687493, "dur": 11,
    "args": {
      "External id": 266852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266852,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266852, "pid": 5, "tid": 7, "ts": 1716454225687493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623042, "dur": 10,
    "args": {
      "External id": 266852,
      "cbid": 211, "correlation": 266852
    }
  },
  {
    "ph": "s", "id": 266852, "pid": 76337, "tid": -914061504, "ts": 1716454225623042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623104, "dur": 0,
    "args": {
      "External id": 266862,
      "cbid": 317, "correlation": 266862
    }
  },
  {
    "ph": "f", "id": 266862, "pid": 76337, "tid": -914061504, "ts": 1716454225623104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623105, "dur": 0,
    "args": {
      "External id": 266863,
      "cbid": 203, "correlation": 266863
    }
  },
  {
    "ph": "f", "id": 266863, "pid": 76337, "tid": -914061504, "ts": 1716454225623105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623105, "dur": 0,
    "args": {
      "External id": 266864,
      "cbid": 205, "correlation": 266864
    }
  },
  {
    "ph": "f", "id": 266864, "pid": 76337, "tid": -914061504, "ts": 1716454225623105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225687506, "dur": 12,
    "args": {
      "External id": 266868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266868, "pid": 5, "tid": 7, "ts": 1716454225687506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623118, "dur": 12,
    "args": {
      "External id": 266868,
      "cbid": 211, "correlation": 266868
    }
  },
  {
    "ph": "s", "id": 266868, "pid": 76337, "tid": -914061504, "ts": 1716454225623118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225687518, "dur": 165,
    "args": {
      "External id": 266870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266870, "pid": 5, "tid": 7, "ts": 1716454225687518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623132, "dur": 6,
    "args": {
      "External id": 266870,
      "cbid": 211, "correlation": 266870
    }
  },
  {
    "ph": "s", "id": 266870, "pid": 76337, "tid": -914061504, "ts": 1716454225623132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225687686, "dur": 1,
    "args": {
      "External id": 266872,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 266872,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 266872, "pid": 5, "tid": 7, "ts": 1716454225687686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623145, "dur": 6,
    "args": {
      "External id": 266872,
      "cbid": 51, "correlation": 266872
    }
  },
  {
    "ph": "s", "id": 266872, "pid": 76337, "tid": -914061504, "ts": 1716454225623145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225687690, "dur": 660,
    "args": {
      "External id": 266873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266873,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 266873, "pid": 5, "tid": 7, "ts": 1716454225687690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623152, "dur": 6,
    "args": {
      "External id": 266873,
      "cbid": 211, "correlation": 266873
    }
  },
  {
    "ph": "s", "id": 266873, "pid": 76337, "tid": -914061504, "ts": 1716454225623152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225688351, "dur": 13,
    "args": {
      "External id": 266875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266875, "pid": 5, "tid": 7, "ts": 1716454225688351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623162, "dur": 5,
    "args": {
      "External id": 266875,
      "cbid": 211, "correlation": 266875
    }
  },
  {
    "ph": "s", "id": 266875, "pid": 76337, "tid": -914061504, "ts": 1716454225623162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225688365, "dur": 15,
    "args": {
      "External id": 266881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266881, "pid": 5, "tid": 7, "ts": 1716454225688365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623191, "dur": 8,
    "args": {
      "External id": 266881,
      "cbid": 211, "correlation": 266881
    }
  },
  {
    "ph": "s", "id": 266881, "pid": 76337, "tid": -914061504, "ts": 1716454225623191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623249, "dur": 0,
    "args": {
      "External id": 266891,
      "cbid": 317, "correlation": 266891
    }
  },
  {
    "ph": "f", "id": 266891, "pid": 76337, "tid": -914061504, "ts": 1716454225623249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623250, "dur": 0,
    "args": {
      "External id": 266892,
      "cbid": 203, "correlation": 266892
    }
  },
  {
    "ph": "f", "id": 266892, "pid": 76337, "tid": -914061504, "ts": 1716454225623250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623250, "dur": 0,
    "args": {
      "External id": 266893,
      "cbid": 205, "correlation": 266893
    }
  },
  {
    "ph": "f", "id": 266893, "pid": 76337, "tid": -914061504, "ts": 1716454225623250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225688382, "dur": 17,
    "args": {
      "External id": 266897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266897,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266897, "pid": 5, "tid": 7, "ts": 1716454225688382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623264, "dur": 11,
    "args": {
      "External id": 266897,
      "cbid": 211, "correlation": 266897
    }
  },
  {
    "ph": "s", "id": 266897, "pid": 76337, "tid": -914061504, "ts": 1716454225623264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225688400, "dur": 4,
    "args": {
      "External id": 266899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266899, "pid": 5, "tid": 7, "ts": 1716454225688400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623280, "dur": 6,
    "args": {
      "External id": 266899,
      "cbid": 211, "correlation": 266899
    }
  },
  {
    "ph": "s", "id": 266899, "pid": 76337, "tid": -914061504, "ts": 1716454225623280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623290, "dur": 0,
    "args": {
      "External id": 266900,
      "cbid": 51, "correlation": 266900
    }
  },
  {
    "ph": "s", "id": 266900, "pid": 76337, "tid": -914061504, "ts": 1716454225623290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225688405, "dur": 135,
    "args": {
      "External id": 266901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266901,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 266901, "pid": 5, "tid": 7, "ts": 1716454225688405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623291, "dur": 5,
    "args": {
      "External id": 266901,
      "cbid": 211, "correlation": 266901
    }
  },
  {
    "ph": "s", "id": 266901, "pid": 76337, "tid": -914061504, "ts": 1716454225623291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225688541, "dur": 15,
    "args": {
      "External id": 266906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266906,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266906, "pid": 5, "tid": 7, "ts": 1716454225688541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623316, "dur": 8,
    "args": {
      "External id": 266906,
      "cbid": 211, "correlation": 266906
    }
  },
  {
    "ph": "s", "id": 266906, "pid": 76337, "tid": -914061504, "ts": 1716454225623316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225688558, "dur": 13,
    "args": {
      "External id": 266914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266914,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266914, "pid": 5, "tid": 7, "ts": 1716454225688558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623345, "dur": 8,
    "args": {
      "External id": 266914,
      "cbid": 211, "correlation": 266914
    }
  },
  {
    "ph": "s", "id": 266914, "pid": 76337, "tid": -914061504, "ts": 1716454225623345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225688572, "dur": 10,
    "args": {
      "External id": 266922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266922,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266922, "pid": 5, "tid": 7, "ts": 1716454225688572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623373, "dur": 8,
    "args": {
      "External id": 266922,
      "cbid": 211, "correlation": 266922
    }
  },
  {
    "ph": "s", "id": 266922, "pid": 76337, "tid": -914061504, "ts": 1716454225623373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225688584, "dur": 19,
    "args": {
      "External id": 266942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266942,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 266942, "pid": 5, "tid": 7, "ts": 1716454225688584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623456, "dur": 12,
    "args": {
      "External id": 266942,
      "cbid": 211, "correlation": 266942
    }
  },
  {
    "ph": "s", "id": 266942, "pid": 76337, "tid": -914061504, "ts": 1716454225623456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225688604, "dur": 5,
    "args": {
      "External id": 266954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266954,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 266954, "pid": 5, "tid": 7, "ts": 1716454225688604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623478, "dur": 6,
    "args": {
      "External id": 266954,
      "cbid": 211, "correlation": 266954
    }
  },
  {
    "ph": "s", "id": 266954, "pid": 76337, "tid": -914061504, "ts": 1716454225623478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225688610, "dur": 18,
    "args": {
      "External id": 266957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266957, "pid": 5, "tid": 7, "ts": 1716454225688610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623496, "dur": 7,
    "args": {
      "External id": 266957,
      "cbid": 211, "correlation": 266957
    }
  },
  {
    "ph": "s", "id": 266957, "pid": 76337, "tid": -914061504, "ts": 1716454225623496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623553, "dur": 0,
    "args": {
      "External id": 266968,
      "cbid": 317, "correlation": 266968
    }
  },
  {
    "ph": "f", "id": 266968, "pid": 76337, "tid": -914061504, "ts": 1716454225623553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623554, "dur": 0,
    "args": {
      "External id": 266969,
      "cbid": 203, "correlation": 266969
    }
  },
  {
    "ph": "f", "id": 266969, "pid": 76337, "tid": -914061504, "ts": 1716454225623554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623555, "dur": 0,
    "args": {
      "External id": 266970,
      "cbid": 205, "correlation": 266970
    }
  },
  {
    "ph": "f", "id": 266970, "pid": 76337, "tid": -914061504, "ts": 1716454225623555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225688629, "dur": 12,
    "args": {
      "External id": 266974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266974,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266974, "pid": 5, "tid": 7, "ts": 1716454225688629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623568, "dur": 11,
    "args": {
      "External id": 266974,
      "cbid": 211, "correlation": 266974
    }
  },
  {
    "ph": "s", "id": 266974, "pid": 76337, "tid": -914061504, "ts": 1716454225623568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225688642, "dur": 3,
    "args": {
      "External id": 266976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266976,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 266976, "pid": 5, "tid": 7, "ts": 1716454225688642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623583, "dur": 6,
    "args": {
      "External id": 266976,
      "cbid": 211, "correlation": 266976
    }
  },
  {
    "ph": "s", "id": 266976, "pid": 76337, "tid": -914061504, "ts": 1716454225623583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623593, "dur": 0,
    "args": {
      "External id": 266977,
      "cbid": 51, "correlation": 266977
    }
  },
  {
    "ph": "s", "id": 266977, "pid": 76337, "tid": -914061504, "ts": 1716454225623593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225688647, "dur": 92,
    "args": {
      "External id": 266978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266978,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 266978, "pid": 5, "tid": 7, "ts": 1716454225688647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623593, "dur": 5,
    "args": {
      "External id": 266978,
      "cbid": 211, "correlation": 266978
    }
  },
  {
    "ph": "s", "id": 266978, "pid": 76337, "tid": -914061504, "ts": 1716454225623593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225688740, "dur": 16,
    "args": {
      "External id": 266983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266983, "pid": 5, "tid": 7, "ts": 1716454225688740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623620, "dur": 8,
    "args": {
      "External id": 266983,
      "cbid": 211, "correlation": 266983
    }
  },
  {
    "ph": "s", "id": 266983, "pid": 76337, "tid": -914061504, "ts": 1716454225623620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225688757, "dur": 84,
    "args": {
      "External id": 266992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 266992,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 266992, "pid": 5, "tid": 7, "ts": 1716454225688757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623700, "dur": 14,
    "args": {
      "External id": 266992,
      "cbid": 211, "correlation": 266992
    }
  },
  {
    "ph": "s", "id": 266992, "pid": 76337, "tid": -914061504, "ts": 1716454225623700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225688843, "dur": 31,
    "args": {
      "External id": 267014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267014,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267014, "pid": 5, "tid": 7, "ts": 1716454225688843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623757, "dur": 10,
    "args": {
      "External id": 267014,
      "cbid": 211, "correlation": 267014
    }
  },
  {
    "ph": "s", "id": 267014, "pid": 76337, "tid": -914061504, "ts": 1716454225623757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623846, "dur": 1,
    "args": {
      "External id": 267025,
      "cbid": 251, "correlation": 267025
    }
  },
  {
    "ph": "f", "id": 267025, "pid": 76337, "tid": -914061504, "ts": 1716454225623846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225688875, "dur": 167,
    "args": {
      "External id": 267026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267026,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267026, "pid": 5, "tid": 7, "ts": 1716454225688875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623852, "dur": 13,
    "args": {
      "External id": 267026,
      "cbid": 211, "correlation": 267026
    }
  },
  {
    "ph": "s", "id": 267026, "pid": 76337, "tid": -914061504, "ts": 1716454225623852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623921, "dur": 1,
    "args": {
      "External id": 267037,
      "cbid": 251, "correlation": 267037
    }
  },
  {
    "ph": "f", "id": 267037, "pid": 76337, "tid": -914061504, "ts": 1716454225623921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225689044, "dur": 161,
    "args": {
      "External id": 267038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267038,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267038, "pid": 5, "tid": 7, "ts": 1716454225689044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623925, "dur": 11,
    "args": {
      "External id": 267038,
      "cbid": 211, "correlation": 267038
    }
  },
  {
    "ph": "s", "id": 267038, "pid": 76337, "tid": -914061504, "ts": 1716454225623925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225623999, "dur": 1,
    "args": {
      "External id": 267049,
      "cbid": 251, "correlation": 267049
    }
  },
  {
    "ph": "f", "id": 267049, "pid": 76337, "tid": -914061504, "ts": 1716454225623999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225689206, "dur": 162,
    "args": {
      "External id": 267050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267050,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267050, "pid": 5, "tid": 7, "ts": 1716454225689206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624003, "dur": 12,
    "args": {
      "External id": 267050,
      "cbid": 211, "correlation": 267050
    }
  },
  {
    "ph": "s", "id": 267050, "pid": 76337, "tid": -914061504, "ts": 1716454225624003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225689369, "dur": 342,
    "args": {
      "External id": 267075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267075,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267075, "pid": 5, "tid": 7, "ts": 1716454225689369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624090, "dur": 13,
    "args": {
      "External id": 267075,
      "cbid": 211, "correlation": 267075
    }
  },
  {
    "ph": "s", "id": 267075, "pid": 76337, "tid": -914061504, "ts": 1716454225624090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624188, "dur": 1,
    "args": {
      "External id": 267093,
      "cbid": 251, "correlation": 267093
    }
  },
  {
    "ph": "f", "id": 267093, "pid": 76337, "tid": -914061504, "ts": 1716454225624188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225689713, "dur": 166,
    "args": {
      "External id": 267095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267095,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267095, "pid": 5, "tid": 7, "ts": 1716454225689713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624194, "dur": 13,
    "args": {
      "External id": 267095,
      "cbid": 211, "correlation": 267095
    }
  },
  {
    "ph": "s", "id": 267095, "pid": 76337, "tid": -914061504, "ts": 1716454225624194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225689881, "dur": 19,
    "args": {
      "External id": 267103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267103,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267103, "pid": 5, "tid": 7, "ts": 1716454225689881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624263, "dur": 13,
    "args": {
      "External id": 267103,
      "cbid": 211, "correlation": 267103
    }
  },
  {
    "ph": "s", "id": 267103, "pid": 76337, "tid": -914061504, "ts": 1716454225624263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225689902, "dur": 27,
    "args": {
      "External id": 267111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267111,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267111, "pid": 5, "tid": 7, "ts": 1716454225689902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624303, "dur": 9,
    "args": {
      "External id": 267111,
      "cbid": 211, "correlation": 267111
    }
  },
  {
    "ph": "s", "id": 267111, "pid": 76337, "tid": -914061504, "ts": 1716454225624303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225689930, "dur": 19,
    "args": {
      "External id": 267122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267122,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267122, "pid": 5, "tid": 7, "ts": 1716454225689930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624373, "dur": 12,
    "args": {
      "External id": 267122,
      "cbid": 211, "correlation": 267122
    }
  },
  {
    "ph": "s", "id": 267122, "pid": 76337, "tid": -914061504, "ts": 1716454225624373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225689951, "dur": 16,
    "args": {
      "External id": 267144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267144,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267144, "pid": 5, "tid": 7, "ts": 1716454225689951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624405, "dur": 7,
    "args": {
      "External id": 267144,
      "cbid": 211, "correlation": 267144
    }
  },
  {
    "ph": "s", "id": 267144, "pid": 76337, "tid": -914061504, "ts": 1716454225624405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624489, "dur": 1,
    "args": {
      "External id": 267155,
      "cbid": 251, "correlation": 267155
    }
  },
  {
    "ph": "f", "id": 267155, "pid": 76337, "tid": -914061504, "ts": 1716454225624489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225689968, "dur": 90,
    "args": {
      "External id": 267156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267156,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 267156, "pid": 5, "tid": 7, "ts": 1716454225689968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624495, "dur": 13,
    "args": {
      "External id": 267156,
      "cbid": 211, "correlation": 267156
    }
  },
  {
    "ph": "s", "id": 267156, "pid": 76337, "tid": -914061504, "ts": 1716454225624495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624563, "dur": 1,
    "args": {
      "External id": 267167,
      "cbid": 251, "correlation": 267167
    }
  },
  {
    "ph": "f", "id": 267167, "pid": 76337, "tid": -914061504, "ts": 1716454225624563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624567, "dur": 0,
    "args": {
      "External id": 267168,
      "cbid": 251, "correlation": 267168
    }
  },
  {
    "ph": "f", "id": 267168, "pid": 76337, "tid": -914061504, "ts": 1716454225624567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225690060, "dur": 12,
    "args": {
      "External id": 267169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267169,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267169, "pid": 5, "tid": 7, "ts": 1716454225690060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624569, "dur": 12,
    "args": {
      "External id": 267169,
      "cbid": 211, "correlation": 267169
    }
  },
  {
    "ph": "s", "id": 267169, "pid": 76337, "tid": -914061504, "ts": 1716454225624569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225690073, "dur": 5,
    "args": {
      "External id": 267171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267171,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267171, "pid": 5, "tid": 7, "ts": 1716454225690073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624582, "dur": 6,
    "args": {
      "External id": 267171,
      "cbid": 211, "correlation": 267171
    }
  },
  {
    "ph": "s", "id": 267171, "pid": 76337, "tid": -914061504, "ts": 1716454225624582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624639, "dur": 1,
    "args": {
      "External id": 267182,
      "cbid": 251, "correlation": 267182
    }
  },
  {
    "ph": "f", "id": 267182, "pid": 76337, "tid": -914061504, "ts": 1716454225624639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624643, "dur": 0,
    "args": {
      "External id": 267183,
      "cbid": 251, "correlation": 267183
    }
  },
  {
    "ph": "f", "id": 267183, "pid": 76337, "tid": -914061504, "ts": 1716454225624643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225690079, "dur": 8,
    "args": {
      "External id": 267184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267184,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267184, "pid": 5, "tid": 7, "ts": 1716454225690079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624644, "dur": 12,
    "args": {
      "External id": 267184,
      "cbid": 211, "correlation": 267184
    }
  },
  {
    "ph": "s", "id": 267184, "pid": 76337, "tid": -914061504, "ts": 1716454225624644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225690089, "dur": 3,
    "args": {
      "External id": 267186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267186,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267186, "pid": 5, "tid": 7, "ts": 1716454225690089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624658, "dur": 6,
    "args": {
      "External id": 267186,
      "cbid": 211, "correlation": 267186
    }
  },
  {
    "ph": "s", "id": 267186, "pid": 76337, "tid": -914061504, "ts": 1716454225624658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225690093, "dur": 56,
    "args": {
      "External id": 267211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267211,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267211, "pid": 5, "tid": 7, "ts": 1716454225690093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624735, "dur": 12,
    "args": {
      "External id": 267211,
      "cbid": 211, "correlation": 267211
    }
  },
  {
    "ph": "s", "id": 267211, "pid": 76337, "tid": -914061504, "ts": 1716454225624735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624832, "dur": 1,
    "args": {
      "External id": 267229,
      "cbid": 251, "correlation": 267229
    }
  },
  {
    "ph": "f", "id": 267229, "pid": 76337, "tid": -914061504, "ts": 1716454225624832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225690151, "dur": 92,
    "args": {
      "External id": 267231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267231,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 267231, "pid": 5, "tid": 7, "ts": 1716454225690151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624838, "dur": 14,
    "args": {
      "External id": 267231,
      "cbid": 211, "correlation": 267231
    }
  },
  {
    "ph": "s", "id": 267231, "pid": 76337, "tid": -914061504, "ts": 1716454225624838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225690244, "dur": 10,
    "args": {
      "External id": 267239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267239,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267239, "pid": 5, "tid": 7, "ts": 1716454225690244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624908, "dur": 12,
    "args": {
      "External id": 267239,
      "cbid": 211, "correlation": 267239
    }
  },
  {
    "ph": "s", "id": 267239, "pid": 76337, "tid": -914061504, "ts": 1716454225624908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225690255, "dur": 21,
    "args": {
      "External id": 267247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267247, "pid": 5, "tid": 7, "ts": 1716454225690255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225624949, "dur": 10,
    "args": {
      "External id": 267247,
      "cbid": 211, "correlation": 267247
    }
  },
  {
    "ph": "s", "id": 267247, "pid": 76337, "tid": -914061504, "ts": 1716454225624949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225690278, "dur": 18,
    "args": {
      "External id": 267269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267269,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267269, "pid": 5, "tid": 7, "ts": 1716454225690278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625008, "dur": 10,
    "args": {
      "External id": 267269,
      "cbid": 211, "correlation": 267269
    }
  },
  {
    "ph": "s", "id": 267269, "pid": 76337, "tid": -914061504, "ts": 1716454225625008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625095, "dur": 1,
    "args": {
      "External id": 267285,
      "cbid": 251, "correlation": 267285
    }
  },
  {
    "ph": "f", "id": 267285, "pid": 76337, "tid": -914061504, "ts": 1716454225625095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625100, "dur": 0,
    "args": {
      "External id": 267287,
      "cbid": 251, "correlation": 267287
    }
  },
  {
    "ph": "f", "id": 267287, "pid": 76337, "tid": -914061504, "ts": 1716454225625100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225690297, "dur": 498,
    "args": {
      "External id": 267288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267288,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267288, "pid": 5, "tid": 7, "ts": 1716454225690297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625102, "dur": 13,
    "args": {
      "External id": 267288,
      "cbid": 211, "correlation": 267288
    }
  },
  {
    "ph": "s", "id": 267288, "pid": 76337, "tid": -914061504, "ts": 1716454225625102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225690797, "dur": 66,
    "args": {
      "External id": 267296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267296,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267296, "pid": 5, "tid": 7, "ts": 1716454225690797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625167, "dur": 12,
    "args": {
      "External id": 267296,
      "cbid": 211, "correlation": 267296
    }
  },
  {
    "ph": "s", "id": 267296, "pid": 76337, "tid": -914061504, "ts": 1716454225625167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225690864, "dur": 66,
    "args": {
      "External id": 267304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267304, "pid": 5, "tid": 7, "ts": 1716454225690864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625197, "dur": 9,
    "args": {
      "External id": 267304,
      "cbid": 211, "correlation": 267304
    }
  },
  {
    "ph": "s", "id": 267304, "pid": 76337, "tid": -914061504, "ts": 1716454225625197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625277, "dur": 1,
    "args": {
      "External id": 267320,
      "cbid": 251, "correlation": 267320
    }
  },
  {
    "ph": "f", "id": 267320, "pid": 76337, "tid": -914061504, "ts": 1716454225625277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225690933, "dur": 1,
    "args": {
      "External id": 267322,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 267322,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 267322, "pid": 5, "tid": 7, "ts": 1716454225690933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625282, "dur": 9,
    "args": {
      "External id": 267322,
      "cbid": 51, "correlation": 267322
    }
  },
  {
    "ph": "s", "id": 267322, "pid": 76337, "tid": -914061504, "ts": 1716454225625282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225690937, "dur": 274,
    "args": {
      "External id": 267323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267323,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 267323, "pid": 5, "tid": 7, "ts": 1716454225690937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625292, "dur": 11,
    "args": {
      "External id": 267323,
      "cbid": 211, "correlation": 267323
    }
  },
  {
    "ph": "s", "id": 267323, "pid": 76337, "tid": -914061504, "ts": 1716454225625292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225691212, "dur": 13,
    "args": {
      "External id": 267331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267331,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267331, "pid": 5, "tid": 7, "ts": 1716454225691212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625335, "dur": 10,
    "args": {
      "External id": 267331,
      "cbid": 211, "correlation": 267331
    }
  },
  {
    "ph": "s", "id": 267331, "pid": 76337, "tid": -914061504, "ts": 1716454225625335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225691227, "dur": 38,
    "args": {
      "External id": 267342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267342,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267342, "pid": 5, "tid": 7, "ts": 1716454225691227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625402, "dur": 12,
    "args": {
      "External id": 267342,
      "cbid": 211, "correlation": 267342
    }
  },
  {
    "ph": "s", "id": 267342, "pid": 76337, "tid": -914061504, "ts": 1716454225625402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625465, "dur": 0,
    "args": {
      "External id": 267354,
      "cbid": 317, "correlation": 267354
    }
  },
  {
    "ph": "f", "id": 267354, "pid": 76337, "tid": -914061504, "ts": 1716454225625465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625466, "dur": 0,
    "args": {
      "External id": 267355,
      "cbid": 203, "correlation": 267355
    }
  },
  {
    "ph": "f", "id": 267355, "pid": 76337, "tid": -914061504, "ts": 1716454225625466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625466, "dur": 0,
    "args": {
      "External id": 267356,
      "cbid": 205, "correlation": 267356
    }
  },
  {
    "ph": "f", "id": 267356, "pid": 76337, "tid": -914061504, "ts": 1716454225625466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225691266, "dur": 13,
    "args": {
      "External id": 267360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267360,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267360, "pid": 5, "tid": 7, "ts": 1716454225691266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625481, "dur": 12,
    "args": {
      "External id": 267360,
      "cbid": 211, "correlation": 267360
    }
  },
  {
    "ph": "s", "id": 267360, "pid": 76337, "tid": -914061504, "ts": 1716454225625481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225691280, "dur": 4,
    "args": {
      "External id": 267362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 267362, "pid": 5, "tid": 7, "ts": 1716454225691280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625497, "dur": 5,
    "args": {
      "External id": 267362,
      "cbid": 211, "correlation": 267362
    }
  },
  {
    "ph": "s", "id": 267362, "pid": 76337, "tid": -914061504, "ts": 1716454225625497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625506, "dur": 0,
    "args": {
      "External id": 267363,
      "cbid": 51, "correlation": 267363
    }
  },
  {
    "ph": "s", "id": 267363, "pid": 76337, "tid": -914061504, "ts": 1716454225625506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225691286, "dur": 99,
    "args": {
      "External id": 267364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267364,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 267364, "pid": 5, "tid": 7, "ts": 1716454225691286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625506, "dur": 6,
    "args": {
      "External id": 267364,
      "cbid": 211, "correlation": 267364
    }
  },
  {
    "ph": "s", "id": 267364, "pid": 76337, "tid": -914061504, "ts": 1716454225625506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225691386, "dur": 16,
    "args": {
      "External id": 267369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267369,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267369, "pid": 5, "tid": 7, "ts": 1716454225691386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625534, "dur": 8,
    "args": {
      "External id": 267369,
      "cbid": 211, "correlation": 267369
    }
  },
  {
    "ph": "s", "id": 267369, "pid": 76337, "tid": -914061504, "ts": 1716454225625534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225691404, "dur": 12,
    "args": {
      "External id": 267377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267377,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267377, "pid": 5, "tid": 7, "ts": 1716454225691404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625565, "dur": 8,
    "args": {
      "External id": 267377,
      "cbid": 211, "correlation": 267377
    }
  },
  {
    "ph": "s", "id": 267377, "pid": 76337, "tid": -914061504, "ts": 1716454225625565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225691418, "dur": 58,
    "args": {
      "External id": 267388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267388,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267388, "pid": 5, "tid": 7, "ts": 1716454225691418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625626, "dur": 11,
    "args": {
      "External id": 267388,
      "cbid": 211, "correlation": 267388
    }
  },
  {
    "ph": "s", "id": 267388, "pid": 76337, "tid": -914061504, "ts": 1716454225625626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625681, "dur": 0,
    "args": {
      "External id": 267398,
      "cbid": 317, "correlation": 267398
    }
  },
  {
    "ph": "f", "id": 267398, "pid": 76337, "tid": -914061504, "ts": 1716454225625681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625682, "dur": 0,
    "args": {
      "External id": 267399,
      "cbid": 203, "correlation": 267399
    }
  },
  {
    "ph": "f", "id": 267399, "pid": 76337, "tid": -914061504, "ts": 1716454225625682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625683, "dur": 0,
    "args": {
      "External id": 267400,
      "cbid": 205, "correlation": 267400
    }
  },
  {
    "ph": "f", "id": 267400, "pid": 76337, "tid": -914061504, "ts": 1716454225625683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225691477, "dur": 39,
    "args": {
      "External id": 267404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267404, "pid": 5, "tid": 7, "ts": 1716454225691477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625699, "dur": 12,
    "args": {
      "External id": 267404,
      "cbid": 211, "correlation": 267404
    }
  },
  {
    "ph": "s", "id": 267404, "pid": 76337, "tid": -914061504, "ts": 1716454225625699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225691518, "dur": 165,
    "args": {
      "External id": 267406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267406,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267406, "pid": 5, "tid": 7, "ts": 1716454225691518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625713, "dur": 6,
    "args": {
      "External id": 267406,
      "cbid": 211, "correlation": 267406
    }
  },
  {
    "ph": "s", "id": 267406, "pid": 76337, "tid": -914061504, "ts": 1716454225625713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225691684, "dur": 1962,
    "args": {
      "External id": 267408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267408,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267408, "pid": 5, "tid": 7, "ts": 1716454225691684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625726, "dur": 8,
    "args": {
      "External id": 267408,
      "cbid": 211, "correlation": 267408
    }
  },
  {
    "ph": "s", "id": 267408, "pid": 76337, "tid": -914061504, "ts": 1716454225625726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225693647, "dur": 40,
    "args": {
      "External id": 267410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267410, "pid": 5, "tid": 7, "ts": 1716454225693647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625737, "dur": 5,
    "args": {
      "External id": 267410,
      "cbid": 211, "correlation": 267410
    }
  },
  {
    "ph": "s", "id": 267410, "pid": 76337, "tid": -914061504, "ts": 1716454225625737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225693688, "dur": 60,
    "args": {
      "External id": 267416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267416, "pid": 5, "tid": 7, "ts": 1716454225693688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625765, "dur": 8,
    "args": {
      "External id": 267416,
      "cbid": 211, "correlation": 267416
    }
  },
  {
    "ph": "s", "id": 267416, "pid": 76337, "tid": -914061504, "ts": 1716454225625765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225693749, "dur": 86,
    "args": {
      "External id": 267425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267425,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267425, "pid": 5, "tid": 7, "ts": 1716454225693749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625855, "dur": 14,
    "args": {
      "External id": 267425,
      "cbid": 211, "correlation": 267425
    }
  },
  {
    "ph": "s", "id": 267425, "pid": 76337, "tid": -914061504, "ts": 1716454225625855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225693837, "dur": 74,
    "args": {
      "External id": 267445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267445,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 267445, "pid": 5, "tid": 7, "ts": 1716454225693837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625926, "dur": 11,
    "args": {
      "External id": 267445,
      "cbid": 211, "correlation": 267445
    }
  },
  {
    "ph": "s", "id": 267445, "pid": 76337, "tid": -914061504, "ts": 1716454225625926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225693912, "dur": 5,
    "args": {
      "External id": 267457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267457,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 267457, "pid": 5, "tid": 7, "ts": 1716454225693912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625947, "dur": 6,
    "args": {
      "External id": 267457,
      "cbid": 211, "correlation": 267457
    }
  },
  {
    "ph": "s", "id": 267457, "pid": 76337, "tid": -914061504, "ts": 1716454225625947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225693918, "dur": 81,
    "args": {
      "External id": 267460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267460, "pid": 5, "tid": 7, "ts": 1716454225693918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225625965, "dur": 16,
    "args": {
      "External id": 267460,
      "cbid": 211, "correlation": 267460
    }
  },
  {
    "ph": "s", "id": 267460, "pid": 76337, "tid": -914061504, "ts": 1716454225625965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225694000, "dur": 55,
    "args": {
      "External id": 267469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267469,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267469, "pid": 5, "tid": 7, "ts": 1716454225694000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626015, "dur": 11,
    "args": {
      "External id": 267469,
      "cbid": 211, "correlation": 267469
    }
  },
  {
    "ph": "s", "id": 267469, "pid": 76337, "tid": -914061504, "ts": 1716454225626015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626069, "dur": 0,
    "args": {
      "External id": 267479,
      "cbid": 317, "correlation": 267479
    }
  },
  {
    "ph": "f", "id": 267479, "pid": 76337, "tid": -914061504, "ts": 1716454225626069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626070, "dur": 0,
    "args": {
      "External id": 267480,
      "cbid": 203, "correlation": 267480
    }
  },
  {
    "ph": "f", "id": 267480, "pid": 76337, "tid": -914061504, "ts": 1716454225626070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626071, "dur": 0,
    "args": {
      "External id": 267481,
      "cbid": 205, "correlation": 267481
    }
  },
  {
    "ph": "f", "id": 267481, "pid": 76337, "tid": -914061504, "ts": 1716454225626071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225694056, "dur": 56,
    "args": {
      "External id": 267485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267485,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267485, "pid": 5, "tid": 7, "ts": 1716454225694056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626087, "dur": 11,
    "args": {
      "External id": 267485,
      "cbid": 211, "correlation": 267485
    }
  },
  {
    "ph": "s", "id": 267485, "pid": 76337, "tid": -914061504, "ts": 1716454225626087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225694114, "dur": 124,
    "args": {
      "External id": 267487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267487,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267487, "pid": 5, "tid": 7, "ts": 1716454225694114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626101, "dur": 5,
    "args": {
      "External id": 267487,
      "cbid": 211, "correlation": 267487
    }
  },
  {
    "ph": "s", "id": 267487, "pid": 76337, "tid": -914061504, "ts": 1716454225626101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225694239, "dur": 1915,
    "args": {
      "External id": 267489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267489,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267489, "pid": 5, "tid": 7, "ts": 1716454225694239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626112, "dur": 6,
    "args": {
      "External id": 267489,
      "cbid": 211, "correlation": 267489
    }
  },
  {
    "ph": "s", "id": 267489, "pid": 76337, "tid": -914061504, "ts": 1716454225626112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225696156, "dur": 20,
    "args": {
      "External id": 267491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267491,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267491, "pid": 5, "tid": 7, "ts": 1716454225696156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626122, "dur": 5,
    "args": {
      "External id": 267491,
      "cbid": 211, "correlation": 267491
    }
  },
  {
    "ph": "s", "id": 267491, "pid": 76337, "tid": -914061504, "ts": 1716454225626122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225696177, "dur": 33,
    "args": {
      "External id": 267497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267497, "pid": 5, "tid": 7, "ts": 1716454225696177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626150, "dur": 8,
    "args": {
      "External id": 267497,
      "cbid": 211, "correlation": 267497
    }
  },
  {
    "ph": "s", "id": 267497, "pid": 76337, "tid": -914061504, "ts": 1716454225626150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225696212, "dur": 3,
    "args": {
      "External id": 267505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267505,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 267505, "pid": 5, "tid": 7, "ts": 1716454225696212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626194, "dur": 9,
    "args": {
      "External id": 267505,
      "cbid": 211, "correlation": 267505
    }
  },
  {
    "ph": "s", "id": 267505, "pid": 76337, "tid": -914061504, "ts": 1716454225626194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626258, "dur": 1,
    "args": {
      "External id": 267521,
      "cbid": 251, "correlation": 267521
    }
  },
  {
    "ph": "f", "id": 267521, "pid": 76337, "tid": -914061504, "ts": 1716454225626258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626263, "dur": 0,
    "args": {
      "External id": 267523,
      "cbid": 251, "correlation": 267523
    }
  },
  {
    "ph": "f", "id": 267523, "pid": 76337, "tid": -914061504, "ts": 1716454225626263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225696217, "dur": 12,
    "args": {
      "External id": 267524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267524,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 267524, "pid": 5, "tid": 7, "ts": 1716454225696217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626265, "dur": 12,
    "args": {
      "External id": 267524,
      "cbid": 211, "correlation": 267524
    }
  },
  {
    "ph": "s", "id": 267524, "pid": 76337, "tid": -914061504, "ts": 1716454225626265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225696230, "dur": 5,
    "args": {
      "External id": 267526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267526,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 267526, "pid": 5, "tid": 7, "ts": 1716454225696230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626279, "dur": 5,
    "args": {
      "External id": 267526,
      "cbid": 211, "correlation": 267526
    }
  },
  {
    "ph": "s", "id": 267526, "pid": 76337, "tid": -914061504, "ts": 1716454225626279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225696236, "dur": 29,
    "args": {
      "External id": 267536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267536, "pid": 5, "tid": 7, "ts": 1716454225696236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626335, "dur": 12,
    "args": {
      "External id": 267536,
      "cbid": 211, "correlation": 267536
    }
  },
  {
    "ph": "s", "id": 267536, "pid": 76337, "tid": -914061504, "ts": 1716454225626335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225696267, "dur": 31,
    "args": {
      "External id": 267556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267556,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 267556, "pid": 5, "tid": 7, "ts": 1716454225696267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626400, "dur": 10,
    "args": {
      "External id": 267556,
      "cbid": 211, "correlation": 267556
    }
  },
  {
    "ph": "s", "id": 267556, "pid": 76337, "tid": -914061504, "ts": 1716454225626400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225696300, "dur": 4,
    "args": {
      "External id": 267568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267568,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 267568, "pid": 5, "tid": 7, "ts": 1716454225696300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626420, "dur": 7,
    "args": {
      "External id": 267568,
      "cbid": 211, "correlation": 267568
    }
  },
  {
    "ph": "s", "id": 267568, "pid": 76337, "tid": -914061504, "ts": 1716454225626420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225696306, "dur": 30,
    "args": {
      "External id": 267571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267571,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267571, "pid": 5, "tid": 7, "ts": 1716454225696306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626438, "dur": 6,
    "args": {
      "External id": 267571,
      "cbid": 211, "correlation": 267571
    }
  },
  {
    "ph": "s", "id": 267571, "pid": 76337, "tid": -914061504, "ts": 1716454225626438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225696337, "dur": 20,
    "args": {
      "External id": 267580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267580,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267580, "pid": 5, "tid": 7, "ts": 1716454225696337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626478, "dur": 9,
    "args": {
      "External id": 267580,
      "cbid": 211, "correlation": 267580
    }
  },
  {
    "ph": "s", "id": 267580, "pid": 76337, "tid": -914061504, "ts": 1716454225626478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626541, "dur": 0,
    "args": {
      "External id": 267590,
      "cbid": 317, "correlation": 267590
    }
  },
  {
    "ph": "f", "id": 267590, "pid": 76337, "tid": -914061504, "ts": 1716454225626541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626542, "dur": 0,
    "args": {
      "External id": 267591,
      "cbid": 203, "correlation": 267591
    }
  },
  {
    "ph": "f", "id": 267591, "pid": 76337, "tid": -914061504, "ts": 1716454225626542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626543, "dur": 0,
    "args": {
      "External id": 267592,
      "cbid": 205, "correlation": 267592
    }
  },
  {
    "ph": "f", "id": 267592, "pid": 76337, "tid": -914061504, "ts": 1716454225626543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225696359, "dur": 23,
    "args": {
      "External id": 267596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267596,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267596, "pid": 5, "tid": 7, "ts": 1716454225696359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626558, "dur": 12,
    "args": {
      "External id": 267596,
      "cbid": 211, "correlation": 267596
    }
  },
  {
    "ph": "s", "id": 267596, "pid": 76337, "tid": -914061504, "ts": 1716454225626558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225696384, "dur": 45,
    "args": {
      "External id": 267598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267598, "pid": 5, "tid": 7, "ts": 1716454225696384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626572, "dur": 6,
    "args": {
      "External id": 267598,
      "cbid": 211, "correlation": 267598
    }
  },
  {
    "ph": "s", "id": 267598, "pid": 76337, "tid": -914061504, "ts": 1716454225626572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225696430, "dur": 654,
    "args": {
      "External id": 267600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267600,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267600, "pid": 5, "tid": 7, "ts": 1716454225696430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626585, "dur": 6,
    "args": {
      "External id": 267600,
      "cbid": 211, "correlation": 267600
    }
  },
  {
    "ph": "s", "id": 267600, "pid": 76337, "tid": -914061504, "ts": 1716454225626585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225697086, "dur": 22,
    "args": {
      "External id": 267602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267602, "pid": 5, "tid": 7, "ts": 1716454225697086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626594, "dur": 5,
    "args": {
      "External id": 267602,
      "cbid": 211, "correlation": 267602
    }
  },
  {
    "ph": "s", "id": 267602, "pid": 76337, "tid": -914061504, "ts": 1716454225626594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225697109, "dur": 33,
    "args": {
      "External id": 267608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267608, "pid": 5, "tid": 7, "ts": 1716454225697109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626622, "dur": 9,
    "args": {
      "External id": 267608,
      "cbid": 211, "correlation": 267608
    }
  },
  {
    "ph": "s", "id": 267608, "pid": 76337, "tid": -914061504, "ts": 1716454225626622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626680, "dur": 0,
    "args": {
      "External id": 267618,
      "cbid": 317, "correlation": 267618
    }
  },
  {
    "ph": "f", "id": 267618, "pid": 76337, "tid": -914061504, "ts": 1716454225626680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626681, "dur": 0,
    "args": {
      "External id": 267619,
      "cbid": 203, "correlation": 267619
    }
  },
  {
    "ph": "f", "id": 267619, "pid": 76337, "tid": -914061504, "ts": 1716454225626681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626682, "dur": 0,
    "args": {
      "External id": 267620,
      "cbid": 205, "correlation": 267620
    }
  },
  {
    "ph": "f", "id": 267620, "pid": 76337, "tid": -914061504, "ts": 1716454225626682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225697143, "dur": 57,
    "args": {
      "External id": 267624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267624, "pid": 5, "tid": 7, "ts": 1716454225697143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626694, "dur": 11,
    "args": {
      "External id": 267624,
      "cbid": 211, "correlation": 267624
    }
  },
  {
    "ph": "s", "id": 267624, "pid": 76337, "tid": -914061504, "ts": 1716454225626694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225697201, "dur": 274,
    "args": {
      "External id": 267626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267626,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267626, "pid": 5, "tid": 7, "ts": 1716454225697201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626712, "dur": 7,
    "args": {
      "External id": 267626,
      "cbid": 211, "correlation": 267626
    }
  },
  {
    "ph": "s", "id": 267626, "pid": 76337, "tid": -914061504, "ts": 1716454225626712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225697477, "dur": 21,
    "args": {
      "External id": 267628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267628,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267628, "pid": 5, "tid": 7, "ts": 1716454225697477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626722, "dur": 6,
    "args": {
      "External id": 267628,
      "cbid": 211, "correlation": 267628
    }
  },
  {
    "ph": "s", "id": 267628, "pid": 76337, "tid": -914061504, "ts": 1716454225626722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225697499, "dur": 34,
    "args": {
      "External id": 267634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267634, "pid": 5, "tid": 7, "ts": 1716454225697499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626749, "dur": 8,
    "args": {
      "External id": 267634,
      "cbid": 211, "correlation": 267634
    }
  },
  {
    "ph": "s", "id": 267634, "pid": 76337, "tid": -914061504, "ts": 1716454225626749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225697534, "dur": 26,
    "args": {
      "External id": 267642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267642,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267642, "pid": 5, "tid": 7, "ts": 1716454225697534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626778, "dur": 8,
    "args": {
      "External id": 267642,
      "cbid": 211, "correlation": 267642
    }
  },
  {
    "ph": "s", "id": 267642, "pid": 76337, "tid": -914061504, "ts": 1716454225626778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225697562, "dur": 20,
    "args": {
      "External id": 267650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267650,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267650, "pid": 5, "tid": 7, "ts": 1716454225697562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626807, "dur": 8,
    "args": {
      "External id": 267650,
      "cbid": 211, "correlation": 267650
    }
  },
  {
    "ph": "s", "id": 267650, "pid": 76337, "tid": -914061504, "ts": 1716454225626807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225697584, "dur": 31,
    "args": {
      "External id": 267670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267670,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 267670, "pid": 5, "tid": 7, "ts": 1716454225697584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626890, "dur": 12,
    "args": {
      "External id": 267670,
      "cbid": 211, "correlation": 267670
    }
  },
  {
    "ph": "s", "id": 267670, "pid": 76337, "tid": -914061504, "ts": 1716454225626890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225697616, "dur": 4,
    "args": {
      "External id": 267682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267682,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 267682, "pid": 5, "tid": 7, "ts": 1716454225697616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626912, "dur": 6,
    "args": {
      "External id": 267682,
      "cbid": 211, "correlation": 267682
    }
  },
  {
    "ph": "s", "id": 267682, "pid": 76337, "tid": -914061504, "ts": 1716454225626912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225697622, "dur": 31,
    "args": {
      "External id": 267685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267685,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267685, "pid": 5, "tid": 7, "ts": 1716454225697622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626929, "dur": 7,
    "args": {
      "External id": 267685,
      "cbid": 211, "correlation": 267685
    }
  },
  {
    "ph": "s", "id": 267685, "pid": 76337, "tid": -914061504, "ts": 1716454225626929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626997, "dur": 0,
    "args": {
      "External id": 267696,
      "cbid": 317, "correlation": 267696
    }
  },
  {
    "ph": "f", "id": 267696, "pid": 76337, "tid": -914061504, "ts": 1716454225626997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626998, "dur": 0,
    "args": {
      "External id": 267697,
      "cbid": 203, "correlation": 267697
    }
  },
  {
    "ph": "f", "id": 267697, "pid": 76337, "tid": -914061504, "ts": 1716454225626998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225626999, "dur": 0,
    "args": {
      "External id": 267698,
      "cbid": 205, "correlation": 267698
    }
  },
  {
    "ph": "f", "id": 267698, "pid": 76337, "tid": -914061504, "ts": 1716454225626999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225697654, "dur": 22,
    "args": {
      "External id": 267702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267702,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267702, "pid": 5, "tid": 7, "ts": 1716454225697654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627013, "dur": 12,
    "args": {
      "External id": 267702,
      "cbid": 211, "correlation": 267702
    }
  },
  {
    "ph": "s", "id": 267702, "pid": 76337, "tid": -914061504, "ts": 1716454225627013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225697677, "dur": 107,
    "args": {
      "External id": 267704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267704,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267704, "pid": 5, "tid": 7, "ts": 1716454225697677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627031, "dur": 7,
    "args": {
      "External id": 267704,
      "cbid": 211, "correlation": 267704
    }
  },
  {
    "ph": "s", "id": 267704, "pid": 76337, "tid": -914061504, "ts": 1716454225627031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225697786, "dur": 22,
    "args": {
      "External id": 267706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267706, "pid": 5, "tid": 7, "ts": 1716454225697786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627042, "dur": 5,
    "args": {
      "External id": 267706,
      "cbid": 211, "correlation": 267706
    }
  },
  {
    "ph": "s", "id": 267706, "pid": 76337, "tid": -914061504, "ts": 1716454225627042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225697809, "dur": 33,
    "args": {
      "External id": 267712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267712, "pid": 5, "tid": 7, "ts": 1716454225697809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627069, "dur": 8,
    "args": {
      "External id": 267712,
      "cbid": 211, "correlation": 267712
    }
  },
  {
    "ph": "s", "id": 267712, "pid": 76337, "tid": -914061504, "ts": 1716454225627069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225697843, "dur": 186,
    "args": {
      "External id": 267721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267721,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267721, "pid": 5, "tid": 7, "ts": 1716454225697843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627151, "dur": 14,
    "args": {
      "External id": 267721,
      "cbid": 211, "correlation": 267721
    }
  },
  {
    "ph": "s", "id": 267721, "pid": 76337, "tid": -914061504, "ts": 1716454225627151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225698030, "dur": 65,
    "args": {
      "External id": 267743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267743,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267743, "pid": 5, "tid": 7, "ts": 1716454225698030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627208, "dur": 10,
    "args": {
      "External id": 267743,
      "cbid": 211, "correlation": 267743
    }
  },
  {
    "ph": "s", "id": 267743, "pid": 76337, "tid": -914061504, "ts": 1716454225627208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627297, "dur": 1,
    "args": {
      "External id": 267754,
      "cbid": 251, "correlation": 267754
    }
  },
  {
    "ph": "f", "id": 267754, "pid": 76337, "tid": -914061504, "ts": 1716454225627297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225698097, "dur": 155,
    "args": {
      "External id": 267755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267755,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267755, "pid": 5, "tid": 7, "ts": 1716454225698097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627302, "dur": 13,
    "args": {
      "External id": 267755,
      "cbid": 211, "correlation": 267755
    }
  },
  {
    "ph": "s", "id": 267755, "pid": 76337, "tid": -914061504, "ts": 1716454225627302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627373, "dur": 1,
    "args": {
      "External id": 267766,
      "cbid": 251, "correlation": 267766
    }
  },
  {
    "ph": "f", "id": 267766, "pid": 76337, "tid": -914061504, "ts": 1716454225627373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225698254, "dur": 148,
    "args": {
      "External id": 267767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267767,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267767, "pid": 5, "tid": 7, "ts": 1716454225698254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627377, "dur": 12,
    "args": {
      "External id": 267767,
      "cbid": 211, "correlation": 267767
    }
  },
  {
    "ph": "s", "id": 267767, "pid": 76337, "tid": -914061504, "ts": 1716454225627377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627440, "dur": 1,
    "args": {
      "External id": 267778,
      "cbid": 251, "correlation": 267778
    }
  },
  {
    "ph": "f", "id": 267778, "pid": 76337, "tid": -914061504, "ts": 1716454225627440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225698403, "dur": 147,
    "args": {
      "External id": 267779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267779,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267779, "pid": 5, "tid": 7, "ts": 1716454225698403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627444, "dur": 11,
    "args": {
      "External id": 267779,
      "cbid": 211, "correlation": 267779
    }
  },
  {
    "ph": "s", "id": 267779, "pid": 76337, "tid": -914061504, "ts": 1716454225627444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225698551, "dur": 1984,
    "args": {
      "External id": 267800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267800,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 267800, "pid": 5, "tid": 7, "ts": 1716454225698551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627526, "dur": 13,
    "args": {
      "External id": 267800,
      "cbid": 211, "correlation": 267800
    }
  },
  {
    "ph": "s", "id": 267800, "pid": 76337, "tid": -914061504, "ts": 1716454225627526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627623, "dur": 1,
    "args": {
      "External id": 267818,
      "cbid": 251, "correlation": 267818
    }
  },
  {
    "ph": "f", "id": 267818, "pid": 76337, "tid": -914061504, "ts": 1716454225627623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225700536, "dur": 150,
    "args": {
      "External id": 267820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267820,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 267820, "pid": 5, "tid": 7, "ts": 1716454225700536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627630, "dur": 14,
    "args": {
      "External id": 267820,
      "cbid": 211, "correlation": 267820
    }
  },
  {
    "ph": "s", "id": 267820, "pid": 76337, "tid": -914061504, "ts": 1716454225627630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225700688, "dur": 35,
    "args": {
      "External id": 267828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267828,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267828, "pid": 5, "tid": 7, "ts": 1716454225700688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627699, "dur": 12,
    "args": {
      "External id": 267828,
      "cbid": 211, "correlation": 267828
    }
  },
  {
    "ph": "s", "id": 267828, "pid": 76337, "tid": -914061504, "ts": 1716454225627699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225700725, "dur": 50,
    "args": {
      "External id": 267836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267836,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267836, "pid": 5, "tid": 7, "ts": 1716454225700725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627738, "dur": 8,
    "args": {
      "External id": 267836,
      "cbid": 211, "correlation": 267836
    }
  },
  {
    "ph": "s", "id": 267836, "pid": 76337, "tid": -914061504, "ts": 1716454225627738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225700776, "dur": 30,
    "args": {
      "External id": 267847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267847, "pid": 5, "tid": 7, "ts": 1716454225700776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627808, "dur": 13,
    "args": {
      "External id": 267847,
      "cbid": 211, "correlation": 267847
    }
  },
  {
    "ph": "s", "id": 267847, "pid": 76337, "tid": -914061504, "ts": 1716454225627808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225700808, "dur": 34,
    "args": {
      "External id": 267869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267869,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267869, "pid": 5, "tid": 7, "ts": 1716454225700808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627839, "dur": 7,
    "args": {
      "External id": 267869,
      "cbid": 211, "correlation": 267869
    }
  },
  {
    "ph": "s", "id": 267869, "pid": 76337, "tid": -914061504, "ts": 1716454225627839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627922, "dur": 1,
    "args": {
      "External id": 267880,
      "cbid": 251, "correlation": 267880
    }
  },
  {
    "ph": "f", "id": 267880, "pid": 76337, "tid": -914061504, "ts": 1716454225627922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225700844, "dur": 92,
    "args": {
      "External id": 267881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267881,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267881, "pid": 5, "tid": 7, "ts": 1716454225700844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225627927, "dur": 14,
    "args": {
      "External id": 267881,
      "cbid": 211, "correlation": 267881
    }
  },
  {
    "ph": "s", "id": 267881, "pid": 76337, "tid": -914061504, "ts": 1716454225627927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628006, "dur": 1,
    "args": {
      "External id": 267892,
      "cbid": 251, "correlation": 267892
    }
  },
  {
    "ph": "f", "id": 267892, "pid": 76337, "tid": -914061504, "ts": 1716454225628006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628010, "dur": 0,
    "args": {
      "External id": 267893,
      "cbid": 251, "correlation": 267893
    }
  },
  {
    "ph": "f", "id": 267893, "pid": 76337, "tid": -914061504, "ts": 1716454225628010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225700937, "dur": 11,
    "args": {
      "External id": 267894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267894,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 267894, "pid": 5, "tid": 7, "ts": 1716454225700937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628011, "dur": 13,
    "args": {
      "External id": 267894,
      "cbid": 211, "correlation": 267894
    }
  },
  {
    "ph": "s", "id": 267894, "pid": 76337, "tid": -914061504, "ts": 1716454225628011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225700949, "dur": 5,
    "args": {
      "External id": 267896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267896,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 267896, "pid": 5, "tid": 7, "ts": 1716454225700949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628026, "dur": 6,
    "args": {
      "External id": 267896,
      "cbid": 211, "correlation": 267896
    }
  },
  {
    "ph": "s", "id": 267896, "pid": 76337, "tid": -914061504, "ts": 1716454225628026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628084, "dur": 1,
    "args": {
      "External id": 267907,
      "cbid": 251, "correlation": 267907
    }
  },
  {
    "ph": "f", "id": 267907, "pid": 76337, "tid": -914061504, "ts": 1716454225628084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628088, "dur": 0,
    "args": {
      "External id": 267908,
      "cbid": 251, "correlation": 267908
    }
  },
  {
    "ph": "f", "id": 267908, "pid": 76337, "tid": -914061504, "ts": 1716454225628088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225700956, "dur": 7,
    "args": {
      "External id": 267909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267909,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 267909, "pid": 5, "tid": 7, "ts": 1716454225700956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628089, "dur": 12,
    "args": {
      "External id": 267909,
      "cbid": 211, "correlation": 267909
    }
  },
  {
    "ph": "s", "id": 267909, "pid": 76337, "tid": -914061504, "ts": 1716454225628089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225700964, "dur": 4,
    "args": {
      "External id": 267911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267911,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 267911, "pid": 5, "tid": 7, "ts": 1716454225700964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628103, "dur": 5,
    "args": {
      "External id": 267911,
      "cbid": 211, "correlation": 267911
    }
  },
  {
    "ph": "s", "id": 267911, "pid": 76337, "tid": -914061504, "ts": 1716454225628103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225700969, "dur": 93,
    "args": {
      "External id": 267932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267932,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 267932, "pid": 5, "tid": 7, "ts": 1716454225700969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628176, "dur": 13,
    "args": {
      "External id": 267932,
      "cbid": 211, "correlation": 267932
    }
  },
  {
    "ph": "s", "id": 267932, "pid": 76337, "tid": -914061504, "ts": 1716454225628176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628272, "dur": 1,
    "args": {
      "External id": 267950,
      "cbid": 251, "correlation": 267950
    }
  },
  {
    "ph": "f", "id": 267950, "pid": 76337, "tid": -914061504, "ts": 1716454225628272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225701064, "dur": 98,
    "args": {
      "External id": 267952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267952,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 267952, "pid": 5, "tid": 7, "ts": 1716454225701064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628278, "dur": 14,
    "args": {
      "External id": 267952,
      "cbid": 211, "correlation": 267952
    }
  },
  {
    "ph": "s", "id": 267952, "pid": 76337, "tid": -914061504, "ts": 1716454225628278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225701163, "dur": 19,
    "args": {
      "External id": 267960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267960,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267960, "pid": 5, "tid": 7, "ts": 1716454225701163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628347, "dur": 12,
    "args": {
      "External id": 267960,
      "cbid": 211, "correlation": 267960
    }
  },
  {
    "ph": "s", "id": 267960, "pid": 76337, "tid": -914061504, "ts": 1716454225628347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225701184, "dur": 38,
    "args": {
      "External id": 267968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267968, "pid": 5, "tid": 7, "ts": 1716454225701184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628388, "dur": 10,
    "args": {
      "External id": 267968,
      "cbid": 211, "correlation": 267968
    }
  },
  {
    "ph": "s", "id": 267968, "pid": 76337, "tid": -914061504, "ts": 1716454225628388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225701223, "dur": 35,
    "args": {
      "External id": 267990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 267990,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 267990, "pid": 5, "tid": 7, "ts": 1716454225701223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628438, "dur": 10,
    "args": {
      "External id": 267990,
      "cbid": 211, "correlation": 267990
    }
  },
  {
    "ph": "s", "id": 267990, "pid": 76337, "tid": -914061504, "ts": 1716454225628438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628526, "dur": 1,
    "args": {
      "External id": 268006,
      "cbid": 251, "correlation": 268006
    }
  },
  {
    "ph": "f", "id": 268006, "pid": 76337, "tid": -914061504, "ts": 1716454225628526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628531, "dur": 0,
    "args": {
      "External id": 268008,
      "cbid": 251, "correlation": 268008
    }
  },
  {
    "ph": "f", "id": 268008, "pid": 76337, "tid": -914061504, "ts": 1716454225628531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225701260, "dur": 549,
    "args": {
      "External id": 268009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268009,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 268009, "pid": 5, "tid": 7, "ts": 1716454225701260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628536, "dur": 13,
    "args": {
      "External id": 268009,
      "cbid": 211, "correlation": 268009
    }
  },
  {
    "ph": "s", "id": 268009, "pid": 76337, "tid": -914061504, "ts": 1716454225628536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225701811, "dur": 128,
    "args": {
      "External id": 268017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268017, "pid": 5, "tid": 7, "ts": 1716454225701811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628602, "dur": 12,
    "args": {
      "External id": 268017,
      "cbid": 211, "correlation": 268017
    }
  },
  {
    "ph": "s", "id": 268017, "pid": 76337, "tid": -914061504, "ts": 1716454225628602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225701940, "dur": 130,
    "args": {
      "External id": 268025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268025, "pid": 5, "tid": 7, "ts": 1716454225701940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628632, "dur": 9,
    "args": {
      "External id": 268025,
      "cbid": 211, "correlation": 268025
    }
  },
  {
    "ph": "s", "id": 268025, "pid": 76337, "tid": -914061504, "ts": 1716454225628632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628709, "dur": 1,
    "args": {
      "External id": 268041,
      "cbid": 251, "correlation": 268041
    }
  },
  {
    "ph": "f", "id": 268041, "pid": 76337, "tid": -914061504, "ts": 1716454225628709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225702071, "dur": 309,
    "args": {
      "External id": 268043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268043,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268043, "pid": 5, "tid": 7, "ts": 1716454225702071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628715, "dur": 12,
    "args": {
      "External id": 268043,
      "cbid": 211, "correlation": 268043
    }
  },
  {
    "ph": "s", "id": 268043, "pid": 76337, "tid": -914061504, "ts": 1716454225628715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225702381, "dur": 27,
    "args": {
      "External id": 268051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268051,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268051, "pid": 5, "tid": 7, "ts": 1716454225702381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628757, "dur": 10,
    "args": {
      "External id": 268051,
      "cbid": 211, "correlation": 268051
    }
  },
  {
    "ph": "s", "id": 268051, "pid": 76337, "tid": -914061504, "ts": 1716454225628757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225702410, "dur": 83,
    "args": {
      "External id": 268062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268062, "pid": 5, "tid": 7, "ts": 1716454225702410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628825, "dur": 12,
    "args": {
      "External id": 268062,
      "cbid": 211, "correlation": 268062
    }
  },
  {
    "ph": "s", "id": 268062, "pid": 76337, "tid": -914061504, "ts": 1716454225628825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628887, "dur": 0,
    "args": {
      "External id": 268074,
      "cbid": 317, "correlation": 268074
    }
  },
  {
    "ph": "f", "id": 268074, "pid": 76337, "tid": -914061504, "ts": 1716454225628887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628888, "dur": 0,
    "args": {
      "External id": 268075,
      "cbid": 203, "correlation": 268075
    }
  },
  {
    "ph": "f", "id": 268075, "pid": 76337, "tid": -914061504, "ts": 1716454225628888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628889, "dur": 0,
    "args": {
      "External id": 268076,
      "cbid": 205, "correlation": 268076
    }
  },
  {
    "ph": "f", "id": 268076, "pid": 76337, "tid": -914061504, "ts": 1716454225628889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225702493, "dur": 22,
    "args": {
      "External id": 268080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268080,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268080, "pid": 5, "tid": 7, "ts": 1716454225702493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628905, "dur": 12,
    "args": {
      "External id": 268080,
      "cbid": 211, "correlation": 268080
    }
  },
  {
    "ph": "s", "id": 268080, "pid": 76337, "tid": -914061504, "ts": 1716454225628905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225702517, "dur": 122,
    "args": {
      "External id": 268082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268082,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268082, "pid": 5, "tid": 7, "ts": 1716454225702517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628922, "dur": 6,
    "args": {
      "External id": 268082,
      "cbid": 211, "correlation": 268082
    }
  },
  {
    "ph": "s", "id": 268082, "pid": 76337, "tid": -914061504, "ts": 1716454225628922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225702640, "dur": 22,
    "args": {
      "External id": 268084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268084, "pid": 5, "tid": 7, "ts": 1716454225702640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628933, "dur": 5,
    "args": {
      "External id": 268084,
      "cbid": 211, "correlation": 268084
    }
  },
  {
    "ph": "s", "id": 268084, "pid": 76337, "tid": -914061504, "ts": 1716454225628933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225702664, "dur": 33,
    "args": {
      "External id": 268090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268090,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268090, "pid": 5, "tid": 7, "ts": 1716454225702664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225628961, "dur": 9,
    "args": {
      "External id": 268090,
      "cbid": 211, "correlation": 268090
    }
  },
  {
    "ph": "s", "id": 268090, "pid": 76337, "tid": -914061504, "ts": 1716454225628961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225702699, "dur": 27,
    "args": {
      "External id": 268098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268098,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268098, "pid": 5, "tid": 7, "ts": 1716454225702699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629003, "dur": 9,
    "args": {
      "External id": 268098,
      "cbid": 211, "correlation": 268098
    }
  },
  {
    "ph": "s", "id": 268098, "pid": 76337, "tid": -914061504, "ts": 1716454225629003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225702727, "dur": 54,
    "args": {
      "External id": 268107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268107,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268107, "pid": 5, "tid": 7, "ts": 1716454225702727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629042, "dur": 11,
    "args": {
      "External id": 268107,
      "cbid": 211, "correlation": 268107
    }
  },
  {
    "ph": "s", "id": 268107, "pid": 76337, "tid": -914061504, "ts": 1716454225629042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225702782, "dur": 52,
    "args": {
      "External id": 268127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268127,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 268127, "pid": 5, "tid": 7, "ts": 1716454225702782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629112, "dur": 11,
    "args": {
      "External id": 268127,
      "cbid": 211, "correlation": 268127
    }
  },
  {
    "ph": "s", "id": 268127, "pid": 76337, "tid": -914061504, "ts": 1716454225629112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225702836, "dur": 5,
    "args": {
      "External id": 268139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268139,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 268139, "pid": 5, "tid": 7, "ts": 1716454225702836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629133, "dur": 6,
    "args": {
      "External id": 268139,
      "cbid": 211, "correlation": 268139
    }
  },
  {
    "ph": "s", "id": 268139, "pid": 76337, "tid": -914061504, "ts": 1716454225629133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225702842, "dur": 58,
    "args": {
      "External id": 268142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268142, "pid": 5, "tid": 7, "ts": 1716454225702842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629152, "dur": 6,
    "args": {
      "External id": 268142,
      "cbid": 211, "correlation": 268142
    }
  },
  {
    "ph": "s", "id": 268142, "pid": 76337, "tid": -914061504, "ts": 1716454225629152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225702902, "dur": 37,
    "args": {
      "External id": 268151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268151,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268151, "pid": 5, "tid": 7, "ts": 1716454225702902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629190, "dur": 11,
    "args": {
      "External id": 268151,
      "cbid": 211, "correlation": 268151
    }
  },
  {
    "ph": "s", "id": 268151, "pid": 76337, "tid": -914061504, "ts": 1716454225629190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629242, "dur": 0,
    "args": {
      "External id": 268161,
      "cbid": 317, "correlation": 268161
    }
  },
  {
    "ph": "f", "id": 268161, "pid": 76337, "tid": -914061504, "ts": 1716454225629242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629243, "dur": 0,
    "args": {
      "External id": 268162,
      "cbid": 203, "correlation": 268162
    }
  },
  {
    "ph": "f", "id": 268162, "pid": 76337, "tid": -914061504, "ts": 1716454225629243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629243, "dur": 0,
    "args": {
      "External id": 268163,
      "cbid": 205, "correlation": 268163
    }
  },
  {
    "ph": "f", "id": 268163, "pid": 76337, "tid": -914061504, "ts": 1716454225629243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225702941, "dur": 40,
    "args": {
      "External id": 268167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268167,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268167, "pid": 5, "tid": 7, "ts": 1716454225702941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629259, "dur": 11,
    "args": {
      "External id": 268167,
      "cbid": 211, "correlation": 268167
    }
  },
  {
    "ph": "s", "id": 268167, "pid": 76337, "tid": -914061504, "ts": 1716454225629259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225702982, "dur": 84,
    "args": {
      "External id": 268169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268169,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268169, "pid": 5, "tid": 7, "ts": 1716454225702982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629273, "dur": 5,
    "args": {
      "External id": 268169,
      "cbid": 211, "correlation": 268169
    }
  },
  {
    "ph": "s", "id": 268169, "pid": 76337, "tid": -914061504, "ts": 1716454225629273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225703067, "dur": 1297,
    "args": {
      "External id": 268171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268171,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268171, "pid": 5, "tid": 7, "ts": 1716454225703067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629284, "dur": 6,
    "args": {
      "External id": 268171,
      "cbid": 211, "correlation": 268171
    }
  },
  {
    "ph": "s", "id": 268171, "pid": 76337, "tid": -914061504, "ts": 1716454225629284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225704366, "dur": 22,
    "args": {
      "External id": 268173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268173, "pid": 5, "tid": 7, "ts": 1716454225704366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629294, "dur": 5,
    "args": {
      "External id": 268173,
      "cbid": 211, "correlation": 268173
    }
  },
  {
    "ph": "s", "id": 268173, "pid": 76337, "tid": -914061504, "ts": 1716454225629294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225704390, "dur": 33,
    "args": {
      "External id": 268179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268179,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268179, "pid": 5, "tid": 7, "ts": 1716454225704390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629322, "dur": 8,
    "args": {
      "External id": 268179,
      "cbid": 211, "correlation": 268179
    }
  },
  {
    "ph": "s", "id": 268179, "pid": 76337, "tid": -914061504, "ts": 1716454225629322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225704424, "dur": 3,
    "args": {
      "External id": 268187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268187,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 268187, "pid": 5, "tid": 7, "ts": 1716454225704424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629365, "dur": 9,
    "args": {
      "External id": 268187,
      "cbid": 211, "correlation": 268187
    }
  },
  {
    "ph": "s", "id": 268187, "pid": 76337, "tid": -914061504, "ts": 1716454225629365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629429, "dur": 1,
    "args": {
      "External id": 268203,
      "cbid": 251, "correlation": 268203
    }
  },
  {
    "ph": "f", "id": 268203, "pid": 76337, "tid": -914061504, "ts": 1716454225629429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629434, "dur": 0,
    "args": {
      "External id": 268205,
      "cbid": 251, "correlation": 268205
    }
  },
  {
    "ph": "f", "id": 268205, "pid": 76337, "tid": -914061504, "ts": 1716454225629434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225704429, "dur": 13,
    "args": {
      "External id": 268206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268206,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 268206, "pid": 5, "tid": 7, "ts": 1716454225704429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629436, "dur": 11,
    "args": {
      "External id": 268206,
      "cbid": 211, "correlation": 268206
    }
  },
  {
    "ph": "s", "id": 268206, "pid": 76337, "tid": -914061504, "ts": 1716454225629436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225704443, "dur": 5,
    "args": {
      "External id": 268208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268208,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 268208, "pid": 5, "tid": 7, "ts": 1716454225704443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629449, "dur": 6,
    "args": {
      "External id": 268208,
      "cbid": 211, "correlation": 268208
    }
  },
  {
    "ph": "s", "id": 268208, "pid": 76337, "tid": -914061504, "ts": 1716454225629449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225704449, "dur": 30,
    "args": {
      "External id": 268218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268218,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268218, "pid": 5, "tid": 7, "ts": 1716454225704449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629508, "dur": 12,
    "args": {
      "External id": 268218,
      "cbid": 211, "correlation": 268218
    }
  },
  {
    "ph": "s", "id": 268218, "pid": 76337, "tid": -914061504, "ts": 1716454225629508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225704480, "dur": 32,
    "args": {
      "External id": 268238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268238,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 268238, "pid": 5, "tid": 7, "ts": 1716454225704480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629573, "dur": 12,
    "args": {
      "External id": 268238,
      "cbid": 211, "correlation": 268238
    }
  },
  {
    "ph": "s", "id": 268238, "pid": 76337, "tid": -914061504, "ts": 1716454225629573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225704513, "dur": 4,
    "args": {
      "External id": 268250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268250,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 268250, "pid": 5, "tid": 7, "ts": 1716454225704513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629595, "dur": 6,
    "args": {
      "External id": 268250,
      "cbid": 211, "correlation": 268250
    }
  },
  {
    "ph": "s", "id": 268250, "pid": 76337, "tid": -914061504, "ts": 1716454225629595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225704519, "dur": 30,
    "args": {
      "External id": 268253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268253, "pid": 5, "tid": 7, "ts": 1716454225704519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629615, "dur": 6,
    "args": {
      "External id": 268253,
      "cbid": 211, "correlation": 268253
    }
  },
  {
    "ph": "s", "id": 268253, "pid": 76337, "tid": -914061504, "ts": 1716454225629615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225704550, "dur": 20,
    "args": {
      "External id": 268262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268262,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268262, "pid": 5, "tid": 7, "ts": 1716454225704550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629654, "dur": 10,
    "args": {
      "External id": 268262,
      "cbid": 211, "correlation": 268262
    }
  },
  {
    "ph": "s", "id": 268262, "pid": 76337, "tid": -914061504, "ts": 1716454225629654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629716, "dur": 0,
    "args": {
      "External id": 268272,
      "cbid": 317, "correlation": 268272
    }
  },
  {
    "ph": "f", "id": 268272, "pid": 76337, "tid": -914061504, "ts": 1716454225629716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629717, "dur": 0,
    "args": {
      "External id": 268273,
      "cbid": 203, "correlation": 268273
    }
  },
  {
    "ph": "f", "id": 268273, "pid": 76337, "tid": -914061504, "ts": 1716454225629717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629717, "dur": 0,
    "args": {
      "External id": 268274,
      "cbid": 205, "correlation": 268274
    }
  },
  {
    "ph": "f", "id": 268274, "pid": 76337, "tid": -914061504, "ts": 1716454225629717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225704571, "dur": 22,
    "args": {
      "External id": 268278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268278,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268278, "pid": 5, "tid": 7, "ts": 1716454225704571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629731, "dur": 12,
    "args": {
      "External id": 268278,
      "cbid": 211, "correlation": 268278
    }
  },
  {
    "ph": "s", "id": 268278, "pid": 76337, "tid": -914061504, "ts": 1716454225629731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225704595, "dur": 45,
    "args": {
      "External id": 268280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268280,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268280, "pid": 5, "tid": 7, "ts": 1716454225704595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629746, "dur": 5,
    "args": {
      "External id": 268280,
      "cbid": 211, "correlation": 268280
    }
  },
  {
    "ph": "s", "id": 268280, "pid": 76337, "tid": -914061504, "ts": 1716454225629746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225704641, "dur": 652,
    "args": {
      "External id": 268282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268282,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268282, "pid": 5, "tid": 7, "ts": 1716454225704641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629757, "dur": 6,
    "args": {
      "External id": 268282,
      "cbid": 211, "correlation": 268282
    }
  },
  {
    "ph": "s", "id": 268282, "pid": 76337, "tid": -914061504, "ts": 1716454225629757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225705294, "dur": 22,
    "args": {
      "External id": 268284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268284,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268284, "pid": 5, "tid": 7, "ts": 1716454225705294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629766, "dur": 5,
    "args": {
      "External id": 268284,
      "cbid": 211, "correlation": 268284
    }
  },
  {
    "ph": "s", "id": 268284, "pid": 76337, "tid": -914061504, "ts": 1716454225629766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225705317, "dur": 33,
    "args": {
      "External id": 268290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268290, "pid": 5, "tid": 7, "ts": 1716454225705317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629794, "dur": 9,
    "args": {
      "External id": 268290,
      "cbid": 211, "correlation": 268290
    }
  },
  {
    "ph": "s", "id": 268290, "pid": 76337, "tid": -914061504, "ts": 1716454225629794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629851, "dur": 0,
    "args": {
      "External id": 268300,
      "cbid": 317, "correlation": 268300
    }
  },
  {
    "ph": "f", "id": 268300, "pid": 76337, "tid": -914061504, "ts": 1716454225629851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629852, "dur": 0,
    "args": {
      "External id": 268301,
      "cbid": 203, "correlation": 268301
    }
  },
  {
    "ph": "f", "id": 268301, "pid": 76337, "tid": -914061504, "ts": 1716454225629852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629852, "dur": 0,
    "args": {
      "External id": 268302,
      "cbid": 205, "correlation": 268302
    }
  },
  {
    "ph": "f", "id": 268302, "pid": 76337, "tid": -914061504, "ts": 1716454225629852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225705352, "dur": 39,
    "args": {
      "External id": 268306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268306,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268306, "pid": 5, "tid": 7, "ts": 1716454225705352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629866, "dur": 12,
    "args": {
      "External id": 268306,
      "cbid": 211, "correlation": 268306
    }
  },
  {
    "ph": "s", "id": 268306, "pid": 76337, "tid": -914061504, "ts": 1716454225629866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225705392, "dur": 193,
    "args": {
      "External id": 268308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268308,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268308, "pid": 5, "tid": 7, "ts": 1716454225705392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629883, "dur": 6,
    "args": {
      "External id": 268308,
      "cbid": 211, "correlation": 268308
    }
  },
  {
    "ph": "s", "id": 268308, "pid": 76337, "tid": -914061504, "ts": 1716454225629883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225705586, "dur": 23,
    "args": {
      "External id": 268310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268310, "pid": 5, "tid": 7, "ts": 1716454225705586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629892, "dur": 5,
    "args": {
      "External id": 268310,
      "cbid": 211, "correlation": 268310
    }
  },
  {
    "ph": "s", "id": 268310, "pid": 76337, "tid": -914061504, "ts": 1716454225629892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225705611, "dur": 33,
    "args": {
      "External id": 268316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268316, "pid": 5, "tid": 7, "ts": 1716454225705611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629918, "dur": 8,
    "args": {
      "External id": 268316,
      "cbid": 211, "correlation": 268316
    }
  },
  {
    "ph": "s", "id": 268316, "pid": 76337, "tid": -914061504, "ts": 1716454225629918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225705645, "dur": 27,
    "args": {
      "External id": 268324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268324,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268324, "pid": 5, "tid": 7, "ts": 1716454225705645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629946, "dur": 8,
    "args": {
      "External id": 268324,
      "cbid": 211, "correlation": 268324
    }
  },
  {
    "ph": "s", "id": 268324, "pid": 76337, "tid": -914061504, "ts": 1716454225629946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225705673, "dur": 20,
    "args": {
      "External id": 268332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268332,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268332, "pid": 5, "tid": 7, "ts": 1716454225705673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225629983, "dur": 9,
    "args": {
      "External id": 268332,
      "cbid": 211, "correlation": 268332
    }
  },
  {
    "ph": "s", "id": 268332, "pid": 76337, "tid": -914061504, "ts": 1716454225629983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225705695, "dur": 31,
    "args": {
      "External id": 268352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268352,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 268352, "pid": 5, "tid": 7, "ts": 1716454225705695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630068, "dur": 12,
    "args": {
      "External id": 268352,
      "cbid": 211, "correlation": 268352
    }
  },
  {
    "ph": "s", "id": 268352, "pid": 76337, "tid": -914061504, "ts": 1716454225630068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225705727, "dur": 4,
    "args": {
      "External id": 268364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268364,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 268364, "pid": 5, "tid": 7, "ts": 1716454225705727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630090, "dur": 6,
    "args": {
      "External id": 268364,
      "cbid": 211, "correlation": 268364
    }
  },
  {
    "ph": "s", "id": 268364, "pid": 76337, "tid": -914061504, "ts": 1716454225630090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225705732, "dur": 30,
    "args": {
      "External id": 268367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268367, "pid": 5, "tid": 7, "ts": 1716454225705732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630108, "dur": 7,
    "args": {
      "External id": 268367,
      "cbid": 211, "correlation": 268367
    }
  },
  {
    "ph": "s", "id": 268367, "pid": 76337, "tid": -914061504, "ts": 1716454225630108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630164, "dur": 0,
    "args": {
      "External id": 268378,
      "cbid": 317, "correlation": 268378
    }
  },
  {
    "ph": "f", "id": 268378, "pid": 76337, "tid": -914061504, "ts": 1716454225630164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630165, "dur": 0,
    "args": {
      "External id": 268379,
      "cbid": 203, "correlation": 268379
    }
  },
  {
    "ph": "f", "id": 268379, "pid": 76337, "tid": -914061504, "ts": 1716454225630165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630166, "dur": 0,
    "args": {
      "External id": 268380,
      "cbid": 205, "correlation": 268380
    }
  },
  {
    "ph": "f", "id": 268380, "pid": 76337, "tid": -914061504, "ts": 1716454225630166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225705764, "dur": 23,
    "args": {
      "External id": 268384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268384,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268384, "pid": 5, "tid": 7, "ts": 1716454225705764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630180, "dur": 12,
    "args": {
      "External id": 268384,
      "cbid": 211, "correlation": 268384
    }
  },
  {
    "ph": "s", "id": 268384, "pid": 76337, "tid": -914061504, "ts": 1716454225630180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225705788, "dur": 107,
    "args": {
      "External id": 268386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268386,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268386, "pid": 5, "tid": 7, "ts": 1716454225705788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630197, "dur": 6,
    "args": {
      "External id": 268386,
      "cbid": 211, "correlation": 268386
    }
  },
  {
    "ph": "s", "id": 268386, "pid": 76337, "tid": -914061504, "ts": 1716454225630197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225705897, "dur": 23,
    "args": {
      "External id": 268388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268388,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268388, "pid": 5, "tid": 7, "ts": 1716454225705897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630207, "dur": 5,
    "args": {
      "External id": 268388,
      "cbid": 211, "correlation": 268388
    }
  },
  {
    "ph": "s", "id": 268388, "pid": 76337, "tid": -914061504, "ts": 1716454225630207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225705921, "dur": 33,
    "args": {
      "External id": 268394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268394, "pid": 5, "tid": 7, "ts": 1716454225705921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630235, "dur": 9,
    "args": {
      "External id": 268394,
      "cbid": 211, "correlation": 268394
    }
  },
  {
    "ph": "s", "id": 268394, "pid": 76337, "tid": -914061504, "ts": 1716454225630235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225705955, "dur": 184,
    "args": {
      "External id": 268403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268403,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268403, "pid": 5, "tid": 7, "ts": 1716454225705955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630316, "dur": 14,
    "args": {
      "External id": 268403,
      "cbid": 211, "correlation": 268403
    }
  },
  {
    "ph": "s", "id": 268403, "pid": 76337, "tid": -914061504, "ts": 1716454225630316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225706141, "dur": 66,
    "args": {
      "External id": 268425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268425,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268425, "pid": 5, "tid": 7, "ts": 1716454225706141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630374, "dur": 10,
    "args": {
      "External id": 268425,
      "cbid": 211, "correlation": 268425
    }
  },
  {
    "ph": "s", "id": 268425, "pid": 76337, "tid": -914061504, "ts": 1716454225630374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630460, "dur": 1,
    "args": {
      "External id": 268436,
      "cbid": 251, "correlation": 268436
    }
  },
  {
    "ph": "f", "id": 268436, "pid": 76337, "tid": -914061504, "ts": 1716454225630460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225706208, "dur": 154,
    "args": {
      "External id": 268437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268437,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268437, "pid": 5, "tid": 7, "ts": 1716454225706208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630466, "dur": 14,
    "args": {
      "External id": 268437,
      "cbid": 211, "correlation": 268437
    }
  },
  {
    "ph": "s", "id": 268437, "pid": 76337, "tid": -914061504, "ts": 1716454225630466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630537, "dur": 1,
    "args": {
      "External id": 268448,
      "cbid": 251, "correlation": 268448
    }
  },
  {
    "ph": "f", "id": 268448, "pid": 76337, "tid": -914061504, "ts": 1716454225630537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225706363, "dur": 149,
    "args": {
      "External id": 268449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268449,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268449, "pid": 5, "tid": 7, "ts": 1716454225706363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630541, "dur": 11,
    "args": {
      "External id": 268449,
      "cbid": 211, "correlation": 268449
    }
  },
  {
    "ph": "s", "id": 268449, "pid": 76337, "tid": -914061504, "ts": 1716454225630541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630606, "dur": 1,
    "args": {
      "External id": 268460,
      "cbid": 251, "correlation": 268460
    }
  },
  {
    "ph": "f", "id": 268460, "pid": 76337, "tid": -914061504, "ts": 1716454225630606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225706514, "dur": 145,
    "args": {
      "External id": 268461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268461,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268461, "pid": 5, "tid": 7, "ts": 1716454225706514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630610, "dur": 12,
    "args": {
      "External id": 268461,
      "cbid": 211, "correlation": 268461
    }
  },
  {
    "ph": "s", "id": 268461, "pid": 76337, "tid": -914061504, "ts": 1716454225630610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225706660, "dur": 1988,
    "args": {
      "External id": 268482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268482,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 268482, "pid": 5, "tid": 7, "ts": 1716454225706660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630692, "dur": 12,
    "args": {
      "External id": 268482,
      "cbid": 211, "correlation": 268482
    }
  },
  {
    "ph": "s", "id": 268482, "pid": 76337, "tid": -914061504, "ts": 1716454225630692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630789, "dur": 1,
    "args": {
      "External id": 268500,
      "cbid": 251, "correlation": 268500
    }
  },
  {
    "ph": "f", "id": 268500, "pid": 76337, "tid": -914061504, "ts": 1716454225630789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225708650, "dur": 150,
    "args": {
      "External id": 268502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268502,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 268502, "pid": 5, "tid": 7, "ts": 1716454225708650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630795, "dur": 13,
    "args": {
      "External id": 268502,
      "cbid": 211, "correlation": 268502
    }
  },
  {
    "ph": "s", "id": 268502, "pid": 76337, "tid": -914061504, "ts": 1716454225630795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225708801, "dur": 35,
    "args": {
      "External id": 268510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268510,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268510, "pid": 5, "tid": 7, "ts": 1716454225708801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630864, "dur": 13,
    "args": {
      "External id": 268510,
      "cbid": 211, "correlation": 268510
    }
  },
  {
    "ph": "s", "id": 268510, "pid": 76337, "tid": -914061504, "ts": 1716454225630864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225708838, "dur": 50,
    "args": {
      "External id": 268518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268518,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268518, "pid": 5, "tid": 7, "ts": 1716454225708838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630903, "dur": 8,
    "args": {
      "External id": 268518,
      "cbid": 211, "correlation": 268518
    }
  },
  {
    "ph": "s", "id": 268518, "pid": 76337, "tid": -914061504, "ts": 1716454225630903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225708889, "dur": 30,
    "args": {
      "External id": 268529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268529, "pid": 5, "tid": 7, "ts": 1716454225708889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225630986, "dur": 13,
    "args": {
      "External id": 268529,
      "cbid": 211, "correlation": 268529
    }
  },
  {
    "ph": "s", "id": 268529, "pid": 76337, "tid": -914061504, "ts": 1716454225630986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225708921, "dur": 35,
    "args": {
      "External id": 268551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268551,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268551, "pid": 5, "tid": 7, "ts": 1716454225708921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631018, "dur": 8,
    "args": {
      "External id": 268551,
      "cbid": 211, "correlation": 268551
    }
  },
  {
    "ph": "s", "id": 268551, "pid": 76337, "tid": -914061504, "ts": 1716454225631018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631104, "dur": 1,
    "args": {
      "External id": 268562,
      "cbid": 251, "correlation": 268562
    }
  },
  {
    "ph": "f", "id": 268562, "pid": 76337, "tid": -914061504, "ts": 1716454225631104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225708957, "dur": 78,
    "args": {
      "External id": 268563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268563,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268563, "pid": 5, "tid": 7, "ts": 1716454225708957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631109, "dur": 13,
    "args": {
      "External id": 268563,
      "cbid": 211, "correlation": 268563
    }
  },
  {
    "ph": "s", "id": 268563, "pid": 76337, "tid": -914061504, "ts": 1716454225631109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631178, "dur": 1,
    "args": {
      "External id": 268574,
      "cbid": 251, "correlation": 268574
    }
  },
  {
    "ph": "f", "id": 268574, "pid": 76337, "tid": -914061504, "ts": 1716454225631178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631182, "dur": 0,
    "args": {
      "External id": 268575,
      "cbid": 251, "correlation": 268575
    }
  },
  {
    "ph": "f", "id": 268575, "pid": 76337, "tid": -914061504, "ts": 1716454225631182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225709036, "dur": 11,
    "args": {
      "External id": 268576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268576,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 268576, "pid": 5, "tid": 7, "ts": 1716454225709036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631183, "dur": 12,
    "args": {
      "External id": 268576,
      "cbid": 211, "correlation": 268576
    }
  },
  {
    "ph": "s", "id": 268576, "pid": 76337, "tid": -914061504, "ts": 1716454225631183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225709049, "dur": 5,
    "args": {
      "External id": 268578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268578,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 268578, "pid": 5, "tid": 7, "ts": 1716454225709049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631197, "dur": 6,
    "args": {
      "External id": 268578,
      "cbid": 211, "correlation": 268578
    }
  },
  {
    "ph": "s", "id": 268578, "pid": 76337, "tid": -914061504, "ts": 1716454225631197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631255, "dur": 1,
    "args": {
      "External id": 268589,
      "cbid": 251, "correlation": 268589
    }
  },
  {
    "ph": "f", "id": 268589, "pid": 76337, "tid": -914061504, "ts": 1716454225631255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631258, "dur": 0,
    "args": {
      "External id": 268590,
      "cbid": 251, "correlation": 268590
    }
  },
  {
    "ph": "f", "id": 268590, "pid": 76337, "tid": -914061504, "ts": 1716454225631258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225709055, "dur": 7,
    "args": {
      "External id": 268591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268591,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 268591, "pid": 5, "tid": 7, "ts": 1716454225709055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631260, "dur": 11,
    "args": {
      "External id": 268591,
      "cbid": 211, "correlation": 268591
    }
  },
  {
    "ph": "s", "id": 268591, "pid": 76337, "tid": -914061504, "ts": 1716454225631260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225709063, "dur": 4,
    "args": {
      "External id": 268593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268593,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 268593, "pid": 5, "tid": 7, "ts": 1716454225709063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631273, "dur": 6,
    "args": {
      "External id": 268593,
      "cbid": 211, "correlation": 268593
    }
  },
  {
    "ph": "s", "id": 268593, "pid": 76337, "tid": -914061504, "ts": 1716454225631273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225709068, "dur": 93,
    "args": {
      "External id": 268614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268614,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 268614, "pid": 5, "tid": 7, "ts": 1716454225709068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631347, "dur": 13,
    "args": {
      "External id": 268614,
      "cbid": 211, "correlation": 268614
    }
  },
  {
    "ph": "s", "id": 268614, "pid": 76337, "tid": -914061504, "ts": 1716454225631347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631444, "dur": 1,
    "args": {
      "External id": 268632,
      "cbid": 251, "correlation": 268632
    }
  },
  {
    "ph": "f", "id": 268632, "pid": 76337, "tid": -914061504, "ts": 1716454225631444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225709163, "dur": 97,
    "args": {
      "External id": 268634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268634,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268634, "pid": 5, "tid": 7, "ts": 1716454225709163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631450, "dur": 13,
    "args": {
      "External id": 268634,
      "cbid": 211, "correlation": 268634
    }
  },
  {
    "ph": "s", "id": 268634, "pid": 76337, "tid": -914061504, "ts": 1716454225631450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225709262, "dur": 19,
    "args": {
      "External id": 268642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268642,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268642, "pid": 5, "tid": 7, "ts": 1716454225709262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631518, "dur": 13,
    "args": {
      "External id": 268642,
      "cbid": 211, "correlation": 268642
    }
  },
  {
    "ph": "s", "id": 268642, "pid": 76337, "tid": -914061504, "ts": 1716454225631518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225709282, "dur": 38,
    "args": {
      "External id": 268650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268650, "pid": 5, "tid": 7, "ts": 1716454225709282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631561, "dur": 9,
    "args": {
      "External id": 268650,
      "cbid": 211, "correlation": 268650
    }
  },
  {
    "ph": "s", "id": 268650, "pid": 76337, "tid": -914061504, "ts": 1716454225631561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225709322, "dur": 36,
    "args": {
      "External id": 268672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268672,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268672, "pid": 5, "tid": 7, "ts": 1716454225709322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631612, "dur": 10,
    "args": {
      "External id": 268672,
      "cbid": 211, "correlation": 268672
    }
  },
  {
    "ph": "s", "id": 268672, "pid": 76337, "tid": -914061504, "ts": 1716454225631612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631701, "dur": 1,
    "args": {
      "External id": 268688,
      "cbid": 251, "correlation": 268688
    }
  },
  {
    "ph": "f", "id": 268688, "pid": 76337, "tid": -914061504, "ts": 1716454225631701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631706, "dur": 0,
    "args": {
      "External id": 268690,
      "cbid": 251, "correlation": 268690
    }
  },
  {
    "ph": "f", "id": 268690, "pid": 76337, "tid": -914061504, "ts": 1716454225631706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225709359, "dur": 548,
    "args": {
      "External id": 268691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268691,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 268691, "pid": 5, "tid": 7, "ts": 1716454225709359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631710, "dur": 12,
    "args": {
      "External id": 268691,
      "cbid": 211, "correlation": 268691
    }
  },
  {
    "ph": "s", "id": 268691, "pid": 76337, "tid": -914061504, "ts": 1716454225631710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225709909, "dur": 126,
    "args": {
      "External id": 268699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268699, "pid": 5, "tid": 7, "ts": 1716454225709909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631774, "dur": 13,
    "args": {
      "External id": 268699,
      "cbid": 211, "correlation": 268699
    }
  },
  {
    "ph": "s", "id": 268699, "pid": 76337, "tid": -914061504, "ts": 1716454225631774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225710036, "dur": 131,
    "args": {
      "External id": 268707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268707,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268707, "pid": 5, "tid": 7, "ts": 1716454225710036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631805, "dur": 8,
    "args": {
      "External id": 268707,
      "cbid": 211, "correlation": 268707
    }
  },
  {
    "ph": "s", "id": 268707, "pid": 76337, "tid": -914061504, "ts": 1716454225631805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631880, "dur": 1,
    "args": {
      "External id": 268723,
      "cbid": 251, "correlation": 268723
    }
  },
  {
    "ph": "f", "id": 268723, "pid": 76337, "tid": -914061504, "ts": 1716454225631880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225710168, "dur": 307,
    "args": {
      "External id": 268725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268725,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268725, "pid": 5, "tid": 7, "ts": 1716454225710168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631887, "dur": 13,
    "args": {
      "External id": 268725,
      "cbid": 211, "correlation": 268725
    }
  },
  {
    "ph": "s", "id": 268725, "pid": 76337, "tid": -914061504, "ts": 1716454225631887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225710477, "dur": 27,
    "args": {
      "External id": 268733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268733,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268733, "pid": 5, "tid": 7, "ts": 1716454225710477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225631928, "dur": 11,
    "args": {
      "External id": 268733,
      "cbid": 211, "correlation": 268733
    }
  },
  {
    "ph": "s", "id": 268733, "pid": 76337, "tid": -914061504, "ts": 1716454225631928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225710505, "dur": 84,
    "args": {
      "External id": 268744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268744, "pid": 5, "tid": 7, "ts": 1716454225710505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632056, "dur": 13,
    "args": {
      "External id": 268744,
      "cbid": 211, "correlation": 268744
    }
  },
  {
    "ph": "s", "id": 268744, "pid": 76337, "tid": -914061504, "ts": 1716454225632056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632121, "dur": 0,
    "args": {
      "External id": 268756,
      "cbid": 317, "correlation": 268756
    }
  },
  {
    "ph": "f", "id": 268756, "pid": 76337, "tid": -914061504, "ts": 1716454225632121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632122, "dur": 0,
    "args": {
      "External id": 268757,
      "cbid": 203, "correlation": 268757
    }
  },
  {
    "ph": "f", "id": 268757, "pid": 76337, "tid": -914061504, "ts": 1716454225632122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632123, "dur": 0,
    "args": {
      "External id": 268758,
      "cbid": 205, "correlation": 268758
    }
  },
  {
    "ph": "f", "id": 268758, "pid": 76337, "tid": -914061504, "ts": 1716454225632123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225710590, "dur": 24,
    "args": {
      "External id": 268762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268762,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268762, "pid": 5, "tid": 7, "ts": 1716454225710590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632138, "dur": 12,
    "args": {
      "External id": 268762,
      "cbid": 211, "correlation": 268762
    }
  },
  {
    "ph": "s", "id": 268762, "pid": 76337, "tid": -914061504, "ts": 1716454225632138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225710615, "dur": 123,
    "args": {
      "External id": 268764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268764,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268764, "pid": 5, "tid": 7, "ts": 1716454225710615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632156, "dur": 7,
    "args": {
      "External id": 268764,
      "cbid": 211, "correlation": 268764
    }
  },
  {
    "ph": "s", "id": 268764, "pid": 76337, "tid": -914061504, "ts": 1716454225632156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225710740, "dur": 24,
    "args": {
      "External id": 268766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268766, "pid": 5, "tid": 7, "ts": 1716454225710740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632167, "dur": 5,
    "args": {
      "External id": 268766,
      "cbid": 211, "correlation": 268766
    }
  },
  {
    "ph": "s", "id": 268766, "pid": 76337, "tid": -914061504, "ts": 1716454225632167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225710765, "dur": 33,
    "args": {
      "External id": 268772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268772, "pid": 5, "tid": 7, "ts": 1716454225710765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632195, "dur": 8,
    "args": {
      "External id": 268772,
      "cbid": 211, "correlation": 268772
    }
  },
  {
    "ph": "s", "id": 268772, "pid": 76337, "tid": -914061504, "ts": 1716454225632195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225710799, "dur": 26,
    "args": {
      "External id": 268780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268780,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268780, "pid": 5, "tid": 7, "ts": 1716454225710799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632225, "dur": 9,
    "args": {
      "External id": 268780,
      "cbid": 211, "correlation": 268780
    }
  },
  {
    "ph": "s", "id": 268780, "pid": 76337, "tid": -914061504, "ts": 1716454225632225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225710827, "dur": 44,
    "args": {
      "External id": 268789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268789,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268789, "pid": 5, "tid": 7, "ts": 1716454225710827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632263, "dur": 10,
    "args": {
      "External id": 268789,
      "cbid": 211, "correlation": 268789
    }
  },
  {
    "ph": "s", "id": 268789, "pid": 76337, "tid": -914061504, "ts": 1716454225632263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225710872, "dur": 43,
    "args": {
      "External id": 268809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268809,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 268809, "pid": 5, "tid": 7, "ts": 1716454225710872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632334, "dur": 11,
    "args": {
      "External id": 268809,
      "cbid": 211, "correlation": 268809
    }
  },
  {
    "ph": "s", "id": 268809, "pid": 76337, "tid": -914061504, "ts": 1716454225632334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225710917, "dur": 5,
    "args": {
      "External id": 268821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268821,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 268821, "pid": 5, "tid": 7, "ts": 1716454225710917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632355, "dur": 6,
    "args": {
      "External id": 268821,
      "cbid": 211, "correlation": 268821
    }
  },
  {
    "ph": "s", "id": 268821, "pid": 76337, "tid": -914061504, "ts": 1716454225632355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225710923, "dur": 44,
    "args": {
      "External id": 268824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268824, "pid": 5, "tid": 7, "ts": 1716454225710923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632375, "dur": 7,
    "args": {
      "External id": 268824,
      "cbid": 211, "correlation": 268824
    }
  },
  {
    "ph": "s", "id": 268824, "pid": 76337, "tid": -914061504, "ts": 1716454225632375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225710969, "dur": 29,
    "args": {
      "External id": 268833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268833,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268833, "pid": 5, "tid": 7, "ts": 1716454225710969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632414, "dur": 10,
    "args": {
      "External id": 268833,
      "cbid": 211, "correlation": 268833
    }
  },
  {
    "ph": "s", "id": 268833, "pid": 76337, "tid": -914061504, "ts": 1716454225632414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632466, "dur": 0,
    "args": {
      "External id": 268843,
      "cbid": 317, "correlation": 268843
    }
  },
  {
    "ph": "f", "id": 268843, "pid": 76337, "tid": -914061504, "ts": 1716454225632466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632467, "dur": 0,
    "args": {
      "External id": 268844,
      "cbid": 203, "correlation": 268844
    }
  },
  {
    "ph": "f", "id": 268844, "pid": 76337, "tid": -914061504, "ts": 1716454225632467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632468, "dur": 0,
    "args": {
      "External id": 268845,
      "cbid": 205, "correlation": 268845
    }
  },
  {
    "ph": "f", "id": 268845, "pid": 76337, "tid": -914061504, "ts": 1716454225632468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225710999, "dur": 30,
    "args": {
      "External id": 268849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268849, "pid": 5, "tid": 7, "ts": 1716454225710999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632482, "dur": 12,
    "args": {
      "External id": 268849,
      "cbid": 211, "correlation": 268849
    }
  },
  {
    "ph": "s", "id": 268849, "pid": 76337, "tid": -914061504, "ts": 1716454225632482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225711030, "dur": 64,
    "args": {
      "External id": 268851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268851, "pid": 5, "tid": 7, "ts": 1716454225711030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632497, "dur": 5,
    "args": {
      "External id": 268851,
      "cbid": 211, "correlation": 268851
    }
  },
  {
    "ph": "s", "id": 268851, "pid": 76337, "tid": -914061504, "ts": 1716454225632497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225711096, "dur": 979,
    "args": {
      "External id": 268853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268853,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268853, "pid": 5, "tid": 7, "ts": 1716454225711096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632508, "dur": 6,
    "args": {
      "External id": 268853,
      "cbid": 211, "correlation": 268853
    }
  },
  {
    "ph": "s", "id": 268853, "pid": 76337, "tid": -914061504, "ts": 1716454225632508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225712077, "dur": 22,
    "args": {
      "External id": 268855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268855,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268855, "pid": 5, "tid": 7, "ts": 1716454225712077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632518, "dur": 5,
    "args": {
      "External id": 268855,
      "cbid": 211, "correlation": 268855
    }
  },
  {
    "ph": "s", "id": 268855, "pid": 76337, "tid": -914061504, "ts": 1716454225632518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225712100, "dur": 33,
    "args": {
      "External id": 268861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268861, "pid": 5, "tid": 7, "ts": 1716454225712100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632546, "dur": 8,
    "args": {
      "External id": 268861,
      "cbid": 211, "correlation": 268861
    }
  },
  {
    "ph": "s", "id": 268861, "pid": 76337, "tid": -914061504, "ts": 1716454225632546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225712134, "dur": 3,
    "args": {
      "External id": 268869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268869,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 268869, "pid": 5, "tid": 7, "ts": 1716454225712134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632590, "dur": 10,
    "args": {
      "External id": 268869,
      "cbid": 211, "correlation": 268869
    }
  },
  {
    "ph": "s", "id": 268869, "pid": 76337, "tid": -914061504, "ts": 1716454225632590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632655, "dur": 1,
    "args": {
      "External id": 268885,
      "cbid": 251, "correlation": 268885
    }
  },
  {
    "ph": "f", "id": 268885, "pid": 76337, "tid": -914061504, "ts": 1716454225632655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632660, "dur": 0,
    "args": {
      "External id": 268887,
      "cbid": 251, "correlation": 268887
    }
  },
  {
    "ph": "f", "id": 268887, "pid": 76337, "tid": -914061504, "ts": 1716454225632660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225712139, "dur": 12,
    "args": {
      "External id": 268888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268888,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 268888, "pid": 5, "tid": 7, "ts": 1716454225712139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632662, "dur": 11,
    "args": {
      "External id": 268888,
      "cbid": 211, "correlation": 268888
    }
  },
  {
    "ph": "s", "id": 268888, "pid": 76337, "tid": -914061504, "ts": 1716454225632662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225712153, "dur": 5,
    "args": {
      "External id": 268890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268890,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 268890, "pid": 5, "tid": 7, "ts": 1716454225712153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632675, "dur": 5,
    "args": {
      "External id": 268890,
      "cbid": 211, "correlation": 268890
    }
  },
  {
    "ph": "s", "id": 268890, "pid": 76337, "tid": -914061504, "ts": 1716454225632675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225712159, "dur": 29,
    "args": {
      "External id": 268900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268900, "pid": 5, "tid": 7, "ts": 1716454225712159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632731, "dur": 13,
    "args": {
      "External id": 268900,
      "cbid": 211, "correlation": 268900
    }
  },
  {
    "ph": "s", "id": 268900, "pid": 76337, "tid": -914061504, "ts": 1716454225632731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225712190, "dur": 31,
    "args": {
      "External id": 268920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268920,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 268920, "pid": 5, "tid": 7, "ts": 1716454225712190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632797, "dur": 10,
    "args": {
      "External id": 268920,
      "cbid": 211, "correlation": 268920
    }
  },
  {
    "ph": "s", "id": 268920, "pid": 76337, "tid": -914061504, "ts": 1716454225632797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225712223, "dur": 4,
    "args": {
      "External id": 268932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268932,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 268932, "pid": 5, "tid": 7, "ts": 1716454225712223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632817, "dur": 6,
    "args": {
      "External id": 268932,
      "cbid": 211, "correlation": 268932
    }
  },
  {
    "ph": "s", "id": 268932, "pid": 76337, "tid": -914061504, "ts": 1716454225632817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225712228, "dur": 31,
    "args": {
      "External id": 268935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268935, "pid": 5, "tid": 7, "ts": 1716454225712228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632836, "dur": 7,
    "args": {
      "External id": 268935,
      "cbid": 211, "correlation": 268935
    }
  },
  {
    "ph": "s", "id": 268935, "pid": 76337, "tid": -914061504, "ts": 1716454225632836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225712260, "dur": 21,
    "args": {
      "External id": 268944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268944,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268944, "pid": 5, "tid": 7, "ts": 1716454225712260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632876, "dur": 10,
    "args": {
      "External id": 268944,
      "cbid": 211, "correlation": 268944
    }
  },
  {
    "ph": "s", "id": 268944, "pid": 76337, "tid": -914061504, "ts": 1716454225632876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632938, "dur": 0,
    "args": {
      "External id": 268954,
      "cbid": 317, "correlation": 268954
    }
  },
  {
    "ph": "f", "id": 268954, "pid": 76337, "tid": -914061504, "ts": 1716454225632938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632939, "dur": 0,
    "args": {
      "External id": 268955,
      "cbid": 203, "correlation": 268955
    }
  },
  {
    "ph": "f", "id": 268955, "pid": 76337, "tid": -914061504, "ts": 1716454225632939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632940, "dur": 0,
    "args": {
      "External id": 268956,
      "cbid": 205, "correlation": 268956
    }
  },
  {
    "ph": "f", "id": 268956, "pid": 76337, "tid": -914061504, "ts": 1716454225632940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225712283, "dur": 23,
    "args": {
      "External id": 268960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268960,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268960, "pid": 5, "tid": 7, "ts": 1716454225712283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632954, "dur": 12,
    "args": {
      "External id": 268960,
      "cbid": 211, "correlation": 268960
    }
  },
  {
    "ph": "s", "id": 268960, "pid": 76337, "tid": -914061504, "ts": 1716454225632954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225712307, "dur": 44,
    "args": {
      "External id": 268962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268962,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268962, "pid": 5, "tid": 7, "ts": 1716454225712307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632969, "dur": 13,
    "args": {
      "External id": 268962,
      "cbid": 211, "correlation": 268962
    }
  },
  {
    "ph": "s", "id": 268962, "pid": 76337, "tid": -914061504, "ts": 1716454225632969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225712352, "dur": 653,
    "args": {
      "External id": 268964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268964,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268964, "pid": 5, "tid": 7, "ts": 1716454225712352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632988, "dur": 7,
    "args": {
      "External id": 268964,
      "cbid": 211, "correlation": 268964
    }
  },
  {
    "ph": "s", "id": 268964, "pid": 76337, "tid": -914061504, "ts": 1716454225632988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713006, "dur": 22,
    "args": {
      "External id": 268966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268966,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268966, "pid": 5, "tid": 7, "ts": 1716454225713006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225632999, "dur": 5,
    "args": {
      "External id": 268966,
      "cbid": 211, "correlation": 268966
    }
  },
  {
    "ph": "s", "id": 268966, "pid": 76337, "tid": -914061504, "ts": 1716454225632999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225713030, "dur": 33,
    "args": {
      "External id": 268972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268972, "pid": 5, "tid": 7, "ts": 1716454225713030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633027, "dur": 8,
    "args": {
      "External id": 268972,
      "cbid": 211, "correlation": 268972
    }
  },
  {
    "ph": "s", "id": 268972, "pid": 76337, "tid": -914061504, "ts": 1716454225633027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633085, "dur": 0,
    "args": {
      "External id": 268982,
      "cbid": 317, "correlation": 268982
    }
  },
  {
    "ph": "f", "id": 268982, "pid": 76337, "tid": -914061504, "ts": 1716454225633085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633086, "dur": 0,
    "args": {
      "External id": 268983,
      "cbid": 203, "correlation": 268983
    }
  },
  {
    "ph": "f", "id": 268983, "pid": 76337, "tid": -914061504, "ts": 1716454225633086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633087, "dur": 0,
    "args": {
      "External id": 268984,
      "cbid": 205, "correlation": 268984
    }
  },
  {
    "ph": "f", "id": 268984, "pid": 76337, "tid": -914061504, "ts": 1716454225633087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713064, "dur": 30,
    "args": {
      "External id": 268988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268988, "pid": 5, "tid": 7, "ts": 1716454225713064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633101, "dur": 11,
    "args": {
      "External id": 268988,
      "cbid": 211, "correlation": 268988
    }
  },
  {
    "ph": "s", "id": 268988, "pid": 76337, "tid": -914061504, "ts": 1716454225633101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225713096, "dur": 154,
    "args": {
      "External id": 268990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268990,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 268990, "pid": 5, "tid": 7, "ts": 1716454225713096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633118, "dur": 6,
    "args": {
      "External id": 268990,
      "cbid": 211, "correlation": 268990
    }
  },
  {
    "ph": "s", "id": 268990, "pid": 76337, "tid": -914061504, "ts": 1716454225633118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713251, "dur": 23,
    "args": {
      "External id": 268992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268992, "pid": 5, "tid": 7, "ts": 1716454225713251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633128, "dur": 5,
    "args": {
      "External id": 268992,
      "cbid": 211, "correlation": 268992
    }
  },
  {
    "ph": "s", "id": 268992, "pid": 76337, "tid": -914061504, "ts": 1716454225633128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225713275, "dur": 33,
    "args": {
      "External id": 268998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 268998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 268998, "pid": 5, "tid": 7, "ts": 1716454225713275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633154, "dur": 8,
    "args": {
      "External id": 268998,
      "cbid": 211, "correlation": 268998
    }
  },
  {
    "ph": "s", "id": 268998, "pid": 76337, "tid": -914061504, "ts": 1716454225633154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225713310, "dur": 27,
    "args": {
      "External id": 269006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269006,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269006, "pid": 5, "tid": 7, "ts": 1716454225713310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633182, "dur": 8,
    "args": {
      "External id": 269006,
      "cbid": 211, "correlation": 269006
    }
  },
  {
    "ph": "s", "id": 269006, "pid": 76337, "tid": -914061504, "ts": 1716454225633182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225713338, "dur": 19,
    "args": {
      "External id": 269014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269014,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269014, "pid": 5, "tid": 7, "ts": 1716454225713338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633211, "dur": 8,
    "args": {
      "External id": 269014,
      "cbid": 211, "correlation": 269014
    }
  },
  {
    "ph": "s", "id": 269014, "pid": 76337, "tid": -914061504, "ts": 1716454225633211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713359, "dur": 32,
    "args": {
      "External id": 269034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269034,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 269034, "pid": 5, "tid": 7, "ts": 1716454225713359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633292, "dur": 13,
    "args": {
      "External id": 269034,
      "cbid": 211, "correlation": 269034
    }
  },
  {
    "ph": "s", "id": 269034, "pid": 76337, "tid": -914061504, "ts": 1716454225633292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225713392, "dur": 4,
    "args": {
      "External id": 269046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269046,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 269046, "pid": 5, "tid": 7, "ts": 1716454225713392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633315, "dur": 6,
    "args": {
      "External id": 269046,
      "cbid": 211, "correlation": 269046
    }
  },
  {
    "ph": "s", "id": 269046, "pid": 76337, "tid": -914061504, "ts": 1716454225633315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225713398, "dur": 32,
    "args": {
      "External id": 269049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269049, "pid": 5, "tid": 7, "ts": 1716454225713398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633332, "dur": 7,
    "args": {
      "External id": 269049,
      "cbid": 211, "correlation": 269049
    }
  },
  {
    "ph": "s", "id": 269049, "pid": 76337, "tid": -914061504, "ts": 1716454225633332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633389, "dur": 0,
    "args": {
      "External id": 269060,
      "cbid": 317, "correlation": 269060
    }
  },
  {
    "ph": "f", "id": 269060, "pid": 76337, "tid": -914061504, "ts": 1716454225633389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633390, "dur": 0,
    "args": {
      "External id": 269061,
      "cbid": 203, "correlation": 269061
    }
  },
  {
    "ph": "f", "id": 269061, "pid": 76337, "tid": -914061504, "ts": 1716454225633390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633391, "dur": 0,
    "args": {
      "External id": 269062,
      "cbid": 205, "correlation": 269062
    }
  },
  {
    "ph": "f", "id": 269062, "pid": 76337, "tid": -914061504, "ts": 1716454225633391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713431, "dur": 22,
    "args": {
      "External id": 269066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269066,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269066, "pid": 5, "tid": 7, "ts": 1716454225713431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633404, "dur": 12,
    "args": {
      "External id": 269066,
      "cbid": 211, "correlation": 269066
    }
  },
  {
    "ph": "s", "id": 269066, "pid": 76337, "tid": -914061504, "ts": 1716454225633404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225713454, "dur": 107,
    "args": {
      "External id": 269068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269068,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269068, "pid": 5, "tid": 7, "ts": 1716454225713454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633422, "dur": 6,
    "args": {
      "External id": 269068,
      "cbid": 211, "correlation": 269068
    }
  },
  {
    "ph": "s", "id": 269068, "pid": 76337, "tid": -914061504, "ts": 1716454225633422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713563, "dur": 23,
    "args": {
      "External id": 269070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269070,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269070, "pid": 5, "tid": 7, "ts": 1716454225713563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633432, "dur": 5,
    "args": {
      "External id": 269070,
      "cbid": 211, "correlation": 269070
    }
  },
  {
    "ph": "s", "id": 269070, "pid": 76337, "tid": -914061504, "ts": 1716454225633432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225713587, "dur": 33,
    "args": {
      "External id": 269076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269076,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269076, "pid": 5, "tid": 7, "ts": 1716454225713587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633459, "dur": 8,
    "args": {
      "External id": 269076,
      "cbid": 211, "correlation": 269076
    }
  },
  {
    "ph": "s", "id": 269076, "pid": 76337, "tid": -914061504, "ts": 1716454225633459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225713621, "dur": 190,
    "args": {
      "External id": 269085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269085,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269085, "pid": 5, "tid": 7, "ts": 1716454225713621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633540, "dur": 15,
    "args": {
      "External id": 269085,
      "cbid": 211, "correlation": 269085
    }
  },
  {
    "ph": "s", "id": 269085, "pid": 76337, "tid": -914061504, "ts": 1716454225633540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225713812, "dur": 65,
    "args": {
      "External id": 269107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269107,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269107, "pid": 5, "tid": 7, "ts": 1716454225713812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633602, "dur": 10,
    "args": {
      "External id": 269107,
      "cbid": 211, "correlation": 269107
    }
  },
  {
    "ph": "s", "id": 269107, "pid": 76337, "tid": -914061504, "ts": 1716454225633602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633688, "dur": 1,
    "args": {
      "External id": 269118,
      "cbid": 251, "correlation": 269118
    }
  },
  {
    "ph": "f", "id": 269118, "pid": 76337, "tid": -914061504, "ts": 1716454225633688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225713879, "dur": 152,
    "args": {
      "External id": 269119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269119,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269119, "pid": 5, "tid": 7, "ts": 1716454225713879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633693, "dur": 14,
    "args": {
      "External id": 269119,
      "cbid": 211, "correlation": 269119
    }
  },
  {
    "ph": "s", "id": 269119, "pid": 76337, "tid": -914061504, "ts": 1716454225633693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633763, "dur": 1,
    "args": {
      "External id": 269130,
      "cbid": 251, "correlation": 269130
    }
  },
  {
    "ph": "f", "id": 269130, "pid": 76337, "tid": -914061504, "ts": 1716454225633763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225714032, "dur": 147,
    "args": {
      "External id": 269131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269131,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269131, "pid": 5, "tid": 7, "ts": 1716454225714032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633767, "dur": 11,
    "args": {
      "External id": 269131,
      "cbid": 211, "correlation": 269131
    }
  },
  {
    "ph": "s", "id": 269131, "pid": 76337, "tid": -914061504, "ts": 1716454225633767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633831, "dur": 1,
    "args": {
      "External id": 269142,
      "cbid": 251, "correlation": 269142
    }
  },
  {
    "ph": "f", "id": 269142, "pid": 76337, "tid": -914061504, "ts": 1716454225633831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225714180, "dur": 148,
    "args": {
      "External id": 269143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269143,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269143, "pid": 5, "tid": 7, "ts": 1716454225714180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633836, "dur": 11,
    "args": {
      "External id": 269143,
      "cbid": 211, "correlation": 269143
    }
  },
  {
    "ph": "s", "id": 269143, "pid": 76337, "tid": -914061504, "ts": 1716454225633836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225714330, "dur": 1993,
    "args": {
      "External id": 269164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269164,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 269164, "pid": 5, "tid": 7, "ts": 1716454225714330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225633917, "dur": 12,
    "args": {
      "External id": 269164,
      "cbid": 211, "correlation": 269164
    }
  },
  {
    "ph": "s", "id": 269164, "pid": 76337, "tid": -914061504, "ts": 1716454225633917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634020, "dur": 1,
    "args": {
      "External id": 269182,
      "cbid": 251, "correlation": 269182
    }
  },
  {
    "ph": "f", "id": 269182, "pid": 76337, "tid": -914061504, "ts": 1716454225634020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225716325, "dur": 152,
    "args": {
      "External id": 269184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269184,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 269184, "pid": 5, "tid": 7, "ts": 1716454225716325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634026, "dur": 13,
    "args": {
      "External id": 269184,
      "cbid": 211, "correlation": 269184
    }
  },
  {
    "ph": "s", "id": 269184, "pid": 76337, "tid": -914061504, "ts": 1716454225634026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225716478, "dur": 36,
    "args": {
      "External id": 269192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269192,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269192, "pid": 5, "tid": 7, "ts": 1716454225716478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634097, "dur": 13,
    "args": {
      "External id": 269192,
      "cbid": 211, "correlation": 269192
    }
  },
  {
    "ph": "s", "id": 269192, "pid": 76337, "tid": -914061504, "ts": 1716454225634097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225716515, "dur": 51,
    "args": {
      "External id": 269200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269200,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269200, "pid": 5, "tid": 7, "ts": 1716454225716515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634136, "dur": 8,
    "args": {
      "External id": 269200,
      "cbid": 211, "correlation": 269200
    }
  },
  {
    "ph": "s", "id": 269200, "pid": 76337, "tid": -914061504, "ts": 1716454225634136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225716567, "dur": 30,
    "args": {
      "External id": 269211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269211, "pid": 5, "tid": 7, "ts": 1716454225716567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634206, "dur": 13,
    "args": {
      "External id": 269211,
      "cbid": 211, "correlation": 269211
    }
  },
  {
    "ph": "s", "id": 269211, "pid": 76337, "tid": -914061504, "ts": 1716454225634206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225716598, "dur": 35,
    "args": {
      "External id": 269233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269233,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269233, "pid": 5, "tid": 7, "ts": 1716454225716598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634237, "dur": 7,
    "args": {
      "External id": 269233,
      "cbid": 211, "correlation": 269233
    }
  },
  {
    "ph": "s", "id": 269233, "pid": 76337, "tid": -914061504, "ts": 1716454225634237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634321, "dur": 1,
    "args": {
      "External id": 269244,
      "cbid": 251, "correlation": 269244
    }
  },
  {
    "ph": "f", "id": 269244, "pid": 76337, "tid": -914061504, "ts": 1716454225634321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225716635, "dur": 91,
    "args": {
      "External id": 269245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269245,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269245, "pid": 5, "tid": 7, "ts": 1716454225716635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634326, "dur": 13,
    "args": {
      "External id": 269245,
      "cbid": 211, "correlation": 269245
    }
  },
  {
    "ph": "s", "id": 269245, "pid": 76337, "tid": -914061504, "ts": 1716454225634326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634393, "dur": 1,
    "args": {
      "External id": 269256,
      "cbid": 251, "correlation": 269256
    }
  },
  {
    "ph": "f", "id": 269256, "pid": 76337, "tid": -914061504, "ts": 1716454225634393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634397, "dur": 0,
    "args": {
      "External id": 269257,
      "cbid": 251, "correlation": 269257
    }
  },
  {
    "ph": "f", "id": 269257, "pid": 76337, "tid": -914061504, "ts": 1716454225634397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225716727, "dur": 12,
    "args": {
      "External id": 269258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269258,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 269258, "pid": 5, "tid": 7, "ts": 1716454225716727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634398, "dur": 13,
    "args": {
      "External id": 269258,
      "cbid": 211, "correlation": 269258
    }
  },
  {
    "ph": "s", "id": 269258, "pid": 76337, "tid": -914061504, "ts": 1716454225634398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225716740, "dur": 5,
    "args": {
      "External id": 269260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269260,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 269260, "pid": 5, "tid": 7, "ts": 1716454225716740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634413, "dur": 6,
    "args": {
      "External id": 269260,
      "cbid": 211, "correlation": 269260
    }
  },
  {
    "ph": "s", "id": 269260, "pid": 76337, "tid": -914061504, "ts": 1716454225634413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634470, "dur": 1,
    "args": {
      "External id": 269271,
      "cbid": 251, "correlation": 269271
    }
  },
  {
    "ph": "f", "id": 269271, "pid": 76337, "tid": -914061504, "ts": 1716454225634470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634474, "dur": 0,
    "args": {
      "External id": 269272,
      "cbid": 251, "correlation": 269272
    }
  },
  {
    "ph": "f", "id": 269272, "pid": 76337, "tid": -914061504, "ts": 1716454225634474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225716747, "dur": 7,
    "args": {
      "External id": 269273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269273,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 269273, "pid": 5, "tid": 7, "ts": 1716454225716747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634475, "dur": 12,
    "args": {
      "External id": 269273,
      "cbid": 211, "correlation": 269273
    }
  },
  {
    "ph": "s", "id": 269273, "pid": 76337, "tid": -914061504, "ts": 1716454225634475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225716756, "dur": 4,
    "args": {
      "External id": 269275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269275,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 269275, "pid": 5, "tid": 7, "ts": 1716454225716756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634488, "dur": 5,
    "args": {
      "External id": 269275,
      "cbid": 211, "correlation": 269275
    }
  },
  {
    "ph": "s", "id": 269275, "pid": 76337, "tid": -914061504, "ts": 1716454225634488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225716760, "dur": 93,
    "args": {
      "External id": 269296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269296,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 269296, "pid": 5, "tid": 7, "ts": 1716454225716760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634563, "dur": 12,
    "args": {
      "External id": 269296,
      "cbid": 211, "correlation": 269296
    }
  },
  {
    "ph": "s", "id": 269296, "pid": 76337, "tid": -914061504, "ts": 1716454225634563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634657, "dur": 1,
    "args": {
      "External id": 269314,
      "cbid": 251, "correlation": 269314
    }
  },
  {
    "ph": "f", "id": 269314, "pid": 76337, "tid": -914061504, "ts": 1716454225634657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225716855, "dur": 98,
    "args": {
      "External id": 269316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269316,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269316, "pid": 5, "tid": 7, "ts": 1716454225716855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634663, "dur": 13,
    "args": {
      "External id": 269316,
      "cbid": 211, "correlation": 269316
    }
  },
  {
    "ph": "s", "id": 269316, "pid": 76337, "tid": -914061504, "ts": 1716454225634663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225716954, "dur": 19,
    "args": {
      "External id": 269324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269324,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269324, "pid": 5, "tid": 7, "ts": 1716454225716954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634732, "dur": 12,
    "args": {
      "External id": 269324,
      "cbid": 211, "correlation": 269324
    }
  },
  {
    "ph": "s", "id": 269324, "pid": 76337, "tid": -914061504, "ts": 1716454225634732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225716974, "dur": 37,
    "args": {
      "External id": 269332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269332,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269332, "pid": 5, "tid": 7, "ts": 1716454225716974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634773, "dur": 9,
    "args": {
      "External id": 269332,
      "cbid": 211, "correlation": 269332
    }
  },
  {
    "ph": "s", "id": 269332, "pid": 76337, "tid": -914061504, "ts": 1716454225634773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225717013, "dur": 34,
    "args": {
      "External id": 269354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269354,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269354, "pid": 5, "tid": 7, "ts": 1716454225717013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634824, "dur": 10,
    "args": {
      "External id": 269354,
      "cbid": 211, "correlation": 269354
    }
  },
  {
    "ph": "s", "id": 269354, "pid": 76337, "tid": -914061504, "ts": 1716454225634824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634912, "dur": 1,
    "args": {
      "External id": 269370,
      "cbid": 251, "correlation": 269370
    }
  },
  {
    "ph": "f", "id": 269370, "pid": 76337, "tid": -914061504, "ts": 1716454225634912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634917, "dur": 0,
    "args": {
      "External id": 269372,
      "cbid": 251, "correlation": 269372
    }
  },
  {
    "ph": "f", "id": 269372, "pid": 76337, "tid": -914061504, "ts": 1716454225634917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225717049, "dur": 549,
    "args": {
      "External id": 269373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269373,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 269373, "pid": 5, "tid": 7, "ts": 1716454225717049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634921, "dur": 13,
    "args": {
      "External id": 269373,
      "cbid": 211, "correlation": 269373
    }
  },
  {
    "ph": "s", "id": 269373, "pid": 76337, "tid": -914061504, "ts": 1716454225634921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225717600, "dur": 126,
    "args": {
      "External id": 269381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269381, "pid": 5, "tid": 7, "ts": 1716454225717600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225634994, "dur": 13,
    "args": {
      "External id": 269381,
      "cbid": 211, "correlation": 269381
    }
  },
  {
    "ph": "s", "id": 269381, "pid": 76337, "tid": -914061504, "ts": 1716454225634994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225717727, "dur": 130,
    "args": {
      "External id": 269389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269389, "pid": 5, "tid": 7, "ts": 1716454225717727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635025, "dur": 8,
    "args": {
      "External id": 269389,
      "cbid": 211, "correlation": 269389
    }
  },
  {
    "ph": "s", "id": 269389, "pid": 76337, "tid": -914061504, "ts": 1716454225635025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635102, "dur": 1,
    "args": {
      "External id": 269405,
      "cbid": 251, "correlation": 269405
    }
  },
  {
    "ph": "f", "id": 269405, "pid": 76337, "tid": -914061504, "ts": 1716454225635102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225717859, "dur": 310,
    "args": {
      "External id": 269407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269407,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269407, "pid": 5, "tid": 7, "ts": 1716454225717859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635107, "dur": 12,
    "args": {
      "External id": 269407,
      "cbid": 211, "correlation": 269407
    }
  },
  {
    "ph": "s", "id": 269407, "pid": 76337, "tid": -914061504, "ts": 1716454225635107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225718170, "dur": 27,
    "args": {
      "External id": 269415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269415,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269415, "pid": 5, "tid": 7, "ts": 1716454225718170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635150, "dur": 9,
    "args": {
      "External id": 269415,
      "cbid": 211, "correlation": 269415
    }
  },
  {
    "ph": "s", "id": 269415, "pid": 76337, "tid": -914061504, "ts": 1716454225635150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225718198, "dur": 82,
    "args": {
      "External id": 269426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269426, "pid": 5, "tid": 7, "ts": 1716454225718198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635217, "dur": 13,
    "args": {
      "External id": 269426,
      "cbid": 211, "correlation": 269426
    }
  },
  {
    "ph": "s", "id": 269426, "pid": 76337, "tid": -914061504, "ts": 1716454225635217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635281, "dur": 0,
    "args": {
      "External id": 269438,
      "cbid": 317, "correlation": 269438
    }
  },
  {
    "ph": "f", "id": 269438, "pid": 76337, "tid": -914061504, "ts": 1716454225635281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635282, "dur": 0,
    "args": {
      "External id": 269439,
      "cbid": 203, "correlation": 269439
    }
  },
  {
    "ph": "f", "id": 269439, "pid": 76337, "tid": -914061504, "ts": 1716454225635282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635283, "dur": 0,
    "args": {
      "External id": 269440,
      "cbid": 205, "correlation": 269440
    }
  },
  {
    "ph": "f", "id": 269440, "pid": 76337, "tid": -914061504, "ts": 1716454225635283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225718282, "dur": 24,
    "args": {
      "External id": 269444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269444, "pid": 5, "tid": 7, "ts": 1716454225718282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635297, "dur": 12,
    "args": {
      "External id": 269444,
      "cbid": 211, "correlation": 269444
    }
  },
  {
    "ph": "s", "id": 269444, "pid": 76337, "tid": -914061504, "ts": 1716454225635297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225718307, "dur": 121,
    "args": {
      "External id": 269446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269446,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269446, "pid": 5, "tid": 7, "ts": 1716454225718307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635316, "dur": 7,
    "args": {
      "External id": 269446,
      "cbid": 211, "correlation": 269446
    }
  },
  {
    "ph": "s", "id": 269446, "pid": 76337, "tid": -914061504, "ts": 1716454225635316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225718430, "dur": 23,
    "args": {
      "External id": 269448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269448,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269448, "pid": 5, "tid": 7, "ts": 1716454225718430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635327, "dur": 5,
    "args": {
      "External id": 269448,
      "cbid": 211, "correlation": 269448
    }
  },
  {
    "ph": "s", "id": 269448, "pid": 76337, "tid": -914061504, "ts": 1716454225635327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225718454, "dur": 33,
    "args": {
      "External id": 269454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269454, "pid": 5, "tid": 7, "ts": 1716454225718454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635354, "dur": 8,
    "args": {
      "External id": 269454,
      "cbid": 211, "correlation": 269454
    }
  },
  {
    "ph": "s", "id": 269454, "pid": 76337, "tid": -914061504, "ts": 1716454225635354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225718489, "dur": 27,
    "args": {
      "External id": 269462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269462,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269462, "pid": 5, "tid": 7, "ts": 1716454225718489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635385, "dur": 8,
    "args": {
      "External id": 269462,
      "cbid": 211, "correlation": 269462
    }
  },
  {
    "ph": "s", "id": 269462, "pid": 76337, "tid": -914061504, "ts": 1716454225635385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225718516, "dur": 102,
    "args": {
      "External id": 269473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269473,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269473, "pid": 5, "tid": 7, "ts": 1716454225718516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635446, "dur": 11,
    "args": {
      "External id": 269473,
      "cbid": 211, "correlation": 269473
    }
  },
  {
    "ph": "s", "id": 269473, "pid": 76337, "tid": -914061504, "ts": 1716454225635446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635500, "dur": 0,
    "args": {
      "External id": 269483,
      "cbid": 317, "correlation": 269483
    }
  },
  {
    "ph": "f", "id": 269483, "pid": 76337, "tid": -914061504, "ts": 1716454225635500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635501, "dur": 0,
    "args": {
      "External id": 269484,
      "cbid": 203, "correlation": 269484
    }
  },
  {
    "ph": "f", "id": 269484, "pid": 76337, "tid": -914061504, "ts": 1716454225635501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635502, "dur": 0,
    "args": {
      "External id": 269485,
      "cbid": 205, "correlation": 269485
    }
  },
  {
    "ph": "f", "id": 269485, "pid": 76337, "tid": -914061504, "ts": 1716454225635502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225718620, "dur": 75,
    "args": {
      "External id": 269489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269489,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269489, "pid": 5, "tid": 7, "ts": 1716454225718620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635516, "dur": 12,
    "args": {
      "External id": 269489,
      "cbid": 211, "correlation": 269489
    }
  },
  {
    "ph": "s", "id": 269489, "pid": 76337, "tid": -914061504, "ts": 1716454225635516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225718697, "dur": 45,
    "args": {
      "External id": 269491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269491,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269491, "pid": 5, "tid": 7, "ts": 1716454225718697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635531, "dur": 5,
    "args": {
      "External id": 269491,
      "cbid": 211, "correlation": 269491
    }
  },
  {
    "ph": "s", "id": 269491, "pid": 76337, "tid": -914061504, "ts": 1716454225635531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225718743, "dur": 4,
    "args": {
      "External id": 269493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269493,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269493, "pid": 5, "tid": 7, "ts": 1716454225718743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635542, "dur": 6,
    "args": {
      "External id": 269493,
      "cbid": 211, "correlation": 269493
    }
  },
  {
    "ph": "s", "id": 269493, "pid": 76337, "tid": -914061504, "ts": 1716454225635542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635551, "dur": 0,
    "args": {
      "External id": 269494,
      "cbid": 51, "correlation": 269494
    }
  },
  {
    "ph": "s", "id": 269494, "pid": 76337, "tid": -914061504, "ts": 1716454225635551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225718748, "dur": 2235,
    "args": {
      "External id": 269495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269495,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269495, "pid": 5, "tid": 7, "ts": 1716454225718748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635552, "dur": 5,
    "args": {
      "External id": 269495,
      "cbid": 211, "correlation": 269495
    }
  },
  {
    "ph": "s", "id": 269495, "pid": 76337, "tid": -914061504, "ts": 1716454225635552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225720984, "dur": 113,
    "args": {
      "External id": 269500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269500, "pid": 5, "tid": 7, "ts": 1716454225720984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635581, "dur": 8,
    "args": {
      "External id": 269500,
      "cbid": 211, "correlation": 269500
    }
  },
  {
    "ph": "s", "id": 269500, "pid": 76337, "tid": -914061504, "ts": 1716454225635581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225721099, "dur": 164,
    "args": {
      "External id": 269509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269509,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269509, "pid": 5, "tid": 7, "ts": 1716454225721099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635672, "dur": 14,
    "args": {
      "External id": 269509,
      "cbid": 211, "correlation": 269509
    }
  },
  {
    "ph": "s", "id": 269509, "pid": 76337, "tid": -914061504, "ts": 1716454225635672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225721264, "dur": 129,
    "args": {
      "External id": 269529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269529,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 269529, "pid": 5, "tid": 7, "ts": 1716454225721264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635741, "dur": 11,
    "args": {
      "External id": 269529,
      "cbid": 211, "correlation": 269529
    }
  },
  {
    "ph": "s", "id": 269529, "pid": 76337, "tid": -914061504, "ts": 1716454225635741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225721394, "dur": 5,
    "args": {
      "External id": 269541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269541,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 269541, "pid": 5, "tid": 7, "ts": 1716454225721394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635763, "dur": 6,
    "args": {
      "External id": 269541,
      "cbid": 211, "correlation": 269541
    }
  },
  {
    "ph": "s", "id": 269541, "pid": 76337, "tid": -914061504, "ts": 1716454225635763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225721400, "dur": 161,
    "args": {
      "External id": 269544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269544,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269544, "pid": 5, "tid": 7, "ts": 1716454225721400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635782, "dur": 7,
    "args": {
      "External id": 269544,
      "cbid": 211, "correlation": 269544
    }
  },
  {
    "ph": "s", "id": 269544, "pid": 76337, "tid": -914061504, "ts": 1716454225635782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225721563, "dur": 102,
    "args": {
      "External id": 269553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269553,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269553, "pid": 5, "tid": 7, "ts": 1716454225721563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635822, "dur": 10,
    "args": {
      "External id": 269553,
      "cbid": 211, "correlation": 269553
    }
  },
  {
    "ph": "s", "id": 269553, "pid": 76337, "tid": -914061504, "ts": 1716454225635822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635874, "dur": 0,
    "args": {
      "External id": 269563,
      "cbid": 317, "correlation": 269563
    }
  },
  {
    "ph": "f", "id": 269563, "pid": 76337, "tid": -914061504, "ts": 1716454225635874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635875, "dur": 0,
    "args": {
      "External id": 269564,
      "cbid": 203, "correlation": 269564
    }
  },
  {
    "ph": "f", "id": 269564, "pid": 76337, "tid": -914061504, "ts": 1716454225635875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635875, "dur": 0,
    "args": {
      "External id": 269565,
      "cbid": 205, "correlation": 269565
    }
  },
  {
    "ph": "f", "id": 269565, "pid": 76337, "tid": -914061504, "ts": 1716454225635875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225721666, "dur": 112,
    "args": {
      "External id": 269569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269569,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269569, "pid": 5, "tid": 7, "ts": 1716454225721666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635892, "dur": 11,
    "args": {
      "External id": 269569,
      "cbid": 211, "correlation": 269569
    }
  },
  {
    "ph": "s", "id": 269569, "pid": 76337, "tid": -914061504, "ts": 1716454225635892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225721779, "dur": 34,
    "args": {
      "External id": 269571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269571,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269571, "pid": 5, "tid": 7, "ts": 1716454225721779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635906, "dur": 5,
    "args": {
      "External id": 269571,
      "cbid": 211, "correlation": 269571
    }
  },
  {
    "ph": "s", "id": 269571, "pid": 76337, "tid": -914061504, "ts": 1716454225635906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225721814, "dur": 3,
    "args": {
      "External id": 269573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269573, "pid": 5, "tid": 7, "ts": 1716454225721814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635915, "dur": 5,
    "args": {
      "External id": 269573,
      "cbid": 211, "correlation": 269573
    }
  },
  {
    "ph": "s", "id": 269573, "pid": 76337, "tid": -914061504, "ts": 1716454225635915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635924, "dur": 0,
    "args": {
      "External id": 269574,
      "cbid": 51, "correlation": 269574
    }
  },
  {
    "ph": "s", "id": 269574, "pid": 76337, "tid": -914061504, "ts": 1716454225635924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225721819, "dur": 2043,
    "args": {
      "External id": 269575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269575,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269575, "pid": 5, "tid": 7, "ts": 1716454225721819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635925, "dur": 7,
    "args": {
      "External id": 269575,
      "cbid": 211, "correlation": 269575
    }
  },
  {
    "ph": "s", "id": 269575, "pid": 76337, "tid": -914061504, "ts": 1716454225635925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225723864, "dur": 60,
    "args": {
      "External id": 269580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269580,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269580, "pid": 5, "tid": 7, "ts": 1716454225723864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225635954, "dur": 9,
    "args": {
      "External id": 269580,
      "cbid": 211, "correlation": 269580
    }
  },
  {
    "ph": "s", "id": 269580, "pid": 76337, "tid": -914061504, "ts": 1716454225635954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225723925, "dur": 3,
    "args": {
      "External id": 269588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269588,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269588, "pid": 5, "tid": 7, "ts": 1716454225723925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636005, "dur": 10,
    "args": {
      "External id": 269588,
      "cbid": 211, "correlation": 269588
    }
  },
  {
    "ph": "s", "id": 269588, "pid": 76337, "tid": -914061504, "ts": 1716454225636005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636070, "dur": 1,
    "args": {
      "External id": 269604,
      "cbid": 251, "correlation": 269604
    }
  },
  {
    "ph": "f", "id": 269604, "pid": 76337, "tid": -914061504, "ts": 1716454225636070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636076, "dur": 0,
    "args": {
      "External id": 269606,
      "cbid": 251, "correlation": 269606
    }
  },
  {
    "ph": "f", "id": 269606, "pid": 76337, "tid": -914061504, "ts": 1716454225636076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225723930, "dur": 12,
    "args": {
      "External id": 269607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269607,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 269607, "pid": 5, "tid": 7, "ts": 1716454225723930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636078, "dur": 12,
    "args": {
      "External id": 269607,
      "cbid": 211, "correlation": 269607
    }
  },
  {
    "ph": "s", "id": 269607, "pid": 76337, "tid": -914061504, "ts": 1716454225636078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225723943, "dur": 5,
    "args": {
      "External id": 269609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269609,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 269609, "pid": 5, "tid": 7, "ts": 1716454225723943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636092, "dur": 5,
    "args": {
      "External id": 269609,
      "cbid": 211, "correlation": 269609
    }
  },
  {
    "ph": "s", "id": 269609, "pid": 76337, "tid": -914061504, "ts": 1716454225636092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225723949, "dur": 55,
    "args": {
      "External id": 269619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269619, "pid": 5, "tid": 7, "ts": 1716454225723949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636149, "dur": 12,
    "args": {
      "External id": 269619,
      "cbid": 211, "correlation": 269619
    }
  },
  {
    "ph": "s", "id": 269619, "pid": 76337, "tid": -914061504, "ts": 1716454225636149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225724005, "dur": 53,
    "args": {
      "External id": 269639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269639,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 269639, "pid": 5, "tid": 7, "ts": 1716454225724005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636216, "dur": 11,
    "args": {
      "External id": 269639,
      "cbid": 211, "correlation": 269639
    }
  },
  {
    "ph": "s", "id": 269639, "pid": 76337, "tid": -914061504, "ts": 1716454225636216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225724059, "dur": 4,
    "args": {
      "External id": 269651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269651,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269651, "pid": 5, "tid": 7, "ts": 1716454225724059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636238, "dur": 6,
    "args": {
      "External id": 269651,
      "cbid": 211, "correlation": 269651
    }
  },
  {
    "ph": "s", "id": 269651, "pid": 76337, "tid": -914061504, "ts": 1716454225636238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225724065, "dur": 56,
    "args": {
      "External id": 269654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269654,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269654, "pid": 5, "tid": 7, "ts": 1716454225724065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636256, "dur": 6,
    "args": {
      "External id": 269654,
      "cbid": 211, "correlation": 269654
    }
  },
  {
    "ph": "s", "id": 269654, "pid": 76337, "tid": -914061504, "ts": 1716454225636256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225724122, "dur": 37,
    "args": {
      "External id": 269663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269663,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269663, "pid": 5, "tid": 7, "ts": 1716454225724122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636296, "dur": 9,
    "args": {
      "External id": 269663,
      "cbid": 211, "correlation": 269663
    }
  },
  {
    "ph": "s", "id": 269663, "pid": 76337, "tid": -914061504, "ts": 1716454225636296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636358, "dur": 0,
    "args": {
      "External id": 269673,
      "cbid": 317, "correlation": 269673
    }
  },
  {
    "ph": "f", "id": 269673, "pid": 76337, "tid": -914061504, "ts": 1716454225636358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636359, "dur": 0,
    "args": {
      "External id": 269674,
      "cbid": 203, "correlation": 269674
    }
  },
  {
    "ph": "f", "id": 269674, "pid": 76337, "tid": -914061504, "ts": 1716454225636359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636360, "dur": 0,
    "args": {
      "External id": 269675,
      "cbid": 205, "correlation": 269675
    }
  },
  {
    "ph": "f", "id": 269675, "pid": 76337, "tid": -914061504, "ts": 1716454225636360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225724161, "dur": 40,
    "args": {
      "External id": 269679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269679,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269679, "pid": 5, "tid": 7, "ts": 1716454225724161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636375, "dur": 12,
    "args": {
      "External id": 269679,
      "cbid": 211, "correlation": 269679
    }
  },
  {
    "ph": "s", "id": 269679, "pid": 76337, "tid": -914061504, "ts": 1716454225636375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225724202, "dur": 14,
    "args": {
      "External id": 269681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269681,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269681, "pid": 5, "tid": 7, "ts": 1716454225724202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636390, "dur": 5,
    "args": {
      "External id": 269681,
      "cbid": 211, "correlation": 269681
    }
  },
  {
    "ph": "s", "id": 269681, "pid": 76337, "tid": -914061504, "ts": 1716454225636390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225724218, "dur": 3,
    "args": {
      "External id": 269683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269683, "pid": 5, "tid": 7, "ts": 1716454225724218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636399, "dur": 5,
    "args": {
      "External id": 269683,
      "cbid": 211, "correlation": 269683
    }
  },
  {
    "ph": "s", "id": 269683, "pid": 76337, "tid": -914061504, "ts": 1716454225636399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636407, "dur": 0,
    "args": {
      "External id": 269684,
      "cbid": 51, "correlation": 269684
    }
  },
  {
    "ph": "s", "id": 269684, "pid": 76337, "tid": -914061504, "ts": 1716454225636407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225724223, "dur": 710,
    "args": {
      "External id": 269685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269685,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269685, "pid": 5, "tid": 7, "ts": 1716454225724223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636408, "dur": 5,
    "args": {
      "External id": 269685,
      "cbid": 211, "correlation": 269685
    }
  },
  {
    "ph": "s", "id": 269685, "pid": 76337, "tid": -914061504, "ts": 1716454225636408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225724933, "dur": 60,
    "args": {
      "External id": 269690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269690,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269690, "pid": 5, "tid": 7, "ts": 1716454225724933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636436, "dur": 8,
    "args": {
      "External id": 269690,
      "cbid": 211, "correlation": 269690
    }
  },
  {
    "ph": "s", "id": 269690, "pid": 76337, "tid": -914061504, "ts": 1716454225636436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636494, "dur": 0,
    "args": {
      "External id": 269700,
      "cbid": 317, "correlation": 269700
    }
  },
  {
    "ph": "f", "id": 269700, "pid": 76337, "tid": -914061504, "ts": 1716454225636494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636494, "dur": 0,
    "args": {
      "External id": 269701,
      "cbid": 203, "correlation": 269701
    }
  },
  {
    "ph": "f", "id": 269701, "pid": 76337, "tid": -914061504, "ts": 1716454225636494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636495, "dur": 0,
    "args": {
      "External id": 269702,
      "cbid": 205, "correlation": 269702
    }
  },
  {
    "ph": "f", "id": 269702, "pid": 76337, "tid": -914061504, "ts": 1716454225636495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225724994, "dur": 4,
    "args": {
      "External id": 269706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269706, "pid": 5, "tid": 7, "ts": 1716454225724994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636511, "dur": 11,
    "args": {
      "External id": 269706,
      "cbid": 211, "correlation": 269706
    }
  },
  {
    "ph": "s", "id": 269706, "pid": 76337, "tid": -914061504, "ts": 1716454225636511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636527, "dur": 0,
    "args": {
      "External id": 269707,
      "cbid": 51, "correlation": 269707
    }
  },
  {
    "ph": "s", "id": 269707, "pid": 76337, "tid": -914061504, "ts": 1716454225636527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225724999, "dur": 268,
    "args": {
      "External id": 269708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269708,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269708, "pid": 5, "tid": 7, "ts": 1716454225724999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636528, "dur": 8,
    "args": {
      "External id": 269708,
      "cbid": 211, "correlation": 269708
    }
  },
  {
    "ph": "s", "id": 269708, "pid": 76337, "tid": -914061504, "ts": 1716454225636528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225725269, "dur": 59,
    "args": {
      "External id": 269713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269713, "pid": 5, "tid": 7, "ts": 1716454225725269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636556, "dur": 8,
    "args": {
      "External id": 269713,
      "cbid": 211, "correlation": 269713
    }
  },
  {
    "ph": "s", "id": 269713, "pid": 76337, "tid": -914061504, "ts": 1716454225636556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225725329, "dur": 50,
    "args": {
      "External id": 269721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269721,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269721, "pid": 5, "tid": 7, "ts": 1716454225725329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636585, "dur": 8,
    "args": {
      "External id": 269721,
      "cbid": 211, "correlation": 269721
    }
  },
  {
    "ph": "s", "id": 269721, "pid": 76337, "tid": -914061504, "ts": 1716454225636585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225725381, "dur": 36,
    "args": {
      "External id": 269729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269729,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269729, "pid": 5, "tid": 7, "ts": 1716454225725381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636613, "dur": 8,
    "args": {
      "External id": 269729,
      "cbid": 211, "correlation": 269729
    }
  },
  {
    "ph": "s", "id": 269729, "pid": 76337, "tid": -914061504, "ts": 1716454225636613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225725418, "dur": 54,
    "args": {
      "External id": 269749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269749,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 269749, "pid": 5, "tid": 7, "ts": 1716454225725418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636695, "dur": 12,
    "args": {
      "External id": 269749,
      "cbid": 211, "correlation": 269749
    }
  },
  {
    "ph": "s", "id": 269749, "pid": 76337, "tid": -914061504, "ts": 1716454225636695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225725473, "dur": 4,
    "args": {
      "External id": 269761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269761,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 269761, "pid": 5, "tid": 7, "ts": 1716454225725473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636717, "dur": 6,
    "args": {
      "External id": 269761,
      "cbid": 211, "correlation": 269761
    }
  },
  {
    "ph": "s", "id": 269761, "pid": 76337, "tid": -914061504, "ts": 1716454225636717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225725478, "dur": 55,
    "args": {
      "External id": 269764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269764,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269764, "pid": 5, "tid": 7, "ts": 1716454225725478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636734, "dur": 7,
    "args": {
      "External id": 269764,
      "cbid": 211, "correlation": 269764
    }
  },
  {
    "ph": "s", "id": 269764, "pid": 76337, "tid": -914061504, "ts": 1716454225636734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636792, "dur": 0,
    "args": {
      "External id": 269775,
      "cbid": 317, "correlation": 269775
    }
  },
  {
    "ph": "f", "id": 269775, "pid": 76337, "tid": -914061504, "ts": 1716454225636792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636793, "dur": 0,
    "args": {
      "External id": 269776,
      "cbid": 203, "correlation": 269776
    }
  },
  {
    "ph": "f", "id": 269776, "pid": 76337, "tid": -914061504, "ts": 1716454225636793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636794, "dur": 0,
    "args": {
      "External id": 269777,
      "cbid": 205, "correlation": 269777
    }
  },
  {
    "ph": "f", "id": 269777, "pid": 76337, "tid": -914061504, "ts": 1716454225636794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636822, "dur": 2,
    "args": {
      "External id": 269781,
      "cbid": 251, "correlation": 269781
    }
  },
  {
    "ph": "f", "id": 269781, "pid": 76337, "tid": -914061504, "ts": 1716454225636822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636825, "dur": 1,
    "args": {
      "External id": 269782,
      "cbid": 251, "correlation": 269782
    }
  },
  {
    "ph": "f", "id": 269782, "pid": 76337, "tid": -914061504, "ts": 1716454225636825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636827, "dur": 0,
    "args": {
      "External id": 269783,
      "cbid": 251, "correlation": 269783
    }
  },
  {
    "ph": "f", "id": 269783, "pid": 76337, "tid": -914061504, "ts": 1716454225636827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636828, "dur": 1,
    "args": {
      "External id": 269784,
      "cbid": 251, "correlation": 269784
    }
  },
  {
    "ph": "f", "id": 269784, "pid": 76337, "tid": -914061504, "ts": 1716454225636828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636831, "dur": 1,
    "args": {
      "External id": 269785,
      "cbid": 251, "correlation": 269785
    }
  },
  {
    "ph": "f", "id": 269785, "pid": 76337, "tid": -914061504, "ts": 1716454225636831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636832, "dur": 1,
    "args": {
      "External id": 269786,
      "cbid": 251, "correlation": 269786
    }
  },
  {
    "ph": "f", "id": 269786, "pid": 76337, "tid": -914061504, "ts": 1716454225636832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636835, "dur": 1,
    "args": {
      "External id": 269787,
      "cbid": 251, "correlation": 269787
    }
  },
  {
    "ph": "f", "id": 269787, "pid": 76337, "tid": -914061504, "ts": 1716454225636835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636836, "dur": 1,
    "args": {
      "External id": 269788,
      "cbid": 251, "correlation": 269788
    }
  },
  {
    "ph": "f", "id": 269788, "pid": 76337, "tid": -914061504, "ts": 1716454225636836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636839, "dur": 0,
    "args": {
      "External id": 269789,
      "cbid": 251, "correlation": 269789
    }
  },
  {
    "ph": "f", "id": 269789, "pid": 76337, "tid": -914061504, "ts": 1716454225636839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225725535, "dur": 116,
    "args": {
      "External id": 269790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269790,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 269790, "pid": 5, "tid": 7, "ts": 1716454225725535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636841, "dur": 12,
    "args": {
      "External id": 269790,
      "cbid": 211, "correlation": 269790
    }
  },
  {
    "ph": "s", "id": 269790, "pid": 76337, "tid": -914061504, "ts": 1716454225636841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225725652, "dur": 60,
    "args": {
      "External id": 269796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269796, "pid": 5, "tid": 7, "ts": 1716454225725652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636876, "dur": 9,
    "args": {
      "External id": 269796,
      "cbid": 211, "correlation": 269796
    }
  },
  {
    "ph": "s", "id": 269796, "pid": 76337, "tid": -914061504, "ts": 1716454225636876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225725713, "dur": 487,
    "args": {
      "External id": 269805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269805,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269805, "pid": 5, "tid": 7, "ts": 1716454225725713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225636958, "dur": 22,
    "args": {
      "External id": 269805,
      "cbid": 211, "correlation": 269805
    }
  },
  {
    "ph": "s", "id": 269805, "pid": 76337, "tid": -914061504, "ts": 1716454225636958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225726202, "dur": 184,
    "args": {
      "External id": 269827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269827,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269827, "pid": 5, "tid": 7, "ts": 1716454225726202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637025, "dur": 11,
    "args": {
      "External id": 269827,
      "cbid": 211, "correlation": 269827
    }
  },
  {
    "ph": "s", "id": 269827, "pid": 76337, "tid": -914061504, "ts": 1716454225637025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637114, "dur": 1,
    "args": {
      "External id": 269838,
      "cbid": 251, "correlation": 269838
    }
  },
  {
    "ph": "f", "id": 269838, "pid": 76337, "tid": -914061504, "ts": 1716454225637114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225726387, "dur": 198,
    "args": {
      "External id": 269839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269839,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269839, "pid": 5, "tid": 7, "ts": 1716454225726387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637119, "dur": 13,
    "args": {
      "External id": 269839,
      "cbid": 211, "correlation": 269839
    }
  },
  {
    "ph": "s", "id": 269839, "pid": 76337, "tid": -914061504, "ts": 1716454225637119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637187, "dur": 1,
    "args": {
      "External id": 269850,
      "cbid": 251, "correlation": 269850
    }
  },
  {
    "ph": "f", "id": 269850, "pid": 76337, "tid": -914061504, "ts": 1716454225637187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225726586, "dur": 190,
    "args": {
      "External id": 269851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269851,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269851, "pid": 5, "tid": 7, "ts": 1716454225726586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637192, "dur": 11,
    "args": {
      "External id": 269851,
      "cbid": 211, "correlation": 269851
    }
  },
  {
    "ph": "s", "id": 269851, "pid": 76337, "tid": -914061504, "ts": 1716454225637192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637254, "dur": 1,
    "args": {
      "External id": 269862,
      "cbid": 251, "correlation": 269862
    }
  },
  {
    "ph": "f", "id": 269862, "pid": 76337, "tid": -914061504, "ts": 1716454225637254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225726778, "dur": 190,
    "args": {
      "External id": 269863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269863,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269863, "pid": 5, "tid": 7, "ts": 1716454225726778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637258, "dur": 12,
    "args": {
      "External id": 269863,
      "cbid": 211, "correlation": 269863
    }
  },
  {
    "ph": "s", "id": 269863, "pid": 76337, "tid": -914061504, "ts": 1716454225637258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225726969, "dur": 18905,
    "args": {
      "External id": 269884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269884,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 269884, "pid": 5, "tid": 7, "ts": 1716454225726969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637342, "dur": 13,
    "args": {
      "External id": 269884,
      "cbid": 211, "correlation": 269884
    }
  },
  {
    "ph": "s", "id": 269884, "pid": 76337, "tid": -914061504, "ts": 1716454225637342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637440, "dur": 1,
    "args": {
      "External id": 269902,
      "cbid": 251, "correlation": 269902
    }
  },
  {
    "ph": "f", "id": 269902, "pid": 76337, "tid": -914061504, "ts": 1716454225637440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225745875, "dur": 207,
    "args": {
      "External id": 269904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269904,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269904, "pid": 5, "tid": 7, "ts": 1716454225745875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637446, "dur": 13,
    "args": {
      "External id": 269904,
      "cbid": 211, "correlation": 269904
    }
  },
  {
    "ph": "s", "id": 269904, "pid": 76337, "tid": -914061504, "ts": 1716454225637446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225746083, "dur": 67,
    "args": {
      "External id": 269912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269912,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269912, "pid": 5, "tid": 7, "ts": 1716454225746083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637514, "dur": 12,
    "args": {
      "External id": 269912,
      "cbid": 211, "correlation": 269912
    }
  },
  {
    "ph": "s", "id": 269912, "pid": 76337, "tid": -914061504, "ts": 1716454225637514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225746152, "dur": 97,
    "args": {
      "External id": 269920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269920,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269920, "pid": 5, "tid": 7, "ts": 1716454225746152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637553, "dur": 8,
    "args": {
      "External id": 269920,
      "cbid": 211, "correlation": 269920
    }
  },
  {
    "ph": "s", "id": 269920, "pid": 76337, "tid": -914061504, "ts": 1716454225637553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225746250, "dur": 53,
    "args": {
      "External id": 269931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269931, "pid": 5, "tid": 7, "ts": 1716454225746250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637622, "dur": 13,
    "args": {
      "External id": 269931,
      "cbid": 211, "correlation": 269931
    }
  },
  {
    "ph": "s", "id": 269931, "pid": 76337, "tid": -914061504, "ts": 1716454225637622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225746304, "dur": 94,
    "args": {
      "External id": 269953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269953,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 269953, "pid": 5, "tid": 7, "ts": 1716454225746304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637654, "dur": 8,
    "args": {
      "External id": 269953,
      "cbid": 211, "correlation": 269953
    }
  },
  {
    "ph": "s", "id": 269953, "pid": 76337, "tid": -914061504, "ts": 1716454225637654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637737, "dur": 1,
    "args": {
      "External id": 269964,
      "cbid": 251, "correlation": 269964
    }
  },
  {
    "ph": "f", "id": 269964, "pid": 76337, "tid": -914061504, "ts": 1716454225637737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225746399, "dur": 107,
    "args": {
      "External id": 269965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269965,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 269965, "pid": 5, "tid": 7, "ts": 1716454225746399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637742, "dur": 14,
    "args": {
      "External id": 269965,
      "cbid": 211, "correlation": 269965
    }
  },
  {
    "ph": "s", "id": 269965, "pid": 76337, "tid": -914061504, "ts": 1716454225637742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637815, "dur": 1,
    "args": {
      "External id": 269976,
      "cbid": 251, "correlation": 269976
    }
  },
  {
    "ph": "f", "id": 269976, "pid": 76337, "tid": -914061504, "ts": 1716454225637815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637819, "dur": 0,
    "args": {
      "External id": 269977,
      "cbid": 251, "correlation": 269977
    }
  },
  {
    "ph": "f", "id": 269977, "pid": 76337, "tid": -914061504, "ts": 1716454225637819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225746508, "dur": 11,
    "args": {
      "External id": 269978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269978,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 269978, "pid": 5, "tid": 7, "ts": 1716454225746508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637821, "dur": 13,
    "args": {
      "External id": 269978,
      "cbid": 211, "correlation": 269978
    }
  },
  {
    "ph": "s", "id": 269978, "pid": 76337, "tid": -914061504, "ts": 1716454225637821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225746520, "dur": 5,
    "args": {
      "External id": 269980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269980,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 269980, "pid": 5, "tid": 7, "ts": 1716454225746520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637836, "dur": 7,
    "args": {
      "External id": 269980,
      "cbid": 211, "correlation": 269980
    }
  },
  {
    "ph": "s", "id": 269980, "pid": 76337, "tid": -914061504, "ts": 1716454225637836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637898, "dur": 1,
    "args": {
      "External id": 269991,
      "cbid": 251, "correlation": 269991
    }
  },
  {
    "ph": "f", "id": 269991, "pid": 76337, "tid": -914061504, "ts": 1716454225637898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637903, "dur": 0,
    "args": {
      "External id": 269992,
      "cbid": 251, "correlation": 269992
    }
  },
  {
    "ph": "f", "id": 269992, "pid": 76337, "tid": -914061504, "ts": 1716454225637903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225746526, "dur": 5,
    "args": {
      "External id": 269993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269993,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 269993, "pid": 5, "tid": 7, "ts": 1716454225746526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637904, "dur": 12,
    "args": {
      "External id": 269993,
      "cbid": 211, "correlation": 269993
    }
  },
  {
    "ph": "s", "id": 269993, "pid": 76337, "tid": -914061504, "ts": 1716454225637904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225746533, "dur": 3,
    "args": {
      "External id": 269995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 269995,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 269995, "pid": 5, "tid": 7, "ts": 1716454225746533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225637917, "dur": 5,
    "args": {
      "External id": 269995,
      "cbid": 211, "correlation": 269995
    }
  },
  {
    "ph": "s", "id": 269995, "pid": 76337, "tid": -914061504, "ts": 1716454225637917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225746537, "dur": 157,
    "args": {
      "External id": 270016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270016,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270016, "pid": 5, "tid": 7, "ts": 1716454225746537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638000, "dur": 13,
    "args": {
      "External id": 270016,
      "cbid": 211, "correlation": 270016
    }
  },
  {
    "ph": "s", "id": 270016, "pid": 76337, "tid": -914061504, "ts": 1716454225638000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638098, "dur": 2,
    "args": {
      "External id": 270034,
      "cbid": 251, "correlation": 270034
    }
  },
  {
    "ph": "f", "id": 270034, "pid": 76337, "tid": -914061504, "ts": 1716454225638098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225746696, "dur": 109,
    "args": {
      "External id": 270036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270036,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270036, "pid": 5, "tid": 7, "ts": 1716454225746696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638105, "dur": 14,
    "args": {
      "External id": 270036,
      "cbid": 211, "correlation": 270036
    }
  },
  {
    "ph": "s", "id": 270036, "pid": 76337, "tid": -914061504, "ts": 1716454225638105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225746806, "dur": 35,
    "args": {
      "External id": 270044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270044,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270044, "pid": 5, "tid": 7, "ts": 1716454225746806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638174, "dur": 12,
    "args": {
      "External id": 270044,
      "cbid": 211, "correlation": 270044
    }
  },
  {
    "ph": "s", "id": 270044, "pid": 76337, "tid": -914061504, "ts": 1716454225638174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225746843, "dur": 68,
    "args": {
      "External id": 270052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270052, "pid": 5, "tid": 7, "ts": 1716454225746843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638215, "dur": 9,
    "args": {
      "External id": 270052,
      "cbid": 211, "correlation": 270052
    }
  },
  {
    "ph": "s", "id": 270052, "pid": 76337, "tid": -914061504, "ts": 1716454225638215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225746912, "dur": 93,
    "args": {
      "External id": 270074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270074,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270074, "pid": 5, "tid": 7, "ts": 1716454225746912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638267, "dur": 10,
    "args": {
      "External id": 270074,
      "cbid": 211, "correlation": 270074
    }
  },
  {
    "ph": "s", "id": 270074, "pid": 76337, "tid": -914061504, "ts": 1716454225638267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638354, "dur": 1,
    "args": {
      "External id": 270090,
      "cbid": 251, "correlation": 270090
    }
  },
  {
    "ph": "f", "id": 270090, "pid": 76337, "tid": -914061504, "ts": 1716454225638354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225747007, "dur": 584,
    "args": {
      "External id": 270092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270092,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270092, "pid": 5, "tid": 7, "ts": 1716454225747007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638360, "dur": 13,
    "args": {
      "External id": 270092,
      "cbid": 211, "correlation": 270092
    }
  },
  {
    "ph": "s", "id": 270092, "pid": 76337, "tid": -914061504, "ts": 1716454225638360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225747592, "dur": 245,
    "args": {
      "External id": 270100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270100, "pid": 5, "tid": 7, "ts": 1716454225747592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638425, "dur": 106,
    "args": {
      "External id": 270100,
      "cbid": 211, "correlation": 270100
    }
  },
  {
    "ph": "s", "id": 270100, "pid": 76337, "tid": -914061504, "ts": 1716454225638425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225747839, "dur": 252,
    "args": {
      "External id": 270108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270108, "pid": 5, "tid": 7, "ts": 1716454225747839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638551, "dur": 342,
    "args": {
      "External id": 270108,
      "cbid": 211, "correlation": 270108
    }
  },
  {
    "ph": "s", "id": 270108, "pid": 76337, "tid": -914061504, "ts": 1716454225638551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638968, "dur": 1,
    "args": {
      "External id": 270124,
      "cbid": 251, "correlation": 270124
    }
  },
  {
    "ph": "f", "id": 270124, "pid": 76337, "tid": -914061504, "ts": 1716454225638968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638980, "dur": 0,
    "args": {
      "External id": 270126,
      "cbid": 251, "correlation": 270126
    }
  },
  {
    "ph": "f", "id": 270126, "pid": 76337, "tid": -914061504, "ts": 1716454225638980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225748092, "dur": 357,
    "args": {
      "External id": 270127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270127,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 270127, "pid": 5, "tid": 7, "ts": 1716454225748092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225638983, "dur": 13,
    "args": {
      "External id": 270127,
      "cbid": 211, "correlation": 270127
    }
  },
  {
    "ph": "s", "id": 270127, "pid": 76337, "tid": -914061504, "ts": 1716454225638983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225748451, "dur": 50,
    "args": {
      "External id": 270135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270135,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270135, "pid": 5, "tid": 7, "ts": 1716454225748451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639027, "dur": 79,
    "args": {
      "External id": 270135,
      "cbid": 211, "correlation": 270135
    }
  },
  {
    "ph": "s", "id": 270135, "pid": 76337, "tid": -914061504, "ts": 1716454225639027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225748502, "dur": 161,
    "args": {
      "External id": 270146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270146, "pid": 5, "tid": 7, "ts": 1716454225748502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639165, "dur": 60,
    "args": {
      "External id": 270146,
      "cbid": 211, "correlation": 270146
    }
  },
  {
    "ph": "s", "id": 270146, "pid": 76337, "tid": -914061504, "ts": 1716454225639165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639277, "dur": 0,
    "args": {
      "External id": 270158,
      "cbid": 317, "correlation": 270158
    }
  },
  {
    "ph": "f", "id": 270158, "pid": 76337, "tid": -914061504, "ts": 1716454225639277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639278, "dur": 0,
    "args": {
      "External id": 270159,
      "cbid": 203, "correlation": 270159
    }
  },
  {
    "ph": "f", "id": 270159, "pid": 76337, "tid": -914061504, "ts": 1716454225639278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639278, "dur": 0,
    "args": {
      "External id": 270160,
      "cbid": 205, "correlation": 270160
    }
  },
  {
    "ph": "f", "id": 270160, "pid": 76337, "tid": -914061504, "ts": 1716454225639278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639303, "dur": 1,
    "args": {
      "External id": 270164,
      "cbid": 251, "correlation": 270164
    }
  },
  {
    "ph": "f", "id": 270164, "pid": 76337, "tid": -914061504, "ts": 1716454225639303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639305, "dur": 0,
    "args": {
      "External id": 270165,
      "cbid": 251, "correlation": 270165
    }
  },
  {
    "ph": "f", "id": 270165, "pid": 76337, "tid": -914061504, "ts": 1716454225639305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639306, "dur": 0,
    "args": {
      "External id": 270166,
      "cbid": 251, "correlation": 270166
    }
  },
  {
    "ph": "f", "id": 270166, "pid": 76337, "tid": -914061504, "ts": 1716454225639306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639306, "dur": 0,
    "args": {
      "External id": 270167,
      "cbid": 251, "correlation": 270167
    }
  },
  {
    "ph": "f", "id": 270167, "pid": 76337, "tid": -914061504, "ts": 1716454225639306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639307, "dur": 0,
    "args": {
      "External id": 270168,
      "cbid": 251, "correlation": 270168
    }
  },
  {
    "ph": "f", "id": 270168, "pid": 76337, "tid": -914061504, "ts": 1716454225639307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639308, "dur": 0,
    "args": {
      "External id": 270169,
      "cbid": 251, "correlation": 270169
    }
  },
  {
    "ph": "f", "id": 270169, "pid": 76337, "tid": -914061504, "ts": 1716454225639308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639309, "dur": 0,
    "args": {
      "External id": 270170,
      "cbid": 251, "correlation": 270170
    }
  },
  {
    "ph": "f", "id": 270170, "pid": 76337, "tid": -914061504, "ts": 1716454225639309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639310, "dur": 0,
    "args": {
      "External id": 270171,
      "cbid": 251, "correlation": 270171
    }
  },
  {
    "ph": "f", "id": 270171, "pid": 76337, "tid": -914061504, "ts": 1716454225639310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639311, "dur": 0,
    "args": {
      "External id": 270172,
      "cbid": 251, "correlation": 270172
    }
  },
  {
    "ph": "f", "id": 270172, "pid": 76337, "tid": -914061504, "ts": 1716454225639311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225748665, "dur": 116,
    "args": {
      "External id": 270173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270173,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270173, "pid": 5, "tid": 7, "ts": 1716454225748665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639313, "dur": 12,
    "args": {
      "External id": 270173,
      "cbid": 211, "correlation": 270173
    }
  },
  {
    "ph": "s", "id": 270173, "pid": 76337, "tid": -914061504, "ts": 1716454225639313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225748782, "dur": 61,
    "args": {
      "External id": 270179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270179,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270179, "pid": 5, "tid": 7, "ts": 1716454225748782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639349, "dur": 9,
    "args": {
      "External id": 270179,
      "cbid": 211, "correlation": 270179
    }
  },
  {
    "ph": "s", "id": 270179, "pid": 76337, "tid": -914061504, "ts": 1716454225639349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225748844, "dur": 50,
    "args": {
      "External id": 270187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270187,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270187, "pid": 5, "tid": 7, "ts": 1716454225748844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639380, "dur": 8,
    "args": {
      "External id": 270187,
      "cbid": 211, "correlation": 270187
    }
  },
  {
    "ph": "s", "id": 270187, "pid": 76337, "tid": -914061504, "ts": 1716454225639380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225748896, "dur": 99,
    "args": {
      "External id": 270196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270196,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270196, "pid": 5, "tid": 7, "ts": 1716454225748896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639419, "dur": 10,
    "args": {
      "External id": 270196,
      "cbid": 211, "correlation": 270196
    }
  },
  {
    "ph": "s", "id": 270196, "pid": 76337, "tid": -914061504, "ts": 1716454225639419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225748996, "dur": 96,
    "args": {
      "External id": 270216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270216,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 270216, "pid": 5, "tid": 7, "ts": 1716454225748996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639489, "dur": 12,
    "args": {
      "External id": 270216,
      "cbid": 211, "correlation": 270216
    }
  },
  {
    "ph": "s", "id": 270216, "pid": 76337, "tid": -914061504, "ts": 1716454225639489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225749093, "dur": 5,
    "args": {
      "External id": 270228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270228,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 270228, "pid": 5, "tid": 7, "ts": 1716454225749093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639512, "dur": 255,
    "args": {
      "External id": 270228,
      "cbid": 211, "correlation": 270228
    }
  },
  {
    "ph": "s", "id": 270228, "pid": 76337, "tid": -914061504, "ts": 1716454225639512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225749099, "dur": 109,
    "args": {
      "External id": 270231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270231, "pid": 5, "tid": 7, "ts": 1716454225749099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639779, "dur": 8,
    "args": {
      "External id": 270231,
      "cbid": 211, "correlation": 270231
    }
  },
  {
    "ph": "s", "id": 270231, "pid": 76337, "tid": -914061504, "ts": 1716454225639779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225749209, "dur": 69,
    "args": {
      "External id": 270240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270240,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270240, "pid": 5, "tid": 7, "ts": 1716454225749209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639819, "dur": 10,
    "args": {
      "External id": 270240,
      "cbid": 211, "correlation": 270240
    }
  },
  {
    "ph": "s", "id": 270240, "pid": 76337, "tid": -914061504, "ts": 1716454225639819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639871, "dur": 0,
    "args": {
      "External id": 270250,
      "cbid": 317, "correlation": 270250
    }
  },
  {
    "ph": "f", "id": 270250, "pid": 76337, "tid": -914061504, "ts": 1716454225639871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639872, "dur": 0,
    "args": {
      "External id": 270251,
      "cbid": 203, "correlation": 270251
    }
  },
  {
    "ph": "f", "id": 270251, "pid": 76337, "tid": -914061504, "ts": 1716454225639872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639873, "dur": 0,
    "args": {
      "External id": 270252,
      "cbid": 205, "correlation": 270252
    }
  },
  {
    "ph": "f", "id": 270252, "pid": 76337, "tid": -914061504, "ts": 1716454225639873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225749280, "dur": 76,
    "args": {
      "External id": 270256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270256,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270256, "pid": 5, "tid": 7, "ts": 1716454225749280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639888, "dur": 12,
    "args": {
      "External id": 270256,
      "cbid": 211, "correlation": 270256
    }
  },
  {
    "ph": "s", "id": 270256, "pid": 76337, "tid": -914061504, "ts": 1716454225639888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225749357, "dur": 25,
    "args": {
      "External id": 270258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270258,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270258, "pid": 5, "tid": 7, "ts": 1716454225749357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639902, "dur": 5,
    "args": {
      "External id": 270258,
      "cbid": 211, "correlation": 270258
    }
  },
  {
    "ph": "s", "id": 270258, "pid": 76337, "tid": -914061504, "ts": 1716454225639902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225749383, "dur": 4,
    "args": {
      "External id": 270260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270260,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270260, "pid": 5, "tid": 7, "ts": 1716454225749383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639912, "dur": 5,
    "args": {
      "External id": 270260,
      "cbid": 211, "correlation": 270260
    }
  },
  {
    "ph": "s", "id": 270260, "pid": 76337, "tid": -914061504, "ts": 1716454225639912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639920, "dur": 0,
    "args": {
      "External id": 270261,
      "cbid": 51, "correlation": 270261
    }
  },
  {
    "ph": "s", "id": 270261, "pid": 76337, "tid": -914061504, "ts": 1716454225639920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225749388, "dur": 1389,
    "args": {
      "External id": 270262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270262,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270262, "pid": 5, "tid": 7, "ts": 1716454225749388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639921, "dur": 5,
    "args": {
      "External id": 270262,
      "cbid": 211, "correlation": 270262
    }
  },
  {
    "ph": "s", "id": 270262, "pid": 76337, "tid": -914061504, "ts": 1716454225639921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225750778, "dur": 61,
    "args": {
      "External id": 270267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270267,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270267, "pid": 5, "tid": 7, "ts": 1716454225750778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225639949, "dur": 8,
    "args": {
      "External id": 270267,
      "cbid": 211, "correlation": 270267
    }
  },
  {
    "ph": "s", "id": 270267, "pid": 76337, "tid": -914061504, "ts": 1716454225639949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225750840, "dur": 4,
    "args": {
      "External id": 270275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270275,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270275, "pid": 5, "tid": 7, "ts": 1716454225750840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640001, "dur": 10,
    "args": {
      "External id": 270275,
      "cbid": 211, "correlation": 270275
    }
  },
  {
    "ph": "s", "id": 270275, "pid": 76337, "tid": -914061504, "ts": 1716454225640001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640067, "dur": 1,
    "args": {
      "External id": 270291,
      "cbid": 251, "correlation": 270291
    }
  },
  {
    "ph": "f", "id": 270291, "pid": 76337, "tid": -914061504, "ts": 1716454225640067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640072, "dur": 0,
    "args": {
      "External id": 270293,
      "cbid": 251, "correlation": 270293
    }
  },
  {
    "ph": "f", "id": 270293, "pid": 76337, "tid": -914061504, "ts": 1716454225640072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225750845, "dur": 11,
    "args": {
      "External id": 270294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270294,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 270294, "pid": 5, "tid": 7, "ts": 1716454225750845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640073, "dur": 12,
    "args": {
      "External id": 270294,
      "cbid": 211, "correlation": 270294
    }
  },
  {
    "ph": "s", "id": 270294, "pid": 76337, "tid": -914061504, "ts": 1716454225640073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225750857, "dur": 5,
    "args": {
      "External id": 270296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270296,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 270296, "pid": 5, "tid": 7, "ts": 1716454225750857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640088, "dur": 176,
    "args": {
      "External id": 270296,
      "cbid": 211, "correlation": 270296
    }
  },
  {
    "ph": "s", "id": 270296, "pid": 76337, "tid": -914061504, "ts": 1716454225640088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225750864, "dur": 54,
    "args": {
      "External id": 270306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270306, "pid": 5, "tid": 7, "ts": 1716454225750864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640319, "dur": 12,
    "args": {
      "External id": 270306,
      "cbid": 211, "correlation": 270306
    }
  },
  {
    "ph": "s", "id": 270306, "pid": 76337, "tid": -914061504, "ts": 1716454225640319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225750920, "dur": 52,
    "args": {
      "External id": 270326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270326,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 270326, "pid": 5, "tid": 7, "ts": 1716454225750920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640384, "dur": 11,
    "args": {
      "External id": 270326,
      "cbid": 211, "correlation": 270326
    }
  },
  {
    "ph": "s", "id": 270326, "pid": 76337, "tid": -914061504, "ts": 1716454225640384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225750973, "dur": 4,
    "args": {
      "External id": 270338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270338,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270338, "pid": 5, "tid": 7, "ts": 1716454225750973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640405, "dur": 7,
    "args": {
      "External id": 270338,
      "cbid": 211, "correlation": 270338
    }
  },
  {
    "ph": "s", "id": 270338, "pid": 76337, "tid": -914061504, "ts": 1716454225640405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225750978, "dur": 56,
    "args": {
      "External id": 270341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270341,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270341, "pid": 5, "tid": 7, "ts": 1716454225750978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640425, "dur": 7,
    "args": {
      "External id": 270341,
      "cbid": 211, "correlation": 270341
    }
  },
  {
    "ph": "s", "id": 270341, "pid": 76337, "tid": -914061504, "ts": 1716454225640425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225751035, "dur": 36,
    "args": {
      "External id": 270350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270350,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270350, "pid": 5, "tid": 7, "ts": 1716454225751035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640466, "dur": 10,
    "args": {
      "External id": 270350,
      "cbid": 211, "correlation": 270350
    }
  },
  {
    "ph": "s", "id": 270350, "pid": 76337, "tid": -914061504, "ts": 1716454225640466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640528, "dur": 0,
    "args": {
      "External id": 270360,
      "cbid": 317, "correlation": 270360
    }
  },
  {
    "ph": "f", "id": 270360, "pid": 76337, "tid": -914061504, "ts": 1716454225640528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640529, "dur": 0,
    "args": {
      "External id": 270361,
      "cbid": 203, "correlation": 270361
    }
  },
  {
    "ph": "f", "id": 270361, "pid": 76337, "tid": -914061504, "ts": 1716454225640529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640529, "dur": 0,
    "args": {
      "External id": 270362,
      "cbid": 205, "correlation": 270362
    }
  },
  {
    "ph": "f", "id": 270362, "pid": 76337, "tid": -914061504, "ts": 1716454225640529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225751073, "dur": 41,
    "args": {
      "External id": 270366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270366,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270366, "pid": 5, "tid": 7, "ts": 1716454225751073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640544, "dur": 12,
    "args": {
      "External id": 270366,
      "cbid": 211, "correlation": 270366
    }
  },
  {
    "ph": "s", "id": 270366, "pid": 76337, "tid": -914061504, "ts": 1716454225640544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225751115, "dur": 15,
    "args": {
      "External id": 270368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270368,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270368, "pid": 5, "tid": 7, "ts": 1716454225751115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640559, "dur": 6,
    "args": {
      "External id": 270368,
      "cbid": 211, "correlation": 270368
    }
  },
  {
    "ph": "s", "id": 270368, "pid": 76337, "tid": -914061504, "ts": 1716454225640559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225751131, "dur": 3,
    "args": {
      "External id": 270370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270370,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270370, "pid": 5, "tid": 7, "ts": 1716454225751131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640568, "dur": 5,
    "args": {
      "External id": 270370,
      "cbid": 211, "correlation": 270370
    }
  },
  {
    "ph": "s", "id": 270370, "pid": 76337, "tid": -914061504, "ts": 1716454225640568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640576, "dur": 0,
    "args": {
      "External id": 270371,
      "cbid": 51, "correlation": 270371
    }
  },
  {
    "ph": "s", "id": 270371, "pid": 76337, "tid": -914061504, "ts": 1716454225640576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225751135, "dur": 707,
    "args": {
      "External id": 270372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270372,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270372, "pid": 5, "tid": 7, "ts": 1716454225751135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640577, "dur": 5,
    "args": {
      "External id": 270372,
      "cbid": 211, "correlation": 270372
    }
  },
  {
    "ph": "s", "id": 270372, "pid": 76337, "tid": -914061504, "ts": 1716454225640577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225751844, "dur": 60,
    "args": {
      "External id": 270377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270377,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270377, "pid": 5, "tid": 7, "ts": 1716454225751844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640604, "dur": 9,
    "args": {
      "External id": 270377,
      "cbid": 211, "correlation": 270377
    }
  },
  {
    "ph": "s", "id": 270377, "pid": 76337, "tid": -914061504, "ts": 1716454225640604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640661, "dur": 0,
    "args": {
      "External id": 270387,
      "cbid": 317, "correlation": 270387
    }
  },
  {
    "ph": "f", "id": 270387, "pid": 76337, "tid": -914061504, "ts": 1716454225640661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640662, "dur": 0,
    "args": {
      "External id": 270388,
      "cbid": 203, "correlation": 270388
    }
  },
  {
    "ph": "f", "id": 270388, "pid": 76337, "tid": -914061504, "ts": 1716454225640662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640663, "dur": 0,
    "args": {
      "External id": 270389,
      "cbid": 205, "correlation": 270389
    }
  },
  {
    "ph": "f", "id": 270389, "pid": 76337, "tid": -914061504, "ts": 1716454225640663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225751905, "dur": 76,
    "args": {
      "External id": 270393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270393,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270393, "pid": 5, "tid": 7, "ts": 1716454225751905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640676, "dur": 11,
    "args": {
      "External id": 270393,
      "cbid": 211, "correlation": 270393
    }
  },
  {
    "ph": "s", "id": 270393, "pid": 76337, "tid": -914061504, "ts": 1716454225640676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225751983, "dur": 208,
    "args": {
      "External id": 270395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270395,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270395, "pid": 5, "tid": 7, "ts": 1716454225751983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640695, "dur": 8,
    "args": {
      "External id": 270395,
      "cbid": 211, "correlation": 270395
    }
  },
  {
    "ph": "s", "id": 270395, "pid": 76337, "tid": -914061504, "ts": 1716454225640695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225752193, "dur": 40,
    "args": {
      "External id": 270397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270397,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270397, "pid": 5, "tid": 7, "ts": 1716454225752193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225640706, "dur": 500,
    "args": {
      "External id": 270397,
      "cbid": 211, "correlation": 270397
    }
  },
  {
    "ph": "s", "id": 270397, "pid": 76337, "tid": -914061504, "ts": 1716454225640706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225752234, "dur": 59,
    "args": {
      "External id": 270403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270403,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270403, "pid": 5, "tid": 7, "ts": 1716454225752234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641228, "dur": 11,
    "args": {
      "External id": 270403,
      "cbid": 211, "correlation": 270403
    }
  },
  {
    "ph": "s", "id": 270403, "pid": 76337, "tid": -914061504, "ts": 1716454225641228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225752295, "dur": 50,
    "args": {
      "External id": 270411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270411,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270411, "pid": 5, "tid": 7, "ts": 1716454225752295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641260, "dur": 8,
    "args": {
      "External id": 270411,
      "cbid": 211, "correlation": 270411
    }
  },
  {
    "ph": "s", "id": 270411, "pid": 76337, "tid": -914061504, "ts": 1716454225641260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225752346, "dur": 36,
    "args": {
      "External id": 270419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270419,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270419, "pid": 5, "tid": 7, "ts": 1716454225752346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641290, "dur": 27,
    "args": {
      "External id": 270419,
      "cbid": 211, "correlation": 270419
    }
  },
  {
    "ph": "s", "id": 270419, "pid": 76337, "tid": -914061504, "ts": 1716454225641290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225752383, "dur": 53,
    "args": {
      "External id": 270439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270439,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 270439, "pid": 5, "tid": 7, "ts": 1716454225752383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641393, "dur": 12,
    "args": {
      "External id": 270439,
      "cbid": 211, "correlation": 270439
    }
  },
  {
    "ph": "s", "id": 270439, "pid": 76337, "tid": -914061504, "ts": 1716454225641393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225752437, "dur": 5,
    "args": {
      "External id": 270451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270451,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270451, "pid": 5, "tid": 7, "ts": 1716454225752437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641415, "dur": 6,
    "args": {
      "External id": 270451,
      "cbid": 211, "correlation": 270451
    }
  },
  {
    "ph": "s", "id": 270451, "pid": 76337, "tid": -914061504, "ts": 1716454225641415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225752443, "dur": 55,
    "args": {
      "External id": 270454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270454, "pid": 5, "tid": 7, "ts": 1716454225752443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641433, "dur": 8,
    "args": {
      "External id": 270454,
      "cbid": 211, "correlation": 270454
    }
  },
  {
    "ph": "s", "id": 270454, "pid": 76337, "tid": -914061504, "ts": 1716454225641433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641490, "dur": 0,
    "args": {
      "External id": 270465,
      "cbid": 317, "correlation": 270465
    }
  },
  {
    "ph": "f", "id": 270465, "pid": 76337, "tid": -914061504, "ts": 1716454225641490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641491, "dur": 0,
    "args": {
      "External id": 270466,
      "cbid": 203, "correlation": 270466
    }
  },
  {
    "ph": "f", "id": 270466, "pid": 76337, "tid": -914061504, "ts": 1716454225641491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641492, "dur": 0,
    "args": {
      "External id": 270467,
      "cbid": 205, "correlation": 270467
    }
  },
  {
    "ph": "f", "id": 270467, "pid": 76337, "tid": -914061504, "ts": 1716454225641492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641515, "dur": 1,
    "args": {
      "External id": 270471,
      "cbid": 251, "correlation": 270471
    }
  },
  {
    "ph": "f", "id": 270471, "pid": 76337, "tid": -914061504, "ts": 1716454225641515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641517, "dur": 0,
    "args": {
      "External id": 270472,
      "cbid": 251, "correlation": 270472
    }
  },
  {
    "ph": "f", "id": 270472, "pid": 76337, "tid": -914061504, "ts": 1716454225641517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641518, "dur": 0,
    "args": {
      "External id": 270473,
      "cbid": 251, "correlation": 270473
    }
  },
  {
    "ph": "f", "id": 270473, "pid": 76337, "tid": -914061504, "ts": 1716454225641518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641519, "dur": 0,
    "args": {
      "External id": 270474,
      "cbid": 251, "correlation": 270474
    }
  },
  {
    "ph": "f", "id": 270474, "pid": 76337, "tid": -914061504, "ts": 1716454225641519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641520, "dur": 0,
    "args": {
      "External id": 270475,
      "cbid": 251, "correlation": 270475
    }
  },
  {
    "ph": "f", "id": 270475, "pid": 76337, "tid": -914061504, "ts": 1716454225641520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641520, "dur": 0,
    "args": {
      "External id": 270476,
      "cbid": 251, "correlation": 270476
    }
  },
  {
    "ph": "f", "id": 270476, "pid": 76337, "tid": -914061504, "ts": 1716454225641520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641521, "dur": 0,
    "args": {
      "External id": 270477,
      "cbid": 251, "correlation": 270477
    }
  },
  {
    "ph": "f", "id": 270477, "pid": 76337, "tid": -914061504, "ts": 1716454225641521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641522, "dur": 0,
    "args": {
      "External id": 270478,
      "cbid": 251, "correlation": 270478
    }
  },
  {
    "ph": "f", "id": 270478, "pid": 76337, "tid": -914061504, "ts": 1716454225641522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641523, "dur": 0,
    "args": {
      "External id": 270479,
      "cbid": 251, "correlation": 270479
    }
  },
  {
    "ph": "f", "id": 270479, "pid": 76337, "tid": -914061504, "ts": 1716454225641523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225752500, "dur": 115,
    "args": {
      "External id": 270480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270480,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270480, "pid": 5, "tid": 7, "ts": 1716454225752500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641525, "dur": 13,
    "args": {
      "External id": 270480,
      "cbid": 211, "correlation": 270480
    }
  },
  {
    "ph": "s", "id": 270480, "pid": 76337, "tid": -914061504, "ts": 1716454225641525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225752616, "dur": 60,
    "args": {
      "External id": 270486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270486, "pid": 5, "tid": 7, "ts": 1716454225752616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641560, "dur": 9,
    "args": {
      "External id": 270486,
      "cbid": 211, "correlation": 270486
    }
  },
  {
    "ph": "s", "id": 270486, "pid": 76337, "tid": -914061504, "ts": 1716454225641560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225752678, "dur": 586,
    "args": {
      "External id": 270495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270495,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270495, "pid": 5, "tid": 7, "ts": 1716454225752678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641643, "dur": 13,
    "args": {
      "External id": 270495,
      "cbid": 211, "correlation": 270495
    }
  },
  {
    "ph": "s", "id": 270495, "pid": 76337, "tid": -914061504, "ts": 1716454225641643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225753265, "dur": 184,
    "args": {
      "External id": 270517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270517,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270517, "pid": 5, "tid": 7, "ts": 1716454225753265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641701, "dur": 10,
    "args": {
      "External id": 270517,
      "cbid": 211, "correlation": 270517
    }
  },
  {
    "ph": "s", "id": 270517, "pid": 76337, "tid": -914061504, "ts": 1716454225641701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641787, "dur": 1,
    "args": {
      "External id": 270528,
      "cbid": 251, "correlation": 270528
    }
  },
  {
    "ph": "f", "id": 270528, "pid": 76337, "tid": -914061504, "ts": 1716454225641787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225753450, "dur": 199,
    "args": {
      "External id": 270529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270529,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270529, "pid": 5, "tid": 7, "ts": 1716454225753450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641792, "dur": 13,
    "args": {
      "External id": 270529,
      "cbid": 211, "correlation": 270529
    }
  },
  {
    "ph": "s", "id": 270529, "pid": 76337, "tid": -914061504, "ts": 1716454225641792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641860, "dur": 1,
    "args": {
      "External id": 270540,
      "cbid": 251, "correlation": 270540
    }
  },
  {
    "ph": "f", "id": 270540, "pid": 76337, "tid": -914061504, "ts": 1716454225641860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225753651, "dur": 191,
    "args": {
      "External id": 270541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270541,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270541, "pid": 5, "tid": 7, "ts": 1716454225753651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641864, "dur": 11,
    "args": {
      "External id": 270541,
      "cbid": 211, "correlation": 270541
    }
  },
  {
    "ph": "s", "id": 270541, "pid": 76337, "tid": -914061504, "ts": 1716454225641864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641927, "dur": 1,
    "args": {
      "External id": 270552,
      "cbid": 251, "correlation": 270552
    }
  },
  {
    "ph": "f", "id": 270552, "pid": 76337, "tid": -914061504, "ts": 1716454225641927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225753843, "dur": 193,
    "args": {
      "External id": 270553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270553,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270553, "pid": 5, "tid": 7, "ts": 1716454225753843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225641931, "dur": 11,
    "args": {
      "External id": 270553,
      "cbid": 211, "correlation": 270553
    }
  },
  {
    "ph": "s", "id": 270553, "pid": 76337, "tid": -914061504, "ts": 1716454225641931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225754038, "dur": 18993,
    "args": {
      "External id": 270574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270574,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270574, "pid": 5, "tid": 7, "ts": 1716454225754038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642019, "dur": 13,
    "args": {
      "External id": 270574,
      "cbid": 211, "correlation": 270574
    }
  },
  {
    "ph": "s", "id": 270574, "pid": 76337, "tid": -914061504, "ts": 1716454225642019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642117, "dur": 1,
    "args": {
      "External id": 270592,
      "cbid": 251, "correlation": 270592
    }
  },
  {
    "ph": "f", "id": 270592, "pid": 76337, "tid": -914061504, "ts": 1716454225642117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225773032, "dur": 206,
    "args": {
      "External id": 270594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270594,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270594, "pid": 5, "tid": 7, "ts": 1716454225773032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642123, "dur": 13,
    "args": {
      "External id": 270594,
      "cbid": 211, "correlation": 270594
    }
  },
  {
    "ph": "s", "id": 270594, "pid": 76337, "tid": -914061504, "ts": 1716454225642123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225773240, "dur": 66,
    "args": {
      "External id": 270602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270602,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270602, "pid": 5, "tid": 7, "ts": 1716454225773240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642192, "dur": 13,
    "args": {
      "External id": 270602,
      "cbid": 211, "correlation": 270602
    }
  },
  {
    "ph": "s", "id": 270602, "pid": 76337, "tid": -914061504, "ts": 1716454225642192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225773307, "dur": 97,
    "args": {
      "External id": 270610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270610,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270610, "pid": 5, "tid": 7, "ts": 1716454225773307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642232, "dur": 9,
    "args": {
      "External id": 270610,
      "cbid": 211, "correlation": 270610
    }
  },
  {
    "ph": "s", "id": 270610, "pid": 76337, "tid": -914061504, "ts": 1716454225642232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225773405, "dur": 54,
    "args": {
      "External id": 270621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270621, "pid": 5, "tid": 7, "ts": 1716454225773405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642303, "dur": 83,
    "args": {
      "External id": 270621,
      "cbid": 211, "correlation": 270621
    }
  },
  {
    "ph": "s", "id": 270621, "pid": 76337, "tid": -914061504, "ts": 1716454225642303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225773461, "dur": 95,
    "args": {
      "External id": 270643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270643,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270643, "pid": 5, "tid": 7, "ts": 1716454225773461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225642406, "dur": 1969,
    "args": {
      "External id": 270643,
      "cbid": 211, "correlation": 270643
    }
  },
  {
    "ph": "s", "id": 270643, "pid": 76337, "tid": -914061504, "ts": 1716454225642406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644454, "dur": 1,
    "args": {
      "External id": 270654,
      "cbid": 251, "correlation": 270654
    }
  },
  {
    "ph": "f", "id": 270654, "pid": 76337, "tid": -914061504, "ts": 1716454225644454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225773557, "dur": 106,
    "args": {
      "External id": 270655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270655,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270655, "pid": 5, "tid": 7, "ts": 1716454225773557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644459, "dur": 67,
    "args": {
      "External id": 270655,
      "cbid": 211, "correlation": 270655
    }
  },
  {
    "ph": "s", "id": 270655, "pid": 76337, "tid": -914061504, "ts": 1716454225644459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644586, "dur": 1,
    "args": {
      "External id": 270666,
      "cbid": 251, "correlation": 270666
    }
  },
  {
    "ph": "f", "id": 270666, "pid": 76337, "tid": -914061504, "ts": 1716454225644586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644589, "dur": 0,
    "args": {
      "External id": 270667,
      "cbid": 251, "correlation": 270667
    }
  },
  {
    "ph": "f", "id": 270667, "pid": 76337, "tid": -914061504, "ts": 1716454225644589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225773664, "dur": 10,
    "args": {
      "External id": 270668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270668,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 270668, "pid": 5, "tid": 7, "ts": 1716454225773664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644591, "dur": 12,
    "args": {
      "External id": 270668,
      "cbid": 211, "correlation": 270668
    }
  },
  {
    "ph": "s", "id": 270668, "pid": 76337, "tid": -914061504, "ts": 1716454225644591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225773675, "dur": 5,
    "args": {
      "External id": 270670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270670,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 270670, "pid": 5, "tid": 7, "ts": 1716454225773675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644605, "dur": 6,
    "args": {
      "External id": 270670,
      "cbid": 211, "correlation": 270670
    }
  },
  {
    "ph": "s", "id": 270670, "pid": 76337, "tid": -914061504, "ts": 1716454225644605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644667, "dur": 1,
    "args": {
      "External id": 270681,
      "cbid": 251, "correlation": 270681
    }
  },
  {
    "ph": "f", "id": 270681, "pid": 76337, "tid": -914061504, "ts": 1716454225644667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644670, "dur": 0,
    "args": {
      "External id": 270682,
      "cbid": 251, "correlation": 270682
    }
  },
  {
    "ph": "f", "id": 270682, "pid": 76337, "tid": -914061504, "ts": 1716454225644670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225773682, "dur": 6,
    "args": {
      "External id": 270683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270683,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 270683, "pid": 5, "tid": 7, "ts": 1716454225773682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644672, "dur": 13,
    "args": {
      "External id": 270683,
      "cbid": 211, "correlation": 270683
    }
  },
  {
    "ph": "s", "id": 270683, "pid": 76337, "tid": -914061504, "ts": 1716454225644672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225773689, "dur": 4,
    "args": {
      "External id": 270685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270685,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 270685, "pid": 5, "tid": 7, "ts": 1716454225773689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644686, "dur": 5,
    "args": {
      "External id": 270685,
      "cbid": 211, "correlation": 270685
    }
  },
  {
    "ph": "s", "id": 270685, "pid": 76337, "tid": -914061504, "ts": 1716454225644686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225773694, "dur": 159,
    "args": {
      "External id": 270706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270706,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270706, "pid": 5, "tid": 7, "ts": 1716454225773694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644760, "dur": 16,
    "args": {
      "External id": 270706,
      "cbid": 211, "correlation": 270706
    }
  },
  {
    "ph": "s", "id": 270706, "pid": 76337, "tid": -914061504, "ts": 1716454225644760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644861, "dur": 1,
    "args": {
      "External id": 270724,
      "cbid": 251, "correlation": 270724
    }
  },
  {
    "ph": "f", "id": 270724, "pid": 76337, "tid": -914061504, "ts": 1716454225644861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225773854, "dur": 108,
    "args": {
      "External id": 270726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270726,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270726, "pid": 5, "tid": 7, "ts": 1716454225773854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644867, "dur": 14,
    "args": {
      "External id": 270726,
      "cbid": 211, "correlation": 270726
    }
  },
  {
    "ph": "s", "id": 270726, "pid": 76337, "tid": -914061504, "ts": 1716454225644867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225773964, "dur": 35,
    "args": {
      "External id": 270734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270734,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270734, "pid": 5, "tid": 7, "ts": 1716454225773964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644937, "dur": 12,
    "args": {
      "External id": 270734,
      "cbid": 211, "correlation": 270734
    }
  },
  {
    "ph": "s", "id": 270734, "pid": 76337, "tid": -914061504, "ts": 1716454225644937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225774000, "dur": 67,
    "args": {
      "External id": 270742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270742, "pid": 5, "tid": 7, "ts": 1716454225774000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225644985, "dur": 10,
    "args": {
      "External id": 270742,
      "cbid": 211, "correlation": 270742
    }
  },
  {
    "ph": "s", "id": 270742, "pid": 76337, "tid": -914061504, "ts": 1716454225644985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225774068, "dur": 95,
    "args": {
      "External id": 270764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270764,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270764, "pid": 5, "tid": 7, "ts": 1716454225774068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645040, "dur": 10,
    "args": {
      "External id": 270764,
      "cbid": 211, "correlation": 270764
    }
  },
  {
    "ph": "s", "id": 270764, "pid": 76337, "tid": -914061504, "ts": 1716454225645040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645126, "dur": 1,
    "args": {
      "External id": 270780,
      "cbid": 251, "correlation": 270780
    }
  },
  {
    "ph": "f", "id": 270780, "pid": 76337, "tid": -914061504, "ts": 1716454225645126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225774164, "dur": 587,
    "args": {
      "External id": 270782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270782,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270782, "pid": 5, "tid": 7, "ts": 1716454225774164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645132, "dur": 14,
    "args": {
      "External id": 270782,
      "cbid": 211, "correlation": 270782
    }
  },
  {
    "ph": "s", "id": 270782, "pid": 76337, "tid": -914061504, "ts": 1716454225645132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225774752, "dur": 248,
    "args": {
      "External id": 270790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270790, "pid": 5, "tid": 7, "ts": 1716454225774752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645198, "dur": 12,
    "args": {
      "External id": 270790,
      "cbid": 211, "correlation": 270790
    }
  },
  {
    "ph": "s", "id": 270790, "pid": 76337, "tid": -914061504, "ts": 1716454225645198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225775001, "dur": 256,
    "args": {
      "External id": 270798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270798,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270798, "pid": 5, "tid": 7, "ts": 1716454225775001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645228, "dur": 9,
    "args": {
      "External id": 270798,
      "cbid": 211, "correlation": 270798
    }
  },
  {
    "ph": "s", "id": 270798, "pid": 76337, "tid": -914061504, "ts": 1716454225645228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645309, "dur": 1,
    "args": {
      "External id": 270814,
      "cbid": 251, "correlation": 270814
    }
  },
  {
    "ph": "f", "id": 270814, "pid": 76337, "tid": -914061504, "ts": 1716454225645309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645314, "dur": 0,
    "args": {
      "External id": 270816,
      "cbid": 251, "correlation": 270816
    }
  },
  {
    "ph": "f", "id": 270816, "pid": 76337, "tid": -914061504, "ts": 1716454225645314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225775259, "dur": 362,
    "args": {
      "External id": 270817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270817,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 270817, "pid": 5, "tid": 7, "ts": 1716454225775259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645317, "dur": 13,
    "args": {
      "External id": 270817,
      "cbid": 211, "correlation": 270817
    }
  },
  {
    "ph": "s", "id": 270817, "pid": 76337, "tid": -914061504, "ts": 1716454225645317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225775623, "dur": 50,
    "args": {
      "External id": 270825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270825,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270825, "pid": 5, "tid": 7, "ts": 1716454225775623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645359, "dur": 10,
    "args": {
      "External id": 270825,
      "cbid": 211, "correlation": 270825
    }
  },
  {
    "ph": "s", "id": 270825, "pid": 76337, "tid": -914061504, "ts": 1716454225645359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225775674, "dur": 161,
    "args": {
      "External id": 270836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270836,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270836, "pid": 5, "tid": 7, "ts": 1716454225775674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645426, "dur": 224,
    "args": {
      "External id": 270836,
      "cbid": 211, "correlation": 270836
    }
  },
  {
    "ph": "s", "id": 270836, "pid": 76337, "tid": -914061504, "ts": 1716454225645426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645702, "dur": 0,
    "args": {
      "External id": 270848,
      "cbid": 317, "correlation": 270848
    }
  },
  {
    "ph": "f", "id": 270848, "pid": 76337, "tid": -914061504, "ts": 1716454225645702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645703, "dur": 0,
    "args": {
      "External id": 270849,
      "cbid": 203, "correlation": 270849
    }
  },
  {
    "ph": "f", "id": 270849, "pid": 76337, "tid": -914061504, "ts": 1716454225645703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645704, "dur": 0,
    "args": {
      "External id": 270850,
      "cbid": 205, "correlation": 270850
    }
  },
  {
    "ph": "f", "id": 270850, "pid": 76337, "tid": -914061504, "ts": 1716454225645704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645728, "dur": 1,
    "args": {
      "External id": 270854,
      "cbid": 251, "correlation": 270854
    }
  },
  {
    "ph": "f", "id": 270854, "pid": 76337, "tid": -914061504, "ts": 1716454225645728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645730, "dur": 0,
    "args": {
      "External id": 270855,
      "cbid": 251, "correlation": 270855
    }
  },
  {
    "ph": "f", "id": 270855, "pid": 76337, "tid": -914061504, "ts": 1716454225645730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645730, "dur": 0,
    "args": {
      "External id": 270856,
      "cbid": 251, "correlation": 270856
    }
  },
  {
    "ph": "f", "id": 270856, "pid": 76337, "tid": -914061504, "ts": 1716454225645730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645731, "dur": 0,
    "args": {
      "External id": 270857,
      "cbid": 251, "correlation": 270857
    }
  },
  {
    "ph": "f", "id": 270857, "pid": 76337, "tid": -914061504, "ts": 1716454225645731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645732, "dur": 0,
    "args": {
      "External id": 270858,
      "cbid": 251, "correlation": 270858
    }
  },
  {
    "ph": "f", "id": 270858, "pid": 76337, "tid": -914061504, "ts": 1716454225645732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645733, "dur": 0,
    "args": {
      "External id": 270859,
      "cbid": 251, "correlation": 270859
    }
  },
  {
    "ph": "f", "id": 270859, "pid": 76337, "tid": -914061504, "ts": 1716454225645733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645734, "dur": 0,
    "args": {
      "External id": 270860,
      "cbid": 251, "correlation": 270860
    }
  },
  {
    "ph": "f", "id": 270860, "pid": 76337, "tid": -914061504, "ts": 1716454225645734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645734, "dur": 0,
    "args": {
      "External id": 270861,
      "cbid": 251, "correlation": 270861
    }
  },
  {
    "ph": "f", "id": 270861, "pid": 76337, "tid": -914061504, "ts": 1716454225645734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645736, "dur": 0,
    "args": {
      "External id": 270862,
      "cbid": 251, "correlation": 270862
    }
  },
  {
    "ph": "f", "id": 270862, "pid": 76337, "tid": -914061504, "ts": 1716454225645736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225775836, "dur": 117,
    "args": {
      "External id": 270863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270863,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 270863, "pid": 5, "tid": 7, "ts": 1716454225775836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645738, "dur": 41,
    "args": {
      "External id": 270863,
      "cbid": 211, "correlation": 270863
    }
  },
  {
    "ph": "s", "id": 270863, "pid": 76337, "tid": -914061504, "ts": 1716454225645738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225775955, "dur": 61,
    "args": {
      "External id": 270869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270869, "pid": 5, "tid": 7, "ts": 1716454225775955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645801, "dur": 106,
    "args": {
      "External id": 270869,
      "cbid": 211, "correlation": 270869
    }
  },
  {
    "ph": "s", "id": 270869, "pid": 76337, "tid": -914061504, "ts": 1716454225645801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225776017, "dur": 50,
    "args": {
      "External id": 270877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270877,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270877, "pid": 5, "tid": 7, "ts": 1716454225776017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225645930, "dur": 291,
    "args": {
      "External id": 270877,
      "cbid": 211, "correlation": 270877
    }
  },
  {
    "ph": "s", "id": 270877, "pid": 76337, "tid": -914061504, "ts": 1716454225645930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225776068, "dur": 100,
    "args": {
      "External id": 270886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270886,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270886, "pid": 5, "tid": 7, "ts": 1716454225776068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646252, "dur": 12,
    "args": {
      "External id": 270886,
      "cbid": 211, "correlation": 270886
    }
  },
  {
    "ph": "s", "id": 270886, "pid": 76337, "tid": -914061504, "ts": 1716454225646252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225776170, "dur": 93,
    "args": {
      "External id": 270906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270906,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 270906, "pid": 5, "tid": 7, "ts": 1716454225776170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646325, "dur": 11,
    "args": {
      "External id": 270906,
      "cbid": 211, "correlation": 270906
    }
  },
  {
    "ph": "s", "id": 270906, "pid": 76337, "tid": -914061504, "ts": 1716454225646325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225776264, "dur": 4,
    "args": {
      "External id": 270918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270918,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 270918, "pid": 5, "tid": 7, "ts": 1716454225776264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646347, "dur": 10,
    "args": {
      "External id": 270918,
      "cbid": 211, "correlation": 270918
    }
  },
  {
    "ph": "s", "id": 270918, "pid": 76337, "tid": -914061504, "ts": 1716454225646347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225776270, "dur": 111,
    "args": {
      "External id": 270921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270921, "pid": 5, "tid": 7, "ts": 1716454225776270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646370, "dur": 111,
    "args": {
      "External id": 270921,
      "cbid": 211, "correlation": 270921
    }
  },
  {
    "ph": "s", "id": 270921, "pid": 76337, "tid": -914061504, "ts": 1716454225646370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225776382, "dur": 69,
    "args": {
      "External id": 270930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270930,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270930, "pid": 5, "tid": 7, "ts": 1716454225776382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646513, "dur": 10,
    "args": {
      "External id": 270930,
      "cbid": 211, "correlation": 270930
    }
  },
  {
    "ph": "s", "id": 270930, "pid": 76337, "tid": -914061504, "ts": 1716454225646513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646565, "dur": 0,
    "args": {
      "External id": 270940,
      "cbid": 317, "correlation": 270940
    }
  },
  {
    "ph": "f", "id": 270940, "pid": 76337, "tid": -914061504, "ts": 1716454225646565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646566, "dur": 0,
    "args": {
      "External id": 270941,
      "cbid": 203, "correlation": 270941
    }
  },
  {
    "ph": "f", "id": 270941, "pid": 76337, "tid": -914061504, "ts": 1716454225646566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646566, "dur": 0,
    "args": {
      "External id": 270942,
      "cbid": 205, "correlation": 270942
    }
  },
  {
    "ph": "f", "id": 270942, "pid": 76337, "tid": -914061504, "ts": 1716454225646566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225776451, "dur": 76,
    "args": {
      "External id": 270946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270946, "pid": 5, "tid": 7, "ts": 1716454225776451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646581, "dur": 12,
    "args": {
      "External id": 270946,
      "cbid": 211, "correlation": 270946
    }
  },
  {
    "ph": "s", "id": 270946, "pid": 76337, "tid": -914061504, "ts": 1716454225646581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225776529, "dur": 25,
    "args": {
      "External id": 270948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270948, "pid": 5, "tid": 7, "ts": 1716454225776529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646595, "dur": 5,
    "args": {
      "External id": 270948,
      "cbid": 211, "correlation": 270948
    }
  },
  {
    "ph": "s", "id": 270948, "pid": 76337, "tid": -914061504, "ts": 1716454225646595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225776555, "dur": 4,
    "args": {
      "External id": 270950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270950, "pid": 5, "tid": 7, "ts": 1716454225776555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646606, "dur": 5,
    "args": {
      "External id": 270950,
      "cbid": 211, "correlation": 270950
    }
  },
  {
    "ph": "s", "id": 270950, "pid": 76337, "tid": -914061504, "ts": 1716454225646606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646614, "dur": 0,
    "args": {
      "External id": 270951,
      "cbid": 51, "correlation": 270951
    }
  },
  {
    "ph": "s", "id": 270951, "pid": 76337, "tid": -914061504, "ts": 1716454225646614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225776560, "dur": 1398,
    "args": {
      "External id": 270952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270952,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 270952, "pid": 5, "tid": 7, "ts": 1716454225776560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646615, "dur": 5,
    "args": {
      "External id": 270952,
      "cbid": 211, "correlation": 270952
    }
  },
  {
    "ph": "s", "id": 270952, "pid": 76337, "tid": -914061504, "ts": 1716454225646615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225777959, "dur": 60,
    "args": {
      "External id": 270957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270957, "pid": 5, "tid": 7, "ts": 1716454225777959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646643, "dur": 9,
    "args": {
      "External id": 270957,
      "cbid": 211, "correlation": 270957
    }
  },
  {
    "ph": "s", "id": 270957, "pid": 76337, "tid": -914061504, "ts": 1716454225646643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225778021, "dur": 4,
    "args": {
      "External id": 270965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270965,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 270965, "pid": 5, "tid": 7, "ts": 1716454225778021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646686, "dur": 10,
    "args": {
      "External id": 270965,
      "cbid": 211, "correlation": 270965
    }
  },
  {
    "ph": "s", "id": 270965, "pid": 76337, "tid": -914061504, "ts": 1716454225646686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646751, "dur": 1,
    "args": {
      "External id": 270981,
      "cbid": 251, "correlation": 270981
    }
  },
  {
    "ph": "f", "id": 270981, "pid": 76337, "tid": -914061504, "ts": 1716454225646751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646757, "dur": 0,
    "args": {
      "External id": 270983,
      "cbid": 251, "correlation": 270983
    }
  },
  {
    "ph": "f", "id": 270983, "pid": 76337, "tid": -914061504, "ts": 1716454225646757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225778026, "dur": 11,
    "args": {
      "External id": 270984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270984,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 270984, "pid": 5, "tid": 7, "ts": 1716454225778026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646758, "dur": 11,
    "args": {
      "External id": 270984,
      "cbid": 211, "correlation": 270984
    }
  },
  {
    "ph": "s", "id": 270984, "pid": 76337, "tid": -914061504, "ts": 1716454225646758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225778039, "dur": 5,
    "args": {
      "External id": 270986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270986,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 270986, "pid": 5, "tid": 7, "ts": 1716454225778039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646771, "dur": 7,
    "args": {
      "External id": 270986,
      "cbid": 211, "correlation": 270986
    }
  },
  {
    "ph": "s", "id": 270986, "pid": 76337, "tid": -914061504, "ts": 1716454225646771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225778045, "dur": 56,
    "args": {
      "External id": 270996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 270996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 270996, "pid": 5, "tid": 7, "ts": 1716454225778045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225646830, "dur": 565,
    "args": {
      "External id": 270996,
      "cbid": 211, "correlation": 270996
    }
  },
  {
    "ph": "s", "id": 270996, "pid": 76337, "tid": -914061504, "ts": 1716454225646830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225778102, "dur": 54,
    "args": {
      "External id": 271016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271016,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 271016, "pid": 5, "tid": 7, "ts": 1716454225778102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647449, "dur": 12,
    "args": {
      "External id": 271016,
      "cbid": 211, "correlation": 271016
    }
  },
  {
    "ph": "s", "id": 271016, "pid": 76337, "tid": -914061504, "ts": 1716454225647449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225778158, "dur": 4,
    "args": {
      "External id": 271028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271028,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 271028, "pid": 5, "tid": 7, "ts": 1716454225778158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647471, "dur": 6,
    "args": {
      "External id": 271028,
      "cbid": 211, "correlation": 271028
    }
  },
  {
    "ph": "s", "id": 271028, "pid": 76337, "tid": -914061504, "ts": 1716454225647471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225778163, "dur": 56,
    "args": {
      "External id": 271031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271031, "pid": 5, "tid": 7, "ts": 1716454225778163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647490, "dur": 6,
    "args": {
      "External id": 271031,
      "cbid": 211, "correlation": 271031
    }
  },
  {
    "ph": "s", "id": 271031, "pid": 76337, "tid": -914061504, "ts": 1716454225647490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225778220, "dur": 36,
    "args": {
      "External id": 271040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271040,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271040, "pid": 5, "tid": 7, "ts": 1716454225778220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647530, "dur": 10,
    "args": {
      "External id": 271040,
      "cbid": 211, "correlation": 271040
    }
  },
  {
    "ph": "s", "id": 271040, "pid": 76337, "tid": -914061504, "ts": 1716454225647530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647593, "dur": 0,
    "args": {
      "External id": 271050,
      "cbid": 317, "correlation": 271050
    }
  },
  {
    "ph": "f", "id": 271050, "pid": 76337, "tid": -914061504, "ts": 1716454225647593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647594, "dur": 0,
    "args": {
      "External id": 271051,
      "cbid": 203, "correlation": 271051
    }
  },
  {
    "ph": "f", "id": 271051, "pid": 76337, "tid": -914061504, "ts": 1716454225647594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647595, "dur": 0,
    "args": {
      "External id": 271052,
      "cbid": 205, "correlation": 271052
    }
  },
  {
    "ph": "f", "id": 271052, "pid": 76337, "tid": -914061504, "ts": 1716454225647595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225778258, "dur": 40,
    "args": {
      "External id": 271056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271056, "pid": 5, "tid": 7, "ts": 1716454225778258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647610, "dur": 12,
    "args": {
      "External id": 271056,
      "cbid": 211, "correlation": 271056
    }
  },
  {
    "ph": "s", "id": 271056, "pid": 76337, "tid": -914061504, "ts": 1716454225647610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225778299, "dur": 14,
    "args": {
      "External id": 271058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271058,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271058, "pid": 5, "tid": 7, "ts": 1716454225778299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647624, "dur": 5,
    "args": {
      "External id": 271058,
      "cbid": 211, "correlation": 271058
    }
  },
  {
    "ph": "s", "id": 271058, "pid": 76337, "tid": -914061504, "ts": 1716454225647624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225778315, "dur": 3,
    "args": {
      "External id": 271060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 271060, "pid": 5, "tid": 7, "ts": 1716454225778315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647633, "dur": 6,
    "args": {
      "External id": 271060,
      "cbid": 211, "correlation": 271060
    }
  },
  {
    "ph": "s", "id": 271060, "pid": 76337, "tid": -914061504, "ts": 1716454225647633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647641, "dur": 0,
    "args": {
      "External id": 271061,
      "cbid": 51, "correlation": 271061
    }
  },
  {
    "ph": "s", "id": 271061, "pid": 76337, "tid": -914061504, "ts": 1716454225647641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225778320, "dur": 715,
    "args": {
      "External id": 271062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271062,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271062, "pid": 5, "tid": 7, "ts": 1716454225778320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647642, "dur": 5,
    "args": {
      "External id": 271062,
      "cbid": 211, "correlation": 271062
    }
  },
  {
    "ph": "s", "id": 271062, "pid": 76337, "tid": -914061504, "ts": 1716454225647642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225779036, "dur": 60,
    "args": {
      "External id": 271067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271067, "pid": 5, "tid": 7, "ts": 1716454225779036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647670, "dur": 9,
    "args": {
      "External id": 271067,
      "cbid": 211, "correlation": 271067
    }
  },
  {
    "ph": "s", "id": 271067, "pid": 76337, "tid": -914061504, "ts": 1716454225647670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647727, "dur": 0,
    "args": {
      "External id": 271077,
      "cbid": 317, "correlation": 271077
    }
  },
  {
    "ph": "f", "id": 271077, "pid": 76337, "tid": -914061504, "ts": 1716454225647727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647728, "dur": 0,
    "args": {
      "External id": 271078,
      "cbid": 203, "correlation": 271078
    }
  },
  {
    "ph": "f", "id": 271078, "pid": 76337, "tid": -914061504, "ts": 1716454225647728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647729, "dur": 0,
    "args": {
      "External id": 271079,
      "cbid": 205, "correlation": 271079
    }
  },
  {
    "ph": "f", "id": 271079, "pid": 76337, "tid": -914061504, "ts": 1716454225647729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225779098, "dur": 75,
    "args": {
      "External id": 271083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271083,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271083, "pid": 5, "tid": 7, "ts": 1716454225779098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647740, "dur": 11,
    "args": {
      "External id": 271083,
      "cbid": 211, "correlation": 271083
    }
  },
  {
    "ph": "s", "id": 271083, "pid": 76337, "tid": -914061504, "ts": 1716454225647740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225779174, "dur": 212,
    "args": {
      "External id": 271085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271085,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271085, "pid": 5, "tid": 7, "ts": 1716454225779174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647758, "dur": 6,
    "args": {
      "External id": 271085,
      "cbid": 211, "correlation": 271085
    }
  },
  {
    "ph": "s", "id": 271085, "pid": 76337, "tid": -914061504, "ts": 1716454225647758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225779387, "dur": 38,
    "args": {
      "External id": 271087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271087,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271087, "pid": 5, "tid": 7, "ts": 1716454225779387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647769, "dur": 5,
    "args": {
      "External id": 271087,
      "cbid": 211, "correlation": 271087
    }
  },
  {
    "ph": "s", "id": 271087, "pid": 76337, "tid": -914061504, "ts": 1716454225647769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225779427, "dur": 60,
    "args": {
      "External id": 271093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271093, "pid": 5, "tid": 7, "ts": 1716454225779427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225647794, "dur": 536,
    "args": {
      "External id": 271093,
      "cbid": 211, "correlation": 271093
    }
  },
  {
    "ph": "s", "id": 271093, "pid": 76337, "tid": -914061504, "ts": 1716454225647794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225779488, "dur": 50,
    "args": {
      "External id": 271101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271101,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271101, "pid": 5, "tid": 7, "ts": 1716454225779488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648352, "dur": 9,
    "args": {
      "External id": 271101,
      "cbid": 211, "correlation": 271101
    }
  },
  {
    "ph": "s", "id": 271101, "pid": 76337, "tid": -914061504, "ts": 1716454225648352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225779540, "dur": 36,
    "args": {
      "External id": 271109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271109,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271109, "pid": 5, "tid": 7, "ts": 1716454225779540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648384, "dur": 8,
    "args": {
      "External id": 271109,
      "cbid": 211, "correlation": 271109
    }
  },
  {
    "ph": "s", "id": 271109, "pid": 76337, "tid": -914061504, "ts": 1716454225648384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225779577, "dur": 53,
    "args": {
      "External id": 271129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271129,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 271129, "pid": 5, "tid": 7, "ts": 1716454225779577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648465, "dur": 13,
    "args": {
      "External id": 271129,
      "cbid": 211, "correlation": 271129
    }
  },
  {
    "ph": "s", "id": 271129, "pid": 76337, "tid": -914061504, "ts": 1716454225648465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225779632, "dur": 4,
    "args": {
      "External id": 271141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271141,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 271141, "pid": 5, "tid": 7, "ts": 1716454225779632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648488, "dur": 7,
    "args": {
      "External id": 271141,
      "cbid": 211, "correlation": 271141
    }
  },
  {
    "ph": "s", "id": 271141, "pid": 76337, "tid": -914061504, "ts": 1716454225648488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225779637, "dur": 57,
    "args": {
      "External id": 271144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271144, "pid": 5, "tid": 7, "ts": 1716454225779637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648506, "dur": 6,
    "args": {
      "External id": 271144,
      "cbid": 211, "correlation": 271144
    }
  },
  {
    "ph": "s", "id": 271144, "pid": 76337, "tid": -914061504, "ts": 1716454225648506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648562, "dur": 0,
    "args": {
      "External id": 271155,
      "cbid": 317, "correlation": 271155
    }
  },
  {
    "ph": "f", "id": 271155, "pid": 76337, "tid": -914061504, "ts": 1716454225648562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648563, "dur": 0,
    "args": {
      "External id": 271156,
      "cbid": 203, "correlation": 271156
    }
  },
  {
    "ph": "f", "id": 271156, "pid": 76337, "tid": -914061504, "ts": 1716454225648563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648564, "dur": 0,
    "args": {
      "External id": 271157,
      "cbid": 205, "correlation": 271157
    }
  },
  {
    "ph": "f", "id": 271157, "pid": 76337, "tid": -914061504, "ts": 1716454225648564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648586, "dur": 1,
    "args": {
      "External id": 271161,
      "cbid": 251, "correlation": 271161
    }
  },
  {
    "ph": "f", "id": 271161, "pid": 76337, "tid": -914061504, "ts": 1716454225648586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648588, "dur": 0,
    "args": {
      "External id": 271162,
      "cbid": 251, "correlation": 271162
    }
  },
  {
    "ph": "f", "id": 271162, "pid": 76337, "tid": -914061504, "ts": 1716454225648588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648589, "dur": 0,
    "args": {
      "External id": 271163,
      "cbid": 251, "correlation": 271163
    }
  },
  {
    "ph": "f", "id": 271163, "pid": 76337, "tid": -914061504, "ts": 1716454225648589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648589, "dur": 0,
    "args": {
      "External id": 271164,
      "cbid": 251, "correlation": 271164
    }
  },
  {
    "ph": "f", "id": 271164, "pid": 76337, "tid": -914061504, "ts": 1716454225648589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648590, "dur": 0,
    "args": {
      "External id": 271165,
      "cbid": 251, "correlation": 271165
    }
  },
  {
    "ph": "f", "id": 271165, "pid": 76337, "tid": -914061504, "ts": 1716454225648590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648591, "dur": 0,
    "args": {
      "External id": 271166,
      "cbid": 251, "correlation": 271166
    }
  },
  {
    "ph": "f", "id": 271166, "pid": 76337, "tid": -914061504, "ts": 1716454225648591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648592, "dur": 0,
    "args": {
      "External id": 271167,
      "cbid": 251, "correlation": 271167
    }
  },
  {
    "ph": "f", "id": 271167, "pid": 76337, "tid": -914061504, "ts": 1716454225648592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648592, "dur": 0,
    "args": {
      "External id": 271168,
      "cbid": 251, "correlation": 271168
    }
  },
  {
    "ph": "f", "id": 271168, "pid": 76337, "tid": -914061504, "ts": 1716454225648592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648594, "dur": 0,
    "args": {
      "External id": 271169,
      "cbid": 251, "correlation": 271169
    }
  },
  {
    "ph": "f", "id": 271169, "pid": 76337, "tid": -914061504, "ts": 1716454225648594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225779695, "dur": 117,
    "args": {
      "External id": 271170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271170,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 271170, "pid": 5, "tid": 7, "ts": 1716454225779695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648596, "dur": 12,
    "args": {
      "External id": 271170,
      "cbid": 211, "correlation": 271170
    }
  },
  {
    "ph": "s", "id": 271170, "pid": 76337, "tid": -914061504, "ts": 1716454225648596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225779814, "dur": 60,
    "args": {
      "External id": 271176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271176, "pid": 5, "tid": 7, "ts": 1716454225779814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648632, "dur": 9,
    "args": {
      "External id": 271176,
      "cbid": 211, "correlation": 271176
    }
  },
  {
    "ph": "s", "id": 271176, "pid": 76337, "tid": -914061504, "ts": 1716454225648632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225779875, "dur": 584,
    "args": {
      "External id": 271185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271185,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271185, "pid": 5, "tid": 7, "ts": 1716454225779875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648713, "dur": 14,
    "args": {
      "External id": 271185,
      "cbid": 211, "correlation": 271185
    }
  },
  {
    "ph": "s", "id": 271185, "pid": 76337, "tid": -914061504, "ts": 1716454225648713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225780460, "dur": 187,
    "args": {
      "External id": 271207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271207,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271207, "pid": 5, "tid": 7, "ts": 1716454225780460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648770, "dur": 11,
    "args": {
      "External id": 271207,
      "cbid": 211, "correlation": 271207
    }
  },
  {
    "ph": "s", "id": 271207, "pid": 76337, "tid": -914061504, "ts": 1716454225648770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648857, "dur": 1,
    "args": {
      "External id": 271218,
      "cbid": 251, "correlation": 271218
    }
  },
  {
    "ph": "f", "id": 271218, "pid": 76337, "tid": -914061504, "ts": 1716454225648857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225780648, "dur": 200,
    "args": {
      "External id": 271219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271219,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271219, "pid": 5, "tid": 7, "ts": 1716454225780648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648862, "dur": 12,
    "args": {
      "External id": 271219,
      "cbid": 211, "correlation": 271219
    }
  },
  {
    "ph": "s", "id": 271219, "pid": 76337, "tid": -914061504, "ts": 1716454225648862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648929, "dur": 1,
    "args": {
      "External id": 271230,
      "cbid": 251, "correlation": 271230
    }
  },
  {
    "ph": "f", "id": 271230, "pid": 76337, "tid": -914061504, "ts": 1716454225648929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225780849, "dur": 198,
    "args": {
      "External id": 271231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271231,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271231, "pid": 5, "tid": 7, "ts": 1716454225780849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225648933, "dur": 12,
    "args": {
      "External id": 271231,
      "cbid": 211, "correlation": 271231
    }
  },
  {
    "ph": "s", "id": 271231, "pid": 76337, "tid": -914061504, "ts": 1716454225648933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649005, "dur": 1,
    "args": {
      "External id": 271242,
      "cbid": 251, "correlation": 271242
    }
  },
  {
    "ph": "f", "id": 271242, "pid": 76337, "tid": -914061504, "ts": 1716454225649005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225781049, "dur": 192,
    "args": {
      "External id": 271243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271243,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271243, "pid": 5, "tid": 7, "ts": 1716454225781049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649009, "dur": 12,
    "args": {
      "External id": 271243,
      "cbid": 211, "correlation": 271243
    }
  },
  {
    "ph": "s", "id": 271243, "pid": 76337, "tid": -914061504, "ts": 1716454225649009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225781243, "dur": 19185,
    "args": {
      "External id": 271264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271264,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 271264, "pid": 5, "tid": 7, "ts": 1716454225781243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649091, "dur": 12,
    "args": {
      "External id": 271264,
      "cbid": 211, "correlation": 271264
    }
  },
  {
    "ph": "s", "id": 271264, "pid": 76337, "tid": -914061504, "ts": 1716454225649091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649187, "dur": 1,
    "args": {
      "External id": 271282,
      "cbid": 251, "correlation": 271282
    }
  },
  {
    "ph": "f", "id": 271282, "pid": 76337, "tid": -914061504, "ts": 1716454225649187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225800429, "dur": 207,
    "args": {
      "External id": 271284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271284,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271284, "pid": 5, "tid": 7, "ts": 1716454225800429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649193, "dur": 13,
    "args": {
      "External id": 271284,
      "cbid": 211, "correlation": 271284
    }
  },
  {
    "ph": "s", "id": 271284, "pid": 76337, "tid": -914061504, "ts": 1716454225649193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225800638, "dur": 67,
    "args": {
      "External id": 271292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271292,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271292, "pid": 5, "tid": 7, "ts": 1716454225800638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649261, "dur": 12,
    "args": {
      "External id": 271292,
      "cbid": 211, "correlation": 271292
    }
  },
  {
    "ph": "s", "id": 271292, "pid": 76337, "tid": -914061504, "ts": 1716454225649261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225800706, "dur": 96,
    "args": {
      "External id": 271300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271300,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271300, "pid": 5, "tid": 7, "ts": 1716454225800706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649299, "dur": 116,
    "args": {
      "External id": 271300,
      "cbid": 211, "correlation": 271300
    }
  },
  {
    "ph": "s", "id": 271300, "pid": 76337, "tid": -914061504, "ts": 1716454225649299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225800803, "dur": 55,
    "args": {
      "External id": 271311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271311, "pid": 5, "tid": 7, "ts": 1716454225800803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225649479, "dur": 1928,
    "args": {
      "External id": 271311,
      "cbid": 211, "correlation": 271311
    }
  },
  {
    "ph": "s", "id": 271311, "pid": 76337, "tid": -914061504, "ts": 1716454225649479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225800860, "dur": 95,
    "args": {
      "External id": 271333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271333,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271333, "pid": 5, "tid": 7, "ts": 1716454225800860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651431, "dur": 123,
    "args": {
      "External id": 271333,
      "cbid": 211, "correlation": 271333
    }
  },
  {
    "ph": "s", "id": 271333, "pid": 76337, "tid": -914061504, "ts": 1716454225651431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651652, "dur": 1,
    "args": {
      "External id": 271344,
      "cbid": 251, "correlation": 271344
    }
  },
  {
    "ph": "f", "id": 271344, "pid": 76337, "tid": -914061504, "ts": 1716454225651652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225800956, "dur": 105,
    "args": {
      "External id": 271345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271345,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271345, "pid": 5, "tid": 7, "ts": 1716454225800956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651658, "dur": 13,
    "args": {
      "External id": 271345,
      "cbid": 211, "correlation": 271345
    }
  },
  {
    "ph": "s", "id": 271345, "pid": 76337, "tid": -914061504, "ts": 1716454225651658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651737, "dur": 1,
    "args": {
      "External id": 271356,
      "cbid": 251, "correlation": 271356
    }
  },
  {
    "ph": "f", "id": 271356, "pid": 76337, "tid": -914061504, "ts": 1716454225651737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651741, "dur": 0,
    "args": {
      "External id": 271357,
      "cbid": 251, "correlation": 271357
    }
  },
  {
    "ph": "f", "id": 271357, "pid": 76337, "tid": -914061504, "ts": 1716454225651741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225801063, "dur": 11,
    "args": {
      "External id": 271358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271358,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271358, "pid": 5, "tid": 7, "ts": 1716454225801063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651743, "dur": 13,
    "args": {
      "External id": 271358,
      "cbid": 211, "correlation": 271358
    }
  },
  {
    "ph": "s", "id": 271358, "pid": 76337, "tid": -914061504, "ts": 1716454225651743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225801075, "dur": 5,
    "args": {
      "External id": 271360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271360,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 271360, "pid": 5, "tid": 7, "ts": 1716454225801075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651759, "dur": 6,
    "args": {
      "External id": 271360,
      "cbid": 211, "correlation": 271360
    }
  },
  {
    "ph": "s", "id": 271360, "pid": 76337, "tid": -914061504, "ts": 1716454225651759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651821, "dur": 1,
    "args": {
      "External id": 271371,
      "cbid": 251, "correlation": 271371
    }
  },
  {
    "ph": "f", "id": 271371, "pid": 76337, "tid": -914061504, "ts": 1716454225651821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651824, "dur": 0,
    "args": {
      "External id": 271372,
      "cbid": 251, "correlation": 271372
    }
  },
  {
    "ph": "f", "id": 271372, "pid": 76337, "tid": -914061504, "ts": 1716454225651824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225801082, "dur": 6,
    "args": {
      "External id": 271373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271373,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271373, "pid": 5, "tid": 7, "ts": 1716454225801082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651826, "dur": 13,
    "args": {
      "External id": 271373,
      "cbid": 211, "correlation": 271373
    }
  },
  {
    "ph": "s", "id": 271373, "pid": 76337, "tid": -914061504, "ts": 1716454225651826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225801089, "dur": 4,
    "args": {
      "External id": 271375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271375,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 271375, "pid": 5, "tid": 7, "ts": 1716454225801089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651840, "dur": 6,
    "args": {
      "External id": 271375,
      "cbid": 211, "correlation": 271375
    }
  },
  {
    "ph": "s", "id": 271375, "pid": 76337, "tid": -914061504, "ts": 1716454225651840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225801094, "dur": 160,
    "args": {
      "External id": 271396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271396,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 271396, "pid": 5, "tid": 7, "ts": 1716454225801094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225651925, "dur": 14,
    "args": {
      "External id": 271396,
      "cbid": 211, "correlation": 271396
    }
  },
  {
    "ph": "s", "id": 271396, "pid": 76337, "tid": -914061504, "ts": 1716454225651925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652041, "dur": 1,
    "args": {
      "External id": 271414,
      "cbid": 251, "correlation": 271414
    }
  },
  {
    "ph": "f", "id": 271414, "pid": 76337, "tid": -914061504, "ts": 1716454225652041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225801256, "dur": 110,
    "args": {
      "External id": 271416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271416,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 271416, "pid": 5, "tid": 7, "ts": 1716454225801256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652047, "dur": 14,
    "args": {
      "External id": 271416,
      "cbid": 211, "correlation": 271416
    }
  },
  {
    "ph": "s", "id": 271416, "pid": 76337, "tid": -914061504, "ts": 1716454225652047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225801367, "dur": 35,
    "args": {
      "External id": 271424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271424,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271424, "pid": 5, "tid": 7, "ts": 1716454225801367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652125, "dur": 12,
    "args": {
      "External id": 271424,
      "cbid": 211, "correlation": 271424
    }
  },
  {
    "ph": "s", "id": 271424, "pid": 76337, "tid": -914061504, "ts": 1716454225652125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225801404, "dur": 67,
    "args": {
      "External id": 271432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271432, "pid": 5, "tid": 7, "ts": 1716454225801404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652169, "dur": 10,
    "args": {
      "External id": 271432,
      "cbid": 211, "correlation": 271432
    }
  },
  {
    "ph": "s", "id": 271432, "pid": 76337, "tid": -914061504, "ts": 1716454225652169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225801473, "dur": 95,
    "args": {
      "External id": 271454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271454,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271454, "pid": 5, "tid": 7, "ts": 1716454225801473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652222, "dur": 11,
    "args": {
      "External id": 271454,
      "cbid": 211, "correlation": 271454
    }
  },
  {
    "ph": "s", "id": 271454, "pid": 76337, "tid": -914061504, "ts": 1716454225652222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652312, "dur": 1,
    "args": {
      "External id": 271470,
      "cbid": 251, "correlation": 271470
    }
  },
  {
    "ph": "f", "id": 271470, "pid": 76337, "tid": -914061504, "ts": 1716454225652312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225801569, "dur": 590,
    "args": {
      "External id": 271472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271472,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 271472, "pid": 5, "tid": 7, "ts": 1716454225801569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652318, "dur": 13,
    "args": {
      "External id": 271472,
      "cbid": 211, "correlation": 271472
    }
  },
  {
    "ph": "s", "id": 271472, "pid": 76337, "tid": -914061504, "ts": 1716454225652318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225802160, "dur": 250,
    "args": {
      "External id": 271480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271480, "pid": 5, "tid": 7, "ts": 1716454225802160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652394, "dur": 13,
    "args": {
      "External id": 271480,
      "cbid": 211, "correlation": 271480
    }
  },
  {
    "ph": "s", "id": 271480, "pid": 76337, "tid": -914061504, "ts": 1716454225652394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225802412, "dur": 255,
    "args": {
      "External id": 271488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271488, "pid": 5, "tid": 7, "ts": 1716454225802412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652429, "dur": 9,
    "args": {
      "External id": 271488,
      "cbid": 211, "correlation": 271488
    }
  },
  {
    "ph": "s", "id": 271488, "pid": 76337, "tid": -914061504, "ts": 1716454225652429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652511, "dur": 2,
    "args": {
      "External id": 271504,
      "cbid": 251, "correlation": 271504
    }
  },
  {
    "ph": "f", "id": 271504, "pid": 76337, "tid": -914061504, "ts": 1716454225652511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652517, "dur": 0,
    "args": {
      "External id": 271506,
      "cbid": 251, "correlation": 271506
    }
  },
  {
    "ph": "f", "id": 271506, "pid": 76337, "tid": -914061504, "ts": 1716454225652517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225802668, "dur": 362,
    "args": {
      "External id": 271507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271507,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271507, "pid": 5, "tid": 7, "ts": 1716454225802668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652520, "dur": 14,
    "args": {
      "External id": 271507,
      "cbid": 211, "correlation": 271507
    }
  },
  {
    "ph": "s", "id": 271507, "pid": 76337, "tid": -914061504, "ts": 1716454225652520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225803031, "dur": 51,
    "args": {
      "External id": 271515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271515,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271515, "pid": 5, "tid": 7, "ts": 1716454225803031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652563, "dur": 112,
    "args": {
      "External id": 271515,
      "cbid": 211, "correlation": 271515
    }
  },
  {
    "ph": "s", "id": 271515, "pid": 76337, "tid": -914061504, "ts": 1716454225652563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225803084, "dur": 162,
    "args": {
      "External id": 271526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271526, "pid": 5, "tid": 7, "ts": 1716454225803084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652741, "dur": 63,
    "args": {
      "External id": 271526,
      "cbid": 211, "correlation": 271526
    }
  },
  {
    "ph": "s", "id": 271526, "pid": 76337, "tid": -914061504, "ts": 1716454225652741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652873, "dur": 0,
    "args": {
      "External id": 271538,
      "cbid": 317, "correlation": 271538
    }
  },
  {
    "ph": "f", "id": 271538, "pid": 76337, "tid": -914061504, "ts": 1716454225652873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652874, "dur": 1,
    "args": {
      "External id": 271539,
      "cbid": 203, "correlation": 271539
    }
  },
  {
    "ph": "f", "id": 271539, "pid": 76337, "tid": -914061504, "ts": 1716454225652874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652876, "dur": 0,
    "args": {
      "External id": 271540,
      "cbid": 205, "correlation": 271540
    }
  },
  {
    "ph": "f", "id": 271540, "pid": 76337, "tid": -914061504, "ts": 1716454225652876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652907, "dur": 1,
    "args": {
      "External id": 271544,
      "cbid": 251, "correlation": 271544
    }
  },
  {
    "ph": "f", "id": 271544, "pid": 76337, "tid": -914061504, "ts": 1716454225652907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652910, "dur": 0,
    "args": {
      "External id": 271545,
      "cbid": 251, "correlation": 271545
    }
  },
  {
    "ph": "f", "id": 271545, "pid": 76337, "tid": -914061504, "ts": 1716454225652910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652910, "dur": 0,
    "args": {
      "External id": 271546,
      "cbid": 251, "correlation": 271546
    }
  },
  {
    "ph": "f", "id": 271546, "pid": 76337, "tid": -914061504, "ts": 1716454225652910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652912, "dur": 0,
    "args": {
      "External id": 271547,
      "cbid": 251, "correlation": 271547
    }
  },
  {
    "ph": "f", "id": 271547, "pid": 76337, "tid": -914061504, "ts": 1716454225652912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652913, "dur": 0,
    "args": {
      "External id": 271548,
      "cbid": 251, "correlation": 271548
    }
  },
  {
    "ph": "f", "id": 271548, "pid": 76337, "tid": -914061504, "ts": 1716454225652913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652914, "dur": 0,
    "args": {
      "External id": 271549,
      "cbid": 251, "correlation": 271549
    }
  },
  {
    "ph": "f", "id": 271549, "pid": 76337, "tid": -914061504, "ts": 1716454225652914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652915, "dur": 0,
    "args": {
      "External id": 271550,
      "cbid": 251, "correlation": 271550
    }
  },
  {
    "ph": "f", "id": 271550, "pid": 76337, "tid": -914061504, "ts": 1716454225652915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652916, "dur": 0,
    "args": {
      "External id": 271551,
      "cbid": 251, "correlation": 271551
    }
  },
  {
    "ph": "f", "id": 271551, "pid": 76337, "tid": -914061504, "ts": 1716454225652916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652918, "dur": 0,
    "args": {
      "External id": 271552,
      "cbid": 251, "correlation": 271552
    }
  },
  {
    "ph": "f", "id": 271552, "pid": 76337, "tid": -914061504, "ts": 1716454225652918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225803246, "dur": 117,
    "args": {
      "External id": 271553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271553,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 271553, "pid": 5, "tid": 7, "ts": 1716454225803246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652920, "dur": 13,
    "args": {
      "External id": 271553,
      "cbid": 211, "correlation": 271553
    }
  },
  {
    "ph": "s", "id": 271553, "pid": 76337, "tid": -914061504, "ts": 1716454225652920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225803365, "dur": 61,
    "args": {
      "External id": 271559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271559, "pid": 5, "tid": 7, "ts": 1716454225803365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225652959, "dur": 285,
    "args": {
      "External id": 271559,
      "cbid": 211, "correlation": 271559
    }
  },
  {
    "ph": "s", "id": 271559, "pid": 76337, "tid": -914061504, "ts": 1716454225652959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225803427, "dur": 50,
    "args": {
      "External id": 271567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271567,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271567, "pid": 5, "tid": 7, "ts": 1716454225803427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653268, "dur": 9,
    "args": {
      "External id": 271567,
      "cbid": 211, "correlation": 271567
    }
  },
  {
    "ph": "s", "id": 271567, "pid": 76337, "tid": -914061504, "ts": 1716454225653268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225803479, "dur": 54,
    "args": {
      "External id": 271587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271587,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 271587, "pid": 5, "tid": 7, "ts": 1716454225803479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653363, "dur": 12,
    "args": {
      "External id": 271587,
      "cbid": 211, "correlation": 271587
    }
  },
  {
    "ph": "s", "id": 271587, "pid": 76337, "tid": -914061504, "ts": 1716454225653363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225803534, "dur": 4,
    "args": {
      "External id": 271599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271599,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 271599, "pid": 5, "tid": 7, "ts": 1716454225803534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653387, "dur": 8,
    "args": {
      "External id": 271599,
      "cbid": 211, "correlation": 271599
    }
  },
  {
    "ph": "s", "id": 271599, "pid": 76337, "tid": -914061504, "ts": 1716454225653387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225803539, "dur": 57,
    "args": {
      "External id": 271602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271602, "pid": 5, "tid": 7, "ts": 1716454225803539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653408, "dur": 97,
    "args": {
      "External id": 271602,
      "cbid": 211, "correlation": 271602
    }
  },
  {
    "ph": "s", "id": 271602, "pid": 76337, "tid": -914061504, "ts": 1716454225653408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225803598, "dur": 38,
    "args": {
      "External id": 271611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271611,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271611, "pid": 5, "tid": 7, "ts": 1716454225803598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653553, "dur": 11,
    "args": {
      "External id": 271611,
      "cbid": 211, "correlation": 271611
    }
  },
  {
    "ph": "s", "id": 271611, "pid": 76337, "tid": -914061504, "ts": 1716454225653553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653609, "dur": 0,
    "args": {
      "External id": 271621,
      "cbid": 317, "correlation": 271621
    }
  },
  {
    "ph": "f", "id": 271621, "pid": 76337, "tid": -914061504, "ts": 1716454225653609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653610, "dur": 0,
    "args": {
      "External id": 271622,
      "cbid": 203, "correlation": 271622
    }
  },
  {
    "ph": "f", "id": 271622, "pid": 76337, "tid": -914061504, "ts": 1716454225653610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653611, "dur": 0,
    "args": {
      "External id": 271623,
      "cbid": 205, "correlation": 271623
    }
  },
  {
    "ph": "f", "id": 271623, "pid": 76337, "tid": -914061504, "ts": 1716454225653611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225803637, "dur": 40,
    "args": {
      "External id": 271627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271627, "pid": 5, "tid": 7, "ts": 1716454225803637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653630, "dur": 13,
    "args": {
      "External id": 271627,
      "cbid": 211, "correlation": 271627
    }
  },
  {
    "ph": "s", "id": 271627, "pid": 76337, "tid": -914061504, "ts": 1716454225653630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225803678, "dur": 3,
    "args": {
      "External id": 271629,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271629,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.100840336134453
    }
  },
  {
    "ph": "f", "id": 271629, "pid": 5, "tid": 7, "ts": 1716454225803678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653646, "dur": 22,
    "args": {
      "External id": 271629,
      "cbid": 51, "correlation": 271629
    }
  },
  {
    "ph": "s", "id": 271629, "pid": 76337, "tid": -914061504, "ts": 1716454225653646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653674, "dur": 2,
    "args": {
      "External id": 271631,
      "cbid": 200, "correlation": 271631
    }
  },
  {
    "ph": "f", "id": 271631, "pid": 76337, "tid": -914061504, "ts": 1716454225653674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653677, "dur": 0,
    "args": {
      "External id": 271632,
      "cbid": 200, "correlation": 271632
    }
  },
  {
    "ph": "f", "id": 271632, "pid": 76337, "tid": -914061504, "ts": 1716454225653677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653678, "dur": 0,
    "args": {
      "External id": 271633,
      "cbid": 200, "correlation": 271633
    }
  },
  {
    "ph": "f", "id": 271633, "pid": 76337, "tid": -914061504, "ts": 1716454225653678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653678, "dur": 0,
    "args": {
      "External id": 271634,
      "cbid": 200, "correlation": 271634
    }
  },
  {
    "ph": "f", "id": 271634, "pid": 76337, "tid": -914061504, "ts": 1716454225653678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653679, "dur": 3,
    "args": {
      "External id": 271635,
      "cbid": 15, "correlation": 271635
    }
  },
  {
    "ph": "f", "id": 271635, "pid": 76337, "tid": -914061504, "ts": 1716454225653679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653684, "dur": 1,
    "args": {
      "External id": 271636,
      "cbid": 251, "correlation": 271636
    }
  },
  {
    "ph": "f", "id": 271636, "pid": 76337, "tid": -914061504, "ts": 1716454225653684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225803683, "dur": 23,
    "args": {
      "External id": 271637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271637,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271637, "pid": 5, "tid": 7, "ts": 1716454225803683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653687, "dur": 9,
    "args": {
      "External id": 271637,
      "cbid": 211, "correlation": 271637
    }
  },
  {
    "ph": "s", "id": 271637, "pid": 76337, "tid": -914061504, "ts": 1716454225653687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225803707, "dur": 4,
    "args": {
      "External id": 271639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271639,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 271639, "pid": 5, "tid": 7, "ts": 1716454225803707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653705, "dur": 7,
    "args": {
      "External id": 271639,
      "cbid": 211, "correlation": 271639
    }
  },
  {
    "ph": "s", "id": 271639, "pid": 76337, "tid": -914061504, "ts": 1716454225653705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653718, "dur": 0,
    "args": {
      "External id": 271640,
      "cbid": 51, "correlation": 271640
    }
  },
  {
    "ph": "s", "id": 271640, "pid": 76337, "tid": -914061504, "ts": 1716454225653718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225803712, "dur": 193,
    "args": {
      "External id": 271641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271641,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271641, "pid": 5, "tid": 7, "ts": 1716454225803712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653719, "dur": 180,
    "args": {
      "External id": 271641,
      "cbid": 211, "correlation": 271641
    }
  },
  {
    "ph": "s", "id": 271641, "pid": 76337, "tid": -914061504, "ts": 1716454225653719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225803907, "dur": 6,
    "args": {
      "External id": 271642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271642,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271642, "pid": 5, "tid": 7, "ts": 1716454225803907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653905, "dur": 7,
    "args": {
      "External id": 271642,
      "cbid": 211, "correlation": 271642
    }
  },
  {
    "ph": "s", "id": 271642, "pid": 76337, "tid": -914061504, "ts": 1716454225653905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225803915, "dur": 5,
    "args": {
      "External id": 271648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 271648, "pid": 5, "tid": 7, "ts": 1716454225803915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225653937, "dur": 8,
    "args": {
      "External id": 271648,
      "cbid": 211, "correlation": 271648
    }
  },
  {
    "ph": "s", "id": 271648, "pid": 76337, "tid": -914061504, "ts": 1716454225653937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225803921, "dur": 3,
    "args": {
      "External id": 271656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271656,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271656, "pid": 5, "tid": 7, "ts": 1716454225803921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225655648, "dur": 16,
    "args": {
      "External id": 271656,
      "cbid": 211, "correlation": 271656
    }
  },
  {
    "ph": "s", "id": 271656, "pid": 76337, "tid": -914061504, "ts": 1716454225655648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225803926, "dur": 3,
    "args": {
      "External id": 271664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271664,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271664, "pid": 5, "tid": 7, "ts": 1716454225803926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225655692, "dur": 10,
    "args": {
      "External id": 271664,
      "cbid": 211, "correlation": 271664
    }
  },
  {
    "ph": "s", "id": 271664, "pid": 76337, "tid": -914061504, "ts": 1716454225655692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225803930, "dur": 3,
    "args": {
      "External id": 271672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271672,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271672, "pid": 5, "tid": 7, "ts": 1716454225803930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225655719, "dur": 8,
    "args": {
      "External id": 271672,
      "cbid": 211, "correlation": 271672
    }
  },
  {
    "ph": "s", "id": 271672, "pid": 76337, "tid": -914061504, "ts": 1716454225655719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225803934, "dur": 3,
    "args": {
      "External id": 271681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271681,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271681, "pid": 5, "tid": 7, "ts": 1716454225803934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225655897, "dur": 14,
    "args": {
      "External id": 271681,
      "cbid": 211, "correlation": 271681
    }
  },
  {
    "ph": "s", "id": 271681, "pid": 76337, "tid": -914061504, "ts": 1716454225655897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225803939, "dur": 3,
    "args": {
      "External id": 271690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271690,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271690, "pid": 5, "tid": 7, "ts": 1716454225803939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225655928, "dur": 7,
    "args": {
      "External id": 271690,
      "cbid": 211, "correlation": 271690
    }
  },
  {
    "ph": "s", "id": 271690, "pid": 76337, "tid": -914061504, "ts": 1716454225655928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225803943, "dur": 3,
    "args": {
      "External id": 271698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271698,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271698, "pid": 5, "tid": 7, "ts": 1716454225803943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225655953, "dur": 8,
    "args": {
      "External id": 271698,
      "cbid": 211, "correlation": 271698
    }
  },
  {
    "ph": "s", "id": 271698, "pid": 76337, "tid": -914061504, "ts": 1716454225655953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225803947, "dur": 3,
    "args": {
      "External id": 271706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271706,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271706, "pid": 5, "tid": 7, "ts": 1716454225803947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225656223, "dur": 16,
    "args": {
      "External id": 271706,
      "cbid": 211, "correlation": 271706
    }
  },
  {
    "ph": "s", "id": 271706, "pid": 76337, "tid": -914061504, "ts": 1716454225656223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225803952, "dur": 3,
    "args": {
      "External id": 271714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271714,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271714, "pid": 5, "tid": 7, "ts": 1716454225803952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225656254, "dur": 8,
    "args": {
      "External id": 271714,
      "cbid": 211, "correlation": 271714
    }
  },
  {
    "ph": "s", "id": 271714, "pid": 76337, "tid": -914061504, "ts": 1716454225656254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225803957, "dur": 1,
    "args": {
      "External id": 271724,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271724,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 271724, "pid": 5, "tid": 7, "ts": 1716454225803957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225656321, "dur": 35,
    "args": {
      "External id": 271724,
      "cbid": 41, "correlation": 271724
    }
  },
  {
    "ph": "s", "id": 271724, "pid": 76337, "tid": -914061504, "ts": 1716454225656321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225656358, "dur": 147623,
    "args": {
      "External id": 271725,
      "cbid": 131, "correlation": 271725
    }
  },
  {
    "ph": "f", "id": 271725, "pid": 76337, "tid": -914061504, "ts": 1716454225656358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225804179, "dur": 3,
    "args": {
      "External id": 271733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271733, "pid": 5, "tid": 7, "ts": 1716454225804179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804149, "dur": 32,
    "args": {
      "External id": 271733,
      "cbid": 211, "correlation": 271733
    }
  },
  {
    "ph": "s", "id": 271733, "pid": 76337, "tid": -914061504, "ts": 1716454225804149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225804293, "dur": 3,
    "args": {
      "External id": 271742,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271742,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 271742, "pid": 5, "tid": 7, "ts": 1716454225804293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804250, "dur": 43,
    "args": {
      "External id": 271742,
      "cbid": 41, "correlation": 271742
    }
  },
  {
    "ph": "s", "id": 271742, "pid": 76337, "tid": -914061504, "ts": 1716454225804250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225804399, "dur": 4,
    "args": {
      "External id": 271752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271752,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 271752, "pid": 5, "tid": 7, "ts": 1716454225804399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804382, "dur": 19,
    "args": {
      "External id": 271752,
      "cbid": 211, "correlation": 271752
    }
  },
  {
    "ph": "s", "id": 271752, "pid": 76337, "tid": -914061504, "ts": 1716454225804382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225804487, "dur": 1,
    "args": {
      "External id": 271762,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271762,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 271762, "pid": 5, "tid": 7, "ts": 1716454225804487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804463, "dur": 23,
    "args": {
      "External id": 271762,
      "cbid": 41, "correlation": 271762
    }
  },
  {
    "ph": "s", "id": 271762, "pid": 76337, "tid": -914061504, "ts": 1716454225804463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804487, "dur": 8,
    "args": {
      "External id": 271763,
      "cbid": 131, "correlation": 271763
    }
  },
  {
    "ph": "f", "id": 271763, "pid": 76337, "tid": -914061504, "ts": 1716454225804487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225804573, "dur": 3,
    "args": {
      "External id": 271770,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271770,
      "bytes": 98304, "memory bandwidth (GB/s)": 29.825242718446603
    }
  },
  {
    "ph": "f", "id": 271770, "pid": 5, "tid": 7, "ts": 1716454225804573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804539, "dur": 34,
    "args": {
      "External id": 271770,
      "cbid": 41, "correlation": 271770
    }
  },
  {
    "ph": "s", "id": 271770, "pid": 76337, "tid": -914061504, "ts": 1716454225804539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225804673, "dur": 3,
    "args": {
      "External id": 271789,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271789,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 271789, "pid": 5, "tid": 7, "ts": 1716454225804673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804652, "dur": 20,
    "args": {
      "External id": 271789,
      "cbid": 41, "correlation": 271789
    }
  },
  {
    "ph": "s", "id": 271789, "pid": 76337, "tid": -914061504, "ts": 1716454225804652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225804713, "dur": 3,
    "args": {
      "External id": 271795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271795,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271795, "pid": 5, "tid": 7, "ts": 1716454225804713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804702, "dur": 11,
    "args": {
      "External id": 271795,
      "cbid": 211, "correlation": 271795
    }
  },
  {
    "ph": "s", "id": 271795, "pid": 76337, "tid": -914061504, "ts": 1716454225804702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225804727, "dur": 6,
    "args": {
      "External id": 271797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271797,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 271797, "pid": 5, "tid": 7, "ts": 1716454225804727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804717, "dur": 9,
    "args": {
      "External id": 271797,
      "cbid": 211, "correlation": 271797
    }
  },
  {
    "ph": "s", "id": 271797, "pid": 76337, "tid": -914061504, "ts": 1716454225804717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454225804735, "dur": 3,
    "args": {
      "External id": 271799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271799, "pid": 5, "tid": 7, "ts": 1716454225804735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804728, "dur": 7,
    "args": {
      "External id": 271799,
      "cbid": 211, "correlation": 271799
    }
  },
  {
    "ph": "s", "id": 271799, "pid": 76337, "tid": -914061504, "ts": 1716454225804728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225804773, "dur": 2,
    "args": {
      "External id": 271807,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271807,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002688172043010753
    }
  },
  {
    "ph": "f", "id": 271807, "pid": 5, "tid": 7, "ts": 1716454225804773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804757, "dur": 15,
    "args": {
      "External id": 271807,
      "cbid": 41, "correlation": 271807
    }
  },
  {
    "ph": "s", "id": 271807, "pid": 76337, "tid": -914061504, "ts": 1716454225804757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225804822, "dur": 3,
    "args": {
      "External id": 271821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271821,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271821, "pid": 5, "tid": 7, "ts": 1716454225804822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804811, "dur": 12,
    "args": {
      "External id": 271821,
      "cbid": 211, "correlation": 271821
    }
  },
  {
    "ph": "s", "id": 271821, "pid": 76337, "tid": -914061504, "ts": 1716454225804811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225804843, "dur": 2,
    "args": {
      "External id": 271835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271835,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271835, "pid": 5, "tid": 7, "ts": 1716454225804843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804835, "dur": 7,
    "args": {
      "External id": 271835,
      "cbid": 211, "correlation": 271835
    }
  },
  {
    "ph": "s", "id": 271835, "pid": 76337, "tid": -914061504, "ts": 1716454225804835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225804878, "dur": 6,
    "args": {
      "External id": 271842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271842,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271842, "pid": 5, "tid": 7, "ts": 1716454225804878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804867, "dur": 11,
    "args": {
      "External id": 271842,
      "cbid": 211, "correlation": 271842
    }
  },
  {
    "ph": "s", "id": 271842, "pid": 76337, "tid": -914061504, "ts": 1716454225804867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454225804890, "dur": 6,
    "args": {
      "External id": 271845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271845,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271845, "pid": 5, "tid": 7, "ts": 1716454225804890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804881, "dur": 8,
    "args": {
      "External id": 271845,
      "cbid": 211, "correlation": 271845
    }
  },
  {
    "ph": "s", "id": 271845, "pid": 76337, "tid": -914061504, "ts": 1716454225804881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225804899, "dur": 3,
    "args": {
      "External id": 271847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271847,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271847, "pid": 5, "tid": 7, "ts": 1716454225804899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804891, "dur": 7,
    "args": {
      "External id": 271847,
      "cbid": 211, "correlation": 271847
    }
  },
  {
    "ph": "s", "id": 271847, "pid": 76337, "tid": -914061504, "ts": 1716454225804891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225804920, "dur": 2,
    "args": {
      "External id": 271850,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271850,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002717391304347826
    }
  },
  {
    "ph": "f", "id": 271850, "pid": 5, "tid": 7, "ts": 1716454225804920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804907, "dur": 12,
    "args": {
      "External id": 271850,
      "cbid": 41, "correlation": 271850
    }
  },
  {
    "ph": "s", "id": 271850, "pid": 76337, "tid": -914061504, "ts": 1716454225804907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225804972, "dur": 4,
    "args": {
      "External id": 271866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271866,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 271866, "pid": 5, "tid": 7, "ts": 1716454225804972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804960, "dur": 12,
    "args": {
      "External id": 271866,
      "cbid": 211, "correlation": 271866
    }
  },
  {
    "ph": "s", "id": 271866, "pid": 76337, "tid": -914061504, "ts": 1716454225804960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225805004, "dur": 3,
    "args": {
      "External id": 271871,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271871,
      "bytes": 1, "memory bandwidth (GB/s)": 0.00030637254901960784
    }
  },
  {
    "ph": "f", "id": 271871, "pid": 5, "tid": 7, "ts": 1716454225805004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225804986, "dur": 17,
    "args": {
      "External id": 271871,
      "cbid": 41, "correlation": 271871
    }
  },
  {
    "ph": "s", "id": 271871, "pid": 76337, "tid": -914061504, "ts": 1716454225804986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225805033, "dur": 1,
    "args": {
      "External id": 271877,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271877,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 271877, "pid": 5, "tid": 7, "ts": 1716454225805033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805013, "dur": 29,
    "args": {
      "External id": 271877,
      "cbid": 41, "correlation": 271877
    }
  },
  {
    "ph": "s", "id": 271877, "pid": 76337, "tid": -914061504, "ts": 1716454225805013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805042, "dur": 3,
    "args": {
      "External id": 271878,
      "cbid": 131, "correlation": 271878
    }
  },
  {
    "ph": "f", "id": 271878, "pid": 76337, "tid": -914061504, "ts": 1716454225805042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225805097, "dur": 3,
    "args": {
      "External id": 271886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271886,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271886, "pid": 5, "tid": 7, "ts": 1716454225805097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805084, "dur": 15,
    "args": {
      "External id": 271886,
      "cbid": 211, "correlation": 271886
    }
  },
  {
    "ph": "s", "id": 271886, "pid": 76337, "tid": -914061504, "ts": 1716454225805084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225805133, "dur": 3,
    "args": {
      "External id": 271896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271896,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271896, "pid": 5, "tid": 7, "ts": 1716454225805133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805120, "dur": 12,
    "args": {
      "External id": 271896,
      "cbid": 211, "correlation": 271896
    }
  },
  {
    "ph": "s", "id": 271896, "pid": 76337, "tid": -914061504, "ts": 1716454225805120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225805159, "dur": 3,
    "args": {
      "External id": 271905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271905,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271905, "pid": 5, "tid": 7, "ts": 1716454225805159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805149, "dur": 9,
    "args": {
      "External id": 271905,
      "cbid": 211, "correlation": 271905
    }
  },
  {
    "ph": "s", "id": 271905, "pid": 76337, "tid": -914061504, "ts": 1716454225805149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225805278, "dur": 12,
    "args": {
      "External id": 271915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271915,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271915, "pid": 5, "tid": 7, "ts": 1716454225805278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805264, "dur": 15,
    "args": {
      "External id": 271915,
      "cbid": 211, "correlation": 271915
    }
  },
  {
    "ph": "s", "id": 271915, "pid": 76337, "tid": -914061504, "ts": 1716454225805264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225805318, "dur": 3,
    "args": {
      "External id": 271923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271923,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271923, "pid": 5, "tid": 7, "ts": 1716454225805318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805308, "dur": 9,
    "args": {
      "External id": 271923,
      "cbid": 211, "correlation": 271923
    }
  },
  {
    "ph": "s", "id": 271923, "pid": 76337, "tid": -914061504, "ts": 1716454225805308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225805369, "dur": 11,
    "args": {
      "External id": 271933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271933,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271933, "pid": 5, "tid": 7, "ts": 1716454225805369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805356, "dur": 14,
    "args": {
      "External id": 271933,
      "cbid": 211, "correlation": 271933
    }
  },
  {
    "ph": "s", "id": 271933, "pid": 76337, "tid": -914061504, "ts": 1716454225805356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225805406, "dur": 10,
    "args": {
      "External id": 271941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271941,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271941, "pid": 5, "tid": 7, "ts": 1716454225805406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805395, "dur": 11,
    "args": {
      "External id": 271941,
      "cbid": 211, "correlation": 271941
    }
  },
  {
    "ph": "s", "id": 271941, "pid": 76337, "tid": -914061504, "ts": 1716454225805395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225805437, "dur": 3,
    "args": {
      "External id": 271950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271950, "pid": 5, "tid": 7, "ts": 1716454225805437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805427, "dur": 9,
    "args": {
      "External id": 271950,
      "cbid": 211, "correlation": 271950
    }
  },
  {
    "ph": "s", "id": 271950, "pid": 76337, "tid": -914061504, "ts": 1716454225805427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225805462, "dur": 5,
    "args": {
      "External id": 271959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271959,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271959, "pid": 5, "tid": 7, "ts": 1716454225805462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805452, "dur": 9,
    "args": {
      "External id": 271959,
      "cbid": 211, "correlation": 271959
    }
  },
  {
    "ph": "s", "id": 271959, "pid": 76337, "tid": -914061504, "ts": 1716454225805452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225805508, "dur": 8,
    "args": {
      "External id": 271969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271969,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271969, "pid": 5, "tid": 7, "ts": 1716454225805508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805495, "dur": 13,
    "args": {
      "External id": 271969,
      "cbid": 211, "correlation": 271969
    }
  },
  {
    "ph": "s", "id": 271969, "pid": 76337, "tid": -914061504, "ts": 1716454225805495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225805843, "dur": 3,
    "args": {
      "External id": 271978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271978,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271978, "pid": 5, "tid": 7, "ts": 1716454225805843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805827, "dur": 16,
    "args": {
      "External id": 271978,
      "cbid": 211, "correlation": 271978
    }
  },
  {
    "ph": "s", "id": 271978, "pid": 76337, "tid": -914061504, "ts": 1716454225805827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225805873, "dur": 3,
    "args": {
      "External id": 271986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 271986,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 271986, "pid": 5, "tid": 7, "ts": 1716454225805873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805862, "dur": 10,
    "args": {
      "External id": 271986,
      "cbid": 211, "correlation": 271986
    }
  },
  {
    "ph": "s", "id": 271986, "pid": 76337, "tid": -914061504, "ts": 1716454225805862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225805928, "dur": 1,
    "args": {
      "External id": 271996,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 271996,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 271996, "pid": 5, "tid": 7, "ts": 1716454225805928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805909, "dur": 17,
    "args": {
      "External id": 271996,
      "cbid": 41, "correlation": 271996
    }
  },
  {
    "ph": "s", "id": 271996, "pid": 76337, "tid": -914061504, "ts": 1716454225805909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225805926, "dur": 8,
    "args": {
      "External id": 271997,
      "cbid": 131, "correlation": 271997
    }
  },
  {
    "ph": "f", "id": 271997, "pid": 76337, "tid": -914061504, "ts": 1716454225805926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454225806027, "dur": 2,
    "args": {
      "External id": 272005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272005, "pid": 5, "tid": 7, "ts": 1716454225806027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806012, "dur": 15,
    "args": {
      "External id": 272005,
      "cbid": 211, "correlation": 272005
    }
  },
  {
    "ph": "s", "id": 272005, "pid": 76337, "tid": -914061504, "ts": 1716454225806012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225806100, "dur": 3,
    "args": {
      "External id": 272014,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272014,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002551020408163265
    }
  },
  {
    "ph": "f", "id": 272014, "pid": 5, "tid": 7, "ts": 1716454225806100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806081, "dur": 19,
    "args": {
      "External id": 272014,
      "cbid": 41, "correlation": 272014
    }
  },
  {
    "ph": "s", "id": 272014, "pid": 76337, "tid": -914061504, "ts": 1716454225806081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225806170, "dur": 3,
    "args": {
      "External id": 272024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272024,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272024, "pid": 5, "tid": 7, "ts": 1716454225806170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806156, "dur": 14,
    "args": {
      "External id": 272024,
      "cbid": 211, "correlation": 272024
    }
  },
  {
    "ph": "s", "id": 272024, "pid": 76337, "tid": -914061504, "ts": 1716454225806156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225806221, "dur": 1,
    "args": {
      "External id": 272034,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272034,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 272034, "pid": 5, "tid": 7, "ts": 1716454225806221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806208, "dur": 11,
    "args": {
      "External id": 272034,
      "cbid": 41, "correlation": 272034
    }
  },
  {
    "ph": "s", "id": 272034, "pid": 76337, "tid": -914061504, "ts": 1716454225806208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806220, "dur": 8,
    "args": {
      "External id": 272035,
      "cbid": 131, "correlation": 272035
    }
  },
  {
    "ph": "f", "id": 272035, "pid": 76337, "tid": -914061504, "ts": 1716454225806220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225806283, "dur": 3,
    "args": {
      "External id": 272042,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272042,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.03030303030303
    }
  },
  {
    "ph": "f", "id": 272042, "pid": 5, "tid": 7, "ts": 1716454225806283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806261, "dur": 21,
    "args": {
      "External id": 272042,
      "cbid": 41, "correlation": 272042
    }
  },
  {
    "ph": "s", "id": 272042, "pid": 76337, "tid": -914061504, "ts": 1716454225806261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225806334, "dur": 1,
    "args": {
      "External id": 272053,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272053,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0012755102040816326
    }
  },
  {
    "ph": "f", "id": 272053, "pid": 5, "tid": 7, "ts": 1716454225806334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806319, "dur": 13,
    "args": {
      "External id": 272053,
      "cbid": 41, "correlation": 272053
    }
  },
  {
    "ph": "s", "id": 272053, "pid": 76337, "tid": -914061504, "ts": 1716454225806319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806333, "dur": 8,
    "args": {
      "External id": 272054,
      "cbid": 131, "correlation": 272054
    }
  },
  {
    "ph": "f", "id": 272054, "pid": 76337, "tid": -914061504, "ts": 1716454225806333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225806383, "dur": 3,
    "args": {
      "External id": 272062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272062,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272062, "pid": 5, "tid": 7, "ts": 1716454225806383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806370, "dur": 14,
    "args": {
      "External id": 272062,
      "cbid": 211, "correlation": 272062
    }
  },
  {
    "ph": "s", "id": 272062, "pid": 76337, "tid": -914061504, "ts": 1716454225806370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225806416, "dur": 3,
    "args": {
      "External id": 272072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272072,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272072, "pid": 5, "tid": 7, "ts": 1716454225806416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806406, "dur": 8,
    "args": {
      "External id": 272072,
      "cbid": 211, "correlation": 272072
    }
  },
  {
    "ph": "s", "id": 272072, "pid": 76337, "tid": -914061504, "ts": 1716454225806406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225806439, "dur": 3,
    "args": {
      "External id": 272081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272081,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272081, "pid": 5, "tid": 7, "ts": 1716454225806439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806430, "dur": 8,
    "args": {
      "External id": 272081,
      "cbid": 211, "correlation": 272081
    }
  },
  {
    "ph": "s", "id": 272081, "pid": 76337, "tid": -914061504, "ts": 1716454225806430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225806507, "dur": 6,
    "args": {
      "External id": 272089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272089, "pid": 5, "tid": 7, "ts": 1716454225806507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806495, "dur": 13,
    "args": {
      "External id": 272089,
      "cbid": 211, "correlation": 272089
    }
  },
  {
    "ph": "s", "id": 272089, "pid": 76337, "tid": -914061504, "ts": 1716454225806495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225806548, "dur": 3,
    "args": {
      "External id": 272098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272098,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272098, "pid": 5, "tid": 7, "ts": 1716454225806548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806538, "dur": 9,
    "args": {
      "External id": 272098,
      "cbid": 211, "correlation": 272098
    }
  },
  {
    "ph": "s", "id": 272098, "pid": 76337, "tid": -914061504, "ts": 1716454225806538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225806570, "dur": 3,
    "args": {
      "External id": 272107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272107,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272107, "pid": 5, "tid": 7, "ts": 1716454225806570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806562, "dur": 7,
    "args": {
      "External id": 272107,
      "cbid": 211, "correlation": 272107
    }
  },
  {
    "ph": "s", "id": 272107, "pid": 76337, "tid": -914061504, "ts": 1716454225806562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225806631, "dur": 3,
    "args": {
      "External id": 272115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272115,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272115, "pid": 5, "tid": 7, "ts": 1716454225806631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806621, "dur": 10,
    "args": {
      "External id": 272115,
      "cbid": 211, "correlation": 272115
    }
  },
  {
    "ph": "s", "id": 272115, "pid": 76337, "tid": -914061504, "ts": 1716454225806621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225806695, "dur": 1,
    "args": {
      "External id": 272123,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272123,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0043859649122807015
    }
  },
  {
    "ph": "f", "id": 272123, "pid": 5, "tid": 7, "ts": 1716454225806695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806674, "dur": 30,
    "args": {
      "External id": 272123,
      "cbid": 41, "correlation": 272123
    }
  },
  {
    "ph": "s", "id": 272123, "pid": 76337, "tid": -914061504, "ts": 1716454225806674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806705, "dur": 3,
    "args": {
      "External id": 272124,
      "cbid": 131, "correlation": 272124
    }
  },
  {
    "ph": "f", "id": 272124, "pid": 76337, "tid": -914061504, "ts": 1716454225806705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225806770, "dur": 1,
    "args": {
      "External id": 272134,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272134,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 272134, "pid": 5, "tid": 7, "ts": 1716454225806770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806754, "dur": 14,
    "args": {
      "External id": 272134,
      "cbid": 41, "correlation": 272134
    }
  },
  {
    "ph": "s", "id": 272134, "pid": 76337, "tid": -914061504, "ts": 1716454225806754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806769, "dur": 8,
    "args": {
      "External id": 272135,
      "cbid": 131, "correlation": 272135
    }
  },
  {
    "ph": "f", "id": 272135, "pid": 76337, "tid": -914061504, "ts": 1716454225806769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225806826, "dur": 1,
    "args": {
      "External id": 272144,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272144,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 272144, "pid": 5, "tid": 7, "ts": 1716454225806826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806815, "dur": 8,
    "args": {
      "External id": 272144,
      "cbid": 41, "correlation": 272144
    }
  },
  {
    "ph": "s", "id": 272144, "pid": 76337, "tid": -914061504, "ts": 1716454225806815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806824, "dur": 8,
    "args": {
      "External id": 272145,
      "cbid": 131, "correlation": 272145
    }
  },
  {
    "ph": "f", "id": 272145, "pid": 76337, "tid": -914061504, "ts": 1716454225806824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225806897, "dur": 4,
    "args": {
      "External id": 272152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272152,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272152, "pid": 5, "tid": 7, "ts": 1716454225806897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806880, "dur": 17,
    "args": {
      "External id": 272152,
      "cbid": 211, "correlation": 272152
    }
  },
  {
    "ph": "s", "id": 272152, "pid": 76337, "tid": -914061504, "ts": 1716454225806880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454225806935, "dur": 4,
    "args": {
      "External id": 272172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272172,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272172, "pid": 5, "tid": 7, "ts": 1716454225806935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806924, "dur": 12,
    "args": {
      "External id": 272172,
      "cbid": 211, "correlation": 272172
    }
  },
  {
    "ph": "s", "id": 272172, "pid": 76337, "tid": -914061504, "ts": 1716454225806924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806937, "dur": 0,
    "args": {
      "External id": 272173,
      "cbid": 11, "correlation": 272173
    }
  },
  {
    "ph": "f", "id": 272173, "pid": 76337, "tid": -914061504, "ts": 1716454225806937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806937, "dur": 0,
    "args": {
      "External id": 272174,
      "cbid": 11, "correlation": 272174
    }
  },
  {
    "ph": "f", "id": 272174, "pid": 76337, "tid": -914061504, "ts": 1716454225806937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225806953, "dur": 1,
    "args": {
      "External id": 272177,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272177,
      "bytes": 4, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 272177, "pid": 5, "tid": 7, "ts": 1716454225806953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806938, "dur": 24,
    "args": {
      "External id": 272177,
      "cbid": 41, "correlation": 272177
    }
  },
  {
    "ph": "s", "id": 272177, "pid": 76337, "tid": -914061504, "ts": 1716454225806938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806963, "dur": 3,
    "args": {
      "External id": 272178,
      "cbid": 131, "correlation": 272178
    }
  },
  {
    "ph": "f", "id": 272178, "pid": 76337, "tid": -914061504, "ts": 1716454225806963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225807001, "dur": 3,
    "args": {
      "External id": 272202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272202, "pid": 5, "tid": 7, "ts": 1716454225807001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225806991, "dur": 11,
    "args": {
      "External id": 272202,
      "cbid": 211, "correlation": 272202
    }
  },
  {
    "ph": "s", "id": 272202, "pid": 76337, "tid": -914061504, "ts": 1716454225806991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807002, "dur": 0,
    "args": {
      "External id": 272203,
      "cbid": 11, "correlation": 272203
    }
  },
  {
    "ph": "f", "id": 272203, "pid": 76337, "tid": -914061504, "ts": 1716454225807002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807003, "dur": 0,
    "args": {
      "External id": 272204,
      "cbid": 11, "correlation": 272204
    }
  },
  {
    "ph": "f", "id": 272204, "pid": 76337, "tid": -914061504, "ts": 1716454225807003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807004, "dur": 1,
    "args": {
      "External id": 272206,
      "cbid": 200, "correlation": 272206
    }
  },
  {
    "ph": "f", "id": 272206, "pid": 76337, "tid": -914061504, "ts": 1716454225807004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454225807015, "dur": 4,
    "args": {
      "External id": 272208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272208,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272208, "pid": 5, "tid": 7, "ts": 1716454225807015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807007, "dur": 8,
    "args": {
      "External id": 272208,
      "cbid": 211, "correlation": 272208
    }
  },
  {
    "ph": "s", "id": 272208, "pid": 76337, "tid": -914061504, "ts": 1716454225807007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807016, "dur": 0,
    "args": {
      "External id": 272209,
      "cbid": 11, "correlation": 272209
    }
  },
  {
    "ph": "f", "id": 272209, "pid": 76337, "tid": -914061504, "ts": 1716454225807016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807016, "dur": 0,
    "args": {
      "External id": 272210,
      "cbid": 11, "correlation": 272210
    }
  },
  {
    "ph": "f", "id": 272210, "pid": 76337, "tid": -914061504, "ts": 1716454225807016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454225807057, "dur": 1,
    "args": {
      "External id": 272217,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272217,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 272217, "pid": 5, "tid": 7, "ts": 1716454225807057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807042, "dur": 23,
    "args": {
      "External id": 272217,
      "cbid": 41, "correlation": 272217
    }
  },
  {
    "ph": "s", "id": 272217, "pid": 76337, "tid": -914061504, "ts": 1716454225807042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807066, "dur": 3,
    "args": {
      "External id": 272218,
      "cbid": 131, "correlation": 272218
    }
  },
  {
    "ph": "f", "id": 272218, "pid": 76337, "tid": -914061504, "ts": 1716454225807066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454225807117, "dur": 1,
    "args": {
      "External id": 272228,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 272228,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 272228, "pid": 5, "tid": 7, "ts": 1716454225807117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807105, "dur": 9,
    "args": {
      "External id": 272228,
      "cbid": 41, "correlation": 272228
    }
  },
  {
    "ph": "s", "id": 272228, "pid": 76337, "tid": -914061504, "ts": 1716454225807105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807116, "dur": 7,
    "args": {
      "External id": 272229,
      "cbid": 131, "correlation": 272229
    }
  },
  {
    "ph": "f", "id": 272229, "pid": 76337, "tid": -914061504, "ts": 1716454225807116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225807186, "dur": 5,
    "args": {
      "External id": 272236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272236,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272236, "pid": 5, "tid": 7, "ts": 1716454225807186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807171, "dur": 15,
    "args": {
      "External id": 272236,
      "cbid": 211, "correlation": 272236
    }
  },
  {
    "ph": "s", "id": 272236, "pid": 76337, "tid": -914061504, "ts": 1716454225807171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225807257, "dur": 3,
    "args": {
      "External id": 272245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272245,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272245, "pid": 5, "tid": 7, "ts": 1716454225807257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807244, "dur": 13,
    "args": {
      "External id": 272245,
      "cbid": 211, "correlation": 272245
    }
  },
  {
    "ph": "s", "id": 272245, "pid": 76337, "tid": -914061504, "ts": 1716454225807244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225807293, "dur": 3,
    "args": {
      "External id": 272253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272253,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272253, "pid": 5, "tid": 7, "ts": 1716454225807293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807283, "dur": 9,
    "args": {
      "External id": 272253,
      "cbid": 211, "correlation": 272253
    }
  },
  {
    "ph": "s", "id": 272253, "pid": 76337, "tid": -914061504, "ts": 1716454225807283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225807324, "dur": 4,
    "args": {
      "External id": 272261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272261,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272261, "pid": 5, "tid": 7, "ts": 1716454225807324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807313, "dur": 11,
    "args": {
      "External id": 272261,
      "cbid": 211, "correlation": 272261
    }
  },
  {
    "ph": "s", "id": 272261, "pid": 76337, "tid": -914061504, "ts": 1716454225807313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225807354, "dur": 4,
    "args": {
      "External id": 272269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272269,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272269, "pid": 5, "tid": 7, "ts": 1716454225807354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807343, "dur": 10,
    "args": {
      "External id": 272269,
      "cbid": 211, "correlation": 272269
    }
  },
  {
    "ph": "s", "id": 272269, "pid": 76337, "tid": -914061504, "ts": 1716454225807343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225807381, "dur": 3,
    "args": {
      "External id": 272277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272277,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272277, "pid": 5, "tid": 7, "ts": 1716454225807381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807372, "dur": 8,
    "args": {
      "External id": 272277,
      "cbid": 211, "correlation": 272277
    }
  },
  {
    "ph": "s", "id": 272277, "pid": 76337, "tid": -914061504, "ts": 1716454225807372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225807407, "dur": 4,
    "args": {
      "External id": 272285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272285,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272285, "pid": 5, "tid": 7, "ts": 1716454225807407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807397, "dur": 10,
    "args": {
      "External id": 272285,
      "cbid": 211, "correlation": 272285
    }
  },
  {
    "ph": "s", "id": 272285, "pid": 76337, "tid": -914061504, "ts": 1716454225807397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225807429, "dur": 4,
    "args": {
      "External id": 272293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272293,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272293, "pid": 5, "tid": 7, "ts": 1716454225807429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807421, "dur": 7,
    "args": {
      "External id": 272293,
      "cbid": 211, "correlation": 272293
    }
  },
  {
    "ph": "s", "id": 272293, "pid": 76337, "tid": -914061504, "ts": 1716454225807421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225807448, "dur": 4,
    "args": {
      "External id": 272301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272301, "pid": 5, "tid": 7, "ts": 1716454225807448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807440, "dur": 7,
    "args": {
      "External id": 272301,
      "cbid": 211, "correlation": 272301
    }
  },
  {
    "ph": "s", "id": 272301, "pid": 76337, "tid": -914061504, "ts": 1716454225807440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225807469, "dur": 3,
    "args": {
      "External id": 272309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272309,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272309, "pid": 5, "tid": 7, "ts": 1716454225807469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807460, "dur": 7,
    "args": {
      "External id": 272309,
      "cbid": 211, "correlation": 272309
    }
  },
  {
    "ph": "s", "id": 272309, "pid": 76337, "tid": -914061504, "ts": 1716454225807460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225807528, "dur": 3,
    "args": {
      "External id": 272317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272317,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272317, "pid": 5, "tid": 7, "ts": 1716454225807528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807517, "dur": 11,
    "args": {
      "External id": 272317,
      "cbid": 211, "correlation": 272317
    }
  },
  {
    "ph": "s", "id": 272317, "pid": 76337, "tid": -914061504, "ts": 1716454225807517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225807555, "dur": 4,
    "args": {
      "External id": 272325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272325, "pid": 5, "tid": 7, "ts": 1716454225807555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807545, "dur": 9,
    "args": {
      "External id": 272325,
      "cbid": 211, "correlation": 272325
    }
  },
  {
    "ph": "s", "id": 272325, "pid": 76337, "tid": -914061504, "ts": 1716454225807545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225807578, "dur": 4,
    "args": {
      "External id": 272333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272333, "pid": 5, "tid": 7, "ts": 1716454225807578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807569, "dur": 7,
    "args": {
      "External id": 272333,
      "cbid": 211, "correlation": 272333
    }
  },
  {
    "ph": "s", "id": 272333, "pid": 76337, "tid": -914061504, "ts": 1716454225807569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225807599, "dur": 3,
    "args": {
      "External id": 272341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272341,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272341, "pid": 5, "tid": 7, "ts": 1716454225807599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225807591, "dur": 7,
    "args": {
      "External id": 272341,
      "cbid": 211, "correlation": 272341
    }
  },
  {
    "ph": "s", "id": 272341, "pid": 76337, "tid": -914061504, "ts": 1716454225807591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225808018, "dur": 5,
    "args": {
      "External id": 272350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272350,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272350, "pid": 5, "tid": 7, "ts": 1716454225808018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808001, "dur": 17,
    "args": {
      "External id": 272350,
      "cbid": 211, "correlation": 272350
    }
  },
  {
    "ph": "s", "id": 272350, "pid": 76337, "tid": -914061504, "ts": 1716454225808001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225808056, "dur": 5,
    "args": {
      "External id": 272359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272359,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272359, "pid": 5, "tid": 7, "ts": 1716454225808056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808045, "dur": 10,
    "args": {
      "External id": 272359,
      "cbid": 211, "correlation": 272359
    }
  },
  {
    "ph": "s", "id": 272359, "pid": 76337, "tid": -914061504, "ts": 1716454225808045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454225808197, "dur": 3,
    "args": {
      "External id": 272375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272375,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272375, "pid": 5, "tid": 7, "ts": 1716454225808197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808182, "dur": 16,
    "args": {
      "External id": 272375,
      "cbid": 211, "correlation": 272375
    }
  },
  {
    "ph": "s", "id": 272375, "pid": 76337, "tid": -914061504, "ts": 1716454225808182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808230, "dur": 3,
    "args": {
      "External id": 272383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272383, "pid": 5, "tid": 7, "ts": 1716454225808230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808220, "dur": 9,
    "args": {
      "External id": 272383,
      "cbid": 211, "correlation": 272383
    }
  },
  {
    "ph": "s", "id": 272383, "pid": 76337, "tid": -914061504, "ts": 1716454225808220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808263, "dur": 3,
    "args": {
      "External id": 272391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272391, "pid": 5, "tid": 7, "ts": 1716454225808263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808254, "dur": 8,
    "args": {
      "External id": 272391,
      "cbid": 211, "correlation": 272391
    }
  },
  {
    "ph": "s", "id": 272391, "pid": 76337, "tid": -914061504, "ts": 1716454225808254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808294, "dur": 4,
    "args": {
      "External id": 272399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272399,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272399, "pid": 5, "tid": 7, "ts": 1716454225808294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808285, "dur": 8,
    "args": {
      "External id": 272399,
      "cbid": 211, "correlation": 272399
    }
  },
  {
    "ph": "s", "id": 272399, "pid": 76337, "tid": -914061504, "ts": 1716454225808285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454225808354, "dur": 4,
    "args": {
      "External id": 272411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272411,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272411, "pid": 5, "tid": 7, "ts": 1716454225808354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808340, "dur": 14,
    "args": {
      "External id": 272411,
      "cbid": 211, "correlation": 272411
    }
  },
  {
    "ph": "s", "id": 272411, "pid": 76337, "tid": -914061504, "ts": 1716454225808340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225808403, "dur": 4,
    "args": {
      "External id": 272422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272422, "pid": 5, "tid": 7, "ts": 1716454225808403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808391, "dur": 12,
    "args": {
      "External id": 272422,
      "cbid": 211, "correlation": 272422
    }
  },
  {
    "ph": "s", "id": 272422, "pid": 76337, "tid": -914061504, "ts": 1716454225808391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808434, "dur": 3,
    "args": {
      "External id": 272430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272430, "pid": 5, "tid": 7, "ts": 1716454225808434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808424, "dur": 9,
    "args": {
      "External id": 272430,
      "cbid": 211, "correlation": 272430
    }
  },
  {
    "ph": "s", "id": 272430, "pid": 76337, "tid": -914061504, "ts": 1716454225808424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808468, "dur": 5,
    "args": {
      "External id": 272438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272438,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272438, "pid": 5, "tid": 7, "ts": 1716454225808468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808457, "dur": 11,
    "args": {
      "External id": 272438,
      "cbid": 211, "correlation": 272438
    }
  },
  {
    "ph": "s", "id": 272438, "pid": 76337, "tid": -914061504, "ts": 1716454225808457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808498, "dur": 5,
    "args": {
      "External id": 272446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272446,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272446, "pid": 5, "tid": 7, "ts": 1716454225808498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808488, "dur": 10,
    "args": {
      "External id": 272446,
      "cbid": 211, "correlation": 272446
    }
  },
  {
    "ph": "s", "id": 272446, "pid": 76337, "tid": -914061504, "ts": 1716454225808488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225808528, "dur": 4,
    "args": {
      "External id": 272455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272455,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272455, "pid": 5, "tid": 7, "ts": 1716454225808528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808518, "dur": 10,
    "args": {
      "External id": 272455,
      "cbid": 211, "correlation": 272455
    }
  },
  {
    "ph": "s", "id": 272455, "pid": 76337, "tid": -914061504, "ts": 1716454225808518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225808590, "dur": 4,
    "args": {
      "External id": 272468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272468,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272468, "pid": 5, "tid": 7, "ts": 1716454225808590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808577, "dur": 13,
    "args": {
      "External id": 272468,
      "cbid": 211, "correlation": 272468
    }
  },
  {
    "ph": "s", "id": 272468, "pid": 76337, "tid": -914061504, "ts": 1716454225808577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454225808631, "dur": 5,
    "args": {
      "External id": 272478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272478,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 272478, "pid": 5, "tid": 7, "ts": 1716454225808631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808619, "dur": 11,
    "args": {
      "External id": 272478,
      "cbid": 211, "correlation": 272478
    }
  },
  {
    "ph": "s", "id": 272478, "pid": 76337, "tid": -914061504, "ts": 1716454225808619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808761, "dur": 8,
    "args": {
      "External id": 272495,
      "cbid": 251, "correlation": 272495
    }
  },
  {
    "ph": "f", "id": 272495, "pid": 76337, "tid": -914061504, "ts": 1716454225808761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454225808796, "dur": 11,
    "args": {
      "External id": 272497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272497,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272497, "pid": 5, "tid": 7, "ts": 1716454225808796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808778, "dur": 20,
    "args": {
      "External id": 272497,
      "cbid": 211, "correlation": 272497
    }
  },
  {
    "ph": "s", "id": 272497, "pid": 76337, "tid": -914061504, "ts": 1716454225808778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225808861, "dur": 4,
    "args": {
      "External id": 272505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272505,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272505, "pid": 5, "tid": 7, "ts": 1716454225808861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808847, "dur": 14,
    "args": {
      "External id": 272505,
      "cbid": 211, "correlation": 272505
    }
  },
  {
    "ph": "s", "id": 272505, "pid": 76337, "tid": -914061504, "ts": 1716454225808847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808922, "dur": 5,
    "args": {
      "External id": 272521,
      "cbid": 251, "correlation": 272521
    }
  },
  {
    "ph": "f", "id": 272521, "pid": 76337, "tid": -914061504, "ts": 1716454225808922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808931, "dur": 0,
    "args": {
      "External id": 272523,
      "cbid": 251, "correlation": 272523
    }
  },
  {
    "ph": "f", "id": 272523, "pid": 76337, "tid": -914061504, "ts": 1716454225808931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225808950, "dur": 14,
    "args": {
      "External id": 272524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272524,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272524, "pid": 5, "tid": 7, "ts": 1716454225808950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808935, "dur": 15,
    "args": {
      "External id": 272524,
      "cbid": 211, "correlation": 272524
    }
  },
  {
    "ph": "s", "id": 272524, "pid": 76337, "tid": -914061504, "ts": 1716454225808935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225808965, "dur": 5,
    "args": {
      "External id": 272526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272526,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272526, "pid": 5, "tid": 7, "ts": 1716454225808965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225808954, "dur": 9,
    "args": {
      "External id": 272526,
      "cbid": 211, "correlation": 272526
    }
  },
  {
    "ph": "s", "id": 272526, "pid": 76337, "tid": -914061504, "ts": 1716454225808954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809078, "dur": 1,
    "args": {
      "External id": 272536,
      "cbid": 317, "correlation": 272536
    }
  },
  {
    "ph": "f", "id": 272536, "pid": 76337, "tid": -914061504, "ts": 1716454225809078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809080, "dur": 1,
    "args": {
      "External id": 272537,
      "cbid": 203, "correlation": 272537
    }
  },
  {
    "ph": "f", "id": 272537, "pid": 76337, "tid": -914061504, "ts": 1716454225809080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809082, "dur": 1,
    "args": {
      "External id": 272538,
      "cbid": 205, "correlation": 272538
    }
  },
  {
    "ph": "f", "id": 272538, "pid": 76337, "tid": -914061504, "ts": 1716454225809082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225809143, "dur": 7,
    "args": {
      "External id": 272542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272542, "pid": 5, "tid": 7, "ts": 1716454225809143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809126, "dur": 17,
    "args": {
      "External id": 272542,
      "cbid": 211, "correlation": 272542
    }
  },
  {
    "ph": "s", "id": 272542, "pid": 76337, "tid": -914061504, "ts": 1716454225809126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225809153, "dur": 4,
    "args": {
      "External id": 272544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272544,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 272544, "pid": 5, "tid": 7, "ts": 1716454225809153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809146, "dur": 6,
    "args": {
      "External id": 272544,
      "cbid": 211, "correlation": 272544
    }
  },
  {
    "ph": "s", "id": 272544, "pid": 76337, "tid": -914061504, "ts": 1716454225809146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225809174, "dur": 3,
    "args": {
      "External id": 272546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272546, "pid": 5, "tid": 7, "ts": 1716454225809174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809164, "dur": 8,
    "args": {
      "External id": 272546,
      "cbid": 211, "correlation": 272546
    }
  },
  {
    "ph": "s", "id": 272546, "pid": 76337, "tid": -914061504, "ts": 1716454225809164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809179, "dur": 0,
    "args": {
      "External id": 272547,
      "cbid": 51, "correlation": 272547
    }
  },
  {
    "ph": "s", "id": 272547, "pid": 76337, "tid": -914061504, "ts": 1716454225809179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225809192, "dur": 90,
    "args": {
      "External id": 272548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272548,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272548, "pid": 5, "tid": 7, "ts": 1716454225809192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809182, "dur": 8,
    "args": {
      "External id": 272548,
      "cbid": 211, "correlation": 272548
    }
  },
  {
    "ph": "s", "id": 272548, "pid": 76337, "tid": -914061504, "ts": 1716454225809182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225809283, "dur": 60,
    "args": {
      "External id": 272553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272553, "pid": 5, "tid": 7, "ts": 1716454225809283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225809221, "dur": 11,
    "args": {
      "External id": 272553,
      "cbid": 211, "correlation": 272553
    }
  },
  {
    "ph": "s", "id": 272553, "pid": 76337, "tid": -914061504, "ts": 1716454225809221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225811067, "dur": 51,
    "args": {
      "External id": 272573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272573,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 272573, "pid": 5, "tid": 7, "ts": 1716454225811067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811050, "dur": 17,
    "args": {
      "External id": 272573,
      "cbid": 211, "correlation": 272573
    }
  },
  {
    "ph": "s", "id": 272573, "pid": 76337, "tid": -914061504, "ts": 1716454225811050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225811120, "dur": 5,
    "args": {
      "External id": 272585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272585,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272585, "pid": 5, "tid": 7, "ts": 1716454225811120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811082, "dur": 9,
    "args": {
      "External id": 272585,
      "cbid": 211, "correlation": 272585
    }
  },
  {
    "ph": "s", "id": 272585, "pid": 76337, "tid": -914061504, "ts": 1716454225811082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225811126, "dur": 58,
    "args": {
      "External id": 272588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272588,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272588, "pid": 5, "tid": 7, "ts": 1716454225811126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811108, "dur": 7,
    "args": {
      "External id": 272588,
      "cbid": 211, "correlation": 272588
    }
  },
  {
    "ph": "s", "id": 272588, "pid": 76337, "tid": -914061504, "ts": 1716454225811108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225811185, "dur": 37,
    "args": {
      "External id": 272597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272597,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272597, "pid": 5, "tid": 7, "ts": 1716454225811185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811155, "dur": 10,
    "args": {
      "External id": 272597,
      "cbid": 211, "correlation": 272597
    }
  },
  {
    "ph": "s", "id": 272597, "pid": 76337, "tid": -914061504, "ts": 1716454225811155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811215, "dur": 0,
    "args": {
      "External id": 272607,
      "cbid": 317, "correlation": 272607
    }
  },
  {
    "ph": "f", "id": 272607, "pid": 76337, "tid": -914061504, "ts": 1716454225811215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811215, "dur": 0,
    "args": {
      "External id": 272608,
      "cbid": 203, "correlation": 272608
    }
  },
  {
    "ph": "f", "id": 272608, "pid": 76337, "tid": -914061504, "ts": 1716454225811215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811216, "dur": 0,
    "args": {
      "External id": 272609,
      "cbid": 205, "correlation": 272609
    }
  },
  {
    "ph": "f", "id": 272609, "pid": 76337, "tid": -914061504, "ts": 1716454225811216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225811246, "dur": 40,
    "args": {
      "External id": 272613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272613,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272613, "pid": 5, "tid": 7, "ts": 1716454225811246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811233, "dur": 13,
    "args": {
      "External id": 272613,
      "cbid": 211, "correlation": 272613
    }
  },
  {
    "ph": "s", "id": 272613, "pid": 76337, "tid": -914061504, "ts": 1716454225811233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225811288, "dur": 15,
    "args": {
      "External id": 272615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272615,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272615, "pid": 5, "tid": 7, "ts": 1716454225811288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811249, "dur": 6,
    "args": {
      "External id": 272615,
      "cbid": 211, "correlation": 272615
    }
  },
  {
    "ph": "s", "id": 272615, "pid": 76337, "tid": -914061504, "ts": 1716454225811249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225811305, "dur": 3,
    "args": {
      "External id": 272617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272617,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272617, "pid": 5, "tid": 7, "ts": 1716454225811305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811261, "dur": 6,
    "args": {
      "External id": 272617,
      "cbid": 211, "correlation": 272617
    }
  },
  {
    "ph": "s", "id": 272617, "pid": 76337, "tid": -914061504, "ts": 1716454225811261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811271, "dur": 0,
    "args": {
      "External id": 272618,
      "cbid": 51, "correlation": 272618
    }
  },
  {
    "ph": "s", "id": 272618, "pid": 76337, "tid": -914061504, "ts": 1716454225811271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225811309, "dur": 720,
    "args": {
      "External id": 272619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272619,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272619, "pid": 5, "tid": 7, "ts": 1716454225811309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811274, "dur": 7,
    "args": {
      "External id": 272619,
      "cbid": 211, "correlation": 272619
    }
  },
  {
    "ph": "s", "id": 272619, "pid": 76337, "tid": -914061504, "ts": 1716454225811274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225812031, "dur": 60,
    "args": {
      "External id": 272624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272624, "pid": 5, "tid": 7, "ts": 1716454225812031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811304, "dur": 9,
    "args": {
      "External id": 272624,
      "cbid": 211, "correlation": 272624
    }
  },
  {
    "ph": "s", "id": 272624, "pid": 76337, "tid": -914061504, "ts": 1716454225811304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225812092, "dur": 4,
    "args": {
      "External id": 272632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272632,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272632, "pid": 5, "tid": 7, "ts": 1716454225812092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811349, "dur": 9,
    "args": {
      "External id": 272632,
      "cbid": 211, "correlation": 272632
    }
  },
  {
    "ph": "s", "id": 272632, "pid": 76337, "tid": -914061504, "ts": 1716454225811349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811418, "dur": 2,
    "args": {
      "External id": 272648,
      "cbid": 251, "correlation": 272648
    }
  },
  {
    "ph": "f", "id": 272648, "pid": 76337, "tid": -914061504, "ts": 1716454225811418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811424, "dur": 0,
    "args": {
      "External id": 272650,
      "cbid": 251, "correlation": 272650
    }
  },
  {
    "ph": "f", "id": 272650, "pid": 76337, "tid": -914061504, "ts": 1716454225811424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225812098, "dur": 9,
    "args": {
      "External id": 272651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272651,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 272651, "pid": 5, "tid": 7, "ts": 1716454225812098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811426, "dur": 12,
    "args": {
      "External id": 272651,
      "cbid": 211, "correlation": 272651
    }
  },
  {
    "ph": "s", "id": 272651, "pid": 76337, "tid": -914061504, "ts": 1716454225811426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225812108, "dur": 4,
    "args": {
      "External id": 272653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272653,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 272653, "pid": 5, "tid": 7, "ts": 1716454225812108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811440, "dur": 6,
    "args": {
      "External id": 272653,
      "cbid": 211, "correlation": 272653
    }
  },
  {
    "ph": "s", "id": 272653, "pid": 76337, "tid": -914061504, "ts": 1716454225811440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225812114, "dur": 54,
    "args": {
      "External id": 272663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272663,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272663, "pid": 5, "tid": 7, "ts": 1716454225812114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811500, "dur": 12,
    "args": {
      "External id": 272663,
      "cbid": 211, "correlation": 272663
    }
  },
  {
    "ph": "s", "id": 272663, "pid": 76337, "tid": -914061504, "ts": 1716454225811500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225812169, "dur": 53,
    "args": {
      "External id": 272683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272683,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 272683, "pid": 5, "tid": 7, "ts": 1716454225812169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811569, "dur": 11,
    "args": {
      "External id": 272683,
      "cbid": 211, "correlation": 272683
    }
  },
  {
    "ph": "s", "id": 272683, "pid": 76337, "tid": -914061504, "ts": 1716454225811569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225812224, "dur": 4,
    "args": {
      "External id": 272695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272695,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272695, "pid": 5, "tid": 7, "ts": 1716454225812224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811590, "dur": 7,
    "args": {
      "External id": 272695,
      "cbid": 211, "correlation": 272695
    }
  },
  {
    "ph": "s", "id": 272695, "pid": 76337, "tid": -914061504, "ts": 1716454225811590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225812229, "dur": 56,
    "args": {
      "External id": 272698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272698, "pid": 5, "tid": 7, "ts": 1716454225812229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811609, "dur": 7,
    "args": {
      "External id": 272698,
      "cbid": 211, "correlation": 272698
    }
  },
  {
    "ph": "s", "id": 272698, "pid": 76337, "tid": -914061504, "ts": 1716454225811609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225812287, "dur": 36,
    "args": {
      "External id": 272707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272707,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272707, "pid": 5, "tid": 7, "ts": 1716454225812287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811650, "dur": 9,
    "args": {
      "External id": 272707,
      "cbid": 211, "correlation": 272707
    }
  },
  {
    "ph": "s", "id": 272707, "pid": 76337, "tid": -914061504, "ts": 1716454225811650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811719, "dur": 0,
    "args": {
      "External id": 272717,
      "cbid": 317, "correlation": 272717
    }
  },
  {
    "ph": "f", "id": 272717, "pid": 76337, "tid": -914061504, "ts": 1716454225811719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811719, "dur": 0,
    "args": {
      "External id": 272718,
      "cbid": 203, "correlation": 272718
    }
  },
  {
    "ph": "f", "id": 272718, "pid": 76337, "tid": -914061504, "ts": 1716454225811719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811720, "dur": 0,
    "args": {
      "External id": 272719,
      "cbid": 205, "correlation": 272719
    }
  },
  {
    "ph": "f", "id": 272719, "pid": 76337, "tid": -914061504, "ts": 1716454225811720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225812324, "dur": 40,
    "args": {
      "External id": 272723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272723,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272723, "pid": 5, "tid": 7, "ts": 1716454225812324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811734, "dur": 12,
    "args": {
      "External id": 272723,
      "cbid": 211, "correlation": 272723
    }
  },
  {
    "ph": "s", "id": 272723, "pid": 76337, "tid": -914061504, "ts": 1716454225811734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225812366, "dur": 15,
    "args": {
      "External id": 272725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272725,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272725, "pid": 5, "tid": 7, "ts": 1716454225812366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811749, "dur": 5,
    "args": {
      "External id": 272725,
      "cbid": 211, "correlation": 272725
    }
  },
  {
    "ph": "s", "id": 272725, "pid": 76337, "tid": -914061504, "ts": 1716454225811749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225812382, "dur": 3,
    "args": {
      "External id": 272727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272727,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272727, "pid": 5, "tid": 7, "ts": 1716454225812382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811758, "dur": 5,
    "args": {
      "External id": 272727,
      "cbid": 211, "correlation": 272727
    }
  },
  {
    "ph": "s", "id": 272727, "pid": 76337, "tid": -914061504, "ts": 1716454225811758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811766, "dur": 0,
    "args": {
      "External id": 272728,
      "cbid": 51, "correlation": 272728
    }
  },
  {
    "ph": "s", "id": 272728, "pid": 76337, "tid": -914061504, "ts": 1716454225811766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225812386, "dur": 715,
    "args": {
      "External id": 272729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272729,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272729, "pid": 5, "tid": 7, "ts": 1716454225812386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811767, "dur": 5,
    "args": {
      "External id": 272729,
      "cbid": 211, "correlation": 272729
    }
  },
  {
    "ph": "s", "id": 272729, "pid": 76337, "tid": -914061504, "ts": 1716454225811767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225813103, "dur": 60,
    "args": {
      "External id": 272734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272734,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272734, "pid": 5, "tid": 7, "ts": 1716454225813103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811795, "dur": 8,
    "args": {
      "External id": 272734,
      "cbid": 211, "correlation": 272734
    }
  },
  {
    "ph": "s", "id": 272734, "pid": 76337, "tid": -914061504, "ts": 1716454225811795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225813164, "dur": 50,
    "args": {
      "External id": 272742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272742,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272742, "pid": 5, "tid": 7, "ts": 1716454225813164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811827, "dur": 9,
    "args": {
      "External id": 272742,
      "cbid": 211, "correlation": 272742
    }
  },
  {
    "ph": "s", "id": 272742, "pid": 76337, "tid": -914061504, "ts": 1716454225811827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225813216, "dur": 36,
    "args": {
      "External id": 272750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272750,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272750, "pid": 5, "tid": 7, "ts": 1716454225813216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811859, "dur": 9,
    "args": {
      "External id": 272750,
      "cbid": 211, "correlation": 272750
    }
  },
  {
    "ph": "s", "id": 272750, "pid": 76337, "tid": -914061504, "ts": 1716454225811859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225813253, "dur": 52,
    "args": {
      "External id": 272770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272770,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 272770, "pid": 5, "tid": 7, "ts": 1716454225813253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811938, "dur": 13,
    "args": {
      "External id": 272770,
      "cbid": 211, "correlation": 272770
    }
  },
  {
    "ph": "s", "id": 272770, "pid": 76337, "tid": -914061504, "ts": 1716454225811938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225813306, "dur": 4,
    "args": {
      "External id": 272782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272782,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 272782, "pid": 5, "tid": 7, "ts": 1716454225813306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811960, "dur": 6,
    "args": {
      "External id": 272782,
      "cbid": 211, "correlation": 272782
    }
  },
  {
    "ph": "s", "id": 272782, "pid": 76337, "tid": -914061504, "ts": 1716454225811960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225813312, "dur": 57,
    "args": {
      "External id": 272785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272785, "pid": 5, "tid": 7, "ts": 1716454225813312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225811986, "dur": 8,
    "args": {
      "External id": 272785,
      "cbid": 211, "correlation": 272785
    }
  },
  {
    "ph": "s", "id": 272785, "pid": 76337, "tid": -914061504, "ts": 1716454225811986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812047, "dur": 0,
    "args": {
      "External id": 272796,
      "cbid": 317, "correlation": 272796
    }
  },
  {
    "ph": "f", "id": 272796, "pid": 76337, "tid": -914061504, "ts": 1716454225812047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812048, "dur": 0,
    "args": {
      "External id": 272797,
      "cbid": 203, "correlation": 272797
    }
  },
  {
    "ph": "f", "id": 272797, "pid": 76337, "tid": -914061504, "ts": 1716454225812048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812049, "dur": 0,
    "args": {
      "External id": 272798,
      "cbid": 205, "correlation": 272798
    }
  },
  {
    "ph": "f", "id": 272798, "pid": 76337, "tid": -914061504, "ts": 1716454225812049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812086, "dur": 4,
    "args": {
      "External id": 272802,
      "cbid": 251, "correlation": 272802
    }
  },
  {
    "ph": "f", "id": 272802, "pid": 76337, "tid": -914061504, "ts": 1716454225812086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812091, "dur": 1,
    "args": {
      "External id": 272803,
      "cbid": 251, "correlation": 272803
    }
  },
  {
    "ph": "f", "id": 272803, "pid": 76337, "tid": -914061504, "ts": 1716454225812091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812094, "dur": 2,
    "args": {
      "External id": 272804,
      "cbid": 251, "correlation": 272804
    }
  },
  {
    "ph": "f", "id": 272804, "pid": 76337, "tid": -914061504, "ts": 1716454225812094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812097, "dur": 2,
    "args": {
      "External id": 272805,
      "cbid": 251, "correlation": 272805
    }
  },
  {
    "ph": "f", "id": 272805, "pid": 76337, "tid": -914061504, "ts": 1716454225812097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812099, "dur": 1,
    "args": {
      "External id": 272806,
      "cbid": 251, "correlation": 272806
    }
  },
  {
    "ph": "f", "id": 272806, "pid": 76337, "tid": -914061504, "ts": 1716454225812099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812101, "dur": 2,
    "args": {
      "External id": 272807,
      "cbid": 251, "correlation": 272807
    }
  },
  {
    "ph": "f", "id": 272807, "pid": 76337, "tid": -914061504, "ts": 1716454225812101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812105, "dur": 1,
    "args": {
      "External id": 272808,
      "cbid": 251, "correlation": 272808
    }
  },
  {
    "ph": "f", "id": 272808, "pid": 76337, "tid": -914061504, "ts": 1716454225812105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812106, "dur": 1,
    "args": {
      "External id": 272809,
      "cbid": 251, "correlation": 272809
    }
  },
  {
    "ph": "f", "id": 272809, "pid": 76337, "tid": -914061504, "ts": 1716454225812106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812109, "dur": 0,
    "args": {
      "External id": 272810,
      "cbid": 251, "correlation": 272810
    }
  },
  {
    "ph": "f", "id": 272810, "pid": 76337, "tid": -914061504, "ts": 1716454225812109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225813370, "dur": 117,
    "args": {
      "External id": 272811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272811,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 272811, "pid": 5, "tid": 7, "ts": 1716454225813370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812113, "dur": 15,
    "args": {
      "External id": 272811,
      "cbid": 211, "correlation": 272811
    }
  },
  {
    "ph": "s", "id": 272811, "pid": 76337, "tid": -914061504, "ts": 1716454225812113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225813489, "dur": 62,
    "args": {
      "External id": 272817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272817,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272817, "pid": 5, "tid": 7, "ts": 1716454225813489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812153, "dur": 9,
    "args": {
      "External id": 272817,
      "cbid": 211, "correlation": 272817
    }
  },
  {
    "ph": "s", "id": 272817, "pid": 76337, "tid": -914061504, "ts": 1716454225812153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225813552, "dur": 612,
    "args": {
      "External id": 272826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272826,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272826, "pid": 5, "tid": 7, "ts": 1716454225813552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812248, "dur": 16,
    "args": {
      "External id": 272826,
      "cbid": 211, "correlation": 272826
    }
  },
  {
    "ph": "s", "id": 272826, "pid": 76337, "tid": -914061504, "ts": 1716454225812248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225814165, "dur": 186,
    "args": {
      "External id": 272848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272848,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272848, "pid": 5, "tid": 7, "ts": 1716454225814165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812324, "dur": 12,
    "args": {
      "External id": 272848,
      "cbid": 211, "correlation": 272848
    }
  },
  {
    "ph": "s", "id": 272848, "pid": 76337, "tid": -914061504, "ts": 1716454225812324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812439, "dur": 2,
    "args": {
      "External id": 272859,
      "cbid": 251, "correlation": 272859
    }
  },
  {
    "ph": "f", "id": 272859, "pid": 76337, "tid": -914061504, "ts": 1716454225812439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225814353, "dur": 198,
    "args": {
      "External id": 272860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272860,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272860, "pid": 5, "tid": 7, "ts": 1716454225814353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812446, "dur": 14,
    "args": {
      "External id": 272860,
      "cbid": 211, "correlation": 272860
    }
  },
  {
    "ph": "s", "id": 272860, "pid": 76337, "tid": -914061504, "ts": 1716454225812446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812518, "dur": 1,
    "args": {
      "External id": 272871,
      "cbid": 251, "correlation": 272871
    }
  },
  {
    "ph": "f", "id": 272871, "pid": 76337, "tid": -914061504, "ts": 1716454225812518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225814553, "dur": 196,
    "args": {
      "External id": 272872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272872,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272872, "pid": 5, "tid": 7, "ts": 1716454225814553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812522, "dur": 11,
    "args": {
      "External id": 272872,
      "cbid": 211, "correlation": 272872
    }
  },
  {
    "ph": "s", "id": 272872, "pid": 76337, "tid": -914061504, "ts": 1716454225812522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812591, "dur": 1,
    "args": {
      "External id": 272883,
      "cbid": 251, "correlation": 272883
    }
  },
  {
    "ph": "f", "id": 272883, "pid": 76337, "tid": -914061504, "ts": 1716454225812591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225814751, "dur": 192,
    "args": {
      "External id": 272884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272884,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272884, "pid": 5, "tid": 7, "ts": 1716454225814751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812595, "dur": 12,
    "args": {
      "External id": 272884,
      "cbid": 211, "correlation": 272884
    }
  },
  {
    "ph": "s", "id": 272884, "pid": 76337, "tid": -914061504, "ts": 1716454225812595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225814944, "dur": 19208,
    "args": {
      "External id": 272905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272905,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 272905, "pid": 5, "tid": 7, "ts": 1716454225814944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812706, "dur": 14,
    "args": {
      "External id": 272905,
      "cbid": 211, "correlation": 272905
    }
  },
  {
    "ph": "s", "id": 272905, "pid": 76337, "tid": -914061504, "ts": 1716454225812706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812816, "dur": 2,
    "args": {
      "External id": 272923,
      "cbid": 251, "correlation": 272923
    }
  },
  {
    "ph": "f", "id": 272923, "pid": 76337, "tid": -914061504, "ts": 1716454225812816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225834153, "dur": 208,
    "args": {
      "External id": 272925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272925,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272925, "pid": 5, "tid": 7, "ts": 1716454225834153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812823, "dur": 13,
    "args": {
      "External id": 272925,
      "cbid": 211, "correlation": 272925
    }
  },
  {
    "ph": "s", "id": 272925, "pid": 76337, "tid": -914061504, "ts": 1716454225812823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225834363, "dur": 67,
    "args": {
      "External id": 272933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272933,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272933, "pid": 5, "tid": 7, "ts": 1716454225834363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812894, "dur": 13,
    "args": {
      "External id": 272933,
      "cbid": 211, "correlation": 272933
    }
  },
  {
    "ph": "s", "id": 272933, "pid": 76337, "tid": -914061504, "ts": 1716454225812894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225834431, "dur": 96,
    "args": {
      "External id": 272941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272941,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272941, "pid": 5, "tid": 7, "ts": 1716454225834431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225812934, "dur": 9,
    "args": {
      "External id": 272941,
      "cbid": 211, "correlation": 272941
    }
  },
  {
    "ph": "s", "id": 272941, "pid": 76337, "tid": -914061504, "ts": 1716454225812934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225834528, "dur": 55,
    "args": {
      "External id": 272952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272952,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272952, "pid": 5, "tid": 7, "ts": 1716454225834528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813030, "dur": 16,
    "args": {
      "External id": 272952,
      "cbid": 211, "correlation": 272952
    }
  },
  {
    "ph": "s", "id": 272952, "pid": 76337, "tid": -914061504, "ts": 1716454225813030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225834585, "dur": 95,
    "args": {
      "External id": 272974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272974,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 272974, "pid": 5, "tid": 7, "ts": 1716454225834585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813067, "dur": 8,
    "args": {
      "External id": 272974,
      "cbid": 211, "correlation": 272974
    }
  },
  {
    "ph": "s", "id": 272974, "pid": 76337, "tid": -914061504, "ts": 1716454225813067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813151, "dur": 1,
    "args": {
      "External id": 272985,
      "cbid": 251, "correlation": 272985
    }
  },
  {
    "ph": "f", "id": 272985, "pid": 76337, "tid": -914061504, "ts": 1716454225813151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225834681, "dur": 109,
    "args": {
      "External id": 272986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272986,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 272986, "pid": 5, "tid": 7, "ts": 1716454225834681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813156, "dur": 13,
    "args": {
      "External id": 272986,
      "cbid": 211, "correlation": 272986
    }
  },
  {
    "ph": "s", "id": 272986, "pid": 76337, "tid": -914061504, "ts": 1716454225813156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813239, "dur": 2,
    "args": {
      "External id": 272997,
      "cbid": 251, "correlation": 272997
    }
  },
  {
    "ph": "f", "id": 272997, "pid": 76337, "tid": -914061504, "ts": 1716454225813239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813243, "dur": 0,
    "args": {
      "External id": 272998,
      "cbid": 251, "correlation": 272998
    }
  },
  {
    "ph": "f", "id": 272998, "pid": 76337, "tid": -914061504, "ts": 1716454225813243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225834792, "dur": 10,
    "args": {
      "External id": 272999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 272999,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 272999, "pid": 5, "tid": 7, "ts": 1716454225834792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813245, "dur": 17,
    "args": {
      "External id": 272999,
      "cbid": 211, "correlation": 272999
    }
  },
  {
    "ph": "s", "id": 272999, "pid": 76337, "tid": -914061504, "ts": 1716454225813245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225834804, "dur": 5,
    "args": {
      "External id": 273001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273001,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 273001, "pid": 5, "tid": 7, "ts": 1716454225834804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813266, "dur": 8,
    "args": {
      "External id": 273001,
      "cbid": 211, "correlation": 273001
    }
  },
  {
    "ph": "s", "id": 273001, "pid": 76337, "tid": -914061504, "ts": 1716454225813266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813334, "dur": 1,
    "args": {
      "External id": 273012,
      "cbid": 251, "correlation": 273012
    }
  },
  {
    "ph": "f", "id": 273012, "pid": 76337, "tid": -914061504, "ts": 1716454225813334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813337, "dur": 0,
    "args": {
      "External id": 273013,
      "cbid": 251, "correlation": 273013
    }
  },
  {
    "ph": "f", "id": 273013, "pid": 76337, "tid": -914061504, "ts": 1716454225813337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225834810, "dur": 7,
    "args": {
      "External id": 273014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273014,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 273014, "pid": 5, "tid": 7, "ts": 1716454225834810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813339, "dur": 12,
    "args": {
      "External id": 273014,
      "cbid": 211, "correlation": 273014
    }
  },
  {
    "ph": "s", "id": 273014, "pid": 76337, "tid": -914061504, "ts": 1716454225813339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225834818, "dur": 4,
    "args": {
      "External id": 273016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273016,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 273016, "pid": 5, "tid": 7, "ts": 1716454225834818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813353, "dur": 6,
    "args": {
      "External id": 273016,
      "cbid": 211, "correlation": 273016
    }
  },
  {
    "ph": "s", "id": 273016, "pid": 76337, "tid": -914061504, "ts": 1716454225813353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225834823, "dur": 159,
    "args": {
      "External id": 273037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273037,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273037, "pid": 5, "tid": 7, "ts": 1716454225834823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813427, "dur": 13,
    "args": {
      "External id": 273037,
      "cbid": 211, "correlation": 273037
    }
  },
  {
    "ph": "s", "id": 273037, "pid": 76337, "tid": -914061504, "ts": 1716454225813427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813524, "dur": 2,
    "args": {
      "External id": 273055,
      "cbid": 251, "correlation": 273055
    }
  },
  {
    "ph": "f", "id": 273055, "pid": 76337, "tid": -914061504, "ts": 1716454225813524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225834984, "dur": 109,
    "args": {
      "External id": 273057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273057,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273057, "pid": 5, "tid": 7, "ts": 1716454225834984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813531, "dur": 15,
    "args": {
      "External id": 273057,
      "cbid": 211, "correlation": 273057
    }
  },
  {
    "ph": "s", "id": 273057, "pid": 76337, "tid": -914061504, "ts": 1716454225813531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225835094, "dur": 35,
    "args": {
      "External id": 273065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273065,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273065, "pid": 5, "tid": 7, "ts": 1716454225835094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813602, "dur": 12,
    "args": {
      "External id": 273065,
      "cbid": 211, "correlation": 273065
    }
  },
  {
    "ph": "s", "id": 273065, "pid": 76337, "tid": -914061504, "ts": 1716454225813602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225835130, "dur": 68,
    "args": {
      "External id": 273073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273073, "pid": 5, "tid": 7, "ts": 1716454225835130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813643, "dur": 9,
    "args": {
      "External id": 273073,
      "cbid": 211, "correlation": 273073
    }
  },
  {
    "ph": "s", "id": 273073, "pid": 76337, "tid": -914061504, "ts": 1716454225813643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225835200, "dur": 95,
    "args": {
      "External id": 273095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273095,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273095, "pid": 5, "tid": 7, "ts": 1716454225835200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813698, "dur": 10,
    "args": {
      "External id": 273095,
      "cbid": 211, "correlation": 273095
    }
  },
  {
    "ph": "s", "id": 273095, "pid": 76337, "tid": -914061504, "ts": 1716454225813698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813789, "dur": 1,
    "args": {
      "External id": 273111,
      "cbid": 251, "correlation": 273111
    }
  },
  {
    "ph": "f", "id": 273111, "pid": 76337, "tid": -914061504, "ts": 1716454225813789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225835296, "dur": 590,
    "args": {
      "External id": 273113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273113,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273113, "pid": 5, "tid": 7, "ts": 1716454225835296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813795, "dur": 12,
    "args": {
      "External id": 273113,
      "cbid": 211, "correlation": 273113
    }
  },
  {
    "ph": "s", "id": 273113, "pid": 76337, "tid": -914061504, "ts": 1716454225813795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225835888, "dur": 248,
    "args": {
      "External id": 273121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273121, "pid": 5, "tid": 7, "ts": 1716454225835888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813872, "dur": 15,
    "args": {
      "External id": 273121,
      "cbid": 211, "correlation": 273121
    }
  },
  {
    "ph": "s", "id": 273121, "pid": 76337, "tid": -914061504, "ts": 1716454225813872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225836137, "dur": 254,
    "args": {
      "External id": 273129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273129, "pid": 5, "tid": 7, "ts": 1716454225836137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225813908, "dur": 9,
    "args": {
      "External id": 273129,
      "cbid": 211, "correlation": 273129
    }
  },
  {
    "ph": "s", "id": 273129, "pid": 76337, "tid": -914061504, "ts": 1716454225813908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814002, "dur": 4,
    "args": {
      "External id": 273145,
      "cbid": 251, "correlation": 273145
    }
  },
  {
    "ph": "f", "id": 273145, "pid": 76337, "tid": -914061504, "ts": 1716454225814002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814009, "dur": 0,
    "args": {
      "External id": 273147,
      "cbid": 251, "correlation": 273147
    }
  },
  {
    "ph": "f", "id": 273147, "pid": 76337, "tid": -914061504, "ts": 1716454225814009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225836392, "dur": 362,
    "args": {
      "External id": 273148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273148,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 273148, "pid": 5, "tid": 7, "ts": 1716454225836392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814015, "dur": 14,
    "args": {
      "External id": 273148,
      "cbid": 211, "correlation": 273148
    }
  },
  {
    "ph": "s", "id": 273148, "pid": 76337, "tid": -914061504, "ts": 1716454225814015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225836755, "dur": 50,
    "args": {
      "External id": 273156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273156,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273156, "pid": 5, "tid": 7, "ts": 1716454225836755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814060, "dur": 10,
    "args": {
      "External id": 273156,
      "cbid": 211, "correlation": 273156
    }
  },
  {
    "ph": "s", "id": 273156, "pid": 76337, "tid": -914061504, "ts": 1716454225814060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225836806, "dur": 162,
    "args": {
      "External id": 273167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273167, "pid": 5, "tid": 7, "ts": 1716454225836806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814132, "dur": 12,
    "args": {
      "External id": 273167,
      "cbid": 211, "correlation": 273167
    }
  },
  {
    "ph": "s", "id": 273167, "pid": 76337, "tid": -914061504, "ts": 1716454225814132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814200, "dur": 0,
    "args": {
      "External id": 273179,
      "cbid": 317, "correlation": 273179
    }
  },
  {
    "ph": "f", "id": 273179, "pid": 76337, "tid": -914061504, "ts": 1716454225814200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814201, "dur": 0,
    "args": {
      "External id": 273180,
      "cbid": 203, "correlation": 273180
    }
  },
  {
    "ph": "f", "id": 273180, "pid": 76337, "tid": -914061504, "ts": 1716454225814201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814202, "dur": 0,
    "args": {
      "External id": 273181,
      "cbid": 205, "correlation": 273181
    }
  },
  {
    "ph": "f", "id": 273181, "pid": 76337, "tid": -914061504, "ts": 1716454225814202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814225, "dur": 1,
    "args": {
      "External id": 273185,
      "cbid": 251, "correlation": 273185
    }
  },
  {
    "ph": "f", "id": 273185, "pid": 76337, "tid": -914061504, "ts": 1716454225814225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814227, "dur": 0,
    "args": {
      "External id": 273186,
      "cbid": 251, "correlation": 273186
    }
  },
  {
    "ph": "f", "id": 273186, "pid": 76337, "tid": -914061504, "ts": 1716454225814227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814228, "dur": 0,
    "args": {
      "External id": 273187,
      "cbid": 251, "correlation": 273187
    }
  },
  {
    "ph": "f", "id": 273187, "pid": 76337, "tid": -914061504, "ts": 1716454225814228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814229, "dur": 0,
    "args": {
      "External id": 273188,
      "cbid": 251, "correlation": 273188
    }
  },
  {
    "ph": "f", "id": 273188, "pid": 76337, "tid": -914061504, "ts": 1716454225814229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814230, "dur": 0,
    "args": {
      "External id": 273189,
      "cbid": 251, "correlation": 273189
    }
  },
  {
    "ph": "f", "id": 273189, "pid": 76337, "tid": -914061504, "ts": 1716454225814230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814231, "dur": 0,
    "args": {
      "External id": 273190,
      "cbid": 251, "correlation": 273190
    }
  },
  {
    "ph": "f", "id": 273190, "pid": 76337, "tid": -914061504, "ts": 1716454225814231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814232, "dur": 0,
    "args": {
      "External id": 273191,
      "cbid": 251, "correlation": 273191
    }
  },
  {
    "ph": "f", "id": 273191, "pid": 76337, "tid": -914061504, "ts": 1716454225814232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814233, "dur": 0,
    "args": {
      "External id": 273192,
      "cbid": 251, "correlation": 273192
    }
  },
  {
    "ph": "f", "id": 273192, "pid": 76337, "tid": -914061504, "ts": 1716454225814233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814234, "dur": 0,
    "args": {
      "External id": 273193,
      "cbid": 251, "correlation": 273193
    }
  },
  {
    "ph": "f", "id": 273193, "pid": 76337, "tid": -914061504, "ts": 1716454225814234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225836970, "dur": 118,
    "args": {
      "External id": 273194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273194,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273194, "pid": 5, "tid": 7, "ts": 1716454225836970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814236, "dur": 13,
    "args": {
      "External id": 273194,
      "cbid": 211, "correlation": 273194
    }
  },
  {
    "ph": "s", "id": 273194, "pid": 76337, "tid": -914061504, "ts": 1716454225814236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225837089, "dur": 60,
    "args": {
      "External id": 273200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273200, "pid": 5, "tid": 7, "ts": 1716454225837089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814271, "dur": 9,
    "args": {
      "External id": 273200,
      "cbid": 211, "correlation": 273200
    }
  },
  {
    "ph": "s", "id": 273200, "pid": 76337, "tid": -914061504, "ts": 1716454225814271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225837151, "dur": 50,
    "args": {
      "External id": 273208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273208,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273208, "pid": 5, "tid": 7, "ts": 1716454225837151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814304, "dur": 8,
    "args": {
      "External id": 273208,
      "cbid": 211, "correlation": 273208
    }
  },
  {
    "ph": "s", "id": 273208, "pid": 76337, "tid": -914061504, "ts": 1716454225814304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225837202, "dur": 53,
    "args": {
      "External id": 273228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273228,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 273228, "pid": 5, "tid": 7, "ts": 1716454225837202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814380, "dur": 12,
    "args": {
      "External id": 273228,
      "cbid": 211, "correlation": 273228
    }
  },
  {
    "ph": "s", "id": 273228, "pid": 76337, "tid": -914061504, "ts": 1716454225814380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225837256, "dur": 5,
    "args": {
      "External id": 273240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273240,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273240, "pid": 5, "tid": 7, "ts": 1716454225837256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814402, "dur": 6,
    "args": {
      "External id": 273240,
      "cbid": 211, "correlation": 273240
    }
  },
  {
    "ph": "s", "id": 273240, "pid": 76337, "tid": -914061504, "ts": 1716454225814402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225837262, "dur": 58,
    "args": {
      "External id": 273243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273243, "pid": 5, "tid": 7, "ts": 1716454225837262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814420, "dur": 7,
    "args": {
      "External id": 273243,
      "cbid": 211, "correlation": 273243
    }
  },
  {
    "ph": "s", "id": 273243, "pid": 76337, "tid": -914061504, "ts": 1716454225814420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225837321, "dur": 38,
    "args": {
      "External id": 273252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273252,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273252, "pid": 5, "tid": 7, "ts": 1716454225837321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814458, "dur": 9,
    "args": {
      "External id": 273252,
      "cbid": 211, "correlation": 273252
    }
  },
  {
    "ph": "s", "id": 273252, "pid": 76337, "tid": -914061504, "ts": 1716454225814458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814509, "dur": 0,
    "args": {
      "External id": 273262,
      "cbid": 317, "correlation": 273262
    }
  },
  {
    "ph": "f", "id": 273262, "pid": 76337, "tid": -914061504, "ts": 1716454225814509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814510, "dur": 0,
    "args": {
      "External id": 273263,
      "cbid": 203, "correlation": 273263
    }
  },
  {
    "ph": "f", "id": 273263, "pid": 76337, "tid": -914061504, "ts": 1716454225814510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814511, "dur": 0,
    "args": {
      "External id": 273264,
      "cbid": 205, "correlation": 273264
    }
  },
  {
    "ph": "f", "id": 273264, "pid": 76337, "tid": -914061504, "ts": 1716454225814511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225837360, "dur": 41,
    "args": {
      "External id": 273268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273268, "pid": 5, "tid": 7, "ts": 1716454225837360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814526, "dur": 12,
    "args": {
      "External id": 273268,
      "cbid": 211, "correlation": 273268
    }
  },
  {
    "ph": "s", "id": 273268, "pid": 76337, "tid": -914061504, "ts": 1716454225814526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225837402, "dur": 14,
    "args": {
      "External id": 273270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273270, "pid": 5, "tid": 7, "ts": 1716454225837402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814540, "dur": 5,
    "args": {
      "External id": 273270,
      "cbid": 211, "correlation": 273270
    }
  },
  {
    "ph": "s", "id": 273270, "pid": 76337, "tid": -914061504, "ts": 1716454225814540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225837418, "dur": 4,
    "args": {
      "External id": 273272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273272, "pid": 5, "tid": 7, "ts": 1716454225837418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814549, "dur": 6,
    "args": {
      "External id": 273272,
      "cbid": 211, "correlation": 273272
    }
  },
  {
    "ph": "s", "id": 273272, "pid": 76337, "tid": -914061504, "ts": 1716454225814549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814558, "dur": 0,
    "args": {
      "External id": 273273,
      "cbid": 51, "correlation": 273273
    }
  },
  {
    "ph": "s", "id": 273273, "pid": 76337, "tid": -914061504, "ts": 1716454225814558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225837423, "dur": 716,
    "args": {
      "External id": 273274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273274,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273274, "pid": 5, "tid": 7, "ts": 1716454225837423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814559, "dur": 5,
    "args": {
      "External id": 273274,
      "cbid": 211, "correlation": 273274
    }
  },
  {
    "ph": "s", "id": 273274, "pid": 76337, "tid": -914061504, "ts": 1716454225814559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225838141, "dur": 60,
    "args": {
      "External id": 273279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273279, "pid": 5, "tid": 7, "ts": 1716454225838141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814587, "dur": 8,
    "args": {
      "External id": 273279,
      "cbid": 211, "correlation": 273279
    }
  },
  {
    "ph": "s", "id": 273279, "pid": 76337, "tid": -914061504, "ts": 1716454225814587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225838203, "dur": 5,
    "args": {
      "External id": 273287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273287,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273287, "pid": 5, "tid": 7, "ts": 1716454225838203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814630, "dur": 9,
    "args": {
      "External id": 273287,
      "cbid": 211, "correlation": 273287
    }
  },
  {
    "ph": "s", "id": 273287, "pid": 76337, "tid": -914061504, "ts": 1716454225814630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814696, "dur": 1,
    "args": {
      "External id": 273303,
      "cbid": 251, "correlation": 273303
    }
  },
  {
    "ph": "f", "id": 273303, "pid": 76337, "tid": -914061504, "ts": 1716454225814696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814701, "dur": 0,
    "args": {
      "External id": 273305,
      "cbid": 251, "correlation": 273305
    }
  },
  {
    "ph": "f", "id": 273305, "pid": 76337, "tid": -914061504, "ts": 1716454225814701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225838209, "dur": 11,
    "args": {
      "External id": 273306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273306,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 273306, "pid": 5, "tid": 7, "ts": 1716454225838209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814703, "dur": 11,
    "args": {
      "External id": 273306,
      "cbid": 211, "correlation": 273306
    }
  },
  {
    "ph": "s", "id": 273306, "pid": 76337, "tid": -914061504, "ts": 1716454225814703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225838222, "dur": 5,
    "args": {
      "External id": 273308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273308,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 273308, "pid": 5, "tid": 7, "ts": 1716454225838222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814716, "dur": 6,
    "args": {
      "External id": 273308,
      "cbid": 211, "correlation": 273308
    }
  },
  {
    "ph": "s", "id": 273308, "pid": 76337, "tid": -914061504, "ts": 1716454225814716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225838228, "dur": 55,
    "args": {
      "External id": 273318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273318, "pid": 5, "tid": 7, "ts": 1716454225838228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814773, "dur": 12,
    "args": {
      "External id": 273318,
      "cbid": 211, "correlation": 273318
    }
  },
  {
    "ph": "s", "id": 273318, "pid": 76337, "tid": -914061504, "ts": 1716454225814773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225838284, "dur": 51,
    "args": {
      "External id": 273338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273338,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 273338, "pid": 5, "tid": 7, "ts": 1716454225838284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814840, "dur": 10,
    "args": {
      "External id": 273338,
      "cbid": 211, "correlation": 273338
    }
  },
  {
    "ph": "s", "id": 273338, "pid": 76337, "tid": -914061504, "ts": 1716454225814840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225838337, "dur": 4,
    "args": {
      "External id": 273350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273350,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273350, "pid": 5, "tid": 7, "ts": 1716454225838337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814863, "dur": 6,
    "args": {
      "External id": 273350,
      "cbid": 211, "correlation": 273350
    }
  },
  {
    "ph": "s", "id": 273350, "pid": 76337, "tid": -914061504, "ts": 1716454225814863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225838342, "dur": 56,
    "args": {
      "External id": 273353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273353, "pid": 5, "tid": 7, "ts": 1716454225838342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814883, "dur": 6,
    "args": {
      "External id": 273353,
      "cbid": 211, "correlation": 273353
    }
  },
  {
    "ph": "s", "id": 273353, "pid": 76337, "tid": -914061504, "ts": 1716454225814883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225838400, "dur": 37,
    "args": {
      "External id": 273362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273362,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273362, "pid": 5, "tid": 7, "ts": 1716454225838400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814923, "dur": 10,
    "args": {
      "External id": 273362,
      "cbid": 211, "correlation": 273362
    }
  },
  {
    "ph": "s", "id": 273362, "pid": 76337, "tid": -914061504, "ts": 1716454225814923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814995, "dur": 0,
    "args": {
      "External id": 273372,
      "cbid": 317, "correlation": 273372
    }
  },
  {
    "ph": "f", "id": 273372, "pid": 76337, "tid": -914061504, "ts": 1716454225814995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814996, "dur": 0,
    "args": {
      "External id": 273373,
      "cbid": 203, "correlation": 273373
    }
  },
  {
    "ph": "f", "id": 273373, "pid": 76337, "tid": -914061504, "ts": 1716454225814996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225814997, "dur": 0,
    "args": {
      "External id": 273374,
      "cbid": 205, "correlation": 273374
    }
  },
  {
    "ph": "f", "id": 273374, "pid": 76337, "tid": -914061504, "ts": 1716454225814997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225838438, "dur": 40,
    "args": {
      "External id": 273378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273378, "pid": 5, "tid": 7, "ts": 1716454225838438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815010, "dur": 13,
    "args": {
      "External id": 273378,
      "cbid": 211, "correlation": 273378
    }
  },
  {
    "ph": "s", "id": 273378, "pid": 76337, "tid": -914061504, "ts": 1716454225815010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225838479, "dur": 14,
    "args": {
      "External id": 273380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273380,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273380, "pid": 5, "tid": 7, "ts": 1716454225838479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815026, "dur": 5,
    "args": {
      "External id": 273380,
      "cbid": 211, "correlation": 273380
    }
  },
  {
    "ph": "s", "id": 273380, "pid": 76337, "tid": -914061504, "ts": 1716454225815026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225838495, "dur": 4,
    "args": {
      "External id": 273382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273382, "pid": 5, "tid": 7, "ts": 1716454225838495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815035, "dur": 5,
    "args": {
      "External id": 273382,
      "cbid": 211, "correlation": 273382
    }
  },
  {
    "ph": "s", "id": 273382, "pid": 76337, "tid": -914061504, "ts": 1716454225815035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815044, "dur": 0,
    "args": {
      "External id": 273383,
      "cbid": 51, "correlation": 273383
    }
  },
  {
    "ph": "s", "id": 273383, "pid": 76337, "tid": -914061504, "ts": 1716454225815044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225838500, "dur": 712,
    "args": {
      "External id": 273384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273384,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273384, "pid": 5, "tid": 7, "ts": 1716454225838500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815044, "dur": 5,
    "args": {
      "External id": 273384,
      "cbid": 211, "correlation": 273384
    }
  },
  {
    "ph": "s", "id": 273384, "pid": 76337, "tid": -914061504, "ts": 1716454225815044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225839213, "dur": 60,
    "args": {
      "External id": 273389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273389, "pid": 5, "tid": 7, "ts": 1716454225839213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815073, "dur": 8,
    "args": {
      "External id": 273389,
      "cbid": 211, "correlation": 273389
    }
  },
  {
    "ph": "s", "id": 273389, "pid": 76337, "tid": -914061504, "ts": 1716454225815073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225839275, "dur": 50,
    "args": {
      "External id": 273397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273397,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273397, "pid": 5, "tid": 7, "ts": 1716454225839275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815105, "dur": 8,
    "args": {
      "External id": 273397,
      "cbid": 211, "correlation": 273397
    }
  },
  {
    "ph": "s", "id": 273397, "pid": 76337, "tid": -914061504, "ts": 1716454225815105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225839326, "dur": 35,
    "args": {
      "External id": 273405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273405,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273405, "pid": 5, "tid": 7, "ts": 1716454225839326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815134, "dur": 8,
    "args": {
      "External id": 273405,
      "cbid": 211, "correlation": 273405
    }
  },
  {
    "ph": "s", "id": 273405, "pid": 76337, "tid": -914061504, "ts": 1716454225815134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225839363, "dur": 52,
    "args": {
      "External id": 273425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273425,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 273425, "pid": 5, "tid": 7, "ts": 1716454225839363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815215, "dur": 13,
    "args": {
      "External id": 273425,
      "cbid": 211, "correlation": 273425
    }
  },
  {
    "ph": "s", "id": 273425, "pid": 76337, "tid": -914061504, "ts": 1716454225815215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225839416, "dur": 4,
    "args": {
      "External id": 273437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273437,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273437, "pid": 5, "tid": 7, "ts": 1716454225839416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815237, "dur": 6,
    "args": {
      "External id": 273437,
      "cbid": 211, "correlation": 273437
    }
  },
  {
    "ph": "s", "id": 273437, "pid": 76337, "tid": -914061504, "ts": 1716454225815237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225839421, "dur": 57,
    "args": {
      "External id": 273440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273440, "pid": 5, "tid": 7, "ts": 1716454225839421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815255, "dur": 6,
    "args": {
      "External id": 273440,
      "cbid": 211, "correlation": 273440
    }
  },
  {
    "ph": "s", "id": 273440, "pid": 76337, "tid": -914061504, "ts": 1716454225815255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815314, "dur": 0,
    "args": {
      "External id": 273451,
      "cbid": 317, "correlation": 273451
    }
  },
  {
    "ph": "f", "id": 273451, "pid": 76337, "tid": -914061504, "ts": 1716454225815314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815315, "dur": 0,
    "args": {
      "External id": 273452,
      "cbid": 203, "correlation": 273452
    }
  },
  {
    "ph": "f", "id": 273452, "pid": 76337, "tid": -914061504, "ts": 1716454225815315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815316, "dur": 0,
    "args": {
      "External id": 273453,
      "cbid": 205, "correlation": 273453
    }
  },
  {
    "ph": "f", "id": 273453, "pid": 76337, "tid": -914061504, "ts": 1716454225815316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815337, "dur": 1,
    "args": {
      "External id": 273457,
      "cbid": 251, "correlation": 273457
    }
  },
  {
    "ph": "f", "id": 273457, "pid": 76337, "tid": -914061504, "ts": 1716454225815337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815339, "dur": 0,
    "args": {
      "External id": 273458,
      "cbid": 251, "correlation": 273458
    }
  },
  {
    "ph": "f", "id": 273458, "pid": 76337, "tid": -914061504, "ts": 1716454225815339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815340, "dur": 0,
    "args": {
      "External id": 273459,
      "cbid": 251, "correlation": 273459
    }
  },
  {
    "ph": "f", "id": 273459, "pid": 76337, "tid": -914061504, "ts": 1716454225815340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815340, "dur": 0,
    "args": {
      "External id": 273460,
      "cbid": 251, "correlation": 273460
    }
  },
  {
    "ph": "f", "id": 273460, "pid": 76337, "tid": -914061504, "ts": 1716454225815340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815341, "dur": 0,
    "args": {
      "External id": 273461,
      "cbid": 251, "correlation": 273461
    }
  },
  {
    "ph": "f", "id": 273461, "pid": 76337, "tid": -914061504, "ts": 1716454225815341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815342, "dur": 0,
    "args": {
      "External id": 273462,
      "cbid": 251, "correlation": 273462
    }
  },
  {
    "ph": "f", "id": 273462, "pid": 76337, "tid": -914061504, "ts": 1716454225815342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815343, "dur": 0,
    "args": {
      "External id": 273463,
      "cbid": 251, "correlation": 273463
    }
  },
  {
    "ph": "f", "id": 273463, "pid": 76337, "tid": -914061504, "ts": 1716454225815343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815344, "dur": 0,
    "args": {
      "External id": 273464,
      "cbid": 251, "correlation": 273464
    }
  },
  {
    "ph": "f", "id": 273464, "pid": 76337, "tid": -914061504, "ts": 1716454225815344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815345, "dur": 0,
    "args": {
      "External id": 273465,
      "cbid": 251, "correlation": 273465
    }
  },
  {
    "ph": "f", "id": 273465, "pid": 76337, "tid": -914061504, "ts": 1716454225815345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225839480, "dur": 112,
    "args": {
      "External id": 273466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273466,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273466, "pid": 5, "tid": 7, "ts": 1716454225839480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815347, "dur": 12,
    "args": {
      "External id": 273466,
      "cbid": 211, "correlation": 273466
    }
  },
  {
    "ph": "s", "id": 273466, "pid": 76337, "tid": -914061504, "ts": 1716454225815347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225839593, "dur": 60,
    "args": {
      "External id": 273472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273472,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273472, "pid": 5, "tid": 7, "ts": 1716454225839593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815383, "dur": 8,
    "args": {
      "External id": 273472,
      "cbid": 211, "correlation": 273472
    }
  },
  {
    "ph": "s", "id": 273472, "pid": 76337, "tid": -914061504, "ts": 1716454225815383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225839654, "dur": 604,
    "args": {
      "External id": 273481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273481,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273481, "pid": 5, "tid": 7, "ts": 1716454225839654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815465, "dur": 14,
    "args": {
      "External id": 273481,
      "cbid": 211, "correlation": 273481
    }
  },
  {
    "ph": "s", "id": 273481, "pid": 76337, "tid": -914061504, "ts": 1716454225815465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225840260, "dur": 185,
    "args": {
      "External id": 273503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273503,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273503, "pid": 5, "tid": 7, "ts": 1716454225840260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815523, "dur": 10,
    "args": {
      "External id": 273503,
      "cbid": 211, "correlation": 273503
    }
  },
  {
    "ph": "s", "id": 273503, "pid": 76337, "tid": -914061504, "ts": 1716454225815523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815611, "dur": 1,
    "args": {
      "External id": 273514,
      "cbid": 251, "correlation": 273514
    }
  },
  {
    "ph": "f", "id": 273514, "pid": 76337, "tid": -914061504, "ts": 1716454225815611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225840446, "dur": 196,
    "args": {
      "External id": 273515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273515,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273515, "pid": 5, "tid": 7, "ts": 1716454225840446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815617, "dur": 13,
    "args": {
      "External id": 273515,
      "cbid": 211, "correlation": 273515
    }
  },
  {
    "ph": "s", "id": 273515, "pid": 76337, "tid": -914061504, "ts": 1716454225815617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815686, "dur": 1,
    "args": {
      "External id": 273526,
      "cbid": 251, "correlation": 273526
    }
  },
  {
    "ph": "f", "id": 273526, "pid": 76337, "tid": -914061504, "ts": 1716454225815686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225840643, "dur": 190,
    "args": {
      "External id": 273527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273527,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273527, "pid": 5, "tid": 7, "ts": 1716454225840643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815690, "dur": 11,
    "args": {
      "External id": 273527,
      "cbid": 211, "correlation": 273527
    }
  },
  {
    "ph": "s", "id": 273527, "pid": 76337, "tid": -914061504, "ts": 1716454225815690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815752, "dur": 1,
    "args": {
      "External id": 273538,
      "cbid": 251, "correlation": 273538
    }
  },
  {
    "ph": "f", "id": 273538, "pid": 76337, "tid": -914061504, "ts": 1716454225815752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225840835, "dur": 189,
    "args": {
      "External id": 273539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273539,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273539, "pid": 5, "tid": 7, "ts": 1716454225840835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815757, "dur": 12,
    "args": {
      "External id": 273539,
      "cbid": 211, "correlation": 273539
    }
  },
  {
    "ph": "s", "id": 273539, "pid": 76337, "tid": -914061504, "ts": 1716454225815757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225841026, "dur": 17715,
    "args": {
      "External id": 273560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273560,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273560, "pid": 5, "tid": 7, "ts": 1716454225841026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815837, "dur": 12,
    "args": {
      "External id": 273560,
      "cbid": 211, "correlation": 273560
    }
  },
  {
    "ph": "s", "id": 273560, "pid": 76337, "tid": -914061504, "ts": 1716454225815837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815934, "dur": 1,
    "args": {
      "External id": 273578,
      "cbid": 251, "correlation": 273578
    }
  },
  {
    "ph": "f", "id": 273578, "pid": 76337, "tid": -914061504, "ts": 1716454225815934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225858742, "dur": 195,
    "args": {
      "External id": 273580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273580,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273580, "pid": 5, "tid": 7, "ts": 1716454225858742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225815940, "dur": 13,
    "args": {
      "External id": 273580,
      "cbid": 211, "correlation": 273580
    }
  },
  {
    "ph": "s", "id": 273580, "pid": 76337, "tid": -914061504, "ts": 1716454225815940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225858939, "dur": 66,
    "args": {
      "External id": 273588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273588,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273588, "pid": 5, "tid": 7, "ts": 1716454225858939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816019, "dur": 13,
    "args": {
      "External id": 273588,
      "cbid": 211, "correlation": 273588
    }
  },
  {
    "ph": "s", "id": 273588, "pid": 76337, "tid": -914061504, "ts": 1716454225816019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225859007, "dur": 98,
    "args": {
      "External id": 273596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273596,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273596, "pid": 5, "tid": 7, "ts": 1716454225859007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816059, "dur": 9,
    "args": {
      "External id": 273596,
      "cbid": 211, "correlation": 273596
    }
  },
  {
    "ph": "s", "id": 273596, "pid": 76337, "tid": -914061504, "ts": 1716454225816059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225859105, "dur": 53,
    "args": {
      "External id": 273607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273607, "pid": 5, "tid": 7, "ts": 1716454225859105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816130, "dur": 13,
    "args": {
      "External id": 273607,
      "cbid": 211, "correlation": 273607
    }
  },
  {
    "ph": "s", "id": 273607, "pid": 76337, "tid": -914061504, "ts": 1716454225816130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225859159, "dur": 87,
    "args": {
      "External id": 273629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273629,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273629, "pid": 5, "tid": 7, "ts": 1716454225859159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816161, "dur": 8,
    "args": {
      "External id": 273629,
      "cbid": 211, "correlation": 273629
    }
  },
  {
    "ph": "s", "id": 273629, "pid": 76337, "tid": -914061504, "ts": 1716454225816161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816244, "dur": 1,
    "args": {
      "External id": 273640,
      "cbid": 251, "correlation": 273640
    }
  },
  {
    "ph": "f", "id": 273640, "pid": 76337, "tid": -914061504, "ts": 1716454225816244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225859248, "dur": 99,
    "args": {
      "External id": 273641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273641,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273641, "pid": 5, "tid": 7, "ts": 1716454225859248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816249, "dur": 13,
    "args": {
      "External id": 273641,
      "cbid": 211, "correlation": 273641
    }
  },
  {
    "ph": "s", "id": 273641, "pid": 76337, "tid": -914061504, "ts": 1716454225816249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816321, "dur": 1,
    "args": {
      "External id": 273652,
      "cbid": 251, "correlation": 273652
    }
  },
  {
    "ph": "f", "id": 273652, "pid": 76337, "tid": -914061504, "ts": 1716454225816321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816324, "dur": 0,
    "args": {
      "External id": 273653,
      "cbid": 251, "correlation": 273653
    }
  },
  {
    "ph": "f", "id": 273653, "pid": 76337, "tid": -914061504, "ts": 1716454225816324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225859349, "dur": 9,
    "args": {
      "External id": 273654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273654,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 273654, "pid": 5, "tid": 7, "ts": 1716454225859349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816326, "dur": 12,
    "args": {
      "External id": 273654,
      "cbid": 211, "correlation": 273654
    }
  },
  {
    "ph": "s", "id": 273654, "pid": 76337, "tid": -914061504, "ts": 1716454225816326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225859360, "dur": 5,
    "args": {
      "External id": 273656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273656,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 273656, "pid": 5, "tid": 7, "ts": 1716454225859360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816340, "dur": 6,
    "args": {
      "External id": 273656,
      "cbid": 211, "correlation": 273656
    }
  },
  {
    "ph": "s", "id": 273656, "pid": 76337, "tid": -914061504, "ts": 1716454225816340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816405, "dur": 1,
    "args": {
      "External id": 273667,
      "cbid": 251, "correlation": 273667
    }
  },
  {
    "ph": "f", "id": 273667, "pid": 76337, "tid": -914061504, "ts": 1716454225816405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816409, "dur": 0,
    "args": {
      "External id": 273668,
      "cbid": 251, "correlation": 273668
    }
  },
  {
    "ph": "f", "id": 273668, "pid": 76337, "tid": -914061504, "ts": 1716454225816409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225859366, "dur": 6,
    "args": {
      "External id": 273669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273669,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 273669, "pid": 5, "tid": 7, "ts": 1716454225859366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816410, "dur": 12,
    "args": {
      "External id": 273669,
      "cbid": 211, "correlation": 273669
    }
  },
  {
    "ph": "s", "id": 273669, "pid": 76337, "tid": -914061504, "ts": 1716454225816410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225859373, "dur": 3,
    "args": {
      "External id": 273671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273671,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 273671, "pid": 5, "tid": 7, "ts": 1716454225859373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816424, "dur": 5,
    "args": {
      "External id": 273671,
      "cbid": 211, "correlation": 273671
    }
  },
  {
    "ph": "s", "id": 273671, "pid": 76337, "tid": -914061504, "ts": 1716454225816424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225859378, "dur": 148,
    "args": {
      "External id": 273692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273692,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273692, "pid": 5, "tid": 7, "ts": 1716454225859378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816497, "dur": 13,
    "args": {
      "External id": 273692,
      "cbid": 211, "correlation": 273692
    }
  },
  {
    "ph": "s", "id": 273692, "pid": 76337, "tid": -914061504, "ts": 1716454225816497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816596, "dur": 1,
    "args": {
      "External id": 273710,
      "cbid": 251, "correlation": 273710
    }
  },
  {
    "ph": "f", "id": 273710, "pid": 76337, "tid": -914061504, "ts": 1716454225816596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225859528, "dur": 103,
    "args": {
      "External id": 273712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273712,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273712, "pid": 5, "tid": 7, "ts": 1716454225859528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816602, "dur": 14,
    "args": {
      "External id": 273712,
      "cbid": 211, "correlation": 273712
    }
  },
  {
    "ph": "s", "id": 273712, "pid": 76337, "tid": -914061504, "ts": 1716454225816602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225859632, "dur": 35,
    "args": {
      "External id": 273720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273720,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273720, "pid": 5, "tid": 7, "ts": 1716454225859632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816671, "dur": 12,
    "args": {
      "External id": 273720,
      "cbid": 211, "correlation": 273720
    }
  },
  {
    "ph": "s", "id": 273720, "pid": 76337, "tid": -914061504, "ts": 1716454225816671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225859668, "dur": 65,
    "args": {
      "External id": 273728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273728,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273728, "pid": 5, "tid": 7, "ts": 1716454225859668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816712, "dur": 9,
    "args": {
      "External id": 273728,
      "cbid": 211, "correlation": 273728
    }
  },
  {
    "ph": "s", "id": 273728, "pid": 76337, "tid": -914061504, "ts": 1716454225816712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225859735, "dur": 87,
    "args": {
      "External id": 273750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273750,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273750, "pid": 5, "tid": 7, "ts": 1716454225859735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816767, "dur": 10,
    "args": {
      "External id": 273750,
      "cbid": 211, "correlation": 273750
    }
  },
  {
    "ph": "s", "id": 273750, "pid": 76337, "tid": -914061504, "ts": 1716454225816767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816857, "dur": 1,
    "args": {
      "External id": 273766,
      "cbid": 251, "correlation": 273766
    }
  },
  {
    "ph": "f", "id": 273766, "pid": 76337, "tid": -914061504, "ts": 1716454225816857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225859823, "dur": 560,
    "args": {
      "External id": 273768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273768,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273768, "pid": 5, "tid": 7, "ts": 1716454225859823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816863, "dur": 12,
    "args": {
      "External id": 273768,
      "cbid": 211, "correlation": 273768
    }
  },
  {
    "ph": "s", "id": 273768, "pid": 76337, "tid": -914061504, "ts": 1716454225816863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225860384, "dur": 237,
    "args": {
      "External id": 273776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273776, "pid": 5, "tid": 7, "ts": 1716454225860384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816927, "dur": 12,
    "args": {
      "External id": 273776,
      "cbid": 211, "correlation": 273776
    }
  },
  {
    "ph": "s", "id": 273776, "pid": 76337, "tid": -914061504, "ts": 1716454225816927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225860623, "dur": 249,
    "args": {
      "External id": 273784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273784, "pid": 5, "tid": 7, "ts": 1716454225860623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225816960, "dur": 8,
    "args": {
      "External id": 273784,
      "cbid": 211, "correlation": 273784
    }
  },
  {
    "ph": "s", "id": 273784, "pid": 76337, "tid": -914061504, "ts": 1716454225816960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817048, "dur": 2,
    "args": {
      "External id": 273800,
      "cbid": 251, "correlation": 273800
    }
  },
  {
    "ph": "f", "id": 273800, "pid": 76337, "tid": -914061504, "ts": 1716454225817048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817054, "dur": 0,
    "args": {
      "External id": 273802,
      "cbid": 251, "correlation": 273802
    }
  },
  {
    "ph": "f", "id": 273802, "pid": 76337, "tid": -914061504, "ts": 1716454225817054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225860872, "dur": 354,
    "args": {
      "External id": 273803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273803,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 273803, "pid": 5, "tid": 7, "ts": 1716454225860872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817057, "dur": 13,
    "args": {
      "External id": 273803,
      "cbid": 211, "correlation": 273803
    }
  },
  {
    "ph": "s", "id": 273803, "pid": 76337, "tid": -914061504, "ts": 1716454225817057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225861228, "dur": 50,
    "args": {
      "External id": 273811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273811,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273811, "pid": 5, "tid": 7, "ts": 1716454225861228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817099, "dur": 10,
    "args": {
      "External id": 273811,
      "cbid": 211, "correlation": 273811
    }
  },
  {
    "ph": "s", "id": 273811, "pid": 76337, "tid": -914061504, "ts": 1716454225817099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225861280, "dur": 151,
    "args": {
      "External id": 273822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273822, "pid": 5, "tid": 7, "ts": 1716454225861280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817166, "dur": 12,
    "args": {
      "External id": 273822,
      "cbid": 211, "correlation": 273822
    }
  },
  {
    "ph": "s", "id": 273822, "pid": 76337, "tid": -914061504, "ts": 1716454225817166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817229, "dur": 0,
    "args": {
      "External id": 273834,
      "cbid": 317, "correlation": 273834
    }
  },
  {
    "ph": "f", "id": 273834, "pid": 76337, "tid": -914061504, "ts": 1716454225817229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817230, "dur": 0,
    "args": {
      "External id": 273835,
      "cbid": 203, "correlation": 273835
    }
  },
  {
    "ph": "f", "id": 273835, "pid": 76337, "tid": -914061504, "ts": 1716454225817230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817231, "dur": 0,
    "args": {
      "External id": 273836,
      "cbid": 205, "correlation": 273836
    }
  },
  {
    "ph": "f", "id": 273836, "pid": 76337, "tid": -914061504, "ts": 1716454225817231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817254, "dur": 1,
    "args": {
      "External id": 273840,
      "cbid": 251, "correlation": 273840
    }
  },
  {
    "ph": "f", "id": 273840, "pid": 76337, "tid": -914061504, "ts": 1716454225817254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817256, "dur": 0,
    "args": {
      "External id": 273841,
      "cbid": 251, "correlation": 273841
    }
  },
  {
    "ph": "f", "id": 273841, "pid": 76337, "tid": -914061504, "ts": 1716454225817256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817256, "dur": 0,
    "args": {
      "External id": 273842,
      "cbid": 251, "correlation": 273842
    }
  },
  {
    "ph": "f", "id": 273842, "pid": 76337, "tid": -914061504, "ts": 1716454225817256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817257, "dur": 0,
    "args": {
      "External id": 273843,
      "cbid": 251, "correlation": 273843
    }
  },
  {
    "ph": "f", "id": 273843, "pid": 76337, "tid": -914061504, "ts": 1716454225817257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817258, "dur": 0,
    "args": {
      "External id": 273844,
      "cbid": 251, "correlation": 273844
    }
  },
  {
    "ph": "f", "id": 273844, "pid": 76337, "tid": -914061504, "ts": 1716454225817258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817258, "dur": 0,
    "args": {
      "External id": 273845,
      "cbid": 251, "correlation": 273845
    }
  },
  {
    "ph": "f", "id": 273845, "pid": 76337, "tid": -914061504, "ts": 1716454225817258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817259, "dur": 0,
    "args": {
      "External id": 273846,
      "cbid": 251, "correlation": 273846
    }
  },
  {
    "ph": "f", "id": 273846, "pid": 76337, "tid": -914061504, "ts": 1716454225817259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817260, "dur": 0,
    "args": {
      "External id": 273847,
      "cbid": 251, "correlation": 273847
    }
  },
  {
    "ph": "f", "id": 273847, "pid": 76337, "tid": -914061504, "ts": 1716454225817260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817261, "dur": 0,
    "args": {
      "External id": 273848,
      "cbid": 251, "correlation": 273848
    }
  },
  {
    "ph": "f", "id": 273848, "pid": 76337, "tid": -914061504, "ts": 1716454225817261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225861432, "dur": 112,
    "args": {
      "External id": 273849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273849,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 273849, "pid": 5, "tid": 7, "ts": 1716454225861432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817263, "dur": 12,
    "args": {
      "External id": 273849,
      "cbid": 211, "correlation": 273849
    }
  },
  {
    "ph": "s", "id": 273849, "pid": 76337, "tid": -914061504, "ts": 1716454225817263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225861546, "dur": 57,
    "args": {
      "External id": 273855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273855, "pid": 5, "tid": 7, "ts": 1716454225861546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817298, "dur": 9,
    "args": {
      "External id": 273855,
      "cbid": 211, "correlation": 273855
    }
  },
  {
    "ph": "s", "id": 273855, "pid": 76337, "tid": -914061504, "ts": 1716454225817298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225861604, "dur": 50,
    "args": {
      "External id": 273863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273863,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273863, "pid": 5, "tid": 7, "ts": 1716454225861604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817329, "dur": 8,
    "args": {
      "External id": 273863,
      "cbid": 211, "correlation": 273863
    }
  },
  {
    "ph": "s", "id": 273863, "pid": 76337, "tid": -914061504, "ts": 1716454225817329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817403, "dur": 0,
    "args": {
      "External id": 273873,
      "cbid": 317, "correlation": 273873
    }
  },
  {
    "ph": "f", "id": 273873, "pid": 76337, "tid": -914061504, "ts": 1716454225817403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817404, "dur": 0,
    "args": {
      "External id": 273874,
      "cbid": 203, "correlation": 273874
    }
  },
  {
    "ph": "f", "id": 273874, "pid": 76337, "tid": -914061504, "ts": 1716454225817404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817404, "dur": 0,
    "args": {
      "External id": 273875,
      "cbid": 205, "correlation": 273875
    }
  },
  {
    "ph": "f", "id": 273875, "pid": 76337, "tid": -914061504, "ts": 1716454225817404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225861655, "dur": 41,
    "args": {
      "External id": 273879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273879,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273879, "pid": 5, "tid": 7, "ts": 1716454225861655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817421, "dur": 12,
    "args": {
      "External id": 273879,
      "cbid": 211, "correlation": 273879
    }
  },
  {
    "ph": "s", "id": 273879, "pid": 76337, "tid": -914061504, "ts": 1716454225817421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225861697, "dur": 13,
    "args": {
      "External id": 273881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273881, "pid": 5, "tid": 7, "ts": 1716454225861697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817436, "dur": 5,
    "args": {
      "External id": 273881,
      "cbid": 211, "correlation": 273881
    }
  },
  {
    "ph": "s", "id": 273881, "pid": 76337, "tid": -914061504, "ts": 1716454225817436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225861713, "dur": 1,
    "args": {
      "External id": 273883,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 273883,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9411764705882353
    }
  },
  {
    "ph": "f", "id": 273883, "pid": 5, "tid": 7, "ts": 1716454225861713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817455, "dur": 17,
    "args": {
      "External id": 273883,
      "cbid": 51, "correlation": 273883
    }
  },
  {
    "ph": "s", "id": 273883, "pid": 76337, "tid": -914061504, "ts": 1716454225817455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225861717, "dur": 348,
    "args": {
      "External id": 273884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273884,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273884, "pid": 5, "tid": 7, "ts": 1716454225861717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817477, "dur": 10,
    "args": {
      "External id": 273884,
      "cbid": 211, "correlation": 273884
    }
  },
  {
    "ph": "s", "id": 273884, "pid": 76337, "tid": -914061504, "ts": 1716454225817477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862066, "dur": 13,
    "args": {
      "External id": 273886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273886,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273886, "pid": 5, "tid": 7, "ts": 1716454225862066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817497, "dur": 7,
    "args": {
      "External id": 273886,
      "cbid": 211, "correlation": 273886
    }
  },
  {
    "ph": "s", "id": 273886, "pid": 76337, "tid": -914061504, "ts": 1716454225817497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225862080, "dur": 14,
    "args": {
      "External id": 273892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273892, "pid": 5, "tid": 7, "ts": 1716454225862080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817527, "dur": 9,
    "args": {
      "External id": 273892,
      "cbid": 211, "correlation": 273892
    }
  },
  {
    "ph": "s", "id": 273892, "pid": 76337, "tid": -914061504, "ts": 1716454225817527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862096, "dur": 19,
    "args": {
      "External id": 273912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273912,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 273912, "pid": 5, "tid": 7, "ts": 1716454225862096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817619, "dur": 12,
    "args": {
      "External id": 273912,
      "cbid": 211, "correlation": 273912
    }
  },
  {
    "ph": "s", "id": 273912, "pid": 76337, "tid": -914061504, "ts": 1716454225817619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225862116, "dur": 5,
    "args": {
      "External id": 273924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273924,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273924, "pid": 5, "tid": 7, "ts": 1716454225862116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817642, "dur": 6,
    "args": {
      "External id": 273924,
      "cbid": 211, "correlation": 273924
    }
  },
  {
    "ph": "s", "id": 273924, "pid": 76337, "tid": -914061504, "ts": 1716454225817642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225862121, "dur": 16,
    "args": {
      "External id": 273927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273927,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273927, "pid": 5, "tid": 7, "ts": 1716454225862121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817660, "dur": 7,
    "args": {
      "External id": 273927,
      "cbid": 211, "correlation": 273927
    }
  },
  {
    "ph": "s", "id": 273927, "pid": 76337, "tid": -914061504, "ts": 1716454225817660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225862139, "dur": 12,
    "args": {
      "External id": 273936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273936,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273936, "pid": 5, "tid": 7, "ts": 1716454225862139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817700, "dur": 10,
    "args": {
      "External id": 273936,
      "cbid": 211, "correlation": 273936
    }
  },
  {
    "ph": "s", "id": 273936, "pid": 76337, "tid": -914061504, "ts": 1716454225817700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817763, "dur": 0,
    "args": {
      "External id": 273946,
      "cbid": 317, "correlation": 273946
    }
  },
  {
    "ph": "f", "id": 273946, "pid": 76337, "tid": -914061504, "ts": 1716454225817763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817763, "dur": 0,
    "args": {
      "External id": 273947,
      "cbid": 203, "correlation": 273947
    }
  },
  {
    "ph": "f", "id": 273947, "pid": 76337, "tid": -914061504, "ts": 1716454225817763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817764, "dur": 0,
    "args": {
      "External id": 273948,
      "cbid": 205, "correlation": 273948
    }
  },
  {
    "ph": "f", "id": 273948, "pid": 76337, "tid": -914061504, "ts": 1716454225817764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862152, "dur": 11,
    "args": {
      "External id": 273952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273952, "pid": 5, "tid": 7, "ts": 1716454225862152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817780, "dur": 13,
    "args": {
      "External id": 273952,
      "cbid": 211, "correlation": 273952
    }
  },
  {
    "ph": "s", "id": 273952, "pid": 76337, "tid": -914061504, "ts": 1716454225817780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862164, "dur": 23,
    "args": {
      "External id": 273954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273954,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273954, "pid": 5, "tid": 7, "ts": 1716454225862164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817796, "dur": 5,
    "args": {
      "External id": 273954,
      "cbid": 211, "correlation": 273954
    }
  },
  {
    "ph": "s", "id": 273954, "pid": 76337, "tid": -914061504, "ts": 1716454225817796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225862189, "dur": 4,
    "args": {
      "External id": 273956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 273956, "pid": 5, "tid": 7, "ts": 1716454225862189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817806, "dur": 6,
    "args": {
      "External id": 273956,
      "cbid": 211, "correlation": 273956
    }
  },
  {
    "ph": "s", "id": 273956, "pid": 76337, "tid": -914061504, "ts": 1716454225817806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817816, "dur": 0,
    "args": {
      "External id": 273957,
      "cbid": 51, "correlation": 273957
    }
  },
  {
    "ph": "s", "id": 273957, "pid": 76337, "tid": -914061504, "ts": 1716454225817816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225862194, "dur": 346,
    "args": {
      "External id": 273958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273958,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 273958, "pid": 5, "tid": 7, "ts": 1716454225862194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817819, "dur": 7,
    "args": {
      "External id": 273958,
      "cbid": 211, "correlation": 273958
    }
  },
  {
    "ph": "s", "id": 273958, "pid": 76337, "tid": -914061504, "ts": 1716454225817819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862542, "dur": 19,
    "args": {
      "External id": 273959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273959,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273959, "pid": 5, "tid": 7, "ts": 1716454225862542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817829, "dur": 5,
    "args": {
      "External id": 273959,
      "cbid": 211, "correlation": 273959
    }
  },
  {
    "ph": "s", "id": 273959, "pid": 76337, "tid": -914061504, "ts": 1716454225817829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225862562, "dur": 32,
    "args": {
      "External id": 273965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 273965, "pid": 5, "tid": 7, "ts": 1716454225862562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817859, "dur": 9,
    "args": {
      "External id": 273965,
      "cbid": 211, "correlation": 273965
    }
  },
  {
    "ph": "s", "id": 273965, "pid": 76337, "tid": -914061504, "ts": 1716454225817859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225862595, "dur": 4,
    "args": {
      "External id": 273973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273973,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 273973, "pid": 5, "tid": 7, "ts": 1716454225862595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817904, "dur": 10,
    "args": {
      "External id": 273973,
      "cbid": 211, "correlation": 273973
    }
  },
  {
    "ph": "s", "id": 273973, "pid": 76337, "tid": -914061504, "ts": 1716454225817904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817969, "dur": 1,
    "args": {
      "External id": 273989,
      "cbid": 251, "correlation": 273989
    }
  },
  {
    "ph": "f", "id": 273989, "pid": 76337, "tid": -914061504, "ts": 1716454225817969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817982, "dur": 0,
    "args": {
      "External id": 273991,
      "cbid": 251, "correlation": 273991
    }
  },
  {
    "ph": "f", "id": 273991, "pid": 76337, "tid": -914061504, "ts": 1716454225817982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225862601, "dur": 11,
    "args": {
      "External id": 273992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273992,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 273992, "pid": 5, "tid": 7, "ts": 1716454225862601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817984, "dur": 12,
    "args": {
      "External id": 273992,
      "cbid": 211, "correlation": 273992
    }
  },
  {
    "ph": "s", "id": 273992, "pid": 76337, "tid": -914061504, "ts": 1716454225817984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225862613, "dur": 5,
    "args": {
      "External id": 273994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 273994,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 273994, "pid": 5, "tid": 7, "ts": 1716454225862613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225817998, "dur": 5,
    "args": {
      "External id": 273994,
      "cbid": 211, "correlation": 273994
    }
  },
  {
    "ph": "s", "id": 273994, "pid": 76337, "tid": -914061504, "ts": 1716454225817998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225862619, "dur": 28,
    "args": {
      "External id": 274004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274004, "pid": 5, "tid": 7, "ts": 1716454225862619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818057, "dur": 12,
    "args": {
      "External id": 274004,
      "cbid": 211, "correlation": 274004
    }
  },
  {
    "ph": "s", "id": 274004, "pid": 76337, "tid": -914061504, "ts": 1716454225818057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862648, "dur": 29,
    "args": {
      "External id": 274024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274024,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 274024, "pid": 5, "tid": 7, "ts": 1716454225862648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818126, "dur": 12,
    "args": {
      "External id": 274024,
      "cbid": 211, "correlation": 274024
    }
  },
  {
    "ph": "s", "id": 274024, "pid": 76337, "tid": -914061504, "ts": 1716454225818126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225862679, "dur": 4,
    "args": {
      "External id": 274036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274036,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 274036, "pid": 5, "tid": 7, "ts": 1716454225862679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818149, "dur": 6,
    "args": {
      "External id": 274036,
      "cbid": 211, "correlation": 274036
    }
  },
  {
    "ph": "s", "id": 274036, "pid": 76337, "tid": -914061504, "ts": 1716454225818149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225862684, "dur": 29,
    "args": {
      "External id": 274039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274039, "pid": 5, "tid": 7, "ts": 1716454225862684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818168, "dur": 6,
    "args": {
      "External id": 274039,
      "cbid": 211, "correlation": 274039
    }
  },
  {
    "ph": "s", "id": 274039, "pid": 76337, "tid": -914061504, "ts": 1716454225818168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225862714, "dur": 21,
    "args": {
      "External id": 274048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274048,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274048, "pid": 5, "tid": 7, "ts": 1716454225862714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818209, "dur": 10,
    "args": {
      "External id": 274048,
      "cbid": 211, "correlation": 274048
    }
  },
  {
    "ph": "s", "id": 274048, "pid": 76337, "tid": -914061504, "ts": 1716454225818209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818272, "dur": 0,
    "args": {
      "External id": 274058,
      "cbid": 317, "correlation": 274058
    }
  },
  {
    "ph": "f", "id": 274058, "pid": 76337, "tid": -914061504, "ts": 1716454225818272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818273, "dur": 0,
    "args": {
      "External id": 274059,
      "cbid": 203, "correlation": 274059
    }
  },
  {
    "ph": "f", "id": 274059, "pid": 76337, "tid": -914061504, "ts": 1716454225818273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818274, "dur": 0,
    "args": {
      "External id": 274060,
      "cbid": 205, "correlation": 274060
    }
  },
  {
    "ph": "f", "id": 274060, "pid": 76337, "tid": -914061504, "ts": 1716454225818274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862736, "dur": 22,
    "args": {
      "External id": 274064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274064,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274064, "pid": 5, "tid": 7, "ts": 1716454225862736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818296, "dur": 12,
    "args": {
      "External id": 274064,
      "cbid": 211, "correlation": 274064
    }
  },
  {
    "ph": "s", "id": 274064, "pid": 76337, "tid": -914061504, "ts": 1716454225818296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225862759, "dur": 41,
    "args": {
      "External id": 274066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274066,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274066, "pid": 5, "tid": 7, "ts": 1716454225862759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818311, "dur": 5,
    "args": {
      "External id": 274066,
      "cbid": 211, "correlation": 274066
    }
  },
  {
    "ph": "s", "id": 274066, "pid": 76337, "tid": -914061504, "ts": 1716454225818311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225862802, "dur": 633,
    "args": {
      "External id": 274068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274068,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274068, "pid": 5, "tid": 7, "ts": 1716454225862802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818328, "dur": 14,
    "args": {
      "External id": 274068,
      "cbid": 211, "correlation": 274068
    }
  },
  {
    "ph": "s", "id": 274068, "pid": 76337, "tid": -914061504, "ts": 1716454225818328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225863436, "dur": 20,
    "args": {
      "External id": 274070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274070,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274070, "pid": 5, "tid": 7, "ts": 1716454225863436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818346, "dur": 5,
    "args": {
      "External id": 274070,
      "cbid": 211, "correlation": 274070
    }
  },
  {
    "ph": "s", "id": 274070, "pid": 76337, "tid": -914061504, "ts": 1716454225818346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225863458, "dur": 32,
    "args": {
      "External id": 274076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274076,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274076, "pid": 5, "tid": 7, "ts": 1716454225863458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818374, "dur": 9,
    "args": {
      "External id": 274076,
      "cbid": 211, "correlation": 274076
    }
  },
  {
    "ph": "s", "id": 274076, "pid": 76337, "tid": -914061504, "ts": 1716454225818374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818434, "dur": 0,
    "args": {
      "External id": 274086,
      "cbid": 317, "correlation": 274086
    }
  },
  {
    "ph": "f", "id": 274086, "pid": 76337, "tid": -914061504, "ts": 1716454225818434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818434, "dur": 0,
    "args": {
      "External id": 274087,
      "cbid": 203, "correlation": 274087
    }
  },
  {
    "ph": "f", "id": 274087, "pid": 76337, "tid": -914061504, "ts": 1716454225818434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818435, "dur": 0,
    "args": {
      "External id": 274088,
      "cbid": 205, "correlation": 274088
    }
  },
  {
    "ph": "f", "id": 274088, "pid": 76337, "tid": -914061504, "ts": 1716454225818435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818458, "dur": 1,
    "args": {
      "External id": 274092,
      "cbid": 251, "correlation": 274092
    }
  },
  {
    "ph": "f", "id": 274092, "pid": 76337, "tid": -914061504, "ts": 1716454225818458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818460, "dur": 0,
    "args": {
      "External id": 274093,
      "cbid": 251, "correlation": 274093
    }
  },
  {
    "ph": "f", "id": 274093, "pid": 76337, "tid": -914061504, "ts": 1716454225818460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818460, "dur": 0,
    "args": {
      "External id": 274094,
      "cbid": 251, "correlation": 274094
    }
  },
  {
    "ph": "f", "id": 274094, "pid": 76337, "tid": -914061504, "ts": 1716454225818460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818461, "dur": 0,
    "args": {
      "External id": 274095,
      "cbid": 251, "correlation": 274095
    }
  },
  {
    "ph": "f", "id": 274095, "pid": 76337, "tid": -914061504, "ts": 1716454225818461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818462, "dur": 0,
    "args": {
      "External id": 274096,
      "cbid": 251, "correlation": 274096
    }
  },
  {
    "ph": "f", "id": 274096, "pid": 76337, "tid": -914061504, "ts": 1716454225818462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818463, "dur": 0,
    "args": {
      "External id": 274097,
      "cbid": 251, "correlation": 274097
    }
  },
  {
    "ph": "f", "id": 274097, "pid": 76337, "tid": -914061504, "ts": 1716454225818463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818463, "dur": 0,
    "args": {
      "External id": 274098,
      "cbid": 251, "correlation": 274098
    }
  },
  {
    "ph": "f", "id": 274098, "pid": 76337, "tid": -914061504, "ts": 1716454225818463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818464, "dur": 0,
    "args": {
      "External id": 274099,
      "cbid": 251, "correlation": 274099
    }
  },
  {
    "ph": "f", "id": 274099, "pid": 76337, "tid": -914061504, "ts": 1716454225818464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818466, "dur": 0,
    "args": {
      "External id": 274100,
      "cbid": 251, "correlation": 274100
    }
  },
  {
    "ph": "f", "id": 274100, "pid": 76337, "tid": -914061504, "ts": 1716454225818466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225863491, "dur": 49,
    "args": {
      "External id": 274101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274101,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 274101, "pid": 5, "tid": 7, "ts": 1716454225863491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818468, "dur": 13,
    "args": {
      "External id": 274101,
      "cbid": 211, "correlation": 274101
    }
  },
  {
    "ph": "s", "id": 274101, "pid": 76337, "tid": -914061504, "ts": 1716454225818468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225863541, "dur": 31,
    "args": {
      "External id": 274107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274107, "pid": 5, "tid": 7, "ts": 1716454225863541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818500, "dur": 8,
    "args": {
      "External id": 274107,
      "cbid": 211, "correlation": 274107
    }
  },
  {
    "ph": "s", "id": 274107, "pid": 76337, "tid": -914061504, "ts": 1716454225818500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225863574, "dur": 27,
    "args": {
      "External id": 274115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274115,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274115, "pid": 5, "tid": 7, "ts": 1716454225863574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818531, "dur": 8,
    "args": {
      "External id": 274115,
      "cbid": 211, "correlation": 274115
    }
  },
  {
    "ph": "s", "id": 274115, "pid": 76337, "tid": -914061504, "ts": 1716454225818531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225863602, "dur": 20,
    "args": {
      "External id": 274123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274123,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274123, "pid": 5, "tid": 7, "ts": 1716454225863602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818559, "dur": 8,
    "args": {
      "External id": 274123,
      "cbid": 211, "correlation": 274123
    }
  },
  {
    "ph": "s", "id": 274123, "pid": 76337, "tid": -914061504, "ts": 1716454225818559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225863623, "dur": 29,
    "args": {
      "External id": 274143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274143,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 274143, "pid": 5, "tid": 7, "ts": 1716454225863623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818640, "dur": 12,
    "args": {
      "External id": 274143,
      "cbid": 211, "correlation": 274143
    }
  },
  {
    "ph": "s", "id": 274143, "pid": 76337, "tid": -914061504, "ts": 1716454225818640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225863653, "dur": 4,
    "args": {
      "External id": 274155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274155,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 274155, "pid": 5, "tid": 7, "ts": 1716454225863653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818662, "dur": 6,
    "args": {
      "External id": 274155,
      "cbid": 211, "correlation": 274155
    }
  },
  {
    "ph": "s", "id": 274155, "pid": 76337, "tid": -914061504, "ts": 1716454225818662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225863658, "dur": 29,
    "args": {
      "External id": 274158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274158, "pid": 5, "tid": 7, "ts": 1716454225863658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818681, "dur": 6,
    "args": {
      "External id": 274158,
      "cbid": 211, "correlation": 274158
    }
  },
  {
    "ph": "s", "id": 274158, "pid": 76337, "tid": -914061504, "ts": 1716454225818681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818738, "dur": 0,
    "args": {
      "External id": 274169,
      "cbid": 317, "correlation": 274169
    }
  },
  {
    "ph": "f", "id": 274169, "pid": 76337, "tid": -914061504, "ts": 1716454225818738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818738, "dur": 0,
    "args": {
      "External id": 274170,
      "cbid": 203, "correlation": 274170
    }
  },
  {
    "ph": "f", "id": 274170, "pid": 76337, "tid": -914061504, "ts": 1716454225818738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818739, "dur": 0,
    "args": {
      "External id": 274171,
      "cbid": 205, "correlation": 274171
    }
  },
  {
    "ph": "f", "id": 274171, "pid": 76337, "tid": -914061504, "ts": 1716454225818739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225863688, "dur": 22,
    "args": {
      "External id": 274175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274175,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274175, "pid": 5, "tid": 7, "ts": 1716454225863688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818754, "dur": 12,
    "args": {
      "External id": 274175,
      "cbid": 211, "correlation": 274175
    }
  },
  {
    "ph": "s", "id": 274175, "pid": 76337, "tid": -914061504, "ts": 1716454225818754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225863712, "dur": 114,
    "args": {
      "External id": 274177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274177,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274177, "pid": 5, "tid": 7, "ts": 1716454225863712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818778, "dur": 9,
    "args": {
      "External id": 274177,
      "cbid": 211, "correlation": 274177
    }
  },
  {
    "ph": "s", "id": 274177, "pid": 76337, "tid": -914061504, "ts": 1716454225818778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225863828, "dur": 22,
    "args": {
      "External id": 274179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274179,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274179, "pid": 5, "tid": 7, "ts": 1716454225863828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818791, "dur": 5,
    "args": {
      "External id": 274179,
      "cbid": 211, "correlation": 274179
    }
  },
  {
    "ph": "s", "id": 274179, "pid": 76337, "tid": -914061504, "ts": 1716454225818791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225863851, "dur": 31,
    "args": {
      "External id": 274185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274185, "pid": 5, "tid": 7, "ts": 1716454225863851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818819, "dur": 9,
    "args": {
      "External id": 274185,
      "cbid": 211, "correlation": 274185
    }
  },
  {
    "ph": "s", "id": 274185, "pid": 76337, "tid": -914061504, "ts": 1716454225818819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225863884, "dur": 184,
    "args": {
      "External id": 274194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274194,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274194, "pid": 5, "tid": 7, "ts": 1716454225863884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818902, "dur": 14,
    "args": {
      "External id": 274194,
      "cbid": 211, "correlation": 274194
    }
  },
  {
    "ph": "s", "id": 274194, "pid": 76337, "tid": -914061504, "ts": 1716454225818902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225864069, "dur": 63,
    "args": {
      "External id": 274216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274216,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274216, "pid": 5, "tid": 7, "ts": 1716454225864069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225818959, "dur": 10,
    "args": {
      "External id": 274216,
      "cbid": 211, "correlation": 274216
    }
  },
  {
    "ph": "s", "id": 274216, "pid": 76337, "tid": -914061504, "ts": 1716454225818959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819058, "dur": 1,
    "args": {
      "External id": 274227,
      "cbid": 251, "correlation": 274227
    }
  },
  {
    "ph": "f", "id": 274227, "pid": 76337, "tid": -914061504, "ts": 1716454225819058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225864133, "dur": 150,
    "args": {
      "External id": 274228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274228,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274228, "pid": 5, "tid": 7, "ts": 1716454225864133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819063, "dur": 14,
    "args": {
      "External id": 274228,
      "cbid": 211, "correlation": 274228
    }
  },
  {
    "ph": "s", "id": 274228, "pid": 76337, "tid": -914061504, "ts": 1716454225819063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819135, "dur": 1,
    "args": {
      "External id": 274239,
      "cbid": 251, "correlation": 274239
    }
  },
  {
    "ph": "f", "id": 274239, "pid": 76337, "tid": -914061504, "ts": 1716454225819135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225864285, "dur": 142,
    "args": {
      "External id": 274240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274240,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274240, "pid": 5, "tid": 7, "ts": 1716454225864285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819138, "dur": 11,
    "args": {
      "External id": 274240,
      "cbid": 211, "correlation": 274240
    }
  },
  {
    "ph": "s", "id": 274240, "pid": 76337, "tid": -914061504, "ts": 1716454225819138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819202, "dur": 1,
    "args": {
      "External id": 274251,
      "cbid": 251, "correlation": 274251
    }
  },
  {
    "ph": "f", "id": 274251, "pid": 76337, "tid": -914061504, "ts": 1716454225819202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225864429, "dur": 143,
    "args": {
      "External id": 274252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274252,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274252, "pid": 5, "tid": 7, "ts": 1716454225864429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819206, "dur": 11,
    "args": {
      "External id": 274252,
      "cbid": 211, "correlation": 274252
    }
  },
  {
    "ph": "s", "id": 274252, "pid": 76337, "tid": -914061504, "ts": 1716454225819206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225864573, "dur": 1848,
    "args": {
      "External id": 274273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274273,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 274273, "pid": 5, "tid": 7, "ts": 1716454225864573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819290, "dur": 13,
    "args": {
      "External id": 274273,
      "cbid": 211, "correlation": 274273
    }
  },
  {
    "ph": "s", "id": 274273, "pid": 76337, "tid": -914061504, "ts": 1716454225819290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819390, "dur": 1,
    "args": {
      "External id": 274291,
      "cbid": 251, "correlation": 274291
    }
  },
  {
    "ph": "f", "id": 274291, "pid": 76337, "tid": -914061504, "ts": 1716454225819390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225866423, "dur": 142,
    "args": {
      "External id": 274293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274293,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 274293, "pid": 5, "tid": 7, "ts": 1716454225866423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819395, "dur": 13,
    "args": {
      "External id": 274293,
      "cbid": 211, "correlation": 274293
    }
  },
  {
    "ph": "s", "id": 274293, "pid": 76337, "tid": -914061504, "ts": 1716454225819395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225866567, "dur": 36,
    "args": {
      "External id": 274301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274301,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274301, "pid": 5, "tid": 7, "ts": 1716454225866567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819466, "dur": 12,
    "args": {
      "External id": 274301,
      "cbid": 211, "correlation": 274301
    }
  },
  {
    "ph": "s", "id": 274301, "pid": 76337, "tid": -914061504, "ts": 1716454225819466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225866604, "dur": 51,
    "args": {
      "External id": 274309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274309,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274309, "pid": 5, "tid": 7, "ts": 1716454225866604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819505, "dur": 9,
    "args": {
      "External id": 274309,
      "cbid": 211, "correlation": 274309
    }
  },
  {
    "ph": "s", "id": 274309, "pid": 76337, "tid": -914061504, "ts": 1716454225819505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225866656, "dur": 30,
    "args": {
      "External id": 274320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274320, "pid": 5, "tid": 7, "ts": 1716454225866656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819578, "dur": 13,
    "args": {
      "External id": 274320,
      "cbid": 211, "correlation": 274320
    }
  },
  {
    "ph": "s", "id": 274320, "pid": 76337, "tid": -914061504, "ts": 1716454225819578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225866687, "dur": 33,
    "args": {
      "External id": 274342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274342,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274342, "pid": 5, "tid": 7, "ts": 1716454225866687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819609, "dur": 8,
    "args": {
      "External id": 274342,
      "cbid": 211, "correlation": 274342
    }
  },
  {
    "ph": "s", "id": 274342, "pid": 76337, "tid": -914061504, "ts": 1716454225819609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819699, "dur": 1,
    "args": {
      "External id": 274353,
      "cbid": 251, "correlation": 274353
    }
  },
  {
    "ph": "f", "id": 274353, "pid": 76337, "tid": -914061504, "ts": 1716454225819699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225866721, "dur": 87,
    "args": {
      "External id": 274354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274354,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274354, "pid": 5, "tid": 7, "ts": 1716454225866721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819704, "dur": 14,
    "args": {
      "External id": 274354,
      "cbid": 211, "correlation": 274354
    }
  },
  {
    "ph": "s", "id": 274354, "pid": 76337, "tid": -914061504, "ts": 1716454225819704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819774, "dur": 1,
    "args": {
      "External id": 274365,
      "cbid": 251, "correlation": 274365
    }
  },
  {
    "ph": "f", "id": 274365, "pid": 76337, "tid": -914061504, "ts": 1716454225819774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819778, "dur": 0,
    "args": {
      "External id": 274366,
      "cbid": 251, "correlation": 274366
    }
  },
  {
    "ph": "f", "id": 274366, "pid": 76337, "tid": -914061504, "ts": 1716454225819778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225866809, "dur": 11,
    "args": {
      "External id": 274367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274367,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 274367, "pid": 5, "tid": 7, "ts": 1716454225866809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819780, "dur": 12,
    "args": {
      "External id": 274367,
      "cbid": 211, "correlation": 274367
    }
  },
  {
    "ph": "s", "id": 274367, "pid": 76337, "tid": -914061504, "ts": 1716454225819780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225866822, "dur": 5,
    "args": {
      "External id": 274369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274369,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 274369, "pid": 5, "tid": 7, "ts": 1716454225866822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819795, "dur": 7,
    "args": {
      "External id": 274369,
      "cbid": 211, "correlation": 274369
    }
  },
  {
    "ph": "s", "id": 274369, "pid": 76337, "tid": -914061504, "ts": 1716454225819795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819856, "dur": 1,
    "args": {
      "External id": 274380,
      "cbid": 251, "correlation": 274380
    }
  },
  {
    "ph": "f", "id": 274380, "pid": 76337, "tid": -914061504, "ts": 1716454225819856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819859, "dur": 0,
    "args": {
      "External id": 274381,
      "cbid": 251, "correlation": 274381
    }
  },
  {
    "ph": "f", "id": 274381, "pid": 76337, "tid": -914061504, "ts": 1716454225819859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225866828, "dur": 7,
    "args": {
      "External id": 274382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274382,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 274382, "pid": 5, "tid": 7, "ts": 1716454225866828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819861, "dur": 12,
    "args": {
      "External id": 274382,
      "cbid": 211, "correlation": 274382
    }
  },
  {
    "ph": "s", "id": 274382, "pid": 76337, "tid": -914061504, "ts": 1716454225819861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225866836, "dur": 3,
    "args": {
      "External id": 274384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274384,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 274384, "pid": 5, "tid": 7, "ts": 1716454225866836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819875, "dur": 5,
    "args": {
      "External id": 274384,
      "cbid": 211, "correlation": 274384
    }
  },
  {
    "ph": "s", "id": 274384, "pid": 76337, "tid": -914061504, "ts": 1716454225819875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225866841, "dur": 88,
    "args": {
      "External id": 274405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274405,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 274405, "pid": 5, "tid": 7, "ts": 1716454225866841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225819949, "dur": 13,
    "args": {
      "External id": 274405,
      "cbid": 211, "correlation": 274405
    }
  },
  {
    "ph": "s", "id": 274405, "pid": 76337, "tid": -914061504, "ts": 1716454225819949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820060, "dur": 1,
    "args": {
      "External id": 274423,
      "cbid": 251, "correlation": 274423
    }
  },
  {
    "ph": "f", "id": 274423, "pid": 76337, "tid": -914061504, "ts": 1716454225820060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225866930, "dur": 95,
    "args": {
      "External id": 274425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274425,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274425, "pid": 5, "tid": 7, "ts": 1716454225866930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820067, "dur": 14,
    "args": {
      "External id": 274425,
      "cbid": 211, "correlation": 274425
    }
  },
  {
    "ph": "s", "id": 274425, "pid": 76337, "tid": -914061504, "ts": 1716454225820067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225867025, "dur": 18,
    "args": {
      "External id": 274433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274433,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274433, "pid": 5, "tid": 7, "ts": 1716454225867025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820137, "dur": 12,
    "args": {
      "External id": 274433,
      "cbid": 211, "correlation": 274433
    }
  },
  {
    "ph": "s", "id": 274433, "pid": 76337, "tid": -914061504, "ts": 1716454225820137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225867045, "dur": 36,
    "args": {
      "External id": 274441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274441,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274441, "pid": 5, "tid": 7, "ts": 1716454225867045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820178, "dur": 9,
    "args": {
      "External id": 274441,
      "cbid": 211, "correlation": 274441
    }
  },
  {
    "ph": "s", "id": 274441, "pid": 76337, "tid": -914061504, "ts": 1716454225820178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225867082, "dur": 33,
    "args": {
      "External id": 274463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274463,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274463, "pid": 5, "tid": 7, "ts": 1716454225867082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820233, "dur": 10,
    "args": {
      "External id": 274463,
      "cbid": 211, "correlation": 274463
    }
  },
  {
    "ph": "s", "id": 274463, "pid": 76337, "tid": -914061504, "ts": 1716454225820233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820324, "dur": 1,
    "args": {
      "External id": 274479,
      "cbid": 251, "correlation": 274479
    }
  },
  {
    "ph": "f", "id": 274479, "pid": 76337, "tid": -914061504, "ts": 1716454225820324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820329, "dur": 0,
    "args": {
      "External id": 274481,
      "cbid": 251, "correlation": 274481
    }
  },
  {
    "ph": "f", "id": 274481, "pid": 76337, "tid": -914061504, "ts": 1716454225820329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225867117, "dur": 523,
    "args": {
      "External id": 274482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274482,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 274482, "pid": 5, "tid": 7, "ts": 1716454225867117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820332, "dur": 13,
    "args": {
      "External id": 274482,
      "cbid": 211, "correlation": 274482
    }
  },
  {
    "ph": "s", "id": 274482, "pid": 76337, "tid": -914061504, "ts": 1716454225820332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225867641, "dur": 120,
    "args": {
      "External id": 274490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274490, "pid": 5, "tid": 7, "ts": 1716454225867641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820397, "dur": 13,
    "args": {
      "External id": 274490,
      "cbid": 211, "correlation": 274490
    }
  },
  {
    "ph": "s", "id": 274490, "pid": 76337, "tid": -914061504, "ts": 1716454225820397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225867763, "dur": 127,
    "args": {
      "External id": 274498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274498, "pid": 5, "tid": 7, "ts": 1716454225867763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820429, "dur": 8,
    "args": {
      "External id": 274498,
      "cbid": 211, "correlation": 274498
    }
  },
  {
    "ph": "s", "id": 274498, "pid": 76337, "tid": -914061504, "ts": 1716454225820429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820511, "dur": 2,
    "args": {
      "External id": 274514,
      "cbid": 251, "correlation": 274514
    }
  },
  {
    "ph": "f", "id": 274514, "pid": 76337, "tid": -914061504, "ts": 1716454225820511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225867891, "dur": 292,
    "args": {
      "External id": 274516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274516,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274516, "pid": 5, "tid": 7, "ts": 1716454225867891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820517, "dur": 12,
    "args": {
      "External id": 274516,
      "cbid": 211, "correlation": 274516
    }
  },
  {
    "ph": "s", "id": 274516, "pid": 76337, "tid": -914061504, "ts": 1716454225820517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225868185, "dur": 28,
    "args": {
      "External id": 274524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274524,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274524, "pid": 5, "tid": 7, "ts": 1716454225868185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820560, "dur": 10,
    "args": {
      "External id": 274524,
      "cbid": 211, "correlation": 274524
    }
  },
  {
    "ph": "s", "id": 274524, "pid": 76337, "tid": -914061504, "ts": 1716454225820560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225868214, "dur": 78,
    "args": {
      "External id": 274535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274535, "pid": 5, "tid": 7, "ts": 1716454225868214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820628, "dur": 12,
    "args": {
      "External id": 274535,
      "cbid": 211, "correlation": 274535
    }
  },
  {
    "ph": "s", "id": 274535, "pid": 76337, "tid": -914061504, "ts": 1716454225820628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820693, "dur": 0,
    "args": {
      "External id": 274547,
      "cbid": 317, "correlation": 274547
    }
  },
  {
    "ph": "f", "id": 274547, "pid": 76337, "tid": -914061504, "ts": 1716454225820693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820694, "dur": 0,
    "args": {
      "External id": 274548,
      "cbid": 203, "correlation": 274548
    }
  },
  {
    "ph": "f", "id": 274548, "pid": 76337, "tid": -914061504, "ts": 1716454225820694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820694, "dur": 0,
    "args": {
      "External id": 274549,
      "cbid": 205, "correlation": 274549
    }
  },
  {
    "ph": "f", "id": 274549, "pid": 76337, "tid": -914061504, "ts": 1716454225820694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225868292, "dur": 22,
    "args": {
      "External id": 274553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274553, "pid": 5, "tid": 7, "ts": 1716454225868292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820713, "dur": 12,
    "args": {
      "External id": 274553,
      "cbid": 211, "correlation": 274553
    }
  },
  {
    "ph": "s", "id": 274553, "pid": 76337, "tid": -914061504, "ts": 1716454225820713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225868315, "dur": 115,
    "args": {
      "External id": 274555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274555,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274555, "pid": 5, "tid": 7, "ts": 1716454225868315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820731, "dur": 7,
    "args": {
      "External id": 274555,
      "cbid": 211, "correlation": 274555
    }
  },
  {
    "ph": "s", "id": 274555, "pid": 76337, "tid": -914061504, "ts": 1716454225820731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225868431, "dur": 22,
    "args": {
      "External id": 274557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274557,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274557, "pid": 5, "tid": 7, "ts": 1716454225868431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820742, "dur": 5,
    "args": {
      "External id": 274557,
      "cbid": 211, "correlation": 274557
    }
  },
  {
    "ph": "s", "id": 274557, "pid": 76337, "tid": -914061504, "ts": 1716454225820742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225868455, "dur": 32,
    "args": {
      "External id": 274563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274563, "pid": 5, "tid": 7, "ts": 1716454225868455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820771, "dur": 8,
    "args": {
      "External id": 274563,
      "cbid": 211, "correlation": 274563
    }
  },
  {
    "ph": "s", "id": 274563, "pid": 76337, "tid": -914061504, "ts": 1716454225820771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225868488, "dur": 27,
    "args": {
      "External id": 274571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274571,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274571, "pid": 5, "tid": 7, "ts": 1716454225868488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820803, "dur": 9,
    "args": {
      "External id": 274571,
      "cbid": 211, "correlation": 274571
    }
  },
  {
    "ph": "s", "id": 274571, "pid": 76337, "tid": -914061504, "ts": 1716454225820803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225868516, "dur": 29,
    "args": {
      "External id": 274591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274591,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 274591, "pid": 5, "tid": 7, "ts": 1716454225868516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820875, "dur": 11,
    "args": {
      "External id": 274591,
      "cbid": 211, "correlation": 274591
    }
  },
  {
    "ph": "s", "id": 274591, "pid": 76337, "tid": -914061504, "ts": 1716454225820875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225868546, "dur": 4,
    "args": {
      "External id": 274603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274603,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 274603, "pid": 5, "tid": 7, "ts": 1716454225868546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820899, "dur": 7,
    "args": {
      "External id": 274603,
      "cbid": 211, "correlation": 274603
    }
  },
  {
    "ph": "s", "id": 274603, "pid": 76337, "tid": -914061504, "ts": 1716454225820899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225868552, "dur": 30,
    "args": {
      "External id": 274606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274606,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274606, "pid": 5, "tid": 7, "ts": 1716454225868552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820918, "dur": 7,
    "args": {
      "External id": 274606,
      "cbid": 211, "correlation": 274606
    }
  },
  {
    "ph": "s", "id": 274606, "pid": 76337, "tid": -914061504, "ts": 1716454225820918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225868583, "dur": 20,
    "args": {
      "External id": 274615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274615,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274615, "pid": 5, "tid": 7, "ts": 1716454225868583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225820958, "dur": 10,
    "args": {
      "External id": 274615,
      "cbid": 211, "correlation": 274615
    }
  },
  {
    "ph": "s", "id": 274615, "pid": 76337, "tid": -914061504, "ts": 1716454225820958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821018, "dur": 0,
    "args": {
      "External id": 274625,
      "cbid": 317, "correlation": 274625
    }
  },
  {
    "ph": "f", "id": 274625, "pid": 76337, "tid": -914061504, "ts": 1716454225821018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821019, "dur": 0,
    "args": {
      "External id": 274626,
      "cbid": 203, "correlation": 274626
    }
  },
  {
    "ph": "f", "id": 274626, "pid": 76337, "tid": -914061504, "ts": 1716454225821019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821019, "dur": 0,
    "args": {
      "External id": 274627,
      "cbid": 205, "correlation": 274627
    }
  },
  {
    "ph": "f", "id": 274627, "pid": 76337, "tid": -914061504, "ts": 1716454225821019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225868605, "dur": 22,
    "args": {
      "External id": 274631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274631, "pid": 5, "tid": 7, "ts": 1716454225868605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821032, "dur": 12,
    "args": {
      "External id": 274631,
      "cbid": 211, "correlation": 274631
    }
  },
  {
    "ph": "s", "id": 274631, "pid": 76337, "tid": -914061504, "ts": 1716454225821032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225868628, "dur": 42,
    "args": {
      "External id": 274633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274633,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274633, "pid": 5, "tid": 7, "ts": 1716454225868628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821046, "dur": 5,
    "args": {
      "External id": 274633,
      "cbid": 211, "correlation": 274633
    }
  },
  {
    "ph": "s", "id": 274633, "pid": 76337, "tid": -914061504, "ts": 1716454225821046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225868671, "dur": 633,
    "args": {
      "External id": 274635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274635,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274635, "pid": 5, "tid": 7, "ts": 1716454225868671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821058, "dur": 6,
    "args": {
      "External id": 274635,
      "cbid": 211, "correlation": 274635
    }
  },
  {
    "ph": "s", "id": 274635, "pid": 76337, "tid": -914061504, "ts": 1716454225821058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225869306, "dur": 20,
    "args": {
      "External id": 274637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274637,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274637, "pid": 5, "tid": 7, "ts": 1716454225869306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821068, "dur": 5,
    "args": {
      "External id": 274637,
      "cbid": 211, "correlation": 274637
    }
  },
  {
    "ph": "s", "id": 274637, "pid": 76337, "tid": -914061504, "ts": 1716454225821068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225869328, "dur": 32,
    "args": {
      "External id": 274643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274643, "pid": 5, "tid": 7, "ts": 1716454225869328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821095, "dur": 9,
    "args": {
      "External id": 274643,
      "cbid": 211, "correlation": 274643
    }
  },
  {
    "ph": "s", "id": 274643, "pid": 76337, "tid": -914061504, "ts": 1716454225821095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225869361, "dur": 4,
    "args": {
      "External id": 274651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274651,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 274651, "pid": 5, "tid": 7, "ts": 1716454225869361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821138, "dur": 9,
    "args": {
      "External id": 274651,
      "cbid": 211, "correlation": 274651
    }
  },
  {
    "ph": "s", "id": 274651, "pid": 76337, "tid": -914061504, "ts": 1716454225821138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821206, "dur": 1,
    "args": {
      "External id": 274667,
      "cbid": 251, "correlation": 274667
    }
  },
  {
    "ph": "f", "id": 274667, "pid": 76337, "tid": -914061504, "ts": 1716454225821206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821211, "dur": 0,
    "args": {
      "External id": 274669,
      "cbid": 251, "correlation": 274669
    }
  },
  {
    "ph": "f", "id": 274669, "pid": 76337, "tid": -914061504, "ts": 1716454225821211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225869366, "dur": 12,
    "args": {
      "External id": 274670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274670,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 274670, "pid": 5, "tid": 7, "ts": 1716454225869366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821213, "dur": 11,
    "args": {
      "External id": 274670,
      "cbid": 211, "correlation": 274670
    }
  },
  {
    "ph": "s", "id": 274670, "pid": 76337, "tid": -914061504, "ts": 1716454225821213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225869380, "dur": 5,
    "args": {
      "External id": 274672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274672,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 274672, "pid": 5, "tid": 7, "ts": 1716454225869380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821225, "dur": 5,
    "args": {
      "External id": 274672,
      "cbid": 211, "correlation": 274672
    }
  },
  {
    "ph": "s", "id": 274672, "pid": 76337, "tid": -914061504, "ts": 1716454225821225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225869386, "dur": 29,
    "args": {
      "External id": 274682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274682, "pid": 5, "tid": 7, "ts": 1716454225869386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821282, "dur": 12,
    "args": {
      "External id": 274682,
      "cbid": 211, "correlation": 274682
    }
  },
  {
    "ph": "s", "id": 274682, "pid": 76337, "tid": -914061504, "ts": 1716454225821282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225869416, "dur": 30,
    "args": {
      "External id": 274702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274702,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 274702, "pid": 5, "tid": 7, "ts": 1716454225869416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821347, "dur": 11,
    "args": {
      "External id": 274702,
      "cbid": 211, "correlation": 274702
    }
  },
  {
    "ph": "s", "id": 274702, "pid": 76337, "tid": -914061504, "ts": 1716454225821347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225869447, "dur": 4,
    "args": {
      "External id": 274714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274714,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 274714, "pid": 5, "tid": 7, "ts": 1716454225869447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821370, "dur": 7,
    "args": {
      "External id": 274714,
      "cbid": 211, "correlation": 274714
    }
  },
  {
    "ph": "s", "id": 274714, "pid": 76337, "tid": -914061504, "ts": 1716454225821370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225869453, "dur": 29,
    "args": {
      "External id": 274717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274717, "pid": 5, "tid": 7, "ts": 1716454225869453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821390, "dur": 6,
    "args": {
      "External id": 274717,
      "cbid": 211, "correlation": 274717
    }
  },
  {
    "ph": "s", "id": 274717, "pid": 76337, "tid": -914061504, "ts": 1716454225821390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225869483, "dur": 21,
    "args": {
      "External id": 274726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274726,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274726, "pid": 5, "tid": 7, "ts": 1716454225869483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821431, "dur": 10,
    "args": {
      "External id": 274726,
      "cbid": 211, "correlation": 274726
    }
  },
  {
    "ph": "s", "id": 274726, "pid": 76337, "tid": -914061504, "ts": 1716454225821431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821494, "dur": 0,
    "args": {
      "External id": 274736,
      "cbid": 317, "correlation": 274736
    }
  },
  {
    "ph": "f", "id": 274736, "pid": 76337, "tid": -914061504, "ts": 1716454225821494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821495, "dur": 0,
    "args": {
      "External id": 274737,
      "cbid": 203, "correlation": 274737
    }
  },
  {
    "ph": "f", "id": 274737, "pid": 76337, "tid": -914061504, "ts": 1716454225821495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821496, "dur": 0,
    "args": {
      "External id": 274738,
      "cbid": 205, "correlation": 274738
    }
  },
  {
    "ph": "f", "id": 274738, "pid": 76337, "tid": -914061504, "ts": 1716454225821496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225869505, "dur": 22,
    "args": {
      "External id": 274742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274742, "pid": 5, "tid": 7, "ts": 1716454225869505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821510, "dur": 12,
    "args": {
      "External id": 274742,
      "cbid": 211, "correlation": 274742
    }
  },
  {
    "ph": "s", "id": 274742, "pid": 76337, "tid": -914061504, "ts": 1716454225821510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225869528, "dur": 42,
    "args": {
      "External id": 274744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274744, "pid": 5, "tid": 7, "ts": 1716454225869528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821524, "dur": 5,
    "args": {
      "External id": 274744,
      "cbid": 211, "correlation": 274744
    }
  },
  {
    "ph": "s", "id": 274744, "pid": 76337, "tid": -914061504, "ts": 1716454225821524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225869572, "dur": 624,
    "args": {
      "External id": 274746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274746,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274746, "pid": 5, "tid": 7, "ts": 1716454225869572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821535, "dur": 5,
    "args": {
      "External id": 274746,
      "cbid": 211, "correlation": 274746
    }
  },
  {
    "ph": "s", "id": 274746, "pid": 76337, "tid": -914061504, "ts": 1716454225821535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225870197, "dur": 21,
    "args": {
      "External id": 274748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274748,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274748, "pid": 5, "tid": 7, "ts": 1716454225870197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821544, "dur": 5,
    "args": {
      "External id": 274748,
      "cbid": 211, "correlation": 274748
    }
  },
  {
    "ph": "s", "id": 274748, "pid": 76337, "tid": -914061504, "ts": 1716454225821544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225870220, "dur": 32,
    "args": {
      "External id": 274754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274754, "pid": 5, "tid": 7, "ts": 1716454225870220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821572, "dur": 8,
    "args": {
      "External id": 274754,
      "cbid": 211, "correlation": 274754
    }
  },
  {
    "ph": "s", "id": 274754, "pid": 76337, "tid": -914061504, "ts": 1716454225821572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225870254, "dur": 27,
    "args": {
      "External id": 274762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274762,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274762, "pid": 5, "tid": 7, "ts": 1716454225870254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821605, "dur": 8,
    "args": {
      "External id": 274762,
      "cbid": 211, "correlation": 274762
    }
  },
  {
    "ph": "s", "id": 274762, "pid": 76337, "tid": -914061504, "ts": 1716454225821605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225870282, "dur": 20,
    "args": {
      "External id": 274770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274770,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274770, "pid": 5, "tid": 7, "ts": 1716454225870282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821633, "dur": 8,
    "args": {
      "External id": 274770,
      "cbid": 211, "correlation": 274770
    }
  },
  {
    "ph": "s", "id": 274770, "pid": 76337, "tid": -914061504, "ts": 1716454225821633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225870303, "dur": 29,
    "args": {
      "External id": 274790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274790,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 274790, "pid": 5, "tid": 7, "ts": 1716454225870303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821712, "dur": 12,
    "args": {
      "External id": 274790,
      "cbid": 211, "correlation": 274790
    }
  },
  {
    "ph": "s", "id": 274790, "pid": 76337, "tid": -914061504, "ts": 1716454225821712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225870333, "dur": 4,
    "args": {
      "External id": 274802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274802,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 274802, "pid": 5, "tid": 7, "ts": 1716454225870333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821733, "dur": 6,
    "args": {
      "External id": 274802,
      "cbid": 211, "correlation": 274802
    }
  },
  {
    "ph": "s", "id": 274802, "pid": 76337, "tid": -914061504, "ts": 1716454225821733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225870338, "dur": 29,
    "args": {
      "External id": 274805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274805, "pid": 5, "tid": 7, "ts": 1716454225870338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821751, "dur": 6,
    "args": {
      "External id": 274805,
      "cbid": 211, "correlation": 274805
    }
  },
  {
    "ph": "s", "id": 274805, "pid": 76337, "tid": -914061504, "ts": 1716454225821751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821809, "dur": 0,
    "args": {
      "External id": 274816,
      "cbid": 317, "correlation": 274816
    }
  },
  {
    "ph": "f", "id": 274816, "pid": 76337, "tid": -914061504, "ts": 1716454225821809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821809, "dur": 0,
    "args": {
      "External id": 274817,
      "cbid": 203, "correlation": 274817
    }
  },
  {
    "ph": "f", "id": 274817, "pid": 76337, "tid": -914061504, "ts": 1716454225821809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821810, "dur": 0,
    "args": {
      "External id": 274818,
      "cbid": 205, "correlation": 274818
    }
  },
  {
    "ph": "f", "id": 274818, "pid": 76337, "tid": -914061504, "ts": 1716454225821810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225870368, "dur": 22,
    "args": {
      "External id": 274822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274822,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274822, "pid": 5, "tid": 7, "ts": 1716454225870368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821823, "dur": 11,
    "args": {
      "External id": 274822,
      "cbid": 211, "correlation": 274822
    }
  },
  {
    "ph": "s", "id": 274822, "pid": 76337, "tid": -914061504, "ts": 1716454225821823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225870391, "dur": 110,
    "args": {
      "External id": 274824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274824,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274824, "pid": 5, "tid": 7, "ts": 1716454225870391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821840, "dur": 6,
    "args": {
      "External id": 274824,
      "cbid": 211, "correlation": 274824
    }
  },
  {
    "ph": "s", "id": 274824, "pid": 76337, "tid": -914061504, "ts": 1716454225821840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225870503, "dur": 21,
    "args": {
      "External id": 274826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274826,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274826, "pid": 5, "tid": 7, "ts": 1716454225870503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821850, "dur": 5,
    "args": {
      "External id": 274826,
      "cbid": 211, "correlation": 274826
    }
  },
  {
    "ph": "s", "id": 274826, "pid": 76337, "tid": -914061504, "ts": 1716454225821850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225870525, "dur": 31,
    "args": {
      "External id": 274832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274832, "pid": 5, "tid": 7, "ts": 1716454225870525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821877, "dur": 8,
    "args": {
      "External id": 274832,
      "cbid": 211, "correlation": 274832
    }
  },
  {
    "ph": "s", "id": 274832, "pid": 76337, "tid": -914061504, "ts": 1716454225821877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225870557, "dur": 184,
    "args": {
      "External id": 274841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274841,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274841, "pid": 5, "tid": 7, "ts": 1716454225870557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225821958, "dur": 14,
    "args": {
      "External id": 274841,
      "cbid": 211, "correlation": 274841
    }
  },
  {
    "ph": "s", "id": 274841, "pid": 76337, "tid": -914061504, "ts": 1716454225821958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225870743, "dur": 63,
    "args": {
      "External id": 274863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274863,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274863, "pid": 5, "tid": 7, "ts": 1716454225870743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822026, "dur": 11,
    "args": {
      "External id": 274863,
      "cbid": 211, "correlation": 274863
    }
  },
  {
    "ph": "s", "id": 274863, "pid": 76337, "tid": -914061504, "ts": 1716454225822026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822115, "dur": 1,
    "args": {
      "External id": 274874,
      "cbid": 251, "correlation": 274874
    }
  },
  {
    "ph": "f", "id": 274874, "pid": 76337, "tid": -914061504, "ts": 1716454225822115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225870807, "dur": 151,
    "args": {
      "External id": 274875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274875,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274875, "pid": 5, "tid": 7, "ts": 1716454225870807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822120, "dur": 12,
    "args": {
      "External id": 274875,
      "cbid": 211, "correlation": 274875
    }
  },
  {
    "ph": "s", "id": 274875, "pid": 76337, "tid": -914061504, "ts": 1716454225822120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822193, "dur": 1,
    "args": {
      "External id": 274886,
      "cbid": 251, "correlation": 274886
    }
  },
  {
    "ph": "f", "id": 274886, "pid": 76337, "tid": -914061504, "ts": 1716454225822193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225870960, "dur": 143,
    "args": {
      "External id": 274887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274887,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274887, "pid": 5, "tid": 7, "ts": 1716454225870960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822198, "dur": 11,
    "args": {
      "External id": 274887,
      "cbid": 211, "correlation": 274887
    }
  },
  {
    "ph": "s", "id": 274887, "pid": 76337, "tid": -914061504, "ts": 1716454225822198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822263, "dur": 1,
    "args": {
      "External id": 274898,
      "cbid": 251, "correlation": 274898
    }
  },
  {
    "ph": "f", "id": 274898, "pid": 76337, "tid": -914061504, "ts": 1716454225822263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225871104, "dur": 144,
    "args": {
      "External id": 274899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274899,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 274899, "pid": 5, "tid": 7, "ts": 1716454225871104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822267, "dur": 11,
    "args": {
      "External id": 274899,
      "cbid": 211, "correlation": 274899
    }
  },
  {
    "ph": "s", "id": 274899, "pid": 76337, "tid": -914061504, "ts": 1716454225822267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225871249, "dur": 1855,
    "args": {
      "External id": 274920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274920,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 274920, "pid": 5, "tid": 7, "ts": 1716454225871249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822349, "dur": 12,
    "args": {
      "External id": 274920,
      "cbid": 211, "correlation": 274920
    }
  },
  {
    "ph": "s", "id": 274920, "pid": 76337, "tid": -914061504, "ts": 1716454225822349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822446, "dur": 1,
    "args": {
      "External id": 274938,
      "cbid": 251, "correlation": 274938
    }
  },
  {
    "ph": "f", "id": 274938, "pid": 76337, "tid": -914061504, "ts": 1716454225822446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225873105, "dur": 143,
    "args": {
      "External id": 274940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274940,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 274940, "pid": 5, "tid": 7, "ts": 1716454225873105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822452, "dur": 13,
    "args": {
      "External id": 274940,
      "cbid": 211, "correlation": 274940
    }
  },
  {
    "ph": "s", "id": 274940, "pid": 76337, "tid": -914061504, "ts": 1716454225822452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225873250, "dur": 35,
    "args": {
      "External id": 274948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274948,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274948, "pid": 5, "tid": 7, "ts": 1716454225873250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822522, "dur": 12,
    "args": {
      "External id": 274948,
      "cbid": 211, "correlation": 274948
    }
  },
  {
    "ph": "s", "id": 274948, "pid": 76337, "tid": -914061504, "ts": 1716454225822522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225873286, "dur": 51,
    "args": {
      "External id": 274956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274956,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274956, "pid": 5, "tid": 7, "ts": 1716454225873286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822561, "dur": 9,
    "args": {
      "External id": 274956,
      "cbid": 211, "correlation": 274956
    }
  },
  {
    "ph": "s", "id": 274956, "pid": 76337, "tid": -914061504, "ts": 1716454225822561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225873338, "dur": 29,
    "args": {
      "External id": 274967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274967,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274967, "pid": 5, "tid": 7, "ts": 1716454225873338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822636, "dur": 14,
    "args": {
      "External id": 274967,
      "cbid": 211, "correlation": 274967
    }
  },
  {
    "ph": "s", "id": 274967, "pid": 76337, "tid": -914061504, "ts": 1716454225822636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225873368, "dur": 33,
    "args": {
      "External id": 274989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 274989,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 274989, "pid": 5, "tid": 7, "ts": 1716454225873368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822668, "dur": 8,
    "args": {
      "External id": 274989,
      "cbid": 211, "correlation": 274989
    }
  },
  {
    "ph": "s", "id": 274989, "pid": 76337, "tid": -914061504, "ts": 1716454225822668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822755, "dur": 1,
    "args": {
      "External id": 275000,
      "cbid": 251, "correlation": 275000
    }
  },
  {
    "ph": "f", "id": 275000, "pid": 76337, "tid": -914061504, "ts": 1716454225822755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225873403, "dur": 84,
    "args": {
      "External id": 275001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275001,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275001, "pid": 5, "tid": 7, "ts": 1716454225873403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822760, "dur": 12,
    "args": {
      "External id": 275001,
      "cbid": 211, "correlation": 275001
    }
  },
  {
    "ph": "s", "id": 275001, "pid": 76337, "tid": -914061504, "ts": 1716454225822760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822829, "dur": 1,
    "args": {
      "External id": 275012,
      "cbid": 251, "correlation": 275012
    }
  },
  {
    "ph": "f", "id": 275012, "pid": 76337, "tid": -914061504, "ts": 1716454225822829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822833, "dur": 0,
    "args": {
      "External id": 275013,
      "cbid": 251, "correlation": 275013
    }
  },
  {
    "ph": "f", "id": 275013, "pid": 76337, "tid": -914061504, "ts": 1716454225822833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225873488, "dur": 11,
    "args": {
      "External id": 275014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275014,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 275014, "pid": 5, "tid": 7, "ts": 1716454225873488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822834, "dur": 13,
    "args": {
      "External id": 275014,
      "cbid": 211, "correlation": 275014
    }
  },
  {
    "ph": "s", "id": 275014, "pid": 76337, "tid": -914061504, "ts": 1716454225822834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225873501, "dur": 5,
    "args": {
      "External id": 275016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275016,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 275016, "pid": 5, "tid": 7, "ts": 1716454225873501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822849, "dur": 6,
    "args": {
      "External id": 275016,
      "cbid": 211, "correlation": 275016
    }
  },
  {
    "ph": "s", "id": 275016, "pid": 76337, "tid": -914061504, "ts": 1716454225822849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822906, "dur": 1,
    "args": {
      "External id": 275027,
      "cbid": 251, "correlation": 275027
    }
  },
  {
    "ph": "f", "id": 275027, "pid": 76337, "tid": -914061504, "ts": 1716454225822906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822910, "dur": 0,
    "args": {
      "External id": 275028,
      "cbid": 251, "correlation": 275028
    }
  },
  {
    "ph": "f", "id": 275028, "pid": 76337, "tid": -914061504, "ts": 1716454225822910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225873508, "dur": 7,
    "args": {
      "External id": 275029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275029,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 275029, "pid": 5, "tid": 7, "ts": 1716454225873508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822911, "dur": 11,
    "args": {
      "External id": 275029,
      "cbid": 211, "correlation": 275029
    }
  },
  {
    "ph": "s", "id": 275029, "pid": 76337, "tid": -914061504, "ts": 1716454225822911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225873516, "dur": 3,
    "args": {
      "External id": 275031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275031,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 275031, "pid": 5, "tid": 7, "ts": 1716454225873516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225822924, "dur": 5,
    "args": {
      "External id": 275031,
      "cbid": 211, "correlation": 275031
    }
  },
  {
    "ph": "s", "id": 275031, "pid": 76337, "tid": -914061504, "ts": 1716454225822924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225873520, "dur": 88,
    "args": {
      "External id": 275052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275052,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 275052, "pid": 5, "tid": 7, "ts": 1716454225873520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823005, "dur": 13,
    "args": {
      "External id": 275052,
      "cbid": 211, "correlation": 275052
    }
  },
  {
    "ph": "s", "id": 275052, "pid": 76337, "tid": -914061504, "ts": 1716454225823005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823104, "dur": 1,
    "args": {
      "External id": 275070,
      "cbid": 251, "correlation": 275070
    }
  },
  {
    "ph": "f", "id": 275070, "pid": 76337, "tid": -914061504, "ts": 1716454225823104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225873609, "dur": 95,
    "args": {
      "External id": 275072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275072,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275072, "pid": 5, "tid": 7, "ts": 1716454225873609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823110, "dur": 13,
    "args": {
      "External id": 275072,
      "cbid": 211, "correlation": 275072
    }
  },
  {
    "ph": "s", "id": 275072, "pid": 76337, "tid": -914061504, "ts": 1716454225823110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225873705, "dur": 19,
    "args": {
      "External id": 275080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275080,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275080, "pid": 5, "tid": 7, "ts": 1716454225873705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823179, "dur": 12,
    "args": {
      "External id": 275080,
      "cbid": 211, "correlation": 275080
    }
  },
  {
    "ph": "s", "id": 275080, "pid": 76337, "tid": -914061504, "ts": 1716454225823179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225873725, "dur": 37,
    "args": {
      "External id": 275088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275088,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275088, "pid": 5, "tid": 7, "ts": 1716454225873725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823222, "dur": 9,
    "args": {
      "External id": 275088,
      "cbid": 211, "correlation": 275088
    }
  },
  {
    "ph": "s", "id": 275088, "pid": 76337, "tid": -914061504, "ts": 1716454225823222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225873764, "dur": 33,
    "args": {
      "External id": 275110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275110,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275110, "pid": 5, "tid": 7, "ts": 1716454225873764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823276, "dur": 10,
    "args": {
      "External id": 275110,
      "cbid": 211, "correlation": 275110
    }
  },
  {
    "ph": "s", "id": 275110, "pid": 76337, "tid": -914061504, "ts": 1716454225823276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823369, "dur": 1,
    "args": {
      "External id": 275126,
      "cbid": 251, "correlation": 275126
    }
  },
  {
    "ph": "f", "id": 275126, "pid": 76337, "tid": -914061504, "ts": 1716454225823369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823373, "dur": 0,
    "args": {
      "External id": 275128,
      "cbid": 251, "correlation": 275128
    }
  },
  {
    "ph": "f", "id": 275128, "pid": 76337, "tid": -914061504, "ts": 1716454225823373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225873798, "dur": 523,
    "args": {
      "External id": 275129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275129,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275129, "pid": 5, "tid": 7, "ts": 1716454225873798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823377, "dur": 13,
    "args": {
      "External id": 275129,
      "cbid": 211, "correlation": 275129
    }
  },
  {
    "ph": "s", "id": 275129, "pid": 76337, "tid": -914061504, "ts": 1716454225823377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225874322, "dur": 123,
    "args": {
      "External id": 275137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275137,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275137, "pid": 5, "tid": 7, "ts": 1716454225874322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823443, "dur": 12,
    "args": {
      "External id": 275137,
      "cbid": 211, "correlation": 275137
    }
  },
  {
    "ph": "s", "id": 275137, "pid": 76337, "tid": -914061504, "ts": 1716454225823443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225874447, "dur": 125,
    "args": {
      "External id": 275145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275145,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275145, "pid": 5, "tid": 7, "ts": 1716454225874447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823473, "dur": 8,
    "args": {
      "External id": 275145,
      "cbid": 211, "correlation": 275145
    }
  },
  {
    "ph": "s", "id": 275145, "pid": 76337, "tid": -914061504, "ts": 1716454225823473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823553, "dur": 1,
    "args": {
      "External id": 275161,
      "cbid": 251, "correlation": 275161
    }
  },
  {
    "ph": "f", "id": 275161, "pid": 76337, "tid": -914061504, "ts": 1716454225823553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225874573, "dur": 300,
    "args": {
      "External id": 275163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275163,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275163, "pid": 5, "tid": 7, "ts": 1716454225874573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823558, "dur": 12,
    "args": {
      "External id": 275163,
      "cbid": 211, "correlation": 275163
    }
  },
  {
    "ph": "s", "id": 275163, "pid": 76337, "tid": -914061504, "ts": 1716454225823558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225874874, "dur": 27,
    "args": {
      "External id": 275171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275171,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275171, "pid": 5, "tid": 7, "ts": 1716454225874874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823602, "dur": 10,
    "args": {
      "External id": 275171,
      "cbid": 211, "correlation": 275171
    }
  },
  {
    "ph": "s", "id": 275171, "pid": 76337, "tid": -914061504, "ts": 1716454225823602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225874902, "dur": 78,
    "args": {
      "External id": 275182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275182,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275182, "pid": 5, "tid": 7, "ts": 1716454225874902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823668, "dur": 12,
    "args": {
      "External id": 275182,
      "cbid": 211, "correlation": 275182
    }
  },
  {
    "ph": "s", "id": 275182, "pid": 76337, "tid": -914061504, "ts": 1716454225823668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823732, "dur": 0,
    "args": {
      "External id": 275194,
      "cbid": 317, "correlation": 275194
    }
  },
  {
    "ph": "f", "id": 275194, "pid": 76337, "tid": -914061504, "ts": 1716454225823732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823733, "dur": 0,
    "args": {
      "External id": 275195,
      "cbid": 203, "correlation": 275195
    }
  },
  {
    "ph": "f", "id": 275195, "pid": 76337, "tid": -914061504, "ts": 1716454225823733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823734, "dur": 0,
    "args": {
      "External id": 275196,
      "cbid": 205, "correlation": 275196
    }
  },
  {
    "ph": "f", "id": 275196, "pid": 76337, "tid": -914061504, "ts": 1716454225823734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225874982, "dur": 23,
    "args": {
      "External id": 275200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275200,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275200, "pid": 5, "tid": 7, "ts": 1716454225874982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823752, "dur": 12,
    "args": {
      "External id": 275200,
      "cbid": 211, "correlation": 275200
    }
  },
  {
    "ph": "s", "id": 275200, "pid": 76337, "tid": -914061504, "ts": 1716454225823752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225875007, "dur": 115,
    "args": {
      "External id": 275202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275202,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275202, "pid": 5, "tid": 7, "ts": 1716454225875007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823770, "dur": 7,
    "args": {
      "External id": 275202,
      "cbid": 211, "correlation": 275202
    }
  },
  {
    "ph": "s", "id": 275202, "pid": 76337, "tid": -914061504, "ts": 1716454225823770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875123, "dur": 22,
    "args": {
      "External id": 275204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275204,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275204, "pid": 5, "tid": 7, "ts": 1716454225875123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823781, "dur": 6,
    "args": {
      "External id": 275204,
      "cbid": 211, "correlation": 275204
    }
  },
  {
    "ph": "s", "id": 275204, "pid": 76337, "tid": -914061504, "ts": 1716454225823781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225875147, "dur": 31,
    "args": {
      "External id": 275210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275210, "pid": 5, "tid": 7, "ts": 1716454225875147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823809, "dur": 8,
    "args": {
      "External id": 275210,
      "cbid": 211, "correlation": 275210
    }
  },
  {
    "ph": "s", "id": 275210, "pid": 76337, "tid": -914061504, "ts": 1716454225823809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225875179, "dur": 27,
    "args": {
      "External id": 275218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275218,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275218, "pid": 5, "tid": 7, "ts": 1716454225875179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823842, "dur": 8,
    "args": {
      "External id": 275218,
      "cbid": 211, "correlation": 275218
    }
  },
  {
    "ph": "s", "id": 275218, "pid": 76337, "tid": -914061504, "ts": 1716454225823842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823919, "dur": 0,
    "args": {
      "External id": 275228,
      "cbid": 317, "correlation": 275228
    }
  },
  {
    "ph": "f", "id": 275228, "pid": 76337, "tid": -914061504, "ts": 1716454225823919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823919, "dur": 0,
    "args": {
      "External id": 275229,
      "cbid": 203, "correlation": 275229
    }
  },
  {
    "ph": "f", "id": 275229, "pid": 76337, "tid": -914061504, "ts": 1716454225823919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823920, "dur": 0,
    "args": {
      "External id": 275230,
      "cbid": 205, "correlation": 275230
    }
  },
  {
    "ph": "f", "id": 275230, "pid": 76337, "tid": -914061504, "ts": 1716454225823920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875207, "dur": 23,
    "args": {
      "External id": 275234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275234, "pid": 5, "tid": 7, "ts": 1716454225875207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823935, "dur": 13,
    "args": {
      "External id": 275234,
      "cbid": 211, "correlation": 275234
    }
  },
  {
    "ph": "s", "id": 275234, "pid": 76337, "tid": -914061504, "ts": 1716454225823935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875232, "dur": 42,
    "args": {
      "External id": 275236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275236, "pid": 5, "tid": 7, "ts": 1716454225875232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823950, "dur": 5,
    "args": {
      "External id": 275236,
      "cbid": 211, "correlation": 275236
    }
  },
  {
    "ph": "s", "id": 275236, "pid": 76337, "tid": -914061504, "ts": 1716454225823950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225875275, "dur": 227,
    "args": {
      "External id": 275238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275238,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 275238, "pid": 5, "tid": 7, "ts": 1716454225875275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823962, "dur": 7,
    "args": {
      "External id": 275238,
      "cbid": 211, "correlation": 275238
    }
  },
  {
    "ph": "s", "id": 275238, "pid": 76337, "tid": -914061504, "ts": 1716454225823962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875504, "dur": 6,
    "args": {
      "External id": 275240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275240, "pid": 5, "tid": 7, "ts": 1716454225875504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225823973, "dur": 13,
    "args": {
      "External id": 275240,
      "cbid": 211, "correlation": 275240
    }
  },
  {
    "ph": "s", "id": 275240, "pid": 76337, "tid": -914061504, "ts": 1716454225823973,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225875511, "dur": 9,
    "args": {
      "External id": 275246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275246, "pid": 5, "tid": 7, "ts": 1716454225875511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824008, "dur": 8,
    "args": {
      "External id": 275246,
      "cbid": 211, "correlation": 275246
    }
  },
  {
    "ph": "s", "id": 275246, "pid": 76337, "tid": -914061504, "ts": 1716454225824008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875522, "dur": 11,
    "args": {
      "External id": 275266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275266,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 275266, "pid": 5, "tid": 7, "ts": 1716454225875522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824101, "dur": 12,
    "args": {
      "External id": 275266,
      "cbid": 211, "correlation": 275266
    }
  },
  {
    "ph": "s", "id": 275266, "pid": 76337, "tid": -914061504, "ts": 1716454225824101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225875534, "dur": 4,
    "args": {
      "External id": 275278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275278,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 275278, "pid": 5, "tid": 7, "ts": 1716454225875534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824127, "dur": 7,
    "args": {
      "External id": 275278,
      "cbid": 211, "correlation": 275278
    }
  },
  {
    "ph": "s", "id": 275278, "pid": 76337, "tid": -914061504, "ts": 1716454225824127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225875540, "dur": 12,
    "args": {
      "External id": 275281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275281, "pid": 5, "tid": 7, "ts": 1716454225875540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824146, "dur": 7,
    "args": {
      "External id": 275281,
      "cbid": 211, "correlation": 275281
    }
  },
  {
    "ph": "s", "id": 275281, "pid": 76337, "tid": -914061504, "ts": 1716454225824146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225875553, "dur": 6,
    "args": {
      "External id": 275290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275290,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275290, "pid": 5, "tid": 7, "ts": 1716454225875553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824186, "dur": 9,
    "args": {
      "External id": 275290,
      "cbid": 211, "correlation": 275290
    }
  },
  {
    "ph": "s", "id": 275290, "pid": 76337, "tid": -914061504, "ts": 1716454225824186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824239, "dur": 0,
    "args": {
      "External id": 275300,
      "cbid": 317, "correlation": 275300
    }
  },
  {
    "ph": "f", "id": 275300, "pid": 76337, "tid": -914061504, "ts": 1716454225824239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824239, "dur": 0,
    "args": {
      "External id": 275301,
      "cbid": 203, "correlation": 275301
    }
  },
  {
    "ph": "f", "id": 275301, "pid": 76337, "tid": -914061504, "ts": 1716454225824239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824240, "dur": 0,
    "args": {
      "External id": 275302,
      "cbid": 205, "correlation": 275302
    }
  },
  {
    "ph": "f", "id": 275302, "pid": 76337, "tid": -914061504, "ts": 1716454225824240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875560, "dur": 5,
    "args": {
      "External id": 275306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275306,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275306, "pid": 5, "tid": 7, "ts": 1716454225875560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824259, "dur": 11,
    "args": {
      "External id": 275306,
      "cbid": 211, "correlation": 275306
    }
  },
  {
    "ph": "s", "id": 275306, "pid": 76337, "tid": -914061504, "ts": 1716454225824259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225875567, "dur": 80,
    "args": {
      "External id": 275308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275308,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275308, "pid": 5, "tid": 7, "ts": 1716454225875567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824274, "dur": 5,
    "args": {
      "External id": 275308,
      "cbid": 211, "correlation": 275308
    }
  },
  {
    "ph": "s", "id": 275308, "pid": 76337, "tid": -914061504, "ts": 1716454225824274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225875649, "dur": 1,
    "args": {
      "External id": 275310,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 275310,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 275310, "pid": 5, "tid": 7, "ts": 1716454225875649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824287, "dur": 8,
    "args": {
      "External id": 275310,
      "cbid": 51, "correlation": 275310
    }
  },
  {
    "ph": "s", "id": 275310, "pid": 76337, "tid": -914061504, "ts": 1716454225824287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225875652, "dur": 525,
    "args": {
      "External id": 275311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275311,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275311, "pid": 5, "tid": 7, "ts": 1716454225875652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824299, "dur": 9,
    "args": {
      "External id": 275311,
      "cbid": 211, "correlation": 275311
    }
  },
  {
    "ph": "s", "id": 275311, "pid": 76337, "tid": -914061504, "ts": 1716454225824299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225876179, "dur": 11,
    "args": {
      "External id": 275313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275313, "pid": 5, "tid": 7, "ts": 1716454225876179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824312, "dur": 5,
    "args": {
      "External id": 275313,
      "cbid": 211, "correlation": 275313
    }
  },
  {
    "ph": "s", "id": 275313, "pid": 76337, "tid": -914061504, "ts": 1716454225824312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225876191, "dur": 14,
    "args": {
      "External id": 275319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275319,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275319, "pid": 5, "tid": 7, "ts": 1716454225876191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824340, "dur": 8,
    "args": {
      "External id": 275319,
      "cbid": 211, "correlation": 275319
    }
  },
  {
    "ph": "s", "id": 275319, "pid": 76337, "tid": -914061504, "ts": 1716454225824340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225876206, "dur": 4,
    "args": {
      "External id": 275327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275327,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 275327, "pid": 5, "tid": 7, "ts": 1716454225876206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824384, "dur": 9,
    "args": {
      "External id": 275327,
      "cbid": 211, "correlation": 275327
    }
  },
  {
    "ph": "s", "id": 275327, "pid": 76337, "tid": -914061504, "ts": 1716454225824384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824452, "dur": 1,
    "args": {
      "External id": 275343,
      "cbid": 251, "correlation": 275343
    }
  },
  {
    "ph": "f", "id": 275343, "pid": 76337, "tid": -914061504, "ts": 1716454225824452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824457, "dur": 0,
    "args": {
      "External id": 275345,
      "cbid": 251, "correlation": 275345
    }
  },
  {
    "ph": "f", "id": 275345, "pid": 76337, "tid": -914061504, "ts": 1716454225824457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225876211, "dur": 13,
    "args": {
      "External id": 275346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275346,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275346, "pid": 5, "tid": 7, "ts": 1716454225876211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824459, "dur": 12,
    "args": {
      "External id": 275346,
      "cbid": 211, "correlation": 275346
    }
  },
  {
    "ph": "s", "id": 275346, "pid": 76337, "tid": -914061504, "ts": 1716454225824459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225876226, "dur": 5,
    "args": {
      "External id": 275348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275348,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275348, "pid": 5, "tid": 7, "ts": 1716454225876226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824472, "dur": 6,
    "args": {
      "External id": 275348,
      "cbid": 211, "correlation": 275348
    }
  },
  {
    "ph": "s", "id": 275348, "pid": 76337, "tid": -914061504, "ts": 1716454225824472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225876232, "dur": 17,
    "args": {
      "External id": 275358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275358, "pid": 5, "tid": 7, "ts": 1716454225876232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824531, "dur": 12,
    "args": {
      "External id": 275358,
      "cbid": 211, "correlation": 275358
    }
  },
  {
    "ph": "s", "id": 275358, "pid": 76337, "tid": -914061504, "ts": 1716454225824531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225876250, "dur": 18,
    "args": {
      "External id": 275378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275378,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 275378, "pid": 5, "tid": 7, "ts": 1716454225876250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824596, "dur": 10,
    "args": {
      "External id": 275378,
      "cbid": 211, "correlation": 275378
    }
  },
  {
    "ph": "s", "id": 275378, "pid": 76337, "tid": -914061504, "ts": 1716454225824596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225876270, "dur": 4,
    "args": {
      "External id": 275390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275390,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 275390, "pid": 5, "tid": 7, "ts": 1716454225876270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824620, "dur": 7,
    "args": {
      "External id": 275390,
      "cbid": 211, "correlation": 275390
    }
  },
  {
    "ph": "s", "id": 275390, "pid": 76337, "tid": -914061504, "ts": 1716454225824620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225876276, "dur": 16,
    "args": {
      "External id": 275393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275393,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275393, "pid": 5, "tid": 7, "ts": 1716454225876276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824641, "dur": 6,
    "args": {
      "External id": 275393,
      "cbid": 211, "correlation": 275393
    }
  },
  {
    "ph": "s", "id": 275393, "pid": 76337, "tid": -914061504, "ts": 1716454225824641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225876293, "dur": 10,
    "args": {
      "External id": 275402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275402,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275402, "pid": 5, "tid": 7, "ts": 1716454225876293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824682, "dur": 9,
    "args": {
      "External id": 275402,
      "cbid": 211, "correlation": 275402
    }
  },
  {
    "ph": "s", "id": 275402, "pid": 76337, "tid": -914061504, "ts": 1716454225824682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824745, "dur": 0,
    "args": {
      "External id": 275412,
      "cbid": 317, "correlation": 275412
    }
  },
  {
    "ph": "f", "id": 275412, "pid": 76337, "tid": -914061504, "ts": 1716454225824745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824746, "dur": 0,
    "args": {
      "External id": 275413,
      "cbid": 203, "correlation": 275413
    }
  },
  {
    "ph": "f", "id": 275413, "pid": 76337, "tid": -914061504, "ts": 1716454225824746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824747, "dur": 0,
    "args": {
      "External id": 275414,
      "cbid": 205, "correlation": 275414
    }
  },
  {
    "ph": "f", "id": 275414, "pid": 76337, "tid": -914061504, "ts": 1716454225824747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225876305, "dur": 11,
    "args": {
      "External id": 275418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275418,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275418, "pid": 5, "tid": 7, "ts": 1716454225876305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824764, "dur": 13,
    "args": {
      "External id": 275418,
      "cbid": 211, "correlation": 275418
    }
  },
  {
    "ph": "s", "id": 275418, "pid": 76337, "tid": -914061504, "ts": 1716454225824764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225876317, "dur": 156,
    "args": {
      "External id": 275420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275420, "pid": 5, "tid": 7, "ts": 1716454225876317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824779, "dur": 5,
    "args": {
      "External id": 275420,
      "cbid": 211, "correlation": 275420
    }
  },
  {
    "ph": "s", "id": 275420, "pid": 76337, "tid": -914061504, "ts": 1716454225824779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225876475, "dur": 1,
    "args": {
      "External id": 275422,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 275422,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 275422, "pid": 5, "tid": 7, "ts": 1716454225876475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824791, "dur": 7,
    "args": {
      "External id": 275422,
      "cbid": 51, "correlation": 275422
    }
  },
  {
    "ph": "s", "id": 275422, "pid": 76337, "tid": -914061504, "ts": 1716454225824791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225876479, "dur": 647,
    "args": {
      "External id": 275423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275423,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275423, "pid": 5, "tid": 7, "ts": 1716454225876479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824799, "dur": 6,
    "args": {
      "External id": 275423,
      "cbid": 211, "correlation": 275423
    }
  },
  {
    "ph": "s", "id": 275423, "pid": 76337, "tid": -914061504, "ts": 1716454225824799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225877127, "dur": 13,
    "args": {
      "External id": 275425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275425,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275425, "pid": 5, "tid": 7, "ts": 1716454225877127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824810, "dur": 5,
    "args": {
      "External id": 275425,
      "cbid": 211, "correlation": 275425
    }
  },
  {
    "ph": "s", "id": 275425, "pid": 76337, "tid": -914061504, "ts": 1716454225824810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225877141, "dur": 14,
    "args": {
      "External id": 275431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275431,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275431, "pid": 5, "tid": 7, "ts": 1716454225877141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824839, "dur": 8,
    "args": {
      "External id": 275431,
      "cbid": 211, "correlation": 275431
    }
  },
  {
    "ph": "s", "id": 275431, "pid": 76337, "tid": -914061504, "ts": 1716454225824839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824896, "dur": 0,
    "args": {
      "External id": 275441,
      "cbid": 317, "correlation": 275441
    }
  },
  {
    "ph": "f", "id": 275441, "pid": 76337, "tid": -914061504, "ts": 1716454225824896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824897, "dur": 0,
    "args": {
      "External id": 275442,
      "cbid": 203, "correlation": 275442
    }
  },
  {
    "ph": "f", "id": 275442, "pid": 76337, "tid": -914061504, "ts": 1716454225824897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824898, "dur": 0,
    "args": {
      "External id": 275443,
      "cbid": 205, "correlation": 275443
    }
  },
  {
    "ph": "f", "id": 275443, "pid": 76337, "tid": -914061504, "ts": 1716454225824898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225877156, "dur": 8,
    "args": {
      "External id": 275447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275447,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275447, "pid": 5, "tid": 7, "ts": 1716454225877156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824913, "dur": 12,
    "args": {
      "External id": 275447,
      "cbid": 211, "correlation": 275447
    }
  },
  {
    "ph": "s", "id": 275447, "pid": 76337, "tid": -914061504, "ts": 1716454225824913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225877166, "dur": 3,
    "args": {
      "External id": 275449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275449, "pid": 5, "tid": 7, "ts": 1716454225877166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824932, "dur": 6,
    "args": {
      "External id": 275449,
      "cbid": 211, "correlation": 275449
    }
  },
  {
    "ph": "s", "id": 275449, "pid": 76337, "tid": -914061504, "ts": 1716454225824932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824941, "dur": 0,
    "args": {
      "External id": 275450,
      "cbid": 51, "correlation": 275450
    }
  },
  {
    "ph": "s", "id": 275450, "pid": 76337, "tid": -914061504, "ts": 1716454225824941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225877170, "dur": 54,
    "args": {
      "External id": 275451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275451,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 275451, "pid": 5, "tid": 7, "ts": 1716454225877170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824942, "dur": 5,
    "args": {
      "External id": 275451,
      "cbid": 211, "correlation": 275451
    }
  },
  {
    "ph": "s", "id": 275451, "pid": 76337, "tid": -914061504, "ts": 1716454225824942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225877226, "dur": 13,
    "args": {
      "External id": 275456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275456, "pid": 5, "tid": 7, "ts": 1716454225877226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225824970, "dur": 17,
    "args": {
      "External id": 275456,
      "cbid": 211, "correlation": 275456
    }
  },
  {
    "ph": "s", "id": 275456, "pid": 76337, "tid": -914061504, "ts": 1716454225824970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225877241, "dur": 12,
    "args": {
      "External id": 275464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275464,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275464, "pid": 5, "tid": 7, "ts": 1716454225877241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825009, "dur": 8,
    "args": {
      "External id": 275464,
      "cbid": 211, "correlation": 275464
    }
  },
  {
    "ph": "s", "id": 275464, "pid": 76337, "tid": -914061504, "ts": 1716454225825009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225877254, "dur": 10,
    "args": {
      "External id": 275472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275472,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275472, "pid": 5, "tid": 7, "ts": 1716454225877254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825038, "dur": 8,
    "args": {
      "External id": 275472,
      "cbid": 211, "correlation": 275472
    }
  },
  {
    "ph": "s", "id": 275472, "pid": 76337, "tid": -914061504, "ts": 1716454225825038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225877265, "dur": 18,
    "args": {
      "External id": 275492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275492,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 275492, "pid": 5, "tid": 7, "ts": 1716454225877265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825119, "dur": 13,
    "args": {
      "External id": 275492,
      "cbid": 211, "correlation": 275492
    }
  },
  {
    "ph": "s", "id": 275492, "pid": 76337, "tid": -914061504, "ts": 1716454225825119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225877284, "dur": 4,
    "args": {
      "External id": 275504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275504,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 275504, "pid": 5, "tid": 7, "ts": 1716454225877284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825142, "dur": 6,
    "args": {
      "External id": 275504,
      "cbid": 211, "correlation": 275504
    }
  },
  {
    "ph": "s", "id": 275504, "pid": 76337, "tid": -914061504, "ts": 1716454225825142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225877289, "dur": 16,
    "args": {
      "External id": 275507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275507, "pid": 5, "tid": 7, "ts": 1716454225877289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825159, "dur": 6,
    "args": {
      "External id": 275507,
      "cbid": 211, "correlation": 275507
    }
  },
  {
    "ph": "s", "id": 275507, "pid": 76337, "tid": -914061504, "ts": 1716454225825159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825216, "dur": 0,
    "args": {
      "External id": 275518,
      "cbid": 317, "correlation": 275518
    }
  },
  {
    "ph": "f", "id": 275518, "pid": 76337, "tid": -914061504, "ts": 1716454225825216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825216, "dur": 0,
    "args": {
      "External id": 275519,
      "cbid": 203, "correlation": 275519
    }
  },
  {
    "ph": "f", "id": 275519, "pid": 76337, "tid": -914061504, "ts": 1716454225825216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825217, "dur": 0,
    "args": {
      "External id": 275520,
      "cbid": 205, "correlation": 275520
    }
  },
  {
    "ph": "f", "id": 275520, "pid": 76337, "tid": -914061504, "ts": 1716454225825217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225877307, "dur": 11,
    "args": {
      "External id": 275524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275524, "pid": 5, "tid": 7, "ts": 1716454225877307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825234, "dur": 11,
    "args": {
      "External id": 275524,
      "cbid": 211, "correlation": 275524
    }
  },
  {
    "ph": "s", "id": 275524, "pid": 76337, "tid": -914061504, "ts": 1716454225825234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225877320, "dur": 3,
    "args": {
      "External id": 275526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275526, "pid": 5, "tid": 7, "ts": 1716454225877320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825251, "dur": 6,
    "args": {
      "External id": 275526,
      "cbid": 211, "correlation": 275526
    }
  },
  {
    "ph": "s", "id": 275526, "pid": 76337, "tid": -914061504, "ts": 1716454225825251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825259, "dur": 0,
    "args": {
      "External id": 275527,
      "cbid": 51, "correlation": 275527
    }
  },
  {
    "ph": "s", "id": 275527, "pid": 76337, "tid": -914061504, "ts": 1716454225825259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225877324, "dur": 94,
    "args": {
      "External id": 275528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275528,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 275528, "pid": 5, "tid": 7, "ts": 1716454225877324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825260, "dur": 6,
    "args": {
      "External id": 275528,
      "cbid": 211, "correlation": 275528
    }
  },
  {
    "ph": "s", "id": 275528, "pid": 76337, "tid": -914061504, "ts": 1716454225825260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225877420, "dur": 15,
    "args": {
      "External id": 275533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275533, "pid": 5, "tid": 7, "ts": 1716454225877420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825289, "dur": 9,
    "args": {
      "External id": 275533,
      "cbid": 211, "correlation": 275533
    }
  },
  {
    "ph": "s", "id": 275533, "pid": 76337, "tid": -914061504, "ts": 1716454225825289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225877436, "dur": 80,
    "args": {
      "External id": 275542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275542,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275542, "pid": 5, "tid": 7, "ts": 1716454225877436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825371, "dur": 14,
    "args": {
      "External id": 275542,
      "cbid": 211, "correlation": 275542
    }
  },
  {
    "ph": "s", "id": 275542, "pid": 76337, "tid": -914061504, "ts": 1716454225825371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225877518, "dur": 30,
    "args": {
      "External id": 275564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275564,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275564, "pid": 5, "tid": 7, "ts": 1716454225877518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825428, "dur": 10,
    "args": {
      "External id": 275564,
      "cbid": 211, "correlation": 275564
    }
  },
  {
    "ph": "s", "id": 275564, "pid": 76337, "tid": -914061504, "ts": 1716454225825428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825517, "dur": 2,
    "args": {
      "External id": 275575,
      "cbid": 251, "correlation": 275575
    }
  },
  {
    "ph": "f", "id": 275575, "pid": 76337, "tid": -914061504, "ts": 1716454225825517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225877549, "dur": 159,
    "args": {
      "External id": 275576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275576,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275576, "pid": 5, "tid": 7, "ts": 1716454225877549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825523, "dur": 14,
    "args": {
      "External id": 275576,
      "cbid": 211, "correlation": 275576
    }
  },
  {
    "ph": "s", "id": 275576, "pid": 76337, "tid": -914061504, "ts": 1716454225825523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825594, "dur": 1,
    "args": {
      "External id": 275587,
      "cbid": 251, "correlation": 275587
    }
  },
  {
    "ph": "f", "id": 275587, "pid": 76337, "tid": -914061504, "ts": 1716454225825594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225877709, "dur": 153,
    "args": {
      "External id": 275588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275588,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275588, "pid": 5, "tid": 7, "ts": 1716454225877709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825599, "dur": 11,
    "args": {
      "External id": 275588,
      "cbid": 211, "correlation": 275588
    }
  },
  {
    "ph": "s", "id": 275588, "pid": 76337, "tid": -914061504, "ts": 1716454225825599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825664, "dur": 1,
    "args": {
      "External id": 275599,
      "cbid": 251, "correlation": 275599
    }
  },
  {
    "ph": "f", "id": 275599, "pid": 76337, "tid": -914061504, "ts": 1716454225825664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225877864, "dur": 152,
    "args": {
      "External id": 275600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275600,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275600, "pid": 5, "tid": 7, "ts": 1716454225877864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825668, "dur": 11,
    "args": {
      "External id": 275600,
      "cbid": 211, "correlation": 275600
    }
  },
  {
    "ph": "s", "id": 275600, "pid": 76337, "tid": -914061504, "ts": 1716454225825668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225878017, "dur": 326,
    "args": {
      "External id": 275625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275625,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275625, "pid": 5, "tid": 7, "ts": 1716454225878017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825753, "dur": 14,
    "args": {
      "External id": 275625,
      "cbid": 211, "correlation": 275625
    }
  },
  {
    "ph": "s", "id": 275625, "pid": 76337, "tid": -914061504, "ts": 1716454225825753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825857, "dur": 1,
    "args": {
      "External id": 275643,
      "cbid": 251, "correlation": 275643
    }
  },
  {
    "ph": "f", "id": 275643, "pid": 76337, "tid": -914061504, "ts": 1716454225825857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225878344, "dur": 160,
    "args": {
      "External id": 275645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275645,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275645, "pid": 5, "tid": 7, "ts": 1716454225878344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825862, "dur": 14,
    "args": {
      "External id": 275645,
      "cbid": 211, "correlation": 275645
    }
  },
  {
    "ph": "s", "id": 275645, "pid": 76337, "tid": -914061504, "ts": 1716454225825862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225878506, "dur": 19,
    "args": {
      "External id": 275653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275653,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275653, "pid": 5, "tid": 7, "ts": 1716454225878506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825932, "dur": 12,
    "args": {
      "External id": 275653,
      "cbid": 211, "correlation": 275653
    }
  },
  {
    "ph": "s", "id": 275653, "pid": 76337, "tid": -914061504, "ts": 1716454225825932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225878526, "dur": 27,
    "args": {
      "External id": 275661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275661,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275661, "pid": 5, "tid": 7, "ts": 1716454225878526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225825972, "dur": 17,
    "args": {
      "External id": 275661,
      "cbid": 211, "correlation": 275661
    }
  },
  {
    "ph": "s", "id": 275661, "pid": 76337, "tid": -914061504, "ts": 1716454225825972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225878555, "dur": 17,
    "args": {
      "External id": 275672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275672, "pid": 5, "tid": 7, "ts": 1716454225878555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826053, "dur": 13,
    "args": {
      "External id": 275672,
      "cbid": 211, "correlation": 275672
    }
  },
  {
    "ph": "s", "id": 275672, "pid": 76337, "tid": -914061504, "ts": 1716454225826053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225878573, "dur": 15,
    "args": {
      "External id": 275694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275694,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275694, "pid": 5, "tid": 7, "ts": 1716454225878573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826084, "dur": 7,
    "args": {
      "External id": 275694,
      "cbid": 211, "correlation": 275694
    }
  },
  {
    "ph": "s", "id": 275694, "pid": 76337, "tid": -914061504, "ts": 1716454225826084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826169, "dur": 2,
    "args": {
      "External id": 275705,
      "cbid": 251, "correlation": 275705
    }
  },
  {
    "ph": "f", "id": 275705, "pid": 76337, "tid": -914061504, "ts": 1716454225826169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225878590, "dur": 85,
    "args": {
      "External id": 275706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275706,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275706, "pid": 5, "tid": 7, "ts": 1716454225878590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826176, "dur": 14,
    "args": {
      "External id": 275706,
      "cbid": 211, "correlation": 275706
    }
  },
  {
    "ph": "s", "id": 275706, "pid": 76337, "tid": -914061504, "ts": 1716454225826176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826246, "dur": 1,
    "args": {
      "External id": 275717,
      "cbid": 251, "correlation": 275717
    }
  },
  {
    "ph": "f", "id": 275717, "pid": 76337, "tid": -914061504, "ts": 1716454225826246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826250, "dur": 0,
    "args": {
      "External id": 275718,
      "cbid": 251, "correlation": 275718
    }
  },
  {
    "ph": "f", "id": 275718, "pid": 76337, "tid": -914061504, "ts": 1716454225826250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225878677, "dur": 11,
    "args": {
      "External id": 275719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275719,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275719, "pid": 5, "tid": 7, "ts": 1716454225878677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826251, "dur": 12,
    "args": {
      "External id": 275719,
      "cbid": 211, "correlation": 275719
    }
  },
  {
    "ph": "s", "id": 275719, "pid": 76337, "tid": -914061504, "ts": 1716454225826251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225878689, "dur": 6,
    "args": {
      "External id": 275721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275721,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275721, "pid": 5, "tid": 7, "ts": 1716454225878689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826265, "dur": 7,
    "args": {
      "External id": 275721,
      "cbid": 211, "correlation": 275721
    }
  },
  {
    "ph": "s", "id": 275721, "pid": 76337, "tid": -914061504, "ts": 1716454225826265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826323, "dur": 1,
    "args": {
      "External id": 275732,
      "cbid": 251, "correlation": 275732
    }
  },
  {
    "ph": "f", "id": 275732, "pid": 76337, "tid": -914061504, "ts": 1716454225826323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826326, "dur": 0,
    "args": {
      "External id": 275733,
      "cbid": 251, "correlation": 275733
    }
  },
  {
    "ph": "f", "id": 275733, "pid": 76337, "tid": -914061504, "ts": 1716454225826326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225878696, "dur": 8,
    "args": {
      "External id": 275734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275734,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275734, "pid": 5, "tid": 7, "ts": 1716454225878696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826328, "dur": 11,
    "args": {
      "External id": 275734,
      "cbid": 211, "correlation": 275734
    }
  },
  {
    "ph": "s", "id": 275734, "pid": 76337, "tid": -914061504, "ts": 1716454225826328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225878706, "dur": 3,
    "args": {
      "External id": 275736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275736,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275736, "pid": 5, "tid": 7, "ts": 1716454225878706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826340, "dur": 5,
    "args": {
      "External id": 275736,
      "cbid": 211, "correlation": 275736
    }
  },
  {
    "ph": "s", "id": 275736, "pid": 76337, "tid": -914061504, "ts": 1716454225826340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225878710, "dur": 52,
    "args": {
      "External id": 275761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275761,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275761, "pid": 5, "tid": 7, "ts": 1716454225878710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826417, "dur": 12,
    "args": {
      "External id": 275761,
      "cbid": 211, "correlation": 275761
    }
  },
  {
    "ph": "s", "id": 275761, "pid": 76337, "tid": -914061504, "ts": 1716454225826417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826514, "dur": 6,
    "args": {
      "External id": 275779,
      "cbid": 251, "correlation": 275779
    }
  },
  {
    "ph": "f", "id": 275779, "pid": 76337, "tid": -914061504, "ts": 1716454225826514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225878763, "dur": 88,
    "args": {
      "External id": 275781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275781,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275781, "pid": 5, "tid": 7, "ts": 1716454225878763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826525, "dur": 15,
    "args": {
      "External id": 275781,
      "cbid": 211, "correlation": 275781
    }
  },
  {
    "ph": "s", "id": 275781, "pid": 76337, "tid": -914061504, "ts": 1716454225826525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225878853, "dur": 9,
    "args": {
      "External id": 275789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275789,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275789, "pid": 5, "tid": 7, "ts": 1716454225878853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826595, "dur": 11,
    "args": {
      "External id": 275789,
      "cbid": 211, "correlation": 275789
    }
  },
  {
    "ph": "s", "id": 275789, "pid": 76337, "tid": -914061504, "ts": 1716454225826595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225878863, "dur": 20,
    "args": {
      "External id": 275797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275797, "pid": 5, "tid": 7, "ts": 1716454225878863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826636, "dur": 9,
    "args": {
      "External id": 275797,
      "cbid": 211, "correlation": 275797
    }
  },
  {
    "ph": "s", "id": 275797, "pid": 76337, "tid": -914061504, "ts": 1716454225826636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225878885, "dur": 17,
    "args": {
      "External id": 275819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275819,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275819, "pid": 5, "tid": 7, "ts": 1716454225878885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826687, "dur": 10,
    "args": {
      "External id": 275819,
      "cbid": 211, "correlation": 275819
    }
  },
  {
    "ph": "s", "id": 275819, "pid": 76337, "tid": -914061504, "ts": 1716454225826687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826778, "dur": 2,
    "args": {
      "External id": 275835,
      "cbid": 251, "correlation": 275835
    }
  },
  {
    "ph": "f", "id": 275835, "pid": 76337, "tid": -914061504, "ts": 1716454225826778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826784, "dur": 0,
    "args": {
      "External id": 275837,
      "cbid": 251, "correlation": 275837
    }
  },
  {
    "ph": "f", "id": 275837, "pid": 76337, "tid": -914061504, "ts": 1716454225826784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225878903, "dur": 483,
    "args": {
      "External id": 275838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275838,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275838, "pid": 5, "tid": 7, "ts": 1716454225878903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826787, "dur": 15,
    "args": {
      "External id": 275838,
      "cbid": 211, "correlation": 275838
    }
  },
  {
    "ph": "s", "id": 275838, "pid": 76337, "tid": -914061504, "ts": 1716454225826787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225879387, "dur": 63,
    "args": {
      "External id": 275846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275846, "pid": 5, "tid": 7, "ts": 1716454225879387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826853, "dur": 12,
    "args": {
      "External id": 275846,
      "cbid": 211, "correlation": 275846
    }
  },
  {
    "ph": "s", "id": 275846, "pid": 76337, "tid": -914061504, "ts": 1716454225826853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225879452, "dur": 66,
    "args": {
      "External id": 275854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275854, "pid": 5, "tid": 7, "ts": 1716454225879452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826884, "dur": 8,
    "args": {
      "External id": 275854,
      "cbid": 211, "correlation": 275854
    }
  },
  {
    "ph": "s", "id": 275854, "pid": 76337, "tid": -914061504, "ts": 1716454225826884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826967, "dur": 1,
    "args": {
      "External id": 275870,
      "cbid": 251, "correlation": 275870
    }
  },
  {
    "ph": "f", "id": 275870, "pid": 76337, "tid": -914061504, "ts": 1716454225826967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225879521, "dur": 1,
    "args": {
      "External id": 275872,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 275872,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1596806387225549
    }
  },
  {
    "ph": "f", "id": 275872, "pid": 5, "tid": 7, "ts": 1716454225879521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826981, "dur": 15,
    "args": {
      "External id": 275872,
      "cbid": 51, "correlation": 275872
    }
  },
  {
    "ph": "s", "id": 275872, "pid": 76337, "tid": -914061504, "ts": 1716454225826981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225879524, "dur": 261,
    "args": {
      "External id": 275873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275873,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275873, "pid": 5, "tid": 7, "ts": 1716454225879524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225826998, "dur": 11,
    "args": {
      "External id": 275873,
      "cbid": 211, "correlation": 275873
    }
  },
  {
    "ph": "s", "id": 275873, "pid": 76337, "tid": -914061504, "ts": 1716454225826998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225879787, "dur": 14,
    "args": {
      "External id": 275881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275881,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275881, "pid": 5, "tid": 7, "ts": 1716454225879787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827041, "dur": 10,
    "args": {
      "External id": 275881,
      "cbid": 211, "correlation": 275881
    }
  },
  {
    "ph": "s", "id": 275881, "pid": 76337, "tid": -914061504, "ts": 1716454225827041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225879802, "dur": 36,
    "args": {
      "External id": 275892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275892, "pid": 5, "tid": 7, "ts": 1716454225879802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827109, "dur": 13,
    "args": {
      "External id": 275892,
      "cbid": 211, "correlation": 275892
    }
  },
  {
    "ph": "s", "id": 275892, "pid": 76337, "tid": -914061504, "ts": 1716454225827109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827174, "dur": 0,
    "args": {
      "External id": 275904,
      "cbid": 317, "correlation": 275904
    }
  },
  {
    "ph": "f", "id": 275904, "pid": 76337, "tid": -914061504, "ts": 1716454225827174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827175, "dur": 0,
    "args": {
      "External id": 275905,
      "cbid": 203, "correlation": 275905
    }
  },
  {
    "ph": "f", "id": 275905, "pid": 76337, "tid": -914061504, "ts": 1716454225827175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827176, "dur": 0,
    "args": {
      "External id": 275906,
      "cbid": 205, "correlation": 275906
    }
  },
  {
    "ph": "f", "id": 275906, "pid": 76337, "tid": -914061504, "ts": 1716454225827176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225879839, "dur": 13,
    "args": {
      "External id": 275910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275910,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275910, "pid": 5, "tid": 7, "ts": 1716454225879839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827194, "dur": 12,
    "args": {
      "External id": 275910,
      "cbid": 211, "correlation": 275910
    }
  },
  {
    "ph": "s", "id": 275910, "pid": 76337, "tid": -914061504, "ts": 1716454225827194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225879853, "dur": 4,
    "args": {
      "External id": 275912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 275912, "pid": 5, "tid": 7, "ts": 1716454225879853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827211, "dur": 7,
    "args": {
      "External id": 275912,
      "cbid": 211, "correlation": 275912
    }
  },
  {
    "ph": "s", "id": 275912, "pid": 76337, "tid": -914061504, "ts": 1716454225827211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827221, "dur": 0,
    "args": {
      "External id": 275913,
      "cbid": 51, "correlation": 275913
    }
  },
  {
    "ph": "s", "id": 275913, "pid": 76337, "tid": -914061504, "ts": 1716454225827221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225879857, "dur": 93,
    "args": {
      "External id": 275914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275914,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 275914, "pid": 5, "tid": 7, "ts": 1716454225879857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827222, "dur": 5,
    "args": {
      "External id": 275914,
      "cbid": 211, "correlation": 275914
    }
  },
  {
    "ph": "s", "id": 275914, "pid": 76337, "tid": -914061504, "ts": 1716454225827222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225879952, "dur": 15,
    "args": {
      "External id": 275919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275919,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275919, "pid": 5, "tid": 7, "ts": 1716454225879952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827252, "dur": 9,
    "args": {
      "External id": 275919,
      "cbid": 211, "correlation": 275919
    }
  },
  {
    "ph": "s", "id": 275919, "pid": 76337, "tid": -914061504, "ts": 1716454225827252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225879968, "dur": 12,
    "args": {
      "External id": 275927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275927,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275927, "pid": 5, "tid": 7, "ts": 1716454225879968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827284, "dur": 8,
    "args": {
      "External id": 275927,
      "cbid": 211, "correlation": 275927
    }
  },
  {
    "ph": "s", "id": 275927, "pid": 76337, "tid": -914061504, "ts": 1716454225827284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225879981, "dur": 17,
    "args": {
      "External id": 275947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275947,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 275947, "pid": 5, "tid": 7, "ts": 1716454225879981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827357, "dur": 13,
    "args": {
      "External id": 275947,
      "cbid": 211, "correlation": 275947
    }
  },
  {
    "ph": "s", "id": 275947, "pid": 76337, "tid": -914061504, "ts": 1716454225827357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225880000, "dur": 5,
    "args": {
      "External id": 275959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275959,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 275959, "pid": 5, "tid": 7, "ts": 1716454225880000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827380, "dur": 6,
    "args": {
      "External id": 275959,
      "cbid": 211, "correlation": 275959
    }
  },
  {
    "ph": "s", "id": 275959, "pid": 76337, "tid": -914061504, "ts": 1716454225827380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225880006, "dur": 17,
    "args": {
      "External id": 275962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275962, "pid": 5, "tid": 7, "ts": 1716454225880006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827397, "dur": 6,
    "args": {
      "External id": 275962,
      "cbid": 211, "correlation": 275962
    }
  },
  {
    "ph": "s", "id": 275962, "pid": 76337, "tid": -914061504, "ts": 1716454225827397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225880024, "dur": 13,
    "args": {
      "External id": 275971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275971,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275971, "pid": 5, "tid": 7, "ts": 1716454225880024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827436, "dur": 9,
    "args": {
      "External id": 275971,
      "cbid": 211, "correlation": 275971
    }
  },
  {
    "ph": "s", "id": 275971, "pid": 76337, "tid": -914061504, "ts": 1716454225827436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827486, "dur": 0,
    "args": {
      "External id": 275981,
      "cbid": 317, "correlation": 275981
    }
  },
  {
    "ph": "f", "id": 275981, "pid": 76337, "tid": -914061504, "ts": 1716454225827486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827487, "dur": 0,
    "args": {
      "External id": 275982,
      "cbid": 203, "correlation": 275982
    }
  },
  {
    "ph": "f", "id": 275982, "pid": 76337, "tid": -914061504, "ts": 1716454225827487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827488, "dur": 0,
    "args": {
      "External id": 275983,
      "cbid": 205, "correlation": 275983
    }
  },
  {
    "ph": "f", "id": 275983, "pid": 76337, "tid": -914061504, "ts": 1716454225827488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225880038, "dur": 11,
    "args": {
      "External id": 275987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275987, "pid": 5, "tid": 7, "ts": 1716454225880038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827501, "dur": 12,
    "args": {
      "External id": 275987,
      "cbid": 211, "correlation": 275987
    }
  },
  {
    "ph": "s", "id": 275987, "pid": 76337, "tid": -914061504, "ts": 1716454225827501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225880050, "dur": 157,
    "args": {
      "External id": 275989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275989,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275989, "pid": 5, "tid": 7, "ts": 1716454225880050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827515, "dur": 6,
    "args": {
      "External id": 275989,
      "cbid": 211, "correlation": 275989
    }
  },
  {
    "ph": "s", "id": 275989, "pid": 76337, "tid": -914061504, "ts": 1716454225827515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225880210, "dur": 1,
    "args": {
      "External id": 275991,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 275991,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 275991, "pid": 5, "tid": 7, "ts": 1716454225880210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827527, "dur": 8,
    "args": {
      "External id": 275991,
      "cbid": 51, "correlation": 275991
    }
  },
  {
    "ph": "s", "id": 275991, "pid": 76337, "tid": -914061504, "ts": 1716454225827527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225880213, "dur": 648,
    "args": {
      "External id": 275992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275992,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 275992, "pid": 5, "tid": 7, "ts": 1716454225880213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827536, "dur": 6,
    "args": {
      "External id": 275992,
      "cbid": 211, "correlation": 275992
    }
  },
  {
    "ph": "s", "id": 275992, "pid": 76337, "tid": -914061504, "ts": 1716454225827536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225880862, "dur": 13,
    "args": {
      "External id": 275994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 275994,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 275994, "pid": 5, "tid": 7, "ts": 1716454225880862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827546, "dur": 5,
    "args": {
      "External id": 275994,
      "cbid": 211, "correlation": 275994
    }
  },
  {
    "ph": "s", "id": 275994, "pid": 76337, "tid": -914061504, "ts": 1716454225827546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225880876, "dur": 14,
    "args": {
      "External id": 276000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276000,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276000, "pid": 5, "tid": 7, "ts": 1716454225880876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827579, "dur": 9,
    "args": {
      "External id": 276000,
      "cbid": 211, "correlation": 276000
    }
  },
  {
    "ph": "s", "id": 276000, "pid": 76337, "tid": -914061504, "ts": 1716454225827579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225880891, "dur": 4,
    "args": {
      "External id": 276008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276008,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 276008, "pid": 5, "tid": 7, "ts": 1716454225880891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827623, "dur": 9,
    "args": {
      "External id": 276008,
      "cbid": 211, "correlation": 276008
    }
  },
  {
    "ph": "s", "id": 276008, "pid": 76337, "tid": -914061504, "ts": 1716454225827623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827690, "dur": 1,
    "args": {
      "External id": 276024,
      "cbid": 251, "correlation": 276024
    }
  },
  {
    "ph": "f", "id": 276024, "pid": 76337, "tid": -914061504, "ts": 1716454225827690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827695, "dur": 0,
    "args": {
      "External id": 276026,
      "cbid": 251, "correlation": 276026
    }
  },
  {
    "ph": "f", "id": 276026, "pid": 76337, "tid": -914061504, "ts": 1716454225827695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225880897, "dur": 13,
    "args": {
      "External id": 276027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276027,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276027, "pid": 5, "tid": 7, "ts": 1716454225880897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827697, "dur": 12,
    "args": {
      "External id": 276027,
      "cbid": 211, "correlation": 276027
    }
  },
  {
    "ph": "s", "id": 276027, "pid": 76337, "tid": -914061504, "ts": 1716454225827697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225880911, "dur": 5,
    "args": {
      "External id": 276029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276029,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276029, "pid": 5, "tid": 7, "ts": 1716454225880911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827711, "dur": 5,
    "args": {
      "External id": 276029,
      "cbid": 211, "correlation": 276029
    }
  },
  {
    "ph": "s", "id": 276029, "pid": 76337, "tid": -914061504, "ts": 1716454225827711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225880917, "dur": 17,
    "args": {
      "External id": 276039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276039, "pid": 5, "tid": 7, "ts": 1716454225880917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827769, "dur": 12,
    "args": {
      "External id": 276039,
      "cbid": 211, "correlation": 276039
    }
  },
  {
    "ph": "s", "id": 276039, "pid": 76337, "tid": -914061504, "ts": 1716454225827769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225880936, "dur": 17,
    "args": {
      "External id": 276059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276059,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 276059, "pid": 5, "tid": 7, "ts": 1716454225880936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827835, "dur": 10,
    "args": {
      "External id": 276059,
      "cbid": 211, "correlation": 276059
    }
  },
  {
    "ph": "s", "id": 276059, "pid": 76337, "tid": -914061504, "ts": 1716454225827835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225880954, "dur": 4,
    "args": {
      "External id": 276071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276071,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 276071, "pid": 5, "tid": 7, "ts": 1716454225880954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827855, "dur": 6,
    "args": {
      "External id": 276071,
      "cbid": 211, "correlation": 276071
    }
  },
  {
    "ph": "s", "id": 276071, "pid": 76337, "tid": -914061504, "ts": 1716454225827855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225880959, "dur": 16,
    "args": {
      "External id": 276074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276074, "pid": 5, "tid": 7, "ts": 1716454225880959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827875, "dur": 6,
    "args": {
      "External id": 276074,
      "cbid": 211, "correlation": 276074
    }
  },
  {
    "ph": "s", "id": 276074, "pid": 76337, "tid": -914061504, "ts": 1716454225827875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225880976, "dur": 11,
    "args": {
      "External id": 276083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276083,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276083, "pid": 5, "tid": 7, "ts": 1716454225880976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827914, "dur": 11,
    "args": {
      "External id": 276083,
      "cbid": 211, "correlation": 276083
    }
  },
  {
    "ph": "s", "id": 276083, "pid": 76337, "tid": -914061504, "ts": 1716454225827914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827983, "dur": 0,
    "args": {
      "External id": 276093,
      "cbid": 317, "correlation": 276093
    }
  },
  {
    "ph": "f", "id": 276093, "pid": 76337, "tid": -914061504, "ts": 1716454225827983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827984, "dur": 0,
    "args": {
      "External id": 276094,
      "cbid": 203, "correlation": 276094
    }
  },
  {
    "ph": "f", "id": 276094, "pid": 76337, "tid": -914061504, "ts": 1716454225827984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827985, "dur": 0,
    "args": {
      "External id": 276095,
      "cbid": 205, "correlation": 276095
    }
  },
  {
    "ph": "f", "id": 276095, "pid": 76337, "tid": -914061504, "ts": 1716454225827985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225880988, "dur": 10,
    "args": {
      "External id": 276099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276099,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276099, "pid": 5, "tid": 7, "ts": 1716454225880988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225827999, "dur": 12,
    "args": {
      "External id": 276099,
      "cbid": 211, "correlation": 276099
    }
  },
  {
    "ph": "s", "id": 276099, "pid": 76337, "tid": -914061504, "ts": 1716454225827999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225881000, "dur": 156,
    "args": {
      "External id": 276101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276101,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276101, "pid": 5, "tid": 7, "ts": 1716454225881000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828013, "dur": 5,
    "args": {
      "External id": 276101,
      "cbid": 211, "correlation": 276101
    }
  },
  {
    "ph": "s", "id": 276101, "pid": 76337, "tid": -914061504, "ts": 1716454225828013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225881159, "dur": 1,
    "args": {
      "External id": 276103,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276103,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 276103, "pid": 5, "tid": 7, "ts": 1716454225881159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828025, "dur": 6,
    "args": {
      "External id": 276103,
      "cbid": 51, "correlation": 276103
    }
  },
  {
    "ph": "s", "id": 276103, "pid": 76337, "tid": -914061504, "ts": 1716454225828025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225881162, "dur": 634,
    "args": {
      "External id": 276104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276104,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276104, "pid": 5, "tid": 7, "ts": 1716454225881162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828032, "dur": 6,
    "args": {
      "External id": 276104,
      "cbid": 211, "correlation": 276104
    }
  },
  {
    "ph": "s", "id": 276104, "pid": 76337, "tid": -914061504, "ts": 1716454225828032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225881798, "dur": 12,
    "args": {
      "External id": 276106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276106,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276106, "pid": 5, "tid": 7, "ts": 1716454225881798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828042, "dur": 5,
    "args": {
      "External id": 276106,
      "cbid": 211, "correlation": 276106
    }
  },
  {
    "ph": "s", "id": 276106, "pid": 76337, "tid": -914061504, "ts": 1716454225828042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225881811, "dur": 14,
    "args": {
      "External id": 276112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276112, "pid": 5, "tid": 7, "ts": 1716454225881811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828071, "dur": 8,
    "args": {
      "External id": 276112,
      "cbid": 211, "correlation": 276112
    }
  },
  {
    "ph": "s", "id": 276112, "pid": 76337, "tid": -914061504, "ts": 1716454225828071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225881826, "dur": 12,
    "args": {
      "External id": 276120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276120,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276120, "pid": 5, "tid": 7, "ts": 1716454225881826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828103, "dur": 8,
    "args": {
      "External id": 276120,
      "cbid": 211, "correlation": 276120
    }
  },
  {
    "ph": "s", "id": 276120, "pid": 76337, "tid": -914061504, "ts": 1716454225828103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225881840, "dur": 10,
    "args": {
      "External id": 276128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276128,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276128, "pid": 5, "tid": 7, "ts": 1716454225881840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828133, "dur": 8,
    "args": {
      "External id": 276128,
      "cbid": 211, "correlation": 276128
    }
  },
  {
    "ph": "s", "id": 276128, "pid": 76337, "tid": -914061504, "ts": 1716454225828133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225881851, "dur": 18,
    "args": {
      "External id": 276148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276148,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 276148, "pid": 5, "tid": 7, "ts": 1716454225881851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828210, "dur": 13,
    "args": {
      "External id": 276148,
      "cbid": 211, "correlation": 276148
    }
  },
  {
    "ph": "s", "id": 276148, "pid": 76337, "tid": -914061504, "ts": 1716454225828210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225881870, "dur": 4,
    "args": {
      "External id": 276160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276160,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 276160, "pid": 5, "tid": 7, "ts": 1716454225881870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828232, "dur": 6,
    "args": {
      "External id": 276160,
      "cbid": 211, "correlation": 276160
    }
  },
  {
    "ph": "s", "id": 276160, "pid": 76337, "tid": -914061504, "ts": 1716454225828232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225881875, "dur": 16,
    "args": {
      "External id": 276163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276163,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276163, "pid": 5, "tid": 7, "ts": 1716454225881875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828250, "dur": 8,
    "args": {
      "External id": 276163,
      "cbid": 211, "correlation": 276163
    }
  },
  {
    "ph": "s", "id": 276163, "pid": 76337, "tid": -914061504, "ts": 1716454225828250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828308, "dur": 0,
    "args": {
      "External id": 276174,
      "cbid": 317, "correlation": 276174
    }
  },
  {
    "ph": "f", "id": 276174, "pid": 76337, "tid": -914061504, "ts": 1716454225828308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828309, "dur": 0,
    "args": {
      "External id": 276175,
      "cbid": 203, "correlation": 276175
    }
  },
  {
    "ph": "f", "id": 276175, "pid": 76337, "tid": -914061504, "ts": 1716454225828309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828309, "dur": 0,
    "args": {
      "External id": 276176,
      "cbid": 205, "correlation": 276176
    }
  },
  {
    "ph": "f", "id": 276176, "pid": 76337, "tid": -914061504, "ts": 1716454225828309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225881892, "dur": 11,
    "args": {
      "External id": 276180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276180,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276180, "pid": 5, "tid": 7, "ts": 1716454225881892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828323, "dur": 12,
    "args": {
      "External id": 276180,
      "cbid": 211, "correlation": 276180
    }
  },
  {
    "ph": "s", "id": 276180, "pid": 76337, "tid": -914061504, "ts": 1716454225828323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225881904, "dur": 4,
    "args": {
      "External id": 276182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276182,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 276182, "pid": 5, "tid": 7, "ts": 1716454225881904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828339, "dur": 5,
    "args": {
      "External id": 276182,
      "cbid": 211, "correlation": 276182
    }
  },
  {
    "ph": "s", "id": 276182, "pid": 76337, "tid": -914061504, "ts": 1716454225828339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828347, "dur": 0,
    "args": {
      "External id": 276183,
      "cbid": 51, "correlation": 276183
    }
  },
  {
    "ph": "s", "id": 276183, "pid": 76337, "tid": -914061504, "ts": 1716454225828347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225881909, "dur": 91,
    "args": {
      "External id": 276184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276184,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 276184, "pid": 5, "tid": 7, "ts": 1716454225881909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828348, "dur": 5,
    "args": {
      "External id": 276184,
      "cbid": 211, "correlation": 276184
    }
  },
  {
    "ph": "s", "id": 276184, "pid": 76337, "tid": -914061504, "ts": 1716454225828348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225882001, "dur": 15,
    "args": {
      "External id": 276189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276189, "pid": 5, "tid": 7, "ts": 1716454225882001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828375, "dur": 9,
    "args": {
      "External id": 276189,
      "cbid": 211, "correlation": 276189
    }
  },
  {
    "ph": "s", "id": 276189, "pid": 76337, "tid": -914061504, "ts": 1716454225828375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225882017, "dur": 80,
    "args": {
      "External id": 276198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276198,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276198, "pid": 5, "tid": 7, "ts": 1716454225882017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828457, "dur": 15,
    "args": {
      "External id": 276198,
      "cbid": 211, "correlation": 276198
    }
  },
  {
    "ph": "s", "id": 276198, "pid": 76337, "tid": -914061504, "ts": 1716454225828457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225882099, "dur": 30,
    "args": {
      "External id": 276220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276220,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276220, "pid": 5, "tid": 7, "ts": 1716454225882099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828517, "dur": 10,
    "args": {
      "External id": 276220,
      "cbid": 211, "correlation": 276220
    }
  },
  {
    "ph": "s", "id": 276220, "pid": 76337, "tid": -914061504, "ts": 1716454225828517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828603, "dur": 1,
    "args": {
      "External id": 276231,
      "cbid": 251, "correlation": 276231
    }
  },
  {
    "ph": "f", "id": 276231, "pid": 76337, "tid": -914061504, "ts": 1716454225828603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225882130, "dur": 140,
    "args": {
      "External id": 276232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276232,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276232, "pid": 5, "tid": 7, "ts": 1716454225882130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828609, "dur": 14,
    "args": {
      "External id": 276232,
      "cbid": 211, "correlation": 276232
    }
  },
  {
    "ph": "s", "id": 276232, "pid": 76337, "tid": -914061504, "ts": 1716454225828609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828679, "dur": 1,
    "args": {
      "External id": 276243,
      "cbid": 251, "correlation": 276243
    }
  },
  {
    "ph": "f", "id": 276243, "pid": 76337, "tid": -914061504, "ts": 1716454225828679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225882271, "dur": 154,
    "args": {
      "External id": 276244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276244,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276244, "pid": 5, "tid": 7, "ts": 1716454225882271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828683, "dur": 11,
    "args": {
      "External id": 276244,
      "cbid": 211, "correlation": 276244
    }
  },
  {
    "ph": "s", "id": 276244, "pid": 76337, "tid": -914061504, "ts": 1716454225828683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828746, "dur": 1,
    "args": {
      "External id": 276255,
      "cbid": 251, "correlation": 276255
    }
  },
  {
    "ph": "f", "id": 276255, "pid": 76337, "tid": -914061504, "ts": 1716454225828746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225882426, "dur": 155,
    "args": {
      "External id": 276256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276256,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276256, "pid": 5, "tid": 7, "ts": 1716454225882426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828750, "dur": 11,
    "args": {
      "External id": 276256,
      "cbid": 211, "correlation": 276256
    }
  },
  {
    "ph": "s", "id": 276256, "pid": 76337, "tid": -914061504, "ts": 1716454225828750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225882582, "dur": 326,
    "args": {
      "External id": 276281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276281,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276281, "pid": 5, "tid": 7, "ts": 1716454225882582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828834, "dur": 13,
    "args": {
      "External id": 276281,
      "cbid": 211, "correlation": 276281
    }
  },
  {
    "ph": "s", "id": 276281, "pid": 76337, "tid": -914061504, "ts": 1716454225828834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828935, "dur": 1,
    "args": {
      "External id": 276299,
      "cbid": 251, "correlation": 276299
    }
  },
  {
    "ph": "f", "id": 276299, "pid": 76337, "tid": -914061504, "ts": 1716454225828935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225882910, "dur": 163,
    "args": {
      "External id": 276301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276301,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276301, "pid": 5, "tid": 7, "ts": 1716454225882910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225828941, "dur": 13,
    "args": {
      "External id": 276301,
      "cbid": 211, "correlation": 276301
    }
  },
  {
    "ph": "s", "id": 276301, "pid": 76337, "tid": -914061504, "ts": 1716454225828941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225883075, "dur": 19,
    "args": {
      "External id": 276309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276309,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276309, "pid": 5, "tid": 7, "ts": 1716454225883075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829019, "dur": 13,
    "args": {
      "External id": 276309,
      "cbid": 211, "correlation": 276309
    }
  },
  {
    "ph": "s", "id": 276309, "pid": 76337, "tid": -914061504, "ts": 1716454225829019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225883095, "dur": 27,
    "args": {
      "External id": 276317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276317,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276317, "pid": 5, "tid": 7, "ts": 1716454225883095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829058, "dur": 9,
    "args": {
      "External id": 276317,
      "cbid": 211, "correlation": 276317
    }
  },
  {
    "ph": "s", "id": 276317, "pid": 76337, "tid": -914061504, "ts": 1716454225829058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225883124, "dur": 18,
    "args": {
      "External id": 276328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276328, "pid": 5, "tid": 7, "ts": 1716454225883124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829129, "dur": 12,
    "args": {
      "External id": 276328,
      "cbid": 211, "correlation": 276328
    }
  },
  {
    "ph": "s", "id": 276328, "pid": 76337, "tid": -914061504, "ts": 1716454225829129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225883143, "dur": 15,
    "args": {
      "External id": 276350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276350,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276350, "pid": 5, "tid": 7, "ts": 1716454225883143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829160, "dur": 8,
    "args": {
      "External id": 276350,
      "cbid": 211, "correlation": 276350
    }
  },
  {
    "ph": "s", "id": 276350, "pid": 76337, "tid": -914061504, "ts": 1716454225829160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829245, "dur": 1,
    "args": {
      "External id": 276361,
      "cbid": 251, "correlation": 276361
    }
  },
  {
    "ph": "f", "id": 276361, "pid": 76337, "tid": -914061504, "ts": 1716454225829245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225883159, "dur": 85,
    "args": {
      "External id": 276362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276362,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 276362, "pid": 5, "tid": 7, "ts": 1716454225883159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829250, "dur": 14,
    "args": {
      "External id": 276362,
      "cbid": 211, "correlation": 276362
    }
  },
  {
    "ph": "s", "id": 276362, "pid": 76337, "tid": -914061504, "ts": 1716454225829250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829319, "dur": 1,
    "args": {
      "External id": 276373,
      "cbid": 251, "correlation": 276373
    }
  },
  {
    "ph": "f", "id": 276373, "pid": 76337, "tid": -914061504, "ts": 1716454225829319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829323, "dur": 0,
    "args": {
      "External id": 276374,
      "cbid": 251, "correlation": 276374
    }
  },
  {
    "ph": "f", "id": 276374, "pid": 76337, "tid": -914061504, "ts": 1716454225829323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225883245, "dur": 12,
    "args": {
      "External id": 276375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276375,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276375, "pid": 5, "tid": 7, "ts": 1716454225883245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829325, "dur": 12,
    "args": {
      "External id": 276375,
      "cbid": 211, "correlation": 276375
    }
  },
  {
    "ph": "s", "id": 276375, "pid": 76337, "tid": -914061504, "ts": 1716454225829325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225883258, "dur": 5,
    "args": {
      "External id": 276377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276377,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276377, "pid": 5, "tid": 7, "ts": 1716454225883258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829338, "dur": 6,
    "args": {
      "External id": 276377,
      "cbid": 211, "correlation": 276377
    }
  },
  {
    "ph": "s", "id": 276377, "pid": 76337, "tid": -914061504, "ts": 1716454225829338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829401, "dur": 1,
    "args": {
      "External id": 276388,
      "cbid": 251, "correlation": 276388
    }
  },
  {
    "ph": "f", "id": 276388, "pid": 76337, "tid": -914061504, "ts": 1716454225829401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829404, "dur": 0,
    "args": {
      "External id": 276389,
      "cbid": 251, "correlation": 276389
    }
  },
  {
    "ph": "f", "id": 276389, "pid": 76337, "tid": -914061504, "ts": 1716454225829404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225883265, "dur": 8,
    "args": {
      "External id": 276390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276390,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276390, "pid": 5, "tid": 7, "ts": 1716454225883265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829406, "dur": 13,
    "args": {
      "External id": 276390,
      "cbid": 211, "correlation": 276390
    }
  },
  {
    "ph": "s", "id": 276390, "pid": 76337, "tid": -914061504, "ts": 1716454225829406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225883274, "dur": 3,
    "args": {
      "External id": 276392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276392,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276392, "pid": 5, "tid": 7, "ts": 1716454225883274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829420, "dur": 6,
    "args": {
      "External id": 276392,
      "cbid": 211, "correlation": 276392
    }
  },
  {
    "ph": "s", "id": 276392, "pid": 76337, "tid": -914061504, "ts": 1716454225829420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225883279, "dur": 53,
    "args": {
      "External id": 276417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276417,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276417, "pid": 5, "tid": 7, "ts": 1716454225883279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829498, "dur": 12,
    "args": {
      "External id": 276417,
      "cbid": 211, "correlation": 276417
    }
  },
  {
    "ph": "s", "id": 276417, "pid": 76337, "tid": -914061504, "ts": 1716454225829498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829600, "dur": 1,
    "args": {
      "External id": 276435,
      "cbid": 251, "correlation": 276435
    }
  },
  {
    "ph": "f", "id": 276435, "pid": 76337, "tid": -914061504, "ts": 1716454225829600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225883333, "dur": 87,
    "args": {
      "External id": 276437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276437,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 276437, "pid": 5, "tid": 7, "ts": 1716454225883333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829606, "dur": 15,
    "args": {
      "External id": 276437,
      "cbid": 211, "correlation": 276437
    }
  },
  {
    "ph": "s", "id": 276437, "pid": 76337, "tid": -914061504, "ts": 1716454225829606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225883421, "dur": 9,
    "args": {
      "External id": 276445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276445,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276445, "pid": 5, "tid": 7, "ts": 1716454225883421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829676, "dur": 12,
    "args": {
      "External id": 276445,
      "cbid": 211, "correlation": 276445
    }
  },
  {
    "ph": "s", "id": 276445, "pid": 76337, "tid": -914061504, "ts": 1716454225829676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225883432, "dur": 20,
    "args": {
      "External id": 276453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276453, "pid": 5, "tid": 7, "ts": 1716454225883432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829718, "dur": 9,
    "args": {
      "External id": 276453,
      "cbid": 211, "correlation": 276453
    }
  },
  {
    "ph": "s", "id": 276453, "pid": 76337, "tid": -914061504, "ts": 1716454225829718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225883453, "dur": 17,
    "args": {
      "External id": 276475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276475,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276475, "pid": 5, "tid": 7, "ts": 1716454225883453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829768, "dur": 10,
    "args": {
      "External id": 276475,
      "cbid": 211, "correlation": 276475
    }
  },
  {
    "ph": "s", "id": 276475, "pid": 76337, "tid": -914061504, "ts": 1716454225829768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829855, "dur": 1,
    "args": {
      "External id": 276491,
      "cbid": 251, "correlation": 276491
    }
  },
  {
    "ph": "f", "id": 276491, "pid": 76337, "tid": -914061504, "ts": 1716454225829855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829860, "dur": 0,
    "args": {
      "External id": 276493,
      "cbid": 251, "correlation": 276493
    }
  },
  {
    "ph": "f", "id": 276493, "pid": 76337, "tid": -914061504, "ts": 1716454225829860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225883471, "dur": 485,
    "args": {
      "External id": 276494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276494,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276494, "pid": 5, "tid": 7, "ts": 1716454225883471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829862, "dur": 13,
    "args": {
      "External id": 276494,
      "cbid": 211, "correlation": 276494
    }
  },
  {
    "ph": "s", "id": 276494, "pid": 76337, "tid": -914061504, "ts": 1716454225829862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225883957, "dur": 63,
    "args": {
      "External id": 276502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276502, "pid": 5, "tid": 7, "ts": 1716454225883957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829927, "dur": 12,
    "args": {
      "External id": 276502,
      "cbid": 211, "correlation": 276502
    }
  },
  {
    "ph": "s", "id": 276502, "pid": 76337, "tid": -914061504, "ts": 1716454225829927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225884022, "dur": 66,
    "args": {
      "External id": 276510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276510, "pid": 5, "tid": 7, "ts": 1716454225884022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225829956, "dur": 8,
    "args": {
      "External id": 276510,
      "cbid": 211, "correlation": 276510
    }
  },
  {
    "ph": "s", "id": 276510, "pid": 76337, "tid": -914061504, "ts": 1716454225829956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830042, "dur": 1,
    "args": {
      "External id": 276526,
      "cbid": 251, "correlation": 276526
    }
  },
  {
    "ph": "f", "id": 276526, "pid": 76337, "tid": -914061504, "ts": 1716454225830042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225884089, "dur": 1,
    "args": {
      "External id": 276528,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276528,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 276528, "pid": 5, "tid": 7, "ts": 1716454225884089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830047, "dur": 9,
    "args": {
      "External id": 276528,
      "cbid": 51, "correlation": 276528
    }
  },
  {
    "ph": "s", "id": 276528, "pid": 76337, "tid": -914061504, "ts": 1716454225830047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225884093, "dur": 261,
    "args": {
      "External id": 276529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276529,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 276529, "pid": 5, "tid": 7, "ts": 1716454225884093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830058, "dur": 12,
    "args": {
      "External id": 276529,
      "cbid": 211, "correlation": 276529
    }
  },
  {
    "ph": "s", "id": 276529, "pid": 76337, "tid": -914061504, "ts": 1716454225830058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225884356, "dur": 13,
    "args": {
      "External id": 276537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276537,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276537, "pid": 5, "tid": 7, "ts": 1716454225884356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830100, "dur": 10,
    "args": {
      "External id": 276537,
      "cbid": 211, "correlation": 276537
    }
  },
  {
    "ph": "s", "id": 276537, "pid": 76337, "tid": -914061504, "ts": 1716454225830100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225884370, "dur": 35,
    "args": {
      "External id": 276548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276548, "pid": 5, "tid": 7, "ts": 1716454225884370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830169, "dur": 12,
    "args": {
      "External id": 276548,
      "cbid": 211, "correlation": 276548
    }
  },
  {
    "ph": "s", "id": 276548, "pid": 76337, "tid": -914061504, "ts": 1716454225830169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830232, "dur": 0,
    "args": {
      "External id": 276560,
      "cbid": 317, "correlation": 276560
    }
  },
  {
    "ph": "f", "id": 276560, "pid": 76337, "tid": -914061504, "ts": 1716454225830232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830233, "dur": 0,
    "args": {
      "External id": 276561,
      "cbid": 203, "correlation": 276561
    }
  },
  {
    "ph": "f", "id": 276561, "pid": 76337, "tid": -914061504, "ts": 1716454225830233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830234, "dur": 0,
    "args": {
      "External id": 276562,
      "cbid": 205, "correlation": 276562
    }
  },
  {
    "ph": "f", "id": 276562, "pid": 76337, "tid": -914061504, "ts": 1716454225830234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884407, "dur": 12,
    "args": {
      "External id": 276566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276566,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276566, "pid": 5, "tid": 7, "ts": 1716454225884407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830249, "dur": 12,
    "args": {
      "External id": 276566,
      "cbid": 211, "correlation": 276566
    }
  },
  {
    "ph": "s", "id": 276566, "pid": 76337, "tid": -914061504, "ts": 1716454225830249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225884420, "dur": 3,
    "args": {
      "External id": 276568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 276568, "pid": 5, "tid": 7, "ts": 1716454225884420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830266, "dur": 6,
    "args": {
      "External id": 276568,
      "cbid": 211, "correlation": 276568
    }
  },
  {
    "ph": "s", "id": 276568, "pid": 76337, "tid": -914061504, "ts": 1716454225830266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830275, "dur": 0,
    "args": {
      "External id": 276569,
      "cbid": 51, "correlation": 276569
    }
  },
  {
    "ph": "s", "id": 276569, "pid": 76337, "tid": -914061504, "ts": 1716454225830275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225884425, "dur": 93,
    "args": {
      "External id": 276570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276570,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 276570, "pid": 5, "tid": 7, "ts": 1716454225884425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830276, "dur": 5,
    "args": {
      "External id": 276570,
      "cbid": 211, "correlation": 276570
    }
  },
  {
    "ph": "s", "id": 276570, "pid": 76337, "tid": -914061504, "ts": 1716454225830276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225884519, "dur": 15,
    "args": {
      "External id": 276575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276575, "pid": 5, "tid": 7, "ts": 1716454225884519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830303, "dur": 9,
    "args": {
      "External id": 276575,
      "cbid": 211, "correlation": 276575
    }
  },
  {
    "ph": "s", "id": 276575, "pid": 76337, "tid": -914061504, "ts": 1716454225830303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225884536, "dur": 11,
    "args": {
      "External id": 276583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276583,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276583, "pid": 5, "tid": 7, "ts": 1716454225884536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830335, "dur": 9,
    "args": {
      "External id": 276583,
      "cbid": 211, "correlation": 276583
    }
  },
  {
    "ph": "s", "id": 276583, "pid": 76337, "tid": -914061504, "ts": 1716454225830335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830404, "dur": 0,
    "args": {
      "External id": 276593,
      "cbid": 317, "correlation": 276593
    }
  },
  {
    "ph": "f", "id": 276593, "pid": 76337, "tid": -914061504, "ts": 1716454225830404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830405, "dur": 0,
    "args": {
      "External id": 276594,
      "cbid": 203, "correlation": 276594
    }
  },
  {
    "ph": "f", "id": 276594, "pid": 76337, "tid": -914061504, "ts": 1716454225830405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830406, "dur": 0,
    "args": {
      "External id": 276595,
      "cbid": 205, "correlation": 276595
    }
  },
  {
    "ph": "f", "id": 276595, "pid": 76337, "tid": -914061504, "ts": 1716454225830406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884549, "dur": 11,
    "args": {
      "External id": 276599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276599,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276599, "pid": 5, "tid": 7, "ts": 1716454225884549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830421, "dur": 12,
    "args": {
      "External id": 276599,
      "cbid": 211, "correlation": 276599
    }
  },
  {
    "ph": "s", "id": 276599, "pid": 76337, "tid": -914061504, "ts": 1716454225830421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884561, "dur": 155,
    "args": {
      "External id": 276601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276601,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276601, "pid": 5, "tid": 7, "ts": 1716454225884561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830435, "dur": 5,
    "args": {
      "External id": 276601,
      "cbid": 211, "correlation": 276601
    }
  },
  {
    "ph": "s", "id": 276601, "pid": 76337, "tid": -914061504, "ts": 1716454225830435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225884718, "dur": 1,
    "args": {
      "External id": 276603,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276603,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 276603, "pid": 5, "tid": 7, "ts": 1716454225884718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830447, "dur": 6,
    "args": {
      "External id": 276603,
      "cbid": 51, "correlation": 276603
    }
  },
  {
    "ph": "s", "id": 276603, "pid": 76337, "tid": -914061504, "ts": 1716454225830447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225884722, "dur": 196,
    "args": {
      "External id": 276604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276604,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 276604, "pid": 5, "tid": 7, "ts": 1716454225884722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830455, "dur": 9,
    "args": {
      "External id": 276604,
      "cbid": 211, "correlation": 276604
    }
  },
  {
    "ph": "s", "id": 276604, "pid": 76337, "tid": -914061504, "ts": 1716454225830455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884919, "dur": 6,
    "args": {
      "External id": 276606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276606,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276606, "pid": 5, "tid": 7, "ts": 1716454225884919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830468, "dur": 5,
    "args": {
      "External id": 276606,
      "cbid": 211, "correlation": 276606
    }
  },
  {
    "ph": "s", "id": 276606, "pid": 76337, "tid": -914061504, "ts": 1716454225830468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225884926, "dur": 6,
    "args": {
      "External id": 276612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276612, "pid": 5, "tid": 7, "ts": 1716454225884926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830495, "dur": 8,
    "args": {
      "External id": 276612,
      "cbid": 211, "correlation": 276612
    }
  },
  {
    "ph": "s", "id": 276612, "pid": 76337, "tid": -914061504, "ts": 1716454225830495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884934, "dur": 11,
    "args": {
      "External id": 276632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276632,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 276632, "pid": 5, "tid": 7, "ts": 1716454225884934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830588, "dur": 12,
    "args": {
      "External id": 276632,
      "cbid": 211, "correlation": 276632
    }
  },
  {
    "ph": "s", "id": 276632, "pid": 76337, "tid": -914061504, "ts": 1716454225830588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225884946, "dur": 4,
    "args": {
      "External id": 276644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276644,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 276644, "pid": 5, "tid": 7, "ts": 1716454225884946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830611, "dur": 6,
    "args": {
      "External id": 276644,
      "cbid": 211, "correlation": 276644
    }
  },
  {
    "ph": "s", "id": 276644, "pid": 76337, "tid": -914061504, "ts": 1716454225830611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225884951, "dur": 8,
    "args": {
      "External id": 276647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276647, "pid": 5, "tid": 7, "ts": 1716454225884951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830630, "dur": 7,
    "args": {
      "External id": 276647,
      "cbid": 211, "correlation": 276647
    }
  },
  {
    "ph": "s", "id": 276647, "pid": 76337, "tid": -914061504, "ts": 1716454225830630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225884960, "dur": 5,
    "args": {
      "External id": 276656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276656,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276656, "pid": 5, "tid": 7, "ts": 1716454225884960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830670, "dur": 9,
    "args": {
      "External id": 276656,
      "cbid": 211, "correlation": 276656
    }
  },
  {
    "ph": "s", "id": 276656, "pid": 76337, "tid": -914061504, "ts": 1716454225830670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830723, "dur": 0,
    "args": {
      "External id": 276666,
      "cbid": 317, "correlation": 276666
    }
  },
  {
    "ph": "f", "id": 276666, "pid": 76337, "tid": -914061504, "ts": 1716454225830723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830724, "dur": 0,
    "args": {
      "External id": 276667,
      "cbid": 203, "correlation": 276667
    }
  },
  {
    "ph": "f", "id": 276667, "pid": 76337, "tid": -914061504, "ts": 1716454225830724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830724, "dur": 0,
    "args": {
      "External id": 276668,
      "cbid": 205, "correlation": 276668
    }
  },
  {
    "ph": "f", "id": 276668, "pid": 76337, "tid": -914061504, "ts": 1716454225830724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884967, "dur": 5,
    "args": {
      "External id": 276672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276672, "pid": 5, "tid": 7, "ts": 1716454225884967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830739, "dur": 11,
    "args": {
      "External id": 276672,
      "cbid": 211, "correlation": 276672
    }
  },
  {
    "ph": "s", "id": 276672, "pid": 76337, "tid": -914061504, "ts": 1716454225830739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225884973, "dur": 155,
    "args": {
      "External id": 276674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276674, "pid": 5, "tid": 7, "ts": 1716454225884973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830753, "dur": 5,
    "args": {
      "External id": 276674,
      "cbid": 211, "correlation": 276674
    }
  },
  {
    "ph": "s", "id": 276674, "pid": 76337, "tid": -914061504, "ts": 1716454225830753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225885130, "dur": 1,
    "args": {
      "External id": 276676,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276676,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 276676, "pid": 5, "tid": 7, "ts": 1716454225885130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830764, "dur": 6,
    "args": {
      "External id": 276676,
      "cbid": 51, "correlation": 276676
    }
  },
  {
    "ph": "s", "id": 276676, "pid": 76337, "tid": -914061504, "ts": 1716454225830764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225885134, "dur": 261,
    "args": {
      "External id": 276677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276677,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276677, "pid": 5, "tid": 7, "ts": 1716454225885134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830771, "dur": 6,
    "args": {
      "External id": 276677,
      "cbid": 211, "correlation": 276677
    }
  },
  {
    "ph": "s", "id": 276677, "pid": 76337, "tid": -914061504, "ts": 1716454225830771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885396, "dur": 5,
    "args": {
      "External id": 276679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276679,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276679, "pid": 5, "tid": 7, "ts": 1716454225885396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830783, "dur": 5,
    "args": {
      "External id": 276679,
      "cbid": 211, "correlation": 276679
    }
  },
  {
    "ph": "s", "id": 276679, "pid": 76337, "tid": -914061504, "ts": 1716454225830783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225885403, "dur": 6,
    "args": {
      "External id": 276685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276685,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276685, "pid": 5, "tid": 7, "ts": 1716454225885403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830812, "dur": 8,
    "args": {
      "External id": 276685,
      "cbid": 211, "correlation": 276685
    }
  },
  {
    "ph": "s", "id": 276685, "pid": 76337, "tid": -914061504, "ts": 1716454225830812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225885410, "dur": 4,
    "args": {
      "External id": 276693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276693,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 276693, "pid": 5, "tid": 7, "ts": 1716454225885410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830854, "dur": 10,
    "args": {
      "External id": 276693,
      "cbid": 211, "correlation": 276693
    }
  },
  {
    "ph": "s", "id": 276693, "pid": 76337, "tid": -914061504, "ts": 1716454225830854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830922, "dur": 1,
    "args": {
      "External id": 276709,
      "cbid": 251, "correlation": 276709
    }
  },
  {
    "ph": "f", "id": 276709, "pid": 76337, "tid": -914061504, "ts": 1716454225830922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830927, "dur": 0,
    "args": {
      "External id": 276711,
      "cbid": 251, "correlation": 276711
    }
  },
  {
    "ph": "f", "id": 276711, "pid": 76337, "tid": -914061504, "ts": 1716454225830927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225885415, "dur": 13,
    "args": {
      "External id": 276712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276712,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276712, "pid": 5, "tid": 7, "ts": 1716454225885415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830929, "dur": 11,
    "args": {
      "External id": 276712,
      "cbid": 211, "correlation": 276712
    }
  },
  {
    "ph": "s", "id": 276712, "pid": 76337, "tid": -914061504, "ts": 1716454225830929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225885429, "dur": 5,
    "args": {
      "External id": 276714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276714,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276714, "pid": 5, "tid": 7, "ts": 1716454225885429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225830942, "dur": 5,
    "args": {
      "External id": 276714,
      "cbid": 211, "correlation": 276714
    }
  },
  {
    "ph": "s", "id": 276714, "pid": 76337, "tid": -914061504, "ts": 1716454225830942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225885436, "dur": 5,
    "args": {
      "External id": 276724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276724,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276724, "pid": 5, "tid": 7, "ts": 1716454225885436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831010, "dur": 13,
    "args": {
      "External id": 276724,
      "cbid": 211, "correlation": 276724
    }
  },
  {
    "ph": "s", "id": 276724, "pid": 76337, "tid": -914061504, "ts": 1716454225831010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885442, "dur": 9,
    "args": {
      "External id": 276744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276744,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 276744, "pid": 5, "tid": 7, "ts": 1716454225885442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831076, "dur": 11,
    "args": {
      "External id": 276744,
      "cbid": 211, "correlation": 276744
    }
  },
  {
    "ph": "s", "id": 276744, "pid": 76337, "tid": -914061504, "ts": 1716454225831076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225885453, "dur": 4,
    "args": {
      "External id": 276756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276756,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 276756, "pid": 5, "tid": 7, "ts": 1716454225885453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831097, "dur": 6,
    "args": {
      "External id": 276756,
      "cbid": 211, "correlation": 276756
    }
  },
  {
    "ph": "s", "id": 276756, "pid": 76337, "tid": -914061504, "ts": 1716454225831097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225885458, "dur": 7,
    "args": {
      "External id": 276759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276759, "pid": 5, "tid": 7, "ts": 1716454225885458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831117, "dur": 6,
    "args": {
      "External id": 276759,
      "cbid": 211, "correlation": 276759
    }
  },
  {
    "ph": "s", "id": 276759, "pid": 76337, "tid": -914061504, "ts": 1716454225831117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225885466, "dur": 4,
    "args": {
      "External id": 276768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276768,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276768, "pid": 5, "tid": 7, "ts": 1716454225885466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831157, "dur": 10,
    "args": {
      "External id": 276768,
      "cbid": 211, "correlation": 276768
    }
  },
  {
    "ph": "s", "id": 276768, "pid": 76337, "tid": -914061504, "ts": 1716454225831157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831221, "dur": 0,
    "args": {
      "External id": 276778,
      "cbid": 317, "correlation": 276778
    }
  },
  {
    "ph": "f", "id": 276778, "pid": 76337, "tid": -914061504, "ts": 1716454225831221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831221, "dur": 0,
    "args": {
      "External id": 276779,
      "cbid": 203, "correlation": 276779
    }
  },
  {
    "ph": "f", "id": 276779, "pid": 76337, "tid": -914061504, "ts": 1716454225831221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831222, "dur": 0,
    "args": {
      "External id": 276780,
      "cbid": 205, "correlation": 276780
    }
  },
  {
    "ph": "f", "id": 276780, "pid": 76337, "tid": -914061504, "ts": 1716454225831222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885472, "dur": 5,
    "args": {
      "External id": 276784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276784,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276784, "pid": 5, "tid": 7, "ts": 1716454225885472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831236, "dur": 12,
    "args": {
      "External id": 276784,
      "cbid": 211, "correlation": 276784
    }
  },
  {
    "ph": "s", "id": 276784, "pid": 76337, "tid": -914061504, "ts": 1716454225831236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885478, "dur": 154,
    "args": {
      "External id": 276786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276786,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276786, "pid": 5, "tid": 7, "ts": 1716454225885478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831251, "dur": 6,
    "args": {
      "External id": 276786,
      "cbid": 211, "correlation": 276786
    }
  },
  {
    "ph": "s", "id": 276786, "pid": 76337, "tid": -914061504, "ts": 1716454225831251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225885634, "dur": 1,
    "args": {
      "External id": 276788,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276788,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 276788, "pid": 5, "tid": 7, "ts": 1716454225885634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831262, "dur": 6,
    "args": {
      "External id": 276788,
      "cbid": 51, "correlation": 276788
    }
  },
  {
    "ph": "s", "id": 276788, "pid": 76337, "tid": -914061504, "ts": 1716454225831262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225885638, "dur": 250,
    "args": {
      "External id": 276789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276789,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276789, "pid": 5, "tid": 7, "ts": 1716454225885638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831269, "dur": 6,
    "args": {
      "External id": 276789,
      "cbid": 211, "correlation": 276789
    }
  },
  {
    "ph": "s", "id": 276789, "pid": 76337, "tid": -914061504, "ts": 1716454225831269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885889, "dur": 6,
    "args": {
      "External id": 276791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276791,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276791, "pid": 5, "tid": 7, "ts": 1716454225885889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831279, "dur": 5,
    "args": {
      "External id": 276791,
      "cbid": 211, "correlation": 276791
    }
  },
  {
    "ph": "s", "id": 276791, "pid": 76337, "tid": -914061504, "ts": 1716454225831279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225885896, "dur": 6,
    "args": {
      "External id": 276797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276797, "pid": 5, "tid": 7, "ts": 1716454225885896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831308, "dur": 8,
    "args": {
      "External id": 276797,
      "cbid": 211, "correlation": 276797
    }
  },
  {
    "ph": "s", "id": 276797, "pid": 76337, "tid": -914061504, "ts": 1716454225831308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225885903, "dur": 5,
    "args": {
      "External id": 276805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276805,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276805, "pid": 5, "tid": 7, "ts": 1716454225885903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831340, "dur": 8,
    "args": {
      "External id": 276805,
      "cbid": 211, "correlation": 276805
    }
  },
  {
    "ph": "s", "id": 276805, "pid": 76337, "tid": -914061504, "ts": 1716454225831340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225885910, "dur": 4,
    "args": {
      "External id": 276813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276813,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276813, "pid": 5, "tid": 7, "ts": 1716454225885910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831369, "dur": 8,
    "args": {
      "External id": 276813,
      "cbid": 211, "correlation": 276813
    }
  },
  {
    "ph": "s", "id": 276813, "pid": 76337, "tid": -914061504, "ts": 1716454225831369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885915, "dur": 9,
    "args": {
      "External id": 276833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276833,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 276833, "pid": 5, "tid": 7, "ts": 1716454225885915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831442, "dur": 13,
    "args": {
      "External id": 276833,
      "cbid": 211, "correlation": 276833
    }
  },
  {
    "ph": "s", "id": 276833, "pid": 76337, "tid": -914061504, "ts": 1716454225831442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225885925, "dur": 4,
    "args": {
      "External id": 276845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276845,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 276845, "pid": 5, "tid": 7, "ts": 1716454225885925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831465, "dur": 6,
    "args": {
      "External id": 276845,
      "cbid": 211, "correlation": 276845
    }
  },
  {
    "ph": "s", "id": 276845, "pid": 76337, "tid": -914061504, "ts": 1716454225831465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225885930, "dur": 6,
    "args": {
      "External id": 276848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276848, "pid": 5, "tid": 7, "ts": 1716454225885930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831482, "dur": 6,
    "args": {
      "External id": 276848,
      "cbid": 211, "correlation": 276848
    }
  },
  {
    "ph": "s", "id": 276848, "pid": 76337, "tid": -914061504, "ts": 1716454225831482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225885938, "dur": 4,
    "args": {
      "External id": 276857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276857,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276857, "pid": 5, "tid": 7, "ts": 1716454225885938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831520, "dur": 9,
    "args": {
      "External id": 276857,
      "cbid": 211, "correlation": 276857
    }
  },
  {
    "ph": "s", "id": 276857, "pid": 76337, "tid": -914061504, "ts": 1716454225831520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831570, "dur": 0,
    "args": {
      "External id": 276867,
      "cbid": 317, "correlation": 276867
    }
  },
  {
    "ph": "f", "id": 276867, "pid": 76337, "tid": -914061504, "ts": 1716454225831570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831571, "dur": 0,
    "args": {
      "External id": 276868,
      "cbid": 203, "correlation": 276868
    }
  },
  {
    "ph": "f", "id": 276868, "pid": 76337, "tid": -914061504, "ts": 1716454225831571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831571, "dur": 0,
    "args": {
      "External id": 276869,
      "cbid": 205, "correlation": 276869
    }
  },
  {
    "ph": "f", "id": 276869, "pid": 76337, "tid": -914061504, "ts": 1716454225831571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885943, "dur": 5,
    "args": {
      "External id": 276873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276873,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276873, "pid": 5, "tid": 7, "ts": 1716454225885943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831584, "dur": 11,
    "args": {
      "External id": 276873,
      "cbid": 211, "correlation": 276873
    }
  },
  {
    "ph": "s", "id": 276873, "pid": 76337, "tid": -914061504, "ts": 1716454225831584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225885949, "dur": 155,
    "args": {
      "External id": 276875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276875, "pid": 5, "tid": 7, "ts": 1716454225885949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831597, "dur": 6,
    "args": {
      "External id": 276875,
      "cbid": 211, "correlation": 276875
    }
  },
  {
    "ph": "s", "id": 276875, "pid": 76337, "tid": -914061504, "ts": 1716454225831597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225886106, "dur": 1,
    "args": {
      "External id": 276877,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276877,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1596806387225549
    }
  },
  {
    "ph": "f", "id": 276877, "pid": 5, "tid": 7, "ts": 1716454225886106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831609, "dur": 6,
    "args": {
      "External id": 276877,
      "cbid": 51, "correlation": 276877
    }
  },
  {
    "ph": "s", "id": 276877, "pid": 76337, "tid": -914061504, "ts": 1716454225831609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225886110, "dur": 250,
    "args": {
      "External id": 276878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276878,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276878, "pid": 5, "tid": 7, "ts": 1716454225886110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831616, "dur": 6,
    "args": {
      "External id": 276878,
      "cbid": 211, "correlation": 276878
    }
  },
  {
    "ph": "s", "id": 276878, "pid": 76337, "tid": -914061504, "ts": 1716454225831616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886361, "dur": 5,
    "args": {
      "External id": 276880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276880, "pid": 5, "tid": 7, "ts": 1716454225886361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831625, "dur": 5,
    "args": {
      "External id": 276880,
      "cbid": 211, "correlation": 276880
    }
  },
  {
    "ph": "s", "id": 276880, "pid": 76337, "tid": -914061504, "ts": 1716454225831625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225886368, "dur": 6,
    "args": {
      "External id": 276886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276886, "pid": 5, "tid": 7, "ts": 1716454225886368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831653, "dur": 9,
    "args": {
      "External id": 276886,
      "cbid": 211, "correlation": 276886
    }
  },
  {
    "ph": "s", "id": 276886, "pid": 76337, "tid": -914061504, "ts": 1716454225831653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225886375, "dur": 3,
    "args": {
      "External id": 276894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276894,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 276894, "pid": 5, "tid": 7, "ts": 1716454225886375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831696, "dur": 10,
    "args": {
      "External id": 276894,
      "cbid": 211, "correlation": 276894
    }
  },
  {
    "ph": "s", "id": 276894, "pid": 76337, "tid": -914061504, "ts": 1716454225831696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831758, "dur": 1,
    "args": {
      "External id": 276910,
      "cbid": 251, "correlation": 276910
    }
  },
  {
    "ph": "f", "id": 276910, "pid": 76337, "tid": -914061504, "ts": 1716454225831758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831763, "dur": 0,
    "args": {
      "External id": 276912,
      "cbid": 251, "correlation": 276912
    }
  },
  {
    "ph": "f", "id": 276912, "pid": 76337, "tid": -914061504, "ts": 1716454225831763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225886380, "dur": 10,
    "args": {
      "External id": 276913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276913,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276913, "pid": 5, "tid": 7, "ts": 1716454225886380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831765, "dur": 11,
    "args": {
      "External id": 276913,
      "cbid": 211, "correlation": 276913
    }
  },
  {
    "ph": "s", "id": 276913, "pid": 76337, "tid": -914061504, "ts": 1716454225831765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225886390, "dur": 3,
    "args": {
      "External id": 276915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276915,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276915, "pid": 5, "tid": 7, "ts": 1716454225886390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831778, "dur": 5,
    "args": {
      "External id": 276915,
      "cbid": 211, "correlation": 276915
    }
  },
  {
    "ph": "s", "id": 276915, "pid": 76337, "tid": -914061504, "ts": 1716454225831778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225886395, "dur": 5,
    "args": {
      "External id": 276925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276925, "pid": 5, "tid": 7, "ts": 1716454225886395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831834, "dur": 12,
    "args": {
      "External id": 276925,
      "cbid": 211, "correlation": 276925
    }
  },
  {
    "ph": "s", "id": 276925, "pid": 76337, "tid": -914061504, "ts": 1716454225831834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886402, "dur": 9,
    "args": {
      "External id": 276945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276945,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 276945, "pid": 5, "tid": 7, "ts": 1716454225886402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831898, "dur": 11,
    "args": {
      "External id": 276945,
      "cbid": 211, "correlation": 276945
    }
  },
  {
    "ph": "s", "id": 276945, "pid": 76337, "tid": -914061504, "ts": 1716454225831898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225886412, "dur": 4,
    "args": {
      "External id": 276957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276957,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 276957, "pid": 5, "tid": 7, "ts": 1716454225886412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831919, "dur": 6,
    "args": {
      "External id": 276957,
      "cbid": 211, "correlation": 276957
    }
  },
  {
    "ph": "s", "id": 276957, "pid": 76337, "tid": -914061504, "ts": 1716454225831919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225886417, "dur": 7,
    "args": {
      "External id": 276960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276960, "pid": 5, "tid": 7, "ts": 1716454225886417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831938, "dur": 6,
    "args": {
      "External id": 276960,
      "cbid": 211, "correlation": 276960
    }
  },
  {
    "ph": "s", "id": 276960, "pid": 76337, "tid": -914061504, "ts": 1716454225831938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225886425, "dur": 4,
    "args": {
      "External id": 276969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276969,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276969, "pid": 5, "tid": 7, "ts": 1716454225886425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225831988, "dur": 11,
    "args": {
      "External id": 276969,
      "cbid": 211, "correlation": 276969
    }
  },
  {
    "ph": "s", "id": 276969, "pid": 76337, "tid": -914061504, "ts": 1716454225831988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832053, "dur": 0,
    "args": {
      "External id": 276979,
      "cbid": 317, "correlation": 276979
    }
  },
  {
    "ph": "f", "id": 276979, "pid": 76337, "tid": -914061504, "ts": 1716454225832053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832054, "dur": 0,
    "args": {
      "External id": 276980,
      "cbid": 203, "correlation": 276980
    }
  },
  {
    "ph": "f", "id": 276980, "pid": 76337, "tid": -914061504, "ts": 1716454225832054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832055, "dur": 0,
    "args": {
      "External id": 276981,
      "cbid": 205, "correlation": 276981
    }
  },
  {
    "ph": "f", "id": 276981, "pid": 76337, "tid": -914061504, "ts": 1716454225832055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886431, "dur": 5,
    "args": {
      "External id": 276985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276985,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276985, "pid": 5, "tid": 7, "ts": 1716454225886431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832069, "dur": 12,
    "args": {
      "External id": 276985,
      "cbid": 211, "correlation": 276985
    }
  },
  {
    "ph": "s", "id": 276985, "pid": 76337, "tid": -914061504, "ts": 1716454225832069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886437, "dur": 155,
    "args": {
      "External id": 276987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276987, "pid": 5, "tid": 7, "ts": 1716454225886437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832083, "dur": 5,
    "args": {
      "External id": 276987,
      "cbid": 211, "correlation": 276987
    }
  },
  {
    "ph": "s", "id": 276987, "pid": 76337, "tid": -914061504, "ts": 1716454225832083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225886594, "dur": 1,
    "args": {
      "External id": 276989,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 276989,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 276989, "pid": 5, "tid": 7, "ts": 1716454225886594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832094, "dur": 6,
    "args": {
      "External id": 276989,
      "cbid": 51, "correlation": 276989
    }
  },
  {
    "ph": "s", "id": 276989, "pid": 76337, "tid": -914061504, "ts": 1716454225832094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225886597, "dur": 250,
    "args": {
      "External id": 276990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276990,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 276990, "pid": 5, "tid": 7, "ts": 1716454225886597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832101, "dur": 6,
    "args": {
      "External id": 276990,
      "cbid": 211, "correlation": 276990
    }
  },
  {
    "ph": "s", "id": 276990, "pid": 76337, "tid": -914061504, "ts": 1716454225832101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886849, "dur": 6,
    "args": {
      "External id": 276992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 276992, "pid": 5, "tid": 7, "ts": 1716454225886849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832112, "dur": 5,
    "args": {
      "External id": 276992,
      "cbid": 211, "correlation": 276992
    }
  },
  {
    "ph": "s", "id": 276992, "pid": 76337, "tid": -914061504, "ts": 1716454225832112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225886856, "dur": 6,
    "args": {
      "External id": 276998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 276998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 276998, "pid": 5, "tid": 7, "ts": 1716454225886856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832140, "dur": 8,
    "args": {
      "External id": 276998,
      "cbid": 211, "correlation": 276998
    }
  },
  {
    "ph": "s", "id": 276998, "pid": 76337, "tid": -914061504, "ts": 1716454225832140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225886863, "dur": 5,
    "args": {
      "External id": 277006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277006,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277006, "pid": 5, "tid": 7, "ts": 1716454225886863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832173, "dur": 8,
    "args": {
      "External id": 277006,
      "cbid": 211, "correlation": 277006
    }
  },
  {
    "ph": "s", "id": 277006, "pid": 76337, "tid": -914061504, "ts": 1716454225832173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225886869, "dur": 5,
    "args": {
      "External id": 277014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277014,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277014, "pid": 5, "tid": 7, "ts": 1716454225886869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832202, "dur": 9,
    "args": {
      "External id": 277014,
      "cbid": 211, "correlation": 277014
    }
  },
  {
    "ph": "s", "id": 277014, "pid": 76337, "tid": -914061504, "ts": 1716454225832202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886875, "dur": 9,
    "args": {
      "External id": 277034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277034,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 277034, "pid": 5, "tid": 7, "ts": 1716454225886875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832305, "dur": 13,
    "args": {
      "External id": 277034,
      "cbid": 211, "correlation": 277034
    }
  },
  {
    "ph": "s", "id": 277034, "pid": 76337, "tid": -914061504, "ts": 1716454225832305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225886886, "dur": 3,
    "args": {
      "External id": 277046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277046,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 277046, "pid": 5, "tid": 7, "ts": 1716454225886886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832328, "dur": 6,
    "args": {
      "External id": 277046,
      "cbid": 211, "correlation": 277046
    }
  },
  {
    "ph": "s", "id": 277046, "pid": 76337, "tid": -914061504, "ts": 1716454225832328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225886890, "dur": 6,
    "args": {
      "External id": 277049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277049, "pid": 5, "tid": 7, "ts": 1716454225886890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832346, "dur": 6,
    "args": {
      "External id": 277049,
      "cbid": 211, "correlation": 277049
    }
  },
  {
    "ph": "s", "id": 277049, "pid": 76337, "tid": -914061504, "ts": 1716454225832346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225886898, "dur": 4,
    "args": {
      "External id": 277058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277058,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277058, "pid": 5, "tid": 7, "ts": 1716454225886898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832384, "dur": 9,
    "args": {
      "External id": 277058,
      "cbid": 211, "correlation": 277058
    }
  },
  {
    "ph": "s", "id": 277058, "pid": 76337, "tid": -914061504, "ts": 1716454225832384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832436, "dur": 0,
    "args": {
      "External id": 277068,
      "cbid": 317, "correlation": 277068
    }
  },
  {
    "ph": "f", "id": 277068, "pid": 76337, "tid": -914061504, "ts": 1716454225832436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832437, "dur": 0,
    "args": {
      "External id": 277069,
      "cbid": 203, "correlation": 277069
    }
  },
  {
    "ph": "f", "id": 277069, "pid": 76337, "tid": -914061504, "ts": 1716454225832437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832437, "dur": 0,
    "args": {
      "External id": 277070,
      "cbid": 205, "correlation": 277070
    }
  },
  {
    "ph": "f", "id": 277070, "pid": 76337, "tid": -914061504, "ts": 1716454225832437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886903, "dur": 5,
    "args": {
      "External id": 277074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277074,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277074, "pid": 5, "tid": 7, "ts": 1716454225886903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832450, "dur": 11,
    "args": {
      "External id": 277074,
      "cbid": 211, "correlation": 277074
    }
  },
  {
    "ph": "s", "id": 277074, "pid": 76337, "tid": -914061504, "ts": 1716454225832450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225886910, "dur": 155,
    "args": {
      "External id": 277076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277076, "pid": 5, "tid": 7, "ts": 1716454225886910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832465, "dur": 6,
    "args": {
      "External id": 277076,
      "cbid": 211, "correlation": 277076
    }
  },
  {
    "ph": "s", "id": 277076, "pid": 76337, "tid": -914061504, "ts": 1716454225832465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225887067, "dur": 1,
    "args": {
      "External id": 277078,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277078,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 277078, "pid": 5, "tid": 7, "ts": 1716454225887067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832476, "dur": 6,
    "args": {
      "External id": 277078,
      "cbid": 51, "correlation": 277078
    }
  },
  {
    "ph": "s", "id": 277078, "pid": 76337, "tid": -914061504, "ts": 1716454225832476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225887070, "dur": 250,
    "args": {
      "External id": 277079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277079,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277079, "pid": 5, "tid": 7, "ts": 1716454225887070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832483, "dur": 6,
    "args": {
      "External id": 277079,
      "cbid": 211, "correlation": 277079
    }
  },
  {
    "ph": "s", "id": 277079, "pid": 76337, "tid": -914061504, "ts": 1716454225832483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887321, "dur": 5,
    "args": {
      "External id": 277081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277081,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277081, "pid": 5, "tid": 7, "ts": 1716454225887321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832493, "dur": 5,
    "args": {
      "External id": 277081,
      "cbid": 211, "correlation": 277081
    }
  },
  {
    "ph": "s", "id": 277081, "pid": 76337, "tid": -914061504, "ts": 1716454225832493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225887328, "dur": 6,
    "args": {
      "External id": 277087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277087, "pid": 5, "tid": 7, "ts": 1716454225887328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832521, "dur": 8,
    "args": {
      "External id": 277087,
      "cbid": 211, "correlation": 277087
    }
  },
  {
    "ph": "s", "id": 277087, "pid": 76337, "tid": -914061504, "ts": 1716454225832521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225887335, "dur": 3,
    "args": {
      "External id": 277095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277095,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 277095, "pid": 5, "tid": 7, "ts": 1716454225887335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832564, "dur": 10,
    "args": {
      "External id": 277095,
      "cbid": 211, "correlation": 277095
    }
  },
  {
    "ph": "s", "id": 277095, "pid": 76337, "tid": -914061504, "ts": 1716454225832564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832625, "dur": 1,
    "args": {
      "External id": 277111,
      "cbid": 251, "correlation": 277111
    }
  },
  {
    "ph": "f", "id": 277111, "pid": 76337, "tid": -914061504, "ts": 1716454225832625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832631, "dur": 0,
    "args": {
      "External id": 277113,
      "cbid": 251, "correlation": 277113
    }
  },
  {
    "ph": "f", "id": 277113, "pid": 76337, "tid": -914061504, "ts": 1716454225832631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225887340, "dur": 10,
    "args": {
      "External id": 277114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277114,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277114, "pid": 5, "tid": 7, "ts": 1716454225887340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832632, "dur": 11,
    "args": {
      "External id": 277114,
      "cbid": 211, "correlation": 277114
    }
  },
  {
    "ph": "s", "id": 277114, "pid": 76337, "tid": -914061504, "ts": 1716454225832632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225887351, "dur": 3,
    "args": {
      "External id": 277116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277116,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277116, "pid": 5, "tid": 7, "ts": 1716454225887351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832645, "dur": 5,
    "args": {
      "External id": 277116,
      "cbid": 211, "correlation": 277116
    }
  },
  {
    "ph": "s", "id": 277116, "pid": 76337, "tid": -914061504, "ts": 1716454225832645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225887356, "dur": 5,
    "args": {
      "External id": 277126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277126,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277126, "pid": 5, "tid": 7, "ts": 1716454225887356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832701, "dur": 12,
    "args": {
      "External id": 277126,
      "cbid": 211, "correlation": 277126
    }
  },
  {
    "ph": "s", "id": 277126, "pid": 76337, "tid": -914061504, "ts": 1716454225832701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887362, "dur": 9,
    "args": {
      "External id": 277146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277146,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 277146, "pid": 5, "tid": 7, "ts": 1716454225887362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832766, "dur": 11,
    "args": {
      "External id": 277146,
      "cbid": 211, "correlation": 277146
    }
  },
  {
    "ph": "s", "id": 277146, "pid": 76337, "tid": -914061504, "ts": 1716454225832766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225887373, "dur": 4,
    "args": {
      "External id": 277158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277158,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 277158, "pid": 5, "tid": 7, "ts": 1716454225887373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832786, "dur": 6,
    "args": {
      "External id": 277158,
      "cbid": 211, "correlation": 277158
    }
  },
  {
    "ph": "s", "id": 277158, "pid": 76337, "tid": -914061504, "ts": 1716454225832786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225887378, "dur": 6,
    "args": {
      "External id": 277161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277161, "pid": 5, "tid": 7, "ts": 1716454225887378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832805, "dur": 6,
    "args": {
      "External id": 277161,
      "cbid": 211, "correlation": 277161
    }
  },
  {
    "ph": "s", "id": 277161, "pid": 76337, "tid": -914061504, "ts": 1716454225832805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225887385, "dur": 4,
    "args": {
      "External id": 277170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277170,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277170, "pid": 5, "tid": 7, "ts": 1716454225887385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832845, "dur": 10,
    "args": {
      "External id": 277170,
      "cbid": 211, "correlation": 277170
    }
  },
  {
    "ph": "s", "id": 277170, "pid": 76337, "tid": -914061504, "ts": 1716454225832845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832908, "dur": 0,
    "args": {
      "External id": 277180,
      "cbid": 317, "correlation": 277180
    }
  },
  {
    "ph": "f", "id": 277180, "pid": 76337, "tid": -914061504, "ts": 1716454225832908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832909, "dur": 0,
    "args": {
      "External id": 277181,
      "cbid": 203, "correlation": 277181
    }
  },
  {
    "ph": "f", "id": 277181, "pid": 76337, "tid": -914061504, "ts": 1716454225832909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832910, "dur": 0,
    "args": {
      "External id": 277182,
      "cbid": 205, "correlation": 277182
    }
  },
  {
    "ph": "f", "id": 277182, "pid": 76337, "tid": -914061504, "ts": 1716454225832910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887391, "dur": 5,
    "args": {
      "External id": 277186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277186, "pid": 5, "tid": 7, "ts": 1716454225887391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832924, "dur": 12,
    "args": {
      "External id": 277186,
      "cbid": 211, "correlation": 277186
    }
  },
  {
    "ph": "s", "id": 277186, "pid": 76337, "tid": -914061504, "ts": 1716454225832924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887397, "dur": 155,
    "args": {
      "External id": 277188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277188,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277188, "pid": 5, "tid": 7, "ts": 1716454225887397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832938, "dur": 5,
    "args": {
      "External id": 277188,
      "cbid": 211, "correlation": 277188
    }
  },
  {
    "ph": "s", "id": 277188, "pid": 76337, "tid": -914061504, "ts": 1716454225832938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225887555, "dur": 1,
    "args": {
      "External id": 277190,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277190,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 277190, "pid": 5, "tid": 7, "ts": 1716454225887555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832949, "dur": 6,
    "args": {
      "External id": 277190,
      "cbid": 51, "correlation": 277190
    }
  },
  {
    "ph": "s", "id": 277190, "pid": 76337, "tid": -914061504, "ts": 1716454225832949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225887558, "dur": 250,
    "args": {
      "External id": 277191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277191,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277191, "pid": 5, "tid": 7, "ts": 1716454225887558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832956, "dur": 6,
    "args": {
      "External id": 277191,
      "cbid": 211, "correlation": 277191
    }
  },
  {
    "ph": "s", "id": 277191, "pid": 76337, "tid": -914061504, "ts": 1716454225832956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887810, "dur": 6,
    "args": {
      "External id": 277193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277193,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277193, "pid": 5, "tid": 7, "ts": 1716454225887810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225832965, "dur": 5,
    "args": {
      "External id": 277193,
      "cbid": 211, "correlation": 277193
    }
  },
  {
    "ph": "s", "id": 277193, "pid": 76337, "tid": -914061504, "ts": 1716454225832965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225887817, "dur": 6,
    "args": {
      "External id": 277199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277199, "pid": 5, "tid": 7, "ts": 1716454225887817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833001, "dur": 9,
    "args": {
      "External id": 277199,
      "cbid": 211, "correlation": 277199
    }
  },
  {
    "ph": "s", "id": 277199, "pid": 76337, "tid": -914061504, "ts": 1716454225833001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225887824, "dur": 5,
    "args": {
      "External id": 277207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277207,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277207, "pid": 5, "tid": 7, "ts": 1716454225887824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833035, "dur": 8,
    "args": {
      "External id": 277207,
      "cbid": 211, "correlation": 277207
    }
  },
  {
    "ph": "s", "id": 277207, "pid": 76337, "tid": -914061504, "ts": 1716454225833035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225887830, "dur": 4,
    "args": {
      "External id": 277215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277215,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277215, "pid": 5, "tid": 7, "ts": 1716454225887830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833064, "dur": 9,
    "args": {
      "External id": 277215,
      "cbid": 211, "correlation": 277215
    }
  },
  {
    "ph": "s", "id": 277215, "pid": 76337, "tid": -914061504, "ts": 1716454225833064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887836, "dur": 9,
    "args": {
      "External id": 277235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277235,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 277235, "pid": 5, "tid": 7, "ts": 1716454225887836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833207, "dur": 14,
    "args": {
      "External id": 277235,
      "cbid": 211, "correlation": 277235
    }
  },
  {
    "ph": "s", "id": 277235, "pid": 76337, "tid": -914061504, "ts": 1716454225833207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225887846, "dur": 4,
    "args": {
      "External id": 277247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277247,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 277247, "pid": 5, "tid": 7, "ts": 1716454225887846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833233, "dur": 6,
    "args": {
      "External id": 277247,
      "cbid": 211, "correlation": 277247
    }
  },
  {
    "ph": "s", "id": 277247, "pid": 76337, "tid": -914061504, "ts": 1716454225833233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225887851, "dur": 6,
    "args": {
      "External id": 277250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277250,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277250, "pid": 5, "tid": 7, "ts": 1716454225887851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833252, "dur": 6,
    "args": {
      "External id": 277250,
      "cbid": 211, "correlation": 277250
    }
  },
  {
    "ph": "s", "id": 277250, "pid": 76337, "tid": -914061504, "ts": 1716454225833252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833311, "dur": 0,
    "args": {
      "External id": 277261,
      "cbid": 317, "correlation": 277261
    }
  },
  {
    "ph": "f", "id": 277261, "pid": 76337, "tid": -914061504, "ts": 1716454225833311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833312, "dur": 0,
    "args": {
      "External id": 277262,
      "cbid": 203, "correlation": 277262
    }
  },
  {
    "ph": "f", "id": 277262, "pid": 76337, "tid": -914061504, "ts": 1716454225833312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833313, "dur": 0,
    "args": {
      "External id": 277263,
      "cbid": 205, "correlation": 277263
    }
  },
  {
    "ph": "f", "id": 277263, "pid": 76337, "tid": -914061504, "ts": 1716454225833313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887858, "dur": 5,
    "args": {
      "External id": 277267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277267,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277267, "pid": 5, "tid": 7, "ts": 1716454225887858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833330, "dur": 12,
    "args": {
      "External id": 277267,
      "cbid": 211, "correlation": 277267
    }
  },
  {
    "ph": "s", "id": 277267, "pid": 76337, "tid": -914061504, "ts": 1716454225833330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225887864, "dur": 35,
    "args": {
      "External id": 277269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277269,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 277269, "pid": 5, "tid": 7, "ts": 1716454225887864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833352, "dur": 10,
    "args": {
      "External id": 277269,
      "cbid": 211, "correlation": 277269
    }
  },
  {
    "ph": "s", "id": 277269, "pid": 76337, "tid": -914061504, "ts": 1716454225833352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887901, "dur": 5,
    "args": {
      "External id": 277271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277271, "pid": 5, "tid": 7, "ts": 1716454225887901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833366, "dur": 5,
    "args": {
      "External id": 277271,
      "cbid": 211, "correlation": 277271
    }
  },
  {
    "ph": "s", "id": 277271, "pid": 76337, "tid": -914061504, "ts": 1716454225833366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225887908, "dur": 6,
    "args": {
      "External id": 277277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277277, "pid": 5, "tid": 7, "ts": 1716454225887908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833393, "dur": 8,
    "args": {
      "External id": 277277,
      "cbid": 211, "correlation": 277277
    }
  },
  {
    "ph": "s", "id": 277277, "pid": 76337, "tid": -914061504, "ts": 1716454225833393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225887915, "dur": 19,
    "args": {
      "External id": 277286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277286,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277286, "pid": 5, "tid": 7, "ts": 1716454225887915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833476, "dur": 14,
    "args": {
      "External id": 277286,
      "cbid": 211, "correlation": 277286
    }
  },
  {
    "ph": "s", "id": 277286, "pid": 76337, "tid": -914061504, "ts": 1716454225833476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225887936, "dur": 10,
    "args": {
      "External id": 277308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277308,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 277308, "pid": 5, "tid": 7, "ts": 1716454225887936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833534, "dur": 10,
    "args": {
      "External id": 277308,
      "cbid": 211, "correlation": 277308
    }
  },
  {
    "ph": "s", "id": 277308, "pid": 76337, "tid": -914061504, "ts": 1716454225833534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833625, "dur": 2,
    "args": {
      "External id": 277319,
      "cbid": 251, "correlation": 277319
    }
  },
  {
    "ph": "f", "id": 277319, "pid": 76337, "tid": -914061504, "ts": 1716454225833625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833630, "dur": 0,
    "args": {
      "External id": 277320,
      "cbid": 251, "correlation": 277320
    }
  },
  {
    "ph": "f", "id": 277320, "pid": 76337, "tid": -914061504, "ts": 1716454225833630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225887947, "dur": 51,
    "args": {
      "External id": 277321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277321,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 277321, "pid": 5, "tid": 7, "ts": 1716454225887947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833633, "dur": 14,
    "args": {
      "External id": 277321,
      "cbid": 211, "correlation": 277321
    }
  },
  {
    "ph": "s", "id": 277321, "pid": 76337, "tid": -914061504, "ts": 1716454225833633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833708, "dur": 1,
    "args": {
      "External id": 277332,
      "cbid": 251, "correlation": 277332
    }
  },
  {
    "ph": "f", "id": 277332, "pid": 76337, "tid": -914061504, "ts": 1716454225833708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833712, "dur": 0,
    "args": {
      "External id": 277333,
      "cbid": 251, "correlation": 277333
    }
  },
  {
    "ph": "f", "id": 277333, "pid": 76337, "tid": -914061504, "ts": 1716454225833712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225887999, "dur": 51,
    "args": {
      "External id": 277334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277334,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 277334, "pid": 5, "tid": 7, "ts": 1716454225887999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833714, "dur": 12,
    "args": {
      "External id": 277334,
      "cbid": 211, "correlation": 277334
    }
  },
  {
    "ph": "s", "id": 277334, "pid": 76337, "tid": -914061504, "ts": 1716454225833714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833780, "dur": 1,
    "args": {
      "External id": 277345,
      "cbid": 251, "correlation": 277345
    }
  },
  {
    "ph": "f", "id": 277345, "pid": 76337, "tid": -914061504, "ts": 1716454225833780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833783, "dur": 0,
    "args": {
      "External id": 277346,
      "cbid": 251, "correlation": 277346
    }
  },
  {
    "ph": "f", "id": 277346, "pid": 76337, "tid": -914061504, "ts": 1716454225833783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225888052, "dur": 51,
    "args": {
      "External id": 277347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277347,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 277347, "pid": 5, "tid": 7, "ts": 1716454225888052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833785, "dur": 11,
    "args": {
      "External id": 277347,
      "cbid": 211, "correlation": 277347
    }
  },
  {
    "ph": "s", "id": 277347, "pid": 76337, "tid": -914061504, "ts": 1716454225833785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225888104, "dur": 53,
    "args": {
      "External id": 277372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277372,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277372, "pid": 5, "tid": 7, "ts": 1716454225888104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833870, "dur": 13,
    "args": {
      "External id": 277372,
      "cbid": 211, "correlation": 277372
    }
  },
  {
    "ph": "s", "id": 277372, "pid": 76337, "tid": -914061504, "ts": 1716454225833870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833972, "dur": 8,
    "args": {
      "External id": 277390,
      "cbid": 251, "correlation": 277390
    }
  },
  {
    "ph": "f", "id": 277390, "pid": 76337, "tid": -914061504, "ts": 1716454225833972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225888159, "dur": 60,
    "args": {
      "External id": 277392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277392,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 277392, "pid": 5, "tid": 7, "ts": 1716454225888159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225833985, "dur": 14,
    "args": {
      "External id": 277392,
      "cbid": 211, "correlation": 277392
    }
  },
  {
    "ph": "s", "id": 277392, "pid": 76337, "tid": -914061504, "ts": 1716454225833985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225888220, "dur": 6,
    "args": {
      "External id": 277400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277400,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277400, "pid": 5, "tid": 7, "ts": 1716454225888220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834057, "dur": 12,
    "args": {
      "External id": 277400,
      "cbid": 211, "correlation": 277400
    }
  },
  {
    "ph": "s", "id": 277400, "pid": 76337, "tid": -914061504, "ts": 1716454225834057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225888227, "dur": 7,
    "args": {
      "External id": 277408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277408,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277408, "pid": 5, "tid": 7, "ts": 1716454225888227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834095, "dur": 9,
    "args": {
      "External id": 277408,
      "cbid": 211, "correlation": 277408
    }
  },
  {
    "ph": "s", "id": 277408, "pid": 76337, "tid": -914061504, "ts": 1716454225834095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888236, "dur": 7,
    "args": {
      "External id": 277419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277419, "pid": 5, "tid": 7, "ts": 1716454225888236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834168, "dur": 12,
    "args": {
      "External id": 277419,
      "cbid": 211, "correlation": 277419
    }
  },
  {
    "ph": "s", "id": 277419, "pid": 76337, "tid": -914061504, "ts": 1716454225834168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888245, "dur": 8,
    "args": {
      "External id": 277441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277441,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 277441, "pid": 5, "tid": 7, "ts": 1716454225888245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834203, "dur": 8,
    "args": {
      "External id": 277441,
      "cbid": 211, "correlation": 277441
    }
  },
  {
    "ph": "s", "id": 277441, "pid": 76337, "tid": -914061504, "ts": 1716454225834203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834290, "dur": 5,
    "args": {
      "External id": 277452,
      "cbid": 251, "correlation": 277452
    }
  },
  {
    "ph": "f", "id": 277452, "pid": 76337, "tid": -914061504, "ts": 1716454225834290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225888255, "dur": 1,
    "args": {
      "External id": 277453,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277453,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 277453, "pid": 5, "tid": 7, "ts": 1716454225888255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834298, "dur": 11,
    "args": {
      "External id": 277453,
      "cbid": 51, "correlation": 277453
    }
  },
  {
    "ph": "s", "id": 277453, "pid": 76337, "tid": -914061504, "ts": 1716454225834298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225888258, "dur": 35,
    "args": {
      "External id": 277454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277454,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 277454, "pid": 5, "tid": 7, "ts": 1716454225888258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834311, "dur": 13,
    "args": {
      "External id": 277454,
      "cbid": 211, "correlation": 277454
    }
  },
  {
    "ph": "s", "id": 277454, "pid": 76337, "tid": -914061504, "ts": 1716454225834311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834383, "dur": 1,
    "args": {
      "External id": 277465,
      "cbid": 251, "correlation": 277465
    }
  },
  {
    "ph": "f", "id": 277465, "pid": 76337, "tid": -914061504, "ts": 1716454225834383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834387, "dur": 0,
    "args": {
      "External id": 277466,
      "cbid": 251, "correlation": 277466
    }
  },
  {
    "ph": "f", "id": 277466, "pid": 76337, "tid": -914061504, "ts": 1716454225834387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225888295, "dur": 12,
    "args": {
      "External id": 277467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277467,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277467, "pid": 5, "tid": 7, "ts": 1716454225888295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834388, "dur": 12,
    "args": {
      "External id": 277467,
      "cbid": 211, "correlation": 277467
    }
  },
  {
    "ph": "s", "id": 277467, "pid": 76337, "tid": -914061504, "ts": 1716454225834388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225888308, "dur": 5,
    "args": {
      "External id": 277469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277469,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277469, "pid": 5, "tid": 7, "ts": 1716454225888308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834403, "dur": 6,
    "args": {
      "External id": 277469,
      "cbid": 211, "correlation": 277469
    }
  },
  {
    "ph": "s", "id": 277469, "pid": 76337, "tid": -914061504, "ts": 1716454225834403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834460, "dur": 1,
    "args": {
      "External id": 277480,
      "cbid": 251, "correlation": 277480
    }
  },
  {
    "ph": "f", "id": 277480, "pid": 76337, "tid": -914061504, "ts": 1716454225834460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834464, "dur": 0,
    "args": {
      "External id": 277481,
      "cbid": 251, "correlation": 277481
    }
  },
  {
    "ph": "f", "id": 277481, "pid": 76337, "tid": -914061504, "ts": 1716454225834464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225888314, "dur": 8,
    "args": {
      "External id": 277482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277482,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277482, "pid": 5, "tid": 7, "ts": 1716454225888314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834465, "dur": 11,
    "args": {
      "External id": 277482,
      "cbid": 211, "correlation": 277482
    }
  },
  {
    "ph": "s", "id": 277482, "pid": 76337, "tid": -914061504, "ts": 1716454225834465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225888324, "dur": 3,
    "args": {
      "External id": 277484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277484,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277484, "pid": 5, "tid": 7, "ts": 1716454225888324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834478, "dur": 5,
    "args": {
      "External id": 277484,
      "cbid": 211, "correlation": 277484
    }
  },
  {
    "ph": "s", "id": 277484, "pid": 76337, "tid": -914061504, "ts": 1716454225834478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225888329, "dur": 20,
    "args": {
      "External id": 277509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277509,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 277509, "pid": 5, "tid": 7, "ts": 1716454225888329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834557, "dur": 12,
    "args": {
      "External id": 277509,
      "cbid": 211, "correlation": 277509
    }
  },
  {
    "ph": "s", "id": 277509, "pid": 76337, "tid": -914061504, "ts": 1716454225834557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834660, "dur": 3,
    "args": {
      "External id": 277527,
      "cbid": 251, "correlation": 277527
    }
  },
  {
    "ph": "f", "id": 277527, "pid": 76337, "tid": -914061504, "ts": 1716454225834660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225888351, "dur": 1,
    "args": {
      "External id": 277529,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277529,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 277529, "pid": 5, "tid": 7, "ts": 1716454225888351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834667, "dur": 10,
    "args": {
      "External id": 277529,
      "cbid": 51, "correlation": 277529
    }
  },
  {
    "ph": "s", "id": 277529, "pid": 76337, "tid": -914061504, "ts": 1716454225834667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225888354, "dur": 34,
    "args": {
      "External id": 277530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277530,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 277530, "pid": 5, "tid": 7, "ts": 1716454225888354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834679, "dur": 12,
    "args": {
      "External id": 277530,
      "cbid": 211, "correlation": 277530
    }
  },
  {
    "ph": "s", "id": 277530, "pid": 76337, "tid": -914061504, "ts": 1716454225834679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225888390, "dur": 4,
    "args": {
      "External id": 277538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277538,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277538, "pid": 5, "tid": 7, "ts": 1716454225888390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834750, "dur": 13,
    "args": {
      "External id": 277538,
      "cbid": 211, "correlation": 277538
    }
  },
  {
    "ph": "s", "id": 277538, "pid": 76337, "tid": -914061504, "ts": 1716454225834750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888395, "dur": 8,
    "args": {
      "External id": 277546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277546, "pid": 5, "tid": 7, "ts": 1716454225888395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834793, "dur": 10,
    "args": {
      "External id": 277546,
      "cbid": 211, "correlation": 277546
    }
  },
  {
    "ph": "s", "id": 277546, "pid": 76337, "tid": -914061504, "ts": 1716454225834793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888405, "dur": 8,
    "args": {
      "External id": 277568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277568,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 277568, "pid": 5, "tid": 7, "ts": 1716454225888405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834844, "dur": 10,
    "args": {
      "External id": 277568,
      "cbid": 211, "correlation": 277568
    }
  },
  {
    "ph": "s", "id": 277568, "pid": 76337, "tid": -914061504, "ts": 1716454225834844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834936, "dur": 1,
    "args": {
      "External id": 277584,
      "cbid": 251, "correlation": 277584
    }
  },
  {
    "ph": "f", "id": 277584, "pid": 76337, "tid": -914061504, "ts": 1716454225834936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834941, "dur": 0,
    "args": {
      "External id": 277586,
      "cbid": 251, "correlation": 277586
    }
  },
  {
    "ph": "f", "id": 277586, "pid": 76337, "tid": -914061504, "ts": 1716454225834941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225888414, "dur": 186,
    "args": {
      "External id": 277587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277587,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277587, "pid": 5, "tid": 7, "ts": 1716454225888414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225834943, "dur": 13,
    "args": {
      "External id": 277587,
      "cbid": 211, "correlation": 277587
    }
  },
  {
    "ph": "s", "id": 277587, "pid": 76337, "tid": -914061504, "ts": 1716454225834943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888601, "dur": 20,
    "args": {
      "External id": 277595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277595,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277595, "pid": 5, "tid": 7, "ts": 1716454225888601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835019, "dur": 13,
    "args": {
      "External id": 277595,
      "cbid": 211, "correlation": 277595
    }
  },
  {
    "ph": "s", "id": 277595, "pid": 76337, "tid": -914061504, "ts": 1716454225835019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888622, "dur": 21,
    "args": {
      "External id": 277603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277603, "pid": 5, "tid": 7, "ts": 1716454225888622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835051, "dur": 9,
    "args": {
      "External id": 277603,
      "cbid": 211, "correlation": 277603
    }
  },
  {
    "ph": "s", "id": 277603, "pid": 76337, "tid": -914061504, "ts": 1716454225835051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835133, "dur": 1,
    "args": {
      "External id": 277619,
      "cbid": 251, "correlation": 277619
    }
  },
  {
    "ph": "f", "id": 277619, "pid": 76337, "tid": -914061504, "ts": 1716454225835133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225888645, "dur": 1,
    "args": {
      "External id": 277621,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277621,
      "bytes": 120, "memory bandwidth (GB/s)": 0.08152173913043478
    }
  },
  {
    "ph": "f", "id": 277621, "pid": 5, "tid": 7, "ts": 1716454225888645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835138, "dur": 8,
    "args": {
      "External id": 277621,
      "cbid": 51, "correlation": 277621
    }
  },
  {
    "ph": "s", "id": 277621, "pid": 76337, "tid": -914061504, "ts": 1716454225835138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225888649, "dur": 107,
    "args": {
      "External id": 277622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277622,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 277622, "pid": 5, "tid": 7, "ts": 1716454225888649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835148, "dur": 12,
    "args": {
      "External id": 277622,
      "cbid": 211, "correlation": 277622
    }
  },
  {
    "ph": "s", "id": 277622, "pid": 76337, "tid": -914061504, "ts": 1716454225835148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225888757, "dur": 5,
    "args": {
      "External id": 277630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277630,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277630, "pid": 5, "tid": 7, "ts": 1716454225888757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835191, "dur": 10,
    "args": {
      "External id": 277630,
      "cbid": 211, "correlation": 277630
    }
  },
  {
    "ph": "s", "id": 277630, "pid": 76337, "tid": -914061504, "ts": 1716454225835191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888764, "dur": 9,
    "args": {
      "External id": 277641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277641,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277641, "pid": 5, "tid": 7, "ts": 1716454225888764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835259, "dur": 12,
    "args": {
      "External id": 277641,
      "cbid": 211, "correlation": 277641
    }
  },
  {
    "ph": "s", "id": 277641, "pid": 76337, "tid": -914061504, "ts": 1716454225835259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835324, "dur": 0,
    "args": {
      "External id": 277653,
      "cbid": 317, "correlation": 277653
    }
  },
  {
    "ph": "f", "id": 277653, "pid": 76337, "tid": -914061504, "ts": 1716454225835324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835325, "dur": 0,
    "args": {
      "External id": 277654,
      "cbid": 203, "correlation": 277654
    }
  },
  {
    "ph": "f", "id": 277654, "pid": 76337, "tid": -914061504, "ts": 1716454225835325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835325, "dur": 0,
    "args": {
      "External id": 277655,
      "cbid": 205, "correlation": 277655
    }
  },
  {
    "ph": "f", "id": 277655, "pid": 76337, "tid": -914061504, "ts": 1716454225835325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888774, "dur": 5,
    "args": {
      "External id": 277659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277659, "pid": 5, "tid": 7, "ts": 1716454225888774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835340, "dur": 12,
    "args": {
      "External id": 277659,
      "cbid": 211, "correlation": 277659
    }
  },
  {
    "ph": "s", "id": 277659, "pid": 76337, "tid": -914061504, "ts": 1716454225835340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225888781, "dur": 36,
    "args": {
      "External id": 277661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277661,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 277661, "pid": 5, "tid": 7, "ts": 1716454225888781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835359, "dur": 7,
    "args": {
      "External id": 277661,
      "cbid": 211, "correlation": 277661
    }
  },
  {
    "ph": "s", "id": 277661, "pid": 76337, "tid": -914061504, "ts": 1716454225835359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888818, "dur": 5,
    "args": {
      "External id": 277663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277663,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277663, "pid": 5, "tid": 7, "ts": 1716454225888818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835370, "dur": 5,
    "args": {
      "External id": 277663,
      "cbid": 211, "correlation": 277663
    }
  },
  {
    "ph": "s", "id": 277663, "pid": 76337, "tid": -914061504, "ts": 1716454225835370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888825, "dur": 7,
    "args": {
      "External id": 277669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277669, "pid": 5, "tid": 7, "ts": 1716454225888825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835397, "dur": 9,
    "args": {
      "External id": 277669,
      "cbid": 211, "correlation": 277669
    }
  },
  {
    "ph": "s", "id": 277669, "pid": 76337, "tid": -914061504, "ts": 1716454225835397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225888833, "dur": 5,
    "args": {
      "External id": 277677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277677,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277677, "pid": 5, "tid": 7, "ts": 1716454225888833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835429, "dur": 8,
    "args": {
      "External id": 277677,
      "cbid": 211, "correlation": 277677
    }
  },
  {
    "ph": "s", "id": 277677, "pid": 76337, "tid": -914061504, "ts": 1716454225835429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888840, "dur": 10,
    "args": {
      "External id": 277697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277697,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 277697, "pid": 5, "tid": 7, "ts": 1716454225888840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835500, "dur": 13,
    "args": {
      "External id": 277697,
      "cbid": 211, "correlation": 277697
    }
  },
  {
    "ph": "s", "id": 277697, "pid": 76337, "tid": -914061504, "ts": 1716454225835500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225888851, "dur": 4,
    "args": {
      "External id": 277709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277709,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 277709, "pid": 5, "tid": 7, "ts": 1716454225888851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835522, "dur": 6,
    "args": {
      "External id": 277709,
      "cbid": 211, "correlation": 277709
    }
  },
  {
    "ph": "s", "id": 277709, "pid": 76337, "tid": -914061504, "ts": 1716454225835522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225888857, "dur": 8,
    "args": {
      "External id": 277712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277712,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277712, "pid": 5, "tid": 7, "ts": 1716454225888857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835540, "dur": 7,
    "args": {
      "External id": 277712,
      "cbid": 211, "correlation": 277712
    }
  },
  {
    "ph": "s", "id": 277712, "pid": 76337, "tid": -914061504, "ts": 1716454225835540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225888867, "dur": 5,
    "args": {
      "External id": 277721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277721,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277721, "pid": 5, "tid": 7, "ts": 1716454225888867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835579, "dur": 10,
    "args": {
      "External id": 277721,
      "cbid": 211, "correlation": 277721
    }
  },
  {
    "ph": "s", "id": 277721, "pid": 76337, "tid": -914061504, "ts": 1716454225835579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835632, "dur": 0,
    "args": {
      "External id": 277731,
      "cbid": 317, "correlation": 277731
    }
  },
  {
    "ph": "f", "id": 277731, "pid": 76337, "tid": -914061504, "ts": 1716454225835632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835633, "dur": 0,
    "args": {
      "External id": 277732,
      "cbid": 203, "correlation": 277732
    }
  },
  {
    "ph": "f", "id": 277732, "pid": 76337, "tid": -914061504, "ts": 1716454225835633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835634, "dur": 0,
    "args": {
      "External id": 277733,
      "cbid": 205, "correlation": 277733
    }
  },
  {
    "ph": "f", "id": 277733, "pid": 76337, "tid": -914061504, "ts": 1716454225835634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888873, "dur": 5,
    "args": {
      "External id": 277737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277737,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277737, "pid": 5, "tid": 7, "ts": 1716454225888873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835647, "dur": 12,
    "args": {
      "External id": 277737,
      "cbid": 211, "correlation": 277737
    }
  },
  {
    "ph": "s", "id": 277737, "pid": 76337, "tid": -914061504, "ts": 1716454225835647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225888879, "dur": 154,
    "args": {
      "External id": 277739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277739, "pid": 5, "tid": 7, "ts": 1716454225888879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835661, "dur": 5,
    "args": {
      "External id": 277739,
      "cbid": 211, "correlation": 277739
    }
  },
  {
    "ph": "s", "id": 277739, "pid": 76337, "tid": -914061504, "ts": 1716454225835661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225889035, "dur": 1,
    "args": {
      "External id": 277741,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277741,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 277741, "pid": 5, "tid": 7, "ts": 1716454225889035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835672, "dur": 6,
    "args": {
      "External id": 277741,
      "cbid": 51, "correlation": 277741
    }
  },
  {
    "ph": "s", "id": 277741, "pid": 76337, "tid": -914061504, "ts": 1716454225835672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225889039, "dur": 260,
    "args": {
      "External id": 277742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277742,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277742, "pid": 5, "tid": 7, "ts": 1716454225889039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835679, "dur": 6,
    "args": {
      "External id": 277742,
      "cbid": 211, "correlation": 277742
    }
  },
  {
    "ph": "s", "id": 277742, "pid": 76337, "tid": -914061504, "ts": 1716454225835679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889300, "dur": 5,
    "args": {
      "External id": 277744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277744, "pid": 5, "tid": 7, "ts": 1716454225889300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835689, "dur": 5,
    "args": {
      "External id": 277744,
      "cbid": 211, "correlation": 277744
    }
  },
  {
    "ph": "s", "id": 277744, "pid": 76337, "tid": -914061504, "ts": 1716454225835689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225889307, "dur": 6,
    "args": {
      "External id": 277750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277750, "pid": 5, "tid": 7, "ts": 1716454225889307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835717, "dur": 8,
    "args": {
      "External id": 277750,
      "cbid": 211, "correlation": 277750
    }
  },
  {
    "ph": "s", "id": 277750, "pid": 76337, "tid": -914061504, "ts": 1716454225835717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225889314, "dur": 4,
    "args": {
      "External id": 277758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277758,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 277758, "pid": 5, "tid": 7, "ts": 1716454225889314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835762, "dur": 10,
    "args": {
      "External id": 277758,
      "cbid": 211, "correlation": 277758
    }
  },
  {
    "ph": "s", "id": 277758, "pid": 76337, "tid": -914061504, "ts": 1716454225835762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835829, "dur": 1,
    "args": {
      "External id": 277774,
      "cbid": 251, "correlation": 277774
    }
  },
  {
    "ph": "f", "id": 277774, "pid": 76337, "tid": -914061504, "ts": 1716454225835829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835834, "dur": 0,
    "args": {
      "External id": 277776,
      "cbid": 251, "correlation": 277776
    }
  },
  {
    "ph": "f", "id": 277776, "pid": 76337, "tid": -914061504, "ts": 1716454225835834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225889319, "dur": 13,
    "args": {
      "External id": 277777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277777,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277777, "pid": 5, "tid": 7, "ts": 1716454225889319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835835, "dur": 11,
    "args": {
      "External id": 277777,
      "cbid": 211, "correlation": 277777
    }
  },
  {
    "ph": "s", "id": 277777, "pid": 76337, "tid": -914061504, "ts": 1716454225835835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225889333, "dur": 5,
    "args": {
      "External id": 277779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277779,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277779, "pid": 5, "tid": 7, "ts": 1716454225889333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835848, "dur": 5,
    "args": {
      "External id": 277779,
      "cbid": 211, "correlation": 277779
    }
  },
  {
    "ph": "s", "id": 277779, "pid": 76337, "tid": -914061504, "ts": 1716454225835848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225889339, "dur": 5,
    "args": {
      "External id": 277789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277789,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277789, "pid": 5, "tid": 7, "ts": 1716454225889339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835905, "dur": 13,
    "args": {
      "External id": 277789,
      "cbid": 211, "correlation": 277789
    }
  },
  {
    "ph": "s", "id": 277789, "pid": 76337, "tid": -914061504, "ts": 1716454225835905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889346, "dur": 10,
    "args": {
      "External id": 277809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277809,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 277809, "pid": 5, "tid": 7, "ts": 1716454225889346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225835972, "dur": 18,
    "args": {
      "External id": 277809,
      "cbid": 211, "correlation": 277809
    }
  },
  {
    "ph": "s", "id": 277809, "pid": 76337, "tid": -914061504, "ts": 1716454225835972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225889357, "dur": 4,
    "args": {
      "External id": 277821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277821,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 277821, "pid": 5, "tid": 7, "ts": 1716454225889357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836000, "dur": 6,
    "args": {
      "External id": 277821,
      "cbid": 211, "correlation": 277821
    }
  },
  {
    "ph": "s", "id": 277821, "pid": 76337, "tid": -914061504, "ts": 1716454225836000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225889362, "dur": 6,
    "args": {
      "External id": 277824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277824, "pid": 5, "tid": 7, "ts": 1716454225889362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836020, "dur": 7,
    "args": {
      "External id": 277824,
      "cbid": 211, "correlation": 277824
    }
  },
  {
    "ph": "s", "id": 277824, "pid": 76337, "tid": -914061504, "ts": 1716454225836020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225889369, "dur": 4,
    "args": {
      "External id": 277833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277833,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277833, "pid": 5, "tid": 7, "ts": 1716454225889369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836062, "dur": 10,
    "args": {
      "External id": 277833,
      "cbid": 211, "correlation": 277833
    }
  },
  {
    "ph": "s", "id": 277833, "pid": 76337, "tid": -914061504, "ts": 1716454225836062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836124, "dur": 0,
    "args": {
      "External id": 277843,
      "cbid": 317, "correlation": 277843
    }
  },
  {
    "ph": "f", "id": 277843, "pid": 76337, "tid": -914061504, "ts": 1716454225836124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836124, "dur": 0,
    "args": {
      "External id": 277844,
      "cbid": 203, "correlation": 277844
    }
  },
  {
    "ph": "f", "id": 277844, "pid": 76337, "tid": -914061504, "ts": 1716454225836124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836125, "dur": 0,
    "args": {
      "External id": 277845,
      "cbid": 205, "correlation": 277845
    }
  },
  {
    "ph": "f", "id": 277845, "pid": 76337, "tid": -914061504, "ts": 1716454225836125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889375, "dur": 5,
    "args": {
      "External id": 277849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277849, "pid": 5, "tid": 7, "ts": 1716454225889375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836138, "dur": 12,
    "args": {
      "External id": 277849,
      "cbid": 211, "correlation": 277849
    }
  },
  {
    "ph": "s", "id": 277849, "pid": 76337, "tid": -914061504, "ts": 1716454225836138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889381, "dur": 155,
    "args": {
      "External id": 277851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277851, "pid": 5, "tid": 7, "ts": 1716454225889381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836153, "dur": 5,
    "args": {
      "External id": 277851,
      "cbid": 211, "correlation": 277851
    }
  },
  {
    "ph": "s", "id": 277851, "pid": 76337, "tid": -914061504, "ts": 1716454225836153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225889538, "dur": 1,
    "args": {
      "External id": 277853,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277853,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 277853, "pid": 5, "tid": 7, "ts": 1716454225889538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836164, "dur": 6,
    "args": {
      "External id": 277853,
      "cbid": 51, "correlation": 277853
    }
  },
  {
    "ph": "s", "id": 277853, "pid": 76337, "tid": -914061504, "ts": 1716454225836164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225889541, "dur": 250,
    "args": {
      "External id": 277854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277854,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277854, "pid": 5, "tid": 7, "ts": 1716454225889541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836171, "dur": 6,
    "args": {
      "External id": 277854,
      "cbid": 211, "correlation": 277854
    }
  },
  {
    "ph": "s", "id": 277854, "pid": 76337, "tid": -914061504, "ts": 1716454225836171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889793, "dur": 6,
    "args": {
      "External id": 277856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277856, "pid": 5, "tid": 7, "ts": 1716454225889793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836181, "dur": 5,
    "args": {
      "External id": 277856,
      "cbid": 211, "correlation": 277856
    }
  },
  {
    "ph": "s", "id": 277856, "pid": 76337, "tid": -914061504, "ts": 1716454225836181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225889800, "dur": 6,
    "args": {
      "External id": 277862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277862,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277862, "pid": 5, "tid": 7, "ts": 1716454225889800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836209, "dur": 9,
    "args": {
      "External id": 277862,
      "cbid": 211, "correlation": 277862
    }
  },
  {
    "ph": "s", "id": 277862, "pid": 76337, "tid": -914061504, "ts": 1716454225836209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225889807, "dur": 5,
    "args": {
      "External id": 277870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277870,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277870, "pid": 5, "tid": 7, "ts": 1716454225889807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836242, "dur": 8,
    "args": {
      "External id": 277870,
      "cbid": 211, "correlation": 277870
    }
  },
  {
    "ph": "s", "id": 277870, "pid": 76337, "tid": -914061504, "ts": 1716454225836242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225889813, "dur": 4,
    "args": {
      "External id": 277878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277878,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277878, "pid": 5, "tid": 7, "ts": 1716454225889813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836271, "dur": 8,
    "args": {
      "External id": 277878,
      "cbid": 211, "correlation": 277878
    }
  },
  {
    "ph": "s", "id": 277878, "pid": 76337, "tid": -914061504, "ts": 1716454225836271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225889819, "dur": 11,
    "args": {
      "External id": 277887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277887,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277887, "pid": 5, "tid": 7, "ts": 1716454225889819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836360, "dur": 13,
    "args": {
      "External id": 277887,
      "cbid": 211, "correlation": 277887
    }
  },
  {
    "ph": "s", "id": 277887, "pid": 76337, "tid": -914061504, "ts": 1716454225836360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889831, "dur": 12,
    "args": {
      "External id": 277907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277907,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 277907, "pid": 5, "tid": 7, "ts": 1716454225889831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836425, "dur": 11,
    "args": {
      "External id": 277907,
      "cbid": 211, "correlation": 277907
    }
  },
  {
    "ph": "s", "id": 277907, "pid": 76337, "tid": -914061504, "ts": 1716454225836425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225889844, "dur": 4,
    "args": {
      "External id": 277919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277919,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277919, "pid": 5, "tid": 7, "ts": 1716454225889844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836446, "dur": 6,
    "args": {
      "External id": 277919,
      "cbid": 211, "correlation": 277919
    }
  },
  {
    "ph": "s", "id": 277919, "pid": 76337, "tid": -914061504, "ts": 1716454225836446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225889849, "dur": 9,
    "args": {
      "External id": 277922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277922, "pid": 5, "tid": 7, "ts": 1716454225889849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836465, "dur": 6,
    "args": {
      "External id": 277922,
      "cbid": 211, "correlation": 277922
    }
  },
  {
    "ph": "s", "id": 277922, "pid": 76337, "tid": -914061504, "ts": 1716454225836465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225889860, "dur": 6,
    "args": {
      "External id": 277931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277931,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277931, "pid": 5, "tid": 7, "ts": 1716454225889860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836503, "dur": 10,
    "args": {
      "External id": 277931,
      "cbid": 211, "correlation": 277931
    }
  },
  {
    "ph": "s", "id": 277931, "pid": 76337, "tid": -914061504, "ts": 1716454225836503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836554, "dur": 0,
    "args": {
      "External id": 277941,
      "cbid": 317, "correlation": 277941
    }
  },
  {
    "ph": "f", "id": 277941, "pid": 76337, "tid": -914061504, "ts": 1716454225836554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836555, "dur": 0,
    "args": {
      "External id": 277942,
      "cbid": 203, "correlation": 277942
    }
  },
  {
    "ph": "f", "id": 277942, "pid": 76337, "tid": -914061504, "ts": 1716454225836555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836556, "dur": 0,
    "args": {
      "External id": 277943,
      "cbid": 205, "correlation": 277943
    }
  },
  {
    "ph": "f", "id": 277943, "pid": 76337, "tid": -914061504, "ts": 1716454225836556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889867, "dur": 6,
    "args": {
      "External id": 277947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277947, "pid": 5, "tid": 7, "ts": 1716454225889867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836571, "dur": 11,
    "args": {
      "External id": 277947,
      "cbid": 211, "correlation": 277947
    }
  },
  {
    "ph": "s", "id": 277947, "pid": 76337, "tid": -914061504, "ts": 1716454225836571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225889875, "dur": 306,
    "args": {
      "External id": 277949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277949, "pid": 5, "tid": 7, "ts": 1716454225889875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836585, "dur": 5,
    "args": {
      "External id": 277949,
      "cbid": 211, "correlation": 277949
    }
  },
  {
    "ph": "s", "id": 277949, "pid": 76337, "tid": -914061504, "ts": 1716454225836585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225890183, "dur": 1,
    "args": {
      "External id": 277951,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 277951,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 277951, "pid": 5, "tid": 7, "ts": 1716454225890183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836597, "dur": 8,
    "args": {
      "External id": 277951,
      "cbid": 51, "correlation": 277951
    }
  },
  {
    "ph": "s", "id": 277951, "pid": 76337, "tid": -914061504, "ts": 1716454225836597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225890186, "dur": 479,
    "args": {
      "External id": 277952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277952,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277952, "pid": 5, "tid": 7, "ts": 1716454225890186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836606, "dur": 7,
    "args": {
      "External id": 277952,
      "cbid": 211, "correlation": 277952
    }
  },
  {
    "ph": "s", "id": 277952, "pid": 76337, "tid": -914061504, "ts": 1716454225836606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225890667, "dur": 6,
    "args": {
      "External id": 277954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277954,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 277954, "pid": 5, "tid": 7, "ts": 1716454225890667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836617, "dur": 5,
    "args": {
      "External id": 277954,
      "cbid": 211, "correlation": 277954
    }
  },
  {
    "ph": "s", "id": 277954, "pid": 76337, "tid": -914061504, "ts": 1716454225836617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225890674, "dur": 6,
    "args": {
      "External id": 277960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277960, "pid": 5, "tid": 7, "ts": 1716454225890674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836649, "dur": 10,
    "args": {
      "External id": 277960,
      "cbid": 211, "correlation": 277960
    }
  },
  {
    "ph": "s", "id": 277960, "pid": 76337, "tid": -914061504, "ts": 1716454225836649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225890681, "dur": 3,
    "args": {
      "External id": 277968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277968,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 277968, "pid": 5, "tid": 7, "ts": 1716454225890681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836694, "dur": 10,
    "args": {
      "External id": 277968,
      "cbid": 211, "correlation": 277968
    }
  },
  {
    "ph": "s", "id": 277968, "pid": 76337, "tid": -914061504, "ts": 1716454225836694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836762, "dur": 1,
    "args": {
      "External id": 277984,
      "cbid": 251, "correlation": 277984
    }
  },
  {
    "ph": "f", "id": 277984, "pid": 76337, "tid": -914061504, "ts": 1716454225836762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836767, "dur": 0,
    "args": {
      "External id": 277986,
      "cbid": 251, "correlation": 277986
    }
  },
  {
    "ph": "f", "id": 277986, "pid": 76337, "tid": -914061504, "ts": 1716454225836767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225890686, "dur": 10,
    "args": {
      "External id": 277987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277987,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277987, "pid": 5, "tid": 7, "ts": 1716454225890686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836769, "dur": 11,
    "args": {
      "External id": 277987,
      "cbid": 211, "correlation": 277987
    }
  },
  {
    "ph": "s", "id": 277987, "pid": 76337, "tid": -914061504, "ts": 1716454225836769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225890697, "dur": 4,
    "args": {
      "External id": 277989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277989,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 277989, "pid": 5, "tid": 7, "ts": 1716454225890697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836782, "dur": 5,
    "args": {
      "External id": 277989,
      "cbid": 211, "correlation": 277989
    }
  },
  {
    "ph": "s", "id": 277989, "pid": 76337, "tid": -914061504, "ts": 1716454225836782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225890703, "dur": 6,
    "args": {
      "External id": 277999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 277999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 277999, "pid": 5, "tid": 7, "ts": 1716454225890703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836839, "dur": 12,
    "args": {
      "External id": 277999,
      "cbid": 211, "correlation": 277999
    }
  },
  {
    "ph": "s", "id": 277999, "pid": 76337, "tid": -914061504, "ts": 1716454225836839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225890710, "dur": 9,
    "args": {
      "External id": 278019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278019,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278019, "pid": 5, "tid": 7, "ts": 1716454225890710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836905, "dur": 12,
    "args": {
      "External id": 278019,
      "cbid": 211, "correlation": 278019
    }
  },
  {
    "ph": "s", "id": 278019, "pid": 76337, "tid": -914061504, "ts": 1716454225836905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225890720, "dur": 4,
    "args": {
      "External id": 278031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278031,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 278031, "pid": 5, "tid": 7, "ts": 1716454225890720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836926, "dur": 6,
    "args": {
      "External id": 278031,
      "cbid": 211, "correlation": 278031
    }
  },
  {
    "ph": "s", "id": 278031, "pid": 76337, "tid": -914061504, "ts": 1716454225836926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225890725, "dur": 6,
    "args": {
      "External id": 278034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278034, "pid": 5, "tid": 7, "ts": 1716454225890725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836945, "dur": 6,
    "args": {
      "External id": 278034,
      "cbid": 211, "correlation": 278034
    }
  },
  {
    "ph": "s", "id": 278034, "pid": 76337, "tid": -914061504, "ts": 1716454225836945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225890733, "dur": 4,
    "args": {
      "External id": 278043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278043,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278043, "pid": 5, "tid": 7, "ts": 1716454225890733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225836993, "dur": 10,
    "args": {
      "External id": 278043,
      "cbid": 211, "correlation": 278043
    }
  },
  {
    "ph": "s", "id": 278043, "pid": 76337, "tid": -914061504, "ts": 1716454225836993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837057, "dur": 0,
    "args": {
      "External id": 278053,
      "cbid": 317, "correlation": 278053
    }
  },
  {
    "ph": "f", "id": 278053, "pid": 76337, "tid": -914061504, "ts": 1716454225837057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837058, "dur": 0,
    "args": {
      "External id": 278054,
      "cbid": 203, "correlation": 278054
    }
  },
  {
    "ph": "f", "id": 278054, "pid": 76337, "tid": -914061504, "ts": 1716454225837058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837058, "dur": 0,
    "args": {
      "External id": 278055,
      "cbid": 205, "correlation": 278055
    }
  },
  {
    "ph": "f", "id": 278055, "pid": 76337, "tid": -914061504, "ts": 1716454225837058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225890738, "dur": 5,
    "args": {
      "External id": 278059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278059,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278059, "pid": 5, "tid": 7, "ts": 1716454225890738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837073, "dur": 11,
    "args": {
      "External id": 278059,
      "cbid": 211, "correlation": 278059
    }
  },
  {
    "ph": "s", "id": 278059, "pid": 76337, "tid": -914061504, "ts": 1716454225837073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225890744, "dur": 155,
    "args": {
      "External id": 278061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278061,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278061, "pid": 5, "tid": 7, "ts": 1716454225890744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837087, "dur": 5,
    "args": {
      "External id": 278061,
      "cbid": 211, "correlation": 278061
    }
  },
  {
    "ph": "s", "id": 278061, "pid": 76337, "tid": -914061504, "ts": 1716454225837087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225890901, "dur": 1,
    "args": {
      "External id": 278063,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278063,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 278063, "pid": 5, "tid": 7, "ts": 1716454225890901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837098, "dur": 6,
    "args": {
      "External id": 278063,
      "cbid": 51, "correlation": 278063
    }
  },
  {
    "ph": "s", "id": 278063, "pid": 76337, "tid": -914061504, "ts": 1716454225837098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225890905, "dur": 250,
    "args": {
      "External id": 278064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278064,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278064, "pid": 5, "tid": 7, "ts": 1716454225890905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837105, "dur": 6,
    "args": {
      "External id": 278064,
      "cbid": 211, "correlation": 278064
    }
  },
  {
    "ph": "s", "id": 278064, "pid": 76337, "tid": -914061504, "ts": 1716454225837105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225891156, "dur": 5,
    "args": {
      "External id": 278066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278066,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278066, "pid": 5, "tid": 7, "ts": 1716454225891156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837116, "dur": 6,
    "args": {
      "External id": 278066,
      "cbid": 211, "correlation": 278066
    }
  },
  {
    "ph": "s", "id": 278066, "pid": 76337, "tid": -914061504, "ts": 1716454225837116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225891163, "dur": 6,
    "args": {
      "External id": 278072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278072, "pid": 5, "tid": 7, "ts": 1716454225891163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837144, "dur": 8,
    "args": {
      "External id": 278072,
      "cbid": 211, "correlation": 278072
    }
  },
  {
    "ph": "s", "id": 278072, "pid": 76337, "tid": -914061504, "ts": 1716454225837144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837204, "dur": 0,
    "args": {
      "External id": 278082,
      "cbid": 317, "correlation": 278082
    }
  },
  {
    "ph": "f", "id": 278082, "pid": 76337, "tid": -914061504, "ts": 1716454225837204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837205, "dur": 0,
    "args": {
      "External id": 278083,
      "cbid": 203, "correlation": 278083
    }
  },
  {
    "ph": "f", "id": 278083, "pid": 76337, "tid": -914061504, "ts": 1716454225837205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837206, "dur": 0,
    "args": {
      "External id": 278084,
      "cbid": 205, "correlation": 278084
    }
  },
  {
    "ph": "f", "id": 278084, "pid": 76337, "tid": -914061504, "ts": 1716454225837206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225891170, "dur": 7,
    "args": {
      "External id": 278088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278088, "pid": 5, "tid": 7, "ts": 1716454225891170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837219, "dur": 12,
    "args": {
      "External id": 278088,
      "cbid": 211, "correlation": 278088
    }
  },
  {
    "ph": "s", "id": 278088, "pid": 76337, "tid": -914061504, "ts": 1716454225837219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225891179, "dur": 3,
    "args": {
      "External id": 278090,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278090,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5625
    }
  },
  {
    "ph": "f", "id": 278090, "pid": 5, "tid": 7, "ts": 1716454225891179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837237, "dur": 14,
    "args": {
      "External id": 278090,
      "cbid": 51, "correlation": 278090
    }
  },
  {
    "ph": "s", "id": 278090, "pid": 76337, "tid": -914061504, "ts": 1716454225837237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225891182, "dur": 96,
    "args": {
      "External id": 278091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278091,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 278091, "pid": 5, "tid": 7, "ts": 1716454225891182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837253, "dur": 6,
    "args": {
      "External id": 278091,
      "cbid": 211, "correlation": 278091
    }
  },
  {
    "ph": "s", "id": 278091, "pid": 76337, "tid": -914061504, "ts": 1716454225837253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225891280, "dur": 5,
    "args": {
      "External id": 278093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278093,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278093, "pid": 5, "tid": 7, "ts": 1716454225891280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837263, "dur": 6,
    "args": {
      "External id": 278093,
      "cbid": 211, "correlation": 278093
    }
  },
  {
    "ph": "s", "id": 278093, "pid": 76337, "tid": -914061504, "ts": 1716454225837263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225891286, "dur": 6,
    "args": {
      "External id": 278099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278099, "pid": 5, "tid": 7, "ts": 1716454225891286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837290, "dur": 8,
    "args": {
      "External id": 278099,
      "cbid": 211, "correlation": 278099
    }
  },
  {
    "ph": "s", "id": 278099, "pid": 76337, "tid": -914061504, "ts": 1716454225837290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225891293, "dur": 5,
    "args": {
      "External id": 278107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278107,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278107, "pid": 5, "tid": 7, "ts": 1716454225891293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837320, "dur": 8,
    "args": {
      "External id": 278107,
      "cbid": 211, "correlation": 278107
    }
  },
  {
    "ph": "s", "id": 278107, "pid": 76337, "tid": -914061504, "ts": 1716454225837320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225891300, "dur": 4,
    "args": {
      "External id": 278115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278115,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278115, "pid": 5, "tid": 7, "ts": 1716454225891300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837348, "dur": 8,
    "args": {
      "External id": 278115,
      "cbid": 211, "correlation": 278115
    }
  },
  {
    "ph": "s", "id": 278115, "pid": 76337, "tid": -914061504, "ts": 1716454225837348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225891305, "dur": 11,
    "args": {
      "External id": 278124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278124,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278124, "pid": 5, "tid": 7, "ts": 1716454225891305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837393, "dur": 10,
    "args": {
      "External id": 278124,
      "cbid": 211, "correlation": 278124
    }
  },
  {
    "ph": "s", "id": 278124, "pid": 76337, "tid": -914061504, "ts": 1716454225837393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225891317, "dur": 11,
    "args": {
      "External id": 278144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278144,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278144, "pid": 5, "tid": 7, "ts": 1716454225891317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837461, "dur": 12,
    "args": {
      "External id": 278144,
      "cbid": 211, "correlation": 278144
    }
  },
  {
    "ph": "s", "id": 278144, "pid": 76337, "tid": -914061504, "ts": 1716454225837461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225891330, "dur": 4,
    "args": {
      "External id": 278156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278156,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278156, "pid": 5, "tid": 7, "ts": 1716454225891330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837483, "dur": 6,
    "args": {
      "External id": 278156,
      "cbid": 211, "correlation": 278156
    }
  },
  {
    "ph": "s", "id": 278156, "pid": 76337, "tid": -914061504, "ts": 1716454225837483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225891335, "dur": 11,
    "args": {
      "External id": 278159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278159,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278159, "pid": 5, "tid": 7, "ts": 1716454225891335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837501, "dur": 6,
    "args": {
      "External id": 278159,
      "cbid": 211, "correlation": 278159
    }
  },
  {
    "ph": "s", "id": 278159, "pid": 76337, "tid": -914061504, "ts": 1716454225837501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225891347, "dur": 6,
    "args": {
      "External id": 278168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278168,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278168, "pid": 5, "tid": 7, "ts": 1716454225891347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837539, "dur": 9,
    "args": {
      "External id": 278168,
      "cbid": 211, "correlation": 278168
    }
  },
  {
    "ph": "s", "id": 278168, "pid": 76337, "tid": -914061504, "ts": 1716454225837539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837590, "dur": 0,
    "args": {
      "External id": 278178,
      "cbid": 317, "correlation": 278178
    }
  },
  {
    "ph": "f", "id": 278178, "pid": 76337, "tid": -914061504, "ts": 1716454225837590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837590, "dur": 0,
    "args": {
      "External id": 278179,
      "cbid": 203, "correlation": 278179
    }
  },
  {
    "ph": "f", "id": 278179, "pid": 76337, "tid": -914061504, "ts": 1716454225837590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837591, "dur": 0,
    "args": {
      "External id": 278180,
      "cbid": 205, "correlation": 278180
    }
  },
  {
    "ph": "f", "id": 278180, "pid": 76337, "tid": -914061504, "ts": 1716454225837591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225891354, "dur": 6,
    "args": {
      "External id": 278184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278184,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278184, "pid": 5, "tid": 7, "ts": 1716454225891354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837605, "dur": 11,
    "args": {
      "External id": 278184,
      "cbid": 211, "correlation": 278184
    }
  },
  {
    "ph": "s", "id": 278184, "pid": 76337, "tid": -914061504, "ts": 1716454225837605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225891362, "dur": 306,
    "args": {
      "External id": 278186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278186, "pid": 5, "tid": 7, "ts": 1716454225891362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837619, "dur": 6,
    "args": {
      "External id": 278186,
      "cbid": 211, "correlation": 278186
    }
  },
  {
    "ph": "s", "id": 278186, "pid": 76337, "tid": -914061504, "ts": 1716454225837619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225891670, "dur": 1,
    "args": {
      "External id": 278188,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278188,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 278188, "pid": 5, "tid": 7, "ts": 1716454225891670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837630, "dur": 6,
    "args": {
      "External id": 278188,
      "cbid": 51, "correlation": 278188
    }
  },
  {
    "ph": "s", "id": 278188, "pid": 76337, "tid": -914061504, "ts": 1716454225837630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225891674, "dur": 480,
    "args": {
      "External id": 278189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278189,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278189, "pid": 5, "tid": 7, "ts": 1716454225891674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837637, "dur": 6,
    "args": {
      "External id": 278189,
      "cbid": 211, "correlation": 278189
    }
  },
  {
    "ph": "s", "id": 278189, "pid": 76337, "tid": -914061504, "ts": 1716454225837637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892156, "dur": 6,
    "args": {
      "External id": 278191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278191,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278191, "pid": 5, "tid": 7, "ts": 1716454225892156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837648, "dur": 5,
    "args": {
      "External id": 278191,
      "cbid": 211, "correlation": 278191
    }
  },
  {
    "ph": "s", "id": 278191, "pid": 76337, "tid": -914061504, "ts": 1716454225837648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225892163, "dur": 6,
    "args": {
      "External id": 278197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278197, "pid": 5, "tid": 7, "ts": 1716454225892163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837677, "dur": 8,
    "args": {
      "External id": 278197,
      "cbid": 211, "correlation": 278197
    }
  },
  {
    "ph": "s", "id": 278197, "pid": 76337, "tid": -914061504, "ts": 1716454225837677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225892170, "dur": 4,
    "args": {
      "External id": 278205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278205,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 278205, "pid": 5, "tid": 7, "ts": 1716454225892170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837719, "dur": 10,
    "args": {
      "External id": 278205,
      "cbid": 211, "correlation": 278205
    }
  },
  {
    "ph": "s", "id": 278205, "pid": 76337, "tid": -914061504, "ts": 1716454225837719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837782, "dur": 1,
    "args": {
      "External id": 278221,
      "cbid": 251, "correlation": 278221
    }
  },
  {
    "ph": "f", "id": 278221, "pid": 76337, "tid": -914061504, "ts": 1716454225837782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837787, "dur": 0,
    "args": {
      "External id": 278223,
      "cbid": 251, "correlation": 278223
    }
  },
  {
    "ph": "f", "id": 278223, "pid": 76337, "tid": -914061504, "ts": 1716454225837787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225892175, "dur": 12,
    "args": {
      "External id": 278224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278224,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278224, "pid": 5, "tid": 7, "ts": 1716454225892175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837789, "dur": 11,
    "args": {
      "External id": 278224,
      "cbid": 211, "correlation": 278224
    }
  },
  {
    "ph": "s", "id": 278224, "pid": 76337, "tid": -914061504, "ts": 1716454225837789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225892188, "dur": 5,
    "args": {
      "External id": 278226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278226,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278226, "pid": 5, "tid": 7, "ts": 1716454225892188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837802, "dur": 5,
    "args": {
      "External id": 278226,
      "cbid": 211, "correlation": 278226
    }
  },
  {
    "ph": "s", "id": 278226, "pid": 76337, "tid": -914061504, "ts": 1716454225837802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225892194, "dur": 6,
    "args": {
      "External id": 278236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278236, "pid": 5, "tid": 7, "ts": 1716454225892194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837858, "dur": 12,
    "args": {
      "External id": 278236,
      "cbid": 211, "correlation": 278236
    }
  },
  {
    "ph": "s", "id": 278236, "pid": 76337, "tid": -914061504, "ts": 1716454225837858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892202, "dur": 9,
    "args": {
      "External id": 278256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278256,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278256, "pid": 5, "tid": 7, "ts": 1716454225892202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837924, "dur": 11,
    "args": {
      "External id": 278256,
      "cbid": 211, "correlation": 278256
    }
  },
  {
    "ph": "s", "id": 278256, "pid": 76337, "tid": -914061504, "ts": 1716454225837924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225892212, "dur": 4,
    "args": {
      "External id": 278268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278268,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 278268, "pid": 5, "tid": 7, "ts": 1716454225892212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837945, "dur": 6,
    "args": {
      "External id": 278268,
      "cbid": 211, "correlation": 278268
    }
  },
  {
    "ph": "s", "id": 278268, "pid": 76337, "tid": -914061504, "ts": 1716454225837945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225892217, "dur": 7,
    "args": {
      "External id": 278271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278271, "pid": 5, "tid": 7, "ts": 1716454225892217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225837963, "dur": 6,
    "args": {
      "External id": 278271,
      "cbid": 211, "correlation": 278271
    }
  },
  {
    "ph": "s", "id": 278271, "pid": 76337, "tid": -914061504, "ts": 1716454225837963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225892225, "dur": 4,
    "args": {
      "External id": 278280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278280,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278280, "pid": 5, "tid": 7, "ts": 1716454225892225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838014, "dur": 10,
    "args": {
      "External id": 278280,
      "cbid": 211, "correlation": 278280
    }
  },
  {
    "ph": "s", "id": 278280, "pid": 76337, "tid": -914061504, "ts": 1716454225838014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838078, "dur": 0,
    "args": {
      "External id": 278290,
      "cbid": 317, "correlation": 278290
    }
  },
  {
    "ph": "f", "id": 278290, "pid": 76337, "tid": -914061504, "ts": 1716454225838078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838079, "dur": 0,
    "args": {
      "External id": 278291,
      "cbid": 203, "correlation": 278291
    }
  },
  {
    "ph": "f", "id": 278291, "pid": 76337, "tid": -914061504, "ts": 1716454225838079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838080, "dur": 0,
    "args": {
      "External id": 278292,
      "cbid": 205, "correlation": 278292
    }
  },
  {
    "ph": "f", "id": 278292, "pid": 76337, "tid": -914061504, "ts": 1716454225838080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892230, "dur": 5,
    "args": {
      "External id": 278296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278296,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278296, "pid": 5, "tid": 7, "ts": 1716454225892230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838093, "dur": 12,
    "args": {
      "External id": 278296,
      "cbid": 211, "correlation": 278296
    }
  },
  {
    "ph": "s", "id": 278296, "pid": 76337, "tid": -914061504, "ts": 1716454225838093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892236, "dur": 155,
    "args": {
      "External id": 278298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278298,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278298, "pid": 5, "tid": 7, "ts": 1716454225892236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838107, "dur": 5,
    "args": {
      "External id": 278298,
      "cbid": 211, "correlation": 278298
    }
  },
  {
    "ph": "s", "id": 278298, "pid": 76337, "tid": -914061504, "ts": 1716454225838107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225892394, "dur": 1,
    "args": {
      "External id": 278300,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278300,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 278300, "pid": 5, "tid": 7, "ts": 1716454225892394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838118, "dur": 7,
    "args": {
      "External id": 278300,
      "cbid": 51, "correlation": 278300
    }
  },
  {
    "ph": "s", "id": 278300, "pid": 76337, "tid": -914061504, "ts": 1716454225838118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225892397, "dur": 250,
    "args": {
      "External id": 278301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278301,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278301, "pid": 5, "tid": 7, "ts": 1716454225892397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838126, "dur": 6,
    "args": {
      "External id": 278301,
      "cbid": 211, "correlation": 278301
    }
  },
  {
    "ph": "s", "id": 278301, "pid": 76337, "tid": -914061504, "ts": 1716454225838126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892649, "dur": 5,
    "args": {
      "External id": 278303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278303, "pid": 5, "tid": 7, "ts": 1716454225892649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838136, "dur": 5,
    "args": {
      "External id": 278303,
      "cbid": 211, "correlation": 278303
    }
  },
  {
    "ph": "s", "id": 278303, "pid": 76337, "tid": -914061504, "ts": 1716454225838136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225892655, "dur": 6,
    "args": {
      "External id": 278309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278309, "pid": 5, "tid": 7, "ts": 1716454225892655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838163, "dur": 9,
    "args": {
      "External id": 278309,
      "cbid": 211, "correlation": 278309
    }
  },
  {
    "ph": "s", "id": 278309, "pid": 76337, "tid": -914061504, "ts": 1716454225838163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838222, "dur": 0,
    "args": {
      "External id": 278319,
      "cbid": 317, "correlation": 278319
    }
  },
  {
    "ph": "f", "id": 278319, "pid": 76337, "tid": -914061504, "ts": 1716454225838222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838223, "dur": 0,
    "args": {
      "External id": 278320,
      "cbid": 203, "correlation": 278320
    }
  },
  {
    "ph": "f", "id": 278320, "pid": 76337, "tid": -914061504, "ts": 1716454225838223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838224, "dur": 0,
    "args": {
      "External id": 278321,
      "cbid": 205, "correlation": 278321
    }
  },
  {
    "ph": "f", "id": 278321, "pid": 76337, "tid": -914061504, "ts": 1716454225838224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892662, "dur": 8,
    "args": {
      "External id": 278325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278325,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278325, "pid": 5, "tid": 7, "ts": 1716454225892662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838236, "dur": 12,
    "args": {
      "External id": 278325,
      "cbid": 211, "correlation": 278325
    }
  },
  {
    "ph": "s", "id": 278325, "pid": 76337, "tid": -914061504, "ts": 1716454225838236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225892671, "dur": 3,
    "args": {
      "External id": 278327,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278327,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5463917525773196
    }
  },
  {
    "ph": "f", "id": 278327, "pid": 5, "tid": 7, "ts": 1716454225892671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838253, "dur": 9,
    "args": {
      "External id": 278327,
      "cbid": 51, "correlation": 278327
    }
  },
  {
    "ph": "s", "id": 278327, "pid": 76337, "tid": -914061504, "ts": 1716454225838253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225892675, "dur": 94,
    "args": {
      "External id": 278328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278328,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 278328, "pid": 5, "tid": 7, "ts": 1716454225892675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838263, "dur": 6,
    "args": {
      "External id": 278328,
      "cbid": 211, "correlation": 278328
    }
  },
  {
    "ph": "s", "id": 278328, "pid": 76337, "tid": -914061504, "ts": 1716454225838263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892770, "dur": 5,
    "args": {
      "External id": 278330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278330,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278330, "pid": 5, "tid": 7, "ts": 1716454225892770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838272, "dur": 6,
    "args": {
      "External id": 278330,
      "cbid": 211, "correlation": 278330
    }
  },
  {
    "ph": "s", "id": 278330, "pid": 76337, "tid": -914061504, "ts": 1716454225838272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225892777, "dur": 6,
    "args": {
      "External id": 278336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278336,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278336, "pid": 5, "tid": 7, "ts": 1716454225892777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838300, "dur": 8,
    "args": {
      "External id": 278336,
      "cbid": 211, "correlation": 278336
    }
  },
  {
    "ph": "s", "id": 278336, "pid": 76337, "tid": -914061504, "ts": 1716454225838300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225892784, "dur": 5,
    "args": {
      "External id": 278344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278344,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278344, "pid": 5, "tid": 7, "ts": 1716454225892784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838330, "dur": 8,
    "args": {
      "External id": 278344,
      "cbid": 211, "correlation": 278344
    }
  },
  {
    "ph": "s", "id": 278344, "pid": 76337, "tid": -914061504, "ts": 1716454225838330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225892791, "dur": 4,
    "args": {
      "External id": 278352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278352,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278352, "pid": 5, "tid": 7, "ts": 1716454225892791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838358, "dur": 8,
    "args": {
      "External id": 278352,
      "cbid": 211, "correlation": 278352
    }
  },
  {
    "ph": "s", "id": 278352, "pid": 76337, "tid": -914061504, "ts": 1716454225838358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225892796, "dur": 10,
    "args": {
      "External id": 278361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278361,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278361, "pid": 5, "tid": 7, "ts": 1716454225892796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838400, "dur": 10,
    "args": {
      "External id": 278361,
      "cbid": 211, "correlation": 278361
    }
  },
  {
    "ph": "s", "id": 278361, "pid": 76337, "tid": -914061504, "ts": 1716454225838400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892808, "dur": 12,
    "args": {
      "External id": 278381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278381,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278381, "pid": 5, "tid": 7, "ts": 1716454225892808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838470, "dur": 12,
    "args": {
      "External id": 278381,
      "cbid": 211, "correlation": 278381
    }
  },
  {
    "ph": "s", "id": 278381, "pid": 76337, "tid": -914061504, "ts": 1716454225838470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225892821, "dur": 4,
    "args": {
      "External id": 278393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278393,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278393, "pid": 5, "tid": 7, "ts": 1716454225892821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838492, "dur": 6,
    "args": {
      "External id": 278393,
      "cbid": 211, "correlation": 278393
    }
  },
  {
    "ph": "s", "id": 278393, "pid": 76337, "tid": -914061504, "ts": 1716454225838492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225892826, "dur": 10,
    "args": {
      "External id": 278396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278396, "pid": 5, "tid": 7, "ts": 1716454225892826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838509, "dur": 7,
    "args": {
      "External id": 278396,
      "cbid": 211, "correlation": 278396
    }
  },
  {
    "ph": "s", "id": 278396, "pid": 76337, "tid": -914061504, "ts": 1716454225838509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225892838, "dur": 6,
    "args": {
      "External id": 278405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278405,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278405, "pid": 5, "tid": 7, "ts": 1716454225892838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838548, "dur": 9,
    "args": {
      "External id": 278405,
      "cbid": 211, "correlation": 278405
    }
  },
  {
    "ph": "s", "id": 278405, "pid": 76337, "tid": -914061504, "ts": 1716454225838548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838599, "dur": 0,
    "args": {
      "External id": 278415,
      "cbid": 317, "correlation": 278415
    }
  },
  {
    "ph": "f", "id": 278415, "pid": 76337, "tid": -914061504, "ts": 1716454225838599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838600, "dur": 0,
    "args": {
      "External id": 278416,
      "cbid": 203, "correlation": 278416
    }
  },
  {
    "ph": "f", "id": 278416, "pid": 76337, "tid": -914061504, "ts": 1716454225838600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838601, "dur": 0,
    "args": {
      "External id": 278417,
      "cbid": 205, "correlation": 278417
    }
  },
  {
    "ph": "f", "id": 278417, "pid": 76337, "tid": -914061504, "ts": 1716454225838601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892845, "dur": 6,
    "args": {
      "External id": 278421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278421,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278421, "pid": 5, "tid": 7, "ts": 1716454225892845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838614, "dur": 11,
    "args": {
      "External id": 278421,
      "cbid": 211, "correlation": 278421
    }
  },
  {
    "ph": "s", "id": 278421, "pid": 76337, "tid": -914061504, "ts": 1716454225838614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225892852, "dur": 307,
    "args": {
      "External id": 278423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278423,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278423, "pid": 5, "tid": 7, "ts": 1716454225892852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838628, "dur": 6,
    "args": {
      "External id": 278423,
      "cbid": 211, "correlation": 278423
    }
  },
  {
    "ph": "s", "id": 278423, "pid": 76337, "tid": -914061504, "ts": 1716454225838628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225893161, "dur": 1,
    "args": {
      "External id": 278425,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278425,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 278425, "pid": 5, "tid": 7, "ts": 1716454225893161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838639, "dur": 6,
    "args": {
      "External id": 278425,
      "cbid": 51, "correlation": 278425
    }
  },
  {
    "ph": "s", "id": 278425, "pid": 76337, "tid": -914061504, "ts": 1716454225838639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225893165, "dur": 480,
    "args": {
      "External id": 278426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278426,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278426, "pid": 5, "tid": 7, "ts": 1716454225893165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838646, "dur": 6,
    "args": {
      "External id": 278426,
      "cbid": 211, "correlation": 278426
    }
  },
  {
    "ph": "s", "id": 278426, "pid": 76337, "tid": -914061504, "ts": 1716454225838646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225893646, "dur": 5,
    "args": {
      "External id": 278428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278428,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278428, "pid": 5, "tid": 7, "ts": 1716454225893646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838655, "dur": 5,
    "args": {
      "External id": 278428,
      "cbid": 211, "correlation": 278428
    }
  },
  {
    "ph": "s", "id": 278428, "pid": 76337, "tid": -914061504, "ts": 1716454225838655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225893653, "dur": 6,
    "args": {
      "External id": 278434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278434,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278434, "pid": 5, "tid": 7, "ts": 1716454225893653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838684, "dur": 9,
    "args": {
      "External id": 278434,
      "cbid": 211, "correlation": 278434
    }
  },
  {
    "ph": "s", "id": 278434, "pid": 76337, "tid": -914061504, "ts": 1716454225838684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225893660, "dur": 4,
    "args": {
      "External id": 278442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278442,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 278442, "pid": 5, "tid": 7, "ts": 1716454225893660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838726, "dur": 10,
    "args": {
      "External id": 278442,
      "cbid": 211, "correlation": 278442
    }
  },
  {
    "ph": "s", "id": 278442, "pid": 76337, "tid": -914061504, "ts": 1716454225838726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838788, "dur": 1,
    "args": {
      "External id": 278458,
      "cbid": 251, "correlation": 278458
    }
  },
  {
    "ph": "f", "id": 278458, "pid": 76337, "tid": -914061504, "ts": 1716454225838788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838793, "dur": 0,
    "args": {
      "External id": 278460,
      "cbid": 251, "correlation": 278460
    }
  },
  {
    "ph": "f", "id": 278460, "pid": 76337, "tid": -914061504, "ts": 1716454225838793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225893665, "dur": 12,
    "args": {
      "External id": 278461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278461,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278461, "pid": 5, "tid": 7, "ts": 1716454225893665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838795, "dur": 11,
    "args": {
      "External id": 278461,
      "cbid": 211, "correlation": 278461
    }
  },
  {
    "ph": "s", "id": 278461, "pid": 76337, "tid": -914061504, "ts": 1716454225838795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225893679, "dur": 5,
    "args": {
      "External id": 278463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278463,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278463, "pid": 5, "tid": 7, "ts": 1716454225893679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838808, "dur": 5,
    "args": {
      "External id": 278463,
      "cbid": 211, "correlation": 278463
    }
  },
  {
    "ph": "s", "id": 278463, "pid": 76337, "tid": -914061504, "ts": 1716454225838808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225893685, "dur": 6,
    "args": {
      "External id": 278473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278473, "pid": 5, "tid": 7, "ts": 1716454225893685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838864, "dur": 11,
    "args": {
      "External id": 278473,
      "cbid": 211, "correlation": 278473
    }
  },
  {
    "ph": "s", "id": 278473, "pid": 76337, "tid": -914061504, "ts": 1716454225838864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225893692, "dur": 9,
    "args": {
      "External id": 278493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278493,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278493, "pid": 5, "tid": 7, "ts": 1716454225893692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838929, "dur": 11,
    "args": {
      "External id": 278493,
      "cbid": 211, "correlation": 278493
    }
  },
  {
    "ph": "s", "id": 278493, "pid": 76337, "tid": -914061504, "ts": 1716454225838929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225893702, "dur": 4,
    "args": {
      "External id": 278505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278505,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 278505, "pid": 5, "tid": 7, "ts": 1716454225893702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838950, "dur": 6,
    "args": {
      "External id": 278505,
      "cbid": 211, "correlation": 278505
    }
  },
  {
    "ph": "s", "id": 278505, "pid": 76337, "tid": -914061504, "ts": 1716454225838950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225893707, "dur": 6,
    "args": {
      "External id": 278508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278508, "pid": 5, "tid": 7, "ts": 1716454225893707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225838969, "dur": 14,
    "args": {
      "External id": 278508,
      "cbid": 211, "correlation": 278508
    }
  },
  {
    "ph": "s", "id": 278508, "pid": 76337, "tid": -914061504, "ts": 1716454225838969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225893715, "dur": 4,
    "args": {
      "External id": 278517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278517,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278517, "pid": 5, "tid": 7, "ts": 1716454225893715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839018, "dur": 10,
    "args": {
      "External id": 278517,
      "cbid": 211, "correlation": 278517
    }
  },
  {
    "ph": "s", "id": 278517, "pid": 76337, "tid": -914061504, "ts": 1716454225839018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839082, "dur": 0,
    "args": {
      "External id": 278527,
      "cbid": 317, "correlation": 278527
    }
  },
  {
    "ph": "f", "id": 278527, "pid": 76337, "tid": -914061504, "ts": 1716454225839082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839083, "dur": 0,
    "args": {
      "External id": 278528,
      "cbid": 203, "correlation": 278528
    }
  },
  {
    "ph": "f", "id": 278528, "pid": 76337, "tid": -914061504, "ts": 1716454225839083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839084, "dur": 0,
    "args": {
      "External id": 278529,
      "cbid": 205, "correlation": 278529
    }
  },
  {
    "ph": "f", "id": 278529, "pid": 76337, "tid": -914061504, "ts": 1716454225839084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225893721, "dur": 5,
    "args": {
      "External id": 278533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278533,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278533, "pid": 5, "tid": 7, "ts": 1716454225893721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839098, "dur": 12,
    "args": {
      "External id": 278533,
      "cbid": 211, "correlation": 278533
    }
  },
  {
    "ph": "s", "id": 278533, "pid": 76337, "tid": -914061504, "ts": 1716454225839098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225893727, "dur": 155,
    "args": {
      "External id": 278535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278535,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278535, "pid": 5, "tid": 7, "ts": 1716454225893727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839112, "dur": 5,
    "args": {
      "External id": 278535,
      "cbid": 211, "correlation": 278535
    }
  },
  {
    "ph": "s", "id": 278535, "pid": 76337, "tid": -914061504, "ts": 1716454225839112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225893884, "dur": 1,
    "args": {
      "External id": 278537,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278537,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 278537, "pid": 5, "tid": 7, "ts": 1716454225893884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839123, "dur": 7,
    "args": {
      "External id": 278537,
      "cbid": 51, "correlation": 278537
    }
  },
  {
    "ph": "s", "id": 278537, "pid": 76337, "tid": -914061504, "ts": 1716454225839123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225893888, "dur": 250,
    "args": {
      "External id": 278538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278538,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278538, "pid": 5, "tid": 7, "ts": 1716454225893888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839131, "dur": 7,
    "args": {
      "External id": 278538,
      "cbid": 211, "correlation": 278538
    }
  },
  {
    "ph": "s", "id": 278538, "pid": 76337, "tid": -914061504, "ts": 1716454225839131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225894139, "dur": 5,
    "args": {
      "External id": 278540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278540,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278540, "pid": 5, "tid": 7, "ts": 1716454225894139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839142, "dur": 5,
    "args": {
      "External id": 278540,
      "cbid": 211, "correlation": 278540
    }
  },
  {
    "ph": "s", "id": 278540, "pid": 76337, "tid": -914061504, "ts": 1716454225839142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225894145, "dur": 6,
    "args": {
      "External id": 278546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278546, "pid": 5, "tid": 7, "ts": 1716454225894145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839170, "dur": 9,
    "args": {
      "External id": 278546,
      "cbid": 211, "correlation": 278546
    }
  },
  {
    "ph": "s", "id": 278546, "pid": 76337, "tid": -914061504, "ts": 1716454225839170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839228, "dur": 0,
    "args": {
      "External id": 278556,
      "cbid": 317, "correlation": 278556
    }
  },
  {
    "ph": "f", "id": 278556, "pid": 76337, "tid": -914061504, "ts": 1716454225839228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839229, "dur": 0,
    "args": {
      "External id": 278557,
      "cbid": 203, "correlation": 278557
    }
  },
  {
    "ph": "f", "id": 278557, "pid": 76337, "tid": -914061504, "ts": 1716454225839229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839230, "dur": 0,
    "args": {
      "External id": 278558,
      "cbid": 205, "correlation": 278558
    }
  },
  {
    "ph": "f", "id": 278558, "pid": 76337, "tid": -914061504, "ts": 1716454225839230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225894153, "dur": 8,
    "args": {
      "External id": 278562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278562,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278562, "pid": 5, "tid": 7, "ts": 1716454225894153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839242, "dur": 11,
    "args": {
      "External id": 278562,
      "cbid": 211, "correlation": 278562
    }
  },
  {
    "ph": "s", "id": 278562, "pid": 76337, "tid": -914061504, "ts": 1716454225839242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225894162, "dur": 3,
    "args": {
      "External id": 278564,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278564,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5625
    }
  },
  {
    "ph": "f", "id": 278564, "pid": 5, "tid": 7, "ts": 1716454225894162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839259, "dur": 9,
    "args": {
      "External id": 278564,
      "cbid": 51, "correlation": 278564
    }
  },
  {
    "ph": "s", "id": 278564, "pid": 76337, "tid": -914061504, "ts": 1716454225839259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225894166, "dur": 93,
    "args": {
      "External id": 278565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278565,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 278565, "pid": 5, "tid": 7, "ts": 1716454225894166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839269, "dur": 6,
    "args": {
      "External id": 278565,
      "cbid": 211, "correlation": 278565
    }
  },
  {
    "ph": "s", "id": 278565, "pid": 76337, "tid": -914061504, "ts": 1716454225839269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225894260, "dur": 6,
    "args": {
      "External id": 278567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278567,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278567, "pid": 5, "tid": 7, "ts": 1716454225894260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839278, "dur": 5,
    "args": {
      "External id": 278567,
      "cbid": 211, "correlation": 278567
    }
  },
  {
    "ph": "s", "id": 278567, "pid": 76337, "tid": -914061504, "ts": 1716454225839278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225894267, "dur": 6,
    "args": {
      "External id": 278573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278573, "pid": 5, "tid": 7, "ts": 1716454225894267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839305, "dur": 8,
    "args": {
      "External id": 278573,
      "cbid": 211, "correlation": 278573
    }
  },
  {
    "ph": "s", "id": 278573, "pid": 76337, "tid": -914061504, "ts": 1716454225839305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225894274, "dur": 5,
    "args": {
      "External id": 278581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278581,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278581, "pid": 5, "tid": 7, "ts": 1716454225894274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839334, "dur": 8,
    "args": {
      "External id": 278581,
      "cbid": 211, "correlation": 278581
    }
  },
  {
    "ph": "s", "id": 278581, "pid": 76337, "tid": -914061504, "ts": 1716454225839334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225894280, "dur": 4,
    "args": {
      "External id": 278589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278589,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 278589, "pid": 5, "tid": 7, "ts": 1716454225894280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839363, "dur": 8,
    "args": {
      "External id": 278589,
      "cbid": 211, "correlation": 278589
    }
  },
  {
    "ph": "s", "id": 278589, "pid": 76337, "tid": -914061504, "ts": 1716454225839363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225894286, "dur": 14,
    "args": {
      "External id": 278600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278600,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278600, "pid": 5, "tid": 7, "ts": 1716454225894286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839442, "dur": 13,
    "args": {
      "External id": 278600,
      "cbid": 211, "correlation": 278600
    }
  },
  {
    "ph": "s", "id": 278600, "pid": 76337, "tid": -914061504, "ts": 1716454225839442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839499, "dur": 0,
    "args": {
      "External id": 278610,
      "cbid": 317, "correlation": 278610
    }
  },
  {
    "ph": "f", "id": 278610, "pid": 76337, "tid": -914061504, "ts": 1716454225839499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839500, "dur": 0,
    "args": {
      "External id": 278611,
      "cbid": 203, "correlation": 278611
    }
  },
  {
    "ph": "f", "id": 278611, "pid": 76337, "tid": -914061504, "ts": 1716454225839500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839500, "dur": 0,
    "args": {
      "External id": 278612,
      "cbid": 205, "correlation": 278612
    }
  },
  {
    "ph": "f", "id": 278612, "pid": 76337, "tid": -914061504, "ts": 1716454225839500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225894301, "dur": 8,
    "args": {
      "External id": 278616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278616,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278616, "pid": 5, "tid": 7, "ts": 1716454225894301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839513, "dur": 11,
    "args": {
      "External id": 278616,
      "cbid": 211, "correlation": 278616
    }
  },
  {
    "ph": "s", "id": 278616, "pid": 76337, "tid": -914061504, "ts": 1716454225839513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225894311, "dur": 157,
    "args": {
      "External id": 278618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278618,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278618, "pid": 5, "tid": 7, "ts": 1716454225894311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839527, "dur": 5,
    "args": {
      "External id": 278618,
      "cbid": 211, "correlation": 278618
    }
  },
  {
    "ph": "s", "id": 278618, "pid": 76337, "tid": -914061504, "ts": 1716454225839527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225894470, "dur": 1,
    "args": {
      "External id": 278620,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278620,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 278620, "pid": 5, "tid": 7, "ts": 1716454225894470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839538, "dur": 6,
    "args": {
      "External id": 278620,
      "cbid": 51, "correlation": 278620
    }
  },
  {
    "ph": "s", "id": 278620, "pid": 76337, "tid": -914061504, "ts": 1716454225839538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225894474, "dur": 634,
    "args": {
      "External id": 278621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278621,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278621, "pid": 5, "tid": 7, "ts": 1716454225894474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839545, "dur": 6,
    "args": {
      "External id": 278621,
      "cbid": 211, "correlation": 278621
    }
  },
  {
    "ph": "s", "id": 278621, "pid": 76337, "tid": -914061504, "ts": 1716454225839545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225895109, "dur": 12,
    "args": {
      "External id": 278623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278623,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278623, "pid": 5, "tid": 7, "ts": 1716454225895109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839555, "dur": 5,
    "args": {
      "External id": 278623,
      "cbid": 211, "correlation": 278623
    }
  },
  {
    "ph": "s", "id": 278623, "pid": 76337, "tid": -914061504, "ts": 1716454225839555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225895122, "dur": 14,
    "args": {
      "External id": 278629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278629, "pid": 5, "tid": 7, "ts": 1716454225895122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839583, "dur": 9,
    "args": {
      "External id": 278629,
      "cbid": 211, "correlation": 278629
    }
  },
  {
    "ph": "s", "id": 278629, "pid": 76337, "tid": -914061504, "ts": 1716454225839583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225895138, "dur": 29,
    "args": {
      "External id": 278638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278638,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278638, "pid": 5, "tid": 7, "ts": 1716454225895138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839673, "dur": 13,
    "args": {
      "External id": 278638,
      "cbid": 211, "correlation": 278638
    }
  },
  {
    "ph": "s", "id": 278638, "pid": 76337, "tid": -914061504, "ts": 1716454225839673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225895168, "dur": 30,
    "args": {
      "External id": 278658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278658,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278658, "pid": 5, "tid": 7, "ts": 1716454225895168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839742, "dur": 11,
    "args": {
      "External id": 278658,
      "cbid": 211, "correlation": 278658
    }
  },
  {
    "ph": "s", "id": 278658, "pid": 76337, "tid": -914061504, "ts": 1716454225839742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225895199, "dur": 4,
    "args": {
      "External id": 278670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278670,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278670, "pid": 5, "tid": 7, "ts": 1716454225895199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839762, "dur": 6,
    "args": {
      "External id": 278670,
      "cbid": 211, "correlation": 278670
    }
  },
  {
    "ph": "s", "id": 278670, "pid": 76337, "tid": -914061504, "ts": 1716454225839762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225895204, "dur": 30,
    "args": {
      "External id": 278673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278673,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278673, "pid": 5, "tid": 7, "ts": 1716454225895204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839780, "dur": 6,
    "args": {
      "External id": 278673,
      "cbid": 211, "correlation": 278673
    }
  },
  {
    "ph": "s", "id": 278673, "pid": 76337, "tid": -914061504, "ts": 1716454225839780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225895235, "dur": 21,
    "args": {
      "External id": 278682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278682,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278682, "pid": 5, "tid": 7, "ts": 1716454225895235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839818, "dur": 10,
    "args": {
      "External id": 278682,
      "cbid": 211, "correlation": 278682
    }
  },
  {
    "ph": "s", "id": 278682, "pid": 76337, "tid": -914061504, "ts": 1716454225839818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839870, "dur": 0,
    "args": {
      "External id": 278692,
      "cbid": 317, "correlation": 278692
    }
  },
  {
    "ph": "f", "id": 278692, "pid": 76337, "tid": -914061504, "ts": 1716454225839870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839871, "dur": 0,
    "args": {
      "External id": 278693,
      "cbid": 203, "correlation": 278693
    }
  },
  {
    "ph": "f", "id": 278693, "pid": 76337, "tid": -914061504, "ts": 1716454225839871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839872, "dur": 0,
    "args": {
      "External id": 278694,
      "cbid": 205, "correlation": 278694
    }
  },
  {
    "ph": "f", "id": 278694, "pid": 76337, "tid": -914061504, "ts": 1716454225839872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225895257, "dur": 22,
    "args": {
      "External id": 278698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278698,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278698, "pid": 5, "tid": 7, "ts": 1716454225895257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839886, "dur": 12,
    "args": {
      "External id": 278698,
      "cbid": 211, "correlation": 278698
    }
  },
  {
    "ph": "s", "id": 278698, "pid": 76337, "tid": -914061504, "ts": 1716454225839886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225895280, "dur": 304,
    "args": {
      "External id": 278700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278700, "pid": 5, "tid": 7, "ts": 1716454225895280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839901, "dur": 5,
    "args": {
      "External id": 278700,
      "cbid": 211, "correlation": 278700
    }
  },
  {
    "ph": "s", "id": 278700, "pid": 76337, "tid": -914061504, "ts": 1716454225839901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225895586, "dur": 1,
    "args": {
      "External id": 278702,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278702,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 278702, "pid": 5, "tid": 7, "ts": 1716454225895586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839912, "dur": 6,
    "args": {
      "External id": 278702,
      "cbid": 51, "correlation": 278702
    }
  },
  {
    "ph": "s", "id": 278702, "pid": 76337, "tid": -914061504, "ts": 1716454225839912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225895590, "dur": 1216,
    "args": {
      "External id": 278703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278703,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278703, "pid": 5, "tid": 7, "ts": 1716454225895590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839919, "dur": 6,
    "args": {
      "External id": 278703,
      "cbid": 211, "correlation": 278703
    }
  },
  {
    "ph": "s", "id": 278703, "pid": 76337, "tid": -914061504, "ts": 1716454225839919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225896807, "dur": 12,
    "args": {
      "External id": 278705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278705,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278705, "pid": 5, "tid": 7, "ts": 1716454225896807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839930, "dur": 5,
    "args": {
      "External id": 278705,
      "cbid": 211, "correlation": 278705
    }
  },
  {
    "ph": "s", "id": 278705, "pid": 76337, "tid": -914061504, "ts": 1716454225839930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225896820, "dur": 14,
    "args": {
      "External id": 278711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278711, "pid": 5, "tid": 7, "ts": 1716454225896820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225839959, "dur": 8,
    "args": {
      "External id": 278711,
      "cbid": 211, "correlation": 278711
    }
  },
  {
    "ph": "s", "id": 278711, "pid": 76337, "tid": -914061504, "ts": 1716454225839959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225896836, "dur": 4,
    "args": {
      "External id": 278719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278719,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 278719, "pid": 5, "tid": 7, "ts": 1716454225896836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840011, "dur": 10,
    "args": {
      "External id": 278719,
      "cbid": 211, "correlation": 278719
    }
  },
  {
    "ph": "s", "id": 278719, "pid": 76337, "tid": -914061504, "ts": 1716454225840011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840075, "dur": 1,
    "args": {
      "External id": 278735,
      "cbid": 251, "correlation": 278735
    }
  },
  {
    "ph": "f", "id": 278735, "pid": 76337, "tid": -914061504, "ts": 1716454225840075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840080, "dur": 0,
    "args": {
      "External id": 278737,
      "cbid": 251, "correlation": 278737
    }
  },
  {
    "ph": "f", "id": 278737, "pid": 76337, "tid": -914061504, "ts": 1716454225840080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225896841, "dur": 12,
    "args": {
      "External id": 278738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278738,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278738, "pid": 5, "tid": 7, "ts": 1716454225896841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840082, "dur": 12,
    "args": {
      "External id": 278738,
      "cbid": 211, "correlation": 278738
    }
  },
  {
    "ph": "s", "id": 278738, "pid": 76337, "tid": -914061504, "ts": 1716454225840082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225896855, "dur": 5,
    "args": {
      "External id": 278740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278740,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278740, "pid": 5, "tid": 7, "ts": 1716454225896855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840095, "dur": 6,
    "args": {
      "External id": 278740,
      "cbid": 211, "correlation": 278740
    }
  },
  {
    "ph": "s", "id": 278740, "pid": 76337, "tid": -914061504, "ts": 1716454225840095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225896861, "dur": 17,
    "args": {
      "External id": 278750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278750, "pid": 5, "tid": 7, "ts": 1716454225896861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840152, "dur": 12,
    "args": {
      "External id": 278750,
      "cbid": 211, "correlation": 278750
    }
  },
  {
    "ph": "s", "id": 278750, "pid": 76337, "tid": -914061504, "ts": 1716454225840152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225896879, "dur": 17,
    "args": {
      "External id": 278770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278770,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278770, "pid": 5, "tid": 7, "ts": 1716454225896879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840218, "dur": 11,
    "args": {
      "External id": 278770,
      "cbid": 211, "correlation": 278770
    }
  },
  {
    "ph": "s", "id": 278770, "pid": 76337, "tid": -914061504, "ts": 1716454225840218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225896897, "dur": 5,
    "args": {
      "External id": 278782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278782,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 278782, "pid": 5, "tid": 7, "ts": 1716454225896897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840238, "dur": 6,
    "args": {
      "External id": 278782,
      "cbid": 211, "correlation": 278782
    }
  },
  {
    "ph": "s", "id": 278782, "pid": 76337, "tid": -914061504, "ts": 1716454225840238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225896903, "dur": 16,
    "args": {
      "External id": 278785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278785, "pid": 5, "tid": 7, "ts": 1716454225896903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840257, "dur": 7,
    "args": {
      "External id": 278785,
      "cbid": 211, "correlation": 278785
    }
  },
  {
    "ph": "s", "id": 278785, "pid": 76337, "tid": -914061504, "ts": 1716454225840257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225896921, "dur": 12,
    "args": {
      "External id": 278794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278794,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278794, "pid": 5, "tid": 7, "ts": 1716454225896921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840297, "dur": 9,
    "args": {
      "External id": 278794,
      "cbid": 211, "correlation": 278794
    }
  },
  {
    "ph": "s", "id": 278794, "pid": 76337, "tid": -914061504, "ts": 1716454225840297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840359, "dur": 0,
    "args": {
      "External id": 278804,
      "cbid": 317, "correlation": 278804
    }
  },
  {
    "ph": "f", "id": 278804, "pid": 76337, "tid": -914061504, "ts": 1716454225840359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840360, "dur": 0,
    "args": {
      "External id": 278805,
      "cbid": 203, "correlation": 278805
    }
  },
  {
    "ph": "f", "id": 278805, "pid": 76337, "tid": -914061504, "ts": 1716454225840360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840361, "dur": 0,
    "args": {
      "External id": 278806,
      "cbid": 205, "correlation": 278806
    }
  },
  {
    "ph": "f", "id": 278806, "pid": 76337, "tid": -914061504, "ts": 1716454225840361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225896934, "dur": 11,
    "args": {
      "External id": 278810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278810, "pid": 5, "tid": 7, "ts": 1716454225896934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840375, "dur": 12,
    "args": {
      "External id": 278810,
      "cbid": 211, "correlation": 278810
    }
  },
  {
    "ph": "s", "id": 278810, "pid": 76337, "tid": -914061504, "ts": 1716454225840375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225896946, "dur": 156,
    "args": {
      "External id": 278812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278812, "pid": 5, "tid": 7, "ts": 1716454225896946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840389, "dur": 5,
    "args": {
      "External id": 278812,
      "cbid": 211, "correlation": 278812
    }
  },
  {
    "ph": "s", "id": 278812, "pid": 76337, "tid": -914061504, "ts": 1716454225840389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225897104, "dur": 1,
    "args": {
      "External id": 278814,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 278814,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 278814, "pid": 5, "tid": 7, "ts": 1716454225897104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840400, "dur": 6,
    "args": {
      "External id": 278814,
      "cbid": 51, "correlation": 278814
    }
  },
  {
    "ph": "s", "id": 278814, "pid": 76337, "tid": -914061504, "ts": 1716454225840400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225897107, "dur": 635,
    "args": {
      "External id": 278815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278815,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278815, "pid": 5, "tid": 7, "ts": 1716454225897107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840407, "dur": 6,
    "args": {
      "External id": 278815,
      "cbid": 211, "correlation": 278815
    }
  },
  {
    "ph": "s", "id": 278815, "pid": 76337, "tid": -914061504, "ts": 1716454225840407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225897744, "dur": 13,
    "args": {
      "External id": 278817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278817,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278817, "pid": 5, "tid": 7, "ts": 1716454225897744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840417, "dur": 5,
    "args": {
      "External id": 278817,
      "cbid": 211, "correlation": 278817
    }
  },
  {
    "ph": "s", "id": 278817, "pid": 76337, "tid": -914061504, "ts": 1716454225840417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225897758, "dur": 14,
    "args": {
      "External id": 278823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278823,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278823, "pid": 5, "tid": 7, "ts": 1716454225897758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840445, "dur": 9,
    "args": {
      "External id": 278823,
      "cbid": 211, "correlation": 278823
    }
  },
  {
    "ph": "s", "id": 278823, "pid": 76337, "tid": -914061504, "ts": 1716454225840445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840505, "dur": 0,
    "args": {
      "External id": 278833,
      "cbid": 317, "correlation": 278833
    }
  },
  {
    "ph": "f", "id": 278833, "pid": 76337, "tid": -914061504, "ts": 1716454225840505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840506, "dur": 0,
    "args": {
      "External id": 278834,
      "cbid": 203, "correlation": 278834
    }
  },
  {
    "ph": "f", "id": 278834, "pid": 76337, "tid": -914061504, "ts": 1716454225840506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840507, "dur": 0,
    "args": {
      "External id": 278835,
      "cbid": 205, "correlation": 278835
    }
  },
  {
    "ph": "f", "id": 278835, "pid": 76337, "tid": -914061504, "ts": 1716454225840507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225897773, "dur": 21,
    "args": {
      "External id": 278839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278839,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278839, "pid": 5, "tid": 7, "ts": 1716454225897773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840520, "dur": 11,
    "args": {
      "External id": 278839,
      "cbid": 211, "correlation": 278839
    }
  },
  {
    "ph": "s", "id": 278839, "pid": 76337, "tid": -914061504, "ts": 1716454225840520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225897795, "dur": 4,
    "args": {
      "External id": 278841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278841,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 278841, "pid": 5, "tid": 7, "ts": 1716454225897795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840537, "dur": 6,
    "args": {
      "External id": 278841,
      "cbid": 211, "correlation": 278841
    }
  },
  {
    "ph": "s", "id": 278841, "pid": 76337, "tid": -914061504, "ts": 1716454225840537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840547, "dur": 0,
    "args": {
      "External id": 278842,
      "cbid": 51, "correlation": 278842
    }
  },
  {
    "ph": "s", "id": 278842, "pid": 76337, "tid": -914061504, "ts": 1716454225840547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225897800, "dur": 166,
    "args": {
      "External id": 278843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278843,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 278843, "pid": 5, "tid": 7, "ts": 1716454225897800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840548, "dur": 6,
    "args": {
      "External id": 278843,
      "cbid": 211, "correlation": 278843
    }
  },
  {
    "ph": "s", "id": 278843, "pid": 76337, "tid": -914061504, "ts": 1716454225840548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225897968, "dur": 15,
    "args": {
      "External id": 278848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278848, "pid": 5, "tid": 7, "ts": 1716454225897968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840574, "dur": 9,
    "args": {
      "External id": 278848,
      "cbid": 211, "correlation": 278848
    }
  },
  {
    "ph": "s", "id": 278848, "pid": 76337, "tid": -914061504, "ts": 1716454225840574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225897984, "dur": 12,
    "args": {
      "External id": 278856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278856,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278856, "pid": 5, "tid": 7, "ts": 1716454225897984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840603, "dur": 7,
    "args": {
      "External id": 278856,
      "cbid": 211, "correlation": 278856
    }
  },
  {
    "ph": "s", "id": 278856, "pid": 76337, "tid": -914061504, "ts": 1716454225840603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225897997, "dur": 10,
    "args": {
      "External id": 278864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278864,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278864, "pid": 5, "tid": 7, "ts": 1716454225897997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840631, "dur": 8,
    "args": {
      "External id": 278864,
      "cbid": 211, "correlation": 278864
    }
  },
  {
    "ph": "s", "id": 278864, "pid": 76337, "tid": -914061504, "ts": 1716454225840631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225898009, "dur": 18,
    "args": {
      "External id": 278884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278884,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 278884, "pid": 5, "tid": 7, "ts": 1716454225898009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840713, "dur": 12,
    "args": {
      "External id": 278884,
      "cbid": 211, "correlation": 278884
    }
  },
  {
    "ph": "s", "id": 278884, "pid": 76337, "tid": -914061504, "ts": 1716454225840713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225898028, "dur": 4,
    "args": {
      "External id": 278896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278896,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 278896, "pid": 5, "tid": 7, "ts": 1716454225898028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840735, "dur": 6,
    "args": {
      "External id": 278896,
      "cbid": 211, "correlation": 278896
    }
  },
  {
    "ph": "s", "id": 278896, "pid": 76337, "tid": -914061504, "ts": 1716454225840735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225898033, "dur": 17,
    "args": {
      "External id": 278899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278899, "pid": 5, "tid": 7, "ts": 1716454225898033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840753, "dur": 7,
    "args": {
      "External id": 278899,
      "cbid": 211, "correlation": 278899
    }
  },
  {
    "ph": "s", "id": 278899, "pid": 76337, "tid": -914061504, "ts": 1716454225840753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840811, "dur": 0,
    "args": {
      "External id": 278910,
      "cbid": 317, "correlation": 278910
    }
  },
  {
    "ph": "f", "id": 278910, "pid": 76337, "tid": -914061504, "ts": 1716454225840811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840811, "dur": 0,
    "args": {
      "External id": 278911,
      "cbid": 203, "correlation": 278911
    }
  },
  {
    "ph": "f", "id": 278911, "pid": 76337, "tid": -914061504, "ts": 1716454225840811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840812, "dur": 0,
    "args": {
      "External id": 278912,
      "cbid": 205, "correlation": 278912
    }
  },
  {
    "ph": "f", "id": 278912, "pid": 76337, "tid": -914061504, "ts": 1716454225840812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225898052, "dur": 12,
    "args": {
      "External id": 278916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278916,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278916, "pid": 5, "tid": 7, "ts": 1716454225898052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840825, "dur": 11,
    "args": {
      "External id": 278916,
      "cbid": 211, "correlation": 278916
    }
  },
  {
    "ph": "s", "id": 278916, "pid": 76337, "tid": -914061504, "ts": 1716454225840825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225898065, "dur": 3,
    "args": {
      "External id": 278918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278918,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 278918, "pid": 5, "tid": 7, "ts": 1716454225898065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840841, "dur": 6,
    "args": {
      "External id": 278918,
      "cbid": 211, "correlation": 278918
    }
  },
  {
    "ph": "s", "id": 278918, "pid": 76337, "tid": -914061504, "ts": 1716454225840841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840850, "dur": 0,
    "args": {
      "External id": 278919,
      "cbid": 51, "correlation": 278919
    }
  },
  {
    "ph": "s", "id": 278919, "pid": 76337, "tid": -914061504, "ts": 1716454225840850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225898070, "dur": 86,
    "args": {
      "External id": 278920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278920,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 278920, "pid": 5, "tid": 7, "ts": 1716454225898070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840851, "dur": 5,
    "args": {
      "External id": 278920,
      "cbid": 211, "correlation": 278920
    }
  },
  {
    "ph": "s", "id": 278920, "pid": 76337, "tid": -914061504, "ts": 1716454225840851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225898157, "dur": 15,
    "args": {
      "External id": 278925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278925, "pid": 5, "tid": 7, "ts": 1716454225898157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840878, "dur": 8,
    "args": {
      "External id": 278925,
      "cbid": 211, "correlation": 278925
    }
  },
  {
    "ph": "s", "id": 278925, "pid": 76337, "tid": -914061504, "ts": 1716454225840878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225898173, "dur": 80,
    "args": {
      "External id": 278934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278934,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278934, "pid": 5, "tid": 7, "ts": 1716454225898173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225840959, "dur": 14,
    "args": {
      "External id": 278934,
      "cbid": 211, "correlation": 278934
    }
  },
  {
    "ph": "s", "id": 278934, "pid": 76337, "tid": -914061504, "ts": 1716454225840959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225898255, "dur": 29,
    "args": {
      "External id": 278956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278956,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 278956, "pid": 5, "tid": 7, "ts": 1716454225898255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841030, "dur": 10,
    "args": {
      "External id": 278956,
      "cbid": 211, "correlation": 278956
    }
  },
  {
    "ph": "s", "id": 278956, "pid": 76337, "tid": -914061504, "ts": 1716454225841030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841126, "dur": 2,
    "args": {
      "External id": 278967,
      "cbid": 251, "correlation": 278967
    }
  },
  {
    "ph": "f", "id": 278967, "pid": 76337, "tid": -914061504, "ts": 1716454225841126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225898285, "dur": 139,
    "args": {
      "External id": 278968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278968,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278968, "pid": 5, "tid": 7, "ts": 1716454225898285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841132, "dur": 14,
    "args": {
      "External id": 278968,
      "cbid": 211, "correlation": 278968
    }
  },
  {
    "ph": "s", "id": 278968, "pid": 76337, "tid": -914061504, "ts": 1716454225841132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841203, "dur": 1,
    "args": {
      "External id": 278979,
      "cbid": 251, "correlation": 278979
    }
  },
  {
    "ph": "f", "id": 278979, "pid": 76337, "tid": -914061504, "ts": 1716454225841203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225898425, "dur": 153,
    "args": {
      "External id": 278980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278980,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278980, "pid": 5, "tid": 7, "ts": 1716454225898425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841207, "dur": 11,
    "args": {
      "External id": 278980,
      "cbid": 211, "correlation": 278980
    }
  },
  {
    "ph": "s", "id": 278980, "pid": 76337, "tid": -914061504, "ts": 1716454225841207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841272, "dur": 1,
    "args": {
      "External id": 278991,
      "cbid": 251, "correlation": 278991
    }
  },
  {
    "ph": "f", "id": 278991, "pid": 76337, "tid": -914061504, "ts": 1716454225841272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225898580, "dur": 134,
    "args": {
      "External id": 278992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 278992,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 278992, "pid": 5, "tid": 7, "ts": 1716454225898580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841276, "dur": 11,
    "args": {
      "External id": 278992,
      "cbid": 211, "correlation": 278992
    }
  },
  {
    "ph": "s", "id": 278992, "pid": 76337, "tid": -914061504, "ts": 1716454225841276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225898715, "dur": 324,
    "args": {
      "External id": 279017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279017,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279017, "pid": 5, "tid": 7, "ts": 1716454225898715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841363, "dur": 13,
    "args": {
      "External id": 279017,
      "cbid": 211, "correlation": 279017
    }
  },
  {
    "ph": "s", "id": 279017, "pid": 76337, "tid": -914061504, "ts": 1716454225841363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841464, "dur": 1,
    "args": {
      "External id": 279035,
      "cbid": 251, "correlation": 279035
    }
  },
  {
    "ph": "f", "id": 279035, "pid": 76337, "tid": -914061504, "ts": 1716454225841464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225899041, "dur": 161,
    "args": {
      "External id": 279037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279037,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279037, "pid": 5, "tid": 7, "ts": 1716454225899041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841470, "dur": 13,
    "args": {
      "External id": 279037,
      "cbid": 211, "correlation": 279037
    }
  },
  {
    "ph": "s", "id": 279037, "pid": 76337, "tid": -914061504, "ts": 1716454225841470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225899203, "dur": 19,
    "args": {
      "External id": 279045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279045,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279045, "pid": 5, "tid": 7, "ts": 1716454225899203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841539, "dur": 13,
    "args": {
      "External id": 279045,
      "cbid": 211, "correlation": 279045
    }
  },
  {
    "ph": "s", "id": 279045, "pid": 76337, "tid": -914061504, "ts": 1716454225841539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225899224, "dur": 28,
    "args": {
      "External id": 279053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279053,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279053, "pid": 5, "tid": 7, "ts": 1716454225899224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841579, "dur": 8,
    "args": {
      "External id": 279053,
      "cbid": 211, "correlation": 279053
    }
  },
  {
    "ph": "s", "id": 279053, "pid": 76337, "tid": -914061504, "ts": 1716454225841579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225899253, "dur": 19,
    "args": {
      "External id": 279064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279064, "pid": 5, "tid": 7, "ts": 1716454225899253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841653, "dur": 13,
    "args": {
      "External id": 279064,
      "cbid": 211, "correlation": 279064
    }
  },
  {
    "ph": "s", "id": 279064, "pid": 76337, "tid": -914061504, "ts": 1716454225841653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225899273, "dur": 15,
    "args": {
      "External id": 279086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279086,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279086, "pid": 5, "tid": 7, "ts": 1716454225899273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841684, "dur": 7,
    "args": {
      "External id": 279086,
      "cbid": 211, "correlation": 279086
    }
  },
  {
    "ph": "s", "id": 279086, "pid": 76337, "tid": -914061504, "ts": 1716454225841684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841771, "dur": 2,
    "args": {
      "External id": 279097,
      "cbid": 251, "correlation": 279097
    }
  },
  {
    "ph": "f", "id": 279097, "pid": 76337, "tid": -914061504, "ts": 1716454225841771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225899289, "dur": 86,
    "args": {
      "External id": 279098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279098,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279098, "pid": 5, "tid": 7, "ts": 1716454225899289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841776, "dur": 13,
    "args": {
      "External id": 279098,
      "cbid": 211, "correlation": 279098
    }
  },
  {
    "ph": "s", "id": 279098, "pid": 76337, "tid": -914061504, "ts": 1716454225841776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841848, "dur": 1,
    "args": {
      "External id": 279109,
      "cbid": 251, "correlation": 279109
    }
  },
  {
    "ph": "f", "id": 279109, "pid": 76337, "tid": -914061504, "ts": 1716454225841848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841852, "dur": 0,
    "args": {
      "External id": 279110,
      "cbid": 251, "correlation": 279110
    }
  },
  {
    "ph": "f", "id": 279110, "pid": 76337, "tid": -914061504, "ts": 1716454225841852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225899376, "dur": 12,
    "args": {
      "External id": 279111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279111,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279111, "pid": 5, "tid": 7, "ts": 1716454225899376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841853, "dur": 11,
    "args": {
      "External id": 279111,
      "cbid": 211, "correlation": 279111
    }
  },
  {
    "ph": "s", "id": 279111, "pid": 76337, "tid": -914061504, "ts": 1716454225841853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225899390, "dur": 5,
    "args": {
      "External id": 279113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279113,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279113, "pid": 5, "tid": 7, "ts": 1716454225899390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841866, "dur": 6,
    "args": {
      "External id": 279113,
      "cbid": 211, "correlation": 279113
    }
  },
  {
    "ph": "s", "id": 279113, "pid": 76337, "tid": -914061504, "ts": 1716454225841866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841922, "dur": 1,
    "args": {
      "External id": 279124,
      "cbid": 251, "correlation": 279124
    }
  },
  {
    "ph": "f", "id": 279124, "pid": 76337, "tid": -914061504, "ts": 1716454225841922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841926, "dur": 0,
    "args": {
      "External id": 279125,
      "cbid": 251, "correlation": 279125
    }
  },
  {
    "ph": "f", "id": 279125, "pid": 76337, "tid": -914061504, "ts": 1716454225841926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225899397, "dur": 8,
    "args": {
      "External id": 279126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279126,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279126, "pid": 5, "tid": 7, "ts": 1716454225899397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841927, "dur": 11,
    "args": {
      "External id": 279126,
      "cbid": 211, "correlation": 279126
    }
  },
  {
    "ph": "s", "id": 279126, "pid": 76337, "tid": -914061504, "ts": 1716454225841927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225899406, "dur": 3,
    "args": {
      "External id": 279128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279128,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279128, "pid": 5, "tid": 7, "ts": 1716454225899406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225841940, "dur": 6,
    "args": {
      "External id": 279128,
      "cbid": 211, "correlation": 279128
    }
  },
  {
    "ph": "s", "id": 279128, "pid": 76337, "tid": -914061504, "ts": 1716454225841940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225899411, "dur": 53,
    "args": {
      "External id": 279153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279153,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279153, "pid": 5, "tid": 7, "ts": 1716454225899411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842027, "dur": 12,
    "args": {
      "External id": 279153,
      "cbid": 211, "correlation": 279153
    }
  },
  {
    "ph": "s", "id": 279153, "pid": 76337, "tid": -914061504, "ts": 1716454225842027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842126, "dur": 2,
    "args": {
      "External id": 279171,
      "cbid": 251, "correlation": 279171
    }
  },
  {
    "ph": "f", "id": 279171, "pid": 76337, "tid": -914061504, "ts": 1716454225842126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225899465, "dur": 87,
    "args": {
      "External id": 279173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279173,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279173, "pid": 5, "tid": 7, "ts": 1716454225899465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842132, "dur": 14,
    "args": {
      "External id": 279173,
      "cbid": 211, "correlation": 279173
    }
  },
  {
    "ph": "s", "id": 279173, "pid": 76337, "tid": -914061504, "ts": 1716454225842132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225899553, "dur": 9,
    "args": {
      "External id": 279181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279181,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279181, "pid": 5, "tid": 7, "ts": 1716454225899553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842201, "dur": 13,
    "args": {
      "External id": 279181,
      "cbid": 211, "correlation": 279181
    }
  },
  {
    "ph": "s", "id": 279181, "pid": 76337, "tid": -914061504, "ts": 1716454225842201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225899563, "dur": 21,
    "args": {
      "External id": 279189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279189, "pid": 5, "tid": 7, "ts": 1716454225899563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842243, "dur": 10,
    "args": {
      "External id": 279189,
      "cbid": 211, "correlation": 279189
    }
  },
  {
    "ph": "s", "id": 279189, "pid": 76337, "tid": -914061504, "ts": 1716454225842243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225899586, "dur": 17,
    "args": {
      "External id": 279211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279211,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279211, "pid": 5, "tid": 7, "ts": 1716454225899586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842294, "dur": 10,
    "args": {
      "External id": 279211,
      "cbid": 211, "correlation": 279211
    }
  },
  {
    "ph": "s", "id": 279211, "pid": 76337, "tid": -914061504, "ts": 1716454225842294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842382, "dur": 1,
    "args": {
      "External id": 279227,
      "cbid": 251, "correlation": 279227
    }
  },
  {
    "ph": "f", "id": 279227, "pid": 76337, "tid": -914061504, "ts": 1716454225842382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842387, "dur": 0,
    "args": {
      "External id": 279229,
      "cbid": 251, "correlation": 279229
    }
  },
  {
    "ph": "f", "id": 279229, "pid": 76337, "tid": -914061504, "ts": 1716454225842387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225899604, "dur": 486,
    "args": {
      "External id": 279230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279230,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279230, "pid": 5, "tid": 7, "ts": 1716454225899604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842389, "dur": 12,
    "args": {
      "External id": 279230,
      "cbid": 211, "correlation": 279230
    }
  },
  {
    "ph": "s", "id": 279230, "pid": 76337, "tid": -914061504, "ts": 1716454225842389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225900092, "dur": 64,
    "args": {
      "External id": 279238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279238, "pid": 5, "tid": 7, "ts": 1716454225900092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842456, "dur": 13,
    "args": {
      "External id": 279238,
      "cbid": 211, "correlation": 279238
    }
  },
  {
    "ph": "s", "id": 279238, "pid": 76337, "tid": -914061504, "ts": 1716454225842456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225900157, "dur": 66,
    "args": {
      "External id": 279246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279246, "pid": 5, "tid": 7, "ts": 1716454225900157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842487, "dur": 8,
    "args": {
      "External id": 279246,
      "cbid": 211, "correlation": 279246
    }
  },
  {
    "ph": "s", "id": 279246, "pid": 76337, "tid": -914061504, "ts": 1716454225842487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842567, "dur": 1,
    "args": {
      "External id": 279262,
      "cbid": 251, "correlation": 279262
    }
  },
  {
    "ph": "f", "id": 279262, "pid": 76337, "tid": -914061504, "ts": 1716454225842567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225900225, "dur": 1,
    "args": {
      "External id": 279264,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 279264,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 279264, "pid": 5, "tid": 7, "ts": 1716454225900225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842572, "dur": 9,
    "args": {
      "External id": 279264,
      "cbid": 51, "correlation": 279264
    }
  },
  {
    "ph": "s", "id": 279264, "pid": 76337, "tid": -914061504, "ts": 1716454225842572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225900229, "dur": 263,
    "args": {
      "External id": 279265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279265,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279265, "pid": 5, "tid": 7, "ts": 1716454225900229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842582, "dur": 11,
    "args": {
      "External id": 279265,
      "cbid": 211, "correlation": 279265
    }
  },
  {
    "ph": "s", "id": 279265, "pid": 76337, "tid": -914061504, "ts": 1716454225842582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225900493, "dur": 14,
    "args": {
      "External id": 279273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279273,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279273, "pid": 5, "tid": 7, "ts": 1716454225900493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842625, "dur": 10,
    "args": {
      "External id": 279273,
      "cbid": 211, "correlation": 279273
    }
  },
  {
    "ph": "s", "id": 279273, "pid": 76337, "tid": -914061504, "ts": 1716454225842625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225900509, "dur": 36,
    "args": {
      "External id": 279284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279284, "pid": 5, "tid": 7, "ts": 1716454225900509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842693, "dur": 13,
    "args": {
      "External id": 279284,
      "cbid": 211, "correlation": 279284
    }
  },
  {
    "ph": "s", "id": 279284, "pid": 76337, "tid": -914061504, "ts": 1716454225842693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842759, "dur": 0,
    "args": {
      "External id": 279296,
      "cbid": 317, "correlation": 279296
    }
  },
  {
    "ph": "f", "id": 279296, "pid": 76337, "tid": -914061504, "ts": 1716454225842759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842760, "dur": 0,
    "args": {
      "External id": 279297,
      "cbid": 203, "correlation": 279297
    }
  },
  {
    "ph": "f", "id": 279297, "pid": 76337, "tid": -914061504, "ts": 1716454225842760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842761, "dur": 0,
    "args": {
      "External id": 279298,
      "cbid": 205, "correlation": 279298
    }
  },
  {
    "ph": "f", "id": 279298, "pid": 76337, "tid": -914061504, "ts": 1716454225842761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225900546, "dur": 12,
    "args": {
      "External id": 279302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279302, "pid": 5, "tid": 7, "ts": 1716454225900546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842775, "dur": 12,
    "args": {
      "External id": 279302,
      "cbid": 211, "correlation": 279302
    }
  },
  {
    "ph": "s", "id": 279302, "pid": 76337, "tid": -914061504, "ts": 1716454225842775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225900559, "dur": 3,
    "args": {
      "External id": 279304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279304, "pid": 5, "tid": 7, "ts": 1716454225900559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842791, "dur": 6,
    "args": {
      "External id": 279304,
      "cbid": 211, "correlation": 279304
    }
  },
  {
    "ph": "s", "id": 279304, "pid": 76337, "tid": -914061504, "ts": 1716454225842791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842800, "dur": 0,
    "args": {
      "External id": 279305,
      "cbid": 51, "correlation": 279305
    }
  },
  {
    "ph": "s", "id": 279305, "pid": 76337, "tid": -914061504, "ts": 1716454225842800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225900564, "dur": 93,
    "args": {
      "External id": 279306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279306,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 279306, "pid": 5, "tid": 7, "ts": 1716454225900564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842801, "dur": 6,
    "args": {
      "External id": 279306,
      "cbid": 211, "correlation": 279306
    }
  },
  {
    "ph": "s", "id": 279306, "pid": 76337, "tid": -914061504, "ts": 1716454225842801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225900658, "dur": 15,
    "args": {
      "External id": 279311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279311, "pid": 5, "tid": 7, "ts": 1716454225900658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842829, "dur": 8,
    "args": {
      "External id": 279311,
      "cbid": 211, "correlation": 279311
    }
  },
  {
    "ph": "s", "id": 279311, "pid": 76337, "tid": -914061504, "ts": 1716454225842829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225900674, "dur": 11,
    "args": {
      "External id": 279319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279319,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279319, "pid": 5, "tid": 7, "ts": 1716454225900674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842862, "dur": 8,
    "args": {
      "External id": 279319,
      "cbid": 211, "correlation": 279319
    }
  },
  {
    "ph": "s", "id": 279319, "pid": 76337, "tid": -914061504, "ts": 1716454225842862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225900687, "dur": 28,
    "args": {
      "External id": 279328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279328,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279328, "pid": 5, "tid": 7, "ts": 1716454225900687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842900, "dur": 11,
    "args": {
      "External id": 279328,
      "cbid": 211, "correlation": 279328
    }
  },
  {
    "ph": "s", "id": 279328, "pid": 76337, "tid": -914061504, "ts": 1716454225842900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225900716, "dur": 30,
    "args": {
      "External id": 279348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279348,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 279348, "pid": 5, "tid": 7, "ts": 1716454225900716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225842973, "dur": 21,
    "args": {
      "External id": 279348,
      "cbid": 211, "correlation": 279348
    }
  },
  {
    "ph": "s", "id": 279348, "pid": 76337, "tid": -914061504, "ts": 1716454225842973,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225900747, "dur": 5,
    "args": {
      "External id": 279360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279360,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279360, "pid": 5, "tid": 7, "ts": 1716454225900747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843004, "dur": 7,
    "args": {
      "External id": 279360,
      "cbid": 211, "correlation": 279360
    }
  },
  {
    "ph": "s", "id": 279360, "pid": 76337, "tid": -914061504, "ts": 1716454225843004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225900753, "dur": 31,
    "args": {
      "External id": 279363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279363,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279363, "pid": 5, "tid": 7, "ts": 1716454225900753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843023, "dur": 7,
    "args": {
      "External id": 279363,
      "cbid": 211, "correlation": 279363
    }
  },
  {
    "ph": "s", "id": 279363, "pid": 76337, "tid": -914061504, "ts": 1716454225843023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225900786, "dur": 21,
    "args": {
      "External id": 279372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279372,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279372, "pid": 5, "tid": 7, "ts": 1716454225900786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843063, "dur": 10,
    "args": {
      "External id": 279372,
      "cbid": 211, "correlation": 279372
    }
  },
  {
    "ph": "s", "id": 279372, "pid": 76337, "tid": -914061504, "ts": 1716454225843063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843115, "dur": 0,
    "args": {
      "External id": 279382,
      "cbid": 317, "correlation": 279382
    }
  },
  {
    "ph": "f", "id": 279382, "pid": 76337, "tid": -914061504, "ts": 1716454225843115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843116, "dur": 0,
    "args": {
      "External id": 279383,
      "cbid": 203, "correlation": 279383
    }
  },
  {
    "ph": "f", "id": 279383, "pid": 76337, "tid": -914061504, "ts": 1716454225843116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843117, "dur": 0,
    "args": {
      "External id": 279384,
      "cbid": 205, "correlation": 279384
    }
  },
  {
    "ph": "f", "id": 279384, "pid": 76337, "tid": -914061504, "ts": 1716454225843117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225900808, "dur": 21,
    "args": {
      "External id": 279388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279388,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279388, "pid": 5, "tid": 7, "ts": 1716454225900808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843130, "dur": 11,
    "args": {
      "External id": 279388,
      "cbid": 211, "correlation": 279388
    }
  },
  {
    "ph": "s", "id": 279388, "pid": 76337, "tid": -914061504, "ts": 1716454225843130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225900830, "dur": 304,
    "args": {
      "External id": 279390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279390,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279390, "pid": 5, "tid": 7, "ts": 1716454225900830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843144, "dur": 5,
    "args": {
      "External id": 279390,
      "cbid": 211, "correlation": 279390
    }
  },
  {
    "ph": "s", "id": 279390, "pid": 76337, "tid": -914061504, "ts": 1716454225843144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225901137, "dur": 1,
    "args": {
      "External id": 279392,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 279392,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 279392, "pid": 5, "tid": 7, "ts": 1716454225901137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843155, "dur": 7,
    "args": {
      "External id": 279392,
      "cbid": 51, "correlation": 279392
    }
  },
  {
    "ph": "s", "id": 279392, "pid": 76337, "tid": -914061504, "ts": 1716454225843155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225901140, "dur": 1228,
    "args": {
      "External id": 279393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279393,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279393, "pid": 5, "tid": 7, "ts": 1716454225901140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843163, "dur": 6,
    "args": {
      "External id": 279393,
      "cbid": 211, "correlation": 279393
    }
  },
  {
    "ph": "s", "id": 279393, "pid": 76337, "tid": -914061504, "ts": 1716454225843163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225902370, "dur": 13,
    "args": {
      "External id": 279395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279395,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279395, "pid": 5, "tid": 7, "ts": 1716454225902370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843173, "dur": 5,
    "args": {
      "External id": 279395,
      "cbid": 211, "correlation": 279395
    }
  },
  {
    "ph": "s", "id": 279395, "pid": 76337, "tid": -914061504, "ts": 1716454225843173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225902384, "dur": 14,
    "args": {
      "External id": 279401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279401, "pid": 5, "tid": 7, "ts": 1716454225902384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843201, "dur": 8,
    "args": {
      "External id": 279401,
      "cbid": 211, "correlation": 279401
    }
  },
  {
    "ph": "s", "id": 279401, "pid": 76337, "tid": -914061504, "ts": 1716454225843201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225902399, "dur": 4,
    "args": {
      "External id": 279409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279409,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 279409, "pid": 5, "tid": 7, "ts": 1716454225902399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843244, "dur": 9,
    "args": {
      "External id": 279409,
      "cbid": 211, "correlation": 279409
    }
  },
  {
    "ph": "s", "id": 279409, "pid": 76337, "tid": -914061504, "ts": 1716454225843244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843307, "dur": 1,
    "args": {
      "External id": 279425,
      "cbid": 251, "correlation": 279425
    }
  },
  {
    "ph": "f", "id": 279425, "pid": 76337, "tid": -914061504, "ts": 1716454225843307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843312, "dur": 0,
    "args": {
      "External id": 279427,
      "cbid": 251, "correlation": 279427
    }
  },
  {
    "ph": "f", "id": 279427, "pid": 76337, "tid": -914061504, "ts": 1716454225843312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225902404, "dur": 13,
    "args": {
      "External id": 279428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279428,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279428, "pid": 5, "tid": 7, "ts": 1716454225902404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843314, "dur": 12,
    "args": {
      "External id": 279428,
      "cbid": 211, "correlation": 279428
    }
  },
  {
    "ph": "s", "id": 279428, "pid": 76337, "tid": -914061504, "ts": 1716454225843314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225902418, "dur": 5,
    "args": {
      "External id": 279430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279430,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279430, "pid": 5, "tid": 7, "ts": 1716454225902418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843328, "dur": 5,
    "args": {
      "External id": 279430,
      "cbid": 211, "correlation": 279430
    }
  },
  {
    "ph": "s", "id": 279430, "pid": 76337, "tid": -914061504, "ts": 1716454225843328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225902424, "dur": 16,
    "args": {
      "External id": 279440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279440, "pid": 5, "tid": 7, "ts": 1716454225902424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843385, "dur": 12,
    "args": {
      "External id": 279440,
      "cbid": 211, "correlation": 279440
    }
  },
  {
    "ph": "s", "id": 279440, "pid": 76337, "tid": -914061504, "ts": 1716454225843385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225902442, "dur": 16,
    "args": {
      "External id": 279460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279460,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 279460, "pid": 5, "tid": 7, "ts": 1716454225902442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843451, "dur": 11,
    "args": {
      "External id": 279460,
      "cbid": 211, "correlation": 279460
    }
  },
  {
    "ph": "s", "id": 279460, "pid": 76337, "tid": -914061504, "ts": 1716454225843451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225902460, "dur": 4,
    "args": {
      "External id": 279472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279472,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 279472, "pid": 5, "tid": 7, "ts": 1716454225902460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843472, "dur": 6,
    "args": {
      "External id": 279472,
      "cbid": 211, "correlation": 279472
    }
  },
  {
    "ph": "s", "id": 279472, "pid": 76337, "tid": -914061504, "ts": 1716454225843472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225902465, "dur": 17,
    "args": {
      "External id": 279475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279475, "pid": 5, "tid": 7, "ts": 1716454225902465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843490, "dur": 6,
    "args": {
      "External id": 279475,
      "cbid": 211, "correlation": 279475
    }
  },
  {
    "ph": "s", "id": 279475, "pid": 76337, "tid": -914061504, "ts": 1716454225843490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225902483, "dur": 11,
    "args": {
      "External id": 279484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279484,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279484, "pid": 5, "tid": 7, "ts": 1716454225902483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843530, "dur": 10,
    "args": {
      "External id": 279484,
      "cbid": 211, "correlation": 279484
    }
  },
  {
    "ph": "s", "id": 279484, "pid": 76337, "tid": -914061504, "ts": 1716454225843530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843592, "dur": 0,
    "args": {
      "External id": 279494,
      "cbid": 317, "correlation": 279494
    }
  },
  {
    "ph": "f", "id": 279494, "pid": 76337, "tid": -914061504, "ts": 1716454225843592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843592, "dur": 0,
    "args": {
      "External id": 279495,
      "cbid": 203, "correlation": 279495
    }
  },
  {
    "ph": "f", "id": 279495, "pid": 76337, "tid": -914061504, "ts": 1716454225843592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843593, "dur": 0,
    "args": {
      "External id": 279496,
      "cbid": 205, "correlation": 279496
    }
  },
  {
    "ph": "f", "id": 279496, "pid": 76337, "tid": -914061504, "ts": 1716454225843593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225902495, "dur": 11,
    "args": {
      "External id": 279500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279500,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279500, "pid": 5, "tid": 7, "ts": 1716454225902495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843607, "dur": 12,
    "args": {
      "External id": 279500,
      "cbid": 211, "correlation": 279500
    }
  },
  {
    "ph": "s", "id": 279500, "pid": 76337, "tid": -914061504, "ts": 1716454225843607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225902508, "dur": 155,
    "args": {
      "External id": 279502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279502,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279502, "pid": 5, "tid": 7, "ts": 1716454225902508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843621, "dur": 5,
    "args": {
      "External id": 279502,
      "cbid": 211, "correlation": 279502
    }
  },
  {
    "ph": "s", "id": 279502, "pid": 76337, "tid": -914061504, "ts": 1716454225843621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225902665, "dur": 1,
    "args": {
      "External id": 279504,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 279504,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 279504, "pid": 5, "tid": 7, "ts": 1716454225902665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843632, "dur": 6,
    "args": {
      "External id": 279504,
      "cbid": 51, "correlation": 279504
    }
  },
  {
    "ph": "s", "id": 279504, "pid": 76337, "tid": -914061504, "ts": 1716454225843632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225902669, "dur": 634,
    "args": {
      "External id": 279505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279505,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279505, "pid": 5, "tid": 7, "ts": 1716454225902669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843640, "dur": 6,
    "args": {
      "External id": 279505,
      "cbid": 211, "correlation": 279505
    }
  },
  {
    "ph": "s", "id": 279505, "pid": 76337, "tid": -914061504, "ts": 1716454225843640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225903304, "dur": 12,
    "args": {
      "External id": 279507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279507,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279507, "pid": 5, "tid": 7, "ts": 1716454225903304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843651, "dur": 5,
    "args": {
      "External id": 279507,
      "cbid": 211, "correlation": 279507
    }
  },
  {
    "ph": "s", "id": 279507, "pid": 76337, "tid": -914061504, "ts": 1716454225843651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225903318, "dur": 14,
    "args": {
      "External id": 279513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279513,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279513, "pid": 5, "tid": 7, "ts": 1716454225903318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843678, "dur": 9,
    "args": {
      "External id": 279513,
      "cbid": 211, "correlation": 279513
    }
  },
  {
    "ph": "s", "id": 279513, "pid": 76337, "tid": -914061504, "ts": 1716454225843678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843737, "dur": 0,
    "args": {
      "External id": 279523,
      "cbid": 317, "correlation": 279523
    }
  },
  {
    "ph": "f", "id": 279523, "pid": 76337, "tid": -914061504, "ts": 1716454225843737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843738, "dur": 0,
    "args": {
      "External id": 279524,
      "cbid": 203, "correlation": 279524
    }
  },
  {
    "ph": "f", "id": 279524, "pid": 76337, "tid": -914061504, "ts": 1716454225843738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843739, "dur": 0,
    "args": {
      "External id": 279525,
      "cbid": 205, "correlation": 279525
    }
  },
  {
    "ph": "f", "id": 279525, "pid": 76337, "tid": -914061504, "ts": 1716454225843739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225903333, "dur": 21,
    "args": {
      "External id": 279529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279529,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279529, "pid": 5, "tid": 7, "ts": 1716454225903333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843751, "dur": 11,
    "args": {
      "External id": 279529,
      "cbid": 211, "correlation": 279529
    }
  },
  {
    "ph": "s", "id": 279529, "pid": 76337, "tid": -914061504, "ts": 1716454225843751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225903355, "dur": 4,
    "args": {
      "External id": 279531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279531, "pid": 5, "tid": 7, "ts": 1716454225903355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843766, "dur": 5,
    "args": {
      "External id": 279531,
      "cbid": 211, "correlation": 279531
    }
  },
  {
    "ph": "s", "id": 279531, "pid": 76337, "tid": -914061504, "ts": 1716454225843766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843774, "dur": 0,
    "args": {
      "External id": 279532,
      "cbid": 51, "correlation": 279532
    }
  },
  {
    "ph": "s", "id": 279532, "pid": 76337, "tid": -914061504, "ts": 1716454225843774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225903361, "dur": 165,
    "args": {
      "External id": 279533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279533,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 279533, "pid": 5, "tid": 7, "ts": 1716454225903361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843775, "dur": 5,
    "args": {
      "External id": 279533,
      "cbid": 211, "correlation": 279533
    }
  },
  {
    "ph": "s", "id": 279533, "pid": 76337, "tid": -914061504, "ts": 1716454225843775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225903528, "dur": 15,
    "args": {
      "External id": 279538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279538, "pid": 5, "tid": 7, "ts": 1716454225903528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843801, "dur": 8,
    "args": {
      "External id": 279538,
      "cbid": 211, "correlation": 279538
    }
  },
  {
    "ph": "s", "id": 279538, "pid": 76337, "tid": -914061504, "ts": 1716454225843801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225903544, "dur": 12,
    "args": {
      "External id": 279546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279546,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279546, "pid": 5, "tid": 7, "ts": 1716454225903544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843829, "dur": 8,
    "args": {
      "External id": 279546,
      "cbid": 211, "correlation": 279546
    }
  },
  {
    "ph": "s", "id": 279546, "pid": 76337, "tid": -914061504, "ts": 1716454225843829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225903557, "dur": 10,
    "args": {
      "External id": 279554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279554,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279554, "pid": 5, "tid": 7, "ts": 1716454225903557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843858, "dur": 9,
    "args": {
      "External id": 279554,
      "cbid": 211, "correlation": 279554
    }
  },
  {
    "ph": "s", "id": 279554, "pid": 76337, "tid": -914061504, "ts": 1716454225843858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225903568, "dur": 18,
    "args": {
      "External id": 279574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279574,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 279574, "pid": 5, "tid": 7, "ts": 1716454225903568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843939, "dur": 12,
    "args": {
      "External id": 279574,
      "cbid": 211, "correlation": 279574
    }
  },
  {
    "ph": "s", "id": 279574, "pid": 76337, "tid": -914061504, "ts": 1716454225843939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225903587, "dur": 4,
    "args": {
      "External id": 279586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279586,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 279586, "pid": 5, "tid": 7, "ts": 1716454225903587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843962, "dur": 6,
    "args": {
      "External id": 279586,
      "cbid": 211, "correlation": 279586
    }
  },
  {
    "ph": "s", "id": 279586, "pid": 76337, "tid": -914061504, "ts": 1716454225843962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225903593, "dur": 16,
    "args": {
      "External id": 279589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279589, "pid": 5, "tid": 7, "ts": 1716454225903593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225843988, "dur": 8,
    "args": {
      "External id": 279589,
      "cbid": 211, "correlation": 279589
    }
  },
  {
    "ph": "s", "id": 279589, "pid": 76337, "tid": -914061504, "ts": 1716454225843988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844048, "dur": 0,
    "args": {
      "External id": 279600,
      "cbid": 317, "correlation": 279600
    }
  },
  {
    "ph": "f", "id": 279600, "pid": 76337, "tid": -914061504, "ts": 1716454225844048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844049, "dur": 0,
    "args": {
      "External id": 279601,
      "cbid": 203, "correlation": 279601
    }
  },
  {
    "ph": "f", "id": 279601, "pid": 76337, "tid": -914061504, "ts": 1716454225844049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844049, "dur": 0,
    "args": {
      "External id": 279602,
      "cbid": 205, "correlation": 279602
    }
  },
  {
    "ph": "f", "id": 279602, "pid": 76337, "tid": -914061504, "ts": 1716454225844049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225903610, "dur": 11,
    "args": {
      "External id": 279606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279606,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279606, "pid": 5, "tid": 7, "ts": 1716454225903610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844063, "dur": 13,
    "args": {
      "External id": 279606,
      "cbid": 211, "correlation": 279606
    }
  },
  {
    "ph": "s", "id": 279606, "pid": 76337, "tid": -914061504, "ts": 1716454225844063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225903622, "dur": 3,
    "args": {
      "External id": 279608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279608, "pid": 5, "tid": 7, "ts": 1716454225903622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844080, "dur": 7,
    "args": {
      "External id": 279608,
      "cbid": 211, "correlation": 279608
    }
  },
  {
    "ph": "s", "id": 279608, "pid": 76337, "tid": -914061504, "ts": 1716454225844080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844090, "dur": 0,
    "args": {
      "External id": 279609,
      "cbid": 51, "correlation": 279609
    }
  },
  {
    "ph": "s", "id": 279609, "pid": 76337, "tid": -914061504, "ts": 1716454225844090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225903627, "dur": 88,
    "args": {
      "External id": 279610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279610,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 279610, "pid": 5, "tid": 7, "ts": 1716454225903627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844090, "dur": 6,
    "args": {
      "External id": 279610,
      "cbid": 211, "correlation": 279610
    }
  },
  {
    "ph": "s", "id": 279610, "pid": 76337, "tid": -914061504, "ts": 1716454225844090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225903716, "dur": 15,
    "args": {
      "External id": 279615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279615,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279615, "pid": 5, "tid": 7, "ts": 1716454225903716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844119, "dur": 8,
    "args": {
      "External id": 279615,
      "cbid": 211, "correlation": 279615
    }
  },
  {
    "ph": "s", "id": 279615, "pid": 76337, "tid": -914061504, "ts": 1716454225844119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225903732, "dur": 82,
    "args": {
      "External id": 279624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279624,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279624, "pid": 5, "tid": 7, "ts": 1716454225903732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844201, "dur": 14,
    "args": {
      "External id": 279624,
      "cbid": 211, "correlation": 279624
    }
  },
  {
    "ph": "s", "id": 279624, "pid": 76337, "tid": -914061504, "ts": 1716454225844201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225903816, "dur": 29,
    "args": {
      "External id": 279646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279646,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279646, "pid": 5, "tid": 7, "ts": 1716454225903816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844260, "dur": 11,
    "args": {
      "External id": 279646,
      "cbid": 211, "correlation": 279646
    }
  },
  {
    "ph": "s", "id": 279646, "pid": 76337, "tid": -914061504, "ts": 1716454225844260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844356, "dur": 2,
    "args": {
      "External id": 279657,
      "cbid": 251, "correlation": 279657
    }
  },
  {
    "ph": "f", "id": 279657, "pid": 76337, "tid": -914061504, "ts": 1716454225844356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225903846, "dur": 156,
    "args": {
      "External id": 279658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279658,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279658, "pid": 5, "tid": 7, "ts": 1716454225903846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844361, "dur": 14,
    "args": {
      "External id": 279658,
      "cbid": 211, "correlation": 279658
    }
  },
  {
    "ph": "s", "id": 279658, "pid": 76337, "tid": -914061504, "ts": 1716454225844361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844432, "dur": 1,
    "args": {
      "External id": 279669,
      "cbid": 251, "correlation": 279669
    }
  },
  {
    "ph": "f", "id": 279669, "pid": 76337, "tid": -914061504, "ts": 1716454225844432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225904003, "dur": 152,
    "args": {
      "External id": 279670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279670,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279670, "pid": 5, "tid": 7, "ts": 1716454225904003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844436, "dur": 12,
    "args": {
      "External id": 279670,
      "cbid": 211, "correlation": 279670
    }
  },
  {
    "ph": "s", "id": 279670, "pid": 76337, "tid": -914061504, "ts": 1716454225844436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844501, "dur": 1,
    "args": {
      "External id": 279681,
      "cbid": 251, "correlation": 279681
    }
  },
  {
    "ph": "f", "id": 279681, "pid": 76337, "tid": -914061504, "ts": 1716454225844501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225904156, "dur": 154,
    "args": {
      "External id": 279682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279682,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279682, "pid": 5, "tid": 7, "ts": 1716454225904156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844505, "dur": 11,
    "args": {
      "External id": 279682,
      "cbid": 211, "correlation": 279682
    }
  },
  {
    "ph": "s", "id": 279682, "pid": 76337, "tid": -914061504, "ts": 1716454225844505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225904312, "dur": 327,
    "args": {
      "External id": 279707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279707,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279707, "pid": 5, "tid": 7, "ts": 1716454225904312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844592, "dur": 16,
    "args": {
      "External id": 279707,
      "cbid": 211, "correlation": 279707
    }
  },
  {
    "ph": "s", "id": 279707, "pid": 76337, "tid": -914061504, "ts": 1716454225844592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844697, "dur": 2,
    "args": {
      "External id": 279725,
      "cbid": 251, "correlation": 279725
    }
  },
  {
    "ph": "f", "id": 279725, "pid": 76337, "tid": -914061504, "ts": 1716454225844697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225904640, "dur": 161,
    "args": {
      "External id": 279727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279727,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279727, "pid": 5, "tid": 7, "ts": 1716454225904640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844703, "dur": 16,
    "args": {
      "External id": 279727,
      "cbid": 211, "correlation": 279727
    }
  },
  {
    "ph": "s", "id": 279727, "pid": 76337, "tid": -914061504, "ts": 1716454225844703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225904802, "dur": 19,
    "args": {
      "External id": 279735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279735,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279735, "pid": 5, "tid": 7, "ts": 1716454225904802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844778, "dur": 13,
    "args": {
      "External id": 279735,
      "cbid": 211, "correlation": 279735
    }
  },
  {
    "ph": "s", "id": 279735, "pid": 76337, "tid": -914061504, "ts": 1716454225844778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225904822, "dur": 28,
    "args": {
      "External id": 279743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279743,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279743, "pid": 5, "tid": 7, "ts": 1716454225904822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844819, "dur": 10,
    "args": {
      "External id": 279743,
      "cbid": 211, "correlation": 279743
    }
  },
  {
    "ph": "s", "id": 279743, "pid": 76337, "tid": -914061504, "ts": 1716454225844819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225904852, "dur": 17,
    "args": {
      "External id": 279754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279754, "pid": 5, "tid": 7, "ts": 1716454225904852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844893, "dur": 16,
    "args": {
      "External id": 279754,
      "cbid": 211, "correlation": 279754
    }
  },
  {
    "ph": "s", "id": 279754, "pid": 76337, "tid": -914061504, "ts": 1716454225844893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225904871, "dur": 15,
    "args": {
      "External id": 279776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279776,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279776, "pid": 5, "tid": 7, "ts": 1716454225904871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225844928, "dur": 10,
    "args": {
      "External id": 279776,
      "cbid": 211, "correlation": 279776
    }
  },
  {
    "ph": "s", "id": 279776, "pid": 76337, "tid": -914061504, "ts": 1716454225844928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845030, "dur": 2,
    "args": {
      "External id": 279787,
      "cbid": 251, "correlation": 279787
    }
  },
  {
    "ph": "f", "id": 279787, "pid": 76337, "tid": -914061504, "ts": 1716454225845030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225904887, "dur": 85,
    "args": {
      "External id": 279788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279788,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279788, "pid": 5, "tid": 7, "ts": 1716454225904887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845036, "dur": 14,
    "args": {
      "External id": 279788,
      "cbid": 211, "correlation": 279788
    }
  },
  {
    "ph": "s", "id": 279788, "pid": 76337, "tid": -914061504, "ts": 1716454225845036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845108, "dur": 1,
    "args": {
      "External id": 279799,
      "cbid": 251, "correlation": 279799
    }
  },
  {
    "ph": "f", "id": 279799, "pid": 76337, "tid": -914061504, "ts": 1716454225845108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845112, "dur": 0,
    "args": {
      "External id": 279800,
      "cbid": 251, "correlation": 279800
    }
  },
  {
    "ph": "f", "id": 279800, "pid": 76337, "tid": -914061504, "ts": 1716454225845112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225904973, "dur": 12,
    "args": {
      "External id": 279801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279801,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279801, "pid": 5, "tid": 7, "ts": 1716454225904973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845114, "dur": 14,
    "args": {
      "External id": 279801,
      "cbid": 211, "correlation": 279801
    }
  },
  {
    "ph": "s", "id": 279801, "pid": 76337, "tid": -914061504, "ts": 1716454225845114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225904987, "dur": 5,
    "args": {
      "External id": 279803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279803,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279803, "pid": 5, "tid": 7, "ts": 1716454225904987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845130, "dur": 8,
    "args": {
      "External id": 279803,
      "cbid": 211, "correlation": 279803
    }
  },
  {
    "ph": "s", "id": 279803, "pid": 76337, "tid": -914061504, "ts": 1716454225845130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845190, "dur": 1,
    "args": {
      "External id": 279814,
      "cbid": 251, "correlation": 279814
    }
  },
  {
    "ph": "f", "id": 279814, "pid": 76337, "tid": -914061504, "ts": 1716454225845190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845193, "dur": 0,
    "args": {
      "External id": 279815,
      "cbid": 251, "correlation": 279815
    }
  },
  {
    "ph": "f", "id": 279815, "pid": 76337, "tid": -914061504, "ts": 1716454225845193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225904993, "dur": 8,
    "args": {
      "External id": 279816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279816,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279816, "pid": 5, "tid": 7, "ts": 1716454225904993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845195, "dur": 12,
    "args": {
      "External id": 279816,
      "cbid": 211, "correlation": 279816
    }
  },
  {
    "ph": "s", "id": 279816, "pid": 76337, "tid": -914061504, "ts": 1716454225845195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225905003, "dur": 3,
    "args": {
      "External id": 279818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279818,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279818, "pid": 5, "tid": 7, "ts": 1716454225905003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845208, "dur": 7,
    "args": {
      "External id": 279818,
      "cbid": 211, "correlation": 279818
    }
  },
  {
    "ph": "s", "id": 279818, "pid": 76337, "tid": -914061504, "ts": 1716454225845208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225905008, "dur": 53,
    "args": {
      "External id": 279843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279843,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279843, "pid": 5, "tid": 7, "ts": 1716454225905008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845286, "dur": 13,
    "args": {
      "External id": 279843,
      "cbid": 211, "correlation": 279843
    }
  },
  {
    "ph": "s", "id": 279843, "pid": 76337, "tid": -914061504, "ts": 1716454225845286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845384, "dur": 1,
    "args": {
      "External id": 279861,
      "cbid": 251, "correlation": 279861
    }
  },
  {
    "ph": "f", "id": 279861, "pid": 76337, "tid": -914061504, "ts": 1716454225845384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225905062, "dur": 87,
    "args": {
      "External id": 279863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279863,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279863, "pid": 5, "tid": 7, "ts": 1716454225905062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845390, "dur": 15,
    "args": {
      "External id": 279863,
      "cbid": 211, "correlation": 279863
    }
  },
  {
    "ph": "s", "id": 279863, "pid": 76337, "tid": -914061504, "ts": 1716454225845390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225905151, "dur": 9,
    "args": {
      "External id": 279871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279871,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279871, "pid": 5, "tid": 7, "ts": 1716454225905151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845460, "dur": 16,
    "args": {
      "External id": 279871,
      "cbid": 211, "correlation": 279871
    }
  },
  {
    "ph": "s", "id": 279871, "pid": 76337, "tid": -914061504, "ts": 1716454225845460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225905161, "dur": 20,
    "args": {
      "External id": 279879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279879,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279879, "pid": 5, "tid": 7, "ts": 1716454225905161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845505, "dur": 10,
    "args": {
      "External id": 279879,
      "cbid": 211, "correlation": 279879
    }
  },
  {
    "ph": "s", "id": 279879, "pid": 76337, "tid": -914061504, "ts": 1716454225845505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225905183, "dur": 17,
    "args": {
      "External id": 279901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279901,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279901, "pid": 5, "tid": 7, "ts": 1716454225905183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845557, "dur": 11,
    "args": {
      "External id": 279901,
      "cbid": 211, "correlation": 279901
    }
  },
  {
    "ph": "s", "id": 279901, "pid": 76337, "tid": -914061504, "ts": 1716454225845557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845647, "dur": 1,
    "args": {
      "External id": 279917,
      "cbid": 251, "correlation": 279917
    }
  },
  {
    "ph": "f", "id": 279917, "pid": 76337, "tid": -914061504, "ts": 1716454225845647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845652, "dur": 0,
    "args": {
      "External id": 279919,
      "cbid": 251, "correlation": 279919
    }
  },
  {
    "ph": "f", "id": 279919, "pid": 76337, "tid": -914061504, "ts": 1716454225845652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225905201, "dur": 488,
    "args": {
      "External id": 279920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279920,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 279920, "pid": 5, "tid": 7, "ts": 1716454225905201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845654, "dur": 13,
    "args": {
      "External id": 279920,
      "cbid": 211, "correlation": 279920
    }
  },
  {
    "ph": "s", "id": 279920, "pid": 76337, "tid": -914061504, "ts": 1716454225845654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225905691, "dur": 64,
    "args": {
      "External id": 279928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279928, "pid": 5, "tid": 7, "ts": 1716454225905691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845725, "dur": 13,
    "args": {
      "External id": 279928,
      "cbid": 211, "correlation": 279928
    }
  },
  {
    "ph": "s", "id": 279928, "pid": 76337, "tid": -914061504, "ts": 1716454225845725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225905756, "dur": 65,
    "args": {
      "External id": 279936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279936, "pid": 5, "tid": 7, "ts": 1716454225905756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845759, "dur": 9,
    "args": {
      "External id": 279936,
      "cbid": 211, "correlation": 279936
    }
  },
  {
    "ph": "s", "id": 279936, "pid": 76337, "tid": -914061504, "ts": 1716454225845759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845839, "dur": 1,
    "args": {
      "External id": 279952,
      "cbid": 251, "correlation": 279952
    }
  },
  {
    "ph": "f", "id": 279952, "pid": 76337, "tid": -914061504, "ts": 1716454225845839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225905823, "dur": 1,
    "args": {
      "External id": 279954,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 279954,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 279954, "pid": 5, "tid": 7, "ts": 1716454225905823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845844, "dur": 14,
    "args": {
      "External id": 279954,
      "cbid": 51, "correlation": 279954
    }
  },
  {
    "ph": "s", "id": 279954, "pid": 76337, "tid": -914061504, "ts": 1716454225845844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225905826, "dur": 261,
    "args": {
      "External id": 279955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279955,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279955, "pid": 5, "tid": 7, "ts": 1716454225905826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845860, "dur": 15,
    "args": {
      "External id": 279955,
      "cbid": 211, "correlation": 279955
    }
  },
  {
    "ph": "s", "id": 279955, "pid": 76337, "tid": -914061504, "ts": 1716454225845860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225906089, "dur": 14,
    "args": {
      "External id": 279963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279963,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279963, "pid": 5, "tid": 7, "ts": 1716454225906089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225845910, "dur": 16,
    "args": {
      "External id": 279963,
      "cbid": 211, "correlation": 279963
    }
  },
  {
    "ph": "s", "id": 279963, "pid": 76337, "tid": -914061504, "ts": 1716454225845910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225906104, "dur": 36,
    "args": {
      "External id": 279974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279974,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279974, "pid": 5, "tid": 7, "ts": 1716454225906104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846092, "dur": 53,
    "args": {
      "External id": 279974,
      "cbid": 211, "correlation": 279974
    }
  },
  {
    "ph": "s", "id": 279974, "pid": 76337, "tid": -914061504, "ts": 1716454225846092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846242, "dur": 0,
    "args": {
      "External id": 279986,
      "cbid": 317, "correlation": 279986
    }
  },
  {
    "ph": "f", "id": 279986, "pid": 76337, "tid": -914061504, "ts": 1716454225846242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846244, "dur": 0,
    "args": {
      "External id": 279987,
      "cbid": 203, "correlation": 279987
    }
  },
  {
    "ph": "f", "id": 279987, "pid": 76337, "tid": -914061504, "ts": 1716454225846244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846246, "dur": 0,
    "args": {
      "External id": 279988,
      "cbid": 205, "correlation": 279988
    }
  },
  {
    "ph": "f", "id": 279988, "pid": 76337, "tid": -914061504, "ts": 1716454225846246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225906142, "dur": 12,
    "args": {
      "External id": 279992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 279992, "pid": 5, "tid": 7, "ts": 1716454225906142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846282, "dur": 22,
    "args": {
      "External id": 279992,
      "cbid": 211, "correlation": 279992
    }
  },
  {
    "ph": "s", "id": 279992, "pid": 76337, "tid": -914061504, "ts": 1716454225846282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225906155, "dur": 4,
    "args": {
      "External id": 279994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 279994, "pid": 5, "tid": 7, "ts": 1716454225906155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846311, "dur": 8,
    "args": {
      "External id": 279994,
      "cbid": 211, "correlation": 279994
    }
  },
  {
    "ph": "s", "id": 279994, "pid": 76337, "tid": -914061504, "ts": 1716454225846311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846322, "dur": 1,
    "args": {
      "External id": 279995,
      "cbid": 51, "correlation": 279995
    }
  },
  {
    "ph": "s", "id": 279995, "pid": 76337, "tid": -914061504, "ts": 1716454225846322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225906160, "dur": 93,
    "args": {
      "External id": 279996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 279996,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 279996, "pid": 5, "tid": 7, "ts": 1716454225906160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846324, "dur": 9,
    "args": {
      "External id": 279996,
      "cbid": 211, "correlation": 279996
    }
  },
  {
    "ph": "s", "id": 279996, "pid": 76337, "tid": -914061504, "ts": 1716454225846324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225906254, "dur": 15,
    "args": {
      "External id": 280001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280001,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280001, "pid": 5, "tid": 7, "ts": 1716454225906254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846379, "dur": 18,
    "args": {
      "External id": 280001,
      "cbid": 211, "correlation": 280001
    }
  },
  {
    "ph": "s", "id": 280001, "pid": 76337, "tid": -914061504, "ts": 1716454225846379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225906271, "dur": 11,
    "args": {
      "External id": 280009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280009,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280009, "pid": 5, "tid": 7, "ts": 1716454225906271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846513, "dur": 14,
    "args": {
      "External id": 280009,
      "cbid": 211, "correlation": 280009
    }
  },
  {
    "ph": "s", "id": 280009, "pid": 76337, "tid": -914061504, "ts": 1716454225846513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225906283, "dur": 24,
    "args": {
      "External id": 280018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280018,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280018, "pid": 5, "tid": 7, "ts": 1716454225906283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846644, "dur": 33,
    "args": {
      "External id": 280018,
      "cbid": 211, "correlation": 280018
    }
  },
  {
    "ph": "s", "id": 280018, "pid": 76337, "tid": -914061504, "ts": 1716454225846644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225906309, "dur": 23,
    "args": {
      "External id": 280038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280038,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 280038, "pid": 5, "tid": 7, "ts": 1716454225906309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846871, "dur": 23,
    "args": {
      "External id": 280038,
      "cbid": 211, "correlation": 280038
    }
  },
  {
    "ph": "s", "id": 280038, "pid": 76337, "tid": -914061504, "ts": 1716454225846871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225906333, "dur": 5,
    "args": {
      "External id": 280050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280050,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 280050, "pid": 5, "tid": 7, "ts": 1716454225906333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846921, "dur": 15,
    "args": {
      "External id": 280050,
      "cbid": 211, "correlation": 280050
    }
  },
  {
    "ph": "s", "id": 280050, "pid": 76337, "tid": -914061504, "ts": 1716454225846921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225906340, "dur": 24,
    "args": {
      "External id": 280053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280053,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280053, "pid": 5, "tid": 7, "ts": 1716454225906340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225846955, "dur": 96,
    "args": {
      "External id": 280053,
      "cbid": 211, "correlation": 280053
    }
  },
  {
    "ph": "s", "id": 280053, "pid": 76337, "tid": -914061504, "ts": 1716454225846955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225906365, "dur": 17,
    "args": {
      "External id": 280062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280062,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280062, "pid": 5, "tid": 7, "ts": 1716454225906365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847128, "dur": 18,
    "args": {
      "External id": 280062,
      "cbid": 211, "correlation": 280062
    }
  },
  {
    "ph": "s", "id": 280062, "pid": 76337, "tid": -914061504, "ts": 1716454225847128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847306, "dur": 0,
    "args": {
      "External id": 280072,
      "cbid": 317, "correlation": 280072
    }
  },
  {
    "ph": "f", "id": 280072, "pid": 76337, "tid": -914061504, "ts": 1716454225847306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847307, "dur": 0,
    "args": {
      "External id": 280073,
      "cbid": 203, "correlation": 280073
    }
  },
  {
    "ph": "f", "id": 280073, "pid": 76337, "tid": -914061504, "ts": 1716454225847307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847308, "dur": 0,
    "args": {
      "External id": 280074,
      "cbid": 205, "correlation": 280074
    }
  },
  {
    "ph": "f", "id": 280074, "pid": 76337, "tid": -914061504, "ts": 1716454225847308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225906383, "dur": 17,
    "args": {
      "External id": 280078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280078,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280078, "pid": 5, "tid": 7, "ts": 1716454225906383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847331, "dur": 18,
    "args": {
      "External id": 280078,
      "cbid": 211, "correlation": 280078
    }
  },
  {
    "ph": "s", "id": 280078, "pid": 76337, "tid": -914061504, "ts": 1716454225847331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225906402, "dur": 230,
    "args": {
      "External id": 280080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280080,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280080, "pid": 5, "tid": 7, "ts": 1716454225906402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847355, "dur": 46,
    "args": {
      "External id": 280080,
      "cbid": 211, "correlation": 280080
    }
  },
  {
    "ph": "s", "id": 280080, "pid": 76337, "tid": -914061504, "ts": 1716454225847355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225906634, "dur": 1,
    "args": {
      "External id": 280082,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 280082,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 280082, "pid": 5, "tid": 7, "ts": 1716454225906634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847460, "dur": 32,
    "args": {
      "External id": 280082,
      "cbid": 51, "correlation": 280082
    }
  },
  {
    "ph": "s", "id": 280082, "pid": 76337, "tid": -914061504, "ts": 1716454225847460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225906638, "dur": 802,
    "args": {
      "External id": 280083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280083,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280083, "pid": 5, "tid": 7, "ts": 1716454225906638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847497, "dur": 21,
    "args": {
      "External id": 280083,
      "cbid": 211, "correlation": 280083
    }
  },
  {
    "ph": "s", "id": 280083, "pid": 76337, "tid": -914061504, "ts": 1716454225847497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225907442, "dur": 13,
    "args": {
      "External id": 280085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280085,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280085, "pid": 5, "tid": 7, "ts": 1716454225907442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847528, "dur": 24,
    "args": {
      "External id": 280085,
      "cbid": 211, "correlation": 280085
    }
  },
  {
    "ph": "s", "id": 280085, "pid": 76337, "tid": -914061504, "ts": 1716454225847528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225907455, "dur": 14,
    "args": {
      "External id": 280091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280091, "pid": 5, "tid": 7, "ts": 1716454225907455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847688, "dur": 42,
    "args": {
      "External id": 280091,
      "cbid": 211, "correlation": 280091
    }
  },
  {
    "ph": "s", "id": 280091, "pid": 76337, "tid": -914061504, "ts": 1716454225847688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225907471, "dur": 4,
    "args": {
      "External id": 280099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280099,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 280099, "pid": 5, "tid": 7, "ts": 1716454225907471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225847896, "dur": 42,
    "args": {
      "External id": 280099,
      "cbid": 211, "correlation": 280099
    }
  },
  {
    "ph": "s", "id": 280099, "pid": 76337, "tid": -914061504, "ts": 1716454225847896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848342, "dur": 10,
    "args": {
      "External id": 280115,
      "cbid": 251, "correlation": 280115
    }
  },
  {
    "ph": "f", "id": 280115, "pid": 76337, "tid": -914061504, "ts": 1716454225848342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848374, "dur": 6,
    "args": {
      "External id": 280117,
      "cbid": 251, "correlation": 280117
    }
  },
  {
    "ph": "f", "id": 280117, "pid": 76337, "tid": -914061504, "ts": 1716454225848374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225907476, "dur": 13,
    "args": {
      "External id": 280118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280118,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280118, "pid": 5, "tid": 7, "ts": 1716454225907476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848401, "dur": 88,
    "args": {
      "External id": 280118,
      "cbid": 211, "correlation": 280118
    }
  },
  {
    "ph": "s", "id": 280118, "pid": 76337, "tid": -914061504, "ts": 1716454225848401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225907490, "dur": 5,
    "args": {
      "External id": 280120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280120,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280120, "pid": 5, "tid": 7, "ts": 1716454225907490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848510, "dur": 31,
    "args": {
      "External id": 280120,
      "cbid": 211, "correlation": 280120
    }
  },
  {
    "ph": "s", "id": 280120, "pid": 76337, "tid": -914061504, "ts": 1716454225848510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225907496, "dur": 17,
    "args": {
      "External id": 280130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280130, "pid": 5, "tid": 7, "ts": 1716454225907496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848669, "dur": 19,
    "args": {
      "External id": 280130,
      "cbid": 211, "correlation": 280130
    }
  },
  {
    "ph": "s", "id": 280130, "pid": 76337, "tid": -914061504, "ts": 1716454225848669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225907515, "dur": 18,
    "args": {
      "External id": 280150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280150,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 280150, "pid": 5, "tid": 7, "ts": 1716454225907515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848814, "dur": 33,
    "args": {
      "External id": 280150,
      "cbid": 211, "correlation": 280150
    }
  },
  {
    "ph": "s", "id": 280150, "pid": 76337, "tid": -914061504, "ts": 1716454225848814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225907533, "dur": 5,
    "args": {
      "External id": 280162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280162,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 280162, "pid": 5, "tid": 7, "ts": 1716454225907533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848876, "dur": 12,
    "args": {
      "External id": 280162,
      "cbid": 211, "correlation": 280162
    }
  },
  {
    "ph": "s", "id": 280162, "pid": 76337, "tid": -914061504, "ts": 1716454225848876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225907540, "dur": 16,
    "args": {
      "External id": 280165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280165, "pid": 5, "tid": 7, "ts": 1716454225907540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848902, "dur": 8,
    "args": {
      "External id": 280165,
      "cbid": 211, "correlation": 280165
    }
  },
  {
    "ph": "s", "id": 280165, "pid": 76337, "tid": -914061504, "ts": 1716454225848902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225907557, "dur": 11,
    "args": {
      "External id": 280174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280174,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280174, "pid": 5, "tid": 7, "ts": 1716454225907557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225848954, "dur": 15,
    "args": {
      "External id": 280174,
      "cbid": 211, "correlation": 280174
    }
  },
  {
    "ph": "s", "id": 280174, "pid": 76337, "tid": -914061504, "ts": 1716454225848954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849059, "dur": 0,
    "args": {
      "External id": 280184,
      "cbid": 317, "correlation": 280184
    }
  },
  {
    "ph": "f", "id": 280184, "pid": 76337, "tid": -914061504, "ts": 1716454225849059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849060, "dur": 0,
    "args": {
      "External id": 280185,
      "cbid": 203, "correlation": 280185
    }
  },
  {
    "ph": "f", "id": 280185, "pid": 76337, "tid": -914061504, "ts": 1716454225849060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849061, "dur": 0,
    "args": {
      "External id": 280186,
      "cbid": 205, "correlation": 280186
    }
  },
  {
    "ph": "f", "id": 280186, "pid": 76337, "tid": -914061504, "ts": 1716454225849061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225907570, "dur": 11,
    "args": {
      "External id": 280190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280190, "pid": 5, "tid": 7, "ts": 1716454225907570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849080, "dur": 15,
    "args": {
      "External id": 280190,
      "cbid": 211, "correlation": 280190
    }
  },
  {
    "ph": "s", "id": 280190, "pid": 76337, "tid": -914061504, "ts": 1716454225849080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225907582, "dur": 155,
    "args": {
      "External id": 280192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280192, "pid": 5, "tid": 7, "ts": 1716454225907582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849098, "dur": 8,
    "args": {
      "External id": 280192,
      "cbid": 211, "correlation": 280192
    }
  },
  {
    "ph": "s", "id": 280192, "pid": 76337, "tid": -914061504, "ts": 1716454225849098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225907740, "dur": 1,
    "args": {
      "External id": 280194,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 280194,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 280194, "pid": 5, "tid": 7, "ts": 1716454225907740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849115, "dur": 11,
    "args": {
      "External id": 280194,
      "cbid": 51, "correlation": 280194
    }
  },
  {
    "ph": "s", "id": 280194, "pid": 76337, "tid": -914061504, "ts": 1716454225849115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225907743, "dur": 635,
    "args": {
      "External id": 280195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280195,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280195, "pid": 5, "tid": 7, "ts": 1716454225907743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849127, "dur": 10,
    "args": {
      "External id": 280195,
      "cbid": 211, "correlation": 280195
    }
  },
  {
    "ph": "s", "id": 280195, "pid": 76337, "tid": -914061504, "ts": 1716454225849127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225908379, "dur": 12,
    "args": {
      "External id": 280197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280197,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280197, "pid": 5, "tid": 7, "ts": 1716454225908379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849144, "dur": 8,
    "args": {
      "External id": 280197,
      "cbid": 211, "correlation": 280197
    }
  },
  {
    "ph": "s", "id": 280197, "pid": 76337, "tid": -914061504, "ts": 1716454225849144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225908393, "dur": 14,
    "args": {
      "External id": 280203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280203, "pid": 5, "tid": 7, "ts": 1716454225908393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849185, "dur": 16,
    "args": {
      "External id": 280203,
      "cbid": 211, "correlation": 280203
    }
  },
  {
    "ph": "s", "id": 280203, "pid": 76337, "tid": -914061504, "ts": 1716454225849185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849270, "dur": 0,
    "args": {
      "External id": 280213,
      "cbid": 317, "correlation": 280213
    }
  },
  {
    "ph": "f", "id": 280213, "pid": 76337, "tid": -914061504, "ts": 1716454225849270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849271, "dur": 0,
    "args": {
      "External id": 280214,
      "cbid": 203, "correlation": 280214
    }
  },
  {
    "ph": "f", "id": 280214, "pid": 76337, "tid": -914061504, "ts": 1716454225849271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849272, "dur": 0,
    "args": {
      "External id": 280215,
      "cbid": 205, "correlation": 280215
    }
  },
  {
    "ph": "f", "id": 280215, "pid": 76337, "tid": -914061504, "ts": 1716454225849272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225908409, "dur": 17,
    "args": {
      "External id": 280219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280219,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280219, "pid": 5, "tid": 7, "ts": 1716454225908409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849288, "dur": 22,
    "args": {
      "External id": 280219,
      "cbid": 211, "correlation": 280219
    }
  },
  {
    "ph": "s", "id": 280219, "pid": 76337, "tid": -914061504, "ts": 1716454225849288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225908427, "dur": 4,
    "args": {
      "External id": 280221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280221,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 280221, "pid": 5, "tid": 7, "ts": 1716454225908427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849317, "dur": 8,
    "args": {
      "External id": 280221,
      "cbid": 211, "correlation": 280221
    }
  },
  {
    "ph": "s", "id": 280221, "pid": 76337, "tid": -914061504, "ts": 1716454225849317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849328, "dur": 0,
    "args": {
      "External id": 280222,
      "cbid": 51, "correlation": 280222
    }
  },
  {
    "ph": "s", "id": 280222, "pid": 76337, "tid": -914061504, "ts": 1716454225849328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225908432, "dur": 125,
    "args": {
      "External id": 280223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280223,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 280223, "pid": 5, "tid": 7, "ts": 1716454225908432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849329, "dur": 6,
    "args": {
      "External id": 280223,
      "cbid": 211, "correlation": 280223
    }
  },
  {
    "ph": "s", "id": 280223, "pid": 76337, "tid": -914061504, "ts": 1716454225849329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225908559, "dur": 15,
    "args": {
      "External id": 280228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280228,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280228, "pid": 5, "tid": 7, "ts": 1716454225908559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849365, "dur": 15,
    "args": {
      "External id": 280228,
      "cbid": 211, "correlation": 280228
    }
  },
  {
    "ph": "s", "id": 280228, "pid": 76337, "tid": -914061504, "ts": 1716454225849365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225908575, "dur": 13,
    "args": {
      "External id": 280236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280236,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280236, "pid": 5, "tid": 7, "ts": 1716454225908575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849411, "dur": 14,
    "args": {
      "External id": 280236,
      "cbid": 211, "correlation": 280236
    }
  },
  {
    "ph": "s", "id": 280236, "pid": 76337, "tid": -914061504, "ts": 1716454225849411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225908589, "dur": 10,
    "args": {
      "External id": 280244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280244,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280244, "pid": 5, "tid": 7, "ts": 1716454225908589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849459, "dur": 12,
    "args": {
      "External id": 280244,
      "cbid": 211, "correlation": 280244
    }
  },
  {
    "ph": "s", "id": 280244, "pid": 76337, "tid": -914061504, "ts": 1716454225849459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225908600, "dur": 17,
    "args": {
      "External id": 280264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280264,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 280264, "pid": 5, "tid": 7, "ts": 1716454225908600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849561, "dur": 14,
    "args": {
      "External id": 280264,
      "cbid": 211, "correlation": 280264
    }
  },
  {
    "ph": "s", "id": 280264, "pid": 76337, "tid": -914061504, "ts": 1716454225849561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225908619, "dur": 4,
    "args": {
      "External id": 280276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280276,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 280276, "pid": 5, "tid": 7, "ts": 1716454225908619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849594, "dur": 18,
    "args": {
      "External id": 280276,
      "cbid": 211, "correlation": 280276
    }
  },
  {
    "ph": "s", "id": 280276, "pid": 76337, "tid": -914061504, "ts": 1716454225849594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225908624, "dur": 16,
    "args": {
      "External id": 280279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280279, "pid": 5, "tid": 7, "ts": 1716454225908624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849632, "dur": 16,
    "args": {
      "External id": 280279,
      "cbid": 211, "correlation": 280279
    }
  },
  {
    "ph": "s", "id": 280279, "pid": 76337, "tid": -914061504, "ts": 1716454225849632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849730, "dur": 1,
    "args": {
      "External id": 280290,
      "cbid": 317, "correlation": 280290
    }
  },
  {
    "ph": "f", "id": 280290, "pid": 76337, "tid": -914061504, "ts": 1716454225849730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849732, "dur": 0,
    "args": {
      "External id": 280291,
      "cbid": 203, "correlation": 280291
    }
  },
  {
    "ph": "f", "id": 280291, "pid": 76337, "tid": -914061504, "ts": 1716454225849732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849733, "dur": 0,
    "args": {
      "External id": 280292,
      "cbid": 205, "correlation": 280292
    }
  },
  {
    "ph": "f", "id": 280292, "pid": 76337, "tid": -914061504, "ts": 1716454225849733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225908642, "dur": 11,
    "args": {
      "External id": 280296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280296,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280296, "pid": 5, "tid": 7, "ts": 1716454225908642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849750, "dur": 20,
    "args": {
      "External id": 280296,
      "cbid": 211, "correlation": 280296
    }
  },
  {
    "ph": "s", "id": 280296, "pid": 76337, "tid": -914061504, "ts": 1716454225849750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225908654, "dur": 3,
    "args": {
      "External id": 280298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280298,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 280298, "pid": 5, "tid": 7, "ts": 1716454225908654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849776, "dur": 10,
    "args": {
      "External id": 280298,
      "cbid": 211, "correlation": 280298
    }
  },
  {
    "ph": "s", "id": 280298, "pid": 76337, "tid": -914061504, "ts": 1716454225849776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849789, "dur": 0,
    "args": {
      "External id": 280299,
      "cbid": 51, "correlation": 280299
    }
  },
  {
    "ph": "s", "id": 280299, "pid": 76337, "tid": -914061504, "ts": 1716454225849789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225908658, "dur": 87,
    "args": {
      "External id": 280300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280300,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 280300, "pid": 5, "tid": 7, "ts": 1716454225908658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849790, "dur": 5,
    "args": {
      "External id": 280300,
      "cbid": 211, "correlation": 280300
    }
  },
  {
    "ph": "s", "id": 280300, "pid": 76337, "tid": -914061504, "ts": 1716454225849790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225908746, "dur": 15,
    "args": {
      "External id": 280305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280305,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280305, "pid": 5, "tid": 7, "ts": 1716454225908746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849822, "dur": 9,
    "args": {
      "External id": 280305,
      "cbid": 211, "correlation": 280305
    }
  },
  {
    "ph": "s", "id": 280305, "pid": 76337, "tid": -914061504, "ts": 1716454225849822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225908763, "dur": 79,
    "args": {
      "External id": 280314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280314,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280314, "pid": 5, "tid": 7, "ts": 1716454225908763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849920, "dur": 16,
    "args": {
      "External id": 280314,
      "cbid": 211, "correlation": 280314
    }
  },
  {
    "ph": "s", "id": 280314, "pid": 76337, "tid": -914061504, "ts": 1716454225849920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225908843, "dur": 29,
    "args": {
      "External id": 280336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280336,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280336, "pid": 5, "tid": 7, "ts": 1716454225908843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225849995, "dur": 11,
    "args": {
      "External id": 280336,
      "cbid": 211, "correlation": 280336
    }
  },
  {
    "ph": "s", "id": 280336, "pid": 76337, "tid": -914061504, "ts": 1716454225849995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850121, "dur": 4,
    "args": {
      "External id": 280347,
      "cbid": 251, "correlation": 280347
    }
  },
  {
    "ph": "f", "id": 280347, "pid": 76337, "tid": -914061504, "ts": 1716454225850121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225908873, "dur": 155,
    "args": {
      "External id": 280348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280348,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280348, "pid": 5, "tid": 7, "ts": 1716454225908873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850134, "dur": 19,
    "args": {
      "External id": 280348,
      "cbid": 211, "correlation": 280348
    }
  },
  {
    "ph": "s", "id": 280348, "pid": 76337, "tid": -914061504, "ts": 1716454225850134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850221, "dur": 1,
    "args": {
      "External id": 280359,
      "cbid": 251, "correlation": 280359
    }
  },
  {
    "ph": "f", "id": 280359, "pid": 76337, "tid": -914061504, "ts": 1716454225850221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225909029, "dur": 153,
    "args": {
      "External id": 280360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280360,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280360, "pid": 5, "tid": 7, "ts": 1716454225909029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850225, "dur": 13,
    "args": {
      "External id": 280360,
      "cbid": 211, "correlation": 280360
    }
  },
  {
    "ph": "s", "id": 280360, "pid": 76337, "tid": -914061504, "ts": 1716454225850225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850292, "dur": 1,
    "args": {
      "External id": 280371,
      "cbid": 251, "correlation": 280371
    }
  },
  {
    "ph": "f", "id": 280371, "pid": 76337, "tid": -914061504, "ts": 1716454225850292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225909184, "dur": 155,
    "args": {
      "External id": 280372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280372,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280372, "pid": 5, "tid": 7, "ts": 1716454225909184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850296, "dur": 11,
    "args": {
      "External id": 280372,
      "cbid": 211, "correlation": 280372
    }
  },
  {
    "ph": "s", "id": 280372, "pid": 76337, "tid": -914061504, "ts": 1716454225850296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225909340, "dur": 326,
    "args": {
      "External id": 280397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280397,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280397, "pid": 5, "tid": 7, "ts": 1716454225909340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850406, "dur": 18,
    "args": {
      "External id": 280397,
      "cbid": 211, "correlation": 280397
    }
  },
  {
    "ph": "s", "id": 280397, "pid": 76337, "tid": -914061504, "ts": 1716454225850406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850577, "dur": 2,
    "args": {
      "External id": 280415,
      "cbid": 251, "correlation": 280415
    }
  },
  {
    "ph": "f", "id": 280415, "pid": 76337, "tid": -914061504, "ts": 1716454225850577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225909667, "dur": 159,
    "args": {
      "External id": 280417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280417,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280417, "pid": 5, "tid": 7, "ts": 1716454225909667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850584, "dur": 16,
    "args": {
      "External id": 280417,
      "cbid": 211, "correlation": 280417
    }
  },
  {
    "ph": "s", "id": 280417, "pid": 76337, "tid": -914061504, "ts": 1716454225850584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225909828, "dur": 19,
    "args": {
      "External id": 280425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280425,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280425, "pid": 5, "tid": 7, "ts": 1716454225909828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850666, "dur": 12,
    "args": {
      "External id": 280425,
      "cbid": 211, "correlation": 280425
    }
  },
  {
    "ph": "s", "id": 280425, "pid": 76337, "tid": -914061504, "ts": 1716454225850666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225909848, "dur": 28,
    "args": {
      "External id": 280433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280433,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280433, "pid": 5, "tid": 7, "ts": 1716454225909848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850706, "dur": 8,
    "args": {
      "External id": 280433,
      "cbid": 211, "correlation": 280433
    }
  },
  {
    "ph": "s", "id": 280433, "pid": 76337, "tid": -914061504, "ts": 1716454225850706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225909877, "dur": 18,
    "args": {
      "External id": 280444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280444,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280444, "pid": 5, "tid": 7, "ts": 1716454225909877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850795, "dur": 14,
    "args": {
      "External id": 280444,
      "cbid": 211, "correlation": 280444
    }
  },
  {
    "ph": "s", "id": 280444, "pid": 76337, "tid": -914061504, "ts": 1716454225850795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225909897, "dur": 15,
    "args": {
      "External id": 280466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280466,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280466, "pid": 5, "tid": 7, "ts": 1716454225909897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850829, "dur": 8,
    "args": {
      "External id": 280466,
      "cbid": 211, "correlation": 280466
    }
  },
  {
    "ph": "s", "id": 280466, "pid": 76337, "tid": -914061504, "ts": 1716454225850829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850917, "dur": 2,
    "args": {
      "External id": 280477,
      "cbid": 251, "correlation": 280477
    }
  },
  {
    "ph": "f", "id": 280477, "pid": 76337, "tid": -914061504, "ts": 1716454225850917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225909913, "dur": 85,
    "args": {
      "External id": 280478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280478,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 280478, "pid": 5, "tid": 7, "ts": 1716454225909913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225850923, "dur": 14,
    "args": {
      "External id": 280478,
      "cbid": 211, "correlation": 280478
    }
  },
  {
    "ph": "s", "id": 280478, "pid": 76337, "tid": -914061504, "ts": 1716454225850923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851014, "dur": 1,
    "args": {
      "External id": 280489,
      "cbid": 251, "correlation": 280489
    }
  },
  {
    "ph": "f", "id": 280489, "pid": 76337, "tid": -914061504, "ts": 1716454225851014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851018, "dur": 0,
    "args": {
      "External id": 280490,
      "cbid": 251, "correlation": 280490
    }
  },
  {
    "ph": "f", "id": 280490, "pid": 76337, "tid": -914061504, "ts": 1716454225851018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225910000, "dur": 12,
    "args": {
      "External id": 280491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280491,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280491, "pid": 5, "tid": 7, "ts": 1716454225910000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851020, "dur": 14,
    "args": {
      "External id": 280491,
      "cbid": 211, "correlation": 280491
    }
  },
  {
    "ph": "s", "id": 280491, "pid": 76337, "tid": -914061504, "ts": 1716454225851020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225910013, "dur": 5,
    "args": {
      "External id": 280493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280493,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280493, "pid": 5, "tid": 7, "ts": 1716454225910013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851037, "dur": 8,
    "args": {
      "External id": 280493,
      "cbid": 211, "correlation": 280493
    }
  },
  {
    "ph": "s", "id": 280493, "pid": 76337, "tid": -914061504, "ts": 1716454225851037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851099, "dur": 1,
    "args": {
      "External id": 280504,
      "cbid": 251, "correlation": 280504
    }
  },
  {
    "ph": "f", "id": 280504, "pid": 76337, "tid": -914061504, "ts": 1716454225851099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851103, "dur": 0,
    "args": {
      "External id": 280505,
      "cbid": 251, "correlation": 280505
    }
  },
  {
    "ph": "f", "id": 280505, "pid": 76337, "tid": -914061504, "ts": 1716454225851103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225910020, "dur": 9,
    "args": {
      "External id": 280506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280506,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280506, "pid": 5, "tid": 7, "ts": 1716454225910020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851105, "dur": 12,
    "args": {
      "External id": 280506,
      "cbid": 211, "correlation": 280506
    }
  },
  {
    "ph": "s", "id": 280506, "pid": 76337, "tid": -914061504, "ts": 1716454225851105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225910030, "dur": 3,
    "args": {
      "External id": 280508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280508,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280508, "pid": 5, "tid": 7, "ts": 1716454225910030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851118, "dur": 5,
    "args": {
      "External id": 280508,
      "cbid": 211, "correlation": 280508
    }
  },
  {
    "ph": "s", "id": 280508, "pid": 76337, "tid": -914061504, "ts": 1716454225851118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225910035, "dur": 53,
    "args": {
      "External id": 280533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280533,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280533, "pid": 5, "tid": 7, "ts": 1716454225910035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851200, "dur": 13,
    "args": {
      "External id": 280533,
      "cbid": 211, "correlation": 280533
    }
  },
  {
    "ph": "s", "id": 280533, "pid": 76337, "tid": -914061504, "ts": 1716454225851200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851306, "dur": 2,
    "args": {
      "External id": 280551,
      "cbid": 251, "correlation": 280551
    }
  },
  {
    "ph": "f", "id": 280551, "pid": 76337, "tid": -914061504, "ts": 1716454225851306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225910089, "dur": 87,
    "args": {
      "External id": 280553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280553,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 280553, "pid": 5, "tid": 7, "ts": 1716454225910089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851312, "dur": 15,
    "args": {
      "External id": 280553,
      "cbid": 211, "correlation": 280553
    }
  },
  {
    "ph": "s", "id": 280553, "pid": 76337, "tid": -914061504, "ts": 1716454225851312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225910178, "dur": 10,
    "args": {
      "External id": 280561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280561,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280561, "pid": 5, "tid": 7, "ts": 1716454225910178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851385, "dur": 12,
    "args": {
      "External id": 280561,
      "cbid": 211, "correlation": 280561
    }
  },
  {
    "ph": "s", "id": 280561, "pid": 76337, "tid": -914061504, "ts": 1716454225851385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225910188, "dur": 21,
    "args": {
      "External id": 280569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280569, "pid": 5, "tid": 7, "ts": 1716454225910188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851428, "dur": 10,
    "args": {
      "External id": 280569,
      "cbid": 211, "correlation": 280569
    }
  },
  {
    "ph": "s", "id": 280569, "pid": 76337, "tid": -914061504, "ts": 1716454225851428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225910211, "dur": 18,
    "args": {
      "External id": 280591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280591,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280591, "pid": 5, "tid": 7, "ts": 1716454225910211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851482, "dur": 10,
    "args": {
      "External id": 280591,
      "cbid": 211, "correlation": 280591
    }
  },
  {
    "ph": "s", "id": 280591, "pid": 76337, "tid": -914061504, "ts": 1716454225851482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851578, "dur": 1,
    "args": {
      "External id": 280607,
      "cbid": 251, "correlation": 280607
    }
  },
  {
    "ph": "f", "id": 280607, "pid": 76337, "tid": -914061504, "ts": 1716454225851578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851583, "dur": 0,
    "args": {
      "External id": 280609,
      "cbid": 251, "correlation": 280609
    }
  },
  {
    "ph": "f", "id": 280609, "pid": 76337, "tid": -914061504, "ts": 1716454225851583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225910230, "dur": 486,
    "args": {
      "External id": 280610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280610,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280610, "pid": 5, "tid": 7, "ts": 1716454225910230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851585, "dur": 14,
    "args": {
      "External id": 280610,
      "cbid": 211, "correlation": 280610
    }
  },
  {
    "ph": "s", "id": 280610, "pid": 76337, "tid": -914061504, "ts": 1716454225851585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225910717, "dur": 64,
    "args": {
      "External id": 280618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280618,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280618, "pid": 5, "tid": 7, "ts": 1716454225910717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851675, "dur": 14,
    "args": {
      "External id": 280618,
      "cbid": 211, "correlation": 280618
    }
  },
  {
    "ph": "s", "id": 280618, "pid": 76337, "tid": -914061504, "ts": 1716454225851675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225910783, "dur": 64,
    "args": {
      "External id": 280626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280626, "pid": 5, "tid": 7, "ts": 1716454225910783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851718, "dur": 11,
    "args": {
      "External id": 280626,
      "cbid": 211, "correlation": 280626
    }
  },
  {
    "ph": "s", "id": 280626, "pid": 76337, "tid": -914061504, "ts": 1716454225851718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851801, "dur": 1,
    "args": {
      "External id": 280642,
      "cbid": 251, "correlation": 280642
    }
  },
  {
    "ph": "f", "id": 280642, "pid": 76337, "tid": -914061504, "ts": 1716454225851801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454225910850, "dur": 1,
    "args": {
      "External id": 280644,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 280644,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 280644, "pid": 5, "tid": 7, "ts": 1716454225910850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851806, "dur": 17,
    "args": {
      "External id": 280644,
      "cbid": 51, "correlation": 280644
    }
  },
  {
    "ph": "s", "id": 280644, "pid": 76337, "tid": -914061504, "ts": 1716454225851806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225910853, "dur": 263,
    "args": {
      "External id": 280645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280645,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 280645, "pid": 5, "tid": 7, "ts": 1716454225910853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851825, "dur": 11,
    "args": {
      "External id": 280645,
      "cbid": 211, "correlation": 280645
    }
  },
  {
    "ph": "s", "id": 280645, "pid": 76337, "tid": -914061504, "ts": 1716454225851825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225911117, "dur": 14,
    "args": {
      "External id": 280653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280653,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280653, "pid": 5, "tid": 7, "ts": 1716454225911117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851868, "dur": 10,
    "args": {
      "External id": 280653,
      "cbid": 211, "correlation": 280653
    }
  },
  {
    "ph": "s", "id": 280653, "pid": 76337, "tid": -914061504, "ts": 1716454225851868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225911132, "dur": 36,
    "args": {
      "External id": 280664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280664,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280664, "pid": 5, "tid": 7, "ts": 1716454225911132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225851947, "dur": 13,
    "args": {
      "External id": 280664,
      "cbid": 211, "correlation": 280664
    }
  },
  {
    "ph": "s", "id": 280664, "pid": 76337, "tid": -914061504, "ts": 1716454225851947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852038, "dur": 0,
    "args": {
      "External id": 280676,
      "cbid": 317, "correlation": 280676
    }
  },
  {
    "ph": "f", "id": 280676, "pid": 76337, "tid": -914061504, "ts": 1716454225852038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852039, "dur": 1,
    "args": {
      "External id": 280677,
      "cbid": 203, "correlation": 280677
    }
  },
  {
    "ph": "f", "id": 280677, "pid": 76337, "tid": -914061504, "ts": 1716454225852039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852040, "dur": 0,
    "args": {
      "External id": 280678,
      "cbid": 205, "correlation": 280678
    }
  },
  {
    "ph": "f", "id": 280678, "pid": 76337, "tid": -914061504, "ts": 1716454225852040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225911170, "dur": 12,
    "args": {
      "External id": 280682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280682,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280682, "pid": 5, "tid": 7, "ts": 1716454225911170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852066, "dur": 14,
    "args": {
      "External id": 280682,
      "cbid": 211, "correlation": 280682
    }
  },
  {
    "ph": "s", "id": 280682, "pid": 76337, "tid": -914061504, "ts": 1716454225852066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225911184, "dur": 4,
    "args": {
      "External id": 280684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 280684, "pid": 5, "tid": 7, "ts": 1716454225911184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852088, "dur": 7,
    "args": {
      "External id": 280684,
      "cbid": 211, "correlation": 280684
    }
  },
  {
    "ph": "s", "id": 280684, "pid": 76337, "tid": -914061504, "ts": 1716454225852088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852099, "dur": 0,
    "args": {
      "External id": 280685,
      "cbid": 51, "correlation": 280685
    }
  },
  {
    "ph": "s", "id": 280685, "pid": 76337, "tid": -914061504, "ts": 1716454225852099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225911188, "dur": 93,
    "args": {
      "External id": 280686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280686,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 280686, "pid": 5, "tid": 7, "ts": 1716454225911188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852100, "dur": 7,
    "args": {
      "External id": 280686,
      "cbid": 211, "correlation": 280686
    }
  },
  {
    "ph": "s", "id": 280686, "pid": 76337, "tid": -914061504, "ts": 1716454225852100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225911283, "dur": 16,
    "args": {
      "External id": 280691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280691, "pid": 5, "tid": 7, "ts": 1716454225911283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852131, "dur": 9,
    "args": {
      "External id": 280691,
      "cbid": 211, "correlation": 280691
    }
  },
  {
    "ph": "s", "id": 280691, "pid": 76337, "tid": -914061504, "ts": 1716454225852131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225911300, "dur": 11,
    "args": {
      "External id": 280699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280699,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280699, "pid": 5, "tid": 7, "ts": 1716454225911300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852164, "dur": 8,
    "args": {
      "External id": 280699,
      "cbid": 211, "correlation": 280699
    }
  },
  {
    "ph": "s", "id": 280699, "pid": 76337, "tid": -914061504, "ts": 1716454225852164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225911312, "dur": 54,
    "args": {
      "External id": 280710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280710,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280710, "pid": 5, "tid": 7, "ts": 1716454225911312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852247, "dur": 14,
    "args": {
      "External id": 280710,
      "cbid": 211, "correlation": 280710
    }
  },
  {
    "ph": "s", "id": 280710, "pid": 76337, "tid": -914061504, "ts": 1716454225852247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852307, "dur": 0,
    "args": {
      "External id": 280720,
      "cbid": 317, "correlation": 280720
    }
  },
  {
    "ph": "f", "id": 280720, "pid": 76337, "tid": -914061504, "ts": 1716454225852307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852308, "dur": 0,
    "args": {
      "External id": 280721,
      "cbid": 203, "correlation": 280721
    }
  },
  {
    "ph": "f", "id": 280721, "pid": 76337, "tid": -914061504, "ts": 1716454225852308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852309, "dur": 0,
    "args": {
      "External id": 280722,
      "cbid": 205, "correlation": 280722
    }
  },
  {
    "ph": "f", "id": 280722, "pid": 76337, "tid": -914061504, "ts": 1716454225852309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225911367, "dur": 39,
    "args": {
      "External id": 280726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280726, "pid": 5, "tid": 7, "ts": 1716454225911367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852328, "dur": 11,
    "args": {
      "External id": 280726,
      "cbid": 211, "correlation": 280726
    }
  },
  {
    "ph": "s", "id": 280726, "pid": 76337, "tid": -914061504, "ts": 1716454225852328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225911408, "dur": 155,
    "args": {
      "External id": 280728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280728,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280728, "pid": 5, "tid": 7, "ts": 1716454225911408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852342, "dur": 5,
    "args": {
      "External id": 280728,
      "cbid": 211, "correlation": 280728
    }
  },
  {
    "ph": "s", "id": 280728, "pid": 76337, "tid": -914061504, "ts": 1716454225852342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225911564, "dur": 1944,
    "args": {
      "External id": 280730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280730,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280730, "pid": 5, "tid": 7, "ts": 1716454225911564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852358, "dur": 10,
    "args": {
      "External id": 280730,
      "cbid": 211, "correlation": 280730
    }
  },
  {
    "ph": "s", "id": 280730, "pid": 76337, "tid": -914061504, "ts": 1716454225852358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225913510, "dur": 39,
    "args": {
      "External id": 280732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280732, "pid": 5, "tid": 7, "ts": 1716454225913510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852375, "dur": 6,
    "args": {
      "External id": 280732,
      "cbid": 211, "correlation": 280732
    }
  },
  {
    "ph": "s", "id": 280732, "pid": 76337, "tid": -914061504, "ts": 1716454225852375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225913550, "dur": 58,
    "args": {
      "External id": 280738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280738, "pid": 5, "tid": 7, "ts": 1716454225913550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852414, "dur": 9,
    "args": {
      "External id": 280738,
      "cbid": 211, "correlation": 280738
    }
  },
  {
    "ph": "s", "id": 280738, "pid": 76337, "tid": -914061504, "ts": 1716454225852414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225913609, "dur": 83,
    "args": {
      "External id": 280747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280747,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280747, "pid": 5, "tid": 7, "ts": 1716454225913609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852547, "dur": 20,
    "args": {
      "External id": 280747,
      "cbid": 211, "correlation": 280747
    }
  },
  {
    "ph": "s", "id": 280747, "pid": 76337, "tid": -914061504, "ts": 1716454225852547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225913694, "dur": 71,
    "args": {
      "External id": 280767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280767,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 280767, "pid": 5, "tid": 7, "ts": 1716454225913694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852647, "dur": 14,
    "args": {
      "External id": 280767,
      "cbid": 211, "correlation": 280767
    }
  },
  {
    "ph": "s", "id": 280767, "pid": 76337, "tid": -914061504, "ts": 1716454225852647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225913767, "dur": 5,
    "args": {
      "External id": 280779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280779,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 280779, "pid": 5, "tid": 7, "ts": 1716454225913767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852672, "dur": 7,
    "args": {
      "External id": 280779,
      "cbid": 211, "correlation": 280779
    }
  },
  {
    "ph": "s", "id": 280779, "pid": 76337, "tid": -914061504, "ts": 1716454225852672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225913772, "dur": 80,
    "args": {
      "External id": 280782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280782, "pid": 5, "tid": 7, "ts": 1716454225913772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852695, "dur": 9,
    "args": {
      "External id": 280782,
      "cbid": 211, "correlation": 280782
    }
  },
  {
    "ph": "s", "id": 280782, "pid": 76337, "tid": -914061504, "ts": 1716454225852695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225913853, "dur": 52,
    "args": {
      "External id": 280791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280791,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280791, "pid": 5, "tid": 7, "ts": 1716454225913853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852746, "dur": 13,
    "args": {
      "External id": 280791,
      "cbid": 211, "correlation": 280791
    }
  },
  {
    "ph": "s", "id": 280791, "pid": 76337, "tid": -914061504, "ts": 1716454225852746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852804, "dur": 0,
    "args": {
      "External id": 280801,
      "cbid": 317, "correlation": 280801
    }
  },
  {
    "ph": "f", "id": 280801, "pid": 76337, "tid": -914061504, "ts": 1716454225852804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852804, "dur": 0,
    "args": {
      "External id": 280802,
      "cbid": 203, "correlation": 280802
    }
  },
  {
    "ph": "f", "id": 280802, "pid": 76337, "tid": -914061504, "ts": 1716454225852804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852805, "dur": 0,
    "args": {
      "External id": 280803,
      "cbid": 205, "correlation": 280803
    }
  },
  {
    "ph": "f", "id": 280803, "pid": 76337, "tid": -914061504, "ts": 1716454225852805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225913907, "dur": 58,
    "args": {
      "External id": 280807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280807,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280807, "pid": 5, "tid": 7, "ts": 1716454225913907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852823, "dur": 12,
    "args": {
      "External id": 280807,
      "cbid": 211, "correlation": 280807
    }
  },
  {
    "ph": "s", "id": 280807, "pid": 76337, "tid": -914061504, "ts": 1716454225852823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225913967, "dur": 116,
    "args": {
      "External id": 280809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280809,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280809, "pid": 5, "tid": 7, "ts": 1716454225913967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852837, "dur": 5,
    "args": {
      "External id": 280809,
      "cbid": 211, "correlation": 280809
    }
  },
  {
    "ph": "s", "id": 280809, "pid": 76337, "tid": -914061504, "ts": 1716454225852837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225914084, "dur": 1841,
    "args": {
      "External id": 280811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280811,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280811, "pid": 5, "tid": 7, "ts": 1716454225914084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852848, "dur": 7,
    "args": {
      "External id": 280811,
      "cbid": 211, "correlation": 280811
    }
  },
  {
    "ph": "s", "id": 280811, "pid": 76337, "tid": -914061504, "ts": 1716454225852848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225915926, "dur": 20,
    "args": {
      "External id": 280813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280813, "pid": 5, "tid": 7, "ts": 1716454225915926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852859, "dur": 5,
    "args": {
      "External id": 280813,
      "cbid": 211, "correlation": 280813
    }
  },
  {
    "ph": "s", "id": 280813, "pid": 76337, "tid": -914061504, "ts": 1716454225852859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225915947, "dur": 32,
    "args": {
      "External id": 280819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280819, "pid": 5, "tid": 7, "ts": 1716454225915947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852887, "dur": 8,
    "args": {
      "External id": 280819,
      "cbid": 211, "correlation": 280819
    }
  },
  {
    "ph": "s", "id": 280819, "pid": 76337, "tid": -914061504, "ts": 1716454225852887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225915980, "dur": 4,
    "args": {
      "External id": 280827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280827,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 280827, "pid": 5, "tid": 7, "ts": 1716454225915980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225852933, "dur": 9,
    "args": {
      "External id": 280827,
      "cbid": 211, "correlation": 280827
    }
  },
  {
    "ph": "s", "id": 280827, "pid": 76337, "tid": -914061504, "ts": 1716454225852933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853012, "dur": 1,
    "args": {
      "External id": 280843,
      "cbid": 251, "correlation": 280843
    }
  },
  {
    "ph": "f", "id": 280843, "pid": 76337, "tid": -914061504, "ts": 1716454225853012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853018, "dur": 0,
    "args": {
      "External id": 280845,
      "cbid": 251, "correlation": 280845
    }
  },
  {
    "ph": "f", "id": 280845, "pid": 76337, "tid": -914061504, "ts": 1716454225853018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225915986, "dur": 12,
    "args": {
      "External id": 280846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280846,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 280846, "pid": 5, "tid": 7, "ts": 1716454225915986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853019, "dur": 12,
    "args": {
      "External id": 280846,
      "cbid": 211, "correlation": 280846
    }
  },
  {
    "ph": "s", "id": 280846, "pid": 76337, "tid": -914061504, "ts": 1716454225853019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225915999, "dur": 5,
    "args": {
      "External id": 280848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280848,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 280848, "pid": 5, "tid": 7, "ts": 1716454225915999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853035, "dur": 7,
    "args": {
      "External id": 280848,
      "cbid": 211, "correlation": 280848
    }
  },
  {
    "ph": "s", "id": 280848, "pid": 76337, "tid": -914061504, "ts": 1716454225853035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225916005, "dur": 29,
    "args": {
      "External id": 280858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280858, "pid": 5, "tid": 7, "ts": 1716454225916005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853102, "dur": 12,
    "args": {
      "External id": 280858,
      "cbid": 211, "correlation": 280858
    }
  },
  {
    "ph": "s", "id": 280858, "pid": 76337, "tid": -914061504, "ts": 1716454225853102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225916036, "dur": 29,
    "args": {
      "External id": 280878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280878,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 280878, "pid": 5, "tid": 7, "ts": 1716454225916036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853167, "dur": 11,
    "args": {
      "External id": 280878,
      "cbid": 211, "correlation": 280878
    }
  },
  {
    "ph": "s", "id": 280878, "pid": 76337, "tid": -914061504, "ts": 1716454225853167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225916066, "dur": 4,
    "args": {
      "External id": 280890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280890,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 280890, "pid": 5, "tid": 7, "ts": 1716454225916066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853188, "dur": 6,
    "args": {
      "External id": 280890,
      "cbid": 211, "correlation": 280890
    }
  },
  {
    "ph": "s", "id": 280890, "pid": 76337, "tid": -914061504, "ts": 1716454225853188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225916072, "dur": 29,
    "args": {
      "External id": 280893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280893,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280893, "pid": 5, "tid": 7, "ts": 1716454225916072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853206, "dur": 6,
    "args": {
      "External id": 280893,
      "cbid": 211, "correlation": 280893
    }
  },
  {
    "ph": "s", "id": 280893, "pid": 76337, "tid": -914061504, "ts": 1716454225853206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225916102, "dur": 20,
    "args": {
      "External id": 280902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280902,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280902, "pid": 5, "tid": 7, "ts": 1716454225916102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853246, "dur": 11,
    "args": {
      "External id": 280902,
      "cbid": 211, "correlation": 280902
    }
  },
  {
    "ph": "s", "id": 280902, "pid": 76337, "tid": -914061504, "ts": 1716454225853246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853311, "dur": 0,
    "args": {
      "External id": 280912,
      "cbid": 317, "correlation": 280912
    }
  },
  {
    "ph": "f", "id": 280912, "pid": 76337, "tid": -914061504, "ts": 1716454225853311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853312, "dur": 0,
    "args": {
      "External id": 280913,
      "cbid": 203, "correlation": 280913
    }
  },
  {
    "ph": "f", "id": 280913, "pid": 76337, "tid": -914061504, "ts": 1716454225853312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853313, "dur": 0,
    "args": {
      "External id": 280914,
      "cbid": 205, "correlation": 280914
    }
  },
  {
    "ph": "f", "id": 280914, "pid": 76337, "tid": -914061504, "ts": 1716454225853313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225916124, "dur": 22,
    "args": {
      "External id": 280918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280918,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280918, "pid": 5, "tid": 7, "ts": 1716454225916124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853329, "dur": 12,
    "args": {
      "External id": 280918,
      "cbid": 211, "correlation": 280918
    }
  },
  {
    "ph": "s", "id": 280918, "pid": 76337, "tid": -914061504, "ts": 1716454225853329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225916147, "dur": 43,
    "args": {
      "External id": 280920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280920,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280920, "pid": 5, "tid": 7, "ts": 1716454225916147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853343, "dur": 5,
    "args": {
      "External id": 280920,
      "cbid": 211, "correlation": 280920
    }
  },
  {
    "ph": "s", "id": 280920, "pid": 76337, "tid": -914061504, "ts": 1716454225853343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225916192, "dur": 625,
    "args": {
      "External id": 280922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280922,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280922, "pid": 5, "tid": 7, "ts": 1716454225916192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853356, "dur": 6,
    "args": {
      "External id": 280922,
      "cbid": 211, "correlation": 280922
    }
  },
  {
    "ph": "s", "id": 280922, "pid": 76337, "tid": -914061504, "ts": 1716454225853356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225916818, "dur": 22,
    "args": {
      "External id": 280924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280924,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280924, "pid": 5, "tid": 7, "ts": 1716454225916818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853366, "dur": 5,
    "args": {
      "External id": 280924,
      "cbid": 211, "correlation": 280924
    }
  },
  {
    "ph": "s", "id": 280924, "pid": 76337, "tid": -914061504, "ts": 1716454225853366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225916841, "dur": 32,
    "args": {
      "External id": 280930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280930,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280930, "pid": 5, "tid": 7, "ts": 1716454225916841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853394, "dur": 9,
    "args": {
      "External id": 280930,
      "cbid": 211, "correlation": 280930
    }
  },
  {
    "ph": "s", "id": 280930, "pid": 76337, "tid": -914061504, "ts": 1716454225853394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853453, "dur": 0,
    "args": {
      "External id": 280940,
      "cbid": 317, "correlation": 280940
    }
  },
  {
    "ph": "f", "id": 280940, "pid": 76337, "tid": -914061504, "ts": 1716454225853453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853453, "dur": 0,
    "args": {
      "External id": 280941,
      "cbid": 203, "correlation": 280941
    }
  },
  {
    "ph": "f", "id": 280941, "pid": 76337, "tid": -914061504, "ts": 1716454225853453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853454, "dur": 0,
    "args": {
      "External id": 280942,
      "cbid": 205, "correlation": 280942
    }
  },
  {
    "ph": "f", "id": 280942, "pid": 76337, "tid": -914061504, "ts": 1716454225853454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225916874, "dur": 55,
    "args": {
      "External id": 280946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280946, "pid": 5, "tid": 7, "ts": 1716454225916874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853468, "dur": 11,
    "args": {
      "External id": 280946,
      "cbid": 211, "correlation": 280946
    }
  },
  {
    "ph": "s", "id": 280946, "pid": 76337, "tid": -914061504, "ts": 1716454225853468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225916931, "dur": 257,
    "args": {
      "External id": 280948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280948,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 280948, "pid": 5, "tid": 7, "ts": 1716454225916931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853486, "dur": 8,
    "args": {
      "External id": 280948,
      "cbid": 211, "correlation": 280948
    }
  },
  {
    "ph": "s", "id": 280948, "pid": 76337, "tid": -914061504, "ts": 1716454225853486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225917189, "dur": 20,
    "args": {
      "External id": 280950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280950,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280950, "pid": 5, "tid": 7, "ts": 1716454225917189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853498, "dur": 5,
    "args": {
      "External id": 280950,
      "cbid": 211, "correlation": 280950
    }
  },
  {
    "ph": "s", "id": 280950, "pid": 76337, "tid": -914061504, "ts": 1716454225853498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225917210, "dur": 31,
    "args": {
      "External id": 280956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280956, "pid": 5, "tid": 7, "ts": 1716454225917210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853524, "dur": 8,
    "args": {
      "External id": 280956,
      "cbid": 211, "correlation": 280956
    }
  },
  {
    "ph": "s", "id": 280956, "pid": 76337, "tid": -914061504, "ts": 1716454225853524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225917243, "dur": 27,
    "args": {
      "External id": 280964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280964,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280964, "pid": 5, "tid": 7, "ts": 1716454225917243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853553, "dur": 8,
    "args": {
      "External id": 280964,
      "cbid": 211, "correlation": 280964
    }
  },
  {
    "ph": "s", "id": 280964, "pid": 76337, "tid": -914061504, "ts": 1716454225853553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225917271, "dur": 19,
    "args": {
      "External id": 280972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280972,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 280972, "pid": 5, "tid": 7, "ts": 1716454225917271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853583, "dur": 9,
    "args": {
      "External id": 280972,
      "cbid": 211, "correlation": 280972
    }
  },
  {
    "ph": "s", "id": 280972, "pid": 76337, "tid": -914061504, "ts": 1716454225853583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225917292, "dur": 29,
    "args": {
      "External id": 280992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 280992,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 280992, "pid": 5, "tid": 7, "ts": 1716454225917292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853668, "dur": 12,
    "args": {
      "External id": 280992,
      "cbid": 211, "correlation": 280992
    }
  },
  {
    "ph": "s", "id": 280992, "pid": 76337, "tid": -914061504, "ts": 1716454225853668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225917323, "dur": 4,
    "args": {
      "External id": 281004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281004,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 281004, "pid": 5, "tid": 7, "ts": 1716454225917323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853689, "dur": 6,
    "args": {
      "External id": 281004,
      "cbid": 211, "correlation": 281004
    }
  },
  {
    "ph": "s", "id": 281004, "pid": 76337, "tid": -914061504, "ts": 1716454225853689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225917328, "dur": 31,
    "args": {
      "External id": 281007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281007,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281007, "pid": 5, "tid": 7, "ts": 1716454225917328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853708, "dur": 7,
    "args": {
      "External id": 281007,
      "cbid": 211, "correlation": 281007
    }
  },
  {
    "ph": "s", "id": 281007, "pid": 76337, "tid": -914061504, "ts": 1716454225853708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853765, "dur": 0,
    "args": {
      "External id": 281018,
      "cbid": 317, "correlation": 281018
    }
  },
  {
    "ph": "f", "id": 281018, "pid": 76337, "tid": -914061504, "ts": 1716454225853765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853766, "dur": 0,
    "args": {
      "External id": 281019,
      "cbid": 203, "correlation": 281019
    }
  },
  {
    "ph": "f", "id": 281019, "pid": 76337, "tid": -914061504, "ts": 1716454225853766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853767, "dur": 0,
    "args": {
      "External id": 281020,
      "cbid": 205, "correlation": 281020
    }
  },
  {
    "ph": "f", "id": 281020, "pid": 76337, "tid": -914061504, "ts": 1716454225853767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225917360, "dur": 22,
    "args": {
      "External id": 281024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281024,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281024, "pid": 5, "tid": 7, "ts": 1716454225917360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853781, "dur": 12,
    "args": {
      "External id": 281024,
      "cbid": 211, "correlation": 281024
    }
  },
  {
    "ph": "s", "id": 281024, "pid": 76337, "tid": -914061504, "ts": 1716454225853781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225917384, "dur": 100,
    "args": {
      "External id": 281026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281026,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281026, "pid": 5, "tid": 7, "ts": 1716454225917384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853799, "dur": 6,
    "args": {
      "External id": 281026,
      "cbid": 211, "correlation": 281026
    }
  },
  {
    "ph": "s", "id": 281026, "pid": 76337, "tid": -914061504, "ts": 1716454225853799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225917485, "dur": 20,
    "args": {
      "External id": 281028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281028, "pid": 5, "tid": 7, "ts": 1716454225917485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853810, "dur": 5,
    "args": {
      "External id": 281028,
      "cbid": 211, "correlation": 281028
    }
  },
  {
    "ph": "s", "id": 281028, "pid": 76337, "tid": -914061504, "ts": 1716454225853810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225917507, "dur": 31,
    "args": {
      "External id": 281034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281034, "pid": 5, "tid": 7, "ts": 1716454225917507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853841, "dur": 9,
    "args": {
      "External id": 281034,
      "cbid": 211, "correlation": 281034
    }
  },
  {
    "ph": "s", "id": 281034, "pid": 76337, "tid": -914061504, "ts": 1716454225853841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225917539, "dur": 179,
    "args": {
      "External id": 281043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281043,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281043, "pid": 5, "tid": 7, "ts": 1716454225917539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225853928, "dur": 16,
    "args": {
      "External id": 281043,
      "cbid": 211, "correlation": 281043
    }
  },
  {
    "ph": "s", "id": 281043, "pid": 76337, "tid": -914061504, "ts": 1716454225853928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225917719, "dur": 62,
    "args": {
      "External id": 281065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281065,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281065, "pid": 5, "tid": 7, "ts": 1716454225917719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854002, "dur": 12,
    "args": {
      "External id": 281065,
      "cbid": 211, "correlation": 281065
    }
  },
  {
    "ph": "s", "id": 281065, "pid": 76337, "tid": -914061504, "ts": 1716454225854002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854100, "dur": 1,
    "args": {
      "External id": 281076,
      "cbid": 251, "correlation": 281076
    }
  },
  {
    "ph": "f", "id": 281076, "pid": 76337, "tid": -914061504, "ts": 1716454225854100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225917782, "dur": 147,
    "args": {
      "External id": 281077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281077,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281077, "pid": 5, "tid": 7, "ts": 1716454225917782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854106, "dur": 14,
    "args": {
      "External id": 281077,
      "cbid": 211, "correlation": 281077
    }
  },
  {
    "ph": "s", "id": 281077, "pid": 76337, "tid": -914061504, "ts": 1716454225854106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854177, "dur": 1,
    "args": {
      "External id": 281088,
      "cbid": 251, "correlation": 281088
    }
  },
  {
    "ph": "f", "id": 281088, "pid": 76337, "tid": -914061504, "ts": 1716454225854177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225917931, "dur": 142,
    "args": {
      "External id": 281089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281089,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281089, "pid": 5, "tid": 7, "ts": 1716454225917931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854181, "dur": 11,
    "args": {
      "External id": 281089,
      "cbid": 211, "correlation": 281089
    }
  },
  {
    "ph": "s", "id": 281089, "pid": 76337, "tid": -914061504, "ts": 1716454225854181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854247, "dur": 1,
    "args": {
      "External id": 281100,
      "cbid": 251, "correlation": 281100
    }
  },
  {
    "ph": "f", "id": 281100, "pid": 76337, "tid": -914061504, "ts": 1716454225854247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225918074, "dur": 139,
    "args": {
      "External id": 281101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281101,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281101, "pid": 5, "tid": 7, "ts": 1716454225918074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854251, "dur": 11,
    "args": {
      "External id": 281101,
      "cbid": 211, "correlation": 281101
    }
  },
  {
    "ph": "s", "id": 281101, "pid": 76337, "tid": -914061504, "ts": 1716454225854251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225918214, "dur": 1853,
    "args": {
      "External id": 281122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281122,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 281122, "pid": 5, "tid": 7, "ts": 1716454225918214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854339, "dur": 14,
    "args": {
      "External id": 281122,
      "cbid": 211, "correlation": 281122
    }
  },
  {
    "ph": "s", "id": 281122, "pid": 76337, "tid": -914061504, "ts": 1716454225854339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854447, "dur": 2,
    "args": {
      "External id": 281140,
      "cbid": 251, "correlation": 281140
    }
  },
  {
    "ph": "f", "id": 281140, "pid": 76337, "tid": -914061504, "ts": 1716454225854447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225920068, "dur": 142,
    "args": {
      "External id": 281142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281142,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 281142, "pid": 5, "tid": 7, "ts": 1716454225920068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854454, "dur": 14,
    "args": {
      "External id": 281142,
      "cbid": 211, "correlation": 281142
    }
  },
  {
    "ph": "s", "id": 281142, "pid": 76337, "tid": -914061504, "ts": 1716454225854454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225920212, "dur": 35,
    "args": {
      "External id": 281150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281150,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281150, "pid": 5, "tid": 7, "ts": 1716454225920212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854525, "dur": 13,
    "args": {
      "External id": 281150,
      "cbid": 211, "correlation": 281150
    }
  },
  {
    "ph": "s", "id": 281150, "pid": 76337, "tid": -914061504, "ts": 1716454225854525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225920249, "dur": 50,
    "args": {
      "External id": 281158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281158,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281158, "pid": 5, "tid": 7, "ts": 1716454225920249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854565, "dur": 9,
    "args": {
      "External id": 281158,
      "cbid": 211, "correlation": 281158
    }
  },
  {
    "ph": "s", "id": 281158, "pid": 76337, "tid": -914061504, "ts": 1716454225854565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225920300, "dur": 29,
    "args": {
      "External id": 281169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281169, "pid": 5, "tid": 7, "ts": 1716454225920300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854639, "dur": 13,
    "args": {
      "External id": 281169,
      "cbid": 211, "correlation": 281169
    }
  },
  {
    "ph": "s", "id": 281169, "pid": 76337, "tid": -914061504, "ts": 1716454225854639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225920331, "dur": 33,
    "args": {
      "External id": 281191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281191,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281191, "pid": 5, "tid": 7, "ts": 1716454225920331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854671, "dur": 7,
    "args": {
      "External id": 281191,
      "cbid": 211, "correlation": 281191
    }
  },
  {
    "ph": "s", "id": 281191, "pid": 76337, "tid": -914061504, "ts": 1716454225854671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854755, "dur": 1,
    "args": {
      "External id": 281202,
      "cbid": 251, "correlation": 281202
    }
  },
  {
    "ph": "f", "id": 281202, "pid": 76337, "tid": -914061504, "ts": 1716454225854755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225920365, "dur": 86,
    "args": {
      "External id": 281203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281203,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281203, "pid": 5, "tid": 7, "ts": 1716454225920365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854761, "dur": 14,
    "args": {
      "External id": 281203,
      "cbid": 211, "correlation": 281203
    }
  },
  {
    "ph": "s", "id": 281203, "pid": 76337, "tid": -914061504, "ts": 1716454225854761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854833, "dur": 1,
    "args": {
      "External id": 281214,
      "cbid": 251, "correlation": 281214
    }
  },
  {
    "ph": "f", "id": 281214, "pid": 76337, "tid": -914061504, "ts": 1716454225854833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854837, "dur": 0,
    "args": {
      "External id": 281215,
      "cbid": 251, "correlation": 281215
    }
  },
  {
    "ph": "f", "id": 281215, "pid": 76337, "tid": -914061504, "ts": 1716454225854837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225920452, "dur": 10,
    "args": {
      "External id": 281216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281216,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 281216, "pid": 5, "tid": 7, "ts": 1716454225920452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854838, "dur": 12,
    "args": {
      "External id": 281216,
      "cbid": 211, "correlation": 281216
    }
  },
  {
    "ph": "s", "id": 281216, "pid": 76337, "tid": -914061504, "ts": 1716454225854838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225920464, "dur": 5,
    "args": {
      "External id": 281218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281218,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 281218, "pid": 5, "tid": 7, "ts": 1716454225920464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854852, "dur": 6,
    "args": {
      "External id": 281218,
      "cbid": 211, "correlation": 281218
    }
  },
  {
    "ph": "s", "id": 281218, "pid": 76337, "tid": -914061504, "ts": 1716454225854852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854910, "dur": 1,
    "args": {
      "External id": 281229,
      "cbid": 251, "correlation": 281229
    }
  },
  {
    "ph": "f", "id": 281229, "pid": 76337, "tid": -914061504, "ts": 1716454225854910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854913, "dur": 0,
    "args": {
      "External id": 281230,
      "cbid": 251, "correlation": 281230
    }
  },
  {
    "ph": "f", "id": 281230, "pid": 76337, "tid": -914061504, "ts": 1716454225854913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225920470, "dur": 7,
    "args": {
      "External id": 281231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281231,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 281231, "pid": 5, "tid": 7, "ts": 1716454225920470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854914, "dur": 13,
    "args": {
      "External id": 281231,
      "cbid": 211, "correlation": 281231
    }
  },
  {
    "ph": "s", "id": 281231, "pid": 76337, "tid": -914061504, "ts": 1716454225854914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225920478, "dur": 3,
    "args": {
      "External id": 281233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281233,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 281233, "pid": 5, "tid": 7, "ts": 1716454225920478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225854928, "dur": 5,
    "args": {
      "External id": 281233,
      "cbid": 211, "correlation": 281233
    }
  },
  {
    "ph": "s", "id": 281233, "pid": 76337, "tid": -914061504, "ts": 1716454225854928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225920483, "dur": 88,
    "args": {
      "External id": 281254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281254,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 281254, "pid": 5, "tid": 7, "ts": 1716454225920483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855009, "dur": 13,
    "args": {
      "External id": 281254,
      "cbid": 211, "correlation": 281254
    }
  },
  {
    "ph": "s", "id": 281254, "pid": 76337, "tid": -914061504, "ts": 1716454225855009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855107, "dur": 2,
    "args": {
      "External id": 281272,
      "cbid": 251, "correlation": 281272
    }
  },
  {
    "ph": "f", "id": 281272, "pid": 76337, "tid": -914061504, "ts": 1716454225855107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225920572, "dur": 94,
    "args": {
      "External id": 281274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281274,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281274, "pid": 5, "tid": 7, "ts": 1716454225920572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855113, "dur": 14,
    "args": {
      "External id": 281274,
      "cbid": 211, "correlation": 281274
    }
  },
  {
    "ph": "s", "id": 281274, "pid": 76337, "tid": -914061504, "ts": 1716454225855113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225920668, "dur": 19,
    "args": {
      "External id": 281282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281282,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281282, "pid": 5, "tid": 7, "ts": 1716454225920668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855183, "dur": 12,
    "args": {
      "External id": 281282,
      "cbid": 211, "correlation": 281282
    }
  },
  {
    "ph": "s", "id": 281282, "pid": 76337, "tid": -914061504, "ts": 1716454225855183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225920688, "dur": 37,
    "args": {
      "External id": 281290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281290, "pid": 5, "tid": 7, "ts": 1716454225920688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855224, "dur": 10,
    "args": {
      "External id": 281290,
      "cbid": 211, "correlation": 281290
    }
  },
  {
    "ph": "s", "id": 281290, "pid": 76337, "tid": -914061504, "ts": 1716454225855224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225920727, "dur": 33,
    "args": {
      "External id": 281312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281312,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281312, "pid": 5, "tid": 7, "ts": 1716454225920727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855275, "dur": 11,
    "args": {
      "External id": 281312,
      "cbid": 211, "correlation": 281312
    }
  },
  {
    "ph": "s", "id": 281312, "pid": 76337, "tid": -914061504, "ts": 1716454225855275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855369, "dur": 1,
    "args": {
      "External id": 281328,
      "cbid": 251, "correlation": 281328
    }
  },
  {
    "ph": "f", "id": 281328, "pid": 76337, "tid": -914061504, "ts": 1716454225855369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855374, "dur": 0,
    "args": {
      "External id": 281330,
      "cbid": 251, "correlation": 281330
    }
  },
  {
    "ph": "f", "id": 281330, "pid": 76337, "tid": -914061504, "ts": 1716454225855374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225920762, "dur": 523,
    "args": {
      "External id": 281331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281331,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 281331, "pid": 5, "tid": 7, "ts": 1716454225920762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855381, "dur": 13,
    "args": {
      "External id": 281331,
      "cbid": 211, "correlation": 281331
    }
  },
  {
    "ph": "s", "id": 281331, "pid": 76337, "tid": -914061504, "ts": 1716454225855381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225921286, "dur": 122,
    "args": {
      "External id": 281339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281339,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281339, "pid": 5, "tid": 7, "ts": 1716454225921286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855452, "dur": 13,
    "args": {
      "External id": 281339,
      "cbid": 211, "correlation": 281339
    }
  },
  {
    "ph": "s", "id": 281339, "pid": 76337, "tid": -914061504, "ts": 1716454225855452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225921409, "dur": 127,
    "args": {
      "External id": 281347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281347, "pid": 5, "tid": 7, "ts": 1716454225921409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855488, "dur": 8,
    "args": {
      "External id": 281347,
      "cbid": 211, "correlation": 281347
    }
  },
  {
    "ph": "s", "id": 281347, "pid": 76337, "tid": -914061504, "ts": 1716454225855488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855565, "dur": 1,
    "args": {
      "External id": 281363,
      "cbid": 251, "correlation": 281363
    }
  },
  {
    "ph": "f", "id": 281363, "pid": 76337, "tid": -914061504, "ts": 1716454225855565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225921537, "dur": 300,
    "args": {
      "External id": 281365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281365,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281365, "pid": 5, "tid": 7, "ts": 1716454225921537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855571, "dur": 13,
    "args": {
      "External id": 281365,
      "cbid": 211, "correlation": 281365
    }
  },
  {
    "ph": "s", "id": 281365, "pid": 76337, "tid": -914061504, "ts": 1716454225855571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225921838, "dur": 27,
    "args": {
      "External id": 281373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281373,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281373, "pid": 5, "tid": 7, "ts": 1716454225921838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855613, "dur": 10,
    "args": {
      "External id": 281373,
      "cbid": 211, "correlation": 281373
    }
  },
  {
    "ph": "s", "id": 281373, "pid": 76337, "tid": -914061504, "ts": 1716454225855613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225921867, "dur": 78,
    "args": {
      "External id": 281384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281384, "pid": 5, "tid": 7, "ts": 1716454225921867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855683, "dur": 12,
    "args": {
      "External id": 281384,
      "cbid": 211, "correlation": 281384
    }
  },
  {
    "ph": "s", "id": 281384, "pid": 76337, "tid": -914061504, "ts": 1716454225855683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855746, "dur": 0,
    "args": {
      "External id": 281396,
      "cbid": 317, "correlation": 281396
    }
  },
  {
    "ph": "f", "id": 281396, "pid": 76337, "tid": -914061504, "ts": 1716454225855746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855747, "dur": 0,
    "args": {
      "External id": 281397,
      "cbid": 203, "correlation": 281397
    }
  },
  {
    "ph": "f", "id": 281397, "pid": 76337, "tid": -914061504, "ts": 1716454225855747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855748, "dur": 0,
    "args": {
      "External id": 281398,
      "cbid": 205, "correlation": 281398
    }
  },
  {
    "ph": "f", "id": 281398, "pid": 76337, "tid": -914061504, "ts": 1716454225855748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225921946, "dur": 23,
    "args": {
      "External id": 281402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281402,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281402, "pid": 5, "tid": 7, "ts": 1716454225921946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855763, "dur": 12,
    "args": {
      "External id": 281402,
      "cbid": 211, "correlation": 281402
    }
  },
  {
    "ph": "s", "id": 281402, "pid": 76337, "tid": -914061504, "ts": 1716454225855763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225921970, "dur": 115,
    "args": {
      "External id": 281404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281404,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281404, "pid": 5, "tid": 7, "ts": 1716454225921970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855781, "dur": 6,
    "args": {
      "External id": 281404,
      "cbid": 211, "correlation": 281404
    }
  },
  {
    "ph": "s", "id": 281404, "pid": 76337, "tid": -914061504, "ts": 1716454225855781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225922087, "dur": 24,
    "args": {
      "External id": 281406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281406,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281406, "pid": 5, "tid": 7, "ts": 1716454225922087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855792, "dur": 5,
    "args": {
      "External id": 281406,
      "cbid": 211, "correlation": 281406
    }
  },
  {
    "ph": "s", "id": 281406, "pid": 76337, "tid": -914061504, "ts": 1716454225855792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225922112, "dur": 31,
    "args": {
      "External id": 281412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281412,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281412, "pid": 5, "tid": 7, "ts": 1716454225922112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855819, "dur": 9,
    "args": {
      "External id": 281412,
      "cbid": 211, "correlation": 281412
    }
  },
  {
    "ph": "s", "id": 281412, "pid": 76337, "tid": -914061504, "ts": 1716454225855819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225922145, "dur": 27,
    "args": {
      "External id": 281420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281420,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281420, "pid": 5, "tid": 7, "ts": 1716454225922145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855851, "dur": 8,
    "args": {
      "External id": 281420,
      "cbid": 211, "correlation": 281420
    }
  },
  {
    "ph": "s", "id": 281420, "pid": 76337, "tid": -914061504, "ts": 1716454225855851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225922173, "dur": 52,
    "args": {
      "External id": 281429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281429,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281429, "pid": 5, "tid": 7, "ts": 1716454225922173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855890, "dur": 10,
    "args": {
      "External id": 281429,
      "cbid": 211, "correlation": 281429
    }
  },
  {
    "ph": "s", "id": 281429, "pid": 76337, "tid": -914061504, "ts": 1716454225855890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225922226, "dur": 51,
    "args": {
      "External id": 281449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281449,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 281449, "pid": 5, "tid": 7, "ts": 1716454225922226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855960, "dur": 11,
    "args": {
      "External id": 281449,
      "cbid": 211, "correlation": 281449
    }
  },
  {
    "ph": "s", "id": 281449, "pid": 76337, "tid": -914061504, "ts": 1716454225855960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225922279, "dur": 4,
    "args": {
      "External id": 281461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281461,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 281461, "pid": 5, "tid": 7, "ts": 1716454225922279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225855991, "dur": 7,
    "args": {
      "External id": 281461,
      "cbid": 211, "correlation": 281461
    }
  },
  {
    "ph": "s", "id": 281461, "pid": 76337, "tid": -914061504, "ts": 1716454225855991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225922284, "dur": 55,
    "args": {
      "External id": 281464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281464, "pid": 5, "tid": 7, "ts": 1716454225922284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856010, "dur": 6,
    "args": {
      "External id": 281464,
      "cbid": 211, "correlation": 281464
    }
  },
  {
    "ph": "s", "id": 281464, "pid": 76337, "tid": -914061504, "ts": 1716454225856010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225922340, "dur": 37,
    "args": {
      "External id": 281473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281473,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281473, "pid": 5, "tid": 7, "ts": 1716454225922340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856049, "dur": 10,
    "args": {
      "External id": 281473,
      "cbid": 211, "correlation": 281473
    }
  },
  {
    "ph": "s", "id": 281473, "pid": 76337, "tid": -914061504, "ts": 1716454225856049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856103, "dur": 0,
    "args": {
      "External id": 281483,
      "cbid": 317, "correlation": 281483
    }
  },
  {
    "ph": "f", "id": 281483, "pid": 76337, "tid": -914061504, "ts": 1716454225856103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856104, "dur": 0,
    "args": {
      "External id": 281484,
      "cbid": 203, "correlation": 281484
    }
  },
  {
    "ph": "f", "id": 281484, "pid": 76337, "tid": -914061504, "ts": 1716454225856104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856105, "dur": 0,
    "args": {
      "External id": 281485,
      "cbid": 205, "correlation": 281485
    }
  },
  {
    "ph": "f", "id": 281485, "pid": 76337, "tid": -914061504, "ts": 1716454225856105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225922378, "dur": 39,
    "args": {
      "External id": 281489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281489,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281489, "pid": 5, "tid": 7, "ts": 1716454225922378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856119, "dur": 11,
    "args": {
      "External id": 281489,
      "cbid": 211, "correlation": 281489
    }
  },
  {
    "ph": "s", "id": 281489, "pid": 76337, "tid": -914061504, "ts": 1716454225856119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225922419, "dur": 80,
    "args": {
      "External id": 281491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281491,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281491, "pid": 5, "tid": 7, "ts": 1716454225922419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856133, "dur": 5,
    "args": {
      "External id": 281491,
      "cbid": 211, "correlation": 281491
    }
  },
  {
    "ph": "s", "id": 281491, "pid": 76337, "tid": -914061504, "ts": 1716454225856133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225922500, "dur": 1249,
    "args": {
      "External id": 281493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281493,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281493, "pid": 5, "tid": 7, "ts": 1716454225922500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856144, "dur": 6,
    "args": {
      "External id": 281493,
      "cbid": 211, "correlation": 281493
    }
  },
  {
    "ph": "s", "id": 281493, "pid": 76337, "tid": -914061504, "ts": 1716454225856144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225923750, "dur": 20,
    "args": {
      "External id": 281495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281495,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281495, "pid": 5, "tid": 7, "ts": 1716454225923750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856154, "dur": 5,
    "args": {
      "External id": 281495,
      "cbid": 211, "correlation": 281495
    }
  },
  {
    "ph": "s", "id": 281495, "pid": 76337, "tid": -914061504, "ts": 1716454225856154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225923771, "dur": 32,
    "args": {
      "External id": 281501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281501, "pid": 5, "tid": 7, "ts": 1716454225923771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856181, "dur": 9,
    "args": {
      "External id": 281501,
      "cbid": 211, "correlation": 281501
    }
  },
  {
    "ph": "s", "id": 281501, "pid": 76337, "tid": -914061504, "ts": 1716454225856181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225923804, "dur": 4,
    "args": {
      "External id": 281509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281509,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 281509, "pid": 5, "tid": 7, "ts": 1716454225923804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856225, "dur": 9,
    "args": {
      "External id": 281509,
      "cbid": 211, "correlation": 281509
    }
  },
  {
    "ph": "s", "id": 281509, "pid": 76337, "tid": -914061504, "ts": 1716454225856225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856291, "dur": 1,
    "args": {
      "External id": 281525,
      "cbid": 251, "correlation": 281525
    }
  },
  {
    "ph": "f", "id": 281525, "pid": 76337, "tid": -914061504, "ts": 1716454225856291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856296, "dur": 0,
    "args": {
      "External id": 281527,
      "cbid": 251, "correlation": 281527
    }
  },
  {
    "ph": "f", "id": 281527, "pid": 76337, "tid": -914061504, "ts": 1716454225856296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225923809, "dur": 11,
    "args": {
      "External id": 281528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281528,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 281528, "pid": 5, "tid": 7, "ts": 1716454225923809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856298, "dur": 11,
    "args": {
      "External id": 281528,
      "cbid": 211, "correlation": 281528
    }
  },
  {
    "ph": "s", "id": 281528, "pid": 76337, "tid": -914061504, "ts": 1716454225856298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225923822, "dur": 5,
    "args": {
      "External id": 281530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281530,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 281530, "pid": 5, "tid": 7, "ts": 1716454225923822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856311, "dur": 5,
    "args": {
      "External id": 281530,
      "cbid": 211, "correlation": 281530
    }
  },
  {
    "ph": "s", "id": 281530, "pid": 76337, "tid": -914061504, "ts": 1716454225856311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225923828, "dur": 28,
    "args": {
      "External id": 281540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281540,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281540, "pid": 5, "tid": 7, "ts": 1716454225923828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856369, "dur": 12,
    "args": {
      "External id": 281540,
      "cbid": 211, "correlation": 281540
    }
  },
  {
    "ph": "s", "id": 281540, "pid": 76337, "tid": -914061504, "ts": 1716454225856369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225923857, "dur": 29,
    "args": {
      "External id": 281560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281560,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 281560, "pid": 5, "tid": 7, "ts": 1716454225923857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856435, "dur": 11,
    "args": {
      "External id": 281560,
      "cbid": 211, "correlation": 281560
    }
  },
  {
    "ph": "s", "id": 281560, "pid": 76337, "tid": -914061504, "ts": 1716454225856435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225923888, "dur": 4,
    "args": {
      "External id": 281572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281572,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 281572, "pid": 5, "tid": 7, "ts": 1716454225923888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856456, "dur": 7,
    "args": {
      "External id": 281572,
      "cbid": 211, "correlation": 281572
    }
  },
  {
    "ph": "s", "id": 281572, "pid": 76337, "tid": -914061504, "ts": 1716454225856456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225923893, "dur": 29,
    "args": {
      "External id": 281575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281575, "pid": 5, "tid": 7, "ts": 1716454225923893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856476, "dur": 8,
    "args": {
      "External id": 281575,
      "cbid": 211, "correlation": 281575
    }
  },
  {
    "ph": "s", "id": 281575, "pid": 76337, "tid": -914061504, "ts": 1716454225856476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225923923, "dur": 21,
    "args": {
      "External id": 281584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281584,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281584, "pid": 5, "tid": 7, "ts": 1716454225923923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856519, "dur": 10,
    "args": {
      "External id": 281584,
      "cbid": 211, "correlation": 281584
    }
  },
  {
    "ph": "s", "id": 281584, "pid": 76337, "tid": -914061504, "ts": 1716454225856519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856581, "dur": 0,
    "args": {
      "External id": 281594,
      "cbid": 317, "correlation": 281594
    }
  },
  {
    "ph": "f", "id": 281594, "pid": 76337, "tid": -914061504, "ts": 1716454225856581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856581, "dur": 0,
    "args": {
      "External id": 281595,
      "cbid": 203, "correlation": 281595
    }
  },
  {
    "ph": "f", "id": 281595, "pid": 76337, "tid": -914061504, "ts": 1716454225856581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856582, "dur": 0,
    "args": {
      "External id": 281596,
      "cbid": 205, "correlation": 281596
    }
  },
  {
    "ph": "f", "id": 281596, "pid": 76337, "tid": -914061504, "ts": 1716454225856582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225923945, "dur": 22,
    "args": {
      "External id": 281600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281600,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281600, "pid": 5, "tid": 7, "ts": 1716454225923945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856597, "dur": 12,
    "args": {
      "External id": 281600,
      "cbid": 211, "correlation": 281600
    }
  },
  {
    "ph": "s", "id": 281600, "pid": 76337, "tid": -914061504, "ts": 1716454225856597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225923968, "dur": 42,
    "args": {
      "External id": 281602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281602, "pid": 5, "tid": 7, "ts": 1716454225923968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856611, "dur": 5,
    "args": {
      "External id": 281602,
      "cbid": 211, "correlation": 281602
    }
  },
  {
    "ph": "s", "id": 281602, "pid": 76337, "tid": -914061504, "ts": 1716454225856611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225924012, "dur": 626,
    "args": {
      "External id": 281604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281604,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281604, "pid": 5, "tid": 7, "ts": 1716454225924012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856622, "dur": 6,
    "args": {
      "External id": 281604,
      "cbid": 211, "correlation": 281604
    }
  },
  {
    "ph": "s", "id": 281604, "pid": 76337, "tid": -914061504, "ts": 1716454225856622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225924639, "dur": 21,
    "args": {
      "External id": 281606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281606,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281606, "pid": 5, "tid": 7, "ts": 1716454225924639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856631, "dur": 5,
    "args": {
      "External id": 281606,
      "cbid": 211, "correlation": 281606
    }
  },
  {
    "ph": "s", "id": 281606, "pid": 76337, "tid": -914061504, "ts": 1716454225856631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225924661, "dur": 32,
    "args": {
      "External id": 281612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281612, "pid": 5, "tid": 7, "ts": 1716454225924661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856659, "dur": 8,
    "args": {
      "External id": 281612,
      "cbid": 211, "correlation": 281612
    }
  },
  {
    "ph": "s", "id": 281612, "pid": 76337, "tid": -914061504, "ts": 1716454225856659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856718, "dur": 0,
    "args": {
      "External id": 281622,
      "cbid": 317, "correlation": 281622
    }
  },
  {
    "ph": "f", "id": 281622, "pid": 76337, "tid": -914061504, "ts": 1716454225856718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856718, "dur": 0,
    "args": {
      "External id": 281623,
      "cbid": 203, "correlation": 281623
    }
  },
  {
    "ph": "f", "id": 281623, "pid": 76337, "tid": -914061504, "ts": 1716454225856718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856719, "dur": 0,
    "args": {
      "External id": 281624,
      "cbid": 205, "correlation": 281624
    }
  },
  {
    "ph": "f", "id": 281624, "pid": 76337, "tid": -914061504, "ts": 1716454225856719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225924694, "dur": 37,
    "args": {
      "External id": 281628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281628,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281628, "pid": 5, "tid": 7, "ts": 1716454225924694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856732, "dur": 12,
    "args": {
      "External id": 281628,
      "cbid": 211, "correlation": 281628
    }
  },
  {
    "ph": "s", "id": 281628, "pid": 76337, "tid": -914061504, "ts": 1716454225856732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225924733, "dur": 183,
    "args": {
      "External id": 281630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281630,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281630, "pid": 5, "tid": 7, "ts": 1716454225924733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856750, "dur": 6,
    "args": {
      "External id": 281630,
      "cbid": 211, "correlation": 281630
    }
  },
  {
    "ph": "s", "id": 281630, "pid": 76337, "tid": -914061504, "ts": 1716454225856750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225924917, "dur": 21,
    "args": {
      "External id": 281632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281632, "pid": 5, "tid": 7, "ts": 1716454225924917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856759, "dur": 5,
    "args": {
      "External id": 281632,
      "cbid": 211, "correlation": 281632
    }
  },
  {
    "ph": "s", "id": 281632, "pid": 76337, "tid": -914061504, "ts": 1716454225856759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225924940, "dur": 32,
    "args": {
      "External id": 281638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281638, "pid": 5, "tid": 7, "ts": 1716454225924940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856785, "dur": 8,
    "args": {
      "External id": 281638,
      "cbid": 211, "correlation": 281638
    }
  },
  {
    "ph": "s", "id": 281638, "pid": 76337, "tid": -914061504, "ts": 1716454225856785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225924973, "dur": 27,
    "args": {
      "External id": 281646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281646,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281646, "pid": 5, "tid": 7, "ts": 1716454225924973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856814, "dur": 8,
    "args": {
      "External id": 281646,
      "cbid": 211, "correlation": 281646
    }
  },
  {
    "ph": "s", "id": 281646, "pid": 76337, "tid": -914061504, "ts": 1716454225856814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225925001, "dur": 20,
    "args": {
      "External id": 281654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281654,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281654, "pid": 5, "tid": 7, "ts": 1716454225925001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856842, "dur": 9,
    "args": {
      "External id": 281654,
      "cbid": 211, "correlation": 281654
    }
  },
  {
    "ph": "s", "id": 281654, "pid": 76337, "tid": -914061504, "ts": 1716454225856842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225925023, "dur": 29,
    "args": {
      "External id": 281674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281674,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 281674, "pid": 5, "tid": 7, "ts": 1716454225925023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856925, "dur": 12,
    "args": {
      "External id": 281674,
      "cbid": 211, "correlation": 281674
    }
  },
  {
    "ph": "s", "id": 281674, "pid": 76337, "tid": -914061504, "ts": 1716454225856925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225925053, "dur": 4,
    "args": {
      "External id": 281686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281686,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 281686, "pid": 5, "tid": 7, "ts": 1716454225925053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856947, "dur": 6,
    "args": {
      "External id": 281686,
      "cbid": 211, "correlation": 281686
    }
  },
  {
    "ph": "s", "id": 281686, "pid": 76337, "tid": -914061504, "ts": 1716454225856947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225925058, "dur": 29,
    "args": {
      "External id": 281689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281689, "pid": 5, "tid": 7, "ts": 1716454225925058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225856966, "dur": 7,
    "args": {
      "External id": 281689,
      "cbid": 211, "correlation": 281689
    }
  },
  {
    "ph": "s", "id": 281689, "pid": 76337, "tid": -914061504, "ts": 1716454225856966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857031, "dur": 0,
    "args": {
      "External id": 281700,
      "cbid": 317, "correlation": 281700
    }
  },
  {
    "ph": "f", "id": 281700, "pid": 76337, "tid": -914061504, "ts": 1716454225857031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857032, "dur": 0,
    "args": {
      "External id": 281701,
      "cbid": 203, "correlation": 281701
    }
  },
  {
    "ph": "f", "id": 281701, "pid": 76337, "tid": -914061504, "ts": 1716454225857032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857032, "dur": 0,
    "args": {
      "External id": 281702,
      "cbid": 205, "correlation": 281702
    }
  },
  {
    "ph": "f", "id": 281702, "pid": 76337, "tid": -914061504, "ts": 1716454225857032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225925088, "dur": 21,
    "args": {
      "External id": 281706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281706, "pid": 5, "tid": 7, "ts": 1716454225925088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857045, "dur": 13,
    "args": {
      "External id": 281706,
      "cbid": 211, "correlation": 281706
    }
  },
  {
    "ph": "s", "id": 281706, "pid": 76337, "tid": -914061504, "ts": 1716454225857045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225925110, "dur": 100,
    "args": {
      "External id": 281708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281708,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281708, "pid": 5, "tid": 7, "ts": 1716454225925110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857064, "dur": 6,
    "args": {
      "External id": 281708,
      "cbid": 211, "correlation": 281708
    }
  },
  {
    "ph": "s", "id": 281708, "pid": 76337, "tid": -914061504, "ts": 1716454225857064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225925211, "dur": 22,
    "args": {
      "External id": 281710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281710,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281710, "pid": 5, "tid": 7, "ts": 1716454225925211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857074, "dur": 5,
    "args": {
      "External id": 281710,
      "cbid": 211, "correlation": 281710
    }
  },
  {
    "ph": "s", "id": 281710, "pid": 76337, "tid": -914061504, "ts": 1716454225857074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225925235, "dur": 31,
    "args": {
      "External id": 281716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281716, "pid": 5, "tid": 7, "ts": 1716454225925235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857101, "dur": 9,
    "args": {
      "External id": 281716,
      "cbid": 211, "correlation": 281716
    }
  },
  {
    "ph": "s", "id": 281716, "pid": 76337, "tid": -914061504, "ts": 1716454225857101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225925267, "dur": 166,
    "args": {
      "External id": 281725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281725,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281725, "pid": 5, "tid": 7, "ts": 1716454225925267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857183, "dur": 14,
    "args": {
      "External id": 281725,
      "cbid": 211, "correlation": 281725
    }
  },
  {
    "ph": "s", "id": 281725, "pid": 76337, "tid": -914061504, "ts": 1716454225857183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225925435, "dur": 62,
    "args": {
      "External id": 281747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281747,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281747, "pid": 5, "tid": 7, "ts": 1716454225925435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857240, "dur": 10,
    "args": {
      "External id": 281747,
      "cbid": 211, "correlation": 281747
    }
  },
  {
    "ph": "s", "id": 281747, "pid": 76337, "tid": -914061504, "ts": 1716454225857240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857327, "dur": 1,
    "args": {
      "External id": 281758,
      "cbid": 251, "correlation": 281758
    }
  },
  {
    "ph": "f", "id": 281758, "pid": 76337, "tid": -914061504, "ts": 1716454225857327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225925499, "dur": 147,
    "args": {
      "External id": 281759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281759,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281759, "pid": 5, "tid": 7, "ts": 1716454225925499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857333, "dur": 13,
    "args": {
      "External id": 281759,
      "cbid": 211, "correlation": 281759
    }
  },
  {
    "ph": "s", "id": 281759, "pid": 76337, "tid": -914061504, "ts": 1716454225857333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857403, "dur": 1,
    "args": {
      "External id": 281770,
      "cbid": 251, "correlation": 281770
    }
  },
  {
    "ph": "f", "id": 281770, "pid": 76337, "tid": -914061504, "ts": 1716454225857403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225925647, "dur": 143,
    "args": {
      "External id": 281771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281771,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281771, "pid": 5, "tid": 7, "ts": 1716454225925647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857406, "dur": 12,
    "args": {
      "External id": 281771,
      "cbid": 211, "correlation": 281771
    }
  },
  {
    "ph": "s", "id": 281771, "pid": 76337, "tid": -914061504, "ts": 1716454225857406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857476, "dur": 1,
    "args": {
      "External id": 281782,
      "cbid": 251, "correlation": 281782
    }
  },
  {
    "ph": "f", "id": 281782, "pid": 76337, "tid": -914061504, "ts": 1716454225857476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225925791, "dur": 143,
    "args": {
      "External id": 281783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281783,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281783, "pid": 5, "tid": 7, "ts": 1716454225925791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857479, "dur": 12,
    "args": {
      "External id": 281783,
      "cbid": 211, "correlation": 281783
    }
  },
  {
    "ph": "s", "id": 281783, "pid": 76337, "tid": -914061504, "ts": 1716454225857479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225925935, "dur": 1849,
    "args": {
      "External id": 281804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281804,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 281804, "pid": 5, "tid": 7, "ts": 1716454225925935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857560, "dur": 13,
    "args": {
      "External id": 281804,
      "cbid": 211, "correlation": 281804
    }
  },
  {
    "ph": "s", "id": 281804, "pid": 76337, "tid": -914061504, "ts": 1716454225857560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857658, "dur": 1,
    "args": {
      "External id": 281822,
      "cbid": 251, "correlation": 281822
    }
  },
  {
    "ph": "f", "id": 281822, "pid": 76337, "tid": -914061504, "ts": 1716454225857658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225927785, "dur": 144,
    "args": {
      "External id": 281824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281824,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 281824, "pid": 5, "tid": 7, "ts": 1716454225927785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857664, "dur": 13,
    "args": {
      "External id": 281824,
      "cbid": 211, "correlation": 281824
    }
  },
  {
    "ph": "s", "id": 281824, "pid": 76337, "tid": -914061504, "ts": 1716454225857664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225927931, "dur": 35,
    "args": {
      "External id": 281832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281832,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281832, "pid": 5, "tid": 7, "ts": 1716454225927931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857734, "dur": 12,
    "args": {
      "External id": 281832,
      "cbid": 211, "correlation": 281832
    }
  },
  {
    "ph": "s", "id": 281832, "pid": 76337, "tid": -914061504, "ts": 1716454225857734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225927967, "dur": 51,
    "args": {
      "External id": 281840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281840,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281840, "pid": 5, "tid": 7, "ts": 1716454225927967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857774, "dur": 9,
    "args": {
      "External id": 281840,
      "cbid": 211, "correlation": 281840
    }
  },
  {
    "ph": "s", "id": 281840, "pid": 76337, "tid": -914061504, "ts": 1716454225857774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225928019, "dur": 29,
    "args": {
      "External id": 281851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281851,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281851, "pid": 5, "tid": 7, "ts": 1716454225928019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857844, "dur": 13,
    "args": {
      "External id": 281851,
      "cbid": 211, "correlation": 281851
    }
  },
  {
    "ph": "s", "id": 281851, "pid": 76337, "tid": -914061504, "ts": 1716454225857844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225928049, "dur": 33,
    "args": {
      "External id": 281873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281873,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281873, "pid": 5, "tid": 7, "ts": 1716454225928049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857876, "dur": 7,
    "args": {
      "External id": 281873,
      "cbid": 211, "correlation": 281873
    }
  },
  {
    "ph": "s", "id": 281873, "pid": 76337, "tid": -914061504, "ts": 1716454225857876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857960, "dur": 1,
    "args": {
      "External id": 281884,
      "cbid": 251, "correlation": 281884
    }
  },
  {
    "ph": "f", "id": 281884, "pid": 76337, "tid": -914061504, "ts": 1716454225857960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225928083, "dur": 86,
    "args": {
      "External id": 281885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281885,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281885, "pid": 5, "tid": 7, "ts": 1716454225928083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225857965, "dur": 21,
    "args": {
      "External id": 281885,
      "cbid": 211, "correlation": 281885
    }
  },
  {
    "ph": "s", "id": 281885, "pid": 76337, "tid": -914061504, "ts": 1716454225857965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858043, "dur": 1,
    "args": {
      "External id": 281896,
      "cbid": 251, "correlation": 281896
    }
  },
  {
    "ph": "f", "id": 281896, "pid": 76337, "tid": -914061504, "ts": 1716454225858043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858047, "dur": 0,
    "args": {
      "External id": 281897,
      "cbid": 251, "correlation": 281897
    }
  },
  {
    "ph": "f", "id": 281897, "pid": 76337, "tid": -914061504, "ts": 1716454225858047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225928171, "dur": 11,
    "args": {
      "External id": 281898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281898,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 281898, "pid": 5, "tid": 7, "ts": 1716454225928171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858048, "dur": 12,
    "args": {
      "External id": 281898,
      "cbid": 211, "correlation": 281898
    }
  },
  {
    "ph": "s", "id": 281898, "pid": 76337, "tid": -914061504, "ts": 1716454225858048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225928183, "dur": 5,
    "args": {
      "External id": 281900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281900,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 281900, "pid": 5, "tid": 7, "ts": 1716454225928183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858062, "dur": 6,
    "args": {
      "External id": 281900,
      "cbid": 211, "correlation": 281900
    }
  },
  {
    "ph": "s", "id": 281900, "pid": 76337, "tid": -914061504, "ts": 1716454225858062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858120, "dur": 1,
    "args": {
      "External id": 281911,
      "cbid": 251, "correlation": 281911
    }
  },
  {
    "ph": "f", "id": 281911, "pid": 76337, "tid": -914061504, "ts": 1716454225858120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858124, "dur": 0,
    "args": {
      "External id": 281912,
      "cbid": 251, "correlation": 281912
    }
  },
  {
    "ph": "f", "id": 281912, "pid": 76337, "tid": -914061504, "ts": 1716454225858124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225928190, "dur": 7,
    "args": {
      "External id": 281913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281913,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 281913, "pid": 5, "tid": 7, "ts": 1716454225928190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858125, "dur": 12,
    "args": {
      "External id": 281913,
      "cbid": 211, "correlation": 281913
    }
  },
  {
    "ph": "s", "id": 281913, "pid": 76337, "tid": -914061504, "ts": 1716454225858125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225928198, "dur": 3,
    "args": {
      "External id": 281915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281915,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 281915, "pid": 5, "tid": 7, "ts": 1716454225928198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858139, "dur": 5,
    "args": {
      "External id": 281915,
      "cbid": 211, "correlation": 281915
    }
  },
  {
    "ph": "s", "id": 281915, "pid": 76337, "tid": -914061504, "ts": 1716454225858139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225928202, "dur": 88,
    "args": {
      "External id": 281936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281936,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 281936, "pid": 5, "tid": 7, "ts": 1716454225928202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858212, "dur": 12,
    "args": {
      "External id": 281936,
      "cbid": 211, "correlation": 281936
    }
  },
  {
    "ph": "s", "id": 281936, "pid": 76337, "tid": -914061504, "ts": 1716454225858212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858308, "dur": 1,
    "args": {
      "External id": 281954,
      "cbid": 251, "correlation": 281954
    }
  },
  {
    "ph": "f", "id": 281954, "pid": 76337, "tid": -914061504, "ts": 1716454225858308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225928292, "dur": 93,
    "args": {
      "External id": 281956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281956,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 281956, "pid": 5, "tid": 7, "ts": 1716454225928292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858314, "dur": 13,
    "args": {
      "External id": 281956,
      "cbid": 211, "correlation": 281956
    }
  },
  {
    "ph": "s", "id": 281956, "pid": 76337, "tid": -914061504, "ts": 1716454225858314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225928386, "dur": 19,
    "args": {
      "External id": 281964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281964,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281964, "pid": 5, "tid": 7, "ts": 1716454225928386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858381, "dur": 13,
    "args": {
      "External id": 281964,
      "cbid": 211, "correlation": 281964
    }
  },
  {
    "ph": "s", "id": 281964, "pid": 76337, "tid": -914061504, "ts": 1716454225858381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225928407, "dur": 37,
    "args": {
      "External id": 281972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281972, "pid": 5, "tid": 7, "ts": 1716454225928407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858422, "dur": 9,
    "args": {
      "External id": 281972,
      "cbid": 211, "correlation": 281972
    }
  },
  {
    "ph": "s", "id": 281972, "pid": 76337, "tid": -914061504, "ts": 1716454225858422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225928445, "dur": 33,
    "args": {
      "External id": 281994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 281994,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 281994, "pid": 5, "tid": 7, "ts": 1716454225928445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858473, "dur": 10,
    "args": {
      "External id": 281994,
      "cbid": 211, "correlation": 281994
    }
  },
  {
    "ph": "s", "id": 281994, "pid": 76337, "tid": -914061504, "ts": 1716454225858473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858563, "dur": 1,
    "args": {
      "External id": 282010,
      "cbid": 251, "correlation": 282010
    }
  },
  {
    "ph": "f", "id": 282010, "pid": 76337, "tid": -914061504, "ts": 1716454225858563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858568, "dur": 0,
    "args": {
      "External id": 282012,
      "cbid": 251, "correlation": 282012
    }
  },
  {
    "ph": "f", "id": 282012, "pid": 76337, "tid": -914061504, "ts": 1716454225858568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225928479, "dur": 520,
    "args": {
      "External id": 282013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282013,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 282013, "pid": 5, "tid": 7, "ts": 1716454225928479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858571, "dur": 13,
    "args": {
      "External id": 282013,
      "cbid": 211, "correlation": 282013
    }
  },
  {
    "ph": "s", "id": 282013, "pid": 76337, "tid": -914061504, "ts": 1716454225858571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225929001, "dur": 121,
    "args": {
      "External id": 282021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282021,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282021, "pid": 5, "tid": 7, "ts": 1716454225929001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858635, "dur": 13,
    "args": {
      "External id": 282021,
      "cbid": 211, "correlation": 282021
    }
  },
  {
    "ph": "s", "id": 282021, "pid": 76337, "tid": -914061504, "ts": 1716454225858635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225929123, "dur": 129,
    "args": {
      "External id": 282029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282029,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282029, "pid": 5, "tid": 7, "ts": 1716454225929123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858666, "dur": 8,
    "args": {
      "External id": 282029,
      "cbid": 211, "correlation": 282029
    }
  },
  {
    "ph": "s", "id": 282029, "pid": 76337, "tid": -914061504, "ts": 1716454225858666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858743, "dur": 1,
    "args": {
      "External id": 282045,
      "cbid": 251, "correlation": 282045
    }
  },
  {
    "ph": "f", "id": 282045, "pid": 76337, "tid": -914061504, "ts": 1716454225858743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225929253, "dur": 299,
    "args": {
      "External id": 282047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282047,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282047, "pid": 5, "tid": 7, "ts": 1716454225929253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858749, "dur": 13,
    "args": {
      "External id": 282047,
      "cbid": 211, "correlation": 282047
    }
  },
  {
    "ph": "s", "id": 282047, "pid": 76337, "tid": -914061504, "ts": 1716454225858749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225929553, "dur": 27,
    "args": {
      "External id": 282055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282055,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282055, "pid": 5, "tid": 7, "ts": 1716454225929553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858791, "dur": 11,
    "args": {
      "External id": 282055,
      "cbid": 211, "correlation": 282055
    }
  },
  {
    "ph": "s", "id": 282055, "pid": 76337, "tid": -914061504, "ts": 1716454225858791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225929582, "dur": 77,
    "args": {
      "External id": 282066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282066, "pid": 5, "tid": 7, "ts": 1716454225929582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858860, "dur": 12,
    "args": {
      "External id": 282066,
      "cbid": 211, "correlation": 282066
    }
  },
  {
    "ph": "s", "id": 282066, "pid": 76337, "tid": -914061504, "ts": 1716454225858860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858923, "dur": 0,
    "args": {
      "External id": 282078,
      "cbid": 317, "correlation": 282078
    }
  },
  {
    "ph": "f", "id": 282078, "pid": 76337, "tid": -914061504, "ts": 1716454225858923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858924, "dur": 0,
    "args": {
      "External id": 282079,
      "cbid": 203, "correlation": 282079
    }
  },
  {
    "ph": "f", "id": 282079, "pid": 76337, "tid": -914061504, "ts": 1716454225858924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858925, "dur": 0,
    "args": {
      "External id": 282080,
      "cbid": 205, "correlation": 282080
    }
  },
  {
    "ph": "f", "id": 282080, "pid": 76337, "tid": -914061504, "ts": 1716454225858925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225929660, "dur": 23,
    "args": {
      "External id": 282084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282084, "pid": 5, "tid": 7, "ts": 1716454225929660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858939, "dur": 13,
    "args": {
      "External id": 282084,
      "cbid": 211, "correlation": 282084
    }
  },
  {
    "ph": "s", "id": 282084, "pid": 76337, "tid": -914061504, "ts": 1716454225858939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225929684, "dur": 116,
    "args": {
      "External id": 282086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282086,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282086, "pid": 5, "tid": 7, "ts": 1716454225929684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858959, "dur": 6,
    "args": {
      "External id": 282086,
      "cbid": 211, "correlation": 282086
    }
  },
  {
    "ph": "s", "id": 282086, "pid": 76337, "tid": -914061504, "ts": 1716454225858959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225929802, "dur": 23,
    "args": {
      "External id": 282088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282088, "pid": 5, "tid": 7, "ts": 1716454225929802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225858969, "dur": 14,
    "args": {
      "External id": 282088,
      "cbid": 211, "correlation": 282088
    }
  },
  {
    "ph": "s", "id": 282088, "pid": 76337, "tid": -914061504, "ts": 1716454225858969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225929826, "dur": 32,
    "args": {
      "External id": 282094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282094, "pid": 5, "tid": 7, "ts": 1716454225929826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859006, "dur": 9,
    "args": {
      "External id": 282094,
      "cbid": 211, "correlation": 282094
    }
  },
  {
    "ph": "s", "id": 282094, "pid": 76337, "tid": -914061504, "ts": 1716454225859006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225929859, "dur": 26,
    "args": {
      "External id": 282102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282102,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282102, "pid": 5, "tid": 7, "ts": 1716454225929859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859038, "dur": 8,
    "args": {
      "External id": 282102,
      "cbid": 211, "correlation": 282102
    }
  },
  {
    "ph": "s", "id": 282102, "pid": 76337, "tid": -914061504, "ts": 1716454225859038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225929887, "dur": 42,
    "args": {
      "External id": 282111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282111,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282111, "pid": 5, "tid": 7, "ts": 1716454225929887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859076, "dur": 10,
    "args": {
      "External id": 282111,
      "cbid": 211, "correlation": 282111
    }
  },
  {
    "ph": "s", "id": 282111, "pid": 76337, "tid": -914061504, "ts": 1716454225859076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225929931, "dur": 42,
    "args": {
      "External id": 282131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282131,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 282131, "pid": 5, "tid": 7, "ts": 1716454225929931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859148, "dur": 12,
    "args": {
      "External id": 282131,
      "cbid": 211, "correlation": 282131
    }
  },
  {
    "ph": "s", "id": 282131, "pid": 76337, "tid": -914061504, "ts": 1716454225859148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225929974, "dur": 5,
    "args": {
      "External id": 282143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282143,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 282143, "pid": 5, "tid": 7, "ts": 1716454225929974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859169, "dur": 6,
    "args": {
      "External id": 282143,
      "cbid": 211, "correlation": 282143
    }
  },
  {
    "ph": "s", "id": 282143, "pid": 76337, "tid": -914061504, "ts": 1716454225859169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225929980, "dur": 41,
    "args": {
      "External id": 282146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282146, "pid": 5, "tid": 7, "ts": 1716454225929980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859189, "dur": 6,
    "args": {
      "External id": 282146,
      "cbid": 211, "correlation": 282146
    }
  },
  {
    "ph": "s", "id": 282146, "pid": 76337, "tid": -914061504, "ts": 1716454225859189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225930023, "dur": 30,
    "args": {
      "External id": 282155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282155,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282155, "pid": 5, "tid": 7, "ts": 1716454225930023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859228, "dur": 10,
    "args": {
      "External id": 282155,
      "cbid": 211, "correlation": 282155
    }
  },
  {
    "ph": "s", "id": 282155, "pid": 76337, "tid": -914061504, "ts": 1716454225859228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859279, "dur": 0,
    "args": {
      "External id": 282165,
      "cbid": 317, "correlation": 282165
    }
  },
  {
    "ph": "f", "id": 282165, "pid": 76337, "tid": -914061504, "ts": 1716454225859279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859279, "dur": 0,
    "args": {
      "External id": 282166,
      "cbid": 203, "correlation": 282166
    }
  },
  {
    "ph": "f", "id": 282166, "pid": 76337, "tid": -914061504, "ts": 1716454225859279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859280, "dur": 0,
    "args": {
      "External id": 282167,
      "cbid": 205, "correlation": 282167
    }
  },
  {
    "ph": "f", "id": 282167, "pid": 76337, "tid": -914061504, "ts": 1716454225859280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225930054, "dur": 31,
    "args": {
      "External id": 282171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282171,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282171, "pid": 5, "tid": 7, "ts": 1716454225930054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859295, "dur": 12,
    "args": {
      "External id": 282171,
      "cbid": 211, "correlation": 282171
    }
  },
  {
    "ph": "s", "id": 282171, "pid": 76337, "tid": -914061504, "ts": 1716454225859295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225930086, "dur": 61,
    "args": {
      "External id": 282173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282173, "pid": 5, "tid": 7, "ts": 1716454225930086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859309, "dur": 6,
    "args": {
      "External id": 282173,
      "cbid": 211, "correlation": 282173
    }
  },
  {
    "ph": "s", "id": 282173, "pid": 76337, "tid": -914061504, "ts": 1716454225859309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225930149, "dur": 937,
    "args": {
      "External id": 282175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282175,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282175, "pid": 5, "tid": 7, "ts": 1716454225930149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859321, "dur": 6,
    "args": {
      "External id": 282175,
      "cbid": 211, "correlation": 282175
    }
  },
  {
    "ph": "s", "id": 282175, "pid": 76337, "tid": -914061504, "ts": 1716454225859321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225931087, "dur": 22,
    "args": {
      "External id": 282177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282177,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282177, "pid": 5, "tid": 7, "ts": 1716454225931087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859331, "dur": 5,
    "args": {
      "External id": 282177,
      "cbid": 211, "correlation": 282177
    }
  },
  {
    "ph": "s", "id": 282177, "pid": 76337, "tid": -914061504, "ts": 1716454225859331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225931110, "dur": 32,
    "args": {
      "External id": 282183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282183,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282183, "pid": 5, "tid": 7, "ts": 1716454225931110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859358, "dur": 9,
    "args": {
      "External id": 282183,
      "cbid": 211, "correlation": 282183
    }
  },
  {
    "ph": "s", "id": 282183, "pid": 76337, "tid": -914061504, "ts": 1716454225859358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225931143, "dur": 4,
    "args": {
      "External id": 282191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282191,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 282191, "pid": 5, "tid": 7, "ts": 1716454225931143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859401, "dur": 10,
    "args": {
      "External id": 282191,
      "cbid": 211, "correlation": 282191
    }
  },
  {
    "ph": "s", "id": 282191, "pid": 76337, "tid": -914061504, "ts": 1716454225859401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859468, "dur": 1,
    "args": {
      "External id": 282207,
      "cbid": 251, "correlation": 282207
    }
  },
  {
    "ph": "f", "id": 282207, "pid": 76337, "tid": -914061504, "ts": 1716454225859468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859473, "dur": 0,
    "args": {
      "External id": 282209,
      "cbid": 251, "correlation": 282209
    }
  },
  {
    "ph": "f", "id": 282209, "pid": 76337, "tid": -914061504, "ts": 1716454225859473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225931148, "dur": 11,
    "args": {
      "External id": 282210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282210,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 282210, "pid": 5, "tid": 7, "ts": 1716454225931148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859475, "dur": 11,
    "args": {
      "External id": 282210,
      "cbid": 211, "correlation": 282210
    }
  },
  {
    "ph": "s", "id": 282210, "pid": 76337, "tid": -914061504, "ts": 1716454225859475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225931161, "dur": 5,
    "args": {
      "External id": 282212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282212,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 282212, "pid": 5, "tid": 7, "ts": 1716454225931161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859488, "dur": 5,
    "args": {
      "External id": 282212,
      "cbid": 211, "correlation": 282212
    }
  },
  {
    "ph": "s", "id": 282212, "pid": 76337, "tid": -914061504, "ts": 1716454225859488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225931167, "dur": 28,
    "args": {
      "External id": 282222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282222, "pid": 5, "tid": 7, "ts": 1716454225931167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859545, "dur": 12,
    "args": {
      "External id": 282222,
      "cbid": 211, "correlation": 282222
    }
  },
  {
    "ph": "s", "id": 282222, "pid": 76337, "tid": -914061504, "ts": 1716454225859545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225931197, "dur": 30,
    "args": {
      "External id": 282242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282242,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 282242, "pid": 5, "tid": 7, "ts": 1716454225931197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859610, "dur": 11,
    "args": {
      "External id": 282242,
      "cbid": 211, "correlation": 282242
    }
  },
  {
    "ph": "s", "id": 282242, "pid": 76337, "tid": -914061504, "ts": 1716454225859610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225931228, "dur": 4,
    "args": {
      "External id": 282254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282254,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 282254, "pid": 5, "tid": 7, "ts": 1716454225931228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859631, "dur": 6,
    "args": {
      "External id": 282254,
      "cbid": 211, "correlation": 282254
    }
  },
  {
    "ph": "s", "id": 282254, "pid": 76337, "tid": -914061504, "ts": 1716454225859631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225931233, "dur": 29,
    "args": {
      "External id": 282257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282257, "pid": 5, "tid": 7, "ts": 1716454225931233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859650, "dur": 7,
    "args": {
      "External id": 282257,
      "cbid": 211, "correlation": 282257
    }
  },
  {
    "ph": "s", "id": 282257, "pid": 76337, "tid": -914061504, "ts": 1716454225859650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225931264, "dur": 20,
    "args": {
      "External id": 282266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282266,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282266, "pid": 5, "tid": 7, "ts": 1716454225931264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859690, "dur": 9,
    "args": {
      "External id": 282266,
      "cbid": 211, "correlation": 282266
    }
  },
  {
    "ph": "s", "id": 282266, "pid": 76337, "tid": -914061504, "ts": 1716454225859690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859752, "dur": 0,
    "args": {
      "External id": 282276,
      "cbid": 317, "correlation": 282276
    }
  },
  {
    "ph": "f", "id": 282276, "pid": 76337, "tid": -914061504, "ts": 1716454225859752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859753, "dur": 0,
    "args": {
      "External id": 282277,
      "cbid": 203, "correlation": 282277
    }
  },
  {
    "ph": "f", "id": 282277, "pid": 76337, "tid": -914061504, "ts": 1716454225859753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859754, "dur": 0,
    "args": {
      "External id": 282278,
      "cbid": 205, "correlation": 282278
    }
  },
  {
    "ph": "f", "id": 282278, "pid": 76337, "tid": -914061504, "ts": 1716454225859754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225931285, "dur": 22,
    "args": {
      "External id": 282282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282282,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282282, "pid": 5, "tid": 7, "ts": 1716454225931285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859768, "dur": 12,
    "args": {
      "External id": 282282,
      "cbid": 211, "correlation": 282282
    }
  },
  {
    "ph": "s", "id": 282282, "pid": 76337, "tid": -914061504, "ts": 1716454225859768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225931309, "dur": 42,
    "args": {
      "External id": 282284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282284,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282284, "pid": 5, "tid": 7, "ts": 1716454225931309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859782, "dur": 5,
    "args": {
      "External id": 282284,
      "cbid": 211, "correlation": 282284
    }
  },
  {
    "ph": "s", "id": 282284, "pid": 76337, "tid": -914061504, "ts": 1716454225859782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225931352, "dur": 627,
    "args": {
      "External id": 282286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282286,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282286, "pid": 5, "tid": 7, "ts": 1716454225931352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859794, "dur": 6,
    "args": {
      "External id": 282286,
      "cbid": 211, "correlation": 282286
    }
  },
  {
    "ph": "s", "id": 282286, "pid": 76337, "tid": -914061504, "ts": 1716454225859794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225931981, "dur": 21,
    "args": {
      "External id": 282288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282288, "pid": 5, "tid": 7, "ts": 1716454225931981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859803, "dur": 5,
    "args": {
      "External id": 282288,
      "cbid": 211, "correlation": 282288
    }
  },
  {
    "ph": "s", "id": 282288, "pid": 76337, "tid": -914061504, "ts": 1716454225859803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225932003, "dur": 32,
    "args": {
      "External id": 282294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282294, "pid": 5, "tid": 7, "ts": 1716454225932003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859831, "dur": 8,
    "args": {
      "External id": 282294,
      "cbid": 211, "correlation": 282294
    }
  },
  {
    "ph": "s", "id": 282294, "pid": 76337, "tid": -914061504, "ts": 1716454225859831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859890, "dur": 0,
    "args": {
      "External id": 282304,
      "cbid": 317, "correlation": 282304
    }
  },
  {
    "ph": "f", "id": 282304, "pid": 76337, "tid": -914061504, "ts": 1716454225859890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859890, "dur": 0,
    "args": {
      "External id": 282305,
      "cbid": 203, "correlation": 282305
    }
  },
  {
    "ph": "f", "id": 282305, "pid": 76337, "tid": -914061504, "ts": 1716454225859890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859891, "dur": 0,
    "args": {
      "External id": 282306,
      "cbid": 205, "correlation": 282306
    }
  },
  {
    "ph": "f", "id": 282306, "pid": 76337, "tid": -914061504, "ts": 1716454225859891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225932036, "dur": 29,
    "args": {
      "External id": 282310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282310, "pid": 5, "tid": 7, "ts": 1716454225932036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859905, "dur": 11,
    "args": {
      "External id": 282310,
      "cbid": 211, "correlation": 282310
    }
  },
  {
    "ph": "s", "id": 282310, "pid": 76337, "tid": -914061504, "ts": 1716454225859905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225932067, "dur": 147,
    "args": {
      "External id": 282312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282312,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282312, "pid": 5, "tid": 7, "ts": 1716454225932067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859924, "dur": 7,
    "args": {
      "External id": 282312,
      "cbid": 211, "correlation": 282312
    }
  },
  {
    "ph": "s", "id": 282312, "pid": 76337, "tid": -914061504, "ts": 1716454225859924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225932215, "dur": 21,
    "args": {
      "External id": 282314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282314,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282314, "pid": 5, "tid": 7, "ts": 1716454225932215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859934, "dur": 5,
    "args": {
      "External id": 282314,
      "cbid": 211, "correlation": 282314
    }
  },
  {
    "ph": "s", "id": 282314, "pid": 76337, "tid": -914061504, "ts": 1716454225859934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225932238, "dur": 32,
    "args": {
      "External id": 282320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282320, "pid": 5, "tid": 7, "ts": 1716454225932238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859960, "dur": 8,
    "args": {
      "External id": 282320,
      "cbid": 211, "correlation": 282320
    }
  },
  {
    "ph": "s", "id": 282320, "pid": 76337, "tid": -914061504, "ts": 1716454225859960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225932271, "dur": 28,
    "args": {
      "External id": 282328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282328,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282328, "pid": 5, "tid": 7, "ts": 1716454225932271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225859997, "dur": 8,
    "args": {
      "External id": 282328,
      "cbid": 211, "correlation": 282328
    }
  },
  {
    "ph": "s", "id": 282328, "pid": 76337, "tid": -914061504, "ts": 1716454225859997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225932300, "dur": 20,
    "args": {
      "External id": 282336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282336,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282336, "pid": 5, "tid": 7, "ts": 1716454225932300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860026, "dur": 8,
    "args": {
      "External id": 282336,
      "cbid": 211, "correlation": 282336
    }
  },
  {
    "ph": "s", "id": 282336, "pid": 76337, "tid": -914061504, "ts": 1716454225860026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225932321, "dur": 28,
    "args": {
      "External id": 282356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282356,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 282356, "pid": 5, "tid": 7, "ts": 1716454225932321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860108, "dur": 12,
    "args": {
      "External id": 282356,
      "cbid": 211, "correlation": 282356
    }
  },
  {
    "ph": "s", "id": 282356, "pid": 76337, "tid": -914061504, "ts": 1716454225860108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225932351, "dur": 5,
    "args": {
      "External id": 282368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282368,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 282368, "pid": 5, "tid": 7, "ts": 1716454225932351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860131, "dur": 6,
    "args": {
      "External id": 282368,
      "cbid": 211, "correlation": 282368
    }
  },
  {
    "ph": "s", "id": 282368, "pid": 76337, "tid": -914061504, "ts": 1716454225860131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225932357, "dur": 30,
    "args": {
      "External id": 282371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282371,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282371, "pid": 5, "tid": 7, "ts": 1716454225932357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860148, "dur": 6,
    "args": {
      "External id": 282371,
      "cbid": 211, "correlation": 282371
    }
  },
  {
    "ph": "s", "id": 282371, "pid": 76337, "tid": -914061504, "ts": 1716454225860148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860205, "dur": 0,
    "args": {
      "External id": 282382,
      "cbid": 317, "correlation": 282382
    }
  },
  {
    "ph": "f", "id": 282382, "pid": 76337, "tid": -914061504, "ts": 1716454225860205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860205, "dur": 0,
    "args": {
      "External id": 282383,
      "cbid": 203, "correlation": 282383
    }
  },
  {
    "ph": "f", "id": 282383, "pid": 76337, "tid": -914061504, "ts": 1716454225860205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860206, "dur": 0,
    "args": {
      "External id": 282384,
      "cbid": 205, "correlation": 282384
    }
  },
  {
    "ph": "f", "id": 282384, "pid": 76337, "tid": -914061504, "ts": 1716454225860206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225932388, "dur": 22,
    "args": {
      "External id": 282388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282388,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282388, "pid": 5, "tid": 7, "ts": 1716454225932388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860219, "dur": 13,
    "args": {
      "External id": 282388,
      "cbid": 211, "correlation": 282388
    }
  },
  {
    "ph": "s", "id": 282388, "pid": 76337, "tid": -914061504, "ts": 1716454225860219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225932411, "dur": 100,
    "args": {
      "External id": 282390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282390,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282390, "pid": 5, "tid": 7, "ts": 1716454225932411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860238, "dur": 6,
    "args": {
      "External id": 282390,
      "cbid": 211, "correlation": 282390
    }
  },
  {
    "ph": "s", "id": 282390, "pid": 76337, "tid": -914061504, "ts": 1716454225860238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225932513, "dur": 23,
    "args": {
      "External id": 282392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282392,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282392, "pid": 5, "tid": 7, "ts": 1716454225932513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860247, "dur": 5,
    "args": {
      "External id": 282392,
      "cbid": 211, "correlation": 282392
    }
  },
  {
    "ph": "s", "id": 282392, "pid": 76337, "tid": -914061504, "ts": 1716454225860247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225932537, "dur": 32,
    "args": {
      "External id": 282398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282398,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282398, "pid": 5, "tid": 7, "ts": 1716454225932537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860274, "dur": 9,
    "args": {
      "External id": 282398,
      "cbid": 211, "correlation": 282398
    }
  },
  {
    "ph": "s", "id": 282398, "pid": 76337, "tid": -914061504, "ts": 1716454225860274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225932570, "dur": 181,
    "args": {
      "External id": 282407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282407,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282407, "pid": 5, "tid": 7, "ts": 1716454225932570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860356, "dur": 14,
    "args": {
      "External id": 282407,
      "cbid": 211, "correlation": 282407
    }
  },
  {
    "ph": "s", "id": 282407, "pid": 76337, "tid": -914061504, "ts": 1716454225860356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225932752, "dur": 62,
    "args": {
      "External id": 282429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282429,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282429, "pid": 5, "tid": 7, "ts": 1716454225932752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860413, "dur": 10,
    "args": {
      "External id": 282429,
      "cbid": 211, "correlation": 282429
    }
  },
  {
    "ph": "s", "id": 282429, "pid": 76337, "tid": -914061504, "ts": 1716454225860413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860500, "dur": 1,
    "args": {
      "External id": 282440,
      "cbid": 251, "correlation": 282440
    }
  },
  {
    "ph": "f", "id": 282440, "pid": 76337, "tid": -914061504, "ts": 1716454225860500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225932816, "dur": 147,
    "args": {
      "External id": 282441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282441,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282441, "pid": 5, "tid": 7, "ts": 1716454225932816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860506, "dur": 12,
    "args": {
      "External id": 282441,
      "cbid": 211, "correlation": 282441
    }
  },
  {
    "ph": "s", "id": 282441, "pid": 76337, "tid": -914061504, "ts": 1716454225860506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860575, "dur": 1,
    "args": {
      "External id": 282452,
      "cbid": 251, "correlation": 282452
    }
  },
  {
    "ph": "f", "id": 282452, "pid": 76337, "tid": -914061504, "ts": 1716454225860575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225932964, "dur": 141,
    "args": {
      "External id": 282453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282453,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282453, "pid": 5, "tid": 7, "ts": 1716454225932964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860579, "dur": 12,
    "args": {
      "External id": 282453,
      "cbid": 211, "correlation": 282453
    }
  },
  {
    "ph": "s", "id": 282453, "pid": 76337, "tid": -914061504, "ts": 1716454225860579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860646, "dur": 1,
    "args": {
      "External id": 282464,
      "cbid": 251, "correlation": 282464
    }
  },
  {
    "ph": "f", "id": 282464, "pid": 76337, "tid": -914061504, "ts": 1716454225860646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225933106, "dur": 142,
    "args": {
      "External id": 282465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282465,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282465, "pid": 5, "tid": 7, "ts": 1716454225933106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860650, "dur": 11,
    "args": {
      "External id": 282465,
      "cbid": 211, "correlation": 282465
    }
  },
  {
    "ph": "s", "id": 282465, "pid": 76337, "tid": -914061504, "ts": 1716454225860650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225933250, "dur": 1855,
    "args": {
      "External id": 282486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282486,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 282486, "pid": 5, "tid": 7, "ts": 1716454225933250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860729, "dur": 12,
    "args": {
      "External id": 282486,
      "cbid": 211, "correlation": 282486
    }
  },
  {
    "ph": "s", "id": 282486, "pid": 76337, "tid": -914061504, "ts": 1716454225860729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860826, "dur": 1,
    "args": {
      "External id": 282504,
      "cbid": 251, "correlation": 282504
    }
  },
  {
    "ph": "f", "id": 282504, "pid": 76337, "tid": -914061504, "ts": 1716454225860826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225935106, "dur": 143,
    "args": {
      "External id": 282506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282506,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 282506, "pid": 5, "tid": 7, "ts": 1716454225935106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860831, "dur": 13,
    "args": {
      "External id": 282506,
      "cbid": 211, "correlation": 282506
    }
  },
  {
    "ph": "s", "id": 282506, "pid": 76337, "tid": -914061504, "ts": 1716454225860831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225935250, "dur": 35,
    "args": {
      "External id": 282514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282514,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282514, "pid": 5, "tid": 7, "ts": 1716454225935250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860900, "dur": 12,
    "args": {
      "External id": 282514,
      "cbid": 211, "correlation": 282514
    }
  },
  {
    "ph": "s", "id": 282514, "pid": 76337, "tid": -914061504, "ts": 1716454225860900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225935287, "dur": 51,
    "args": {
      "External id": 282522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282522,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282522, "pid": 5, "tid": 7, "ts": 1716454225935287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225860939, "dur": 9,
    "args": {
      "External id": 282522,
      "cbid": 211, "correlation": 282522
    }
  },
  {
    "ph": "s", "id": 282522, "pid": 76337, "tid": -914061504, "ts": 1716454225860939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225935339, "dur": 29,
    "args": {
      "External id": 282533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282533, "pid": 5, "tid": 7, "ts": 1716454225935339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861019, "dur": 13,
    "args": {
      "External id": 282533,
      "cbid": 211, "correlation": 282533
    }
  },
  {
    "ph": "s", "id": 282533, "pid": 76337, "tid": -914061504, "ts": 1716454225861019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225935370, "dur": 33,
    "args": {
      "External id": 282555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282555,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282555, "pid": 5, "tid": 7, "ts": 1716454225935370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861049, "dur": 8,
    "args": {
      "External id": 282555,
      "cbid": 211, "correlation": 282555
    }
  },
  {
    "ph": "s", "id": 282555, "pid": 76337, "tid": -914061504, "ts": 1716454225861049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861134, "dur": 1,
    "args": {
      "External id": 282566,
      "cbid": 251, "correlation": 282566
    }
  },
  {
    "ph": "f", "id": 282566, "pid": 76337, "tid": -914061504, "ts": 1716454225861134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225935403, "dur": 87,
    "args": {
      "External id": 282567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282567,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282567, "pid": 5, "tid": 7, "ts": 1716454225935403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861139, "dur": 13,
    "args": {
      "External id": 282567,
      "cbid": 211, "correlation": 282567
    }
  },
  {
    "ph": "s", "id": 282567, "pid": 76337, "tid": -914061504, "ts": 1716454225861139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861207, "dur": 1,
    "args": {
      "External id": 282578,
      "cbid": 251, "correlation": 282578
    }
  },
  {
    "ph": "f", "id": 282578, "pid": 76337, "tid": -914061504, "ts": 1716454225861207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861211, "dur": 0,
    "args": {
      "External id": 282579,
      "cbid": 251, "correlation": 282579
    }
  },
  {
    "ph": "f", "id": 282579, "pid": 76337, "tid": -914061504, "ts": 1716454225861211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225935491, "dur": 10,
    "args": {
      "External id": 282580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282580,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 282580, "pid": 5, "tid": 7, "ts": 1716454225935491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861213, "dur": 12,
    "args": {
      "External id": 282580,
      "cbid": 211, "correlation": 282580
    }
  },
  {
    "ph": "s", "id": 282580, "pid": 76337, "tid": -914061504, "ts": 1716454225861213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225935503, "dur": 5,
    "args": {
      "External id": 282582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282582,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 282582, "pid": 5, "tid": 7, "ts": 1716454225935503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861226, "dur": 6,
    "args": {
      "External id": 282582,
      "cbid": 211, "correlation": 282582
    }
  },
  {
    "ph": "s", "id": 282582, "pid": 76337, "tid": -914061504, "ts": 1716454225861226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861285, "dur": 1,
    "args": {
      "External id": 282593,
      "cbid": 251, "correlation": 282593
    }
  },
  {
    "ph": "f", "id": 282593, "pid": 76337, "tid": -914061504, "ts": 1716454225861285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861288, "dur": 0,
    "args": {
      "External id": 282594,
      "cbid": 251, "correlation": 282594
    }
  },
  {
    "ph": "f", "id": 282594, "pid": 76337, "tid": -914061504, "ts": 1716454225861288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225935509, "dur": 7,
    "args": {
      "External id": 282595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282595,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 282595, "pid": 5, "tid": 7, "ts": 1716454225935509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861290, "dur": 11,
    "args": {
      "External id": 282595,
      "cbid": 211, "correlation": 282595
    }
  },
  {
    "ph": "s", "id": 282595, "pid": 76337, "tid": -914061504, "ts": 1716454225861290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225935518, "dur": 3,
    "args": {
      "External id": 282597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282597,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 282597, "pid": 5, "tid": 7, "ts": 1716454225935518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861303, "dur": 6,
    "args": {
      "External id": 282597,
      "cbid": 211, "correlation": 282597
    }
  },
  {
    "ph": "s", "id": 282597, "pid": 76337, "tid": -914061504, "ts": 1716454225861303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225935522, "dur": 90,
    "args": {
      "External id": 282618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282618,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 282618, "pid": 5, "tid": 7, "ts": 1716454225935522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861377, "dur": 12,
    "args": {
      "External id": 282618,
      "cbid": 211, "correlation": 282618
    }
  },
  {
    "ph": "s", "id": 282618, "pid": 76337, "tid": -914061504, "ts": 1716454225861377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861474, "dur": 1,
    "args": {
      "External id": 282636,
      "cbid": 251, "correlation": 282636
    }
  },
  {
    "ph": "f", "id": 282636, "pid": 76337, "tid": -914061504, "ts": 1716454225861474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225935613, "dur": 92,
    "args": {
      "External id": 282638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282638,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282638, "pid": 5, "tid": 7, "ts": 1716454225935613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861479, "dur": 13,
    "args": {
      "External id": 282638,
      "cbid": 211, "correlation": 282638
    }
  },
  {
    "ph": "s", "id": 282638, "pid": 76337, "tid": -914061504, "ts": 1716454225861479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225935707, "dur": 19,
    "args": {
      "External id": 282646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282646,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282646, "pid": 5, "tid": 7, "ts": 1716454225935707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861548, "dur": 13,
    "args": {
      "External id": 282646,
      "cbid": 211, "correlation": 282646
    }
  },
  {
    "ph": "s", "id": 282646, "pid": 76337, "tid": -914061504, "ts": 1716454225861548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225935727, "dur": 37,
    "args": {
      "External id": 282654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282654,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282654, "pid": 5, "tid": 7, "ts": 1716454225935727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861589, "dur": 9,
    "args": {
      "External id": 282654,
      "cbid": 211, "correlation": 282654
    }
  },
  {
    "ph": "s", "id": 282654, "pid": 76337, "tid": -914061504, "ts": 1716454225861589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225935765, "dur": 33,
    "args": {
      "External id": 282676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282676,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282676, "pid": 5, "tid": 7, "ts": 1716454225935765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861639, "dur": 10,
    "args": {
      "External id": 282676,
      "cbid": 211, "correlation": 282676
    }
  },
  {
    "ph": "s", "id": 282676, "pid": 76337, "tid": -914061504, "ts": 1716454225861639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861729, "dur": 1,
    "args": {
      "External id": 282692,
      "cbid": 251, "correlation": 282692
    }
  },
  {
    "ph": "f", "id": 282692, "pid": 76337, "tid": -914061504, "ts": 1716454225861729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861734, "dur": 0,
    "args": {
      "External id": 282694,
      "cbid": 251, "correlation": 282694
    }
  },
  {
    "ph": "f", "id": 282694, "pid": 76337, "tid": -914061504, "ts": 1716454225861734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225935800, "dur": 522,
    "args": {
      "External id": 282695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282695,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 282695, "pid": 5, "tid": 7, "ts": 1716454225935800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861737, "dur": 13,
    "args": {
      "External id": 282695,
      "cbid": 211, "correlation": 282695
    }
  },
  {
    "ph": "s", "id": 282695, "pid": 76337, "tid": -914061504, "ts": 1716454225861737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225936323, "dur": 122,
    "args": {
      "External id": 282703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282703, "pid": 5, "tid": 7, "ts": 1716454225936323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861802, "dur": 13,
    "args": {
      "External id": 282703,
      "cbid": 211, "correlation": 282703
    }
  },
  {
    "ph": "s", "id": 282703, "pid": 76337, "tid": -914061504, "ts": 1716454225861802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225936446, "dur": 129,
    "args": {
      "External id": 282711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282711, "pid": 5, "tid": 7, "ts": 1716454225936446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861832, "dur": 8,
    "args": {
      "External id": 282711,
      "cbid": 211, "correlation": 282711
    }
  },
  {
    "ph": "s", "id": 282711, "pid": 76337, "tid": -914061504, "ts": 1716454225861832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861908, "dur": 1,
    "args": {
      "External id": 282727,
      "cbid": 251, "correlation": 282727
    }
  },
  {
    "ph": "f", "id": 282727, "pid": 76337, "tid": -914061504, "ts": 1716454225861908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225936576, "dur": 301,
    "args": {
      "External id": 282729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282729,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282729, "pid": 5, "tid": 7, "ts": 1716454225936576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861914, "dur": 12,
    "args": {
      "External id": 282729,
      "cbid": 211, "correlation": 282729
    }
  },
  {
    "ph": "s", "id": 282729, "pid": 76337, "tid": -914061504, "ts": 1716454225861914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225936879, "dur": 28,
    "args": {
      "External id": 282737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282737,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282737, "pid": 5, "tid": 7, "ts": 1716454225936879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225861956, "dur": 10,
    "args": {
      "External id": 282737,
      "cbid": 211, "correlation": 282737
    }
  },
  {
    "ph": "s", "id": 282737, "pid": 76337, "tid": -914061504, "ts": 1716454225861956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225936908, "dur": 78,
    "args": {
      "External id": 282748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282748,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282748, "pid": 5, "tid": 7, "ts": 1716454225936908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862031, "dur": 13,
    "args": {
      "External id": 282748,
      "cbid": 211, "correlation": 282748
    }
  },
  {
    "ph": "s", "id": 282748, "pid": 76337, "tid": -914061504, "ts": 1716454225862031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862095, "dur": 0,
    "args": {
      "External id": 282760,
      "cbid": 317, "correlation": 282760
    }
  },
  {
    "ph": "f", "id": 282760, "pid": 76337, "tid": -914061504, "ts": 1716454225862095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862096, "dur": 0,
    "args": {
      "External id": 282761,
      "cbid": 203, "correlation": 282761
    }
  },
  {
    "ph": "f", "id": 282761, "pid": 76337, "tid": -914061504, "ts": 1716454225862096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862096, "dur": 0,
    "args": {
      "External id": 282762,
      "cbid": 205, "correlation": 282762
    }
  },
  {
    "ph": "f", "id": 282762, "pid": 76337, "tid": -914061504, "ts": 1716454225862096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225936987, "dur": 22,
    "args": {
      "External id": 282766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282766, "pid": 5, "tid": 7, "ts": 1716454225936987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862111, "dur": 13,
    "args": {
      "External id": 282766,
      "cbid": 211, "correlation": 282766
    }
  },
  {
    "ph": "s", "id": 282766, "pid": 76337, "tid": -914061504, "ts": 1716454225862111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225937010, "dur": 115,
    "args": {
      "External id": 282768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282768,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282768, "pid": 5, "tid": 7, "ts": 1716454225937010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862130, "dur": 6,
    "args": {
      "External id": 282768,
      "cbid": 211, "correlation": 282768
    }
  },
  {
    "ph": "s", "id": 282768, "pid": 76337, "tid": -914061504, "ts": 1716454225862130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225937126, "dur": 23,
    "args": {
      "External id": 282770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282770, "pid": 5, "tid": 7, "ts": 1716454225937126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862140, "dur": 5,
    "args": {
      "External id": 282770,
      "cbid": 211, "correlation": 282770
    }
  },
  {
    "ph": "s", "id": 282770, "pid": 76337, "tid": -914061504, "ts": 1716454225862140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225937151, "dur": 31,
    "args": {
      "External id": 282776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282776, "pid": 5, "tid": 7, "ts": 1716454225937151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862167, "dur": 9,
    "args": {
      "External id": 282776,
      "cbid": 211, "correlation": 282776
    }
  },
  {
    "ph": "s", "id": 282776, "pid": 76337, "tid": -914061504, "ts": 1716454225862167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225937183, "dur": 28,
    "args": {
      "External id": 282784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282784,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282784, "pid": 5, "tid": 7, "ts": 1716454225937183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862199, "dur": 8,
    "args": {
      "External id": 282784,
      "cbid": 211, "correlation": 282784
    }
  },
  {
    "ph": "s", "id": 282784, "pid": 76337, "tid": -914061504, "ts": 1716454225862199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454225937212, "dur": 98,
    "args": {
      "External id": 282795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282795,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282795, "pid": 5, "tid": 7, "ts": 1716454225937212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862265, "dur": 13,
    "args": {
      "External id": 282795,
      "cbid": 211, "correlation": 282795
    }
  },
  {
    "ph": "s", "id": 282795, "pid": 76337, "tid": -914061504, "ts": 1716454225862265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862322, "dur": 0,
    "args": {
      "External id": 282805,
      "cbid": 317, "correlation": 282805
    }
  },
  {
    "ph": "f", "id": 282805, "pid": 76337, "tid": -914061504, "ts": 1716454225862322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862323, "dur": 0,
    "args": {
      "External id": 282806,
      "cbid": 203, "correlation": 282806
    }
  },
  {
    "ph": "f", "id": 282806, "pid": 76337, "tid": -914061504, "ts": 1716454225862323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862324, "dur": 0,
    "args": {
      "External id": 282807,
      "cbid": 205, "correlation": 282807
    }
  },
  {
    "ph": "f", "id": 282807, "pid": 76337, "tid": -914061504, "ts": 1716454225862324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225937311, "dur": 74,
    "args": {
      "External id": 282811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282811,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282811, "pid": 5, "tid": 7, "ts": 1716454225937311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862338, "dur": 11,
    "args": {
      "External id": 282811,
      "cbid": 211, "correlation": 282811
    }
  },
  {
    "ph": "s", "id": 282811, "pid": 76337, "tid": -914061504, "ts": 1716454225862338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225937387, "dur": 42,
    "args": {
      "External id": 282813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282813, "pid": 5, "tid": 7, "ts": 1716454225937387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862352, "dur": 5,
    "args": {
      "External id": 282813,
      "cbid": 211, "correlation": 282813
    }
  },
  {
    "ph": "s", "id": 282813, "pid": 76337, "tid": -914061504, "ts": 1716454225862352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225937430, "dur": 4,
    "args": {
      "External id": 282815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 282815, "pid": 5, "tid": 7, "ts": 1716454225937430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862363, "dur": 7,
    "args": {
      "External id": 282815,
      "cbid": 211, "correlation": 282815
    }
  },
  {
    "ph": "s", "id": 282815, "pid": 76337, "tid": -914061504, "ts": 1716454225862363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862373, "dur": 0,
    "args": {
      "External id": 282816,
      "cbid": 51, "correlation": 282816
    }
  },
  {
    "ph": "s", "id": 282816, "pid": 76337, "tid": -914061504, "ts": 1716454225862373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225937435, "dur": 2195,
    "args": {
      "External id": 282817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282817,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282817, "pid": 5, "tid": 7, "ts": 1716454225937435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862374, "dur": 6,
    "args": {
      "External id": 282817,
      "cbid": 211, "correlation": 282817
    }
  },
  {
    "ph": "s", "id": 282817, "pid": 76337, "tid": -914061504, "ts": 1716454225862374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225939631, "dur": 109,
    "args": {
      "External id": 282822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282822, "pid": 5, "tid": 7, "ts": 1716454225939631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862402, "dur": 8,
    "args": {
      "External id": 282822,
      "cbid": 211, "correlation": 282822
    }
  },
  {
    "ph": "s", "id": 282822, "pid": 76337, "tid": -914061504, "ts": 1716454225862402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225939742, "dur": 162,
    "args": {
      "External id": 282831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282831,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282831, "pid": 5, "tid": 7, "ts": 1716454225939742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862497, "dur": 13,
    "args": {
      "External id": 282831,
      "cbid": 211, "correlation": 282831
    }
  },
  {
    "ph": "s", "id": 282831, "pid": 76337, "tid": -914061504, "ts": 1716454225862497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225939905, "dur": 125,
    "args": {
      "External id": 282851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282851,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 282851, "pid": 5, "tid": 7, "ts": 1716454225939905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862568, "dur": 12,
    "args": {
      "External id": 282851,
      "cbid": 211, "correlation": 282851
    }
  },
  {
    "ph": "s", "id": 282851, "pid": 76337, "tid": -914061504, "ts": 1716454225862568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225940031, "dur": 4,
    "args": {
      "External id": 282863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282863,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 282863, "pid": 5, "tid": 7, "ts": 1716454225940031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862602, "dur": 7,
    "args": {
      "External id": 282863,
      "cbid": 211, "correlation": 282863
    }
  },
  {
    "ph": "s", "id": 282863, "pid": 76337, "tid": -914061504, "ts": 1716454225862602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225940037, "dur": 157,
    "args": {
      "External id": 282866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282866, "pid": 5, "tid": 7, "ts": 1716454225940037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862621, "dur": 7,
    "args": {
      "External id": 282866,
      "cbid": 211, "correlation": 282866
    }
  },
  {
    "ph": "s", "id": 282866, "pid": 76337, "tid": -914061504, "ts": 1716454225862621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225940195, "dur": 101,
    "args": {
      "External id": 282875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282875,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282875, "pid": 5, "tid": 7, "ts": 1716454225940195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862662, "dur": 10,
    "args": {
      "External id": 282875,
      "cbid": 211, "correlation": 282875
    }
  },
  {
    "ph": "s", "id": 282875, "pid": 76337, "tid": -914061504, "ts": 1716454225862662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862716, "dur": 0,
    "args": {
      "External id": 282885,
      "cbid": 317, "correlation": 282885
    }
  },
  {
    "ph": "f", "id": 282885, "pid": 76337, "tid": -914061504, "ts": 1716454225862716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862716, "dur": 0,
    "args": {
      "External id": 282886,
      "cbid": 203, "correlation": 282886
    }
  },
  {
    "ph": "f", "id": 282886, "pid": 76337, "tid": -914061504, "ts": 1716454225862716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862717, "dur": 0,
    "args": {
      "External id": 282887,
      "cbid": 205, "correlation": 282887
    }
  },
  {
    "ph": "f", "id": 282887, "pid": 76337, "tid": -914061504, "ts": 1716454225862717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225940297, "dur": 109,
    "args": {
      "External id": 282891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282891,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282891, "pid": 5, "tid": 7, "ts": 1716454225940297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862734, "dur": 12,
    "args": {
      "External id": 282891,
      "cbid": 211, "correlation": 282891
    }
  },
  {
    "ph": "s", "id": 282891, "pid": 76337, "tid": -914061504, "ts": 1716454225862734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225940407, "dur": 32,
    "args": {
      "External id": 282893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282893,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282893, "pid": 5, "tid": 7, "ts": 1716454225940407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862748, "dur": 5,
    "args": {
      "External id": 282893,
      "cbid": 211, "correlation": 282893
    }
  },
  {
    "ph": "s", "id": 282893, "pid": 76337, "tid": -914061504, "ts": 1716454225862748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225940441, "dur": 4,
    "args": {
      "External id": 282895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282895,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 282895, "pid": 5, "tid": 7, "ts": 1716454225940441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862758, "dur": 6,
    "args": {
      "External id": 282895,
      "cbid": 211, "correlation": 282895
    }
  },
  {
    "ph": "s", "id": 282895, "pid": 76337, "tid": -914061504, "ts": 1716454225862758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862767, "dur": 0,
    "args": {
      "External id": 282896,
      "cbid": 51, "correlation": 282896
    }
  },
  {
    "ph": "s", "id": 282896, "pid": 76337, "tid": -914061504, "ts": 1716454225862767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225940445, "dur": 1958,
    "args": {
      "External id": 282897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282897,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 282897, "pid": 5, "tid": 7, "ts": 1716454225940445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862768, "dur": 8,
    "args": {
      "External id": 282897,
      "cbid": 211, "correlation": 282897
    }
  },
  {
    "ph": "s", "id": 282897, "pid": 76337, "tid": -914061504, "ts": 1716454225862768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225942404, "dur": 58,
    "args": {
      "External id": 282902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282902, "pid": 5, "tid": 7, "ts": 1716454225942404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862803, "dur": 9,
    "args": {
      "External id": 282902,
      "cbid": 211, "correlation": 282902
    }
  },
  {
    "ph": "s", "id": 282902, "pid": 76337, "tid": -914061504, "ts": 1716454225862803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225942463, "dur": 4,
    "args": {
      "External id": 282910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282910,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 282910, "pid": 5, "tid": 7, "ts": 1716454225942463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862847, "dur": 9,
    "args": {
      "External id": 282910,
      "cbid": 211, "correlation": 282910
    }
  },
  {
    "ph": "s", "id": 282910, "pid": 76337, "tid": -914061504, "ts": 1716454225862847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862910, "dur": 1,
    "args": {
      "External id": 282926,
      "cbid": 251, "correlation": 282926
    }
  },
  {
    "ph": "f", "id": 282926, "pid": 76337, "tid": -914061504, "ts": 1716454225862910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862915, "dur": 0,
    "args": {
      "External id": 282928,
      "cbid": 251, "correlation": 282928
    }
  },
  {
    "ph": "f", "id": 282928, "pid": 76337, "tid": -914061504, "ts": 1716454225862915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225942469, "dur": 11,
    "args": {
      "External id": 282929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282929,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 282929, "pid": 5, "tid": 7, "ts": 1716454225942469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862917, "dur": 12,
    "args": {
      "External id": 282929,
      "cbid": 211, "correlation": 282929
    }
  },
  {
    "ph": "s", "id": 282929, "pid": 76337, "tid": -914061504, "ts": 1716454225862917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225942481, "dur": 5,
    "args": {
      "External id": 282931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282931,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 282931, "pid": 5, "tid": 7, "ts": 1716454225942481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862931, "dur": 6,
    "args": {
      "External id": 282931,
      "cbid": 211, "correlation": 282931
    }
  },
  {
    "ph": "s", "id": 282931, "pid": 76337, "tid": -914061504, "ts": 1716454225862931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225942487, "dur": 51,
    "args": {
      "External id": 282941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282941, "pid": 5, "tid": 7, "ts": 1716454225942487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225862995, "dur": 13,
    "args": {
      "External id": 282941,
      "cbid": 211, "correlation": 282941
    }
  },
  {
    "ph": "s", "id": 282941, "pid": 76337, "tid": -914061504, "ts": 1716454225862995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225942540, "dur": 49,
    "args": {
      "External id": 282961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282961,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 282961, "pid": 5, "tid": 7, "ts": 1716454225942540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863064, "dur": 11,
    "args": {
      "External id": 282961,
      "cbid": 211, "correlation": 282961
    }
  },
  {
    "ph": "s", "id": 282961, "pid": 76337, "tid": -914061504, "ts": 1716454225863064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225942590, "dur": 4,
    "args": {
      "External id": 282973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282973,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 282973, "pid": 5, "tid": 7, "ts": 1716454225942590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863085, "dur": 6,
    "args": {
      "External id": 282973,
      "cbid": 211, "correlation": 282973
    }
  },
  {
    "ph": "s", "id": 282973, "pid": 76337, "tid": -914061504, "ts": 1716454225863085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225942595, "dur": 54,
    "args": {
      "External id": 282976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282976,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282976, "pid": 5, "tid": 7, "ts": 1716454225942595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863104, "dur": 6,
    "args": {
      "External id": 282976,
      "cbid": 211, "correlation": 282976
    }
  },
  {
    "ph": "s", "id": 282976, "pid": 76337, "tid": -914061504, "ts": 1716454225863104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225942650, "dur": 36,
    "args": {
      "External id": 282985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 282985,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 282985, "pid": 5, "tid": 7, "ts": 1716454225942650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863144, "dur": 9,
    "args": {
      "External id": 282985,
      "cbid": 211, "correlation": 282985
    }
  },
  {
    "ph": "s", "id": 282985, "pid": 76337, "tid": -914061504, "ts": 1716454225863144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863206, "dur": 0,
    "args": {
      "External id": 282995,
      "cbid": 317, "correlation": 282995
    }
  },
  {
    "ph": "f", "id": 282995, "pid": 76337, "tid": -914061504, "ts": 1716454225863206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863207, "dur": 0,
    "args": {
      "External id": 282996,
      "cbid": 203, "correlation": 282996
    }
  },
  {
    "ph": "f", "id": 282996, "pid": 76337, "tid": -914061504, "ts": 1716454225863207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863208, "dur": 0,
    "args": {
      "External id": 282997,
      "cbid": 205, "correlation": 282997
    }
  },
  {
    "ph": "f", "id": 282997, "pid": 76337, "tid": -914061504, "ts": 1716454225863208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225942687, "dur": 39,
    "args": {
      "External id": 283001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283001,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283001, "pid": 5, "tid": 7, "ts": 1716454225942687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863226, "dur": 13,
    "args": {
      "External id": 283001,
      "cbid": 211, "correlation": 283001
    }
  },
  {
    "ph": "s", "id": 283001, "pid": 76337, "tid": -914061504, "ts": 1716454225863226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225942728, "dur": 13,
    "args": {
      "External id": 283003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283003,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283003, "pid": 5, "tid": 7, "ts": 1716454225942728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863242, "dur": 5,
    "args": {
      "External id": 283003,
      "cbid": 211, "correlation": 283003
    }
  },
  {
    "ph": "s", "id": 283003, "pid": 76337, "tid": -914061504, "ts": 1716454225863242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225942743, "dur": 3,
    "args": {
      "External id": 283005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283005, "pid": 5, "tid": 7, "ts": 1716454225942743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863252, "dur": 6,
    "args": {
      "External id": 283005,
      "cbid": 211, "correlation": 283005
    }
  },
  {
    "ph": "s", "id": 283005, "pid": 76337, "tid": -914061504, "ts": 1716454225863252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863261, "dur": 0,
    "args": {
      "External id": 283006,
      "cbid": 51, "correlation": 283006
    }
  },
  {
    "ph": "s", "id": 283006, "pid": 76337, "tid": -914061504, "ts": 1716454225863261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225942747, "dur": 676,
    "args": {
      "External id": 283007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283007,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283007, "pid": 5, "tid": 7, "ts": 1716454225942747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863262, "dur": 5,
    "args": {
      "External id": 283007,
      "cbid": 211, "correlation": 283007
    }
  },
  {
    "ph": "s", "id": 283007, "pid": 76337, "tid": -914061504, "ts": 1716454225863262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225943424, "dur": 58,
    "args": {
      "External id": 283012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283012,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283012, "pid": 5, "tid": 7, "ts": 1716454225943424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863290, "dur": 8,
    "args": {
      "External id": 283012,
      "cbid": 211, "correlation": 283012
    }
  },
  {
    "ph": "s", "id": 283012, "pid": 76337, "tid": -914061504, "ts": 1716454225863290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863346, "dur": 0,
    "args": {
      "External id": 283022,
      "cbid": 317, "correlation": 283022
    }
  },
  {
    "ph": "f", "id": 283022, "pid": 76337, "tid": -914061504, "ts": 1716454225863346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863347, "dur": 0,
    "args": {
      "External id": 283023,
      "cbid": 203, "correlation": 283023
    }
  },
  {
    "ph": "f", "id": 283023, "pid": 76337, "tid": -914061504, "ts": 1716454225863347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863348, "dur": 0,
    "args": {
      "External id": 283024,
      "cbid": 205, "correlation": 283024
    }
  },
  {
    "ph": "f", "id": 283024, "pid": 76337, "tid": -914061504, "ts": 1716454225863348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225943484, "dur": 3,
    "args": {
      "External id": 283028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283028, "pid": 5, "tid": 7, "ts": 1716454225943484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863364, "dur": 11,
    "args": {
      "External id": 283028,
      "cbid": 211, "correlation": 283028
    }
  },
  {
    "ph": "s", "id": 283028, "pid": 76337, "tid": -914061504, "ts": 1716454225863364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863379, "dur": 0,
    "args": {
      "External id": 283029,
      "cbid": 51, "correlation": 283029
    }
  },
  {
    "ph": "s", "id": 283029, "pid": 76337, "tid": -914061504, "ts": 1716454225863379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225943488, "dur": 257,
    "args": {
      "External id": 283030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283030,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283030, "pid": 5, "tid": 7, "ts": 1716454225943488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863381, "dur": 8,
    "args": {
      "External id": 283030,
      "cbid": 211, "correlation": 283030
    }
  },
  {
    "ph": "s", "id": 283030, "pid": 76337, "tid": -914061504, "ts": 1716454225863381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225943746, "dur": 57,
    "args": {
      "External id": 283035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283035,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283035, "pid": 5, "tid": 7, "ts": 1716454225943746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863408, "dur": 8,
    "args": {
      "External id": 283035,
      "cbid": 211, "correlation": 283035
    }
  },
  {
    "ph": "s", "id": 283035, "pid": 76337, "tid": -914061504, "ts": 1716454225863408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225943805, "dur": 50,
    "args": {
      "External id": 283043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283043,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283043, "pid": 5, "tid": 7, "ts": 1716454225943805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863437, "dur": 8,
    "args": {
      "External id": 283043,
      "cbid": 211, "correlation": 283043
    }
  },
  {
    "ph": "s", "id": 283043, "pid": 76337, "tid": -914061504, "ts": 1716454225863437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225943856, "dur": 35,
    "args": {
      "External id": 283051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283051,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283051, "pid": 5, "tid": 7, "ts": 1716454225943856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863465, "dur": 8,
    "args": {
      "External id": 283051,
      "cbid": 211, "correlation": 283051
    }
  },
  {
    "ph": "s", "id": 283051, "pid": 76337, "tid": -914061504, "ts": 1716454225863465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225943892, "dur": 50,
    "args": {
      "External id": 283071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283071,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 283071, "pid": 5, "tid": 7, "ts": 1716454225943892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863547, "dur": 12,
    "args": {
      "External id": 283071,
      "cbid": 211, "correlation": 283071
    }
  },
  {
    "ph": "s", "id": 283071, "pid": 76337, "tid": -914061504, "ts": 1716454225863547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225943944, "dur": 4,
    "args": {
      "External id": 283083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283083,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283083, "pid": 5, "tid": 7, "ts": 1716454225943944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863572, "dur": 7,
    "args": {
      "External id": 283083,
      "cbid": 211, "correlation": 283083
    }
  },
  {
    "ph": "s", "id": 283083, "pid": 76337, "tid": -914061504, "ts": 1716454225863572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225943949, "dur": 53,
    "args": {
      "External id": 283086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283086, "pid": 5, "tid": 7, "ts": 1716454225943949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863590, "dur": 8,
    "args": {
      "External id": 283086,
      "cbid": 211, "correlation": 283086
    }
  },
  {
    "ph": "s", "id": 283086, "pid": 76337, "tid": -914061504, "ts": 1716454225863590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863649, "dur": 0,
    "args": {
      "External id": 283097,
      "cbid": 317, "correlation": 283097
    }
  },
  {
    "ph": "f", "id": 283097, "pid": 76337, "tid": -914061504, "ts": 1716454225863649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863650, "dur": 0,
    "args": {
      "External id": 283098,
      "cbid": 203, "correlation": 283098
    }
  },
  {
    "ph": "f", "id": 283098, "pid": 76337, "tid": -914061504, "ts": 1716454225863650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863651, "dur": 0,
    "args": {
      "External id": 283099,
      "cbid": 205, "correlation": 283099
    }
  },
  {
    "ph": "f", "id": 283099, "pid": 76337, "tid": -914061504, "ts": 1716454225863651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863684, "dur": 2,
    "args": {
      "External id": 283103,
      "cbid": 251, "correlation": 283103
    }
  },
  {
    "ph": "f", "id": 283103, "pid": 76337, "tid": -914061504, "ts": 1716454225863684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863688, "dur": 1,
    "args": {
      "External id": 283104,
      "cbid": 251, "correlation": 283104
    }
  },
  {
    "ph": "f", "id": 283104, "pid": 76337, "tid": -914061504, "ts": 1716454225863688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863690, "dur": 1,
    "args": {
      "External id": 283105,
      "cbid": 251, "correlation": 283105
    }
  },
  {
    "ph": "f", "id": 283105, "pid": 76337, "tid": -914061504, "ts": 1716454225863690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863692, "dur": 1,
    "args": {
      "External id": 283106,
      "cbid": 251, "correlation": 283106
    }
  },
  {
    "ph": "f", "id": 283106, "pid": 76337, "tid": -914061504, "ts": 1716454225863692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863694, "dur": 1,
    "args": {
      "External id": 283107,
      "cbid": 251, "correlation": 283107
    }
  },
  {
    "ph": "f", "id": 283107, "pid": 76337, "tid": -914061504, "ts": 1716454225863694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863695, "dur": 1,
    "args": {
      "External id": 283108,
      "cbid": 251, "correlation": 283108
    }
  },
  {
    "ph": "f", "id": 283108, "pid": 76337, "tid": -914061504, "ts": 1716454225863695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863697, "dur": 1,
    "args": {
      "External id": 283109,
      "cbid": 251, "correlation": 283109
    }
  },
  {
    "ph": "f", "id": 283109, "pid": 76337, "tid": -914061504, "ts": 1716454225863697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863699, "dur": 1,
    "args": {
      "External id": 283110,
      "cbid": 251, "correlation": 283110
    }
  },
  {
    "ph": "f", "id": 283110, "pid": 76337, "tid": -914061504, "ts": 1716454225863699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863702, "dur": 0,
    "args": {
      "External id": 283111,
      "cbid": 251, "correlation": 283111
    }
  },
  {
    "ph": "f", "id": 283111, "pid": 76337, "tid": -914061504, "ts": 1716454225863702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225944003, "dur": 110,
    "args": {
      "External id": 283112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283112,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283112, "pid": 5, "tid": 7, "ts": 1716454225944003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863705, "dur": 13,
    "args": {
      "External id": 283112,
      "cbid": 211, "correlation": 283112
    }
  },
  {
    "ph": "s", "id": 283112, "pid": 76337, "tid": -914061504, "ts": 1716454225863705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225944114, "dur": 58,
    "args": {
      "External id": 283118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283118, "pid": 5, "tid": 7, "ts": 1716454225944114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863744, "dur": 9,
    "args": {
      "External id": 283118,
      "cbid": 211, "correlation": 283118
    }
  },
  {
    "ph": "s", "id": 283118, "pid": 76337, "tid": -914061504, "ts": 1716454225863744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225944173, "dur": 604,
    "args": {
      "External id": 283127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283127,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283127, "pid": 5, "tid": 7, "ts": 1716454225944173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863827, "dur": 14,
    "args": {
      "External id": 283127,
      "cbid": 211, "correlation": 283127
    }
  },
  {
    "ph": "s", "id": 283127, "pid": 76337, "tid": -914061504, "ts": 1716454225863827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225944779, "dur": 172,
    "args": {
      "External id": 283149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283149,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283149, "pid": 5, "tid": 7, "ts": 1716454225944779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863887, "dur": 11,
    "args": {
      "External id": 283149,
      "cbid": 211, "correlation": 283149
    }
  },
  {
    "ph": "s", "id": 283149, "pid": 76337, "tid": -914061504, "ts": 1716454225863887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863987, "dur": 1,
    "args": {
      "External id": 283160,
      "cbid": 251, "correlation": 283160
    }
  },
  {
    "ph": "f", "id": 283160, "pid": 76337, "tid": -914061504, "ts": 1716454225863987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225944952, "dur": 195,
    "args": {
      "External id": 283161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283161,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283161, "pid": 5, "tid": 7, "ts": 1716454225944952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225863993, "dur": 13,
    "args": {
      "External id": 283161,
      "cbid": 211, "correlation": 283161
    }
  },
  {
    "ph": "s", "id": 283161, "pid": 76337, "tid": -914061504, "ts": 1716454225863993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864067, "dur": 1,
    "args": {
      "External id": 283172,
      "cbid": 251, "correlation": 283172
    }
  },
  {
    "ph": "f", "id": 283172, "pid": 76337, "tid": -914061504, "ts": 1716454225864067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225945148, "dur": 181,
    "args": {
      "External id": 283173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283173,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283173, "pid": 5, "tid": 7, "ts": 1716454225945148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864070, "dur": 12,
    "args": {
      "External id": 283173,
      "cbid": 211, "correlation": 283173
    }
  },
  {
    "ph": "s", "id": 283173, "pid": 76337, "tid": -914061504, "ts": 1716454225864070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864134, "dur": 1,
    "args": {
      "External id": 283184,
      "cbid": 251, "correlation": 283184
    }
  },
  {
    "ph": "f", "id": 283184, "pid": 76337, "tid": -914061504, "ts": 1716454225864134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225945330, "dur": 178,
    "args": {
      "External id": 283185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283185,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283185, "pid": 5, "tid": 7, "ts": 1716454225945330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864138, "dur": 12,
    "args": {
      "External id": 283185,
      "cbid": 211, "correlation": 283185
    }
  },
  {
    "ph": "s", "id": 283185, "pid": 76337, "tid": -914061504, "ts": 1716454225864138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225945509, "dur": 17570,
    "args": {
      "External id": 283206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283206,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283206, "pid": 5, "tid": 7, "ts": 1716454225945509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864222, "dur": 15,
    "args": {
      "External id": 283206,
      "cbid": 211, "correlation": 283206
    }
  },
  {
    "ph": "s", "id": 283206, "pid": 76337, "tid": -914061504, "ts": 1716454225864222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864325, "dur": 1,
    "args": {
      "External id": 283224,
      "cbid": 251, "correlation": 283224
    }
  },
  {
    "ph": "f", "id": 283224, "pid": 76337, "tid": -914061504, "ts": 1716454225864325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225963080, "dur": 193,
    "args": {
      "External id": 283226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283226,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283226, "pid": 5, "tid": 7, "ts": 1716454225963080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864331, "dur": 14,
    "args": {
      "External id": 283226,
      "cbid": 211, "correlation": 283226
    }
  },
  {
    "ph": "s", "id": 283226, "pid": 76337, "tid": -914061504, "ts": 1716454225864331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225963274, "dur": 67,
    "args": {
      "External id": 283234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283234,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283234, "pid": 5, "tid": 7, "ts": 1716454225963274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864401, "dur": 12,
    "args": {
      "External id": 283234,
      "cbid": 211, "correlation": 283234
    }
  },
  {
    "ph": "s", "id": 283234, "pid": 76337, "tid": -914061504, "ts": 1716454225864401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225963342, "dur": 98,
    "args": {
      "External id": 283242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283242,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283242, "pid": 5, "tid": 7, "ts": 1716454225963342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864440, "dur": 9,
    "args": {
      "External id": 283242,
      "cbid": 211, "correlation": 283242
    }
  },
  {
    "ph": "s", "id": 283242, "pid": 76337, "tid": -914061504, "ts": 1716454225864440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225963441, "dur": 53,
    "args": {
      "External id": 283253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283253, "pid": 5, "tid": 7, "ts": 1716454225963441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864511, "dur": 12,
    "args": {
      "External id": 283253,
      "cbid": 211, "correlation": 283253
    }
  },
  {
    "ph": "s", "id": 283253, "pid": 76337, "tid": -914061504, "ts": 1716454225864511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225963496, "dur": 87,
    "args": {
      "External id": 283275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283275,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283275, "pid": 5, "tid": 7, "ts": 1716454225963496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864545, "dur": 8,
    "args": {
      "External id": 283275,
      "cbid": 211, "correlation": 283275
    }
  },
  {
    "ph": "s", "id": 283275, "pid": 76337, "tid": -914061504, "ts": 1716454225864545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864630, "dur": 1,
    "args": {
      "External id": 283286,
      "cbid": 251, "correlation": 283286
    }
  },
  {
    "ph": "f", "id": 283286, "pid": 76337, "tid": -914061504, "ts": 1716454225864630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225963584, "dur": 99,
    "args": {
      "External id": 283287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283287,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283287, "pid": 5, "tid": 7, "ts": 1716454225963584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864636, "dur": 14,
    "args": {
      "External id": 283287,
      "cbid": 211, "correlation": 283287
    }
  },
  {
    "ph": "s", "id": 283287, "pid": 76337, "tid": -914061504, "ts": 1716454225864636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864716, "dur": 1,
    "args": {
      "External id": 283298,
      "cbid": 251, "correlation": 283298
    }
  },
  {
    "ph": "f", "id": 283298, "pid": 76337, "tid": -914061504, "ts": 1716454225864716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864721, "dur": 0,
    "args": {
      "External id": 283299,
      "cbid": 251, "correlation": 283299
    }
  },
  {
    "ph": "f", "id": 283299, "pid": 76337, "tid": -914061504, "ts": 1716454225864721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225963684, "dur": 10,
    "args": {
      "External id": 283300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283300,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 283300, "pid": 5, "tid": 7, "ts": 1716454225963684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864723, "dur": 14,
    "args": {
      "External id": 283300,
      "cbid": 211, "correlation": 283300
    }
  },
  {
    "ph": "s", "id": 283300, "pid": 76337, "tid": -914061504, "ts": 1716454225864723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225963695, "dur": 5,
    "args": {
      "External id": 283302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283302,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 283302, "pid": 5, "tid": 7, "ts": 1716454225963695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864740, "dur": 9,
    "args": {
      "External id": 283302,
      "cbid": 211, "correlation": 283302
    }
  },
  {
    "ph": "s", "id": 283302, "pid": 76337, "tid": -914061504, "ts": 1716454225864740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864807, "dur": 1,
    "args": {
      "External id": 283313,
      "cbid": 251, "correlation": 283313
    }
  },
  {
    "ph": "f", "id": 283313, "pid": 76337, "tid": -914061504, "ts": 1716454225864807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864811, "dur": 0,
    "args": {
      "External id": 283314,
      "cbid": 251, "correlation": 283314
    }
  },
  {
    "ph": "f", "id": 283314, "pid": 76337, "tid": -914061504, "ts": 1716454225864811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225963701, "dur": 6,
    "args": {
      "External id": 283315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283315,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 283315, "pid": 5, "tid": 7, "ts": 1716454225963701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864812, "dur": 12,
    "args": {
      "External id": 283315,
      "cbid": 211, "correlation": 283315
    }
  },
  {
    "ph": "s", "id": 283315, "pid": 76337, "tid": -914061504, "ts": 1716454225864812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225963708, "dur": 3,
    "args": {
      "External id": 283317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283317,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 283317, "pid": 5, "tid": 7, "ts": 1716454225963708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864826, "dur": 6,
    "args": {
      "External id": 283317,
      "cbid": 211, "correlation": 283317
    }
  },
  {
    "ph": "s", "id": 283317, "pid": 76337, "tid": -914061504, "ts": 1716454225864826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225963713, "dur": 151,
    "args": {
      "External id": 283338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283338,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283338, "pid": 5, "tid": 7, "ts": 1716454225963713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225864901, "dur": 12,
    "args": {
      "External id": 283338,
      "cbid": 211, "correlation": 283338
    }
  },
  {
    "ph": "s", "id": 283338, "pid": 76337, "tid": -914061504, "ts": 1716454225864901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865007, "dur": 2,
    "args": {
      "External id": 283356,
      "cbid": 251, "correlation": 283356
    }
  },
  {
    "ph": "f", "id": 283356, "pid": 76337, "tid": -914061504, "ts": 1716454225865007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225963865, "dur": 102,
    "args": {
      "External id": 283358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283358,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283358, "pid": 5, "tid": 7, "ts": 1716454225963865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865013, "dur": 14,
    "args": {
      "External id": 283358,
      "cbid": 211, "correlation": 283358
    }
  },
  {
    "ph": "s", "id": 283358, "pid": 76337, "tid": -914061504, "ts": 1716454225865013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225963969, "dur": 35,
    "args": {
      "External id": 283366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283366,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283366, "pid": 5, "tid": 7, "ts": 1716454225963969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865083, "dur": 12,
    "args": {
      "External id": 283366,
      "cbid": 211, "correlation": 283366
    }
  },
  {
    "ph": "s", "id": 283366, "pid": 76337, "tid": -914061504, "ts": 1716454225865083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225964005, "dur": 65,
    "args": {
      "External id": 283374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283374, "pid": 5, "tid": 7, "ts": 1716454225964005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865124, "dur": 9,
    "args": {
      "External id": 283374,
      "cbid": 211, "correlation": 283374
    }
  },
  {
    "ph": "s", "id": 283374, "pid": 76337, "tid": -914061504, "ts": 1716454225865124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225964071, "dur": 87,
    "args": {
      "External id": 283396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283396,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283396, "pid": 5, "tid": 7, "ts": 1716454225964071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865175, "dur": 10,
    "args": {
      "External id": 283396,
      "cbid": 211, "correlation": 283396
    }
  },
  {
    "ph": "s", "id": 283396, "pid": 76337, "tid": -914061504, "ts": 1716454225865175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865260, "dur": 1,
    "args": {
      "External id": 283412,
      "cbid": 251, "correlation": 283412
    }
  },
  {
    "ph": "f", "id": 283412, "pid": 76337, "tid": -914061504, "ts": 1716454225865260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225964159, "dur": 558,
    "args": {
      "External id": 283414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283414,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283414, "pid": 5, "tid": 7, "ts": 1716454225964159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865266, "dur": 12,
    "args": {
      "External id": 283414,
      "cbid": 211, "correlation": 283414
    }
  },
  {
    "ph": "s", "id": 283414, "pid": 76337, "tid": -914061504, "ts": 1716454225865266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225964718, "dur": 236,
    "args": {
      "External id": 283422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283422, "pid": 5, "tid": 7, "ts": 1716454225964718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865331, "dur": 12,
    "args": {
      "External id": 283422,
      "cbid": 211, "correlation": 283422
    }
  },
  {
    "ph": "s", "id": 283422, "pid": 76337, "tid": -914061504, "ts": 1716454225865331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225964956, "dur": 251,
    "args": {
      "External id": 283430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283430, "pid": 5, "tid": 7, "ts": 1716454225964956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865363, "dur": 8,
    "args": {
      "External id": 283430,
      "cbid": 211, "correlation": 283430
    }
  },
  {
    "ph": "s", "id": 283430, "pid": 76337, "tid": -914061504, "ts": 1716454225865363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865444, "dur": 1,
    "args": {
      "External id": 283446,
      "cbid": 251, "correlation": 283446
    }
  },
  {
    "ph": "f", "id": 283446, "pid": 76337, "tid": -914061504, "ts": 1716454225865444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865449, "dur": 0,
    "args": {
      "External id": 283448,
      "cbid": 251, "correlation": 283448
    }
  },
  {
    "ph": "f", "id": 283448, "pid": 76337, "tid": -914061504, "ts": 1716454225865449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225965208, "dur": 353,
    "args": {
      "External id": 283449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283449,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 283449, "pid": 5, "tid": 7, "ts": 1716454225965208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865452, "dur": 13,
    "args": {
      "External id": 283449,
      "cbid": 211, "correlation": 283449
    }
  },
  {
    "ph": "s", "id": 283449, "pid": 76337, "tid": -914061504, "ts": 1716454225865452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225965562, "dur": 50,
    "args": {
      "External id": 283457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283457,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283457, "pid": 5, "tid": 7, "ts": 1716454225965562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865493, "dur": 11,
    "args": {
      "External id": 283457,
      "cbid": 211, "correlation": 283457
    }
  },
  {
    "ph": "s", "id": 283457, "pid": 76337, "tid": -914061504, "ts": 1716454225865493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225965614, "dur": 151,
    "args": {
      "External id": 283468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283468, "pid": 5, "tid": 7, "ts": 1716454225965614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865562, "dur": 12,
    "args": {
      "External id": 283468,
      "cbid": 211, "correlation": 283468
    }
  },
  {
    "ph": "s", "id": 283468, "pid": 76337, "tid": -914061504, "ts": 1716454225865562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865625, "dur": 0,
    "args": {
      "External id": 283480,
      "cbid": 317, "correlation": 283480
    }
  },
  {
    "ph": "f", "id": 283480, "pid": 76337, "tid": -914061504, "ts": 1716454225865625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865626, "dur": 0,
    "args": {
      "External id": 283481,
      "cbid": 203, "correlation": 283481
    }
  },
  {
    "ph": "f", "id": 283481, "pid": 76337, "tid": -914061504, "ts": 1716454225865626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865627, "dur": 0,
    "args": {
      "External id": 283482,
      "cbid": 205, "correlation": 283482
    }
  },
  {
    "ph": "f", "id": 283482, "pid": 76337, "tid": -914061504, "ts": 1716454225865627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865650, "dur": 1,
    "args": {
      "External id": 283486,
      "cbid": 251, "correlation": 283486
    }
  },
  {
    "ph": "f", "id": 283486, "pid": 76337, "tid": -914061504, "ts": 1716454225865650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865652, "dur": 1,
    "args": {
      "External id": 283487,
      "cbid": 251, "correlation": 283487
    }
  },
  {
    "ph": "f", "id": 283487, "pid": 76337, "tid": -914061504, "ts": 1716454225865652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865654, "dur": 0,
    "args": {
      "External id": 283488,
      "cbid": 251, "correlation": 283488
    }
  },
  {
    "ph": "f", "id": 283488, "pid": 76337, "tid": -914061504, "ts": 1716454225865654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865655, "dur": 0,
    "args": {
      "External id": 283489,
      "cbid": 251, "correlation": 283489
    }
  },
  {
    "ph": "f", "id": 283489, "pid": 76337, "tid": -914061504, "ts": 1716454225865655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865656, "dur": 0,
    "args": {
      "External id": 283490,
      "cbid": 251, "correlation": 283490
    }
  },
  {
    "ph": "f", "id": 283490, "pid": 76337, "tid": -914061504, "ts": 1716454225865656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865657, "dur": 0,
    "args": {
      "External id": 283491,
      "cbid": 251, "correlation": 283491
    }
  },
  {
    "ph": "f", "id": 283491, "pid": 76337, "tid": -914061504, "ts": 1716454225865657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865658, "dur": 0,
    "args": {
      "External id": 283492,
      "cbid": 251, "correlation": 283492
    }
  },
  {
    "ph": "f", "id": 283492, "pid": 76337, "tid": -914061504, "ts": 1716454225865658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865659, "dur": 0,
    "args": {
      "External id": 283493,
      "cbid": 251, "correlation": 283493
    }
  },
  {
    "ph": "f", "id": 283493, "pid": 76337, "tid": -914061504, "ts": 1716454225865659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865660, "dur": 0,
    "args": {
      "External id": 283494,
      "cbid": 251, "correlation": 283494
    }
  },
  {
    "ph": "f", "id": 283494, "pid": 76337, "tid": -914061504, "ts": 1716454225865660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225965766, "dur": 111,
    "args": {
      "External id": 283495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283495,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283495, "pid": 5, "tid": 7, "ts": 1716454225965766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865662, "dur": 12,
    "args": {
      "External id": 283495,
      "cbid": 211, "correlation": 283495
    }
  },
  {
    "ph": "s", "id": 283495, "pid": 76337, "tid": -914061504, "ts": 1716454225865662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225965878, "dur": 58,
    "args": {
      "External id": 283501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283501,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283501, "pid": 5, "tid": 7, "ts": 1716454225965878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865696, "dur": 9,
    "args": {
      "External id": 283501,
      "cbid": 211, "correlation": 283501
    }
  },
  {
    "ph": "s", "id": 283501, "pid": 76337, "tid": -914061504, "ts": 1716454225865696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225965937, "dur": 50,
    "args": {
      "External id": 283509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283509,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283509, "pid": 5, "tid": 7, "ts": 1716454225965937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865728, "dur": 8,
    "args": {
      "External id": 283509,
      "cbid": 211, "correlation": 283509
    }
  },
  {
    "ph": "s", "id": 283509, "pid": 76337, "tid": -914061504, "ts": 1716454225865728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225965989, "dur": 96,
    "args": {
      "External id": 283518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283518,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283518, "pid": 5, "tid": 7, "ts": 1716454225965989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865767, "dur": 10,
    "args": {
      "External id": 283518,
      "cbid": 211, "correlation": 283518
    }
  },
  {
    "ph": "s", "id": 283518, "pid": 76337, "tid": -914061504, "ts": 1716454225865767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225966086, "dur": 88,
    "args": {
      "External id": 283538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283538,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 283538, "pid": 5, "tid": 7, "ts": 1716454225966086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865838, "dur": 11,
    "args": {
      "External id": 283538,
      "cbid": 211, "correlation": 283538
    }
  },
  {
    "ph": "s", "id": 283538, "pid": 76337, "tid": -914061504, "ts": 1716454225865838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225966175, "dur": 4,
    "args": {
      "External id": 283550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283550,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 283550, "pid": 5, "tid": 7, "ts": 1716454225966175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865859, "dur": 7,
    "args": {
      "External id": 283550,
      "cbid": 211, "correlation": 283550
    }
  },
  {
    "ph": "s", "id": 283550, "pid": 76337, "tid": -914061504, "ts": 1716454225865859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225966181, "dur": 105,
    "args": {
      "External id": 283553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283553, "pid": 5, "tid": 7, "ts": 1716454225966181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865878, "dur": 7,
    "args": {
      "External id": 283553,
      "cbid": 211, "correlation": 283553
    }
  },
  {
    "ph": "s", "id": 283553, "pid": 76337, "tid": -914061504, "ts": 1716454225865878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225966287, "dur": 68,
    "args": {
      "External id": 283562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283562,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283562, "pid": 5, "tid": 7, "ts": 1716454225966287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865917, "dur": 9,
    "args": {
      "External id": 283562,
      "cbid": 211, "correlation": 283562
    }
  },
  {
    "ph": "s", "id": 283562, "pid": 76337, "tid": -914061504, "ts": 1716454225865917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865968, "dur": 0,
    "args": {
      "External id": 283572,
      "cbid": 317, "correlation": 283572
    }
  },
  {
    "ph": "f", "id": 283572, "pid": 76337, "tid": -914061504, "ts": 1716454225865968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865969, "dur": 0,
    "args": {
      "External id": 283573,
      "cbid": 203, "correlation": 283573
    }
  },
  {
    "ph": "f", "id": 283573, "pid": 76337, "tid": -914061504, "ts": 1716454225865969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865970, "dur": 0,
    "args": {
      "External id": 283574,
      "cbid": 205, "correlation": 283574
    }
  },
  {
    "ph": "f", "id": 283574, "pid": 76337, "tid": -914061504, "ts": 1716454225865970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225966356, "dur": 76,
    "args": {
      "External id": 283578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283578,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283578, "pid": 5, "tid": 7, "ts": 1716454225966356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225865992, "dur": 12,
    "args": {
      "External id": 283578,
      "cbid": 211, "correlation": 283578
    }
  },
  {
    "ph": "s", "id": 283578, "pid": 76337, "tid": -914061504, "ts": 1716454225865992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225966433, "dur": 23,
    "args": {
      "External id": 283580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283580,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283580, "pid": 5, "tid": 7, "ts": 1716454225966433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866006, "dur": 5,
    "args": {
      "External id": 283580,
      "cbid": 211, "correlation": 283580
    }
  },
  {
    "ph": "s", "id": 283580, "pid": 76337, "tid": -914061504, "ts": 1716454225866006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225966458, "dur": 4,
    "args": {
      "External id": 283582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283582, "pid": 5, "tid": 7, "ts": 1716454225966458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866016, "dur": 6,
    "args": {
      "External id": 283582,
      "cbid": 211, "correlation": 283582
    }
  },
  {
    "ph": "s", "id": 283582, "pid": 76337, "tid": -914061504, "ts": 1716454225866016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866025, "dur": 0,
    "args": {
      "External id": 283583,
      "cbid": 51, "correlation": 283583
    }
  },
  {
    "ph": "s", "id": 283583, "pid": 76337, "tid": -914061504, "ts": 1716454225866025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225966463, "dur": 1316,
    "args": {
      "External id": 283584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283584,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283584, "pid": 5, "tid": 7, "ts": 1716454225966463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866026, "dur": 5,
    "args": {
      "External id": 283584,
      "cbid": 211, "correlation": 283584
    }
  },
  {
    "ph": "s", "id": 283584, "pid": 76337, "tid": -914061504, "ts": 1716454225866026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225967780, "dur": 57,
    "args": {
      "External id": 283589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283589, "pid": 5, "tid": 7, "ts": 1716454225967780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866053, "dur": 8,
    "args": {
      "External id": 283589,
      "cbid": 211, "correlation": 283589
    }
  },
  {
    "ph": "s", "id": 283589, "pid": 76337, "tid": -914061504, "ts": 1716454225866053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225967839, "dur": 4,
    "args": {
      "External id": 283597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283597,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283597, "pid": 5, "tid": 7, "ts": 1716454225967839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866096, "dur": 9,
    "args": {
      "External id": 283597,
      "cbid": 211, "correlation": 283597
    }
  },
  {
    "ph": "s", "id": 283597, "pid": 76337, "tid": -914061504, "ts": 1716454225866096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866161, "dur": 2,
    "args": {
      "External id": 283613,
      "cbid": 251, "correlation": 283613
    }
  },
  {
    "ph": "f", "id": 283613, "pid": 76337, "tid": -914061504, "ts": 1716454225866161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866167, "dur": 0,
    "args": {
      "External id": 283615,
      "cbid": 251, "correlation": 283615
    }
  },
  {
    "ph": "f", "id": 283615, "pid": 76337, "tid": -914061504, "ts": 1716454225866167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225967844, "dur": 11,
    "args": {
      "External id": 283616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283616,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 283616, "pid": 5, "tid": 7, "ts": 1716454225967844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866170, "dur": 11,
    "args": {
      "External id": 283616,
      "cbid": 211, "correlation": 283616
    }
  },
  {
    "ph": "s", "id": 283616, "pid": 76337, "tid": -914061504, "ts": 1716454225866170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225967857, "dur": 5,
    "args": {
      "External id": 283618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283618,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 283618, "pid": 5, "tid": 7, "ts": 1716454225967857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866183, "dur": 5,
    "args": {
      "External id": 283618,
      "cbid": 211, "correlation": 283618
    }
  },
  {
    "ph": "s", "id": 283618, "pid": 76337, "tid": -914061504, "ts": 1716454225866183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225967863, "dur": 53,
    "args": {
      "External id": 283628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283628, "pid": 5, "tid": 7, "ts": 1716454225967863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866239, "dur": 12,
    "args": {
      "External id": 283628,
      "cbid": 211, "correlation": 283628
    }
  },
  {
    "ph": "s", "id": 283628, "pid": 76337, "tid": -914061504, "ts": 1716454225866239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225967918, "dur": 49,
    "args": {
      "External id": 283648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283648,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 283648, "pid": 5, "tid": 7, "ts": 1716454225967918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866305, "dur": 10,
    "args": {
      "External id": 283648,
      "cbid": 211, "correlation": 283648
    }
  },
  {
    "ph": "s", "id": 283648, "pid": 76337, "tid": -914061504, "ts": 1716454225866305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225967968, "dur": 4,
    "args": {
      "External id": 283660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283660,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283660, "pid": 5, "tid": 7, "ts": 1716454225967968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866326, "dur": 6,
    "args": {
      "External id": 283660,
      "cbid": 211, "correlation": 283660
    }
  },
  {
    "ph": "s", "id": 283660, "pid": 76337, "tid": -914061504, "ts": 1716454225866326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225967972, "dur": 53,
    "args": {
      "External id": 283663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283663,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283663, "pid": 5, "tid": 7, "ts": 1716454225967972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866345, "dur": 6,
    "args": {
      "External id": 283663,
      "cbid": 211, "correlation": 283663
    }
  },
  {
    "ph": "s", "id": 283663, "pid": 76337, "tid": -914061504, "ts": 1716454225866345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225968027, "dur": 36,
    "args": {
      "External id": 283672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283672,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283672, "pid": 5, "tid": 7, "ts": 1716454225968027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866385, "dur": 10,
    "args": {
      "External id": 283672,
      "cbid": 211, "correlation": 283672
    }
  },
  {
    "ph": "s", "id": 283672, "pid": 76337, "tid": -914061504, "ts": 1716454225866385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866446, "dur": 0,
    "args": {
      "External id": 283682,
      "cbid": 317, "correlation": 283682
    }
  },
  {
    "ph": "f", "id": 283682, "pid": 76337, "tid": -914061504, "ts": 1716454225866446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866447, "dur": 0,
    "args": {
      "External id": 283683,
      "cbid": 203, "correlation": 283683
    }
  },
  {
    "ph": "f", "id": 283683, "pid": 76337, "tid": -914061504, "ts": 1716454225866447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866448, "dur": 0,
    "args": {
      "External id": 283684,
      "cbid": 205, "correlation": 283684
    }
  },
  {
    "ph": "f", "id": 283684, "pid": 76337, "tid": -914061504, "ts": 1716454225866448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225968064, "dur": 39,
    "args": {
      "External id": 283688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283688,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283688, "pid": 5, "tid": 7, "ts": 1716454225968064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866462, "dur": 12,
    "args": {
      "External id": 283688,
      "cbid": 211, "correlation": 283688
    }
  },
  {
    "ph": "s", "id": 283688, "pid": 76337, "tid": -914061504, "ts": 1716454225866462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225968105, "dur": 14,
    "args": {
      "External id": 283690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283690,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283690, "pid": 5, "tid": 7, "ts": 1716454225968105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866477, "dur": 5,
    "args": {
      "External id": 283690,
      "cbid": 211, "correlation": 283690
    }
  },
  {
    "ph": "s", "id": 283690, "pid": 76337, "tid": -914061504, "ts": 1716454225866477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225968120, "dur": 3,
    "args": {
      "External id": 283692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283692, "pid": 5, "tid": 7, "ts": 1716454225968120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866486, "dur": 5,
    "args": {
      "External id": 283692,
      "cbid": 211, "correlation": 283692
    }
  },
  {
    "ph": "s", "id": 283692, "pid": 76337, "tid": -914061504, "ts": 1716454225866486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866494, "dur": 0,
    "args": {
      "External id": 283693,
      "cbid": 51, "correlation": 283693
    }
  },
  {
    "ph": "s", "id": 283693, "pid": 76337, "tid": -914061504, "ts": 1716454225866494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225968124, "dur": 675,
    "args": {
      "External id": 283694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283694,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283694, "pid": 5, "tid": 7, "ts": 1716454225968124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866495, "dur": 5,
    "args": {
      "External id": 283694,
      "cbid": 211, "correlation": 283694
    }
  },
  {
    "ph": "s", "id": 283694, "pid": 76337, "tid": -914061504, "ts": 1716454225866495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225968801, "dur": 57,
    "args": {
      "External id": 283699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283699, "pid": 5, "tid": 7, "ts": 1716454225968801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866522, "dur": 9,
    "args": {
      "External id": 283699,
      "cbid": 211, "correlation": 283699
    }
  },
  {
    "ph": "s", "id": 283699, "pid": 76337, "tid": -914061504, "ts": 1716454225866522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866579, "dur": 0,
    "args": {
      "External id": 283709,
      "cbid": 317, "correlation": 283709
    }
  },
  {
    "ph": "f", "id": 283709, "pid": 76337, "tid": -914061504, "ts": 1716454225866579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866580, "dur": 0,
    "args": {
      "External id": 283710,
      "cbid": 203, "correlation": 283710
    }
  },
  {
    "ph": "f", "id": 283710, "pid": 76337, "tid": -914061504, "ts": 1716454225866580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866581, "dur": 0,
    "args": {
      "External id": 283711,
      "cbid": 205, "correlation": 283711
    }
  },
  {
    "ph": "f", "id": 283711, "pid": 76337, "tid": -914061504, "ts": 1716454225866581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225968859, "dur": 75,
    "args": {
      "External id": 283715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283715,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283715, "pid": 5, "tid": 7, "ts": 1716454225968859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866593, "dur": 11,
    "args": {
      "External id": 283715,
      "cbid": 211, "correlation": 283715
    }
  },
  {
    "ph": "s", "id": 283715, "pid": 76337, "tid": -914061504, "ts": 1716454225866593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225968935, "dur": 201,
    "args": {
      "External id": 283717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283717,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283717, "pid": 5, "tid": 7, "ts": 1716454225968935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866612, "dur": 8,
    "args": {
      "External id": 283717,
      "cbid": 211, "correlation": 283717
    }
  },
  {
    "ph": "s", "id": 283717, "pid": 76337, "tid": -914061504, "ts": 1716454225866612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225969137, "dur": 39,
    "args": {
      "External id": 283719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283719,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283719, "pid": 5, "tid": 7, "ts": 1716454225969137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866625, "dur": 7,
    "args": {
      "External id": 283719,
      "cbid": 211, "correlation": 283719
    }
  },
  {
    "ph": "s", "id": 283719, "pid": 76337, "tid": -914061504, "ts": 1716454225866625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225969178, "dur": 57,
    "args": {
      "External id": 283725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283725,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283725, "pid": 5, "tid": 7, "ts": 1716454225969178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866653, "dur": 9,
    "args": {
      "External id": 283725,
      "cbid": 211, "correlation": 283725
    }
  },
  {
    "ph": "s", "id": 283725, "pid": 76337, "tid": -914061504, "ts": 1716454225866653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225969236, "dur": 51,
    "args": {
      "External id": 283733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283733,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283733, "pid": 5, "tid": 7, "ts": 1716454225969236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866683, "dur": 8,
    "args": {
      "External id": 283733,
      "cbid": 211, "correlation": 283733
    }
  },
  {
    "ph": "s", "id": 283733, "pid": 76337, "tid": -914061504, "ts": 1716454225866683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225969288, "dur": 35,
    "args": {
      "External id": 283741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283741,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283741, "pid": 5, "tid": 7, "ts": 1716454225969288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866711, "dur": 8,
    "args": {
      "External id": 283741,
      "cbid": 211, "correlation": 283741
    }
  },
  {
    "ph": "s", "id": 283741, "pid": 76337, "tid": -914061504, "ts": 1716454225866711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225969325, "dur": 49,
    "args": {
      "External id": 283761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283761,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 283761, "pid": 5, "tid": 7, "ts": 1716454225969325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866793, "dur": 12,
    "args": {
      "External id": 283761,
      "cbid": 211, "correlation": 283761
    }
  },
  {
    "ph": "s", "id": 283761, "pid": 76337, "tid": -914061504, "ts": 1716454225866793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225969375, "dur": 4,
    "args": {
      "External id": 283773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283773,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 283773, "pid": 5, "tid": 7, "ts": 1716454225969375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866818, "dur": 7,
    "args": {
      "External id": 283773,
      "cbid": 211, "correlation": 283773
    }
  },
  {
    "ph": "s", "id": 283773, "pid": 76337, "tid": -914061504, "ts": 1716454225866818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225969381, "dur": 56,
    "args": {
      "External id": 283776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283776, "pid": 5, "tid": 7, "ts": 1716454225969381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866837, "dur": 7,
    "args": {
      "External id": 283776,
      "cbid": 211, "correlation": 283776
    }
  },
  {
    "ph": "s", "id": 283776, "pid": 76337, "tid": -914061504, "ts": 1716454225866837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866895, "dur": 0,
    "args": {
      "External id": 283787,
      "cbid": 317, "correlation": 283787
    }
  },
  {
    "ph": "f", "id": 283787, "pid": 76337, "tid": -914061504, "ts": 1716454225866895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866896, "dur": 0,
    "args": {
      "External id": 283788,
      "cbid": 203, "correlation": 283788
    }
  },
  {
    "ph": "f", "id": 283788, "pid": 76337, "tid": -914061504, "ts": 1716454225866896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866897, "dur": 0,
    "args": {
      "External id": 283789,
      "cbid": 205, "correlation": 283789
    }
  },
  {
    "ph": "f", "id": 283789, "pid": 76337, "tid": -914061504, "ts": 1716454225866897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866919, "dur": 1,
    "args": {
      "External id": 283793,
      "cbid": 251, "correlation": 283793
    }
  },
  {
    "ph": "f", "id": 283793, "pid": 76337, "tid": -914061504, "ts": 1716454225866919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866921, "dur": 0,
    "args": {
      "External id": 283794,
      "cbid": 251, "correlation": 283794
    }
  },
  {
    "ph": "f", "id": 283794, "pid": 76337, "tid": -914061504, "ts": 1716454225866921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866921, "dur": 0,
    "args": {
      "External id": 283795,
      "cbid": 251, "correlation": 283795
    }
  },
  {
    "ph": "f", "id": 283795, "pid": 76337, "tid": -914061504, "ts": 1716454225866921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866922, "dur": 0,
    "args": {
      "External id": 283796,
      "cbid": 251, "correlation": 283796
    }
  },
  {
    "ph": "f", "id": 283796, "pid": 76337, "tid": -914061504, "ts": 1716454225866922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866923, "dur": 0,
    "args": {
      "External id": 283797,
      "cbid": 251, "correlation": 283797
    }
  },
  {
    "ph": "f", "id": 283797, "pid": 76337, "tid": -914061504, "ts": 1716454225866923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866923, "dur": 0,
    "args": {
      "External id": 283798,
      "cbid": 251, "correlation": 283798
    }
  },
  {
    "ph": "f", "id": 283798, "pid": 76337, "tid": -914061504, "ts": 1716454225866923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866924, "dur": 0,
    "args": {
      "External id": 283799,
      "cbid": 251, "correlation": 283799
    }
  },
  {
    "ph": "f", "id": 283799, "pid": 76337, "tid": -914061504, "ts": 1716454225866924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866925, "dur": 0,
    "args": {
      "External id": 283800,
      "cbid": 251, "correlation": 283800
    }
  },
  {
    "ph": "f", "id": 283800, "pid": 76337, "tid": -914061504, "ts": 1716454225866925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866926, "dur": 0,
    "args": {
      "External id": 283801,
      "cbid": 251, "correlation": 283801
    }
  },
  {
    "ph": "f", "id": 283801, "pid": 76337, "tid": -914061504, "ts": 1716454225866926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225969438, "dur": 108,
    "args": {
      "External id": 283802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283802,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283802, "pid": 5, "tid": 7, "ts": 1716454225969438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866928, "dur": 13,
    "args": {
      "External id": 283802,
      "cbid": 211, "correlation": 283802
    }
  },
  {
    "ph": "s", "id": 283802, "pid": 76337, "tid": -914061504, "ts": 1716454225866928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225969548, "dur": 57,
    "args": {
      "External id": 283808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283808, "pid": 5, "tid": 7, "ts": 1716454225969548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225866963, "dur": 9,
    "args": {
      "External id": 283808,
      "cbid": 211, "correlation": 283808
    }
  },
  {
    "ph": "s", "id": 283808, "pid": 76337, "tid": -914061504, "ts": 1716454225866963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225969606, "dur": 536,
    "args": {
      "External id": 283817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283817,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283817, "pid": 5, "tid": 7, "ts": 1716454225969606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867054, "dur": 14,
    "args": {
      "External id": 283817,
      "cbid": 211, "correlation": 283817
    }
  },
  {
    "ph": "s", "id": 283817, "pid": 76337, "tid": -914061504, "ts": 1716454225867054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225970144, "dur": 172,
    "args": {
      "External id": 283839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283839,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283839, "pid": 5, "tid": 7, "ts": 1716454225970144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867114, "dur": 11,
    "args": {
      "External id": 283839,
      "cbid": 211, "correlation": 283839
    }
  },
  {
    "ph": "s", "id": 283839, "pid": 76337, "tid": -914061504, "ts": 1716454225867114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867205, "dur": 1,
    "args": {
      "External id": 283850,
      "cbid": 251, "correlation": 283850
    }
  },
  {
    "ph": "f", "id": 283850, "pid": 76337, "tid": -914061504, "ts": 1716454225867205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225970317, "dur": 190,
    "args": {
      "External id": 283851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283851,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283851, "pid": 5, "tid": 7, "ts": 1716454225970317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867210, "dur": 13,
    "args": {
      "External id": 283851,
      "cbid": 211, "correlation": 283851
    }
  },
  {
    "ph": "s", "id": 283851, "pid": 76337, "tid": -914061504, "ts": 1716454225867210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867281, "dur": 1,
    "args": {
      "External id": 283862,
      "cbid": 251, "correlation": 283862
    }
  },
  {
    "ph": "f", "id": 283862, "pid": 76337, "tid": -914061504, "ts": 1716454225867281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225970509, "dur": 182,
    "args": {
      "External id": 283863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283863,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283863, "pid": 5, "tid": 7, "ts": 1716454225970509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867285, "dur": 12,
    "args": {
      "External id": 283863,
      "cbid": 211, "correlation": 283863
    }
  },
  {
    "ph": "s", "id": 283863, "pid": 76337, "tid": -914061504, "ts": 1716454225867285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867351, "dur": 1,
    "args": {
      "External id": 283874,
      "cbid": 251, "correlation": 283874
    }
  },
  {
    "ph": "f", "id": 283874, "pid": 76337, "tid": -914061504, "ts": 1716454225867351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225970692, "dur": 183,
    "args": {
      "External id": 283875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283875,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283875, "pid": 5, "tid": 7, "ts": 1716454225970692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867355, "dur": 11,
    "args": {
      "External id": 283875,
      "cbid": 211, "correlation": 283875
    }
  },
  {
    "ph": "s", "id": 283875, "pid": 76337, "tid": -914061504, "ts": 1716454225867355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225970876, "dur": 17654,
    "args": {
      "External id": 283896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283896,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 283896, "pid": 5, "tid": 7, "ts": 1716454225970876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867435, "dur": 13,
    "args": {
      "External id": 283896,
      "cbid": 211, "correlation": 283896
    }
  },
  {
    "ph": "s", "id": 283896, "pid": 76337, "tid": -914061504, "ts": 1716454225867435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867534, "dur": 1,
    "args": {
      "External id": 283914,
      "cbid": 251, "correlation": 283914
    }
  },
  {
    "ph": "f", "id": 283914, "pid": 76337, "tid": -914061504, "ts": 1716454225867534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225988531, "dur": 193,
    "args": {
      "External id": 283916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283916,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283916, "pid": 5, "tid": 7, "ts": 1716454225988531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867540, "dur": 14,
    "args": {
      "External id": 283916,
      "cbid": 211, "correlation": 283916
    }
  },
  {
    "ph": "s", "id": 283916, "pid": 76337, "tid": -914061504, "ts": 1716454225867540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225988726, "dur": 66,
    "args": {
      "External id": 283924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283924,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283924, "pid": 5, "tid": 7, "ts": 1716454225988726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867610, "dur": 12,
    "args": {
      "External id": 283924,
      "cbid": 211, "correlation": 283924
    }
  },
  {
    "ph": "s", "id": 283924, "pid": 76337, "tid": -914061504, "ts": 1716454225867610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225988793, "dur": 98,
    "args": {
      "External id": 283932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283932,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283932, "pid": 5, "tid": 7, "ts": 1716454225988793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867649, "dur": 9,
    "args": {
      "External id": 283932,
      "cbid": 211, "correlation": 283932
    }
  },
  {
    "ph": "s", "id": 283932, "pid": 76337, "tid": -914061504, "ts": 1716454225867649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225988892, "dur": 53,
    "args": {
      "External id": 283943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283943, "pid": 5, "tid": 7, "ts": 1716454225988892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867719, "dur": 12,
    "args": {
      "External id": 283943,
      "cbid": 211, "correlation": 283943
    }
  },
  {
    "ph": "s", "id": 283943, "pid": 76337, "tid": -914061504, "ts": 1716454225867719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225988946, "dur": 87,
    "args": {
      "External id": 283965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283965,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 283965, "pid": 5, "tid": 7, "ts": 1716454225988946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867750, "dur": 8,
    "args": {
      "External id": 283965,
      "cbid": 211, "correlation": 283965
    }
  },
  {
    "ph": "s", "id": 283965, "pid": 76337, "tid": -914061504, "ts": 1716454225867750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867838, "dur": 1,
    "args": {
      "External id": 283976,
      "cbid": 251, "correlation": 283976
    }
  },
  {
    "ph": "f", "id": 283976, "pid": 76337, "tid": -914061504, "ts": 1716454225867838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225989034, "dur": 97,
    "args": {
      "External id": 283977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283977,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 283977, "pid": 5, "tid": 7, "ts": 1716454225989034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867843, "dur": 14,
    "args": {
      "External id": 283977,
      "cbid": 211, "correlation": 283977
    }
  },
  {
    "ph": "s", "id": 283977, "pid": 76337, "tid": -914061504, "ts": 1716454225867843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867915, "dur": 1,
    "args": {
      "External id": 283988,
      "cbid": 251, "correlation": 283988
    }
  },
  {
    "ph": "f", "id": 283988, "pid": 76337, "tid": -914061504, "ts": 1716454225867915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867918, "dur": 0,
    "args": {
      "External id": 283989,
      "cbid": 251, "correlation": 283989
    }
  },
  {
    "ph": "f", "id": 283989, "pid": 76337, "tid": -914061504, "ts": 1716454225867918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225989133, "dur": 10,
    "args": {
      "External id": 283990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283990,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 283990, "pid": 5, "tid": 7, "ts": 1716454225989133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867920, "dur": 12,
    "args": {
      "External id": 283990,
      "cbid": 211, "correlation": 283990
    }
  },
  {
    "ph": "s", "id": 283990, "pid": 76337, "tid": -914061504, "ts": 1716454225867920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225989144, "dur": 5,
    "args": {
      "External id": 283992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 283992,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 283992, "pid": 5, "tid": 7, "ts": 1716454225989144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225867934, "dur": 6,
    "args": {
      "External id": 283992,
      "cbid": 211, "correlation": 283992
    }
  },
  {
    "ph": "s", "id": 283992, "pid": 76337, "tid": -914061504, "ts": 1716454225867934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868008, "dur": 1,
    "args": {
      "External id": 284003,
      "cbid": 251, "correlation": 284003
    }
  },
  {
    "ph": "f", "id": 284003, "pid": 76337, "tid": -914061504, "ts": 1716454225868008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868012, "dur": 0,
    "args": {
      "External id": 284004,
      "cbid": 251, "correlation": 284004
    }
  },
  {
    "ph": "f", "id": 284004, "pid": 76337, "tid": -914061504, "ts": 1716454225868012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225989151, "dur": 6,
    "args": {
      "External id": 284005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284005,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284005, "pid": 5, "tid": 7, "ts": 1716454225989151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868014, "dur": 12,
    "args": {
      "External id": 284005,
      "cbid": 211, "correlation": 284005
    }
  },
  {
    "ph": "s", "id": 284005, "pid": 76337, "tid": -914061504, "ts": 1716454225868014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225989158, "dur": 3,
    "args": {
      "External id": 284007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284007,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 284007, "pid": 5, "tid": 7, "ts": 1716454225989158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868028, "dur": 5,
    "args": {
      "External id": 284007,
      "cbid": 211, "correlation": 284007
    }
  },
  {
    "ph": "s", "id": 284007, "pid": 76337, "tid": -914061504, "ts": 1716454225868028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225989162, "dur": 148,
    "args": {
      "External id": 284028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284028,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284028, "pid": 5, "tid": 7, "ts": 1716454225989162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868102, "dur": 13,
    "args": {
      "External id": 284028,
      "cbid": 211, "correlation": 284028
    }
  },
  {
    "ph": "s", "id": 284028, "pid": 76337, "tid": -914061504, "ts": 1716454225868102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868198, "dur": 1,
    "args": {
      "External id": 284046,
      "cbid": 251, "correlation": 284046
    }
  },
  {
    "ph": "f", "id": 284046, "pid": 76337, "tid": -914061504, "ts": 1716454225868198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225989312, "dur": 102,
    "args": {
      "External id": 284048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284048,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284048, "pid": 5, "tid": 7, "ts": 1716454225989312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868204, "dur": 14,
    "args": {
      "External id": 284048,
      "cbid": 211, "correlation": 284048
    }
  },
  {
    "ph": "s", "id": 284048, "pid": 76337, "tid": -914061504, "ts": 1716454225868204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225989415, "dur": 34,
    "args": {
      "External id": 284056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284056,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284056, "pid": 5, "tid": 7, "ts": 1716454225989415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868273, "dur": 12,
    "args": {
      "External id": 284056,
      "cbid": 211, "correlation": 284056
    }
  },
  {
    "ph": "s", "id": 284056, "pid": 76337, "tid": -914061504, "ts": 1716454225868273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225989451, "dur": 65,
    "args": {
      "External id": 284064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284064, "pid": 5, "tid": 7, "ts": 1716454225989451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868314, "dur": 9,
    "args": {
      "External id": 284064,
      "cbid": 211, "correlation": 284064
    }
  },
  {
    "ph": "s", "id": 284064, "pid": 76337, "tid": -914061504, "ts": 1716454225868314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225989517, "dur": 87,
    "args": {
      "External id": 284086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284086,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284086, "pid": 5, "tid": 7, "ts": 1716454225989517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868364, "dur": 10,
    "args": {
      "External id": 284086,
      "cbid": 211, "correlation": 284086
    }
  },
  {
    "ph": "s", "id": 284086, "pid": 76337, "tid": -914061504, "ts": 1716454225868364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868450, "dur": 1,
    "args": {
      "External id": 284102,
      "cbid": 251, "correlation": 284102
    }
  },
  {
    "ph": "f", "id": 284102, "pid": 76337, "tid": -914061504, "ts": 1716454225868450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225989606, "dur": 558,
    "args": {
      "External id": 284104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284104,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284104, "pid": 5, "tid": 7, "ts": 1716454225989606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868457, "dur": 14,
    "args": {
      "External id": 284104,
      "cbid": 211, "correlation": 284104
    }
  },
  {
    "ph": "s", "id": 284104, "pid": 76337, "tid": -914061504, "ts": 1716454225868457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225990165, "dur": 237,
    "args": {
      "External id": 284112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284112, "pid": 5, "tid": 7, "ts": 1716454225990165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868525, "dur": 12,
    "args": {
      "External id": 284112,
      "cbid": 211, "correlation": 284112
    }
  },
  {
    "ph": "s", "id": 284112, "pid": 76337, "tid": -914061504, "ts": 1716454225868525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225990403, "dur": 253,
    "args": {
      "External id": 284120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284120, "pid": 5, "tid": 7, "ts": 1716454225990403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868556, "dur": 9,
    "args": {
      "External id": 284120,
      "cbid": 211, "correlation": 284120
    }
  },
  {
    "ph": "s", "id": 284120, "pid": 76337, "tid": -914061504, "ts": 1716454225868556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868637, "dur": 1,
    "args": {
      "External id": 284136,
      "cbid": 251, "correlation": 284136
    }
  },
  {
    "ph": "f", "id": 284136, "pid": 76337, "tid": -914061504, "ts": 1716454225868637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868642, "dur": 0,
    "args": {
      "External id": 284138,
      "cbid": 251, "correlation": 284138
    }
  },
  {
    "ph": "f", "id": 284138, "pid": 76337, "tid": -914061504, "ts": 1716454225868642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454225990658, "dur": 348,
    "args": {
      "External id": 284139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284139,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 284139, "pid": 5, "tid": 7, "ts": 1716454225990658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868645, "dur": 13,
    "args": {
      "External id": 284139,
      "cbid": 211, "correlation": 284139
    }
  },
  {
    "ph": "s", "id": 284139, "pid": 76337, "tid": -914061504, "ts": 1716454225868645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225991008, "dur": 50,
    "args": {
      "External id": 284147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284147,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284147, "pid": 5, "tid": 7, "ts": 1716454225991008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868686, "dur": 10,
    "args": {
      "External id": 284147,
      "cbid": 211, "correlation": 284147
    }
  },
  {
    "ph": "s", "id": 284147, "pid": 76337, "tid": -914061504, "ts": 1716454225868686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225991059, "dur": 150,
    "args": {
      "External id": 284158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284158, "pid": 5, "tid": 7, "ts": 1716454225991059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868753, "dur": 13,
    "args": {
      "External id": 284158,
      "cbid": 211, "correlation": 284158
    }
  },
  {
    "ph": "s", "id": 284158, "pid": 76337, "tid": -914061504, "ts": 1716454225868753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868817, "dur": 0,
    "args": {
      "External id": 284170,
      "cbid": 317, "correlation": 284170
    }
  },
  {
    "ph": "f", "id": 284170, "pid": 76337, "tid": -914061504, "ts": 1716454225868817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868818, "dur": 0,
    "args": {
      "External id": 284171,
      "cbid": 203, "correlation": 284171
    }
  },
  {
    "ph": "f", "id": 284171, "pid": 76337, "tid": -914061504, "ts": 1716454225868818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868818, "dur": 0,
    "args": {
      "External id": 284172,
      "cbid": 205, "correlation": 284172
    }
  },
  {
    "ph": "f", "id": 284172, "pid": 76337, "tid": -914061504, "ts": 1716454225868818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868841, "dur": 1,
    "args": {
      "External id": 284176,
      "cbid": 251, "correlation": 284176
    }
  },
  {
    "ph": "f", "id": 284176, "pid": 76337, "tid": -914061504, "ts": 1716454225868841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868843, "dur": 0,
    "args": {
      "External id": 284177,
      "cbid": 251, "correlation": 284177
    }
  },
  {
    "ph": "f", "id": 284177, "pid": 76337, "tid": -914061504, "ts": 1716454225868843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868843, "dur": 0,
    "args": {
      "External id": 284178,
      "cbid": 251, "correlation": 284178
    }
  },
  {
    "ph": "f", "id": 284178, "pid": 76337, "tid": -914061504, "ts": 1716454225868843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868844, "dur": 0,
    "args": {
      "External id": 284179,
      "cbid": 251, "correlation": 284179
    }
  },
  {
    "ph": "f", "id": 284179, "pid": 76337, "tid": -914061504, "ts": 1716454225868844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868845, "dur": 0,
    "args": {
      "External id": 284180,
      "cbid": 251, "correlation": 284180
    }
  },
  {
    "ph": "f", "id": 284180, "pid": 76337, "tid": -914061504, "ts": 1716454225868845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868845, "dur": 0,
    "args": {
      "External id": 284181,
      "cbid": 251, "correlation": 284181
    }
  },
  {
    "ph": "f", "id": 284181, "pid": 76337, "tid": -914061504, "ts": 1716454225868845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868846, "dur": 0,
    "args": {
      "External id": 284182,
      "cbid": 251, "correlation": 284182
    }
  },
  {
    "ph": "f", "id": 284182, "pid": 76337, "tid": -914061504, "ts": 1716454225868846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868847, "dur": 0,
    "args": {
      "External id": 284183,
      "cbid": 251, "correlation": 284183
    }
  },
  {
    "ph": "f", "id": 284183, "pid": 76337, "tid": -914061504, "ts": 1716454225868847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868848, "dur": 0,
    "args": {
      "External id": 284184,
      "cbid": 251, "correlation": 284184
    }
  },
  {
    "ph": "f", "id": 284184, "pid": 76337, "tid": -914061504, "ts": 1716454225868848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225991210, "dur": 112,
    "args": {
      "External id": 284185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284185,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284185, "pid": 5, "tid": 7, "ts": 1716454225991210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868850, "dur": 12,
    "args": {
      "External id": 284185,
      "cbid": 211, "correlation": 284185
    }
  },
  {
    "ph": "s", "id": 284185, "pid": 76337, "tid": -914061504, "ts": 1716454225868850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225991323, "dur": 58,
    "args": {
      "External id": 284191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284191, "pid": 5, "tid": 7, "ts": 1716454225991323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868885, "dur": 9,
    "args": {
      "External id": 284191,
      "cbid": 211, "correlation": 284191
    }
  },
  {
    "ph": "s", "id": 284191, "pid": 76337, "tid": -914061504, "ts": 1716454225868885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225991382, "dur": 50,
    "args": {
      "External id": 284199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284199,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284199, "pid": 5, "tid": 7, "ts": 1716454225991382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868916, "dur": 9,
    "args": {
      "External id": 284199,
      "cbid": 211, "correlation": 284199
    }
  },
  {
    "ph": "s", "id": 284199, "pid": 76337, "tid": -914061504, "ts": 1716454225868916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225991433, "dur": 95,
    "args": {
      "External id": 284208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284208,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284208, "pid": 5, "tid": 7, "ts": 1716454225991433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225868955, "dur": 10,
    "args": {
      "External id": 284208,
      "cbid": 211, "correlation": 284208
    }
  },
  {
    "ph": "s", "id": 284208, "pid": 76337, "tid": -914061504, "ts": 1716454225868955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225991529, "dur": 88,
    "args": {
      "External id": 284228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284228,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 284228, "pid": 5, "tid": 7, "ts": 1716454225991529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869032, "dur": 12,
    "args": {
      "External id": 284228,
      "cbid": 211, "correlation": 284228
    }
  },
  {
    "ph": "s", "id": 284228, "pid": 76337, "tid": -914061504, "ts": 1716454225869032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225991619, "dur": 4,
    "args": {
      "External id": 284240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284240,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 284240, "pid": 5, "tid": 7, "ts": 1716454225991619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869055, "dur": 6,
    "args": {
      "External id": 284240,
      "cbid": 211, "correlation": 284240
    }
  },
  {
    "ph": "s", "id": 284240, "pid": 76337, "tid": -914061504, "ts": 1716454225869055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225991625, "dur": 105,
    "args": {
      "External id": 284243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284243, "pid": 5, "tid": 7, "ts": 1716454225991625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869073, "dur": 7,
    "args": {
      "External id": 284243,
      "cbid": 211, "correlation": 284243
    }
  },
  {
    "ph": "s", "id": 284243, "pid": 76337, "tid": -914061504, "ts": 1716454225869073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225991731, "dur": 69,
    "args": {
      "External id": 284252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284252,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284252, "pid": 5, "tid": 7, "ts": 1716454225991731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869112, "dur": 9,
    "args": {
      "External id": 284252,
      "cbid": 211, "correlation": 284252
    }
  },
  {
    "ph": "s", "id": 284252, "pid": 76337, "tid": -914061504, "ts": 1716454225869112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869163, "dur": 0,
    "args": {
      "External id": 284262,
      "cbid": 317, "correlation": 284262
    }
  },
  {
    "ph": "f", "id": 284262, "pid": 76337, "tid": -914061504, "ts": 1716454225869163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869164, "dur": 0,
    "args": {
      "External id": 284263,
      "cbid": 203, "correlation": 284263
    }
  },
  {
    "ph": "f", "id": 284263, "pid": 76337, "tid": -914061504, "ts": 1716454225869164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869165, "dur": 0,
    "args": {
      "External id": 284264,
      "cbid": 205, "correlation": 284264
    }
  },
  {
    "ph": "f", "id": 284264, "pid": 76337, "tid": -914061504, "ts": 1716454225869165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225991801, "dur": 76,
    "args": {
      "External id": 284268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284268, "pid": 5, "tid": 7, "ts": 1716454225991801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869179, "dur": 12,
    "args": {
      "External id": 284268,
      "cbid": 211, "correlation": 284268
    }
  },
  {
    "ph": "s", "id": 284268, "pid": 76337, "tid": -914061504, "ts": 1716454225869179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225991878, "dur": 23,
    "args": {
      "External id": 284270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284270, "pid": 5, "tid": 7, "ts": 1716454225991878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869194, "dur": 5,
    "args": {
      "External id": 284270,
      "cbid": 211, "correlation": 284270
    }
  },
  {
    "ph": "s", "id": 284270, "pid": 76337, "tid": -914061504, "ts": 1716454225869194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225991903, "dur": 3,
    "args": {
      "External id": 284272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284272, "pid": 5, "tid": 7, "ts": 1716454225991903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869203, "dur": 5,
    "args": {
      "External id": 284272,
      "cbid": 211, "correlation": 284272
    }
  },
  {
    "ph": "s", "id": 284272, "pid": 76337, "tid": -914061504, "ts": 1716454225869203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869211, "dur": 0,
    "args": {
      "External id": 284273,
      "cbid": 51, "correlation": 284273
    }
  },
  {
    "ph": "s", "id": 284273, "pid": 76337, "tid": -914061504, "ts": 1716454225869211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225991907, "dur": 1317,
    "args": {
      "External id": 284274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284274,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284274, "pid": 5, "tid": 7, "ts": 1716454225991907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869212, "dur": 5,
    "args": {
      "External id": 284274,
      "cbid": 211, "correlation": 284274
    }
  },
  {
    "ph": "s", "id": 284274, "pid": 76337, "tid": -914061504, "ts": 1716454225869212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225993225, "dur": 57,
    "args": {
      "External id": 284279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284279, "pid": 5, "tid": 7, "ts": 1716454225993225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869240, "dur": 8,
    "args": {
      "External id": 284279,
      "cbid": 211, "correlation": 284279
    }
  },
  {
    "ph": "s", "id": 284279, "pid": 76337, "tid": -914061504, "ts": 1716454225869240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225993284, "dur": 3,
    "args": {
      "External id": 284287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284287,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284287, "pid": 5, "tid": 7, "ts": 1716454225993284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869283, "dur": 10,
    "args": {
      "External id": 284287,
      "cbid": 211, "correlation": 284287
    }
  },
  {
    "ph": "s", "id": 284287, "pid": 76337, "tid": -914061504, "ts": 1716454225869283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869352, "dur": 1,
    "args": {
      "External id": 284303,
      "cbid": 251, "correlation": 284303
    }
  },
  {
    "ph": "f", "id": 284303, "pid": 76337, "tid": -914061504, "ts": 1716454225869352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869357, "dur": 0,
    "args": {
      "External id": 284305,
      "cbid": 251, "correlation": 284305
    }
  },
  {
    "ph": "f", "id": 284305, "pid": 76337, "tid": -914061504, "ts": 1716454225869357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225993288, "dur": 11,
    "args": {
      "External id": 284306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284306,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 284306, "pid": 5, "tid": 7, "ts": 1716454225993288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869359, "dur": 11,
    "args": {
      "External id": 284306,
      "cbid": 211, "correlation": 284306
    }
  },
  {
    "ph": "s", "id": 284306, "pid": 76337, "tid": -914061504, "ts": 1716454225869359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454225993300, "dur": 5,
    "args": {
      "External id": 284308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284308,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 284308, "pid": 5, "tid": 7, "ts": 1716454225993300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869372, "dur": 5,
    "args": {
      "External id": 284308,
      "cbid": 211, "correlation": 284308
    }
  },
  {
    "ph": "s", "id": 284308, "pid": 76337, "tid": -914061504, "ts": 1716454225869372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225993307, "dur": 52,
    "args": {
      "External id": 284318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284318, "pid": 5, "tid": 7, "ts": 1716454225993307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869429, "dur": 13,
    "args": {
      "External id": 284318,
      "cbid": 211, "correlation": 284318
    }
  },
  {
    "ph": "s", "id": 284318, "pid": 76337, "tid": -914061504, "ts": 1716454225869429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225993360, "dur": 48,
    "args": {
      "External id": 284338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284338,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 284338, "pid": 5, "tid": 7, "ts": 1716454225993360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869495, "dur": 11,
    "args": {
      "External id": 284338,
      "cbid": 211, "correlation": 284338
    }
  },
  {
    "ph": "s", "id": 284338, "pid": 76337, "tid": -914061504, "ts": 1716454225869495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225993409, "dur": 4,
    "args": {
      "External id": 284350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284350,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284350, "pid": 5, "tid": 7, "ts": 1716454225993409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869516, "dur": 6,
    "args": {
      "External id": 284350,
      "cbid": 211, "correlation": 284350
    }
  },
  {
    "ph": "s", "id": 284350, "pid": 76337, "tid": -914061504, "ts": 1716454225869516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225993414, "dur": 52,
    "args": {
      "External id": 284353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284353, "pid": 5, "tid": 7, "ts": 1716454225993414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869536, "dur": 6,
    "args": {
      "External id": 284353,
      "cbid": 211, "correlation": 284353
    }
  },
  {
    "ph": "s", "id": 284353, "pid": 76337, "tid": -914061504, "ts": 1716454225869536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225993468, "dur": 36,
    "args": {
      "External id": 284362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284362,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284362, "pid": 5, "tid": 7, "ts": 1716454225993468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869576, "dur": 10,
    "args": {
      "External id": 284362,
      "cbid": 211, "correlation": 284362
    }
  },
  {
    "ph": "s", "id": 284362, "pid": 76337, "tid": -914061504, "ts": 1716454225869576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869637, "dur": 0,
    "args": {
      "External id": 284372,
      "cbid": 317, "correlation": 284372
    }
  },
  {
    "ph": "f", "id": 284372, "pid": 76337, "tid": -914061504, "ts": 1716454225869637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869638, "dur": 0,
    "args": {
      "External id": 284373,
      "cbid": 203, "correlation": 284373
    }
  },
  {
    "ph": "f", "id": 284373, "pid": 76337, "tid": -914061504, "ts": 1716454225869638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869638, "dur": 0,
    "args": {
      "External id": 284374,
      "cbid": 205, "correlation": 284374
    }
  },
  {
    "ph": "f", "id": 284374, "pid": 76337, "tid": -914061504, "ts": 1716454225869638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225993505, "dur": 39,
    "args": {
      "External id": 284378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284378, "pid": 5, "tid": 7, "ts": 1716454225993505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869652, "dur": 12,
    "args": {
      "External id": 284378,
      "cbid": 211, "correlation": 284378
    }
  },
  {
    "ph": "s", "id": 284378, "pid": 76337, "tid": -914061504, "ts": 1716454225869652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225993545, "dur": 13,
    "args": {
      "External id": 284380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284380,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284380, "pid": 5, "tid": 7, "ts": 1716454225993545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869666, "dur": 5,
    "args": {
      "External id": 284380,
      "cbid": 211, "correlation": 284380
    }
  },
  {
    "ph": "s", "id": 284380, "pid": 76337, "tid": -914061504, "ts": 1716454225869666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454225993560, "dur": 3,
    "args": {
      "External id": 284382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284382, "pid": 5, "tid": 7, "ts": 1716454225993560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869675, "dur": 6,
    "args": {
      "External id": 284382,
      "cbid": 211, "correlation": 284382
    }
  },
  {
    "ph": "s", "id": 284382, "pid": 76337, "tid": -914061504, "ts": 1716454225869675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869684, "dur": 0,
    "args": {
      "External id": 284383,
      "cbid": 51, "correlation": 284383
    }
  },
  {
    "ph": "s", "id": 284383, "pid": 76337, "tid": -914061504, "ts": 1716454225869684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454225993564, "dur": 674,
    "args": {
      "External id": 284384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284384,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284384, "pid": 5, "tid": 7, "ts": 1716454225993564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869685, "dur": 6,
    "args": {
      "External id": 284384,
      "cbid": 211, "correlation": 284384
    }
  },
  {
    "ph": "s", "id": 284384, "pid": 76337, "tid": -914061504, "ts": 1716454225869685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225994240, "dur": 58,
    "args": {
      "External id": 284389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284389, "pid": 5, "tid": 7, "ts": 1716454225994240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869713, "dur": 8,
    "args": {
      "External id": 284389,
      "cbid": 211, "correlation": 284389
    }
  },
  {
    "ph": "s", "id": 284389, "pid": 76337, "tid": -914061504, "ts": 1716454225869713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869770, "dur": 0,
    "args": {
      "External id": 284399,
      "cbid": 317, "correlation": 284399
    }
  },
  {
    "ph": "f", "id": 284399, "pid": 76337, "tid": -914061504, "ts": 1716454225869770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869771, "dur": 0,
    "args": {
      "External id": 284400,
      "cbid": 203, "correlation": 284400
    }
  },
  {
    "ph": "f", "id": 284400, "pid": 76337, "tid": -914061504, "ts": 1716454225869771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869771, "dur": 0,
    "args": {
      "External id": 284401,
      "cbid": 205, "correlation": 284401
    }
  },
  {
    "ph": "f", "id": 284401, "pid": 76337, "tid": -914061504, "ts": 1716454225869771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225994299, "dur": 75,
    "args": {
      "External id": 284405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284405,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284405, "pid": 5, "tid": 7, "ts": 1716454225994299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869783, "dur": 12,
    "args": {
      "External id": 284405,
      "cbid": 211, "correlation": 284405
    }
  },
  {
    "ph": "s", "id": 284405, "pid": 76337, "tid": -914061504, "ts": 1716454225869783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454225994375, "dur": 202,
    "args": {
      "External id": 284407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284407,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284407, "pid": 5, "tid": 7, "ts": 1716454225994375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869801, "dur": 6,
    "args": {
      "External id": 284407,
      "cbid": 211, "correlation": 284407
    }
  },
  {
    "ph": "s", "id": 284407, "pid": 76337, "tid": -914061504, "ts": 1716454225869801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454225994578, "dur": 38,
    "args": {
      "External id": 284409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284409,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284409, "pid": 5, "tid": 7, "ts": 1716454225994578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869812, "dur": 5,
    "args": {
      "External id": 284409,
      "cbid": 211, "correlation": 284409
    }
  },
  {
    "ph": "s", "id": 284409, "pid": 76337, "tid": -914061504, "ts": 1716454225869812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225994617, "dur": 57,
    "args": {
      "External id": 284415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284415, "pid": 5, "tid": 7, "ts": 1716454225994617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225869837, "dur": 368,
    "args": {
      "External id": 284415,
      "cbid": 211, "correlation": 284415
    }
  },
  {
    "ph": "s", "id": 284415, "pid": 76337, "tid": -914061504, "ts": 1716454225869837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454225994676, "dur": 50,
    "args": {
      "External id": 284423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284423,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284423, "pid": 5, "tid": 7, "ts": 1716454225994676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870227, "dur": 8,
    "args": {
      "External id": 284423,
      "cbid": 211, "correlation": 284423
    }
  },
  {
    "ph": "s", "id": 284423, "pid": 76337, "tid": -914061504, "ts": 1716454225870227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454225994727, "dur": 35,
    "args": {
      "External id": 284431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284431,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284431, "pid": 5, "tid": 7, "ts": 1716454225994727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870256, "dur": 8,
    "args": {
      "External id": 284431,
      "cbid": 211, "correlation": 284431
    }
  },
  {
    "ph": "s", "id": 284431, "pid": 76337, "tid": -914061504, "ts": 1716454225870256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225994763, "dur": 49,
    "args": {
      "External id": 284451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284451,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 284451, "pid": 5, "tid": 7, "ts": 1716454225994763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870337, "dur": 13,
    "args": {
      "External id": 284451,
      "cbid": 211, "correlation": 284451
    }
  },
  {
    "ph": "s", "id": 284451, "pid": 76337, "tid": -914061504, "ts": 1716454225870337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454225994813, "dur": 4,
    "args": {
      "External id": 284463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284463,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284463, "pid": 5, "tid": 7, "ts": 1716454225994813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870360, "dur": 6,
    "args": {
      "External id": 284463,
      "cbid": 211, "correlation": 284463
    }
  },
  {
    "ph": "s", "id": 284463, "pid": 76337, "tid": -914061504, "ts": 1716454225870360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225994819, "dur": 52,
    "args": {
      "External id": 284466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284466, "pid": 5, "tid": 7, "ts": 1716454225994819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870378, "dur": 6,
    "args": {
      "External id": 284466,
      "cbid": 211, "correlation": 284466
    }
  },
  {
    "ph": "s", "id": 284466, "pid": 76337, "tid": -914061504, "ts": 1716454225870378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870434, "dur": 0,
    "args": {
      "External id": 284477,
      "cbid": 317, "correlation": 284477
    }
  },
  {
    "ph": "f", "id": 284477, "pid": 76337, "tid": -914061504, "ts": 1716454225870434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870435, "dur": 0,
    "args": {
      "External id": 284478,
      "cbid": 203, "correlation": 284478
    }
  },
  {
    "ph": "f", "id": 284478, "pid": 76337, "tid": -914061504, "ts": 1716454225870435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870436, "dur": 0,
    "args": {
      "External id": 284479,
      "cbid": 205, "correlation": 284479
    }
  },
  {
    "ph": "f", "id": 284479, "pid": 76337, "tid": -914061504, "ts": 1716454225870436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870458, "dur": 1,
    "args": {
      "External id": 284483,
      "cbid": 251, "correlation": 284483
    }
  },
  {
    "ph": "f", "id": 284483, "pid": 76337, "tid": -914061504, "ts": 1716454225870458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870460, "dur": 0,
    "args": {
      "External id": 284484,
      "cbid": 251, "correlation": 284484
    }
  },
  {
    "ph": "f", "id": 284484, "pid": 76337, "tid": -914061504, "ts": 1716454225870460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870461, "dur": 0,
    "args": {
      "External id": 284485,
      "cbid": 251, "correlation": 284485
    }
  },
  {
    "ph": "f", "id": 284485, "pid": 76337, "tid": -914061504, "ts": 1716454225870461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870462, "dur": 0,
    "args": {
      "External id": 284486,
      "cbid": 251, "correlation": 284486
    }
  },
  {
    "ph": "f", "id": 284486, "pid": 76337, "tid": -914061504, "ts": 1716454225870462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870462, "dur": 0,
    "args": {
      "External id": 284487,
      "cbid": 251, "correlation": 284487
    }
  },
  {
    "ph": "f", "id": 284487, "pid": 76337, "tid": -914061504, "ts": 1716454225870462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870463, "dur": 0,
    "args": {
      "External id": 284488,
      "cbid": 251, "correlation": 284488
    }
  },
  {
    "ph": "f", "id": 284488, "pid": 76337, "tid": -914061504, "ts": 1716454225870463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870464, "dur": 0,
    "args": {
      "External id": 284489,
      "cbid": 251, "correlation": 284489
    }
  },
  {
    "ph": "f", "id": 284489, "pid": 76337, "tid": -914061504, "ts": 1716454225870464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870465, "dur": 0,
    "args": {
      "External id": 284490,
      "cbid": 251, "correlation": 284490
    }
  },
  {
    "ph": "f", "id": 284490, "pid": 76337, "tid": -914061504, "ts": 1716454225870465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870466, "dur": 0,
    "args": {
      "External id": 284491,
      "cbid": 251, "correlation": 284491
    }
  },
  {
    "ph": "f", "id": 284491, "pid": 76337, "tid": -914061504, "ts": 1716454225870466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454225994872, "dur": 109,
    "args": {
      "External id": 284492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284492,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284492, "pid": 5, "tid": 7, "ts": 1716454225994872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870468, "dur": 12,
    "args": {
      "External id": 284492,
      "cbid": 211, "correlation": 284492
    }
  },
  {
    "ph": "s", "id": 284492, "pid": 76337, "tid": -914061504, "ts": 1716454225870468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454225994983, "dur": 57,
    "args": {
      "External id": 284498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284498, "pid": 5, "tid": 7, "ts": 1716454225994983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870504, "dur": 9,
    "args": {
      "External id": 284498,
      "cbid": 211, "correlation": 284498
    }
  },
  {
    "ph": "s", "id": 284498, "pid": 76337, "tid": -914061504, "ts": 1716454225870504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454225995041, "dur": 503,
    "args": {
      "External id": 284507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284507,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284507, "pid": 5, "tid": 7, "ts": 1716454225995041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870585, "dur": 14,
    "args": {
      "External id": 284507,
      "cbid": 211, "correlation": 284507
    }
  },
  {
    "ph": "s", "id": 284507, "pid": 76337, "tid": -914061504, "ts": 1716454225870585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454225995545, "dur": 171,
    "args": {
      "External id": 284529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284529,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284529, "pid": 5, "tid": 7, "ts": 1716454225995545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870642, "dur": 11,
    "args": {
      "External id": 284529,
      "cbid": 211, "correlation": 284529
    }
  },
  {
    "ph": "s", "id": 284529, "pid": 76337, "tid": -914061504, "ts": 1716454225870642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870732, "dur": 1,
    "args": {
      "External id": 284540,
      "cbid": 251, "correlation": 284540
    }
  },
  {
    "ph": "f", "id": 284540, "pid": 76337, "tid": -914061504, "ts": 1716454225870732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225995718, "dur": 191,
    "args": {
      "External id": 284541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284541,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284541, "pid": 5, "tid": 7, "ts": 1716454225995718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870737, "dur": 15,
    "args": {
      "External id": 284541,
      "cbid": 211, "correlation": 284541
    }
  },
  {
    "ph": "s", "id": 284541, "pid": 76337, "tid": -914061504, "ts": 1716454225870737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870807, "dur": 1,
    "args": {
      "External id": 284552,
      "cbid": 251, "correlation": 284552
    }
  },
  {
    "ph": "f", "id": 284552, "pid": 76337, "tid": -914061504, "ts": 1716454225870807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225995910, "dur": 182,
    "args": {
      "External id": 284553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284553,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284553, "pid": 5, "tid": 7, "ts": 1716454225995910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870811, "dur": 11,
    "args": {
      "External id": 284553,
      "cbid": 211, "correlation": 284553
    }
  },
  {
    "ph": "s", "id": 284553, "pid": 76337, "tid": -914061504, "ts": 1716454225870811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870877, "dur": 1,
    "args": {
      "External id": 284564,
      "cbid": 251, "correlation": 284564
    }
  },
  {
    "ph": "f", "id": 284564, "pid": 76337, "tid": -914061504, "ts": 1716454225870877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454225996093, "dur": 180,
    "args": {
      "External id": 284565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284565,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284565, "pid": 5, "tid": 7, "ts": 1716454225996093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870881, "dur": 11,
    "args": {
      "External id": 284565,
      "cbid": 211, "correlation": 284565
    }
  },
  {
    "ph": "s", "id": 284565, "pid": 76337, "tid": -914061504, "ts": 1716454225870881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454225996274, "dur": 17603,
    "args": {
      "External id": 284586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284586,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284586, "pid": 5, "tid": 7, "ts": 1716454225996274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225870962, "dur": 19,
    "args": {
      "External id": 284586,
      "cbid": 211, "correlation": 284586
    }
  },
  {
    "ph": "s", "id": 284586, "pid": 76337, "tid": -914061504, "ts": 1716454225870962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225871066, "dur": 1,
    "args": {
      "External id": 284604,
      "cbid": 251, "correlation": 284604
    }
  },
  {
    "ph": "f", "id": 284604, "pid": 76337, "tid": -914061504, "ts": 1716454225871066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454226013878, "dur": 195,
    "args": {
      "External id": 284606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284606,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284606, "pid": 5, "tid": 7, "ts": 1716454226013878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225871071, "dur": 13,
    "args": {
      "External id": 284606,
      "cbid": 211, "correlation": 284606
    }
  },
  {
    "ph": "s", "id": 284606, "pid": 76337, "tid": -914061504, "ts": 1716454225871071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226014074, "dur": 66,
    "args": {
      "External id": 284614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284614,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284614, "pid": 5, "tid": 7, "ts": 1716454226014074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225871140, "dur": 12,
    "args": {
      "External id": 284614,
      "cbid": 211, "correlation": 284614
    }
  },
  {
    "ph": "s", "id": 284614, "pid": 76337, "tid": -914061504, "ts": 1716454225871140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226014142, "dur": 98,
    "args": {
      "External id": 284622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284622,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284622, "pid": 5, "tid": 7, "ts": 1716454226014142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225871179, "dur": 77,
    "args": {
      "External id": 284622,
      "cbid": 211, "correlation": 284622
    }
  },
  {
    "ph": "s", "id": 284622, "pid": 76337, "tid": -914061504, "ts": 1716454225871179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226014241, "dur": 52,
    "args": {
      "External id": 284633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284633, "pid": 5, "tid": 7, "ts": 1716454226014241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225871321, "dur": 1794,
    "args": {
      "External id": 284633,
      "cbid": 211, "correlation": 284633
    }
  },
  {
    "ph": "s", "id": 284633, "pid": 76337, "tid": -914061504, "ts": 1716454225871321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226014294, "dur": 87,
    "args": {
      "External id": 284655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284655,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284655, "pid": 5, "tid": 7, "ts": 1716454226014294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873135, "dur": 122,
    "args": {
      "External id": 284655,
      "cbid": 211, "correlation": 284655
    }
  },
  {
    "ph": "s", "id": 284655, "pid": 76337, "tid": -914061504, "ts": 1716454225873135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873336, "dur": 1,
    "args": {
      "External id": 284666,
      "cbid": 251, "correlation": 284666
    }
  },
  {
    "ph": "f", "id": 284666, "pid": 76337, "tid": -914061504, "ts": 1716454225873336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454226014383, "dur": 102,
    "args": {
      "External id": 284667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284667,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284667, "pid": 5, "tid": 7, "ts": 1716454226014383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873341, "dur": 13,
    "args": {
      "External id": 284667,
      "cbid": 211, "correlation": 284667
    }
  },
  {
    "ph": "s", "id": 284667, "pid": 76337, "tid": -914061504, "ts": 1716454225873341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873415, "dur": 1,
    "args": {
      "External id": 284678,
      "cbid": 251, "correlation": 284678
    }
  },
  {
    "ph": "f", "id": 284678, "pid": 76337, "tid": -914061504, "ts": 1716454225873415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873419, "dur": 0,
    "args": {
      "External id": 284679,
      "cbid": 251, "correlation": 284679
    }
  },
  {
    "ph": "f", "id": 284679, "pid": 76337, "tid": -914061504, "ts": 1716454225873419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454226014486, "dur": 10,
    "args": {
      "External id": 284680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284680,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284680, "pid": 5, "tid": 7, "ts": 1716454226014486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873421, "dur": 12,
    "args": {
      "External id": 284680,
      "cbid": 211, "correlation": 284680
    }
  },
  {
    "ph": "s", "id": 284680, "pid": 76337, "tid": -914061504, "ts": 1716454225873421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454226014497, "dur": 5,
    "args": {
      "External id": 284682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284682,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 284682, "pid": 5, "tid": 7, "ts": 1716454226014497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873435, "dur": 6,
    "args": {
      "External id": 284682,
      "cbid": 211, "correlation": 284682
    }
  },
  {
    "ph": "s", "id": 284682, "pid": 76337, "tid": -914061504, "ts": 1716454225873435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873499, "dur": 1,
    "args": {
      "External id": 284693,
      "cbid": 251, "correlation": 284693
    }
  },
  {
    "ph": "f", "id": 284693, "pid": 76337, "tid": -914061504, "ts": 1716454225873499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873503, "dur": 0,
    "args": {
      "External id": 284694,
      "cbid": 251, "correlation": 284694
    }
  },
  {
    "ph": "f", "id": 284694, "pid": 76337, "tid": -914061504, "ts": 1716454225873503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454226014503, "dur": 6,
    "args": {
      "External id": 284695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284695,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284695, "pid": 5, "tid": 7, "ts": 1716454226014503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873504, "dur": 12,
    "args": {
      "External id": 284695,
      "cbid": 211, "correlation": 284695
    }
  },
  {
    "ph": "s", "id": 284695, "pid": 76337, "tid": -914061504, "ts": 1716454225873504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454226014510, "dur": 3,
    "args": {
      "External id": 284697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284697,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 284697, "pid": 5, "tid": 7, "ts": 1716454226014510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873518, "dur": 6,
    "args": {
      "External id": 284697,
      "cbid": 211, "correlation": 284697
    }
  },
  {
    "ph": "s", "id": 284697, "pid": 76337, "tid": -914061504, "ts": 1716454225873518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454226014515, "dur": 149,
    "args": {
      "External id": 284718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284718,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284718, "pid": 5, "tid": 7, "ts": 1716454226014515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873592, "dur": 13,
    "args": {
      "External id": 284718,
      "cbid": 211, "correlation": 284718
    }
  },
  {
    "ph": "s", "id": 284718, "pid": 76337, "tid": -914061504, "ts": 1716454225873592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873691, "dur": 1,
    "args": {
      "External id": 284736,
      "cbid": 251, "correlation": 284736
    }
  },
  {
    "ph": "f", "id": 284736, "pid": 76337, "tid": -914061504, "ts": 1716454225873691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454226014665, "dur": 103,
    "args": {
      "External id": 284738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284738,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284738, "pid": 5, "tid": 7, "ts": 1716454226014665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873697, "dur": 14,
    "args": {
      "External id": 284738,
      "cbid": 211, "correlation": 284738
    }
  },
  {
    "ph": "s", "id": 284738, "pid": 76337, "tid": -914061504, "ts": 1716454225873697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226014770, "dur": 35,
    "args": {
      "External id": 284746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284746,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284746, "pid": 5, "tid": 7, "ts": 1716454226014770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873767, "dur": 12,
    "args": {
      "External id": 284746,
      "cbid": 211, "correlation": 284746
    }
  },
  {
    "ph": "s", "id": 284746, "pid": 76337, "tid": -914061504, "ts": 1716454225873767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226014806, "dur": 64,
    "args": {
      "External id": 284754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284754, "pid": 5, "tid": 7, "ts": 1716454226014806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873807, "dur": 10,
    "args": {
      "External id": 284754,
      "cbid": 211, "correlation": 284754
    }
  },
  {
    "ph": "s", "id": 284754, "pid": 76337, "tid": -914061504, "ts": 1716454225873807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226014872, "dur": 87,
    "args": {
      "External id": 284776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284776,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284776, "pid": 5, "tid": 7, "ts": 1716454226014872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873857, "dur": 11,
    "args": {
      "External id": 284776,
      "cbid": 211, "correlation": 284776
    }
  },
  {
    "ph": "s", "id": 284776, "pid": 76337, "tid": -914061504, "ts": 1716454225873857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873941, "dur": 1,
    "args": {
      "External id": 284792,
      "cbid": 251, "correlation": 284792
    }
  },
  {
    "ph": "f", "id": 284792, "pid": 76337, "tid": -914061504, "ts": 1716454225873941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454226014960, "dur": 561,
    "args": {
      "External id": 284794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284794,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 284794, "pid": 5, "tid": 7, "ts": 1716454226014960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225873947, "dur": 12,
    "args": {
      "External id": 284794,
      "cbid": 211, "correlation": 284794
    }
  },
  {
    "ph": "s", "id": 284794, "pid": 76337, "tid": -914061504, "ts": 1716454225873947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226015522, "dur": 238,
    "args": {
      "External id": 284802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284802,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284802, "pid": 5, "tid": 7, "ts": 1716454226015522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874021, "dur": 13,
    "args": {
      "External id": 284802,
      "cbid": 211, "correlation": 284802
    }
  },
  {
    "ph": "s", "id": 284802, "pid": 76337, "tid": -914061504, "ts": 1716454225874021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226015762, "dur": 247,
    "args": {
      "External id": 284810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284810, "pid": 5, "tid": 7, "ts": 1716454226015762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874051, "dur": 8,
    "args": {
      "External id": 284810,
      "cbid": 211, "correlation": 284810
    }
  },
  {
    "ph": "s", "id": 284810, "pid": 76337, "tid": -914061504, "ts": 1716454225874051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874133, "dur": 1,
    "args": {
      "External id": 284826,
      "cbid": 251, "correlation": 284826
    }
  },
  {
    "ph": "f", "id": 284826, "pid": 76337, "tid": -914061504, "ts": 1716454225874133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874138, "dur": 0,
    "args": {
      "External id": 284828,
      "cbid": 251, "correlation": 284828
    }
  },
  {
    "ph": "f", "id": 284828, "pid": 76337, "tid": -914061504, "ts": 1716454225874138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454226016010, "dur": 351,
    "args": {
      "External id": 284829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284829,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 284829, "pid": 5, "tid": 7, "ts": 1716454226016010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874141, "dur": 13,
    "args": {
      "External id": 284829,
      "cbid": 211, "correlation": 284829
    }
  },
  {
    "ph": "s", "id": 284829, "pid": 76337, "tid": -914061504, "ts": 1716454225874141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226016362, "dur": 50,
    "args": {
      "External id": 284837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284837,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284837, "pid": 5, "tid": 7, "ts": 1716454226016362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874183, "dur": 148,
    "args": {
      "External id": 284837,
      "cbid": 211, "correlation": 284837
    }
  },
  {
    "ph": "s", "id": 284837, "pid": 76337, "tid": -914061504, "ts": 1716454225874183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226016414, "dur": 150,
    "args": {
      "External id": 284848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284848, "pid": 5, "tid": 7, "ts": 1716454226016414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874389, "dur": 67,
    "args": {
      "External id": 284848,
      "cbid": 211, "correlation": 284848
    }
  },
  {
    "ph": "s", "id": 284848, "pid": 76337, "tid": -914061504, "ts": 1716454225874389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874509, "dur": 0,
    "args": {
      "External id": 284860,
      "cbid": 317, "correlation": 284860
    }
  },
  {
    "ph": "f", "id": 284860, "pid": 76337, "tid": -914061504, "ts": 1716454225874509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874509, "dur": 0,
    "args": {
      "External id": 284861,
      "cbid": 203, "correlation": 284861
    }
  },
  {
    "ph": "f", "id": 284861, "pid": 76337, "tid": -914061504, "ts": 1716454225874509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874510, "dur": 0,
    "args": {
      "External id": 284862,
      "cbid": 205, "correlation": 284862
    }
  },
  {
    "ph": "f", "id": 284862, "pid": 76337, "tid": -914061504, "ts": 1716454225874510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874533, "dur": 1,
    "args": {
      "External id": 284866,
      "cbid": 251, "correlation": 284866
    }
  },
  {
    "ph": "f", "id": 284866, "pid": 76337, "tid": -914061504, "ts": 1716454225874533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874534, "dur": 0,
    "args": {
      "External id": 284867,
      "cbid": 251, "correlation": 284867
    }
  },
  {
    "ph": "f", "id": 284867, "pid": 76337, "tid": -914061504, "ts": 1716454225874534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874535, "dur": 0,
    "args": {
      "External id": 284868,
      "cbid": 251, "correlation": 284868
    }
  },
  {
    "ph": "f", "id": 284868, "pid": 76337, "tid": -914061504, "ts": 1716454225874535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874536, "dur": 0,
    "args": {
      "External id": 284869,
      "cbid": 251, "correlation": 284869
    }
  },
  {
    "ph": "f", "id": 284869, "pid": 76337, "tid": -914061504, "ts": 1716454225874536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874537, "dur": 0,
    "args": {
      "External id": 284870,
      "cbid": 251, "correlation": 284870
    }
  },
  {
    "ph": "f", "id": 284870, "pid": 76337, "tid": -914061504, "ts": 1716454225874537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874538, "dur": 0,
    "args": {
      "External id": 284871,
      "cbid": 251, "correlation": 284871
    }
  },
  {
    "ph": "f", "id": 284871, "pid": 76337, "tid": -914061504, "ts": 1716454225874538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874538, "dur": 0,
    "args": {
      "External id": 284872,
      "cbid": 251, "correlation": 284872
    }
  },
  {
    "ph": "f", "id": 284872, "pid": 76337, "tid": -914061504, "ts": 1716454225874538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874539, "dur": 0,
    "args": {
      "External id": 284873,
      "cbid": 251, "correlation": 284873
    }
  },
  {
    "ph": "f", "id": 284873, "pid": 76337, "tid": -914061504, "ts": 1716454225874539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874541, "dur": 0,
    "args": {
      "External id": 284874,
      "cbid": 251, "correlation": 284874
    }
  },
  {
    "ph": "f", "id": 284874, "pid": 76337, "tid": -914061504, "ts": 1716454225874541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454226016565, "dur": 109,
    "args": {
      "External id": 284875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284875,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 284875, "pid": 5, "tid": 7, "ts": 1716454226016565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874543, "dur": 38,
    "args": {
      "External id": 284875,
      "cbid": 211, "correlation": 284875
    }
  },
  {
    "ph": "s", "id": 284875, "pid": 76337, "tid": -914061504, "ts": 1716454225874543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226016675, "dur": 57,
    "args": {
      "External id": 284881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284881, "pid": 5, "tid": 7, "ts": 1716454226016675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874605, "dur": 277,
    "args": {
      "External id": 284881,
      "cbid": 211, "correlation": 284881
    }
  },
  {
    "ph": "s", "id": 284881, "pid": 76337, "tid": -914061504, "ts": 1716454225874605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226016733, "dur": 50,
    "args": {
      "External id": 284889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284889,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284889, "pid": 5, "tid": 7, "ts": 1716454226016733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874905, "dur": 8,
    "args": {
      "External id": 284889,
      "cbid": 211, "correlation": 284889
    }
  },
  {
    "ph": "s", "id": 284889, "pid": 76337, "tid": -914061504, "ts": 1716454225874905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226016785, "dur": 50,
    "args": {
      "External id": 284909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284909,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 284909, "pid": 5, "tid": 7, "ts": 1716454226016785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225874998, "dur": 13,
    "args": {
      "External id": 284909,
      "cbid": 211, "correlation": 284909
    }
  },
  {
    "ph": "s", "id": 284909, "pid": 76337, "tid": -914061504, "ts": 1716454225874998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226016836, "dur": 5,
    "args": {
      "External id": 284921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284921,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284921, "pid": 5, "tid": 7, "ts": 1716454226016836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875022, "dur": 7,
    "args": {
      "External id": 284921,
      "cbid": 211, "correlation": 284921
    }
  },
  {
    "ph": "s", "id": 284921, "pid": 76337, "tid": -914061504, "ts": 1716454225875022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226016842, "dur": 55,
    "args": {
      "External id": 284924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284924, "pid": 5, "tid": 7, "ts": 1716454226016842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875042, "dur": 88,
    "args": {
      "External id": 284924,
      "cbid": 211, "correlation": 284924
    }
  },
  {
    "ph": "s", "id": 284924, "pid": 76337, "tid": -914061504, "ts": 1716454225875042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226016898, "dur": 37,
    "args": {
      "External id": 284933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284933,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284933, "pid": 5, "tid": 7, "ts": 1716454226016898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875170, "dur": 10,
    "args": {
      "External id": 284933,
      "cbid": 211, "correlation": 284933
    }
  },
  {
    "ph": "s", "id": 284933, "pid": 76337, "tid": -914061504, "ts": 1716454225875170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875226, "dur": 0,
    "args": {
      "External id": 284943,
      "cbid": 317, "correlation": 284943
    }
  },
  {
    "ph": "f", "id": 284943, "pid": 76337, "tid": -914061504, "ts": 1716454225875226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875226, "dur": 0,
    "args": {
      "External id": 284944,
      "cbid": 203, "correlation": 284944
    }
  },
  {
    "ph": "f", "id": 284944, "pid": 76337, "tid": -914061504, "ts": 1716454225875226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875227, "dur": 0,
    "args": {
      "External id": 284945,
      "cbid": 205, "correlation": 284945
    }
  },
  {
    "ph": "f", "id": 284945, "pid": 76337, "tid": -914061504, "ts": 1716454225875227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226016936, "dur": 41,
    "args": {
      "External id": 284949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284949, "pid": 5, "tid": 7, "ts": 1716454226016936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875244, "dur": 12,
    "args": {
      "External id": 284949,
      "cbid": 211, "correlation": 284949
    }
  },
  {
    "ph": "s", "id": 284949, "pid": 76337, "tid": -914061504, "ts": 1716454225875244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454226016978, "dur": 3,
    "args": {
      "External id": 284951,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 284951,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.80327868852459
    }
  },
  {
    "ph": "f", "id": 284951, "pid": 5, "tid": 7, "ts": 1716454226016978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875259, "dur": 22,
    "args": {
      "External id": 284951,
      "cbid": 51, "correlation": 284951
    }
  },
  {
    "ph": "s", "id": 284951, "pid": 76337, "tid": -914061504, "ts": 1716454225875259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875292, "dur": 2,
    "args": {
      "External id": 284953,
      "cbid": 200, "correlation": 284953
    }
  },
  {
    "ph": "f", "id": 284953, "pid": 76337, "tid": -914061504, "ts": 1716454225875292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875295, "dur": 0,
    "args": {
      "External id": 284954,
      "cbid": 200, "correlation": 284954
    }
  },
  {
    "ph": "f", "id": 284954, "pid": 76337, "tid": -914061504, "ts": 1716454225875295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875295, "dur": 0,
    "args": {
      "External id": 284955,
      "cbid": 200, "correlation": 284955
    }
  },
  {
    "ph": "f", "id": 284955, "pid": 76337, "tid": -914061504, "ts": 1716454225875295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875296, "dur": 0,
    "args": {
      "External id": 284956,
      "cbid": 200, "correlation": 284956
    }
  },
  {
    "ph": "f", "id": 284956, "pid": 76337, "tid": -914061504, "ts": 1716454225875296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875297, "dur": 5,
    "args": {
      "External id": 284957,
      "cbid": 15, "correlation": 284957
    }
  },
  {
    "ph": "f", "id": 284957, "pid": 76337, "tid": -914061504, "ts": 1716454225875297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875302, "dur": 1,
    "args": {
      "External id": 284958,
      "cbid": 251, "correlation": 284958
    }
  },
  {
    "ph": "f", "id": 284958, "pid": 76337, "tid": -914061504, "ts": 1716454225875302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454226016983, "dur": 23,
    "args": {
      "External id": 284959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284959,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284959, "pid": 5, "tid": 7, "ts": 1716454226016983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875306, "dur": 11,
    "args": {
      "External id": 284959,
      "cbid": 211, "correlation": 284959
    }
  },
  {
    "ph": "s", "id": 284959, "pid": 76337, "tid": -914061504, "ts": 1716454225875306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226017007, "dur": 4,
    "args": {
      "External id": 284961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 284961, "pid": 5, "tid": 7, "ts": 1716454226017007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875322, "dur": 6,
    "args": {
      "External id": 284961,
      "cbid": 211, "correlation": 284961
    }
  },
  {
    "ph": "s", "id": 284961, "pid": 76337, "tid": -914061504, "ts": 1716454225875322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875331, "dur": 0,
    "args": {
      "External id": 284962,
      "cbid": 51, "correlation": 284962
    }
  },
  {
    "ph": "s", "id": 284962, "pid": 76337, "tid": -914061504, "ts": 1716454225875331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226017013, "dur": 182,
    "args": {
      "External id": 284963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284963,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284963, "pid": 5, "tid": 7, "ts": 1716454226017013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875332, "dur": 179,
    "args": {
      "External id": 284963,
      "cbid": 211, "correlation": 284963
    }
  },
  {
    "ph": "s", "id": 284963, "pid": 76337, "tid": -914061504, "ts": 1716454225875332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226017196, "dur": 6,
    "args": {
      "External id": 284964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284964,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 284964, "pid": 5, "tid": 7, "ts": 1716454226017196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875515, "dur": 6,
    "args": {
      "External id": 284964,
      "cbid": 211, "correlation": 284964
    }
  },
  {
    "ph": "s", "id": 284964, "pid": 76337, "tid": -914061504, "ts": 1716454225875515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226017203, "dur": 4,
    "args": {
      "External id": 284970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 284970, "pid": 5, "tid": 7, "ts": 1716454226017203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225875545, "dur": 9,
    "args": {
      "External id": 284970,
      "cbid": 211, "correlation": 284970
    }
  },
  {
    "ph": "s", "id": 284970, "pid": 76337, "tid": -914061504, "ts": 1716454225875545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226017209, "dur": 3,
    "args": {
      "External id": 284978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284978,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284978, "pid": 5, "tid": 7, "ts": 1716454226017209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877323, "dur": 16,
    "args": {
      "External id": 284978,
      "cbid": 211, "correlation": 284978
    }
  },
  {
    "ph": "s", "id": 284978, "pid": 76337, "tid": -914061504, "ts": 1716454225877323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226017213, "dur": 3,
    "args": {
      "External id": 284986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284986,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284986, "pid": 5, "tid": 7, "ts": 1716454226017213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877364, "dur": 11,
    "args": {
      "External id": 284986,
      "cbid": 211, "correlation": 284986
    }
  },
  {
    "ph": "s", "id": 284986, "pid": 76337, "tid": -914061504, "ts": 1716454225877364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226017217, "dur": 3,
    "args": {
      "External id": 284994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 284994,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 284994, "pid": 5, "tid": 7, "ts": 1716454226017217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877391, "dur": 9,
    "args": {
      "External id": 284994,
      "cbid": 211, "correlation": 284994
    }
  },
  {
    "ph": "s", "id": 284994, "pid": 76337, "tid": -914061504, "ts": 1716454225877391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226017221, "dur": 3,
    "args": {
      "External id": 285003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285003,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285003, "pid": 5, "tid": 7, "ts": 1716454226017221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877567, "dur": 14,
    "args": {
      "External id": 285003,
      "cbid": 211, "correlation": 285003
    }
  },
  {
    "ph": "s", "id": 285003, "pid": 76337, "tid": -914061504, "ts": 1716454225877567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226017225, "dur": 3,
    "args": {
      "External id": 285012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285012,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285012, "pid": 5, "tid": 7, "ts": 1716454226017225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877596, "dur": 7,
    "args": {
      "External id": 285012,
      "cbid": 211, "correlation": 285012
    }
  },
  {
    "ph": "s", "id": 285012, "pid": 76337, "tid": -914061504, "ts": 1716454225877596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226017229, "dur": 3,
    "args": {
      "External id": 285020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285020,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285020, "pid": 5, "tid": 7, "ts": 1716454226017229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877621, "dur": 8,
    "args": {
      "External id": 285020,
      "cbid": 211, "correlation": 285020
    }
  },
  {
    "ph": "s", "id": 285020, "pid": 76337, "tid": -914061504, "ts": 1716454225877621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226017234, "dur": 3,
    "args": {
      "External id": 285028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285028,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285028, "pid": 5, "tid": 7, "ts": 1716454226017234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877877, "dur": 16,
    "args": {
      "External id": 285028,
      "cbid": 211, "correlation": 285028
    }
  },
  {
    "ph": "s", "id": 285028, "pid": 76337, "tid": -914061504, "ts": 1716454225877877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226017238, "dur": 3,
    "args": {
      "External id": 285036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285036,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285036, "pid": 5, "tid": 7, "ts": 1716454226017238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877908, "dur": 7,
    "args": {
      "External id": 285036,
      "cbid": 211, "correlation": 285036
    }
  },
  {
    "ph": "s", "id": 285036, "pid": 76337, "tid": -914061504, "ts": 1716454225877908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226017242, "dur": 1,
    "args": {
      "External id": 285046,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285046,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 285046, "pid": 5, "tid": 7, "ts": 1716454226017242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454225877982, "dur": 38,
    "args": {
      "External id": 285046,
      "cbid": 41, "correlation": 285046
    }
  },
  {
    "ph": "s", "id": 285046, "pid": 76337, "tid": -914061504, "ts": 1716454225877982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454225878021, "dur": 139241,
    "args": {
      "External id": 285047,
      "cbid": 131, "correlation": 285047
    }
  },
  {
    "ph": "f", "id": 285047, "pid": 76337, "tid": -914061504, "ts": 1716454225878021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454226017449, "dur": 2,
    "args": {
      "External id": 285055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285055, "pid": 5, "tid": 7, "ts": 1716454226017449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017422, "dur": 29,
    "args": {
      "External id": 285055,
      "cbid": 211, "correlation": 285055
    }
  },
  {
    "ph": "s", "id": 285055, "pid": 76337, "tid": -914061504, "ts": 1716454226017422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226017555, "dur": 3,
    "args": {
      "External id": 285064,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285064,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0025
    }
  },
  {
    "ph": "f", "id": 285064, "pid": 5, "tid": 7, "ts": 1716454226017555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017513, "dur": 42,
    "args": {
      "External id": 285064,
      "cbid": 41, "correlation": 285064
    }
  },
  {
    "ph": "s", "id": 285064, "pid": 76337, "tid": -914061504, "ts": 1716454226017513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454226017654, "dur": 4,
    "args": {
      "External id": 285074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285074,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285074, "pid": 5, "tid": 7, "ts": 1716454226017654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017638, "dur": 18,
    "args": {
      "External id": 285074,
      "cbid": 211, "correlation": 285074
    }
  },
  {
    "ph": "s", "id": 285074, "pid": 76337, "tid": -914061504, "ts": 1716454226017638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226017740, "dur": 1,
    "args": {
      "External id": 285084,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285084,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 285084, "pid": 5, "tid": 7, "ts": 1716454226017740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017714, "dur": 24,
    "args": {
      "External id": 285084,
      "cbid": 41, "correlation": 285084
    }
  },
  {
    "ph": "s", "id": 285084, "pid": 76337, "tid": -914061504, "ts": 1716454226017714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017739, "dur": 8,
    "args": {
      "External id": 285085,
      "cbid": 131, "correlation": 285085
    }
  },
  {
    "ph": "f", "id": 285085, "pid": 76337, "tid": -914061504, "ts": 1716454226017739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226017813, "dur": 3,
    "args": {
      "External id": 285092,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285092,
      "bytes": 98304, "memory bandwidth (GB/s)": 32
    }
  },
  {
    "ph": "f", "id": 285092, "pid": 5, "tid": 7, "ts": 1716454226017813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017789, "dur": 23,
    "args": {
      "External id": 285092,
      "cbid": 41, "correlation": 285092
    }
  },
  {
    "ph": "s", "id": 285092, "pid": 76337, "tid": -914061504, "ts": 1716454226017789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226017902, "dur": 2,
    "args": {
      "External id": 285111,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285111,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005555555555555556
    }
  },
  {
    "ph": "f", "id": 285111, "pid": 5, "tid": 7, "ts": 1716454226017902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017882, "dur": 20,
    "args": {
      "External id": 285111,
      "cbid": 41, "correlation": 285111
    }
  },
  {
    "ph": "s", "id": 285111, "pid": 76337, "tid": -914061504, "ts": 1716454226017882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454226017942, "dur": 2,
    "args": {
      "External id": 285117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285117,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285117, "pid": 5, "tid": 7, "ts": 1716454226017942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017929, "dur": 13,
    "args": {
      "External id": 285117,
      "cbid": 211, "correlation": 285117
    }
  },
  {
    "ph": "s", "id": 285117, "pid": 76337, "tid": -914061504, "ts": 1716454226017929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454226017959, "dur": 6,
    "args": {
      "External id": 285119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285119,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 285119, "pid": 5, "tid": 7, "ts": 1716454226017959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017946, "dur": 12,
    "args": {
      "External id": 285119,
      "cbid": 211, "correlation": 285119
    }
  },
  {
    "ph": "s", "id": 285119, "pid": 76337, "tid": -914061504, "ts": 1716454226017946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454226017968, "dur": 3,
    "args": {
      "External id": 285121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285121, "pid": 5, "tid": 7, "ts": 1716454226017968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017960, "dur": 7,
    "args": {
      "External id": 285121,
      "cbid": 211, "correlation": 285121
    }
  },
  {
    "ph": "s", "id": 285121, "pid": 76337, "tid": -914061504, "ts": 1716454226017960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226018013, "dur": 2,
    "args": {
      "External id": 285129,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285129,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0028735632183908046
    }
  },
  {
    "ph": "f", "id": 285129, "pid": 5, "tid": 7, "ts": 1716454226018013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226017997, "dur": 15,
    "args": {
      "External id": 285129,
      "cbid": 41, "correlation": 285129
    }
  },
  {
    "ph": "s", "id": 285129, "pid": 76337, "tid": -914061504, "ts": 1716454226017997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454226018060, "dur": 2,
    "args": {
      "External id": 285143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285143,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285143, "pid": 5, "tid": 7, "ts": 1716454226018060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018050, "dur": 12,
    "args": {
      "External id": 285143,
      "cbid": 211, "correlation": 285143
    }
  },
  {
    "ph": "s", "id": 285143, "pid": 76337, "tid": -914061504, "ts": 1716454226018050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454226018080, "dur": 2,
    "args": {
      "External id": 285157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285157,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285157, "pid": 5, "tid": 7, "ts": 1716454226018080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018073, "dur": 7,
    "args": {
      "External id": 285157,
      "cbid": 211, "correlation": 285157
    }
  },
  {
    "ph": "s", "id": 285157, "pid": 76337, "tid": -914061504, "ts": 1716454226018073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454226018117, "dur": 6,
    "args": {
      "External id": 285164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285164,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285164, "pid": 5, "tid": 7, "ts": 1716454226018117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018106, "dur": 11,
    "args": {
      "External id": 285164,
      "cbid": 211, "correlation": 285164
    }
  },
  {
    "ph": "s", "id": 285164, "pid": 76337, "tid": -914061504, "ts": 1716454226018106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454226018129, "dur": 6,
    "args": {
      "External id": 285167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285167,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285167, "pid": 5, "tid": 7, "ts": 1716454226018129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018121, "dur": 8,
    "args": {
      "External id": 285167,
      "cbid": 211, "correlation": 285167
    }
  },
  {
    "ph": "s", "id": 285167, "pid": 76337, "tid": -914061504, "ts": 1716454226018121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454226018140, "dur": 3,
    "args": {
      "External id": 285169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285169,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285169, "pid": 5, "tid": 7, "ts": 1716454226018140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018131, "dur": 6,
    "args": {
      "External id": 285169,
      "cbid": 211, "correlation": 285169
    }
  },
  {
    "ph": "s", "id": 285169, "pid": 76337, "tid": -914061504, "ts": 1716454226018131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226018159, "dur": 2,
    "args": {
      "External id": 285172,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285172,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0029080334423845873
    }
  },
  {
    "ph": "f", "id": 285172, "pid": 5, "tid": 7, "ts": 1716454226018159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018146, "dur": 12,
    "args": {
      "External id": 285172,
      "cbid": 41, "correlation": 285172
    }
  },
  {
    "ph": "s", "id": 285172, "pid": 76337, "tid": -914061504, "ts": 1716454226018146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454226018215, "dur": 4,
    "args": {
      "External id": 285188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285188,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285188, "pid": 5, "tid": 7, "ts": 1716454226018215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018199, "dur": 16,
    "args": {
      "External id": 285188,
      "cbid": 211, "correlation": 285188
    }
  },
  {
    "ph": "s", "id": 285188, "pid": 76337, "tid": -914061504, "ts": 1716454226018199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226018236, "dur": 3,
    "args": {
      "External id": 285193,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285193,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003255208333333333
    }
  },
  {
    "ph": "f", "id": 285193, "pid": 5, "tid": 7, "ts": 1716454226018236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018221, "dur": 14,
    "args": {
      "External id": 285193,
      "cbid": 41, "correlation": 285193
    }
  },
  {
    "ph": "s", "id": 285193, "pid": 76337, "tid": -914061504, "ts": 1716454226018221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454226018261, "dur": 1,
    "args": {
      "External id": 285199,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285199,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0006009615384615385
    }
  },
  {
    "ph": "f", "id": 285199, "pid": 5, "tid": 7, "ts": 1716454226018261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018244, "dur": 26,
    "args": {
      "External id": 285199,
      "cbid": 41, "correlation": 285199
    }
  },
  {
    "ph": "s", "id": 285199, "pid": 76337, "tid": -914061504, "ts": 1716454226018244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018271, "dur": 4,
    "args": {
      "External id": 285200,
      "cbid": 131, "correlation": 285200
    }
  },
  {
    "ph": "f", "id": 285200, "pid": 76337, "tid": -914061504, "ts": 1716454226018271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226018325, "dur": 3,
    "args": {
      "External id": 285208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285208,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285208, "pid": 5, "tid": 7, "ts": 1716454226018325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018312, "dur": 14,
    "args": {
      "External id": 285208,
      "cbid": 211, "correlation": 285208
    }
  },
  {
    "ph": "s", "id": 285208, "pid": 76337, "tid": -914061504, "ts": 1716454226018312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226018356, "dur": 3,
    "args": {
      "External id": 285218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285218,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285218, "pid": 5, "tid": 7, "ts": 1716454226018356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018347, "dur": 8,
    "args": {
      "External id": 285218,
      "cbid": 211, "correlation": 285218
    }
  },
  {
    "ph": "s", "id": 285218, "pid": 76337, "tid": -914061504, "ts": 1716454226018347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226018383, "dur": 3,
    "args": {
      "External id": 285227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285227,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285227, "pid": 5, "tid": 7, "ts": 1716454226018383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018373, "dur": 8,
    "args": {
      "External id": 285227,
      "cbid": 211, "correlation": 285227
    }
  },
  {
    "ph": "s", "id": 285227, "pid": 76337, "tid": -914061504, "ts": 1716454226018373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454226018498, "dur": 11,
    "args": {
      "External id": 285237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285237,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285237, "pid": 5, "tid": 7, "ts": 1716454226018498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018484, "dur": 15,
    "args": {
      "External id": 285237,
      "cbid": 211, "correlation": 285237
    }
  },
  {
    "ph": "s", "id": 285237, "pid": 76337, "tid": -914061504, "ts": 1716454226018484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226018539, "dur": 3,
    "args": {
      "External id": 285245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285245,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285245, "pid": 5, "tid": 7, "ts": 1716454226018539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018530, "dur": 9,
    "args": {
      "External id": 285245,
      "cbid": 211, "correlation": 285245
    }
  },
  {
    "ph": "s", "id": 285245, "pid": 76337, "tid": -914061504, "ts": 1716454226018530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454226018585, "dur": 11,
    "args": {
      "External id": 285255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285255,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285255, "pid": 5, "tid": 7, "ts": 1716454226018585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018573, "dur": 11,
    "args": {
      "External id": 285255,
      "cbid": 211, "correlation": 285255
    }
  },
  {
    "ph": "s", "id": 285255, "pid": 76337, "tid": -914061504, "ts": 1716454226018573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454226018616, "dur": 10,
    "args": {
      "External id": 285263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285263,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285263, "pid": 5, "tid": 7, "ts": 1716454226018616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018606, "dur": 9,
    "args": {
      "External id": 285263,
      "cbid": 211, "correlation": 285263
    }
  },
  {
    "ph": "s", "id": 285263, "pid": 76337, "tid": -914061504, "ts": 1716454226018606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226018643, "dur": 3,
    "args": {
      "External id": 285272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285272, "pid": 5, "tid": 7, "ts": 1716454226018643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018633, "dur": 8,
    "args": {
      "External id": 285272,
      "cbid": 211, "correlation": 285272
    }
  },
  {
    "ph": "s", "id": 285272, "pid": 76337, "tid": -914061504, "ts": 1716454226018633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454226018672, "dur": 5,
    "args": {
      "External id": 285281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285281,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285281, "pid": 5, "tid": 7, "ts": 1716454226018672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018659, "dur": 12,
    "args": {
      "External id": 285281,
      "cbid": 211, "correlation": 285281
    }
  },
  {
    "ph": "s", "id": 285281, "pid": 76337, "tid": -914061504, "ts": 1716454226018659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454226018710, "dur": 8,
    "args": {
      "External id": 285291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285291,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285291, "pid": 5, "tid": 7, "ts": 1716454226018710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018700, "dur": 10,
    "args": {
      "External id": 285291,
      "cbid": 211, "correlation": 285291
    }
  },
  {
    "ph": "s", "id": 285291, "pid": 76337, "tid": -914061504, "ts": 1716454226018700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226018984, "dur": 1,
    "args": {
      "External id": 285302,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285302,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0026595744680851063
    }
  },
  {
    "ph": "f", "id": 285302, "pid": 5, "tid": 7, "ts": 1716454226018984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018961, "dur": 22,
    "args": {
      "External id": 285302,
      "cbid": 41, "correlation": 285302
    }
  },
  {
    "ph": "s", "id": 285302, "pid": 76337, "tid": -914061504, "ts": 1716454226018961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454226018983, "dur": 8,
    "args": {
      "External id": 285303,
      "cbid": 131, "correlation": 285303
    }
  },
  {
    "ph": "f", "id": 285303, "pid": 76337, "tid": -914061504, "ts": 1716454226018983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454226019074, "dur": 2,
    "args": {
      "External id": 285311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 285311, "pid": 5, "tid": 7, "ts": 1716454226019074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019058, "dur": 16,
    "args": {
      "External id": 285311,
      "cbid": 211, "correlation": 285311
    }
  },
  {
    "ph": "s", "id": 285311, "pid": 76337, "tid": -914061504, "ts": 1716454226019058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226019157, "dur": 2,
    "args": {
      "External id": 285320,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285320,
      "bytes": 4, "memory bandwidth (GB/s)": 0.001358695652173913
    }
  },
  {
    "ph": "f", "id": 285320, "pid": 5, "tid": 7, "ts": 1716454226019157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019138, "dur": 19,
    "args": {
      "External id": 285320,
      "cbid": 41, "correlation": 285320
    }
  },
  {
    "ph": "s", "id": 285320, "pid": 76337, "tid": -914061504, "ts": 1716454226019138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454226019213, "dur": 1,
    "args": {
      "External id": 285331,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285331,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0026595744680851063
    }
  },
  {
    "ph": "f", "id": 285331, "pid": 5, "tid": 7, "ts": 1716454226019213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019197, "dur": 14,
    "args": {
      "External id": 285331,
      "cbid": 41, "correlation": 285331
    }
  },
  {
    "ph": "s", "id": 285331, "pid": 76337, "tid": -914061504, "ts": 1716454226019197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019211, "dur": 8,
    "args": {
      "External id": 285332,
      "cbid": 131, "correlation": 285332
    }
  },
  {
    "ph": "f", "id": 285332, "pid": 76337, "tid": -914061504, "ts": 1716454226019211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226019266, "dur": 3,
    "args": {
      "External id": 285340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285340,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285340, "pid": 5, "tid": 7, "ts": 1716454226019266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019252, "dur": 13,
    "args": {
      "External id": 285340,
      "cbid": 211, "correlation": 285340
    }
  },
  {
    "ph": "s", "id": 285340, "pid": 76337, "tid": -914061504, "ts": 1716454226019252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226019295, "dur": 3,
    "args": {
      "External id": 285350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285350,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285350, "pid": 5, "tid": 7, "ts": 1716454226019295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019285, "dur": 8,
    "args": {
      "External id": 285350,
      "cbid": 211, "correlation": 285350
    }
  },
  {
    "ph": "s", "id": 285350, "pid": 76337, "tid": -914061504, "ts": 1716454226019285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226019316, "dur": 3,
    "args": {
      "External id": 285359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285359,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285359, "pid": 5, "tid": 7, "ts": 1716454226019316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019308, "dur": 7,
    "args": {
      "External id": 285359,
      "cbid": 211, "correlation": 285359
    }
  },
  {
    "ph": "s", "id": 285359, "pid": 76337, "tid": -914061504, "ts": 1716454226019308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226019367, "dur": 3,
    "args": {
      "External id": 285367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285367,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285367, "pid": 5, "tid": 7, "ts": 1716454226019367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019356, "dur": 11,
    "args": {
      "External id": 285367,
      "cbid": 211, "correlation": 285367
    }
  },
  {
    "ph": "s", "id": 285367, "pid": 76337, "tid": -914061504, "ts": 1716454226019356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226019424, "dur": 3,
    "args": {
      "External id": 285375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285375,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285375, "pid": 5, "tid": 7, "ts": 1716454226019424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019414, "dur": 9,
    "args": {
      "External id": 285375,
      "cbid": 211, "correlation": 285375
    }
  },
  {
    "ph": "s", "id": 285375, "pid": 76337, "tid": -914061504, "ts": 1716454226019414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226019477, "dur": 3,
    "args": {
      "External id": 285383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285383,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 285383, "pid": 5, "tid": 7, "ts": 1716454226019477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019466, "dur": 12,
    "args": {
      "External id": 285383,
      "cbid": 211, "correlation": 285383
    }
  },
  {
    "ph": "s", "id": 285383, "pid": 76337, "tid": -914061504, "ts": 1716454226019466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226019505, "dur": 5,
    "args": {
      "External id": 285391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285391, "pid": 5, "tid": 7, "ts": 1716454226019505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019496, "dur": 9,
    "args": {
      "External id": 285391,
      "cbid": 211, "correlation": 285391
    }
  },
  {
    "ph": "s", "id": 285391, "pid": 76337, "tid": -914061504, "ts": 1716454226019496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226019529, "dur": 4,
    "args": {
      "External id": 285399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285399,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285399, "pid": 5, "tid": 7, "ts": 1716454226019529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019520, "dur": 8,
    "args": {
      "External id": 285399,
      "cbid": 211, "correlation": 285399
    }
  },
  {
    "ph": "s", "id": 285399, "pid": 76337, "tid": -914061504, "ts": 1716454226019520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226019549, "dur": 3,
    "args": {
      "External id": 285407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285407,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285407, "pid": 5, "tid": 7, "ts": 1716454226019549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226019541, "dur": 7,
    "args": {
      "External id": 285407,
      "cbid": 211, "correlation": 285407
    }
  },
  {
    "ph": "s", "id": 285407, "pid": 76337, "tid": -914061504, "ts": 1716454226019541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226020300, "dur": 3,
    "args": {
      "External id": 285415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285415,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285415, "pid": 5, "tid": 7, "ts": 1716454226020300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226020281, "dur": 20,
    "args": {
      "External id": 285415,
      "cbid": 211, "correlation": 285415
    }
  },
  {
    "ph": "s", "id": 285415, "pid": 76337, "tid": -914061504, "ts": 1716454226020281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226020462, "dur": 1,
    "args": {
      "External id": 285425,
      "cbid": 317, "correlation": 285425
    }
  },
  {
    "ph": "f", "id": 285425, "pid": 76337, "tid": -914061504, "ts": 1716454226020462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226020465, "dur": 1,
    "args": {
      "External id": 285426,
      "cbid": 203, "correlation": 285426
    }
  },
  {
    "ph": "f", "id": 285426, "pid": 76337, "tid": -914061504, "ts": 1716454226020465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226020466, "dur": 1,
    "args": {
      "External id": 285427,
      "cbid": 205, "correlation": 285427
    }
  },
  {
    "ph": "f", "id": 285427, "pid": 76337, "tid": -914061504, "ts": 1716454226020466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226024001, "dur": 171,
    "args": {
      "External id": 285431,
      "cbid": 251, "correlation": 285431
    }
  },
  {
    "ph": "f", "id": 285431, "pid": 76337, "tid": -914061504, "ts": 1716454226024001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226024174, "dur": 45,
    "args": {
      "External id": 285432,
      "cbid": 251, "correlation": 285432
    }
  },
  {
    "ph": "f", "id": 285432, "pid": 76337, "tid": -914061504, "ts": 1716454226024174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226024221, "dur": 0,
    "args": {
      "External id": 285433,
      "cbid": 251, "correlation": 285433
    }
  },
  {
    "ph": "f", "id": 285433, "pid": 76337, "tid": -914061504, "ts": 1716454226024221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_NN_kernel<float, 256, 4, 2, 8, 4, 4, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454226024251, "dur": 5,
    "args": {
      "External id": 285434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285434,
      "registers per thread": 64,
      "shared memory": 12288,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [48, 1, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 285434, "pid": 5, "tid": 7, "ts": 1716454226024251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226024227, "dur": 99,
    "args": {
      "External id": 285434,
      "cbid": 211, "correlation": 285434
    }
  },
  {
    "ph": "s", "id": 285434, "pid": 76337, "tid": -914061504, "ts": 1716454226024227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226025037, "dur": 5,
    "args": {
      "External id": 285440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 285440, "pid": 5, "tid": 7, "ts": 1716454226025037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226025020, "dur": 16,
    "args": {
      "External id": 285440,
      "cbid": 211, "correlation": 285440
    }
  },
  {
    "ph": "s", "id": 285440, "pid": 76337, "tid": -914061504, "ts": 1716454226025020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226025211, "dur": 0,
    "args": {
      "External id": 285450,
      "cbid": 317, "correlation": 285450
    }
  },
  {
    "ph": "f", "id": 285450, "pid": 76337, "tid": -914061504, "ts": 1716454226025211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226025212, "dur": 1,
    "args": {
      "External id": 285451,
      "cbid": 203, "correlation": 285451
    }
  },
  {
    "ph": "f", "id": 285451, "pid": 76337, "tid": -914061504, "ts": 1716454226025212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226025214, "dur": 0,
    "args": {
      "External id": 285452,
      "cbid": 205, "correlation": 285452
    }
  },
  {
    "ph": "f", "id": 285452, "pid": 76337, "tid": -914061504, "ts": 1716454226025214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454226028781, "dur": 3,
    "args": {
      "External id": 285456,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285456,
      "bytes": 196608, "memory bandwidth (GB/s)": 60.23529411764706
    }
  },
  {
    "ph": "f", "id": 285456, "pid": 5, "tid": 7, "ts": 1716454226028781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226028738, "dur": 43,
    "args": {
      "External id": 285456,
      "cbid": 51, "correlation": 285456
    }
  },
  {
    "ph": "s", "id": 285456, "pid": 76337, "tid": -914061504, "ts": 1716454226028738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226028801, "dur": 5,
    "args": {
      "External id": 285457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285457,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4.8,
      "warps per SM": 38.4,
      "grid": [96, 1, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 285457, "pid": 5, "tid": 7, "ts": 1716454226028801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226028788, "dur": 13,
    "args": {
      "External id": 285457,
      "cbid": 211, "correlation": 285457
    }
  },
  {
    "ph": "s", "id": 285457, "pid": 76337, "tid": -914061504, "ts": 1716454226028788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454226028813, "dur": 2,
    "args": {
      "External id": 285459,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 285459,
      "bytes": 73728, "memory bandwidth (GB/s)": 25.318681318681318
    }
  },
  {
    "ph": "f", "id": 285459, "pid": 5, "tid": 7, "ts": 1716454226028813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226028804, "dur": 7,
    "args": {
      "External id": 285459,
      "cbid": 51, "correlation": 285459
    }
  },
  {
    "ph": "s", "id": 285459, "pid": 76337, "tid": -914061504, "ts": 1716454226028804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226028820, "dur": 4,
    "args": {
      "External id": 285460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285460,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 1, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 285460, "pid": 5, "tid": 7, "ts": 1716454226028820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226028813, "dur": 6,
    "args": {
      "External id": 285460,
      "cbid": 211, "correlation": 285460
    }
  },
  {
    "ph": "s", "id": 285460, "pid": 76337, "tid": -914061504, "ts": 1716454226028813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x64x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226029081, "dur": 72,
    "args": {
      "External id": 285462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285462,
      "registers per thread": 184,
      "shared memory": 12288,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [8, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285462, "pid": 5, "tid": 7, "ts": 1716454226029081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226028838, "dur": 245,
    "args": {
      "External id": 285462,
      "cbid": 211, "correlation": 285462
    }
  },
  {
    "ph": "s", "id": 285462, "pid": 76337, "tid": -914061504, "ts": 1716454226028838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226029155, "dur": 31,
    "args": {
      "External id": 285464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285464,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285464, "pid": 5, "tid": 7, "ts": 1716454226029155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226029092, "dur": 8,
    "args": {
      "External id": 285464,
      "cbid": 211, "correlation": 285464
    }
  },
  {
    "ph": "s", "id": 285464, "pid": 76337, "tid": -914061504, "ts": 1716454226029092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226029986, "dur": 48,
    "args": {
      "External id": 285470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285470,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285470, "pid": 5, "tid": 7, "ts": 1716454226029986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226029964, "dur": 21,
    "args": {
      "External id": 285470,
      "cbid": 211, "correlation": 285470
    }
  },
  {
    "ph": "s", "id": 285470, "pid": 76337, "tid": -914061504, "ts": 1716454226029964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226030343, "dur": 57,
    "args": {
      "External id": 285490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285490,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 285490, "pid": 5, "tid": 7, "ts": 1716454226030343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030326, "dur": 16,
    "args": {
      "External id": 285490,
      "cbid": 211, "correlation": 285490
    }
  },
  {
    "ph": "s", "id": 285490, "pid": 76337, "tid": -914061504, "ts": 1716454226030326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226030401, "dur": 5,
    "args": {
      "External id": 285502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285502,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 285502, "pid": 5, "tid": 7, "ts": 1716454226030401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030355, "dur": 10,
    "args": {
      "External id": 285502,
      "cbid": 211, "correlation": 285502
    }
  },
  {
    "ph": "s", "id": 285502, "pid": 76337, "tid": -914061504, "ts": 1716454226030355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226030407, "dur": 43,
    "args": {
      "External id": 285505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285505,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285505, "pid": 5, "tid": 7, "ts": 1716454226030407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030389, "dur": 9,
    "args": {
      "External id": 285505,
      "cbid": 211, "correlation": 285505
    }
  },
  {
    "ph": "s", "id": 285505, "pid": 76337, "tid": -914061504, "ts": 1716454226030389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226030469, "dur": 31,
    "args": {
      "External id": 285514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285514,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285514, "pid": 5, "tid": 7, "ts": 1716454226030469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030457, "dur": 12,
    "args": {
      "External id": 285514,
      "cbid": 211, "correlation": 285514
    }
  },
  {
    "ph": "s", "id": 285514, "pid": 76337, "tid": -914061504, "ts": 1716454226030457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030541, "dur": 0,
    "args": {
      "External id": 285524,
      "cbid": 317, "correlation": 285524
    }
  },
  {
    "ph": "f", "id": 285524, "pid": 76337, "tid": -914061504, "ts": 1716454226030541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030543, "dur": 0,
    "args": {
      "External id": 285525,
      "cbid": 203, "correlation": 285525
    }
  },
  {
    "ph": "f", "id": 285525, "pid": 76337, "tid": -914061504, "ts": 1716454226030543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226030544, "dur": 0,
    "args": {
      "External id": 285526,
      "cbid": 205, "correlation": 285526
    }
  },
  {
    "ph": "f", "id": 285526, "pid": 76337, "tid": -914061504, "ts": 1716454226030544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226033677, "dur": 34,
    "args": {
      "External id": 285530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285530,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285530, "pid": 5, "tid": 7, "ts": 1716454226033677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226033652, "dur": 26,
    "args": {
      "External id": 285530,
      "cbid": 211, "correlation": 285530
    }
  },
  {
    "ph": "s", "id": 285530, "pid": 76337, "tid": -914061504, "ts": 1716454226033652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226033713, "dur": 28,
    "args": {
      "External id": 285532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285532,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285532, "pid": 5, "tid": 7, "ts": 1716454226033713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226033682, "dur": 6,
    "args": {
      "External id": 285532,
      "cbid": 211, "correlation": 285532
    }
  },
  {
    "ph": "s", "id": 285532, "pid": 76337, "tid": -914061504, "ts": 1716454226033682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226033742, "dur": 746,
    "args": {
      "External id": 285534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285534,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285534, "pid": 5, "tid": 7, "ts": 1716454226033742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226033706, "dur": 11,
    "args": {
      "External id": 285534,
      "cbid": 211, "correlation": 285534
    }
  },
  {
    "ph": "s", "id": 285534, "pid": 76337, "tid": -914061504, "ts": 1716454226033706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226034490, "dur": 30,
    "args": {
      "External id": 285536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285536,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285536, "pid": 5, "tid": 7, "ts": 1716454226034490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226033723, "dur": 7,
    "args": {
      "External id": 285536,
      "cbid": 211, "correlation": 285536
    }
  },
  {
    "ph": "s", "id": 285536, "pid": 76337, "tid": -914061504, "ts": 1716454226033723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226034541, "dur": 47,
    "args": {
      "External id": 285542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285542, "pid": 5, "tid": 7, "ts": 1716454226034541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034526, "dur": 14,
    "args": {
      "External id": 285542,
      "cbid": 211, "correlation": 285542
    }
  },
  {
    "ph": "s", "id": 285542, "pid": 76337, "tid": -914061504, "ts": 1716454226034526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226034684, "dur": 57,
    "args": {
      "External id": 285562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285562,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 285562, "pid": 5, "tid": 7, "ts": 1716454226034684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034671, "dur": 13,
    "args": {
      "External id": 285562,
      "cbid": 211, "correlation": 285562
    }
  },
  {
    "ph": "s", "id": 285562, "pid": 76337, "tid": -914061504, "ts": 1716454226034671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226034743, "dur": 4,
    "args": {
      "External id": 285574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285574,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 285574, "pid": 5, "tid": 7, "ts": 1716454226034743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034698, "dur": 9,
    "args": {
      "External id": 285574,
      "cbid": 211, "correlation": 285574
    }
  },
  {
    "ph": "s", "id": 285574, "pid": 76337, "tid": -914061504, "ts": 1716454226034698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226034748, "dur": 43,
    "args": {
      "External id": 285577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285577, "pid": 5, "tid": 7, "ts": 1716454226034748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034735, "dur": 8,
    "args": {
      "External id": 285577,
      "cbid": 211, "correlation": 285577
    }
  },
  {
    "ph": "s", "id": 285577, "pid": 76337, "tid": -914061504, "ts": 1716454226034735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226034804, "dur": 30,
    "args": {
      "External id": 285586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285586,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285586, "pid": 5, "tid": 7, "ts": 1716454226034804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034791, "dur": 12,
    "args": {
      "External id": 285586,
      "cbid": 211, "correlation": 285586
    }
  },
  {
    "ph": "s", "id": 285586, "pid": 76337, "tid": -914061504, "ts": 1716454226034791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034891, "dur": 0,
    "args": {
      "External id": 285596,
      "cbid": 317, "correlation": 285596
    }
  },
  {
    "ph": "f", "id": 285596, "pid": 76337, "tid": -914061504, "ts": 1716454226034891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034892, "dur": 1,
    "args": {
      "External id": 285597,
      "cbid": 203, "correlation": 285597
    }
  },
  {
    "ph": "f", "id": 285597, "pid": 76337, "tid": -914061504, "ts": 1716454226034892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034894, "dur": 0,
    "args": {
      "External id": 285598,
      "cbid": 205, "correlation": 285598
    }
  },
  {
    "ph": "f", "id": 285598, "pid": 76337, "tid": -914061504, "ts": 1716454226034894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226034923, "dur": 32,
    "args": {
      "External id": 285602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285602, "pid": 5, "tid": 7, "ts": 1716454226034923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034910, "dur": 12,
    "args": {
      "External id": 285602,
      "cbid": 211, "correlation": 285602
    }
  },
  {
    "ph": "s", "id": 285602, "pid": 76337, "tid": -914061504, "ts": 1716454226034910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226034956, "dur": 29,
    "args": {
      "External id": 285604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285604,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285604, "pid": 5, "tid": 7, "ts": 1716454226034956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034925, "dur": 6,
    "args": {
      "External id": 285604,
      "cbid": 211, "correlation": 285604
    }
  },
  {
    "ph": "s", "id": 285604, "pid": 76337, "tid": -914061504, "ts": 1716454226034925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226034986, "dur": 725,
    "args": {
      "External id": 285606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285606,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285606, "pid": 5, "tid": 7, "ts": 1716454226034986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034937, "dur": 6,
    "args": {
      "External id": 285606,
      "cbid": 211, "correlation": 285606
    }
  },
  {
    "ph": "s", "id": 285606, "pid": 76337, "tid": -914061504, "ts": 1716454226034937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226035712, "dur": 30,
    "args": {
      "External id": 285608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285608,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285608, "pid": 5, "tid": 7, "ts": 1716454226035712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034946, "dur": 6,
    "args": {
      "External id": 285608,
      "cbid": 211, "correlation": 285608
    }
  },
  {
    "ph": "s", "id": 285608, "pid": 76337, "tid": -914061504, "ts": 1716454226034946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226035744, "dur": 48,
    "args": {
      "External id": 285614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285614, "pid": 5, "tid": 7, "ts": 1716454226035744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226034983, "dur": 9,
    "args": {
      "External id": 285614,
      "cbid": 211, "correlation": 285614
    }
  },
  {
    "ph": "s", "id": 285614, "pid": 76337, "tid": -914061504, "ts": 1716454226034983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454226035793, "dur": 41,
    "args": {
      "External id": 285622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285622,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285622, "pid": 5, "tid": 7, "ts": 1716454226035793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035026, "dur": 10,
    "args": {
      "External id": 285622,
      "cbid": 211, "correlation": 285622
    }
  },
  {
    "ph": "s", "id": 285622, "pid": 76337, "tid": -914061504, "ts": 1716454226035026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226035836, "dur": 29,
    "args": {
      "External id": 285630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285630,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285630, "pid": 5, "tid": 7, "ts": 1716454226035836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035077, "dur": 10,
    "args": {
      "External id": 285630,
      "cbid": 211, "correlation": 285630
    }
  },
  {
    "ph": "s", "id": 285630, "pid": 76337, "tid": -914061504, "ts": 1716454226035077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226035866, "dur": 60,
    "args": {
      "External id": 285650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285650,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 285650, "pid": 5, "tid": 7, "ts": 1716454226035866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035310, "dur": 15,
    "args": {
      "External id": 285650,
      "cbid": 211, "correlation": 285650
    }
  },
  {
    "ph": "s", "id": 285650, "pid": 76337, "tid": -914061504, "ts": 1716454226035310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226035928, "dur": 4,
    "args": {
      "External id": 285662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285662,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 285662, "pid": 5, "tid": 7, "ts": 1716454226035928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035339, "dur": 7,
    "args": {
      "External id": 285662,
      "cbid": 211, "correlation": 285662
    }
  },
  {
    "ph": "s", "id": 285662, "pid": 76337, "tid": -914061504, "ts": 1716454226035339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226035933, "dur": 43,
    "args": {
      "External id": 285665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285665, "pid": 5, "tid": 7, "ts": 1716454226035933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035358, "dur": 7,
    "args": {
      "External id": 285665,
      "cbid": 211, "correlation": 285665
    }
  },
  {
    "ph": "s", "id": 285665, "pid": 76337, "tid": -914061504, "ts": 1716454226035358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035527, "dur": 104,
    "args": {
      "External id": 285677,
      "cbid": 251, "correlation": 285677
    }
  },
  {
    "ph": "f", "id": 285677, "pid": 76337, "tid": -914061504, "ts": 1716454226035527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035633, "dur": 1,
    "args": {
      "External id": 285678,
      "cbid": 251, "correlation": 285678
    }
  },
  {
    "ph": "f", "id": 285678, "pid": 76337, "tid": -914061504, "ts": 1716454226035633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035635, "dur": 45,
    "args": {
      "External id": 285679,
      "cbid": 251, "correlation": 285679
    }
  },
  {
    "ph": "f", "id": 285679, "pid": 76337, "tid": -914061504, "ts": 1716454226035635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035681, "dur": 3,
    "args": {
      "External id": 285680,
      "cbid": 251, "correlation": 285680
    }
  },
  {
    "ph": "f", "id": 285680, "pid": 76337, "tid": -914061504, "ts": 1716454226035681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035686, "dur": 51,
    "args": {
      "External id": 285681,
      "cbid": 251, "correlation": 285681
    }
  },
  {
    "ph": "f", "id": 285681, "pid": 76337, "tid": -914061504, "ts": 1716454226035686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035737, "dur": 1,
    "args": {
      "External id": 285682,
      "cbid": 251, "correlation": 285682
    }
  },
  {
    "ph": "f", "id": 285682, "pid": 76337, "tid": -914061504, "ts": 1716454226035737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035739, "dur": 1,
    "args": {
      "External id": 285683,
      "cbid": 251, "correlation": 285683
    }
  },
  {
    "ph": "f", "id": 285683, "pid": 76337, "tid": -914061504, "ts": 1716454226035739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035741, "dur": 40,
    "args": {
      "External id": 285684,
      "cbid": 251, "correlation": 285684
    }
  },
  {
    "ph": "f", "id": 285684, "pid": 76337, "tid": -914061504, "ts": 1716454226035741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035783, "dur": 0,
    "args": {
      "External id": 285685,
      "cbid": 251, "correlation": 285685
    }
  },
  {
    "ph": "f", "id": 285685, "pid": 76337, "tid": -914061504, "ts": 1716454226035783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454226036003, "dur": 104,
    "args": {
      "External id": 285686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285686,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [8, 24, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 285686, "pid": 5, "tid": 7, "ts": 1716454226036003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035793, "dur": 16,
    "args": {
      "External id": 285686,
      "cbid": 211, "correlation": 285686
    }
  },
  {
    "ph": "s", "id": 285686, "pid": 76337, "tid": -914061504, "ts": 1716454226035793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226036109, "dur": 46,
    "args": {
      "External id": 285691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285691, "pid": 5, "tid": 7, "ts": 1716454226036109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035842, "dur": 10,
    "args": {
      "External id": 285691,
      "cbid": 211, "correlation": 285691
    }
  },
  {
    "ph": "s", "id": 285691, "pid": 76337, "tid": -914061504, "ts": 1716454226035842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035926, "dur": 1,
    "args": {
      "External id": 285702,
      "cbid": 251, "correlation": 285702
    }
  },
  {
    "ph": "f", "id": 285702, "pid": 76337, "tid": -914061504, "ts": 1716454226035926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035928, "dur": 0,
    "args": {
      "External id": 285703,
      "cbid": 251, "correlation": 285703
    }
  },
  {
    "ph": "f", "id": 285703, "pid": 76337, "tid": -914061504, "ts": 1716454226035928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035928, "dur": 0,
    "args": {
      "External id": 285704,
      "cbid": 251, "correlation": 285704
    }
  },
  {
    "ph": "f", "id": 285704, "pid": 76337, "tid": -914061504, "ts": 1716454226035928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035929, "dur": 0,
    "args": {
      "External id": 285705,
      "cbid": 251, "correlation": 285705
    }
  },
  {
    "ph": "f", "id": 285705, "pid": 76337, "tid": -914061504, "ts": 1716454226035929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035930, "dur": 0,
    "args": {
      "External id": 285706,
      "cbid": 251, "correlation": 285706
    }
  },
  {
    "ph": "f", "id": 285706, "pid": 76337, "tid": -914061504, "ts": 1716454226035930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035930, "dur": 0,
    "args": {
      "External id": 285707,
      "cbid": 251, "correlation": 285707
    }
  },
  {
    "ph": "f", "id": 285707, "pid": 76337, "tid": -914061504, "ts": 1716454226035930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035931, "dur": 0,
    "args": {
      "External id": 285708,
      "cbid": 251, "correlation": 285708
    }
  },
  {
    "ph": "f", "id": 285708, "pid": 76337, "tid": -914061504, "ts": 1716454226035931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035932, "dur": 0,
    "args": {
      "External id": 285709,
      "cbid": 251, "correlation": 285709
    }
  },
  {
    "ph": "f", "id": 285709, "pid": 76337, "tid": -914061504, "ts": 1716454226035932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035933, "dur": 0,
    "args": {
      "External id": 285710,
      "cbid": 251, "correlation": 285710
    }
  },
  {
    "ph": "f", "id": 285710, "pid": 76337, "tid": -914061504, "ts": 1716454226035933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454226036157, "dur": 101,
    "args": {
      "External id": 285711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285711,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [8, 24, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 285711, "pid": 5, "tid": 7, "ts": 1716454226036157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035936, "dur": 13,
    "args": {
      "External id": 285711,
      "cbid": 211, "correlation": 285711
    }
  },
  {
    "ph": "s", "id": 285711, "pid": 76337, "tid": -914061504, "ts": 1716454226035936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226036259, "dur": 46,
    "args": {
      "External id": 285716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285716, "pid": 5, "tid": 7, "ts": 1716454226036259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226035967, "dur": 18,
    "args": {
      "External id": 285716,
      "cbid": 211, "correlation": 285716
    }
  },
  {
    "ph": "s", "id": 285716, "pid": 76337, "tid": -914061504, "ts": 1716454226035967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036048, "dur": 1,
    "args": {
      "External id": 285727,
      "cbid": 251, "correlation": 285727
    }
  },
  {
    "ph": "f", "id": 285727, "pid": 76337, "tid": -914061504, "ts": 1716454226036048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036050, "dur": 0,
    "args": {
      "External id": 285728,
      "cbid": 251, "correlation": 285728
    }
  },
  {
    "ph": "f", "id": 285728, "pid": 76337, "tid": -914061504, "ts": 1716454226036050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036050, "dur": 0,
    "args": {
      "External id": 285729,
      "cbid": 251, "correlation": 285729
    }
  },
  {
    "ph": "f", "id": 285729, "pid": 76337, "tid": -914061504, "ts": 1716454226036050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036051, "dur": 0,
    "args": {
      "External id": 285730,
      "cbid": 251, "correlation": 285730
    }
  },
  {
    "ph": "f", "id": 285730, "pid": 76337, "tid": -914061504, "ts": 1716454226036051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036051, "dur": 0,
    "args": {
      "External id": 285731,
      "cbid": 251, "correlation": 285731
    }
  },
  {
    "ph": "f", "id": 285731, "pid": 76337, "tid": -914061504, "ts": 1716454226036051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036052, "dur": 0,
    "args": {
      "External id": 285732,
      "cbid": 251, "correlation": 285732
    }
  },
  {
    "ph": "f", "id": 285732, "pid": 76337, "tid": -914061504, "ts": 1716454226036052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036053, "dur": 0,
    "args": {
      "External id": 285733,
      "cbid": 251, "correlation": 285733
    }
  },
  {
    "ph": "f", "id": 285733, "pid": 76337, "tid": -914061504, "ts": 1716454226036053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036053, "dur": 0,
    "args": {
      "External id": 285734,
      "cbid": 251, "correlation": 285734
    }
  },
  {
    "ph": "f", "id": 285734, "pid": 76337, "tid": -914061504, "ts": 1716454226036053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036054, "dur": 0,
    "args": {
      "External id": 285735,
      "cbid": 251, "correlation": 285735
    }
  },
  {
    "ph": "f", "id": 285735, "pid": 76337, "tid": -914061504, "ts": 1716454226036054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_tt", "pid": 5, "tid": 7,
    "ts": 1716454226036306, "dur": 104,
    "args": {
      "External id": 285736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285736,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [8, 24, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 285736, "pid": 5, "tid": 7, "ts": 1716454226036306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036058, "dur": 13,
    "args": {
      "External id": 285736,
      "cbid": 211, "correlation": 285736
    }
  },
  {
    "ph": "s", "id": 285736, "pid": 76337, "tid": -914061504, "ts": 1716454226036058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226036412, "dur": 46,
    "args": {
      "External id": 285741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285741, "pid": 5, "tid": 7, "ts": 1716454226036412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036090, "dur": 9,
    "args": {
      "External id": 285741,
      "cbid": 211, "correlation": 285741
    }
  },
  {
    "ph": "s", "id": 285741, "pid": 76337, "tid": -914061504, "ts": 1716454226036090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454226036459, "dur": 3561,
    "args": {
      "External id": 285766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285766,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [96, 1, 4],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285766, "pid": 5, "tid": 7, "ts": 1716454226036459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036195, "dur": 15,
    "args": {
      "External id": 285766,
      "cbid": 211, "correlation": 285766
    }
  },
  {
    "ph": "s", "id": 285766, "pid": 76337, "tid": -914061504, "ts": 1716454226036195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036349, "dur": 2,
    "args": {
      "External id": 285784,
      "cbid": 251, "correlation": 285784
    }
  },
  {
    "ph": "f", "id": 285784, "pid": 76337, "tid": -914061504, "ts": 1716454226036349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036355, "dur": 0,
    "args": {
      "External id": 285786,
      "cbid": 251, "correlation": 285786
    }
  },
  {
    "ph": "f", "id": 285786, "pid": 76337, "tid": -914061504, "ts": 1716454226036355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454226040021, "dur": 112,
    "args": {
      "External id": 285787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285787,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [4, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285787, "pid": 5, "tid": 7, "ts": 1716454226040021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036361, "dur": 15,
    "args": {
      "External id": 285787,
      "cbid": 211, "correlation": 285787
    }
  },
  {
    "ph": "s", "id": 285787, "pid": 76337, "tid": -914061504, "ts": 1716454226036361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226040135, "dur": 122,
    "args": {
      "External id": 285795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285795,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285795, "pid": 5, "tid": 7, "ts": 1716454226040135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036461, "dur": 14,
    "args": {
      "External id": 285795,
      "cbid": 211, "correlation": 285795
    }
  },
  {
    "ph": "s", "id": 285795, "pid": 76337, "tid": -914061504, "ts": 1716454226036461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226040259, "dur": 28,
    "args": {
      "External id": 285803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285803,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285803, "pid": 5, "tid": 7, "ts": 1716454226040259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036501, "dur": 10,
    "args": {
      "External id": 285803,
      "cbid": 211, "correlation": 285803
    }
  },
  {
    "ph": "s", "id": 285803, "pid": 76337, "tid": -914061504, "ts": 1716454226036501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226040288, "dur": 117,
    "args": {
      "External id": 285813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285813, "pid": 5, "tid": 7, "ts": 1716454226040288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036600, "dur": 16,
    "args": {
      "External id": 285813,
      "cbid": 211, "correlation": 285813
    }
  },
  {
    "ph": "s", "id": 285813, "pid": 76337, "tid": -914061504, "ts": 1716454226036600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226040406, "dur": 61,
    "args": {
      "External id": 285834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285834,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 285834, "pid": 5, "tid": 7, "ts": 1716454226040406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036639, "dur": 8,
    "args": {
      "External id": 285834,
      "cbid": 211, "correlation": 285834
    }
  },
  {
    "ph": "s", "id": 285834, "pid": 76337, "tid": -914061504, "ts": 1716454226036639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226040469, "dur": 4,
    "args": {
      "External id": 285846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285846,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 285846, "pid": 5, "tid": 7, "ts": 1716454226040469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036659, "dur": 6,
    "args": {
      "External id": 285846,
      "cbid": 211, "correlation": 285846
    }
  },
  {
    "ph": "s", "id": 285846, "pid": 76337, "tid": -914061504, "ts": 1716454226036659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226040474, "dur": 45,
    "args": {
      "External id": 285849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285849, "pid": 5, "tid": 7, "ts": 1716454226040474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036675, "dur": 6,
    "args": {
      "External id": 285849,
      "cbid": 211, "correlation": 285849
    }
  },
  {
    "ph": "s", "id": 285849, "pid": 76337, "tid": -914061504, "ts": 1716454226036675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226040521, "dur": 30,
    "args": {
      "External id": 285858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285858,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285858, "pid": 5, "tid": 7, "ts": 1716454226040521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036717, "dur": 9,
    "args": {
      "External id": 285858,
      "cbid": 211, "correlation": 285858
    }
  },
  {
    "ph": "s", "id": 285858, "pid": 76337, "tid": -914061504, "ts": 1716454226036717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036780, "dur": 0,
    "args": {
      "External id": 285868,
      "cbid": 317, "correlation": 285868
    }
  },
  {
    "ph": "f", "id": 285868, "pid": 76337, "tid": -914061504, "ts": 1716454226036780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036781, "dur": 0,
    "args": {
      "External id": 285869,
      "cbid": 203, "correlation": 285869
    }
  },
  {
    "ph": "f", "id": 285869, "pid": 76337, "tid": -914061504, "ts": 1716454226036781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036782, "dur": 0,
    "args": {
      "External id": 285870,
      "cbid": 205, "correlation": 285870
    }
  },
  {
    "ph": "f", "id": 285870, "pid": 76337, "tid": -914061504, "ts": 1716454226036782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226040552, "dur": 34,
    "args": {
      "External id": 285874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285874,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285874, "pid": 5, "tid": 7, "ts": 1716454226040552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036796, "dur": 13,
    "args": {
      "External id": 285874,
      "cbid": 211, "correlation": 285874
    }
  },
  {
    "ph": "s", "id": 285874, "pid": 76337, "tid": -914061504, "ts": 1716454226036796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226040588, "dur": 28,
    "args": {
      "External id": 285876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285876,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285876, "pid": 5, "tid": 7, "ts": 1716454226040588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036811, "dur": 5,
    "args": {
      "External id": 285876,
      "cbid": 211, "correlation": 285876
    }
  },
  {
    "ph": "s", "id": 285876, "pid": 76337, "tid": -914061504, "ts": 1716454226036811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226040618, "dur": 743,
    "args": {
      "External id": 285878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285878,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285878, "pid": 5, "tid": 7, "ts": 1716454226040618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036823, "dur": 6,
    "args": {
      "External id": 285878,
      "cbid": 211, "correlation": 285878
    }
  },
  {
    "ph": "s", "id": 285878, "pid": 76337, "tid": -914061504, "ts": 1716454226036823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226041362, "dur": 30,
    "args": {
      "External id": 285880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285880, "pid": 5, "tid": 7, "ts": 1716454226041362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036833, "dur": 5,
    "args": {
      "External id": 285880,
      "cbid": 211, "correlation": 285880
    }
  },
  {
    "ph": "s", "id": 285880, "pid": 76337, "tid": -914061504, "ts": 1716454226036833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226041394, "dur": 47,
    "args": {
      "External id": 285886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285886, "pid": 5, "tid": 7, "ts": 1716454226041394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036862, "dur": 9,
    "args": {
      "External id": 285886,
      "cbid": 211, "correlation": 285886
    }
  },
  {
    "ph": "s", "id": 285886, "pid": 76337, "tid": -914061504, "ts": 1716454226036862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226041442, "dur": 59,
    "args": {
      "External id": 285906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285906,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 285906, "pid": 5, "tid": 7, "ts": 1716454226041442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036936, "dur": 12,
    "args": {
      "External id": 285906,
      "cbid": 211, "correlation": 285906
    }
  },
  {
    "ph": "s", "id": 285906, "pid": 76337, "tid": -914061504, "ts": 1716454226036936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226041502, "dur": 4,
    "args": {
      "External id": 285918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285918,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 285918, "pid": 5, "tid": 7, "ts": 1716454226041502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036959, "dur": 6,
    "args": {
      "External id": 285918,
      "cbid": 211, "correlation": 285918
    }
  },
  {
    "ph": "s", "id": 285918, "pid": 76337, "tid": -914061504, "ts": 1716454226036959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226041507, "dur": 44,
    "args": {
      "External id": 285921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285921, "pid": 5, "tid": 7, "ts": 1716454226041507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226036987, "dur": 7,
    "args": {
      "External id": 285921,
      "cbid": 211, "correlation": 285921
    }
  },
  {
    "ph": "s", "id": 285921, "pid": 76337, "tid": -914061504, "ts": 1716454226036987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226041552, "dur": 29,
    "args": {
      "External id": 285930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285930,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285930, "pid": 5, "tid": 7, "ts": 1716454226041552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037029, "dur": 10,
    "args": {
      "External id": 285930,
      "cbid": 211, "correlation": 285930
    }
  },
  {
    "ph": "s", "id": 285930, "pid": 76337, "tid": -914061504, "ts": 1716454226037029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037094, "dur": 0,
    "args": {
      "External id": 285940,
      "cbid": 317, "correlation": 285940
    }
  },
  {
    "ph": "f", "id": 285940, "pid": 76337, "tid": -914061504, "ts": 1716454226037094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037095, "dur": 0,
    "args": {
      "External id": 285941,
      "cbid": 203, "correlation": 285941
    }
  },
  {
    "ph": "f", "id": 285941, "pid": 76337, "tid": -914061504, "ts": 1716454226037095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037096, "dur": 0,
    "args": {
      "External id": 285942,
      "cbid": 205, "correlation": 285942
    }
  },
  {
    "ph": "f", "id": 285942, "pid": 76337, "tid": -914061504, "ts": 1716454226037096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226041583, "dur": 33,
    "args": {
      "External id": 285946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285946, "pid": 5, "tid": 7, "ts": 1716454226041583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037109, "dur": 12,
    "args": {
      "External id": 285946,
      "cbid": 211, "correlation": 285946
    }
  },
  {
    "ph": "s", "id": 285946, "pid": 76337, "tid": -914061504, "ts": 1716454226037109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226041617, "dur": 28,
    "args": {
      "External id": 285948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285948, "pid": 5, "tid": 7, "ts": 1716454226041617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037124, "dur": 5,
    "args": {
      "External id": 285948,
      "cbid": 211, "correlation": 285948
    }
  },
  {
    "ph": "s", "id": 285948, "pid": 76337, "tid": -914061504, "ts": 1716454226037124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226041647, "dur": 727,
    "args": {
      "External id": 285950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285950,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 285950, "pid": 5, "tid": 7, "ts": 1716454226041647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037135, "dur": 6,
    "args": {
      "External id": 285950,
      "cbid": 211, "correlation": 285950
    }
  },
  {
    "ph": "s", "id": 285950, "pid": 76337, "tid": -914061504, "ts": 1716454226037135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226042375, "dur": 30,
    "args": {
      "External id": 285952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285952, "pid": 5, "tid": 7, "ts": 1716454226042375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037144, "dur": 5,
    "args": {
      "External id": 285952,
      "cbid": 211, "correlation": 285952
    }
  },
  {
    "ph": "s", "id": 285952, "pid": 76337, "tid": -914061504, "ts": 1716454226037144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226042406, "dur": 47,
    "args": {
      "External id": 285958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285958, "pid": 5, "tid": 7, "ts": 1716454226042406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037172, "dur": 8,
    "args": {
      "External id": 285958,
      "cbid": 211, "correlation": 285958
    }
  },
  {
    "ph": "s", "id": 285958, "pid": 76337, "tid": -914061504, "ts": 1716454226037172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226042454, "dur": 123,
    "args": {
      "External id": 285966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285966,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285966, "pid": 5, "tid": 7, "ts": 1716454226042454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037207, "dur": 9,
    "args": {
      "External id": 285966,
      "cbid": 211, "correlation": 285966
    }
  },
  {
    "ph": "s", "id": 285966, "pid": 76337, "tid": -914061504, "ts": 1716454226037207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226042578, "dur": 28,
    "args": {
      "External id": 285974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285974,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285974, "pid": 5, "tid": 7, "ts": 1716454226042578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037237, "dur": 8,
    "args": {
      "External id": 285974,
      "cbid": 211, "correlation": 285974
    }
  },
  {
    "ph": "s", "id": 285974, "pid": 76337, "tid": -914061504, "ts": 1716454226037237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226042607, "dur": 117,
    "args": {
      "External id": 285984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 285984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 285984, "pid": 5, "tid": 7, "ts": 1716454226042607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037343, "dur": 14,
    "args": {
      "External id": 285984,
      "cbid": 211, "correlation": 285984
    }
  },
  {
    "ph": "s", "id": 285984, "pid": 76337, "tid": -914061504, "ts": 1716454226037343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226042726, "dur": 61,
    "args": {
      "External id": 286005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286005,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286005, "pid": 5, "tid": 7, "ts": 1716454226042726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037381, "dur": 8,
    "args": {
      "External id": 286005,
      "cbid": 211, "correlation": 286005
    }
  },
  {
    "ph": "s", "id": 286005, "pid": 76337, "tid": -914061504, "ts": 1716454226037381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226042788, "dur": 4,
    "args": {
      "External id": 286017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286017,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286017, "pid": 5, "tid": 7, "ts": 1716454226042788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037398, "dur": 6,
    "args": {
      "External id": 286017,
      "cbid": 211, "correlation": 286017
    }
  },
  {
    "ph": "s", "id": 286017, "pid": 76337, "tid": -914061504, "ts": 1716454226037398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226042793, "dur": 43,
    "args": {
      "External id": 286020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286020,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286020, "pid": 5, "tid": 7, "ts": 1716454226042793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037416, "dur": 6,
    "args": {
      "External id": 286020,
      "cbid": 211, "correlation": 286020
    }
  },
  {
    "ph": "s", "id": 286020, "pid": 76337, "tid": -914061504, "ts": 1716454226037416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226042838, "dur": 30,
    "args": {
      "External id": 286029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286029,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286029, "pid": 5, "tid": 7, "ts": 1716454226042838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037455, "dur": 10,
    "args": {
      "External id": 286029,
      "cbid": 211, "correlation": 286029
    }
  },
  {
    "ph": "s", "id": 286029, "pid": 76337, "tid": -914061504, "ts": 1716454226037455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037517, "dur": 0,
    "args": {
      "External id": 286039,
      "cbid": 317, "correlation": 286039
    }
  },
  {
    "ph": "f", "id": 286039, "pid": 76337, "tid": -914061504, "ts": 1716454226037517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037518, "dur": 0,
    "args": {
      "External id": 286040,
      "cbid": 203, "correlation": 286040
    }
  },
  {
    "ph": "f", "id": 286040, "pid": 76337, "tid": -914061504, "ts": 1716454226037518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037519, "dur": 0,
    "args": {
      "External id": 286041,
      "cbid": 205, "correlation": 286041
    }
  },
  {
    "ph": "f", "id": 286041, "pid": 76337, "tid": -914061504, "ts": 1716454226037519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226042869, "dur": 34,
    "args": {
      "External id": 286045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286045,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286045, "pid": 5, "tid": 7, "ts": 1716454226042869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037535, "dur": 12,
    "args": {
      "External id": 286045,
      "cbid": 211, "correlation": 286045
    }
  },
  {
    "ph": "s", "id": 286045, "pid": 76337, "tid": -914061504, "ts": 1716454226037535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226042904, "dur": 29,
    "args": {
      "External id": 286047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286047,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286047, "pid": 5, "tid": 7, "ts": 1716454226042904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037549, "dur": 5,
    "args": {
      "External id": 286047,
      "cbid": 211, "correlation": 286047
    }
  },
  {
    "ph": "s", "id": 286047, "pid": 76337, "tid": -914061504, "ts": 1716454226037549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226042934, "dur": 726,
    "args": {
      "External id": 286049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286049,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286049, "pid": 5, "tid": 7, "ts": 1716454226042934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037561, "dur": 6,
    "args": {
      "External id": 286049,
      "cbid": 211, "correlation": 286049
    }
  },
  {
    "ph": "s", "id": 286049, "pid": 76337, "tid": -914061504, "ts": 1716454226037561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226043661, "dur": 30,
    "args": {
      "External id": 286051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286051,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286051, "pid": 5, "tid": 7, "ts": 1716454226043661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037570, "dur": 5,
    "args": {
      "External id": 286051,
      "cbid": 211, "correlation": 286051
    }
  },
  {
    "ph": "s", "id": 286051, "pid": 76337, "tid": -914061504, "ts": 1716454226037570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226043693, "dur": 48,
    "args": {
      "External id": 286057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286057, "pid": 5, "tid": 7, "ts": 1716454226043693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037597, "dur": 9,
    "args": {
      "External id": 286057,
      "cbid": 211, "correlation": 286057
    }
  },
  {
    "ph": "s", "id": 286057, "pid": 76337, "tid": -914061504, "ts": 1716454226037597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226043742, "dur": 57,
    "args": {
      "External id": 286077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286077,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286077, "pid": 5, "tid": 7, "ts": 1716454226043742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037672, "dur": 12,
    "args": {
      "External id": 286077,
      "cbid": 211, "correlation": 286077
    }
  },
  {
    "ph": "s", "id": 286077, "pid": 76337, "tid": -914061504, "ts": 1716454226037672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226043800, "dur": 4,
    "args": {
      "External id": 286089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286089,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286089, "pid": 5, "tid": 7, "ts": 1716454226043800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037696, "dur": 6,
    "args": {
      "External id": 286089,
      "cbid": 211, "correlation": 286089
    }
  },
  {
    "ph": "s", "id": 286089, "pid": 76337, "tid": -914061504, "ts": 1716454226037696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226043805, "dur": 43,
    "args": {
      "External id": 286092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286092, "pid": 5, "tid": 7, "ts": 1716454226043805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037714, "dur": 6,
    "args": {
      "External id": 286092,
      "cbid": 211, "correlation": 286092
    }
  },
  {
    "ph": "s", "id": 286092, "pid": 76337, "tid": -914061504, "ts": 1716454226037714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226043849, "dur": 30,
    "args": {
      "External id": 286101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286101,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286101, "pid": 5, "tid": 7, "ts": 1716454226043849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037753, "dur": 11,
    "args": {
      "External id": 286101,
      "cbid": 211, "correlation": 286101
    }
  },
  {
    "ph": "s", "id": 286101, "pid": 76337, "tid": -914061504, "ts": 1716454226037753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037821, "dur": 0,
    "args": {
      "External id": 286111,
      "cbid": 317, "correlation": 286111
    }
  },
  {
    "ph": "f", "id": 286111, "pid": 76337, "tid": -914061504, "ts": 1716454226037821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037821, "dur": 0,
    "args": {
      "External id": 286112,
      "cbid": 203, "correlation": 286112
    }
  },
  {
    "ph": "f", "id": 286112, "pid": 76337, "tid": -914061504, "ts": 1716454226037821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037822, "dur": 0,
    "args": {
      "External id": 286113,
      "cbid": 205, "correlation": 286113
    }
  },
  {
    "ph": "f", "id": 286113, "pid": 76337, "tid": -914061504, "ts": 1716454226037822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226043880, "dur": 32,
    "args": {
      "External id": 286117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286117, "pid": 5, "tid": 7, "ts": 1716454226043880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037837, "dur": 12,
    "args": {
      "External id": 286117,
      "cbid": 211, "correlation": 286117
    }
  },
  {
    "ph": "s", "id": 286117, "pid": 76337, "tid": -914061504, "ts": 1716454226037837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226043914, "dur": 28,
    "args": {
      "External id": 286119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286119,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286119, "pid": 5, "tid": 7, "ts": 1716454226043914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037851, "dur": 5,
    "args": {
      "External id": 286119,
      "cbid": 211, "correlation": 286119
    }
  },
  {
    "ph": "s", "id": 286119, "pid": 76337, "tid": -914061504, "ts": 1716454226037851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226043943, "dur": 726,
    "args": {
      "External id": 286121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286121,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286121, "pid": 5, "tid": 7, "ts": 1716454226043943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037863, "dur": 6,
    "args": {
      "External id": 286121,
      "cbid": 211, "correlation": 286121
    }
  },
  {
    "ph": "s", "id": 286121, "pid": 76337, "tid": -914061504, "ts": 1716454226037863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226044671, "dur": 32,
    "args": {
      "External id": 286123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286123, "pid": 5, "tid": 7, "ts": 1716454226044671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037872, "dur": 5,
    "args": {
      "External id": 286123,
      "cbid": 211, "correlation": 286123
    }
  },
  {
    "ph": "s", "id": 286123, "pid": 76337, "tid": -914061504, "ts": 1716454226037872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226044705, "dur": 48,
    "args": {
      "External id": 286129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286129, "pid": 5, "tid": 7, "ts": 1716454226044705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037899, "dur": 8,
    "args": {
      "External id": 286129,
      "cbid": 211, "correlation": 286129
    }
  },
  {
    "ph": "s", "id": 286129, "pid": 76337, "tid": -914061504, "ts": 1716454226037899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226044754, "dur": 123,
    "args": {
      "External id": 286137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286137,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286137, "pid": 5, "tid": 7, "ts": 1716454226044754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037933, "dur": 9,
    "args": {
      "External id": 286137,
      "cbid": 211, "correlation": 286137
    }
  },
  {
    "ph": "s", "id": 286137, "pid": 76337, "tid": -914061504, "ts": 1716454226037933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226044878, "dur": 28,
    "args": {
      "External id": 286145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286145,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286145, "pid": 5, "tid": 7, "ts": 1716454226044878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226037964, "dur": 8,
    "args": {
      "External id": 286145,
      "cbid": 211, "correlation": 286145
    }
  },
  {
    "ph": "s", "id": 286145, "pid": 76337, "tid": -914061504, "ts": 1716454226037964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226044907, "dur": 117,
    "args": {
      "External id": 286155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286155, "pid": 5, "tid": 7, "ts": 1716454226044907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038046, "dur": 13,
    "args": {
      "External id": 286155,
      "cbid": 211, "correlation": 286155
    }
  },
  {
    "ph": "s", "id": 286155, "pid": 76337, "tid": -914061504, "ts": 1716454226038046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226045025, "dur": 59,
    "args": {
      "External id": 286176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286176,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286176, "pid": 5, "tid": 7, "ts": 1716454226045025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038083, "dur": 8,
    "args": {
      "External id": 286176,
      "cbid": 211, "correlation": 286176
    }
  },
  {
    "ph": "s", "id": 286176, "pid": 76337, "tid": -914061504, "ts": 1716454226038083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226045085, "dur": 4,
    "args": {
      "External id": 286188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286188,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286188, "pid": 5, "tid": 7, "ts": 1716454226045085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038100, "dur": 6,
    "args": {
      "External id": 286188,
      "cbid": 211, "correlation": 286188
    }
  },
  {
    "ph": "s", "id": 286188, "pid": 76337, "tid": -914061504, "ts": 1716454226038100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226045091, "dur": 44,
    "args": {
      "External id": 286191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286191, "pid": 5, "tid": 7, "ts": 1716454226045091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038118, "dur": 6,
    "args": {
      "External id": 286191,
      "cbid": 211, "correlation": 286191
    }
  },
  {
    "ph": "s", "id": 286191, "pid": 76337, "tid": -914061504, "ts": 1716454226038118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226045136, "dur": 30,
    "args": {
      "External id": 286200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286200,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286200, "pid": 5, "tid": 7, "ts": 1716454226045136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038157, "dur": 10,
    "args": {
      "External id": 286200,
      "cbid": 211, "correlation": 286200
    }
  },
  {
    "ph": "s", "id": 286200, "pid": 76337, "tid": -914061504, "ts": 1716454226038157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038209, "dur": 0,
    "args": {
      "External id": 286210,
      "cbid": 317, "correlation": 286210
    }
  },
  {
    "ph": "f", "id": 286210, "pid": 76337, "tid": -914061504, "ts": 1716454226038209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038209, "dur": 0,
    "args": {
      "External id": 286211,
      "cbid": 203, "correlation": 286211
    }
  },
  {
    "ph": "f", "id": 286211, "pid": 76337, "tid": -914061504, "ts": 1716454226038209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038210, "dur": 0,
    "args": {
      "External id": 286212,
      "cbid": 205, "correlation": 286212
    }
  },
  {
    "ph": "f", "id": 286212, "pid": 76337, "tid": -914061504, "ts": 1716454226038210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226045168, "dur": 32,
    "args": {
      "External id": 286216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286216,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286216, "pid": 5, "tid": 7, "ts": 1716454226045168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038224, "dur": 11,
    "args": {
      "External id": 286216,
      "cbid": 211, "correlation": 286216
    }
  },
  {
    "ph": "s", "id": 286216, "pid": 76337, "tid": -914061504, "ts": 1716454226038224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226045201, "dur": 28,
    "args": {
      "External id": 286218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286218,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286218, "pid": 5, "tid": 7, "ts": 1716454226045201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038238, "dur": 5,
    "args": {
      "External id": 286218,
      "cbid": 211, "correlation": 286218
    }
  },
  {
    "ph": "s", "id": 286218, "pid": 76337, "tid": -914061504, "ts": 1716454226038238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226045230, "dur": 728,
    "args": {
      "External id": 286220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286220,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286220, "pid": 5, "tid": 7, "ts": 1716454226045230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038249, "dur": 6,
    "args": {
      "External id": 286220,
      "cbid": 211, "correlation": 286220
    }
  },
  {
    "ph": "s", "id": 286220, "pid": 76337, "tid": -914061504, "ts": 1716454226038249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226045959, "dur": 30,
    "args": {
      "External id": 286222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286222,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286222, "pid": 5, "tid": 7, "ts": 1716454226045959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038258, "dur": 6,
    "args": {
      "External id": 286222,
      "cbid": 211, "correlation": 286222
    }
  },
  {
    "ph": "s", "id": 286222, "pid": 76337, "tid": -914061504, "ts": 1716454226038258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226045991, "dur": 47,
    "args": {
      "External id": 286228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286228,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286228, "pid": 5, "tid": 7, "ts": 1716454226045991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038286, "dur": 8,
    "args": {
      "External id": 286228,
      "cbid": 211, "correlation": 286228
    }
  },
  {
    "ph": "s", "id": 286228, "pid": 76337, "tid": -914061504, "ts": 1716454226038286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226046039, "dur": 57,
    "args": {
      "External id": 286248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286248,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286248, "pid": 5, "tid": 7, "ts": 1716454226046039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038360, "dur": 12,
    "args": {
      "External id": 286248,
      "cbid": 211, "correlation": 286248
    }
  },
  {
    "ph": "s", "id": 286248, "pid": 76337, "tid": -914061504, "ts": 1716454226038360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226046097, "dur": 4,
    "args": {
      "External id": 286260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286260,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286260, "pid": 5, "tid": 7, "ts": 1716454226046097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038382, "dur": 6,
    "args": {
      "External id": 286260,
      "cbid": 211, "correlation": 286260
    }
  },
  {
    "ph": "s", "id": 286260, "pid": 76337, "tid": -914061504, "ts": 1716454226038382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226046102, "dur": 44,
    "args": {
      "External id": 286263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286263, "pid": 5, "tid": 7, "ts": 1716454226046102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038399, "dur": 6,
    "args": {
      "External id": 286263,
      "cbid": 211, "correlation": 286263
    }
  },
  {
    "ph": "s", "id": 286263, "pid": 76337, "tid": -914061504, "ts": 1716454226038399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226046147, "dur": 29,
    "args": {
      "External id": 286272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286272,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286272, "pid": 5, "tid": 7, "ts": 1716454226046147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038438, "dur": 10,
    "args": {
      "External id": 286272,
      "cbid": 211, "correlation": 286272
    }
  },
  {
    "ph": "s", "id": 286272, "pid": 76337, "tid": -914061504, "ts": 1716454226038438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038504, "dur": 0,
    "args": {
      "External id": 286282,
      "cbid": 317, "correlation": 286282
    }
  },
  {
    "ph": "f", "id": 286282, "pid": 76337, "tid": -914061504, "ts": 1716454226038504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038505, "dur": 0,
    "args": {
      "External id": 286283,
      "cbid": 203, "correlation": 286283
    }
  },
  {
    "ph": "f", "id": 286283, "pid": 76337, "tid": -914061504, "ts": 1716454226038505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038506, "dur": 0,
    "args": {
      "External id": 286284,
      "cbid": 205, "correlation": 286284
    }
  },
  {
    "ph": "f", "id": 286284, "pid": 76337, "tid": -914061504, "ts": 1716454226038506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226046178, "dur": 32,
    "args": {
      "External id": 286288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286288, "pid": 5, "tid": 7, "ts": 1716454226046178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038520, "dur": 12,
    "args": {
      "External id": 286288,
      "cbid": 211, "correlation": 286288
    }
  },
  {
    "ph": "s", "id": 286288, "pid": 76337, "tid": -914061504, "ts": 1716454226038520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226046211, "dur": 28,
    "args": {
      "External id": 286290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286290,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286290, "pid": 5, "tid": 7, "ts": 1716454226046211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038534, "dur": 5,
    "args": {
      "External id": 286290,
      "cbid": 211, "correlation": 286290
    }
  },
  {
    "ph": "s", "id": 286290, "pid": 76337, "tid": -914061504, "ts": 1716454226038534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226046241, "dur": 726,
    "args": {
      "External id": 286292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286292,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286292, "pid": 5, "tid": 7, "ts": 1716454226046241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038545, "dur": 6,
    "args": {
      "External id": 286292,
      "cbid": 211, "correlation": 286292
    }
  },
  {
    "ph": "s", "id": 286292, "pid": 76337, "tid": -914061504, "ts": 1716454226038545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226046969, "dur": 31,
    "args": {
      "External id": 286294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286294,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286294, "pid": 5, "tid": 7, "ts": 1716454226046969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038555, "dur": 5,
    "args": {
      "External id": 286294,
      "cbid": 211, "correlation": 286294
    }
  },
  {
    "ph": "s", "id": 286294, "pid": 76337, "tid": -914061504, "ts": 1716454226038555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226047001, "dur": 47,
    "args": {
      "External id": 286300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286300,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286300, "pid": 5, "tid": 7, "ts": 1716454226047001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038582, "dur": 8,
    "args": {
      "External id": 286300,
      "cbid": 211, "correlation": 286300
    }
  },
  {
    "ph": "s", "id": 286300, "pid": 76337, "tid": -914061504, "ts": 1716454226038582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226047050, "dur": 121,
    "args": {
      "External id": 286308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286308,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286308, "pid": 5, "tid": 7, "ts": 1716454226047050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038617, "dur": 9,
    "args": {
      "External id": 286308,
      "cbid": 211, "correlation": 286308
    }
  },
  {
    "ph": "s", "id": 286308, "pid": 76337, "tid": -914061504, "ts": 1716454226038617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226047172, "dur": 28,
    "args": {
      "External id": 286316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286316,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286316, "pid": 5, "tid": 7, "ts": 1716454226047172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038645, "dur": 8,
    "args": {
      "External id": 286316,
      "cbid": 211, "correlation": 286316
    }
  },
  {
    "ph": "s", "id": 286316, "pid": 76337, "tid": -914061504, "ts": 1716454226038645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226047201, "dur": 117,
    "args": {
      "External id": 286326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286326, "pid": 5, "tid": 7, "ts": 1716454226047201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038722, "dur": 12,
    "args": {
      "External id": 286326,
      "cbid": 211, "correlation": 286326
    }
  },
  {
    "ph": "s", "id": 286326, "pid": 76337, "tid": -914061504, "ts": 1716454226038722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226047320, "dur": 60,
    "args": {
      "External id": 286347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286347,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286347, "pid": 5, "tid": 7, "ts": 1716454226047320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038757, "dur": 8,
    "args": {
      "External id": 286347,
      "cbid": 211, "correlation": 286347
    }
  },
  {
    "ph": "s", "id": 286347, "pid": 76337, "tid": -914061504, "ts": 1716454226038757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226047381, "dur": 4,
    "args": {
      "External id": 286359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286359,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286359, "pid": 5, "tid": 7, "ts": 1716454226047381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038775, "dur": 6,
    "args": {
      "External id": 286359,
      "cbid": 211, "correlation": 286359
    }
  },
  {
    "ph": "s", "id": 286359, "pid": 76337, "tid": -914061504, "ts": 1716454226038775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226047386, "dur": 43,
    "args": {
      "External id": 286362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286362, "pid": 5, "tid": 7, "ts": 1716454226047386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038792, "dur": 6,
    "args": {
      "External id": 286362,
      "cbid": 211, "correlation": 286362
    }
  },
  {
    "ph": "s", "id": 286362, "pid": 76337, "tid": -914061504, "ts": 1716454226038792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226047431, "dur": 30,
    "args": {
      "External id": 286371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286371,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286371, "pid": 5, "tid": 7, "ts": 1716454226047431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038831, "dur": 9,
    "args": {
      "External id": 286371,
      "cbid": 211, "correlation": 286371
    }
  },
  {
    "ph": "s", "id": 286371, "pid": 76337, "tid": -914061504, "ts": 1716454226038831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038889, "dur": 0,
    "args": {
      "External id": 286381,
      "cbid": 317, "correlation": 286381
    }
  },
  {
    "ph": "f", "id": 286381, "pid": 76337, "tid": -914061504, "ts": 1716454226038889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038889, "dur": 0,
    "args": {
      "External id": 286382,
      "cbid": 203, "correlation": 286382
    }
  },
  {
    "ph": "f", "id": 286382, "pid": 76337, "tid": -914061504, "ts": 1716454226038889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038890, "dur": 0,
    "args": {
      "External id": 286383,
      "cbid": 205, "correlation": 286383
    }
  },
  {
    "ph": "f", "id": 286383, "pid": 76337, "tid": -914061504, "ts": 1716454226038890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226047462, "dur": 33,
    "args": {
      "External id": 286387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286387,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286387, "pid": 5, "tid": 7, "ts": 1716454226047462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038903, "dur": 12,
    "args": {
      "External id": 286387,
      "cbid": 211, "correlation": 286387
    }
  },
  {
    "ph": "s", "id": 286387, "pid": 76337, "tid": -914061504, "ts": 1716454226038903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226047496, "dur": 29,
    "args": {
      "External id": 286389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286389,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286389, "pid": 5, "tid": 7, "ts": 1716454226047496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038919, "dur": 5,
    "args": {
      "External id": 286389,
      "cbid": 211, "correlation": 286389
    }
  },
  {
    "ph": "s", "id": 286389, "pid": 76337, "tid": -914061504, "ts": 1716454226038919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226047526, "dur": 728,
    "args": {
      "External id": 286391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286391,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286391, "pid": 5, "tid": 7, "ts": 1716454226047526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038930, "dur": 6,
    "args": {
      "External id": 286391,
      "cbid": 211, "correlation": 286391
    }
  },
  {
    "ph": "s", "id": 286391, "pid": 76337, "tid": -914061504, "ts": 1716454226038930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226048256, "dur": 31,
    "args": {
      "External id": 286393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286393,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286393, "pid": 5, "tid": 7, "ts": 1716454226048256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038939, "dur": 5,
    "args": {
      "External id": 286393,
      "cbid": 211, "correlation": 286393
    }
  },
  {
    "ph": "s", "id": 286393, "pid": 76337, "tid": -914061504, "ts": 1716454226038939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226048288, "dur": 47,
    "args": {
      "External id": 286399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286399,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286399, "pid": 5, "tid": 7, "ts": 1716454226048288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226038967, "dur": 18,
    "args": {
      "External id": 286399,
      "cbid": 211, "correlation": 286399
    }
  },
  {
    "ph": "s", "id": 286399, "pid": 76337, "tid": -914061504, "ts": 1716454226038967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226048337, "dur": 58,
    "args": {
      "External id": 286419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286419,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286419, "pid": 5, "tid": 7, "ts": 1716454226048337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039048, "dur": 11,
    "args": {
      "External id": 286419,
      "cbid": 211, "correlation": 286419
    }
  },
  {
    "ph": "s", "id": 286419, "pid": 76337, "tid": -914061504, "ts": 1716454226039048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226048396, "dur": 4,
    "args": {
      "External id": 286431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286431,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286431, "pid": 5, "tid": 7, "ts": 1716454226048396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039069, "dur": 7,
    "args": {
      "External id": 286431,
      "cbid": 211, "correlation": 286431
    }
  },
  {
    "ph": "s", "id": 286431, "pid": 76337, "tid": -914061504, "ts": 1716454226039069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226048401, "dur": 44,
    "args": {
      "External id": 286434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286434,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286434, "pid": 5, "tid": 7, "ts": 1716454226048401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039087, "dur": 6,
    "args": {
      "External id": 286434,
      "cbid": 211, "correlation": 286434
    }
  },
  {
    "ph": "s", "id": 286434, "pid": 76337, "tid": -914061504, "ts": 1716454226039087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226048446, "dur": 29,
    "args": {
      "External id": 286443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286443,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286443, "pid": 5, "tid": 7, "ts": 1716454226048446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039126, "dur": 10,
    "args": {
      "External id": 286443,
      "cbid": 211, "correlation": 286443
    }
  },
  {
    "ph": "s", "id": 286443, "pid": 76337, "tid": -914061504, "ts": 1716454226039126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039191, "dur": 0,
    "args": {
      "External id": 286453,
      "cbid": 317, "correlation": 286453
    }
  },
  {
    "ph": "f", "id": 286453, "pid": 76337, "tid": -914061504, "ts": 1716454226039191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039192, "dur": 0,
    "args": {
      "External id": 286454,
      "cbid": 203, "correlation": 286454
    }
  },
  {
    "ph": "f", "id": 286454, "pid": 76337, "tid": -914061504, "ts": 1716454226039192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039193, "dur": 0,
    "args": {
      "External id": 286455,
      "cbid": 205, "correlation": 286455
    }
  },
  {
    "ph": "f", "id": 286455, "pid": 76337, "tid": -914061504, "ts": 1716454226039193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226048477, "dur": 33,
    "args": {
      "External id": 286459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286459,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286459, "pid": 5, "tid": 7, "ts": 1716454226048477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039207, "dur": 13,
    "args": {
      "External id": 286459,
      "cbid": 211, "correlation": 286459
    }
  },
  {
    "ph": "s", "id": 286459, "pid": 76337, "tid": -914061504, "ts": 1716454226039207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226048511, "dur": 29,
    "args": {
      "External id": 286461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286461,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286461, "pid": 5, "tid": 7, "ts": 1716454226048511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039223, "dur": 5,
    "args": {
      "External id": 286461,
      "cbid": 211, "correlation": 286461
    }
  },
  {
    "ph": "s", "id": 286461, "pid": 76337, "tid": -914061504, "ts": 1716454226039223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226048541, "dur": 727,
    "args": {
      "External id": 286463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286463,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 2.4,
      "warps per SM": 19.2,
      "grid": [2, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286463, "pid": 5, "tid": 7, "ts": 1716454226048541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039233, "dur": 6,
    "args": {
      "External id": 286463,
      "cbid": 211, "correlation": 286463
    }
  },
  {
    "ph": "s", "id": 286463, "pid": 76337, "tid": -914061504, "ts": 1716454226039233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226049270, "dur": 31,
    "args": {
      "External id": 286465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286465,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 76.8,
      "warps per SM": 614.4,
      "grid": [96, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286465, "pid": 5, "tid": 7, "ts": 1716454226049270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039243, "dur": 5,
    "args": {
      "External id": 286465,
      "cbid": 211, "correlation": 286465
    }
  },
  {
    "ph": "s", "id": 286465, "pid": 76337, "tid": -914061504, "ts": 1716454226039243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226049302, "dur": 47,
    "args": {
      "External id": 286471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286471,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286471, "pid": 5, "tid": 7, "ts": 1716454226049302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039270, "dur": 8,
    "args": {
      "External id": 286471,
      "cbid": 211, "correlation": 286471
    }
  },
  {
    "ph": "s", "id": 286471, "pid": 76337, "tid": -914061504, "ts": 1716454226039270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226049351, "dur": 123,
    "args": {
      "External id": 286479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286479, "pid": 5, "tid": 7, "ts": 1716454226049351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039304, "dur": 9,
    "args": {
      "External id": 286479,
      "cbid": 211, "correlation": 286479
    }
  },
  {
    "ph": "s", "id": 286479, "pid": 76337, "tid": -914061504, "ts": 1716454226039304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226049475, "dur": 27,
    "args": {
      "External id": 286487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286487,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286487, "pid": 5, "tid": 7, "ts": 1716454226049475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039334, "dur": 8,
    "args": {
      "External id": 286487,
      "cbid": 211, "correlation": 286487
    }
  },
  {
    "ph": "s", "id": 286487, "pid": 76337, "tid": -914061504, "ts": 1716454226039334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_nhwc_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float, unsigned long)", "pid": 5, "tid": 7,
    "ts": 1716454226049509, "dur": 324,
    "args": {
      "External id": 286496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286496,
      "registers per thread": 26,
      "shared memory": 0,
      "blocks per SM": 307.2,
      "warps per SM": 9830.4,
      "grid": [24576, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286496, "pid": 5, "tid": 7, "ts": 1716454226049509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039419, "dur": 51,
    "args": {
      "External id": 286496,
      "cbid": 211, "correlation": 286496
    }
  },
  {
    "ph": "s", "id": 286496, "pid": 76337, "tid": -914061504, "ts": 1716454226039419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226049834, "dur": 24,
    "args": {
      "External id": 286510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 57.6,
      "warps per SM": 230.4,
      "grid": [4608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286510, "pid": 5, "tid": 7, "ts": 1716454226049834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039536, "dur": 12,
    "args": {
      "External id": 286510,
      "cbid": 211, "correlation": 286510
    }
  },
  {
    "ph": "s", "id": 286510, "pid": 76337, "tid": -914061504, "ts": 1716454226039536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039556, "dur": 0,
    "args": {
      "External id": 286517,
      "cbid": 317, "correlation": 286517
    }
  },
  {
    "ph": "f", "id": 286517, "pid": 76337, "tid": -914061504, "ts": 1716454226039556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039557, "dur": 0,
    "args": {
      "External id": 286518,
      "cbid": 203, "correlation": 286518
    }
  },
  {
    "ph": "f", "id": 286518, "pid": 76337, "tid": -914061504, "ts": 1716454226039557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226039558, "dur": 0,
    "args": {
      "External id": 286519,
      "cbid": 205, "correlation": 286519
    }
  },
  {
    "ph": "f", "id": 286519, "pid": 76337, "tid": -914061504, "ts": 1716454226039558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226049860, "dur": 2553,
    "args": {
      "External id": 286523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286523,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286523, "pid": 5, "tid": 7, "ts": 1716454226049860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226042491, "dur": 40,
    "args": {
      "External id": 286523,
      "cbid": 211, "correlation": 286523
    }
  },
  {
    "ph": "s", "id": 286523, "pid": 76337, "tid": -914061504, "ts": 1716454226042491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226052414, "dur": 166,
    "args": {
      "External id": 286529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286529, "pid": 5, "tid": 7, "ts": 1716454226052414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043311, "dur": 14,
    "args": {
      "External id": 286529,
      "cbid": 211, "correlation": 286529
    }
  },
  {
    "ph": "s", "id": 286529, "pid": 76337, "tid": -914061504, "ts": 1716454226043311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226052581, "dur": 453,
    "args": {
      "External id": 286539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286539,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286539, "pid": 5, "tid": 7, "ts": 1716454226052581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043479, "dur": 15,
    "args": {
      "External id": 286539,
      "cbid": 211, "correlation": 286539
    }
  },
  {
    "ph": "s", "id": 286539, "pid": 76337, "tid": -914061504, "ts": 1716454226043479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226053036, "dur": 203,
    "args": {
      "External id": 286560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286560,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286560, "pid": 5, "tid": 7, "ts": 1716454226053036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043523, "dur": 8,
    "args": {
      "External id": 286560,
      "cbid": 211, "correlation": 286560
    }
  },
  {
    "ph": "s", "id": 286560, "pid": 76337, "tid": -914061504, "ts": 1716454226043523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226053240, "dur": 4,
    "args": {
      "External id": 286572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286572,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286572, "pid": 5, "tid": 7, "ts": 1716454226053240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043545, "dur": 9,
    "args": {
      "External id": 286572,
      "cbid": 211, "correlation": 286572
    }
  },
  {
    "ph": "s", "id": 286572, "pid": 76337, "tid": -914061504, "ts": 1716454226043545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226053245, "dur": 164,
    "args": {
      "External id": 286575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286575, "pid": 5, "tid": 7, "ts": 1716454226053245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043577, "dur": 8,
    "args": {
      "External id": 286575,
      "cbid": 211, "correlation": 286575
    }
  },
  {
    "ph": "s", "id": 286575, "pid": 76337, "tid": -914061504, "ts": 1716454226043577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226053411, "dur": 107,
    "args": {
      "External id": 286584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286584,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286584, "pid": 5, "tid": 7, "ts": 1716454226053411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043631, "dur": 10,
    "args": {
      "External id": 286584,
      "cbid": 211, "correlation": 286584
    }
  },
  {
    "ph": "s", "id": 286584, "pid": 76337, "tid": -914061504, "ts": 1716454226043631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043705, "dur": 0,
    "args": {
      "External id": 286594,
      "cbid": 317, "correlation": 286594
    }
  },
  {
    "ph": "f", "id": 286594, "pid": 76337, "tid": -914061504, "ts": 1716454226043705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043706, "dur": 1,
    "args": {
      "External id": 286595,
      "cbid": 203, "correlation": 286595
    }
  },
  {
    "ph": "f", "id": 286595, "pid": 76337, "tid": -914061504, "ts": 1716454226043706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226043708, "dur": 0,
    "args": {
      "External id": 286596,
      "cbid": 205, "correlation": 286596
    }
  },
  {
    "ph": "f", "id": 286596, "pid": 76337, "tid": -914061504, "ts": 1716454226043708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226053519, "dur": 117,
    "args": {
      "External id": 286600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286600,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286600, "pid": 5, "tid": 7, "ts": 1716454226053519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226046332, "dur": 26,
    "args": {
      "External id": 286600,
      "cbid": 211, "correlation": 286600
    }
  },
  {
    "ph": "s", "id": 286600, "pid": 76337, "tid": -914061504, "ts": 1716454226046332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226053638, "dur": 29,
    "args": {
      "External id": 286602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286602, "pid": 5, "tid": 7, "ts": 1716454226053638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226046362, "dur": 6,
    "args": {
      "External id": 286602,
      "cbid": 211, "correlation": 286602
    }
  },
  {
    "ph": "s", "id": 286602, "pid": 76337, "tid": -914061504, "ts": 1716454226046362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226053668, "dur": 2533,
    "args": {
      "External id": 286604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286604,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286604, "pid": 5, "tid": 7, "ts": 1716454226053668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226046377, "dur": 9,
    "args": {
      "External id": 286604,
      "cbid": 211, "correlation": 286604
    }
  },
  {
    "ph": "s", "id": 286604, "pid": 76337, "tid": -914061504, "ts": 1716454226046377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226056202, "dur": 135,
    "args": {
      "External id": 286606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286606,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286606, "pid": 5, "tid": 7, "ts": 1716454226056202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226046392, "dur": 8,
    "args": {
      "External id": 286606,
      "cbid": 211, "correlation": 286606
    }
  },
  {
    "ph": "s", "id": 286606, "pid": 76337, "tid": -914061504, "ts": 1716454226046392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226056338, "dur": 175,
    "args": {
      "External id": 286612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286612, "pid": 5, "tid": 7, "ts": 1716454226056338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047128, "dur": 14,
    "args": {
      "External id": 286612,
      "cbid": 211, "correlation": 286612
    }
  },
  {
    "ph": "s", "id": 286612, "pid": 76337, "tid": -914061504, "ts": 1716454226047128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226056514, "dur": 208,
    "args": {
      "External id": 286632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286632,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286632, "pid": 5, "tid": 7, "ts": 1716454226056514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047256, "dur": 13,
    "args": {
      "External id": 286632,
      "cbid": 211, "correlation": 286632
    }
  },
  {
    "ph": "s", "id": 286632, "pid": 76337, "tid": -914061504, "ts": 1716454226047256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226056723, "dur": 4,
    "args": {
      "External id": 286644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286644,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286644, "pid": 5, "tid": 7, "ts": 1716454226056723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047280, "dur": 8,
    "args": {
      "External id": 286644,
      "cbid": 211, "correlation": 286644
    }
  },
  {
    "ph": "s", "id": 286644, "pid": 76337, "tid": -914061504, "ts": 1716454226047280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226056728, "dur": 163,
    "args": {
      "External id": 286647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286647, "pid": 5, "tid": 7, "ts": 1716454226056728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047309, "dur": 7,
    "args": {
      "External id": 286647,
      "cbid": 211, "correlation": 286647
    }
  },
  {
    "ph": "s", "id": 286647, "pid": 76337, "tid": -914061504, "ts": 1716454226047309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226056892, "dur": 107,
    "args": {
      "External id": 286656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286656,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286656, "pid": 5, "tid": 7, "ts": 1716454226056892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047359, "dur": 10,
    "args": {
      "External id": 286656,
      "cbid": 211, "correlation": 286656
    }
  },
  {
    "ph": "s", "id": 286656, "pid": 76337, "tid": -914061504, "ts": 1716454226047359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047450, "dur": 0,
    "args": {
      "External id": 286666,
      "cbid": 317, "correlation": 286666
    }
  },
  {
    "ph": "f", "id": 286666, "pid": 76337, "tid": -914061504, "ts": 1716454226047450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047451, "dur": 0,
    "args": {
      "External id": 286667,
      "cbid": 203, "correlation": 286667
    }
  },
  {
    "ph": "f", "id": 286667, "pid": 76337, "tid": -914061504, "ts": 1716454226047451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047452, "dur": 0,
    "args": {
      "External id": 286668,
      "cbid": 205, "correlation": 286668
    }
  },
  {
    "ph": "f", "id": 286668, "pid": 76337, "tid": -914061504, "ts": 1716454226047452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226057000, "dur": 117,
    "args": {
      "External id": 286672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286672, "pid": 5, "tid": 7, "ts": 1716454226057000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047466, "dur": 13,
    "args": {
      "External id": 286672,
      "cbid": 211, "correlation": 286672
    }
  },
  {
    "ph": "s", "id": 286672, "pid": 76337, "tid": -914061504, "ts": 1716454226047466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226057119, "dur": 29,
    "args": {
      "External id": 286674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286674, "pid": 5, "tid": 7, "ts": 1716454226057119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047482, "dur": 6,
    "args": {
      "External id": 286674,
      "cbid": 211, "correlation": 286674
    }
  },
  {
    "ph": "s", "id": 286674, "pid": 76337, "tid": -914061504, "ts": 1716454226047482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226057149, "dur": 2538,
    "args": {
      "External id": 286676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286676,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286676, "pid": 5, "tid": 7, "ts": 1716454226057149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047493, "dur": 6,
    "args": {
      "External id": 286676,
      "cbid": 211, "correlation": 286676
    }
  },
  {
    "ph": "s", "id": 286676, "pid": 76337, "tid": -914061504, "ts": 1716454226047493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226059689, "dur": 144,
    "args": {
      "External id": 286678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286678, "pid": 5, "tid": 7, "ts": 1716454226059689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047503, "dur": 5,
    "args": {
      "External id": 286678,
      "cbid": 211, "correlation": 286678
    }
  },
  {
    "ph": "s", "id": 286678, "pid": 76337, "tid": -914061504, "ts": 1716454226047503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226059834, "dur": 176,
    "args": {
      "External id": 286684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286684, "pid": 5, "tid": 7, "ts": 1716454226059834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047532, "dur": 9,
    "args": {
      "External id": 286684,
      "cbid": 211, "correlation": 286684
    }
  },
  {
    "ph": "s", "id": 286684, "pid": 76337, "tid": -914061504, "ts": 1716454226047532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226060011, "dur": 474,
    "args": {
      "External id": 286692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286692, "pid": 5, "tid": 7, "ts": 1716454226060011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047576, "dur": 9,
    "args": {
      "External id": 286692,
      "cbid": 211, "correlation": 286692
    }
  },
  {
    "ph": "s", "id": 286692, "pid": 76337, "tid": -914061504, "ts": 1716454226047576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226060486, "dur": 103,
    "args": {
      "External id": 286700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286700,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286700, "pid": 5, "tid": 7, "ts": 1716454226060486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047621, "dur": 12,
    "args": {
      "External id": 286700,
      "cbid": 211, "correlation": 286700
    }
  },
  {
    "ph": "s", "id": 286700, "pid": 76337, "tid": -914061504, "ts": 1716454226047621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226060591, "dur": 454,
    "args": {
      "External id": 286710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286710, "pid": 5, "tid": 7, "ts": 1716454226060591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047710, "dur": 14,
    "args": {
      "External id": 286710,
      "cbid": 211, "correlation": 286710
    }
  },
  {
    "ph": "s", "id": 286710, "pid": 76337, "tid": -914061504, "ts": 1716454226047710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226061046, "dur": 204,
    "args": {
      "External id": 286731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286731,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286731, "pid": 5, "tid": 7, "ts": 1716454226061046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047746, "dur": 7,
    "args": {
      "External id": 286731,
      "cbid": 211, "correlation": 286731
    }
  },
  {
    "ph": "s", "id": 286731, "pid": 76337, "tid": -914061504, "ts": 1716454226047746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226061251, "dur": 4,
    "args": {
      "External id": 286743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286743,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286743, "pid": 5, "tid": 7, "ts": 1716454226061251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047765, "dur": 6,
    "args": {
      "External id": 286743,
      "cbid": 211, "correlation": 286743
    }
  },
  {
    "ph": "s", "id": 286743, "pid": 76337, "tid": -914061504, "ts": 1716454226047765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226061256, "dur": 163,
    "args": {
      "External id": 286746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286746,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286746, "pid": 5, "tid": 7, "ts": 1716454226061256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047783, "dur": 7,
    "args": {
      "External id": 286746,
      "cbid": 211, "correlation": 286746
    }
  },
  {
    "ph": "s", "id": 286746, "pid": 76337, "tid": -914061504, "ts": 1716454226047783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226061420, "dur": 106,
    "args": {
      "External id": 286755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286755,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286755, "pid": 5, "tid": 7, "ts": 1716454226061420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047823, "dur": 10,
    "args": {
      "External id": 286755,
      "cbid": 211, "correlation": 286755
    }
  },
  {
    "ph": "s", "id": 286755, "pid": 76337, "tid": -914061504, "ts": 1716454226047823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047879, "dur": 0,
    "args": {
      "External id": 286765,
      "cbid": 317, "correlation": 286765
    }
  },
  {
    "ph": "f", "id": 286765, "pid": 76337, "tid": -914061504, "ts": 1716454226047879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047880, "dur": 0,
    "args": {
      "External id": 286766,
      "cbid": 203, "correlation": 286766
    }
  },
  {
    "ph": "f", "id": 286766, "pid": 76337, "tid": -914061504, "ts": 1716454226047880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047881, "dur": 0,
    "args": {
      "External id": 286767,
      "cbid": 205, "correlation": 286767
    }
  },
  {
    "ph": "f", "id": 286767, "pid": 76337, "tid": -914061504, "ts": 1716454226047881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226061528, "dur": 116,
    "args": {
      "External id": 286771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286771,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286771, "pid": 5, "tid": 7, "ts": 1716454226061528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047900, "dur": 12,
    "args": {
      "External id": 286771,
      "cbid": 211, "correlation": 286771
    }
  },
  {
    "ph": "s", "id": 286771, "pid": 76337, "tid": -914061504, "ts": 1716454226047900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226061645, "dur": 28,
    "args": {
      "External id": 286773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286773, "pid": 5, "tid": 7, "ts": 1716454226061645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047914, "dur": 5,
    "args": {
      "External id": 286773,
      "cbid": 211, "correlation": 286773
    }
  },
  {
    "ph": "s", "id": 286773, "pid": 76337, "tid": -914061504, "ts": 1716454226047914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226061674, "dur": 2528,
    "args": {
      "External id": 286775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286775,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286775, "pid": 5, "tid": 7, "ts": 1716454226061674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047925, "dur": 6,
    "args": {
      "External id": 286775,
      "cbid": 211, "correlation": 286775
    }
  },
  {
    "ph": "s", "id": 286775, "pid": 76337, "tid": -914061504, "ts": 1716454226047925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226064203, "dur": 135,
    "args": {
      "External id": 286777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286777,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286777, "pid": 5, "tid": 7, "ts": 1716454226064203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047935, "dur": 6,
    "args": {
      "External id": 286777,
      "cbid": 211, "correlation": 286777
    }
  },
  {
    "ph": "s", "id": 286777, "pid": 76337, "tid": -914061504, "ts": 1716454226047935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226064340, "dur": 175,
    "args": {
      "External id": 286783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286783, "pid": 5, "tid": 7, "ts": 1716454226064340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226047963, "dur": 8,
    "args": {
      "External id": 286783,
      "cbid": 211, "correlation": 286783
    }
  },
  {
    "ph": "s", "id": 286783, "pid": 76337, "tid": -914061504, "ts": 1716454226047963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226064516, "dur": 209,
    "args": {
      "External id": 286803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286803,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286803, "pid": 5, "tid": 7, "ts": 1716454226064516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048051, "dur": 12,
    "args": {
      "External id": 286803,
      "cbid": 211, "correlation": 286803
    }
  },
  {
    "ph": "s", "id": 286803, "pid": 76337, "tid": -914061504, "ts": 1716454226048051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226064727, "dur": 4,
    "args": {
      "External id": 286815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286815,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286815, "pid": 5, "tid": 7, "ts": 1716454226064727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048072, "dur": 6,
    "args": {
      "External id": 286815,
      "cbid": 211, "correlation": 286815
    }
  },
  {
    "ph": "s", "id": 286815, "pid": 76337, "tid": -914061504, "ts": 1716454226048072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226064731, "dur": 163,
    "args": {
      "External id": 286818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286818, "pid": 5, "tid": 7, "ts": 1716454226064731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048091, "dur": 6,
    "args": {
      "External id": 286818,
      "cbid": 211, "correlation": 286818
    }
  },
  {
    "ph": "s", "id": 286818, "pid": 76337, "tid": -914061504, "ts": 1716454226048091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226064895, "dur": 106,
    "args": {
      "External id": 286827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286827,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286827, "pid": 5, "tid": 7, "ts": 1716454226064895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048131, "dur": 11,
    "args": {
      "External id": 286827,
      "cbid": 211, "correlation": 286827
    }
  },
  {
    "ph": "s", "id": 286827, "pid": 76337, "tid": -914061504, "ts": 1716454226048131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048199, "dur": 0,
    "args": {
      "External id": 286837,
      "cbid": 317, "correlation": 286837
    }
  },
  {
    "ph": "f", "id": 286837, "pid": 76337, "tid": -914061504, "ts": 1716454226048199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048200, "dur": 0,
    "args": {
      "External id": 286838,
      "cbid": 203, "correlation": 286838
    }
  },
  {
    "ph": "f", "id": 286838, "pid": 76337, "tid": -914061504, "ts": 1716454226048200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048201, "dur": 0,
    "args": {
      "External id": 286839,
      "cbid": 205, "correlation": 286839
    }
  },
  {
    "ph": "f", "id": 286839, "pid": 76337, "tid": -914061504, "ts": 1716454226048201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226065003, "dur": 118,
    "args": {
      "External id": 286843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286843,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286843, "pid": 5, "tid": 7, "ts": 1716454226065003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048214, "dur": 12,
    "args": {
      "External id": 286843,
      "cbid": 211, "correlation": 286843
    }
  },
  {
    "ph": "s", "id": 286843, "pid": 76337, "tid": -914061504, "ts": 1716454226048214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226065122, "dur": 28,
    "args": {
      "External id": 286845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286845,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286845, "pid": 5, "tid": 7, "ts": 1716454226065122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048228, "dur": 5,
    "args": {
      "External id": 286845,
      "cbid": 211, "correlation": 286845
    }
  },
  {
    "ph": "s", "id": 286845, "pid": 76337, "tid": -914061504, "ts": 1716454226048228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226065151, "dur": 2537,
    "args": {
      "External id": 286847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286847,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286847, "pid": 5, "tid": 7, "ts": 1716454226065151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048239, "dur": 6,
    "args": {
      "External id": 286847,
      "cbid": 211, "correlation": 286847
    }
  },
  {
    "ph": "s", "id": 286847, "pid": 76337, "tid": -914061504, "ts": 1716454226048239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226067690, "dur": 132,
    "args": {
      "External id": 286849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286849, "pid": 5, "tid": 7, "ts": 1716454226067690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048248, "dur": 5,
    "args": {
      "External id": 286849,
      "cbid": 211, "correlation": 286849
    }
  },
  {
    "ph": "s", "id": 286849, "pid": 76337, "tid": -914061504, "ts": 1716454226048248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226067823, "dur": 175,
    "args": {
      "External id": 286855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286855, "pid": 5, "tid": 7, "ts": 1716454226067823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048276, "dur": 8,
    "args": {
      "External id": 286855,
      "cbid": 211, "correlation": 286855
    }
  },
  {
    "ph": "s", "id": 286855, "pid": 76337, "tid": -914061504, "ts": 1716454226048276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226067999, "dur": 473,
    "args": {
      "External id": 286863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286863, "pid": 5, "tid": 7, "ts": 1716454226067999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048311, "dur": 9,
    "args": {
      "External id": 286863,
      "cbid": 211, "correlation": 286863
    }
  },
  {
    "ph": "s", "id": 286863, "pid": 76337, "tid": -914061504, "ts": 1716454226048311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226068474, "dur": 102,
    "args": {
      "External id": 286871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286871,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286871, "pid": 5, "tid": 7, "ts": 1716454226068474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048341, "dur": 9,
    "args": {
      "External id": 286871,
      "cbid": 211, "correlation": 286871
    }
  },
  {
    "ph": "s", "id": 286871, "pid": 76337, "tid": -914061504, "ts": 1716454226048341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226068577, "dur": 455,
    "args": {
      "External id": 286881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286881, "pid": 5, "tid": 7, "ts": 1716454226068577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048419, "dur": 14,
    "args": {
      "External id": 286881,
      "cbid": 211, "correlation": 286881
    }
  },
  {
    "ph": "s", "id": 286881, "pid": 76337, "tid": -914061504, "ts": 1716454226048419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226069033, "dur": 202,
    "args": {
      "External id": 286902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286902,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286902, "pid": 5, "tid": 7, "ts": 1716454226069033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048458, "dur": 8,
    "args": {
      "External id": 286902,
      "cbid": 211, "correlation": 286902
    }
  },
  {
    "ph": "s", "id": 286902, "pid": 76337, "tid": -914061504, "ts": 1716454226048458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226069236, "dur": 4,
    "args": {
      "External id": 286914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286914,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286914, "pid": 5, "tid": 7, "ts": 1716454226069236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048475, "dur": 6,
    "args": {
      "External id": 286914,
      "cbid": 211, "correlation": 286914
    }
  },
  {
    "ph": "s", "id": 286914, "pid": 76337, "tid": -914061504, "ts": 1716454226048475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226069241, "dur": 164,
    "args": {
      "External id": 286917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286917,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286917, "pid": 5, "tid": 7, "ts": 1716454226069241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048492, "dur": 7,
    "args": {
      "External id": 286917,
      "cbid": 211, "correlation": 286917
    }
  },
  {
    "ph": "s", "id": 286917, "pid": 76337, "tid": -914061504, "ts": 1716454226048492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226069407, "dur": 107,
    "args": {
      "External id": 286926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286926,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286926, "pid": 5, "tid": 7, "ts": 1716454226069407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048532, "dur": 11,
    "args": {
      "External id": 286926,
      "cbid": 211, "correlation": 286926
    }
  },
  {
    "ph": "s", "id": 286926, "pid": 76337, "tid": -914061504, "ts": 1716454226048532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048590, "dur": 0,
    "args": {
      "External id": 286936,
      "cbid": 317, "correlation": 286936
    }
  },
  {
    "ph": "f", "id": 286936, "pid": 76337, "tid": -914061504, "ts": 1716454226048590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048591, "dur": 0,
    "args": {
      "External id": 286937,
      "cbid": 203, "correlation": 286937
    }
  },
  {
    "ph": "f", "id": 286937, "pid": 76337, "tid": -914061504, "ts": 1716454226048591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048592, "dur": 0,
    "args": {
      "External id": 286938,
      "cbid": 205, "correlation": 286938
    }
  },
  {
    "ph": "f", "id": 286938, "pid": 76337, "tid": -914061504, "ts": 1716454226048592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226069515, "dur": 119,
    "args": {
      "External id": 286942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286942,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286942, "pid": 5, "tid": 7, "ts": 1716454226069515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048606, "dur": 12,
    "args": {
      "External id": 286942,
      "cbid": 211, "correlation": 286942
    }
  },
  {
    "ph": "s", "id": 286942, "pid": 76337, "tid": -914061504, "ts": 1716454226048606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226069636, "dur": 29,
    "args": {
      "External id": 286944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286944,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286944, "pid": 5, "tid": 7, "ts": 1716454226069636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048620, "dur": 5,
    "args": {
      "External id": 286944,
      "cbid": 211, "correlation": 286944
    }
  },
  {
    "ph": "s", "id": 286944, "pid": 76337, "tid": -914061504, "ts": 1716454226048620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226069666, "dur": 2532,
    "args": {
      "External id": 286946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286946,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 286946, "pid": 5, "tid": 7, "ts": 1716454226069666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048631, "dur": 6,
    "args": {
      "External id": 286946,
      "cbid": 211, "correlation": 286946
    }
  },
  {
    "ph": "s", "id": 286946, "pid": 76337, "tid": -914061504, "ts": 1716454226048631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226072200, "dur": 132,
    "args": {
      "External id": 286948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286948, "pid": 5, "tid": 7, "ts": 1716454226072200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048640, "dur": 6,
    "args": {
      "External id": 286948,
      "cbid": 211, "correlation": 286948
    }
  },
  {
    "ph": "s", "id": 286948, "pid": 76337, "tid": -914061504, "ts": 1716454226048640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226072333, "dur": 175,
    "args": {
      "External id": 286954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286954, "pid": 5, "tid": 7, "ts": 1716454226072333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048669, "dur": 8,
    "args": {
      "External id": 286954,
      "cbid": 211, "correlation": 286954
    }
  },
  {
    "ph": "s", "id": 286954, "pid": 76337, "tid": -914061504, "ts": 1716454226048669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226072510, "dur": 207,
    "args": {
      "External id": 286974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286974,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 286974, "pid": 5, "tid": 7, "ts": 1716454226072510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048740, "dur": 12,
    "args": {
      "External id": 286974,
      "cbid": 211, "correlation": 286974
    }
  },
  {
    "ph": "s", "id": 286974, "pid": 76337, "tid": -914061504, "ts": 1716454226048740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226072718, "dur": 4,
    "args": {
      "External id": 286986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286986,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 286986, "pid": 5, "tid": 7, "ts": 1716454226072718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048762, "dur": 6,
    "args": {
      "External id": 286986,
      "cbid": 211, "correlation": 286986
    }
  },
  {
    "ph": "s", "id": 286986, "pid": 76337, "tid": -914061504, "ts": 1716454226048762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226072723, "dur": 163,
    "args": {
      "External id": 286989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286989, "pid": 5, "tid": 7, "ts": 1716454226072723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048779, "dur": 6,
    "args": {
      "External id": 286989,
      "cbid": 211, "correlation": 286989
    }
  },
  {
    "ph": "s", "id": 286989, "pid": 76337, "tid": -914061504, "ts": 1716454226048779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226072888, "dur": 106,
    "args": {
      "External id": 286998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 286998,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 286998, "pid": 5, "tid": 7, "ts": 1716454226072888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048819, "dur": 10,
    "args": {
      "External id": 286998,
      "cbid": 211, "correlation": 286998
    }
  },
  {
    "ph": "s", "id": 286998, "pid": 76337, "tid": -914061504, "ts": 1716454226048819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048884, "dur": 0,
    "args": {
      "External id": 287008,
      "cbid": 317, "correlation": 287008
    }
  },
  {
    "ph": "f", "id": 287008, "pid": 76337, "tid": -914061504, "ts": 1716454226048884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048885, "dur": 0,
    "args": {
      "External id": 287009,
      "cbid": 203, "correlation": 287009
    }
  },
  {
    "ph": "f", "id": 287009, "pid": 76337, "tid": -914061504, "ts": 1716454226048885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048885, "dur": 0,
    "args": {
      "External id": 287010,
      "cbid": 205, "correlation": 287010
    }
  },
  {
    "ph": "f", "id": 287010, "pid": 76337, "tid": -914061504, "ts": 1716454226048885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226072995, "dur": 117,
    "args": {
      "External id": 287014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287014, "pid": 5, "tid": 7, "ts": 1716454226072995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048898, "dur": 13,
    "args": {
      "External id": 287014,
      "cbid": 211, "correlation": 287014
    }
  },
  {
    "ph": "s", "id": 287014, "pid": 76337, "tid": -914061504, "ts": 1716454226048898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226073114, "dur": 28,
    "args": {
      "External id": 287016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287016,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 102.4,
      "warps per SM": 819.2,
      "grid": [1, 16, 512],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287016, "pid": 5, "tid": 7, "ts": 1716454226073114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048914, "dur": 5,
    "args": {
      "External id": 287016,
      "cbid": 211, "correlation": 287016
    }
  },
  {
    "ph": "s", "id": 287016, "pid": 76337, "tid": -914061504, "ts": 1716454226048914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226073143, "dur": 2537,
    "args": {
      "External id": 287018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287018,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [4, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287018, "pid": 5, "tid": 7, "ts": 1716454226073143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048925, "dur": 6,
    "args": {
      "External id": 287018,
      "cbid": 211, "correlation": 287018
    }
  },
  {
    "ph": "s", "id": 287018, "pid": 76337, "tid": -914061504, "ts": 1716454226048925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226075682, "dur": 131,
    "args": {
      "External id": 287020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [384, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287020, "pid": 5, "tid": 7, "ts": 1716454226075682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048934, "dur": 5,
    "args": {
      "External id": 287020,
      "cbid": 211, "correlation": 287020
    }
  },
  {
    "ph": "s", "id": 287020, "pid": 76337, "tid": -914061504, "ts": 1716454226048934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226075814, "dur": 174,
    "args": {
      "External id": 287026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287026, "pid": 5, "tid": 7, "ts": 1716454226075814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226048961, "dur": 8,
    "args": {
      "External id": 287026,
      "cbid": 211, "correlation": 287026
    }
  },
  {
    "ph": "s", "id": 287026, "pid": 76337, "tid": -914061504, "ts": 1716454226048961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226075990, "dur": 473,
    "args": {
      "External id": 287034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287034,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287034, "pid": 5, "tid": 7, "ts": 1716454226075990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226049005, "dur": 10,
    "args": {
      "External id": 287034,
      "cbid": 211, "correlation": 287034
    }
  },
  {
    "ph": "s", "id": 287034, "pid": 76337, "tid": -914061504, "ts": 1716454226049005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226076464, "dur": 103,
    "args": {
      "External id": 287042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287042,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287042, "pid": 5, "tid": 7, "ts": 1716454226076464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226049036, "dur": 8,
    "args": {
      "External id": 287042,
      "cbid": 211, "correlation": 287042
    }
  },
  {
    "ph": "s", "id": 287042, "pid": 76337, "tid": -914061504, "ts": 1716454226049036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226049126, "dur": 1,
    "args": {
      "External id": 287050,
      "cbid": 317, "correlation": 287050
    }
  },
  {
    "ph": "f", "id": 287050, "pid": 76337, "tid": -914061504, "ts": 1716454226049126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226049128, "dur": 114586,
    "args": {
      "External id": 287051,
      "cbid": 20, "correlation": 287051
    }
  },
  {
    "ph": "f", "id": 287051, "pid": 76337, "tid": -914061504, "ts": 1716454226049128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_nhwc_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float, unsigned long)", "pid": 5, "tid": 7,
    "ts": 1716454226163767, "dur": 1285,
    "args": {
      "External id": 287055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287055,
      "registers per thread": 26,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 39321.6,
      "grid": [98304, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287055, "pid": 5, "tid": 7, "ts": 1716454226163767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226163739, "dur": 27,
    "args": {
      "External id": 287055,
      "cbid": 211, "correlation": 287055
    }
  },
  {
    "ph": "s", "id": 287055, "pid": 76337, "tid": -914061504, "ts": 1716454226163739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226163851, "dur": 1,
    "args": {
      "External id": 287063,
      "cbid": 317, "correlation": 287063
    }
  },
  {
    "ph": "f", "id": 287063, "pid": 76337, "tid": -914061504, "ts": 1716454226163851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226163853, "dur": 2693,
    "args": {
      "External id": 287064,
      "cbid": 20, "correlation": 287064
    }
  },
  {
    "ph": "f", "id": 287064, "pid": 76337, "tid": -914061504, "ts": 1716454226163853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226166614, "dur": 24,
    "args": {
      "External id": 287073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 57.6,
      "warps per SM": 230.4,
      "grid": [4608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287073, "pid": 5, "tid": 7, "ts": 1716454226166614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226166596, "dur": 19,
    "args": {
      "External id": 287073,
      "cbid": 211, "correlation": 287073
    }
  },
  {
    "ph": "s", "id": 287073, "pid": 76337, "tid": -914061504, "ts": 1716454226166596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226166628, "dur": 0,
    "args": {
      "External id": 287080,
      "cbid": 317, "correlation": 287080
    }
  },
  {
    "ph": "f", "id": 287080, "pid": 76337, "tid": -914061504, "ts": 1716454226166628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226166629, "dur": 1,
    "args": {
      "External id": 287081,
      "cbid": 203, "correlation": 287081
    }
  },
  {
    "ph": "f", "id": 287081, "pid": 76337, "tid": -914061504, "ts": 1716454226166629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226166631, "dur": 0,
    "args": {
      "External id": 287082,
      "cbid": 205, "correlation": 287082
    }
  },
  {
    "ph": "f", "id": 287082, "pid": 76337, "tid": -914061504, "ts": 1716454226166631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize256x128x32_stage1_warpsize4x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226169243, "dur": 10145,
    "args": {
      "External id": 287086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287086,
      "registers per thread": 224,
      "shared memory": 34816,
      "blocks per SM": 38.4,
      "warps per SM": 307.2,
      "grid": [4, 768, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287086, "pid": 5, "tid": 7, "ts": 1716454226169243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226168999, "dur": 247,
    "args": {
      "External id": 287086,
      "cbid": 211, "correlation": 287086
    }
  },
  {
    "ph": "s", "id": 287086, "pid": 76337, "tid": -914061504, "ts": 1716454226168999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226179390, "dur": 642,
    "args": {
      "External id": 287092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287092, "pid": 5, "tid": 7, "ts": 1716454226179390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226169899, "dur": 15,
    "args": {
      "External id": 287092,
      "cbid": 211, "correlation": 287092
    }
  },
  {
    "ph": "s", "id": 287092, "pid": 76337, "tid": -914061504, "ts": 1716454226169899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226180033, "dur": 2840,
    "args": {
      "External id": 287102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287102, "pid": 5, "tid": 7, "ts": 1716454226180033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226170099, "dur": 16,
    "args": {
      "External id": 287102,
      "cbid": 211, "correlation": 287102
    }
  },
  {
    "ph": "s", "id": 287102, "pid": 76337, "tid": -914061504, "ts": 1716454226170099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226170136, "dur": 1,
    "args": {
      "External id": 287112,
      "cbid": 317, "correlation": 287112
    }
  },
  {
    "ph": "f", "id": 287112, "pid": 76337, "tid": -914061504, "ts": 1716454226170136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226170137, "dur": 862,
    "args": {
      "External id": 287113,
      "cbid": 20, "correlation": 287113
    }
  },
  {
    "ph": "f", "id": 287113, "pid": 76337, "tid": -914061504, "ts": 1716454226170137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226182874, "dur": 808,
    "args": {
      "External id": 287127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287127,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287127, "pid": 5, "tid": 7, "ts": 1716454226182874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171024, "dur": 18,
    "args": {
      "External id": 287127,
      "cbid": 211, "correlation": 287127
    }
  },
  {
    "ph": "s", "id": 287127, "pid": 76337, "tid": -914061504, "ts": 1716454226171024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226183683, "dur": 5,
    "args": {
      "External id": 287139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287139,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.1,
      "warps per SM": 0.8,
      "grid": [8, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287139, "pid": 5, "tid": 7, "ts": 1716454226183683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171055, "dur": 9,
    "args": {
      "External id": 287139,
      "cbid": 211, "correlation": 287139
    }
  },
  {
    "ph": "s", "id": 287139, "pid": 76337, "tid": -914061504, "ts": 1716454226171055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226183689, "dur": 641,
    "args": {
      "External id": 287142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287142, "pid": 5, "tid": 7, "ts": 1716454226183689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171084, "dur": 9,
    "args": {
      "External id": 287142,
      "cbid": 211, "correlation": 287142
    }
  },
  {
    "ph": "s", "id": 287142, "pid": 76337, "tid": -914061504, "ts": 1716454226171084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226184332, "dur": 414,
    "args": {
      "External id": 287151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287151,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287151, "pid": 5, "tid": 7, "ts": 1716454226184332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171152, "dur": 14,
    "args": {
      "External id": 287151,
      "cbid": 211, "correlation": 287151
    }
  },
  {
    "ph": "s", "id": 287151, "pid": 76337, "tid": -914061504, "ts": 1716454226171152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171236, "dur": 0,
    "args": {
      "External id": 287161,
      "cbid": 317, "correlation": 287161
    }
  },
  {
    "ph": "f", "id": 287161, "pid": 76337, "tid": -914061504, "ts": 1716454226171236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171237, "dur": 1,
    "args": {
      "External id": 287162,
      "cbid": 203, "correlation": 287162
    }
  },
  {
    "ph": "f", "id": 287162, "pid": 76337, "tid": -914061504, "ts": 1716454226171237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226171239, "dur": 0,
    "args": {
      "External id": 287163,
      "cbid": 205, "correlation": 287163
    }
  },
  {
    "ph": "f", "id": 287163, "pid": 76337, "tid": -914061504, "ts": 1716454226171239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226173434, "dur": 3,
    "args": {
      "External id": 287169,
      "cbid": 317, "correlation": 287169
    }
  },
  {
    "ph": "f", "id": 287169, "pid": 76337, "tid": -914061504, "ts": 1716454226173434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226173438, "dur": 8481,
    "args": {
      "External id": 287170,
      "cbid": 20, "correlation": 287170
    }
  },
  {
    "ph": "f", "id": 287170, "pid": 76337, "tid": -914061504, "ts": 1716454226173438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226184747, "dur": 1080,
    "args": {
      "External id": 287171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287171,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [1536, 16, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287171, "pid": 5, "tid": 7, "ts": 1716454226184747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226181947, "dur": 36,
    "args": {
      "External id": 287171,
      "cbid": 211, "correlation": 287171
    }
  },
  {
    "ph": "s", "id": 287171, "pid": 76337, "tid": -914061504, "ts": 1716454226181947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226185829, "dur": 17,
    "args": {
      "External id": 287173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 51.2,
      "warps per SM": 409.6,
      "grid": [1, 16, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287173, "pid": 5, "tid": 7, "ts": 1716454226185829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226181986, "dur": 6,
    "args": {
      "External id": 287173,
      "cbid": 211, "correlation": 287173
    }
  },
  {
    "ph": "s", "id": 287173, "pid": 76337, "tid": -914061504, "ts": 1716454226181986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x256x32_stage1_warpsize2x4x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226185847, "dur": 4885,
    "args": {
      "External id": 287175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287175,
      "registers per thread": 254,
      "shared memory": 33792,
      "blocks per SM": 19.2,
      "warps per SM": 153.6,
      "grid": [1, 1536, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287175, "pid": 5, "tid": 7, "ts": 1716454226185847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182006, "dur": 11,
    "args": {
      "External id": 287175,
      "cbid": 211, "correlation": 287175
    }
  },
  {
    "ph": "s", "id": 287175, "pid": 76337, "tid": -914061504, "ts": 1716454226182006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226190734, "dur": 207,
    "args": {
      "External id": 287177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287177,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 614.4,
      "warps per SM": 4915.2,
      "grid": [1536, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287177, "pid": 5, "tid": 7, "ts": 1716454226190734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182024, "dur": 7,
    "args": {
      "External id": 287177,
      "cbid": 211, "correlation": 287177
    }
  },
  {
    "ph": "s", "id": 287177, "pid": 76337, "tid": -914061504, "ts": 1716454226182024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226190942, "dur": 340,
    "args": {
      "External id": 287183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287183,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287183, "pid": 5, "tid": 7, "ts": 1716454226190942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182653, "dur": 14,
    "args": {
      "External id": 287183,
      "cbid": 211, "correlation": 287183
    }
  },
  {
    "ph": "s", "id": 287183, "pid": 76337, "tid": -914061504, "ts": 1716454226182653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226191284, "dur": 401,
    "args": {
      "External id": 287203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287203,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287203, "pid": 5, "tid": 7, "ts": 1716454226191284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182791, "dur": 14,
    "args": {
      "External id": 287203,
      "cbid": 211, "correlation": 287203
    }
  },
  {
    "ph": "s", "id": 287203, "pid": 76337, "tid": -914061504, "ts": 1716454226182791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226191685, "dur": 4,
    "args": {
      "External id": 287215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287215,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287215, "pid": 5, "tid": 7, "ts": 1716454226191685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182820, "dur": 8,
    "args": {
      "External id": 287215,
      "cbid": 211, "correlation": 287215
    }
  },
  {
    "ph": "s", "id": 287215, "pid": 76337, "tid": -914061504, "ts": 1716454226182820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226191691, "dur": 323,
    "args": {
      "External id": 287218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287218,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287218, "pid": 5, "tid": 7, "ts": 1716454226191691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182849, "dur": 9,
    "args": {
      "External id": 287218,
      "cbid": 211, "correlation": 287218
    }
  },
  {
    "ph": "s", "id": 287218, "pid": 76337, "tid": -914061504, "ts": 1716454226182849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226192015, "dur": 210,
    "args": {
      "External id": 287227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287227,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287227, "pid": 5, "tid": 7, "ts": 1716454226192015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226182905, "dur": 11,
    "args": {
      "External id": 287227,
      "cbid": 211, "correlation": 287227
    }
  },
  {
    "ph": "s", "id": 287227, "pid": 76337, "tid": -914061504, "ts": 1716454226182905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226183010, "dur": 0,
    "args": {
      "External id": 287237,
      "cbid": 317, "correlation": 287237
    }
  },
  {
    "ph": "f", "id": 287237, "pid": 76337, "tid": -914061504, "ts": 1716454226183010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226183011, "dur": 1,
    "args": {
      "External id": 287238,
      "cbid": 203, "correlation": 287238
    }
  },
  {
    "ph": "f", "id": 287238, "pid": 76337, "tid": -914061504, "ts": 1716454226183011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226183012, "dur": 0,
    "args": {
      "External id": 287239,
      "cbid": 205, "correlation": 287239
    }
  },
  {
    "ph": "f", "id": 287239, "pid": 76337, "tid": -914061504, "ts": 1716454226183012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226192227, "dur": 369,
    "args": {
      "External id": 287243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287243,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 614.4,
      "warps per SM": 4915.2,
      "grid": [1536, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287243, "pid": 5, "tid": 7, "ts": 1716454226192227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226185332, "dur": 23,
    "args": {
      "External id": 287243,
      "cbid": 211, "correlation": 287243
    }
  },
  {
    "ph": "s", "id": 287243, "pid": 76337, "tid": -914061504, "ts": 1716454226185332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226192597, "dur": 10,
    "args": {
      "External id": 287245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287245,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287245, "pid": 5, "tid": 7, "ts": 1716454226192597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226185358, "dur": 6,
    "args": {
      "External id": 287245,
      "cbid": 211, "correlation": 287245
    }
  },
  {
    "ph": "s", "id": 287245, "pid": 76337, "tid": -914061504, "ts": 1716454226185358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226192609, "dur": 4,
    "args": {
      "External id": 287247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 287247, "pid": 5, "tid": 7, "ts": 1716454226192609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226185371, "dur": 12,
    "args": {
      "External id": 287247,
      "cbid": 211, "correlation": 287247
    }
  },
  {
    "ph": "s", "id": 287247, "pid": 76337, "tid": -914061504, "ts": 1716454226185371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226185386, "dur": 0,
    "args": {
      "External id": 287248,
      "cbid": 51, "correlation": 287248
    }
  },
  {
    "ph": "s", "id": 287248, "pid": 76337, "tid": -914061504, "ts": 1716454226185386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226192614, "dur": 2669,
    "args": {
      "External id": 287249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287249,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [1536, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287249, "pid": 5, "tid": 7, "ts": 1716454226192614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226185388, "dur": 10,
    "args": {
      "External id": 287249,
      "cbid": 211, "correlation": 287249
    }
  },
  {
    "ph": "s", "id": 287249, "pid": 76337, "tid": -914061504, "ts": 1716454226185388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226195285, "dur": 338,
    "args": {
      "External id": 287254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287254, "pid": 5, "tid": 7, "ts": 1716454226195285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226186041, "dur": 13,
    "args": {
      "External id": 287254,
      "cbid": 211, "correlation": 287254
    }
  },
  {
    "ph": "s", "id": 287254, "pid": 76337, "tid": -914061504, "ts": 1716454226186041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226186139, "dur": 0,
    "args": {
      "External id": 287264,
      "cbid": 317, "correlation": 287264
    }
  },
  {
    "ph": "f", "id": 287264, "pid": 76337, "tid": -914061504, "ts": 1716454226186139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226186140, "dur": 0,
    "args": {
      "External id": 287265,
      "cbid": 203, "correlation": 287265
    }
  },
  {
    "ph": "f", "id": 287265, "pid": 76337, "tid": -914061504, "ts": 1716454226186140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226186141, "dur": 0,
    "args": {
      "External id": 287266,
      "cbid": 205, "correlation": 287266
    }
  },
  {
    "ph": "f", "id": 287266, "pid": 76337, "tid": -914061504, "ts": 1716454226186141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226195624, "dur": 776,
    "args": {
      "External id": 287270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287270,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [2, 1536, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287270, "pid": 5, "tid": 7, "ts": 1716454226195624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226187750, "dur": 22,
    "args": {
      "External id": 287270,
      "cbid": 211, "correlation": 287270
    }
  },
  {
    "ph": "s", "id": 287270, "pid": 76337, "tid": -914061504, "ts": 1716454226187750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226196401, "dur": 321,
    "args": {
      "External id": 287276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287276, "pid": 5, "tid": 7, "ts": 1716454226196401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188202, "dur": 12,
    "args": {
      "External id": 287276,
      "cbid": 211, "correlation": 287276
    }
  },
  {
    "ph": "s", "id": 287276, "pid": 76337, "tid": -914061504, "ts": 1716454226188202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226196724, "dur": 889,
    "args": {
      "External id": 287284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287284, "pid": 5, "tid": 7, "ts": 1716454226196724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188260, "dur": 9,
    "args": {
      "External id": 287284,
      "cbid": 211, "correlation": 287284
    }
  },
  {
    "ph": "s", "id": 287284, "pid": 76337, "tid": -914061504, "ts": 1716454226188260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226197614, "dur": 203,
    "args": {
      "External id": 287292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287292,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287292, "pid": 5, "tid": 7, "ts": 1716454226197614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188316, "dur": 11,
    "args": {
      "External id": 287292,
      "cbid": 211, "correlation": 287292
    }
  },
  {
    "ph": "s", "id": 287292, "pid": 76337, "tid": -914061504, "ts": 1716454226188316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226197819, "dur": 1258,
    "args": {
      "External id": 287302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287302,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287302, "pid": 5, "tid": 7, "ts": 1716454226197819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188443, "dur": 16,
    "args": {
      "External id": 287302,
      "cbid": 211, "correlation": 287302
    }
  },
  {
    "ph": "s", "id": 287302, "pid": 76337, "tid": -914061504, "ts": 1716454226188443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226199078, "dur": 419,
    "args": {
      "External id": 287323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287323,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287323, "pid": 5, "tid": 7, "ts": 1716454226199078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188490, "dur": 8,
    "args": {
      "External id": 287323,
      "cbid": 211, "correlation": 287323
    }
  },
  {
    "ph": "s", "id": 287323, "pid": 76337, "tid": -914061504, "ts": 1716454226188490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226199499, "dur": 4,
    "args": {
      "External id": 287335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287335,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287335, "pid": 5, "tid": 7, "ts": 1716454226199499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188514, "dur": 8,
    "args": {
      "External id": 287335,
      "cbid": 211, "correlation": 287335
    }
  },
  {
    "ph": "s", "id": 287335, "pid": 76337, "tid": -914061504, "ts": 1716454226188514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226199504, "dur": 322,
    "args": {
      "External id": 287338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287338, "pid": 5, "tid": 7, "ts": 1716454226199504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188538, "dur": 7,
    "args": {
      "External id": 287338,
      "cbid": 211, "correlation": 287338
    }
  },
  {
    "ph": "s", "id": 287338, "pid": 76337, "tid": -914061504, "ts": 1716454226188538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226199827, "dur": 209,
    "args": {
      "External id": 287347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287347,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287347, "pid": 5, "tid": 7, "ts": 1716454226199827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188592, "dur": 11,
    "args": {
      "External id": 287347,
      "cbid": 211, "correlation": 287347
    }
  },
  {
    "ph": "s", "id": 287347, "pid": 76337, "tid": -914061504, "ts": 1716454226188592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188658, "dur": 0,
    "args": {
      "External id": 287357,
      "cbid": 317, "correlation": 287357
    }
  },
  {
    "ph": "f", "id": 287357, "pid": 76337, "tid": -914061504, "ts": 1716454226188658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188659, "dur": 0,
    "args": {
      "External id": 287358,
      "cbid": 203, "correlation": 287358
    }
  },
  {
    "ph": "f", "id": 287358, "pid": 76337, "tid": -914061504, "ts": 1716454226188659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188660, "dur": 0,
    "args": {
      "External id": 287359,
      "cbid": 205, "correlation": 287359
    }
  },
  {
    "ph": "f", "id": 287359, "pid": 76337, "tid": -914061504, "ts": 1716454226188660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226200038, "dur": 367,
    "args": {
      "External id": 287363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287363,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 614.4,
      "warps per SM": 4915.2,
      "grid": [1536, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287363, "pid": 5, "tid": 7, "ts": 1716454226200038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188677, "dur": 13,
    "args": {
      "External id": 287363,
      "cbid": 211, "correlation": 287363
    }
  },
  {
    "ph": "s", "id": 287363, "pid": 76337, "tid": -914061504, "ts": 1716454226188677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226200406, "dur": 10,
    "args": {
      "External id": 287365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287365,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287365, "pid": 5, "tid": 7, "ts": 1716454226200406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188693, "dur": 5,
    "args": {
      "External id": 287365,
      "cbid": 211, "correlation": 287365
    }
  },
  {
    "ph": "s", "id": 287365, "pid": 76337, "tid": -914061504, "ts": 1716454226188693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226200418, "dur": 4,
    "args": {
      "External id": 287367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 287367, "pid": 5, "tid": 7, "ts": 1716454226200418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188703, "dur": 8,
    "args": {
      "External id": 287367,
      "cbid": 211, "correlation": 287367
    }
  },
  {
    "ph": "s", "id": 287367, "pid": 76337, "tid": -914061504, "ts": 1716454226188703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188714, "dur": 0,
    "args": {
      "External id": 287368,
      "cbid": 51, "correlation": 287368
    }
  },
  {
    "ph": "s", "id": 287368, "pid": 76337, "tid": -914061504, "ts": 1716454226188714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226200423, "dur": 2663,
    "args": {
      "External id": 287369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287369,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [1536, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287369, "pid": 5, "tid": 7, "ts": 1716454226200423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188715, "dur": 6,
    "args": {
      "External id": 287369,
      "cbid": 211, "correlation": 287369
    }
  },
  {
    "ph": "s", "id": 287369, "pid": 76337, "tid": -914061504, "ts": 1716454226188715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226203087, "dur": 339,
    "args": {
      "External id": 287374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287374, "pid": 5, "tid": 7, "ts": 1716454226203087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188745, "dur": 9,
    "args": {
      "External id": 287374,
      "cbid": 211, "correlation": 287374
    }
  },
  {
    "ph": "s", "id": 287374, "pid": 76337, "tid": -914061504, "ts": 1716454226188745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226203427, "dur": 398,
    "args": {
      "External id": 287394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287394,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287394, "pid": 5, "tid": 7, "ts": 1716454226203427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188819, "dur": 12,
    "args": {
      "External id": 287394,
      "cbid": 211, "correlation": 287394
    }
  },
  {
    "ph": "s", "id": 287394, "pid": 76337, "tid": -914061504, "ts": 1716454226188819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226203826, "dur": 4,
    "args": {
      "External id": 287406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287406,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287406, "pid": 5, "tid": 7, "ts": 1716454226203826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188841, "dur": 6,
    "args": {
      "External id": 287406,
      "cbid": 211, "correlation": 287406
    }
  },
  {
    "ph": "s", "id": 287406, "pid": 76337, "tid": -914061504, "ts": 1716454226188841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226203831, "dur": 324,
    "args": {
      "External id": 287409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287409, "pid": 5, "tid": 7, "ts": 1716454226203831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188859, "dur": 6,
    "args": {
      "External id": 287409,
      "cbid": 211, "correlation": 287409
    }
  },
  {
    "ph": "s", "id": 287409, "pid": 76337, "tid": -914061504, "ts": 1716454226188859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226204156, "dur": 209,
    "args": {
      "External id": 287418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287418,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287418, "pid": 5, "tid": 7, "ts": 1716454226204156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188900, "dur": 10,
    "args": {
      "External id": 287418,
      "cbid": 211, "correlation": 287418
    }
  },
  {
    "ph": "s", "id": 287418, "pid": 76337, "tid": -914061504, "ts": 1716454226188900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188981, "dur": 0,
    "args": {
      "External id": 287428,
      "cbid": 317, "correlation": 287428
    }
  },
  {
    "ph": "f", "id": 287428, "pid": 76337, "tid": -914061504, "ts": 1716454226188981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188982, "dur": 0,
    "args": {
      "External id": 287429,
      "cbid": 203, "correlation": 287429
    }
  },
  {
    "ph": "f", "id": 287429, "pid": 76337, "tid": -914061504, "ts": 1716454226188982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188983, "dur": 0,
    "args": {
      "External id": 287430,
      "cbid": 205, "correlation": 287430
    }
  },
  {
    "ph": "f", "id": 287430, "pid": 76337, "tid": -914061504, "ts": 1716454226188983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226204366, "dur": 372,
    "args": {
      "External id": 287434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287434,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 614.4,
      "warps per SM": 4915.2,
      "grid": [1536, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287434, "pid": 5, "tid": 7, "ts": 1716454226204366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226188997, "dur": 13,
    "args": {
      "External id": 287434,
      "cbid": 211, "correlation": 287434
    }
  },
  {
    "ph": "s", "id": 287434, "pid": 76337, "tid": -914061504, "ts": 1716454226188997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226204739, "dur": 11,
    "args": {
      "External id": 287436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287436,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287436, "pid": 5, "tid": 7, "ts": 1716454226204739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189012, "dur": 5,
    "args": {
      "External id": 287436,
      "cbid": 211, "correlation": 287436
    }
  },
  {
    "ph": "s", "id": 287436, "pid": 76337, "tid": -914061504, "ts": 1716454226189012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226204751, "dur": 3,
    "args": {
      "External id": 287438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287438,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 287438, "pid": 5, "tid": 7, "ts": 1716454226204751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189021, "dur": 6,
    "args": {
      "External id": 287438,
      "cbid": 211, "correlation": 287438
    }
  },
  {
    "ph": "s", "id": 287438, "pid": 76337, "tid": -914061504, "ts": 1716454226189021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189029, "dur": 0,
    "args": {
      "External id": 287439,
      "cbid": 51, "correlation": 287439
    }
  },
  {
    "ph": "s", "id": 287439, "pid": 76337, "tid": -914061504, "ts": 1716454226189029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226204756, "dur": 2676,
    "args": {
      "External id": 287440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287440,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [1536, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287440, "pid": 5, "tid": 7, "ts": 1716454226204756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189030, "dur": 6,
    "args": {
      "External id": 287440,
      "cbid": 211, "correlation": 287440
    }
  },
  {
    "ph": "s", "id": 287440, "pid": 76337, "tid": -914061504, "ts": 1716454226189030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226207433, "dur": 340,
    "args": {
      "External id": 287445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287445, "pid": 5, "tid": 7, "ts": 1716454226207433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189061, "dur": 9,
    "args": {
      "External id": 287445,
      "cbid": 211, "correlation": 287445
    }
  },
  {
    "ph": "s", "id": 287445, "pid": 76337, "tid": -914061504, "ts": 1716454226189061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226207774, "dur": 886,
    "args": {
      "External id": 287453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287453, "pid": 5, "tid": 7, "ts": 1716454226207774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189097, "dur": 9,
    "args": {
      "External id": 287453,
      "cbid": 211, "correlation": 287453
    }
  },
  {
    "ph": "s", "id": 287453, "pid": 76337, "tid": -914061504, "ts": 1716454226189097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226208661, "dur": 202,
    "args": {
      "External id": 287461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287461,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287461, "pid": 5, "tid": 7, "ts": 1716454226208661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189127, "dur": 10,
    "args": {
      "External id": 287461,
      "cbid": 211, "correlation": 287461
    }
  },
  {
    "ph": "s", "id": 287461, "pid": 76337, "tid": -914061504, "ts": 1716454226189127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226208865, "dur": 1269,
    "args": {
      "External id": 287471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287471,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287471, "pid": 5, "tid": 7, "ts": 1716454226208865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189204, "dur": 13,
    "args": {
      "External id": 287471,
      "cbid": 211, "correlation": 287471
    }
  },
  {
    "ph": "s", "id": 287471, "pid": 76337, "tid": -914061504, "ts": 1716454226189204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226210135, "dur": 426,
    "args": {
      "External id": 287492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287492,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287492, "pid": 5, "tid": 7, "ts": 1716454226210135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189238, "dur": 7,
    "args": {
      "External id": 287492,
      "cbid": 211, "correlation": 287492
    }
  },
  {
    "ph": "s", "id": 287492, "pid": 76337, "tid": -914061504, "ts": 1716454226189238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226210563, "dur": 4,
    "args": {
      "External id": 287504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287504,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287504, "pid": 5, "tid": 7, "ts": 1716454226210563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189255, "dur": 6,
    "args": {
      "External id": 287504,
      "cbid": 211, "correlation": 287504
    }
  },
  {
    "ph": "s", "id": 287504, "pid": 76337, "tid": -914061504, "ts": 1716454226189255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226210567, "dur": 325,
    "args": {
      "External id": 287507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287507, "pid": 5, "tid": 7, "ts": 1716454226210567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189271, "dur": 6,
    "args": {
      "External id": 287507,
      "cbid": 211, "correlation": 287507
    }
  },
  {
    "ph": "s", "id": 287507, "pid": 76337, "tid": -914061504, "ts": 1716454226189271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226210894, "dur": 210,
    "args": {
      "External id": 287516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287516,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287516, "pid": 5, "tid": 7, "ts": 1716454226210894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189311, "dur": 9,
    "args": {
      "External id": 287516,
      "cbid": 211, "correlation": 287516
    }
  },
  {
    "ph": "s", "id": 287516, "pid": 76337, "tid": -914061504, "ts": 1716454226189311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189367, "dur": 0,
    "args": {
      "External id": 287526,
      "cbid": 317, "correlation": 287526
    }
  },
  {
    "ph": "f", "id": 287526, "pid": 76337, "tid": -914061504, "ts": 1716454226189367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189367, "dur": 0,
    "args": {
      "External id": 287527,
      "cbid": 203, "correlation": 287527
    }
  },
  {
    "ph": "f", "id": 287527, "pid": 76337, "tid": -914061504, "ts": 1716454226189367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189368, "dur": 0,
    "args": {
      "External id": 287528,
      "cbid": 205, "correlation": 287528
    }
  },
  {
    "ph": "f", "id": 287528, "pid": 76337, "tid": -914061504, "ts": 1716454226189368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226211105, "dur": 381,
    "args": {
      "External id": 287532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287532,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 614.4,
      "warps per SM": 4915.2,
      "grid": [1536, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287532, "pid": 5, "tid": 7, "ts": 1716454226211105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189381, "dur": 13,
    "args": {
      "External id": 287532,
      "cbid": 211, "correlation": 287532
    }
  },
  {
    "ph": "s", "id": 287532, "pid": 76337, "tid": -914061504, "ts": 1716454226189381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226211487, "dur": 11,
    "args": {
      "External id": 287534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287534,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287534, "pid": 5, "tid": 7, "ts": 1716454226211487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189397, "dur": 5,
    "args": {
      "External id": 287534,
      "cbid": 211, "correlation": 287534
    }
  },
  {
    "ph": "s", "id": 287534, "pid": 76337, "tid": -914061504, "ts": 1716454226189397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226211499, "dur": 4,
    "args": {
      "External id": 287536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 287536, "pid": 5, "tid": 7, "ts": 1716454226211499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189405, "dur": 5,
    "args": {
      "External id": 287536,
      "cbid": 211, "correlation": 287536
    }
  },
  {
    "ph": "s", "id": 287536, "pid": 76337, "tid": -914061504, "ts": 1716454226189405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189413, "dur": 0,
    "args": {
      "External id": 287537,
      "cbid": 51, "correlation": 287537
    }
  },
  {
    "ph": "s", "id": 287537, "pid": 76337, "tid": -914061504, "ts": 1716454226189413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226211504, "dur": 2699,
    "args": {
      "External id": 287538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287538,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [1536, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287538, "pid": 5, "tid": 7, "ts": 1716454226211504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189414, "dur": 5,
    "args": {
      "External id": 287538,
      "cbid": 211, "correlation": 287538
    }
  },
  {
    "ph": "s", "id": 287538, "pid": 76337, "tid": -914061504, "ts": 1716454226189414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226214205, "dur": 343,
    "args": {
      "External id": 287543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287543,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287543, "pid": 5, "tid": 7, "ts": 1716454226214205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189442, "dur": 8,
    "args": {
      "External id": 287543,
      "cbid": 211, "correlation": 287543
    }
  },
  {
    "ph": "s", "id": 287543, "pid": 76337, "tid": -914061504, "ts": 1716454226189442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226214549, "dur": 403,
    "args": {
      "External id": 287563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287563,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287563, "pid": 5, "tid": 7, "ts": 1716454226214549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189514, "dur": 12,
    "args": {
      "External id": 287563,
      "cbid": 211, "correlation": 287563
    }
  },
  {
    "ph": "s", "id": 287563, "pid": 76337, "tid": -914061504, "ts": 1716454226189514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226214953, "dur": 4,
    "args": {
      "External id": 287575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287575,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287575, "pid": 5, "tid": 7, "ts": 1716454226214953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189535, "dur": 7,
    "args": {
      "External id": 287575,
      "cbid": 211, "correlation": 287575
    }
  },
  {
    "ph": "s", "id": 287575, "pid": 76337, "tid": -914061504, "ts": 1716454226189535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226214958, "dur": 327,
    "args": {
      "External id": 287578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287578, "pid": 5, "tid": 7, "ts": 1716454226214958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189553, "dur": 6,
    "args": {
      "External id": 287578,
      "cbid": 211, "correlation": 287578
    }
  },
  {
    "ph": "s", "id": 287578, "pid": 76337, "tid": -914061504, "ts": 1716454226189553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226215286, "dur": 210,
    "args": {
      "External id": 287587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287587,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287587, "pid": 5, "tid": 7, "ts": 1716454226215286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189594, "dur": 10,
    "args": {
      "External id": 287587,
      "cbid": 211, "correlation": 287587
    }
  },
  {
    "ph": "s", "id": 287587, "pid": 76337, "tid": -914061504, "ts": 1716454226189594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189661, "dur": 0,
    "args": {
      "External id": 287597,
      "cbid": 317, "correlation": 287597
    }
  },
  {
    "ph": "f", "id": 287597, "pid": 76337, "tid": -914061504, "ts": 1716454226189661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189662, "dur": 0,
    "args": {
      "External id": 287598,
      "cbid": 203, "correlation": 287598
    }
  },
  {
    "ph": "f", "id": 287598, "pid": 76337, "tid": -914061504, "ts": 1716454226189662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189663, "dur": 0,
    "args": {
      "External id": 287599,
      "cbid": 205, "correlation": 287599
    }
  },
  {
    "ph": "f", "id": 287599, "pid": 76337, "tid": -914061504, "ts": 1716454226189663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226215497, "dur": 380,
    "args": {
      "External id": 287603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287603,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 614.4,
      "warps per SM": 4915.2,
      "grid": [1536, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287603, "pid": 5, "tid": 7, "ts": 1716454226215497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189675, "dur": 12,
    "args": {
      "External id": 287603,
      "cbid": 211, "correlation": 287603
    }
  },
  {
    "ph": "s", "id": 287603, "pid": 76337, "tid": -914061504, "ts": 1716454226189675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226215878, "dur": 11,
    "args": {
      "External id": 287605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287605,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 25.6,
      "warps per SM": 204.8,
      "grid": [1, 8, 256],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287605, "pid": 5, "tid": 7, "ts": 1716454226215878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189690, "dur": 6,
    "args": {
      "External id": 287605,
      "cbid": 211, "correlation": 287605
    }
  },
  {
    "ph": "s", "id": 287605, "pid": 76337, "tid": -914061504, "ts": 1716454226189690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226215890, "dur": 4,
    "args": {
      "External id": 287607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4125,
      "warps per SM": 19.3,
      "grid": [193, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 287607, "pid": 5, "tid": 7, "ts": 1716454226215890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189700, "dur": 6,
    "args": {
      "External id": 287607,
      "cbid": 211, "correlation": 287607
    }
  },
  {
    "ph": "s", "id": 287607, "pid": 76337, "tid": -914061504, "ts": 1716454226189700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189708, "dur": 0,
    "args": {
      "External id": 287608,
      "cbid": 51, "correlation": 287608
    }
  },
  {
    "ph": "s", "id": 287608, "pid": 76337, "tid": -914061504, "ts": 1716454226189708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226215895, "dur": 2759,
    "args": {
      "External id": 287609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287609,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [1536, 2, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287609, "pid": 5, "tid": 7, "ts": 1716454226215895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189709, "dur": 5,
    "args": {
      "External id": 287609,
      "cbid": 211, "correlation": 287609
    }
  },
  {
    "ph": "s", "id": 287609, "pid": 76337, "tid": -914061504, "ts": 1716454226189709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226218656, "dur": 347,
    "args": {
      "External id": 287614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287614, "pid": 5, "tid": 7, "ts": 1716454226218656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189736, "dur": 9,
    "args": {
      "External id": 287614,
      "cbid": 211, "correlation": 287614
    }
  },
  {
    "ph": "s", "id": 287614, "pid": 76337, "tid": -914061504, "ts": 1716454226189736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226219004, "dur": 906,
    "args": {
      "External id": 287622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287622,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287622, "pid": 5, "tid": 7, "ts": 1716454226219004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189771, "dur": 9,
    "args": {
      "External id": 287622,
      "cbid": 211, "correlation": 287622
    }
  },
  {
    "ph": "s", "id": 287622, "pid": 76337, "tid": -914061504, "ts": 1716454226189771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226219911, "dur": 203,
    "args": {
      "External id": 287630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287630,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1228.8,
      "warps per SM": 4915.2,
      "grid": [98304, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287630, "pid": 5, "tid": 7, "ts": 1716454226219911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189801, "dur": 8,
    "args": {
      "External id": 287630,
      "cbid": 211, "correlation": 287630
    }
  },
  {
    "ph": "s", "id": 287630, "pid": 76337, "tid": -914061504, "ts": 1716454226189801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189885, "dur": 1,
    "args": {
      "External id": 287638,
      "cbid": 317, "correlation": 287638
    }
  },
  {
    "ph": "f", "id": 287638, "pid": 76337, "tid": -914061504, "ts": 1716454226189885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226189887, "dur": 1499,
    "args": {
      "External id": 287639,
      "cbid": 20, "correlation": 287639
    }
  },
  {
    "ph": "f", "id": 287639, "pid": 76337, "tid": -914061504, "ts": 1716454226189887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_nhwc_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float, unsigned long)", "pid": 5, "tid": 7,
    "ts": 1716454226220116, "dur": 2652,
    "args": {
      "External id": 287643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287643,
      "registers per thread": 26,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 78643.2,
      "grid": [196608, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287643, "pid": 5, "tid": 7, "ts": 1716454226220116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226191403, "dur": 19,
    "args": {
      "External id": 287643,
      "cbid": 211, "correlation": 287643
    }
  },
  {
    "ph": "s", "id": 287643, "pid": 76337, "tid": -914061504, "ts": 1716454226191403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226191471, "dur": 1,
    "args": {
      "External id": 287651,
      "cbid": 317, "correlation": 287651
    }
  },
  {
    "ph": "f", "id": 287651, "pid": 76337, "tid": -914061504, "ts": 1716454226191471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226191473, "dur": 9562,
    "args": {
      "External id": 287652,
      "cbid": 20, "correlation": 287652
    }
  },
  {
    "ph": "f", "id": 287652, "pid": 76337, "tid": -914061504, "ts": 1716454226191473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226222768, "dur": 9,
    "args": {
      "External id": 287661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 287661, "pid": 5, "tid": 7, "ts": 1716454226222768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226201107, "dur": 28,
    "args": {
      "External id": 287661,
      "cbid": 211, "correlation": 287661
    }
  },
  {
    "ph": "s", "id": 287661, "pid": 76337, "tid": -914061504, "ts": 1716454226201107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226201152, "dur": 1,
    "args": {
      "External id": 287668,
      "cbid": 317, "correlation": 287668
    }
  },
  {
    "ph": "f", "id": 287668, "pid": 76337, "tid": -914061504, "ts": 1716454226201152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226201153, "dur": 1,
    "args": {
      "External id": 287669,
      "cbid": 203, "correlation": 287669
    }
  },
  {
    "ph": "f", "id": 287669, "pid": 76337, "tid": -914061504, "ts": 1716454226201153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226201155, "dur": 1,
    "args": {
      "External id": 287670,
      "cbid": 205, "correlation": 287670
    }
  },
  {
    "ph": "f", "id": 287670, "pid": 76337, "tid": -914061504, "ts": 1716454226201155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226222779, "dur": 10232,
    "args": {
      "External id": 287674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287674,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [2, 6144, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287674, "pid": 5, "tid": 7, "ts": 1716454226222779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226203468, "dur": 22,
    "args": {
      "External id": 287674,
      "cbid": 211, "correlation": 287674
    }
  },
  {
    "ph": "s", "id": 287674, "pid": 76337, "tid": -914061504, "ts": 1716454226203468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226233012, "dur": 1292,
    "args": {
      "External id": 287680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 4915.2,
      "warps per SM": 19660.8,
      "grid": [393216, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287680, "pid": 5, "tid": 7, "ts": 1716454226233012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226204082, "dur": 14,
    "args": {
      "External id": 287680,
      "cbid": 211, "correlation": 287680
    }
  },
  {
    "ph": "s", "id": 287680, "pid": 76337, "tid": -914061504, "ts": 1716454226204082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226234305, "dur": 20760,
    "args": {
      "External id": 287690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287690,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 4915.2,
      "warps per SM": 19660.8,
      "grid": [393216, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287690, "pid": 5, "tid": 7, "ts": 1716454226234305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226204327, "dur": 14,
    "args": {
      "External id": 287690,
      "cbid": 211, "correlation": 287690
    }
  },
  {
    "ph": "s", "id": 287690, "pid": 76337, "tid": -914061504, "ts": 1716454226204327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226204368, "dur": 1,
    "args": {
      "External id": 287700,
      "cbid": 317, "correlation": 287700
    }
  },
  {
    "ph": "f", "id": 287700, "pid": 76337, "tid": -914061504, "ts": 1716454226204368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226204369, "dur": 12140,
    "args": {
      "External id": 287701,
      "cbid": 20, "correlation": 287701
    }
  },
  {
    "ph": "f", "id": 287701, "pid": 76337, "tid": -914061504, "ts": 1716454226204369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226255066, "dur": 1688,
    "args": {
      "External id": 287715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287715,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287715, "pid": 5, "tid": 7, "ts": 1716454226255066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216535, "dur": 18,
    "args": {
      "External id": 287715,
      "cbid": 211, "correlation": 287715
    }
  },
  {
    "ph": "s", "id": 287715, "pid": 76337, "tid": -914061504, "ts": 1716454226216535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226256755, "dur": 4,
    "args": {
      "External id": 287727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287727,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 287727, "pid": 5, "tid": 7, "ts": 1716454226256755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216573, "dur": 10,
    "args": {
      "External id": 287727,
      "cbid": 211, "correlation": 287727
    }
  },
  {
    "ph": "s", "id": 287727, "pid": 76337, "tid": -914061504, "ts": 1716454226216573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226256760, "dur": 1298,
    "args": {
      "External id": 287730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 4915.2,
      "warps per SM": 19660.8,
      "grid": [393216, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287730, "pid": 5, "tid": 7, "ts": 1716454226256760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216603, "dur": 8,
    "args": {
      "External id": 287730,
      "cbid": 211, "correlation": 287730
    }
  },
  {
    "ph": "s", "id": 287730, "pid": 76337, "tid": -914061504, "ts": 1716454226216603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226258060, "dur": 827,
    "args": {
      "External id": 287739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287739,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4915.2,
      "warps per SM": 19660.8,
      "grid": [393216, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287739, "pid": 5, "tid": 7, "ts": 1716454226258060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216672, "dur": 12,
    "args": {
      "External id": 287739,
      "cbid": 211, "correlation": 287739
    }
  },
  {
    "ph": "s", "id": 287739, "pid": 76337, "tid": -914061504, "ts": 1716454226216672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216766, "dur": 0,
    "args": {
      "External id": 287749,
      "cbid": 317, "correlation": 287749
    }
  },
  {
    "ph": "f", "id": 287749, "pid": 76337, "tid": -914061504, "ts": 1716454226216766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216767, "dur": 0,
    "args": {
      "External id": 287750,
      "cbid": 203, "correlation": 287750
    }
  },
  {
    "ph": "f", "id": 287750, "pid": 76337, "tid": -914061504, "ts": 1716454226216767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226216768, "dur": 0,
    "args": {
      "External id": 287751,
      "cbid": 205, "correlation": 287751
    }
  },
  {
    "ph": "f", "id": 287751, "pid": 76337, "tid": -914061504, "ts": 1716454226216768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226219182, "dur": 2,
    "args": {
      "External id": 287757,
      "cbid": 317, "correlation": 287757
    }
  },
  {
    "ph": "f", "id": 287757, "pid": 76337, "tid": -914061504, "ts": 1716454226219182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMalloc", "pid": 76337, "tid": -914061504,
    "ts": 1716454226219184, "dur": 669555,
    "args": {
      "External id": 287758,
      "cbid": 20, "correlation": 287758
    }
  },
  {
    "ph": "f", "id": 287758, "pid": 76337, "tid": -914061504, "ts": 1716454226219184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226888850, "dur": 1608,
    "args": {
      "External id": 287759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287759,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 2457.6,
      "warps per SM": 19660.8,
      "grid": [6144, 8, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287759, "pid": 5, "tid": 7, "ts": 1716454226888850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226888792, "dur": 61,
    "args": {
      "External id": 287759,
      "cbid": 211, "correlation": 287759
    }
  },
  {
    "ph": "s", "id": 287759, "pid": 76337, "tid": -914061504, "ts": 1716454226888792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226890460, "dur": 8,
    "args": {
      "External id": 287761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12.8,
      "warps per SM": 102.4,
      "grid": [1, 8, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287761, "pid": 5, "tid": 7, "ts": 1716454226890460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226888856, "dur": 7,
    "args": {
      "External id": 287761,
      "cbid": 211, "correlation": 287761
    }
  },
  {
    "ph": "s", "id": 287761, "pid": 76337, "tid": -914061504, "ts": 1716454226888856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226890469, "dur": 5,
    "args": {
      "External id": 287763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287763, "pid": 5, "tid": 7, "ts": 1716454226890469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226888875, "dur": 13,
    "args": {
      "External id": 287763,
      "cbid": 211, "correlation": 287763
    }
  },
  {
    "ph": "s", "id": 287763, "pid": 76337, "tid": -914061504, "ts": 1716454226888875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226888895, "dur": 0,
    "args": {
      "External id": 287764,
      "cbid": 51, "correlation": 287764
    }
  },
  {
    "ph": "s", "id": 287764, "pid": 76337, "tid": -914061504, "ts": 1716454226888895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226890475, "dur": 5320,
    "args": {
      "External id": 287765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287765,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287765, "pid": 5, "tid": 7, "ts": 1716454226890475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226888896, "dur": 11,
    "args": {
      "External id": 287765,
      "cbid": 211, "correlation": 287765
    }
  },
  {
    "ph": "s", "id": 287765, "pid": 76337, "tid": -914061504, "ts": 1716454226888896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226895796, "dur": 683,
    "args": {
      "External id": 287770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287770, "pid": 5, "tid": 7, "ts": 1716454226895796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226889825, "dur": 18,
    "args": {
      "External id": 287770,
      "cbid": 211, "correlation": 287770
    }
  },
  {
    "ph": "s", "id": 287770, "pid": 76337, "tid": -914061504, "ts": 1716454226889825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226896481, "dur": 798,
    "args": {
      "External id": 287790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287790,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287790, "pid": 5, "tid": 7, "ts": 1716454226896481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890088, "dur": 15,
    "args": {
      "External id": 287790,
      "cbid": 211, "correlation": 287790
    }
  },
  {
    "ph": "s", "id": 287790, "pid": 76337, "tid": -914061504, "ts": 1716454226890088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226897281, "dur": 4,
    "args": {
      "External id": 287802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287802,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 287802, "pid": 5, "tid": 7, "ts": 1716454226897281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890117, "dur": 9,
    "args": {
      "External id": 287802,
      "cbid": 211, "correlation": 287802
    }
  },
  {
    "ph": "s", "id": 287802, "pid": 76337, "tid": -914061504, "ts": 1716454226890117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226897286, "dur": 655,
    "args": {
      "External id": 287805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287805, "pid": 5, "tid": 7, "ts": 1716454226897286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890150, "dur": 8,
    "args": {
      "External id": 287805,
      "cbid": 211, "correlation": 287805
    }
  },
  {
    "ph": "s", "id": 287805, "pid": 76337, "tid": -914061504, "ts": 1716454226890150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226897942, "dur": 416,
    "args": {
      "External id": 287814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287814,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287814, "pid": 5, "tid": 7, "ts": 1716454226897942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890217, "dur": 14,
    "args": {
      "External id": 287814,
      "cbid": 211, "correlation": 287814
    }
  },
  {
    "ph": "s", "id": 287814, "pid": 76337, "tid": -914061504, "ts": 1716454226890217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890353, "dur": 1,
    "args": {
      "External id": 287824,
      "cbid": 317, "correlation": 287824
    }
  },
  {
    "ph": "f", "id": 287824, "pid": 76337, "tid": -914061504, "ts": 1716454226890353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890355, "dur": 1,
    "args": {
      "External id": 287825,
      "cbid": 203, "correlation": 287825
    }
  },
  {
    "ph": "f", "id": 287825, "pid": 76337, "tid": -914061504, "ts": 1716454226890355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226890357, "dur": 1,
    "args": {
      "External id": 287826,
      "cbid": 205, "correlation": 287826
    }
  },
  {
    "ph": "f", "id": 287826, "pid": 76337, "tid": -914061504, "ts": 1716454226890357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226898359, "dur": 514,
    "args": {
      "External id": 287830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287830,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [6144, 4, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287830, "pid": 5, "tid": 7, "ts": 1716454226898359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893084, "dur": 24,
    "args": {
      "External id": 287830,
      "cbid": 211, "correlation": 287830
    }
  },
  {
    "ph": "s", "id": 287830, "pid": 76337, "tid": -914061504, "ts": 1716454226893084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226898874, "dur": 6,
    "args": {
      "External id": 287832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287832,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 287832, "pid": 5, "tid": 7, "ts": 1716454226898874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893111, "dur": 6,
    "args": {
      "External id": 287832,
      "cbid": 211, "correlation": 287832
    }
  },
  {
    "ph": "s", "id": 287832, "pid": 76337, "tid": -914061504, "ts": 1716454226893111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226898881, "dur": 5,
    "args": {
      "External id": 287834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287834,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287834, "pid": 5, "tid": 7, "ts": 1716454226898881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893124, "dur": 9,
    "args": {
      "External id": 287834,
      "cbid": 211, "correlation": 287834
    }
  },
  {
    "ph": "s", "id": 287834, "pid": 76337, "tid": -914061504, "ts": 1716454226893124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893136, "dur": 1,
    "args": {
      "External id": 287835,
      "cbid": 51, "correlation": 287835
    }
  },
  {
    "ph": "s", "id": 287835, "pid": 76337, "tid": -914061504, "ts": 1716454226893136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226898888, "dur": 2852,
    "args": {
      "External id": 287836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287836,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287836, "pid": 5, "tid": 7, "ts": 1716454226898888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893138, "dur": 9,
    "args": {
      "External id": 287836,
      "cbid": 211, "correlation": 287836
    }
  },
  {
    "ph": "s", "id": 287836, "pid": 76337, "tid": -914061504, "ts": 1716454226893138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226901741, "dur": 687,
    "args": {
      "External id": 287841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287841,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287841, "pid": 5, "tid": 7, "ts": 1716454226901741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893846, "dur": 13,
    "args": {
      "External id": 287841,
      "cbid": 211, "correlation": 287841
    }
  },
  {
    "ph": "s", "id": 287841, "pid": 76337, "tid": -914061504, "ts": 1716454226893846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893944, "dur": 0,
    "args": {
      "External id": 287851,
      "cbid": 317, "correlation": 287851
    }
  },
  {
    "ph": "f", "id": 287851, "pid": 76337, "tid": -914061504, "ts": 1716454226893944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893945, "dur": 0,
    "args": {
      "External id": 287852,
      "cbid": 203, "correlation": 287852
    }
  },
  {
    "ph": "f", "id": 287852, "pid": 76337, "tid": -914061504, "ts": 1716454226893945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226893946, "dur": 0,
    "args": {
      "External id": 287853,
      "cbid": 205, "correlation": 287853
    }
  },
  {
    "ph": "f", "id": 287853, "pid": 76337, "tid": -914061504, "ts": 1716454226893946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226902429, "dur": 900,
    "args": {
      "External id": 287857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287857,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [1, 6144, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287857, "pid": 5, "tid": 7, "ts": 1716454226902429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226895635, "dur": 70,
    "args": {
      "External id": 287857,
      "cbid": 211, "correlation": 287857
    }
  },
  {
    "ph": "s", "id": 287857, "pid": 76337, "tid": -914061504, "ts": 1716454226895635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226903331, "dur": 651,
    "args": {
      "External id": 287863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287863, "pid": 5, "tid": 7, "ts": 1716454226903331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896170, "dur": 12,
    "args": {
      "External id": 287863,
      "cbid": 211, "correlation": 287863
    }
  },
  {
    "ph": "s", "id": 287863, "pid": 76337, "tid": -914061504, "ts": 1716454226896170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226903983, "dur": 2699,
    "args": {
      "External id": 287871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287871,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287871, "pid": 5, "tid": 7, "ts": 1716454226903983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896224, "dur": 9,
    "args": {
      "External id": 287871,
      "cbid": 211, "correlation": 287871
    }
  },
  {
    "ph": "s", "id": 287871, "pid": 76337, "tid": -914061504, "ts": 1716454226896224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226906683, "dur": 404,
    "args": {
      "External id": 287879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287879,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287879, "pid": 5, "tid": 7, "ts": 1716454226906683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896278, "dur": 14,
    "args": {
      "External id": 287879,
      "cbid": 211, "correlation": 287879
    }
  },
  {
    "ph": "s", "id": 287879, "pid": 76337, "tid": -914061504, "ts": 1716454226896278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226907088, "dur": 7129,
    "args": {
      "External id": 287889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287889,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287889, "pid": 5, "tid": 7, "ts": 1716454226907088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896429, "dur": 16,
    "args": {
      "External id": 287889,
      "cbid": 211, "correlation": 287889
    }
  },
  {
    "ph": "s", "id": 287889, "pid": 76337, "tid": -914061504, "ts": 1716454226896429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226914219, "dur": 791,
    "args": {
      "External id": 287910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287910,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287910, "pid": 5, "tid": 7, "ts": 1716454226914219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896477, "dur": 9,
    "args": {
      "External id": 287910,
      "cbid": 211, "correlation": 287910
    }
  },
  {
    "ph": "s", "id": 287910, "pid": 76337, "tid": -914061504, "ts": 1716454226896477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226915011, "dur": 4,
    "args": {
      "External id": 287922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287922,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 287922, "pid": 5, "tid": 7, "ts": 1716454226915011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896498, "dur": 8,
    "args": {
      "External id": 287922,
      "cbid": 211, "correlation": 287922
    }
  },
  {
    "ph": "s", "id": 287922, "pid": 76337, "tid": -914061504, "ts": 1716454226896498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226915017, "dur": 650,
    "args": {
      "External id": 287925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287925, "pid": 5, "tid": 7, "ts": 1716454226915017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896525, "dur": 8,
    "args": {
      "External id": 287925,
      "cbid": 211, "correlation": 287925
    }
  },
  {
    "ph": "s", "id": 287925, "pid": 76337, "tid": -914061504, "ts": 1716454226896525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226915668, "dur": 415,
    "args": {
      "External id": 287934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287934,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287934, "pid": 5, "tid": 7, "ts": 1716454226915668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896580, "dur": 11,
    "args": {
      "External id": 287934,
      "cbid": 211, "correlation": 287934
    }
  },
  {
    "ph": "s", "id": 287934, "pid": 76337, "tid": -914061504, "ts": 1716454226896580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896646, "dur": 0,
    "args": {
      "External id": 287944,
      "cbid": 317, "correlation": 287944
    }
  },
  {
    "ph": "f", "id": 287944, "pid": 76337, "tid": -914061504, "ts": 1716454226896646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896647, "dur": 0,
    "args": {
      "External id": 287945,
      "cbid": 203, "correlation": 287945
    }
  },
  {
    "ph": "f", "id": 287945, "pid": 76337, "tid": -914061504, "ts": 1716454226896647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896648, "dur": 0,
    "args": {
      "External id": 287946,
      "cbid": 205, "correlation": 287946
    }
  },
  {
    "ph": "f", "id": 287946, "pid": 76337, "tid": -914061504, "ts": 1716454226896648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226916085, "dur": 516,
    "args": {
      "External id": 287950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287950,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [6144, 4, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287950, "pid": 5, "tid": 7, "ts": 1716454226916085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896666, "dur": 13,
    "args": {
      "External id": 287950,
      "cbid": 211, "correlation": 287950
    }
  },
  {
    "ph": "s", "id": 287950, "pid": 76337, "tid": -914061504, "ts": 1716454226896666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226916602, "dur": 5,
    "args": {
      "External id": 287952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 287952, "pid": 5, "tid": 7, "ts": 1716454226916602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896682, "dur": 5,
    "args": {
      "External id": 287952,
      "cbid": 211, "correlation": 287952
    }
  },
  {
    "ph": "s", "id": 287952, "pid": 76337, "tid": -914061504, "ts": 1716454226896682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226916609, "dur": 5,
    "args": {
      "External id": 287954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287954, "pid": 5, "tid": 7, "ts": 1716454226916609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896691, "dur": 6,
    "args": {
      "External id": 287954,
      "cbid": 211, "correlation": 287954
    }
  },
  {
    "ph": "s", "id": 287954, "pid": 76337, "tid": -914061504, "ts": 1716454226896691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896701, "dur": 0,
    "args": {
      "External id": 287955,
      "cbid": 51, "correlation": 287955
    }
  },
  {
    "ph": "s", "id": 287955, "pid": 76337, "tid": -914061504, "ts": 1716454226896701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226916615, "dur": 2835,
    "args": {
      "External id": 287956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287956,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 287956, "pid": 5, "tid": 7, "ts": 1716454226916615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896702, "dur": 7,
    "args": {
      "External id": 287956,
      "cbid": 211, "correlation": 287956
    }
  },
  {
    "ph": "s", "id": 287956, "pid": 76337, "tid": -914061504, "ts": 1716454226896702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226919452, "dur": 682,
    "args": {
      "External id": 287961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287961, "pid": 5, "tid": 7, "ts": 1716454226919452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896734, "dur": 9,
    "args": {
      "External id": 287961,
      "cbid": 211, "correlation": 287961
    }
  },
  {
    "ph": "s", "id": 287961, "pid": 76337, "tid": -914061504, "ts": 1716454226896734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226920136, "dur": 794,
    "args": {
      "External id": 287981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287981,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 287981, "pid": 5, "tid": 7, "ts": 1716454226920136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896807, "dur": 12,
    "args": {
      "External id": 287981,
      "cbid": 211, "correlation": 287981
    }
  },
  {
    "ph": "s", "id": 287981, "pid": 76337, "tid": -914061504, "ts": 1716454226896807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226920931, "dur": 4,
    "args": {
      "External id": 287993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287993,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 287993, "pid": 5, "tid": 7, "ts": 1716454226920931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896830, "dur": 6,
    "args": {
      "External id": 287993,
      "cbid": 211, "correlation": 287993
    }
  },
  {
    "ph": "s", "id": 287993, "pid": 76337, "tid": -914061504, "ts": 1716454226896830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226920936, "dur": 652,
    "args": {
      "External id": 287996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 287996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 287996, "pid": 5, "tid": 7, "ts": 1716454226920936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896847, "dur": 7,
    "args": {
      "External id": 287996,
      "cbid": 211, "correlation": 287996
    }
  },
  {
    "ph": "s", "id": 287996, "pid": 76337, "tid": -914061504, "ts": 1716454226896847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226921590, "dur": 415,
    "args": {
      "External id": 288005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288005, "pid": 5, "tid": 7, "ts": 1716454226921590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896888, "dur": 10,
    "args": {
      "External id": 288005,
      "cbid": 211, "correlation": 288005
    }
  },
  {
    "ph": "s", "id": 288005, "pid": 76337, "tid": -914061504, "ts": 1716454226896888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896961, "dur": 0,
    "args": {
      "External id": 288015,
      "cbid": 317, "correlation": 288015
    }
  },
  {
    "ph": "f", "id": 288015, "pid": 76337, "tid": -914061504, "ts": 1716454226896961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896962, "dur": 0,
    "args": {
      "External id": 288016,
      "cbid": 203, "correlation": 288016
    }
  },
  {
    "ph": "f", "id": 288016, "pid": 76337, "tid": -914061504, "ts": 1716454226896962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896963, "dur": 0,
    "args": {
      "External id": 288017,
      "cbid": 205, "correlation": 288017
    }
  },
  {
    "ph": "f", "id": 288017, "pid": 76337, "tid": -914061504, "ts": 1716454226896963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226922006, "dur": 524,
    "args": {
      "External id": 288021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [6144, 4, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288021, "pid": 5, "tid": 7, "ts": 1716454226922006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896983, "dur": 13,
    "args": {
      "External id": 288021,
      "cbid": 211, "correlation": 288021
    }
  },
  {
    "ph": "s", "id": 288021, "pid": 76337, "tid": -914061504, "ts": 1716454226896983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226922531, "dur": 6,
    "args": {
      "External id": 288023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 288023, "pid": 5, "tid": 7, "ts": 1716454226922531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226896998, "dur": 5,
    "args": {
      "External id": 288023,
      "cbid": 211, "correlation": 288023
    }
  },
  {
    "ph": "s", "id": 288023, "pid": 76337, "tid": -914061504, "ts": 1716454226896998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226922538, "dur": 5,
    "args": {
      "External id": 288025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288025, "pid": 5, "tid": 7, "ts": 1716454226922538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897007, "dur": 7,
    "args": {
      "External id": 288025,
      "cbid": 211, "correlation": 288025
    }
  },
  {
    "ph": "s", "id": 288025, "pid": 76337, "tid": -914061504, "ts": 1716454226897007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897017, "dur": 0,
    "args": {
      "External id": 288026,
      "cbid": 51, "correlation": 288026
    }
  },
  {
    "ph": "s", "id": 288026, "pid": 76337, "tid": -914061504, "ts": 1716454226897017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226922544, "dur": 2844,
    "args": {
      "External id": 288027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288027,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 288027, "pid": 5, "tid": 7, "ts": 1716454226922544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897017, "dur": 5,
    "args": {
      "External id": 288027,
      "cbid": 211, "correlation": 288027
    }
  },
  {
    "ph": "s", "id": 288027, "pid": 76337, "tid": -914061504, "ts": 1716454226897017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226925389, "dur": 685,
    "args": {
      "External id": 288032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288032, "pid": 5, "tid": 7, "ts": 1716454226925389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897045, "dur": 8,
    "args": {
      "External id": 288032,
      "cbid": 211, "correlation": 288032
    }
  },
  {
    "ph": "s", "id": 288032, "pid": 76337, "tid": -914061504, "ts": 1716454226897045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226926075, "dur": 2703,
    "args": {
      "External id": 288040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288040, "pid": 5, "tid": 7, "ts": 1716454226926075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897080, "dur": 9,
    "args": {
      "External id": 288040,
      "cbid": 211, "correlation": 288040
    }
  },
  {
    "ph": "s", "id": 288040, "pid": 76337, "tid": -914061504, "ts": 1716454226897080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226928780, "dur": 402,
    "args": {
      "External id": 288048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288048,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288048, "pid": 5, "tid": 7, "ts": 1716454226928780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897112, "dur": 9,
    "args": {
      "External id": 288048,
      "cbid": 211, "correlation": 288048
    }
  },
  {
    "ph": "s", "id": 288048, "pid": 76337, "tid": -914061504, "ts": 1716454226897112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226929183, "dur": 7125,
    "args": {
      "External id": 288058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288058,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288058, "pid": 5, "tid": 7, "ts": 1716454226929183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897189, "dur": 13,
    "args": {
      "External id": 288058,
      "cbid": 211, "correlation": 288058
    }
  },
  {
    "ph": "s", "id": 288058, "pid": 76337, "tid": -914061504, "ts": 1716454226897189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226936310, "dur": 793,
    "args": {
      "External id": 288079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288079,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 288079, "pid": 5, "tid": 7, "ts": 1716454226936310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897223, "dur": 8,
    "args": {
      "External id": 288079,
      "cbid": 211, "correlation": 288079
    }
  },
  {
    "ph": "s", "id": 288079, "pid": 76337, "tid": -914061504, "ts": 1716454226897223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226937103, "dur": 4,
    "args": {
      "External id": 288091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288091,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 288091, "pid": 5, "tid": 7, "ts": 1716454226937103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897240, "dur": 6,
    "args": {
      "External id": 288091,
      "cbid": 211, "correlation": 288091
    }
  },
  {
    "ph": "s", "id": 288091, "pid": 76337, "tid": -914061504, "ts": 1716454226897240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226937108, "dur": 655,
    "args": {
      "External id": 288094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288094, "pid": 5, "tid": 7, "ts": 1716454226937108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897257, "dur": 7,
    "args": {
      "External id": 288094,
      "cbid": 211, "correlation": 288094
    }
  },
  {
    "ph": "s", "id": 288094, "pid": 76337, "tid": -914061504, "ts": 1716454226897257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226937764, "dur": 416,
    "args": {
      "External id": 288103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288103,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288103, "pid": 5, "tid": 7, "ts": 1716454226937764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897297, "dur": 10,
    "args": {
      "External id": 288103,
      "cbid": 211, "correlation": 288103
    }
  },
  {
    "ph": "s", "id": 288103, "pid": 76337, "tid": -914061504, "ts": 1716454226897297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897351, "dur": 0,
    "args": {
      "External id": 288113,
      "cbid": 317, "correlation": 288113
    }
  },
  {
    "ph": "f", "id": 288113, "pid": 76337, "tid": -914061504, "ts": 1716454226897351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897352, "dur": 0,
    "args": {
      "External id": 288114,
      "cbid": 203, "correlation": 288114
    }
  },
  {
    "ph": "f", "id": 288114, "pid": 76337, "tid": -914061504, "ts": 1716454226897352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897352, "dur": 0,
    "args": {
      "External id": 288115,
      "cbid": 205, "correlation": 288115
    }
  },
  {
    "ph": "f", "id": 288115, "pid": 76337, "tid": -914061504, "ts": 1716454226897352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226938181, "dur": 521,
    "args": {
      "External id": 288119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288119,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [6144, 4, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288119, "pid": 5, "tid": 7, "ts": 1716454226938181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897366, "dur": 12,
    "args": {
      "External id": 288119,
      "cbid": 211, "correlation": 288119
    }
  },
  {
    "ph": "s", "id": 288119, "pid": 76337, "tid": -914061504, "ts": 1716454226897366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226938704, "dur": 6,
    "args": {
      "External id": 288121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288121,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 288121, "pid": 5, "tid": 7, "ts": 1716454226938704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897380, "dur": 5,
    "args": {
      "External id": 288121,
      "cbid": 211, "correlation": 288121
    }
  },
  {
    "ph": "s", "id": 288121, "pid": 76337, "tid": -914061504, "ts": 1716454226897380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226938711, "dur": 5,
    "args": {
      "External id": 288123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288123, "pid": 5, "tid": 7, "ts": 1716454226938711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897389, "dur": 6,
    "args": {
      "External id": 288123,
      "cbid": 211, "correlation": 288123
    }
  },
  {
    "ph": "s", "id": 288123, "pid": 76337, "tid": -914061504, "ts": 1716454226897389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897398, "dur": 0,
    "args": {
      "External id": 288124,
      "cbid": 51, "correlation": 288124
    }
  },
  {
    "ph": "s", "id": 288124, "pid": 76337, "tid": -914061504, "ts": 1716454226897398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226938717, "dur": 2828,
    "args": {
      "External id": 288125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288125,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 288125, "pid": 5, "tid": 7, "ts": 1716454226938717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897398, "dur": 5,
    "args": {
      "External id": 288125,
      "cbid": 211, "correlation": 288125
    }
  },
  {
    "ph": "s", "id": 288125, "pid": 76337, "tid": -914061504, "ts": 1716454226897398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226941546, "dur": 685,
    "args": {
      "External id": 288130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288130, "pid": 5, "tid": 7, "ts": 1716454226941546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897426, "dur": 9,
    "args": {
      "External id": 288130,
      "cbid": 211, "correlation": 288130
    }
  },
  {
    "ph": "s", "id": 288130, "pid": 76337, "tid": -914061504, "ts": 1716454226897426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226942232, "dur": 798,
    "args": {
      "External id": 288150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288150,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 288150, "pid": 5, "tid": 7, "ts": 1716454226942232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897498, "dur": 12,
    "args": {
      "External id": 288150,
      "cbid": 211, "correlation": 288150
    }
  },
  {
    "ph": "s", "id": 288150, "pid": 76337, "tid": -914061504, "ts": 1716454226897498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226943032, "dur": 4,
    "args": {
      "External id": 288162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288162,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 288162, "pid": 5, "tid": 7, "ts": 1716454226943032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897520, "dur": 6,
    "args": {
      "External id": 288162,
      "cbid": 211, "correlation": 288162
    }
  },
  {
    "ph": "s", "id": 288162, "pid": 76337, "tid": -914061504, "ts": 1716454226897520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226943036, "dur": 653,
    "args": {
      "External id": 288165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288165, "pid": 5, "tid": 7, "ts": 1716454226943036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897537, "dur": 6,
    "args": {
      "External id": 288165,
      "cbid": 211, "correlation": 288165
    }
  },
  {
    "ph": "s", "id": 288165, "pid": 76337, "tid": -914061504, "ts": 1716454226897537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226943691, "dur": 414,
    "args": {
      "External id": 288174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288174,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288174, "pid": 5, "tid": 7, "ts": 1716454226943691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897577, "dur": 10,
    "args": {
      "External id": 288174,
      "cbid": 211, "correlation": 288174
    }
  },
  {
    "ph": "s", "id": 288174, "pid": 76337, "tid": -914061504, "ts": 1716454226897577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897643, "dur": 0,
    "args": {
      "External id": 288184,
      "cbid": 317, "correlation": 288184
    }
  },
  {
    "ph": "f", "id": 288184, "pid": 76337, "tid": -914061504, "ts": 1716454226897643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897644, "dur": 0,
    "args": {
      "External id": 288185,
      "cbid": 203, "correlation": 288185
    }
  },
  {
    "ph": "f", "id": 288185, "pid": 76337, "tid": -914061504, "ts": 1716454226897644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897645, "dur": 0,
    "args": {
      "External id": 288186,
      "cbid": 205, "correlation": 288186
    }
  },
  {
    "ph": "f", "id": 288186, "pid": 76337, "tid": -914061504, "ts": 1716454226897645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226944107, "dur": 523,
    "args": {
      "External id": 288190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [6144, 4, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288190, "pid": 5, "tid": 7, "ts": 1716454226944107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897658, "dur": 12,
    "args": {
      "External id": 288190,
      "cbid": 211, "correlation": 288190
    }
  },
  {
    "ph": "s", "id": 288190, "pid": 76337, "tid": -914061504, "ts": 1716454226897658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226944631, "dur": 5,
    "args": {
      "External id": 288192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 6.4,
      "warps per SM": 51.2,
      "grid": [1, 4, 128],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 288192, "pid": 5, "tid": 7, "ts": 1716454226944631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897672, "dur": 5,
    "args": {
      "External id": 288192,
      "cbid": 211, "correlation": 288192
    }
  },
  {
    "ph": "s", "id": 288192, "pid": 76337, "tid": -914061504, "ts": 1716454226897672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454226944638, "dur": 5,
    "args": {
      "External id": 288194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288194,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 9.6125,
      "warps per SM": 76.9,
      "grid": [769, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288194, "pid": 5, "tid": 7, "ts": 1716454226944638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897681, "dur": 5,
    "args": {
      "External id": 288194,
      "cbid": 211, "correlation": 288194
    }
  },
  {
    "ph": "s", "id": 288194, "pid": 76337, "tid": -914061504, "ts": 1716454226897681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897690, "dur": 0,
    "args": {
      "External id": 288195,
      "cbid": 51, "correlation": 288195
    }
  },
  {
    "ph": "s", "id": 288195, "pid": 76337, "tid": -914061504, "ts": 1716454226897690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454226944644, "dur": 2833,
    "args": {
      "External id": 288196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288196,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 288196, "pid": 5, "tid": 7, "ts": 1716454226944644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897690, "dur": 5,
    "args": {
      "External id": 288196,
      "cbid": 211, "correlation": 288196
    }
  },
  {
    "ph": "s", "id": 288196, "pid": 76337, "tid": -914061504, "ts": 1716454226897690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226947478, "dur": 685,
    "args": {
      "External id": 288201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288201,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288201, "pid": 5, "tid": 7, "ts": 1716454226947478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897718, "dur": 9,
    "args": {
      "External id": 288201,
      "cbid": 211, "correlation": 288201
    }
  },
  {
    "ph": "s", "id": 288201, "pid": 76337, "tid": -914061504, "ts": 1716454226897718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226948164, "dur": 2702,
    "args": {
      "External id": 288209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288209, "pid": 5, "tid": 7, "ts": 1716454226948164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897752, "dur": 9,
    "args": {
      "External id": 288209,
      "cbid": 211, "correlation": 288209
    }
  },
  {
    "ph": "s", "id": 288209, "pid": 76337, "tid": -914061504, "ts": 1716454226897752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226950867, "dur": 401,
    "args": {
      "External id": 288217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288217,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288217, "pid": 5, "tid": 7, "ts": 1716454226950867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897782, "dur": 8,
    "args": {
      "External id": 288217,
      "cbid": 211, "correlation": 288217
    }
  },
  {
    "ph": "s", "id": 288217, "pid": 76337, "tid": -914061504, "ts": 1716454226897782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226951270, "dur": 7119,
    "args": {
      "External id": 288227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288227,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288227, "pid": 5, "tid": 7, "ts": 1716454226951270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897855, "dur": 12,
    "args": {
      "External id": 288227,
      "cbid": 211, "correlation": 288227
    }
  },
  {
    "ph": "s", "id": 288227, "pid": 76337, "tid": -914061504, "ts": 1716454226897855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454226958390, "dur": 784,
    "args": {
      "External id": 288248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288248,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 1.6,
      "warps per SM": 25.6,
      "grid": [128, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 40
    }
  },
  {
    "ph": "f", "id": 288248, "pid": 5, "tid": 7, "ts": 1716454226958390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897887, "dur": 7,
    "args": {
      "External id": 288248,
      "cbid": 211, "correlation": 288248
    }
  },
  {
    "ph": "s", "id": 288248, "pid": 76337, "tid": -914061504, "ts": 1716454226897887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454226959176, "dur": 4,
    "args": {
      "External id": 288260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288260,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.025,
      "warps per SM": 0.2,
      "grid": [2, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 288260, "pid": 5, "tid": 7, "ts": 1716454226959176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897903, "dur": 7,
    "args": {
      "External id": 288260,
      "cbid": 211, "correlation": 288260
    }
  },
  {
    "ph": "s", "id": 288260, "pid": 76337, "tid": -914061504, "ts": 1716454226897903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226959181, "dur": 651,
    "args": {
      "External id": 288263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288263, "pid": 5, "tid": 7, "ts": 1716454226959181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897920, "dur": 6,
    "args": {
      "External id": 288263,
      "cbid": 211, "correlation": 288263
    }
  },
  {
    "ph": "s", "id": 288263, "pid": 76337, "tid": -914061504, "ts": 1716454226897920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226959833, "dur": 418,
    "args": {
      "External id": 288272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288272,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 2457.6,
      "warps per SM": 9830.4,
      "grid": [196608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288272, "pid": 5, "tid": 7, "ts": 1716454226959833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226897967, "dur": 18,
    "args": {
      "External id": 288272,
      "cbid": 211, "correlation": 288272
    }
  },
  {
    "ph": "s", "id": 288272, "pid": 76337, "tid": -914061504, "ts": 1716454226897967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454226898030, "dur": 0,
    "args": {
      "External id": 288282,
      "cbid": 317, "correlation": 288282
    }
  },
  {
    "ph": "f", "id": 288282, "pid": 76337, "tid": -914061504, "ts": 1716454226898030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454226898031, "dur": 0,
    "args": {
      "External id": 288283,
      "cbid": 203, "correlation": 288283
    }
  },
  {
    "ph": "f", "id": 288283, "pid": 76337, "tid": -914061504, "ts": 1716454226898031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454226898032, "dur": 0,
    "args": {
      "External id": 288284,
      "cbid": 205, "correlation": 288284
    }
  },
  {
    "ph": "f", "id": 288284, "pid": 76337, "tid": -914061504, "ts": 1716454226898032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226960253, "dur": 541,
    "args": {
      "External id": 288288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1228.8,
      "warps per SM": 9830.4,
      "grid": [6144, 4, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288288, "pid": 5, "tid": 7, "ts": 1716454226960253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900409, "dur": 24,
    "args": {
      "External id": 288288,
      "cbid": 211, "correlation": 288288
    }
  },
  {
    "ph": "s", "id": 288288, "pid": 76337, "tid": -914061504, "ts": 1716454226900409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454226960795, "dur": 3,
    "args": {
      "External id": 288290,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 288290,
      "bytes": 18432, "memory bandwidth (GB/s)": 5.052631578947368
    }
  },
  {
    "ph": "f", "id": 288290, "pid": 5, "tid": 7, "ts": 1716454226960795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900437, "dur": 25,
    "args": {
      "External id": 288290,
      "cbid": 51, "correlation": 288290
    }
  },
  {
    "ph": "s", "id": 288290, "pid": 76337, "tid": -914061504, "ts": 1716454226900437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900468, "dur": 2,
    "args": {
      "External id": 288292,
      "cbid": 200, "correlation": 288292
    }
  },
  {
    "ph": "f", "id": 288292, "pid": 76337, "tid": -914061504, "ts": 1716454226900468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900470, "dur": 0,
    "args": {
      "External id": 288293,
      "cbid": 200, "correlation": 288293
    }
  },
  {
    "ph": "f", "id": 288293, "pid": 76337, "tid": -914061504, "ts": 1716454226900470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900471, "dur": 0,
    "args": {
      "External id": 288294,
      "cbid": 200, "correlation": 288294
    }
  },
  {
    "ph": "f", "id": 288294, "pid": 76337, "tid": -914061504, "ts": 1716454226900471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900471, "dur": 0,
    "args": {
      "External id": 288295,
      "cbid": 200, "correlation": 288295
    }
  },
  {
    "ph": "f", "id": 288295, "pid": 76337, "tid": -914061504, "ts": 1716454226900471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900472, "dur": 4,
    "args": {
      "External id": 288296,
      "cbid": 15, "correlation": 288296
    }
  },
  {
    "ph": "f", "id": 288296, "pid": 76337, "tid": -914061504, "ts": 1716454226900472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900477, "dur": 3,
    "args": {
      "External id": 288297,
      "cbid": 251, "correlation": 288297
    }
  },
  {
    "ph": "f", "id": 288297, "pid": 76337, "tid": -914061504, "ts": 1716454226900477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454226960800, "dur": 21,
    "args": {
      "External id": 288298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288298,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288298, "pid": 5, "tid": 7, "ts": 1716454226960800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900482, "dur": 10,
    "args": {
      "External id": 288298,
      "cbid": 211, "correlation": 288298
    }
  },
  {
    "ph": "s", "id": 288298, "pid": 76337, "tid": -914061504, "ts": 1716454226900482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x32x32_stage1_warpsize4x1x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454226960833, "dur": 1072,
    "args": {
      "External id": 288300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288300,
      "registers per thread": 128,
      "shared memory": 10240,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [1, 6144, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 25
    }
  },
  {
    "ph": "f", "id": 288300, "pid": 5, "tid": 7, "ts": 1716454226960833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900509, "dur": 190,
    "args": {
      "External id": 288300,
      "cbid": 211, "correlation": 288300
    }
  },
  {
    "ph": "s", "id": 288300, "pid": 76337, "tid": -914061504, "ts": 1716454226900509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454226961907, "dur": 74,
    "args": {
      "External id": 288302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 307.2,
      "warps per SM": 2457.6,
      "grid": [6144, 1, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288302, "pid": 5, "tid": 7, "ts": 1716454226961907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226900707, "dur": 9,
    "args": {
      "External id": 288302,
      "cbid": 211, "correlation": 288302
    }
  },
  {
    "ph": "s", "id": 288302, "pid": 76337, "tid": -914061504, "ts": 1716454226900707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454226961982, "dur": 20,
    "args": {
      "External id": 288308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288308,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 57.6,
      "warps per SM": 230.4,
      "grid": [4608, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288308, "pid": 5, "tid": 7, "ts": 1716454226961982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901387, "dur": 13,
    "args": {
      "External id": 288308,
      "cbid": 211, "correlation": 288308
    }
  },
  {
    "ph": "s", "id": 288308, "pid": 76337, "tid": -914061504, "ts": 1716454226901387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962004, "dur": 5,
    "args": {
      "External id": 288318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288318,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288318, "pid": 5, "tid": 7, "ts": 1716454226962004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901689, "dur": 16,
    "args": {
      "External id": 288318,
      "cbid": 211, "correlation": 288318
    }
  },
  {
    "ph": "s", "id": 288318, "pid": 76337, "tid": -914061504, "ts": 1716454226901689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962011, "dur": 5,
    "args": {
      "External id": 288326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288326,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288326, "pid": 5, "tid": 7, "ts": 1716454226962011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901731, "dur": 9,
    "args": {
      "External id": 288326,
      "cbid": 211, "correlation": 288326
    }
  },
  {
    "ph": "s", "id": 288326, "pid": 76337, "tid": -914061504, "ts": 1716454226901731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962017, "dur": 6,
    "args": {
      "External id": 288334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288334,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288334, "pid": 5, "tid": 7, "ts": 1716454226962017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901825, "dur": 16,
    "args": {
      "External id": 288334,
      "cbid": 211, "correlation": 288334
    }
  },
  {
    "ph": "s", "id": 288334, "pid": 76337, "tid": -914061504, "ts": 1716454226901825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962024, "dur": 5,
    "args": {
      "External id": 288344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288344,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288344, "pid": 5, "tid": 7, "ts": 1716454226962024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901877, "dur": 10,
    "args": {
      "External id": 288344,
      "cbid": 211, "correlation": 288344
    }
  },
  {
    "ph": "s", "id": 288344, "pid": 76337, "tid": -914061504, "ts": 1716454226901877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962030, "dur": 4,
    "args": {
      "External id": 288352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288352,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288352, "pid": 5, "tid": 7, "ts": 1716454226962030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901905, "dur": 8,
    "args": {
      "External id": 288352,
      "cbid": 211, "correlation": 288352
    }
  },
  {
    "ph": "s", "id": 288352, "pid": 76337, "tid": -914061504, "ts": 1716454226901905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962036, "dur": 4,
    "args": {
      "External id": 288360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288360,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288360, "pid": 5, "tid": 7, "ts": 1716454226962036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901930, "dur": 7,
    "args": {
      "External id": 288360,
      "cbid": 211, "correlation": 288360
    }
  },
  {
    "ph": "s", "id": 288360, "pid": 76337, "tid": -914061504, "ts": 1716454226901930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962041, "dur": 5,
    "args": {
      "External id": 288370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288370,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288370, "pid": 5, "tid": 7, "ts": 1716454226962041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226901967, "dur": 18,
    "args": {
      "External id": 288370,
      "cbid": 211, "correlation": 288370
    }
  },
  {
    "ph": "s", "id": 288370, "pid": 76337, "tid": -914061504, "ts": 1716454226901967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962047, "dur": 4,
    "args": {
      "External id": 288378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288378,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288378, "pid": 5, "tid": 7, "ts": 1716454226962047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902006, "dur": 8,
    "args": {
      "External id": 288378,
      "cbid": 211, "correlation": 288378
    }
  },
  {
    "ph": "s", "id": 288378, "pid": 76337, "tid": -914061504, "ts": 1716454226902006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962053, "dur": 4,
    "args": {
      "External id": 288386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288386,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288386, "pid": 5, "tid": 7, "ts": 1716454226962053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902032, "dur": 8,
    "args": {
      "External id": 288386,
      "cbid": 211, "correlation": 288386
    }
  },
  {
    "ph": "s", "id": 288386, "pid": 76337, "tid": -914061504, "ts": 1716454226902032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962059, "dur": 5,
    "args": {
      "External id": 288396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288396,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288396, "pid": 5, "tid": 7, "ts": 1716454226962059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902073, "dur": 10,
    "args": {
      "External id": 288396,
      "cbid": 211, "correlation": 288396
    }
  },
  {
    "ph": "s", "id": 288396, "pid": 76337, "tid": -914061504, "ts": 1716454226902073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962065, "dur": 4,
    "args": {
      "External id": 288404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288404,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288404, "pid": 5, "tid": 7, "ts": 1716454226962065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902100, "dur": 9,
    "args": {
      "External id": 288404,
      "cbid": 211, "correlation": 288404
    }
  },
  {
    "ph": "s", "id": 288404, "pid": 76337, "tid": -914061504, "ts": 1716454226902100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::launch_clamp_scalar(at::TensorIteratorBase&, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::{lambda()#1}::operator()() const::{lambda()#8}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454226962070, "dur": 4,
    "args": {
      "External id": 288412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288412,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [1152, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 90
    }
  },
  {
    "ph": "f", "id": 288412, "pid": 5, "tid": 7, "ts": 1716454226962070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902129, "dur": 8,
    "args": {
      "External id": 288412,
      "cbid": 211, "correlation": 288412
    }
  },
  {
    "ph": "s", "id": 288412, "pid": 76337, "tid": -914061504, "ts": 1716454226902129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454226962076, "dur": 18,
    "args": {
      "External id": 288421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 288421,
      "registers per thread": 26,
      "shared memory": 0,
      "blocks per SM": 8,
      "warps per SM": 128,
      "grid": [160, 4, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 288421, "pid": 5, "tid": 7, "ts": 1716454226962076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902190, "dur": 17,
    "args": {
      "External id": 288421,
      "cbid": 211, "correlation": 288421
    }
  },
  {
    "ph": "s", "id": 288421, "pid": 76337, "tid": -914061504, "ts": 1716454226902190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454226962096, "dur": 852,
    "args": {
      "External id": 288426,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 288426,
      "bytes": 4718592, "memory bandwidth (GB/s)": 5.536434035762895
    }
  },
  {
    "ph": "f", "id": 288426, "pid": 5, "tid": 7, "ts": 1716454226962096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454226902270, "dur": 61007,
    "args": {
      "External id": 288426,
      "cbid": 41, "correlation": 288426
    }
  },
  {
    "ph": "s", "id": 288426, "pid": 76337, "tid": -914061504, "ts": 1716454226902270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454226963279, "dur": 12,
    "args": {
      "External id": 288427,
      "cbid": 131, "correlation": 288427
    }
  },
  {
    "ph": "f", "id": 288427, "pid": 76337, "tid": -914061504, "ts": 1716454226963279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454227004833, "dur": 81,
    "args": {
      "External id": 288431,
      "cbid": 165, "correlation": 288431
    }
  },
  {
    "ph": "f", "id": 288431, "pid": 76337, "tid": -914061504, "ts": 1716454227004833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222500365, "dur": 1,
    "args": {
      "External id": 72204,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72204,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00423728813559322
    }
  },
  {
    "ph": "f", "id": 72204, "pid": 5, "tid": 7, "ts": 1716454222500365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500334, "dur": 29,
    "args": {
      "External id": 72204,
      "cbid": 41, "correlation": 72204
    }
  },
  {
    "ph": "s", "id": 72204, "pid": 76337, "tid": -914061504, "ts": 1716454222500334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500365, "dur": 11,
    "args": {
      "External id": 72205,
      "cbid": 131, "correlation": 72205
    }
  },
  {
    "ph": "f", "id": 72205, "pid": 76337, "tid": -914061504, "ts": 1716454222500365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222500459, "dur": 3,
    "args": {
      "External id": 72212,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72212,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 72212, "pid": 5, "tid": 7, "ts": 1716454222500459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500428, "dur": 31,
    "args": {
      "External id": 72212,
      "cbid": 41, "correlation": 72212
    }
  },
  {
    "ph": "s", "id": 72212, "pid": 76337, "tid": -914061504, "ts": 1716454222500428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222500512, "dur": 1,
    "args": {
      "External id": 72223,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72223,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 72223, "pid": 5, "tid": 7, "ts": 1716454222500512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500498, "dur": 11,
    "args": {
      "External id": 72223,
      "cbid": 41, "correlation": 72223
    }
  },
  {
    "ph": "s", "id": 72223, "pid": 76337, "tid": -914061504, "ts": 1716454222500498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500510, "dur": 9,
    "args": {
      "External id": 72224,
      "cbid": 131, "correlation": 72224
    }
  },
  {
    "ph": "f", "id": 72224, "pid": 76337, "tid": -914061504, "ts": 1716454222500510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222500575, "dur": 3,
    "args": {
      "External id": 72232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72232,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72232, "pid": 5, "tid": 7, "ts": 1716454222500575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500557, "dur": 20,
    "args": {
      "External id": 72232,
      "cbid": 211, "correlation": 72232
    }
  },
  {
    "ph": "s", "id": 72232, "pid": 76337, "tid": -914061504, "ts": 1716454222500557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222500620, "dur": 3,
    "args": {
      "External id": 72242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72242,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72242, "pid": 5, "tid": 7, "ts": 1716454222500620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500607, "dur": 13,
    "args": {
      "External id": 72242,
      "cbid": 211, "correlation": 72242
    }
  },
  {
    "ph": "s", "id": 72242, "pid": 76337, "tid": -914061504, "ts": 1716454222500607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222500663, "dur": 3,
    "args": {
      "External id": 72251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72251,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72251, "pid": 5, "tid": 7, "ts": 1716454222500663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500648, "dur": 14,
    "args": {
      "External id": 72251,
      "cbid": 211, "correlation": 72251
    }
  },
  {
    "ph": "s", "id": 72251, "pid": 76337, "tid": -914061504, "ts": 1716454222500648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222500783, "dur": 5,
    "args": {
      "External id": 72259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72259,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72259, "pid": 5, "tid": 7, "ts": 1716454222500783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500767, "dur": 16,
    "args": {
      "External id": 72259,
      "cbid": 211, "correlation": 72259
    }
  },
  {
    "ph": "s", "id": 72259, "pid": 76337, "tid": -914061504, "ts": 1716454222500767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222500830, "dur": 3,
    "args": {
      "External id": 72268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72268,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72268, "pid": 5, "tid": 7, "ts": 1716454222500830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500819, "dur": 9,
    "args": {
      "External id": 72268,
      "cbid": 211, "correlation": 72268
    }
  },
  {
    "ph": "s", "id": 72268, "pid": 76337, "tid": -914061504, "ts": 1716454222500819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222500854, "dur": 3,
    "args": {
      "External id": 72277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72277,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72277, "pid": 5, "tid": 7, "ts": 1716454222500854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500845, "dur": 8,
    "args": {
      "External id": 72277,
      "cbid": 211, "correlation": 72277
    }
  },
  {
    "ph": "s", "id": 72277, "pid": 76337, "tid": -914061504, "ts": 1716454222500845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222500925, "dur": 3,
    "args": {
      "External id": 72285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72285,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72285, "pid": 5, "tid": 7, "ts": 1716454222500925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500914, "dur": 9,
    "args": {
      "External id": 72285,
      "cbid": 211, "correlation": 72285
    }
  },
  {
    "ph": "s", "id": 72285, "pid": 76337, "tid": -914061504, "ts": 1716454222500914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222501011, "dur": 1,
    "args": {
      "External id": 72293,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72293,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 72293, "pid": 5, "tid": 7, "ts": 1716454222501011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222500988, "dur": 32,
    "args": {
      "External id": 72293,
      "cbid": 41, "correlation": 72293
    }
  },
  {
    "ph": "s", "id": 72293, "pid": 76337, "tid": -914061504, "ts": 1716454222500988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501021, "dur": 4,
    "args": {
      "External id": 72294,
      "cbid": 131, "correlation": 72294
    }
  },
  {
    "ph": "f", "id": 72294, "pid": 76337, "tid": -914061504, "ts": 1716454222501021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222501089, "dur": 1,
    "args": {
      "External id": 72304,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72304,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 72304, "pid": 5, "tid": 7, "ts": 1716454222501089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501075, "dur": 11,
    "args": {
      "External id": 72304,
      "cbid": 41, "correlation": 72304
    }
  },
  {
    "ph": "s", "id": 72304, "pid": 76337, "tid": -914061504, "ts": 1716454222501075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501087, "dur": 8,
    "args": {
      "External id": 72305,
      "cbid": 131, "correlation": 72305
    }
  },
  {
    "ph": "f", "id": 72305, "pid": 76337, "tid": -914061504, "ts": 1716454222501087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222501148, "dur": 1,
    "args": {
      "External id": 72314,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72314,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 72314, "pid": 5, "tid": 7, "ts": 1716454222501148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501136, "dur": 10,
    "args": {
      "External id": 72314,
      "cbid": 41, "correlation": 72314
    }
  },
  {
    "ph": "s", "id": 72314, "pid": 76337, "tid": -914061504, "ts": 1716454222501136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501146, "dur": 8,
    "args": {
      "External id": 72315,
      "cbid": 131, "correlation": 72315
    }
  },
  {
    "ph": "f", "id": 72315, "pid": 76337, "tid": -914061504, "ts": 1716454222501146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222501222, "dur": 4,
    "args": {
      "External id": 72322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72322,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72322, "pid": 5, "tid": 7, "ts": 1716454222501222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501205, "dur": 17,
    "args": {
      "External id": 72322,
      "cbid": 211, "correlation": 72322
    }
  },
  {
    "ph": "s", "id": 72322, "pid": 76337, "tid": -914061504, "ts": 1716454222501205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454222501259, "dur": 4,
    "args": {
      "External id": 72342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72342,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72342, "pid": 5, "tid": 7, "ts": 1716454222501259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501248, "dur": 12,
    "args": {
      "External id": 72342,
      "cbid": 211, "correlation": 72342
    }
  },
  {
    "ph": "s", "id": 72342, "pid": 76337, "tid": -914061504, "ts": 1716454222501248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501261, "dur": 0,
    "args": {
      "External id": 72343,
      "cbid": 11, "correlation": 72343
    }
  },
  {
    "ph": "f", "id": 72343, "pid": 76337, "tid": -914061504, "ts": 1716454222501261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501261, "dur": 0,
    "args": {
      "External id": 72344,
      "cbid": 11, "correlation": 72344
    }
  },
  {
    "ph": "f", "id": 72344, "pid": 76337, "tid": -914061504, "ts": 1716454222501261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222501277, "dur": 1,
    "args": {
      "External id": 72347,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72347,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 72347, "pid": 5, "tid": 7, "ts": 1716454222501277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501263, "dur": 24,
    "args": {
      "External id": 72347,
      "cbid": 41, "correlation": 72347
    }
  },
  {
    "ph": "s", "id": 72347, "pid": 76337, "tid": -914061504, "ts": 1716454222501263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501287, "dur": 3,
    "args": {
      "External id": 72348,
      "cbid": 131, "correlation": 72348
    }
  },
  {
    "ph": "f", "id": 72348, "pid": 76337, "tid": -914061504, "ts": 1716454222501287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222501318, "dur": 3,
    "args": {
      "External id": 72372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72372,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72372, "pid": 5, "tid": 7, "ts": 1716454222501318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501308, "dur": 9,
    "args": {
      "External id": 72372,
      "cbid": 211, "correlation": 72372
    }
  },
  {
    "ph": "s", "id": 72372, "pid": 76337, "tid": -914061504, "ts": 1716454222501308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501318, "dur": 0,
    "args": {
      "External id": 72373,
      "cbid": 11, "correlation": 72373
    }
  },
  {
    "ph": "f", "id": 72373, "pid": 76337, "tid": -914061504, "ts": 1716454222501318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501318, "dur": 0,
    "args": {
      "External id": 72374,
      "cbid": 11, "correlation": 72374
    }
  },
  {
    "ph": "f", "id": 72374, "pid": 76337, "tid": -914061504, "ts": 1716454222501318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501320, "dur": 1,
    "args": {
      "External id": 72376,
      "cbid": 200, "correlation": 72376
    }
  },
  {
    "ph": "f", "id": 72376, "pid": 76337, "tid": -914061504, "ts": 1716454222501320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222501331, "dur": 4,
    "args": {
      "External id": 72378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72378,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72378, "pid": 5, "tid": 7, "ts": 1716454222501331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501323, "dur": 9,
    "args": {
      "External id": 72378,
      "cbid": 211, "correlation": 72378
    }
  },
  {
    "ph": "s", "id": 72378, "pid": 76337, "tid": -914061504, "ts": 1716454222501323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501333, "dur": 0,
    "args": {
      "External id": 72379,
      "cbid": 11, "correlation": 72379
    }
  },
  {
    "ph": "f", "id": 72379, "pid": 76337, "tid": -914061504, "ts": 1716454222501333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501334, "dur": 0,
    "args": {
      "External id": 72380,
      "cbid": 11, "correlation": 72380
    }
  },
  {
    "ph": "f", "id": 72380, "pid": 76337, "tid": -914061504, "ts": 1716454222501334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222501374, "dur": 1,
    "args": {
      "External id": 72387,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72387,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0047169811320754715
    }
  },
  {
    "ph": "f", "id": 72387, "pid": 5, "tid": 7, "ts": 1716454222501374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501361, "dur": 21,
    "args": {
      "External id": 72387,
      "cbid": 41, "correlation": 72387
    }
  },
  {
    "ph": "s", "id": 72387, "pid": 76337, "tid": -914061504, "ts": 1716454222501361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501383, "dur": 3,
    "args": {
      "External id": 72388,
      "cbid": 131, "correlation": 72388
    }
  },
  {
    "ph": "f", "id": 72388, "pid": 76337, "tid": -914061504, "ts": 1716454222501383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222501434, "dur": 1,
    "args": {
      "External id": 72398,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 72398,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 72398, "pid": 5, "tid": 7, "ts": 1716454222501434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501421, "dur": 11,
    "args": {
      "External id": 72398,
      "cbid": 41, "correlation": 72398
    }
  },
  {
    "ph": "s", "id": 72398, "pid": 76337, "tid": -914061504, "ts": 1716454222501421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501433, "dur": 8,
    "args": {
      "External id": 72399,
      "cbid": 131, "correlation": 72399
    }
  },
  {
    "ph": "f", "id": 72399, "pid": 76337, "tid": -914061504, "ts": 1716454222501433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222501508, "dur": 5,
    "args": {
      "External id": 72406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72406,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72406, "pid": 5, "tid": 7, "ts": 1716454222501508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501492, "dur": 16,
    "args": {
      "External id": 72406,
      "cbid": 211, "correlation": 72406
    }
  },
  {
    "ph": "s", "id": 72406, "pid": 76337, "tid": -914061504, "ts": 1716454222501492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222501581, "dur": 3,
    "args": {
      "External id": 72415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72415,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72415, "pid": 5, "tid": 7, "ts": 1716454222501581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501568, "dur": 13,
    "args": {
      "External id": 72415,
      "cbid": 211, "correlation": 72415
    }
  },
  {
    "ph": "s", "id": 72415, "pid": 76337, "tid": -914061504, "ts": 1716454222501568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222501620, "dur": 3,
    "args": {
      "External id": 72423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72423,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72423, "pid": 5, "tid": 7, "ts": 1716454222501620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501609, "dur": 11,
    "args": {
      "External id": 72423,
      "cbid": 211, "correlation": 72423
    }
  },
  {
    "ph": "s", "id": 72423, "pid": 76337, "tid": -914061504, "ts": 1716454222501609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222501655, "dur": 4,
    "args": {
      "External id": 72431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72431,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72431, "pid": 5, "tid": 7, "ts": 1716454222501655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501642, "dur": 12,
    "args": {
      "External id": 72431,
      "cbid": 211, "correlation": 72431
    }
  },
  {
    "ph": "s", "id": 72431, "pid": 76337, "tid": -914061504, "ts": 1716454222501642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222501687, "dur": 4,
    "args": {
      "External id": 72439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72439,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72439, "pid": 5, "tid": 7, "ts": 1716454222501687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501676, "dur": 10,
    "args": {
      "External id": 72439,
      "cbid": 211, "correlation": 72439
    }
  },
  {
    "ph": "s", "id": 72439, "pid": 76337, "tid": -914061504, "ts": 1716454222501676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222501714, "dur": 3,
    "args": {
      "External id": 72447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72447,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72447, "pid": 5, "tid": 7, "ts": 1716454222501714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501704, "dur": 8,
    "args": {
      "External id": 72447,
      "cbid": 211, "correlation": 72447
    }
  },
  {
    "ph": "s", "id": 72447, "pid": 76337, "tid": -914061504, "ts": 1716454222501704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222501740, "dur": 3,
    "args": {
      "External id": 72455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72455,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72455, "pid": 5, "tid": 7, "ts": 1716454222501740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501730, "dur": 8,
    "args": {
      "External id": 72455,
      "cbid": 211, "correlation": 72455
    }
  },
  {
    "ph": "s", "id": 72455, "pid": 76337, "tid": -914061504, "ts": 1716454222501730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222501762, "dur": 4,
    "args": {
      "External id": 72463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72463, "pid": 5, "tid": 7, "ts": 1716454222501762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501753, "dur": 7,
    "args": {
      "External id": 72463,
      "cbid": 211, "correlation": 72463
    }
  },
  {
    "ph": "s", "id": 72463, "pid": 76337, "tid": -914061504, "ts": 1716454222501753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222501780, "dur": 4,
    "args": {
      "External id": 72471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72471,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72471, "pid": 5, "tid": 7, "ts": 1716454222501780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501771, "dur": 7,
    "args": {
      "External id": 72471,
      "cbid": 211, "correlation": 72471
    }
  },
  {
    "ph": "s", "id": 72471, "pid": 76337, "tid": -914061504, "ts": 1716454222501771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222501800, "dur": 3,
    "args": {
      "External id": 72479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72479,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72479, "pid": 5, "tid": 7, "ts": 1716454222501800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501791, "dur": 7,
    "args": {
      "External id": 72479,
      "cbid": 211, "correlation": 72479
    }
  },
  {
    "ph": "s", "id": 72479, "pid": 76337, "tid": -914061504, "ts": 1716454222501791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222501860, "dur": 3,
    "args": {
      "External id": 72487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72487,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72487, "pid": 5, "tid": 7, "ts": 1716454222501860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501847, "dur": 12,
    "args": {
      "External id": 72487,
      "cbid": 211, "correlation": 72487
    }
  },
  {
    "ph": "s", "id": 72487, "pid": 76337, "tid": -914061504, "ts": 1716454222501847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222501885, "dur": 4,
    "args": {
      "External id": 72495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72495,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72495, "pid": 5, "tid": 7, "ts": 1716454222501885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501876, "dur": 7,
    "args": {
      "External id": 72495,
      "cbid": 211, "correlation": 72495
    }
  },
  {
    "ph": "s", "id": 72495, "pid": 76337, "tid": -914061504, "ts": 1716454222501876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222501908, "dur": 4,
    "args": {
      "External id": 72503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72503,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72503, "pid": 5, "tid": 7, "ts": 1716454222501908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501898, "dur": 8,
    "args": {
      "External id": 72503,
      "cbid": 211, "correlation": 72503
    }
  },
  {
    "ph": "s", "id": 72503, "pid": 76337, "tid": -914061504, "ts": 1716454222501898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222501926, "dur": 3,
    "args": {
      "External id": 72511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72511,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 72511, "pid": 5, "tid": 7, "ts": 1716454222501926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222501918, "dur": 7,
    "args": {
      "External id": 72511,
      "cbid": 211, "correlation": 72511
    }
  },
  {
    "ph": "s", "id": 72511, "pid": 76337, "tid": -914061504, "ts": 1716454222501918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222502324, "dur": 5,
    "args": {
      "External id": 72520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72520,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72520, "pid": 5, "tid": 7, "ts": 1716454222502324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502306, "dur": 18,
    "args": {
      "External id": 72520,
      "cbid": 211, "correlation": 72520
    }
  },
  {
    "ph": "s", "id": 72520, "pid": 76337, "tid": -914061504, "ts": 1716454222502306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222502363, "dur": 5,
    "args": {
      "External id": 72529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72529,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72529, "pid": 5, "tid": 7, "ts": 1716454222502363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502352, "dur": 9,
    "args": {
      "External id": 72529,
      "cbid": 211, "correlation": 72529
    }
  },
  {
    "ph": "s", "id": 72529, "pid": 76337, "tid": -914061504, "ts": 1716454222502352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222502501, "dur": 3,
    "args": {
      "External id": 72545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72545, "pid": 5, "tid": 7, "ts": 1716454222502501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502486, "dur": 15,
    "args": {
      "External id": 72545,
      "cbid": 211, "correlation": 72545
    }
  },
  {
    "ph": "s", "id": 72545, "pid": 76337, "tid": -914061504, "ts": 1716454222502486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222502537, "dur": 3,
    "args": {
      "External id": 72553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72553, "pid": 5, "tid": 7, "ts": 1716454222502537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502528, "dur": 9,
    "args": {
      "External id": 72553,
      "cbid": 211, "correlation": 72553
    }
  },
  {
    "ph": "s", "id": 72553, "pid": 76337, "tid": -914061504, "ts": 1716454222502528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222502571, "dur": 3,
    "args": {
      "External id": 72561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72561,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72561, "pid": 5, "tid": 7, "ts": 1716454222502571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502560, "dur": 10,
    "args": {
      "External id": 72561,
      "cbid": 211, "correlation": 72561
    }
  },
  {
    "ph": "s", "id": 72561, "pid": 76337, "tid": -914061504, "ts": 1716454222502560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222502604, "dur": 4,
    "args": {
      "External id": 72569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72569,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72569, "pid": 5, "tid": 7, "ts": 1716454222502604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502594, "dur": 9,
    "args": {
      "External id": 72569,
      "cbid": 211, "correlation": 72569
    }
  },
  {
    "ph": "s", "id": 72569, "pid": 76337, "tid": -914061504, "ts": 1716454222502594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222502662, "dur": 4,
    "args": {
      "External id": 72581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72581,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72581, "pid": 5, "tid": 7, "ts": 1716454222502662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502649, "dur": 13,
    "args": {
      "External id": 72581,
      "cbid": 211, "correlation": 72581
    }
  },
  {
    "ph": "s", "id": 72581, "pid": 76337, "tid": -914061504, "ts": 1716454222502649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222502710, "dur": 4,
    "args": {
      "External id": 72592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72592, "pid": 5, "tid": 7, "ts": 1716454222502710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502698, "dur": 12,
    "args": {
      "External id": 72592,
      "cbid": 211, "correlation": 72592
    }
  },
  {
    "ph": "s", "id": 72592, "pid": 76337, "tid": -914061504, "ts": 1716454222502698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222502741, "dur": 3,
    "args": {
      "External id": 72600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72600, "pid": 5, "tid": 7, "ts": 1716454222502741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502732, "dur": 8,
    "args": {
      "External id": 72600,
      "cbid": 211, "correlation": 72600
    }
  },
  {
    "ph": "s", "id": 72600, "pid": 76337, "tid": -914061504, "ts": 1716454222502732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222502773, "dur": 5,
    "args": {
      "External id": 72608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72608,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72608, "pid": 5, "tid": 7, "ts": 1716454222502773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502763, "dur": 10,
    "args": {
      "External id": 72608,
      "cbid": 211, "correlation": 72608
    }
  },
  {
    "ph": "s", "id": 72608, "pid": 76337, "tid": -914061504, "ts": 1716454222502763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222502802, "dur": 5,
    "args": {
      "External id": 72616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72616,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72616, "pid": 5, "tid": 7, "ts": 1716454222502802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502792, "dur": 9,
    "args": {
      "External id": 72616,
      "cbid": 211, "correlation": 72616
    }
  },
  {
    "ph": "s", "id": 72616, "pid": 76337, "tid": -914061504, "ts": 1716454222502792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222502833, "dur": 4,
    "args": {
      "External id": 72625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72625,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72625, "pid": 5, "tid": 7, "ts": 1716454222502833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502822, "dur": 11,
    "args": {
      "External id": 72625,
      "cbid": 211, "correlation": 72625
    }
  },
  {
    "ph": "s", "id": 72625, "pid": 76337, "tid": -914061504, "ts": 1716454222502822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222502894, "dur": 4,
    "args": {
      "External id": 72638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72638,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72638, "pid": 5, "tid": 7, "ts": 1716454222502894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502881, "dur": 13,
    "args": {
      "External id": 72638,
      "cbid": 211, "correlation": 72638
    }
  },
  {
    "ph": "s", "id": 72638, "pid": 76337, "tid": -914061504, "ts": 1716454222502881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222502937, "dur": 5,
    "args": {
      "External id": 72648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72648,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 72648, "pid": 5, "tid": 7, "ts": 1716454222502937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222502924, "dur": 12,
    "args": {
      "External id": 72648,
      "cbid": 211, "correlation": 72648
    }
  },
  {
    "ph": "s", "id": 72648, "pid": 76337, "tid": -914061504, "ts": 1716454222502924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503079, "dur": 5,
    "args": {
      "External id": 72665,
      "cbid": 251, "correlation": 72665
    }
  },
  {
    "ph": "f", "id": 72665, "pid": 76337, "tid": -914061504, "ts": 1716454222503079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454222503110, "dur": 12,
    "args": {
      "External id": 72667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72667,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 72667, "pid": 5, "tid": 7, "ts": 1716454222503110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503094, "dur": 17,
    "args": {
      "External id": 72667,
      "cbid": 211, "correlation": 72667
    }
  },
  {
    "ph": "s", "id": 72667, "pid": 76337, "tid": -914061504, "ts": 1716454222503094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222503174, "dur": 4,
    "args": {
      "External id": 72675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72675,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72675, "pid": 5, "tid": 7, "ts": 1716454222503174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503161, "dur": 13,
    "args": {
      "External id": 72675,
      "cbid": 211, "correlation": 72675
    }
  },
  {
    "ph": "s", "id": 72675, "pid": 76337, "tid": -914061504, "ts": 1716454222503161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503232, "dur": 2,
    "args": {
      "External id": 72691,
      "cbid": 251, "correlation": 72691
    }
  },
  {
    "ph": "f", "id": 72691, "pid": 76337, "tid": -914061504, "ts": 1716454222503232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503238, "dur": 0,
    "args": {
      "External id": 72693,
      "cbid": 251, "correlation": 72693
    }
  },
  {
    "ph": "f", "id": 72693, "pid": 76337, "tid": -914061504, "ts": 1716454222503238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222503255, "dur": 13,
    "args": {
      "External id": 72694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72694,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 72694, "pid": 5, "tid": 7, "ts": 1716454222503255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503241, "dur": 14,
    "args": {
      "External id": 72694,
      "cbid": 211, "correlation": 72694
    }
  },
  {
    "ph": "s", "id": 72694, "pid": 76337, "tid": -914061504, "ts": 1716454222503241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222503270, "dur": 5,
    "args": {
      "External id": 72696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72696,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 72696, "pid": 5, "tid": 7, "ts": 1716454222503270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503259, "dur": 9,
    "args": {
      "External id": 72696,
      "cbid": 211, "correlation": 72696
    }
  },
  {
    "ph": "s", "id": 72696, "pid": 76337, "tid": -914061504, "ts": 1716454222503259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503419, "dur": 4,
    "args": {
      "External id": 72706,
      "cbid": 317, "correlation": 72706
    }
  },
  {
    "ph": "f", "id": 72706, "pid": 76337, "tid": -914061504, "ts": 1716454222503419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503423, "dur": 1,
    "args": {
      "External id": 72707,
      "cbid": 203, "correlation": 72707
    }
  },
  {
    "ph": "f", "id": 72707, "pid": 76337, "tid": -914061504, "ts": 1716454222503423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503425, "dur": 1,
    "args": {
      "External id": 72708,
      "cbid": 205, "correlation": 72708
    }
  },
  {
    "ph": "f", "id": 72708, "pid": 76337, "tid": -914061504, "ts": 1716454222503425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222503494, "dur": 7,
    "args": {
      "External id": 72712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72712,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72712, "pid": 5, "tid": 7, "ts": 1716454222503494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503478, "dur": 15,
    "args": {
      "External id": 72712,
      "cbid": 211, "correlation": 72712
    }
  },
  {
    "ph": "s", "id": 72712, "pid": 76337, "tid": -914061504, "ts": 1716454222503478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222503504, "dur": 4,
    "args": {
      "External id": 72714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72714,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 72714, "pid": 5, "tid": 7, "ts": 1716454222503504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503497, "dur": 6,
    "args": {
      "External id": 72714,
      "cbid": 211, "correlation": 72714
    }
  },
  {
    "ph": "s", "id": 72714, "pid": 76337, "tid": -914061504, "ts": 1716454222503497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222503547, "dur": 4,
    "args": {
      "External id": 72716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72716, "pid": 5, "tid": 7, "ts": 1716454222503547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503536, "dur": 9,
    "args": {
      "External id": 72716,
      "cbid": 211, "correlation": 72716
    }
  },
  {
    "ph": "s", "id": 72716, "pid": 76337, "tid": -914061504, "ts": 1716454222503536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503557, "dur": 1,
    "args": {
      "External id": 72717,
      "cbid": 51, "correlation": 72717
    }
  },
  {
    "ph": "s", "id": 72717, "pid": 76337, "tid": -914061504, "ts": 1716454222503557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222503568, "dur": 87,
    "args": {
      "External id": 72718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72718,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 72718, "pid": 5, "tid": 7, "ts": 1716454222503568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503559, "dur": 8,
    "args": {
      "External id": 72718,
      "cbid": 211, "correlation": 72718
    }
  },
  {
    "ph": "s", "id": 72718, "pid": 76337, "tid": -914061504, "ts": 1716454222503559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222503656, "dur": 60,
    "args": {
      "External id": 72723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72723,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72723, "pid": 5, "tid": 7, "ts": 1716454222503656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222503598, "dur": 11,
    "args": {
      "External id": 72723,
      "cbid": 211, "correlation": 72723
    }
  },
  {
    "ph": "s", "id": 72723, "pid": 76337, "tid": -914061504, "ts": 1716454222503598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222505498, "dur": 52,
    "args": {
      "External id": 72743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72743,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 72743, "pid": 5, "tid": 7, "ts": 1716454222505498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505481, "dur": 17,
    "args": {
      "External id": 72743,
      "cbid": 211, "correlation": 72743
    }
  },
  {
    "ph": "s", "id": 72743, "pid": 76337, "tid": -914061504, "ts": 1716454222505481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222505551, "dur": 4,
    "args": {
      "External id": 72755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72755,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72755, "pid": 5, "tid": 7, "ts": 1716454222505551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505511, "dur": 8,
    "args": {
      "External id": 72755,
      "cbid": 211, "correlation": 72755
    }
  },
  {
    "ph": "s", "id": 72755, "pid": 76337, "tid": -914061504, "ts": 1716454222505511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222505557, "dur": 58,
    "args": {
      "External id": 72758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72758,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72758, "pid": 5, "tid": 7, "ts": 1716454222505557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505536, "dur": 8,
    "args": {
      "External id": 72758,
      "cbid": 211, "correlation": 72758
    }
  },
  {
    "ph": "s", "id": 72758, "pid": 76337, "tid": -914061504, "ts": 1716454222505536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222505616, "dur": 37,
    "args": {
      "External id": 72767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72767,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72767, "pid": 5, "tid": 7, "ts": 1716454222505616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505582, "dur": 10,
    "args": {
      "External id": 72767,
      "cbid": 211, "correlation": 72767
    }
  },
  {
    "ph": "s", "id": 72767, "pid": 76337, "tid": -914061504, "ts": 1716454222505582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505642, "dur": 0,
    "args": {
      "External id": 72777,
      "cbid": 317, "correlation": 72777
    }
  },
  {
    "ph": "f", "id": 72777, "pid": 76337, "tid": -914061504, "ts": 1716454222505642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505642, "dur": 0,
    "args": {
      "External id": 72778,
      "cbid": 203, "correlation": 72778
    }
  },
  {
    "ph": "f", "id": 72778, "pid": 76337, "tid": -914061504, "ts": 1716454222505642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505643, "dur": 0,
    "args": {
      "External id": 72779,
      "cbid": 205, "correlation": 72779
    }
  },
  {
    "ph": "f", "id": 72779, "pid": 76337, "tid": -914061504, "ts": 1716454222505643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222505675, "dur": 40,
    "args": {
      "External id": 72783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72783,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72783, "pid": 5, "tid": 7, "ts": 1716454222505675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505662, "dur": 12,
    "args": {
      "External id": 72783,
      "cbid": 211, "correlation": 72783
    }
  },
  {
    "ph": "s", "id": 72783, "pid": 76337, "tid": -914061504, "ts": 1716454222505662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222505717, "dur": 14,
    "args": {
      "External id": 72785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72785,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72785, "pid": 5, "tid": 7, "ts": 1716454222505717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505677, "dur": 6,
    "args": {
      "External id": 72785,
      "cbid": 211, "correlation": 72785
    }
  },
  {
    "ph": "s", "id": 72785, "pid": 76337, "tid": -914061504, "ts": 1716454222505677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222505733, "dur": 3,
    "args": {
      "External id": 72787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72787,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72787, "pid": 5, "tid": 7, "ts": 1716454222505733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505701, "dur": 8,
    "args": {
      "External id": 72787,
      "cbid": 211, "correlation": 72787
    }
  },
  {
    "ph": "s", "id": 72787, "pid": 76337, "tid": -914061504, "ts": 1716454222505701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505718, "dur": 0,
    "args": {
      "External id": 72788,
      "cbid": 51, "correlation": 72788
    }
  },
  {
    "ph": "s", "id": 72788, "pid": 76337, "tid": -914061504, "ts": 1716454222505718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222505738, "dur": 712,
    "args": {
      "External id": 72789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72789,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 72789, "pid": 5, "tid": 7, "ts": 1716454222505738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505719, "dur": 8,
    "args": {
      "External id": 72789,
      "cbid": 211, "correlation": 72789
    }
  },
  {
    "ph": "s", "id": 72789, "pid": 76337, "tid": -914061504, "ts": 1716454222505719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222506451, "dur": 59,
    "args": {
      "External id": 72794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72794,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72794, "pid": 5, "tid": 7, "ts": 1716454222506451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505751, "dur": 10,
    "args": {
      "External id": 72794,
      "cbid": 211, "correlation": 72794
    }
  },
  {
    "ph": "s", "id": 72794, "pid": 76337, "tid": -914061504, "ts": 1716454222505751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222506511, "dur": 3,
    "args": {
      "External id": 72802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72802,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72802, "pid": 5, "tid": 7, "ts": 1716454222506511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505797, "dur": 9,
    "args": {
      "External id": 72802,
      "cbid": 211, "correlation": 72802
    }
  },
  {
    "ph": "s", "id": 72802, "pid": 76337, "tid": -914061504, "ts": 1716454222505797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505867, "dur": 2,
    "args": {
      "External id": 72818,
      "cbid": 251, "correlation": 72818
    }
  },
  {
    "ph": "f", "id": 72818, "pid": 76337, "tid": -914061504, "ts": 1716454222505867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505872, "dur": 0,
    "args": {
      "External id": 72820,
      "cbid": 251, "correlation": 72820
    }
  },
  {
    "ph": "f", "id": 72820, "pid": 76337, "tid": -914061504, "ts": 1716454222505872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222506516, "dur": 9,
    "args": {
      "External id": 72821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72821,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 72821, "pid": 5, "tid": 7, "ts": 1716454222506516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505874, "dur": 12,
    "args": {
      "External id": 72821,
      "cbid": 211, "correlation": 72821
    }
  },
  {
    "ph": "s", "id": 72821, "pid": 76337, "tid": -914061504, "ts": 1716454222505874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222506526, "dur": 4,
    "args": {
      "External id": 72823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72823,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 72823, "pid": 5, "tid": 7, "ts": 1716454222506526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505888, "dur": 6,
    "args": {
      "External id": 72823,
      "cbid": 211, "correlation": 72823
    }
  },
  {
    "ph": "s", "id": 72823, "pid": 76337, "tid": -914061504, "ts": 1716454222505888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222506531, "dur": 56,
    "args": {
      "External id": 72833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72833, "pid": 5, "tid": 7, "ts": 1716454222506531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222505948, "dur": 12,
    "args": {
      "External id": 72833,
      "cbid": 211, "correlation": 72833
    }
  },
  {
    "ph": "s", "id": 72833, "pid": 76337, "tid": -914061504, "ts": 1716454222505948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222506589, "dur": 53,
    "args": {
      "External id": 72853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72853,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 72853, "pid": 5, "tid": 7, "ts": 1716454222506589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506027, "dur": 12,
    "args": {
      "External id": 72853,
      "cbid": 211, "correlation": 72853
    }
  },
  {
    "ph": "s", "id": 72853, "pid": 76337, "tid": -914061504, "ts": 1716454222506027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222506643, "dur": 4,
    "args": {
      "External id": 72865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72865,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72865, "pid": 5, "tid": 7, "ts": 1716454222506643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506050, "dur": 6,
    "args": {
      "External id": 72865,
      "cbid": 211, "correlation": 72865
    }
  },
  {
    "ph": "s", "id": 72865, "pid": 76337, "tid": -914061504, "ts": 1716454222506050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222506648, "dur": 56,
    "args": {
      "External id": 72868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72868,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72868, "pid": 5, "tid": 7, "ts": 1716454222506648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506070, "dur": 6,
    "args": {
      "External id": 72868,
      "cbid": 211, "correlation": 72868
    }
  },
  {
    "ph": "s", "id": 72868, "pid": 76337, "tid": -914061504, "ts": 1716454222506070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222506706, "dur": 37,
    "args": {
      "External id": 72877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72877,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72877, "pid": 5, "tid": 7, "ts": 1716454222506706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506111, "dur": 10,
    "args": {
      "External id": 72877,
      "cbid": 211, "correlation": 72877
    }
  },
  {
    "ph": "s", "id": 72877, "pid": 76337, "tid": -914061504, "ts": 1716454222506111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506182, "dur": 0,
    "args": {
      "External id": 72887,
      "cbid": 317, "correlation": 72887
    }
  },
  {
    "ph": "f", "id": 72887, "pid": 76337, "tid": -914061504, "ts": 1716454222506182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506183, "dur": 0,
    "args": {
      "External id": 72888,
      "cbid": 203, "correlation": 72888
    }
  },
  {
    "ph": "f", "id": 72888, "pid": 76337, "tid": -914061504, "ts": 1716454222506183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506184, "dur": 0,
    "args": {
      "External id": 72889,
      "cbid": 205, "correlation": 72889
    }
  },
  {
    "ph": "f", "id": 72889, "pid": 76337, "tid": -914061504, "ts": 1716454222506184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222506744, "dur": 39,
    "args": {
      "External id": 72893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72893,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72893, "pid": 5, "tid": 7, "ts": 1716454222506744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506198, "dur": 12,
    "args": {
      "External id": 72893,
      "cbid": 211, "correlation": 72893
    }
  },
  {
    "ph": "s", "id": 72893, "pid": 76337, "tid": -914061504, "ts": 1716454222506198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222506785, "dur": 15,
    "args": {
      "External id": 72895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72895,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72895, "pid": 5, "tid": 7, "ts": 1716454222506785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506213, "dur": 5,
    "args": {
      "External id": 72895,
      "cbid": 211, "correlation": 72895
    }
  },
  {
    "ph": "s", "id": 72895, "pid": 76337, "tid": -914061504, "ts": 1716454222506213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222506801, "dur": 4,
    "args": {
      "External id": 72897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72897,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72897, "pid": 5, "tid": 7, "ts": 1716454222506801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506222, "dur": 6,
    "args": {
      "External id": 72897,
      "cbid": 211, "correlation": 72897
    }
  },
  {
    "ph": "s", "id": 72897, "pid": 76337, "tid": -914061504, "ts": 1716454222506222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506231, "dur": 0,
    "args": {
      "External id": 72898,
      "cbid": 51, "correlation": 72898
    }
  },
  {
    "ph": "s", "id": 72898, "pid": 76337, "tid": -914061504, "ts": 1716454222506231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222506806, "dur": 705,
    "args": {
      "External id": 72899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72899,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 72899, "pid": 5, "tid": 7, "ts": 1716454222506806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506231, "dur": 5,
    "args": {
      "External id": 72899,
      "cbid": 211, "correlation": 72899
    }
  },
  {
    "ph": "s", "id": 72899, "pid": 76337, "tid": -914061504, "ts": 1716454222506231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222507513, "dur": 60,
    "args": {
      "External id": 72904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72904,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72904, "pid": 5, "tid": 7, "ts": 1716454222507513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506259, "dur": 8,
    "args": {
      "External id": 72904,
      "cbid": 211, "correlation": 72904
    }
  },
  {
    "ph": "s", "id": 72904, "pid": 76337, "tid": -914061504, "ts": 1716454222506259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222507574, "dur": 50,
    "args": {
      "External id": 72912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72912,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72912, "pid": 5, "tid": 7, "ts": 1716454222507574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506291, "dur": 9,
    "args": {
      "External id": 72912,
      "cbid": 211, "correlation": 72912
    }
  },
  {
    "ph": "s", "id": 72912, "pid": 76337, "tid": -914061504, "ts": 1716454222506291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222507626, "dur": 35,
    "args": {
      "External id": 72920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72920,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72920, "pid": 5, "tid": 7, "ts": 1716454222507626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506324, "dur": 10,
    "args": {
      "External id": 72920,
      "cbid": 211, "correlation": 72920
    }
  },
  {
    "ph": "s", "id": 72920, "pid": 76337, "tid": -914061504, "ts": 1716454222506324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222507662, "dur": 51,
    "args": {
      "External id": 72940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72940,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 72940, "pid": 5, "tid": 7, "ts": 1716454222507662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506407, "dur": 12,
    "args": {
      "External id": 72940,
      "cbid": 211, "correlation": 72940
    }
  },
  {
    "ph": "s", "id": 72940, "pid": 76337, "tid": -914061504, "ts": 1716454222506407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222507715, "dur": 4,
    "args": {
      "External id": 72952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72952,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 72952, "pid": 5, "tid": 7, "ts": 1716454222507715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506429, "dur": 7,
    "args": {
      "External id": 72952,
      "cbid": 211, "correlation": 72952
    }
  },
  {
    "ph": "s", "id": 72952, "pid": 76337, "tid": -914061504, "ts": 1716454222506429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222507720, "dur": 56,
    "args": {
      "External id": 72955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72955, "pid": 5, "tid": 7, "ts": 1716454222507720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506447, "dur": 6,
    "args": {
      "External id": 72955,
      "cbid": 211, "correlation": 72955
    }
  },
  {
    "ph": "s", "id": 72955, "pid": 76337, "tid": -914061504, "ts": 1716454222506447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506504, "dur": 0,
    "args": {
      "External id": 72966,
      "cbid": 317, "correlation": 72966
    }
  },
  {
    "ph": "f", "id": 72966, "pid": 76337, "tid": -914061504, "ts": 1716454222506504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506505, "dur": 0,
    "args": {
      "External id": 72967,
      "cbid": 203, "correlation": 72967
    }
  },
  {
    "ph": "f", "id": 72967, "pid": 76337, "tid": -914061504, "ts": 1716454222506505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506506, "dur": 0,
    "args": {
      "External id": 72968,
      "cbid": 205, "correlation": 72968
    }
  },
  {
    "ph": "f", "id": 72968, "pid": 76337, "tid": -914061504, "ts": 1716454222506506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506539, "dur": 2,
    "args": {
      "External id": 72972,
      "cbid": 251, "correlation": 72972
    }
  },
  {
    "ph": "f", "id": 72972, "pid": 76337, "tid": -914061504, "ts": 1716454222506539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506543, "dur": 1,
    "args": {
      "External id": 72973,
      "cbid": 251, "correlation": 72973
    }
  },
  {
    "ph": "f", "id": 72973, "pid": 76337, "tid": -914061504, "ts": 1716454222506543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506545, "dur": 1,
    "args": {
      "External id": 72974,
      "cbid": 251, "correlation": 72974
    }
  },
  {
    "ph": "f", "id": 72974, "pid": 76337, "tid": -914061504, "ts": 1716454222506545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506547, "dur": 1,
    "args": {
      "External id": 72975,
      "cbid": 251, "correlation": 72975
    }
  },
  {
    "ph": "f", "id": 72975, "pid": 76337, "tid": -914061504, "ts": 1716454222506547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506549, "dur": 1,
    "args": {
      "External id": 72976,
      "cbid": 251, "correlation": 72976
    }
  },
  {
    "ph": "f", "id": 72976, "pid": 76337, "tid": -914061504, "ts": 1716454222506549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506551, "dur": 1,
    "args": {
      "External id": 72977,
      "cbid": 251, "correlation": 72977
    }
  },
  {
    "ph": "f", "id": 72977, "pid": 76337, "tid": -914061504, "ts": 1716454222506551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506553, "dur": 1,
    "args": {
      "External id": 72978,
      "cbid": 251, "correlation": 72978
    }
  },
  {
    "ph": "f", "id": 72978, "pid": 76337, "tid": -914061504, "ts": 1716454222506553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506555, "dur": 1,
    "args": {
      "External id": 72979,
      "cbid": 251, "correlation": 72979
    }
  },
  {
    "ph": "f", "id": 72979, "pid": 76337, "tid": -914061504, "ts": 1716454222506555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506558, "dur": 0,
    "args": {
      "External id": 72980,
      "cbid": 251, "correlation": 72980
    }
  },
  {
    "ph": "f", "id": 72980, "pid": 76337, "tid": -914061504, "ts": 1716454222506558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222507777, "dur": 118,
    "args": {
      "External id": 72981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72981,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 72981, "pid": 5, "tid": 7, "ts": 1716454222507777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506562, "dur": 14,
    "args": {
      "External id": 72981,
      "cbid": 211, "correlation": 72981
    }
  },
  {
    "ph": "s", "id": 72981, "pid": 76337, "tid": -914061504, "ts": 1716454222506562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222507896, "dur": 60,
    "args": {
      "External id": 72987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72987, "pid": 5, "tid": 7, "ts": 1716454222507896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506600, "dur": 10,
    "args": {
      "External id": 72987,
      "cbid": 211, "correlation": 72987
    }
  },
  {
    "ph": "s", "id": 72987, "pid": 76337, "tid": -914061504, "ts": 1716454222506600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222507958, "dur": 552,
    "args": {
      "External id": 72996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 72996,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 72996, "pid": 5, "tid": 7, "ts": 1716454222507958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506698, "dur": 16,
    "args": {
      "External id": 72996,
      "cbid": 211, "correlation": 72996
    }
  },
  {
    "ph": "s", "id": 72996, "pid": 76337, "tid": -914061504, "ts": 1716454222506698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222508511, "dur": 184,
    "args": {
      "External id": 73018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73018,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73018, "pid": 5, "tid": 7, "ts": 1716454222508511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506772, "dur": 11,
    "args": {
      "External id": 73018,
      "cbid": 211, "correlation": 73018
    }
  },
  {
    "ph": "s", "id": 73018, "pid": 76337, "tid": -914061504, "ts": 1716454222506772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506889, "dur": 2,
    "args": {
      "External id": 73029,
      "cbid": 251, "correlation": 73029
    }
  },
  {
    "ph": "f", "id": 73029, "pid": 76337, "tid": -914061504, "ts": 1716454222506889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222508697, "dur": 199,
    "args": {
      "External id": 73030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73030,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73030, "pid": 5, "tid": 7, "ts": 1716454222508697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506896, "dur": 15,
    "args": {
      "External id": 73030,
      "cbid": 211, "correlation": 73030
    }
  },
  {
    "ph": "s", "id": 73030, "pid": 76337, "tid": -914061504, "ts": 1716454222506896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506970, "dur": 1,
    "args": {
      "External id": 73041,
      "cbid": 251, "correlation": 73041
    }
  },
  {
    "ph": "f", "id": 73041, "pid": 76337, "tid": -914061504, "ts": 1716454222506970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222508897, "dur": 191,
    "args": {
      "External id": 73042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73042,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73042, "pid": 5, "tid": 7, "ts": 1716454222508897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222506981, "dur": 13,
    "args": {
      "External id": 73042,
      "cbid": 211, "correlation": 73042
    }
  },
  {
    "ph": "s", "id": 73042, "pid": 76337, "tid": -914061504, "ts": 1716454222506981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507050, "dur": 1,
    "args": {
      "External id": 73053,
      "cbid": 251, "correlation": 73053
    }
  },
  {
    "ph": "f", "id": 73053, "pid": 76337, "tid": -914061504, "ts": 1716454222507050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222509089, "dur": 196,
    "args": {
      "External id": 73054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73054,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73054, "pid": 5, "tid": 7, "ts": 1716454222509089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507054, "dur": 11,
    "args": {
      "External id": 73054,
      "cbid": 211, "correlation": 73054
    }
  },
  {
    "ph": "s", "id": 73054, "pid": 76337, "tid": -914061504, "ts": 1716454222507054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222509286, "dur": 19050,
    "args": {
      "External id": 73075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73075,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73075, "pid": 5, "tid": 7, "ts": 1716454222509286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507163, "dur": 15,
    "args": {
      "External id": 73075,
      "cbid": 211, "correlation": 73075
    }
  },
  {
    "ph": "s", "id": 73075, "pid": 76337, "tid": -914061504, "ts": 1716454222507163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507277, "dur": 2,
    "args": {
      "External id": 73093,
      "cbid": 251, "correlation": 73093
    }
  },
  {
    "ph": "f", "id": 73093, "pid": 76337, "tid": -914061504, "ts": 1716454222507277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222528338, "dur": 206,
    "args": {
      "External id": 73095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73095,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73095, "pid": 5, "tid": 7, "ts": 1716454222528338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507284, "dur": 14,
    "args": {
      "External id": 73095,
      "cbid": 211, "correlation": 73095
    }
  },
  {
    "ph": "s", "id": 73095, "pid": 76337, "tid": -914061504, "ts": 1716454222507284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222528545, "dur": 66,
    "args": {
      "External id": 73103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73103,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73103, "pid": 5, "tid": 7, "ts": 1716454222528545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507360, "dur": 13,
    "args": {
      "External id": 73103,
      "cbid": 211, "correlation": 73103
    }
  },
  {
    "ph": "s", "id": 73103, "pid": 76337, "tid": -914061504, "ts": 1716454222507360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222528613, "dur": 97,
    "args": {
      "External id": 73111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73111,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73111, "pid": 5, "tid": 7, "ts": 1716454222528613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507400, "dur": 10,
    "args": {
      "External id": 73111,
      "cbid": 211, "correlation": 73111
    }
  },
  {
    "ph": "s", "id": 73111, "pid": 76337, "tid": -914061504, "ts": 1716454222507400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222528711, "dur": 56,
    "args": {
      "External id": 73122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73122,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73122, "pid": 5, "tid": 7, "ts": 1716454222528711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507487, "dur": 14,
    "args": {
      "External id": 73122,
      "cbid": 211, "correlation": 73122
    }
  },
  {
    "ph": "s", "id": 73122, "pid": 76337, "tid": -914061504, "ts": 1716454222507487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222528768, "dur": 94,
    "args": {
      "External id": 73144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73144,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73144, "pid": 5, "tid": 7, "ts": 1716454222528768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507521, "dur": 10,
    "args": {
      "External id": 73144,
      "cbid": 211, "correlation": 73144
    }
  },
  {
    "ph": "s", "id": 73144, "pid": 76337, "tid": -914061504, "ts": 1716454222507521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507606, "dur": 1,
    "args": {
      "External id": 73155,
      "cbid": 251, "correlation": 73155
    }
  },
  {
    "ph": "f", "id": 73155, "pid": 76337, "tid": -914061504, "ts": 1716454222507606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222528864, "dur": 107,
    "args": {
      "External id": 73156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73156,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73156, "pid": 5, "tid": 7, "ts": 1716454222528864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507611, "dur": 12,
    "args": {
      "External id": 73156,
      "cbid": 211, "correlation": 73156
    }
  },
  {
    "ph": "s", "id": 73156, "pid": 76337, "tid": -914061504, "ts": 1716454222507611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507694, "dur": 1,
    "args": {
      "External id": 73167,
      "cbid": 251, "correlation": 73167
    }
  },
  {
    "ph": "f", "id": 73167, "pid": 76337, "tid": -914061504, "ts": 1716454222507694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507698, "dur": 0,
    "args": {
      "External id": 73168,
      "cbid": 251, "correlation": 73168
    }
  },
  {
    "ph": "f", "id": 73168, "pid": 76337, "tid": -914061504, "ts": 1716454222507698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222528972, "dur": 10,
    "args": {
      "External id": 73169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73169,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 73169, "pid": 5, "tid": 7, "ts": 1716454222528972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507700, "dur": 14,
    "args": {
      "External id": 73169,
      "cbid": 211, "correlation": 73169
    }
  },
  {
    "ph": "s", "id": 73169, "pid": 76337, "tid": -914061504, "ts": 1716454222507700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222528984, "dur": 5,
    "args": {
      "External id": 73171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73171,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 73171, "pid": 5, "tid": 7, "ts": 1716454222528984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507718, "dur": 7,
    "args": {
      "External id": 73171,
      "cbid": 211, "correlation": 73171
    }
  },
  {
    "ph": "s", "id": 73171, "pid": 76337, "tid": -914061504, "ts": 1716454222507718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507779, "dur": 1,
    "args": {
      "External id": 73182,
      "cbid": 251, "correlation": 73182
    }
  },
  {
    "ph": "f", "id": 73182, "pid": 76337, "tid": -914061504, "ts": 1716454222507779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507783, "dur": 0,
    "args": {
      "External id": 73183,
      "cbid": 251, "correlation": 73183
    }
  },
  {
    "ph": "f", "id": 73183, "pid": 76337, "tid": -914061504, "ts": 1716454222507783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222528990, "dur": 6,
    "args": {
      "External id": 73184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73184,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 73184, "pid": 5, "tid": 7, "ts": 1716454222528990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507784, "dur": 12,
    "args": {
      "External id": 73184,
      "cbid": 211, "correlation": 73184
    }
  },
  {
    "ph": "s", "id": 73184, "pid": 76337, "tid": -914061504, "ts": 1716454222507784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222528998, "dur": 4,
    "args": {
      "External id": 73186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73186,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 73186, "pid": 5, "tid": 7, "ts": 1716454222528998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507797, "dur": 5,
    "args": {
      "External id": 73186,
      "cbid": 211, "correlation": 73186
    }
  },
  {
    "ph": "s", "id": 73186, "pid": 76337, "tid": -914061504, "ts": 1716454222507797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222529003, "dur": 159,
    "args": {
      "External id": 73207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73207,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73207, "pid": 5, "tid": 7, "ts": 1716454222529003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507872, "dur": 12,
    "args": {
      "External id": 73207,
      "cbid": 211, "correlation": 73207
    }
  },
  {
    "ph": "s", "id": 73207, "pid": 76337, "tid": -914061504, "ts": 1716454222507872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507971, "dur": 9,
    "args": {
      "External id": 73225,
      "cbid": 251, "correlation": 73225
    }
  },
  {
    "ph": "f", "id": 73225, "pid": 76337, "tid": -914061504, "ts": 1716454222507971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222529163, "dur": 109,
    "args": {
      "External id": 73227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73227,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73227, "pid": 5, "tid": 7, "ts": 1716454222529163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222507985, "dur": 14,
    "args": {
      "External id": 73227,
      "cbid": 211, "correlation": 73227
    }
  },
  {
    "ph": "s", "id": 73227, "pid": 76337, "tid": -914061504, "ts": 1716454222507985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222529274, "dur": 35,
    "args": {
      "External id": 73235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73235,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73235, "pid": 5, "tid": 7, "ts": 1716454222529274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508059, "dur": 12,
    "args": {
      "External id": 73235,
      "cbid": 211, "correlation": 73235
    }
  },
  {
    "ph": "s", "id": 73235, "pid": 76337, "tid": -914061504, "ts": 1716454222508059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222529310, "dur": 69,
    "args": {
      "External id": 73243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73243, "pid": 5, "tid": 7, "ts": 1716454222529310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508101, "dur": 9,
    "args": {
      "External id": 73243,
      "cbid": 211, "correlation": 73243
    }
  },
  {
    "ph": "s", "id": 73243, "pid": 76337, "tid": -914061504, "ts": 1716454222508101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222529380, "dur": 94,
    "args": {
      "External id": 73265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73265,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73265, "pid": 5, "tid": 7, "ts": 1716454222529380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508153, "dur": 10,
    "args": {
      "External id": 73265,
      "cbid": 211, "correlation": 73265
    }
  },
  {
    "ph": "s", "id": 73265, "pid": 76337, "tid": -914061504, "ts": 1716454222508153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508245, "dur": 1,
    "args": {
      "External id": 73281,
      "cbid": 251, "correlation": 73281
    }
  },
  {
    "ph": "f", "id": 73281, "pid": 76337, "tid": -914061504, "ts": 1716454222508245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222529475, "dur": 585,
    "args": {
      "External id": 73283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73283,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73283, "pid": 5, "tid": 7, "ts": 1716454222529475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508250, "dur": 13,
    "args": {
      "External id": 73283,
      "cbid": 211, "correlation": 73283
    }
  },
  {
    "ph": "s", "id": 73283, "pid": 76337, "tid": -914061504, "ts": 1716454222508250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222530061, "dur": 244,
    "args": {
      "External id": 73291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73291,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73291, "pid": 5, "tid": 7, "ts": 1716454222530061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508329, "dur": 14,
    "args": {
      "External id": 73291,
      "cbid": 211, "correlation": 73291
    }
  },
  {
    "ph": "s", "id": 73291, "pid": 76337, "tid": -914061504, "ts": 1716454222508329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222530306, "dur": 253,
    "args": {
      "External id": 73299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73299, "pid": 5, "tid": 7, "ts": 1716454222530306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508363, "dur": 9,
    "args": {
      "External id": 73299,
      "cbid": 211, "correlation": 73299
    }
  },
  {
    "ph": "s", "id": 73299, "pid": 76337, "tid": -914061504, "ts": 1716454222508363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508451, "dur": 2,
    "args": {
      "External id": 73315,
      "cbid": 251, "correlation": 73315
    }
  },
  {
    "ph": "f", "id": 73315, "pid": 76337, "tid": -914061504, "ts": 1716454222508451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508457, "dur": 0,
    "args": {
      "External id": 73317,
      "cbid": 251, "correlation": 73317
    }
  },
  {
    "ph": "f", "id": 73317, "pid": 76337, "tid": -914061504, "ts": 1716454222508457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222530560, "dur": 362,
    "args": {
      "External id": 73318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73318,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 73318, "pid": 5, "tid": 7, "ts": 1716454222530560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508462, "dur": 13,
    "args": {
      "External id": 73318,
      "cbid": 211, "correlation": 73318
    }
  },
  {
    "ph": "s", "id": 73318, "pid": 76337, "tid": -914061504, "ts": 1716454222508462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222530923, "dur": 50,
    "args": {
      "External id": 73326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73326,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73326, "pid": 5, "tid": 7, "ts": 1716454222530923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508505, "dur": 10,
    "args": {
      "External id": 73326,
      "cbid": 211, "correlation": 73326
    }
  },
  {
    "ph": "s", "id": 73326, "pid": 76337, "tid": -914061504, "ts": 1716454222508505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222530974, "dur": 160,
    "args": {
      "External id": 73337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73337,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73337, "pid": 5, "tid": 7, "ts": 1716454222530974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508579, "dur": 13,
    "args": {
      "External id": 73337,
      "cbid": 211, "correlation": 73337
    }
  },
  {
    "ph": "s", "id": 73337, "pid": 76337, "tid": -914061504, "ts": 1716454222508579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508649, "dur": 0,
    "args": {
      "External id": 73349,
      "cbid": 317, "correlation": 73349
    }
  },
  {
    "ph": "f", "id": 73349, "pid": 76337, "tid": -914061504, "ts": 1716454222508649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508650, "dur": 0,
    "args": {
      "External id": 73350,
      "cbid": 203, "correlation": 73350
    }
  },
  {
    "ph": "f", "id": 73350, "pid": 76337, "tid": -914061504, "ts": 1716454222508650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222508650, "dur": 0,
    "args": {
      "External id": 73351,
      "cbid": 205, "correlation": 73351
    }
  },
  {
    "ph": "f", "id": 73351, "pid": 76337, "tid": -914061504, "ts": 1716454222508650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509301, "dur": 2,
    "args": {
      "External id": 73355,
      "cbid": 251, "correlation": 73355
    }
  },
  {
    "ph": "f", "id": 73355, "pid": 76337, "tid": -914061504, "ts": 1716454222509301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509303, "dur": 0,
    "args": {
      "External id": 73356,
      "cbid": 251, "correlation": 73356
    }
  },
  {
    "ph": "f", "id": 73356, "pid": 76337, "tid": -914061504, "ts": 1716454222509303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509304, "dur": 0,
    "args": {
      "External id": 73357,
      "cbid": 251, "correlation": 73357
    }
  },
  {
    "ph": "f", "id": 73357, "pid": 76337, "tid": -914061504, "ts": 1716454222509304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509306, "dur": 0,
    "args": {
      "External id": 73358,
      "cbid": 251, "correlation": 73358
    }
  },
  {
    "ph": "f", "id": 73358, "pid": 76337, "tid": -914061504, "ts": 1716454222509306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509307, "dur": 1,
    "args": {
      "External id": 73359,
      "cbid": 251, "correlation": 73359
    }
  },
  {
    "ph": "f", "id": 73359, "pid": 76337, "tid": -914061504, "ts": 1716454222509307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509308, "dur": 0,
    "args": {
      "External id": 73360,
      "cbid": 251, "correlation": 73360
    }
  },
  {
    "ph": "f", "id": 73360, "pid": 76337, "tid": -914061504, "ts": 1716454222509308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509309, "dur": 0,
    "args": {
      "External id": 73361,
      "cbid": 251, "correlation": 73361
    }
  },
  {
    "ph": "f", "id": 73361, "pid": 76337, "tid": -914061504, "ts": 1716454222509309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509310, "dur": 0,
    "args": {
      "External id": 73362,
      "cbid": 251, "correlation": 73362
    }
  },
  {
    "ph": "f", "id": 73362, "pid": 76337, "tid": -914061504, "ts": 1716454222509310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509312, "dur": 0,
    "args": {
      "External id": 73363,
      "cbid": 251, "correlation": 73363
    }
  },
  {
    "ph": "f", "id": 73363, "pid": 76337, "tid": -914061504, "ts": 1716454222509312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222531136, "dur": 115,
    "args": {
      "External id": 73364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73364,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73364, "pid": 5, "tid": 7, "ts": 1716454222531136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509314, "dur": 14,
    "args": {
      "External id": 73364,
      "cbid": 211, "correlation": 73364
    }
  },
  {
    "ph": "s", "id": 73364, "pid": 76337, "tid": -914061504, "ts": 1716454222509314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222531252, "dur": 60,
    "args": {
      "External id": 73370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73370,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73370, "pid": 5, "tid": 7, "ts": 1716454222531252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509354, "dur": 9,
    "args": {
      "External id": 73370,
      "cbid": 211, "correlation": 73370
    }
  },
  {
    "ph": "s", "id": 73370, "pid": 76337, "tid": -914061504, "ts": 1716454222509354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222531314, "dur": 50,
    "args": {
      "External id": 73378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73378,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73378, "pid": 5, "tid": 7, "ts": 1716454222531314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509389, "dur": 9,
    "args": {
      "External id": 73378,
      "cbid": 211, "correlation": 73378
    }
  },
  {
    "ph": "s", "id": 73378, "pid": 76337, "tid": -914061504, "ts": 1716454222509389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222531365, "dur": 53,
    "args": {
      "External id": 73398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73398,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 73398, "pid": 5, "tid": 7, "ts": 1716454222531365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509467, "dur": 11,
    "args": {
      "External id": 73398,
      "cbid": 211, "correlation": 73398
    }
  },
  {
    "ph": "s", "id": 73398, "pid": 76337, "tid": -914061504, "ts": 1716454222509467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222531419, "dur": 4,
    "args": {
      "External id": 73410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73410,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 73410, "pid": 5, "tid": 7, "ts": 1716454222531419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509490, "dur": 7,
    "args": {
      "External id": 73410,
      "cbid": 211, "correlation": 73410
    }
  },
  {
    "ph": "s", "id": 73410, "pid": 76337, "tid": -914061504, "ts": 1716454222509490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222531425, "dur": 58,
    "args": {
      "External id": 73413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73413, "pid": 5, "tid": 7, "ts": 1716454222531425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509509, "dur": 7,
    "args": {
      "External id": 73413,
      "cbid": 211, "correlation": 73413
    }
  },
  {
    "ph": "s", "id": 73413, "pid": 76337, "tid": -914061504, "ts": 1716454222509509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222531484, "dur": 38,
    "args": {
      "External id": 73422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73422,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73422, "pid": 5, "tid": 7, "ts": 1716454222531484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509551, "dur": 10,
    "args": {
      "External id": 73422,
      "cbid": 211, "correlation": 73422
    }
  },
  {
    "ph": "s", "id": 73422, "pid": 76337, "tid": -914061504, "ts": 1716454222509551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509605, "dur": 0,
    "args": {
      "External id": 73432,
      "cbid": 317, "correlation": 73432
    }
  },
  {
    "ph": "f", "id": 73432, "pid": 76337, "tid": -914061504, "ts": 1716454222509605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509606, "dur": 0,
    "args": {
      "External id": 73433,
      "cbid": 203, "correlation": 73433
    }
  },
  {
    "ph": "f", "id": 73433, "pid": 76337, "tid": -914061504, "ts": 1716454222509606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509607, "dur": 0,
    "args": {
      "External id": 73434,
      "cbid": 205, "correlation": 73434
    }
  },
  {
    "ph": "f", "id": 73434, "pid": 76337, "tid": -914061504, "ts": 1716454222509607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222531523, "dur": 42,
    "args": {
      "External id": 73438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73438, "pid": 5, "tid": 7, "ts": 1716454222531523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509640, "dur": 12,
    "args": {
      "External id": 73438,
      "cbid": 211, "correlation": 73438
    }
  },
  {
    "ph": "s", "id": 73438, "pid": 76337, "tid": -914061504, "ts": 1716454222509640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222531567, "dur": 15,
    "args": {
      "External id": 73440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73440,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73440, "pid": 5, "tid": 7, "ts": 1716454222531567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509655, "dur": 6,
    "args": {
      "External id": 73440,
      "cbid": 211, "correlation": 73440
    }
  },
  {
    "ph": "s", "id": 73440, "pid": 76337, "tid": -914061504, "ts": 1716454222509655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222531582, "dur": 4,
    "args": {
      "External id": 73442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 73442, "pid": 5, "tid": 7, "ts": 1716454222531582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509666, "dur": 6,
    "args": {
      "External id": 73442,
      "cbid": 211, "correlation": 73442
    }
  },
  {
    "ph": "s", "id": 73442, "pid": 76337, "tid": -914061504, "ts": 1716454222509666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509675, "dur": 0,
    "args": {
      "External id": 73443,
      "cbid": 51, "correlation": 73443
    }
  },
  {
    "ph": "s", "id": 73443, "pid": 76337, "tid": -914061504, "ts": 1716454222509675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222531587, "dur": 712,
    "args": {
      "External id": 73444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73444,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73444, "pid": 5, "tid": 7, "ts": 1716454222531587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509676, "dur": 5,
    "args": {
      "External id": 73444,
      "cbid": 211, "correlation": 73444
    }
  },
  {
    "ph": "s", "id": 73444, "pid": 76337, "tid": -914061504, "ts": 1716454222509676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222532301, "dur": 60,
    "args": {
      "External id": 73449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73449, "pid": 5, "tid": 7, "ts": 1716454222532301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509703, "dur": 9,
    "args": {
      "External id": 73449,
      "cbid": 211, "correlation": 73449
    }
  },
  {
    "ph": "s", "id": 73449, "pid": 76337, "tid": -914061504, "ts": 1716454222509703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222532362, "dur": 3,
    "args": {
      "External id": 73457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73457,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 73457, "pid": 5, "tid": 7, "ts": 1716454222532362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509746, "dur": 9,
    "args": {
      "External id": 73457,
      "cbid": 211, "correlation": 73457
    }
  },
  {
    "ph": "s", "id": 73457, "pid": 76337, "tid": -914061504, "ts": 1716454222509746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509813, "dur": 2,
    "args": {
      "External id": 73473,
      "cbid": 251, "correlation": 73473
    }
  },
  {
    "ph": "f", "id": 73473, "pid": 76337, "tid": -914061504, "ts": 1716454222509813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509819, "dur": 0,
    "args": {
      "External id": 73475,
      "cbid": 251, "correlation": 73475
    }
  },
  {
    "ph": "f", "id": 73475, "pid": 76337, "tid": -914061504, "ts": 1716454222509819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222532367, "dur": 11,
    "args": {
      "External id": 73476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73476,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 73476, "pid": 5, "tid": 7, "ts": 1716454222532367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509821, "dur": 11,
    "args": {
      "External id": 73476,
      "cbid": 211, "correlation": 73476
    }
  },
  {
    "ph": "s", "id": 73476, "pid": 76337, "tid": -914061504, "ts": 1716454222509821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222532379, "dur": 5,
    "args": {
      "External id": 73478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73478,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 73478, "pid": 5, "tid": 7, "ts": 1716454222532379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509835, "dur": 5,
    "args": {
      "External id": 73478,
      "cbid": 211, "correlation": 73478
    }
  },
  {
    "ph": "s", "id": 73478, "pid": 76337, "tid": -914061504, "ts": 1716454222509835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222532386, "dur": 55,
    "args": {
      "External id": 73488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73488, "pid": 5, "tid": 7, "ts": 1716454222532386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509893, "dur": 12,
    "args": {
      "External id": 73488,
      "cbid": 211, "correlation": 73488
    }
  },
  {
    "ph": "s", "id": 73488, "pid": 76337, "tid": -914061504, "ts": 1716454222509893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222532442, "dur": 51,
    "args": {
      "External id": 73508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73508,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 73508, "pid": 5, "tid": 7, "ts": 1716454222532442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509960, "dur": 11,
    "args": {
      "External id": 73508,
      "cbid": 211, "correlation": 73508
    }
  },
  {
    "ph": "s", "id": 73508, "pid": 76337, "tid": -914061504, "ts": 1716454222509960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222532494, "dur": 4,
    "args": {
      "External id": 73520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73520,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 73520, "pid": 5, "tid": 7, "ts": 1716454222532494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222509989, "dur": 7,
    "args": {
      "External id": 73520,
      "cbid": 211, "correlation": 73520
    }
  },
  {
    "ph": "s", "id": 73520, "pid": 76337, "tid": -914061504, "ts": 1716454222509989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222532500, "dur": 57,
    "args": {
      "External id": 73523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73523,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73523, "pid": 5, "tid": 7, "ts": 1716454222532500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510008, "dur": 6,
    "args": {
      "External id": 73523,
      "cbid": 211, "correlation": 73523
    }
  },
  {
    "ph": "s", "id": 73523, "pid": 76337, "tid": -914061504, "ts": 1716454222510008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222532558, "dur": 36,
    "args": {
      "External id": 73532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73532,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73532, "pid": 5, "tid": 7, "ts": 1716454222532558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510050, "dur": 9,
    "args": {
      "External id": 73532,
      "cbid": 211, "correlation": 73532
    }
  },
  {
    "ph": "s", "id": 73532, "pid": 76337, "tid": -914061504, "ts": 1716454222510050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510113, "dur": 0,
    "args": {
      "External id": 73542,
      "cbid": 317, "correlation": 73542
    }
  },
  {
    "ph": "f", "id": 73542, "pid": 76337, "tid": -914061504, "ts": 1716454222510113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510113, "dur": 0,
    "args": {
      "External id": 73543,
      "cbid": 203, "correlation": 73543
    }
  },
  {
    "ph": "f", "id": 73543, "pid": 76337, "tid": -914061504, "ts": 1716454222510113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510114, "dur": 0,
    "args": {
      "External id": 73544,
      "cbid": 205, "correlation": 73544
    }
  },
  {
    "ph": "f", "id": 73544, "pid": 76337, "tid": -914061504, "ts": 1716454222510114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222532595, "dur": 40,
    "args": {
      "External id": 73548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73548,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73548, "pid": 5, "tid": 7, "ts": 1716454222532595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510129, "dur": 12,
    "args": {
      "External id": 73548,
      "cbid": 211, "correlation": 73548
    }
  },
  {
    "ph": "s", "id": 73548, "pid": 76337, "tid": -914061504, "ts": 1716454222510129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222532636, "dur": 15,
    "args": {
      "External id": 73550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73550,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73550, "pid": 5, "tid": 7, "ts": 1716454222532636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510143, "dur": 5,
    "args": {
      "External id": 73550,
      "cbid": 211, "correlation": 73550
    }
  },
  {
    "ph": "s", "id": 73550, "pid": 76337, "tid": -914061504, "ts": 1716454222510143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222532652, "dur": 3,
    "args": {
      "External id": 73552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73552,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 73552, "pid": 5, "tid": 7, "ts": 1716454222532652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510152, "dur": 5,
    "args": {
      "External id": 73552,
      "cbid": 211, "correlation": 73552
    }
  },
  {
    "ph": "s", "id": 73552, "pid": 76337, "tid": -914061504, "ts": 1716454222510152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510160, "dur": 0,
    "args": {
      "External id": 73553,
      "cbid": 51, "correlation": 73553
    }
  },
  {
    "ph": "s", "id": 73553, "pid": 76337, "tid": -914061504, "ts": 1716454222510160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222532656, "dur": 707,
    "args": {
      "External id": 73554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73554,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73554, "pid": 5, "tid": 7, "ts": 1716454222532656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510161, "dur": 5,
    "args": {
      "External id": 73554,
      "cbid": 211, "correlation": 73554
    }
  },
  {
    "ph": "s", "id": 73554, "pid": 76337, "tid": -914061504, "ts": 1716454222510161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222533365, "dur": 60,
    "args": {
      "External id": 73559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73559, "pid": 5, "tid": 7, "ts": 1716454222533365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510188, "dur": 8,
    "args": {
      "External id": 73559,
      "cbid": 211, "correlation": 73559
    }
  },
  {
    "ph": "s", "id": 73559, "pid": 76337, "tid": -914061504, "ts": 1716454222510188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222533426, "dur": 50,
    "args": {
      "External id": 73567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73567,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73567, "pid": 5, "tid": 7, "ts": 1716454222533426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510219, "dur": 8,
    "args": {
      "External id": 73567,
      "cbid": 211, "correlation": 73567
    }
  },
  {
    "ph": "s", "id": 73567, "pid": 76337, "tid": -914061504, "ts": 1716454222510219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222533478, "dur": 36,
    "args": {
      "External id": 73575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73575,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73575, "pid": 5, "tid": 7, "ts": 1716454222533478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510249, "dur": 8,
    "args": {
      "External id": 73575,
      "cbid": 211, "correlation": 73575
    }
  },
  {
    "ph": "s", "id": 73575, "pid": 76337, "tid": -914061504, "ts": 1716454222510249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222533515, "dur": 53,
    "args": {
      "External id": 73595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73595,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 73595, "pid": 5, "tid": 7, "ts": 1716454222533515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510329, "dur": 13,
    "args": {
      "External id": 73595,
      "cbid": 211, "correlation": 73595
    }
  },
  {
    "ph": "s", "id": 73595, "pid": 76337, "tid": -914061504, "ts": 1716454222510329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222533569, "dur": 4,
    "args": {
      "External id": 73607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73607,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 73607, "pid": 5, "tid": 7, "ts": 1716454222533569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510351, "dur": 6,
    "args": {
      "External id": 73607,
      "cbid": 211, "correlation": 73607
    }
  },
  {
    "ph": "s", "id": 73607, "pid": 76337, "tid": -914061504, "ts": 1716454222510351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222533574, "dur": 55,
    "args": {
      "External id": 73610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73610,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73610, "pid": 5, "tid": 7, "ts": 1716454222533574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510368, "dur": 6,
    "args": {
      "External id": 73610,
      "cbid": 211, "correlation": 73610
    }
  },
  {
    "ph": "s", "id": 73610, "pid": 76337, "tid": -914061504, "ts": 1716454222510368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510424, "dur": 0,
    "args": {
      "External id": 73621,
      "cbid": 317, "correlation": 73621
    }
  },
  {
    "ph": "f", "id": 73621, "pid": 76337, "tid": -914061504, "ts": 1716454222510424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510424, "dur": 0,
    "args": {
      "External id": 73622,
      "cbid": 203, "correlation": 73622
    }
  },
  {
    "ph": "f", "id": 73622, "pid": 76337, "tid": -914061504, "ts": 1716454222510424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510425, "dur": 0,
    "args": {
      "External id": 73623,
      "cbid": 205, "correlation": 73623
    }
  },
  {
    "ph": "f", "id": 73623, "pid": 76337, "tid": -914061504, "ts": 1716454222510425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510447, "dur": 1,
    "args": {
      "External id": 73627,
      "cbid": 251, "correlation": 73627
    }
  },
  {
    "ph": "f", "id": 73627, "pid": 76337, "tid": -914061504, "ts": 1716454222510447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510449, "dur": 0,
    "args": {
      "External id": 73628,
      "cbid": 251, "correlation": 73628
    }
  },
  {
    "ph": "f", "id": 73628, "pid": 76337, "tid": -914061504, "ts": 1716454222510449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510450, "dur": 0,
    "args": {
      "External id": 73629,
      "cbid": 251, "correlation": 73629
    }
  },
  {
    "ph": "f", "id": 73629, "pid": 76337, "tid": -914061504, "ts": 1716454222510450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510451, "dur": 0,
    "args": {
      "External id": 73630,
      "cbid": 251, "correlation": 73630
    }
  },
  {
    "ph": "f", "id": 73630, "pid": 76337, "tid": -914061504, "ts": 1716454222510451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510451, "dur": 0,
    "args": {
      "External id": 73631,
      "cbid": 251, "correlation": 73631
    }
  },
  {
    "ph": "f", "id": 73631, "pid": 76337, "tid": -914061504, "ts": 1716454222510451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510452, "dur": 0,
    "args": {
      "External id": 73632,
      "cbid": 251, "correlation": 73632
    }
  },
  {
    "ph": "f", "id": 73632, "pid": 76337, "tid": -914061504, "ts": 1716454222510452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510453, "dur": 0,
    "args": {
      "External id": 73633,
      "cbid": 251, "correlation": 73633
    }
  },
  {
    "ph": "f", "id": 73633, "pid": 76337, "tid": -914061504, "ts": 1716454222510453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510454, "dur": 0,
    "args": {
      "External id": 73634,
      "cbid": 251, "correlation": 73634
    }
  },
  {
    "ph": "f", "id": 73634, "pid": 76337, "tid": -914061504, "ts": 1716454222510454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510455, "dur": 0,
    "args": {
      "External id": 73635,
      "cbid": 251, "correlation": 73635
    }
  },
  {
    "ph": "f", "id": 73635, "pid": 76337, "tid": -914061504, "ts": 1716454222510455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222533631, "dur": 114,
    "args": {
      "External id": 73636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73636,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73636, "pid": 5, "tid": 7, "ts": 1716454222533631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510457, "dur": 12,
    "args": {
      "External id": 73636,
      "cbid": 211, "correlation": 73636
    }
  },
  {
    "ph": "s", "id": 73636, "pid": 76337, "tid": -914061504, "ts": 1716454222510457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222533746, "dur": 60,
    "args": {
      "External id": 73642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73642,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73642, "pid": 5, "tid": 7, "ts": 1716454222533746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510492, "dur": 9,
    "args": {
      "External id": 73642,
      "cbid": 211, "correlation": 73642
    }
  },
  {
    "ph": "s", "id": 73642, "pid": 76337, "tid": -914061504, "ts": 1716454222510492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222533808, "dur": 663,
    "args": {
      "External id": 73651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73651,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73651, "pid": 5, "tid": 7, "ts": 1716454222533808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510577, "dur": 14,
    "args": {
      "External id": 73651,
      "cbid": 211, "correlation": 73651
    }
  },
  {
    "ph": "s", "id": 73651, "pid": 76337, "tid": -914061504, "ts": 1716454222510577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222534472, "dur": 184,
    "args": {
      "External id": 73673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73673,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73673, "pid": 5, "tid": 7, "ts": 1716454222534472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510636, "dur": 10,
    "args": {
      "External id": 73673,
      "cbid": 211, "correlation": 73673
    }
  },
  {
    "ph": "s", "id": 73673, "pid": 76337, "tid": -914061504, "ts": 1716454222510636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510726, "dur": 1,
    "args": {
      "External id": 73684,
      "cbid": 251, "correlation": 73684
    }
  },
  {
    "ph": "f", "id": 73684, "pid": 76337, "tid": -914061504, "ts": 1716454222510726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222534658, "dur": 198,
    "args": {
      "External id": 73685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73685,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73685, "pid": 5, "tid": 7, "ts": 1716454222534658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510731, "dur": 12,
    "args": {
      "External id": 73685,
      "cbid": 211, "correlation": 73685
    }
  },
  {
    "ph": "s", "id": 73685, "pid": 76337, "tid": -914061504, "ts": 1716454222510731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510799, "dur": 1,
    "args": {
      "External id": 73696,
      "cbid": 251, "correlation": 73696
    }
  },
  {
    "ph": "f", "id": 73696, "pid": 76337, "tid": -914061504, "ts": 1716454222510799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222534857, "dur": 188,
    "args": {
      "External id": 73697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73697,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73697, "pid": 5, "tid": 7, "ts": 1716454222534857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510803, "dur": 12,
    "args": {
      "External id": 73697,
      "cbid": 211, "correlation": 73697
    }
  },
  {
    "ph": "s", "id": 73697, "pid": 76337, "tid": -914061504, "ts": 1716454222510803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510866, "dur": 1,
    "args": {
      "External id": 73708,
      "cbid": 251, "correlation": 73708
    }
  },
  {
    "ph": "f", "id": 73708, "pid": 76337, "tid": -914061504, "ts": 1716454222510866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222535046, "dur": 188,
    "args": {
      "External id": 73709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73709,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73709, "pid": 5, "tid": 7, "ts": 1716454222535046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510871, "dur": 11,
    "args": {
      "External id": 73709,
      "cbid": 211, "correlation": 73709
    }
  },
  {
    "ph": "s", "id": 73709, "pid": 76337, "tid": -914061504, "ts": 1716454222510871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222535235, "dur": 18980,
    "args": {
      "External id": 73730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73730,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73730, "pid": 5, "tid": 7, "ts": 1716454222535235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222510952, "dur": 13,
    "args": {
      "External id": 73730,
      "cbid": 211, "correlation": 73730
    }
  },
  {
    "ph": "s", "id": 73730, "pid": 76337, "tid": -914061504, "ts": 1716454222510952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511061, "dur": 1,
    "args": {
      "External id": 73748,
      "cbid": 251, "correlation": 73748
    }
  },
  {
    "ph": "f", "id": 73748, "pid": 76337, "tid": -914061504, "ts": 1716454222511061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222554216, "dur": 207,
    "args": {
      "External id": 73750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73750,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73750, "pid": 5, "tid": 7, "ts": 1716454222554216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511066, "dur": 14,
    "args": {
      "External id": 73750,
      "cbid": 211, "correlation": 73750
    }
  },
  {
    "ph": "s", "id": 73750, "pid": 76337, "tid": -914061504, "ts": 1716454222511066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222554425, "dur": 66,
    "args": {
      "External id": 73758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73758,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73758, "pid": 5, "tid": 7, "ts": 1716454222554425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511138, "dur": 12,
    "args": {
      "External id": 73758,
      "cbid": 211, "correlation": 73758
    }
  },
  {
    "ph": "s", "id": 73758, "pid": 76337, "tid": -914061504, "ts": 1716454222511138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222554492, "dur": 97,
    "args": {
      "External id": 73766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73766,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73766, "pid": 5, "tid": 7, "ts": 1716454222554492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511178, "dur": 9,
    "args": {
      "External id": 73766,
      "cbid": 211, "correlation": 73766
    }
  },
  {
    "ph": "s", "id": 73766, "pid": 76337, "tid": -914061504, "ts": 1716454222511178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222554590, "dur": 56,
    "args": {
      "External id": 73777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73777, "pid": 5, "tid": 7, "ts": 1716454222554590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511251, "dur": 12,
    "args": {
      "External id": 73777,
      "cbid": 211, "correlation": 73777
    }
  },
  {
    "ph": "s", "id": 73777, "pid": 76337, "tid": -914061504, "ts": 1716454222511251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222554648, "dur": 94,
    "args": {
      "External id": 73799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73799,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73799, "pid": 5, "tid": 7, "ts": 1716454222554648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511282, "dur": 8,
    "args": {
      "External id": 73799,
      "cbid": 211, "correlation": 73799
    }
  },
  {
    "ph": "s", "id": 73799, "pid": 76337, "tid": -914061504, "ts": 1716454222511282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511366, "dur": 1,
    "args": {
      "External id": 73810,
      "cbid": 251, "correlation": 73810
    }
  },
  {
    "ph": "f", "id": 73810, "pid": 76337, "tid": -914061504, "ts": 1716454222511366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222554743, "dur": 105,
    "args": {
      "External id": 73811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73811,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73811, "pid": 5, "tid": 7, "ts": 1716454222554743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511371, "dur": 12,
    "args": {
      "External id": 73811,
      "cbid": 211, "correlation": 73811
    }
  },
  {
    "ph": "s", "id": 73811, "pid": 76337, "tid": -914061504, "ts": 1716454222511371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511444, "dur": 1,
    "args": {
      "External id": 73822,
      "cbid": 251, "correlation": 73822
    }
  },
  {
    "ph": "f", "id": 73822, "pid": 76337, "tid": -914061504, "ts": 1716454222511444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511448, "dur": 0,
    "args": {
      "External id": 73823,
      "cbid": 251, "correlation": 73823
    }
  },
  {
    "ph": "f", "id": 73823, "pid": 76337, "tid": -914061504, "ts": 1716454222511448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222554849, "dur": 10,
    "args": {
      "External id": 73824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73824,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 73824, "pid": 5, "tid": 7, "ts": 1716454222554849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511449, "dur": 12,
    "args": {
      "External id": 73824,
      "cbid": 211, "correlation": 73824
    }
  },
  {
    "ph": "s", "id": 73824, "pid": 76337, "tid": -914061504, "ts": 1716454222511449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222554860, "dur": 5,
    "args": {
      "External id": 73826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73826,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 73826, "pid": 5, "tid": 7, "ts": 1716454222554860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511464, "dur": 6,
    "args": {
      "External id": 73826,
      "cbid": 211, "correlation": 73826
    }
  },
  {
    "ph": "s", "id": 73826, "pid": 76337, "tid": -914061504, "ts": 1716454222511464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511524, "dur": 1,
    "args": {
      "External id": 73837,
      "cbid": 251, "correlation": 73837
    }
  },
  {
    "ph": "f", "id": 73837, "pid": 76337, "tid": -914061504, "ts": 1716454222511524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511527, "dur": 0,
    "args": {
      "External id": 73838,
      "cbid": 251, "correlation": 73838
    }
  },
  {
    "ph": "f", "id": 73838, "pid": 76337, "tid": -914061504, "ts": 1716454222511527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222554867, "dur": 6,
    "args": {
      "External id": 73839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73839,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 73839, "pid": 5, "tid": 7, "ts": 1716454222554867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511529, "dur": 11,
    "args": {
      "External id": 73839,
      "cbid": 211, "correlation": 73839
    }
  },
  {
    "ph": "s", "id": 73839, "pid": 76337, "tid": -914061504, "ts": 1716454222511529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222554874, "dur": 4,
    "args": {
      "External id": 73841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73841,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 73841, "pid": 5, "tid": 7, "ts": 1716454222554874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511542, "dur": 6,
    "args": {
      "External id": 73841,
      "cbid": 211, "correlation": 73841
    }
  },
  {
    "ph": "s", "id": 73841, "pid": 76337, "tid": -914061504, "ts": 1716454222511542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222554879, "dur": 158,
    "args": {
      "External id": 73862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73862,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73862, "pid": 5, "tid": 7, "ts": 1716454222554879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511617, "dur": 12,
    "args": {
      "External id": 73862,
      "cbid": 211, "correlation": 73862
    }
  },
  {
    "ph": "s", "id": 73862, "pid": 76337, "tid": -914061504, "ts": 1716454222511617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511712, "dur": 1,
    "args": {
      "External id": 73880,
      "cbid": 251, "correlation": 73880
    }
  },
  {
    "ph": "f", "id": 73880, "pid": 76337, "tid": -914061504, "ts": 1716454222511712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222555039, "dur": 106,
    "args": {
      "External id": 73882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73882,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 73882, "pid": 5, "tid": 7, "ts": 1716454222555039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511718, "dur": 13,
    "args": {
      "External id": 73882,
      "cbid": 211, "correlation": 73882
    }
  },
  {
    "ph": "s", "id": 73882, "pid": 76337, "tid": -914061504, "ts": 1716454222511718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222555146, "dur": 34,
    "args": {
      "External id": 73890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73890,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73890, "pid": 5, "tid": 7, "ts": 1716454222555146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511787, "dur": 12,
    "args": {
      "External id": 73890,
      "cbid": 211, "correlation": 73890
    }
  },
  {
    "ph": "s", "id": 73890, "pid": 76337, "tid": -914061504, "ts": 1716454222511787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222555182, "dur": 68,
    "args": {
      "External id": 73898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73898, "pid": 5, "tid": 7, "ts": 1716454222555182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511829, "dur": 9,
    "args": {
      "External id": 73898,
      "cbid": 211, "correlation": 73898
    }
  },
  {
    "ph": "s", "id": 73898, "pid": 76337, "tid": -914061504, "ts": 1716454222511829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222555251, "dur": 94,
    "args": {
      "External id": 73920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73920,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73920, "pid": 5, "tid": 7, "ts": 1716454222555251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511881, "dur": 10,
    "args": {
      "External id": 73920,
      "cbid": 211, "correlation": 73920
    }
  },
  {
    "ph": "s", "id": 73920, "pid": 76337, "tid": -914061504, "ts": 1716454222511881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511967, "dur": 1,
    "args": {
      "External id": 73936,
      "cbid": 251, "correlation": 73936
    }
  },
  {
    "ph": "f", "id": 73936, "pid": 76337, "tid": -914061504, "ts": 1716454222511967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222555346, "dur": 588,
    "args": {
      "External id": 73938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73938,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 73938, "pid": 5, "tid": 7, "ts": 1716454222555346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222511980, "dur": 12,
    "args": {
      "External id": 73938,
      "cbid": 211, "correlation": 73938
    }
  },
  {
    "ph": "s", "id": 73938, "pid": 76337, "tid": -914061504, "ts": 1716454222511980,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222555936, "dur": 246,
    "args": {
      "External id": 73946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73946, "pid": 5, "tid": 7, "ts": 1716454222555936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512049, "dur": 12,
    "args": {
      "External id": 73946,
      "cbid": 211, "correlation": 73946
    }
  },
  {
    "ph": "s", "id": 73946, "pid": 76337, "tid": -914061504, "ts": 1716454222512049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222556183, "dur": 253,
    "args": {
      "External id": 73954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73954, "pid": 5, "tid": 7, "ts": 1716454222556183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512081, "dur": 8,
    "args": {
      "External id": 73954,
      "cbid": 211, "correlation": 73954
    }
  },
  {
    "ph": "s", "id": 73954, "pid": 76337, "tid": -914061504, "ts": 1716454222512081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512163, "dur": 2,
    "args": {
      "External id": 73970,
      "cbid": 251, "correlation": 73970
    }
  },
  {
    "ph": "f", "id": 73970, "pid": 76337, "tid": -914061504, "ts": 1716454222512163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512168, "dur": 0,
    "args": {
      "External id": 73972,
      "cbid": 251, "correlation": 73972
    }
  },
  {
    "ph": "f", "id": 73972, "pid": 76337, "tid": -914061504, "ts": 1716454222512168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222556437, "dur": 362,
    "args": {
      "External id": 73973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73973,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 73973, "pid": 5, "tid": 7, "ts": 1716454222556437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512172, "dur": 13,
    "args": {
      "External id": 73973,
      "cbid": 211, "correlation": 73973
    }
  },
  {
    "ph": "s", "id": 73973, "pid": 76337, "tid": -914061504, "ts": 1716454222512172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222556800, "dur": 50,
    "args": {
      "External id": 73981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73981,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73981, "pid": 5, "tid": 7, "ts": 1716454222556800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512214, "dur": 10,
    "args": {
      "External id": 73981,
      "cbid": 211, "correlation": 73981
    }
  },
  {
    "ph": "s", "id": 73981, "pid": 76337, "tid": -914061504, "ts": 1716454222512214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222556851, "dur": 161,
    "args": {
      "External id": 73992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 73992,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 73992, "pid": 5, "tid": 7, "ts": 1716454222556851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512284, "dur": 12,
    "args": {
      "External id": 73992,
      "cbid": 211, "correlation": 73992
    }
  },
  {
    "ph": "s", "id": 73992, "pid": 76337, "tid": -914061504, "ts": 1716454222512284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512348, "dur": 0,
    "args": {
      "External id": 74004,
      "cbid": 317, "correlation": 74004
    }
  },
  {
    "ph": "f", "id": 74004, "pid": 76337, "tid": -914061504, "ts": 1716454222512348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512349, "dur": 0,
    "args": {
      "External id": 74005,
      "cbid": 203, "correlation": 74005
    }
  },
  {
    "ph": "f", "id": 74005, "pid": 76337, "tid": -914061504, "ts": 1716454222512349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512350, "dur": 0,
    "args": {
      "External id": 74006,
      "cbid": 205, "correlation": 74006
    }
  },
  {
    "ph": "f", "id": 74006, "pid": 76337, "tid": -914061504, "ts": 1716454222512350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512376, "dur": 1,
    "args": {
      "External id": 74010,
      "cbid": 251, "correlation": 74010
    }
  },
  {
    "ph": "f", "id": 74010, "pid": 76337, "tid": -914061504, "ts": 1716454222512376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512377, "dur": 0,
    "args": {
      "External id": 74011,
      "cbid": 251, "correlation": 74011
    }
  },
  {
    "ph": "f", "id": 74011, "pid": 76337, "tid": -914061504, "ts": 1716454222512377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512378, "dur": 0,
    "args": {
      "External id": 74012,
      "cbid": 251, "correlation": 74012
    }
  },
  {
    "ph": "f", "id": 74012, "pid": 76337, "tid": -914061504, "ts": 1716454222512378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512379, "dur": 0,
    "args": {
      "External id": 74013,
      "cbid": 251, "correlation": 74013
    }
  },
  {
    "ph": "f", "id": 74013, "pid": 76337, "tid": -914061504, "ts": 1716454222512379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512380, "dur": 0,
    "args": {
      "External id": 74014,
      "cbid": 251, "correlation": 74014
    }
  },
  {
    "ph": "f", "id": 74014, "pid": 76337, "tid": -914061504, "ts": 1716454222512380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512380, "dur": 0,
    "args": {
      "External id": 74015,
      "cbid": 251, "correlation": 74015
    }
  },
  {
    "ph": "f", "id": 74015, "pid": 76337, "tid": -914061504, "ts": 1716454222512380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512381, "dur": 0,
    "args": {
      "External id": 74016,
      "cbid": 251, "correlation": 74016
    }
  },
  {
    "ph": "f", "id": 74016, "pid": 76337, "tid": -914061504, "ts": 1716454222512381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512382, "dur": 0,
    "args": {
      "External id": 74017,
      "cbid": 251, "correlation": 74017
    }
  },
  {
    "ph": "f", "id": 74017, "pid": 76337, "tid": -914061504, "ts": 1716454222512382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512383, "dur": 0,
    "args": {
      "External id": 74018,
      "cbid": 251, "correlation": 74018
    }
  },
  {
    "ph": "f", "id": 74018, "pid": 76337, "tid": -914061504, "ts": 1716454222512383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222557014, "dur": 117,
    "args": {
      "External id": 74019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74019,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 74019, "pid": 5, "tid": 7, "ts": 1716454222557014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512385, "dur": 12,
    "args": {
      "External id": 74019,
      "cbid": 211, "correlation": 74019
    }
  },
  {
    "ph": "s", "id": 74019, "pid": 76337, "tid": -914061504, "ts": 1716454222512385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222557132, "dur": 60,
    "args": {
      "External id": 74025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74025, "pid": 5, "tid": 7, "ts": 1716454222557132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512420, "dur": 9,
    "args": {
      "External id": 74025,
      "cbid": 211, "correlation": 74025
    }
  },
  {
    "ph": "s", "id": 74025, "pid": 76337, "tid": -914061504, "ts": 1716454222512420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222557194, "dur": 49,
    "args": {
      "External id": 74033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74033,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74033, "pid": 5, "tid": 7, "ts": 1716454222557194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512452, "dur": 8,
    "args": {
      "External id": 74033,
      "cbid": 211, "correlation": 74033
    }
  },
  {
    "ph": "s", "id": 74033, "pid": 76337, "tid": -914061504, "ts": 1716454222512452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512526, "dur": 0,
    "args": {
      "External id": 74043,
      "cbid": 317, "correlation": 74043
    }
  },
  {
    "ph": "f", "id": 74043, "pid": 76337, "tid": -914061504, "ts": 1716454222512526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512526, "dur": 0,
    "args": {
      "External id": 74044,
      "cbid": 203, "correlation": 74044
    }
  },
  {
    "ph": "f", "id": 74044, "pid": 76337, "tid": -914061504, "ts": 1716454222512526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512527, "dur": 0,
    "args": {
      "External id": 74045,
      "cbid": 205, "correlation": 74045
    }
  },
  {
    "ph": "f", "id": 74045, "pid": 76337, "tid": -914061504, "ts": 1716454222512527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222557244, "dur": 41,
    "args": {
      "External id": 74049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74049,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74049, "pid": 5, "tid": 7, "ts": 1716454222557244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512544, "dur": 12,
    "args": {
      "External id": 74049,
      "cbid": 211, "correlation": 74049
    }
  },
  {
    "ph": "s", "id": 74049, "pid": 76337, "tid": -914061504, "ts": 1716454222512544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222557287, "dur": 14,
    "args": {
      "External id": 74051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74051,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74051, "pid": 5, "tid": 7, "ts": 1716454222557287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512559, "dur": 5,
    "args": {
      "External id": 74051,
      "cbid": 211, "correlation": 74051
    }
  },
  {
    "ph": "s", "id": 74051, "pid": 76337, "tid": -914061504, "ts": 1716454222512559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222557303, "dur": 1,
    "args": {
      "External id": 74053,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 74053,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.8894035900405327
    }
  },
  {
    "ph": "f", "id": 74053, "pid": 5, "tid": 7, "ts": 1716454222557303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512601, "dur": 18,
    "args": {
      "External id": 74053,
      "cbid": 51, "correlation": 74053
    }
  },
  {
    "ph": "s", "id": 74053, "pid": 76337, "tid": -914061504, "ts": 1716454222512601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222557307, "dur": 367,
    "args": {
      "External id": 74054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74054,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74054, "pid": 5, "tid": 7, "ts": 1716454222557307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512620, "dur": 10,
    "args": {
      "External id": 74054,
      "cbid": 211, "correlation": 74054
    }
  },
  {
    "ph": "s", "id": 74054, "pid": 76337, "tid": -914061504, "ts": 1716454222512620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222557676, "dur": 14,
    "args": {
      "External id": 74056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74056, "pid": 5, "tid": 7, "ts": 1716454222557676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512638, "dur": 7,
    "args": {
      "External id": 74056,
      "cbid": 211, "correlation": 74056
    }
  },
  {
    "ph": "s", "id": 74056, "pid": 76337, "tid": -914061504, "ts": 1716454222512638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222557691, "dur": 16,
    "args": {
      "External id": 74062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74062, "pid": 5, "tid": 7, "ts": 1716454222557691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512669, "dur": 9,
    "args": {
      "External id": 74062,
      "cbid": 211, "correlation": 74062
    }
  },
  {
    "ph": "s", "id": 74062, "pid": 76337, "tid": -914061504, "ts": 1716454222512669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222557708, "dur": 19,
    "args": {
      "External id": 74082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74082,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 74082, "pid": 5, "tid": 7, "ts": 1716454222557708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512765, "dur": 13,
    "args": {
      "External id": 74082,
      "cbid": 211, "correlation": 74082
    }
  },
  {
    "ph": "s", "id": 74082, "pid": 76337, "tid": -914061504, "ts": 1716454222512765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222557728, "dur": 4,
    "args": {
      "External id": 74094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74094,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 74094, "pid": 5, "tid": 7, "ts": 1716454222557728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512788, "dur": 6,
    "args": {
      "External id": 74094,
      "cbid": 211, "correlation": 74094
    }
  },
  {
    "ph": "s", "id": 74094, "pid": 76337, "tid": -914061504, "ts": 1716454222512788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222557734, "dur": 18,
    "args": {
      "External id": 74097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74097, "pid": 5, "tid": 7, "ts": 1716454222557734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512806, "dur": 6,
    "args": {
      "External id": 74097,
      "cbid": 211, "correlation": 74097
    }
  },
  {
    "ph": "s", "id": 74097, "pid": 76337, "tid": -914061504, "ts": 1716454222512806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222557753, "dur": 12,
    "args": {
      "External id": 74106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74106,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74106, "pid": 5, "tid": 7, "ts": 1716454222557753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512845, "dur": 9,
    "args": {
      "External id": 74106,
      "cbid": 211, "correlation": 74106
    }
  },
  {
    "ph": "s", "id": 74106, "pid": 76337, "tid": -914061504, "ts": 1716454222512845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512900, "dur": 0,
    "args": {
      "External id": 74116,
      "cbid": 317, "correlation": 74116
    }
  },
  {
    "ph": "f", "id": 74116, "pid": 76337, "tid": -914061504, "ts": 1716454222512900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512901, "dur": 0,
    "args": {
      "External id": 74117,
      "cbid": 203, "correlation": 74117
    }
  },
  {
    "ph": "f", "id": 74117, "pid": 76337, "tid": -914061504, "ts": 1716454222512901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512902, "dur": 0,
    "args": {
      "External id": 74118,
      "cbid": 205, "correlation": 74118
    }
  },
  {
    "ph": "f", "id": 74118, "pid": 76337, "tid": -914061504, "ts": 1716454222512902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222557766, "dur": 12,
    "args": {
      "External id": 74122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74122, "pid": 5, "tid": 7, "ts": 1716454222557766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512917, "dur": 12,
    "args": {
      "External id": 74122,
      "cbid": 211, "correlation": 74122
    }
  },
  {
    "ph": "s", "id": 74122, "pid": 76337, "tid": -914061504, "ts": 1716454222512917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222557779, "dur": 25,
    "args": {
      "External id": 74124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74124, "pid": 5, "tid": 7, "ts": 1716454222557779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512932, "dur": 5,
    "args": {
      "External id": 74124,
      "cbid": 211, "correlation": 74124
    }
  },
  {
    "ph": "s", "id": 74124, "pid": 76337, "tid": -914061504, "ts": 1716454222512932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222557805, "dur": 4,
    "args": {
      "External id": 74126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74126,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 74126, "pid": 5, "tid": 7, "ts": 1716454222557805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512943, "dur": 6,
    "args": {
      "External id": 74126,
      "cbid": 211, "correlation": 74126
    }
  },
  {
    "ph": "s", "id": 74126, "pid": 76337, "tid": -914061504, "ts": 1716454222512943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512965, "dur": 0,
    "args": {
      "External id": 74127,
      "cbid": 51, "correlation": 74127
    }
  },
  {
    "ph": "s", "id": 74127, "pid": 76337, "tid": -914061504, "ts": 1716454222512965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222557810, "dur": 362,
    "args": {
      "External id": 74128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74128,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74128, "pid": 5, "tid": 7, "ts": 1716454222557810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512966, "dur": 22,
    "args": {
      "External id": 74128,
      "cbid": 211, "correlation": 74128
    }
  },
  {
    "ph": "s", "id": 74128, "pid": 76337, "tid": -914061504, "ts": 1716454222512966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222558174, "dur": 21,
    "args": {
      "External id": 74129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74129, "pid": 5, "tid": 7, "ts": 1716454222558174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222512992, "dur": 6,
    "args": {
      "External id": 74129,
      "cbid": 211, "correlation": 74129
    }
  },
  {
    "ph": "s", "id": 74129, "pid": 76337, "tid": -914061504, "ts": 1716454222512992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222558197, "dur": 33,
    "args": {
      "External id": 74135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74135, "pid": 5, "tid": 7, "ts": 1716454222558197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513022, "dur": 8,
    "args": {
      "External id": 74135,
      "cbid": 211, "correlation": 74135
    }
  },
  {
    "ph": "s", "id": 74135, "pid": 76337, "tid": -914061504, "ts": 1716454222513022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222558231, "dur": 4,
    "args": {
      "External id": 74143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74143,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 74143, "pid": 5, "tid": 7, "ts": 1716454222558231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513067, "dur": 9,
    "args": {
      "External id": 74143,
      "cbid": 211, "correlation": 74143
    }
  },
  {
    "ph": "s", "id": 74143, "pid": 76337, "tid": -914061504, "ts": 1716454222513067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513134, "dur": 1,
    "args": {
      "External id": 74159,
      "cbid": 251, "correlation": 74159
    }
  },
  {
    "ph": "f", "id": 74159, "pid": 76337, "tid": -914061504, "ts": 1716454222513134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513139, "dur": 0,
    "args": {
      "External id": 74161,
      "cbid": 251, "correlation": 74161
    }
  },
  {
    "ph": "f", "id": 74161, "pid": 76337, "tid": -914061504, "ts": 1716454222513139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222558236, "dur": 12,
    "args": {
      "External id": 74162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74162,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 74162, "pid": 5, "tid": 7, "ts": 1716454222558236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513141, "dur": 12,
    "args": {
      "External id": 74162,
      "cbid": 211, "correlation": 74162
    }
  },
  {
    "ph": "s", "id": 74162, "pid": 76337, "tid": -914061504, "ts": 1716454222513141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222558249, "dur": 5,
    "args": {
      "External id": 74164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74164,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 74164, "pid": 5, "tid": 7, "ts": 1716454222558249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513155, "dur": 5,
    "args": {
      "External id": 74164,
      "cbid": 211, "correlation": 74164
    }
  },
  {
    "ph": "s", "id": 74164, "pid": 76337, "tid": -914061504, "ts": 1716454222513155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222558255, "dur": 29,
    "args": {
      "External id": 74174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74174, "pid": 5, "tid": 7, "ts": 1716454222558255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513214, "dur": 12,
    "args": {
      "External id": 74174,
      "cbid": 211, "correlation": 74174
    }
  },
  {
    "ph": "s", "id": 74174, "pid": 76337, "tid": -914061504, "ts": 1716454222513214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222558286, "dur": 30,
    "args": {
      "External id": 74194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74194,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 74194, "pid": 5, "tid": 7, "ts": 1716454222558286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513281, "dur": 10,
    "args": {
      "External id": 74194,
      "cbid": 211, "correlation": 74194
    }
  },
  {
    "ph": "s", "id": 74194, "pid": 76337, "tid": -914061504, "ts": 1716454222513281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222558317, "dur": 4,
    "args": {
      "External id": 74206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74206,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 74206, "pid": 5, "tid": 7, "ts": 1716454222558317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513301, "dur": 6,
    "args": {
      "External id": 74206,
      "cbid": 211, "correlation": 74206
    }
  },
  {
    "ph": "s", "id": 74206, "pid": 76337, "tid": -914061504, "ts": 1716454222513301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222558322, "dur": 30,
    "args": {
      "External id": 74209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74209, "pid": 5, "tid": 7, "ts": 1716454222558322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513319, "dur": 7,
    "args": {
      "External id": 74209,
      "cbid": 211, "correlation": 74209
    }
  },
  {
    "ph": "s", "id": 74209, "pid": 76337, "tid": -914061504, "ts": 1716454222513319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222558354, "dur": 20,
    "args": {
      "External id": 74218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74218,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74218, "pid": 5, "tid": 7, "ts": 1716454222558354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513359, "dur": 10,
    "args": {
      "External id": 74218,
      "cbid": 211, "correlation": 74218
    }
  },
  {
    "ph": "s", "id": 74218, "pid": 76337, "tid": -914061504, "ts": 1716454222513359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513423, "dur": 0,
    "args": {
      "External id": 74228,
      "cbid": 317, "correlation": 74228
    }
  },
  {
    "ph": "f", "id": 74228, "pid": 76337, "tid": -914061504, "ts": 1716454222513423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513424, "dur": 0,
    "args": {
      "External id": 74229,
      "cbid": 203, "correlation": 74229
    }
  },
  {
    "ph": "f", "id": 74229, "pid": 76337, "tid": -914061504, "ts": 1716454222513424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513425, "dur": 0,
    "args": {
      "External id": 74230,
      "cbid": 205, "correlation": 74230
    }
  },
  {
    "ph": "f", "id": 74230, "pid": 76337, "tid": -914061504, "ts": 1716454222513425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222558376, "dur": 23,
    "args": {
      "External id": 74234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74234, "pid": 5, "tid": 7, "ts": 1716454222558376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513442, "dur": 12,
    "args": {
      "External id": 74234,
      "cbid": 211, "correlation": 74234
    }
  },
  {
    "ph": "s", "id": 74234, "pid": 76337, "tid": -914061504, "ts": 1716454222513442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222558400, "dur": 45,
    "args": {
      "External id": 74236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74236, "pid": 5, "tid": 7, "ts": 1716454222558400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513457, "dur": 5,
    "args": {
      "External id": 74236,
      "cbid": 211, "correlation": 74236
    }
  },
  {
    "ph": "s", "id": 74236, "pid": 76337, "tid": -914061504, "ts": 1716454222513457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222558447, "dur": 664,
    "args": {
      "External id": 74238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74238,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74238, "pid": 5, "tid": 7, "ts": 1716454222558447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513471, "dur": 10,
    "args": {
      "External id": 74238,
      "cbid": 211, "correlation": 74238
    }
  },
  {
    "ph": "s", "id": 74238, "pid": 76337, "tid": -914061504, "ts": 1716454222513471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222559112, "dur": 22,
    "args": {
      "External id": 74240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74240, "pid": 5, "tid": 7, "ts": 1716454222559112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513485, "dur": 5,
    "args": {
      "External id": 74240,
      "cbid": 211, "correlation": 74240
    }
  },
  {
    "ph": "s", "id": 74240, "pid": 76337, "tid": -914061504, "ts": 1716454222513485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222559136, "dur": 33,
    "args": {
      "External id": 74246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74246, "pid": 5, "tid": 7, "ts": 1716454222559136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513512, "dur": 8,
    "args": {
      "External id": 74246,
      "cbid": 211, "correlation": 74246
    }
  },
  {
    "ph": "s", "id": 74246, "pid": 76337, "tid": -914061504, "ts": 1716454222513512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513570, "dur": 0,
    "args": {
      "External id": 74256,
      "cbid": 317, "correlation": 74256
    }
  },
  {
    "ph": "f", "id": 74256, "pid": 76337, "tid": -914061504, "ts": 1716454222513570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513571, "dur": 0,
    "args": {
      "External id": 74257,
      "cbid": 203, "correlation": 74257
    }
  },
  {
    "ph": "f", "id": 74257, "pid": 76337, "tid": -914061504, "ts": 1716454222513571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513571, "dur": 0,
    "args": {
      "External id": 74258,
      "cbid": 205, "correlation": 74258
    }
  },
  {
    "ph": "f", "id": 74258, "pid": 76337, "tid": -914061504, "ts": 1716454222513571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513594, "dur": 1,
    "args": {
      "External id": 74262,
      "cbid": 251, "correlation": 74262
    }
  },
  {
    "ph": "f", "id": 74262, "pid": 76337, "tid": -914061504, "ts": 1716454222513594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513596, "dur": 0,
    "args": {
      "External id": 74263,
      "cbid": 251, "correlation": 74263
    }
  },
  {
    "ph": "f", "id": 74263, "pid": 76337, "tid": -914061504, "ts": 1716454222513596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513596, "dur": 0,
    "args": {
      "External id": 74264,
      "cbid": 251, "correlation": 74264
    }
  },
  {
    "ph": "f", "id": 74264, "pid": 76337, "tid": -914061504, "ts": 1716454222513596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513597, "dur": 0,
    "args": {
      "External id": 74265,
      "cbid": 251, "correlation": 74265
    }
  },
  {
    "ph": "f", "id": 74265, "pid": 76337, "tid": -914061504, "ts": 1716454222513597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513598, "dur": 0,
    "args": {
      "External id": 74266,
      "cbid": 251, "correlation": 74266
    }
  },
  {
    "ph": "f", "id": 74266, "pid": 76337, "tid": -914061504, "ts": 1716454222513598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513599, "dur": 0,
    "args": {
      "External id": 74267,
      "cbid": 251, "correlation": 74267
    }
  },
  {
    "ph": "f", "id": 74267, "pid": 76337, "tid": -914061504, "ts": 1716454222513599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513600, "dur": 0,
    "args": {
      "External id": 74268,
      "cbid": 251, "correlation": 74268
    }
  },
  {
    "ph": "f", "id": 74268, "pid": 76337, "tid": -914061504, "ts": 1716454222513600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513600, "dur": 0,
    "args": {
      "External id": 74269,
      "cbid": 251, "correlation": 74269
    }
  },
  {
    "ph": "f", "id": 74269, "pid": 76337, "tid": -914061504, "ts": 1716454222513600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513602, "dur": 0,
    "args": {
      "External id": 74270,
      "cbid": 251, "correlation": 74270
    }
  },
  {
    "ph": "f", "id": 74270, "pid": 76337, "tid": -914061504, "ts": 1716454222513602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222559170, "dur": 51,
    "args": {
      "External id": 74271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74271,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 74271, "pid": 5, "tid": 7, "ts": 1716454222559170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513604, "dur": 12,
    "args": {
      "External id": 74271,
      "cbid": 211, "correlation": 74271
    }
  },
  {
    "ph": "s", "id": 74271, "pid": 76337, "tid": -914061504, "ts": 1716454222513604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222559223, "dur": 33,
    "args": {
      "External id": 74277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74277, "pid": 5, "tid": 7, "ts": 1716454222559223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513635, "dur": 9,
    "args": {
      "External id": 74277,
      "cbid": 211, "correlation": 74277
    }
  },
  {
    "ph": "s", "id": 74277, "pid": 76337, "tid": -914061504, "ts": 1716454222513635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222559257, "dur": 27,
    "args": {
      "External id": 74285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74285,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74285, "pid": 5, "tid": 7, "ts": 1716454222559257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513665, "dur": 8,
    "args": {
      "External id": 74285,
      "cbid": 211, "correlation": 74285
    }
  },
  {
    "ph": "s", "id": 74285, "pid": 76337, "tid": -914061504, "ts": 1716454222513665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222559285, "dur": 19,
    "args": {
      "External id": 74293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74293,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74293, "pid": 5, "tid": 7, "ts": 1716454222559285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513694, "dur": 9,
    "args": {
      "External id": 74293,
      "cbid": 211, "correlation": 74293
    }
  },
  {
    "ph": "s", "id": 74293, "pid": 76337, "tid": -914061504, "ts": 1716454222513694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222559306, "dur": 31,
    "args": {
      "External id": 74313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74313,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 74313, "pid": 5, "tid": 7, "ts": 1716454222559306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513778, "dur": 12,
    "args": {
      "External id": 74313,
      "cbid": 211, "correlation": 74313
    }
  },
  {
    "ph": "s", "id": 74313, "pid": 76337, "tid": -914061504, "ts": 1716454222513778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222559338, "dur": 4,
    "args": {
      "External id": 74325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74325,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 74325, "pid": 5, "tid": 7, "ts": 1716454222559338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513800, "dur": 6,
    "args": {
      "External id": 74325,
      "cbid": 211, "correlation": 74325
    }
  },
  {
    "ph": "s", "id": 74325, "pid": 76337, "tid": -914061504, "ts": 1716454222513800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222559343, "dur": 30,
    "args": {
      "External id": 74328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74328, "pid": 5, "tid": 7, "ts": 1716454222559343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513817, "dur": 7,
    "args": {
      "External id": 74328,
      "cbid": 211, "correlation": 74328
    }
  },
  {
    "ph": "s", "id": 74328, "pid": 76337, "tid": -914061504, "ts": 1716454222513817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513875, "dur": 0,
    "args": {
      "External id": 74339,
      "cbid": 317, "correlation": 74339
    }
  },
  {
    "ph": "f", "id": 74339, "pid": 76337, "tid": -914061504, "ts": 1716454222513875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513876, "dur": 0,
    "args": {
      "External id": 74340,
      "cbid": 203, "correlation": 74340
    }
  },
  {
    "ph": "f", "id": 74340, "pid": 76337, "tid": -914061504, "ts": 1716454222513876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513877, "dur": 0,
    "args": {
      "External id": 74341,
      "cbid": 205, "correlation": 74341
    }
  },
  {
    "ph": "f", "id": 74341, "pid": 76337, "tid": -914061504, "ts": 1716454222513877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222559375, "dur": 23,
    "args": {
      "External id": 74345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74345,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74345, "pid": 5, "tid": 7, "ts": 1716454222559375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513892, "dur": 12,
    "args": {
      "External id": 74345,
      "cbid": 211, "correlation": 74345
    }
  },
  {
    "ph": "s", "id": 74345, "pid": 76337, "tid": -914061504, "ts": 1716454222513892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222559400, "dur": 123,
    "args": {
      "External id": 74347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74347,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74347, "pid": 5, "tid": 7, "ts": 1716454222559400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513912, "dur": 8,
    "args": {
      "External id": 74347,
      "cbid": 211, "correlation": 74347
    }
  },
  {
    "ph": "s", "id": 74347, "pid": 76337, "tid": -914061504, "ts": 1716454222513912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222559525, "dur": 22,
    "args": {
      "External id": 74349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74349, "pid": 5, "tid": 7, "ts": 1716454222559525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513924, "dur": 5,
    "args": {
      "External id": 74349,
      "cbid": 211, "correlation": 74349
    }
  },
  {
    "ph": "s", "id": 74349, "pid": 76337, "tid": -914061504, "ts": 1716454222513924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222559548, "dur": 33,
    "args": {
      "External id": 74355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74355, "pid": 5, "tid": 7, "ts": 1716454222559548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222513952, "dur": 8,
    "args": {
      "External id": 74355,
      "cbid": 211, "correlation": 74355
    }
  },
  {
    "ph": "s", "id": 74355, "pid": 76337, "tid": -914061504, "ts": 1716454222513952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222559581, "dur": 174,
    "args": {
      "External id": 74364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74364,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74364, "pid": 5, "tid": 7, "ts": 1716454222559581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514046, "dur": 15,
    "args": {
      "External id": 74364,
      "cbid": 211, "correlation": 74364
    }
  },
  {
    "ph": "s", "id": 74364, "pid": 76337, "tid": -914061504, "ts": 1716454222514046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222559757, "dur": 66,
    "args": {
      "External id": 74386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74386,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74386, "pid": 5, "tid": 7, "ts": 1716454222559757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514104, "dur": 10,
    "args": {
      "External id": 74386,
      "cbid": 211, "correlation": 74386
    }
  },
  {
    "ph": "s", "id": 74386, "pid": 76337, "tid": -914061504, "ts": 1716454222514104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514194, "dur": 1,
    "args": {
      "External id": 74397,
      "cbid": 251, "correlation": 74397
    }
  },
  {
    "ph": "f", "id": 74397, "pid": 76337, "tid": -914061504, "ts": 1716454222514194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222559824, "dur": 157,
    "args": {
      "External id": 74398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74398,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74398, "pid": 5, "tid": 7, "ts": 1716454222559824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514199, "dur": 12,
    "args": {
      "External id": 74398,
      "cbid": 211, "correlation": 74398
    }
  },
  {
    "ph": "s", "id": 74398, "pid": 76337, "tid": -914061504, "ts": 1716454222514199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514268, "dur": 1,
    "args": {
      "External id": 74409,
      "cbid": 251, "correlation": 74409
    }
  },
  {
    "ph": "f", "id": 74409, "pid": 76337, "tid": -914061504, "ts": 1716454222514268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222559982, "dur": 151,
    "args": {
      "External id": 74410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74410,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74410, "pid": 5, "tid": 7, "ts": 1716454222559982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514272, "dur": 12,
    "args": {
      "External id": 74410,
      "cbid": 211, "correlation": 74410
    }
  },
  {
    "ph": "s", "id": 74410, "pid": 76337, "tid": -914061504, "ts": 1716454222514272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514339, "dur": 1,
    "args": {
      "External id": 74421,
      "cbid": 251, "correlation": 74421
    }
  },
  {
    "ph": "f", "id": 74421, "pid": 76337, "tid": -914061504, "ts": 1716454222514339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222560134, "dur": 147,
    "args": {
      "External id": 74422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74422,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74422, "pid": 5, "tid": 7, "ts": 1716454222560134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514343, "dur": 11,
    "args": {
      "External id": 74422,
      "cbid": 211, "correlation": 74422
    }
  },
  {
    "ph": "s", "id": 74422, "pid": 76337, "tid": -914061504, "ts": 1716454222514343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222560283, "dur": 1990,
    "args": {
      "External id": 74443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74443,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 74443, "pid": 5, "tid": 7, "ts": 1716454222560283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514425, "dur": 13,
    "args": {
      "External id": 74443,
      "cbid": 211, "correlation": 74443
    }
  },
  {
    "ph": "s", "id": 74443, "pid": 76337, "tid": -914061504, "ts": 1716454222514425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514526, "dur": 1,
    "args": {
      "External id": 74461,
      "cbid": 251, "correlation": 74461
    }
  },
  {
    "ph": "f", "id": 74461, "pid": 76337, "tid": -914061504, "ts": 1716454222514526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222562275, "dur": 149,
    "args": {
      "External id": 74463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74463,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 74463, "pid": 5, "tid": 7, "ts": 1716454222562275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514532, "dur": 13,
    "args": {
      "External id": 74463,
      "cbid": 211, "correlation": 74463
    }
  },
  {
    "ph": "s", "id": 74463, "pid": 76337, "tid": -914061504, "ts": 1716454222514532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222562425, "dur": 35,
    "args": {
      "External id": 74471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74471,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74471, "pid": 5, "tid": 7, "ts": 1716454222562425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514603, "dur": 12,
    "args": {
      "External id": 74471,
      "cbid": 211, "correlation": 74471
    }
  },
  {
    "ph": "s", "id": 74471, "pid": 76337, "tid": -914061504, "ts": 1716454222514603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222562462, "dur": 51,
    "args": {
      "External id": 74479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74479,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74479, "pid": 5, "tid": 7, "ts": 1716454222562462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514642, "dur": 9,
    "args": {
      "External id": 74479,
      "cbid": 211, "correlation": 74479
    }
  },
  {
    "ph": "s", "id": 74479, "pid": 76337, "tid": -914061504, "ts": 1716454222514642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222562514, "dur": 30,
    "args": {
      "External id": 74490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74490, "pid": 5, "tid": 7, "ts": 1716454222562514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514715, "dur": 13,
    "args": {
      "External id": 74490,
      "cbid": 211, "correlation": 74490
    }
  },
  {
    "ph": "s", "id": 74490, "pid": 76337, "tid": -914061504, "ts": 1716454222514715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222562545, "dur": 35,
    "args": {
      "External id": 74512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74512,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74512, "pid": 5, "tid": 7, "ts": 1716454222562545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514746, "dur": 8,
    "args": {
      "External id": 74512,
      "cbid": 211, "correlation": 74512
    }
  },
  {
    "ph": "s", "id": 74512, "pid": 76337, "tid": -914061504, "ts": 1716454222514746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514830, "dur": 1,
    "args": {
      "External id": 74523,
      "cbid": 251, "correlation": 74523
    }
  },
  {
    "ph": "f", "id": 74523, "pid": 76337, "tid": -914061504, "ts": 1716454222514830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222562582, "dur": 91,
    "args": {
      "External id": 74524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74524,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74524, "pid": 5, "tid": 7, "ts": 1716454222562582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514836, "dur": 12,
    "args": {
      "External id": 74524,
      "cbid": 211, "correlation": 74524
    }
  },
  {
    "ph": "s", "id": 74524, "pid": 76337, "tid": -914061504, "ts": 1716454222514836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514905, "dur": 1,
    "args": {
      "External id": 74535,
      "cbid": 251, "correlation": 74535
    }
  },
  {
    "ph": "f", "id": 74535, "pid": 76337, "tid": -914061504, "ts": 1716454222514905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514908, "dur": 0,
    "args": {
      "External id": 74536,
      "cbid": 251, "correlation": 74536
    }
  },
  {
    "ph": "f", "id": 74536, "pid": 76337, "tid": -914061504, "ts": 1716454222514908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222562674, "dur": 11,
    "args": {
      "External id": 74537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74537,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 74537, "pid": 5, "tid": 7, "ts": 1716454222562674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514910, "dur": 12,
    "args": {
      "External id": 74537,
      "cbid": 211, "correlation": 74537
    }
  },
  {
    "ph": "s", "id": 74537, "pid": 76337, "tid": -914061504, "ts": 1716454222514910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222562686, "dur": 5,
    "args": {
      "External id": 74539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74539,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 74539, "pid": 5, "tid": 7, "ts": 1716454222562686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514925, "dur": 6,
    "args": {
      "External id": 74539,
      "cbid": 211, "correlation": 74539
    }
  },
  {
    "ph": "s", "id": 74539, "pid": 76337, "tid": -914061504, "ts": 1716454222514925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514991, "dur": 1,
    "args": {
      "External id": 74550,
      "cbid": 251, "correlation": 74550
    }
  },
  {
    "ph": "f", "id": 74550, "pid": 76337, "tid": -914061504, "ts": 1716454222514991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514995, "dur": 0,
    "args": {
      "External id": 74551,
      "cbid": 251, "correlation": 74551
    }
  },
  {
    "ph": "f", "id": 74551, "pid": 76337, "tid": -914061504, "ts": 1716454222514995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222562692, "dur": 7,
    "args": {
      "External id": 74552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74552,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 74552, "pid": 5, "tid": 7, "ts": 1716454222562692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222514996, "dur": 12,
    "args": {
      "External id": 74552,
      "cbid": 211, "correlation": 74552
    }
  },
  {
    "ph": "s", "id": 74552, "pid": 76337, "tid": -914061504, "ts": 1716454222514996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222562700, "dur": 3,
    "args": {
      "External id": 74554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74554,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 74554, "pid": 5, "tid": 7, "ts": 1716454222562700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515010, "dur": 5,
    "args": {
      "External id": 74554,
      "cbid": 211, "correlation": 74554
    }
  },
  {
    "ph": "s", "id": 74554, "pid": 76337, "tid": -914061504, "ts": 1716454222515010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222562705, "dur": 94,
    "args": {
      "External id": 74575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74575,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 74575, "pid": 5, "tid": 7, "ts": 1716454222562705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515085, "dur": 14,
    "args": {
      "External id": 74575,
      "cbid": 211, "correlation": 74575
    }
  },
  {
    "ph": "s", "id": 74575, "pid": 76337, "tid": -914061504, "ts": 1716454222515085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515181, "dur": 1,
    "args": {
      "External id": 74593,
      "cbid": 251, "correlation": 74593
    }
  },
  {
    "ph": "f", "id": 74593, "pid": 76337, "tid": -914061504, "ts": 1716454222515181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222562800, "dur": 99,
    "args": {
      "External id": 74595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74595,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74595, "pid": 5, "tid": 7, "ts": 1716454222562800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515187, "dur": 13,
    "args": {
      "External id": 74595,
      "cbid": 211, "correlation": 74595
    }
  },
  {
    "ph": "s", "id": 74595, "pid": 76337, "tid": -914061504, "ts": 1716454222515187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222562901, "dur": 19,
    "args": {
      "External id": 74603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74603,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74603, "pid": 5, "tid": 7, "ts": 1716454222562901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515256, "dur": 13,
    "args": {
      "External id": 74603,
      "cbid": 211, "correlation": 74603
    }
  },
  {
    "ph": "s", "id": 74603, "pid": 76337, "tid": -914061504, "ts": 1716454222515256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222562921, "dur": 37,
    "args": {
      "External id": 74611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74611, "pid": 5, "tid": 7, "ts": 1716454222562921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515298, "dur": 9,
    "args": {
      "External id": 74611,
      "cbid": 211, "correlation": 74611
    }
  },
  {
    "ph": "s", "id": 74611, "pid": 76337, "tid": -914061504, "ts": 1716454222515298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222562960, "dur": 35,
    "args": {
      "External id": 74633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74633,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74633, "pid": 5, "tid": 7, "ts": 1716454222562960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515351, "dur": 10,
    "args": {
      "External id": 74633,
      "cbid": 211, "correlation": 74633
    }
  },
  {
    "ph": "s", "id": 74633, "pid": 76337, "tid": -914061504, "ts": 1716454222515351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515440, "dur": 1,
    "args": {
      "External id": 74649,
      "cbid": 251, "correlation": 74649
    }
  },
  {
    "ph": "f", "id": 74649, "pid": 76337, "tid": -914061504, "ts": 1716454222515440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515446, "dur": 0,
    "args": {
      "External id": 74651,
      "cbid": 251, "correlation": 74651
    }
  },
  {
    "ph": "f", "id": 74651, "pid": 76337, "tid": -914061504, "ts": 1716454222515446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222562997, "dur": 551,
    "args": {
      "External id": 74652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74652,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 74652, "pid": 5, "tid": 7, "ts": 1716454222562997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515449, "dur": 12,
    "args": {
      "External id": 74652,
      "cbid": 211, "correlation": 74652
    }
  },
  {
    "ph": "s", "id": 74652, "pid": 76337, "tid": -914061504, "ts": 1716454222515449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222563549, "dur": 127,
    "args": {
      "External id": 74660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74660, "pid": 5, "tid": 7, "ts": 1716454222563549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515516, "dur": 13,
    "args": {
      "External id": 74660,
      "cbid": 211, "correlation": 74660
    }
  },
  {
    "ph": "s", "id": 74660, "pid": 76337, "tid": -914061504, "ts": 1716454222515516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222563677, "dur": 129,
    "args": {
      "External id": 74668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74668, "pid": 5, "tid": 7, "ts": 1716454222563677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515548, "dur": 8,
    "args": {
      "External id": 74668,
      "cbid": 211, "correlation": 74668
    }
  },
  {
    "ph": "s", "id": 74668, "pid": 76337, "tid": -914061504, "ts": 1716454222515548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515626, "dur": 1,
    "args": {
      "External id": 74684,
      "cbid": 251, "correlation": 74684
    }
  },
  {
    "ph": "f", "id": 74684, "pid": 76337, "tid": -914061504, "ts": 1716454222515626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222563807, "dur": 308,
    "args": {
      "External id": 74686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74686,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74686, "pid": 5, "tid": 7, "ts": 1716454222563807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515631, "dur": 13,
    "args": {
      "External id": 74686,
      "cbid": 211, "correlation": 74686
    }
  },
  {
    "ph": "s", "id": 74686, "pid": 76337, "tid": -914061504, "ts": 1716454222515631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222564117, "dur": 27,
    "args": {
      "External id": 74694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74694,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74694, "pid": 5, "tid": 7, "ts": 1716454222564117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515674, "dur": 10,
    "args": {
      "External id": 74694,
      "cbid": 211, "correlation": 74694
    }
  },
  {
    "ph": "s", "id": 74694, "pid": 76337, "tid": -914061504, "ts": 1716454222515674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222564145, "dur": 83,
    "args": {
      "External id": 74705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74705, "pid": 5, "tid": 7, "ts": 1716454222564145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515743, "dur": 12,
    "args": {
      "External id": 74705,
      "cbid": 211, "correlation": 74705
    }
  },
  {
    "ph": "s", "id": 74705, "pid": 76337, "tid": -914061504, "ts": 1716454222515743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515807, "dur": 0,
    "args": {
      "External id": 74717,
      "cbid": 317, "correlation": 74717
    }
  },
  {
    "ph": "f", "id": 74717, "pid": 76337, "tid": -914061504, "ts": 1716454222515807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515808, "dur": 0,
    "args": {
      "External id": 74718,
      "cbid": 203, "correlation": 74718
    }
  },
  {
    "ph": "f", "id": 74718, "pid": 76337, "tid": -914061504, "ts": 1716454222515808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515809, "dur": 0,
    "args": {
      "External id": 74719,
      "cbid": 205, "correlation": 74719
    }
  },
  {
    "ph": "f", "id": 74719, "pid": 76337, "tid": -914061504, "ts": 1716454222515809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222564230, "dur": 23,
    "args": {
      "External id": 74723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74723,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74723, "pid": 5, "tid": 7, "ts": 1716454222564230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515826, "dur": 12,
    "args": {
      "External id": 74723,
      "cbid": 211, "correlation": 74723
    }
  },
  {
    "ph": "s", "id": 74723, "pid": 76337, "tid": -914061504, "ts": 1716454222515826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222564254, "dur": 122,
    "args": {
      "External id": 74725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74725,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74725, "pid": 5, "tid": 7, "ts": 1716454222564254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515845, "dur": 7,
    "args": {
      "External id": 74725,
      "cbid": 211, "correlation": 74725
    }
  },
  {
    "ph": "s", "id": 74725, "pid": 76337, "tid": -914061504, "ts": 1716454222515845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222564377, "dur": 23,
    "args": {
      "External id": 74727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74727,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74727, "pid": 5, "tid": 7, "ts": 1716454222564377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515856, "dur": 5,
    "args": {
      "External id": 74727,
      "cbid": 211, "correlation": 74727
    }
  },
  {
    "ph": "s", "id": 74727, "pid": 76337, "tid": -914061504, "ts": 1716454222515856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222564401, "dur": 33,
    "args": {
      "External id": 74733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74733, "pid": 5, "tid": 7, "ts": 1716454222564401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515883, "dur": 8,
    "args": {
      "External id": 74733,
      "cbid": 211, "correlation": 74733
    }
  },
  {
    "ph": "s", "id": 74733, "pid": 76337, "tid": -914061504, "ts": 1716454222515883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222564435, "dur": 27,
    "args": {
      "External id": 74741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74741,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74741, "pid": 5, "tid": 7, "ts": 1716454222564435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515915, "dur": 8,
    "args": {
      "External id": 74741,
      "cbid": 211, "correlation": 74741
    }
  },
  {
    "ph": "s", "id": 74741, "pid": 76337, "tid": -914061504, "ts": 1716454222515915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222564463, "dur": 31,
    "args": {
      "External id": 74761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74761,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 74761, "pid": 5, "tid": 7, "ts": 1716454222564463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222515995, "dur": 13,
    "args": {
      "External id": 74761,
      "cbid": 211, "correlation": 74761
    }
  },
  {
    "ph": "s", "id": 74761, "pid": 76337, "tid": -914061504, "ts": 1716454222515995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222564495, "dur": 4,
    "args": {
      "External id": 74773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74773,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 74773, "pid": 5, "tid": 7, "ts": 1716454222564495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516018, "dur": 6,
    "args": {
      "External id": 74773,
      "cbid": 211, "correlation": 74773
    }
  },
  {
    "ph": "s", "id": 74773, "pid": 76337, "tid": -914061504, "ts": 1716454222516018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222564501, "dur": 31,
    "args": {
      "External id": 74776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74776, "pid": 5, "tid": 7, "ts": 1716454222564501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516037, "dur": 6,
    "args": {
      "External id": 74776,
      "cbid": 211, "correlation": 74776
    }
  },
  {
    "ph": "s", "id": 74776, "pid": 76337, "tid": -914061504, "ts": 1716454222516037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222564533, "dur": 21,
    "args": {
      "External id": 74785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74785,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74785, "pid": 5, "tid": 7, "ts": 1716454222564533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516077, "dur": 9,
    "args": {
      "External id": 74785,
      "cbid": 211, "correlation": 74785
    }
  },
  {
    "ph": "s", "id": 74785, "pid": 76337, "tid": -914061504, "ts": 1716454222516077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516127, "dur": 0,
    "args": {
      "External id": 74795,
      "cbid": 317, "correlation": 74795
    }
  },
  {
    "ph": "f", "id": 74795, "pid": 76337, "tid": -914061504, "ts": 1716454222516127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516128, "dur": 0,
    "args": {
      "External id": 74796,
      "cbid": 203, "correlation": 74796
    }
  },
  {
    "ph": "f", "id": 74796, "pid": 76337, "tid": -914061504, "ts": 1716454222516128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516129, "dur": 0,
    "args": {
      "External id": 74797,
      "cbid": 205, "correlation": 74797
    }
  },
  {
    "ph": "f", "id": 74797, "pid": 76337, "tid": -914061504, "ts": 1716454222516129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222564555, "dur": 23,
    "args": {
      "External id": 74801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74801,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74801, "pid": 5, "tid": 7, "ts": 1716454222564555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516142, "dur": 11,
    "args": {
      "External id": 74801,
      "cbid": 211, "correlation": 74801
    }
  },
  {
    "ph": "s", "id": 74801, "pid": 76337, "tid": -914061504, "ts": 1716454222516142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222564579, "dur": 45,
    "args": {
      "External id": 74803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74803,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74803, "pid": 5, "tid": 7, "ts": 1716454222564579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516156, "dur": 6,
    "args": {
      "External id": 74803,
      "cbid": 211, "correlation": 74803
    }
  },
  {
    "ph": "s", "id": 74803, "pid": 76337, "tid": -914061504, "ts": 1716454222516156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222564626, "dur": 665,
    "args": {
      "External id": 74805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74805,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74805, "pid": 5, "tid": 7, "ts": 1716454222564626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516168, "dur": 6,
    "args": {
      "External id": 74805,
      "cbid": 211, "correlation": 74805
    }
  },
  {
    "ph": "s", "id": 74805, "pid": 76337, "tid": -914061504, "ts": 1716454222516168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222565292, "dur": 22,
    "args": {
      "External id": 74807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74807,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74807, "pid": 5, "tid": 7, "ts": 1716454222565292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516178, "dur": 5,
    "args": {
      "External id": 74807,
      "cbid": 211, "correlation": 74807
    }
  },
  {
    "ph": "s", "id": 74807, "pid": 76337, "tid": -914061504, "ts": 1716454222516178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222565315, "dur": 33,
    "args": {
      "External id": 74813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74813, "pid": 5, "tid": 7, "ts": 1716454222565315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516205, "dur": 9,
    "args": {
      "External id": 74813,
      "cbid": 211, "correlation": 74813
    }
  },
  {
    "ph": "s", "id": 74813, "pid": 76337, "tid": -914061504, "ts": 1716454222516205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222565349, "dur": 3,
    "args": {
      "External id": 74821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74821,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 74821, "pid": 5, "tid": 7, "ts": 1716454222565349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516247, "dur": 9,
    "args": {
      "External id": 74821,
      "cbid": 211, "correlation": 74821
    }
  },
  {
    "ph": "s", "id": 74821, "pid": 76337, "tid": -914061504, "ts": 1716454222516247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516313, "dur": 1,
    "args": {
      "External id": 74837,
      "cbid": 251, "correlation": 74837
    }
  },
  {
    "ph": "f", "id": 74837, "pid": 76337, "tid": -914061504, "ts": 1716454222516313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516318, "dur": 0,
    "args": {
      "External id": 74839,
      "cbid": 251, "correlation": 74839
    }
  },
  {
    "ph": "f", "id": 74839, "pid": 76337, "tid": -914061504, "ts": 1716454222516318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222565354, "dur": 12,
    "args": {
      "External id": 74840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74840,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 74840, "pid": 5, "tid": 7, "ts": 1716454222565354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516320, "dur": 11,
    "args": {
      "External id": 74840,
      "cbid": 211, "correlation": 74840
    }
  },
  {
    "ph": "s", "id": 74840, "pid": 76337, "tid": -914061504, "ts": 1716454222516320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222565368, "dur": 5,
    "args": {
      "External id": 74842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74842,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 74842, "pid": 5, "tid": 7, "ts": 1716454222565368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516333, "dur": 5,
    "args": {
      "External id": 74842,
      "cbid": 211, "correlation": 74842
    }
  },
  {
    "ph": "s", "id": 74842, "pid": 76337, "tid": -914061504, "ts": 1716454222516333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222565375, "dur": 29,
    "args": {
      "External id": 74852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74852, "pid": 5, "tid": 7, "ts": 1716454222565375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516391, "dur": 12,
    "args": {
      "External id": 74852,
      "cbid": 211, "correlation": 74852
    }
  },
  {
    "ph": "s", "id": 74852, "pid": 76337, "tid": -914061504, "ts": 1716454222516391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222565405, "dur": 30,
    "args": {
      "External id": 74872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74872,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 74872, "pid": 5, "tid": 7, "ts": 1716454222565405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516456, "dur": 11,
    "args": {
      "External id": 74872,
      "cbid": 211, "correlation": 74872
    }
  },
  {
    "ph": "s", "id": 74872, "pid": 76337, "tid": -914061504, "ts": 1716454222516456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222565437, "dur": 4,
    "args": {
      "External id": 74884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74884,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 74884, "pid": 5, "tid": 7, "ts": 1716454222565437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516477, "dur": 6,
    "args": {
      "External id": 74884,
      "cbid": 211, "correlation": 74884
    }
  },
  {
    "ph": "s", "id": 74884, "pid": 76337, "tid": -914061504, "ts": 1716454222516477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222565442, "dur": 30,
    "args": {
      "External id": 74887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74887, "pid": 5, "tid": 7, "ts": 1716454222565442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516495, "dur": 6,
    "args": {
      "External id": 74887,
      "cbid": 211, "correlation": 74887
    }
  },
  {
    "ph": "s", "id": 74887, "pid": 76337, "tid": -914061504, "ts": 1716454222516495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222565474, "dur": 21,
    "args": {
      "External id": 74896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74896,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74896, "pid": 5, "tid": 7, "ts": 1716454222565474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516536, "dur": 9,
    "args": {
      "External id": 74896,
      "cbid": 211, "correlation": 74896
    }
  },
  {
    "ph": "s", "id": 74896, "pid": 76337, "tid": -914061504, "ts": 1716454222516536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516598, "dur": 0,
    "args": {
      "External id": 74906,
      "cbid": 317, "correlation": 74906
    }
  },
  {
    "ph": "f", "id": 74906, "pid": 76337, "tid": -914061504, "ts": 1716454222516598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516599, "dur": 0,
    "args": {
      "External id": 74907,
      "cbid": 203, "correlation": 74907
    }
  },
  {
    "ph": "f", "id": 74907, "pid": 76337, "tid": -914061504, "ts": 1716454222516599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516600, "dur": 0,
    "args": {
      "External id": 74908,
      "cbid": 205, "correlation": 74908
    }
  },
  {
    "ph": "f", "id": 74908, "pid": 76337, "tid": -914061504, "ts": 1716454222516600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222565496, "dur": 24,
    "args": {
      "External id": 74912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74912, "pid": 5, "tid": 7, "ts": 1716454222565496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516614, "dur": 12,
    "args": {
      "External id": 74912,
      "cbid": 211, "correlation": 74912
    }
  },
  {
    "ph": "s", "id": 74912, "pid": 76337, "tid": -914061504, "ts": 1716454222516614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222565521, "dur": 44,
    "args": {
      "External id": 74914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74914, "pid": 5, "tid": 7, "ts": 1716454222565521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516628, "dur": 5,
    "args": {
      "External id": 74914,
      "cbid": 211, "correlation": 74914
    }
  },
  {
    "ph": "s", "id": 74914, "pid": 76337, "tid": -914061504, "ts": 1716454222516628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222565567, "dur": 655,
    "args": {
      "External id": 74916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74916,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74916, "pid": 5, "tid": 7, "ts": 1716454222565567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516640, "dur": 6,
    "args": {
      "External id": 74916,
      "cbid": 211, "correlation": 74916
    }
  },
  {
    "ph": "s", "id": 74916, "pid": 76337, "tid": -914061504, "ts": 1716454222516640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222566223, "dur": 22,
    "args": {
      "External id": 74918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74918,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74918, "pid": 5, "tid": 7, "ts": 1716454222566223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516649, "dur": 5,
    "args": {
      "External id": 74918,
      "cbid": 211, "correlation": 74918
    }
  },
  {
    "ph": "s", "id": 74918, "pid": 76337, "tid": -914061504, "ts": 1716454222516649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222566246, "dur": 34,
    "args": {
      "External id": 74924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74924, "pid": 5, "tid": 7, "ts": 1716454222566246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516677, "dur": 8,
    "args": {
      "External id": 74924,
      "cbid": 211, "correlation": 74924
    }
  },
  {
    "ph": "s", "id": 74924, "pid": 76337, "tid": -914061504, "ts": 1716454222516677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222566282, "dur": 27,
    "args": {
      "External id": 74932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74932,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74932, "pid": 5, "tid": 7, "ts": 1716454222566282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516708, "dur": 8,
    "args": {
      "External id": 74932,
      "cbid": 211, "correlation": 74932
    }
  },
  {
    "ph": "s", "id": 74932, "pid": 76337, "tid": -914061504, "ts": 1716454222516708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222566310, "dur": 20,
    "args": {
      "External id": 74940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74940,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74940, "pid": 5, "tid": 7, "ts": 1716454222566310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516738, "dur": 8,
    "args": {
      "External id": 74940,
      "cbid": 211, "correlation": 74940
    }
  },
  {
    "ph": "s", "id": 74940, "pid": 76337, "tid": -914061504, "ts": 1716454222516738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222566331, "dur": 31,
    "args": {
      "External id": 74960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74960,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 74960, "pid": 5, "tid": 7, "ts": 1716454222566331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516816, "dur": 12,
    "args": {
      "External id": 74960,
      "cbid": 211, "correlation": 74960
    }
  },
  {
    "ph": "s", "id": 74960, "pid": 76337, "tid": -914061504, "ts": 1716454222516816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222566364, "dur": 4,
    "args": {
      "External id": 74972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74972,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 74972, "pid": 5, "tid": 7, "ts": 1716454222566364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516837, "dur": 6,
    "args": {
      "External id": 74972,
      "cbid": 211, "correlation": 74972
    }
  },
  {
    "ph": "s", "id": 74972, "pid": 76337, "tid": -914061504, "ts": 1716454222516837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222566369, "dur": 31,
    "args": {
      "External id": 74975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74975,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74975, "pid": 5, "tid": 7, "ts": 1716454222566369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516855, "dur": 6,
    "args": {
      "External id": 74975,
      "cbid": 211, "correlation": 74975
    }
  },
  {
    "ph": "s", "id": 74975, "pid": 76337, "tid": -914061504, "ts": 1716454222516855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516912, "dur": 0,
    "args": {
      "External id": 74986,
      "cbid": 317, "correlation": 74986
    }
  },
  {
    "ph": "f", "id": 74986, "pid": 76337, "tid": -914061504, "ts": 1716454222516912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516912, "dur": 0,
    "args": {
      "External id": 74987,
      "cbid": 203, "correlation": 74987
    }
  },
  {
    "ph": "f", "id": 74987, "pid": 76337, "tid": -914061504, "ts": 1716454222516912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516913, "dur": 0,
    "args": {
      "External id": 74988,
      "cbid": 205, "correlation": 74988
    }
  },
  {
    "ph": "f", "id": 74988, "pid": 76337, "tid": -914061504, "ts": 1716454222516913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222566401, "dur": 22,
    "args": {
      "External id": 74992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74992, "pid": 5, "tid": 7, "ts": 1716454222566401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516926, "dur": 11,
    "args": {
      "External id": 74992,
      "cbid": 211, "correlation": 74992
    }
  },
  {
    "ph": "s", "id": 74992, "pid": 76337, "tid": -914061504, "ts": 1716454222516926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222566424, "dur": 119,
    "args": {
      "External id": 74994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74994,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 74994, "pid": 5, "tid": 7, "ts": 1716454222566424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516944, "dur": 7,
    "args": {
      "External id": 74994,
      "cbid": 211, "correlation": 74994
    }
  },
  {
    "ph": "s", "id": 74994, "pid": 76337, "tid": -914061504, "ts": 1716454222516944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222566544, "dur": 22,
    "args": {
      "External id": 74996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 74996,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 74996, "pid": 5, "tid": 7, "ts": 1716454222566544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516954, "dur": 5,
    "args": {
      "External id": 74996,
      "cbid": 211, "correlation": 74996
    }
  },
  {
    "ph": "s", "id": 74996, "pid": 76337, "tid": -914061504, "ts": 1716454222516954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222566567, "dur": 33,
    "args": {
      "External id": 75002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75002, "pid": 5, "tid": 7, "ts": 1716454222566567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222516989, "dur": 10,
    "args": {
      "External id": 75002,
      "cbid": 211, "correlation": 75002
    }
  },
  {
    "ph": "s", "id": 75002, "pid": 76337, "tid": -914061504, "ts": 1716454222516989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222566602, "dur": 202,
    "args": {
      "External id": 75011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75011,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75011, "pid": 5, "tid": 7, "ts": 1716454222566602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517072, "dur": 13,
    "args": {
      "External id": 75011,
      "cbid": 211, "correlation": 75011
    }
  },
  {
    "ph": "s", "id": 75011, "pid": 76337, "tid": -914061504, "ts": 1716454222517072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222566805, "dur": 67,
    "args": {
      "External id": 75033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75033,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75033, "pid": 5, "tid": 7, "ts": 1716454222566805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517129, "dur": 10,
    "args": {
      "External id": 75033,
      "cbid": 211, "correlation": 75033
    }
  },
  {
    "ph": "s", "id": 75033, "pid": 76337, "tid": -914061504, "ts": 1716454222517129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517216, "dur": 1,
    "args": {
      "External id": 75044,
      "cbid": 251, "correlation": 75044
    }
  },
  {
    "ph": "f", "id": 75044, "pid": 76337, "tid": -914061504, "ts": 1716454222517216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222566873, "dur": 155,
    "args": {
      "External id": 75045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75045,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75045, "pid": 5, "tid": 7, "ts": 1716454222566873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517222, "dur": 12,
    "args": {
      "External id": 75045,
      "cbid": 211, "correlation": 75045
    }
  },
  {
    "ph": "s", "id": 75045, "pid": 76337, "tid": -914061504, "ts": 1716454222517222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517291, "dur": 1,
    "args": {
      "External id": 75056,
      "cbid": 251, "correlation": 75056
    }
  },
  {
    "ph": "f", "id": 75056, "pid": 76337, "tid": -914061504, "ts": 1716454222517291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222567030, "dur": 148,
    "args": {
      "External id": 75057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75057,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75057, "pid": 5, "tid": 7, "ts": 1716454222567030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517295, "dur": 12,
    "args": {
      "External id": 75057,
      "cbid": 211, "correlation": 75057
    }
  },
  {
    "ph": "s", "id": 75057, "pid": 76337, "tid": -914061504, "ts": 1716454222517295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517360, "dur": 1,
    "args": {
      "External id": 75068,
      "cbid": 251, "correlation": 75068
    }
  },
  {
    "ph": "f", "id": 75068, "pid": 76337, "tid": -914061504, "ts": 1716454222517360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222567180, "dur": 148,
    "args": {
      "External id": 75069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75069,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75069, "pid": 5, "tid": 7, "ts": 1716454222567180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517364, "dur": 11,
    "args": {
      "External id": 75069,
      "cbid": 211, "correlation": 75069
    }
  },
  {
    "ph": "s", "id": 75069, "pid": 76337, "tid": -914061504, "ts": 1716454222517364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222567329, "dur": 1988,
    "args": {
      "External id": 75090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75090,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 75090, "pid": 5, "tid": 7, "ts": 1716454222567329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517442, "dur": 14,
    "args": {
      "External id": 75090,
      "cbid": 211, "correlation": 75090
    }
  },
  {
    "ph": "s", "id": 75090, "pid": 76337, "tid": -914061504, "ts": 1716454222517442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517541, "dur": 1,
    "args": {
      "External id": 75108,
      "cbid": 251, "correlation": 75108
    }
  },
  {
    "ph": "f", "id": 75108, "pid": 76337, "tid": -914061504, "ts": 1716454222517541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222569319, "dur": 151,
    "args": {
      "External id": 75110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75110,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 75110, "pid": 5, "tid": 7, "ts": 1716454222569319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517547, "dur": 13,
    "args": {
      "External id": 75110,
      "cbid": 211, "correlation": 75110
    }
  },
  {
    "ph": "s", "id": 75110, "pid": 76337, "tid": -914061504, "ts": 1716454222517547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222569471, "dur": 35,
    "args": {
      "External id": 75118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75118,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75118, "pid": 5, "tid": 7, "ts": 1716454222569471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517619, "dur": 12,
    "args": {
      "External id": 75118,
      "cbid": 211, "correlation": 75118
    }
  },
  {
    "ph": "s", "id": 75118, "pid": 76337, "tid": -914061504, "ts": 1716454222517619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222569508, "dur": 50,
    "args": {
      "External id": 75126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75126,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75126, "pid": 5, "tid": 7, "ts": 1716454222569508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517657, "dur": 8,
    "args": {
      "External id": 75126,
      "cbid": 211, "correlation": 75126
    }
  },
  {
    "ph": "s", "id": 75126, "pid": 76337, "tid": -914061504, "ts": 1716454222517657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222569559, "dur": 31,
    "args": {
      "External id": 75137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75137,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75137, "pid": 5, "tid": 7, "ts": 1716454222569559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517729, "dur": 13,
    "args": {
      "External id": 75137,
      "cbid": 211, "correlation": 75137
    }
  },
  {
    "ph": "s", "id": 75137, "pid": 76337, "tid": -914061504, "ts": 1716454222517729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222569591, "dur": 35,
    "args": {
      "External id": 75159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75159,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75159, "pid": 5, "tid": 7, "ts": 1716454222569591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517760, "dur": 7,
    "args": {
      "External id": 75159,
      "cbid": 211, "correlation": 75159
    }
  },
  {
    "ph": "s", "id": 75159, "pid": 76337, "tid": -914061504, "ts": 1716454222517760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517845, "dur": 1,
    "args": {
      "External id": 75170,
      "cbid": 251, "correlation": 75170
    }
  },
  {
    "ph": "f", "id": 75170, "pid": 76337, "tid": -914061504, "ts": 1716454222517845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222569628, "dur": 90,
    "args": {
      "External id": 75171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75171,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75171, "pid": 5, "tid": 7, "ts": 1716454222569628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517850, "dur": 12,
    "args": {
      "External id": 75171,
      "cbid": 211, "correlation": 75171
    }
  },
  {
    "ph": "s", "id": 75171, "pid": 76337, "tid": -914061504, "ts": 1716454222517850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517918, "dur": 1,
    "args": {
      "External id": 75182,
      "cbid": 251, "correlation": 75182
    }
  },
  {
    "ph": "f", "id": 75182, "pid": 76337, "tid": -914061504, "ts": 1716454222517918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517921, "dur": 0,
    "args": {
      "External id": 75183,
      "cbid": 251, "correlation": 75183
    }
  },
  {
    "ph": "f", "id": 75183, "pid": 76337, "tid": -914061504, "ts": 1716454222517921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222569720, "dur": 11,
    "args": {
      "External id": 75184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75184,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 75184, "pid": 5, "tid": 7, "ts": 1716454222569720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517923, "dur": 12,
    "args": {
      "External id": 75184,
      "cbid": 211, "correlation": 75184
    }
  },
  {
    "ph": "s", "id": 75184, "pid": 76337, "tid": -914061504, "ts": 1716454222517923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222569732, "dur": 5,
    "args": {
      "External id": 75186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75186,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 75186, "pid": 5, "tid": 7, "ts": 1716454222569732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222517936, "dur": 6,
    "args": {
      "External id": 75186,
      "cbid": 211, "correlation": 75186
    }
  },
  {
    "ph": "s", "id": 75186, "pid": 76337, "tid": -914061504, "ts": 1716454222517936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518003, "dur": 1,
    "args": {
      "External id": 75197,
      "cbid": 251, "correlation": 75197
    }
  },
  {
    "ph": "f", "id": 75197, "pid": 76337, "tid": -914061504, "ts": 1716454222518003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518007, "dur": 0,
    "args": {
      "External id": 75198,
      "cbid": 251, "correlation": 75198
    }
  },
  {
    "ph": "f", "id": 75198, "pid": 76337, "tid": -914061504, "ts": 1716454222518007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222569738, "dur": 7,
    "args": {
      "External id": 75199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75199,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 75199, "pid": 5, "tid": 7, "ts": 1716454222569738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518008, "dur": 12,
    "args": {
      "External id": 75199,
      "cbid": 211, "correlation": 75199
    }
  },
  {
    "ph": "s", "id": 75199, "pid": 76337, "tid": -914061504, "ts": 1716454222518008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222569747, "dur": 3,
    "args": {
      "External id": 75201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75201,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 75201, "pid": 5, "tid": 7, "ts": 1716454222569747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518022, "dur": 5,
    "args": {
      "External id": 75201,
      "cbid": 211, "correlation": 75201
    }
  },
  {
    "ph": "s", "id": 75201, "pid": 76337, "tid": -914061504, "ts": 1716454222518022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222569751, "dur": 94,
    "args": {
      "External id": 75222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75222,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 75222, "pid": 5, "tid": 7, "ts": 1716454222569751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518096, "dur": 12,
    "args": {
      "External id": 75222,
      "cbid": 211, "correlation": 75222
    }
  },
  {
    "ph": "s", "id": 75222, "pid": 76337, "tid": -914061504, "ts": 1716454222518096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518194, "dur": 1,
    "args": {
      "External id": 75240,
      "cbid": 251, "correlation": 75240
    }
  },
  {
    "ph": "f", "id": 75240, "pid": 76337, "tid": -914061504, "ts": 1716454222518194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222569846, "dur": 100,
    "args": {
      "External id": 75242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75242,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75242, "pid": 5, "tid": 7, "ts": 1716454222569846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518200, "dur": 13,
    "args": {
      "External id": 75242,
      "cbid": 211, "correlation": 75242
    }
  },
  {
    "ph": "s", "id": 75242, "pid": 76337, "tid": -914061504, "ts": 1716454222518200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222569948, "dur": 19,
    "args": {
      "External id": 75250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75250,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75250, "pid": 5, "tid": 7, "ts": 1716454222569948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518268, "dur": 12,
    "args": {
      "External id": 75250,
      "cbid": 211, "correlation": 75250
    }
  },
  {
    "ph": "s", "id": 75250, "pid": 76337, "tid": -914061504, "ts": 1716454222518268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222569969, "dur": 37,
    "args": {
      "External id": 75258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75258, "pid": 5, "tid": 7, "ts": 1716454222569969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518310, "dur": 9,
    "args": {
      "External id": 75258,
      "cbid": 211, "correlation": 75258
    }
  },
  {
    "ph": "s", "id": 75258, "pid": 76337, "tid": -914061504, "ts": 1716454222518310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222570007, "dur": 36,
    "args": {
      "External id": 75280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75280,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75280, "pid": 5, "tid": 7, "ts": 1716454222570007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518361, "dur": 10,
    "args": {
      "External id": 75280,
      "cbid": 211, "correlation": 75280
    }
  },
  {
    "ph": "s", "id": 75280, "pid": 76337, "tid": -914061504, "ts": 1716454222518361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518450, "dur": 1,
    "args": {
      "External id": 75296,
      "cbid": 251, "correlation": 75296
    }
  },
  {
    "ph": "f", "id": 75296, "pid": 76337, "tid": -914061504, "ts": 1716454222518450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518455, "dur": 0,
    "args": {
      "External id": 75298,
      "cbid": 251, "correlation": 75298
    }
  },
  {
    "ph": "f", "id": 75298, "pid": 76337, "tid": -914061504, "ts": 1716454222518455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222570044, "dur": 549,
    "args": {
      "External id": 75299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75299,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 75299, "pid": 5, "tid": 7, "ts": 1716454222570044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518458, "dur": 13,
    "args": {
      "External id": 75299,
      "cbid": 211, "correlation": 75299
    }
  },
  {
    "ph": "s", "id": 75299, "pid": 76337, "tid": -914061504, "ts": 1716454222518458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222570594, "dur": 127,
    "args": {
      "External id": 75307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75307,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75307, "pid": 5, "tid": 7, "ts": 1716454222570594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518524, "dur": 13,
    "args": {
      "External id": 75307,
      "cbid": 211, "correlation": 75307
    }
  },
  {
    "ph": "s", "id": 75307, "pid": 76337, "tid": -914061504, "ts": 1716454222518524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222570723, "dur": 130,
    "args": {
      "External id": 75315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75315, "pid": 5, "tid": 7, "ts": 1716454222570723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518555, "dur": 8,
    "args": {
      "External id": 75315,
      "cbid": 211, "correlation": 75315
    }
  },
  {
    "ph": "s", "id": 75315, "pid": 76337, "tid": -914061504, "ts": 1716454222518555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518631, "dur": 1,
    "args": {
      "External id": 75331,
      "cbid": 251, "correlation": 75331
    }
  },
  {
    "ph": "f", "id": 75331, "pid": 76337, "tid": -914061504, "ts": 1716454222518631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222570854, "dur": 308,
    "args": {
      "External id": 75333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75333,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75333, "pid": 5, "tid": 7, "ts": 1716454222570854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518636, "dur": 12,
    "args": {
      "External id": 75333,
      "cbid": 211, "correlation": 75333
    }
  },
  {
    "ph": "s", "id": 75333, "pid": 76337, "tid": -914061504, "ts": 1716454222518636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222571163, "dur": 27,
    "args": {
      "External id": 75341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75341,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75341, "pid": 5, "tid": 7, "ts": 1716454222571163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518678, "dur": 10,
    "args": {
      "External id": 75341,
      "cbid": 211, "correlation": 75341
    }
  },
  {
    "ph": "s", "id": 75341, "pid": 76337, "tid": -914061504, "ts": 1716454222518678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222571191, "dur": 83,
    "args": {
      "External id": 75352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75352, "pid": 5, "tid": 7, "ts": 1716454222571191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518745, "dur": 12,
    "args": {
      "External id": 75352,
      "cbid": 211, "correlation": 75352
    }
  },
  {
    "ph": "s", "id": 75352, "pid": 76337, "tid": -914061504, "ts": 1716454222518745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518809, "dur": 0,
    "args": {
      "External id": 75364,
      "cbid": 317, "correlation": 75364
    }
  },
  {
    "ph": "f", "id": 75364, "pid": 76337, "tid": -914061504, "ts": 1716454222518809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518810, "dur": 0,
    "args": {
      "External id": 75365,
      "cbid": 203, "correlation": 75365
    }
  },
  {
    "ph": "f", "id": 75365, "pid": 76337, "tid": -914061504, "ts": 1716454222518810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518811, "dur": 0,
    "args": {
      "External id": 75366,
      "cbid": 205, "correlation": 75366
    }
  },
  {
    "ph": "f", "id": 75366, "pid": 76337, "tid": -914061504, "ts": 1716454222518811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571276, "dur": 23,
    "args": {
      "External id": 75370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75370,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75370, "pid": 5, "tid": 7, "ts": 1716454222571276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518826, "dur": 12,
    "args": {
      "External id": 75370,
      "cbid": 211, "correlation": 75370
    }
  },
  {
    "ph": "s", "id": 75370, "pid": 76337, "tid": -914061504, "ts": 1716454222518826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222571300, "dur": 123,
    "args": {
      "External id": 75372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75372,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75372, "pid": 5, "tid": 7, "ts": 1716454222571300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518846, "dur": 6,
    "args": {
      "External id": 75372,
      "cbid": 211, "correlation": 75372
    }
  },
  {
    "ph": "s", "id": 75372, "pid": 76337, "tid": -914061504, "ts": 1716454222518846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571425, "dur": 21,
    "args": {
      "External id": 75374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75374,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75374, "pid": 5, "tid": 7, "ts": 1716454222571425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518856, "dur": 5,
    "args": {
      "External id": 75374,
      "cbid": 211, "correlation": 75374
    }
  },
  {
    "ph": "s", "id": 75374, "pid": 76337, "tid": -914061504, "ts": 1716454222518856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222571447, "dur": 33,
    "args": {
      "External id": 75380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75380, "pid": 5, "tid": 7, "ts": 1716454222571447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518884, "dur": 8,
    "args": {
      "External id": 75380,
      "cbid": 211, "correlation": 75380
    }
  },
  {
    "ph": "s", "id": 75380, "pid": 76337, "tid": -914061504, "ts": 1716454222518884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222571482, "dur": 27,
    "args": {
      "External id": 75388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75388,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75388, "pid": 5, "tid": 7, "ts": 1716454222571482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518916, "dur": 8,
    "args": {
      "External id": 75388,
      "cbid": 211, "correlation": 75388
    }
  },
  {
    "ph": "s", "id": 75388, "pid": 76337, "tid": -914061504, "ts": 1716454222518916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518996, "dur": 0,
    "args": {
      "External id": 75398,
      "cbid": 317, "correlation": 75398
    }
  },
  {
    "ph": "f", "id": 75398, "pid": 76337, "tid": -914061504, "ts": 1716454222518996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518997, "dur": 0,
    "args": {
      "External id": 75399,
      "cbid": 203, "correlation": 75399
    }
  },
  {
    "ph": "f", "id": 75399, "pid": 76337, "tid": -914061504, "ts": 1716454222518997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222518997, "dur": 0,
    "args": {
      "External id": 75400,
      "cbid": 205, "correlation": 75400
    }
  },
  {
    "ph": "f", "id": 75400, "pid": 76337, "tid": -914061504, "ts": 1716454222518997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571510, "dur": 22,
    "args": {
      "External id": 75404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75404, "pid": 5, "tid": 7, "ts": 1716454222571510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519011, "dur": 12,
    "args": {
      "External id": 75404,
      "cbid": 211, "correlation": 75404
    }
  },
  {
    "ph": "s", "id": 75404, "pid": 76337, "tid": -914061504, "ts": 1716454222519011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571533, "dur": 46,
    "args": {
      "External id": 75406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75406,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75406, "pid": 5, "tid": 7, "ts": 1716454222571533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519026, "dur": 5,
    "args": {
      "External id": 75406,
      "cbid": 211, "correlation": 75406
    }
  },
  {
    "ph": "s", "id": 75406, "pid": 76337, "tid": -914061504, "ts": 1716454222519026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222571580, "dur": 239,
    "args": {
      "External id": 75408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75408,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 75408, "pid": 5, "tid": 7, "ts": 1716454222571580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519038, "dur": 7,
    "args": {
      "External id": 75408,
      "cbid": 211, "correlation": 75408
    }
  },
  {
    "ph": "s", "id": 75408, "pid": 76337, "tid": -914061504, "ts": 1716454222519038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571820, "dur": 6,
    "args": {
      "External id": 75410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75410, "pid": 5, "tid": 7, "ts": 1716454222571820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519048, "dur": 5,
    "args": {
      "External id": 75410,
      "cbid": 211, "correlation": 75410
    }
  },
  {
    "ph": "s", "id": 75410, "pid": 76337, "tid": -914061504, "ts": 1716454222519048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222571828, "dur": 9,
    "args": {
      "External id": 75416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75416, "pid": 5, "tid": 7, "ts": 1716454222571828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519075, "dur": 8,
    "args": {
      "External id": 75416,
      "cbid": 211, "correlation": 75416
    }
  },
  {
    "ph": "s", "id": 75416, "pid": 76337, "tid": -914061504, "ts": 1716454222519075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571838, "dur": 12,
    "args": {
      "External id": 75436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75436,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 75436, "pid": 5, "tid": 7, "ts": 1716454222571838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519169, "dur": 13,
    "args": {
      "External id": 75436,
      "cbid": 211, "correlation": 75436
    }
  },
  {
    "ph": "s", "id": 75436, "pid": 76337, "tid": -914061504, "ts": 1716454222519169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222571852, "dur": 4,
    "args": {
      "External id": 75448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75448,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 75448, "pid": 5, "tid": 7, "ts": 1716454222571852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519192, "dur": 6,
    "args": {
      "External id": 75448,
      "cbid": 211, "correlation": 75448
    }
  },
  {
    "ph": "s", "id": 75448, "pid": 76337, "tid": -914061504, "ts": 1716454222519192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222571858, "dur": 12,
    "args": {
      "External id": 75451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75451,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75451, "pid": 5, "tid": 7, "ts": 1716454222571858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519210, "dur": 7,
    "args": {
      "External id": 75451,
      "cbid": 211, "correlation": 75451
    }
  },
  {
    "ph": "s", "id": 75451, "pid": 76337, "tid": -914061504, "ts": 1716454222519210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222571872, "dur": 7,
    "args": {
      "External id": 75460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75460,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75460, "pid": 5, "tid": 7, "ts": 1716454222571872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519250, "dur": 9,
    "args": {
      "External id": 75460,
      "cbid": 211, "correlation": 75460
    }
  },
  {
    "ph": "s", "id": 75460, "pid": 76337, "tid": -914061504, "ts": 1716454222519250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519302, "dur": 0,
    "args": {
      "External id": 75470,
      "cbid": 317, "correlation": 75470
    }
  },
  {
    "ph": "f", "id": 75470, "pid": 76337, "tid": -914061504, "ts": 1716454222519302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519303, "dur": 0,
    "args": {
      "External id": 75471,
      "cbid": 203, "correlation": 75471
    }
  },
  {
    "ph": "f", "id": 75471, "pid": 76337, "tid": -914061504, "ts": 1716454222519303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519304, "dur": 0,
    "args": {
      "External id": 75472,
      "cbid": 205, "correlation": 75472
    }
  },
  {
    "ph": "f", "id": 75472, "pid": 76337, "tid": -914061504, "ts": 1716454222519304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571880, "dur": 6,
    "args": {
      "External id": 75476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75476,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75476, "pid": 5, "tid": 7, "ts": 1716454222571880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519320, "dur": 11,
    "args": {
      "External id": 75476,
      "cbid": 211, "correlation": 75476
    }
  },
  {
    "ph": "s", "id": 75476, "pid": 76337, "tid": -914061504, "ts": 1716454222519320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222571887, "dur": 85,
    "args": {
      "External id": 75478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75478, "pid": 5, "tid": 7, "ts": 1716454222571887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519334, "dur": 5,
    "args": {
      "External id": 75478,
      "cbid": 211, "correlation": 75478
    }
  },
  {
    "ph": "s", "id": 75478, "pid": 76337, "tid": -914061504, "ts": 1716454222519334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222571974, "dur": 1,
    "args": {
      "External id": 75480,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 75480,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5454545454545454
    }
  },
  {
    "ph": "f", "id": 75480, "pid": 5, "tid": 7, "ts": 1716454222571974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519347, "dur": 9,
    "args": {
      "External id": 75480,
      "cbid": 51, "correlation": 75480
    }
  },
  {
    "ph": "s", "id": 75480, "pid": 76337, "tid": -914061504, "ts": 1716454222519347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222571978, "dur": 544,
    "args": {
      "External id": 75481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75481,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75481, "pid": 5, "tid": 7, "ts": 1716454222571978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519357, "dur": 8,
    "args": {
      "External id": 75481,
      "cbid": 211, "correlation": 75481
    }
  },
  {
    "ph": "s", "id": 75481, "pid": 76337, "tid": -914061504, "ts": 1716454222519357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222572523, "dur": 12,
    "args": {
      "External id": 75483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75483,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75483, "pid": 5, "tid": 7, "ts": 1716454222572523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519368, "dur": 5,
    "args": {
      "External id": 75483,
      "cbid": 211, "correlation": 75483
    }
  },
  {
    "ph": "s", "id": 75483, "pid": 76337, "tid": -914061504, "ts": 1716454222519368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222572536, "dur": 15,
    "args": {
      "External id": 75489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75489,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75489, "pid": 5, "tid": 7, "ts": 1716454222572536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519396, "dur": 9,
    "args": {
      "External id": 75489,
      "cbid": 211, "correlation": 75489
    }
  },
  {
    "ph": "s", "id": 75489, "pid": 76337, "tid": -914061504, "ts": 1716454222519396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222572552, "dur": 3,
    "args": {
      "External id": 75497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75497,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 75497, "pid": 5, "tid": 7, "ts": 1716454222572552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519440, "dur": 10,
    "args": {
      "External id": 75497,
      "cbid": 211, "correlation": 75497
    }
  },
  {
    "ph": "s", "id": 75497, "pid": 76337, "tid": -914061504, "ts": 1716454222519440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519504, "dur": 1,
    "args": {
      "External id": 75513,
      "cbid": 251, "correlation": 75513
    }
  },
  {
    "ph": "f", "id": 75513, "pid": 76337, "tid": -914061504, "ts": 1716454222519504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519510, "dur": 0,
    "args": {
      "External id": 75515,
      "cbid": 251, "correlation": 75515
    }
  },
  {
    "ph": "f", "id": 75515, "pid": 76337, "tid": -914061504, "ts": 1716454222519510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222572557, "dur": 13,
    "args": {
      "External id": 75516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75516,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75516, "pid": 5, "tid": 7, "ts": 1716454222572557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519512, "dur": 11,
    "args": {
      "External id": 75516,
      "cbid": 211, "correlation": 75516
    }
  },
  {
    "ph": "s", "id": 75516, "pid": 76337, "tid": -914061504, "ts": 1716454222519512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222572572, "dur": 5,
    "args": {
      "External id": 75518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75518,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75518, "pid": 5, "tid": 7, "ts": 1716454222572572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519524, "dur": 5,
    "args": {
      "External id": 75518,
      "cbid": 211, "correlation": 75518
    }
  },
  {
    "ph": "s", "id": 75518, "pid": 76337, "tid": -914061504, "ts": 1716454222519524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222572578, "dur": 17,
    "args": {
      "External id": 75528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75528,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75528, "pid": 5, "tid": 7, "ts": 1716454222572578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519581, "dur": 12,
    "args": {
      "External id": 75528,
      "cbid": 211, "correlation": 75528
    }
  },
  {
    "ph": "s", "id": 75528, "pid": 76337, "tid": -914061504, "ts": 1716454222519581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222572596, "dur": 19,
    "args": {
      "External id": 75548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75548,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 75548, "pid": 5, "tid": 7, "ts": 1716454222572596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519647, "dur": 11,
    "args": {
      "External id": 75548,
      "cbid": 211, "correlation": 75548
    }
  },
  {
    "ph": "s", "id": 75548, "pid": 76337, "tid": -914061504, "ts": 1716454222519647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222572617, "dur": 5,
    "args": {
      "External id": 75560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75560,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 75560, "pid": 5, "tid": 7, "ts": 1716454222572617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519669, "dur": 6,
    "args": {
      "External id": 75560,
      "cbid": 211, "correlation": 75560
    }
  },
  {
    "ph": "s", "id": 75560, "pid": 76337, "tid": -914061504, "ts": 1716454222519669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222572623, "dur": 17,
    "args": {
      "External id": 75563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75563, "pid": 5, "tid": 7, "ts": 1716454222572623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519687, "dur": 6,
    "args": {
      "External id": 75563,
      "cbid": 211, "correlation": 75563
    }
  },
  {
    "ph": "s", "id": 75563, "pid": 76337, "tid": -914061504, "ts": 1716454222519687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222572641, "dur": 11,
    "args": {
      "External id": 75572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75572,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75572, "pid": 5, "tid": 7, "ts": 1716454222572641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519727, "dur": 10,
    "args": {
      "External id": 75572,
      "cbid": 211, "correlation": 75572
    }
  },
  {
    "ph": "s", "id": 75572, "pid": 76337, "tid": -914061504, "ts": 1716454222519727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519790, "dur": 0,
    "args": {
      "External id": 75582,
      "cbid": 317, "correlation": 75582
    }
  },
  {
    "ph": "f", "id": 75582, "pid": 76337, "tid": -914061504, "ts": 1716454222519790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519791, "dur": 0,
    "args": {
      "External id": 75583,
      "cbid": 203, "correlation": 75583
    }
  },
  {
    "ph": "f", "id": 75583, "pid": 76337, "tid": -914061504, "ts": 1716454222519791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519792, "dur": 0,
    "args": {
      "External id": 75584,
      "cbid": 205, "correlation": 75584
    }
  },
  {
    "ph": "f", "id": 75584, "pid": 76337, "tid": -914061504, "ts": 1716454222519792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222572654, "dur": 11,
    "args": {
      "External id": 75588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75588,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75588, "pid": 5, "tid": 7, "ts": 1716454222572654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519810, "dur": 12,
    "args": {
      "External id": 75588,
      "cbid": 211, "correlation": 75588
    }
  },
  {
    "ph": "s", "id": 75588, "pid": 76337, "tid": -914061504, "ts": 1716454222519810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222572666, "dur": 165,
    "args": {
      "External id": 75590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75590,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75590, "pid": 5, "tid": 7, "ts": 1716454222572666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519824, "dur": 5,
    "args": {
      "External id": 75590,
      "cbid": 211, "correlation": 75590
    }
  },
  {
    "ph": "s", "id": 75590, "pid": 76337, "tid": -914061504, "ts": 1716454222519824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222572833, "dur": 1,
    "args": {
      "External id": 75592,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 75592,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 75592, "pid": 5, "tid": 7, "ts": 1716454222572833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519836, "dur": 7,
    "args": {
      "External id": 75592,
      "cbid": 51, "correlation": 75592
    }
  },
  {
    "ph": "s", "id": 75592, "pid": 76337, "tid": -914061504, "ts": 1716454222519836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222572837, "dur": 672,
    "args": {
      "External id": 75593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75593,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75593, "pid": 5, "tid": 7, "ts": 1716454222572837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519844, "dur": 7,
    "args": {
      "External id": 75593,
      "cbid": 211, "correlation": 75593
    }
  },
  {
    "ph": "s", "id": 75593, "pid": 76337, "tid": -914061504, "ts": 1716454222519844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222573511, "dur": 13,
    "args": {
      "External id": 75595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75595,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75595, "pid": 5, "tid": 7, "ts": 1716454222573511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519855, "dur": 6,
    "args": {
      "External id": 75595,
      "cbid": 211, "correlation": 75595
    }
  },
  {
    "ph": "s", "id": 75595, "pid": 76337, "tid": -914061504, "ts": 1716454222519855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222573526, "dur": 15,
    "args": {
      "External id": 75601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75601,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75601, "pid": 5, "tid": 7, "ts": 1716454222573526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519883, "dur": 8,
    "args": {
      "External id": 75601,
      "cbid": 211, "correlation": 75601
    }
  },
  {
    "ph": "s", "id": 75601, "pid": 76337, "tid": -914061504, "ts": 1716454222519883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519941, "dur": 0,
    "args": {
      "External id": 75611,
      "cbid": 317, "correlation": 75611
    }
  },
  {
    "ph": "f", "id": 75611, "pid": 76337, "tid": -914061504, "ts": 1716454222519941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519942, "dur": 0,
    "args": {
      "External id": 75612,
      "cbid": 203, "correlation": 75612
    }
  },
  {
    "ph": "f", "id": 75612, "pid": 76337, "tid": -914061504, "ts": 1716454222519942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519942, "dur": 0,
    "args": {
      "External id": 75613,
      "cbid": 205, "correlation": 75613
    }
  },
  {
    "ph": "f", "id": 75613, "pid": 76337, "tid": -914061504, "ts": 1716454222519942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222573542, "dur": 8,
    "args": {
      "External id": 75617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75617,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75617, "pid": 5, "tid": 7, "ts": 1716454222573542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519957, "dur": 12,
    "args": {
      "External id": 75617,
      "cbid": 211, "correlation": 75617
    }
  },
  {
    "ph": "s", "id": 75617, "pid": 76337, "tid": -914061504, "ts": 1716454222519957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222573551, "dur": 3,
    "args": {
      "External id": 75619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 75619, "pid": 5, "tid": 7, "ts": 1716454222573551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519983, "dur": 6,
    "args": {
      "External id": 75619,
      "cbid": 211, "correlation": 75619
    }
  },
  {
    "ph": "s", "id": 75619, "pid": 76337, "tid": -914061504, "ts": 1716454222519983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519993, "dur": 0,
    "args": {
      "External id": 75620,
      "cbid": 51, "correlation": 75620
    }
  },
  {
    "ph": "s", "id": 75620, "pid": 76337, "tid": -914061504, "ts": 1716454222519993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222573556, "dur": 58,
    "args": {
      "External id": 75621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75621,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 75621, "pid": 5, "tid": 7, "ts": 1716454222573556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222519994, "dur": 6,
    "args": {
      "External id": 75621,
      "cbid": 211, "correlation": 75621
    }
  },
  {
    "ph": "s", "id": 75621, "pid": 76337, "tid": -914061504, "ts": 1716454222519994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222573616, "dur": 15,
    "args": {
      "External id": 75626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75626, "pid": 5, "tid": 7, "ts": 1716454222573616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520021, "dur": 8,
    "args": {
      "External id": 75626,
      "cbid": 211, "correlation": 75626
    }
  },
  {
    "ph": "s", "id": 75626, "pid": 76337, "tid": -914061504, "ts": 1716454222520021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222573632, "dur": 11,
    "args": {
      "External id": 75634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75634,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75634, "pid": 5, "tid": 7, "ts": 1716454222573632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520049, "dur": 8,
    "args": {
      "External id": 75634,
      "cbid": 211, "correlation": 75634
    }
  },
  {
    "ph": "s", "id": 75634, "pid": 76337, "tid": -914061504, "ts": 1716454222520049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222573644, "dur": 11,
    "args": {
      "External id": 75642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75642,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75642, "pid": 5, "tid": 7, "ts": 1716454222573644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520078, "dur": 8,
    "args": {
      "External id": 75642,
      "cbid": 211, "correlation": 75642
    }
  },
  {
    "ph": "s", "id": 75642, "pid": 76337, "tid": -914061504, "ts": 1716454222520078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222573657, "dur": 19,
    "args": {
      "External id": 75662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75662,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 75662, "pid": 5, "tid": 7, "ts": 1716454222573657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520158, "dur": 12,
    "args": {
      "External id": 75662,
      "cbid": 211, "correlation": 75662
    }
  },
  {
    "ph": "s", "id": 75662, "pid": 76337, "tid": -914061504, "ts": 1716454222520158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222573677, "dur": 4,
    "args": {
      "External id": 75674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75674,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 75674, "pid": 5, "tid": 7, "ts": 1716454222573677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520180, "dur": 6,
    "args": {
      "External id": 75674,
      "cbid": 211, "correlation": 75674
    }
  },
  {
    "ph": "s", "id": 75674, "pid": 76337, "tid": -914061504, "ts": 1716454222520180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222573683, "dur": 17,
    "args": {
      "External id": 75677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75677, "pid": 5, "tid": 7, "ts": 1716454222573683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520197, "dur": 6,
    "args": {
      "External id": 75677,
      "cbid": 211, "correlation": 75677
    }
  },
  {
    "ph": "s", "id": 75677, "pid": 76337, "tid": -914061504, "ts": 1716454222520197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520254, "dur": 0,
    "args": {
      "External id": 75688,
      "cbid": 317, "correlation": 75688
    }
  },
  {
    "ph": "f", "id": 75688, "pid": 76337, "tid": -914061504, "ts": 1716454222520254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520255, "dur": 0,
    "args": {
      "External id": 75689,
      "cbid": 203, "correlation": 75689
    }
  },
  {
    "ph": "f", "id": 75689, "pid": 76337, "tid": -914061504, "ts": 1716454222520255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520256, "dur": 0,
    "args": {
      "External id": 75690,
      "cbid": 205, "correlation": 75690
    }
  },
  {
    "ph": "f", "id": 75690, "pid": 76337, "tid": -914061504, "ts": 1716454222520256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222573701, "dur": 12,
    "args": {
      "External id": 75694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75694,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75694, "pid": 5, "tid": 7, "ts": 1716454222573701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520273, "dur": 11,
    "args": {
      "External id": 75694,
      "cbid": 211, "correlation": 75694
    }
  },
  {
    "ph": "s", "id": 75694, "pid": 76337, "tid": -914061504, "ts": 1716454222520273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222573714, "dur": 3,
    "args": {
      "External id": 75696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 75696, "pid": 5, "tid": 7, "ts": 1716454222573714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520289, "dur": 5,
    "args": {
      "External id": 75696,
      "cbid": 211, "correlation": 75696
    }
  },
  {
    "ph": "s", "id": 75696, "pid": 76337, "tid": -914061504, "ts": 1716454222520289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520297, "dur": 0,
    "args": {
      "External id": 75697,
      "cbid": 51, "correlation": 75697
    }
  },
  {
    "ph": "s", "id": 75697, "pid": 76337, "tid": -914061504, "ts": 1716454222520297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222573719, "dur": 99,
    "args": {
      "External id": 75698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75698,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 75698, "pid": 5, "tid": 7, "ts": 1716454222573719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520298, "dur": 7,
    "args": {
      "External id": 75698,
      "cbid": 211, "correlation": 75698
    }
  },
  {
    "ph": "s", "id": 75698, "pid": 76337, "tid": -914061504, "ts": 1716454222520298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222573819, "dur": 16,
    "args": {
      "External id": 75703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75703, "pid": 5, "tid": 7, "ts": 1716454222573819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520328, "dur": 8,
    "args": {
      "External id": 75703,
      "cbid": 211, "correlation": 75703
    }
  },
  {
    "ph": "s", "id": 75703, "pid": 76337, "tid": -914061504, "ts": 1716454222520328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222573836, "dur": 86,
    "args": {
      "External id": 75712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75712,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75712, "pid": 5, "tid": 7, "ts": 1716454222573836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520410, "dur": 14,
    "args": {
      "External id": 75712,
      "cbid": 211, "correlation": 75712
    }
  },
  {
    "ph": "s", "id": 75712, "pid": 76337, "tid": -914061504, "ts": 1716454222520410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222573923, "dur": 31,
    "args": {
      "External id": 75734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75734,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75734, "pid": 5, "tid": 7, "ts": 1716454222573923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520466, "dur": 10,
    "args": {
      "External id": 75734,
      "cbid": 211, "correlation": 75734
    }
  },
  {
    "ph": "s", "id": 75734, "pid": 76337, "tid": -914061504, "ts": 1716454222520466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520555, "dur": 1,
    "args": {
      "External id": 75745,
      "cbid": 251, "correlation": 75745
    }
  },
  {
    "ph": "f", "id": 75745, "pid": 76337, "tid": -914061504, "ts": 1716454222520555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222573956, "dur": 166,
    "args": {
      "External id": 75746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75746,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75746, "pid": 5, "tid": 7, "ts": 1716454222573956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520560, "dur": 13,
    "args": {
      "External id": 75746,
      "cbid": 211, "correlation": 75746
    }
  },
  {
    "ph": "s", "id": 75746, "pid": 76337, "tid": -914061504, "ts": 1716454222520560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520630, "dur": 1,
    "args": {
      "External id": 75757,
      "cbid": 251, "correlation": 75757
    }
  },
  {
    "ph": "f", "id": 75757, "pid": 76337, "tid": -914061504, "ts": 1716454222520630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222574123, "dur": 161,
    "args": {
      "External id": 75758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75758,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75758, "pid": 5, "tid": 7, "ts": 1716454222574123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520634, "dur": 11,
    "args": {
      "External id": 75758,
      "cbid": 211, "correlation": 75758
    }
  },
  {
    "ph": "s", "id": 75758, "pid": 76337, "tid": -914061504, "ts": 1716454222520634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520699, "dur": 1,
    "args": {
      "External id": 75769,
      "cbid": 251, "correlation": 75769
    }
  },
  {
    "ph": "f", "id": 75769, "pid": 76337, "tid": -914061504, "ts": 1716454222520699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222574286, "dur": 160,
    "args": {
      "External id": 75770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75770,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75770, "pid": 5, "tid": 7, "ts": 1716454222574286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520703, "dur": 11,
    "args": {
      "External id": 75770,
      "cbid": 211, "correlation": 75770
    }
  },
  {
    "ph": "s", "id": 75770, "pid": 76337, "tid": -914061504, "ts": 1716454222520703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222574448, "dur": 342,
    "args": {
      "External id": 75795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75795,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75795, "pid": 5, "tid": 7, "ts": 1716454222574448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520790, "dur": 12,
    "args": {
      "External id": 75795,
      "cbid": 211, "correlation": 75795
    }
  },
  {
    "ph": "s", "id": 75795, "pid": 76337, "tid": -914061504, "ts": 1716454222520790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520891, "dur": 1,
    "args": {
      "External id": 75813,
      "cbid": 251, "correlation": 75813
    }
  },
  {
    "ph": "f", "id": 75813, "pid": 76337, "tid": -914061504, "ts": 1716454222520891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222574792, "dur": 168,
    "args": {
      "External id": 75815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75815,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75815, "pid": 5, "tid": 7, "ts": 1716454222574792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520897, "dur": 14,
    "args": {
      "External id": 75815,
      "cbid": 211, "correlation": 75815
    }
  },
  {
    "ph": "s", "id": 75815, "pid": 76337, "tid": -914061504, "ts": 1716454222520897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222574961, "dur": 19,
    "args": {
      "External id": 75823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75823,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75823, "pid": 5, "tid": 7, "ts": 1716454222574961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222520967, "dur": 19,
    "args": {
      "External id": 75823,
      "cbid": 211, "correlation": 75823
    }
  },
  {
    "ph": "s", "id": 75823, "pid": 76337, "tid": -914061504, "ts": 1716454222520967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222574981, "dur": 28,
    "args": {
      "External id": 75831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75831,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75831, "pid": 5, "tid": 7, "ts": 1716454222574981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521016, "dur": 9,
    "args": {
      "External id": 75831,
      "cbid": 211, "correlation": 75831
    }
  },
  {
    "ph": "s", "id": 75831, "pid": 76337, "tid": -914061504, "ts": 1716454222521016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222575010, "dur": 18,
    "args": {
      "External id": 75842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75842, "pid": 5, "tid": 7, "ts": 1716454222575010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521088, "dur": 12,
    "args": {
      "External id": 75842,
      "cbid": 211, "correlation": 75842
    }
  },
  {
    "ph": "s", "id": 75842, "pid": 76337, "tid": -914061504, "ts": 1716454222521088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222575029, "dur": 17,
    "args": {
      "External id": 75864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75864,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75864, "pid": 5, "tid": 7, "ts": 1716454222575029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521120, "dur": 7,
    "args": {
      "External id": 75864,
      "cbid": 211, "correlation": 75864
    }
  },
  {
    "ph": "s", "id": 75864, "pid": 76337, "tid": -914061504, "ts": 1716454222521120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521205, "dur": 2,
    "args": {
      "External id": 75875,
      "cbid": 251, "correlation": 75875
    }
  },
  {
    "ph": "f", "id": 75875, "pid": 76337, "tid": -914061504, "ts": 1716454222521205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222575047, "dur": 91,
    "args": {
      "External id": 75876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75876,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 75876, "pid": 5, "tid": 7, "ts": 1716454222575047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521212, "dur": 14,
    "args": {
      "External id": 75876,
      "cbid": 211, "correlation": 75876
    }
  },
  {
    "ph": "s", "id": 75876, "pid": 76337, "tid": -914061504, "ts": 1716454222521212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521282, "dur": 1,
    "args": {
      "External id": 75887,
      "cbid": 251, "correlation": 75887
    }
  },
  {
    "ph": "f", "id": 75887, "pid": 76337, "tid": -914061504, "ts": 1716454222521282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521286, "dur": 0,
    "args": {
      "External id": 75888,
      "cbid": 251, "correlation": 75888
    }
  },
  {
    "ph": "f", "id": 75888, "pid": 76337, "tid": -914061504, "ts": 1716454222521286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222575140, "dur": 13,
    "args": {
      "External id": 75889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75889,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75889, "pid": 5, "tid": 7, "ts": 1716454222575140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521288, "dur": 12,
    "args": {
      "External id": 75889,
      "cbid": 211, "correlation": 75889
    }
  },
  {
    "ph": "s", "id": 75889, "pid": 76337, "tid": -914061504, "ts": 1716454222521288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222575154, "dur": 6,
    "args": {
      "External id": 75891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75891,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75891, "pid": 5, "tid": 7, "ts": 1716454222575154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521301, "dur": 5,
    "args": {
      "External id": 75891,
      "cbid": 211, "correlation": 75891
    }
  },
  {
    "ph": "s", "id": 75891, "pid": 76337, "tid": -914061504, "ts": 1716454222521301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521358, "dur": 1,
    "args": {
      "External id": 75902,
      "cbid": 251, "correlation": 75902
    }
  },
  {
    "ph": "f", "id": 75902, "pid": 76337, "tid": -914061504, "ts": 1716454222521358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521361, "dur": 0,
    "args": {
      "External id": 75903,
      "cbid": 251, "correlation": 75903
    }
  },
  {
    "ph": "f", "id": 75903, "pid": 76337, "tid": -914061504, "ts": 1716454222521361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222575161, "dur": 8,
    "args": {
      "External id": 75904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75904,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75904, "pid": 5, "tid": 7, "ts": 1716454222575161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521363, "dur": 12,
    "args": {
      "External id": 75904,
      "cbid": 211, "correlation": 75904
    }
  },
  {
    "ph": "s", "id": 75904, "pid": 76337, "tid": -914061504, "ts": 1716454222521363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222575171, "dur": 3,
    "args": {
      "External id": 75906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75906,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75906, "pid": 5, "tid": 7, "ts": 1716454222575171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521376, "dur": 5,
    "args": {
      "External id": 75906,
      "cbid": 211, "correlation": 75906
    }
  },
  {
    "ph": "s", "id": 75906, "pid": 76337, "tid": -914061504, "ts": 1716454222521376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222575175, "dur": 56,
    "args": {
      "External id": 75931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75931,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 75931, "pid": 5, "tid": 7, "ts": 1716454222575175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521452, "dur": 12,
    "args": {
      "External id": 75931,
      "cbid": 211, "correlation": 75931
    }
  },
  {
    "ph": "s", "id": 75931, "pid": 76337, "tid": -914061504, "ts": 1716454222521452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521551, "dur": 2,
    "args": {
      "External id": 75949,
      "cbid": 251, "correlation": 75949
    }
  },
  {
    "ph": "f", "id": 75949, "pid": 76337, "tid": -914061504, "ts": 1716454222521551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222575233, "dur": 93,
    "args": {
      "External id": 75951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75951,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 75951, "pid": 5, "tid": 7, "ts": 1716454222575233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521558, "dur": 14,
    "args": {
      "External id": 75951,
      "cbid": 211, "correlation": 75951
    }
  },
  {
    "ph": "s", "id": 75951, "pid": 76337, "tid": -914061504, "ts": 1716454222521558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222575327, "dur": 10,
    "args": {
      "External id": 75959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75959,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75959, "pid": 5, "tid": 7, "ts": 1716454222575327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521627, "dur": 12,
    "args": {
      "External id": 75959,
      "cbid": 211, "correlation": 75959
    }
  },
  {
    "ph": "s", "id": 75959, "pid": 76337, "tid": -914061504, "ts": 1716454222521627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222575338, "dur": 21,
    "args": {
      "External id": 75967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75967,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75967, "pid": 5, "tid": 7, "ts": 1716454222575338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521669, "dur": 10,
    "args": {
      "External id": 75967,
      "cbid": 211, "correlation": 75967
    }
  },
  {
    "ph": "s", "id": 75967, "pid": 76337, "tid": -914061504, "ts": 1716454222521669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222575360, "dur": 18,
    "args": {
      "External id": 75989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 75989,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 75989, "pid": 5, "tid": 7, "ts": 1716454222575360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521720, "dur": 10,
    "args": {
      "External id": 75989,
      "cbid": 211, "correlation": 75989
    }
  },
  {
    "ph": "s", "id": 75989, "pid": 76337, "tid": -914061504, "ts": 1716454222521720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521809, "dur": 2,
    "args": {
      "External id": 76005,
      "cbid": 251, "correlation": 76005
    }
  },
  {
    "ph": "f", "id": 76005, "pid": 76337, "tid": -914061504, "ts": 1716454222521809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521815, "dur": 0,
    "args": {
      "External id": 76007,
      "cbid": 251, "correlation": 76007
    }
  },
  {
    "ph": "f", "id": 76007, "pid": 76337, "tid": -914061504, "ts": 1716454222521815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222575380, "dur": 496,
    "args": {
      "External id": 76008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76008,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76008, "pid": 5, "tid": 7, "ts": 1716454222575380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521818, "dur": 14,
    "args": {
      "External id": 76008,
      "cbid": 211, "correlation": 76008
    }
  },
  {
    "ph": "s", "id": 76008, "pid": 76337, "tid": -914061504, "ts": 1716454222521818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222575877, "dur": 68,
    "args": {
      "External id": 76016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76016, "pid": 5, "tid": 7, "ts": 1716454222575877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521884, "dur": 13,
    "args": {
      "External id": 76016,
      "cbid": 211, "correlation": 76016
    }
  },
  {
    "ph": "s", "id": 76016, "pid": 76337, "tid": -914061504, "ts": 1716454222521884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222575947, "dur": 69,
    "args": {
      "External id": 76024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76024,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76024, "pid": 5, "tid": 7, "ts": 1716454222575947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222521916, "dur": 8,
    "args": {
      "External id": 76024,
      "cbid": 211, "correlation": 76024
    }
  },
  {
    "ph": "s", "id": 76024, "pid": 76337, "tid": -914061504, "ts": 1716454222521916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522004, "dur": 1,
    "args": {
      "External id": 76040,
      "cbid": 251, "correlation": 76040
    }
  },
  {
    "ph": "f", "id": 76040, "pid": 76337, "tid": -914061504, "ts": 1716454222522004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222576018, "dur": 1,
    "args": {
      "External id": 76042,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76042,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 76042, "pid": 5, "tid": 7, "ts": 1716454222576018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522010, "dur": 12,
    "args": {
      "External id": 76042,
      "cbid": 51, "correlation": 76042
    }
  },
  {
    "ph": "s", "id": 76042, "pid": 76337, "tid": -914061504, "ts": 1716454222522010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222576021, "dur": 272,
    "args": {
      "External id": 76043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76043,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76043, "pid": 5, "tid": 7, "ts": 1716454222576021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522024, "dur": 11,
    "args": {
      "External id": 76043,
      "cbid": 211, "correlation": 76043
    }
  },
  {
    "ph": "s", "id": 76043, "pid": 76337, "tid": -914061504, "ts": 1716454222522024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222576295, "dur": 14,
    "args": {
      "External id": 76051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76051,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76051, "pid": 5, "tid": 7, "ts": 1716454222576295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522066, "dur": 11,
    "args": {
      "External id": 76051,
      "cbid": 211, "correlation": 76051
    }
  },
  {
    "ph": "s", "id": 76051, "pid": 76337, "tid": -914061504, "ts": 1716454222522066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222576311, "dur": 39,
    "args": {
      "External id": 76062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76062, "pid": 5, "tid": 7, "ts": 1716454222576311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522134, "dur": 12,
    "args": {
      "External id": 76062,
      "cbid": 211, "correlation": 76062
    }
  },
  {
    "ph": "s", "id": 76062, "pid": 76337, "tid": -914061504, "ts": 1716454222522134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522199, "dur": 0,
    "args": {
      "External id": 76074,
      "cbid": 317, "correlation": 76074
    }
  },
  {
    "ph": "f", "id": 76074, "pid": 76337, "tid": -914061504, "ts": 1716454222522199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522199, "dur": 0,
    "args": {
      "External id": 76075,
      "cbid": 203, "correlation": 76075
    }
  },
  {
    "ph": "f", "id": 76075, "pid": 76337, "tid": -914061504, "ts": 1716454222522199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522200, "dur": 0,
    "args": {
      "External id": 76076,
      "cbid": 205, "correlation": 76076
    }
  },
  {
    "ph": "f", "id": 76076, "pid": 76337, "tid": -914061504, "ts": 1716454222522200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222576351, "dur": 13,
    "args": {
      "External id": 76080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76080,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76080, "pid": 5, "tid": 7, "ts": 1716454222576351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522217, "dur": 12,
    "args": {
      "External id": 76080,
      "cbid": 211, "correlation": 76080
    }
  },
  {
    "ph": "s", "id": 76080, "pid": 76337, "tid": -914061504, "ts": 1716454222522217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222576365, "dur": 4,
    "args": {
      "External id": 76082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76082,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76082, "pid": 5, "tid": 7, "ts": 1716454222576365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522234, "dur": 6,
    "args": {
      "External id": 76082,
      "cbid": 211, "correlation": 76082
    }
  },
  {
    "ph": "s", "id": 76082, "pid": 76337, "tid": -914061504, "ts": 1716454222522234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522242, "dur": 0,
    "args": {
      "External id": 76083,
      "cbid": 51, "correlation": 76083
    }
  },
  {
    "ph": "s", "id": 76083, "pid": 76337, "tid": -914061504, "ts": 1716454222522242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222576371, "dur": 99,
    "args": {
      "External id": 76084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76084,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 76084, "pid": 5, "tid": 7, "ts": 1716454222576371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522243, "dur": 5,
    "args": {
      "External id": 76084,
      "cbid": 211, "correlation": 76084
    }
  },
  {
    "ph": "s", "id": 76084, "pid": 76337, "tid": -914061504, "ts": 1716454222522243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222576471, "dur": 17,
    "args": {
      "External id": 76089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76089, "pid": 5, "tid": 7, "ts": 1716454222576471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522271, "dur": 8,
    "args": {
      "External id": 76089,
      "cbid": 211, "correlation": 76089
    }
  },
  {
    "ph": "s", "id": 76089, "pid": 76337, "tid": -914061504, "ts": 1716454222522271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222576490, "dur": 13,
    "args": {
      "External id": 76097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76097,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76097, "pid": 5, "tid": 7, "ts": 1716454222576490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522301, "dur": 8,
    "args": {
      "External id": 76097,
      "cbid": 211, "correlation": 76097
    }
  },
  {
    "ph": "s", "id": 76097, "pid": 76337, "tid": -914061504, "ts": 1716454222522301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222576504, "dur": 18,
    "args": {
      "External id": 76117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76117,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 76117, "pid": 5, "tid": 7, "ts": 1716454222576504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522374, "dur": 12,
    "args": {
      "External id": 76117,
      "cbid": 211, "correlation": 76117
    }
  },
  {
    "ph": "s", "id": 76117, "pid": 76337, "tid": -914061504, "ts": 1716454222522374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222576524, "dur": 5,
    "args": {
      "External id": 76129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76129,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 76129, "pid": 5, "tid": 7, "ts": 1716454222576524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522397, "dur": 6,
    "args": {
      "External id": 76129,
      "cbid": 211, "correlation": 76129
    }
  },
  {
    "ph": "s", "id": 76129, "pid": 76337, "tid": -914061504, "ts": 1716454222522397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222576530, "dur": 18,
    "args": {
      "External id": 76132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76132, "pid": 5, "tid": 7, "ts": 1716454222576530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522414, "dur": 7,
    "args": {
      "External id": 76132,
      "cbid": 211, "correlation": 76132
    }
  },
  {
    "ph": "s", "id": 76132, "pid": 76337, "tid": -914061504, "ts": 1716454222522414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222576549, "dur": 11,
    "args": {
      "External id": 76141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76141,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76141, "pid": 5, "tid": 7, "ts": 1716454222576549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522452, "dur": 10,
    "args": {
      "External id": 76141,
      "cbid": 211, "correlation": 76141
    }
  },
  {
    "ph": "s", "id": 76141, "pid": 76337, "tid": -914061504, "ts": 1716454222522452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522504, "dur": 0,
    "args": {
      "External id": 76151,
      "cbid": 317, "correlation": 76151
    }
  },
  {
    "ph": "f", "id": 76151, "pid": 76337, "tid": -914061504, "ts": 1716454222522504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522505, "dur": 0,
    "args": {
      "External id": 76152,
      "cbid": 203, "correlation": 76152
    }
  },
  {
    "ph": "f", "id": 76152, "pid": 76337, "tid": -914061504, "ts": 1716454222522505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522506, "dur": 0,
    "args": {
      "External id": 76153,
      "cbid": 205, "correlation": 76153
    }
  },
  {
    "ph": "f", "id": 76153, "pid": 76337, "tid": -914061504, "ts": 1716454222522506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222576562, "dur": 11,
    "args": {
      "External id": 76157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76157,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76157, "pid": 5, "tid": 7, "ts": 1716454222576562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522520, "dur": 11,
    "args": {
      "External id": 76157,
      "cbid": 211, "correlation": 76157
    }
  },
  {
    "ph": "s", "id": 76157, "pid": 76337, "tid": -914061504, "ts": 1716454222522520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222576575, "dur": 165,
    "args": {
      "External id": 76159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76159,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76159, "pid": 5, "tid": 7, "ts": 1716454222576575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522533, "dur": 6,
    "args": {
      "External id": 76159,
      "cbid": 211, "correlation": 76159
    }
  },
  {
    "ph": "s", "id": 76159, "pid": 76337, "tid": -914061504, "ts": 1716454222522533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222576742, "dur": 1,
    "args": {
      "External id": 76161,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76161,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 76161, "pid": 5, "tid": 7, "ts": 1716454222576742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522545, "dur": 6,
    "args": {
      "External id": 76161,
      "cbid": 51, "correlation": 76161
    }
  },
  {
    "ph": "s", "id": 76161, "pid": 76337, "tid": -914061504, "ts": 1716454222522545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222576746, "dur": 674,
    "args": {
      "External id": 76162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76162,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76162, "pid": 5, "tid": 7, "ts": 1716454222576746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522553, "dur": 6,
    "args": {
      "External id": 76162,
      "cbid": 211, "correlation": 76162
    }
  },
  {
    "ph": "s", "id": 76162, "pid": 76337, "tid": -914061504, "ts": 1716454222522553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222577421, "dur": 13,
    "args": {
      "External id": 76164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76164,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76164, "pid": 5, "tid": 7, "ts": 1716454222577421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522563, "dur": 5,
    "args": {
      "External id": 76164,
      "cbid": 211, "correlation": 76164
    }
  },
  {
    "ph": "s", "id": 76164, "pid": 76337, "tid": -914061504, "ts": 1716454222522563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222577435, "dur": 15,
    "args": {
      "External id": 76170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76170,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76170, "pid": 5, "tid": 7, "ts": 1716454222577435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522591, "dur": 9,
    "args": {
      "External id": 76170,
      "cbid": 211, "correlation": 76170
    }
  },
  {
    "ph": "s", "id": 76170, "pid": 76337, "tid": -914061504, "ts": 1716454222522591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222577451, "dur": 3,
    "args": {
      "External id": 76178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76178,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 76178, "pid": 5, "tid": 7, "ts": 1716454222577451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522635, "dur": 9,
    "args": {
      "External id": 76178,
      "cbid": 211, "correlation": 76178
    }
  },
  {
    "ph": "s", "id": 76178, "pid": 76337, "tid": -914061504, "ts": 1716454222522635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522699, "dur": 1,
    "args": {
      "External id": 76194,
      "cbid": 251, "correlation": 76194
    }
  },
  {
    "ph": "f", "id": 76194, "pid": 76337, "tid": -914061504, "ts": 1716454222522699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522704, "dur": 0,
    "args": {
      "External id": 76196,
      "cbid": 251, "correlation": 76196
    }
  },
  {
    "ph": "f", "id": 76196, "pid": 76337, "tid": -914061504, "ts": 1716454222522704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222577456, "dur": 13,
    "args": {
      "External id": 76197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76197,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76197, "pid": 5, "tid": 7, "ts": 1716454222577456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522706, "dur": 11,
    "args": {
      "External id": 76197,
      "cbid": 211, "correlation": 76197
    }
  },
  {
    "ph": "s", "id": 76197, "pid": 76337, "tid": -914061504, "ts": 1716454222522706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222577471, "dur": 5,
    "args": {
      "External id": 76199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76199,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76199, "pid": 5, "tid": 7, "ts": 1716454222577471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522719, "dur": 5,
    "args": {
      "External id": 76199,
      "cbid": 211, "correlation": 76199
    }
  },
  {
    "ph": "s", "id": 76199, "pid": 76337, "tid": -914061504, "ts": 1716454222522719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222577477, "dur": 17,
    "args": {
      "External id": 76209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76209, "pid": 5, "tid": 7, "ts": 1716454222577477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522777, "dur": 12,
    "args": {
      "External id": 76209,
      "cbid": 211, "correlation": 76209
    }
  },
  {
    "ph": "s", "id": 76209, "pid": 76337, "tid": -914061504, "ts": 1716454222522777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222577496, "dur": 18,
    "args": {
      "External id": 76229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76229,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 76229, "pid": 5, "tid": 7, "ts": 1716454222577496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522842, "dur": 11,
    "args": {
      "External id": 76229,
      "cbid": 211, "correlation": 76229
    }
  },
  {
    "ph": "s", "id": 76229, "pid": 76337, "tid": -914061504, "ts": 1716454222522842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222577515, "dur": 4,
    "args": {
      "External id": 76241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76241,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 76241, "pid": 5, "tid": 7, "ts": 1716454222577515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522862, "dur": 6,
    "args": {
      "External id": 76241,
      "cbid": 211, "correlation": 76241
    }
  },
  {
    "ph": "s", "id": 76241, "pid": 76337, "tid": -914061504, "ts": 1716454222522862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222577520, "dur": 17,
    "args": {
      "External id": 76244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76244, "pid": 5, "tid": 7, "ts": 1716454222577520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522881, "dur": 6,
    "args": {
      "External id": 76244,
      "cbid": 211, "correlation": 76244
    }
  },
  {
    "ph": "s", "id": 76244, "pid": 76337, "tid": -914061504, "ts": 1716454222522881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222577538, "dur": 11,
    "args": {
      "External id": 76253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76253,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76253, "pid": 5, "tid": 7, "ts": 1716454222577538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522920, "dur": 9,
    "args": {
      "External id": 76253,
      "cbid": 211, "correlation": 76253
    }
  },
  {
    "ph": "s", "id": 76253, "pid": 76337, "tid": -914061504, "ts": 1716454222522920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522988, "dur": 0,
    "args": {
      "External id": 76263,
      "cbid": 317, "correlation": 76263
    }
  },
  {
    "ph": "f", "id": 76263, "pid": 76337, "tid": -914061504, "ts": 1716454222522988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522988, "dur": 0,
    "args": {
      "External id": 76264,
      "cbid": 203, "correlation": 76264
    }
  },
  {
    "ph": "f", "id": 76264, "pid": 76337, "tid": -914061504, "ts": 1716454222522988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222522989, "dur": 0,
    "args": {
      "External id": 76265,
      "cbid": 205, "correlation": 76265
    }
  },
  {
    "ph": "f", "id": 76265, "pid": 76337, "tid": -914061504, "ts": 1716454222522989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222577550, "dur": 11,
    "args": {
      "External id": 76269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76269,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76269, "pid": 5, "tid": 7, "ts": 1716454222577550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523004, "dur": 12,
    "args": {
      "External id": 76269,
      "cbid": 211, "correlation": 76269
    }
  },
  {
    "ph": "s", "id": 76269, "pid": 76337, "tid": -914061504, "ts": 1716454222523004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222577563, "dur": 166,
    "args": {
      "External id": 76271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76271, "pid": 5, "tid": 7, "ts": 1716454222577563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523019, "dur": 5,
    "args": {
      "External id": 76271,
      "cbid": 211, "correlation": 76271
    }
  },
  {
    "ph": "s", "id": 76271, "pid": 76337, "tid": -914061504, "ts": 1716454222523019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222577731, "dur": 1,
    "args": {
      "External id": 76273,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76273,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 76273, "pid": 5, "tid": 7, "ts": 1716454222577731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523030, "dur": 6,
    "args": {
      "External id": 76273,
      "cbid": 51, "correlation": 76273
    }
  },
  {
    "ph": "s", "id": 76273, "pid": 76337, "tid": -914061504, "ts": 1716454222523030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222577735, "dur": 659,
    "args": {
      "External id": 76274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76274,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76274, "pid": 5, "tid": 7, "ts": 1716454222577735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523037, "dur": 6,
    "args": {
      "External id": 76274,
      "cbid": 211, "correlation": 76274
    }
  },
  {
    "ph": "s", "id": 76274, "pid": 76337, "tid": -914061504, "ts": 1716454222523037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222578396, "dur": 12,
    "args": {
      "External id": 76276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76276,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76276, "pid": 5, "tid": 7, "ts": 1716454222578396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523047, "dur": 5,
    "args": {
      "External id": 76276,
      "cbid": 211, "correlation": 76276
    }
  },
  {
    "ph": "s", "id": 76276, "pid": 76337, "tid": -914061504, "ts": 1716454222523047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222578410, "dur": 15,
    "args": {
      "External id": 76282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76282, "pid": 5, "tid": 7, "ts": 1716454222578410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523075, "dur": 8,
    "args": {
      "External id": 76282,
      "cbid": 211, "correlation": 76282
    }
  },
  {
    "ph": "s", "id": 76282, "pid": 76337, "tid": -914061504, "ts": 1716454222523075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222578426, "dur": 12,
    "args": {
      "External id": 76290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76290,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76290, "pid": 5, "tid": 7, "ts": 1716454222578426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523107, "dur": 9,
    "args": {
      "External id": 76290,
      "cbid": 211, "correlation": 76290
    }
  },
  {
    "ph": "s", "id": 76290, "pid": 76337, "tid": -914061504, "ts": 1716454222523107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222578439, "dur": 10,
    "args": {
      "External id": 76298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76298,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76298, "pid": 5, "tid": 7, "ts": 1716454222578439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523136, "dur": 8,
    "args": {
      "External id": 76298,
      "cbid": 211, "correlation": 76298
    }
  },
  {
    "ph": "s", "id": 76298, "pid": 76337, "tid": -914061504, "ts": 1716454222523136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222578451, "dur": 19,
    "args": {
      "External id": 76318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76318,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 76318, "pid": 5, "tid": 7, "ts": 1716454222578451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523213, "dur": 12,
    "args": {
      "External id": 76318,
      "cbid": 211, "correlation": 76318
    }
  },
  {
    "ph": "s", "id": 76318, "pid": 76337, "tid": -914061504, "ts": 1716454222523213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222578471, "dur": 4,
    "args": {
      "External id": 76330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76330,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 76330, "pid": 5, "tid": 7, "ts": 1716454222578471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523235, "dur": 7,
    "args": {
      "External id": 76330,
      "cbid": 211, "correlation": 76330
    }
  },
  {
    "ph": "s", "id": 76330, "pid": 76337, "tid": -914061504, "ts": 1716454222523235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222578476, "dur": 17,
    "args": {
      "External id": 76333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76333, "pid": 5, "tid": 7, "ts": 1716454222578476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523254, "dur": 6,
    "args": {
      "External id": 76333,
      "cbid": 211, "correlation": 76333
    }
  },
  {
    "ph": "s", "id": 76333, "pid": 76337, "tid": -914061504, "ts": 1716454222523254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523310, "dur": 0,
    "args": {
      "External id": 76344,
      "cbid": 317, "correlation": 76344
    }
  },
  {
    "ph": "f", "id": 76344, "pid": 76337, "tid": -914061504, "ts": 1716454222523310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523311, "dur": 0,
    "args": {
      "External id": 76345,
      "cbid": 203, "correlation": 76345
    }
  },
  {
    "ph": "f", "id": 76345, "pid": 76337, "tid": -914061504, "ts": 1716454222523311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523311, "dur": 0,
    "args": {
      "External id": 76346,
      "cbid": 205, "correlation": 76346
    }
  },
  {
    "ph": "f", "id": 76346, "pid": 76337, "tid": -914061504, "ts": 1716454222523311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222578494, "dur": 11,
    "args": {
      "External id": 76350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76350, "pid": 5, "tid": 7, "ts": 1716454222578494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523324, "dur": 12,
    "args": {
      "External id": 76350,
      "cbid": 211, "correlation": 76350
    }
  },
  {
    "ph": "s", "id": 76350, "pid": 76337, "tid": -914061504, "ts": 1716454222523324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222578506, "dur": 4,
    "args": {
      "External id": 76352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76352, "pid": 5, "tid": 7, "ts": 1716454222578506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523340, "dur": 6,
    "args": {
      "External id": 76352,
      "cbid": 211, "correlation": 76352
    }
  },
  {
    "ph": "s", "id": 76352, "pid": 76337, "tid": -914061504, "ts": 1716454222523340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523349, "dur": 0,
    "args": {
      "External id": 76353,
      "cbid": 51, "correlation": 76353
    }
  },
  {
    "ph": "s", "id": 76353, "pid": 76337, "tid": -914061504, "ts": 1716454222523349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222578512, "dur": 96,
    "args": {
      "External id": 76354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76354,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 76354, "pid": 5, "tid": 7, "ts": 1716454222578512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523350, "dur": 5,
    "args": {
      "External id": 76354,
      "cbid": 211, "correlation": 76354
    }
  },
  {
    "ph": "s", "id": 76354, "pid": 76337, "tid": -914061504, "ts": 1716454222523350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222578609, "dur": 16,
    "args": {
      "External id": 76359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76359, "pid": 5, "tid": 7, "ts": 1716454222578609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523376, "dur": 8,
    "args": {
      "External id": 76359,
      "cbid": 211, "correlation": 76359
    }
  },
  {
    "ph": "s", "id": 76359, "pid": 76337, "tid": -914061504, "ts": 1716454222523376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222578627, "dur": 84,
    "args": {
      "External id": 76368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76368,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76368, "pid": 5, "tid": 7, "ts": 1716454222578627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523458, "dur": 14,
    "args": {
      "External id": 76368,
      "cbid": 211, "correlation": 76368
    }
  },
  {
    "ph": "s", "id": 76368, "pid": 76337, "tid": -914061504, "ts": 1716454222523458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222578713, "dur": 31,
    "args": {
      "External id": 76390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76390,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76390, "pid": 5, "tid": 7, "ts": 1716454222578713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523514, "dur": 10,
    "args": {
      "External id": 76390,
      "cbid": 211, "correlation": 76390
    }
  },
  {
    "ph": "s", "id": 76390, "pid": 76337, "tid": -914061504, "ts": 1716454222523514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523602, "dur": 1,
    "args": {
      "External id": 76401,
      "cbid": 251, "correlation": 76401
    }
  },
  {
    "ph": "f", "id": 76401, "pid": 76337, "tid": -914061504, "ts": 1716454222523602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222578746, "dur": 166,
    "args": {
      "External id": 76402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76402,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76402, "pid": 5, "tid": 7, "ts": 1716454222578746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523607, "dur": 13,
    "args": {
      "External id": 76402,
      "cbid": 211, "correlation": 76402
    }
  },
  {
    "ph": "s", "id": 76402, "pid": 76337, "tid": -914061504, "ts": 1716454222523607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523677, "dur": 1,
    "args": {
      "External id": 76413,
      "cbid": 251, "correlation": 76413
    }
  },
  {
    "ph": "f", "id": 76413, "pid": 76337, "tid": -914061504, "ts": 1716454222523677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222578914, "dur": 160,
    "args": {
      "External id": 76414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76414,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76414, "pid": 5, "tid": 7, "ts": 1716454222578914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523681, "dur": 12,
    "args": {
      "External id": 76414,
      "cbid": 211, "correlation": 76414
    }
  },
  {
    "ph": "s", "id": 76414, "pid": 76337, "tid": -914061504, "ts": 1716454222523681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523745, "dur": 1,
    "args": {
      "External id": 76425,
      "cbid": 251, "correlation": 76425
    }
  },
  {
    "ph": "f", "id": 76425, "pid": 76337, "tid": -914061504, "ts": 1716454222523745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222579075, "dur": 163,
    "args": {
      "External id": 76426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76426,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76426, "pid": 5, "tid": 7, "ts": 1716454222579075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523749, "dur": 11,
    "args": {
      "External id": 76426,
      "cbid": 211, "correlation": 76426
    }
  },
  {
    "ph": "s", "id": 76426, "pid": 76337, "tid": -914061504, "ts": 1716454222523749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222579240, "dur": 343,
    "args": {
      "External id": 76451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76451,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76451, "pid": 5, "tid": 7, "ts": 1716454222579240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523831, "dur": 13,
    "args": {
      "External id": 76451,
      "cbid": 211, "correlation": 76451
    }
  },
  {
    "ph": "s", "id": 76451, "pid": 76337, "tid": -914061504, "ts": 1716454222523831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523931, "dur": 1,
    "args": {
      "External id": 76469,
      "cbid": 251, "correlation": 76469
    }
  },
  {
    "ph": "f", "id": 76469, "pid": 76337, "tid": -914061504, "ts": 1716454222523931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222579584, "dur": 170,
    "args": {
      "External id": 76471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76471,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76471, "pid": 5, "tid": 7, "ts": 1716454222579584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222523936, "dur": 13,
    "args": {
      "External id": 76471,
      "cbid": 211, "correlation": 76471
    }
  },
  {
    "ph": "s", "id": 76471, "pid": 76337, "tid": -914061504, "ts": 1716454222523936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222579756, "dur": 19,
    "args": {
      "External id": 76479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76479,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76479, "pid": 5, "tid": 7, "ts": 1716454222579756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524015, "dur": 13,
    "args": {
      "External id": 76479,
      "cbid": 211, "correlation": 76479
    }
  },
  {
    "ph": "s", "id": 76479, "pid": 76337, "tid": -914061504, "ts": 1716454222524015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222579776, "dur": 28,
    "args": {
      "External id": 76487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76487,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76487, "pid": 5, "tid": 7, "ts": 1716454222579776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524055, "dur": 9,
    "args": {
      "External id": 76487,
      "cbid": 211, "correlation": 76487
    }
  },
  {
    "ph": "s", "id": 76487, "pid": 76337, "tid": -914061504, "ts": 1716454222524055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222579805, "dur": 19,
    "args": {
      "External id": 76498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76498, "pid": 5, "tid": 7, "ts": 1716454222579805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524127, "dur": 12,
    "args": {
      "External id": 76498,
      "cbid": 211, "correlation": 76498
    }
  },
  {
    "ph": "s", "id": 76498, "pid": 76337, "tid": -914061504, "ts": 1716454222524127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222579825, "dur": 16,
    "args": {
      "External id": 76520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76520,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76520, "pid": 5, "tid": 7, "ts": 1716454222579825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524158, "dur": 8,
    "args": {
      "External id": 76520,
      "cbid": 211, "correlation": 76520
    }
  },
  {
    "ph": "s", "id": 76520, "pid": 76337, "tid": -914061504, "ts": 1716454222524158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524244, "dur": 1,
    "args": {
      "External id": 76531,
      "cbid": 251, "correlation": 76531
    }
  },
  {
    "ph": "f", "id": 76531, "pid": 76337, "tid": -914061504, "ts": 1716454222524244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222579843, "dur": 90,
    "args": {
      "External id": 76532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76532,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76532, "pid": 5, "tid": 7, "ts": 1716454222579843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524249, "dur": 13,
    "args": {
      "External id": 76532,
      "cbid": 211, "correlation": 76532
    }
  },
  {
    "ph": "s", "id": 76532, "pid": 76337, "tid": -914061504, "ts": 1716454222524249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524318, "dur": 1,
    "args": {
      "External id": 76543,
      "cbid": 251, "correlation": 76543
    }
  },
  {
    "ph": "f", "id": 76543, "pid": 76337, "tid": -914061504, "ts": 1716454222524318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524321, "dur": 0,
    "args": {
      "External id": 76544,
      "cbid": 251, "correlation": 76544
    }
  },
  {
    "ph": "f", "id": 76544, "pid": 76337, "tid": -914061504, "ts": 1716454222524321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222579934, "dur": 12,
    "args": {
      "External id": 76545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76545,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76545, "pid": 5, "tid": 7, "ts": 1716454222579934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524323, "dur": 12,
    "args": {
      "External id": 76545,
      "cbid": 211, "correlation": 76545
    }
  },
  {
    "ph": "s", "id": 76545, "pid": 76337, "tid": -914061504, "ts": 1716454222524323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222579948, "dur": 5,
    "args": {
      "External id": 76547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76547,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76547, "pid": 5, "tid": 7, "ts": 1716454222579948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524337, "dur": 5,
    "args": {
      "External id": 76547,
      "cbid": 211, "correlation": 76547
    }
  },
  {
    "ph": "s", "id": 76547, "pid": 76337, "tid": -914061504, "ts": 1716454222524337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524394, "dur": 1,
    "args": {
      "External id": 76558,
      "cbid": 251, "correlation": 76558
    }
  },
  {
    "ph": "f", "id": 76558, "pid": 76337, "tid": -914061504, "ts": 1716454222524394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524397, "dur": 0,
    "args": {
      "External id": 76559,
      "cbid": 251, "correlation": 76559
    }
  },
  {
    "ph": "f", "id": 76559, "pid": 76337, "tid": -914061504, "ts": 1716454222524397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222579955, "dur": 8,
    "args": {
      "External id": 76560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76560,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76560, "pid": 5, "tid": 7, "ts": 1716454222579955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524399, "dur": 11,
    "args": {
      "External id": 76560,
      "cbid": 211, "correlation": 76560
    }
  },
  {
    "ph": "s", "id": 76560, "pid": 76337, "tid": -914061504, "ts": 1716454222524399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222579965, "dur": 3,
    "args": {
      "External id": 76562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76562,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76562, "pid": 5, "tid": 7, "ts": 1716454222579965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524411, "dur": 5,
    "args": {
      "External id": 76562,
      "cbid": 211, "correlation": 76562
    }
  },
  {
    "ph": "s", "id": 76562, "pid": 76337, "tid": -914061504, "ts": 1716454222524411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222579969, "dur": 57,
    "args": {
      "External id": 76587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76587,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76587, "pid": 5, "tid": 7, "ts": 1716454222579969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524487, "dur": 13,
    "args": {
      "External id": 76587,
      "cbid": 211, "correlation": 76587
    }
  },
  {
    "ph": "s", "id": 76587, "pid": 76337, "tid": -914061504, "ts": 1716454222524487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524586, "dur": 1,
    "args": {
      "External id": 76605,
      "cbid": 251, "correlation": 76605
    }
  },
  {
    "ph": "f", "id": 76605, "pid": 76337, "tid": -914061504, "ts": 1716454222524586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222580027, "dur": 93,
    "args": {
      "External id": 76607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76607,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76607, "pid": 5, "tid": 7, "ts": 1716454222580027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524592, "dur": 14,
    "args": {
      "External id": 76607,
      "cbid": 211, "correlation": 76607
    }
  },
  {
    "ph": "s", "id": 76607, "pid": 76337, "tid": -914061504, "ts": 1716454222524592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222580122, "dur": 10,
    "args": {
      "External id": 76615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76615,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76615, "pid": 5, "tid": 7, "ts": 1716454222580122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524662, "dur": 12,
    "args": {
      "External id": 76615,
      "cbid": 211, "correlation": 76615
    }
  },
  {
    "ph": "s", "id": 76615, "pid": 76337, "tid": -914061504, "ts": 1716454222524662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222580133, "dur": 22,
    "args": {
      "External id": 76623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76623,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76623, "pid": 5, "tid": 7, "ts": 1716454222580133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524703, "dur": 9,
    "args": {
      "External id": 76623,
      "cbid": 211, "correlation": 76623
    }
  },
  {
    "ph": "s", "id": 76623, "pid": 76337, "tid": -914061504, "ts": 1716454222524703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222580156, "dur": 18,
    "args": {
      "External id": 76645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76645,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76645, "pid": 5, "tid": 7, "ts": 1716454222580156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524754, "dur": 10,
    "args": {
      "External id": 76645,
      "cbid": 211, "correlation": 76645
    }
  },
  {
    "ph": "s", "id": 76645, "pid": 76337, "tid": -914061504, "ts": 1716454222524754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524840, "dur": 1,
    "args": {
      "External id": 76661,
      "cbid": 251, "correlation": 76661
    }
  },
  {
    "ph": "f", "id": 76661, "pid": 76337, "tid": -914061504, "ts": 1716454222524840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524845, "dur": 0,
    "args": {
      "External id": 76663,
      "cbid": 251, "correlation": 76663
    }
  },
  {
    "ph": "f", "id": 76663, "pid": 76337, "tid": -914061504, "ts": 1716454222524845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222580176, "dur": 496,
    "args": {
      "External id": 76664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76664,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76664, "pid": 5, "tid": 7, "ts": 1716454222580176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524847, "dur": 14,
    "args": {
      "External id": 76664,
      "cbid": 211, "correlation": 76664
    }
  },
  {
    "ph": "s", "id": 76664, "pid": 76337, "tid": -914061504, "ts": 1716454222524847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222580673, "dur": 67,
    "args": {
      "External id": 76672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76672, "pid": 5, "tid": 7, "ts": 1716454222580673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524913, "dur": 12,
    "args": {
      "External id": 76672,
      "cbid": 211, "correlation": 76672
    }
  },
  {
    "ph": "s", "id": 76672, "pid": 76337, "tid": -914061504, "ts": 1716454222524913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222580741, "dur": 67,
    "args": {
      "External id": 76680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76680, "pid": 5, "tid": 7, "ts": 1716454222580741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222524942, "dur": 8,
    "args": {
      "External id": 76680,
      "cbid": 211, "correlation": 76680
    }
  },
  {
    "ph": "s", "id": 76680, "pid": 76337, "tid": -914061504, "ts": 1716454222524942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525030, "dur": 1,
    "args": {
      "External id": 76696,
      "cbid": 251, "correlation": 76696
    }
  },
  {
    "ph": "f", "id": 76696, "pid": 76337, "tid": -914061504, "ts": 1716454222525030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222580811, "dur": 1,
    "args": {
      "External id": 76698,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76698,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 76698, "pid": 5, "tid": 7, "ts": 1716454222580811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525035, "dur": 9,
    "args": {
      "External id": 76698,
      "cbid": 51, "correlation": 76698
    }
  },
  {
    "ph": "s", "id": 76698, "pid": 76337, "tid": -914061504, "ts": 1716454222525035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222580814, "dur": 278,
    "args": {
      "External id": 76699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76699,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76699, "pid": 5, "tid": 7, "ts": 1716454222580814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525046, "dur": 11,
    "args": {
      "External id": 76699,
      "cbid": 211, "correlation": 76699
    }
  },
  {
    "ph": "s", "id": 76699, "pid": 76337, "tid": -914061504, "ts": 1716454222525046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222581093, "dur": 14,
    "args": {
      "External id": 76707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76707,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76707, "pid": 5, "tid": 7, "ts": 1716454222581093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525088, "dur": 10,
    "args": {
      "External id": 76707,
      "cbid": 211, "correlation": 76707
    }
  },
  {
    "ph": "s", "id": 76707, "pid": 76337, "tid": -914061504, "ts": 1716454222525088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222581109, "dur": 38,
    "args": {
      "External id": 76718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76718, "pid": 5, "tid": 7, "ts": 1716454222581109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525155, "dur": 12,
    "args": {
      "External id": 76718,
      "cbid": 211, "correlation": 76718
    }
  },
  {
    "ph": "s", "id": 76718, "pid": 76337, "tid": -914061504, "ts": 1716454222525155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525220, "dur": 0,
    "args": {
      "External id": 76730,
      "cbid": 317, "correlation": 76730
    }
  },
  {
    "ph": "f", "id": 76730, "pid": 76337, "tid": -914061504, "ts": 1716454222525220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525221, "dur": 0,
    "args": {
      "External id": 76731,
      "cbid": 203, "correlation": 76731
    }
  },
  {
    "ph": "f", "id": 76731, "pid": 76337, "tid": -914061504, "ts": 1716454222525221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525221, "dur": 0,
    "args": {
      "External id": 76732,
      "cbid": 205, "correlation": 76732
    }
  },
  {
    "ph": "f", "id": 76732, "pid": 76337, "tid": -914061504, "ts": 1716454222525221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581149, "dur": 14,
    "args": {
      "External id": 76736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76736,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76736, "pid": 5, "tid": 7, "ts": 1716454222581149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525238, "dur": 12,
    "args": {
      "External id": 76736,
      "cbid": 211, "correlation": 76736
    }
  },
  {
    "ph": "s", "id": 76736, "pid": 76337, "tid": -914061504, "ts": 1716454222525238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222581163, "dur": 4,
    "args": {
      "External id": 76738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 76738, "pid": 5, "tid": 7, "ts": 1716454222581163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525254, "dur": 6,
    "args": {
      "External id": 76738,
      "cbid": 211, "correlation": 76738
    }
  },
  {
    "ph": "s", "id": 76738, "pid": 76337, "tid": -914061504, "ts": 1716454222525254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525263, "dur": 0,
    "args": {
      "External id": 76739,
      "cbid": 51, "correlation": 76739
    }
  },
  {
    "ph": "s", "id": 76739, "pid": 76337, "tid": -914061504, "ts": 1716454222525263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222581169, "dur": 100,
    "args": {
      "External id": 76740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76740,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 76740, "pid": 5, "tid": 7, "ts": 1716454222581169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525264, "dur": 5,
    "args": {
      "External id": 76740,
      "cbid": 211, "correlation": 76740
    }
  },
  {
    "ph": "s", "id": 76740, "pid": 76337, "tid": -914061504, "ts": 1716454222525264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222581270, "dur": 17,
    "args": {
      "External id": 76745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76745, "pid": 5, "tid": 7, "ts": 1716454222581270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525291, "dur": 8,
    "args": {
      "External id": 76745,
      "cbid": 211, "correlation": 76745
    }
  },
  {
    "ph": "s", "id": 76745, "pid": 76337, "tid": -914061504, "ts": 1716454222525291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222581288, "dur": 11,
    "args": {
      "External id": 76753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76753,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76753, "pid": 5, "tid": 7, "ts": 1716454222581288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525322, "dur": 9,
    "args": {
      "External id": 76753,
      "cbid": 211, "correlation": 76753
    }
  },
  {
    "ph": "s", "id": 76753, "pid": 76337, "tid": -914061504, "ts": 1716454222525322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525391, "dur": 0,
    "args": {
      "External id": 76763,
      "cbid": 317, "correlation": 76763
    }
  },
  {
    "ph": "f", "id": 76763, "pid": 76337, "tid": -914061504, "ts": 1716454222525391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525392, "dur": 0,
    "args": {
      "External id": 76764,
      "cbid": 203, "correlation": 76764
    }
  },
  {
    "ph": "f", "id": 76764, "pid": 76337, "tid": -914061504, "ts": 1716454222525392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525393, "dur": 0,
    "args": {
      "External id": 76765,
      "cbid": 205, "correlation": 76765
    }
  },
  {
    "ph": "f", "id": 76765, "pid": 76337, "tid": -914061504, "ts": 1716454222525393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581301, "dur": 12,
    "args": {
      "External id": 76769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76769,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76769, "pid": 5, "tid": 7, "ts": 1716454222581301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525407, "dur": 12,
    "args": {
      "External id": 76769,
      "cbid": 211, "correlation": 76769
    }
  },
  {
    "ph": "s", "id": 76769, "pid": 76337, "tid": -914061504, "ts": 1716454222525407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581315, "dur": 165,
    "args": {
      "External id": 76771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76771,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76771, "pid": 5, "tid": 7, "ts": 1716454222581315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525421, "dur": 5,
    "args": {
      "External id": 76771,
      "cbid": 211, "correlation": 76771
    }
  },
  {
    "ph": "s", "id": 76771, "pid": 76337, "tid": -914061504, "ts": 1716454222525421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222581482, "dur": 1,
    "args": {
      "External id": 76773,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76773,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 76773, "pid": 5, "tid": 7, "ts": 1716454222581482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525432, "dur": 6,
    "args": {
      "External id": 76773,
      "cbid": 51, "correlation": 76773
    }
  },
  {
    "ph": "s", "id": 76773, "pid": 76337, "tid": -914061504, "ts": 1716454222525432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222581486, "dur": 201,
    "args": {
      "External id": 76774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76774,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 76774, "pid": 5, "tid": 7, "ts": 1716454222581486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525440, "dur": 8,
    "args": {
      "External id": 76774,
      "cbid": 211, "correlation": 76774
    }
  },
  {
    "ph": "s", "id": 76774, "pid": 76337, "tid": -914061504, "ts": 1716454222525440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581688, "dur": 6,
    "args": {
      "External id": 76776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76776, "pid": 5, "tid": 7, "ts": 1716454222581688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525452, "dur": 5,
    "args": {
      "External id": 76776,
      "cbid": 211, "correlation": 76776
    }
  },
  {
    "ph": "s", "id": 76776, "pid": 76337, "tid": -914061504, "ts": 1716454222525452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222581696, "dur": 7,
    "args": {
      "External id": 76782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76782, "pid": 5, "tid": 7, "ts": 1716454222581696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525481, "dur": 8,
    "args": {
      "External id": 76782,
      "cbid": 211, "correlation": 76782
    }
  },
  {
    "ph": "s", "id": 76782, "pid": 76337, "tid": -914061504, "ts": 1716454222525481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581704, "dur": 11,
    "args": {
      "External id": 76802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76802,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 76802, "pid": 5, "tid": 7, "ts": 1716454222581704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525574, "dur": 12,
    "args": {
      "External id": 76802,
      "cbid": 211, "correlation": 76802
    }
  },
  {
    "ph": "s", "id": 76802, "pid": 76337, "tid": -914061504, "ts": 1716454222525574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222581716, "dur": 4,
    "args": {
      "External id": 76814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76814,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 76814, "pid": 5, "tid": 7, "ts": 1716454222581716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525596, "dur": 6,
    "args": {
      "External id": 76814,
      "cbid": 211, "correlation": 76814
    }
  },
  {
    "ph": "s", "id": 76814, "pid": 76337, "tid": -914061504, "ts": 1716454222525596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222581722, "dur": 9,
    "args": {
      "External id": 76817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76817,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76817, "pid": 5, "tid": 7, "ts": 1716454222581722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525614, "dur": 6,
    "args": {
      "External id": 76817,
      "cbid": 211, "correlation": 76817
    }
  },
  {
    "ph": "s", "id": 76817, "pid": 76337, "tid": -914061504, "ts": 1716454222525614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222581731, "dur": 5,
    "args": {
      "External id": 76826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76826,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76826, "pid": 5, "tid": 7, "ts": 1716454222581731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525654, "dur": 9,
    "args": {
      "External id": 76826,
      "cbid": 211, "correlation": 76826
    }
  },
  {
    "ph": "s", "id": 76826, "pid": 76337, "tid": -914061504, "ts": 1716454222525654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525705, "dur": 0,
    "args": {
      "External id": 76836,
      "cbid": 317, "correlation": 76836
    }
  },
  {
    "ph": "f", "id": 76836, "pid": 76337, "tid": -914061504, "ts": 1716454222525705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525706, "dur": 0,
    "args": {
      "External id": 76837,
      "cbid": 203, "correlation": 76837
    }
  },
  {
    "ph": "f", "id": 76837, "pid": 76337, "tid": -914061504, "ts": 1716454222525706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525707, "dur": 0,
    "args": {
      "External id": 76838,
      "cbid": 205, "correlation": 76838
    }
  },
  {
    "ph": "f", "id": 76838, "pid": 76337, "tid": -914061504, "ts": 1716454222525707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581738, "dur": 5,
    "args": {
      "External id": 76842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76842,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76842, "pid": 5, "tid": 7, "ts": 1716454222581738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525722, "dur": 11,
    "args": {
      "External id": 76842,
      "cbid": 211, "correlation": 76842
    }
  },
  {
    "ph": "s", "id": 76842, "pid": 76337, "tid": -914061504, "ts": 1716454222525722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222581744, "dur": 165,
    "args": {
      "External id": 76844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76844,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76844, "pid": 5, "tid": 7, "ts": 1716454222581744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525736, "dur": 6,
    "args": {
      "External id": 76844,
      "cbid": 211, "correlation": 76844
    }
  },
  {
    "ph": "s", "id": 76844, "pid": 76337, "tid": -914061504, "ts": 1716454222525736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222581912, "dur": 1,
    "args": {
      "External id": 76846,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76846,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15315890236119975
    }
  },
  {
    "ph": "f", "id": 76846, "pid": 5, "tid": 7, "ts": 1716454222581912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525747, "dur": 7,
    "args": {
      "External id": 76846,
      "cbid": 51, "correlation": 76846
    }
  },
  {
    "ph": "s", "id": 76846, "pid": 76337, "tid": -914061504, "ts": 1716454222525747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222581915, "dur": 275,
    "args": {
      "External id": 76847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76847,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76847, "pid": 5, "tid": 7, "ts": 1716454222581915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525755, "dur": 6,
    "args": {
      "External id": 76847,
      "cbid": 211, "correlation": 76847
    }
  },
  {
    "ph": "s", "id": 76847, "pid": 76337, "tid": -914061504, "ts": 1716454222525755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582192, "dur": 6,
    "args": {
      "External id": 76849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76849, "pid": 5, "tid": 7, "ts": 1716454222582192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525765, "dur": 6,
    "args": {
      "External id": 76849,
      "cbid": 211, "correlation": 76849
    }
  },
  {
    "ph": "s", "id": 76849, "pid": 76337, "tid": -914061504, "ts": 1716454222525765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222582199, "dur": 6,
    "args": {
      "External id": 76855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76855, "pid": 5, "tid": 7, "ts": 1716454222582199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525794, "dur": 8,
    "args": {
      "External id": 76855,
      "cbid": 211, "correlation": 76855
    }
  },
  {
    "ph": "s", "id": 76855, "pid": 76337, "tid": -914061504, "ts": 1716454222525794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222582207, "dur": 3,
    "args": {
      "External id": 76863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76863,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 76863, "pid": 5, "tid": 7, "ts": 1716454222582207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525836, "dur": 10,
    "args": {
      "External id": 76863,
      "cbid": 211, "correlation": 76863
    }
  },
  {
    "ph": "s", "id": 76863, "pid": 76337, "tid": -914061504, "ts": 1716454222525836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525900, "dur": 1,
    "args": {
      "External id": 76879,
      "cbid": 251, "correlation": 76879
    }
  },
  {
    "ph": "f", "id": 76879, "pid": 76337, "tid": -914061504, "ts": 1716454222525900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525905, "dur": 0,
    "args": {
      "External id": 76881,
      "cbid": 251, "correlation": 76881
    }
  },
  {
    "ph": "f", "id": 76881, "pid": 76337, "tid": -914061504, "ts": 1716454222525905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222582212, "dur": 13,
    "args": {
      "External id": 76882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76882,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76882, "pid": 5, "tid": 7, "ts": 1716454222582212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525907, "dur": 11,
    "args": {
      "External id": 76882,
      "cbid": 211, "correlation": 76882
    }
  },
  {
    "ph": "s", "id": 76882, "pid": 76337, "tid": -914061504, "ts": 1716454222525907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222582226, "dur": 5,
    "args": {
      "External id": 76884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76884,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76884, "pid": 5, "tid": 7, "ts": 1716454222582226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525920, "dur": 5,
    "args": {
      "External id": 76884,
      "cbid": 211, "correlation": 76884
    }
  },
  {
    "ph": "s", "id": 76884, "pid": 76337, "tid": -914061504, "ts": 1716454222525920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222582233, "dur": 6,
    "args": {
      "External id": 76894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76894,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76894, "pid": 5, "tid": 7, "ts": 1716454222582233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222525987, "dur": 13,
    "args": {
      "External id": 76894,
      "cbid": 211, "correlation": 76894
    }
  },
  {
    "ph": "s", "id": 76894, "pid": 76337, "tid": -914061504, "ts": 1716454222525987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582240, "dur": 10,
    "args": {
      "External id": 76914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76914,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 76914, "pid": 5, "tid": 7, "ts": 1716454222582240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526055, "dur": 11,
    "args": {
      "External id": 76914,
      "cbid": 211, "correlation": 76914
    }
  },
  {
    "ph": "s", "id": 76914, "pid": 76337, "tid": -914061504, "ts": 1716454222526055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222582251, "dur": 4,
    "args": {
      "External id": 76926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76926,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 76926, "pid": 5, "tid": 7, "ts": 1716454222582251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526076, "dur": 6,
    "args": {
      "External id": 76926,
      "cbid": 211, "correlation": 76926
    }
  },
  {
    "ph": "s", "id": 76926, "pid": 76337, "tid": -914061504, "ts": 1716454222526076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222582257, "dur": 7,
    "args": {
      "External id": 76929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76929,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76929, "pid": 5, "tid": 7, "ts": 1716454222582257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526094, "dur": 7,
    "args": {
      "External id": 76929,
      "cbid": 211, "correlation": 76929
    }
  },
  {
    "ph": "s", "id": 76929, "pid": 76337, "tid": -914061504, "ts": 1716454222526094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222582265, "dur": 5,
    "args": {
      "External id": 76938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76938,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76938, "pid": 5, "tid": 7, "ts": 1716454222582265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526134, "dur": 10,
    "args": {
      "External id": 76938,
      "cbid": 211, "correlation": 76938
    }
  },
  {
    "ph": "s", "id": 76938, "pid": 76337, "tid": -914061504, "ts": 1716454222526134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526197, "dur": 0,
    "args": {
      "External id": 76948,
      "cbid": 317, "correlation": 76948
    }
  },
  {
    "ph": "f", "id": 76948, "pid": 76337, "tid": -914061504, "ts": 1716454222526197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526198, "dur": 0,
    "args": {
      "External id": 76949,
      "cbid": 203, "correlation": 76949
    }
  },
  {
    "ph": "f", "id": 76949, "pid": 76337, "tid": -914061504, "ts": 1716454222526198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526199, "dur": 0,
    "args": {
      "External id": 76950,
      "cbid": 205, "correlation": 76950
    }
  },
  {
    "ph": "f", "id": 76950, "pid": 76337, "tid": -914061504, "ts": 1716454222526199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582271, "dur": 5,
    "args": {
      "External id": 76954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76954,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76954, "pid": 5, "tid": 7, "ts": 1716454222582271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526213, "dur": 11,
    "args": {
      "External id": 76954,
      "cbid": 211, "correlation": 76954
    }
  },
  {
    "ph": "s", "id": 76954, "pid": 76337, "tid": -914061504, "ts": 1716454222526213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582278, "dur": 164,
    "args": {
      "External id": 76956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76956,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76956, "pid": 5, "tid": 7, "ts": 1716454222582278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526227, "dur": 5,
    "args": {
      "External id": 76956,
      "cbid": 211, "correlation": 76956
    }
  },
  {
    "ph": "s", "id": 76956, "pid": 76337, "tid": -914061504, "ts": 1716454222526227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222582444, "dur": 1,
    "args": {
      "External id": 76958,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 76958,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 76958, "pid": 5, "tid": 7, "ts": 1716454222582444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526237, "dur": 7,
    "args": {
      "External id": 76958,
      "cbid": 51, "correlation": 76958
    }
  },
  {
    "ph": "s", "id": 76958, "pid": 76337, "tid": -914061504, "ts": 1716454222526237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222582448, "dur": 263,
    "args": {
      "External id": 76959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76959,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 76959, "pid": 5, "tid": 7, "ts": 1716454222582448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526246, "dur": 6,
    "args": {
      "External id": 76959,
      "cbid": 211, "correlation": 76959
    }
  },
  {
    "ph": "s", "id": 76959, "pid": 76337, "tid": -914061504, "ts": 1716454222526246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582712, "dur": 6,
    "args": {
      "External id": 76961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76961,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 76961, "pid": 5, "tid": 7, "ts": 1716454222582712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526255, "dur": 5,
    "args": {
      "External id": 76961,
      "cbid": 211, "correlation": 76961
    }
  },
  {
    "ph": "s", "id": 76961, "pid": 76337, "tid": -914061504, "ts": 1716454222526255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222582720, "dur": 6,
    "args": {
      "External id": 76967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76967,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76967, "pid": 5, "tid": 7, "ts": 1716454222582720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526282, "dur": 8,
    "args": {
      "External id": 76967,
      "cbid": 211, "correlation": 76967
    }
  },
  {
    "ph": "s", "id": 76967, "pid": 76337, "tid": -914061504, "ts": 1716454222526282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222582727, "dur": 5,
    "args": {
      "External id": 76975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76975,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76975, "pid": 5, "tid": 7, "ts": 1716454222582727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526315, "dur": 8,
    "args": {
      "External id": 76975,
      "cbid": 211, "correlation": 76975
    }
  },
  {
    "ph": "s", "id": 76975, "pid": 76337, "tid": -914061504, "ts": 1716454222526315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222582734, "dur": 5,
    "args": {
      "External id": 76983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 76983,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 76983, "pid": 5, "tid": 7, "ts": 1716454222582734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526344, "dur": 9,
    "args": {
      "External id": 76983,
      "cbid": 211, "correlation": 76983
    }
  },
  {
    "ph": "s", "id": 76983, "pid": 76337, "tid": -914061504, "ts": 1716454222526344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582740, "dur": 10,
    "args": {
      "External id": 77003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77003,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77003, "pid": 5, "tid": 7, "ts": 1716454222582740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526419, "dur": 13,
    "args": {
      "External id": 77003,
      "cbid": 211, "correlation": 77003
    }
  },
  {
    "ph": "s", "id": 77003, "pid": 76337, "tid": -914061504, "ts": 1716454222526419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222582751, "dur": 4,
    "args": {
      "External id": 77015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77015,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77015, "pid": 5, "tid": 7, "ts": 1716454222582751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526441, "dur": 6,
    "args": {
      "External id": 77015,
      "cbid": 211, "correlation": 77015
    }
  },
  {
    "ph": "s", "id": 77015, "pid": 76337, "tid": -914061504, "ts": 1716454222526441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222582756, "dur": 6,
    "args": {
      "External id": 77018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77018, "pid": 5, "tid": 7, "ts": 1716454222582756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526459, "dur": 6,
    "args": {
      "External id": 77018,
      "cbid": 211, "correlation": 77018
    }
  },
  {
    "ph": "s", "id": 77018, "pid": 76337, "tid": -914061504, "ts": 1716454222526459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222582764, "dur": 5,
    "args": {
      "External id": 77027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77027,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77027, "pid": 5, "tid": 7, "ts": 1716454222582764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526496, "dur": 10,
    "args": {
      "External id": 77027,
      "cbid": 211, "correlation": 77027
    }
  },
  {
    "ph": "s", "id": 77027, "pid": 76337, "tid": -914061504, "ts": 1716454222526496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526547, "dur": 0,
    "args": {
      "External id": 77037,
      "cbid": 317, "correlation": 77037
    }
  },
  {
    "ph": "f", "id": 77037, "pid": 76337, "tid": -914061504, "ts": 1716454222526547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526548, "dur": 0,
    "args": {
      "External id": 77038,
      "cbid": 203, "correlation": 77038
    }
  },
  {
    "ph": "f", "id": 77038, "pid": 76337, "tid": -914061504, "ts": 1716454222526548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526548, "dur": 0,
    "args": {
      "External id": 77039,
      "cbid": 205, "correlation": 77039
    }
  },
  {
    "ph": "f", "id": 77039, "pid": 76337, "tid": -914061504, "ts": 1716454222526548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582770, "dur": 5,
    "args": {
      "External id": 77043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77043,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77043, "pid": 5, "tid": 7, "ts": 1716454222582770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526563, "dur": 11,
    "args": {
      "External id": 77043,
      "cbid": 211, "correlation": 77043
    }
  },
  {
    "ph": "s", "id": 77043, "pid": 76337, "tid": -914061504, "ts": 1716454222526563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222582776, "dur": 166,
    "args": {
      "External id": 77045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77045,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77045, "pid": 5, "tid": 7, "ts": 1716454222582776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526577, "dur": 5,
    "args": {
      "External id": 77045,
      "cbid": 211, "correlation": 77045
    }
  },
  {
    "ph": "s", "id": 77045, "pid": 76337, "tid": -914061504, "ts": 1716454222526577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222582944, "dur": 1,
    "args": {
      "External id": 77047,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77047,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 77047, "pid": 5, "tid": 7, "ts": 1716454222582944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526587, "dur": 6,
    "args": {
      "External id": 77047,
      "cbid": 51, "correlation": 77047
    }
  },
  {
    "ph": "s", "id": 77047, "pid": 76337, "tid": -914061504, "ts": 1716454222526587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222582948, "dur": 263,
    "args": {
      "External id": 77048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77048,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77048, "pid": 5, "tid": 7, "ts": 1716454222582948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526594, "dur": 6,
    "args": {
      "External id": 77048,
      "cbid": 211, "correlation": 77048
    }
  },
  {
    "ph": "s", "id": 77048, "pid": 76337, "tid": -914061504, "ts": 1716454222526594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583212, "dur": 6,
    "args": {
      "External id": 77050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77050, "pid": 5, "tid": 7, "ts": 1716454222583212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526604, "dur": 5,
    "args": {
      "External id": 77050,
      "cbid": 211, "correlation": 77050
    }
  },
  {
    "ph": "s", "id": 77050, "pid": 76337, "tid": -914061504, "ts": 1716454222526604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222583219, "dur": 6,
    "args": {
      "External id": 77056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77056,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77056, "pid": 5, "tid": 7, "ts": 1716454222583219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526632, "dur": 8,
    "args": {
      "External id": 77056,
      "cbid": 211, "correlation": 77056
    }
  },
  {
    "ph": "s", "id": 77056, "pid": 76337, "tid": -914061504, "ts": 1716454222526632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222583227, "dur": 3,
    "args": {
      "External id": 77064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77064,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 77064, "pid": 5, "tid": 7, "ts": 1716454222583227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526676, "dur": 9,
    "args": {
      "External id": 77064,
      "cbid": 211, "correlation": 77064
    }
  },
  {
    "ph": "s", "id": 77064, "pid": 76337, "tid": -914061504, "ts": 1716454222526676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526738, "dur": 1,
    "args": {
      "External id": 77080,
      "cbid": 251, "correlation": 77080
    }
  },
  {
    "ph": "f", "id": 77080, "pid": 76337, "tid": -914061504, "ts": 1716454222526738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526743, "dur": 0,
    "args": {
      "External id": 77082,
      "cbid": 251, "correlation": 77082
    }
  },
  {
    "ph": "f", "id": 77082, "pid": 76337, "tid": -914061504, "ts": 1716454222526743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222583231, "dur": 10,
    "args": {
      "External id": 77083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77083,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77083, "pid": 5, "tid": 7, "ts": 1716454222583231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526745, "dur": 11,
    "args": {
      "External id": 77083,
      "cbid": 211, "correlation": 77083
    }
  },
  {
    "ph": "s", "id": 77083, "pid": 76337, "tid": -914061504, "ts": 1716454222526745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222583242, "dur": 4,
    "args": {
      "External id": 77085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77085,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77085, "pid": 5, "tid": 7, "ts": 1716454222583242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526758, "dur": 6,
    "args": {
      "External id": 77085,
      "cbid": 211, "correlation": 77085
    }
  },
  {
    "ph": "s", "id": 77085, "pid": 76337, "tid": -914061504, "ts": 1716454222526758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222583247, "dur": 6,
    "args": {
      "External id": 77095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77095,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77095, "pid": 5, "tid": 7, "ts": 1716454222583247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526815, "dur": 12,
    "args": {
      "External id": 77095,
      "cbid": 211, "correlation": 77095
    }
  },
  {
    "ph": "s", "id": 77095, "pid": 76337, "tid": -914061504, "ts": 1716454222526815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583254, "dur": 10,
    "args": {
      "External id": 77115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77115,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77115, "pid": 5, "tid": 7, "ts": 1716454222583254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526880, "dur": 10,
    "args": {
      "External id": 77115,
      "cbid": 211, "correlation": 77115
    }
  },
  {
    "ph": "s", "id": 77115, "pid": 76337, "tid": -914061504, "ts": 1716454222526880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222583265, "dur": 4,
    "args": {
      "External id": 77127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77127,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77127, "pid": 5, "tid": 7, "ts": 1716454222583265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526901, "dur": 6,
    "args": {
      "External id": 77127,
      "cbid": 211, "correlation": 77127
    }
  },
  {
    "ph": "s", "id": 77127, "pid": 76337, "tid": -914061504, "ts": 1716454222526901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222583271, "dur": 7,
    "args": {
      "External id": 77130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77130, "pid": 5, "tid": 7, "ts": 1716454222583271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526918, "dur": 7,
    "args": {
      "External id": 77130,
      "cbid": 211, "correlation": 77130
    }
  },
  {
    "ph": "s", "id": 77130, "pid": 76337, "tid": -914061504, "ts": 1716454222526918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222583279, "dur": 5,
    "args": {
      "External id": 77139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77139,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77139, "pid": 5, "tid": 7, "ts": 1716454222583279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222526958, "dur": 10,
    "args": {
      "External id": 77139,
      "cbid": 211, "correlation": 77139
    }
  },
  {
    "ph": "s", "id": 77139, "pid": 76337, "tid": -914061504, "ts": 1716454222526958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527030, "dur": 0,
    "args": {
      "External id": 77149,
      "cbid": 317, "correlation": 77149
    }
  },
  {
    "ph": "f", "id": 77149, "pid": 76337, "tid": -914061504, "ts": 1716454222527030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527031, "dur": 0,
    "args": {
      "External id": 77150,
      "cbid": 203, "correlation": 77150
    }
  },
  {
    "ph": "f", "id": 77150, "pid": 76337, "tid": -914061504, "ts": 1716454222527031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527032, "dur": 0,
    "args": {
      "External id": 77151,
      "cbid": 205, "correlation": 77151
    }
  },
  {
    "ph": "f", "id": 77151, "pid": 76337, "tid": -914061504, "ts": 1716454222527032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583285, "dur": 5,
    "args": {
      "External id": 77155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77155,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77155, "pid": 5, "tid": 7, "ts": 1716454222583285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527046, "dur": 12,
    "args": {
      "External id": 77155,
      "cbid": 211, "correlation": 77155
    }
  },
  {
    "ph": "s", "id": 77155, "pid": 76337, "tid": -914061504, "ts": 1716454222527046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583292, "dur": 166,
    "args": {
      "External id": 77157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77157,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77157, "pid": 5, "tid": 7, "ts": 1716454222583292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527061, "dur": 5,
    "args": {
      "External id": 77157,
      "cbid": 211, "correlation": 77157
    }
  },
  {
    "ph": "s", "id": 77157, "pid": 76337, "tid": -914061504, "ts": 1716454222527061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222583460, "dur": 1,
    "args": {
      "External id": 77159,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77159,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 77159, "pid": 5, "tid": 7, "ts": 1716454222583460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527071, "dur": 6,
    "args": {
      "External id": 77159,
      "cbid": 51, "correlation": 77159
    }
  },
  {
    "ph": "s", "id": 77159, "pid": 76337, "tid": -914061504, "ts": 1716454222527071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222583463, "dur": 263,
    "args": {
      "External id": 77160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77160,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77160, "pid": 5, "tid": 7, "ts": 1716454222583463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527079, "dur": 7,
    "args": {
      "External id": 77160,
      "cbid": 211, "correlation": 77160
    }
  },
  {
    "ph": "s", "id": 77160, "pid": 76337, "tid": -914061504, "ts": 1716454222527079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583728, "dur": 6,
    "args": {
      "External id": 77162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77162,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77162, "pid": 5, "tid": 7, "ts": 1716454222583728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527090, "dur": 5,
    "args": {
      "External id": 77162,
      "cbid": 211, "correlation": 77162
    }
  },
  {
    "ph": "s", "id": 77162, "pid": 76337, "tid": -914061504, "ts": 1716454222527090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222583735, "dur": 6,
    "args": {
      "External id": 77168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77168, "pid": 5, "tid": 7, "ts": 1716454222583735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527118, "dur": 8,
    "args": {
      "External id": 77168,
      "cbid": 211, "correlation": 77168
    }
  },
  {
    "ph": "s", "id": 77168, "pid": 76337, "tid": -914061504, "ts": 1716454222527118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222583743, "dur": 5,
    "args": {
      "External id": 77176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77176,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77176, "pid": 5, "tid": 7, "ts": 1716454222583743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527151, "dur": 8,
    "args": {
      "External id": 77176,
      "cbid": 211, "correlation": 77176
    }
  },
  {
    "ph": "s", "id": 77176, "pid": 76337, "tid": -914061504, "ts": 1716454222527151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222583749, "dur": 5,
    "args": {
      "External id": 77184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77184,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77184, "pid": 5, "tid": 7, "ts": 1716454222583749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527179, "dur": 8,
    "args": {
      "External id": 77184,
      "cbid": 211, "correlation": 77184
    }
  },
  {
    "ph": "s", "id": 77184, "pid": 76337, "tid": -914061504, "ts": 1716454222527179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583755, "dur": 10,
    "args": {
      "External id": 77204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77204,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77204, "pid": 5, "tid": 7, "ts": 1716454222583755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527284, "dur": 14,
    "args": {
      "External id": 77204,
      "cbid": 211, "correlation": 77204
    }
  },
  {
    "ph": "s", "id": 77204, "pid": 76337, "tid": -914061504, "ts": 1716454222527284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222583766, "dur": 4,
    "args": {
      "External id": 77216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77216,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77216, "pid": 5, "tid": 7, "ts": 1716454222583766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527308, "dur": 6,
    "args": {
      "External id": 77216,
      "cbid": 211, "correlation": 77216
    }
  },
  {
    "ph": "s", "id": 77216, "pid": 76337, "tid": -914061504, "ts": 1716454222527308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222583771, "dur": 7,
    "args": {
      "External id": 77219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77219, "pid": 5, "tid": 7, "ts": 1716454222583771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527326, "dur": 6,
    "args": {
      "External id": 77219,
      "cbid": 211, "correlation": 77219
    }
  },
  {
    "ph": "s", "id": 77219, "pid": 76337, "tid": -914061504, "ts": 1716454222527326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222583779, "dur": 5,
    "args": {
      "External id": 77228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77228,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77228, "pid": 5, "tid": 7, "ts": 1716454222583779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527365, "dur": 9,
    "args": {
      "External id": 77228,
      "cbid": 211, "correlation": 77228
    }
  },
  {
    "ph": "s", "id": 77228, "pid": 76337, "tid": -914061504, "ts": 1716454222527365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527416, "dur": 0,
    "args": {
      "External id": 77238,
      "cbid": 317, "correlation": 77238
    }
  },
  {
    "ph": "f", "id": 77238, "pid": 76337, "tid": -914061504, "ts": 1716454222527416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527417, "dur": 0,
    "args": {
      "External id": 77239,
      "cbid": 203, "correlation": 77239
    }
  },
  {
    "ph": "f", "id": 77239, "pid": 76337, "tid": -914061504, "ts": 1716454222527417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527418, "dur": 0,
    "args": {
      "External id": 77240,
      "cbid": 205, "correlation": 77240
    }
  },
  {
    "ph": "f", "id": 77240, "pid": 76337, "tid": -914061504, "ts": 1716454222527418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583785, "dur": 5,
    "args": {
      "External id": 77244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77244,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77244, "pid": 5, "tid": 7, "ts": 1716454222583785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527431, "dur": 11,
    "args": {
      "External id": 77244,
      "cbid": 211, "correlation": 77244
    }
  },
  {
    "ph": "s", "id": 77244, "pid": 76337, "tid": -914061504, "ts": 1716454222527431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222583791, "dur": 165,
    "args": {
      "External id": 77246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77246,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77246, "pid": 5, "tid": 7, "ts": 1716454222583791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527445, "dur": 5,
    "args": {
      "External id": 77246,
      "cbid": 211, "correlation": 77246
    }
  },
  {
    "ph": "s", "id": 77246, "pid": 76337, "tid": -914061504, "ts": 1716454222527445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222583958, "dur": 1,
    "args": {
      "External id": 77248,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77248,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 77248, "pid": 5, "tid": 7, "ts": 1716454222583958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527455, "dur": 6,
    "args": {
      "External id": 77248,
      "cbid": 51, "correlation": 77248
    }
  },
  {
    "ph": "s", "id": 77248, "pid": 76337, "tid": -914061504, "ts": 1716454222527455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222583962, "dur": 263,
    "args": {
      "External id": 77249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77249,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77249, "pid": 5, "tid": 7, "ts": 1716454222583962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527462, "dur": 6,
    "args": {
      "External id": 77249,
      "cbid": 211, "correlation": 77249
    }
  },
  {
    "ph": "s", "id": 77249, "pid": 76337, "tid": -914061504, "ts": 1716454222527462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584227, "dur": 6,
    "args": {
      "External id": 77251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77251,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77251, "pid": 5, "tid": 7, "ts": 1716454222584227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527472, "dur": 5,
    "args": {
      "External id": 77251,
      "cbid": 211, "correlation": 77251
    }
  },
  {
    "ph": "s", "id": 77251, "pid": 76337, "tid": -914061504, "ts": 1716454222527472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222584233, "dur": 6,
    "args": {
      "External id": 77257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77257, "pid": 5, "tid": 7, "ts": 1716454222584233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527500, "dur": 8,
    "args": {
      "External id": 77257,
      "cbid": 211, "correlation": 77257
    }
  },
  {
    "ph": "s", "id": 77257, "pid": 76337, "tid": -914061504, "ts": 1716454222527500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222584241, "dur": 3,
    "args": {
      "External id": 77265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77265,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 77265, "pid": 5, "tid": 7, "ts": 1716454222584241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527543, "dur": 10,
    "args": {
      "External id": 77265,
      "cbid": 211, "correlation": 77265
    }
  },
  {
    "ph": "s", "id": 77265, "pid": 76337, "tid": -914061504, "ts": 1716454222527543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527605, "dur": 1,
    "args": {
      "External id": 77281,
      "cbid": 251, "correlation": 77281
    }
  },
  {
    "ph": "f", "id": 77281, "pid": 76337, "tid": -914061504, "ts": 1716454222527605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527611, "dur": 0,
    "args": {
      "External id": 77283,
      "cbid": 251, "correlation": 77283
    }
  },
  {
    "ph": "f", "id": 77283, "pid": 76337, "tid": -914061504, "ts": 1716454222527611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222584245, "dur": 11,
    "args": {
      "External id": 77284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77284,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77284, "pid": 5, "tid": 7, "ts": 1716454222584245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527613, "dur": 11,
    "args": {
      "External id": 77284,
      "cbid": 211, "correlation": 77284
    }
  },
  {
    "ph": "s", "id": 77284, "pid": 76337, "tid": -914061504, "ts": 1716454222527613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222584258, "dur": 4,
    "args": {
      "External id": 77286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77286,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77286, "pid": 5, "tid": 7, "ts": 1716454222584258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527625, "dur": 5,
    "args": {
      "External id": 77286,
      "cbid": 211, "correlation": 77286
    }
  },
  {
    "ph": "s", "id": 77286, "pid": 76337, "tid": -914061504, "ts": 1716454222527625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222584263, "dur": 6,
    "args": {
      "External id": 77296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77296,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77296, "pid": 5, "tid": 7, "ts": 1716454222584263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527682, "dur": 12,
    "args": {
      "External id": 77296,
      "cbid": 211, "correlation": 77296
    }
  },
  {
    "ph": "s", "id": 77296, "pid": 76337, "tid": -914061504, "ts": 1716454222527682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584270, "dur": 10,
    "args": {
      "External id": 77316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77316,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77316, "pid": 5, "tid": 7, "ts": 1716454222584270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527748, "dur": 10,
    "args": {
      "External id": 77316,
      "cbid": 211, "correlation": 77316
    }
  },
  {
    "ph": "s", "id": 77316, "pid": 76337, "tid": -914061504, "ts": 1716454222527748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222584281, "dur": 4,
    "args": {
      "External id": 77328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77328,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77328, "pid": 5, "tid": 7, "ts": 1716454222584281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527769, "dur": 6,
    "args": {
      "External id": 77328,
      "cbid": 211, "correlation": 77328
    }
  },
  {
    "ph": "s", "id": 77328, "pid": 76337, "tid": -914061504, "ts": 1716454222527769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222584286, "dur": 7,
    "args": {
      "External id": 77331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77331,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77331, "pid": 5, "tid": 7, "ts": 1716454222584286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527787, "dur": 6,
    "args": {
      "External id": 77331,
      "cbid": 211, "correlation": 77331
    }
  },
  {
    "ph": "s", "id": 77331, "pid": 76337, "tid": -914061504, "ts": 1716454222527787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222584295, "dur": 5,
    "args": {
      "External id": 77340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77340,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77340, "pid": 5, "tid": 7, "ts": 1716454222584295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527826, "dur": 10,
    "args": {
      "External id": 77340,
      "cbid": 211, "correlation": 77340
    }
  },
  {
    "ph": "s", "id": 77340, "pid": 76337, "tid": -914061504, "ts": 1716454222527826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527889, "dur": 0,
    "args": {
      "External id": 77350,
      "cbid": 317, "correlation": 77350
    }
  },
  {
    "ph": "f", "id": 77350, "pid": 76337, "tid": -914061504, "ts": 1716454222527889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527890, "dur": 0,
    "args": {
      "External id": 77351,
      "cbid": 203, "correlation": 77351
    }
  },
  {
    "ph": "f", "id": 77351, "pid": 76337, "tid": -914061504, "ts": 1716454222527890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527891, "dur": 0,
    "args": {
      "External id": 77352,
      "cbid": 205, "correlation": 77352
    }
  },
  {
    "ph": "f", "id": 77352, "pid": 76337, "tid": -914061504, "ts": 1716454222527891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584301, "dur": 5,
    "args": {
      "External id": 77356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77356, "pid": 5, "tid": 7, "ts": 1716454222584301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527905, "dur": 12,
    "args": {
      "External id": 77356,
      "cbid": 211, "correlation": 77356
    }
  },
  {
    "ph": "s", "id": 77356, "pid": 76337, "tid": -914061504, "ts": 1716454222527905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584307, "dur": 165,
    "args": {
      "External id": 77358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77358,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77358, "pid": 5, "tid": 7, "ts": 1716454222584307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527919, "dur": 5,
    "args": {
      "External id": 77358,
      "cbid": 211, "correlation": 77358
    }
  },
  {
    "ph": "s", "id": 77358, "pid": 76337, "tid": -914061504, "ts": 1716454222527919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222584474, "dur": 1,
    "args": {
      "External id": 77360,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77360,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 77360, "pid": 5, "tid": 7, "ts": 1716454222584474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527930, "dur": 6,
    "args": {
      "External id": 77360,
      "cbid": 51, "correlation": 77360
    }
  },
  {
    "ph": "s", "id": 77360, "pid": 76337, "tid": -914061504, "ts": 1716454222527930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222584478, "dur": 264,
    "args": {
      "External id": 77361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77361,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77361, "pid": 5, "tid": 7, "ts": 1716454222584478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527937, "dur": 6,
    "args": {
      "External id": 77361,
      "cbid": 211, "correlation": 77361
    }
  },
  {
    "ph": "s", "id": 77361, "pid": 76337, "tid": -914061504, "ts": 1716454222527937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584744, "dur": 6,
    "args": {
      "External id": 77363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77363,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77363, "pid": 5, "tid": 7, "ts": 1716454222584744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527946, "dur": 5,
    "args": {
      "External id": 77363,
      "cbid": 211, "correlation": 77363
    }
  },
  {
    "ph": "s", "id": 77363, "pid": 76337, "tid": -914061504, "ts": 1716454222527946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222584751, "dur": 6,
    "args": {
      "External id": 77369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77369,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77369, "pid": 5, "tid": 7, "ts": 1716454222584751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222527983, "dur": 9,
    "args": {
      "External id": 77369,
      "cbid": 211, "correlation": 77369
    }
  },
  {
    "ph": "s", "id": 77369, "pid": 76337, "tid": -914061504, "ts": 1716454222527983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222584758, "dur": 5,
    "args": {
      "External id": 77377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77377,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77377, "pid": 5, "tid": 7, "ts": 1716454222584758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528016, "dur": 8,
    "args": {
      "External id": 77377,
      "cbid": 211, "correlation": 77377
    }
  },
  {
    "ph": "s", "id": 77377, "pid": 76337, "tid": -914061504, "ts": 1716454222528016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222584765, "dur": 5,
    "args": {
      "External id": 77385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77385,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77385, "pid": 5, "tid": 7, "ts": 1716454222584765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528046, "dur": 8,
    "args": {
      "External id": 77385,
      "cbid": 211, "correlation": 77385
    }
  },
  {
    "ph": "s", "id": 77385, "pid": 76337, "tid": -914061504, "ts": 1716454222528046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584771, "dur": 10,
    "args": {
      "External id": 77405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77405,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77405, "pid": 5, "tid": 7, "ts": 1716454222584771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528187, "dur": 13,
    "args": {
      "External id": 77405,
      "cbid": 211, "correlation": 77405
    }
  },
  {
    "ph": "s", "id": 77405, "pid": 76337, "tid": -914061504, "ts": 1716454222528187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222584782, "dur": 4,
    "args": {
      "External id": 77417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77417,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77417, "pid": 5, "tid": 7, "ts": 1716454222584782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528210, "dur": 6,
    "args": {
      "External id": 77417,
      "cbid": 211, "correlation": 77417
    }
  },
  {
    "ph": "s", "id": 77417, "pid": 76337, "tid": -914061504, "ts": 1716454222528210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222584787, "dur": 7,
    "args": {
      "External id": 77420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77420, "pid": 5, "tid": 7, "ts": 1716454222584787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528228, "dur": 7,
    "args": {
      "External id": 77420,
      "cbid": 211, "correlation": 77420
    }
  },
  {
    "ph": "s", "id": 77420, "pid": 76337, "tid": -914061504, "ts": 1716454222528228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528287, "dur": 0,
    "args": {
      "External id": 77431,
      "cbid": 317, "correlation": 77431
    }
  },
  {
    "ph": "f", "id": 77431, "pid": 76337, "tid": -914061504, "ts": 1716454222528287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528288, "dur": 0,
    "args": {
      "External id": 77432,
      "cbid": 203, "correlation": 77432
    }
  },
  {
    "ph": "f", "id": 77432, "pid": 76337, "tid": -914061504, "ts": 1716454222528288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528289, "dur": 0,
    "args": {
      "External id": 77433,
      "cbid": 205, "correlation": 77433
    }
  },
  {
    "ph": "f", "id": 77433, "pid": 76337, "tid": -914061504, "ts": 1716454222528289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584795, "dur": 5,
    "args": {
      "External id": 77437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77437,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77437, "pid": 5, "tid": 7, "ts": 1716454222584795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528306, "dur": 11,
    "args": {
      "External id": 77437,
      "cbid": 211, "correlation": 77437
    }
  },
  {
    "ph": "s", "id": 77437, "pid": 76337, "tid": -914061504, "ts": 1716454222528306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222584801, "dur": 38,
    "args": {
      "External id": 77439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77439,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 77439, "pid": 5, "tid": 7, "ts": 1716454222584801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528325, "dur": 9,
    "args": {
      "External id": 77439,
      "cbid": 211, "correlation": 77439
    }
  },
  {
    "ph": "s", "id": 77439, "pid": 76337, "tid": -914061504, "ts": 1716454222528325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584840, "dur": 5,
    "args": {
      "External id": 77441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77441,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77441, "pid": 5, "tid": 7, "ts": 1716454222584840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528338, "dur": 5,
    "args": {
      "External id": 77441,
      "cbid": 211, "correlation": 77441
    }
  },
  {
    "ph": "s", "id": 77441, "pid": 76337, "tid": -914061504, "ts": 1716454222528338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222584846, "dur": 6,
    "args": {
      "External id": 77447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77447, "pid": 5, "tid": 7, "ts": 1716454222584846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528365, "dur": 8,
    "args": {
      "External id": 77447,
      "cbid": 211, "correlation": 77447
    }
  },
  {
    "ph": "s", "id": 77447, "pid": 76337, "tid": -914061504, "ts": 1716454222528365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222584854, "dur": 21,
    "args": {
      "External id": 77456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77456,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77456, "pid": 5, "tid": 7, "ts": 1716454222584854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528448, "dur": 14,
    "args": {
      "External id": 77456,
      "cbid": 211, "correlation": 77456
    }
  },
  {
    "ph": "s", "id": 77456, "pid": 76337, "tid": -914061504, "ts": 1716454222528448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222584876, "dur": 11,
    "args": {
      "External id": 77478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77478,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 77478, "pid": 5, "tid": 7, "ts": 1716454222584876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528506, "dur": 10,
    "args": {
      "External id": 77478,
      "cbid": 211, "correlation": 77478
    }
  },
  {
    "ph": "s", "id": 77478, "pid": 76337, "tid": -914061504, "ts": 1716454222528506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528598, "dur": 2,
    "args": {
      "External id": 77489,
      "cbid": 251, "correlation": 77489
    }
  },
  {
    "ph": "f", "id": 77489, "pid": 76337, "tid": -914061504, "ts": 1716454222528598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528602, "dur": 0,
    "args": {
      "External id": 77490,
      "cbid": 251, "correlation": 77490
    }
  },
  {
    "ph": "f", "id": 77490, "pid": 76337, "tid": -914061504, "ts": 1716454222528602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222584888, "dur": 55,
    "args": {
      "External id": 77491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77491,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 77491, "pid": 5, "tid": 7, "ts": 1716454222584888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528606, "dur": 13,
    "args": {
      "External id": 77491,
      "cbid": 211, "correlation": 77491
    }
  },
  {
    "ph": "s", "id": 77491, "pid": 76337, "tid": -914061504, "ts": 1716454222528606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528675, "dur": 1,
    "args": {
      "External id": 77502,
      "cbid": 251, "correlation": 77502
    }
  },
  {
    "ph": "f", "id": 77502, "pid": 76337, "tid": -914061504, "ts": 1716454222528675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528678, "dur": 0,
    "args": {
      "External id": 77503,
      "cbid": 251, "correlation": 77503
    }
  },
  {
    "ph": "f", "id": 77503, "pid": 76337, "tid": -914061504, "ts": 1716454222528678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222584944, "dur": 53,
    "args": {
      "External id": 77504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77504,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 77504, "pid": 5, "tid": 7, "ts": 1716454222584944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528680, "dur": 11,
    "args": {
      "External id": 77504,
      "cbid": 211, "correlation": 77504
    }
  },
  {
    "ph": "s", "id": 77504, "pid": 76337, "tid": -914061504, "ts": 1716454222528680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528745, "dur": 1,
    "args": {
      "External id": 77515,
      "cbid": 251, "correlation": 77515
    }
  },
  {
    "ph": "f", "id": 77515, "pid": 76337, "tid": -914061504, "ts": 1716454222528745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528749, "dur": 0,
    "args": {
      "External id": 77516,
      "cbid": 251, "correlation": 77516
    }
  },
  {
    "ph": "f", "id": 77516, "pid": 76337, "tid": -914061504, "ts": 1716454222528749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222584999, "dur": 54,
    "args": {
      "External id": 77517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77517,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 77517, "pid": 5, "tid": 7, "ts": 1716454222584999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528751, "dur": 11,
    "args": {
      "External id": 77517,
      "cbid": 211, "correlation": 77517
    }
  },
  {
    "ph": "s", "id": 77517, "pid": 76337, "tid": -914061504, "ts": 1716454222528751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222585054, "dur": 58,
    "args": {
      "External id": 77542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77542,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77542, "pid": 5, "tid": 7, "ts": 1716454222585054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528834, "dur": 13,
    "args": {
      "External id": 77542,
      "cbid": 211, "correlation": 77542
    }
  },
  {
    "ph": "s", "id": 77542, "pid": 76337, "tid": -914061504, "ts": 1716454222528834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528933, "dur": 1,
    "args": {
      "External id": 77560,
      "cbid": 251, "correlation": 77560
    }
  },
  {
    "ph": "f", "id": 77560, "pid": 76337, "tid": -914061504, "ts": 1716454222528933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222585113, "dur": 64,
    "args": {
      "External id": 77562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77562,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 77562, "pid": 5, "tid": 7, "ts": 1716454222585113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222528939, "dur": 13,
    "args": {
      "External id": 77562,
      "cbid": 211, "correlation": 77562
    }
  },
  {
    "ph": "s", "id": 77562, "pid": 76337, "tid": -914061504, "ts": 1716454222528939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222585178, "dur": 6,
    "args": {
      "External id": 77570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77570,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77570, "pid": 5, "tid": 7, "ts": 1716454222585178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529019, "dur": 12,
    "args": {
      "External id": 77570,
      "cbid": 211, "correlation": 77570
    }
  },
  {
    "ph": "s", "id": 77570, "pid": 76337, "tid": -914061504, "ts": 1716454222529019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222585186, "dur": 7,
    "args": {
      "External id": 77578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77578,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77578, "pid": 5, "tid": 7, "ts": 1716454222585186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529057, "dur": 9,
    "args": {
      "External id": 77578,
      "cbid": 211, "correlation": 77578
    }
  },
  {
    "ph": "s", "id": 77578, "pid": 76337, "tid": -914061504, "ts": 1716454222529057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585194, "dur": 8,
    "args": {
      "External id": 77589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77589, "pid": 5, "tid": 7, "ts": 1716454222585194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529131, "dur": 13,
    "args": {
      "External id": 77589,
      "cbid": 211, "correlation": 77589
    }
  },
  {
    "ph": "s", "id": 77589, "pid": 76337, "tid": -914061504, "ts": 1716454222529131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585203, "dur": 9,
    "args": {
      "External id": 77611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77611,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 77611, "pid": 5, "tid": 7, "ts": 1716454222585203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529163, "dur": 8,
    "args": {
      "External id": 77611,
      "cbid": 211, "correlation": 77611
    }
  },
  {
    "ph": "s", "id": 77611, "pid": 76337, "tid": -914061504, "ts": 1716454222529163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529248, "dur": 2,
    "args": {
      "External id": 77622,
      "cbid": 251, "correlation": 77622
    }
  },
  {
    "ph": "f", "id": 77622, "pid": 76337, "tid": -914061504, "ts": 1716454222529248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222585214, "dur": 1,
    "args": {
      "External id": 77623,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77623,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 77623, "pid": 5, "tid": 7, "ts": 1716454222585214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529253, "dur": 11,
    "args": {
      "External id": 77623,
      "cbid": 51, "correlation": 77623
    }
  },
  {
    "ph": "s", "id": 77623, "pid": 76337, "tid": -914061504, "ts": 1716454222529253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222585218, "dur": 37,
    "args": {
      "External id": 77624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77624,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 77624, "pid": 5, "tid": 7, "ts": 1716454222585218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529266, "dur": 13,
    "args": {
      "External id": 77624,
      "cbid": 211, "correlation": 77624
    }
  },
  {
    "ph": "s", "id": 77624, "pid": 76337, "tid": -914061504, "ts": 1716454222529266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529337, "dur": 1,
    "args": {
      "External id": 77635,
      "cbid": 251, "correlation": 77635
    }
  },
  {
    "ph": "f", "id": 77635, "pid": 76337, "tid": -914061504, "ts": 1716454222529337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529341, "dur": 0,
    "args": {
      "External id": 77636,
      "cbid": 251, "correlation": 77636
    }
  },
  {
    "ph": "f", "id": 77636, "pid": 76337, "tid": -914061504, "ts": 1716454222529341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222585256, "dur": 12,
    "args": {
      "External id": 77637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77637,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77637, "pid": 5, "tid": 7, "ts": 1716454222585256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529343, "dur": 12,
    "args": {
      "External id": 77637,
      "cbid": 211, "correlation": 77637
    }
  },
  {
    "ph": "s", "id": 77637, "pid": 76337, "tid": -914061504, "ts": 1716454222529343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222585269, "dur": 5,
    "args": {
      "External id": 77639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77639,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77639, "pid": 5, "tid": 7, "ts": 1716454222585269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529356, "dur": 5,
    "args": {
      "External id": 77639,
      "cbid": 211, "correlation": 77639
    }
  },
  {
    "ph": "s", "id": 77639, "pid": 76337, "tid": -914061504, "ts": 1716454222529356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529413, "dur": 1,
    "args": {
      "External id": 77650,
      "cbid": 251, "correlation": 77650
    }
  },
  {
    "ph": "f", "id": 77650, "pid": 76337, "tid": -914061504, "ts": 1716454222529413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529416, "dur": 0,
    "args": {
      "External id": 77651,
      "cbid": 251, "correlation": 77651
    }
  },
  {
    "ph": "f", "id": 77651, "pid": 76337, "tid": -914061504, "ts": 1716454222529416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222585276, "dur": 8,
    "args": {
      "External id": 77652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77652,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77652, "pid": 5, "tid": 7, "ts": 1716454222585276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529417, "dur": 12,
    "args": {
      "External id": 77652,
      "cbid": 211, "correlation": 77652
    }
  },
  {
    "ph": "s", "id": 77652, "pid": 76337, "tid": -914061504, "ts": 1716454222529417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222585285, "dur": 3,
    "args": {
      "External id": 77654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77654,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77654, "pid": 5, "tid": 7, "ts": 1716454222585285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529431, "dur": 5,
    "args": {
      "External id": 77654,
      "cbid": 211, "correlation": 77654
    }
  },
  {
    "ph": "s", "id": 77654, "pid": 76337, "tid": -914061504, "ts": 1716454222529431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222585290, "dur": 20,
    "args": {
      "External id": 77679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77679,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 77679, "pid": 5, "tid": 7, "ts": 1716454222585290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529508, "dur": 13,
    "args": {
      "External id": 77679,
      "cbid": 211, "correlation": 77679
    }
  },
  {
    "ph": "s", "id": 77679, "pid": 76337, "tid": -914061504, "ts": 1716454222529508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529609, "dur": 2,
    "args": {
      "External id": 77697,
      "cbid": 251, "correlation": 77697
    }
  },
  {
    "ph": "f", "id": 77697, "pid": 76337, "tid": -914061504, "ts": 1716454222529609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222585312, "dur": 1,
    "args": {
      "External id": 77699,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77699,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 77699, "pid": 5, "tid": 7, "ts": 1716454222585312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529615, "dur": 10,
    "args": {
      "External id": 77699,
      "cbid": 51, "correlation": 77699
    }
  },
  {
    "ph": "s", "id": 77699, "pid": 76337, "tid": -914061504, "ts": 1716454222529615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222585315, "dur": 38,
    "args": {
      "External id": 77700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77700,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 77700, "pid": 5, "tid": 7, "ts": 1716454222585315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529626, "dur": 13,
    "args": {
      "External id": 77700,
      "cbid": 211, "correlation": 77700
    }
  },
  {
    "ph": "s", "id": 77700, "pid": 76337, "tid": -914061504, "ts": 1716454222529626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222585354, "dur": 4,
    "args": {
      "External id": 77708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77708,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77708, "pid": 5, "tid": 7, "ts": 1716454222585354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529698, "dur": 12,
    "args": {
      "External id": 77708,
      "cbid": 211, "correlation": 77708
    }
  },
  {
    "ph": "s", "id": 77708, "pid": 76337, "tid": -914061504, "ts": 1716454222529698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585360, "dur": 8,
    "args": {
      "External id": 77716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77716, "pid": 5, "tid": 7, "ts": 1716454222585360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529739, "dur": 9,
    "args": {
      "External id": 77716,
      "cbid": 211, "correlation": 77716
    }
  },
  {
    "ph": "s", "id": 77716, "pid": 76337, "tid": -914061504, "ts": 1716454222529739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585369, "dur": 8,
    "args": {
      "External id": 77738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77738,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 77738, "pid": 5, "tid": 7, "ts": 1716454222585369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529790, "dur": 10,
    "args": {
      "External id": 77738,
      "cbid": 211, "correlation": 77738
    }
  },
  {
    "ph": "s", "id": 77738, "pid": 76337, "tid": -914061504, "ts": 1716454222529790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529880, "dur": 1,
    "args": {
      "External id": 77754,
      "cbid": 251, "correlation": 77754
    }
  },
  {
    "ph": "f", "id": 77754, "pid": 76337, "tid": -914061504, "ts": 1716454222529880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529885, "dur": 0,
    "args": {
      "External id": 77756,
      "cbid": 251, "correlation": 77756
    }
  },
  {
    "ph": "f", "id": 77756, "pid": 76337, "tid": -914061504, "ts": 1716454222529885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222585379, "dur": 193,
    "args": {
      "External id": 77757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77757,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77757, "pid": 5, "tid": 7, "ts": 1716454222585379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529887, "dur": 13,
    "args": {
      "External id": 77757,
      "cbid": 211, "correlation": 77757
    }
  },
  {
    "ph": "s", "id": 77757, "pid": 76337, "tid": -914061504, "ts": 1716454222529887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585574, "dur": 21,
    "args": {
      "External id": 77765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77765, "pid": 5, "tid": 7, "ts": 1716454222585574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529952, "dur": 12,
    "args": {
      "External id": 77765,
      "cbid": 211, "correlation": 77765
    }
  },
  {
    "ph": "s", "id": 77765, "pid": 76337, "tid": -914061504, "ts": 1716454222529952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585596, "dur": 20,
    "args": {
      "External id": 77773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77773,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77773, "pid": 5, "tid": 7, "ts": 1716454222585596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222529991, "dur": 10,
    "args": {
      "External id": 77773,
      "cbid": 211, "correlation": 77773
    }
  },
  {
    "ph": "s", "id": 77773, "pid": 76337, "tid": -914061504, "ts": 1716454222529991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530075, "dur": 1,
    "args": {
      "External id": 77789,
      "cbid": 251, "correlation": 77789
    }
  },
  {
    "ph": "f", "id": 77789, "pid": 76337, "tid": -914061504, "ts": 1716454222530075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222585618, "dur": 1,
    "args": {
      "External id": 77791,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77791,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 77791, "pid": 5, "tid": 7, "ts": 1716454222585618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530080, "dur": 8,
    "args": {
      "External id": 77791,
      "cbid": 51, "correlation": 77791
    }
  },
  {
    "ph": "s", "id": 77791, "pid": 76337, "tid": -914061504, "ts": 1716454222530080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222585622, "dur": 110,
    "args": {
      "External id": 77792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77792,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 77792, "pid": 5, "tid": 7, "ts": 1716454222585622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530090, "dur": 12,
    "args": {
      "External id": 77792,
      "cbid": 211, "correlation": 77792
    }
  },
  {
    "ph": "s", "id": 77792, "pid": 76337, "tid": -914061504, "ts": 1716454222530090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222585733, "dur": 5,
    "args": {
      "External id": 77800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77800,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77800, "pid": 5, "tid": 7, "ts": 1716454222585733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530133, "dur": 10,
    "args": {
      "External id": 77800,
      "cbid": 211, "correlation": 77800
    }
  },
  {
    "ph": "s", "id": 77800, "pid": 76337, "tid": -914061504, "ts": 1716454222530133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585740, "dur": 10,
    "args": {
      "External id": 77811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77811,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77811, "pid": 5, "tid": 7, "ts": 1716454222585740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530202, "dur": 13,
    "args": {
      "External id": 77811,
      "cbid": 211, "correlation": 77811
    }
  },
  {
    "ph": "s", "id": 77811, "pid": 76337, "tid": -914061504, "ts": 1716454222530202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530268, "dur": 0,
    "args": {
      "External id": 77823,
      "cbid": 317, "correlation": 77823
    }
  },
  {
    "ph": "f", "id": 77823, "pid": 76337, "tid": -914061504, "ts": 1716454222530268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530269, "dur": 0,
    "args": {
      "External id": 77824,
      "cbid": 203, "correlation": 77824
    }
  },
  {
    "ph": "f", "id": 77824, "pid": 76337, "tid": -914061504, "ts": 1716454222530269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530270, "dur": 0,
    "args": {
      "External id": 77825,
      "cbid": 205, "correlation": 77825
    }
  },
  {
    "ph": "f", "id": 77825, "pid": 76337, "tid": -914061504, "ts": 1716454222530270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585751, "dur": 6,
    "args": {
      "External id": 77829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77829,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77829, "pid": 5, "tid": 7, "ts": 1716454222585751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530286, "dur": 12,
    "args": {
      "External id": 77829,
      "cbid": 211, "correlation": 77829
    }
  },
  {
    "ph": "s", "id": 77829, "pid": 76337, "tid": -914061504, "ts": 1716454222530286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222585758, "dur": 38,
    "args": {
      "External id": 77831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77831,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 77831, "pid": 5, "tid": 7, "ts": 1716454222585758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530305, "dur": 7,
    "args": {
      "External id": 77831,
      "cbid": 211, "correlation": 77831
    }
  },
  {
    "ph": "s", "id": 77831, "pid": 76337, "tid": -914061504, "ts": 1716454222530305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585798, "dur": 6,
    "args": {
      "External id": 77833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77833,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77833, "pid": 5, "tid": 7, "ts": 1716454222585798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530316, "dur": 5,
    "args": {
      "External id": 77833,
      "cbid": 211, "correlation": 77833
    }
  },
  {
    "ph": "s", "id": 77833, "pid": 76337, "tid": -914061504, "ts": 1716454222530316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585805, "dur": 7,
    "args": {
      "External id": 77839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77839, "pid": 5, "tid": 7, "ts": 1716454222585805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530343, "dur": 8,
    "args": {
      "External id": 77839,
      "cbid": 211, "correlation": 77839
    }
  },
  {
    "ph": "s", "id": 77839, "pid": 76337, "tid": -914061504, "ts": 1716454222530343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222585813, "dur": 5,
    "args": {
      "External id": 77847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77847,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77847, "pid": 5, "tid": 7, "ts": 1716454222585813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530374, "dur": 8,
    "args": {
      "External id": 77847,
      "cbid": 211, "correlation": 77847
    }
  },
  {
    "ph": "s", "id": 77847, "pid": 76337, "tid": -914061504, "ts": 1716454222530374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585820, "dur": 11,
    "args": {
      "External id": 77867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77867,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77867, "pid": 5, "tid": 7, "ts": 1716454222585820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530445, "dur": 12,
    "args": {
      "External id": 77867,
      "cbid": 211, "correlation": 77867
    }
  },
  {
    "ph": "s", "id": 77867, "pid": 76337, "tid": -914061504, "ts": 1716454222530445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222585832, "dur": 4,
    "args": {
      "External id": 77879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77879,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77879, "pid": 5, "tid": 7, "ts": 1716454222585832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530467, "dur": 7,
    "args": {
      "External id": 77879,
      "cbid": 211, "correlation": 77879
    }
  },
  {
    "ph": "s", "id": 77879, "pid": 76337, "tid": -914061504, "ts": 1716454222530467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222585838, "dur": 8,
    "args": {
      "External id": 77882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77882,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77882, "pid": 5, "tid": 7, "ts": 1716454222585838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530485, "dur": 6,
    "args": {
      "External id": 77882,
      "cbid": 211, "correlation": 77882
    }
  },
  {
    "ph": "s", "id": 77882, "pid": 76337, "tid": -914061504, "ts": 1716454222530485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222585847, "dur": 5,
    "args": {
      "External id": 77891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77891,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77891, "pid": 5, "tid": 7, "ts": 1716454222585847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530524, "dur": 10,
    "args": {
      "External id": 77891,
      "cbid": 211, "correlation": 77891
    }
  },
  {
    "ph": "s", "id": 77891, "pid": 76337, "tid": -914061504, "ts": 1716454222530524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530576, "dur": 0,
    "args": {
      "External id": 77901,
      "cbid": 317, "correlation": 77901
    }
  },
  {
    "ph": "f", "id": 77901, "pid": 76337, "tid": -914061504, "ts": 1716454222530576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530576, "dur": 0,
    "args": {
      "External id": 77902,
      "cbid": 203, "correlation": 77902
    }
  },
  {
    "ph": "f", "id": 77902, "pid": 76337, "tid": -914061504, "ts": 1716454222530576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530577, "dur": 0,
    "args": {
      "External id": 77903,
      "cbid": 205, "correlation": 77903
    }
  },
  {
    "ph": "f", "id": 77903, "pid": 76337, "tid": -914061504, "ts": 1716454222530577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585854, "dur": 5,
    "args": {
      "External id": 77907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77907,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77907, "pid": 5, "tid": 7, "ts": 1716454222585854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530591, "dur": 11,
    "args": {
      "External id": 77907,
      "cbid": 211, "correlation": 77907
    }
  },
  {
    "ph": "s", "id": 77907, "pid": 76337, "tid": -914061504, "ts": 1716454222530591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222585860, "dur": 165,
    "args": {
      "External id": 77909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77909, "pid": 5, "tid": 7, "ts": 1716454222585860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530605, "dur": 5,
    "args": {
      "External id": 77909,
      "cbid": 211, "correlation": 77909
    }
  },
  {
    "ph": "s", "id": 77909, "pid": 76337, "tid": -914061504, "ts": 1716454222530605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222586028, "dur": 1,
    "args": {
      "External id": 77911,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 77911,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 77911, "pid": 5, "tid": 7, "ts": 1716454222586028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530615, "dur": 7,
    "args": {
      "External id": 77911,
      "cbid": 51, "correlation": 77911
    }
  },
  {
    "ph": "s", "id": 77911, "pid": 76337, "tid": -914061504, "ts": 1716454222530615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222586031, "dur": 275,
    "args": {
      "External id": 77912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77912,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77912, "pid": 5, "tid": 7, "ts": 1716454222586031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530624, "dur": 6,
    "args": {
      "External id": 77912,
      "cbid": 211, "correlation": 77912
    }
  },
  {
    "ph": "s", "id": 77912, "pid": 76337, "tid": -914061504, "ts": 1716454222530624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586308, "dur": 6,
    "args": {
      "External id": 77914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 77914, "pid": 5, "tid": 7, "ts": 1716454222586308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530633, "dur": 5,
    "args": {
      "External id": 77914,
      "cbid": 211, "correlation": 77914
    }
  },
  {
    "ph": "s", "id": 77914, "pid": 76337, "tid": -914061504, "ts": 1716454222530633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222586315, "dur": 6,
    "args": {
      "External id": 77920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77920,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77920, "pid": 5, "tid": 7, "ts": 1716454222586315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530661, "dur": 9,
    "args": {
      "External id": 77920,
      "cbid": 211, "correlation": 77920
    }
  },
  {
    "ph": "s", "id": 77920, "pid": 76337, "tid": -914061504, "ts": 1716454222530661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222586322, "dur": 3,
    "args": {
      "External id": 77928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77928,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 77928, "pid": 5, "tid": 7, "ts": 1716454222586322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530706, "dur": 9,
    "args": {
      "External id": 77928,
      "cbid": 211, "correlation": 77928
    }
  },
  {
    "ph": "s", "id": 77928, "pid": 76337, "tid": -914061504, "ts": 1716454222530706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530769, "dur": 1,
    "args": {
      "External id": 77944,
      "cbid": 251, "correlation": 77944
    }
  },
  {
    "ph": "f", "id": 77944, "pid": 76337, "tid": -914061504, "ts": 1716454222530769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530774, "dur": 0,
    "args": {
      "External id": 77946,
      "cbid": 251, "correlation": 77946
    }
  },
  {
    "ph": "f", "id": 77946, "pid": 76337, "tid": -914061504, "ts": 1716454222530774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222586327, "dur": 12,
    "args": {
      "External id": 77947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77947,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77947, "pid": 5, "tid": 7, "ts": 1716454222586327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530776, "dur": 12,
    "args": {
      "External id": 77947,
      "cbid": 211, "correlation": 77947
    }
  },
  {
    "ph": "s", "id": 77947, "pid": 76337, "tid": -914061504, "ts": 1716454222530776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222586341, "dur": 5,
    "args": {
      "External id": 77949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77949,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 77949, "pid": 5, "tid": 7, "ts": 1716454222586341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530790, "dur": 5,
    "args": {
      "External id": 77949,
      "cbid": 211, "correlation": 77949
    }
  },
  {
    "ph": "s", "id": 77949, "pid": 76337, "tid": -914061504, "ts": 1716454222530790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222586347, "dur": 6,
    "args": {
      "External id": 77959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77959, "pid": 5, "tid": 7, "ts": 1716454222586347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530847, "dur": 13,
    "args": {
      "External id": 77959,
      "cbid": 211, "correlation": 77959
    }
  },
  {
    "ph": "s", "id": 77959, "pid": 76337, "tid": -914061504, "ts": 1716454222530847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586355, "dur": 10,
    "args": {
      "External id": 77979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77979,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 77979, "pid": 5, "tid": 7, "ts": 1716454222586355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530914, "dur": 11,
    "args": {
      "External id": 77979,
      "cbid": 211, "correlation": 77979
    }
  },
  {
    "ph": "s", "id": 77979, "pid": 76337, "tid": -914061504, "ts": 1716454222530914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222586366, "dur": 4,
    "args": {
      "External id": 77991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77991,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 77991, "pid": 5, "tid": 7, "ts": 1716454222586366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530934, "dur": 6,
    "args": {
      "External id": 77991,
      "cbid": 211, "correlation": 77991
    }
  },
  {
    "ph": "s", "id": 77991, "pid": 76337, "tid": -914061504, "ts": 1716454222530934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222586371, "dur": 7,
    "args": {
      "External id": 77994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 77994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 77994, "pid": 5, "tid": 7, "ts": 1716454222586371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222530953, "dur": 6,
    "args": {
      "External id": 77994,
      "cbid": 211, "correlation": 77994
    }
  },
  {
    "ph": "s", "id": 77994, "pid": 76337, "tid": -914061504, "ts": 1716454222530953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222586379, "dur": 5,
    "args": {
      "External id": 78003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78003,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78003, "pid": 5, "tid": 7, "ts": 1716454222586379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531001, "dur": 11,
    "args": {
      "External id": 78003,
      "cbid": 211, "correlation": 78003
    }
  },
  {
    "ph": "s", "id": 78003, "pid": 76337, "tid": -914061504, "ts": 1716454222531001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531065, "dur": 0,
    "args": {
      "External id": 78013,
      "cbid": 317, "correlation": 78013
    }
  },
  {
    "ph": "f", "id": 78013, "pid": 76337, "tid": -914061504, "ts": 1716454222531065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531066, "dur": 0,
    "args": {
      "External id": 78014,
      "cbid": 203, "correlation": 78014
    }
  },
  {
    "ph": "f", "id": 78014, "pid": 76337, "tid": -914061504, "ts": 1716454222531066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531067, "dur": 0,
    "args": {
      "External id": 78015,
      "cbid": 205, "correlation": 78015
    }
  },
  {
    "ph": "f", "id": 78015, "pid": 76337, "tid": -914061504, "ts": 1716454222531067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586385, "dur": 5,
    "args": {
      "External id": 78019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78019,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78019, "pid": 5, "tid": 7, "ts": 1716454222586385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531080, "dur": 12,
    "args": {
      "External id": 78019,
      "cbid": 211, "correlation": 78019
    }
  },
  {
    "ph": "s", "id": 78019, "pid": 76337, "tid": -914061504, "ts": 1716454222531080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586392, "dur": 164,
    "args": {
      "External id": 78021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78021, "pid": 5, "tid": 7, "ts": 1716454222586392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531094, "dur": 6,
    "args": {
      "External id": 78021,
      "cbid": 211, "correlation": 78021
    }
  },
  {
    "ph": "s", "id": 78021, "pid": 76337, "tid": -914061504, "ts": 1716454222531094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222586559, "dur": 1,
    "args": {
      "External id": 78023,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78023,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 78023, "pid": 5, "tid": 7, "ts": 1716454222586559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531106, "dur": 6,
    "args": {
      "External id": 78023,
      "cbid": 51, "correlation": 78023
    }
  },
  {
    "ph": "s", "id": 78023, "pid": 76337, "tid": -914061504, "ts": 1716454222531106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222586562, "dur": 265,
    "args": {
      "External id": 78024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78024,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78024, "pid": 5, "tid": 7, "ts": 1716454222586562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531113, "dur": 6,
    "args": {
      "External id": 78024,
      "cbid": 211, "correlation": 78024
    }
  },
  {
    "ph": "s", "id": 78024, "pid": 76337, "tid": -914061504, "ts": 1716454222531113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586828, "dur": 6,
    "args": {
      "External id": 78026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78026, "pid": 5, "tid": 7, "ts": 1716454222586828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531123, "dur": 5,
    "args": {
      "External id": 78026,
      "cbid": 211, "correlation": 78026
    }
  },
  {
    "ph": "s", "id": 78026, "pid": 76337, "tid": -914061504, "ts": 1716454222531123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222586836, "dur": 6,
    "args": {
      "External id": 78032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78032, "pid": 5, "tid": 7, "ts": 1716454222586836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531151, "dur": 9,
    "args": {
      "External id": 78032,
      "cbid": 211, "correlation": 78032
    }
  },
  {
    "ph": "s", "id": 78032, "pid": 76337, "tid": -914061504, "ts": 1716454222531151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222586844, "dur": 5,
    "args": {
      "External id": 78040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78040,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78040, "pid": 5, "tid": 7, "ts": 1716454222586844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531184, "dur": 8,
    "args": {
      "External id": 78040,
      "cbid": 211, "correlation": 78040
    }
  },
  {
    "ph": "s", "id": 78040, "pid": 76337, "tid": -914061504, "ts": 1716454222531184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222586849, "dur": 5,
    "args": {
      "External id": 78048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78048,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78048, "pid": 5, "tid": 7, "ts": 1716454222586849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531213, "dur": 8,
    "args": {
      "External id": 78048,
      "cbid": 211, "correlation": 78048
    }
  },
  {
    "ph": "s", "id": 78048, "pid": 76337, "tid": -914061504, "ts": 1716454222531213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222586855, "dur": 11,
    "args": {
      "External id": 78057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78057,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78057, "pid": 5, "tid": 7, "ts": 1716454222586855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531301, "dur": 15,
    "args": {
      "External id": 78057,
      "cbid": 211, "correlation": 78057
    }
  },
  {
    "ph": "s", "id": 78057, "pid": 76337, "tid": -914061504, "ts": 1716454222531301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586868, "dur": 12,
    "args": {
      "External id": 78077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78077,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78077, "pid": 5, "tid": 7, "ts": 1716454222586868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531371, "dur": 11,
    "args": {
      "External id": 78077,
      "cbid": 211, "correlation": 78077
    }
  },
  {
    "ph": "s", "id": 78077, "pid": 76337, "tid": -914061504, "ts": 1716454222531371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222586881, "dur": 4,
    "args": {
      "External id": 78089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78089,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78089, "pid": 5, "tid": 7, "ts": 1716454222586881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531392, "dur": 6,
    "args": {
      "External id": 78089,
      "cbid": 211, "correlation": 78089
    }
  },
  {
    "ph": "s", "id": 78089, "pid": 76337, "tid": -914061504, "ts": 1716454222531392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222586887, "dur": 10,
    "args": {
      "External id": 78092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78092, "pid": 5, "tid": 7, "ts": 1716454222586887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531410, "dur": 8,
    "args": {
      "External id": 78092,
      "cbid": 211, "correlation": 78092
    }
  },
  {
    "ph": "s", "id": 78092, "pid": 76337, "tid": -914061504, "ts": 1716454222531410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222586899, "dur": 6,
    "args": {
      "External id": 78101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78101,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78101, "pid": 5, "tid": 7, "ts": 1716454222586899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531449, "dur": 9,
    "args": {
      "External id": 78101,
      "cbid": 211, "correlation": 78101
    }
  },
  {
    "ph": "s", "id": 78101, "pid": 76337, "tid": -914061504, "ts": 1716454222531449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531502, "dur": 0,
    "args": {
      "External id": 78111,
      "cbid": 317, "correlation": 78111
    }
  },
  {
    "ph": "f", "id": 78111, "pid": 76337, "tid": -914061504, "ts": 1716454222531502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531503, "dur": 0,
    "args": {
      "External id": 78112,
      "cbid": 203, "correlation": 78112
    }
  },
  {
    "ph": "f", "id": 78112, "pid": 76337, "tid": -914061504, "ts": 1716454222531503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531503, "dur": 0,
    "args": {
      "External id": 78113,
      "cbid": 205, "correlation": 78113
    }
  },
  {
    "ph": "f", "id": 78113, "pid": 76337, "tid": -914061504, "ts": 1716454222531503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586906, "dur": 7,
    "args": {
      "External id": 78117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78117, "pid": 5, "tid": 7, "ts": 1716454222586906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531520, "dur": 12,
    "args": {
      "External id": 78117,
      "cbid": 211, "correlation": 78117
    }
  },
  {
    "ph": "s", "id": 78117, "pid": 76337, "tid": -914061504, "ts": 1716454222531520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222586914, "dur": 324,
    "args": {
      "External id": 78119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78119,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78119, "pid": 5, "tid": 7, "ts": 1716454222586914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531534, "dur": 5,
    "args": {
      "External id": 78119,
      "cbid": 211, "correlation": 78119
    }
  },
  {
    "ph": "s", "id": 78119, "pid": 76337, "tid": -914061504, "ts": 1716454222531534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222587241, "dur": 1,
    "args": {
      "External id": 78121,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78121,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 78121, "pid": 5, "tid": 7, "ts": 1716454222587241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531545, "dur": 6,
    "args": {
      "External id": 78121,
      "cbid": 51, "correlation": 78121
    }
  },
  {
    "ph": "s", "id": 78121, "pid": 76337, "tid": -914061504, "ts": 1716454222531545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222587245, "dur": 504,
    "args": {
      "External id": 78122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78122,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78122, "pid": 5, "tid": 7, "ts": 1716454222587245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531552, "dur": 6,
    "args": {
      "External id": 78122,
      "cbid": 211, "correlation": 78122
    }
  },
  {
    "ph": "s", "id": 78122, "pid": 76337, "tid": -914061504, "ts": 1716454222531552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222587750, "dur": 6,
    "args": {
      "External id": 78124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78124, "pid": 5, "tid": 7, "ts": 1716454222587750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531563, "dur": 5,
    "args": {
      "External id": 78124,
      "cbid": 211, "correlation": 78124
    }
  },
  {
    "ph": "s", "id": 78124, "pid": 76337, "tid": -914061504, "ts": 1716454222531563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222587757, "dur": 6,
    "args": {
      "External id": 78130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78130, "pid": 5, "tid": 7, "ts": 1716454222587757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531591, "dur": 8,
    "args": {
      "External id": 78130,
      "cbid": 211, "correlation": 78130
    }
  },
  {
    "ph": "s", "id": 78130, "pid": 76337, "tid": -914061504, "ts": 1716454222531591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222587765, "dur": 3,
    "args": {
      "External id": 78138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78138,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 78138, "pid": 5, "tid": 7, "ts": 1716454222587765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531635, "dur": 9,
    "args": {
      "External id": 78138,
      "cbid": 211, "correlation": 78138
    }
  },
  {
    "ph": "s", "id": 78138, "pid": 76337, "tid": -914061504, "ts": 1716454222531635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531697, "dur": 1,
    "args": {
      "External id": 78154,
      "cbid": 251, "correlation": 78154
    }
  },
  {
    "ph": "f", "id": 78154, "pid": 76337, "tid": -914061504, "ts": 1716454222531697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531702, "dur": 0,
    "args": {
      "External id": 78156,
      "cbid": 251, "correlation": 78156
    }
  },
  {
    "ph": "f", "id": 78156, "pid": 76337, "tid": -914061504, "ts": 1716454222531702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222587769, "dur": 11,
    "args": {
      "External id": 78157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78157,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78157, "pid": 5, "tid": 7, "ts": 1716454222587769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531704, "dur": 11,
    "args": {
      "External id": 78157,
      "cbid": 211, "correlation": 78157
    }
  },
  {
    "ph": "s", "id": 78157, "pid": 76337, "tid": -914061504, "ts": 1716454222531704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222587782, "dur": 5,
    "args": {
      "External id": 78159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78159,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78159, "pid": 5, "tid": 7, "ts": 1716454222587782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531717, "dur": 5,
    "args": {
      "External id": 78159,
      "cbid": 211, "correlation": 78159
    }
  },
  {
    "ph": "s", "id": 78159, "pid": 76337, "tid": -914061504, "ts": 1716454222531717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222587788, "dur": 6,
    "args": {
      "External id": 78169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78169, "pid": 5, "tid": 7, "ts": 1716454222587788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531774, "dur": 12,
    "args": {
      "External id": 78169,
      "cbid": 211, "correlation": 78169
    }
  },
  {
    "ph": "s", "id": 78169, "pid": 76337, "tid": -914061504, "ts": 1716454222531774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222587795, "dur": 10,
    "args": {
      "External id": 78189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78189,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78189, "pid": 5, "tid": 7, "ts": 1716454222587795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531840, "dur": 11,
    "args": {
      "External id": 78189,
      "cbid": 211, "correlation": 78189
    }
  },
  {
    "ph": "s", "id": 78189, "pid": 76337, "tid": -914061504, "ts": 1716454222531840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222587806, "dur": 4,
    "args": {
      "External id": 78201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78201,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 78201, "pid": 5, "tid": 7, "ts": 1716454222587806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531861, "dur": 6,
    "args": {
      "External id": 78201,
      "cbid": 211, "correlation": 78201
    }
  },
  {
    "ph": "s", "id": 78201, "pid": 76337, "tid": -914061504, "ts": 1716454222531861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222587811, "dur": 7,
    "args": {
      "External id": 78204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78204, "pid": 5, "tid": 7, "ts": 1716454222587811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531879, "dur": 7,
    "args": {
      "External id": 78204,
      "cbid": 211, "correlation": 78204
    }
  },
  {
    "ph": "s", "id": 78204, "pid": 76337, "tid": -914061504, "ts": 1716454222531879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222587819, "dur": 5,
    "args": {
      "External id": 78213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78213,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78213, "pid": 5, "tid": 7, "ts": 1716454222587819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531919, "dur": 10,
    "args": {
      "External id": 78213,
      "cbid": 211, "correlation": 78213
    }
  },
  {
    "ph": "s", "id": 78213, "pid": 76337, "tid": -914061504, "ts": 1716454222531919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531988, "dur": 0,
    "args": {
      "External id": 78223,
      "cbid": 317, "correlation": 78223
    }
  },
  {
    "ph": "f", "id": 78223, "pid": 76337, "tid": -914061504, "ts": 1716454222531988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531989, "dur": 0,
    "args": {
      "External id": 78224,
      "cbid": 203, "correlation": 78224
    }
  },
  {
    "ph": "f", "id": 78224, "pid": 76337, "tid": -914061504, "ts": 1716454222531989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222531990, "dur": 0,
    "args": {
      "External id": 78225,
      "cbid": 205, "correlation": 78225
    }
  },
  {
    "ph": "f", "id": 78225, "pid": 76337, "tid": -914061504, "ts": 1716454222531990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222587825, "dur": 5,
    "args": {
      "External id": 78229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78229, "pid": 5, "tid": 7, "ts": 1716454222587825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532004, "dur": 12,
    "args": {
      "External id": 78229,
      "cbid": 211, "correlation": 78229
    }
  },
  {
    "ph": "s", "id": 78229, "pid": 76337, "tid": -914061504, "ts": 1716454222532004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222587832, "dur": 166,
    "args": {
      "External id": 78231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78231, "pid": 5, "tid": 7, "ts": 1716454222587832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532018, "dur": 5,
    "args": {
      "External id": 78231,
      "cbid": 211, "correlation": 78231
    }
  },
  {
    "ph": "s", "id": 78231, "pid": 76337, "tid": -914061504, "ts": 1716454222532018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222588000, "dur": 1,
    "args": {
      "External id": 78233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78233,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 78233, "pid": 5, "tid": 7, "ts": 1716454222588000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532029, "dur": 7,
    "args": {
      "External id": 78233,
      "cbid": 51, "correlation": 78233
    }
  },
  {
    "ph": "s", "id": 78233, "pid": 76337, "tid": -914061504, "ts": 1716454222532029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222588003, "dur": 264,
    "args": {
      "External id": 78234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78234,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78234, "pid": 5, "tid": 7, "ts": 1716454222588003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532036, "dur": 7,
    "args": {
      "External id": 78234,
      "cbid": 211, "correlation": 78234
    }
  },
  {
    "ph": "s", "id": 78234, "pid": 76337, "tid": -914061504, "ts": 1716454222532036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222588268, "dur": 6,
    "args": {
      "External id": 78236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78236, "pid": 5, "tid": 7, "ts": 1716454222588268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532046, "dur": 5,
    "args": {
      "External id": 78236,
      "cbid": 211, "correlation": 78236
    }
  },
  {
    "ph": "s", "id": 78236, "pid": 76337, "tid": -914061504, "ts": 1716454222532046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222588276, "dur": 6,
    "args": {
      "External id": 78242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78242, "pid": 5, "tid": 7, "ts": 1716454222588276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532074, "dur": 8,
    "args": {
      "External id": 78242,
      "cbid": 211, "correlation": 78242
    }
  },
  {
    "ph": "s", "id": 78242, "pid": 76337, "tid": -914061504, "ts": 1716454222532074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532133, "dur": 0,
    "args": {
      "External id": 78252,
      "cbid": 317, "correlation": 78252
    }
  },
  {
    "ph": "f", "id": 78252, "pid": 76337, "tid": -914061504, "ts": 1716454222532133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532133, "dur": 0,
    "args": {
      "External id": 78253,
      "cbid": 203, "correlation": 78253
    }
  },
  {
    "ph": "f", "id": 78253, "pid": 76337, "tid": -914061504, "ts": 1716454222532133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532134, "dur": 0,
    "args": {
      "External id": 78254,
      "cbid": 205, "correlation": 78254
    }
  },
  {
    "ph": "f", "id": 78254, "pid": 76337, "tid": -914061504, "ts": 1716454222532134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222588283, "dur": 8,
    "args": {
      "External id": 78258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78258,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78258, "pid": 5, "tid": 7, "ts": 1716454222588283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532149, "dur": 12,
    "args": {
      "External id": 78258,
      "cbid": 211, "correlation": 78258
    }
  },
  {
    "ph": "s", "id": 78258, "pid": 76337, "tid": -914061504, "ts": 1716454222532149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222588292, "dur": 3,
    "args": {
      "External id": 78260,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78260,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4846891432106404
    }
  },
  {
    "ph": "f", "id": 78260, "pid": 5, "tid": 7, "ts": 1716454222588292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532167, "dur": 14,
    "args": {
      "External id": 78260,
      "cbid": 51, "correlation": 78260
    }
  },
  {
    "ph": "s", "id": 78260, "pid": 76337, "tid": -914061504, "ts": 1716454222532167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222588296, "dur": 98,
    "args": {
      "External id": 78261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78261,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 78261, "pid": 5, "tid": 7, "ts": 1716454222588296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532181, "dur": 7,
    "args": {
      "External id": 78261,
      "cbid": 211, "correlation": 78261
    }
  },
  {
    "ph": "s", "id": 78261, "pid": 76337, "tid": -914061504, "ts": 1716454222532181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222588396, "dur": 5,
    "args": {
      "External id": 78263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78263,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78263, "pid": 5, "tid": 7, "ts": 1716454222588396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532193, "dur": 5,
    "args": {
      "External id": 78263,
      "cbid": 211, "correlation": 78263
    }
  },
  {
    "ph": "s", "id": 78263, "pid": 76337, "tid": -914061504, "ts": 1716454222532193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222588403, "dur": 6,
    "args": {
      "External id": 78269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78269,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78269, "pid": 5, "tid": 7, "ts": 1716454222588403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532219, "dur": 8,
    "args": {
      "External id": 78269,
      "cbid": 211, "correlation": 78269
    }
  },
  {
    "ph": "s", "id": 78269, "pid": 76337, "tid": -914061504, "ts": 1716454222532219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222588411, "dur": 5,
    "args": {
      "External id": 78277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78277,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78277, "pid": 5, "tid": 7, "ts": 1716454222588411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532249, "dur": 7,
    "args": {
      "External id": 78277,
      "cbid": 211, "correlation": 78277
    }
  },
  {
    "ph": "s", "id": 78277, "pid": 76337, "tid": -914061504, "ts": 1716454222532249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222588417, "dur": 4,
    "args": {
      "External id": 78285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78285,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78285, "pid": 5, "tid": 7, "ts": 1716454222588417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532277, "dur": 8,
    "args": {
      "External id": 78285,
      "cbid": 211, "correlation": 78285
    }
  },
  {
    "ph": "s", "id": 78285, "pid": 76337, "tid": -914061504, "ts": 1716454222532277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222588423, "dur": 11,
    "args": {
      "External id": 78294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78294,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78294, "pid": 5, "tid": 7, "ts": 1716454222588423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532320, "dur": 10,
    "args": {
      "External id": 78294,
      "cbid": 211, "correlation": 78294
    }
  },
  {
    "ph": "s", "id": 78294, "pid": 76337, "tid": -914061504, "ts": 1716454222532320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222588435, "dur": 12,
    "args": {
      "External id": 78314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78314,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78314, "pid": 5, "tid": 7, "ts": 1716454222588435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532389, "dur": 11,
    "args": {
      "External id": 78314,
      "cbid": 211, "correlation": 78314
    }
  },
  {
    "ph": "s", "id": 78314, "pid": 76337, "tid": -914061504, "ts": 1716454222532389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222588449, "dur": 4,
    "args": {
      "External id": 78326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78326,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78326, "pid": 5, "tid": 7, "ts": 1716454222588449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532410, "dur": 6,
    "args": {
      "External id": 78326,
      "cbid": 211, "correlation": 78326
    }
  },
  {
    "ph": "s", "id": 78326, "pid": 76337, "tid": -914061504, "ts": 1716454222532410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222588455, "dur": 11,
    "args": {
      "External id": 78329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78329,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78329, "pid": 5, "tid": 7, "ts": 1716454222588455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532428, "dur": 7,
    "args": {
      "External id": 78329,
      "cbid": 211, "correlation": 78329
    }
  },
  {
    "ph": "s", "id": 78329, "pid": 76337, "tid": -914061504, "ts": 1716454222532428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222588467, "dur": 6,
    "args": {
      "External id": 78338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78338,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78338, "pid": 5, "tid": 7, "ts": 1716454222588467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532467, "dur": 9,
    "args": {
      "External id": 78338,
      "cbid": 211, "correlation": 78338
    }
  },
  {
    "ph": "s", "id": 78338, "pid": 76337, "tid": -914061504, "ts": 1716454222532467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532518, "dur": 0,
    "args": {
      "External id": 78348,
      "cbid": 317, "correlation": 78348
    }
  },
  {
    "ph": "f", "id": 78348, "pid": 76337, "tid": -914061504, "ts": 1716454222532518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532518, "dur": 0,
    "args": {
      "External id": 78349,
      "cbid": 203, "correlation": 78349
    }
  },
  {
    "ph": "f", "id": 78349, "pid": 76337, "tid": -914061504, "ts": 1716454222532518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532519, "dur": 0,
    "args": {
      "External id": 78350,
      "cbid": 205, "correlation": 78350
    }
  },
  {
    "ph": "f", "id": 78350, "pid": 76337, "tid": -914061504, "ts": 1716454222532519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222588475, "dur": 7,
    "args": {
      "External id": 78354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78354,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78354, "pid": 5, "tid": 7, "ts": 1716454222588475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532533, "dur": 11,
    "args": {
      "External id": 78354,
      "cbid": 211, "correlation": 78354
    }
  },
  {
    "ph": "s", "id": 78354, "pid": 76337, "tid": -914061504, "ts": 1716454222532533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222588483, "dur": 326,
    "args": {
      "External id": 78356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78356, "pid": 5, "tid": 7, "ts": 1716454222588483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532547, "dur": 5,
    "args": {
      "External id": 78356,
      "cbid": 211, "correlation": 78356
    }
  },
  {
    "ph": "s", "id": 78356, "pid": 76337, "tid": -914061504, "ts": 1716454222532547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222588811, "dur": 1,
    "args": {
      "External id": 78358,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78358,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 78358, "pid": 5, "tid": 7, "ts": 1716454222588811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532558, "dur": 6,
    "args": {
      "External id": 78358,
      "cbid": 51, "correlation": 78358
    }
  },
  {
    "ph": "s", "id": 78358, "pid": 76337, "tid": -914061504, "ts": 1716454222532558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222588814, "dur": 505,
    "args": {
      "External id": 78359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78359,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78359, "pid": 5, "tid": 7, "ts": 1716454222588814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532565, "dur": 6,
    "args": {
      "External id": 78359,
      "cbid": 211, "correlation": 78359
    }
  },
  {
    "ph": "s", "id": 78359, "pid": 76337, "tid": -914061504, "ts": 1716454222532565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589321, "dur": 6,
    "args": {
      "External id": 78361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78361,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78361, "pid": 5, "tid": 7, "ts": 1716454222589321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532576, "dur": 5,
    "args": {
      "External id": 78361,
      "cbid": 211, "correlation": 78361
    }
  },
  {
    "ph": "s", "id": 78361, "pid": 76337, "tid": -914061504, "ts": 1716454222532576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222589328, "dur": 6,
    "args": {
      "External id": 78367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78367, "pid": 5, "tid": 7, "ts": 1716454222589328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532604, "dur": 8,
    "args": {
      "External id": 78367,
      "cbid": 211, "correlation": 78367
    }
  },
  {
    "ph": "s", "id": 78367, "pid": 76337, "tid": -914061504, "ts": 1716454222532604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222589336, "dur": 3,
    "args": {
      "External id": 78375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78375,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 78375, "pid": 5, "tid": 7, "ts": 1716454222589336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532646, "dur": 9,
    "args": {
      "External id": 78375,
      "cbid": 211, "correlation": 78375
    }
  },
  {
    "ph": "s", "id": 78375, "pid": 76337, "tid": -914061504, "ts": 1716454222532646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532709, "dur": 1,
    "args": {
      "External id": 78391,
      "cbid": 251, "correlation": 78391
    }
  },
  {
    "ph": "f", "id": 78391, "pid": 76337, "tid": -914061504, "ts": 1716454222532709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532714, "dur": 0,
    "args": {
      "External id": 78393,
      "cbid": 251, "correlation": 78393
    }
  },
  {
    "ph": "f", "id": 78393, "pid": 76337, "tid": -914061504, "ts": 1716454222532714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222589340, "dur": 13,
    "args": {
      "External id": 78394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78394,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78394, "pid": 5, "tid": 7, "ts": 1716454222589340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532716, "dur": 11,
    "args": {
      "External id": 78394,
      "cbid": 211, "correlation": 78394
    }
  },
  {
    "ph": "s", "id": 78394, "pid": 76337, "tid": -914061504, "ts": 1716454222532716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222589354, "dur": 5,
    "args": {
      "External id": 78396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78396,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78396, "pid": 5, "tid": 7, "ts": 1716454222589354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532729, "dur": 5,
    "args": {
      "External id": 78396,
      "cbid": 211, "correlation": 78396
    }
  },
  {
    "ph": "s", "id": 78396, "pid": 76337, "tid": -914061504, "ts": 1716454222532729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222589361, "dur": 6,
    "args": {
      "External id": 78406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78406,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78406, "pid": 5, "tid": 7, "ts": 1716454222589361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532785, "dur": 11,
    "args": {
      "External id": 78406,
      "cbid": 211, "correlation": 78406
    }
  },
  {
    "ph": "s", "id": 78406, "pid": 76337, "tid": -914061504, "ts": 1716454222532785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589368, "dur": 10,
    "args": {
      "External id": 78426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78426,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78426, "pid": 5, "tid": 7, "ts": 1716454222589368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532849, "dur": 10,
    "args": {
      "External id": 78426,
      "cbid": 211, "correlation": 78426
    }
  },
  {
    "ph": "s", "id": 78426, "pid": 76337, "tid": -914061504, "ts": 1716454222532849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222589379, "dur": 4,
    "args": {
      "External id": 78438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78438,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 78438, "pid": 5, "tid": 7, "ts": 1716454222589379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532869, "dur": 6,
    "args": {
      "External id": 78438,
      "cbid": 211, "correlation": 78438
    }
  },
  {
    "ph": "s", "id": 78438, "pid": 76337, "tid": -914061504, "ts": 1716454222532869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222589384, "dur": 7,
    "args": {
      "External id": 78441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78441,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78441, "pid": 5, "tid": 7, "ts": 1716454222589384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532888, "dur": 6,
    "args": {
      "External id": 78441,
      "cbid": 211, "correlation": 78441
    }
  },
  {
    "ph": "s", "id": 78441, "pid": 76337, "tid": -914061504, "ts": 1716454222532888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222589392, "dur": 5,
    "args": {
      "External id": 78450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78450,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78450, "pid": 5, "tid": 7, "ts": 1716454222589392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532927, "dur": 10,
    "args": {
      "External id": 78450,
      "cbid": 211, "correlation": 78450
    }
  },
  {
    "ph": "s", "id": 78450, "pid": 76337, "tid": -914061504, "ts": 1716454222532927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222532999, "dur": 0,
    "args": {
      "External id": 78460,
      "cbid": 317, "correlation": 78460
    }
  },
  {
    "ph": "f", "id": 78460, "pid": 76337, "tid": -914061504, "ts": 1716454222532999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533000, "dur": 0,
    "args": {
      "External id": 78461,
      "cbid": 203, "correlation": 78461
    }
  },
  {
    "ph": "f", "id": 78461, "pid": 76337, "tid": -914061504, "ts": 1716454222533000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533001, "dur": 0,
    "args": {
      "External id": 78462,
      "cbid": 205, "correlation": 78462
    }
  },
  {
    "ph": "f", "id": 78462, "pid": 76337, "tid": -914061504, "ts": 1716454222533001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589398, "dur": 5,
    "args": {
      "External id": 78466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78466,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78466, "pid": 5, "tid": 7, "ts": 1716454222589398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533014, "dur": 12,
    "args": {
      "External id": 78466,
      "cbid": 211, "correlation": 78466
    }
  },
  {
    "ph": "s", "id": 78466, "pid": 76337, "tid": -914061504, "ts": 1716454222533014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589405, "dur": 165,
    "args": {
      "External id": 78468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78468,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78468, "pid": 5, "tid": 7, "ts": 1716454222589405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533029, "dur": 5,
    "args": {
      "External id": 78468,
      "cbid": 211, "correlation": 78468
    }
  },
  {
    "ph": "s", "id": 78468, "pid": 76337, "tid": -914061504, "ts": 1716454222533029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222589572, "dur": 1,
    "args": {
      "External id": 78470,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78470,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 78470, "pid": 5, "tid": 7, "ts": 1716454222589572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533040, "dur": 6,
    "args": {
      "External id": 78470,
      "cbid": 51, "correlation": 78470
    }
  },
  {
    "ph": "s", "id": 78470, "pid": 76337, "tid": -914061504, "ts": 1716454222533040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222589576, "dur": 264,
    "args": {
      "External id": 78471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78471,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78471, "pid": 5, "tid": 7, "ts": 1716454222589576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533047, "dur": 6,
    "args": {
      "External id": 78471,
      "cbid": 211, "correlation": 78471
    }
  },
  {
    "ph": "s", "id": 78471, "pid": 76337, "tid": -914061504, "ts": 1716454222533047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589841, "dur": 6,
    "args": {
      "External id": 78473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78473,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78473, "pid": 5, "tid": 7, "ts": 1716454222589841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533057, "dur": 5,
    "args": {
      "External id": 78473,
      "cbid": 211, "correlation": 78473
    }
  },
  {
    "ph": "s", "id": 78473, "pid": 76337, "tid": -914061504, "ts": 1716454222533057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222589848, "dur": 6,
    "args": {
      "External id": 78479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78479, "pid": 5, "tid": 7, "ts": 1716454222589848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533085, "dur": 8,
    "args": {
      "External id": 78479,
      "cbid": 211, "correlation": 78479
    }
  },
  {
    "ph": "s", "id": 78479, "pid": 76337, "tid": -914061504, "ts": 1716454222533085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533145, "dur": 0,
    "args": {
      "External id": 78489,
      "cbid": 317, "correlation": 78489
    }
  },
  {
    "ph": "f", "id": 78489, "pid": 76337, "tid": -914061504, "ts": 1716454222533145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533146, "dur": 0,
    "args": {
      "External id": 78490,
      "cbid": 203, "correlation": 78490
    }
  },
  {
    "ph": "f", "id": 78490, "pid": 76337, "tid": -914061504, "ts": 1716454222533146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533147, "dur": 0,
    "args": {
      "External id": 78491,
      "cbid": 205, "correlation": 78491
    }
  },
  {
    "ph": "f", "id": 78491, "pid": 76337, "tid": -914061504, "ts": 1716454222533147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589856, "dur": 8,
    "args": {
      "External id": 78495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78495,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78495, "pid": 5, "tid": 7, "ts": 1716454222589856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533158, "dur": 11,
    "args": {
      "External id": 78495,
      "cbid": 211, "correlation": 78495
    }
  },
  {
    "ph": "s", "id": 78495, "pid": 76337, "tid": -914061504, "ts": 1716454222533158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222589866, "dur": 3,
    "args": {
      "External id": 78497,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78497,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 78497, "pid": 5, "tid": 7, "ts": 1716454222589866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533174, "dur": 9,
    "args": {
      "External id": 78497,
      "cbid": 51, "correlation": 78497
    }
  },
  {
    "ph": "s", "id": 78497, "pid": 76337, "tid": -914061504, "ts": 1716454222533174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222589870, "dur": 96,
    "args": {
      "External id": 78498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78498,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 78498, "pid": 5, "tid": 7, "ts": 1716454222589870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533184, "dur": 7,
    "args": {
      "External id": 78498,
      "cbid": 211, "correlation": 78498
    }
  },
  {
    "ph": "s", "id": 78498, "pid": 76337, "tid": -914061504, "ts": 1716454222533184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222589968, "dur": 6,
    "args": {
      "External id": 78500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78500,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78500, "pid": 5, "tid": 7, "ts": 1716454222589968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533196, "dur": 5,
    "args": {
      "External id": 78500,
      "cbid": 211, "correlation": 78500
    }
  },
  {
    "ph": "s", "id": 78500, "pid": 76337, "tid": -914061504, "ts": 1716454222533196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222589975, "dur": 7,
    "args": {
      "External id": 78506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78506,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78506, "pid": 5, "tid": 7, "ts": 1716454222589975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533221, "dur": 8,
    "args": {
      "External id": 78506,
      "cbid": 211, "correlation": 78506
    }
  },
  {
    "ph": "s", "id": 78506, "pid": 76337, "tid": -914061504, "ts": 1716454222533221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222589983, "dur": 5,
    "args": {
      "External id": 78514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78514,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78514, "pid": 5, "tid": 7, "ts": 1716454222589983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533251, "dur": 8,
    "args": {
      "External id": 78514,
      "cbid": 211, "correlation": 78514
    }
  },
  {
    "ph": "s", "id": 78514, "pid": 76337, "tid": -914061504, "ts": 1716454222533251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222589990, "dur": 5,
    "args": {
      "External id": 78522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78522,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78522, "pid": 5, "tid": 7, "ts": 1716454222589990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533279, "dur": 8,
    "args": {
      "External id": 78522,
      "cbid": 211, "correlation": 78522
    }
  },
  {
    "ph": "s", "id": 78522, "pid": 76337, "tid": -914061504, "ts": 1716454222533279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222589995, "dur": 11,
    "args": {
      "External id": 78531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78531,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78531, "pid": 5, "tid": 7, "ts": 1716454222589995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533322, "dur": 10,
    "args": {
      "External id": 78531,
      "cbid": 211, "correlation": 78531
    }
  },
  {
    "ph": "s", "id": 78531, "pid": 76337, "tid": -914061504, "ts": 1716454222533322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590008, "dur": 12,
    "args": {
      "External id": 78551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78551,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78551, "pid": 5, "tid": 7, "ts": 1716454222590008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533390, "dur": 12,
    "args": {
      "External id": 78551,
      "cbid": 211, "correlation": 78551
    }
  },
  {
    "ph": "s", "id": 78551, "pid": 76337, "tid": -914061504, "ts": 1716454222533390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222590022, "dur": 4,
    "args": {
      "External id": 78563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78563,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78563, "pid": 5, "tid": 7, "ts": 1716454222590022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533411, "dur": 6,
    "args": {
      "External id": 78563,
      "cbid": 211, "correlation": 78563
    }
  },
  {
    "ph": "s", "id": 78563, "pid": 76337, "tid": -914061504, "ts": 1716454222533411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222590027, "dur": 11,
    "args": {
      "External id": 78566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78566, "pid": 5, "tid": 7, "ts": 1716454222590027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533429, "dur": 8,
    "args": {
      "External id": 78566,
      "cbid": 211, "correlation": 78566
    }
  },
  {
    "ph": "s", "id": 78566, "pid": 76337, "tid": -914061504, "ts": 1716454222533429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222590039, "dur": 6,
    "args": {
      "External id": 78575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78575,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78575, "pid": 5, "tid": 7, "ts": 1716454222590039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533469, "dur": 9,
    "args": {
      "External id": 78575,
      "cbid": 211, "correlation": 78575
    }
  },
  {
    "ph": "s", "id": 78575, "pid": 76337, "tid": -914061504, "ts": 1716454222533469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533519, "dur": 0,
    "args": {
      "External id": 78585,
      "cbid": 317, "correlation": 78585
    }
  },
  {
    "ph": "f", "id": 78585, "pid": 76337, "tid": -914061504, "ts": 1716454222533519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533520, "dur": 0,
    "args": {
      "External id": 78586,
      "cbid": 203, "correlation": 78586
    }
  },
  {
    "ph": "f", "id": 78586, "pid": 76337, "tid": -914061504, "ts": 1716454222533520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533521, "dur": 0,
    "args": {
      "External id": 78587,
      "cbid": 205, "correlation": 78587
    }
  },
  {
    "ph": "f", "id": 78587, "pid": 76337, "tid": -914061504, "ts": 1716454222533521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590047, "dur": 7,
    "args": {
      "External id": 78591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78591, "pid": 5, "tid": 7, "ts": 1716454222590047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533534, "dur": 12,
    "args": {
      "External id": 78591,
      "cbid": 211, "correlation": 78591
    }
  },
  {
    "ph": "s", "id": 78591, "pid": 76337, "tid": -914061504, "ts": 1716454222533534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590055, "dur": 325,
    "args": {
      "External id": 78593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78593,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78593, "pid": 5, "tid": 7, "ts": 1716454222590055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533548, "dur": 5,
    "args": {
      "External id": 78593,
      "cbid": 211, "correlation": 78593
    }
  },
  {
    "ph": "s", "id": 78593, "pid": 76337, "tid": -914061504, "ts": 1716454222533548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222590383, "dur": 1,
    "args": {
      "External id": 78595,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78595,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15315890236119975
    }
  },
  {
    "ph": "f", "id": 78595, "pid": 5, "tid": 7, "ts": 1716454222590383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533559, "dur": 6,
    "args": {
      "External id": 78595,
      "cbid": 51, "correlation": 78595
    }
  },
  {
    "ph": "s", "id": 78595, "pid": 76337, "tid": -914061504, "ts": 1716454222533559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222590386, "dur": 505,
    "args": {
      "External id": 78596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78596,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78596, "pid": 5, "tid": 7, "ts": 1716454222590386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533566, "dur": 6,
    "args": {
      "External id": 78596,
      "cbid": 211, "correlation": 78596
    }
  },
  {
    "ph": "s", "id": 78596, "pid": 76337, "tid": -914061504, "ts": 1716454222533566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590893, "dur": 6,
    "args": {
      "External id": 78598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78598, "pid": 5, "tid": 7, "ts": 1716454222590893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533575, "dur": 5,
    "args": {
      "External id": 78598,
      "cbid": 211, "correlation": 78598
    }
  },
  {
    "ph": "s", "id": 78598, "pid": 76337, "tid": -914061504, "ts": 1716454222533575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222590900, "dur": 6,
    "args": {
      "External id": 78604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78604, "pid": 5, "tid": 7, "ts": 1716454222590900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533604, "dur": 8,
    "args": {
      "External id": 78604,
      "cbid": 211, "correlation": 78604
    }
  },
  {
    "ph": "s", "id": 78604, "pid": 76337, "tid": -914061504, "ts": 1716454222533604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222590908, "dur": 3,
    "args": {
      "External id": 78612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78612,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 78612, "pid": 5, "tid": 7, "ts": 1716454222590908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533646, "dur": 10,
    "args": {
      "External id": 78612,
      "cbid": 211, "correlation": 78612
    }
  },
  {
    "ph": "s", "id": 78612, "pid": 76337, "tid": -914061504, "ts": 1716454222533646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533709, "dur": 1,
    "args": {
      "External id": 78628,
      "cbid": 251, "correlation": 78628
    }
  },
  {
    "ph": "f", "id": 78628, "pid": 76337, "tid": -914061504, "ts": 1716454222533709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533714, "dur": 0,
    "args": {
      "External id": 78630,
      "cbid": 251, "correlation": 78630
    }
  },
  {
    "ph": "f", "id": 78630, "pid": 76337, "tid": -914061504, "ts": 1716454222533714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222590912, "dur": 13,
    "args": {
      "External id": 78631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78631,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78631, "pid": 5, "tid": 7, "ts": 1716454222590912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533716, "dur": 11,
    "args": {
      "External id": 78631,
      "cbid": 211, "correlation": 78631
    }
  },
  {
    "ph": "s", "id": 78631, "pid": 76337, "tid": -914061504, "ts": 1716454222533716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222590926, "dur": 5,
    "args": {
      "External id": 78633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78633,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78633, "pid": 5, "tid": 7, "ts": 1716454222590926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533728, "dur": 5,
    "args": {
      "External id": 78633,
      "cbid": 211, "correlation": 78633
    }
  },
  {
    "ph": "s", "id": 78633, "pid": 76337, "tid": -914061504, "ts": 1716454222533728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222590933, "dur": 6,
    "args": {
      "External id": 78643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78643, "pid": 5, "tid": 7, "ts": 1716454222590933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533785, "dur": 11,
    "args": {
      "External id": 78643,
      "cbid": 211, "correlation": 78643
    }
  },
  {
    "ph": "s", "id": 78643, "pid": 76337, "tid": -914061504, "ts": 1716454222533785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590940, "dur": 10,
    "args": {
      "External id": 78663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78663,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78663, "pid": 5, "tid": 7, "ts": 1716454222590940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533849, "dur": 11,
    "args": {
      "External id": 78663,
      "cbid": 211, "correlation": 78663
    }
  },
  {
    "ph": "s", "id": 78663, "pid": 76337, "tid": -914061504, "ts": 1716454222533849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222590951, "dur": 4,
    "args": {
      "External id": 78675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78675,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 78675, "pid": 5, "tid": 7, "ts": 1716454222590951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533871, "dur": 6,
    "args": {
      "External id": 78675,
      "cbid": 211, "correlation": 78675
    }
  },
  {
    "ph": "s", "id": 78675, "pid": 76337, "tid": -914061504, "ts": 1716454222533871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222590956, "dur": 7,
    "args": {
      "External id": 78678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78678,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78678, "pid": 5, "tid": 7, "ts": 1716454222590956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533889, "dur": 6,
    "args": {
      "External id": 78678,
      "cbid": 211, "correlation": 78678
    }
  },
  {
    "ph": "s", "id": 78678, "pid": 76337, "tid": -914061504, "ts": 1716454222533889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222590964, "dur": 5,
    "args": {
      "External id": 78687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78687,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78687, "pid": 5, "tid": 7, "ts": 1716454222590964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533928, "dur": 10,
    "args": {
      "External id": 78687,
      "cbid": 211, "correlation": 78687
    }
  },
  {
    "ph": "s", "id": 78687, "pid": 76337, "tid": -914061504, "ts": 1716454222533928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533999, "dur": 0,
    "args": {
      "External id": 78697,
      "cbid": 317, "correlation": 78697
    }
  },
  {
    "ph": "f", "id": 78697, "pid": 76337, "tid": -914061504, "ts": 1716454222533999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222533999, "dur": 0,
    "args": {
      "External id": 78698,
      "cbid": 203, "correlation": 78698
    }
  },
  {
    "ph": "f", "id": 78698, "pid": 76337, "tid": -914061504, "ts": 1716454222533999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534000, "dur": 0,
    "args": {
      "External id": 78699,
      "cbid": 205, "correlation": 78699
    }
  },
  {
    "ph": "f", "id": 78699, "pid": 76337, "tid": -914061504, "ts": 1716454222534000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590970, "dur": 5,
    "args": {
      "External id": 78703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78703,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78703, "pid": 5, "tid": 7, "ts": 1716454222590970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534013, "dur": 13,
    "args": {
      "External id": 78703,
      "cbid": 211, "correlation": 78703
    }
  },
  {
    "ph": "s", "id": 78703, "pid": 76337, "tid": -914061504, "ts": 1716454222534013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222590977, "dur": 167,
    "args": {
      "External id": 78705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78705,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78705, "pid": 5, "tid": 7, "ts": 1716454222590977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534029, "dur": 5,
    "args": {
      "External id": 78705,
      "cbid": 211, "correlation": 78705
    }
  },
  {
    "ph": "s", "id": 78705, "pid": 76337, "tid": -914061504, "ts": 1716454222534029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222591146, "dur": 1,
    "args": {
      "External id": 78707,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78707,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 78707, "pid": 5, "tid": 7, "ts": 1716454222591146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534040, "dur": 7,
    "args": {
      "External id": 78707,
      "cbid": 51, "correlation": 78707
    }
  },
  {
    "ph": "s", "id": 78707, "pid": 76337, "tid": -914061504, "ts": 1716454222534040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222591149, "dur": 264,
    "args": {
      "External id": 78708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78708,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78708, "pid": 5, "tid": 7, "ts": 1716454222591149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534047, "dur": 6,
    "args": {
      "External id": 78708,
      "cbid": 211, "correlation": 78708
    }
  },
  {
    "ph": "s", "id": 78708, "pid": 76337, "tid": -914061504, "ts": 1716454222534047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222591415, "dur": 6,
    "args": {
      "External id": 78710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78710,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78710, "pid": 5, "tid": 7, "ts": 1716454222591415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534057, "dur": 5,
    "args": {
      "External id": 78710,
      "cbid": 211, "correlation": 78710
    }
  },
  {
    "ph": "s", "id": 78710, "pid": 76337, "tid": -914061504, "ts": 1716454222534057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222591422, "dur": 6,
    "args": {
      "External id": 78716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78716, "pid": 5, "tid": 7, "ts": 1716454222591422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534086, "dur": 8,
    "args": {
      "External id": 78716,
      "cbid": 211, "correlation": 78716
    }
  },
  {
    "ph": "s", "id": 78716, "pid": 76337, "tid": -914061504, "ts": 1716454222534086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534145, "dur": 0,
    "args": {
      "External id": 78726,
      "cbid": 317, "correlation": 78726
    }
  },
  {
    "ph": "f", "id": 78726, "pid": 76337, "tid": -914061504, "ts": 1716454222534145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534145, "dur": 0,
    "args": {
      "External id": 78727,
      "cbid": 203, "correlation": 78727
    }
  },
  {
    "ph": "f", "id": 78727, "pid": 76337, "tid": -914061504, "ts": 1716454222534145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534146, "dur": 0,
    "args": {
      "External id": 78728,
      "cbid": 205, "correlation": 78728
    }
  },
  {
    "ph": "f", "id": 78728, "pid": 76337, "tid": -914061504, "ts": 1716454222534146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222591429, "dur": 8,
    "args": {
      "External id": 78732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78732, "pid": 5, "tid": 7, "ts": 1716454222591429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534157, "dur": 11,
    "args": {
      "External id": 78732,
      "cbid": 211, "correlation": 78732
    }
  },
  {
    "ph": "s", "id": 78732, "pid": 76337, "tid": -914061504, "ts": 1716454222534157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222591439, "dur": 3,
    "args": {
      "External id": 78734,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78734,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4285714285714286
    }
  },
  {
    "ph": "f", "id": 78734, "pid": 5, "tid": 7, "ts": 1716454222591439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534174, "dur": 10,
    "args": {
      "External id": 78734,
      "cbid": 51, "correlation": 78734
    }
  },
  {
    "ph": "s", "id": 78734, "pid": 76337, "tid": -914061504, "ts": 1716454222534174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222591443, "dur": 96,
    "args": {
      "External id": 78735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78735,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 78735, "pid": 5, "tid": 7, "ts": 1716454222591443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534184, "dur": 6,
    "args": {
      "External id": 78735,
      "cbid": 211, "correlation": 78735
    }
  },
  {
    "ph": "s", "id": 78735, "pid": 76337, "tid": -914061504, "ts": 1716454222534184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222591541, "dur": 6,
    "args": {
      "External id": 78737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78737,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78737, "pid": 5, "tid": 7, "ts": 1716454222591541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534194, "dur": 5,
    "args": {
      "External id": 78737,
      "cbid": 211, "correlation": 78737
    }
  },
  {
    "ph": "s", "id": 78737, "pid": 76337, "tid": -914061504, "ts": 1716454222534194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222591548, "dur": 6,
    "args": {
      "External id": 78743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78743, "pid": 5, "tid": 7, "ts": 1716454222591548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534220, "dur": 9,
    "args": {
      "External id": 78743,
      "cbid": 211, "correlation": 78743
    }
  },
  {
    "ph": "s", "id": 78743, "pid": 76337, "tid": -914061504, "ts": 1716454222534220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222591555, "dur": 5,
    "args": {
      "External id": 78751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78751,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78751, "pid": 5, "tid": 7, "ts": 1716454222591555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534250, "dur": 8,
    "args": {
      "External id": 78751,
      "cbid": 211, "correlation": 78751
    }
  },
  {
    "ph": "s", "id": 78751, "pid": 76337, "tid": -914061504, "ts": 1716454222534250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222591562, "dur": 4,
    "args": {
      "External id": 78759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78759,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 78759, "pid": 5, "tid": 7, "ts": 1716454222591562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534278, "dur": 8,
    "args": {
      "External id": 78759,
      "cbid": 211, "correlation": 78759
    }
  },
  {
    "ph": "s", "id": 78759, "pid": 76337, "tid": -914061504, "ts": 1716454222534278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222591567, "dur": 15,
    "args": {
      "External id": 78770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78770,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78770, "pid": 5, "tid": 7, "ts": 1716454222591567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534358, "dur": 12,
    "args": {
      "External id": 78770,
      "cbid": 211, "correlation": 78770
    }
  },
  {
    "ph": "s", "id": 78770, "pid": 76337, "tid": -914061504, "ts": 1716454222534358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534415, "dur": 0,
    "args": {
      "External id": 78780,
      "cbid": 317, "correlation": 78780
    }
  },
  {
    "ph": "f", "id": 78780, "pid": 76337, "tid": -914061504, "ts": 1716454222534415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534416, "dur": 0,
    "args": {
      "External id": 78781,
      "cbid": 203, "correlation": 78781
    }
  },
  {
    "ph": "f", "id": 78781, "pid": 76337, "tid": -914061504, "ts": 1716454222534416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534416, "dur": 0,
    "args": {
      "External id": 78782,
      "cbid": 205, "correlation": 78782
    }
  },
  {
    "ph": "f", "id": 78782, "pid": 76337, "tid": -914061504, "ts": 1716454222534416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222591583, "dur": 9,
    "args": {
      "External id": 78786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78786,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78786, "pid": 5, "tid": 7, "ts": 1716454222591583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534430, "dur": 11,
    "args": {
      "External id": 78786,
      "cbid": 211, "correlation": 78786
    }
  },
  {
    "ph": "s", "id": 78786, "pid": 76337, "tid": -914061504, "ts": 1716454222534430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222591593, "dur": 165,
    "args": {
      "External id": 78788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78788, "pid": 5, "tid": 7, "ts": 1716454222591593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534444, "dur": 5,
    "args": {
      "External id": 78788,
      "cbid": 211, "correlation": 78788
    }
  },
  {
    "ph": "s", "id": 78788, "pid": 76337, "tid": -914061504, "ts": 1716454222534444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222591761, "dur": 1,
    "args": {
      "External id": 78790,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78790,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5772699939867709
    }
  },
  {
    "ph": "f", "id": 78790, "pid": 5, "tid": 7, "ts": 1716454222591761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534454, "dur": 6,
    "args": {
      "External id": 78790,
      "cbid": 51, "correlation": 78790
    }
  },
  {
    "ph": "s", "id": 78790, "pid": 76337, "tid": -914061504, "ts": 1716454222534454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222591764, "dur": 659,
    "args": {
      "External id": 78791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78791,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78791, "pid": 5, "tid": 7, "ts": 1716454222591764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534461, "dur": 6,
    "args": {
      "External id": 78791,
      "cbid": 211, "correlation": 78791
    }
  },
  {
    "ph": "s", "id": 78791, "pid": 76337, "tid": -914061504, "ts": 1716454222534461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222592425, "dur": 12,
    "args": {
      "External id": 78793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78793,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78793, "pid": 5, "tid": 7, "ts": 1716454222592425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534471, "dur": 7,
    "args": {
      "External id": 78793,
      "cbid": 211, "correlation": 78793
    }
  },
  {
    "ph": "s", "id": 78793, "pid": 76337, "tid": -914061504, "ts": 1716454222534471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222592438, "dur": 15,
    "args": {
      "External id": 78799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78799, "pid": 5, "tid": 7, "ts": 1716454222592438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534501, "dur": 8,
    "args": {
      "External id": 78799,
      "cbid": 211, "correlation": 78799
    }
  },
  {
    "ph": "s", "id": 78799, "pid": 76337, "tid": -914061504, "ts": 1716454222534501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222592454, "dur": 32,
    "args": {
      "External id": 78808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78808,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78808, "pid": 5, "tid": 7, "ts": 1716454222592454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534592, "dur": 12,
    "args": {
      "External id": 78808,
      "cbid": 211, "correlation": 78808
    }
  },
  {
    "ph": "s", "id": 78808, "pid": 76337, "tid": -914061504, "ts": 1716454222534592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222592487, "dur": 31,
    "args": {
      "External id": 78828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78828,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78828, "pid": 5, "tid": 7, "ts": 1716454222592487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534660, "dur": 10,
    "args": {
      "External id": 78828,
      "cbid": 211, "correlation": 78828
    }
  },
  {
    "ph": "s", "id": 78828, "pid": 76337, "tid": -914061504, "ts": 1716454222534660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222592519, "dur": 5,
    "args": {
      "External id": 78840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78840,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78840, "pid": 5, "tid": 7, "ts": 1716454222592519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534679, "dur": 6,
    "args": {
      "External id": 78840,
      "cbid": 211, "correlation": 78840
    }
  },
  {
    "ph": "s", "id": 78840, "pid": 76337, "tid": -914061504, "ts": 1716454222534679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222592525, "dur": 30,
    "args": {
      "External id": 78843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78843,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78843, "pid": 5, "tid": 7, "ts": 1716454222592525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534698, "dur": 7,
    "args": {
      "External id": 78843,
      "cbid": 211, "correlation": 78843
    }
  },
  {
    "ph": "s", "id": 78843, "pid": 76337, "tid": -914061504, "ts": 1716454222534698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222592556, "dur": 20,
    "args": {
      "External id": 78852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78852,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78852, "pid": 5, "tid": 7, "ts": 1716454222592556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534736, "dur": 9,
    "args": {
      "External id": 78852,
      "cbid": 211, "correlation": 78852
    }
  },
  {
    "ph": "s", "id": 78852, "pid": 76337, "tid": -914061504, "ts": 1716454222534736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534788, "dur": 0,
    "args": {
      "External id": 78862,
      "cbid": 317, "correlation": 78862
    }
  },
  {
    "ph": "f", "id": 78862, "pid": 76337, "tid": -914061504, "ts": 1716454222534788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534789, "dur": 0,
    "args": {
      "External id": 78863,
      "cbid": 203, "correlation": 78863
    }
  },
  {
    "ph": "f", "id": 78863, "pid": 76337, "tid": -914061504, "ts": 1716454222534789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534790, "dur": 0,
    "args": {
      "External id": 78864,
      "cbid": 205, "correlation": 78864
    }
  },
  {
    "ph": "f", "id": 78864, "pid": 76337, "tid": -914061504, "ts": 1716454222534790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222592577, "dur": 21,
    "args": {
      "External id": 78868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78868, "pid": 5, "tid": 7, "ts": 1716454222592577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534807, "dur": 11,
    "args": {
      "External id": 78868,
      "cbid": 211, "correlation": 78868
    }
  },
  {
    "ph": "s", "id": 78868, "pid": 76337, "tid": -914061504, "ts": 1716454222534807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222592600, "dur": 325,
    "args": {
      "External id": 78870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78870, "pid": 5, "tid": 7, "ts": 1716454222592600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534820, "dur": 6,
    "args": {
      "External id": 78870,
      "cbid": 211, "correlation": 78870
    }
  },
  {
    "ph": "s", "id": 78870, "pid": 76337, "tid": -914061504, "ts": 1716454222534820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222592927, "dur": 1,
    "args": {
      "External id": 78872,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78872,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 78872, "pid": 5, "tid": 7, "ts": 1716454222592927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534832, "dur": 6,
    "args": {
      "External id": 78872,
      "cbid": 51, "correlation": 78872
    }
  },
  {
    "ph": "s", "id": 78872, "pid": 76337, "tid": -914061504, "ts": 1716454222534832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222592931, "dur": 1261,
    "args": {
      "External id": 78873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78873,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78873, "pid": 5, "tid": 7, "ts": 1716454222592931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534839, "dur": 7,
    "args": {
      "External id": 78873,
      "cbid": 211, "correlation": 78873
    }
  },
  {
    "ph": "s", "id": 78873, "pid": 76337, "tid": -914061504, "ts": 1716454222534839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222594193, "dur": 13,
    "args": {
      "External id": 78875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78875, "pid": 5, "tid": 7, "ts": 1716454222594193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534849, "dur": 5,
    "args": {
      "External id": 78875,
      "cbid": 211, "correlation": 78875
    }
  },
  {
    "ph": "s", "id": 78875, "pid": 76337, "tid": -914061504, "ts": 1716454222534849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222594207, "dur": 16,
    "args": {
      "External id": 78881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78881, "pid": 5, "tid": 7, "ts": 1716454222594207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534877, "dur": 8,
    "args": {
      "External id": 78881,
      "cbid": 211, "correlation": 78881
    }
  },
  {
    "ph": "s", "id": 78881, "pid": 76337, "tid": -914061504, "ts": 1716454222534877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222594224, "dur": 3,
    "args": {
      "External id": 78889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78889,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 78889, "pid": 5, "tid": 7, "ts": 1716454222594224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534922, "dur": 9,
    "args": {
      "External id": 78889,
      "cbid": 211, "correlation": 78889
    }
  },
  {
    "ph": "s", "id": 78889, "pid": 76337, "tid": -914061504, "ts": 1716454222534922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534992, "dur": 2,
    "args": {
      "External id": 78905,
      "cbid": 251, "correlation": 78905
    }
  },
  {
    "ph": "f", "id": 78905, "pid": 76337, "tid": -914061504, "ts": 1716454222534992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222534998, "dur": 0,
    "args": {
      "External id": 78907,
      "cbid": 251, "correlation": 78907
    }
  },
  {
    "ph": "f", "id": 78907, "pid": 76337, "tid": -914061504, "ts": 1716454222534998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222594229, "dur": 12,
    "args": {
      "External id": 78908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78908,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78908, "pid": 5, "tid": 7, "ts": 1716454222594229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535000, "dur": 12,
    "args": {
      "External id": 78908,
      "cbid": 211, "correlation": 78908
    }
  },
  {
    "ph": "s", "id": 78908, "pid": 76337, "tid": -914061504, "ts": 1716454222535000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222594243, "dur": 5,
    "args": {
      "External id": 78910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78910,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78910, "pid": 5, "tid": 7, "ts": 1716454222594243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535014, "dur": 5,
    "args": {
      "External id": 78910,
      "cbid": 211, "correlation": 78910
    }
  },
  {
    "ph": "s", "id": 78910, "pid": 76337, "tid": -914061504, "ts": 1716454222535014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222594249, "dur": 17,
    "args": {
      "External id": 78920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78920,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78920, "pid": 5, "tid": 7, "ts": 1716454222594249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535072, "dur": 12,
    "args": {
      "External id": 78920,
      "cbid": 211, "correlation": 78920
    }
  },
  {
    "ph": "s", "id": 78920, "pid": 76337, "tid": -914061504, "ts": 1716454222535072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222594267, "dur": 18,
    "args": {
      "External id": 78940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78940,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 78940, "pid": 5, "tid": 7, "ts": 1716454222594267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535138, "dur": 11,
    "args": {
      "External id": 78940,
      "cbid": 211, "correlation": 78940
    }
  },
  {
    "ph": "s", "id": 78940, "pid": 76337, "tid": -914061504, "ts": 1716454222535138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222594286, "dur": 4,
    "args": {
      "External id": 78952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78952,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 78952, "pid": 5, "tid": 7, "ts": 1716454222594286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535160, "dur": 6,
    "args": {
      "External id": 78952,
      "cbid": 211, "correlation": 78952
    }
  },
  {
    "ph": "s", "id": 78952, "pid": 76337, "tid": -914061504, "ts": 1716454222535160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222594292, "dur": 17,
    "args": {
      "External id": 78955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78955, "pid": 5, "tid": 7, "ts": 1716454222594292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535178, "dur": 6,
    "args": {
      "External id": 78955,
      "cbid": 211, "correlation": 78955
    }
  },
  {
    "ph": "s", "id": 78955, "pid": 76337, "tid": -914061504, "ts": 1716454222535178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222594310, "dur": 11,
    "args": {
      "External id": 78964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78964,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78964, "pid": 5, "tid": 7, "ts": 1716454222594310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535218, "dur": 9,
    "args": {
      "External id": 78964,
      "cbid": 211, "correlation": 78964
    }
  },
  {
    "ph": "s", "id": 78964, "pid": 76337, "tid": -914061504, "ts": 1716454222535218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535281, "dur": 0,
    "args": {
      "External id": 78974,
      "cbid": 317, "correlation": 78974
    }
  },
  {
    "ph": "f", "id": 78974, "pid": 76337, "tid": -914061504, "ts": 1716454222535281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535281, "dur": 0,
    "args": {
      "External id": 78975,
      "cbid": 203, "correlation": 78975
    }
  },
  {
    "ph": "f", "id": 78975, "pid": 76337, "tid": -914061504, "ts": 1716454222535281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535282, "dur": 0,
    "args": {
      "External id": 78976,
      "cbid": 205, "correlation": 78976
    }
  },
  {
    "ph": "f", "id": 78976, "pid": 76337, "tid": -914061504, "ts": 1716454222535282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222594322, "dur": 12,
    "args": {
      "External id": 78980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78980, "pid": 5, "tid": 7, "ts": 1716454222594322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535296, "dur": 12,
    "args": {
      "External id": 78980,
      "cbid": 211, "correlation": 78980
    }
  },
  {
    "ph": "s", "id": 78980, "pid": 76337, "tid": -914061504, "ts": 1716454222535296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222594335, "dur": 167,
    "args": {
      "External id": 78982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78982, "pid": 5, "tid": 7, "ts": 1716454222594335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535310, "dur": 6,
    "args": {
      "External id": 78982,
      "cbid": 211, "correlation": 78982
    }
  },
  {
    "ph": "s", "id": 78982, "pid": 76337, "tid": -914061504, "ts": 1716454222535310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222594504, "dur": 1,
    "args": {
      "External id": 78984,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 78984,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 78984, "pid": 5, "tid": 7, "ts": 1716454222594504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535322, "dur": 7,
    "args": {
      "External id": 78984,
      "cbid": 51, "correlation": 78984
    }
  },
  {
    "ph": "s", "id": 78984, "pid": 76337, "tid": -914061504, "ts": 1716454222535322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222594508, "dur": 659,
    "args": {
      "External id": 78985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78985,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 78985, "pid": 5, "tid": 7, "ts": 1716454222594508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535330, "dur": 6,
    "args": {
      "External id": 78985,
      "cbid": 211, "correlation": 78985
    }
  },
  {
    "ph": "s", "id": 78985, "pid": 76337, "tid": -914061504, "ts": 1716454222535330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222595169, "dur": 13,
    "args": {
      "External id": 78987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78987, "pid": 5, "tid": 7, "ts": 1716454222595169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535340, "dur": 5,
    "args": {
      "External id": 78987,
      "cbid": 211, "correlation": 78987
    }
  },
  {
    "ph": "s", "id": 78987, "pid": 76337, "tid": -914061504, "ts": 1716454222535340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222595183, "dur": 15,
    "args": {
      "External id": 78993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 78993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 78993, "pid": 5, "tid": 7, "ts": 1716454222595183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535368, "dur": 9,
    "args": {
      "External id": 78993,
      "cbid": 211, "correlation": 78993
    }
  },
  {
    "ph": "s", "id": 78993, "pid": 76337, "tid": -914061504, "ts": 1716454222535368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535428, "dur": 0,
    "args": {
      "External id": 79003,
      "cbid": 317, "correlation": 79003
    }
  },
  {
    "ph": "f", "id": 79003, "pid": 76337, "tid": -914061504, "ts": 1716454222535428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535428, "dur": 0,
    "args": {
      "External id": 79004,
      "cbid": 203, "correlation": 79004
    }
  },
  {
    "ph": "f", "id": 79004, "pid": 76337, "tid": -914061504, "ts": 1716454222535428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535429, "dur": 0,
    "args": {
      "External id": 79005,
      "cbid": 205, "correlation": 79005
    }
  },
  {
    "ph": "f", "id": 79005, "pid": 76337, "tid": -914061504, "ts": 1716454222535429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222595199, "dur": 21,
    "args": {
      "External id": 79009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79009,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79009, "pid": 5, "tid": 7, "ts": 1716454222595199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535442, "dur": 11,
    "args": {
      "External id": 79009,
      "cbid": 211, "correlation": 79009
    }
  },
  {
    "ph": "s", "id": 79009, "pid": 76337, "tid": -914061504, "ts": 1716454222535442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222595222, "dur": 4,
    "args": {
      "External id": 79011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79011, "pid": 5, "tid": 7, "ts": 1716454222595222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535459, "dur": 6,
    "args": {
      "External id": 79011,
      "cbid": 211, "correlation": 79011
    }
  },
  {
    "ph": "s", "id": 79011, "pid": 76337, "tid": -914061504, "ts": 1716454222535459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535468, "dur": 0,
    "args": {
      "External id": 79012,
      "cbid": 51, "correlation": 79012
    }
  },
  {
    "ph": "s", "id": 79012, "pid": 76337, "tid": -914061504, "ts": 1716454222535468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222595227, "dur": 179,
    "args": {
      "External id": 79013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79013,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 79013, "pid": 5, "tid": 7, "ts": 1716454222595227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535469, "dur": 6,
    "args": {
      "External id": 79013,
      "cbid": 211, "correlation": 79013
    }
  },
  {
    "ph": "s", "id": 79013, "pid": 76337, "tid": -914061504, "ts": 1716454222535469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222595407, "dur": 16,
    "args": {
      "External id": 79018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79018, "pid": 5, "tid": 7, "ts": 1716454222595407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535496, "dur": 8,
    "args": {
      "External id": 79018,
      "cbid": 211, "correlation": 79018
    }
  },
  {
    "ph": "s", "id": 79018, "pid": 76337, "tid": -914061504, "ts": 1716454222535496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222595425, "dur": 12,
    "args": {
      "External id": 79026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79026,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79026, "pid": 5, "tid": 7, "ts": 1716454222595425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535524, "dur": 8,
    "args": {
      "External id": 79026,
      "cbid": 211, "correlation": 79026
    }
  },
  {
    "ph": "s", "id": 79026, "pid": 76337, "tid": -914061504, "ts": 1716454222535524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222595438, "dur": 10,
    "args": {
      "External id": 79034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79034,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79034, "pid": 5, "tid": 7, "ts": 1716454222595438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535554, "dur": 9,
    "args": {
      "External id": 79034,
      "cbid": 211, "correlation": 79034
    }
  },
  {
    "ph": "s", "id": 79034, "pid": 76337, "tid": -914061504, "ts": 1716454222535554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222595450, "dur": 19,
    "args": {
      "External id": 79054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79054,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 79054, "pid": 5, "tid": 7, "ts": 1716454222595450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535638, "dur": 12,
    "args": {
      "External id": 79054,
      "cbid": 211, "correlation": 79054
    }
  },
  {
    "ph": "s", "id": 79054, "pid": 76337, "tid": -914061504, "ts": 1716454222535638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222595470, "dur": 5,
    "args": {
      "External id": 79066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79066,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 79066, "pid": 5, "tid": 7, "ts": 1716454222595470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535660, "dur": 6,
    "args": {
      "External id": 79066,
      "cbid": 211, "correlation": 79066
    }
  },
  {
    "ph": "s", "id": 79066, "pid": 76337, "tid": -914061504, "ts": 1716454222535660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222595476, "dur": 17,
    "args": {
      "External id": 79069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79069, "pid": 5, "tid": 7, "ts": 1716454222595476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535677, "dur": 6,
    "args": {
      "External id": 79069,
      "cbid": 211, "correlation": 79069
    }
  },
  {
    "ph": "s", "id": 79069, "pid": 76337, "tid": -914061504, "ts": 1716454222535677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535734, "dur": 0,
    "args": {
      "External id": 79080,
      "cbid": 317, "correlation": 79080
    }
  },
  {
    "ph": "f", "id": 79080, "pid": 76337, "tid": -914061504, "ts": 1716454222535734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535735, "dur": 0,
    "args": {
      "External id": 79081,
      "cbid": 203, "correlation": 79081
    }
  },
  {
    "ph": "f", "id": 79081, "pid": 76337, "tid": -914061504, "ts": 1716454222535735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535735, "dur": 0,
    "args": {
      "External id": 79082,
      "cbid": 205, "correlation": 79082
    }
  },
  {
    "ph": "f", "id": 79082, "pid": 76337, "tid": -914061504, "ts": 1716454222535735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222595494, "dur": 12,
    "args": {
      "External id": 79086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79086,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79086, "pid": 5, "tid": 7, "ts": 1716454222595494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535749, "dur": 12,
    "args": {
      "External id": 79086,
      "cbid": 211, "correlation": 79086
    }
  },
  {
    "ph": "s", "id": 79086, "pid": 76337, "tid": -914061504, "ts": 1716454222535749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222595507, "dur": 3,
    "args": {
      "External id": 79088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79088,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79088, "pid": 5, "tid": 7, "ts": 1716454222595507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535765, "dur": 6,
    "args": {
      "External id": 79088,
      "cbid": 211, "correlation": 79088
    }
  },
  {
    "ph": "s", "id": 79088, "pid": 76337, "tid": -914061504, "ts": 1716454222535765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535774, "dur": 0,
    "args": {
      "External id": 79089,
      "cbid": 51, "correlation": 79089
    }
  },
  {
    "ph": "s", "id": 79089, "pid": 76337, "tid": -914061504, "ts": 1716454222535774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222595512, "dur": 93,
    "args": {
      "External id": 79090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79090,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 79090, "pid": 5, "tid": 7, "ts": 1716454222595512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535775, "dur": 5,
    "args": {
      "External id": 79090,
      "cbid": 211, "correlation": 79090
    }
  },
  {
    "ph": "s", "id": 79090, "pid": 76337, "tid": -914061504, "ts": 1716454222535775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222595606, "dur": 16,
    "args": {
      "External id": 79095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79095,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79095, "pid": 5, "tid": 7, "ts": 1716454222595606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535802, "dur": 9,
    "args": {
      "External id": 79095,
      "cbid": 211, "correlation": 79095
    }
  },
  {
    "ph": "s", "id": 79095, "pid": 76337, "tid": -914061504, "ts": 1716454222535802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222595623, "dur": 85,
    "args": {
      "External id": 79104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79104,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79104, "pid": 5, "tid": 7, "ts": 1716454222595623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535884, "dur": 16,
    "args": {
      "External id": 79104,
      "cbid": 211, "correlation": 79104
    }
  },
  {
    "ph": "s", "id": 79104, "pid": 76337, "tid": -914061504, "ts": 1716454222535884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222595710, "dur": 29,
    "args": {
      "External id": 79126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79126,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79126, "pid": 5, "tid": 7, "ts": 1716454222595710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222535945, "dur": 10,
    "args": {
      "External id": 79126,
      "cbid": 211, "correlation": 79126
    }
  },
  {
    "ph": "s", "id": 79126, "pid": 76337, "tid": -914061504, "ts": 1716454222535945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536048, "dur": 2,
    "args": {
      "External id": 79137,
      "cbid": 251, "correlation": 79137
    }
  },
  {
    "ph": "f", "id": 79137, "pid": 76337, "tid": -914061504, "ts": 1716454222536048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222595740, "dur": 167,
    "args": {
      "External id": 79138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79138,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79138, "pid": 5, "tid": 7, "ts": 1716454222595740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536054, "dur": 13,
    "args": {
      "External id": 79138,
      "cbid": 211, "correlation": 79138
    }
  },
  {
    "ph": "s", "id": 79138, "pid": 76337, "tid": -914061504, "ts": 1716454222536054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536124, "dur": 1,
    "args": {
      "External id": 79149,
      "cbid": 251, "correlation": 79149
    }
  },
  {
    "ph": "f", "id": 79149, "pid": 76337, "tid": -914061504, "ts": 1716454222536124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222595909, "dur": 162,
    "args": {
      "External id": 79150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79150,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79150, "pid": 5, "tid": 7, "ts": 1716454222595909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536128, "dur": 11,
    "args": {
      "External id": 79150,
      "cbid": 211, "correlation": 79150
    }
  },
  {
    "ph": "s", "id": 79150, "pid": 76337, "tid": -914061504, "ts": 1716454222536128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536192, "dur": 1,
    "args": {
      "External id": 79161,
      "cbid": 251, "correlation": 79161
    }
  },
  {
    "ph": "f", "id": 79161, "pid": 76337, "tid": -914061504, "ts": 1716454222536192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222596072, "dur": 161,
    "args": {
      "External id": 79162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79162,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79162, "pid": 5, "tid": 7, "ts": 1716454222596072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536196, "dur": 11,
    "args": {
      "External id": 79162,
      "cbid": 211, "correlation": 79162
    }
  },
  {
    "ph": "s", "id": 79162, "pid": 76337, "tid": -914061504, "ts": 1716454222536196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222596234, "dur": 343,
    "args": {
      "External id": 79187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79187,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79187, "pid": 5, "tid": 7, "ts": 1716454222596234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536283, "dur": 12,
    "args": {
      "External id": 79187,
      "cbid": 211, "correlation": 79187
    }
  },
  {
    "ph": "s", "id": 79187, "pid": 76337, "tid": -914061504, "ts": 1716454222536283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536385, "dur": 1,
    "args": {
      "External id": 79205,
      "cbid": 251, "correlation": 79205
    }
  },
  {
    "ph": "f", "id": 79205, "pid": 76337, "tid": -914061504, "ts": 1716454222536385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222596578, "dur": 170,
    "args": {
      "External id": 79207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79207,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79207, "pid": 5, "tid": 7, "ts": 1716454222596578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536391, "dur": 14,
    "args": {
      "External id": 79207,
      "cbid": 211, "correlation": 79207
    }
  },
  {
    "ph": "s", "id": 79207, "pid": 76337, "tid": -914061504, "ts": 1716454222536391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222596749, "dur": 19,
    "args": {
      "External id": 79215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79215,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79215, "pid": 5, "tid": 7, "ts": 1716454222596749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536461, "dur": 12,
    "args": {
      "External id": 79215,
      "cbid": 211, "correlation": 79215
    }
  },
  {
    "ph": "s", "id": 79215, "pid": 76337, "tid": -914061504, "ts": 1716454222536461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222596770, "dur": 28,
    "args": {
      "External id": 79223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79223,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79223, "pid": 5, "tid": 7, "ts": 1716454222596770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536502, "dur": 9,
    "args": {
      "External id": 79223,
      "cbid": 211, "correlation": 79223
    }
  },
  {
    "ph": "s", "id": 79223, "pid": 76337, "tid": -914061504, "ts": 1716454222536502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222596799, "dur": 19,
    "args": {
      "External id": 79234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79234, "pid": 5, "tid": 7, "ts": 1716454222596799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536576, "dur": 12,
    "args": {
      "External id": 79234,
      "cbid": 211, "correlation": 79234
    }
  },
  {
    "ph": "s", "id": 79234, "pid": 76337, "tid": -914061504, "ts": 1716454222536576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222596819, "dur": 17,
    "args": {
      "External id": 79256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79256,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79256, "pid": 5, "tid": 7, "ts": 1716454222596819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536607, "dur": 7,
    "args": {
      "External id": 79256,
      "cbid": 211, "correlation": 79256
    }
  },
  {
    "ph": "s", "id": 79256, "pid": 76337, "tid": -914061504, "ts": 1716454222536607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536692, "dur": 2,
    "args": {
      "External id": 79267,
      "cbid": 251, "correlation": 79267
    }
  },
  {
    "ph": "f", "id": 79267, "pid": 76337, "tid": -914061504, "ts": 1716454222536692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222596837, "dur": 91,
    "args": {
      "External id": 79268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79268,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79268, "pid": 5, "tid": 7, "ts": 1716454222596837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536698, "dur": 14,
    "args": {
      "External id": 79268,
      "cbid": 211, "correlation": 79268
    }
  },
  {
    "ph": "s", "id": 79268, "pid": 76337, "tid": -914061504, "ts": 1716454222536698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536768, "dur": 1,
    "args": {
      "External id": 79279,
      "cbid": 251, "correlation": 79279
    }
  },
  {
    "ph": "f", "id": 79279, "pid": 76337, "tid": -914061504, "ts": 1716454222536768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536772, "dur": 0,
    "args": {
      "External id": 79280,
      "cbid": 251, "correlation": 79280
    }
  },
  {
    "ph": "f", "id": 79280, "pid": 76337, "tid": -914061504, "ts": 1716454222536772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222596929, "dur": 13,
    "args": {
      "External id": 79281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79281,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79281, "pid": 5, "tid": 7, "ts": 1716454222596929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536774, "dur": 11,
    "args": {
      "External id": 79281,
      "cbid": 211, "correlation": 79281
    }
  },
  {
    "ph": "s", "id": 79281, "pid": 76337, "tid": -914061504, "ts": 1716454222536774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222596943, "dur": 6,
    "args": {
      "External id": 79283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79283,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79283, "pid": 5, "tid": 7, "ts": 1716454222596943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536787, "dur": 5,
    "args": {
      "External id": 79283,
      "cbid": 211, "correlation": 79283
    }
  },
  {
    "ph": "s", "id": 79283, "pid": 76337, "tid": -914061504, "ts": 1716454222536787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536844, "dur": 1,
    "args": {
      "External id": 79294,
      "cbid": 251, "correlation": 79294
    }
  },
  {
    "ph": "f", "id": 79294, "pid": 76337, "tid": -914061504, "ts": 1716454222536844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536847, "dur": 0,
    "args": {
      "External id": 79295,
      "cbid": 251, "correlation": 79295
    }
  },
  {
    "ph": "f", "id": 79295, "pid": 76337, "tid": -914061504, "ts": 1716454222536847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222596950, "dur": 8,
    "args": {
      "External id": 79296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79296,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79296, "pid": 5, "tid": 7, "ts": 1716454222596950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536848, "dur": 12,
    "args": {
      "External id": 79296,
      "cbid": 211, "correlation": 79296
    }
  },
  {
    "ph": "s", "id": 79296, "pid": 76337, "tid": -914061504, "ts": 1716454222536848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222596960, "dur": 3,
    "args": {
      "External id": 79298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79298,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79298, "pid": 5, "tid": 7, "ts": 1716454222596960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536861, "dur": 5,
    "args": {
      "External id": 79298,
      "cbid": 211, "correlation": 79298
    }
  },
  {
    "ph": "s", "id": 79298, "pid": 76337, "tid": -914061504, "ts": 1716454222536861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222596965, "dur": 56,
    "args": {
      "External id": 79323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79323,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79323, "pid": 5, "tid": 7, "ts": 1716454222596965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222536939, "dur": 12,
    "args": {
      "External id": 79323,
      "cbid": 211, "correlation": 79323
    }
  },
  {
    "ph": "s", "id": 79323, "pid": 76337, "tid": -914061504, "ts": 1716454222536939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537046, "dur": 2,
    "args": {
      "External id": 79341,
      "cbid": 251, "correlation": 79341
    }
  },
  {
    "ph": "f", "id": 79341, "pid": 76337, "tid": -914061504, "ts": 1716454222537046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222597022, "dur": 93,
    "args": {
      "External id": 79343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79343,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79343, "pid": 5, "tid": 7, "ts": 1716454222597022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537052, "dur": 14,
    "args": {
      "External id": 79343,
      "cbid": 211, "correlation": 79343
    }
  },
  {
    "ph": "s", "id": 79343, "pid": 76337, "tid": -914061504, "ts": 1716454222537052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222597116, "dur": 9,
    "args": {
      "External id": 79351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79351,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79351, "pid": 5, "tid": 7, "ts": 1716454222597116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537122, "dur": 13,
    "args": {
      "External id": 79351,
      "cbid": 211, "correlation": 79351
    }
  },
  {
    "ph": "s", "id": 79351, "pid": 76337, "tid": -914061504, "ts": 1716454222537122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222597127, "dur": 22,
    "args": {
      "External id": 79359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79359, "pid": 5, "tid": 7, "ts": 1716454222597127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537164, "dur": 10,
    "args": {
      "External id": 79359,
      "cbid": 211, "correlation": 79359
    }
  },
  {
    "ph": "s", "id": 79359, "pid": 76337, "tid": -914061504, "ts": 1716454222537164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222597150, "dur": 17,
    "args": {
      "External id": 79381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79381,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79381, "pid": 5, "tid": 7, "ts": 1716454222597150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537217, "dur": 10,
    "args": {
      "External id": 79381,
      "cbid": 211, "correlation": 79381
    }
  },
  {
    "ph": "s", "id": 79381, "pid": 76337, "tid": -914061504, "ts": 1716454222537217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537306, "dur": 1,
    "args": {
      "External id": 79397,
      "cbid": 251, "correlation": 79397
    }
  },
  {
    "ph": "f", "id": 79397, "pid": 76337, "tid": -914061504, "ts": 1716454222537306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537311, "dur": 0,
    "args": {
      "External id": 79399,
      "cbid": 251, "correlation": 79399
    }
  },
  {
    "ph": "f", "id": 79399, "pid": 76337, "tid": -914061504, "ts": 1716454222537311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222597169, "dur": 498,
    "args": {
      "External id": 79400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79400,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79400, "pid": 5, "tid": 7, "ts": 1716454222597169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537313, "dur": 13,
    "args": {
      "External id": 79400,
      "cbid": 211, "correlation": 79400
    }
  },
  {
    "ph": "s", "id": 79400, "pid": 76337, "tid": -914061504, "ts": 1716454222537313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222597668, "dur": 66,
    "args": {
      "External id": 79408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79408, "pid": 5, "tid": 7, "ts": 1716454222597668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537379, "dur": 13,
    "args": {
      "External id": 79408,
      "cbid": 211, "correlation": 79408
    }
  },
  {
    "ph": "s", "id": 79408, "pid": 76337, "tid": -914061504, "ts": 1716454222537379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222597736, "dur": 66,
    "args": {
      "External id": 79416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79416, "pid": 5, "tid": 7, "ts": 1716454222597736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537411, "dur": 8,
    "args": {
      "External id": 79416,
      "cbid": 211, "correlation": 79416
    }
  },
  {
    "ph": "s", "id": 79416, "pid": 76337, "tid": -914061504, "ts": 1716454222537411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537493, "dur": 1,
    "args": {
      "External id": 79432,
      "cbid": 251, "correlation": 79432
    }
  },
  {
    "ph": "f", "id": 79432, "pid": 76337, "tid": -914061504, "ts": 1716454222537493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222597805, "dur": 1,
    "args": {
      "External id": 79434,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 79434,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 79434, "pid": 5, "tid": 7, "ts": 1716454222597805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537498, "dur": 10,
    "args": {
      "External id": 79434,
      "cbid": 51, "correlation": 79434
    }
  },
  {
    "ph": "s", "id": 79434, "pid": 76337, "tid": -914061504, "ts": 1716454222537498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222597808, "dur": 276,
    "args": {
      "External id": 79435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79435,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79435, "pid": 5, "tid": 7, "ts": 1716454222597808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537510, "dur": 11,
    "args": {
      "External id": 79435,
      "cbid": 211, "correlation": 79435
    }
  },
  {
    "ph": "s", "id": 79435, "pid": 76337, "tid": -914061504, "ts": 1716454222537510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222598086, "dur": 14,
    "args": {
      "External id": 79443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79443,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79443, "pid": 5, "tid": 7, "ts": 1716454222598086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537553, "dur": 10,
    "args": {
      "External id": 79443,
      "cbid": 211, "correlation": 79443
    }
  },
  {
    "ph": "s", "id": 79443, "pid": 76337, "tid": -914061504, "ts": 1716454222537553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222598101, "dur": 38,
    "args": {
      "External id": 79454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79454, "pid": 5, "tid": 7, "ts": 1716454222598101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537622, "dur": 13,
    "args": {
      "External id": 79454,
      "cbid": 211, "correlation": 79454
    }
  },
  {
    "ph": "s", "id": 79454, "pid": 76337, "tid": -914061504, "ts": 1716454222537622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537688, "dur": 0,
    "args": {
      "External id": 79466,
      "cbid": 317, "correlation": 79466
    }
  },
  {
    "ph": "f", "id": 79466, "pid": 76337, "tid": -914061504, "ts": 1716454222537688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537689, "dur": 0,
    "args": {
      "External id": 79467,
      "cbid": 203, "correlation": 79467
    }
  },
  {
    "ph": "f", "id": 79467, "pid": 76337, "tid": -914061504, "ts": 1716454222537689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537690, "dur": 0,
    "args": {
      "External id": 79468,
      "cbid": 205, "correlation": 79468
    }
  },
  {
    "ph": "f", "id": 79468, "pid": 76337, "tid": -914061504, "ts": 1716454222537690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222598140, "dur": 13,
    "args": {
      "External id": 79472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79472,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79472, "pid": 5, "tid": 7, "ts": 1716454222598140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537707, "dur": 12,
    "args": {
      "External id": 79472,
      "cbid": 211, "correlation": 79472
    }
  },
  {
    "ph": "s", "id": 79472, "pid": 76337, "tid": -914061504, "ts": 1716454222537707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222598154, "dur": 4,
    "args": {
      "External id": 79474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79474, "pid": 5, "tid": 7, "ts": 1716454222598154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537723, "dur": 6,
    "args": {
      "External id": 79474,
      "cbid": 211, "correlation": 79474
    }
  },
  {
    "ph": "s", "id": 79474, "pid": 76337, "tid": -914061504, "ts": 1716454222537723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537732, "dur": 0,
    "args": {
      "External id": 79475,
      "cbid": 51, "correlation": 79475
    }
  },
  {
    "ph": "s", "id": 79475, "pid": 76337, "tid": -914061504, "ts": 1716454222537732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222598159, "dur": 99,
    "args": {
      "External id": 79476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79476,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 79476, "pid": 5, "tid": 7, "ts": 1716454222598159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537733, "dur": 5,
    "args": {
      "External id": 79476,
      "cbid": 211, "correlation": 79476
    }
  },
  {
    "ph": "s", "id": 79476, "pid": 76337, "tid": -914061504, "ts": 1716454222537733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222598260, "dur": 17,
    "args": {
      "External id": 79481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79481,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79481, "pid": 5, "tid": 7, "ts": 1716454222598260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537761, "dur": 9,
    "args": {
      "External id": 79481,
      "cbid": 211, "correlation": 79481
    }
  },
  {
    "ph": "s", "id": 79481, "pid": 76337, "tid": -914061504, "ts": 1716454222537761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222598278, "dur": 12,
    "args": {
      "External id": 79489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79489,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79489, "pid": 5, "tid": 7, "ts": 1716454222598278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537793, "dur": 9,
    "args": {
      "External id": 79489,
      "cbid": 211, "correlation": 79489
    }
  },
  {
    "ph": "s", "id": 79489, "pid": 76337, "tid": -914061504, "ts": 1716454222537793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222598291, "dur": 32,
    "args": {
      "External id": 79498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79498,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79498, "pid": 5, "tid": 7, "ts": 1716454222598291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537833, "dur": 10,
    "args": {
      "External id": 79498,
      "cbid": 211, "correlation": 79498
    }
  },
  {
    "ph": "s", "id": 79498, "pid": 76337, "tid": -914061504, "ts": 1716454222537833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222598324, "dur": 31,
    "args": {
      "External id": 79518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79518,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 79518, "pid": 5, "tid": 7, "ts": 1716454222598324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537903, "dur": 12,
    "args": {
      "External id": 79518,
      "cbid": 211, "correlation": 79518
    }
  },
  {
    "ph": "s", "id": 79518, "pid": 76337, "tid": -914061504, "ts": 1716454222537903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222598356, "dur": 5,
    "args": {
      "External id": 79530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79530,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79530, "pid": 5, "tid": 7, "ts": 1716454222598356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537925, "dur": 6,
    "args": {
      "External id": 79530,
      "cbid": 211, "correlation": 79530
    }
  },
  {
    "ph": "s", "id": 79530, "pid": 76337, "tid": -914061504, "ts": 1716454222537925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222598363, "dur": 32,
    "args": {
      "External id": 79533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79533, "pid": 5, "tid": 7, "ts": 1716454222598363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537943, "dur": 6,
    "args": {
      "External id": 79533,
      "cbid": 211, "correlation": 79533
    }
  },
  {
    "ph": "s", "id": 79533, "pid": 76337, "tid": -914061504, "ts": 1716454222537943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222598396, "dur": 21,
    "args": {
      "External id": 79542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79542,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79542, "pid": 5, "tid": 7, "ts": 1716454222598396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222537989, "dur": 11,
    "args": {
      "External id": 79542,
      "cbid": 211, "correlation": 79542
    }
  },
  {
    "ph": "s", "id": 79542, "pid": 76337, "tid": -914061504, "ts": 1716454222537989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538042, "dur": 0,
    "args": {
      "External id": 79552,
      "cbid": 317, "correlation": 79552
    }
  },
  {
    "ph": "f", "id": 79552, "pid": 76337, "tid": -914061504, "ts": 1716454222538042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538043, "dur": 0,
    "args": {
      "External id": 79553,
      "cbid": 203, "correlation": 79553
    }
  },
  {
    "ph": "f", "id": 79553, "pid": 76337, "tid": -914061504, "ts": 1716454222538043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538044, "dur": 0,
    "args": {
      "External id": 79554,
      "cbid": 205, "correlation": 79554
    }
  },
  {
    "ph": "f", "id": 79554, "pid": 76337, "tid": -914061504, "ts": 1716454222538044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222598418, "dur": 22,
    "args": {
      "External id": 79558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79558, "pid": 5, "tid": 7, "ts": 1716454222598418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538058, "dur": 12,
    "args": {
      "External id": 79558,
      "cbid": 211, "correlation": 79558
    }
  },
  {
    "ph": "s", "id": 79558, "pid": 76337, "tid": -914061504, "ts": 1716454222538058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222598441, "dur": 325,
    "args": {
      "External id": 79560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79560,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79560, "pid": 5, "tid": 7, "ts": 1716454222598441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538072, "dur": 5,
    "args": {
      "External id": 79560,
      "cbid": 211, "correlation": 79560
    }
  },
  {
    "ph": "s", "id": 79560, "pid": 76337, "tid": -914061504, "ts": 1716454222538072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222598768, "dur": 1,
    "args": {
      "External id": 79562,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 79562,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 79562, "pid": 5, "tid": 7, "ts": 1716454222598768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538084, "dur": 6,
    "args": {
      "External id": 79562,
      "cbid": 51, "correlation": 79562
    }
  },
  {
    "ph": "s", "id": 79562, "pid": 76337, "tid": -914061504, "ts": 1716454222538084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222598772, "dur": 1279,
    "args": {
      "External id": 79563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79563,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79563, "pid": 5, "tid": 7, "ts": 1716454222598772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538091, "dur": 6,
    "args": {
      "External id": 79563,
      "cbid": 211, "correlation": 79563
    }
  },
  {
    "ph": "s", "id": 79563, "pid": 76337, "tid": -914061504, "ts": 1716454222538091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222600052, "dur": 13,
    "args": {
      "External id": 79565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79565,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79565, "pid": 5, "tid": 7, "ts": 1716454222600052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538101, "dur": 5,
    "args": {
      "External id": 79565,
      "cbid": 211, "correlation": 79565
    }
  },
  {
    "ph": "s", "id": 79565, "pid": 76337, "tid": -914061504, "ts": 1716454222538101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222600066, "dur": 15,
    "args": {
      "External id": 79571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79571,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79571, "pid": 5, "tid": 7, "ts": 1716454222600066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538130, "dur": 8,
    "args": {
      "External id": 79571,
      "cbid": 211, "correlation": 79571
    }
  },
  {
    "ph": "s", "id": 79571, "pid": 76337, "tid": -914061504, "ts": 1716454222538130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222600083, "dur": 4,
    "args": {
      "External id": 79579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79579,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 79579, "pid": 5, "tid": 7, "ts": 1716454222600083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538174, "dur": 10,
    "args": {
      "External id": 79579,
      "cbid": 211, "correlation": 79579
    }
  },
  {
    "ph": "s", "id": 79579, "pid": 76337, "tid": -914061504, "ts": 1716454222538174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538240, "dur": 1,
    "args": {
      "External id": 79595,
      "cbid": 251, "correlation": 79595
    }
  },
  {
    "ph": "f", "id": 79595, "pid": 76337, "tid": -914061504, "ts": 1716454222538240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538245, "dur": 0,
    "args": {
      "External id": 79597,
      "cbid": 251, "correlation": 79597
    }
  },
  {
    "ph": "f", "id": 79597, "pid": 76337, "tid": -914061504, "ts": 1716454222538245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222600088, "dur": 14,
    "args": {
      "External id": 79598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79598,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79598, "pid": 5, "tid": 7, "ts": 1716454222600088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538247, "dur": 11,
    "args": {
      "External id": 79598,
      "cbid": 211, "correlation": 79598
    }
  },
  {
    "ph": "s", "id": 79598, "pid": 76337, "tid": -914061504, "ts": 1716454222538247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222600103, "dur": 5,
    "args": {
      "External id": 79600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79600,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79600, "pid": 5, "tid": 7, "ts": 1716454222600103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538260, "dur": 5,
    "args": {
      "External id": 79600,
      "cbid": 211, "correlation": 79600
    }
  },
  {
    "ph": "s", "id": 79600, "pid": 76337, "tid": -914061504, "ts": 1716454222538260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222600109, "dur": 17,
    "args": {
      "External id": 79610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79610,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79610, "pid": 5, "tid": 7, "ts": 1716454222600109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538318, "dur": 13,
    "args": {
      "External id": 79610,
      "cbid": 211, "correlation": 79610
    }
  },
  {
    "ph": "s", "id": 79610, "pid": 76337, "tid": -914061504, "ts": 1716454222538318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222600127, "dur": 18,
    "args": {
      "External id": 79630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79630,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 79630, "pid": 5, "tid": 7, "ts": 1716454222600127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538385, "dur": 11,
    "args": {
      "External id": 79630,
      "cbid": 211, "correlation": 79630
    }
  },
  {
    "ph": "s", "id": 79630, "pid": 76337, "tid": -914061504, "ts": 1716454222538385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222600147, "dur": 4,
    "args": {
      "External id": 79642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79642,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 79642, "pid": 5, "tid": 7, "ts": 1716454222600147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538406, "dur": 6,
    "args": {
      "External id": 79642,
      "cbid": 211, "correlation": 79642
    }
  },
  {
    "ph": "s", "id": 79642, "pid": 76337, "tid": -914061504, "ts": 1716454222538406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222600152, "dur": 17,
    "args": {
      "External id": 79645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79645,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79645, "pid": 5, "tid": 7, "ts": 1716454222600152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538424, "dur": 7,
    "args": {
      "External id": 79645,
      "cbid": 211, "correlation": 79645
    }
  },
  {
    "ph": "s", "id": 79645, "pid": 76337, "tid": -914061504, "ts": 1716454222538424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222600170, "dur": 11,
    "args": {
      "External id": 79654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79654,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79654, "pid": 5, "tid": 7, "ts": 1716454222600170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538464, "dur": 10,
    "args": {
      "External id": 79654,
      "cbid": 211, "correlation": 79654
    }
  },
  {
    "ph": "s", "id": 79654, "pid": 76337, "tid": -914061504, "ts": 1716454222538464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538526, "dur": 0,
    "args": {
      "External id": 79664,
      "cbid": 317, "correlation": 79664
    }
  },
  {
    "ph": "f", "id": 79664, "pid": 76337, "tid": -914061504, "ts": 1716454222538526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538527, "dur": 0,
    "args": {
      "External id": 79665,
      "cbid": 203, "correlation": 79665
    }
  },
  {
    "ph": "f", "id": 79665, "pid": 76337, "tid": -914061504, "ts": 1716454222538527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538528, "dur": 0,
    "args": {
      "External id": 79666,
      "cbid": 205, "correlation": 79666
    }
  },
  {
    "ph": "f", "id": 79666, "pid": 76337, "tid": -914061504, "ts": 1716454222538528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222600183, "dur": 11,
    "args": {
      "External id": 79670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79670, "pid": 5, "tid": 7, "ts": 1716454222600183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538542, "dur": 12,
    "args": {
      "External id": 79670,
      "cbid": 211, "correlation": 79670
    }
  },
  {
    "ph": "s", "id": 79670, "pid": 76337, "tid": -914061504, "ts": 1716454222538542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222600195, "dur": 166,
    "args": {
      "External id": 79672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79672, "pid": 5, "tid": 7, "ts": 1716454222600195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538556, "dur": 5,
    "args": {
      "External id": 79672,
      "cbid": 211, "correlation": 79672
    }
  },
  {
    "ph": "s", "id": 79672, "pid": 76337, "tid": -914061504, "ts": 1716454222538556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222600363, "dur": 1,
    "args": {
      "External id": 79674,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 79674,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 79674, "pid": 5, "tid": 7, "ts": 1716454222600363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538567, "dur": 6,
    "args": {
      "External id": 79674,
      "cbid": 51, "correlation": 79674
    }
  },
  {
    "ph": "s", "id": 79674, "pid": 76337, "tid": -914061504, "ts": 1716454222538567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222600367, "dur": 661,
    "args": {
      "External id": 79675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79675,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79675, "pid": 5, "tid": 7, "ts": 1716454222600367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538575, "dur": 7,
    "args": {
      "External id": 79675,
      "cbid": 211, "correlation": 79675
    }
  },
  {
    "ph": "s", "id": 79675, "pid": 76337, "tid": -914061504, "ts": 1716454222538575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222601029, "dur": 13,
    "args": {
      "External id": 79677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79677,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79677, "pid": 5, "tid": 7, "ts": 1716454222601029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538586, "dur": 5,
    "args": {
      "External id": 79677,
      "cbid": 211, "correlation": 79677
    }
  },
  {
    "ph": "s", "id": 79677, "pid": 76337, "tid": -914061504, "ts": 1716454222538586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222601043, "dur": 15,
    "args": {
      "External id": 79683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79683, "pid": 5, "tid": 7, "ts": 1716454222601043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538614, "dur": 8,
    "args": {
      "External id": 79683,
      "cbid": 211, "correlation": 79683
    }
  },
  {
    "ph": "s", "id": 79683, "pid": 76337, "tid": -914061504, "ts": 1716454222538614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538672, "dur": 0,
    "args": {
      "External id": 79693,
      "cbid": 317, "correlation": 79693
    }
  },
  {
    "ph": "f", "id": 79693, "pid": 76337, "tid": -914061504, "ts": 1716454222538672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538673, "dur": 0,
    "args": {
      "External id": 79694,
      "cbid": 203, "correlation": 79694
    }
  },
  {
    "ph": "f", "id": 79694, "pid": 76337, "tid": -914061504, "ts": 1716454222538673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538673, "dur": 0,
    "args": {
      "External id": 79695,
      "cbid": 205, "correlation": 79695
    }
  },
  {
    "ph": "f", "id": 79695, "pid": 76337, "tid": -914061504, "ts": 1716454222538673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222601060, "dur": 21,
    "args": {
      "External id": 79699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79699, "pid": 5, "tid": 7, "ts": 1716454222601060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538686, "dur": 12,
    "args": {
      "External id": 79699,
      "cbid": 211, "correlation": 79699
    }
  },
  {
    "ph": "s", "id": 79699, "pid": 76337, "tid": -914061504, "ts": 1716454222538686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222601082, "dur": 4,
    "args": {
      "External id": 79701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79701, "pid": 5, "tid": 7, "ts": 1716454222601082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538702, "dur": 6,
    "args": {
      "External id": 79701,
      "cbid": 211, "correlation": 79701
    }
  },
  {
    "ph": "s", "id": 79701, "pid": 76337, "tid": -914061504, "ts": 1716454222538702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538711, "dur": 0,
    "args": {
      "External id": 79702,
      "cbid": 51, "correlation": 79702
    }
  },
  {
    "ph": "s", "id": 79702, "pid": 76337, "tid": -914061504, "ts": 1716454222538711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222601087, "dur": 174,
    "args": {
      "External id": 79703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79703,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 79703, "pid": 5, "tid": 7, "ts": 1716454222601087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538711, "dur": 5,
    "args": {
      "External id": 79703,
      "cbid": 211, "correlation": 79703
    }
  },
  {
    "ph": "s", "id": 79703, "pid": 76337, "tid": -914061504, "ts": 1716454222538711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222601263, "dur": 17,
    "args": {
      "External id": 79708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79708, "pid": 5, "tid": 7, "ts": 1716454222601263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538736, "dur": 9,
    "args": {
      "External id": 79708,
      "cbid": 211, "correlation": 79708
    }
  },
  {
    "ph": "s", "id": 79708, "pid": 76337, "tid": -914061504, "ts": 1716454222538736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222601281, "dur": 12,
    "args": {
      "External id": 79716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79716,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79716, "pid": 5, "tid": 7, "ts": 1716454222601281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538766, "dur": 8,
    "args": {
      "External id": 79716,
      "cbid": 211, "correlation": 79716
    }
  },
  {
    "ph": "s", "id": 79716, "pid": 76337, "tid": -914061504, "ts": 1716454222538766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222601294, "dur": 10,
    "args": {
      "External id": 79724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79724,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79724, "pid": 5, "tid": 7, "ts": 1716454222601294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538794, "dur": 9,
    "args": {
      "External id": 79724,
      "cbid": 211, "correlation": 79724
    }
  },
  {
    "ph": "s", "id": 79724, "pid": 76337, "tid": -914061504, "ts": 1716454222538794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222601305, "dur": 19,
    "args": {
      "External id": 79744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79744,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 79744, "pid": 5, "tid": 7, "ts": 1716454222601305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538876, "dur": 13,
    "args": {
      "External id": 79744,
      "cbid": 211, "correlation": 79744
    }
  },
  {
    "ph": "s", "id": 79744, "pid": 76337, "tid": -914061504, "ts": 1716454222538876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222601325, "dur": 4,
    "args": {
      "External id": 79756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79756,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 79756, "pid": 5, "tid": 7, "ts": 1716454222601325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538899, "dur": 6,
    "args": {
      "External id": 79756,
      "cbid": 211, "correlation": 79756
    }
  },
  {
    "ph": "s", "id": 79756, "pid": 76337, "tid": -914061504, "ts": 1716454222538899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222601331, "dur": 18,
    "args": {
      "External id": 79759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79759, "pid": 5, "tid": 7, "ts": 1716454222601331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538916, "dur": 7,
    "args": {
      "External id": 79759,
      "cbid": 211, "correlation": 79759
    }
  },
  {
    "ph": "s", "id": 79759, "pid": 76337, "tid": -914061504, "ts": 1716454222538916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538982, "dur": 0,
    "args": {
      "External id": 79770,
      "cbid": 317, "correlation": 79770
    }
  },
  {
    "ph": "f", "id": 79770, "pid": 76337, "tid": -914061504, "ts": 1716454222538982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538983, "dur": 0,
    "args": {
      "External id": 79771,
      "cbid": 203, "correlation": 79771
    }
  },
  {
    "ph": "f", "id": 79771, "pid": 76337, "tid": -914061504, "ts": 1716454222538983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538984, "dur": 0,
    "args": {
      "External id": 79772,
      "cbid": 205, "correlation": 79772
    }
  },
  {
    "ph": "f", "id": 79772, "pid": 76337, "tid": -914061504, "ts": 1716454222538984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222601350, "dur": 13,
    "args": {
      "External id": 79776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79776, "pid": 5, "tid": 7, "ts": 1716454222601350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222538997, "dur": 12,
    "args": {
      "External id": 79776,
      "cbid": 211, "correlation": 79776
    }
  },
  {
    "ph": "s", "id": 79776, "pid": 76337, "tid": -914061504, "ts": 1716454222538997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222601364, "dur": 3,
    "args": {
      "External id": 79778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79778, "pid": 5, "tid": 7, "ts": 1716454222601364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539013, "dur": 6,
    "args": {
      "External id": 79778,
      "cbid": 211, "correlation": 79778
    }
  },
  {
    "ph": "s", "id": 79778, "pid": 76337, "tid": -914061504, "ts": 1716454222539013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539022, "dur": 0,
    "args": {
      "External id": 79779,
      "cbid": 51, "correlation": 79779
    }
  },
  {
    "ph": "s", "id": 79779, "pid": 76337, "tid": -914061504, "ts": 1716454222539022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222601369, "dur": 93,
    "args": {
      "External id": 79780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79780,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 79780, "pid": 5, "tid": 7, "ts": 1716454222601369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539023, "dur": 5,
    "args": {
      "External id": 79780,
      "cbid": 211, "correlation": 79780
    }
  },
  {
    "ph": "s", "id": 79780, "pid": 76337, "tid": -914061504, "ts": 1716454222539023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222601463, "dur": 16,
    "args": {
      "External id": 79785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79785, "pid": 5, "tid": 7, "ts": 1716454222601463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539050, "dur": 8,
    "args": {
      "External id": 79785,
      "cbid": 211, "correlation": 79785
    }
  },
  {
    "ph": "s", "id": 79785, "pid": 76337, "tid": -914061504, "ts": 1716454222539050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222601480, "dur": 86,
    "args": {
      "External id": 79794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79794,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79794, "pid": 5, "tid": 7, "ts": 1716454222601480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539132, "dur": 14,
    "args": {
      "External id": 79794,
      "cbid": 211, "correlation": 79794
    }
  },
  {
    "ph": "s", "id": 79794, "pid": 76337, "tid": -914061504, "ts": 1716454222539132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222601567, "dur": 30,
    "args": {
      "External id": 79816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79816,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79816, "pid": 5, "tid": 7, "ts": 1716454222601567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539189, "dur": 10,
    "args": {
      "External id": 79816,
      "cbid": 211, "correlation": 79816
    }
  },
  {
    "ph": "s", "id": 79816, "pid": 76337, "tid": -914061504, "ts": 1716454222539189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539277, "dur": 1,
    "args": {
      "External id": 79827,
      "cbid": 251, "correlation": 79827
    }
  },
  {
    "ph": "f", "id": 79827, "pid": 76337, "tid": -914061504, "ts": 1716454222539277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222601598, "dur": 168,
    "args": {
      "External id": 79828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79828,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79828, "pid": 5, "tid": 7, "ts": 1716454222601598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539283, "dur": 13,
    "args": {
      "External id": 79828,
      "cbid": 211, "correlation": 79828
    }
  },
  {
    "ph": "s", "id": 79828, "pid": 76337, "tid": -914061504, "ts": 1716454222539283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539353, "dur": 1,
    "args": {
      "External id": 79839,
      "cbid": 251, "correlation": 79839
    }
  },
  {
    "ph": "f", "id": 79839, "pid": 76337, "tid": -914061504, "ts": 1716454222539353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222601767, "dur": 161,
    "args": {
      "External id": 79840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79840,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79840, "pid": 5, "tid": 7, "ts": 1716454222601767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539357, "dur": 12,
    "args": {
      "External id": 79840,
      "cbid": 211, "correlation": 79840
    }
  },
  {
    "ph": "s", "id": 79840, "pid": 76337, "tid": -914061504, "ts": 1716454222539357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539424, "dur": 1,
    "args": {
      "External id": 79851,
      "cbid": 251, "correlation": 79851
    }
  },
  {
    "ph": "f", "id": 79851, "pid": 76337, "tid": -914061504, "ts": 1716454222539424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222601930, "dur": 161,
    "args": {
      "External id": 79852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79852,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79852, "pid": 5, "tid": 7, "ts": 1716454222601930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539428, "dur": 12,
    "args": {
      "External id": 79852,
      "cbid": 211, "correlation": 79852
    }
  },
  {
    "ph": "s", "id": 79852, "pid": 76337, "tid": -914061504, "ts": 1716454222539428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222602093, "dur": 342,
    "args": {
      "External id": 79877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79877,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79877, "pid": 5, "tid": 7, "ts": 1716454222602093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539512, "dur": 12,
    "args": {
      "External id": 79877,
      "cbid": 211, "correlation": 79877
    }
  },
  {
    "ph": "s", "id": 79877, "pid": 76337, "tid": -914061504, "ts": 1716454222539512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539612, "dur": 1,
    "args": {
      "External id": 79895,
      "cbid": 251, "correlation": 79895
    }
  },
  {
    "ph": "f", "id": 79895, "pid": 76337, "tid": -914061504, "ts": 1716454222539612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222602436, "dur": 169,
    "args": {
      "External id": 79897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79897,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79897, "pid": 5, "tid": 7, "ts": 1716454222602436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539617, "dur": 13,
    "args": {
      "External id": 79897,
      "cbid": 211, "correlation": 79897
    }
  },
  {
    "ph": "s", "id": 79897, "pid": 76337, "tid": -914061504, "ts": 1716454222539617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222602607, "dur": 19,
    "args": {
      "External id": 79905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79905,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79905, "pid": 5, "tid": 7, "ts": 1716454222602607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539688, "dur": 12,
    "args": {
      "External id": 79905,
      "cbid": 211, "correlation": 79905
    }
  },
  {
    "ph": "s", "id": 79905, "pid": 76337, "tid": -914061504, "ts": 1716454222539688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222602627, "dur": 27,
    "args": {
      "External id": 79913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79913,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79913, "pid": 5, "tid": 7, "ts": 1716454222602627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539728, "dur": 8,
    "args": {
      "External id": 79913,
      "cbid": 211, "correlation": 79913
    }
  },
  {
    "ph": "s", "id": 79913, "pid": 76337, "tid": -914061504, "ts": 1716454222539728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222602656, "dur": 19,
    "args": {
      "External id": 79924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79924, "pid": 5, "tid": 7, "ts": 1716454222602656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539801, "dur": 13,
    "args": {
      "External id": 79924,
      "cbid": 211, "correlation": 79924
    }
  },
  {
    "ph": "s", "id": 79924, "pid": 76337, "tid": -914061504, "ts": 1716454222539801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222602676, "dur": 16,
    "args": {
      "External id": 79946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79946,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 79946, "pid": 5, "tid": 7, "ts": 1716454222602676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539832, "dur": 8,
    "args": {
      "External id": 79946,
      "cbid": 211, "correlation": 79946
    }
  },
  {
    "ph": "s", "id": 79946, "pid": 76337, "tid": -914061504, "ts": 1716454222539832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539919, "dur": 1,
    "args": {
      "External id": 79957,
      "cbid": 251, "correlation": 79957
    }
  },
  {
    "ph": "f", "id": 79957, "pid": 76337, "tid": -914061504, "ts": 1716454222539919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222602694, "dur": 92,
    "args": {
      "External id": 79958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79958,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 79958, "pid": 5, "tid": 7, "ts": 1716454222602694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222539925, "dur": 13,
    "args": {
      "External id": 79958,
      "cbid": 211, "correlation": 79958
    }
  },
  {
    "ph": "s", "id": 79958, "pid": 76337, "tid": -914061504, "ts": 1716454222539925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540001, "dur": 1,
    "args": {
      "External id": 79969,
      "cbid": 251, "correlation": 79969
    }
  },
  {
    "ph": "f", "id": 79969, "pid": 76337, "tid": -914061504, "ts": 1716454222540001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540005, "dur": 0,
    "args": {
      "External id": 79970,
      "cbid": 251, "correlation": 79970
    }
  },
  {
    "ph": "f", "id": 79970, "pid": 76337, "tid": -914061504, "ts": 1716454222540005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222602787, "dur": 13,
    "args": {
      "External id": 79971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79971,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79971, "pid": 5, "tid": 7, "ts": 1716454222602787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540007, "dur": 12,
    "args": {
      "External id": 79971,
      "cbid": 211, "correlation": 79971
    }
  },
  {
    "ph": "s", "id": 79971, "pid": 76337, "tid": -914061504, "ts": 1716454222540007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222602802, "dur": 6,
    "args": {
      "External id": 79973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79973,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79973, "pid": 5, "tid": 7, "ts": 1716454222602802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540021, "dur": 6,
    "args": {
      "External id": 79973,
      "cbid": 211, "correlation": 79973
    }
  },
  {
    "ph": "s", "id": 79973, "pid": 76337, "tid": -914061504, "ts": 1716454222540021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540080, "dur": 1,
    "args": {
      "External id": 79984,
      "cbid": 251, "correlation": 79984
    }
  },
  {
    "ph": "f", "id": 79984, "pid": 76337, "tid": -914061504, "ts": 1716454222540080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540083, "dur": 0,
    "args": {
      "External id": 79985,
      "cbid": 251, "correlation": 79985
    }
  },
  {
    "ph": "f", "id": 79985, "pid": 76337, "tid": -914061504, "ts": 1716454222540083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222602809, "dur": 9,
    "args": {
      "External id": 79986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79986,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79986, "pid": 5, "tid": 7, "ts": 1716454222602809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540085, "dur": 12,
    "args": {
      "External id": 79986,
      "cbid": 211, "correlation": 79986
    }
  },
  {
    "ph": "s", "id": 79986, "pid": 76337, "tid": -914061504, "ts": 1716454222540085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222602819, "dur": 3,
    "args": {
      "External id": 79988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 79988,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 79988, "pid": 5, "tid": 7, "ts": 1716454222602819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540098, "dur": 5,
    "args": {
      "External id": 79988,
      "cbid": 211, "correlation": 79988
    }
  },
  {
    "ph": "s", "id": 79988, "pid": 76337, "tid": -914061504, "ts": 1716454222540098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222602823, "dur": 55,
    "args": {
      "External id": 80013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80013,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80013, "pid": 5, "tid": 7, "ts": 1716454222602823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540175, "dur": 12,
    "args": {
      "External id": 80013,
      "cbid": 211, "correlation": 80013
    }
  },
  {
    "ph": "s", "id": 80013, "pid": 76337, "tid": -914061504, "ts": 1716454222540175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540273, "dur": 1,
    "args": {
      "External id": 80031,
      "cbid": 251, "correlation": 80031
    }
  },
  {
    "ph": "f", "id": 80031, "pid": 76337, "tid": -914061504, "ts": 1716454222540273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222602880, "dur": 92,
    "args": {
      "External id": 80033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80033,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80033, "pid": 5, "tid": 7, "ts": 1716454222602880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540279, "dur": 13,
    "args": {
      "External id": 80033,
      "cbid": 211, "correlation": 80033
    }
  },
  {
    "ph": "s", "id": 80033, "pid": 76337, "tid": -914061504, "ts": 1716454222540279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222602973, "dur": 11,
    "args": {
      "External id": 80041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80041,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80041, "pid": 5, "tid": 7, "ts": 1716454222602973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540348, "dur": 13,
    "args": {
      "External id": 80041,
      "cbid": 211, "correlation": 80041
    }
  },
  {
    "ph": "s", "id": 80041, "pid": 76337, "tid": -914061504, "ts": 1716454222540348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222602986, "dur": 21,
    "args": {
      "External id": 80049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80049,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80049, "pid": 5, "tid": 7, "ts": 1716454222602986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540391, "dur": 10,
    "args": {
      "External id": 80049,
      "cbid": 211, "correlation": 80049
    }
  },
  {
    "ph": "s", "id": 80049, "pid": 76337, "tid": -914061504, "ts": 1716454222540391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222603008, "dur": 18,
    "args": {
      "External id": 80071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80071,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80071, "pid": 5, "tid": 7, "ts": 1716454222603008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540442, "dur": 10,
    "args": {
      "External id": 80071,
      "cbid": 211, "correlation": 80071
    }
  },
  {
    "ph": "s", "id": 80071, "pid": 76337, "tid": -914061504, "ts": 1716454222540442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540530, "dur": 1,
    "args": {
      "External id": 80087,
      "cbid": 251, "correlation": 80087
    }
  },
  {
    "ph": "f", "id": 80087, "pid": 76337, "tid": -914061504, "ts": 1716454222540530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540535, "dur": 0,
    "args": {
      "External id": 80089,
      "cbid": 251, "correlation": 80089
    }
  },
  {
    "ph": "f", "id": 80089, "pid": 76337, "tid": -914061504, "ts": 1716454222540535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222603027, "dur": 496,
    "args": {
      "External id": 80090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80090,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80090, "pid": 5, "tid": 7, "ts": 1716454222603027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540537, "dur": 12,
    "args": {
      "External id": 80090,
      "cbid": 211, "correlation": 80090
    }
  },
  {
    "ph": "s", "id": 80090, "pid": 76337, "tid": -914061504, "ts": 1716454222540537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222603525, "dur": 66,
    "args": {
      "External id": 80098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80098, "pid": 5, "tid": 7, "ts": 1716454222603525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540602, "dur": 12,
    "args": {
      "External id": 80098,
      "cbid": 211, "correlation": 80098
    }
  },
  {
    "ph": "s", "id": 80098, "pid": 76337, "tid": -914061504, "ts": 1716454222540602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222603592, "dur": 66,
    "args": {
      "External id": 80106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80106, "pid": 5, "tid": 7, "ts": 1716454222603592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540634, "dur": 8,
    "args": {
      "External id": 80106,
      "cbid": 211, "correlation": 80106
    }
  },
  {
    "ph": "s", "id": 80106, "pid": 76337, "tid": -914061504, "ts": 1716454222540634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540713, "dur": 1,
    "args": {
      "External id": 80122,
      "cbid": 251, "correlation": 80122
    }
  },
  {
    "ph": "f", "id": 80122, "pid": 76337, "tid": -914061504, "ts": 1716454222540713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222603661, "dur": 1,
    "args": {
      "External id": 80124,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 80124,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 80124, "pid": 5, "tid": 7, "ts": 1716454222603661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540718, "dur": 9,
    "args": {
      "External id": 80124,
      "cbid": 51, "correlation": 80124
    }
  },
  {
    "ph": "s", "id": 80124, "pid": 76337, "tid": -914061504, "ts": 1716454222540718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222603665, "dur": 273,
    "args": {
      "External id": 80125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80125,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80125, "pid": 5, "tid": 7, "ts": 1716454222603665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540728, "dur": 11,
    "args": {
      "External id": 80125,
      "cbid": 211, "correlation": 80125
    }
  },
  {
    "ph": "s", "id": 80125, "pid": 76337, "tid": -914061504, "ts": 1716454222540728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222603940, "dur": 14,
    "args": {
      "External id": 80133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80133,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80133, "pid": 5, "tid": 7, "ts": 1716454222603940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540770, "dur": 10,
    "args": {
      "External id": 80133,
      "cbid": 211, "correlation": 80133
    }
  },
  {
    "ph": "s", "id": 80133, "pid": 76337, "tid": -914061504, "ts": 1716454222540770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222603955, "dur": 38,
    "args": {
      "External id": 80144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80144, "pid": 5, "tid": 7, "ts": 1716454222603955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540839, "dur": 12,
    "args": {
      "External id": 80144,
      "cbid": 211, "correlation": 80144
    }
  },
  {
    "ph": "s", "id": 80144, "pid": 76337, "tid": -914061504, "ts": 1716454222540839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540903, "dur": 0,
    "args": {
      "External id": 80156,
      "cbid": 317, "correlation": 80156
    }
  },
  {
    "ph": "f", "id": 80156, "pid": 76337, "tid": -914061504, "ts": 1716454222540903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540904, "dur": 0,
    "args": {
      "External id": 80157,
      "cbid": 203, "correlation": 80157
    }
  },
  {
    "ph": "f", "id": 80157, "pid": 76337, "tid": -914061504, "ts": 1716454222540904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540904, "dur": 0,
    "args": {
      "External id": 80158,
      "cbid": 205, "correlation": 80158
    }
  },
  {
    "ph": "f", "id": 80158, "pid": 76337, "tid": -914061504, "ts": 1716454222540904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222603995, "dur": 13,
    "args": {
      "External id": 80162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80162,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80162, "pid": 5, "tid": 7, "ts": 1716454222603995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540920, "dur": 12,
    "args": {
      "External id": 80162,
      "cbid": 211, "correlation": 80162
    }
  },
  {
    "ph": "s", "id": 80162, "pid": 76337, "tid": -914061504, "ts": 1716454222540920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222604009, "dur": 4,
    "args": {
      "External id": 80164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80164, "pid": 5, "tid": 7, "ts": 1716454222604009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540936, "dur": 6,
    "args": {
      "External id": 80164,
      "cbid": 211, "correlation": 80164
    }
  },
  {
    "ph": "s", "id": 80164, "pid": 76337, "tid": -914061504, "ts": 1716454222540936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540946, "dur": 0,
    "args": {
      "External id": 80165,
      "cbid": 51, "correlation": 80165
    }
  },
  {
    "ph": "s", "id": 80165, "pid": 76337, "tid": -914061504, "ts": 1716454222540946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222604014, "dur": 99,
    "args": {
      "External id": 80166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80166,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 80166, "pid": 5, "tid": 7, "ts": 1716454222604014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540947, "dur": 5,
    "args": {
      "External id": 80166,
      "cbid": 211, "correlation": 80166
    }
  },
  {
    "ph": "s", "id": 80166, "pid": 76337, "tid": -914061504, "ts": 1716454222540947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222604115, "dur": 16,
    "args": {
      "External id": 80171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80171,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80171, "pid": 5, "tid": 7, "ts": 1716454222604115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222540981, "dur": 10,
    "args": {
      "External id": 80171,
      "cbid": 211, "correlation": 80171
    }
  },
  {
    "ph": "s", "id": 80171, "pid": 76337, "tid": -914061504, "ts": 1716454222540981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222604133, "dur": 12,
    "args": {
      "External id": 80179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80179,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80179, "pid": 5, "tid": 7, "ts": 1716454222604133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541014, "dur": 8,
    "args": {
      "External id": 80179,
      "cbid": 211, "correlation": 80179
    }
  },
  {
    "ph": "s", "id": 80179, "pid": 76337, "tid": -914061504, "ts": 1716454222541014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222604146, "dur": 25,
    "args": {
      "External id": 80188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80188,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80188, "pid": 5, "tid": 7, "ts": 1716454222604146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541052, "dur": 10,
    "args": {
      "External id": 80188,
      "cbid": 211, "correlation": 80188
    }
  },
  {
    "ph": "s", "id": 80188, "pid": 76337, "tid": -914061504, "ts": 1716454222541052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222604172, "dur": 25,
    "args": {
      "External id": 80208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80208,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 80208, "pid": 5, "tid": 7, "ts": 1716454222604172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541123, "dur": 11,
    "args": {
      "External id": 80208,
      "cbid": 211, "correlation": 80208
    }
  },
  {
    "ph": "s", "id": 80208, "pid": 76337, "tid": -914061504, "ts": 1716454222541123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222604199, "dur": 5,
    "args": {
      "External id": 80220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80220,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 80220, "pid": 5, "tid": 7, "ts": 1716454222604199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541144, "dur": 7,
    "args": {
      "External id": 80220,
      "cbid": 211, "correlation": 80220
    }
  },
  {
    "ph": "s", "id": 80220, "pid": 76337, "tid": -914061504, "ts": 1716454222541144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222604205, "dur": 25,
    "args": {
      "External id": 80223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80223, "pid": 5, "tid": 7, "ts": 1716454222604205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541163, "dur": 7,
    "args": {
      "External id": 80223,
      "cbid": 211, "correlation": 80223
    }
  },
  {
    "ph": "s", "id": 80223, "pid": 76337, "tid": -914061504, "ts": 1716454222541163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222604231, "dur": 18,
    "args": {
      "External id": 80232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80232,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80232, "pid": 5, "tid": 7, "ts": 1716454222604231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541202, "dur": 9,
    "args": {
      "External id": 80232,
      "cbid": 211, "correlation": 80232
    }
  },
  {
    "ph": "s", "id": 80232, "pid": 76337, "tid": -914061504, "ts": 1716454222541202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541254, "dur": 0,
    "args": {
      "External id": 80242,
      "cbid": 317, "correlation": 80242
    }
  },
  {
    "ph": "f", "id": 80242, "pid": 76337, "tid": -914061504, "ts": 1716454222541254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541254, "dur": 0,
    "args": {
      "External id": 80243,
      "cbid": 203, "correlation": 80243
    }
  },
  {
    "ph": "f", "id": 80243, "pid": 76337, "tid": -914061504, "ts": 1716454222541254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541255, "dur": 0,
    "args": {
      "External id": 80244,
      "cbid": 205, "correlation": 80244
    }
  },
  {
    "ph": "f", "id": 80244, "pid": 76337, "tid": -914061504, "ts": 1716454222541255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222604250, "dur": 18,
    "args": {
      "External id": 80248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80248,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80248, "pid": 5, "tid": 7, "ts": 1716454222604250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541270, "dur": 11,
    "args": {
      "External id": 80248,
      "cbid": 211, "correlation": 80248
    }
  },
  {
    "ph": "s", "id": 80248, "pid": 76337, "tid": -914061504, "ts": 1716454222541270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222604269, "dur": 246,
    "args": {
      "External id": 80250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80250,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80250, "pid": 5, "tid": 7, "ts": 1716454222604269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541284, "dur": 5,
    "args": {
      "External id": 80250,
      "cbid": 211, "correlation": 80250
    }
  },
  {
    "ph": "s", "id": 80250, "pid": 76337, "tid": -914061504, "ts": 1716454222541284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222604518, "dur": 1,
    "args": {
      "External id": 80252,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 80252,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 80252, "pid": 5, "tid": 7, "ts": 1716454222604518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541295, "dur": 8,
    "args": {
      "External id": 80252,
      "cbid": 51, "correlation": 80252
    }
  },
  {
    "ph": "s", "id": 80252, "pid": 76337, "tid": -914061504, "ts": 1716454222541295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222604522, "dur": 824,
    "args": {
      "External id": 80253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80253,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80253, "pid": 5, "tid": 7, "ts": 1716454222604522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541304, "dur": 6,
    "args": {
      "External id": 80253,
      "cbid": 211, "correlation": 80253
    }
  },
  {
    "ph": "s", "id": 80253, "pid": 76337, "tid": -914061504, "ts": 1716454222541304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222605346, "dur": 13,
    "args": {
      "External id": 80255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80255,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80255, "pid": 5, "tid": 7, "ts": 1716454222605346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541315, "dur": 5,
    "args": {
      "External id": 80255,
      "cbid": 211, "correlation": 80255
    }
  },
  {
    "ph": "s", "id": 80255, "pid": 76337, "tid": -914061504, "ts": 1716454222541315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222605361, "dur": 15,
    "args": {
      "External id": 80261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80261,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80261, "pid": 5, "tid": 7, "ts": 1716454222605361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541343, "dur": 9,
    "args": {
      "External id": 80261,
      "cbid": 211, "correlation": 80261
    }
  },
  {
    "ph": "s", "id": 80261, "pid": 76337, "tid": -914061504, "ts": 1716454222541343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222605377, "dur": 3,
    "args": {
      "External id": 80269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80269,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 80269, "pid": 5, "tid": 7, "ts": 1716454222605377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541387, "dur": 9,
    "args": {
      "External id": 80269,
      "cbid": 211, "correlation": 80269
    }
  },
  {
    "ph": "s", "id": 80269, "pid": 76337, "tid": -914061504, "ts": 1716454222541387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541453, "dur": 2,
    "args": {
      "External id": 80285,
      "cbid": 251, "correlation": 80285
    }
  },
  {
    "ph": "f", "id": 80285, "pid": 76337, "tid": -914061504, "ts": 1716454222541453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541459, "dur": 0,
    "args": {
      "External id": 80287,
      "cbid": 251, "correlation": 80287
    }
  },
  {
    "ph": "f", "id": 80287, "pid": 76337, "tid": -914061504, "ts": 1716454222541459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222605382, "dur": 13,
    "args": {
      "External id": 80288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80288,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80288, "pid": 5, "tid": 7, "ts": 1716454222605382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541461, "dur": 11,
    "args": {
      "External id": 80288,
      "cbid": 211, "correlation": 80288
    }
  },
  {
    "ph": "s", "id": 80288, "pid": 76337, "tid": -914061504, "ts": 1716454222541461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222605397, "dur": 5,
    "args": {
      "External id": 80290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80290,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80290, "pid": 5, "tid": 7, "ts": 1716454222605397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541473, "dur": 6,
    "args": {
      "External id": 80290,
      "cbid": 211, "correlation": 80290
    }
  },
  {
    "ph": "s", "id": 80290, "pid": 76337, "tid": -914061504, "ts": 1716454222541473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222605403, "dur": 17,
    "args": {
      "External id": 80300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80300,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80300, "pid": 5, "tid": 7, "ts": 1716454222605403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541532, "dur": 12,
    "args": {
      "External id": 80300,
      "cbid": 211, "correlation": 80300
    }
  },
  {
    "ph": "s", "id": 80300, "pid": 76337, "tid": -914061504, "ts": 1716454222541532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222605421, "dur": 18,
    "args": {
      "External id": 80320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80320,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 80320, "pid": 5, "tid": 7, "ts": 1716454222605421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541597, "dur": 11,
    "args": {
      "External id": 80320,
      "cbid": 211, "correlation": 80320
    }
  },
  {
    "ph": "s", "id": 80320, "pid": 76337, "tid": -914061504, "ts": 1716454222541597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222605440, "dur": 4,
    "args": {
      "External id": 80332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80332,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 80332, "pid": 5, "tid": 7, "ts": 1716454222605440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541618, "dur": 6,
    "args": {
      "External id": 80332,
      "cbid": 211, "correlation": 80332
    }
  },
  {
    "ph": "s", "id": 80332, "pid": 76337, "tid": -914061504, "ts": 1716454222541618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222605446, "dur": 17,
    "args": {
      "External id": 80335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80335,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80335, "pid": 5, "tid": 7, "ts": 1716454222605446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541636, "dur": 6,
    "args": {
      "External id": 80335,
      "cbid": 211, "correlation": 80335
    }
  },
  {
    "ph": "s", "id": 80335, "pid": 76337, "tid": -914061504, "ts": 1716454222541636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222605464, "dur": 11,
    "args": {
      "External id": 80344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80344,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80344, "pid": 5, "tid": 7, "ts": 1716454222605464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541677, "dur": 11,
    "args": {
      "External id": 80344,
      "cbid": 211, "correlation": 80344
    }
  },
  {
    "ph": "s", "id": 80344, "pid": 76337, "tid": -914061504, "ts": 1716454222541677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541740, "dur": 0,
    "args": {
      "External id": 80354,
      "cbid": 317, "correlation": 80354
    }
  },
  {
    "ph": "f", "id": 80354, "pid": 76337, "tid": -914061504, "ts": 1716454222541740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541740, "dur": 0,
    "args": {
      "External id": 80355,
      "cbid": 203, "correlation": 80355
    }
  },
  {
    "ph": "f", "id": 80355, "pid": 76337, "tid": -914061504, "ts": 1716454222541740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541741, "dur": 0,
    "args": {
      "External id": 80356,
      "cbid": 205, "correlation": 80356
    }
  },
  {
    "ph": "f", "id": 80356, "pid": 76337, "tid": -914061504, "ts": 1716454222541741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222605476, "dur": 11,
    "args": {
      "External id": 80360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80360,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80360, "pid": 5, "tid": 7, "ts": 1716454222605476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541755, "dur": 12,
    "args": {
      "External id": 80360,
      "cbid": 211, "correlation": 80360
    }
  },
  {
    "ph": "s", "id": 80360, "pid": 76337, "tid": -914061504, "ts": 1716454222541755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222605489, "dur": 165,
    "args": {
      "External id": 80362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80362,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80362, "pid": 5, "tid": 7, "ts": 1716454222605489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541769, "dur": 6,
    "args": {
      "External id": 80362,
      "cbid": 211, "correlation": 80362
    }
  },
  {
    "ph": "s", "id": 80362, "pid": 76337, "tid": -914061504, "ts": 1716454222541769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222605657, "dur": 1,
    "args": {
      "External id": 80364,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 80364,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 80364, "pid": 5, "tid": 7, "ts": 1716454222605657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541780, "dur": 6,
    "args": {
      "External id": 80364,
      "cbid": 51, "correlation": 80364
    }
  },
  {
    "ph": "s", "id": 80364, "pid": 76337, "tid": -914061504, "ts": 1716454222541780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222605660, "dur": 660,
    "args": {
      "External id": 80365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80365,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80365, "pid": 5, "tid": 7, "ts": 1716454222605660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541788, "dur": 6,
    "args": {
      "External id": 80365,
      "cbid": 211, "correlation": 80365
    }
  },
  {
    "ph": "s", "id": 80365, "pid": 76337, "tid": -914061504, "ts": 1716454222541788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222606322, "dur": 13,
    "args": {
      "External id": 80367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80367,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80367, "pid": 5, "tid": 7, "ts": 1716454222606322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541798, "dur": 5,
    "args": {
      "External id": 80367,
      "cbid": 211, "correlation": 80367
    }
  },
  {
    "ph": "s", "id": 80367, "pid": 76337, "tid": -914061504, "ts": 1716454222541798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222606335, "dur": 15,
    "args": {
      "External id": 80373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80373, "pid": 5, "tid": 7, "ts": 1716454222606335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541826, "dur": 9,
    "args": {
      "External id": 80373,
      "cbid": 211, "correlation": 80373
    }
  },
  {
    "ph": "s", "id": 80373, "pid": 76337, "tid": -914061504, "ts": 1716454222541826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541884, "dur": 0,
    "args": {
      "External id": 80383,
      "cbid": 317, "correlation": 80383
    }
  },
  {
    "ph": "f", "id": 80383, "pid": 76337, "tid": -914061504, "ts": 1716454222541884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541885, "dur": 0,
    "args": {
      "External id": 80384,
      "cbid": 203, "correlation": 80384
    }
  },
  {
    "ph": "f", "id": 80384, "pid": 76337, "tid": -914061504, "ts": 1716454222541885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541886, "dur": 0,
    "args": {
      "External id": 80385,
      "cbid": 205, "correlation": 80385
    }
  },
  {
    "ph": "f", "id": 80385, "pid": 76337, "tid": -914061504, "ts": 1716454222541886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222606352, "dur": 18,
    "args": {
      "External id": 80389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80389,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80389, "pid": 5, "tid": 7, "ts": 1716454222606352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541900, "dur": 11,
    "args": {
      "External id": 80389,
      "cbid": 211, "correlation": 80389
    }
  },
  {
    "ph": "s", "id": 80389, "pid": 76337, "tid": -914061504, "ts": 1716454222541900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222606371, "dur": 4,
    "args": {
      "External id": 80391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80391, "pid": 5, "tid": 7, "ts": 1716454222606371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541916, "dur": 6,
    "args": {
      "External id": 80391,
      "cbid": 211, "correlation": 80391
    }
  },
  {
    "ph": "s", "id": 80391, "pid": 76337, "tid": -914061504, "ts": 1716454222541916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541924, "dur": 0,
    "args": {
      "External id": 80392,
      "cbid": 51, "correlation": 80392
    }
  },
  {
    "ph": "s", "id": 80392, "pid": 76337, "tid": -914061504, "ts": 1716454222541924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222606376, "dur": 135,
    "args": {
      "External id": 80393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80393,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 80393, "pid": 5, "tid": 7, "ts": 1716454222606376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541925, "dur": 5,
    "args": {
      "External id": 80393,
      "cbid": 211, "correlation": 80393
    }
  },
  {
    "ph": "s", "id": 80393, "pid": 76337, "tid": -914061504, "ts": 1716454222541925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222606513, "dur": 16,
    "args": {
      "External id": 80398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80398,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80398, "pid": 5, "tid": 7, "ts": 1716454222606513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541951, "dur": 8,
    "args": {
      "External id": 80398,
      "cbid": 211, "correlation": 80398
    }
  },
  {
    "ph": "s", "id": 80398, "pid": 76337, "tid": -914061504, "ts": 1716454222541951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222606530, "dur": 12,
    "args": {
      "External id": 80406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80406,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80406, "pid": 5, "tid": 7, "ts": 1716454222606530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222541990, "dur": 9,
    "args": {
      "External id": 80406,
      "cbid": 211, "correlation": 80406
    }
  },
  {
    "ph": "s", "id": 80406, "pid": 76337, "tid": -914061504, "ts": 1716454222541990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222606544, "dur": 10,
    "args": {
      "External id": 80414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80414,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80414, "pid": 5, "tid": 7, "ts": 1716454222606544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542021, "dur": 8,
    "args": {
      "External id": 80414,
      "cbid": 211, "correlation": 80414
    }
  },
  {
    "ph": "s", "id": 80414, "pid": 76337, "tid": -914061504, "ts": 1716454222542021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222606555, "dur": 19,
    "args": {
      "External id": 80434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80434,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 80434, "pid": 5, "tid": 7, "ts": 1716454222606555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542105, "dur": 12,
    "args": {
      "External id": 80434,
      "cbid": 211, "correlation": 80434
    }
  },
  {
    "ph": "s", "id": 80434, "pid": 76337, "tid": -914061504, "ts": 1716454222542105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222606575, "dur": 4,
    "args": {
      "External id": 80446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80446,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 80446, "pid": 5, "tid": 7, "ts": 1716454222606575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542127, "dur": 6,
    "args": {
      "External id": 80446,
      "cbid": 211, "correlation": 80446
    }
  },
  {
    "ph": "s", "id": 80446, "pid": 76337, "tid": -914061504, "ts": 1716454222542127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222606581, "dur": 18,
    "args": {
      "External id": 80449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80449, "pid": 5, "tid": 7, "ts": 1716454222606581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542145, "dur": 7,
    "args": {
      "External id": 80449,
      "cbid": 211, "correlation": 80449
    }
  },
  {
    "ph": "s", "id": 80449, "pid": 76337, "tid": -914061504, "ts": 1716454222542145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542202, "dur": 0,
    "args": {
      "External id": 80460,
      "cbid": 317, "correlation": 80460
    }
  },
  {
    "ph": "f", "id": 80460, "pid": 76337, "tid": -914061504, "ts": 1716454222542202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542203, "dur": 0,
    "args": {
      "External id": 80461,
      "cbid": 203, "correlation": 80461
    }
  },
  {
    "ph": "f", "id": 80461, "pid": 76337, "tid": -914061504, "ts": 1716454222542203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542204, "dur": 0,
    "args": {
      "External id": 80462,
      "cbid": 205, "correlation": 80462
    }
  },
  {
    "ph": "f", "id": 80462, "pid": 76337, "tid": -914061504, "ts": 1716454222542204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222606600, "dur": 11,
    "args": {
      "External id": 80466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80466,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80466, "pid": 5, "tid": 7, "ts": 1716454222606600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542216, "dur": 11,
    "args": {
      "External id": 80466,
      "cbid": 211, "correlation": 80466
    }
  },
  {
    "ph": "s", "id": 80466, "pid": 76337, "tid": -914061504, "ts": 1716454222542216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222606613, "dur": 3,
    "args": {
      "External id": 80468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80468, "pid": 5, "tid": 7, "ts": 1716454222606613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542232, "dur": 6,
    "args": {
      "External id": 80468,
      "cbid": 211, "correlation": 80468
    }
  },
  {
    "ph": "s", "id": 80468, "pid": 76337, "tid": -914061504, "ts": 1716454222542232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542240, "dur": 0,
    "args": {
      "External id": 80469,
      "cbid": 51, "correlation": 80469
    }
  },
  {
    "ph": "s", "id": 80469, "pid": 76337, "tid": -914061504, "ts": 1716454222542240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222606618, "dur": 94,
    "args": {
      "External id": 80470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80470,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 80470, "pid": 5, "tid": 7, "ts": 1716454222606618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542241, "dur": 5,
    "args": {
      "External id": 80470,
      "cbid": 211, "correlation": 80470
    }
  },
  {
    "ph": "s", "id": 80470, "pid": 76337, "tid": -914061504, "ts": 1716454222542241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222606713, "dur": 16,
    "args": {
      "External id": 80475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80475, "pid": 5, "tid": 7, "ts": 1716454222606713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542268, "dur": 8,
    "args": {
      "External id": 80475,
      "cbid": 211, "correlation": 80475
    }
  },
  {
    "ph": "s", "id": 80475, "pid": 76337, "tid": -914061504, "ts": 1716454222542268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222606731, "dur": 85,
    "args": {
      "External id": 80484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80484,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80484, "pid": 5, "tid": 7, "ts": 1716454222606731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542348, "dur": 15,
    "args": {
      "External id": 80484,
      "cbid": 211, "correlation": 80484
    }
  },
  {
    "ph": "s", "id": 80484, "pid": 76337, "tid": -914061504, "ts": 1716454222542348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222606818, "dur": 30,
    "args": {
      "External id": 80506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80506,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80506, "pid": 5, "tid": 7, "ts": 1716454222606818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542404, "dur": 11,
    "args": {
      "External id": 80506,
      "cbid": 211, "correlation": 80506
    }
  },
  {
    "ph": "s", "id": 80506, "pid": 76337, "tid": -914061504, "ts": 1716454222542404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542493, "dur": 1,
    "args": {
      "External id": 80517,
      "cbid": 251, "correlation": 80517
    }
  },
  {
    "ph": "f", "id": 80517, "pid": 76337, "tid": -914061504, "ts": 1716454222542493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222606849, "dur": 167,
    "args": {
      "External id": 80518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80518,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80518, "pid": 5, "tid": 7, "ts": 1716454222606849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542499, "dur": 14,
    "args": {
      "External id": 80518,
      "cbid": 211, "correlation": 80518
    }
  },
  {
    "ph": "s", "id": 80518, "pid": 76337, "tid": -914061504, "ts": 1716454222542499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542568, "dur": 1,
    "args": {
      "External id": 80529,
      "cbid": 251, "correlation": 80529
    }
  },
  {
    "ph": "f", "id": 80529, "pid": 76337, "tid": -914061504, "ts": 1716454222542568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222607017, "dur": 162,
    "args": {
      "External id": 80530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80530,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80530, "pid": 5, "tid": 7, "ts": 1716454222607017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542572, "dur": 11,
    "args": {
      "External id": 80530,
      "cbid": 211, "correlation": 80530
    }
  },
  {
    "ph": "s", "id": 80530, "pid": 76337, "tid": -914061504, "ts": 1716454222542572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542637, "dur": 1,
    "args": {
      "External id": 80541,
      "cbid": 251, "correlation": 80541
    }
  },
  {
    "ph": "f", "id": 80541, "pid": 76337, "tid": -914061504, "ts": 1716454222542637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222607181, "dur": 160,
    "args": {
      "External id": 80542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80542,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80542, "pid": 5, "tid": 7, "ts": 1716454222607181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542641, "dur": 11,
    "args": {
      "External id": 80542,
      "cbid": 211, "correlation": 80542
    }
  },
  {
    "ph": "s", "id": 80542, "pid": 76337, "tid": -914061504, "ts": 1716454222542641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222607342, "dur": 340,
    "args": {
      "External id": 80567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80567,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80567, "pid": 5, "tid": 7, "ts": 1716454222607342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542725, "dur": 13,
    "args": {
      "External id": 80567,
      "cbid": 211, "correlation": 80567
    }
  },
  {
    "ph": "s", "id": 80567, "pid": 76337, "tid": -914061504, "ts": 1716454222542725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542824, "dur": 1,
    "args": {
      "External id": 80585,
      "cbid": 251, "correlation": 80585
    }
  },
  {
    "ph": "f", "id": 80585, "pid": 76337, "tid": -914061504, "ts": 1716454222542824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222607683, "dur": 170,
    "args": {
      "External id": 80587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80587,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80587, "pid": 5, "tid": 7, "ts": 1716454222607683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542830, "dur": 13,
    "args": {
      "External id": 80587,
      "cbid": 211, "correlation": 80587
    }
  },
  {
    "ph": "s", "id": 80587, "pid": 76337, "tid": -914061504, "ts": 1716454222542830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222607855, "dur": 20,
    "args": {
      "External id": 80595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80595,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80595, "pid": 5, "tid": 7, "ts": 1716454222607855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542901, "dur": 13,
    "args": {
      "External id": 80595,
      "cbid": 211, "correlation": 80595
    }
  },
  {
    "ph": "s", "id": 80595, "pid": 76337, "tid": -914061504, "ts": 1716454222542901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222607876, "dur": 28,
    "args": {
      "External id": 80603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80603,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80603, "pid": 5, "tid": 7, "ts": 1716454222607876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222542942, "dur": 9,
    "args": {
      "External id": 80603,
      "cbid": 211, "correlation": 80603
    }
  },
  {
    "ph": "s", "id": 80603, "pid": 76337, "tid": -914061504, "ts": 1716454222542942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222607905, "dur": 18,
    "args": {
      "External id": 80614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80614, "pid": 5, "tid": 7, "ts": 1716454222607905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543023, "dur": 14,
    "args": {
      "External id": 80614,
      "cbid": 211, "correlation": 80614
    }
  },
  {
    "ph": "s", "id": 80614, "pid": 76337, "tid": -914061504, "ts": 1716454222543023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222607925, "dur": 16,
    "args": {
      "External id": 80636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80636,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80636, "pid": 5, "tid": 7, "ts": 1716454222607925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543056, "dur": 8,
    "args": {
      "External id": 80636,
      "cbid": 211, "correlation": 80636
    }
  },
  {
    "ph": "s", "id": 80636, "pid": 76337, "tid": -914061504, "ts": 1716454222543056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543142, "dur": 1,
    "args": {
      "External id": 80647,
      "cbid": 251, "correlation": 80647
    }
  },
  {
    "ph": "f", "id": 80647, "pid": 76337, "tid": -914061504, "ts": 1716454222543142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222607942, "dur": 91,
    "args": {
      "External id": 80648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80648,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80648, "pid": 5, "tid": 7, "ts": 1716454222607942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543147, "dur": 13,
    "args": {
      "External id": 80648,
      "cbid": 211, "correlation": 80648
    }
  },
  {
    "ph": "s", "id": 80648, "pid": 76337, "tid": -914061504, "ts": 1716454222543147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543216, "dur": 1,
    "args": {
      "External id": 80659,
      "cbid": 251, "correlation": 80659
    }
  },
  {
    "ph": "f", "id": 80659, "pid": 76337, "tid": -914061504, "ts": 1716454222543216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543220, "dur": 0,
    "args": {
      "External id": 80660,
      "cbid": 251, "correlation": 80660
    }
  },
  {
    "ph": "f", "id": 80660, "pid": 76337, "tid": -914061504, "ts": 1716454222543220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222608034, "dur": 12,
    "args": {
      "External id": 80661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80661,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80661, "pid": 5, "tid": 7, "ts": 1716454222608034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543222, "dur": 12,
    "args": {
      "External id": 80661,
      "cbid": 211, "correlation": 80661
    }
  },
  {
    "ph": "s", "id": 80661, "pid": 76337, "tid": -914061504, "ts": 1716454222543222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222608047, "dur": 5,
    "args": {
      "External id": 80663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80663,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80663, "pid": 5, "tid": 7, "ts": 1716454222608047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543235, "dur": 6,
    "args": {
      "External id": 80663,
      "cbid": 211, "correlation": 80663
    }
  },
  {
    "ph": "s", "id": 80663, "pid": 76337, "tid": -914061504, "ts": 1716454222543235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543293, "dur": 1,
    "args": {
      "External id": 80674,
      "cbid": 251, "correlation": 80674
    }
  },
  {
    "ph": "f", "id": 80674, "pid": 76337, "tid": -914061504, "ts": 1716454222543293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543296, "dur": 0,
    "args": {
      "External id": 80675,
      "cbid": 251, "correlation": 80675
    }
  },
  {
    "ph": "f", "id": 80675, "pid": 76337, "tid": -914061504, "ts": 1716454222543296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222608054, "dur": 8,
    "args": {
      "External id": 80676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80676,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80676, "pid": 5, "tid": 7, "ts": 1716454222608054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543298, "dur": 12,
    "args": {
      "External id": 80676,
      "cbid": 211, "correlation": 80676
    }
  },
  {
    "ph": "s", "id": 80676, "pid": 76337, "tid": -914061504, "ts": 1716454222543298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222608064, "dur": 3,
    "args": {
      "External id": 80678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80678,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80678, "pid": 5, "tid": 7, "ts": 1716454222608064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543311, "dur": 6,
    "args": {
      "External id": 80678,
      "cbid": 211, "correlation": 80678
    }
  },
  {
    "ph": "s", "id": 80678, "pid": 76337, "tid": -914061504, "ts": 1716454222543311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222608068, "dur": 57,
    "args": {
      "External id": 80703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80703,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80703, "pid": 5, "tid": 7, "ts": 1716454222608068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543388, "dur": 13,
    "args": {
      "External id": 80703,
      "cbid": 211, "correlation": 80703
    }
  },
  {
    "ph": "s", "id": 80703, "pid": 76337, "tid": -914061504, "ts": 1716454222543388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543486, "dur": 1,
    "args": {
      "External id": 80721,
      "cbid": 251, "correlation": 80721
    }
  },
  {
    "ph": "f", "id": 80721, "pid": 76337, "tid": -914061504, "ts": 1716454222543486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222608127, "dur": 93,
    "args": {
      "External id": 80723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80723,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80723, "pid": 5, "tid": 7, "ts": 1716454222608127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543493, "dur": 14,
    "args": {
      "External id": 80723,
      "cbid": 211, "correlation": 80723
    }
  },
  {
    "ph": "s", "id": 80723, "pid": 76337, "tid": -914061504, "ts": 1716454222543493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222608221, "dur": 9,
    "args": {
      "External id": 80731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80731,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80731, "pid": 5, "tid": 7, "ts": 1716454222608221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543563, "dur": 12,
    "args": {
      "External id": 80731,
      "cbid": 211, "correlation": 80731
    }
  },
  {
    "ph": "s", "id": 80731, "pid": 76337, "tid": -914061504, "ts": 1716454222543563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222608232, "dur": 21,
    "args": {
      "External id": 80739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80739, "pid": 5, "tid": 7, "ts": 1716454222608232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543605, "dur": 9,
    "args": {
      "External id": 80739,
      "cbid": 211, "correlation": 80739
    }
  },
  {
    "ph": "s", "id": 80739, "pid": 76337, "tid": -914061504, "ts": 1716454222543605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222608254, "dur": 18,
    "args": {
      "External id": 80761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80761,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80761, "pid": 5, "tid": 7, "ts": 1716454222608254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543655, "dur": 10,
    "args": {
      "External id": 80761,
      "cbid": 211, "correlation": 80761
    }
  },
  {
    "ph": "s", "id": 80761, "pid": 76337, "tid": -914061504, "ts": 1716454222543655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543742, "dur": 1,
    "args": {
      "External id": 80777,
      "cbid": 251, "correlation": 80777
    }
  },
  {
    "ph": "f", "id": 80777, "pid": 76337, "tid": -914061504, "ts": 1716454222543742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543747, "dur": 0,
    "args": {
      "External id": 80779,
      "cbid": 251, "correlation": 80779
    }
  },
  {
    "ph": "f", "id": 80779, "pid": 76337, "tid": -914061504, "ts": 1716454222543747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222608274, "dur": 500,
    "args": {
      "External id": 80780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80780,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80780, "pid": 5, "tid": 7, "ts": 1716454222608274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543749, "dur": 13,
    "args": {
      "External id": 80780,
      "cbid": 211, "correlation": 80780
    }
  },
  {
    "ph": "s", "id": 80780, "pid": 76337, "tid": -914061504, "ts": 1716454222543749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222608775, "dur": 67,
    "args": {
      "External id": 80788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80788, "pid": 5, "tid": 7, "ts": 1716454222608775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543815, "dur": 12,
    "args": {
      "External id": 80788,
      "cbid": 211, "correlation": 80788
    }
  },
  {
    "ph": "s", "id": 80788, "pid": 76337, "tid": -914061504, "ts": 1716454222543815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222608843, "dur": 67,
    "args": {
      "External id": 80796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80796, "pid": 5, "tid": 7, "ts": 1716454222608843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543845, "dur": 8,
    "args": {
      "External id": 80796,
      "cbid": 211, "correlation": 80796
    }
  },
  {
    "ph": "s", "id": 80796, "pid": 76337, "tid": -914061504, "ts": 1716454222543845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543924, "dur": 1,
    "args": {
      "External id": 80812,
      "cbid": 251, "correlation": 80812
    }
  },
  {
    "ph": "f", "id": 80812, "pid": 76337, "tid": -914061504, "ts": 1716454222543924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222608912, "dur": 1,
    "args": {
      "External id": 80814,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 80814,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 80814, "pid": 5, "tid": 7, "ts": 1716454222608912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543929, "dur": 9,
    "args": {
      "External id": 80814,
      "cbid": 51, "correlation": 80814
    }
  },
  {
    "ph": "s", "id": 80814, "pid": 76337, "tid": -914061504, "ts": 1716454222543929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222608916, "dur": 276,
    "args": {
      "External id": 80815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80815,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80815, "pid": 5, "tid": 7, "ts": 1716454222608916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543940, "dur": 11,
    "args": {
      "External id": 80815,
      "cbid": 211, "correlation": 80815
    }
  },
  {
    "ph": "s", "id": 80815, "pid": 76337, "tid": -914061504, "ts": 1716454222543940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222609193, "dur": 14,
    "args": {
      "External id": 80823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80823,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80823, "pid": 5, "tid": 7, "ts": 1716454222609193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222543989, "dur": 11,
    "args": {
      "External id": 80823,
      "cbid": 211, "correlation": 80823
    }
  },
  {
    "ph": "s", "id": 80823, "pid": 76337, "tid": -914061504, "ts": 1716454222543989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222609208, "dur": 38,
    "args": {
      "External id": 80834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80834,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80834, "pid": 5, "tid": 7, "ts": 1716454222609208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544059, "dur": 12,
    "args": {
      "External id": 80834,
      "cbid": 211, "correlation": 80834
    }
  },
  {
    "ph": "s", "id": 80834, "pid": 76337, "tid": -914061504, "ts": 1716454222544059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544122, "dur": 0,
    "args": {
      "External id": 80846,
      "cbid": 317, "correlation": 80846
    }
  },
  {
    "ph": "f", "id": 80846, "pid": 76337, "tid": -914061504, "ts": 1716454222544122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544123, "dur": 0,
    "args": {
      "External id": 80847,
      "cbid": 203, "correlation": 80847
    }
  },
  {
    "ph": "f", "id": 80847, "pid": 76337, "tid": -914061504, "ts": 1716454222544123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544124, "dur": 0,
    "args": {
      "External id": 80848,
      "cbid": 205, "correlation": 80848
    }
  },
  {
    "ph": "f", "id": 80848, "pid": 76337, "tid": -914061504, "ts": 1716454222544124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222609248, "dur": 13,
    "args": {
      "External id": 80852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80852,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80852, "pid": 5, "tid": 7, "ts": 1716454222609248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544139, "dur": 12,
    "args": {
      "External id": 80852,
      "cbid": 211, "correlation": 80852
    }
  },
  {
    "ph": "s", "id": 80852, "pid": 76337, "tid": -914061504, "ts": 1716454222544139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222609262, "dur": 4,
    "args": {
      "External id": 80854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 80854, "pid": 5, "tid": 7, "ts": 1716454222609262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544156, "dur": 5,
    "args": {
      "External id": 80854,
      "cbid": 211, "correlation": 80854
    }
  },
  {
    "ph": "s", "id": 80854, "pid": 76337, "tid": -914061504, "ts": 1716454222544156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544164, "dur": 0,
    "args": {
      "External id": 80855,
      "cbid": 51, "correlation": 80855
    }
  },
  {
    "ph": "s", "id": 80855, "pid": 76337, "tid": -914061504, "ts": 1716454222544164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222609267, "dur": 98,
    "args": {
      "External id": 80856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80856,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 80856, "pid": 5, "tid": 7, "ts": 1716454222609267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544165, "dur": 5,
    "args": {
      "External id": 80856,
      "cbid": 211, "correlation": 80856
    }
  },
  {
    "ph": "s", "id": 80856, "pid": 76337, "tid": -914061504, "ts": 1716454222544165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222609367, "dur": 16,
    "args": {
      "External id": 80861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80861, "pid": 5, "tid": 7, "ts": 1716454222609367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544192, "dur": 9,
    "args": {
      "External id": 80861,
      "cbid": 211, "correlation": 80861
    }
  },
  {
    "ph": "s", "id": 80861, "pid": 76337, "tid": -914061504, "ts": 1716454222544192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222609385, "dur": 11,
    "args": {
      "External id": 80869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80869,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80869, "pid": 5, "tid": 7, "ts": 1716454222609385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544224, "dur": 8,
    "args": {
      "External id": 80869,
      "cbid": 211, "correlation": 80869
    }
  },
  {
    "ph": "s", "id": 80869, "pid": 76337, "tid": -914061504, "ts": 1716454222544224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222609397, "dur": 57,
    "args": {
      "External id": 80880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80880,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80880, "pid": 5, "tid": 7, "ts": 1716454222609397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544289, "dur": 11,
    "args": {
      "External id": 80880,
      "cbid": 211, "correlation": 80880
    }
  },
  {
    "ph": "s", "id": 80880, "pid": 76337, "tid": -914061504, "ts": 1716454222544289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544344, "dur": 0,
    "args": {
      "External id": 80890,
      "cbid": 317, "correlation": 80890
    }
  },
  {
    "ph": "f", "id": 80890, "pid": 76337, "tid": -914061504, "ts": 1716454222544344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544345, "dur": 0,
    "args": {
      "External id": 80891,
      "cbid": 203, "correlation": 80891
    }
  },
  {
    "ph": "f", "id": 80891, "pid": 76337, "tid": -914061504, "ts": 1716454222544345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544346, "dur": 0,
    "args": {
      "External id": 80892,
      "cbid": 205, "correlation": 80892
    }
  },
  {
    "ph": "f", "id": 80892, "pid": 76337, "tid": -914061504, "ts": 1716454222544346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222609456, "dur": 39,
    "args": {
      "External id": 80896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80896,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80896, "pid": 5, "tid": 7, "ts": 1716454222609456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544362, "dur": 12,
    "args": {
      "External id": 80896,
      "cbid": 211, "correlation": 80896
    }
  },
  {
    "ph": "s", "id": 80896, "pid": 76337, "tid": -914061504, "ts": 1716454222544362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222609497, "dur": 165,
    "args": {
      "External id": 80898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80898,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80898, "pid": 5, "tid": 7, "ts": 1716454222609497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544376, "dur": 5,
    "args": {
      "External id": 80898,
      "cbid": 211, "correlation": 80898
    }
  },
  {
    "ph": "s", "id": 80898, "pid": 76337, "tid": -914061504, "ts": 1716454222544376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222609663, "dur": 1960,
    "args": {
      "External id": 80900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80900,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80900, "pid": 5, "tid": 7, "ts": 1716454222609663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544388, "dur": 8,
    "args": {
      "External id": 80900,
      "cbid": 211, "correlation": 80900
    }
  },
  {
    "ph": "s", "id": 80900, "pid": 76337, "tid": -914061504, "ts": 1716454222544388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222611624, "dur": 40,
    "args": {
      "External id": 80902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80902,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80902, "pid": 5, "tid": 7, "ts": 1716454222611624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544400, "dur": 5,
    "args": {
      "External id": 80902,
      "cbid": 211, "correlation": 80902
    }
  },
  {
    "ph": "s", "id": 80902, "pid": 76337, "tid": -914061504, "ts": 1716454222544400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222611665, "dur": 60,
    "args": {
      "External id": 80908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80908,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80908, "pid": 5, "tid": 7, "ts": 1716454222611665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544428, "dur": 9,
    "args": {
      "External id": 80908,
      "cbid": 211, "correlation": 80908
    }
  },
  {
    "ph": "s", "id": 80908, "pid": 76337, "tid": -914061504, "ts": 1716454222544428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222611726, "dur": 85,
    "args": {
      "External id": 80917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80917,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80917, "pid": 5, "tid": 7, "ts": 1716454222611726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544520, "dur": 13,
    "args": {
      "External id": 80917,
      "cbid": 211, "correlation": 80917
    }
  },
  {
    "ph": "s", "id": 80917, "pid": 76337, "tid": -914061504, "ts": 1716454222544520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222611813, "dur": 74,
    "args": {
      "External id": 80937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80937,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 80937, "pid": 5, "tid": 7, "ts": 1716454222611813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544591, "dur": 12,
    "args": {
      "External id": 80937,
      "cbid": 211, "correlation": 80937
    }
  },
  {
    "ph": "s", "id": 80937, "pid": 76337, "tid": -914061504, "ts": 1716454222544591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222611888, "dur": 5,
    "args": {
      "External id": 80949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80949,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 80949, "pid": 5, "tid": 7, "ts": 1716454222611888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544613, "dur": 6,
    "args": {
      "External id": 80949,
      "cbid": 211, "correlation": 80949
    }
  },
  {
    "ph": "s", "id": 80949, "pid": 76337, "tid": -914061504, "ts": 1716454222544613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222611894, "dur": 82,
    "args": {
      "External id": 80952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80952,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80952, "pid": 5, "tid": 7, "ts": 1716454222611894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544631, "dur": 7,
    "args": {
      "External id": 80952,
      "cbid": 211, "correlation": 80952
    }
  },
  {
    "ph": "s", "id": 80952, "pid": 76337, "tid": -914061504, "ts": 1716454222544631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222611978, "dur": 54,
    "args": {
      "External id": 80961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80961,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80961, "pid": 5, "tid": 7, "ts": 1716454222611978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544670, "dur": 10,
    "args": {
      "External id": 80961,
      "cbid": 211, "correlation": 80961
    }
  },
  {
    "ph": "s", "id": 80961, "pid": 76337, "tid": -914061504, "ts": 1716454222544670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544722, "dur": 0,
    "args": {
      "External id": 80971,
      "cbid": 317, "correlation": 80971
    }
  },
  {
    "ph": "f", "id": 80971, "pid": 76337, "tid": -914061504, "ts": 1716454222544722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544723, "dur": 0,
    "args": {
      "External id": 80972,
      "cbid": 203, "correlation": 80972
    }
  },
  {
    "ph": "f", "id": 80972, "pid": 76337, "tid": -914061504, "ts": 1716454222544723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544723, "dur": 0,
    "args": {
      "External id": 80973,
      "cbid": 205, "correlation": 80973
    }
  },
  {
    "ph": "f", "id": 80973, "pid": 76337, "tid": -914061504, "ts": 1716454222544723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222612033, "dur": 56,
    "args": {
      "External id": 80977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80977,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80977, "pid": 5, "tid": 7, "ts": 1716454222612033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544738, "dur": 11,
    "args": {
      "External id": 80977,
      "cbid": 211, "correlation": 80977
    }
  },
  {
    "ph": "s", "id": 80977, "pid": 76337, "tid": -914061504, "ts": 1716454222544738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222612091, "dur": 125,
    "args": {
      "External id": 80979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80979,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80979, "pid": 5, "tid": 7, "ts": 1716454222612091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544752, "dur": 5,
    "args": {
      "External id": 80979,
      "cbid": 211, "correlation": 80979
    }
  },
  {
    "ph": "s", "id": 80979, "pid": 76337, "tid": -914061504, "ts": 1716454222544752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222612216, "dur": 1920,
    "args": {
      "External id": 80981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80981,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 80981, "pid": 5, "tid": 7, "ts": 1716454222612216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544764, "dur": 6,
    "args": {
      "External id": 80981,
      "cbid": 211, "correlation": 80981
    }
  },
  {
    "ph": "s", "id": 80981, "pid": 76337, "tid": -914061504, "ts": 1716454222544764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222614138, "dur": 20,
    "args": {
      "External id": 80983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80983, "pid": 5, "tid": 7, "ts": 1716454222614138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544774, "dur": 5,
    "args": {
      "External id": 80983,
      "cbid": 211, "correlation": 80983
    }
  },
  {
    "ph": "s", "id": 80983, "pid": 76337, "tid": -914061504, "ts": 1716454222544774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222614159, "dur": 33,
    "args": {
      "External id": 80989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 80989, "pid": 5, "tid": 7, "ts": 1716454222614159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544801, "dur": 8,
    "args": {
      "External id": 80989,
      "cbid": 211, "correlation": 80989
    }
  },
  {
    "ph": "s", "id": 80989, "pid": 76337, "tid": -914061504, "ts": 1716454222544801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222614194, "dur": 3,
    "args": {
      "External id": 80997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 80997,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 80997, "pid": 5, "tid": 7, "ts": 1716454222614194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544845, "dur": 10,
    "args": {
      "External id": 80997,
      "cbid": 211, "correlation": 80997
    }
  },
  {
    "ph": "s", "id": 80997, "pid": 76337, "tid": -914061504, "ts": 1716454222544845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544912, "dur": 1,
    "args": {
      "External id": 81013,
      "cbid": 251, "correlation": 81013
    }
  },
  {
    "ph": "f", "id": 81013, "pid": 76337, "tid": -914061504, "ts": 1716454222544912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544917, "dur": 0,
    "args": {
      "External id": 81015,
      "cbid": 251, "correlation": 81015
    }
  },
  {
    "ph": "f", "id": 81015, "pid": 76337, "tid": -914061504, "ts": 1716454222544917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222614198, "dur": 12,
    "args": {
      "External id": 81016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81016,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 81016, "pid": 5, "tid": 7, "ts": 1716454222614198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544919, "dur": 12,
    "args": {
      "External id": 81016,
      "cbid": 211, "correlation": 81016
    }
  },
  {
    "ph": "s", "id": 81016, "pid": 76337, "tid": -914061504, "ts": 1716454222544919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222614212, "dur": 5,
    "args": {
      "External id": 81018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81018,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 81018, "pid": 5, "tid": 7, "ts": 1716454222614212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544933, "dur": 5,
    "args": {
      "External id": 81018,
      "cbid": 211, "correlation": 81018
    }
  },
  {
    "ph": "s", "id": 81018, "pid": 76337, "tid": -914061504, "ts": 1716454222544933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222614219, "dur": 29,
    "args": {
      "External id": 81028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81028, "pid": 5, "tid": 7, "ts": 1716454222614219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222544999, "dur": 12,
    "args": {
      "External id": 81028,
      "cbid": 211, "correlation": 81028
    }
  },
  {
    "ph": "s", "id": 81028, "pid": 76337, "tid": -914061504, "ts": 1716454222544999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222614249, "dur": 30,
    "args": {
      "External id": 81048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81048,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 81048, "pid": 5, "tid": 7, "ts": 1716454222614249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545065, "dur": 11,
    "args": {
      "External id": 81048,
      "cbid": 211, "correlation": 81048
    }
  },
  {
    "ph": "s", "id": 81048, "pid": 76337, "tid": -914061504, "ts": 1716454222545065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222614281, "dur": 4,
    "args": {
      "External id": 81060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81060,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 81060, "pid": 5, "tid": 7, "ts": 1716454222614281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545087, "dur": 6,
    "args": {
      "External id": 81060,
      "cbid": 211, "correlation": 81060
    }
  },
  {
    "ph": "s", "id": 81060, "pid": 76337, "tid": -914061504, "ts": 1716454222545087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222614287, "dur": 31,
    "args": {
      "External id": 81063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81063,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81063, "pid": 5, "tid": 7, "ts": 1716454222614287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545104, "dur": 7,
    "args": {
      "External id": 81063,
      "cbid": 211, "correlation": 81063
    }
  },
  {
    "ph": "s", "id": 81063, "pid": 76337, "tid": -914061504, "ts": 1716454222545104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222614319, "dur": 21,
    "args": {
      "External id": 81072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81072,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81072, "pid": 5, "tid": 7, "ts": 1716454222614319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545146, "dur": 10,
    "args": {
      "External id": 81072,
      "cbid": 211, "correlation": 81072
    }
  },
  {
    "ph": "s", "id": 81072, "pid": 76337, "tid": -914061504, "ts": 1716454222545146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545208, "dur": 0,
    "args": {
      "External id": 81082,
      "cbid": 317, "correlation": 81082
    }
  },
  {
    "ph": "f", "id": 81082, "pid": 76337, "tid": -914061504, "ts": 1716454222545208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545208, "dur": 0,
    "args": {
      "External id": 81083,
      "cbid": 203, "correlation": 81083
    }
  },
  {
    "ph": "f", "id": 81083, "pid": 76337, "tid": -914061504, "ts": 1716454222545208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545209, "dur": 0,
    "args": {
      "External id": 81084,
      "cbid": 205, "correlation": 81084
    }
  },
  {
    "ph": "f", "id": 81084, "pid": 76337, "tid": -914061504, "ts": 1716454222545209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222614341, "dur": 22,
    "args": {
      "External id": 81088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81088, "pid": 5, "tid": 7, "ts": 1716454222614341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545224, "dur": 12,
    "args": {
      "External id": 81088,
      "cbid": 211, "correlation": 81088
    }
  },
  {
    "ph": "s", "id": 81088, "pid": 76337, "tid": -914061504, "ts": 1716454222545224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222614364, "dur": 45,
    "args": {
      "External id": 81090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81090, "pid": 5, "tid": 7, "ts": 1716454222614364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545239, "dur": 6,
    "args": {
      "External id": 81090,
      "cbid": 211, "correlation": 81090
    }
  },
  {
    "ph": "s", "id": 81090, "pid": 76337, "tid": -914061504, "ts": 1716454222545239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222614410, "dur": 653,
    "args": {
      "External id": 81092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81092,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81092, "pid": 5, "tid": 7, "ts": 1716454222614410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545251, "dur": 6,
    "args": {
      "External id": 81092,
      "cbid": 211, "correlation": 81092
    }
  },
  {
    "ph": "s", "id": 81092, "pid": 76337, "tid": -914061504, "ts": 1716454222545251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222615065, "dur": 22,
    "args": {
      "External id": 81094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81094,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81094, "pid": 5, "tid": 7, "ts": 1716454222615065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545261, "dur": 5,
    "args": {
      "External id": 81094,
      "cbid": 211, "correlation": 81094
    }
  },
  {
    "ph": "s", "id": 81094, "pid": 76337, "tid": -914061504, "ts": 1716454222545261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222615088, "dur": 33,
    "args": {
      "External id": 81100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81100, "pid": 5, "tid": 7, "ts": 1716454222615088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545289, "dur": 9,
    "args": {
      "External id": 81100,
      "cbid": 211, "correlation": 81100
    }
  },
  {
    "ph": "s", "id": 81100, "pid": 76337, "tid": -914061504, "ts": 1716454222545289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545347, "dur": 0,
    "args": {
      "External id": 81110,
      "cbid": 317, "correlation": 81110
    }
  },
  {
    "ph": "f", "id": 81110, "pid": 76337, "tid": -914061504, "ts": 1716454222545347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545348, "dur": 0,
    "args": {
      "External id": 81111,
      "cbid": 203, "correlation": 81111
    }
  },
  {
    "ph": "f", "id": 81111, "pid": 76337, "tid": -914061504, "ts": 1716454222545348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545348, "dur": 0,
    "args": {
      "External id": 81112,
      "cbid": 205, "correlation": 81112
    }
  },
  {
    "ph": "f", "id": 81112, "pid": 76337, "tid": -914061504, "ts": 1716454222545348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222615123, "dur": 55,
    "args": {
      "External id": 81116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81116, "pid": 5, "tid": 7, "ts": 1716454222615123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545361, "dur": 12,
    "args": {
      "External id": 81116,
      "cbid": 211, "correlation": 81116
    }
  },
  {
    "ph": "s", "id": 81116, "pid": 76337, "tid": -914061504, "ts": 1716454222545361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222615179, "dur": 274,
    "args": {
      "External id": 81118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81118,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81118, "pid": 5, "tid": 7, "ts": 1716454222615179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545378, "dur": 7,
    "args": {
      "External id": 81118,
      "cbid": 211, "correlation": 81118
    }
  },
  {
    "ph": "s", "id": 81118, "pid": 76337, "tid": -914061504, "ts": 1716454222545378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222615455, "dur": 21,
    "args": {
      "External id": 81120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81120,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81120, "pid": 5, "tid": 7, "ts": 1716454222615455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545390, "dur": 6,
    "args": {
      "External id": 81120,
      "cbid": 211, "correlation": 81120
    }
  },
  {
    "ph": "s", "id": 81120, "pid": 76337, "tid": -914061504, "ts": 1716454222545390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222615478, "dur": 33,
    "args": {
      "External id": 81126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81126,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81126, "pid": 5, "tid": 7, "ts": 1716454222615478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545416, "dur": 9,
    "args": {
      "External id": 81126,
      "cbid": 211, "correlation": 81126
    }
  },
  {
    "ph": "s", "id": 81126, "pid": 76337, "tid": -914061504, "ts": 1716454222545416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222615512, "dur": 27,
    "args": {
      "External id": 81134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81134,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81134, "pid": 5, "tid": 7, "ts": 1716454222615512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545445, "dur": 8,
    "args": {
      "External id": 81134,
      "cbid": 211, "correlation": 81134
    }
  },
  {
    "ph": "s", "id": 81134, "pid": 76337, "tid": -914061504, "ts": 1716454222545445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222615540, "dur": 20,
    "args": {
      "External id": 81142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81142,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81142, "pid": 5, "tid": 7, "ts": 1716454222615540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545473, "dur": 8,
    "args": {
      "External id": 81142,
      "cbid": 211, "correlation": 81142
    }
  },
  {
    "ph": "s", "id": 81142, "pid": 76337, "tid": -914061504, "ts": 1716454222545473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222615562, "dur": 31,
    "args": {
      "External id": 81162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81162,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 81162, "pid": 5, "tid": 7, "ts": 1716454222615562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545555, "dur": 13,
    "args": {
      "External id": 81162,
      "cbid": 211, "correlation": 81162
    }
  },
  {
    "ph": "s", "id": 81162, "pid": 76337, "tid": -914061504, "ts": 1716454222545555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222615594, "dur": 4,
    "args": {
      "External id": 81174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81174,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 81174, "pid": 5, "tid": 7, "ts": 1716454222615594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545578, "dur": 6,
    "args": {
      "External id": 81174,
      "cbid": 211, "correlation": 81174
    }
  },
  {
    "ph": "s", "id": 81174, "pid": 76337, "tid": -914061504, "ts": 1716454222545578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222615599, "dur": 31,
    "args": {
      "External id": 81177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81177, "pid": 5, "tid": 7, "ts": 1716454222615599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545595, "dur": 6,
    "args": {
      "External id": 81177,
      "cbid": 211, "correlation": 81177
    }
  },
  {
    "ph": "s", "id": 81177, "pid": 76337, "tid": -914061504, "ts": 1716454222545595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545652, "dur": 0,
    "args": {
      "External id": 81188,
      "cbid": 317, "correlation": 81188
    }
  },
  {
    "ph": "f", "id": 81188, "pid": 76337, "tid": -914061504, "ts": 1716454222545652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545653, "dur": 0,
    "args": {
      "External id": 81189,
      "cbid": 203, "correlation": 81189
    }
  },
  {
    "ph": "f", "id": 81189, "pid": 76337, "tid": -914061504, "ts": 1716454222545653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545653, "dur": 0,
    "args": {
      "External id": 81190,
      "cbid": 205, "correlation": 81190
    }
  },
  {
    "ph": "f", "id": 81190, "pid": 76337, "tid": -914061504, "ts": 1716454222545653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222615632, "dur": 22,
    "args": {
      "External id": 81194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81194,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81194, "pid": 5, "tid": 7, "ts": 1716454222615632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545668, "dur": 11,
    "args": {
      "External id": 81194,
      "cbid": 211, "correlation": 81194
    }
  },
  {
    "ph": "s", "id": 81194, "pid": 76337, "tid": -914061504, "ts": 1716454222545668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222615655, "dur": 107,
    "args": {
      "External id": 81196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81196,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81196, "pid": 5, "tid": 7, "ts": 1716454222615655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545685, "dur": 6,
    "args": {
      "External id": 81196,
      "cbid": 211, "correlation": 81196
    }
  },
  {
    "ph": "s", "id": 81196, "pid": 76337, "tid": -914061504, "ts": 1716454222545685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222615763, "dur": 23,
    "args": {
      "External id": 81198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81198,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81198, "pid": 5, "tid": 7, "ts": 1716454222615763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545695, "dur": 5,
    "args": {
      "External id": 81198,
      "cbid": 211, "correlation": 81198
    }
  },
  {
    "ph": "s", "id": 81198, "pid": 76337, "tid": -914061504, "ts": 1716454222545695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222615787, "dur": 33,
    "args": {
      "External id": 81204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81204, "pid": 5, "tid": 7, "ts": 1716454222615787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545723, "dur": 8,
    "args": {
      "External id": 81204,
      "cbid": 211, "correlation": 81204
    }
  },
  {
    "ph": "s", "id": 81204, "pid": 76337, "tid": -914061504, "ts": 1716454222545723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222615822, "dur": 188,
    "args": {
      "External id": 81213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81213,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81213, "pid": 5, "tid": 7, "ts": 1716454222615822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545805, "dur": 14,
    "args": {
      "External id": 81213,
      "cbid": 211, "correlation": 81213
    }
  },
  {
    "ph": "s", "id": 81213, "pid": 76337, "tid": -914061504, "ts": 1716454222545805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222616011, "dur": 66,
    "args": {
      "External id": 81235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81235,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81235, "pid": 5, "tid": 7, "ts": 1716454222616011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545862, "dur": 10,
    "args": {
      "External id": 81235,
      "cbid": 211, "correlation": 81235
    }
  },
  {
    "ph": "s", "id": 81235, "pid": 76337, "tid": -914061504, "ts": 1716454222545862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545952, "dur": 1,
    "args": {
      "External id": 81246,
      "cbid": 251, "correlation": 81246
    }
  },
  {
    "ph": "f", "id": 81246, "pid": 76337, "tid": -914061504, "ts": 1716454222545952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222616078, "dur": 157,
    "args": {
      "External id": 81247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81247,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81247, "pid": 5, "tid": 7, "ts": 1716454222616078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222545957, "dur": 13,
    "args": {
      "External id": 81247,
      "cbid": 211, "correlation": 81247
    }
  },
  {
    "ph": "s", "id": 81247, "pid": 76337, "tid": -914061504, "ts": 1716454222545957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546034, "dur": 1,
    "args": {
      "External id": 81258,
      "cbid": 251, "correlation": 81258
    }
  },
  {
    "ph": "f", "id": 81258, "pid": 76337, "tid": -914061504, "ts": 1716454222546034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222616236, "dur": 148,
    "args": {
      "External id": 81259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81259,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81259, "pid": 5, "tid": 7, "ts": 1716454222616236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546039, "dur": 12,
    "args": {
      "External id": 81259,
      "cbid": 211, "correlation": 81259
    }
  },
  {
    "ph": "s", "id": 81259, "pid": 76337, "tid": -914061504, "ts": 1716454222546039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546105, "dur": 1,
    "args": {
      "External id": 81270,
      "cbid": 251, "correlation": 81270
    }
  },
  {
    "ph": "f", "id": 81270, "pid": 76337, "tid": -914061504, "ts": 1716454222546105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222616386, "dur": 146,
    "args": {
      "External id": 81271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81271,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81271, "pid": 5, "tid": 7, "ts": 1716454222616386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546110, "dur": 11,
    "args": {
      "External id": 81271,
      "cbid": 211, "correlation": 81271
    }
  },
  {
    "ph": "s", "id": 81271, "pid": 76337, "tid": -914061504, "ts": 1716454222546110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222616533, "dur": 1991,
    "args": {
      "External id": 81292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81292,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 81292, "pid": 5, "tid": 7, "ts": 1716454222616533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546191, "dur": 13,
    "args": {
      "External id": 81292,
      "cbid": 211, "correlation": 81292
    }
  },
  {
    "ph": "s", "id": 81292, "pid": 76337, "tid": -914061504, "ts": 1716454222546191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546291, "dur": 1,
    "args": {
      "External id": 81310,
      "cbid": 251, "correlation": 81310
    }
  },
  {
    "ph": "f", "id": 81310, "pid": 76337, "tid": -914061504, "ts": 1716454222546291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222618526, "dur": 153,
    "args": {
      "External id": 81312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81312,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 81312, "pid": 5, "tid": 7, "ts": 1716454222618526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546297, "dur": 14,
    "args": {
      "External id": 81312,
      "cbid": 211, "correlation": 81312
    }
  },
  {
    "ph": "s", "id": 81312, "pid": 76337, "tid": -914061504, "ts": 1716454222546297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222618680, "dur": 35,
    "args": {
      "External id": 81320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81320,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81320, "pid": 5, "tid": 7, "ts": 1716454222618680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546369, "dur": 12,
    "args": {
      "External id": 81320,
      "cbid": 211, "correlation": 81320
    }
  },
  {
    "ph": "s", "id": 81320, "pid": 76337, "tid": -914061504, "ts": 1716454222546369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222618717, "dur": 51,
    "args": {
      "External id": 81328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81328,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81328, "pid": 5, "tid": 7, "ts": 1716454222618717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546408, "dur": 9,
    "args": {
      "External id": 81328,
      "cbid": 211, "correlation": 81328
    }
  },
  {
    "ph": "s", "id": 81328, "pid": 76337, "tid": -914061504, "ts": 1716454222546408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222618769, "dur": 30,
    "args": {
      "External id": 81339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81339,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81339, "pid": 5, "tid": 7, "ts": 1716454222618769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546480, "dur": 13,
    "args": {
      "External id": 81339,
      "cbid": 211, "correlation": 81339
    }
  },
  {
    "ph": "s", "id": 81339, "pid": 76337, "tid": -914061504, "ts": 1716454222546480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222618801, "dur": 35,
    "args": {
      "External id": 81361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81361,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81361, "pid": 5, "tid": 7, "ts": 1716454222618801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546511, "dur": 8,
    "args": {
      "External id": 81361,
      "cbid": 211, "correlation": 81361
    }
  },
  {
    "ph": "s", "id": 81361, "pid": 76337, "tid": -914061504, "ts": 1716454222546511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546596, "dur": 1,
    "args": {
      "External id": 81372,
      "cbid": 251, "correlation": 81372
    }
  },
  {
    "ph": "f", "id": 81372, "pid": 76337, "tid": -914061504, "ts": 1716454222546596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222618837, "dur": 92,
    "args": {
      "External id": 81373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81373,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81373, "pid": 5, "tid": 7, "ts": 1716454222618837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546601, "dur": 13,
    "args": {
      "External id": 81373,
      "cbid": 211, "correlation": 81373
    }
  },
  {
    "ph": "s", "id": 81373, "pid": 76337, "tid": -914061504, "ts": 1716454222546601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546671, "dur": 1,
    "args": {
      "External id": 81384,
      "cbid": 251, "correlation": 81384
    }
  },
  {
    "ph": "f", "id": 81384, "pid": 76337, "tid": -914061504, "ts": 1716454222546671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546675, "dur": 0,
    "args": {
      "External id": 81385,
      "cbid": 251, "correlation": 81385
    }
  },
  {
    "ph": "f", "id": 81385, "pid": 76337, "tid": -914061504, "ts": 1716454222546675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222618930, "dur": 12,
    "args": {
      "External id": 81386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81386,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 81386, "pid": 5, "tid": 7, "ts": 1716454222618930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546677, "dur": 12,
    "args": {
      "External id": 81386,
      "cbid": 211, "correlation": 81386
    }
  },
  {
    "ph": "s", "id": 81386, "pid": 76337, "tid": -914061504, "ts": 1716454222546677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222618943, "dur": 5,
    "args": {
      "External id": 81388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81388,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 81388, "pid": 5, "tid": 7, "ts": 1716454222618943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546691, "dur": 6,
    "args": {
      "External id": 81388,
      "cbid": 211, "correlation": 81388
    }
  },
  {
    "ph": "s", "id": 81388, "pid": 76337, "tid": -914061504, "ts": 1716454222546691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546748, "dur": 1,
    "args": {
      "External id": 81399,
      "cbid": 251, "correlation": 81399
    }
  },
  {
    "ph": "f", "id": 81399, "pid": 76337, "tid": -914061504, "ts": 1716454222546748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546751, "dur": 0,
    "args": {
      "External id": 81400,
      "cbid": 251, "correlation": 81400
    }
  },
  {
    "ph": "f", "id": 81400, "pid": 76337, "tid": -914061504, "ts": 1716454222546751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222618949, "dur": 7,
    "args": {
      "External id": 81401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81401,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 81401, "pid": 5, "tid": 7, "ts": 1716454222618949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546753, "dur": 11,
    "args": {
      "External id": 81401,
      "cbid": 211, "correlation": 81401
    }
  },
  {
    "ph": "s", "id": 81401, "pid": 76337, "tid": -914061504, "ts": 1716454222546753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222618958, "dur": 4,
    "args": {
      "External id": 81403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81403,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 81403, "pid": 5, "tid": 7, "ts": 1716454222618958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546765, "dur": 5,
    "args": {
      "External id": 81403,
      "cbid": 211, "correlation": 81403
    }
  },
  {
    "ph": "s", "id": 81403, "pid": 76337, "tid": -914061504, "ts": 1716454222546765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222618963, "dur": 95,
    "args": {
      "External id": 81424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81424,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 81424, "pid": 5, "tid": 7, "ts": 1716454222618963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546839, "dur": 13,
    "args": {
      "External id": 81424,
      "cbid": 211, "correlation": 81424
    }
  },
  {
    "ph": "s", "id": 81424, "pid": 76337, "tid": -914061504, "ts": 1716454222546839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546936, "dur": 1,
    "args": {
      "External id": 81442,
      "cbid": 251, "correlation": 81442
    }
  },
  {
    "ph": "f", "id": 81442, "pid": 76337, "tid": -914061504, "ts": 1716454222546936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222619058, "dur": 101,
    "args": {
      "External id": 81444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81444,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81444, "pid": 5, "tid": 7, "ts": 1716454222619058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222546942, "dur": 13,
    "args": {
      "External id": 81444,
      "cbid": 211, "correlation": 81444
    }
  },
  {
    "ph": "s", "id": 81444, "pid": 76337, "tid": -914061504, "ts": 1716454222546942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222619160, "dur": 20,
    "args": {
      "External id": 81452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81452,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81452, "pid": 5, "tid": 7, "ts": 1716454222619160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547020, "dur": 12,
    "args": {
      "External id": 81452,
      "cbid": 211, "correlation": 81452
    }
  },
  {
    "ph": "s", "id": 81452, "pid": 76337, "tid": -914061504, "ts": 1716454222547020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222619181, "dur": 37,
    "args": {
      "External id": 81460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81460, "pid": 5, "tid": 7, "ts": 1716454222619181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547062, "dur": 9,
    "args": {
      "External id": 81460,
      "cbid": 211, "correlation": 81460
    }
  },
  {
    "ph": "s", "id": 81460, "pid": 76337, "tid": -914061504, "ts": 1716454222547062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222619220, "dur": 35,
    "args": {
      "External id": 81482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81482,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81482, "pid": 5, "tid": 7, "ts": 1716454222619220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547113, "dur": 10,
    "args": {
      "External id": 81482,
      "cbid": 211, "correlation": 81482
    }
  },
  {
    "ph": "s", "id": 81482, "pid": 76337, "tid": -914061504, "ts": 1716454222547113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547204, "dur": 1,
    "args": {
      "External id": 81498,
      "cbid": 251, "correlation": 81498
    }
  },
  {
    "ph": "f", "id": 81498, "pid": 76337, "tid": -914061504, "ts": 1716454222547204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547209, "dur": 0,
    "args": {
      "External id": 81500,
      "cbid": 251, "correlation": 81500
    }
  },
  {
    "ph": "f", "id": 81500, "pid": 76337, "tid": -914061504, "ts": 1716454222547209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222619257, "dur": 551,
    "args": {
      "External id": 81501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81501,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 81501, "pid": 5, "tid": 7, "ts": 1716454222619257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547213, "dur": 13,
    "args": {
      "External id": 81501,
      "cbid": 211, "correlation": 81501
    }
  },
  {
    "ph": "s", "id": 81501, "pid": 76337, "tid": -914061504, "ts": 1716454222547213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222619809, "dur": 128,
    "args": {
      "External id": 81509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81509, "pid": 5, "tid": 7, "ts": 1716454222619809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547278, "dur": 13,
    "args": {
      "External id": 81509,
      "cbid": 211, "correlation": 81509
    }
  },
  {
    "ph": "s", "id": 81509, "pid": 76337, "tid": -914061504, "ts": 1716454222547278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222619938, "dur": 130,
    "args": {
      "External id": 81517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81517, "pid": 5, "tid": 7, "ts": 1716454222619938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547310, "dur": 8,
    "args": {
      "External id": 81517,
      "cbid": 211, "correlation": 81517
    }
  },
  {
    "ph": "s", "id": 81517, "pid": 76337, "tid": -914061504, "ts": 1716454222547310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547386, "dur": 1,
    "args": {
      "External id": 81533,
      "cbid": 251, "correlation": 81533
    }
  },
  {
    "ph": "f", "id": 81533, "pid": 76337, "tid": -914061504, "ts": 1716454222547386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222620070, "dur": 306,
    "args": {
      "External id": 81535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81535,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81535, "pid": 5, "tid": 7, "ts": 1716454222620070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547392, "dur": 12,
    "args": {
      "External id": 81535,
      "cbid": 211, "correlation": 81535
    }
  },
  {
    "ph": "s", "id": 81535, "pid": 76337, "tid": -914061504, "ts": 1716454222547392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222620377, "dur": 27,
    "args": {
      "External id": 81543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81543,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81543, "pid": 5, "tid": 7, "ts": 1716454222620377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547434, "dur": 10,
    "args": {
      "External id": 81543,
      "cbid": 211, "correlation": 81543
    }
  },
  {
    "ph": "s", "id": 81543, "pid": 76337, "tid": -914061504, "ts": 1716454222547434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222620405, "dur": 83,
    "args": {
      "External id": 81554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81554,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81554, "pid": 5, "tid": 7, "ts": 1716454222620405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547502, "dur": 13,
    "args": {
      "External id": 81554,
      "cbid": 211, "correlation": 81554
    }
  },
  {
    "ph": "s", "id": 81554, "pid": 76337, "tid": -914061504, "ts": 1716454222547502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547566, "dur": 0,
    "args": {
      "External id": 81566,
      "cbid": 317, "correlation": 81566
    }
  },
  {
    "ph": "f", "id": 81566, "pid": 76337, "tid": -914061504, "ts": 1716454222547566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547567, "dur": 0,
    "args": {
      "External id": 81567,
      "cbid": 203, "correlation": 81567
    }
  },
  {
    "ph": "f", "id": 81567, "pid": 76337, "tid": -914061504, "ts": 1716454222547567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547568, "dur": 0,
    "args": {
      "External id": 81568,
      "cbid": 205, "correlation": 81568
    }
  },
  {
    "ph": "f", "id": 81568, "pid": 76337, "tid": -914061504, "ts": 1716454222547568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222620490, "dur": 23,
    "args": {
      "External id": 81572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81572,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81572, "pid": 5, "tid": 7, "ts": 1716454222620490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547583, "dur": 12,
    "args": {
      "External id": 81572,
      "cbid": 211, "correlation": 81572
    }
  },
  {
    "ph": "s", "id": 81572, "pid": 76337, "tid": -914061504, "ts": 1716454222547583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222620514, "dur": 122,
    "args": {
      "External id": 81574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81574,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81574, "pid": 5, "tid": 7, "ts": 1716454222620514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547601, "dur": 6,
    "args": {
      "External id": 81574,
      "cbid": 211, "correlation": 81574
    }
  },
  {
    "ph": "s", "id": 81574, "pid": 76337, "tid": -914061504, "ts": 1716454222547601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222620638, "dur": 23,
    "args": {
      "External id": 81576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81576,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81576, "pid": 5, "tid": 7, "ts": 1716454222620638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547611, "dur": 6,
    "args": {
      "External id": 81576,
      "cbid": 211, "correlation": 81576
    }
  },
  {
    "ph": "s", "id": 81576, "pid": 76337, "tid": -914061504, "ts": 1716454222547611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222620663, "dur": 33,
    "args": {
      "External id": 81582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81582, "pid": 5, "tid": 7, "ts": 1716454222620663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547639, "dur": 8,
    "args": {
      "External id": 81582,
      "cbid": 211, "correlation": 81582
    }
  },
  {
    "ph": "s", "id": 81582, "pid": 76337, "tid": -914061504, "ts": 1716454222547639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222620697, "dur": 27,
    "args": {
      "External id": 81590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81590,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81590, "pid": 5, "tid": 7, "ts": 1716454222620697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547671, "dur": 8,
    "args": {
      "External id": 81590,
      "cbid": 211, "correlation": 81590
    }
  },
  {
    "ph": "s", "id": 81590, "pid": 76337, "tid": -914061504, "ts": 1716454222547671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222620725, "dur": 55,
    "args": {
      "External id": 81599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81599,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81599, "pid": 5, "tid": 7, "ts": 1716454222620725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547708, "dur": 11,
    "args": {
      "External id": 81599,
      "cbid": 211, "correlation": 81599
    }
  },
  {
    "ph": "s", "id": 81599, "pid": 76337, "tid": -914061504, "ts": 1716454222547708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222620781, "dur": 52,
    "args": {
      "External id": 81619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81619,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 81619, "pid": 5, "tid": 7, "ts": 1716454222620781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547780, "dur": 12,
    "args": {
      "External id": 81619,
      "cbid": 211, "correlation": 81619
    }
  },
  {
    "ph": "s", "id": 81619, "pid": 76337, "tid": -914061504, "ts": 1716454222547780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222620835, "dur": 5,
    "args": {
      "External id": 81631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81631,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 81631, "pid": 5, "tid": 7, "ts": 1716454222620835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547802, "dur": 6,
    "args": {
      "External id": 81631,
      "cbid": 211, "correlation": 81631
    }
  },
  {
    "ph": "s", "id": 81631, "pid": 76337, "tid": -914061504, "ts": 1716454222547802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222620841, "dur": 57,
    "args": {
      "External id": 81634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81634, "pid": 5, "tid": 7, "ts": 1716454222620841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547821, "dur": 7,
    "args": {
      "External id": 81634,
      "cbid": 211, "correlation": 81634
    }
  },
  {
    "ph": "s", "id": 81634, "pid": 76337, "tid": -914061504, "ts": 1716454222547821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222620899, "dur": 37,
    "args": {
      "External id": 81643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81643,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81643, "pid": 5, "tid": 7, "ts": 1716454222620899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547859, "dur": 10,
    "args": {
      "External id": 81643,
      "cbid": 211, "correlation": 81643
    }
  },
  {
    "ph": "s", "id": 81643, "pid": 76337, "tid": -914061504, "ts": 1716454222547859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547910, "dur": 0,
    "args": {
      "External id": 81653,
      "cbid": 317, "correlation": 81653
    }
  },
  {
    "ph": "f", "id": 81653, "pid": 76337, "tid": -914061504, "ts": 1716454222547910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547911, "dur": 0,
    "args": {
      "External id": 81654,
      "cbid": 203, "correlation": 81654
    }
  },
  {
    "ph": "f", "id": 81654, "pid": 76337, "tid": -914061504, "ts": 1716454222547911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547912, "dur": 0,
    "args": {
      "External id": 81655,
      "cbid": 205, "correlation": 81655
    }
  },
  {
    "ph": "f", "id": 81655, "pid": 76337, "tid": -914061504, "ts": 1716454222547912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222620938, "dur": 39,
    "args": {
      "External id": 81659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81659, "pid": 5, "tid": 7, "ts": 1716454222620938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547927, "dur": 11,
    "args": {
      "External id": 81659,
      "cbid": 211, "correlation": 81659
    }
  },
  {
    "ph": "s", "id": 81659, "pid": 76337, "tid": -914061504, "ts": 1716454222547927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222620978, "dur": 85,
    "args": {
      "External id": 81661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81661, "pid": 5, "tid": 7, "ts": 1716454222620978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547941, "dur": 5,
    "args": {
      "External id": 81661,
      "cbid": 211, "correlation": 81661
    }
  },
  {
    "ph": "s", "id": 81661, "pid": 76337, "tid": -914061504, "ts": 1716454222547941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222621065, "dur": 1295,
    "args": {
      "External id": 81663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81663,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81663, "pid": 5, "tid": 7, "ts": 1716454222621065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547952, "dur": 6,
    "args": {
      "External id": 81663,
      "cbid": 211, "correlation": 81663
    }
  },
  {
    "ph": "s", "id": 81663, "pid": 76337, "tid": -914061504, "ts": 1716454222547952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222622361, "dur": 22,
    "args": {
      "External id": 81665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81665, "pid": 5, "tid": 7, "ts": 1716454222622361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547962, "dur": 5,
    "args": {
      "External id": 81665,
      "cbid": 211, "correlation": 81665
    }
  },
  {
    "ph": "s", "id": 81665, "pid": 76337, "tid": -914061504, "ts": 1716454222547962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222622384, "dur": 33,
    "args": {
      "External id": 81671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81671, "pid": 5, "tid": 7, "ts": 1716454222622384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222547997, "dur": 9,
    "args": {
      "External id": 81671,
      "cbid": 211, "correlation": 81671
    }
  },
  {
    "ph": "s", "id": 81671, "pid": 76337, "tid": -914061504, "ts": 1716454222547997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222622419, "dur": 3,
    "args": {
      "External id": 81679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81679,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 81679, "pid": 5, "tid": 7, "ts": 1716454222622419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548042, "dur": 10,
    "args": {
      "External id": 81679,
      "cbid": 211, "correlation": 81679
    }
  },
  {
    "ph": "s", "id": 81679, "pid": 76337, "tid": -914061504, "ts": 1716454222548042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548106, "dur": 1,
    "args": {
      "External id": 81695,
      "cbid": 251, "correlation": 81695
    }
  },
  {
    "ph": "f", "id": 81695, "pid": 76337, "tid": -914061504, "ts": 1716454222548106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548111, "dur": 0,
    "args": {
      "External id": 81697,
      "cbid": 251, "correlation": 81697
    }
  },
  {
    "ph": "f", "id": 81697, "pid": 76337, "tid": -914061504, "ts": 1716454222548111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222622423, "dur": 12,
    "args": {
      "External id": 81698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81698,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 81698, "pid": 5, "tid": 7, "ts": 1716454222622423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548113, "dur": 11,
    "args": {
      "External id": 81698,
      "cbid": 211, "correlation": 81698
    }
  },
  {
    "ph": "s", "id": 81698, "pid": 76337, "tid": -914061504, "ts": 1716454222548113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222622437, "dur": 5,
    "args": {
      "External id": 81700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81700,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 81700, "pid": 5, "tid": 7, "ts": 1716454222622437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548126, "dur": 6,
    "args": {
      "External id": 81700,
      "cbid": 211, "correlation": 81700
    }
  },
  {
    "ph": "s", "id": 81700, "pid": 76337, "tid": -914061504, "ts": 1716454222548126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222622443, "dur": 30,
    "args": {
      "External id": 81710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81710, "pid": 5, "tid": 7, "ts": 1716454222622443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548185, "dur": 12,
    "args": {
      "External id": 81710,
      "cbid": 211, "correlation": 81710
    }
  },
  {
    "ph": "s", "id": 81710, "pid": 76337, "tid": -914061504, "ts": 1716454222548185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222622474, "dur": 31,
    "args": {
      "External id": 81730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81730,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 81730, "pid": 5, "tid": 7, "ts": 1716454222622474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548251, "dur": 12,
    "args": {
      "External id": 81730,
      "cbid": 211, "correlation": 81730
    }
  },
  {
    "ph": "s", "id": 81730, "pid": 76337, "tid": -914061504, "ts": 1716454222548251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222622507, "dur": 5,
    "args": {
      "External id": 81742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81742,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 81742, "pid": 5, "tid": 7, "ts": 1716454222622507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548273, "dur": 6,
    "args": {
      "External id": 81742,
      "cbid": 211, "correlation": 81742
    }
  },
  {
    "ph": "s", "id": 81742, "pid": 76337, "tid": -914061504, "ts": 1716454222548273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222622512, "dur": 31,
    "args": {
      "External id": 81745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81745, "pid": 5, "tid": 7, "ts": 1716454222622512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548291, "dur": 7,
    "args": {
      "External id": 81745,
      "cbid": 211, "correlation": 81745
    }
  },
  {
    "ph": "s", "id": 81745, "pid": 76337, "tid": -914061504, "ts": 1716454222548291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222622544, "dur": 20,
    "args": {
      "External id": 81754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81754,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81754, "pid": 5, "tid": 7, "ts": 1716454222622544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548331, "dur": 10,
    "args": {
      "External id": 81754,
      "cbid": 211, "correlation": 81754
    }
  },
  {
    "ph": "s", "id": 81754, "pid": 76337, "tid": -914061504, "ts": 1716454222548331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548393, "dur": 0,
    "args": {
      "External id": 81764,
      "cbid": 317, "correlation": 81764
    }
  },
  {
    "ph": "f", "id": 81764, "pid": 76337, "tid": -914061504, "ts": 1716454222548393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548393, "dur": 0,
    "args": {
      "External id": 81765,
      "cbid": 203, "correlation": 81765
    }
  },
  {
    "ph": "f", "id": 81765, "pid": 76337, "tid": -914061504, "ts": 1716454222548393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548394, "dur": 0,
    "args": {
      "External id": 81766,
      "cbid": 205, "correlation": 81766
    }
  },
  {
    "ph": "f", "id": 81766, "pid": 76337, "tid": -914061504, "ts": 1716454222548394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222622566, "dur": 22,
    "args": {
      "External id": 81770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81770, "pid": 5, "tid": 7, "ts": 1716454222622566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548408, "dur": 11,
    "args": {
      "External id": 81770,
      "cbid": 211, "correlation": 81770
    }
  },
  {
    "ph": "s", "id": 81770, "pid": 76337, "tid": -914061504, "ts": 1716454222548408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222622589, "dur": 44,
    "args": {
      "External id": 81772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81772, "pid": 5, "tid": 7, "ts": 1716454222622589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548422, "dur": 5,
    "args": {
      "External id": 81772,
      "cbid": 211, "correlation": 81772
    }
  },
  {
    "ph": "s", "id": 81772, "pid": 76337, "tid": -914061504, "ts": 1716454222548422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222622635, "dur": 654,
    "args": {
      "External id": 81774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81774,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81774, "pid": 5, "tid": 7, "ts": 1716454222622635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548433, "dur": 6,
    "args": {
      "External id": 81774,
      "cbid": 211, "correlation": 81774
    }
  },
  {
    "ph": "s", "id": 81774, "pid": 76337, "tid": -914061504, "ts": 1716454222548433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222623290, "dur": 21,
    "args": {
      "External id": 81776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81776, "pid": 5, "tid": 7, "ts": 1716454222623290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548443, "dur": 6,
    "args": {
      "External id": 81776,
      "cbid": 211, "correlation": 81776
    }
  },
  {
    "ph": "s", "id": 81776, "pid": 76337, "tid": -914061504, "ts": 1716454222548443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222623312, "dur": 33,
    "args": {
      "External id": 81782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81782, "pid": 5, "tid": 7, "ts": 1716454222623312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548471, "dur": 9,
    "args": {
      "External id": 81782,
      "cbid": 211, "correlation": 81782
    }
  },
  {
    "ph": "s", "id": 81782, "pid": 76337, "tid": -914061504, "ts": 1716454222548471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548529, "dur": 0,
    "args": {
      "External id": 81792,
      "cbid": 317, "correlation": 81792
    }
  },
  {
    "ph": "f", "id": 81792, "pid": 76337, "tid": -914061504, "ts": 1716454222548529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548530, "dur": 0,
    "args": {
      "External id": 81793,
      "cbid": 203, "correlation": 81793
    }
  },
  {
    "ph": "f", "id": 81793, "pid": 76337, "tid": -914061504, "ts": 1716454222548530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548531, "dur": 0,
    "args": {
      "External id": 81794,
      "cbid": 205, "correlation": 81794
    }
  },
  {
    "ph": "f", "id": 81794, "pid": 76337, "tid": -914061504, "ts": 1716454222548531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222623347, "dur": 38,
    "args": {
      "External id": 81798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81798,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81798, "pid": 5, "tid": 7, "ts": 1716454222623347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548543, "dur": 12,
    "args": {
      "External id": 81798,
      "cbid": 211, "correlation": 81798
    }
  },
  {
    "ph": "s", "id": 81798, "pid": 76337, "tid": -914061504, "ts": 1716454222548543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222623386, "dur": 193,
    "args": {
      "External id": 81800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81800,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81800, "pid": 5, "tid": 7, "ts": 1716454222623386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548561, "dur": 6,
    "args": {
      "External id": 81800,
      "cbid": 211, "correlation": 81800
    }
  },
  {
    "ph": "s", "id": 81800, "pid": 76337, "tid": -914061504, "ts": 1716454222548561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222623581, "dur": 22,
    "args": {
      "External id": 81802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81802, "pid": 5, "tid": 7, "ts": 1716454222623581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548570, "dur": 5,
    "args": {
      "External id": 81802,
      "cbid": 211, "correlation": 81802
    }
  },
  {
    "ph": "s", "id": 81802, "pid": 76337, "tid": -914061504, "ts": 1716454222548570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222623605, "dur": 33,
    "args": {
      "External id": 81808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81808, "pid": 5, "tid": 7, "ts": 1716454222623605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548595, "dur": 9,
    "args": {
      "External id": 81808,
      "cbid": 211, "correlation": 81808
    }
  },
  {
    "ph": "s", "id": 81808, "pid": 76337, "tid": -914061504, "ts": 1716454222548595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222623640, "dur": 27,
    "args": {
      "External id": 81816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81816,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81816, "pid": 5, "tid": 7, "ts": 1716454222623640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548625, "dur": 8,
    "args": {
      "External id": 81816,
      "cbid": 211, "correlation": 81816
    }
  },
  {
    "ph": "s", "id": 81816, "pid": 76337, "tid": -914061504, "ts": 1716454222548625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222623668, "dur": 19,
    "args": {
      "External id": 81824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81824,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81824, "pid": 5, "tid": 7, "ts": 1716454222623668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548653, "dur": 9,
    "args": {
      "External id": 81824,
      "cbid": 211, "correlation": 81824
    }
  },
  {
    "ph": "s", "id": 81824, "pid": 76337, "tid": -914061504, "ts": 1716454222548653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222623689, "dur": 30,
    "args": {
      "External id": 81844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81844,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 81844, "pid": 5, "tid": 7, "ts": 1716454222623689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548736, "dur": 12,
    "args": {
      "External id": 81844,
      "cbid": 211, "correlation": 81844
    }
  },
  {
    "ph": "s", "id": 81844, "pid": 76337, "tid": -914061504, "ts": 1716454222548736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222623720, "dur": 4,
    "args": {
      "External id": 81856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81856,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 81856, "pid": 5, "tid": 7, "ts": 1716454222623720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548758, "dur": 7,
    "args": {
      "External id": 81856,
      "cbid": 211, "correlation": 81856
    }
  },
  {
    "ph": "s", "id": 81856, "pid": 76337, "tid": -914061504, "ts": 1716454222548758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222623726, "dur": 31,
    "args": {
      "External id": 81859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81859,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81859, "pid": 5, "tid": 7, "ts": 1716454222623726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548776, "dur": 6,
    "args": {
      "External id": 81859,
      "cbid": 211, "correlation": 81859
    }
  },
  {
    "ph": "s", "id": 81859, "pid": 76337, "tid": -914061504, "ts": 1716454222548776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548833, "dur": 0,
    "args": {
      "External id": 81870,
      "cbid": 317, "correlation": 81870
    }
  },
  {
    "ph": "f", "id": 81870, "pid": 76337, "tid": -914061504, "ts": 1716454222548833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548833, "dur": 0,
    "args": {
      "External id": 81871,
      "cbid": 203, "correlation": 81871
    }
  },
  {
    "ph": "f", "id": 81871, "pid": 76337, "tid": -914061504, "ts": 1716454222548833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548834, "dur": 0,
    "args": {
      "External id": 81872,
      "cbid": 205, "correlation": 81872
    }
  },
  {
    "ph": "f", "id": 81872, "pid": 76337, "tid": -914061504, "ts": 1716454222548834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222623758, "dur": 23,
    "args": {
      "External id": 81876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81876,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81876, "pid": 5, "tid": 7, "ts": 1716454222623758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548847, "dur": 12,
    "args": {
      "External id": 81876,
      "cbid": 211, "correlation": 81876
    }
  },
  {
    "ph": "s", "id": 81876, "pid": 76337, "tid": -914061504, "ts": 1716454222548847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222623782, "dur": 108,
    "args": {
      "External id": 81878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81878,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81878, "pid": 5, "tid": 7, "ts": 1716454222623782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548865, "dur": 7,
    "args": {
      "External id": 81878,
      "cbid": 211, "correlation": 81878
    }
  },
  {
    "ph": "s", "id": 81878, "pid": 76337, "tid": -914061504, "ts": 1716454222548865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222623891, "dur": 21,
    "args": {
      "External id": 81880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81880, "pid": 5, "tid": 7, "ts": 1716454222623891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548875, "dur": 5,
    "args": {
      "External id": 81880,
      "cbid": 211, "correlation": 81880
    }
  },
  {
    "ph": "s", "id": 81880, "pid": 76337, "tid": -914061504, "ts": 1716454222548875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222623913, "dur": 33,
    "args": {
      "External id": 81886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81886, "pid": 5, "tid": 7, "ts": 1716454222623913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548903, "dur": 8,
    "args": {
      "External id": 81886,
      "cbid": 211, "correlation": 81886
    }
  },
  {
    "ph": "s", "id": 81886, "pid": 76337, "tid": -914061504, "ts": 1716454222548903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222623947, "dur": 185,
    "args": {
      "External id": 81895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81895,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81895, "pid": 5, "tid": 7, "ts": 1716454222623947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222548992, "dur": 14,
    "args": {
      "External id": 81895,
      "cbid": 211, "correlation": 81895
    }
  },
  {
    "ph": "s", "id": 81895, "pid": 76337, "tid": -914061504, "ts": 1716454222548992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222624134, "dur": 65,
    "args": {
      "External id": 81917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81917,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 81917, "pid": 5, "tid": 7, "ts": 1716454222624134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549050, "dur": 10,
    "args": {
      "External id": 81917,
      "cbid": 211, "correlation": 81917
    }
  },
  {
    "ph": "s", "id": 81917, "pid": 76337, "tid": -914061504, "ts": 1716454222549050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549139, "dur": 1,
    "args": {
      "External id": 81928,
      "cbid": 251, "correlation": 81928
    }
  },
  {
    "ph": "f", "id": 81928, "pid": 76337, "tid": -914061504, "ts": 1716454222549139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222624201, "dur": 158,
    "args": {
      "External id": 81929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81929,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81929, "pid": 5, "tid": 7, "ts": 1716454222624201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549144, "dur": 13,
    "args": {
      "External id": 81929,
      "cbid": 211, "correlation": 81929
    }
  },
  {
    "ph": "s", "id": 81929, "pid": 76337, "tid": -914061504, "ts": 1716454222549144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549214, "dur": 1,
    "args": {
      "External id": 81940,
      "cbid": 251, "correlation": 81940
    }
  },
  {
    "ph": "f", "id": 81940, "pid": 76337, "tid": -914061504, "ts": 1716454222549214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222624360, "dur": 148,
    "args": {
      "External id": 81941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81941,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81941, "pid": 5, "tid": 7, "ts": 1716454222624360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549218, "dur": 12,
    "args": {
      "External id": 81941,
      "cbid": 211, "correlation": 81941
    }
  },
  {
    "ph": "s", "id": 81941, "pid": 76337, "tid": -914061504, "ts": 1716454222549218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549284, "dur": 1,
    "args": {
      "External id": 81952,
      "cbid": 251, "correlation": 81952
    }
  },
  {
    "ph": "f", "id": 81952, "pid": 76337, "tid": -914061504, "ts": 1716454222549284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222624509, "dur": 145,
    "args": {
      "External id": 81953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81953,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 81953, "pid": 5, "tid": 7, "ts": 1716454222624509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549288, "dur": 11,
    "args": {
      "External id": 81953,
      "cbid": 211, "correlation": 81953
    }
  },
  {
    "ph": "s", "id": 81953, "pid": 76337, "tid": -914061504, "ts": 1716454222549288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222624656, "dur": 1988,
    "args": {
      "External id": 81974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81974,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 81974, "pid": 5, "tid": 7, "ts": 1716454222624656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549366, "dur": 12,
    "args": {
      "External id": 81974,
      "cbid": 211, "correlation": 81974
    }
  },
  {
    "ph": "s", "id": 81974, "pid": 76337, "tid": -914061504, "ts": 1716454222549366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549463, "dur": 1,
    "args": {
      "External id": 81992,
      "cbid": 251, "correlation": 81992
    }
  },
  {
    "ph": "f", "id": 81992, "pid": 76337, "tid": -914061504, "ts": 1716454222549463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222626646, "dur": 152,
    "args": {
      "External id": 81994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 81994,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 81994, "pid": 5, "tid": 7, "ts": 1716454222626646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549469, "dur": 13,
    "args": {
      "External id": 81994,
      "cbid": 211, "correlation": 81994
    }
  },
  {
    "ph": "s", "id": 81994, "pid": 76337, "tid": -914061504, "ts": 1716454222549469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222626799, "dur": 36,
    "args": {
      "External id": 82002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82002,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82002, "pid": 5, "tid": 7, "ts": 1716454222626799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549539, "dur": 12,
    "args": {
      "External id": 82002,
      "cbid": 211, "correlation": 82002
    }
  },
  {
    "ph": "s", "id": 82002, "pid": 76337, "tid": -914061504, "ts": 1716454222549539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222626835, "dur": 50,
    "args": {
      "External id": 82010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82010,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82010, "pid": 5, "tid": 7, "ts": 1716454222626835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549577, "dur": 8,
    "args": {
      "External id": 82010,
      "cbid": 211, "correlation": 82010
    }
  },
  {
    "ph": "s", "id": 82010, "pid": 76337, "tid": -914061504, "ts": 1716454222549577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222626887, "dur": 30,
    "args": {
      "External id": 82021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82021,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82021, "pid": 5, "tid": 7, "ts": 1716454222626887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549648, "dur": 12,
    "args": {
      "External id": 82021,
      "cbid": 211, "correlation": 82021
    }
  },
  {
    "ph": "s", "id": 82021, "pid": 76337, "tid": -914061504, "ts": 1716454222549648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222626919, "dur": 35,
    "args": {
      "External id": 82043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82043,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82043, "pid": 5, "tid": 7, "ts": 1716454222626919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549679, "dur": 7,
    "args": {
      "External id": 82043,
      "cbid": 211, "correlation": 82043
    }
  },
  {
    "ph": "s", "id": 82043, "pid": 76337, "tid": -914061504, "ts": 1716454222549679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549763, "dur": 1,
    "args": {
      "External id": 82054,
      "cbid": 251, "correlation": 82054
    }
  },
  {
    "ph": "f", "id": 82054, "pid": 76337, "tid": -914061504, "ts": 1716454222549763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222626956, "dur": 92,
    "args": {
      "External id": 82055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82055,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82055, "pid": 5, "tid": 7, "ts": 1716454222626956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549768, "dur": 13,
    "args": {
      "External id": 82055,
      "cbid": 211, "correlation": 82055
    }
  },
  {
    "ph": "s", "id": 82055, "pid": 76337, "tid": -914061504, "ts": 1716454222549768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549837, "dur": 1,
    "args": {
      "External id": 82066,
      "cbid": 251, "correlation": 82066
    }
  },
  {
    "ph": "f", "id": 82066, "pid": 76337, "tid": -914061504, "ts": 1716454222549837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549840, "dur": 0,
    "args": {
      "External id": 82067,
      "cbid": 251, "correlation": 82067
    }
  },
  {
    "ph": "f", "id": 82067, "pid": 76337, "tid": -914061504, "ts": 1716454222549840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222627049, "dur": 12,
    "args": {
      "External id": 82068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82068,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 82068, "pid": 5, "tid": 7, "ts": 1716454222627049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549842, "dur": 13,
    "args": {
      "External id": 82068,
      "cbid": 211, "correlation": 82068
    }
  },
  {
    "ph": "s", "id": 82068, "pid": 76337, "tid": -914061504, "ts": 1716454222549842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222627062, "dur": 5,
    "args": {
      "External id": 82070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82070,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 82070, "pid": 5, "tid": 7, "ts": 1716454222627062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549857, "dur": 6,
    "args": {
      "External id": 82070,
      "cbid": 211, "correlation": 82070
    }
  },
  {
    "ph": "s", "id": 82070, "pid": 76337, "tid": -914061504, "ts": 1716454222549857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549913, "dur": 1,
    "args": {
      "External id": 82081,
      "cbid": 251, "correlation": 82081
    }
  },
  {
    "ph": "f", "id": 82081, "pid": 76337, "tid": -914061504, "ts": 1716454222549913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549916, "dur": 0,
    "args": {
      "External id": 82082,
      "cbid": 251, "correlation": 82082
    }
  },
  {
    "ph": "f", "id": 82082, "pid": 76337, "tid": -914061504, "ts": 1716454222549916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222627068, "dur": 7,
    "args": {
      "External id": 82083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82083,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 82083, "pid": 5, "tid": 7, "ts": 1716454222627068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549918, "dur": 11,
    "args": {
      "External id": 82083,
      "cbid": 211, "correlation": 82083
    }
  },
  {
    "ph": "s", "id": 82083, "pid": 76337, "tid": -914061504, "ts": 1716454222549918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222627077, "dur": 4,
    "args": {
      "External id": 82085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82085,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 82085, "pid": 5, "tid": 7, "ts": 1716454222627077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222549931, "dur": 5,
    "args": {
      "External id": 82085,
      "cbid": 211, "correlation": 82085
    }
  },
  {
    "ph": "s", "id": 82085, "pid": 76337, "tid": -914061504, "ts": 1716454222549931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222627081, "dur": 94,
    "args": {
      "External id": 82106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82106,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 82106, "pid": 5, "tid": 7, "ts": 1716454222627081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550016, "dur": 13,
    "args": {
      "External id": 82106,
      "cbid": 211, "correlation": 82106
    }
  },
  {
    "ph": "s", "id": 82106, "pid": 76337, "tid": -914061504, "ts": 1716454222550016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550114, "dur": 1,
    "args": {
      "External id": 82124,
      "cbid": 251, "correlation": 82124
    }
  },
  {
    "ph": "f", "id": 82124, "pid": 76337, "tid": -914061504, "ts": 1716454222550114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222627176, "dur": 85,
    "args": {
      "External id": 82126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82126,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82126, "pid": 5, "tid": 7, "ts": 1716454222627176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550120, "dur": 13,
    "args": {
      "External id": 82126,
      "cbid": 211, "correlation": 82126
    }
  },
  {
    "ph": "s", "id": 82126, "pid": 76337, "tid": -914061504, "ts": 1716454222550120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222627263, "dur": 19,
    "args": {
      "External id": 82134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82134,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82134, "pid": 5, "tid": 7, "ts": 1716454222627263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550190, "dur": 12,
    "args": {
      "External id": 82134,
      "cbid": 211, "correlation": 82134
    }
  },
  {
    "ph": "s", "id": 82134, "pid": 76337, "tid": -914061504, "ts": 1716454222550190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222627283, "dur": 40,
    "args": {
      "External id": 82142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82142, "pid": 5, "tid": 7, "ts": 1716454222627283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550232, "dur": 9,
    "args": {
      "External id": 82142,
      "cbid": 211, "correlation": 82142
    }
  },
  {
    "ph": "s", "id": 82142, "pid": 76337, "tid": -914061504, "ts": 1716454222550232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222627324, "dur": 35,
    "args": {
      "External id": 82164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82164,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82164, "pid": 5, "tid": 7, "ts": 1716454222627324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550283, "dur": 10,
    "args": {
      "External id": 82164,
      "cbid": 211, "correlation": 82164
    }
  },
  {
    "ph": "s", "id": 82164, "pid": 76337, "tid": -914061504, "ts": 1716454222550283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550372, "dur": 1,
    "args": {
      "External id": 82180,
      "cbid": 251, "correlation": 82180
    }
  },
  {
    "ph": "f", "id": 82180, "pid": 76337, "tid": -914061504, "ts": 1716454222550372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550377, "dur": 0,
    "args": {
      "External id": 82182,
      "cbid": 251, "correlation": 82182
    }
  },
  {
    "ph": "f", "id": 82182, "pid": 76337, "tid": -914061504, "ts": 1716454222550377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222627361, "dur": 550,
    "args": {
      "External id": 82183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82183,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 82183, "pid": 5, "tid": 7, "ts": 1716454222627361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550381, "dur": 13,
    "args": {
      "External id": 82183,
      "cbid": 211, "correlation": 82183
    }
  },
  {
    "ph": "s", "id": 82183, "pid": 76337, "tid": -914061504, "ts": 1716454222550381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222627912, "dur": 128,
    "args": {
      "External id": 82191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82191, "pid": 5, "tid": 7, "ts": 1716454222627912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550445, "dur": 13,
    "args": {
      "External id": 82191,
      "cbid": 211, "correlation": 82191
    }
  },
  {
    "ph": "s", "id": 82191, "pid": 76337, "tid": -914061504, "ts": 1716454222550445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222628041, "dur": 131,
    "args": {
      "External id": 82199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82199, "pid": 5, "tid": 7, "ts": 1716454222628041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550475, "dur": 9,
    "args": {
      "External id": 82199,
      "cbid": 211, "correlation": 82199
    }
  },
  {
    "ph": "s", "id": 82199, "pid": 76337, "tid": -914061504, "ts": 1716454222550475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550553, "dur": 1,
    "args": {
      "External id": 82215,
      "cbid": 251, "correlation": 82215
    }
  },
  {
    "ph": "f", "id": 82215, "pid": 76337, "tid": -914061504, "ts": 1716454222550553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222628173, "dur": 317,
    "args": {
      "External id": 82217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82217,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82217, "pid": 5, "tid": 7, "ts": 1716454222628173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550558, "dur": 12,
    "args": {
      "External id": 82217,
      "cbid": 211, "correlation": 82217
    }
  },
  {
    "ph": "s", "id": 82217, "pid": 76337, "tid": -914061504, "ts": 1716454222550558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222628492, "dur": 27,
    "args": {
      "External id": 82225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82225,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82225, "pid": 5, "tid": 7, "ts": 1716454222628492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550600, "dur": 10,
    "args": {
      "External id": 82225,
      "cbid": 211, "correlation": 82225
    }
  },
  {
    "ph": "s", "id": 82225, "pid": 76337, "tid": -914061504, "ts": 1716454222550600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222628520, "dur": 83,
    "args": {
      "External id": 82236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82236, "pid": 5, "tid": 7, "ts": 1716454222628520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550668, "dur": 12,
    "args": {
      "External id": 82236,
      "cbid": 211, "correlation": 82236
    }
  },
  {
    "ph": "s", "id": 82236, "pid": 76337, "tid": -914061504, "ts": 1716454222550668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550730, "dur": 0,
    "args": {
      "External id": 82248,
      "cbid": 317, "correlation": 82248
    }
  },
  {
    "ph": "f", "id": 82248, "pid": 76337, "tid": -914061504, "ts": 1716454222550730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550731, "dur": 0,
    "args": {
      "External id": 82249,
      "cbid": 203, "correlation": 82249
    }
  },
  {
    "ph": "f", "id": 82249, "pid": 76337, "tid": -914061504, "ts": 1716454222550731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550732, "dur": 0,
    "args": {
      "External id": 82250,
      "cbid": 205, "correlation": 82250
    }
  },
  {
    "ph": "f", "id": 82250, "pid": 76337, "tid": -914061504, "ts": 1716454222550732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222628604, "dur": 24,
    "args": {
      "External id": 82254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82254,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82254, "pid": 5, "tid": 7, "ts": 1716454222628604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550748, "dur": 12,
    "args": {
      "External id": 82254,
      "cbid": 211, "correlation": 82254
    }
  },
  {
    "ph": "s", "id": 82254, "pid": 76337, "tid": -914061504, "ts": 1716454222550748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222628629, "dur": 124,
    "args": {
      "External id": 82256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82256,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82256, "pid": 5, "tid": 7, "ts": 1716454222628629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550765, "dur": 6,
    "args": {
      "External id": 82256,
      "cbid": 211, "correlation": 82256
    }
  },
  {
    "ph": "s", "id": 82256, "pid": 76337, "tid": -914061504, "ts": 1716454222550765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222628754, "dur": 24,
    "args": {
      "External id": 82258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82258,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82258, "pid": 5, "tid": 7, "ts": 1716454222628754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550776, "dur": 5,
    "args": {
      "External id": 82258,
      "cbid": 211, "correlation": 82258
    }
  },
  {
    "ph": "s", "id": 82258, "pid": 76337, "tid": -914061504, "ts": 1716454222550776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222628780, "dur": 33,
    "args": {
      "External id": 82264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82264,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82264, "pid": 5, "tid": 7, "ts": 1716454222628780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550803, "dur": 8,
    "args": {
      "External id": 82264,
      "cbid": 211, "correlation": 82264
    }
  },
  {
    "ph": "s", "id": 82264, "pid": 76337, "tid": -914061504, "ts": 1716454222550803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222628815, "dur": 26,
    "args": {
      "External id": 82272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82272,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82272, "pid": 5, "tid": 7, "ts": 1716454222628815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550834, "dur": 8,
    "args": {
      "External id": 82272,
      "cbid": 211, "correlation": 82272
    }
  },
  {
    "ph": "s", "id": 82272, "pid": 76337, "tid": -914061504, "ts": 1716454222550834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222628842, "dur": 44,
    "args": {
      "External id": 82281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82281,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82281, "pid": 5, "tid": 7, "ts": 1716454222628842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550872, "dur": 10,
    "args": {
      "External id": 82281,
      "cbid": 211, "correlation": 82281
    }
  },
  {
    "ph": "s", "id": 82281, "pid": 76337, "tid": -914061504, "ts": 1716454222550872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222628887, "dur": 43,
    "args": {
      "External id": 82301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82301,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 82301, "pid": 5, "tid": 7, "ts": 1716454222628887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550942, "dur": 12,
    "args": {
      "External id": 82301,
      "cbid": 211, "correlation": 82301
    }
  },
  {
    "ph": "s", "id": 82301, "pid": 76337, "tid": -914061504, "ts": 1716454222550942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222628932, "dur": 5,
    "args": {
      "External id": 82313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82313,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 82313, "pid": 5, "tid": 7, "ts": 1716454222628932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550964, "dur": 6,
    "args": {
      "External id": 82313,
      "cbid": 211, "correlation": 82313
    }
  },
  {
    "ph": "s", "id": 82313, "pid": 76337, "tid": -914061504, "ts": 1716454222550964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222628938, "dur": 45,
    "args": {
      "External id": 82316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82316, "pid": 5, "tid": 7, "ts": 1716454222628938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222550991, "dur": 8,
    "args": {
      "External id": 82316,
      "cbid": 211, "correlation": 82316
    }
  },
  {
    "ph": "s", "id": 82316, "pid": 76337, "tid": -914061504, "ts": 1716454222550991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222628984, "dur": 31,
    "args": {
      "External id": 82325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82325,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82325, "pid": 5, "tid": 7, "ts": 1716454222628984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551032, "dur": 10,
    "args": {
      "External id": 82325,
      "cbid": 211, "correlation": 82325
    }
  },
  {
    "ph": "s", "id": 82325, "pid": 76337, "tid": -914061504, "ts": 1716454222551032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551085, "dur": 0,
    "args": {
      "External id": 82335,
      "cbid": 317, "correlation": 82335
    }
  },
  {
    "ph": "f", "id": 82335, "pid": 76337, "tid": -914061504, "ts": 1716454222551085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551086, "dur": 0,
    "args": {
      "External id": 82336,
      "cbid": 203, "correlation": 82336
    }
  },
  {
    "ph": "f", "id": 82336, "pid": 76337, "tid": -914061504, "ts": 1716454222551086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551086, "dur": 0,
    "args": {
      "External id": 82337,
      "cbid": 205, "correlation": 82337
    }
  },
  {
    "ph": "f", "id": 82337, "pid": 76337, "tid": -914061504, "ts": 1716454222551086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222629016, "dur": 30,
    "args": {
      "External id": 82341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82341,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82341, "pid": 5, "tid": 7, "ts": 1716454222629016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551101, "dur": 12,
    "args": {
      "External id": 82341,
      "cbid": 211, "correlation": 82341
    }
  },
  {
    "ph": "s", "id": 82341, "pid": 76337, "tid": -914061504, "ts": 1716454222551101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222629048, "dur": 64,
    "args": {
      "External id": 82343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82343, "pid": 5, "tid": 7, "ts": 1716454222629048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551116, "dur": 5,
    "args": {
      "External id": 82343,
      "cbid": 211, "correlation": 82343
    }
  },
  {
    "ph": "s", "id": 82343, "pid": 76337, "tid": -914061504, "ts": 1716454222551116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222629113, "dur": 979,
    "args": {
      "External id": 82345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82345,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82345, "pid": 5, "tid": 7, "ts": 1716454222629113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551127, "dur": 6,
    "args": {
      "External id": 82345,
      "cbid": 211, "correlation": 82345
    }
  },
  {
    "ph": "s", "id": 82345, "pid": 76337, "tid": -914061504, "ts": 1716454222551127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222630094, "dur": 22,
    "args": {
      "External id": 82347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82347,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82347, "pid": 5, "tid": 7, "ts": 1716454222630094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551137, "dur": 5,
    "args": {
      "External id": 82347,
      "cbid": 211, "correlation": 82347
    }
  },
  {
    "ph": "s", "id": 82347, "pid": 76337, "tid": -914061504, "ts": 1716454222551137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222630117, "dur": 34,
    "args": {
      "External id": 82353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82353, "pid": 5, "tid": 7, "ts": 1716454222630117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551164, "dur": 9,
    "args": {
      "External id": 82353,
      "cbid": 211, "correlation": 82353
    }
  },
  {
    "ph": "s", "id": 82353, "pid": 76337, "tid": -914061504, "ts": 1716454222551164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222630152, "dur": 3,
    "args": {
      "External id": 82361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82361,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 82361, "pid": 5, "tid": 7, "ts": 1716454222630152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551209, "dur": 9,
    "args": {
      "External id": 82361,
      "cbid": 211, "correlation": 82361
    }
  },
  {
    "ph": "s", "id": 82361, "pid": 76337, "tid": -914061504, "ts": 1716454222551209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551272, "dur": 1,
    "args": {
      "External id": 82377,
      "cbid": 251, "correlation": 82377
    }
  },
  {
    "ph": "f", "id": 82377, "pid": 76337, "tid": -914061504, "ts": 1716454222551272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551277, "dur": 0,
    "args": {
      "External id": 82379,
      "cbid": 251, "correlation": 82379
    }
  },
  {
    "ph": "f", "id": 82379, "pid": 76337, "tid": -914061504, "ts": 1716454222551277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222630157, "dur": 12,
    "args": {
      "External id": 82380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82380,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 82380, "pid": 5, "tid": 7, "ts": 1716454222630157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551279, "dur": 11,
    "args": {
      "External id": 82380,
      "cbid": 211, "correlation": 82380
    }
  },
  {
    "ph": "s", "id": 82380, "pid": 76337, "tid": -914061504, "ts": 1716454222551279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222630170, "dur": 5,
    "args": {
      "External id": 82382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82382,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 82382, "pid": 5, "tid": 7, "ts": 1716454222630170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551291, "dur": 5,
    "args": {
      "External id": 82382,
      "cbid": 211, "correlation": 82382
    }
  },
  {
    "ph": "s", "id": 82382, "pid": 76337, "tid": -914061504, "ts": 1716454222551291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222630177, "dur": 29,
    "args": {
      "External id": 82392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82392,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82392, "pid": 5, "tid": 7, "ts": 1716454222630177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551349, "dur": 13,
    "args": {
      "External id": 82392,
      "cbid": 211, "correlation": 82392
    }
  },
  {
    "ph": "s", "id": 82392, "pid": 76337, "tid": -914061504, "ts": 1716454222551349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222630208, "dur": 32,
    "args": {
      "External id": 82412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82412,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 82412, "pid": 5, "tid": 7, "ts": 1716454222630208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551417, "dur": 11,
    "args": {
      "External id": 82412,
      "cbid": 211, "correlation": 82412
    }
  },
  {
    "ph": "s", "id": 82412, "pid": 76337, "tid": -914061504, "ts": 1716454222551417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222630241, "dur": 4,
    "args": {
      "External id": 82424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82424,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 82424, "pid": 5, "tid": 7, "ts": 1716454222630241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551437, "dur": 6,
    "args": {
      "External id": 82424,
      "cbid": 211, "correlation": 82424
    }
  },
  {
    "ph": "s", "id": 82424, "pid": 76337, "tid": -914061504, "ts": 1716454222551437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222630246, "dur": 30,
    "args": {
      "External id": 82427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82427,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82427, "pid": 5, "tid": 7, "ts": 1716454222630246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551455, "dur": 7,
    "args": {
      "External id": 82427,
      "cbid": 211, "correlation": 82427
    }
  },
  {
    "ph": "s", "id": 82427, "pid": 76337, "tid": -914061504, "ts": 1716454222551455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222630278, "dur": 21,
    "args": {
      "External id": 82436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82436,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82436, "pid": 5, "tid": 7, "ts": 1716454222630278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551495, "dur": 10,
    "args": {
      "External id": 82436,
      "cbid": 211, "correlation": 82436
    }
  },
  {
    "ph": "s", "id": 82436, "pid": 76337, "tid": -914061504, "ts": 1716454222551495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551558, "dur": 0,
    "args": {
      "External id": 82446,
      "cbid": 317, "correlation": 82446
    }
  },
  {
    "ph": "f", "id": 82446, "pid": 76337, "tid": -914061504, "ts": 1716454222551558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551559, "dur": 0,
    "args": {
      "External id": 82447,
      "cbid": 203, "correlation": 82447
    }
  },
  {
    "ph": "f", "id": 82447, "pid": 76337, "tid": -914061504, "ts": 1716454222551559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551559, "dur": 0,
    "args": {
      "External id": 82448,
      "cbid": 205, "correlation": 82448
    }
  },
  {
    "ph": "f", "id": 82448, "pid": 76337, "tid": -914061504, "ts": 1716454222551559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222630300, "dur": 22,
    "args": {
      "External id": 82452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82452, "pid": 5, "tid": 7, "ts": 1716454222630300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551574, "dur": 12,
    "args": {
      "External id": 82452,
      "cbid": 211, "correlation": 82452
    }
  },
  {
    "ph": "s", "id": 82452, "pid": 76337, "tid": -914061504, "ts": 1716454222551574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222630323, "dur": 45,
    "args": {
      "External id": 82454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82454, "pid": 5, "tid": 7, "ts": 1716454222630323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551588, "dur": 5,
    "args": {
      "External id": 82454,
      "cbid": 211, "correlation": 82454
    }
  },
  {
    "ph": "s", "id": 82454, "pid": 76337, "tid": -914061504, "ts": 1716454222551588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222630369, "dur": 654,
    "args": {
      "External id": 82456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82456,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82456, "pid": 5, "tid": 7, "ts": 1716454222630369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551599, "dur": 6,
    "args": {
      "External id": 82456,
      "cbid": 211, "correlation": 82456
    }
  },
  {
    "ph": "s", "id": 82456, "pid": 76337, "tid": -914061504, "ts": 1716454222551599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631025, "dur": 23,
    "args": {
      "External id": 82458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82458,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82458, "pid": 5, "tid": 7, "ts": 1716454222631025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551609, "dur": 5,
    "args": {
      "External id": 82458,
      "cbid": 211, "correlation": 82458
    }
  },
  {
    "ph": "s", "id": 82458, "pid": 76337, "tid": -914061504, "ts": 1716454222551609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222631049, "dur": 34,
    "args": {
      "External id": 82464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82464, "pid": 5, "tid": 7, "ts": 1716454222631049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551636, "dur": 8,
    "args": {
      "External id": 82464,
      "cbid": 211, "correlation": 82464
    }
  },
  {
    "ph": "s", "id": 82464, "pid": 76337, "tid": -914061504, "ts": 1716454222551636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551694, "dur": 0,
    "args": {
      "External id": 82474,
      "cbid": 317, "correlation": 82474
    }
  },
  {
    "ph": "f", "id": 82474, "pid": 76337, "tid": -914061504, "ts": 1716454222551694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551694, "dur": 0,
    "args": {
      "External id": 82475,
      "cbid": 203, "correlation": 82475
    }
  },
  {
    "ph": "f", "id": 82475, "pid": 76337, "tid": -914061504, "ts": 1716454222551694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551695, "dur": 0,
    "args": {
      "External id": 82476,
      "cbid": 205, "correlation": 82476
    }
  },
  {
    "ph": "f", "id": 82476, "pid": 76337, "tid": -914061504, "ts": 1716454222551695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631085, "dur": 30,
    "args": {
      "External id": 82480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82480,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82480, "pid": 5, "tid": 7, "ts": 1716454222631085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551708, "dur": 11,
    "args": {
      "External id": 82480,
      "cbid": 211, "correlation": 82480
    }
  },
  {
    "ph": "s", "id": 82480, "pid": 76337, "tid": -914061504, "ts": 1716454222551708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222631116, "dur": 154,
    "args": {
      "External id": 82482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82482,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82482, "pid": 5, "tid": 7, "ts": 1716454222631116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551726, "dur": 6,
    "args": {
      "External id": 82482,
      "cbid": 211, "correlation": 82482
    }
  },
  {
    "ph": "s", "id": 82482, "pid": 76337, "tid": -914061504, "ts": 1716454222551726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631272, "dur": 22,
    "args": {
      "External id": 82484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82484,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82484, "pid": 5, "tid": 7, "ts": 1716454222631272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551735, "dur": 5,
    "args": {
      "External id": 82484,
      "cbid": 211, "correlation": 82484
    }
  },
  {
    "ph": "s", "id": 82484, "pid": 76337, "tid": -914061504, "ts": 1716454222551735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222631296, "dur": 33,
    "args": {
      "External id": 82490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82490, "pid": 5, "tid": 7, "ts": 1716454222631296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551761, "dur": 8,
    "args": {
      "External id": 82490,
      "cbid": 211, "correlation": 82490
    }
  },
  {
    "ph": "s", "id": 82490, "pid": 76337, "tid": -914061504, "ts": 1716454222551761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222631330, "dur": 28,
    "args": {
      "External id": 82498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82498,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82498, "pid": 5, "tid": 7, "ts": 1716454222631330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551790, "dur": 7,
    "args": {
      "External id": 82498,
      "cbid": 211, "correlation": 82498
    }
  },
  {
    "ph": "s", "id": 82498, "pid": 76337, "tid": -914061504, "ts": 1716454222551790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222631358, "dur": 19,
    "args": {
      "External id": 82506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82506,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82506, "pid": 5, "tid": 7, "ts": 1716454222631358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551817, "dur": 8,
    "args": {
      "External id": 82506,
      "cbid": 211, "correlation": 82506
    }
  },
  {
    "ph": "s", "id": 82506, "pid": 76337, "tid": -914061504, "ts": 1716454222551817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631379, "dur": 31,
    "args": {
      "External id": 82526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82526,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 82526, "pid": 5, "tid": 7, "ts": 1716454222631379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551900, "dur": 12,
    "args": {
      "External id": 82526,
      "cbid": 211, "correlation": 82526
    }
  },
  {
    "ph": "s", "id": 82526, "pid": 76337, "tid": -914061504, "ts": 1716454222551900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222631411, "dur": 4,
    "args": {
      "External id": 82538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82538,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 82538, "pid": 5, "tid": 7, "ts": 1716454222631411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551922, "dur": 6,
    "args": {
      "External id": 82538,
      "cbid": 211, "correlation": 82538
    }
  },
  {
    "ph": "s", "id": 82538, "pid": 76337, "tid": -914061504, "ts": 1716454222551922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222631416, "dur": 31,
    "args": {
      "External id": 82541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82541, "pid": 5, "tid": 7, "ts": 1716454222631416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222551940, "dur": 6,
    "args": {
      "External id": 82541,
      "cbid": 211, "correlation": 82541
    }
  },
  {
    "ph": "s", "id": 82541, "pid": 76337, "tid": -914061504, "ts": 1716454222551940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552006, "dur": 0,
    "args": {
      "External id": 82552,
      "cbid": 317, "correlation": 82552
    }
  },
  {
    "ph": "f", "id": 82552, "pid": 76337, "tid": -914061504, "ts": 1716454222552006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552007, "dur": 0,
    "args": {
      "External id": 82553,
      "cbid": 203, "correlation": 82553
    }
  },
  {
    "ph": "f", "id": 82553, "pid": 76337, "tid": -914061504, "ts": 1716454222552007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552008, "dur": 0,
    "args": {
      "External id": 82554,
      "cbid": 205, "correlation": 82554
    }
  },
  {
    "ph": "f", "id": 82554, "pid": 76337, "tid": -914061504, "ts": 1716454222552008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631449, "dur": 23,
    "args": {
      "External id": 82558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82558, "pid": 5, "tid": 7, "ts": 1716454222631449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552021, "dur": 12,
    "args": {
      "External id": 82558,
      "cbid": 211, "correlation": 82558
    }
  },
  {
    "ph": "s", "id": 82558, "pid": 76337, "tid": -914061504, "ts": 1716454222552021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222631473, "dur": 107,
    "args": {
      "External id": 82560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82560,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82560, "pid": 5, "tid": 7, "ts": 1716454222631473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552039, "dur": 7,
    "args": {
      "External id": 82560,
      "cbid": 211, "correlation": 82560
    }
  },
  {
    "ph": "s", "id": 82560, "pid": 76337, "tid": -914061504, "ts": 1716454222552039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631581, "dur": 23,
    "args": {
      "External id": 82562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82562,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82562, "pid": 5, "tid": 7, "ts": 1716454222631581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552049, "dur": 5,
    "args": {
      "External id": 82562,
      "cbid": 211, "correlation": 82562
    }
  },
  {
    "ph": "s", "id": 82562, "pid": 76337, "tid": -914061504, "ts": 1716454222552049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222631605, "dur": 33,
    "args": {
      "External id": 82568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82568, "pid": 5, "tid": 7, "ts": 1716454222631605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552077, "dur": 8,
    "args": {
      "External id": 82568,
      "cbid": 211, "correlation": 82568
    }
  },
  {
    "ph": "s", "id": 82568, "pid": 76337, "tid": -914061504, "ts": 1716454222552077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222631639, "dur": 205,
    "args": {
      "External id": 82577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82577,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82577, "pid": 5, "tid": 7, "ts": 1716454222631639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552158, "dur": 14,
    "args": {
      "External id": 82577,
      "cbid": 211, "correlation": 82577
    }
  },
  {
    "ph": "s", "id": 82577, "pid": 76337, "tid": -914061504, "ts": 1716454222552158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222631845, "dur": 66,
    "args": {
      "External id": 82599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82599,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82599, "pid": 5, "tid": 7, "ts": 1716454222631845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552215, "dur": 10,
    "args": {
      "External id": 82599,
      "cbid": 211, "correlation": 82599
    }
  },
  {
    "ph": "s", "id": 82599, "pid": 76337, "tid": -914061504, "ts": 1716454222552215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552305, "dur": 2,
    "args": {
      "External id": 82610,
      "cbid": 251, "correlation": 82610
    }
  },
  {
    "ph": "f", "id": 82610, "pid": 76337, "tid": -914061504, "ts": 1716454222552305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222631912, "dur": 156,
    "args": {
      "External id": 82611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82611,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82611, "pid": 5, "tid": 7, "ts": 1716454222631912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552310, "dur": 13,
    "args": {
      "External id": 82611,
      "cbid": 211, "correlation": 82611
    }
  },
  {
    "ph": "s", "id": 82611, "pid": 76337, "tid": -914061504, "ts": 1716454222552310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552381, "dur": 1,
    "args": {
      "External id": 82622,
      "cbid": 251, "correlation": 82622
    }
  },
  {
    "ph": "f", "id": 82622, "pid": 76337, "tid": -914061504, "ts": 1716454222552381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222632069, "dur": 147,
    "args": {
      "External id": 82623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82623,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82623, "pid": 5, "tid": 7, "ts": 1716454222632069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552385, "dur": 12,
    "args": {
      "External id": 82623,
      "cbid": 211, "correlation": 82623
    }
  },
  {
    "ph": "s", "id": 82623, "pid": 76337, "tid": -914061504, "ts": 1716454222552385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552451, "dur": 1,
    "args": {
      "External id": 82634,
      "cbid": 251, "correlation": 82634
    }
  },
  {
    "ph": "f", "id": 82634, "pid": 76337, "tid": -914061504, "ts": 1716454222552451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222632218, "dur": 147,
    "args": {
      "External id": 82635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82635,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82635, "pid": 5, "tid": 7, "ts": 1716454222632218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552455, "dur": 12,
    "args": {
      "External id": 82635,
      "cbid": 211, "correlation": 82635
    }
  },
  {
    "ph": "s", "id": 82635, "pid": 76337, "tid": -914061504, "ts": 1716454222552455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222632367, "dur": 1989,
    "args": {
      "External id": 82656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82656,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 82656, "pid": 5, "tid": 7, "ts": 1716454222632367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552536, "dur": 12,
    "args": {
      "External id": 82656,
      "cbid": 211, "correlation": 82656
    }
  },
  {
    "ph": "s", "id": 82656, "pid": 76337, "tid": -914061504, "ts": 1716454222552536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552634, "dur": 1,
    "args": {
      "External id": 82674,
      "cbid": 251, "correlation": 82674
    }
  },
  {
    "ph": "f", "id": 82674, "pid": 76337, "tid": -914061504, "ts": 1716454222552634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222634357, "dur": 151,
    "args": {
      "External id": 82676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82676,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 82676, "pid": 5, "tid": 7, "ts": 1716454222634357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552640, "dur": 13,
    "args": {
      "External id": 82676,
      "cbid": 211, "correlation": 82676
    }
  },
  {
    "ph": "s", "id": 82676, "pid": 76337, "tid": -914061504, "ts": 1716454222552640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222634510, "dur": 35,
    "args": {
      "External id": 82684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82684,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82684, "pid": 5, "tid": 7, "ts": 1716454222634510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552709, "dur": 12,
    "args": {
      "External id": 82684,
      "cbid": 211, "correlation": 82684
    }
  },
  {
    "ph": "s", "id": 82684, "pid": 76337, "tid": -914061504, "ts": 1716454222552709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222634547, "dur": 51,
    "args": {
      "External id": 82692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82692,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82692, "pid": 5, "tid": 7, "ts": 1716454222634547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552749, "dur": 9,
    "args": {
      "External id": 82692,
      "cbid": 211, "correlation": 82692
    }
  },
  {
    "ph": "s", "id": 82692, "pid": 76337, "tid": -914061504, "ts": 1716454222552749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222634599, "dur": 31,
    "args": {
      "External id": 82703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82703, "pid": 5, "tid": 7, "ts": 1716454222634599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552820, "dur": 13,
    "args": {
      "External id": 82703,
      "cbid": 211, "correlation": 82703
    }
  },
  {
    "ph": "s", "id": 82703, "pid": 76337, "tid": -914061504, "ts": 1716454222552820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222634632, "dur": 35,
    "args": {
      "External id": 82725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82725,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82725, "pid": 5, "tid": 7, "ts": 1716454222634632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552852, "dur": 7,
    "args": {
      "External id": 82725,
      "cbid": 211, "correlation": 82725
    }
  },
  {
    "ph": "s", "id": 82725, "pid": 76337, "tid": -914061504, "ts": 1716454222552852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552936, "dur": 1,
    "args": {
      "External id": 82736,
      "cbid": 251, "correlation": 82736
    }
  },
  {
    "ph": "f", "id": 82736, "pid": 76337, "tid": -914061504, "ts": 1716454222552936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222634668, "dur": 91,
    "args": {
      "External id": 82737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82737,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82737, "pid": 5, "tid": 7, "ts": 1716454222634668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222552942, "dur": 13,
    "args": {
      "External id": 82737,
      "cbid": 211, "correlation": 82737
    }
  },
  {
    "ph": "s", "id": 82737, "pid": 76337, "tid": -914061504, "ts": 1716454222552942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553019, "dur": 1,
    "args": {
      "External id": 82748,
      "cbid": 251, "correlation": 82748
    }
  },
  {
    "ph": "f", "id": 82748, "pid": 76337, "tid": -914061504, "ts": 1716454222553019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553022, "dur": 0,
    "args": {
      "External id": 82749,
      "cbid": 251, "correlation": 82749
    }
  },
  {
    "ph": "f", "id": 82749, "pid": 76337, "tid": -914061504, "ts": 1716454222553022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222634760, "dur": 11,
    "args": {
      "External id": 82750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82750,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 82750, "pid": 5, "tid": 7, "ts": 1716454222634760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553024, "dur": 12,
    "args": {
      "External id": 82750,
      "cbid": 211, "correlation": 82750
    }
  },
  {
    "ph": "s", "id": 82750, "pid": 76337, "tid": -914061504, "ts": 1716454222553024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222634773, "dur": 5,
    "args": {
      "External id": 82752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82752,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 82752, "pid": 5, "tid": 7, "ts": 1716454222634773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553038, "dur": 6,
    "args": {
      "External id": 82752,
      "cbid": 211, "correlation": 82752
    }
  },
  {
    "ph": "s", "id": 82752, "pid": 76337, "tid": -914061504, "ts": 1716454222553038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553096, "dur": 1,
    "args": {
      "External id": 82763,
      "cbid": 251, "correlation": 82763
    }
  },
  {
    "ph": "f", "id": 82763, "pid": 76337, "tid": -914061504, "ts": 1716454222553096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553100, "dur": 0,
    "args": {
      "External id": 82764,
      "cbid": 251, "correlation": 82764
    }
  },
  {
    "ph": "f", "id": 82764, "pid": 76337, "tid": -914061504, "ts": 1716454222553100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222634779, "dur": 7,
    "args": {
      "External id": 82765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82765,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 82765, "pid": 5, "tid": 7, "ts": 1716454222634779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553102, "dur": 12,
    "args": {
      "External id": 82765,
      "cbid": 211, "correlation": 82765
    }
  },
  {
    "ph": "s", "id": 82765, "pid": 76337, "tid": -914061504, "ts": 1716454222553102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222634788, "dur": 4,
    "args": {
      "External id": 82767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82767,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 82767, "pid": 5, "tid": 7, "ts": 1716454222634788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553115, "dur": 5,
    "args": {
      "External id": 82767,
      "cbid": 211, "correlation": 82767
    }
  },
  {
    "ph": "s", "id": 82767, "pid": 76337, "tid": -914061504, "ts": 1716454222553115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222634792, "dur": 92,
    "args": {
      "External id": 82788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82788,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 82788, "pid": 5, "tid": 7, "ts": 1716454222634792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553189, "dur": 13,
    "args": {
      "External id": 82788,
      "cbid": 211, "correlation": 82788
    }
  },
  {
    "ph": "s", "id": 82788, "pid": 76337, "tid": -914061504, "ts": 1716454222553189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553285, "dur": 1,
    "args": {
      "External id": 82806,
      "cbid": 251, "correlation": 82806
    }
  },
  {
    "ph": "f", "id": 82806, "pid": 76337, "tid": -914061504, "ts": 1716454222553285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222634886, "dur": 99,
    "args": {
      "External id": 82808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82808,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82808, "pid": 5, "tid": 7, "ts": 1716454222634886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553291, "dur": 13,
    "args": {
      "External id": 82808,
      "cbid": 211, "correlation": 82808
    }
  },
  {
    "ph": "s", "id": 82808, "pid": 76337, "tid": -914061504, "ts": 1716454222553291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222634987, "dur": 19,
    "args": {
      "External id": 82816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82816,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82816, "pid": 5, "tid": 7, "ts": 1716454222634987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553360, "dur": 12,
    "args": {
      "External id": 82816,
      "cbid": 211, "correlation": 82816
    }
  },
  {
    "ph": "s", "id": 82816, "pid": 76337, "tid": -914061504, "ts": 1716454222553360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222635007, "dur": 38,
    "args": {
      "External id": 82824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82824, "pid": 5, "tid": 7, "ts": 1716454222635007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553402, "dur": 10,
    "args": {
      "External id": 82824,
      "cbid": 211, "correlation": 82824
    }
  },
  {
    "ph": "s", "id": 82824, "pid": 76337, "tid": -914061504, "ts": 1716454222553402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222635046, "dur": 35,
    "args": {
      "External id": 82846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82846,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82846, "pid": 5, "tid": 7, "ts": 1716454222635046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553454, "dur": 10,
    "args": {
      "External id": 82846,
      "cbid": 211, "correlation": 82846
    }
  },
  {
    "ph": "s", "id": 82846, "pid": 76337, "tid": -914061504, "ts": 1716454222553454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553542, "dur": 1,
    "args": {
      "External id": 82862,
      "cbid": 251, "correlation": 82862
    }
  },
  {
    "ph": "f", "id": 82862, "pid": 76337, "tid": -914061504, "ts": 1716454222553542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553547, "dur": 0,
    "args": {
      "External id": 82864,
      "cbid": 251, "correlation": 82864
    }
  },
  {
    "ph": "f", "id": 82864, "pid": 76337, "tid": -914061504, "ts": 1716454222553547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222635083, "dur": 549,
    "args": {
      "External id": 82865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82865,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 82865, "pid": 5, "tid": 7, "ts": 1716454222635083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553550, "dur": 12,
    "args": {
      "External id": 82865,
      "cbid": 211, "correlation": 82865
    }
  },
  {
    "ph": "s", "id": 82865, "pid": 76337, "tid": -914061504, "ts": 1716454222553550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222635634, "dur": 128,
    "args": {
      "External id": 82873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82873,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82873, "pid": 5, "tid": 7, "ts": 1716454222635634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553615, "dur": 14,
    "args": {
      "External id": 82873,
      "cbid": 211, "correlation": 82873
    }
  },
  {
    "ph": "s", "id": 82873, "pid": 76337, "tid": -914061504, "ts": 1716454222553615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222635763, "dur": 131,
    "args": {
      "External id": 82881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82881, "pid": 5, "tid": 7, "ts": 1716454222635763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553646, "dur": 8,
    "args": {
      "External id": 82881,
      "cbid": 211, "correlation": 82881
    }
  },
  {
    "ph": "s", "id": 82881, "pid": 76337, "tid": -914061504, "ts": 1716454222553646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553722, "dur": 1,
    "args": {
      "External id": 82897,
      "cbid": 251, "correlation": 82897
    }
  },
  {
    "ph": "f", "id": 82897, "pid": 76337, "tid": -914061504, "ts": 1716454222553722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222635895, "dur": 309,
    "args": {
      "External id": 82899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82899,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82899, "pid": 5, "tid": 7, "ts": 1716454222635895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553728, "dur": 13,
    "args": {
      "External id": 82899,
      "cbid": 211, "correlation": 82899
    }
  },
  {
    "ph": "s", "id": 82899, "pid": 76337, "tid": -914061504, "ts": 1716454222553728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222636205, "dur": 27,
    "args": {
      "External id": 82907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82907,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82907, "pid": 5, "tid": 7, "ts": 1716454222636205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553770, "dur": 10,
    "args": {
      "External id": 82907,
      "cbid": 211, "correlation": 82907
    }
  },
  {
    "ph": "s", "id": 82907, "pid": 76337, "tid": -914061504, "ts": 1716454222553770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222636234, "dur": 83,
    "args": {
      "External id": 82918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82918,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82918, "pid": 5, "tid": 7, "ts": 1716454222636234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553837, "dur": 13,
    "args": {
      "External id": 82918,
      "cbid": 211, "correlation": 82918
    }
  },
  {
    "ph": "s", "id": 82918, "pid": 76337, "tid": -914061504, "ts": 1716454222553837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553901, "dur": 0,
    "args": {
      "External id": 82930,
      "cbid": 317, "correlation": 82930
    }
  },
  {
    "ph": "f", "id": 82930, "pid": 76337, "tid": -914061504, "ts": 1716454222553901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553902, "dur": 0,
    "args": {
      "External id": 82931,
      "cbid": 203, "correlation": 82931
    }
  },
  {
    "ph": "f", "id": 82931, "pid": 76337, "tid": -914061504, "ts": 1716454222553902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553903, "dur": 0,
    "args": {
      "External id": 82932,
      "cbid": 205, "correlation": 82932
    }
  },
  {
    "ph": "f", "id": 82932, "pid": 76337, "tid": -914061504, "ts": 1716454222553903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222636318, "dur": 23,
    "args": {
      "External id": 82936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82936,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82936, "pid": 5, "tid": 7, "ts": 1716454222636318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553918, "dur": 11,
    "args": {
      "External id": 82936,
      "cbid": 211, "correlation": 82936
    }
  },
  {
    "ph": "s", "id": 82936, "pid": 76337, "tid": -914061504, "ts": 1716454222553918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222636343, "dur": 123,
    "args": {
      "External id": 82938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82938,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82938, "pid": 5, "tid": 7, "ts": 1716454222636343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553936, "dur": 7,
    "args": {
      "External id": 82938,
      "cbid": 211, "correlation": 82938
    }
  },
  {
    "ph": "s", "id": 82938, "pid": 76337, "tid": -914061504, "ts": 1716454222553936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222636467, "dur": 23,
    "args": {
      "External id": 82940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82940,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82940, "pid": 5, "tid": 7, "ts": 1716454222636467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553947, "dur": 6,
    "args": {
      "External id": 82940,
      "cbid": 211, "correlation": 82940
    }
  },
  {
    "ph": "s", "id": 82940, "pid": 76337, "tid": -914061504, "ts": 1716454222553947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222636492, "dur": 33,
    "args": {
      "External id": 82946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82946, "pid": 5, "tid": 7, "ts": 1716454222636492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222553982, "dur": 9,
    "args": {
      "External id": 82946,
      "cbid": 211, "correlation": 82946
    }
  },
  {
    "ph": "s", "id": 82946, "pid": 76337, "tid": -914061504, "ts": 1716454222553982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222636526, "dur": 27,
    "args": {
      "External id": 82954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82954,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82954, "pid": 5, "tid": 7, "ts": 1716454222636526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554015, "dur": 8,
    "args": {
      "External id": 82954,
      "cbid": 211, "correlation": 82954
    }
  },
  {
    "ph": "s", "id": 82954, "pid": 76337, "tid": -914061504, "ts": 1716454222554015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222636554, "dur": 102,
    "args": {
      "External id": 82965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82965,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82965, "pid": 5, "tid": 7, "ts": 1716454222636554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554078, "dur": 11,
    "args": {
      "External id": 82965,
      "cbid": 211, "correlation": 82965
    }
  },
  {
    "ph": "s", "id": 82965, "pid": 76337, "tid": -914061504, "ts": 1716454222554078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554132, "dur": 0,
    "args": {
      "External id": 82975,
      "cbid": 317, "correlation": 82975
    }
  },
  {
    "ph": "f", "id": 82975, "pid": 76337, "tid": -914061504, "ts": 1716454222554132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554133, "dur": 0,
    "args": {
      "External id": 82976,
      "cbid": 203, "correlation": 82976
    }
  },
  {
    "ph": "f", "id": 82976, "pid": 76337, "tid": -914061504, "ts": 1716454222554133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554133, "dur": 0,
    "args": {
      "External id": 82977,
      "cbid": 205, "correlation": 82977
    }
  },
  {
    "ph": "f", "id": 82977, "pid": 76337, "tid": -914061504, "ts": 1716454222554133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222636658, "dur": 75,
    "args": {
      "External id": 82981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82981, "pid": 5, "tid": 7, "ts": 1716454222636658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554147, "dur": 12,
    "args": {
      "External id": 82981,
      "cbid": 211, "correlation": 82981
    }
  },
  {
    "ph": "s", "id": 82981, "pid": 76337, "tid": -914061504, "ts": 1716454222554147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222636735, "dur": 44,
    "args": {
      "External id": 82983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82983, "pid": 5, "tid": 7, "ts": 1716454222636735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554162, "dur": 5,
    "args": {
      "External id": 82983,
      "cbid": 211, "correlation": 82983
    }
  },
  {
    "ph": "s", "id": 82983, "pid": 76337, "tid": -914061504, "ts": 1716454222554162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222636781, "dur": 4,
    "args": {
      "External id": 82985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82985,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 82985, "pid": 5, "tid": 7, "ts": 1716454222636781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554172, "dur": 6,
    "args": {
      "External id": 82985,
      "cbid": 211, "correlation": 82985
    }
  },
  {
    "ph": "s", "id": 82985, "pid": 76337, "tid": -914061504, "ts": 1716454222554172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554181, "dur": 0,
    "args": {
      "External id": 82986,
      "cbid": 51, "correlation": 82986
    }
  },
  {
    "ph": "s", "id": 82986, "pid": 76337, "tid": -914061504, "ts": 1716454222554181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222636786, "dur": 2226,
    "args": {
      "External id": 82987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82987,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 82987, "pid": 5, "tid": 7, "ts": 1716454222636786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554182, "dur": 5,
    "args": {
      "External id": 82987,
      "cbid": 211, "correlation": 82987
    }
  },
  {
    "ph": "s", "id": 82987, "pid": 76337, "tid": -914061504, "ts": 1716454222554182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222639013, "dur": 113,
    "args": {
      "External id": 82992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 82992,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 82992, "pid": 5, "tid": 7, "ts": 1716454222639013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554210, "dur": 9,
    "args": {
      "External id": 82992,
      "cbid": 211, "correlation": 82992
    }
  },
  {
    "ph": "s", "id": 82992, "pid": 76337, "tid": -914061504, "ts": 1716454222554210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222639128, "dur": 168,
    "args": {
      "External id": 83001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83001,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83001, "pid": 5, "tid": 7, "ts": 1716454222639128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554300, "dur": 13,
    "args": {
      "External id": 83001,
      "cbid": 211, "correlation": 83001
    }
  },
  {
    "ph": "s", "id": 83001, "pid": 76337, "tid": -914061504, "ts": 1716454222554300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222639297, "dur": 132,
    "args": {
      "External id": 83021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83021,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 83021, "pid": 5, "tid": 7, "ts": 1716454222639297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554373, "dur": 11,
    "args": {
      "External id": 83021,
      "cbid": 211, "correlation": 83021
    }
  },
  {
    "ph": "s", "id": 83021, "pid": 76337, "tid": -914061504, "ts": 1716454222554373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222639431, "dur": 4,
    "args": {
      "External id": 83033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83033,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 83033, "pid": 5, "tid": 7, "ts": 1716454222639431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554395, "dur": 7,
    "args": {
      "External id": 83033,
      "cbid": 211, "correlation": 83033
    }
  },
  {
    "ph": "s", "id": 83033, "pid": 76337, "tid": -914061504, "ts": 1716454222554395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222639437, "dur": 162,
    "args": {
      "External id": 83036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83036, "pid": 5, "tid": 7, "ts": 1716454222639437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554413, "dur": 7,
    "args": {
      "External id": 83036,
      "cbid": 211, "correlation": 83036
    }
  },
  {
    "ph": "s", "id": 83036, "pid": 76337, "tid": -914061504, "ts": 1716454222554413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222639600, "dur": 101,
    "args": {
      "External id": 83045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83045,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83045, "pid": 5, "tid": 7, "ts": 1716454222639600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554454, "dur": 10,
    "args": {
      "External id": 83045,
      "cbid": 211, "correlation": 83045
    }
  },
  {
    "ph": "s", "id": 83045, "pid": 76337, "tid": -914061504, "ts": 1716454222554454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554508, "dur": 0,
    "args": {
      "External id": 83055,
      "cbid": 317, "correlation": 83055
    }
  },
  {
    "ph": "f", "id": 83055, "pid": 76337, "tid": -914061504, "ts": 1716454222554508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554508, "dur": 0,
    "args": {
      "External id": 83056,
      "cbid": 203, "correlation": 83056
    }
  },
  {
    "ph": "f", "id": 83056, "pid": 76337, "tid": -914061504, "ts": 1716454222554508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554509, "dur": 0,
    "args": {
      "External id": 83057,
      "cbid": 205, "correlation": 83057
    }
  },
  {
    "ph": "f", "id": 83057, "pid": 76337, "tid": -914061504, "ts": 1716454222554509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222639702, "dur": 112,
    "args": {
      "External id": 83061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83061,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83061, "pid": 5, "tid": 7, "ts": 1716454222639702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554525, "dur": 11,
    "args": {
      "External id": 83061,
      "cbid": 211, "correlation": 83061
    }
  },
  {
    "ph": "s", "id": 83061, "pid": 76337, "tid": -914061504, "ts": 1716454222554525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222639815, "dur": 34,
    "args": {
      "External id": 83063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83063,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83063, "pid": 5, "tid": 7, "ts": 1716454222639815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554539, "dur": 5,
    "args": {
      "External id": 83063,
      "cbid": 211, "correlation": 83063
    }
  },
  {
    "ph": "s", "id": 83063, "pid": 76337, "tid": -914061504, "ts": 1716454222554539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222639851, "dur": 3,
    "args": {
      "External id": 83065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83065, "pid": 5, "tid": 7, "ts": 1716454222639851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554549, "dur": 5,
    "args": {
      "External id": 83065,
      "cbid": 211, "correlation": 83065
    }
  },
  {
    "ph": "s", "id": 83065, "pid": 76337, "tid": -914061504, "ts": 1716454222554549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554558, "dur": 0,
    "args": {
      "External id": 83066,
      "cbid": 51, "correlation": 83066
    }
  },
  {
    "ph": "s", "id": 83066, "pid": 76337, "tid": -914061504, "ts": 1716454222554558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222639856, "dur": 2045,
    "args": {
      "External id": 83067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83067,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83067, "pid": 5, "tid": 7, "ts": 1716454222639856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554558, "dur": 6,
    "args": {
      "External id": 83067,
      "cbid": 211, "correlation": 83067
    }
  },
  {
    "ph": "s", "id": 83067, "pid": 76337, "tid": -914061504, "ts": 1716454222554558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222641902, "dur": 59,
    "args": {
      "External id": 83072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83072, "pid": 5, "tid": 7, "ts": 1716454222641902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554587, "dur": 8,
    "args": {
      "External id": 83072,
      "cbid": 211, "correlation": 83072
    }
  },
  {
    "ph": "s", "id": 83072, "pid": 76337, "tid": -914061504, "ts": 1716454222554587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222641963, "dur": 3,
    "args": {
      "External id": 83080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83080,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83080, "pid": 5, "tid": 7, "ts": 1716454222641963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554631, "dur": 9,
    "args": {
      "External id": 83080,
      "cbid": 211, "correlation": 83080
    }
  },
  {
    "ph": "s", "id": 83080, "pid": 76337, "tid": -914061504, "ts": 1716454222554631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554696, "dur": 1,
    "args": {
      "External id": 83096,
      "cbid": 251, "correlation": 83096
    }
  },
  {
    "ph": "f", "id": 83096, "pid": 76337, "tid": -914061504, "ts": 1716454222554696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554701, "dur": 0,
    "args": {
      "External id": 83098,
      "cbid": 251, "correlation": 83098
    }
  },
  {
    "ph": "f", "id": 83098, "pid": 76337, "tid": -914061504, "ts": 1716454222554701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222641968, "dur": 11,
    "args": {
      "External id": 83099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83099,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 83099, "pid": 5, "tid": 7, "ts": 1716454222641968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554703, "dur": 11,
    "args": {
      "External id": 83099,
      "cbid": 211, "correlation": 83099
    }
  },
  {
    "ph": "s", "id": 83099, "pid": 76337, "tid": -914061504, "ts": 1716454222554703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222641980, "dur": 5,
    "args": {
      "External id": 83101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83101,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 83101, "pid": 5, "tid": 7, "ts": 1716454222641980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554716, "dur": 5,
    "args": {
      "External id": 83101,
      "cbid": 211, "correlation": 83101
    }
  },
  {
    "ph": "s", "id": 83101, "pid": 76337, "tid": -914061504, "ts": 1716454222554716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222641987, "dur": 54,
    "args": {
      "External id": 83111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83111,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83111, "pid": 5, "tid": 7, "ts": 1716454222641987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554773, "dur": 13,
    "args": {
      "External id": 83111,
      "cbid": 211, "correlation": 83111
    }
  },
  {
    "ph": "s", "id": 83111, "pid": 76337, "tid": -914061504, "ts": 1716454222554773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222642042, "dur": 52,
    "args": {
      "External id": 83131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83131,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 83131, "pid": 5, "tid": 7, "ts": 1716454222642042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554840, "dur": 12,
    "args": {
      "External id": 83131,
      "cbid": 211, "correlation": 83131
    }
  },
  {
    "ph": "s", "id": 83131, "pid": 76337, "tid": -914061504, "ts": 1716454222554840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222642096, "dur": 4,
    "args": {
      "External id": 83143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83143,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83143, "pid": 5, "tid": 7, "ts": 1716454222642096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554861, "dur": 6,
    "args": {
      "External id": 83143,
      "cbid": 211, "correlation": 83143
    }
  },
  {
    "ph": "s", "id": 83143, "pid": 76337, "tid": -914061504, "ts": 1716454222554861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222642101, "dur": 57,
    "args": {
      "External id": 83146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83146, "pid": 5, "tid": 7, "ts": 1716454222642101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554879, "dur": 6,
    "args": {
      "External id": 83146,
      "cbid": 211, "correlation": 83146
    }
  },
  {
    "ph": "s", "id": 83146, "pid": 76337, "tid": -914061504, "ts": 1716454222554879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222642159, "dur": 37,
    "args": {
      "External id": 83155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83155,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83155, "pid": 5, "tid": 7, "ts": 1716454222642159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554920, "dur": 10,
    "args": {
      "External id": 83155,
      "cbid": 211, "correlation": 83155
    }
  },
  {
    "ph": "s", "id": 83155, "pid": 76337, "tid": -914061504, "ts": 1716454222554920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554991, "dur": 0,
    "args": {
      "External id": 83165,
      "cbid": 317, "correlation": 83165
    }
  },
  {
    "ph": "f", "id": 83165, "pid": 76337, "tid": -914061504, "ts": 1716454222554991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554992, "dur": 0,
    "args": {
      "External id": 83166,
      "cbid": 203, "correlation": 83166
    }
  },
  {
    "ph": "f", "id": 83166, "pid": 76337, "tid": -914061504, "ts": 1716454222554992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222554993, "dur": 0,
    "args": {
      "External id": 83167,
      "cbid": 205, "correlation": 83167
    }
  },
  {
    "ph": "f", "id": 83167, "pid": 76337, "tid": -914061504, "ts": 1716454222554993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222642198, "dur": 40,
    "args": {
      "External id": 83171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83171,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83171, "pid": 5, "tid": 7, "ts": 1716454222642198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555010, "dur": 13,
    "args": {
      "External id": 83171,
      "cbid": 211, "correlation": 83171
    }
  },
  {
    "ph": "s", "id": 83171, "pid": 76337, "tid": -914061504, "ts": 1716454222555010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222642239, "dur": 14,
    "args": {
      "External id": 83173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83173, "pid": 5, "tid": 7, "ts": 1716454222642239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555025, "dur": 5,
    "args": {
      "External id": 83173,
      "cbid": 211, "correlation": 83173
    }
  },
  {
    "ph": "s", "id": 83173, "pid": 76337, "tid": -914061504, "ts": 1716454222555025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222642255, "dur": 4,
    "args": {
      "External id": 83175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83175, "pid": 5, "tid": 7, "ts": 1716454222642255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555035, "dur": 5,
    "args": {
      "External id": 83175,
      "cbid": 211, "correlation": 83175
    }
  },
  {
    "ph": "s", "id": 83175, "pid": 76337, "tid": -914061504, "ts": 1716454222555035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555043, "dur": 0,
    "args": {
      "External id": 83176,
      "cbid": 51, "correlation": 83176
    }
  },
  {
    "ph": "s", "id": 83176, "pid": 76337, "tid": -914061504, "ts": 1716454222555043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222642260, "dur": 711,
    "args": {
      "External id": 83177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83177,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83177, "pid": 5, "tid": 7, "ts": 1716454222642260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555044, "dur": 5,
    "args": {
      "External id": 83177,
      "cbid": 211, "correlation": 83177
    }
  },
  {
    "ph": "s", "id": 83177, "pid": 76337, "tid": -914061504, "ts": 1716454222555044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222642972, "dur": 59,
    "args": {
      "External id": 83182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83182,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83182, "pid": 5, "tid": 7, "ts": 1716454222642972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555073, "dur": 9,
    "args": {
      "External id": 83182,
      "cbid": 211, "correlation": 83182
    }
  },
  {
    "ph": "s", "id": 83182, "pid": 76337, "tid": -914061504, "ts": 1716454222555073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555130, "dur": 0,
    "args": {
      "External id": 83192,
      "cbid": 317, "correlation": 83192
    }
  },
  {
    "ph": "f", "id": 83192, "pid": 76337, "tid": -914061504, "ts": 1716454222555130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555131, "dur": 0,
    "args": {
      "External id": 83193,
      "cbid": 203, "correlation": 83193
    }
  },
  {
    "ph": "f", "id": 83193, "pid": 76337, "tid": -914061504, "ts": 1716454222555131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555131, "dur": 0,
    "args": {
      "External id": 83194,
      "cbid": 205, "correlation": 83194
    }
  },
  {
    "ph": "f", "id": 83194, "pid": 76337, "tid": -914061504, "ts": 1716454222555131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222643032, "dur": 3,
    "args": {
      "External id": 83198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83198, "pid": 5, "tid": 7, "ts": 1716454222643032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555186, "dur": 12,
    "args": {
      "External id": 83198,
      "cbid": 211, "correlation": 83198
    }
  },
  {
    "ph": "s", "id": 83198, "pid": 76337, "tid": -914061504, "ts": 1716454222555186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555209, "dur": 0,
    "args": {
      "External id": 83199,
      "cbid": 51, "correlation": 83199
    }
  },
  {
    "ph": "s", "id": 83199, "pid": 76337, "tid": -914061504, "ts": 1716454222555209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222643037, "dur": 269,
    "args": {
      "External id": 83200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83200,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83200, "pid": 5, "tid": 7, "ts": 1716454222643037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555210, "dur": 7,
    "args": {
      "External id": 83200,
      "cbid": 211, "correlation": 83200
    }
  },
  {
    "ph": "s", "id": 83200, "pid": 76337, "tid": -914061504, "ts": 1716454222555210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222643307, "dur": 59,
    "args": {
      "External id": 83205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83205, "pid": 5, "tid": 7, "ts": 1716454222643307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555240, "dur": 9,
    "args": {
      "External id": 83205,
      "cbid": 211, "correlation": 83205
    }
  },
  {
    "ph": "s", "id": 83205, "pid": 76337, "tid": -914061504, "ts": 1716454222555240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222643368, "dur": 50,
    "args": {
      "External id": 83213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83213,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83213, "pid": 5, "tid": 7, "ts": 1716454222643368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555270, "dur": 8,
    "args": {
      "External id": 83213,
      "cbid": 211, "correlation": 83213
    }
  },
  {
    "ph": "s", "id": 83213, "pid": 76337, "tid": -914061504, "ts": 1716454222555270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222643419, "dur": 35,
    "args": {
      "External id": 83221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83221,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83221, "pid": 5, "tid": 7, "ts": 1716454222643419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555300, "dur": 8,
    "args": {
      "External id": 83221,
      "cbid": 211, "correlation": 83221
    }
  },
  {
    "ph": "s", "id": 83221, "pid": 76337, "tid": -914061504, "ts": 1716454222555300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222643456, "dur": 52,
    "args": {
      "External id": 83241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83241,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 83241, "pid": 5, "tid": 7, "ts": 1716454222643456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555382, "dur": 12,
    "args": {
      "External id": 83241,
      "cbid": 211, "correlation": 83241
    }
  },
  {
    "ph": "s", "id": 83241, "pid": 76337, "tid": -914061504, "ts": 1716454222555382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222643509, "dur": 4,
    "args": {
      "External id": 83253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83253,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83253, "pid": 5, "tid": 7, "ts": 1716454222643509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555404, "dur": 6,
    "args": {
      "External id": 83253,
      "cbid": 211, "correlation": 83253
    }
  },
  {
    "ph": "s", "id": 83253, "pid": 76337, "tid": -914061504, "ts": 1716454222555404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222643514, "dur": 57,
    "args": {
      "External id": 83256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83256, "pid": 5, "tid": 7, "ts": 1716454222643514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555423, "dur": 7,
    "args": {
      "External id": 83256,
      "cbid": 211, "correlation": 83256
    }
  },
  {
    "ph": "s", "id": 83256, "pid": 76337, "tid": -914061504, "ts": 1716454222555423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555479, "dur": 0,
    "args": {
      "External id": 83267,
      "cbid": 317, "correlation": 83267
    }
  },
  {
    "ph": "f", "id": 83267, "pid": 76337, "tid": -914061504, "ts": 1716454222555479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555480, "dur": 0,
    "args": {
      "External id": 83268,
      "cbid": 203, "correlation": 83268
    }
  },
  {
    "ph": "f", "id": 83268, "pid": 76337, "tid": -914061504, "ts": 1716454222555480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555481, "dur": 0,
    "args": {
      "External id": 83269,
      "cbid": 205, "correlation": 83269
    }
  },
  {
    "ph": "f", "id": 83269, "pid": 76337, "tid": -914061504, "ts": 1716454222555481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555511, "dur": 2,
    "args": {
      "External id": 83273,
      "cbid": 251, "correlation": 83273
    }
  },
  {
    "ph": "f", "id": 83273, "pid": 76337, "tid": -914061504, "ts": 1716454222555511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555514, "dur": 0,
    "args": {
      "External id": 83274,
      "cbid": 251, "correlation": 83274
    }
  },
  {
    "ph": "f", "id": 83274, "pid": 76337, "tid": -914061504, "ts": 1716454222555514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555515, "dur": 1,
    "args": {
      "External id": 83275,
      "cbid": 251, "correlation": 83275
    }
  },
  {
    "ph": "f", "id": 83275, "pid": 76337, "tid": -914061504, "ts": 1716454222555515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555517, "dur": 0,
    "args": {
      "External id": 83276,
      "cbid": 251, "correlation": 83276
    }
  },
  {
    "ph": "f", "id": 83276, "pid": 76337, "tid": -914061504, "ts": 1716454222555517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555519, "dur": 1,
    "args": {
      "External id": 83277,
      "cbid": 251, "correlation": 83277
    }
  },
  {
    "ph": "f", "id": 83277, "pid": 76337, "tid": -914061504, "ts": 1716454222555519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555520, "dur": 1,
    "args": {
      "External id": 83278,
      "cbid": 251, "correlation": 83278
    }
  },
  {
    "ph": "f", "id": 83278, "pid": 76337, "tid": -914061504, "ts": 1716454222555520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555522, "dur": 1,
    "args": {
      "External id": 83279,
      "cbid": 251, "correlation": 83279
    }
  },
  {
    "ph": "f", "id": 83279, "pid": 76337, "tid": -914061504, "ts": 1716454222555522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555524, "dur": 1,
    "args": {
      "External id": 83280,
      "cbid": 251, "correlation": 83280
    }
  },
  {
    "ph": "f", "id": 83280, "pid": 76337, "tid": -914061504, "ts": 1716454222555524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555526, "dur": 0,
    "args": {
      "External id": 83281,
      "cbid": 251, "correlation": 83281
    }
  },
  {
    "ph": "f", "id": 83281, "pid": 76337, "tid": -914061504, "ts": 1716454222555526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222643573, "dur": 116,
    "args": {
      "External id": 83282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83282,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 83282, "pid": 5, "tid": 7, "ts": 1716454222643573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555528, "dur": 13,
    "args": {
      "External id": 83282,
      "cbid": 211, "correlation": 83282
    }
  },
  {
    "ph": "s", "id": 83282, "pid": 76337, "tid": -914061504, "ts": 1716454222555528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222643690, "dur": 60,
    "args": {
      "External id": 83288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83288, "pid": 5, "tid": 7, "ts": 1716454222643690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555565, "dur": 9,
    "args": {
      "External id": 83288,
      "cbid": 211, "correlation": 83288
    }
  },
  {
    "ph": "s", "id": 83288, "pid": 76337, "tid": -914061504, "ts": 1716454222555565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222643752, "dur": 552,
    "args": {
      "External id": 83297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83297,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83297, "pid": 5, "tid": 7, "ts": 1716454222643752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555648, "dur": 14,
    "args": {
      "External id": 83297,
      "cbid": 211, "correlation": 83297
    }
  },
  {
    "ph": "s", "id": 83297, "pid": 76337, "tid": -914061504, "ts": 1716454222555648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222644305, "dur": 184,
    "args": {
      "External id": 83319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83319,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83319, "pid": 5, "tid": 7, "ts": 1716454222644305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555706, "dur": 10,
    "args": {
      "External id": 83319,
      "cbid": 211, "correlation": 83319
    }
  },
  {
    "ph": "s", "id": 83319, "pid": 76337, "tid": -914061504, "ts": 1716454222555706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555796, "dur": 1,
    "args": {
      "External id": 83330,
      "cbid": 251, "correlation": 83330
    }
  },
  {
    "ph": "f", "id": 83330, "pid": 76337, "tid": -914061504, "ts": 1716454222555796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222644491, "dur": 198,
    "args": {
      "External id": 83331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83331,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83331, "pid": 5, "tid": 7, "ts": 1716454222644491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555801, "dur": 13,
    "args": {
      "External id": 83331,
      "cbid": 211, "correlation": 83331
    }
  },
  {
    "ph": "s", "id": 83331, "pid": 76337, "tid": -914061504, "ts": 1716454222555801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555870, "dur": 1,
    "args": {
      "External id": 83342,
      "cbid": 251, "correlation": 83342
    }
  },
  {
    "ph": "f", "id": 83342, "pid": 76337, "tid": -914061504, "ts": 1716454222555870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222644690, "dur": 191,
    "args": {
      "External id": 83343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83343,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83343, "pid": 5, "tid": 7, "ts": 1716454222644690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555875, "dur": 12,
    "args": {
      "External id": 83343,
      "cbid": 211, "correlation": 83343
    }
  },
  {
    "ph": "s", "id": 83343, "pid": 76337, "tid": -914061504, "ts": 1716454222555875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555939, "dur": 1,
    "args": {
      "External id": 83354,
      "cbid": 251, "correlation": 83354
    }
  },
  {
    "ph": "f", "id": 83354, "pid": 76337, "tid": -914061504, "ts": 1716454222555939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222644882, "dur": 190,
    "args": {
      "External id": 83355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83355,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83355, "pid": 5, "tid": 7, "ts": 1716454222644882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222555943, "dur": 11,
    "args": {
      "External id": 83355,
      "cbid": 211, "correlation": 83355
    }
  },
  {
    "ph": "s", "id": 83355, "pid": 76337, "tid": -914061504, "ts": 1716454222555943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222645073, "dur": 18937,
    "args": {
      "External id": 83376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83376,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 83376, "pid": 5, "tid": 7, "ts": 1716454222645073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556033, "dur": 15,
    "args": {
      "External id": 83376,
      "cbid": 211, "correlation": 83376
    }
  },
  {
    "ph": "s", "id": 83376, "pid": 76337, "tid": -914061504, "ts": 1716454222556033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556133, "dur": 1,
    "args": {
      "External id": 83394,
      "cbid": 251, "correlation": 83394
    }
  },
  {
    "ph": "f", "id": 83394, "pid": 76337, "tid": -914061504, "ts": 1716454222556133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222664011, "dur": 206,
    "args": {
      "External id": 83396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83396,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83396, "pid": 5, "tid": 7, "ts": 1716454222664011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556139, "dur": 14,
    "args": {
      "External id": 83396,
      "cbid": 211, "correlation": 83396
    }
  },
  {
    "ph": "s", "id": 83396, "pid": 76337, "tid": -914061504, "ts": 1716454222556139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222664218, "dur": 66,
    "args": {
      "External id": 83404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83404,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83404, "pid": 5, "tid": 7, "ts": 1716454222664218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556211, "dur": 12,
    "args": {
      "External id": 83404,
      "cbid": 211, "correlation": 83404
    }
  },
  {
    "ph": "s", "id": 83404, "pid": 76337, "tid": -914061504, "ts": 1716454222556211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222664286, "dur": 97,
    "args": {
      "External id": 83412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83412,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83412, "pid": 5, "tid": 7, "ts": 1716454222664286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556250, "dur": 9,
    "args": {
      "External id": 83412,
      "cbid": 211, "correlation": 83412
    }
  },
  {
    "ph": "s", "id": 83412, "pid": 76337, "tid": -914061504, "ts": 1716454222556250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222664384, "dur": 55,
    "args": {
      "External id": 83423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83423, "pid": 5, "tid": 7, "ts": 1716454222664384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556323, "dur": 13,
    "args": {
      "External id": 83423,
      "cbid": 211, "correlation": 83423
    }
  },
  {
    "ph": "s", "id": 83423, "pid": 76337, "tid": -914061504, "ts": 1716454222556323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222664440, "dur": 94,
    "args": {
      "External id": 83445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83445,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83445, "pid": 5, "tid": 7, "ts": 1716454222664440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556356, "dur": 7,
    "args": {
      "External id": 83445,
      "cbid": 211, "correlation": 83445
    }
  },
  {
    "ph": "s", "id": 83445, "pid": 76337, "tid": -914061504, "ts": 1716454222556356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556439, "dur": 1,
    "args": {
      "External id": 83456,
      "cbid": 251, "correlation": 83456
    }
  },
  {
    "ph": "f", "id": 83456, "pid": 76337, "tid": -914061504, "ts": 1716454222556439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222664535, "dur": 104,
    "args": {
      "External id": 83457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83457,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83457, "pid": 5, "tid": 7, "ts": 1716454222664535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556444, "dur": 14,
    "args": {
      "External id": 83457,
      "cbid": 211, "correlation": 83457
    }
  },
  {
    "ph": "s", "id": 83457, "pid": 76337, "tid": -914061504, "ts": 1716454222556444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556518, "dur": 1,
    "args": {
      "External id": 83468,
      "cbid": 251, "correlation": 83468
    }
  },
  {
    "ph": "f", "id": 83468, "pid": 76337, "tid": -914061504, "ts": 1716454222556518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556522, "dur": 0,
    "args": {
      "External id": 83469,
      "cbid": 251, "correlation": 83469
    }
  },
  {
    "ph": "f", "id": 83469, "pid": 76337, "tid": -914061504, "ts": 1716454222556522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222664641, "dur": 10,
    "args": {
      "External id": 83470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83470,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 83470, "pid": 5, "tid": 7, "ts": 1716454222664641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556524, "dur": 13,
    "args": {
      "External id": 83470,
      "cbid": 211, "correlation": 83470
    }
  },
  {
    "ph": "s", "id": 83470, "pid": 76337, "tid": -914061504, "ts": 1716454222556524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222664652, "dur": 5,
    "args": {
      "External id": 83472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83472,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 83472, "pid": 5, "tid": 7, "ts": 1716454222664652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556540, "dur": 7,
    "args": {
      "External id": 83472,
      "cbid": 211, "correlation": 83472
    }
  },
  {
    "ph": "s", "id": 83472, "pid": 76337, "tid": -914061504, "ts": 1716454222556540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556602, "dur": 1,
    "args": {
      "External id": 83483,
      "cbid": 251, "correlation": 83483
    }
  },
  {
    "ph": "f", "id": 83483, "pid": 76337, "tid": -914061504, "ts": 1716454222556602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556605, "dur": 0,
    "args": {
      "External id": 83484,
      "cbid": 251, "correlation": 83484
    }
  },
  {
    "ph": "f", "id": 83484, "pid": 76337, "tid": -914061504, "ts": 1716454222556605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222664659, "dur": 6,
    "args": {
      "External id": 83485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83485,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 83485, "pid": 5, "tid": 7, "ts": 1716454222664659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556607, "dur": 12,
    "args": {
      "External id": 83485,
      "cbid": 211, "correlation": 83485
    }
  },
  {
    "ph": "s", "id": 83485, "pid": 76337, "tid": -914061504, "ts": 1716454222556607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222664666, "dur": 4,
    "args": {
      "External id": 83487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83487,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 83487, "pid": 5, "tid": 7, "ts": 1716454222664666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556620, "dur": 5,
    "args": {
      "External id": 83487,
      "cbid": 211, "correlation": 83487
    }
  },
  {
    "ph": "s", "id": 83487, "pid": 76337, "tid": -914061504, "ts": 1716454222556620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222664671, "dur": 157,
    "args": {
      "External id": 83508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83508,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 83508, "pid": 5, "tid": 7, "ts": 1716454222664671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556693, "dur": 12,
    "args": {
      "External id": 83508,
      "cbid": 211, "correlation": 83508
    }
  },
  {
    "ph": "s", "id": 83508, "pid": 76337, "tid": -914061504, "ts": 1716454222556693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556790, "dur": 1,
    "args": {
      "External id": 83526,
      "cbid": 251, "correlation": 83526
    }
  },
  {
    "ph": "f", "id": 83526, "pid": 76337, "tid": -914061504, "ts": 1716454222556790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222664830, "dur": 109,
    "args": {
      "External id": 83528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83528,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 83528, "pid": 5, "tid": 7, "ts": 1716454222664830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556796, "dur": 13,
    "args": {
      "External id": 83528,
      "cbid": 211, "correlation": 83528
    }
  },
  {
    "ph": "s", "id": 83528, "pid": 76337, "tid": -914061504, "ts": 1716454222556796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222664941, "dur": 35,
    "args": {
      "External id": 83536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83536,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83536, "pid": 5, "tid": 7, "ts": 1716454222664941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556865, "dur": 12,
    "args": {
      "External id": 83536,
      "cbid": 211, "correlation": 83536
    }
  },
  {
    "ph": "s", "id": 83536, "pid": 76337, "tid": -914061504, "ts": 1716454222556865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222664977, "dur": 67,
    "args": {
      "External id": 83544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83544,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83544, "pid": 5, "tid": 7, "ts": 1716454222664977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556906, "dur": 9,
    "args": {
      "External id": 83544,
      "cbid": 211, "correlation": 83544
    }
  },
  {
    "ph": "s", "id": 83544, "pid": 76337, "tid": -914061504, "ts": 1716454222556906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222665045, "dur": 94,
    "args": {
      "External id": 83566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83566,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83566, "pid": 5, "tid": 7, "ts": 1716454222665045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222556958, "dur": 10,
    "args": {
      "External id": 83566,
      "cbid": 211, "correlation": 83566
    }
  },
  {
    "ph": "s", "id": 83566, "pid": 76337, "tid": -914061504, "ts": 1716454222556958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557055, "dur": 1,
    "args": {
      "External id": 83582,
      "cbid": 251, "correlation": 83582
    }
  },
  {
    "ph": "f", "id": 83582, "pid": 76337, "tid": -914061504, "ts": 1716454222557055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222665141, "dur": 583,
    "args": {
      "External id": 83584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83584,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83584, "pid": 5, "tid": 7, "ts": 1716454222665141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557061, "dur": 13,
    "args": {
      "External id": 83584,
      "cbid": 211, "correlation": 83584
    }
  },
  {
    "ph": "s", "id": 83584, "pid": 76337, "tid": -914061504, "ts": 1716454222557061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222665725, "dur": 245,
    "args": {
      "External id": 83592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83592, "pid": 5, "tid": 7, "ts": 1716454222665725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557127, "dur": 14,
    "args": {
      "External id": 83592,
      "cbid": 211, "correlation": 83592
    }
  },
  {
    "ph": "s", "id": 83592, "pid": 76337, "tid": -914061504, "ts": 1716454222557127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222665972, "dur": 255,
    "args": {
      "External id": 83600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83600, "pid": 5, "tid": 7, "ts": 1716454222665972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557159, "dur": 8,
    "args": {
      "External id": 83600,
      "cbid": 211, "correlation": 83600
    }
  },
  {
    "ph": "s", "id": 83600, "pid": 76337, "tid": -914061504, "ts": 1716454222557159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557242, "dur": 2,
    "args": {
      "External id": 83616,
      "cbid": 251, "correlation": 83616
    }
  },
  {
    "ph": "f", "id": 83616, "pid": 76337, "tid": -914061504, "ts": 1716454222557242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557248, "dur": 0,
    "args": {
      "External id": 83618,
      "cbid": 251, "correlation": 83618
    }
  },
  {
    "ph": "f", "id": 83618, "pid": 76337, "tid": -914061504, "ts": 1716454222557248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222666229, "dur": 359,
    "args": {
      "External id": 83619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83619,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 83619, "pid": 5, "tid": 7, "ts": 1716454222666229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557250, "dur": 13,
    "args": {
      "External id": 83619,
      "cbid": 211, "correlation": 83619
    }
  },
  {
    "ph": "s", "id": 83619, "pid": 76337, "tid": -914061504, "ts": 1716454222557250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222666589, "dur": 50,
    "args": {
      "External id": 83627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83627,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83627, "pid": 5, "tid": 7, "ts": 1716454222666589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557292, "dur": 10,
    "args": {
      "External id": 83627,
      "cbid": 211, "correlation": 83627
    }
  },
  {
    "ph": "s", "id": 83627, "pid": 76337, "tid": -914061504, "ts": 1716454222557292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222666641, "dur": 161,
    "args": {
      "External id": 83638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83638, "pid": 5, "tid": 7, "ts": 1716454222666641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557360, "dur": 12,
    "args": {
      "External id": 83638,
      "cbid": 211, "correlation": 83638
    }
  },
  {
    "ph": "s", "id": 83638, "pid": 76337, "tid": -914061504, "ts": 1716454222557360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557424, "dur": 0,
    "args": {
      "External id": 83650,
      "cbid": 317, "correlation": 83650
    }
  },
  {
    "ph": "f", "id": 83650, "pid": 76337, "tid": -914061504, "ts": 1716454222557424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557424, "dur": 0,
    "args": {
      "External id": 83651,
      "cbid": 203, "correlation": 83651
    }
  },
  {
    "ph": "f", "id": 83651, "pid": 76337, "tid": -914061504, "ts": 1716454222557424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557425, "dur": 0,
    "args": {
      "External id": 83652,
      "cbid": 205, "correlation": 83652
    }
  },
  {
    "ph": "f", "id": 83652, "pid": 76337, "tid": -914061504, "ts": 1716454222557425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557449, "dur": 1,
    "args": {
      "External id": 83656,
      "cbid": 251, "correlation": 83656
    }
  },
  {
    "ph": "f", "id": 83656, "pid": 76337, "tid": -914061504, "ts": 1716454222557449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557451, "dur": 0,
    "args": {
      "External id": 83657,
      "cbid": 251, "correlation": 83657
    }
  },
  {
    "ph": "f", "id": 83657, "pid": 76337, "tid": -914061504, "ts": 1716454222557451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557452, "dur": 0,
    "args": {
      "External id": 83658,
      "cbid": 251, "correlation": 83658
    }
  },
  {
    "ph": "f", "id": 83658, "pid": 76337, "tid": -914061504, "ts": 1716454222557452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557453, "dur": 0,
    "args": {
      "External id": 83659,
      "cbid": 251, "correlation": 83659
    }
  },
  {
    "ph": "f", "id": 83659, "pid": 76337, "tid": -914061504, "ts": 1716454222557453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557454, "dur": 1,
    "args": {
      "External id": 83660,
      "cbid": 251, "correlation": 83660
    }
  },
  {
    "ph": "f", "id": 83660, "pid": 76337, "tid": -914061504, "ts": 1716454222557454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557455, "dur": 0,
    "args": {
      "External id": 83661,
      "cbid": 251, "correlation": 83661
    }
  },
  {
    "ph": "f", "id": 83661, "pid": 76337, "tid": -914061504, "ts": 1716454222557455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557456, "dur": 0,
    "args": {
      "External id": 83662,
      "cbid": 251, "correlation": 83662
    }
  },
  {
    "ph": "f", "id": 83662, "pid": 76337, "tid": -914061504, "ts": 1716454222557456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557457, "dur": 0,
    "args": {
      "External id": 83663,
      "cbid": 251, "correlation": 83663
    }
  },
  {
    "ph": "f", "id": 83663, "pid": 76337, "tid": -914061504, "ts": 1716454222557457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557458, "dur": 0,
    "args": {
      "External id": 83664,
      "cbid": 251, "correlation": 83664
    }
  },
  {
    "ph": "f", "id": 83664, "pid": 76337, "tid": -914061504, "ts": 1716454222557458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222666804, "dur": 115,
    "args": {
      "External id": 83665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83665,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 83665, "pid": 5, "tid": 7, "ts": 1716454222666804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557460, "dur": 12,
    "args": {
      "External id": 83665,
      "cbid": 211, "correlation": 83665
    }
  },
  {
    "ph": "s", "id": 83665, "pid": 76337, "tid": -914061504, "ts": 1716454222557460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222666920, "dur": 60,
    "args": {
      "External id": 83671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83671, "pid": 5, "tid": 7, "ts": 1716454222666920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557495, "dur": 9,
    "args": {
      "External id": 83671,
      "cbid": 211, "correlation": 83671
    }
  },
  {
    "ph": "s", "id": 83671, "pid": 76337, "tid": -914061504, "ts": 1716454222557495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222666981, "dur": 50,
    "args": {
      "External id": 83679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83679,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83679, "pid": 5, "tid": 7, "ts": 1716454222666981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557526, "dur": 8,
    "args": {
      "External id": 83679,
      "cbid": 211, "correlation": 83679
    }
  },
  {
    "ph": "s", "id": 83679, "pid": 76337, "tid": -914061504, "ts": 1716454222557526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222667032, "dur": 99,
    "args": {
      "External id": 83688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83688,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83688, "pid": 5, "tid": 7, "ts": 1716454222667032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557565, "dur": 10,
    "args": {
      "External id": 83688,
      "cbid": 211, "correlation": 83688
    }
  },
  {
    "ph": "s", "id": 83688, "pid": 76337, "tid": -914061504, "ts": 1716454222557565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222667132, "dur": 92,
    "args": {
      "External id": 83708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83708,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 83708, "pid": 5, "tid": 7, "ts": 1716454222667132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557637, "dur": 11,
    "args": {
      "External id": 83708,
      "cbid": 211, "correlation": 83708
    }
  },
  {
    "ph": "s", "id": 83708, "pid": 76337, "tid": -914061504, "ts": 1716454222557637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222667226, "dur": 5,
    "args": {
      "External id": 83720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83720,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 83720, "pid": 5, "tid": 7, "ts": 1716454222667226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557659, "dur": 23,
    "args": {
      "External id": 83720,
      "cbid": 211, "correlation": 83720
    }
  },
  {
    "ph": "s", "id": 83720, "pid": 76337, "tid": -914061504, "ts": 1716454222557659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222667232, "dur": 109,
    "args": {
      "External id": 83723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83723,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83723, "pid": 5, "tid": 7, "ts": 1716454222667232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557695, "dur": 7,
    "args": {
      "External id": 83723,
      "cbid": 211, "correlation": 83723
    }
  },
  {
    "ph": "s", "id": 83723, "pid": 76337, "tid": -914061504, "ts": 1716454222557695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222667343, "dur": 69,
    "args": {
      "External id": 83732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83732,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83732, "pid": 5, "tid": 7, "ts": 1716454222667343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557735, "dur": 10,
    "args": {
      "External id": 83732,
      "cbid": 211, "correlation": 83732
    }
  },
  {
    "ph": "s", "id": 83732, "pid": 76337, "tid": -914061504, "ts": 1716454222557735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557787, "dur": 0,
    "args": {
      "External id": 83742,
      "cbid": 317, "correlation": 83742
    }
  },
  {
    "ph": "f", "id": 83742, "pid": 76337, "tid": -914061504, "ts": 1716454222557787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557788, "dur": 0,
    "args": {
      "External id": 83743,
      "cbid": 203, "correlation": 83743
    }
  },
  {
    "ph": "f", "id": 83743, "pid": 76337, "tid": -914061504, "ts": 1716454222557788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557788, "dur": 0,
    "args": {
      "External id": 83744,
      "cbid": 205, "correlation": 83744
    }
  },
  {
    "ph": "f", "id": 83744, "pid": 76337, "tid": -914061504, "ts": 1716454222557788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222667414, "dur": 77,
    "args": {
      "External id": 83748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83748,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83748, "pid": 5, "tid": 7, "ts": 1716454222667414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557803, "dur": 12,
    "args": {
      "External id": 83748,
      "cbid": 211, "correlation": 83748
    }
  },
  {
    "ph": "s", "id": 83748, "pid": 76337, "tid": -914061504, "ts": 1716454222557803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222667492, "dur": 24,
    "args": {
      "External id": 83750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83750,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83750, "pid": 5, "tid": 7, "ts": 1716454222667492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557817, "dur": 5,
    "args": {
      "External id": 83750,
      "cbid": 211, "correlation": 83750
    }
  },
  {
    "ph": "s", "id": 83750, "pid": 76337, "tid": -914061504, "ts": 1716454222557817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222667518, "dur": 4,
    "args": {
      "External id": 83752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83752, "pid": 5, "tid": 7, "ts": 1716454222667518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557829, "dur": 6,
    "args": {
      "External id": 83752,
      "cbid": 211, "correlation": 83752
    }
  },
  {
    "ph": "s", "id": 83752, "pid": 76337, "tid": -914061504, "ts": 1716454222557829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557837, "dur": 0,
    "args": {
      "External id": 83753,
      "cbid": 51, "correlation": 83753
    }
  },
  {
    "ph": "s", "id": 83753, "pid": 76337, "tid": -914061504, "ts": 1716454222557837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222667523, "dur": 1385,
    "args": {
      "External id": 83754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83754,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83754, "pid": 5, "tid": 7, "ts": 1716454222667523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557838, "dur": 5,
    "args": {
      "External id": 83754,
      "cbid": 211, "correlation": 83754
    }
  },
  {
    "ph": "s", "id": 83754, "pid": 76337, "tid": -914061504, "ts": 1716454222557838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222668910, "dur": 60,
    "args": {
      "External id": 83759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83759, "pid": 5, "tid": 7, "ts": 1716454222668910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557865, "dur": 9,
    "args": {
      "External id": 83759,
      "cbid": 211, "correlation": 83759
    }
  },
  {
    "ph": "s", "id": 83759, "pid": 76337, "tid": -914061504, "ts": 1716454222557865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222668971, "dur": 3,
    "args": {
      "External id": 83767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83767,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83767, "pid": 5, "tid": 7, "ts": 1716454222668971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557909, "dur": 9,
    "args": {
      "External id": 83767,
      "cbid": 211, "correlation": 83767
    }
  },
  {
    "ph": "s", "id": 83767, "pid": 76337, "tid": -914061504, "ts": 1716454222557909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557982, "dur": 2,
    "args": {
      "External id": 83783,
      "cbid": 251, "correlation": 83783
    }
  },
  {
    "ph": "f", "id": 83783, "pid": 76337, "tid": -914061504, "ts": 1716454222557982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557988, "dur": 0,
    "args": {
      "External id": 83785,
      "cbid": 251, "correlation": 83785
    }
  },
  {
    "ph": "f", "id": 83785, "pid": 76337, "tid": -914061504, "ts": 1716454222557988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222668976, "dur": 12,
    "args": {
      "External id": 83786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83786,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 83786, "pid": 5, "tid": 7, "ts": 1716454222668976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222557990, "dur": 12,
    "args": {
      "External id": 83786,
      "cbid": 211, "correlation": 83786
    }
  },
  {
    "ph": "s", "id": 83786, "pid": 76337, "tid": -914061504, "ts": 1716454222557990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222668989, "dur": 5,
    "args": {
      "External id": 83788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83788,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 83788, "pid": 5, "tid": 7, "ts": 1716454222668989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558004, "dur": 176,
    "args": {
      "External id": 83788,
      "cbid": 211, "correlation": 83788
    }
  },
  {
    "ph": "s", "id": 83788, "pid": 76337, "tid": -914061504, "ts": 1716454222558004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222668996, "dur": 56,
    "args": {
      "External id": 83798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83798,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83798, "pid": 5, "tid": 7, "ts": 1716454222668996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558234, "dur": 12,
    "args": {
      "External id": 83798,
      "cbid": 211, "correlation": 83798
    }
  },
  {
    "ph": "s", "id": 83798, "pid": 76337, "tid": -914061504, "ts": 1716454222558234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222669053, "dur": 53,
    "args": {
      "External id": 83818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83818,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 83818, "pid": 5, "tid": 7, "ts": 1716454222669053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558301, "dur": 11,
    "args": {
      "External id": 83818,
      "cbid": 211, "correlation": 83818
    }
  },
  {
    "ph": "s", "id": 83818, "pid": 76337, "tid": -914061504, "ts": 1716454222558301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222669107, "dur": 4,
    "args": {
      "External id": 83830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83830,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83830, "pid": 5, "tid": 7, "ts": 1716454222669107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558321, "dur": 7,
    "args": {
      "External id": 83830,
      "cbid": 211, "correlation": 83830
    }
  },
  {
    "ph": "s", "id": 83830, "pid": 76337, "tid": -914061504, "ts": 1716454222558321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222669112, "dur": 55,
    "args": {
      "External id": 83833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83833, "pid": 5, "tid": 7, "ts": 1716454222669112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558341, "dur": 7,
    "args": {
      "External id": 83833,
      "cbid": 211, "correlation": 83833
    }
  },
  {
    "ph": "s", "id": 83833, "pid": 76337, "tid": -914061504, "ts": 1716454222558341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222669169, "dur": 36,
    "args": {
      "External id": 83842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83842,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83842, "pid": 5, "tid": 7, "ts": 1716454222669169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558382, "dur": 10,
    "args": {
      "External id": 83842,
      "cbid": 211, "correlation": 83842
    }
  },
  {
    "ph": "s", "id": 83842, "pid": 76337, "tid": -914061504, "ts": 1716454222558382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558445, "dur": 0,
    "args": {
      "External id": 83852,
      "cbid": 317, "correlation": 83852
    }
  },
  {
    "ph": "f", "id": 83852, "pid": 76337, "tid": -914061504, "ts": 1716454222558445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558446, "dur": 0,
    "args": {
      "External id": 83853,
      "cbid": 203, "correlation": 83853
    }
  },
  {
    "ph": "f", "id": 83853, "pid": 76337, "tid": -914061504, "ts": 1716454222558446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558447, "dur": 0,
    "args": {
      "External id": 83854,
      "cbid": 205, "correlation": 83854
    }
  },
  {
    "ph": "f", "id": 83854, "pid": 76337, "tid": -914061504, "ts": 1716454222558447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222669206, "dur": 40,
    "args": {
      "External id": 83858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83858,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83858, "pid": 5, "tid": 7, "ts": 1716454222669206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558460, "dur": 12,
    "args": {
      "External id": 83858,
      "cbid": 211, "correlation": 83858
    }
  },
  {
    "ph": "s", "id": 83858, "pid": 76337, "tid": -914061504, "ts": 1716454222558460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222669248, "dur": 14,
    "args": {
      "External id": 83860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83860,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83860, "pid": 5, "tid": 7, "ts": 1716454222669248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558475, "dur": 5,
    "args": {
      "External id": 83860,
      "cbid": 211, "correlation": 83860
    }
  },
  {
    "ph": "s", "id": 83860, "pid": 76337, "tid": -914061504, "ts": 1716454222558475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222669264, "dur": 3,
    "args": {
      "External id": 83862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83862,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83862, "pid": 5, "tid": 7, "ts": 1716454222669264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558485, "dur": 6,
    "args": {
      "External id": 83862,
      "cbid": 211, "correlation": 83862
    }
  },
  {
    "ph": "s", "id": 83862, "pid": 76337, "tid": -914061504, "ts": 1716454222558485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558493, "dur": 0,
    "args": {
      "External id": 83863,
      "cbid": 51, "correlation": 83863
    }
  },
  {
    "ph": "s", "id": 83863, "pid": 76337, "tid": -914061504, "ts": 1716454222558493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222669269, "dur": 709,
    "args": {
      "External id": 83864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83864,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83864, "pid": 5, "tid": 7, "ts": 1716454222669269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558494, "dur": 5,
    "args": {
      "External id": 83864,
      "cbid": 211, "correlation": 83864
    }
  },
  {
    "ph": "s", "id": 83864, "pid": 76337, "tid": -914061504, "ts": 1716454222558494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222669979, "dur": 60,
    "args": {
      "External id": 83869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83869, "pid": 5, "tid": 7, "ts": 1716454222669979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558521, "dur": 9,
    "args": {
      "External id": 83869,
      "cbid": 211, "correlation": 83869
    }
  },
  {
    "ph": "s", "id": 83869, "pid": 76337, "tid": -914061504, "ts": 1716454222558521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558579, "dur": 0,
    "args": {
      "External id": 83879,
      "cbid": 317, "correlation": 83879
    }
  },
  {
    "ph": "f", "id": 83879, "pid": 76337, "tid": -914061504, "ts": 1716454222558579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558579, "dur": 0,
    "args": {
      "External id": 83880,
      "cbid": 203, "correlation": 83880
    }
  },
  {
    "ph": "f", "id": 83880, "pid": 76337, "tid": -914061504, "ts": 1716454222558579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558580, "dur": 0,
    "args": {
      "External id": 83881,
      "cbid": 205, "correlation": 83881
    }
  },
  {
    "ph": "f", "id": 83881, "pid": 76337, "tid": -914061504, "ts": 1716454222558580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222670040, "dur": 74,
    "args": {
      "External id": 83885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83885, "pid": 5, "tid": 7, "ts": 1716454222670040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558593, "dur": 12,
    "args": {
      "External id": 83885,
      "cbid": 211, "correlation": 83885
    }
  },
  {
    "ph": "s", "id": 83885, "pid": 76337, "tid": -914061504, "ts": 1716454222558593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222670115, "dur": 209,
    "args": {
      "External id": 83887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83887,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 83887, "pid": 5, "tid": 7, "ts": 1716454222670115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558613, "dur": 8,
    "args": {
      "External id": 83887,
      "cbid": 211, "correlation": 83887
    }
  },
  {
    "ph": "s", "id": 83887, "pid": 76337, "tid": -914061504, "ts": 1716454222558613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222670325, "dur": 40,
    "args": {
      "External id": 83889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83889,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83889, "pid": 5, "tid": 7, "ts": 1716454222670325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222558625, "dur": 495,
    "args": {
      "External id": 83889,
      "cbid": 211, "correlation": 83889
    }
  },
  {
    "ph": "s", "id": 83889, "pid": 76337, "tid": -914061504, "ts": 1716454222558625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222670366, "dur": 61,
    "args": {
      "External id": 83895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83895,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83895, "pid": 5, "tid": 7, "ts": 1716454222670366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559142, "dur": 10,
    "args": {
      "External id": 83895,
      "cbid": 211, "correlation": 83895
    }
  },
  {
    "ph": "s", "id": 83895, "pid": 76337, "tid": -914061504, "ts": 1716454222559142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222670429, "dur": 50,
    "args": {
      "External id": 83903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83903,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83903, "pid": 5, "tid": 7, "ts": 1716454222670429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559174, "dur": 8,
    "args": {
      "External id": 83903,
      "cbid": 211, "correlation": 83903
    }
  },
  {
    "ph": "s", "id": 83903, "pid": 76337, "tid": -914061504, "ts": 1716454222559174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222670481, "dur": 35,
    "args": {
      "External id": 83911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83911,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83911, "pid": 5, "tid": 7, "ts": 1716454222670481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559202, "dur": 28,
    "args": {
      "External id": 83911,
      "cbid": 211, "correlation": 83911
    }
  },
  {
    "ph": "s", "id": 83911, "pid": 76337, "tid": -914061504, "ts": 1716454222559202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222670517, "dur": 53,
    "args": {
      "External id": 83931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83931,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 83931, "pid": 5, "tid": 7, "ts": 1716454222670517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559306, "dur": 13,
    "args": {
      "External id": 83931,
      "cbid": 211, "correlation": 83931
    }
  },
  {
    "ph": "s", "id": 83931, "pid": 76337, "tid": -914061504, "ts": 1716454222559306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222670571, "dur": 4,
    "args": {
      "External id": 83943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83943,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 83943, "pid": 5, "tid": 7, "ts": 1716454222670571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559328, "dur": 6,
    "args": {
      "External id": 83943,
      "cbid": 211, "correlation": 83943
    }
  },
  {
    "ph": "s", "id": 83943, "pid": 76337, "tid": -914061504, "ts": 1716454222559328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222670577, "dur": 55,
    "args": {
      "External id": 83946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83946, "pid": 5, "tid": 7, "ts": 1716454222670577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559346, "dur": 6,
    "args": {
      "External id": 83946,
      "cbid": 211, "correlation": 83946
    }
  },
  {
    "ph": "s", "id": 83946, "pid": 76337, "tid": -914061504, "ts": 1716454222559346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559404, "dur": 0,
    "args": {
      "External id": 83957,
      "cbid": 317, "correlation": 83957
    }
  },
  {
    "ph": "f", "id": 83957, "pid": 76337, "tid": -914061504, "ts": 1716454222559404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559405, "dur": 0,
    "args": {
      "External id": 83958,
      "cbid": 203, "correlation": 83958
    }
  },
  {
    "ph": "f", "id": 83958, "pid": 76337, "tid": -914061504, "ts": 1716454222559405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559405, "dur": 0,
    "args": {
      "External id": 83959,
      "cbid": 205, "correlation": 83959
    }
  },
  {
    "ph": "f", "id": 83959, "pid": 76337, "tid": -914061504, "ts": 1716454222559405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559429, "dur": 1,
    "args": {
      "External id": 83963,
      "cbid": 251, "correlation": 83963
    }
  },
  {
    "ph": "f", "id": 83963, "pid": 76337, "tid": -914061504, "ts": 1716454222559429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559431, "dur": 0,
    "args": {
      "External id": 83964,
      "cbid": 251, "correlation": 83964
    }
  },
  {
    "ph": "f", "id": 83964, "pid": 76337, "tid": -914061504, "ts": 1716454222559431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559432, "dur": 0,
    "args": {
      "External id": 83965,
      "cbid": 251, "correlation": 83965
    }
  },
  {
    "ph": "f", "id": 83965, "pid": 76337, "tid": -914061504, "ts": 1716454222559432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559433, "dur": 0,
    "args": {
      "External id": 83966,
      "cbid": 251, "correlation": 83966
    }
  },
  {
    "ph": "f", "id": 83966, "pid": 76337, "tid": -914061504, "ts": 1716454222559433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559433, "dur": 0,
    "args": {
      "External id": 83967,
      "cbid": 251, "correlation": 83967
    }
  },
  {
    "ph": "f", "id": 83967, "pid": 76337, "tid": -914061504, "ts": 1716454222559433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559434, "dur": 0,
    "args": {
      "External id": 83968,
      "cbid": 251, "correlation": 83968
    }
  },
  {
    "ph": "f", "id": 83968, "pid": 76337, "tid": -914061504, "ts": 1716454222559434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559435, "dur": 0,
    "args": {
      "External id": 83969,
      "cbid": 251, "correlation": 83969
    }
  },
  {
    "ph": "f", "id": 83969, "pid": 76337, "tid": -914061504, "ts": 1716454222559435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559436, "dur": 0,
    "args": {
      "External id": 83970,
      "cbid": 251, "correlation": 83970
    }
  },
  {
    "ph": "f", "id": 83970, "pid": 76337, "tid": -914061504, "ts": 1716454222559436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559437, "dur": 0,
    "args": {
      "External id": 83971,
      "cbid": 251, "correlation": 83971
    }
  },
  {
    "ph": "f", "id": 83971, "pid": 76337, "tid": -914061504, "ts": 1716454222559437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222670633, "dur": 114,
    "args": {
      "External id": 83972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83972,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 83972, "pid": 5, "tid": 7, "ts": 1716454222670633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559439, "dur": 13,
    "args": {
      "External id": 83972,
      "cbid": 211, "correlation": 83972
    }
  },
  {
    "ph": "s", "id": 83972, "pid": 76337, "tid": -914061504, "ts": 1716454222559439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222670749, "dur": 60,
    "args": {
      "External id": 83978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83978,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83978, "pid": 5, "tid": 7, "ts": 1716454222670749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559475, "dur": 8,
    "args": {
      "External id": 83978,
      "cbid": 211, "correlation": 83978
    }
  },
  {
    "ph": "s", "id": 83978, "pid": 76337, "tid": -914061504, "ts": 1716454222559475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222670811, "dur": 456,
    "args": {
      "External id": 83987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 83987,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 83987, "pid": 5, "tid": 7, "ts": 1716454222670811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559556, "dur": 15,
    "args": {
      "External id": 83987,
      "cbid": 211, "correlation": 83987
    }
  },
  {
    "ph": "s", "id": 83987, "pid": 76337, "tid": -914061504, "ts": 1716454222559556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222671269, "dur": 185,
    "args": {
      "External id": 84009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84009,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84009, "pid": 5, "tid": 7, "ts": 1716454222671269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559614, "dur": 10,
    "args": {
      "External id": 84009,
      "cbid": 211, "correlation": 84009
    }
  },
  {
    "ph": "s", "id": 84009, "pid": 76337, "tid": -914061504, "ts": 1716454222559614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559701, "dur": 1,
    "args": {
      "External id": 84020,
      "cbid": 251, "correlation": 84020
    }
  },
  {
    "ph": "f", "id": 84020, "pid": 76337, "tid": -914061504, "ts": 1716454222559701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222671455, "dur": 200,
    "args": {
      "External id": 84021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84021,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84021, "pid": 5, "tid": 7, "ts": 1716454222671455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559706, "dur": 13,
    "args": {
      "External id": 84021,
      "cbid": 211, "correlation": 84021
    }
  },
  {
    "ph": "s", "id": 84021, "pid": 76337, "tid": -914061504, "ts": 1716454222559706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559774, "dur": 1,
    "args": {
      "External id": 84032,
      "cbid": 251, "correlation": 84032
    }
  },
  {
    "ph": "f", "id": 84032, "pid": 76337, "tid": -914061504, "ts": 1716454222559774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222671656, "dur": 195,
    "args": {
      "External id": 84033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84033,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84033, "pid": 5, "tid": 7, "ts": 1716454222671656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559779, "dur": 11,
    "args": {
      "External id": 84033,
      "cbid": 211, "correlation": 84033
    }
  },
  {
    "ph": "s", "id": 84033, "pid": 76337, "tid": -914061504, "ts": 1716454222559779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559842, "dur": 1,
    "args": {
      "External id": 84044,
      "cbid": 251, "correlation": 84044
    }
  },
  {
    "ph": "f", "id": 84044, "pid": 76337, "tid": -914061504, "ts": 1716454222559842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222671853, "dur": 194,
    "args": {
      "External id": 84045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84045,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84045, "pid": 5, "tid": 7, "ts": 1716454222671853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559846, "dur": 11,
    "args": {
      "External id": 84045,
      "cbid": 211, "correlation": 84045
    }
  },
  {
    "ph": "s", "id": 84045, "pid": 76337, "tid": -914061504, "ts": 1716454222559846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222672048, "dur": 18957,
    "args": {
      "External id": 84066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84066,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84066, "pid": 5, "tid": 7, "ts": 1716454222672048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222559926, "dur": 13,
    "args": {
      "External id": 84066,
      "cbid": 211, "correlation": 84066
    }
  },
  {
    "ph": "s", "id": 84066, "pid": 76337, "tid": -914061504, "ts": 1716454222559926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222560033, "dur": 1,
    "args": {
      "External id": 84084,
      "cbid": 251, "correlation": 84084
    }
  },
  {
    "ph": "f", "id": 84084, "pid": 76337, "tid": -914061504, "ts": 1716454222560033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222691007, "dur": 204,
    "args": {
      "External id": 84086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84086,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84086, "pid": 5, "tid": 7, "ts": 1716454222691007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222560039, "dur": 14,
    "args": {
      "External id": 84086,
      "cbid": 211, "correlation": 84086
    }
  },
  {
    "ph": "s", "id": 84086, "pid": 76337, "tid": -914061504, "ts": 1716454222560039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222691212, "dur": 66,
    "args": {
      "External id": 84094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84094,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84094, "pid": 5, "tid": 7, "ts": 1716454222691212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222560111, "dur": 12,
    "args": {
      "External id": 84094,
      "cbid": 211, "correlation": 84094
    }
  },
  {
    "ph": "s", "id": 84094, "pid": 76337, "tid": -914061504, "ts": 1716454222560111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222691280, "dur": 97,
    "args": {
      "External id": 84102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84102,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84102, "pid": 5, "tid": 7, "ts": 1716454222691280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222560151, "dur": 9,
    "args": {
      "External id": 84102,
      "cbid": 211, "correlation": 84102
    }
  },
  {
    "ph": "s", "id": 84102, "pid": 76337, "tid": -914061504, "ts": 1716454222560151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222691378, "dur": 56,
    "args": {
      "External id": 84113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84113,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84113, "pid": 5, "tid": 7, "ts": 1716454222691378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222560222, "dur": 70,
    "args": {
      "External id": 84113,
      "cbid": 211, "correlation": 84113
    }
  },
  {
    "ph": "s", "id": 84113, "pid": 76337, "tid": -914061504, "ts": 1716454222560222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222691435, "dur": 94,
    "args": {
      "External id": 84135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84135,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84135, "pid": 5, "tid": 7, "ts": 1716454222691435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222560312, "dur": 1972,
    "args": {
      "External id": 84135,
      "cbid": 211, "correlation": 84135
    }
  },
  {
    "ph": "s", "id": 84135, "pid": 76337, "tid": -914061504, "ts": 1716454222560312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562360, "dur": 1,
    "args": {
      "External id": 84146,
      "cbid": 251, "correlation": 84146
    }
  },
  {
    "ph": "f", "id": 84146, "pid": 76337, "tid": -914061504, "ts": 1716454222562360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222691530, "dur": 105,
    "args": {
      "External id": 84147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84147,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84147, "pid": 5, "tid": 7, "ts": 1716454222691530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562366, "dur": 69,
    "args": {
      "External id": 84147,
      "cbid": 211, "correlation": 84147
    }
  },
  {
    "ph": "s", "id": 84147, "pid": 76337, "tid": -914061504, "ts": 1716454222562366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562495, "dur": 1,
    "args": {
      "External id": 84158,
      "cbid": 251, "correlation": 84158
    }
  },
  {
    "ph": "f", "id": 84158, "pid": 76337, "tid": -914061504, "ts": 1716454222562495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562500, "dur": 0,
    "args": {
      "External id": 84159,
      "cbid": 251, "correlation": 84159
    }
  },
  {
    "ph": "f", "id": 84159, "pid": 76337, "tid": -914061504, "ts": 1716454222562500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222691637, "dur": 10,
    "args": {
      "External id": 84160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84160,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 84160, "pid": 5, "tid": 7, "ts": 1716454222691637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562501, "dur": 13,
    "args": {
      "External id": 84160,
      "cbid": 211, "correlation": 84160
    }
  },
  {
    "ph": "s", "id": 84160, "pid": 76337, "tid": -914061504, "ts": 1716454222562501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222691648, "dur": 5,
    "args": {
      "External id": 84162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84162,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 84162, "pid": 5, "tid": 7, "ts": 1716454222691648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562516, "dur": 6,
    "args": {
      "External id": 84162,
      "cbid": 211, "correlation": 84162
    }
  },
  {
    "ph": "s", "id": 84162, "pid": 76337, "tid": -914061504, "ts": 1716454222562516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562577, "dur": 1,
    "args": {
      "External id": 84173,
      "cbid": 251, "correlation": 84173
    }
  },
  {
    "ph": "f", "id": 84173, "pid": 76337, "tid": -914061504, "ts": 1716454222562577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562580, "dur": 0,
    "args": {
      "External id": 84174,
      "cbid": 251, "correlation": 84174
    }
  },
  {
    "ph": "f", "id": 84174, "pid": 76337, "tid": -914061504, "ts": 1716454222562580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222691655, "dur": 6,
    "args": {
      "External id": 84175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84175,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 84175, "pid": 5, "tid": 7, "ts": 1716454222691655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562582, "dur": 12,
    "args": {
      "External id": 84175,
      "cbid": 211, "correlation": 84175
    }
  },
  {
    "ph": "s", "id": 84175, "pid": 76337, "tid": -914061504, "ts": 1716454222562582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222691662, "dur": 4,
    "args": {
      "External id": 84177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84177,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 84177, "pid": 5, "tid": 7, "ts": 1716454222691662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562596, "dur": 6,
    "args": {
      "External id": 84177,
      "cbid": 211, "correlation": 84177
    }
  },
  {
    "ph": "s", "id": 84177, "pid": 76337, "tid": -914061504, "ts": 1716454222562596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222691667, "dur": 159,
    "args": {
      "External id": 84198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84198,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84198, "pid": 5, "tid": 7, "ts": 1716454222691667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562671, "dur": 13,
    "args": {
      "External id": 84198,
      "cbid": 211, "correlation": 84198
    }
  },
  {
    "ph": "s", "id": 84198, "pid": 76337, "tid": -914061504, "ts": 1716454222562671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562771, "dur": 1,
    "args": {
      "External id": 84216,
      "cbid": 251, "correlation": 84216
    }
  },
  {
    "ph": "f", "id": 84216, "pid": 76337, "tid": -914061504, "ts": 1716454222562771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222691828, "dur": 107,
    "args": {
      "External id": 84218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84218,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84218, "pid": 5, "tid": 7, "ts": 1716454222691828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562777, "dur": 13,
    "args": {
      "External id": 84218,
      "cbid": 211, "correlation": 84218
    }
  },
  {
    "ph": "s", "id": 84218, "pid": 76337, "tid": -914061504, "ts": 1716454222562777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222691937, "dur": 34,
    "args": {
      "External id": 84226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84226,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84226, "pid": 5, "tid": 7, "ts": 1716454222691937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562847, "dur": 12,
    "args": {
      "External id": 84226,
      "cbid": 211, "correlation": 84226
    }
  },
  {
    "ph": "s", "id": 84226, "pid": 76337, "tid": -914061504, "ts": 1716454222562847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222691973, "dur": 67,
    "args": {
      "External id": 84234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84234, "pid": 5, "tid": 7, "ts": 1716454222691973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562889, "dur": 9,
    "args": {
      "External id": 84234,
      "cbid": 211, "correlation": 84234
    }
  },
  {
    "ph": "s", "id": 84234, "pid": 76337, "tid": -914061504, "ts": 1716454222562889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222692041, "dur": 94,
    "args": {
      "External id": 84256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84256,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84256, "pid": 5, "tid": 7, "ts": 1716454222692041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222562940, "dur": 10,
    "args": {
      "External id": 84256,
      "cbid": 211, "correlation": 84256
    }
  },
  {
    "ph": "s", "id": 84256, "pid": 76337, "tid": -914061504, "ts": 1716454222562940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563036, "dur": 1,
    "args": {
      "External id": 84272,
      "cbid": 251, "correlation": 84272
    }
  },
  {
    "ph": "f", "id": 84272, "pid": 76337, "tid": -914061504, "ts": 1716454222563036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222692136, "dur": 586,
    "args": {
      "External id": 84274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84274,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84274, "pid": 5, "tid": 7, "ts": 1716454222692136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563042, "dur": 14,
    "args": {
      "External id": 84274,
      "cbid": 211, "correlation": 84274
    }
  },
  {
    "ph": "s", "id": 84274, "pid": 76337, "tid": -914061504, "ts": 1716454222563042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222692723, "dur": 249,
    "args": {
      "External id": 84282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84282, "pid": 5, "tid": 7, "ts": 1716454222692723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563109, "dur": 12,
    "args": {
      "External id": 84282,
      "cbid": 211, "correlation": 84282
    }
  },
  {
    "ph": "s", "id": 84282, "pid": 76337, "tid": -914061504, "ts": 1716454222563109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222692973, "dur": 251,
    "args": {
      "External id": 84290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84290, "pid": 5, "tid": 7, "ts": 1716454222692973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563140, "dur": 8,
    "args": {
      "External id": 84290,
      "cbid": 211, "correlation": 84290
    }
  },
  {
    "ph": "s", "id": 84290, "pid": 76337, "tid": -914061504, "ts": 1716454222563140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563221, "dur": 1,
    "args": {
      "External id": 84306,
      "cbid": 251, "correlation": 84306
    }
  },
  {
    "ph": "f", "id": 84306, "pid": 76337, "tid": -914061504, "ts": 1716454222563221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563226, "dur": 0,
    "args": {
      "External id": 84308,
      "cbid": 251, "correlation": 84308
    }
  },
  {
    "ph": "f", "id": 84308, "pid": 76337, "tid": -914061504, "ts": 1716454222563226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222693226, "dur": 362,
    "args": {
      "External id": 84309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84309,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 84309, "pid": 5, "tid": 7, "ts": 1716454222693226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563229, "dur": 13,
    "args": {
      "External id": 84309,
      "cbid": 211, "correlation": 84309
    }
  },
  {
    "ph": "s", "id": 84309, "pid": 76337, "tid": -914061504, "ts": 1716454222563229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222693590, "dur": 50,
    "args": {
      "External id": 84317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84317,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84317, "pid": 5, "tid": 7, "ts": 1716454222693590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563272, "dur": 10,
    "args": {
      "External id": 84317,
      "cbid": 211, "correlation": 84317
    }
  },
  {
    "ph": "s", "id": 84317, "pid": 76337, "tid": -914061504, "ts": 1716454222563272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222693641, "dur": 162,
    "args": {
      "External id": 84328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84328, "pid": 5, "tid": 7, "ts": 1716454222693641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563339, "dur": 219,
    "args": {
      "External id": 84328,
      "cbid": 211, "correlation": 84328
    }
  },
  {
    "ph": "s", "id": 84328, "pid": 76337, "tid": -914061504, "ts": 1716454222563339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563611, "dur": 0,
    "args": {
      "External id": 84340,
      "cbid": 317, "correlation": 84340
    }
  },
  {
    "ph": "f", "id": 84340, "pid": 76337, "tid": -914061504, "ts": 1716454222563611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563611, "dur": 0,
    "args": {
      "External id": 84341,
      "cbid": 203, "correlation": 84341
    }
  },
  {
    "ph": "f", "id": 84341, "pid": 76337, "tid": -914061504, "ts": 1716454222563611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563612, "dur": 0,
    "args": {
      "External id": 84342,
      "cbid": 205, "correlation": 84342
    }
  },
  {
    "ph": "f", "id": 84342, "pid": 76337, "tid": -914061504, "ts": 1716454222563612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563635, "dur": 1,
    "args": {
      "External id": 84346,
      "cbid": 251, "correlation": 84346
    }
  },
  {
    "ph": "f", "id": 84346, "pid": 76337, "tid": -914061504, "ts": 1716454222563635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563637, "dur": 0,
    "args": {
      "External id": 84347,
      "cbid": 251, "correlation": 84347
    }
  },
  {
    "ph": "f", "id": 84347, "pid": 76337, "tid": -914061504, "ts": 1716454222563637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563637, "dur": 0,
    "args": {
      "External id": 84348,
      "cbid": 251, "correlation": 84348
    }
  },
  {
    "ph": "f", "id": 84348, "pid": 76337, "tid": -914061504, "ts": 1716454222563637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563638, "dur": 0,
    "args": {
      "External id": 84349,
      "cbid": 251, "correlation": 84349
    }
  },
  {
    "ph": "f", "id": 84349, "pid": 76337, "tid": -914061504, "ts": 1716454222563638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563639, "dur": 0,
    "args": {
      "External id": 84350,
      "cbid": 251, "correlation": 84350
    }
  },
  {
    "ph": "f", "id": 84350, "pid": 76337, "tid": -914061504, "ts": 1716454222563639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563640, "dur": 0,
    "args": {
      "External id": 84351,
      "cbid": 251, "correlation": 84351
    }
  },
  {
    "ph": "f", "id": 84351, "pid": 76337, "tid": -914061504, "ts": 1716454222563640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563640, "dur": 0,
    "args": {
      "External id": 84352,
      "cbid": 251, "correlation": 84352
    }
  },
  {
    "ph": "f", "id": 84352, "pid": 76337, "tid": -914061504, "ts": 1716454222563640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563641, "dur": 0,
    "args": {
      "External id": 84353,
      "cbid": 251, "correlation": 84353
    }
  },
  {
    "ph": "f", "id": 84353, "pid": 76337, "tid": -914061504, "ts": 1716454222563641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563642, "dur": 0,
    "args": {
      "External id": 84354,
      "cbid": 251, "correlation": 84354
    }
  },
  {
    "ph": "f", "id": 84354, "pid": 76337, "tid": -914061504, "ts": 1716454222563642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222693804, "dur": 117,
    "args": {
      "External id": 84355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84355,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84355, "pid": 5, "tid": 7, "ts": 1716454222693804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563644, "dur": 42,
    "args": {
      "External id": 84355,
      "cbid": 211, "correlation": 84355
    }
  },
  {
    "ph": "s", "id": 84355, "pid": 76337, "tid": -914061504, "ts": 1716454222563644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222693923, "dur": 60,
    "args": {
      "External id": 84361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84361, "pid": 5, "tid": 7, "ts": 1716454222693923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563709, "dur": 107,
    "args": {
      "External id": 84361,
      "cbid": 211, "correlation": 84361
    }
  },
  {
    "ph": "s", "id": 84361, "pid": 76337, "tid": -914061504, "ts": 1716454222563709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222693984, "dur": 50,
    "args": {
      "External id": 84369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84369,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84369, "pid": 5, "tid": 7, "ts": 1716454222693984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222563838, "dur": 286,
    "args": {
      "External id": 84369,
      "cbid": 211, "correlation": 84369
    }
  },
  {
    "ph": "s", "id": 84369, "pid": 76337, "tid": -914061504, "ts": 1716454222563838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222694036, "dur": 98,
    "args": {
      "External id": 84378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84378,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84378, "pid": 5, "tid": 7, "ts": 1716454222694036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564156, "dur": 11,
    "args": {
      "External id": 84378,
      "cbid": 211, "correlation": 84378
    }
  },
  {
    "ph": "s", "id": 84378, "pid": 76337, "tid": -914061504, "ts": 1716454222564156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222694135, "dur": 94,
    "args": {
      "External id": 84398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84398,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 84398, "pid": 5, "tid": 7, "ts": 1716454222694135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564230, "dur": 12,
    "args": {
      "External id": 84398,
      "cbid": 211, "correlation": 84398
    }
  },
  {
    "ph": "s", "id": 84398, "pid": 76337, "tid": -914061504, "ts": 1716454222564230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222694230, "dur": 5,
    "args": {
      "External id": 84410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84410,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 84410, "pid": 5, "tid": 7, "ts": 1716454222694230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564252, "dur": 10,
    "args": {
      "External id": 84410,
      "cbid": 211, "correlation": 84410
    }
  },
  {
    "ph": "s", "id": 84410, "pid": 76337, "tid": -914061504, "ts": 1716454222564252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222694236, "dur": 109,
    "args": {
      "External id": 84413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84413, "pid": 5, "tid": 7, "ts": 1716454222694236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564275, "dur": 109,
    "args": {
      "External id": 84413,
      "cbid": 211, "correlation": 84413
    }
  },
  {
    "ph": "s", "id": 84413, "pid": 76337, "tid": -914061504, "ts": 1716454222564275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222694347, "dur": 70,
    "args": {
      "External id": 84422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84422,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84422, "pid": 5, "tid": 7, "ts": 1716454222694347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564417, "dur": 10,
    "args": {
      "External id": 84422,
      "cbid": 211, "correlation": 84422
    }
  },
  {
    "ph": "s", "id": 84422, "pid": 76337, "tid": -914061504, "ts": 1716454222564417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564470, "dur": 0,
    "args": {
      "External id": 84432,
      "cbid": 317, "correlation": 84432
    }
  },
  {
    "ph": "f", "id": 84432, "pid": 76337, "tid": -914061504, "ts": 1716454222564470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564471, "dur": 0,
    "args": {
      "External id": 84433,
      "cbid": 203, "correlation": 84433
    }
  },
  {
    "ph": "f", "id": 84433, "pid": 76337, "tid": -914061504, "ts": 1716454222564471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564471, "dur": 0,
    "args": {
      "External id": 84434,
      "cbid": 205, "correlation": 84434
    }
  },
  {
    "ph": "f", "id": 84434, "pid": 76337, "tid": -914061504, "ts": 1716454222564471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222694418, "dur": 76,
    "args": {
      "External id": 84438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84438, "pid": 5, "tid": 7, "ts": 1716454222694418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564486, "dur": 12,
    "args": {
      "External id": 84438,
      "cbid": 211, "correlation": 84438
    }
  },
  {
    "ph": "s", "id": 84438, "pid": 76337, "tid": -914061504, "ts": 1716454222564486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222694496, "dur": 24,
    "args": {
      "External id": 84440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84440,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84440, "pid": 5, "tid": 7, "ts": 1716454222694496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564500, "dur": 5,
    "args": {
      "External id": 84440,
      "cbid": 211, "correlation": 84440
    }
  },
  {
    "ph": "s", "id": 84440, "pid": 76337, "tid": -914061504, "ts": 1716454222564500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222694521, "dur": 4,
    "args": {
      "External id": 84442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 84442, "pid": 5, "tid": 7, "ts": 1716454222694521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564510, "dur": 6,
    "args": {
      "External id": 84442,
      "cbid": 211, "correlation": 84442
    }
  },
  {
    "ph": "s", "id": 84442, "pid": 76337, "tid": -914061504, "ts": 1716454222564510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564519, "dur": 0,
    "args": {
      "External id": 84443,
      "cbid": 51, "correlation": 84443
    }
  },
  {
    "ph": "s", "id": 84443, "pid": 76337, "tid": -914061504, "ts": 1716454222564519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222694526, "dur": 1387,
    "args": {
      "External id": 84444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84444,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84444, "pid": 5, "tid": 7, "ts": 1716454222694526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564520, "dur": 5,
    "args": {
      "External id": 84444,
      "cbid": 211, "correlation": 84444
    }
  },
  {
    "ph": "s", "id": 84444, "pid": 76337, "tid": -914061504, "ts": 1716454222564520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222695915, "dur": 60,
    "args": {
      "External id": 84449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84449, "pid": 5, "tid": 7, "ts": 1716454222695915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564549, "dur": 8,
    "args": {
      "External id": 84449,
      "cbid": 211, "correlation": 84449
    }
  },
  {
    "ph": "s", "id": 84449, "pid": 76337, "tid": -914061504, "ts": 1716454222564549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222695976, "dur": 4,
    "args": {
      "External id": 84457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84457,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 84457, "pid": 5, "tid": 7, "ts": 1716454222695976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564592, "dur": 10,
    "args": {
      "External id": 84457,
      "cbid": 211, "correlation": 84457
    }
  },
  {
    "ph": "s", "id": 84457, "pid": 76337, "tid": -914061504, "ts": 1716454222564592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564659, "dur": 1,
    "args": {
      "External id": 84473,
      "cbid": 251, "correlation": 84473
    }
  },
  {
    "ph": "f", "id": 84473, "pid": 76337, "tid": -914061504, "ts": 1716454222564659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564664, "dur": 0,
    "args": {
      "External id": 84475,
      "cbid": 251, "correlation": 84475
    }
  },
  {
    "ph": "f", "id": 84475, "pid": 76337, "tid": -914061504, "ts": 1716454222564664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222695981, "dur": 12,
    "args": {
      "External id": 84476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84476,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 84476, "pid": 5, "tid": 7, "ts": 1716454222695981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564666, "dur": 12,
    "args": {
      "External id": 84476,
      "cbid": 211, "correlation": 84476
    }
  },
  {
    "ph": "s", "id": 84476, "pid": 76337, "tid": -914061504, "ts": 1716454222564666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222695994, "dur": 5,
    "args": {
      "External id": 84478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84478,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 84478, "pid": 5, "tid": 7, "ts": 1716454222695994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564680, "dur": 6,
    "args": {
      "External id": 84478,
      "cbid": 211, "correlation": 84478
    }
  },
  {
    "ph": "s", "id": 84478, "pid": 76337, "tid": -914061504, "ts": 1716454222564680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222696000, "dur": 55,
    "args": {
      "External id": 84488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84488, "pid": 5, "tid": 7, "ts": 1716454222696000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222564738, "dur": 563,
    "args": {
      "External id": 84488,
      "cbid": 211, "correlation": 84488
    }
  },
  {
    "ph": "s", "id": 84488, "pid": 76337, "tid": -914061504, "ts": 1716454222564738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222696057, "dur": 53,
    "args": {
      "External id": 84508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84508,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 84508, "pid": 5, "tid": 7, "ts": 1716454222696057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565357, "dur": 11,
    "args": {
      "External id": 84508,
      "cbid": 211, "correlation": 84508
    }
  },
  {
    "ph": "s", "id": 84508, "pid": 76337, "tid": -914061504, "ts": 1716454222565357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222696111, "dur": 4,
    "args": {
      "External id": 84520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84520,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 84520, "pid": 5, "tid": 7, "ts": 1716454222696111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565378, "dur": 6,
    "args": {
      "External id": 84520,
      "cbid": 211, "correlation": 84520
    }
  },
  {
    "ph": "s", "id": 84520, "pid": 76337, "tid": -914061504, "ts": 1716454222565378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222696116, "dur": 56,
    "args": {
      "External id": 84523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84523,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84523, "pid": 5, "tid": 7, "ts": 1716454222696116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565396, "dur": 7,
    "args": {
      "External id": 84523,
      "cbid": 211, "correlation": 84523
    }
  },
  {
    "ph": "s", "id": 84523, "pid": 76337, "tid": -914061504, "ts": 1716454222565396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222696173, "dur": 36,
    "args": {
      "External id": 84532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84532,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84532, "pid": 5, "tid": 7, "ts": 1716454222696173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565438, "dur": 10,
    "args": {
      "External id": 84532,
      "cbid": 211, "correlation": 84532
    }
  },
  {
    "ph": "s", "id": 84532, "pid": 76337, "tid": -914061504, "ts": 1716454222565438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565501, "dur": 0,
    "args": {
      "External id": 84542,
      "cbid": 317, "correlation": 84542
    }
  },
  {
    "ph": "f", "id": 84542, "pid": 76337, "tid": -914061504, "ts": 1716454222565501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565501, "dur": 0,
    "args": {
      "External id": 84543,
      "cbid": 203, "correlation": 84543
    }
  },
  {
    "ph": "f", "id": 84543, "pid": 76337, "tid": -914061504, "ts": 1716454222565501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565502, "dur": 0,
    "args": {
      "External id": 84544,
      "cbid": 205, "correlation": 84544
    }
  },
  {
    "ph": "f", "id": 84544, "pid": 76337, "tid": -914061504, "ts": 1716454222565502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222696211, "dur": 40,
    "args": {
      "External id": 84548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84548,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84548, "pid": 5, "tid": 7, "ts": 1716454222696211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565517, "dur": 12,
    "args": {
      "External id": 84548,
      "cbid": 211, "correlation": 84548
    }
  },
  {
    "ph": "s", "id": 84548, "pid": 76337, "tid": -914061504, "ts": 1716454222565517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222696252, "dur": 14,
    "args": {
      "External id": 84550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84550,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84550, "pid": 5, "tid": 7, "ts": 1716454222696252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565531, "dur": 6,
    "args": {
      "External id": 84550,
      "cbid": 211, "correlation": 84550
    }
  },
  {
    "ph": "s", "id": 84550, "pid": 76337, "tid": -914061504, "ts": 1716454222565531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222696268, "dur": 3,
    "args": {
      "External id": 84552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84552,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 84552, "pid": 5, "tid": 7, "ts": 1716454222696268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565541, "dur": 5,
    "args": {
      "External id": 84552,
      "cbid": 211, "correlation": 84552
    }
  },
  {
    "ph": "s", "id": 84552, "pid": 76337, "tid": -914061504, "ts": 1716454222565541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565549, "dur": 0,
    "args": {
      "External id": 84553,
      "cbid": 51, "correlation": 84553
    }
  },
  {
    "ph": "s", "id": 84553, "pid": 76337, "tid": -914061504, "ts": 1716454222565549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222696272, "dur": 708,
    "args": {
      "External id": 84554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84554,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84554, "pid": 5, "tid": 7, "ts": 1716454222696272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565550, "dur": 5,
    "args": {
      "External id": 84554,
      "cbid": 211, "correlation": 84554
    }
  },
  {
    "ph": "s", "id": 84554, "pid": 76337, "tid": -914061504, "ts": 1716454222565550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222696982, "dur": 60,
    "args": {
      "External id": 84559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84559, "pid": 5, "tid": 7, "ts": 1716454222696982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565578, "dur": 9,
    "args": {
      "External id": 84559,
      "cbid": 211, "correlation": 84559
    }
  },
  {
    "ph": "s", "id": 84559, "pid": 76337, "tid": -914061504, "ts": 1716454222565578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565636, "dur": 0,
    "args": {
      "External id": 84569,
      "cbid": 317, "correlation": 84569
    }
  },
  {
    "ph": "f", "id": 84569, "pid": 76337, "tid": -914061504, "ts": 1716454222565636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565637, "dur": 0,
    "args": {
      "External id": 84570,
      "cbid": 203, "correlation": 84570
    }
  },
  {
    "ph": "f", "id": 84570, "pid": 76337, "tid": -914061504, "ts": 1716454222565637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565637, "dur": 0,
    "args": {
      "External id": 84571,
      "cbid": 205, "correlation": 84571
    }
  },
  {
    "ph": "f", "id": 84571, "pid": 76337, "tid": -914061504, "ts": 1716454222565637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222697043, "dur": 76,
    "args": {
      "External id": 84575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84575,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84575, "pid": 5, "tid": 7, "ts": 1716454222697043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565649, "dur": 12,
    "args": {
      "External id": 84575,
      "cbid": 211, "correlation": 84575
    }
  },
  {
    "ph": "s", "id": 84575, "pid": 76337, "tid": -914061504, "ts": 1716454222565649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222697120, "dur": 211,
    "args": {
      "External id": 84577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84577,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84577, "pid": 5, "tid": 7, "ts": 1716454222697120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565667, "dur": 7,
    "args": {
      "External id": 84577,
      "cbid": 211, "correlation": 84577
    }
  },
  {
    "ph": "s", "id": 84577, "pid": 76337, "tid": -914061504, "ts": 1716454222565667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222697332, "dur": 39,
    "args": {
      "External id": 84579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84579,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84579, "pid": 5, "tid": 7, "ts": 1716454222697332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565679, "dur": 6,
    "args": {
      "External id": 84579,
      "cbid": 211, "correlation": 84579
    }
  },
  {
    "ph": "s", "id": 84579, "pid": 76337, "tid": -914061504, "ts": 1716454222565679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222697373, "dur": 59,
    "args": {
      "External id": 84585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84585,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84585, "pid": 5, "tid": 7, "ts": 1716454222697373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222565705, "dur": 526,
    "args": {
      "External id": 84585,
      "cbid": 211, "correlation": 84585
    }
  },
  {
    "ph": "s", "id": 84585, "pid": 76337, "tid": -914061504, "ts": 1716454222565705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222697433, "dur": 50,
    "args": {
      "External id": 84593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84593,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84593, "pid": 5, "tid": 7, "ts": 1716454222697433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566253, "dur": 9,
    "args": {
      "External id": 84593,
      "cbid": 211, "correlation": 84593
    }
  },
  {
    "ph": "s", "id": 84593, "pid": 76337, "tid": -914061504, "ts": 1716454222566253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222697484, "dur": 35,
    "args": {
      "External id": 84601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84601,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84601, "pid": 5, "tid": 7, "ts": 1716454222697484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566283, "dur": 9,
    "args": {
      "External id": 84601,
      "cbid": 211, "correlation": 84601
    }
  },
  {
    "ph": "s", "id": 84601, "pid": 76337, "tid": -914061504, "ts": 1716454222566283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222697520, "dur": 52,
    "args": {
      "External id": 84621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84621,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 84621, "pid": 5, "tid": 7, "ts": 1716454222697520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566365, "dur": 13,
    "args": {
      "External id": 84621,
      "cbid": 211, "correlation": 84621
    }
  },
  {
    "ph": "s", "id": 84621, "pid": 76337, "tid": -914061504, "ts": 1716454222566365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222697574, "dur": 5,
    "args": {
      "External id": 84633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84633,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 84633, "pid": 5, "tid": 7, "ts": 1716454222697574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566389, "dur": 6,
    "args": {
      "External id": 84633,
      "cbid": 211, "correlation": 84633
    }
  },
  {
    "ph": "s", "id": 84633, "pid": 76337, "tid": -914061504, "ts": 1716454222566389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222697580, "dur": 56,
    "args": {
      "External id": 84636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84636, "pid": 5, "tid": 7, "ts": 1716454222697580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566407, "dur": 7,
    "args": {
      "External id": 84636,
      "cbid": 211, "correlation": 84636
    }
  },
  {
    "ph": "s", "id": 84636, "pid": 76337, "tid": -914061504, "ts": 1716454222566407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566465, "dur": 0,
    "args": {
      "External id": 84647,
      "cbid": 317, "correlation": 84647
    }
  },
  {
    "ph": "f", "id": 84647, "pid": 76337, "tid": -914061504, "ts": 1716454222566465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566465, "dur": 0,
    "args": {
      "External id": 84648,
      "cbid": 203, "correlation": 84648
    }
  },
  {
    "ph": "f", "id": 84648, "pid": 76337, "tid": -914061504, "ts": 1716454222566465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566466, "dur": 0,
    "args": {
      "External id": 84649,
      "cbid": 205, "correlation": 84649
    }
  },
  {
    "ph": "f", "id": 84649, "pid": 76337, "tid": -914061504, "ts": 1716454222566466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566490, "dur": 1,
    "args": {
      "External id": 84653,
      "cbid": 251, "correlation": 84653
    }
  },
  {
    "ph": "f", "id": 84653, "pid": 76337, "tid": -914061504, "ts": 1716454222566490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566492, "dur": 0,
    "args": {
      "External id": 84654,
      "cbid": 251, "correlation": 84654
    }
  },
  {
    "ph": "f", "id": 84654, "pid": 76337, "tid": -914061504, "ts": 1716454222566492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566493, "dur": 0,
    "args": {
      "External id": 84655,
      "cbid": 251, "correlation": 84655
    }
  },
  {
    "ph": "f", "id": 84655, "pid": 76337, "tid": -914061504, "ts": 1716454222566493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566493, "dur": 0,
    "args": {
      "External id": 84656,
      "cbid": 251, "correlation": 84656
    }
  },
  {
    "ph": "f", "id": 84656, "pid": 76337, "tid": -914061504, "ts": 1716454222566493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566494, "dur": 0,
    "args": {
      "External id": 84657,
      "cbid": 251, "correlation": 84657
    }
  },
  {
    "ph": "f", "id": 84657, "pid": 76337, "tid": -914061504, "ts": 1716454222566494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566495, "dur": 0,
    "args": {
      "External id": 84658,
      "cbid": 251, "correlation": 84658
    }
  },
  {
    "ph": "f", "id": 84658, "pid": 76337, "tid": -914061504, "ts": 1716454222566495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566496, "dur": 0,
    "args": {
      "External id": 84659,
      "cbid": 251, "correlation": 84659
    }
  },
  {
    "ph": "f", "id": 84659, "pid": 76337, "tid": -914061504, "ts": 1716454222566496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566496, "dur": 0,
    "args": {
      "External id": 84660,
      "cbid": 251, "correlation": 84660
    }
  },
  {
    "ph": "f", "id": 84660, "pid": 76337, "tid": -914061504, "ts": 1716454222566496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566498, "dur": 0,
    "args": {
      "External id": 84661,
      "cbid": 251, "correlation": 84661
    }
  },
  {
    "ph": "f", "id": 84661, "pid": 76337, "tid": -914061504, "ts": 1716454222566498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222697637, "dur": 114,
    "args": {
      "External id": 84662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84662,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84662, "pid": 5, "tid": 7, "ts": 1716454222697637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566500, "dur": 13,
    "args": {
      "External id": 84662,
      "cbid": 211, "correlation": 84662
    }
  },
  {
    "ph": "s", "id": 84662, "pid": 76337, "tid": -914061504, "ts": 1716454222566500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222697753, "dur": 60,
    "args": {
      "External id": 84668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84668, "pid": 5, "tid": 7, "ts": 1716454222697753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566535, "dur": 9,
    "args": {
      "External id": 84668,
      "cbid": 211, "correlation": 84668
    }
  },
  {
    "ph": "s", "id": 84668, "pid": 76337, "tid": -914061504, "ts": 1716454222566535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222697815, "dur": 599,
    "args": {
      "External id": 84677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84677,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84677, "pid": 5, "tid": 7, "ts": 1716454222697815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566617, "dur": 13,
    "args": {
      "External id": 84677,
      "cbid": 211, "correlation": 84677
    }
  },
  {
    "ph": "s", "id": 84677, "pid": 76337, "tid": -914061504, "ts": 1716454222566617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222698416, "dur": 184,
    "args": {
      "External id": 84699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84699,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84699, "pid": 5, "tid": 7, "ts": 1716454222698416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566675, "dur": 10,
    "args": {
      "External id": 84699,
      "cbid": 211, "correlation": 84699
    }
  },
  {
    "ph": "s", "id": 84699, "pid": 76337, "tid": -914061504, "ts": 1716454222566675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566761, "dur": 1,
    "args": {
      "External id": 84710,
      "cbid": 251, "correlation": 84710
    }
  },
  {
    "ph": "f", "id": 84710, "pid": 76337, "tid": -914061504, "ts": 1716454222566761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222698602, "dur": 196,
    "args": {
      "External id": 84711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84711,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84711, "pid": 5, "tid": 7, "ts": 1716454222698602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566766, "dur": 13,
    "args": {
      "External id": 84711,
      "cbid": 211, "correlation": 84711
    }
  },
  {
    "ph": "s", "id": 84711, "pid": 76337, "tid": -914061504, "ts": 1716454222566766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566834, "dur": 1,
    "args": {
      "External id": 84722,
      "cbid": 251, "correlation": 84722
    }
  },
  {
    "ph": "f", "id": 84722, "pid": 76337, "tid": -914061504, "ts": 1716454222566834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222698799, "dur": 191,
    "args": {
      "External id": 84723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84723,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84723, "pid": 5, "tid": 7, "ts": 1716454222698799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566838, "dur": 11,
    "args": {
      "External id": 84723,
      "cbid": 211, "correlation": 84723
    }
  },
  {
    "ph": "s", "id": 84723, "pid": 76337, "tid": -914061504, "ts": 1716454222566838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566902, "dur": 1,
    "args": {
      "External id": 84734,
      "cbid": 251, "correlation": 84734
    }
  },
  {
    "ph": "f", "id": 84734, "pid": 76337, "tid": -914061504, "ts": 1716454222566902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222698992, "dur": 188,
    "args": {
      "External id": 84735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84735,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84735, "pid": 5, "tid": 7, "ts": 1716454222698992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566906, "dur": 11,
    "args": {
      "External id": 84735,
      "cbid": 211, "correlation": 84735
    }
  },
  {
    "ph": "s", "id": 84735, "pid": 76337, "tid": -914061504, "ts": 1716454222566906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222699181, "dur": 18956,
    "args": {
      "External id": 84756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84756,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84756, "pid": 5, "tid": 7, "ts": 1716454222699181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222566992, "dur": 13,
    "args": {
      "External id": 84756,
      "cbid": 211, "correlation": 84756
    }
  },
  {
    "ph": "s", "id": 84756, "pid": 76337, "tid": -914061504, "ts": 1716454222566992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222567090, "dur": 1,
    "args": {
      "External id": 84774,
      "cbid": 251, "correlation": 84774
    }
  },
  {
    "ph": "f", "id": 84774, "pid": 76337, "tid": -914061504, "ts": 1716454222567090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222718139, "dur": 202,
    "args": {
      "External id": 84776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84776,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84776, "pid": 5, "tid": 7, "ts": 1716454222718139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222567096, "dur": 13,
    "args": {
      "External id": 84776,
      "cbid": 211, "correlation": 84776
    }
  },
  {
    "ph": "s", "id": 84776, "pid": 76337, "tid": -914061504, "ts": 1716454222567096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222718342, "dur": 66,
    "args": {
      "External id": 84784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84784,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84784, "pid": 5, "tid": 7, "ts": 1716454222718342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222567167, "dur": 21,
    "args": {
      "External id": 84784,
      "cbid": 211, "correlation": 84784
    }
  },
  {
    "ph": "s", "id": 84784, "pid": 76337, "tid": -914061504, "ts": 1716454222567167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222718410, "dur": 96,
    "args": {
      "External id": 84792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84792,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84792, "pid": 5, "tid": 7, "ts": 1716454222718410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222567217, "dur": 119,
    "args": {
      "External id": 84792,
      "cbid": 211, "correlation": 84792
    }
  },
  {
    "ph": "s", "id": 84792, "pid": 76337, "tid": -914061504, "ts": 1716454222567217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222718508, "dur": 54,
    "args": {
      "External id": 84803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84803,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84803, "pid": 5, "tid": 7, "ts": 1716454222718508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222567400, "dur": 1929,
    "args": {
      "External id": 84803,
      "cbid": 211, "correlation": 84803
    }
  },
  {
    "ph": "s", "id": 84803, "pid": 76337, "tid": -914061504, "ts": 1716454222567400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222718563, "dur": 94,
    "args": {
      "External id": 84825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84825,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84825, "pid": 5, "tid": 7, "ts": 1716454222718563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569349, "dur": 130,
    "args": {
      "External id": 84825,
      "cbid": 211, "correlation": 84825
    }
  },
  {
    "ph": "s", "id": 84825, "pid": 76337, "tid": -914061504, "ts": 1716454222569349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569556, "dur": 1,
    "args": {
      "External id": 84836,
      "cbid": 251, "correlation": 84836
    }
  },
  {
    "ph": "f", "id": 84836, "pid": 76337, "tid": -914061504, "ts": 1716454222569556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222718658, "dur": 105,
    "args": {
      "External id": 84837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84837,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84837, "pid": 5, "tid": 7, "ts": 1716454222718658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569561, "dur": 14,
    "args": {
      "External id": 84837,
      "cbid": 211, "correlation": 84837
    }
  },
  {
    "ph": "s", "id": 84837, "pid": 76337, "tid": -914061504, "ts": 1716454222569561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569635, "dur": 1,
    "args": {
      "External id": 84848,
      "cbid": 251, "correlation": 84848
    }
  },
  {
    "ph": "f", "id": 84848, "pid": 76337, "tid": -914061504, "ts": 1716454222569635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569639, "dur": 0,
    "args": {
      "External id": 84849,
      "cbid": 251, "correlation": 84849
    }
  },
  {
    "ph": "f", "id": 84849, "pid": 76337, "tid": -914061504, "ts": 1716454222569639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222718765, "dur": 10,
    "args": {
      "External id": 84850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84850,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 84850, "pid": 5, "tid": 7, "ts": 1716454222718765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569641, "dur": 13,
    "args": {
      "External id": 84850,
      "cbid": 211, "correlation": 84850
    }
  },
  {
    "ph": "s", "id": 84850, "pid": 76337, "tid": -914061504, "ts": 1716454222569641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222718776, "dur": 5,
    "args": {
      "External id": 84852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84852,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 84852, "pid": 5, "tid": 7, "ts": 1716454222718776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569655, "dur": 6,
    "args": {
      "External id": 84852,
      "cbid": 211, "correlation": 84852
    }
  },
  {
    "ph": "s", "id": 84852, "pid": 76337, "tid": -914061504, "ts": 1716454222569655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569716, "dur": 1,
    "args": {
      "External id": 84863,
      "cbid": 251, "correlation": 84863
    }
  },
  {
    "ph": "f", "id": 84863, "pid": 76337, "tid": -914061504, "ts": 1716454222569716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569720, "dur": 0,
    "args": {
      "External id": 84864,
      "cbid": 251, "correlation": 84864
    }
  },
  {
    "ph": "f", "id": 84864, "pid": 76337, "tid": -914061504, "ts": 1716454222569720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222718782, "dur": 6,
    "args": {
      "External id": 84865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84865,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 84865, "pid": 5, "tid": 7, "ts": 1716454222718782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569721, "dur": 12,
    "args": {
      "External id": 84865,
      "cbid": 211, "correlation": 84865
    }
  },
  {
    "ph": "s", "id": 84865, "pid": 76337, "tid": -914061504, "ts": 1716454222569721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222718790, "dur": 3,
    "args": {
      "External id": 84867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84867,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 84867, "pid": 5, "tid": 7, "ts": 1716454222718790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569734, "dur": 7,
    "args": {
      "External id": 84867,
      "cbid": 211, "correlation": 84867
    }
  },
  {
    "ph": "s", "id": 84867, "pid": 76337, "tid": -914061504, "ts": 1716454222569734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222718795, "dur": 159,
    "args": {
      "External id": 84888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84888,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84888, "pid": 5, "tid": 7, "ts": 1716454222718795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569810, "dur": 12,
    "args": {
      "External id": 84888,
      "cbid": 211, "correlation": 84888
    }
  },
  {
    "ph": "s", "id": 84888, "pid": 76337, "tid": -914061504, "ts": 1716454222569810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569908, "dur": 1,
    "args": {
      "External id": 84906,
      "cbid": 251, "correlation": 84906
    }
  },
  {
    "ph": "f", "id": 84906, "pid": 76337, "tid": -914061504, "ts": 1716454222569908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222718955, "dur": 109,
    "args": {
      "External id": 84908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84908,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 84908, "pid": 5, "tid": 7, "ts": 1716454222718955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569914, "dur": 13,
    "args": {
      "External id": 84908,
      "cbid": 211, "correlation": 84908
    }
  },
  {
    "ph": "s", "id": 84908, "pid": 76337, "tid": -914061504, "ts": 1716454222569914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222719066, "dur": 35,
    "args": {
      "External id": 84916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84916,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84916, "pid": 5, "tid": 7, "ts": 1716454222719066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222569992, "dur": 13,
    "args": {
      "External id": 84916,
      "cbid": 211, "correlation": 84916
    }
  },
  {
    "ph": "s", "id": 84916, "pid": 76337, "tid": -914061504, "ts": 1716454222569992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222719102, "dur": 67,
    "args": {
      "External id": 84924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84924, "pid": 5, "tid": 7, "ts": 1716454222719102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570034, "dur": 9,
    "args": {
      "External id": 84924,
      "cbid": 211, "correlation": 84924
    }
  },
  {
    "ph": "s", "id": 84924, "pid": 76337, "tid": -914061504, "ts": 1716454222570034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222719170, "dur": 94,
    "args": {
      "External id": 84946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84946,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84946, "pid": 5, "tid": 7, "ts": 1716454222719170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570086, "dur": 10,
    "args": {
      "External id": 84946,
      "cbid": 211, "correlation": 84946
    }
  },
  {
    "ph": "s", "id": 84946, "pid": 76337, "tid": -914061504, "ts": 1716454222570086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570174, "dur": 1,
    "args": {
      "External id": 84962,
      "cbid": 251, "correlation": 84962
    }
  },
  {
    "ph": "f", "id": 84962, "pid": 76337, "tid": -914061504, "ts": 1716454222570174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222719265, "dur": 584,
    "args": {
      "External id": 84964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84964,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 84964, "pid": 5, "tid": 7, "ts": 1716454222719265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570180, "dur": 13,
    "args": {
      "External id": 84964,
      "cbid": 211, "correlation": 84964
    }
  },
  {
    "ph": "s", "id": 84964, "pid": 76337, "tid": -914061504, "ts": 1716454222570180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222719851, "dur": 247,
    "args": {
      "External id": 84972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84972, "pid": 5, "tid": 7, "ts": 1716454222719851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570246, "dur": 13,
    "args": {
      "External id": 84972,
      "cbid": 211, "correlation": 84972
    }
  },
  {
    "ph": "s", "id": 84972, "pid": 76337, "tid": -914061504, "ts": 1716454222570246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222720099, "dur": 256,
    "args": {
      "External id": 84980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84980,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 84980, "pid": 5, "tid": 7, "ts": 1716454222720099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570276, "dur": 8,
    "args": {
      "External id": 84980,
      "cbid": 211, "correlation": 84980
    }
  },
  {
    "ph": "s", "id": 84980, "pid": 76337, "tid": -914061504, "ts": 1716454222570276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570357, "dur": 1,
    "args": {
      "External id": 84996,
      "cbid": 251, "correlation": 84996
    }
  },
  {
    "ph": "f", "id": 84996, "pid": 76337, "tid": -914061504, "ts": 1716454222570357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570362, "dur": 0,
    "args": {
      "External id": 84998,
      "cbid": 251, "correlation": 84998
    }
  },
  {
    "ph": "f", "id": 84998, "pid": 76337, "tid": -914061504, "ts": 1716454222570362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222720356, "dur": 360,
    "args": {
      "External id": 84999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 84999,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 84999, "pid": 5, "tid": 7, "ts": 1716454222720356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570365, "dur": 13,
    "args": {
      "External id": 84999,
      "cbid": 211, "correlation": 84999
    }
  },
  {
    "ph": "s", "id": 84999, "pid": 76337, "tid": -914061504, "ts": 1716454222570365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222720718, "dur": 50,
    "args": {
      "External id": 85007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85007,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85007, "pid": 5, "tid": 7, "ts": 1716454222720718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570407, "dur": 196,
    "args": {
      "External id": 85007,
      "cbid": 211, "correlation": 85007
    }
  },
  {
    "ph": "s", "id": 85007, "pid": 76337, "tid": -914061504, "ts": 1716454222570407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222720769, "dur": 161,
    "args": {
      "External id": 85018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85018, "pid": 5, "tid": 7, "ts": 1716454222720769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570660, "dur": 72,
    "args": {
      "External id": 85018,
      "cbid": 211, "correlation": 85018
    }
  },
  {
    "ph": "s", "id": 85018, "pid": 76337, "tid": -914061504, "ts": 1716454222570660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570786, "dur": 0,
    "args": {
      "External id": 85030,
      "cbid": 317, "correlation": 85030
    }
  },
  {
    "ph": "f", "id": 85030, "pid": 76337, "tid": -914061504, "ts": 1716454222570786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570787, "dur": 0,
    "args": {
      "External id": 85031,
      "cbid": 203, "correlation": 85031
    }
  },
  {
    "ph": "f", "id": 85031, "pid": 76337, "tid": -914061504, "ts": 1716454222570787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570788, "dur": 0,
    "args": {
      "External id": 85032,
      "cbid": 205, "correlation": 85032
    }
  },
  {
    "ph": "f", "id": 85032, "pid": 76337, "tid": -914061504, "ts": 1716454222570788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570810, "dur": 1,
    "args": {
      "External id": 85036,
      "cbid": 251, "correlation": 85036
    }
  },
  {
    "ph": "f", "id": 85036, "pid": 76337, "tid": -914061504, "ts": 1716454222570810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570812, "dur": 0,
    "args": {
      "External id": 85037,
      "cbid": 251, "correlation": 85037
    }
  },
  {
    "ph": "f", "id": 85037, "pid": 76337, "tid": -914061504, "ts": 1716454222570812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570813, "dur": 0,
    "args": {
      "External id": 85038,
      "cbid": 251, "correlation": 85038
    }
  },
  {
    "ph": "f", "id": 85038, "pid": 76337, "tid": -914061504, "ts": 1716454222570813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570814, "dur": 0,
    "args": {
      "External id": 85039,
      "cbid": 251, "correlation": 85039
    }
  },
  {
    "ph": "f", "id": 85039, "pid": 76337, "tid": -914061504, "ts": 1716454222570814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570815, "dur": 0,
    "args": {
      "External id": 85040,
      "cbid": 251, "correlation": 85040
    }
  },
  {
    "ph": "f", "id": 85040, "pid": 76337, "tid": -914061504, "ts": 1716454222570815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570815, "dur": 0,
    "args": {
      "External id": 85041,
      "cbid": 251, "correlation": 85041
    }
  },
  {
    "ph": "f", "id": 85041, "pid": 76337, "tid": -914061504, "ts": 1716454222570815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570816, "dur": 0,
    "args": {
      "External id": 85042,
      "cbid": 251, "correlation": 85042
    }
  },
  {
    "ph": "f", "id": 85042, "pid": 76337, "tid": -914061504, "ts": 1716454222570816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570817, "dur": 0,
    "args": {
      "External id": 85043,
      "cbid": 251, "correlation": 85043
    }
  },
  {
    "ph": "f", "id": 85043, "pid": 76337, "tid": -914061504, "ts": 1716454222570817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570818, "dur": 0,
    "args": {
      "External id": 85044,
      "cbid": 251, "correlation": 85044
    }
  },
  {
    "ph": "f", "id": 85044, "pid": 76337, "tid": -914061504, "ts": 1716454222570818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222720931, "dur": 117,
    "args": {
      "External id": 85045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85045,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 85045, "pid": 5, "tid": 7, "ts": 1716454222720931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570820, "dur": 43,
    "args": {
      "External id": 85045,
      "cbid": 211, "correlation": 85045
    }
  },
  {
    "ph": "s", "id": 85045, "pid": 76337, "tid": -914061504, "ts": 1716454222570820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222721050, "dur": 60,
    "args": {
      "External id": 85051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85051, "pid": 5, "tid": 7, "ts": 1716454222721050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222570885, "dur": 286,
    "args": {
      "External id": 85051,
      "cbid": 211, "correlation": 85051
    }
  },
  {
    "ph": "s", "id": 85051, "pid": 76337, "tid": -914061504, "ts": 1716454222570885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222721111, "dur": 50,
    "args": {
      "External id": 85059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85059,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85059, "pid": 5, "tid": 7, "ts": 1716454222721111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571195, "dur": 9,
    "args": {
      "External id": 85059,
      "cbid": 211, "correlation": 85059
    }
  },
  {
    "ph": "s", "id": 85059, "pid": 76337, "tid": -914061504, "ts": 1716454222571195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222721162, "dur": 53,
    "args": {
      "External id": 85079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85079,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 85079, "pid": 5, "tid": 7, "ts": 1716454222721162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571276, "dur": 12,
    "args": {
      "External id": 85079,
      "cbid": 211, "correlation": 85079
    }
  },
  {
    "ph": "s", "id": 85079, "pid": 76337, "tid": -914061504, "ts": 1716454222571276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222721217, "dur": 4,
    "args": {
      "External id": 85091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85091,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 85091, "pid": 5, "tid": 7, "ts": 1716454222721217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571298, "dur": 10,
    "args": {
      "External id": 85091,
      "cbid": 211, "correlation": 85091
    }
  },
  {
    "ph": "s", "id": 85091, "pid": 76337, "tid": -914061504, "ts": 1716454222571298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222721223, "dur": 56,
    "args": {
      "External id": 85094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85094,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85094, "pid": 5, "tid": 7, "ts": 1716454222721223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571320, "dur": 111,
    "args": {
      "External id": 85094,
      "cbid": 211, "correlation": 85094
    }
  },
  {
    "ph": "s", "id": 85094, "pid": 76337, "tid": -914061504, "ts": 1716454222571320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222721280, "dur": 37,
    "args": {
      "External id": 85103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85103,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85103, "pid": 5, "tid": 7, "ts": 1716454222721280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571471, "dur": 10,
    "args": {
      "External id": 85103,
      "cbid": 211, "correlation": 85103
    }
  },
  {
    "ph": "s", "id": 85103, "pid": 76337, "tid": -914061504, "ts": 1716454222571471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571526, "dur": 0,
    "args": {
      "External id": 85113,
      "cbid": 317, "correlation": 85113
    }
  },
  {
    "ph": "f", "id": 85113, "pid": 76337, "tid": -914061504, "ts": 1716454222571526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571527, "dur": 0,
    "args": {
      "External id": 85114,
      "cbid": 203, "correlation": 85114
    }
  },
  {
    "ph": "f", "id": 85114, "pid": 76337, "tid": -914061504, "ts": 1716454222571527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571528, "dur": 0,
    "args": {
      "External id": 85115,
      "cbid": 205, "correlation": 85115
    }
  },
  {
    "ph": "f", "id": 85115, "pid": 76337, "tid": -914061504, "ts": 1716454222571528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222721318, "dur": 43,
    "args": {
      "External id": 85119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85119,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85119, "pid": 5, "tid": 7, "ts": 1716454222721318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571544, "dur": 12,
    "args": {
      "External id": 85119,
      "cbid": 211, "correlation": 85119
    }
  },
  {
    "ph": "s", "id": 85119, "pid": 76337, "tid": -914061504, "ts": 1716454222571544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222721362, "dur": 4,
    "args": {
      "External id": 85121,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85121,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.52
    }
  },
  {
    "ph": "f", "id": 85121, "pid": 5, "tid": 7, "ts": 1716454222721362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571559, "dur": 17,
    "args": {
      "External id": 85121,
      "cbid": 51, "correlation": 85121
    }
  },
  {
    "ph": "s", "id": 85121, "pid": 76337, "tid": -914061504, "ts": 1716454222571559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571581, "dur": 1,
    "args": {
      "External id": 85123,
      "cbid": 200, "correlation": 85123
    }
  },
  {
    "ph": "f", "id": 85123, "pid": 76337, "tid": -914061504, "ts": 1716454222571581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571583, "dur": 0,
    "args": {
      "External id": 85124,
      "cbid": 200, "correlation": 85124
    }
  },
  {
    "ph": "f", "id": 85124, "pid": 76337, "tid": -914061504, "ts": 1716454222571583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571583, "dur": 0,
    "args": {
      "External id": 85125,
      "cbid": 200, "correlation": 85125
    }
  },
  {
    "ph": "f", "id": 85125, "pid": 76337, "tid": -914061504, "ts": 1716454222571583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571584, "dur": 0,
    "args": {
      "External id": 85126,
      "cbid": 200, "correlation": 85126
    }
  },
  {
    "ph": "f", "id": 85126, "pid": 76337, "tid": -914061504, "ts": 1716454222571584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571585, "dur": 4,
    "args": {
      "External id": 85127,
      "cbid": 15, "correlation": 85127
    }
  },
  {
    "ph": "f", "id": 85127, "pid": 76337, "tid": -914061504, "ts": 1716454222571585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571590, "dur": 1,
    "args": {
      "External id": 85128,
      "cbid": 251, "correlation": 85128
    }
  },
  {
    "ph": "f", "id": 85128, "pid": 76337, "tid": -914061504, "ts": 1716454222571590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222721367, "dur": 24,
    "args": {
      "External id": 85129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85129,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85129, "pid": 5, "tid": 7, "ts": 1716454222721367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571593, "dur": 8,
    "args": {
      "External id": 85129,
      "cbid": 211, "correlation": 85129
    }
  },
  {
    "ph": "s", "id": 85129, "pid": 76337, "tid": -914061504, "ts": 1716454222571593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222721392, "dur": 4,
    "args": {
      "External id": 85131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 85131, "pid": 5, "tid": 7, "ts": 1716454222721392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571606, "dur": 7,
    "args": {
      "External id": 85131,
      "cbid": 211, "correlation": 85131
    }
  },
  {
    "ph": "s", "id": 85131, "pid": 76337, "tid": -914061504, "ts": 1716454222571606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571617, "dur": 0,
    "args": {
      "External id": 85132,
      "cbid": 51, "correlation": 85132
    }
  },
  {
    "ph": "s", "id": 85132, "pid": 76337, "tid": -914061504, "ts": 1716454222571617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222721397, "dur": 191,
    "args": {
      "External id": 85133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85133,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85133, "pid": 5, "tid": 7, "ts": 1716454222721397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571618, "dur": 211,
    "args": {
      "External id": 85133,
      "cbid": 211, "correlation": 85133
    }
  },
  {
    "ph": "s", "id": 85133, "pid": 76337, "tid": -914061504, "ts": 1716454222571618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222721589, "dur": 6,
    "args": {
      "External id": 85134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85134, "pid": 5, "tid": 7, "ts": 1716454222721589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571832, "dur": 6,
    "args": {
      "External id": 85134,
      "cbid": 211, "correlation": 85134
    }
  },
  {
    "ph": "s", "id": 85134, "pid": 76337, "tid": -914061504, "ts": 1716454222571832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222721597, "dur": 5,
    "args": {
      "External id": 85140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 85140, "pid": 5, "tid": 7, "ts": 1716454222721597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222571862, "dur": 9,
    "args": {
      "External id": 85140,
      "cbid": 211, "correlation": 85140
    }
  },
  {
    "ph": "s", "id": 85140, "pid": 76337, "tid": -914061504, "ts": 1716454222571862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222721603, "dur": 3,
    "args": {
      "External id": 85148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85148,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85148, "pid": 5, "tid": 7, "ts": 1716454222721603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222573552, "dur": 14,
    "args": {
      "External id": 85148,
      "cbid": 211, "correlation": 85148
    }
  },
  {
    "ph": "s", "id": 85148, "pid": 76337, "tid": -914061504, "ts": 1716454222573552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222721607, "dur": 3,
    "args": {
      "External id": 85156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85156,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85156, "pid": 5, "tid": 7, "ts": 1716454222721607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222573593, "dur": 11,
    "args": {
      "External id": 85156,
      "cbid": 211, "correlation": 85156
    }
  },
  {
    "ph": "s", "id": 85156, "pid": 76337, "tid": -914061504, "ts": 1716454222573593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222721611, "dur": 3,
    "args": {
      "External id": 85164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85164,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85164, "pid": 5, "tid": 7, "ts": 1716454222721611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222573621, "dur": 8,
    "args": {
      "External id": 85164,
      "cbid": 211, "correlation": 85164
    }
  },
  {
    "ph": "s", "id": 85164, "pid": 76337, "tid": -914061504, "ts": 1716454222573621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222721615, "dur": 3,
    "args": {
      "External id": 85173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85173,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85173, "pid": 5, "tid": 7, "ts": 1716454222721615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222573798, "dur": 14,
    "args": {
      "External id": 85173,
      "cbid": 211, "correlation": 85173
    }
  },
  {
    "ph": "s", "id": 85173, "pid": 76337, "tid": -914061504, "ts": 1716454222573798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222721619, "dur": 3,
    "args": {
      "External id": 85182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85182,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85182, "pid": 5, "tid": 7, "ts": 1716454222721619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222573827, "dur": 7,
    "args": {
      "External id": 85182,
      "cbid": 211, "correlation": 85182
    }
  },
  {
    "ph": "s", "id": 85182, "pid": 76337, "tid": -914061504, "ts": 1716454222573827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222721623, "dur": 3,
    "args": {
      "External id": 85190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85190,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85190, "pid": 5, "tid": 7, "ts": 1716454222721623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222573852, "dur": 8,
    "args": {
      "External id": 85190,
      "cbid": 211, "correlation": 85190
    }
  },
  {
    "ph": "s", "id": 85190, "pid": 76337, "tid": -914061504, "ts": 1716454222573852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222721628, "dur": 3,
    "args": {
      "External id": 85198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85198,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85198, "pid": 5, "tid": 7, "ts": 1716454222721628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222574120, "dur": 16,
    "args": {
      "External id": 85198,
      "cbid": 211, "correlation": 85198
    }
  },
  {
    "ph": "s", "id": 85198, "pid": 76337, "tid": -914061504, "ts": 1716454222574120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222721632, "dur": 3,
    "args": {
      "External id": 85206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85206,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85206, "pid": 5, "tid": 7, "ts": 1716454222721632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222574152, "dur": 7,
    "args": {
      "External id": 85206,
      "cbid": 211, "correlation": 85206
    }
  },
  {
    "ph": "s", "id": 85206, "pid": 76337, "tid": -914061504, "ts": 1716454222574152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222721637, "dur": 1,
    "args": {
      "External id": 85216,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85216,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85216, "pid": 5, "tid": 7, "ts": 1716454222721637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222574215, "dur": 36,
    "args": {
      "External id": 85216,
      "cbid": 41, "correlation": 85216
    }
  },
  {
    "ph": "s", "id": 85216, "pid": 76337, "tid": -914061504, "ts": 1716454222574215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222574252, "dur": 147413,
    "args": {
      "External id": 85217,
      "cbid": 131, "correlation": 85217
    }
  },
  {
    "ph": "f", "id": 85217, "pid": 76337, "tid": -914061504, "ts": 1716454222574252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222722022, "dur": 3,
    "args": {
      "External id": 85225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85225,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85225, "pid": 5, "tid": 7, "ts": 1716454222722022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222721958, "dur": 67,
    "args": {
      "External id": 85225,
      "cbid": 211, "correlation": 85225
    }
  },
  {
    "ph": "s", "id": 85225, "pid": 76337, "tid": -914061504, "ts": 1716454222721958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722148, "dur": 3,
    "args": {
      "External id": 85234,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85234,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 85234, "pid": 5, "tid": 7, "ts": 1716454222722148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722103, "dur": 45,
    "args": {
      "External id": 85234,
      "cbid": 41, "correlation": 85234
    }
  },
  {
    "ph": "s", "id": 85234, "pid": 76337, "tid": -914061504, "ts": 1716454222722103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222722260, "dur": 4,
    "args": {
      "External id": 85244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85244,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85244, "pid": 5, "tid": 7, "ts": 1716454222722260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722242, "dur": 19,
    "args": {
      "External id": 85244,
      "cbid": 211, "correlation": 85244
    }
  },
  {
    "ph": "s", "id": 85244, "pid": 76337, "tid": -914061504, "ts": 1716454222722242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722362, "dur": 1,
    "args": {
      "External id": 85254,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85254,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85254, "pid": 5, "tid": 7, "ts": 1716454222722362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722339, "dur": 21,
    "args": {
      "External id": 85254,
      "cbid": 41, "correlation": 85254
    }
  },
  {
    "ph": "s", "id": 85254, "pid": 76337, "tid": -914061504, "ts": 1716454222722339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722362, "dur": 9,
    "args": {
      "External id": 85255,
      "cbid": 131, "correlation": 85255
    }
  },
  {
    "ph": "f", "id": 85255, "pid": 76337, "tid": -914061504, "ts": 1716454222722362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722444, "dur": 3,
    "args": {
      "External id": 85262,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85262,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.415841584158414
    }
  },
  {
    "ph": "f", "id": 85262, "pid": 5, "tid": 7, "ts": 1716454222722444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722420, "dur": 23,
    "args": {
      "External id": 85262,
      "cbid": 41, "correlation": 85262
    }
  },
  {
    "ph": "s", "id": 85262, "pid": 76337, "tid": -914061504, "ts": 1716454222722420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722541, "dur": 3,
    "args": {
      "External id": 85281,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85281,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85281, "pid": 5, "tid": 7, "ts": 1716454222722541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722521, "dur": 19,
    "args": {
      "External id": 85281,
      "cbid": 41, "correlation": 85281
    }
  },
  {
    "ph": "s", "id": 85281, "pid": 76337, "tid": -914061504, "ts": 1716454222722521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222722580, "dur": 3,
    "args": {
      "External id": 85287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85287, "pid": 5, "tid": 7, "ts": 1716454222722580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722569, "dur": 11,
    "args": {
      "External id": 85287,
      "cbid": 211, "correlation": 85287
    }
  },
  {
    "ph": "s", "id": 85287, "pid": 76337, "tid": -914061504, "ts": 1716454222722569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222722595, "dur": 6,
    "args": {
      "External id": 85289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85289,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 85289, "pid": 5, "tid": 7, "ts": 1716454222722595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722584, "dur": 9,
    "args": {
      "External id": 85289,
      "cbid": 211, "correlation": 85289
    }
  },
  {
    "ph": "s", "id": 85289, "pid": 76337, "tid": -914061504, "ts": 1716454222722584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454222722604, "dur": 3,
    "args": {
      "External id": 85291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85291,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85291, "pid": 5, "tid": 7, "ts": 1716454222722604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722595, "dur": 7,
    "args": {
      "External id": 85291,
      "cbid": 211, "correlation": 85291
    }
  },
  {
    "ph": "s", "id": 85291, "pid": 76337, "tid": -914061504, "ts": 1716454222722595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722638, "dur": 3,
    "args": {
      "External id": 85299,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85299,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002631578947368421
    }
  },
  {
    "ph": "f", "id": 85299, "pid": 5, "tid": 7, "ts": 1716454222722638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722623, "dur": 14,
    "args": {
      "External id": 85299,
      "cbid": 41, "correlation": 85299
    }
  },
  {
    "ph": "s", "id": 85299, "pid": 76337, "tid": -914061504, "ts": 1716454222722623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222722688, "dur": 3,
    "args": {
      "External id": 85313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85313, "pid": 5, "tid": 7, "ts": 1716454222722688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722676, "dur": 13,
    "args": {
      "External id": 85313,
      "cbid": 211, "correlation": 85313
    }
  },
  {
    "ph": "s", "id": 85313, "pid": 76337, "tid": -914061504, "ts": 1716454222722676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222722708, "dur": 2,
    "args": {
      "External id": 85327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85327, "pid": 5, "tid": 7, "ts": 1716454222722708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722700, "dur": 6,
    "args": {
      "External id": 85327,
      "cbid": 211, "correlation": 85327
    }
  },
  {
    "ph": "s", "id": 85327, "pid": 76337, "tid": -914061504, "ts": 1716454222722700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222722744, "dur": 6,
    "args": {
      "External id": 85334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85334,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85334, "pid": 5, "tid": 7, "ts": 1716454222722744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722734, "dur": 11,
    "args": {
      "External id": 85334,
      "cbid": 211, "correlation": 85334
    }
  },
  {
    "ph": "s", "id": 85334, "pid": 76337, "tid": -914061504, "ts": 1716454222722734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222722755, "dur": 6,
    "args": {
      "External id": 85337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85337,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85337, "pid": 5, "tid": 7, "ts": 1716454222722755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722747, "dur": 7,
    "args": {
      "External id": 85337,
      "cbid": 211, "correlation": 85337
    }
  },
  {
    "ph": "s", "id": 85337, "pid": 76337, "tid": -914061504, "ts": 1716454222722747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222722763, "dur": 3,
    "args": {
      "External id": 85339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85339,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85339, "pid": 5, "tid": 7, "ts": 1716454222722763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722756, "dur": 6,
    "args": {
      "External id": 85339,
      "cbid": 211, "correlation": 85339
    }
  },
  {
    "ph": "s", "id": 85339, "pid": 76337, "tid": -914061504, "ts": 1716454222722756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722784, "dur": 2,
    "args": {
      "External id": 85342,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85342,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 85342, "pid": 5, "tid": 7, "ts": 1716454222722784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722771, "dur": 12,
    "args": {
      "External id": 85342,
      "cbid": 41, "correlation": 85342
    }
  },
  {
    "ph": "s", "id": 85342, "pid": 76337, "tid": -914061504, "ts": 1716454222722771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222722838, "dur": 4,
    "args": {
      "External id": 85358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85358,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85358, "pid": 5, "tid": 7, "ts": 1716454222722838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722824, "dur": 14,
    "args": {
      "External id": 85358,
      "cbid": 211, "correlation": 85358
    }
  },
  {
    "ph": "s", "id": 85358, "pid": 76337, "tid": -914061504, "ts": 1716454222722824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222722858, "dur": 3,
    "args": {
      "External id": 85363,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85363,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003125
    }
  },
  {
    "ph": "f", "id": 85363, "pid": 5, "tid": 7, "ts": 1716454222722858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722844, "dur": 13,
    "args": {
      "External id": 85363,
      "cbid": 41, "correlation": 85363
    }
  },
  {
    "ph": "s", "id": 85363, "pid": 76337, "tid": -914061504, "ts": 1716454222722844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222722886, "dur": 1,
    "args": {
      "External id": 85369,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85369,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 85369, "pid": 5, "tid": 7, "ts": 1716454222722886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722867, "dur": 28,
    "args": {
      "External id": 85369,
      "cbid": 41, "correlation": 85369
    }
  },
  {
    "ph": "s", "id": 85369, "pid": 76337, "tid": -914061504, "ts": 1716454222722867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722896, "dur": 4,
    "args": {
      "External id": 85370,
      "cbid": 131, "correlation": 85370
    }
  },
  {
    "ph": "f", "id": 85370, "pid": 76337, "tid": -914061504, "ts": 1716454222722896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222722956, "dur": 3,
    "args": {
      "External id": 85378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85378,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85378, "pid": 5, "tid": 7, "ts": 1716454222722956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722939, "dur": 17,
    "args": {
      "External id": 85378,
      "cbid": 211, "correlation": 85378
    }
  },
  {
    "ph": "s", "id": 85378, "pid": 76337, "tid": -914061504, "ts": 1716454222722939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222722997, "dur": 3,
    "args": {
      "External id": 85388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85388,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85388, "pid": 5, "tid": 7, "ts": 1716454222722997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222722985, "dur": 11,
    "args": {
      "External id": 85388,
      "cbid": 211, "correlation": 85388
    }
  },
  {
    "ph": "s", "id": 85388, "pid": 76337, "tid": -914061504, "ts": 1716454222722985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222723027, "dur": 3,
    "args": {
      "External id": 85397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85397,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85397, "pid": 5, "tid": 7, "ts": 1716454222723027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723017, "dur": 9,
    "args": {
      "External id": 85397,
      "cbid": 211, "correlation": 85397
    }
  },
  {
    "ph": "s", "id": 85397, "pid": 76337, "tid": -914061504, "ts": 1716454222723017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222723158, "dur": 12,
    "args": {
      "External id": 85407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85407,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85407, "pid": 5, "tid": 7, "ts": 1716454222723158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723143, "dur": 16,
    "args": {
      "External id": 85407,
      "cbid": 211, "correlation": 85407
    }
  },
  {
    "ph": "s", "id": 85407, "pid": 76337, "tid": -914061504, "ts": 1716454222723143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222723199, "dur": 3,
    "args": {
      "External id": 85415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85415,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85415, "pid": 5, "tid": 7, "ts": 1716454222723199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723189, "dur": 8,
    "args": {
      "External id": 85415,
      "cbid": 211, "correlation": 85415
    }
  },
  {
    "ph": "s", "id": 85415, "pid": 76337, "tid": -914061504, "ts": 1716454222723189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222723247, "dur": 12,
    "args": {
      "External id": 85425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85425,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85425, "pid": 5, "tid": 7, "ts": 1716454222723247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723236, "dur": 11,
    "args": {
      "External id": 85425,
      "cbid": 211, "correlation": 85425
    }
  },
  {
    "ph": "s", "id": 85425, "pid": 76337, "tid": -914061504, "ts": 1716454222723236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222723284, "dur": 10,
    "args": {
      "External id": 85433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85433,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85433, "pid": 5, "tid": 7, "ts": 1716454222723284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723274, "dur": 9,
    "args": {
      "External id": 85433,
      "cbid": 211, "correlation": 85433
    }
  },
  {
    "ph": "s", "id": 85433, "pid": 76337, "tid": -914061504, "ts": 1716454222723274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222723313, "dur": 3,
    "args": {
      "External id": 85442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85442, "pid": 5, "tid": 7, "ts": 1716454222723313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723303, "dur": 9,
    "args": {
      "External id": 85442,
      "cbid": 211, "correlation": 85442
    }
  },
  {
    "ph": "s", "id": 85442, "pid": 76337, "tid": -914061504, "ts": 1716454222723303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222723337, "dur": 5,
    "args": {
      "External id": 85451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85451,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85451, "pid": 5, "tid": 7, "ts": 1716454222723337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723329, "dur": 8,
    "args": {
      "External id": 85451,
      "cbid": 211, "correlation": 85451
    }
  },
  {
    "ph": "s", "id": 85451, "pid": 76337, "tid": -914061504, "ts": 1716454222723329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222723380, "dur": 8,
    "args": {
      "External id": 85461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85461,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85461, "pid": 5, "tid": 7, "ts": 1716454222723380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723368, "dur": 12,
    "args": {
      "External id": 85461,
      "cbid": 211, "correlation": 85461
    }
  },
  {
    "ph": "s", "id": 85461, "pid": 76337, "tid": -914061504, "ts": 1716454222723368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222723729, "dur": 3,
    "args": {
      "External id": 85470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85470,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85470, "pid": 5, "tid": 7, "ts": 1716454222723729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723714, "dur": 16,
    "args": {
      "External id": 85470,
      "cbid": 211, "correlation": 85470
    }
  },
  {
    "ph": "s", "id": 85470, "pid": 76337, "tid": -914061504, "ts": 1716454222723714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222723761, "dur": 3,
    "args": {
      "External id": 85478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85478,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85478, "pid": 5, "tid": 7, "ts": 1716454222723761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723750, "dur": 11,
    "args": {
      "External id": 85478,
      "cbid": 211, "correlation": 85478
    }
  },
  {
    "ph": "s", "id": 85478, "pid": 76337, "tid": -914061504, "ts": 1716454222723750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222723814, "dur": 1,
    "args": {
      "External id": 85488,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85488,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85488, "pid": 5, "tid": 7, "ts": 1716454222723814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723799, "dur": 13,
    "args": {
      "External id": 85488,
      "cbid": 41, "correlation": 85488
    }
  },
  {
    "ph": "s", "id": 85488, "pid": 76337, "tid": -914061504, "ts": 1716454222723799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723813, "dur": 9,
    "args": {
      "External id": 85489,
      "cbid": 131, "correlation": 85489
    }
  },
  {
    "ph": "f", "id": 85489, "pid": 76337, "tid": -914061504, "ts": 1716454222723813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222723907, "dur": 2,
    "args": {
      "External id": 85497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85497, "pid": 5, "tid": 7, "ts": 1716454222723907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723893, "dur": 15,
    "args": {
      "External id": 85497,
      "cbid": 211, "correlation": 85497
    }
  },
  {
    "ph": "s", "id": 85497, "pid": 76337, "tid": -914061504, "ts": 1716454222723893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222723990, "dur": 3,
    "args": {
      "External id": 85506,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85506,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002577319587628866
    }
  },
  {
    "ph": "f", "id": 85506, "pid": 5, "tid": 7, "ts": 1716454222723990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222723964, "dur": 26,
    "args": {
      "External id": 85506,
      "cbid": 41, "correlation": 85506
    }
  },
  {
    "ph": "s", "id": 85506, "pid": 76337, "tid": -914061504, "ts": 1716454222723964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222724067, "dur": 3,
    "args": {
      "External id": 85516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85516,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85516, "pid": 5, "tid": 7, "ts": 1716454222724067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724052, "dur": 15,
    "args": {
      "External id": 85516,
      "cbid": 211, "correlation": 85516
    }
  },
  {
    "ph": "s", "id": 85516, "pid": 76337, "tid": -914061504, "ts": 1716454222724052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222724120, "dur": 1,
    "args": {
      "External id": 85526,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85526,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85526, "pid": 5, "tid": 7, "ts": 1716454222724120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724104, "dur": 14,
    "args": {
      "External id": 85526,
      "cbid": 41, "correlation": 85526
    }
  },
  {
    "ph": "s", "id": 85526, "pid": 76337, "tid": -914061504, "ts": 1716454222724104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724119, "dur": 8,
    "args": {
      "External id": 85527,
      "cbid": 131, "correlation": 85527
    }
  },
  {
    "ph": "f", "id": 85527, "pid": 76337, "tid": -914061504, "ts": 1716454222724119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222724181, "dur": 3,
    "args": {
      "External id": 85534,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85534,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 85534, "pid": 5, "tid": 7, "ts": 1716454222724181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724160, "dur": 20,
    "args": {
      "External id": 85534,
      "cbid": 41, "correlation": 85534
    }
  },
  {
    "ph": "s", "id": 85534, "pid": 76337, "tid": -914061504, "ts": 1716454222724160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222724228, "dur": 1,
    "args": {
      "External id": 85545,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85545,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 85545, "pid": 5, "tid": 7, "ts": 1716454222724228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724216, "dur": 10,
    "args": {
      "External id": 85545,
      "cbid": 41, "correlation": 85545
    }
  },
  {
    "ph": "s", "id": 85545, "pid": 76337, "tid": -914061504, "ts": 1716454222724216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724227, "dur": 8,
    "args": {
      "External id": 85546,
      "cbid": 131, "correlation": 85546
    }
  },
  {
    "ph": "f", "id": 85546, "pid": 76337, "tid": -914061504, "ts": 1716454222724227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222724276, "dur": 3,
    "args": {
      "External id": 85554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85554,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85554, "pid": 5, "tid": 7, "ts": 1716454222724276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724263, "dur": 13,
    "args": {
      "External id": 85554,
      "cbid": 211, "correlation": 85554
    }
  },
  {
    "ph": "s", "id": 85554, "pid": 76337, "tid": -914061504, "ts": 1716454222724263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222724306, "dur": 3,
    "args": {
      "External id": 85564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85564,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85564, "pid": 5, "tid": 7, "ts": 1716454222724306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724297, "dur": 8,
    "args": {
      "External id": 85564,
      "cbid": 211, "correlation": 85564
    }
  },
  {
    "ph": "s", "id": 85564, "pid": 76337, "tid": -914061504, "ts": 1716454222724297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222724328, "dur": 3,
    "args": {
      "External id": 85573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85573,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85573, "pid": 5, "tid": 7, "ts": 1716454222724328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724320, "dur": 7,
    "args": {
      "External id": 85573,
      "cbid": 211, "correlation": 85573
    }
  },
  {
    "ph": "s", "id": 85573, "pid": 76337, "tid": -914061504, "ts": 1716454222724320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222724397, "dur": 6,
    "args": {
      "External id": 85581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85581,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85581, "pid": 5, "tid": 7, "ts": 1716454222724397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724384, "dur": 14,
    "args": {
      "External id": 85581,
      "cbid": 211, "correlation": 85581
    }
  },
  {
    "ph": "s", "id": 85581, "pid": 76337, "tid": -914061504, "ts": 1716454222724384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222724437, "dur": 3,
    "args": {
      "External id": 85590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85590,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85590, "pid": 5, "tid": 7, "ts": 1716454222724437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724427, "dur": 9,
    "args": {
      "External id": 85590,
      "cbid": 211, "correlation": 85590
    }
  },
  {
    "ph": "s", "id": 85590, "pid": 76337, "tid": -914061504, "ts": 1716454222724427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222724460, "dur": 3,
    "args": {
      "External id": 85599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85599,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85599, "pid": 5, "tid": 7, "ts": 1716454222724460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724451, "dur": 8,
    "args": {
      "External id": 85599,
      "cbid": 211, "correlation": 85599
    }
  },
  {
    "ph": "s", "id": 85599, "pid": 76337, "tid": -914061504, "ts": 1716454222724451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222724526, "dur": 3,
    "args": {
      "External id": 85607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85607,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85607, "pid": 5, "tid": 7, "ts": 1716454222724526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724515, "dur": 10,
    "args": {
      "External id": 85607,
      "cbid": 211, "correlation": 85607
    }
  },
  {
    "ph": "s", "id": 85607, "pid": 76337, "tid": -914061504, "ts": 1716454222724515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222724584, "dur": 1,
    "args": {
      "External id": 85615,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85615,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0043859649122807015
    }
  },
  {
    "ph": "f", "id": 85615, "pid": 5, "tid": 7, "ts": 1716454222724584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724568, "dur": 26,
    "args": {
      "External id": 85615,
      "cbid": 41, "correlation": 85615
    }
  },
  {
    "ph": "s", "id": 85615, "pid": 76337, "tid": -914061504, "ts": 1716454222724568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724594, "dur": 3,
    "args": {
      "External id": 85616,
      "cbid": 131, "correlation": 85616
    }
  },
  {
    "ph": "f", "id": 85616, "pid": 76337, "tid": -914061504, "ts": 1716454222724594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222724656, "dur": 1,
    "args": {
      "External id": 85626,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85626,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 85626, "pid": 5, "tid": 7, "ts": 1716454222724656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724643, "dur": 11,
    "args": {
      "External id": 85626,
      "cbid": 41, "correlation": 85626
    }
  },
  {
    "ph": "s", "id": 85626, "pid": 76337, "tid": -914061504, "ts": 1716454222724643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724655, "dur": 8,
    "args": {
      "External id": 85627,
      "cbid": 131, "correlation": 85627
    }
  },
  {
    "ph": "f", "id": 85627, "pid": 76337, "tid": -914061504, "ts": 1716454222724655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222724759, "dur": 1,
    "args": {
      "External id": 85636,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85636,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85636, "pid": 5, "tid": 7, "ts": 1716454222724759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724704, "dur": 53,
    "args": {
      "External id": 85636,
      "cbid": 41, "correlation": 85636
    }
  },
  {
    "ph": "s", "id": 85636, "pid": 76337, "tid": -914061504, "ts": 1716454222724704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724758, "dur": 8,
    "args": {
      "External id": 85637,
      "cbid": 131, "correlation": 85637
    }
  },
  {
    "ph": "f", "id": 85637, "pid": 76337, "tid": -914061504, "ts": 1716454222724758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222724834, "dur": 4,
    "args": {
      "External id": 85644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85644,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85644, "pid": 5, "tid": 7, "ts": 1716454222724834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724817, "dur": 17,
    "args": {
      "External id": 85644,
      "cbid": 211, "correlation": 85644
    }
  },
  {
    "ph": "s", "id": 85644, "pid": 76337, "tid": -914061504, "ts": 1716454222724817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454222724871, "dur": 4,
    "args": {
      "External id": 85664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85664,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85664, "pid": 5, "tid": 7, "ts": 1716454222724871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724859, "dur": 12,
    "args": {
      "External id": 85664,
      "cbid": 211, "correlation": 85664
    }
  },
  {
    "ph": "s", "id": 85664, "pid": 76337, "tid": -914061504, "ts": 1716454222724859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724872, "dur": 0,
    "args": {
      "External id": 85665,
      "cbid": 11, "correlation": 85665
    }
  },
  {
    "ph": "f", "id": 85665, "pid": 76337, "tid": -914061504, "ts": 1716454222724872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724872, "dur": 0,
    "args": {
      "External id": 85666,
      "cbid": 11, "correlation": 85666
    }
  },
  {
    "ph": "f", "id": 85666, "pid": 76337, "tid": -914061504, "ts": 1716454222724872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222724886, "dur": 1,
    "args": {
      "External id": 85669,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85669,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 85669, "pid": 5, "tid": 7, "ts": 1716454222724886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724873, "dur": 21,
    "args": {
      "External id": 85669,
      "cbid": 41, "correlation": 85669
    }
  },
  {
    "ph": "s", "id": 85669, "pid": 76337, "tid": -914061504, "ts": 1716454222724873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724896, "dur": 3,
    "args": {
      "External id": 85670,
      "cbid": 131, "correlation": 85670
    }
  },
  {
    "ph": "f", "id": 85670, "pid": 76337, "tid": -914061504, "ts": 1716454222724896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222724924, "dur": 3,
    "args": {
      "External id": 85694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85694, "pid": 5, "tid": 7, "ts": 1716454222724924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724914, "dur": 9,
    "args": {
      "External id": 85694,
      "cbid": 211, "correlation": 85694
    }
  },
  {
    "ph": "s", "id": 85694, "pid": 76337, "tid": -914061504, "ts": 1716454222724914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724924, "dur": 0,
    "args": {
      "External id": 85695,
      "cbid": 11, "correlation": 85695
    }
  },
  {
    "ph": "f", "id": 85695, "pid": 76337, "tid": -914061504, "ts": 1716454222724924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724924, "dur": 0,
    "args": {
      "External id": 85696,
      "cbid": 11, "correlation": 85696
    }
  },
  {
    "ph": "f", "id": 85696, "pid": 76337, "tid": -914061504, "ts": 1716454222724924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724926, "dur": 1,
    "args": {
      "External id": 85698,
      "cbid": 200, "correlation": 85698
    }
  },
  {
    "ph": "f", "id": 85698, "pid": 76337, "tid": -914061504, "ts": 1716454222724926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222724937, "dur": 4,
    "args": {
      "External id": 85700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85700,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85700, "pid": 5, "tid": 7, "ts": 1716454222724937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724929, "dur": 8,
    "args": {
      "External id": 85700,
      "cbid": 211, "correlation": 85700
    }
  },
  {
    "ph": "s", "id": 85700, "pid": 76337, "tid": -914061504, "ts": 1716454222724929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724938, "dur": 0,
    "args": {
      "External id": 85701,
      "cbid": 11, "correlation": 85701
    }
  },
  {
    "ph": "f", "id": 85701, "pid": 76337, "tid": -914061504, "ts": 1716454222724938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724938, "dur": 0,
    "args": {
      "External id": 85702,
      "cbid": 11, "correlation": 85702
    }
  },
  {
    "ph": "f", "id": 85702, "pid": 76337, "tid": -914061504, "ts": 1716454222724938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222724984, "dur": 1,
    "args": {
      "External id": 85709,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85709,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0047169811320754715
    }
  },
  {
    "ph": "f", "id": 85709, "pid": 5, "tid": 7, "ts": 1716454222724984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724965, "dur": 28,
    "args": {
      "External id": 85709,
      "cbid": 41, "correlation": 85709
    }
  },
  {
    "ph": "s", "id": 85709, "pid": 76337, "tid": -914061504, "ts": 1716454222724965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222724993, "dur": 3,
    "args": {
      "External id": 85710,
      "cbid": 131, "correlation": 85710
    }
  },
  {
    "ph": "f", "id": 85710, "pid": 76337, "tid": -914061504, "ts": 1716454222724993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222725046, "dur": 1,
    "args": {
      "External id": 85720,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 85720,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 85720, "pid": 5, "tid": 7, "ts": 1716454222725046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725034, "dur": 10,
    "args": {
      "External id": 85720,
      "cbid": 41, "correlation": 85720
    }
  },
  {
    "ph": "s", "id": 85720, "pid": 76337, "tid": -914061504, "ts": 1716454222725034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725045, "dur": 7,
    "args": {
      "External id": 85721,
      "cbid": 131, "correlation": 85721
    }
  },
  {
    "ph": "f", "id": 85721, "pid": 76337, "tid": -914061504, "ts": 1716454222725045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222725115, "dur": 5,
    "args": {
      "External id": 85728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85728,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85728, "pid": 5, "tid": 7, "ts": 1716454222725115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725100, "dur": 16,
    "args": {
      "External id": 85728,
      "cbid": 211, "correlation": 85728
    }
  },
  {
    "ph": "s", "id": 85728, "pid": 76337, "tid": -914061504, "ts": 1716454222725100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222725185, "dur": 3,
    "args": {
      "External id": 85737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85737,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85737, "pid": 5, "tid": 7, "ts": 1716454222725185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725171, "dur": 13,
    "args": {
      "External id": 85737,
      "cbid": 211, "correlation": 85737
    }
  },
  {
    "ph": "s", "id": 85737, "pid": 76337, "tid": -914061504, "ts": 1716454222725171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222725220, "dur": 3,
    "args": {
      "External id": 85745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85745,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85745, "pid": 5, "tid": 7, "ts": 1716454222725220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725210, "dur": 9,
    "args": {
      "External id": 85745,
      "cbid": 211, "correlation": 85745
    }
  },
  {
    "ph": "s", "id": 85745, "pid": 76337, "tid": -914061504, "ts": 1716454222725210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222725252, "dur": 4,
    "args": {
      "External id": 85753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85753,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85753, "pid": 5, "tid": 7, "ts": 1716454222725252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725241, "dur": 11,
    "args": {
      "External id": 85753,
      "cbid": 211, "correlation": 85753
    }
  },
  {
    "ph": "s", "id": 85753, "pid": 76337, "tid": -914061504, "ts": 1716454222725241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222725283, "dur": 4,
    "args": {
      "External id": 85761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85761,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85761, "pid": 5, "tid": 7, "ts": 1716454222725283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725271, "dur": 11,
    "args": {
      "External id": 85761,
      "cbid": 211, "correlation": 85761
    }
  },
  {
    "ph": "s", "id": 85761, "pid": 76337, "tid": -914061504, "ts": 1716454222725271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222725309, "dur": 3,
    "args": {
      "External id": 85769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85769,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85769, "pid": 5, "tid": 7, "ts": 1716454222725309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725300, "dur": 8,
    "args": {
      "External id": 85769,
      "cbid": 211, "correlation": 85769
    }
  },
  {
    "ph": "s", "id": 85769, "pid": 76337, "tid": -914061504, "ts": 1716454222725300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222725334, "dur": 3,
    "args": {
      "External id": 85777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85777,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85777, "pid": 5, "tid": 7, "ts": 1716454222725334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725325, "dur": 8,
    "args": {
      "External id": 85777,
      "cbid": 211, "correlation": 85777
    }
  },
  {
    "ph": "s", "id": 85777, "pid": 76337, "tid": -914061504, "ts": 1716454222725325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222725356, "dur": 4,
    "args": {
      "External id": 85785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85785, "pid": 5, "tid": 7, "ts": 1716454222725356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725348, "dur": 7,
    "args": {
      "External id": 85785,
      "cbid": 211, "correlation": 85785
    }
  },
  {
    "ph": "s", "id": 85785, "pid": 76337, "tid": -914061504, "ts": 1716454222725348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222725375, "dur": 5,
    "args": {
      "External id": 85793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85793,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85793, "pid": 5, "tid": 7, "ts": 1716454222725375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725367, "dur": 7,
    "args": {
      "External id": 85793,
      "cbid": 211, "correlation": 85793
    }
  },
  {
    "ph": "s", "id": 85793, "pid": 76337, "tid": -914061504, "ts": 1716454222725367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222725394, "dur": 3,
    "args": {
      "External id": 85801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85801,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85801, "pid": 5, "tid": 7, "ts": 1716454222725394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725386, "dur": 7,
    "args": {
      "External id": 85801,
      "cbid": 211, "correlation": 85801
    }
  },
  {
    "ph": "s", "id": 85801, "pid": 76337, "tid": -914061504, "ts": 1716454222725386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222725453, "dur": 4,
    "args": {
      "External id": 85809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85809,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 85809, "pid": 5, "tid": 7, "ts": 1716454222725453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725442, "dur": 11,
    "args": {
      "External id": 85809,
      "cbid": 211, "correlation": 85809
    }
  },
  {
    "ph": "s", "id": 85809, "pid": 76337, "tid": -914061504, "ts": 1716454222725442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222725479, "dur": 4,
    "args": {
      "External id": 85817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85817,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85817, "pid": 5, "tid": 7, "ts": 1716454222725479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725470, "dur": 8,
    "args": {
      "External id": 85817,
      "cbid": 211, "correlation": 85817
    }
  },
  {
    "ph": "s", "id": 85817, "pid": 76337, "tid": -914061504, "ts": 1716454222725470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222725501, "dur": 4,
    "args": {
      "External id": 85825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85825,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85825, "pid": 5, "tid": 7, "ts": 1716454222725501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725492, "dur": 7,
    "args": {
      "External id": 85825,
      "cbid": 211, "correlation": 85825
    }
  },
  {
    "ph": "s", "id": 85825, "pid": 76337, "tid": -914061504, "ts": 1716454222725492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222725521, "dur": 3,
    "args": {
      "External id": 85833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85833,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 85833, "pid": 5, "tid": 7, "ts": 1716454222725521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725512, "dur": 7,
    "args": {
      "External id": 85833,
      "cbid": 211, "correlation": 85833
    }
  },
  {
    "ph": "s", "id": 85833, "pid": 76337, "tid": -914061504, "ts": 1716454222725512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222725933, "dur": 5,
    "args": {
      "External id": 85842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85842,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85842, "pid": 5, "tid": 7, "ts": 1716454222725933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725916, "dur": 17,
    "args": {
      "External id": 85842,
      "cbid": 211, "correlation": 85842
    }
  },
  {
    "ph": "s", "id": 85842, "pid": 76337, "tid": -914061504, "ts": 1716454222725916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222725970, "dur": 5,
    "args": {
      "External id": 85851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85851,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85851, "pid": 5, "tid": 7, "ts": 1716454222725970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222725960, "dur": 9,
    "args": {
      "External id": 85851,
      "cbid": 211, "correlation": 85851
    }
  },
  {
    "ph": "s", "id": 85851, "pid": 76337, "tid": -914061504, "ts": 1716454222725960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222726113, "dur": 3,
    "args": {
      "External id": 85867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85867, "pid": 5, "tid": 7, "ts": 1716454222726113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726097, "dur": 16,
    "args": {
      "External id": 85867,
      "cbid": 211, "correlation": 85867
    }
  },
  {
    "ph": "s", "id": 85867, "pid": 76337, "tid": -914061504, "ts": 1716454222726097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726147, "dur": 3,
    "args": {
      "External id": 85875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85875, "pid": 5, "tid": 7, "ts": 1716454222726147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726137, "dur": 9,
    "args": {
      "External id": 85875,
      "cbid": 211, "correlation": 85875
    }
  },
  {
    "ph": "s", "id": 85875, "pid": 76337, "tid": -914061504, "ts": 1716454222726137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726179, "dur": 3,
    "args": {
      "External id": 85883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85883, "pid": 5, "tid": 7, "ts": 1716454222726179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726170, "dur": 8,
    "args": {
      "External id": 85883,
      "cbid": 211, "correlation": 85883
    }
  },
  {
    "ph": "s", "id": 85883, "pid": 76337, "tid": -914061504, "ts": 1716454222726170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726209, "dur": 4,
    "args": {
      "External id": 85891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85891,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85891, "pid": 5, "tid": 7, "ts": 1716454222726209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726200, "dur": 8,
    "args": {
      "External id": 85891,
      "cbid": 211, "correlation": 85891
    }
  },
  {
    "ph": "s", "id": 85891, "pid": 76337, "tid": -914061504, "ts": 1716454222726200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222726266, "dur": 4,
    "args": {
      "External id": 85903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85903,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85903, "pid": 5, "tid": 7, "ts": 1716454222726266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726253, "dur": 13,
    "args": {
      "External id": 85903,
      "cbid": 211, "correlation": 85903
    }
  },
  {
    "ph": "s", "id": 85903, "pid": 76337, "tid": -914061504, "ts": 1716454222726253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222726312, "dur": 4,
    "args": {
      "External id": 85914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85914, "pid": 5, "tid": 7, "ts": 1716454222726312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726300, "dur": 12,
    "args": {
      "External id": 85914,
      "cbid": 211, "correlation": 85914
    }
  },
  {
    "ph": "s", "id": 85914, "pid": 76337, "tid": -914061504, "ts": 1716454222726300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726344, "dur": 3,
    "args": {
      "External id": 85922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85922, "pid": 5, "tid": 7, "ts": 1716454222726344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726334, "dur": 8,
    "args": {
      "External id": 85922,
      "cbid": 211, "correlation": 85922
    }
  },
  {
    "ph": "s", "id": 85922, "pid": 76337, "tid": -914061504, "ts": 1716454222726334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726376, "dur": 5,
    "args": {
      "External id": 85930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85930,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85930, "pid": 5, "tid": 7, "ts": 1716454222726376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726366, "dur": 10,
    "args": {
      "External id": 85930,
      "cbid": 211, "correlation": 85930
    }
  },
  {
    "ph": "s", "id": 85930, "pid": 76337, "tid": -914061504, "ts": 1716454222726366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726404, "dur": 5,
    "args": {
      "External id": 85938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85938,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85938, "pid": 5, "tid": 7, "ts": 1716454222726404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726394, "dur": 9,
    "args": {
      "External id": 85938,
      "cbid": 211, "correlation": 85938
    }
  },
  {
    "ph": "s", "id": 85938, "pid": 76337, "tid": -914061504, "ts": 1716454222726394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222726435, "dur": 4,
    "args": {
      "External id": 85947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85947,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85947, "pid": 5, "tid": 7, "ts": 1716454222726435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726424, "dur": 10,
    "args": {
      "External id": 85947,
      "cbid": 211, "correlation": 85947
    }
  },
  {
    "ph": "s", "id": 85947, "pid": 76337, "tid": -914061504, "ts": 1716454222726424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222726494, "dur": 5,
    "args": {
      "External id": 85960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85960,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 85960, "pid": 5, "tid": 7, "ts": 1716454222726494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726482, "dur": 13,
    "args": {
      "External id": 85960,
      "cbid": 211, "correlation": 85960
    }
  },
  {
    "ph": "s", "id": 85960, "pid": 76337, "tid": -914061504, "ts": 1716454222726482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222726534, "dur": 5,
    "args": {
      "External id": 85970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85970,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 85970, "pid": 5, "tid": 7, "ts": 1716454222726534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726523, "dur": 10,
    "args": {
      "External id": 85970,
      "cbid": 211, "correlation": 85970
    }
  },
  {
    "ph": "s", "id": 85970, "pid": 76337, "tid": -914061504, "ts": 1716454222726523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726665, "dur": 5,
    "args": {
      "External id": 85987,
      "cbid": 251, "correlation": 85987
    }
  },
  {
    "ph": "f", "id": 85987, "pid": 76337, "tid": -914061504, "ts": 1716454222726665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454222726694, "dur": 11,
    "args": {
      "External id": 85989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85989,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 85989, "pid": 5, "tid": 7, "ts": 1716454222726694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726679, "dur": 16,
    "args": {
      "External id": 85989,
      "cbid": 211, "correlation": 85989
    }
  },
  {
    "ph": "s", "id": 85989, "pid": 76337, "tid": -914061504, "ts": 1716454222726679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222726755, "dur": 3,
    "args": {
      "External id": 85997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 85997,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 85997, "pid": 5, "tid": 7, "ts": 1716454222726755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726743, "dur": 11,
    "args": {
      "External id": 85997,
      "cbid": 211, "correlation": 85997
    }
  },
  {
    "ph": "s", "id": 85997, "pid": 76337, "tid": -914061504, "ts": 1716454222726743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726813, "dur": 2,
    "args": {
      "External id": 86013,
      "cbid": 251, "correlation": 86013
    }
  },
  {
    "ph": "f", "id": 86013, "pid": 76337, "tid": -914061504, "ts": 1716454222726813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726819, "dur": 0,
    "args": {
      "External id": 86015,
      "cbid": 251, "correlation": 86015
    }
  },
  {
    "ph": "f", "id": 86015, "pid": 76337, "tid": -914061504, "ts": 1716454222726819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222726836, "dur": 14,
    "args": {
      "External id": 86016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86016,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86016, "pid": 5, "tid": 7, "ts": 1716454222726836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726822, "dur": 14,
    "args": {
      "External id": 86016,
      "cbid": 211, "correlation": 86016
    }
  },
  {
    "ph": "s", "id": 86016, "pid": 76337, "tid": -914061504, "ts": 1716454222726822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222726852, "dur": 5,
    "args": {
      "External id": 86018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86018,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86018, "pid": 5, "tid": 7, "ts": 1716454222726852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726841, "dur": 9,
    "args": {
      "External id": 86018,
      "cbid": 211, "correlation": 86018
    }
  },
  {
    "ph": "s", "id": 86018, "pid": 76337, "tid": -914061504, "ts": 1716454222726841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726955, "dur": 1,
    "args": {
      "External id": 86028,
      "cbid": 317, "correlation": 86028
    }
  },
  {
    "ph": "f", "id": 86028, "pid": 76337, "tid": -914061504, "ts": 1716454222726955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726957, "dur": 1,
    "args": {
      "External id": 86029,
      "cbid": 203, "correlation": 86029
    }
  },
  {
    "ph": "f", "id": 86029, "pid": 76337, "tid": -914061504, "ts": 1716454222726957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222726959, "dur": 1,
    "args": {
      "External id": 86030,
      "cbid": 205, "correlation": 86030
    }
  },
  {
    "ph": "f", "id": 86030, "pid": 76337, "tid": -914061504, "ts": 1716454222726959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222727026, "dur": 6,
    "args": {
      "External id": 86034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86034,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86034, "pid": 5, "tid": 7, "ts": 1716454222727026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222727010, "dur": 15,
    "args": {
      "External id": 86034,
      "cbid": 211, "correlation": 86034
    }
  },
  {
    "ph": "s", "id": 86034, "pid": 76337, "tid": -914061504, "ts": 1716454222727010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222727036, "dur": 4,
    "args": {
      "External id": 86036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86036,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 86036, "pid": 5, "tid": 7, "ts": 1716454222727036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222727029, "dur": 6,
    "args": {
      "External id": 86036,
      "cbid": 211, "correlation": 86036
    }
  },
  {
    "ph": "s", "id": 86036, "pid": 76337, "tid": -914061504, "ts": 1716454222727029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222727057, "dur": 3,
    "args": {
      "External id": 86038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86038, "pid": 5, "tid": 7, "ts": 1716454222727057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222727047, "dur": 8,
    "args": {
      "External id": 86038,
      "cbid": 211, "correlation": 86038
    }
  },
  {
    "ph": "s", "id": 86038, "pid": 76337, "tid": -914061504, "ts": 1716454222727047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222727062, "dur": 0,
    "args": {
      "External id": 86039,
      "cbid": 51, "correlation": 86039
    }
  },
  {
    "ph": "s", "id": 86039, "pid": 76337, "tid": -914061504, "ts": 1716454222727062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222727072, "dur": 86,
    "args": {
      "External id": 86040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86040,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86040, "pid": 5, "tid": 7, "ts": 1716454222727072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222727063, "dur": 7,
    "args": {
      "External id": 86040,
      "cbid": 211, "correlation": 86040
    }
  },
  {
    "ph": "s", "id": 86040, "pid": 76337, "tid": -914061504, "ts": 1716454222727063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222727159, "dur": 59,
    "args": {
      "External id": 86045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86045, "pid": 5, "tid": 7, "ts": 1716454222727159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222727103, "dur": 10,
    "args": {
      "External id": 86045,
      "cbid": 211, "correlation": 86045
    }
  },
  {
    "ph": "s", "id": 86045, "pid": 76337, "tid": -914061504, "ts": 1716454222727103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222728927, "dur": 53,
    "args": {
      "External id": 86065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86065,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 86065, "pid": 5, "tid": 7, "ts": 1716454222728927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222728910, "dur": 17,
    "args": {
      "External id": 86065,
      "cbid": 211, "correlation": 86065
    }
  },
  {
    "ph": "s", "id": 86065, "pid": 76337, "tid": -914061504, "ts": 1716454222728910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222728981, "dur": 4,
    "args": {
      "External id": 86077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86077,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86077, "pid": 5, "tid": 7, "ts": 1716454222728981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222728939, "dur": 8,
    "args": {
      "External id": 86077,
      "cbid": 211, "correlation": 86077
    }
  },
  {
    "ph": "s", "id": 86077, "pid": 76337, "tid": -914061504, "ts": 1716454222728939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222728987, "dur": 56,
    "args": {
      "External id": 86080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86080,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86080, "pid": 5, "tid": 7, "ts": 1716454222728987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222728963, "dur": 7,
    "args": {
      "External id": 86080,
      "cbid": 211, "correlation": 86080
    }
  },
  {
    "ph": "s", "id": 86080, "pid": 76337, "tid": -914061504, "ts": 1716454222728963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222729044, "dur": 37,
    "args": {
      "External id": 86089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86089,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86089, "pid": 5, "tid": 7, "ts": 1716454222729044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729016, "dur": 11,
    "args": {
      "External id": 86089,
      "cbid": 211, "correlation": 86089
    }
  },
  {
    "ph": "s", "id": 86089, "pid": 76337, "tid": -914061504, "ts": 1716454222729016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729076, "dur": 0,
    "args": {
      "External id": 86099,
      "cbid": 317, "correlation": 86099
    }
  },
  {
    "ph": "f", "id": 86099, "pid": 76337, "tid": -914061504, "ts": 1716454222729076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729077, "dur": 0,
    "args": {
      "External id": 86100,
      "cbid": 203, "correlation": 86100
    }
  },
  {
    "ph": "f", "id": 86100, "pid": 76337, "tid": -914061504, "ts": 1716454222729077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729078, "dur": 0,
    "args": {
      "External id": 86101,
      "cbid": 205, "correlation": 86101
    }
  },
  {
    "ph": "f", "id": 86101, "pid": 76337, "tid": -914061504, "ts": 1716454222729078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222729109, "dur": 41,
    "args": {
      "External id": 86105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86105,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86105, "pid": 5, "tid": 7, "ts": 1716454222729109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729097, "dur": 12,
    "args": {
      "External id": 86105,
      "cbid": 211, "correlation": 86105
    }
  },
  {
    "ph": "s", "id": 86105, "pid": 76337, "tid": -914061504, "ts": 1716454222729097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222729151, "dur": 15,
    "args": {
      "External id": 86107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86107,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86107, "pid": 5, "tid": 7, "ts": 1716454222729151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729111, "dur": 6,
    "args": {
      "External id": 86107,
      "cbid": 211, "correlation": 86107
    }
  },
  {
    "ph": "s", "id": 86107, "pid": 76337, "tid": -914061504, "ts": 1716454222729111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222729167, "dur": 3,
    "args": {
      "External id": 86109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86109,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86109, "pid": 5, "tid": 7, "ts": 1716454222729167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729123, "dur": 6,
    "args": {
      "External id": 86109,
      "cbid": 211, "correlation": 86109
    }
  },
  {
    "ph": "s", "id": 86109, "pid": 76337, "tid": -914061504, "ts": 1716454222729123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729133, "dur": 0,
    "args": {
      "External id": 86110,
      "cbid": 51, "correlation": 86110
    }
  },
  {
    "ph": "s", "id": 86110, "pid": 76337, "tid": -914061504, "ts": 1716454222729133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222729172, "dur": 713,
    "args": {
      "External id": 86111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86111,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86111, "pid": 5, "tid": 7, "ts": 1716454222729172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729134, "dur": 6,
    "args": {
      "External id": 86111,
      "cbid": 211, "correlation": 86111
    }
  },
  {
    "ph": "s", "id": 86111, "pid": 76337, "tid": -914061504, "ts": 1716454222729134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222729886, "dur": 60,
    "args": {
      "External id": 86116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86116,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86116, "pid": 5, "tid": 7, "ts": 1716454222729886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729164, "dur": 9,
    "args": {
      "External id": 86116,
      "cbid": 211, "correlation": 86116
    }
  },
  {
    "ph": "s", "id": 86116, "pid": 76337, "tid": -914061504, "ts": 1716454222729164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222729947, "dur": 4,
    "args": {
      "External id": 86124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86124,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86124, "pid": 5, "tid": 7, "ts": 1716454222729947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729210, "dur": 9,
    "args": {
      "External id": 86124,
      "cbid": 211, "correlation": 86124
    }
  },
  {
    "ph": "s", "id": 86124, "pid": 76337, "tid": -914061504, "ts": 1716454222729210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729278, "dur": 2,
    "args": {
      "External id": 86140,
      "cbid": 251, "correlation": 86140
    }
  },
  {
    "ph": "f", "id": 86140, "pid": 76337, "tid": -914061504, "ts": 1716454222729278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729284, "dur": 0,
    "args": {
      "External id": 86142,
      "cbid": 251, "correlation": 86142
    }
  },
  {
    "ph": "f", "id": 86142, "pid": 76337, "tid": -914061504, "ts": 1716454222729284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222729952, "dur": 9,
    "args": {
      "External id": 86143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86143,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 86143, "pid": 5, "tid": 7, "ts": 1716454222729952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729286, "dur": 11,
    "args": {
      "External id": 86143,
      "cbid": 211, "correlation": 86143
    }
  },
  {
    "ph": "s", "id": 86143, "pid": 76337, "tid": -914061504, "ts": 1716454222729286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222729962, "dur": 4,
    "args": {
      "External id": 86145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86145,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 86145, "pid": 5, "tid": 7, "ts": 1716454222729962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729299, "dur": 7,
    "args": {
      "External id": 86145,
      "cbid": 211, "correlation": 86145
    }
  },
  {
    "ph": "s", "id": 86145, "pid": 76337, "tid": -914061504, "ts": 1716454222729299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222729967, "dur": 55,
    "args": {
      "External id": 86155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86155, "pid": 5, "tid": 7, "ts": 1716454222729967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729363, "dur": 12,
    "args": {
      "External id": 86155,
      "cbid": 211, "correlation": 86155
    }
  },
  {
    "ph": "s", "id": 86155, "pid": 76337, "tid": -914061504, "ts": 1716454222729363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222730023, "dur": 53,
    "args": {
      "External id": 86175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86175,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 86175, "pid": 5, "tid": 7, "ts": 1716454222730023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729430, "dur": 12,
    "args": {
      "External id": 86175,
      "cbid": 211, "correlation": 86175
    }
  },
  {
    "ph": "s", "id": 86175, "pid": 76337, "tid": -914061504, "ts": 1716454222729430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222730078, "dur": 4,
    "args": {
      "External id": 86187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86187,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86187, "pid": 5, "tid": 7, "ts": 1716454222730078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729451, "dur": 6,
    "args": {
      "External id": 86187,
      "cbid": 211, "correlation": 86187
    }
  },
  {
    "ph": "s", "id": 86187, "pid": 76337, "tid": -914061504, "ts": 1716454222729451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222730083, "dur": 56,
    "args": {
      "External id": 86190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86190,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86190, "pid": 5, "tid": 7, "ts": 1716454222730083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729470, "dur": 7,
    "args": {
      "External id": 86190,
      "cbid": 211, "correlation": 86190
    }
  },
  {
    "ph": "s", "id": 86190, "pid": 76337, "tid": -914061504, "ts": 1716454222729470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222730140, "dur": 37,
    "args": {
      "External id": 86199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86199,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86199, "pid": 5, "tid": 7, "ts": 1716454222730140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729511, "dur": 9,
    "args": {
      "External id": 86199,
      "cbid": 211, "correlation": 86199
    }
  },
  {
    "ph": "s", "id": 86199, "pid": 76337, "tid": -914061504, "ts": 1716454222729511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729583, "dur": 0,
    "args": {
      "External id": 86209,
      "cbid": 317, "correlation": 86209
    }
  },
  {
    "ph": "f", "id": 86209, "pid": 76337, "tid": -914061504, "ts": 1716454222729583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729584, "dur": 0,
    "args": {
      "External id": 86210,
      "cbid": 203, "correlation": 86210
    }
  },
  {
    "ph": "f", "id": 86210, "pid": 76337, "tid": -914061504, "ts": 1716454222729584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729585, "dur": 0,
    "args": {
      "External id": 86211,
      "cbid": 205, "correlation": 86211
    }
  },
  {
    "ph": "f", "id": 86211, "pid": 76337, "tid": -914061504, "ts": 1716454222729585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222730179, "dur": 39,
    "args": {
      "External id": 86215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86215,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86215, "pid": 5, "tid": 7, "ts": 1716454222730179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729599, "dur": 13,
    "args": {
      "External id": 86215,
      "cbid": 211, "correlation": 86215
    }
  },
  {
    "ph": "s", "id": 86215, "pid": 76337, "tid": -914061504, "ts": 1716454222729599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222730219, "dur": 15,
    "args": {
      "External id": 86217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86217,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86217, "pid": 5, "tid": 7, "ts": 1716454222730219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729614, "dur": 5,
    "args": {
      "External id": 86217,
      "cbid": 211, "correlation": 86217
    }
  },
  {
    "ph": "s", "id": 86217, "pid": 76337, "tid": -914061504, "ts": 1716454222729614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222730235, "dur": 4,
    "args": {
      "External id": 86219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86219, "pid": 5, "tid": 7, "ts": 1716454222730235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729623, "dur": 5,
    "args": {
      "External id": 86219,
      "cbid": 211, "correlation": 86219
    }
  },
  {
    "ph": "s", "id": 86219, "pid": 76337, "tid": -914061504, "ts": 1716454222729623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729632, "dur": 0,
    "args": {
      "External id": 86220,
      "cbid": 51, "correlation": 86220
    }
  },
  {
    "ph": "s", "id": 86220, "pid": 76337, "tid": -914061504, "ts": 1716454222729632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222730240, "dur": 704,
    "args": {
      "External id": 86221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86221,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86221, "pid": 5, "tid": 7, "ts": 1716454222730240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729632, "dur": 5,
    "args": {
      "External id": 86221,
      "cbid": 211, "correlation": 86221
    }
  },
  {
    "ph": "s", "id": 86221, "pid": 76337, "tid": -914061504, "ts": 1716454222729632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222730945, "dur": 59,
    "args": {
      "External id": 86226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86226, "pid": 5, "tid": 7, "ts": 1716454222730945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729660, "dur": 9,
    "args": {
      "External id": 86226,
      "cbid": 211, "correlation": 86226
    }
  },
  {
    "ph": "s", "id": 86226, "pid": 76337, "tid": -914061504, "ts": 1716454222729660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222731006, "dur": 50,
    "args": {
      "External id": 86234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86234,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86234, "pid": 5, "tid": 7, "ts": 1716454222731006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729693, "dur": 8,
    "args": {
      "External id": 86234,
      "cbid": 211, "correlation": 86234
    }
  },
  {
    "ph": "s", "id": 86234, "pid": 76337, "tid": -914061504, "ts": 1716454222729693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222731058, "dur": 36,
    "args": {
      "External id": 86242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86242,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86242, "pid": 5, "tid": 7, "ts": 1716454222731058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729726, "dur": 10,
    "args": {
      "External id": 86242,
      "cbid": 211, "correlation": 86242
    }
  },
  {
    "ph": "s", "id": 86242, "pid": 76337, "tid": -914061504, "ts": 1716454222729726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222731095, "dur": 53,
    "args": {
      "External id": 86262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86262,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 86262, "pid": 5, "tid": 7, "ts": 1716454222731095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729809, "dur": 12,
    "args": {
      "External id": 86262,
      "cbid": 211, "correlation": 86262
    }
  },
  {
    "ph": "s", "id": 86262, "pid": 76337, "tid": -914061504, "ts": 1716454222729809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222731149, "dur": 4,
    "args": {
      "External id": 86274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86274,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86274, "pid": 5, "tid": 7, "ts": 1716454222731149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729831, "dur": 6,
    "args": {
      "External id": 86274,
      "cbid": 211, "correlation": 86274
    }
  },
  {
    "ph": "s", "id": 86274, "pid": 76337, "tid": -914061504, "ts": 1716454222729831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222731154, "dur": 55,
    "args": {
      "External id": 86277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86277, "pid": 5, "tid": 7, "ts": 1716454222731154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729850, "dur": 6,
    "args": {
      "External id": 86277,
      "cbid": 211, "correlation": 86277
    }
  },
  {
    "ph": "s", "id": 86277, "pid": 76337, "tid": -914061504, "ts": 1716454222729850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729907, "dur": 0,
    "args": {
      "External id": 86288,
      "cbid": 317, "correlation": 86288
    }
  },
  {
    "ph": "f", "id": 86288, "pid": 76337, "tid": -914061504, "ts": 1716454222729907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729908, "dur": 0,
    "args": {
      "External id": 86289,
      "cbid": 203, "correlation": 86289
    }
  },
  {
    "ph": "f", "id": 86289, "pid": 76337, "tid": -914061504, "ts": 1716454222729908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729908, "dur": 0,
    "args": {
      "External id": 86290,
      "cbid": 205, "correlation": 86290
    }
  },
  {
    "ph": "f", "id": 86290, "pid": 76337, "tid": -914061504, "ts": 1716454222729908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729943, "dur": 2,
    "args": {
      "External id": 86294,
      "cbid": 251, "correlation": 86294
    }
  },
  {
    "ph": "f", "id": 86294, "pid": 76337, "tid": -914061504, "ts": 1716454222729943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729947, "dur": 1,
    "args": {
      "External id": 86295,
      "cbid": 251, "correlation": 86295
    }
  },
  {
    "ph": "f", "id": 86295, "pid": 76337, "tid": -914061504, "ts": 1716454222729947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729949, "dur": 1,
    "args": {
      "External id": 86296,
      "cbid": 251, "correlation": 86296
    }
  },
  {
    "ph": "f", "id": 86296, "pid": 76337, "tid": -914061504, "ts": 1716454222729949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729952, "dur": 1,
    "args": {
      "External id": 86297,
      "cbid": 251, "correlation": 86297
    }
  },
  {
    "ph": "f", "id": 86297, "pid": 76337, "tid": -914061504, "ts": 1716454222729952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729954, "dur": 1,
    "args": {
      "External id": 86298,
      "cbid": 251, "correlation": 86298
    }
  },
  {
    "ph": "f", "id": 86298, "pid": 76337, "tid": -914061504, "ts": 1716454222729954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729956, "dur": 1,
    "args": {
      "External id": 86299,
      "cbid": 251, "correlation": 86299
    }
  },
  {
    "ph": "f", "id": 86299, "pid": 76337, "tid": -914061504, "ts": 1716454222729956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729958, "dur": 1,
    "args": {
      "External id": 86300,
      "cbid": 251, "correlation": 86300
    }
  },
  {
    "ph": "f", "id": 86300, "pid": 76337, "tid": -914061504, "ts": 1716454222729958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729959, "dur": 1,
    "args": {
      "External id": 86301,
      "cbid": 251, "correlation": 86301
    }
  },
  {
    "ph": "f", "id": 86301, "pid": 76337, "tid": -914061504, "ts": 1716454222729959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729962, "dur": 0,
    "args": {
      "External id": 86302,
      "cbid": 251, "correlation": 86302
    }
  },
  {
    "ph": "f", "id": 86302, "pid": 76337, "tid": -914061504, "ts": 1716454222729962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222731211, "dur": 116,
    "args": {
      "External id": 86303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86303,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 86303, "pid": 5, "tid": 7, "ts": 1716454222731211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222729966, "dur": 22,
    "args": {
      "External id": 86303,
      "cbid": 211, "correlation": 86303
    }
  },
  {
    "ph": "s", "id": 86303, "pid": 76337, "tid": -914061504, "ts": 1716454222729966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222731328, "dur": 60,
    "args": {
      "External id": 86309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86309, "pid": 5, "tid": 7, "ts": 1716454222731328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730012, "dur": 9,
    "args": {
      "External id": 86309,
      "cbid": 211, "correlation": 86309
    }
  },
  {
    "ph": "s", "id": 86309, "pid": 76337, "tid": -914061504, "ts": 1716454222730012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222731390, "dur": 586,
    "args": {
      "External id": 86318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86318,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86318, "pid": 5, "tid": 7, "ts": 1716454222731390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730109, "dur": 16,
    "args": {
      "External id": 86318,
      "cbid": 211, "correlation": 86318
    }
  },
  {
    "ph": "s", "id": 86318, "pid": 76337, "tid": -914061504, "ts": 1716454222730109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222731978, "dur": 185,
    "args": {
      "External id": 86340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86340,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86340, "pid": 5, "tid": 7, "ts": 1716454222731978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730184, "dur": 12,
    "args": {
      "External id": 86340,
      "cbid": 211, "correlation": 86340
    }
  },
  {
    "ph": "s", "id": 86340, "pid": 76337, "tid": -914061504, "ts": 1716454222730184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730300, "dur": 2,
    "args": {
      "External id": 86351,
      "cbid": 251, "correlation": 86351
    }
  },
  {
    "ph": "f", "id": 86351, "pid": 76337, "tid": -914061504, "ts": 1716454222730300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222732164, "dur": 199,
    "args": {
      "External id": 86352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86352,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86352, "pid": 5, "tid": 7, "ts": 1716454222732164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730308, "dur": 14,
    "args": {
      "External id": 86352,
      "cbid": 211, "correlation": 86352
    }
  },
  {
    "ph": "s", "id": 86352, "pid": 76337, "tid": -914061504, "ts": 1716454222730308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730380, "dur": 1,
    "args": {
      "External id": 86363,
      "cbid": 251, "correlation": 86363
    }
  },
  {
    "ph": "f", "id": 86363, "pid": 76337, "tid": -914061504, "ts": 1716454222730380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222732364, "dur": 193,
    "args": {
      "External id": 86364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86364,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86364, "pid": 5, "tid": 7, "ts": 1716454222732364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730384, "dur": 11,
    "args": {
      "External id": 86364,
      "cbid": 211, "correlation": 86364
    }
  },
  {
    "ph": "s", "id": 86364, "pid": 76337, "tid": -914061504, "ts": 1716454222730384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730448, "dur": 1,
    "args": {
      "External id": 86375,
      "cbid": 251, "correlation": 86375
    }
  },
  {
    "ph": "f", "id": 86375, "pid": 76337, "tid": -914061504, "ts": 1716454222730448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222732558, "dur": 189,
    "args": {
      "External id": 86376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86376,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86376, "pid": 5, "tid": 7, "ts": 1716454222732558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730453, "dur": 12,
    "args": {
      "External id": 86376,
      "cbid": 211, "correlation": 86376
    }
  },
  {
    "ph": "s", "id": 86376, "pid": 76337, "tid": -914061504, "ts": 1716454222730453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222732749, "dur": 18986,
    "args": {
      "External id": 86397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86397,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 86397, "pid": 5, "tid": 7, "ts": 1716454222732749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730563, "dur": 15,
    "args": {
      "External id": 86397,
      "cbid": 211, "correlation": 86397
    }
  },
  {
    "ph": "s", "id": 86397, "pid": 76337, "tid": -914061504, "ts": 1716454222730563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730680, "dur": 2,
    "args": {
      "External id": 86415,
      "cbid": 251, "correlation": 86415
    }
  },
  {
    "ph": "f", "id": 86415, "pid": 76337, "tid": -914061504, "ts": 1716454222730680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222751736, "dur": 203,
    "args": {
      "External id": 86417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86417,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86417, "pid": 5, "tid": 7, "ts": 1716454222751736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730686, "dur": 13,
    "args": {
      "External id": 86417,
      "cbid": 211, "correlation": 86417
    }
  },
  {
    "ph": "s", "id": 86417, "pid": 76337, "tid": -914061504, "ts": 1716454222730686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222751940, "dur": 66,
    "args": {
      "External id": 86425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86425,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86425, "pid": 5, "tid": 7, "ts": 1716454222751940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730760, "dur": 13,
    "args": {
      "External id": 86425,
      "cbid": 211, "correlation": 86425
    }
  },
  {
    "ph": "s", "id": 86425, "pid": 76337, "tid": -914061504, "ts": 1716454222730760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222752008, "dur": 97,
    "args": {
      "External id": 86433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86433,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86433, "pid": 5, "tid": 7, "ts": 1716454222752008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730802, "dur": 9,
    "args": {
      "External id": 86433,
      "cbid": 211, "correlation": 86433
    }
  },
  {
    "ph": "s", "id": 86433, "pid": 76337, "tid": -914061504, "ts": 1716454222730802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222752106, "dur": 55,
    "args": {
      "External id": 86444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86444,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86444, "pid": 5, "tid": 7, "ts": 1716454222752106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730889, "dur": 15,
    "args": {
      "External id": 86444,
      "cbid": 211, "correlation": 86444
    }
  },
  {
    "ph": "s", "id": 86444, "pid": 76337, "tid": -914061504, "ts": 1716454222730889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222752163, "dur": 94,
    "args": {
      "External id": 86466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86466,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86466, "pid": 5, "tid": 7, "ts": 1716454222752163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222730925, "dur": 7,
    "args": {
      "External id": 86466,
      "cbid": 211, "correlation": 86466
    }
  },
  {
    "ph": "s", "id": 86466, "pid": 76337, "tid": -914061504, "ts": 1716454222730925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731016, "dur": 1,
    "args": {
      "External id": 86477,
      "cbid": 251, "correlation": 86477
    }
  },
  {
    "ph": "f", "id": 86477, "pid": 76337, "tid": -914061504, "ts": 1716454222731016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222752258, "dur": 104,
    "args": {
      "External id": 86478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86478,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86478, "pid": 5, "tid": 7, "ts": 1716454222752258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731020, "dur": 14,
    "args": {
      "External id": 86478,
      "cbid": 211, "correlation": 86478
    }
  },
  {
    "ph": "s", "id": 86478, "pid": 76337, "tid": -914061504, "ts": 1716454222731020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731102, "dur": 1,
    "args": {
      "External id": 86489,
      "cbid": 251, "correlation": 86489
    }
  },
  {
    "ph": "f", "id": 86489, "pid": 76337, "tid": -914061504, "ts": 1716454222731102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731106, "dur": 0,
    "args": {
      "External id": 86490,
      "cbid": 251, "correlation": 86490
    }
  },
  {
    "ph": "f", "id": 86490, "pid": 76337, "tid": -914061504, "ts": 1716454222731106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222752364, "dur": 10,
    "args": {
      "External id": 86491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86491,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 86491, "pid": 5, "tid": 7, "ts": 1716454222752364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731108, "dur": 14,
    "args": {
      "External id": 86491,
      "cbid": 211, "correlation": 86491
    }
  },
  {
    "ph": "s", "id": 86491, "pid": 76337, "tid": -914061504, "ts": 1716454222731108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222752375, "dur": 5,
    "args": {
      "External id": 86493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86493,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 86493, "pid": 5, "tid": 7, "ts": 1716454222752375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731125, "dur": 8,
    "args": {
      "External id": 86493,
      "cbid": 211, "correlation": 86493
    }
  },
  {
    "ph": "s", "id": 86493, "pid": 76337, "tid": -914061504, "ts": 1716454222731125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731189, "dur": 1,
    "args": {
      "External id": 86504,
      "cbid": 251, "correlation": 86504
    }
  },
  {
    "ph": "f", "id": 86504, "pid": 76337, "tid": -914061504, "ts": 1716454222731189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731193, "dur": 0,
    "args": {
      "External id": 86505,
      "cbid": 251, "correlation": 86505
    }
  },
  {
    "ph": "f", "id": 86505, "pid": 76337, "tid": -914061504, "ts": 1716454222731193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222752382, "dur": 6,
    "args": {
      "External id": 86506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86506,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 86506, "pid": 5, "tid": 7, "ts": 1716454222752382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731194, "dur": 13,
    "args": {
      "External id": 86506,
      "cbid": 211, "correlation": 86506
    }
  },
  {
    "ph": "s", "id": 86506, "pid": 76337, "tid": -914061504, "ts": 1716454222731194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222752389, "dur": 4,
    "args": {
      "External id": 86508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86508,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 86508, "pid": 5, "tid": 7, "ts": 1716454222752389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731208, "dur": 5,
    "args": {
      "External id": 86508,
      "cbid": 211, "correlation": 86508
    }
  },
  {
    "ph": "s", "id": 86508, "pid": 76337, "tid": -914061504, "ts": 1716454222731208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222752394, "dur": 159,
    "args": {
      "External id": 86529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86529,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 86529, "pid": 5, "tid": 7, "ts": 1716454222752394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731282, "dur": 13,
    "args": {
      "External id": 86529,
      "cbid": 211, "correlation": 86529
    }
  },
  {
    "ph": "s", "id": 86529, "pid": 76337, "tid": -914061504, "ts": 1716454222731282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731383, "dur": 2,
    "args": {
      "External id": 86547,
      "cbid": 251, "correlation": 86547
    }
  },
  {
    "ph": "f", "id": 86547, "pid": 76337, "tid": -914061504, "ts": 1716454222731383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222752554, "dur": 109,
    "args": {
      "External id": 86549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86549,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 86549, "pid": 5, "tid": 7, "ts": 1716454222752554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731389, "dur": 13,
    "args": {
      "External id": 86549,
      "cbid": 211, "correlation": 86549
    }
  },
  {
    "ph": "s", "id": 86549, "pid": 76337, "tid": -914061504, "ts": 1716454222731389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222752665, "dur": 35,
    "args": {
      "External id": 86557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86557,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86557, "pid": 5, "tid": 7, "ts": 1716454222752665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731461, "dur": 12,
    "args": {
      "External id": 86557,
      "cbid": 211, "correlation": 86557
    }
  },
  {
    "ph": "s", "id": 86557, "pid": 76337, "tid": -914061504, "ts": 1716454222731461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222752702, "dur": 68,
    "args": {
      "External id": 86565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86565, "pid": 5, "tid": 7, "ts": 1716454222752702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731504, "dur": 10,
    "args": {
      "External id": 86565,
      "cbid": 211, "correlation": 86565
    }
  },
  {
    "ph": "s", "id": 86565, "pid": 76337, "tid": -914061504, "ts": 1716454222731504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222752771, "dur": 94,
    "args": {
      "External id": 86587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86587,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86587, "pid": 5, "tid": 7, "ts": 1716454222752771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731558, "dur": 10,
    "args": {
      "External id": 86587,
      "cbid": 211, "correlation": 86587
    }
  },
  {
    "ph": "s", "id": 86587, "pid": 76337, "tid": -914061504, "ts": 1716454222731558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731649, "dur": 1,
    "args": {
      "External id": 86603,
      "cbid": 251, "correlation": 86603
    }
  },
  {
    "ph": "f", "id": 86603, "pid": 76337, "tid": -914061504, "ts": 1716454222731649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222752866, "dur": 587,
    "args": {
      "External id": 86605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86605,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86605, "pid": 5, "tid": 7, "ts": 1716454222752866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731655, "dur": 13,
    "args": {
      "External id": 86605,
      "cbid": 211, "correlation": 86605
    }
  },
  {
    "ph": "s", "id": 86605, "pid": 76337, "tid": -914061504, "ts": 1716454222731655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222753454, "dur": 245,
    "args": {
      "External id": 86613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86613,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86613, "pid": 5, "tid": 7, "ts": 1716454222753454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731741, "dur": 15,
    "args": {
      "External id": 86613,
      "cbid": 211, "correlation": 86613
    }
  },
  {
    "ph": "s", "id": 86613, "pid": 76337, "tid": -914061504, "ts": 1716454222731741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222753701, "dur": 254,
    "args": {
      "External id": 86621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86621, "pid": 5, "tid": 7, "ts": 1716454222753701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731780, "dur": 10,
    "args": {
      "External id": 86621,
      "cbid": 211, "correlation": 86621
    }
  },
  {
    "ph": "s", "id": 86621, "pid": 76337, "tid": -914061504, "ts": 1716454222731780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731865, "dur": 2,
    "args": {
      "External id": 86637,
      "cbid": 251, "correlation": 86637
    }
  },
  {
    "ph": "f", "id": 86637, "pid": 76337, "tid": -914061504, "ts": 1716454222731865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731870, "dur": 0,
    "args": {
      "External id": 86639,
      "cbid": 251, "correlation": 86639
    }
  },
  {
    "ph": "f", "id": 86639, "pid": 76337, "tid": -914061504, "ts": 1716454222731870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222753956, "dur": 361,
    "args": {
      "External id": 86640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86640,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 86640, "pid": 5, "tid": 7, "ts": 1716454222753956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731875, "dur": 14,
    "args": {
      "External id": 86640,
      "cbid": 211, "correlation": 86640
    }
  },
  {
    "ph": "s", "id": 86640, "pid": 76337, "tid": -914061504, "ts": 1716454222731875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222754319, "dur": 51,
    "args": {
      "External id": 86648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86648,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86648, "pid": 5, "tid": 7, "ts": 1716454222754319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222731920, "dur": 11,
    "args": {
      "External id": 86648,
      "cbid": 211, "correlation": 86648
    }
  },
  {
    "ph": "s", "id": 86648, "pid": 76337, "tid": -914061504, "ts": 1716454222731920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222754372, "dur": 160,
    "args": {
      "External id": 86659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86659,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86659, "pid": 5, "tid": 7, "ts": 1716454222754372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732002, "dur": 15,
    "args": {
      "External id": 86659,
      "cbid": 211, "correlation": 86659
    }
  },
  {
    "ph": "s", "id": 86659, "pid": 76337, "tid": -914061504, "ts": 1716454222732002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732075, "dur": 0,
    "args": {
      "External id": 86671,
      "cbid": 317, "correlation": 86671
    }
  },
  {
    "ph": "f", "id": 86671, "pid": 76337, "tid": -914061504, "ts": 1716454222732075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732077, "dur": 0,
    "args": {
      "External id": 86672,
      "cbid": 203, "correlation": 86672
    }
  },
  {
    "ph": "f", "id": 86672, "pid": 76337, "tid": -914061504, "ts": 1716454222732077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732077, "dur": 0,
    "args": {
      "External id": 86673,
      "cbid": 205, "correlation": 86673
    }
  },
  {
    "ph": "f", "id": 86673, "pid": 76337, "tid": -914061504, "ts": 1716454222732077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732105, "dur": 1,
    "args": {
      "External id": 86677,
      "cbid": 251, "correlation": 86677
    }
  },
  {
    "ph": "f", "id": 86677, "pid": 76337, "tid": -914061504, "ts": 1716454222732105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732108, "dur": 0,
    "args": {
      "External id": 86678,
      "cbid": 251, "correlation": 86678
    }
  },
  {
    "ph": "f", "id": 86678, "pid": 76337, "tid": -914061504, "ts": 1716454222732108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732109, "dur": 0,
    "args": {
      "External id": 86679,
      "cbid": 251, "correlation": 86679
    }
  },
  {
    "ph": "f", "id": 86679, "pid": 76337, "tid": -914061504, "ts": 1716454222732109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732110, "dur": 0,
    "args": {
      "External id": 86680,
      "cbid": 251, "correlation": 86680
    }
  },
  {
    "ph": "f", "id": 86680, "pid": 76337, "tid": -914061504, "ts": 1716454222732110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732111, "dur": 1,
    "args": {
      "External id": 86681,
      "cbid": 251, "correlation": 86681
    }
  },
  {
    "ph": "f", "id": 86681, "pid": 76337, "tid": -914061504, "ts": 1716454222732111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732112, "dur": 0,
    "args": {
      "External id": 86682,
      "cbid": 251, "correlation": 86682
    }
  },
  {
    "ph": "f", "id": 86682, "pid": 76337, "tid": -914061504, "ts": 1716454222732112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732114, "dur": 0,
    "args": {
      "External id": 86683,
      "cbid": 251, "correlation": 86683
    }
  },
  {
    "ph": "f", "id": 86683, "pid": 76337, "tid": -914061504, "ts": 1716454222732114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732114, "dur": 0,
    "args": {
      "External id": 86684,
      "cbid": 251, "correlation": 86684
    }
  },
  {
    "ph": "f", "id": 86684, "pid": 76337, "tid": -914061504, "ts": 1716454222732114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732116, "dur": 0,
    "args": {
      "External id": 86685,
      "cbid": 251, "correlation": 86685
    }
  },
  {
    "ph": "f", "id": 86685, "pid": 76337, "tid": -914061504, "ts": 1716454222732116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222754533, "dur": 117,
    "args": {
      "External id": 86686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86686,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 86686, "pid": 5, "tid": 7, "ts": 1716454222754533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732118, "dur": 13,
    "args": {
      "External id": 86686,
      "cbid": 211, "correlation": 86686
    }
  },
  {
    "ph": "s", "id": 86686, "pid": 76337, "tid": -914061504, "ts": 1716454222732118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222754651, "dur": 60,
    "args": {
      "External id": 86692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86692, "pid": 5, "tid": 7, "ts": 1716454222754651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732154, "dur": 9,
    "args": {
      "External id": 86692,
      "cbid": 211, "correlation": 86692
    }
  },
  {
    "ph": "s", "id": 86692, "pid": 76337, "tid": -914061504, "ts": 1716454222732154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222754713, "dur": 50,
    "args": {
      "External id": 86700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86700,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86700, "pid": 5, "tid": 7, "ts": 1716454222754713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732189, "dur": 8,
    "args": {
      "External id": 86700,
      "cbid": 211, "correlation": 86700
    }
  },
  {
    "ph": "s", "id": 86700, "pid": 76337, "tid": -914061504, "ts": 1716454222732189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222754764, "dur": 54,
    "args": {
      "External id": 86720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86720,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 86720, "pid": 5, "tid": 7, "ts": 1716454222754764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732267, "dur": 12,
    "args": {
      "External id": 86720,
      "cbid": 211, "correlation": 86720
    }
  },
  {
    "ph": "s", "id": 86720, "pid": 76337, "tid": -914061504, "ts": 1716454222732267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222754819, "dur": 4,
    "args": {
      "External id": 86732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86732,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86732, "pid": 5, "tid": 7, "ts": 1716454222754819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732289, "dur": 7,
    "args": {
      "External id": 86732,
      "cbid": 211, "correlation": 86732
    }
  },
  {
    "ph": "s", "id": 86732, "pid": 76337, "tid": -914061504, "ts": 1716454222732289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222754825, "dur": 57,
    "args": {
      "External id": 86735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86735, "pid": 5, "tid": 7, "ts": 1716454222754825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732309, "dur": 7,
    "args": {
      "External id": 86735,
      "cbid": 211, "correlation": 86735
    }
  },
  {
    "ph": "s", "id": 86735, "pid": 76337, "tid": -914061504, "ts": 1716454222732309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222754883, "dur": 36,
    "args": {
      "External id": 86744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86744,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86744, "pid": 5, "tid": 7, "ts": 1716454222754883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732349, "dur": 11,
    "args": {
      "External id": 86744,
      "cbid": 211, "correlation": 86744
    }
  },
  {
    "ph": "s", "id": 86744, "pid": 76337, "tid": -914061504, "ts": 1716454222732349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732402, "dur": 0,
    "args": {
      "External id": 86754,
      "cbid": 317, "correlation": 86754
    }
  },
  {
    "ph": "f", "id": 86754, "pid": 76337, "tid": -914061504, "ts": 1716454222732402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732403, "dur": 0,
    "args": {
      "External id": 86755,
      "cbid": 203, "correlation": 86755
    }
  },
  {
    "ph": "f", "id": 86755, "pid": 76337, "tid": -914061504, "ts": 1716454222732403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732404, "dur": 0,
    "args": {
      "External id": 86756,
      "cbid": 205, "correlation": 86756
    }
  },
  {
    "ph": "f", "id": 86756, "pid": 76337, "tid": -914061504, "ts": 1716454222732404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222754921, "dur": 42,
    "args": {
      "External id": 86760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86760,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86760, "pid": 5, "tid": 7, "ts": 1716454222754921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732420, "dur": 12,
    "args": {
      "External id": 86760,
      "cbid": 211, "correlation": 86760
    }
  },
  {
    "ph": "s", "id": 86760, "pid": 76337, "tid": -914061504, "ts": 1716454222732420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222754964, "dur": 14,
    "args": {
      "External id": 86762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86762,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86762, "pid": 5, "tid": 7, "ts": 1716454222754964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732434, "dur": 5,
    "args": {
      "External id": 86762,
      "cbid": 211, "correlation": 86762
    }
  },
  {
    "ph": "s", "id": 86762, "pid": 76337, "tid": -914061504, "ts": 1716454222732434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222754980, "dur": 4,
    "args": {
      "External id": 86764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86764,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86764, "pid": 5, "tid": 7, "ts": 1716454222754980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732445, "dur": 6,
    "args": {
      "External id": 86764,
      "cbid": 211, "correlation": 86764
    }
  },
  {
    "ph": "s", "id": 86764, "pid": 76337, "tid": -914061504, "ts": 1716454222732445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732454, "dur": 0,
    "args": {
      "External id": 86765,
      "cbid": 51, "correlation": 86765
    }
  },
  {
    "ph": "s", "id": 86765, "pid": 76337, "tid": -914061504, "ts": 1716454222732454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222754985, "dur": 715,
    "args": {
      "External id": 86766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86766,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86766, "pid": 5, "tid": 7, "ts": 1716454222754985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732455, "dur": 5,
    "args": {
      "External id": 86766,
      "cbid": 211, "correlation": 86766
    }
  },
  {
    "ph": "s", "id": 86766, "pid": 76337, "tid": -914061504, "ts": 1716454222732455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222755702, "dur": 60,
    "args": {
      "External id": 86771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86771, "pid": 5, "tid": 7, "ts": 1716454222755702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732484, "dur": 8,
    "args": {
      "External id": 86771,
      "cbid": 211, "correlation": 86771
    }
  },
  {
    "ph": "s", "id": 86771, "pid": 76337, "tid": -914061504, "ts": 1716454222732484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222755763, "dur": 3,
    "args": {
      "External id": 86779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86779,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86779, "pid": 5, "tid": 7, "ts": 1716454222755763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732527, "dur": 10,
    "args": {
      "External id": 86779,
      "cbid": 211, "correlation": 86779
    }
  },
  {
    "ph": "s", "id": 86779, "pid": 76337, "tid": -914061504, "ts": 1716454222732527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732594, "dur": 2,
    "args": {
      "External id": 86795,
      "cbid": 251, "correlation": 86795
    }
  },
  {
    "ph": "f", "id": 86795, "pid": 76337, "tid": -914061504, "ts": 1716454222732594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732600, "dur": 0,
    "args": {
      "External id": 86797,
      "cbid": 251, "correlation": 86797
    }
  },
  {
    "ph": "f", "id": 86797, "pid": 76337, "tid": -914061504, "ts": 1716454222732600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222755768, "dur": 11,
    "args": {
      "External id": 86798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86798,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 86798, "pid": 5, "tid": 7, "ts": 1716454222755768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732602, "dur": 12,
    "args": {
      "External id": 86798,
      "cbid": 211, "correlation": 86798
    }
  },
  {
    "ph": "s", "id": 86798, "pid": 76337, "tid": -914061504, "ts": 1716454222732602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222755781, "dur": 5,
    "args": {
      "External id": 86800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86800,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 86800, "pid": 5, "tid": 7, "ts": 1716454222755781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732616, "dur": 6,
    "args": {
      "External id": 86800,
      "cbid": 211, "correlation": 86800
    }
  },
  {
    "ph": "s", "id": 86800, "pid": 76337, "tid": -914061504, "ts": 1716454222732616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222755787, "dur": 53,
    "args": {
      "External id": 86810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86810, "pid": 5, "tid": 7, "ts": 1716454222755787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732677, "dur": 12,
    "args": {
      "External id": 86810,
      "cbid": 211, "correlation": 86810
    }
  },
  {
    "ph": "s", "id": 86810, "pid": 76337, "tid": -914061504, "ts": 1716454222732677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222755841, "dur": 51,
    "args": {
      "External id": 86830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86830,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 86830, "pid": 5, "tid": 7, "ts": 1716454222755841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732744, "dur": 11,
    "args": {
      "External id": 86830,
      "cbid": 211, "correlation": 86830
    }
  },
  {
    "ph": "s", "id": 86830, "pid": 76337, "tid": -914061504, "ts": 1716454222732744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222755894, "dur": 4,
    "args": {
      "External id": 86842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86842,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86842, "pid": 5, "tid": 7, "ts": 1716454222755894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732765, "dur": 6,
    "args": {
      "External id": 86842,
      "cbid": 211, "correlation": 86842
    }
  },
  {
    "ph": "s", "id": 86842, "pid": 76337, "tid": -914061504, "ts": 1716454222732765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222755899, "dur": 57,
    "args": {
      "External id": 86845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86845, "pid": 5, "tid": 7, "ts": 1716454222755899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732784, "dur": 6,
    "args": {
      "External id": 86845,
      "cbid": 211, "correlation": 86845
    }
  },
  {
    "ph": "s", "id": 86845, "pid": 76337, "tid": -914061504, "ts": 1716454222732784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222755957, "dur": 38,
    "args": {
      "External id": 86854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86854,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86854, "pid": 5, "tid": 7, "ts": 1716454222755957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732824, "dur": 10,
    "args": {
      "External id": 86854,
      "cbid": 211, "correlation": 86854
    }
  },
  {
    "ph": "s", "id": 86854, "pid": 76337, "tid": -914061504, "ts": 1716454222732824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732889, "dur": 0,
    "args": {
      "External id": 86864,
      "cbid": 317, "correlation": 86864
    }
  },
  {
    "ph": "f", "id": 86864, "pid": 76337, "tid": -914061504, "ts": 1716454222732889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732889, "dur": 0,
    "args": {
      "External id": 86865,
      "cbid": 203, "correlation": 86865
    }
  },
  {
    "ph": "f", "id": 86865, "pid": 76337, "tid": -914061504, "ts": 1716454222732889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732890, "dur": 0,
    "args": {
      "External id": 86866,
      "cbid": 205, "correlation": 86866
    }
  },
  {
    "ph": "f", "id": 86866, "pid": 76337, "tid": -914061504, "ts": 1716454222732890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222755996, "dur": 40,
    "args": {
      "External id": 86870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86870, "pid": 5, "tid": 7, "ts": 1716454222755996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732904, "dur": 12,
    "args": {
      "External id": 86870,
      "cbid": 211, "correlation": 86870
    }
  },
  {
    "ph": "s", "id": 86870, "pid": 76337, "tid": -914061504, "ts": 1716454222732904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222756038, "dur": 15,
    "args": {
      "External id": 86872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86872,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86872, "pid": 5, "tid": 7, "ts": 1716454222756038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732918, "dur": 5,
    "args": {
      "External id": 86872,
      "cbid": 211, "correlation": 86872
    }
  },
  {
    "ph": "s", "id": 86872, "pid": 76337, "tid": -914061504, "ts": 1716454222732918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222756054, "dur": 3,
    "args": {
      "External id": 86874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86874,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86874, "pid": 5, "tid": 7, "ts": 1716454222756054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732927, "dur": 6,
    "args": {
      "External id": 86874,
      "cbid": 211, "correlation": 86874
    }
  },
  {
    "ph": "s", "id": 86874, "pid": 76337, "tid": -914061504, "ts": 1716454222732927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732936, "dur": 0,
    "args": {
      "External id": 86875,
      "cbid": 51, "correlation": 86875
    }
  },
  {
    "ph": "s", "id": 86875, "pid": 76337, "tid": -914061504, "ts": 1716454222732936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222756058, "dur": 705,
    "args": {
      "External id": 86876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86876,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 86876, "pid": 5, "tid": 7, "ts": 1716454222756058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732937, "dur": 5,
    "args": {
      "External id": 86876,
      "cbid": 211, "correlation": 86876
    }
  },
  {
    "ph": "s", "id": 86876, "pid": 76337, "tid": -914061504, "ts": 1716454222732937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222756765, "dur": 60,
    "args": {
      "External id": 86881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86881, "pid": 5, "tid": 7, "ts": 1716454222756765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222732964, "dur": 19,
    "args": {
      "External id": 86881,
      "cbid": 211, "correlation": 86881
    }
  },
  {
    "ph": "s", "id": 86881, "pid": 76337, "tid": -914061504, "ts": 1716454222732964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222756827, "dur": 50,
    "args": {
      "External id": 86889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86889,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86889, "pid": 5, "tid": 7, "ts": 1716454222756827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733008, "dur": 9,
    "args": {
      "External id": 86889,
      "cbid": 211, "correlation": 86889
    }
  },
  {
    "ph": "s", "id": 86889, "pid": 76337, "tid": -914061504, "ts": 1716454222733008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222756878, "dur": 35,
    "args": {
      "External id": 86897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86897,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86897, "pid": 5, "tid": 7, "ts": 1716454222756878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733038, "dur": 9,
    "args": {
      "External id": 86897,
      "cbid": 211, "correlation": 86897
    }
  },
  {
    "ph": "s", "id": 86897, "pid": 76337, "tid": -914061504, "ts": 1716454222733038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222756915, "dur": 52,
    "args": {
      "External id": 86917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86917,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 86917, "pid": 5, "tid": 7, "ts": 1716454222756915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733119, "dur": 12,
    "args": {
      "External id": 86917,
      "cbid": 211, "correlation": 86917
    }
  },
  {
    "ph": "s", "id": 86917, "pid": 76337, "tid": -914061504, "ts": 1716454222733119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222756968, "dur": 4,
    "args": {
      "External id": 86929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86929,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 86929, "pid": 5, "tid": 7, "ts": 1716454222756968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733141, "dur": 6,
    "args": {
      "External id": 86929,
      "cbid": 211, "correlation": 86929
    }
  },
  {
    "ph": "s", "id": 86929, "pid": 76337, "tid": -914061504, "ts": 1716454222733141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222756973, "dur": 55,
    "args": {
      "External id": 86932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86932, "pid": 5, "tid": 7, "ts": 1716454222756973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733159, "dur": 8,
    "args": {
      "External id": 86932,
      "cbid": 211, "correlation": 86932
    }
  },
  {
    "ph": "s", "id": 86932, "pid": 76337, "tid": -914061504, "ts": 1716454222733159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733217, "dur": 0,
    "args": {
      "External id": 86943,
      "cbid": 317, "correlation": 86943
    }
  },
  {
    "ph": "f", "id": 86943, "pid": 76337, "tid": -914061504, "ts": 1716454222733217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733218, "dur": 0,
    "args": {
      "External id": 86944,
      "cbid": 203, "correlation": 86944
    }
  },
  {
    "ph": "f", "id": 86944, "pid": 76337, "tid": -914061504, "ts": 1716454222733218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733218, "dur": 0,
    "args": {
      "External id": 86945,
      "cbid": 205, "correlation": 86945
    }
  },
  {
    "ph": "f", "id": 86945, "pid": 76337, "tid": -914061504, "ts": 1716454222733218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733243, "dur": 1,
    "args": {
      "External id": 86949,
      "cbid": 251, "correlation": 86949
    }
  },
  {
    "ph": "f", "id": 86949, "pid": 76337, "tid": -914061504, "ts": 1716454222733243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733245, "dur": 0,
    "args": {
      "External id": 86950,
      "cbid": 251, "correlation": 86950
    }
  },
  {
    "ph": "f", "id": 86950, "pid": 76337, "tid": -914061504, "ts": 1716454222733245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733246, "dur": 0,
    "args": {
      "External id": 86951,
      "cbid": 251, "correlation": 86951
    }
  },
  {
    "ph": "f", "id": 86951, "pid": 76337, "tid": -914061504, "ts": 1716454222733246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733247, "dur": 0,
    "args": {
      "External id": 86952,
      "cbid": 251, "correlation": 86952
    }
  },
  {
    "ph": "f", "id": 86952, "pid": 76337, "tid": -914061504, "ts": 1716454222733247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733248, "dur": 0,
    "args": {
      "External id": 86953,
      "cbid": 251, "correlation": 86953
    }
  },
  {
    "ph": "f", "id": 86953, "pid": 76337, "tid": -914061504, "ts": 1716454222733248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733249, "dur": 0,
    "args": {
      "External id": 86954,
      "cbid": 251, "correlation": 86954
    }
  },
  {
    "ph": "f", "id": 86954, "pid": 76337, "tid": -914061504, "ts": 1716454222733249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733249, "dur": 0,
    "args": {
      "External id": 86955,
      "cbid": 251, "correlation": 86955
    }
  },
  {
    "ph": "f", "id": 86955, "pid": 76337, "tid": -914061504, "ts": 1716454222733249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733250, "dur": 0,
    "args": {
      "External id": 86956,
      "cbid": 251, "correlation": 86956
    }
  },
  {
    "ph": "f", "id": 86956, "pid": 76337, "tid": -914061504, "ts": 1716454222733250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733251, "dur": 0,
    "args": {
      "External id": 86957,
      "cbid": 251, "correlation": 86957
    }
  },
  {
    "ph": "f", "id": 86957, "pid": 76337, "tid": -914061504, "ts": 1716454222733251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222757030, "dur": 114,
    "args": {
      "External id": 86958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86958,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 86958, "pid": 5, "tid": 7, "ts": 1716454222757030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733253, "dur": 12,
    "args": {
      "External id": 86958,
      "cbid": 211, "correlation": 86958
    }
  },
  {
    "ph": "s", "id": 86958, "pid": 76337, "tid": -914061504, "ts": 1716454222733253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222757145, "dur": 60,
    "args": {
      "External id": 86964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86964,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86964, "pid": 5, "tid": 7, "ts": 1716454222757145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733289, "dur": 9,
    "args": {
      "External id": 86964,
      "cbid": 211, "correlation": 86964
    }
  },
  {
    "ph": "s", "id": 86964, "pid": 76337, "tid": -914061504, "ts": 1716454222733289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222757206, "dur": 449,
    "args": {
      "External id": 86973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86973,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86973, "pid": 5, "tid": 7, "ts": 1716454222757206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733375, "dur": 15,
    "args": {
      "External id": 86973,
      "cbid": 211, "correlation": 86973
    }
  },
  {
    "ph": "s", "id": 86973, "pid": 76337, "tid": -914061504, "ts": 1716454222733375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222757657, "dur": 184,
    "args": {
      "External id": 86995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 86995,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 86995, "pid": 5, "tid": 7, "ts": 1716454222757657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733434, "dur": 10,
    "args": {
      "External id": 86995,
      "cbid": 211, "correlation": 86995
    }
  },
  {
    "ph": "s", "id": 86995, "pid": 76337, "tid": -914061504, "ts": 1716454222733434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733527, "dur": 1,
    "args": {
      "External id": 87006,
      "cbid": 251, "correlation": 87006
    }
  },
  {
    "ph": "f", "id": 87006, "pid": 76337, "tid": -914061504, "ts": 1716454222733527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222757842, "dur": 197,
    "args": {
      "External id": 87007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87007,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87007, "pid": 5, "tid": 7, "ts": 1716454222757842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733532, "dur": 13,
    "args": {
      "External id": 87007,
      "cbid": 211, "correlation": 87007
    }
  },
  {
    "ph": "s", "id": 87007, "pid": 76337, "tid": -914061504, "ts": 1716454222733532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733603, "dur": 1,
    "args": {
      "External id": 87018,
      "cbid": 251, "correlation": 87018
    }
  },
  {
    "ph": "f", "id": 87018, "pid": 76337, "tid": -914061504, "ts": 1716454222733603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222758040, "dur": 188,
    "args": {
      "External id": 87019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87019,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87019, "pid": 5, "tid": 7, "ts": 1716454222758040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733607, "dur": 12,
    "args": {
      "External id": 87019,
      "cbid": 211, "correlation": 87019
    }
  },
  {
    "ph": "s", "id": 87019, "pid": 76337, "tid": -914061504, "ts": 1716454222733607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733669, "dur": 1,
    "args": {
      "External id": 87030,
      "cbid": 251, "correlation": 87030
    }
  },
  {
    "ph": "f", "id": 87030, "pid": 76337, "tid": -914061504, "ts": 1716454222733669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222758229, "dur": 188,
    "args": {
      "External id": 87031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87031,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87031, "pid": 5, "tid": 7, "ts": 1716454222758229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733673, "dur": 12,
    "args": {
      "External id": 87031,
      "cbid": 211, "correlation": 87031
    }
  },
  {
    "ph": "s", "id": 87031, "pid": 76337, "tid": -914061504, "ts": 1716454222733673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222758419, "dur": 19004,
    "args": {
      "External id": 87052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87052,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87052, "pid": 5, "tid": 7, "ts": 1716454222758419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733759, "dur": 14,
    "args": {
      "External id": 87052,
      "cbid": 211, "correlation": 87052
    }
  },
  {
    "ph": "s", "id": 87052, "pid": 76337, "tid": -914061504, "ts": 1716454222733759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733862, "dur": 1,
    "args": {
      "External id": 87070,
      "cbid": 251, "correlation": 87070
    }
  },
  {
    "ph": "f", "id": 87070, "pid": 76337, "tid": -914061504, "ts": 1716454222733862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222777424, "dur": 204,
    "args": {
      "External id": 87072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87072,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87072, "pid": 5, "tid": 7, "ts": 1716454222777424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733868, "dur": 13,
    "args": {
      "External id": 87072,
      "cbid": 211, "correlation": 87072
    }
  },
  {
    "ph": "s", "id": 87072, "pid": 76337, "tid": -914061504, "ts": 1716454222733868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222777630, "dur": 66,
    "args": {
      "External id": 87080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87080,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87080, "pid": 5, "tid": 7, "ts": 1716454222777630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733939, "dur": 13,
    "args": {
      "External id": 87080,
      "cbid": 211, "correlation": 87080
    }
  },
  {
    "ph": "s", "id": 87080, "pid": 76337, "tid": -914061504, "ts": 1716454222733939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222777697, "dur": 98,
    "args": {
      "External id": 87088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87088,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87088, "pid": 5, "tid": 7, "ts": 1716454222777697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222733988, "dur": 9,
    "args": {
      "External id": 87088,
      "cbid": 211, "correlation": 87088
    }
  },
  {
    "ph": "s", "id": 87088, "pid": 76337, "tid": -914061504, "ts": 1716454222733988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222777796, "dur": 55,
    "args": {
      "External id": 87099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87099, "pid": 5, "tid": 7, "ts": 1716454222777796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734063, "dur": 13,
    "args": {
      "External id": 87099,
      "cbid": 211, "correlation": 87099
    }
  },
  {
    "ph": "s", "id": 87099, "pid": 76337, "tid": -914061504, "ts": 1716454222734063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222777852, "dur": 94,
    "args": {
      "External id": 87121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87121,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87121, "pid": 5, "tid": 7, "ts": 1716454222777852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734094, "dur": 8,
    "args": {
      "External id": 87121,
      "cbid": 211, "correlation": 87121
    }
  },
  {
    "ph": "s", "id": 87121, "pid": 76337, "tid": -914061504, "ts": 1716454222734094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734179, "dur": 1,
    "args": {
      "External id": 87132,
      "cbid": 251, "correlation": 87132
    }
  },
  {
    "ph": "f", "id": 87132, "pid": 76337, "tid": -914061504, "ts": 1716454222734179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222777948, "dur": 107,
    "args": {
      "External id": 87133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87133,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87133, "pid": 5, "tid": 7, "ts": 1716454222777948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734184, "dur": 14,
    "args": {
      "External id": 87133,
      "cbid": 211, "correlation": 87133
    }
  },
  {
    "ph": "s", "id": 87133, "pid": 76337, "tid": -914061504, "ts": 1716454222734184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734258, "dur": 1,
    "args": {
      "External id": 87144,
      "cbid": 251, "correlation": 87144
    }
  },
  {
    "ph": "f", "id": 87144, "pid": 76337, "tid": -914061504, "ts": 1716454222734258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734261, "dur": 0,
    "args": {
      "External id": 87145,
      "cbid": 251, "correlation": 87145
    }
  },
  {
    "ph": "f", "id": 87145, "pid": 76337, "tid": -914061504, "ts": 1716454222734261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222778055, "dur": 10,
    "args": {
      "External id": 87146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87146,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 87146, "pid": 5, "tid": 7, "ts": 1716454222778055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734264, "dur": 12,
    "args": {
      "External id": 87146,
      "cbid": 211, "correlation": 87146
    }
  },
  {
    "ph": "s", "id": 87146, "pid": 76337, "tid": -914061504, "ts": 1716454222734264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222778067, "dur": 5,
    "args": {
      "External id": 87148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87148,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 87148, "pid": 5, "tid": 7, "ts": 1716454222778067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734278, "dur": 6,
    "args": {
      "External id": 87148,
      "cbid": 211, "correlation": 87148
    }
  },
  {
    "ph": "s", "id": 87148, "pid": 76337, "tid": -914061504, "ts": 1716454222734278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734339, "dur": 1,
    "args": {
      "External id": 87159,
      "cbid": 251, "correlation": 87159
    }
  },
  {
    "ph": "f", "id": 87159, "pid": 76337, "tid": -914061504, "ts": 1716454222734339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734342, "dur": 0,
    "args": {
      "External id": 87160,
      "cbid": 251, "correlation": 87160
    }
  },
  {
    "ph": "f", "id": 87160, "pid": 76337, "tid": -914061504, "ts": 1716454222734342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222778073, "dur": 6,
    "args": {
      "External id": 87161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87161,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 87161, "pid": 5, "tid": 7, "ts": 1716454222778073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734344, "dur": 12,
    "args": {
      "External id": 87161,
      "cbid": 211, "correlation": 87161
    }
  },
  {
    "ph": "s", "id": 87161, "pid": 76337, "tid": -914061504, "ts": 1716454222734344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222778081, "dur": 4,
    "args": {
      "External id": 87163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87163,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 87163, "pid": 5, "tid": 7, "ts": 1716454222778081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734358, "dur": 6,
    "args": {
      "External id": 87163,
      "cbid": 211, "correlation": 87163
    }
  },
  {
    "ph": "s", "id": 87163, "pid": 76337, "tid": -914061504, "ts": 1716454222734358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222778086, "dur": 158,
    "args": {
      "External id": 87184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87184,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87184, "pid": 5, "tid": 7, "ts": 1716454222778086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734432, "dur": 12,
    "args": {
      "External id": 87184,
      "cbid": 211, "correlation": 87184
    }
  },
  {
    "ph": "s", "id": 87184, "pid": 76337, "tid": -914061504, "ts": 1716454222734432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734530, "dur": 1,
    "args": {
      "External id": 87202,
      "cbid": 251, "correlation": 87202
    }
  },
  {
    "ph": "f", "id": 87202, "pid": 76337, "tid": -914061504, "ts": 1716454222734530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222778245, "dur": 108,
    "args": {
      "External id": 87204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87204,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87204, "pid": 5, "tid": 7, "ts": 1716454222778245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734536, "dur": 13,
    "args": {
      "External id": 87204,
      "cbid": 211, "correlation": 87204
    }
  },
  {
    "ph": "s", "id": 87204, "pid": 76337, "tid": -914061504, "ts": 1716454222734536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222778355, "dur": 35,
    "args": {
      "External id": 87212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87212,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87212, "pid": 5, "tid": 7, "ts": 1716454222778355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734606, "dur": 12,
    "args": {
      "External id": 87212,
      "cbid": 211, "correlation": 87212
    }
  },
  {
    "ph": "s", "id": 87212, "pid": 76337, "tid": -914061504, "ts": 1716454222734606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222778391, "dur": 68,
    "args": {
      "External id": 87220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87220,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87220, "pid": 5, "tid": 7, "ts": 1716454222778391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734647, "dur": 9,
    "args": {
      "External id": 87220,
      "cbid": 211, "correlation": 87220
    }
  },
  {
    "ph": "s", "id": 87220, "pid": 76337, "tid": -914061504, "ts": 1716454222734647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222778460, "dur": 93,
    "args": {
      "External id": 87242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87242,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87242, "pid": 5, "tid": 7, "ts": 1716454222778460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734699, "dur": 10,
    "args": {
      "External id": 87242,
      "cbid": 211, "correlation": 87242
    }
  },
  {
    "ph": "s", "id": 87242, "pid": 76337, "tid": -914061504, "ts": 1716454222734699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734786, "dur": 1,
    "args": {
      "External id": 87258,
      "cbid": 251, "correlation": 87258
    }
  },
  {
    "ph": "f", "id": 87258, "pid": 76337, "tid": -914061504, "ts": 1716454222734786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222778555, "dur": 586,
    "args": {
      "External id": 87260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87260,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87260, "pid": 5, "tid": 7, "ts": 1716454222778555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734792, "dur": 13,
    "args": {
      "External id": 87260,
      "cbid": 211, "correlation": 87260
    }
  },
  {
    "ph": "s", "id": 87260, "pid": 76337, "tid": -914061504, "ts": 1716454222734792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222779142, "dur": 245,
    "args": {
      "External id": 87268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87268,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87268, "pid": 5, "tid": 7, "ts": 1716454222779142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734860, "dur": 14,
    "args": {
      "External id": 87268,
      "cbid": 211, "correlation": 87268
    }
  },
  {
    "ph": "s", "id": 87268, "pid": 76337, "tid": -914061504, "ts": 1716454222734860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222779389, "dur": 252,
    "args": {
      "External id": 87276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87276, "pid": 5, "tid": 7, "ts": 1716454222779389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734893, "dur": 8,
    "args": {
      "External id": 87276,
      "cbid": 211, "correlation": 87276
    }
  },
  {
    "ph": "s", "id": 87276, "pid": 76337, "tid": -914061504, "ts": 1716454222734893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734985, "dur": 1,
    "args": {
      "External id": 87292,
      "cbid": 251, "correlation": 87292
    }
  },
  {
    "ph": "f", "id": 87292, "pid": 76337, "tid": -914061504, "ts": 1716454222734985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734990, "dur": 0,
    "args": {
      "External id": 87294,
      "cbid": 251, "correlation": 87294
    }
  },
  {
    "ph": "f", "id": 87294, "pid": 76337, "tid": -914061504, "ts": 1716454222734990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222779642, "dur": 364,
    "args": {
      "External id": 87295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87295,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 87295, "pid": 5, "tid": 7, "ts": 1716454222779642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222734993, "dur": 13,
    "args": {
      "External id": 87295,
      "cbid": 211, "correlation": 87295
    }
  },
  {
    "ph": "s", "id": 87295, "pid": 76337, "tid": -914061504, "ts": 1716454222734993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222780007, "dur": 50,
    "args": {
      "External id": 87303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87303,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87303, "pid": 5, "tid": 7, "ts": 1716454222780007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735037, "dur": 10,
    "args": {
      "External id": 87303,
      "cbid": 211, "correlation": 87303
    }
  },
  {
    "ph": "s", "id": 87303, "pid": 76337, "tid": -914061504, "ts": 1716454222735037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222780058, "dur": 159,
    "args": {
      "External id": 87314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87314, "pid": 5, "tid": 7, "ts": 1716454222780058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735105, "dur": 12,
    "args": {
      "External id": 87314,
      "cbid": 211, "correlation": 87314
    }
  },
  {
    "ph": "s", "id": 87314, "pid": 76337, "tid": -914061504, "ts": 1716454222735105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735172, "dur": 0,
    "args": {
      "External id": 87326,
      "cbid": 317, "correlation": 87326
    }
  },
  {
    "ph": "f", "id": 87326, "pid": 76337, "tid": -914061504, "ts": 1716454222735172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735173, "dur": 0,
    "args": {
      "External id": 87327,
      "cbid": 203, "correlation": 87327
    }
  },
  {
    "ph": "f", "id": 87327, "pid": 76337, "tid": -914061504, "ts": 1716454222735173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735174, "dur": 0,
    "args": {
      "External id": 87328,
      "cbid": 205, "correlation": 87328
    }
  },
  {
    "ph": "f", "id": 87328, "pid": 76337, "tid": -914061504, "ts": 1716454222735174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735199, "dur": 1,
    "args": {
      "External id": 87332,
      "cbid": 251, "correlation": 87332
    }
  },
  {
    "ph": "f", "id": 87332, "pid": 76337, "tid": -914061504, "ts": 1716454222735199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735201, "dur": 0,
    "args": {
      "External id": 87333,
      "cbid": 251, "correlation": 87333
    }
  },
  {
    "ph": "f", "id": 87333, "pid": 76337, "tid": -914061504, "ts": 1716454222735201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735202, "dur": 0,
    "args": {
      "External id": 87334,
      "cbid": 251, "correlation": 87334
    }
  },
  {
    "ph": "f", "id": 87334, "pid": 76337, "tid": -914061504, "ts": 1716454222735202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735203, "dur": 0,
    "args": {
      "External id": 87335,
      "cbid": 251, "correlation": 87335
    }
  },
  {
    "ph": "f", "id": 87335, "pid": 76337, "tid": -914061504, "ts": 1716454222735203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735204, "dur": 0,
    "args": {
      "External id": 87336,
      "cbid": 251, "correlation": 87336
    }
  },
  {
    "ph": "f", "id": 87336, "pid": 76337, "tid": -914061504, "ts": 1716454222735204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735204, "dur": 0,
    "args": {
      "External id": 87337,
      "cbid": 251, "correlation": 87337
    }
  },
  {
    "ph": "f", "id": 87337, "pid": 76337, "tid": -914061504, "ts": 1716454222735204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735205, "dur": 0,
    "args": {
      "External id": 87338,
      "cbid": 251, "correlation": 87338
    }
  },
  {
    "ph": "f", "id": 87338, "pid": 76337, "tid": -914061504, "ts": 1716454222735205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735206, "dur": 0,
    "args": {
      "External id": 87339,
      "cbid": 251, "correlation": 87339
    }
  },
  {
    "ph": "f", "id": 87339, "pid": 76337, "tid": -914061504, "ts": 1716454222735206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735207, "dur": 0,
    "args": {
      "External id": 87340,
      "cbid": 251, "correlation": 87340
    }
  },
  {
    "ph": "f", "id": 87340, "pid": 76337, "tid": -914061504, "ts": 1716454222735207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222780218, "dur": 114,
    "args": {
      "External id": 87341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87341,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87341, "pid": 5, "tid": 7, "ts": 1716454222780218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735209, "dur": 12,
    "args": {
      "External id": 87341,
      "cbid": 211, "correlation": 87341
    }
  },
  {
    "ph": "s", "id": 87341, "pid": 76337, "tid": -914061504, "ts": 1716454222735209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222780334, "dur": 59,
    "args": {
      "External id": 87347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87347, "pid": 5, "tid": 7, "ts": 1716454222780334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735244, "dur": 10,
    "args": {
      "External id": 87347,
      "cbid": 211, "correlation": 87347
    }
  },
  {
    "ph": "s", "id": 87347, "pid": 76337, "tid": -914061504, "ts": 1716454222735244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222780394, "dur": 50,
    "args": {
      "External id": 87355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87355,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87355, "pid": 5, "tid": 7, "ts": 1716454222780394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735276, "dur": 8,
    "args": {
      "External id": 87355,
      "cbid": 211, "correlation": 87355
    }
  },
  {
    "ph": "s", "id": 87355, "pid": 76337, "tid": -914061504, "ts": 1716454222735276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735351, "dur": 0,
    "args": {
      "External id": 87365,
      "cbid": 317, "correlation": 87365
    }
  },
  {
    "ph": "f", "id": 87365, "pid": 76337, "tid": -914061504, "ts": 1716454222735351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735352, "dur": 0,
    "args": {
      "External id": 87366,
      "cbid": 203, "correlation": 87366
    }
  },
  {
    "ph": "f", "id": 87366, "pid": 76337, "tid": -914061504, "ts": 1716454222735352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735352, "dur": 0,
    "args": {
      "External id": 87367,
      "cbid": 205, "correlation": 87367
    }
  },
  {
    "ph": "f", "id": 87367, "pid": 76337, "tid": -914061504, "ts": 1716454222735352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222780445, "dur": 41,
    "args": {
      "External id": 87371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87371, "pid": 5, "tid": 7, "ts": 1716454222780445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735371, "dur": 12,
    "args": {
      "External id": 87371,
      "cbid": 211, "correlation": 87371
    }
  },
  {
    "ph": "s", "id": 87371, "pid": 76337, "tid": -914061504, "ts": 1716454222735371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222780487, "dur": 14,
    "args": {
      "External id": 87373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87373,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87373, "pid": 5, "tid": 7, "ts": 1716454222780487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735386, "dur": 5,
    "args": {
      "External id": 87373,
      "cbid": 211, "correlation": 87373
    }
  },
  {
    "ph": "s", "id": 87373, "pid": 76337, "tid": -914061504, "ts": 1716454222735386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222780504, "dur": 1,
    "args": {
      "External id": 87375,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 87375,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9230769230769231
    }
  },
  {
    "ph": "f", "id": 87375, "pid": 5, "tid": 7, "ts": 1716454222780504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735406, "dur": 18,
    "args": {
      "External id": 87375,
      "cbid": 51, "correlation": 87375
    }
  },
  {
    "ph": "s", "id": 87375, "pid": 76337, "tid": -914061504, "ts": 1716454222735406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222780507, "dur": 357,
    "args": {
      "External id": 87376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87376,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87376, "pid": 5, "tid": 7, "ts": 1716454222780507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735426, "dur": 10,
    "args": {
      "External id": 87376,
      "cbid": 211, "correlation": 87376
    }
  },
  {
    "ph": "s", "id": 87376, "pid": 76337, "tid": -914061504, "ts": 1716454222735426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222780865, "dur": 13,
    "args": {
      "External id": 87378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87378, "pid": 5, "tid": 7, "ts": 1716454222780865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735444, "dur": 7,
    "args": {
      "External id": 87378,
      "cbid": 211, "correlation": 87378
    }
  },
  {
    "ph": "s", "id": 87378, "pid": 76337, "tid": -914061504, "ts": 1716454222735444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222780880, "dur": 15,
    "args": {
      "External id": 87384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87384, "pid": 5, "tid": 7, "ts": 1716454222780880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735474, "dur": 8,
    "args": {
      "External id": 87384,
      "cbid": 211, "correlation": 87384
    }
  },
  {
    "ph": "s", "id": 87384, "pid": 76337, "tid": -914061504, "ts": 1716454222735474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222780896, "dur": 18,
    "args": {
      "External id": 87404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87404,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 87404, "pid": 5, "tid": 7, "ts": 1716454222780896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735572, "dur": 13,
    "args": {
      "External id": 87404,
      "cbid": 211, "correlation": 87404
    }
  },
  {
    "ph": "s", "id": 87404, "pid": 76337, "tid": -914061504, "ts": 1716454222735572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222780915, "dur": 4,
    "args": {
      "External id": 87416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87416,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 87416, "pid": 5, "tid": 7, "ts": 1716454222780915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735595, "dur": 6,
    "args": {
      "External id": 87416,
      "cbid": 211, "correlation": 87416
    }
  },
  {
    "ph": "s", "id": 87416, "pid": 76337, "tid": -914061504, "ts": 1716454222735595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222780921, "dur": 18,
    "args": {
      "External id": 87419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87419, "pid": 5, "tid": 7, "ts": 1716454222780921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735613, "dur": 7,
    "args": {
      "External id": 87419,
      "cbid": 211, "correlation": 87419
    }
  },
  {
    "ph": "s", "id": 87419, "pid": 76337, "tid": -914061504, "ts": 1716454222735613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222780940, "dur": 11,
    "args": {
      "External id": 87428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87428,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87428, "pid": 5, "tid": 7, "ts": 1716454222780940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735654, "dur": 10,
    "args": {
      "External id": 87428,
      "cbid": 211, "correlation": 87428
    }
  },
  {
    "ph": "s", "id": 87428, "pid": 76337, "tid": -914061504, "ts": 1716454222735654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735710, "dur": 0,
    "args": {
      "External id": 87438,
      "cbid": 317, "correlation": 87438
    }
  },
  {
    "ph": "f", "id": 87438, "pid": 76337, "tid": -914061504, "ts": 1716454222735710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735711, "dur": 0,
    "args": {
      "External id": 87439,
      "cbid": 203, "correlation": 87439
    }
  },
  {
    "ph": "f", "id": 87439, "pid": 76337, "tid": -914061504, "ts": 1716454222735711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735712, "dur": 0,
    "args": {
      "External id": 87440,
      "cbid": 205, "correlation": 87440
    }
  },
  {
    "ph": "f", "id": 87440, "pid": 76337, "tid": -914061504, "ts": 1716454222735712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222780953, "dur": 11,
    "args": {
      "External id": 87444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87444, "pid": 5, "tid": 7, "ts": 1716454222780953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735727, "dur": 11,
    "args": {
      "External id": 87444,
      "cbid": 211, "correlation": 87444
    }
  },
  {
    "ph": "s", "id": 87444, "pid": 76337, "tid": -914061504, "ts": 1716454222735727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222780965, "dur": 24,
    "args": {
      "External id": 87446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87446, "pid": 5, "tid": 7, "ts": 1716454222780965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735741, "dur": 5,
    "args": {
      "External id": 87446,
      "cbid": 211, "correlation": 87446
    }
  },
  {
    "ph": "s", "id": 87446, "pid": 76337, "tid": -914061504, "ts": 1716454222735741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222780991, "dur": 3,
    "args": {
      "External id": 87448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87448,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 87448, "pid": 5, "tid": 7, "ts": 1716454222780991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735753, "dur": 6,
    "args": {
      "External id": 87448,
      "cbid": 211, "correlation": 87448
    }
  },
  {
    "ph": "s", "id": 87448, "pid": 76337, "tid": -914061504, "ts": 1716454222735753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735763, "dur": 0,
    "args": {
      "External id": 87449,
      "cbid": 51, "correlation": 87449
    }
  },
  {
    "ph": "s", "id": 87449, "pid": 76337, "tid": -914061504, "ts": 1716454222735763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222780995, "dur": 356,
    "args": {
      "External id": 87450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87450,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87450, "pid": 5, "tid": 7, "ts": 1716454222780995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735764, "dur": 7,
    "args": {
      "External id": 87450,
      "cbid": 211, "correlation": 87450
    }
  },
  {
    "ph": "s", "id": 87450, "pid": 76337, "tid": -914061504, "ts": 1716454222735764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222781353, "dur": 21,
    "args": {
      "External id": 87451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87451, "pid": 5, "tid": 7, "ts": 1716454222781353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735773, "dur": 5,
    "args": {
      "External id": 87451,
      "cbid": 211, "correlation": 87451
    }
  },
  {
    "ph": "s", "id": 87451, "pid": 76337, "tid": -914061504, "ts": 1716454222735773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222781375, "dur": 32,
    "args": {
      "External id": 87457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87457, "pid": 5, "tid": 7, "ts": 1716454222781375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735801, "dur": 8,
    "args": {
      "External id": 87457,
      "cbid": 211, "correlation": 87457
    }
  },
  {
    "ph": "s", "id": 87457, "pid": 76337, "tid": -914061504, "ts": 1716454222735801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222781409, "dur": 3,
    "args": {
      "External id": 87465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87465,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 87465, "pid": 5, "tid": 7, "ts": 1716454222781409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735845, "dur": 9,
    "args": {
      "External id": 87465,
      "cbid": 211, "correlation": 87465
    }
  },
  {
    "ph": "s", "id": 87465, "pid": 76337, "tid": -914061504, "ts": 1716454222735845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735914, "dur": 2,
    "args": {
      "External id": 87481,
      "cbid": 251, "correlation": 87481
    }
  },
  {
    "ph": "f", "id": 87481, "pid": 76337, "tid": -914061504, "ts": 1716454222735914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735919, "dur": 0,
    "args": {
      "External id": 87483,
      "cbid": 251, "correlation": 87483
    }
  },
  {
    "ph": "f", "id": 87483, "pid": 76337, "tid": -914061504, "ts": 1716454222735919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222781413, "dur": 12,
    "args": {
      "External id": 87484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87484,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 87484, "pid": 5, "tid": 7, "ts": 1716454222781413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735921, "dur": 12,
    "args": {
      "External id": 87484,
      "cbid": 211, "correlation": 87484
    }
  },
  {
    "ph": "s", "id": 87484, "pid": 76337, "tid": -914061504, "ts": 1716454222735921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222781426, "dur": 5,
    "args": {
      "External id": 87486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87486,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 87486, "pid": 5, "tid": 7, "ts": 1716454222781426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222735935, "dur": 6,
    "args": {
      "External id": 87486,
      "cbid": 211, "correlation": 87486
    }
  },
  {
    "ph": "s", "id": 87486, "pid": 76337, "tid": -914061504, "ts": 1716454222735935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222781433, "dur": 28,
    "args": {
      "External id": 87496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87496, "pid": 5, "tid": 7, "ts": 1716454222781433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736004, "dur": 12,
    "args": {
      "External id": 87496,
      "cbid": 211, "correlation": 87496
    }
  },
  {
    "ph": "s", "id": 87496, "pid": 76337, "tid": -914061504, "ts": 1716454222736004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222781462, "dur": 30,
    "args": {
      "External id": 87516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87516,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 87516, "pid": 5, "tid": 7, "ts": 1716454222781462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736072, "dur": 11,
    "args": {
      "External id": 87516,
      "cbid": 211, "correlation": 87516
    }
  },
  {
    "ph": "s", "id": 87516, "pid": 76337, "tid": -914061504, "ts": 1716454222736072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222781493, "dur": 4,
    "args": {
      "External id": 87528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87528,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 87528, "pid": 5, "tid": 7, "ts": 1716454222781493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736093, "dur": 6,
    "args": {
      "External id": 87528,
      "cbid": 211, "correlation": 87528
    }
  },
  {
    "ph": "s", "id": 87528, "pid": 76337, "tid": -914061504, "ts": 1716454222736093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222781498, "dur": 29,
    "args": {
      "External id": 87531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87531, "pid": 5, "tid": 7, "ts": 1716454222781498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736112, "dur": 6,
    "args": {
      "External id": 87531,
      "cbid": 211, "correlation": 87531
    }
  },
  {
    "ph": "s", "id": 87531, "pid": 76337, "tid": -914061504, "ts": 1716454222736112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222781528, "dur": 20,
    "args": {
      "External id": 87540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87540,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87540, "pid": 5, "tid": 7, "ts": 1716454222781528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736154, "dur": 10,
    "args": {
      "External id": 87540,
      "cbid": 211, "correlation": 87540
    }
  },
  {
    "ph": "s", "id": 87540, "pid": 76337, "tid": -914061504, "ts": 1716454222736154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736219, "dur": 0,
    "args": {
      "External id": 87550,
      "cbid": 317, "correlation": 87550
    }
  },
  {
    "ph": "f", "id": 87550, "pid": 76337, "tid": -914061504, "ts": 1716454222736219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736220, "dur": 0,
    "args": {
      "External id": 87551,
      "cbid": 203, "correlation": 87551
    }
  },
  {
    "ph": "f", "id": 87551, "pid": 76337, "tid": -914061504, "ts": 1716454222736220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736220, "dur": 0,
    "args": {
      "External id": 87552,
      "cbid": 205, "correlation": 87552
    }
  },
  {
    "ph": "f", "id": 87552, "pid": 76337, "tid": -914061504, "ts": 1716454222736220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222781550, "dur": 22,
    "args": {
      "External id": 87556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87556,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87556, "pid": 5, "tid": 7, "ts": 1716454222781550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736237, "dur": 12,
    "args": {
      "External id": 87556,
      "cbid": 211, "correlation": 87556
    }
  },
  {
    "ph": "s", "id": 87556, "pid": 76337, "tid": -914061504, "ts": 1716454222736237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222781573, "dur": 43,
    "args": {
      "External id": 87558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87558, "pid": 5, "tid": 7, "ts": 1716454222781573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736252, "dur": 5,
    "args": {
      "External id": 87558,
      "cbid": 211, "correlation": 87558
    }
  },
  {
    "ph": "s", "id": 87558, "pid": 76337, "tid": -914061504, "ts": 1716454222736252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222781618, "dur": 647,
    "args": {
      "External id": 87560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87560,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87560, "pid": 5, "tid": 7, "ts": 1716454222781618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736267, "dur": 10,
    "args": {
      "External id": 87560,
      "cbid": 211, "correlation": 87560
    }
  },
  {
    "ph": "s", "id": 87560, "pid": 76337, "tid": -914061504, "ts": 1716454222736267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222782266, "dur": 20,
    "args": {
      "External id": 87562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87562,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87562, "pid": 5, "tid": 7, "ts": 1716454222782266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736281, "dur": 5,
    "args": {
      "External id": 87562,
      "cbid": 211, "correlation": 87562
    }
  },
  {
    "ph": "s", "id": 87562, "pid": 76337, "tid": -914061504, "ts": 1716454222736281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222782287, "dur": 33,
    "args": {
      "External id": 87568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87568, "pid": 5, "tid": 7, "ts": 1716454222782287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736309, "dur": 8,
    "args": {
      "External id": 87568,
      "cbid": 211, "correlation": 87568
    }
  },
  {
    "ph": "s", "id": 87568, "pid": 76337, "tid": -914061504, "ts": 1716454222736309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736368, "dur": 0,
    "args": {
      "External id": 87578,
      "cbid": 317, "correlation": 87578
    }
  },
  {
    "ph": "f", "id": 87578, "pid": 76337, "tid": -914061504, "ts": 1716454222736368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736368, "dur": 0,
    "args": {
      "External id": 87579,
      "cbid": 203, "correlation": 87579
    }
  },
  {
    "ph": "f", "id": 87579, "pid": 76337, "tid": -914061504, "ts": 1716454222736368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736369, "dur": 0,
    "args": {
      "External id": 87580,
      "cbid": 205, "correlation": 87580
    }
  },
  {
    "ph": "f", "id": 87580, "pid": 76337, "tid": -914061504, "ts": 1716454222736369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736390, "dur": 1,
    "args": {
      "External id": 87584,
      "cbid": 251, "correlation": 87584
    }
  },
  {
    "ph": "f", "id": 87584, "pid": 76337, "tid": -914061504, "ts": 1716454222736390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736392, "dur": 0,
    "args": {
      "External id": 87585,
      "cbid": 251, "correlation": 87585
    }
  },
  {
    "ph": "f", "id": 87585, "pid": 76337, "tid": -914061504, "ts": 1716454222736392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736393, "dur": 0,
    "args": {
      "External id": 87586,
      "cbid": 251, "correlation": 87586
    }
  },
  {
    "ph": "f", "id": 87586, "pid": 76337, "tid": -914061504, "ts": 1716454222736393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736394, "dur": 0,
    "args": {
      "External id": 87587,
      "cbid": 251, "correlation": 87587
    }
  },
  {
    "ph": "f", "id": 87587, "pid": 76337, "tid": -914061504, "ts": 1716454222736394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736395, "dur": 0,
    "args": {
      "External id": 87588,
      "cbid": 251, "correlation": 87588
    }
  },
  {
    "ph": "f", "id": 87588, "pid": 76337, "tid": -914061504, "ts": 1716454222736395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736395, "dur": 0,
    "args": {
      "External id": 87589,
      "cbid": 251, "correlation": 87589
    }
  },
  {
    "ph": "f", "id": 87589, "pid": 76337, "tid": -914061504, "ts": 1716454222736395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736396, "dur": 0,
    "args": {
      "External id": 87590,
      "cbid": 251, "correlation": 87590
    }
  },
  {
    "ph": "f", "id": 87590, "pid": 76337, "tid": -914061504, "ts": 1716454222736396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736397, "dur": 0,
    "args": {
      "External id": 87591,
      "cbid": 251, "correlation": 87591
    }
  },
  {
    "ph": "f", "id": 87591, "pid": 76337, "tid": -914061504, "ts": 1716454222736397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736398, "dur": 0,
    "args": {
      "External id": 87592,
      "cbid": 251, "correlation": 87592
    }
  },
  {
    "ph": "f", "id": 87592, "pid": 76337, "tid": -914061504, "ts": 1716454222736398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222782321, "dur": 51,
    "args": {
      "External id": 87593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87593,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87593, "pid": 5, "tid": 7, "ts": 1716454222782321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736400, "dur": 13,
    "args": {
      "External id": 87593,
      "cbid": 211, "correlation": 87593
    }
  },
  {
    "ph": "s", "id": 87593, "pid": 76337, "tid": -914061504, "ts": 1716454222736400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222782373, "dur": 32,
    "args": {
      "External id": 87599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87599, "pid": 5, "tid": 7, "ts": 1716454222782373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736433, "dur": 9,
    "args": {
      "External id": 87599,
      "cbid": 211, "correlation": 87599
    }
  },
  {
    "ph": "s", "id": 87599, "pid": 76337, "tid": -914061504, "ts": 1716454222736433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222782406, "dur": 27,
    "args": {
      "External id": 87607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87607,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87607, "pid": 5, "tid": 7, "ts": 1716454222782406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736462, "dur": 8,
    "args": {
      "External id": 87607,
      "cbid": 211, "correlation": 87607
    }
  },
  {
    "ph": "s", "id": 87607, "pid": 76337, "tid": -914061504, "ts": 1716454222736462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222782434, "dur": 20,
    "args": {
      "External id": 87615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87615,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87615, "pid": 5, "tid": 7, "ts": 1716454222782434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736493, "dur": 9,
    "args": {
      "External id": 87615,
      "cbid": 211, "correlation": 87615
    }
  },
  {
    "ph": "s", "id": 87615, "pid": 76337, "tid": -914061504, "ts": 1716454222736493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222782456, "dur": 30,
    "args": {
      "External id": 87635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87635,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 87635, "pid": 5, "tid": 7, "ts": 1716454222782456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736578, "dur": 12,
    "args": {
      "External id": 87635,
      "cbid": 211, "correlation": 87635
    }
  },
  {
    "ph": "s", "id": 87635, "pid": 76337, "tid": -914061504, "ts": 1716454222736578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222782487, "dur": 4,
    "args": {
      "External id": 87647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87647,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 87647, "pid": 5, "tid": 7, "ts": 1716454222782487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736600, "dur": 6,
    "args": {
      "External id": 87647,
      "cbid": 211, "correlation": 87647
    }
  },
  {
    "ph": "s", "id": 87647, "pid": 76337, "tid": -914061504, "ts": 1716454222736600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222782492, "dur": 30,
    "args": {
      "External id": 87650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87650, "pid": 5, "tid": 7, "ts": 1716454222782492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736618, "dur": 6,
    "args": {
      "External id": 87650,
      "cbid": 211, "correlation": 87650
    }
  },
  {
    "ph": "s", "id": 87650, "pid": 76337, "tid": -914061504, "ts": 1716454222736618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736675, "dur": 0,
    "args": {
      "External id": 87661,
      "cbid": 317, "correlation": 87661
    }
  },
  {
    "ph": "f", "id": 87661, "pid": 76337, "tid": -914061504, "ts": 1716454222736675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736676, "dur": 0,
    "args": {
      "External id": 87662,
      "cbid": 203, "correlation": 87662
    }
  },
  {
    "ph": "f", "id": 87662, "pid": 76337, "tid": -914061504, "ts": 1716454222736676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736677, "dur": 0,
    "args": {
      "External id": 87663,
      "cbid": 205, "correlation": 87663
    }
  },
  {
    "ph": "f", "id": 87663, "pid": 76337, "tid": -914061504, "ts": 1716454222736677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222782523, "dur": 24,
    "args": {
      "External id": 87667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87667,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87667, "pid": 5, "tid": 7, "ts": 1716454222782523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736692, "dur": 12,
    "args": {
      "External id": 87667,
      "cbid": 211, "correlation": 87667
    }
  },
  {
    "ph": "s", "id": 87667, "pid": 76337, "tid": -914061504, "ts": 1716454222736692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222782549, "dur": 118,
    "args": {
      "External id": 87669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87669,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87669, "pid": 5, "tid": 7, "ts": 1716454222782549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736712, "dur": 8,
    "args": {
      "External id": 87669,
      "cbid": 211, "correlation": 87669
    }
  },
  {
    "ph": "s", "id": 87669, "pid": 76337, "tid": -914061504, "ts": 1716454222736712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222782668, "dur": 22,
    "args": {
      "External id": 87671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87671, "pid": 5, "tid": 7, "ts": 1716454222782668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736724, "dur": 5,
    "args": {
      "External id": 87671,
      "cbid": 211, "correlation": 87671
    }
  },
  {
    "ph": "s", "id": 87671, "pid": 76337, "tid": -914061504, "ts": 1716454222736724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222782691, "dur": 32,
    "args": {
      "External id": 87677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87677, "pid": 5, "tid": 7, "ts": 1716454222782691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736752, "dur": 8,
    "args": {
      "External id": 87677,
      "cbid": 211, "correlation": 87677
    }
  },
  {
    "ph": "s", "id": 87677, "pid": 76337, "tid": -914061504, "ts": 1716454222736752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222782724, "dur": 160,
    "args": {
      "External id": 87686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87686,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87686, "pid": 5, "tid": 7, "ts": 1716454222782724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736837, "dur": 16,
    "args": {
      "External id": 87686,
      "cbid": 211, "correlation": 87686
    }
  },
  {
    "ph": "s", "id": 87686, "pid": 76337, "tid": -914061504, "ts": 1716454222736837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222782886, "dur": 64,
    "args": {
      "External id": 87708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87708,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87708, "pid": 5, "tid": 7, "ts": 1716454222782886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222736897, "dur": 10,
    "args": {
      "External id": 87708,
      "cbid": 211, "correlation": 87708
    }
  },
  {
    "ph": "s", "id": 87708, "pid": 76337, "tid": -914061504, "ts": 1716454222736897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737002, "dur": 1,
    "args": {
      "External id": 87719,
      "cbid": 251, "correlation": 87719
    }
  },
  {
    "ph": "f", "id": 87719, "pid": 76337, "tid": -914061504, "ts": 1716454222737002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222782952, "dur": 154,
    "args": {
      "External id": 87720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87720,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87720, "pid": 5, "tid": 7, "ts": 1716454222782952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737007, "dur": 14,
    "args": {
      "External id": 87720,
      "cbid": 211, "correlation": 87720
    }
  },
  {
    "ph": "s", "id": 87720, "pid": 76337, "tid": -914061504, "ts": 1716454222737007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737079, "dur": 1,
    "args": {
      "External id": 87731,
      "cbid": 251, "correlation": 87731
    }
  },
  {
    "ph": "f", "id": 87731, "pid": 76337, "tid": -914061504, "ts": 1716454222737079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222783107, "dur": 143,
    "args": {
      "External id": 87732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87732,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87732, "pid": 5, "tid": 7, "ts": 1716454222783107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737083, "dur": 12,
    "args": {
      "External id": 87732,
      "cbid": 211, "correlation": 87732
    }
  },
  {
    "ph": "s", "id": 87732, "pid": 76337, "tid": -914061504, "ts": 1716454222737083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737148, "dur": 1,
    "args": {
      "External id": 87743,
      "cbid": 251, "correlation": 87743
    }
  },
  {
    "ph": "f", "id": 87743, "pid": 76337, "tid": -914061504, "ts": 1716454222737148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222783252, "dur": 142,
    "args": {
      "External id": 87744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87744,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87744, "pid": 5, "tid": 7, "ts": 1716454222783252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737153, "dur": 11,
    "args": {
      "External id": 87744,
      "cbid": 211, "correlation": 87744
    }
  },
  {
    "ph": "s", "id": 87744, "pid": 76337, "tid": -914061504, "ts": 1716454222737153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222783395, "dur": 1907,
    "args": {
      "External id": 87765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87765,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87765, "pid": 5, "tid": 7, "ts": 1716454222783395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737240, "dur": 14,
    "args": {
      "External id": 87765,
      "cbid": 211, "correlation": 87765
    }
  },
  {
    "ph": "s", "id": 87765, "pid": 76337, "tid": -914061504, "ts": 1716454222737240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737343, "dur": 1,
    "args": {
      "External id": 87783,
      "cbid": 251, "correlation": 87783
    }
  },
  {
    "ph": "f", "id": 87783, "pid": 76337, "tid": -914061504, "ts": 1716454222737343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222785304, "dur": 143,
    "args": {
      "External id": 87785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87785,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 87785, "pid": 5, "tid": 7, "ts": 1716454222785304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737348, "dur": 14,
    "args": {
      "External id": 87785,
      "cbid": 211, "correlation": 87785
    }
  },
  {
    "ph": "s", "id": 87785, "pid": 76337, "tid": -914061504, "ts": 1716454222737348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222785448, "dur": 35,
    "args": {
      "External id": 87793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87793,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87793, "pid": 5, "tid": 7, "ts": 1716454222785448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737421, "dur": 12,
    "args": {
      "External id": 87793,
      "cbid": 211, "correlation": 87793
    }
  },
  {
    "ph": "s", "id": 87793, "pid": 76337, "tid": -914061504, "ts": 1716454222737421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222785485, "dur": 50,
    "args": {
      "External id": 87801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87801,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87801, "pid": 5, "tid": 7, "ts": 1716454222785485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737460, "dur": 8,
    "args": {
      "External id": 87801,
      "cbid": 211, "correlation": 87801
    }
  },
  {
    "ph": "s", "id": 87801, "pid": 76337, "tid": -914061504, "ts": 1716454222737460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222785536, "dur": 29,
    "args": {
      "External id": 87812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87812, "pid": 5, "tid": 7, "ts": 1716454222785536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737537, "dur": 13,
    "args": {
      "External id": 87812,
      "cbid": 211, "correlation": 87812
    }
  },
  {
    "ph": "s", "id": 87812, "pid": 76337, "tid": -914061504, "ts": 1716454222737537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222785567, "dur": 33,
    "args": {
      "External id": 87834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87834,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87834, "pid": 5, "tid": 7, "ts": 1716454222785567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737569, "dur": 8,
    "args": {
      "External id": 87834,
      "cbid": 211, "correlation": 87834
    }
  },
  {
    "ph": "s", "id": 87834, "pid": 76337, "tid": -914061504, "ts": 1716454222737569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737655, "dur": 1,
    "args": {
      "External id": 87845,
      "cbid": 251, "correlation": 87845
    }
  },
  {
    "ph": "f", "id": 87845, "pid": 76337, "tid": -914061504, "ts": 1716454222737655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222785601, "dur": 75,
    "args": {
      "External id": 87846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87846,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87846, "pid": 5, "tid": 7, "ts": 1716454222785601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737660, "dur": 13,
    "args": {
      "External id": 87846,
      "cbid": 211, "correlation": 87846
    }
  },
  {
    "ph": "s", "id": 87846, "pid": 76337, "tid": -914061504, "ts": 1716454222737660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737732, "dur": 1,
    "args": {
      "External id": 87857,
      "cbid": 251, "correlation": 87857
    }
  },
  {
    "ph": "f", "id": 87857, "pid": 76337, "tid": -914061504, "ts": 1716454222737732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737736, "dur": 0,
    "args": {
      "External id": 87858,
      "cbid": 251, "correlation": 87858
    }
  },
  {
    "ph": "f", "id": 87858, "pid": 76337, "tid": -914061504, "ts": 1716454222737736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222785678, "dur": 11,
    "args": {
      "External id": 87859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87859,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 87859, "pid": 5, "tid": 7, "ts": 1716454222785678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737737, "dur": 13,
    "args": {
      "External id": 87859,
      "cbid": 211, "correlation": 87859
    }
  },
  {
    "ph": "s", "id": 87859, "pid": 76337, "tid": -914061504, "ts": 1716454222737737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222785691, "dur": 5,
    "args": {
      "External id": 87861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87861,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 87861, "pid": 5, "tid": 7, "ts": 1716454222785691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737753, "dur": 7,
    "args": {
      "External id": 87861,
      "cbid": 211, "correlation": 87861
    }
  },
  {
    "ph": "s", "id": 87861, "pid": 76337, "tid": -914061504, "ts": 1716454222737753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737813, "dur": 1,
    "args": {
      "External id": 87872,
      "cbid": 251, "correlation": 87872
    }
  },
  {
    "ph": "f", "id": 87872, "pid": 76337, "tid": -914061504, "ts": 1716454222737813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737816, "dur": 0,
    "args": {
      "External id": 87873,
      "cbid": 251, "correlation": 87873
    }
  },
  {
    "ph": "f", "id": 87873, "pid": 76337, "tid": -914061504, "ts": 1716454222737816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222785697, "dur": 7,
    "args": {
      "External id": 87874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87874,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 87874, "pid": 5, "tid": 7, "ts": 1716454222785697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737818, "dur": 11,
    "args": {
      "External id": 87874,
      "cbid": 211, "correlation": 87874
    }
  },
  {
    "ph": "s", "id": 87874, "pid": 76337, "tid": -914061504, "ts": 1716454222737818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222785705, "dur": 3,
    "args": {
      "External id": 87876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87876,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 87876, "pid": 5, "tid": 7, "ts": 1716454222785705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737831, "dur": 5,
    "args": {
      "External id": 87876,
      "cbid": 211, "correlation": 87876
    }
  },
  {
    "ph": "s", "id": 87876, "pid": 76337, "tid": -914061504, "ts": 1716454222737831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222785710, "dur": 90,
    "args": {
      "External id": 87897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87897,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 87897, "pid": 5, "tid": 7, "ts": 1716454222785710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222737905, "dur": 13,
    "args": {
      "External id": 87897,
      "cbid": 211, "correlation": 87897
    }
  },
  {
    "ph": "s", "id": 87897, "pid": 76337, "tid": -914061504, "ts": 1716454222737905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738012, "dur": 1,
    "args": {
      "External id": 87915,
      "cbid": 251, "correlation": 87915
    }
  },
  {
    "ph": "f", "id": 87915, "pid": 76337, "tid": -914061504, "ts": 1716454222738012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222785801, "dur": 96,
    "args": {
      "External id": 87917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87917,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 87917, "pid": 5, "tid": 7, "ts": 1716454222785801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738018, "dur": 14,
    "args": {
      "External id": 87917,
      "cbid": 211, "correlation": 87917
    }
  },
  {
    "ph": "s", "id": 87917, "pid": 76337, "tid": -914061504, "ts": 1716454222738018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222785899, "dur": 19,
    "args": {
      "External id": 87925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87925,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87925, "pid": 5, "tid": 7, "ts": 1716454222785899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738089, "dur": 13,
    "args": {
      "External id": 87925,
      "cbid": 211, "correlation": 87925
    }
  },
  {
    "ph": "s", "id": 87925, "pid": 76337, "tid": -914061504, "ts": 1716454222738089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222785919, "dur": 37,
    "args": {
      "External id": 87933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87933,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87933, "pid": 5, "tid": 7, "ts": 1716454222785919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738132, "dur": 9,
    "args": {
      "External id": 87933,
      "cbid": 211, "correlation": 87933
    }
  },
  {
    "ph": "s", "id": 87933, "pid": 76337, "tid": -914061504, "ts": 1716454222738132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222785958, "dur": 35,
    "args": {
      "External id": 87955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87955,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87955, "pid": 5, "tid": 7, "ts": 1716454222785958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738185, "dur": 11,
    "args": {
      "External id": 87955,
      "cbid": 211, "correlation": 87955
    }
  },
  {
    "ph": "s", "id": 87955, "pid": 76337, "tid": -914061504, "ts": 1716454222738185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738277, "dur": 1,
    "args": {
      "External id": 87971,
      "cbid": 251, "correlation": 87971
    }
  },
  {
    "ph": "f", "id": 87971, "pid": 76337, "tid": -914061504, "ts": 1716454222738277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738282, "dur": 0,
    "args": {
      "External id": 87973,
      "cbid": 251, "correlation": 87973
    }
  },
  {
    "ph": "f", "id": 87973, "pid": 76337, "tid": -914061504, "ts": 1716454222738282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222785994, "dur": 532,
    "args": {
      "External id": 87974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87974,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 87974, "pid": 5, "tid": 7, "ts": 1716454222785994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738287, "dur": 13,
    "args": {
      "External id": 87974,
      "cbid": 211, "correlation": 87974
    }
  },
  {
    "ph": "s", "id": 87974, "pid": 76337, "tid": -914061504, "ts": 1716454222738287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222786528, "dur": 122,
    "args": {
      "External id": 87982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87982,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87982, "pid": 5, "tid": 7, "ts": 1716454222786528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738360, "dur": 15,
    "args": {
      "External id": 87982,
      "cbid": 211, "correlation": 87982
    }
  },
  {
    "ph": "s", "id": 87982, "pid": 76337, "tid": -914061504, "ts": 1716454222738360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222786651, "dur": 128,
    "args": {
      "External id": 87990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 87990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 87990, "pid": 5, "tid": 7, "ts": 1716454222786651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738395, "dur": 8,
    "args": {
      "External id": 87990,
      "cbid": 211, "correlation": 87990
    }
  },
  {
    "ph": "s", "id": 87990, "pid": 76337, "tid": -914061504, "ts": 1716454222738395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738474, "dur": 1,
    "args": {
      "External id": 88006,
      "cbid": 251, "correlation": 88006
    }
  },
  {
    "ph": "f", "id": 88006, "pid": 76337, "tid": -914061504, "ts": 1716454222738474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222786780, "dur": 311,
    "args": {
      "External id": 88008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88008,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88008, "pid": 5, "tid": 7, "ts": 1716454222786780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738480, "dur": 13,
    "args": {
      "External id": 88008,
      "cbid": 211, "correlation": 88008
    }
  },
  {
    "ph": "s", "id": 88008, "pid": 76337, "tid": -914061504, "ts": 1716454222738480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222787092, "dur": 27,
    "args": {
      "External id": 88016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88016,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88016, "pid": 5, "tid": 7, "ts": 1716454222787092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738523, "dur": 9,
    "args": {
      "External id": 88016,
      "cbid": 211, "correlation": 88016
    }
  },
  {
    "ph": "s", "id": 88016, "pid": 76337, "tid": -914061504, "ts": 1716454222738523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222787120, "dur": 80,
    "args": {
      "External id": 88027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88027, "pid": 5, "tid": 7, "ts": 1716454222787120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738595, "dur": 12,
    "args": {
      "External id": 88027,
      "cbid": 211, "correlation": 88027
    }
  },
  {
    "ph": "s", "id": 88027, "pid": 76337, "tid": -914061504, "ts": 1716454222738595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738662, "dur": 0,
    "args": {
      "External id": 88039,
      "cbid": 317, "correlation": 88039
    }
  },
  {
    "ph": "f", "id": 88039, "pid": 76337, "tid": -914061504, "ts": 1716454222738662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738663, "dur": 0,
    "args": {
      "External id": 88040,
      "cbid": 203, "correlation": 88040
    }
  },
  {
    "ph": "f", "id": 88040, "pid": 76337, "tid": -914061504, "ts": 1716454222738663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738664, "dur": 0,
    "args": {
      "External id": 88041,
      "cbid": 205, "correlation": 88041
    }
  },
  {
    "ph": "f", "id": 88041, "pid": 76337, "tid": -914061504, "ts": 1716454222738664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222787202, "dur": 22,
    "args": {
      "External id": 88045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88045,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88045, "pid": 5, "tid": 7, "ts": 1716454222787202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738683, "dur": 12,
    "args": {
      "External id": 88045,
      "cbid": 211, "correlation": 88045
    }
  },
  {
    "ph": "s", "id": 88045, "pid": 76337, "tid": -914061504, "ts": 1716454222738683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222787225, "dur": 118,
    "args": {
      "External id": 88047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88047,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88047, "pid": 5, "tid": 7, "ts": 1716454222787225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738703, "dur": 7,
    "args": {
      "External id": 88047,
      "cbid": 211, "correlation": 88047
    }
  },
  {
    "ph": "s", "id": 88047, "pid": 76337, "tid": -914061504, "ts": 1716454222738703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222787345, "dur": 23,
    "args": {
      "External id": 88049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88049,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88049, "pid": 5, "tid": 7, "ts": 1716454222787345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738714, "dur": 5,
    "args": {
      "External id": 88049,
      "cbid": 211, "correlation": 88049
    }
  },
  {
    "ph": "s", "id": 88049, "pid": 76337, "tid": -914061504, "ts": 1716454222738714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222787369, "dur": 32,
    "args": {
      "External id": 88055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88055, "pid": 5, "tid": 7, "ts": 1716454222787369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738742, "dur": 8,
    "args": {
      "External id": 88055,
      "cbid": 211, "correlation": 88055
    }
  },
  {
    "ph": "s", "id": 88055, "pid": 76337, "tid": -914061504, "ts": 1716454222738742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222787402, "dur": 27,
    "args": {
      "External id": 88063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88063,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88063, "pid": 5, "tid": 7, "ts": 1716454222787402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738774, "dur": 8,
    "args": {
      "External id": 88063,
      "cbid": 211, "correlation": 88063
    }
  },
  {
    "ph": "s", "id": 88063, "pid": 76337, "tid": -914061504, "ts": 1716454222738774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222787430, "dur": 29,
    "args": {
      "External id": 88083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88083,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 88083, "pid": 5, "tid": 7, "ts": 1716454222787430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738851, "dur": 12,
    "args": {
      "External id": 88083,
      "cbid": 211, "correlation": 88083
    }
  },
  {
    "ph": "s", "id": 88083, "pid": 76337, "tid": -914061504, "ts": 1716454222738851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222787461, "dur": 5,
    "args": {
      "External id": 88095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88095,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 88095, "pid": 5, "tid": 7, "ts": 1716454222787461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738874, "dur": 6,
    "args": {
      "External id": 88095,
      "cbid": 211, "correlation": 88095
    }
  },
  {
    "ph": "s", "id": 88095, "pid": 76337, "tid": -914061504, "ts": 1716454222738874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222787467, "dur": 30,
    "args": {
      "External id": 88098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88098, "pid": 5, "tid": 7, "ts": 1716454222787467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738892, "dur": 7,
    "args": {
      "External id": 88098,
      "cbid": 211, "correlation": 88098
    }
  },
  {
    "ph": "s", "id": 88098, "pid": 76337, "tid": -914061504, "ts": 1716454222738892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222787498, "dur": 22,
    "args": {
      "External id": 88107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88107,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88107, "pid": 5, "tid": 7, "ts": 1716454222787498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738933, "dur": 10,
    "args": {
      "External id": 88107,
      "cbid": 211, "correlation": 88107
    }
  },
  {
    "ph": "s", "id": 88107, "pid": 76337, "tid": -914061504, "ts": 1716454222738933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738994, "dur": 0,
    "args": {
      "External id": 88117,
      "cbid": 317, "correlation": 88117
    }
  },
  {
    "ph": "f", "id": 88117, "pid": 76337, "tid": -914061504, "ts": 1716454222738994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738994, "dur": 0,
    "args": {
      "External id": 88118,
      "cbid": 203, "correlation": 88118
    }
  },
  {
    "ph": "f", "id": 88118, "pid": 76337, "tid": -914061504, "ts": 1716454222738994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222738995, "dur": 0,
    "args": {
      "External id": 88119,
      "cbid": 205, "correlation": 88119
    }
  },
  {
    "ph": "f", "id": 88119, "pid": 76337, "tid": -914061504, "ts": 1716454222738995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222787521, "dur": 22,
    "args": {
      "External id": 88123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88123, "pid": 5, "tid": 7, "ts": 1716454222787521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739009, "dur": 12,
    "args": {
      "External id": 88123,
      "cbid": 211, "correlation": 88123
    }
  },
  {
    "ph": "s", "id": 88123, "pid": 76337, "tid": -914061504, "ts": 1716454222739009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222787544, "dur": 44,
    "args": {
      "External id": 88125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88125,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88125, "pid": 5, "tid": 7, "ts": 1716454222787544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739024, "dur": 5,
    "args": {
      "External id": 88125,
      "cbid": 211, "correlation": 88125
    }
  },
  {
    "ph": "s", "id": 88125, "pid": 76337, "tid": -914061504, "ts": 1716454222739024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222787590, "dur": 646,
    "args": {
      "External id": 88127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88127,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88127, "pid": 5, "tid": 7, "ts": 1716454222787590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739036, "dur": 7,
    "args": {
      "External id": 88127,
      "cbid": 211, "correlation": 88127
    }
  },
  {
    "ph": "s", "id": 88127, "pid": 76337, "tid": -914061504, "ts": 1716454222739036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222788237, "dur": 23,
    "args": {
      "External id": 88129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88129, "pid": 5, "tid": 7, "ts": 1716454222788237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739046, "dur": 5,
    "args": {
      "External id": 88129,
      "cbid": 211, "correlation": 88129
    }
  },
  {
    "ph": "s", "id": 88129, "pid": 76337, "tid": -914061504, "ts": 1716454222739046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222788261, "dur": 32,
    "args": {
      "External id": 88135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88135, "pid": 5, "tid": 7, "ts": 1716454222788261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739073, "dur": 8,
    "args": {
      "External id": 88135,
      "cbid": 211, "correlation": 88135
    }
  },
  {
    "ph": "s", "id": 88135, "pid": 76337, "tid": -914061504, "ts": 1716454222739073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222788294, "dur": 3,
    "args": {
      "External id": 88143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88143,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 88143, "pid": 5, "tid": 7, "ts": 1716454222788294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739117, "dur": 9,
    "args": {
      "External id": 88143,
      "cbid": 211, "correlation": 88143
    }
  },
  {
    "ph": "s", "id": 88143, "pid": 76337, "tid": -914061504, "ts": 1716454222739117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739185, "dur": 1,
    "args": {
      "External id": 88159,
      "cbid": 251, "correlation": 88159
    }
  },
  {
    "ph": "f", "id": 88159, "pid": 76337, "tid": -914061504, "ts": 1716454222739185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739190, "dur": 0,
    "args": {
      "External id": 88161,
      "cbid": 251, "correlation": 88161
    }
  },
  {
    "ph": "f", "id": 88161, "pid": 76337, "tid": -914061504, "ts": 1716454222739190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222788299, "dur": 12,
    "args": {
      "External id": 88162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88162,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 88162, "pid": 5, "tid": 7, "ts": 1716454222788299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739192, "dur": 13,
    "args": {
      "External id": 88162,
      "cbid": 211, "correlation": 88162
    }
  },
  {
    "ph": "s", "id": 88162, "pid": 76337, "tid": -914061504, "ts": 1716454222739192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222788312, "dur": 5,
    "args": {
      "External id": 88164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88164,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 88164, "pid": 5, "tid": 7, "ts": 1716454222788312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739207, "dur": 5,
    "args": {
      "External id": 88164,
      "cbid": 211, "correlation": 88164
    }
  },
  {
    "ph": "s", "id": 88164, "pid": 76337, "tid": -914061504, "ts": 1716454222739207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222788319, "dur": 29,
    "args": {
      "External id": 88174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88174, "pid": 5, "tid": 7, "ts": 1716454222788319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739266, "dur": 12,
    "args": {
      "External id": 88174,
      "cbid": 211, "correlation": 88174
    }
  },
  {
    "ph": "s", "id": 88174, "pid": 76337, "tid": -914061504, "ts": 1716454222739266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222788349, "dur": 29,
    "args": {
      "External id": 88194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88194,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 88194, "pid": 5, "tid": 7, "ts": 1716454222788349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739333, "dur": 10,
    "args": {
      "External id": 88194,
      "cbid": 211, "correlation": 88194
    }
  },
  {
    "ph": "s", "id": 88194, "pid": 76337, "tid": -914061504, "ts": 1716454222739333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222788380, "dur": 4,
    "args": {
      "External id": 88206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88206,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 88206, "pid": 5, "tid": 7, "ts": 1716454222788380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739353, "dur": 6,
    "args": {
      "External id": 88206,
      "cbid": 211, "correlation": 88206
    }
  },
  {
    "ph": "s", "id": 88206, "pid": 76337, "tid": -914061504, "ts": 1716454222739353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222788385, "dur": 30,
    "args": {
      "External id": 88209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88209, "pid": 5, "tid": 7, "ts": 1716454222788385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739372, "dur": 7,
    "args": {
      "External id": 88209,
      "cbid": 211, "correlation": 88209
    }
  },
  {
    "ph": "s", "id": 88209, "pid": 76337, "tid": -914061504, "ts": 1716454222739372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222788416, "dur": 20,
    "args": {
      "External id": 88218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88218,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88218, "pid": 5, "tid": 7, "ts": 1716454222788416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739412, "dur": 10,
    "args": {
      "External id": 88218,
      "cbid": 211, "correlation": 88218
    }
  },
  {
    "ph": "s", "id": 88218, "pid": 76337, "tid": -914061504, "ts": 1716454222739412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739476, "dur": 0,
    "args": {
      "External id": 88228,
      "cbid": 317, "correlation": 88228
    }
  },
  {
    "ph": "f", "id": 88228, "pid": 76337, "tid": -914061504, "ts": 1716454222739476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739477, "dur": 0,
    "args": {
      "External id": 88229,
      "cbid": 203, "correlation": 88229
    }
  },
  {
    "ph": "f", "id": 88229, "pid": 76337, "tid": -914061504, "ts": 1716454222739477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739478, "dur": 0,
    "args": {
      "External id": 88230,
      "cbid": 205, "correlation": 88230
    }
  },
  {
    "ph": "f", "id": 88230, "pid": 76337, "tid": -914061504, "ts": 1716454222739478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222788437, "dur": 22,
    "args": {
      "External id": 88234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88234, "pid": 5, "tid": 7, "ts": 1716454222788437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739492, "dur": 12,
    "args": {
      "External id": 88234,
      "cbid": 211, "correlation": 88234
    }
  },
  {
    "ph": "s", "id": 88234, "pid": 76337, "tid": -914061504, "ts": 1716454222739492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222788460, "dur": 44,
    "args": {
      "External id": 88236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88236, "pid": 5, "tid": 7, "ts": 1716454222788460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739506, "dur": 5,
    "args": {
      "External id": 88236,
      "cbid": 211, "correlation": 88236
    }
  },
  {
    "ph": "s", "id": 88236, "pid": 76337, "tid": -914061504, "ts": 1716454222739506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222788505, "dur": 638,
    "args": {
      "External id": 88238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88238,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88238, "pid": 5, "tid": 7, "ts": 1716454222788505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739518, "dur": 6,
    "args": {
      "External id": 88238,
      "cbid": 211, "correlation": 88238
    }
  },
  {
    "ph": "s", "id": 88238, "pid": 76337, "tid": -914061504, "ts": 1716454222739518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222789144, "dur": 24,
    "args": {
      "External id": 88240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88240,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88240, "pid": 5, "tid": 7, "ts": 1716454222789144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739527, "dur": 6,
    "args": {
      "External id": 88240,
      "cbid": 211, "correlation": 88240
    }
  },
  {
    "ph": "s", "id": 88240, "pid": 76337, "tid": -914061504, "ts": 1716454222739527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222789169, "dur": 33,
    "args": {
      "External id": 88246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88246, "pid": 5, "tid": 7, "ts": 1716454222789169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739555, "dur": 8,
    "args": {
      "External id": 88246,
      "cbid": 211, "correlation": 88246
    }
  },
  {
    "ph": "s", "id": 88246, "pid": 76337, "tid": -914061504, "ts": 1716454222739555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222789203, "dur": 27,
    "args": {
      "External id": 88254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88254,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88254, "pid": 5, "tid": 7, "ts": 1716454222789203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739587, "dur": 9,
    "args": {
      "External id": 88254,
      "cbid": 211, "correlation": 88254
    }
  },
  {
    "ph": "s", "id": 88254, "pid": 76337, "tid": -914061504, "ts": 1716454222739587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222789231, "dur": 20,
    "args": {
      "External id": 88262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88262,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88262, "pid": 5, "tid": 7, "ts": 1716454222789231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739618, "dur": 9,
    "args": {
      "External id": 88262,
      "cbid": 211, "correlation": 88262
    }
  },
  {
    "ph": "s", "id": 88262, "pid": 76337, "tid": -914061504, "ts": 1716454222739618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222789252, "dur": 30,
    "args": {
      "External id": 88282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88282,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 88282, "pid": 5, "tid": 7, "ts": 1716454222789252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739699, "dur": 13,
    "args": {
      "External id": 88282,
      "cbid": 211, "correlation": 88282
    }
  },
  {
    "ph": "s", "id": 88282, "pid": 76337, "tid": -914061504, "ts": 1716454222739699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222789284, "dur": 5,
    "args": {
      "External id": 88294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88294,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 88294, "pid": 5, "tid": 7, "ts": 1716454222789284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739722, "dur": 6,
    "args": {
      "External id": 88294,
      "cbid": 211, "correlation": 88294
    }
  },
  {
    "ph": "s", "id": 88294, "pid": 76337, "tid": -914061504, "ts": 1716454222739722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222789290, "dur": 30,
    "args": {
      "External id": 88297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88297,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88297, "pid": 5, "tid": 7, "ts": 1716454222789290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739739, "dur": 6,
    "args": {
      "External id": 88297,
      "cbid": 211, "correlation": 88297
    }
  },
  {
    "ph": "s", "id": 88297, "pid": 76337, "tid": -914061504, "ts": 1716454222739739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739796, "dur": 0,
    "args": {
      "External id": 88308,
      "cbid": 317, "correlation": 88308
    }
  },
  {
    "ph": "f", "id": 88308, "pid": 76337, "tid": -914061504, "ts": 1716454222739796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739797, "dur": 0,
    "args": {
      "External id": 88309,
      "cbid": 203, "correlation": 88309
    }
  },
  {
    "ph": "f", "id": 88309, "pid": 76337, "tid": -914061504, "ts": 1716454222739797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739797, "dur": 0,
    "args": {
      "External id": 88310,
      "cbid": 205, "correlation": 88310
    }
  },
  {
    "ph": "f", "id": 88310, "pid": 76337, "tid": -914061504, "ts": 1716454222739797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222789321, "dur": 21,
    "args": {
      "External id": 88314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88314,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88314, "pid": 5, "tid": 7, "ts": 1716454222789321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739811, "dur": 12,
    "args": {
      "External id": 88314,
      "cbid": 211, "correlation": 88314
    }
  },
  {
    "ph": "s", "id": 88314, "pid": 76337, "tid": -914061504, "ts": 1716454222739811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222789343, "dur": 115,
    "args": {
      "External id": 88316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88316,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88316, "pid": 5, "tid": 7, "ts": 1716454222789343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739829, "dur": 7,
    "args": {
      "External id": 88316,
      "cbid": 211, "correlation": 88316
    }
  },
  {
    "ph": "s", "id": 88316, "pid": 76337, "tid": -914061504, "ts": 1716454222739829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222789459, "dur": 21,
    "args": {
      "External id": 88318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88318,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88318, "pid": 5, "tid": 7, "ts": 1716454222789459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739839, "dur": 5,
    "args": {
      "External id": 88318,
      "cbid": 211, "correlation": 88318
    }
  },
  {
    "ph": "s", "id": 88318, "pid": 76337, "tid": -914061504, "ts": 1716454222739839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222789482, "dur": 32,
    "args": {
      "External id": 88324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88324, "pid": 5, "tid": 7, "ts": 1716454222789482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739867, "dur": 8,
    "args": {
      "External id": 88324,
      "cbid": 211, "correlation": 88324
    }
  },
  {
    "ph": "s", "id": 88324, "pid": 76337, "tid": -914061504, "ts": 1716454222739867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222789515, "dur": 188,
    "args": {
      "External id": 88333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88333,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88333, "pid": 5, "tid": 7, "ts": 1716454222789515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222739952, "dur": 15,
    "args": {
      "External id": 88333,
      "cbid": 211, "correlation": 88333
    }
  },
  {
    "ph": "s", "id": 88333, "pid": 76337, "tid": -914061504, "ts": 1716454222739952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222789705, "dur": 64,
    "args": {
      "External id": 88355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88355,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88355, "pid": 5, "tid": 7, "ts": 1716454222789705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740021, "dur": 11,
    "args": {
      "External id": 88355,
      "cbid": 211, "correlation": 88355
    }
  },
  {
    "ph": "s", "id": 88355, "pid": 76337, "tid": -914061504, "ts": 1716454222740021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740115, "dur": 1,
    "args": {
      "External id": 88366,
      "cbid": 251, "correlation": 88366
    }
  },
  {
    "ph": "f", "id": 88366, "pid": 76337, "tid": -914061504, "ts": 1716454222740115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222789770, "dur": 154,
    "args": {
      "External id": 88367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88367,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88367, "pid": 5, "tid": 7, "ts": 1716454222789770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740121, "dur": 13,
    "args": {
      "External id": 88367,
      "cbid": 211, "correlation": 88367
    }
  },
  {
    "ph": "s", "id": 88367, "pid": 76337, "tid": -914061504, "ts": 1716454222740121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740191, "dur": 1,
    "args": {
      "External id": 88378,
      "cbid": 251, "correlation": 88378
    }
  },
  {
    "ph": "f", "id": 88378, "pid": 76337, "tid": -914061504, "ts": 1716454222740191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222789925, "dur": 145,
    "args": {
      "External id": 88379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88379,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88379, "pid": 5, "tid": 7, "ts": 1716454222789925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740195, "dur": 11,
    "args": {
      "External id": 88379,
      "cbid": 211, "correlation": 88379
    }
  },
  {
    "ph": "s", "id": 88379, "pid": 76337, "tid": -914061504, "ts": 1716454222740195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740261, "dur": 1,
    "args": {
      "External id": 88390,
      "cbid": 251, "correlation": 88390
    }
  },
  {
    "ph": "f", "id": 88390, "pid": 76337, "tid": -914061504, "ts": 1716454222740261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222790071, "dur": 144,
    "args": {
      "External id": 88391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88391,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88391, "pid": 5, "tid": 7, "ts": 1716454222790071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740265, "dur": 11,
    "args": {
      "External id": 88391,
      "cbid": 211, "correlation": 88391
    }
  },
  {
    "ph": "s", "id": 88391, "pid": 76337, "tid": -914061504, "ts": 1716454222740265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222790217, "dur": 1909,
    "args": {
      "External id": 88412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88412,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 88412, "pid": 5, "tid": 7, "ts": 1716454222790217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740351, "dur": 14,
    "args": {
      "External id": 88412,
      "cbid": 211, "correlation": 88412
    }
  },
  {
    "ph": "s", "id": 88412, "pid": 76337, "tid": -914061504, "ts": 1716454222740351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740457, "dur": 1,
    "args": {
      "External id": 88430,
      "cbid": 251, "correlation": 88430
    }
  },
  {
    "ph": "f", "id": 88430, "pid": 76337, "tid": -914061504, "ts": 1716454222740457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222792127, "dur": 145,
    "args": {
      "External id": 88432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88432,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 88432, "pid": 5, "tid": 7, "ts": 1716454222792127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740463, "dur": 14,
    "args": {
      "External id": 88432,
      "cbid": 211, "correlation": 88432
    }
  },
  {
    "ph": "s", "id": 88432, "pid": 76337, "tid": -914061504, "ts": 1716454222740463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222792274, "dur": 35,
    "args": {
      "External id": 88440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88440,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88440, "pid": 5, "tid": 7, "ts": 1716454222792274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740536, "dur": 12,
    "args": {
      "External id": 88440,
      "cbid": 211, "correlation": 88440
    }
  },
  {
    "ph": "s", "id": 88440, "pid": 76337, "tid": -914061504, "ts": 1716454222740536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222792310, "dur": 51,
    "args": {
      "External id": 88448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88448,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88448, "pid": 5, "tid": 7, "ts": 1716454222792310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740575, "dur": 9,
    "args": {
      "External id": 88448,
      "cbid": 211, "correlation": 88448
    }
  },
  {
    "ph": "s", "id": 88448, "pid": 76337, "tid": -914061504, "ts": 1716454222740575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222792362, "dur": 30,
    "args": {
      "External id": 88459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88459,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88459, "pid": 5, "tid": 7, "ts": 1716454222792362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740651, "dur": 13,
    "args": {
      "External id": 88459,
      "cbid": 211, "correlation": 88459
    }
  },
  {
    "ph": "s", "id": 88459, "pid": 76337, "tid": -914061504, "ts": 1716454222740651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222792393, "dur": 33,
    "args": {
      "External id": 88481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88481,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88481, "pid": 5, "tid": 7, "ts": 1716454222792393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740682, "dur": 8,
    "args": {
      "External id": 88481,
      "cbid": 211, "correlation": 88481
    }
  },
  {
    "ph": "s", "id": 88481, "pid": 76337, "tid": -914061504, "ts": 1716454222740682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740768, "dur": 1,
    "args": {
      "External id": 88492,
      "cbid": 251, "correlation": 88492
    }
  },
  {
    "ph": "f", "id": 88492, "pid": 76337, "tid": -914061504, "ts": 1716454222740768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222792428, "dur": 74,
    "args": {
      "External id": 88493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88493,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88493, "pid": 5, "tid": 7, "ts": 1716454222792428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740773, "dur": 13,
    "args": {
      "External id": 88493,
      "cbid": 211, "correlation": 88493
    }
  },
  {
    "ph": "s", "id": 88493, "pid": 76337, "tid": -914061504, "ts": 1716454222740773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740843, "dur": 1,
    "args": {
      "External id": 88504,
      "cbid": 251, "correlation": 88504
    }
  },
  {
    "ph": "f", "id": 88504, "pid": 76337, "tid": -914061504, "ts": 1716454222740843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740847, "dur": 0,
    "args": {
      "External id": 88505,
      "cbid": 251, "correlation": 88505
    }
  },
  {
    "ph": "f", "id": 88505, "pid": 76337, "tid": -914061504, "ts": 1716454222740847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222792503, "dur": 11,
    "args": {
      "External id": 88506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88506,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 88506, "pid": 5, "tid": 7, "ts": 1716454222792503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740849, "dur": 13,
    "args": {
      "External id": 88506,
      "cbid": 211, "correlation": 88506
    }
  },
  {
    "ph": "s", "id": 88506, "pid": 76337, "tid": -914061504, "ts": 1716454222740849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222792516, "dur": 5,
    "args": {
      "External id": 88508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88508,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 88508, "pid": 5, "tid": 7, "ts": 1716454222792516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740865, "dur": 6,
    "args": {
      "External id": 88508,
      "cbid": 211, "correlation": 88508
    }
  },
  {
    "ph": "s", "id": 88508, "pid": 76337, "tid": -914061504, "ts": 1716454222740865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740922, "dur": 1,
    "args": {
      "External id": 88519,
      "cbid": 251, "correlation": 88519
    }
  },
  {
    "ph": "f", "id": 88519, "pid": 76337, "tid": -914061504, "ts": 1716454222740922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740926, "dur": 0,
    "args": {
      "External id": 88520,
      "cbid": 251, "correlation": 88520
    }
  },
  {
    "ph": "f", "id": 88520, "pid": 76337, "tid": -914061504, "ts": 1716454222740926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222792522, "dur": 7,
    "args": {
      "External id": 88521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88521,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 88521, "pid": 5, "tid": 7, "ts": 1716454222792522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740927, "dur": 11,
    "args": {
      "External id": 88521,
      "cbid": 211, "correlation": 88521
    }
  },
  {
    "ph": "s", "id": 88521, "pid": 76337, "tid": -914061504, "ts": 1716454222740927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222792531, "dur": 3,
    "args": {
      "External id": 88523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88523,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 88523, "pid": 5, "tid": 7, "ts": 1716454222792531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222740940, "dur": 6,
    "args": {
      "External id": 88523,
      "cbid": 211, "correlation": 88523
    }
  },
  {
    "ph": "s", "id": 88523, "pid": 76337, "tid": -914061504, "ts": 1716454222740940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222792535, "dur": 90,
    "args": {
      "External id": 88544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88544,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 88544, "pid": 5, "tid": 7, "ts": 1716454222792535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741025, "dur": 13,
    "args": {
      "External id": 88544,
      "cbid": 211, "correlation": 88544
    }
  },
  {
    "ph": "s", "id": 88544, "pid": 76337, "tid": -914061504, "ts": 1716454222741025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741124, "dur": 1,
    "args": {
      "External id": 88562,
      "cbid": 251, "correlation": 88562
    }
  },
  {
    "ph": "f", "id": 88562, "pid": 76337, "tid": -914061504, "ts": 1716454222741124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222792626, "dur": 97,
    "args": {
      "External id": 88564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88564,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88564, "pid": 5, "tid": 7, "ts": 1716454222792626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741130, "dur": 14,
    "args": {
      "External id": 88564,
      "cbid": 211, "correlation": 88564
    }
  },
  {
    "ph": "s", "id": 88564, "pid": 76337, "tid": -914061504, "ts": 1716454222741130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222792724, "dur": 19,
    "args": {
      "External id": 88572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88572,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88572, "pid": 5, "tid": 7, "ts": 1716454222792724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741202, "dur": 12,
    "args": {
      "External id": 88572,
      "cbid": 211, "correlation": 88572
    }
  },
  {
    "ph": "s", "id": 88572, "pid": 76337, "tid": -914061504, "ts": 1716454222741202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222792745, "dur": 38,
    "args": {
      "External id": 88580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88580,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88580, "pid": 5, "tid": 7, "ts": 1716454222792745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741243, "dur": 10,
    "args": {
      "External id": 88580,
      "cbid": 211, "correlation": 88580
    }
  },
  {
    "ph": "s", "id": 88580, "pid": 76337, "tid": -914061504, "ts": 1716454222741243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222792784, "dur": 34,
    "args": {
      "External id": 88602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88602,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88602, "pid": 5, "tid": 7, "ts": 1716454222792784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741295, "dur": 10,
    "args": {
      "External id": 88602,
      "cbid": 211, "correlation": 88602
    }
  },
  {
    "ph": "s", "id": 88602, "pid": 76337, "tid": -914061504, "ts": 1716454222741295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741386, "dur": 1,
    "args": {
      "External id": 88618,
      "cbid": 251, "correlation": 88618
    }
  },
  {
    "ph": "f", "id": 88618, "pid": 76337, "tid": -914061504, "ts": 1716454222741386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741391, "dur": 0,
    "args": {
      "External id": 88620,
      "cbid": 251, "correlation": 88620
    }
  },
  {
    "ph": "f", "id": 88620, "pid": 76337, "tid": -914061504, "ts": 1716454222741391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222792819, "dur": 529,
    "args": {
      "External id": 88621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88621,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 88621, "pid": 5, "tid": 7, "ts": 1716454222792819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741395, "dur": 13,
    "args": {
      "External id": 88621,
      "cbid": 211, "correlation": 88621
    }
  },
  {
    "ph": "s", "id": 88621, "pid": 76337, "tid": -914061504, "ts": 1716454222741395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222793349, "dur": 125,
    "args": {
      "External id": 88629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88629, "pid": 5, "tid": 7, "ts": 1716454222793349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741463, "dur": 12,
    "args": {
      "External id": 88629,
      "cbid": 211, "correlation": 88629
    }
  },
  {
    "ph": "s", "id": 88629, "pid": 76337, "tid": -914061504, "ts": 1716454222741463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222793475, "dur": 127,
    "args": {
      "External id": 88637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88637, "pid": 5, "tid": 7, "ts": 1716454222793475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741495, "dur": 9,
    "args": {
      "External id": 88637,
      "cbid": 211, "correlation": 88637
    }
  },
  {
    "ph": "s", "id": 88637, "pid": 76337, "tid": -914061504, "ts": 1716454222741495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741571, "dur": 1,
    "args": {
      "External id": 88653,
      "cbid": 251, "correlation": 88653
    }
  },
  {
    "ph": "f", "id": 88653, "pid": 76337, "tid": -914061504, "ts": 1716454222741571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222793603, "dur": 300,
    "args": {
      "External id": 88655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88655,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88655, "pid": 5, "tid": 7, "ts": 1716454222793603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741577, "dur": 12,
    "args": {
      "External id": 88655,
      "cbid": 211, "correlation": 88655
    }
  },
  {
    "ph": "s", "id": 88655, "pid": 76337, "tid": -914061504, "ts": 1716454222741577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222793905, "dur": 28,
    "args": {
      "External id": 88663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88663,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88663, "pid": 5, "tid": 7, "ts": 1716454222793905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741619, "dur": 10,
    "args": {
      "External id": 88663,
      "cbid": 211, "correlation": 88663
    }
  },
  {
    "ph": "s", "id": 88663, "pid": 76337, "tid": -914061504, "ts": 1716454222741619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222793933, "dur": 80,
    "args": {
      "External id": 88674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88674, "pid": 5, "tid": 7, "ts": 1716454222793933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741689, "dur": 12,
    "args": {
      "External id": 88674,
      "cbid": 211, "correlation": 88674
    }
  },
  {
    "ph": "s", "id": 88674, "pid": 76337, "tid": -914061504, "ts": 1716454222741689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741755, "dur": 0,
    "args": {
      "External id": 88686,
      "cbid": 317, "correlation": 88686
    }
  },
  {
    "ph": "f", "id": 88686, "pid": 76337, "tid": -914061504, "ts": 1716454222741755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741756, "dur": 0,
    "args": {
      "External id": 88687,
      "cbid": 203, "correlation": 88687
    }
  },
  {
    "ph": "f", "id": 88687, "pid": 76337, "tid": -914061504, "ts": 1716454222741756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741757, "dur": 0,
    "args": {
      "External id": 88688,
      "cbid": 205, "correlation": 88688
    }
  },
  {
    "ph": "f", "id": 88688, "pid": 76337, "tid": -914061504, "ts": 1716454222741757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794014, "dur": 23,
    "args": {
      "External id": 88692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88692, "pid": 5, "tid": 7, "ts": 1716454222794014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741774, "dur": 12,
    "args": {
      "External id": 88692,
      "cbid": 211, "correlation": 88692
    }
  },
  {
    "ph": "s", "id": 88692, "pid": 76337, "tid": -914061504, "ts": 1716454222741774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222794038, "dur": 118,
    "args": {
      "External id": 88694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88694,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88694, "pid": 5, "tid": 7, "ts": 1716454222794038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741793, "dur": 7,
    "args": {
      "External id": 88694,
      "cbid": 211, "correlation": 88694
    }
  },
  {
    "ph": "s", "id": 88694, "pid": 76337, "tid": -914061504, "ts": 1716454222741793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794158, "dur": 22,
    "args": {
      "External id": 88696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88696,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88696, "pid": 5, "tid": 7, "ts": 1716454222794158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741804, "dur": 5,
    "args": {
      "External id": 88696,
      "cbid": 211, "correlation": 88696
    }
  },
  {
    "ph": "s", "id": 88696, "pid": 76337, "tid": -914061504, "ts": 1716454222741804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222794182, "dur": 32,
    "args": {
      "External id": 88702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88702, "pid": 5, "tid": 7, "ts": 1716454222794182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741831, "dur": 8,
    "args": {
      "External id": 88702,
      "cbid": 211, "correlation": 88702
    }
  },
  {
    "ph": "s", "id": 88702, "pid": 76337, "tid": -914061504, "ts": 1716454222741831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222794215, "dur": 27,
    "args": {
      "External id": 88710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88710,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88710, "pid": 5, "tid": 7, "ts": 1716454222794215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741864, "dur": 9,
    "args": {
      "External id": 88710,
      "cbid": 211, "correlation": 88710
    }
  },
  {
    "ph": "s", "id": 88710, "pid": 76337, "tid": -914061504, "ts": 1716454222741864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741938, "dur": 0,
    "args": {
      "External id": 88720,
      "cbid": 317, "correlation": 88720
    }
  },
  {
    "ph": "f", "id": 88720, "pid": 76337, "tid": -914061504, "ts": 1716454222741938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741939, "dur": 0,
    "args": {
      "External id": 88721,
      "cbid": 203, "correlation": 88721
    }
  },
  {
    "ph": "f", "id": 88721, "pid": 76337, "tid": -914061504, "ts": 1716454222741939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741939, "dur": 0,
    "args": {
      "External id": 88722,
      "cbid": 205, "correlation": 88722
    }
  },
  {
    "ph": "f", "id": 88722, "pid": 76337, "tid": -914061504, "ts": 1716454222741939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794243, "dur": 23,
    "args": {
      "External id": 88726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88726, "pid": 5, "tid": 7, "ts": 1716454222794243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741954, "dur": 12,
    "args": {
      "External id": 88726,
      "cbid": 211, "correlation": 88726
    }
  },
  {
    "ph": "s", "id": 88726, "pid": 76337, "tid": -914061504, "ts": 1716454222741954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794267, "dur": 43,
    "args": {
      "External id": 88728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88728,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88728, "pid": 5, "tid": 7, "ts": 1716454222794267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741968, "dur": 14,
    "args": {
      "External id": 88728,
      "cbid": 211, "correlation": 88728
    }
  },
  {
    "ph": "s", "id": 88728, "pid": 76337, "tid": -914061504, "ts": 1716454222741968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222794312, "dur": 232,
    "args": {
      "External id": 88730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88730,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 88730, "pid": 5, "tid": 7, "ts": 1716454222794312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222741989, "dur": 7,
    "args": {
      "External id": 88730,
      "cbid": 211, "correlation": 88730
    }
  },
  {
    "ph": "s", "id": 88730, "pid": 76337, "tid": -914061504, "ts": 1716454222741989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794545, "dur": 6,
    "args": {
      "External id": 88732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88732, "pid": 5, "tid": 7, "ts": 1716454222794545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742000, "dur": 5,
    "args": {
      "External id": 88732,
      "cbid": 211, "correlation": 88732
    }
  },
  {
    "ph": "s", "id": 88732, "pid": 76337, "tid": -914061504, "ts": 1716454222742000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222794552, "dur": 9,
    "args": {
      "External id": 88738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88738, "pid": 5, "tid": 7, "ts": 1716454222794552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742027, "dur": 9,
    "args": {
      "External id": 88738,
      "cbid": 211, "correlation": 88738
    }
  },
  {
    "ph": "s", "id": 88738, "pid": 76337, "tid": -914061504, "ts": 1716454222742027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794563, "dur": 12,
    "args": {
      "External id": 88758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88758,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 88758, "pid": 5, "tid": 7, "ts": 1716454222794563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742125, "dur": 12,
    "args": {
      "External id": 88758,
      "cbid": 211, "correlation": 88758
    }
  },
  {
    "ph": "s", "id": 88758, "pid": 76337, "tid": -914061504, "ts": 1716454222742125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222794576, "dur": 4,
    "args": {
      "External id": 88770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88770,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 88770, "pid": 5, "tid": 7, "ts": 1716454222794576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742148, "dur": 7,
    "args": {
      "External id": 88770,
      "cbid": 211, "correlation": 88770
    }
  },
  {
    "ph": "s", "id": 88770, "pid": 76337, "tid": -914061504, "ts": 1716454222742148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222794581, "dur": 11,
    "args": {
      "External id": 88773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88773,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88773, "pid": 5, "tid": 7, "ts": 1716454222794581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742168, "dur": 7,
    "args": {
      "External id": 88773,
      "cbid": 211, "correlation": 88773
    }
  },
  {
    "ph": "s", "id": 88773, "pid": 76337, "tid": -914061504, "ts": 1716454222742168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222794594, "dur": 7,
    "args": {
      "External id": 88782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88782,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88782, "pid": 5, "tid": 7, "ts": 1716454222794594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742210, "dur": 10,
    "args": {
      "External id": 88782,
      "cbid": 211, "correlation": 88782
    }
  },
  {
    "ph": "s", "id": 88782, "pid": 76337, "tid": -914061504, "ts": 1716454222742210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742261, "dur": 0,
    "args": {
      "External id": 88792,
      "cbid": 317, "correlation": 88792
    }
  },
  {
    "ph": "f", "id": 88792, "pid": 76337, "tid": -914061504, "ts": 1716454222742261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742263, "dur": 0,
    "args": {
      "External id": 88793,
      "cbid": 203, "correlation": 88793
    }
  },
  {
    "ph": "f", "id": 88793, "pid": 76337, "tid": -914061504, "ts": 1716454222742263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742263, "dur": 0,
    "args": {
      "External id": 88794,
      "cbid": 205, "correlation": 88794
    }
  },
  {
    "ph": "f", "id": 88794, "pid": 76337, "tid": -914061504, "ts": 1716454222742263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794602, "dur": 5,
    "args": {
      "External id": 88798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88798,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88798, "pid": 5, "tid": 7, "ts": 1716454222794602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742279, "dur": 11,
    "args": {
      "External id": 88798,
      "cbid": 211, "correlation": 88798
    }
  },
  {
    "ph": "s", "id": 88798, "pid": 76337, "tid": -914061504, "ts": 1716454222742279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222794608, "dur": 83,
    "args": {
      "External id": 88800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88800, "pid": 5, "tid": 7, "ts": 1716454222794608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742293, "dur": 5,
    "args": {
      "External id": 88800,
      "cbid": 211, "correlation": 88800
    }
  },
  {
    "ph": "s", "id": 88800, "pid": 76337, "tid": -914061504, "ts": 1716454222742293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222794694, "dur": 1,
    "args": {
      "External id": 88802,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 88802,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 88802, "pid": 5, "tid": 7, "ts": 1716454222794694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742306, "dur": 13,
    "args": {
      "External id": 88802,
      "cbid": 51, "correlation": 88802
    }
  },
  {
    "ph": "s", "id": 88802, "pid": 76337, "tid": -914061504, "ts": 1716454222742306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222794698, "dur": 535,
    "args": {
      "External id": 88803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88803,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88803, "pid": 5, "tid": 7, "ts": 1716454222794698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742320, "dur": 9,
    "args": {
      "External id": 88803,
      "cbid": 211, "correlation": 88803
    }
  },
  {
    "ph": "s", "id": 88803, "pid": 76337, "tid": -914061504, "ts": 1716454222742320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222795234, "dur": 11,
    "args": {
      "External id": 88805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88805,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88805, "pid": 5, "tid": 7, "ts": 1716454222795234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742333, "dur": 5,
    "args": {
      "External id": 88805,
      "cbid": 211, "correlation": 88805
    }
  },
  {
    "ph": "s", "id": 88805, "pid": 76337, "tid": -914061504, "ts": 1716454222742333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222795247, "dur": 14,
    "args": {
      "External id": 88811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88811,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88811, "pid": 5, "tid": 7, "ts": 1716454222795247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742361, "dur": 9,
    "args": {
      "External id": 88811,
      "cbid": 211, "correlation": 88811
    }
  },
  {
    "ph": "s", "id": 88811, "pid": 76337, "tid": -914061504, "ts": 1716454222742361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222795262, "dur": 3,
    "args": {
      "External id": 88819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88819,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 88819, "pid": 5, "tid": 7, "ts": 1716454222795262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742405, "dur": 9,
    "args": {
      "External id": 88819,
      "cbid": 211, "correlation": 88819
    }
  },
  {
    "ph": "s", "id": 88819, "pid": 76337, "tid": -914061504, "ts": 1716454222742405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742473, "dur": 1,
    "args": {
      "External id": 88835,
      "cbid": 251, "correlation": 88835
    }
  },
  {
    "ph": "f", "id": 88835, "pid": 76337, "tid": -914061504, "ts": 1716454222742473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742479, "dur": 0,
    "args": {
      "External id": 88837,
      "cbid": 251, "correlation": 88837
    }
  },
  {
    "ph": "f", "id": 88837, "pid": 76337, "tid": -914061504, "ts": 1716454222742479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222795267, "dur": 13,
    "args": {
      "External id": 88838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88838,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88838, "pid": 5, "tid": 7, "ts": 1716454222795267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742481, "dur": 12,
    "args": {
      "External id": 88838,
      "cbid": 211, "correlation": 88838
    }
  },
  {
    "ph": "s", "id": 88838, "pid": 76337, "tid": -914061504, "ts": 1716454222742481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222795281, "dur": 5,
    "args": {
      "External id": 88840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88840,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88840, "pid": 5, "tid": 7, "ts": 1716454222795281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742495, "dur": 5,
    "args": {
      "External id": 88840,
      "cbid": 211, "correlation": 88840
    }
  },
  {
    "ph": "s", "id": 88840, "pid": 76337, "tid": -914061504, "ts": 1716454222742495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222795287, "dur": 16,
    "args": {
      "External id": 88850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88850,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88850, "pid": 5, "tid": 7, "ts": 1716454222795287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742555, "dur": 12,
    "args": {
      "External id": 88850,
      "cbid": 211, "correlation": 88850
    }
  },
  {
    "ph": "s", "id": 88850, "pid": 76337, "tid": -914061504, "ts": 1716454222742555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222795305, "dur": 17,
    "args": {
      "External id": 88870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88870,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 88870, "pid": 5, "tid": 7, "ts": 1716454222795305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742622, "dur": 11,
    "args": {
      "External id": 88870,
      "cbid": 211, "correlation": 88870
    }
  },
  {
    "ph": "s", "id": 88870, "pid": 76337, "tid": -914061504, "ts": 1716454222742622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222795323, "dur": 5,
    "args": {
      "External id": 88882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88882,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 88882, "pid": 5, "tid": 7, "ts": 1716454222795323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742644, "dur": 6,
    "args": {
      "External id": 88882,
      "cbid": 211, "correlation": 88882
    }
  },
  {
    "ph": "s", "id": 88882, "pid": 76337, "tid": -914061504, "ts": 1716454222742644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222795329, "dur": 17,
    "args": {
      "External id": 88885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88885,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88885, "pid": 5, "tid": 7, "ts": 1716454222795329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742663, "dur": 7,
    "args": {
      "External id": 88885,
      "cbid": 211, "correlation": 88885
    }
  },
  {
    "ph": "s", "id": 88885, "pid": 76337, "tid": -914061504, "ts": 1716454222742663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222795347, "dur": 11,
    "args": {
      "External id": 88894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88894,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88894, "pid": 5, "tid": 7, "ts": 1716454222795347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742704, "dur": 10,
    "args": {
      "External id": 88894,
      "cbid": 211, "correlation": 88894
    }
  },
  {
    "ph": "s", "id": 88894, "pid": 76337, "tid": -914061504, "ts": 1716454222742704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742767, "dur": 0,
    "args": {
      "External id": 88904,
      "cbid": 317, "correlation": 88904
    }
  },
  {
    "ph": "f", "id": 88904, "pid": 76337, "tid": -914061504, "ts": 1716454222742767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742768, "dur": 0,
    "args": {
      "External id": 88905,
      "cbid": 203, "correlation": 88905
    }
  },
  {
    "ph": "f", "id": 88905, "pid": 76337, "tid": -914061504, "ts": 1716454222742768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742768, "dur": 0,
    "args": {
      "External id": 88906,
      "cbid": 205, "correlation": 88906
    }
  },
  {
    "ph": "f", "id": 88906, "pid": 76337, "tid": -914061504, "ts": 1716454222742768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222795359, "dur": 10,
    "args": {
      "External id": 88910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88910,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88910, "pid": 5, "tid": 7, "ts": 1716454222795359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742785, "dur": 12,
    "args": {
      "External id": 88910,
      "cbid": 211, "correlation": 88910
    }
  },
  {
    "ph": "s", "id": 88910, "pid": 76337, "tid": -914061504, "ts": 1716454222742785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222795371, "dur": 160,
    "args": {
      "External id": 88912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88912, "pid": 5, "tid": 7, "ts": 1716454222795371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742800, "dur": 5,
    "args": {
      "External id": 88912,
      "cbid": 211, "correlation": 88912
    }
  },
  {
    "ph": "s", "id": 88912, "pid": 76337, "tid": -914061504, "ts": 1716454222742800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222795533, "dur": 1,
    "args": {
      "External id": 88914,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 88914,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 88914, "pid": 5, "tid": 7, "ts": 1716454222795533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742812, "dur": 8,
    "args": {
      "External id": 88914,
      "cbid": 51, "correlation": 88914
    }
  },
  {
    "ph": "s", "id": 88914, "pid": 76337, "tid": -914061504, "ts": 1716454222742812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222795537, "dur": 652,
    "args": {
      "External id": 88915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88915,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 88915, "pid": 5, "tid": 7, "ts": 1716454222795537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742822, "dur": 8,
    "args": {
      "External id": 88915,
      "cbid": 211, "correlation": 88915
    }
  },
  {
    "ph": "s", "id": 88915, "pid": 76337, "tid": -914061504, "ts": 1716454222742822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222796191, "dur": 13,
    "args": {
      "External id": 88917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88917,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88917, "pid": 5, "tid": 7, "ts": 1716454222796191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742834, "dur": 5,
    "args": {
      "External id": 88917,
      "cbid": 211, "correlation": 88917
    }
  },
  {
    "ph": "s", "id": 88917, "pid": 76337, "tid": -914061504, "ts": 1716454222742834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222796205, "dur": 14,
    "args": {
      "External id": 88923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88923, "pid": 5, "tid": 7, "ts": 1716454222796205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742862, "dur": 8,
    "args": {
      "External id": 88923,
      "cbid": 211, "correlation": 88923
    }
  },
  {
    "ph": "s", "id": 88923, "pid": 76337, "tid": -914061504, "ts": 1716454222742862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742920, "dur": 0,
    "args": {
      "External id": 88933,
      "cbid": 317, "correlation": 88933
    }
  },
  {
    "ph": "f", "id": 88933, "pid": 76337, "tid": -914061504, "ts": 1716454222742920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742921, "dur": 0,
    "args": {
      "External id": 88934,
      "cbid": 203, "correlation": 88934
    }
  },
  {
    "ph": "f", "id": 88934, "pid": 76337, "tid": -914061504, "ts": 1716454222742921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742922, "dur": 0,
    "args": {
      "External id": 88935,
      "cbid": 205, "correlation": 88935
    }
  },
  {
    "ph": "f", "id": 88935, "pid": 76337, "tid": -914061504, "ts": 1716454222742922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222796220, "dur": 8,
    "args": {
      "External id": 88939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88939,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88939, "pid": 5, "tid": 7, "ts": 1716454222796220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742936, "dur": 13,
    "args": {
      "External id": 88939,
      "cbid": 211, "correlation": 88939
    }
  },
  {
    "ph": "s", "id": 88939, "pid": 76337, "tid": -914061504, "ts": 1716454222742936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222796230, "dur": 4,
    "args": {
      "External id": 88941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 88941, "pid": 5, "tid": 7, "ts": 1716454222796230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742955, "dur": 6,
    "args": {
      "External id": 88941,
      "cbid": 211, "correlation": 88941
    }
  },
  {
    "ph": "s", "id": 88941, "pid": 76337, "tid": -914061504, "ts": 1716454222742955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742966, "dur": 0,
    "args": {
      "External id": 88942,
      "cbid": 51, "correlation": 88942
    }
  },
  {
    "ph": "s", "id": 88942, "pid": 76337, "tid": -914061504, "ts": 1716454222742966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222796235, "dur": 56,
    "args": {
      "External id": 88943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88943,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 88943, "pid": 5, "tid": 7, "ts": 1716454222796235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222742966, "dur": 6,
    "args": {
      "External id": 88943,
      "cbid": 211, "correlation": 88943
    }
  },
  {
    "ph": "s", "id": 88943, "pid": 76337, "tid": -914061504, "ts": 1716454222742966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222796292, "dur": 14,
    "args": {
      "External id": 88948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88948, "pid": 5, "tid": 7, "ts": 1716454222796292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743000, "dur": 10,
    "args": {
      "External id": 88948,
      "cbid": 211, "correlation": 88948
    }
  },
  {
    "ph": "s", "id": 88948, "pid": 76337, "tid": -914061504, "ts": 1716454222743000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222796307, "dur": 12,
    "args": {
      "External id": 88956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88956,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88956, "pid": 5, "tid": 7, "ts": 1716454222796307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743030, "dur": 8,
    "args": {
      "External id": 88956,
      "cbid": 211, "correlation": 88956
    }
  },
  {
    "ph": "s", "id": 88956, "pid": 76337, "tid": -914061504, "ts": 1716454222743030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222796321, "dur": 10,
    "args": {
      "External id": 88964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88964,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88964, "pid": 5, "tid": 7, "ts": 1716454222796321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743061, "dur": 9,
    "args": {
      "External id": 88964,
      "cbid": 211, "correlation": 88964
    }
  },
  {
    "ph": "s", "id": 88964, "pid": 76337, "tid": -914061504, "ts": 1716454222743061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222796332, "dur": 18,
    "args": {
      "External id": 88984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88984,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 88984, "pid": 5, "tid": 7, "ts": 1716454222796332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743143, "dur": 12,
    "args": {
      "External id": 88984,
      "cbid": 211, "correlation": 88984
    }
  },
  {
    "ph": "s", "id": 88984, "pid": 76337, "tid": -914061504, "ts": 1716454222743143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222796351, "dur": 4,
    "args": {
      "External id": 88996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88996,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 88996, "pid": 5, "tid": 7, "ts": 1716454222796351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743165, "dur": 7,
    "args": {
      "External id": 88996,
      "cbid": 211, "correlation": 88996
    }
  },
  {
    "ph": "s", "id": 88996, "pid": 76337, "tid": -914061504, "ts": 1716454222743165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222796357, "dur": 17,
    "args": {
      "External id": 88999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 88999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 88999, "pid": 5, "tid": 7, "ts": 1716454222796357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743184, "dur": 6,
    "args": {
      "External id": 88999,
      "cbid": 211, "correlation": 88999
    }
  },
  {
    "ph": "s", "id": 88999, "pid": 76337, "tid": -914061504, "ts": 1716454222743184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743241, "dur": 0,
    "args": {
      "External id": 89010,
      "cbid": 317, "correlation": 89010
    }
  },
  {
    "ph": "f", "id": 89010, "pid": 76337, "tid": -914061504, "ts": 1716454222743241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743242, "dur": 0,
    "args": {
      "External id": 89011,
      "cbid": 203, "correlation": 89011
    }
  },
  {
    "ph": "f", "id": 89011, "pid": 76337, "tid": -914061504, "ts": 1716454222743242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743243, "dur": 0,
    "args": {
      "External id": 89012,
      "cbid": 205, "correlation": 89012
    }
  },
  {
    "ph": "f", "id": 89012, "pid": 76337, "tid": -914061504, "ts": 1716454222743243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222796375, "dur": 12,
    "args": {
      "External id": 89016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89016,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89016, "pid": 5, "tid": 7, "ts": 1716454222796375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743258, "dur": 11,
    "args": {
      "External id": 89016,
      "cbid": 211, "correlation": 89016
    }
  },
  {
    "ph": "s", "id": 89016, "pid": 76337, "tid": -914061504, "ts": 1716454222743258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222796389, "dur": 3,
    "args": {
      "External id": 89018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89018, "pid": 5, "tid": 7, "ts": 1716454222796389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743275, "dur": 6,
    "args": {
      "External id": 89018,
      "cbid": 211, "correlation": 89018
    }
  },
  {
    "ph": "s", "id": 89018, "pid": 76337, "tid": -914061504, "ts": 1716454222743275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743285, "dur": 0,
    "args": {
      "External id": 89019,
      "cbid": 51, "correlation": 89019
    }
  },
  {
    "ph": "s", "id": 89019, "pid": 76337, "tid": -914061504, "ts": 1716454222743285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222796393, "dur": 96,
    "args": {
      "External id": 89020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89020,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 89020, "pid": 5, "tid": 7, "ts": 1716454222796393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743286, "dur": 6,
    "args": {
      "External id": 89020,
      "cbid": 211, "correlation": 89020
    }
  },
  {
    "ph": "s", "id": 89020, "pid": 76337, "tid": -914061504, "ts": 1716454222743286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222796490, "dur": 15,
    "args": {
      "External id": 89025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89025, "pid": 5, "tid": 7, "ts": 1716454222796490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743314, "dur": 8,
    "args": {
      "External id": 89025,
      "cbid": 211, "correlation": 89025
    }
  },
  {
    "ph": "s", "id": 89025, "pid": 76337, "tid": -914061504, "ts": 1716454222743314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222796506, "dur": 81,
    "args": {
      "External id": 89034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89034,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89034, "pid": 5, "tid": 7, "ts": 1716454222796506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743400, "dur": 16,
    "args": {
      "External id": 89034,
      "cbid": 211, "correlation": 89034
    }
  },
  {
    "ph": "s", "id": 89034, "pid": 76337, "tid": -914061504, "ts": 1716454222743400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222796589, "dur": 30,
    "args": {
      "External id": 89056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89056,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89056, "pid": 5, "tid": 7, "ts": 1716454222796589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743460, "dur": 10,
    "args": {
      "External id": 89056,
      "cbid": 211, "correlation": 89056
    }
  },
  {
    "ph": "s", "id": 89056, "pid": 76337, "tid": -914061504, "ts": 1716454222743460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743554, "dur": 1,
    "args": {
      "External id": 89067,
      "cbid": 251, "correlation": 89067
    }
  },
  {
    "ph": "f", "id": 89067, "pid": 76337, "tid": -914061504, "ts": 1716454222743554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222796621, "dur": 164,
    "args": {
      "External id": 89068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89068,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89068, "pid": 5, "tid": 7, "ts": 1716454222796621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743560, "dur": 14,
    "args": {
      "External id": 89068,
      "cbid": 211, "correlation": 89068
    }
  },
  {
    "ph": "s", "id": 89068, "pid": 76337, "tid": -914061504, "ts": 1716454222743560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743632, "dur": 1,
    "args": {
      "External id": 89079,
      "cbid": 251, "correlation": 89079
    }
  },
  {
    "ph": "f", "id": 89079, "pid": 76337, "tid": -914061504, "ts": 1716454222743632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222796785, "dur": 157,
    "args": {
      "External id": 89080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89080,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89080, "pid": 5, "tid": 7, "ts": 1716454222796785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743636, "dur": 11,
    "args": {
      "External id": 89080,
      "cbid": 211, "correlation": 89080
    }
  },
  {
    "ph": "s", "id": 89080, "pid": 76337, "tid": -914061504, "ts": 1716454222743636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743701, "dur": 1,
    "args": {
      "External id": 89091,
      "cbid": 251, "correlation": 89091
    }
  },
  {
    "ph": "f", "id": 89091, "pid": 76337, "tid": -914061504, "ts": 1716454222743701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222796944, "dur": 158,
    "args": {
      "External id": 89092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89092,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89092, "pid": 5, "tid": 7, "ts": 1716454222796944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743705, "dur": 11,
    "args": {
      "External id": 89092,
      "cbid": 211, "correlation": 89092
    }
  },
  {
    "ph": "s", "id": 89092, "pid": 76337, "tid": -914061504, "ts": 1716454222743705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222797103, "dur": 334,
    "args": {
      "External id": 89117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89117,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89117, "pid": 5, "tid": 7, "ts": 1716454222797103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743797, "dur": 14,
    "args": {
      "External id": 89117,
      "cbid": 211, "correlation": 89117
    }
  },
  {
    "ph": "s", "id": 89117, "pid": 76337, "tid": -914061504, "ts": 1716454222743797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743904, "dur": 1,
    "args": {
      "External id": 89135,
      "cbid": 251, "correlation": 89135
    }
  },
  {
    "ph": "f", "id": 89135, "pid": 76337, "tid": -914061504, "ts": 1716454222743904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222797439, "dur": 165,
    "args": {
      "External id": 89137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89137,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89137, "pid": 5, "tid": 7, "ts": 1716454222797439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743910, "dur": 13,
    "args": {
      "External id": 89137,
      "cbid": 211, "correlation": 89137
    }
  },
  {
    "ph": "s", "id": 89137, "pid": 76337, "tid": -914061504, "ts": 1716454222743910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222797605, "dur": 19,
    "args": {
      "External id": 89145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89145,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89145, "pid": 5, "tid": 7, "ts": 1716454222797605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222743991, "dur": 13,
    "args": {
      "External id": 89145,
      "cbid": 211, "correlation": 89145
    }
  },
  {
    "ph": "s", "id": 89145, "pid": 76337, "tid": -914061504, "ts": 1716454222743991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222797625, "dur": 28,
    "args": {
      "External id": 89153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89153,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89153, "pid": 5, "tid": 7, "ts": 1716454222797625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744033, "dur": 9,
    "args": {
      "External id": 89153,
      "cbid": 211, "correlation": 89153
    }
  },
  {
    "ph": "s", "id": 89153, "pid": 76337, "tid": -914061504, "ts": 1716454222744033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222797654, "dur": 18,
    "args": {
      "External id": 89164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89164, "pid": 5, "tid": 7, "ts": 1716454222797654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744109, "dur": 13,
    "args": {
      "External id": 89164,
      "cbid": 211, "correlation": 89164
    }
  },
  {
    "ph": "s", "id": 89164, "pid": 76337, "tid": -914061504, "ts": 1716454222744109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222797673, "dur": 16,
    "args": {
      "External id": 89186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89186,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89186, "pid": 5, "tid": 7, "ts": 1716454222797673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744142, "dur": 7,
    "args": {
      "External id": 89186,
      "cbid": 211, "correlation": 89186
    }
  },
  {
    "ph": "s", "id": 89186, "pid": 76337, "tid": -914061504, "ts": 1716454222744142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744228, "dur": 2,
    "args": {
      "External id": 89197,
      "cbid": 251, "correlation": 89197
    }
  },
  {
    "ph": "f", "id": 89197, "pid": 76337, "tid": -914061504, "ts": 1716454222744228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222797690, "dur": 88,
    "args": {
      "External id": 89198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89198,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89198, "pid": 5, "tid": 7, "ts": 1716454222797690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744234, "dur": 15,
    "args": {
      "External id": 89198,
      "cbid": 211, "correlation": 89198
    }
  },
  {
    "ph": "s", "id": 89198, "pid": 76337, "tid": -914061504, "ts": 1716454222744234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744309, "dur": 1,
    "args": {
      "External id": 89209,
      "cbid": 251, "correlation": 89209
    }
  },
  {
    "ph": "f", "id": 89209, "pid": 76337, "tid": -914061504, "ts": 1716454222744309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744313, "dur": 0,
    "args": {
      "External id": 89210,
      "cbid": 251, "correlation": 89210
    }
  },
  {
    "ph": "f", "id": 89210, "pid": 76337, "tid": -914061504, "ts": 1716454222744313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222797780, "dur": 11,
    "args": {
      "External id": 89211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89211,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89211, "pid": 5, "tid": 7, "ts": 1716454222797780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744314, "dur": 12,
    "args": {
      "External id": 89211,
      "cbid": 211, "correlation": 89211
    }
  },
  {
    "ph": "s", "id": 89211, "pid": 76337, "tid": -914061504, "ts": 1716454222744314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222797792, "dur": 5,
    "args": {
      "External id": 89213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89213,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89213, "pid": 5, "tid": 7, "ts": 1716454222797792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744329, "dur": 6,
    "args": {
      "External id": 89213,
      "cbid": 211, "correlation": 89213
    }
  },
  {
    "ph": "s", "id": 89213, "pid": 76337, "tid": -914061504, "ts": 1716454222744329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744387, "dur": 1,
    "args": {
      "External id": 89224,
      "cbid": 251, "correlation": 89224
    }
  },
  {
    "ph": "f", "id": 89224, "pid": 76337, "tid": -914061504, "ts": 1716454222744387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744390, "dur": 0,
    "args": {
      "External id": 89225,
      "cbid": 251, "correlation": 89225
    }
  },
  {
    "ph": "f", "id": 89225, "pid": 76337, "tid": -914061504, "ts": 1716454222744390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222797799, "dur": 8,
    "args": {
      "External id": 89226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89226,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89226, "pid": 5, "tid": 7, "ts": 1716454222797799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744391, "dur": 11,
    "args": {
      "External id": 89226,
      "cbid": 211, "correlation": 89226
    }
  },
  {
    "ph": "s", "id": 89226, "pid": 76337, "tid": -914061504, "ts": 1716454222744391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222797809, "dur": 3,
    "args": {
      "External id": 89228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89228,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89228, "pid": 5, "tid": 7, "ts": 1716454222797809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744404, "dur": 6,
    "args": {
      "External id": 89228,
      "cbid": 211, "correlation": 89228
    }
  },
  {
    "ph": "s", "id": 89228, "pid": 76337, "tid": -914061504, "ts": 1716454222744404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222797813, "dur": 54,
    "args": {
      "External id": 89253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89253,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89253, "pid": 5, "tid": 7, "ts": 1716454222797813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744483, "dur": 13,
    "args": {
      "External id": 89253,
      "cbid": 211, "correlation": 89253
    }
  },
  {
    "ph": "s", "id": 89253, "pid": 76337, "tid": -914061504, "ts": 1716454222744483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744583, "dur": 2,
    "args": {
      "External id": 89271,
      "cbid": 251, "correlation": 89271
    }
  },
  {
    "ph": "f", "id": 89271, "pid": 76337, "tid": -914061504, "ts": 1716454222744583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222797869, "dur": 90,
    "args": {
      "External id": 89273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89273,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89273, "pid": 5, "tid": 7, "ts": 1716454222797869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744590, "dur": 14,
    "args": {
      "External id": 89273,
      "cbid": 211, "correlation": 89273
    }
  },
  {
    "ph": "s", "id": 89273, "pid": 76337, "tid": -914061504, "ts": 1716454222744590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222797960, "dur": 10,
    "args": {
      "External id": 89281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89281,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89281, "pid": 5, "tid": 7, "ts": 1716454222797960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744661, "dur": 12,
    "args": {
      "External id": 89281,
      "cbid": 211, "correlation": 89281
    }
  },
  {
    "ph": "s", "id": 89281, "pid": 76337, "tid": -914061504, "ts": 1716454222744661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222797971, "dur": 22,
    "args": {
      "External id": 89289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89289, "pid": 5, "tid": 7, "ts": 1716454222797971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744704, "dur": 9,
    "args": {
      "External id": 89289,
      "cbid": 211, "correlation": 89289
    }
  },
  {
    "ph": "s", "id": 89289, "pid": 76337, "tid": -914061504, "ts": 1716454222744704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222797994, "dur": 18,
    "args": {
      "External id": 89311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89311,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89311, "pid": 5, "tid": 7, "ts": 1716454222797994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744755, "dur": 11,
    "args": {
      "External id": 89311,
      "cbid": 211, "correlation": 89311
    }
  },
  {
    "ph": "s", "id": 89311, "pid": 76337, "tid": -914061504, "ts": 1716454222744755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744849, "dur": 2,
    "args": {
      "External id": 89327,
      "cbid": 251, "correlation": 89327
    }
  },
  {
    "ph": "f", "id": 89327, "pid": 76337, "tid": -914061504, "ts": 1716454222744849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744854, "dur": 0,
    "args": {
      "External id": 89329,
      "cbid": 251, "correlation": 89329
    }
  },
  {
    "ph": "f", "id": 89329, "pid": 76337, "tid": -914061504, "ts": 1716454222744854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222798013, "dur": 493,
    "args": {
      "External id": 89330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89330,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89330, "pid": 5, "tid": 7, "ts": 1716454222798013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744857, "dur": 16,
    "args": {
      "External id": 89330,
      "cbid": 211, "correlation": 89330
    }
  },
  {
    "ph": "s", "id": 89330, "pid": 76337, "tid": -914061504, "ts": 1716454222744857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222798508, "dur": 66,
    "args": {
      "External id": 89338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89338, "pid": 5, "tid": 7, "ts": 1716454222798508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744932, "dur": 13,
    "args": {
      "External id": 89338,
      "cbid": 211, "correlation": 89338
    }
  },
  {
    "ph": "s", "id": 89338, "pid": 76337, "tid": -914061504, "ts": 1716454222744932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222798575, "dur": 69,
    "args": {
      "External id": 89346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89346, "pid": 5, "tid": 7, "ts": 1716454222798575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222744964, "dur": 8,
    "args": {
      "External id": 89346,
      "cbid": 211, "correlation": 89346
    }
  },
  {
    "ph": "s", "id": 89346, "pid": 76337, "tid": -914061504, "ts": 1716454222744964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745054, "dur": 1,
    "args": {
      "External id": 89362,
      "cbid": 251, "correlation": 89362
    }
  },
  {
    "ph": "f", "id": 89362, "pid": 76337, "tid": -914061504, "ts": 1716454222745054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222798647, "dur": 1,
    "args": {
      "External id": 89364,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 89364,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 89364, "pid": 5, "tid": 7, "ts": 1716454222798647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745059, "dur": 12,
    "args": {
      "External id": 89364,
      "cbid": 51, "correlation": 89364
    }
  },
  {
    "ph": "s", "id": 89364, "pid": 76337, "tid": -914061504, "ts": 1716454222745059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222798650, "dur": 266,
    "args": {
      "External id": 89365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89365,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89365, "pid": 5, "tid": 7, "ts": 1716454222798650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745073, "dur": 11,
    "args": {
      "External id": 89365,
      "cbid": 211, "correlation": 89365
    }
  },
  {
    "ph": "s", "id": 89365, "pid": 76337, "tid": -914061504, "ts": 1716454222745073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222798918, "dur": 13,
    "args": {
      "External id": 89373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89373,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89373, "pid": 5, "tid": 7, "ts": 1716454222798918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745117, "dur": 10,
    "args": {
      "External id": 89373,
      "cbid": 211, "correlation": 89373
    }
  },
  {
    "ph": "s", "id": 89373, "pid": 76337, "tid": -914061504, "ts": 1716454222745117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222798933, "dur": 38,
    "args": {
      "External id": 89384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89384, "pid": 5, "tid": 7, "ts": 1716454222798933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745187, "dur": 13,
    "args": {
      "External id": 89384,
      "cbid": 211, "correlation": 89384
    }
  },
  {
    "ph": "s", "id": 89384, "pid": 76337, "tid": -914061504, "ts": 1716454222745187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745254, "dur": 0,
    "args": {
      "External id": 89396,
      "cbid": 317, "correlation": 89396
    }
  },
  {
    "ph": "f", "id": 89396, "pid": 76337, "tid": -914061504, "ts": 1716454222745254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745255, "dur": 0,
    "args": {
      "External id": 89397,
      "cbid": 203, "correlation": 89397
    }
  },
  {
    "ph": "f", "id": 89397, "pid": 76337, "tid": -914061504, "ts": 1716454222745255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745256, "dur": 0,
    "args": {
      "External id": 89398,
      "cbid": 205, "correlation": 89398
    }
  },
  {
    "ph": "f", "id": 89398, "pid": 76337, "tid": -914061504, "ts": 1716454222745256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222798972, "dur": 12,
    "args": {
      "External id": 89402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89402,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89402, "pid": 5, "tid": 7, "ts": 1716454222798972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745275, "dur": 12,
    "args": {
      "External id": 89402,
      "cbid": 211, "correlation": 89402
    }
  },
  {
    "ph": "s", "id": 89402, "pid": 76337, "tid": -914061504, "ts": 1716454222745275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222798986, "dur": 4,
    "args": {
      "External id": 89404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89404,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89404, "pid": 5, "tid": 7, "ts": 1716454222798986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745292, "dur": 6,
    "args": {
      "External id": 89404,
      "cbid": 211, "correlation": 89404
    }
  },
  {
    "ph": "s", "id": 89404, "pid": 76337, "tid": -914061504, "ts": 1716454222745292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745301, "dur": 0,
    "args": {
      "External id": 89405,
      "cbid": 51, "correlation": 89405
    }
  },
  {
    "ph": "s", "id": 89405, "pid": 76337, "tid": -914061504, "ts": 1716454222745301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222798991, "dur": 96,
    "args": {
      "External id": 89406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89406,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 89406, "pid": 5, "tid": 7, "ts": 1716454222798991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745302, "dur": 6,
    "args": {
      "External id": 89406,
      "cbid": 211, "correlation": 89406
    }
  },
  {
    "ph": "s", "id": 89406, "pid": 76337, "tid": -914061504, "ts": 1716454222745302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222799089, "dur": 16,
    "args": {
      "External id": 89411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89411, "pid": 5, "tid": 7, "ts": 1716454222799089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745331, "dur": 9,
    "args": {
      "External id": 89411,
      "cbid": 211, "correlation": 89411
    }
  },
  {
    "ph": "s", "id": 89411, "pid": 76337, "tid": -914061504, "ts": 1716454222745331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222799106, "dur": 11,
    "args": {
      "External id": 89419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89419,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89419, "pid": 5, "tid": 7, "ts": 1716454222799106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745363, "dur": 8,
    "args": {
      "External id": 89419,
      "cbid": 211, "correlation": 89419
    }
  },
  {
    "ph": "s", "id": 89419, "pid": 76337, "tid": -914061504, "ts": 1716454222745363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222799118, "dur": 18,
    "args": {
      "External id": 89439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89439,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 89439, "pid": 5, "tid": 7, "ts": 1716454222799118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745440, "dur": 12,
    "args": {
      "External id": 89439,
      "cbid": 211, "correlation": 89439
    }
  },
  {
    "ph": "s", "id": 89439, "pid": 76337, "tid": -914061504, "ts": 1716454222745440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222799138, "dur": 5,
    "args": {
      "External id": 89451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89451,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 89451, "pid": 5, "tid": 7, "ts": 1716454222799138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745462, "dur": 6,
    "args": {
      "External id": 89451,
      "cbid": 211, "correlation": 89451
    }
  },
  {
    "ph": "s", "id": 89451, "pid": 76337, "tid": -914061504, "ts": 1716454222745462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222799144, "dur": 18,
    "args": {
      "External id": 89454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89454, "pid": 5, "tid": 7, "ts": 1716454222799144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745480, "dur": 6,
    "args": {
      "External id": 89454,
      "cbid": 211, "correlation": 89454
    }
  },
  {
    "ph": "s", "id": 89454, "pid": 76337, "tid": -914061504, "ts": 1716454222745480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222799163, "dur": 12,
    "args": {
      "External id": 89463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89463,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89463, "pid": 5, "tid": 7, "ts": 1716454222799163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745522, "dur": 10,
    "args": {
      "External id": 89463,
      "cbid": 211, "correlation": 89463
    }
  },
  {
    "ph": "s", "id": 89463, "pid": 76337, "tid": -914061504, "ts": 1716454222745522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745573, "dur": 0,
    "args": {
      "External id": 89473,
      "cbid": 317, "correlation": 89473
    }
  },
  {
    "ph": "f", "id": 89473, "pid": 76337, "tid": -914061504, "ts": 1716454222745573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745574, "dur": 0,
    "args": {
      "External id": 89474,
      "cbid": 203, "correlation": 89474
    }
  },
  {
    "ph": "f", "id": 89474, "pid": 76337, "tid": -914061504, "ts": 1716454222745574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745575, "dur": 0,
    "args": {
      "External id": 89475,
      "cbid": 205, "correlation": 89475
    }
  },
  {
    "ph": "f", "id": 89475, "pid": 76337, "tid": -914061504, "ts": 1716454222745575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222799176, "dur": 11,
    "args": {
      "External id": 89479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89479,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89479, "pid": 5, "tid": 7, "ts": 1716454222799176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745588, "dur": 11,
    "args": {
      "External id": 89479,
      "cbid": 211, "correlation": 89479
    }
  },
  {
    "ph": "s", "id": 89479, "pid": 76337, "tid": -914061504, "ts": 1716454222745588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222799188, "dur": 160,
    "args": {
      "External id": 89481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89481,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89481, "pid": 5, "tid": 7, "ts": 1716454222799188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745602, "dur": 5,
    "args": {
      "External id": 89481,
      "cbid": 211, "correlation": 89481
    }
  },
  {
    "ph": "s", "id": 89481, "pid": 76337, "tid": -914061504, "ts": 1716454222745602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222799351, "dur": 1,
    "args": {
      "External id": 89483,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 89483,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 89483, "pid": 5, "tid": 7, "ts": 1716454222799351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745614, "dur": 7,
    "args": {
      "External id": 89483,
      "cbid": 51, "correlation": 89483
    }
  },
  {
    "ph": "s", "id": 89483, "pid": 76337, "tid": -914061504, "ts": 1716454222745614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222799354, "dur": 656,
    "args": {
      "External id": 89484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89484,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89484, "pid": 5, "tid": 7, "ts": 1716454222799354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745623, "dur": 7,
    "args": {
      "External id": 89484,
      "cbid": 211, "correlation": 89484
    }
  },
  {
    "ph": "s", "id": 89484, "pid": 76337, "tid": -914061504, "ts": 1716454222745623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222800011, "dur": 13,
    "args": {
      "External id": 89486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89486,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89486, "pid": 5, "tid": 7, "ts": 1716454222800011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745634, "dur": 5,
    "args": {
      "External id": 89486,
      "cbid": 211, "correlation": 89486
    }
  },
  {
    "ph": "s", "id": 89486, "pid": 76337, "tid": -914061504, "ts": 1716454222745634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222800026, "dur": 14,
    "args": {
      "External id": 89492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89492,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89492, "pid": 5, "tid": 7, "ts": 1716454222800026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745663, "dur": 10,
    "args": {
      "External id": 89492,
      "cbid": 211, "correlation": 89492
    }
  },
  {
    "ph": "s", "id": 89492, "pid": 76337, "tid": -914061504, "ts": 1716454222745663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222800041, "dur": 3,
    "args": {
      "External id": 89500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89500,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 89500, "pid": 5, "tid": 7, "ts": 1716454222800041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745707, "dur": 9,
    "args": {
      "External id": 89500,
      "cbid": 211, "correlation": 89500
    }
  },
  {
    "ph": "s", "id": 89500, "pid": 76337, "tid": -914061504, "ts": 1716454222745707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745776, "dur": 1,
    "args": {
      "External id": 89516,
      "cbid": 251, "correlation": 89516
    }
  },
  {
    "ph": "f", "id": 89516, "pid": 76337, "tid": -914061504, "ts": 1716454222745776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745781, "dur": 0,
    "args": {
      "External id": 89518,
      "cbid": 251, "correlation": 89518
    }
  },
  {
    "ph": "f", "id": 89518, "pid": 76337, "tid": -914061504, "ts": 1716454222745781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222800046, "dur": 13,
    "args": {
      "External id": 89519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89519,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89519, "pid": 5, "tid": 7, "ts": 1716454222800046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745783, "dur": 12,
    "args": {
      "External id": 89519,
      "cbid": 211, "correlation": 89519
    }
  },
  {
    "ph": "s", "id": 89519, "pid": 76337, "tid": -914061504, "ts": 1716454222745783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222800060, "dur": 5,
    "args": {
      "External id": 89521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89521,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89521, "pid": 5, "tid": 7, "ts": 1716454222800060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745797, "dur": 6,
    "args": {
      "External id": 89521,
      "cbid": 211, "correlation": 89521
    }
  },
  {
    "ph": "s", "id": 89521, "pid": 76337, "tid": -914061504, "ts": 1716454222745797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222800067, "dur": 17,
    "args": {
      "External id": 89531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89531, "pid": 5, "tid": 7, "ts": 1716454222800067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745857, "dur": 12,
    "args": {
      "External id": 89531,
      "cbid": 211, "correlation": 89531
    }
  },
  {
    "ph": "s", "id": 89531, "pid": 76337, "tid": -914061504, "ts": 1716454222745857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222800085, "dur": 17,
    "args": {
      "External id": 89551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89551,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 89551, "pid": 5, "tid": 7, "ts": 1716454222800085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745923, "dur": 11,
    "args": {
      "External id": 89551,
      "cbid": 211, "correlation": 89551
    }
  },
  {
    "ph": "s", "id": 89551, "pid": 76337, "tid": -914061504, "ts": 1716454222745923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222800104, "dur": 4,
    "args": {
      "External id": 89563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89563,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 89563, "pid": 5, "tid": 7, "ts": 1716454222800104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745944, "dur": 6,
    "args": {
      "External id": 89563,
      "cbid": 211, "correlation": 89563
    }
  },
  {
    "ph": "s", "id": 89563, "pid": 76337, "tid": -914061504, "ts": 1716454222745944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222800109, "dur": 16,
    "args": {
      "External id": 89566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89566, "pid": 5, "tid": 7, "ts": 1716454222800109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222745964, "dur": 6,
    "args": {
      "External id": 89566,
      "cbid": 211, "correlation": 89566
    }
  },
  {
    "ph": "s", "id": 89566, "pid": 76337, "tid": -914061504, "ts": 1716454222745964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222800127, "dur": 11,
    "args": {
      "External id": 89575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89575,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89575, "pid": 5, "tid": 7, "ts": 1716454222800127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746013, "dur": 11,
    "args": {
      "External id": 89575,
      "cbid": 211, "correlation": 89575
    }
  },
  {
    "ph": "s", "id": 89575, "pid": 76337, "tid": -914061504, "ts": 1716454222746013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746078, "dur": 0,
    "args": {
      "External id": 89585,
      "cbid": 317, "correlation": 89585
    }
  },
  {
    "ph": "f", "id": 89585, "pid": 76337, "tid": -914061504, "ts": 1716454222746078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746079, "dur": 0,
    "args": {
      "External id": 89586,
      "cbid": 203, "correlation": 89586
    }
  },
  {
    "ph": "f", "id": 89586, "pid": 76337, "tid": -914061504, "ts": 1716454222746079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746080, "dur": 0,
    "args": {
      "External id": 89587,
      "cbid": 205, "correlation": 89587
    }
  },
  {
    "ph": "f", "id": 89587, "pid": 76337, "tid": -914061504, "ts": 1716454222746080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222800139, "dur": 10,
    "args": {
      "External id": 89591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89591, "pid": 5, "tid": 7, "ts": 1716454222800139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746094, "dur": 12,
    "args": {
      "External id": 89591,
      "cbid": 211, "correlation": 89591
    }
  },
  {
    "ph": "s", "id": 89591, "pid": 76337, "tid": -914061504, "ts": 1716454222746094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222800150, "dur": 159,
    "args": {
      "External id": 89593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89593,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89593, "pid": 5, "tid": 7, "ts": 1716454222800150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746108, "dur": 5,
    "args": {
      "External id": 89593,
      "cbid": 211, "correlation": 89593
    }
  },
  {
    "ph": "s", "id": 89593, "pid": 76337, "tid": -914061504, "ts": 1716454222746108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222800311, "dur": 1,
    "args": {
      "External id": 89595,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 89595,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 89595, "pid": 5, "tid": 7, "ts": 1716454222800311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746119, "dur": 6,
    "args": {
      "External id": 89595,
      "cbid": 51, "correlation": 89595
    }
  },
  {
    "ph": "s", "id": 89595, "pid": 76337, "tid": -914061504, "ts": 1716454222746119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222800315, "dur": 637,
    "args": {
      "External id": 89596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89596,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89596, "pid": 5, "tid": 7, "ts": 1716454222800315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746127, "dur": 7,
    "args": {
      "External id": 89596,
      "cbid": 211, "correlation": 89596
    }
  },
  {
    "ph": "s", "id": 89596, "pid": 76337, "tid": -914061504, "ts": 1716454222746127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222800954, "dur": 12,
    "args": {
      "External id": 89598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89598, "pid": 5, "tid": 7, "ts": 1716454222800954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746139, "dur": 5,
    "args": {
      "External id": 89598,
      "cbid": 211, "correlation": 89598
    }
  },
  {
    "ph": "s", "id": 89598, "pid": 76337, "tid": -914061504, "ts": 1716454222746139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222800967, "dur": 14,
    "args": {
      "External id": 89604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89604, "pid": 5, "tid": 7, "ts": 1716454222800967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746168, "dur": 8,
    "args": {
      "External id": 89604,
      "cbid": 211, "correlation": 89604
    }
  },
  {
    "ph": "s", "id": 89604, "pid": 76337, "tid": -914061504, "ts": 1716454222746168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222800983, "dur": 12,
    "args": {
      "External id": 89612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89612,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89612, "pid": 5, "tid": 7, "ts": 1716454222800983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746201, "dur": 8,
    "args": {
      "External id": 89612,
      "cbid": 211, "correlation": 89612
    }
  },
  {
    "ph": "s", "id": 89612, "pid": 76337, "tid": -914061504, "ts": 1716454222746201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222800996, "dur": 10,
    "args": {
      "External id": 89620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89620,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89620, "pid": 5, "tid": 7, "ts": 1716454222800996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746230, "dur": 9,
    "args": {
      "External id": 89620,
      "cbid": 211, "correlation": 89620
    }
  },
  {
    "ph": "s", "id": 89620, "pid": 76337, "tid": -914061504, "ts": 1716454222746230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222801008, "dur": 18,
    "args": {
      "External id": 89640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89640,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 89640, "pid": 5, "tid": 7, "ts": 1716454222801008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746312, "dur": 12,
    "args": {
      "External id": 89640,
      "cbid": 211, "correlation": 89640
    }
  },
  {
    "ph": "s", "id": 89640, "pid": 76337, "tid": -914061504, "ts": 1716454222746312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222801027, "dur": 4,
    "args": {
      "External id": 89652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89652,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 89652, "pid": 5, "tid": 7, "ts": 1716454222801027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746334, "dur": 6,
    "args": {
      "External id": 89652,
      "cbid": 211, "correlation": 89652
    }
  },
  {
    "ph": "s", "id": 89652, "pid": 76337, "tid": -914061504, "ts": 1716454222746334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222801032, "dur": 16,
    "args": {
      "External id": 89655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89655, "pid": 5, "tid": 7, "ts": 1716454222801032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746352, "dur": 7,
    "args": {
      "External id": 89655,
      "cbid": 211, "correlation": 89655
    }
  },
  {
    "ph": "s", "id": 89655, "pid": 76337, "tid": -914061504, "ts": 1716454222746352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746409, "dur": 0,
    "args": {
      "External id": 89666,
      "cbid": 317, "correlation": 89666
    }
  },
  {
    "ph": "f", "id": 89666, "pid": 76337, "tid": -914061504, "ts": 1716454222746409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746410, "dur": 0,
    "args": {
      "External id": 89667,
      "cbid": 203, "correlation": 89667
    }
  },
  {
    "ph": "f", "id": 89667, "pid": 76337, "tid": -914061504, "ts": 1716454222746410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746411, "dur": 0,
    "args": {
      "External id": 89668,
      "cbid": 205, "correlation": 89668
    }
  },
  {
    "ph": "f", "id": 89668, "pid": 76337, "tid": -914061504, "ts": 1716454222746411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222801049, "dur": 11,
    "args": {
      "External id": 89672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89672, "pid": 5, "tid": 7, "ts": 1716454222801049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746425, "dur": 11,
    "args": {
      "External id": 89672,
      "cbid": 211, "correlation": 89672
    }
  },
  {
    "ph": "s", "id": 89672, "pid": 76337, "tid": -914061504, "ts": 1716454222746425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222801062, "dur": 4,
    "args": {
      "External id": 89674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89674, "pid": 5, "tid": 7, "ts": 1716454222801062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746441, "dur": 7,
    "args": {
      "External id": 89674,
      "cbid": 211, "correlation": 89674
    }
  },
  {
    "ph": "s", "id": 89674, "pid": 76337, "tid": -914061504, "ts": 1716454222746441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746450, "dur": 0,
    "args": {
      "External id": 89675,
      "cbid": 51, "correlation": 89675
    }
  },
  {
    "ph": "s", "id": 89675, "pid": 76337, "tid": -914061504, "ts": 1716454222746450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222801067, "dur": 93,
    "args": {
      "External id": 89676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89676,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 89676, "pid": 5, "tid": 7, "ts": 1716454222801067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746451, "dur": 5,
    "args": {
      "External id": 89676,
      "cbid": 211, "correlation": 89676
    }
  },
  {
    "ph": "s", "id": 89676, "pid": 76337, "tid": -914061504, "ts": 1716454222746451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222801161, "dur": 16,
    "args": {
      "External id": 89681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89681,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89681, "pid": 5, "tid": 7, "ts": 1716454222801161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746479, "dur": 8,
    "args": {
      "External id": 89681,
      "cbid": 211, "correlation": 89681
    }
  },
  {
    "ph": "s", "id": 89681, "pid": 76337, "tid": -914061504, "ts": 1716454222746479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222801178, "dur": 83,
    "args": {
      "External id": 89690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89690,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89690, "pid": 5, "tid": 7, "ts": 1716454222801178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746565, "dur": 16,
    "args": {
      "External id": 89690,
      "cbid": 211, "correlation": 89690
    }
  },
  {
    "ph": "s", "id": 89690, "pid": 76337, "tid": -914061504, "ts": 1716454222746565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222801262, "dur": 29,
    "args": {
      "External id": 89712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89712,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89712, "pid": 5, "tid": 7, "ts": 1716454222801262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746624, "dur": 10,
    "args": {
      "External id": 89712,
      "cbid": 211, "correlation": 89712
    }
  },
  {
    "ph": "s", "id": 89712, "pid": 76337, "tid": -914061504, "ts": 1716454222746624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746717, "dur": 2,
    "args": {
      "External id": 89723,
      "cbid": 251, "correlation": 89723
    }
  },
  {
    "ph": "f", "id": 89723, "pid": 76337, "tid": -914061504, "ts": 1716454222746717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222801293, "dur": 160,
    "args": {
      "External id": 89724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89724,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89724, "pid": 5, "tid": 7, "ts": 1716454222801293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746722, "dur": 13,
    "args": {
      "External id": 89724,
      "cbid": 211, "correlation": 89724
    }
  },
  {
    "ph": "s", "id": 89724, "pid": 76337, "tid": -914061504, "ts": 1716454222746722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746793, "dur": 1,
    "args": {
      "External id": 89735,
      "cbid": 251, "correlation": 89735
    }
  },
  {
    "ph": "f", "id": 89735, "pid": 76337, "tid": -914061504, "ts": 1716454222746793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222801454, "dur": 156,
    "args": {
      "External id": 89736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89736,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89736, "pid": 5, "tid": 7, "ts": 1716454222801454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746797, "dur": 11,
    "args": {
      "External id": 89736,
      "cbid": 211, "correlation": 89736
    }
  },
  {
    "ph": "s", "id": 89736, "pid": 76337, "tid": -914061504, "ts": 1716454222746797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746862, "dur": 1,
    "args": {
      "External id": 89747,
      "cbid": 251, "correlation": 89747
    }
  },
  {
    "ph": "f", "id": 89747, "pid": 76337, "tid": -914061504, "ts": 1716454222746862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222801612, "dur": 156,
    "args": {
      "External id": 89748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89748,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89748, "pid": 5, "tid": 7, "ts": 1716454222801612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746866, "dur": 12,
    "args": {
      "External id": 89748,
      "cbid": 211, "correlation": 89748
    }
  },
  {
    "ph": "s", "id": 89748, "pid": 76337, "tid": -914061504, "ts": 1716454222746866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222801769, "dur": 332,
    "args": {
      "External id": 89773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89773,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89773, "pid": 5, "tid": 7, "ts": 1716454222801769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222746956, "dur": 14,
    "args": {
      "External id": 89773,
      "cbid": 211, "correlation": 89773
    }
  },
  {
    "ph": "s", "id": 89773, "pid": 76337, "tid": -914061504, "ts": 1716454222746956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747073, "dur": 1,
    "args": {
      "External id": 89791,
      "cbid": 251, "correlation": 89791
    }
  },
  {
    "ph": "f", "id": 89791, "pid": 76337, "tid": -914061504, "ts": 1716454222747073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222802103, "dur": 164,
    "args": {
      "External id": 89793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89793,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89793, "pid": 5, "tid": 7, "ts": 1716454222802103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747079, "dur": 14,
    "args": {
      "External id": 89793,
      "cbid": 211, "correlation": 89793
    }
  },
  {
    "ph": "s", "id": 89793, "pid": 76337, "tid": -914061504, "ts": 1716454222747079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222802268, "dur": 20,
    "args": {
      "External id": 89801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89801,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89801, "pid": 5, "tid": 7, "ts": 1716454222802268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747151, "dur": 13,
    "args": {
      "External id": 89801,
      "cbid": 211, "correlation": 89801
    }
  },
  {
    "ph": "s", "id": 89801, "pid": 76337, "tid": -914061504, "ts": 1716454222747151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222802289, "dur": 27,
    "args": {
      "External id": 89809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89809,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89809, "pid": 5, "tid": 7, "ts": 1716454222802289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747192, "dur": 8,
    "args": {
      "External id": 89809,
      "cbid": 211, "correlation": 89809
    }
  },
  {
    "ph": "s", "id": 89809, "pid": 76337, "tid": -914061504, "ts": 1716454222747192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222802318, "dur": 18,
    "args": {
      "External id": 89820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89820, "pid": 5, "tid": 7, "ts": 1716454222802318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747265, "dur": 14,
    "args": {
      "External id": 89820,
      "cbid": 211, "correlation": 89820
    }
  },
  {
    "ph": "s", "id": 89820, "pid": 76337, "tid": -914061504, "ts": 1716454222747265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222802337, "dur": 16,
    "args": {
      "External id": 89842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89842,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89842, "pid": 5, "tid": 7, "ts": 1716454222802337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747298, "dur": 7,
    "args": {
      "External id": 89842,
      "cbid": 211, "correlation": 89842
    }
  },
  {
    "ph": "s", "id": 89842, "pid": 76337, "tid": -914061504, "ts": 1716454222747298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747384, "dur": 2,
    "args": {
      "External id": 89853,
      "cbid": 251, "correlation": 89853
    }
  },
  {
    "ph": "f", "id": 89853, "pid": 76337, "tid": -914061504, "ts": 1716454222747384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222802355, "dur": 87,
    "args": {
      "External id": 89854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89854,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89854, "pid": 5, "tid": 7, "ts": 1716454222802355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747390, "dur": 14,
    "args": {
      "External id": 89854,
      "cbid": 211, "correlation": 89854
    }
  },
  {
    "ph": "s", "id": 89854, "pid": 76337, "tid": -914061504, "ts": 1716454222747390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747461, "dur": 1,
    "args": {
      "External id": 89865,
      "cbid": 251, "correlation": 89865
    }
  },
  {
    "ph": "f", "id": 89865, "pid": 76337, "tid": -914061504, "ts": 1716454222747461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747465, "dur": 0,
    "args": {
      "External id": 89866,
      "cbid": 251, "correlation": 89866
    }
  },
  {
    "ph": "f", "id": 89866, "pid": 76337, "tid": -914061504, "ts": 1716454222747465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222802443, "dur": 11,
    "args": {
      "External id": 89867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89867,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89867, "pid": 5, "tid": 7, "ts": 1716454222802443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747466, "dur": 12,
    "args": {
      "External id": 89867,
      "cbid": 211, "correlation": 89867
    }
  },
  {
    "ph": "s", "id": 89867, "pid": 76337, "tid": -914061504, "ts": 1716454222747466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222802456, "dur": 5,
    "args": {
      "External id": 89869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89869,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89869, "pid": 5, "tid": 7, "ts": 1716454222802456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747480, "dur": 7,
    "args": {
      "External id": 89869,
      "cbid": 211, "correlation": 89869
    }
  },
  {
    "ph": "s", "id": 89869, "pid": 76337, "tid": -914061504, "ts": 1716454222747480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747539, "dur": 1,
    "args": {
      "External id": 89880,
      "cbid": 251, "correlation": 89880
    }
  },
  {
    "ph": "f", "id": 89880, "pid": 76337, "tid": -914061504, "ts": 1716454222747539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747542, "dur": 0,
    "args": {
      "External id": 89881,
      "cbid": 251, "correlation": 89881
    }
  },
  {
    "ph": "f", "id": 89881, "pid": 76337, "tid": -914061504, "ts": 1716454222747542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222802463, "dur": 8,
    "args": {
      "External id": 89882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89882,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89882, "pid": 5, "tid": 7, "ts": 1716454222802463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747544, "dur": 12,
    "args": {
      "External id": 89882,
      "cbid": 211, "correlation": 89882
    }
  },
  {
    "ph": "s", "id": 89882, "pid": 76337, "tid": -914061504, "ts": 1716454222747544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222802472, "dur": 4,
    "args": {
      "External id": 89884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89884,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89884, "pid": 5, "tid": 7, "ts": 1716454222802472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747557, "dur": 6,
    "args": {
      "External id": 89884,
      "cbid": 211, "correlation": 89884
    }
  },
  {
    "ph": "s", "id": 89884, "pid": 76337, "tid": -914061504, "ts": 1716454222747557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222802477, "dur": 54,
    "args": {
      "External id": 89909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89909,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89909, "pid": 5, "tid": 7, "ts": 1716454222802477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747636, "dur": 13,
    "args": {
      "External id": 89909,
      "cbid": 211, "correlation": 89909
    }
  },
  {
    "ph": "s", "id": 89909, "pid": 76337, "tid": -914061504, "ts": 1716454222747636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747739, "dur": 2,
    "args": {
      "External id": 89927,
      "cbid": 251, "correlation": 89927
    }
  },
  {
    "ph": "f", "id": 89927, "pid": 76337, "tid": -914061504, "ts": 1716454222747739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222802532, "dur": 90,
    "args": {
      "External id": 89929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89929,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 89929, "pid": 5, "tid": 7, "ts": 1716454222802532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747746, "dur": 14,
    "args": {
      "External id": 89929,
      "cbid": 211, "correlation": 89929
    }
  },
  {
    "ph": "s", "id": 89929, "pid": 76337, "tid": -914061504, "ts": 1716454222747746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222802623, "dur": 10,
    "args": {
      "External id": 89937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89937,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89937, "pid": 5, "tid": 7, "ts": 1716454222802623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747816, "dur": 13,
    "args": {
      "External id": 89937,
      "cbid": 211, "correlation": 89937
    }
  },
  {
    "ph": "s", "id": 89937, "pid": 76337, "tid": -914061504, "ts": 1716454222747816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222802634, "dur": 22,
    "args": {
      "External id": 89945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89945, "pid": 5, "tid": 7, "ts": 1716454222802634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747859, "dur": 10,
    "args": {
      "External id": 89945,
      "cbid": 211, "correlation": 89945
    }
  },
  {
    "ph": "s", "id": 89945, "pid": 76337, "tid": -914061504, "ts": 1716454222747859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222802657, "dur": 17,
    "args": {
      "External id": 89967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89967,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89967, "pid": 5, "tid": 7, "ts": 1716454222802657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222747912, "dur": 10,
    "args": {
      "External id": 89967,
      "cbid": 211, "correlation": 89967
    }
  },
  {
    "ph": "s", "id": 89967, "pid": 76337, "tid": -914061504, "ts": 1716454222747912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748007, "dur": 2,
    "args": {
      "External id": 89983,
      "cbid": 251, "correlation": 89983
    }
  },
  {
    "ph": "f", "id": 89983, "pid": 76337, "tid": -914061504, "ts": 1716454222748007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748013, "dur": 0,
    "args": {
      "External id": 89985,
      "cbid": 251, "correlation": 89985
    }
  },
  {
    "ph": "f", "id": 89985, "pid": 76337, "tid": -914061504, "ts": 1716454222748013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222802675, "dur": 490,
    "args": {
      "External id": 89986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89986,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 89986, "pid": 5, "tid": 7, "ts": 1716454222802675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748015, "dur": 14,
    "args": {
      "External id": 89986,
      "cbid": 211, "correlation": 89986
    }
  },
  {
    "ph": "s", "id": 89986, "pid": 76337, "tid": -914061504, "ts": 1716454222748015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222803167, "dur": 65,
    "args": {
      "External id": 89994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 89994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 89994, "pid": 5, "tid": 7, "ts": 1716454222803167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748087, "dur": 13,
    "args": {
      "External id": 89994,
      "cbid": 211, "correlation": 89994
    }
  },
  {
    "ph": "s", "id": 89994, "pid": 76337, "tid": -914061504, "ts": 1716454222748087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222803233, "dur": 67,
    "args": {
      "External id": 90002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90002, "pid": 5, "tid": 7, "ts": 1716454222803233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748120, "dur": 9,
    "args": {
      "External id": 90002,
      "cbid": 211, "correlation": 90002
    }
  },
  {
    "ph": "s", "id": 90002, "pid": 76337, "tid": -914061504, "ts": 1716454222748120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748203, "dur": 1,
    "args": {
      "External id": 90018,
      "cbid": 251, "correlation": 90018
    }
  },
  {
    "ph": "f", "id": 90018, "pid": 76337, "tid": -914061504, "ts": 1716454222748203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222803302, "dur": 1,
    "args": {
      "External id": 90020,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90020,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 90020, "pid": 5, "tid": 7, "ts": 1716454222803302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748208, "dur": 11,
    "args": {
      "External id": 90020,
      "cbid": 51, "correlation": 90020
    }
  },
  {
    "ph": "s", "id": 90020, "pid": 76337, "tid": -914061504, "ts": 1716454222748208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222803305, "dur": 271,
    "args": {
      "External id": 90021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90021,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 90021, "pid": 5, "tid": 7, "ts": 1716454222803305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748220, "dur": 11,
    "args": {
      "External id": 90021,
      "cbid": 211, "correlation": 90021
    }
  },
  {
    "ph": "s", "id": 90021, "pid": 76337, "tid": -914061504, "ts": 1716454222748220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222803578, "dur": 13,
    "args": {
      "External id": 90029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90029,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90029, "pid": 5, "tid": 7, "ts": 1716454222803578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748264, "dur": 10,
    "args": {
      "External id": 90029,
      "cbid": 211, "correlation": 90029
    }
  },
  {
    "ph": "s", "id": 90029, "pid": 76337, "tid": -914061504, "ts": 1716454222748264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222803593, "dur": 37,
    "args": {
      "External id": 90040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90040, "pid": 5, "tid": 7, "ts": 1716454222803593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748334, "dur": 13,
    "args": {
      "External id": 90040,
      "cbid": 211, "correlation": 90040
    }
  },
  {
    "ph": "s", "id": 90040, "pid": 76337, "tid": -914061504, "ts": 1716454222748334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748401, "dur": 0,
    "args": {
      "External id": 90052,
      "cbid": 317, "correlation": 90052
    }
  },
  {
    "ph": "f", "id": 90052, "pid": 76337, "tid": -914061504, "ts": 1716454222748401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748402, "dur": 0,
    "args": {
      "External id": 90053,
      "cbid": 203, "correlation": 90053
    }
  },
  {
    "ph": "f", "id": 90053, "pid": 76337, "tid": -914061504, "ts": 1716454222748402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748403, "dur": 0,
    "args": {
      "External id": 90054,
      "cbid": 205, "correlation": 90054
    }
  },
  {
    "ph": "f", "id": 90054, "pid": 76337, "tid": -914061504, "ts": 1716454222748403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222803631, "dur": 12,
    "args": {
      "External id": 90058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90058,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90058, "pid": 5, "tid": 7, "ts": 1716454222803631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748420, "dur": 12,
    "args": {
      "External id": 90058,
      "cbid": 211, "correlation": 90058
    }
  },
  {
    "ph": "s", "id": 90058, "pid": 76337, "tid": -914061504, "ts": 1716454222748420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222803645, "dur": 4,
    "args": {
      "External id": 90060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 90060, "pid": 5, "tid": 7, "ts": 1716454222803645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748437, "dur": 6,
    "args": {
      "External id": 90060,
      "cbid": 211, "correlation": 90060
    }
  },
  {
    "ph": "s", "id": 90060, "pid": 76337, "tid": -914061504, "ts": 1716454222748437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748446, "dur": 0,
    "args": {
      "External id": 90061,
      "cbid": 51, "correlation": 90061
    }
  },
  {
    "ph": "s", "id": 90061, "pid": 76337, "tid": -914061504, "ts": 1716454222748446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222803650, "dur": 96,
    "args": {
      "External id": 90062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90062,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 90062, "pid": 5, "tid": 7, "ts": 1716454222803650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748447, "dur": 5,
    "args": {
      "External id": 90062,
      "cbid": 211, "correlation": 90062
    }
  },
  {
    "ph": "s", "id": 90062, "pid": 76337, "tid": -914061504, "ts": 1716454222748447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222803747, "dur": 16,
    "args": {
      "External id": 90067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90067, "pid": 5, "tid": 7, "ts": 1716454222803747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748474, "dur": 8,
    "args": {
      "External id": 90067,
      "cbid": 211, "correlation": 90067
    }
  },
  {
    "ph": "s", "id": 90067, "pid": 76337, "tid": -914061504, "ts": 1716454222748474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222803765, "dur": 11,
    "args": {
      "External id": 90075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90075,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90075, "pid": 5, "tid": 7, "ts": 1716454222803765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748506, "dur": 8,
    "args": {
      "External id": 90075,
      "cbid": 211, "correlation": 90075
    }
  },
  {
    "ph": "s", "id": 90075, "pid": 76337, "tid": -914061504, "ts": 1716454222748506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748579, "dur": 0,
    "args": {
      "External id": 90085,
      "cbid": 317, "correlation": 90085
    }
  },
  {
    "ph": "f", "id": 90085, "pid": 76337, "tid": -914061504, "ts": 1716454222748579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748580, "dur": 0,
    "args": {
      "External id": 90086,
      "cbid": 203, "correlation": 90086
    }
  },
  {
    "ph": "f", "id": 90086, "pid": 76337, "tid": -914061504, "ts": 1716454222748580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748581, "dur": 0,
    "args": {
      "External id": 90087,
      "cbid": 205, "correlation": 90087
    }
  },
  {
    "ph": "f", "id": 90087, "pid": 76337, "tid": -914061504, "ts": 1716454222748581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222803777, "dur": 12,
    "args": {
      "External id": 90091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90091,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90091, "pid": 5, "tid": 7, "ts": 1716454222803777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748596, "dur": 12,
    "args": {
      "External id": 90091,
      "cbid": 211, "correlation": 90091
    }
  },
  {
    "ph": "s", "id": 90091, "pid": 76337, "tid": -914061504, "ts": 1716454222748596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222803790, "dur": 159,
    "args": {
      "External id": 90093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90093,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90093, "pid": 5, "tid": 7, "ts": 1716454222803790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748610, "dur": 5,
    "args": {
      "External id": 90093,
      "cbid": 211, "correlation": 90093
    }
  },
  {
    "ph": "s", "id": 90093, "pid": 76337, "tid": -914061504, "ts": 1716454222748610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222803951, "dur": 1,
    "args": {
      "External id": 90095,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90095,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 90095, "pid": 5, "tid": 7, "ts": 1716454222803951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748623, "dur": 7,
    "args": {
      "External id": 90095,
      "cbid": 51, "correlation": 90095
    }
  },
  {
    "ph": "s", "id": 90095, "pid": 76337, "tid": -914061504, "ts": 1716454222748623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222803955, "dur": 196,
    "args": {
      "External id": 90096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90096,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 90096, "pid": 5, "tid": 7, "ts": 1716454222803955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748632, "dur": 7,
    "args": {
      "External id": 90096,
      "cbid": 211, "correlation": 90096
    }
  },
  {
    "ph": "s", "id": 90096, "pid": 76337, "tid": -914061504, "ts": 1716454222748632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804153, "dur": 6,
    "args": {
      "External id": 90098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90098, "pid": 5, "tid": 7, "ts": 1716454222804153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748644, "dur": 5,
    "args": {
      "External id": 90098,
      "cbid": 211, "correlation": 90098
    }
  },
  {
    "ph": "s", "id": 90098, "pid": 76337, "tid": -914061504, "ts": 1716454222748644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222804160, "dur": 6,
    "args": {
      "External id": 90104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90104, "pid": 5, "tid": 7, "ts": 1716454222804160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748671, "dur": 8,
    "args": {
      "External id": 90104,
      "cbid": 211, "correlation": 90104
    }
  },
  {
    "ph": "s", "id": 90104, "pid": 76337, "tid": -914061504, "ts": 1716454222748671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804168, "dur": 11,
    "args": {
      "External id": 90124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90124,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90124, "pid": 5, "tid": 7, "ts": 1716454222804168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748770, "dur": 12,
    "args": {
      "External id": 90124,
      "cbid": 211, "correlation": 90124
    }
  },
  {
    "ph": "s", "id": 90124, "pid": 76337, "tid": -914061504, "ts": 1716454222748770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222804180, "dur": 4,
    "args": {
      "External id": 90136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90136,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90136, "pid": 5, "tid": 7, "ts": 1716454222804180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748793, "dur": 7,
    "args": {
      "External id": 90136,
      "cbid": 211, "correlation": 90136
    }
  },
  {
    "ph": "s", "id": 90136, "pid": 76337, "tid": -914061504, "ts": 1716454222748793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222804185, "dur": 8,
    "args": {
      "External id": 90139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90139,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90139, "pid": 5, "tid": 7, "ts": 1716454222804185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748813, "dur": 6,
    "args": {
      "External id": 90139,
      "cbid": 211, "correlation": 90139
    }
  },
  {
    "ph": "s", "id": 90139, "pid": 76337, "tid": -914061504, "ts": 1716454222748813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222804195, "dur": 5,
    "args": {
      "External id": 90148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90148,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90148, "pid": 5, "tid": 7, "ts": 1716454222804195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748855, "dur": 11,
    "args": {
      "External id": 90148,
      "cbid": 211, "correlation": 90148
    }
  },
  {
    "ph": "s", "id": 90148, "pid": 76337, "tid": -914061504, "ts": 1716454222748855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748908, "dur": 0,
    "args": {
      "External id": 90158,
      "cbid": 317, "correlation": 90158
    }
  },
  {
    "ph": "f", "id": 90158, "pid": 76337, "tid": -914061504, "ts": 1716454222748908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748909, "dur": 0,
    "args": {
      "External id": 90159,
      "cbid": 203, "correlation": 90159
    }
  },
  {
    "ph": "f", "id": 90159, "pid": 76337, "tid": -914061504, "ts": 1716454222748909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748909, "dur": 0,
    "args": {
      "External id": 90160,
      "cbid": 205, "correlation": 90160
    }
  },
  {
    "ph": "f", "id": 90160, "pid": 76337, "tid": -914061504, "ts": 1716454222748909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804201, "dur": 5,
    "args": {
      "External id": 90164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90164,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90164, "pid": 5, "tid": 7, "ts": 1716454222804201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748925, "dur": 12,
    "args": {
      "External id": 90164,
      "cbid": 211, "correlation": 90164
    }
  },
  {
    "ph": "s", "id": 90164, "pid": 76337, "tid": -914061504, "ts": 1716454222748925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804208, "dur": 159,
    "args": {
      "External id": 90166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90166,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90166, "pid": 5, "tid": 7, "ts": 1716454222804208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748940, "dur": 5,
    "args": {
      "External id": 90166,
      "cbid": 211, "correlation": 90166
    }
  },
  {
    "ph": "s", "id": 90166, "pid": 76337, "tid": -914061504, "ts": 1716454222748940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222804369, "dur": 1,
    "args": {
      "External id": 90168,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90168,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 90168, "pid": 5, "tid": 7, "ts": 1716454222804369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748951, "dur": 6,
    "args": {
      "External id": 90168,
      "cbid": 51, "correlation": 90168
    }
  },
  {
    "ph": "s", "id": 90168, "pid": 76337, "tid": -914061504, "ts": 1716454222748951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222804373, "dur": 266,
    "args": {
      "External id": 90169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90169,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90169, "pid": 5, "tid": 7, "ts": 1716454222804373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748959, "dur": 6,
    "args": {
      "External id": 90169,
      "cbid": 211, "correlation": 90169
    }
  },
  {
    "ph": "s", "id": 90169, "pid": 76337, "tid": -914061504, "ts": 1716454222748959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804640, "dur": 5,
    "args": {
      "External id": 90171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90171,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90171, "pid": 5, "tid": 7, "ts": 1716454222804640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222748970, "dur": 13,
    "args": {
      "External id": 90171,
      "cbid": 211, "correlation": 90171
    }
  },
  {
    "ph": "s", "id": 90171, "pid": 76337, "tid": -914061504, "ts": 1716454222748970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222804646, "dur": 6,
    "args": {
      "External id": 90177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90177, "pid": 5, "tid": 7, "ts": 1716454222804646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749007, "dur": 9,
    "args": {
      "External id": 90177,
      "cbid": 211, "correlation": 90177
    }
  },
  {
    "ph": "s", "id": 90177, "pid": 76337, "tid": -914061504, "ts": 1716454222749007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222804654, "dur": 3,
    "args": {
      "External id": 90185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90185,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 90185, "pid": 5, "tid": 7, "ts": 1716454222804654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749052, "dur": 9,
    "args": {
      "External id": 90185,
      "cbid": 211, "correlation": 90185
    }
  },
  {
    "ph": "s", "id": 90185, "pid": 76337, "tid": -914061504, "ts": 1716454222749052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749119, "dur": 1,
    "args": {
      "External id": 90201,
      "cbid": 251, "correlation": 90201
    }
  },
  {
    "ph": "f", "id": 90201, "pid": 76337, "tid": -914061504, "ts": 1716454222749119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749124, "dur": 0,
    "args": {
      "External id": 90203,
      "cbid": 251, "correlation": 90203
    }
  },
  {
    "ph": "f", "id": 90203, "pid": 76337, "tid": -914061504, "ts": 1716454222749124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222804658, "dur": 13,
    "args": {
      "External id": 90204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90204,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90204, "pid": 5, "tid": 7, "ts": 1716454222804658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749126, "dur": 12,
    "args": {
      "External id": 90204,
      "cbid": 211, "correlation": 90204
    }
  },
  {
    "ph": "s", "id": 90204, "pid": 76337, "tid": -914061504, "ts": 1716454222749126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222804672, "dur": 5,
    "args": {
      "External id": 90206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90206,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90206, "pid": 5, "tid": 7, "ts": 1716454222804672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749140, "dur": 6,
    "args": {
      "External id": 90206,
      "cbid": 211, "correlation": 90206
    }
  },
  {
    "ph": "s", "id": 90206, "pid": 76337, "tid": -914061504, "ts": 1716454222749140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222804679, "dur": 6,
    "args": {
      "External id": 90216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90216,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90216, "pid": 5, "tid": 7, "ts": 1716454222804679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749200, "dur": 12,
    "args": {
      "External id": 90216,
      "cbid": 211, "correlation": 90216
    }
  },
  {
    "ph": "s", "id": 90216, "pid": 76337, "tid": -914061504, "ts": 1716454222749200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804686, "dur": 10,
    "args": {
      "External id": 90236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90236,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90236, "pid": 5, "tid": 7, "ts": 1716454222804686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749267, "dur": 11,
    "args": {
      "External id": 90236,
      "cbid": 211, "correlation": 90236
    }
  },
  {
    "ph": "s", "id": 90236, "pid": 76337, "tid": -914061504, "ts": 1716454222749267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222804697, "dur": 4,
    "args": {
      "External id": 90248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90248,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90248, "pid": 5, "tid": 7, "ts": 1716454222804697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749288, "dur": 6,
    "args": {
      "External id": 90248,
      "cbid": 211, "correlation": 90248
    }
  },
  {
    "ph": "s", "id": 90248, "pid": 76337, "tid": -914061504, "ts": 1716454222749288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222804702, "dur": 7,
    "args": {
      "External id": 90251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90251,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90251, "pid": 5, "tid": 7, "ts": 1716454222804702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749307, "dur": 7,
    "args": {
      "External id": 90251,
      "cbid": 211, "correlation": 90251
    }
  },
  {
    "ph": "s", "id": 90251, "pid": 76337, "tid": -914061504, "ts": 1716454222749307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222804709, "dur": 4,
    "args": {
      "External id": 90260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90260,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90260, "pid": 5, "tid": 7, "ts": 1716454222804709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749348, "dur": 10,
    "args": {
      "External id": 90260,
      "cbid": 211, "correlation": 90260
    }
  },
  {
    "ph": "s", "id": 90260, "pid": 76337, "tid": -914061504, "ts": 1716454222749348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749412, "dur": 0,
    "args": {
      "External id": 90270,
      "cbid": 317, "correlation": 90270
    }
  },
  {
    "ph": "f", "id": 90270, "pid": 76337, "tid": -914061504, "ts": 1716454222749412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749413, "dur": 0,
    "args": {
      "External id": 90271,
      "cbid": 203, "correlation": 90271
    }
  },
  {
    "ph": "f", "id": 90271, "pid": 76337, "tid": -914061504, "ts": 1716454222749413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749414, "dur": 0,
    "args": {
      "External id": 90272,
      "cbid": 205, "correlation": 90272
    }
  },
  {
    "ph": "f", "id": 90272, "pid": 76337, "tid": -914061504, "ts": 1716454222749414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804715, "dur": 5,
    "args": {
      "External id": 90276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90276,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90276, "pid": 5, "tid": 7, "ts": 1716454222804715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749429, "dur": 12,
    "args": {
      "External id": 90276,
      "cbid": 211, "correlation": 90276
    }
  },
  {
    "ph": "s", "id": 90276, "pid": 76337, "tid": -914061504, "ts": 1716454222749429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222804721, "dur": 159,
    "args": {
      "External id": 90278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90278,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90278, "pid": 5, "tid": 7, "ts": 1716454222804721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749443, "dur": 5,
    "args": {
      "External id": 90278,
      "cbid": 211, "correlation": 90278
    }
  },
  {
    "ph": "s", "id": 90278, "pid": 76337, "tid": -914061504, "ts": 1716454222749443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222804882, "dur": 1,
    "args": {
      "External id": 90280,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90280,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 90280, "pid": 5, "tid": 7, "ts": 1716454222804882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749454, "dur": 6,
    "args": {
      "External id": 90280,
      "cbid": 51, "correlation": 90280
    }
  },
  {
    "ph": "s", "id": 90280, "pid": 76337, "tid": -914061504, "ts": 1716454222749454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222804886, "dur": 255,
    "args": {
      "External id": 90281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90281,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90281, "pid": 5, "tid": 7, "ts": 1716454222804886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749461, "dur": 6,
    "args": {
      "External id": 90281,
      "cbid": 211, "correlation": 90281
    }
  },
  {
    "ph": "s", "id": 90281, "pid": 76337, "tid": -914061504, "ts": 1716454222749461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805142, "dur": 6,
    "args": {
      "External id": 90283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90283,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90283, "pid": 5, "tid": 7, "ts": 1716454222805142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749472, "dur": 6,
    "args": {
      "External id": 90283,
      "cbid": 211, "correlation": 90283
    }
  },
  {
    "ph": "s", "id": 90283, "pid": 76337, "tid": -914061504, "ts": 1716454222749472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222805149, "dur": 6,
    "args": {
      "External id": 90289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90289, "pid": 5, "tid": 7, "ts": 1716454222805149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749500, "dur": 8,
    "args": {
      "External id": 90289,
      "cbid": 211, "correlation": 90289
    }
  },
  {
    "ph": "s", "id": 90289, "pid": 76337, "tid": -914061504, "ts": 1716454222749500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222805157, "dur": 5,
    "args": {
      "External id": 90297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90297,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90297, "pid": 5, "tid": 7, "ts": 1716454222805157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749533, "dur": 9,
    "args": {
      "External id": 90297,
      "cbid": 211, "correlation": 90297
    }
  },
  {
    "ph": "s", "id": 90297, "pid": 76337, "tid": -914061504, "ts": 1716454222749533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222805163, "dur": 4,
    "args": {
      "External id": 90305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90305,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90305, "pid": 5, "tid": 7, "ts": 1716454222805163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749562, "dur": 9,
    "args": {
      "External id": 90305,
      "cbid": 211, "correlation": 90305
    }
  },
  {
    "ph": "s", "id": 90305, "pid": 76337, "tid": -914061504, "ts": 1716454222749562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805169, "dur": 9,
    "args": {
      "External id": 90325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90325,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90325, "pid": 5, "tid": 7, "ts": 1716454222805169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749637, "dur": 13,
    "args": {
      "External id": 90325,
      "cbid": 211, "correlation": 90325
    }
  },
  {
    "ph": "s", "id": 90325, "pid": 76337, "tid": -914061504, "ts": 1716454222749637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222805179, "dur": 4,
    "args": {
      "External id": 90337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90337,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90337, "pid": 5, "tid": 7, "ts": 1716454222805179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749660, "dur": 6,
    "args": {
      "External id": 90337,
      "cbid": 211, "correlation": 90337
    }
  },
  {
    "ph": "s", "id": 90337, "pid": 76337, "tid": -914061504, "ts": 1716454222749660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222805184, "dur": 7,
    "args": {
      "External id": 90340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90340, "pid": 5, "tid": 7, "ts": 1716454222805184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749678, "dur": 7,
    "args": {
      "External id": 90340,
      "cbid": 211, "correlation": 90340
    }
  },
  {
    "ph": "s", "id": 90340, "pid": 76337, "tid": -914061504, "ts": 1716454222749678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222805192, "dur": 4,
    "args": {
      "External id": 90349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90349,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90349, "pid": 5, "tid": 7, "ts": 1716454222805192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749717, "dur": 9,
    "args": {
      "External id": 90349,
      "cbid": 211, "correlation": 90349
    }
  },
  {
    "ph": "s", "id": 90349, "pid": 76337, "tid": -914061504, "ts": 1716454222749717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749768, "dur": 0,
    "args": {
      "External id": 90359,
      "cbid": 317, "correlation": 90359
    }
  },
  {
    "ph": "f", "id": 90359, "pid": 76337, "tid": -914061504, "ts": 1716454222749768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749769, "dur": 0,
    "args": {
      "External id": 90360,
      "cbid": 203, "correlation": 90360
    }
  },
  {
    "ph": "f", "id": 90360, "pid": 76337, "tid": -914061504, "ts": 1716454222749769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749770, "dur": 0,
    "args": {
      "External id": 90361,
      "cbid": 205, "correlation": 90361
    }
  },
  {
    "ph": "f", "id": 90361, "pid": 76337, "tid": -914061504, "ts": 1716454222749770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805198, "dur": 5,
    "args": {
      "External id": 90365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90365,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90365, "pid": 5, "tid": 7, "ts": 1716454222805198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749783, "dur": 11,
    "args": {
      "External id": 90365,
      "cbid": 211, "correlation": 90365
    }
  },
  {
    "ph": "s", "id": 90365, "pid": 76337, "tid": -914061504, "ts": 1716454222749783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805204, "dur": 160,
    "args": {
      "External id": 90367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90367,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90367, "pid": 5, "tid": 7, "ts": 1716454222805204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749797, "dur": 5,
    "args": {
      "External id": 90367,
      "cbid": 211, "correlation": 90367
    }
  },
  {
    "ph": "s", "id": 90367, "pid": 76337, "tid": -914061504, "ts": 1716454222749797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222805367, "dur": 1,
    "args": {
      "External id": 90369,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90369,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 90369, "pid": 5, "tid": 7, "ts": 1716454222805367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749808, "dur": 7,
    "args": {
      "External id": 90369,
      "cbid": 51, "correlation": 90369
    }
  },
  {
    "ph": "s", "id": 90369, "pid": 76337, "tid": -914061504, "ts": 1716454222749808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222805370, "dur": 255,
    "args": {
      "External id": 90370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90370,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90370, "pid": 5, "tid": 7, "ts": 1716454222805370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749815, "dur": 6,
    "args": {
      "External id": 90370,
      "cbid": 211, "correlation": 90370
    }
  },
  {
    "ph": "s", "id": 90370, "pid": 76337, "tid": -914061504, "ts": 1716454222749815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805626, "dur": 6,
    "args": {
      "External id": 90372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90372,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90372, "pid": 5, "tid": 7, "ts": 1716454222805626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749826, "dur": 5,
    "args": {
      "External id": 90372,
      "cbid": 211, "correlation": 90372
    }
  },
  {
    "ph": "s", "id": 90372, "pid": 76337, "tid": -914061504, "ts": 1716454222749826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222805633, "dur": 6,
    "args": {
      "External id": 90378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90378,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90378, "pid": 5, "tid": 7, "ts": 1716454222805633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749853, "dur": 9,
    "args": {
      "External id": 90378,
      "cbid": 211, "correlation": 90378
    }
  },
  {
    "ph": "s", "id": 90378, "pid": 76337, "tid": -914061504, "ts": 1716454222749853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222805640, "dur": 3,
    "args": {
      "External id": 90386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90386,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 90386, "pid": 5, "tid": 7, "ts": 1716454222805640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749897, "dur": 9,
    "args": {
      "External id": 90386,
      "cbid": 211, "correlation": 90386
    }
  },
  {
    "ph": "s", "id": 90386, "pid": 76337, "tid": -914061504, "ts": 1716454222749897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749961, "dur": 1,
    "args": {
      "External id": 90402,
      "cbid": 251, "correlation": 90402
    }
  },
  {
    "ph": "f", "id": 90402, "pid": 76337, "tid": -914061504, "ts": 1716454222749961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749966, "dur": 0,
    "args": {
      "External id": 90404,
      "cbid": 251, "correlation": 90404
    }
  },
  {
    "ph": "f", "id": 90404, "pid": 76337, "tid": -914061504, "ts": 1716454222749966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222805645, "dur": 10,
    "args": {
      "External id": 90405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90405,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90405, "pid": 5, "tid": 7, "ts": 1716454222805645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749968, "dur": 21,
    "args": {
      "External id": 90405,
      "cbid": 211, "correlation": 90405
    }
  },
  {
    "ph": "s", "id": 90405, "pid": 76337, "tid": -914061504, "ts": 1716454222749968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222805656, "dur": 4,
    "args": {
      "External id": 90407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90407,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90407, "pid": 5, "tid": 7, "ts": 1716454222805656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222749991, "dur": 6,
    "args": {
      "External id": 90407,
      "cbid": 211, "correlation": 90407
    }
  },
  {
    "ph": "s", "id": 90407, "pid": 76337, "tid": -914061504, "ts": 1716454222749991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222805661, "dur": 6,
    "args": {
      "External id": 90417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90417, "pid": 5, "tid": 7, "ts": 1716454222805661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750049, "dur": 12,
    "args": {
      "External id": 90417,
      "cbid": 211, "correlation": 90417
    }
  },
  {
    "ph": "s", "id": 90417, "pid": 76337, "tid": -914061504, "ts": 1716454222750049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805667, "dur": 9,
    "args": {
      "External id": 90437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90437,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90437, "pid": 5, "tid": 7, "ts": 1716454222805667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750115, "dur": 11,
    "args": {
      "External id": 90437,
      "cbid": 211, "correlation": 90437
    }
  },
  {
    "ph": "s", "id": 90437, "pid": 76337, "tid": -914061504, "ts": 1716454222750115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222805678, "dur": 4,
    "args": {
      "External id": 90449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90449,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90449, "pid": 5, "tid": 7, "ts": 1716454222805678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750136, "dur": 6,
    "args": {
      "External id": 90449,
      "cbid": 211, "correlation": 90449
    }
  },
  {
    "ph": "s", "id": 90449, "pid": 76337, "tid": -914061504, "ts": 1716454222750136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222805683, "dur": 6,
    "args": {
      "External id": 90452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90452, "pid": 5, "tid": 7, "ts": 1716454222805683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750157, "dur": 6,
    "args": {
      "External id": 90452,
      "cbid": 211, "correlation": 90452
    }
  },
  {
    "ph": "s", "id": 90452, "pid": 76337, "tid": -914061504, "ts": 1716454222750157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222805691, "dur": 4,
    "args": {
      "External id": 90461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90461,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90461, "pid": 5, "tid": 7, "ts": 1716454222805691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750197, "dur": 10,
    "args": {
      "External id": 90461,
      "cbid": 211, "correlation": 90461
    }
  },
  {
    "ph": "s", "id": 90461, "pid": 76337, "tid": -914061504, "ts": 1716454222750197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750260, "dur": 0,
    "args": {
      "External id": 90471,
      "cbid": 317, "correlation": 90471
    }
  },
  {
    "ph": "f", "id": 90471, "pid": 76337, "tid": -914061504, "ts": 1716454222750260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750261, "dur": 0,
    "args": {
      "External id": 90472,
      "cbid": 203, "correlation": 90472
    }
  },
  {
    "ph": "f", "id": 90472, "pid": 76337, "tid": -914061504, "ts": 1716454222750261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750262, "dur": 0,
    "args": {
      "External id": 90473,
      "cbid": 205, "correlation": 90473
    }
  },
  {
    "ph": "f", "id": 90473, "pid": 76337, "tid": -914061504, "ts": 1716454222750262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805696, "dur": 5,
    "args": {
      "External id": 90477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90477,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90477, "pid": 5, "tid": 7, "ts": 1716454222805696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750275, "dur": 12,
    "args": {
      "External id": 90477,
      "cbid": 211, "correlation": 90477
    }
  },
  {
    "ph": "s", "id": 90477, "pid": 76337, "tid": -914061504, "ts": 1716454222750275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222805703, "dur": 159,
    "args": {
      "External id": 90479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90479,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90479, "pid": 5, "tid": 7, "ts": 1716454222805703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750290, "dur": 5,
    "args": {
      "External id": 90479,
      "cbid": 211, "correlation": 90479
    }
  },
  {
    "ph": "s", "id": 90479, "pid": 76337, "tid": -914061504, "ts": 1716454222750290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222805864, "dur": 1,
    "args": {
      "External id": 90481,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90481,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 90481, "pid": 5, "tid": 7, "ts": 1716454222805864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750301, "dur": 6,
    "args": {
      "External id": 90481,
      "cbid": 51, "correlation": 90481
    }
  },
  {
    "ph": "s", "id": 90481, "pid": 76337, "tid": -914061504, "ts": 1716454222750301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222805868, "dur": 253,
    "args": {
      "External id": 90482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90482,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90482, "pid": 5, "tid": 7, "ts": 1716454222805868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750308, "dur": 6,
    "args": {
      "External id": 90482,
      "cbid": 211, "correlation": 90482
    }
  },
  {
    "ph": "s", "id": 90482, "pid": 76337, "tid": -914061504, "ts": 1716454222750308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806122, "dur": 6,
    "args": {
      "External id": 90484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90484,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90484, "pid": 5, "tid": 7, "ts": 1716454222806122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750320, "dur": 5,
    "args": {
      "External id": 90484,
      "cbid": 211, "correlation": 90484
    }
  },
  {
    "ph": "s", "id": 90484, "pid": 76337, "tid": -914061504, "ts": 1716454222750320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222806129, "dur": 6,
    "args": {
      "External id": 90490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90490, "pid": 5, "tid": 7, "ts": 1716454222806129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750348, "dur": 9,
    "args": {
      "External id": 90490,
      "cbid": 211, "correlation": 90490
    }
  },
  {
    "ph": "s", "id": 90490, "pid": 76337, "tid": -914061504, "ts": 1716454222750348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222806136, "dur": 5,
    "args": {
      "External id": 90498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90498,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90498, "pid": 5, "tid": 7, "ts": 1716454222806136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750382, "dur": 9,
    "args": {
      "External id": 90498,
      "cbid": 211, "correlation": 90498
    }
  },
  {
    "ph": "s", "id": 90498, "pid": 76337, "tid": -914061504, "ts": 1716454222750382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222806142, "dur": 4,
    "args": {
      "External id": 90506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90506,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90506, "pid": 5, "tid": 7, "ts": 1716454222806142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750411, "dur": 10,
    "args": {
      "External id": 90506,
      "cbid": 211, "correlation": 90506
    }
  },
  {
    "ph": "s", "id": 90506, "pid": 76337, "tid": -914061504, "ts": 1716454222750411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806148, "dur": 10,
    "args": {
      "External id": 90526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90526,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90526, "pid": 5, "tid": 7, "ts": 1716454222806148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750519, "dur": 13,
    "args": {
      "External id": 90526,
      "cbid": 211, "correlation": 90526
    }
  },
  {
    "ph": "s", "id": 90526, "pid": 76337, "tid": -914061504, "ts": 1716454222750519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222806159, "dur": 3,
    "args": {
      "External id": 90538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90538,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90538, "pid": 5, "tid": 7, "ts": 1716454222806159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750542, "dur": 6,
    "args": {
      "External id": 90538,
      "cbid": 211, "correlation": 90538
    }
  },
  {
    "ph": "s", "id": 90538, "pid": 76337, "tid": -914061504, "ts": 1716454222750542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222806164, "dur": 6,
    "args": {
      "External id": 90541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90541, "pid": 5, "tid": 7, "ts": 1716454222806164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750559, "dur": 7,
    "args": {
      "External id": 90541,
      "cbid": 211, "correlation": 90541
    }
  },
  {
    "ph": "s", "id": 90541, "pid": 76337, "tid": -914061504, "ts": 1716454222750559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222806171, "dur": 4,
    "args": {
      "External id": 90550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90550,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90550, "pid": 5, "tid": 7, "ts": 1716454222806171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750597, "dur": 9,
    "args": {
      "External id": 90550,
      "cbid": 211, "correlation": 90550
    }
  },
  {
    "ph": "s", "id": 90550, "pid": 76337, "tid": -914061504, "ts": 1716454222750597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750650, "dur": 0,
    "args": {
      "External id": 90560,
      "cbid": 317, "correlation": 90560
    }
  },
  {
    "ph": "f", "id": 90560, "pid": 76337, "tid": -914061504, "ts": 1716454222750650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750651, "dur": 0,
    "args": {
      "External id": 90561,
      "cbid": 203, "correlation": 90561
    }
  },
  {
    "ph": "f", "id": 90561, "pid": 76337, "tid": -914061504, "ts": 1716454222750651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750652, "dur": 0,
    "args": {
      "External id": 90562,
      "cbid": 205, "correlation": 90562
    }
  },
  {
    "ph": "f", "id": 90562, "pid": 76337, "tid": -914061504, "ts": 1716454222750652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806177, "dur": 5,
    "args": {
      "External id": 90566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90566,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90566, "pid": 5, "tid": 7, "ts": 1716454222806177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750665, "dur": 12,
    "args": {
      "External id": 90566,
      "cbid": 211, "correlation": 90566
    }
  },
  {
    "ph": "s", "id": 90566, "pid": 76337, "tid": -914061504, "ts": 1716454222750665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806183, "dur": 159,
    "args": {
      "External id": 90568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90568,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90568, "pid": 5, "tid": 7, "ts": 1716454222806183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750679, "dur": 5,
    "args": {
      "External id": 90568,
      "cbid": 211, "correlation": 90568
    }
  },
  {
    "ph": "s", "id": 90568, "pid": 76337, "tid": -914061504, "ts": 1716454222750679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222806345, "dur": 1,
    "args": {
      "External id": 90570,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90570,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 90570, "pid": 5, "tid": 7, "ts": 1716454222806345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750690, "dur": 6,
    "args": {
      "External id": 90570,
      "cbid": 51, "correlation": 90570
    }
  },
  {
    "ph": "s", "id": 90570, "pid": 76337, "tid": -914061504, "ts": 1716454222750690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222806348, "dur": 253,
    "args": {
      "External id": 90571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90571,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90571, "pid": 5, "tid": 7, "ts": 1716454222806348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750697, "dur": 7,
    "args": {
      "External id": 90571,
      "cbid": 211, "correlation": 90571
    }
  },
  {
    "ph": "s", "id": 90571, "pid": 76337, "tid": -914061504, "ts": 1716454222750697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806603, "dur": 6,
    "args": {
      "External id": 90573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90573,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90573, "pid": 5, "tid": 7, "ts": 1716454222806603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750708, "dur": 5,
    "args": {
      "External id": 90573,
      "cbid": 211, "correlation": 90573
    }
  },
  {
    "ph": "s", "id": 90573, "pid": 76337, "tid": -914061504, "ts": 1716454222750708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222806609, "dur": 6,
    "args": {
      "External id": 90579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90579,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90579, "pid": 5, "tid": 7, "ts": 1716454222806609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750736, "dur": 8,
    "args": {
      "External id": 90579,
      "cbid": 211, "correlation": 90579
    }
  },
  {
    "ph": "s", "id": 90579, "pid": 76337, "tid": -914061504, "ts": 1716454222750736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222806617, "dur": 3,
    "args": {
      "External id": 90587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90587,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 90587, "pid": 5, "tid": 7, "ts": 1716454222806617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750779, "dur": 9,
    "args": {
      "External id": 90587,
      "cbid": 211, "correlation": 90587
    }
  },
  {
    "ph": "s", "id": 90587, "pid": 76337, "tid": -914061504, "ts": 1716454222750779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750843, "dur": 1,
    "args": {
      "External id": 90603,
      "cbid": 251, "correlation": 90603
    }
  },
  {
    "ph": "f", "id": 90603, "pid": 76337, "tid": -914061504, "ts": 1716454222750843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750848, "dur": 0,
    "args": {
      "External id": 90605,
      "cbid": 251, "correlation": 90605
    }
  },
  {
    "ph": "f", "id": 90605, "pid": 76337, "tid": -914061504, "ts": 1716454222750848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222806621, "dur": 10,
    "args": {
      "External id": 90606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90606,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90606, "pid": 5, "tid": 7, "ts": 1716454222806621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750850, "dur": 11,
    "args": {
      "External id": 90606,
      "cbid": 211, "correlation": 90606
    }
  },
  {
    "ph": "s", "id": 90606, "pid": 76337, "tid": -914061504, "ts": 1716454222750850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222806632, "dur": 4,
    "args": {
      "External id": 90608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90608,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90608, "pid": 5, "tid": 7, "ts": 1716454222806632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750863, "dur": 6,
    "args": {
      "External id": 90608,
      "cbid": 211, "correlation": 90608
    }
  },
  {
    "ph": "s", "id": 90608, "pid": 76337, "tid": -914061504, "ts": 1716454222750863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222806638, "dur": 5,
    "args": {
      "External id": 90618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90618,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90618, "pid": 5, "tid": 7, "ts": 1716454222806638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750919, "dur": 13,
    "args": {
      "External id": 90618,
      "cbid": 211, "correlation": 90618
    }
  },
  {
    "ph": "s", "id": 90618, "pid": 76337, "tid": -914061504, "ts": 1716454222750919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806644, "dur": 9,
    "args": {
      "External id": 90638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90638,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90638, "pid": 5, "tid": 7, "ts": 1716454222806644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222750995, "dur": 11,
    "args": {
      "External id": 90638,
      "cbid": 211, "correlation": 90638
    }
  },
  {
    "ph": "s", "id": 90638, "pid": 76337, "tid": -914061504, "ts": 1716454222750995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222806655, "dur": 3,
    "args": {
      "External id": 90650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90650,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90650, "pid": 5, "tid": 7, "ts": 1716454222806655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751016, "dur": 6,
    "args": {
      "External id": 90650,
      "cbid": 211, "correlation": 90650
    }
  },
  {
    "ph": "s", "id": 90650, "pid": 76337, "tid": -914061504, "ts": 1716454222751016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222806660, "dur": 6,
    "args": {
      "External id": 90653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90653,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90653, "pid": 5, "tid": 7, "ts": 1716454222806660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751036, "dur": 7,
    "args": {
      "External id": 90653,
      "cbid": 211, "correlation": 90653
    }
  },
  {
    "ph": "s", "id": 90653, "pid": 76337, "tid": -914061504, "ts": 1716454222751036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222806667, "dur": 4,
    "args": {
      "External id": 90662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90662,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90662, "pid": 5, "tid": 7, "ts": 1716454222806667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751077, "dur": 10,
    "args": {
      "External id": 90662,
      "cbid": 211, "correlation": 90662
    }
  },
  {
    "ph": "s", "id": 90662, "pid": 76337, "tid": -914061504, "ts": 1716454222751077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751141, "dur": 0,
    "args": {
      "External id": 90672,
      "cbid": 317, "correlation": 90672
    }
  },
  {
    "ph": "f", "id": 90672, "pid": 76337, "tid": -914061504, "ts": 1716454222751141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751142, "dur": 0,
    "args": {
      "External id": 90673,
      "cbid": 203, "correlation": 90673
    }
  },
  {
    "ph": "f", "id": 90673, "pid": 76337, "tid": -914061504, "ts": 1716454222751142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751142, "dur": 0,
    "args": {
      "External id": 90674,
      "cbid": 205, "correlation": 90674
    }
  },
  {
    "ph": "f", "id": 90674, "pid": 76337, "tid": -914061504, "ts": 1716454222751142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806673, "dur": 5,
    "args": {
      "External id": 90678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90678, "pid": 5, "tid": 7, "ts": 1716454222806673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751157, "dur": 12,
    "args": {
      "External id": 90678,
      "cbid": 211, "correlation": 90678
    }
  },
  {
    "ph": "s", "id": 90678, "pid": 76337, "tid": -914061504, "ts": 1716454222751157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222806679, "dur": 159,
    "args": {
      "External id": 90680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90680,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90680, "pid": 5, "tid": 7, "ts": 1716454222806679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751172, "dur": 5,
    "args": {
      "External id": 90680,
      "cbid": 211, "correlation": 90680
    }
  },
  {
    "ph": "s", "id": 90680, "pid": 76337, "tid": -914061504, "ts": 1716454222751172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222806841, "dur": 1,
    "args": {
      "External id": 90682,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90682,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 90682, "pid": 5, "tid": 7, "ts": 1716454222806841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751182, "dur": 6,
    "args": {
      "External id": 90682,
      "cbid": 51, "correlation": 90682
    }
  },
  {
    "ph": "s", "id": 90682, "pid": 76337, "tid": -914061504, "ts": 1716454222751182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222806845, "dur": 254,
    "args": {
      "External id": 90683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90683,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90683, "pid": 5, "tid": 7, "ts": 1716454222806845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751190, "dur": 6,
    "args": {
      "External id": 90683,
      "cbid": 211, "correlation": 90683
    }
  },
  {
    "ph": "s", "id": 90683, "pid": 76337, "tid": -914061504, "ts": 1716454222751190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807100, "dur": 6,
    "args": {
      "External id": 90685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90685,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90685, "pid": 5, "tid": 7, "ts": 1716454222807100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751199, "dur": 5,
    "args": {
      "External id": 90685,
      "cbid": 211, "correlation": 90685
    }
  },
  {
    "ph": "s", "id": 90685, "pid": 76337, "tid": -914061504, "ts": 1716454222751199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807107, "dur": 6,
    "args": {
      "External id": 90691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90691, "pid": 5, "tid": 7, "ts": 1716454222807107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751228, "dur": 9,
    "args": {
      "External id": 90691,
      "cbid": 211, "correlation": 90691
    }
  },
  {
    "ph": "s", "id": 90691, "pid": 76337, "tid": -914061504, "ts": 1716454222751228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222807115, "dur": 5,
    "args": {
      "External id": 90699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90699,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90699, "pid": 5, "tid": 7, "ts": 1716454222807115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751260, "dur": 9,
    "args": {
      "External id": 90699,
      "cbid": 211, "correlation": 90699
    }
  },
  {
    "ph": "s", "id": 90699, "pid": 76337, "tid": -914061504, "ts": 1716454222751260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222807121, "dur": 4,
    "args": {
      "External id": 90707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90707,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90707, "pid": 5, "tid": 7, "ts": 1716454222807121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751291, "dur": 9,
    "args": {
      "External id": 90707,
      "cbid": 211, "correlation": 90707
    }
  },
  {
    "ph": "s", "id": 90707, "pid": 76337, "tid": -914061504, "ts": 1716454222751291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807126, "dur": 9,
    "args": {
      "External id": 90727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90727,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 90727, "pid": 5, "tid": 7, "ts": 1716454222807126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751439, "dur": 14,
    "args": {
      "External id": 90727,
      "cbid": 211, "correlation": 90727
    }
  },
  {
    "ph": "s", "id": 90727, "pid": 76337, "tid": -914061504, "ts": 1716454222751439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222807137, "dur": 4,
    "args": {
      "External id": 90739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90739,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 90739, "pid": 5, "tid": 7, "ts": 1716454222807137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751462, "dur": 7,
    "args": {
      "External id": 90739,
      "cbid": 211, "correlation": 90739
    }
  },
  {
    "ph": "s", "id": 90739, "pid": 76337, "tid": -914061504, "ts": 1716454222751462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807142, "dur": 6,
    "args": {
      "External id": 90742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90742, "pid": 5, "tid": 7, "ts": 1716454222807142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751480, "dur": 7,
    "args": {
      "External id": 90742,
      "cbid": 211, "correlation": 90742
    }
  },
  {
    "ph": "s", "id": 90742, "pid": 76337, "tid": -914061504, "ts": 1716454222751480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751539, "dur": 0,
    "args": {
      "External id": 90753,
      "cbid": 317, "correlation": 90753
    }
  },
  {
    "ph": "f", "id": 90753, "pid": 76337, "tid": -914061504, "ts": 1716454222751539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751540, "dur": 0,
    "args": {
      "External id": 90754,
      "cbid": 203, "correlation": 90754
    }
  },
  {
    "ph": "f", "id": 90754, "pid": 76337, "tid": -914061504, "ts": 1716454222751540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751541, "dur": 0,
    "args": {
      "External id": 90755,
      "cbid": 205, "correlation": 90755
    }
  },
  {
    "ph": "f", "id": 90755, "pid": 76337, "tid": -914061504, "ts": 1716454222751541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807150, "dur": 5,
    "args": {
      "External id": 90759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90759,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90759, "pid": 5, "tid": 7, "ts": 1716454222807150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751559, "dur": 12,
    "args": {
      "External id": 90759,
      "cbid": 211, "correlation": 90759
    }
  },
  {
    "ph": "s", "id": 90759, "pid": 76337, "tid": -914061504, "ts": 1716454222751559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222807156, "dur": 36,
    "args": {
      "External id": 90761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90761,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 90761, "pid": 5, "tid": 7, "ts": 1716454222807156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751579, "dur": 9,
    "args": {
      "External id": 90761,
      "cbid": 211, "correlation": 90761
    }
  },
  {
    "ph": "s", "id": 90761, "pid": 76337, "tid": -914061504, "ts": 1716454222751579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807194, "dur": 5,
    "args": {
      "External id": 90763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90763,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90763, "pid": 5, "tid": 7, "ts": 1716454222807194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751591, "dur": 5,
    "args": {
      "External id": 90763,
      "cbid": 211, "correlation": 90763
    }
  },
  {
    "ph": "s", "id": 90763, "pid": 76337, "tid": -914061504, "ts": 1716454222751591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807200, "dur": 6,
    "args": {
      "External id": 90769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90769,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90769, "pid": 5, "tid": 7, "ts": 1716454222807200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751619, "dur": 8,
    "args": {
      "External id": 90769,
      "cbid": 211, "correlation": 90769
    }
  },
  {
    "ph": "s", "id": 90769, "pid": 76337, "tid": -914061504, "ts": 1716454222751619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222807207, "dur": 20,
    "args": {
      "External id": 90778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90778,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90778, "pid": 5, "tid": 7, "ts": 1716454222807207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751705, "dur": 16,
    "args": {
      "External id": 90778,
      "cbid": 211, "correlation": 90778
    }
  },
  {
    "ph": "s", "id": 90778, "pid": 76337, "tid": -914061504, "ts": 1716454222751705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807229, "dur": 11,
    "args": {
      "External id": 90800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90800,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 90800, "pid": 5, "tid": 7, "ts": 1716454222807229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751764, "dur": 12,
    "args": {
      "External id": 90800,
      "cbid": 211, "correlation": 90800
    }
  },
  {
    "ph": "s", "id": 90800, "pid": 76337, "tid": -914061504, "ts": 1716454222751764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751861, "dur": 2,
    "args": {
      "External id": 90811,
      "cbid": 251, "correlation": 90811
    }
  },
  {
    "ph": "f", "id": 90811, "pid": 76337, "tid": -914061504, "ts": 1716454222751861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751867, "dur": 0,
    "args": {
      "External id": 90812,
      "cbid": 251, "correlation": 90812
    }
  },
  {
    "ph": "f", "id": 90812, "pid": 76337, "tid": -914061504, "ts": 1716454222751867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807241, "dur": 53,
    "args": {
      "External id": 90813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90813,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 90813, "pid": 5, "tid": 7, "ts": 1716454222807241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751871, "dur": 15,
    "args": {
      "External id": 90813,
      "cbid": 211, "correlation": 90813
    }
  },
  {
    "ph": "s", "id": 90813, "pid": 76337, "tid": -914061504, "ts": 1716454222751871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751944, "dur": 1,
    "args": {
      "External id": 90824,
      "cbid": 251, "correlation": 90824
    }
  },
  {
    "ph": "f", "id": 90824, "pid": 76337, "tid": -914061504, "ts": 1716454222751944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751948, "dur": 0,
    "args": {
      "External id": 90825,
      "cbid": 251, "correlation": 90825
    }
  },
  {
    "ph": "f", "id": 90825, "pid": 76337, "tid": -914061504, "ts": 1716454222751948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807295, "dur": 52,
    "args": {
      "External id": 90826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90826,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 90826, "pid": 5, "tid": 7, "ts": 1716454222807295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222751950, "dur": 12,
    "args": {
      "External id": 90826,
      "cbid": 211, "correlation": 90826
    }
  },
  {
    "ph": "s", "id": 90826, "pid": 76337, "tid": -914061504, "ts": 1716454222751950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752026, "dur": 1,
    "args": {
      "External id": 90837,
      "cbid": 251, "correlation": 90837
    }
  },
  {
    "ph": "f", "id": 90837, "pid": 76337, "tid": -914061504, "ts": 1716454222752026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752030, "dur": 0,
    "args": {
      "External id": 90838,
      "cbid": 251, "correlation": 90838
    }
  },
  {
    "ph": "f", "id": 90838, "pid": 76337, "tid": -914061504, "ts": 1716454222752030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807348, "dur": 53,
    "args": {
      "External id": 90839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90839,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 90839, "pid": 5, "tid": 7, "ts": 1716454222807348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752032, "dur": 12,
    "args": {
      "External id": 90839,
      "cbid": 211, "correlation": 90839
    }
  },
  {
    "ph": "s", "id": 90839, "pid": 76337, "tid": -914061504, "ts": 1716454222752032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222807402, "dur": 56,
    "args": {
      "External id": 90864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90864,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90864, "pid": 5, "tid": 7, "ts": 1716454222807402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752124, "dur": 13,
    "args": {
      "External id": 90864,
      "cbid": 211, "correlation": 90864
    }
  },
  {
    "ph": "s", "id": 90864, "pid": 76337, "tid": -914061504, "ts": 1716454222752124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752229, "dur": 1,
    "args": {
      "External id": 90882,
      "cbid": 251, "correlation": 90882
    }
  },
  {
    "ph": "f", "id": 90882, "pid": 76337, "tid": -914061504, "ts": 1716454222752229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807460, "dur": 62,
    "args": {
      "External id": 90884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90884,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 90884, "pid": 5, "tid": 7, "ts": 1716454222807460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752235, "dur": 14,
    "args": {
      "External id": 90884,
      "cbid": 211, "correlation": 90884
    }
  },
  {
    "ph": "s", "id": 90884, "pid": 76337, "tid": -914061504, "ts": 1716454222752235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222807523, "dur": 6,
    "args": {
      "External id": 90892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90892,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90892, "pid": 5, "tid": 7, "ts": 1716454222807523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752308, "dur": 14,
    "args": {
      "External id": 90892,
      "cbid": 211, "correlation": 90892
    }
  },
  {
    "ph": "s", "id": 90892, "pid": 76337, "tid": -914061504, "ts": 1716454222752308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222807531, "dur": 7,
    "args": {
      "External id": 90900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90900,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 90900, "pid": 5, "tid": 7, "ts": 1716454222807531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752348, "dur": 9,
    "args": {
      "External id": 90900,
      "cbid": 211, "correlation": 90900
    }
  },
  {
    "ph": "s", "id": 90900, "pid": 76337, "tid": -914061504, "ts": 1716454222752348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807539, "dur": 7,
    "args": {
      "External id": 90911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90911,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 90911, "pid": 5, "tid": 7, "ts": 1716454222807539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752427, "dur": 13,
    "args": {
      "External id": 90911,
      "cbid": 211, "correlation": 90911
    }
  },
  {
    "ph": "s", "id": 90911, "pid": 76337, "tid": -914061504, "ts": 1716454222752427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807548, "dur": 8,
    "args": {
      "External id": 90933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90933,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 90933, "pid": 5, "tid": 7, "ts": 1716454222807548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752460, "dur": 8,
    "args": {
      "External id": 90933,
      "cbid": 211, "correlation": 90933
    }
  },
  {
    "ph": "s", "id": 90933, "pid": 76337, "tid": -914061504, "ts": 1716454222752460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752547, "dur": 2,
    "args": {
      "External id": 90944,
      "cbid": 251, "correlation": 90944
    }
  },
  {
    "ph": "f", "id": 90944, "pid": 76337, "tid": -914061504, "ts": 1716454222752547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222807558, "dur": 1,
    "args": {
      "External id": 90945,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 90945,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 90945, "pid": 5, "tid": 7, "ts": 1716454222807558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752553, "dur": 11,
    "args": {
      "External id": 90945,
      "cbid": 51, "correlation": 90945
    }
  },
  {
    "ph": "s", "id": 90945, "pid": 76337, "tid": -914061504, "ts": 1716454222752553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807562, "dur": 36,
    "args": {
      "External id": 90946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90946,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 90946, "pid": 5, "tid": 7, "ts": 1716454222807562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752565, "dur": 12,
    "args": {
      "External id": 90946,
      "cbid": 211, "correlation": 90946
    }
  },
  {
    "ph": "s", "id": 90946, "pid": 76337, "tid": -914061504, "ts": 1716454222752565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752638, "dur": 1,
    "args": {
      "External id": 90957,
      "cbid": 251, "correlation": 90957
    }
  },
  {
    "ph": "f", "id": 90957, "pid": 76337, "tid": -914061504, "ts": 1716454222752638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752642, "dur": 0,
    "args": {
      "External id": 90958,
      "cbid": 251, "correlation": 90958
    }
  },
  {
    "ph": "f", "id": 90958, "pid": 76337, "tid": -914061504, "ts": 1716454222752642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222807599, "dur": 12,
    "args": {
      "External id": 90959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90959,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90959, "pid": 5, "tid": 7, "ts": 1716454222807599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752644, "dur": 13,
    "args": {
      "External id": 90959,
      "cbid": 211, "correlation": 90959
    }
  },
  {
    "ph": "s", "id": 90959, "pid": 76337, "tid": -914061504, "ts": 1716454222752644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222807612, "dur": 5,
    "args": {
      "External id": 90961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90961,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90961, "pid": 5, "tid": 7, "ts": 1716454222807612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752660, "dur": 6,
    "args": {
      "External id": 90961,
      "cbid": 211, "correlation": 90961
    }
  },
  {
    "ph": "s", "id": 90961, "pid": 76337, "tid": -914061504, "ts": 1716454222752660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752717, "dur": 1,
    "args": {
      "External id": 90972,
      "cbid": 251, "correlation": 90972
    }
  },
  {
    "ph": "f", "id": 90972, "pid": 76337, "tid": -914061504, "ts": 1716454222752717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752721, "dur": 0,
    "args": {
      "External id": 90973,
      "cbid": 251, "correlation": 90973
    }
  },
  {
    "ph": "f", "id": 90973, "pid": 76337, "tid": -914061504, "ts": 1716454222752721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222807619, "dur": 8,
    "args": {
      "External id": 90974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90974,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90974, "pid": 5, "tid": 7, "ts": 1716454222807619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752722, "dur": 12,
    "args": {
      "External id": 90974,
      "cbid": 211, "correlation": 90974
    }
  },
  {
    "ph": "s", "id": 90974, "pid": 76337, "tid": -914061504, "ts": 1716454222752722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222807628, "dur": 4,
    "args": {
      "External id": 90976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 90976,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 90976, "pid": 5, "tid": 7, "ts": 1716454222807628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752735, "dur": 6,
    "args": {
      "External id": 90976,
      "cbid": 211, "correlation": 90976
    }
  },
  {
    "ph": "s", "id": 90976, "pid": 76337, "tid": -914061504, "ts": 1716454222752735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222807633, "dur": 19,
    "args": {
      "External id": 91001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91001,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 91001, "pid": 5, "tid": 7, "ts": 1716454222807633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752815, "dur": 14,
    "args": {
      "External id": 91001,
      "cbid": 211, "correlation": 91001
    }
  },
  {
    "ph": "s", "id": 91001, "pid": 76337, "tid": -914061504, "ts": 1716454222752815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752917, "dur": 2,
    "args": {
      "External id": 91019,
      "cbid": 251, "correlation": 91019
    }
  },
  {
    "ph": "f", "id": 91019, "pid": 76337, "tid": -914061504, "ts": 1716454222752917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222807655, "dur": 1,
    "args": {
      "External id": 91021,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91021,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 91021, "pid": 5, "tid": 7, "ts": 1716454222807655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752922, "dur": 11,
    "args": {
      "External id": 91021,
      "cbid": 51, "correlation": 91021
    }
  },
  {
    "ph": "s", "id": 91021, "pid": 76337, "tid": -914061504, "ts": 1716454222752922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807658, "dur": 36,
    "args": {
      "External id": 91022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91022,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 91022, "pid": 5, "tid": 7, "ts": 1716454222807658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222752935, "dur": 13,
    "args": {
      "External id": 91022,
      "cbid": 211, "correlation": 91022
    }
  },
  {
    "ph": "s", "id": 91022, "pid": 76337, "tid": -914061504, "ts": 1716454222752935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222807695, "dur": 4,
    "args": {
      "External id": 91030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91030,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91030, "pid": 5, "tid": 7, "ts": 1716454222807695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753017, "dur": 13,
    "args": {
      "External id": 91030,
      "cbid": 211, "correlation": 91030
    }
  },
  {
    "ph": "s", "id": 91030, "pid": 76337, "tid": -914061504, "ts": 1716454222753017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807700, "dur": 8,
    "args": {
      "External id": 91038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91038, "pid": 5, "tid": 7, "ts": 1716454222807700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753060, "dur": 9,
    "args": {
      "External id": 91038,
      "cbid": 211, "correlation": 91038
    }
  },
  {
    "ph": "s", "id": 91038, "pid": 76337, "tid": -914061504, "ts": 1716454222753060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222807710, "dur": 8,
    "args": {
      "External id": 91060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91060,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 91060, "pid": 5, "tid": 7, "ts": 1716454222807710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753113, "dur": 10,
    "args": {
      "External id": 91060,
      "cbid": 211, "correlation": 91060
    }
  },
  {
    "ph": "s", "id": 91060, "pid": 76337, "tid": -914061504, "ts": 1716454222753113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753206, "dur": 1,
    "args": {
      "External id": 91076,
      "cbid": 251, "correlation": 91076
    }
  },
  {
    "ph": "f", "id": 91076, "pid": 76337, "tid": -914061504, "ts": 1716454222753206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753212, "dur": 0,
    "args": {
      "External id": 91078,
      "cbid": 251, "correlation": 91078
    }
  },
  {
    "ph": "f", "id": 91078, "pid": 76337, "tid": -914061504, "ts": 1716454222753212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222807719, "dur": 188,
    "args": {
      "External id": 91079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91079,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91079, "pid": 5, "tid": 7, "ts": 1716454222807719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753215, "dur": 13,
    "args": {
      "External id": 91079,
      "cbid": 211, "correlation": 91079
    }
  },
  {
    "ph": "s", "id": 91079, "pid": 76337, "tid": -914061504, "ts": 1716454222753215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807908, "dur": 21,
    "args": {
      "External id": 91087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91087, "pid": 5, "tid": 7, "ts": 1716454222807908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753287, "dur": 13,
    "args": {
      "External id": 91087,
      "cbid": 211, "correlation": 91087
    }
  },
  {
    "ph": "s", "id": 91087, "pid": 76337, "tid": -914061504, "ts": 1716454222753287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222807931, "dur": 21,
    "args": {
      "External id": 91095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91095,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91095, "pid": 5, "tid": 7, "ts": 1716454222807931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753321, "dur": 8,
    "args": {
      "External id": 91095,
      "cbid": 211, "correlation": 91095
    }
  },
  {
    "ph": "s", "id": 91095, "pid": 76337, "tid": -914061504, "ts": 1716454222753321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753402, "dur": 1,
    "args": {
      "External id": 91111,
      "cbid": 251, "correlation": 91111
    }
  },
  {
    "ph": "f", "id": 91111, "pid": 76337, "tid": -914061504, "ts": 1716454222753402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222807954, "dur": 1,
    "args": {
      "External id": 91113,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91113,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 91113, "pid": 5, "tid": 7, "ts": 1716454222807954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753407, "dur": 9,
    "args": {
      "External id": 91113,
      "cbid": 51, "correlation": 91113
    }
  },
  {
    "ph": "s", "id": 91113, "pid": 76337, "tid": -914061504, "ts": 1716454222753407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222807958, "dur": 110,
    "args": {
      "External id": 91114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91114,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 91114, "pid": 5, "tid": 7, "ts": 1716454222807958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753417, "dur": 12,
    "args": {
      "External id": 91114,
      "cbid": 211, "correlation": 91114
    }
  },
  {
    "ph": "s", "id": 91114, "pid": 76337, "tid": -914061504, "ts": 1716454222753417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222808068, "dur": 5,
    "args": {
      "External id": 91122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91122,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91122, "pid": 5, "tid": 7, "ts": 1716454222808068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753459, "dur": 10,
    "args": {
      "External id": 91122,
      "cbid": 211, "correlation": 91122
    }
  },
  {
    "ph": "s", "id": 91122, "pid": 76337, "tid": -914061504, "ts": 1716454222753459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222808075, "dur": 9,
    "args": {
      "External id": 91133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91133, "pid": 5, "tid": 7, "ts": 1716454222808075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753533, "dur": 12,
    "args": {
      "External id": 91133,
      "cbid": 211, "correlation": 91133
    }
  },
  {
    "ph": "s", "id": 91133, "pid": 76337, "tid": -914061504, "ts": 1716454222753533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753602, "dur": 0,
    "args": {
      "External id": 91145,
      "cbid": 317, "correlation": 91145
    }
  },
  {
    "ph": "f", "id": 91145, "pid": 76337, "tid": -914061504, "ts": 1716454222753602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753603, "dur": 0,
    "args": {
      "External id": 91146,
      "cbid": 203, "correlation": 91146
    }
  },
  {
    "ph": "f", "id": 91146, "pid": 76337, "tid": -914061504, "ts": 1716454222753603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753603, "dur": 0,
    "args": {
      "External id": 91147,
      "cbid": 205, "correlation": 91147
    }
  },
  {
    "ph": "f", "id": 91147, "pid": 76337, "tid": -914061504, "ts": 1716454222753603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808085, "dur": 5,
    "args": {
      "External id": 91151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91151,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91151, "pid": 5, "tid": 7, "ts": 1716454222808085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753621, "dur": 12,
    "args": {
      "External id": 91151,
      "cbid": 211, "correlation": 91151
    }
  },
  {
    "ph": "s", "id": 91151, "pid": 76337, "tid": -914061504, "ts": 1716454222753621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222808092, "dur": 37,
    "args": {
      "External id": 91153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91153,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 91153, "pid": 5, "tid": 7, "ts": 1716454222808092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753642, "dur": 8,
    "args": {
      "External id": 91153,
      "cbid": 211, "correlation": 91153
    }
  },
  {
    "ph": "s", "id": 91153, "pid": 76337, "tid": -914061504, "ts": 1716454222753642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808130, "dur": 6,
    "args": {
      "External id": 91155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91155,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91155, "pid": 5, "tid": 7, "ts": 1716454222808130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753654, "dur": 5,
    "args": {
      "External id": 91155,
      "cbid": 211, "correlation": 91155
    }
  },
  {
    "ph": "s", "id": 91155, "pid": 76337, "tid": -914061504, "ts": 1716454222753654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222808137, "dur": 7,
    "args": {
      "External id": 91161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91161, "pid": 5, "tid": 7, "ts": 1716454222808137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753681, "dur": 9,
    "args": {
      "External id": 91161,
      "cbid": 211, "correlation": 91161
    }
  },
  {
    "ph": "s", "id": 91161, "pid": 76337, "tid": -914061504, "ts": 1716454222753681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222808145, "dur": 5,
    "args": {
      "External id": 91169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91169,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91169, "pid": 5, "tid": 7, "ts": 1716454222808145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753714, "dur": 8,
    "args": {
      "External id": 91169,
      "cbid": 211, "correlation": 91169
    }
  },
  {
    "ph": "s", "id": 91169, "pid": 76337, "tid": -914061504, "ts": 1716454222753714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808151, "dur": 11,
    "args": {
      "External id": 91189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91189,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91189, "pid": 5, "tid": 7, "ts": 1716454222808151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753792, "dur": 12,
    "args": {
      "External id": 91189,
      "cbid": 211, "correlation": 91189
    }
  },
  {
    "ph": "s", "id": 91189, "pid": 76337, "tid": -914061504, "ts": 1716454222753792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222808163, "dur": 4,
    "args": {
      "External id": 91201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91201,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 91201, "pid": 5, "tid": 7, "ts": 1716454222808163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753815, "dur": 7,
    "args": {
      "External id": 91201,
      "cbid": 211, "correlation": 91201
    }
  },
  {
    "ph": "s", "id": 91201, "pid": 76337, "tid": -914061504, "ts": 1716454222753815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222808169, "dur": 8,
    "args": {
      "External id": 91204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91204,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91204, "pid": 5, "tid": 7, "ts": 1716454222808169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753834, "dur": 6,
    "args": {
      "External id": 91204,
      "cbid": 211, "correlation": 91204
    }
  },
  {
    "ph": "s", "id": 91204, "pid": 76337, "tid": -914061504, "ts": 1716454222753834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222808178, "dur": 5,
    "args": {
      "External id": 91213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91213,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91213, "pid": 5, "tid": 7, "ts": 1716454222808178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753875, "dur": 10,
    "args": {
      "External id": 91213,
      "cbid": 211, "correlation": 91213
    }
  },
  {
    "ph": "s", "id": 91213, "pid": 76337, "tid": -914061504, "ts": 1716454222753875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753928, "dur": 0,
    "args": {
      "External id": 91223,
      "cbid": 317, "correlation": 91223
    }
  },
  {
    "ph": "f", "id": 91223, "pid": 76337, "tid": -914061504, "ts": 1716454222753928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753929, "dur": 0,
    "args": {
      "External id": 91224,
      "cbid": 203, "correlation": 91224
    }
  },
  {
    "ph": "f", "id": 91224, "pid": 76337, "tid": -914061504, "ts": 1716454222753929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753930, "dur": 0,
    "args": {
      "External id": 91225,
      "cbid": 205, "correlation": 91225
    }
  },
  {
    "ph": "f", "id": 91225, "pid": 76337, "tid": -914061504, "ts": 1716454222753930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808185, "dur": 5,
    "args": {
      "External id": 91229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91229, "pid": 5, "tid": 7, "ts": 1716454222808185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753944, "dur": 11,
    "args": {
      "External id": 91229,
      "cbid": 211, "correlation": 91229
    }
  },
  {
    "ph": "s", "id": 91229, "pid": 76337, "tid": -914061504, "ts": 1716454222753944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808191, "dur": 160,
    "args": {
      "External id": 91231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91231, "pid": 5, "tid": 7, "ts": 1716454222808191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753958, "dur": 5,
    "args": {
      "External id": 91231,
      "cbid": 211, "correlation": 91231
    }
  },
  {
    "ph": "s", "id": 91231, "pid": 76337, "tid": -914061504, "ts": 1716454222753958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222808353, "dur": 1,
    "args": {
      "External id": 91233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91233,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 91233, "pid": 5, "tid": 7, "ts": 1716454222808353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753969, "dur": 16,
    "args": {
      "External id": 91233,
      "cbid": 51, "correlation": 91233
    }
  },
  {
    "ph": "s", "id": 91233, "pid": 76337, "tid": -914061504, "ts": 1716454222753969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222808357, "dur": 265,
    "args": {
      "External id": 91234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91234,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91234, "pid": 5, "tid": 7, "ts": 1716454222808357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753986, "dur": 7,
    "args": {
      "External id": 91234,
      "cbid": 211, "correlation": 91234
    }
  },
  {
    "ph": "s", "id": 91234, "pid": 76337, "tid": -914061504, "ts": 1716454222753986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808623, "dur": 6,
    "args": {
      "External id": 91236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91236, "pid": 5, "tid": 7, "ts": 1716454222808623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222753997, "dur": 5,
    "args": {
      "External id": 91236,
      "cbid": 211, "correlation": 91236
    }
  },
  {
    "ph": "s", "id": 91236, "pid": 76337, "tid": -914061504, "ts": 1716454222753997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222808630, "dur": 6,
    "args": {
      "External id": 91242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91242, "pid": 5, "tid": 7, "ts": 1716454222808630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754025, "dur": 10,
    "args": {
      "External id": 91242,
      "cbid": 211, "correlation": 91242
    }
  },
  {
    "ph": "s", "id": 91242, "pid": 76337, "tid": -914061504, "ts": 1716454222754025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222808638, "dur": 3,
    "args": {
      "External id": 91250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91250,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 91250, "pid": 5, "tid": 7, "ts": 1716454222808638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754071, "dur": 9,
    "args": {
      "External id": 91250,
      "cbid": 211, "correlation": 91250
    }
  },
  {
    "ph": "s", "id": 91250, "pid": 76337, "tid": -914061504, "ts": 1716454222754071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754138, "dur": 1,
    "args": {
      "External id": 91266,
      "cbid": 251, "correlation": 91266
    }
  },
  {
    "ph": "f", "id": 91266, "pid": 76337, "tid": -914061504, "ts": 1716454222754138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754144, "dur": 0,
    "args": {
      "External id": 91268,
      "cbid": 251, "correlation": 91268
    }
  },
  {
    "ph": "f", "id": 91268, "pid": 76337, "tid": -914061504, "ts": 1716454222754144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222808642, "dur": 12,
    "args": {
      "External id": 91269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91269,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91269, "pid": 5, "tid": 7, "ts": 1716454222808642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754146, "dur": 12,
    "args": {
      "External id": 91269,
      "cbid": 211, "correlation": 91269
    }
  },
  {
    "ph": "s", "id": 91269, "pid": 76337, "tid": -914061504, "ts": 1716454222754146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222808656, "dur": 5,
    "args": {
      "External id": 91271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91271,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91271, "pid": 5, "tid": 7, "ts": 1716454222808656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754160, "dur": 5,
    "args": {
      "External id": 91271,
      "cbid": 211, "correlation": 91271
    }
  },
  {
    "ph": "s", "id": 91271, "pid": 76337, "tid": -914061504, "ts": 1716454222754160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222808662, "dur": 6,
    "args": {
      "External id": 91281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91281, "pid": 5, "tid": 7, "ts": 1716454222808662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754218, "dur": 12,
    "args": {
      "External id": 91281,
      "cbid": 211, "correlation": 91281
    }
  },
  {
    "ph": "s", "id": 91281, "pid": 76337, "tid": -914061504, "ts": 1716454222754218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808669, "dur": 10,
    "args": {
      "External id": 91301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91301,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91301, "pid": 5, "tid": 7, "ts": 1716454222808669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754285, "dur": 11,
    "args": {
      "External id": 91301,
      "cbid": 211, "correlation": 91301
    }
  },
  {
    "ph": "s", "id": 91301, "pid": 76337, "tid": -914061504, "ts": 1716454222754285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222808680, "dur": 4,
    "args": {
      "External id": 91313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91313,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 91313, "pid": 5, "tid": 7, "ts": 1716454222808680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754305, "dur": 6,
    "args": {
      "External id": 91313,
      "cbid": 211, "correlation": 91313
    }
  },
  {
    "ph": "s", "id": 91313, "pid": 76337, "tid": -914061504, "ts": 1716454222754305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222808685, "dur": 7,
    "args": {
      "External id": 91316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91316, "pid": 5, "tid": 7, "ts": 1716454222808685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754324, "dur": 7,
    "args": {
      "External id": 91316,
      "cbid": 211, "correlation": 91316
    }
  },
  {
    "ph": "s", "id": 91316, "pid": 76337, "tid": -914061504, "ts": 1716454222754324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222808693, "dur": 4,
    "args": {
      "External id": 91325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91325,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91325, "pid": 5, "tid": 7, "ts": 1716454222808693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754365, "dur": 10,
    "args": {
      "External id": 91325,
      "cbid": 211, "correlation": 91325
    }
  },
  {
    "ph": "s", "id": 91325, "pid": 76337, "tid": -914061504, "ts": 1716454222754365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754429, "dur": 0,
    "args": {
      "External id": 91335,
      "cbid": 317, "correlation": 91335
    }
  },
  {
    "ph": "f", "id": 91335, "pid": 76337, "tid": -914061504, "ts": 1716454222754429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754430, "dur": 0,
    "args": {
      "External id": 91336,
      "cbid": 203, "correlation": 91336
    }
  },
  {
    "ph": "f", "id": 91336, "pid": 76337, "tid": -914061504, "ts": 1716454222754430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754431, "dur": 0,
    "args": {
      "External id": 91337,
      "cbid": 205, "correlation": 91337
    }
  },
  {
    "ph": "f", "id": 91337, "pid": 76337, "tid": -914061504, "ts": 1716454222754431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808699, "dur": 5,
    "args": {
      "External id": 91341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91341,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91341, "pid": 5, "tid": 7, "ts": 1716454222808699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754445, "dur": 12,
    "args": {
      "External id": 91341,
      "cbid": 211, "correlation": 91341
    }
  },
  {
    "ph": "s", "id": 91341, "pid": 76337, "tid": -914061504, "ts": 1716454222754445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222808705, "dur": 159,
    "args": {
      "External id": 91343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91343, "pid": 5, "tid": 7, "ts": 1716454222808705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754459, "dur": 5,
    "args": {
      "External id": 91343,
      "cbid": 211, "correlation": 91343
    }
  },
  {
    "ph": "s", "id": 91343, "pid": 76337, "tid": -914061504, "ts": 1716454222754459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222808866, "dur": 1,
    "args": {
      "External id": 91345,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91345,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 91345, "pid": 5, "tid": 7, "ts": 1716454222808866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754470, "dur": 6,
    "args": {
      "External id": 91345,
      "cbid": 51, "correlation": 91345
    }
  },
  {
    "ph": "s", "id": 91345, "pid": 76337, "tid": -914061504, "ts": 1716454222754470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222808870, "dur": 255,
    "args": {
      "External id": 91346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91346,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91346, "pid": 5, "tid": 7, "ts": 1716454222808870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754477, "dur": 7,
    "args": {
      "External id": 91346,
      "cbid": 211, "correlation": 91346
    }
  },
  {
    "ph": "s", "id": 91346, "pid": 76337, "tid": -914061504, "ts": 1716454222754477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222809126, "dur": 6,
    "args": {
      "External id": 91348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91348, "pid": 5, "tid": 7, "ts": 1716454222809126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754489, "dur": 5,
    "args": {
      "External id": 91348,
      "cbid": 211, "correlation": 91348
    }
  },
  {
    "ph": "s", "id": 91348, "pid": 76337, "tid": -914061504, "ts": 1716454222754489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222809133, "dur": 6,
    "args": {
      "External id": 91354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91354, "pid": 5, "tid": 7, "ts": 1716454222809133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754517, "dur": 9,
    "args": {
      "External id": 91354,
      "cbid": 211, "correlation": 91354
    }
  },
  {
    "ph": "s", "id": 91354, "pid": 76337, "tid": -914061504, "ts": 1716454222754517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222809141, "dur": 5,
    "args": {
      "External id": 91362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91362,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91362, "pid": 5, "tid": 7, "ts": 1716454222809141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754550, "dur": 8,
    "args": {
      "External id": 91362,
      "cbid": 211, "correlation": 91362
    }
  },
  {
    "ph": "s", "id": 91362, "pid": 76337, "tid": -914061504, "ts": 1716454222754550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222809146, "dur": 4,
    "args": {
      "External id": 91370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91370,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91370, "pid": 5, "tid": 7, "ts": 1716454222809146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754580, "dur": 8,
    "args": {
      "External id": 91370,
      "cbid": 211, "correlation": 91370
    }
  },
  {
    "ph": "s", "id": 91370, "pid": 76337, "tid": -914061504, "ts": 1716454222754580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222809152, "dur": 11,
    "args": {
      "External id": 91379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91379,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91379, "pid": 5, "tid": 7, "ts": 1716454222809152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754676, "dur": 14,
    "args": {
      "External id": 91379,
      "cbid": 211, "correlation": 91379
    }
  },
  {
    "ph": "s", "id": 91379, "pid": 76337, "tid": -914061504, "ts": 1716454222754676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222809164, "dur": 12,
    "args": {
      "External id": 91399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91399,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91399, "pid": 5, "tid": 7, "ts": 1716454222809164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754745, "dur": 11,
    "args": {
      "External id": 91399,
      "cbid": 211, "correlation": 91399
    }
  },
  {
    "ph": "s", "id": 91399, "pid": 76337, "tid": -914061504, "ts": 1716454222754745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222809178, "dur": 4,
    "args": {
      "External id": 91411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91411,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91411, "pid": 5, "tid": 7, "ts": 1716454222809178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754768, "dur": 6,
    "args": {
      "External id": 91411,
      "cbid": 211, "correlation": 91411
    }
  },
  {
    "ph": "s", "id": 91411, "pid": 76337, "tid": -914061504, "ts": 1716454222754768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222809183, "dur": 10,
    "args": {
      "External id": 91414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91414,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91414, "pid": 5, "tid": 7, "ts": 1716454222809183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754786, "dur": 6,
    "args": {
      "External id": 91414,
      "cbid": 211, "correlation": 91414
    }
  },
  {
    "ph": "s", "id": 91414, "pid": 76337, "tid": -914061504, "ts": 1716454222754786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222809194, "dur": 6,
    "args": {
      "External id": 91423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91423,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91423, "pid": 5, "tid": 7, "ts": 1716454222809194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754825, "dur": 10,
    "args": {
      "External id": 91423,
      "cbid": 211, "correlation": 91423
    }
  },
  {
    "ph": "s", "id": 91423, "pid": 76337, "tid": -914061504, "ts": 1716454222754825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754878, "dur": 0,
    "args": {
      "External id": 91433,
      "cbid": 317, "correlation": 91433
    }
  },
  {
    "ph": "f", "id": 91433, "pid": 76337, "tid": -914061504, "ts": 1716454222754878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754879, "dur": 0,
    "args": {
      "External id": 91434,
      "cbid": 203, "correlation": 91434
    }
  },
  {
    "ph": "f", "id": 91434, "pid": 76337, "tid": -914061504, "ts": 1716454222754879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754880, "dur": 0,
    "args": {
      "External id": 91435,
      "cbid": 205, "correlation": 91435
    }
  },
  {
    "ph": "f", "id": 91435, "pid": 76337, "tid": -914061504, "ts": 1716454222754880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222809201, "dur": 6,
    "args": {
      "External id": 91439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91439,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91439, "pid": 5, "tid": 7, "ts": 1716454222809201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754896, "dur": 11,
    "args": {
      "External id": 91439,
      "cbid": 211, "correlation": 91439
    }
  },
  {
    "ph": "s", "id": 91439, "pid": 76337, "tid": -914061504, "ts": 1716454222754896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222809209, "dur": 314,
    "args": {
      "External id": 91441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91441,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91441, "pid": 5, "tid": 7, "ts": 1716454222809209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754910, "dur": 5,
    "args": {
      "External id": 91441,
      "cbid": 211, "correlation": 91441
    }
  },
  {
    "ph": "s", "id": 91441, "pid": 76337, "tid": -914061504, "ts": 1716454222754910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222809525, "dur": 1,
    "args": {
      "External id": 91443,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91443,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 91443, "pid": 5, "tid": 7, "ts": 1716454222809525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754921, "dur": 7,
    "args": {
      "External id": 91443,
      "cbid": 51, "correlation": 91443
    }
  },
  {
    "ph": "s", "id": 91443, "pid": 76337, "tid": -914061504, "ts": 1716454222754921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222809529, "dur": 488,
    "args": {
      "External id": 91444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91444,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91444, "pid": 5, "tid": 7, "ts": 1716454222809529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754930, "dur": 7,
    "args": {
      "External id": 91444,
      "cbid": 211, "correlation": 91444
    }
  },
  {
    "ph": "s", "id": 91444, "pid": 76337, "tid": -914061504, "ts": 1716454222754930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810018, "dur": 6,
    "args": {
      "External id": 91446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91446, "pid": 5, "tid": 7, "ts": 1716454222810018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754940, "dur": 5,
    "args": {
      "External id": 91446,
      "cbid": 211, "correlation": 91446
    }
  },
  {
    "ph": "s", "id": 91446, "pid": 76337, "tid": -914061504, "ts": 1716454222754940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222810025, "dur": 6,
    "args": {
      "External id": 91452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91452, "pid": 5, "tid": 7, "ts": 1716454222810025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222754969, "dur": 16,
    "args": {
      "External id": 91452,
      "cbid": 211, "correlation": 91452
    }
  },
  {
    "ph": "s", "id": 91452, "pid": 76337, "tid": -914061504, "ts": 1716454222754969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222810032, "dur": 3,
    "args": {
      "External id": 91460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91460,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 91460, "pid": 5, "tid": 7, "ts": 1716454222810032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755022, "dur": 10,
    "args": {
      "External id": 91460,
      "cbid": 211, "correlation": 91460
    }
  },
  {
    "ph": "s", "id": 91460, "pid": 76337, "tid": -914061504, "ts": 1716454222755022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755086, "dur": 1,
    "args": {
      "External id": 91476,
      "cbid": 251, "correlation": 91476
    }
  },
  {
    "ph": "f", "id": 91476, "pid": 76337, "tid": -914061504, "ts": 1716454222755086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755092, "dur": 0,
    "args": {
      "External id": 91478,
      "cbid": 251, "correlation": 91478
    }
  },
  {
    "ph": "f", "id": 91478, "pid": 76337, "tid": -914061504, "ts": 1716454222755092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222810036, "dur": 11,
    "args": {
      "External id": 91479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91479,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91479, "pid": 5, "tid": 7, "ts": 1716454222810036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755094, "dur": 12,
    "args": {
      "External id": 91479,
      "cbid": 211, "correlation": 91479
    }
  },
  {
    "ph": "s", "id": 91479, "pid": 76337, "tid": -914061504, "ts": 1716454222755094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222810048, "dur": 4,
    "args": {
      "External id": 91481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91481,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91481, "pid": 5, "tid": 7, "ts": 1716454222810048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755107, "dur": 5,
    "args": {
      "External id": 91481,
      "cbid": 211, "correlation": 91481
    }
  },
  {
    "ph": "s", "id": 91481, "pid": 76337, "tid": -914061504, "ts": 1716454222755107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222810054, "dur": 6,
    "args": {
      "External id": 91491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91491, "pid": 5, "tid": 7, "ts": 1716454222810054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755165, "dur": 12,
    "args": {
      "External id": 91491,
      "cbid": 211, "correlation": 91491
    }
  },
  {
    "ph": "s", "id": 91491, "pid": 76337, "tid": -914061504, "ts": 1716454222755165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810061, "dur": 9,
    "args": {
      "External id": 91511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91511,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91511, "pid": 5, "tid": 7, "ts": 1716454222810061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755231, "dur": 10,
    "args": {
      "External id": 91511,
      "cbid": 211, "correlation": 91511
    }
  },
  {
    "ph": "s", "id": 91511, "pid": 76337, "tid": -914061504, "ts": 1716454222755231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222810072, "dur": 4,
    "args": {
      "External id": 91523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91523,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 91523, "pid": 5, "tid": 7, "ts": 1716454222810072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755252, "dur": 6,
    "args": {
      "External id": 91523,
      "cbid": 211, "correlation": 91523
    }
  },
  {
    "ph": "s", "id": 91523, "pid": 76337, "tid": -914061504, "ts": 1716454222755252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222810077, "dur": 6,
    "args": {
      "External id": 91526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91526, "pid": 5, "tid": 7, "ts": 1716454222810077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755272, "dur": 7,
    "args": {
      "External id": 91526,
      "cbid": 211, "correlation": 91526
    }
  },
  {
    "ph": "s", "id": 91526, "pid": 76337, "tid": -914061504, "ts": 1716454222755272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222810085, "dur": 4,
    "args": {
      "External id": 91535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91535,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91535, "pid": 5, "tid": 7, "ts": 1716454222810085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755312, "dur": 11,
    "args": {
      "External id": 91535,
      "cbid": 211, "correlation": 91535
    }
  },
  {
    "ph": "s", "id": 91535, "pid": 76337, "tid": -914061504, "ts": 1716454222755312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755376, "dur": 0,
    "args": {
      "External id": 91545,
      "cbid": 317, "correlation": 91545
    }
  },
  {
    "ph": "f", "id": 91545, "pid": 76337, "tid": -914061504, "ts": 1716454222755376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755377, "dur": 0,
    "args": {
      "External id": 91546,
      "cbid": 203, "correlation": 91546
    }
  },
  {
    "ph": "f", "id": 91546, "pid": 76337, "tid": -914061504, "ts": 1716454222755377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755378, "dur": 0,
    "args": {
      "External id": 91547,
      "cbid": 205, "correlation": 91547
    }
  },
  {
    "ph": "f", "id": 91547, "pid": 76337, "tid": -914061504, "ts": 1716454222755378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810090, "dur": 5,
    "args": {
      "External id": 91551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91551,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91551, "pid": 5, "tid": 7, "ts": 1716454222810090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755392, "dur": 12,
    "args": {
      "External id": 91551,
      "cbid": 211, "correlation": 91551
    }
  },
  {
    "ph": "s", "id": 91551, "pid": 76337, "tid": -914061504, "ts": 1716454222755392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810097, "dur": 159,
    "args": {
      "External id": 91553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91553, "pid": 5, "tid": 7, "ts": 1716454222810097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755407, "dur": 5,
    "args": {
      "External id": 91553,
      "cbid": 211, "correlation": 91553
    }
  },
  {
    "ph": "s", "id": 91553, "pid": 76337, "tid": -914061504, "ts": 1716454222755407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222810258, "dur": 1,
    "args": {
      "External id": 91555,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91555,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 91555, "pid": 5, "tid": 7, "ts": 1716454222810258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755418, "dur": 6,
    "args": {
      "External id": 91555,
      "cbid": 51, "correlation": 91555
    }
  },
  {
    "ph": "s", "id": 91555, "pid": 76337, "tid": -914061504, "ts": 1716454222755418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222810261, "dur": 254,
    "args": {
      "External id": 91556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91556,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91556, "pid": 5, "tid": 7, "ts": 1716454222810261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755425, "dur": 6,
    "args": {
      "External id": 91556,
      "cbid": 211, "correlation": 91556
    }
  },
  {
    "ph": "s", "id": 91556, "pid": 76337, "tid": -914061504, "ts": 1716454222755425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810517, "dur": 6,
    "args": {
      "External id": 91558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91558, "pid": 5, "tid": 7, "ts": 1716454222810517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755435, "dur": 6,
    "args": {
      "External id": 91558,
      "cbid": 211, "correlation": 91558
    }
  },
  {
    "ph": "s", "id": 91558, "pid": 76337, "tid": -914061504, "ts": 1716454222755435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222810524, "dur": 6,
    "args": {
      "External id": 91564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91564, "pid": 5, "tid": 7, "ts": 1716454222810524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755464, "dur": 8,
    "args": {
      "External id": 91564,
      "cbid": 211, "correlation": 91564
    }
  },
  {
    "ph": "s", "id": 91564, "pid": 76337, "tid": -914061504, "ts": 1716454222755464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755523, "dur": 0,
    "args": {
      "External id": 91574,
      "cbid": 317, "correlation": 91574
    }
  },
  {
    "ph": "f", "id": 91574, "pid": 76337, "tid": -914061504, "ts": 1716454222755523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755523, "dur": 0,
    "args": {
      "External id": 91575,
      "cbid": 203, "correlation": 91575
    }
  },
  {
    "ph": "f", "id": 91575, "pid": 76337, "tid": -914061504, "ts": 1716454222755523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755524, "dur": 0,
    "args": {
      "External id": 91576,
      "cbid": 205, "correlation": 91576
    }
  },
  {
    "ph": "f", "id": 91576, "pid": 76337, "tid": -914061504, "ts": 1716454222755524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810531, "dur": 8,
    "args": {
      "External id": 91580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91580,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91580, "pid": 5, "tid": 7, "ts": 1716454222810531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755538, "dur": 12,
    "args": {
      "External id": 91580,
      "cbid": 211, "correlation": 91580
    }
  },
  {
    "ph": "s", "id": 91580, "pid": 76337, "tid": -914061504, "ts": 1716454222755538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222810540, "dur": 3,
    "args": {
      "External id": 91582,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91582,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.530612244897959
    }
  },
  {
    "ph": "f", "id": 91582, "pid": 5, "tid": 7, "ts": 1716454222810540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755557, "dur": 16,
    "args": {
      "External id": 91582,
      "cbid": 51, "correlation": 91582
    }
  },
  {
    "ph": "s", "id": 91582, "pid": 76337, "tid": -914061504, "ts": 1716454222755557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222810544, "dur": 95,
    "args": {
      "External id": 91583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91583,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 91583, "pid": 5, "tid": 7, "ts": 1716454222810544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755574, "dur": 7,
    "args": {
      "External id": 91583,
      "cbid": 211, "correlation": 91583
    }
  },
  {
    "ph": "s", "id": 91583, "pid": 76337, "tid": -914061504, "ts": 1716454222755574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810641, "dur": 6,
    "args": {
      "External id": 91585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91585, "pid": 5, "tid": 7, "ts": 1716454222810641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755585, "dur": 5,
    "args": {
      "External id": 91585,
      "cbid": 211, "correlation": 91585
    }
  },
  {
    "ph": "s", "id": 91585, "pid": 76337, "tid": -914061504, "ts": 1716454222755585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222810648, "dur": 6,
    "args": {
      "External id": 91591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91591,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91591, "pid": 5, "tid": 7, "ts": 1716454222810648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755613, "dur": 8,
    "args": {
      "External id": 91591,
      "cbid": 211, "correlation": 91591
    }
  },
  {
    "ph": "s", "id": 91591, "pid": 76337, "tid": -914061504, "ts": 1716454222755613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222810655, "dur": 5,
    "args": {
      "External id": 91599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91599,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91599, "pid": 5, "tid": 7, "ts": 1716454222810655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755641, "dur": 8,
    "args": {
      "External id": 91599,
      "cbid": 211, "correlation": 91599
    }
  },
  {
    "ph": "s", "id": 91599, "pid": 76337, "tid": -914061504, "ts": 1716454222755641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222810661, "dur": 4,
    "args": {
      "External id": 91607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91607,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91607, "pid": 5, "tid": 7, "ts": 1716454222810661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755670, "dur": 8,
    "args": {
      "External id": 91607,
      "cbid": 211, "correlation": 91607
    }
  },
  {
    "ph": "s", "id": 91607, "pid": 76337, "tid": -914061504, "ts": 1716454222755670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222810667, "dur": 11,
    "args": {
      "External id": 91616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91616,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91616, "pid": 5, "tid": 7, "ts": 1716454222810667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755715, "dur": 11,
    "args": {
      "External id": 91616,
      "cbid": 211, "correlation": 91616
    }
  },
  {
    "ph": "s", "id": 91616, "pid": 76337, "tid": -914061504, "ts": 1716454222755715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810679, "dur": 12,
    "args": {
      "External id": 91636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91636,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91636, "pid": 5, "tid": 7, "ts": 1716454222810679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755786, "dur": 11,
    "args": {
      "External id": 91636,
      "cbid": 211, "correlation": 91636
    }
  },
  {
    "ph": "s", "id": 91636, "pid": 76337, "tid": -914061504, "ts": 1716454222755786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222810693, "dur": 4,
    "args": {
      "External id": 91648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91648,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91648, "pid": 5, "tid": 7, "ts": 1716454222810693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755807, "dur": 6,
    "args": {
      "External id": 91648,
      "cbid": 211, "correlation": 91648
    }
  },
  {
    "ph": "s", "id": 91648, "pid": 76337, "tid": -914061504, "ts": 1716454222755807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222810698, "dur": 11,
    "args": {
      "External id": 91651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91651,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91651, "pid": 5, "tid": 7, "ts": 1716454222810698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755825, "dur": 7,
    "args": {
      "External id": 91651,
      "cbid": 211, "correlation": 91651
    }
  },
  {
    "ph": "s", "id": 91651, "pid": 76337, "tid": -914061504, "ts": 1716454222755825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222810710, "dur": 6,
    "args": {
      "External id": 91660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91660,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91660, "pid": 5, "tid": 7, "ts": 1716454222810710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755863, "dur": 9,
    "args": {
      "External id": 91660,
      "cbid": 211, "correlation": 91660
    }
  },
  {
    "ph": "s", "id": 91660, "pid": 76337, "tid": -914061504, "ts": 1716454222755863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755914, "dur": 0,
    "args": {
      "External id": 91670,
      "cbid": 317, "correlation": 91670
    }
  },
  {
    "ph": "f", "id": 91670, "pid": 76337, "tid": -914061504, "ts": 1716454222755914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755915, "dur": 0,
    "args": {
      "External id": 91671,
      "cbid": 203, "correlation": 91671
    }
  },
  {
    "ph": "f", "id": 91671, "pid": 76337, "tid": -914061504, "ts": 1716454222755915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755916, "dur": 0,
    "args": {
      "External id": 91672,
      "cbid": 205, "correlation": 91672
    }
  },
  {
    "ph": "f", "id": 91672, "pid": 76337, "tid": -914061504, "ts": 1716454222755916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810718, "dur": 6,
    "args": {
      "External id": 91676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91676,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91676, "pid": 5, "tid": 7, "ts": 1716454222810718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755930, "dur": 11,
    "args": {
      "External id": 91676,
      "cbid": 211, "correlation": 91676
    }
  },
  {
    "ph": "s", "id": 91676, "pid": 76337, "tid": -914061504, "ts": 1716454222755930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222810725, "dur": 315,
    "args": {
      "External id": 91678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91678, "pid": 5, "tid": 7, "ts": 1716454222810725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755944, "dur": 6,
    "args": {
      "External id": 91678,
      "cbid": 211, "correlation": 91678
    }
  },
  {
    "ph": "s", "id": 91678, "pid": 76337, "tid": -914061504, "ts": 1716454222755944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222811043, "dur": 1,
    "args": {
      "External id": 91680,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91680,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 91680, "pid": 5, "tid": 7, "ts": 1716454222811043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755955, "dur": 6,
    "args": {
      "External id": 91680,
      "cbid": 51, "correlation": 91680
    }
  },
  {
    "ph": "s", "id": 91680, "pid": 76337, "tid": -914061504, "ts": 1716454222755955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222811046, "dur": 487,
    "args": {
      "External id": 91681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91681,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91681, "pid": 5, "tid": 7, "ts": 1716454222811046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755963, "dur": 6,
    "args": {
      "External id": 91681,
      "cbid": 211, "correlation": 91681
    }
  },
  {
    "ph": "s", "id": 91681, "pid": 76337, "tid": -914061504, "ts": 1716454222755963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222811535, "dur": 6,
    "args": {
      "External id": 91683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91683,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91683, "pid": 5, "tid": 7, "ts": 1716454222811535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222755972, "dur": 14,
    "args": {
      "External id": 91683,
      "cbid": 211, "correlation": 91683
    }
  },
  {
    "ph": "s", "id": 91683, "pid": 76337, "tid": -914061504, "ts": 1716454222755972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222811542, "dur": 6,
    "args": {
      "External id": 91689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91689, "pid": 5, "tid": 7, "ts": 1716454222811542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756011, "dur": 9,
    "args": {
      "External id": 91689,
      "cbid": 211, "correlation": 91689
    }
  },
  {
    "ph": "s", "id": 91689, "pid": 76337, "tid": -914061504, "ts": 1716454222756011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222811549, "dur": 3,
    "args": {
      "External id": 91697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91697,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 91697, "pid": 5, "tid": 7, "ts": 1716454222811549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756055, "dur": 10,
    "args": {
      "External id": 91697,
      "cbid": 211, "correlation": 91697
    }
  },
  {
    "ph": "s", "id": 91697, "pid": 76337, "tid": -914061504, "ts": 1716454222756055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756119, "dur": 1,
    "args": {
      "External id": 91713,
      "cbid": 251, "correlation": 91713
    }
  },
  {
    "ph": "f", "id": 91713, "pid": 76337, "tid": -914061504, "ts": 1716454222756119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756125, "dur": 0,
    "args": {
      "External id": 91715,
      "cbid": 251, "correlation": 91715
    }
  },
  {
    "ph": "f", "id": 91715, "pid": 76337, "tid": -914061504, "ts": 1716454222756125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222811554, "dur": 12,
    "args": {
      "External id": 91716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91716,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91716, "pid": 5, "tid": 7, "ts": 1716454222811554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756127, "dur": 11,
    "args": {
      "External id": 91716,
      "cbid": 211, "correlation": 91716
    }
  },
  {
    "ph": "s", "id": 91716, "pid": 76337, "tid": -914061504, "ts": 1716454222756127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222811567, "dur": 5,
    "args": {
      "External id": 91718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91718,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91718, "pid": 5, "tid": 7, "ts": 1716454222811567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756140, "dur": 6,
    "args": {
      "External id": 91718,
      "cbid": 211, "correlation": 91718
    }
  },
  {
    "ph": "s", "id": 91718, "pid": 76337, "tid": -914061504, "ts": 1716454222756140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222811574, "dur": 6,
    "args": {
      "External id": 91728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91728,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91728, "pid": 5, "tid": 7, "ts": 1716454222811574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756197, "dur": 12,
    "args": {
      "External id": 91728,
      "cbid": 211, "correlation": 91728
    }
  },
  {
    "ph": "s", "id": 91728, "pid": 76337, "tid": -914061504, "ts": 1716454222756197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222811581, "dur": 9,
    "args": {
      "External id": 91748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91748,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91748, "pid": 5, "tid": 7, "ts": 1716454222811581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756262, "dur": 11,
    "args": {
      "External id": 91748,
      "cbid": 211, "correlation": 91748
    }
  },
  {
    "ph": "s", "id": 91748, "pid": 76337, "tid": -914061504, "ts": 1716454222756262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222811592, "dur": 4,
    "args": {
      "External id": 91760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91760,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 91760, "pid": 5, "tid": 7, "ts": 1716454222811592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756283, "dur": 6,
    "args": {
      "External id": 91760,
      "cbid": 211, "correlation": 91760
    }
  },
  {
    "ph": "s", "id": 91760, "pid": 76337, "tid": -914061504, "ts": 1716454222756283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222811597, "dur": 6,
    "args": {
      "External id": 91763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91763, "pid": 5, "tid": 7, "ts": 1716454222811597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756303, "dur": 7,
    "args": {
      "External id": 91763,
      "cbid": 211, "correlation": 91763
    }
  },
  {
    "ph": "s", "id": 91763, "pid": 76337, "tid": -914061504, "ts": 1716454222756303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222811605, "dur": 4,
    "args": {
      "External id": 91772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91772,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91772, "pid": 5, "tid": 7, "ts": 1716454222811605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756344, "dur": 10,
    "args": {
      "External id": 91772,
      "cbid": 211, "correlation": 91772
    }
  },
  {
    "ph": "s", "id": 91772, "pid": 76337, "tid": -914061504, "ts": 1716454222756344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756407, "dur": 0,
    "args": {
      "External id": 91782,
      "cbid": 317, "correlation": 91782
    }
  },
  {
    "ph": "f", "id": 91782, "pid": 76337, "tid": -914061504, "ts": 1716454222756407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756408, "dur": 0,
    "args": {
      "External id": 91783,
      "cbid": 203, "correlation": 91783
    }
  },
  {
    "ph": "f", "id": 91783, "pid": 76337, "tid": -914061504, "ts": 1716454222756408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756408, "dur": 0,
    "args": {
      "External id": 91784,
      "cbid": 205, "correlation": 91784
    }
  },
  {
    "ph": "f", "id": 91784, "pid": 76337, "tid": -914061504, "ts": 1716454222756408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222811611, "dur": 5,
    "args": {
      "External id": 91788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91788, "pid": 5, "tid": 7, "ts": 1716454222811611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756423, "dur": 12,
    "args": {
      "External id": 91788,
      "cbid": 211, "correlation": 91788
    }
  },
  {
    "ph": "s", "id": 91788, "pid": 76337, "tid": -914061504, "ts": 1716454222756423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222811617, "dur": 159,
    "args": {
      "External id": 91790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91790, "pid": 5, "tid": 7, "ts": 1716454222811617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756437, "dur": 6,
    "args": {
      "External id": 91790,
      "cbid": 211, "correlation": 91790
    }
  },
  {
    "ph": "s", "id": 91790, "pid": 76337, "tid": -914061504, "ts": 1716454222756437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222811778, "dur": 1,
    "args": {
      "External id": 91792,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91792,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 91792, "pid": 5, "tid": 7, "ts": 1716454222811778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756449, "dur": 6,
    "args": {
      "External id": 91792,
      "cbid": 51, "correlation": 91792
    }
  },
  {
    "ph": "s", "id": 91792, "pid": 76337, "tid": -914061504, "ts": 1716454222756449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222811781, "dur": 254,
    "args": {
      "External id": 91793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91793,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91793, "pid": 5, "tid": 7, "ts": 1716454222811781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756456, "dur": 6,
    "args": {
      "External id": 91793,
      "cbid": 211, "correlation": 91793
    }
  },
  {
    "ph": "s", "id": 91793, "pid": 76337, "tid": -914061504, "ts": 1716454222756456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222812037, "dur": 6,
    "args": {
      "External id": 91795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91795,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91795, "pid": 5, "tid": 7, "ts": 1716454222812037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756466, "dur": 5,
    "args": {
      "External id": 91795,
      "cbid": 211, "correlation": 91795
    }
  },
  {
    "ph": "s", "id": 91795, "pid": 76337, "tid": -914061504, "ts": 1716454222756466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222812044, "dur": 6,
    "args": {
      "External id": 91801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91801, "pid": 5, "tid": 7, "ts": 1716454222812044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756495, "dur": 8,
    "args": {
      "External id": 91801,
      "cbid": 211, "correlation": 91801
    }
  },
  {
    "ph": "s", "id": 91801, "pid": 76337, "tid": -914061504, "ts": 1716454222756495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756554, "dur": 0,
    "args": {
      "External id": 91811,
      "cbid": 317, "correlation": 91811
    }
  },
  {
    "ph": "f", "id": 91811, "pid": 76337, "tid": -914061504, "ts": 1716454222756554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756555, "dur": 0,
    "args": {
      "External id": 91812,
      "cbid": 203, "correlation": 91812
    }
  },
  {
    "ph": "f", "id": 91812, "pid": 76337, "tid": -914061504, "ts": 1716454222756555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756556, "dur": 0,
    "args": {
      "External id": 91813,
      "cbid": 205, "correlation": 91813
    }
  },
  {
    "ph": "f", "id": 91813, "pid": 76337, "tid": -914061504, "ts": 1716454222756556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222812051, "dur": 7,
    "args": {
      "External id": 91817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91817,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91817, "pid": 5, "tid": 7, "ts": 1716454222812051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756567, "dur": 11,
    "args": {
      "External id": 91817,
      "cbid": 211, "correlation": 91817
    }
  },
  {
    "ph": "s", "id": 91817, "pid": 76337, "tid": -914061504, "ts": 1716454222756567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222812060, "dur": 3,
    "args": {
      "External id": 91819,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91819,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5151515151515151
    }
  },
  {
    "ph": "f", "id": 91819, "pid": 5, "tid": 7, "ts": 1716454222812060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756584, "dur": 11,
    "args": {
      "External id": 91819,
      "cbid": 51, "correlation": 91819
    }
  },
  {
    "ph": "s", "id": 91819, "pid": 76337, "tid": -914061504, "ts": 1716454222756584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222812064, "dur": 95,
    "args": {
      "External id": 91820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91820,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 91820, "pid": 5, "tid": 7, "ts": 1716454222812064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756597, "dur": 7,
    "args": {
      "External id": 91820,
      "cbid": 211, "correlation": 91820
    }
  },
  {
    "ph": "s", "id": 91820, "pid": 76337, "tid": -914061504, "ts": 1716454222756597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222812160, "dur": 6,
    "args": {
      "External id": 91822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91822,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91822, "pid": 5, "tid": 7, "ts": 1716454222812160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756607, "dur": 5,
    "args": {
      "External id": 91822,
      "cbid": 211, "correlation": 91822
    }
  },
  {
    "ph": "s", "id": 91822, "pid": 76337, "tid": -914061504, "ts": 1716454222756607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222812167, "dur": 6,
    "args": {
      "External id": 91828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91828,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91828, "pid": 5, "tid": 7, "ts": 1716454222812167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756633, "dur": 10,
    "args": {
      "External id": 91828,
      "cbid": 211, "correlation": 91828
    }
  },
  {
    "ph": "s", "id": 91828, "pid": 76337, "tid": -914061504, "ts": 1716454222756633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222812175, "dur": 5,
    "args": {
      "External id": 91836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91836,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91836, "pid": 5, "tid": 7, "ts": 1716454222812175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756665, "dur": 9,
    "args": {
      "External id": 91836,
      "cbid": 211, "correlation": 91836
    }
  },
  {
    "ph": "s", "id": 91836, "pid": 76337, "tid": -914061504, "ts": 1716454222756665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222812181, "dur": 4,
    "args": {
      "External id": 91844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91844,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91844, "pid": 5, "tid": 7, "ts": 1716454222812181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756695, "dur": 8,
    "args": {
      "External id": 91844,
      "cbid": 211, "correlation": 91844
    }
  },
  {
    "ph": "s", "id": 91844, "pid": 76337, "tid": -914061504, "ts": 1716454222756695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222812186, "dur": 11,
    "args": {
      "External id": 91853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91853,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91853, "pid": 5, "tid": 7, "ts": 1716454222812186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756740, "dur": 11,
    "args": {
      "External id": 91853,
      "cbid": 211, "correlation": 91853
    }
  },
  {
    "ph": "s", "id": 91853, "pid": 76337, "tid": -914061504, "ts": 1716454222756740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222812198, "dur": 12,
    "args": {
      "External id": 91873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91873,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91873, "pid": 5, "tid": 7, "ts": 1716454222812198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756810, "dur": 12,
    "args": {
      "External id": 91873,
      "cbid": 211, "correlation": 91873
    }
  },
  {
    "ph": "s", "id": 91873, "pid": 76337, "tid": -914061504, "ts": 1716454222756810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222812212, "dur": 4,
    "args": {
      "External id": 91885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91885,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91885, "pid": 5, "tid": 7, "ts": 1716454222812212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756832, "dur": 6,
    "args": {
      "External id": 91885,
      "cbid": 211, "correlation": 91885
    }
  },
  {
    "ph": "s", "id": 91885, "pid": 76337, "tid": -914061504, "ts": 1716454222756832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222812217, "dur": 11,
    "args": {
      "External id": 91888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91888, "pid": 5, "tid": 7, "ts": 1716454222812217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756850, "dur": 7,
    "args": {
      "External id": 91888,
      "cbid": 211, "correlation": 91888
    }
  },
  {
    "ph": "s", "id": 91888, "pid": 76337, "tid": -914061504, "ts": 1716454222756850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222812230, "dur": 6,
    "args": {
      "External id": 91897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91897,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91897, "pid": 5, "tid": 7, "ts": 1716454222812230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756888, "dur": 10,
    "args": {
      "External id": 91897,
      "cbid": 211, "correlation": 91897
    }
  },
  {
    "ph": "s", "id": 91897, "pid": 76337, "tid": -914061504, "ts": 1716454222756888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756940, "dur": 0,
    "args": {
      "External id": 91907,
      "cbid": 317, "correlation": 91907
    }
  },
  {
    "ph": "f", "id": 91907, "pid": 76337, "tid": -914061504, "ts": 1716454222756940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756941, "dur": 0,
    "args": {
      "External id": 91908,
      "cbid": 203, "correlation": 91908
    }
  },
  {
    "ph": "f", "id": 91908, "pid": 76337, "tid": -914061504, "ts": 1716454222756941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756942, "dur": 0,
    "args": {
      "External id": 91909,
      "cbid": 205, "correlation": 91909
    }
  },
  {
    "ph": "f", "id": 91909, "pid": 76337, "tid": -914061504, "ts": 1716454222756942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222812237, "dur": 6,
    "args": {
      "External id": 91913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91913, "pid": 5, "tid": 7, "ts": 1716454222812237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756955, "dur": 11,
    "args": {
      "External id": 91913,
      "cbid": 211, "correlation": 91913
    }
  },
  {
    "ph": "s", "id": 91913, "pid": 76337, "tid": -914061504, "ts": 1716454222756955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222812245, "dur": 314,
    "args": {
      "External id": 91915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91915, "pid": 5, "tid": 7, "ts": 1716454222812245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756969, "dur": 13,
    "args": {
      "External id": 91915,
      "cbid": 211, "correlation": 91915
    }
  },
  {
    "ph": "s", "id": 91915, "pid": 76337, "tid": -914061504, "ts": 1716454222756969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222812561, "dur": 1,
    "args": {
      "External id": 91917,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 91917,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 91917, "pid": 5, "tid": 7, "ts": 1716454222812561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756987, "dur": 7,
    "args": {
      "External id": 91917,
      "cbid": 51, "correlation": 91917
    }
  },
  {
    "ph": "s", "id": 91917, "pid": 76337, "tid": -914061504, "ts": 1716454222756987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222812564, "dur": 489,
    "args": {
      "External id": 91918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91918,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91918, "pid": 5, "tid": 7, "ts": 1716454222812564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222756995, "dur": 7,
    "args": {
      "External id": 91918,
      "cbid": 211, "correlation": 91918
    }
  },
  {
    "ph": "s", "id": 91918, "pid": 76337, "tid": -914061504, "ts": 1716454222756995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813055, "dur": 6,
    "args": {
      "External id": 91920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91920,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 91920, "pid": 5, "tid": 7, "ts": 1716454222813055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757006, "dur": 5,
    "args": {
      "External id": 91920,
      "cbid": 211, "correlation": 91920
    }
  },
  {
    "ph": "s", "id": 91920, "pid": 76337, "tid": -914061504, "ts": 1716454222757006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222813062, "dur": 6,
    "args": {
      "External id": 91926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91926, "pid": 5, "tid": 7, "ts": 1716454222813062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757035, "dur": 10,
    "args": {
      "External id": 91926,
      "cbid": 211, "correlation": 91926
    }
  },
  {
    "ph": "s", "id": 91926, "pid": 76337, "tid": -914061504, "ts": 1716454222757035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222813069, "dur": 3,
    "args": {
      "External id": 91934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91934,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 91934, "pid": 5, "tid": 7, "ts": 1716454222813069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757079, "dur": 9,
    "args": {
      "External id": 91934,
      "cbid": 211, "correlation": 91934
    }
  },
  {
    "ph": "s", "id": 91934, "pid": 76337, "tid": -914061504, "ts": 1716454222757079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757142, "dur": 1,
    "args": {
      "External id": 91950,
      "cbid": 251, "correlation": 91950
    }
  },
  {
    "ph": "f", "id": 91950, "pid": 76337, "tid": -914061504, "ts": 1716454222757142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757148, "dur": 0,
    "args": {
      "External id": 91952,
      "cbid": 251, "correlation": 91952
    }
  },
  {
    "ph": "f", "id": 91952, "pid": 76337, "tid": -914061504, "ts": 1716454222757148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222813073, "dur": 13,
    "args": {
      "External id": 91953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91953,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91953, "pid": 5, "tid": 7, "ts": 1716454222813073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757150, "dur": 11,
    "args": {
      "External id": 91953,
      "cbid": 211, "correlation": 91953
    }
  },
  {
    "ph": "s", "id": 91953, "pid": 76337, "tid": -914061504, "ts": 1716454222757150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222813088, "dur": 5,
    "args": {
      "External id": 91955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91955,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 91955, "pid": 5, "tid": 7, "ts": 1716454222813088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757163, "dur": 5,
    "args": {
      "External id": 91955,
      "cbid": 211, "correlation": 91955
    }
  },
  {
    "ph": "s", "id": 91955, "pid": 76337, "tid": -914061504, "ts": 1716454222757163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222813094, "dur": 6,
    "args": {
      "External id": 91965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 91965, "pid": 5, "tid": 7, "ts": 1716454222813094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757220, "dur": 12,
    "args": {
      "External id": 91965,
      "cbid": 211, "correlation": 91965
    }
  },
  {
    "ph": "s", "id": 91965, "pid": 76337, "tid": -914061504, "ts": 1716454222757220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813101, "dur": 9,
    "args": {
      "External id": 91985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91985,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 91985, "pid": 5, "tid": 7, "ts": 1716454222813101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757286, "dur": 11,
    "args": {
      "External id": 91985,
      "cbid": 211, "correlation": 91985
    }
  },
  {
    "ph": "s", "id": 91985, "pid": 76337, "tid": -914061504, "ts": 1716454222757286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222813111, "dur": 3,
    "args": {
      "External id": 91997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 91997,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 91997, "pid": 5, "tid": 7, "ts": 1716454222813111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757307, "dur": 7,
    "args": {
      "External id": 91997,
      "cbid": 211, "correlation": 91997
    }
  },
  {
    "ph": "s", "id": 91997, "pid": 76337, "tid": -914061504, "ts": 1716454222757307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222813116, "dur": 6,
    "args": {
      "External id": 92000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92000,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 92000, "pid": 5, "tid": 7, "ts": 1716454222813116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757327, "dur": 7,
    "args": {
      "External id": 92000,
      "cbid": 211, "correlation": 92000
    }
  },
  {
    "ph": "s", "id": 92000, "pid": 76337, "tid": -914061504, "ts": 1716454222757327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222813124, "dur": 4,
    "args": {
      "External id": 92009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92009,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 92009, "pid": 5, "tid": 7, "ts": 1716454222813124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757369, "dur": 10,
    "args": {
      "External id": 92009,
      "cbid": 211, "correlation": 92009
    }
  },
  {
    "ph": "s", "id": 92009, "pid": 76337, "tid": -914061504, "ts": 1716454222757369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757433, "dur": 0,
    "args": {
      "External id": 92019,
      "cbid": 317, "correlation": 92019
    }
  },
  {
    "ph": "f", "id": 92019, "pid": 76337, "tid": -914061504, "ts": 1716454222757433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757434, "dur": 0,
    "args": {
      "External id": 92020,
      "cbid": 203, "correlation": 92020
    }
  },
  {
    "ph": "f", "id": 92020, "pid": 76337, "tid": -914061504, "ts": 1716454222757434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757435, "dur": 0,
    "args": {
      "External id": 92021,
      "cbid": 205, "correlation": 92021
    }
  },
  {
    "ph": "f", "id": 92021, "pid": 76337, "tid": -914061504, "ts": 1716454222757435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813130, "dur": 5,
    "args": {
      "External id": 92025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92025,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92025, "pid": 5, "tid": 7, "ts": 1716454222813130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757449, "dur": 12,
    "args": {
      "External id": 92025,
      "cbid": 211, "correlation": 92025
    }
  },
  {
    "ph": "s", "id": 92025, "pid": 76337, "tid": -914061504, "ts": 1716454222757449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813136, "dur": 160,
    "args": {
      "External id": 92027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92027,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92027, "pid": 5, "tid": 7, "ts": 1716454222813136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757463, "dur": 5,
    "args": {
      "External id": 92027,
      "cbid": 211, "correlation": 92027
    }
  },
  {
    "ph": "s", "id": 92027, "pid": 76337, "tid": -914061504, "ts": 1716454222757463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222813298, "dur": 1,
    "args": {
      "External id": 92029,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92029,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 92029, "pid": 5, "tid": 7, "ts": 1716454222813298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757474, "dur": 7,
    "args": {
      "External id": 92029,
      "cbid": 51, "correlation": 92029
    }
  },
  {
    "ph": "s", "id": 92029, "pid": 76337, "tid": -914061504, "ts": 1716454222757474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222813301, "dur": 254,
    "args": {
      "External id": 92030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92030,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92030, "pid": 5, "tid": 7, "ts": 1716454222813301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757482, "dur": 6,
    "args": {
      "External id": 92030,
      "cbid": 211, "correlation": 92030
    }
  },
  {
    "ph": "s", "id": 92030, "pid": 76337, "tid": -914061504, "ts": 1716454222757482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813557, "dur": 6,
    "args": {
      "External id": 92032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92032,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92032, "pid": 5, "tid": 7, "ts": 1716454222813557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757492, "dur": 5,
    "args": {
      "External id": 92032,
      "cbid": 211, "correlation": 92032
    }
  },
  {
    "ph": "s", "id": 92032, "pid": 76337, "tid": -914061504, "ts": 1716454222757492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222813564, "dur": 6,
    "args": {
      "External id": 92038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 92038, "pid": 5, "tid": 7, "ts": 1716454222813564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757520, "dur": 9,
    "args": {
      "External id": 92038,
      "cbid": 211, "correlation": 92038
    }
  },
  {
    "ph": "s", "id": 92038, "pid": 76337, "tid": -914061504, "ts": 1716454222757520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757578, "dur": 0,
    "args": {
      "External id": 92048,
      "cbid": 317, "correlation": 92048
    }
  },
  {
    "ph": "f", "id": 92048, "pid": 76337, "tid": -914061504, "ts": 1716454222757578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757579, "dur": 0,
    "args": {
      "External id": 92049,
      "cbid": 203, "correlation": 92049
    }
  },
  {
    "ph": "f", "id": 92049, "pid": 76337, "tid": -914061504, "ts": 1716454222757579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757580, "dur": 0,
    "args": {
      "External id": 92050,
      "cbid": 205, "correlation": 92050
    }
  },
  {
    "ph": "f", "id": 92050, "pid": 76337, "tid": -914061504, "ts": 1716454222757580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813571, "dur": 8,
    "args": {
      "External id": 92054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92054, "pid": 5, "tid": 7, "ts": 1716454222813571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757592, "dur": 11,
    "args": {
      "External id": 92054,
      "cbid": 211, "correlation": 92054
    }
  },
  {
    "ph": "s", "id": 92054, "pid": 76337, "tid": -914061504, "ts": 1716454222757592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222813580, "dur": 3,
    "args": {
      "External id": 92056,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92056,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.530612244897959
    }
  },
  {
    "ph": "f", "id": 92056, "pid": 5, "tid": 7, "ts": 1716454222813580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757608, "dur": 10,
    "args": {
      "External id": 92056,
      "cbid": 51, "correlation": 92056
    }
  },
  {
    "ph": "s", "id": 92056, "pid": 76337, "tid": -914061504, "ts": 1716454222757608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222813584, "dur": 92,
    "args": {
      "External id": 92057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92057,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 92057, "pid": 5, "tid": 7, "ts": 1716454222813584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757619, "dur": 6,
    "args": {
      "External id": 92057,
      "cbid": 211, "correlation": 92057
    }
  },
  {
    "ph": "s", "id": 92057, "pid": 76337, "tid": -914061504, "ts": 1716454222757619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813678, "dur": 6,
    "args": {
      "External id": 92059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92059,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92059, "pid": 5, "tid": 7, "ts": 1716454222813678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757629, "dur": 5,
    "args": {
      "External id": 92059,
      "cbid": 211, "correlation": 92059
    }
  },
  {
    "ph": "s", "id": 92059, "pid": 76337, "tid": -914061504, "ts": 1716454222757629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222813685, "dur": 6,
    "args": {
      "External id": 92065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 92065, "pid": 5, "tid": 7, "ts": 1716454222813685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757656, "dur": 8,
    "args": {
      "External id": 92065,
      "cbid": 211, "correlation": 92065
    }
  },
  {
    "ph": "s", "id": 92065, "pid": 76337, "tid": -914061504, "ts": 1716454222757656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222813692, "dur": 5,
    "args": {
      "External id": 92073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92073,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 92073, "pid": 5, "tid": 7, "ts": 1716454222813692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757685, "dur": 8,
    "args": {
      "External id": 92073,
      "cbid": 211, "correlation": 92073
    }
  },
  {
    "ph": "s", "id": 92073, "pid": 76337, "tid": -914061504, "ts": 1716454222757685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222813699, "dur": 4,
    "args": {
      "External id": 92081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92081,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 92081, "pid": 5, "tid": 7, "ts": 1716454222813699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757715, "dur": 8,
    "args": {
      "External id": 92081,
      "cbid": 211, "correlation": 92081
    }
  },
  {
    "ph": "s", "id": 92081, "pid": 76337, "tid": -914061504, "ts": 1716454222757715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222813704, "dur": 14,
    "args": {
      "External id": 92092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92092,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92092, "pid": 5, "tid": 7, "ts": 1716454222813704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757798, "dur": 14,
    "args": {
      "External id": 92092,
      "cbid": 211, "correlation": 92092
    }
  },
  {
    "ph": "s", "id": 92092, "pid": 76337, "tid": -914061504, "ts": 1716454222757798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757855, "dur": 0,
    "args": {
      "External id": 92102,
      "cbid": 317, "correlation": 92102
    }
  },
  {
    "ph": "f", "id": 92102, "pid": 76337, "tid": -914061504, "ts": 1716454222757855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757856, "dur": 0,
    "args": {
      "External id": 92103,
      "cbid": 203, "correlation": 92103
    }
  },
  {
    "ph": "f", "id": 92103, "pid": 76337, "tid": -914061504, "ts": 1716454222757856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757857, "dur": 0,
    "args": {
      "External id": 92104,
      "cbid": 205, "correlation": 92104
    }
  },
  {
    "ph": "f", "id": 92104, "pid": 76337, "tid": -914061504, "ts": 1716454222757857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813719, "dur": 8,
    "args": {
      "External id": 92108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92108,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92108, "pid": 5, "tid": 7, "ts": 1716454222813719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757872, "dur": 12,
    "args": {
      "External id": 92108,
      "cbid": 211, "correlation": 92108
    }
  },
  {
    "ph": "s", "id": 92108, "pid": 76337, "tid": -914061504, "ts": 1716454222757872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222813729, "dur": 160,
    "args": {
      "External id": 92110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92110, "pid": 5, "tid": 7, "ts": 1716454222813729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757887, "dur": 5,
    "args": {
      "External id": 92110,
      "cbid": 211, "correlation": 92110
    }
  },
  {
    "ph": "s", "id": 92110, "pid": 76337, "tid": -914061504, "ts": 1716454222757887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222813891, "dur": 1,
    "args": {
      "External id": 92112,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92112,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 92112, "pid": 5, "tid": 7, "ts": 1716454222813891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757898, "dur": 6,
    "args": {
      "External id": 92112,
      "cbid": 51, "correlation": 92112
    }
  },
  {
    "ph": "s", "id": 92112, "pid": 76337, "tid": -914061504, "ts": 1716454222757898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222813895, "dur": 642,
    "args": {
      "External id": 92113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92113,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92113, "pid": 5, "tid": 7, "ts": 1716454222813895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757905, "dur": 6,
    "args": {
      "External id": 92113,
      "cbid": 211, "correlation": 92113
    }
  },
  {
    "ph": "s", "id": 92113, "pid": 76337, "tid": -914061504, "ts": 1716454222757905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222814538, "dur": 12,
    "args": {
      "External id": 92115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92115,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92115, "pid": 5, "tid": 7, "ts": 1716454222814538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757915, "dur": 6,
    "args": {
      "External id": 92115,
      "cbid": 211, "correlation": 92115
    }
  },
  {
    "ph": "s", "id": 92115, "pid": 76337, "tid": -914061504, "ts": 1716454222757915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222814552, "dur": 14,
    "args": {
      "External id": 92121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92121, "pid": 5, "tid": 7, "ts": 1716454222814552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222757944, "dur": 9,
    "args": {
      "External id": 92121,
      "cbid": 211, "correlation": 92121
    }
  },
  {
    "ph": "s", "id": 92121, "pid": 76337, "tid": -914061504, "ts": 1716454222757944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222814567, "dur": 29,
    "args": {
      "External id": 92130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92130,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92130, "pid": 5, "tid": 7, "ts": 1716454222814567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758046, "dur": 13,
    "args": {
      "External id": 92130,
      "cbid": 211, "correlation": 92130
    }
  },
  {
    "ph": "s", "id": 92130, "pid": 76337, "tid": -914061504, "ts": 1716454222758046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222814598, "dur": 29,
    "args": {
      "External id": 92150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92150,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 92150, "pid": 5, "tid": 7, "ts": 1716454222814598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758115, "dur": 11,
    "args": {
      "External id": 92150,
      "cbid": 211, "correlation": 92150
    }
  },
  {
    "ph": "s", "id": 92150, "pid": 76337, "tid": -914061504, "ts": 1716454222758115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222814629, "dur": 4,
    "args": {
      "External id": 92162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92162,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92162, "pid": 5, "tid": 7, "ts": 1716454222814629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758136, "dur": 7,
    "args": {
      "External id": 92162,
      "cbid": 211, "correlation": 92162
    }
  },
  {
    "ph": "s", "id": 92162, "pid": 76337, "tid": -914061504, "ts": 1716454222758136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222814634, "dur": 30,
    "args": {
      "External id": 92165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92165, "pid": 5, "tid": 7, "ts": 1716454222814634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758155, "dur": 6,
    "args": {
      "External id": 92165,
      "cbid": 211, "correlation": 92165
    }
  },
  {
    "ph": "s", "id": 92165, "pid": 76337, "tid": -914061504, "ts": 1716454222758155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222814665, "dur": 21,
    "args": {
      "External id": 92174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92174,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92174, "pid": 5, "tid": 7, "ts": 1716454222814665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758194, "dur": 10,
    "args": {
      "External id": 92174,
      "cbid": 211, "correlation": 92174
    }
  },
  {
    "ph": "s", "id": 92174, "pid": 76337, "tid": -914061504, "ts": 1716454222758194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758247, "dur": 0,
    "args": {
      "External id": 92184,
      "cbid": 317, "correlation": 92184
    }
  },
  {
    "ph": "f", "id": 92184, "pid": 76337, "tid": -914061504, "ts": 1716454222758247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758248, "dur": 0,
    "args": {
      "External id": 92185,
      "cbid": 203, "correlation": 92185
    }
  },
  {
    "ph": "f", "id": 92185, "pid": 76337, "tid": -914061504, "ts": 1716454222758248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758249, "dur": 0,
    "args": {
      "External id": 92186,
      "cbid": 205, "correlation": 92186
    }
  },
  {
    "ph": "f", "id": 92186, "pid": 76337, "tid": -914061504, "ts": 1716454222758249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222814687, "dur": 22,
    "args": {
      "External id": 92190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92190, "pid": 5, "tid": 7, "ts": 1716454222814687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758264, "dur": 11,
    "args": {
      "External id": 92190,
      "cbid": 211, "correlation": 92190
    }
  },
  {
    "ph": "s", "id": 92190, "pid": 76337, "tid": -914061504, "ts": 1716454222758264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222814711, "dur": 315,
    "args": {
      "External id": 92192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92192, "pid": 5, "tid": 7, "ts": 1716454222814711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758278, "dur": 5,
    "args": {
      "External id": 92192,
      "cbid": 211, "correlation": 92192
    }
  },
  {
    "ph": "s", "id": 92192, "pid": 76337, "tid": -914061504, "ts": 1716454222758278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222815027, "dur": 1,
    "args": {
      "External id": 92194,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92194,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 92194, "pid": 5, "tid": 7, "ts": 1716454222815027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758289, "dur": 6,
    "args": {
      "External id": 92194,
      "cbid": 51, "correlation": 92194
    }
  },
  {
    "ph": "s", "id": 92194, "pid": 76337, "tid": -914061504, "ts": 1716454222758289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222815031, "dur": 1227,
    "args": {
      "External id": 92195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92195,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92195, "pid": 5, "tid": 7, "ts": 1716454222815031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758296, "dur": 6,
    "args": {
      "External id": 92195,
      "cbid": 211, "correlation": 92195
    }
  },
  {
    "ph": "s", "id": 92195, "pid": 76337, "tid": -914061504, "ts": 1716454222758296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222816259, "dur": 12,
    "args": {
      "External id": 92197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92197,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92197, "pid": 5, "tid": 7, "ts": 1716454222816259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758306, "dur": 5,
    "args": {
      "External id": 92197,
      "cbid": 211, "correlation": 92197
    }
  },
  {
    "ph": "s", "id": 92197, "pid": 76337, "tid": -914061504, "ts": 1716454222758306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222816272, "dur": 15,
    "args": {
      "External id": 92203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92203, "pid": 5, "tid": 7, "ts": 1716454222816272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758335, "dur": 8,
    "args": {
      "External id": 92203,
      "cbid": 211, "correlation": 92203
    }
  },
  {
    "ph": "s", "id": 92203, "pid": 76337, "tid": -914061504, "ts": 1716454222758335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222816288, "dur": 3,
    "args": {
      "External id": 92211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92211,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 92211, "pid": 5, "tid": 7, "ts": 1716454222816288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758379, "dur": 10,
    "args": {
      "External id": 92211,
      "cbid": 211, "correlation": 92211
    }
  },
  {
    "ph": "s", "id": 92211, "pid": 76337, "tid": -914061504, "ts": 1716454222758379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758446, "dur": 1,
    "args": {
      "External id": 92227,
      "cbid": 251, "correlation": 92227
    }
  },
  {
    "ph": "f", "id": 92227, "pid": 76337, "tid": -914061504, "ts": 1716454222758446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758451, "dur": 0,
    "args": {
      "External id": 92229,
      "cbid": 251, "correlation": 92229
    }
  },
  {
    "ph": "f", "id": 92229, "pid": 76337, "tid": -914061504, "ts": 1716454222758451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222816293, "dur": 12,
    "args": {
      "External id": 92230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92230,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92230, "pid": 5, "tid": 7, "ts": 1716454222816293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758453, "dur": 12,
    "args": {
      "External id": 92230,
      "cbid": 211, "correlation": 92230
    }
  },
  {
    "ph": "s", "id": 92230, "pid": 76337, "tid": -914061504, "ts": 1716454222758453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222816306, "dur": 5,
    "args": {
      "External id": 92232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92232,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92232, "pid": 5, "tid": 7, "ts": 1716454222816306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758467, "dur": 6,
    "args": {
      "External id": 92232,
      "cbid": 211, "correlation": 92232
    }
  },
  {
    "ph": "s", "id": 92232, "pid": 76337, "tid": -914061504, "ts": 1716454222758467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222816312, "dur": 17,
    "args": {
      "External id": 92242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92242, "pid": 5, "tid": 7, "ts": 1716454222816312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758525, "dur": 13,
    "args": {
      "External id": 92242,
      "cbid": 211, "correlation": 92242
    }
  },
  {
    "ph": "s", "id": 92242, "pid": 76337, "tid": -914061504, "ts": 1716454222758525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222816330, "dur": 17,
    "args": {
      "External id": 92262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92262,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 92262, "pid": 5, "tid": 7, "ts": 1716454222816330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758592, "dur": 11,
    "args": {
      "External id": 92262,
      "cbid": 211, "correlation": 92262
    }
  },
  {
    "ph": "s", "id": 92262, "pid": 76337, "tid": -914061504, "ts": 1716454222758592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222816349, "dur": 4,
    "args": {
      "External id": 92274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92274,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 92274, "pid": 5, "tid": 7, "ts": 1716454222816349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758613, "dur": 6,
    "args": {
      "External id": 92274,
      "cbid": 211, "correlation": 92274
    }
  },
  {
    "ph": "s", "id": 92274, "pid": 76337, "tid": -914061504, "ts": 1716454222758613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222816354, "dur": 16,
    "args": {
      "External id": 92277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92277, "pid": 5, "tid": 7, "ts": 1716454222816354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758632, "dur": 7,
    "args": {
      "External id": 92277,
      "cbid": 211, "correlation": 92277
    }
  },
  {
    "ph": "s", "id": 92277, "pid": 76337, "tid": -914061504, "ts": 1716454222758632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222816372, "dur": 11,
    "args": {
      "External id": 92286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92286,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92286, "pid": 5, "tid": 7, "ts": 1716454222816372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758672, "dur": 10,
    "args": {
      "External id": 92286,
      "cbid": 211, "correlation": 92286
    }
  },
  {
    "ph": "s", "id": 92286, "pid": 76337, "tid": -914061504, "ts": 1716454222758672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758734, "dur": 0,
    "args": {
      "External id": 92296,
      "cbid": 317, "correlation": 92296
    }
  },
  {
    "ph": "f", "id": 92296, "pid": 76337, "tid": -914061504, "ts": 1716454222758734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758735, "dur": 0,
    "args": {
      "External id": 92297,
      "cbid": 203, "correlation": 92297
    }
  },
  {
    "ph": "f", "id": 92297, "pid": 76337, "tid": -914061504, "ts": 1716454222758735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758736, "dur": 0,
    "args": {
      "External id": 92298,
      "cbid": 205, "correlation": 92298
    }
  },
  {
    "ph": "f", "id": 92298, "pid": 76337, "tid": -914061504, "ts": 1716454222758736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222816384, "dur": 11,
    "args": {
      "External id": 92302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92302, "pid": 5, "tid": 7, "ts": 1716454222816384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758750, "dur": 12,
    "args": {
      "External id": 92302,
      "cbid": 211, "correlation": 92302
    }
  },
  {
    "ph": "s", "id": 92302, "pid": 76337, "tid": -914061504, "ts": 1716454222758750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222816396, "dur": 160,
    "args": {
      "External id": 92304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92304, "pid": 5, "tid": 7, "ts": 1716454222816396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758764, "dur": 5,
    "args": {
      "External id": 92304,
      "cbid": 211, "correlation": 92304
    }
  },
  {
    "ph": "s", "id": 92304, "pid": 76337, "tid": -914061504, "ts": 1716454222758764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222816557, "dur": 1,
    "args": {
      "External id": 92306,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92306,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 92306, "pid": 5, "tid": 7, "ts": 1716454222816557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758775, "dur": 7,
    "args": {
      "External id": 92306,
      "cbid": 51, "correlation": 92306
    }
  },
  {
    "ph": "s", "id": 92306, "pid": 76337, "tid": -914061504, "ts": 1716454222758775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222816561, "dur": 639,
    "args": {
      "External id": 92307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92307,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92307, "pid": 5, "tid": 7, "ts": 1716454222816561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758783, "dur": 6,
    "args": {
      "External id": 92307,
      "cbid": 211, "correlation": 92307
    }
  },
  {
    "ph": "s", "id": 92307, "pid": 76337, "tid": -914061504, "ts": 1716454222758783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222817201, "dur": 12,
    "args": {
      "External id": 92309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92309, "pid": 5, "tid": 7, "ts": 1716454222817201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758793, "dur": 5,
    "args": {
      "External id": 92309,
      "cbid": 211, "correlation": 92309
    }
  },
  {
    "ph": "s", "id": 92309, "pid": 76337, "tid": -914061504, "ts": 1716454222758793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222817214, "dur": 15,
    "args": {
      "External id": 92315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92315, "pid": 5, "tid": 7, "ts": 1716454222817214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758822, "dur": 9,
    "args": {
      "External id": 92315,
      "cbid": 211, "correlation": 92315
    }
  },
  {
    "ph": "s", "id": 92315, "pid": 76337, "tid": -914061504, "ts": 1716454222758822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758880, "dur": 0,
    "args": {
      "External id": 92325,
      "cbid": 317, "correlation": 92325
    }
  },
  {
    "ph": "f", "id": 92325, "pid": 76337, "tid": -914061504, "ts": 1716454222758880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758881, "dur": 0,
    "args": {
      "External id": 92326,
      "cbid": 203, "correlation": 92326
    }
  },
  {
    "ph": "f", "id": 92326, "pid": 76337, "tid": -914061504, "ts": 1716454222758881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758882, "dur": 0,
    "args": {
      "External id": 92327,
      "cbid": 205, "correlation": 92327
    }
  },
  {
    "ph": "f", "id": 92327, "pid": 76337, "tid": -914061504, "ts": 1716454222758882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222817230, "dur": 21,
    "args": {
      "External id": 92331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92331,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92331, "pid": 5, "tid": 7, "ts": 1716454222817230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758895, "dur": 11,
    "args": {
      "External id": 92331,
      "cbid": 211, "correlation": 92331
    }
  },
  {
    "ph": "s", "id": 92331, "pid": 76337, "tid": -914061504, "ts": 1716454222758895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222817252, "dur": 4,
    "args": {
      "External id": 92333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 92333, "pid": 5, "tid": 7, "ts": 1716454222817252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758913, "dur": 7,
    "args": {
      "External id": 92333,
      "cbid": 211, "correlation": 92333
    }
  },
  {
    "ph": "s", "id": 92333, "pid": 76337, "tid": -914061504, "ts": 1716454222758913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758925, "dur": 0,
    "args": {
      "External id": 92334,
      "cbid": 51, "correlation": 92334
    }
  },
  {
    "ph": "s", "id": 92334, "pid": 76337, "tid": -914061504, "ts": 1716454222758925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222817257, "dur": 172,
    "args": {
      "External id": 92335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92335,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 92335, "pid": 5, "tid": 7, "ts": 1716454222817257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758926, "dur": 7,
    "args": {
      "External id": 92335,
      "cbid": 211, "correlation": 92335
    }
  },
  {
    "ph": "s", "id": 92335, "pid": 76337, "tid": -914061504, "ts": 1716454222758926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222817430, "dur": 16,
    "args": {
      "External id": 92340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92340, "pid": 5, "tid": 7, "ts": 1716454222817430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758953, "dur": 8,
    "args": {
      "External id": 92340,
      "cbid": 211, "correlation": 92340
    }
  },
  {
    "ph": "s", "id": 92340, "pid": 76337, "tid": -914061504, "ts": 1716454222758953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222817447, "dur": 11,
    "args": {
      "External id": 92348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92348,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92348, "pid": 5, "tid": 7, "ts": 1716454222817447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222758991, "dur": 9,
    "args": {
      "External id": 92348,
      "cbid": 211, "correlation": 92348
    }
  },
  {
    "ph": "s", "id": 92348, "pid": 76337, "tid": -914061504, "ts": 1716454222758991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222817460, "dur": 10,
    "args": {
      "External id": 92356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92356,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92356, "pid": 5, "tid": 7, "ts": 1716454222817460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759021, "dur": 9,
    "args": {
      "External id": 92356,
      "cbid": 211, "correlation": 92356
    }
  },
  {
    "ph": "s", "id": 92356, "pid": 76337, "tid": -914061504, "ts": 1716454222759021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222817472, "dur": 18,
    "args": {
      "External id": 92376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92376,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 92376, "pid": 5, "tid": 7, "ts": 1716454222817472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759106, "dur": 12,
    "args": {
      "External id": 92376,
      "cbid": 211, "correlation": 92376
    }
  },
  {
    "ph": "s", "id": 92376, "pid": 76337, "tid": -914061504, "ts": 1716454222759106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222817491, "dur": 5,
    "args": {
      "External id": 92388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92388,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 92388, "pid": 5, "tid": 7, "ts": 1716454222817491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759129, "dur": 7,
    "args": {
      "External id": 92388,
      "cbid": 211, "correlation": 92388
    }
  },
  {
    "ph": "s", "id": 92388, "pid": 76337, "tid": -914061504, "ts": 1716454222759129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222817497, "dur": 16,
    "args": {
      "External id": 92391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92391, "pid": 5, "tid": 7, "ts": 1716454222817497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759148, "dur": 7,
    "args": {
      "External id": 92391,
      "cbid": 211, "correlation": 92391
    }
  },
  {
    "ph": "s", "id": 92391, "pid": 76337, "tid": -914061504, "ts": 1716454222759148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759205, "dur": 0,
    "args": {
      "External id": 92402,
      "cbid": 317, "correlation": 92402
    }
  },
  {
    "ph": "f", "id": 92402, "pid": 76337, "tid": -914061504, "ts": 1716454222759205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759205, "dur": 0,
    "args": {
      "External id": 92403,
      "cbid": 203, "correlation": 92403
    }
  },
  {
    "ph": "f", "id": 92403, "pid": 76337, "tid": -914061504, "ts": 1716454222759205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759206, "dur": 0,
    "args": {
      "External id": 92404,
      "cbid": 205, "correlation": 92404
    }
  },
  {
    "ph": "f", "id": 92404, "pid": 76337, "tid": -914061504, "ts": 1716454222759206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222817514, "dur": 11,
    "args": {
      "External id": 92408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92408,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92408, "pid": 5, "tid": 7, "ts": 1716454222817514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759220, "dur": 12,
    "args": {
      "External id": 92408,
      "cbid": 211, "correlation": 92408
    }
  },
  {
    "ph": "s", "id": 92408, "pid": 76337, "tid": -914061504, "ts": 1716454222759220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222817526, "dur": 3,
    "args": {
      "External id": 92410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92410,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 92410, "pid": 5, "tid": 7, "ts": 1716454222817526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759237, "dur": 6,
    "args": {
      "External id": 92410,
      "cbid": 211, "correlation": 92410
    }
  },
  {
    "ph": "s", "id": 92410, "pid": 76337, "tid": -914061504, "ts": 1716454222759237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759245, "dur": 0,
    "args": {
      "External id": 92411,
      "cbid": 51, "correlation": 92411
    }
  },
  {
    "ph": "s", "id": 92411, "pid": 76337, "tid": -914061504, "ts": 1716454222759245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222817531, "dur": 88,
    "args": {
      "External id": 92412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92412,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 92412, "pid": 5, "tid": 7, "ts": 1716454222817531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759246, "dur": 5,
    "args": {
      "External id": 92412,
      "cbid": 211, "correlation": 92412
    }
  },
  {
    "ph": "s", "id": 92412, "pid": 76337, "tid": -914061504, "ts": 1716454222759246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222817621, "dur": 15,
    "args": {
      "External id": 92417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92417, "pid": 5, "tid": 7, "ts": 1716454222817621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759273, "dur": 8,
    "args": {
      "External id": 92417,
      "cbid": 211, "correlation": 92417
    }
  },
  {
    "ph": "s", "id": 92417, "pid": 76337, "tid": -914061504, "ts": 1716454222759273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222817637, "dur": 82,
    "args": {
      "External id": 92426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92426,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92426, "pid": 5, "tid": 7, "ts": 1716454222817637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759360, "dur": 15,
    "args": {
      "External id": 92426,
      "cbid": 211, "correlation": 92426
    }
  },
  {
    "ph": "s", "id": 92426, "pid": 76337, "tid": -914061504, "ts": 1716454222759360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222817720, "dur": 30,
    "args": {
      "External id": 92448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92448,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92448, "pid": 5, "tid": 7, "ts": 1716454222817720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759422, "dur": 10,
    "args": {
      "External id": 92448,
      "cbid": 211, "correlation": 92448
    }
  },
  {
    "ph": "s", "id": 92448, "pid": 76337, "tid": -914061504, "ts": 1716454222759422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759525, "dur": 2,
    "args": {
      "External id": 92459,
      "cbid": 251, "correlation": 92459
    }
  },
  {
    "ph": "f", "id": 92459, "pid": 76337, "tid": -914061504, "ts": 1716454222759525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222817751, "dur": 161,
    "args": {
      "External id": 92460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92460,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92460, "pid": 5, "tid": 7, "ts": 1716454222817751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759531, "dur": 13,
    "args": {
      "External id": 92460,
      "cbid": 211, "correlation": 92460
    }
  },
  {
    "ph": "s", "id": 92460, "pid": 76337, "tid": -914061504, "ts": 1716454222759531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759603, "dur": 1,
    "args": {
      "External id": 92471,
      "cbid": 251, "correlation": 92471
    }
  },
  {
    "ph": "f", "id": 92471, "pid": 76337, "tid": -914061504, "ts": 1716454222759603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222817913, "dur": 157,
    "args": {
      "External id": 92472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92472,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92472, "pid": 5, "tid": 7, "ts": 1716454222817913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759607, "dur": 12,
    "args": {
      "External id": 92472,
      "cbid": 211, "correlation": 92472
    }
  },
  {
    "ph": "s", "id": 92472, "pid": 76337, "tid": -914061504, "ts": 1716454222759607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759674, "dur": 1,
    "args": {
      "External id": 92483,
      "cbid": 251, "correlation": 92483
    }
  },
  {
    "ph": "f", "id": 92483, "pid": 76337, "tid": -914061504, "ts": 1716454222759674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222818072, "dur": 133,
    "args": {
      "External id": 92484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92484,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92484, "pid": 5, "tid": 7, "ts": 1716454222818072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759679, "dur": 11,
    "args": {
      "External id": 92484,
      "cbid": 211, "correlation": 92484
    }
  },
  {
    "ph": "s", "id": 92484, "pid": 76337, "tid": -914061504, "ts": 1716454222759679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222818207, "dur": 330,
    "args": {
      "External id": 92509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92509,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92509, "pid": 5, "tid": 7, "ts": 1716454222818207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759771, "dur": 14,
    "args": {
      "External id": 92509,
      "cbid": 211, "correlation": 92509
    }
  },
  {
    "ph": "s", "id": 92509, "pid": 76337, "tid": -914061504, "ts": 1716454222759771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759879, "dur": 2,
    "args": {
      "External id": 92527,
      "cbid": 251, "correlation": 92527
    }
  },
  {
    "ph": "f", "id": 92527, "pid": 76337, "tid": -914061504, "ts": 1716454222759879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222818539, "dur": 165,
    "args": {
      "External id": 92529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92529,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92529, "pid": 5, "tid": 7, "ts": 1716454222818539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759885, "dur": 14,
    "args": {
      "External id": 92529,
      "cbid": 211, "correlation": 92529
    }
  },
  {
    "ph": "s", "id": 92529, "pid": 76337, "tid": -914061504, "ts": 1716454222759885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222818705, "dur": 19,
    "args": {
      "External id": 92537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92537,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92537, "pid": 5, "tid": 7, "ts": 1716454222818705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222759959, "dur": 12,
    "args": {
      "External id": 92537,
      "cbid": 211, "correlation": 92537
    }
  },
  {
    "ph": "s", "id": 92537, "pid": 76337, "tid": -914061504, "ts": 1716454222759959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222818726, "dur": 28,
    "args": {
      "External id": 92545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92545,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92545, "pid": 5, "tid": 7, "ts": 1716454222818726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760008, "dur": 10,
    "args": {
      "External id": 92545,
      "cbid": 211, "correlation": 92545
    }
  },
  {
    "ph": "s", "id": 92545, "pid": 76337, "tid": -914061504, "ts": 1716454222760008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222818755, "dur": 19,
    "args": {
      "External id": 92556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92556, "pid": 5, "tid": 7, "ts": 1716454222818755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760088, "dur": 13,
    "args": {
      "External id": 92556,
      "cbid": 211, "correlation": 92556
    }
  },
  {
    "ph": "s", "id": 92556, "pid": 76337, "tid": -914061504, "ts": 1716454222760088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222818775, "dur": 16,
    "args": {
      "External id": 92578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92578,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92578, "pid": 5, "tid": 7, "ts": 1716454222818775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760122, "dur": 8,
    "args": {
      "External id": 92578,
      "cbid": 211, "correlation": 92578
    }
  },
  {
    "ph": "s", "id": 92578, "pid": 76337, "tid": -914061504, "ts": 1716454222760122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760209, "dur": 2,
    "args": {
      "External id": 92589,
      "cbid": 251, "correlation": 92589
    }
  },
  {
    "ph": "f", "id": 92589, "pid": 76337, "tid": -914061504, "ts": 1716454222760209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222818792, "dur": 87,
    "args": {
      "External id": 92590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92590,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 92590, "pid": 5, "tid": 7, "ts": 1716454222818792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760215, "dur": 14,
    "args": {
      "External id": 92590,
      "cbid": 211, "correlation": 92590
    }
  },
  {
    "ph": "s", "id": 92590, "pid": 76337, "tid": -914061504, "ts": 1716454222760215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760290, "dur": 1,
    "args": {
      "External id": 92601,
      "cbid": 251, "correlation": 92601
    }
  },
  {
    "ph": "f", "id": 92601, "pid": 76337, "tid": -914061504, "ts": 1716454222760290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760294, "dur": 0,
    "args": {
      "External id": 92602,
      "cbid": 251, "correlation": 92602
    }
  },
  {
    "ph": "f", "id": 92602, "pid": 76337, "tid": -914061504, "ts": 1716454222760294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222818881, "dur": 12,
    "args": {
      "External id": 92603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92603,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92603, "pid": 5, "tid": 7, "ts": 1716454222818881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760296, "dur": 12,
    "args": {
      "External id": 92603,
      "cbid": 211, "correlation": 92603
    }
  },
  {
    "ph": "s", "id": 92603, "pid": 76337, "tid": -914061504, "ts": 1716454222760296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222818894, "dur": 6,
    "args": {
      "External id": 92605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92605,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92605, "pid": 5, "tid": 7, "ts": 1716454222818894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760310, "dur": 6,
    "args": {
      "External id": 92605,
      "cbid": 211, "correlation": 92605
    }
  },
  {
    "ph": "s", "id": 92605, "pid": 76337, "tid": -914061504, "ts": 1716454222760310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760369, "dur": 1,
    "args": {
      "External id": 92616,
      "cbid": 251, "correlation": 92616
    }
  },
  {
    "ph": "f", "id": 92616, "pid": 76337, "tid": -914061504, "ts": 1716454222760369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760372, "dur": 0,
    "args": {
      "External id": 92617,
      "cbid": 251, "correlation": 92617
    }
  },
  {
    "ph": "f", "id": 92617, "pid": 76337, "tid": -914061504, "ts": 1716454222760372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222818901, "dur": 8,
    "args": {
      "External id": 92618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92618,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92618, "pid": 5, "tid": 7, "ts": 1716454222818901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760373, "dur": 11,
    "args": {
      "External id": 92618,
      "cbid": 211, "correlation": 92618
    }
  },
  {
    "ph": "s", "id": 92618, "pid": 76337, "tid": -914061504, "ts": 1716454222760373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222818911, "dur": 4,
    "args": {
      "External id": 92620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92620,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92620, "pid": 5, "tid": 7, "ts": 1716454222818911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760386, "dur": 6,
    "args": {
      "External id": 92620,
      "cbid": 211, "correlation": 92620
    }
  },
  {
    "ph": "s", "id": 92620, "pid": 76337, "tid": -914061504, "ts": 1716454222760386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222818916, "dur": 54,
    "args": {
      "External id": 92645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92645,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92645, "pid": 5, "tid": 7, "ts": 1716454222818916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760465, "dur": 12,
    "args": {
      "External id": 92645,
      "cbid": 211, "correlation": 92645
    }
  },
  {
    "ph": "s", "id": 92645, "pid": 76337, "tid": -914061504, "ts": 1716454222760465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760567, "dur": 2,
    "args": {
      "External id": 92663,
      "cbid": 251, "correlation": 92663
    }
  },
  {
    "ph": "f", "id": 92663, "pid": 76337, "tid": -914061504, "ts": 1716454222760567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222818971, "dur": 89,
    "args": {
      "External id": 92665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92665,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 92665, "pid": 5, "tid": 7, "ts": 1716454222818971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760574, "dur": 14,
    "args": {
      "External id": 92665,
      "cbid": 211, "correlation": 92665
    }
  },
  {
    "ph": "s", "id": 92665, "pid": 76337, "tid": -914061504, "ts": 1716454222760574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222819062, "dur": 10,
    "args": {
      "External id": 92673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92673,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92673, "pid": 5, "tid": 7, "ts": 1716454222819062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760645, "dur": 13,
    "args": {
      "External id": 92673,
      "cbid": 211, "correlation": 92673
    }
  },
  {
    "ph": "s", "id": 92673, "pid": 76337, "tid": -914061504, "ts": 1716454222760645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222819073, "dur": 20,
    "args": {
      "External id": 92681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92681,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92681, "pid": 5, "tid": 7, "ts": 1716454222819073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760687, "dur": 10,
    "args": {
      "External id": 92681,
      "cbid": 211, "correlation": 92681
    }
  },
  {
    "ph": "s", "id": 92681, "pid": 76337, "tid": -914061504, "ts": 1716454222760687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222819095, "dur": 17,
    "args": {
      "External id": 92703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92703,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92703, "pid": 5, "tid": 7, "ts": 1716454222819095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760739, "dur": 11,
    "args": {
      "External id": 92703,
      "cbid": 211, "correlation": 92703
    }
  },
  {
    "ph": "s", "id": 92703, "pid": 76337, "tid": -914061504, "ts": 1716454222760739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760829, "dur": 1,
    "args": {
      "External id": 92719,
      "cbid": 251, "correlation": 92719
    }
  },
  {
    "ph": "f", "id": 92719, "pid": 76337, "tid": -914061504, "ts": 1716454222760829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760834, "dur": 0,
    "args": {
      "External id": 92721,
      "cbid": 251, "correlation": 92721
    }
  },
  {
    "ph": "f", "id": 92721, "pid": 76337, "tid": -914061504, "ts": 1716454222760834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222819113, "dur": 492,
    "args": {
      "External id": 92722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92722,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92722, "pid": 5, "tid": 7, "ts": 1716454222819113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760836, "dur": 13,
    "args": {
      "External id": 92722,
      "cbid": 211, "correlation": 92722
    }
  },
  {
    "ph": "s", "id": 92722, "pid": 76337, "tid": -914061504, "ts": 1716454222760836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222819606, "dur": 65,
    "args": {
      "External id": 92730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92730, "pid": 5, "tid": 7, "ts": 1716454222819606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760910, "dur": 14,
    "args": {
      "External id": 92730,
      "cbid": 211, "correlation": 92730
    }
  },
  {
    "ph": "s", "id": 92730, "pid": 76337, "tid": -914061504, "ts": 1716454222760910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222819672, "dur": 65,
    "args": {
      "External id": 92738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92738, "pid": 5, "tid": 7, "ts": 1716454222819672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222760945, "dur": 8,
    "args": {
      "External id": 92738,
      "cbid": 211, "correlation": 92738
    }
  },
  {
    "ph": "s", "id": 92738, "pid": 76337, "tid": -914061504, "ts": 1716454222760945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761037, "dur": 1,
    "args": {
      "External id": 92754,
      "cbid": 251, "correlation": 92754
    }
  },
  {
    "ph": "f", "id": 92754, "pid": 76337, "tid": -914061504, "ts": 1716454222761037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222819740, "dur": 1,
    "args": {
      "External id": 92756,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92756,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 92756, "pid": 5, "tid": 7, "ts": 1716454222819740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761042, "dur": 12,
    "args": {
      "External id": 92756,
      "cbid": 51, "correlation": 92756
    }
  },
  {
    "ph": "s", "id": 92756, "pid": 76337, "tid": -914061504, "ts": 1716454222761042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222819743, "dur": 270,
    "args": {
      "External id": 92757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92757,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 92757, "pid": 5, "tid": 7, "ts": 1716454222819743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761055, "dur": 12,
    "args": {
      "External id": 92757,
      "cbid": 211, "correlation": 92757
    }
  },
  {
    "ph": "s", "id": 92757, "pid": 76337, "tid": -914061504, "ts": 1716454222761055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222820014, "dur": 14,
    "args": {
      "External id": 92765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92765,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92765, "pid": 5, "tid": 7, "ts": 1716454222820014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761099, "dur": 11,
    "args": {
      "External id": 92765,
      "cbid": 211, "correlation": 92765
    }
  },
  {
    "ph": "s", "id": 92765, "pid": 76337, "tid": -914061504, "ts": 1716454222761099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222820030, "dur": 37,
    "args": {
      "External id": 92776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92776, "pid": 5, "tid": 7, "ts": 1716454222820030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761172, "dur": 13,
    "args": {
      "External id": 92776,
      "cbid": 211, "correlation": 92776
    }
  },
  {
    "ph": "s", "id": 92776, "pid": 76337, "tid": -914061504, "ts": 1716454222761172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761241, "dur": 0,
    "args": {
      "External id": 92788,
      "cbid": 317, "correlation": 92788
    }
  },
  {
    "ph": "f", "id": 92788, "pid": 76337, "tid": -914061504, "ts": 1716454222761241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761242, "dur": 0,
    "args": {
      "External id": 92789,
      "cbid": 203, "correlation": 92789
    }
  },
  {
    "ph": "f", "id": 92789, "pid": 76337, "tid": -914061504, "ts": 1716454222761242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761243, "dur": 0,
    "args": {
      "External id": 92790,
      "cbid": 205, "correlation": 92790
    }
  },
  {
    "ph": "f", "id": 92790, "pid": 76337, "tid": -914061504, "ts": 1716454222761243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222820068, "dur": 13,
    "args": {
      "External id": 92794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92794,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92794, "pid": 5, "tid": 7, "ts": 1716454222820068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761261, "dur": 12,
    "args": {
      "External id": 92794,
      "cbid": 211, "correlation": 92794
    }
  },
  {
    "ph": "s", "id": 92794, "pid": 76337, "tid": -914061504, "ts": 1716454222761261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222820082, "dur": 4,
    "args": {
      "External id": 92796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 92796, "pid": 5, "tid": 7, "ts": 1716454222820082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761279, "dur": 6,
    "args": {
      "External id": 92796,
      "cbid": 211, "correlation": 92796
    }
  },
  {
    "ph": "s", "id": 92796, "pid": 76337, "tid": -914061504, "ts": 1716454222761279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761288, "dur": 0,
    "args": {
      "External id": 92797,
      "cbid": 51, "correlation": 92797
    }
  },
  {
    "ph": "s", "id": 92797, "pid": 76337, "tid": -914061504, "ts": 1716454222761288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222820088, "dur": 96,
    "args": {
      "External id": 92798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92798,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 92798, "pid": 5, "tid": 7, "ts": 1716454222820088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761289, "dur": 5,
    "args": {
      "External id": 92798,
      "cbid": 211, "correlation": 92798
    }
  },
  {
    "ph": "s", "id": 92798, "pid": 76337, "tid": -914061504, "ts": 1716454222761289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222820186, "dur": 16,
    "args": {
      "External id": 92803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92803,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92803, "pid": 5, "tid": 7, "ts": 1716454222820186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761317, "dur": 9,
    "args": {
      "External id": 92803,
      "cbid": 211, "correlation": 92803
    }
  },
  {
    "ph": "s", "id": 92803, "pid": 76337, "tid": -914061504, "ts": 1716454222761317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222820203, "dur": 11,
    "args": {
      "External id": 92811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92811,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92811, "pid": 5, "tid": 7, "ts": 1716454222820203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761349, "dur": 8,
    "args": {
      "External id": 92811,
      "cbid": 211, "correlation": 92811
    }
  },
  {
    "ph": "s", "id": 92811, "pid": 76337, "tid": -914061504, "ts": 1716454222761349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222820216, "dur": 30,
    "args": {
      "External id": 92820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92820,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92820, "pid": 5, "tid": 7, "ts": 1716454222820216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761392, "dur": 11,
    "args": {
      "External id": 92820,
      "cbid": 211, "correlation": 92820
    }
  },
  {
    "ph": "s", "id": 92820, "pid": 76337, "tid": -914061504, "ts": 1716454222761392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222820247, "dur": 30,
    "args": {
      "External id": 92840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92840,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 92840, "pid": 5, "tid": 7, "ts": 1716454222820247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761469, "dur": 12,
    "args": {
      "External id": 92840,
      "cbid": 211, "correlation": 92840
    }
  },
  {
    "ph": "s", "id": 92840, "pid": 76337, "tid": -914061504, "ts": 1716454222761469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222820279, "dur": 5,
    "args": {
      "External id": 92852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92852,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92852, "pid": 5, "tid": 7, "ts": 1716454222820279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761491, "dur": 7,
    "args": {
      "External id": 92852,
      "cbid": 211, "correlation": 92852
    }
  },
  {
    "ph": "s", "id": 92852, "pid": 76337, "tid": -914061504, "ts": 1716454222761491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222820285, "dur": 30,
    "args": {
      "External id": 92855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92855, "pid": 5, "tid": 7, "ts": 1716454222820285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761510, "dur": 6,
    "args": {
      "External id": 92855,
      "cbid": 211, "correlation": 92855
    }
  },
  {
    "ph": "s", "id": 92855, "pid": 76337, "tid": -914061504, "ts": 1716454222761510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222820316, "dur": 21,
    "args": {
      "External id": 92864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92864,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92864, "pid": 5, "tid": 7, "ts": 1716454222820316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761551, "dur": 10,
    "args": {
      "External id": 92864,
      "cbid": 211, "correlation": 92864
    }
  },
  {
    "ph": "s", "id": 92864, "pid": 76337, "tid": -914061504, "ts": 1716454222761551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761605, "dur": 0,
    "args": {
      "External id": 92874,
      "cbid": 317, "correlation": 92874
    }
  },
  {
    "ph": "f", "id": 92874, "pid": 76337, "tid": -914061504, "ts": 1716454222761605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761606, "dur": 0,
    "args": {
      "External id": 92875,
      "cbid": 203, "correlation": 92875
    }
  },
  {
    "ph": "f", "id": 92875, "pid": 76337, "tid": -914061504, "ts": 1716454222761606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761607, "dur": 0,
    "args": {
      "External id": 92876,
      "cbid": 205, "correlation": 92876
    }
  },
  {
    "ph": "f", "id": 92876, "pid": 76337, "tid": -914061504, "ts": 1716454222761607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222820339, "dur": 22,
    "args": {
      "External id": 92880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92880, "pid": 5, "tid": 7, "ts": 1716454222820339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761620, "dur": 11,
    "args": {
      "External id": 92880,
      "cbid": 211, "correlation": 92880
    }
  },
  {
    "ph": "s", "id": 92880, "pid": 76337, "tid": -914061504, "ts": 1716454222761620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222820362, "dur": 314,
    "args": {
      "External id": 92882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92882,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92882, "pid": 5, "tid": 7, "ts": 1716454222820362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761634, "dur": 5,
    "args": {
      "External id": 92882,
      "cbid": 211, "correlation": 92882
    }
  },
  {
    "ph": "s", "id": 92882, "pid": 76337, "tid": -914061504, "ts": 1716454222761634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222820678, "dur": 1,
    "args": {
      "External id": 92884,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92884,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 92884, "pid": 5, "tid": 7, "ts": 1716454222820678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761646, "dur": 8,
    "args": {
      "External id": 92884,
      "cbid": 51, "correlation": 92884
    }
  },
  {
    "ph": "s", "id": 92884, "pid": 76337, "tid": -914061504, "ts": 1716454222761646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222820682, "dur": 1238,
    "args": {
      "External id": 92885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92885,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92885, "pid": 5, "tid": 7, "ts": 1716454222820682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761655, "dur": 7,
    "args": {
      "External id": 92885,
      "cbid": 211, "correlation": 92885
    }
  },
  {
    "ph": "s", "id": 92885, "pid": 76337, "tid": -914061504, "ts": 1716454222761655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222821921, "dur": 13,
    "args": {
      "External id": 92887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92887,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92887, "pid": 5, "tid": 7, "ts": 1716454222821921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761666, "dur": 5,
    "args": {
      "External id": 92887,
      "cbid": 211, "correlation": 92887
    }
  },
  {
    "ph": "s", "id": 92887, "pid": 76337, "tid": -914061504, "ts": 1716454222761666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222821935, "dur": 14,
    "args": {
      "External id": 92893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92893,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92893, "pid": 5, "tid": 7, "ts": 1716454222821935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761694, "dur": 9,
    "args": {
      "External id": 92893,
      "cbid": 211, "correlation": 92893
    }
  },
  {
    "ph": "s", "id": 92893, "pid": 76337, "tid": -914061504, "ts": 1716454222761694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222821951, "dur": 3,
    "args": {
      "External id": 92901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92901,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 92901, "pid": 5, "tid": 7, "ts": 1716454222821951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761740, "dur": 9,
    "args": {
      "External id": 92901,
      "cbid": 211, "correlation": 92901
    }
  },
  {
    "ph": "s", "id": 92901, "pid": 76337, "tid": -914061504, "ts": 1716454222761740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761808, "dur": 1,
    "args": {
      "External id": 92917,
      "cbid": 251, "correlation": 92917
    }
  },
  {
    "ph": "f", "id": 92917, "pid": 76337, "tid": -914061504, "ts": 1716454222761808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761813, "dur": 0,
    "args": {
      "External id": 92919,
      "cbid": 251, "correlation": 92919
    }
  },
  {
    "ph": "f", "id": 92919, "pid": 76337, "tid": -914061504, "ts": 1716454222761813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222821955, "dur": 13,
    "args": {
      "External id": 92920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92920,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92920, "pid": 5, "tid": 7, "ts": 1716454222821955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761815, "dur": 12,
    "args": {
      "External id": 92920,
      "cbid": 211, "correlation": 92920
    }
  },
  {
    "ph": "s", "id": 92920, "pid": 76337, "tid": -914061504, "ts": 1716454222761815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222821969, "dur": 5,
    "args": {
      "External id": 92922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92922,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92922, "pid": 5, "tid": 7, "ts": 1716454222821969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761830, "dur": 6,
    "args": {
      "External id": 92922,
      "cbid": 211, "correlation": 92922
    }
  },
  {
    "ph": "s", "id": 92922, "pid": 76337, "tid": -914061504, "ts": 1716454222761830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222821976, "dur": 16,
    "args": {
      "External id": 92932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92932, "pid": 5, "tid": 7, "ts": 1716454222821976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761890, "dur": 12,
    "args": {
      "External id": 92932,
      "cbid": 211, "correlation": 92932
    }
  },
  {
    "ph": "s", "id": 92932, "pid": 76337, "tid": -914061504, "ts": 1716454222761890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222821993, "dur": 17,
    "args": {
      "External id": 92952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92952,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 92952, "pid": 5, "tid": 7, "ts": 1716454222821993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761957, "dur": 11,
    "args": {
      "External id": 92952,
      "cbid": 211, "correlation": 92952
    }
  },
  {
    "ph": "s", "id": 92952, "pid": 76337, "tid": -914061504, "ts": 1716454222761957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222822011, "dur": 4,
    "args": {
      "External id": 92964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92964,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 92964, "pid": 5, "tid": 7, "ts": 1716454222822011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222761986, "dur": 7,
    "args": {
      "External id": 92964,
      "cbid": 211, "correlation": 92964
    }
  },
  {
    "ph": "s", "id": 92964, "pid": 76337, "tid": -914061504, "ts": 1716454222761986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222822016, "dur": 16,
    "args": {
      "External id": 92967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92967,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92967, "pid": 5, "tid": 7, "ts": 1716454222822016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762005, "dur": 7,
    "args": {
      "External id": 92967,
      "cbid": 211, "correlation": 92967
    }
  },
  {
    "ph": "s", "id": 92967, "pid": 76337, "tid": -914061504, "ts": 1716454222762005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222822034, "dur": 11,
    "args": {
      "External id": 92976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92976,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92976, "pid": 5, "tid": 7, "ts": 1716454222822034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762047, "dur": 11,
    "args": {
      "External id": 92976,
      "cbid": 211, "correlation": 92976
    }
  },
  {
    "ph": "s", "id": 92976, "pid": 76337, "tid": -914061504, "ts": 1716454222762047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762112, "dur": 0,
    "args": {
      "External id": 92986,
      "cbid": 317, "correlation": 92986
    }
  },
  {
    "ph": "f", "id": 92986, "pid": 76337, "tid": -914061504, "ts": 1716454222762112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762113, "dur": 0,
    "args": {
      "External id": 92987,
      "cbid": 203, "correlation": 92987
    }
  },
  {
    "ph": "f", "id": 92987, "pid": 76337, "tid": -914061504, "ts": 1716454222762113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762114, "dur": 0,
    "args": {
      "External id": 92988,
      "cbid": 205, "correlation": 92988
    }
  },
  {
    "ph": "f", "id": 92988, "pid": 76337, "tid": -914061504, "ts": 1716454222762114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222822046, "dur": 11,
    "args": {
      "External id": 92992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92992, "pid": 5, "tid": 7, "ts": 1716454222822046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762128, "dur": 12,
    "args": {
      "External id": 92992,
      "cbid": 211, "correlation": 92992
    }
  },
  {
    "ph": "s", "id": 92992, "pid": 76337, "tid": -914061504, "ts": 1716454222762128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222822058, "dur": 160,
    "args": {
      "External id": 92994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92994,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92994, "pid": 5, "tid": 7, "ts": 1716454222822058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762142, "dur": 5,
    "args": {
      "External id": 92994,
      "cbid": 211, "correlation": 92994
    }
  },
  {
    "ph": "s", "id": 92994, "pid": 76337, "tid": -914061504, "ts": 1716454222762142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222822221, "dur": 1,
    "args": {
      "External id": 92996,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 92996,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 92996, "pid": 5, "tid": 7, "ts": 1716454222822221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762153, "dur": 6,
    "args": {
      "External id": 92996,
      "cbid": 51, "correlation": 92996
    }
  },
  {
    "ph": "s", "id": 92996, "pid": 76337, "tid": -914061504, "ts": 1716454222762153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222822224, "dur": 639,
    "args": {
      "External id": 92997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92997,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 92997, "pid": 5, "tid": 7, "ts": 1716454222822224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762161, "dur": 7,
    "args": {
      "External id": 92997,
      "cbid": 211, "correlation": 92997
    }
  },
  {
    "ph": "s", "id": 92997, "pid": 76337, "tid": -914061504, "ts": 1716454222762161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222822865, "dur": 12,
    "args": {
      "External id": 92999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 92999,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 92999, "pid": 5, "tid": 7, "ts": 1716454222822865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762172, "dur": 5,
    "args": {
      "External id": 92999,
      "cbid": 211, "correlation": 92999
    }
  },
  {
    "ph": "s", "id": 92999, "pid": 76337, "tid": -914061504, "ts": 1716454222762172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222822878, "dur": 14,
    "args": {
      "External id": 93005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93005, "pid": 5, "tid": 7, "ts": 1716454222822878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762200, "dur": 8,
    "args": {
      "External id": 93005,
      "cbid": 211, "correlation": 93005
    }
  },
  {
    "ph": "s", "id": 93005, "pid": 76337, "tid": -914061504, "ts": 1716454222762200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762259, "dur": 0,
    "args": {
      "External id": 93015,
      "cbid": 317, "correlation": 93015
    }
  },
  {
    "ph": "f", "id": 93015, "pid": 76337, "tid": -914061504, "ts": 1716454222762259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762260, "dur": 0,
    "args": {
      "External id": 93016,
      "cbid": 203, "correlation": 93016
    }
  },
  {
    "ph": "f", "id": 93016, "pid": 76337, "tid": -914061504, "ts": 1716454222762260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762261, "dur": 0,
    "args": {
      "External id": 93017,
      "cbid": 205, "correlation": 93017
    }
  },
  {
    "ph": "f", "id": 93017, "pid": 76337, "tid": -914061504, "ts": 1716454222762261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222822894, "dur": 21,
    "args": {
      "External id": 93021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93021, "pid": 5, "tid": 7, "ts": 1716454222822894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762274, "dur": 12,
    "args": {
      "External id": 93021,
      "cbid": 211, "correlation": 93021
    }
  },
  {
    "ph": "s", "id": 93021, "pid": 76337, "tid": -914061504, "ts": 1716454222762274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222822916, "dur": 4,
    "args": {
      "External id": 93023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93023, "pid": 5, "tid": 7, "ts": 1716454222822916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762290, "dur": 6,
    "args": {
      "External id": 93023,
      "cbid": 211, "correlation": 93023
    }
  },
  {
    "ph": "s", "id": 93023, "pid": 76337, "tid": -914061504, "ts": 1716454222762290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762299, "dur": 0,
    "args": {
      "External id": 93024,
      "cbid": 51, "correlation": 93024
    }
  },
  {
    "ph": "s", "id": 93024, "pid": 76337, "tid": -914061504, "ts": 1716454222762299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222822921, "dur": 167,
    "args": {
      "External id": 93025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93025,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 93025, "pid": 5, "tid": 7, "ts": 1716454222822921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762299, "dur": 5,
    "args": {
      "External id": 93025,
      "cbid": 211, "correlation": 93025
    }
  },
  {
    "ph": "s", "id": 93025, "pid": 76337, "tid": -914061504, "ts": 1716454222762299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222823090, "dur": 16,
    "args": {
      "External id": 93030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93030, "pid": 5, "tid": 7, "ts": 1716454222823090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762324, "dur": 8,
    "args": {
      "External id": 93030,
      "cbid": 211, "correlation": 93030
    }
  },
  {
    "ph": "s", "id": 93030, "pid": 76337, "tid": -914061504, "ts": 1716454222762324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222823106, "dur": 13,
    "args": {
      "External id": 93038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93038,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93038, "pid": 5, "tid": 7, "ts": 1716454222823106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762353, "dur": 8,
    "args": {
      "External id": 93038,
      "cbid": 211, "correlation": 93038
    }
  },
  {
    "ph": "s", "id": 93038, "pid": 76337, "tid": -914061504, "ts": 1716454222762353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222823120, "dur": 10,
    "args": {
      "External id": 93046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93046,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93046, "pid": 5, "tid": 7, "ts": 1716454222823120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762382, "dur": 9,
    "args": {
      "External id": 93046,
      "cbid": 211, "correlation": 93046
    }
  },
  {
    "ph": "s", "id": 93046, "pid": 76337, "tid": -914061504, "ts": 1716454222762382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222823132, "dur": 18,
    "args": {
      "External id": 93066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93066,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 93066, "pid": 5, "tid": 7, "ts": 1716454222823132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762467, "dur": 13,
    "args": {
      "External id": 93066,
      "cbid": 211, "correlation": 93066
    }
  },
  {
    "ph": "s", "id": 93066, "pid": 76337, "tid": -914061504, "ts": 1716454222762467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222823151, "dur": 4,
    "args": {
      "External id": 93078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93078,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 93078, "pid": 5, "tid": 7, "ts": 1716454222823151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762490, "dur": 6,
    "args": {
      "External id": 93078,
      "cbid": 211, "correlation": 93078
    }
  },
  {
    "ph": "s", "id": 93078, "pid": 76337, "tid": -914061504, "ts": 1716454222762490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222823157, "dur": 16,
    "args": {
      "External id": 93081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93081,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93081, "pid": 5, "tid": 7, "ts": 1716454222823157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762508, "dur": 6,
    "args": {
      "External id": 93081,
      "cbid": 211, "correlation": 93081
    }
  },
  {
    "ph": "s", "id": 93081, "pid": 76337, "tid": -914061504, "ts": 1716454222762508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762566, "dur": 0,
    "args": {
      "External id": 93092,
      "cbid": 317, "correlation": 93092
    }
  },
  {
    "ph": "f", "id": 93092, "pid": 76337, "tid": -914061504, "ts": 1716454222762566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762567, "dur": 0,
    "args": {
      "External id": 93093,
      "cbid": 203, "correlation": 93093
    }
  },
  {
    "ph": "f", "id": 93093, "pid": 76337, "tid": -914061504, "ts": 1716454222762567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762567, "dur": 0,
    "args": {
      "External id": 93094,
      "cbid": 205, "correlation": 93094
    }
  },
  {
    "ph": "f", "id": 93094, "pid": 76337, "tid": -914061504, "ts": 1716454222762567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222823175, "dur": 13,
    "args": {
      "External id": 93098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93098, "pid": 5, "tid": 7, "ts": 1716454222823175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762581, "dur": 12,
    "args": {
      "External id": 93098,
      "cbid": 211, "correlation": 93098
    }
  },
  {
    "ph": "s", "id": 93098, "pid": 76337, "tid": -914061504, "ts": 1716454222762581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222823189, "dur": 3,
    "args": {
      "External id": 93100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93100, "pid": 5, "tid": 7, "ts": 1716454222823189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762596, "dur": 6,
    "args": {
      "External id": 93100,
      "cbid": 211, "correlation": 93100
    }
  },
  {
    "ph": "s", "id": 93100, "pid": 76337, "tid": -914061504, "ts": 1716454222762596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762604, "dur": 0,
    "args": {
      "External id": 93101,
      "cbid": 51, "correlation": 93101
    }
  },
  {
    "ph": "s", "id": 93101, "pid": 76337, "tid": -914061504, "ts": 1716454222762604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222823194, "dur": 91,
    "args": {
      "External id": 93102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93102,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 93102, "pid": 5, "tid": 7, "ts": 1716454222823194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762605, "dur": 5,
    "args": {
      "External id": 93102,
      "cbid": 211, "correlation": 93102
    }
  },
  {
    "ph": "s", "id": 93102, "pid": 76337, "tid": -914061504, "ts": 1716454222762605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222823286, "dur": 16,
    "args": {
      "External id": 93107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93107, "pid": 5, "tid": 7, "ts": 1716454222823286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762632, "dur": 8,
    "args": {
      "External id": 93107,
      "cbid": 211, "correlation": 93107
    }
  },
  {
    "ph": "s", "id": 93107, "pid": 76337, "tid": -914061504, "ts": 1716454222762632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222823302, "dur": 81,
    "args": {
      "External id": 93116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93116,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93116, "pid": 5, "tid": 7, "ts": 1716454222823302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762715, "dur": 15,
    "args": {
      "External id": 93116,
      "cbid": 211, "correlation": 93116
    }
  },
  {
    "ph": "s", "id": 93116, "pid": 76337, "tid": -914061504, "ts": 1716454222762715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222823385, "dur": 30,
    "args": {
      "External id": 93138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93138,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93138, "pid": 5, "tid": 7, "ts": 1716454222823385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762773, "dur": 10,
    "args": {
      "External id": 93138,
      "cbid": 211, "correlation": 93138
    }
  },
  {
    "ph": "s", "id": 93138, "pid": 76337, "tid": -914061504, "ts": 1716454222762773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762868, "dur": 1,
    "args": {
      "External id": 93149,
      "cbid": 251, "correlation": 93149
    }
  },
  {
    "ph": "f", "id": 93149, "pid": 76337, "tid": -914061504, "ts": 1716454222762868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222823416, "dur": 140,
    "args": {
      "External id": 93150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93150,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93150, "pid": 5, "tid": 7, "ts": 1716454222823416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762873, "dur": 13,
    "args": {
      "External id": 93150,
      "cbid": 211, "correlation": 93150
    }
  },
  {
    "ph": "s", "id": 93150, "pid": 76337, "tid": -914061504, "ts": 1716454222762873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762944, "dur": 1,
    "args": {
      "External id": 93161,
      "cbid": 251, "correlation": 93161
    }
  },
  {
    "ph": "f", "id": 93161, "pid": 76337, "tid": -914061504, "ts": 1716454222762944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222823558, "dur": 153,
    "args": {
      "External id": 93162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93162,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93162, "pid": 5, "tid": 7, "ts": 1716454222823558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222762948, "dur": 11,
    "args": {
      "External id": 93162,
      "cbid": 211, "correlation": 93162
    }
  },
  {
    "ph": "s", "id": 93162, "pid": 76337, "tid": -914061504, "ts": 1716454222762948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763024, "dur": 1,
    "args": {
      "External id": 93173,
      "cbid": 251, "correlation": 93173
    }
  },
  {
    "ph": "f", "id": 93173, "pid": 76337, "tid": -914061504, "ts": 1716454222763024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222823712, "dur": 157,
    "args": {
      "External id": 93174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93174,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93174, "pid": 5, "tid": 7, "ts": 1716454222823712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763028, "dur": 12,
    "args": {
      "External id": 93174,
      "cbid": 211, "correlation": 93174
    }
  },
  {
    "ph": "s", "id": 93174, "pid": 76337, "tid": -914061504, "ts": 1716454222763028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222823871, "dur": 331,
    "args": {
      "External id": 93199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93199,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93199, "pid": 5, "tid": 7, "ts": 1716454222823871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763117, "dur": 14,
    "args": {
      "External id": 93199,
      "cbid": 211, "correlation": 93199
    }
  },
  {
    "ph": "s", "id": 93199, "pid": 76337, "tid": -914061504, "ts": 1716454222763117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763220, "dur": 1,
    "args": {
      "External id": 93217,
      "cbid": 251, "correlation": 93217
    }
  },
  {
    "ph": "f", "id": 93217, "pid": 76337, "tid": -914061504, "ts": 1716454222763220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222824203, "dur": 162,
    "args": {
      "External id": 93219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93219,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93219, "pid": 5, "tid": 7, "ts": 1716454222824203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763226, "dur": 13,
    "args": {
      "External id": 93219,
      "cbid": 211, "correlation": 93219
    }
  },
  {
    "ph": "s", "id": 93219, "pid": 76337, "tid": -914061504, "ts": 1716454222763226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222824366, "dur": 19,
    "args": {
      "External id": 93227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93227,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93227, "pid": 5, "tid": 7, "ts": 1716454222824366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763297, "dur": 13,
    "args": {
      "External id": 93227,
      "cbid": 211, "correlation": 93227
    }
  },
  {
    "ph": "s", "id": 93227, "pid": 76337, "tid": -914061504, "ts": 1716454222763297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222824386, "dur": 27,
    "args": {
      "External id": 93235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93235,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93235, "pid": 5, "tid": 7, "ts": 1716454222824386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763337, "dur": 9,
    "args": {
      "External id": 93235,
      "cbid": 211, "correlation": 93235
    }
  },
  {
    "ph": "s", "id": 93235, "pid": 76337, "tid": -914061504, "ts": 1716454222763337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222824415, "dur": 18,
    "args": {
      "External id": 93246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93246, "pid": 5, "tid": 7, "ts": 1716454222824415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763411, "dur": 14,
    "args": {
      "External id": 93246,
      "cbid": 211, "correlation": 93246
    }
  },
  {
    "ph": "s", "id": 93246, "pid": 76337, "tid": -914061504, "ts": 1716454222763411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222824435, "dur": 16,
    "args": {
      "External id": 93268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93268,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93268, "pid": 5, "tid": 7, "ts": 1716454222824435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763443, "dur": 7,
    "args": {
      "External id": 93268,
      "cbid": 211, "correlation": 93268
    }
  },
  {
    "ph": "s", "id": 93268, "pid": 76337, "tid": -914061504, "ts": 1716454222763443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763529, "dur": 1,
    "args": {
      "External id": 93279,
      "cbid": 251, "correlation": 93279
    }
  },
  {
    "ph": "f", "id": 93279, "pid": 76337, "tid": -914061504, "ts": 1716454222763529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222824452, "dur": 88,
    "args": {
      "External id": 93280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93280,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93280, "pid": 5, "tid": 7, "ts": 1716454222824452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763535, "dur": 13,
    "args": {
      "External id": 93280,
      "cbid": 211, "correlation": 93280
    }
  },
  {
    "ph": "s", "id": 93280, "pid": 76337, "tid": -914061504, "ts": 1716454222763535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763607, "dur": 1,
    "args": {
      "External id": 93291,
      "cbid": 251, "correlation": 93291
    }
  },
  {
    "ph": "f", "id": 93291, "pid": 76337, "tid": -914061504, "ts": 1716454222763607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763610, "dur": 0,
    "args": {
      "External id": 93292,
      "cbid": 251, "correlation": 93292
    }
  },
  {
    "ph": "f", "id": 93292, "pid": 76337, "tid": -914061504, "ts": 1716454222763610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222824542, "dur": 13,
    "args": {
      "External id": 93293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93293,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93293, "pid": 5, "tid": 7, "ts": 1716454222824542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763612, "dur": 12,
    "args": {
      "External id": 93293,
      "cbid": 211, "correlation": 93293
    }
  },
  {
    "ph": "s", "id": 93293, "pid": 76337, "tid": -914061504, "ts": 1716454222763612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222824556, "dur": 6,
    "args": {
      "External id": 93295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93295,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93295, "pid": 5, "tid": 7, "ts": 1716454222824556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763626, "dur": 6,
    "args": {
      "External id": 93295,
      "cbid": 211, "correlation": 93295
    }
  },
  {
    "ph": "s", "id": 93295, "pid": 76337, "tid": -914061504, "ts": 1716454222763626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763684, "dur": 1,
    "args": {
      "External id": 93306,
      "cbid": 251, "correlation": 93306
    }
  },
  {
    "ph": "f", "id": 93306, "pid": 76337, "tid": -914061504, "ts": 1716454222763684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763688, "dur": 0,
    "args": {
      "External id": 93307,
      "cbid": 251, "correlation": 93307
    }
  },
  {
    "ph": "f", "id": 93307, "pid": 76337, "tid": -914061504, "ts": 1716454222763688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222824563, "dur": 9,
    "args": {
      "External id": 93308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93308,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93308, "pid": 5, "tid": 7, "ts": 1716454222824563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763689, "dur": 11,
    "args": {
      "External id": 93308,
      "cbid": 211, "correlation": 93308
    }
  },
  {
    "ph": "s", "id": 93308, "pid": 76337, "tid": -914061504, "ts": 1716454222763689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222824573, "dur": 3,
    "args": {
      "External id": 93310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93310,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93310, "pid": 5, "tid": 7, "ts": 1716454222824573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763702, "dur": 5,
    "args": {
      "External id": 93310,
      "cbid": 211, "correlation": 93310
    }
  },
  {
    "ph": "s", "id": 93310, "pid": 76337, "tid": -914061504, "ts": 1716454222763702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222824578, "dur": 55,
    "args": {
      "External id": 93335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93335,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93335, "pid": 5, "tid": 7, "ts": 1716454222824578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763781, "dur": 12,
    "args": {
      "External id": 93335,
      "cbid": 211, "correlation": 93335
    }
  },
  {
    "ph": "s", "id": 93335, "pid": 76337, "tid": -914061504, "ts": 1716454222763781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763881, "dur": 1,
    "args": {
      "External id": 93353,
      "cbid": 251, "correlation": 93353
    }
  },
  {
    "ph": "f", "id": 93353, "pid": 76337, "tid": -914061504, "ts": 1716454222763881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222824634, "dur": 89,
    "args": {
      "External id": 93355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93355,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93355, "pid": 5, "tid": 7, "ts": 1716454222824634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763887, "dur": 14,
    "args": {
      "External id": 93355,
      "cbid": 211, "correlation": 93355
    }
  },
  {
    "ph": "s", "id": 93355, "pid": 76337, "tid": -914061504, "ts": 1716454222763887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222824725, "dur": 9,
    "args": {
      "External id": 93363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93363,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93363, "pid": 5, "tid": 7, "ts": 1716454222824725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222763957, "dur": 12,
    "args": {
      "External id": 93363,
      "cbid": 211, "correlation": 93363
    }
  },
  {
    "ph": "s", "id": 93363, "pid": 76337, "tid": -914061504, "ts": 1716454222763957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222824735, "dur": 20,
    "args": {
      "External id": 93371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93371,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93371, "pid": 5, "tid": 7, "ts": 1716454222824735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764009, "dur": 10,
    "args": {
      "External id": 93371,
      "cbid": 211, "correlation": 93371
    }
  },
  {
    "ph": "s", "id": 93371, "pid": 76337, "tid": -914061504, "ts": 1716454222764009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222824756, "dur": 17,
    "args": {
      "External id": 93393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93393,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93393, "pid": 5, "tid": 7, "ts": 1716454222824756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764061, "dur": 11,
    "args": {
      "External id": 93393,
      "cbid": 211, "correlation": 93393
    }
  },
  {
    "ph": "s", "id": 93393, "pid": 76337, "tid": -914061504, "ts": 1716454222764061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764149, "dur": 1,
    "args": {
      "External id": 93409,
      "cbid": 251, "correlation": 93409
    }
  },
  {
    "ph": "f", "id": 93409, "pid": 76337, "tid": -914061504, "ts": 1716454222764149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764154, "dur": 0,
    "args": {
      "External id": 93411,
      "cbid": 251, "correlation": 93411
    }
  },
  {
    "ph": "f", "id": 93411, "pid": 76337, "tid": -914061504, "ts": 1716454222764154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222824775, "dur": 493,
    "args": {
      "External id": 93412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93412,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93412, "pid": 5, "tid": 7, "ts": 1716454222824775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764156, "dur": 13,
    "args": {
      "External id": 93412,
      "cbid": 211, "correlation": 93412
    }
  },
  {
    "ph": "s", "id": 93412, "pid": 76337, "tid": -914061504, "ts": 1716454222764156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222825269, "dur": 66,
    "args": {
      "External id": 93420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93420, "pid": 5, "tid": 7, "ts": 1716454222825269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764226, "dur": 14,
    "args": {
      "External id": 93420,
      "cbid": 211, "correlation": 93420
    }
  },
  {
    "ph": "s", "id": 93420, "pid": 76337, "tid": -914061504, "ts": 1716454222764226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222825336, "dur": 65,
    "args": {
      "External id": 93428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93428, "pid": 5, "tid": 7, "ts": 1716454222825336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764260, "dur": 8,
    "args": {
      "External id": 93428,
      "cbid": 211, "correlation": 93428
    }
  },
  {
    "ph": "s", "id": 93428, "pid": 76337, "tid": -914061504, "ts": 1716454222764260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764339, "dur": 1,
    "args": {
      "External id": 93444,
      "cbid": 251, "correlation": 93444
    }
  },
  {
    "ph": "f", "id": 93444, "pid": 76337, "tid": -914061504, "ts": 1716454222764339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222825403, "dur": 1,
    "args": {
      "External id": 93446,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 93446,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 93446, "pid": 5, "tid": 7, "ts": 1716454222825403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764344, "dur": 11,
    "args": {
      "External id": 93446,
      "cbid": 51, "correlation": 93446
    }
  },
  {
    "ph": "s", "id": 93446, "pid": 76337, "tid": -914061504, "ts": 1716454222764344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222825407, "dur": 265,
    "args": {
      "External id": 93447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93447,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93447, "pid": 5, "tid": 7, "ts": 1716454222825407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764356, "dur": 11,
    "args": {
      "External id": 93447,
      "cbid": 211, "correlation": 93447
    }
  },
  {
    "ph": "s", "id": 93447, "pid": 76337, "tid": -914061504, "ts": 1716454222764356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222825673, "dur": 14,
    "args": {
      "External id": 93455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93455,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93455, "pid": 5, "tid": 7, "ts": 1716454222825673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764398, "dur": 11,
    "args": {
      "External id": 93455,
      "cbid": 211, "correlation": 93455
    }
  },
  {
    "ph": "s", "id": 93455, "pid": 76337, "tid": -914061504, "ts": 1716454222764398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222825688, "dur": 37,
    "args": {
      "External id": 93466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93466, "pid": 5, "tid": 7, "ts": 1716454222825688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764470, "dur": 12,
    "args": {
      "External id": 93466,
      "cbid": 211, "correlation": 93466
    }
  },
  {
    "ph": "s", "id": 93466, "pid": 76337, "tid": -914061504, "ts": 1716454222764470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764537, "dur": 0,
    "args": {
      "External id": 93478,
      "cbid": 317, "correlation": 93478
    }
  },
  {
    "ph": "f", "id": 93478, "pid": 76337, "tid": -914061504, "ts": 1716454222764537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764538, "dur": 0,
    "args": {
      "External id": 93479,
      "cbid": 203, "correlation": 93479
    }
  },
  {
    "ph": "f", "id": 93479, "pid": 76337, "tid": -914061504, "ts": 1716454222764538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764539, "dur": 0,
    "args": {
      "External id": 93480,
      "cbid": 205, "correlation": 93480
    }
  },
  {
    "ph": "f", "id": 93480, "pid": 76337, "tid": -914061504, "ts": 1716454222764539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222825726, "dur": 14,
    "args": {
      "External id": 93484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93484,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93484, "pid": 5, "tid": 7, "ts": 1716454222825726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764556, "dur": 12,
    "args": {
      "External id": 93484,
      "cbid": 211, "correlation": 93484
    }
  },
  {
    "ph": "s", "id": 93484, "pid": 76337, "tid": -914061504, "ts": 1716454222764556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222825742, "dur": 4,
    "args": {
      "External id": 93486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93486, "pid": 5, "tid": 7, "ts": 1716454222825742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764574, "dur": 6,
    "args": {
      "External id": 93486,
      "cbid": 211, "correlation": 93486
    }
  },
  {
    "ph": "s", "id": 93486, "pid": 76337, "tid": -914061504, "ts": 1716454222764574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764583, "dur": 0,
    "args": {
      "External id": 93487,
      "cbid": 51, "correlation": 93487
    }
  },
  {
    "ph": "s", "id": 93487, "pid": 76337, "tid": -914061504, "ts": 1716454222764583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222825747, "dur": 95,
    "args": {
      "External id": 93488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93488,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 93488, "pid": 5, "tid": 7, "ts": 1716454222825747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764584, "dur": 5,
    "args": {
      "External id": 93488,
      "cbid": 211, "correlation": 93488
    }
  },
  {
    "ph": "s", "id": 93488, "pid": 76337, "tid": -914061504, "ts": 1716454222764584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222825843, "dur": 16,
    "args": {
      "External id": 93493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93493,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93493, "pid": 5, "tid": 7, "ts": 1716454222825843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764611, "dur": 9,
    "args": {
      "External id": 93493,
      "cbid": 211, "correlation": 93493
    }
  },
  {
    "ph": "s", "id": 93493, "pid": 76337, "tid": -914061504, "ts": 1716454222764611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222825860, "dur": 12,
    "args": {
      "External id": 93501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93501,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93501, "pid": 5, "tid": 7, "ts": 1716454222825860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764643, "dur": 8,
    "args": {
      "External id": 93501,
      "cbid": 211, "correlation": 93501
    }
  },
  {
    "ph": "s", "id": 93501, "pid": 76337, "tid": -914061504, "ts": 1716454222764643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222825873, "dur": 24,
    "args": {
      "External id": 93510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93510,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93510, "pid": 5, "tid": 7, "ts": 1716454222825873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764684, "dur": 10,
    "args": {
      "External id": 93510,
      "cbid": 211, "correlation": 93510
    }
  },
  {
    "ph": "s", "id": 93510, "pid": 76337, "tid": -914061504, "ts": 1716454222764684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222825898, "dur": 23,
    "args": {
      "External id": 93530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93530,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 93530, "pid": 5, "tid": 7, "ts": 1716454222825898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764759, "dur": 12,
    "args": {
      "External id": 93530,
      "cbid": 211, "correlation": 93530
    }
  },
  {
    "ph": "s", "id": 93530, "pid": 76337, "tid": -914061504, "ts": 1716454222764759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222825923, "dur": 5,
    "args": {
      "External id": 93542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93542,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 93542, "pid": 5, "tid": 7, "ts": 1716454222825923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764780, "dur": 7,
    "args": {
      "External id": 93542,
      "cbid": 211, "correlation": 93542
    }
  },
  {
    "ph": "s", "id": 93542, "pid": 76337, "tid": -914061504, "ts": 1716454222764780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222825929, "dur": 24,
    "args": {
      "External id": 93545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93545, "pid": 5, "tid": 7, "ts": 1716454222825929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764799, "dur": 7,
    "args": {
      "External id": 93545,
      "cbid": 211, "correlation": 93545
    }
  },
  {
    "ph": "s", "id": 93545, "pid": 76337, "tid": -914061504, "ts": 1716454222764799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222825954, "dur": 17,
    "args": {
      "External id": 93554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93554,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93554, "pid": 5, "tid": 7, "ts": 1716454222825954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764839, "dur": 10,
    "args": {
      "External id": 93554,
      "cbid": 211, "correlation": 93554
    }
  },
  {
    "ph": "s", "id": 93554, "pid": 76337, "tid": -914061504, "ts": 1716454222764839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764892, "dur": 0,
    "args": {
      "External id": 93564,
      "cbid": 317, "correlation": 93564
    }
  },
  {
    "ph": "f", "id": 93564, "pid": 76337, "tid": -914061504, "ts": 1716454222764892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764893, "dur": 0,
    "args": {
      "External id": 93565,
      "cbid": 203, "correlation": 93565
    }
  },
  {
    "ph": "f", "id": 93565, "pid": 76337, "tid": -914061504, "ts": 1716454222764893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764893, "dur": 0,
    "args": {
      "External id": 93566,
      "cbid": 205, "correlation": 93566
    }
  },
  {
    "ph": "f", "id": 93566, "pid": 76337, "tid": -914061504, "ts": 1716454222764893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222825972, "dur": 17,
    "args": {
      "External id": 93570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93570,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93570, "pid": 5, "tid": 7, "ts": 1716454222825972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764908, "dur": 12,
    "args": {
      "External id": 93570,
      "cbid": 211, "correlation": 93570
    }
  },
  {
    "ph": "s", "id": 93570, "pid": 76337, "tid": -914061504, "ts": 1716454222764908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222825991, "dur": 236,
    "args": {
      "External id": 93572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93572,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93572, "pid": 5, "tid": 7, "ts": 1716454222825991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764923, "dur": 5,
    "args": {
      "External id": 93572,
      "cbid": 211, "correlation": 93572
    }
  },
  {
    "ph": "s", "id": 93572, "pid": 76337, "tid": -914061504, "ts": 1716454222764923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222826230, "dur": 1,
    "args": {
      "External id": 93574,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 93574,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 93574, "pid": 5, "tid": 7, "ts": 1716454222826230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764935, "dur": 8,
    "args": {
      "External id": 93574,
      "cbid": 51, "correlation": 93574
    }
  },
  {
    "ph": "s", "id": 93574, "pid": 76337, "tid": -914061504, "ts": 1716454222764935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222826233, "dur": 804,
    "args": {
      "External id": 93575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93575,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93575, "pid": 5, "tid": 7, "ts": 1716454222826233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764944, "dur": 6,
    "args": {
      "External id": 93575,
      "cbid": 211, "correlation": 93575
    }
  },
  {
    "ph": "s", "id": 93575, "pid": 76337, "tid": -914061504, "ts": 1716454222764944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222827039, "dur": 13,
    "args": {
      "External id": 93577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93577,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93577, "pid": 5, "tid": 7, "ts": 1716454222827039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764955, "dur": 6,
    "args": {
      "External id": 93577,
      "cbid": 211, "correlation": 93577
    }
  },
  {
    "ph": "s", "id": 93577, "pid": 76337, "tid": -914061504, "ts": 1716454222764955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222827053, "dur": 14,
    "args": {
      "External id": 93583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93583,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93583, "pid": 5, "tid": 7, "ts": 1716454222827053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222764991, "dur": 9,
    "args": {
      "External id": 93583,
      "cbid": 211, "correlation": 93583
    }
  },
  {
    "ph": "s", "id": 93583, "pid": 76337, "tid": -914061504, "ts": 1716454222764991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222827069, "dur": 3,
    "args": {
      "External id": 93591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93591,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 93591, "pid": 5, "tid": 7, "ts": 1716454222827069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765038, "dur": 10,
    "args": {
      "External id": 93591,
      "cbid": 211, "correlation": 93591
    }
  },
  {
    "ph": "s", "id": 93591, "pid": 76337, "tid": -914061504, "ts": 1716454222765038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765105, "dur": 1,
    "args": {
      "External id": 93607,
      "cbid": 251, "correlation": 93607
    }
  },
  {
    "ph": "f", "id": 93607, "pid": 76337, "tid": -914061504, "ts": 1716454222765105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765110, "dur": 0,
    "args": {
      "External id": 93609,
      "cbid": 251, "correlation": 93609
    }
  },
  {
    "ph": "f", "id": 93609, "pid": 76337, "tid": -914061504, "ts": 1716454222765110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222827073, "dur": 13,
    "args": {
      "External id": 93610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93610,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93610, "pid": 5, "tid": 7, "ts": 1716454222827073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765112, "dur": 11,
    "args": {
      "External id": 93610,
      "cbid": 211, "correlation": 93610
    }
  },
  {
    "ph": "s", "id": 93610, "pid": 76337, "tid": -914061504, "ts": 1716454222765112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222827088, "dur": 5,
    "args": {
      "External id": 93612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93612,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93612, "pid": 5, "tid": 7, "ts": 1716454222827088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765125, "dur": 6,
    "args": {
      "External id": 93612,
      "cbid": 211, "correlation": 93612
    }
  },
  {
    "ph": "s", "id": 93612, "pid": 76337, "tid": -914061504, "ts": 1716454222765125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222827094, "dur": 17,
    "args": {
      "External id": 93622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93622,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93622, "pid": 5, "tid": 7, "ts": 1716454222827094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765185, "dur": 12,
    "args": {
      "External id": 93622,
      "cbid": 211, "correlation": 93622
    }
  },
  {
    "ph": "s", "id": 93622, "pid": 76337, "tid": -914061504, "ts": 1716454222765185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222827113, "dur": 19,
    "args": {
      "External id": 93642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93642,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 93642, "pid": 5, "tid": 7, "ts": 1716454222827113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765252, "dur": 11,
    "args": {
      "External id": 93642,
      "cbid": 211, "correlation": 93642
    }
  },
  {
    "ph": "s", "id": 93642, "pid": 76337, "tid": -914061504, "ts": 1716454222765252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222827132, "dur": 4,
    "args": {
      "External id": 93654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93654,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 93654, "pid": 5, "tid": 7, "ts": 1716454222827132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765273, "dur": 6,
    "args": {
      "External id": 93654,
      "cbid": 211, "correlation": 93654
    }
  },
  {
    "ph": "s", "id": 93654, "pid": 76337, "tid": -914061504, "ts": 1716454222765273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222827137, "dur": 16,
    "args": {
      "External id": 93657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93657,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93657, "pid": 5, "tid": 7, "ts": 1716454222827137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765290, "dur": 6,
    "args": {
      "External id": 93657,
      "cbid": 211, "correlation": 93657
    }
  },
  {
    "ph": "s", "id": 93657, "pid": 76337, "tid": -914061504, "ts": 1716454222765290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222827155, "dur": 11,
    "args": {
      "External id": 93666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93666,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93666, "pid": 5, "tid": 7, "ts": 1716454222827155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765331, "dur": 10,
    "args": {
      "External id": 93666,
      "cbid": 211, "correlation": 93666
    }
  },
  {
    "ph": "s", "id": 93666, "pid": 76337, "tid": -914061504, "ts": 1716454222765331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765394, "dur": 0,
    "args": {
      "External id": 93676,
      "cbid": 317, "correlation": 93676
    }
  },
  {
    "ph": "f", "id": 93676, "pid": 76337, "tid": -914061504, "ts": 1716454222765394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765395, "dur": 0,
    "args": {
      "External id": 93677,
      "cbid": 203, "correlation": 93677
    }
  },
  {
    "ph": "f", "id": 93677, "pid": 76337, "tid": -914061504, "ts": 1716454222765395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765395, "dur": 0,
    "args": {
      "External id": 93678,
      "cbid": 205, "correlation": 93678
    }
  },
  {
    "ph": "f", "id": 93678, "pid": 76337, "tid": -914061504, "ts": 1716454222765395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222827167, "dur": 11,
    "args": {
      "External id": 93682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93682,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93682, "pid": 5, "tid": 7, "ts": 1716454222827167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765409, "dur": 12,
    "args": {
      "External id": 93682,
      "cbid": 211, "correlation": 93682
    }
  },
  {
    "ph": "s", "id": 93682, "pid": 76337, "tid": -914061504, "ts": 1716454222765409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222827180, "dur": 161,
    "args": {
      "External id": 93684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93684,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93684, "pid": 5, "tid": 7, "ts": 1716454222827180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765423, "dur": 5,
    "args": {
      "External id": 93684,
      "cbid": 211, "correlation": 93684
    }
  },
  {
    "ph": "s", "id": 93684, "pid": 76337, "tid": -914061504, "ts": 1716454222765423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222827342, "dur": 1,
    "args": {
      "External id": 93686,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 93686,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5996252342286071
    }
  },
  {
    "ph": "f", "id": 93686, "pid": 5, "tid": 7, "ts": 1716454222827342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765434, "dur": 6,
    "args": {
      "External id": 93686,
      "cbid": 51, "correlation": 93686
    }
  },
  {
    "ph": "s", "id": 93686, "pid": 76337, "tid": -914061504, "ts": 1716454222765434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222827346, "dur": 639,
    "args": {
      "External id": 93687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93687,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93687, "pid": 5, "tid": 7, "ts": 1716454222827346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765442, "dur": 6,
    "args": {
      "External id": 93687,
      "cbid": 211, "correlation": 93687
    }
  },
  {
    "ph": "s", "id": 93687, "pid": 76337, "tid": -914061504, "ts": 1716454222765442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222827987, "dur": 12,
    "args": {
      "External id": 93689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93689,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93689, "pid": 5, "tid": 7, "ts": 1716454222827987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765452, "dur": 5,
    "args": {
      "External id": 93689,
      "cbid": 211, "correlation": 93689
    }
  },
  {
    "ph": "s", "id": 93689, "pid": 76337, "tid": -914061504, "ts": 1716454222765452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222828000, "dur": 14,
    "args": {
      "External id": 93695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93695,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93695, "pid": 5, "tid": 7, "ts": 1716454222828000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765480, "dur": 9,
    "args": {
      "External id": 93695,
      "cbid": 211, "correlation": 93695
    }
  },
  {
    "ph": "s", "id": 93695, "pid": 76337, "tid": -914061504, "ts": 1716454222765480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765539, "dur": 0,
    "args": {
      "External id": 93705,
      "cbid": 317, "correlation": 93705
    }
  },
  {
    "ph": "f", "id": 93705, "pid": 76337, "tid": -914061504, "ts": 1716454222765539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765539, "dur": 0,
    "args": {
      "External id": 93706,
      "cbid": 203, "correlation": 93706
    }
  },
  {
    "ph": "f", "id": 93706, "pid": 76337, "tid": -914061504, "ts": 1716454222765539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765540, "dur": 0,
    "args": {
      "External id": 93707,
      "cbid": 205, "correlation": 93707
    }
  },
  {
    "ph": "f", "id": 93707, "pid": 76337, "tid": -914061504, "ts": 1716454222765540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222828015, "dur": 18,
    "args": {
      "External id": 93711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93711,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93711, "pid": 5, "tid": 7, "ts": 1716454222828015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765554, "dur": 12,
    "args": {
      "External id": 93711,
      "cbid": 211, "correlation": 93711
    }
  },
  {
    "ph": "s", "id": 93711, "pid": 76337, "tid": -914061504, "ts": 1716454222765554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222828035, "dur": 4,
    "args": {
      "External id": 93713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93713, "pid": 5, "tid": 7, "ts": 1716454222828035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765570, "dur": 6,
    "args": {
      "External id": 93713,
      "cbid": 211, "correlation": 93713
    }
  },
  {
    "ph": "s", "id": 93713, "pid": 76337, "tid": -914061504, "ts": 1716454222765570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765579, "dur": 0,
    "args": {
      "External id": 93714,
      "cbid": 51, "correlation": 93714
    }
  },
  {
    "ph": "s", "id": 93714, "pid": 76337, "tid": -914061504, "ts": 1716454222765579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222828040, "dur": 132,
    "args": {
      "External id": 93715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93715,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 93715, "pid": 5, "tid": 7, "ts": 1716454222828040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765580, "dur": 5,
    "args": {
      "External id": 93715,
      "cbid": 211, "correlation": 93715
    }
  },
  {
    "ph": "s", "id": 93715, "pid": 76337, "tid": -914061504, "ts": 1716454222765580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222828174, "dur": 15,
    "args": {
      "External id": 93720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93720,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93720, "pid": 5, "tid": 7, "ts": 1716454222828174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765606, "dur": 8,
    "args": {
      "External id": 93720,
      "cbid": 211, "correlation": 93720
    }
  },
  {
    "ph": "s", "id": 93720, "pid": 76337, "tid": -914061504, "ts": 1716454222765606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222828189, "dur": 13,
    "args": {
      "External id": 93728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93728,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93728, "pid": 5, "tid": 7, "ts": 1716454222828189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765635, "dur": 8,
    "args": {
      "External id": 93728,
      "cbid": 211, "correlation": 93728
    }
  },
  {
    "ph": "s", "id": 93728, "pid": 76337, "tid": -914061504, "ts": 1716454222765635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222828204, "dur": 10,
    "args": {
      "External id": 93736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93736,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93736, "pid": 5, "tid": 7, "ts": 1716454222828204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765664, "dur": 9,
    "args": {
      "External id": 93736,
      "cbid": 211, "correlation": 93736
    }
  },
  {
    "ph": "s", "id": 93736, "pid": 76337, "tid": -914061504, "ts": 1716454222765664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222828215, "dur": 18,
    "args": {
      "External id": 93756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93756,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 93756, "pid": 5, "tid": 7, "ts": 1716454222828215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765748, "dur": 12,
    "args": {
      "External id": 93756,
      "cbid": 211, "correlation": 93756
    }
  },
  {
    "ph": "s", "id": 93756, "pid": 76337, "tid": -914061504, "ts": 1716454222765748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222828235, "dur": 4,
    "args": {
      "External id": 93768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93768,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 93768, "pid": 5, "tid": 7, "ts": 1716454222828235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765770, "dur": 7,
    "args": {
      "External id": 93768,
      "cbid": 211, "correlation": 93768
    }
  },
  {
    "ph": "s", "id": 93768, "pid": 76337, "tid": -914061504, "ts": 1716454222765770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222828240, "dur": 16,
    "args": {
      "External id": 93771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93771, "pid": 5, "tid": 7, "ts": 1716454222828240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765788, "dur": 6,
    "args": {
      "External id": 93771,
      "cbid": 211, "correlation": 93771
    }
  },
  {
    "ph": "s", "id": 93771, "pid": 76337, "tid": -914061504, "ts": 1716454222765788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765845, "dur": 0,
    "args": {
      "External id": 93782,
      "cbid": 317, "correlation": 93782
    }
  },
  {
    "ph": "f", "id": 93782, "pid": 76337, "tid": -914061504, "ts": 1716454222765845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765845, "dur": 0,
    "args": {
      "External id": 93783,
      "cbid": 203, "correlation": 93783
    }
  },
  {
    "ph": "f", "id": 93783, "pid": 76337, "tid": -914061504, "ts": 1716454222765845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765846, "dur": 0,
    "args": {
      "External id": 93784,
      "cbid": 205, "correlation": 93784
    }
  },
  {
    "ph": "f", "id": 93784, "pid": 76337, "tid": -914061504, "ts": 1716454222765846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222828258, "dur": 12,
    "args": {
      "External id": 93788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93788, "pid": 5, "tid": 7, "ts": 1716454222828258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765859, "dur": 12,
    "args": {
      "External id": 93788,
      "cbid": 211, "correlation": 93788
    }
  },
  {
    "ph": "s", "id": 93788, "pid": 76337, "tid": -914061504, "ts": 1716454222765859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222828271, "dur": 3,
    "args": {
      "External id": 93790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93790, "pid": 5, "tid": 7, "ts": 1716454222828271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765876, "dur": 6,
    "args": {
      "External id": 93790,
      "cbid": 211, "correlation": 93790
    }
  },
  {
    "ph": "s", "id": 93790, "pid": 76337, "tid": -914061504, "ts": 1716454222765876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765884, "dur": 0,
    "args": {
      "External id": 93791,
      "cbid": 51, "correlation": 93791
    }
  },
  {
    "ph": "s", "id": 93791, "pid": 76337, "tid": -914061504, "ts": 1716454222765884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222828276, "dur": 89,
    "args": {
      "External id": 93792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93792,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 93792, "pid": 5, "tid": 7, "ts": 1716454222828276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765884, "dur": 5,
    "args": {
      "External id": 93792,
      "cbid": 211, "correlation": 93792
    }
  },
  {
    "ph": "s", "id": 93792, "pid": 76337, "tid": -914061504, "ts": 1716454222765884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222828366, "dur": 16,
    "args": {
      "External id": 93797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93797, "pid": 5, "tid": 7, "ts": 1716454222828366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222765912, "dur": 9,
    "args": {
      "External id": 93797,
      "cbid": 211, "correlation": 93797
    }
  },
  {
    "ph": "s", "id": 93797, "pid": 76337, "tid": -914061504, "ts": 1716454222765912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222828383, "dur": 83,
    "args": {
      "External id": 93806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93806,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93806, "pid": 5, "tid": 7, "ts": 1716454222828383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766009, "dur": 15,
    "args": {
      "External id": 93806,
      "cbid": 211, "correlation": 93806
    }
  },
  {
    "ph": "s", "id": 93806, "pid": 76337, "tid": -914061504, "ts": 1716454222766009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222828467, "dur": 30,
    "args": {
      "External id": 93828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93828,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93828, "pid": 5, "tid": 7, "ts": 1716454222828467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766070, "dur": 11,
    "args": {
      "External id": 93828,
      "cbid": 211, "correlation": 93828
    }
  },
  {
    "ph": "s", "id": 93828, "pid": 76337, "tid": -914061504, "ts": 1716454222766070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766165, "dur": 2,
    "args": {
      "External id": 93839,
      "cbid": 251, "correlation": 93839
    }
  },
  {
    "ph": "f", "id": 93839, "pid": 76337, "tid": -914061504, "ts": 1716454222766165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222828498, "dur": 162,
    "args": {
      "External id": 93840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93840,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93840, "pid": 5, "tid": 7, "ts": 1716454222828498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766170, "dur": 14,
    "args": {
      "External id": 93840,
      "cbid": 211, "correlation": 93840
    }
  },
  {
    "ph": "s", "id": 93840, "pid": 76337, "tid": -914061504, "ts": 1716454222766170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766241, "dur": 1,
    "args": {
      "External id": 93851,
      "cbid": 251, "correlation": 93851
    }
  },
  {
    "ph": "f", "id": 93851, "pid": 76337, "tid": -914061504, "ts": 1716454222766241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222828661, "dur": 155,
    "args": {
      "External id": 93852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93852,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93852, "pid": 5, "tid": 7, "ts": 1716454222828661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766245, "dur": 12,
    "args": {
      "External id": 93852,
      "cbid": 211, "correlation": 93852
    }
  },
  {
    "ph": "s", "id": 93852, "pid": 76337, "tid": -914061504, "ts": 1716454222766245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766310, "dur": 1,
    "args": {
      "External id": 93863,
      "cbid": 251, "correlation": 93863
    }
  },
  {
    "ph": "f", "id": 93863, "pid": 76337, "tid": -914061504, "ts": 1716454222766310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222828817, "dur": 157,
    "args": {
      "External id": 93864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93864,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93864, "pid": 5, "tid": 7, "ts": 1716454222828817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766314, "dur": 11,
    "args": {
      "External id": 93864,
      "cbid": 211, "correlation": 93864
    }
  },
  {
    "ph": "s", "id": 93864, "pid": 76337, "tid": -914061504, "ts": 1716454222766314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222828975, "dur": 329,
    "args": {
      "External id": 93889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93889,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93889, "pid": 5, "tid": 7, "ts": 1716454222828975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766403, "dur": 14,
    "args": {
      "External id": 93889,
      "cbid": 211, "correlation": 93889
    }
  },
  {
    "ph": "s", "id": 93889, "pid": 76337, "tid": -914061504, "ts": 1716454222766403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766507, "dur": 1,
    "args": {
      "External id": 93907,
      "cbid": 251, "correlation": 93907
    }
  },
  {
    "ph": "f", "id": 93907, "pid": 76337, "tid": -914061504, "ts": 1716454222766507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222829306, "dur": 164,
    "args": {
      "External id": 93909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93909,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93909, "pid": 5, "tid": 7, "ts": 1716454222829306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766513, "dur": 14,
    "args": {
      "External id": 93909,
      "cbid": 211, "correlation": 93909
    }
  },
  {
    "ph": "s", "id": 93909, "pid": 76337, "tid": -914061504, "ts": 1716454222766513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222829471, "dur": 19,
    "args": {
      "External id": 93917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93917,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93917, "pid": 5, "tid": 7, "ts": 1716454222829471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766583, "dur": 12,
    "args": {
      "External id": 93917,
      "cbid": 211, "correlation": 93917
    }
  },
  {
    "ph": "s", "id": 93917, "pid": 76337, "tid": -914061504, "ts": 1716454222766583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222829492, "dur": 28,
    "args": {
      "External id": 93925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93925,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93925, "pid": 5, "tid": 7, "ts": 1716454222829492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766622, "dur": 9,
    "args": {
      "External id": 93925,
      "cbid": 211, "correlation": 93925
    }
  },
  {
    "ph": "s", "id": 93925, "pid": 76337, "tid": -914061504, "ts": 1716454222766622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222829521, "dur": 18,
    "args": {
      "External id": 93936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93936, "pid": 5, "tid": 7, "ts": 1716454222829521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766697, "dur": 13,
    "args": {
      "External id": 93936,
      "cbid": 211, "correlation": 93936
    }
  },
  {
    "ph": "s", "id": 93936, "pid": 76337, "tid": -914061504, "ts": 1716454222766697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222829540, "dur": 16,
    "args": {
      "External id": 93958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93958,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 93958, "pid": 5, "tid": 7, "ts": 1716454222829540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766729, "dur": 10,
    "args": {
      "External id": 93958,
      "cbid": 211, "correlation": 93958
    }
  },
  {
    "ph": "s", "id": 93958, "pid": 76337, "tid": -914061504, "ts": 1716454222766729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766817, "dur": 1,
    "args": {
      "External id": 93969,
      "cbid": 251, "correlation": 93969
    }
  },
  {
    "ph": "f", "id": 93969, "pid": 76337, "tid": -914061504, "ts": 1716454222766817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222829558, "dur": 88,
    "args": {
      "External id": 93970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93970,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 93970, "pid": 5, "tid": 7, "ts": 1716454222829558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766822, "dur": 13,
    "args": {
      "External id": 93970,
      "cbid": 211, "correlation": 93970
    }
  },
  {
    "ph": "s", "id": 93970, "pid": 76337, "tid": -914061504, "ts": 1716454222766822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766895, "dur": 1,
    "args": {
      "External id": 93981,
      "cbid": 251, "correlation": 93981
    }
  },
  {
    "ph": "f", "id": 93981, "pid": 76337, "tid": -914061504, "ts": 1716454222766895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766899, "dur": 0,
    "args": {
      "External id": 93982,
      "cbid": 251, "correlation": 93982
    }
  },
  {
    "ph": "f", "id": 93982, "pid": 76337, "tid": -914061504, "ts": 1716454222766899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222829647, "dur": 12,
    "args": {
      "External id": 93983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93983,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93983, "pid": 5, "tid": 7, "ts": 1716454222829647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766901, "dur": 12,
    "args": {
      "External id": 93983,
      "cbid": 211, "correlation": 93983
    }
  },
  {
    "ph": "s", "id": 93983, "pid": 76337, "tid": -914061504, "ts": 1716454222766901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222829660, "dur": 6,
    "args": {
      "External id": 93985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93985,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93985, "pid": 5, "tid": 7, "ts": 1716454222829660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766915, "dur": 7,
    "args": {
      "External id": 93985,
      "cbid": 211, "correlation": 93985
    }
  },
  {
    "ph": "s", "id": 93985, "pid": 76337, "tid": -914061504, "ts": 1716454222766915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766981, "dur": 1,
    "args": {
      "External id": 93996,
      "cbid": 251, "correlation": 93996
    }
  },
  {
    "ph": "f", "id": 93996, "pid": 76337, "tid": -914061504, "ts": 1716454222766981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766985, "dur": 0,
    "args": {
      "External id": 93997,
      "cbid": 251, "correlation": 93997
    }
  },
  {
    "ph": "f", "id": 93997, "pid": 76337, "tid": -914061504, "ts": 1716454222766985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222829667, "dur": 8,
    "args": {
      "External id": 93998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 93998,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 93998, "pid": 5, "tid": 7, "ts": 1716454222829667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222766987, "dur": 12,
    "args": {
      "External id": 93998,
      "cbid": 211, "correlation": 93998
    }
  },
  {
    "ph": "s", "id": 93998, "pid": 76337, "tid": -914061504, "ts": 1716454222766987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222829677, "dur": 4,
    "args": {
      "External id": 94000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94000,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94000, "pid": 5, "tid": 7, "ts": 1716454222829677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767001, "dur": 6,
    "args": {
      "External id": 94000,
      "cbid": 211, "correlation": 94000
    }
  },
  {
    "ph": "s", "id": 94000, "pid": 76337, "tid": -914061504, "ts": 1716454222767001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222829681, "dur": 55,
    "args": {
      "External id": 94025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94025,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94025, "pid": 5, "tid": 7, "ts": 1716454222829681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767079, "dur": 13,
    "args": {
      "External id": 94025,
      "cbid": 211, "correlation": 94025
    }
  },
  {
    "ph": "s", "id": 94025, "pid": 76337, "tid": -914061504, "ts": 1716454222767079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767180, "dur": 2,
    "args": {
      "External id": 94043,
      "cbid": 251, "correlation": 94043
    }
  },
  {
    "ph": "f", "id": 94043, "pid": 76337, "tid": -914061504, "ts": 1716454222767180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222829738, "dur": 90,
    "args": {
      "External id": 94045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94045,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 94045, "pid": 5, "tid": 7, "ts": 1716454222829738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767186, "dur": 15,
    "args": {
      "External id": 94045,
      "cbid": 211, "correlation": 94045
    }
  },
  {
    "ph": "s", "id": 94045, "pid": 76337, "tid": -914061504, "ts": 1716454222767186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222829829, "dur": 9,
    "args": {
      "External id": 94053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94053,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94053, "pid": 5, "tid": 7, "ts": 1716454222829829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767259, "dur": 12,
    "args": {
      "External id": 94053,
      "cbid": 211, "correlation": 94053
    }
  },
  {
    "ph": "s", "id": 94053, "pid": 76337, "tid": -914061504, "ts": 1716454222767259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222829840, "dur": 21,
    "args": {
      "External id": 94061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94061, "pid": 5, "tid": 7, "ts": 1716454222829840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767300, "dur": 10,
    "args": {
      "External id": 94061,
      "cbid": 211, "correlation": 94061
    }
  },
  {
    "ph": "s", "id": 94061, "pid": 76337, "tid": -914061504, "ts": 1716454222767300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222829863, "dur": 19,
    "args": {
      "External id": 94083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94083,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94083, "pid": 5, "tid": 7, "ts": 1716454222829863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767351, "dur": 10,
    "args": {
      "External id": 94083,
      "cbid": 211, "correlation": 94083
    }
  },
  {
    "ph": "s", "id": 94083, "pid": 76337, "tid": -914061504, "ts": 1716454222767351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767440, "dur": 1,
    "args": {
      "External id": 94099,
      "cbid": 251, "correlation": 94099
    }
  },
  {
    "ph": "f", "id": 94099, "pid": 76337, "tid": -914061504, "ts": 1716454222767440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767444, "dur": 0,
    "args": {
      "External id": 94101,
      "cbid": 251, "correlation": 94101
    }
  },
  {
    "ph": "f", "id": 94101, "pid": 76337, "tid": -914061504, "ts": 1716454222767444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222829883, "dur": 494,
    "args": {
      "External id": 94102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94102,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94102, "pid": 5, "tid": 7, "ts": 1716454222829883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767446, "dur": 13,
    "args": {
      "External id": 94102,
      "cbid": 211, "correlation": 94102
    }
  },
  {
    "ph": "s", "id": 94102, "pid": 76337, "tid": -914061504, "ts": 1716454222767446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222830378, "dur": 64,
    "args": {
      "External id": 94110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94110, "pid": 5, "tid": 7, "ts": 1716454222830378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767516, "dur": 13,
    "args": {
      "External id": 94110,
      "cbid": 211, "correlation": 94110
    }
  },
  {
    "ph": "s", "id": 94110, "pid": 76337, "tid": -914061504, "ts": 1716454222767516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222830443, "dur": 66,
    "args": {
      "External id": 94118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94118, "pid": 5, "tid": 7, "ts": 1716454222830443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767550, "dur": 9,
    "args": {
      "External id": 94118,
      "cbid": 211, "correlation": 94118
    }
  },
  {
    "ph": "s", "id": 94118, "pid": 76337, "tid": -914061504, "ts": 1716454222767550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767629, "dur": 1,
    "args": {
      "External id": 94134,
      "cbid": 251, "correlation": 94134
    }
  },
  {
    "ph": "f", "id": 94134, "pid": 76337, "tid": -914061504, "ts": 1716454222767629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222830512, "dur": 1,
    "args": {
      "External id": 94136,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 94136,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 94136, "pid": 5, "tid": 7, "ts": 1716454222830512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767634, "dur": 11,
    "args": {
      "External id": 94136,
      "cbid": 51, "correlation": 94136
    }
  },
  {
    "ph": "s", "id": 94136, "pid": 76337, "tid": -914061504, "ts": 1716454222767634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222830516, "dur": 267,
    "args": {
      "External id": 94137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94137,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 94137, "pid": 5, "tid": 7, "ts": 1716454222830516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767647, "dur": 11,
    "args": {
      "External id": 94137,
      "cbid": 211, "correlation": 94137
    }
  },
  {
    "ph": "s", "id": 94137, "pid": 76337, "tid": -914061504, "ts": 1716454222767647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222830784, "dur": 15,
    "args": {
      "External id": 94145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94145,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94145, "pid": 5, "tid": 7, "ts": 1716454222830784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767689, "dur": 11,
    "args": {
      "External id": 94145,
      "cbid": 211, "correlation": 94145
    }
  },
  {
    "ph": "s", "id": 94145, "pid": 76337, "tid": -914061504, "ts": 1716454222767689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222830800, "dur": 37,
    "args": {
      "External id": 94156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94156,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94156, "pid": 5, "tid": 7, "ts": 1716454222830800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767760, "dur": 12,
    "args": {
      "External id": 94156,
      "cbid": 211, "correlation": 94156
    }
  },
  {
    "ph": "s", "id": 94156, "pid": 76337, "tid": -914061504, "ts": 1716454222767760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767828, "dur": 0,
    "args": {
      "External id": 94168,
      "cbid": 317, "correlation": 94168
    }
  },
  {
    "ph": "f", "id": 94168, "pid": 76337, "tid": -914061504, "ts": 1716454222767828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767829, "dur": 0,
    "args": {
      "External id": 94169,
      "cbid": 203, "correlation": 94169
    }
  },
  {
    "ph": "f", "id": 94169, "pid": 76337, "tid": -914061504, "ts": 1716454222767829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767830, "dur": 0,
    "args": {
      "External id": 94170,
      "cbid": 205, "correlation": 94170
    }
  },
  {
    "ph": "f", "id": 94170, "pid": 76337, "tid": -914061504, "ts": 1716454222767830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222830838, "dur": 13,
    "args": {
      "External id": 94174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94174,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94174, "pid": 5, "tid": 7, "ts": 1716454222830838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767846, "dur": 12,
    "args": {
      "External id": 94174,
      "cbid": 211, "correlation": 94174
    }
  },
  {
    "ph": "s", "id": 94174, "pid": 76337, "tid": -914061504, "ts": 1716454222767846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222830852, "dur": 4,
    "args": {
      "External id": 94176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 94176, "pid": 5, "tid": 7, "ts": 1716454222830852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767864, "dur": 7,
    "args": {
      "External id": 94176,
      "cbid": 211, "correlation": 94176
    }
  },
  {
    "ph": "s", "id": 94176, "pid": 76337, "tid": -914061504, "ts": 1716454222767864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767873, "dur": 0,
    "args": {
      "External id": 94177,
      "cbid": 51, "correlation": 94177
    }
  },
  {
    "ph": "s", "id": 94177, "pid": 76337, "tid": -914061504, "ts": 1716454222767873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222830857, "dur": 95,
    "args": {
      "External id": 94178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94178,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 94178, "pid": 5, "tid": 7, "ts": 1716454222830857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767874, "dur": 5,
    "args": {
      "External id": 94178,
      "cbid": 211, "correlation": 94178
    }
  },
  {
    "ph": "s", "id": 94178, "pid": 76337, "tid": -914061504, "ts": 1716454222767874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222830953, "dur": 16,
    "args": {
      "External id": 94183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94183,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94183, "pid": 5, "tid": 7, "ts": 1716454222830953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767901, "dur": 8,
    "args": {
      "External id": 94183,
      "cbid": 211, "correlation": 94183
    }
  },
  {
    "ph": "s", "id": 94183, "pid": 76337, "tid": -914061504, "ts": 1716454222767901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222830971, "dur": 12,
    "args": {
      "External id": 94191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94191,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94191, "pid": 5, "tid": 7, "ts": 1716454222830971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222767933, "dur": 8,
    "args": {
      "External id": 94191,
      "cbid": 211, "correlation": 94191
    }
  },
  {
    "ph": "s", "id": 94191, "pid": 76337, "tid": -914061504, "ts": 1716454222767933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222830984, "dur": 55,
    "args": {
      "External id": 94202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94202,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94202, "pid": 5, "tid": 7, "ts": 1716454222830984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768010, "dur": 13,
    "args": {
      "External id": 94202,
      "cbid": 211, "correlation": 94202
    }
  },
  {
    "ph": "s", "id": 94202, "pid": 76337, "tid": -914061504, "ts": 1716454222768010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768069, "dur": 0,
    "args": {
      "External id": 94212,
      "cbid": 317, "correlation": 94212
    }
  },
  {
    "ph": "f", "id": 94212, "pid": 76337, "tid": -914061504, "ts": 1716454222768069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768070, "dur": 0,
    "args": {
      "External id": 94213,
      "cbid": 203, "correlation": 94213
    }
  },
  {
    "ph": "f", "id": 94213, "pid": 76337, "tid": -914061504, "ts": 1716454222768070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768070, "dur": 0,
    "args": {
      "External id": 94214,
      "cbid": 205, "correlation": 94214
    }
  },
  {
    "ph": "f", "id": 94214, "pid": 76337, "tid": -914061504, "ts": 1716454222768070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222831041, "dur": 39,
    "args": {
      "External id": 94218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94218,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94218, "pid": 5, "tid": 7, "ts": 1716454222831041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768088, "dur": 12,
    "args": {
      "External id": 94218,
      "cbid": 211, "correlation": 94218
    }
  },
  {
    "ph": "s", "id": 94218, "pid": 76337, "tid": -914061504, "ts": 1716454222768088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222831081, "dur": 160,
    "args": {
      "External id": 94220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94220, "pid": 5, "tid": 7, "ts": 1716454222831081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768103, "dur": 5,
    "args": {
      "External id": 94220,
      "cbid": 211, "correlation": 94220
    }
  },
  {
    "ph": "s", "id": 94220, "pid": 76337, "tid": -914061504, "ts": 1716454222768103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222831242, "dur": 1965,
    "args": {
      "External id": 94222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94222,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94222, "pid": 5, "tid": 7, "ts": 1716454222831242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768116, "dur": 8,
    "args": {
      "External id": 94222,
      "cbid": 211, "correlation": 94222
    }
  },
  {
    "ph": "s", "id": 94222, "pid": 76337, "tid": -914061504, "ts": 1716454222768116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222833209, "dur": 39,
    "args": {
      "External id": 94224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94224, "pid": 5, "tid": 7, "ts": 1716454222833209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768129, "dur": 5,
    "args": {
      "External id": 94224,
      "cbid": 211, "correlation": 94224
    }
  },
  {
    "ph": "s", "id": 94224, "pid": 76337, "tid": -914061504, "ts": 1716454222768129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222833249, "dur": 58,
    "args": {
      "External id": 94230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94230, "pid": 5, "tid": 7, "ts": 1716454222833249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768158, "dur": 8,
    "args": {
      "External id": 94230,
      "cbid": 211, "correlation": 94230
    }
  },
  {
    "ph": "s", "id": 94230, "pid": 76337, "tid": -914061504, "ts": 1716454222768158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222833308, "dur": 84,
    "args": {
      "External id": 94239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94239,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94239, "pid": 5, "tid": 7, "ts": 1716454222833308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768256, "dur": 14,
    "args": {
      "External id": 94239,
      "cbid": 211, "correlation": 94239
    }
  },
  {
    "ph": "s", "id": 94239, "pid": 76337, "tid": -914061504, "ts": 1716454222768256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222833394, "dur": 73,
    "args": {
      "External id": 94259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94259,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 94259, "pid": 5, "tid": 7, "ts": 1716454222833394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768331, "dur": 11,
    "args": {
      "External id": 94259,
      "cbid": 211, "correlation": 94259
    }
  },
  {
    "ph": "s", "id": 94259, "pid": 76337, "tid": -914061504, "ts": 1716454222768331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222833468, "dur": 5,
    "args": {
      "External id": 94271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94271,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 94271, "pid": 5, "tid": 7, "ts": 1716454222833468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768352, "dur": 7,
    "args": {
      "External id": 94271,
      "cbid": 211, "correlation": 94271
    }
  },
  {
    "ph": "s", "id": 94271, "pid": 76337, "tid": -914061504, "ts": 1716454222768352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222833474, "dur": 81,
    "args": {
      "External id": 94274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94274,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94274, "pid": 5, "tid": 7, "ts": 1716454222833474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768372, "dur": 7,
    "args": {
      "External id": 94274,
      "cbid": 211, "correlation": 94274
    }
  },
  {
    "ph": "s", "id": 94274, "pid": 76337, "tid": -914061504, "ts": 1716454222768372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222833556, "dur": 53,
    "args": {
      "External id": 94283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94283,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94283, "pid": 5, "tid": 7, "ts": 1716454222833556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768413, "dur": 10,
    "args": {
      "External id": 94283,
      "cbid": 211, "correlation": 94283
    }
  },
  {
    "ph": "s", "id": 94283, "pid": 76337, "tid": -914061504, "ts": 1716454222768413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768466, "dur": 0,
    "args": {
      "External id": 94293,
      "cbid": 317, "correlation": 94293
    }
  },
  {
    "ph": "f", "id": 94293, "pid": 76337, "tid": -914061504, "ts": 1716454222768466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768466, "dur": 0,
    "args": {
      "External id": 94294,
      "cbid": 203, "correlation": 94294
    }
  },
  {
    "ph": "f", "id": 94294, "pid": 76337, "tid": -914061504, "ts": 1716454222768466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768467, "dur": 0,
    "args": {
      "External id": 94295,
      "cbid": 205, "correlation": 94295
    }
  },
  {
    "ph": "f", "id": 94295, "pid": 76337, "tid": -914061504, "ts": 1716454222768467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222833611, "dur": 57,
    "args": {
      "External id": 94299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94299,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94299, "pid": 5, "tid": 7, "ts": 1716454222833611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768484, "dur": 11,
    "args": {
      "External id": 94299,
      "cbid": 211, "correlation": 94299
    }
  },
  {
    "ph": "s", "id": 94299, "pid": 76337, "tid": -914061504, "ts": 1716454222768484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222833669, "dur": 121,
    "args": {
      "External id": 94301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94301,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94301, "pid": 5, "tid": 7, "ts": 1716454222833669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768498, "dur": 5,
    "args": {
      "External id": 94301,
      "cbid": 211, "correlation": 94301
    }
  },
  {
    "ph": "s", "id": 94301, "pid": 76337, "tid": -914061504, "ts": 1716454222768498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222833791, "dur": 1873,
    "args": {
      "External id": 94303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94303,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94303, "pid": 5, "tid": 7, "ts": 1716454222833791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768509, "dur": 6,
    "args": {
      "External id": 94303,
      "cbid": 211, "correlation": 94303
    }
  },
  {
    "ph": "s", "id": 94303, "pid": 76337, "tid": -914061504, "ts": 1716454222768509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222835665, "dur": 20,
    "args": {
      "External id": 94305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94305, "pid": 5, "tid": 7, "ts": 1716454222835665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768519, "dur": 5,
    "args": {
      "External id": 94305,
      "cbid": 211, "correlation": 94305
    }
  },
  {
    "ph": "s", "id": 94305, "pid": 76337, "tid": -914061504, "ts": 1716454222768519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222835686, "dur": 32,
    "args": {
      "External id": 94311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94311, "pid": 5, "tid": 7, "ts": 1716454222835686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768548, "dur": 9,
    "args": {
      "External id": 94311,
      "cbid": 211, "correlation": 94311
    }
  },
  {
    "ph": "s", "id": 94311, "pid": 76337, "tid": -914061504, "ts": 1716454222768548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222835720, "dur": 3,
    "args": {
      "External id": 94319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94319,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 94319, "pid": 5, "tid": 7, "ts": 1716454222835720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768592, "dur": 10,
    "args": {
      "External id": 94319,
      "cbid": 211, "correlation": 94319
    }
  },
  {
    "ph": "s", "id": 94319, "pid": 76337, "tid": -914061504, "ts": 1716454222768592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768662, "dur": 1,
    "args": {
      "External id": 94335,
      "cbid": 251, "correlation": 94335
    }
  },
  {
    "ph": "f", "id": 94335, "pid": 76337, "tid": -914061504, "ts": 1716454222768662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768668, "dur": 0,
    "args": {
      "External id": 94337,
      "cbid": 251, "correlation": 94337
    }
  },
  {
    "ph": "f", "id": 94337, "pid": 76337, "tid": -914061504, "ts": 1716454222768668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222835724, "dur": 12,
    "args": {
      "External id": 94338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94338,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 94338, "pid": 5, "tid": 7, "ts": 1716454222835724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768670, "dur": 12,
    "args": {
      "External id": 94338,
      "cbid": 211, "correlation": 94338
    }
  },
  {
    "ph": "s", "id": 94338, "pid": 76337, "tid": -914061504, "ts": 1716454222768670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222835738, "dur": 5,
    "args": {
      "External id": 94340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94340,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 94340, "pid": 5, "tid": 7, "ts": 1716454222835738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768684, "dur": 6,
    "args": {
      "External id": 94340,
      "cbid": 211, "correlation": 94340
    }
  },
  {
    "ph": "s", "id": 94340, "pid": 76337, "tid": -914061504, "ts": 1716454222768684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222835744, "dur": 29,
    "args": {
      "External id": 94350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94350, "pid": 5, "tid": 7, "ts": 1716454222835744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768745, "dur": 13,
    "args": {
      "External id": 94350,
      "cbid": 211, "correlation": 94350
    }
  },
  {
    "ph": "s", "id": 94350, "pid": 76337, "tid": -914061504, "ts": 1716454222768745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222835774, "dur": 31,
    "args": {
      "External id": 94370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94370,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 94370, "pid": 5, "tid": 7, "ts": 1716454222835774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768812, "dur": 11,
    "args": {
      "External id": 94370,
      "cbid": 211, "correlation": 94370
    }
  },
  {
    "ph": "s", "id": 94370, "pid": 76337, "tid": -914061504, "ts": 1716454222768812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222835807, "dur": 4,
    "args": {
      "External id": 94382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94382,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 94382, "pid": 5, "tid": 7, "ts": 1716454222835807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768833, "dur": 6,
    "args": {
      "External id": 94382,
      "cbid": 211, "correlation": 94382
    }
  },
  {
    "ph": "s", "id": 94382, "pid": 76337, "tid": -914061504, "ts": 1716454222768833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222835812, "dur": 30,
    "args": {
      "External id": 94385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94385,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94385, "pid": 5, "tid": 7, "ts": 1716454222835812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768851, "dur": 6,
    "args": {
      "External id": 94385,
      "cbid": 211, "correlation": 94385
    }
  },
  {
    "ph": "s", "id": 94385, "pid": 76337, "tid": -914061504, "ts": 1716454222768851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222835843, "dur": 21,
    "args": {
      "External id": 94394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94394,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94394, "pid": 5, "tid": 7, "ts": 1716454222835843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768892, "dur": 10,
    "args": {
      "External id": 94394,
      "cbid": 211, "correlation": 94394
    }
  },
  {
    "ph": "s", "id": 94394, "pid": 76337, "tid": -914061504, "ts": 1716454222768892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768956, "dur": 0,
    "args": {
      "External id": 94404,
      "cbid": 317, "correlation": 94404
    }
  },
  {
    "ph": "f", "id": 94404, "pid": 76337, "tid": -914061504, "ts": 1716454222768956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768957, "dur": 0,
    "args": {
      "External id": 94405,
      "cbid": 203, "correlation": 94405
    }
  },
  {
    "ph": "f", "id": 94405, "pid": 76337, "tid": -914061504, "ts": 1716454222768957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768958, "dur": 0,
    "args": {
      "External id": 94406,
      "cbid": 205, "correlation": 94406
    }
  },
  {
    "ph": "f", "id": 94406, "pid": 76337, "tid": -914061504, "ts": 1716454222768958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222835865, "dur": 23,
    "args": {
      "External id": 94410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94410, "pid": 5, "tid": 7, "ts": 1716454222835865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768981, "dur": 13,
    "args": {
      "External id": 94410,
      "cbid": 211, "correlation": 94410
    }
  },
  {
    "ph": "s", "id": 94410, "pid": 76337, "tid": -914061504, "ts": 1716454222768981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222835889, "dur": 43,
    "args": {
      "External id": 94412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94412, "pid": 5, "tid": 7, "ts": 1716454222835889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222768998, "dur": 5,
    "args": {
      "External id": 94412,
      "cbid": 211, "correlation": 94412
    }
  },
  {
    "ph": "s", "id": 94412, "pid": 76337, "tid": -914061504, "ts": 1716454222768998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222835933, "dur": 640,
    "args": {
      "External id": 94414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94414,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94414, "pid": 5, "tid": 7, "ts": 1716454222835933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769009, "dur": 6,
    "args": {
      "External id": 94414,
      "cbid": 211, "correlation": 94414
    }
  },
  {
    "ph": "s", "id": 94414, "pid": 76337, "tid": -914061504, "ts": 1716454222769009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222836575, "dur": 21,
    "args": {
      "External id": 94416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94416,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94416, "pid": 5, "tid": 7, "ts": 1716454222836575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769019, "dur": 5,
    "args": {
      "External id": 94416,
      "cbid": 211, "correlation": 94416
    }
  },
  {
    "ph": "s", "id": 94416, "pid": 76337, "tid": -914061504, "ts": 1716454222769019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222836597, "dur": 32,
    "args": {
      "External id": 94422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94422, "pid": 5, "tid": 7, "ts": 1716454222836597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769048, "dur": 9,
    "args": {
      "External id": 94422,
      "cbid": 211, "correlation": 94422
    }
  },
  {
    "ph": "s", "id": 94422, "pid": 76337, "tid": -914061504, "ts": 1716454222769048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769106, "dur": 0,
    "args": {
      "External id": 94432,
      "cbid": 317, "correlation": 94432
    }
  },
  {
    "ph": "f", "id": 94432, "pid": 76337, "tid": -914061504, "ts": 1716454222769106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769107, "dur": 0,
    "args": {
      "External id": 94433,
      "cbid": 203, "correlation": 94433
    }
  },
  {
    "ph": "f", "id": 94433, "pid": 76337, "tid": -914061504, "ts": 1716454222769107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769108, "dur": 0,
    "args": {
      "External id": 94434,
      "cbid": 205, "correlation": 94434
    }
  },
  {
    "ph": "f", "id": 94434, "pid": 76337, "tid": -914061504, "ts": 1716454222769108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222836631, "dur": 56,
    "args": {
      "External id": 94438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94438, "pid": 5, "tid": 7, "ts": 1716454222836631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769120, "dur": 11,
    "args": {
      "External id": 94438,
      "cbid": 211, "correlation": 94438
    }
  },
  {
    "ph": "s", "id": 94438, "pid": 76337, "tid": -914061504, "ts": 1716454222769120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222836688, "dur": 262,
    "args": {
      "External id": 94440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94440,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94440, "pid": 5, "tid": 7, "ts": 1716454222836688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769138, "dur": 8,
    "args": {
      "External id": 94440,
      "cbid": 211, "correlation": 94440
    }
  },
  {
    "ph": "s", "id": 94440, "pid": 76337, "tid": -914061504, "ts": 1716454222769138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222836951, "dur": 21,
    "args": {
      "External id": 94442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94442,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94442, "pid": 5, "tid": 7, "ts": 1716454222836951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769150, "dur": 5,
    "args": {
      "External id": 94442,
      "cbid": 211, "correlation": 94442
    }
  },
  {
    "ph": "s", "id": 94442, "pid": 76337, "tid": -914061504, "ts": 1716454222769150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222836973, "dur": 32,
    "args": {
      "External id": 94448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94448,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94448, "pid": 5, "tid": 7, "ts": 1716454222836973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769176, "dur": 8,
    "args": {
      "External id": 94448,
      "cbid": 211, "correlation": 94448
    }
  },
  {
    "ph": "s", "id": 94448, "pid": 76337, "tid": -914061504, "ts": 1716454222769176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222837007, "dur": 27,
    "args": {
      "External id": 94456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94456,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94456, "pid": 5, "tid": 7, "ts": 1716454222837007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769204, "dur": 8,
    "args": {
      "External id": 94456,
      "cbid": 211, "correlation": 94456
    }
  },
  {
    "ph": "s", "id": 94456, "pid": 76337, "tid": -914061504, "ts": 1716454222769204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222837035, "dur": 19,
    "args": {
      "External id": 94464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94464,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94464, "pid": 5, "tid": 7, "ts": 1716454222837035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769234, "dur": 9,
    "args": {
      "External id": 94464,
      "cbid": 211, "correlation": 94464
    }
  },
  {
    "ph": "s", "id": 94464, "pid": 76337, "tid": -914061504, "ts": 1716454222769234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222837055, "dur": 29,
    "args": {
      "External id": 94484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94484,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 94484, "pid": 5, "tid": 7, "ts": 1716454222837055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769317, "dur": 12,
    "args": {
      "External id": 94484,
      "cbid": 211, "correlation": 94484
    }
  },
  {
    "ph": "s", "id": 94484, "pid": 76337, "tid": -914061504, "ts": 1716454222769317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222837086, "dur": 4,
    "args": {
      "External id": 94496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94496,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 94496, "pid": 5, "tid": 7, "ts": 1716454222837086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769340, "dur": 7,
    "args": {
      "External id": 94496,
      "cbid": 211, "correlation": 94496
    }
  },
  {
    "ph": "s", "id": 94496, "pid": 76337, "tid": -914061504, "ts": 1716454222769340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222837092, "dur": 31,
    "args": {
      "External id": 94499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94499, "pid": 5, "tid": 7, "ts": 1716454222837092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769358, "dur": 6,
    "args": {
      "External id": 94499,
      "cbid": 211, "correlation": 94499
    }
  },
  {
    "ph": "s", "id": 94499, "pid": 76337, "tid": -914061504, "ts": 1716454222769358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769415, "dur": 0,
    "args": {
      "External id": 94510,
      "cbid": 317, "correlation": 94510
    }
  },
  {
    "ph": "f", "id": 94510, "pid": 76337, "tid": -914061504, "ts": 1716454222769415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769416, "dur": 0,
    "args": {
      "External id": 94511,
      "cbid": 203, "correlation": 94511
    }
  },
  {
    "ph": "f", "id": 94511, "pid": 76337, "tid": -914061504, "ts": 1716454222769416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769416, "dur": 0,
    "args": {
      "External id": 94512,
      "cbid": 205, "correlation": 94512
    }
  },
  {
    "ph": "f", "id": 94512, "pid": 76337, "tid": -914061504, "ts": 1716454222769416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222837124, "dur": 21,
    "args": {
      "External id": 94516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94516,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94516, "pid": 5, "tid": 7, "ts": 1716454222837124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769430, "dur": 11,
    "args": {
      "External id": 94516,
      "cbid": 211, "correlation": 94516
    }
  },
  {
    "ph": "s", "id": 94516, "pid": 76337, "tid": -914061504, "ts": 1716454222769430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222837147, "dur": 103,
    "args": {
      "External id": 94518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94518,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94518, "pid": 5, "tid": 7, "ts": 1716454222837147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769448, "dur": 7,
    "args": {
      "External id": 94518,
      "cbid": 211, "correlation": 94518
    }
  },
  {
    "ph": "s", "id": 94518, "pid": 76337, "tid": -914061504, "ts": 1716454222769448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222837251, "dur": 22,
    "args": {
      "External id": 94520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94520,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94520, "pid": 5, "tid": 7, "ts": 1716454222837251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769458, "dur": 5,
    "args": {
      "External id": 94520,
      "cbid": 211, "correlation": 94520
    }
  },
  {
    "ph": "s", "id": 94520, "pid": 76337, "tid": -914061504, "ts": 1716454222769458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222837274, "dur": 32,
    "args": {
      "External id": 94526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94526, "pid": 5, "tid": 7, "ts": 1716454222837274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769486, "dur": 8,
    "args": {
      "External id": 94526,
      "cbid": 211, "correlation": 94526
    }
  },
  {
    "ph": "s", "id": 94526, "pid": 76337, "tid": -914061504, "ts": 1716454222769486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222837307, "dur": 194,
    "args": {
      "External id": 94535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94535,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94535, "pid": 5, "tid": 7, "ts": 1716454222837307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769570, "dur": 14,
    "args": {
      "External id": 94535,
      "cbid": 211, "correlation": 94535
    }
  },
  {
    "ph": "s", "id": 94535, "pid": 76337, "tid": -914061504, "ts": 1716454222769570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222837503, "dur": 63,
    "args": {
      "External id": 94557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94557,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94557, "pid": 5, "tid": 7, "ts": 1716454222837503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769631, "dur": 10,
    "args": {
      "External id": 94557,
      "cbid": 211, "correlation": 94557
    }
  },
  {
    "ph": "s", "id": 94557, "pid": 76337, "tid": -914061504, "ts": 1716454222769631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769726, "dur": 2,
    "args": {
      "External id": 94568,
      "cbid": 251, "correlation": 94568
    }
  },
  {
    "ph": "f", "id": 94568, "pid": 76337, "tid": -914061504, "ts": 1716454222769726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222837567, "dur": 152,
    "args": {
      "External id": 94569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94569,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94569, "pid": 5, "tid": 7, "ts": 1716454222837567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769732, "dur": 14,
    "args": {
      "External id": 94569,
      "cbid": 211, "correlation": 94569
    }
  },
  {
    "ph": "s", "id": 94569, "pid": 76337, "tid": -914061504, "ts": 1716454222769732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769802, "dur": 1,
    "args": {
      "External id": 94580,
      "cbid": 251, "correlation": 94580
    }
  },
  {
    "ph": "f", "id": 94580, "pid": 76337, "tid": -914061504, "ts": 1716454222769802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222837721, "dur": 143,
    "args": {
      "External id": 94581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94581,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94581, "pid": 5, "tid": 7, "ts": 1716454222837721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769806, "dur": 11,
    "args": {
      "External id": 94581,
      "cbid": 211, "correlation": 94581
    }
  },
  {
    "ph": "s", "id": 94581, "pid": 76337, "tid": -914061504, "ts": 1716454222769806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769871, "dur": 1,
    "args": {
      "External id": 94592,
      "cbid": 251, "correlation": 94592
    }
  },
  {
    "ph": "f", "id": 94592, "pid": 76337, "tid": -914061504, "ts": 1716454222769871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222837866, "dur": 144,
    "args": {
      "External id": 94593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94593,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94593, "pid": 5, "tid": 7, "ts": 1716454222837866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769875, "dur": 11,
    "args": {
      "External id": 94593,
      "cbid": 211, "correlation": 94593
    }
  },
  {
    "ph": "s", "id": 94593, "pid": 76337, "tid": -914061504, "ts": 1716454222769875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222838011, "dur": 1909,
    "args": {
      "External id": 94614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94614,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 94614, "pid": 5, "tid": 7, "ts": 1716454222838011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222769962, "dur": 23,
    "args": {
      "External id": 94614,
      "cbid": 211, "correlation": 94614
    }
  },
  {
    "ph": "s", "id": 94614, "pid": 76337, "tid": -914061504, "ts": 1716454222769962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770079, "dur": 2,
    "args": {
      "External id": 94632,
      "cbid": 251, "correlation": 94632
    }
  },
  {
    "ph": "f", "id": 94632, "pid": 76337, "tid": -914061504, "ts": 1716454222770079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222839922, "dur": 147,
    "args": {
      "External id": 94634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94634,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 94634, "pid": 5, "tid": 7, "ts": 1716454222839922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770085, "dur": 14,
    "args": {
      "External id": 94634,
      "cbid": 211, "correlation": 94634
    }
  },
  {
    "ph": "s", "id": 94634, "pid": 76337, "tid": -914061504, "ts": 1716454222770085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222840070, "dur": 36,
    "args": {
      "External id": 94642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94642,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94642, "pid": 5, "tid": 7, "ts": 1716454222840070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770157, "dur": 12,
    "args": {
      "External id": 94642,
      "cbid": 211, "correlation": 94642
    }
  },
  {
    "ph": "s", "id": 94642, "pid": 76337, "tid": -914061504, "ts": 1716454222770157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222840108, "dur": 51,
    "args": {
      "External id": 94650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94650,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94650, "pid": 5, "tid": 7, "ts": 1716454222840108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770197, "dur": 8,
    "args": {
      "External id": 94650,
      "cbid": 211, "correlation": 94650
    }
  },
  {
    "ph": "s", "id": 94650, "pid": 76337, "tid": -914061504, "ts": 1716454222770197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222840160, "dur": 30,
    "args": {
      "External id": 94661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94661, "pid": 5, "tid": 7, "ts": 1716454222840160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770272, "dur": 13,
    "args": {
      "External id": 94661,
      "cbid": 211, "correlation": 94661
    }
  },
  {
    "ph": "s", "id": 94661, "pid": 76337, "tid": -914061504, "ts": 1716454222770272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222840191, "dur": 34,
    "args": {
      "External id": 94683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94683,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94683, "pid": 5, "tid": 7, "ts": 1716454222840191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770304, "dur": 8,
    "args": {
      "External id": 94683,
      "cbid": 211, "correlation": 94683
    }
  },
  {
    "ph": "s", "id": 94683, "pid": 76337, "tid": -914061504, "ts": 1716454222770304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770389, "dur": 1,
    "args": {
      "External id": 94694,
      "cbid": 251, "correlation": 94694
    }
  },
  {
    "ph": "f", "id": 94694, "pid": 76337, "tid": -914061504, "ts": 1716454222770389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222840226, "dur": 88,
    "args": {
      "External id": 94695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94695,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94695, "pid": 5, "tid": 7, "ts": 1716454222840226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770395, "dur": 13,
    "args": {
      "External id": 94695,
      "cbid": 211, "correlation": 94695
    }
  },
  {
    "ph": "s", "id": 94695, "pid": 76337, "tid": -914061504, "ts": 1716454222770395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770467, "dur": 1,
    "args": {
      "External id": 94706,
      "cbid": 251, "correlation": 94706
    }
  },
  {
    "ph": "f", "id": 94706, "pid": 76337, "tid": -914061504, "ts": 1716454222770467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770471, "dur": 0,
    "args": {
      "External id": 94707,
      "cbid": 251, "correlation": 94707
    }
  },
  {
    "ph": "f", "id": 94707, "pid": 76337, "tid": -914061504, "ts": 1716454222770471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222840315, "dur": 11,
    "args": {
      "External id": 94708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94708,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 94708, "pid": 5, "tid": 7, "ts": 1716454222840315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770473, "dur": 12,
    "args": {
      "External id": 94708,
      "cbid": 211, "correlation": 94708
    }
  },
  {
    "ph": "s", "id": 94708, "pid": 76337, "tid": -914061504, "ts": 1716454222770473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222840327, "dur": 5,
    "args": {
      "External id": 94710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94710,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 94710, "pid": 5, "tid": 7, "ts": 1716454222840327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770487, "dur": 6,
    "args": {
      "External id": 94710,
      "cbid": 211, "correlation": 94710
    }
  },
  {
    "ph": "s", "id": 94710, "pid": 76337, "tid": -914061504, "ts": 1716454222770487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770545, "dur": 1,
    "args": {
      "External id": 94721,
      "cbid": 251, "correlation": 94721
    }
  },
  {
    "ph": "f", "id": 94721, "pid": 76337, "tid": -914061504, "ts": 1716454222770545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770548, "dur": 0,
    "args": {
      "External id": 94722,
      "cbid": 251, "correlation": 94722
    }
  },
  {
    "ph": "f", "id": 94722, "pid": 76337, "tid": -914061504, "ts": 1716454222770548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222840333, "dur": 7,
    "args": {
      "External id": 94723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94723,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 94723, "pid": 5, "tid": 7, "ts": 1716454222840333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770550, "dur": 11,
    "args": {
      "External id": 94723,
      "cbid": 211, "correlation": 94723
    }
  },
  {
    "ph": "s", "id": 94723, "pid": 76337, "tid": -914061504, "ts": 1716454222770550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222840341, "dur": 3,
    "args": {
      "External id": 94725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94725,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 94725, "pid": 5, "tid": 7, "ts": 1716454222840341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770563, "dur": 5,
    "args": {
      "External id": 94725,
      "cbid": 211, "correlation": 94725
    }
  },
  {
    "ph": "s", "id": 94725, "pid": 76337, "tid": -914061504, "ts": 1716454222770563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222840346, "dur": 90,
    "args": {
      "External id": 94746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94746,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 94746, "pid": 5, "tid": 7, "ts": 1716454222840346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770637, "dur": 13,
    "args": {
      "External id": 94746,
      "cbid": 211, "correlation": 94746
    }
  },
  {
    "ph": "s", "id": 94746, "pid": 76337, "tid": -914061504, "ts": 1716454222770637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770735, "dur": 1,
    "args": {
      "External id": 94764,
      "cbid": 251, "correlation": 94764
    }
  },
  {
    "ph": "f", "id": 94764, "pid": 76337, "tid": -914061504, "ts": 1716454222770735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222840437, "dur": 96,
    "args": {
      "External id": 94766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94766,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94766, "pid": 5, "tid": 7, "ts": 1716454222840437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770741, "dur": 14,
    "args": {
      "External id": 94766,
      "cbid": 211, "correlation": 94766
    }
  },
  {
    "ph": "s", "id": 94766, "pid": 76337, "tid": -914061504, "ts": 1716454222770741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222840534, "dur": 19,
    "args": {
      "External id": 94774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94774,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94774, "pid": 5, "tid": 7, "ts": 1716454222840534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770812, "dur": 12,
    "args": {
      "External id": 94774,
      "cbid": 211, "correlation": 94774
    }
  },
  {
    "ph": "s", "id": 94774, "pid": 76337, "tid": -914061504, "ts": 1716454222770812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222840555, "dur": 38,
    "args": {
      "External id": 94782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94782, "pid": 5, "tid": 7, "ts": 1716454222840555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770853, "dur": 10,
    "args": {
      "External id": 94782,
      "cbid": 211, "correlation": 94782
    }
  },
  {
    "ph": "s", "id": 94782, "pid": 76337, "tid": -914061504, "ts": 1716454222770853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222840594, "dur": 34,
    "args": {
      "External id": 94804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94804,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94804, "pid": 5, "tid": 7, "ts": 1716454222840594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222770905, "dur": 10,
    "args": {
      "External id": 94804,
      "cbid": 211, "correlation": 94804
    }
  },
  {
    "ph": "s", "id": 94804, "pid": 76337, "tid": -914061504, "ts": 1716454222770905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771004, "dur": 1,
    "args": {
      "External id": 94820,
      "cbid": 251, "correlation": 94820
    }
  },
  {
    "ph": "f", "id": 94820, "pid": 76337, "tid": -914061504, "ts": 1716454222771004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771009, "dur": 0,
    "args": {
      "External id": 94822,
      "cbid": 251, "correlation": 94822
    }
  },
  {
    "ph": "f", "id": 94822, "pid": 76337, "tid": -914061504, "ts": 1716454222771009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222840630, "dur": 531,
    "args": {
      "External id": 94823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94823,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 94823, "pid": 5, "tid": 7, "ts": 1716454222840630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771014, "dur": 14,
    "args": {
      "External id": 94823,
      "cbid": 211, "correlation": 94823
    }
  },
  {
    "ph": "s", "id": 94823, "pid": 76337, "tid": -914061504, "ts": 1716454222771014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222841162, "dur": 123,
    "args": {
      "External id": 94831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94831,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94831, "pid": 5, "tid": 7, "ts": 1716454222841162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771085, "dur": 13,
    "args": {
      "External id": 94831,
      "cbid": 211, "correlation": 94831
    }
  },
  {
    "ph": "s", "id": 94831, "pid": 76337, "tid": -914061504, "ts": 1716454222771085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222841286, "dur": 129,
    "args": {
      "External id": 94839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94839, "pid": 5, "tid": 7, "ts": 1716454222841286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771119, "dur": 9,
    "args": {
      "External id": 94839,
      "cbid": 211, "correlation": 94839
    }
  },
  {
    "ph": "s", "id": 94839, "pid": 76337, "tid": -914061504, "ts": 1716454222771119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771199, "dur": 1,
    "args": {
      "External id": 94855,
      "cbid": 251, "correlation": 94855
    }
  },
  {
    "ph": "f", "id": 94855, "pid": 76337, "tid": -914061504, "ts": 1716454222771199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222841416, "dur": 301,
    "args": {
      "External id": 94857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94857,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94857, "pid": 5, "tid": 7, "ts": 1716454222841416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771205, "dur": 12,
    "args": {
      "External id": 94857,
      "cbid": 211, "correlation": 94857
    }
  },
  {
    "ph": "s", "id": 94857, "pid": 76337, "tid": -914061504, "ts": 1716454222771205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222841719, "dur": 27,
    "args": {
      "External id": 94865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94865,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94865, "pid": 5, "tid": 7, "ts": 1716454222841719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771247, "dur": 11,
    "args": {
      "External id": 94865,
      "cbid": 211, "correlation": 94865
    }
  },
  {
    "ph": "s", "id": 94865, "pid": 76337, "tid": -914061504, "ts": 1716454222771247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222841748, "dur": 80,
    "args": {
      "External id": 94876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94876, "pid": 5, "tid": 7, "ts": 1716454222841748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771318, "dur": 12,
    "args": {
      "External id": 94876,
      "cbid": 211, "correlation": 94876
    }
  },
  {
    "ph": "s", "id": 94876, "pid": 76337, "tid": -914061504, "ts": 1716454222771318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771385, "dur": 0,
    "args": {
      "External id": 94888,
      "cbid": 317, "correlation": 94888
    }
  },
  {
    "ph": "f", "id": 94888, "pid": 76337, "tid": -914061504, "ts": 1716454222771385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771386, "dur": 0,
    "args": {
      "External id": 94889,
      "cbid": 203, "correlation": 94889
    }
  },
  {
    "ph": "f", "id": 94889, "pid": 76337, "tid": -914061504, "ts": 1716454222771386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771387, "dur": 0,
    "args": {
      "External id": 94890,
      "cbid": 205, "correlation": 94890
    }
  },
  {
    "ph": "f", "id": 94890, "pid": 76337, "tid": -914061504, "ts": 1716454222771387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222841828, "dur": 22,
    "args": {
      "External id": 94894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94894,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94894, "pid": 5, "tid": 7, "ts": 1716454222841828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771403, "dur": 13,
    "args": {
      "External id": 94894,
      "cbid": 211, "correlation": 94894
    }
  },
  {
    "ph": "s", "id": 94894, "pid": 76337, "tid": -914061504, "ts": 1716454222771403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222841852, "dur": 118,
    "args": {
      "External id": 94896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94896,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94896, "pid": 5, "tid": 7, "ts": 1716454222841852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771424, "dur": 7,
    "args": {
      "External id": 94896,
      "cbid": 211, "correlation": 94896
    }
  },
  {
    "ph": "s", "id": 94896, "pid": 76337, "tid": -914061504, "ts": 1716454222771424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222841971, "dur": 24,
    "args": {
      "External id": 94898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94898,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94898, "pid": 5, "tid": 7, "ts": 1716454222841971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771435, "dur": 5,
    "args": {
      "External id": 94898,
      "cbid": 211, "correlation": 94898
    }
  },
  {
    "ph": "s", "id": 94898, "pid": 76337, "tid": -914061504, "ts": 1716454222771435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222841996, "dur": 33,
    "args": {
      "External id": 94904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94904,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94904, "pid": 5, "tid": 7, "ts": 1716454222841996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771462, "dur": 10,
    "args": {
      "External id": 94904,
      "cbid": 211, "correlation": 94904
    }
  },
  {
    "ph": "s", "id": 94904, "pid": 76337, "tid": -914061504, "ts": 1716454222771462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222842030, "dur": 27,
    "args": {
      "External id": 94912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94912,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94912, "pid": 5, "tid": 7, "ts": 1716454222842030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771494, "dur": 8,
    "args": {
      "External id": 94912,
      "cbid": 211, "correlation": 94912
    }
  },
  {
    "ph": "s", "id": 94912, "pid": 76337, "tid": -914061504, "ts": 1716454222771494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222842058, "dur": 53,
    "args": {
      "External id": 94921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94921,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94921, "pid": 5, "tid": 7, "ts": 1716454222842058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771535, "dur": 10,
    "args": {
      "External id": 94921,
      "cbid": 211, "correlation": 94921
    }
  },
  {
    "ph": "s", "id": 94921, "pid": 76337, "tid": -914061504, "ts": 1716454222771535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222842113, "dur": 51,
    "args": {
      "External id": 94941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94941,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 94941, "pid": 5, "tid": 7, "ts": 1716454222842113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771611, "dur": 11,
    "args": {
      "External id": 94941,
      "cbid": 211, "correlation": 94941
    }
  },
  {
    "ph": "s", "id": 94941, "pid": 76337, "tid": -914061504, "ts": 1716454222771611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222842165, "dur": 5,
    "args": {
      "External id": 94953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94953,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 94953, "pid": 5, "tid": 7, "ts": 1716454222842165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771632, "dur": 7,
    "args": {
      "External id": 94953,
      "cbid": 211, "correlation": 94953
    }
  },
  {
    "ph": "s", "id": 94953, "pid": 76337, "tid": -914061504, "ts": 1716454222771632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222842171, "dur": 55,
    "args": {
      "External id": 94956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94956, "pid": 5, "tid": 7, "ts": 1716454222842171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771652, "dur": 6,
    "args": {
      "External id": 94956,
      "cbid": 211, "correlation": 94956
    }
  },
  {
    "ph": "s", "id": 94956, "pid": 76337, "tid": -914061504, "ts": 1716454222771652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222842227, "dur": 37,
    "args": {
      "External id": 94965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94965,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94965, "pid": 5, "tid": 7, "ts": 1716454222842227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771691, "dur": 10,
    "args": {
      "External id": 94965,
      "cbid": 211, "correlation": 94965
    }
  },
  {
    "ph": "s", "id": 94965, "pid": 76337, "tid": -914061504, "ts": 1716454222771691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771743, "dur": 0,
    "args": {
      "External id": 94975,
      "cbid": 317, "correlation": 94975
    }
  },
  {
    "ph": "f", "id": 94975, "pid": 76337, "tid": -914061504, "ts": 1716454222771743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771744, "dur": 0,
    "args": {
      "External id": 94976,
      "cbid": 203, "correlation": 94976
    }
  },
  {
    "ph": "f", "id": 94976, "pid": 76337, "tid": -914061504, "ts": 1716454222771744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771745, "dur": 0,
    "args": {
      "External id": 94977,
      "cbid": 205, "correlation": 94977
    }
  },
  {
    "ph": "f", "id": 94977, "pid": 76337, "tid": -914061504, "ts": 1716454222771745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222842265, "dur": 39,
    "args": {
      "External id": 94981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94981, "pid": 5, "tid": 7, "ts": 1716454222842265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771760, "dur": 11,
    "args": {
      "External id": 94981,
      "cbid": 211, "correlation": 94981
    }
  },
  {
    "ph": "s", "id": 94981, "pid": 76337, "tid": -914061504, "ts": 1716454222771760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222842305, "dur": 81,
    "args": {
      "External id": 94983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94983, "pid": 5, "tid": 7, "ts": 1716454222842305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771774, "dur": 6,
    "args": {
      "External id": 94983,
      "cbid": 211, "correlation": 94983
    }
  },
  {
    "ph": "s", "id": 94983, "pid": 76337, "tid": -914061504, "ts": 1716454222771774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222842388, "dur": 1263,
    "args": {
      "External id": 94985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94985,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 94985, "pid": 5, "tid": 7, "ts": 1716454222842388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771786, "dur": 7,
    "args": {
      "External id": 94985,
      "cbid": 211, "correlation": 94985
    }
  },
  {
    "ph": "s", "id": 94985, "pid": 76337, "tid": -914061504, "ts": 1716454222771786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222843652, "dur": 20,
    "args": {
      "External id": 94987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94987, "pid": 5, "tid": 7, "ts": 1716454222843652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771796, "dur": 5,
    "args": {
      "External id": 94987,
      "cbid": 211, "correlation": 94987
    }
  },
  {
    "ph": "s", "id": 94987, "pid": 76337, "tid": -914061504, "ts": 1716454222771796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222843674, "dur": 33,
    "args": {
      "External id": 94993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 94993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 94993, "pid": 5, "tid": 7, "ts": 1716454222843674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771824, "dur": 8,
    "args": {
      "External id": 94993,
      "cbid": 211, "correlation": 94993
    }
  },
  {
    "ph": "s", "id": 94993, "pid": 76337, "tid": -914061504, "ts": 1716454222771824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222843708, "dur": 3,
    "args": {
      "External id": 95001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95001,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 95001, "pid": 5, "tid": 7, "ts": 1716454222843708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771868, "dur": 10,
    "args": {
      "External id": 95001,
      "cbid": 211, "correlation": 95001
    }
  },
  {
    "ph": "s", "id": 95001, "pid": 76337, "tid": -914061504, "ts": 1716454222771868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771934, "dur": 1,
    "args": {
      "External id": 95017,
      "cbid": 251, "correlation": 95017
    }
  },
  {
    "ph": "f", "id": 95017, "pid": 76337, "tid": -914061504, "ts": 1716454222771934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771940, "dur": 0,
    "args": {
      "External id": 95019,
      "cbid": 251, "correlation": 95019
    }
  },
  {
    "ph": "f", "id": 95019, "pid": 76337, "tid": -914061504, "ts": 1716454222771940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222843713, "dur": 12,
    "args": {
      "External id": 95020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95020,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 95020, "pid": 5, "tid": 7, "ts": 1716454222843713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771941, "dur": 12,
    "args": {
      "External id": 95020,
      "cbid": 211, "correlation": 95020
    }
  },
  {
    "ph": "s", "id": 95020, "pid": 76337, "tid": -914061504, "ts": 1716454222771941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222843726, "dur": 5,
    "args": {
      "External id": 95022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95022,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 95022, "pid": 5, "tid": 7, "ts": 1716454222843726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222771956, "dur": 6,
    "args": {
      "External id": 95022,
      "cbid": 211, "correlation": 95022
    }
  },
  {
    "ph": "s", "id": 95022, "pid": 76337, "tid": -914061504, "ts": 1716454222771956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222843732, "dur": 29,
    "args": {
      "External id": 95032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95032, "pid": 5, "tid": 7, "ts": 1716454222843732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772024, "dur": 13,
    "args": {
      "External id": 95032,
      "cbid": 211, "correlation": 95032
    }
  },
  {
    "ph": "s", "id": 95032, "pid": 76337, "tid": -914061504, "ts": 1716454222772024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222843762, "dur": 32,
    "args": {
      "External id": 95052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95052,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 95052, "pid": 5, "tid": 7, "ts": 1716454222843762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772091, "dur": 11,
    "args": {
      "External id": 95052,
      "cbid": 211, "correlation": 95052
    }
  },
  {
    "ph": "s", "id": 95052, "pid": 76337, "tid": -914061504, "ts": 1716454222772091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222843795, "dur": 4,
    "args": {
      "External id": 95064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95064,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 95064, "pid": 5, "tid": 7, "ts": 1716454222843795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772112, "dur": 8,
    "args": {
      "External id": 95064,
      "cbid": 211, "correlation": 95064
    }
  },
  {
    "ph": "s", "id": 95064, "pid": 76337, "tid": -914061504, "ts": 1716454222772112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222843800, "dur": 29,
    "args": {
      "External id": 95067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95067, "pid": 5, "tid": 7, "ts": 1716454222843800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772131, "dur": 6,
    "args": {
      "External id": 95067,
      "cbid": 211, "correlation": 95067
    }
  },
  {
    "ph": "s", "id": 95067, "pid": 76337, "tid": -914061504, "ts": 1716454222772131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222843831, "dur": 21,
    "args": {
      "External id": 95076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95076,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95076, "pid": 5, "tid": 7, "ts": 1716454222843831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772172, "dur": 10,
    "args": {
      "External id": 95076,
      "cbid": 211, "correlation": 95076
    }
  },
  {
    "ph": "s", "id": 95076, "pid": 76337, "tid": -914061504, "ts": 1716454222772172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772236, "dur": 0,
    "args": {
      "External id": 95086,
      "cbid": 317, "correlation": 95086
    }
  },
  {
    "ph": "f", "id": 95086, "pid": 76337, "tid": -914061504, "ts": 1716454222772236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772237, "dur": 0,
    "args": {
      "External id": 95087,
      "cbid": 203, "correlation": 95087
    }
  },
  {
    "ph": "f", "id": 95087, "pid": 76337, "tid": -914061504, "ts": 1716454222772237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772238, "dur": 0,
    "args": {
      "External id": 95088,
      "cbid": 205, "correlation": 95088
    }
  },
  {
    "ph": "f", "id": 95088, "pid": 76337, "tid": -914061504, "ts": 1716454222772238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222843853, "dur": 22,
    "args": {
      "External id": 95092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95092,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95092, "pid": 5, "tid": 7, "ts": 1716454222843853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772252, "dur": 12,
    "args": {
      "External id": 95092,
      "cbid": 211, "correlation": 95092
    }
  },
  {
    "ph": "s", "id": 95092, "pid": 76337, "tid": -914061504, "ts": 1716454222772252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222843876, "dur": 43,
    "args": {
      "External id": 95094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95094,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95094, "pid": 5, "tid": 7, "ts": 1716454222843876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772267, "dur": 5,
    "args": {
      "External id": 95094,
      "cbid": 211, "correlation": 95094
    }
  },
  {
    "ph": "s", "id": 95094, "pid": 76337, "tid": -914061504, "ts": 1716454222772267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222843920, "dur": 637,
    "args": {
      "External id": 95096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95096,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95096, "pid": 5, "tid": 7, "ts": 1716454222843920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772279, "dur": 6,
    "args": {
      "External id": 95096,
      "cbid": 211, "correlation": 95096
    }
  },
  {
    "ph": "s", "id": 95096, "pid": 76337, "tid": -914061504, "ts": 1716454222772279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222844558, "dur": 21,
    "args": {
      "External id": 95098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95098, "pid": 5, "tid": 7, "ts": 1716454222844558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772289, "dur": 5,
    "args": {
      "External id": 95098,
      "cbid": 211, "correlation": 95098
    }
  },
  {
    "ph": "s", "id": 95098, "pid": 76337, "tid": -914061504, "ts": 1716454222772289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222844581, "dur": 32,
    "args": {
      "External id": 95104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95104, "pid": 5, "tid": 7, "ts": 1716454222844581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772316, "dur": 10,
    "args": {
      "External id": 95104,
      "cbid": 211, "correlation": 95104
    }
  },
  {
    "ph": "s", "id": 95104, "pid": 76337, "tid": -914061504, "ts": 1716454222772316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772375, "dur": 0,
    "args": {
      "External id": 95114,
      "cbid": 317, "correlation": 95114
    }
  },
  {
    "ph": "f", "id": 95114, "pid": 76337, "tid": -914061504, "ts": 1716454222772375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772376, "dur": 0,
    "args": {
      "External id": 95115,
      "cbid": 203, "correlation": 95115
    }
  },
  {
    "ph": "f", "id": 95115, "pid": 76337, "tid": -914061504, "ts": 1716454222772376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772377, "dur": 0,
    "args": {
      "External id": 95116,
      "cbid": 205, "correlation": 95116
    }
  },
  {
    "ph": "f", "id": 95116, "pid": 76337, "tid": -914061504, "ts": 1716454222772377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222844614, "dur": 38,
    "args": {
      "External id": 95120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95120,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95120, "pid": 5, "tid": 7, "ts": 1716454222844614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772390, "dur": 12,
    "args": {
      "External id": 95120,
      "cbid": 211, "correlation": 95120
    }
  },
  {
    "ph": "s", "id": 95120, "pid": 76337, "tid": -914061504, "ts": 1716454222772390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222844654, "dur": 188,
    "args": {
      "External id": 95122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95122,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95122, "pid": 5, "tid": 7, "ts": 1716454222844654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772407, "dur": 6,
    "args": {
      "External id": 95122,
      "cbid": 211, "correlation": 95122
    }
  },
  {
    "ph": "s", "id": 95122, "pid": 76337, "tid": -914061504, "ts": 1716454222772407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222844843, "dur": 22,
    "args": {
      "External id": 95124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95124, "pid": 5, "tid": 7, "ts": 1716454222844843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772417, "dur": 5,
    "args": {
      "External id": 95124,
      "cbid": 211, "correlation": 95124
    }
  },
  {
    "ph": "s", "id": 95124, "pid": 76337, "tid": -914061504, "ts": 1716454222772417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222844866, "dur": 32,
    "args": {
      "External id": 95130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95130, "pid": 5, "tid": 7, "ts": 1716454222844866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772443, "dur": 8,
    "args": {
      "External id": 95130,
      "cbid": 211, "correlation": 95130
    }
  },
  {
    "ph": "s", "id": 95130, "pid": 76337, "tid": -914061504, "ts": 1716454222772443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222844900, "dur": 28,
    "args": {
      "External id": 95138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95138,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95138, "pid": 5, "tid": 7, "ts": 1716454222844900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772472, "dur": 8,
    "args": {
      "External id": 95138,
      "cbid": 211, "correlation": 95138
    }
  },
  {
    "ph": "s", "id": 95138, "pid": 76337, "tid": -914061504, "ts": 1716454222772472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222844929, "dur": 20,
    "args": {
      "External id": 95146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95146,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95146, "pid": 5, "tid": 7, "ts": 1716454222844929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772501, "dur": 8,
    "args": {
      "External id": 95146,
      "cbid": 211, "correlation": 95146
    }
  },
  {
    "ph": "s", "id": 95146, "pid": 76337, "tid": -914061504, "ts": 1716454222772501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222844950, "dur": 30,
    "args": {
      "External id": 95166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95166,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 95166, "pid": 5, "tid": 7, "ts": 1716454222844950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772584, "dur": 12,
    "args": {
      "External id": 95166,
      "cbid": 211, "correlation": 95166
    }
  },
  {
    "ph": "s", "id": 95166, "pid": 76337, "tid": -914061504, "ts": 1716454222772584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222844981, "dur": 4,
    "args": {
      "External id": 95178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95178,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 95178, "pid": 5, "tid": 7, "ts": 1716454222844981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772605, "dur": 7,
    "args": {
      "External id": 95178,
      "cbid": 211, "correlation": 95178
    }
  },
  {
    "ph": "s", "id": 95178, "pid": 76337, "tid": -914061504, "ts": 1716454222772605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222844987, "dur": 29,
    "args": {
      "External id": 95181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95181,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95181, "pid": 5, "tid": 7, "ts": 1716454222844987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772624, "dur": 7,
    "args": {
      "External id": 95181,
      "cbid": 211, "correlation": 95181
    }
  },
  {
    "ph": "s", "id": 95181, "pid": 76337, "tid": -914061504, "ts": 1716454222772624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772681, "dur": 0,
    "args": {
      "External id": 95192,
      "cbid": 317, "correlation": 95192
    }
  },
  {
    "ph": "f", "id": 95192, "pid": 76337, "tid": -914061504, "ts": 1716454222772681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772682, "dur": 0,
    "args": {
      "External id": 95193,
      "cbid": 203, "correlation": 95193
    }
  },
  {
    "ph": "f", "id": 95193, "pid": 76337, "tid": -914061504, "ts": 1716454222772682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772682, "dur": 0,
    "args": {
      "External id": 95194,
      "cbid": 205, "correlation": 95194
    }
  },
  {
    "ph": "f", "id": 95194, "pid": 76337, "tid": -914061504, "ts": 1716454222772682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222845018, "dur": 23,
    "args": {
      "External id": 95198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95198,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95198, "pid": 5, "tid": 7, "ts": 1716454222845018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772696, "dur": 11,
    "args": {
      "External id": 95198,
      "cbid": 211, "correlation": 95198
    }
  },
  {
    "ph": "s", "id": 95198, "pid": 76337, "tid": -914061504, "ts": 1716454222772696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222845042, "dur": 103,
    "args": {
      "External id": 95200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95200,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95200, "pid": 5, "tid": 7, "ts": 1716454222845042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772714, "dur": 6,
    "args": {
      "External id": 95200,
      "cbid": 211, "correlation": 95200
    }
  },
  {
    "ph": "s", "id": 95200, "pid": 76337, "tid": -914061504, "ts": 1716454222772714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222845146, "dur": 21,
    "args": {
      "External id": 95202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95202, "pid": 5, "tid": 7, "ts": 1716454222845146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772724, "dur": 6,
    "args": {
      "External id": 95202,
      "cbid": 211, "correlation": 95202
    }
  },
  {
    "ph": "s", "id": 95202, "pid": 76337, "tid": -914061504, "ts": 1716454222772724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222845169, "dur": 32,
    "args": {
      "External id": 95208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95208,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95208, "pid": 5, "tid": 7, "ts": 1716454222845169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772751, "dur": 9,
    "args": {
      "External id": 95208,
      "cbid": 211, "correlation": 95208
    }
  },
  {
    "ph": "s", "id": 95208, "pid": 76337, "tid": -914061504, "ts": 1716454222772751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222845202, "dur": 198,
    "args": {
      "External id": 95217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95217,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95217, "pid": 5, "tid": 7, "ts": 1716454222845202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772834, "dur": 15,
    "args": {
      "External id": 95217,
      "cbid": 211, "correlation": 95217
    }
  },
  {
    "ph": "s", "id": 95217, "pid": 76337, "tid": -914061504, "ts": 1716454222772834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222845401, "dur": 64,
    "args": {
      "External id": 95239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95239,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95239, "pid": 5, "tid": 7, "ts": 1716454222845401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772892, "dur": 11,
    "args": {
      "External id": 95239,
      "cbid": 211, "correlation": 95239
    }
  },
  {
    "ph": "s", "id": 95239, "pid": 76337, "tid": -914061504, "ts": 1716454222772892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772993, "dur": 1,
    "args": {
      "External id": 95250,
      "cbid": 251, "correlation": 95250
    }
  },
  {
    "ph": "f", "id": 95250, "pid": 76337, "tid": -914061504, "ts": 1716454222772993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222845466, "dur": 150,
    "args": {
      "External id": 95251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95251,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95251, "pid": 5, "tid": 7, "ts": 1716454222845466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222772999, "dur": 13,
    "args": {
      "External id": 95251,
      "cbid": 211, "correlation": 95251
    }
  },
  {
    "ph": "s", "id": 95251, "pid": 76337, "tid": -914061504, "ts": 1716454222772999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773071, "dur": 1,
    "args": {
      "External id": 95262,
      "cbid": 251, "correlation": 95262
    }
  },
  {
    "ph": "f", "id": 95262, "pid": 76337, "tid": -914061504, "ts": 1716454222773071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222845618, "dur": 143,
    "args": {
      "External id": 95263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95263,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95263, "pid": 5, "tid": 7, "ts": 1716454222845618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773075, "dur": 12,
    "args": {
      "External id": 95263,
      "cbid": 211, "correlation": 95263
    }
  },
  {
    "ph": "s", "id": 95263, "pid": 76337, "tid": -914061504, "ts": 1716454222773075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773140, "dur": 1,
    "args": {
      "External id": 95274,
      "cbid": 251, "correlation": 95274
    }
  },
  {
    "ph": "f", "id": 95274, "pid": 76337, "tid": -914061504, "ts": 1716454222773140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222845763, "dur": 144,
    "args": {
      "External id": 95275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95275,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95275, "pid": 5, "tid": 7, "ts": 1716454222845763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773144, "dur": 11,
    "args": {
      "External id": 95275,
      "cbid": 211, "correlation": 95275
    }
  },
  {
    "ph": "s", "id": 95275, "pid": 76337, "tid": -914061504, "ts": 1716454222773144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222845908, "dur": 1903,
    "args": {
      "External id": 95296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95296,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 95296, "pid": 5, "tid": 7, "ts": 1716454222845908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773226, "dur": 13,
    "args": {
      "External id": 95296,
      "cbid": 211, "correlation": 95296
    }
  },
  {
    "ph": "s", "id": 95296, "pid": 76337, "tid": -914061504, "ts": 1716454222773226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773329, "dur": 1,
    "args": {
      "External id": 95314,
      "cbid": 251, "correlation": 95314
    }
  },
  {
    "ph": "f", "id": 95314, "pid": 76337, "tid": -914061504, "ts": 1716454222773329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222847813, "dur": 147,
    "args": {
      "External id": 95316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95316,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 95316, "pid": 5, "tid": 7, "ts": 1716454222847813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773335, "dur": 13,
    "args": {
      "External id": 95316,
      "cbid": 211, "correlation": 95316
    }
  },
  {
    "ph": "s", "id": 95316, "pid": 76337, "tid": -914061504, "ts": 1716454222773335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222847961, "dur": 35,
    "args": {
      "External id": 95324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95324,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95324, "pid": 5, "tid": 7, "ts": 1716454222847961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773406, "dur": 14,
    "args": {
      "External id": 95324,
      "cbid": 211, "correlation": 95324
    }
  },
  {
    "ph": "s", "id": 95324, "pid": 76337, "tid": -914061504, "ts": 1716454222773406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222847997, "dur": 51,
    "args": {
      "External id": 95332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95332,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95332, "pid": 5, "tid": 7, "ts": 1716454222847997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773447, "dur": 8,
    "args": {
      "External id": 95332,
      "cbid": 211, "correlation": 95332
    }
  },
  {
    "ph": "s", "id": 95332, "pid": 76337, "tid": -914061504, "ts": 1716454222773447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222848049, "dur": 29,
    "args": {
      "External id": 95343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95343,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95343, "pid": 5, "tid": 7, "ts": 1716454222848049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773521, "dur": 13,
    "args": {
      "External id": 95343,
      "cbid": 211, "correlation": 95343
    }
  },
  {
    "ph": "s", "id": 95343, "pid": 76337, "tid": -914061504, "ts": 1716454222773521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222848079, "dur": 34,
    "args": {
      "External id": 95365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95365,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95365, "pid": 5, "tid": 7, "ts": 1716454222848079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773553, "dur": 7,
    "args": {
      "External id": 95365,
      "cbid": 211, "correlation": 95365
    }
  },
  {
    "ph": "s", "id": 95365, "pid": 76337, "tid": -914061504, "ts": 1716454222773553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773640, "dur": 1,
    "args": {
      "External id": 95376,
      "cbid": 251, "correlation": 95376
    }
  },
  {
    "ph": "f", "id": 95376, "pid": 76337, "tid": -914061504, "ts": 1716454222773640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222848115, "dur": 87,
    "args": {
      "External id": 95377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95377,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95377, "pid": 5, "tid": 7, "ts": 1716454222848115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773645, "dur": 13,
    "args": {
      "External id": 95377,
      "cbid": 211, "correlation": 95377
    }
  },
  {
    "ph": "s", "id": 95377, "pid": 76337, "tid": -914061504, "ts": 1716454222773645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773717, "dur": 1,
    "args": {
      "External id": 95388,
      "cbid": 251, "correlation": 95388
    }
  },
  {
    "ph": "f", "id": 95388, "pid": 76337, "tid": -914061504, "ts": 1716454222773717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773720, "dur": 0,
    "args": {
      "External id": 95389,
      "cbid": 251, "correlation": 95389
    }
  },
  {
    "ph": "f", "id": 95389, "pid": 76337, "tid": -914061504, "ts": 1716454222773720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222848204, "dur": 11,
    "args": {
      "External id": 95390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95390,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 95390, "pid": 5, "tid": 7, "ts": 1716454222848204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773722, "dur": 12,
    "args": {
      "External id": 95390,
      "cbid": 211, "correlation": 95390
    }
  },
  {
    "ph": "s", "id": 95390, "pid": 76337, "tid": -914061504, "ts": 1716454222773722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222848216, "dur": 5,
    "args": {
      "External id": 95392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95392,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 95392, "pid": 5, "tid": 7, "ts": 1716454222848216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773737, "dur": 6,
    "args": {
      "External id": 95392,
      "cbid": 211, "correlation": 95392
    }
  },
  {
    "ph": "s", "id": 95392, "pid": 76337, "tid": -914061504, "ts": 1716454222773737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773794, "dur": 1,
    "args": {
      "External id": 95403,
      "cbid": 251, "correlation": 95403
    }
  },
  {
    "ph": "f", "id": 95403, "pid": 76337, "tid": -914061504, "ts": 1716454222773794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773797, "dur": 0,
    "args": {
      "External id": 95404,
      "cbid": 251, "correlation": 95404
    }
  },
  {
    "ph": "f", "id": 95404, "pid": 76337, "tid": -914061504, "ts": 1716454222773797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222848222, "dur": 7,
    "args": {
      "External id": 95405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95405,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 95405, "pid": 5, "tid": 7, "ts": 1716454222848222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773799, "dur": 12,
    "args": {
      "External id": 95405,
      "cbid": 211, "correlation": 95405
    }
  },
  {
    "ph": "s", "id": 95405, "pid": 76337, "tid": -914061504, "ts": 1716454222773799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222848231, "dur": 3,
    "args": {
      "External id": 95407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95407,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 95407, "pid": 5, "tid": 7, "ts": 1716454222848231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773813, "dur": 6,
    "args": {
      "External id": 95407,
      "cbid": 211, "correlation": 95407
    }
  },
  {
    "ph": "s", "id": 95407, "pid": 76337, "tid": -914061504, "ts": 1716454222773813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222848235, "dur": 89,
    "args": {
      "External id": 95428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95428,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 95428, "pid": 5, "tid": 7, "ts": 1716454222848235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773887, "dur": 12,
    "args": {
      "External id": 95428,
      "cbid": 211, "correlation": 95428
    }
  },
  {
    "ph": "s", "id": 95428, "pid": 76337, "tid": -914061504, "ts": 1716454222773887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222773994, "dur": 1,
    "args": {
      "External id": 95446,
      "cbid": 251, "correlation": 95446
    }
  },
  {
    "ph": "f", "id": 95446, "pid": 76337, "tid": -914061504, "ts": 1716454222773994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222848326, "dur": 98,
    "args": {
      "External id": 95448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95448,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95448, "pid": 5, "tid": 7, "ts": 1716454222848326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774000, "dur": 14,
    "args": {
      "External id": 95448,
      "cbid": 211, "correlation": 95448
    }
  },
  {
    "ph": "s", "id": 95448, "pid": 76337, "tid": -914061504, "ts": 1716454222774000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222848425, "dur": 19,
    "args": {
      "External id": 95456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95456,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95456, "pid": 5, "tid": 7, "ts": 1716454222848425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774070, "dur": 12,
    "args": {
      "External id": 95456,
      "cbid": 211, "correlation": 95456
    }
  },
  {
    "ph": "s", "id": 95456, "pid": 76337, "tid": -914061504, "ts": 1716454222774070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222848446, "dur": 38,
    "args": {
      "External id": 95464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95464, "pid": 5, "tid": 7, "ts": 1716454222848446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774112, "dur": 9,
    "args": {
      "External id": 95464,
      "cbid": 211, "correlation": 95464
    }
  },
  {
    "ph": "s", "id": 95464, "pid": 76337, "tid": -914061504, "ts": 1716454222774112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222848484, "dur": 34,
    "args": {
      "External id": 95486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95486,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95486, "pid": 5, "tid": 7, "ts": 1716454222848484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774163, "dur": 10,
    "args": {
      "External id": 95486,
      "cbid": 211, "correlation": 95486
    }
  },
  {
    "ph": "s", "id": 95486, "pid": 76337, "tid": -914061504, "ts": 1716454222774163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774253, "dur": 1,
    "args": {
      "External id": 95502,
      "cbid": 251, "correlation": 95502
    }
  },
  {
    "ph": "f", "id": 95502, "pid": 76337, "tid": -914061504, "ts": 1716454222774253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774258, "dur": 0,
    "args": {
      "External id": 95504,
      "cbid": 251, "correlation": 95504
    }
  },
  {
    "ph": "f", "id": 95504, "pid": 76337, "tid": -914061504, "ts": 1716454222774258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222848520, "dur": 531,
    "args": {
      "External id": 95505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95505,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 95505, "pid": 5, "tid": 7, "ts": 1716454222848520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774262, "dur": 13,
    "args": {
      "External id": 95505,
      "cbid": 211, "correlation": 95505
    }
  },
  {
    "ph": "s", "id": 95505, "pid": 76337, "tid": -914061504, "ts": 1716454222774262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222849053, "dur": 123,
    "args": {
      "External id": 95513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95513,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95513, "pid": 5, "tid": 7, "ts": 1716454222849053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774331, "dur": 13,
    "args": {
      "External id": 95513,
      "cbid": 211, "correlation": 95513
    }
  },
  {
    "ph": "s", "id": 95513, "pid": 76337, "tid": -914061504, "ts": 1716454222774331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222849177, "dur": 130,
    "args": {
      "External id": 95521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95521, "pid": 5, "tid": 7, "ts": 1716454222849177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774363, "dur": 8,
    "args": {
      "External id": 95521,
      "cbid": 211, "correlation": 95521
    }
  },
  {
    "ph": "s", "id": 95521, "pid": 76337, "tid": -914061504, "ts": 1716454222774363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774440, "dur": 2,
    "args": {
      "External id": 95537,
      "cbid": 251, "correlation": 95537
    }
  },
  {
    "ph": "f", "id": 95537, "pid": 76337, "tid": -914061504, "ts": 1716454222774440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222849309, "dur": 305,
    "args": {
      "External id": 95539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95539,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95539, "pid": 5, "tid": 7, "ts": 1716454222849309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774446, "dur": 12,
    "args": {
      "External id": 95539,
      "cbid": 211, "correlation": 95539
    }
  },
  {
    "ph": "s", "id": 95539, "pid": 76337, "tid": -914061504, "ts": 1716454222774446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222849615, "dur": 27,
    "args": {
      "External id": 95547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95547,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95547, "pid": 5, "tid": 7, "ts": 1716454222849615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774487, "dur": 10,
    "args": {
      "External id": 95547,
      "cbid": 211, "correlation": 95547
    }
  },
  {
    "ph": "s", "id": 95547, "pid": 76337, "tid": -914061504, "ts": 1716454222774487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222849643, "dur": 80,
    "args": {
      "External id": 95558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95558, "pid": 5, "tid": 7, "ts": 1716454222849643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774557, "dur": 12,
    "args": {
      "External id": 95558,
      "cbid": 211, "correlation": 95558
    }
  },
  {
    "ph": "s", "id": 95558, "pid": 76337, "tid": -914061504, "ts": 1716454222774557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774625, "dur": 0,
    "args": {
      "External id": 95570,
      "cbid": 317, "correlation": 95570
    }
  },
  {
    "ph": "f", "id": 95570, "pid": 76337, "tid": -914061504, "ts": 1716454222774625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774626, "dur": 0,
    "args": {
      "External id": 95571,
      "cbid": 203, "correlation": 95571
    }
  },
  {
    "ph": "f", "id": 95571, "pid": 76337, "tid": -914061504, "ts": 1716454222774626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774627, "dur": 0,
    "args": {
      "External id": 95572,
      "cbid": 205, "correlation": 95572
    }
  },
  {
    "ph": "f", "id": 95572, "pid": 76337, "tid": -914061504, "ts": 1716454222774627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222849725, "dur": 22,
    "args": {
      "External id": 95576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95576,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95576, "pid": 5, "tid": 7, "ts": 1716454222849725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774644, "dur": 12,
    "args": {
      "External id": 95576,
      "cbid": 211, "correlation": 95576
    }
  },
  {
    "ph": "s", "id": 95576, "pid": 76337, "tid": -914061504, "ts": 1716454222774644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222849748, "dur": 117,
    "args": {
      "External id": 95578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95578,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95578, "pid": 5, "tid": 7, "ts": 1716454222849748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774663, "dur": 8,
    "args": {
      "External id": 95578,
      "cbid": 211, "correlation": 95578
    }
  },
  {
    "ph": "s", "id": 95578, "pid": 76337, "tid": -914061504, "ts": 1716454222774663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222849867, "dur": 23,
    "args": {
      "External id": 95580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95580,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95580, "pid": 5, "tid": 7, "ts": 1716454222849867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774675, "dur": 5,
    "args": {
      "External id": 95580,
      "cbid": 211, "correlation": 95580
    }
  },
  {
    "ph": "s", "id": 95580, "pid": 76337, "tid": -914061504, "ts": 1716454222774675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222849891, "dur": 33,
    "args": {
      "External id": 95586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95586,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95586, "pid": 5, "tid": 7, "ts": 1716454222849891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774702, "dur": 8,
    "args": {
      "External id": 95586,
      "cbid": 211, "correlation": 95586
    }
  },
  {
    "ph": "s", "id": 95586, "pid": 76337, "tid": -914061504, "ts": 1716454222774702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222849925, "dur": 26,
    "args": {
      "External id": 95594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95594,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95594, "pid": 5, "tid": 7, "ts": 1716454222849925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774734, "dur": 8,
    "args": {
      "External id": 95594,
      "cbid": 211, "correlation": 95594
    }
  },
  {
    "ph": "s", "id": 95594, "pid": 76337, "tid": -914061504, "ts": 1716454222774734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222849953, "dur": 45,
    "args": {
      "External id": 95603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95603,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95603, "pid": 5, "tid": 7, "ts": 1716454222849953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774772, "dur": 12,
    "args": {
      "External id": 95603,
      "cbid": 211, "correlation": 95603
    }
  },
  {
    "ph": "s", "id": 95603, "pid": 76337, "tid": -914061504, "ts": 1716454222774772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222849999, "dur": 42,
    "args": {
      "External id": 95623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95623,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 95623, "pid": 5, "tid": 7, "ts": 1716454222849999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774849, "dur": 11,
    "args": {
      "External id": 95623,
      "cbid": 211, "correlation": 95623
    }
  },
  {
    "ph": "s", "id": 95623, "pid": 76337, "tid": -914061504, "ts": 1716454222774849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222850043, "dur": 5,
    "args": {
      "External id": 95635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95635,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 95635, "pid": 5, "tid": 7, "ts": 1716454222850043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774870, "dur": 6,
    "args": {
      "External id": 95635,
      "cbid": 211, "correlation": 95635
    }
  },
  {
    "ph": "s", "id": 95635, "pid": 76337, "tid": -914061504, "ts": 1716454222774870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222850049, "dur": 43,
    "args": {
      "External id": 95638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95638, "pid": 5, "tid": 7, "ts": 1716454222850049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774889, "dur": 8,
    "args": {
      "External id": 95638,
      "cbid": 211, "correlation": 95638
    }
  },
  {
    "ph": "s", "id": 95638, "pid": 76337, "tid": -914061504, "ts": 1716454222774889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222850094, "dur": 29,
    "args": {
      "External id": 95647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95647,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95647, "pid": 5, "tid": 7, "ts": 1716454222850094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774930, "dur": 10,
    "args": {
      "External id": 95647,
      "cbid": 211, "correlation": 95647
    }
  },
  {
    "ph": "s", "id": 95647, "pid": 76337, "tid": -914061504, "ts": 1716454222774930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774990, "dur": 0,
    "args": {
      "External id": 95657,
      "cbid": 317, "correlation": 95657
    }
  },
  {
    "ph": "f", "id": 95657, "pid": 76337, "tid": -914061504, "ts": 1716454222774990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774991, "dur": 0,
    "args": {
      "External id": 95658,
      "cbid": 203, "correlation": 95658
    }
  },
  {
    "ph": "f", "id": 95658, "pid": 76337, "tid": -914061504, "ts": 1716454222774991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222774992, "dur": 0,
    "args": {
      "External id": 95659,
      "cbid": 205, "correlation": 95659
    }
  },
  {
    "ph": "f", "id": 95659, "pid": 76337, "tid": -914061504, "ts": 1716454222774992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222850124, "dur": 31,
    "args": {
      "External id": 95663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95663,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95663, "pid": 5, "tid": 7, "ts": 1716454222850124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775008, "dur": 12,
    "args": {
      "External id": 95663,
      "cbid": 211, "correlation": 95663
    }
  },
  {
    "ph": "s", "id": 95663, "pid": 76337, "tid": -914061504, "ts": 1716454222775008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222850157, "dur": 63,
    "args": {
      "External id": 95665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95665, "pid": 5, "tid": 7, "ts": 1716454222850157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775022, "dur": 5,
    "args": {
      "External id": 95665,
      "cbid": 211, "correlation": 95665
    }
  },
  {
    "ph": "s", "id": 95665, "pid": 76337, "tid": -914061504, "ts": 1716454222775022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222850221, "dur": 955,
    "args": {
      "External id": 95667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95667,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95667, "pid": 5, "tid": 7, "ts": 1716454222850221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775033, "dur": 6,
    "args": {
      "External id": 95667,
      "cbid": 211, "correlation": 95667
    }
  },
  {
    "ph": "s", "id": 95667, "pid": 76337, "tid": -914061504, "ts": 1716454222775033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222851177, "dur": 21,
    "args": {
      "External id": 95669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95669,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95669, "pid": 5, "tid": 7, "ts": 1716454222851177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775043, "dur": 5,
    "args": {
      "External id": 95669,
      "cbid": 211, "correlation": 95669
    }
  },
  {
    "ph": "s", "id": 95669, "pid": 76337, "tid": -914061504, "ts": 1716454222775043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222851200, "dur": 32,
    "args": {
      "External id": 95675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95675, "pid": 5, "tid": 7, "ts": 1716454222851200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775072, "dur": 8,
    "args": {
      "External id": 95675,
      "cbid": 211, "correlation": 95675
    }
  },
  {
    "ph": "s", "id": 95675, "pid": 76337, "tid": -914061504, "ts": 1716454222775072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222851234, "dur": 3,
    "args": {
      "External id": 95683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95683,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 95683, "pid": 5, "tid": 7, "ts": 1716454222851234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775116, "dur": 9,
    "args": {
      "External id": 95683,
      "cbid": 211, "correlation": 95683
    }
  },
  {
    "ph": "s", "id": 95683, "pid": 76337, "tid": -914061504, "ts": 1716454222775116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775184, "dur": 1,
    "args": {
      "External id": 95699,
      "cbid": 251, "correlation": 95699
    }
  },
  {
    "ph": "f", "id": 95699, "pid": 76337, "tid": -914061504, "ts": 1716454222775184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775190, "dur": 0,
    "args": {
      "External id": 95701,
      "cbid": 251, "correlation": 95701
    }
  },
  {
    "ph": "f", "id": 95701, "pid": 76337, "tid": -914061504, "ts": 1716454222775190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222851238, "dur": 12,
    "args": {
      "External id": 95702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95702,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 95702, "pid": 5, "tid": 7, "ts": 1716454222851238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775192, "dur": 12,
    "args": {
      "External id": 95702,
      "cbid": 211, "correlation": 95702
    }
  },
  {
    "ph": "s", "id": 95702, "pid": 76337, "tid": -914061504, "ts": 1716454222775192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222851252, "dur": 5,
    "args": {
      "External id": 95704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95704,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 95704, "pid": 5, "tid": 7, "ts": 1716454222851252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775205, "dur": 6,
    "args": {
      "External id": 95704,
      "cbid": 211, "correlation": 95704
    }
  },
  {
    "ph": "s", "id": 95704, "pid": 76337, "tid": -914061504, "ts": 1716454222775205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222851258, "dur": 30,
    "args": {
      "External id": 95714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95714,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95714, "pid": 5, "tid": 7, "ts": 1716454222851258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775264, "dur": 13,
    "args": {
      "External id": 95714,
      "cbid": 211, "correlation": 95714
    }
  },
  {
    "ph": "s", "id": 95714, "pid": 76337, "tid": -914061504, "ts": 1716454222775264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222851289, "dur": 31,
    "args": {
      "External id": 95734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95734,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 95734, "pid": 5, "tid": 7, "ts": 1716454222851289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775331, "dur": 11,
    "args": {
      "External id": 95734,
      "cbid": 211, "correlation": 95734
    }
  },
  {
    "ph": "s", "id": 95734, "pid": 76337, "tid": -914061504, "ts": 1716454222775331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222851321, "dur": 4,
    "args": {
      "External id": 95746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95746,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 95746, "pid": 5, "tid": 7, "ts": 1716454222851321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775351, "dur": 6,
    "args": {
      "External id": 95746,
      "cbid": 211, "correlation": 95746
    }
  },
  {
    "ph": "s", "id": 95746, "pid": 76337, "tid": -914061504, "ts": 1716454222775351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222851326, "dur": 29,
    "args": {
      "External id": 95749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95749,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95749, "pid": 5, "tid": 7, "ts": 1716454222851326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775370, "dur": 6,
    "args": {
      "External id": 95749,
      "cbid": 211, "correlation": 95749
    }
  },
  {
    "ph": "s", "id": 95749, "pid": 76337, "tid": -914061504, "ts": 1716454222775370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222851357, "dur": 20,
    "args": {
      "External id": 95758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95758,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95758, "pid": 5, "tid": 7, "ts": 1716454222851357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775411, "dur": 10,
    "args": {
      "External id": 95758,
      "cbid": 211, "correlation": 95758
    }
  },
  {
    "ph": "s", "id": 95758, "pid": 76337, "tid": -914061504, "ts": 1716454222775411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775475, "dur": 0,
    "args": {
      "External id": 95768,
      "cbid": 317, "correlation": 95768
    }
  },
  {
    "ph": "f", "id": 95768, "pid": 76337, "tid": -914061504, "ts": 1716454222775475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775476, "dur": 0,
    "args": {
      "External id": 95769,
      "cbid": 203, "correlation": 95769
    }
  },
  {
    "ph": "f", "id": 95769, "pid": 76337, "tid": -914061504, "ts": 1716454222775476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775476, "dur": 0,
    "args": {
      "External id": 95770,
      "cbid": 205, "correlation": 95770
    }
  },
  {
    "ph": "f", "id": 95770, "pid": 76337, "tid": -914061504, "ts": 1716454222775476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222851378, "dur": 24,
    "args": {
      "External id": 95774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95774,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95774, "pid": 5, "tid": 7, "ts": 1716454222851378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775491, "dur": 13,
    "args": {
      "External id": 95774,
      "cbid": 211, "correlation": 95774
    }
  },
  {
    "ph": "s", "id": 95774, "pid": 76337, "tid": -914061504, "ts": 1716454222775491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222851403, "dur": 43,
    "args": {
      "External id": 95776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95776, "pid": 5, "tid": 7, "ts": 1716454222851403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775506, "dur": 5,
    "args": {
      "External id": 95776,
      "cbid": 211, "correlation": 95776
    }
  },
  {
    "ph": "s", "id": 95776, "pid": 76337, "tid": -914061504, "ts": 1716454222775506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222851447, "dur": 637,
    "args": {
      "External id": 95778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95778,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95778, "pid": 5, "tid": 7, "ts": 1716454222851447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775518, "dur": 6,
    "args": {
      "External id": 95778,
      "cbid": 211, "correlation": 95778
    }
  },
  {
    "ph": "s", "id": 95778, "pid": 76337, "tid": -914061504, "ts": 1716454222775518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852085, "dur": 21,
    "args": {
      "External id": 95780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95780,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95780, "pid": 5, "tid": 7, "ts": 1716454222852085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775527, "dur": 5,
    "args": {
      "External id": 95780,
      "cbid": 211, "correlation": 95780
    }
  },
  {
    "ph": "s", "id": 95780, "pid": 76337, "tid": -914061504, "ts": 1716454222775527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222852108, "dur": 33,
    "args": {
      "External id": 95786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95786, "pid": 5, "tid": 7, "ts": 1716454222852108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775555, "dur": 9,
    "args": {
      "External id": 95786,
      "cbid": 211, "correlation": 95786
    }
  },
  {
    "ph": "s", "id": 95786, "pid": 76337, "tid": -914061504, "ts": 1716454222775555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775612, "dur": 0,
    "args": {
      "External id": 95796,
      "cbid": 317, "correlation": 95796
    }
  },
  {
    "ph": "f", "id": 95796, "pid": 76337, "tid": -914061504, "ts": 1716454222775612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775612, "dur": 0,
    "args": {
      "External id": 95797,
      "cbid": 203, "correlation": 95797
    }
  },
  {
    "ph": "f", "id": 95797, "pid": 76337, "tid": -914061504, "ts": 1716454222775612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775613, "dur": 0,
    "args": {
      "External id": 95798,
      "cbid": 205, "correlation": 95798
    }
  },
  {
    "ph": "f", "id": 95798, "pid": 76337, "tid": -914061504, "ts": 1716454222775613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852142, "dur": 29,
    "args": {
      "External id": 95802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95802, "pid": 5, "tid": 7, "ts": 1716454222852142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775628, "dur": 12,
    "args": {
      "External id": 95802,
      "cbid": 211, "correlation": 95802
    }
  },
  {
    "ph": "s", "id": 95802, "pid": 76337, "tid": -914061504, "ts": 1716454222775628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222852173, "dur": 150,
    "args": {
      "External id": 95804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95804,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95804, "pid": 5, "tid": 7, "ts": 1716454222852173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775646, "dur": 6,
    "args": {
      "External id": 95804,
      "cbid": 211, "correlation": 95804
    }
  },
  {
    "ph": "s", "id": 95804, "pid": 76337, "tid": -914061504, "ts": 1716454222775646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852324, "dur": 22,
    "args": {
      "External id": 95806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95806,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95806, "pid": 5, "tid": 7, "ts": 1716454222852324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775655, "dur": 5,
    "args": {
      "External id": 95806,
      "cbid": 211, "correlation": 95806
    }
  },
  {
    "ph": "s", "id": 95806, "pid": 76337, "tid": -914061504, "ts": 1716454222775655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222852348, "dur": 32,
    "args": {
      "External id": 95812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95812, "pid": 5, "tid": 7, "ts": 1716454222852348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775681, "dur": 8,
    "args": {
      "External id": 95812,
      "cbid": 211, "correlation": 95812
    }
  },
  {
    "ph": "s", "id": 95812, "pid": 76337, "tid": -914061504, "ts": 1716454222775681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222852382, "dur": 27,
    "args": {
      "External id": 95820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95820,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95820, "pid": 5, "tid": 7, "ts": 1716454222852382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775709, "dur": 7,
    "args": {
      "External id": 95820,
      "cbid": 211, "correlation": 95820
    }
  },
  {
    "ph": "s", "id": 95820, "pid": 76337, "tid": -914061504, "ts": 1716454222775709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222852410, "dur": 19,
    "args": {
      "External id": 95828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95828,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95828, "pid": 5, "tid": 7, "ts": 1716454222852410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775738, "dur": 8,
    "args": {
      "External id": 95828,
      "cbid": 211, "correlation": 95828
    }
  },
  {
    "ph": "s", "id": 95828, "pid": 76337, "tid": -914061504, "ts": 1716454222775738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852430, "dur": 30,
    "args": {
      "External id": 95848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95848,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 95848, "pid": 5, "tid": 7, "ts": 1716454222852430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775820, "dur": 13,
    "args": {
      "External id": 95848,
      "cbid": 211, "correlation": 95848
    }
  },
  {
    "ph": "s", "id": 95848, "pid": 76337, "tid": -914061504, "ts": 1716454222775820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222852462, "dur": 4,
    "args": {
      "External id": 95860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95860,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 95860, "pid": 5, "tid": 7, "ts": 1716454222852462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775843, "dur": 6,
    "args": {
      "External id": 95860,
      "cbid": 211, "correlation": 95860
    }
  },
  {
    "ph": "s", "id": 95860, "pid": 76337, "tid": -914061504, "ts": 1716454222775843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222852467, "dur": 30,
    "args": {
      "External id": 95863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95863, "pid": 5, "tid": 7, "ts": 1716454222852467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775861, "dur": 7,
    "args": {
      "External id": 95863,
      "cbid": 211, "correlation": 95863
    }
  },
  {
    "ph": "s", "id": 95863, "pid": 76337, "tid": -914061504, "ts": 1716454222775861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775917, "dur": 0,
    "args": {
      "External id": 95874,
      "cbid": 317, "correlation": 95874
    }
  },
  {
    "ph": "f", "id": 95874, "pid": 76337, "tid": -914061504, "ts": 1716454222775917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775918, "dur": 0,
    "args": {
      "External id": 95875,
      "cbid": 203, "correlation": 95875
    }
  },
  {
    "ph": "f", "id": 95875, "pid": 76337, "tid": -914061504, "ts": 1716454222775918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775919, "dur": 0,
    "args": {
      "External id": 95876,
      "cbid": 205, "correlation": 95876
    }
  },
  {
    "ph": "f", "id": 95876, "pid": 76337, "tid": -914061504, "ts": 1716454222775919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852499, "dur": 22,
    "args": {
      "External id": 95880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95880, "pid": 5, "tid": 7, "ts": 1716454222852499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775933, "dur": 12,
    "args": {
      "External id": 95880,
      "cbid": 211, "correlation": 95880
    }
  },
  {
    "ph": "s", "id": 95880, "pid": 76337, "tid": -914061504, "ts": 1716454222775933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222852521, "dur": 102,
    "args": {
      "External id": 95882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95882,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95882, "pid": 5, "tid": 7, "ts": 1716454222852521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775950, "dur": 6,
    "args": {
      "External id": 95882,
      "cbid": 211, "correlation": 95882
    }
  },
  {
    "ph": "s", "id": 95882, "pid": 76337, "tid": -914061504, "ts": 1716454222775950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852625, "dur": 22,
    "args": {
      "External id": 95884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95884,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95884, "pid": 5, "tid": 7, "ts": 1716454222852625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775960, "dur": 5,
    "args": {
      "External id": 95884,
      "cbid": 211, "correlation": 95884
    }
  },
  {
    "ph": "s", "id": 95884, "pid": 76337, "tid": -914061504, "ts": 1716454222775960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222852649, "dur": 32,
    "args": {
      "External id": 95890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95890, "pid": 5, "tid": 7, "ts": 1716454222852649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222775998, "dur": 10,
    "args": {
      "External id": 95890,
      "cbid": 211, "correlation": 95890
    }
  },
  {
    "ph": "s", "id": 95890, "pid": 76337, "tid": -914061504, "ts": 1716454222775998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222852682, "dur": 191,
    "args": {
      "External id": 95899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95899,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95899, "pid": 5, "tid": 7, "ts": 1716454222852682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776084, "dur": 15,
    "args": {
      "External id": 95899,
      "cbid": 211, "correlation": 95899
    }
  },
  {
    "ph": "s", "id": 95899, "pid": 76337, "tid": -914061504, "ts": 1716454222776084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222852874, "dur": 63,
    "args": {
      "External id": 95921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95921,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 95921, "pid": 5, "tid": 7, "ts": 1716454222852874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776143, "dur": 10,
    "args": {
      "External id": 95921,
      "cbid": 211, "correlation": 95921
    }
  },
  {
    "ph": "s", "id": 95921, "pid": 76337, "tid": -914061504, "ts": 1716454222776143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776237, "dur": 2,
    "args": {
      "External id": 95932,
      "cbid": 251, "correlation": 95932
    }
  },
  {
    "ph": "f", "id": 95932, "pid": 76337, "tid": -914061504, "ts": 1716454222776237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222852938, "dur": 149,
    "args": {
      "External id": 95933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95933,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95933, "pid": 5, "tid": 7, "ts": 1716454222852938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776243, "dur": 13,
    "args": {
      "External id": 95933,
      "cbid": 211, "correlation": 95933
    }
  },
  {
    "ph": "s", "id": 95933, "pid": 76337, "tid": -914061504, "ts": 1716454222776243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776313, "dur": 1,
    "args": {
      "External id": 95944,
      "cbid": 251, "correlation": 95944
    }
  },
  {
    "ph": "f", "id": 95944, "pid": 76337, "tid": -914061504, "ts": 1716454222776313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222853089, "dur": 144,
    "args": {
      "External id": 95945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95945,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95945, "pid": 5, "tid": 7, "ts": 1716454222853089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776317, "dur": 12,
    "args": {
      "External id": 95945,
      "cbid": 211, "correlation": 95945
    }
  },
  {
    "ph": "s", "id": 95945, "pid": 76337, "tid": -914061504, "ts": 1716454222776317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776382, "dur": 1,
    "args": {
      "External id": 95956,
      "cbid": 251, "correlation": 95956
    }
  },
  {
    "ph": "f", "id": 95956, "pid": 76337, "tid": -914061504, "ts": 1716454222776382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222853234, "dur": 143,
    "args": {
      "External id": 95957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95957,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 95957, "pid": 5, "tid": 7, "ts": 1716454222853234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776387, "dur": 12,
    "args": {
      "External id": 95957,
      "cbid": 211, "correlation": 95957
    }
  },
  {
    "ph": "s", "id": 95957, "pid": 76337, "tid": -914061504, "ts": 1716454222776387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222853378, "dur": 1911,
    "args": {
      "External id": 95978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95978,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 95978, "pid": 5, "tid": 7, "ts": 1716454222853378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776471, "dur": 13,
    "args": {
      "External id": 95978,
      "cbid": 211, "correlation": 95978
    }
  },
  {
    "ph": "s", "id": 95978, "pid": 76337, "tid": -914061504, "ts": 1716454222776471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776574, "dur": 1,
    "args": {
      "External id": 95996,
      "cbid": 251, "correlation": 95996
    }
  },
  {
    "ph": "f", "id": 95996, "pid": 76337, "tid": -914061504, "ts": 1716454222776574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222855290, "dur": 147,
    "args": {
      "External id": 95998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 95998,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 95998, "pid": 5, "tid": 7, "ts": 1716454222855290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776580, "dur": 13,
    "args": {
      "External id": 95998,
      "cbid": 211, "correlation": 95998
    }
  },
  {
    "ph": "s", "id": 95998, "pid": 76337, "tid": -914061504, "ts": 1716454222776580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222855439, "dur": 35,
    "args": {
      "External id": 96006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96006,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96006, "pid": 5, "tid": 7, "ts": 1716454222855439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776651, "dur": 13,
    "args": {
      "External id": 96006,
      "cbid": 211, "correlation": 96006
    }
  },
  {
    "ph": "s", "id": 96006, "pid": 76337, "tid": -914061504, "ts": 1716454222776651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222855475, "dur": 51,
    "args": {
      "External id": 96014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96014,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96014, "pid": 5, "tid": 7, "ts": 1716454222855475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776691, "dur": 9,
    "args": {
      "External id": 96014,
      "cbid": 211, "correlation": 96014
    }
  },
  {
    "ph": "s", "id": 96014, "pid": 76337, "tid": -914061504, "ts": 1716454222776691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222855528, "dur": 30,
    "args": {
      "External id": 96025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96025, "pid": 5, "tid": 7, "ts": 1716454222855528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776767, "dur": 13,
    "args": {
      "External id": 96025,
      "cbid": 211, "correlation": 96025
    }
  },
  {
    "ph": "s", "id": 96025, "pid": 76337, "tid": -914061504, "ts": 1716454222776767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222855559, "dur": 34,
    "args": {
      "External id": 96047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96047,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96047, "pid": 5, "tid": 7, "ts": 1716454222855559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776800, "dur": 8,
    "args": {
      "External id": 96047,
      "cbid": 211, "correlation": 96047
    }
  },
  {
    "ph": "s", "id": 96047, "pid": 76337, "tid": -914061504, "ts": 1716454222776800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776885, "dur": 1,
    "args": {
      "External id": 96058,
      "cbid": 251, "correlation": 96058
    }
  },
  {
    "ph": "f", "id": 96058, "pid": 76337, "tid": -914061504, "ts": 1716454222776885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222855595, "dur": 88,
    "args": {
      "External id": 96059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96059,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96059, "pid": 5, "tid": 7, "ts": 1716454222855595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776890, "dur": 13,
    "args": {
      "External id": 96059,
      "cbid": 211, "correlation": 96059
    }
  },
  {
    "ph": "s", "id": 96059, "pid": 76337, "tid": -914061504, "ts": 1716454222776890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776962, "dur": 1,
    "args": {
      "External id": 96070,
      "cbid": 251, "correlation": 96070
    }
  },
  {
    "ph": "f", "id": 96070, "pid": 76337, "tid": -914061504, "ts": 1716454222776962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776966, "dur": 0,
    "args": {
      "External id": 96071,
      "cbid": 251, "correlation": 96071
    }
  },
  {
    "ph": "f", "id": 96071, "pid": 76337, "tid": -914061504, "ts": 1716454222776966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222855684, "dur": 11,
    "args": {
      "External id": 96072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96072,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 96072, "pid": 5, "tid": 7, "ts": 1716454222855684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776967, "dur": 22,
    "args": {
      "External id": 96072,
      "cbid": 211, "correlation": 96072
    }
  },
  {
    "ph": "s", "id": 96072, "pid": 76337, "tid": -914061504, "ts": 1716454222776967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222855696, "dur": 5,
    "args": {
      "External id": 96074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96074,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 96074, "pid": 5, "tid": 7, "ts": 1716454222855696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222776991, "dur": 7,
    "args": {
      "External id": 96074,
      "cbid": 211, "correlation": 96074
    }
  },
  {
    "ph": "s", "id": 96074, "pid": 76337, "tid": -914061504, "ts": 1716454222776991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777052, "dur": 1,
    "args": {
      "External id": 96085,
      "cbid": 251, "correlation": 96085
    }
  },
  {
    "ph": "f", "id": 96085, "pid": 76337, "tid": -914061504, "ts": 1716454222777052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777056, "dur": 0,
    "args": {
      "External id": 96086,
      "cbid": 251, "correlation": 96086
    }
  },
  {
    "ph": "f", "id": 96086, "pid": 76337, "tid": -914061504, "ts": 1716454222777056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222855702, "dur": 7,
    "args": {
      "External id": 96087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96087,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 96087, "pid": 5, "tid": 7, "ts": 1716454222855702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777057, "dur": 11,
    "args": {
      "External id": 96087,
      "cbid": 211, "correlation": 96087
    }
  },
  {
    "ph": "s", "id": 96087, "pid": 76337, "tid": -914061504, "ts": 1716454222777057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222855710, "dur": 3,
    "args": {
      "External id": 96089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96089,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 96089, "pid": 5, "tid": 7, "ts": 1716454222855710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777070, "dur": 6,
    "args": {
      "External id": 96089,
      "cbid": 211, "correlation": 96089
    }
  },
  {
    "ph": "s", "id": 96089, "pid": 76337, "tid": -914061504, "ts": 1716454222777070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222855715, "dur": 90,
    "args": {
      "External id": 96110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96110,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 96110, "pid": 5, "tid": 7, "ts": 1716454222855715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777144, "dur": 12,
    "args": {
      "External id": 96110,
      "cbid": 211, "correlation": 96110
    }
  },
  {
    "ph": "s", "id": 96110, "pid": 76337, "tid": -914061504, "ts": 1716454222777144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777242, "dur": 1,
    "args": {
      "External id": 96128,
      "cbid": 251, "correlation": 96128
    }
  },
  {
    "ph": "f", "id": 96128, "pid": 76337, "tid": -914061504, "ts": 1716454222777242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222855806, "dur": 96,
    "args": {
      "External id": 96130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96130,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96130, "pid": 5, "tid": 7, "ts": 1716454222855806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777247, "dur": 13,
    "args": {
      "External id": 96130,
      "cbid": 211, "correlation": 96130
    }
  },
  {
    "ph": "s", "id": 96130, "pid": 76337, "tid": -914061504, "ts": 1716454222777247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222855903, "dur": 20,
    "args": {
      "External id": 96138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96138,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96138, "pid": 5, "tid": 7, "ts": 1716454222855903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777317, "dur": 13,
    "args": {
      "External id": 96138,
      "cbid": 211, "correlation": 96138
    }
  },
  {
    "ph": "s", "id": 96138, "pid": 76337, "tid": -914061504, "ts": 1716454222777317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222855924, "dur": 38,
    "args": {
      "External id": 96146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96146,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96146, "pid": 5, "tid": 7, "ts": 1716454222855924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777358, "dur": 9,
    "args": {
      "External id": 96146,
      "cbid": 211, "correlation": 96146
    }
  },
  {
    "ph": "s", "id": 96146, "pid": 76337, "tid": -914061504, "ts": 1716454222777358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222855964, "dur": 34,
    "args": {
      "External id": 96168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96168,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96168, "pid": 5, "tid": 7, "ts": 1716454222855964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777409, "dur": 10,
    "args": {
      "External id": 96168,
      "cbid": 211, "correlation": 96168
    }
  },
  {
    "ph": "s", "id": 96168, "pid": 76337, "tid": -914061504, "ts": 1716454222777409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777500, "dur": 1,
    "args": {
      "External id": 96184,
      "cbid": 251, "correlation": 96184
    }
  },
  {
    "ph": "f", "id": 96184, "pid": 76337, "tid": -914061504, "ts": 1716454222777500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777505, "dur": 0,
    "args": {
      "External id": 96186,
      "cbid": 251, "correlation": 96186
    }
  },
  {
    "ph": "f", "id": 96186, "pid": 76337, "tid": -914061504, "ts": 1716454222777505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222855999, "dur": 532,
    "args": {
      "External id": 96187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96187,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 96187, "pid": 5, "tid": 7, "ts": 1716454222855999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777509, "dur": 13,
    "args": {
      "External id": 96187,
      "cbid": 211, "correlation": 96187
    }
  },
  {
    "ph": "s", "id": 96187, "pid": 76337, "tid": -914061504, "ts": 1716454222777509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222856532, "dur": 124,
    "args": {
      "External id": 96195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96195,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96195, "pid": 5, "tid": 7, "ts": 1716454222856532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777578, "dur": 13,
    "args": {
      "External id": 96195,
      "cbid": 211, "correlation": 96195
    }
  },
  {
    "ph": "s", "id": 96195, "pid": 76337, "tid": -914061504, "ts": 1716454222777578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222856658, "dur": 126,
    "args": {
      "External id": 96203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96203, "pid": 5, "tid": 7, "ts": 1716454222856658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777611, "dur": 8,
    "args": {
      "External id": 96203,
      "cbid": 211, "correlation": 96203
    }
  },
  {
    "ph": "s", "id": 96203, "pid": 76337, "tid": -914061504, "ts": 1716454222777611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777689, "dur": 1,
    "args": {
      "External id": 96219,
      "cbid": 251, "correlation": 96219
    }
  },
  {
    "ph": "f", "id": 96219, "pid": 76337, "tid": -914061504, "ts": 1716454222777689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222856785, "dur": 306,
    "args": {
      "External id": 96221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96221,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96221, "pid": 5, "tid": 7, "ts": 1716454222856785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777694, "dur": 13,
    "args": {
      "External id": 96221,
      "cbid": 211, "correlation": 96221
    }
  },
  {
    "ph": "s", "id": 96221, "pid": 76337, "tid": -914061504, "ts": 1716454222777694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222857092, "dur": 27,
    "args": {
      "External id": 96229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96229,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96229, "pid": 5, "tid": 7, "ts": 1716454222857092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777737, "dur": 10,
    "args": {
      "External id": 96229,
      "cbid": 211, "correlation": 96229
    }
  },
  {
    "ph": "s", "id": 96229, "pid": 76337, "tid": -914061504, "ts": 1716454222777737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222857121, "dur": 80,
    "args": {
      "External id": 96240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96240,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96240, "pid": 5, "tid": 7, "ts": 1716454222857121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777806, "dur": 12,
    "args": {
      "External id": 96240,
      "cbid": 211, "correlation": 96240
    }
  },
  {
    "ph": "s", "id": 96240, "pid": 76337, "tid": -914061504, "ts": 1716454222777806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777873, "dur": 0,
    "args": {
      "External id": 96252,
      "cbid": 317, "correlation": 96252
    }
  },
  {
    "ph": "f", "id": 96252, "pid": 76337, "tid": -914061504, "ts": 1716454222777873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777874, "dur": 0,
    "args": {
      "External id": 96253,
      "cbid": 203, "correlation": 96253
    }
  },
  {
    "ph": "f", "id": 96253, "pid": 76337, "tid": -914061504, "ts": 1716454222777874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777875, "dur": 0,
    "args": {
      "External id": 96254,
      "cbid": 205, "correlation": 96254
    }
  },
  {
    "ph": "f", "id": 96254, "pid": 76337, "tid": -914061504, "ts": 1716454222777875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222857202, "dur": 22,
    "args": {
      "External id": 96258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96258,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96258, "pid": 5, "tid": 7, "ts": 1716454222857202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777892, "dur": 12,
    "args": {
      "External id": 96258,
      "cbid": 211, "correlation": 96258
    }
  },
  {
    "ph": "s", "id": 96258, "pid": 76337, "tid": -914061504, "ts": 1716454222777892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222857226, "dur": 119,
    "args": {
      "External id": 96260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96260,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96260, "pid": 5, "tid": 7, "ts": 1716454222857226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777911, "dur": 7,
    "args": {
      "External id": 96260,
      "cbid": 211, "correlation": 96260
    }
  },
  {
    "ph": "s", "id": 96260, "pid": 76337, "tid": -914061504, "ts": 1716454222777911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222857346, "dur": 24,
    "args": {
      "External id": 96262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96262,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96262, "pid": 5, "tid": 7, "ts": 1716454222857346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777922, "dur": 5,
    "args": {
      "External id": 96262,
      "cbid": 211, "correlation": 96262
    }
  },
  {
    "ph": "s", "id": 96262, "pid": 76337, "tid": -914061504, "ts": 1716454222777922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222857371, "dur": 33,
    "args": {
      "External id": 96268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96268,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96268, "pid": 5, "tid": 7, "ts": 1716454222857371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777950, "dur": 9,
    "args": {
      "External id": 96268,
      "cbid": 211, "correlation": 96268
    }
  },
  {
    "ph": "s", "id": 96268, "pid": 76337, "tid": -914061504, "ts": 1716454222777950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222857405, "dur": 26,
    "args": {
      "External id": 96276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96276,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96276, "pid": 5, "tid": 7, "ts": 1716454222857405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222777991, "dur": 9,
    "args": {
      "External id": 96276,
      "cbid": 211, "correlation": 96276
    }
  },
  {
    "ph": "s", "id": 96276, "pid": 76337, "tid": -914061504, "ts": 1716454222777991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222857433, "dur": 99,
    "args": {
      "External id": 96287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96287,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96287, "pid": 5, "tid": 7, "ts": 1716454222857433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778059, "dur": 12,
    "args": {
      "External id": 96287,
      "cbid": 211, "correlation": 96287
    }
  },
  {
    "ph": "s", "id": 96287, "pid": 76337, "tid": -914061504, "ts": 1716454222778059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778117, "dur": 0,
    "args": {
      "External id": 96297,
      "cbid": 317, "correlation": 96297
    }
  },
  {
    "ph": "f", "id": 96297, "pid": 76337, "tid": -914061504, "ts": 1716454222778117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778118, "dur": 0,
    "args": {
      "External id": 96298,
      "cbid": 203, "correlation": 96298
    }
  },
  {
    "ph": "f", "id": 96298, "pid": 76337, "tid": -914061504, "ts": 1716454222778118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778118, "dur": 0,
    "args": {
      "External id": 96299,
      "cbid": 205, "correlation": 96299
    }
  },
  {
    "ph": "f", "id": 96299, "pid": 76337, "tid": -914061504, "ts": 1716454222778118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222857534, "dur": 75,
    "args": {
      "External id": 96303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96303, "pid": 5, "tid": 7, "ts": 1716454222857534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778133, "dur": 11,
    "args": {
      "External id": 96303,
      "cbid": 211, "correlation": 96303
    }
  },
  {
    "ph": "s", "id": 96303, "pid": 76337, "tid": -914061504, "ts": 1716454222778133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222857610, "dur": 44,
    "args": {
      "External id": 96305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96305, "pid": 5, "tid": 7, "ts": 1716454222857610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778147, "dur": 6,
    "args": {
      "External id": 96305,
      "cbid": 211, "correlation": 96305
    }
  },
  {
    "ph": "s", "id": 96305, "pid": 76337, "tid": -914061504, "ts": 1716454222778147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222857655, "dur": 4,
    "args": {
      "External id": 96307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96307,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96307, "pid": 5, "tid": 7, "ts": 1716454222857655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778159, "dur": 6,
    "args": {
      "External id": 96307,
      "cbid": 211, "correlation": 96307
    }
  },
  {
    "ph": "s", "id": 96307, "pid": 76337, "tid": -914061504, "ts": 1716454222778159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778170, "dur": 0,
    "args": {
      "External id": 96308,
      "cbid": 51, "correlation": 96308
    }
  },
  {
    "ph": "s", "id": 96308, "pid": 76337, "tid": -914061504, "ts": 1716454222778170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222857660, "dur": 2220,
    "args": {
      "External id": 96309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96309,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96309, "pid": 5, "tid": 7, "ts": 1716454222857660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778171, "dur": 6,
    "args": {
      "External id": 96309,
      "cbid": 211, "correlation": 96309
    }
  },
  {
    "ph": "s", "id": 96309, "pid": 76337, "tid": -914061504, "ts": 1716454222778171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222859881, "dur": 111,
    "args": {
      "External id": 96314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96314, "pid": 5, "tid": 7, "ts": 1716454222859881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778200, "dur": 9,
    "args": {
      "External id": 96314,
      "cbid": 211, "correlation": 96314
    }
  },
  {
    "ph": "s", "id": 96314, "pid": 76337, "tid": -914061504, "ts": 1716454222778200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222859994, "dur": 166,
    "args": {
      "External id": 96323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96323,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96323, "pid": 5, "tid": 7, "ts": 1716454222859994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778297, "dur": 14,
    "args": {
      "External id": 96323,
      "cbid": 211, "correlation": 96323
    }
  },
  {
    "ph": "s", "id": 96323, "pid": 76337, "tid": -914061504, "ts": 1716454222778297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222860161, "dur": 129,
    "args": {
      "External id": 96343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96343,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 96343, "pid": 5, "tid": 7, "ts": 1716454222860161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778372, "dur": 11,
    "args": {
      "External id": 96343,
      "cbid": 211, "correlation": 96343
    }
  },
  {
    "ph": "s", "id": 96343, "pid": 76337, "tid": -914061504, "ts": 1716454222778372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222860292, "dur": 5,
    "args": {
      "External id": 96355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96355,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 96355, "pid": 5, "tid": 7, "ts": 1716454222860292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778393, "dur": 7,
    "args": {
      "External id": 96355,
      "cbid": 211, "correlation": 96355
    }
  },
  {
    "ph": "s", "id": 96355, "pid": 76337, "tid": -914061504, "ts": 1716454222778393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222860297, "dur": 158,
    "args": {
      "External id": 96358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96358, "pid": 5, "tid": 7, "ts": 1716454222860297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778412, "dur": 7,
    "args": {
      "External id": 96358,
      "cbid": 211, "correlation": 96358
    }
  },
  {
    "ph": "s", "id": 96358, "pid": 76337, "tid": -914061504, "ts": 1716454222778412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222860457, "dur": 101,
    "args": {
      "External id": 96367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96367,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96367, "pid": 5, "tid": 7, "ts": 1716454222860457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778454, "dur": 10,
    "args": {
      "External id": 96367,
      "cbid": 211, "correlation": 96367
    }
  },
  {
    "ph": "s", "id": 96367, "pid": 76337, "tid": -914061504, "ts": 1716454222778454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778508, "dur": 0,
    "args": {
      "External id": 96377,
      "cbid": 317, "correlation": 96377
    }
  },
  {
    "ph": "f", "id": 96377, "pid": 76337, "tid": -914061504, "ts": 1716454222778508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778509, "dur": 0,
    "args": {
      "External id": 96378,
      "cbid": 203, "correlation": 96378
    }
  },
  {
    "ph": "f", "id": 96378, "pid": 76337, "tid": -914061504, "ts": 1716454222778509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778510, "dur": 0,
    "args": {
      "External id": 96379,
      "cbid": 205, "correlation": 96379
    }
  },
  {
    "ph": "f", "id": 96379, "pid": 76337, "tid": -914061504, "ts": 1716454222778510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222860559, "dur": 110,
    "args": {
      "External id": 96383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96383,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96383, "pid": 5, "tid": 7, "ts": 1716454222860559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778526, "dur": 12,
    "args": {
      "External id": 96383,
      "cbid": 211, "correlation": 96383
    }
  },
  {
    "ph": "s", "id": 96383, "pid": 76337, "tid": -914061504, "ts": 1716454222778526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222860671, "dur": 34,
    "args": {
      "External id": 96385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96385,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96385, "pid": 5, "tid": 7, "ts": 1716454222860671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778540, "dur": 5,
    "args": {
      "External id": 96385,
      "cbid": 211, "correlation": 96385
    }
  },
  {
    "ph": "s", "id": 96385, "pid": 76337, "tid": -914061504, "ts": 1716454222778540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222860706, "dur": 4,
    "args": {
      "External id": 96387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96387, "pid": 5, "tid": 7, "ts": 1716454222860706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778551, "dur": 6,
    "args": {
      "External id": 96387,
      "cbid": 211, "correlation": 96387
    }
  },
  {
    "ph": "s", "id": 96387, "pid": 76337, "tid": -914061504, "ts": 1716454222778551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778561, "dur": 0,
    "args": {
      "External id": 96388,
      "cbid": 51, "correlation": 96388
    }
  },
  {
    "ph": "s", "id": 96388, "pid": 76337, "tid": -914061504, "ts": 1716454222778561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222860711, "dur": 1981,
    "args": {
      "External id": 96389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96389,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96389, "pid": 5, "tid": 7, "ts": 1716454222860711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778562, "dur": 6,
    "args": {
      "External id": 96389,
      "cbid": 211, "correlation": 96389
    }
  },
  {
    "ph": "s", "id": 96389, "pid": 76337, "tid": -914061504, "ts": 1716454222778562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222862693, "dur": 58,
    "args": {
      "External id": 96394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96394, "pid": 5, "tid": 7, "ts": 1716454222862693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778590, "dur": 10,
    "args": {
      "External id": 96394,
      "cbid": 211, "correlation": 96394
    }
  },
  {
    "ph": "s", "id": 96394, "pid": 76337, "tid": -914061504, "ts": 1716454222778590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222862752, "dur": 4,
    "args": {
      "External id": 96402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96402,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96402, "pid": 5, "tid": 7, "ts": 1716454222862752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778634, "dur": 10,
    "args": {
      "External id": 96402,
      "cbid": 211, "correlation": 96402
    }
  },
  {
    "ph": "s", "id": 96402, "pid": 76337, "tid": -914061504, "ts": 1716454222778634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778703, "dur": 1,
    "args": {
      "External id": 96418,
      "cbid": 251, "correlation": 96418
    }
  },
  {
    "ph": "f", "id": 96418, "pid": 76337, "tid": -914061504, "ts": 1716454222778703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778708, "dur": 0,
    "args": {
      "External id": 96420,
      "cbid": 251, "correlation": 96420
    }
  },
  {
    "ph": "f", "id": 96420, "pid": 76337, "tid": -914061504, "ts": 1716454222778708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222862757, "dur": 11,
    "args": {
      "External id": 96421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96421,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 96421, "pid": 5, "tid": 7, "ts": 1716454222862757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778711, "dur": 12,
    "args": {
      "External id": 96421,
      "cbid": 211, "correlation": 96421
    }
  },
  {
    "ph": "s", "id": 96421, "pid": 76337, "tid": -914061504, "ts": 1716454222778711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222862770, "dur": 5,
    "args": {
      "External id": 96423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96423,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 96423, "pid": 5, "tid": 7, "ts": 1716454222862770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778725, "dur": 6,
    "args": {
      "External id": 96423,
      "cbid": 211, "correlation": 96423
    }
  },
  {
    "ph": "s", "id": 96423, "pid": 76337, "tid": -914061504, "ts": 1716454222778725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222862776, "dur": 53,
    "args": {
      "External id": 96433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96433,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96433, "pid": 5, "tid": 7, "ts": 1716454222862776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778784, "dur": 12,
    "args": {
      "External id": 96433,
      "cbid": 211, "correlation": 96433
    }
  },
  {
    "ph": "s", "id": 96433, "pid": 76337, "tid": -914061504, "ts": 1716454222778784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222862830, "dur": 50,
    "args": {
      "External id": 96453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96453,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 96453, "pid": 5, "tid": 7, "ts": 1716454222862830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778851, "dur": 11,
    "args": {
      "External id": 96453,
      "cbid": 211, "correlation": 96453
    }
  },
  {
    "ph": "s", "id": 96453, "pid": 76337, "tid": -914061504, "ts": 1716454222778851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222862881, "dur": 4,
    "args": {
      "External id": 96465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96465,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96465, "pid": 5, "tid": 7, "ts": 1716454222862881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778872, "dur": 6,
    "args": {
      "External id": 96465,
      "cbid": 211, "correlation": 96465
    }
  },
  {
    "ph": "s", "id": 96465, "pid": 76337, "tid": -914061504, "ts": 1716454222778872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222862886, "dur": 55,
    "args": {
      "External id": 96468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96468, "pid": 5, "tid": 7, "ts": 1716454222862886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778890, "dur": 7,
    "args": {
      "External id": 96468,
      "cbid": 211, "correlation": 96468
    }
  },
  {
    "ph": "s", "id": 96468, "pid": 76337, "tid": -914061504, "ts": 1716454222778890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222862943, "dur": 36,
    "args": {
      "External id": 96477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96477,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96477, "pid": 5, "tid": 7, "ts": 1716454222862943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222778930, "dur": 11,
    "args": {
      "External id": 96477,
      "cbid": 211, "correlation": 96477
    }
  },
  {
    "ph": "s", "id": 96477, "pid": 76337, "tid": -914061504, "ts": 1716454222778930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779003, "dur": 0,
    "args": {
      "External id": 96487,
      "cbid": 317, "correlation": 96487
    }
  },
  {
    "ph": "f", "id": 96487, "pid": 76337, "tid": -914061504, "ts": 1716454222779003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779004, "dur": 0,
    "args": {
      "External id": 96488,
      "cbid": 203, "correlation": 96488
    }
  },
  {
    "ph": "f", "id": 96488, "pid": 76337, "tid": -914061504, "ts": 1716454222779004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779005, "dur": 0,
    "args": {
      "External id": 96489,
      "cbid": 205, "correlation": 96489
    }
  },
  {
    "ph": "f", "id": 96489, "pid": 76337, "tid": -914061504, "ts": 1716454222779005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222862980, "dur": 39,
    "args": {
      "External id": 96493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96493,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96493, "pid": 5, "tid": 7, "ts": 1716454222862980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779021, "dur": 12,
    "args": {
      "External id": 96493,
      "cbid": 211, "correlation": 96493
    }
  },
  {
    "ph": "s", "id": 96493, "pid": 76337, "tid": -914061504, "ts": 1716454222779021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222863021, "dur": 14,
    "args": {
      "External id": 96495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96495,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96495, "pid": 5, "tid": 7, "ts": 1716454222863021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779036, "dur": 5,
    "args": {
      "External id": 96495,
      "cbid": 211, "correlation": 96495
    }
  },
  {
    "ph": "s", "id": 96495, "pid": 76337, "tid": -914061504, "ts": 1716454222779036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222863036, "dur": 3,
    "args": {
      "External id": 96497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96497, "pid": 5, "tid": 7, "ts": 1716454222863036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779046, "dur": 6,
    "args": {
      "External id": 96497,
      "cbid": 211, "correlation": 96497
    }
  },
  {
    "ph": "s", "id": 96497, "pid": 76337, "tid": -914061504, "ts": 1716454222779046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779055, "dur": 0,
    "args": {
      "External id": 96498,
      "cbid": 51, "correlation": 96498
    }
  },
  {
    "ph": "s", "id": 96498, "pid": 76337, "tid": -914061504, "ts": 1716454222779055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222863040, "dur": 689,
    "args": {
      "External id": 96499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96499,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96499, "pid": 5, "tid": 7, "ts": 1716454222863040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779056, "dur": 5,
    "args": {
      "External id": 96499,
      "cbid": 211, "correlation": 96499
    }
  },
  {
    "ph": "s", "id": 96499, "pid": 76337, "tid": -914061504, "ts": 1716454222779056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222863731, "dur": 58,
    "args": {
      "External id": 96504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96504,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96504, "pid": 5, "tid": 7, "ts": 1716454222863731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779084, "dur": 8,
    "args": {
      "External id": 96504,
      "cbid": 211, "correlation": 96504
    }
  },
  {
    "ph": "s", "id": 96504, "pid": 76337, "tid": -914061504, "ts": 1716454222779084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779141, "dur": 0,
    "args": {
      "External id": 96514,
      "cbid": 317, "correlation": 96514
    }
  },
  {
    "ph": "f", "id": 96514, "pid": 76337, "tid": -914061504, "ts": 1716454222779141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779141, "dur": 0,
    "args": {
      "External id": 96515,
      "cbid": 203, "correlation": 96515
    }
  },
  {
    "ph": "f", "id": 96515, "pid": 76337, "tid": -914061504, "ts": 1716454222779141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779142, "dur": 0,
    "args": {
      "External id": 96516,
      "cbid": 205, "correlation": 96516
    }
  },
  {
    "ph": "f", "id": 96516, "pid": 76337, "tid": -914061504, "ts": 1716454222779142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222863790, "dur": 3,
    "args": {
      "External id": 96520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96520, "pid": 5, "tid": 7, "ts": 1716454222863790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779158, "dur": 11,
    "args": {
      "External id": 96520,
      "cbid": 211, "correlation": 96520
    }
  },
  {
    "ph": "s", "id": 96520, "pid": 76337, "tid": -914061504, "ts": 1716454222779158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779174, "dur": 0,
    "args": {
      "External id": 96521,
      "cbid": 51, "correlation": 96521
    }
  },
  {
    "ph": "s", "id": 96521, "pid": 76337, "tid": -914061504, "ts": 1716454222779174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222863795, "dur": 264,
    "args": {
      "External id": 96522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96522,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96522, "pid": 5, "tid": 7, "ts": 1716454222863795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779175, "dur": 7,
    "args": {
      "External id": 96522,
      "cbid": 211, "correlation": 96522
    }
  },
  {
    "ph": "s", "id": 96522, "pid": 76337, "tid": -914061504, "ts": 1716454222779175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222864060, "dur": 58,
    "args": {
      "External id": 96527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96527,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96527, "pid": 5, "tid": 7, "ts": 1716454222864060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779201, "dur": 9,
    "args": {
      "External id": 96527,
      "cbid": 211, "correlation": 96527
    }
  },
  {
    "ph": "s", "id": 96527, "pid": 76337, "tid": -914061504, "ts": 1716454222779201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222864119, "dur": 50,
    "args": {
      "External id": 96535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96535,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96535, "pid": 5, "tid": 7, "ts": 1716454222864119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779230, "dur": 8,
    "args": {
      "External id": 96535,
      "cbid": 211, "correlation": 96535
    }
  },
  {
    "ph": "s", "id": 96535, "pid": 76337, "tid": -914061504, "ts": 1716454222779230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222864170, "dur": 35,
    "args": {
      "External id": 96543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96543,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96543, "pid": 5, "tid": 7, "ts": 1716454222864170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779260, "dur": 10,
    "args": {
      "External id": 96543,
      "cbid": 211, "correlation": 96543
    }
  },
  {
    "ph": "s", "id": 96543, "pid": 76337, "tid": -914061504, "ts": 1716454222779260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222864207, "dur": 51,
    "args": {
      "External id": 96563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96563,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 96563, "pid": 5, "tid": 7, "ts": 1716454222864207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779343, "dur": 12,
    "args": {
      "External id": 96563,
      "cbid": 211, "correlation": 96563
    }
  },
  {
    "ph": "s", "id": 96563, "pid": 76337, "tid": -914061504, "ts": 1716454222779343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222864259, "dur": 4,
    "args": {
      "External id": 96575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96575,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 96575, "pid": 5, "tid": 7, "ts": 1716454222864259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779365, "dur": 6,
    "args": {
      "External id": 96575,
      "cbid": 211, "correlation": 96575
    }
  },
  {
    "ph": "s", "id": 96575, "pid": 76337, "tid": -914061504, "ts": 1716454222779365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222864264, "dur": 54,
    "args": {
      "External id": 96578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96578, "pid": 5, "tid": 7, "ts": 1716454222864264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779383, "dur": 7,
    "args": {
      "External id": 96578,
      "cbid": 211, "correlation": 96578
    }
  },
  {
    "ph": "s", "id": 96578, "pid": 76337, "tid": -914061504, "ts": 1716454222779383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779441, "dur": 0,
    "args": {
      "External id": 96589,
      "cbid": 317, "correlation": 96589
    }
  },
  {
    "ph": "f", "id": 96589, "pid": 76337, "tid": -914061504, "ts": 1716454222779441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779442, "dur": 0,
    "args": {
      "External id": 96590,
      "cbid": 203, "correlation": 96590
    }
  },
  {
    "ph": "f", "id": 96590, "pid": 76337, "tid": -914061504, "ts": 1716454222779442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779442, "dur": 0,
    "args": {
      "External id": 96591,
      "cbid": 205, "correlation": 96591
    }
  },
  {
    "ph": "f", "id": 96591, "pid": 76337, "tid": -914061504, "ts": 1716454222779442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779475, "dur": 2,
    "args": {
      "External id": 96595,
      "cbid": 251, "correlation": 96595
    }
  },
  {
    "ph": "f", "id": 96595, "pid": 76337, "tid": -914061504, "ts": 1716454222779475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779479, "dur": 1,
    "args": {
      "External id": 96596,
      "cbid": 251, "correlation": 96596
    }
  },
  {
    "ph": "f", "id": 96596, "pid": 76337, "tid": -914061504, "ts": 1716454222779479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779481, "dur": 1,
    "args": {
      "External id": 96597,
      "cbid": 251, "correlation": 96597
    }
  },
  {
    "ph": "f", "id": 96597, "pid": 76337, "tid": -914061504, "ts": 1716454222779481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779483, "dur": 1,
    "args": {
      "External id": 96598,
      "cbid": 251, "correlation": 96598
    }
  },
  {
    "ph": "f", "id": 96598, "pid": 76337, "tid": -914061504, "ts": 1716454222779483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779485, "dur": 1,
    "args": {
      "External id": 96599,
      "cbid": 251, "correlation": 96599
    }
  },
  {
    "ph": "f", "id": 96599, "pid": 76337, "tid": -914061504, "ts": 1716454222779485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779487, "dur": 1,
    "args": {
      "External id": 96600,
      "cbid": 251, "correlation": 96600
    }
  },
  {
    "ph": "f", "id": 96600, "pid": 76337, "tid": -914061504, "ts": 1716454222779487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779489, "dur": 1,
    "args": {
      "External id": 96601,
      "cbid": 251, "correlation": 96601
    }
  },
  {
    "ph": "f", "id": 96601, "pid": 76337, "tid": -914061504, "ts": 1716454222779489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779491, "dur": 1,
    "args": {
      "External id": 96602,
      "cbid": 251, "correlation": 96602
    }
  },
  {
    "ph": "f", "id": 96602, "pid": 76337, "tid": -914061504, "ts": 1716454222779491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779493, "dur": 0,
    "args": {
      "External id": 96603,
      "cbid": 251, "correlation": 96603
    }
  },
  {
    "ph": "f", "id": 96603, "pid": 76337, "tid": -914061504, "ts": 1716454222779493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222864319, "dur": 113,
    "args": {
      "External id": 96604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96604,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 96604, "pid": 5, "tid": 7, "ts": 1716454222864319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779496, "dur": 13,
    "args": {
      "External id": 96604,
      "cbid": 211, "correlation": 96604
    }
  },
  {
    "ph": "s", "id": 96604, "pid": 76337, "tid": -914061504, "ts": 1716454222779496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222864433, "dur": 60,
    "args": {
      "External id": 96610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96610,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96610, "pid": 5, "tid": 7, "ts": 1716454222864433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779532, "dur": 10,
    "args": {
      "External id": 96610,
      "cbid": 211, "correlation": 96610
    }
  },
  {
    "ph": "s", "id": 96610, "pid": 76337, "tid": -914061504, "ts": 1716454222779532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222864494, "dur": 489,
    "args": {
      "External id": 96619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96619,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96619, "pid": 5, "tid": 7, "ts": 1716454222864494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779617, "dur": 15,
    "args": {
      "External id": 96619,
      "cbid": 211, "correlation": 96619
    }
  },
  {
    "ph": "s", "id": 96619, "pid": 76337, "tid": -914061504, "ts": 1716454222779617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222864984, "dur": 177,
    "args": {
      "External id": 96641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96641,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96641, "pid": 5, "tid": 7, "ts": 1716454222864984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779678, "dur": 11,
    "args": {
      "External id": 96641,
      "cbid": 211, "correlation": 96641
    }
  },
  {
    "ph": "s", "id": 96641, "pid": 76337, "tid": -914061504, "ts": 1716454222779678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779773, "dur": 2,
    "args": {
      "External id": 96652,
      "cbid": 251, "correlation": 96652
    }
  },
  {
    "ph": "f", "id": 96652, "pid": 76337, "tid": -914061504, "ts": 1716454222779773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222865162, "dur": 196,
    "args": {
      "External id": 96653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96653,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96653, "pid": 5, "tid": 7, "ts": 1716454222865162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779779, "dur": 12,
    "args": {
      "External id": 96653,
      "cbid": 211, "correlation": 96653
    }
  },
  {
    "ph": "s", "id": 96653, "pid": 76337, "tid": -914061504, "ts": 1716454222779779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779848, "dur": 1,
    "args": {
      "External id": 96664,
      "cbid": 251, "correlation": 96664
    }
  },
  {
    "ph": "f", "id": 96664, "pid": 76337, "tid": -914061504, "ts": 1716454222779848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222865360, "dur": 187,
    "args": {
      "External id": 96665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96665,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96665, "pid": 5, "tid": 7, "ts": 1716454222865360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779852, "dur": 11,
    "args": {
      "External id": 96665,
      "cbid": 211, "correlation": 96665
    }
  },
  {
    "ph": "s", "id": 96665, "pid": 76337, "tid": -914061504, "ts": 1716454222779852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779915, "dur": 1,
    "args": {
      "External id": 96676,
      "cbid": 251, "correlation": 96676
    }
  },
  {
    "ph": "f", "id": 96676, "pid": 76337, "tid": -914061504, "ts": 1716454222779915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222865548, "dur": 183,
    "args": {
      "External id": 96677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96677,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96677, "pid": 5, "tid": 7, "ts": 1716454222865548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222779919, "dur": 12,
    "args": {
      "External id": 96677,
      "cbid": 211, "correlation": 96677
    }
  },
  {
    "ph": "s", "id": 96677, "pid": 76337, "tid": -914061504, "ts": 1716454222779919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222865733, "dur": 18125,
    "args": {
      "External id": 96698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96698,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 96698, "pid": 5, "tid": 7, "ts": 1716454222865733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780012, "dur": 15,
    "args": {
      "External id": 96698,
      "cbid": 211, "correlation": 96698
    }
  },
  {
    "ph": "s", "id": 96698, "pid": 76337, "tid": -914061504, "ts": 1716454222780012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780116, "dur": 1,
    "args": {
      "External id": 96716,
      "cbid": 251, "correlation": 96716
    }
  },
  {
    "ph": "f", "id": 96716, "pid": 76337, "tid": -914061504, "ts": 1716454222780116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222883859, "dur": 198,
    "args": {
      "External id": 96718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96718,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96718, "pid": 5, "tid": 7, "ts": 1716454222883859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780122, "dur": 13,
    "args": {
      "External id": 96718,
      "cbid": 211, "correlation": 96718
    }
  },
  {
    "ph": "s", "id": 96718, "pid": 76337, "tid": -914061504, "ts": 1716454222780122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222884059, "dur": 67,
    "args": {
      "External id": 96726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96726,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96726, "pid": 5, "tid": 7, "ts": 1716454222884059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780192, "dur": 13,
    "args": {
      "External id": 96726,
      "cbid": 211, "correlation": 96726
    }
  },
  {
    "ph": "s", "id": 96726, "pid": 76337, "tid": -914061504, "ts": 1716454222780192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222884127, "dur": 97,
    "args": {
      "External id": 96734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96734,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96734, "pid": 5, "tid": 7, "ts": 1716454222884127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780232, "dur": 9,
    "args": {
      "External id": 96734,
      "cbid": 211, "correlation": 96734
    }
  },
  {
    "ph": "s", "id": 96734, "pid": 76337, "tid": -914061504, "ts": 1716454222780232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222884225, "dur": 54,
    "args": {
      "External id": 96745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96745, "pid": 5, "tid": 7, "ts": 1716454222884225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780308, "dur": 14,
    "args": {
      "External id": 96745,
      "cbid": 211, "correlation": 96745
    }
  },
  {
    "ph": "s", "id": 96745, "pid": 76337, "tid": -914061504, "ts": 1716454222780308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222884280, "dur": 90,
    "args": {
      "External id": 96767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96767,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96767, "pid": 5, "tid": 7, "ts": 1716454222884280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780340, "dur": 8,
    "args": {
      "External id": 96767,
      "cbid": 211, "correlation": 96767
    }
  },
  {
    "ph": "s", "id": 96767, "pid": 76337, "tid": -914061504, "ts": 1716454222780340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780424, "dur": 1,
    "args": {
      "External id": 96778,
      "cbid": 251, "correlation": 96778
    }
  },
  {
    "ph": "f", "id": 96778, "pid": 76337, "tid": -914061504, "ts": 1716454222780424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222884371, "dur": 103,
    "args": {
      "External id": 96779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96779,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96779, "pid": 5, "tid": 7, "ts": 1716454222884371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780429, "dur": 12,
    "args": {
      "External id": 96779,
      "cbid": 211, "correlation": 96779
    }
  },
  {
    "ph": "s", "id": 96779, "pid": 76337, "tid": -914061504, "ts": 1716454222780429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780506, "dur": 1,
    "args": {
      "External id": 96790,
      "cbid": 251, "correlation": 96790
    }
  },
  {
    "ph": "f", "id": 96790, "pid": 76337, "tid": -914061504, "ts": 1716454222780506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780510, "dur": 0,
    "args": {
      "External id": 96791,
      "cbid": 251, "correlation": 96791
    }
  },
  {
    "ph": "f", "id": 96791, "pid": 76337, "tid": -914061504, "ts": 1716454222780510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222884476, "dur": 10,
    "args": {
      "External id": 96792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96792,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 96792, "pid": 5, "tid": 7, "ts": 1716454222884476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780512, "dur": 15,
    "args": {
      "External id": 96792,
      "cbid": 211, "correlation": 96792
    }
  },
  {
    "ph": "s", "id": 96792, "pid": 76337, "tid": -914061504, "ts": 1716454222780512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222884487, "dur": 5,
    "args": {
      "External id": 96794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96794,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 96794, "pid": 5, "tid": 7, "ts": 1716454222884487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780530, "dur": 8,
    "args": {
      "External id": 96794,
      "cbid": 211, "correlation": 96794
    }
  },
  {
    "ph": "s", "id": 96794, "pid": 76337, "tid": -914061504, "ts": 1716454222780530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780593, "dur": 1,
    "args": {
      "External id": 96805,
      "cbid": 251, "correlation": 96805
    }
  },
  {
    "ph": "f", "id": 96805, "pid": 76337, "tid": -914061504, "ts": 1716454222780593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780597, "dur": 0,
    "args": {
      "External id": 96806,
      "cbid": 251, "correlation": 96806
    }
  },
  {
    "ph": "f", "id": 96806, "pid": 76337, "tid": -914061504, "ts": 1716454222780597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222884493, "dur": 6,
    "args": {
      "External id": 96807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96807,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 96807, "pid": 5, "tid": 7, "ts": 1716454222884493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780598, "dur": 12,
    "args": {
      "External id": 96807,
      "cbid": 211, "correlation": 96807
    }
  },
  {
    "ph": "s", "id": 96807, "pid": 76337, "tid": -914061504, "ts": 1716454222780598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222884501, "dur": 3,
    "args": {
      "External id": 96809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96809,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 96809, "pid": 5, "tid": 7, "ts": 1716454222884501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780611, "dur": 5,
    "args": {
      "External id": 96809,
      "cbid": 211, "correlation": 96809
    }
  },
  {
    "ph": "s", "id": 96809, "pid": 76337, "tid": -914061504, "ts": 1716454222780611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222884505, "dur": 152,
    "args": {
      "External id": 96830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96830,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 96830, "pid": 5, "tid": 7, "ts": 1716454222884505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780685, "dur": 13,
    "args": {
      "External id": 96830,
      "cbid": 211, "correlation": 96830
    }
  },
  {
    "ph": "s", "id": 96830, "pid": 76337, "tid": -914061504, "ts": 1716454222780685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780782, "dur": 2,
    "args": {
      "External id": 96848,
      "cbid": 251, "correlation": 96848
    }
  },
  {
    "ph": "f", "id": 96848, "pid": 76337, "tid": -914061504, "ts": 1716454222780782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222884659, "dur": 106,
    "args": {
      "External id": 96850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96850,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 96850, "pid": 5, "tid": 7, "ts": 1716454222884659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780788, "dur": 14,
    "args": {
      "External id": 96850,
      "cbid": 211, "correlation": 96850
    }
  },
  {
    "ph": "s", "id": 96850, "pid": 76337, "tid": -914061504, "ts": 1716454222780788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222884767, "dur": 35,
    "args": {
      "External id": 96858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96858,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96858, "pid": 5, "tid": 7, "ts": 1716454222884767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780858, "dur": 12,
    "args": {
      "External id": 96858,
      "cbid": 211, "correlation": 96858
    }
  },
  {
    "ph": "s", "id": 96858, "pid": 76337, "tid": -914061504, "ts": 1716454222780858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222884803, "dur": 67,
    "args": {
      "External id": 96866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96866, "pid": 5, "tid": 7, "ts": 1716454222884803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780899, "dur": 9,
    "args": {
      "External id": 96866,
      "cbid": 211, "correlation": 96866
    }
  },
  {
    "ph": "s", "id": 96866, "pid": 76337, "tid": -914061504, "ts": 1716454222780899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222884871, "dur": 90,
    "args": {
      "External id": 96888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96888,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96888, "pid": 5, "tid": 7, "ts": 1716454222884871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222780950, "dur": 10,
    "args": {
      "External id": 96888,
      "cbid": 211, "correlation": 96888
    }
  },
  {
    "ph": "s", "id": 96888, "pid": 76337, "tid": -914061504, "ts": 1716454222780950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781048, "dur": 1,
    "args": {
      "External id": 96904,
      "cbid": 251, "correlation": 96904
    }
  },
  {
    "ph": "f", "id": 96904, "pid": 76337, "tid": -914061504, "ts": 1716454222781048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222884962, "dur": 564,
    "args": {
      "External id": 96906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96906,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 96906, "pid": 5, "tid": 7, "ts": 1716454222884962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781054, "dur": 13,
    "args": {
      "External id": 96906,
      "cbid": 211, "correlation": 96906
    }
  },
  {
    "ph": "s", "id": 96906, "pid": 76337, "tid": -914061504, "ts": 1716454222781054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222885527, "dur": 239,
    "args": {
      "External id": 96914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96914, "pid": 5, "tid": 7, "ts": 1716454222885527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781125, "dur": 13,
    "args": {
      "External id": 96914,
      "cbid": 211, "correlation": 96914
    }
  },
  {
    "ph": "s", "id": 96914, "pid": 76337, "tid": -914061504, "ts": 1716454222781125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222885767, "dur": 249,
    "args": {
      "External id": 96922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96922, "pid": 5, "tid": 7, "ts": 1716454222885767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781158, "dur": 9,
    "args": {
      "External id": 96922,
      "cbid": 211, "correlation": 96922
    }
  },
  {
    "ph": "s", "id": 96922, "pid": 76337, "tid": -914061504, "ts": 1716454222781158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781242, "dur": 1,
    "args": {
      "External id": 96938,
      "cbid": 251, "correlation": 96938
    }
  },
  {
    "ph": "f", "id": 96938, "pid": 76337, "tid": -914061504, "ts": 1716454222781242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781247, "dur": 0,
    "args": {
      "External id": 96940,
      "cbid": 251, "correlation": 96940
    }
  },
  {
    "ph": "f", "id": 96940, "pid": 76337, "tid": -914061504, "ts": 1716454222781247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222886017, "dur": 351,
    "args": {
      "External id": 96941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96941,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 96941, "pid": 5, "tid": 7, "ts": 1716454222886017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781252, "dur": 13,
    "args": {
      "External id": 96941,
      "cbid": 211, "correlation": 96941
    }
  },
  {
    "ph": "s", "id": 96941, "pid": 76337, "tid": -914061504, "ts": 1716454222781252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222886369, "dur": 50,
    "args": {
      "External id": 96949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96949,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96949, "pid": 5, "tid": 7, "ts": 1716454222886369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781294, "dur": 11,
    "args": {
      "External id": 96949,
      "cbid": 211, "correlation": 96949
    }
  },
  {
    "ph": "s", "id": 96949, "pid": 76337, "tid": -914061504, "ts": 1716454222781294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222886421, "dur": 155,
    "args": {
      "External id": 96960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96960, "pid": 5, "tid": 7, "ts": 1716454222886421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781363, "dur": 12,
    "args": {
      "External id": 96960,
      "cbid": 211, "correlation": 96960
    }
  },
  {
    "ph": "s", "id": 96960, "pid": 76337, "tid": -914061504, "ts": 1716454222781363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781429, "dur": 0,
    "args": {
      "External id": 96972,
      "cbid": 317, "correlation": 96972
    }
  },
  {
    "ph": "f", "id": 96972, "pid": 76337, "tid": -914061504, "ts": 1716454222781429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781430, "dur": 0,
    "args": {
      "External id": 96973,
      "cbid": 203, "correlation": 96973
    }
  },
  {
    "ph": "f", "id": 96973, "pid": 76337, "tid": -914061504, "ts": 1716454222781430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781431, "dur": 0,
    "args": {
      "External id": 96974,
      "cbid": 205, "correlation": 96974
    }
  },
  {
    "ph": "f", "id": 96974, "pid": 76337, "tid": -914061504, "ts": 1716454222781431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781456, "dur": 1,
    "args": {
      "External id": 96978,
      "cbid": 251, "correlation": 96978
    }
  },
  {
    "ph": "f", "id": 96978, "pid": 76337, "tid": -914061504, "ts": 1716454222781456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781458, "dur": 1,
    "args": {
      "External id": 96979,
      "cbid": 251, "correlation": 96979
    }
  },
  {
    "ph": "f", "id": 96979, "pid": 76337, "tid": -914061504, "ts": 1716454222781458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781460, "dur": 1,
    "args": {
      "External id": 96980,
      "cbid": 251, "correlation": 96980
    }
  },
  {
    "ph": "f", "id": 96980, "pid": 76337, "tid": -914061504, "ts": 1716454222781460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781461, "dur": 0,
    "args": {
      "External id": 96981,
      "cbid": 251, "correlation": 96981
    }
  },
  {
    "ph": "f", "id": 96981, "pid": 76337, "tid": -914061504, "ts": 1716454222781461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781463, "dur": 1,
    "args": {
      "External id": 96982,
      "cbid": 251, "correlation": 96982
    }
  },
  {
    "ph": "f", "id": 96982, "pid": 76337, "tid": -914061504, "ts": 1716454222781463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781464, "dur": 0,
    "args": {
      "External id": 96983,
      "cbid": 251, "correlation": 96983
    }
  },
  {
    "ph": "f", "id": 96983, "pid": 76337, "tid": -914061504, "ts": 1716454222781464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781465, "dur": 0,
    "args": {
      "External id": 96984,
      "cbid": 251, "correlation": 96984
    }
  },
  {
    "ph": "f", "id": 96984, "pid": 76337, "tid": -914061504, "ts": 1716454222781465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781466, "dur": 0,
    "args": {
      "External id": 96985,
      "cbid": 251, "correlation": 96985
    }
  },
  {
    "ph": "f", "id": 96985, "pid": 76337, "tid": -914061504, "ts": 1716454222781466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781468, "dur": 0,
    "args": {
      "External id": 96986,
      "cbid": 251, "correlation": 96986
    }
  },
  {
    "ph": "f", "id": 96986, "pid": 76337, "tid": -914061504, "ts": 1716454222781468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222886578, "dur": 113,
    "args": {
      "External id": 96987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96987,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 96987, "pid": 5, "tid": 7, "ts": 1716454222886578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781470, "dur": 12,
    "args": {
      "External id": 96987,
      "cbid": 211, "correlation": 96987
    }
  },
  {
    "ph": "s", "id": 96987, "pid": 76337, "tid": -914061504, "ts": 1716454222781470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222886692, "dur": 60,
    "args": {
      "External id": 96993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 96993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 96993, "pid": 5, "tid": 7, "ts": 1716454222886692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781504, "dur": 10,
    "args": {
      "External id": 96993,
      "cbid": 211, "correlation": 96993
    }
  },
  {
    "ph": "s", "id": 96993, "pid": 76337, "tid": -914061504, "ts": 1716454222781504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222886753, "dur": 50,
    "args": {
      "External id": 97001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97001,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97001, "pid": 5, "tid": 7, "ts": 1716454222886753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781536, "dur": 8,
    "args": {
      "External id": 97001,
      "cbid": 211, "correlation": 97001
    }
  },
  {
    "ph": "s", "id": 97001, "pid": 76337, "tid": -914061504, "ts": 1716454222781536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222886804, "dur": 98,
    "args": {
      "External id": 97010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97010,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97010, "pid": 5, "tid": 7, "ts": 1716454222886804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781578, "dur": 10,
    "args": {
      "External id": 97010,
      "cbid": 211, "correlation": 97010
    }
  },
  {
    "ph": "s", "id": 97010, "pid": 76337, "tid": -914061504, "ts": 1716454222781578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222886903, "dur": 90,
    "args": {
      "External id": 97030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97030,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 97030, "pid": 5, "tid": 7, "ts": 1716454222886903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781654, "dur": 11,
    "args": {
      "External id": 97030,
      "cbid": 211, "correlation": 97030
    }
  },
  {
    "ph": "s", "id": 97030, "pid": 76337, "tid": -914061504, "ts": 1716454222781654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222886995, "dur": 4,
    "args": {
      "External id": 97042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97042,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 97042, "pid": 5, "tid": 7, "ts": 1716454222886995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781675, "dur": 7,
    "args": {
      "External id": 97042,
      "cbid": 211, "correlation": 97042
    }
  },
  {
    "ph": "s", "id": 97042, "pid": 76337, "tid": -914061504, "ts": 1716454222781675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222887001, "dur": 107,
    "args": {
      "External id": 97045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97045, "pid": 5, "tid": 7, "ts": 1716454222887001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781695, "dur": 7,
    "args": {
      "External id": 97045,
      "cbid": 211, "correlation": 97045
    }
  },
  {
    "ph": "s", "id": 97045, "pid": 76337, "tid": -914061504, "ts": 1716454222781695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222887109, "dur": 69,
    "args": {
      "External id": 97054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97054,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97054, "pid": 5, "tid": 7, "ts": 1716454222887109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781735, "dur": 10,
    "args": {
      "External id": 97054,
      "cbid": 211, "correlation": 97054
    }
  },
  {
    "ph": "s", "id": 97054, "pid": 76337, "tid": -914061504, "ts": 1716454222781735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781787, "dur": 0,
    "args": {
      "External id": 97064,
      "cbid": 317, "correlation": 97064
    }
  },
  {
    "ph": "f", "id": 97064, "pid": 76337, "tid": -914061504, "ts": 1716454222781787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781788, "dur": 0,
    "args": {
      "External id": 97065,
      "cbid": 203, "correlation": 97065
    }
  },
  {
    "ph": "f", "id": 97065, "pid": 76337, "tid": -914061504, "ts": 1716454222781788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781788, "dur": 0,
    "args": {
      "External id": 97066,
      "cbid": 205, "correlation": 97066
    }
  },
  {
    "ph": "f", "id": 97066, "pid": 76337, "tid": -914061504, "ts": 1716454222781788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222887180, "dur": 74,
    "args": {
      "External id": 97070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97070,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97070, "pid": 5, "tid": 7, "ts": 1716454222887180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781805, "dur": 11,
    "args": {
      "External id": 97070,
      "cbid": 211, "correlation": 97070
    }
  },
  {
    "ph": "s", "id": 97070, "pid": 76337, "tid": -914061504, "ts": 1716454222781805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222887255, "dur": 23,
    "args": {
      "External id": 97072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97072,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97072, "pid": 5, "tid": 7, "ts": 1716454222887255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781819, "dur": 5,
    "args": {
      "External id": 97072,
      "cbid": 211, "correlation": 97072
    }
  },
  {
    "ph": "s", "id": 97072, "pid": 76337, "tid": -914061504, "ts": 1716454222781819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222887280, "dur": 4,
    "args": {
      "External id": 97074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97074, "pid": 5, "tid": 7, "ts": 1716454222887280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781830, "dur": 6,
    "args": {
      "External id": 97074,
      "cbid": 211, "correlation": 97074
    }
  },
  {
    "ph": "s", "id": 97074, "pid": 76337, "tid": -914061504, "ts": 1716454222781830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781839, "dur": 0,
    "args": {
      "External id": 97075,
      "cbid": 51, "correlation": 97075
    }
  },
  {
    "ph": "s", "id": 97075, "pid": 76337, "tid": -914061504, "ts": 1716454222781839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222887285, "dur": 1340,
    "args": {
      "External id": 97076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97076,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97076, "pid": 5, "tid": 7, "ts": 1716454222887285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781840, "dur": 5,
    "args": {
      "External id": 97076,
      "cbid": 211, "correlation": 97076
    }
  },
  {
    "ph": "s", "id": 97076, "pid": 76337, "tid": -914061504, "ts": 1716454222781840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222888626, "dur": 58,
    "args": {
      "External id": 97081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97081,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97081, "pid": 5, "tid": 7, "ts": 1716454222888626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781868, "dur": 8,
    "args": {
      "External id": 97081,
      "cbid": 211, "correlation": 97081
    }
  },
  {
    "ph": "s", "id": 97081, "pid": 76337, "tid": -914061504, "ts": 1716454222781868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222888685, "dur": 3,
    "args": {
      "External id": 97089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97089,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97089, "pid": 5, "tid": 7, "ts": 1716454222888685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781910, "dur": 10,
    "args": {
      "External id": 97089,
      "cbid": 211, "correlation": 97089
    }
  },
  {
    "ph": "s", "id": 97089, "pid": 76337, "tid": -914061504, "ts": 1716454222781910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781984, "dur": 1,
    "args": {
      "External id": 97105,
      "cbid": 251, "correlation": 97105
    }
  },
  {
    "ph": "f", "id": 97105, "pid": 76337, "tid": -914061504, "ts": 1716454222781984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781989, "dur": 0,
    "args": {
      "External id": 97107,
      "cbid": 251, "correlation": 97107
    }
  },
  {
    "ph": "f", "id": 97107, "pid": 76337, "tid": -914061504, "ts": 1716454222781989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222888690, "dur": 11,
    "args": {
      "External id": 97108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97108,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 97108, "pid": 5, "tid": 7, "ts": 1716454222888690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222781991, "dur": 12,
    "args": {
      "External id": 97108,
      "cbid": 211, "correlation": 97108
    }
  },
  {
    "ph": "s", "id": 97108, "pid": 76337, "tid": -914061504, "ts": 1716454222781991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222888702, "dur": 5,
    "args": {
      "External id": 97110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97110,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 97110, "pid": 5, "tid": 7, "ts": 1716454222888702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782006, "dur": 6,
    "args": {
      "External id": 97110,
      "cbid": 211, "correlation": 97110
    }
  },
  {
    "ph": "s", "id": 97110, "pid": 76337, "tid": -914061504, "ts": 1716454222782006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222888708, "dur": 54,
    "args": {
      "External id": 97120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97120, "pid": 5, "tid": 7, "ts": 1716454222888708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782065, "dur": 13,
    "args": {
      "External id": 97120,
      "cbid": 211, "correlation": 97120
    }
  },
  {
    "ph": "s", "id": 97120, "pid": 76337, "tid": -914061504, "ts": 1716454222782065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222888764, "dur": 52,
    "args": {
      "External id": 97140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97140,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 97140, "pid": 5, "tid": 7, "ts": 1716454222888764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782132, "dur": 12,
    "args": {
      "External id": 97140,
      "cbid": 211, "correlation": 97140
    }
  },
  {
    "ph": "s", "id": 97140, "pid": 76337, "tid": -914061504, "ts": 1716454222782132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222888817, "dur": 4,
    "args": {
      "External id": 97152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97152,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97152, "pid": 5, "tid": 7, "ts": 1716454222888817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782153, "dur": 6,
    "args": {
      "External id": 97152,
      "cbid": 211, "correlation": 97152
    }
  },
  {
    "ph": "s", "id": 97152, "pid": 76337, "tid": -914061504, "ts": 1716454222782153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222888822, "dur": 53,
    "args": {
      "External id": 97155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97155, "pid": 5, "tid": 7, "ts": 1716454222888822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782172, "dur": 6,
    "args": {
      "External id": 97155,
      "cbid": 211, "correlation": 97155
    }
  },
  {
    "ph": "s", "id": 97155, "pid": 76337, "tid": -914061504, "ts": 1716454222782172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222888877, "dur": 36,
    "args": {
      "External id": 97164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97164,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97164, "pid": 5, "tid": 7, "ts": 1716454222888877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782214, "dur": 10,
    "args": {
      "External id": 97164,
      "cbid": 211, "correlation": 97164
    }
  },
  {
    "ph": "s", "id": 97164, "pid": 76337, "tid": -914061504, "ts": 1716454222782214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782278, "dur": 0,
    "args": {
      "External id": 97174,
      "cbid": 317, "correlation": 97174
    }
  },
  {
    "ph": "f", "id": 97174, "pid": 76337, "tid": -914061504, "ts": 1716454222782278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782279, "dur": 0,
    "args": {
      "External id": 97175,
      "cbid": 203, "correlation": 97175
    }
  },
  {
    "ph": "f", "id": 97175, "pid": 76337, "tid": -914061504, "ts": 1716454222782279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782280, "dur": 0,
    "args": {
      "External id": 97176,
      "cbid": 205, "correlation": 97176
    }
  },
  {
    "ph": "f", "id": 97176, "pid": 76337, "tid": -914061504, "ts": 1716454222782280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222888915, "dur": 40,
    "args": {
      "External id": 97180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97180,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97180, "pid": 5, "tid": 7, "ts": 1716454222888915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782294, "dur": 12,
    "args": {
      "External id": 97180,
      "cbid": 211, "correlation": 97180
    }
  },
  {
    "ph": "s", "id": 97180, "pid": 76337, "tid": -914061504, "ts": 1716454222782294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222888956, "dur": 14,
    "args": {
      "External id": 97182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97182,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97182, "pid": 5, "tid": 7, "ts": 1716454222888956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782308, "dur": 6,
    "args": {
      "External id": 97182,
      "cbid": 211, "correlation": 97182
    }
  },
  {
    "ph": "s", "id": 97182, "pid": 76337, "tid": -914061504, "ts": 1716454222782308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222888971, "dur": 3,
    "args": {
      "External id": 97184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97184,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97184, "pid": 5, "tid": 7, "ts": 1716454222888971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782318, "dur": 5,
    "args": {
      "External id": 97184,
      "cbid": 211, "correlation": 97184
    }
  },
  {
    "ph": "s", "id": 97184, "pid": 76337, "tid": -914061504, "ts": 1716454222782318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782327, "dur": 0,
    "args": {
      "External id": 97185,
      "cbid": 51, "correlation": 97185
    }
  },
  {
    "ph": "s", "id": 97185, "pid": 76337, "tid": -914061504, "ts": 1716454222782327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222888976, "dur": 684,
    "args": {
      "External id": 97186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97186,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97186, "pid": 5, "tid": 7, "ts": 1716454222888976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782328, "dur": 5,
    "args": {
      "External id": 97186,
      "cbid": 211, "correlation": 97186
    }
  },
  {
    "ph": "s", "id": 97186, "pid": 76337, "tid": -914061504, "ts": 1716454222782328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222889662, "dur": 58,
    "args": {
      "External id": 97191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97191, "pid": 5, "tid": 7, "ts": 1716454222889662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782354, "dur": 8,
    "args": {
      "External id": 97191,
      "cbid": 211, "correlation": 97191
    }
  },
  {
    "ph": "s", "id": 97191, "pid": 76337, "tid": -914061504, "ts": 1716454222782354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782412, "dur": 0,
    "args": {
      "External id": 97201,
      "cbid": 317, "correlation": 97201
    }
  },
  {
    "ph": "f", "id": 97201, "pid": 76337, "tid": -914061504, "ts": 1716454222782412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782413, "dur": 0,
    "args": {
      "External id": 97202,
      "cbid": 203, "correlation": 97202
    }
  },
  {
    "ph": "f", "id": 97202, "pid": 76337, "tid": -914061504, "ts": 1716454222782413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782413, "dur": 0,
    "args": {
      "External id": 97203,
      "cbid": 205, "correlation": 97203
    }
  },
  {
    "ph": "f", "id": 97203, "pid": 76337, "tid": -914061504, "ts": 1716454222782413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222889721, "dur": 75,
    "args": {
      "External id": 97207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97207,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97207, "pid": 5, "tid": 7, "ts": 1716454222889721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782426, "dur": 11,
    "args": {
      "External id": 97207,
      "cbid": 211, "correlation": 97207
    }
  },
  {
    "ph": "s", "id": 97207, "pid": 76337, "tid": -914061504, "ts": 1716454222782426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222889798, "dur": 205,
    "args": {
      "External id": 97209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97209,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97209, "pid": 5, "tid": 7, "ts": 1716454222889798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782447, "dur": 8,
    "args": {
      "External id": 97209,
      "cbid": 211, "correlation": 97209
    }
  },
  {
    "ph": "s", "id": 97209, "pid": 76337, "tid": -914061504, "ts": 1716454222782447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222890004, "dur": 38,
    "args": {
      "External id": 97211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97211,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97211, "pid": 5, "tid": 7, "ts": 1716454222890004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782460, "dur": 6,
    "args": {
      "External id": 97211,
      "cbid": 211, "correlation": 97211
    }
  },
  {
    "ph": "s", "id": 97211, "pid": 76337, "tid": -914061504, "ts": 1716454222782460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222890043, "dur": 58,
    "args": {
      "External id": 97217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97217,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97217, "pid": 5, "tid": 7, "ts": 1716454222890043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782487, "dur": 8,
    "args": {
      "External id": 97217,
      "cbid": 211, "correlation": 97217
    }
  },
  {
    "ph": "s", "id": 97217, "pid": 76337, "tid": -914061504, "ts": 1716454222782487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222890102, "dur": 50,
    "args": {
      "External id": 97225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97225,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97225, "pid": 5, "tid": 7, "ts": 1716454222890102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782516, "dur": 8,
    "args": {
      "External id": 97225,
      "cbid": 211, "correlation": 97225
    }
  },
  {
    "ph": "s", "id": 97225, "pid": 76337, "tid": -914061504, "ts": 1716454222782516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222890154, "dur": 35,
    "args": {
      "External id": 97233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97233,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97233, "pid": 5, "tid": 7, "ts": 1716454222890154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782545, "dur": 9,
    "args": {
      "External id": 97233,
      "cbid": 211, "correlation": 97233
    }
  },
  {
    "ph": "s", "id": 97233, "pid": 76337, "tid": -914061504, "ts": 1716454222782545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222890190, "dur": 51,
    "args": {
      "External id": 97253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97253,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 97253, "pid": 5, "tid": 7, "ts": 1716454222890190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782627, "dur": 12,
    "args": {
      "External id": 97253,
      "cbid": 211, "correlation": 97253
    }
  },
  {
    "ph": "s", "id": 97253, "pid": 76337, "tid": -914061504, "ts": 1716454222782627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222890243, "dur": 4,
    "args": {
      "External id": 97265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97265,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97265, "pid": 5, "tid": 7, "ts": 1716454222890243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782649, "dur": 7,
    "args": {
      "External id": 97265,
      "cbid": 211, "correlation": 97265
    }
  },
  {
    "ph": "s", "id": 97265, "pid": 76337, "tid": -914061504, "ts": 1716454222782649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222890248, "dur": 54,
    "args": {
      "External id": 97268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97268,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97268, "pid": 5, "tid": 7, "ts": 1716454222890248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782667, "dur": 6,
    "args": {
      "External id": 97268,
      "cbid": 211, "correlation": 97268
    }
  },
  {
    "ph": "s", "id": 97268, "pid": 76337, "tid": -914061504, "ts": 1716454222782667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782724, "dur": 0,
    "args": {
      "External id": 97279,
      "cbid": 317, "correlation": 97279
    }
  },
  {
    "ph": "f", "id": 97279, "pid": 76337, "tid": -914061504, "ts": 1716454222782724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782725, "dur": 0,
    "args": {
      "External id": 97280,
      "cbid": 203, "correlation": 97280
    }
  },
  {
    "ph": "f", "id": 97280, "pid": 76337, "tid": -914061504, "ts": 1716454222782725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782725, "dur": 0,
    "args": {
      "External id": 97281,
      "cbid": 205, "correlation": 97281
    }
  },
  {
    "ph": "f", "id": 97281, "pid": 76337, "tid": -914061504, "ts": 1716454222782725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782749, "dur": 1,
    "args": {
      "External id": 97285,
      "cbid": 251, "correlation": 97285
    }
  },
  {
    "ph": "f", "id": 97285, "pid": 76337, "tid": -914061504, "ts": 1716454222782749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782751, "dur": 0,
    "args": {
      "External id": 97286,
      "cbid": 251, "correlation": 97286
    }
  },
  {
    "ph": "f", "id": 97286, "pid": 76337, "tid": -914061504, "ts": 1716454222782751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782752, "dur": 0,
    "args": {
      "External id": 97287,
      "cbid": 251, "correlation": 97287
    }
  },
  {
    "ph": "f", "id": 97287, "pid": 76337, "tid": -914061504, "ts": 1716454222782752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782753, "dur": 0,
    "args": {
      "External id": 97288,
      "cbid": 251, "correlation": 97288
    }
  },
  {
    "ph": "f", "id": 97288, "pid": 76337, "tid": -914061504, "ts": 1716454222782753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782753, "dur": 0,
    "args": {
      "External id": 97289,
      "cbid": 251, "correlation": 97289
    }
  },
  {
    "ph": "f", "id": 97289, "pid": 76337, "tid": -914061504, "ts": 1716454222782753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782754, "dur": 0,
    "args": {
      "External id": 97290,
      "cbid": 251, "correlation": 97290
    }
  },
  {
    "ph": "f", "id": 97290, "pid": 76337, "tid": -914061504, "ts": 1716454222782754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782755, "dur": 0,
    "args": {
      "External id": 97291,
      "cbid": 251, "correlation": 97291
    }
  },
  {
    "ph": "f", "id": 97291, "pid": 76337, "tid": -914061504, "ts": 1716454222782755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782756, "dur": 0,
    "args": {
      "External id": 97292,
      "cbid": 251, "correlation": 97292
    }
  },
  {
    "ph": "f", "id": 97292, "pid": 76337, "tid": -914061504, "ts": 1716454222782756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782758, "dur": 0,
    "args": {
      "External id": 97293,
      "cbid": 251, "correlation": 97293
    }
  },
  {
    "ph": "f", "id": 97293, "pid": 76337, "tid": -914061504, "ts": 1716454222782758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222890303, "dur": 111,
    "args": {
      "External id": 97294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97294,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 97294, "pid": 5, "tid": 7, "ts": 1716454222890303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782760, "dur": 13,
    "args": {
      "External id": 97294,
      "cbid": 211, "correlation": 97294
    }
  },
  {
    "ph": "s", "id": 97294, "pid": 76337, "tid": -914061504, "ts": 1716454222782760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222890416, "dur": 59,
    "args": {
      "External id": 97300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97300,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97300, "pid": 5, "tid": 7, "ts": 1716454222890416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782795, "dur": 9,
    "args": {
      "External id": 97300,
      "cbid": 211, "correlation": 97300
    }
  },
  {
    "ph": "s", "id": 97300, "pid": 76337, "tid": -914061504, "ts": 1716454222782795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222890476, "dur": 566,
    "args": {
      "External id": 97309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97309,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97309, "pid": 5, "tid": 7, "ts": 1716454222890476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782879, "dur": 15,
    "args": {
      "External id": 97309,
      "cbid": 211, "correlation": 97309
    }
  },
  {
    "ph": "s", "id": 97309, "pid": 76337, "tid": -914061504, "ts": 1716454222782879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222891044, "dur": 177,
    "args": {
      "External id": 97331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97331,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97331, "pid": 5, "tid": 7, "ts": 1716454222891044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222782940, "dur": 10,
    "args": {
      "External id": 97331,
      "cbid": 211, "correlation": 97331
    }
  },
  {
    "ph": "s", "id": 97331, "pid": 76337, "tid": -914061504, "ts": 1716454222782940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783038, "dur": 1,
    "args": {
      "External id": 97342,
      "cbid": 251, "correlation": 97342
    }
  },
  {
    "ph": "f", "id": 97342, "pid": 76337, "tid": -914061504, "ts": 1716454222783038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222891222, "dur": 193,
    "args": {
      "External id": 97343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97343,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97343, "pid": 5, "tid": 7, "ts": 1716454222891222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783043, "dur": 14,
    "args": {
      "External id": 97343,
      "cbid": 211, "correlation": 97343
    }
  },
  {
    "ph": "s", "id": 97343, "pid": 76337, "tid": -914061504, "ts": 1716454222783043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783112, "dur": 1,
    "args": {
      "External id": 97354,
      "cbid": 251, "correlation": 97354
    }
  },
  {
    "ph": "f", "id": 97354, "pid": 76337, "tid": -914061504, "ts": 1716454222783112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222891417, "dur": 187,
    "args": {
      "External id": 97355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97355,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97355, "pid": 5, "tid": 7, "ts": 1716454222891417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783116, "dur": 11,
    "args": {
      "External id": 97355,
      "cbid": 211, "correlation": 97355
    }
  },
  {
    "ph": "s", "id": 97355, "pid": 76337, "tid": -914061504, "ts": 1716454222783116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783178, "dur": 1,
    "args": {
      "External id": 97366,
      "cbid": 251, "correlation": 97366
    }
  },
  {
    "ph": "f", "id": 97366, "pid": 76337, "tid": -914061504, "ts": 1716454222783178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222891605, "dur": 187,
    "args": {
      "External id": 97367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97367,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97367, "pid": 5, "tid": 7, "ts": 1716454222891605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783182, "dur": 11,
    "args": {
      "External id": 97367,
      "cbid": 211, "correlation": 97367
    }
  },
  {
    "ph": "s", "id": 97367, "pid": 76337, "tid": -914061504, "ts": 1716454222783182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222891794, "dur": 18153,
    "args": {
      "External id": 97388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97388,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 97388, "pid": 5, "tid": 7, "ts": 1716454222891794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783269, "dur": 13,
    "args": {
      "External id": 97388,
      "cbid": 211, "correlation": 97388
    }
  },
  {
    "ph": "s", "id": 97388, "pid": 76337, "tid": -914061504, "ts": 1716454222783269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783369, "dur": 1,
    "args": {
      "External id": 97406,
      "cbid": 251, "correlation": 97406
    }
  },
  {
    "ph": "f", "id": 97406, "pid": 76337, "tid": -914061504, "ts": 1716454222783369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222909948, "dur": 202,
    "args": {
      "External id": 97408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97408,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97408, "pid": 5, "tid": 7, "ts": 1716454222909948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783375, "dur": 14,
    "args": {
      "External id": 97408,
      "cbid": 211, "correlation": 97408
    }
  },
  {
    "ph": "s", "id": 97408, "pid": 76337, "tid": -914061504, "ts": 1716454222783375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222910152, "dur": 66,
    "args": {
      "External id": 97416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97416,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97416, "pid": 5, "tid": 7, "ts": 1716454222910152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783446, "dur": 12,
    "args": {
      "External id": 97416,
      "cbid": 211, "correlation": 97416
    }
  },
  {
    "ph": "s", "id": 97416, "pid": 76337, "tid": -914061504, "ts": 1716454222783446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222910219, "dur": 97,
    "args": {
      "External id": 97424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97424,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97424, "pid": 5, "tid": 7, "ts": 1716454222910219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783484, "dur": 9,
    "args": {
      "External id": 97424,
      "cbid": 211, "correlation": 97424
    }
  },
  {
    "ph": "s", "id": 97424, "pid": 76337, "tid": -914061504, "ts": 1716454222783484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222910317, "dur": 55,
    "args": {
      "External id": 97435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97435, "pid": 5, "tid": 7, "ts": 1716454222910317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783560, "dur": 13,
    "args": {
      "External id": 97435,
      "cbid": 211, "correlation": 97435
    }
  },
  {
    "ph": "s", "id": 97435, "pid": 76337, "tid": -914061504, "ts": 1716454222783560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222910374, "dur": 90,
    "args": {
      "External id": 97457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97457,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97457, "pid": 5, "tid": 7, "ts": 1716454222910374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222783592, "dur": 1720,
    "args": {
      "External id": 97457,
      "cbid": 211, "correlation": 97457
    }
  },
  {
    "ph": "s", "id": 97457, "pid": 76337, "tid": -914061504, "ts": 1716454222783592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785392, "dur": 1,
    "args": {
      "External id": 97468,
      "cbid": 251, "correlation": 97468
    }
  },
  {
    "ph": "f", "id": 97468, "pid": 76337, "tid": -914061504, "ts": 1716454222785392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222910465, "dur": 101,
    "args": {
      "External id": 97469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97469,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97469, "pid": 5, "tid": 7, "ts": 1716454222910465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785397, "dur": 61,
    "args": {
      "External id": 97469,
      "cbid": 211, "correlation": 97469
    }
  },
  {
    "ph": "s", "id": 97469, "pid": 76337, "tid": -914061504, "ts": 1716454222785397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785518, "dur": 1,
    "args": {
      "External id": 97480,
      "cbid": 251, "correlation": 97480
    }
  },
  {
    "ph": "f", "id": 97480, "pid": 76337, "tid": -914061504, "ts": 1716454222785518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785523, "dur": 0,
    "args": {
      "External id": 97481,
      "cbid": 251, "correlation": 97481
    }
  },
  {
    "ph": "f", "id": 97481, "pid": 76337, "tid": -914061504, "ts": 1716454222785523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222910567, "dur": 10,
    "args": {
      "External id": 97482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97482,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 97482, "pid": 5, "tid": 7, "ts": 1716454222910567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785524, "dur": 12,
    "args": {
      "External id": 97482,
      "cbid": 211, "correlation": 97482
    }
  },
  {
    "ph": "s", "id": 97482, "pid": 76337, "tid": -914061504, "ts": 1716454222785524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222910578, "dur": 5,
    "args": {
      "External id": 97484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97484,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 97484, "pid": 5, "tid": 7, "ts": 1716454222910578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785538, "dur": 6,
    "args": {
      "External id": 97484,
      "cbid": 211, "correlation": 97484
    }
  },
  {
    "ph": "s", "id": 97484, "pid": 76337, "tid": -914061504, "ts": 1716454222785538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785599, "dur": 1,
    "args": {
      "External id": 97495,
      "cbid": 251, "correlation": 97495
    }
  },
  {
    "ph": "f", "id": 97495, "pid": 76337, "tid": -914061504, "ts": 1716454222785599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785602, "dur": 0,
    "args": {
      "External id": 97496,
      "cbid": 251, "correlation": 97496
    }
  },
  {
    "ph": "f", "id": 97496, "pid": 76337, "tid": -914061504, "ts": 1716454222785602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222910584, "dur": 6,
    "args": {
      "External id": 97497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97497,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 97497, "pid": 5, "tid": 7, "ts": 1716454222910584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785604, "dur": 11,
    "args": {
      "External id": 97497,
      "cbid": 211, "correlation": 97497
    }
  },
  {
    "ph": "s", "id": 97497, "pid": 76337, "tid": -914061504, "ts": 1716454222785604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222910591, "dur": 3,
    "args": {
      "External id": 97499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97499,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 97499, "pid": 5, "tid": 7, "ts": 1716454222910591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785617, "dur": 5,
    "args": {
      "External id": 97499,
      "cbid": 211, "correlation": 97499
    }
  },
  {
    "ph": "s", "id": 97499, "pid": 76337, "tid": -914061504, "ts": 1716454222785617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222910596, "dur": 154,
    "args": {
      "External id": 97520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97520,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 97520, "pid": 5, "tid": 7, "ts": 1716454222910596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785692, "dur": 12,
    "args": {
      "External id": 97520,
      "cbid": 211, "correlation": 97520
    }
  },
  {
    "ph": "s", "id": 97520, "pid": 76337, "tid": -914061504, "ts": 1716454222785692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785790, "dur": 1,
    "args": {
      "External id": 97538,
      "cbid": 251, "correlation": 97538
    }
  },
  {
    "ph": "f", "id": 97538, "pid": 76337, "tid": -914061504, "ts": 1716454222785790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222910751, "dur": 105,
    "args": {
      "External id": 97540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97540,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 97540, "pid": 5, "tid": 7, "ts": 1716454222910751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785797, "dur": 14,
    "args": {
      "External id": 97540,
      "cbid": 211, "correlation": 97540
    }
  },
  {
    "ph": "s", "id": 97540, "pid": 76337, "tid": -914061504, "ts": 1716454222785797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222910858, "dur": 34,
    "args": {
      "External id": 97548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97548,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97548, "pid": 5, "tid": 7, "ts": 1716454222910858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785868, "dur": 12,
    "args": {
      "External id": 97548,
      "cbid": 211, "correlation": 97548
    }
  },
  {
    "ph": "s", "id": 97548, "pid": 76337, "tid": -914061504, "ts": 1716454222785868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222910893, "dur": 67,
    "args": {
      "External id": 97556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97556, "pid": 5, "tid": 7, "ts": 1716454222910893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785907, "dur": 10,
    "args": {
      "External id": 97556,
      "cbid": 211, "correlation": 97556
    }
  },
  {
    "ph": "s", "id": 97556, "pid": 76337, "tid": -914061504, "ts": 1716454222785907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222910961, "dur": 90,
    "args": {
      "External id": 97578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97578,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97578, "pid": 5, "tid": 7, "ts": 1716454222910961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222785958, "dur": 10,
    "args": {
      "External id": 97578,
      "cbid": 211, "correlation": 97578
    }
  },
  {
    "ph": "s", "id": 97578, "pid": 76337, "tid": -914061504, "ts": 1716454222785958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786057, "dur": 1,
    "args": {
      "External id": 97594,
      "cbid": 251, "correlation": 97594
    }
  },
  {
    "ph": "f", "id": 97594, "pid": 76337, "tid": -914061504, "ts": 1716454222786057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222911052, "dur": 564,
    "args": {
      "External id": 97596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97596,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97596, "pid": 5, "tid": 7, "ts": 1716454222911052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786063, "dur": 13,
    "args": {
      "External id": 97596,
      "cbid": 211, "correlation": 97596
    }
  },
  {
    "ph": "s", "id": 97596, "pid": 76337, "tid": -914061504, "ts": 1716454222786063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222911618, "dur": 242,
    "args": {
      "External id": 97604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97604, "pid": 5, "tid": 7, "ts": 1716454222911618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786133, "dur": 13,
    "args": {
      "External id": 97604,
      "cbid": 211, "correlation": 97604
    }
  },
  {
    "ph": "s", "id": 97604, "pid": 76337, "tid": -914061504, "ts": 1716454222786133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222911861, "dur": 250,
    "args": {
      "External id": 97612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97612, "pid": 5, "tid": 7, "ts": 1716454222911861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786166, "dur": 8,
    "args": {
      "External id": 97612,
      "cbid": 211, "correlation": 97612
    }
  },
  {
    "ph": "s", "id": 97612, "pid": 76337, "tid": -914061504, "ts": 1716454222786166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786249, "dur": 2,
    "args": {
      "External id": 97628,
      "cbid": 251, "correlation": 97628
    }
  },
  {
    "ph": "f", "id": 97628, "pid": 76337, "tid": -914061504, "ts": 1716454222786249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786254, "dur": 0,
    "args": {
      "External id": 97630,
      "cbid": 251, "correlation": 97630
    }
  },
  {
    "ph": "f", "id": 97630, "pid": 76337, "tid": -914061504, "ts": 1716454222786254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222912112, "dur": 355,
    "args": {
      "External id": 97631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97631,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 97631, "pid": 5, "tid": 7, "ts": 1716454222912112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786259, "dur": 14,
    "args": {
      "External id": 97631,
      "cbid": 211, "correlation": 97631
    }
  },
  {
    "ph": "s", "id": 97631, "pid": 76337, "tid": -914061504, "ts": 1716454222786259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222912469, "dur": 50,
    "args": {
      "External id": 97639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97639,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97639, "pid": 5, "tid": 7, "ts": 1716454222912469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786302, "dur": 10,
    "args": {
      "External id": 97639,
      "cbid": 211, "correlation": 97639
    }
  },
  {
    "ph": "s", "id": 97639, "pid": 76337, "tid": -914061504, "ts": 1716454222786302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222912520, "dur": 155,
    "args": {
      "External id": 97650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97650, "pid": 5, "tid": 7, "ts": 1716454222912520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786371, "dur": 165,
    "args": {
      "External id": 97650,
      "cbid": 211, "correlation": 97650
    }
  },
  {
    "ph": "s", "id": 97650, "pid": 76337, "tid": -914061504, "ts": 1716454222786371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786593, "dur": 0,
    "args": {
      "External id": 97662,
      "cbid": 317, "correlation": 97662
    }
  },
  {
    "ph": "f", "id": 97662, "pid": 76337, "tid": -914061504, "ts": 1716454222786593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786593, "dur": 0,
    "args": {
      "External id": 97663,
      "cbid": 203, "correlation": 97663
    }
  },
  {
    "ph": "f", "id": 97663, "pid": 76337, "tid": -914061504, "ts": 1716454222786593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786594, "dur": 0,
    "args": {
      "External id": 97664,
      "cbid": 205, "correlation": 97664
    }
  },
  {
    "ph": "f", "id": 97664, "pid": 76337, "tid": -914061504, "ts": 1716454222786594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786620, "dur": 1,
    "args": {
      "External id": 97668,
      "cbid": 251, "correlation": 97668
    }
  },
  {
    "ph": "f", "id": 97668, "pid": 76337, "tid": -914061504, "ts": 1716454222786620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786622, "dur": 0,
    "args": {
      "External id": 97669,
      "cbid": 251, "correlation": 97669
    }
  },
  {
    "ph": "f", "id": 97669, "pid": 76337, "tid": -914061504, "ts": 1716454222786622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786623, "dur": 0,
    "args": {
      "External id": 97670,
      "cbid": 251, "correlation": 97670
    }
  },
  {
    "ph": "f", "id": 97670, "pid": 76337, "tid": -914061504, "ts": 1716454222786623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786624, "dur": 0,
    "args": {
      "External id": 97671,
      "cbid": 251, "correlation": 97671
    }
  },
  {
    "ph": "f", "id": 97671, "pid": 76337, "tid": -914061504, "ts": 1716454222786624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786625, "dur": 1,
    "args": {
      "External id": 97672,
      "cbid": 251, "correlation": 97672
    }
  },
  {
    "ph": "f", "id": 97672, "pid": 76337, "tid": -914061504, "ts": 1716454222786625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786626, "dur": 0,
    "args": {
      "External id": 97673,
      "cbid": 251, "correlation": 97673
    }
  },
  {
    "ph": "f", "id": 97673, "pid": 76337, "tid": -914061504, "ts": 1716454222786626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786627, "dur": 0,
    "args": {
      "External id": 97674,
      "cbid": 251, "correlation": 97674
    }
  },
  {
    "ph": "f", "id": 97674, "pid": 76337, "tid": -914061504, "ts": 1716454222786627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786628, "dur": 0,
    "args": {
      "External id": 97675,
      "cbid": 251, "correlation": 97675
    }
  },
  {
    "ph": "f", "id": 97675, "pid": 76337, "tid": -914061504, "ts": 1716454222786628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786629, "dur": 0,
    "args": {
      "External id": 97676,
      "cbid": 251, "correlation": 97676
    }
  },
  {
    "ph": "f", "id": 97676, "pid": 76337, "tid": -914061504, "ts": 1716454222786629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222912676, "dur": 113,
    "args": {
      "External id": 97677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97677,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 97677, "pid": 5, "tid": 7, "ts": 1716454222912676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786631, "dur": 29,
    "args": {
      "External id": 97677,
      "cbid": 211, "correlation": 97677
    }
  },
  {
    "ph": "s", "id": 97677, "pid": 76337, "tid": -914061504, "ts": 1716454222786631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222912790, "dur": 59,
    "args": {
      "External id": 97683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97683, "pid": 5, "tid": 7, "ts": 1716454222912790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786683, "dur": 105,
    "args": {
      "External id": 97683,
      "cbid": 211, "correlation": 97683
    }
  },
  {
    "ph": "s", "id": 97683, "pid": 76337, "tid": -914061504, "ts": 1716454222786683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222912850, "dur": 50,
    "args": {
      "External id": 97691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97691,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97691, "pid": 5, "tid": 7, "ts": 1716454222912850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222786811, "dur": 289,
    "args": {
      "External id": 97691,
      "cbid": 211, "correlation": 97691
    }
  },
  {
    "ph": "s", "id": 97691, "pid": 76337, "tid": -914061504, "ts": 1716454222786811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222912902, "dur": 97,
    "args": {
      "External id": 97700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97700,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97700, "pid": 5, "tid": 7, "ts": 1716454222912902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787135, "dur": 11,
    "args": {
      "External id": 97700,
      "cbid": 211, "correlation": 97700
    }
  },
  {
    "ph": "s", "id": 97700, "pid": 76337, "tid": -914061504, "ts": 1716454222787135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222913000, "dur": 90,
    "args": {
      "External id": 97720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97720,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 97720, "pid": 5, "tid": 7, "ts": 1716454222913000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787214, "dur": 11,
    "args": {
      "External id": 97720,
      "cbid": 211, "correlation": 97720
    }
  },
  {
    "ph": "s", "id": 97720, "pid": 76337, "tid": -914061504, "ts": 1716454222787214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222913092, "dur": 4,
    "args": {
      "External id": 97732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97732,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 97732, "pid": 5, "tid": 7, "ts": 1716454222913092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787235, "dur": 7,
    "args": {
      "External id": 97732,
      "cbid": 211, "correlation": 97732
    }
  },
  {
    "ph": "s", "id": 97732, "pid": 76337, "tid": -914061504, "ts": 1716454222787235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222913097, "dur": 108,
    "args": {
      "External id": 97735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97735, "pid": 5, "tid": 7, "ts": 1716454222913097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787254, "dur": 97,
    "args": {
      "External id": 97735,
      "cbid": 211, "correlation": 97735
    }
  },
  {
    "ph": "s", "id": 97735, "pid": 76337, "tid": -914061504, "ts": 1716454222787254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222913206, "dur": 68,
    "args": {
      "External id": 97744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97744,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97744, "pid": 5, "tid": 7, "ts": 1716454222913206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787386, "dur": 10,
    "args": {
      "External id": 97744,
      "cbid": 211, "correlation": 97744
    }
  },
  {
    "ph": "s", "id": 97744, "pid": 76337, "tid": -914061504, "ts": 1716454222787386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787440, "dur": 0,
    "args": {
      "External id": 97754,
      "cbid": 317, "correlation": 97754
    }
  },
  {
    "ph": "f", "id": 97754, "pid": 76337, "tid": -914061504, "ts": 1716454222787440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787440, "dur": 0,
    "args": {
      "External id": 97755,
      "cbid": 203, "correlation": 97755
    }
  },
  {
    "ph": "f", "id": 97755, "pid": 76337, "tid": -914061504, "ts": 1716454222787440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787441, "dur": 0,
    "args": {
      "External id": 97756,
      "cbid": 205, "correlation": 97756
    }
  },
  {
    "ph": "f", "id": 97756, "pid": 76337, "tid": -914061504, "ts": 1716454222787441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222913276, "dur": 76,
    "args": {
      "External id": 97760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97760,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97760, "pid": 5, "tid": 7, "ts": 1716454222913276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787458, "dur": 13,
    "args": {
      "External id": 97760,
      "cbid": 211, "correlation": 97760
    }
  },
  {
    "ph": "s", "id": 97760, "pid": 76337, "tid": -914061504, "ts": 1716454222787458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222913353, "dur": 23,
    "args": {
      "External id": 97762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97762,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97762, "pid": 5, "tid": 7, "ts": 1716454222913353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787473, "dur": 5,
    "args": {
      "External id": 97762,
      "cbid": 211, "correlation": 97762
    }
  },
  {
    "ph": "s", "id": 97762, "pid": 76337, "tid": -914061504, "ts": 1716454222787473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222913378, "dur": 4,
    "args": {
      "External id": 97764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97764,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97764, "pid": 5, "tid": 7, "ts": 1716454222913378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787485, "dur": 6,
    "args": {
      "External id": 97764,
      "cbid": 211, "correlation": 97764
    }
  },
  {
    "ph": "s", "id": 97764, "pid": 76337, "tid": -914061504, "ts": 1716454222787485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787494, "dur": 0,
    "args": {
      "External id": 97765,
      "cbid": 51, "correlation": 97765
    }
  },
  {
    "ph": "s", "id": 97765, "pid": 76337, "tid": -914061504, "ts": 1716454222787494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222913383, "dur": 1339,
    "args": {
      "External id": 97766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97766,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97766, "pid": 5, "tid": 7, "ts": 1716454222913383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787495, "dur": 6,
    "args": {
      "External id": 97766,
      "cbid": 211, "correlation": 97766
    }
  },
  {
    "ph": "s", "id": 97766, "pid": 76337, "tid": -914061504, "ts": 1716454222787495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222914723, "dur": 58,
    "args": {
      "External id": 97771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97771, "pid": 5, "tid": 7, "ts": 1716454222914723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787524, "dur": 9,
    "args": {
      "External id": 97771,
      "cbid": 211, "correlation": 97771
    }
  },
  {
    "ph": "s", "id": 97771, "pid": 76337, "tid": -914061504, "ts": 1716454222787524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222914782, "dur": 3,
    "args": {
      "External id": 97779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97779,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97779, "pid": 5, "tid": 7, "ts": 1716454222914782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787567, "dur": 10,
    "args": {
      "External id": 97779,
      "cbid": 211, "correlation": 97779
    }
  },
  {
    "ph": "s", "id": 97779, "pid": 76337, "tid": -914061504, "ts": 1716454222787567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787637, "dur": 2,
    "args": {
      "External id": 97795,
      "cbid": 251, "correlation": 97795
    }
  },
  {
    "ph": "f", "id": 97795, "pid": 76337, "tid": -914061504, "ts": 1716454222787637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787643, "dur": 0,
    "args": {
      "External id": 97797,
      "cbid": 251, "correlation": 97797
    }
  },
  {
    "ph": "f", "id": 97797, "pid": 76337, "tid": -914061504, "ts": 1716454222787643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222914787, "dur": 11,
    "args": {
      "External id": 97798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97798,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 97798, "pid": 5, "tid": 7, "ts": 1716454222914787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787645, "dur": 12,
    "args": {
      "External id": 97798,
      "cbid": 211, "correlation": 97798
    }
  },
  {
    "ph": "s", "id": 97798, "pid": 76337, "tid": -914061504, "ts": 1716454222787645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222914800, "dur": 5,
    "args": {
      "External id": 97800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97800,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 97800, "pid": 5, "tid": 7, "ts": 1716454222914800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787659, "dur": 5,
    "args": {
      "External id": 97800,
      "cbid": 211, "correlation": 97800
    }
  },
  {
    "ph": "s", "id": 97800, "pid": 76337, "tid": -914061504, "ts": 1716454222787659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222914806, "dur": 54,
    "args": {
      "External id": 97810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97810, "pid": 5, "tid": 7, "ts": 1716454222914806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222787720, "dur": 526,
    "args": {
      "External id": 97810,
      "cbid": 211, "correlation": 97810
    }
  },
  {
    "ph": "s", "id": 97810, "pid": 76337, "tid": -914061504, "ts": 1716454222787720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222914861, "dur": 51,
    "args": {
      "External id": 97830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97830,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 97830, "pid": 5, "tid": 7, "ts": 1716454222914861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788303, "dur": 12,
    "args": {
      "External id": 97830,
      "cbid": 211, "correlation": 97830
    }
  },
  {
    "ph": "s", "id": 97830, "pid": 76337, "tid": -914061504, "ts": 1716454222788303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222914914, "dur": 4,
    "args": {
      "External id": 97842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97842,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97842, "pid": 5, "tid": 7, "ts": 1716454222914914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788325, "dur": 6,
    "args": {
      "External id": 97842,
      "cbid": 211, "correlation": 97842
    }
  },
  {
    "ph": "s", "id": 97842, "pid": 76337, "tid": -914061504, "ts": 1716454222788325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222914919, "dur": 54,
    "args": {
      "External id": 97845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97845, "pid": 5, "tid": 7, "ts": 1716454222914919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788345, "dur": 7,
    "args": {
      "External id": 97845,
      "cbid": 211, "correlation": 97845
    }
  },
  {
    "ph": "s", "id": 97845, "pid": 76337, "tid": -914061504, "ts": 1716454222788345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222914974, "dur": 37,
    "args": {
      "External id": 97854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97854,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97854, "pid": 5, "tid": 7, "ts": 1716454222914974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788386, "dur": 10,
    "args": {
      "External id": 97854,
      "cbid": 211, "correlation": 97854
    }
  },
  {
    "ph": "s", "id": 97854, "pid": 76337, "tid": -914061504, "ts": 1716454222788386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788452, "dur": 0,
    "args": {
      "External id": 97864,
      "cbid": 317, "correlation": 97864
    }
  },
  {
    "ph": "f", "id": 97864, "pid": 76337, "tid": -914061504, "ts": 1716454222788452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788453, "dur": 0,
    "args": {
      "External id": 97865,
      "cbid": 203, "correlation": 97865
    }
  },
  {
    "ph": "f", "id": 97865, "pid": 76337, "tid": -914061504, "ts": 1716454222788453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788453, "dur": 0,
    "args": {
      "External id": 97866,
      "cbid": 205, "correlation": 97866
    }
  },
  {
    "ph": "f", "id": 97866, "pid": 76337, "tid": -914061504, "ts": 1716454222788453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222915012, "dur": 40,
    "args": {
      "External id": 97870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97870,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97870, "pid": 5, "tid": 7, "ts": 1716454222915012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788469, "dur": 12,
    "args": {
      "External id": 97870,
      "cbid": 211, "correlation": 97870
    }
  },
  {
    "ph": "s", "id": 97870, "pid": 76337, "tid": -914061504, "ts": 1716454222788469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222915053, "dur": 14,
    "args": {
      "External id": 97872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97872,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97872, "pid": 5, "tid": 7, "ts": 1716454222915053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788483, "dur": 5,
    "args": {
      "External id": 97872,
      "cbid": 211, "correlation": 97872
    }
  },
  {
    "ph": "s", "id": 97872, "pid": 76337, "tid": -914061504, "ts": 1716454222788483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222915068, "dur": 3,
    "args": {
      "External id": 97874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97874,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97874, "pid": 5, "tid": 7, "ts": 1716454222915068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788493, "dur": 6,
    "args": {
      "External id": 97874,
      "cbid": 211, "correlation": 97874
    }
  },
  {
    "ph": "s", "id": 97874, "pid": 76337, "tid": -914061504, "ts": 1716454222788493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788501, "dur": 0,
    "args": {
      "External id": 97875,
      "cbid": 51, "correlation": 97875
    }
  },
  {
    "ph": "s", "id": 97875, "pid": 76337, "tid": -914061504, "ts": 1716454222788501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222915073, "dur": 683,
    "args": {
      "External id": 97876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97876,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97876, "pid": 5, "tid": 7, "ts": 1716454222915073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788502, "dur": 5,
    "args": {
      "External id": 97876,
      "cbid": 211, "correlation": 97876
    }
  },
  {
    "ph": "s", "id": 97876, "pid": 76337, "tid": -914061504, "ts": 1716454222788502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222915758, "dur": 59,
    "args": {
      "External id": 97881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97881,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97881, "pid": 5, "tid": 7, "ts": 1716454222915758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788530, "dur": 8,
    "args": {
      "External id": 97881,
      "cbid": 211, "correlation": 97881
    }
  },
  {
    "ph": "s", "id": 97881, "pid": 76337, "tid": -914061504, "ts": 1716454222788530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788587, "dur": 0,
    "args": {
      "External id": 97891,
      "cbid": 317, "correlation": 97891
    }
  },
  {
    "ph": "f", "id": 97891, "pid": 76337, "tid": -914061504, "ts": 1716454222788587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788588, "dur": 0,
    "args": {
      "External id": 97892,
      "cbid": 203, "correlation": 97892
    }
  },
  {
    "ph": "f", "id": 97892, "pid": 76337, "tid": -914061504, "ts": 1716454222788588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788588, "dur": 0,
    "args": {
      "External id": 97893,
      "cbid": 205, "correlation": 97893
    }
  },
  {
    "ph": "f", "id": 97893, "pid": 76337, "tid": -914061504, "ts": 1716454222788588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222915817, "dur": 75,
    "args": {
      "External id": 97897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97897,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97897, "pid": 5, "tid": 7, "ts": 1716454222915817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788601, "dur": 11,
    "args": {
      "External id": 97897,
      "cbid": 211, "correlation": 97897
    }
  },
  {
    "ph": "s", "id": 97897, "pid": 76337, "tid": -914061504, "ts": 1716454222788601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222915893, "dur": 204,
    "args": {
      "External id": 97899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97899,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 97899, "pid": 5, "tid": 7, "ts": 1716454222915893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788621, "dur": 7,
    "args": {
      "External id": 97899,
      "cbid": 211, "correlation": 97899
    }
  },
  {
    "ph": "s", "id": 97899, "pid": 76337, "tid": -914061504, "ts": 1716454222788621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222916099, "dur": 40,
    "args": {
      "External id": 97901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97901,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97901, "pid": 5, "tid": 7, "ts": 1716454222916099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788633, "dur": 6,
    "args": {
      "External id": 97901,
      "cbid": 211, "correlation": 97901
    }
  },
  {
    "ph": "s", "id": 97901, "pid": 76337, "tid": -914061504, "ts": 1716454222788633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222916140, "dur": 58,
    "args": {
      "External id": 97907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97907,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97907, "pid": 5, "tid": 7, "ts": 1716454222916140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222788660, "dur": 491,
    "args": {
      "External id": 97907,
      "cbid": 211, "correlation": 97907
    }
  },
  {
    "ph": "s", "id": 97907, "pid": 76337, "tid": -914061504, "ts": 1716454222788660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222916199, "dur": 50,
    "args": {
      "External id": 97915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97915,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97915, "pid": 5, "tid": 7, "ts": 1716454222916199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789174, "dur": 9,
    "args": {
      "External id": 97915,
      "cbid": 211, "correlation": 97915
    }
  },
  {
    "ph": "s", "id": 97915, "pid": 76337, "tid": -914061504, "ts": 1716454222789174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222916250, "dur": 35,
    "args": {
      "External id": 97923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97923,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97923, "pid": 5, "tid": 7, "ts": 1716454222916250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789207, "dur": 10,
    "args": {
      "External id": 97923,
      "cbid": 211, "correlation": 97923
    }
  },
  {
    "ph": "s", "id": 97923, "pid": 76337, "tid": -914061504, "ts": 1716454222789207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222916287, "dur": 52,
    "args": {
      "External id": 97943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97943,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 97943, "pid": 5, "tid": 7, "ts": 1716454222916287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789293, "dur": 12,
    "args": {
      "External id": 97943,
      "cbid": 211, "correlation": 97943
    }
  },
  {
    "ph": "s", "id": 97943, "pid": 76337, "tid": -914061504, "ts": 1716454222789293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222916340, "dur": 4,
    "args": {
      "External id": 97955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97955,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 97955, "pid": 5, "tid": 7, "ts": 1716454222916340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789315, "dur": 7,
    "args": {
      "External id": 97955,
      "cbid": 211, "correlation": 97955
    }
  },
  {
    "ph": "s", "id": 97955, "pid": 76337, "tid": -914061504, "ts": 1716454222789315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222916346, "dur": 54,
    "args": {
      "External id": 97958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97958, "pid": 5, "tid": 7, "ts": 1716454222916346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789334, "dur": 7,
    "args": {
      "External id": 97958,
      "cbid": 211, "correlation": 97958
    }
  },
  {
    "ph": "s", "id": 97958, "pid": 76337, "tid": -914061504, "ts": 1716454222789334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789391, "dur": 0,
    "args": {
      "External id": 97969,
      "cbid": 317, "correlation": 97969
    }
  },
  {
    "ph": "f", "id": 97969, "pid": 76337, "tid": -914061504, "ts": 1716454222789391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789392, "dur": 0,
    "args": {
      "External id": 97970,
      "cbid": 203, "correlation": 97970
    }
  },
  {
    "ph": "f", "id": 97970, "pid": 76337, "tid": -914061504, "ts": 1716454222789392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789393, "dur": 0,
    "args": {
      "External id": 97971,
      "cbid": 205, "correlation": 97971
    }
  },
  {
    "ph": "f", "id": 97971, "pid": 76337, "tid": -914061504, "ts": 1716454222789393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789418, "dur": 1,
    "args": {
      "External id": 97975,
      "cbid": 251, "correlation": 97975
    }
  },
  {
    "ph": "f", "id": 97975, "pid": 76337, "tid": -914061504, "ts": 1716454222789418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789420, "dur": 0,
    "args": {
      "External id": 97976,
      "cbid": 251, "correlation": 97976
    }
  },
  {
    "ph": "f", "id": 97976, "pid": 76337, "tid": -914061504, "ts": 1716454222789420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789422, "dur": 0,
    "args": {
      "External id": 97977,
      "cbid": 251, "correlation": 97977
    }
  },
  {
    "ph": "f", "id": 97977, "pid": 76337, "tid": -914061504, "ts": 1716454222789422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789423, "dur": 0,
    "args": {
      "External id": 97978,
      "cbid": 251, "correlation": 97978
    }
  },
  {
    "ph": "f", "id": 97978, "pid": 76337, "tid": -914061504, "ts": 1716454222789423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789424, "dur": 0,
    "args": {
      "External id": 97979,
      "cbid": 251, "correlation": 97979
    }
  },
  {
    "ph": "f", "id": 97979, "pid": 76337, "tid": -914061504, "ts": 1716454222789424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789424, "dur": 0,
    "args": {
      "External id": 97980,
      "cbid": 251, "correlation": 97980
    }
  },
  {
    "ph": "f", "id": 97980, "pid": 76337, "tid": -914061504, "ts": 1716454222789424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789426, "dur": 0,
    "args": {
      "External id": 97981,
      "cbid": 251, "correlation": 97981
    }
  },
  {
    "ph": "f", "id": 97981, "pid": 76337, "tid": -914061504, "ts": 1716454222789426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789427, "dur": 0,
    "args": {
      "External id": 97982,
      "cbid": 251, "correlation": 97982
    }
  },
  {
    "ph": "f", "id": 97982, "pid": 76337, "tid": -914061504, "ts": 1716454222789427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789428, "dur": 0,
    "args": {
      "External id": 97983,
      "cbid": 251, "correlation": 97983
    }
  },
  {
    "ph": "f", "id": 97983, "pid": 76337, "tid": -914061504, "ts": 1716454222789428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222916401, "dur": 112,
    "args": {
      "External id": 97984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97984,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 97984, "pid": 5, "tid": 7, "ts": 1716454222916401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789431, "dur": 14,
    "args": {
      "External id": 97984,
      "cbid": 211, "correlation": 97984
    }
  },
  {
    "ph": "s", "id": 97984, "pid": 76337, "tid": -914061504, "ts": 1716454222789431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222916515, "dur": 59,
    "args": {
      "External id": 97990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97990, "pid": 5, "tid": 7, "ts": 1716454222916515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789467, "dur": 9,
    "args": {
      "External id": 97990,
      "cbid": 211, "correlation": 97990
    }
  },
  {
    "ph": "s", "id": 97990, "pid": 76337, "tid": -914061504, "ts": 1716454222789467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222916575, "dur": 607,
    "args": {
      "External id": 97999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 97999,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 97999, "pid": 5, "tid": 7, "ts": 1716454222916575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789557, "dur": 15,
    "args": {
      "External id": 97999,
      "cbid": 211, "correlation": 97999
    }
  },
  {
    "ph": "s", "id": 97999, "pid": 76337, "tid": -914061504, "ts": 1716454222789557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222917183, "dur": 177,
    "args": {
      "External id": 98021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98021,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98021, "pid": 5, "tid": 7, "ts": 1716454222917183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789622, "dur": 11,
    "args": {
      "External id": 98021,
      "cbid": 211, "correlation": 98021
    }
  },
  {
    "ph": "s", "id": 98021, "pid": 76337, "tid": -914061504, "ts": 1716454222789622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789722, "dur": 1,
    "args": {
      "External id": 98032,
      "cbid": 251, "correlation": 98032
    }
  },
  {
    "ph": "f", "id": 98032, "pid": 76337, "tid": -914061504, "ts": 1716454222789722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222917361, "dur": 194,
    "args": {
      "External id": 98033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98033,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 98033, "pid": 5, "tid": 7, "ts": 1716454222917361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789727, "dur": 13,
    "args": {
      "External id": 98033,
      "cbid": 211, "correlation": 98033
    }
  },
  {
    "ph": "s", "id": 98033, "pid": 76337, "tid": -914061504, "ts": 1716454222789727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789797, "dur": 1,
    "args": {
      "External id": 98044,
      "cbid": 251, "correlation": 98044
    }
  },
  {
    "ph": "f", "id": 98044, "pid": 76337, "tid": -914061504, "ts": 1716454222789797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222917556, "dur": 184,
    "args": {
      "External id": 98045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98045,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 98045, "pid": 5, "tid": 7, "ts": 1716454222917556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789801, "dur": 11,
    "args": {
      "External id": 98045,
      "cbid": 211, "correlation": 98045
    }
  },
  {
    "ph": "s", "id": 98045, "pid": 76337, "tid": -914061504, "ts": 1716454222789801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789864, "dur": 1,
    "args": {
      "External id": 98056,
      "cbid": 251, "correlation": 98056
    }
  },
  {
    "ph": "f", "id": 98056, "pid": 76337, "tid": -914061504, "ts": 1716454222789864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222917742, "dur": 185,
    "args": {
      "External id": 98057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98057,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 98057, "pid": 5, "tid": 7, "ts": 1716454222917742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789868, "dur": 11,
    "args": {
      "External id": 98057,
      "cbid": 211, "correlation": 98057
    }
  },
  {
    "ph": "s", "id": 98057, "pid": 76337, "tid": -914061504, "ts": 1716454222789868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222917928, "dur": 18122,
    "args": {
      "External id": 98078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98078,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 98078, "pid": 5, "tid": 7, "ts": 1716454222917928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222789961, "dur": 20,
    "args": {
      "External id": 98078,
      "cbid": 211, "correlation": 98078
    }
  },
  {
    "ph": "s", "id": 98078, "pid": 76337, "tid": -914061504, "ts": 1716454222789961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222790079, "dur": 2,
    "args": {
      "External id": 98096,
      "cbid": 251, "correlation": 98096
    }
  },
  {
    "ph": "f", "id": 98096, "pid": 76337, "tid": -914061504, "ts": 1716454222790079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222936051, "dur": 200,
    "args": {
      "External id": 98098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98098,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 98098, "pid": 5, "tid": 7, "ts": 1716454222936051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222790085, "dur": 13,
    "args": {
      "External id": 98098,
      "cbid": 211, "correlation": 98098
    }
  },
  {
    "ph": "s", "id": 98098, "pid": 76337, "tid": -914061504, "ts": 1716454222790085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222936252, "dur": 66,
    "args": {
      "External id": 98106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98106,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98106, "pid": 5, "tid": 7, "ts": 1716454222936252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222790159, "dur": 13,
    "args": {
      "External id": 98106,
      "cbid": 211, "correlation": 98106
    }
  },
  {
    "ph": "s", "id": 98106, "pid": 76337, "tid": -914061504, "ts": 1716454222790159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222936319, "dur": 97,
    "args": {
      "External id": 98114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98114,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98114, "pid": 5, "tid": 7, "ts": 1716454222936319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222790200, "dur": 26,
    "args": {
      "External id": 98114,
      "cbid": 211, "correlation": 98114
    }
  },
  {
    "ph": "s", "id": 98114, "pid": 76337, "tid": -914061504, "ts": 1716454222790200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222936418, "dur": 54,
    "args": {
      "External id": 98125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98125, "pid": 5, "tid": 7, "ts": 1716454222936418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222790298, "dur": 1838,
    "args": {
      "External id": 98125,
      "cbid": 211, "correlation": 98125
    }
  },
  {
    "ph": "s", "id": 98125, "pid": 76337, "tid": -914061504, "ts": 1716454222790298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222936472, "dur": 90,
    "args": {
      "External id": 98147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98147,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98147, "pid": 5, "tid": 7, "ts": 1716454222936472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792157, "dur": 124,
    "args": {
      "External id": 98147,
      "cbid": 211, "correlation": 98147
    }
  },
  {
    "ph": "s", "id": 98147, "pid": 76337, "tid": -914061504, "ts": 1716454222792157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792359, "dur": 1,
    "args": {
      "External id": 98158,
      "cbid": 251, "correlation": 98158
    }
  },
  {
    "ph": "f", "id": 98158, "pid": 76337, "tid": -914061504, "ts": 1716454222792359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222936564, "dur": 101,
    "args": {
      "External id": 98159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98159,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 98159, "pid": 5, "tid": 7, "ts": 1716454222936564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792364, "dur": 13,
    "args": {
      "External id": 98159,
      "cbid": 211, "correlation": 98159
    }
  },
  {
    "ph": "s", "id": 98159, "pid": 76337, "tid": -914061504, "ts": 1716454222792364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792445, "dur": 1,
    "args": {
      "External id": 98170,
      "cbid": 251, "correlation": 98170
    }
  },
  {
    "ph": "f", "id": 98170, "pid": 76337, "tid": -914061504, "ts": 1716454222792445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792449, "dur": 0,
    "args": {
      "External id": 98171,
      "cbid": 251, "correlation": 98171
    }
  },
  {
    "ph": "f", "id": 98171, "pid": 76337, "tid": -914061504, "ts": 1716454222792449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222936666, "dur": 10,
    "args": {
      "External id": 98172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98172,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98172, "pid": 5, "tid": 7, "ts": 1716454222936666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792451, "dur": 13,
    "args": {
      "External id": 98172,
      "cbid": 211, "correlation": 98172
    }
  },
  {
    "ph": "s", "id": 98172, "pid": 76337, "tid": -914061504, "ts": 1716454222792451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222936677, "dur": 5,
    "args": {
      "External id": 98174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98174,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 98174, "pid": 5, "tid": 7, "ts": 1716454222936677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792466, "dur": 7,
    "args": {
      "External id": 98174,
      "cbid": 211, "correlation": 98174
    }
  },
  {
    "ph": "s", "id": 98174, "pid": 76337, "tid": -914061504, "ts": 1716454222792466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792529, "dur": 1,
    "args": {
      "External id": 98185,
      "cbid": 251, "correlation": 98185
    }
  },
  {
    "ph": "f", "id": 98185, "pid": 76337, "tid": -914061504, "ts": 1716454222792529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792532, "dur": 0,
    "args": {
      "External id": 98186,
      "cbid": 251, "correlation": 98186
    }
  },
  {
    "ph": "f", "id": 98186, "pid": 76337, "tid": -914061504, "ts": 1716454222792532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222936684, "dur": 6,
    "args": {
      "External id": 98187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98187,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98187, "pid": 5, "tid": 7, "ts": 1716454222936684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792534, "dur": 12,
    "args": {
      "External id": 98187,
      "cbid": 211, "correlation": 98187
    }
  },
  {
    "ph": "s", "id": 98187, "pid": 76337, "tid": -914061504, "ts": 1716454222792534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222936691, "dur": 3,
    "args": {
      "External id": 98189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98189,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 98189, "pid": 5, "tid": 7, "ts": 1716454222936691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792547, "dur": 6,
    "args": {
      "External id": 98189,
      "cbid": 211, "correlation": 98189
    }
  },
  {
    "ph": "s", "id": 98189, "pid": 76337, "tid": -914061504, "ts": 1716454222792547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222936696, "dur": 152,
    "args": {
      "External id": 98210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98210,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 98210, "pid": 5, "tid": 7, "ts": 1716454222936696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792623, "dur": 12,
    "args": {
      "External id": 98210,
      "cbid": 211, "correlation": 98210
    }
  },
  {
    "ph": "s", "id": 98210, "pid": 76337, "tid": -914061504, "ts": 1716454222792623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792721, "dur": 1,
    "args": {
      "External id": 98228,
      "cbid": 251, "correlation": 98228
    }
  },
  {
    "ph": "f", "id": 98228, "pid": 76337, "tid": -914061504, "ts": 1716454222792721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222936850, "dur": 104,
    "args": {
      "External id": 98230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98230,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 98230, "pid": 5, "tid": 7, "ts": 1716454222936850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792728, "dur": 14,
    "args": {
      "External id": 98230,
      "cbid": 211, "correlation": 98230
    }
  },
  {
    "ph": "s", "id": 98230, "pid": 76337, "tid": -914061504, "ts": 1716454222792728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222936955, "dur": 35,
    "args": {
      "External id": 98238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98238,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98238, "pid": 5, "tid": 7, "ts": 1716454222936955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792798, "dur": 12,
    "args": {
      "External id": 98238,
      "cbid": 211, "correlation": 98238
    }
  },
  {
    "ph": "s", "id": 98238, "pid": 76337, "tid": -914061504, "ts": 1716454222792798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222936991, "dur": 67,
    "args": {
      "External id": 98246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98246, "pid": 5, "tid": 7, "ts": 1716454222936991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792839, "dur": 10,
    "args": {
      "External id": 98246,
      "cbid": 211, "correlation": 98246
    }
  },
  {
    "ph": "s", "id": 98246, "pid": 76337, "tid": -914061504, "ts": 1716454222792839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222937059, "dur": 90,
    "args": {
      "External id": 98268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98268,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98268, "pid": 5, "tid": 7, "ts": 1716454222937059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792891, "dur": 10,
    "args": {
      "External id": 98268,
      "cbid": 211, "correlation": 98268
    }
  },
  {
    "ph": "s", "id": 98268, "pid": 76337, "tid": -914061504, "ts": 1716454222792891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792987, "dur": 1,
    "args": {
      "External id": 98284,
      "cbid": 251, "correlation": 98284
    }
  },
  {
    "ph": "f", "id": 98284, "pid": 76337, "tid": -914061504, "ts": 1716454222792987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222937151, "dur": 562,
    "args": {
      "External id": 98286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98286,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 98286, "pid": 5, "tid": 7, "ts": 1716454222937151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222792993, "dur": 13,
    "args": {
      "External id": 98286,
      "cbid": 211, "correlation": 98286
    }
  },
  {
    "ph": "s", "id": 98286, "pid": 76337, "tid": -914061504, "ts": 1716454222792993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222937714, "dur": 240,
    "args": {
      "External id": 98294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98294, "pid": 5, "tid": 7, "ts": 1716454222937714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793069, "dur": 15,
    "args": {
      "External id": 98294,
      "cbid": 211, "correlation": 98294
    }
  },
  {
    "ph": "s", "id": 98294, "pid": 76337, "tid": -914061504, "ts": 1716454222793069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222937955, "dur": 251,
    "args": {
      "External id": 98302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98302,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98302, "pid": 5, "tid": 7, "ts": 1716454222937955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793106, "dur": 9,
    "args": {
      "External id": 98302,
      "cbid": 211, "correlation": 98302
    }
  },
  {
    "ph": "s", "id": 98302, "pid": 76337, "tid": -914061504, "ts": 1716454222793106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793189, "dur": 1,
    "args": {
      "External id": 98318,
      "cbid": 251, "correlation": 98318
    }
  },
  {
    "ph": "f", "id": 98318, "pid": 76337, "tid": -914061504, "ts": 1716454222793189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793195, "dur": 0,
    "args": {
      "External id": 98320,
      "cbid": 251, "correlation": 98320
    }
  },
  {
    "ph": "f", "id": 98320, "pid": 76337, "tid": -914061504, "ts": 1716454222793195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222938208, "dur": 359,
    "args": {
      "External id": 98321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98321,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98321, "pid": 5, "tid": 7, "ts": 1716454222938208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793199, "dur": 13,
    "args": {
      "External id": 98321,
      "cbid": 211, "correlation": 98321
    }
  },
  {
    "ph": "s", "id": 98321, "pid": 76337, "tid": -914061504, "ts": 1716454222793199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222938568, "dur": 50,
    "args": {
      "External id": 98329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98329,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98329, "pid": 5, "tid": 7, "ts": 1716454222938568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793242, "dur": 115,
    "args": {
      "External id": 98329,
      "cbid": 211, "correlation": 98329
    }
  },
  {
    "ph": "s", "id": 98329, "pid": 76337, "tid": -914061504, "ts": 1716454222793242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222938619, "dur": 155,
    "args": {
      "External id": 98340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98340, "pid": 5, "tid": 7, "ts": 1716454222938619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793419, "dur": 66,
    "args": {
      "External id": 98340,
      "cbid": 211, "correlation": 98340
    }
  },
  {
    "ph": "s", "id": 98340, "pid": 76337, "tid": -914061504, "ts": 1716454222793419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793541, "dur": 0,
    "args": {
      "External id": 98352,
      "cbid": 317, "correlation": 98352
    }
  },
  {
    "ph": "f", "id": 98352, "pid": 76337, "tid": -914061504, "ts": 1716454222793541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793542, "dur": 0,
    "args": {
      "External id": 98353,
      "cbid": 203, "correlation": 98353
    }
  },
  {
    "ph": "f", "id": 98353, "pid": 76337, "tid": -914061504, "ts": 1716454222793542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793543, "dur": 0,
    "args": {
      "External id": 98354,
      "cbid": 205, "correlation": 98354
    }
  },
  {
    "ph": "f", "id": 98354, "pid": 76337, "tid": -914061504, "ts": 1716454222793543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793570, "dur": 1,
    "args": {
      "External id": 98358,
      "cbid": 251, "correlation": 98358
    }
  },
  {
    "ph": "f", "id": 98358, "pid": 76337, "tid": -914061504, "ts": 1716454222793570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793572, "dur": 0,
    "args": {
      "External id": 98359,
      "cbid": 251, "correlation": 98359
    }
  },
  {
    "ph": "f", "id": 98359, "pid": 76337, "tid": -914061504, "ts": 1716454222793572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793573, "dur": 0,
    "args": {
      "External id": 98360,
      "cbid": 251, "correlation": 98360
    }
  },
  {
    "ph": "f", "id": 98360, "pid": 76337, "tid": -914061504, "ts": 1716454222793573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793574, "dur": 0,
    "args": {
      "External id": 98361,
      "cbid": 251, "correlation": 98361
    }
  },
  {
    "ph": "f", "id": 98361, "pid": 76337, "tid": -914061504, "ts": 1716454222793574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793575, "dur": 1,
    "args": {
      "External id": 98362,
      "cbid": 251, "correlation": 98362
    }
  },
  {
    "ph": "f", "id": 98362, "pid": 76337, "tid": -914061504, "ts": 1716454222793575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793576, "dur": 0,
    "args": {
      "External id": 98363,
      "cbid": 251, "correlation": 98363
    }
  },
  {
    "ph": "f", "id": 98363, "pid": 76337, "tid": -914061504, "ts": 1716454222793576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793577, "dur": 0,
    "args": {
      "External id": 98364,
      "cbid": 251, "correlation": 98364
    }
  },
  {
    "ph": "f", "id": 98364, "pid": 76337, "tid": -914061504, "ts": 1716454222793577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793578, "dur": 0,
    "args": {
      "External id": 98365,
      "cbid": 251, "correlation": 98365
    }
  },
  {
    "ph": "f", "id": 98365, "pid": 76337, "tid": -914061504, "ts": 1716454222793578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793579, "dur": 0,
    "args": {
      "External id": 98366,
      "cbid": 251, "correlation": 98366
    }
  },
  {
    "ph": "f", "id": 98366, "pid": 76337, "tid": -914061504, "ts": 1716454222793579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222938776, "dur": 114,
    "args": {
      "External id": 98367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98367,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 98367, "pid": 5, "tid": 7, "ts": 1716454222938776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793582, "dur": 31,
    "args": {
      "External id": 98367,
      "cbid": 211, "correlation": 98367
    }
  },
  {
    "ph": "s", "id": 98367, "pid": 76337, "tid": -914061504, "ts": 1716454222793582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222938891, "dur": 59,
    "args": {
      "External id": 98373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98373, "pid": 5, "tid": 7, "ts": 1716454222938891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793635, "dur": 276,
    "args": {
      "External id": 98373,
      "cbid": 211, "correlation": 98373
    }
  },
  {
    "ph": "s", "id": 98373, "pid": 76337, "tid": -914061504, "ts": 1716454222793635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222938952, "dur": 50,
    "args": {
      "External id": 98381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98381,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98381, "pid": 5, "tid": 7, "ts": 1716454222938952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222793935, "dur": 9,
    "args": {
      "External id": 98381,
      "cbid": 211, "correlation": 98381
    }
  },
  {
    "ph": "s", "id": 98381, "pid": 76337, "tid": -914061504, "ts": 1716454222793935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222939003, "dur": 51,
    "args": {
      "External id": 98401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98401,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 98401, "pid": 5, "tid": 7, "ts": 1716454222939003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794030, "dur": 13,
    "args": {
      "External id": 98401,
      "cbid": 211, "correlation": 98401
    }
  },
  {
    "ph": "s", "id": 98401, "pid": 76337, "tid": -914061504, "ts": 1716454222794030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222939056, "dur": 4,
    "args": {
      "External id": 98413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98413,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 98413, "pid": 5, "tid": 7, "ts": 1716454222939056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794053, "dur": 6,
    "args": {
      "External id": 98413,
      "cbid": 211, "correlation": 98413
    }
  },
  {
    "ph": "s", "id": 98413, "pid": 76337, "tid": -914061504, "ts": 1716454222794053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222939061, "dur": 56,
    "args": {
      "External id": 98416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98416,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98416, "pid": 5, "tid": 7, "ts": 1716454222939061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794072, "dur": 93,
    "args": {
      "External id": 98416,
      "cbid": 211, "correlation": 98416
    }
  },
  {
    "ph": "s", "id": 98416, "pid": 76337, "tid": -914061504, "ts": 1716454222794072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222939118, "dur": 37,
    "args": {
      "External id": 98425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98425,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98425, "pid": 5, "tid": 7, "ts": 1716454222939118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794207, "dur": 11,
    "args": {
      "External id": 98425,
      "cbid": 211, "correlation": 98425
    }
  },
  {
    "ph": "s", "id": 98425, "pid": 76337, "tid": -914061504, "ts": 1716454222794207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794264, "dur": 0,
    "args": {
      "External id": 98435,
      "cbid": 317, "correlation": 98435
    }
  },
  {
    "ph": "f", "id": 98435, "pid": 76337, "tid": -914061504, "ts": 1716454222794264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794264, "dur": 0,
    "args": {
      "External id": 98436,
      "cbid": 203, "correlation": 98436
    }
  },
  {
    "ph": "f", "id": 98436, "pid": 76337, "tid": -914061504, "ts": 1716454222794264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794265, "dur": 0,
    "args": {
      "External id": 98437,
      "cbid": 205, "correlation": 98437
    }
  },
  {
    "ph": "f", "id": 98437, "pid": 76337, "tid": -914061504, "ts": 1716454222794265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222939156, "dur": 41,
    "args": {
      "External id": 98441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98441,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98441, "pid": 5, "tid": 7, "ts": 1716454222939156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794284, "dur": 12,
    "args": {
      "External id": 98441,
      "cbid": 211, "correlation": 98441
    }
  },
  {
    "ph": "s", "id": 98441, "pid": 76337, "tid": -914061504, "ts": 1716454222794284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222939199, "dur": 3,
    "args": {
      "External id": 98443,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98443,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.413793103448276
    }
  },
  {
    "ph": "f", "id": 98443, "pid": 5, "tid": 7, "ts": 1716454222939199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794300, "dur": 23,
    "args": {
      "External id": 98443,
      "cbid": 51, "correlation": 98443
    }
  },
  {
    "ph": "s", "id": 98443, "pid": 76337, "tid": -914061504, "ts": 1716454222794300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794328, "dur": 2,
    "args": {
      "External id": 98445,
      "cbid": 200, "correlation": 98445
    }
  },
  {
    "ph": "f", "id": 98445, "pid": 76337, "tid": -914061504, "ts": 1716454222794328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794331, "dur": 0,
    "args": {
      "External id": 98446,
      "cbid": 200, "correlation": 98446
    }
  },
  {
    "ph": "f", "id": 98446, "pid": 76337, "tid": -914061504, "ts": 1716454222794331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794331, "dur": 0,
    "args": {
      "External id": 98447,
      "cbid": 200, "correlation": 98447
    }
  },
  {
    "ph": "f", "id": 98447, "pid": 76337, "tid": -914061504, "ts": 1716454222794331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794331, "dur": 0,
    "args": {
      "External id": 98448,
      "cbid": 200, "correlation": 98448
    }
  },
  {
    "ph": "f", "id": 98448, "pid": 76337, "tid": -914061504, "ts": 1716454222794331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794333, "dur": 4,
    "args": {
      "External id": 98449,
      "cbid": 15, "correlation": 98449
    }
  },
  {
    "ph": "f", "id": 98449, "pid": 76337, "tid": -914061504, "ts": 1716454222794333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794337, "dur": 1,
    "args": {
      "External id": 98450,
      "cbid": 251, "correlation": 98450
    }
  },
  {
    "ph": "f", "id": 98450, "pid": 76337, "tid": -914061504, "ts": 1716454222794337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454222939203, "dur": 25,
    "args": {
      "External id": 98451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98451,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98451, "pid": 5, "tid": 7, "ts": 1716454222939203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794341, "dur": 8,
    "args": {
      "External id": 98451,
      "cbid": 211, "correlation": 98451
    }
  },
  {
    "ph": "s", "id": 98451, "pid": 76337, "tid": -914061504, "ts": 1716454222794341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222939230, "dur": 4,
    "args": {
      "External id": 98453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 98453, "pid": 5, "tid": 7, "ts": 1716454222939230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794356, "dur": 6,
    "args": {
      "External id": 98453,
      "cbid": 211, "correlation": 98453
    }
  },
  {
    "ph": "s", "id": 98453, "pid": 76337, "tid": -914061504, "ts": 1716454222794356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794367, "dur": 0,
    "args": {
      "External id": 98454,
      "cbid": 51, "correlation": 98454
    }
  },
  {
    "ph": "s", "id": 98454, "pid": 76337, "tid": -914061504, "ts": 1716454222794367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222939235, "dur": 185,
    "args": {
      "External id": 98455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98455,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98455, "pid": 5, "tid": 7, "ts": 1716454222939235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794368, "dur": 184,
    "args": {
      "External id": 98455,
      "cbid": 211, "correlation": 98455
    }
  },
  {
    "ph": "s", "id": 98455, "pid": 76337, "tid": -914061504, "ts": 1716454222794368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222939421, "dur": 6,
    "args": {
      "External id": 98456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98456, "pid": 5, "tid": 7, "ts": 1716454222939421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794556, "dur": 6,
    "args": {
      "External id": 98456,
      "cbid": 211, "correlation": 98456
    }
  },
  {
    "ph": "s", "id": 98456, "pid": 76337, "tid": -914061504, "ts": 1716454222794556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222939428, "dur": 5,
    "args": {
      "External id": 98462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 98462, "pid": 5, "tid": 7, "ts": 1716454222939428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222794586, "dur": 9,
    "args": {
      "External id": 98462,
      "cbid": 211, "correlation": 98462
    }
  },
  {
    "ph": "s", "id": 98462, "pid": 76337, "tid": -914061504, "ts": 1716454222794586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222939434, "dur": 3,
    "args": {
      "External id": 98470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98470,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98470, "pid": 5, "tid": 7, "ts": 1716454222939434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796327, "dur": 16,
    "args": {
      "External id": 98470,
      "cbid": 211, "correlation": 98470
    }
  },
  {
    "ph": "s", "id": 98470, "pid": 76337, "tid": -914061504, "ts": 1716454222796327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222939438, "dur": 3,
    "args": {
      "External id": 98478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98478,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98478, "pid": 5, "tid": 7, "ts": 1716454222939438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796373, "dur": 10,
    "args": {
      "External id": 98478,
      "cbid": 211, "correlation": 98478
    }
  },
  {
    "ph": "s", "id": 98478, "pid": 76337, "tid": -914061504, "ts": 1716454222796373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222939443, "dur": 3,
    "args": {
      "External id": 98486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98486,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98486, "pid": 5, "tid": 7, "ts": 1716454222939443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796400, "dur": 9,
    "args": {
      "External id": 98486,
      "cbid": 211, "correlation": 98486
    }
  },
  {
    "ph": "s", "id": 98486, "pid": 76337, "tid": -914061504, "ts": 1716454222796400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222939447, "dur": 3,
    "args": {
      "External id": 98495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98495,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98495, "pid": 5, "tid": 7, "ts": 1716454222939447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796583, "dur": 14,
    "args": {
      "External id": 98495,
      "cbid": 211, "correlation": 98495
    }
  },
  {
    "ph": "s", "id": 98495, "pid": 76337, "tid": -914061504, "ts": 1716454222796583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222939450, "dur": 3,
    "args": {
      "External id": 98504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98504,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98504, "pid": 5, "tid": 7, "ts": 1716454222939450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796613, "dur": 7,
    "args": {
      "External id": 98504,
      "cbid": 211, "correlation": 98504
    }
  },
  {
    "ph": "s", "id": 98504, "pid": 76337, "tid": -914061504, "ts": 1716454222796613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222939455, "dur": 3,
    "args": {
      "External id": 98512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98512,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98512, "pid": 5, "tid": 7, "ts": 1716454222939455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796641, "dur": 9,
    "args": {
      "External id": 98512,
      "cbid": 211, "correlation": 98512
    }
  },
  {
    "ph": "s", "id": 98512, "pid": 76337, "tid": -914061504, "ts": 1716454222796641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222939459, "dur": 3,
    "args": {
      "External id": 98520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98520,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98520, "pid": 5, "tid": 7, "ts": 1716454222939459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796903, "dur": 15,
    "args": {
      "External id": 98520,
      "cbid": 211, "correlation": 98520
    }
  },
  {
    "ph": "s", "id": 98520, "pid": 76337, "tid": -914061504, "ts": 1716454222796903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222939463, "dur": 3,
    "args": {
      "External id": 98528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98528,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98528, "pid": 5, "tid": 7, "ts": 1716454222939463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222796935, "dur": 8,
    "args": {
      "External id": 98528,
      "cbid": 211, "correlation": 98528
    }
  },
  {
    "ph": "s", "id": 98528, "pid": 76337, "tid": -914061504, "ts": 1716454222796935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222939468, "dur": 1,
    "args": {
      "External id": 98538,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98538,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 98538, "pid": 5, "tid": 7, "ts": 1716454222939468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222797010, "dur": 36,
    "args": {
      "External id": 98538,
      "cbid": 41, "correlation": 98538
    }
  },
  {
    "ph": "s", "id": 98538, "pid": 76337, "tid": -914061504, "ts": 1716454222797010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222797047, "dur": 142451,
    "args": {
      "External id": 98539,
      "cbid": 131, "correlation": 98539
    }
  },
  {
    "ph": "f", "id": 98539, "pid": 76337, "tid": -914061504, "ts": 1716454222797047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222939931, "dur": 3,
    "args": {
      "External id": 98547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98547,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98547, "pid": 5, "tid": 7, "ts": 1716454222939931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222939862, "dur": 75,
    "args": {
      "External id": 98547,
      "cbid": 211, "correlation": 98547
    }
  },
  {
    "ph": "s", "id": 98547, "pid": 76337, "tid": -914061504, "ts": 1716454222939862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940075, "dur": 3,
    "args": {
      "External id": 98556,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98556,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0024279210925644916
    }
  },
  {
    "ph": "f", "id": 98556, "pid": 5, "tid": 7, "ts": 1716454222940075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940026, "dur": 50,
    "args": {
      "External id": 98556,
      "cbid": 41, "correlation": 98556
    }
  },
  {
    "ph": "s", "id": 98556, "pid": 76337, "tid": -914061504, "ts": 1716454222940026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222940193, "dur": 4,
    "args": {
      "External id": 98566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98566,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98566, "pid": 5, "tid": 7, "ts": 1716454222940193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940175, "dur": 19,
    "args": {
      "External id": 98566,
      "cbid": 211, "correlation": 98566
    }
  },
  {
    "ph": "s", "id": 98566, "pid": 76337, "tid": -914061504, "ts": 1716454222940175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940301, "dur": 1,
    "args": {
      "External id": 98576,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98576,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 98576, "pid": 5, "tid": 7, "ts": 1716454222940301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940278, "dur": 21,
    "args": {
      "External id": 98576,
      "cbid": 41, "correlation": 98576
    }
  },
  {
    "ph": "s", "id": 98576, "pid": 76337, "tid": -914061504, "ts": 1716454222940278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940300, "dur": 7,
    "args": {
      "External id": 98577,
      "cbid": 131, "correlation": 98577
    }
  },
  {
    "ph": "f", "id": 98577, "pid": 76337, "tid": -914061504, "ts": 1716454222940300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940384, "dur": 3,
    "args": {
      "External id": 98584,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98584,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.03030303030303
    }
  },
  {
    "ph": "f", "id": 98584, "pid": 5, "tid": 7, "ts": 1716454222940384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940361, "dur": 23,
    "args": {
      "External id": 98584,
      "cbid": 41, "correlation": 98584
    }
  },
  {
    "ph": "s", "id": 98584, "pid": 76337, "tid": -914061504, "ts": 1716454222940361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940484, "dur": 2,
    "args": {
      "External id": 98603,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98603,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005376344086021506
    }
  },
  {
    "ph": "f", "id": 98603, "pid": 5, "tid": 7, "ts": 1716454222940484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940464, "dur": 19,
    "args": {
      "External id": 98603,
      "cbid": 41, "correlation": 98603
    }
  },
  {
    "ph": "s", "id": 98603, "pid": 76337, "tid": -914061504, "ts": 1716454222940464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222940526, "dur": 3,
    "args": {
      "External id": 98609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98609, "pid": 5, "tid": 7, "ts": 1716454222940526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940513, "dur": 14,
    "args": {
      "External id": 98609,
      "cbid": 211, "correlation": 98609
    }
  },
  {
    "ph": "s", "id": 98609, "pid": 76337, "tid": -914061504, "ts": 1716454222940513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222940543, "dur": 6,
    "args": {
      "External id": 98611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98611,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 98611, "pid": 5, "tid": 7, "ts": 1716454222940543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940532, "dur": 10,
    "args": {
      "External id": 98611,
      "cbid": 211, "correlation": 98611
    }
  },
  {
    "ph": "s", "id": 98611, "pid": 76337, "tid": -914061504, "ts": 1716454222940532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454222940552, "dur": 3,
    "args": {
      "External id": 98613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98613,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98613, "pid": 5, "tid": 7, "ts": 1716454222940552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940544, "dur": 7,
    "args": {
      "External id": 98613,
      "cbid": 211, "correlation": 98613
    }
  },
  {
    "ph": "s", "id": 98613, "pid": 76337, "tid": -914061504, "ts": 1716454222940544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940588, "dur": 2,
    "args": {
      "External id": 98621,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98621,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002808002808002808
    }
  },
  {
    "ph": "f", "id": 98621, "pid": 5, "tid": 7, "ts": 1716454222940588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940572, "dur": 15,
    "args": {
      "External id": 98621,
      "cbid": 41, "correlation": 98621
    }
  },
  {
    "ph": "s", "id": 98621, "pid": 76337, "tid": -914061504, "ts": 1716454222940572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222940642, "dur": 3,
    "args": {
      "External id": 98635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98635, "pid": 5, "tid": 7, "ts": 1716454222940642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940630, "dur": 13,
    "args": {
      "External id": 98635,
      "cbid": 211, "correlation": 98635
    }
  },
  {
    "ph": "s", "id": 98635, "pid": 76337, "tid": -914061504, "ts": 1716454222940630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222940664, "dur": 2,
    "args": {
      "External id": 98649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98649,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98649, "pid": 5, "tid": 7, "ts": 1716454222940664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940655, "dur": 8,
    "args": {
      "External id": 98649,
      "cbid": 211, "correlation": 98649
    }
  },
  {
    "ph": "s", "id": 98649, "pid": 76337, "tid": -914061504, "ts": 1716454222940655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222940703, "dur": 6,
    "args": {
      "External id": 98656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98656,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98656, "pid": 5, "tid": 7, "ts": 1716454222940703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940692, "dur": 11,
    "args": {
      "External id": 98656,
      "cbid": 211, "correlation": 98656
    }
  },
  {
    "ph": "s", "id": 98656, "pid": 76337, "tid": -914061504, "ts": 1716454222940692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454222940714, "dur": 6,
    "args": {
      "External id": 98659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98659,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98659, "pid": 5, "tid": 7, "ts": 1716454222940714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940706, "dur": 7,
    "args": {
      "External id": 98659,
      "cbid": 211, "correlation": 98659
    }
  },
  {
    "ph": "s", "id": 98659, "pid": 76337, "tid": -914061504, "ts": 1716454222940706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222940723, "dur": 3,
    "args": {
      "External id": 98661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98661,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98661, "pid": 5, "tid": 7, "ts": 1716454222940723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940715, "dur": 8,
    "args": {
      "External id": 98661,
      "cbid": 211, "correlation": 98661
    }
  },
  {
    "ph": "s", "id": 98661, "pid": 76337, "tid": -914061504, "ts": 1716454222940715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940745, "dur": 2,
    "args": {
      "External id": 98664,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98664,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0028089887640449437
    }
  },
  {
    "ph": "f", "id": 98664, "pid": 5, "tid": 7, "ts": 1716454222940745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940732, "dur": 12,
    "args": {
      "External id": 98664,
      "cbid": 41, "correlation": 98664
    }
  },
  {
    "ph": "s", "id": 98664, "pid": 76337, "tid": -914061504, "ts": 1716454222940732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222940800, "dur": 4,
    "args": {
      "External id": 98680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98680,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98680, "pid": 5, "tid": 7, "ts": 1716454222940800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940786, "dur": 15,
    "args": {
      "External id": 98680,
      "cbid": 211, "correlation": 98680
    }
  },
  {
    "ph": "s", "id": 98680, "pid": 76337, "tid": -914061504, "ts": 1716454222940786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222940822, "dur": 3,
    "args": {
      "External id": 98685,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98685,
      "bytes": 1, "memory bandwidth (GB/s)": 0.00030637254901960784
    }
  },
  {
    "ph": "f", "id": 98685, "pid": 5, "tid": 7, "ts": 1716454222940822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940806, "dur": 15,
    "args": {
      "External id": 98685,
      "cbid": 41, "correlation": 98685
    }
  },
  {
    "ph": "s", "id": 98685, "pid": 76337, "tid": -914061504, "ts": 1716454222940806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222940849, "dur": 1,
    "args": {
      "External id": 98691,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98691,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005896226415094339
    }
  },
  {
    "ph": "f", "id": 98691, "pid": 5, "tid": 7, "ts": 1716454222940849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940831, "dur": 27,
    "args": {
      "External id": 98691,
      "cbid": 41, "correlation": 98691
    }
  },
  {
    "ph": "s", "id": 98691, "pid": 76337, "tid": -914061504, "ts": 1716454222940831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940859, "dur": 4,
    "args": {
      "External id": 98692,
      "cbid": 131, "correlation": 98692
    }
  },
  {
    "ph": "f", "id": 98692, "pid": 76337, "tid": -914061504, "ts": 1716454222940859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222940918, "dur": 3,
    "args": {
      "External id": 98700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98700,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98700, "pid": 5, "tid": 7, "ts": 1716454222940918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940902, "dur": 16,
    "args": {
      "External id": 98700,
      "cbid": 211, "correlation": 98700
    }
  },
  {
    "ph": "s", "id": 98700, "pid": 76337, "tid": -914061504, "ts": 1716454222940902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222940951, "dur": 3,
    "args": {
      "External id": 98710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98710,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98710, "pid": 5, "tid": 7, "ts": 1716454222940951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940941, "dur": 9,
    "args": {
      "External id": 98710,
      "cbid": 211, "correlation": 98710
    }
  },
  {
    "ph": "s", "id": 98710, "pid": 76337, "tid": -914061504, "ts": 1716454222940941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222940992, "dur": 3,
    "args": {
      "External id": 98719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98719,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98719, "pid": 5, "tid": 7, "ts": 1716454222940992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222940980, "dur": 11,
    "args": {
      "External id": 98719,
      "cbid": 211, "correlation": 98719
    }
  },
  {
    "ph": "s", "id": 98719, "pid": 76337, "tid": -914061504, "ts": 1716454222940980,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222941126, "dur": 12,
    "args": {
      "External id": 98729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98729,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98729, "pid": 5, "tid": 7, "ts": 1716454222941126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941112, "dur": 15,
    "args": {
      "External id": 98729,
      "cbid": 211, "correlation": 98729
    }
  },
  {
    "ph": "s", "id": 98729, "pid": 76337, "tid": -914061504, "ts": 1716454222941112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222941168, "dur": 3,
    "args": {
      "External id": 98737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98737,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98737, "pid": 5, "tid": 7, "ts": 1716454222941168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941158, "dur": 9,
    "args": {
      "External id": 98737,
      "cbid": 211, "correlation": 98737
    }
  },
  {
    "ph": "s", "id": 98737, "pid": 76337, "tid": -914061504, "ts": 1716454222941158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222941219, "dur": 11,
    "args": {
      "External id": 98747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98747,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98747, "pid": 5, "tid": 7, "ts": 1716454222941219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941208, "dur": 12,
    "args": {
      "External id": 98747,
      "cbid": 211, "correlation": 98747
    }
  },
  {
    "ph": "s", "id": 98747, "pid": 76337, "tid": -914061504, "ts": 1716454222941208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222941257, "dur": 10,
    "args": {
      "External id": 98755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98755,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98755, "pid": 5, "tid": 7, "ts": 1716454222941257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941248, "dur": 9,
    "args": {
      "External id": 98755,
      "cbid": 211, "correlation": 98755
    }
  },
  {
    "ph": "s", "id": 98755, "pid": 76337, "tid": -914061504, "ts": 1716454222941248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222941284, "dur": 3,
    "args": {
      "External id": 98764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98764,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98764, "pid": 5, "tid": 7, "ts": 1716454222941284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941275, "dur": 9,
    "args": {
      "External id": 98764,
      "cbid": 211, "correlation": 98764
    }
  },
  {
    "ph": "s", "id": 98764, "pid": 76337, "tid": -914061504, "ts": 1716454222941275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222941313, "dur": 5,
    "args": {
      "External id": 98773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98773,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98773, "pid": 5, "tid": 7, "ts": 1716454222941313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941300, "dur": 12,
    "args": {
      "External id": 98773,
      "cbid": 211, "correlation": 98773
    }
  },
  {
    "ph": "s", "id": 98773, "pid": 76337, "tid": -914061504, "ts": 1716454222941300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222941354, "dur": 8,
    "args": {
      "External id": 98783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98783,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98783, "pid": 5, "tid": 7, "ts": 1716454222941354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941343, "dur": 12,
    "args": {
      "External id": 98783,
      "cbid": 211, "correlation": 98783
    }
  },
  {
    "ph": "s", "id": 98783, "pid": 76337, "tid": -914061504, "ts": 1716454222941343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222941723, "dur": 3,
    "args": {
      "External id": 98792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98792,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98792, "pid": 5, "tid": 7, "ts": 1716454222941723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941708, "dur": 15,
    "args": {
      "External id": 98792,
      "cbid": 211, "correlation": 98792
    }
  },
  {
    "ph": "s", "id": 98792, "pid": 76337, "tid": -914061504, "ts": 1716454222941708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222941756, "dur": 3,
    "args": {
      "External id": 98800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98800,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98800, "pid": 5, "tid": 7, "ts": 1716454222941756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941746, "dur": 11,
    "args": {
      "External id": 98800,
      "cbid": 211, "correlation": 98800
    }
  },
  {
    "ph": "s", "id": 98800, "pid": 76337, "tid": -914061504, "ts": 1716454222941746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222941849, "dur": 1,
    "args": {
      "External id": 98810,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98810,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 98810, "pid": 5, "tid": 7, "ts": 1716454222941849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941794, "dur": 53,
    "args": {
      "External id": 98810,
      "cbid": 41, "correlation": 98810
    }
  },
  {
    "ph": "s", "id": 98810, "pid": 76337, "tid": -914061504, "ts": 1716454222941794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941849, "dur": 8,
    "args": {
      "External id": 98811,
      "cbid": 131, "correlation": 98811
    }
  },
  {
    "ph": "f", "id": 98811, "pid": 76337, "tid": -914061504, "ts": 1716454222941849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454222941945, "dur": 2,
    "args": {
      "External id": 98819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98819, "pid": 5, "tid": 7, "ts": 1716454222941945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222941930, "dur": 15,
    "args": {
      "External id": 98819,
      "cbid": 211, "correlation": 98819
    }
  },
  {
    "ph": "s", "id": 98819, "pid": 76337, "tid": -914061504, "ts": 1716454222941930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222942031, "dur": 3,
    "args": {
      "External id": 98828,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98828,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0026595744680851063
    }
  },
  {
    "ph": "f", "id": 98828, "pid": 5, "tid": 7, "ts": 1716454222942031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942012, "dur": 18,
    "args": {
      "External id": 98828,
      "cbid": 41, "correlation": 98828
    }
  },
  {
    "ph": "s", "id": 98828, "pid": 76337, "tid": -914061504, "ts": 1716454222942012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222942104, "dur": 3,
    "args": {
      "External id": 98838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98838,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 98838, "pid": 5, "tid": 7, "ts": 1716454222942104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942089, "dur": 14,
    "args": {
      "External id": 98838,
      "cbid": 211, "correlation": 98838
    }
  },
  {
    "ph": "s", "id": 98838, "pid": 76337, "tid": -914061504, "ts": 1716454222942089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222942157, "dur": 1,
    "args": {
      "External id": 98848,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98848,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 98848, "pid": 5, "tid": 7, "ts": 1716454222942157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942142, "dur": 13,
    "args": {
      "External id": 98848,
      "cbid": 41, "correlation": 98848
    }
  },
  {
    "ph": "s", "id": 98848, "pid": 76337, "tid": -914061504, "ts": 1716454222942142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942156, "dur": 8,
    "args": {
      "External id": 98849,
      "cbid": 131, "correlation": 98849
    }
  },
  {
    "ph": "f", "id": 98849, "pid": 76337, "tid": -914061504, "ts": 1716454222942156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222942217, "dur": 3,
    "args": {
      "External id": 98856,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98856,
      "bytes": 98304, "memory bandwidth (GB/s)": 32.33684210526316
    }
  },
  {
    "ph": "f", "id": 98856, "pid": 5, "tid": 7, "ts": 1716454222942217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942198, "dur": 19,
    "args": {
      "External id": 98856,
      "cbid": 41, "correlation": 98856
    }
  },
  {
    "ph": "s", "id": 98856, "pid": 76337, "tid": -914061504, "ts": 1716454222942198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222942264, "dur": 1,
    "args": {
      "External id": 98867,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98867,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013297872340425532
    }
  },
  {
    "ph": "f", "id": 98867, "pid": 5, "tid": 7, "ts": 1716454222942264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942252, "dur": 10,
    "args": {
      "External id": 98867,
      "cbid": 41, "correlation": 98867
    }
  },
  {
    "ph": "s", "id": 98867, "pid": 76337, "tid": -914061504, "ts": 1716454222942252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942263, "dur": 7,
    "args": {
      "External id": 98868,
      "cbid": 131, "correlation": 98868
    }
  },
  {
    "ph": "f", "id": 98868, "pid": 76337, "tid": -914061504, "ts": 1716454222942263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222942312, "dur": 3,
    "args": {
      "External id": 98876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98876,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98876, "pid": 5, "tid": 7, "ts": 1716454222942312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942299, "dur": 13,
    "args": {
      "External id": 98876,
      "cbid": 211, "correlation": 98876
    }
  },
  {
    "ph": "s", "id": 98876, "pid": 76337, "tid": -914061504, "ts": 1716454222942299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222942342, "dur": 3,
    "args": {
      "External id": 98886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98886,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98886, "pid": 5, "tid": 7, "ts": 1716454222942342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942332, "dur": 8,
    "args": {
      "External id": 98886,
      "cbid": 211, "correlation": 98886
    }
  },
  {
    "ph": "s", "id": 98886, "pid": 76337, "tid": -914061504, "ts": 1716454222942332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222942363, "dur": 3,
    "args": {
      "External id": 98895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98895,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98895, "pid": 5, "tid": 7, "ts": 1716454222942363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942355, "dur": 7,
    "args": {
      "External id": 98895,
      "cbid": 211, "correlation": 98895
    }
  },
  {
    "ph": "s", "id": 98895, "pid": 76337, "tid": -914061504, "ts": 1716454222942355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222942434, "dur": 5,
    "args": {
      "External id": 98903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98903,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98903, "pid": 5, "tid": 7, "ts": 1716454222942434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942421, "dur": 13,
    "args": {
      "External id": 98903,
      "cbid": 211, "correlation": 98903
    }
  },
  {
    "ph": "s", "id": 98903, "pid": 76337, "tid": -914061504, "ts": 1716454222942421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222942474, "dur": 3,
    "args": {
      "External id": 98912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98912,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98912, "pid": 5, "tid": 7, "ts": 1716454222942474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942464, "dur": 9,
    "args": {
      "External id": 98912,
      "cbid": 211, "correlation": 98912
    }
  },
  {
    "ph": "s", "id": 98912, "pid": 76337, "tid": -914061504, "ts": 1716454222942464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222942497, "dur": 3,
    "args": {
      "External id": 98921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98921,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98921, "pid": 5, "tid": 7, "ts": 1716454222942497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942488, "dur": 7,
    "args": {
      "External id": 98921,
      "cbid": 211, "correlation": 98921
    }
  },
  {
    "ph": "s", "id": 98921, "pid": 76337, "tid": -914061504, "ts": 1716454222942488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222942562, "dur": 3,
    "args": {
      "External id": 98929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98929,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 98929, "pid": 5, "tid": 7, "ts": 1716454222942562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942551, "dur": 10,
    "args": {
      "External id": 98929,
      "cbid": 211, "correlation": 98929
    }
  },
  {
    "ph": "s", "id": 98929, "pid": 76337, "tid": -914061504, "ts": 1716454222942551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222942622, "dur": 1,
    "args": {
      "External id": 98937,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98937,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 98937, "pid": 5, "tid": 7, "ts": 1716454222942622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942605, "dur": 26,
    "args": {
      "External id": 98937,
      "cbid": 41, "correlation": 98937
    }
  },
  {
    "ph": "s", "id": 98937, "pid": 76337, "tid": -914061504, "ts": 1716454222942605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942633, "dur": 3,
    "args": {
      "External id": 98938,
      "cbid": 131, "correlation": 98938
    }
  },
  {
    "ph": "f", "id": 98938, "pid": 76337, "tid": -914061504, "ts": 1716454222942633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222942693, "dur": 1,
    "args": {
      "External id": 98948,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98948,
      "bytes": 42, "memory bandwidth (GB/s)": 0.0234375
    }
  },
  {
    "ph": "f", "id": 98948, "pid": 5, "tid": 7, "ts": 1716454222942693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942681, "dur": 10,
    "args": {
      "External id": 98948,
      "cbid": 41, "correlation": 98948
    }
  },
  {
    "ph": "s", "id": 98948, "pid": 76337, "tid": -914061504, "ts": 1716454222942681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942692, "dur": 8,
    "args": {
      "External id": 98949,
      "cbid": 131, "correlation": 98949
    }
  },
  {
    "ph": "f", "id": 98949, "pid": 76337, "tid": -914061504, "ts": 1716454222942692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222942752, "dur": 1,
    "args": {
      "External id": 98958,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98958,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 98958, "pid": 5, "tid": 7, "ts": 1716454222942752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942741, "dur": 8,
    "args": {
      "External id": 98958,
      "cbid": 41, "correlation": 98958
    }
  },
  {
    "ph": "s", "id": 98958, "pid": 76337, "tid": -914061504, "ts": 1716454222942741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942750, "dur": 8,
    "args": {
      "External id": 98959,
      "cbid": 131, "correlation": 98959
    }
  },
  {
    "ph": "f", "id": 98959, "pid": 76337, "tid": -914061504, "ts": 1716454222942750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222942827, "dur": 4,
    "args": {
      "External id": 98966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98966,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98966, "pid": 5, "tid": 7, "ts": 1716454222942827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942810, "dur": 17,
    "args": {
      "External id": 98966,
      "cbid": 211, "correlation": 98966
    }
  },
  {
    "ph": "s", "id": 98966, "pid": 76337, "tid": -914061504, "ts": 1716454222942810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454222942864, "dur": 4,
    "args": {
      "External id": 98986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 98986,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 98986, "pid": 5, "tid": 7, "ts": 1716454222942864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942853, "dur": 12,
    "args": {
      "External id": 98986,
      "cbid": 211, "correlation": 98986
    }
  },
  {
    "ph": "s", "id": 98986, "pid": 76337, "tid": -914061504, "ts": 1716454222942853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942866, "dur": 0,
    "args": {
      "External id": 98987,
      "cbid": 11, "correlation": 98987
    }
  },
  {
    "ph": "f", "id": 98987, "pid": 76337, "tid": -914061504, "ts": 1716454222942866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942866, "dur": 0,
    "args": {
      "External id": 98988,
      "cbid": 11, "correlation": 98988
    }
  },
  {
    "ph": "f", "id": 98988, "pid": 76337, "tid": -914061504, "ts": 1716454222942866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222942880, "dur": 1,
    "args": {
      "External id": 98991,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 98991,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0024509803921568627
    }
  },
  {
    "ph": "f", "id": 98991, "pid": 5, "tid": 7, "ts": 1716454222942880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942867, "dur": 21,
    "args": {
      "External id": 98991,
      "cbid": 41, "correlation": 98991
    }
  },
  {
    "ph": "s", "id": 98991, "pid": 76337, "tid": -914061504, "ts": 1716454222942867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942889, "dur": 3,
    "args": {
      "External id": 98992,
      "cbid": 131, "correlation": 98992
    }
  },
  {
    "ph": "f", "id": 98992, "pid": 76337, "tid": -914061504, "ts": 1716454222942889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222942917, "dur": 3,
    "args": {
      "External id": 99016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99016, "pid": 5, "tid": 7, "ts": 1716454222942917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942908, "dur": 9,
    "args": {
      "External id": 99016,
      "cbid": 211, "correlation": 99016
    }
  },
  {
    "ph": "s", "id": 99016, "pid": 76337, "tid": -914061504, "ts": 1716454222942908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942918, "dur": 0,
    "args": {
      "External id": 99017,
      "cbid": 11, "correlation": 99017
    }
  },
  {
    "ph": "f", "id": 99017, "pid": 76337, "tid": -914061504, "ts": 1716454222942918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942918, "dur": 0,
    "args": {
      "External id": 99018,
      "cbid": 11, "correlation": 99018
    }
  },
  {
    "ph": "f", "id": 99018, "pid": 76337, "tid": -914061504, "ts": 1716454222942918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942920, "dur": 1,
    "args": {
      "External id": 99020,
      "cbid": 200, "correlation": 99020
    }
  },
  {
    "ph": "f", "id": 99020, "pid": 76337, "tid": -914061504, "ts": 1716454222942920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454222942930, "dur": 4,
    "args": {
      "External id": 99022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99022,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99022, "pid": 5, "tid": 7, "ts": 1716454222942930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942923, "dur": 8,
    "args": {
      "External id": 99022,
      "cbid": 211, "correlation": 99022
    }
  },
  {
    "ph": "s", "id": 99022, "pid": 76337, "tid": -914061504, "ts": 1716454222942923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942931, "dur": 0,
    "args": {
      "External id": 99023,
      "cbid": 11, "correlation": 99023
    }
  },
  {
    "ph": "f", "id": 99023, "pid": 76337, "tid": -914061504, "ts": 1716454222942931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942932, "dur": 0,
    "args": {
      "External id": 99024,
      "cbid": 11, "correlation": 99024
    }
  },
  {
    "ph": "f", "id": 99024, "pid": 76337, "tid": -914061504, "ts": 1716454222942932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454222942970, "dur": 1,
    "args": {
      "External id": 99031,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 99031,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004901960784313725
    }
  },
  {
    "ph": "f", "id": 99031, "pid": 5, "tid": 7, "ts": 1716454222942970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942959, "dur": 25,
    "args": {
      "External id": 99031,
      "cbid": 41, "correlation": 99031
    }
  },
  {
    "ph": "s", "id": 99031, "pid": 76337, "tid": -914061504, "ts": 1716454222942959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222942985, "dur": 3,
    "args": {
      "External id": 99032,
      "cbid": 131, "correlation": 99032
    }
  },
  {
    "ph": "f", "id": 99032, "pid": 76337, "tid": -914061504, "ts": 1716454222942985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454222943036, "dur": 1,
    "args": {
      "External id": 99042,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 99042,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 99042, "pid": 5, "tid": 7, "ts": 1716454222943036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943024, "dur": 10,
    "args": {
      "External id": 99042,
      "cbid": 41, "correlation": 99042
    }
  },
  {
    "ph": "s", "id": 99042, "pid": 76337, "tid": -914061504, "ts": 1716454222943024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943035, "dur": 8,
    "args": {
      "External id": 99043,
      "cbid": 131, "correlation": 99043
    }
  },
  {
    "ph": "f", "id": 99043, "pid": 76337, "tid": -914061504, "ts": 1716454222943035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222943106, "dur": 5,
    "args": {
      "External id": 99050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99050,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99050, "pid": 5, "tid": 7, "ts": 1716454222943106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943091, "dur": 16,
    "args": {
      "External id": 99050,
      "cbid": 211, "correlation": 99050
    }
  },
  {
    "ph": "s", "id": 99050, "pid": 76337, "tid": -914061504, "ts": 1716454222943091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222943176, "dur": 3,
    "args": {
      "External id": 99059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99059,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99059, "pid": 5, "tid": 7, "ts": 1716454222943176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943163, "dur": 12,
    "args": {
      "External id": 99059,
      "cbid": 211, "correlation": 99059
    }
  },
  {
    "ph": "s", "id": 99059, "pid": 76337, "tid": -914061504, "ts": 1716454222943163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222943213, "dur": 3,
    "args": {
      "External id": 99067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99067,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99067, "pid": 5, "tid": 7, "ts": 1716454222943213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943202, "dur": 10,
    "args": {
      "External id": 99067,
      "cbid": 211, "correlation": 99067
    }
  },
  {
    "ph": "s", "id": 99067, "pid": 76337, "tid": -914061504, "ts": 1716454222943202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222943247, "dur": 4,
    "args": {
      "External id": 99075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99075,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99075, "pid": 5, "tid": 7, "ts": 1716454222943247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943235, "dur": 12,
    "args": {
      "External id": 99075,
      "cbid": 211, "correlation": 99075
    }
  },
  {
    "ph": "s", "id": 99075, "pid": 76337, "tid": -914061504, "ts": 1716454222943235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222943276, "dur": 4,
    "args": {
      "External id": 99083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99083,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99083, "pid": 5, "tid": 7, "ts": 1716454222943276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943266, "dur": 10,
    "args": {
      "External id": 99083,
      "cbid": 211, "correlation": 99083
    }
  },
  {
    "ph": "s", "id": 99083, "pid": 76337, "tid": -914061504, "ts": 1716454222943266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222943303, "dur": 3,
    "args": {
      "External id": 99091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99091,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99091, "pid": 5, "tid": 7, "ts": 1716454222943303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943294, "dur": 8,
    "args": {
      "External id": 99091,
      "cbid": 211, "correlation": 99091
    }
  },
  {
    "ph": "s", "id": 99091, "pid": 76337, "tid": -914061504, "ts": 1716454222943294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222943328, "dur": 3,
    "args": {
      "External id": 99099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99099,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99099, "pid": 5, "tid": 7, "ts": 1716454222943328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943319, "dur": 9,
    "args": {
      "External id": 99099,
      "cbid": 211, "correlation": 99099
    }
  },
  {
    "ph": "s", "id": 99099, "pid": 76337, "tid": -914061504, "ts": 1716454222943319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222943351, "dur": 4,
    "args": {
      "External id": 99107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99107, "pid": 5, "tid": 7, "ts": 1716454222943351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943343, "dur": 7,
    "args": {
      "External id": 99107,
      "cbid": 211, "correlation": 99107
    }
  },
  {
    "ph": "s", "id": 99107, "pid": 76337, "tid": -914061504, "ts": 1716454222943343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222943369, "dur": 4,
    "args": {
      "External id": 99115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99115,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99115, "pid": 5, "tid": 7, "ts": 1716454222943369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943361, "dur": 7,
    "args": {
      "External id": 99115,
      "cbid": 211, "correlation": 99115
    }
  },
  {
    "ph": "s", "id": 99115, "pid": 76337, "tid": -914061504, "ts": 1716454222943361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222943388, "dur": 3,
    "args": {
      "External id": 99123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99123,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99123, "pid": 5, "tid": 7, "ts": 1716454222943388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943380, "dur": 7,
    "args": {
      "External id": 99123,
      "cbid": 211, "correlation": 99123
    }
  },
  {
    "ph": "s", "id": 99123, "pid": 76337, "tid": -914061504, "ts": 1716454222943380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222943446, "dur": 3,
    "args": {
      "External id": 99131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99131,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99131, "pid": 5, "tid": 7, "ts": 1716454222943446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943435, "dur": 11,
    "args": {
      "External id": 99131,
      "cbid": 211, "correlation": 99131
    }
  },
  {
    "ph": "s", "id": 99131, "pid": 76337, "tid": -914061504, "ts": 1716454222943435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222943470, "dur": 4,
    "args": {
      "External id": 99139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99139,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99139, "pid": 5, "tid": 7, "ts": 1716454222943470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943462, "dur": 8,
    "args": {
      "External id": 99139,
      "cbid": 211, "correlation": 99139
    }
  },
  {
    "ph": "s", "id": 99139, "pid": 76337, "tid": -914061504, "ts": 1716454222943462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222943494, "dur": 4,
    "args": {
      "External id": 99147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99147,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99147, "pid": 5, "tid": 7, "ts": 1716454222943494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943484, "dur": 8,
    "args": {
      "External id": 99147,
      "cbid": 211, "correlation": 99147
    }
  },
  {
    "ph": "s", "id": 99147, "pid": 76337, "tid": -914061504, "ts": 1716454222943484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222943512, "dur": 3,
    "args": {
      "External id": 99155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99155,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99155, "pid": 5, "tid": 7, "ts": 1716454222943512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943505, "dur": 6,
    "args": {
      "External id": 99155,
      "cbid": 211, "correlation": 99155
    }
  },
  {
    "ph": "s", "id": 99155, "pid": 76337, "tid": -914061504, "ts": 1716454222943505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222943920, "dur": 5,
    "args": {
      "External id": 99164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99164,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99164, "pid": 5, "tid": 7, "ts": 1716454222943920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943904, "dur": 17,
    "args": {
      "External id": 99164,
      "cbid": 211, "correlation": 99164
    }
  },
  {
    "ph": "s", "id": 99164, "pid": 76337, "tid": -914061504, "ts": 1716454222943904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222943957, "dur": 5,
    "args": {
      "External id": 99173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99173,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99173, "pid": 5, "tid": 7, "ts": 1716454222943957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222943947, "dur": 9,
    "args": {
      "External id": 99173,
      "cbid": 211, "correlation": 99173
    }
  },
  {
    "ph": "s", "id": 99173, "pid": 76337, "tid": -914061504, "ts": 1716454222943947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454222944100, "dur": 2,
    "args": {
      "External id": 99189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99189, "pid": 5, "tid": 7, "ts": 1716454222944100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944085, "dur": 15,
    "args": {
      "External id": 99189,
      "cbid": 211, "correlation": 99189
    }
  },
  {
    "ph": "s", "id": 99189, "pid": 76337, "tid": -914061504, "ts": 1716454222944085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944133, "dur": 3,
    "args": {
      "External id": 99197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99197, "pid": 5, "tid": 7, "ts": 1716454222944133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944124, "dur": 8,
    "args": {
      "External id": 99197,
      "cbid": 211, "correlation": 99197
    }
  },
  {
    "ph": "s", "id": 99197, "pid": 76337, "tid": -914061504, "ts": 1716454222944124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944165, "dur": 3,
    "args": {
      "External id": 99205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99205, "pid": 5, "tid": 7, "ts": 1716454222944165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944155, "dur": 10,
    "args": {
      "External id": 99205,
      "cbid": 211, "correlation": 99205
    }
  },
  {
    "ph": "s", "id": 99205, "pid": 76337, "tid": -914061504, "ts": 1716454222944155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944195, "dur": 4,
    "args": {
      "External id": 99213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99213,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99213, "pid": 5, "tid": 7, "ts": 1716454222944195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944186, "dur": 8,
    "args": {
      "External id": 99213,
      "cbid": 211, "correlation": 99213
    }
  },
  {
    "ph": "s", "id": 99213, "pid": 76337, "tid": -914061504, "ts": 1716454222944186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454222944251, "dur": 4,
    "args": {
      "External id": 99225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99225,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99225, "pid": 5, "tid": 7, "ts": 1716454222944251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944238, "dur": 13,
    "args": {
      "External id": 99225,
      "cbid": 211, "correlation": 99225
    }
  },
  {
    "ph": "s", "id": 99225, "pid": 76337, "tid": -914061504, "ts": 1716454222944238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222944297, "dur": 4,
    "args": {
      "External id": 99236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99236, "pid": 5, "tid": 7, "ts": 1716454222944297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944285, "dur": 11,
    "args": {
      "External id": 99236,
      "cbid": 211, "correlation": 99236
    }
  },
  {
    "ph": "s", "id": 99236, "pid": 76337, "tid": -914061504, "ts": 1716454222944285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944329, "dur": 2,
    "args": {
      "External id": 99244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99244, "pid": 5, "tid": 7, "ts": 1716454222944329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944319, "dur": 8,
    "args": {
      "External id": 99244,
      "cbid": 211, "correlation": 99244
    }
  },
  {
    "ph": "s", "id": 99244, "pid": 76337, "tid": -914061504, "ts": 1716454222944319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944362, "dur": 5,
    "args": {
      "External id": 99252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99252,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99252, "pid": 5, "tid": 7, "ts": 1716454222944362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944351, "dur": 10,
    "args": {
      "External id": 99252,
      "cbid": 211, "correlation": 99252
    }
  },
  {
    "ph": "s", "id": 99252, "pid": 76337, "tid": -914061504, "ts": 1716454222944351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944391, "dur": 5,
    "args": {
      "External id": 99260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99260,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99260, "pid": 5, "tid": 7, "ts": 1716454222944391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944380, "dur": 10,
    "args": {
      "External id": 99260,
      "cbid": 211, "correlation": 99260
    }
  },
  {
    "ph": "s", "id": 99260, "pid": 76337, "tid": -914061504, "ts": 1716454222944380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222944421, "dur": 4,
    "args": {
      "External id": 99269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99269,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99269, "pid": 5, "tid": 7, "ts": 1716454222944421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944411, "dur": 9,
    "args": {
      "External id": 99269,
      "cbid": 211, "correlation": 99269
    }
  },
  {
    "ph": "s", "id": 99269, "pid": 76337, "tid": -914061504, "ts": 1716454222944411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222944481, "dur": 4,
    "args": {
      "External id": 99282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99282,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99282, "pid": 5, "tid": 7, "ts": 1716454222944481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944469, "dur": 13,
    "args": {
      "External id": 99282,
      "cbid": 211, "correlation": 99282
    }
  },
  {
    "ph": "s", "id": 99282, "pid": 76337, "tid": -914061504, "ts": 1716454222944469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454222944523, "dur": 5,
    "args": {
      "External id": 99292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99292,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99292, "pid": 5, "tid": 7, "ts": 1716454222944523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944511, "dur": 11,
    "args": {
      "External id": 99292,
      "cbid": 211, "correlation": 99292
    }
  },
  {
    "ph": "s", "id": 99292, "pid": 76337, "tid": -914061504, "ts": 1716454222944511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944657, "dur": 5,
    "args": {
      "External id": 99309,
      "cbid": 251, "correlation": 99309
    }
  },
  {
    "ph": "f", "id": 99309, "pid": 76337, "tid": -914061504, "ts": 1716454222944657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454222944688, "dur": 12,
    "args": {
      "External id": 99311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99311,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99311, "pid": 5, "tid": 7, "ts": 1716454222944688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944672, "dur": 17,
    "args": {
      "External id": 99311,
      "cbid": 211, "correlation": 99311
    }
  },
  {
    "ph": "s", "id": 99311, "pid": 76337, "tid": -914061504, "ts": 1716454222944672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222944754, "dur": 4,
    "args": {
      "External id": 99319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99319,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99319, "pid": 5, "tid": 7, "ts": 1716454222944754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944742, "dur": 12,
    "args": {
      "External id": 99319,
      "cbid": 211, "correlation": 99319
    }
  },
  {
    "ph": "s", "id": 99319, "pid": 76337, "tid": -914061504, "ts": 1716454222944742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944814, "dur": 2,
    "args": {
      "External id": 99335,
      "cbid": 251, "correlation": 99335
    }
  },
  {
    "ph": "f", "id": 99335, "pid": 76337, "tid": -914061504, "ts": 1716454222944814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944820, "dur": 0,
    "args": {
      "External id": 99337,
      "cbid": 251, "correlation": 99337
    }
  },
  {
    "ph": "f", "id": 99337, "pid": 76337, "tid": -914061504, "ts": 1716454222944820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222944838, "dur": 13,
    "args": {
      "External id": 99338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99338,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99338, "pid": 5, "tid": 7, "ts": 1716454222944838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944823, "dur": 15,
    "args": {
      "External id": 99338,
      "cbid": 211, "correlation": 99338
    }
  },
  {
    "ph": "s", "id": 99338, "pid": 76337, "tid": -914061504, "ts": 1716454222944823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222944853, "dur": 5,
    "args": {
      "External id": 99340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99340,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99340, "pid": 5, "tid": 7, "ts": 1716454222944853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944843, "dur": 9,
    "args": {
      "External id": 99340,
      "cbid": 211, "correlation": 99340
    }
  },
  {
    "ph": "s", "id": 99340, "pid": 76337, "tid": -914061504, "ts": 1716454222944843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944957, "dur": 1,
    "args": {
      "External id": 99350,
      "cbid": 317, "correlation": 99350
    }
  },
  {
    "ph": "f", "id": 99350, "pid": 76337, "tid": -914061504, "ts": 1716454222944957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944959, "dur": 1,
    "args": {
      "External id": 99351,
      "cbid": 203, "correlation": 99351
    }
  },
  {
    "ph": "f", "id": 99351, "pid": 76337, "tid": -914061504, "ts": 1716454222944959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222944961, "dur": 1,
    "args": {
      "External id": 99352,
      "cbid": 205, "correlation": 99352
    }
  },
  {
    "ph": "f", "id": 99352, "pid": 76337, "tid": -914061504, "ts": 1716454222944961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222945030, "dur": 6,
    "args": {
      "External id": 99356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99356, "pid": 5, "tid": 7, "ts": 1716454222945030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222945014, "dur": 16,
    "args": {
      "External id": 99356,
      "cbid": 211, "correlation": 99356
    }
  },
  {
    "ph": "s", "id": 99356, "pid": 76337, "tid": -914061504, "ts": 1716454222945014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222945041, "dur": 4,
    "args": {
      "External id": 99358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99358,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 99358, "pid": 5, "tid": 7, "ts": 1716454222945041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222945033, "dur": 6,
    "args": {
      "External id": 99358,
      "cbid": 211, "correlation": 99358
    }
  },
  {
    "ph": "s", "id": 99358, "pid": 76337, "tid": -914061504, "ts": 1716454222945033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222945062, "dur": 3,
    "args": {
      "External id": 99360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99360, "pid": 5, "tid": 7, "ts": 1716454222945062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222945052, "dur": 8,
    "args": {
      "External id": 99360,
      "cbid": 211, "correlation": 99360
    }
  },
  {
    "ph": "s", "id": 99360, "pid": 76337, "tid": -914061504, "ts": 1716454222945052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222945067, "dur": 0,
    "args": {
      "External id": 99361,
      "cbid": 51, "correlation": 99361
    }
  },
  {
    "ph": "s", "id": 99361, "pid": 76337, "tid": -914061504, "ts": 1716454222945067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222945078, "dur": 82,
    "args": {
      "External id": 99362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99362,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99362, "pid": 5, "tid": 7, "ts": 1716454222945078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222945069, "dur": 7,
    "args": {
      "External id": 99362,
      "cbid": 211, "correlation": 99362
    }
  },
  {
    "ph": "s", "id": 99362, "pid": 76337, "tid": -914061504, "ts": 1716454222945069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222945161, "dur": 59,
    "args": {
      "External id": 99367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99367, "pid": 5, "tid": 7, "ts": 1716454222945161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222945106, "dur": 11,
    "args": {
      "External id": 99367,
      "cbid": 211, "correlation": 99367
    }
  },
  {
    "ph": "s", "id": 99367, "pid": 76337, "tid": -914061504, "ts": 1716454222945106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222946949, "dur": 51,
    "args": {
      "External id": 99387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99387,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 99387, "pid": 5, "tid": 7, "ts": 1716454222946949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222946932, "dur": 16,
    "args": {
      "External id": 99387,
      "cbid": 211, "correlation": 99387
    }
  },
  {
    "ph": "s", "id": 99387, "pid": 76337, "tid": -914061504, "ts": 1716454222946932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222947001, "dur": 4,
    "args": {
      "External id": 99399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99399,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99399, "pid": 5, "tid": 7, "ts": 1716454222947001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222946961, "dur": 8,
    "args": {
      "External id": 99399,
      "cbid": 211, "correlation": 99399
    }
  },
  {
    "ph": "s", "id": 99399, "pid": 76337, "tid": -914061504, "ts": 1716454222946961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222947006, "dur": 55,
    "args": {
      "External id": 99402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99402,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99402, "pid": 5, "tid": 7, "ts": 1716454222947006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222946992, "dur": 10,
    "args": {
      "External id": 99402,
      "cbid": 211, "correlation": 99402
    }
  },
  {
    "ph": "s", "id": 99402, "pid": 76337, "tid": -914061504, "ts": 1716454222946992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222947063, "dur": 36,
    "args": {
      "External id": 99411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99411,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99411, "pid": 5, "tid": 7, "ts": 1716454222947063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947039, "dur": 11,
    "args": {
      "External id": 99411,
      "cbid": 211, "correlation": 99411
    }
  },
  {
    "ph": "s", "id": 99411, "pid": 76337, "tid": -914061504, "ts": 1716454222947039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947100, "dur": 0,
    "args": {
      "External id": 99421,
      "cbid": 317, "correlation": 99421
    }
  },
  {
    "ph": "f", "id": 99421, "pid": 76337, "tid": -914061504, "ts": 1716454222947100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947100, "dur": 0,
    "args": {
      "External id": 99422,
      "cbid": 203, "correlation": 99422
    }
  },
  {
    "ph": "f", "id": 99422, "pid": 76337, "tid": -914061504, "ts": 1716454222947100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947101, "dur": 0,
    "args": {
      "External id": 99423,
      "cbid": 205, "correlation": 99423
    }
  },
  {
    "ph": "f", "id": 99423, "pid": 76337, "tid": -914061504, "ts": 1716454222947101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222947133, "dur": 39,
    "args": {
      "External id": 99427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99427,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99427, "pid": 5, "tid": 7, "ts": 1716454222947133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947120, "dur": 12,
    "args": {
      "External id": 99427,
      "cbid": 211, "correlation": 99427
    }
  },
  {
    "ph": "s", "id": 99427, "pid": 76337, "tid": -914061504, "ts": 1716454222947120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222947173, "dur": 14,
    "args": {
      "External id": 99429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99429,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99429, "pid": 5, "tid": 7, "ts": 1716454222947173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947135, "dur": 6,
    "args": {
      "External id": 99429,
      "cbid": 211, "correlation": 99429
    }
  },
  {
    "ph": "s", "id": 99429, "pid": 76337, "tid": -914061504, "ts": 1716454222947135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222947188, "dur": 3,
    "args": {
      "External id": 99431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99431,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99431, "pid": 5, "tid": 7, "ts": 1716454222947188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947148, "dur": 6,
    "args": {
      "External id": 99431,
      "cbid": 211, "correlation": 99431
    }
  },
  {
    "ph": "s", "id": 99431, "pid": 76337, "tid": -914061504, "ts": 1716454222947148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947158, "dur": 0,
    "args": {
      "External id": 99432,
      "cbid": 51, "correlation": 99432
    }
  },
  {
    "ph": "s", "id": 99432, "pid": 76337, "tid": -914061504, "ts": 1716454222947158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222947193, "dur": 688,
    "args": {
      "External id": 99433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99433,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99433, "pid": 5, "tid": 7, "ts": 1716454222947193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947159, "dur": 6,
    "args": {
      "External id": 99433,
      "cbid": 211, "correlation": 99433
    }
  },
  {
    "ph": "s", "id": 99433, "pid": 76337, "tid": -914061504, "ts": 1716454222947159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222947882, "dur": 58,
    "args": {
      "External id": 99438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99438,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99438, "pid": 5, "tid": 7, "ts": 1716454222947882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947189, "dur": 10,
    "args": {
      "External id": 99438,
      "cbid": 211, "correlation": 99438
    }
  },
  {
    "ph": "s", "id": 99438, "pid": 76337, "tid": -914061504, "ts": 1716454222947189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222947941, "dur": 3,
    "args": {
      "External id": 99446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99446,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99446, "pid": 5, "tid": 7, "ts": 1716454222947941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947233, "dur": 9,
    "args": {
      "External id": 99446,
      "cbid": 211, "correlation": 99446
    }
  },
  {
    "ph": "s", "id": 99446, "pid": 76337, "tid": -914061504, "ts": 1716454222947233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947301, "dur": 1,
    "args": {
      "External id": 99462,
      "cbid": 251, "correlation": 99462
    }
  },
  {
    "ph": "f", "id": 99462, "pid": 76337, "tid": -914061504, "ts": 1716454222947301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947306, "dur": 0,
    "args": {
      "External id": 99464,
      "cbid": 251, "correlation": 99464
    }
  },
  {
    "ph": "f", "id": 99464, "pid": 76337, "tid": -914061504, "ts": 1716454222947306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222947946, "dur": 8,
    "args": {
      "External id": 99465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99465,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 99465, "pid": 5, "tid": 7, "ts": 1716454222947946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947308, "dur": 12,
    "args": {
      "External id": 99465,
      "cbid": 211, "correlation": 99465
    }
  },
  {
    "ph": "s", "id": 99465, "pid": 76337, "tid": -914061504, "ts": 1716454222947308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222947955, "dur": 4,
    "args": {
      "External id": 99467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99467,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 99467, "pid": 5, "tid": 7, "ts": 1716454222947955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947322, "dur": 7,
    "args": {
      "External id": 99467,
      "cbid": 211, "correlation": 99467
    }
  },
  {
    "ph": "s", "id": 99467, "pid": 76337, "tid": -914061504, "ts": 1716454222947322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222947961, "dur": 54,
    "args": {
      "External id": 99477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99477, "pid": 5, "tid": 7, "ts": 1716454222947961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947383, "dur": 13,
    "args": {
      "External id": 99477,
      "cbid": 211, "correlation": 99477
    }
  },
  {
    "ph": "s", "id": 99477, "pid": 76337, "tid": -914061504, "ts": 1716454222947383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222948016, "dur": 52,
    "args": {
      "External id": 99497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99497,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 99497, "pid": 5, "tid": 7, "ts": 1716454222948016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947451, "dur": 11,
    "args": {
      "External id": 99497,
      "cbid": 211, "correlation": 99497
    }
  },
  {
    "ph": "s", "id": 99497, "pid": 76337, "tid": -914061504, "ts": 1716454222947451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222948069, "dur": 4,
    "args": {
      "External id": 99509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99509,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99509, "pid": 5, "tid": 7, "ts": 1716454222948069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947472, "dur": 6,
    "args": {
      "External id": 99509,
      "cbid": 211, "correlation": 99509
    }
  },
  {
    "ph": "s", "id": 99509, "pid": 76337, "tid": -914061504, "ts": 1716454222947472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222948074, "dur": 54,
    "args": {
      "External id": 99512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99512,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99512, "pid": 5, "tid": 7, "ts": 1716454222948074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947490, "dur": 7,
    "args": {
      "External id": 99512,
      "cbid": 211, "correlation": 99512
    }
  },
  {
    "ph": "s", "id": 99512, "pid": 76337, "tid": -914061504, "ts": 1716454222947490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222948130, "dur": 37,
    "args": {
      "External id": 99521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99521,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99521, "pid": 5, "tid": 7, "ts": 1716454222948130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947530, "dur": 10,
    "args": {
      "External id": 99521,
      "cbid": 211, "correlation": 99521
    }
  },
  {
    "ph": "s", "id": 99521, "pid": 76337, "tid": -914061504, "ts": 1716454222947530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947602, "dur": 0,
    "args": {
      "External id": 99531,
      "cbid": 317, "correlation": 99531
    }
  },
  {
    "ph": "f", "id": 99531, "pid": 76337, "tid": -914061504, "ts": 1716454222947602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947603, "dur": 0,
    "args": {
      "External id": 99532,
      "cbid": 203, "correlation": 99532
    }
  },
  {
    "ph": "f", "id": 99532, "pid": 76337, "tid": -914061504, "ts": 1716454222947603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947604, "dur": 0,
    "args": {
      "External id": 99533,
      "cbid": 205, "correlation": 99533
    }
  },
  {
    "ph": "f", "id": 99533, "pid": 76337, "tid": -914061504, "ts": 1716454222947604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222948168, "dur": 42,
    "args": {
      "External id": 99537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99537,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99537, "pid": 5, "tid": 7, "ts": 1716454222948168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947619, "dur": 12,
    "args": {
      "External id": 99537,
      "cbid": 211, "correlation": 99537
    }
  },
  {
    "ph": "s", "id": 99537, "pid": 76337, "tid": -914061504, "ts": 1716454222947619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222948211, "dur": 14,
    "args": {
      "External id": 99539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99539,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99539, "pid": 5, "tid": 7, "ts": 1716454222948211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947633, "dur": 5,
    "args": {
      "External id": 99539,
      "cbid": 211, "correlation": 99539
    }
  },
  {
    "ph": "s", "id": 99539, "pid": 76337, "tid": -914061504, "ts": 1716454222947633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222948227, "dur": 3,
    "args": {
      "External id": 99541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99541, "pid": 5, "tid": 7, "ts": 1716454222948227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947643, "dur": 6,
    "args": {
      "External id": 99541,
      "cbid": 211, "correlation": 99541
    }
  },
  {
    "ph": "s", "id": 99541, "pid": 76337, "tid": -914061504, "ts": 1716454222947643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947652, "dur": 0,
    "args": {
      "External id": 99542,
      "cbid": 51, "correlation": 99542
    }
  },
  {
    "ph": "s", "id": 99542, "pid": 76337, "tid": -914061504, "ts": 1716454222947652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222948231, "dur": 684,
    "args": {
      "External id": 99543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99543,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99543, "pid": 5, "tid": 7, "ts": 1716454222948231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947653, "dur": 5,
    "args": {
      "External id": 99543,
      "cbid": 211, "correlation": 99543
    }
  },
  {
    "ph": "s", "id": 99543, "pid": 76337, "tid": -914061504, "ts": 1716454222947653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222948916, "dur": 58,
    "args": {
      "External id": 99548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99548, "pid": 5, "tid": 7, "ts": 1716454222948916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947681, "dur": 9,
    "args": {
      "External id": 99548,
      "cbid": 211, "correlation": 99548
    }
  },
  {
    "ph": "s", "id": 99548, "pid": 76337, "tid": -914061504, "ts": 1716454222947681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222948976, "dur": 50,
    "args": {
      "External id": 99556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99556,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99556, "pid": 5, "tid": 7, "ts": 1716454222948976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947714, "dur": 8,
    "args": {
      "External id": 99556,
      "cbid": 211, "correlation": 99556
    }
  },
  {
    "ph": "s", "id": 99556, "pid": 76337, "tid": -914061504, "ts": 1716454222947714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222949028, "dur": 35,
    "args": {
      "External id": 99564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99564,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99564, "pid": 5, "tid": 7, "ts": 1716454222949028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947746, "dur": 10,
    "args": {
      "External id": 99564,
      "cbid": 211, "correlation": 99564
    }
  },
  {
    "ph": "s", "id": 99564, "pid": 76337, "tid": -914061504, "ts": 1716454222947746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222949064, "dur": 51,
    "args": {
      "External id": 99584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99584,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 99584, "pid": 5, "tid": 7, "ts": 1716454222949064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947831, "dur": 13,
    "args": {
      "External id": 99584,
      "cbid": 211, "correlation": 99584
    }
  },
  {
    "ph": "s", "id": 99584, "pid": 76337, "tid": -914061504, "ts": 1716454222947831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222949116, "dur": 4,
    "args": {
      "External id": 99596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99596,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 99596, "pid": 5, "tid": 7, "ts": 1716454222949116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947854, "dur": 6,
    "args": {
      "External id": 99596,
      "cbid": 211, "correlation": 99596
    }
  },
  {
    "ph": "s", "id": 99596, "pid": 76337, "tid": -914061504, "ts": 1716454222947854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222949121, "dur": 54,
    "args": {
      "External id": 99599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99599, "pid": 5, "tid": 7, "ts": 1716454222949121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947872, "dur": 7,
    "args": {
      "External id": 99599,
      "cbid": 211, "correlation": 99599
    }
  },
  {
    "ph": "s", "id": 99599, "pid": 76337, "tid": -914061504, "ts": 1716454222947872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947931, "dur": 0,
    "args": {
      "External id": 99610,
      "cbid": 317, "correlation": 99610
    }
  },
  {
    "ph": "f", "id": 99610, "pid": 76337, "tid": -914061504, "ts": 1716454222947931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947932, "dur": 0,
    "args": {
      "External id": 99611,
      "cbid": 203, "correlation": 99611
    }
  },
  {
    "ph": "f", "id": 99611, "pid": 76337, "tid": -914061504, "ts": 1716454222947932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947933, "dur": 0,
    "args": {
      "External id": 99612,
      "cbid": 205, "correlation": 99612
    }
  },
  {
    "ph": "f", "id": 99612, "pid": 76337, "tid": -914061504, "ts": 1716454222947933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947968, "dur": 2,
    "args": {
      "External id": 99616,
      "cbid": 251, "correlation": 99616
    }
  },
  {
    "ph": "f", "id": 99616, "pid": 76337, "tid": -914061504, "ts": 1716454222947968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947972, "dur": 9,
    "args": {
      "External id": 99617,
      "cbid": 251, "correlation": 99617
    }
  },
  {
    "ph": "f", "id": 99617, "pid": 76337, "tid": -914061504, "ts": 1716454222947972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947982, "dur": 1,
    "args": {
      "External id": 99618,
      "cbid": 251, "correlation": 99618
    }
  },
  {
    "ph": "f", "id": 99618, "pid": 76337, "tid": -914061504, "ts": 1716454222947982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947984, "dur": 1,
    "args": {
      "External id": 99619,
      "cbid": 251, "correlation": 99619
    }
  },
  {
    "ph": "f", "id": 99619, "pid": 76337, "tid": -914061504, "ts": 1716454222947984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947986, "dur": 1,
    "args": {
      "External id": 99620,
      "cbid": 251, "correlation": 99620
    }
  },
  {
    "ph": "f", "id": 99620, "pid": 76337, "tid": -914061504, "ts": 1716454222947986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947988, "dur": 1,
    "args": {
      "External id": 99621,
      "cbid": 251, "correlation": 99621
    }
  },
  {
    "ph": "f", "id": 99621, "pid": 76337, "tid": -914061504, "ts": 1716454222947988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947990, "dur": 1,
    "args": {
      "External id": 99622,
      "cbid": 251, "correlation": 99622
    }
  },
  {
    "ph": "f", "id": 99622, "pid": 76337, "tid": -914061504, "ts": 1716454222947990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947991, "dur": 1,
    "args": {
      "External id": 99623,
      "cbid": 251, "correlation": 99623
    }
  },
  {
    "ph": "f", "id": 99623, "pid": 76337, "tid": -914061504, "ts": 1716454222947991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947994, "dur": 0,
    "args": {
      "External id": 99624,
      "cbid": 251, "correlation": 99624
    }
  },
  {
    "ph": "f", "id": 99624, "pid": 76337, "tid": -914061504, "ts": 1716454222947994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222949177, "dur": 115,
    "args": {
      "External id": 99625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99625,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 99625, "pid": 5, "tid": 7, "ts": 1716454222949177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222947998, "dur": 14,
    "args": {
      "External id": 99625,
      "cbid": 211, "correlation": 99625
    }
  },
  {
    "ph": "s", "id": 99625, "pid": 76337, "tid": -914061504, "ts": 1716454222947998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222949293, "dur": 59,
    "args": {
      "External id": 99631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99631,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99631, "pid": 5, "tid": 7, "ts": 1716454222949293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948036, "dur": 10,
    "args": {
      "External id": 99631,
      "cbid": 211, "correlation": 99631
    }
  },
  {
    "ph": "s", "id": 99631, "pid": 76337, "tid": -914061504, "ts": 1716454222948036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222949353, "dur": 526,
    "args": {
      "External id": 99640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99640,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99640, "pid": 5, "tid": 7, "ts": 1716454222949353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948133, "dur": 16,
    "args": {
      "External id": 99640,
      "cbid": 211, "correlation": 99640
    }
  },
  {
    "ph": "s", "id": 99640, "pid": 76337, "tid": -914061504, "ts": 1716454222948133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222949880, "dur": 177,
    "args": {
      "External id": 99662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99662,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99662, "pid": 5, "tid": 7, "ts": 1716454222949880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948207, "dur": 12,
    "args": {
      "External id": 99662,
      "cbid": 211, "correlation": 99662
    }
  },
  {
    "ph": "s", "id": 99662, "pid": 76337, "tid": -914061504, "ts": 1716454222948207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948325, "dur": 2,
    "args": {
      "External id": 99673,
      "cbid": 251, "correlation": 99673
    }
  },
  {
    "ph": "f", "id": 99673, "pid": 76337, "tid": -914061504, "ts": 1716454222948325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222950058, "dur": 195,
    "args": {
      "External id": 99674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99674,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99674, "pid": 5, "tid": 7, "ts": 1716454222950058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948332, "dur": 14,
    "args": {
      "External id": 99674,
      "cbid": 211, "correlation": 99674
    }
  },
  {
    "ph": "s", "id": 99674, "pid": 76337, "tid": -914061504, "ts": 1716454222948332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948406, "dur": 1,
    "args": {
      "External id": 99685,
      "cbid": 251, "correlation": 99685
    }
  },
  {
    "ph": "f", "id": 99685, "pid": 76337, "tid": -914061504, "ts": 1716454222948406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222950254, "dur": 186,
    "args": {
      "External id": 99686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99686,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99686, "pid": 5, "tid": 7, "ts": 1716454222950254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948410, "dur": 12,
    "args": {
      "External id": 99686,
      "cbid": 211, "correlation": 99686
    }
  },
  {
    "ph": "s", "id": 99686, "pid": 76337, "tid": -914061504, "ts": 1716454222948410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948476, "dur": 1,
    "args": {
      "External id": 99697,
      "cbid": 251, "correlation": 99697
    }
  },
  {
    "ph": "f", "id": 99697, "pid": 76337, "tid": -914061504, "ts": 1716454222948476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222950441, "dur": 185,
    "args": {
      "External id": 99698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99698,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99698, "pid": 5, "tid": 7, "ts": 1716454222950441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948480, "dur": 11,
    "args": {
      "External id": 99698,
      "cbid": 211, "correlation": 99698
    }
  },
  {
    "ph": "s", "id": 99698, "pid": 76337, "tid": -914061504, "ts": 1716454222948480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222950628, "dur": 18157,
    "args": {
      "External id": 99719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99719,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 99719, "pid": 5, "tid": 7, "ts": 1716454222950628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948591, "dur": 15,
    "args": {
      "External id": 99719,
      "cbid": 211, "correlation": 99719
    }
  },
  {
    "ph": "s", "id": 99719, "pid": 76337, "tid": -914061504, "ts": 1716454222948591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948706, "dur": 2,
    "args": {
      "External id": 99737,
      "cbid": 251, "correlation": 99737
    }
  },
  {
    "ph": "f", "id": 99737, "pid": 76337, "tid": -914061504, "ts": 1716454222948706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222968786, "dur": 198,
    "args": {
      "External id": 99739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99739,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99739, "pid": 5, "tid": 7, "ts": 1716454222968786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948714, "dur": 14,
    "args": {
      "External id": 99739,
      "cbid": 211, "correlation": 99739
    }
  },
  {
    "ph": "s", "id": 99739, "pid": 76337, "tid": -914061504, "ts": 1716454222948714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222968985, "dur": 66,
    "args": {
      "External id": 99747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99747,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99747, "pid": 5, "tid": 7, "ts": 1716454222968985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948788, "dur": 12,
    "args": {
      "External id": 99747,
      "cbid": 211, "correlation": 99747
    }
  },
  {
    "ph": "s", "id": 99747, "pid": 76337, "tid": -914061504, "ts": 1716454222948788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222969053, "dur": 97,
    "args": {
      "External id": 99755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99755,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99755, "pid": 5, "tid": 7, "ts": 1716454222969053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948866, "dur": 10,
    "args": {
      "External id": 99755,
      "cbid": 211, "correlation": 99755
    }
  },
  {
    "ph": "s", "id": 99755, "pid": 76337, "tid": -914061504, "ts": 1716454222948866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222969151, "dur": 53,
    "args": {
      "External id": 99766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99766,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99766, "pid": 5, "tid": 7, "ts": 1716454222969151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948955, "dur": 15,
    "args": {
      "External id": 99766,
      "cbid": 211, "correlation": 99766
    }
  },
  {
    "ph": "s", "id": 99766, "pid": 76337, "tid": -914061504, "ts": 1716454222948955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222969206, "dur": 90,
    "args": {
      "External id": 99788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99788,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99788, "pid": 5, "tid": 7, "ts": 1716454222969206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222948996, "dur": 9,
    "args": {
      "External id": 99788,
      "cbid": 211, "correlation": 99788
    }
  },
  {
    "ph": "s", "id": 99788, "pid": 76337, "tid": -914061504, "ts": 1716454222948996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949083, "dur": 1,
    "args": {
      "External id": 99799,
      "cbid": 251, "correlation": 99799
    }
  },
  {
    "ph": "f", "id": 99799, "pid": 76337, "tid": -914061504, "ts": 1716454222949083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222969297, "dur": 103,
    "args": {
      "External id": 99800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99800,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99800, "pid": 5, "tid": 7, "ts": 1716454222969297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949089, "dur": 13,
    "args": {
      "External id": 99800,
      "cbid": 211, "correlation": 99800
    }
  },
  {
    "ph": "s", "id": 99800, "pid": 76337, "tid": -914061504, "ts": 1716454222949089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949170, "dur": 1,
    "args": {
      "External id": 99811,
      "cbid": 251, "correlation": 99811
    }
  },
  {
    "ph": "f", "id": 99811, "pid": 76337, "tid": -914061504, "ts": 1716454222949170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949175, "dur": 0,
    "args": {
      "External id": 99812,
      "cbid": 251, "correlation": 99812
    }
  },
  {
    "ph": "f", "id": 99812, "pid": 76337, "tid": -914061504, "ts": 1716454222949175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222969401, "dur": 10,
    "args": {
      "External id": 99813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99813,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99813, "pid": 5, "tid": 7, "ts": 1716454222969401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949177, "dur": 14,
    "args": {
      "External id": 99813,
      "cbid": 211, "correlation": 99813
    }
  },
  {
    "ph": "s", "id": 99813, "pid": 76337, "tid": -914061504, "ts": 1716454222949177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222969412, "dur": 5,
    "args": {
      "External id": 99815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99815,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 99815, "pid": 5, "tid": 7, "ts": 1716454222969412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949194, "dur": 9,
    "args": {
      "External id": 99815,
      "cbid": 211, "correlation": 99815
    }
  },
  {
    "ph": "s", "id": 99815, "pid": 76337, "tid": -914061504, "ts": 1716454222949194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949258, "dur": 1,
    "args": {
      "External id": 99826,
      "cbid": 251, "correlation": 99826
    }
  },
  {
    "ph": "f", "id": 99826, "pid": 76337, "tid": -914061504, "ts": 1716454222949258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949262, "dur": 0,
    "args": {
      "External id": 99827,
      "cbid": 251, "correlation": 99827
    }
  },
  {
    "ph": "f", "id": 99827, "pid": 76337, "tid": -914061504, "ts": 1716454222949262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222969418, "dur": 6,
    "args": {
      "External id": 99828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99828,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 99828, "pid": 5, "tid": 7, "ts": 1716454222969418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949263, "dur": 12,
    "args": {
      "External id": 99828,
      "cbid": 211, "correlation": 99828
    }
  },
  {
    "ph": "s", "id": 99828, "pid": 76337, "tid": -914061504, "ts": 1716454222949263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222969425, "dur": 3,
    "args": {
      "External id": 99830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99830,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 99830, "pid": 5, "tid": 7, "ts": 1716454222969425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949277, "dur": 6,
    "args": {
      "External id": 99830,
      "cbid": 211, "correlation": 99830
    }
  },
  {
    "ph": "s", "id": 99830, "pid": 76337, "tid": -914061504, "ts": 1716454222949277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222969430, "dur": 152,
    "args": {
      "External id": 99851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99851,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 99851, "pid": 5, "tid": 7, "ts": 1716454222969430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949352, "dur": 12,
    "args": {
      "External id": 99851,
      "cbid": 211, "correlation": 99851
    }
  },
  {
    "ph": "s", "id": 99851, "pid": 76337, "tid": -914061504, "ts": 1716454222949352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949452, "dur": 2,
    "args": {
      "External id": 99869,
      "cbid": 251, "correlation": 99869
    }
  },
  {
    "ph": "f", "id": 99869, "pid": 76337, "tid": -914061504, "ts": 1716454222949452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222969583, "dur": 106,
    "args": {
      "External id": 99871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99871,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 99871, "pid": 5, "tid": 7, "ts": 1716454222969583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949459, "dur": 14,
    "args": {
      "External id": 99871,
      "cbid": 211, "correlation": 99871
    }
  },
  {
    "ph": "s", "id": 99871, "pid": 76337, "tid": -914061504, "ts": 1716454222949459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222969690, "dur": 35,
    "args": {
      "External id": 99879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99879,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99879, "pid": 5, "tid": 7, "ts": 1716454222969690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949530, "dur": 12,
    "args": {
      "External id": 99879,
      "cbid": 211, "correlation": 99879
    }
  },
  {
    "ph": "s", "id": 99879, "pid": 76337, "tid": -914061504, "ts": 1716454222949530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222969726, "dur": 68,
    "args": {
      "External id": 99887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99887, "pid": 5, "tid": 7, "ts": 1716454222969726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949572, "dur": 9,
    "args": {
      "External id": 99887,
      "cbid": 211, "correlation": 99887
    }
  },
  {
    "ph": "s", "id": 99887, "pid": 76337, "tid": -914061504, "ts": 1716454222949572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222969796, "dur": 90,
    "args": {
      "External id": 99909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99909,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99909, "pid": 5, "tid": 7, "ts": 1716454222969796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949624, "dur": 10,
    "args": {
      "External id": 99909,
      "cbid": 211, "correlation": 99909
    }
  },
  {
    "ph": "s", "id": 99909, "pid": 76337, "tid": -914061504, "ts": 1716454222949624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949715, "dur": 1,
    "args": {
      "External id": 99925,
      "cbid": 251, "correlation": 99925
    }
  },
  {
    "ph": "f", "id": 99925, "pid": 76337, "tid": -914061504, "ts": 1716454222949715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222969887, "dur": 567,
    "args": {
      "External id": 99927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99927,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 99927, "pid": 5, "tid": 7, "ts": 1716454222969887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949720, "dur": 12,
    "args": {
      "External id": 99927,
      "cbid": 211, "correlation": 99927
    }
  },
  {
    "ph": "s", "id": 99927, "pid": 76337, "tid": -914061504, "ts": 1716454222949720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222970456, "dur": 242,
    "args": {
      "External id": 99935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99935, "pid": 5, "tid": 7, "ts": 1716454222970456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949801, "dur": 14,
    "args": {
      "External id": 99935,
      "cbid": 211, "correlation": 99935
    }
  },
  {
    "ph": "s", "id": 99935, "pid": 76337, "tid": -914061504, "ts": 1716454222949801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222970699, "dur": 250,
    "args": {
      "External id": 99943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99943, "pid": 5, "tid": 7, "ts": 1716454222970699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949838, "dur": 9,
    "args": {
      "External id": 99943,
      "cbid": 211, "correlation": 99943
    }
  },
  {
    "ph": "s", "id": 99943, "pid": 76337, "tid": -914061504, "ts": 1716454222949838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949922, "dur": 2,
    "args": {
      "External id": 99959,
      "cbid": 251, "correlation": 99959
    }
  },
  {
    "ph": "f", "id": 99959, "pid": 76337, "tid": -914061504, "ts": 1716454222949922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949927, "dur": 0,
    "args": {
      "External id": 99961,
      "cbid": 251, "correlation": 99961
    }
  },
  {
    "ph": "f", "id": 99961, "pid": 76337, "tid": -914061504, "ts": 1716454222949927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222970950, "dur": 354,
    "args": {
      "External id": 99962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99962,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 99962, "pid": 5, "tid": 7, "ts": 1716454222970950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949932, "dur": 14,
    "args": {
      "External id": 99962,
      "cbid": 211, "correlation": 99962
    }
  },
  {
    "ph": "s", "id": 99962, "pid": 76337, "tid": -914061504, "ts": 1716454222949932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222971305, "dur": 51,
    "args": {
      "External id": 99970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99970,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99970, "pid": 5, "tid": 7, "ts": 1716454222971305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222949984, "dur": 11,
    "args": {
      "External id": 99970,
      "cbid": 211, "correlation": 99970
    }
  },
  {
    "ph": "s", "id": 99970, "pid": 76337, "tid": -914061504, "ts": 1716454222949984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222971357, "dur": 154,
    "args": {
      "External id": 99981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 99981,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 99981, "pid": 5, "tid": 7, "ts": 1716454222971357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950059, "dur": 12,
    "args": {
      "External id": 99981,
      "cbid": 211, "correlation": 99981
    }
  },
  {
    "ph": "s", "id": 99981, "pid": 76337, "tid": -914061504, "ts": 1716454222950059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950124, "dur": 0,
    "args": {
      "External id": 99993,
      "cbid": 317, "correlation": 99993
    }
  },
  {
    "ph": "f", "id": 99993, "pid": 76337, "tid": -914061504, "ts": 1716454222950124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950125, "dur": 0,
    "args": {
      "External id": 99994,
      "cbid": 203, "correlation": 99994
    }
  },
  {
    "ph": "f", "id": 99994, "pid": 76337, "tid": -914061504, "ts": 1716454222950125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950125, "dur": 0,
    "args": {
      "External id": 99995,
      "cbid": 205, "correlation": 99995
    }
  },
  {
    "ph": "f", "id": 99995, "pid": 76337, "tid": -914061504, "ts": 1716454222950125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950150, "dur": 1,
    "args": {
      "External id": 99999,
      "cbid": 251, "correlation": 99999
    }
  },
  {
    "ph": "f", "id": 99999, "pid": 76337, "tid": -914061504, "ts": 1716454222950150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950152, "dur": 0,
    "args": {
      "External id": 100000,
      "cbid": 251, "correlation": 100000
    }
  },
  {
    "ph": "f", "id": 100000, "pid": 76337, "tid": -914061504, "ts": 1716454222950152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950152, "dur": 0,
    "args": {
      "External id": 100001,
      "cbid": 251, "correlation": 100001
    }
  },
  {
    "ph": "f", "id": 100001, "pid": 76337, "tid": -914061504, "ts": 1716454222950152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950153, "dur": 0,
    "args": {
      "External id": 100002,
      "cbid": 251, "correlation": 100002
    }
  },
  {
    "ph": "f", "id": 100002, "pid": 76337, "tid": -914061504, "ts": 1716454222950153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950154, "dur": 0,
    "args": {
      "External id": 100003,
      "cbid": 251, "correlation": 100003
    }
  },
  {
    "ph": "f", "id": 100003, "pid": 76337, "tid": -914061504, "ts": 1716454222950154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950155, "dur": 0,
    "args": {
      "External id": 100004,
      "cbid": 251, "correlation": 100004
    }
  },
  {
    "ph": "f", "id": 100004, "pid": 76337, "tid": -914061504, "ts": 1716454222950155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950156, "dur": 0,
    "args": {
      "External id": 100005,
      "cbid": 251, "correlation": 100005
    }
  },
  {
    "ph": "f", "id": 100005, "pid": 76337, "tid": -914061504, "ts": 1716454222950156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950157, "dur": 0,
    "args": {
      "External id": 100006,
      "cbid": 251, "correlation": 100006
    }
  },
  {
    "ph": "f", "id": 100006, "pid": 76337, "tid": -914061504, "ts": 1716454222950157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950158, "dur": 0,
    "args": {
      "External id": 100007,
      "cbid": 251, "correlation": 100007
    }
  },
  {
    "ph": "f", "id": 100007, "pid": 76337, "tid": -914061504, "ts": 1716454222950158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222971513, "dur": 113,
    "args": {
      "External id": 100008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100008,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100008, "pid": 5, "tid": 7, "ts": 1716454222971513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950160, "dur": 12,
    "args": {
      "External id": 100008,
      "cbid": 211, "correlation": 100008
    }
  },
  {
    "ph": "s", "id": 100008, "pid": 76337, "tid": -914061504, "ts": 1716454222950160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222971628, "dur": 59,
    "args": {
      "External id": 100014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100014,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100014, "pid": 5, "tid": 7, "ts": 1716454222971628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950195, "dur": 9,
    "args": {
      "External id": 100014,
      "cbid": 211, "correlation": 100014
    }
  },
  {
    "ph": "s", "id": 100014, "pid": 76337, "tid": -914061504, "ts": 1716454222950195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222971688, "dur": 50,
    "args": {
      "External id": 100022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100022,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100022, "pid": 5, "tid": 7, "ts": 1716454222971688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950227, "dur": 8,
    "args": {
      "External id": 100022,
      "cbid": 211, "correlation": 100022
    }
  },
  {
    "ph": "s", "id": 100022, "pid": 76337, "tid": -914061504, "ts": 1716454222950227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222971739, "dur": 51,
    "args": {
      "External id": 100042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100042,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 100042, "pid": 5, "tid": 7, "ts": 1716454222971739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950302, "dur": 12,
    "args": {
      "External id": 100042,
      "cbid": 211, "correlation": 100042
    }
  },
  {
    "ph": "s", "id": 100042, "pid": 76337, "tid": -914061504, "ts": 1716454222950302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222971792, "dur": 5,
    "args": {
      "External id": 100054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100054,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100054, "pid": 5, "tid": 7, "ts": 1716454222971792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950324, "dur": 6,
    "args": {
      "External id": 100054,
      "cbid": 211, "correlation": 100054
    }
  },
  {
    "ph": "s", "id": 100054, "pid": 76337, "tid": -914061504, "ts": 1716454222950324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222971798, "dur": 55,
    "args": {
      "External id": 100057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100057, "pid": 5, "tid": 7, "ts": 1716454222971798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950342, "dur": 7,
    "args": {
      "External id": 100057,
      "cbid": 211, "correlation": 100057
    }
  },
  {
    "ph": "s", "id": 100057, "pid": 76337, "tid": -914061504, "ts": 1716454222950342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222971855, "dur": 37,
    "args": {
      "External id": 100066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100066,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100066, "pid": 5, "tid": 7, "ts": 1716454222971855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950382, "dur": 10,
    "args": {
      "External id": 100066,
      "cbid": 211, "correlation": 100066
    }
  },
  {
    "ph": "s", "id": 100066, "pid": 76337, "tid": -914061504, "ts": 1716454222950382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950434, "dur": 0,
    "args": {
      "External id": 100076,
      "cbid": 317, "correlation": 100076
    }
  },
  {
    "ph": "f", "id": 100076, "pid": 76337, "tid": -914061504, "ts": 1716454222950434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950435, "dur": 0,
    "args": {
      "External id": 100077,
      "cbid": 203, "correlation": 100077
    }
  },
  {
    "ph": "f", "id": 100077, "pid": 76337, "tid": -914061504, "ts": 1716454222950435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950435, "dur": 0,
    "args": {
      "External id": 100078,
      "cbid": 205, "correlation": 100078
    }
  },
  {
    "ph": "f", "id": 100078, "pid": 76337, "tid": -914061504, "ts": 1716454222950435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222971893, "dur": 41,
    "args": {
      "External id": 100082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100082, "pid": 5, "tid": 7, "ts": 1716454222971893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950451, "dur": 11,
    "args": {
      "External id": 100082,
      "cbid": 211, "correlation": 100082
    }
  },
  {
    "ph": "s", "id": 100082, "pid": 76337, "tid": -914061504, "ts": 1716454222950451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222971935, "dur": 14,
    "args": {
      "External id": 100084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100084, "pid": 5, "tid": 7, "ts": 1716454222971935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950464, "dur": 5,
    "args": {
      "External id": 100084,
      "cbid": 211, "correlation": 100084
    }
  },
  {
    "ph": "s", "id": 100084, "pid": 76337, "tid": -914061504, "ts": 1716454222950464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222971950, "dur": 4,
    "args": {
      "External id": 100086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100086, "pid": 5, "tid": 7, "ts": 1716454222971950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950474, "dur": 6,
    "args": {
      "External id": 100086,
      "cbid": 211, "correlation": 100086
    }
  },
  {
    "ph": "s", "id": 100086, "pid": 76337, "tid": -914061504, "ts": 1716454222950474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950482, "dur": 0,
    "args": {
      "External id": 100087,
      "cbid": 51, "correlation": 100087
    }
  },
  {
    "ph": "s", "id": 100087, "pid": 76337, "tid": -914061504, "ts": 1716454222950482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222971955, "dur": 689,
    "args": {
      "External id": 100088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100088,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100088, "pid": 5, "tid": 7, "ts": 1716454222971955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950483, "dur": 5,
    "args": {
      "External id": 100088,
      "cbid": 211, "correlation": 100088
    }
  },
  {
    "ph": "s", "id": 100088, "pid": 76337, "tid": -914061504, "ts": 1716454222950483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222972645, "dur": 59,
    "args": {
      "External id": 100093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100093, "pid": 5, "tid": 7, "ts": 1716454222972645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950511, "dur": 9,
    "args": {
      "External id": 100093,
      "cbid": 211, "correlation": 100093
    }
  },
  {
    "ph": "s", "id": 100093, "pid": 76337, "tid": -914061504, "ts": 1716454222950511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222972706, "dur": 4,
    "args": {
      "External id": 100101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100101,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100101, "pid": 5, "tid": 7, "ts": 1716454222972706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950554, "dur": 9,
    "args": {
      "External id": 100101,
      "cbid": 211, "correlation": 100101
    }
  },
  {
    "ph": "s", "id": 100101, "pid": 76337, "tid": -914061504, "ts": 1716454222950554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950619, "dur": 1,
    "args": {
      "External id": 100117,
      "cbid": 251, "correlation": 100117
    }
  },
  {
    "ph": "f", "id": 100117, "pid": 76337, "tid": -914061504, "ts": 1716454222950619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950625, "dur": 0,
    "args": {
      "External id": 100119,
      "cbid": 251, "correlation": 100119
    }
  },
  {
    "ph": "f", "id": 100119, "pid": 76337, "tid": -914061504, "ts": 1716454222950625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222972710, "dur": 11,
    "args": {
      "External id": 100120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100120,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 100120, "pid": 5, "tid": 7, "ts": 1716454222972710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950626, "dur": 11,
    "args": {
      "External id": 100120,
      "cbid": 211, "correlation": 100120
    }
  },
  {
    "ph": "s", "id": 100120, "pid": 76337, "tid": -914061504, "ts": 1716454222950626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222972723, "dur": 5,
    "args": {
      "External id": 100122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100122,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 100122, "pid": 5, "tid": 7, "ts": 1716454222972723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950640, "dur": 6,
    "args": {
      "External id": 100122,
      "cbid": 211, "correlation": 100122
    }
  },
  {
    "ph": "s", "id": 100122, "pid": 76337, "tid": -914061504, "ts": 1716454222950640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222972729, "dur": 52,
    "args": {
      "External id": 100132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100132, "pid": 5, "tid": 7, "ts": 1716454222972729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950698, "dur": 12,
    "args": {
      "External id": 100132,
      "cbid": 211, "correlation": 100132
    }
  },
  {
    "ph": "s", "id": 100132, "pid": 76337, "tid": -914061504, "ts": 1716454222950698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222972782, "dur": 50,
    "args": {
      "External id": 100152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100152,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 100152, "pid": 5, "tid": 7, "ts": 1716454222972782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950764, "dur": 11,
    "args": {
      "External id": 100152,
      "cbid": 211, "correlation": 100152
    }
  },
  {
    "ph": "s", "id": 100152, "pid": 76337, "tid": -914061504, "ts": 1716454222950764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222972834, "dur": 4,
    "args": {
      "External id": 100164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100164,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100164, "pid": 5, "tid": 7, "ts": 1716454222972834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950785, "dur": 6,
    "args": {
      "External id": 100164,
      "cbid": 211, "correlation": 100164
    }
  },
  {
    "ph": "s", "id": 100164, "pid": 76337, "tid": -914061504, "ts": 1716454222950785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222972839, "dur": 55,
    "args": {
      "External id": 100167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100167, "pid": 5, "tid": 7, "ts": 1716454222972839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950803, "dur": 7,
    "args": {
      "External id": 100167,
      "cbid": 211, "correlation": 100167
    }
  },
  {
    "ph": "s", "id": 100167, "pid": 76337, "tid": -914061504, "ts": 1716454222950803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222972895, "dur": 37,
    "args": {
      "External id": 100176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100176,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100176, "pid": 5, "tid": 7, "ts": 1716454222972895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950843, "dur": 10,
    "args": {
      "External id": 100176,
      "cbid": 211, "correlation": 100176
    }
  },
  {
    "ph": "s", "id": 100176, "pid": 76337, "tid": -914061504, "ts": 1716454222950843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950908, "dur": 0,
    "args": {
      "External id": 100186,
      "cbid": 317, "correlation": 100186
    }
  },
  {
    "ph": "f", "id": 100186, "pid": 76337, "tid": -914061504, "ts": 1716454222950908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950908, "dur": 0,
    "args": {
      "External id": 100187,
      "cbid": 203, "correlation": 100187
    }
  },
  {
    "ph": "f", "id": 100187, "pid": 76337, "tid": -914061504, "ts": 1716454222950908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950909, "dur": 0,
    "args": {
      "External id": 100188,
      "cbid": 205, "correlation": 100188
    }
  },
  {
    "ph": "f", "id": 100188, "pid": 76337, "tid": -914061504, "ts": 1716454222950909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222972933, "dur": 41,
    "args": {
      "External id": 100192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100192, "pid": 5, "tid": 7, "ts": 1716454222972933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950923, "dur": 12,
    "args": {
      "External id": 100192,
      "cbid": 211, "correlation": 100192
    }
  },
  {
    "ph": "s", "id": 100192, "pid": 76337, "tid": -914061504, "ts": 1716454222950923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222972975, "dur": 14,
    "args": {
      "External id": 100194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100194,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100194, "pid": 5, "tid": 7, "ts": 1716454222972975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950937, "dur": 5,
    "args": {
      "External id": 100194,
      "cbid": 211, "correlation": 100194
    }
  },
  {
    "ph": "s", "id": 100194, "pid": 76337, "tid": -914061504, "ts": 1716454222950937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222972990, "dur": 3,
    "args": {
      "External id": 100196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100196,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100196, "pid": 5, "tid": 7, "ts": 1716454222972990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950946, "dur": 6,
    "args": {
      "External id": 100196,
      "cbid": 211, "correlation": 100196
    }
  },
  {
    "ph": "s", "id": 100196, "pid": 76337, "tid": -914061504, "ts": 1716454222950946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950955, "dur": 0,
    "args": {
      "External id": 100197,
      "cbid": 51, "correlation": 100197
    }
  },
  {
    "ph": "s", "id": 100197, "pid": 76337, "tid": -914061504, "ts": 1716454222950955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222972994, "dur": 685,
    "args": {
      "External id": 100198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100198,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100198, "pid": 5, "tid": 7, "ts": 1716454222972994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950956, "dur": 5,
    "args": {
      "External id": 100198,
      "cbid": 211, "correlation": 100198
    }
  },
  {
    "ph": "s", "id": 100198, "pid": 76337, "tid": -914061504, "ts": 1716454222950956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222973680, "dur": 57,
    "args": {
      "External id": 100203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100203, "pid": 5, "tid": 7, "ts": 1716454222973680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222950992, "dur": 11,
    "args": {
      "External id": 100203,
      "cbid": 211, "correlation": 100203
    }
  },
  {
    "ph": "s", "id": 100203, "pid": 76337, "tid": -914061504, "ts": 1716454222950992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222973739, "dur": 50,
    "args": {
      "External id": 100211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100211,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100211, "pid": 5, "tid": 7, "ts": 1716454222973739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951026, "dur": 9,
    "args": {
      "External id": 100211,
      "cbid": 211, "correlation": 100211
    }
  },
  {
    "ph": "s", "id": 100211, "pid": 76337, "tid": -914061504, "ts": 1716454222951026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222973790, "dur": 36,
    "args": {
      "External id": 100219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100219,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100219, "pid": 5, "tid": 7, "ts": 1716454222973790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951056, "dur": 8,
    "args": {
      "External id": 100219,
      "cbid": 211, "correlation": 100219
    }
  },
  {
    "ph": "s", "id": 100219, "pid": 76337, "tid": -914061504, "ts": 1716454222951056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222973827, "dur": 50,
    "args": {
      "External id": 100239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100239,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 100239, "pid": 5, "tid": 7, "ts": 1716454222973827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951135, "dur": 13,
    "args": {
      "External id": 100239,
      "cbid": 211, "correlation": 100239
    }
  },
  {
    "ph": "s", "id": 100239, "pid": 76337, "tid": -914061504, "ts": 1716454222951135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222973879, "dur": 4,
    "args": {
      "External id": 100251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100251,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100251, "pid": 5, "tid": 7, "ts": 1716454222973879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951158, "dur": 6,
    "args": {
      "External id": 100251,
      "cbid": 211, "correlation": 100251
    }
  },
  {
    "ph": "s", "id": 100251, "pid": 76337, "tid": -914061504, "ts": 1716454222951158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222973884, "dur": 54,
    "args": {
      "External id": 100254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100254, "pid": 5, "tid": 7, "ts": 1716454222973884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951175, "dur": 6,
    "args": {
      "External id": 100254,
      "cbid": 211, "correlation": 100254
    }
  },
  {
    "ph": "s", "id": 100254, "pid": 76337, "tid": -914061504, "ts": 1716454222951175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951233, "dur": 0,
    "args": {
      "External id": 100265,
      "cbid": 317, "correlation": 100265
    }
  },
  {
    "ph": "f", "id": 100265, "pid": 76337, "tid": -914061504, "ts": 1716454222951233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951234, "dur": 0,
    "args": {
      "External id": 100266,
      "cbid": 203, "correlation": 100266
    }
  },
  {
    "ph": "f", "id": 100266, "pid": 76337, "tid": -914061504, "ts": 1716454222951234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951234, "dur": 0,
    "args": {
      "External id": 100267,
      "cbid": 205, "correlation": 100267
    }
  },
  {
    "ph": "f", "id": 100267, "pid": 76337, "tid": -914061504, "ts": 1716454222951234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951257, "dur": 1,
    "args": {
      "External id": 100271,
      "cbid": 251, "correlation": 100271
    }
  },
  {
    "ph": "f", "id": 100271, "pid": 76337, "tid": -914061504, "ts": 1716454222951257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951259, "dur": 0,
    "args": {
      "External id": 100272,
      "cbid": 251, "correlation": 100272
    }
  },
  {
    "ph": "f", "id": 100272, "pid": 76337, "tid": -914061504, "ts": 1716454222951259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951260, "dur": 0,
    "args": {
      "External id": 100273,
      "cbid": 251, "correlation": 100273
    }
  },
  {
    "ph": "f", "id": 100273, "pid": 76337, "tid": -914061504, "ts": 1716454222951260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951260, "dur": 0,
    "args": {
      "External id": 100274,
      "cbid": 251, "correlation": 100274
    }
  },
  {
    "ph": "f", "id": 100274, "pid": 76337, "tid": -914061504, "ts": 1716454222951260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951261, "dur": 0,
    "args": {
      "External id": 100275,
      "cbid": 251, "correlation": 100275
    }
  },
  {
    "ph": "f", "id": 100275, "pid": 76337, "tid": -914061504, "ts": 1716454222951261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951261, "dur": 0,
    "args": {
      "External id": 100276,
      "cbid": 251, "correlation": 100276
    }
  },
  {
    "ph": "f", "id": 100276, "pid": 76337, "tid": -914061504, "ts": 1716454222951261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951262, "dur": 0,
    "args": {
      "External id": 100277,
      "cbid": 251, "correlation": 100277
    }
  },
  {
    "ph": "f", "id": 100277, "pid": 76337, "tid": -914061504, "ts": 1716454222951262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951263, "dur": 0,
    "args": {
      "External id": 100278,
      "cbid": 251, "correlation": 100278
    }
  },
  {
    "ph": "f", "id": 100278, "pid": 76337, "tid": -914061504, "ts": 1716454222951263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951264, "dur": 0,
    "args": {
      "External id": 100279,
      "cbid": 251, "correlation": 100279
    }
  },
  {
    "ph": "f", "id": 100279, "pid": 76337, "tid": -914061504, "ts": 1716454222951264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222973939, "dur": 111,
    "args": {
      "External id": 100280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100280,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100280, "pid": 5, "tid": 7, "ts": 1716454222973939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951266, "dur": 12,
    "args": {
      "External id": 100280,
      "cbid": 211, "correlation": 100280
    }
  },
  {
    "ph": "s", "id": 100280, "pid": 76337, "tid": -914061504, "ts": 1716454222951266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222974051, "dur": 59,
    "args": {
      "External id": 100286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100286, "pid": 5, "tid": 7, "ts": 1716454222974051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951302, "dur": 9,
    "args": {
      "External id": 100286,
      "cbid": 211, "correlation": 100286
    }
  },
  {
    "ph": "s", "id": 100286, "pid": 76337, "tid": -914061504, "ts": 1716454222951302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222974111, "dur": 693,
    "args": {
      "External id": 100295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100295,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100295, "pid": 5, "tid": 7, "ts": 1716454222974111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951384, "dur": 15,
    "args": {
      "External id": 100295,
      "cbid": 211, "correlation": 100295
    }
  },
  {
    "ph": "s", "id": 100295, "pid": 76337, "tid": -914061504, "ts": 1716454222951384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222974805, "dur": 176,
    "args": {
      "External id": 100317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100317,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100317, "pid": 5, "tid": 7, "ts": 1716454222974805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951440, "dur": 11,
    "args": {
      "External id": 100317,
      "cbid": 211, "correlation": 100317
    }
  },
  {
    "ph": "s", "id": 100317, "pid": 76337, "tid": -914061504, "ts": 1716454222951440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951527, "dur": 1,
    "args": {
      "External id": 100328,
      "cbid": 251, "correlation": 100328
    }
  },
  {
    "ph": "f", "id": 100328, "pid": 76337, "tid": -914061504, "ts": 1716454222951527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222974983, "dur": 193,
    "args": {
      "External id": 100329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100329,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100329, "pid": 5, "tid": 7, "ts": 1716454222974983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951532, "dur": 14,
    "args": {
      "External id": 100329,
      "cbid": 211, "correlation": 100329
    }
  },
  {
    "ph": "s", "id": 100329, "pid": 76337, "tid": -914061504, "ts": 1716454222951532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951601, "dur": 1,
    "args": {
      "External id": 100340,
      "cbid": 251, "correlation": 100340
    }
  },
  {
    "ph": "f", "id": 100340, "pid": 76337, "tid": -914061504, "ts": 1716454222951601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222975177, "dur": 186,
    "args": {
      "External id": 100341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100341,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100341, "pid": 5, "tid": 7, "ts": 1716454222975177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951605, "dur": 11,
    "args": {
      "External id": 100341,
      "cbid": 211, "correlation": 100341
    }
  },
  {
    "ph": "s", "id": 100341, "pid": 76337, "tid": -914061504, "ts": 1716454222951605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951668, "dur": 1,
    "args": {
      "External id": 100352,
      "cbid": 251, "correlation": 100352
    }
  },
  {
    "ph": "f", "id": 100352, "pid": 76337, "tid": -914061504, "ts": 1716454222951668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222975364, "dur": 187,
    "args": {
      "External id": 100353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100353,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100353, "pid": 5, "tid": 7, "ts": 1716454222975364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951672, "dur": 11,
    "args": {
      "External id": 100353,
      "cbid": 211, "correlation": 100353
    }
  },
  {
    "ph": "s", "id": 100353, "pid": 76337, "tid": -914061504, "ts": 1716454222951672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222975552, "dur": 18220,
    "args": {
      "External id": 100374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100374,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100374, "pid": 5, "tid": 7, "ts": 1716454222975552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951752, "dur": 12,
    "args": {
      "External id": 100374,
      "cbid": 211, "correlation": 100374
    }
  },
  {
    "ph": "s", "id": 100374, "pid": 76337, "tid": -914061504, "ts": 1716454222951752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951851, "dur": 1,
    "args": {
      "External id": 100392,
      "cbid": 251, "correlation": 100392
    }
  },
  {
    "ph": "f", "id": 100392, "pid": 76337, "tid": -914061504, "ts": 1716454222951851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222993773, "dur": 197,
    "args": {
      "External id": 100394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100394,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100394, "pid": 5, "tid": 7, "ts": 1716454222993773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951856, "dur": 13,
    "args": {
      "External id": 100394,
      "cbid": 211, "correlation": 100394
    }
  },
  {
    "ph": "s", "id": 100394, "pid": 76337, "tid": -914061504, "ts": 1716454222951856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222993972, "dur": 66,
    "args": {
      "External id": 100402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100402,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100402, "pid": 5, "tid": 7, "ts": 1716454222993972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951927, "dur": 12,
    "args": {
      "External id": 100402,
      "cbid": 211, "correlation": 100402
    }
  },
  {
    "ph": "s", "id": 100402, "pid": 76337, "tid": -914061504, "ts": 1716454222951927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222994040, "dur": 97,
    "args": {
      "External id": 100410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100410,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100410, "pid": 5, "tid": 7, "ts": 1716454222994040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222951966, "dur": 16,
    "args": {
      "External id": 100410,
      "cbid": 211, "correlation": 100410
    }
  },
  {
    "ph": "s", "id": 100410, "pid": 76337, "tid": -914061504, "ts": 1716454222951966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222994137, "dur": 53,
    "args": {
      "External id": 100421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100421, "pid": 5, "tid": 7, "ts": 1716454222994137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952047, "dur": 13,
    "args": {
      "External id": 100421,
      "cbid": 211, "correlation": 100421
    }
  },
  {
    "ph": "s", "id": 100421, "pid": 76337, "tid": -914061504, "ts": 1716454222952047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222994192, "dur": 90,
    "args": {
      "External id": 100443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100443,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100443, "pid": 5, "tid": 7, "ts": 1716454222994192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952078, "dur": 8,
    "args": {
      "External id": 100443,
      "cbid": 211, "correlation": 100443
    }
  },
  {
    "ph": "s", "id": 100443, "pid": 76337, "tid": -914061504, "ts": 1716454222952078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952163, "dur": 1,
    "args": {
      "External id": 100454,
      "cbid": 251, "correlation": 100454
    }
  },
  {
    "ph": "f", "id": 100454, "pid": 76337, "tid": -914061504, "ts": 1716454222952163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222994283, "dur": 103,
    "args": {
      "External id": 100455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100455,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100455, "pid": 5, "tid": 7, "ts": 1716454222994283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952168, "dur": 12,
    "args": {
      "External id": 100455,
      "cbid": 211, "correlation": 100455
    }
  },
  {
    "ph": "s", "id": 100455, "pid": 76337, "tid": -914061504, "ts": 1716454222952168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952239, "dur": 1,
    "args": {
      "External id": 100466,
      "cbid": 251, "correlation": 100466
    }
  },
  {
    "ph": "f", "id": 100466, "pid": 76337, "tid": -914061504, "ts": 1716454222952239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952242, "dur": 0,
    "args": {
      "External id": 100467,
      "cbid": 251, "correlation": 100467
    }
  },
  {
    "ph": "f", "id": 100467, "pid": 76337, "tid": -914061504, "ts": 1716454222952242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222994388, "dur": 10,
    "args": {
      "External id": 100468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100468,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 100468, "pid": 5, "tid": 7, "ts": 1716454222994388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952244, "dur": 12,
    "args": {
      "External id": 100468,
      "cbid": 211, "correlation": 100468
    }
  },
  {
    "ph": "s", "id": 100468, "pid": 76337, "tid": -914061504, "ts": 1716454222952244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222994400, "dur": 5,
    "args": {
      "External id": 100470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100470,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 100470, "pid": 5, "tid": 7, "ts": 1716454222994400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952259, "dur": 6,
    "args": {
      "External id": 100470,
      "cbid": 211, "correlation": 100470
    }
  },
  {
    "ph": "s", "id": 100470, "pid": 76337, "tid": -914061504, "ts": 1716454222952259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952320, "dur": 1,
    "args": {
      "External id": 100481,
      "cbid": 251, "correlation": 100481
    }
  },
  {
    "ph": "f", "id": 100481, "pid": 76337, "tid": -914061504, "ts": 1716454222952320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952323, "dur": 0,
    "args": {
      "External id": 100482,
      "cbid": 251, "correlation": 100482
    }
  },
  {
    "ph": "f", "id": 100482, "pid": 76337, "tid": -914061504, "ts": 1716454222952323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222994406, "dur": 6,
    "args": {
      "External id": 100483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100483,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 100483, "pid": 5, "tid": 7, "ts": 1716454222994406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952325, "dur": 11,
    "args": {
      "External id": 100483,
      "cbid": 211, "correlation": 100483
    }
  },
  {
    "ph": "s", "id": 100483, "pid": 76337, "tid": -914061504, "ts": 1716454222952325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222994413, "dur": 3,
    "args": {
      "External id": 100485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100485,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 100485, "pid": 5, "tid": 7, "ts": 1716454222994413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952338, "dur": 6,
    "args": {
      "External id": 100485,
      "cbid": 211, "correlation": 100485
    }
  },
  {
    "ph": "s", "id": 100485, "pid": 76337, "tid": -914061504, "ts": 1716454222952338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222994418, "dur": 152,
    "args": {
      "External id": 100506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100506,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100506, "pid": 5, "tid": 7, "ts": 1716454222994418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952412, "dur": 12,
    "args": {
      "External id": 100506,
      "cbid": 211, "correlation": 100506
    }
  },
  {
    "ph": "s", "id": 100506, "pid": 76337, "tid": -914061504, "ts": 1716454222952412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952508, "dur": 1,
    "args": {
      "External id": 100524,
      "cbid": 251, "correlation": 100524
    }
  },
  {
    "ph": "f", "id": 100524, "pid": 76337, "tid": -914061504, "ts": 1716454222952508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222994571, "dur": 107,
    "args": {
      "External id": 100526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100526,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100526, "pid": 5, "tid": 7, "ts": 1716454222994571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952514, "dur": 13,
    "args": {
      "External id": 100526,
      "cbid": 211, "correlation": 100526
    }
  },
  {
    "ph": "s", "id": 100526, "pid": 76337, "tid": -914061504, "ts": 1716454222952514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222994679, "dur": 34,
    "args": {
      "External id": 100534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100534,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100534, "pid": 5, "tid": 7, "ts": 1716454222994679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952584, "dur": 13,
    "args": {
      "External id": 100534,
      "cbid": 211, "correlation": 100534
    }
  },
  {
    "ph": "s", "id": 100534, "pid": 76337, "tid": -914061504, "ts": 1716454222952584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222994715, "dur": 66,
    "args": {
      "External id": 100542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100542, "pid": 5, "tid": 7, "ts": 1716454222994715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952625, "dur": 9,
    "args": {
      "External id": 100542,
      "cbid": 211, "correlation": 100542
    }
  },
  {
    "ph": "s", "id": 100542, "pid": 76337, "tid": -914061504, "ts": 1716454222952625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222994782, "dur": 90,
    "args": {
      "External id": 100564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100564,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100564, "pid": 5, "tid": 7, "ts": 1716454222994782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952675, "dur": 10,
    "args": {
      "External id": 100564,
      "cbid": 211, "correlation": 100564
    }
  },
  {
    "ph": "s", "id": 100564, "pid": 76337, "tid": -914061504, "ts": 1716454222952675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952762, "dur": 1,
    "args": {
      "External id": 100580,
      "cbid": 251, "correlation": 100580
    }
  },
  {
    "ph": "f", "id": 100580, "pid": 76337, "tid": -914061504, "ts": 1716454222952762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222994874, "dur": 564,
    "args": {
      "External id": 100582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100582,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100582, "pid": 5, "tid": 7, "ts": 1716454222994874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952768, "dur": 13,
    "args": {
      "External id": 100582,
      "cbid": 211, "correlation": 100582
    }
  },
  {
    "ph": "s", "id": 100582, "pid": 76337, "tid": -914061504, "ts": 1716454222952768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222995439, "dur": 240,
    "args": {
      "External id": 100590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100590, "pid": 5, "tid": 7, "ts": 1716454222995439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952833, "dur": 13,
    "args": {
      "External id": 100590,
      "cbid": 211, "correlation": 100590
    }
  },
  {
    "ph": "s", "id": 100590, "pid": 76337, "tid": -914061504, "ts": 1716454222952833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222995681, "dur": 251,
    "args": {
      "External id": 100598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100598,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100598, "pid": 5, "tid": 7, "ts": 1716454222995681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952864, "dur": 8,
    "args": {
      "External id": 100598,
      "cbid": 211, "correlation": 100598
    }
  },
  {
    "ph": "s", "id": 100598, "pid": 76337, "tid": -914061504, "ts": 1716454222952864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952947, "dur": 2,
    "args": {
      "External id": 100614,
      "cbid": 251, "correlation": 100614
    }
  },
  {
    "ph": "f", "id": 100614, "pid": 76337, "tid": -914061504, "ts": 1716454222952947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952953, "dur": 0,
    "args": {
      "External id": 100616,
      "cbid": 251, "correlation": 100616
    }
  },
  {
    "ph": "f", "id": 100616, "pid": 76337, "tid": -914061504, "ts": 1716454222952953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454222995933, "dur": 356,
    "args": {
      "External id": 100617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100617,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 100617, "pid": 5, "tid": 7, "ts": 1716454222995933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222952955, "dur": 13,
    "args": {
      "External id": 100617,
      "cbid": 211, "correlation": 100617
    }
  },
  {
    "ph": "s", "id": 100617, "pid": 76337, "tid": -914061504, "ts": 1716454222952955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222996291, "dur": 50,
    "args": {
      "External id": 100625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100625,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100625, "pid": 5, "tid": 7, "ts": 1716454222996291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953007, "dur": 10,
    "args": {
      "External id": 100625,
      "cbid": 211, "correlation": 100625
    }
  },
  {
    "ph": "s", "id": 100625, "pid": 76337, "tid": -914061504, "ts": 1716454222953007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222996342, "dur": 154,
    "args": {
      "External id": 100636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100636, "pid": 5, "tid": 7, "ts": 1716454222996342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953075, "dur": 12,
    "args": {
      "External id": 100636,
      "cbid": 211, "correlation": 100636
    }
  },
  {
    "ph": "s", "id": 100636, "pid": 76337, "tid": -914061504, "ts": 1716454222953075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953139, "dur": 0,
    "args": {
      "External id": 100648,
      "cbid": 317, "correlation": 100648
    }
  },
  {
    "ph": "f", "id": 100648, "pid": 76337, "tid": -914061504, "ts": 1716454222953139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953140, "dur": 0,
    "args": {
      "External id": 100649,
      "cbid": 203, "correlation": 100649
    }
  },
  {
    "ph": "f", "id": 100649, "pid": 76337, "tid": -914061504, "ts": 1716454222953140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953141, "dur": 0,
    "args": {
      "External id": 100650,
      "cbid": 205, "correlation": 100650
    }
  },
  {
    "ph": "f", "id": 100650, "pid": 76337, "tid": -914061504, "ts": 1716454222953141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953164, "dur": 1,
    "args": {
      "External id": 100654,
      "cbid": 251, "correlation": 100654
    }
  },
  {
    "ph": "f", "id": 100654, "pid": 76337, "tid": -914061504, "ts": 1716454222953164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953166, "dur": 0,
    "args": {
      "External id": 100655,
      "cbid": 251, "correlation": 100655
    }
  },
  {
    "ph": "f", "id": 100655, "pid": 76337, "tid": -914061504, "ts": 1716454222953166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953166, "dur": 0,
    "args": {
      "External id": 100656,
      "cbid": 251, "correlation": 100656
    }
  },
  {
    "ph": "f", "id": 100656, "pid": 76337, "tid": -914061504, "ts": 1716454222953166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953167, "dur": 0,
    "args": {
      "External id": 100657,
      "cbid": 251, "correlation": 100657
    }
  },
  {
    "ph": "f", "id": 100657, "pid": 76337, "tid": -914061504, "ts": 1716454222953167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953167, "dur": 0,
    "args": {
      "External id": 100658,
      "cbid": 251, "correlation": 100658
    }
  },
  {
    "ph": "f", "id": 100658, "pid": 76337, "tid": -914061504, "ts": 1716454222953167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953168, "dur": 0,
    "args": {
      "External id": 100659,
      "cbid": 251, "correlation": 100659
    }
  },
  {
    "ph": "f", "id": 100659, "pid": 76337, "tid": -914061504, "ts": 1716454222953168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953169, "dur": 0,
    "args": {
      "External id": 100660,
      "cbid": 251, "correlation": 100660
    }
  },
  {
    "ph": "f", "id": 100660, "pid": 76337, "tid": -914061504, "ts": 1716454222953169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953170, "dur": 0,
    "args": {
      "External id": 100661,
      "cbid": 251, "correlation": 100661
    }
  },
  {
    "ph": "f", "id": 100661, "pid": 76337, "tid": -914061504, "ts": 1716454222953170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953171, "dur": 0,
    "args": {
      "External id": 100662,
      "cbid": 251, "correlation": 100662
    }
  },
  {
    "ph": "f", "id": 100662, "pid": 76337, "tid": -914061504, "ts": 1716454222953171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222996498, "dur": 112,
    "args": {
      "External id": 100663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100663,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100663, "pid": 5, "tid": 7, "ts": 1716454222996498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953173, "dur": 12,
    "args": {
      "External id": 100663,
      "cbid": 211, "correlation": 100663
    }
  },
  {
    "ph": "s", "id": 100663, "pid": 76337, "tid": -914061504, "ts": 1716454222953173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222996612, "dur": 59,
    "args": {
      "External id": 100669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100669, "pid": 5, "tid": 7, "ts": 1716454222996612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953208, "dur": 9,
    "args": {
      "External id": 100669,
      "cbid": 211, "correlation": 100669
    }
  },
  {
    "ph": "s", "id": 100669, "pid": 76337, "tid": -914061504, "ts": 1716454222953208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222996672, "dur": 50,
    "args": {
      "External id": 100677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100677,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100677, "pid": 5, "tid": 7, "ts": 1716454222996672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953240, "dur": 9,
    "args": {
      "External id": 100677,
      "cbid": 211, "correlation": 100677
    }
  },
  {
    "ph": "s", "id": 100677, "pid": 76337, "tid": -914061504, "ts": 1716454222953240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953313, "dur": 0,
    "args": {
      "External id": 100687,
      "cbid": 317, "correlation": 100687
    }
  },
  {
    "ph": "f", "id": 100687, "pid": 76337, "tid": -914061504, "ts": 1716454222953313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953314, "dur": 0,
    "args": {
      "External id": 100688,
      "cbid": 203, "correlation": 100688
    }
  },
  {
    "ph": "f", "id": 100688, "pid": 76337, "tid": -914061504, "ts": 1716454222953314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953314, "dur": 0,
    "args": {
      "External id": 100689,
      "cbid": 205, "correlation": 100689
    }
  },
  {
    "ph": "f", "id": 100689, "pid": 76337, "tid": -914061504, "ts": 1716454222953314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222996723, "dur": 41,
    "args": {
      "External id": 100693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100693,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100693, "pid": 5, "tid": 7, "ts": 1716454222996723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953330, "dur": 12,
    "args": {
      "External id": 100693,
      "cbid": 211, "correlation": 100693
    }
  },
  {
    "ph": "s", "id": 100693, "pid": 76337, "tid": -914061504, "ts": 1716454222953330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222996765, "dur": 14,
    "args": {
      "External id": 100695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100695,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100695, "pid": 5, "tid": 7, "ts": 1716454222996765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953346, "dur": 6,
    "args": {
      "External id": 100695,
      "cbid": 211, "correlation": 100695
    }
  },
  {
    "ph": "s", "id": 100695, "pid": 76337, "tid": -914061504, "ts": 1716454222953346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454222996782, "dur": 1,
    "args": {
      "External id": 100697,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 100697,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 100697, "pid": 5, "tid": 7, "ts": 1716454222996782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953366, "dur": 18,
    "args": {
      "External id": 100697,
      "cbid": 51, "correlation": 100697
    }
  },
  {
    "ph": "s", "id": 100697, "pid": 76337, "tid": -914061504, "ts": 1716454222953366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222996786, "dur": 352,
    "args": {
      "External id": 100698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100698,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100698, "pid": 5, "tid": 7, "ts": 1716454222996786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953386, "dur": 11,
    "args": {
      "External id": 100698,
      "cbid": 211, "correlation": 100698
    }
  },
  {
    "ph": "s", "id": 100698, "pid": 76337, "tid": -914061504, "ts": 1716454222953386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997139, "dur": 12,
    "args": {
      "External id": 100700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100700, "pid": 5, "tid": 7, "ts": 1716454222997139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953405, "dur": 7,
    "args": {
      "External id": 100700,
      "cbid": 211, "correlation": 100700
    }
  },
  {
    "ph": "s", "id": 100700, "pid": 76337, "tid": -914061504, "ts": 1716454222953405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222997153, "dur": 15,
    "args": {
      "External id": 100706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100706, "pid": 5, "tid": 7, "ts": 1716454222997153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953435, "dur": 10,
    "args": {
      "External id": 100706,
      "cbid": 211, "correlation": 100706
    }
  },
  {
    "ph": "s", "id": 100706, "pid": 76337, "tid": -914061504, "ts": 1716454222953435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997169, "dur": 18,
    "args": {
      "External id": 100726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100726,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 100726, "pid": 5, "tid": 7, "ts": 1716454222997169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953528, "dur": 13,
    "args": {
      "External id": 100726,
      "cbid": 211, "correlation": 100726
    }
  },
  {
    "ph": "s", "id": 100726, "pid": 76337, "tid": -914061504, "ts": 1716454222953528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222997188, "dur": 4,
    "args": {
      "External id": 100738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100738,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100738, "pid": 5, "tid": 7, "ts": 1716454222997188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953551, "dur": 6,
    "args": {
      "External id": 100738,
      "cbid": 211, "correlation": 100738
    }
  },
  {
    "ph": "s", "id": 100738, "pid": 76337, "tid": -914061504, "ts": 1716454222953551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222997194, "dur": 18,
    "args": {
      "External id": 100741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100741, "pid": 5, "tid": 7, "ts": 1716454222997194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953569, "dur": 6,
    "args": {
      "External id": 100741,
      "cbid": 211, "correlation": 100741
    }
  },
  {
    "ph": "s", "id": 100741, "pid": 76337, "tid": -914061504, "ts": 1716454222953569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222997213, "dur": 11,
    "args": {
      "External id": 100750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100750,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100750, "pid": 5, "tid": 7, "ts": 1716454222997213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953610, "dur": 9,
    "args": {
      "External id": 100750,
      "cbid": 211, "correlation": 100750
    }
  },
  {
    "ph": "s", "id": 100750, "pid": 76337, "tid": -914061504, "ts": 1716454222953610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953664, "dur": 0,
    "args": {
      "External id": 100760,
      "cbid": 317, "correlation": 100760
    }
  },
  {
    "ph": "f", "id": 100760, "pid": 76337, "tid": -914061504, "ts": 1716454222953664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953665, "dur": 0,
    "args": {
      "External id": 100761,
      "cbid": 203, "correlation": 100761
    }
  },
  {
    "ph": "f", "id": 100761, "pid": 76337, "tid": -914061504, "ts": 1716454222953665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953666, "dur": 0,
    "args": {
      "External id": 100762,
      "cbid": 205, "correlation": 100762
    }
  },
  {
    "ph": "f", "id": 100762, "pid": 76337, "tid": -914061504, "ts": 1716454222953666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997225, "dur": 11,
    "args": {
      "External id": 100766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100766, "pid": 5, "tid": 7, "ts": 1716454222997225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953681, "dur": 12,
    "args": {
      "External id": 100766,
      "cbid": 211, "correlation": 100766
    }
  },
  {
    "ph": "s", "id": 100766, "pid": 76337, "tid": -914061504, "ts": 1716454222953681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997237, "dur": 24,
    "args": {
      "External id": 100768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100768, "pid": 5, "tid": 7, "ts": 1716454222997237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953696, "dur": 5,
    "args": {
      "External id": 100768,
      "cbid": 211, "correlation": 100768
    }
  },
  {
    "ph": "s", "id": 100768, "pid": 76337, "tid": -914061504, "ts": 1716454222953696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454222997262, "dur": 4,
    "args": {
      "External id": 100770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 100770, "pid": 5, "tid": 7, "ts": 1716454222997262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953707, "dur": 6,
    "args": {
      "External id": 100770,
      "cbid": 211, "correlation": 100770
    }
  },
  {
    "ph": "s", "id": 100770, "pid": 76337, "tid": -914061504, "ts": 1716454222953707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953716, "dur": 0,
    "args": {
      "External id": 100771,
      "cbid": 51, "correlation": 100771
    }
  },
  {
    "ph": "s", "id": 100771, "pid": 76337, "tid": -914061504, "ts": 1716454222953716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454222997267, "dur": 354,
    "args": {
      "External id": 100772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100772,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100772, "pid": 5, "tid": 7, "ts": 1716454222997267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953717, "dur": 7,
    "args": {
      "External id": 100772,
      "cbid": 211, "correlation": 100772
    }
  },
  {
    "ph": "s", "id": 100772, "pid": 76337, "tid": -914061504, "ts": 1716454222953717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997623, "dur": 20,
    "args": {
      "External id": 100773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100773, "pid": 5, "tid": 7, "ts": 1716454222997623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953727, "dur": 5,
    "args": {
      "External id": 100773,
      "cbid": 211, "correlation": 100773
    }
  },
  {
    "ph": "s", "id": 100773, "pid": 76337, "tid": -914061504, "ts": 1716454222953727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222997644, "dur": 32,
    "args": {
      "External id": 100779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100779, "pid": 5, "tid": 7, "ts": 1716454222997644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953755, "dur": 8,
    "args": {
      "External id": 100779,
      "cbid": 211, "correlation": 100779
    }
  },
  {
    "ph": "s", "id": 100779, "pid": 76337, "tid": -914061504, "ts": 1716454222953755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222997677, "dur": 3,
    "args": {
      "External id": 100787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100787,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 100787, "pid": 5, "tid": 7, "ts": 1716454222997677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953798, "dur": 9,
    "args": {
      "External id": 100787,
      "cbid": 211, "correlation": 100787
    }
  },
  {
    "ph": "s", "id": 100787, "pid": 76337, "tid": -914061504, "ts": 1716454222953798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953863, "dur": 1,
    "args": {
      "External id": 100803,
      "cbid": 251, "correlation": 100803
    }
  },
  {
    "ph": "f", "id": 100803, "pid": 76337, "tid": -914061504, "ts": 1716454222953863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953868, "dur": 0,
    "args": {
      "External id": 100805,
      "cbid": 251, "correlation": 100805
    }
  },
  {
    "ph": "f", "id": 100805, "pid": 76337, "tid": -914061504, "ts": 1716454222953868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222997682, "dur": 12,
    "args": {
      "External id": 100806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100806,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 100806, "pid": 5, "tid": 7, "ts": 1716454222997682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953870, "dur": 11,
    "args": {
      "External id": 100806,
      "cbid": 211, "correlation": 100806
    }
  },
  {
    "ph": "s", "id": 100806, "pid": 76337, "tid": -914061504, "ts": 1716454222953870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454222997696, "dur": 5,
    "args": {
      "External id": 100808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100808,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 100808, "pid": 5, "tid": 7, "ts": 1716454222997696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953883, "dur": 6,
    "args": {
      "External id": 100808,
      "cbid": 211, "correlation": 100808
    }
  },
  {
    "ph": "s", "id": 100808, "pid": 76337, "tid": -914061504, "ts": 1716454222953883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222997702, "dur": 30,
    "args": {
      "External id": 100818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100818, "pid": 5, "tid": 7, "ts": 1716454222997702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222953941, "dur": 12,
    "args": {
      "External id": 100818,
      "cbid": 211, "correlation": 100818
    }
  },
  {
    "ph": "s", "id": 100818, "pid": 76337, "tid": -914061504, "ts": 1716454222953941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997733, "dur": 31,
    "args": {
      "External id": 100838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100838,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 100838, "pid": 5, "tid": 7, "ts": 1716454222997733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954016, "dur": 11,
    "args": {
      "External id": 100838,
      "cbid": 211, "correlation": 100838
    }
  },
  {
    "ph": "s", "id": 100838, "pid": 76337, "tid": -914061504, "ts": 1716454222954016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222997765, "dur": 5,
    "args": {
      "External id": 100850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100850,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 100850, "pid": 5, "tid": 7, "ts": 1716454222997765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954037, "dur": 7,
    "args": {
      "External id": 100850,
      "cbid": 211, "correlation": 100850
    }
  },
  {
    "ph": "s", "id": 100850, "pid": 76337, "tid": -914061504, "ts": 1716454222954037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222997772, "dur": 30,
    "args": {
      "External id": 100853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100853, "pid": 5, "tid": 7, "ts": 1716454222997772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954058, "dur": 6,
    "args": {
      "External id": 100853,
      "cbid": 211, "correlation": 100853
    }
  },
  {
    "ph": "s", "id": 100853, "pid": 76337, "tid": -914061504, "ts": 1716454222954058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222997803, "dur": 20,
    "args": {
      "External id": 100862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100862,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100862, "pid": 5, "tid": 7, "ts": 1716454222997803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954099, "dur": 10,
    "args": {
      "External id": 100862,
      "cbid": 211, "correlation": 100862
    }
  },
  {
    "ph": "s", "id": 100862, "pid": 76337, "tid": -914061504, "ts": 1716454222954099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954162, "dur": 0,
    "args": {
      "External id": 100872,
      "cbid": 317, "correlation": 100872
    }
  },
  {
    "ph": "f", "id": 100872, "pid": 76337, "tid": -914061504, "ts": 1716454222954162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954163, "dur": 0,
    "args": {
      "External id": 100873,
      "cbid": 203, "correlation": 100873
    }
  },
  {
    "ph": "f", "id": 100873, "pid": 76337, "tid": -914061504, "ts": 1716454222954163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954163, "dur": 0,
    "args": {
      "External id": 100874,
      "cbid": 205, "correlation": 100874
    }
  },
  {
    "ph": "f", "id": 100874, "pid": 76337, "tid": -914061504, "ts": 1716454222954163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997824, "dur": 23,
    "args": {
      "External id": 100878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100878,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100878, "pid": 5, "tid": 7, "ts": 1716454222997824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954182, "dur": 13,
    "args": {
      "External id": 100878,
      "cbid": 211, "correlation": 100878
    }
  },
  {
    "ph": "s", "id": 100878, "pid": 76337, "tid": -914061504, "ts": 1716454222954182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222997848, "dur": 43,
    "args": {
      "External id": 100880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100880, "pid": 5, "tid": 7, "ts": 1716454222997848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954197, "dur": 5,
    "args": {
      "External id": 100880,
      "cbid": 211, "correlation": 100880
    }
  },
  {
    "ph": "s", "id": 100880, "pid": 76337, "tid": -914061504, "ts": 1716454222954197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222997893, "dur": 647,
    "args": {
      "External id": 100882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100882,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100882, "pid": 5, "tid": 7, "ts": 1716454222997893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954211, "dur": 9,
    "args": {
      "External id": 100882,
      "cbid": 211, "correlation": 100882
    }
  },
  {
    "ph": "s", "id": 100882, "pid": 76337, "tid": -914061504, "ts": 1716454222954211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222998541, "dur": 20,
    "args": {
      "External id": 100884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100884,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100884, "pid": 5, "tid": 7, "ts": 1716454222998541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954224, "dur": 5,
    "args": {
      "External id": 100884,
      "cbid": 211, "correlation": 100884
    }
  },
  {
    "ph": "s", "id": 100884, "pid": 76337, "tid": -914061504, "ts": 1716454222954224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222998562, "dur": 33,
    "args": {
      "External id": 100890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100890, "pid": 5, "tid": 7, "ts": 1716454222998562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954253, "dur": 8,
    "args": {
      "External id": 100890,
      "cbid": 211, "correlation": 100890
    }
  },
  {
    "ph": "s", "id": 100890, "pid": 76337, "tid": -914061504, "ts": 1716454222954253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954311, "dur": 0,
    "args": {
      "External id": 100900,
      "cbid": 317, "correlation": 100900
    }
  },
  {
    "ph": "f", "id": 100900, "pid": 76337, "tid": -914061504, "ts": 1716454222954311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954312, "dur": 0,
    "args": {
      "External id": 100901,
      "cbid": 203, "correlation": 100901
    }
  },
  {
    "ph": "f", "id": 100901, "pid": 76337, "tid": -914061504, "ts": 1716454222954312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954313, "dur": 0,
    "args": {
      "External id": 100902,
      "cbid": 205, "correlation": 100902
    }
  },
  {
    "ph": "f", "id": 100902, "pid": 76337, "tid": -914061504, "ts": 1716454222954313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954335, "dur": 1,
    "args": {
      "External id": 100906,
      "cbid": 251, "correlation": 100906
    }
  },
  {
    "ph": "f", "id": 100906, "pid": 76337, "tid": -914061504, "ts": 1716454222954335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954337, "dur": 0,
    "args": {
      "External id": 100907,
      "cbid": 251, "correlation": 100907
    }
  },
  {
    "ph": "f", "id": 100907, "pid": 76337, "tid": -914061504, "ts": 1716454222954337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954337, "dur": 0,
    "args": {
      "External id": 100908,
      "cbid": 251, "correlation": 100908
    }
  },
  {
    "ph": "f", "id": 100908, "pid": 76337, "tid": -914061504, "ts": 1716454222954337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954338, "dur": 0,
    "args": {
      "External id": 100909,
      "cbid": 251, "correlation": 100909
    }
  },
  {
    "ph": "f", "id": 100909, "pid": 76337, "tid": -914061504, "ts": 1716454222954338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954339, "dur": 0,
    "args": {
      "External id": 100910,
      "cbid": 251, "correlation": 100910
    }
  },
  {
    "ph": "f", "id": 100910, "pid": 76337, "tid": -914061504, "ts": 1716454222954339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954340, "dur": 0,
    "args": {
      "External id": 100911,
      "cbid": 251, "correlation": 100911
    }
  },
  {
    "ph": "f", "id": 100911, "pid": 76337, "tid": -914061504, "ts": 1716454222954340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954341, "dur": 0,
    "args": {
      "External id": 100912,
      "cbid": 251, "correlation": 100912
    }
  },
  {
    "ph": "f", "id": 100912, "pid": 76337, "tid": -914061504, "ts": 1716454222954341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954342, "dur": 0,
    "args": {
      "External id": 100913,
      "cbid": 251, "correlation": 100913
    }
  },
  {
    "ph": "f", "id": 100913, "pid": 76337, "tid": -914061504, "ts": 1716454222954342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954343, "dur": 0,
    "args": {
      "External id": 100914,
      "cbid": 251, "correlation": 100914
    }
  },
  {
    "ph": "f", "id": 100914, "pid": 76337, "tid": -914061504, "ts": 1716454222954343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454222998596, "dur": 51,
    "args": {
      "External id": 100915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100915,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 100915, "pid": 5, "tid": 7, "ts": 1716454222998596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954345, "dur": 12,
    "args": {
      "External id": 100915,
      "cbid": 211, "correlation": 100915
    }
  },
  {
    "ph": "s", "id": 100915, "pid": 76337, "tid": -914061504, "ts": 1716454222954345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222998648, "dur": 31,
    "args": {
      "External id": 100921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100921, "pid": 5, "tid": 7, "ts": 1716454222998648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954377, "dur": 8,
    "args": {
      "External id": 100921,
      "cbid": 211, "correlation": 100921
    }
  },
  {
    "ph": "s", "id": 100921, "pid": 76337, "tid": -914061504, "ts": 1716454222954377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454222998681, "dur": 27,
    "args": {
      "External id": 100929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100929,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100929, "pid": 5, "tid": 7, "ts": 1716454222998681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954407, "dur": 8,
    "args": {
      "External id": 100929,
      "cbid": 211, "correlation": 100929
    }
  },
  {
    "ph": "s", "id": 100929, "pid": 76337, "tid": -914061504, "ts": 1716454222954407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454222998709, "dur": 20,
    "args": {
      "External id": 100937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100937,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100937, "pid": 5, "tid": 7, "ts": 1716454222998709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954436, "dur": 9,
    "args": {
      "External id": 100937,
      "cbid": 211, "correlation": 100937
    }
  },
  {
    "ph": "s", "id": 100937, "pid": 76337, "tid": -914061504, "ts": 1716454222954436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222998730, "dur": 29,
    "args": {
      "External id": 100957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100957,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 100957, "pid": 5, "tid": 7, "ts": 1716454222998730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954518, "dur": 12,
    "args": {
      "External id": 100957,
      "cbid": 211, "correlation": 100957
    }
  },
  {
    "ph": "s", "id": 100957, "pid": 76337, "tid": -914061504, "ts": 1716454222954518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454222998760, "dur": 5,
    "args": {
      "External id": 100969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100969,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 100969, "pid": 5, "tid": 7, "ts": 1716454222998760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954539, "dur": 7,
    "args": {
      "External id": 100969,
      "cbid": 211, "correlation": 100969
    }
  },
  {
    "ph": "s", "id": 100969, "pid": 76337, "tid": -914061504, "ts": 1716454222954539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222998767, "dur": 29,
    "args": {
      "External id": 100972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100972, "pid": 5, "tid": 7, "ts": 1716454222998767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954557, "dur": 6,
    "args": {
      "External id": 100972,
      "cbid": 211, "correlation": 100972
    }
  },
  {
    "ph": "s", "id": 100972, "pid": 76337, "tid": -914061504, "ts": 1716454222954557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954615, "dur": 0,
    "args": {
      "External id": 100983,
      "cbid": 317, "correlation": 100983
    }
  },
  {
    "ph": "f", "id": 100983, "pid": 76337, "tid": -914061504, "ts": 1716454222954615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954616, "dur": 0,
    "args": {
      "External id": 100984,
      "cbid": 203, "correlation": 100984
    }
  },
  {
    "ph": "f", "id": 100984, "pid": 76337, "tid": -914061504, "ts": 1716454222954616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954617, "dur": 0,
    "args": {
      "External id": 100985,
      "cbid": 205, "correlation": 100985
    }
  },
  {
    "ph": "f", "id": 100985, "pid": 76337, "tid": -914061504, "ts": 1716454222954617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222998797, "dur": 22,
    "args": {
      "External id": 100989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100989,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100989, "pid": 5, "tid": 7, "ts": 1716454222998797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954632, "dur": 12,
    "args": {
      "External id": 100989,
      "cbid": 211, "correlation": 100989
    }
  },
  {
    "ph": "s", "id": 100989, "pid": 76337, "tid": -914061504, "ts": 1716454222954632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454222998820, "dur": 117,
    "args": {
      "External id": 100991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100991,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 100991, "pid": 5, "tid": 7, "ts": 1716454222998820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954653, "dur": 9,
    "args": {
      "External id": 100991,
      "cbid": 211, "correlation": 100991
    }
  },
  {
    "ph": "s", "id": 100991, "pid": 76337, "tid": -914061504, "ts": 1716454222954653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454222998939, "dur": 22,
    "args": {
      "External id": 100993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100993,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100993, "pid": 5, "tid": 7, "ts": 1716454222998939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954665, "dur": 5,
    "args": {
      "External id": 100993,
      "cbid": 211, "correlation": 100993
    }
  },
  {
    "ph": "s", "id": 100993, "pid": 76337, "tid": -914061504, "ts": 1716454222954665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454222998962, "dur": 32,
    "args": {
      "External id": 100999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 100999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 100999, "pid": 5, "tid": 7, "ts": 1716454222998962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954693, "dur": 9,
    "args": {
      "External id": 100999,
      "cbid": 211, "correlation": 100999
    }
  },
  {
    "ph": "s", "id": 100999, "pid": 76337, "tid": -914061504, "ts": 1716454222954693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454222998995, "dur": 197,
    "args": {
      "External id": 101008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101008,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101008, "pid": 5, "tid": 7, "ts": 1716454222998995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954774, "dur": 14,
    "args": {
      "External id": 101008,
      "cbid": 211, "correlation": 101008
    }
  },
  {
    "ph": "s", "id": 101008, "pid": 76337, "tid": -914061504, "ts": 1716454222954774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454222999193, "dur": 64,
    "args": {
      "External id": 101030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101030,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101030, "pid": 5, "tid": 7, "ts": 1716454222999193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954831, "dur": 10,
    "args": {
      "External id": 101030,
      "cbid": 211, "correlation": 101030
    }
  },
  {
    "ph": "s", "id": 101030, "pid": 76337, "tid": -914061504, "ts": 1716454222954831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954922, "dur": 1,
    "args": {
      "External id": 101041,
      "cbid": 251, "correlation": 101041
    }
  },
  {
    "ph": "f", "id": 101041, "pid": 76337, "tid": -914061504, "ts": 1716454222954922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222999258, "dur": 152,
    "args": {
      "External id": 101042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101042,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101042, "pid": 5, "tid": 7, "ts": 1716454222999258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222954927, "dur": 13,
    "args": {
      "External id": 101042,
      "cbid": 211, "correlation": 101042
    }
  },
  {
    "ph": "s", "id": 101042, "pid": 76337, "tid": -914061504, "ts": 1716454222954927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955007, "dur": 1,
    "args": {
      "External id": 101053,
      "cbid": 251, "correlation": 101053
    }
  },
  {
    "ph": "f", "id": 101053, "pid": 76337, "tid": -914061504, "ts": 1716454222955007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222999411, "dur": 144,
    "args": {
      "External id": 101054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101054,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101054, "pid": 5, "tid": 7, "ts": 1716454222999411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955012, "dur": 12,
    "args": {
      "External id": 101054,
      "cbid": 211, "correlation": 101054
    }
  },
  {
    "ph": "s", "id": 101054, "pid": 76337, "tid": -914061504, "ts": 1716454222955012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955079, "dur": 1,
    "args": {
      "External id": 101065,
      "cbid": 251, "correlation": 101065
    }
  },
  {
    "ph": "f", "id": 101065, "pid": 76337, "tid": -914061504, "ts": 1716454222955079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454222999557, "dur": 143,
    "args": {
      "External id": 101066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101066,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101066, "pid": 5, "tid": 7, "ts": 1716454222999557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955084, "dur": 12,
    "args": {
      "External id": 101066,
      "cbid": 211, "correlation": 101066
    }
  },
  {
    "ph": "s", "id": 101066, "pid": 76337, "tid": -914061504, "ts": 1716454222955084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454222999701, "dur": 1909,
    "args": {
      "External id": 101087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101087,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 101087, "pid": 5, "tid": 7, "ts": 1716454222999701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955170, "dur": 14,
    "args": {
      "External id": 101087,
      "cbid": 211, "correlation": 101087
    }
  },
  {
    "ph": "s", "id": 101087, "pid": 76337, "tid": -914061504, "ts": 1716454222955170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955273, "dur": 1,
    "args": {
      "External id": 101105,
      "cbid": 251, "correlation": 101105
    }
  },
  {
    "ph": "f", "id": 101105, "pid": 76337, "tid": -914061504, "ts": 1716454222955273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223001612, "dur": 145,
    "args": {
      "External id": 101107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101107,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 101107, "pid": 5, "tid": 7, "ts": 1716454223001612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955279, "dur": 14,
    "args": {
      "External id": 101107,
      "cbid": 211, "correlation": 101107
    }
  },
  {
    "ph": "s", "id": 101107, "pid": 76337, "tid": -914061504, "ts": 1716454222955279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223001758, "dur": 36,
    "args": {
      "External id": 101115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101115,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101115, "pid": 5, "tid": 7, "ts": 1716454223001758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955352, "dur": 12,
    "args": {
      "External id": 101115,
      "cbid": 211, "correlation": 101115
    }
  },
  {
    "ph": "s", "id": 101115, "pid": 76337, "tid": -914061504, "ts": 1716454222955352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223001796, "dur": 51,
    "args": {
      "External id": 101123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101123,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101123, "pid": 5, "tid": 7, "ts": 1716454223001796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955391, "dur": 9,
    "args": {
      "External id": 101123,
      "cbid": 211, "correlation": 101123
    }
  },
  {
    "ph": "s", "id": 101123, "pid": 76337, "tid": -914061504, "ts": 1716454222955391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223001848, "dur": 30,
    "args": {
      "External id": 101134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101134, "pid": 5, "tid": 7, "ts": 1716454223001848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955464, "dur": 13,
    "args": {
      "External id": 101134,
      "cbid": 211, "correlation": 101134
    }
  },
  {
    "ph": "s", "id": 101134, "pid": 76337, "tid": -914061504, "ts": 1716454222955464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223001879, "dur": 34,
    "args": {
      "External id": 101156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101156,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101156, "pid": 5, "tid": 7, "ts": 1716454223001879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955496, "dur": 8,
    "args": {
      "External id": 101156,
      "cbid": 211, "correlation": 101156
    }
  },
  {
    "ph": "s", "id": 101156, "pid": 76337, "tid": -914061504, "ts": 1716454222955496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955580, "dur": 1,
    "args": {
      "External id": 101167,
      "cbid": 251, "correlation": 101167
    }
  },
  {
    "ph": "f", "id": 101167, "pid": 76337, "tid": -914061504, "ts": 1716454222955580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223001914, "dur": 86,
    "args": {
      "External id": 101168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101168,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101168, "pid": 5, "tid": 7, "ts": 1716454223001914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955585, "dur": 14,
    "args": {
      "External id": 101168,
      "cbid": 211, "correlation": 101168
    }
  },
  {
    "ph": "s", "id": 101168, "pid": 76337, "tid": -914061504, "ts": 1716454222955585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955656, "dur": 1,
    "args": {
      "External id": 101179,
      "cbid": 251, "correlation": 101179
    }
  },
  {
    "ph": "f", "id": 101179, "pid": 76337, "tid": -914061504, "ts": 1716454222955656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955659, "dur": 0,
    "args": {
      "External id": 101180,
      "cbid": 251, "correlation": 101180
    }
  },
  {
    "ph": "f", "id": 101180, "pid": 76337, "tid": -914061504, "ts": 1716454222955659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223002002, "dur": 11,
    "args": {
      "External id": 101181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101181,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 101181, "pid": 5, "tid": 7, "ts": 1716454223002002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955661, "dur": 11,
    "args": {
      "External id": 101181,
      "cbid": 211, "correlation": 101181
    }
  },
  {
    "ph": "s", "id": 101181, "pid": 76337, "tid": -914061504, "ts": 1716454222955661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223002014, "dur": 5,
    "args": {
      "External id": 101183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101183,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 101183, "pid": 5, "tid": 7, "ts": 1716454223002014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955675, "dur": 7,
    "args": {
      "External id": 101183,
      "cbid": 211, "correlation": 101183
    }
  },
  {
    "ph": "s", "id": 101183, "pid": 76337, "tid": -914061504, "ts": 1716454222955675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955735, "dur": 1,
    "args": {
      "External id": 101194,
      "cbid": 251, "correlation": 101194
    }
  },
  {
    "ph": "f", "id": 101194, "pid": 76337, "tid": -914061504, "ts": 1716454222955735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955738, "dur": 0,
    "args": {
      "External id": 101195,
      "cbid": 251, "correlation": 101195
    }
  },
  {
    "ph": "f", "id": 101195, "pid": 76337, "tid": -914061504, "ts": 1716454222955738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223002020, "dur": 7,
    "args": {
      "External id": 101196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101196,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 101196, "pid": 5, "tid": 7, "ts": 1716454223002020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955740, "dur": 12,
    "args": {
      "External id": 101196,
      "cbid": 211, "correlation": 101196
    }
  },
  {
    "ph": "s", "id": 101196, "pid": 76337, "tid": -914061504, "ts": 1716454222955740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223002028, "dur": 3,
    "args": {
      "External id": 101198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101198,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 101198, "pid": 5, "tid": 7, "ts": 1716454223002028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955754, "dur": 5,
    "args": {
      "External id": 101198,
      "cbid": 211, "correlation": 101198
    }
  },
  {
    "ph": "s", "id": 101198, "pid": 76337, "tid": -914061504, "ts": 1716454222955754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223002033, "dur": 90,
    "args": {
      "External id": 101219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101219,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 101219, "pid": 5, "tid": 7, "ts": 1716454223002033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955827, "dur": 13,
    "args": {
      "External id": 101219,
      "cbid": 211, "correlation": 101219
    }
  },
  {
    "ph": "s", "id": 101219, "pid": 76337, "tid": -914061504, "ts": 1716454222955827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955925, "dur": 1,
    "args": {
      "External id": 101237,
      "cbid": 251, "correlation": 101237
    }
  },
  {
    "ph": "f", "id": 101237, "pid": 76337, "tid": -914061504, "ts": 1716454222955925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223002124, "dur": 97,
    "args": {
      "External id": 101239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101239,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101239, "pid": 5, "tid": 7, "ts": 1716454223002124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222955930, "dur": 13,
    "args": {
      "External id": 101239,
      "cbid": 211, "correlation": 101239
    }
  },
  {
    "ph": "s", "id": 101239, "pid": 76337, "tid": -914061504, "ts": 1716454222955930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223002223, "dur": 19,
    "args": {
      "External id": 101247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101247,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101247, "pid": 5, "tid": 7, "ts": 1716454223002223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956009, "dur": 12,
    "args": {
      "External id": 101247,
      "cbid": 211, "correlation": 101247
    }
  },
  {
    "ph": "s", "id": 101247, "pid": 76337, "tid": -914061504, "ts": 1716454222956009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223002243, "dur": 37,
    "args": {
      "External id": 101255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101255, "pid": 5, "tid": 7, "ts": 1716454223002243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956053, "dur": 9,
    "args": {
      "External id": 101255,
      "cbid": 211, "correlation": 101255
    }
  },
  {
    "ph": "s", "id": 101255, "pid": 76337, "tid": -914061504, "ts": 1716454222956053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223002281, "dur": 34,
    "args": {
      "External id": 101277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101277,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101277, "pid": 5, "tid": 7, "ts": 1716454223002281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956105, "dur": 10,
    "args": {
      "External id": 101277,
      "cbid": 211, "correlation": 101277
    }
  },
  {
    "ph": "s", "id": 101277, "pid": 76337, "tid": -914061504, "ts": 1716454222956105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956194, "dur": 1,
    "args": {
      "External id": 101293,
      "cbid": 251, "correlation": 101293
    }
  },
  {
    "ph": "f", "id": 101293, "pid": 76337, "tid": -914061504, "ts": 1716454222956194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956199, "dur": 0,
    "args": {
      "External id": 101295,
      "cbid": 251, "correlation": 101295
    }
  },
  {
    "ph": "f", "id": 101295, "pid": 76337, "tid": -914061504, "ts": 1716454222956199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223002316, "dur": 535,
    "args": {
      "External id": 101296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101296,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 101296, "pid": 5, "tid": 7, "ts": 1716454223002316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956203, "dur": 13,
    "args": {
      "External id": 101296,
      "cbid": 211, "correlation": 101296
    }
  },
  {
    "ph": "s", "id": 101296, "pid": 76337, "tid": -914061504, "ts": 1716454222956203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223002852, "dur": 122,
    "args": {
      "External id": 101304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101304, "pid": 5, "tid": 7, "ts": 1716454223002852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956267, "dur": 13,
    "args": {
      "External id": 101304,
      "cbid": 211, "correlation": 101304
    }
  },
  {
    "ph": "s", "id": 101304, "pid": 76337, "tid": -914061504, "ts": 1716454222956267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223002976, "dur": 128,
    "args": {
      "External id": 101312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101312, "pid": 5, "tid": 7, "ts": 1716454223002976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956299, "dur": 8,
    "args": {
      "External id": 101312,
      "cbid": 211, "correlation": 101312
    }
  },
  {
    "ph": "s", "id": 101312, "pid": 76337, "tid": -914061504, "ts": 1716454222956299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956376, "dur": 1,
    "args": {
      "External id": 101328,
      "cbid": 251, "correlation": 101328
    }
  },
  {
    "ph": "f", "id": 101328, "pid": 76337, "tid": -914061504, "ts": 1716454222956376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223003105, "dur": 300,
    "args": {
      "External id": 101330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101330,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101330, "pid": 5, "tid": 7, "ts": 1716454223003105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956382, "dur": 13,
    "args": {
      "External id": 101330,
      "cbid": 211, "correlation": 101330
    }
  },
  {
    "ph": "s", "id": 101330, "pid": 76337, "tid": -914061504, "ts": 1716454222956382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223003406, "dur": 28,
    "args": {
      "External id": 101338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101338,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101338, "pid": 5, "tid": 7, "ts": 1716454223003406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956424, "dur": 10,
    "args": {
      "External id": 101338,
      "cbid": 211, "correlation": 101338
    }
  },
  {
    "ph": "s", "id": 101338, "pid": 76337, "tid": -914061504, "ts": 1716454222956424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223003435, "dur": 81,
    "args": {
      "External id": 101349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101349, "pid": 5, "tid": 7, "ts": 1716454223003435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956493, "dur": 13,
    "args": {
      "External id": 101349,
      "cbid": 211, "correlation": 101349
    }
  },
  {
    "ph": "s", "id": 101349, "pid": 76337, "tid": -914061504, "ts": 1716454222956493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956558, "dur": 0,
    "args": {
      "External id": 101361,
      "cbid": 317, "correlation": 101361
    }
  },
  {
    "ph": "f", "id": 101361, "pid": 76337, "tid": -914061504, "ts": 1716454222956558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956559, "dur": 0,
    "args": {
      "External id": 101362,
      "cbid": 203, "correlation": 101362
    }
  },
  {
    "ph": "f", "id": 101362, "pid": 76337, "tid": -914061504, "ts": 1716454222956559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956559, "dur": 0,
    "args": {
      "External id": 101363,
      "cbid": 205, "correlation": 101363
    }
  },
  {
    "ph": "f", "id": 101363, "pid": 76337, "tid": -914061504, "ts": 1716454222956559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223003517, "dur": 22,
    "args": {
      "External id": 101367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101367,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101367, "pid": 5, "tid": 7, "ts": 1716454223003517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956574, "dur": 12,
    "args": {
      "External id": 101367,
      "cbid": 211, "correlation": 101367
    }
  },
  {
    "ph": "s", "id": 101367, "pid": 76337, "tid": -914061504, "ts": 1716454222956574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223003541, "dur": 118,
    "args": {
      "External id": 101369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101369,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101369, "pid": 5, "tid": 7, "ts": 1716454223003541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956593, "dur": 6,
    "args": {
      "External id": 101369,
      "cbid": 211, "correlation": 101369
    }
  },
  {
    "ph": "s", "id": 101369, "pid": 76337, "tid": -914061504, "ts": 1716454222956593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223003660, "dur": 22,
    "args": {
      "External id": 101371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101371, "pid": 5, "tid": 7, "ts": 1716454223003660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956603, "dur": 5,
    "args": {
      "External id": 101371,
      "cbid": 211, "correlation": 101371
    }
  },
  {
    "ph": "s", "id": 101371, "pid": 76337, "tid": -914061504, "ts": 1716454222956603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223003684, "dur": 33,
    "args": {
      "External id": 101377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101377,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101377, "pid": 5, "tid": 7, "ts": 1716454223003684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956630, "dur": 9,
    "args": {
      "External id": 101377,
      "cbid": 211, "correlation": 101377
    }
  },
  {
    "ph": "s", "id": 101377, "pid": 76337, "tid": -914061504, "ts": 1716454222956630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223003718, "dur": 26,
    "args": {
      "External id": 101385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101385,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101385, "pid": 5, "tid": 7, "ts": 1716454223003718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956662, "dur": 8,
    "args": {
      "External id": 101385,
      "cbid": 211, "correlation": 101385
    }
  },
  {
    "ph": "s", "id": 101385, "pid": 76337, "tid": -914061504, "ts": 1716454222956662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223003745, "dur": 30,
    "args": {
      "External id": 101405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101405,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 101405, "pid": 5, "tid": 7, "ts": 1716454223003745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956734, "dur": 12,
    "args": {
      "External id": 101405,
      "cbid": 211, "correlation": 101405
    }
  },
  {
    "ph": "s", "id": 101405, "pid": 76337, "tid": -914061504, "ts": 1716454222956734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223003777, "dur": 5,
    "args": {
      "External id": 101417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101417,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 101417, "pid": 5, "tid": 7, "ts": 1716454223003777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956756, "dur": 6,
    "args": {
      "External id": 101417,
      "cbid": 211, "correlation": 101417
    }
  },
  {
    "ph": "s", "id": 101417, "pid": 76337, "tid": -914061504, "ts": 1716454222956756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223003783, "dur": 31,
    "args": {
      "External id": 101420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101420, "pid": 5, "tid": 7, "ts": 1716454223003783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956774, "dur": 7,
    "args": {
      "External id": 101420,
      "cbid": 211, "correlation": 101420
    }
  },
  {
    "ph": "s", "id": 101420, "pid": 76337, "tid": -914061504, "ts": 1716454222956774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223003815, "dur": 21,
    "args": {
      "External id": 101429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101429,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101429, "pid": 5, "tid": 7, "ts": 1716454223003815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956813, "dur": 10,
    "args": {
      "External id": 101429,
      "cbid": 211, "correlation": 101429
    }
  },
  {
    "ph": "s", "id": 101429, "pid": 76337, "tid": -914061504, "ts": 1716454222956813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956864, "dur": 0,
    "args": {
      "External id": 101439,
      "cbid": 317, "correlation": 101439
    }
  },
  {
    "ph": "f", "id": 101439, "pid": 76337, "tid": -914061504, "ts": 1716454222956864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956865, "dur": 0,
    "args": {
      "External id": 101440,
      "cbid": 203, "correlation": 101440
    }
  },
  {
    "ph": "f", "id": 101440, "pid": 76337, "tid": -914061504, "ts": 1716454222956865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956866, "dur": 0,
    "args": {
      "External id": 101441,
      "cbid": 205, "correlation": 101441
    }
  },
  {
    "ph": "f", "id": 101441, "pid": 76337, "tid": -914061504, "ts": 1716454222956866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223003838, "dur": 21,
    "args": {
      "External id": 101445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101445,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101445, "pid": 5, "tid": 7, "ts": 1716454223003838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956879, "dur": 11,
    "args": {
      "External id": 101445,
      "cbid": 211, "correlation": 101445
    }
  },
  {
    "ph": "s", "id": 101445, "pid": 76337, "tid": -914061504, "ts": 1716454222956879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223003860, "dur": 43,
    "args": {
      "External id": 101447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101447,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101447, "pid": 5, "tid": 7, "ts": 1716454223003860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956893, "dur": 6,
    "args": {
      "External id": 101447,
      "cbid": 211, "correlation": 101447
    }
  },
  {
    "ph": "s", "id": 101447, "pid": 76337, "tid": -914061504, "ts": 1716454222956893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223003904, "dur": 647,
    "args": {
      "External id": 101449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101449,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101449, "pid": 5, "tid": 7, "ts": 1716454223003904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956904, "dur": 6,
    "args": {
      "External id": 101449,
      "cbid": 211, "correlation": 101449
    }
  },
  {
    "ph": "s", "id": 101449, "pid": 76337, "tid": -914061504, "ts": 1716454222956904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223004553, "dur": 22,
    "args": {
      "External id": 101451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101451, "pid": 5, "tid": 7, "ts": 1716454223004553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956914, "dur": 5,
    "args": {
      "External id": 101451,
      "cbid": 211, "correlation": 101451
    }
  },
  {
    "ph": "s", "id": 101451, "pid": 76337, "tid": -914061504, "ts": 1716454222956914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223004577, "dur": 33,
    "args": {
      "External id": 101457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101457, "pid": 5, "tid": 7, "ts": 1716454223004577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956941, "dur": 9,
    "args": {
      "External id": 101457,
      "cbid": 211, "correlation": 101457
    }
  },
  {
    "ph": "s", "id": 101457, "pid": 76337, "tid": -914061504, "ts": 1716454222956941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223004611, "dur": 3,
    "args": {
      "External id": 101465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101465,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 101465, "pid": 5, "tid": 7, "ts": 1716454223004611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222956993, "dur": 11,
    "args": {
      "External id": 101465,
      "cbid": 211, "correlation": 101465
    }
  },
  {
    "ph": "s", "id": 101465, "pid": 76337, "tid": -914061504, "ts": 1716454222956993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957059, "dur": 1,
    "args": {
      "External id": 101481,
      "cbid": 251, "correlation": 101481
    }
  },
  {
    "ph": "f", "id": 101481, "pid": 76337, "tid": -914061504, "ts": 1716454222957059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957064, "dur": 0,
    "args": {
      "External id": 101483,
      "cbid": 251, "correlation": 101483
    }
  },
  {
    "ph": "f", "id": 101483, "pid": 76337, "tid": -914061504, "ts": 1716454222957064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223004615, "dur": 12,
    "args": {
      "External id": 101484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101484,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 101484, "pid": 5, "tid": 7, "ts": 1716454223004615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957066, "dur": 11,
    "args": {
      "External id": 101484,
      "cbid": 211, "correlation": 101484
    }
  },
  {
    "ph": "s", "id": 101484, "pid": 76337, "tid": -914061504, "ts": 1716454222957066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223004629, "dur": 5,
    "args": {
      "External id": 101486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101486,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 101486, "pid": 5, "tid": 7, "ts": 1716454223004629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957079, "dur": 5,
    "args": {
      "External id": 101486,
      "cbid": 211, "correlation": 101486
    }
  },
  {
    "ph": "s", "id": 101486, "pid": 76337, "tid": -914061504, "ts": 1716454222957079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223004635, "dur": 29,
    "args": {
      "External id": 101496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101496, "pid": 5, "tid": 7, "ts": 1716454223004635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957137, "dur": 13,
    "args": {
      "External id": 101496,
      "cbid": 211, "correlation": 101496
    }
  },
  {
    "ph": "s", "id": 101496, "pid": 76337, "tid": -914061504, "ts": 1716454222957137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223004665, "dur": 30,
    "args": {
      "External id": 101516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101516,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 101516, "pid": 5, "tid": 7, "ts": 1716454223004665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957204, "dur": 11,
    "args": {
      "External id": 101516,
      "cbid": 211, "correlation": 101516
    }
  },
  {
    "ph": "s", "id": 101516, "pid": 76337, "tid": -914061504, "ts": 1716454222957204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223004696, "dur": 5,
    "args": {
      "External id": 101528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101528,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 101528, "pid": 5, "tid": 7, "ts": 1716454223004696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957225, "dur": 6,
    "args": {
      "External id": 101528,
      "cbid": 211, "correlation": 101528
    }
  },
  {
    "ph": "s", "id": 101528, "pid": 76337, "tid": -914061504, "ts": 1716454222957225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223004702, "dur": 29,
    "args": {
      "External id": 101531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101531, "pid": 5, "tid": 7, "ts": 1716454223004702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957245, "dur": 6,
    "args": {
      "External id": 101531,
      "cbid": 211, "correlation": 101531
    }
  },
  {
    "ph": "s", "id": 101531, "pid": 76337, "tid": -914061504, "ts": 1716454222957245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223004733, "dur": 20,
    "args": {
      "External id": 101540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101540,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101540, "pid": 5, "tid": 7, "ts": 1716454223004733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957286, "dur": 10,
    "args": {
      "External id": 101540,
      "cbid": 211, "correlation": 101540
    }
  },
  {
    "ph": "s", "id": 101540, "pid": 76337, "tid": -914061504, "ts": 1716454222957286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957349, "dur": 0,
    "args": {
      "External id": 101550,
      "cbid": 317, "correlation": 101550
    }
  },
  {
    "ph": "f", "id": 101550, "pid": 76337, "tid": -914061504, "ts": 1716454222957349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957350, "dur": 0,
    "args": {
      "External id": 101551,
      "cbid": 203, "correlation": 101551
    }
  },
  {
    "ph": "f", "id": 101551, "pid": 76337, "tid": -914061504, "ts": 1716454222957350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957351, "dur": 0,
    "args": {
      "External id": 101552,
      "cbid": 205, "correlation": 101552
    }
  },
  {
    "ph": "f", "id": 101552, "pid": 76337, "tid": -914061504, "ts": 1716454222957351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223004754, "dur": 23,
    "args": {
      "External id": 101556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101556,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101556, "pid": 5, "tid": 7, "ts": 1716454223004754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957364, "dur": 12,
    "args": {
      "External id": 101556,
      "cbid": 211, "correlation": 101556
    }
  },
  {
    "ph": "s", "id": 101556, "pid": 76337, "tid": -914061504, "ts": 1716454222957364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223004779, "dur": 43,
    "args": {
      "External id": 101558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101558, "pid": 5, "tid": 7, "ts": 1716454223004779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957378, "dur": 5,
    "args": {
      "External id": 101558,
      "cbid": 211, "correlation": 101558
    }
  },
  {
    "ph": "s", "id": 101558, "pid": 76337, "tid": -914061504, "ts": 1716454222957378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223004823, "dur": 640,
    "args": {
      "External id": 101560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101560,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101560, "pid": 5, "tid": 7, "ts": 1716454223004823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957390, "dur": 7,
    "args": {
      "External id": 101560,
      "cbid": 211, "correlation": 101560
    }
  },
  {
    "ph": "s", "id": 101560, "pid": 76337, "tid": -914061504, "ts": 1716454222957390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223005464, "dur": 22,
    "args": {
      "External id": 101562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101562,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101562, "pid": 5, "tid": 7, "ts": 1716454223005464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957400, "dur": 5,
    "args": {
      "External id": 101562,
      "cbid": 211, "correlation": 101562
    }
  },
  {
    "ph": "s", "id": 101562, "pid": 76337, "tid": -914061504, "ts": 1716454222957400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223005487, "dur": 32,
    "args": {
      "External id": 101568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101568, "pid": 5, "tid": 7, "ts": 1716454223005487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957427, "dur": 8,
    "args": {
      "External id": 101568,
      "cbid": 211, "correlation": 101568
    }
  },
  {
    "ph": "s", "id": 101568, "pid": 76337, "tid": -914061504, "ts": 1716454222957427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223005521, "dur": 26,
    "args": {
      "External id": 101576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101576,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101576, "pid": 5, "tid": 7, "ts": 1716454223005521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957459, "dur": 8,
    "args": {
      "External id": 101576,
      "cbid": 211, "correlation": 101576
    }
  },
  {
    "ph": "s", "id": 101576, "pid": 76337, "tid": -914061504, "ts": 1716454222957459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223005548, "dur": 20,
    "args": {
      "External id": 101584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101584,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101584, "pid": 5, "tid": 7, "ts": 1716454223005548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957488, "dur": 9,
    "args": {
      "External id": 101584,
      "cbid": 211, "correlation": 101584
    }
  },
  {
    "ph": "s", "id": 101584, "pid": 76337, "tid": -914061504, "ts": 1716454222957488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223005569, "dur": 30,
    "args": {
      "External id": 101604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101604,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 101604, "pid": 5, "tid": 7, "ts": 1716454223005569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957567, "dur": 12,
    "args": {
      "External id": 101604,
      "cbid": 211, "correlation": 101604
    }
  },
  {
    "ph": "s", "id": 101604, "pid": 76337, "tid": -914061504, "ts": 1716454222957567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223005600, "dur": 5,
    "args": {
      "External id": 101616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101616,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 101616, "pid": 5, "tid": 7, "ts": 1716454223005600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957590, "dur": 7,
    "args": {
      "External id": 101616,
      "cbid": 211, "correlation": 101616
    }
  },
  {
    "ph": "s", "id": 101616, "pid": 76337, "tid": -914061504, "ts": 1716454222957590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223005606, "dur": 29,
    "args": {
      "External id": 101619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101619, "pid": 5, "tid": 7, "ts": 1716454223005606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957607, "dur": 6,
    "args": {
      "External id": 101619,
      "cbid": 211, "correlation": 101619
    }
  },
  {
    "ph": "s", "id": 101619, "pid": 76337, "tid": -914061504, "ts": 1716454222957607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957664, "dur": 0,
    "args": {
      "External id": 101630,
      "cbid": 317, "correlation": 101630
    }
  },
  {
    "ph": "f", "id": 101630, "pid": 76337, "tid": -914061504, "ts": 1716454222957664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957665, "dur": 0,
    "args": {
      "External id": 101631,
      "cbid": 203, "correlation": 101631
    }
  },
  {
    "ph": "f", "id": 101631, "pid": 76337, "tid": -914061504, "ts": 1716454222957665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957666, "dur": 0,
    "args": {
      "External id": 101632,
      "cbid": 205, "correlation": 101632
    }
  },
  {
    "ph": "f", "id": 101632, "pid": 76337, "tid": -914061504, "ts": 1716454222957666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223005637, "dur": 22,
    "args": {
      "External id": 101636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101636,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101636, "pid": 5, "tid": 7, "ts": 1716454223005637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957679, "dur": 12,
    "args": {
      "External id": 101636,
      "cbid": 211, "correlation": 101636
    }
  },
  {
    "ph": "s", "id": 101636, "pid": 76337, "tid": -914061504, "ts": 1716454222957679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223005660, "dur": 115,
    "args": {
      "External id": 101638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101638,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101638, "pid": 5, "tid": 7, "ts": 1716454223005660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957697, "dur": 6,
    "args": {
      "External id": 101638,
      "cbid": 211, "correlation": 101638
    }
  },
  {
    "ph": "s", "id": 101638, "pid": 76337, "tid": -914061504, "ts": 1716454222957697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223005776, "dur": 22,
    "args": {
      "External id": 101640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101640,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101640, "pid": 5, "tid": 7, "ts": 1716454223005776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957707, "dur": 5,
    "args": {
      "External id": 101640,
      "cbid": 211, "correlation": 101640
    }
  },
  {
    "ph": "s", "id": 101640, "pid": 76337, "tid": -914061504, "ts": 1716454222957707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223005799, "dur": 32,
    "args": {
      "External id": 101646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101646, "pid": 5, "tid": 7, "ts": 1716454223005799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957733, "dur": 9,
    "args": {
      "External id": 101646,
      "cbid": 211, "correlation": 101646
    }
  },
  {
    "ph": "s", "id": 101646, "pid": 76337, "tid": -914061504, "ts": 1716454222957733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223005833, "dur": 178,
    "args": {
      "External id": 101655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101655,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101655, "pid": 5, "tid": 7, "ts": 1716454223005833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957816, "dur": 14,
    "args": {
      "External id": 101655,
      "cbid": 211, "correlation": 101655
    }
  },
  {
    "ph": "s", "id": 101655, "pid": 76337, "tid": -914061504, "ts": 1716454222957816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223006012, "dur": 64,
    "args": {
      "External id": 101677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101677,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101677, "pid": 5, "tid": 7, "ts": 1716454223006012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957873, "dur": 10,
    "args": {
      "External id": 101677,
      "cbid": 211, "correlation": 101677
    }
  },
  {
    "ph": "s", "id": 101677, "pid": 76337, "tid": -914061504, "ts": 1716454222957873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957961, "dur": 1,
    "args": {
      "External id": 101688,
      "cbid": 251, "correlation": 101688
    }
  },
  {
    "ph": "f", "id": 101688, "pid": 76337, "tid": -914061504, "ts": 1716454222957961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223006077, "dur": 149,
    "args": {
      "External id": 101689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101689,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101689, "pid": 5, "tid": 7, "ts": 1716454223006077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222957966, "dur": 21,
    "args": {
      "External id": 101689,
      "cbid": 211, "correlation": 101689
    }
  },
  {
    "ph": "s", "id": 101689, "pid": 76337, "tid": -914061504, "ts": 1716454222957966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958049, "dur": 1,
    "args": {
      "External id": 101700,
      "cbid": 251, "correlation": 101700
    }
  },
  {
    "ph": "f", "id": 101700, "pid": 76337, "tid": -914061504, "ts": 1716454222958049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223006227, "dur": 146,
    "args": {
      "External id": 101701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101701,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101701, "pid": 5, "tid": 7, "ts": 1716454223006227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958054, "dur": 12,
    "args": {
      "External id": 101701,
      "cbid": 211, "correlation": 101701
    }
  },
  {
    "ph": "s", "id": 101701, "pid": 76337, "tid": -914061504, "ts": 1716454222958054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958120, "dur": 1,
    "args": {
      "External id": 101712,
      "cbid": 251, "correlation": 101712
    }
  },
  {
    "ph": "f", "id": 101712, "pid": 76337, "tid": -914061504, "ts": 1716454222958120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223006374, "dur": 144,
    "args": {
      "External id": 101713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101713,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101713, "pid": 5, "tid": 7, "ts": 1716454223006374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958124, "dur": 11,
    "args": {
      "External id": 101713,
      "cbid": 211, "correlation": 101713
    }
  },
  {
    "ph": "s", "id": 101713, "pid": 76337, "tid": -914061504, "ts": 1716454222958124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223006518, "dur": 1908,
    "args": {
      "External id": 101734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101734,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 101734, "pid": 5, "tid": 7, "ts": 1716454223006518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958205, "dur": 13,
    "args": {
      "External id": 101734,
      "cbid": 211, "correlation": 101734
    }
  },
  {
    "ph": "s", "id": 101734, "pid": 76337, "tid": -914061504, "ts": 1716454222958205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958305, "dur": 1,
    "args": {
      "External id": 101752,
      "cbid": 251, "correlation": 101752
    }
  },
  {
    "ph": "f", "id": 101752, "pid": 76337, "tid": -914061504, "ts": 1716454222958305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223008428, "dur": 144,
    "args": {
      "External id": 101754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101754,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 101754, "pid": 5, "tid": 7, "ts": 1716454223008428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958311, "dur": 13,
    "args": {
      "External id": 101754,
      "cbid": 211, "correlation": 101754
    }
  },
  {
    "ph": "s", "id": 101754, "pid": 76337, "tid": -914061504, "ts": 1716454222958311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223008574, "dur": 35,
    "args": {
      "External id": 101762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101762,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101762, "pid": 5, "tid": 7, "ts": 1716454223008574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958382, "dur": 13,
    "args": {
      "External id": 101762,
      "cbid": 211, "correlation": 101762
    }
  },
  {
    "ph": "s", "id": 101762, "pid": 76337, "tid": -914061504, "ts": 1716454222958382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223008610, "dur": 50,
    "args": {
      "External id": 101770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101770,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101770, "pid": 5, "tid": 7, "ts": 1716454223008610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958422, "dur": 8,
    "args": {
      "External id": 101770,
      "cbid": 211, "correlation": 101770
    }
  },
  {
    "ph": "s", "id": 101770, "pid": 76337, "tid": -914061504, "ts": 1716454222958422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223008662, "dur": 30,
    "args": {
      "External id": 101781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101781,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101781, "pid": 5, "tid": 7, "ts": 1716454223008662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958495, "dur": 12,
    "args": {
      "External id": 101781,
      "cbid": 211, "correlation": 101781
    }
  },
  {
    "ph": "s", "id": 101781, "pid": 76337, "tid": -914061504, "ts": 1716454222958495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223008693, "dur": 34,
    "args": {
      "External id": 101803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101803,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101803, "pid": 5, "tid": 7, "ts": 1716454223008693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958526, "dur": 8,
    "args": {
      "External id": 101803,
      "cbid": 211, "correlation": 101803
    }
  },
  {
    "ph": "s", "id": 101803, "pid": 76337, "tid": -914061504, "ts": 1716454222958526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958610, "dur": 1,
    "args": {
      "External id": 101814,
      "cbid": 251, "correlation": 101814
    }
  },
  {
    "ph": "f", "id": 101814, "pid": 76337, "tid": -914061504, "ts": 1716454222958610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223008729, "dur": 89,
    "args": {
      "External id": 101815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101815,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101815, "pid": 5, "tid": 7, "ts": 1716454223008729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958616, "dur": 13,
    "args": {
      "External id": 101815,
      "cbid": 211, "correlation": 101815
    }
  },
  {
    "ph": "s", "id": 101815, "pid": 76337, "tid": -914061504, "ts": 1716454222958616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958684, "dur": 1,
    "args": {
      "External id": 101826,
      "cbid": 251, "correlation": 101826
    }
  },
  {
    "ph": "f", "id": 101826, "pid": 76337, "tid": -914061504, "ts": 1716454222958684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958688, "dur": 0,
    "args": {
      "External id": 101827,
      "cbid": 251, "correlation": 101827
    }
  },
  {
    "ph": "f", "id": 101827, "pid": 76337, "tid": -914061504, "ts": 1716454222958688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223008819, "dur": 11,
    "args": {
      "External id": 101828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101828,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 101828, "pid": 5, "tid": 7, "ts": 1716454223008819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958690, "dur": 12,
    "args": {
      "External id": 101828,
      "cbid": 211, "correlation": 101828
    }
  },
  {
    "ph": "s", "id": 101828, "pid": 76337, "tid": -914061504, "ts": 1716454222958690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223008832, "dur": 5,
    "args": {
      "External id": 101830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101830,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 101830, "pid": 5, "tid": 7, "ts": 1716454223008832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958704, "dur": 6,
    "args": {
      "External id": 101830,
      "cbid": 211, "correlation": 101830
    }
  },
  {
    "ph": "s", "id": 101830, "pid": 76337, "tid": -914061504, "ts": 1716454222958704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958762, "dur": 1,
    "args": {
      "External id": 101841,
      "cbid": 251, "correlation": 101841
    }
  },
  {
    "ph": "f", "id": 101841, "pid": 76337, "tid": -914061504, "ts": 1716454222958762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958765, "dur": 0,
    "args": {
      "External id": 101842,
      "cbid": 251, "correlation": 101842
    }
  },
  {
    "ph": "f", "id": 101842, "pid": 76337, "tid": -914061504, "ts": 1716454222958765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223008839, "dur": 7,
    "args": {
      "External id": 101843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101843,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 101843, "pid": 5, "tid": 7, "ts": 1716454223008839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958767, "dur": 11,
    "args": {
      "External id": 101843,
      "cbid": 211, "correlation": 101843
    }
  },
  {
    "ph": "s", "id": 101843, "pid": 76337, "tid": -914061504, "ts": 1716454222958767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223008847, "dur": 4,
    "args": {
      "External id": 101845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101845,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 101845, "pid": 5, "tid": 7, "ts": 1716454223008847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958780, "dur": 6,
    "args": {
      "External id": 101845,
      "cbid": 211, "correlation": 101845
    }
  },
  {
    "ph": "s", "id": 101845, "pid": 76337, "tid": -914061504, "ts": 1716454222958780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223008852, "dur": 90,
    "args": {
      "External id": 101866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101866,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 101866, "pid": 5, "tid": 7, "ts": 1716454223008852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958854, "dur": 12,
    "args": {
      "External id": 101866,
      "cbid": 211, "correlation": 101866
    }
  },
  {
    "ph": "s", "id": 101866, "pid": 76337, "tid": -914061504, "ts": 1716454222958854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958953, "dur": 2,
    "args": {
      "External id": 101884,
      "cbid": 251, "correlation": 101884
    }
  },
  {
    "ph": "f", "id": 101884, "pid": 76337, "tid": -914061504, "ts": 1716454222958953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223008944, "dur": 96,
    "args": {
      "External id": 101886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101886,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101886, "pid": 5, "tid": 7, "ts": 1716454223008944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222958960, "dur": 21,
    "args": {
      "External id": 101886,
      "cbid": 211, "correlation": 101886
    }
  },
  {
    "ph": "s", "id": 101886, "pid": 76337, "tid": -914061504, "ts": 1716454222958960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223009041, "dur": 19,
    "args": {
      "External id": 101894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101894,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101894, "pid": 5, "tid": 7, "ts": 1716454223009041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959038, "dur": 13,
    "args": {
      "External id": 101894,
      "cbid": 211, "correlation": 101894
    }
  },
  {
    "ph": "s", "id": 101894, "pid": 76337, "tid": -914061504, "ts": 1716454222959038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223009061, "dur": 36,
    "args": {
      "External id": 101902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101902, "pid": 5, "tid": 7, "ts": 1716454223009061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959081, "dur": 10,
    "args": {
      "External id": 101902,
      "cbid": 211, "correlation": 101902
    }
  },
  {
    "ph": "s", "id": 101902, "pid": 76337, "tid": -914061504, "ts": 1716454222959081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223009099, "dur": 33,
    "args": {
      "External id": 101924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101924,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101924, "pid": 5, "tid": 7, "ts": 1716454223009099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959132, "dur": 11,
    "args": {
      "External id": 101924,
      "cbid": 211, "correlation": 101924
    }
  },
  {
    "ph": "s", "id": 101924, "pid": 76337, "tid": -914061504, "ts": 1716454222959132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959222, "dur": 1,
    "args": {
      "External id": 101940,
      "cbid": 251, "correlation": 101940
    }
  },
  {
    "ph": "f", "id": 101940, "pid": 76337, "tid": -914061504, "ts": 1716454222959222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959227, "dur": 0,
    "args": {
      "External id": 101942,
      "cbid": 251, "correlation": 101942
    }
  },
  {
    "ph": "f", "id": 101942, "pid": 76337, "tid": -914061504, "ts": 1716454222959227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223009133, "dur": 531,
    "args": {
      "External id": 101943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101943,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 101943, "pid": 5, "tid": 7, "ts": 1716454223009133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959230, "dur": 13,
    "args": {
      "External id": 101943,
      "cbid": 211, "correlation": 101943
    }
  },
  {
    "ph": "s", "id": 101943, "pid": 76337, "tid": -914061504, "ts": 1716454222959230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223009665, "dur": 123,
    "args": {
      "External id": 101951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101951,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101951, "pid": 5, "tid": 7, "ts": 1716454223009665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959296, "dur": 12,
    "args": {
      "External id": 101951,
      "cbid": 211, "correlation": 101951
    }
  },
  {
    "ph": "s", "id": 101951, "pid": 76337, "tid": -914061504, "ts": 1716454222959296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223009790, "dur": 129,
    "args": {
      "External id": 101959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101959, "pid": 5, "tid": 7, "ts": 1716454223009790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959326, "dur": 8,
    "args": {
      "External id": 101959,
      "cbid": 211, "correlation": 101959
    }
  },
  {
    "ph": "s", "id": 101959, "pid": 76337, "tid": -914061504, "ts": 1716454222959326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959403, "dur": 1,
    "args": {
      "External id": 101975,
      "cbid": 251, "correlation": 101975
    }
  },
  {
    "ph": "f", "id": 101975, "pid": 76337, "tid": -914061504, "ts": 1716454222959403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223009919, "dur": 301,
    "args": {
      "External id": 101977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101977,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 101977, "pid": 5, "tid": 7, "ts": 1716454223009919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959408, "dur": 12,
    "args": {
      "External id": 101977,
      "cbid": 211, "correlation": 101977
    }
  },
  {
    "ph": "s", "id": 101977, "pid": 76337, "tid": -914061504, "ts": 1716454222959408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223010222, "dur": 27,
    "args": {
      "External id": 101985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101985,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101985, "pid": 5, "tid": 7, "ts": 1716454223010222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959451, "dur": 10,
    "args": {
      "External id": 101985,
      "cbid": 211, "correlation": 101985
    }
  },
  {
    "ph": "s", "id": 101985, "pid": 76337, "tid": -914061504, "ts": 1716454222959451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223010250, "dur": 80,
    "args": {
      "External id": 101996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 101996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 101996, "pid": 5, "tid": 7, "ts": 1716454223010250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959518, "dur": 12,
    "args": {
      "External id": 101996,
      "cbid": 211, "correlation": 101996
    }
  },
  {
    "ph": "s", "id": 101996, "pid": 76337, "tid": -914061504, "ts": 1716454222959518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959581, "dur": 0,
    "args": {
      "External id": 102008,
      "cbid": 317, "correlation": 102008
    }
  },
  {
    "ph": "f", "id": 102008, "pid": 76337, "tid": -914061504, "ts": 1716454222959581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959582, "dur": 0,
    "args": {
      "External id": 102009,
      "cbid": 203, "correlation": 102009
    }
  },
  {
    "ph": "f", "id": 102009, "pid": 76337, "tid": -914061504, "ts": 1716454222959582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959583, "dur": 0,
    "args": {
      "External id": 102010,
      "cbid": 205, "correlation": 102010
    }
  },
  {
    "ph": "f", "id": 102010, "pid": 76337, "tid": -914061504, "ts": 1716454222959583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010331, "dur": 23,
    "args": {
      "External id": 102014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102014, "pid": 5, "tid": 7, "ts": 1716454223010331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959599, "dur": 12,
    "args": {
      "External id": 102014,
      "cbid": 211, "correlation": 102014
    }
  },
  {
    "ph": "s", "id": 102014, "pid": 76337, "tid": -914061504, "ts": 1716454222959599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223010355, "dur": 119,
    "args": {
      "External id": 102016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102016,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102016, "pid": 5, "tid": 7, "ts": 1716454223010355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959617, "dur": 6,
    "args": {
      "External id": 102016,
      "cbid": 211, "correlation": 102016
    }
  },
  {
    "ph": "s", "id": 102016, "pid": 76337, "tid": -914061504, "ts": 1716454222959617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010475, "dur": 23,
    "args": {
      "External id": 102018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102018,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102018, "pid": 5, "tid": 7, "ts": 1716454223010475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959627, "dur": 5,
    "args": {
      "External id": 102018,
      "cbid": 211, "correlation": 102018
    }
  },
  {
    "ph": "s", "id": 102018, "pid": 76337, "tid": -914061504, "ts": 1716454222959627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223010500, "dur": 32,
    "args": {
      "External id": 102024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102024,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102024, "pid": 5, "tid": 7, "ts": 1716454223010500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959655, "dur": 8,
    "args": {
      "External id": 102024,
      "cbid": 211, "correlation": 102024
    }
  },
  {
    "ph": "s", "id": 102024, "pid": 76337, "tid": -914061504, "ts": 1716454222959655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223010533, "dur": 27,
    "args": {
      "External id": 102032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102032,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102032, "pid": 5, "tid": 7, "ts": 1716454223010533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959687, "dur": 9,
    "args": {
      "External id": 102032,
      "cbid": 211, "correlation": 102032
    }
  },
  {
    "ph": "s", "id": 102032, "pid": 76337, "tid": -914061504, "ts": 1716454222959687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959758, "dur": 0,
    "args": {
      "External id": 102042,
      "cbid": 317, "correlation": 102042
    }
  },
  {
    "ph": "f", "id": 102042, "pid": 76337, "tid": -914061504, "ts": 1716454222959758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959759, "dur": 0,
    "args": {
      "External id": 102043,
      "cbid": 203, "correlation": 102043
    }
  },
  {
    "ph": "f", "id": 102043, "pid": 76337, "tid": -914061504, "ts": 1716454222959759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959760, "dur": 0,
    "args": {
      "External id": 102044,
      "cbid": 205, "correlation": 102044
    }
  },
  {
    "ph": "f", "id": 102044, "pid": 76337, "tid": -914061504, "ts": 1716454222959760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010561, "dur": 23,
    "args": {
      "External id": 102048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102048, "pid": 5, "tid": 7, "ts": 1716454223010561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959774, "dur": 12,
    "args": {
      "External id": 102048,
      "cbid": 211, "correlation": 102048
    }
  },
  {
    "ph": "s", "id": 102048, "pid": 76337, "tid": -914061504, "ts": 1716454222959774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010586, "dur": 44,
    "args": {
      "External id": 102050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102050, "pid": 5, "tid": 7, "ts": 1716454223010586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959789, "dur": 6,
    "args": {
      "External id": 102050,
      "cbid": 211, "correlation": 102050
    }
  },
  {
    "ph": "s", "id": 102050, "pid": 76337, "tid": -914061504, "ts": 1716454222959789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223010631, "dur": 232,
    "args": {
      "External id": 102052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102052,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 102052, "pid": 5, "tid": 7, "ts": 1716454223010631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959801, "dur": 7,
    "args": {
      "External id": 102052,
      "cbid": 211, "correlation": 102052
    }
  },
  {
    "ph": "s", "id": 102052, "pid": 76337, "tid": -914061504, "ts": 1716454222959801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010864, "dur": 6,
    "args": {
      "External id": 102054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102054, "pid": 5, "tid": 7, "ts": 1716454223010864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959812, "dur": 5,
    "args": {
      "External id": 102054,
      "cbid": 211, "correlation": 102054
    }
  },
  {
    "ph": "s", "id": 102054, "pid": 76337, "tid": -914061504, "ts": 1716454222959812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223010872, "dur": 9,
    "args": {
      "External id": 102060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102060, "pid": 5, "tid": 7, "ts": 1716454223010872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959839, "dur": 9,
    "args": {
      "External id": 102060,
      "cbid": 211, "correlation": 102060
    }
  },
  {
    "ph": "s", "id": 102060, "pid": 76337, "tid": -914061504, "ts": 1716454222959839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010882, "dur": 12,
    "args": {
      "External id": 102080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102080,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 102080, "pid": 5, "tid": 7, "ts": 1716454223010882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959931, "dur": 13,
    "args": {
      "External id": 102080,
      "cbid": 211, "correlation": 102080
    }
  },
  {
    "ph": "s", "id": 102080, "pid": 76337, "tid": -914061504, "ts": 1716454222959931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223010895, "dur": 5,
    "args": {
      "External id": 102092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102092,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 102092, "pid": 5, "tid": 7, "ts": 1716454223010895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959954, "dur": 6,
    "args": {
      "External id": 102092,
      "cbid": 211, "correlation": 102092
    }
  },
  {
    "ph": "s", "id": 102092, "pid": 76337, "tid": -914061504, "ts": 1716454222959954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223010901, "dur": 12,
    "args": {
      "External id": 102095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102095,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102095, "pid": 5, "tid": 7, "ts": 1716454223010901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222959972, "dur": 16,
    "args": {
      "External id": 102095,
      "cbid": 211, "correlation": 102095
    }
  },
  {
    "ph": "s", "id": 102095, "pid": 76337, "tid": -914061504, "ts": 1716454222959972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223010914, "dur": 7,
    "args": {
      "External id": 102104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102104,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102104, "pid": 5, "tid": 7, "ts": 1716454223010914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960022, "dur": 10,
    "args": {
      "External id": 102104,
      "cbid": 211, "correlation": 102104
    }
  },
  {
    "ph": "s", "id": 102104, "pid": 76337, "tid": -914061504, "ts": 1716454222960022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960075, "dur": 0,
    "args": {
      "External id": 102114,
      "cbid": 317, "correlation": 102114
    }
  },
  {
    "ph": "f", "id": 102114, "pid": 76337, "tid": -914061504, "ts": 1716454222960075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960076, "dur": 0,
    "args": {
      "External id": 102115,
      "cbid": 203, "correlation": 102115
    }
  },
  {
    "ph": "f", "id": 102115, "pid": 76337, "tid": -914061504, "ts": 1716454222960076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960076, "dur": 0,
    "args": {
      "External id": 102116,
      "cbid": 205, "correlation": 102116
    }
  },
  {
    "ph": "f", "id": 102116, "pid": 76337, "tid": -914061504, "ts": 1716454222960076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010922, "dur": 5,
    "args": {
      "External id": 102120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102120,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102120, "pid": 5, "tid": 7, "ts": 1716454223010922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960093, "dur": 11,
    "args": {
      "External id": 102120,
      "cbid": 211, "correlation": 102120
    }
  },
  {
    "ph": "s", "id": 102120, "pid": 76337, "tid": -914061504, "ts": 1716454222960093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223010929, "dur": 82,
    "args": {
      "External id": 102122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102122, "pid": 5, "tid": 7, "ts": 1716454223010929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960108, "dur": 5,
    "args": {
      "External id": 102122,
      "cbid": 211, "correlation": 102122
    }
  },
  {
    "ph": "s", "id": 102122, "pid": 76337, "tid": -914061504, "ts": 1716454222960108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223011014, "dur": 1,
    "args": {
      "External id": 102124,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 102124,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 102124, "pid": 5, "tid": 7, "ts": 1716454223011014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960120, "dur": 8,
    "args": {
      "External id": 102124,
      "cbid": 51, "correlation": 102124
    }
  },
  {
    "ph": "s", "id": 102124, "pid": 76337, "tid": -914061504, "ts": 1716454222960120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223011017, "dur": 535,
    "args": {
      "External id": 102125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102125,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102125, "pid": 5, "tid": 7, "ts": 1716454223011017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960130, "dur": 9,
    "args": {
      "External id": 102125,
      "cbid": 211, "correlation": 102125
    }
  },
  {
    "ph": "s", "id": 102125, "pid": 76337, "tid": -914061504, "ts": 1716454222960130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223011554, "dur": 12,
    "args": {
      "External id": 102127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102127,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102127, "pid": 5, "tid": 7, "ts": 1716454223011554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960143, "dur": 5,
    "args": {
      "External id": 102127,
      "cbid": 211, "correlation": 102127
    }
  },
  {
    "ph": "s", "id": 102127, "pid": 76337, "tid": -914061504, "ts": 1716454222960143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223011567, "dur": 14,
    "args": {
      "External id": 102133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102133, "pid": 5, "tid": 7, "ts": 1716454223011567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960171, "dur": 8,
    "args": {
      "External id": 102133,
      "cbid": 211, "correlation": 102133
    }
  },
  {
    "ph": "s", "id": 102133, "pid": 76337, "tid": -914061504, "ts": 1716454222960171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223011583, "dur": 3,
    "args": {
      "External id": 102141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102141,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 102141, "pid": 5, "tid": 7, "ts": 1716454223011583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960214, "dur": 9,
    "args": {
      "External id": 102141,
      "cbid": 211, "correlation": 102141
    }
  },
  {
    "ph": "s", "id": 102141, "pid": 76337, "tid": -914061504, "ts": 1716454222960214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960281, "dur": 1,
    "args": {
      "External id": 102157,
      "cbid": 251, "correlation": 102157
    }
  },
  {
    "ph": "f", "id": 102157, "pid": 76337, "tid": -914061504, "ts": 1716454222960281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960287, "dur": 0,
    "args": {
      "External id": 102159,
      "cbid": 251, "correlation": 102159
    }
  },
  {
    "ph": "f", "id": 102159, "pid": 76337, "tid": -914061504, "ts": 1716454222960287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223011587, "dur": 14,
    "args": {
      "External id": 102160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102160,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102160, "pid": 5, "tid": 7, "ts": 1716454223011587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960289, "dur": 11,
    "args": {
      "External id": 102160,
      "cbid": 211, "correlation": 102160
    }
  },
  {
    "ph": "s", "id": 102160, "pid": 76337, "tid": -914061504, "ts": 1716454222960289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223011602, "dur": 5,
    "args": {
      "External id": 102162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102162,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102162, "pid": 5, "tid": 7, "ts": 1716454223011602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960302, "dur": 5,
    "args": {
      "External id": 102162,
      "cbid": 211, "correlation": 102162
    }
  },
  {
    "ph": "s", "id": 102162, "pid": 76337, "tid": -914061504, "ts": 1716454222960302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223011609, "dur": 18,
    "args": {
      "External id": 102172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102172,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102172, "pid": 5, "tid": 7, "ts": 1716454223011609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960360, "dur": 12,
    "args": {
      "External id": 102172,
      "cbid": 211, "correlation": 102172
    }
  },
  {
    "ph": "s", "id": 102172, "pid": 76337, "tid": -914061504, "ts": 1716454222960360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223011628, "dur": 17,
    "args": {
      "External id": 102192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102192,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 102192, "pid": 5, "tid": 7, "ts": 1716454223011628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960426, "dur": 12,
    "args": {
      "External id": 102192,
      "cbid": 211, "correlation": 102192
    }
  },
  {
    "ph": "s", "id": 102192, "pid": 76337, "tid": -914061504, "ts": 1716454222960426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223011646, "dur": 4,
    "args": {
      "External id": 102204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102204,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 102204, "pid": 5, "tid": 7, "ts": 1716454223011646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960447, "dur": 6,
    "args": {
      "External id": 102204,
      "cbid": 211, "correlation": 102204
    }
  },
  {
    "ph": "s", "id": 102204, "pid": 76337, "tid": -914061504, "ts": 1716454222960447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223011652, "dur": 18,
    "args": {
      "External id": 102207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102207,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102207, "pid": 5, "tid": 7, "ts": 1716454223011652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960466, "dur": 6,
    "args": {
      "External id": 102207,
      "cbid": 211, "correlation": 102207
    }
  },
  {
    "ph": "s", "id": 102207, "pid": 76337, "tid": -914061504, "ts": 1716454222960466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223011671, "dur": 10,
    "args": {
      "External id": 102216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102216,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102216, "pid": 5, "tid": 7, "ts": 1716454223011671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960506, "dur": 9,
    "args": {
      "External id": 102216,
      "cbid": 211, "correlation": 102216
    }
  },
  {
    "ph": "s", "id": 102216, "pid": 76337, "tid": -914061504, "ts": 1716454222960506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960568, "dur": 0,
    "args": {
      "External id": 102226,
      "cbid": 317, "correlation": 102226
    }
  },
  {
    "ph": "f", "id": 102226, "pid": 76337, "tid": -914061504, "ts": 1716454222960568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960569, "dur": 0,
    "args": {
      "External id": 102227,
      "cbid": 203, "correlation": 102227
    }
  },
  {
    "ph": "f", "id": 102227, "pid": 76337, "tid": -914061504, "ts": 1716454222960569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960570, "dur": 0,
    "args": {
      "External id": 102228,
      "cbid": 205, "correlation": 102228
    }
  },
  {
    "ph": "f", "id": 102228, "pid": 76337, "tid": -914061504, "ts": 1716454222960570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223011682, "dur": 11,
    "args": {
      "External id": 102232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102232,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102232, "pid": 5, "tid": 7, "ts": 1716454223011682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960588, "dur": 12,
    "args": {
      "External id": 102232,
      "cbid": 211, "correlation": 102232
    }
  },
  {
    "ph": "s", "id": 102232, "pid": 76337, "tid": -914061504, "ts": 1716454222960588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223011694, "dur": 161,
    "args": {
      "External id": 102234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102234, "pid": 5, "tid": 7, "ts": 1716454223011694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960602, "dur": 5,
    "args": {
      "External id": 102234,
      "cbid": 211, "correlation": 102234
    }
  },
  {
    "ph": "s", "id": 102234, "pid": 76337, "tid": -914061504, "ts": 1716454222960602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223011858, "dur": 1,
    "args": {
      "External id": 102236,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 102236,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 102236, "pid": 5, "tid": 7, "ts": 1716454223011858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960614, "dur": 7,
    "args": {
      "External id": 102236,
      "cbid": 51, "correlation": 102236
    }
  },
  {
    "ph": "s", "id": 102236, "pid": 76337, "tid": -914061504, "ts": 1716454222960614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223011861, "dur": 655,
    "args": {
      "External id": 102237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102237,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102237, "pid": 5, "tid": 7, "ts": 1716454223011861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960622, "dur": 7,
    "args": {
      "External id": 102237,
      "cbid": 211, "correlation": 102237
    }
  },
  {
    "ph": "s", "id": 102237, "pid": 76337, "tid": -914061504, "ts": 1716454222960622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223012518, "dur": 11,
    "args": {
      "External id": 102239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102239,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102239, "pid": 5, "tid": 7, "ts": 1716454223012518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960634, "dur": 6,
    "args": {
      "External id": 102239,
      "cbid": 211, "correlation": 102239
    }
  },
  {
    "ph": "s", "id": 102239, "pid": 76337, "tid": -914061504, "ts": 1716454222960634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223012530, "dur": 14,
    "args": {
      "External id": 102245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102245,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102245, "pid": 5, "tid": 7, "ts": 1716454223012530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960662, "dur": 8,
    "args": {
      "External id": 102245,
      "cbid": 211, "correlation": 102245
    }
  },
  {
    "ph": "s", "id": 102245, "pid": 76337, "tid": -914061504, "ts": 1716454222960662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960720, "dur": 0,
    "args": {
      "External id": 102255,
      "cbid": 317, "correlation": 102255
    }
  },
  {
    "ph": "f", "id": 102255, "pid": 76337, "tid": -914061504, "ts": 1716454222960720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960720, "dur": 0,
    "args": {
      "External id": 102256,
      "cbid": 203, "correlation": 102256
    }
  },
  {
    "ph": "f", "id": 102256, "pid": 76337, "tid": -914061504, "ts": 1716454222960720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960721, "dur": 0,
    "args": {
      "External id": 102257,
      "cbid": 205, "correlation": 102257
    }
  },
  {
    "ph": "f", "id": 102257, "pid": 76337, "tid": -914061504, "ts": 1716454222960721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223012546, "dur": 8,
    "args": {
      "External id": 102261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102261,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102261, "pid": 5, "tid": 7, "ts": 1716454223012546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960736, "dur": 11,
    "args": {
      "External id": 102261,
      "cbid": 211, "correlation": 102261
    }
  },
  {
    "ph": "s", "id": 102261, "pid": 76337, "tid": -914061504, "ts": 1716454222960736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223012556, "dur": 3,
    "args": {
      "External id": 102263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102263, "pid": 5, "tid": 7, "ts": 1716454223012556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960753, "dur": 6,
    "args": {
      "External id": 102263,
      "cbid": 211, "correlation": 102263
    }
  },
  {
    "ph": "s", "id": 102263, "pid": 76337, "tid": -914061504, "ts": 1716454222960753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960762, "dur": 0,
    "args": {
      "External id": 102264,
      "cbid": 51, "correlation": 102264
    }
  },
  {
    "ph": "s", "id": 102264, "pid": 76337, "tid": -914061504, "ts": 1716454222960762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223012560, "dur": 56,
    "args": {
      "External id": 102265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102265,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 102265, "pid": 5, "tid": 7, "ts": 1716454223012560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960763, "dur": 5,
    "args": {
      "External id": 102265,
      "cbid": 211, "correlation": 102265
    }
  },
  {
    "ph": "s", "id": 102265, "pid": 76337, "tid": -914061504, "ts": 1716454222960763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223012618, "dur": 14,
    "args": {
      "External id": 102270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102270, "pid": 5, "tid": 7, "ts": 1716454223012618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960789, "dur": 8,
    "args": {
      "External id": 102270,
      "cbid": 211, "correlation": 102270
    }
  },
  {
    "ph": "s", "id": 102270, "pid": 76337, "tid": -914061504, "ts": 1716454222960789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223012633, "dur": 11,
    "args": {
      "External id": 102278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102278,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102278, "pid": 5, "tid": 7, "ts": 1716454223012633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960816, "dur": 7,
    "args": {
      "External id": 102278,
      "cbid": 211, "correlation": 102278
    }
  },
  {
    "ph": "s", "id": 102278, "pid": 76337, "tid": -914061504, "ts": 1716454222960816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223012645, "dur": 10,
    "args": {
      "External id": 102286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102286,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102286, "pid": 5, "tid": 7, "ts": 1716454223012645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960846, "dur": 8,
    "args": {
      "External id": 102286,
      "cbid": 211, "correlation": 102286
    }
  },
  {
    "ph": "s", "id": 102286, "pid": 76337, "tid": -914061504, "ts": 1716454222960846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223012656, "dur": 18,
    "args": {
      "External id": 102306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102306,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 102306, "pid": 5, "tid": 7, "ts": 1716454223012656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960925, "dur": 13,
    "args": {
      "External id": 102306,
      "cbid": 211, "correlation": 102306
    }
  },
  {
    "ph": "s", "id": 102306, "pid": 76337, "tid": -914061504, "ts": 1716454222960925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223012675, "dur": 5,
    "args": {
      "External id": 102318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102318,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 102318, "pid": 5, "tid": 7, "ts": 1716454223012675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960948, "dur": 6,
    "args": {
      "External id": 102318,
      "cbid": 211, "correlation": 102318
    }
  },
  {
    "ph": "s", "id": 102318, "pid": 76337, "tid": -914061504, "ts": 1716454222960948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223012681, "dur": 17,
    "args": {
      "External id": 102321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102321, "pid": 5, "tid": 7, "ts": 1716454223012681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222960966, "dur": 6,
    "args": {
      "External id": 102321,
      "cbid": 211, "correlation": 102321
    }
  },
  {
    "ph": "s", "id": 102321, "pid": 76337, "tid": -914061504, "ts": 1716454222960966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961032, "dur": 0,
    "args": {
      "External id": 102332,
      "cbid": 317, "correlation": 102332
    }
  },
  {
    "ph": "f", "id": 102332, "pid": 76337, "tid": -914061504, "ts": 1716454222961032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961033, "dur": 0,
    "args": {
      "External id": 102333,
      "cbid": 203, "correlation": 102333
    }
  },
  {
    "ph": "f", "id": 102333, "pid": 76337, "tid": -914061504, "ts": 1716454222961033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961034, "dur": 0,
    "args": {
      "External id": 102334,
      "cbid": 205, "correlation": 102334
    }
  },
  {
    "ph": "f", "id": 102334, "pid": 76337, "tid": -914061504, "ts": 1716454222961034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223012700, "dur": 12,
    "args": {
      "External id": 102338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102338,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102338, "pid": 5, "tid": 7, "ts": 1716454223012700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961050, "dur": 12,
    "args": {
      "External id": 102338,
      "cbid": 211, "correlation": 102338
    }
  },
  {
    "ph": "s", "id": 102338, "pid": 76337, "tid": -914061504, "ts": 1716454222961050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223012713, "dur": 3,
    "args": {
      "External id": 102340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102340, "pid": 5, "tid": 7, "ts": 1716454223012713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961066, "dur": 6,
    "args": {
      "External id": 102340,
      "cbid": 211, "correlation": 102340
    }
  },
  {
    "ph": "s", "id": 102340, "pid": 76337, "tid": -914061504, "ts": 1716454222961066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961076, "dur": 0,
    "args": {
      "External id": 102341,
      "cbid": 51, "correlation": 102341
    }
  },
  {
    "ph": "s", "id": 102341, "pid": 76337, "tid": -914061504, "ts": 1716454222961076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223012718, "dur": 95,
    "args": {
      "External id": 102342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102342,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 102342, "pid": 5, "tid": 7, "ts": 1716454223012718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961077, "dur": 7,
    "args": {
      "External id": 102342,
      "cbid": 211, "correlation": 102342
    }
  },
  {
    "ph": "s", "id": 102342, "pid": 76337, "tid": -914061504, "ts": 1716454222961077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223012814, "dur": 15,
    "args": {
      "External id": 102347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102347, "pid": 5, "tid": 7, "ts": 1716454223012814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961105, "dur": 8,
    "args": {
      "External id": 102347,
      "cbid": 211, "correlation": 102347
    }
  },
  {
    "ph": "s", "id": 102347, "pid": 76337, "tid": -914061504, "ts": 1716454222961105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223012831, "dur": 83,
    "args": {
      "External id": 102356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102356,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102356, "pid": 5, "tid": 7, "ts": 1716454223012831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961188, "dur": 14,
    "args": {
      "External id": 102356,
      "cbid": 211, "correlation": 102356
    }
  },
  {
    "ph": "s", "id": 102356, "pid": 76337, "tid": -914061504, "ts": 1716454222961188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223012915, "dur": 31,
    "args": {
      "External id": 102378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102378,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102378, "pid": 5, "tid": 7, "ts": 1716454223012915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961244, "dur": 10,
    "args": {
      "External id": 102378,
      "cbid": 211, "correlation": 102378
    }
  },
  {
    "ph": "s", "id": 102378, "pid": 76337, "tid": -914061504, "ts": 1716454222961244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961338, "dur": 1,
    "args": {
      "External id": 102389,
      "cbid": 251, "correlation": 102389
    }
  },
  {
    "ph": "f", "id": 102389, "pid": 76337, "tid": -914061504, "ts": 1716454222961338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223012947, "dur": 163,
    "args": {
      "External id": 102390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102390,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102390, "pid": 5, "tid": 7, "ts": 1716454223012947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961343, "dur": 13,
    "args": {
      "External id": 102390,
      "cbid": 211, "correlation": 102390
    }
  },
  {
    "ph": "s", "id": 102390, "pid": 76337, "tid": -914061504, "ts": 1716454222961343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961413, "dur": 1,
    "args": {
      "External id": 102401,
      "cbid": 251, "correlation": 102401
    }
  },
  {
    "ph": "f", "id": 102401, "pid": 76337, "tid": -914061504, "ts": 1716454222961413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223013110, "dur": 158,
    "args": {
      "External id": 102402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102402,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102402, "pid": 5, "tid": 7, "ts": 1716454223013110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961417, "dur": 11,
    "args": {
      "External id": 102402,
      "cbid": 211, "correlation": 102402
    }
  },
  {
    "ph": "s", "id": 102402, "pid": 76337, "tid": -914061504, "ts": 1716454222961417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961484, "dur": 1,
    "args": {
      "External id": 102413,
      "cbid": 251, "correlation": 102413
    }
  },
  {
    "ph": "f", "id": 102413, "pid": 76337, "tid": -914061504, "ts": 1716454222961484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223013270, "dur": 156,
    "args": {
      "External id": 102414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102414,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102414, "pid": 5, "tid": 7, "ts": 1716454223013270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961488, "dur": 11,
    "args": {
      "External id": 102414,
      "cbid": 211, "correlation": 102414
    }
  },
  {
    "ph": "s", "id": 102414, "pid": 76337, "tid": -914061504, "ts": 1716454222961488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223013427, "dur": 332,
    "args": {
      "External id": 102439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102439,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102439, "pid": 5, "tid": 7, "ts": 1716454223013427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961576, "dur": 14,
    "args": {
      "External id": 102439,
      "cbid": 211, "correlation": 102439
    }
  },
  {
    "ph": "s", "id": 102439, "pid": 76337, "tid": -914061504, "ts": 1716454222961576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961677, "dur": 1,
    "args": {
      "External id": 102457,
      "cbid": 251, "correlation": 102457
    }
  },
  {
    "ph": "f", "id": 102457, "pid": 76337, "tid": -914061504, "ts": 1716454222961677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223013760, "dur": 164,
    "args": {
      "External id": 102459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102459,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102459, "pid": 5, "tid": 7, "ts": 1716454223013760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961683, "dur": 13,
    "args": {
      "External id": 102459,
      "cbid": 211, "correlation": 102459
    }
  },
  {
    "ph": "s", "id": 102459, "pid": 76337, "tid": -914061504, "ts": 1716454222961683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223013926, "dur": 19,
    "args": {
      "External id": 102467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102467,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102467, "pid": 5, "tid": 7, "ts": 1716454223013926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961754, "dur": 12,
    "args": {
      "External id": 102467,
      "cbid": 211, "correlation": 102467
    }
  },
  {
    "ph": "s", "id": 102467, "pid": 76337, "tid": -914061504, "ts": 1716454222961754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223013946, "dur": 29,
    "args": {
      "External id": 102475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102475,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102475, "pid": 5, "tid": 7, "ts": 1716454223013946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961792, "dur": 8,
    "args": {
      "External id": 102475,
      "cbid": 211, "correlation": 102475
    }
  },
  {
    "ph": "s", "id": 102475, "pid": 76337, "tid": -914061504, "ts": 1716454222961792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223013976, "dur": 18,
    "args": {
      "External id": 102486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102486, "pid": 5, "tid": 7, "ts": 1716454223013976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961864, "dur": 12,
    "args": {
      "External id": 102486,
      "cbid": 211, "correlation": 102486
    }
  },
  {
    "ph": "s", "id": 102486, "pid": 76337, "tid": -914061504, "ts": 1716454222961864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223013995, "dur": 16,
    "args": {
      "External id": 102508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102508,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102508, "pid": 5, "tid": 7, "ts": 1716454223013995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961896, "dur": 8,
    "args": {
      "External id": 102508,
      "cbid": 211, "correlation": 102508
    }
  },
  {
    "ph": "s", "id": 102508, "pid": 76337, "tid": -914061504, "ts": 1716454222961896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961991, "dur": 2,
    "args": {
      "External id": 102519,
      "cbid": 251, "correlation": 102519
    }
  },
  {
    "ph": "f", "id": 102519, "pid": 76337, "tid": -914061504, "ts": 1716454222961991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223014012, "dur": 88,
    "args": {
      "External id": 102520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102520,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102520, "pid": 5, "tid": 7, "ts": 1716454223014012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222961998, "dur": 15,
    "args": {
      "External id": 102520,
      "cbid": 211, "correlation": 102520
    }
  },
  {
    "ph": "s", "id": 102520, "pid": 76337, "tid": -914061504, "ts": 1716454222961998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962071, "dur": 1,
    "args": {
      "External id": 102531,
      "cbid": 251, "correlation": 102531
    }
  },
  {
    "ph": "f", "id": 102531, "pid": 76337, "tid": -914061504, "ts": 1716454222962071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962075, "dur": 0,
    "args": {
      "External id": 102532,
      "cbid": 251, "correlation": 102532
    }
  },
  {
    "ph": "f", "id": 102532, "pid": 76337, "tid": -914061504, "ts": 1716454222962075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223014102, "dur": 12,
    "args": {
      "External id": 102533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102533,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102533, "pid": 5, "tid": 7, "ts": 1716454223014102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962076, "dur": 12,
    "args": {
      "External id": 102533,
      "cbid": 211, "correlation": 102533
    }
  },
  {
    "ph": "s", "id": 102533, "pid": 76337, "tid": -914061504, "ts": 1716454222962076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223014115, "dur": 6,
    "args": {
      "External id": 102535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102535,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102535, "pid": 5, "tid": 7, "ts": 1716454223014115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962091, "dur": 6,
    "args": {
      "External id": 102535,
      "cbid": 211, "correlation": 102535
    }
  },
  {
    "ph": "s", "id": 102535, "pid": 76337, "tid": -914061504, "ts": 1716454222962091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962149, "dur": 1,
    "args": {
      "External id": 102546,
      "cbid": 251, "correlation": 102546
    }
  },
  {
    "ph": "f", "id": 102546, "pid": 76337, "tid": -914061504, "ts": 1716454222962149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962152, "dur": 0,
    "args": {
      "External id": 102547,
      "cbid": 251, "correlation": 102547
    }
  },
  {
    "ph": "f", "id": 102547, "pid": 76337, "tid": -914061504, "ts": 1716454222962152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223014122, "dur": 8,
    "args": {
      "External id": 102548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102548,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102548, "pid": 5, "tid": 7, "ts": 1716454223014122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962153, "dur": 11,
    "args": {
      "External id": 102548,
      "cbid": 211, "correlation": 102548
    }
  },
  {
    "ph": "s", "id": 102548, "pid": 76337, "tid": -914061504, "ts": 1716454222962153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223014131, "dur": 3,
    "args": {
      "External id": 102550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102550,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102550, "pid": 5, "tid": 7, "ts": 1716454223014131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962166, "dur": 6,
    "args": {
      "External id": 102550,
      "cbid": 211, "correlation": 102550
    }
  },
  {
    "ph": "s", "id": 102550, "pid": 76337, "tid": -914061504, "ts": 1716454222962166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223014136, "dur": 54,
    "args": {
      "External id": 102575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102575,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102575, "pid": 5, "tid": 7, "ts": 1716454223014136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962244, "dur": 12,
    "args": {
      "External id": 102575,
      "cbid": 211, "correlation": 102575
    }
  },
  {
    "ph": "s", "id": 102575, "pid": 76337, "tid": -914061504, "ts": 1716454222962244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962344, "dur": 2,
    "args": {
      "External id": 102593,
      "cbid": 251, "correlation": 102593
    }
  },
  {
    "ph": "f", "id": 102593, "pid": 76337, "tid": -914061504, "ts": 1716454222962344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223014191, "dur": 90,
    "args": {
      "External id": 102595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102595,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102595, "pid": 5, "tid": 7, "ts": 1716454223014191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962351, "dur": 14,
    "args": {
      "External id": 102595,
      "cbid": 211, "correlation": 102595
    }
  },
  {
    "ph": "s", "id": 102595, "pid": 76337, "tid": -914061504, "ts": 1716454222962351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223014282, "dur": 9,
    "args": {
      "External id": 102603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102603,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102603, "pid": 5, "tid": 7, "ts": 1716454223014282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962421, "dur": 12,
    "args": {
      "External id": 102603,
      "cbid": 211, "correlation": 102603
    }
  },
  {
    "ph": "s", "id": 102603, "pid": 76337, "tid": -914061504, "ts": 1716454222962421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223014293, "dur": 22,
    "args": {
      "External id": 102611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102611, "pid": 5, "tid": 7, "ts": 1716454223014293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962463, "dur": 9,
    "args": {
      "External id": 102611,
      "cbid": 211, "correlation": 102611
    }
  },
  {
    "ph": "s", "id": 102611, "pid": 76337, "tid": -914061504, "ts": 1716454222962463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223014316, "dur": 18,
    "args": {
      "External id": 102633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102633,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102633, "pid": 5, "tid": 7, "ts": 1716454223014316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962515, "dur": 10,
    "args": {
      "External id": 102633,
      "cbid": 211, "correlation": 102633
    }
  },
  {
    "ph": "s", "id": 102633, "pid": 76337, "tid": -914061504, "ts": 1716454222962515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962605, "dur": 2,
    "args": {
      "External id": 102649,
      "cbid": 251, "correlation": 102649
    }
  },
  {
    "ph": "f", "id": 102649, "pid": 76337, "tid": -914061504, "ts": 1716454222962605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962610, "dur": 0,
    "args": {
      "External id": 102651,
      "cbid": 251, "correlation": 102651
    }
  },
  {
    "ph": "f", "id": 102651, "pid": 76337, "tid": -914061504, "ts": 1716454222962610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223014336, "dur": 490,
    "args": {
      "External id": 102652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102652,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102652, "pid": 5, "tid": 7, "ts": 1716454223014336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962613, "dur": 14,
    "args": {
      "External id": 102652,
      "cbid": 211, "correlation": 102652
    }
  },
  {
    "ph": "s", "id": 102652, "pid": 76337, "tid": -914061504, "ts": 1716454222962613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223014827, "dur": 65,
    "args": {
      "External id": 102660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102660, "pid": 5, "tid": 7, "ts": 1716454223014827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962678, "dur": 12,
    "args": {
      "External id": 102660,
      "cbid": 211, "correlation": 102660
    }
  },
  {
    "ph": "s", "id": 102660, "pid": 76337, "tid": -914061504, "ts": 1716454222962678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223014893, "dur": 69,
    "args": {
      "External id": 102668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102668, "pid": 5, "tid": 7, "ts": 1716454223014893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962710, "dur": 8,
    "args": {
      "External id": 102668,
      "cbid": 211, "correlation": 102668
    }
  },
  {
    "ph": "s", "id": 102668, "pid": 76337, "tid": -914061504, "ts": 1716454222962710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962791, "dur": 1,
    "args": {
      "External id": 102684,
      "cbid": 251, "correlation": 102684
    }
  },
  {
    "ph": "f", "id": 102684, "pid": 76337, "tid": -914061504, "ts": 1716454222962791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223014964, "dur": 1,
    "args": {
      "External id": 102686,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 102686,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 102686, "pid": 5, "tid": 7, "ts": 1716454223014964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962797, "dur": 11,
    "args": {
      "External id": 102686,
      "cbid": 51, "correlation": 102686
    }
  },
  {
    "ph": "s", "id": 102686, "pid": 76337, "tid": -914061504, "ts": 1716454222962797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223014968, "dur": 266,
    "args": {
      "External id": 102687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102687,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102687, "pid": 5, "tid": 7, "ts": 1716454223014968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962809, "dur": 11,
    "args": {
      "External id": 102687,
      "cbid": 211, "correlation": 102687
    }
  },
  {
    "ph": "s", "id": 102687, "pid": 76337, "tid": -914061504, "ts": 1716454222962809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223015236, "dur": 14,
    "args": {
      "External id": 102695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102695,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102695, "pid": 5, "tid": 7, "ts": 1716454223015236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962851, "dur": 10,
    "args": {
      "External id": 102695,
      "cbid": 211, "correlation": 102695
    }
  },
  {
    "ph": "s", "id": 102695, "pid": 76337, "tid": -914061504, "ts": 1716454222962851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223015251, "dur": 37,
    "args": {
      "External id": 102706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102706, "pid": 5, "tid": 7, "ts": 1716454223015251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962918, "dur": 12,
    "args": {
      "External id": 102706,
      "cbid": 211, "correlation": 102706
    }
  },
  {
    "ph": "s", "id": 102706, "pid": 76337, "tid": -914061504, "ts": 1716454222962918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962990, "dur": 0,
    "args": {
      "External id": 102718,
      "cbid": 317, "correlation": 102718
    }
  },
  {
    "ph": "f", "id": 102718, "pid": 76337, "tid": -914061504, "ts": 1716454222962990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962991, "dur": 0,
    "args": {
      "External id": 102719,
      "cbid": 203, "correlation": 102719
    }
  },
  {
    "ph": "f", "id": 102719, "pid": 76337, "tid": -914061504, "ts": 1716454222962991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222962992, "dur": 0,
    "args": {
      "External id": 102720,
      "cbid": 205, "correlation": 102720
    }
  },
  {
    "ph": "f", "id": 102720, "pid": 76337, "tid": -914061504, "ts": 1716454222962992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223015289, "dur": 14,
    "args": {
      "External id": 102724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102724,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102724, "pid": 5, "tid": 7, "ts": 1716454223015289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963008, "dur": 12,
    "args": {
      "External id": 102724,
      "cbid": 211, "correlation": 102724
    }
  },
  {
    "ph": "s", "id": 102724, "pid": 76337, "tid": -914061504, "ts": 1716454222963008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223015304, "dur": 4,
    "args": {
      "External id": 102726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102726,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102726, "pid": 5, "tid": 7, "ts": 1716454223015304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963025, "dur": 6,
    "args": {
      "External id": 102726,
      "cbid": 211, "correlation": 102726
    }
  },
  {
    "ph": "s", "id": 102726, "pid": 76337, "tid": -914061504, "ts": 1716454222963025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963033, "dur": 0,
    "args": {
      "External id": 102727,
      "cbid": 51, "correlation": 102727
    }
  },
  {
    "ph": "s", "id": 102727, "pid": 76337, "tid": -914061504, "ts": 1716454222963033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223015309, "dur": 96,
    "args": {
      "External id": 102728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102728,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 102728, "pid": 5, "tid": 7, "ts": 1716454223015309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963034, "dur": 5,
    "args": {
      "External id": 102728,
      "cbid": 211, "correlation": 102728
    }
  },
  {
    "ph": "s", "id": 102728, "pid": 76337, "tid": -914061504, "ts": 1716454222963034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223015406, "dur": 16,
    "args": {
      "External id": 102733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102733, "pid": 5, "tid": 7, "ts": 1716454223015406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963062, "dur": 8,
    "args": {
      "External id": 102733,
      "cbid": 211, "correlation": 102733
    }
  },
  {
    "ph": "s", "id": 102733, "pid": 76337, "tid": -914061504, "ts": 1716454222963062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223015423, "dur": 13,
    "args": {
      "External id": 102741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102741,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102741, "pid": 5, "tid": 7, "ts": 1716454223015423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963095, "dur": 8,
    "args": {
      "External id": 102741,
      "cbid": 211, "correlation": 102741
    }
  },
  {
    "ph": "s", "id": 102741, "pid": 76337, "tid": -914061504, "ts": 1716454222963095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223015437, "dur": 17,
    "args": {
      "External id": 102761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102761,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 102761, "pid": 5, "tid": 7, "ts": 1716454223015437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963167, "dur": 12,
    "args": {
      "External id": 102761,
      "cbid": 211, "correlation": 102761
    }
  },
  {
    "ph": "s", "id": 102761, "pid": 76337, "tid": -914061504, "ts": 1716454222963167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223015456, "dur": 4,
    "args": {
      "External id": 102773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102773,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 102773, "pid": 5, "tid": 7, "ts": 1716454223015456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963189, "dur": 7,
    "args": {
      "External id": 102773,
      "cbid": 211, "correlation": 102773
    }
  },
  {
    "ph": "s", "id": 102773, "pid": 76337, "tid": -914061504, "ts": 1716454222963189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223015461, "dur": 18,
    "args": {
      "External id": 102776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102776, "pid": 5, "tid": 7, "ts": 1716454223015461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963207, "dur": 6,
    "args": {
      "External id": 102776,
      "cbid": 211, "correlation": 102776
    }
  },
  {
    "ph": "s", "id": 102776, "pid": 76337, "tid": -914061504, "ts": 1716454222963207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223015480, "dur": 12,
    "args": {
      "External id": 102785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102785,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102785, "pid": 5, "tid": 7, "ts": 1716454223015480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963246, "dur": 10,
    "args": {
      "External id": 102785,
      "cbid": 211, "correlation": 102785
    }
  },
  {
    "ph": "s", "id": 102785, "pid": 76337, "tid": -914061504, "ts": 1716454222963246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963298, "dur": 0,
    "args": {
      "External id": 102795,
      "cbid": 317, "correlation": 102795
    }
  },
  {
    "ph": "f", "id": 102795, "pid": 76337, "tid": -914061504, "ts": 1716454222963298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963299, "dur": 0,
    "args": {
      "External id": 102796,
      "cbid": 203, "correlation": 102796
    }
  },
  {
    "ph": "f", "id": 102796, "pid": 76337, "tid": -914061504, "ts": 1716454222963299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963300, "dur": 0,
    "args": {
      "External id": 102797,
      "cbid": 205, "correlation": 102797
    }
  },
  {
    "ph": "f", "id": 102797, "pid": 76337, "tid": -914061504, "ts": 1716454222963300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223015494, "dur": 11,
    "args": {
      "External id": 102801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102801,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102801, "pid": 5, "tid": 7, "ts": 1716454223015494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963313, "dur": 11,
    "args": {
      "External id": 102801,
      "cbid": 211, "correlation": 102801
    }
  },
  {
    "ph": "s", "id": 102801, "pid": 76337, "tid": -914061504, "ts": 1716454222963313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223015506, "dur": 160,
    "args": {
      "External id": 102803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102803,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102803, "pid": 5, "tid": 7, "ts": 1716454223015506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963327, "dur": 5,
    "args": {
      "External id": 102803,
      "cbid": 211, "correlation": 102803
    }
  },
  {
    "ph": "s", "id": 102803, "pid": 76337, "tid": -914061504, "ts": 1716454222963327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223015668, "dur": 1,
    "args": {
      "External id": 102805,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 102805,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 102805, "pid": 5, "tid": 7, "ts": 1716454223015668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963338, "dur": 7,
    "args": {
      "External id": 102805,
      "cbid": 51, "correlation": 102805
    }
  },
  {
    "ph": "s", "id": 102805, "pid": 76337, "tid": -914061504, "ts": 1716454222963338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223015672, "dur": 658,
    "args": {
      "External id": 102806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102806,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102806, "pid": 5, "tid": 7, "ts": 1716454223015672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963346, "dur": 6,
    "args": {
      "External id": 102806,
      "cbid": 211, "correlation": 102806
    }
  },
  {
    "ph": "s", "id": 102806, "pid": 76337, "tid": -914061504, "ts": 1716454222963346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223016331, "dur": 13,
    "args": {
      "External id": 102808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102808,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102808, "pid": 5, "tid": 7, "ts": 1716454223016331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963356, "dur": 5,
    "args": {
      "External id": 102808,
      "cbid": 211, "correlation": 102808
    }
  },
  {
    "ph": "s", "id": 102808, "pid": 76337, "tid": -914061504, "ts": 1716454222963356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223016346, "dur": 15,
    "args": {
      "External id": 102814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102814,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102814, "pid": 5, "tid": 7, "ts": 1716454223016346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963384, "dur": 9,
    "args": {
      "External id": 102814,
      "cbid": 211, "correlation": 102814
    }
  },
  {
    "ph": "s", "id": 102814, "pid": 76337, "tid": -914061504, "ts": 1716454222963384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223016362, "dur": 3,
    "args": {
      "External id": 102822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102822,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 102822, "pid": 5, "tid": 7, "ts": 1716454223016362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963428, "dur": 9,
    "args": {
      "External id": 102822,
      "cbid": 211, "correlation": 102822
    }
  },
  {
    "ph": "s", "id": 102822, "pid": 76337, "tid": -914061504, "ts": 1716454222963428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963495, "dur": 1,
    "args": {
      "External id": 102838,
      "cbid": 251, "correlation": 102838
    }
  },
  {
    "ph": "f", "id": 102838, "pid": 76337, "tid": -914061504, "ts": 1716454222963495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963500, "dur": 0,
    "args": {
      "External id": 102840,
      "cbid": 251, "correlation": 102840
    }
  },
  {
    "ph": "f", "id": 102840, "pid": 76337, "tid": -914061504, "ts": 1716454222963500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223016367, "dur": 14,
    "args": {
      "External id": 102841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102841,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102841, "pid": 5, "tid": 7, "ts": 1716454223016367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963502, "dur": 11,
    "args": {
      "External id": 102841,
      "cbid": 211, "correlation": 102841
    }
  },
  {
    "ph": "s", "id": 102841, "pid": 76337, "tid": -914061504, "ts": 1716454222963502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223016381, "dur": 5,
    "args": {
      "External id": 102843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102843,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102843, "pid": 5, "tid": 7, "ts": 1716454223016381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963515, "dur": 6,
    "args": {
      "External id": 102843,
      "cbid": 211, "correlation": 102843
    }
  },
  {
    "ph": "s", "id": 102843, "pid": 76337, "tid": -914061504, "ts": 1716454222963515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223016388, "dur": 17,
    "args": {
      "External id": 102853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102853, "pid": 5, "tid": 7, "ts": 1716454223016388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963573, "dur": 12,
    "args": {
      "External id": 102853,
      "cbid": 211, "correlation": 102853
    }
  },
  {
    "ph": "s", "id": 102853, "pid": 76337, "tid": -914061504, "ts": 1716454222963573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223016406, "dur": 17,
    "args": {
      "External id": 102873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102873,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 102873, "pid": 5, "tid": 7, "ts": 1716454223016406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963639, "dur": 11,
    "args": {
      "External id": 102873,
      "cbid": 211, "correlation": 102873
    }
  },
  {
    "ph": "s", "id": 102873, "pid": 76337, "tid": -914061504, "ts": 1716454222963639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223016425, "dur": 4,
    "args": {
      "External id": 102885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102885,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 102885, "pid": 5, "tid": 7, "ts": 1716454223016425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963660, "dur": 6,
    "args": {
      "External id": 102885,
      "cbid": 211, "correlation": 102885
    }
  },
  {
    "ph": "s", "id": 102885, "pid": 76337, "tid": -914061504, "ts": 1716454222963660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223016430, "dur": 17,
    "args": {
      "External id": 102888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102888, "pid": 5, "tid": 7, "ts": 1716454223016430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963679, "dur": 6,
    "args": {
      "External id": 102888,
      "cbid": 211, "correlation": 102888
    }
  },
  {
    "ph": "s", "id": 102888, "pid": 76337, "tid": -914061504, "ts": 1716454222963679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223016448, "dur": 10,
    "args": {
      "External id": 102897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102897,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102897, "pid": 5, "tid": 7, "ts": 1716454223016448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963719, "dur": 9,
    "args": {
      "External id": 102897,
      "cbid": 211, "correlation": 102897
    }
  },
  {
    "ph": "s", "id": 102897, "pid": 76337, "tid": -914061504, "ts": 1716454222963719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963780, "dur": 0,
    "args": {
      "External id": 102907,
      "cbid": 317, "correlation": 102907
    }
  },
  {
    "ph": "f", "id": 102907, "pid": 76337, "tid": -914061504, "ts": 1716454222963780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963781, "dur": 0,
    "args": {
      "External id": 102908,
      "cbid": 203, "correlation": 102908
    }
  },
  {
    "ph": "f", "id": 102908, "pid": 76337, "tid": -914061504, "ts": 1716454222963781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963782, "dur": 0,
    "args": {
      "External id": 102909,
      "cbid": 205, "correlation": 102909
    }
  },
  {
    "ph": "f", "id": 102909, "pid": 76337, "tid": -914061504, "ts": 1716454222963782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223016460, "dur": 10,
    "args": {
      "External id": 102913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102913, "pid": 5, "tid": 7, "ts": 1716454223016460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963796, "dur": 12,
    "args": {
      "External id": 102913,
      "cbid": 211, "correlation": 102913
    }
  },
  {
    "ph": "s", "id": 102913, "pid": 76337, "tid": -914061504, "ts": 1716454222963796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223016471, "dur": 159,
    "args": {
      "External id": 102915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102915, "pid": 5, "tid": 7, "ts": 1716454223016471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963811, "dur": 5,
    "args": {
      "External id": 102915,
      "cbid": 211, "correlation": 102915
    }
  },
  {
    "ph": "s", "id": 102915, "pid": 76337, "tid": -914061504, "ts": 1716454222963811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223016633, "dur": 1,
    "args": {
      "External id": 102917,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 102917,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 102917, "pid": 5, "tid": 7, "ts": 1716454223016633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963822, "dur": 8,
    "args": {
      "External id": 102917,
      "cbid": 51, "correlation": 102917
    }
  },
  {
    "ph": "s", "id": 102917, "pid": 76337, "tid": -914061504, "ts": 1716454222963822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223016637, "dur": 640,
    "args": {
      "External id": 102918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102918,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 102918, "pid": 5, "tid": 7, "ts": 1716454223016637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963831, "dur": 6,
    "args": {
      "External id": 102918,
      "cbid": 211, "correlation": 102918
    }
  },
  {
    "ph": "s", "id": 102918, "pid": 76337, "tid": -914061504, "ts": 1716454222963831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223017279, "dur": 13,
    "args": {
      "External id": 102920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102920,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102920, "pid": 5, "tid": 7, "ts": 1716454223017279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963841, "dur": 6,
    "args": {
      "External id": 102920,
      "cbid": 211, "correlation": 102920
    }
  },
  {
    "ph": "s", "id": 102920, "pid": 76337, "tid": -914061504, "ts": 1716454222963841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223017293, "dur": 15,
    "args": {
      "External id": 102926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102926, "pid": 5, "tid": 7, "ts": 1716454223017293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963870, "dur": 9,
    "args": {
      "External id": 102926,
      "cbid": 211, "correlation": 102926
    }
  },
  {
    "ph": "s", "id": 102926, "pid": 76337, "tid": -914061504, "ts": 1716454222963870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223017309, "dur": 12,
    "args": {
      "External id": 102934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102934,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102934, "pid": 5, "tid": 7, "ts": 1716454223017309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963902, "dur": 8,
    "args": {
      "External id": 102934,
      "cbid": 211, "correlation": 102934
    }
  },
  {
    "ph": "s", "id": 102934, "pid": 76337, "tid": -914061504, "ts": 1716454222963902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223017323, "dur": 10,
    "args": {
      "External id": 102942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102942,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102942, "pid": 5, "tid": 7, "ts": 1716454223017323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222963930, "dur": 8,
    "args": {
      "External id": 102942,
      "cbid": 211, "correlation": 102942
    }
  },
  {
    "ph": "s", "id": 102942, "pid": 76337, "tid": -914061504, "ts": 1716454222963930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223017334, "dur": 18,
    "args": {
      "External id": 102962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102962,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 102962, "pid": 5, "tid": 7, "ts": 1716454223017334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964020, "dur": 12,
    "args": {
      "External id": 102962,
      "cbid": 211, "correlation": 102962
    }
  },
  {
    "ph": "s", "id": 102962, "pid": 76337, "tid": -914061504, "ts": 1716454222964020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223017353, "dur": 4,
    "args": {
      "External id": 102974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102974,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 102974, "pid": 5, "tid": 7, "ts": 1716454223017353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964042, "dur": 7,
    "args": {
      "External id": 102974,
      "cbid": 211, "correlation": 102974
    }
  },
  {
    "ph": "s", "id": 102974, "pid": 76337, "tid": -914061504, "ts": 1716454222964042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223017358, "dur": 16,
    "args": {
      "External id": 102977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102977, "pid": 5, "tid": 7, "ts": 1716454223017358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964060, "dur": 6,
    "args": {
      "External id": 102977,
      "cbid": 211, "correlation": 102977
    }
  },
  {
    "ph": "s", "id": 102977, "pid": 76337, "tid": -914061504, "ts": 1716454222964060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964117, "dur": 0,
    "args": {
      "External id": 102988,
      "cbid": 317, "correlation": 102988
    }
  },
  {
    "ph": "f", "id": 102988, "pid": 76337, "tid": -914061504, "ts": 1716454222964117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964118, "dur": 0,
    "args": {
      "External id": 102989,
      "cbid": 203, "correlation": 102989
    }
  },
  {
    "ph": "f", "id": 102989, "pid": 76337, "tid": -914061504, "ts": 1716454222964118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964119, "dur": 0,
    "args": {
      "External id": 102990,
      "cbid": 205, "correlation": 102990
    }
  },
  {
    "ph": "f", "id": 102990, "pid": 76337, "tid": -914061504, "ts": 1716454222964119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223017376, "dur": 11,
    "args": {
      "External id": 102994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102994,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 102994, "pid": 5, "tid": 7, "ts": 1716454223017376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964132, "dur": 12,
    "args": {
      "External id": 102994,
      "cbid": 211, "correlation": 102994
    }
  },
  {
    "ph": "s", "id": 102994, "pid": 76337, "tid": -914061504, "ts": 1716454222964132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223017388, "dur": 4,
    "args": {
      "External id": 102996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 102996, "pid": 5, "tid": 7, "ts": 1716454223017388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964150, "dur": 6,
    "args": {
      "External id": 102996,
      "cbid": 211, "correlation": 102996
    }
  },
  {
    "ph": "s", "id": 102996, "pid": 76337, "tid": -914061504, "ts": 1716454222964150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964158, "dur": 0,
    "args": {
      "External id": 102997,
      "cbid": 51, "correlation": 102997
    }
  },
  {
    "ph": "s", "id": 102997, "pid": 76337, "tid": -914061504, "ts": 1716454222964158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223017393, "dur": 93,
    "args": {
      "External id": 102998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 102998,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 102998, "pid": 5, "tid": 7, "ts": 1716454223017393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964159, "dur": 5,
    "args": {
      "External id": 102998,
      "cbid": 211, "correlation": 102998
    }
  },
  {
    "ph": "s", "id": 102998, "pid": 76337, "tid": -914061504, "ts": 1716454222964159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223017488, "dur": 15,
    "args": {
      "External id": 103003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103003, "pid": 5, "tid": 7, "ts": 1716454223017488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964185, "dur": 9,
    "args": {
      "External id": 103003,
      "cbid": 211, "correlation": 103003
    }
  },
  {
    "ph": "s", "id": 103003, "pid": 76337, "tid": -914061504, "ts": 1716454222964185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223017504, "dur": 81,
    "args": {
      "External id": 103012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103012,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103012, "pid": 5, "tid": 7, "ts": 1716454223017504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964268, "dur": 14,
    "args": {
      "External id": 103012,
      "cbid": 211, "correlation": 103012
    }
  },
  {
    "ph": "s", "id": 103012, "pid": 76337, "tid": -914061504, "ts": 1716454222964268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223017586, "dur": 30,
    "args": {
      "External id": 103034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103034,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103034, "pid": 5, "tid": 7, "ts": 1716454223017586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964324, "dur": 10,
    "args": {
      "External id": 103034,
      "cbid": 211, "correlation": 103034
    }
  },
  {
    "ph": "s", "id": 103034, "pid": 76337, "tid": -914061504, "ts": 1716454222964324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964412, "dur": 1,
    "args": {
      "External id": 103045,
      "cbid": 251, "correlation": 103045
    }
  },
  {
    "ph": "f", "id": 103045, "pid": 76337, "tid": -914061504, "ts": 1716454222964412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223017618, "dur": 141,
    "args": {
      "External id": 103046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103046,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103046, "pid": 5, "tid": 7, "ts": 1716454223017618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964417, "dur": 12,
    "args": {
      "External id": 103046,
      "cbid": 211, "correlation": 103046
    }
  },
  {
    "ph": "s", "id": 103046, "pid": 76337, "tid": -914061504, "ts": 1716454222964417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964486, "dur": 1,
    "args": {
      "External id": 103057,
      "cbid": 251, "correlation": 103057
    }
  },
  {
    "ph": "f", "id": 103057, "pid": 76337, "tid": -914061504, "ts": 1716454222964486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223017760, "dur": 155,
    "args": {
      "External id": 103058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103058,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103058, "pid": 5, "tid": 7, "ts": 1716454223017760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964490, "dur": 12,
    "args": {
      "External id": 103058,
      "cbid": 211, "correlation": 103058
    }
  },
  {
    "ph": "s", "id": 103058, "pid": 76337, "tid": -914061504, "ts": 1716454222964490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964556, "dur": 1,
    "args": {
      "External id": 103069,
      "cbid": 251, "correlation": 103069
    }
  },
  {
    "ph": "f", "id": 103069, "pid": 76337, "tid": -914061504, "ts": 1716454222964556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223017916, "dur": 156,
    "args": {
      "External id": 103070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103070,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103070, "pid": 5, "tid": 7, "ts": 1716454223017916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964560, "dur": 11,
    "args": {
      "External id": 103070,
      "cbid": 211, "correlation": 103070
    }
  },
  {
    "ph": "s", "id": 103070, "pid": 76337, "tid": -914061504, "ts": 1716454222964560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223018074, "dur": 330,
    "args": {
      "External id": 103095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103095,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103095, "pid": 5, "tid": 7, "ts": 1716454223018074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964644, "dur": 13,
    "args": {
      "External id": 103095,
      "cbid": 211, "correlation": 103095
    }
  },
  {
    "ph": "s", "id": 103095, "pid": 76337, "tid": -914061504, "ts": 1716454222964644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964742, "dur": 1,
    "args": {
      "External id": 103113,
      "cbid": 251, "correlation": 103113
    }
  },
  {
    "ph": "f", "id": 103113, "pid": 76337, "tid": -914061504, "ts": 1716454222964742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223018405, "dur": 163,
    "args": {
      "External id": 103115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103115,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103115, "pid": 5, "tid": 7, "ts": 1716454223018405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964748, "dur": 14,
    "args": {
      "External id": 103115,
      "cbid": 211, "correlation": 103115
    }
  },
  {
    "ph": "s", "id": 103115, "pid": 76337, "tid": -914061504, "ts": 1716454222964748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223018570, "dur": 19,
    "args": {
      "External id": 103123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103123,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103123, "pid": 5, "tid": 7, "ts": 1716454223018570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964819, "dur": 12,
    "args": {
      "External id": 103123,
      "cbid": 211, "correlation": 103123
    }
  },
  {
    "ph": "s", "id": 103123, "pid": 76337, "tid": -914061504, "ts": 1716454222964819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223018591, "dur": 27,
    "args": {
      "External id": 103131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103131,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103131, "pid": 5, "tid": 7, "ts": 1716454223018591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964858, "dur": 9,
    "args": {
      "External id": 103131,
      "cbid": 211, "correlation": 103131
    }
  },
  {
    "ph": "s", "id": 103131, "pid": 76337, "tid": -914061504, "ts": 1716454222964858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223018619, "dur": 18,
    "args": {
      "External id": 103142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103142, "pid": 5, "tid": 7, "ts": 1716454223018619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964929, "dur": 12,
    "args": {
      "External id": 103142,
      "cbid": 211, "correlation": 103142
    }
  },
  {
    "ph": "s", "id": 103142, "pid": 76337, "tid": -914061504, "ts": 1716454222964929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223018638, "dur": 16,
    "args": {
      "External id": 103164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103164,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103164, "pid": 5, "tid": 7, "ts": 1716454223018638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222964962, "dur": 7,
    "args": {
      "External id": 103164,
      "cbid": 211, "correlation": 103164
    }
  },
  {
    "ph": "s", "id": 103164, "pid": 76337, "tid": -914061504, "ts": 1716454222964962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965055, "dur": 1,
    "args": {
      "External id": 103175,
      "cbid": 251, "correlation": 103175
    }
  },
  {
    "ph": "f", "id": 103175, "pid": 76337, "tid": -914061504, "ts": 1716454222965055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223018655, "dur": 87,
    "args": {
      "External id": 103176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103176,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 103176, "pid": 5, "tid": 7, "ts": 1716454223018655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965061, "dur": 14,
    "args": {
      "External id": 103176,
      "cbid": 211, "correlation": 103176
    }
  },
  {
    "ph": "s", "id": 103176, "pid": 76337, "tid": -914061504, "ts": 1716454222965061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965130, "dur": 1,
    "args": {
      "External id": 103187,
      "cbid": 251, "correlation": 103187
    }
  },
  {
    "ph": "f", "id": 103187, "pid": 76337, "tid": -914061504, "ts": 1716454222965130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965133, "dur": 0,
    "args": {
      "External id": 103188,
      "cbid": 251, "correlation": 103188
    }
  },
  {
    "ph": "f", "id": 103188, "pid": 76337, "tid": -914061504, "ts": 1716454222965133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223018744, "dur": 12,
    "args": {
      "External id": 103189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103189,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103189, "pid": 5, "tid": 7, "ts": 1716454223018744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965135, "dur": 12,
    "args": {
      "External id": 103189,
      "cbid": 211, "correlation": 103189
    }
  },
  {
    "ph": "s", "id": 103189, "pid": 76337, "tid": -914061504, "ts": 1716454222965135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223018757, "dur": 6,
    "args": {
      "External id": 103191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103191,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103191, "pid": 5, "tid": 7, "ts": 1716454223018757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965148, "dur": 6,
    "args": {
      "External id": 103191,
      "cbid": 211, "correlation": 103191
    }
  },
  {
    "ph": "s", "id": 103191, "pid": 76337, "tid": -914061504, "ts": 1716454222965148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965205, "dur": 1,
    "args": {
      "External id": 103202,
      "cbid": 251, "correlation": 103202
    }
  },
  {
    "ph": "f", "id": 103202, "pid": 76337, "tid": -914061504, "ts": 1716454222965205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965209, "dur": 0,
    "args": {
      "External id": 103203,
      "cbid": 251, "correlation": 103203
    }
  },
  {
    "ph": "f", "id": 103203, "pid": 76337, "tid": -914061504, "ts": 1716454222965209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223018764, "dur": 8,
    "args": {
      "External id": 103204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103204,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103204, "pid": 5, "tid": 7, "ts": 1716454223018764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965211, "dur": 11,
    "args": {
      "External id": 103204,
      "cbid": 211, "correlation": 103204
    }
  },
  {
    "ph": "s", "id": 103204, "pid": 76337, "tid": -914061504, "ts": 1716454222965211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223018774, "dur": 3,
    "args": {
      "External id": 103206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103206,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103206, "pid": 5, "tid": 7, "ts": 1716454223018774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965224, "dur": 5,
    "args": {
      "External id": 103206,
      "cbid": 211, "correlation": 103206
    }
  },
  {
    "ph": "s", "id": 103206, "pid": 76337, "tid": -914061504, "ts": 1716454222965224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223018778, "dur": 54,
    "args": {
      "External id": 103231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103231,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103231, "pid": 5, "tid": 7, "ts": 1716454223018778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965300, "dur": 13,
    "args": {
      "External id": 103231,
      "cbid": 211, "correlation": 103231
    }
  },
  {
    "ph": "s", "id": 103231, "pid": 76337, "tid": -914061504, "ts": 1716454222965300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965398, "dur": 1,
    "args": {
      "External id": 103249,
      "cbid": 251, "correlation": 103249
    }
  },
  {
    "ph": "f", "id": 103249, "pid": 76337, "tid": -914061504, "ts": 1716454222965398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223018833, "dur": 88,
    "args": {
      "External id": 103251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103251,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 103251, "pid": 5, "tid": 7, "ts": 1716454223018833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965404, "dur": 14,
    "args": {
      "External id": 103251,
      "cbid": 211, "correlation": 103251
    }
  },
  {
    "ph": "s", "id": 103251, "pid": 76337, "tid": -914061504, "ts": 1716454222965404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223018923, "dur": 9,
    "args": {
      "External id": 103259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103259,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103259, "pid": 5, "tid": 7, "ts": 1716454223018923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965475, "dur": 12,
    "args": {
      "External id": 103259,
      "cbid": 211, "correlation": 103259
    }
  },
  {
    "ph": "s", "id": 103259, "pid": 76337, "tid": -914061504, "ts": 1716454222965475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223018934, "dur": 20,
    "args": {
      "External id": 103267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103267,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103267, "pid": 5, "tid": 7, "ts": 1716454223018934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965516, "dur": 9,
    "args": {
      "External id": 103267,
      "cbid": 211, "correlation": 103267
    }
  },
  {
    "ph": "s", "id": 103267, "pid": 76337, "tid": -914061504, "ts": 1716454222965516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223018955, "dur": 17,
    "args": {
      "External id": 103289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103289,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103289, "pid": 5, "tid": 7, "ts": 1716454223018955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965567, "dur": 10,
    "args": {
      "External id": 103289,
      "cbid": 211, "correlation": 103289
    }
  },
  {
    "ph": "s", "id": 103289, "pid": 76337, "tid": -914061504, "ts": 1716454222965567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965655, "dur": 1,
    "args": {
      "External id": 103305,
      "cbid": 251, "correlation": 103305
    }
  },
  {
    "ph": "f", "id": 103305, "pid": 76337, "tid": -914061504, "ts": 1716454222965655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965660, "dur": 0,
    "args": {
      "External id": 103307,
      "cbid": 251, "correlation": 103307
    }
  },
  {
    "ph": "f", "id": 103307, "pid": 76337, "tid": -914061504, "ts": 1716454222965660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223018974, "dur": 490,
    "args": {
      "External id": 103308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103308,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103308, "pid": 5, "tid": 7, "ts": 1716454223018974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965662, "dur": 12,
    "args": {
      "External id": 103308,
      "cbid": 211, "correlation": 103308
    }
  },
  {
    "ph": "s", "id": 103308, "pid": 76337, "tid": -914061504, "ts": 1716454222965662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223019465, "dur": 65,
    "args": {
      "External id": 103316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103316, "pid": 5, "tid": 7, "ts": 1716454223019465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965726, "dur": 12,
    "args": {
      "External id": 103316,
      "cbid": 211, "correlation": 103316
    }
  },
  {
    "ph": "s", "id": 103316, "pid": 76337, "tid": -914061504, "ts": 1716454222965726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223019531, "dur": 67,
    "args": {
      "External id": 103324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103324, "pid": 5, "tid": 7, "ts": 1716454223019531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965755, "dur": 9,
    "args": {
      "External id": 103324,
      "cbid": 211, "correlation": 103324
    }
  },
  {
    "ph": "s", "id": 103324, "pid": 76337, "tid": -914061504, "ts": 1716454222965755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965835, "dur": 1,
    "args": {
      "External id": 103340,
      "cbid": 251, "correlation": 103340
    }
  },
  {
    "ph": "f", "id": 103340, "pid": 76337, "tid": -914061504, "ts": 1716454222965835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223019601, "dur": 1,
    "args": {
      "External id": 103342,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103342,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1561483409238777
    }
  },
  {
    "ph": "f", "id": 103342, "pid": 5, "tid": 7, "ts": 1716454223019601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965840, "dur": 10,
    "args": {
      "External id": 103342,
      "cbid": 51, "correlation": 103342
    }
  },
  {
    "ph": "s", "id": 103342, "pid": 76337, "tid": -914061504, "ts": 1716454222965840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223019605, "dur": 268,
    "args": {
      "External id": 103343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103343,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 103343, "pid": 5, "tid": 7, "ts": 1716454223019605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965851, "dur": 12,
    "args": {
      "External id": 103343,
      "cbid": 211, "correlation": 103343
    }
  },
  {
    "ph": "s", "id": 103343, "pid": 76337, "tid": -914061504, "ts": 1716454222965851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223019874, "dur": 14,
    "args": {
      "External id": 103351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103351,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103351, "pid": 5, "tid": 7, "ts": 1716454223019874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965893, "dur": 10,
    "args": {
      "External id": 103351,
      "cbid": 211, "correlation": 103351
    }
  },
  {
    "ph": "s", "id": 103351, "pid": 76337, "tid": -914061504, "ts": 1716454222965893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223019889, "dur": 37,
    "args": {
      "External id": 103362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103362, "pid": 5, "tid": 7, "ts": 1716454223019889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222965961, "dur": 20,
    "args": {
      "External id": 103362,
      "cbid": 211, "correlation": 103362
    }
  },
  {
    "ph": "s", "id": 103362, "pid": 76337, "tid": -914061504, "ts": 1716454222965961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966035, "dur": 0,
    "args": {
      "External id": 103374,
      "cbid": 317, "correlation": 103374
    }
  },
  {
    "ph": "f", "id": 103374, "pid": 76337, "tid": -914061504, "ts": 1716454222966035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966036, "dur": 0,
    "args": {
      "External id": 103375,
      "cbid": 203, "correlation": 103375
    }
  },
  {
    "ph": "f", "id": 103375, "pid": 76337, "tid": -914061504, "ts": 1716454222966036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966036, "dur": 0,
    "args": {
      "External id": 103376,
      "cbid": 205, "correlation": 103376
    }
  },
  {
    "ph": "f", "id": 103376, "pid": 76337, "tid": -914061504, "ts": 1716454222966036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223019928, "dur": 13,
    "args": {
      "External id": 103380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103380,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103380, "pid": 5, "tid": 7, "ts": 1716454223019928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966051, "dur": 13,
    "args": {
      "External id": 103380,
      "cbid": 211, "correlation": 103380
    }
  },
  {
    "ph": "s", "id": 103380, "pid": 76337, "tid": -914061504, "ts": 1716454222966051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223019942, "dur": 4,
    "args": {
      "External id": 103382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 103382, "pid": 5, "tid": 7, "ts": 1716454223019942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966069, "dur": 5,
    "args": {
      "External id": 103382,
      "cbid": 211, "correlation": 103382
    }
  },
  {
    "ph": "s", "id": 103382, "pid": 76337, "tid": -914061504, "ts": 1716454222966069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966078, "dur": 0,
    "args": {
      "External id": 103383,
      "cbid": 51, "correlation": 103383
    }
  },
  {
    "ph": "s", "id": 103383, "pid": 76337, "tid": -914061504, "ts": 1716454222966078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223019947, "dur": 96,
    "args": {
      "External id": 103384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103384,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 103384, "pid": 5, "tid": 7, "ts": 1716454223019947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966078, "dur": 5,
    "args": {
      "External id": 103384,
      "cbid": 211, "correlation": 103384
    }
  },
  {
    "ph": "s", "id": 103384, "pid": 76337, "tid": -914061504, "ts": 1716454222966078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223020045, "dur": 16,
    "args": {
      "External id": 103389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103389, "pid": 5, "tid": 7, "ts": 1716454223020045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966106, "dur": 9,
    "args": {
      "External id": 103389,
      "cbid": 211, "correlation": 103389
    }
  },
  {
    "ph": "s", "id": 103389, "pid": 76337, "tid": -914061504, "ts": 1716454222966106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223020062, "dur": 11,
    "args": {
      "External id": 103397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103397,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103397, "pid": 5, "tid": 7, "ts": 1716454223020062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966138, "dur": 8,
    "args": {
      "External id": 103397,
      "cbid": 211, "correlation": 103397
    }
  },
  {
    "ph": "s", "id": 103397, "pid": 76337, "tid": -914061504, "ts": 1716454222966138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966208, "dur": 0,
    "args": {
      "External id": 103407,
      "cbid": 317, "correlation": 103407
    }
  },
  {
    "ph": "f", "id": 103407, "pid": 76337, "tid": -914061504, "ts": 1716454222966208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966209, "dur": 0,
    "args": {
      "External id": 103408,
      "cbid": 203, "correlation": 103408
    }
  },
  {
    "ph": "f", "id": 103408, "pid": 76337, "tid": -914061504, "ts": 1716454222966209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966210, "dur": 0,
    "args": {
      "External id": 103409,
      "cbid": 205, "correlation": 103409
    }
  },
  {
    "ph": "f", "id": 103409, "pid": 76337, "tid": -914061504, "ts": 1716454222966210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020075, "dur": 11,
    "args": {
      "External id": 103413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103413,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103413, "pid": 5, "tid": 7, "ts": 1716454223020075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966224, "dur": 12,
    "args": {
      "External id": 103413,
      "cbid": 211, "correlation": 103413
    }
  },
  {
    "ph": "s", "id": 103413, "pid": 76337, "tid": -914061504, "ts": 1716454222966224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020086, "dur": 159,
    "args": {
      "External id": 103415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103415,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103415, "pid": 5, "tid": 7, "ts": 1716454223020086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966239, "dur": 5,
    "args": {
      "External id": 103415,
      "cbid": 211, "correlation": 103415
    }
  },
  {
    "ph": "s", "id": 103415, "pid": 76337, "tid": -914061504, "ts": 1716454222966239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223020248, "dur": 1,
    "args": {
      "External id": 103417,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103417,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 103417, "pid": 5, "tid": 7, "ts": 1716454223020248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966250, "dur": 7,
    "args": {
      "External id": 103417,
      "cbid": 51, "correlation": 103417
    }
  },
  {
    "ph": "s", "id": 103417, "pid": 76337, "tid": -914061504, "ts": 1716454222966250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223020252, "dur": 196,
    "args": {
      "External id": 103418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103418,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 103418, "pid": 5, "tid": 7, "ts": 1716454223020252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966259, "dur": 8,
    "args": {
      "External id": 103418,
      "cbid": 211, "correlation": 103418
    }
  },
  {
    "ph": "s", "id": 103418, "pid": 76337, "tid": -914061504, "ts": 1716454222966259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020449, "dur": 6,
    "args": {
      "External id": 103420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103420, "pid": 5, "tid": 7, "ts": 1716454223020449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966270, "dur": 5,
    "args": {
      "External id": 103420,
      "cbid": 211, "correlation": 103420
    }
  },
  {
    "ph": "s", "id": 103420, "pid": 76337, "tid": -914061504, "ts": 1716454222966270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223020456, "dur": 6,
    "args": {
      "External id": 103426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103426, "pid": 5, "tid": 7, "ts": 1716454223020456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966298, "dur": 9,
    "args": {
      "External id": 103426,
      "cbid": 211, "correlation": 103426
    }
  },
  {
    "ph": "s", "id": 103426, "pid": 76337, "tid": -914061504, "ts": 1716454222966298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020464, "dur": 10,
    "args": {
      "External id": 103446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103446,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 103446, "pid": 5, "tid": 7, "ts": 1716454223020464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966391, "dur": 12,
    "args": {
      "External id": 103446,
      "cbid": 211, "correlation": 103446
    }
  },
  {
    "ph": "s", "id": 103446, "pid": 76337, "tid": -914061504, "ts": 1716454222966391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223020476, "dur": 4,
    "args": {
      "External id": 103458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103458,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 103458, "pid": 5, "tid": 7, "ts": 1716454223020476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966415, "dur": 6,
    "args": {
      "External id": 103458,
      "cbid": 211, "correlation": 103458
    }
  },
  {
    "ph": "s", "id": 103458, "pid": 76337, "tid": -914061504, "ts": 1716454222966415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223020481, "dur": 8,
    "args": {
      "External id": 103461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103461,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103461, "pid": 5, "tid": 7, "ts": 1716454223020481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966434, "dur": 6,
    "args": {
      "External id": 103461,
      "cbid": 211, "correlation": 103461
    }
  },
  {
    "ph": "s", "id": 103461, "pid": 76337, "tid": -914061504, "ts": 1716454222966434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223020491, "dur": 5,
    "args": {
      "External id": 103470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103470,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103470, "pid": 5, "tid": 7, "ts": 1716454223020491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966475, "dur": 9,
    "args": {
      "External id": 103470,
      "cbid": 211, "correlation": 103470
    }
  },
  {
    "ph": "s", "id": 103470, "pid": 76337, "tid": -914061504, "ts": 1716454222966475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966526, "dur": 0,
    "args": {
      "External id": 103480,
      "cbid": 317, "correlation": 103480
    }
  },
  {
    "ph": "f", "id": 103480, "pid": 76337, "tid": -914061504, "ts": 1716454222966526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966527, "dur": 0,
    "args": {
      "External id": 103481,
      "cbid": 203, "correlation": 103481
    }
  },
  {
    "ph": "f", "id": 103481, "pid": 76337, "tid": -914061504, "ts": 1716454222966527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966528, "dur": 0,
    "args": {
      "External id": 103482,
      "cbid": 205, "correlation": 103482
    }
  },
  {
    "ph": "f", "id": 103482, "pid": 76337, "tid": -914061504, "ts": 1716454222966528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020497, "dur": 5,
    "args": {
      "External id": 103486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103486,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103486, "pid": 5, "tid": 7, "ts": 1716454223020497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966543, "dur": 12,
    "args": {
      "External id": 103486,
      "cbid": 211, "correlation": 103486
    }
  },
  {
    "ph": "s", "id": 103486, "pid": 76337, "tid": -914061504, "ts": 1716454222966543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020503, "dur": 159,
    "args": {
      "External id": 103488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103488,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103488, "pid": 5, "tid": 7, "ts": 1716454223020503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966557, "dur": 6,
    "args": {
      "External id": 103488,
      "cbid": 211, "correlation": 103488
    }
  },
  {
    "ph": "s", "id": 103488, "pid": 76337, "tid": -914061504, "ts": 1716454222966557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223020665, "dur": 1,
    "args": {
      "External id": 103490,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103490,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 103490, "pid": 5, "tid": 7, "ts": 1716454223020665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966568, "dur": 6,
    "args": {
      "External id": 103490,
      "cbid": 51, "correlation": 103490
    }
  },
  {
    "ph": "s", "id": 103490, "pid": 76337, "tid": -914061504, "ts": 1716454222966568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223020668, "dur": 264,
    "args": {
      "External id": 103491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103491,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103491, "pid": 5, "tid": 7, "ts": 1716454223020668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966576, "dur": 6,
    "args": {
      "External id": 103491,
      "cbid": 211, "correlation": 103491
    }
  },
  {
    "ph": "s", "id": 103491, "pid": 76337, "tid": -914061504, "ts": 1716454222966576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020934, "dur": 6,
    "args": {
      "External id": 103493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103493,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103493, "pid": 5, "tid": 7, "ts": 1716454223020934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966587, "dur": 5,
    "args": {
      "External id": 103493,
      "cbid": 211, "correlation": 103493
    }
  },
  {
    "ph": "s", "id": 103493, "pid": 76337, "tid": -914061504, "ts": 1716454222966587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223020941, "dur": 6,
    "args": {
      "External id": 103499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103499, "pid": 5, "tid": 7, "ts": 1716454223020941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966615, "dur": 9,
    "args": {
      "External id": 103499,
      "cbid": 211, "correlation": 103499
    }
  },
  {
    "ph": "s", "id": 103499, "pid": 76337, "tid": -914061504, "ts": 1716454222966615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223020948, "dur": 3,
    "args": {
      "External id": 103507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103507,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 103507, "pid": 5, "tid": 7, "ts": 1716454223020948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966660, "dur": 9,
    "args": {
      "External id": 103507,
      "cbid": 211, "correlation": 103507
    }
  },
  {
    "ph": "s", "id": 103507, "pid": 76337, "tid": -914061504, "ts": 1716454222966660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966727, "dur": 1,
    "args": {
      "External id": 103523,
      "cbid": 251, "correlation": 103523
    }
  },
  {
    "ph": "f", "id": 103523, "pid": 76337, "tid": -914061504, "ts": 1716454222966727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966732, "dur": 0,
    "args": {
      "External id": 103525,
      "cbid": 251, "correlation": 103525
    }
  },
  {
    "ph": "f", "id": 103525, "pid": 76337, "tid": -914061504, "ts": 1716454222966732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223020953, "dur": 13,
    "args": {
      "External id": 103526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103526,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103526, "pid": 5, "tid": 7, "ts": 1716454223020953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966734, "dur": 11,
    "args": {
      "External id": 103526,
      "cbid": 211, "correlation": 103526
    }
  },
  {
    "ph": "s", "id": 103526, "pid": 76337, "tid": -914061504, "ts": 1716454222966734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223020967, "dur": 5,
    "args": {
      "External id": 103528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103528,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103528, "pid": 5, "tid": 7, "ts": 1716454223020967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966747, "dur": 6,
    "args": {
      "External id": 103528,
      "cbid": 211, "correlation": 103528
    }
  },
  {
    "ph": "s", "id": 103528, "pid": 76337, "tid": -914061504, "ts": 1716454222966747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223020973, "dur": 5,
    "args": {
      "External id": 103538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103538,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103538, "pid": 5, "tid": 7, "ts": 1716454223020973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966804, "dur": 13,
    "args": {
      "External id": 103538,
      "cbid": 211, "correlation": 103538
    }
  },
  {
    "ph": "s", "id": 103538, "pid": 76337, "tid": -914061504, "ts": 1716454222966804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223020980, "dur": 10,
    "args": {
      "External id": 103558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103558,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 103558, "pid": 5, "tid": 7, "ts": 1716454223020980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966873, "dur": 11,
    "args": {
      "External id": 103558,
      "cbid": 211, "correlation": 103558
    }
  },
  {
    "ph": "s", "id": 103558, "pid": 76337, "tid": -914061504, "ts": 1716454222966873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223020991, "dur": 4,
    "args": {
      "External id": 103570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103570,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 103570, "pid": 5, "tid": 7, "ts": 1716454223020991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966893, "dur": 6,
    "args": {
      "External id": 103570,
      "cbid": 211, "correlation": 103570
    }
  },
  {
    "ph": "s", "id": 103570, "pid": 76337, "tid": -914061504, "ts": 1716454222966893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223020996, "dur": 7,
    "args": {
      "External id": 103573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103573, "pid": 5, "tid": 7, "ts": 1716454223020996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966913, "dur": 6,
    "args": {
      "External id": 103573,
      "cbid": 211, "correlation": 103573
    }
  },
  {
    "ph": "s", "id": 103573, "pid": 76337, "tid": -914061504, "ts": 1716454222966913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223021004, "dur": 4,
    "args": {
      "External id": 103582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103582,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103582, "pid": 5, "tid": 7, "ts": 1716454223021004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222966953, "dur": 10,
    "args": {
      "External id": 103582,
      "cbid": 211, "correlation": 103582
    }
  },
  {
    "ph": "s", "id": 103582, "pid": 76337, "tid": -914061504, "ts": 1716454222966953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967028, "dur": 0,
    "args": {
      "External id": 103592,
      "cbid": 317, "correlation": 103592
    }
  },
  {
    "ph": "f", "id": 103592, "pid": 76337, "tid": -914061504, "ts": 1716454222967028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967029, "dur": 0,
    "args": {
      "External id": 103593,
      "cbid": 203, "correlation": 103593
    }
  },
  {
    "ph": "f", "id": 103593, "pid": 76337, "tid": -914061504, "ts": 1716454222967029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967029, "dur": 0,
    "args": {
      "External id": 103594,
      "cbid": 205, "correlation": 103594
    }
  },
  {
    "ph": "f", "id": 103594, "pid": 76337, "tid": -914061504, "ts": 1716454222967029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021010, "dur": 5,
    "args": {
      "External id": 103598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103598,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103598, "pid": 5, "tid": 7, "ts": 1716454223021010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967043, "dur": 14,
    "args": {
      "External id": 103598,
      "cbid": 211, "correlation": 103598
    }
  },
  {
    "ph": "s", "id": 103598, "pid": 76337, "tid": -914061504, "ts": 1716454222967043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021016, "dur": 160,
    "args": {
      "External id": 103600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103600,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103600, "pid": 5, "tid": 7, "ts": 1716454223021016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967060, "dur": 5,
    "args": {
      "External id": 103600,
      "cbid": 211, "correlation": 103600
    }
  },
  {
    "ph": "s", "id": 103600, "pid": 76337, "tid": -914061504, "ts": 1716454222967060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223021178, "dur": 1,
    "args": {
      "External id": 103602,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103602,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 103602, "pid": 5, "tid": 7, "ts": 1716454223021178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967071, "dur": 6,
    "args": {
      "External id": 103602,
      "cbid": 51, "correlation": 103602
    }
  },
  {
    "ph": "s", "id": 103602, "pid": 76337, "tid": -914061504, "ts": 1716454222967071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223021182, "dur": 255,
    "args": {
      "External id": 103603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103603,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103603, "pid": 5, "tid": 7, "ts": 1716454223021182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967078, "dur": 6,
    "args": {
      "External id": 103603,
      "cbid": 211, "correlation": 103603
    }
  },
  {
    "ph": "s", "id": 103603, "pid": 76337, "tid": -914061504, "ts": 1716454222967078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021438, "dur": 6,
    "args": {
      "External id": 103605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103605,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103605, "pid": 5, "tid": 7, "ts": 1716454223021438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967087, "dur": 5,
    "args": {
      "External id": 103605,
      "cbid": 211, "correlation": 103605
    }
  },
  {
    "ph": "s", "id": 103605, "pid": 76337, "tid": -914061504, "ts": 1716454222967087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223021445, "dur": 6,
    "args": {
      "External id": 103611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103611, "pid": 5, "tid": 7, "ts": 1716454223021445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967116, "dur": 8,
    "args": {
      "External id": 103611,
      "cbid": 211, "correlation": 103611
    }
  },
  {
    "ph": "s", "id": 103611, "pid": 76337, "tid": -914061504, "ts": 1716454222967116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223021452, "dur": 5,
    "args": {
      "External id": 103619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103619,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103619, "pid": 5, "tid": 7, "ts": 1716454223021452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967148, "dur": 9,
    "args": {
      "External id": 103619,
      "cbid": 211, "correlation": 103619
    }
  },
  {
    "ph": "s", "id": 103619, "pid": 76337, "tid": -914061504, "ts": 1716454222967148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223021458, "dur": 4,
    "args": {
      "External id": 103627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103627,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103627, "pid": 5, "tid": 7, "ts": 1716454223021458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967178, "dur": 8,
    "args": {
      "External id": 103627,
      "cbid": 211, "correlation": 103627
    }
  },
  {
    "ph": "s", "id": 103627, "pid": 76337, "tid": -914061504, "ts": 1716454222967178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021464, "dur": 9,
    "args": {
      "External id": 103647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103647,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 103647, "pid": 5, "tid": 7, "ts": 1716454223021464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967253, "dur": 13,
    "args": {
      "External id": 103647,
      "cbid": 211, "correlation": 103647
    }
  },
  {
    "ph": "s", "id": 103647, "pid": 76337, "tid": -914061504, "ts": 1716454222967253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223021474, "dur": 4,
    "args": {
      "External id": 103659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103659,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 103659, "pid": 5, "tid": 7, "ts": 1716454223021474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967275, "dur": 6,
    "args": {
      "External id": 103659,
      "cbid": 211, "correlation": 103659
    }
  },
  {
    "ph": "s", "id": 103659, "pid": 76337, "tid": -914061504, "ts": 1716454222967275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223021479, "dur": 6,
    "args": {
      "External id": 103662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103662, "pid": 5, "tid": 7, "ts": 1716454223021479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967293, "dur": 6,
    "args": {
      "External id": 103662,
      "cbid": 211, "correlation": 103662
    }
  },
  {
    "ph": "s", "id": 103662, "pid": 76337, "tid": -914061504, "ts": 1716454222967293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223021487, "dur": 4,
    "args": {
      "External id": 103671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103671,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103671, "pid": 5, "tid": 7, "ts": 1716454223021487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967331, "dur": 10,
    "args": {
      "External id": 103671,
      "cbid": 211, "correlation": 103671
    }
  },
  {
    "ph": "s", "id": 103671, "pid": 76337, "tid": -914061504, "ts": 1716454222967331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967383, "dur": 0,
    "args": {
      "External id": 103681,
      "cbid": 317, "correlation": 103681
    }
  },
  {
    "ph": "f", "id": 103681, "pid": 76337, "tid": -914061504, "ts": 1716454222967383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967384, "dur": 0,
    "args": {
      "External id": 103682,
      "cbid": 203, "correlation": 103682
    }
  },
  {
    "ph": "f", "id": 103682, "pid": 76337, "tid": -914061504, "ts": 1716454222967384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967384, "dur": 0,
    "args": {
      "External id": 103683,
      "cbid": 205, "correlation": 103683
    }
  },
  {
    "ph": "f", "id": 103683, "pid": 76337, "tid": -914061504, "ts": 1716454222967384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021493, "dur": 5,
    "args": {
      "External id": 103687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103687,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103687, "pid": 5, "tid": 7, "ts": 1716454223021493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967397, "dur": 13,
    "args": {
      "External id": 103687,
      "cbid": 211, "correlation": 103687
    }
  },
  {
    "ph": "s", "id": 103687, "pid": 76337, "tid": -914061504, "ts": 1716454222967397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021499, "dur": 160,
    "args": {
      "External id": 103689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103689,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103689, "pid": 5, "tid": 7, "ts": 1716454223021499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967413, "dur": 5,
    "args": {
      "External id": 103689,
      "cbid": 211, "correlation": 103689
    }
  },
  {
    "ph": "s", "id": 103689, "pid": 76337, "tid": -914061504, "ts": 1716454222967413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223021661, "dur": 1,
    "args": {
      "External id": 103691,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103691,
      "bytes": 240, "memory bandwidth (GB/s)": 0.13392857142857142
    }
  },
  {
    "ph": "f", "id": 103691, "pid": 5, "tid": 7, "ts": 1716454223021661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967423, "dur": 6,
    "args": {
      "External id": 103691,
      "cbid": 51, "correlation": 103691
    }
  },
  {
    "ph": "s", "id": 103691, "pid": 76337, "tid": -914061504, "ts": 1716454222967423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223021665, "dur": 254,
    "args": {
      "External id": 103692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103692,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103692, "pid": 5, "tid": 7, "ts": 1716454223021665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967430, "dur": 6,
    "args": {
      "External id": 103692,
      "cbid": 211, "correlation": 103692
    }
  },
  {
    "ph": "s", "id": 103692, "pid": 76337, "tid": -914061504, "ts": 1716454222967430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021920, "dur": 6,
    "args": {
      "External id": 103694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103694,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103694, "pid": 5, "tid": 7, "ts": 1716454223021920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967439, "dur": 5,
    "args": {
      "External id": 103694,
      "cbid": 211, "correlation": 103694
    }
  },
  {
    "ph": "s", "id": 103694, "pid": 76337, "tid": -914061504, "ts": 1716454222967439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223021927, "dur": 6,
    "args": {
      "External id": 103700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103700,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103700, "pid": 5, "tid": 7, "ts": 1716454223021927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967468, "dur": 8,
    "args": {
      "External id": 103700,
      "cbid": 211, "correlation": 103700
    }
  },
  {
    "ph": "s", "id": 103700, "pid": 76337, "tid": -914061504, "ts": 1716454222967468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223021934, "dur": 3,
    "args": {
      "External id": 103708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103708,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 103708, "pid": 5, "tid": 7, "ts": 1716454223021934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967512, "dur": 9,
    "args": {
      "External id": 103708,
      "cbid": 211, "correlation": 103708
    }
  },
  {
    "ph": "s", "id": 103708, "pid": 76337, "tid": -914061504, "ts": 1716454222967512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967575, "dur": 1,
    "args": {
      "External id": 103724,
      "cbid": 251, "correlation": 103724
    }
  },
  {
    "ph": "f", "id": 103724, "pid": 76337, "tid": -914061504, "ts": 1716454222967575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967580, "dur": 0,
    "args": {
      "External id": 103726,
      "cbid": 251, "correlation": 103726
    }
  },
  {
    "ph": "f", "id": 103726, "pid": 76337, "tid": -914061504, "ts": 1716454222967580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223021939, "dur": 10,
    "args": {
      "External id": 103727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103727,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103727, "pid": 5, "tid": 7, "ts": 1716454223021939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967581, "dur": 11,
    "args": {
      "External id": 103727,
      "cbid": 211, "correlation": 103727
    }
  },
  {
    "ph": "s", "id": 103727, "pid": 76337, "tid": -914061504, "ts": 1716454222967581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223021950, "dur": 4,
    "args": {
      "External id": 103729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103729,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103729, "pid": 5, "tid": 7, "ts": 1716454223021950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967594, "dur": 6,
    "args": {
      "External id": 103729,
      "cbid": 211, "correlation": 103729
    }
  },
  {
    "ph": "s", "id": 103729, "pid": 76337, "tid": -914061504, "ts": 1716454222967594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223021955, "dur": 6,
    "args": {
      "External id": 103739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103739, "pid": 5, "tid": 7, "ts": 1716454223021955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967650, "dur": 13,
    "args": {
      "External id": 103739,
      "cbid": 211, "correlation": 103739
    }
  },
  {
    "ph": "s", "id": 103739, "pid": 76337, "tid": -914061504, "ts": 1716454222967650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021962, "dur": 9,
    "args": {
      "External id": 103759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103759,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 103759, "pid": 5, "tid": 7, "ts": 1716454223021962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967718, "dur": 10,
    "args": {
      "External id": 103759,
      "cbid": 211, "correlation": 103759
    }
  },
  {
    "ph": "s", "id": 103759, "pid": 76337, "tid": -914061504, "ts": 1716454222967718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223021973, "dur": 3,
    "args": {
      "External id": 103771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103771,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 103771, "pid": 5, "tid": 7, "ts": 1716454223021973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967738, "dur": 6,
    "args": {
      "External id": 103771,
      "cbid": 211, "correlation": 103771
    }
  },
  {
    "ph": "s", "id": 103771, "pid": 76337, "tid": -914061504, "ts": 1716454222967738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223021977, "dur": 7,
    "args": {
      "External id": 103774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103774, "pid": 5, "tid": 7, "ts": 1716454223021977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967757, "dur": 7,
    "args": {
      "External id": 103774,
      "cbid": 211, "correlation": 103774
    }
  },
  {
    "ph": "s", "id": 103774, "pid": 76337, "tid": -914061504, "ts": 1716454222967757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223021985, "dur": 4,
    "args": {
      "External id": 103783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103783,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103783, "pid": 5, "tid": 7, "ts": 1716454223021985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967798, "dur": 10,
    "args": {
      "External id": 103783,
      "cbid": 211, "correlation": 103783
    }
  },
  {
    "ph": "s", "id": 103783, "pid": 76337, "tid": -914061504, "ts": 1716454222967798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967861, "dur": 0,
    "args": {
      "External id": 103793,
      "cbid": 317, "correlation": 103793
    }
  },
  {
    "ph": "f", "id": 103793, "pid": 76337, "tid": -914061504, "ts": 1716454222967861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967862, "dur": 0,
    "args": {
      "External id": 103794,
      "cbid": 203, "correlation": 103794
    }
  },
  {
    "ph": "f", "id": 103794, "pid": 76337, "tid": -914061504, "ts": 1716454222967862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967862, "dur": 0,
    "args": {
      "External id": 103795,
      "cbid": 205, "correlation": 103795
    }
  },
  {
    "ph": "f", "id": 103795, "pid": 76337, "tid": -914061504, "ts": 1716454222967862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021991, "dur": 5,
    "args": {
      "External id": 103799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103799,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103799, "pid": 5, "tid": 7, "ts": 1716454223021991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967876, "dur": 12,
    "args": {
      "External id": 103799,
      "cbid": 211, "correlation": 103799
    }
  },
  {
    "ph": "s", "id": 103799, "pid": 76337, "tid": -914061504, "ts": 1716454222967876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223021998, "dur": 158,
    "args": {
      "External id": 103801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103801,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103801, "pid": 5, "tid": 7, "ts": 1716454223021998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967890, "dur": 5,
    "args": {
      "External id": 103801,
      "cbid": 211, "correlation": 103801
    }
  },
  {
    "ph": "s", "id": 103801, "pid": 76337, "tid": -914061504, "ts": 1716454222967890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223022158, "dur": 1,
    "args": {
      "External id": 103803,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103803,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 103803, "pid": 5, "tid": 7, "ts": 1716454223022158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967901, "dur": 6,
    "args": {
      "External id": 103803,
      "cbid": 51, "correlation": 103803
    }
  },
  {
    "ph": "s", "id": 103803, "pid": 76337, "tid": -914061504, "ts": 1716454222967901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223022162, "dur": 254,
    "args": {
      "External id": 103804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103804,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103804, "pid": 5, "tid": 7, "ts": 1716454223022162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967909, "dur": 7,
    "args": {
      "External id": 103804,
      "cbid": 211, "correlation": 103804
    }
  },
  {
    "ph": "s", "id": 103804, "pid": 76337, "tid": -914061504, "ts": 1716454222967909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022417, "dur": 6,
    "args": {
      "External id": 103806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103806,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103806, "pid": 5, "tid": 7, "ts": 1716454223022417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967919, "dur": 5,
    "args": {
      "External id": 103806,
      "cbid": 211, "correlation": 103806
    }
  },
  {
    "ph": "s", "id": 103806, "pid": 76337, "tid": -914061504, "ts": 1716454222967919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223022424, "dur": 6,
    "args": {
      "External id": 103812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103812, "pid": 5, "tid": 7, "ts": 1716454223022424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967947, "dur": 8,
    "args": {
      "External id": 103812,
      "cbid": 211, "correlation": 103812
    }
  },
  {
    "ph": "s", "id": 103812, "pid": 76337, "tid": -914061504, "ts": 1716454222967947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223022432, "dur": 5,
    "args": {
      "External id": 103820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103820,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103820, "pid": 5, "tid": 7, "ts": 1716454223022432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222967988, "dur": 9,
    "args": {
      "External id": 103820,
      "cbid": 211, "correlation": 103820
    }
  },
  {
    "ph": "s", "id": 103820, "pid": 76337, "tid": -914061504, "ts": 1716454222967988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223022438, "dur": 4,
    "args": {
      "External id": 103828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103828,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103828, "pid": 5, "tid": 7, "ts": 1716454223022438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968019, "dur": 8,
    "args": {
      "External id": 103828,
      "cbid": 211, "correlation": 103828
    }
  },
  {
    "ph": "s", "id": 103828, "pid": 76337, "tid": -914061504, "ts": 1716454222968019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022444, "dur": 9,
    "args": {
      "External id": 103848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103848,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 103848, "pid": 5, "tid": 7, "ts": 1716454223022444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968124, "dur": 12,
    "args": {
      "External id": 103848,
      "cbid": 211, "correlation": 103848
    }
  },
  {
    "ph": "s", "id": 103848, "pid": 76337, "tid": -914061504, "ts": 1716454222968124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223022454, "dur": 4,
    "args": {
      "External id": 103860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103860,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 103860, "pid": 5, "tid": 7, "ts": 1716454223022454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968146, "dur": 6,
    "args": {
      "External id": 103860,
      "cbid": 211, "correlation": 103860
    }
  },
  {
    "ph": "s", "id": 103860, "pid": 76337, "tid": -914061504, "ts": 1716454222968146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223022459, "dur": 6,
    "args": {
      "External id": 103863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103863, "pid": 5, "tid": 7, "ts": 1716454223022459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968164, "dur": 7,
    "args": {
      "External id": 103863,
      "cbid": 211, "correlation": 103863
    }
  },
  {
    "ph": "s", "id": 103863, "pid": 76337, "tid": -914061504, "ts": 1716454222968164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223022467, "dur": 4,
    "args": {
      "External id": 103872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103872,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103872, "pid": 5, "tid": 7, "ts": 1716454223022467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968203, "dur": 10,
    "args": {
      "External id": 103872,
      "cbid": 211, "correlation": 103872
    }
  },
  {
    "ph": "s", "id": 103872, "pid": 76337, "tid": -914061504, "ts": 1716454222968203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968255, "dur": 0,
    "args": {
      "External id": 103882,
      "cbid": 317, "correlation": 103882
    }
  },
  {
    "ph": "f", "id": 103882, "pid": 76337, "tid": -914061504, "ts": 1716454222968255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968256, "dur": 0,
    "args": {
      "External id": 103883,
      "cbid": 203, "correlation": 103883
    }
  },
  {
    "ph": "f", "id": 103883, "pid": 76337, "tid": -914061504, "ts": 1716454222968256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968257, "dur": 0,
    "args": {
      "External id": 103884,
      "cbid": 205, "correlation": 103884
    }
  },
  {
    "ph": "f", "id": 103884, "pid": 76337, "tid": -914061504, "ts": 1716454222968257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022473, "dur": 5,
    "args": {
      "External id": 103888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103888,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103888, "pid": 5, "tid": 7, "ts": 1716454223022473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968271, "dur": 12,
    "args": {
      "External id": 103888,
      "cbid": 211, "correlation": 103888
    }
  },
  {
    "ph": "s", "id": 103888, "pid": 76337, "tid": -914061504, "ts": 1716454222968271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022479, "dur": 160,
    "args": {
      "External id": 103890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103890,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103890, "pid": 5, "tid": 7, "ts": 1716454223022479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968285, "dur": 5,
    "args": {
      "External id": 103890,
      "cbid": 211, "correlation": 103890
    }
  },
  {
    "ph": "s", "id": 103890, "pid": 76337, "tid": -914061504, "ts": 1716454222968285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223022641, "dur": 1,
    "args": {
      "External id": 103892,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 103892,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 103892, "pid": 5, "tid": 7, "ts": 1716454223022641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968295, "dur": 7,
    "args": {
      "External id": 103892,
      "cbid": 51, "correlation": 103892
    }
  },
  {
    "ph": "s", "id": 103892, "pid": 76337, "tid": -914061504, "ts": 1716454222968295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223022645, "dur": 253,
    "args": {
      "External id": 103893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103893,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103893, "pid": 5, "tid": 7, "ts": 1716454223022645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968303, "dur": 7,
    "args": {
      "External id": 103893,
      "cbid": 211, "correlation": 103893
    }
  },
  {
    "ph": "s", "id": 103893, "pid": 76337, "tid": -914061504, "ts": 1716454222968303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022900, "dur": 6,
    "args": {
      "External id": 103895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103895,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 103895, "pid": 5, "tid": 7, "ts": 1716454223022900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968313, "dur": 5,
    "args": {
      "External id": 103895,
      "cbid": 211, "correlation": 103895
    }
  },
  {
    "ph": "s", "id": 103895, "pid": 76337, "tid": -914061504, "ts": 1716454222968313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223022907, "dur": 6,
    "args": {
      "External id": 103901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103901, "pid": 5, "tid": 7, "ts": 1716454223022907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968341, "dur": 8,
    "args": {
      "External id": 103901,
      "cbid": 211, "correlation": 103901
    }
  },
  {
    "ph": "s", "id": 103901, "pid": 76337, "tid": -914061504, "ts": 1716454222968341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223022914, "dur": 3,
    "args": {
      "External id": 103909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103909,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 103909, "pid": 5, "tid": 7, "ts": 1716454223022914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968385, "dur": 9,
    "args": {
      "External id": 103909,
      "cbid": 211, "correlation": 103909
    }
  },
  {
    "ph": "s", "id": 103909, "pid": 76337, "tid": -914061504, "ts": 1716454222968385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968446, "dur": 1,
    "args": {
      "External id": 103925,
      "cbid": 251, "correlation": 103925
    }
  },
  {
    "ph": "f", "id": 103925, "pid": 76337, "tid": -914061504, "ts": 1716454222968446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968452, "dur": 0,
    "args": {
      "External id": 103927,
      "cbid": 251, "correlation": 103927
    }
  },
  {
    "ph": "f", "id": 103927, "pid": 76337, "tid": -914061504, "ts": 1716454222968452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223022918, "dur": 11,
    "args": {
      "External id": 103928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103928,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103928, "pid": 5, "tid": 7, "ts": 1716454223022918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968454, "dur": 11,
    "args": {
      "External id": 103928,
      "cbid": 211, "correlation": 103928
    }
  },
  {
    "ph": "s", "id": 103928, "pid": 76337, "tid": -914061504, "ts": 1716454222968454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223022931, "dur": 4,
    "args": {
      "External id": 103930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103930,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 103930, "pid": 5, "tid": 7, "ts": 1716454223022931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968467, "dur": 6,
    "args": {
      "External id": 103930,
      "cbid": 211, "correlation": 103930
    }
  },
  {
    "ph": "s", "id": 103930, "pid": 76337, "tid": -914061504, "ts": 1716454222968467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223022936, "dur": 5,
    "args": {
      "External id": 103940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103940,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103940, "pid": 5, "tid": 7, "ts": 1716454223022936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968524, "dur": 12,
    "args": {
      "External id": 103940,
      "cbid": 211, "correlation": 103940
    }
  },
  {
    "ph": "s", "id": 103940, "pid": 76337, "tid": -914061504, "ts": 1716454222968524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022943, "dur": 10,
    "args": {
      "External id": 103960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103960,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 103960, "pid": 5, "tid": 7, "ts": 1716454223022943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968590, "dur": 11,
    "args": {
      "External id": 103960,
      "cbid": 211, "correlation": 103960
    }
  },
  {
    "ph": "s", "id": 103960, "pid": 76337, "tid": -914061504, "ts": 1716454222968590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223022954, "dur": 4,
    "args": {
      "External id": 103972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103972,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 103972, "pid": 5, "tid": 7, "ts": 1716454223022954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968612, "dur": 6,
    "args": {
      "External id": 103972,
      "cbid": 211, "correlation": 103972
    }
  },
  {
    "ph": "s", "id": 103972, "pid": 76337, "tid": -914061504, "ts": 1716454222968612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223022959, "dur": 7,
    "args": {
      "External id": 103975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103975,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103975, "pid": 5, "tid": 7, "ts": 1716454223022959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968631, "dur": 6,
    "args": {
      "External id": 103975,
      "cbid": 211, "correlation": 103975
    }
  },
  {
    "ph": "s", "id": 103975, "pid": 76337, "tid": -914061504, "ts": 1716454222968631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223022967, "dur": 4,
    "args": {
      "External id": 103984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 103984,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 103984, "pid": 5, "tid": 7, "ts": 1716454223022967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968671, "dur": 10,
    "args": {
      "External id": 103984,
      "cbid": 211, "correlation": 103984
    }
  },
  {
    "ph": "s", "id": 103984, "pid": 76337, "tid": -914061504, "ts": 1716454222968671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968734, "dur": 0,
    "args": {
      "External id": 103994,
      "cbid": 317, "correlation": 103994
    }
  },
  {
    "ph": "f", "id": 103994, "pid": 76337, "tid": -914061504, "ts": 1716454222968734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968735, "dur": 0,
    "args": {
      "External id": 103995,
      "cbid": 203, "correlation": 103995
    }
  },
  {
    "ph": "f", "id": 103995, "pid": 76337, "tid": -914061504, "ts": 1716454222968735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968735, "dur": 0,
    "args": {
      "External id": 103996,
      "cbid": 205, "correlation": 103996
    }
  },
  {
    "ph": "f", "id": 103996, "pid": 76337, "tid": -914061504, "ts": 1716454222968735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022972, "dur": 5,
    "args": {
      "External id": 104000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104000,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104000, "pid": 5, "tid": 7, "ts": 1716454223022972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968749, "dur": 13,
    "args": {
      "External id": 104000,
      "cbid": 211, "correlation": 104000
    }
  },
  {
    "ph": "s", "id": 104000, "pid": 76337, "tid": -914061504, "ts": 1716454222968749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223022979, "dur": 159,
    "args": {
      "External id": 104002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104002,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104002, "pid": 5, "tid": 7, "ts": 1716454223022979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968764, "dur": 5,
    "args": {
      "External id": 104002,
      "cbid": 211, "correlation": 104002
    }
  },
  {
    "ph": "s", "id": 104002, "pid": 76337, "tid": -914061504, "ts": 1716454222968764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223023140, "dur": 1,
    "args": {
      "External id": 104004,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104004,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 104004, "pid": 5, "tid": 7, "ts": 1716454223023140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968775, "dur": 6,
    "args": {
      "External id": 104004,
      "cbid": 51, "correlation": 104004
    }
  },
  {
    "ph": "s", "id": 104004, "pid": 76337, "tid": -914061504, "ts": 1716454222968775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223023143, "dur": 254,
    "args": {
      "External id": 104005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104005,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104005, "pid": 5, "tid": 7, "ts": 1716454223023143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968782, "dur": 6,
    "args": {
      "External id": 104005,
      "cbid": 211, "correlation": 104005
    }
  },
  {
    "ph": "s", "id": 104005, "pid": 76337, "tid": -914061504, "ts": 1716454222968782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223023398, "dur": 5,
    "args": {
      "External id": 104007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104007,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104007, "pid": 5, "tid": 7, "ts": 1716454223023398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968792, "dur": 5,
    "args": {
      "External id": 104007,
      "cbid": 211, "correlation": 104007
    }
  },
  {
    "ph": "s", "id": 104007, "pid": 76337, "tid": -914061504, "ts": 1716454222968792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223023405, "dur": 6,
    "args": {
      "External id": 104013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104013, "pid": 5, "tid": 7, "ts": 1716454223023405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968820, "dur": 8,
    "args": {
      "External id": 104013,
      "cbid": 211, "correlation": 104013
    }
  },
  {
    "ph": "s", "id": 104013, "pid": 76337, "tid": -914061504, "ts": 1716454222968820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223023413, "dur": 5,
    "args": {
      "External id": 104021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104021,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104021, "pid": 5, "tid": 7, "ts": 1716454223023413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968852, "dur": 9,
    "args": {
      "External id": 104021,
      "cbid": 211, "correlation": 104021
    }
  },
  {
    "ph": "s", "id": 104021, "pid": 76337, "tid": -914061504, "ts": 1716454222968852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223023419, "dur": 4,
    "args": {
      "External id": 104029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104029,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104029, "pid": 5, "tid": 7, "ts": 1716454223023419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222968882, "dur": 8,
    "args": {
      "External id": 104029,
      "cbid": 211, "correlation": 104029
    }
  },
  {
    "ph": "s", "id": 104029, "pid": 76337, "tid": -914061504, "ts": 1716454222968882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223023424, "dur": 9,
    "args": {
      "External id": 104049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104049,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 104049, "pid": 5, "tid": 7, "ts": 1716454223023424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969033, "dur": 14,
    "args": {
      "External id": 104049,
      "cbid": 211, "correlation": 104049
    }
  },
  {
    "ph": "s", "id": 104049, "pid": 76337, "tid": -914061504, "ts": 1716454222969033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223023435, "dur": 4,
    "args": {
      "External id": 104061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104061,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 104061, "pid": 5, "tid": 7, "ts": 1716454223023435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969057, "dur": 7,
    "args": {
      "External id": 104061,
      "cbid": 211, "correlation": 104061
    }
  },
  {
    "ph": "s", "id": 104061, "pid": 76337, "tid": -914061504, "ts": 1716454222969057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223023440, "dur": 6,
    "args": {
      "External id": 104064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104064, "pid": 5, "tid": 7, "ts": 1716454223023440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969076, "dur": 7,
    "args": {
      "External id": 104064,
      "cbid": 211, "correlation": 104064
    }
  },
  {
    "ph": "s", "id": 104064, "pid": 76337, "tid": -914061504, "ts": 1716454222969076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969135, "dur": 0,
    "args": {
      "External id": 104075,
      "cbid": 317, "correlation": 104075
    }
  },
  {
    "ph": "f", "id": 104075, "pid": 76337, "tid": -914061504, "ts": 1716454222969135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969135, "dur": 0,
    "args": {
      "External id": 104076,
      "cbid": 203, "correlation": 104076
    }
  },
  {
    "ph": "f", "id": 104076, "pid": 76337, "tid": -914061504, "ts": 1716454222969135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969136, "dur": 0,
    "args": {
      "External id": 104077,
      "cbid": 205, "correlation": 104077
    }
  },
  {
    "ph": "f", "id": 104077, "pid": 76337, "tid": -914061504, "ts": 1716454222969136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223023447, "dur": 5,
    "args": {
      "External id": 104081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104081,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104081, "pid": 5, "tid": 7, "ts": 1716454223023447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969153, "dur": 12,
    "args": {
      "External id": 104081,
      "cbid": 211, "correlation": 104081
    }
  },
  {
    "ph": "s", "id": 104081, "pid": 76337, "tid": -914061504, "ts": 1716454222969153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223023453, "dur": 36,
    "args": {
      "External id": 104083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104083,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 104083, "pid": 5, "tid": 7, "ts": 1716454223023453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969172, "dur": 9,
    "args": {
      "External id": 104083,
      "cbid": 211, "correlation": 104083
    }
  },
  {
    "ph": "s", "id": 104083, "pid": 76337, "tid": -914061504, "ts": 1716454222969172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223023491, "dur": 5,
    "args": {
      "External id": 104085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104085,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104085, "pid": 5, "tid": 7, "ts": 1716454223023491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969185, "dur": 5,
    "args": {
      "External id": 104085,
      "cbid": 211, "correlation": 104085
    }
  },
  {
    "ph": "s", "id": 104085, "pid": 76337, "tid": -914061504, "ts": 1716454222969185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223023497, "dur": 6,
    "args": {
      "External id": 104091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104091, "pid": 5, "tid": 7, "ts": 1716454223023497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969213, "dur": 8,
    "args": {
      "External id": 104091,
      "cbid": 211, "correlation": 104091
    }
  },
  {
    "ph": "s", "id": 104091, "pid": 76337, "tid": -914061504, "ts": 1716454222969213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223023504, "dur": 20,
    "args": {
      "External id": 104100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104100,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104100, "pid": 5, "tid": 7, "ts": 1716454223023504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969294, "dur": 14,
    "args": {
      "External id": 104100,
      "cbid": 211, "correlation": 104100
    }
  },
  {
    "ph": "s", "id": 104100, "pid": 76337, "tid": -914061504, "ts": 1716454222969294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223023525, "dur": 10,
    "args": {
      "External id": 104122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104122,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 104122, "pid": 5, "tid": 7, "ts": 1716454223023525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969351, "dur": 11,
    "args": {
      "External id": 104122,
      "cbid": 211, "correlation": 104122
    }
  },
  {
    "ph": "s", "id": 104122, "pid": 76337, "tid": -914061504, "ts": 1716454222969351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969441, "dur": 2,
    "args": {
      "External id": 104133,
      "cbid": 251, "correlation": 104133
    }
  },
  {
    "ph": "f", "id": 104133, "pid": 76337, "tid": -914061504, "ts": 1716454222969441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969445, "dur": 0,
    "args": {
      "External id": 104134,
      "cbid": 251, "correlation": 104134
    }
  },
  {
    "ph": "f", "id": 104134, "pid": 76337, "tid": -914061504, "ts": 1716454222969445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223023537, "dur": 53,
    "args": {
      "External id": 104135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104135,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 104135, "pid": 5, "tid": 7, "ts": 1716454223023537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969448, "dur": 14,
    "args": {
      "External id": 104135,
      "cbid": 211, "correlation": 104135
    }
  },
  {
    "ph": "s", "id": 104135, "pid": 76337, "tid": -914061504, "ts": 1716454222969448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969518, "dur": 1,
    "args": {
      "External id": 104146,
      "cbid": 251, "correlation": 104146
    }
  },
  {
    "ph": "f", "id": 104146, "pid": 76337, "tid": -914061504, "ts": 1716454222969518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969522, "dur": 0,
    "args": {
      "External id": 104147,
      "cbid": 251, "correlation": 104147
    }
  },
  {
    "ph": "f", "id": 104147, "pid": 76337, "tid": -914061504, "ts": 1716454222969522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223023591, "dur": 53,
    "args": {
      "External id": 104148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104148,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 104148, "pid": 5, "tid": 7, "ts": 1716454223023591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969524, "dur": 11,
    "args": {
      "External id": 104148,
      "cbid": 211, "correlation": 104148
    }
  },
  {
    "ph": "s", "id": 104148, "pid": 76337, "tid": -914061504, "ts": 1716454222969524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969589, "dur": 1,
    "args": {
      "External id": 104159,
      "cbid": 251, "correlation": 104159
    }
  },
  {
    "ph": "f", "id": 104159, "pid": 76337, "tid": -914061504, "ts": 1716454222969589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969592, "dur": 0,
    "args": {
      "External id": 104160,
      "cbid": 251, "correlation": 104160
    }
  },
  {
    "ph": "f", "id": 104160, "pid": 76337, "tid": -914061504, "ts": 1716454222969592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223023645, "dur": 53,
    "args": {
      "External id": 104161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104161,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 104161, "pid": 5, "tid": 7, "ts": 1716454223023645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969594, "dur": 11,
    "args": {
      "External id": 104161,
      "cbid": 211, "correlation": 104161
    }
  },
  {
    "ph": "s", "id": 104161, "pid": 76337, "tid": -914061504, "ts": 1716454222969594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223023700, "dur": 55,
    "args": {
      "External id": 104186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104186,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104186, "pid": 5, "tid": 7, "ts": 1716454223023700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969678, "dur": 13,
    "args": {
      "External id": 104186,
      "cbid": 211, "correlation": 104186
    }
  },
  {
    "ph": "s", "id": 104186, "pid": 76337, "tid": -914061504, "ts": 1716454222969678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969779, "dur": 1,
    "args": {
      "External id": 104204,
      "cbid": 251, "correlation": 104204
    }
  },
  {
    "ph": "f", "id": 104204, "pid": 76337, "tid": -914061504, "ts": 1716454222969779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223023757, "dur": 61,
    "args": {
      "External id": 104206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104206,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 104206, "pid": 5, "tid": 7, "ts": 1716454223023757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969784, "dur": 13,
    "args": {
      "External id": 104206,
      "cbid": 211, "correlation": 104206
    }
  },
  {
    "ph": "s", "id": 104206, "pid": 76337, "tid": -914061504, "ts": 1716454222969784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223023819, "dur": 6,
    "args": {
      "External id": 104214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104214,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104214, "pid": 5, "tid": 7, "ts": 1716454223023819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969855, "dur": 13,
    "args": {
      "External id": 104214,
      "cbid": 211, "correlation": 104214
    }
  },
  {
    "ph": "s", "id": 104214, "pid": 76337, "tid": -914061504, "ts": 1716454222969855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223023826, "dur": 7,
    "args": {
      "External id": 104222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104222,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104222, "pid": 5, "tid": 7, "ts": 1716454223023826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969893, "dur": 8,
    "args": {
      "External id": 104222,
      "cbid": 211, "correlation": 104222
    }
  },
  {
    "ph": "s", "id": 104222, "pid": 76337, "tid": -914061504, "ts": 1716454222969893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223023835, "dur": 7,
    "args": {
      "External id": 104233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104233,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104233, "pid": 5, "tid": 7, "ts": 1716454223023835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222969966, "dur": 21,
    "args": {
      "External id": 104233,
      "cbid": 211, "correlation": 104233
    }
  },
  {
    "ph": "s", "id": 104233, "pid": 76337, "tid": -914061504, "ts": 1716454222969966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223023843, "dur": 8,
    "args": {
      "External id": 104255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104255,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 104255, "pid": 5, "tid": 7, "ts": 1716454223023843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970007, "dur": 9,
    "args": {
      "External id": 104255,
      "cbid": 211, "correlation": 104255
    }
  },
  {
    "ph": "s", "id": 104255, "pid": 76337, "tid": -914061504, "ts": 1716454222970007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970095, "dur": 2,
    "args": {
      "External id": 104266,
      "cbid": 251, "correlation": 104266
    }
  },
  {
    "ph": "f", "id": 104266, "pid": 76337, "tid": -914061504, "ts": 1716454222970095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223023854, "dur": 1,
    "args": {
      "External id": 104267,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104267,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 104267, "pid": 5, "tid": 7, "ts": 1716454223023854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970100, "dur": 11,
    "args": {
      "External id": 104267,
      "cbid": 51, "correlation": 104267
    }
  },
  {
    "ph": "s", "id": 104267, "pid": 76337, "tid": -914061504, "ts": 1716454222970100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223023857, "dur": 36,
    "args": {
      "External id": 104268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104268,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 104268, "pid": 5, "tid": 7, "ts": 1716454223023857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970113, "dur": 13,
    "args": {
      "External id": 104268,
      "cbid": 211, "correlation": 104268
    }
  },
  {
    "ph": "s", "id": 104268, "pid": 76337, "tid": -914061504, "ts": 1716454222970113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970185, "dur": 1,
    "args": {
      "External id": 104279,
      "cbid": 251, "correlation": 104279
    }
  },
  {
    "ph": "f", "id": 104279, "pid": 76337, "tid": -914061504, "ts": 1716454222970185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970189, "dur": 0,
    "args": {
      "External id": 104280,
      "cbid": 251, "correlation": 104280
    }
  },
  {
    "ph": "f", "id": 104280, "pid": 76337, "tid": -914061504, "ts": 1716454222970189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223023895, "dur": 12,
    "args": {
      "External id": 104281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104281,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104281, "pid": 5, "tid": 7, "ts": 1716454223023895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970190, "dur": 12,
    "args": {
      "External id": 104281,
      "cbid": 211, "correlation": 104281
    }
  },
  {
    "ph": "s", "id": 104281, "pid": 76337, "tid": -914061504, "ts": 1716454222970190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223023908, "dur": 5,
    "args": {
      "External id": 104283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104283,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104283, "pid": 5, "tid": 7, "ts": 1716454223023908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970204, "dur": 7,
    "args": {
      "External id": 104283,
      "cbid": 211, "correlation": 104283
    }
  },
  {
    "ph": "s", "id": 104283, "pid": 76337, "tid": -914061504, "ts": 1716454222970204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970263, "dur": 1,
    "args": {
      "External id": 104294,
      "cbid": 251, "correlation": 104294
    }
  },
  {
    "ph": "f", "id": 104294, "pid": 76337, "tid": -914061504, "ts": 1716454222970263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970266, "dur": 0,
    "args": {
      "External id": 104295,
      "cbid": 251, "correlation": 104295
    }
  },
  {
    "ph": "f", "id": 104295, "pid": 76337, "tid": -914061504, "ts": 1716454222970266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223023915, "dur": 8,
    "args": {
      "External id": 104296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104296,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104296, "pid": 5, "tid": 7, "ts": 1716454223023915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970268, "dur": 11,
    "args": {
      "External id": 104296,
      "cbid": 211, "correlation": 104296
    }
  },
  {
    "ph": "s", "id": 104296, "pid": 76337, "tid": -914061504, "ts": 1716454222970268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223023924, "dur": 3,
    "args": {
      "External id": 104298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104298,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104298, "pid": 5, "tid": 7, "ts": 1716454223023924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970280, "dur": 5,
    "args": {
      "External id": 104298,
      "cbid": 211, "correlation": 104298
    }
  },
  {
    "ph": "s", "id": 104298, "pid": 76337, "tid": -914061504, "ts": 1716454222970280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223023929, "dur": 19,
    "args": {
      "External id": 104323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104323,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 104323, "pid": 5, "tid": 7, "ts": 1716454223023929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970361, "dur": 12,
    "args": {
      "External id": 104323,
      "cbid": 211, "correlation": 104323
    }
  },
  {
    "ph": "s", "id": 104323, "pid": 76337, "tid": -914061504, "ts": 1716454222970361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970462, "dur": 2,
    "args": {
      "External id": 104341,
      "cbid": 251, "correlation": 104341
    }
  },
  {
    "ph": "f", "id": 104341, "pid": 76337, "tid": -914061504, "ts": 1716454222970462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223023950, "dur": 1,
    "args": {
      "External id": 104343,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104343,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 104343, "pid": 5, "tid": 7, "ts": 1716454223023950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970467, "dur": 10,
    "args": {
      "External id": 104343,
      "cbid": 51, "correlation": 104343
    }
  },
  {
    "ph": "s", "id": 104343, "pid": 76337, "tid": -914061504, "ts": 1716454222970467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223023954, "dur": 36,
    "args": {
      "External id": 104344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104344,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 104344, "pid": 5, "tid": 7, "ts": 1716454223023954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970479, "dur": 12,
    "args": {
      "External id": 104344,
      "cbid": 211, "correlation": 104344
    }
  },
  {
    "ph": "s", "id": 104344, "pid": 76337, "tid": -914061504, "ts": 1716454222970479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223023991, "dur": 4,
    "args": {
      "External id": 104352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104352,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104352, "pid": 5, "tid": 7, "ts": 1716454223023991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970550, "dur": 13,
    "args": {
      "External id": 104352,
      "cbid": 211, "correlation": 104352
    }
  },
  {
    "ph": "s", "id": 104352, "pid": 76337, "tid": -914061504, "ts": 1716454222970550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223023996, "dur": 8,
    "args": {
      "External id": 104360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104360, "pid": 5, "tid": 7, "ts": 1716454223023996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970592, "dur": 9,
    "args": {
      "External id": 104360,
      "cbid": 211, "correlation": 104360
    }
  },
  {
    "ph": "s", "id": 104360, "pid": 76337, "tid": -914061504, "ts": 1716454222970592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024006, "dur": 8,
    "args": {
      "External id": 104382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104382,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 104382, "pid": 5, "tid": 7, "ts": 1716454223024006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970643, "dur": 10,
    "args": {
      "External id": 104382,
      "cbid": 211, "correlation": 104382
    }
  },
  {
    "ph": "s", "id": 104382, "pid": 76337, "tid": -914061504, "ts": 1716454222970643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970733, "dur": 1,
    "args": {
      "External id": 104398,
      "cbid": 251, "correlation": 104398
    }
  },
  {
    "ph": "f", "id": 104398, "pid": 76337, "tid": -914061504, "ts": 1716454222970733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970738, "dur": 0,
    "args": {
      "External id": 104400,
      "cbid": 251, "correlation": 104400
    }
  },
  {
    "ph": "f", "id": 104400, "pid": 76337, "tid": -914061504, "ts": 1716454222970738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223024015, "dur": 186,
    "args": {
      "External id": 104401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104401,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104401, "pid": 5, "tid": 7, "ts": 1716454223024015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970740, "dur": 13,
    "args": {
      "External id": 104401,
      "cbid": 211, "correlation": 104401
    }
  },
  {
    "ph": "s", "id": 104401, "pid": 76337, "tid": -914061504, "ts": 1716454222970740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024202, "dur": 21,
    "args": {
      "External id": 104409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104409, "pid": 5, "tid": 7, "ts": 1716454223024202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970807, "dur": 12,
    "args": {
      "External id": 104409,
      "cbid": 211, "correlation": 104409
    }
  },
  {
    "ph": "s", "id": 104409, "pid": 76337, "tid": -914061504, "ts": 1716454222970807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024225, "dur": 22,
    "args": {
      "External id": 104417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104417, "pid": 5, "tid": 7, "ts": 1716454223024225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970837, "dur": 8,
    "args": {
      "External id": 104417,
      "cbid": 211, "correlation": 104417
    }
  },
  {
    "ph": "s", "id": 104417, "pid": 76337, "tid": -914061504, "ts": 1716454222970837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970919, "dur": 1,
    "args": {
      "External id": 104433,
      "cbid": 251, "correlation": 104433
    }
  },
  {
    "ph": "f", "id": 104433, "pid": 76337, "tid": -914061504, "ts": 1716454222970919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223024249, "dur": 1,
    "args": {
      "External id": 104435,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104435,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 104435, "pid": 5, "tid": 7, "ts": 1716454223024249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970924, "dur": 8,
    "args": {
      "External id": 104435,
      "cbid": 51, "correlation": 104435
    }
  },
  {
    "ph": "s", "id": 104435, "pid": 76337, "tid": -914061504, "ts": 1716454222970924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223024252, "dur": 109,
    "args": {
      "External id": 104436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104436,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 104436, "pid": 5, "tid": 7, "ts": 1716454223024252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970934, "dur": 11,
    "args": {
      "External id": 104436,
      "cbid": 211, "correlation": 104436
    }
  },
  {
    "ph": "s", "id": 104436, "pid": 76337, "tid": -914061504, "ts": 1716454222970934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223024362, "dur": 5,
    "args": {
      "External id": 104444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104444,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104444, "pid": 5, "tid": 7, "ts": 1716454223024362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222970985, "dur": 10,
    "args": {
      "External id": 104444,
      "cbid": 211, "correlation": 104444
    }
  },
  {
    "ph": "s", "id": 104444, "pid": 76337, "tid": -914061504, "ts": 1716454222970985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024369, "dur": 10,
    "args": {
      "External id": 104455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104455, "pid": 5, "tid": 7, "ts": 1716454223024369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971056, "dur": 12,
    "args": {
      "External id": 104455,
      "cbid": 211, "correlation": 104455
    }
  },
  {
    "ph": "s", "id": 104455, "pid": 76337, "tid": -914061504, "ts": 1716454222971056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971121, "dur": 0,
    "args": {
      "External id": 104467,
      "cbid": 317, "correlation": 104467
    }
  },
  {
    "ph": "f", "id": 104467, "pid": 76337, "tid": -914061504, "ts": 1716454222971121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971122, "dur": 0,
    "args": {
      "External id": 104468,
      "cbid": 203, "correlation": 104468
    }
  },
  {
    "ph": "f", "id": 104468, "pid": 76337, "tid": -914061504, "ts": 1716454222971122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971123, "dur": 0,
    "args": {
      "External id": 104469,
      "cbid": 205, "correlation": 104469
    }
  },
  {
    "ph": "f", "id": 104469, "pid": 76337, "tid": -914061504, "ts": 1716454222971123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024380, "dur": 5,
    "args": {
      "External id": 104473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104473,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104473, "pid": 5, "tid": 7, "ts": 1716454223024380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971139, "dur": 13,
    "args": {
      "External id": 104473,
      "cbid": 211, "correlation": 104473
    }
  },
  {
    "ph": "s", "id": 104473, "pid": 76337, "tid": -914061504, "ts": 1716454222971139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223024386, "dur": 37,
    "args": {
      "External id": 104475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104475,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 104475, "pid": 5, "tid": 7, "ts": 1716454223024386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971158, "dur": 7,
    "args": {
      "External id": 104475,
      "cbid": 211, "correlation": 104475
    }
  },
  {
    "ph": "s", "id": 104475, "pid": 76337, "tid": -914061504, "ts": 1716454222971158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024424, "dur": 6,
    "args": {
      "External id": 104477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104477,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104477, "pid": 5, "tid": 7, "ts": 1716454223024424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971168, "dur": 5,
    "args": {
      "External id": 104477,
      "cbid": 211, "correlation": 104477
    }
  },
  {
    "ph": "s", "id": 104477, "pid": 76337, "tid": -914061504, "ts": 1716454222971168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024431, "dur": 7,
    "args": {
      "External id": 104483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104483, "pid": 5, "tid": 7, "ts": 1716454223024431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971195, "dur": 9,
    "args": {
      "External id": 104483,
      "cbid": 211, "correlation": 104483
    }
  },
  {
    "ph": "s", "id": 104483, "pid": 76337, "tid": -914061504, "ts": 1716454222971195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223024440, "dur": 5,
    "args": {
      "External id": 104491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104491,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104491, "pid": 5, "tid": 7, "ts": 1716454223024440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971227, "dur": 8,
    "args": {
      "External id": 104491,
      "cbid": 211, "correlation": 104491
    }
  },
  {
    "ph": "s", "id": 104491, "pid": 76337, "tid": -914061504, "ts": 1716454222971227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024446, "dur": 11,
    "args": {
      "External id": 104511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104511,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 104511, "pid": 5, "tid": 7, "ts": 1716454223024446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971301, "dur": 12,
    "args": {
      "External id": 104511,
      "cbid": 211, "correlation": 104511
    }
  },
  {
    "ph": "s", "id": 104511, "pid": 76337, "tid": -914061504, "ts": 1716454222971301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223024458, "dur": 4,
    "args": {
      "External id": 104523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104523,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 104523, "pid": 5, "tid": 7, "ts": 1716454223024458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971323, "dur": 6,
    "args": {
      "External id": 104523,
      "cbid": 211, "correlation": 104523
    }
  },
  {
    "ph": "s", "id": 104523, "pid": 76337, "tid": -914061504, "ts": 1716454222971323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024464, "dur": 8,
    "args": {
      "External id": 104526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104526,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104526, "pid": 5, "tid": 7, "ts": 1716454223024464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971341, "dur": 7,
    "args": {
      "External id": 104526,
      "cbid": 211, "correlation": 104526
    }
  },
  {
    "ph": "s", "id": 104526, "pid": 76337, "tid": -914061504, "ts": 1716454222971341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223024473, "dur": 5,
    "args": {
      "External id": 104535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104535,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104535, "pid": 5, "tid": 7, "ts": 1716454223024473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971380, "dur": 10,
    "args": {
      "External id": 104535,
      "cbid": 211, "correlation": 104535
    }
  },
  {
    "ph": "s", "id": 104535, "pid": 76337, "tid": -914061504, "ts": 1716454222971380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971432, "dur": 0,
    "args": {
      "External id": 104545,
      "cbid": 317, "correlation": 104545
    }
  },
  {
    "ph": "f", "id": 104545, "pid": 76337, "tid": -914061504, "ts": 1716454222971432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971432, "dur": 0,
    "args": {
      "External id": 104546,
      "cbid": 203, "correlation": 104546
    }
  },
  {
    "ph": "f", "id": 104546, "pid": 76337, "tid": -914061504, "ts": 1716454222971432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971433, "dur": 0,
    "args": {
      "External id": 104547,
      "cbid": 205, "correlation": 104547
    }
  },
  {
    "ph": "f", "id": 104547, "pid": 76337, "tid": -914061504, "ts": 1716454222971433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024479, "dur": 5,
    "args": {
      "External id": 104551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104551,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104551, "pid": 5, "tid": 7, "ts": 1716454223024479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971446, "dur": 12,
    "args": {
      "External id": 104551,
      "cbid": 211, "correlation": 104551
    }
  },
  {
    "ph": "s", "id": 104551, "pid": 76337, "tid": -914061504, "ts": 1716454222971446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024486, "dur": 158,
    "args": {
      "External id": 104553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104553, "pid": 5, "tid": 7, "ts": 1716454223024486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971461, "dur": 5,
    "args": {
      "External id": 104553,
      "cbid": 211, "correlation": 104553
    }
  },
  {
    "ph": "s", "id": 104553, "pid": 76337, "tid": -914061504, "ts": 1716454222971461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223024646, "dur": 1,
    "args": {
      "External id": 104555,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104555,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14423076923076922
    }
  },
  {
    "ph": "f", "id": 104555, "pid": 5, "tid": 7, "ts": 1716454223024646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971471, "dur": 6,
    "args": {
      "External id": 104555,
      "cbid": 51, "correlation": 104555
    }
  },
  {
    "ph": "s", "id": 104555, "pid": 76337, "tid": -914061504, "ts": 1716454222971471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223024649, "dur": 266,
    "args": {
      "External id": 104556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104556,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104556, "pid": 5, "tid": 7, "ts": 1716454223024649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971479, "dur": 6,
    "args": {
      "External id": 104556,
      "cbid": 211, "correlation": 104556
    }
  },
  {
    "ph": "s", "id": 104556, "pid": 76337, "tid": -914061504, "ts": 1716454222971479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024917, "dur": 6,
    "args": {
      "External id": 104558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104558, "pid": 5, "tid": 7, "ts": 1716454223024917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971488, "dur": 5,
    "args": {
      "External id": 104558,
      "cbid": 211, "correlation": 104558
    }
  },
  {
    "ph": "s", "id": 104558, "pid": 76337, "tid": -914061504, "ts": 1716454222971488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024924, "dur": 6,
    "args": {
      "External id": 104564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104564, "pid": 5, "tid": 7, "ts": 1716454223024924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971516, "dur": 8,
    "args": {
      "External id": 104564,
      "cbid": 211, "correlation": 104564
    }
  },
  {
    "ph": "s", "id": 104564, "pid": 76337, "tid": -914061504, "ts": 1716454222971516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223024931, "dur": 3,
    "args": {
      "External id": 104572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104572,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 104572, "pid": 5, "tid": 7, "ts": 1716454223024931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971560, "dur": 9,
    "args": {
      "External id": 104572,
      "cbid": 211, "correlation": 104572
    }
  },
  {
    "ph": "s", "id": 104572, "pid": 76337, "tid": -914061504, "ts": 1716454222971560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971624, "dur": 1,
    "args": {
      "External id": 104588,
      "cbid": 251, "correlation": 104588
    }
  },
  {
    "ph": "f", "id": 104588, "pid": 76337, "tid": -914061504, "ts": 1716454222971624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971629, "dur": 0,
    "args": {
      "External id": 104590,
      "cbid": 251, "correlation": 104590
    }
  },
  {
    "ph": "f", "id": 104590, "pid": 76337, "tid": -914061504, "ts": 1716454222971629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223024936, "dur": 12,
    "args": {
      "External id": 104591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104591,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104591, "pid": 5, "tid": 7, "ts": 1716454223024936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971631, "dur": 11,
    "args": {
      "External id": 104591,
      "cbid": 211, "correlation": 104591
    }
  },
  {
    "ph": "s", "id": 104591, "pid": 76337, "tid": -914061504, "ts": 1716454222971631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223024949, "dur": 5,
    "args": {
      "External id": 104593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104593,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104593, "pid": 5, "tid": 7, "ts": 1716454223024949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971644, "dur": 6,
    "args": {
      "External id": 104593,
      "cbid": 211, "correlation": 104593
    }
  },
  {
    "ph": "s", "id": 104593, "pid": 76337, "tid": -914061504, "ts": 1716454222971644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024956, "dur": 6,
    "args": {
      "External id": 104603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104603, "pid": 5, "tid": 7, "ts": 1716454223024956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971701, "dur": 12,
    "args": {
      "External id": 104603,
      "cbid": 211, "correlation": 104603
    }
  },
  {
    "ph": "s", "id": 104603, "pid": 76337, "tid": -914061504, "ts": 1716454222971701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024963, "dur": 10,
    "args": {
      "External id": 104623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104623,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 104623, "pid": 5, "tid": 7, "ts": 1716454223024963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971767, "dur": 10,
    "args": {
      "External id": 104623,
      "cbid": 211, "correlation": 104623
    }
  },
  {
    "ph": "s", "id": 104623, "pid": 76337, "tid": -914061504, "ts": 1716454222971767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223024974, "dur": 4,
    "args": {
      "External id": 104635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104635,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 104635, "pid": 5, "tid": 7, "ts": 1716454223024974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971787, "dur": 6,
    "args": {
      "External id": 104635,
      "cbid": 211, "correlation": 104635
    }
  },
  {
    "ph": "s", "id": 104635, "pid": 76337, "tid": -914061504, "ts": 1716454222971787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223024979, "dur": 7,
    "args": {
      "External id": 104638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104638, "pid": 5, "tid": 7, "ts": 1716454223024979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971806, "dur": 7,
    "args": {
      "External id": 104638,
      "cbid": 211, "correlation": 104638
    }
  },
  {
    "ph": "s", "id": 104638, "pid": 76337, "tid": -914061504, "ts": 1716454222971806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223024987, "dur": 4,
    "args": {
      "External id": 104647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104647,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104647, "pid": 5, "tid": 7, "ts": 1716454223024987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971846, "dur": 10,
    "args": {
      "External id": 104647,
      "cbid": 211, "correlation": 104647
    }
  },
  {
    "ph": "s", "id": 104647, "pid": 76337, "tid": -914061504, "ts": 1716454222971846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971909, "dur": 0,
    "args": {
      "External id": 104657,
      "cbid": 317, "correlation": 104657
    }
  },
  {
    "ph": "f", "id": 104657, "pid": 76337, "tid": -914061504, "ts": 1716454222971909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971910, "dur": 0,
    "args": {
      "External id": 104658,
      "cbid": 203, "correlation": 104658
    }
  },
  {
    "ph": "f", "id": 104658, "pid": 76337, "tid": -914061504, "ts": 1716454222971910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971910, "dur": 0,
    "args": {
      "External id": 104659,
      "cbid": 205, "correlation": 104659
    }
  },
  {
    "ph": "f", "id": 104659, "pid": 76337, "tid": -914061504, "ts": 1716454222971910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024993, "dur": 5,
    "args": {
      "External id": 104663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104663,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104663, "pid": 5, "tid": 7, "ts": 1716454223024993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971924, "dur": 12,
    "args": {
      "External id": 104663,
      "cbid": 211, "correlation": 104663
    }
  },
  {
    "ph": "s", "id": 104663, "pid": 76337, "tid": -914061504, "ts": 1716454222971924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223024999, "dur": 160,
    "args": {
      "External id": 104665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104665, "pid": 5, "tid": 7, "ts": 1716454223024999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971938, "dur": 5,
    "args": {
      "External id": 104665,
      "cbid": 211, "correlation": 104665
    }
  },
  {
    "ph": "s", "id": 104665, "pid": 76337, "tid": -914061504, "ts": 1716454222971938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223025161, "dur": 1,
    "args": {
      "External id": 104667,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104667,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 104667, "pid": 5, "tid": 7, "ts": 1716454223025161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971949, "dur": 7,
    "args": {
      "External id": 104667,
      "cbid": 51, "correlation": 104667
    }
  },
  {
    "ph": "s", "id": 104667, "pid": 76337, "tid": -914061504, "ts": 1716454222971949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223025164, "dur": 254,
    "args": {
      "External id": 104668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104668,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104668, "pid": 5, "tid": 7, "ts": 1716454223025164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971957, "dur": 6,
    "args": {
      "External id": 104668,
      "cbid": 211, "correlation": 104668
    }
  },
  {
    "ph": "s", "id": 104668, "pid": 76337, "tid": -914061504, "ts": 1716454222971957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223025420, "dur": 6,
    "args": {
      "External id": 104670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104670, "pid": 5, "tid": 7, "ts": 1716454223025420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222971966, "dur": 5,
    "args": {
      "External id": 104670,
      "cbid": 211, "correlation": 104670
    }
  },
  {
    "ph": "s", "id": 104670, "pid": 76337, "tid": -914061504, "ts": 1716454222971966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223025427, "dur": 6,
    "args": {
      "External id": 104676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104676, "pid": 5, "tid": 7, "ts": 1716454223025427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972004, "dur": 9,
    "args": {
      "External id": 104676,
      "cbid": 211, "correlation": 104676
    }
  },
  {
    "ph": "s", "id": 104676, "pid": 76337, "tid": -914061504, "ts": 1716454222972004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223025434, "dur": 5,
    "args": {
      "External id": 104684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104684,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104684, "pid": 5, "tid": 7, "ts": 1716454223025434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972037, "dur": 8,
    "args": {
      "External id": 104684,
      "cbid": 211, "correlation": 104684
    }
  },
  {
    "ph": "s", "id": 104684, "pid": 76337, "tid": -914061504, "ts": 1716454222972037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223025440, "dur": 4,
    "args": {
      "External id": 104692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104692,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104692, "pid": 5, "tid": 7, "ts": 1716454223025440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972067, "dur": 8,
    "args": {
      "External id": 104692,
      "cbid": 211, "correlation": 104692
    }
  },
  {
    "ph": "s", "id": 104692, "pid": 76337, "tid": -914061504, "ts": 1716454222972067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223025446, "dur": 11,
    "args": {
      "External id": 104701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104701,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104701, "pid": 5, "tid": 7, "ts": 1716454223025446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972156, "dur": 14,
    "args": {
      "External id": 104701,
      "cbid": 211, "correlation": 104701
    }
  },
  {
    "ph": "s", "id": 104701, "pid": 76337, "tid": -914061504, "ts": 1716454222972156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223025458, "dur": 12,
    "args": {
      "External id": 104721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104721,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 104721, "pid": 5, "tid": 7, "ts": 1716454223025458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972224, "dur": 11,
    "args": {
      "External id": 104721,
      "cbid": 211, "correlation": 104721
    }
  },
  {
    "ph": "s", "id": 104721, "pid": 76337, "tid": -914061504, "ts": 1716454222972224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223025472, "dur": 4,
    "args": {
      "External id": 104733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104733,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104733, "pid": 5, "tid": 7, "ts": 1716454223025472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972245, "dur": 7,
    "args": {
      "External id": 104733,
      "cbid": 211, "correlation": 104733
    }
  },
  {
    "ph": "s", "id": 104733, "pid": 76337, "tid": -914061504, "ts": 1716454222972245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223025477, "dur": 10,
    "args": {
      "External id": 104736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104736,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104736, "pid": 5, "tid": 7, "ts": 1716454223025477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972264, "dur": 6,
    "args": {
      "External id": 104736,
      "cbid": 211, "correlation": 104736
    }
  },
  {
    "ph": "s", "id": 104736, "pid": 76337, "tid": -914061504, "ts": 1716454222972264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223025488, "dur": 6,
    "args": {
      "External id": 104745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104745,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104745, "pid": 5, "tid": 7, "ts": 1716454223025488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972304, "dur": 9,
    "args": {
      "External id": 104745,
      "cbid": 211, "correlation": 104745
    }
  },
  {
    "ph": "s", "id": 104745, "pid": 76337, "tid": -914061504, "ts": 1716454222972304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972357, "dur": 0,
    "args": {
      "External id": 104755,
      "cbid": 317, "correlation": 104755
    }
  },
  {
    "ph": "f", "id": 104755, "pid": 76337, "tid": -914061504, "ts": 1716454222972357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972358, "dur": 0,
    "args": {
      "External id": 104756,
      "cbid": 203, "correlation": 104756
    }
  },
  {
    "ph": "f", "id": 104756, "pid": 76337, "tid": -914061504, "ts": 1716454222972358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972359, "dur": 0,
    "args": {
      "External id": 104757,
      "cbid": 205, "correlation": 104757
    }
  },
  {
    "ph": "f", "id": 104757, "pid": 76337, "tid": -914061504, "ts": 1716454222972359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223025495, "dur": 6,
    "args": {
      "External id": 104761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104761, "pid": 5, "tid": 7, "ts": 1716454223025495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972375, "dur": 11,
    "args": {
      "External id": 104761,
      "cbid": 211, "correlation": 104761
    }
  },
  {
    "ph": "s", "id": 104761, "pid": 76337, "tid": -914061504, "ts": 1716454222972375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223025503, "dur": 313,
    "args": {
      "External id": 104763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104763,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104763, "pid": 5, "tid": 7, "ts": 1716454223025503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972388, "dur": 5,
    "args": {
      "External id": 104763,
      "cbid": 211, "correlation": 104763
    }
  },
  {
    "ph": "s", "id": 104763, "pid": 76337, "tid": -914061504, "ts": 1716454222972388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223025819, "dur": 1,
    "args": {
      "External id": 104765,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104765,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 104765, "pid": 5, "tid": 7, "ts": 1716454223025819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972399, "dur": 8,
    "args": {
      "External id": 104765,
      "cbid": 51, "correlation": 104765
    }
  },
  {
    "ph": "s", "id": 104765, "pid": 76337, "tid": -914061504, "ts": 1716454222972399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223025822, "dur": 488,
    "args": {
      "External id": 104766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104766,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104766, "pid": 5, "tid": 7, "ts": 1716454223025822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972408, "dur": 6,
    "args": {
      "External id": 104766,
      "cbid": 211, "correlation": 104766
    }
  },
  {
    "ph": "s", "id": 104766, "pid": 76337, "tid": -914061504, "ts": 1716454222972408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026312, "dur": 5,
    "args": {
      "External id": 104768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104768, "pid": 5, "tid": 7, "ts": 1716454223026312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972420, "dur": 5,
    "args": {
      "External id": 104768,
      "cbid": 211, "correlation": 104768
    }
  },
  {
    "ph": "s", "id": 104768, "pid": 76337, "tid": -914061504, "ts": 1716454222972420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223026318, "dur": 6,
    "args": {
      "External id": 104774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104774, "pid": 5, "tid": 7, "ts": 1716454223026318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972450, "dur": 9,
    "args": {
      "External id": 104774,
      "cbid": 211, "correlation": 104774
    }
  },
  {
    "ph": "s", "id": 104774, "pid": 76337, "tid": -914061504, "ts": 1716454222972450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223026325, "dur": 3,
    "args": {
      "External id": 104782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104782,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 104782, "pid": 5, "tid": 7, "ts": 1716454223026325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972492, "dur": 10,
    "args": {
      "External id": 104782,
      "cbid": 211, "correlation": 104782
    }
  },
  {
    "ph": "s", "id": 104782, "pid": 76337, "tid": -914061504, "ts": 1716454222972492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972554, "dur": 1,
    "args": {
      "External id": 104798,
      "cbid": 251, "correlation": 104798
    }
  },
  {
    "ph": "f", "id": 104798, "pid": 76337, "tid": -914061504, "ts": 1716454222972554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972559, "dur": 0,
    "args": {
      "External id": 104800,
      "cbid": 251, "correlation": 104800
    }
  },
  {
    "ph": "f", "id": 104800, "pid": 76337, "tid": -914061504, "ts": 1716454222972559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223026330, "dur": 12,
    "args": {
      "External id": 104801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104801,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104801, "pid": 5, "tid": 7, "ts": 1716454223026330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972561, "dur": 11,
    "args": {
      "External id": 104801,
      "cbid": 211, "correlation": 104801
    }
  },
  {
    "ph": "s", "id": 104801, "pid": 76337, "tid": -914061504, "ts": 1716454222972561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223026343, "dur": 5,
    "args": {
      "External id": 104803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104803,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104803, "pid": 5, "tid": 7, "ts": 1716454223026343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972574, "dur": 5,
    "args": {
      "External id": 104803,
      "cbid": 211, "correlation": 104803
    }
  },
  {
    "ph": "s", "id": 104803, "pid": 76337, "tid": -914061504, "ts": 1716454222972574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223026349, "dur": 6,
    "args": {
      "External id": 104813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104813, "pid": 5, "tid": 7, "ts": 1716454223026349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972630, "dur": 12,
    "args": {
      "External id": 104813,
      "cbid": 211, "correlation": 104813
    }
  },
  {
    "ph": "s", "id": 104813, "pid": 76337, "tid": -914061504, "ts": 1716454222972630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026356, "dur": 9,
    "args": {
      "External id": 104833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104833,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 104833, "pid": 5, "tid": 7, "ts": 1716454223026356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972696, "dur": 11,
    "args": {
      "External id": 104833,
      "cbid": 211, "correlation": 104833
    }
  },
  {
    "ph": "s", "id": 104833, "pid": 76337, "tid": -914061504, "ts": 1716454222972696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223026366, "dur": 4,
    "args": {
      "External id": 104845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104845,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 104845, "pid": 5, "tid": 7, "ts": 1716454223026366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972716, "dur": 6,
    "args": {
      "External id": 104845,
      "cbid": 211, "correlation": 104845
    }
  },
  {
    "ph": "s", "id": 104845, "pid": 76337, "tid": -914061504, "ts": 1716454222972716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223026371, "dur": 6,
    "args": {
      "External id": 104848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104848, "pid": 5, "tid": 7, "ts": 1716454223026371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972736, "dur": 7,
    "args": {
      "External id": 104848,
      "cbid": 211, "correlation": 104848
    }
  },
  {
    "ph": "s", "id": 104848, "pid": 76337, "tid": -914061504, "ts": 1716454222972736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223026379, "dur": 4,
    "args": {
      "External id": 104857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104857,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104857, "pid": 5, "tid": 7, "ts": 1716454223026379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972776, "dur": 10,
    "args": {
      "External id": 104857,
      "cbid": 211, "correlation": 104857
    }
  },
  {
    "ph": "s", "id": 104857, "pid": 76337, "tid": -914061504, "ts": 1716454222972776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972839, "dur": 0,
    "args": {
      "External id": 104867,
      "cbid": 317, "correlation": 104867
    }
  },
  {
    "ph": "f", "id": 104867, "pid": 76337, "tid": -914061504, "ts": 1716454222972839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972840, "dur": 0,
    "args": {
      "External id": 104868,
      "cbid": 203, "correlation": 104868
    }
  },
  {
    "ph": "f", "id": 104868, "pid": 76337, "tid": -914061504, "ts": 1716454222972840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972841, "dur": 0,
    "args": {
      "External id": 104869,
      "cbid": 205, "correlation": 104869
    }
  },
  {
    "ph": "f", "id": 104869, "pid": 76337, "tid": -914061504, "ts": 1716454222972841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026385, "dur": 5,
    "args": {
      "External id": 104873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104873,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104873, "pid": 5, "tid": 7, "ts": 1716454223026385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972855, "dur": 12,
    "args": {
      "External id": 104873,
      "cbid": 211, "correlation": 104873
    }
  },
  {
    "ph": "s", "id": 104873, "pid": 76337, "tid": -914061504, "ts": 1716454222972855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026391, "dur": 159,
    "args": {
      "External id": 104875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104875, "pid": 5, "tid": 7, "ts": 1716454223026391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972870, "dur": 5,
    "args": {
      "External id": 104875,
      "cbid": 211, "correlation": 104875
    }
  },
  {
    "ph": "s", "id": 104875, "pid": 76337, "tid": -914061504, "ts": 1716454222972870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223026553, "dur": 1,
    "args": {
      "External id": 104877,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104877,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 104877, "pid": 5, "tid": 7, "ts": 1716454223026553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972880, "dur": 6,
    "args": {
      "External id": 104877,
      "cbid": 51, "correlation": 104877
    }
  },
  {
    "ph": "s", "id": 104877, "pid": 76337, "tid": -914061504, "ts": 1716454222972880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223026556, "dur": 254,
    "args": {
      "External id": 104878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104878,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104878, "pid": 5, "tid": 7, "ts": 1716454223026556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972887, "dur": 6,
    "args": {
      "External id": 104878,
      "cbid": 211, "correlation": 104878
    }
  },
  {
    "ph": "s", "id": 104878, "pid": 76337, "tid": -914061504, "ts": 1716454222972887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026811, "dur": 6,
    "args": {
      "External id": 104880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104880, "pid": 5, "tid": 7, "ts": 1716454223026811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972897, "dur": 6,
    "args": {
      "External id": 104880,
      "cbid": 211, "correlation": 104880
    }
  },
  {
    "ph": "s", "id": 104880, "pid": 76337, "tid": -914061504, "ts": 1716454222972897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223026818, "dur": 6,
    "args": {
      "External id": 104886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104886, "pid": 5, "tid": 7, "ts": 1716454223026818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972926, "dur": 8,
    "args": {
      "External id": 104886,
      "cbid": 211, "correlation": 104886
    }
  },
  {
    "ph": "s", "id": 104886, "pid": 76337, "tid": -914061504, "ts": 1716454222972926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972994, "dur": 0,
    "args": {
      "External id": 104896,
      "cbid": 317, "correlation": 104896
    }
  },
  {
    "ph": "f", "id": 104896, "pid": 76337, "tid": -914061504, "ts": 1716454222972994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972994, "dur": 0,
    "args": {
      "External id": 104897,
      "cbid": 203, "correlation": 104897
    }
  },
  {
    "ph": "f", "id": 104897, "pid": 76337, "tid": -914061504, "ts": 1716454222972994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222972995, "dur": 0,
    "args": {
      "External id": 104898,
      "cbid": 205, "correlation": 104898
    }
  },
  {
    "ph": "f", "id": 104898, "pid": 76337, "tid": -914061504, "ts": 1716454222972995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026826, "dur": 8,
    "args": {
      "External id": 104902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104902,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104902, "pid": 5, "tid": 7, "ts": 1716454223026826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973010, "dur": 12,
    "args": {
      "External id": 104902,
      "cbid": 211, "correlation": 104902
    }
  },
  {
    "ph": "s", "id": 104902, "pid": 76337, "tid": -914061504, "ts": 1716454222973010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223026835, "dur": 3,
    "args": {
      "External id": 104904,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 104904,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 104904, "pid": 5, "tid": 7, "ts": 1716454223026835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973028, "dur": 15,
    "args": {
      "External id": 104904,
      "cbid": 51, "correlation": 104904
    }
  },
  {
    "ph": "s", "id": 104904, "pid": 76337, "tid": -914061504, "ts": 1716454222973028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223026839, "dur": 94,
    "args": {
      "External id": 104905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104905,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 104905, "pid": 5, "tid": 7, "ts": 1716454223026839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973045, "dur": 8,
    "args": {
      "External id": 104905,
      "cbid": 211, "correlation": 104905
    }
  },
  {
    "ph": "s", "id": 104905, "pid": 76337, "tid": -914061504, "ts": 1716454222973045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026935, "dur": 6,
    "args": {
      "External id": 104907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104907,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104907, "pid": 5, "tid": 7, "ts": 1716454223026935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973056, "dur": 6,
    "args": {
      "External id": 104907,
      "cbid": 211, "correlation": 104907
    }
  },
  {
    "ph": "s", "id": 104907, "pid": 76337, "tid": -914061504, "ts": 1716454222973056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223026942, "dur": 6,
    "args": {
      "External id": 104913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104913,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104913, "pid": 5, "tid": 7, "ts": 1716454223026942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973084, "dur": 9,
    "args": {
      "External id": 104913,
      "cbid": 211, "correlation": 104913
    }
  },
  {
    "ph": "s", "id": 104913, "pid": 76337, "tid": -914061504, "ts": 1716454222973084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223026949, "dur": 5,
    "args": {
      "External id": 104921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104921,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104921, "pid": 5, "tid": 7, "ts": 1716454223026949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973113, "dur": 8,
    "args": {
      "External id": 104921,
      "cbid": 211, "correlation": 104921
    }
  },
  {
    "ph": "s", "id": 104921, "pid": 76337, "tid": -914061504, "ts": 1716454222973113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223026955, "dur": 4,
    "args": {
      "External id": 104929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104929,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 104929, "pid": 5, "tid": 7, "ts": 1716454223026955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973142, "dur": 9,
    "args": {
      "External id": 104929,
      "cbid": 211, "correlation": 104929
    }
  },
  {
    "ph": "s", "id": 104929, "pid": 76337, "tid": -914061504, "ts": 1716454222973142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223026961, "dur": 11,
    "args": {
      "External id": 104938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104938,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104938, "pid": 5, "tid": 7, "ts": 1716454223026961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973185, "dur": 11,
    "args": {
      "External id": 104938,
      "cbid": 211, "correlation": 104938
    }
  },
  {
    "ph": "s", "id": 104938, "pid": 76337, "tid": -914061504, "ts": 1716454222973185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223026973, "dur": 12,
    "args": {
      "External id": 104958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104958,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 104958, "pid": 5, "tid": 7, "ts": 1716454223026973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973256, "dur": 11,
    "args": {
      "External id": 104958,
      "cbid": 211, "correlation": 104958
    }
  },
  {
    "ph": "s", "id": 104958, "pid": 76337, "tid": -914061504, "ts": 1716454222973256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223026986, "dur": 4,
    "args": {
      "External id": 104970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104970,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 104970, "pid": 5, "tid": 7, "ts": 1716454223026986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973277, "dur": 6,
    "args": {
      "External id": 104970,
      "cbid": 211, "correlation": 104970
    }
  },
  {
    "ph": "s", "id": 104970, "pid": 76337, "tid": -914061504, "ts": 1716454222973277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223026992, "dur": 10,
    "args": {
      "External id": 104973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104973,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104973, "pid": 5, "tid": 7, "ts": 1716454223026992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973296, "dur": 6,
    "args": {
      "External id": 104973,
      "cbid": 211, "correlation": 104973
    }
  },
  {
    "ph": "s", "id": 104973, "pid": 76337, "tid": -914061504, "ts": 1716454222973296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223027003, "dur": 6,
    "args": {
      "External id": 104982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104982,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104982, "pid": 5, "tid": 7, "ts": 1716454223027003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973333, "dur": 10,
    "args": {
      "External id": 104982,
      "cbid": 211, "correlation": 104982
    }
  },
  {
    "ph": "s", "id": 104982, "pid": 76337, "tid": -914061504, "ts": 1716454222973333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973385, "dur": 0,
    "args": {
      "External id": 104992,
      "cbid": 317, "correlation": 104992
    }
  },
  {
    "ph": "f", "id": 104992, "pid": 76337, "tid": -914061504, "ts": 1716454222973385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973386, "dur": 0,
    "args": {
      "External id": 104993,
      "cbid": 203, "correlation": 104993
    }
  },
  {
    "ph": "f", "id": 104993, "pid": 76337, "tid": -914061504, "ts": 1716454222973386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973387, "dur": 0,
    "args": {
      "External id": 104994,
      "cbid": 205, "correlation": 104994
    }
  },
  {
    "ph": "f", "id": 104994, "pid": 76337, "tid": -914061504, "ts": 1716454222973387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223027011, "dur": 6,
    "args": {
      "External id": 104998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 104998,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 104998, "pid": 5, "tid": 7, "ts": 1716454223027011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973401, "dur": 11,
    "args": {
      "External id": 104998,
      "cbid": 211, "correlation": 104998
    }
  },
  {
    "ph": "s", "id": 104998, "pid": 76337, "tid": -914061504, "ts": 1716454222973401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223027018, "dur": 312,
    "args": {
      "External id": 105000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105000,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105000, "pid": 5, "tid": 7, "ts": 1716454223027018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973414, "dur": 5,
    "args": {
      "External id": 105000,
      "cbid": 211, "correlation": 105000
    }
  },
  {
    "ph": "s", "id": 105000, "pid": 76337, "tid": -914061504, "ts": 1716454222973414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223027333, "dur": 1,
    "args": {
      "External id": 105002,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105002,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 105002, "pid": 5, "tid": 7, "ts": 1716454223027333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973425, "dur": 6,
    "args": {
      "External id": 105002,
      "cbid": 51, "correlation": 105002
    }
  },
  {
    "ph": "s", "id": 105002, "pid": 76337, "tid": -914061504, "ts": 1716454222973425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223027337, "dur": 489,
    "args": {
      "External id": 105003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105003,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105003, "pid": 5, "tid": 7, "ts": 1716454223027337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973432, "dur": 6,
    "args": {
      "External id": 105003,
      "cbid": 211, "correlation": 105003
    }
  },
  {
    "ph": "s", "id": 105003, "pid": 76337, "tid": -914061504, "ts": 1716454222973432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223027827, "dur": 5,
    "args": {
      "External id": 105005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105005,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105005, "pid": 5, "tid": 7, "ts": 1716454223027827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973443, "dur": 6,
    "args": {
      "External id": 105005,
      "cbid": 211, "correlation": 105005
    }
  },
  {
    "ph": "s", "id": 105005, "pid": 76337, "tid": -914061504, "ts": 1716454222973443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223027834, "dur": 6,
    "args": {
      "External id": 105011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105011, "pid": 5, "tid": 7, "ts": 1716454223027834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973471, "dur": 8,
    "args": {
      "External id": 105011,
      "cbid": 211, "correlation": 105011
    }
  },
  {
    "ph": "s", "id": 105011, "pid": 76337, "tid": -914061504, "ts": 1716454222973471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223027841, "dur": 3,
    "args": {
      "External id": 105019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105019,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 105019, "pid": 5, "tid": 7, "ts": 1716454223027841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973515, "dur": 9,
    "args": {
      "External id": 105019,
      "cbid": 211, "correlation": 105019
    }
  },
  {
    "ph": "s", "id": 105019, "pid": 76337, "tid": -914061504, "ts": 1716454222973515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973577, "dur": 1,
    "args": {
      "External id": 105035,
      "cbid": 251, "correlation": 105035
    }
  },
  {
    "ph": "f", "id": 105035, "pid": 76337, "tid": -914061504, "ts": 1716454222973577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973582, "dur": 0,
    "args": {
      "External id": 105037,
      "cbid": 251, "correlation": 105037
    }
  },
  {
    "ph": "f", "id": 105037, "pid": 76337, "tid": -914061504, "ts": 1716454222973582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223027846, "dur": 12,
    "args": {
      "External id": 105038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105038,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105038, "pid": 5, "tid": 7, "ts": 1716454223027846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973584, "dur": 12,
    "args": {
      "External id": 105038,
      "cbid": 211, "correlation": 105038
    }
  },
  {
    "ph": "s", "id": 105038, "pid": 76337, "tid": -914061504, "ts": 1716454222973584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223027859, "dur": 5,
    "args": {
      "External id": 105040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105040,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105040, "pid": 5, "tid": 7, "ts": 1716454223027859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973598, "dur": 5,
    "args": {
      "External id": 105040,
      "cbid": 211, "correlation": 105040
    }
  },
  {
    "ph": "s", "id": 105040, "pid": 76337, "tid": -914061504, "ts": 1716454222973598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223027866, "dur": 6,
    "args": {
      "External id": 105050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105050, "pid": 5, "tid": 7, "ts": 1716454223027866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973654, "dur": 11,
    "args": {
      "External id": 105050,
      "cbid": 211, "correlation": 105050
    }
  },
  {
    "ph": "s", "id": 105050, "pid": 76337, "tid": -914061504, "ts": 1716454222973654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223027873, "dur": 9,
    "args": {
      "External id": 105070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105070,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 105070, "pid": 5, "tid": 7, "ts": 1716454223027873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973718, "dur": 10,
    "args": {
      "External id": 105070,
      "cbid": 211, "correlation": 105070
    }
  },
  {
    "ph": "s", "id": 105070, "pid": 76337, "tid": -914061504, "ts": 1716454222973718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223027883, "dur": 4,
    "args": {
      "External id": 105082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105082,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 105082, "pid": 5, "tid": 7, "ts": 1716454223027883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973738, "dur": 6,
    "args": {
      "External id": 105082,
      "cbid": 211, "correlation": 105082
    }
  },
  {
    "ph": "s", "id": 105082, "pid": 76337, "tid": -914061504, "ts": 1716454222973738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223027888, "dur": 7,
    "args": {
      "External id": 105085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105085, "pid": 5, "tid": 7, "ts": 1716454223027888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973758, "dur": 6,
    "args": {
      "External id": 105085,
      "cbid": 211, "correlation": 105085
    }
  },
  {
    "ph": "s", "id": 105085, "pid": 76337, "tid": -914061504, "ts": 1716454222973758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223027896, "dur": 4,
    "args": {
      "External id": 105094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105094,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105094, "pid": 5, "tid": 7, "ts": 1716454223027896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973799, "dur": 10,
    "args": {
      "External id": 105094,
      "cbid": 211, "correlation": 105094
    }
  },
  {
    "ph": "s", "id": 105094, "pid": 76337, "tid": -914061504, "ts": 1716454222973799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973862, "dur": 0,
    "args": {
      "External id": 105104,
      "cbid": 317, "correlation": 105104
    }
  },
  {
    "ph": "f", "id": 105104, "pid": 76337, "tid": -914061504, "ts": 1716454222973862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973863, "dur": 0,
    "args": {
      "External id": 105105,
      "cbid": 203, "correlation": 105105
    }
  },
  {
    "ph": "f", "id": 105105, "pid": 76337, "tid": -914061504, "ts": 1716454222973863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973864, "dur": 0,
    "args": {
      "External id": 105106,
      "cbid": 205, "correlation": 105106
    }
  },
  {
    "ph": "f", "id": 105106, "pid": 76337, "tid": -914061504, "ts": 1716454222973864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223027902, "dur": 5,
    "args": {
      "External id": 105110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105110, "pid": 5, "tid": 7, "ts": 1716454223027902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973877, "dur": 12,
    "args": {
      "External id": 105110,
      "cbid": 211, "correlation": 105110
    }
  },
  {
    "ph": "s", "id": 105110, "pid": 76337, "tid": -914061504, "ts": 1716454222973877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223027908, "dur": 159,
    "args": {
      "External id": 105112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105112,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105112, "pid": 5, "tid": 7, "ts": 1716454223027908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973892, "dur": 6,
    "args": {
      "External id": 105112,
      "cbid": 211, "correlation": 105112
    }
  },
  {
    "ph": "s", "id": 105112, "pid": 76337, "tid": -914061504, "ts": 1716454222973892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223028070, "dur": 1,
    "args": {
      "External id": 105114,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105114,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 105114, "pid": 5, "tid": 7, "ts": 1716454223028070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973903, "dur": 6,
    "args": {
      "External id": 105114,
      "cbid": 51, "correlation": 105114
    }
  },
  {
    "ph": "s", "id": 105114, "pid": 76337, "tid": -914061504, "ts": 1716454222973903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223028073, "dur": 254,
    "args": {
      "External id": 105115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105115,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105115, "pid": 5, "tid": 7, "ts": 1716454223028073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973910, "dur": 6,
    "args": {
      "External id": 105115,
      "cbid": 211, "correlation": 105115
    }
  },
  {
    "ph": "s", "id": 105115, "pid": 76337, "tid": -914061504, "ts": 1716454222973910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223028329, "dur": 6,
    "args": {
      "External id": 105117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105117, "pid": 5, "tid": 7, "ts": 1716454223028329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973920, "dur": 5,
    "args": {
      "External id": 105117,
      "cbid": 211, "correlation": 105117
    }
  },
  {
    "ph": "s", "id": 105117, "pid": 76337, "tid": -914061504, "ts": 1716454222973920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223028336, "dur": 6,
    "args": {
      "External id": 105123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105123, "pid": 5, "tid": 7, "ts": 1716454223028336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222973949, "dur": 9,
    "args": {
      "External id": 105123,
      "cbid": 211, "correlation": 105123
    }
  },
  {
    "ph": "s", "id": 105123, "pid": 76337, "tid": -914061504, "ts": 1716454222973949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974018, "dur": 0,
    "args": {
      "External id": 105133,
      "cbid": 317, "correlation": 105133
    }
  },
  {
    "ph": "f", "id": 105133, "pid": 76337, "tid": -914061504, "ts": 1716454222974018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974019, "dur": 0,
    "args": {
      "External id": 105134,
      "cbid": 203, "correlation": 105134
    }
  },
  {
    "ph": "f", "id": 105134, "pid": 76337, "tid": -914061504, "ts": 1716454222974019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974020, "dur": 0,
    "args": {
      "External id": 105135,
      "cbid": 205, "correlation": 105135
    }
  },
  {
    "ph": "f", "id": 105135, "pid": 76337, "tid": -914061504, "ts": 1716454222974020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223028343, "dur": 7,
    "args": {
      "External id": 105139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105139,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105139, "pid": 5, "tid": 7, "ts": 1716454223028343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974031, "dur": 12,
    "args": {
      "External id": 105139,
      "cbid": 211, "correlation": 105139
    }
  },
  {
    "ph": "s", "id": 105139, "pid": 76337, "tid": -914061504, "ts": 1716454222974031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223028352, "dur": 3,
    "args": {
      "External id": 105141,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105141,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3888888888888888
    }
  },
  {
    "ph": "f", "id": 105141, "pid": 5, "tid": 7, "ts": 1716454223028352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974049, "dur": 10,
    "args": {
      "External id": 105141,
      "cbid": 51, "correlation": 105141
    }
  },
  {
    "ph": "s", "id": 105141, "pid": 76337, "tid": -914061504, "ts": 1716454222974049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223028356, "dur": 94,
    "args": {
      "External id": 105142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105142,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 105142, "pid": 5, "tid": 7, "ts": 1716454223028356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974060, "dur": 6,
    "args": {
      "External id": 105142,
      "cbid": 211, "correlation": 105142
    }
  },
  {
    "ph": "s", "id": 105142, "pid": 76337, "tid": -914061504, "ts": 1716454222974060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223028452, "dur": 6,
    "args": {
      "External id": 105144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105144,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105144, "pid": 5, "tid": 7, "ts": 1716454223028452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974070, "dur": 5,
    "args": {
      "External id": 105144,
      "cbid": 211, "correlation": 105144
    }
  },
  {
    "ph": "s", "id": 105144, "pid": 76337, "tid": -914061504, "ts": 1716454222974070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223028459, "dur": 6,
    "args": {
      "External id": 105150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105150,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105150, "pid": 5, "tid": 7, "ts": 1716454223028459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974098, "dur": 9,
    "args": {
      "External id": 105150,
      "cbid": 211, "correlation": 105150
    }
  },
  {
    "ph": "s", "id": 105150, "pid": 76337, "tid": -914061504, "ts": 1716454222974098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223028466, "dur": 5,
    "args": {
      "External id": 105158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105158,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105158, "pid": 5, "tid": 7, "ts": 1716454223028466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974127, "dur": 8,
    "args": {
      "External id": 105158,
      "cbid": 211, "correlation": 105158
    }
  },
  {
    "ph": "s", "id": 105158, "pid": 76337, "tid": -914061504, "ts": 1716454222974127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223028472, "dur": 4,
    "args": {
      "External id": 105166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105166,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105166, "pid": 5, "tid": 7, "ts": 1716454223028472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974156, "dur": 9,
    "args": {
      "External id": 105166,
      "cbid": 211, "correlation": 105166
    }
  },
  {
    "ph": "s", "id": 105166, "pid": 76337, "tid": -914061504, "ts": 1716454222974156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223028478, "dur": 11,
    "args": {
      "External id": 105175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105175,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105175, "pid": 5, "tid": 7, "ts": 1716454223028478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974201, "dur": 10,
    "args": {
      "External id": 105175,
      "cbid": 211, "correlation": 105175
    }
  },
  {
    "ph": "s", "id": 105175, "pid": 76337, "tid": -914061504, "ts": 1716454222974201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223028490, "dur": 12,
    "args": {
      "External id": 105195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105195,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 105195, "pid": 5, "tid": 7, "ts": 1716454223028490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974270, "dur": 11,
    "args": {
      "External id": 105195,
      "cbid": 211, "correlation": 105195
    }
  },
  {
    "ph": "s", "id": 105195, "pid": 76337, "tid": -914061504, "ts": 1716454222974270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223028504, "dur": 5,
    "args": {
      "External id": 105207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105207,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105207, "pid": 5, "tid": 7, "ts": 1716454223028504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974291, "dur": 7,
    "args": {
      "External id": 105207,
      "cbid": 211, "correlation": 105207
    }
  },
  {
    "ph": "s", "id": 105207, "pid": 76337, "tid": -914061504, "ts": 1716454222974291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223028510, "dur": 11,
    "args": {
      "External id": 105210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105210, "pid": 5, "tid": 7, "ts": 1716454223028510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974309, "dur": 7,
    "args": {
      "External id": 105210,
      "cbid": 211, "correlation": 105210
    }
  },
  {
    "ph": "s", "id": 105210, "pid": 76337, "tid": -914061504, "ts": 1716454222974309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223028522, "dur": 6,
    "args": {
      "External id": 105219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105219,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105219, "pid": 5, "tid": 7, "ts": 1716454223028522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974348, "dur": 9,
    "args": {
      "External id": 105219,
      "cbid": 211, "correlation": 105219
    }
  },
  {
    "ph": "s", "id": 105219, "pid": 76337, "tid": -914061504, "ts": 1716454222974348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974399, "dur": 0,
    "args": {
      "External id": 105229,
      "cbid": 317, "correlation": 105229
    }
  },
  {
    "ph": "f", "id": 105229, "pid": 76337, "tid": -914061504, "ts": 1716454222974399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974399, "dur": 0,
    "args": {
      "External id": 105230,
      "cbid": 203, "correlation": 105230
    }
  },
  {
    "ph": "f", "id": 105230, "pid": 76337, "tid": -914061504, "ts": 1716454222974399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974400, "dur": 0,
    "args": {
      "External id": 105231,
      "cbid": 205, "correlation": 105231
    }
  },
  {
    "ph": "f", "id": 105231, "pid": 76337, "tid": -914061504, "ts": 1716454222974400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223028529, "dur": 7,
    "args": {
      "External id": 105235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105235, "pid": 5, "tid": 7, "ts": 1716454223028529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974414, "dur": 11,
    "args": {
      "External id": 105235,
      "cbid": 211, "correlation": 105235
    }
  },
  {
    "ph": "s", "id": 105235, "pid": 76337, "tid": -914061504, "ts": 1716454222974414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223028537, "dur": 312,
    "args": {
      "External id": 105237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105237, "pid": 5, "tid": 7, "ts": 1716454223028537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974427, "dur": 5,
    "args": {
      "External id": 105237,
      "cbid": 211, "correlation": 105237
    }
  },
  {
    "ph": "s", "id": 105237, "pid": 76337, "tid": -914061504, "ts": 1716454222974427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223028851, "dur": 1,
    "args": {
      "External id": 105239,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105239,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 105239, "pid": 5, "tid": 7, "ts": 1716454223028851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974438, "dur": 6,
    "args": {
      "External id": 105239,
      "cbid": 51, "correlation": 105239
    }
  },
  {
    "ph": "s", "id": 105239, "pid": 76337, "tid": -914061504, "ts": 1716454222974438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223028855, "dur": 488,
    "args": {
      "External id": 105240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105240,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105240, "pid": 5, "tid": 7, "ts": 1716454223028855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974445, "dur": 7,
    "args": {
      "External id": 105240,
      "cbid": 211, "correlation": 105240
    }
  },
  {
    "ph": "s", "id": 105240, "pid": 76337, "tid": -914061504, "ts": 1716454222974445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029345, "dur": 5,
    "args": {
      "External id": 105242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105242,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105242, "pid": 5, "tid": 7, "ts": 1716454223029345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974455, "dur": 5,
    "args": {
      "External id": 105242,
      "cbid": 211, "correlation": 105242
    }
  },
  {
    "ph": "s", "id": 105242, "pid": 76337, "tid": -914061504, "ts": 1716454222974455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223029351, "dur": 6,
    "args": {
      "External id": 105248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105248, "pid": 5, "tid": 7, "ts": 1716454223029351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974483, "dur": 8,
    "args": {
      "External id": 105248,
      "cbid": 211, "correlation": 105248
    }
  },
  {
    "ph": "s", "id": 105248, "pid": 76337, "tid": -914061504, "ts": 1716454222974483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223029359, "dur": 3,
    "args": {
      "External id": 105256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105256,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 105256, "pid": 5, "tid": 7, "ts": 1716454223029359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974525, "dur": 9,
    "args": {
      "External id": 105256,
      "cbid": 211, "correlation": 105256
    }
  },
  {
    "ph": "s", "id": 105256, "pid": 76337, "tid": -914061504, "ts": 1716454222974525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974586, "dur": 1,
    "args": {
      "External id": 105272,
      "cbid": 251, "correlation": 105272
    }
  },
  {
    "ph": "f", "id": 105272, "pid": 76337, "tid": -914061504, "ts": 1716454222974586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974591, "dur": 0,
    "args": {
      "External id": 105274,
      "cbid": 251, "correlation": 105274
    }
  },
  {
    "ph": "f", "id": 105274, "pid": 76337, "tid": -914061504, "ts": 1716454222974591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223029363, "dur": 13,
    "args": {
      "External id": 105275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105275,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105275, "pid": 5, "tid": 7, "ts": 1716454223029363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974593, "dur": 12,
    "args": {
      "External id": 105275,
      "cbid": 211, "correlation": 105275
    }
  },
  {
    "ph": "s", "id": 105275, "pid": 76337, "tid": -914061504, "ts": 1716454222974593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223029378, "dur": 5,
    "args": {
      "External id": 105277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105277,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105277, "pid": 5, "tid": 7, "ts": 1716454223029378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974606, "dur": 5,
    "args": {
      "External id": 105277,
      "cbid": 211, "correlation": 105277
    }
  },
  {
    "ph": "s", "id": 105277, "pid": 76337, "tid": -914061504, "ts": 1716454222974606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223029384, "dur": 6,
    "args": {
      "External id": 105287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105287, "pid": 5, "tid": 7, "ts": 1716454223029384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974662, "dur": 11,
    "args": {
      "External id": 105287,
      "cbid": 211, "correlation": 105287
    }
  },
  {
    "ph": "s", "id": 105287, "pid": 76337, "tid": -914061504, "ts": 1716454222974662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029391, "dur": 9,
    "args": {
      "External id": 105307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105307,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 105307, "pid": 5, "tid": 7, "ts": 1716454223029391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974727, "dur": 11,
    "args": {
      "External id": 105307,
      "cbid": 211, "correlation": 105307
    }
  },
  {
    "ph": "s", "id": 105307, "pid": 76337, "tid": -914061504, "ts": 1716454222974727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223029402, "dur": 4,
    "args": {
      "External id": 105319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105319,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 105319, "pid": 5, "tid": 7, "ts": 1716454223029402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974748, "dur": 6,
    "args": {
      "External id": 105319,
      "cbid": 211, "correlation": 105319
    }
  },
  {
    "ph": "s", "id": 105319, "pid": 76337, "tid": -914061504, "ts": 1716454222974748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223029406, "dur": 6,
    "args": {
      "External id": 105322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105322,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105322, "pid": 5, "tid": 7, "ts": 1716454223029406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974767, "dur": 6,
    "args": {
      "External id": 105322,
      "cbid": 211, "correlation": 105322
    }
  },
  {
    "ph": "s", "id": 105322, "pid": 76337, "tid": -914061504, "ts": 1716454222974767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223029414, "dur": 4,
    "args": {
      "External id": 105331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105331,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105331, "pid": 5, "tid": 7, "ts": 1716454223029414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974808, "dur": 10,
    "args": {
      "External id": 105331,
      "cbid": 211, "correlation": 105331
    }
  },
  {
    "ph": "s", "id": 105331, "pid": 76337, "tid": -914061504, "ts": 1716454222974808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974870, "dur": 0,
    "args": {
      "External id": 105341,
      "cbid": 317, "correlation": 105341
    }
  },
  {
    "ph": "f", "id": 105341, "pid": 76337, "tid": -914061504, "ts": 1716454222974870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974871, "dur": 0,
    "args": {
      "External id": 105342,
      "cbid": 203, "correlation": 105342
    }
  },
  {
    "ph": "f", "id": 105342, "pid": 76337, "tid": -914061504, "ts": 1716454222974871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974872, "dur": 0,
    "args": {
      "External id": 105343,
      "cbid": 205, "correlation": 105343
    }
  },
  {
    "ph": "f", "id": 105343, "pid": 76337, "tid": -914061504, "ts": 1716454222974872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029420, "dur": 5,
    "args": {
      "External id": 105347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105347,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105347, "pid": 5, "tid": 7, "ts": 1716454223029420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974886, "dur": 12,
    "args": {
      "External id": 105347,
      "cbid": 211, "correlation": 105347
    }
  },
  {
    "ph": "s", "id": 105347, "pid": 76337, "tid": -914061504, "ts": 1716454222974886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029426, "dur": 160,
    "args": {
      "External id": 105349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105349, "pid": 5, "tid": 7, "ts": 1716454223029426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974900, "dur": 6,
    "args": {
      "External id": 105349,
      "cbid": 211, "correlation": 105349
    }
  },
  {
    "ph": "s", "id": 105349, "pid": 76337, "tid": -914061504, "ts": 1716454222974900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223029588, "dur": 1,
    "args": {
      "External id": 105351,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105351,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 105351, "pid": 5, "tid": 7, "ts": 1716454223029588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974911, "dur": 7,
    "args": {
      "External id": 105351,
      "cbid": 51, "correlation": 105351
    }
  },
  {
    "ph": "s", "id": 105351, "pid": 76337, "tid": -914061504, "ts": 1716454222974911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223029592, "dur": 254,
    "args": {
      "External id": 105352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105352,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105352, "pid": 5, "tid": 7, "ts": 1716454223029592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974919, "dur": 6,
    "args": {
      "External id": 105352,
      "cbid": 211, "correlation": 105352
    }
  },
  {
    "ph": "s", "id": 105352, "pid": 76337, "tid": -914061504, "ts": 1716454222974919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029847, "dur": 6,
    "args": {
      "External id": 105354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105354,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105354, "pid": 5, "tid": 7, "ts": 1716454223029847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974929, "dur": 5,
    "args": {
      "External id": 105354,
      "cbid": 211, "correlation": 105354
    }
  },
  {
    "ph": "s", "id": 105354, "pid": 76337, "tid": -914061504, "ts": 1716454222974929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223029854, "dur": 6,
    "args": {
      "External id": 105360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105360, "pid": 5, "tid": 7, "ts": 1716454223029854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222974958, "dur": 8,
    "args": {
      "External id": 105360,
      "cbid": 211, "correlation": 105360
    }
  },
  {
    "ph": "s", "id": 105360, "pid": 76337, "tid": -914061504, "ts": 1716454222974958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975027, "dur": 0,
    "args": {
      "External id": 105370,
      "cbid": 317, "correlation": 105370
    }
  },
  {
    "ph": "f", "id": 105370, "pid": 76337, "tid": -914061504, "ts": 1716454222975027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975028, "dur": 0,
    "args": {
      "External id": 105371,
      "cbid": 203, "correlation": 105371
    }
  },
  {
    "ph": "f", "id": 105371, "pid": 76337, "tid": -914061504, "ts": 1716454222975028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975028, "dur": 0,
    "args": {
      "External id": 105372,
      "cbid": 205, "correlation": 105372
    }
  },
  {
    "ph": "f", "id": 105372, "pid": 76337, "tid": -914061504, "ts": 1716454222975028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029861, "dur": 8,
    "args": {
      "External id": 105376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105376,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105376, "pid": 5, "tid": 7, "ts": 1716454223029861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975040, "dur": 13,
    "args": {
      "External id": 105376,
      "cbid": 211, "correlation": 105376
    }
  },
  {
    "ph": "s", "id": 105376, "pid": 76337, "tid": -914061504, "ts": 1716454222975040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223029871, "dur": 3,
    "args": {
      "External id": 105378,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105378,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.514673398548438
    }
  },
  {
    "ph": "f", "id": 105378, "pid": 5, "tid": 7, "ts": 1716454223029871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975058, "dur": 9,
    "args": {
      "External id": 105378,
      "cbid": 51, "correlation": 105378
    }
  },
  {
    "ph": "s", "id": 105378, "pid": 76337, "tid": -914061504, "ts": 1716454222975058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223029875, "dur": 93,
    "args": {
      "External id": 105379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105379,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 105379, "pid": 5, "tid": 7, "ts": 1716454223029875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975068, "dur": 6,
    "args": {
      "External id": 105379,
      "cbid": 211, "correlation": 105379
    }
  },
  {
    "ph": "s", "id": 105379, "pid": 76337, "tid": -914061504, "ts": 1716454222975068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223029969, "dur": 5,
    "args": {
      "External id": 105381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105381,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105381, "pid": 5, "tid": 7, "ts": 1716454223029969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975078, "dur": 5,
    "args": {
      "External id": 105381,
      "cbid": 211, "correlation": 105381
    }
  },
  {
    "ph": "s", "id": 105381, "pid": 76337, "tid": -914061504, "ts": 1716454222975078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223029976, "dur": 6,
    "args": {
      "External id": 105387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105387, "pid": 5, "tid": 7, "ts": 1716454223029976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975105, "dur": 8,
    "args": {
      "External id": 105387,
      "cbid": 211, "correlation": 105387
    }
  },
  {
    "ph": "s", "id": 105387, "pid": 76337, "tid": -914061504, "ts": 1716454222975105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223029983, "dur": 5,
    "args": {
      "External id": 105395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105395,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105395, "pid": 5, "tid": 7, "ts": 1716454223029983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975134, "dur": 8,
    "args": {
      "External id": 105395,
      "cbid": 211, "correlation": 105395
    }
  },
  {
    "ph": "s", "id": 105395, "pid": 76337, "tid": -914061504, "ts": 1716454222975134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223029990, "dur": 4,
    "args": {
      "External id": 105403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105403,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 105403, "pid": 5, "tid": 7, "ts": 1716454223029990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975163, "dur": 8,
    "args": {
      "External id": 105403,
      "cbid": 211, "correlation": 105403
    }
  },
  {
    "ph": "s", "id": 105403, "pid": 76337, "tid": -914061504, "ts": 1716454222975163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223029995, "dur": 14,
    "args": {
      "External id": 105414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105414,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105414, "pid": 5, "tid": 7, "ts": 1716454223029995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975245, "dur": 13,
    "args": {
      "External id": 105414,
      "cbid": 211, "correlation": 105414
    }
  },
  {
    "ph": "s", "id": 105414, "pid": 76337, "tid": -914061504, "ts": 1716454222975245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975302, "dur": 0,
    "args": {
      "External id": 105424,
      "cbid": 317, "correlation": 105424
    }
  },
  {
    "ph": "f", "id": 105424, "pid": 76337, "tid": -914061504, "ts": 1716454222975302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975303, "dur": 0,
    "args": {
      "External id": 105425,
      "cbid": 203, "correlation": 105425
    }
  },
  {
    "ph": "f", "id": 105425, "pid": 76337, "tid": -914061504, "ts": 1716454222975303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975304, "dur": 0,
    "args": {
      "External id": 105426,
      "cbid": 205, "correlation": 105426
    }
  },
  {
    "ph": "f", "id": 105426, "pid": 76337, "tid": -914061504, "ts": 1716454222975304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223030011, "dur": 8,
    "args": {
      "External id": 105430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105430,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105430, "pid": 5, "tid": 7, "ts": 1716454223030011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975317, "dur": 11,
    "args": {
      "External id": 105430,
      "cbid": 211, "correlation": 105430
    }
  },
  {
    "ph": "s", "id": 105430, "pid": 76337, "tid": -914061504, "ts": 1716454222975317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223030020, "dur": 161,
    "args": {
      "External id": 105432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105432, "pid": 5, "tid": 7, "ts": 1716454223030020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975330, "dur": 5,
    "args": {
      "External id": 105432,
      "cbid": 211, "correlation": 105432
    }
  },
  {
    "ph": "s", "id": 105432, "pid": 76337, "tid": -914061504, "ts": 1716454222975330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223030183, "dur": 1,
    "args": {
      "External id": 105434,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105434,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 105434, "pid": 5, "tid": 7, "ts": 1716454223030183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975341, "dur": 7,
    "args": {
      "External id": 105434,
      "cbid": 51, "correlation": 105434
    }
  },
  {
    "ph": "s", "id": 105434, "pid": 76337, "tid": -914061504, "ts": 1716454222975341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223030187, "dur": 643,
    "args": {
      "External id": 105435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105435,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105435, "pid": 5, "tid": 7, "ts": 1716454223030187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975349, "dur": 6,
    "args": {
      "External id": 105435,
      "cbid": 211, "correlation": 105435
    }
  },
  {
    "ph": "s", "id": 105435, "pid": 76337, "tid": -914061504, "ts": 1716454222975349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223030831, "dur": 11,
    "args": {
      "External id": 105437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105437,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105437, "pid": 5, "tid": 7, "ts": 1716454223030831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975359, "dur": 5,
    "args": {
      "External id": 105437,
      "cbid": 211, "correlation": 105437
    }
  },
  {
    "ph": "s", "id": 105437, "pid": 76337, "tid": -914061504, "ts": 1716454222975359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223030843, "dur": 14,
    "args": {
      "External id": 105443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105443, "pid": 5, "tid": 7, "ts": 1716454223030843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975387, "dur": 9,
    "args": {
      "External id": 105443,
      "cbid": 211, "correlation": 105443
    }
  },
  {
    "ph": "s", "id": 105443, "pid": 76337, "tid": -914061504, "ts": 1716454222975387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223030859, "dur": 30,
    "args": {
      "External id": 105452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105452,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105452, "pid": 5, "tid": 7, "ts": 1716454223030859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975476, "dur": 12,
    "args": {
      "External id": 105452,
      "cbid": 211, "correlation": 105452
    }
  },
  {
    "ph": "s", "id": 105452, "pid": 76337, "tid": -914061504, "ts": 1716454222975476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223030890, "dur": 30,
    "args": {
      "External id": 105472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105472,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 105472, "pid": 5, "tid": 7, "ts": 1716454223030890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975544, "dur": 12,
    "args": {
      "External id": 105472,
      "cbid": 211, "correlation": 105472
    }
  },
  {
    "ph": "s", "id": 105472, "pid": 76337, "tid": -914061504, "ts": 1716454222975544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223030921, "dur": 4,
    "args": {
      "External id": 105484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105484,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105484, "pid": 5, "tid": 7, "ts": 1716454223030921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975566, "dur": 6,
    "args": {
      "External id": 105484,
      "cbid": 211, "correlation": 105484
    }
  },
  {
    "ph": "s", "id": 105484, "pid": 76337, "tid": -914061504, "ts": 1716454222975566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223030927, "dur": 31,
    "args": {
      "External id": 105487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105487,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105487, "pid": 5, "tid": 7, "ts": 1716454223030927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975584, "dur": 8,
    "args": {
      "External id": 105487,
      "cbid": 211, "correlation": 105487
    }
  },
  {
    "ph": "s", "id": 105487, "pid": 76337, "tid": -914061504, "ts": 1716454222975584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223030960, "dur": 21,
    "args": {
      "External id": 105496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105496,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105496, "pid": 5, "tid": 7, "ts": 1716454223030960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975624, "dur": 10,
    "args": {
      "External id": 105496,
      "cbid": 211, "correlation": 105496
    }
  },
  {
    "ph": "s", "id": 105496, "pid": 76337, "tid": -914061504, "ts": 1716454222975624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975676, "dur": 0,
    "args": {
      "External id": 105506,
      "cbid": 317, "correlation": 105506
    }
  },
  {
    "ph": "f", "id": 105506, "pid": 76337, "tid": -914061504, "ts": 1716454222975676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975677, "dur": 0,
    "args": {
      "External id": 105507,
      "cbid": 203, "correlation": 105507
    }
  },
  {
    "ph": "f", "id": 105507, "pid": 76337, "tid": -914061504, "ts": 1716454222975677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975678, "dur": 0,
    "args": {
      "External id": 105508,
      "cbid": 205, "correlation": 105508
    }
  },
  {
    "ph": "f", "id": 105508, "pid": 76337, "tid": -914061504, "ts": 1716454222975678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223030981, "dur": 23,
    "args": {
      "External id": 105512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105512,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105512, "pid": 5, "tid": 7, "ts": 1716454223030981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975692, "dur": 13,
    "args": {
      "External id": 105512,
      "cbid": 211, "correlation": 105512
    }
  },
  {
    "ph": "s", "id": 105512, "pid": 76337, "tid": -914061504, "ts": 1716454222975692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223031006, "dur": 314,
    "args": {
      "External id": 105514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105514,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105514, "pid": 5, "tid": 7, "ts": 1716454223031006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975708, "dur": 5,
    "args": {
      "External id": 105514,
      "cbid": 211, "correlation": 105514
    }
  },
  {
    "ph": "s", "id": 105514, "pid": 76337, "tid": -914061504, "ts": 1716454222975708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223031322, "dur": 1,
    "args": {
      "External id": 105516,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105516,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 105516, "pid": 5, "tid": 7, "ts": 1716454223031322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975719, "dur": 7,
    "args": {
      "External id": 105516,
      "cbid": 51, "correlation": 105516
    }
  },
  {
    "ph": "s", "id": 105516, "pid": 76337, "tid": -914061504, "ts": 1716454222975719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223031325, "dur": 1227,
    "args": {
      "External id": 105517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105517,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105517, "pid": 5, "tid": 7, "ts": 1716454223031325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975727, "dur": 6,
    "args": {
      "External id": 105517,
      "cbid": 211, "correlation": 105517
    }
  },
  {
    "ph": "s", "id": 105517, "pid": 76337, "tid": -914061504, "ts": 1716454222975727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223032553, "dur": 12,
    "args": {
      "External id": 105519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105519,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105519, "pid": 5, "tid": 7, "ts": 1716454223032553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975737, "dur": 5,
    "args": {
      "External id": 105519,
      "cbid": 211, "correlation": 105519
    }
  },
  {
    "ph": "s", "id": 105519, "pid": 76337, "tid": -914061504, "ts": 1716454222975737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223032567, "dur": 14,
    "args": {
      "External id": 105525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105525,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105525, "pid": 5, "tid": 7, "ts": 1716454223032567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975765, "dur": 9,
    "args": {
      "External id": 105525,
      "cbid": 211, "correlation": 105525
    }
  },
  {
    "ph": "s", "id": 105525, "pid": 76337, "tid": -914061504, "ts": 1716454222975765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223032583, "dur": 3,
    "args": {
      "External id": 105533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105533,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 105533, "pid": 5, "tid": 7, "ts": 1716454223032583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975810, "dur": 9,
    "args": {
      "External id": 105533,
      "cbid": 211, "correlation": 105533
    }
  },
  {
    "ph": "s", "id": 105533, "pid": 76337, "tid": -914061504, "ts": 1716454222975810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975872, "dur": 1,
    "args": {
      "External id": 105549,
      "cbid": 251, "correlation": 105549
    }
  },
  {
    "ph": "f", "id": 105549, "pid": 76337, "tid": -914061504, "ts": 1716454222975872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975878, "dur": 0,
    "args": {
      "External id": 105551,
      "cbid": 251, "correlation": 105551
    }
  },
  {
    "ph": "f", "id": 105551, "pid": 76337, "tid": -914061504, "ts": 1716454222975878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223032587, "dur": 12,
    "args": {
      "External id": 105552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105552,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105552, "pid": 5, "tid": 7, "ts": 1716454223032587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975880, "dur": 11,
    "args": {
      "External id": 105552,
      "cbid": 211, "correlation": 105552
    }
  },
  {
    "ph": "s", "id": 105552, "pid": 76337, "tid": -914061504, "ts": 1716454222975880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223032601, "dur": 5,
    "args": {
      "External id": 105554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105554,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105554, "pid": 5, "tid": 7, "ts": 1716454223032601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975893, "dur": 6,
    "args": {
      "External id": 105554,
      "cbid": 211, "correlation": 105554
    }
  },
  {
    "ph": "s", "id": 105554, "pid": 76337, "tid": -914061504, "ts": 1716454222975893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223032607, "dur": 17,
    "args": {
      "External id": 105564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105564, "pid": 5, "tid": 7, "ts": 1716454223032607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222975950, "dur": 12,
    "args": {
      "External id": 105564,
      "cbid": 211, "correlation": 105564
    }
  },
  {
    "ph": "s", "id": 105564, "pid": 76337, "tid": -914061504, "ts": 1716454222975950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223032625, "dur": 17,
    "args": {
      "External id": 105584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105584,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 105584, "pid": 5, "tid": 7, "ts": 1716454223032625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976026, "dur": 11,
    "args": {
      "External id": 105584,
      "cbid": 211, "correlation": 105584
    }
  },
  {
    "ph": "s", "id": 105584, "pid": 76337, "tid": -914061504, "ts": 1716454222976026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223032644, "dur": 4,
    "args": {
      "External id": 105596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105596,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 105596, "pid": 5, "tid": 7, "ts": 1716454223032644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976047, "dur": 7,
    "args": {
      "External id": 105596,
      "cbid": 211, "correlation": 105596
    }
  },
  {
    "ph": "s", "id": 105596, "pid": 76337, "tid": -914061504, "ts": 1716454222976047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223032649, "dur": 16,
    "args": {
      "External id": 105599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105599, "pid": 5, "tid": 7, "ts": 1716454223032649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976067, "dur": 6,
    "args": {
      "External id": 105599,
      "cbid": 211, "correlation": 105599
    }
  },
  {
    "ph": "s", "id": 105599, "pid": 76337, "tid": -914061504, "ts": 1716454222976067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223032666, "dur": 11,
    "args": {
      "External id": 105608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105608,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105608, "pid": 5, "tid": 7, "ts": 1716454223032666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976108, "dur": 10,
    "args": {
      "External id": 105608,
      "cbid": 211, "correlation": 105608
    }
  },
  {
    "ph": "s", "id": 105608, "pid": 76337, "tid": -914061504, "ts": 1716454222976108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976172, "dur": 0,
    "args": {
      "External id": 105618,
      "cbid": 317, "correlation": 105618
    }
  },
  {
    "ph": "f", "id": 105618, "pid": 76337, "tid": -914061504, "ts": 1716454222976172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976173, "dur": 0,
    "args": {
      "External id": 105619,
      "cbid": 203, "correlation": 105619
    }
  },
  {
    "ph": "f", "id": 105619, "pid": 76337, "tid": -914061504, "ts": 1716454222976173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976174, "dur": 0,
    "args": {
      "External id": 105620,
      "cbid": 205, "correlation": 105620
    }
  },
  {
    "ph": "f", "id": 105620, "pid": 76337, "tid": -914061504, "ts": 1716454222976174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223032678, "dur": 12,
    "args": {
      "External id": 105624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105624, "pid": 5, "tid": 7, "ts": 1716454223032678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976187, "dur": 12,
    "args": {
      "External id": 105624,
      "cbid": 211, "correlation": 105624
    }
  },
  {
    "ph": "s", "id": 105624, "pid": 76337, "tid": -914061504, "ts": 1716454222976187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223032692, "dur": 159,
    "args": {
      "External id": 105626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105626, "pid": 5, "tid": 7, "ts": 1716454223032692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976201, "dur": 6,
    "args": {
      "External id": 105626,
      "cbid": 211, "correlation": 105626
    }
  },
  {
    "ph": "s", "id": 105626, "pid": 76337, "tid": -914061504, "ts": 1716454222976201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223032853, "dur": 1,
    "args": {
      "External id": 105628,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 105628,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 105628, "pid": 5, "tid": 7, "ts": 1716454223032853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976213, "dur": 6,
    "args": {
      "External id": 105628,
      "cbid": 51, "correlation": 105628
    }
  },
  {
    "ph": "s", "id": 105628, "pid": 76337, "tid": -914061504, "ts": 1716454222976213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223032857, "dur": 638,
    "args": {
      "External id": 105629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105629,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105629, "pid": 5, "tid": 7, "ts": 1716454223032857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976220, "dur": 6,
    "args": {
      "External id": 105629,
      "cbid": 211, "correlation": 105629
    }
  },
  {
    "ph": "s", "id": 105629, "pid": 76337, "tid": -914061504, "ts": 1716454222976220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223033497, "dur": 12,
    "args": {
      "External id": 105631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105631, "pid": 5, "tid": 7, "ts": 1716454223033497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976230, "dur": 5,
    "args": {
      "External id": 105631,
      "cbid": 211, "correlation": 105631
    }
  },
  {
    "ph": "s", "id": 105631, "pid": 76337, "tid": -914061504, "ts": 1716454222976230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223033511, "dur": 14,
    "args": {
      "External id": 105637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105637, "pid": 5, "tid": 7, "ts": 1716454223033511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976259, "dur": 8,
    "args": {
      "External id": 105637,
      "cbid": 211, "correlation": 105637
    }
  },
  {
    "ph": "s", "id": 105637, "pid": 76337, "tid": -914061504, "ts": 1716454222976259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976318, "dur": 0,
    "args": {
      "External id": 105647,
      "cbid": 317, "correlation": 105647
    }
  },
  {
    "ph": "f", "id": 105647, "pid": 76337, "tid": -914061504, "ts": 1716454222976318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976319, "dur": 0,
    "args": {
      "External id": 105648,
      "cbid": 203, "correlation": 105648
    }
  },
  {
    "ph": "f", "id": 105648, "pid": 76337, "tid": -914061504, "ts": 1716454222976319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976319, "dur": 0,
    "args": {
      "External id": 105649,
      "cbid": 205, "correlation": 105649
    }
  },
  {
    "ph": "f", "id": 105649, "pid": 76337, "tid": -914061504, "ts": 1716454222976319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223033526, "dur": 21,
    "args": {
      "External id": 105653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105653,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105653, "pid": 5, "tid": 7, "ts": 1716454223033526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976333, "dur": 12,
    "args": {
      "External id": 105653,
      "cbid": 211, "correlation": 105653
    }
  },
  {
    "ph": "s", "id": 105653, "pid": 76337, "tid": -914061504, "ts": 1716454222976333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223033549, "dur": 4,
    "args": {
      "External id": 105655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 105655, "pid": 5, "tid": 7, "ts": 1716454223033549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976350, "dur": 7,
    "args": {
      "External id": 105655,
      "cbid": 211, "correlation": 105655
    }
  },
  {
    "ph": "s", "id": 105655, "pid": 76337, "tid": -914061504, "ts": 1716454222976350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976361, "dur": 0,
    "args": {
      "External id": 105656,
      "cbid": 51, "correlation": 105656
    }
  },
  {
    "ph": "s", "id": 105656, "pid": 76337, "tid": -914061504, "ts": 1716454222976361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223033554, "dur": 172,
    "args": {
      "External id": 105657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105657,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 105657, "pid": 5, "tid": 7, "ts": 1716454223033554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976362, "dur": 5,
    "args": {
      "External id": 105657,
      "cbid": 211, "correlation": 105657
    }
  },
  {
    "ph": "s", "id": 105657, "pid": 76337, "tid": -914061504, "ts": 1716454222976362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223033728, "dur": 15,
    "args": {
      "External id": 105662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105662, "pid": 5, "tid": 7, "ts": 1716454223033728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976388, "dur": 8,
    "args": {
      "External id": 105662,
      "cbid": 211, "correlation": 105662
    }
  },
  {
    "ph": "s", "id": 105662, "pid": 76337, "tid": -914061504, "ts": 1716454222976388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223033745, "dur": 12,
    "args": {
      "External id": 105670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105670,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105670, "pid": 5, "tid": 7, "ts": 1716454223033745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976417, "dur": 8,
    "args": {
      "External id": 105670,
      "cbid": 211, "correlation": 105670
    }
  },
  {
    "ph": "s", "id": 105670, "pid": 76337, "tid": -914061504, "ts": 1716454222976417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223033758, "dur": 10,
    "args": {
      "External id": 105678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105678,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105678, "pid": 5, "tid": 7, "ts": 1716454223033758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976445, "dur": 8,
    "args": {
      "External id": 105678,
      "cbid": 211, "correlation": 105678
    }
  },
  {
    "ph": "s", "id": 105678, "pid": 76337, "tid": -914061504, "ts": 1716454222976445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223033769, "dur": 18,
    "args": {
      "External id": 105698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105698,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 105698, "pid": 5, "tid": 7, "ts": 1716454223033769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976527, "dur": 12,
    "args": {
      "External id": 105698,
      "cbid": 211, "correlation": 105698
    }
  },
  {
    "ph": "s", "id": 105698, "pid": 76337, "tid": -914061504, "ts": 1716454222976527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223033787, "dur": 4,
    "args": {
      "External id": 105710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105710,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 105710, "pid": 5, "tid": 7, "ts": 1716454223033787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976549, "dur": 7,
    "args": {
      "External id": 105710,
      "cbid": 211, "correlation": 105710
    }
  },
  {
    "ph": "s", "id": 105710, "pid": 76337, "tid": -914061504, "ts": 1716454222976549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223033793, "dur": 16,
    "args": {
      "External id": 105713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105713, "pid": 5, "tid": 7, "ts": 1716454223033793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976568, "dur": 7,
    "args": {
      "External id": 105713,
      "cbid": 211, "correlation": 105713
    }
  },
  {
    "ph": "s", "id": 105713, "pid": 76337, "tid": -914061504, "ts": 1716454222976568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976624, "dur": 0,
    "args": {
      "External id": 105724,
      "cbid": 317, "correlation": 105724
    }
  },
  {
    "ph": "f", "id": 105724, "pid": 76337, "tid": -914061504, "ts": 1716454222976624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976625, "dur": 0,
    "args": {
      "External id": 105725,
      "cbid": 203, "correlation": 105725
    }
  },
  {
    "ph": "f", "id": 105725, "pid": 76337, "tid": -914061504, "ts": 1716454222976625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976626, "dur": 0,
    "args": {
      "External id": 105726,
      "cbid": 205, "correlation": 105726
    }
  },
  {
    "ph": "f", "id": 105726, "pid": 76337, "tid": -914061504, "ts": 1716454222976626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223033811, "dur": 11,
    "args": {
      "External id": 105730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105730,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105730, "pid": 5, "tid": 7, "ts": 1716454223033811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976639, "dur": 12,
    "args": {
      "External id": 105730,
      "cbid": 211, "correlation": 105730
    }
  },
  {
    "ph": "s", "id": 105730, "pid": 76337, "tid": -914061504, "ts": 1716454222976639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223033823, "dur": 3,
    "args": {
      "External id": 105732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105732,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 105732, "pid": 5, "tid": 7, "ts": 1716454223033823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976655, "dur": 6,
    "args": {
      "External id": 105732,
      "cbid": 211, "correlation": 105732
    }
  },
  {
    "ph": "s", "id": 105732, "pid": 76337, "tid": -914061504, "ts": 1716454222976655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976664, "dur": 0,
    "args": {
      "External id": 105733,
      "cbid": 51, "correlation": 105733
    }
  },
  {
    "ph": "s", "id": 105733, "pid": 76337, "tid": -914061504, "ts": 1716454222976664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223033828, "dur": 88,
    "args": {
      "External id": 105734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105734,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 105734, "pid": 5, "tid": 7, "ts": 1716454223033828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976665, "dur": 5,
    "args": {
      "External id": 105734,
      "cbid": 211, "correlation": 105734
    }
  },
  {
    "ph": "s", "id": 105734, "pid": 76337, "tid": -914061504, "ts": 1716454222976665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223033917, "dur": 15,
    "args": {
      "External id": 105739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105739, "pid": 5, "tid": 7, "ts": 1716454223033917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976691, "dur": 8,
    "args": {
      "External id": 105739,
      "cbid": 211, "correlation": 105739
    }
  },
  {
    "ph": "s", "id": 105739, "pid": 76337, "tid": -914061504, "ts": 1716454222976691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223033934, "dur": 82,
    "args": {
      "External id": 105748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105748,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105748, "pid": 5, "tid": 7, "ts": 1716454223033934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976773, "dur": 15,
    "args": {
      "External id": 105748,
      "cbid": 211, "correlation": 105748
    }
  },
  {
    "ph": "s", "id": 105748, "pid": 76337, "tid": -914061504, "ts": 1716454222976773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223034017, "dur": 30,
    "args": {
      "External id": 105770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105770,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105770, "pid": 5, "tid": 7, "ts": 1716454223034017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976831, "dur": 10,
    "args": {
      "External id": 105770,
      "cbid": 211, "correlation": 105770
    }
  },
  {
    "ph": "s", "id": 105770, "pid": 76337, "tid": -914061504, "ts": 1716454222976831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976924, "dur": 1,
    "args": {
      "External id": 105781,
      "cbid": 251, "correlation": 105781
    }
  },
  {
    "ph": "f", "id": 105781, "pid": 76337, "tid": -914061504, "ts": 1716454222976924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223034048, "dur": 162,
    "args": {
      "External id": 105782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105782,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105782, "pid": 5, "tid": 7, "ts": 1716454223034048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222976929, "dur": 13,
    "args": {
      "External id": 105782,
      "cbid": 211, "correlation": 105782
    }
  },
  {
    "ph": "s", "id": 105782, "pid": 76337, "tid": -914061504, "ts": 1716454222976929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977011, "dur": 1,
    "args": {
      "External id": 105793,
      "cbid": 251, "correlation": 105793
    }
  },
  {
    "ph": "f", "id": 105793, "pid": 76337, "tid": -914061504, "ts": 1716454222977011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223034212, "dur": 156,
    "args": {
      "External id": 105794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105794,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105794, "pid": 5, "tid": 7, "ts": 1716454223034212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977015, "dur": 12,
    "args": {
      "External id": 105794,
      "cbid": 211, "correlation": 105794
    }
  },
  {
    "ph": "s", "id": 105794, "pid": 76337, "tid": -914061504, "ts": 1716454222977015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977082, "dur": 1,
    "args": {
      "External id": 105805,
      "cbid": 251, "correlation": 105805
    }
  },
  {
    "ph": "f", "id": 105805, "pid": 76337, "tid": -914061504, "ts": 1716454222977082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223034369, "dur": 157,
    "args": {
      "External id": 105806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105806,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105806, "pid": 5, "tid": 7, "ts": 1716454223034369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977086, "dur": 11,
    "args": {
      "External id": 105806,
      "cbid": 211, "correlation": 105806
    }
  },
  {
    "ph": "s", "id": 105806, "pid": 76337, "tid": -914061504, "ts": 1716454222977086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223034527, "dur": 330,
    "args": {
      "External id": 105831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105831,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105831, "pid": 5, "tid": 7, "ts": 1716454223034527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977174, "dur": 13,
    "args": {
      "External id": 105831,
      "cbid": 211, "correlation": 105831
    }
  },
  {
    "ph": "s", "id": 105831, "pid": 76337, "tid": -914061504, "ts": 1716454222977174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977276, "dur": 1,
    "args": {
      "External id": 105849,
      "cbid": 251, "correlation": 105849
    }
  },
  {
    "ph": "f", "id": 105849, "pid": 76337, "tid": -914061504, "ts": 1716454222977276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223034859, "dur": 165,
    "args": {
      "External id": 105851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105851,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105851, "pid": 5, "tid": 7, "ts": 1716454223034859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977282, "dur": 13,
    "args": {
      "External id": 105851,
      "cbid": 211, "correlation": 105851
    }
  },
  {
    "ph": "s", "id": 105851, "pid": 76337, "tid": -914061504, "ts": 1716454222977282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223035025, "dur": 19,
    "args": {
      "External id": 105859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105859,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105859, "pid": 5, "tid": 7, "ts": 1716454223035025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977353, "dur": 13,
    "args": {
      "External id": 105859,
      "cbid": 211, "correlation": 105859
    }
  },
  {
    "ph": "s", "id": 105859, "pid": 76337, "tid": -914061504, "ts": 1716454222977353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223035046, "dur": 27,
    "args": {
      "External id": 105867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105867,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105867, "pid": 5, "tid": 7, "ts": 1716454223035046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977392, "dur": 8,
    "args": {
      "External id": 105867,
      "cbid": 211, "correlation": 105867
    }
  },
  {
    "ph": "s", "id": 105867, "pid": 76337, "tid": -914061504, "ts": 1716454222977392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223035074, "dur": 19,
    "args": {
      "External id": 105878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105878, "pid": 5, "tid": 7, "ts": 1716454223035074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977467, "dur": 12,
    "args": {
      "External id": 105878,
      "cbid": 211, "correlation": 105878
    }
  },
  {
    "ph": "s", "id": 105878, "pid": 76337, "tid": -914061504, "ts": 1716454222977467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223035094, "dur": 16,
    "args": {
      "External id": 105900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105900,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105900, "pid": 5, "tid": 7, "ts": 1716454223035094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977497, "dur": 8,
    "args": {
      "External id": 105900,
      "cbid": 211, "correlation": 105900
    }
  },
  {
    "ph": "s", "id": 105900, "pid": 76337, "tid": -914061504, "ts": 1716454222977497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977584, "dur": 1,
    "args": {
      "External id": 105911,
      "cbid": 251, "correlation": 105911
    }
  },
  {
    "ph": "f", "id": 105911, "pid": 76337, "tid": -914061504, "ts": 1716454222977584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223035111, "dur": 87,
    "args": {
      "External id": 105912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105912,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 105912, "pid": 5, "tid": 7, "ts": 1716454223035111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977589, "dur": 13,
    "args": {
      "External id": 105912,
      "cbid": 211, "correlation": 105912
    }
  },
  {
    "ph": "s", "id": 105912, "pid": 76337, "tid": -914061504, "ts": 1716454222977589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977659, "dur": 1,
    "args": {
      "External id": 105923,
      "cbid": 251, "correlation": 105923
    }
  },
  {
    "ph": "f", "id": 105923, "pid": 76337, "tid": -914061504, "ts": 1716454222977659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977663, "dur": 0,
    "args": {
      "External id": 105924,
      "cbid": 251, "correlation": 105924
    }
  },
  {
    "ph": "f", "id": 105924, "pid": 76337, "tid": -914061504, "ts": 1716454222977663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223035200, "dur": 12,
    "args": {
      "External id": 105925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105925,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105925, "pid": 5, "tid": 7, "ts": 1716454223035200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977664, "dur": 12,
    "args": {
      "External id": 105925,
      "cbid": 211, "correlation": 105925
    }
  },
  {
    "ph": "s", "id": 105925, "pid": 76337, "tid": -914061504, "ts": 1716454222977664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223035213, "dur": 6,
    "args": {
      "External id": 105927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105927,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105927, "pid": 5, "tid": 7, "ts": 1716454223035213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977678, "dur": 6,
    "args": {
      "External id": 105927,
      "cbid": 211, "correlation": 105927
    }
  },
  {
    "ph": "s", "id": 105927, "pid": 76337, "tid": -914061504, "ts": 1716454222977678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977735, "dur": 1,
    "args": {
      "External id": 105938,
      "cbid": 251, "correlation": 105938
    }
  },
  {
    "ph": "f", "id": 105938, "pid": 76337, "tid": -914061504, "ts": 1716454222977735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977738, "dur": 0,
    "args": {
      "External id": 105939,
      "cbid": 251, "correlation": 105939
    }
  },
  {
    "ph": "f", "id": 105939, "pid": 76337, "tid": -914061504, "ts": 1716454222977738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223035220, "dur": 8,
    "args": {
      "External id": 105940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105940,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105940, "pid": 5, "tid": 7, "ts": 1716454223035220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977740, "dur": 12,
    "args": {
      "External id": 105940,
      "cbid": 211, "correlation": 105940
    }
  },
  {
    "ph": "s", "id": 105940, "pid": 76337, "tid": -914061504, "ts": 1716454222977740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223035230, "dur": 3,
    "args": {
      "External id": 105942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105942,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105942, "pid": 5, "tid": 7, "ts": 1716454223035230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977753, "dur": 6,
    "args": {
      "External id": 105942,
      "cbid": 211, "correlation": 105942
    }
  },
  {
    "ph": "s", "id": 105942, "pid": 76337, "tid": -914061504, "ts": 1716454222977753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223035234, "dur": 54,
    "args": {
      "External id": 105967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105967,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 105967, "pid": 5, "tid": 7, "ts": 1716454223035234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977831, "dur": 12,
    "args": {
      "External id": 105967,
      "cbid": 211, "correlation": 105967
    }
  },
  {
    "ph": "s", "id": 105967, "pid": 76337, "tid": -914061504, "ts": 1716454222977831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977931, "dur": 1,
    "args": {
      "External id": 105985,
      "cbid": 251, "correlation": 105985
    }
  },
  {
    "ph": "f", "id": 105985, "pid": 76337, "tid": -914061504, "ts": 1716454222977931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223035289, "dur": 89,
    "args": {
      "External id": 105987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105987,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 105987, "pid": 5, "tid": 7, "ts": 1716454223035289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222977937, "dur": 14,
    "args": {
      "External id": 105987,
      "cbid": 211, "correlation": 105987
    }
  },
  {
    "ph": "s", "id": 105987, "pid": 76337, "tid": -914061504, "ts": 1716454222977937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223035379, "dur": 9,
    "args": {
      "External id": 105995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 105995,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 105995, "pid": 5, "tid": 7, "ts": 1716454223035379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978016, "dur": 13,
    "args": {
      "External id": 105995,
      "cbid": 211, "correlation": 105995
    }
  },
  {
    "ph": "s", "id": 105995, "pid": 76337, "tid": -914061504, "ts": 1716454222978016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223035390, "dur": 21,
    "args": {
      "External id": 106003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106003, "pid": 5, "tid": 7, "ts": 1716454223035390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978060, "dur": 9,
    "args": {
      "External id": 106003,
      "cbid": 211, "correlation": 106003
    }
  },
  {
    "ph": "s", "id": 106003, "pid": 76337, "tid": -914061504, "ts": 1716454222978060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223035412, "dur": 17,
    "args": {
      "External id": 106025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106025,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106025, "pid": 5, "tid": 7, "ts": 1716454223035412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978112, "dur": 10,
    "args": {
      "External id": 106025,
      "cbid": 211, "correlation": 106025
    }
  },
  {
    "ph": "s", "id": 106025, "pid": 76337, "tid": -914061504, "ts": 1716454222978112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978201, "dur": 1,
    "args": {
      "External id": 106041,
      "cbid": 251, "correlation": 106041
    }
  },
  {
    "ph": "f", "id": 106041, "pid": 76337, "tid": -914061504, "ts": 1716454222978201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978206, "dur": 0,
    "args": {
      "External id": 106043,
      "cbid": 251, "correlation": 106043
    }
  },
  {
    "ph": "f", "id": 106043, "pid": 76337, "tid": -914061504, "ts": 1716454222978206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223035430, "dur": 491,
    "args": {
      "External id": 106044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106044,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106044, "pid": 5, "tid": 7, "ts": 1716454223035430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978209, "dur": 12,
    "args": {
      "External id": 106044,
      "cbid": 211, "correlation": 106044
    }
  },
  {
    "ph": "s", "id": 106044, "pid": 76337, "tid": -914061504, "ts": 1716454222978209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223035923, "dur": 64,
    "args": {
      "External id": 106052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106052, "pid": 5, "tid": 7, "ts": 1716454223035923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978277, "dur": 13,
    "args": {
      "External id": 106052,
      "cbid": 211, "correlation": 106052
    }
  },
  {
    "ph": "s", "id": 106052, "pid": 76337, "tid": -914061504, "ts": 1716454222978277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223035989, "dur": 67,
    "args": {
      "External id": 106060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106060, "pid": 5, "tid": 7, "ts": 1716454223035989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978310, "dur": 8,
    "args": {
      "External id": 106060,
      "cbid": 211, "correlation": 106060
    }
  },
  {
    "ph": "s", "id": 106060, "pid": 76337, "tid": -914061504, "ts": 1716454222978310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978390, "dur": 1,
    "args": {
      "External id": 106076,
      "cbid": 251, "correlation": 106076
    }
  },
  {
    "ph": "f", "id": 106076, "pid": 76337, "tid": -914061504, "ts": 1716454222978390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223036058, "dur": 1,
    "args": {
      "External id": 106078,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 106078,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 106078, "pid": 5, "tid": 7, "ts": 1716454223036058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978395, "dur": 10,
    "args": {
      "External id": 106078,
      "cbid": 51, "correlation": 106078
    }
  },
  {
    "ph": "s", "id": 106078, "pid": 76337, "tid": -914061504, "ts": 1716454222978395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223036062, "dur": 270,
    "args": {
      "External id": 106079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106079,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106079, "pid": 5, "tid": 7, "ts": 1716454223036062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978407, "dur": 11,
    "args": {
      "External id": 106079,
      "cbid": 211, "correlation": 106079
    }
  },
  {
    "ph": "s", "id": 106079, "pid": 76337, "tid": -914061504, "ts": 1716454222978407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223036333, "dur": 13,
    "args": {
      "External id": 106087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106087,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106087, "pid": 5, "tid": 7, "ts": 1716454223036333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978448, "dur": 11,
    "args": {
      "External id": 106087,
      "cbid": 211, "correlation": 106087
    }
  },
  {
    "ph": "s", "id": 106087, "pid": 76337, "tid": -914061504, "ts": 1716454222978448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223036348, "dur": 37,
    "args": {
      "External id": 106098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106098, "pid": 5, "tid": 7, "ts": 1716454223036348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978517, "dur": 12,
    "args": {
      "External id": 106098,
      "cbid": 211, "correlation": 106098
    }
  },
  {
    "ph": "s", "id": 106098, "pid": 76337, "tid": -914061504, "ts": 1716454222978517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978583, "dur": 0,
    "args": {
      "External id": 106110,
      "cbid": 317, "correlation": 106110
    }
  },
  {
    "ph": "f", "id": 106110, "pid": 76337, "tid": -914061504, "ts": 1716454222978583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978584, "dur": 0,
    "args": {
      "External id": 106111,
      "cbid": 203, "correlation": 106111
    }
  },
  {
    "ph": "f", "id": 106111, "pid": 76337, "tid": -914061504, "ts": 1716454222978584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978585, "dur": 0,
    "args": {
      "External id": 106112,
      "cbid": 205, "correlation": 106112
    }
  },
  {
    "ph": "f", "id": 106112, "pid": 76337, "tid": -914061504, "ts": 1716454222978585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223036386, "dur": 12,
    "args": {
      "External id": 106116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106116, "pid": 5, "tid": 7, "ts": 1716454223036386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978601, "dur": 12,
    "args": {
      "External id": 106116,
      "cbid": 211, "correlation": 106116
    }
  },
  {
    "ph": "s", "id": 106116, "pid": 76337, "tid": -914061504, "ts": 1716454222978601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223036400, "dur": 4,
    "args": {
      "External id": 106118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106118, "pid": 5, "tid": 7, "ts": 1716454223036400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978618, "dur": 6,
    "args": {
      "External id": 106118,
      "cbid": 211, "correlation": 106118
    }
  },
  {
    "ph": "s", "id": 106118, "pid": 76337, "tid": -914061504, "ts": 1716454222978618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978627, "dur": 0,
    "args": {
      "External id": 106119,
      "cbid": 51, "correlation": 106119
    }
  },
  {
    "ph": "s", "id": 106119, "pid": 76337, "tid": -914061504, "ts": 1716454222978627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223036406, "dur": 95,
    "args": {
      "External id": 106120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106120,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 106120, "pid": 5, "tid": 7, "ts": 1716454223036406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978628, "dur": 5,
    "args": {
      "External id": 106120,
      "cbid": 211, "correlation": 106120
    }
  },
  {
    "ph": "s", "id": 106120, "pid": 76337, "tid": -914061504, "ts": 1716454222978628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223036502, "dur": 16,
    "args": {
      "External id": 106125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106125,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106125, "pid": 5, "tid": 7, "ts": 1716454223036502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978656, "dur": 9,
    "args": {
      "External id": 106125,
      "cbid": 211, "correlation": 106125
    }
  },
  {
    "ph": "s", "id": 106125, "pid": 76337, "tid": -914061504, "ts": 1716454222978656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223036520, "dur": 12,
    "args": {
      "External id": 106133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106133,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106133, "pid": 5, "tid": 7, "ts": 1716454223036520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978687, "dur": 8,
    "args": {
      "External id": 106133,
      "cbid": 211, "correlation": 106133
    }
  },
  {
    "ph": "s", "id": 106133, "pid": 76337, "tid": -914061504, "ts": 1716454222978687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223036533, "dur": 30,
    "args": {
      "External id": 106142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106142,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106142, "pid": 5, "tid": 7, "ts": 1716454223036533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978727, "dur": 10,
    "args": {
      "External id": 106142,
      "cbid": 211, "correlation": 106142
    }
  },
  {
    "ph": "s", "id": 106142, "pid": 76337, "tid": -914061504, "ts": 1716454222978727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223036565, "dur": 30,
    "args": {
      "External id": 106162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106162,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 106162, "pid": 5, "tid": 7, "ts": 1716454223036565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978799, "dur": 12,
    "args": {
      "External id": 106162,
      "cbid": 211, "correlation": 106162
    }
  },
  {
    "ph": "s", "id": 106162, "pid": 76337, "tid": -914061504, "ts": 1716454222978799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223036596, "dur": 5,
    "args": {
      "External id": 106174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106174,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106174, "pid": 5, "tid": 7, "ts": 1716454223036596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978822, "dur": 6,
    "args": {
      "External id": 106174,
      "cbid": 211, "correlation": 106174
    }
  },
  {
    "ph": "s", "id": 106174, "pid": 76337, "tid": -914061504, "ts": 1716454222978822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223036602, "dur": 31,
    "args": {
      "External id": 106177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106177, "pid": 5, "tid": 7, "ts": 1716454223036602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978840, "dur": 6,
    "args": {
      "External id": 106177,
      "cbid": 211, "correlation": 106177
    }
  },
  {
    "ph": "s", "id": 106177, "pid": 76337, "tid": -914061504, "ts": 1716454222978840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223036635, "dur": 20,
    "args": {
      "External id": 106186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106186,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106186, "pid": 5, "tid": 7, "ts": 1716454223036635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978880, "dur": 10,
    "args": {
      "External id": 106186,
      "cbid": 211, "correlation": 106186
    }
  },
  {
    "ph": "s", "id": 106186, "pid": 76337, "tid": -914061504, "ts": 1716454222978880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978932, "dur": 0,
    "args": {
      "External id": 106196,
      "cbid": 317, "correlation": 106196
    }
  },
  {
    "ph": "f", "id": 106196, "pid": 76337, "tid": -914061504, "ts": 1716454222978932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978933, "dur": 0,
    "args": {
      "External id": 106197,
      "cbid": 203, "correlation": 106197
    }
  },
  {
    "ph": "f", "id": 106197, "pid": 76337, "tid": -914061504, "ts": 1716454222978933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978933, "dur": 0,
    "args": {
      "External id": 106198,
      "cbid": 205, "correlation": 106198
    }
  },
  {
    "ph": "f", "id": 106198, "pid": 76337, "tid": -914061504, "ts": 1716454222978933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223036656, "dur": 23,
    "args": {
      "External id": 106202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106202, "pid": 5, "tid": 7, "ts": 1716454223036656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978947, "dur": 12,
    "args": {
      "External id": 106202,
      "cbid": 211, "correlation": 106202
    }
  },
  {
    "ph": "s", "id": 106202, "pid": 76337, "tid": -914061504, "ts": 1716454222978947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223036680, "dur": 314,
    "args": {
      "External id": 106204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106204,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106204, "pid": 5, "tid": 7, "ts": 1716454223036680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978961, "dur": 5,
    "args": {
      "External id": 106204,
      "cbid": 211, "correlation": 106204
    }
  },
  {
    "ph": "s", "id": 106204, "pid": 76337, "tid": -914061504, "ts": 1716454222978961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223036995, "dur": 1,
    "args": {
      "External id": 106206,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 106206,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 106206, "pid": 5, "tid": 7, "ts": 1716454223036995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978972, "dur": 15,
    "args": {
      "External id": 106206,
      "cbid": 51, "correlation": 106206
    }
  },
  {
    "ph": "s", "id": 106206, "pid": 76337, "tid": -914061504, "ts": 1716454222978972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223036999, "dur": 1241,
    "args": {
      "External id": 106207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106207,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106207, "pid": 5, "tid": 7, "ts": 1716454223036999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978988, "dur": 7,
    "args": {
      "External id": 106207,
      "cbid": 211, "correlation": 106207
    }
  },
  {
    "ph": "s", "id": 106207, "pid": 76337, "tid": -914061504, "ts": 1716454222978988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223038241, "dur": 13,
    "args": {
      "External id": 106209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106209,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106209, "pid": 5, "tid": 7, "ts": 1716454223038241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222978999, "dur": 6,
    "args": {
      "External id": 106209,
      "cbid": 211, "correlation": 106209
    }
  },
  {
    "ph": "s", "id": 106209, "pid": 76337, "tid": -914061504, "ts": 1716454222978999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223038256, "dur": 15,
    "args": {
      "External id": 106215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106215,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106215, "pid": 5, "tid": 7, "ts": 1716454223038256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979028, "dur": 9,
    "args": {
      "External id": 106215,
      "cbid": 211, "correlation": 106215
    }
  },
  {
    "ph": "s", "id": 106215, "pid": 76337, "tid": -914061504, "ts": 1716454222979028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223038272, "dur": 3,
    "args": {
      "External id": 106223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106223,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 106223, "pid": 5, "tid": 7, "ts": 1716454223038272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979074, "dur": 9,
    "args": {
      "External id": 106223,
      "cbid": 211, "correlation": 106223
    }
  },
  {
    "ph": "s", "id": 106223, "pid": 76337, "tid": -914061504, "ts": 1716454222979074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979139, "dur": 1,
    "args": {
      "External id": 106239,
      "cbid": 251, "correlation": 106239
    }
  },
  {
    "ph": "f", "id": 106239, "pid": 76337, "tid": -914061504, "ts": 1716454222979139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979145, "dur": 0,
    "args": {
      "External id": 106241,
      "cbid": 251, "correlation": 106241
    }
  },
  {
    "ph": "f", "id": 106241, "pid": 76337, "tid": -914061504, "ts": 1716454222979145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223038276, "dur": 13,
    "args": {
      "External id": 106242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106242,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106242, "pid": 5, "tid": 7, "ts": 1716454223038276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979146, "dur": 12,
    "args": {
      "External id": 106242,
      "cbid": 211, "correlation": 106242
    }
  },
  {
    "ph": "s", "id": 106242, "pid": 76337, "tid": -914061504, "ts": 1716454222979146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223038290, "dur": 5,
    "args": {
      "External id": 106244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106244,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106244, "pid": 5, "tid": 7, "ts": 1716454223038290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979160, "dur": 5,
    "args": {
      "External id": 106244,
      "cbid": 211, "correlation": 106244
    }
  },
  {
    "ph": "s", "id": 106244, "pid": 76337, "tid": -914061504, "ts": 1716454222979160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223038297, "dur": 17,
    "args": {
      "External id": 106254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106254, "pid": 5, "tid": 7, "ts": 1716454223038297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979217, "dur": 12,
    "args": {
      "External id": 106254,
      "cbid": 211, "correlation": 106254
    }
  },
  {
    "ph": "s", "id": 106254, "pid": 76337, "tid": -914061504, "ts": 1716454222979217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223038315, "dur": 17,
    "args": {
      "External id": 106274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106274,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 106274, "pid": 5, "tid": 7, "ts": 1716454223038315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979283, "dur": 11,
    "args": {
      "External id": 106274,
      "cbid": 211, "correlation": 106274
    }
  },
  {
    "ph": "s", "id": 106274, "pid": 76337, "tid": -914061504, "ts": 1716454222979283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223038333, "dur": 4,
    "args": {
      "External id": 106286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106286,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 106286, "pid": 5, "tid": 7, "ts": 1716454223038333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979304, "dur": 7,
    "args": {
      "External id": 106286,
      "cbid": 211, "correlation": 106286
    }
  },
  {
    "ph": "s", "id": 106286, "pid": 76337, "tid": -914061504, "ts": 1716454222979304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223038338, "dur": 17,
    "args": {
      "External id": 106289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106289, "pid": 5, "tid": 7, "ts": 1716454223038338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979323, "dur": 7,
    "args": {
      "External id": 106289,
      "cbid": 211, "correlation": 106289
    }
  },
  {
    "ph": "s", "id": 106289, "pid": 76337, "tid": -914061504, "ts": 1716454222979323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223038356, "dur": 11,
    "args": {
      "External id": 106298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106298,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106298, "pid": 5, "tid": 7, "ts": 1716454223038356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979364, "dur": 9,
    "args": {
      "External id": 106298,
      "cbid": 211, "correlation": 106298
    }
  },
  {
    "ph": "s", "id": 106298, "pid": 76337, "tid": -914061504, "ts": 1716454222979364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979426, "dur": 0,
    "args": {
      "External id": 106308,
      "cbid": 317, "correlation": 106308
    }
  },
  {
    "ph": "f", "id": 106308, "pid": 76337, "tid": -914061504, "ts": 1716454222979426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979427, "dur": 0,
    "args": {
      "External id": 106309,
      "cbid": 203, "correlation": 106309
    }
  },
  {
    "ph": "f", "id": 106309, "pid": 76337, "tid": -914061504, "ts": 1716454222979427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979428, "dur": 0,
    "args": {
      "External id": 106310,
      "cbid": 205, "correlation": 106310
    }
  },
  {
    "ph": "f", "id": 106310, "pid": 76337, "tid": -914061504, "ts": 1716454222979428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223038369, "dur": 11,
    "args": {
      "External id": 106314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106314,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106314, "pid": 5, "tid": 7, "ts": 1716454223038369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979441, "dur": 12,
    "args": {
      "External id": 106314,
      "cbid": 211, "correlation": 106314
    }
  },
  {
    "ph": "s", "id": 106314, "pid": 76337, "tid": -914061504, "ts": 1716454222979441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223038381, "dur": 160,
    "args": {
      "External id": 106316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106316,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106316, "pid": 5, "tid": 7, "ts": 1716454223038381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979456, "dur": 6,
    "args": {
      "External id": 106316,
      "cbid": 211, "correlation": 106316
    }
  },
  {
    "ph": "s", "id": 106316, "pid": 76337, "tid": -914061504, "ts": 1716454222979456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223038544, "dur": 1,
    "args": {
      "External id": 106318,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 106318,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 106318, "pid": 5, "tid": 7, "ts": 1716454223038544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979467, "dur": 6,
    "args": {
      "External id": 106318,
      "cbid": 51, "correlation": 106318
    }
  },
  {
    "ph": "s", "id": 106318, "pid": 76337, "tid": -914061504, "ts": 1716454222979467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223038547, "dur": 639,
    "args": {
      "External id": 106319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106319,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106319, "pid": 5, "tid": 7, "ts": 1716454223038547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979475, "dur": 6,
    "args": {
      "External id": 106319,
      "cbid": 211, "correlation": 106319
    }
  },
  {
    "ph": "s", "id": 106319, "pid": 76337, "tid": -914061504, "ts": 1716454222979475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223039187, "dur": 12,
    "args": {
      "External id": 106321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106321,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106321, "pid": 5, "tid": 7, "ts": 1716454223039187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979485, "dur": 5,
    "args": {
      "External id": 106321,
      "cbid": 211, "correlation": 106321
    }
  },
  {
    "ph": "s", "id": 106321, "pid": 76337, "tid": -914061504, "ts": 1716454222979485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223039200, "dur": 15,
    "args": {
      "External id": 106327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106327, "pid": 5, "tid": 7, "ts": 1716454223039200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979514, "dur": 8,
    "args": {
      "External id": 106327,
      "cbid": 211, "correlation": 106327
    }
  },
  {
    "ph": "s", "id": 106327, "pid": 76337, "tid": -914061504, "ts": 1716454222979514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979572, "dur": 0,
    "args": {
      "External id": 106337,
      "cbid": 317, "correlation": 106337
    }
  },
  {
    "ph": "f", "id": 106337, "pid": 76337, "tid": -914061504, "ts": 1716454222979572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979572, "dur": 0,
    "args": {
      "External id": 106338,
      "cbid": 203, "correlation": 106338
    }
  },
  {
    "ph": "f", "id": 106338, "pid": 76337, "tid": -914061504, "ts": 1716454222979572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979573, "dur": 0,
    "args": {
      "External id": 106339,
      "cbid": 205, "correlation": 106339
    }
  },
  {
    "ph": "f", "id": 106339, "pid": 76337, "tid": -914061504, "ts": 1716454222979573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223039216, "dur": 21,
    "args": {
      "External id": 106343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106343,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106343, "pid": 5, "tid": 7, "ts": 1716454223039216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979585, "dur": 11,
    "args": {
      "External id": 106343,
      "cbid": 211, "correlation": 106343
    }
  },
  {
    "ph": "s", "id": 106343, "pid": 76337, "tid": -914061504, "ts": 1716454222979585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223039238, "dur": 4,
    "args": {
      "External id": 106345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106345,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106345, "pid": 5, "tid": 7, "ts": 1716454223039238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979600, "dur": 6,
    "args": {
      "External id": 106345,
      "cbid": 211, "correlation": 106345
    }
  },
  {
    "ph": "s", "id": 106345, "pid": 76337, "tid": -914061504, "ts": 1716454222979600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979610, "dur": 0,
    "args": {
      "External id": 106346,
      "cbid": 51, "correlation": 106346
    }
  },
  {
    "ph": "s", "id": 106346, "pid": 76337, "tid": -914061504, "ts": 1716454222979610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223039244, "dur": 168,
    "args": {
      "External id": 106347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106347,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 106347, "pid": 5, "tid": 7, "ts": 1716454223039244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979610, "dur": 5,
    "args": {
      "External id": 106347,
      "cbid": 211, "correlation": 106347
    }
  },
  {
    "ph": "s", "id": 106347, "pid": 76337, "tid": -914061504, "ts": 1716454222979610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223039413, "dur": 16,
    "args": {
      "External id": 106352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106352, "pid": 5, "tid": 7, "ts": 1716454223039413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979636, "dur": 8,
    "args": {
      "External id": 106352,
      "cbid": 211, "correlation": 106352
    }
  },
  {
    "ph": "s", "id": 106352, "pid": 76337, "tid": -914061504, "ts": 1716454222979636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223039430, "dur": 13,
    "args": {
      "External id": 106360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106360,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106360, "pid": 5, "tid": 7, "ts": 1716454223039430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979665, "dur": 8,
    "args": {
      "External id": 106360,
      "cbid": 211, "correlation": 106360
    }
  },
  {
    "ph": "s", "id": 106360, "pid": 76337, "tid": -914061504, "ts": 1716454222979665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223039445, "dur": 10,
    "args": {
      "External id": 106368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106368,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106368, "pid": 5, "tid": 7, "ts": 1716454223039445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979693, "dur": 8,
    "args": {
      "External id": 106368,
      "cbid": 211, "correlation": 106368
    }
  },
  {
    "ph": "s", "id": 106368, "pid": 76337, "tid": -914061504, "ts": 1716454222979693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223039456, "dur": 18,
    "args": {
      "External id": 106388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106388,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 106388, "pid": 5, "tid": 7, "ts": 1716454223039456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979775, "dur": 12,
    "args": {
      "External id": 106388,
      "cbid": 211, "correlation": 106388
    }
  },
  {
    "ph": "s", "id": 106388, "pid": 76337, "tid": -914061504, "ts": 1716454222979775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223039475, "dur": 5,
    "args": {
      "External id": 106400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106400,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 106400, "pid": 5, "tid": 7, "ts": 1716454223039475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979796, "dur": 6,
    "args": {
      "External id": 106400,
      "cbid": 211, "correlation": 106400
    }
  },
  {
    "ph": "s", "id": 106400, "pid": 76337, "tid": -914061504, "ts": 1716454222979796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223039481, "dur": 17,
    "args": {
      "External id": 106403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106403,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106403, "pid": 5, "tid": 7, "ts": 1716454223039481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979815, "dur": 6,
    "args": {
      "External id": 106403,
      "cbid": 211, "correlation": 106403
    }
  },
  {
    "ph": "s", "id": 106403, "pid": 76337, "tid": -914061504, "ts": 1716454222979815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979873, "dur": 0,
    "args": {
      "External id": 106414,
      "cbid": 317, "correlation": 106414
    }
  },
  {
    "ph": "f", "id": 106414, "pid": 76337, "tid": -914061504, "ts": 1716454222979873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979874, "dur": 0,
    "args": {
      "External id": 106415,
      "cbid": 203, "correlation": 106415
    }
  },
  {
    "ph": "f", "id": 106415, "pid": 76337, "tid": -914061504, "ts": 1716454222979874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979875, "dur": 0,
    "args": {
      "External id": 106416,
      "cbid": 205, "correlation": 106416
    }
  },
  {
    "ph": "f", "id": 106416, "pid": 76337, "tid": -914061504, "ts": 1716454222979875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223039499, "dur": 11,
    "args": {
      "External id": 106420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106420, "pid": 5, "tid": 7, "ts": 1716454223039499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979887, "dur": 12,
    "args": {
      "External id": 106420,
      "cbid": 211, "correlation": 106420
    }
  },
  {
    "ph": "s", "id": 106420, "pid": 76337, "tid": -914061504, "ts": 1716454222979887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223039511, "dur": 4,
    "args": {
      "External id": 106422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106422, "pid": 5, "tid": 7, "ts": 1716454223039511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979903, "dur": 7,
    "args": {
      "External id": 106422,
      "cbid": 211, "correlation": 106422
    }
  },
  {
    "ph": "s", "id": 106422, "pid": 76337, "tid": -914061504, "ts": 1716454222979903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979913, "dur": 0,
    "args": {
      "External id": 106423,
      "cbid": 51, "correlation": 106423
    }
  },
  {
    "ph": "s", "id": 106423, "pid": 76337, "tid": -914061504, "ts": 1716454222979913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223039516, "dur": 89,
    "args": {
      "External id": 106424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106424,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 106424, "pid": 5, "tid": 7, "ts": 1716454223039516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979914, "dur": 5,
    "args": {
      "External id": 106424,
      "cbid": 211, "correlation": 106424
    }
  },
  {
    "ph": "s", "id": 106424, "pid": 76337, "tid": -914061504, "ts": 1716454222979914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223039606, "dur": 16,
    "args": {
      "External id": 106429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106429, "pid": 5, "tid": 7, "ts": 1716454223039606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222979940, "dur": 8,
    "args": {
      "External id": 106429,
      "cbid": 211, "correlation": 106429
    }
  },
  {
    "ph": "s", "id": 106429, "pid": 76337, "tid": -914061504, "ts": 1716454222979940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223039623, "dur": 83,
    "args": {
      "External id": 106438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106438,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106438, "pid": 5, "tid": 7, "ts": 1716454223039623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980030, "dur": 14,
    "args": {
      "External id": 106438,
      "cbid": 211, "correlation": 106438
    }
  },
  {
    "ph": "s", "id": 106438, "pid": 76337, "tid": -914061504, "ts": 1716454222980030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223039708, "dur": 29,
    "args": {
      "External id": 106460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106460,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106460, "pid": 5, "tid": 7, "ts": 1716454223039708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980087, "dur": 10,
    "args": {
      "External id": 106460,
      "cbid": 211, "correlation": 106460
    }
  },
  {
    "ph": "s", "id": 106460, "pid": 76337, "tid": -914061504, "ts": 1716454222980087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980175, "dur": 1,
    "args": {
      "External id": 106471,
      "cbid": 251, "correlation": 106471
    }
  },
  {
    "ph": "f", "id": 106471, "pid": 76337, "tid": -914061504, "ts": 1716454222980175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223039738, "dur": 163,
    "args": {
      "External id": 106472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106472,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106472, "pid": 5, "tid": 7, "ts": 1716454223039738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980180, "dur": 13,
    "args": {
      "External id": 106472,
      "cbid": 211, "correlation": 106472
    }
  },
  {
    "ph": "s", "id": 106472, "pid": 76337, "tid": -914061504, "ts": 1716454222980180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980252, "dur": 1,
    "args": {
      "External id": 106483,
      "cbid": 251, "correlation": 106483
    }
  },
  {
    "ph": "f", "id": 106483, "pid": 76337, "tid": -914061504, "ts": 1716454222980252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223039901, "dur": 156,
    "args": {
      "External id": 106484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106484,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106484, "pid": 5, "tid": 7, "ts": 1716454223039901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980256, "dur": 12,
    "args": {
      "External id": 106484,
      "cbid": 211, "correlation": 106484
    }
  },
  {
    "ph": "s", "id": 106484, "pid": 76337, "tid": -914061504, "ts": 1716454222980256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980322, "dur": 1,
    "args": {
      "External id": 106495,
      "cbid": 251, "correlation": 106495
    }
  },
  {
    "ph": "f", "id": 106495, "pid": 76337, "tid": -914061504, "ts": 1716454222980322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223040059, "dur": 154,
    "args": {
      "External id": 106496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106496,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106496, "pid": 5, "tid": 7, "ts": 1716454223040059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980326, "dur": 11,
    "args": {
      "External id": 106496,
      "cbid": 211, "correlation": 106496
    }
  },
  {
    "ph": "s", "id": 106496, "pid": 76337, "tid": -914061504, "ts": 1716454222980326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223040215, "dur": 328,
    "args": {
      "External id": 106521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106521,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106521, "pid": 5, "tid": 7, "ts": 1716454223040215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980410, "dur": 13,
    "args": {
      "External id": 106521,
      "cbid": 211, "correlation": 106521
    }
  },
  {
    "ph": "s", "id": 106521, "pid": 76337, "tid": -914061504, "ts": 1716454222980410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980509, "dur": 1,
    "args": {
      "External id": 106539,
      "cbid": 251, "correlation": 106539
    }
  },
  {
    "ph": "f", "id": 106539, "pid": 76337, "tid": -914061504, "ts": 1716454222980509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223040544, "dur": 141,
    "args": {
      "External id": 106541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106541,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106541, "pid": 5, "tid": 7, "ts": 1716454223040544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980515, "dur": 13,
    "args": {
      "External id": 106541,
      "cbid": 211, "correlation": 106541
    }
  },
  {
    "ph": "s", "id": 106541, "pid": 76337, "tid": -914061504, "ts": 1716454222980515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223040687, "dur": 19,
    "args": {
      "External id": 106549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106549,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106549, "pid": 5, "tid": 7, "ts": 1716454223040687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980585, "dur": 12,
    "args": {
      "External id": 106549,
      "cbid": 211, "correlation": 106549
    }
  },
  {
    "ph": "s", "id": 106549, "pid": 76337, "tid": -914061504, "ts": 1716454222980585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223040707, "dur": 28,
    "args": {
      "External id": 106557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106557,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106557, "pid": 5, "tid": 7, "ts": 1716454223040707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980624, "dur": 8,
    "args": {
      "External id": 106557,
      "cbid": 211, "correlation": 106557
    }
  },
  {
    "ph": "s", "id": 106557, "pid": 76337, "tid": -914061504, "ts": 1716454222980624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223040736, "dur": 18,
    "args": {
      "External id": 106568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106568, "pid": 5, "tid": 7, "ts": 1716454223040736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980696, "dur": 13,
    "args": {
      "External id": 106568,
      "cbid": 211, "correlation": 106568
    }
  },
  {
    "ph": "s", "id": 106568, "pid": 76337, "tid": -914061504, "ts": 1716454222980696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223040756, "dur": 16,
    "args": {
      "External id": 106590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106590,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106590, "pid": 5, "tid": 7, "ts": 1716454223040756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980727, "dur": 8,
    "args": {
      "External id": 106590,
      "cbid": 211, "correlation": 106590
    }
  },
  {
    "ph": "s", "id": 106590, "pid": 76337, "tid": -914061504, "ts": 1716454222980727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980812, "dur": 1,
    "args": {
      "External id": 106601,
      "cbid": 251, "correlation": 106601
    }
  },
  {
    "ph": "f", "id": 106601, "pid": 76337, "tid": -914061504, "ts": 1716454222980812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223040773, "dur": 87,
    "args": {
      "External id": 106602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106602,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106602, "pid": 5, "tid": 7, "ts": 1716454223040773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980818, "dur": 13,
    "args": {
      "External id": 106602,
      "cbid": 211, "correlation": 106602
    }
  },
  {
    "ph": "s", "id": 106602, "pid": 76337, "tid": -914061504, "ts": 1716454222980818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980888, "dur": 1,
    "args": {
      "External id": 106613,
      "cbid": 251, "correlation": 106613
    }
  },
  {
    "ph": "f", "id": 106613, "pid": 76337, "tid": -914061504, "ts": 1716454222980888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980892, "dur": 0,
    "args": {
      "External id": 106614,
      "cbid": 251, "correlation": 106614
    }
  },
  {
    "ph": "f", "id": 106614, "pid": 76337, "tid": -914061504, "ts": 1716454222980892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223040862, "dur": 12,
    "args": {
      "External id": 106615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106615,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106615, "pid": 5, "tid": 7, "ts": 1716454223040862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980893, "dur": 11,
    "args": {
      "External id": 106615,
      "cbid": 211, "correlation": 106615
    }
  },
  {
    "ph": "s", "id": 106615, "pid": 76337, "tid": -914061504, "ts": 1716454222980893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223040876, "dur": 5,
    "args": {
      "External id": 106617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106617,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106617, "pid": 5, "tid": 7, "ts": 1716454223040876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980906, "dur": 5,
    "args": {
      "External id": 106617,
      "cbid": 211, "correlation": 106617
    }
  },
  {
    "ph": "s", "id": 106617, "pid": 76337, "tid": -914061504, "ts": 1716454222980906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980964, "dur": 1,
    "args": {
      "External id": 106628,
      "cbid": 251, "correlation": 106628
    }
  },
  {
    "ph": "f", "id": 106628, "pid": 76337, "tid": -914061504, "ts": 1716454222980964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980967, "dur": 0,
    "args": {
      "External id": 106629,
      "cbid": 251, "correlation": 106629
    }
  },
  {
    "ph": "f", "id": 106629, "pid": 76337, "tid": -914061504, "ts": 1716454222980967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223040882, "dur": 8,
    "args": {
      "External id": 106630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106630,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106630, "pid": 5, "tid": 7, "ts": 1716454223040882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980969, "dur": 20,
    "args": {
      "External id": 106630,
      "cbid": 211, "correlation": 106630
    }
  },
  {
    "ph": "s", "id": 106630, "pid": 76337, "tid": -914061504, "ts": 1716454222980969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223040892, "dur": 3,
    "args": {
      "External id": 106632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106632,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106632, "pid": 5, "tid": 7, "ts": 1716454223040892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222980990, "dur": 6,
    "args": {
      "External id": 106632,
      "cbid": 211, "correlation": 106632
    }
  },
  {
    "ph": "s", "id": 106632, "pid": 76337, "tid": -914061504, "ts": 1716454222980990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223040896, "dur": 53,
    "args": {
      "External id": 106657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106657,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106657, "pid": 5, "tid": 7, "ts": 1716454223040896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981069, "dur": 12,
    "args": {
      "External id": 106657,
      "cbid": 211, "correlation": 106657
    }
  },
  {
    "ph": "s", "id": 106657, "pid": 76337, "tid": -914061504, "ts": 1716454222981069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981168, "dur": 1,
    "args": {
      "External id": 106675,
      "cbid": 251, "correlation": 106675
    }
  },
  {
    "ph": "f", "id": 106675, "pid": 76337, "tid": -914061504, "ts": 1716454222981168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223040951, "dur": 88,
    "args": {
      "External id": 106677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106677,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106677, "pid": 5, "tid": 7, "ts": 1716454223040951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981174, "dur": 13,
    "args": {
      "External id": 106677,
      "cbid": 211, "correlation": 106677
    }
  },
  {
    "ph": "s", "id": 106677, "pid": 76337, "tid": -914061504, "ts": 1716454222981174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223041041, "dur": 9,
    "args": {
      "External id": 106685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106685,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106685, "pid": 5, "tid": 7, "ts": 1716454223041041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981243, "dur": 12,
    "args": {
      "External id": 106685,
      "cbid": 211, "correlation": 106685
    }
  },
  {
    "ph": "s", "id": 106685, "pid": 76337, "tid": -914061504, "ts": 1716454222981243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223041052, "dur": 20,
    "args": {
      "External id": 106693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106693,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106693, "pid": 5, "tid": 7, "ts": 1716454223041052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981284, "dur": 9,
    "args": {
      "External id": 106693,
      "cbid": 211, "correlation": 106693
    }
  },
  {
    "ph": "s", "id": 106693, "pid": 76337, "tid": -914061504, "ts": 1716454222981284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223041073, "dur": 18,
    "args": {
      "External id": 106715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106715,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106715, "pid": 5, "tid": 7, "ts": 1716454223041073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981336, "dur": 10,
    "args": {
      "External id": 106715,
      "cbid": 211, "correlation": 106715
    }
  },
  {
    "ph": "s", "id": 106715, "pid": 76337, "tid": -914061504, "ts": 1716454222981336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981423, "dur": 1,
    "args": {
      "External id": 106731,
      "cbid": 251, "correlation": 106731
    }
  },
  {
    "ph": "f", "id": 106731, "pid": 76337, "tid": -914061504, "ts": 1716454222981423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981428, "dur": 0,
    "args": {
      "External id": 106733,
      "cbid": 251, "correlation": 106733
    }
  },
  {
    "ph": "f", "id": 106733, "pid": 76337, "tid": -914061504, "ts": 1716454222981428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223041092, "dur": 493,
    "args": {
      "External id": 106734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106734,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106734, "pid": 5, "tid": 7, "ts": 1716454223041092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981429, "dur": 13,
    "args": {
      "External id": 106734,
      "cbid": 211, "correlation": 106734
    }
  },
  {
    "ph": "s", "id": 106734, "pid": 76337, "tid": -914061504, "ts": 1716454222981429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223041587, "dur": 66,
    "args": {
      "External id": 106742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106742, "pid": 5, "tid": 7, "ts": 1716454223041587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981495, "dur": 12,
    "args": {
      "External id": 106742,
      "cbid": 211, "correlation": 106742
    }
  },
  {
    "ph": "s", "id": 106742, "pid": 76337, "tid": -914061504, "ts": 1716454222981495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223041654, "dur": 66,
    "args": {
      "External id": 106750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106750, "pid": 5, "tid": 7, "ts": 1716454223041654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981526, "dur": 8,
    "args": {
      "External id": 106750,
      "cbid": 211, "correlation": 106750
    }
  },
  {
    "ph": "s", "id": 106750, "pid": 76337, "tid": -914061504, "ts": 1716454222981526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981605, "dur": 1,
    "args": {
      "External id": 106766,
      "cbid": 251, "correlation": 106766
    }
  },
  {
    "ph": "f", "id": 106766, "pid": 76337, "tid": -914061504, "ts": 1716454222981605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223041722, "dur": 1,
    "args": {
      "External id": 106768,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 106768,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 106768, "pid": 5, "tid": 7, "ts": 1716454223041722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981610, "dur": 9,
    "args": {
      "External id": 106768,
      "cbid": 51, "correlation": 106768
    }
  },
  {
    "ph": "s", "id": 106768, "pid": 76337, "tid": -914061504, "ts": 1716454222981610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223041726, "dur": 264,
    "args": {
      "External id": 106769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106769,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106769, "pid": 5, "tid": 7, "ts": 1716454223041726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981622, "dur": 11,
    "args": {
      "External id": 106769,
      "cbid": 211, "correlation": 106769
    }
  },
  {
    "ph": "s", "id": 106769, "pid": 76337, "tid": -914061504, "ts": 1716454222981622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223041991, "dur": 14,
    "args": {
      "External id": 106777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106777,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106777, "pid": 5, "tid": 7, "ts": 1716454223041991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981663, "dur": 11,
    "args": {
      "External id": 106777,
      "cbid": 211, "correlation": 106777
    }
  },
  {
    "ph": "s", "id": 106777, "pid": 76337, "tid": -914061504, "ts": 1716454222981663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223042006, "dur": 37,
    "args": {
      "External id": 106788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106788, "pid": 5, "tid": 7, "ts": 1716454223042006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981732, "dur": 12,
    "args": {
      "External id": 106788,
      "cbid": 211, "correlation": 106788
    }
  },
  {
    "ph": "s", "id": 106788, "pid": 76337, "tid": -914061504, "ts": 1716454222981732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981796, "dur": 0,
    "args": {
      "External id": 106800,
      "cbid": 317, "correlation": 106800
    }
  },
  {
    "ph": "f", "id": 106800, "pid": 76337, "tid": -914061504, "ts": 1716454222981796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981797, "dur": 0,
    "args": {
      "External id": 106801,
      "cbid": 203, "correlation": 106801
    }
  },
  {
    "ph": "f", "id": 106801, "pid": 76337, "tid": -914061504, "ts": 1716454222981797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981798, "dur": 0,
    "args": {
      "External id": 106802,
      "cbid": 205, "correlation": 106802
    }
  },
  {
    "ph": "f", "id": 106802, "pid": 76337, "tid": -914061504, "ts": 1716454222981798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223042045, "dur": 13,
    "args": {
      "External id": 106806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106806,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106806, "pid": 5, "tid": 7, "ts": 1716454223042045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981812, "dur": 13,
    "args": {
      "External id": 106806,
      "cbid": 211, "correlation": 106806
    }
  },
  {
    "ph": "s", "id": 106806, "pid": 76337, "tid": -914061504, "ts": 1716454222981812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223042060, "dur": 4,
    "args": {
      "External id": 106808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 106808, "pid": 5, "tid": 7, "ts": 1716454223042060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981829, "dur": 6,
    "args": {
      "External id": 106808,
      "cbid": 211, "correlation": 106808
    }
  },
  {
    "ph": "s", "id": 106808, "pid": 76337, "tid": -914061504, "ts": 1716454222981829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981838, "dur": 0,
    "args": {
      "External id": 106809,
      "cbid": 51, "correlation": 106809
    }
  },
  {
    "ph": "s", "id": 106809, "pid": 76337, "tid": -914061504, "ts": 1716454222981838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223042065, "dur": 95,
    "args": {
      "External id": 106810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106810,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 106810, "pid": 5, "tid": 7, "ts": 1716454223042065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981838, "dur": 5,
    "args": {
      "External id": 106810,
      "cbid": 211, "correlation": 106810
    }
  },
  {
    "ph": "s", "id": 106810, "pid": 76337, "tid": -914061504, "ts": 1716454222981838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223042161, "dur": 16,
    "args": {
      "External id": 106815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106815, "pid": 5, "tid": 7, "ts": 1716454223042161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981865, "dur": 9,
    "args": {
      "External id": 106815,
      "cbid": 211, "correlation": 106815
    }
  },
  {
    "ph": "s", "id": 106815, "pid": 76337, "tid": -914061504, "ts": 1716454222981865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223042179, "dur": 12,
    "args": {
      "External id": 106823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106823,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106823, "pid": 5, "tid": 7, "ts": 1716454223042179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981897, "dur": 8,
    "args": {
      "External id": 106823,
      "cbid": 211, "correlation": 106823
    }
  },
  {
    "ph": "s", "id": 106823, "pid": 76337, "tid": -914061504, "ts": 1716454222981897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223042192, "dur": 26,
    "args": {
      "External id": 106832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106832,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106832, "pid": 5, "tid": 7, "ts": 1716454223042192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222981936, "dur": 10,
    "args": {
      "External id": 106832,
      "cbid": 211, "correlation": 106832
    }
  },
  {
    "ph": "s", "id": 106832, "pid": 76337, "tid": -914061504, "ts": 1716454222981936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223042219, "dur": 24,
    "args": {
      "External id": 106852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106852,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 106852, "pid": 5, "tid": 7, "ts": 1716454223042219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982023, "dur": 13,
    "args": {
      "External id": 106852,
      "cbid": 211, "correlation": 106852
    }
  },
  {
    "ph": "s", "id": 106852, "pid": 76337, "tid": -914061504, "ts": 1716454222982023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223042244, "dur": 5,
    "args": {
      "External id": 106864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106864,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 106864, "pid": 5, "tid": 7, "ts": 1716454223042244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982047, "dur": 6,
    "args": {
      "External id": 106864,
      "cbid": 211, "correlation": 106864
    }
  },
  {
    "ph": "s", "id": 106864, "pid": 76337, "tid": -914061504, "ts": 1716454222982047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223042250, "dur": 24,
    "args": {
      "External id": 106867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106867, "pid": 5, "tid": 7, "ts": 1716454223042250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982065, "dur": 7,
    "args": {
      "External id": 106867,
      "cbid": 211, "correlation": 106867
    }
  },
  {
    "ph": "s", "id": 106867, "pid": 76337, "tid": -914061504, "ts": 1716454222982065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223042276, "dur": 17,
    "args": {
      "External id": 106876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106876,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106876, "pid": 5, "tid": 7, "ts": 1716454223042276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982106, "dur": 10,
    "args": {
      "External id": 106876,
      "cbid": 211, "correlation": 106876
    }
  },
  {
    "ph": "s", "id": 106876, "pid": 76337, "tid": -914061504, "ts": 1716454222982106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982158, "dur": 0,
    "args": {
      "External id": 106886,
      "cbid": 317, "correlation": 106886
    }
  },
  {
    "ph": "f", "id": 106886, "pid": 76337, "tid": -914061504, "ts": 1716454222982158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982159, "dur": 0,
    "args": {
      "External id": 106887,
      "cbid": 203, "correlation": 106887
    }
  },
  {
    "ph": "f", "id": 106887, "pid": 76337, "tid": -914061504, "ts": 1716454222982159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982160, "dur": 0,
    "args": {
      "External id": 106888,
      "cbid": 205, "correlation": 106888
    }
  },
  {
    "ph": "f", "id": 106888, "pid": 76337, "tid": -914061504, "ts": 1716454222982160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223042294, "dur": 17,
    "args": {
      "External id": 106892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106892,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106892, "pid": 5, "tid": 7, "ts": 1716454223042294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982175, "dur": 12,
    "args": {
      "External id": 106892,
      "cbid": 211, "correlation": 106892
    }
  },
  {
    "ph": "s", "id": 106892, "pid": 76337, "tid": -914061504, "ts": 1716454222982175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223042312, "dur": 236,
    "args": {
      "External id": 106894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106894,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106894, "pid": 5, "tid": 7, "ts": 1716454223042312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982189, "dur": 5,
    "args": {
      "External id": 106894,
      "cbid": 211, "correlation": 106894
    }
  },
  {
    "ph": "s", "id": 106894, "pid": 76337, "tid": -914061504, "ts": 1716454222982189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223042551, "dur": 1,
    "args": {
      "External id": 106896,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 106896,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 106896, "pid": 5, "tid": 7, "ts": 1716454223042551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982200, "dur": 8,
    "args": {
      "External id": 106896,
      "cbid": 51, "correlation": 106896
    }
  },
  {
    "ph": "s", "id": 106896, "pid": 76337, "tid": -914061504, "ts": 1716454222982200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223042554, "dur": 807,
    "args": {
      "External id": 106897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106897,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106897, "pid": 5, "tid": 7, "ts": 1716454223042554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982209, "dur": 6,
    "args": {
      "External id": 106897,
      "cbid": 211, "correlation": 106897
    }
  },
  {
    "ph": "s", "id": 106897, "pid": 76337, "tid": -914061504, "ts": 1716454222982209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223043362, "dur": 13,
    "args": {
      "External id": 106899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106899,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106899, "pid": 5, "tid": 7, "ts": 1716454223043362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982220, "dur": 6,
    "args": {
      "External id": 106899,
      "cbid": 211, "correlation": 106899
    }
  },
  {
    "ph": "s", "id": 106899, "pid": 76337, "tid": -914061504, "ts": 1716454222982220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223043377, "dur": 15,
    "args": {
      "External id": 106905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106905,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106905, "pid": 5, "tid": 7, "ts": 1716454223043377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982248, "dur": 8,
    "args": {
      "External id": 106905,
      "cbid": 211, "correlation": 106905
    }
  },
  {
    "ph": "s", "id": 106905, "pid": 76337, "tid": -914061504, "ts": 1716454222982248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223043393, "dur": 3,
    "args": {
      "External id": 106913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106913,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 106913, "pid": 5, "tid": 7, "ts": 1716454223043393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982292, "dur": 9,
    "args": {
      "External id": 106913,
      "cbid": 211, "correlation": 106913
    }
  },
  {
    "ph": "s", "id": 106913, "pid": 76337, "tid": -914061504, "ts": 1716454222982292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982357, "dur": 1,
    "args": {
      "External id": 106929,
      "cbid": 251, "correlation": 106929
    }
  },
  {
    "ph": "f", "id": 106929, "pid": 76337, "tid": -914061504, "ts": 1716454222982357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982362, "dur": 0,
    "args": {
      "External id": 106931,
      "cbid": 251, "correlation": 106931
    }
  },
  {
    "ph": "f", "id": 106931, "pid": 76337, "tid": -914061504, "ts": 1716454222982362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223043398, "dur": 14,
    "args": {
      "External id": 106932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106932,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106932, "pid": 5, "tid": 7, "ts": 1716454223043398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982364, "dur": 12,
    "args": {
      "External id": 106932,
      "cbid": 211, "correlation": 106932
    }
  },
  {
    "ph": "s", "id": 106932, "pid": 76337, "tid": -914061504, "ts": 1716454222982364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223043412, "dur": 5,
    "args": {
      "External id": 106934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106934,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 106934, "pid": 5, "tid": 7, "ts": 1716454223043412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982378, "dur": 6,
    "args": {
      "External id": 106934,
      "cbid": 211, "correlation": 106934
    }
  },
  {
    "ph": "s", "id": 106934, "pid": 76337, "tid": -914061504, "ts": 1716454222982378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223043419, "dur": 16,
    "args": {
      "External id": 106944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106944,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106944, "pid": 5, "tid": 7, "ts": 1716454223043419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982436, "dur": 12,
    "args": {
      "External id": 106944,
      "cbid": 211, "correlation": 106944
    }
  },
  {
    "ph": "s", "id": 106944, "pid": 76337, "tid": -914061504, "ts": 1716454222982436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223043436, "dur": 17,
    "args": {
      "External id": 106964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106964,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 106964, "pid": 5, "tid": 7, "ts": 1716454223043436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982502, "dur": 11,
    "args": {
      "External id": 106964,
      "cbid": 211, "correlation": 106964
    }
  },
  {
    "ph": "s", "id": 106964, "pid": 76337, "tid": -914061504, "ts": 1716454222982502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223043455, "dur": 4,
    "args": {
      "External id": 106976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106976,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 106976, "pid": 5, "tid": 7, "ts": 1716454223043455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982523, "dur": 7,
    "args": {
      "External id": 106976,
      "cbid": 211, "correlation": 106976
    }
  },
  {
    "ph": "s", "id": 106976, "pid": 76337, "tid": -914061504, "ts": 1716454222982523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223043460, "dur": 16,
    "args": {
      "External id": 106979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106979, "pid": 5, "tid": 7, "ts": 1716454223043460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982543, "dur": 6,
    "args": {
      "External id": 106979,
      "cbid": 211, "correlation": 106979
    }
  },
  {
    "ph": "s", "id": 106979, "pid": 76337, "tid": -914061504, "ts": 1716454222982543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223043478, "dur": 11,
    "args": {
      "External id": 106988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 106988,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 106988, "pid": 5, "tid": 7, "ts": 1716454223043478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982584, "dur": 10,
    "args": {
      "External id": 106988,
      "cbid": 211, "correlation": 106988
    }
  },
  {
    "ph": "s", "id": 106988, "pid": 76337, "tid": -914061504, "ts": 1716454222982584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982646, "dur": 0,
    "args": {
      "External id": 106998,
      "cbid": 317, "correlation": 106998
    }
  },
  {
    "ph": "f", "id": 106998, "pid": 76337, "tid": -914061504, "ts": 1716454222982646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982647, "dur": 0,
    "args": {
      "External id": 106999,
      "cbid": 203, "correlation": 106999
    }
  },
  {
    "ph": "f", "id": 106999, "pid": 76337, "tid": -914061504, "ts": 1716454222982647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982648, "dur": 0,
    "args": {
      "External id": 107000,
      "cbid": 205, "correlation": 107000
    }
  },
  {
    "ph": "f", "id": 107000, "pid": 76337, "tid": -914061504, "ts": 1716454222982648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223043490, "dur": 11,
    "args": {
      "External id": 107004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107004,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107004, "pid": 5, "tid": 7, "ts": 1716454223043490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982661, "dur": 12,
    "args": {
      "External id": 107004,
      "cbid": 211, "correlation": 107004
    }
  },
  {
    "ph": "s", "id": 107004, "pid": 76337, "tid": -914061504, "ts": 1716454222982661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223043502, "dur": 160,
    "args": {
      "External id": 107006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107006,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107006, "pid": 5, "tid": 7, "ts": 1716454223043502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982676, "dur": 6,
    "args": {
      "External id": 107006,
      "cbid": 211, "correlation": 107006
    }
  },
  {
    "ph": "s", "id": 107006, "pid": 76337, "tid": -914061504, "ts": 1716454222982676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223043664, "dur": 1,
    "args": {
      "External id": 107008,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 107008,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 107008, "pid": 5, "tid": 7, "ts": 1716454223043664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982687, "dur": 6,
    "args": {
      "External id": 107008,
      "cbid": 51, "correlation": 107008
    }
  },
  {
    "ph": "s", "id": 107008, "pid": 76337, "tid": -914061504, "ts": 1716454222982687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223043668, "dur": 639,
    "args": {
      "External id": 107009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107009,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107009, "pid": 5, "tid": 7, "ts": 1716454223043668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982695, "dur": 6,
    "args": {
      "External id": 107009,
      "cbid": 211, "correlation": 107009
    }
  },
  {
    "ph": "s", "id": 107009, "pid": 76337, "tid": -914061504, "ts": 1716454222982695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223044308, "dur": 12,
    "args": {
      "External id": 107011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107011,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107011, "pid": 5, "tid": 7, "ts": 1716454223044308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982705, "dur": 5,
    "args": {
      "External id": 107011,
      "cbid": 211, "correlation": 107011
    }
  },
  {
    "ph": "s", "id": 107011, "pid": 76337, "tid": -914061504, "ts": 1716454222982705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223044322, "dur": 14,
    "args": {
      "External id": 107017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107017, "pid": 5, "tid": 7, "ts": 1716454223044322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982734, "dur": 9,
    "args": {
      "External id": 107017,
      "cbid": 211, "correlation": 107017
    }
  },
  {
    "ph": "s", "id": 107017, "pid": 76337, "tid": -914061504, "ts": 1716454222982734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982793, "dur": 0,
    "args": {
      "External id": 107027,
      "cbid": 317, "correlation": 107027
    }
  },
  {
    "ph": "f", "id": 107027, "pid": 76337, "tid": -914061504, "ts": 1716454222982793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982793, "dur": 0,
    "args": {
      "External id": 107028,
      "cbid": 203, "correlation": 107028
    }
  },
  {
    "ph": "f", "id": 107028, "pid": 76337, "tid": -914061504, "ts": 1716454222982793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982794, "dur": 0,
    "args": {
      "External id": 107029,
      "cbid": 205, "correlation": 107029
    }
  },
  {
    "ph": "f", "id": 107029, "pid": 76337, "tid": -914061504, "ts": 1716454222982794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223044337, "dur": 17,
    "args": {
      "External id": 107033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107033,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107033, "pid": 5, "tid": 7, "ts": 1716454223044337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982807, "dur": 12,
    "args": {
      "External id": 107033,
      "cbid": 211, "correlation": 107033
    }
  },
  {
    "ph": "s", "id": 107033, "pid": 76337, "tid": -914061504, "ts": 1716454222982807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223044355, "dur": 4,
    "args": {
      "External id": 107035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107035,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107035, "pid": 5, "tid": 7, "ts": 1716454223044355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982824, "dur": 7,
    "args": {
      "External id": 107035,
      "cbid": 211, "correlation": 107035
    }
  },
  {
    "ph": "s", "id": 107035, "pid": 76337, "tid": -914061504, "ts": 1716454222982824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982834, "dur": 0,
    "args": {
      "External id": 107036,
      "cbid": 51, "correlation": 107036
    }
  },
  {
    "ph": "s", "id": 107036, "pid": 76337, "tid": -914061504, "ts": 1716454222982834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223044361, "dur": 130,
    "args": {
      "External id": 107037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107037,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 107037, "pid": 5, "tid": 7, "ts": 1716454223044361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982834, "dur": 5,
    "args": {
      "External id": 107037,
      "cbid": 211, "correlation": 107037
    }
  },
  {
    "ph": "s", "id": 107037, "pid": 76337, "tid": -914061504, "ts": 1716454222982834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223044492, "dur": 15,
    "args": {
      "External id": 107042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107042,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107042, "pid": 5, "tid": 7, "ts": 1716454223044492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982860, "dur": 8,
    "args": {
      "External id": 107042,
      "cbid": 211, "correlation": 107042
    }
  },
  {
    "ph": "s", "id": 107042, "pid": 76337, "tid": -914061504, "ts": 1716454222982860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223044509, "dur": 12,
    "args": {
      "External id": 107050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107050,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107050, "pid": 5, "tid": 7, "ts": 1716454223044509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982889, "dur": 8,
    "args": {
      "External id": 107050,
      "cbid": 211, "correlation": 107050
    }
  },
  {
    "ph": "s", "id": 107050, "pid": 76337, "tid": -914061504, "ts": 1716454222982889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223044521, "dur": 10,
    "args": {
      "External id": 107058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107058,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107058, "pid": 5, "tid": 7, "ts": 1716454223044521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222982917, "dur": 9,
    "args": {
      "External id": 107058,
      "cbid": 211, "correlation": 107058
    }
  },
  {
    "ph": "s", "id": 107058, "pid": 76337, "tid": -914061504, "ts": 1716454222982917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223044533, "dur": 19,
    "args": {
      "External id": 107078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107078,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 107078, "pid": 5, "tid": 7, "ts": 1716454223044533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983008, "dur": 13,
    "args": {
      "External id": 107078,
      "cbid": 211, "correlation": 107078
    }
  },
  {
    "ph": "s", "id": 107078, "pid": 76337, "tid": -914061504, "ts": 1716454222983008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223044553, "dur": 4,
    "args": {
      "External id": 107090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107090,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 107090, "pid": 5, "tid": 7, "ts": 1716454223044553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983032, "dur": 6,
    "args": {
      "External id": 107090,
      "cbid": 211, "correlation": 107090
    }
  },
  {
    "ph": "s", "id": 107090, "pid": 76337, "tid": -914061504, "ts": 1716454222983032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223044558, "dur": 17,
    "args": {
      "External id": 107093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107093, "pid": 5, "tid": 7, "ts": 1716454223044558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983049, "dur": 7,
    "args": {
      "External id": 107093,
      "cbid": 211, "correlation": 107093
    }
  },
  {
    "ph": "s", "id": 107093, "pid": 76337, "tid": -914061504, "ts": 1716454222983049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983107, "dur": 0,
    "args": {
      "External id": 107104,
      "cbid": 317, "correlation": 107104
    }
  },
  {
    "ph": "f", "id": 107104, "pid": 76337, "tid": -914061504, "ts": 1716454222983107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983108, "dur": 0,
    "args": {
      "External id": 107105,
      "cbid": 203, "correlation": 107105
    }
  },
  {
    "ph": "f", "id": 107105, "pid": 76337, "tid": -914061504, "ts": 1716454222983108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983108, "dur": 0,
    "args": {
      "External id": 107106,
      "cbid": 205, "correlation": 107106
    }
  },
  {
    "ph": "f", "id": 107106, "pid": 76337, "tid": -914061504, "ts": 1716454222983108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223044576, "dur": 11,
    "args": {
      "External id": 107110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107110, "pid": 5, "tid": 7, "ts": 1716454223044576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983121, "dur": 13,
    "args": {
      "External id": 107110,
      "cbid": 211, "correlation": 107110
    }
  },
  {
    "ph": "s", "id": 107110, "pid": 76337, "tid": -914061504, "ts": 1716454222983121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223044588, "dur": 3,
    "args": {
      "External id": 107112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107112, "pid": 5, "tid": 7, "ts": 1716454223044588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983138, "dur": 5,
    "args": {
      "External id": 107112,
      "cbid": 211, "correlation": 107112
    }
  },
  {
    "ph": "s", "id": 107112, "pid": 76337, "tid": -914061504, "ts": 1716454222983138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983146, "dur": 0,
    "args": {
      "External id": 107113,
      "cbid": 51, "correlation": 107113
    }
  },
  {
    "ph": "s", "id": 107113, "pid": 76337, "tid": -914061504, "ts": 1716454222983146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223044592, "dur": 88,
    "args": {
      "External id": 107114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107114,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 107114, "pid": 5, "tid": 7, "ts": 1716454223044592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983147, "dur": 5,
    "args": {
      "External id": 107114,
      "cbid": 211, "correlation": 107114
    }
  },
  {
    "ph": "s", "id": 107114, "pid": 76337, "tid": -914061504, "ts": 1716454222983147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223044682, "dur": 15,
    "args": {
      "External id": 107119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107119, "pid": 5, "tid": 7, "ts": 1716454223044682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983173, "dur": 10,
    "args": {
      "External id": 107119,
      "cbid": 211, "correlation": 107119
    }
  },
  {
    "ph": "s", "id": 107119, "pid": 76337, "tid": -914061504, "ts": 1716454222983173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223044698, "dur": 82,
    "args": {
      "External id": 107128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107128,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107128, "pid": 5, "tid": 7, "ts": 1716454223044698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983254, "dur": 13,
    "args": {
      "External id": 107128,
      "cbid": 211, "correlation": 107128
    }
  },
  {
    "ph": "s", "id": 107128, "pid": 76337, "tid": -914061504, "ts": 1716454222983254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223044782, "dur": 29,
    "args": {
      "External id": 107150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107150,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107150, "pid": 5, "tid": 7, "ts": 1716454223044782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983311, "dur": 10,
    "args": {
      "External id": 107150,
      "cbid": 211, "correlation": 107150
    }
  },
  {
    "ph": "s", "id": 107150, "pid": 76337, "tid": -914061504, "ts": 1716454222983311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983399, "dur": 1,
    "args": {
      "External id": 107161,
      "cbid": 251, "correlation": 107161
    }
  },
  {
    "ph": "f", "id": 107161, "pid": 76337, "tid": -914061504, "ts": 1716454222983399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223044813, "dur": 161,
    "args": {
      "External id": 107162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107162,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107162, "pid": 5, "tid": 7, "ts": 1716454223044813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983405, "dur": 13,
    "args": {
      "External id": 107162,
      "cbid": 211, "correlation": 107162
    }
  },
  {
    "ph": "s", "id": 107162, "pid": 76337, "tid": -914061504, "ts": 1716454222983405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983475, "dur": 1,
    "args": {
      "External id": 107173,
      "cbid": 251, "correlation": 107173
    }
  },
  {
    "ph": "f", "id": 107173, "pid": 76337, "tid": -914061504, "ts": 1716454222983475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223044975, "dur": 155,
    "args": {
      "External id": 107174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107174,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107174, "pid": 5, "tid": 7, "ts": 1716454223044975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983480, "dur": 11,
    "args": {
      "External id": 107174,
      "cbid": 211, "correlation": 107174
    }
  },
  {
    "ph": "s", "id": 107174, "pid": 76337, "tid": -914061504, "ts": 1716454222983480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983544, "dur": 1,
    "args": {
      "External id": 107185,
      "cbid": 251, "correlation": 107185
    }
  },
  {
    "ph": "f", "id": 107185, "pid": 76337, "tid": -914061504, "ts": 1716454222983544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223045131, "dur": 156,
    "args": {
      "External id": 107186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107186,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107186, "pid": 5, "tid": 7, "ts": 1716454223045131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983548, "dur": 11,
    "args": {
      "External id": 107186,
      "cbid": 211, "correlation": 107186
    }
  },
  {
    "ph": "s", "id": 107186, "pid": 76337, "tid": -914061504, "ts": 1716454222983548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223045289, "dur": 329,
    "args": {
      "External id": 107211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107211,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107211, "pid": 5, "tid": 7, "ts": 1716454223045289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983631, "dur": 12,
    "args": {
      "External id": 107211,
      "cbid": 211, "correlation": 107211
    }
  },
  {
    "ph": "s", "id": 107211, "pid": 76337, "tid": -914061504, "ts": 1716454222983631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983730, "dur": 1,
    "args": {
      "External id": 107229,
      "cbid": 251, "correlation": 107229
    }
  },
  {
    "ph": "f", "id": 107229, "pid": 76337, "tid": -914061504, "ts": 1716454222983730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223045620, "dur": 161,
    "args": {
      "External id": 107231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107231,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107231, "pid": 5, "tid": 7, "ts": 1716454223045620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983737, "dur": 13,
    "args": {
      "External id": 107231,
      "cbid": 211, "correlation": 107231
    }
  },
  {
    "ph": "s", "id": 107231, "pid": 76337, "tid": -914061504, "ts": 1716454222983737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223045782, "dur": 19,
    "args": {
      "External id": 107239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107239,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107239, "pid": 5, "tid": 7, "ts": 1716454223045782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983805, "dur": 12,
    "args": {
      "External id": 107239,
      "cbid": 211, "correlation": 107239
    }
  },
  {
    "ph": "s", "id": 107239, "pid": 76337, "tid": -914061504, "ts": 1716454222983805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223045802, "dur": 27,
    "args": {
      "External id": 107247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107247,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107247, "pid": 5, "tid": 7, "ts": 1716454223045802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983844, "dur": 8,
    "args": {
      "External id": 107247,
      "cbid": 211, "correlation": 107247
    }
  },
  {
    "ph": "s", "id": 107247, "pid": 76337, "tid": -914061504, "ts": 1716454222983844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223045831, "dur": 19,
    "args": {
      "External id": 107258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107258,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107258, "pid": 5, "tid": 7, "ts": 1716454223045831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983915, "dur": 12,
    "args": {
      "External id": 107258,
      "cbid": 211, "correlation": 107258
    }
  },
  {
    "ph": "s", "id": 107258, "pid": 76337, "tid": -914061504, "ts": 1716454222983915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223045851, "dur": 16,
    "args": {
      "External id": 107280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107280,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107280, "pid": 5, "tid": 7, "ts": 1716454223045851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222983947, "dur": 8,
    "args": {
      "External id": 107280,
      "cbid": 211, "correlation": 107280
    }
  },
  {
    "ph": "s", "id": 107280, "pid": 76337, "tid": -914061504, "ts": 1716454222983947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984043, "dur": 1,
    "args": {
      "External id": 107291,
      "cbid": 251, "correlation": 107291
    }
  },
  {
    "ph": "f", "id": 107291, "pid": 76337, "tid": -914061504, "ts": 1716454222984043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223045868, "dur": 87,
    "args": {
      "External id": 107292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107292,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107292, "pid": 5, "tid": 7, "ts": 1716454223045868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984048, "dur": 13,
    "args": {
      "External id": 107292,
      "cbid": 211, "correlation": 107292
    }
  },
  {
    "ph": "s", "id": 107292, "pid": 76337, "tid": -914061504, "ts": 1716454222984048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984116, "dur": 1,
    "args": {
      "External id": 107303,
      "cbid": 251, "correlation": 107303
    }
  },
  {
    "ph": "f", "id": 107303, "pid": 76337, "tid": -914061504, "ts": 1716454222984116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984120, "dur": 0,
    "args": {
      "External id": 107304,
      "cbid": 251, "correlation": 107304
    }
  },
  {
    "ph": "f", "id": 107304, "pid": 76337, "tid": -914061504, "ts": 1716454222984120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223045957, "dur": 12,
    "args": {
      "External id": 107305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107305,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107305, "pid": 5, "tid": 7, "ts": 1716454223045957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984122, "dur": 12,
    "args": {
      "External id": 107305,
      "cbid": 211, "correlation": 107305
    }
  },
  {
    "ph": "s", "id": 107305, "pid": 76337, "tid": -914061504, "ts": 1716454222984122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223045970, "dur": 5,
    "args": {
      "External id": 107307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107307,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107307, "pid": 5, "tid": 7, "ts": 1716454223045970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984136, "dur": 6,
    "args": {
      "External id": 107307,
      "cbid": 211, "correlation": 107307
    }
  },
  {
    "ph": "s", "id": 107307, "pid": 76337, "tid": -914061504, "ts": 1716454222984136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984193, "dur": 1,
    "args": {
      "External id": 107318,
      "cbid": 251, "correlation": 107318
    }
  },
  {
    "ph": "f", "id": 107318, "pid": 76337, "tid": -914061504, "ts": 1716454222984193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984197, "dur": 0,
    "args": {
      "External id": 107319,
      "cbid": 251, "correlation": 107319
    }
  },
  {
    "ph": "f", "id": 107319, "pid": 76337, "tid": -914061504, "ts": 1716454222984197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223045977, "dur": 8,
    "args": {
      "External id": 107320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107320,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107320, "pid": 5, "tid": 7, "ts": 1716454223045977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984198, "dur": 12,
    "args": {
      "External id": 107320,
      "cbid": 211, "correlation": 107320
    }
  },
  {
    "ph": "s", "id": 107320, "pid": 76337, "tid": -914061504, "ts": 1716454222984198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223045986, "dur": 3,
    "args": {
      "External id": 107322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107322,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107322, "pid": 5, "tid": 7, "ts": 1716454223045986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984211, "dur": 5,
    "args": {
      "External id": 107322,
      "cbid": 211, "correlation": 107322
    }
  },
  {
    "ph": "s", "id": 107322, "pid": 76337, "tid": -914061504, "ts": 1716454222984211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223045991, "dur": 53,
    "args": {
      "External id": 107347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107347,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107347, "pid": 5, "tid": 7, "ts": 1716454223045991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984290, "dur": 13,
    "args": {
      "External id": 107347,
      "cbid": 211, "correlation": 107347
    }
  },
  {
    "ph": "s", "id": 107347, "pid": 76337, "tid": -914061504, "ts": 1716454222984290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984392, "dur": 1,
    "args": {
      "External id": 107365,
      "cbid": 251, "correlation": 107365
    }
  },
  {
    "ph": "f", "id": 107365, "pid": 76337, "tid": -914061504, "ts": 1716454222984392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223046045, "dur": 90,
    "args": {
      "External id": 107367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107367,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107367, "pid": 5, "tid": 7, "ts": 1716454223046045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984398, "dur": 13,
    "args": {
      "External id": 107367,
      "cbid": 211, "correlation": 107367
    }
  },
  {
    "ph": "s", "id": 107367, "pid": 76337, "tid": -914061504, "ts": 1716454222984398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223046136, "dur": 9,
    "args": {
      "External id": 107375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107375,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107375, "pid": 5, "tid": 7, "ts": 1716454223046136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984468, "dur": 12,
    "args": {
      "External id": 107375,
      "cbid": 211, "correlation": 107375
    }
  },
  {
    "ph": "s", "id": 107375, "pid": 76337, "tid": -914061504, "ts": 1716454222984468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223046147, "dur": 21,
    "args": {
      "External id": 107383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107383, "pid": 5, "tid": 7, "ts": 1716454223046147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984509, "dur": 9,
    "args": {
      "External id": 107383,
      "cbid": 211, "correlation": 107383
    }
  },
  {
    "ph": "s", "id": 107383, "pid": 76337, "tid": -914061504, "ts": 1716454222984509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223046170, "dur": 17,
    "args": {
      "External id": 107405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107405,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107405, "pid": 5, "tid": 7, "ts": 1716454223046170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984561, "dur": 10,
    "args": {
      "External id": 107405,
      "cbid": 211, "correlation": 107405
    }
  },
  {
    "ph": "s", "id": 107405, "pid": 76337, "tid": -914061504, "ts": 1716454222984561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984649, "dur": 1,
    "args": {
      "External id": 107421,
      "cbid": 251, "correlation": 107421
    }
  },
  {
    "ph": "f", "id": 107421, "pid": 76337, "tid": -914061504, "ts": 1716454222984649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984654, "dur": 0,
    "args": {
      "External id": 107423,
      "cbid": 251, "correlation": 107423
    }
  },
  {
    "ph": "f", "id": 107423, "pid": 76337, "tid": -914061504, "ts": 1716454222984654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223046188, "dur": 493,
    "args": {
      "External id": 107424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107424,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107424, "pid": 5, "tid": 7, "ts": 1716454223046188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984656, "dur": 12,
    "args": {
      "External id": 107424,
      "cbid": 211, "correlation": 107424
    }
  },
  {
    "ph": "s", "id": 107424, "pid": 76337, "tid": -914061504, "ts": 1716454222984656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223046683, "dur": 66,
    "args": {
      "External id": 107432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107432, "pid": 5, "tid": 7, "ts": 1716454223046683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984720, "dur": 13,
    "args": {
      "External id": 107432,
      "cbid": 211, "correlation": 107432
    }
  },
  {
    "ph": "s", "id": 107432, "pid": 76337, "tid": -914061504, "ts": 1716454222984720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223046750, "dur": 64,
    "args": {
      "External id": 107440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107440, "pid": 5, "tid": 7, "ts": 1716454223046750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984752, "dur": 8,
    "args": {
      "External id": 107440,
      "cbid": 211, "correlation": 107440
    }
  },
  {
    "ph": "s", "id": 107440, "pid": 76337, "tid": -914061504, "ts": 1716454222984752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984831, "dur": 1,
    "args": {
      "External id": 107456,
      "cbid": 251, "correlation": 107456
    }
  },
  {
    "ph": "f", "id": 107456, "pid": 76337, "tid": -914061504, "ts": 1716454222984831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223046816, "dur": 1,
    "args": {
      "External id": 107458,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 107458,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 107458, "pid": 5, "tid": 7, "ts": 1716454223046816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984836, "dur": 9,
    "args": {
      "External id": 107458,
      "cbid": 51, "correlation": 107458
    }
  },
  {
    "ph": "s", "id": 107458, "pid": 76337, "tid": -914061504, "ts": 1716454222984836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223046820, "dur": 269,
    "args": {
      "External id": 107459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107459,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107459, "pid": 5, "tid": 7, "ts": 1716454223046820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984847, "dur": 11,
    "args": {
      "External id": 107459,
      "cbid": 211, "correlation": 107459
    }
  },
  {
    "ph": "s", "id": 107459, "pid": 76337, "tid": -914061504, "ts": 1716454222984847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223047090, "dur": 13,
    "args": {
      "External id": 107467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107467,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107467, "pid": 5, "tid": 7, "ts": 1716454223047090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984889, "dur": 10,
    "args": {
      "External id": 107467,
      "cbid": 211, "correlation": 107467
    }
  },
  {
    "ph": "s", "id": 107467, "pid": 76337, "tid": -914061504, "ts": 1716454222984889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223047105, "dur": 37,
    "args": {
      "External id": 107478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107478,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107478, "pid": 5, "tid": 7, "ts": 1716454223047105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222984957, "dur": 12,
    "args": {
      "External id": 107478,
      "cbid": 211, "correlation": 107478
    }
  },
  {
    "ph": "s", "id": 107478, "pid": 76337, "tid": -914061504, "ts": 1716454222984957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985029, "dur": 0,
    "args": {
      "External id": 107490,
      "cbid": 317, "correlation": 107490
    }
  },
  {
    "ph": "f", "id": 107490, "pid": 76337, "tid": -914061504, "ts": 1716454222985029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985030, "dur": 0,
    "args": {
      "External id": 107491,
      "cbid": 203, "correlation": 107491
    }
  },
  {
    "ph": "f", "id": 107491, "pid": 76337, "tid": -914061504, "ts": 1716454222985030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985031, "dur": 0,
    "args": {
      "External id": 107492,
      "cbid": 205, "correlation": 107492
    }
  },
  {
    "ph": "f", "id": 107492, "pid": 76337, "tid": -914061504, "ts": 1716454222985031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223047143, "dur": 12,
    "args": {
      "External id": 107496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107496,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107496, "pid": 5, "tid": 7, "ts": 1716454223047143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985046, "dur": 12,
    "args": {
      "External id": 107496,
      "cbid": 211, "correlation": 107496
    }
  },
  {
    "ph": "s", "id": 107496, "pid": 76337, "tid": -914061504, "ts": 1716454222985046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223047156, "dur": 4,
    "args": {
      "External id": 107498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107498, "pid": 5, "tid": 7, "ts": 1716454223047156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985062, "dur": 6,
    "args": {
      "External id": 107498,
      "cbid": 211, "correlation": 107498
    }
  },
  {
    "ph": "s", "id": 107498, "pid": 76337, "tid": -914061504, "ts": 1716454222985062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985071, "dur": 0,
    "args": {
      "External id": 107499,
      "cbid": 51, "correlation": 107499
    }
  },
  {
    "ph": "s", "id": 107499, "pid": 76337, "tid": -914061504, "ts": 1716454222985071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223047161, "dur": 96,
    "args": {
      "External id": 107500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107500,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 107500, "pid": 5, "tid": 7, "ts": 1716454223047161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985072, "dur": 5,
    "args": {
      "External id": 107500,
      "cbid": 211, "correlation": 107500
    }
  },
  {
    "ph": "s", "id": 107500, "pid": 76337, "tid": -914061504, "ts": 1716454222985072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223047259, "dur": 16,
    "args": {
      "External id": 107505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107505,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107505, "pid": 5, "tid": 7, "ts": 1716454223047259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985100, "dur": 9,
    "args": {
      "External id": 107505,
      "cbid": 211, "correlation": 107505
    }
  },
  {
    "ph": "s", "id": 107505, "pid": 76337, "tid": -914061504, "ts": 1716454222985100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223047276, "dur": 13,
    "args": {
      "External id": 107513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107513,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107513, "pid": 5, "tid": 7, "ts": 1716454223047276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985131, "dur": 9,
    "args": {
      "External id": 107513,
      "cbid": 211, "correlation": 107513
    }
  },
  {
    "ph": "s", "id": 107513, "pid": 76337, "tid": -914061504, "ts": 1716454222985131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223047290, "dur": 56,
    "args": {
      "External id": 107524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107524,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107524, "pid": 5, "tid": 7, "ts": 1716454223047290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985195, "dur": 11,
    "args": {
      "External id": 107524,
      "cbid": 211, "correlation": 107524
    }
  },
  {
    "ph": "s", "id": 107524, "pid": 76337, "tid": -914061504, "ts": 1716454222985195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985249, "dur": 0,
    "args": {
      "External id": 107534,
      "cbid": 317, "correlation": 107534
    }
  },
  {
    "ph": "f", "id": 107534, "pid": 76337, "tid": -914061504, "ts": 1716454222985249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985250, "dur": 0,
    "args": {
      "External id": 107535,
      "cbid": 203, "correlation": 107535
    }
  },
  {
    "ph": "f", "id": 107535, "pid": 76337, "tid": -914061504, "ts": 1716454222985250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985251, "dur": 0,
    "args": {
      "External id": 107536,
      "cbid": 205, "correlation": 107536
    }
  },
  {
    "ph": "f", "id": 107536, "pid": 76337, "tid": -914061504, "ts": 1716454222985251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223047348, "dur": 39,
    "args": {
      "External id": 107540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107540,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107540, "pid": 5, "tid": 7, "ts": 1716454223047348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985267, "dur": 12,
    "args": {
      "External id": 107540,
      "cbid": 211, "correlation": 107540
    }
  },
  {
    "ph": "s", "id": 107540, "pid": 76337, "tid": -914061504, "ts": 1716454222985267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223047388, "dur": 160,
    "args": {
      "External id": 107542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107542, "pid": 5, "tid": 7, "ts": 1716454223047388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985281, "dur": 6,
    "args": {
      "External id": 107542,
      "cbid": 211, "correlation": 107542
    }
  },
  {
    "ph": "s", "id": 107542, "pid": 76337, "tid": -914061504, "ts": 1716454222985281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223047549, "dur": 1962,
    "args": {
      "External id": 107544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107544,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107544, "pid": 5, "tid": 7, "ts": 1716454223047549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985294, "dur": 8,
    "args": {
      "External id": 107544,
      "cbid": 211, "correlation": 107544
    }
  },
  {
    "ph": "s", "id": 107544, "pid": 76337, "tid": -914061504, "ts": 1716454222985294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223049513, "dur": 40,
    "args": {
      "External id": 107546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107546, "pid": 5, "tid": 7, "ts": 1716454223049513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985305, "dur": 5,
    "args": {
      "External id": 107546,
      "cbid": 211, "correlation": 107546
    }
  },
  {
    "ph": "s", "id": 107546, "pid": 76337, "tid": -914061504, "ts": 1716454222985305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223049553, "dur": 58,
    "args": {
      "External id": 107552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107552,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107552, "pid": 5, "tid": 7, "ts": 1716454223049553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985334, "dur": 8,
    "args": {
      "External id": 107552,
      "cbid": 211, "correlation": 107552
    }
  },
  {
    "ph": "s", "id": 107552, "pid": 76337, "tid": -914061504, "ts": 1716454222985334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223049613, "dur": 86,
    "args": {
      "External id": 107561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107561,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107561, "pid": 5, "tid": 7, "ts": 1716454223049613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985423, "dur": 13,
    "args": {
      "External id": 107561,
      "cbid": 211, "correlation": 107561
    }
  },
  {
    "ph": "s", "id": 107561, "pid": 76337, "tid": -914061504, "ts": 1716454222985423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223049700, "dur": 72,
    "args": {
      "External id": 107581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107581,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 107581, "pid": 5, "tid": 7, "ts": 1716454223049700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985494, "dur": 11,
    "args": {
      "External id": 107581,
      "cbid": 211, "correlation": 107581
    }
  },
  {
    "ph": "s", "id": 107581, "pid": 76337, "tid": -914061504, "ts": 1716454222985494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223049774, "dur": 5,
    "args": {
      "External id": 107593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107593,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 107593, "pid": 5, "tid": 7, "ts": 1716454223049774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985515, "dur": 6,
    "args": {
      "External id": 107593,
      "cbid": 211, "correlation": 107593
    }
  },
  {
    "ph": "s", "id": 107593, "pid": 76337, "tid": -914061504, "ts": 1716454222985515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223049780, "dur": 80,
    "args": {
      "External id": 107596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107596,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107596, "pid": 5, "tid": 7, "ts": 1716454223049780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985534, "dur": 7,
    "args": {
      "External id": 107596,
      "cbid": 211, "correlation": 107596
    }
  },
  {
    "ph": "s", "id": 107596, "pid": 76337, "tid": -914061504, "ts": 1716454222985534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223049862, "dur": 53,
    "args": {
      "External id": 107605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107605,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107605, "pid": 5, "tid": 7, "ts": 1716454223049862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985575, "dur": 11,
    "args": {
      "External id": 107605,
      "cbid": 211, "correlation": 107605
    }
  },
  {
    "ph": "s", "id": 107605, "pid": 76337, "tid": -914061504, "ts": 1716454222985575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985627, "dur": 0,
    "args": {
      "External id": 107615,
      "cbid": 317, "correlation": 107615
    }
  },
  {
    "ph": "f", "id": 107615, "pid": 76337, "tid": -914061504, "ts": 1716454222985627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985628, "dur": 0,
    "args": {
      "External id": 107616,
      "cbid": 203, "correlation": 107616
    }
  },
  {
    "ph": "f", "id": 107616, "pid": 76337, "tid": -914061504, "ts": 1716454222985628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985629, "dur": 0,
    "args": {
      "External id": 107617,
      "cbid": 205, "correlation": 107617
    }
  },
  {
    "ph": "f", "id": 107617, "pid": 76337, "tid": -914061504, "ts": 1716454222985629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223049916, "dur": 57,
    "args": {
      "External id": 107621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107621,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107621, "pid": 5, "tid": 7, "ts": 1716454223049916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985646, "dur": 11,
    "args": {
      "External id": 107621,
      "cbid": 211, "correlation": 107621
    }
  },
  {
    "ph": "s", "id": 107621, "pid": 76337, "tid": -914061504, "ts": 1716454222985646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223049974, "dur": 119,
    "args": {
      "External id": 107623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107623,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107623, "pid": 5, "tid": 7, "ts": 1716454223049974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985660, "dur": 5,
    "args": {
      "External id": 107623,
      "cbid": 211, "correlation": 107623
    }
  },
  {
    "ph": "s", "id": 107623, "pid": 76337, "tid": -914061504, "ts": 1716454222985660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223050094, "dur": 1877,
    "args": {
      "External id": 107625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107625,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107625, "pid": 5, "tid": 7, "ts": 1716454223050094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985671, "dur": 6,
    "args": {
      "External id": 107625,
      "cbid": 211, "correlation": 107625
    }
  },
  {
    "ph": "s", "id": 107625, "pid": 76337, "tid": -914061504, "ts": 1716454222985671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223051973, "dur": 19,
    "args": {
      "External id": 107627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107627, "pid": 5, "tid": 7, "ts": 1716454223051973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985680, "dur": 6,
    "args": {
      "External id": 107627,
      "cbid": 211, "correlation": 107627
    }
  },
  {
    "ph": "s", "id": 107627, "pid": 76337, "tid": -914061504, "ts": 1716454222985680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223051993, "dur": 32,
    "args": {
      "External id": 107633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107633, "pid": 5, "tid": 7, "ts": 1716454223051993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985709, "dur": 8,
    "args": {
      "External id": 107633,
      "cbid": 211, "correlation": 107633
    }
  },
  {
    "ph": "s", "id": 107633, "pid": 76337, "tid": -914061504, "ts": 1716454222985709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223052027, "dur": 3,
    "args": {
      "External id": 107641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107641,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 107641, "pid": 5, "tid": 7, "ts": 1716454223052027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985753, "dur": 9,
    "args": {
      "External id": 107641,
      "cbid": 211, "correlation": 107641
    }
  },
  {
    "ph": "s", "id": 107641, "pid": 76337, "tid": -914061504, "ts": 1716454222985753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985819, "dur": 1,
    "args": {
      "External id": 107657,
      "cbid": 251, "correlation": 107657
    }
  },
  {
    "ph": "f", "id": 107657, "pid": 76337, "tid": -914061504, "ts": 1716454222985819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985824, "dur": 0,
    "args": {
      "External id": 107659,
      "cbid": 251, "correlation": 107659
    }
  },
  {
    "ph": "f", "id": 107659, "pid": 76337, "tid": -914061504, "ts": 1716454222985824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223052032, "dur": 12,
    "args": {
      "External id": 107660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107660,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 107660, "pid": 5, "tid": 7, "ts": 1716454223052032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985826, "dur": 12,
    "args": {
      "External id": 107660,
      "cbid": 211, "correlation": 107660
    }
  },
  {
    "ph": "s", "id": 107660, "pid": 76337, "tid": -914061504, "ts": 1716454222985826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223052045, "dur": 5,
    "args": {
      "External id": 107662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107662,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 107662, "pid": 5, "tid": 7, "ts": 1716454223052045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985840, "dur": 5,
    "args": {
      "External id": 107662,
      "cbid": 211, "correlation": 107662
    }
  },
  {
    "ph": "s", "id": 107662, "pid": 76337, "tid": -914061504, "ts": 1716454222985840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223052051, "dur": 29,
    "args": {
      "External id": 107672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107672, "pid": 5, "tid": 7, "ts": 1716454223052051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985897, "dur": 12,
    "args": {
      "External id": 107672,
      "cbid": 211, "correlation": 107672
    }
  },
  {
    "ph": "s", "id": 107672, "pid": 76337, "tid": -914061504, "ts": 1716454222985897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223052081, "dur": 30,
    "args": {
      "External id": 107692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107692,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 107692, "pid": 5, "tid": 7, "ts": 1716454223052081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985963, "dur": 19,
    "args": {
      "External id": 107692,
      "cbid": 211, "correlation": 107692
    }
  },
  {
    "ph": "s", "id": 107692, "pid": 76337, "tid": -914061504, "ts": 1716454222985963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223052113, "dur": 5,
    "args": {
      "External id": 107704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107704,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 107704, "pid": 5, "tid": 7, "ts": 1716454223052113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222985993, "dur": 7,
    "args": {
      "External id": 107704,
      "cbid": 211, "correlation": 107704
    }
  },
  {
    "ph": "s", "id": 107704, "pid": 76337, "tid": -914061504, "ts": 1716454222985993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223052119, "dur": 30,
    "args": {
      "External id": 107707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107707,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107707, "pid": 5, "tid": 7, "ts": 1716454223052119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986011, "dur": 6,
    "args": {
      "External id": 107707,
      "cbid": 211, "correlation": 107707
    }
  },
  {
    "ph": "s", "id": 107707, "pid": 76337, "tid": -914061504, "ts": 1716454222986011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223052150, "dur": 20,
    "args": {
      "External id": 107716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107716,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107716, "pid": 5, "tid": 7, "ts": 1716454223052150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986053, "dur": 10,
    "args": {
      "External id": 107716,
      "cbid": 211, "correlation": 107716
    }
  },
  {
    "ph": "s", "id": 107716, "pid": 76337, "tid": -914061504, "ts": 1716454222986053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986116, "dur": 0,
    "args": {
      "External id": 107726,
      "cbid": 317, "correlation": 107726
    }
  },
  {
    "ph": "f", "id": 107726, "pid": 76337, "tid": -914061504, "ts": 1716454222986116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986117, "dur": 0,
    "args": {
      "External id": 107727,
      "cbid": 203, "correlation": 107727
    }
  },
  {
    "ph": "f", "id": 107727, "pid": 76337, "tid": -914061504, "ts": 1716454222986117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986117, "dur": 0,
    "args": {
      "External id": 107728,
      "cbid": 205, "correlation": 107728
    }
  },
  {
    "ph": "f", "id": 107728, "pid": 76337, "tid": -914061504, "ts": 1716454222986117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223052171, "dur": 22,
    "args": {
      "External id": 107732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107732, "pid": 5, "tid": 7, "ts": 1716454223052171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986133, "dur": 13,
    "args": {
      "External id": 107732,
      "cbid": 211, "correlation": 107732
    }
  },
  {
    "ph": "s", "id": 107732, "pid": 76337, "tid": -914061504, "ts": 1716454222986133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223052195, "dur": 43,
    "args": {
      "External id": 107734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107734, "pid": 5, "tid": 7, "ts": 1716454223052195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986148, "dur": 5,
    "args": {
      "External id": 107734,
      "cbid": 211, "correlation": 107734
    }
  },
  {
    "ph": "s", "id": 107734, "pid": 76337, "tid": -914061504, "ts": 1716454222986148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223052239, "dur": 641,
    "args": {
      "External id": 107736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107736,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107736, "pid": 5, "tid": 7, "ts": 1716454223052239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986159, "dur": 6,
    "args": {
      "External id": 107736,
      "cbid": 211, "correlation": 107736
    }
  },
  {
    "ph": "s", "id": 107736, "pid": 76337, "tid": -914061504, "ts": 1716454222986159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223052882, "dur": 22,
    "args": {
      "External id": 107738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107738, "pid": 5, "tid": 7, "ts": 1716454223052882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986169, "dur": 5,
    "args": {
      "External id": 107738,
      "cbid": 211, "correlation": 107738
    }
  },
  {
    "ph": "s", "id": 107738, "pid": 76337, "tid": -914061504, "ts": 1716454222986169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223052905, "dur": 32,
    "args": {
      "External id": 107744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107744, "pid": 5, "tid": 7, "ts": 1716454223052905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986197, "dur": 8,
    "args": {
      "External id": 107744,
      "cbid": 211, "correlation": 107744
    }
  },
  {
    "ph": "s", "id": 107744, "pid": 76337, "tid": -914061504, "ts": 1716454222986197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986255, "dur": 0,
    "args": {
      "External id": 107754,
      "cbid": 317, "correlation": 107754
    }
  },
  {
    "ph": "f", "id": 107754, "pid": 76337, "tid": -914061504, "ts": 1716454222986255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986255, "dur": 0,
    "args": {
      "External id": 107755,
      "cbid": 203, "correlation": 107755
    }
  },
  {
    "ph": "f", "id": 107755, "pid": 76337, "tid": -914061504, "ts": 1716454222986255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986256, "dur": 0,
    "args": {
      "External id": 107756,
      "cbid": 205, "correlation": 107756
    }
  },
  {
    "ph": "f", "id": 107756, "pid": 76337, "tid": -914061504, "ts": 1716454222986256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223052938, "dur": 56,
    "args": {
      "External id": 107760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107760,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107760, "pid": 5, "tid": 7, "ts": 1716454223052938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986268, "dur": 11,
    "args": {
      "External id": 107760,
      "cbid": 211, "correlation": 107760
    }
  },
  {
    "ph": "s", "id": 107760, "pid": 76337, "tid": -914061504, "ts": 1716454222986268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223052996, "dur": 264,
    "args": {
      "External id": 107762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107762,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107762, "pid": 5, "tid": 7, "ts": 1716454223052996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986287, "dur": 8,
    "args": {
      "External id": 107762,
      "cbid": 211, "correlation": 107762
    }
  },
  {
    "ph": "s", "id": 107762, "pid": 76337, "tid": -914061504, "ts": 1716454222986287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223053261, "dur": 21,
    "args": {
      "External id": 107764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107764,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107764, "pid": 5, "tid": 7, "ts": 1716454223053261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986298, "dur": 5,
    "args": {
      "External id": 107764,
      "cbid": 211, "correlation": 107764
    }
  },
  {
    "ph": "s", "id": 107764, "pid": 76337, "tid": -914061504, "ts": 1716454222986298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223053283, "dur": 31,
    "args": {
      "External id": 107770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107770,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107770, "pid": 5, "tid": 7, "ts": 1716454223053283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986324, "dur": 9,
    "args": {
      "External id": 107770,
      "cbid": 211, "correlation": 107770
    }
  },
  {
    "ph": "s", "id": 107770, "pid": 76337, "tid": -914061504, "ts": 1716454222986324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223053316, "dur": 27,
    "args": {
      "External id": 107778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107778,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107778, "pid": 5, "tid": 7, "ts": 1716454223053316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986352, "dur": 8,
    "args": {
      "External id": 107778,
      "cbid": 211, "correlation": 107778
    }
  },
  {
    "ph": "s", "id": 107778, "pid": 76337, "tid": -914061504, "ts": 1716454222986352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223053344, "dur": 20,
    "args": {
      "External id": 107786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107786,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107786, "pid": 5, "tid": 7, "ts": 1716454223053344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986381, "dur": 8,
    "args": {
      "External id": 107786,
      "cbid": 211, "correlation": 107786
    }
  },
  {
    "ph": "s", "id": 107786, "pid": 76337, "tid": -914061504, "ts": 1716454222986381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223053365, "dur": 29,
    "args": {
      "External id": 107806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107806,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 107806, "pid": 5, "tid": 7, "ts": 1716454223053365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986462, "dur": 12,
    "args": {
      "External id": 107806,
      "cbid": 211, "correlation": 107806
    }
  },
  {
    "ph": "s", "id": 107806, "pid": 76337, "tid": -914061504, "ts": 1716454222986462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223053396, "dur": 5,
    "args": {
      "External id": 107818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107818,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 107818, "pid": 5, "tid": 7, "ts": 1716454223053396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986484, "dur": 6,
    "args": {
      "External id": 107818,
      "cbid": 211, "correlation": 107818
    }
  },
  {
    "ph": "s", "id": 107818, "pid": 76337, "tid": -914061504, "ts": 1716454222986484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223053402, "dur": 31,
    "args": {
      "External id": 107821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107821,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107821, "pid": 5, "tid": 7, "ts": 1716454223053402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986502, "dur": 7,
    "args": {
      "External id": 107821,
      "cbid": 211, "correlation": 107821
    }
  },
  {
    "ph": "s", "id": 107821, "pid": 76337, "tid": -914061504, "ts": 1716454222986502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986560, "dur": 0,
    "args": {
      "External id": 107832,
      "cbid": 317, "correlation": 107832
    }
  },
  {
    "ph": "f", "id": 107832, "pid": 76337, "tid": -914061504, "ts": 1716454222986560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986560, "dur": 0,
    "args": {
      "External id": 107833,
      "cbid": 203, "correlation": 107833
    }
  },
  {
    "ph": "f", "id": 107833, "pid": 76337, "tid": -914061504, "ts": 1716454222986560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986561, "dur": 0,
    "args": {
      "External id": 107834,
      "cbid": 205, "correlation": 107834
    }
  },
  {
    "ph": "f", "id": 107834, "pid": 76337, "tid": -914061504, "ts": 1716454222986561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223053434, "dur": 22,
    "args": {
      "External id": 107838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107838, "pid": 5, "tid": 7, "ts": 1716454223053434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986574, "dur": 13,
    "args": {
      "External id": 107838,
      "cbid": 211, "correlation": 107838
    }
  },
  {
    "ph": "s", "id": 107838, "pid": 76337, "tid": -914061504, "ts": 1716454222986574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223053457, "dur": 104,
    "args": {
      "External id": 107840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107840,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107840, "pid": 5, "tid": 7, "ts": 1716454223053457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986594, "dur": 6,
    "args": {
      "External id": 107840,
      "cbid": 211, "correlation": 107840
    }
  },
  {
    "ph": "s", "id": 107840, "pid": 76337, "tid": -914061504, "ts": 1716454222986594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223053563, "dur": 20,
    "args": {
      "External id": 107842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107842,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107842, "pid": 5, "tid": 7, "ts": 1716454223053563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986603, "dur": 5,
    "args": {
      "External id": 107842,
      "cbid": 211, "correlation": 107842
    }
  },
  {
    "ph": "s", "id": 107842, "pid": 76337, "tid": -914061504, "ts": 1716454222986603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223053584, "dur": 32,
    "args": {
      "External id": 107848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107848, "pid": 5, "tid": 7, "ts": 1716454223053584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986631, "dur": 10,
    "args": {
      "External id": 107848,
      "cbid": 211, "correlation": 107848
    }
  },
  {
    "ph": "s", "id": 107848, "pid": 76337, "tid": -914061504, "ts": 1716454222986631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223053618, "dur": 198,
    "args": {
      "External id": 107857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107857,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107857, "pid": 5, "tid": 7, "ts": 1716454223053618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986714, "dur": 14,
    "args": {
      "External id": 107857,
      "cbid": 211, "correlation": 107857
    }
  },
  {
    "ph": "s", "id": 107857, "pid": 76337, "tid": -914061504, "ts": 1716454222986714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223053817, "dur": 63,
    "args": {
      "External id": 107879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107879,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107879, "pid": 5, "tid": 7, "ts": 1716454223053817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986770, "dur": 10,
    "args": {
      "External id": 107879,
      "cbid": 211, "correlation": 107879
    }
  },
  {
    "ph": "s", "id": 107879, "pid": 76337, "tid": -914061504, "ts": 1716454222986770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986861, "dur": 2,
    "args": {
      "External id": 107890,
      "cbid": 251, "correlation": 107890
    }
  },
  {
    "ph": "f", "id": 107890, "pid": 76337, "tid": -914061504, "ts": 1716454222986861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223053881, "dur": 156,
    "args": {
      "External id": 107891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107891,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107891, "pid": 5, "tid": 7, "ts": 1716454223053881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986867, "dur": 13,
    "args": {
      "External id": 107891,
      "cbid": 211, "correlation": 107891
    }
  },
  {
    "ph": "s", "id": 107891, "pid": 76337, "tid": -914061504, "ts": 1716454222986867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986938, "dur": 1,
    "args": {
      "External id": 107902,
      "cbid": 251, "correlation": 107902
    }
  },
  {
    "ph": "f", "id": 107902, "pid": 76337, "tid": -914061504, "ts": 1716454222986938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223054039, "dur": 143,
    "args": {
      "External id": 107903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107903,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107903, "pid": 5, "tid": 7, "ts": 1716454223054039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222986942, "dur": 11,
    "args": {
      "External id": 107903,
      "cbid": 211, "correlation": 107903
    }
  },
  {
    "ph": "s", "id": 107903, "pid": 76337, "tid": -914061504, "ts": 1716454222986942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987016, "dur": 1,
    "args": {
      "External id": 107914,
      "cbid": 251, "correlation": 107914
    }
  },
  {
    "ph": "f", "id": 107914, "pid": 76337, "tid": -914061504, "ts": 1716454222987016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223054184, "dur": 145,
    "args": {
      "External id": 107915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107915,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 107915, "pid": 5, "tid": 7, "ts": 1716454223054184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987020, "dur": 11,
    "args": {
      "External id": 107915,
      "cbid": 211, "correlation": 107915
    }
  },
  {
    "ph": "s", "id": 107915, "pid": 76337, "tid": -914061504, "ts": 1716454222987020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223054330, "dur": 1907,
    "args": {
      "External id": 107936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107936,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 107936, "pid": 5, "tid": 7, "ts": 1716454223054330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987106, "dur": 13,
    "args": {
      "External id": 107936,
      "cbid": 211, "correlation": 107936
    }
  },
  {
    "ph": "s", "id": 107936, "pid": 76337, "tid": -914061504, "ts": 1716454222987106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987206, "dur": 2,
    "args": {
      "External id": 107954,
      "cbid": 251, "correlation": 107954
    }
  },
  {
    "ph": "f", "id": 107954, "pid": 76337, "tid": -914061504, "ts": 1716454222987206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223056238, "dur": 148,
    "args": {
      "External id": 107956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107956,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 107956, "pid": 5, "tid": 7, "ts": 1716454223056238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987213, "dur": 13,
    "args": {
      "External id": 107956,
      "cbid": 211, "correlation": 107956
    }
  },
  {
    "ph": "s", "id": 107956, "pid": 76337, "tid": -914061504, "ts": 1716454222987213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223056388, "dur": 35,
    "args": {
      "External id": 107964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107964,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107964, "pid": 5, "tid": 7, "ts": 1716454223056388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987283, "dur": 13,
    "args": {
      "External id": 107964,
      "cbid": 211, "correlation": 107964
    }
  },
  {
    "ph": "s", "id": 107964, "pid": 76337, "tid": -914061504, "ts": 1716454222987283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223056424, "dur": 51,
    "args": {
      "External id": 107972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107972,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107972, "pid": 5, "tid": 7, "ts": 1716454223056424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987322, "dur": 8,
    "args": {
      "External id": 107972,
      "cbid": 211, "correlation": 107972
    }
  },
  {
    "ph": "s", "id": 107972, "pid": 76337, "tid": -914061504, "ts": 1716454222987322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223056477, "dur": 30,
    "args": {
      "External id": 107983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 107983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 107983, "pid": 5, "tid": 7, "ts": 1716454223056477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987392, "dur": 12,
    "args": {
      "External id": 107983,
      "cbid": 211, "correlation": 107983
    }
  },
  {
    "ph": "s", "id": 107983, "pid": 76337, "tid": -914061504, "ts": 1716454222987392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223056508, "dur": 33,
    "args": {
      "External id": 108005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108005,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108005, "pid": 5, "tid": 7, "ts": 1716454223056508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987423, "dur": 7,
    "args": {
      "External id": 108005,
      "cbid": 211, "correlation": 108005
    }
  },
  {
    "ph": "s", "id": 108005, "pid": 76337, "tid": -914061504, "ts": 1716454222987423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987508, "dur": 1,
    "args": {
      "External id": 108016,
      "cbid": 251, "correlation": 108016
    }
  },
  {
    "ph": "f", "id": 108016, "pid": 76337, "tid": -914061504, "ts": 1716454222987508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223056543, "dur": 75,
    "args": {
      "External id": 108017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108017,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108017, "pid": 5, "tid": 7, "ts": 1716454223056543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987513, "dur": 13,
    "args": {
      "External id": 108017,
      "cbid": 211, "correlation": 108017
    }
  },
  {
    "ph": "s", "id": 108017, "pid": 76337, "tid": -914061504, "ts": 1716454222987513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987583, "dur": 1,
    "args": {
      "External id": 108028,
      "cbid": 251, "correlation": 108028
    }
  },
  {
    "ph": "f", "id": 108028, "pid": 76337, "tid": -914061504, "ts": 1716454222987583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987587, "dur": 0,
    "args": {
      "External id": 108029,
      "cbid": 251, "correlation": 108029
    }
  },
  {
    "ph": "f", "id": 108029, "pid": 76337, "tid": -914061504, "ts": 1716454222987587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223056619, "dur": 11,
    "args": {
      "External id": 108030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108030,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 108030, "pid": 5, "tid": 7, "ts": 1716454223056619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987589, "dur": 12,
    "args": {
      "External id": 108030,
      "cbid": 211, "correlation": 108030
    }
  },
  {
    "ph": "s", "id": 108030, "pid": 76337, "tid": -914061504, "ts": 1716454222987589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223056631, "dur": 5,
    "args": {
      "External id": 108032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108032,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 108032, "pid": 5, "tid": 7, "ts": 1716454223056631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987602, "dur": 6,
    "args": {
      "External id": 108032,
      "cbid": 211, "correlation": 108032
    }
  },
  {
    "ph": "s", "id": 108032, "pid": 76337, "tid": -914061504, "ts": 1716454222987602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987660, "dur": 1,
    "args": {
      "External id": 108043,
      "cbid": 251, "correlation": 108043
    }
  },
  {
    "ph": "f", "id": 108043, "pid": 76337, "tid": -914061504, "ts": 1716454222987660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987664, "dur": 0,
    "args": {
      "External id": 108044,
      "cbid": 251, "correlation": 108044
    }
  },
  {
    "ph": "f", "id": 108044, "pid": 76337, "tid": -914061504, "ts": 1716454222987664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223056638, "dur": 7,
    "args": {
      "External id": 108045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108045,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 108045, "pid": 5, "tid": 7, "ts": 1716454223056638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987665, "dur": 11,
    "args": {
      "External id": 108045,
      "cbid": 211, "correlation": 108045
    }
  },
  {
    "ph": "s", "id": 108045, "pid": 76337, "tid": -914061504, "ts": 1716454222987665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223056646, "dur": 3,
    "args": {
      "External id": 108047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108047,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 108047, "pid": 5, "tid": 7, "ts": 1716454223056646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987678, "dur": 6,
    "args": {
      "External id": 108047,
      "cbid": 211, "correlation": 108047
    }
  },
  {
    "ph": "s", "id": 108047, "pid": 76337, "tid": -914061504, "ts": 1716454222987678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223056651, "dur": 89,
    "args": {
      "External id": 108068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108068,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 108068, "pid": 5, "tid": 7, "ts": 1716454223056651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987752, "dur": 12,
    "args": {
      "External id": 108068,
      "cbid": 211, "correlation": 108068
    }
  },
  {
    "ph": "s", "id": 108068, "pid": 76337, "tid": -914061504, "ts": 1716454222987752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987849, "dur": 1,
    "args": {
      "External id": 108086,
      "cbid": 251, "correlation": 108086
    }
  },
  {
    "ph": "f", "id": 108086, "pid": 76337, "tid": -914061504, "ts": 1716454222987849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223056741, "dur": 96,
    "args": {
      "External id": 108088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108088,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108088, "pid": 5, "tid": 7, "ts": 1716454223056741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987855, "dur": 13,
    "args": {
      "External id": 108088,
      "cbid": 211, "correlation": 108088
    }
  },
  {
    "ph": "s", "id": 108088, "pid": 76337, "tid": -914061504, "ts": 1716454222987855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223056838, "dur": 19,
    "args": {
      "External id": 108096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108096,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108096, "pid": 5, "tid": 7, "ts": 1716454223056838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987924, "dur": 12,
    "args": {
      "External id": 108096,
      "cbid": 211, "correlation": 108096
    }
  },
  {
    "ph": "s", "id": 108096, "pid": 76337, "tid": -914061504, "ts": 1716454222987924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223056859, "dur": 37,
    "args": {
      "External id": 108104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108104, "pid": 5, "tid": 7, "ts": 1716454223056859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222987966, "dur": 17,
    "args": {
      "External id": 108104,
      "cbid": 211, "correlation": 108104
    }
  },
  {
    "ph": "s", "id": 108104, "pid": 76337, "tid": -914061504, "ts": 1716454222987966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223056897, "dur": 34,
    "args": {
      "External id": 108126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108126,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108126, "pid": 5, "tid": 7, "ts": 1716454223056897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988025, "dur": 11,
    "args": {
      "External id": 108126,
      "cbid": 211, "correlation": 108126
    }
  },
  {
    "ph": "s", "id": 108126, "pid": 76337, "tid": -914061504, "ts": 1716454222988025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988116, "dur": 1,
    "args": {
      "External id": 108142,
      "cbid": 251, "correlation": 108142
    }
  },
  {
    "ph": "f", "id": 108142, "pid": 76337, "tid": -914061504, "ts": 1716454222988116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988121, "dur": 0,
    "args": {
      "External id": 108144,
      "cbid": 251, "correlation": 108144
    }
  },
  {
    "ph": "f", "id": 108144, "pid": 76337, "tid": -914061504, "ts": 1716454222988121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223056932, "dur": 531,
    "args": {
      "External id": 108145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108145,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 108145, "pid": 5, "tid": 7, "ts": 1716454223056932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988126, "dur": 13,
    "args": {
      "External id": 108145,
      "cbid": 211, "correlation": 108145
    }
  },
  {
    "ph": "s", "id": 108145, "pid": 76337, "tid": -914061504, "ts": 1716454222988126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223057465, "dur": 123,
    "args": {
      "External id": 108153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108153,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108153, "pid": 5, "tid": 7, "ts": 1716454223057465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988192, "dur": 12,
    "args": {
      "External id": 108153,
      "cbid": 211, "correlation": 108153
    }
  },
  {
    "ph": "s", "id": 108153, "pid": 76337, "tid": -914061504, "ts": 1716454222988192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223057589, "dur": 128,
    "args": {
      "External id": 108161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108161, "pid": 5, "tid": 7, "ts": 1716454223057589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988222, "dur": 9,
    "args": {
      "External id": 108161,
      "cbid": 211, "correlation": 108161
    }
  },
  {
    "ph": "s", "id": 108161, "pid": 76337, "tid": -914061504, "ts": 1716454222988222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988300, "dur": 1,
    "args": {
      "External id": 108177,
      "cbid": 251, "correlation": 108177
    }
  },
  {
    "ph": "f", "id": 108177, "pid": 76337, "tid": -914061504, "ts": 1716454222988300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223057718, "dur": 301,
    "args": {
      "External id": 108179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108179,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108179, "pid": 5, "tid": 7, "ts": 1716454223057718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988305, "dur": 12,
    "args": {
      "External id": 108179,
      "cbid": 211, "correlation": 108179
    }
  },
  {
    "ph": "s", "id": 108179, "pid": 76337, "tid": -914061504, "ts": 1716454222988305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223058021, "dur": 27,
    "args": {
      "External id": 108187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108187,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108187, "pid": 5, "tid": 7, "ts": 1716454223058021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988348, "dur": 10,
    "args": {
      "External id": 108187,
      "cbid": 211, "correlation": 108187
    }
  },
  {
    "ph": "s", "id": 108187, "pid": 76337, "tid": -914061504, "ts": 1716454222988348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223058049, "dur": 79,
    "args": {
      "External id": 108198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108198, "pid": 5, "tid": 7, "ts": 1716454223058049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988415, "dur": 12,
    "args": {
      "External id": 108198,
      "cbid": 211, "correlation": 108198
    }
  },
  {
    "ph": "s", "id": 108198, "pid": 76337, "tid": -914061504, "ts": 1716454222988415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988479, "dur": 0,
    "args": {
      "External id": 108210,
      "cbid": 317, "correlation": 108210
    }
  },
  {
    "ph": "f", "id": 108210, "pid": 76337, "tid": -914061504, "ts": 1716454222988479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988480, "dur": 0,
    "args": {
      "External id": 108211,
      "cbid": 203, "correlation": 108211
    }
  },
  {
    "ph": "f", "id": 108211, "pid": 76337, "tid": -914061504, "ts": 1716454222988480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988481, "dur": 0,
    "args": {
      "External id": 108212,
      "cbid": 205, "correlation": 108212
    }
  },
  {
    "ph": "f", "id": 108212, "pid": 76337, "tid": -914061504, "ts": 1716454222988481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223058130, "dur": 22,
    "args": {
      "External id": 108216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108216,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108216, "pid": 5, "tid": 7, "ts": 1716454223058130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988496, "dur": 12,
    "args": {
      "External id": 108216,
      "cbid": 211, "correlation": 108216
    }
  },
  {
    "ph": "s", "id": 108216, "pid": 76337, "tid": -914061504, "ts": 1716454222988496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223058153, "dur": 117,
    "args": {
      "External id": 108218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108218,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108218, "pid": 5, "tid": 7, "ts": 1716454223058153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988514, "dur": 6,
    "args": {
      "External id": 108218,
      "cbid": 211, "correlation": 108218
    }
  },
  {
    "ph": "s", "id": 108218, "pid": 76337, "tid": -914061504, "ts": 1716454222988514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223058271, "dur": 23,
    "args": {
      "External id": 108220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108220, "pid": 5, "tid": 7, "ts": 1716454223058271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988525, "dur": 5,
    "args": {
      "External id": 108220,
      "cbid": 211, "correlation": 108220
    }
  },
  {
    "ph": "s", "id": 108220, "pid": 76337, "tid": -914061504, "ts": 1716454222988525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223058296, "dur": 33,
    "args": {
      "External id": 108226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108226, "pid": 5, "tid": 7, "ts": 1716454223058296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988553, "dur": 9,
    "args": {
      "External id": 108226,
      "cbid": 211, "correlation": 108226
    }
  },
  {
    "ph": "s", "id": 108226, "pid": 76337, "tid": -914061504, "ts": 1716454222988553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223058330, "dur": 27,
    "args": {
      "External id": 108234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108234,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108234, "pid": 5, "tid": 7, "ts": 1716454223058330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988586, "dur": 8,
    "args": {
      "External id": 108234,
      "cbid": 211, "correlation": 108234
    }
  },
  {
    "ph": "s", "id": 108234, "pid": 76337, "tid": -914061504, "ts": 1716454222988586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223058358, "dur": 53,
    "args": {
      "External id": 108243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108243,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108243, "pid": 5, "tid": 7, "ts": 1716454223058358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988624, "dur": 11,
    "args": {
      "External id": 108243,
      "cbid": 211, "correlation": 108243
    }
  },
  {
    "ph": "s", "id": 108243, "pid": 76337, "tid": -914061504, "ts": 1716454222988624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223058412, "dur": 52,
    "args": {
      "External id": 108263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108263,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 108263, "pid": 5, "tid": 7, "ts": 1716454223058412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988696, "dur": 11,
    "args": {
      "External id": 108263,
      "cbid": 211, "correlation": 108263
    }
  },
  {
    "ph": "s", "id": 108263, "pid": 76337, "tid": -914061504, "ts": 1716454222988696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223058466, "dur": 5,
    "args": {
      "External id": 108275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108275,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 108275, "pid": 5, "tid": 7, "ts": 1716454223058466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988717, "dur": 6,
    "args": {
      "External id": 108275,
      "cbid": 211, "correlation": 108275
    }
  },
  {
    "ph": "s", "id": 108275, "pid": 76337, "tid": -914061504, "ts": 1716454222988717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223058472, "dur": 56,
    "args": {
      "External id": 108278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108278, "pid": 5, "tid": 7, "ts": 1716454223058472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988735, "dur": 7,
    "args": {
      "External id": 108278,
      "cbid": 211, "correlation": 108278
    }
  },
  {
    "ph": "s", "id": 108278, "pid": 76337, "tid": -914061504, "ts": 1716454222988735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223058530, "dur": 36,
    "args": {
      "External id": 108287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108287,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108287, "pid": 5, "tid": 7, "ts": 1716454223058530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988773, "dur": 10,
    "args": {
      "External id": 108287,
      "cbid": 211, "correlation": 108287
    }
  },
  {
    "ph": "s", "id": 108287, "pid": 76337, "tid": -914061504, "ts": 1716454222988773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988825, "dur": 0,
    "args": {
      "External id": 108297,
      "cbid": 317, "correlation": 108297
    }
  },
  {
    "ph": "f", "id": 108297, "pid": 76337, "tid": -914061504, "ts": 1716454222988825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988826, "dur": 0,
    "args": {
      "External id": 108298,
      "cbid": 203, "correlation": 108298
    }
  },
  {
    "ph": "f", "id": 108298, "pid": 76337, "tid": -914061504, "ts": 1716454222988826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988827, "dur": 0,
    "args": {
      "External id": 108299,
      "cbid": 205, "correlation": 108299
    }
  },
  {
    "ph": "f", "id": 108299, "pid": 76337, "tid": -914061504, "ts": 1716454222988827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223058567, "dur": 39,
    "args": {
      "External id": 108303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108303, "pid": 5, "tid": 7, "ts": 1716454223058567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988841, "dur": 11,
    "args": {
      "External id": 108303,
      "cbid": 211, "correlation": 108303
    }
  },
  {
    "ph": "s", "id": 108303, "pid": 76337, "tid": -914061504, "ts": 1716454222988841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223058608, "dur": 81,
    "args": {
      "External id": 108305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108305, "pid": 5, "tid": 7, "ts": 1716454223058608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988855, "dur": 5,
    "args": {
      "External id": 108305,
      "cbid": 211, "correlation": 108305
    }
  },
  {
    "ph": "s", "id": 108305, "pid": 76337, "tid": -914061504, "ts": 1716454222988855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223058691, "dur": 1267,
    "args": {
      "External id": 108307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108307,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108307, "pid": 5, "tid": 7, "ts": 1716454223058691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988866, "dur": 6,
    "args": {
      "External id": 108307,
      "cbid": 211, "correlation": 108307
    }
  },
  {
    "ph": "s", "id": 108307, "pid": 76337, "tid": -914061504, "ts": 1716454222988866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223059958, "dur": 22,
    "args": {
      "External id": 108309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108309, "pid": 5, "tid": 7, "ts": 1716454223059958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988876, "dur": 5,
    "args": {
      "External id": 108309,
      "cbid": 211, "correlation": 108309
    }
  },
  {
    "ph": "s", "id": 108309, "pid": 76337, "tid": -914061504, "ts": 1716454222988876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223059982, "dur": 32,
    "args": {
      "External id": 108315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108315, "pid": 5, "tid": 7, "ts": 1716454223059982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988904, "dur": 8,
    "args": {
      "External id": 108315,
      "cbid": 211, "correlation": 108315
    }
  },
  {
    "ph": "s", "id": 108315, "pid": 76337, "tid": -914061504, "ts": 1716454222988904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223060015, "dur": 3,
    "args": {
      "External id": 108323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108323,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 108323, "pid": 5, "tid": 7, "ts": 1716454223060015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222988946, "dur": 9,
    "args": {
      "External id": 108323,
      "cbid": 211, "correlation": 108323
    }
  },
  {
    "ph": "s", "id": 108323, "pid": 76337, "tid": -914061504, "ts": 1716454222988946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989017, "dur": 1,
    "args": {
      "External id": 108339,
      "cbid": 251, "correlation": 108339
    }
  },
  {
    "ph": "f", "id": 108339, "pid": 76337, "tid": -914061504, "ts": 1716454222989017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989022, "dur": 0,
    "args": {
      "External id": 108341,
      "cbid": 251, "correlation": 108341
    }
  },
  {
    "ph": "f", "id": 108341, "pid": 76337, "tid": -914061504, "ts": 1716454222989022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223060020, "dur": 12,
    "args": {
      "External id": 108342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108342,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 108342, "pid": 5, "tid": 7, "ts": 1716454223060020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989024, "dur": 12,
    "args": {
      "External id": 108342,
      "cbid": 211, "correlation": 108342
    }
  },
  {
    "ph": "s", "id": 108342, "pid": 76337, "tid": -914061504, "ts": 1716454222989024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223060033, "dur": 5,
    "args": {
      "External id": 108344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108344,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 108344, "pid": 5, "tid": 7, "ts": 1716454223060033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989038, "dur": 5,
    "args": {
      "External id": 108344,
      "cbid": 211, "correlation": 108344
    }
  },
  {
    "ph": "s", "id": 108344, "pid": 76337, "tid": -914061504, "ts": 1716454222989038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223060040, "dur": 30,
    "args": {
      "External id": 108354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108354, "pid": 5, "tid": 7, "ts": 1716454223060040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989095, "dur": 12,
    "args": {
      "External id": 108354,
      "cbid": 211, "correlation": 108354
    }
  },
  {
    "ph": "s", "id": 108354, "pid": 76337, "tid": -914061504, "ts": 1716454222989095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223060071, "dur": 30,
    "args": {
      "External id": 108374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108374,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 108374, "pid": 5, "tid": 7, "ts": 1716454223060071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989160, "dur": 11,
    "args": {
      "External id": 108374,
      "cbid": 211, "correlation": 108374
    }
  },
  {
    "ph": "s", "id": 108374, "pid": 76337, "tid": -914061504, "ts": 1716454222989160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223060102, "dur": 4,
    "args": {
      "External id": 108386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108386,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 108386, "pid": 5, "tid": 7, "ts": 1716454223060102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989180, "dur": 6,
    "args": {
      "External id": 108386,
      "cbid": 211, "correlation": 108386
    }
  },
  {
    "ph": "s", "id": 108386, "pid": 76337, "tid": -914061504, "ts": 1716454222989180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223060108, "dur": 30,
    "args": {
      "External id": 108389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108389, "pid": 5, "tid": 7, "ts": 1716454223060108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989199, "dur": 6,
    "args": {
      "External id": 108389,
      "cbid": 211, "correlation": 108389
    }
  },
  {
    "ph": "s", "id": 108389, "pid": 76337, "tid": -914061504, "ts": 1716454222989199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223060139, "dur": 21,
    "args": {
      "External id": 108398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108398,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108398, "pid": 5, "tid": 7, "ts": 1716454223060139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989240, "dur": 9,
    "args": {
      "External id": 108398,
      "cbid": 211, "correlation": 108398
    }
  },
  {
    "ph": "s", "id": 108398, "pid": 76337, "tid": -914061504, "ts": 1716454222989240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989302, "dur": 0,
    "args": {
      "External id": 108408,
      "cbid": 317, "correlation": 108408
    }
  },
  {
    "ph": "f", "id": 108408, "pid": 76337, "tid": -914061504, "ts": 1716454222989302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989303, "dur": 0,
    "args": {
      "External id": 108409,
      "cbid": 203, "correlation": 108409
    }
  },
  {
    "ph": "f", "id": 108409, "pid": 76337, "tid": -914061504, "ts": 1716454222989303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989303, "dur": 0,
    "args": {
      "External id": 108410,
      "cbid": 205, "correlation": 108410
    }
  },
  {
    "ph": "f", "id": 108410, "pid": 76337, "tid": -914061504, "ts": 1716454222989303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223060161, "dur": 22,
    "args": {
      "External id": 108414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108414,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108414, "pid": 5, "tid": 7, "ts": 1716454223060161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989317, "dur": 11,
    "args": {
      "External id": 108414,
      "cbid": 211, "correlation": 108414
    }
  },
  {
    "ph": "s", "id": 108414, "pid": 76337, "tid": -914061504, "ts": 1716454222989317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223060184, "dur": 43,
    "args": {
      "External id": 108416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108416,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108416, "pid": 5, "tid": 7, "ts": 1716454223060184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989331, "dur": 6,
    "args": {
      "External id": 108416,
      "cbid": 211, "correlation": 108416
    }
  },
  {
    "ph": "s", "id": 108416, "pid": 76337, "tid": -914061504, "ts": 1716454222989331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223060229, "dur": 638,
    "args": {
      "External id": 108418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108418,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108418, "pid": 5, "tid": 7, "ts": 1716454223060229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989344, "dur": 6,
    "args": {
      "External id": 108418,
      "cbid": 211, "correlation": 108418
    }
  },
  {
    "ph": "s", "id": 108418, "pid": 76337, "tid": -914061504, "ts": 1716454222989344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223060868, "dur": 21,
    "args": {
      "External id": 108420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108420, "pid": 5, "tid": 7, "ts": 1716454223060868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989353, "dur": 5,
    "args": {
      "External id": 108420,
      "cbid": 211, "correlation": 108420
    }
  },
  {
    "ph": "s", "id": 108420, "pid": 76337, "tid": -914061504, "ts": 1716454222989353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223060891, "dur": 32,
    "args": {
      "External id": 108426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108426, "pid": 5, "tid": 7, "ts": 1716454223060891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989380, "dur": 10,
    "args": {
      "External id": 108426,
      "cbid": 211, "correlation": 108426
    }
  },
  {
    "ph": "s", "id": 108426, "pid": 76337, "tid": -914061504, "ts": 1716454222989380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989440, "dur": 0,
    "args": {
      "External id": 108436,
      "cbid": 317, "correlation": 108436
    }
  },
  {
    "ph": "f", "id": 108436, "pid": 76337, "tid": -914061504, "ts": 1716454222989440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989441, "dur": 0,
    "args": {
      "External id": 108437,
      "cbid": 203, "correlation": 108437
    }
  },
  {
    "ph": "f", "id": 108437, "pid": 76337, "tid": -914061504, "ts": 1716454222989441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989442, "dur": 0,
    "args": {
      "External id": 108438,
      "cbid": 205, "correlation": 108438
    }
  },
  {
    "ph": "f", "id": 108438, "pid": 76337, "tid": -914061504, "ts": 1716454222989442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223060925, "dur": 38,
    "args": {
      "External id": 108442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108442,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108442, "pid": 5, "tid": 7, "ts": 1716454223060925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989453, "dur": 13,
    "args": {
      "External id": 108442,
      "cbid": 211, "correlation": 108442
    }
  },
  {
    "ph": "s", "id": 108442, "pid": 76337, "tid": -914061504, "ts": 1716454222989453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223060964, "dur": 186,
    "args": {
      "External id": 108444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108444,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108444, "pid": 5, "tid": 7, "ts": 1716454223060964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989472, "dur": 6,
    "args": {
      "External id": 108444,
      "cbid": 211, "correlation": 108444
    }
  },
  {
    "ph": "s", "id": 108444, "pid": 76337, "tid": -914061504, "ts": 1716454222989472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223061152, "dur": 21,
    "args": {
      "External id": 108446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108446, "pid": 5, "tid": 7, "ts": 1716454223061152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989481, "dur": 6,
    "args": {
      "External id": 108446,
      "cbid": 211, "correlation": 108446
    }
  },
  {
    "ph": "s", "id": 108446, "pid": 76337, "tid": -914061504, "ts": 1716454222989481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223061174, "dur": 32,
    "args": {
      "External id": 108452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108452, "pid": 5, "tid": 7, "ts": 1716454223061174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989508, "dur": 8,
    "args": {
      "External id": 108452,
      "cbid": 211, "correlation": 108452
    }
  },
  {
    "ph": "s", "id": 108452, "pid": 76337, "tid": -914061504, "ts": 1716454222989508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223061208, "dur": 27,
    "args": {
      "External id": 108460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108460,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108460, "pid": 5, "tid": 7, "ts": 1716454223061208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989537, "dur": 8,
    "args": {
      "External id": 108460,
      "cbid": 211, "correlation": 108460
    }
  },
  {
    "ph": "s", "id": 108460, "pid": 76337, "tid": -914061504, "ts": 1716454222989537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223061236, "dur": 20,
    "args": {
      "External id": 108468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108468,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108468, "pid": 5, "tid": 7, "ts": 1716454223061236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989565, "dur": 8,
    "args": {
      "External id": 108468,
      "cbid": 211, "correlation": 108468
    }
  },
  {
    "ph": "s", "id": 108468, "pid": 76337, "tid": -914061504, "ts": 1716454222989565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223061257, "dur": 30,
    "args": {
      "External id": 108488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108488,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 108488, "pid": 5, "tid": 7, "ts": 1716454223061257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989648, "dur": 12,
    "args": {
      "External id": 108488,
      "cbid": 211, "correlation": 108488
    }
  },
  {
    "ph": "s", "id": 108488, "pid": 76337, "tid": -914061504, "ts": 1716454222989648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223061288, "dur": 5,
    "args": {
      "External id": 108500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108500,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 108500, "pid": 5, "tid": 7, "ts": 1716454223061288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989670, "dur": 6,
    "args": {
      "External id": 108500,
      "cbid": 211, "correlation": 108500
    }
  },
  {
    "ph": "s", "id": 108500, "pid": 76337, "tid": -914061504, "ts": 1716454222989670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223061294, "dur": 29,
    "args": {
      "External id": 108503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108503,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108503, "pid": 5, "tid": 7, "ts": 1716454223061294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989688, "dur": 7,
    "args": {
      "External id": 108503,
      "cbid": 211, "correlation": 108503
    }
  },
  {
    "ph": "s", "id": 108503, "pid": 76337, "tid": -914061504, "ts": 1716454222989688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989744, "dur": 0,
    "args": {
      "External id": 108514,
      "cbid": 317, "correlation": 108514
    }
  },
  {
    "ph": "f", "id": 108514, "pid": 76337, "tid": -914061504, "ts": 1716454222989744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989745, "dur": 0,
    "args": {
      "External id": 108515,
      "cbid": 203, "correlation": 108515
    }
  },
  {
    "ph": "f", "id": 108515, "pid": 76337, "tid": -914061504, "ts": 1716454222989745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989745, "dur": 0,
    "args": {
      "External id": 108516,
      "cbid": 205, "correlation": 108516
    }
  },
  {
    "ph": "f", "id": 108516, "pid": 76337, "tid": -914061504, "ts": 1716454222989745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223061325, "dur": 23,
    "args": {
      "External id": 108520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108520,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108520, "pid": 5, "tid": 7, "ts": 1716454223061325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989758, "dur": 12,
    "args": {
      "External id": 108520,
      "cbid": 211, "correlation": 108520
    }
  },
  {
    "ph": "s", "id": 108520, "pid": 76337, "tid": -914061504, "ts": 1716454222989758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223061349, "dur": 103,
    "args": {
      "External id": 108522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108522,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108522, "pid": 5, "tid": 7, "ts": 1716454223061349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989775, "dur": 6,
    "args": {
      "External id": 108522,
      "cbid": 211, "correlation": 108522
    }
  },
  {
    "ph": "s", "id": 108522, "pid": 76337, "tid": -914061504, "ts": 1716454222989775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223061454, "dur": 23,
    "args": {
      "External id": 108524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108524, "pid": 5, "tid": 7, "ts": 1716454223061454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989785, "dur": 6,
    "args": {
      "External id": 108524,
      "cbid": 211, "correlation": 108524
    }
  },
  {
    "ph": "s", "id": 108524, "pid": 76337, "tid": -914061504, "ts": 1716454222989785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223061478, "dur": 33,
    "args": {
      "External id": 108530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108530, "pid": 5, "tid": 7, "ts": 1716454223061478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989813, "dur": 9,
    "args": {
      "External id": 108530,
      "cbid": 211, "correlation": 108530
    }
  },
  {
    "ph": "s", "id": 108530, "pid": 76337, "tid": -914061504, "ts": 1716454222989813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223061512, "dur": 176,
    "args": {
      "External id": 108539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108539,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108539, "pid": 5, "tid": 7, "ts": 1716454223061512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989897, "dur": 14,
    "args": {
      "External id": 108539,
      "cbid": 211, "correlation": 108539
    }
  },
  {
    "ph": "s", "id": 108539, "pid": 76337, "tid": -914061504, "ts": 1716454222989897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223061688, "dur": 64,
    "args": {
      "External id": 108561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108561,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108561, "pid": 5, "tid": 7, "ts": 1716454223061688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222989955, "dur": 10,
    "args": {
      "External id": 108561,
      "cbid": 211, "correlation": 108561
    }
  },
  {
    "ph": "s", "id": 108561, "pid": 76337, "tid": -914061504, "ts": 1716454222989955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990054, "dur": 1,
    "args": {
      "External id": 108572,
      "cbid": 251, "correlation": 108572
    }
  },
  {
    "ph": "f", "id": 108572, "pid": 76337, "tid": -914061504, "ts": 1716454222990054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223061754, "dur": 151,
    "args": {
      "External id": 108573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108573,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108573, "pid": 5, "tid": 7, "ts": 1716454223061754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990059, "dur": 13,
    "args": {
      "External id": 108573,
      "cbid": 211, "correlation": 108573
    }
  },
  {
    "ph": "s", "id": 108573, "pid": 76337, "tid": -914061504, "ts": 1716454222990059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990129, "dur": 1,
    "args": {
      "External id": 108584,
      "cbid": 251, "correlation": 108584
    }
  },
  {
    "ph": "f", "id": 108584, "pid": 76337, "tid": -914061504, "ts": 1716454222990129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223061906, "dur": 145,
    "args": {
      "External id": 108585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108585,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108585, "pid": 5, "tid": 7, "ts": 1716454223061906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990133, "dur": 12,
    "args": {
      "External id": 108585,
      "cbid": 211, "correlation": 108585
    }
  },
  {
    "ph": "s", "id": 108585, "pid": 76337, "tid": -914061504, "ts": 1716454222990133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990198, "dur": 1,
    "args": {
      "External id": 108596,
      "cbid": 251, "correlation": 108596
    }
  },
  {
    "ph": "f", "id": 108596, "pid": 76337, "tid": -914061504, "ts": 1716454222990198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223062053, "dur": 141,
    "args": {
      "External id": 108597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108597,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108597, "pid": 5, "tid": 7, "ts": 1716454223062053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990202, "dur": 11,
    "args": {
      "External id": 108597,
      "cbid": 211, "correlation": 108597
    }
  },
  {
    "ph": "s", "id": 108597, "pid": 76337, "tid": -914061504, "ts": 1716454222990202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223062196, "dur": 1906,
    "args": {
      "External id": 108618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108618,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 108618, "pid": 5, "tid": 7, "ts": 1716454223062196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990281, "dur": 14,
    "args": {
      "External id": 108618,
      "cbid": 211, "correlation": 108618
    }
  },
  {
    "ph": "s", "id": 108618, "pid": 76337, "tid": -914061504, "ts": 1716454222990281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990379, "dur": 1,
    "args": {
      "External id": 108636,
      "cbid": 251, "correlation": 108636
    }
  },
  {
    "ph": "f", "id": 108636, "pid": 76337, "tid": -914061504, "ts": 1716454222990379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223064103, "dur": 146,
    "args": {
      "External id": 108638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108638,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 108638, "pid": 5, "tid": 7, "ts": 1716454223064103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990385, "dur": 14,
    "args": {
      "External id": 108638,
      "cbid": 211, "correlation": 108638
    }
  },
  {
    "ph": "s", "id": 108638, "pid": 76337, "tid": -914061504, "ts": 1716454222990385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223064250, "dur": 36,
    "args": {
      "External id": 108646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108646,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108646, "pid": 5, "tid": 7, "ts": 1716454223064250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990456, "dur": 12,
    "args": {
      "External id": 108646,
      "cbid": 211, "correlation": 108646
    }
  },
  {
    "ph": "s", "id": 108646, "pid": 76337, "tid": -914061504, "ts": 1716454222990456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223064287, "dur": 50,
    "args": {
      "External id": 108654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108654,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108654, "pid": 5, "tid": 7, "ts": 1716454223064287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990495, "dur": 8,
    "args": {
      "External id": 108654,
      "cbid": 211, "correlation": 108654
    }
  },
  {
    "ph": "s", "id": 108654, "pid": 76337, "tid": -914061504, "ts": 1716454222990495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223064339, "dur": 29,
    "args": {
      "External id": 108665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108665, "pid": 5, "tid": 7, "ts": 1716454223064339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990564, "dur": 13,
    "args": {
      "External id": 108665,
      "cbid": 211, "correlation": 108665
    }
  },
  {
    "ph": "s", "id": 108665, "pid": 76337, "tid": -914061504, "ts": 1716454222990564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223064369, "dur": 34,
    "args": {
      "External id": 108687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108687,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108687, "pid": 5, "tid": 7, "ts": 1716454223064369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990597, "dur": 7,
    "args": {
      "External id": 108687,
      "cbid": 211, "correlation": 108687
    }
  },
  {
    "ph": "s", "id": 108687, "pid": 76337, "tid": -914061504, "ts": 1716454222990597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990679, "dur": 1,
    "args": {
      "External id": 108698,
      "cbid": 251, "correlation": 108698
    }
  },
  {
    "ph": "f", "id": 108698, "pid": 76337, "tid": -914061504, "ts": 1716454222990679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223064404, "dur": 88,
    "args": {
      "External id": 108699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108699,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108699, "pid": 5, "tid": 7, "ts": 1716454223064404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990684, "dur": 14,
    "args": {
      "External id": 108699,
      "cbid": 211, "correlation": 108699
    }
  },
  {
    "ph": "s", "id": 108699, "pid": 76337, "tid": -914061504, "ts": 1716454222990684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990753, "dur": 1,
    "args": {
      "External id": 108710,
      "cbid": 251, "correlation": 108710
    }
  },
  {
    "ph": "f", "id": 108710, "pid": 76337, "tid": -914061504, "ts": 1716454222990753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990757, "dur": 0,
    "args": {
      "External id": 108711,
      "cbid": 251, "correlation": 108711
    }
  },
  {
    "ph": "f", "id": 108711, "pid": 76337, "tid": -914061504, "ts": 1716454222990757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223064493, "dur": 11,
    "args": {
      "External id": 108712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108712,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 108712, "pid": 5, "tid": 7, "ts": 1716454223064493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990759, "dur": 12,
    "args": {
      "External id": 108712,
      "cbid": 211, "correlation": 108712
    }
  },
  {
    "ph": "s", "id": 108712, "pid": 76337, "tid": -914061504, "ts": 1716454222990759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223064506, "dur": 5,
    "args": {
      "External id": 108714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108714,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 108714, "pid": 5, "tid": 7, "ts": 1716454223064506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990772, "dur": 6,
    "args": {
      "External id": 108714,
      "cbid": 211, "correlation": 108714
    }
  },
  {
    "ph": "s", "id": 108714, "pid": 76337, "tid": -914061504, "ts": 1716454222990772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990830, "dur": 1,
    "args": {
      "External id": 108725,
      "cbid": 251, "correlation": 108725
    }
  },
  {
    "ph": "f", "id": 108725, "pid": 76337, "tid": -914061504, "ts": 1716454222990830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990834, "dur": 0,
    "args": {
      "External id": 108726,
      "cbid": 251, "correlation": 108726
    }
  },
  {
    "ph": "f", "id": 108726, "pid": 76337, "tid": -914061504, "ts": 1716454222990834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223064512, "dur": 7,
    "args": {
      "External id": 108727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108727,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 108727, "pid": 5, "tid": 7, "ts": 1716454223064512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990835, "dur": 12,
    "args": {
      "External id": 108727,
      "cbid": 211, "correlation": 108727
    }
  },
  {
    "ph": "s", "id": 108727, "pid": 76337, "tid": -914061504, "ts": 1716454222990835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223064521, "dur": 3,
    "args": {
      "External id": 108729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108729,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 108729, "pid": 5, "tid": 7, "ts": 1716454223064521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990849, "dur": 5,
    "args": {
      "External id": 108729,
      "cbid": 211, "correlation": 108729
    }
  },
  {
    "ph": "s", "id": 108729, "pid": 76337, "tid": -914061504, "ts": 1716454222990849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223064526, "dur": 89,
    "args": {
      "External id": 108750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108750,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 108750, "pid": 5, "tid": 7, "ts": 1716454223064526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222990921, "dur": 12,
    "args": {
      "External id": 108750,
      "cbid": 211, "correlation": 108750
    }
  },
  {
    "ph": "s", "id": 108750, "pid": 76337, "tid": -914061504, "ts": 1716454222990921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991026, "dur": 1,
    "args": {
      "External id": 108768,
      "cbid": 251, "correlation": 108768
    }
  },
  {
    "ph": "f", "id": 108768, "pid": 76337, "tid": -914061504, "ts": 1716454222991026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223064616, "dur": 82,
    "args": {
      "External id": 108770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108770,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108770, "pid": 5, "tid": 7, "ts": 1716454223064616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991032, "dur": 14,
    "args": {
      "External id": 108770,
      "cbid": 211, "correlation": 108770
    }
  },
  {
    "ph": "s", "id": 108770, "pid": 76337, "tid": -914061504, "ts": 1716454222991032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223064700, "dur": 20,
    "args": {
      "External id": 108778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108778,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108778, "pid": 5, "tid": 7, "ts": 1716454223064700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991103, "dur": 12,
    "args": {
      "External id": 108778,
      "cbid": 211, "correlation": 108778
    }
  },
  {
    "ph": "s", "id": 108778, "pid": 76337, "tid": -914061504, "ts": 1716454222991103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223064721, "dur": 37,
    "args": {
      "External id": 108786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108786, "pid": 5, "tid": 7, "ts": 1716454223064721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991146, "dur": 9,
    "args": {
      "External id": 108786,
      "cbid": 211, "correlation": 108786
    }
  },
  {
    "ph": "s", "id": 108786, "pid": 76337, "tid": -914061504, "ts": 1716454222991146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223064759, "dur": 34,
    "args": {
      "External id": 108808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108808,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108808, "pid": 5, "tid": 7, "ts": 1716454223064759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991198, "dur": 10,
    "args": {
      "External id": 108808,
      "cbid": 211, "correlation": 108808
    }
  },
  {
    "ph": "s", "id": 108808, "pid": 76337, "tid": -914061504, "ts": 1716454222991198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991288, "dur": 1,
    "args": {
      "External id": 108824,
      "cbid": 251, "correlation": 108824
    }
  },
  {
    "ph": "f", "id": 108824, "pid": 76337, "tid": -914061504, "ts": 1716454222991288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991293, "dur": 0,
    "args": {
      "External id": 108826,
      "cbid": 251, "correlation": 108826
    }
  },
  {
    "ph": "f", "id": 108826, "pid": 76337, "tid": -914061504, "ts": 1716454222991293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223064794, "dur": 528,
    "args": {
      "External id": 108827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108827,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 108827, "pid": 5, "tid": 7, "ts": 1716454223064794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991297, "dur": 13,
    "args": {
      "External id": 108827,
      "cbid": 211, "correlation": 108827
    }
  },
  {
    "ph": "s", "id": 108827, "pid": 76337, "tid": -914061504, "ts": 1716454222991297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223065324, "dur": 122,
    "args": {
      "External id": 108835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108835,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108835, "pid": 5, "tid": 7, "ts": 1716454223065324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991362, "dur": 12,
    "args": {
      "External id": 108835,
      "cbid": 211, "correlation": 108835
    }
  },
  {
    "ph": "s", "id": 108835, "pid": 76337, "tid": -914061504, "ts": 1716454222991362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223065447, "dur": 128,
    "args": {
      "External id": 108843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108843,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108843, "pid": 5, "tid": 7, "ts": 1716454223065447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991393, "dur": 8,
    "args": {
      "External id": 108843,
      "cbid": 211, "correlation": 108843
    }
  },
  {
    "ph": "s", "id": 108843, "pid": 76337, "tid": -914061504, "ts": 1716454222991393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991469, "dur": 1,
    "args": {
      "External id": 108859,
      "cbid": 251, "correlation": 108859
    }
  },
  {
    "ph": "f", "id": 108859, "pid": 76337, "tid": -914061504, "ts": 1716454222991469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223065576, "dur": 305,
    "args": {
      "External id": 108861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108861,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108861, "pid": 5, "tid": 7, "ts": 1716454223065576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991475, "dur": 13,
    "args": {
      "External id": 108861,
      "cbid": 211, "correlation": 108861
    }
  },
  {
    "ph": "s", "id": 108861, "pid": 76337, "tid": -914061504, "ts": 1716454222991475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223065882, "dur": 27,
    "args": {
      "External id": 108869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108869,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108869, "pid": 5, "tid": 7, "ts": 1716454223065882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991517, "dur": 10,
    "args": {
      "External id": 108869,
      "cbid": 211, "correlation": 108869
    }
  },
  {
    "ph": "s", "id": 108869, "pid": 76337, "tid": -914061504, "ts": 1716454222991517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223065910, "dur": 80,
    "args": {
      "External id": 108880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108880, "pid": 5, "tid": 7, "ts": 1716454223065910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991586, "dur": 13,
    "args": {
      "External id": 108880,
      "cbid": 211, "correlation": 108880
    }
  },
  {
    "ph": "s", "id": 108880, "pid": 76337, "tid": -914061504, "ts": 1716454222991586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991650, "dur": 0,
    "args": {
      "External id": 108892,
      "cbid": 317, "correlation": 108892
    }
  },
  {
    "ph": "f", "id": 108892, "pid": 76337, "tid": -914061504, "ts": 1716454222991650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991651, "dur": 0,
    "args": {
      "External id": 108893,
      "cbid": 203, "correlation": 108893
    }
  },
  {
    "ph": "f", "id": 108893, "pid": 76337, "tid": -914061504, "ts": 1716454222991651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991652, "dur": 0,
    "args": {
      "External id": 108894,
      "cbid": 205, "correlation": 108894
    }
  },
  {
    "ph": "f", "id": 108894, "pid": 76337, "tid": -914061504, "ts": 1716454222991652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223065991, "dur": 24,
    "args": {
      "External id": 108898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108898,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108898, "pid": 5, "tid": 7, "ts": 1716454223065991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991666, "dur": 12,
    "args": {
      "External id": 108898,
      "cbid": 211, "correlation": 108898
    }
  },
  {
    "ph": "s", "id": 108898, "pid": 76337, "tid": -914061504, "ts": 1716454222991666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223066016, "dur": 117,
    "args": {
      "External id": 108900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108900,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108900, "pid": 5, "tid": 7, "ts": 1716454223066016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991685, "dur": 7,
    "args": {
      "External id": 108900,
      "cbid": 211, "correlation": 108900
    }
  },
  {
    "ph": "s", "id": 108900, "pid": 76337, "tid": -914061504, "ts": 1716454222991685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223066135, "dur": 23,
    "args": {
      "External id": 108902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108902,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108902, "pid": 5, "tid": 7, "ts": 1716454223066135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991696, "dur": 5,
    "args": {
      "External id": 108902,
      "cbid": 211, "correlation": 108902
    }
  },
  {
    "ph": "s", "id": 108902, "pid": 76337, "tid": -914061504, "ts": 1716454222991696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223066160, "dur": 33,
    "args": {
      "External id": 108908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108908,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108908, "pid": 5, "tid": 7, "ts": 1716454223066160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991723, "dur": 8,
    "args": {
      "External id": 108908,
      "cbid": 211, "correlation": 108908
    }
  },
  {
    "ph": "s", "id": 108908, "pid": 76337, "tid": -914061504, "ts": 1716454222991723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223066194, "dur": 26,
    "args": {
      "External id": 108916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108916,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108916, "pid": 5, "tid": 7, "ts": 1716454223066194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991755, "dur": 8,
    "args": {
      "External id": 108916,
      "cbid": 211, "correlation": 108916
    }
  },
  {
    "ph": "s", "id": 108916, "pid": 76337, "tid": -914061504, "ts": 1716454222991755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223066222, "dur": 43,
    "args": {
      "External id": 108925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108925,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108925, "pid": 5, "tid": 7, "ts": 1716454223066222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991795, "dur": 10,
    "args": {
      "External id": 108925,
      "cbid": 211, "correlation": 108925
    }
  },
  {
    "ph": "s", "id": 108925, "pid": 76337, "tid": -914061504, "ts": 1716454222991795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223066266, "dur": 44,
    "args": {
      "External id": 108945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108945,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 108945, "pid": 5, "tid": 7, "ts": 1716454223066266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991865, "dur": 11,
    "args": {
      "External id": 108945,
      "cbid": 211, "correlation": 108945
    }
  },
  {
    "ph": "s", "id": 108945, "pid": 76337, "tid": -914061504, "ts": 1716454222991865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223066311, "dur": 5,
    "args": {
      "External id": 108957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108957,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 108957, "pid": 5, "tid": 7, "ts": 1716454223066311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991886, "dur": 7,
    "args": {
      "External id": 108957,
      "cbid": 211, "correlation": 108957
    }
  },
  {
    "ph": "s", "id": 108957, "pid": 76337, "tid": -914061504, "ts": 1716454222991886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223066317, "dur": 42,
    "args": {
      "External id": 108960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108960, "pid": 5, "tid": 7, "ts": 1716454223066317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991905, "dur": 7,
    "args": {
      "External id": 108960,
      "cbid": 211, "correlation": 108960
    }
  },
  {
    "ph": "s", "id": 108960, "pid": 76337, "tid": -914061504, "ts": 1716454222991905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223066361, "dur": 30,
    "args": {
      "External id": 108969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108969,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108969, "pid": 5, "tid": 7, "ts": 1716454223066361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222991945, "dur": 10,
    "args": {
      "External id": 108969,
      "cbid": 211, "correlation": 108969
    }
  },
  {
    "ph": "s", "id": 108969, "pid": 76337, "tid": -914061504, "ts": 1716454222991945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992005, "dur": 0,
    "args": {
      "External id": 108979,
      "cbid": 317, "correlation": 108979
    }
  },
  {
    "ph": "f", "id": 108979, "pid": 76337, "tid": -914061504, "ts": 1716454222992005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992006, "dur": 0,
    "args": {
      "External id": 108980,
      "cbid": 203, "correlation": 108980
    }
  },
  {
    "ph": "f", "id": 108980, "pid": 76337, "tid": -914061504, "ts": 1716454222992006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992007, "dur": 0,
    "args": {
      "External id": 108981,
      "cbid": 205, "correlation": 108981
    }
  },
  {
    "ph": "f", "id": 108981, "pid": 76337, "tid": -914061504, "ts": 1716454222992007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223066392, "dur": 30,
    "args": {
      "External id": 108985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108985,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108985, "pid": 5, "tid": 7, "ts": 1716454223066392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992022, "dur": 12,
    "args": {
      "External id": 108985,
      "cbid": 211, "correlation": 108985
    }
  },
  {
    "ph": "s", "id": 108985, "pid": 76337, "tid": -914061504, "ts": 1716454222992022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223066423, "dur": 62,
    "args": {
      "External id": 108987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108987, "pid": 5, "tid": 7, "ts": 1716454223066423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992036, "dur": 6,
    "args": {
      "External id": 108987,
      "cbid": 211, "correlation": 108987
    }
  },
  {
    "ph": "s", "id": 108987, "pid": 76337, "tid": -914061504, "ts": 1716454222992036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223066487, "dur": 958,
    "args": {
      "External id": 108989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108989,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 108989, "pid": 5, "tid": 7, "ts": 1716454223066487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992048, "dur": 6,
    "args": {
      "External id": 108989,
      "cbid": 211, "correlation": 108989
    }
  },
  {
    "ph": "s", "id": 108989, "pid": 76337, "tid": -914061504, "ts": 1716454222992048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223067446, "dur": 20,
    "args": {
      "External id": 108991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108991,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108991, "pid": 5, "tid": 7, "ts": 1716454223067446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992058, "dur": 5,
    "args": {
      "External id": 108991,
      "cbid": 211, "correlation": 108991
    }
  },
  {
    "ph": "s", "id": 108991, "pid": 76337, "tid": -914061504, "ts": 1716454222992058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223067468, "dur": 33,
    "args": {
      "External id": 108997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 108997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 108997, "pid": 5, "tid": 7, "ts": 1716454223067468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992085, "dur": 9,
    "args": {
      "External id": 108997,
      "cbid": 211, "correlation": 108997
    }
  },
  {
    "ph": "s", "id": 108997, "pid": 76337, "tid": -914061504, "ts": 1716454222992085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223067502, "dur": 3,
    "args": {
      "External id": 109005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109005, "pid": 5, "tid": 7, "ts": 1716454223067502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992129, "dur": 10,
    "args": {
      "External id": 109005,
      "cbid": 211, "correlation": 109005
    }
  },
  {
    "ph": "s", "id": 109005, "pid": 76337, "tid": -914061504, "ts": 1716454222992129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992194, "dur": 1,
    "args": {
      "External id": 109021,
      "cbid": 251, "correlation": 109021
    }
  },
  {
    "ph": "f", "id": 109021, "pid": 76337, "tid": -914061504, "ts": 1716454222992194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992200, "dur": 0,
    "args": {
      "External id": 109023,
      "cbid": 251, "correlation": 109023
    }
  },
  {
    "ph": "f", "id": 109023, "pid": 76337, "tid": -914061504, "ts": 1716454222992200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223067507, "dur": 12,
    "args": {
      "External id": 109024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109024,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 109024, "pid": 5, "tid": 7, "ts": 1716454223067507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992201, "dur": 11,
    "args": {
      "External id": 109024,
      "cbid": 211, "correlation": 109024
    }
  },
  {
    "ph": "s", "id": 109024, "pid": 76337, "tid": -914061504, "ts": 1716454222992201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223067521, "dur": 5,
    "args": {
      "External id": 109026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109026,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 109026, "pid": 5, "tid": 7, "ts": 1716454223067521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992214, "dur": 5,
    "args": {
      "External id": 109026,
      "cbid": 211, "correlation": 109026
    }
  },
  {
    "ph": "s", "id": 109026, "pid": 76337, "tid": -914061504, "ts": 1716454222992214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223067527, "dur": 28,
    "args": {
      "External id": 109036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109036, "pid": 5, "tid": 7, "ts": 1716454223067527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992272, "dur": 12,
    "args": {
      "External id": 109036,
      "cbid": 211, "correlation": 109036
    }
  },
  {
    "ph": "s", "id": 109036, "pid": 76337, "tid": -914061504, "ts": 1716454222992272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223067556, "dur": 31,
    "args": {
      "External id": 109056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109056,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 109056, "pid": 5, "tid": 7, "ts": 1716454223067556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992337, "dur": 11,
    "args": {
      "External id": 109056,
      "cbid": 211, "correlation": 109056
    }
  },
  {
    "ph": "s", "id": 109056, "pid": 76337, "tid": -914061504, "ts": 1716454222992337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223067589, "dur": 5,
    "args": {
      "External id": 109068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109068,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 109068, "pid": 5, "tid": 7, "ts": 1716454223067589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992357, "dur": 6,
    "args": {
      "External id": 109068,
      "cbid": 211, "correlation": 109068
    }
  },
  {
    "ph": "s", "id": 109068, "pid": 76337, "tid": -914061504, "ts": 1716454222992357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223067595, "dur": 30,
    "args": {
      "External id": 109071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109071,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109071, "pid": 5, "tid": 7, "ts": 1716454223067595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992375, "dur": 7,
    "args": {
      "External id": 109071,
      "cbid": 211, "correlation": 109071
    }
  },
  {
    "ph": "s", "id": 109071, "pid": 76337, "tid": -914061504, "ts": 1716454222992375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223067626, "dur": 21,
    "args": {
      "External id": 109080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109080,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109080, "pid": 5, "tid": 7, "ts": 1716454223067626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992416, "dur": 10,
    "args": {
      "External id": 109080,
      "cbid": 211, "correlation": 109080
    }
  },
  {
    "ph": "s", "id": 109080, "pid": 76337, "tid": -914061504, "ts": 1716454222992416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992479, "dur": 0,
    "args": {
      "External id": 109090,
      "cbid": 317, "correlation": 109090
    }
  },
  {
    "ph": "f", "id": 109090, "pid": 76337, "tid": -914061504, "ts": 1716454222992479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992480, "dur": 0,
    "args": {
      "External id": 109091,
      "cbid": 203, "correlation": 109091
    }
  },
  {
    "ph": "f", "id": 109091, "pid": 76337, "tid": -914061504, "ts": 1716454222992480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992480, "dur": 0,
    "args": {
      "External id": 109092,
      "cbid": 205, "correlation": 109092
    }
  },
  {
    "ph": "f", "id": 109092, "pid": 76337, "tid": -914061504, "ts": 1716454222992480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223067648, "dur": 23,
    "args": {
      "External id": 109096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109096, "pid": 5, "tid": 7, "ts": 1716454223067648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992495, "dur": 12,
    "args": {
      "External id": 109096,
      "cbid": 211, "correlation": 109096
    }
  },
  {
    "ph": "s", "id": 109096, "pid": 76337, "tid": -914061504, "ts": 1716454222992495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223067672, "dur": 43,
    "args": {
      "External id": 109098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109098, "pid": 5, "tid": 7, "ts": 1716454223067672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992509, "dur": 5,
    "args": {
      "External id": 109098,
      "cbid": 211, "correlation": 109098
    }
  },
  {
    "ph": "s", "id": 109098, "pid": 76337, "tid": -914061504, "ts": 1716454222992509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223067717, "dur": 638,
    "args": {
      "External id": 109100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109100,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109100, "pid": 5, "tid": 7, "ts": 1716454223067717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992520, "dur": 5,
    "args": {
      "External id": 109100,
      "cbid": 211, "correlation": 109100
    }
  },
  {
    "ph": "s", "id": 109100, "pid": 76337, "tid": -914061504, "ts": 1716454222992520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223068356, "dur": 23,
    "args": {
      "External id": 109102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109102,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109102, "pid": 5, "tid": 7, "ts": 1716454223068356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992530, "dur": 5,
    "args": {
      "External id": 109102,
      "cbid": 211, "correlation": 109102
    }
  },
  {
    "ph": "s", "id": 109102, "pid": 76337, "tid": -914061504, "ts": 1716454222992530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223068380, "dur": 32,
    "args": {
      "External id": 109108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109108, "pid": 5, "tid": 7, "ts": 1716454223068380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992557, "dur": 8,
    "args": {
      "External id": 109108,
      "cbid": 211, "correlation": 109108
    }
  },
  {
    "ph": "s", "id": 109108, "pid": 76337, "tid": -914061504, "ts": 1716454222992557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992614, "dur": 0,
    "args": {
      "External id": 109118,
      "cbid": 317, "correlation": 109118
    }
  },
  {
    "ph": "f", "id": 109118, "pid": 76337, "tid": -914061504, "ts": 1716454222992614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992615, "dur": 0,
    "args": {
      "External id": 109119,
      "cbid": 203, "correlation": 109119
    }
  },
  {
    "ph": "f", "id": 109119, "pid": 76337, "tid": -914061504, "ts": 1716454222992615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992616, "dur": 0,
    "args": {
      "External id": 109120,
      "cbid": 205, "correlation": 109120
    }
  },
  {
    "ph": "f", "id": 109120, "pid": 76337, "tid": -914061504, "ts": 1716454222992616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223068414, "dur": 30,
    "args": {
      "External id": 109124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109124, "pid": 5, "tid": 7, "ts": 1716454223068414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992630, "dur": 12,
    "args": {
      "External id": 109124,
      "cbid": 211, "correlation": 109124
    }
  },
  {
    "ph": "s", "id": 109124, "pid": 76337, "tid": -914061504, "ts": 1716454222992630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223068445, "dur": 150,
    "args": {
      "External id": 109126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109126,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109126, "pid": 5, "tid": 7, "ts": 1716454223068445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992648, "dur": 6,
    "args": {
      "External id": 109126,
      "cbid": 211, "correlation": 109126
    }
  },
  {
    "ph": "s", "id": 109126, "pid": 76337, "tid": -914061504, "ts": 1716454222992648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223068597, "dur": 24,
    "args": {
      "External id": 109128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109128,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109128, "pid": 5, "tid": 7, "ts": 1716454223068597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992657, "dur": 5,
    "args": {
      "External id": 109128,
      "cbid": 211, "correlation": 109128
    }
  },
  {
    "ph": "s", "id": 109128, "pid": 76337, "tid": -914061504, "ts": 1716454222992657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223068622, "dur": 32,
    "args": {
      "External id": 109134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109134, "pid": 5, "tid": 7, "ts": 1716454223068622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992683, "dur": 9,
    "args": {
      "External id": 109134,
      "cbid": 211, "correlation": 109134
    }
  },
  {
    "ph": "s", "id": 109134, "pid": 76337, "tid": -914061504, "ts": 1716454222992683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223068655, "dur": 27,
    "args": {
      "External id": 109142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109142,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109142, "pid": 5, "tid": 7, "ts": 1716454223068655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992712, "dur": 8,
    "args": {
      "External id": 109142,
      "cbid": 211, "correlation": 109142
    }
  },
  {
    "ph": "s", "id": 109142, "pid": 76337, "tid": -914061504, "ts": 1716454222992712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223068683, "dur": 19,
    "args": {
      "External id": 109150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109150,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109150, "pid": 5, "tid": 7, "ts": 1716454223068683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992741, "dur": 8,
    "args": {
      "External id": 109150,
      "cbid": 211, "correlation": 109150
    }
  },
  {
    "ph": "s", "id": 109150, "pid": 76337, "tid": -914061504, "ts": 1716454222992741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223068704, "dur": 30,
    "args": {
      "External id": 109170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109170,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 109170, "pid": 5, "tid": 7, "ts": 1716454223068704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992823, "dur": 13,
    "args": {
      "External id": 109170,
      "cbid": 211, "correlation": 109170
    }
  },
  {
    "ph": "s", "id": 109170, "pid": 76337, "tid": -914061504, "ts": 1716454222992823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223068735, "dur": 5,
    "args": {
      "External id": 109182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109182,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 109182, "pid": 5, "tid": 7, "ts": 1716454223068735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992847, "dur": 6,
    "args": {
      "External id": 109182,
      "cbid": 211, "correlation": 109182
    }
  },
  {
    "ph": "s", "id": 109182, "pid": 76337, "tid": -914061504, "ts": 1716454222992847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223068741, "dur": 30,
    "args": {
      "External id": 109185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109185, "pid": 5, "tid": 7, "ts": 1716454223068741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992864, "dur": 6,
    "args": {
      "External id": 109185,
      "cbid": 211, "correlation": 109185
    }
  },
  {
    "ph": "s", "id": 109185, "pid": 76337, "tid": -914061504, "ts": 1716454222992864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992921, "dur": 0,
    "args": {
      "External id": 109196,
      "cbid": 317, "correlation": 109196
    }
  },
  {
    "ph": "f", "id": 109196, "pid": 76337, "tid": -914061504, "ts": 1716454222992921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992922, "dur": 0,
    "args": {
      "External id": 109197,
      "cbid": 203, "correlation": 109197
    }
  },
  {
    "ph": "f", "id": 109197, "pid": 76337, "tid": -914061504, "ts": 1716454222992922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992922, "dur": 0,
    "args": {
      "External id": 109198,
      "cbid": 205, "correlation": 109198
    }
  },
  {
    "ph": "f", "id": 109198, "pid": 76337, "tid": -914061504, "ts": 1716454222992922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223068773, "dur": 21,
    "args": {
      "External id": 109202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109202, "pid": 5, "tid": 7, "ts": 1716454223068773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992935, "dur": 12,
    "args": {
      "External id": 109202,
      "cbid": 211, "correlation": 109202
    }
  },
  {
    "ph": "s", "id": 109202, "pid": 76337, "tid": -914061504, "ts": 1716454222992935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223068795, "dur": 102,
    "args": {
      "External id": 109204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109204,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109204, "pid": 5, "tid": 7, "ts": 1716454223068795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992953, "dur": 6,
    "args": {
      "External id": 109204,
      "cbid": 211, "correlation": 109204
    }
  },
  {
    "ph": "s", "id": 109204, "pid": 76337, "tid": -914061504, "ts": 1716454222992953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223068899, "dur": 22,
    "args": {
      "External id": 109206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109206,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109206, "pid": 5, "tid": 7, "ts": 1716454223068899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992963, "dur": 5,
    "args": {
      "External id": 109206,
      "cbid": 211, "correlation": 109206
    }
  },
  {
    "ph": "s", "id": 109206, "pid": 76337, "tid": -914061504, "ts": 1716454222992963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223068922, "dur": 32,
    "args": {
      "External id": 109212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109212, "pid": 5, "tid": 7, "ts": 1716454223068922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222992998, "dur": 9,
    "args": {
      "External id": 109212,
      "cbid": 211, "correlation": 109212
    }
  },
  {
    "ph": "s", "id": 109212, "pid": 76337, "tid": -914061504, "ts": 1716454222992998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223068955, "dur": 170,
    "args": {
      "External id": 109221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109221,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109221, "pid": 5, "tid": 7, "ts": 1716454223068955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993080, "dur": 14,
    "args": {
      "External id": 109221,
      "cbid": 211, "correlation": 109221
    }
  },
  {
    "ph": "s", "id": 109221, "pid": 76337, "tid": -914061504, "ts": 1716454222993080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223069127, "dur": 64,
    "args": {
      "External id": 109243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109243,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109243, "pid": 5, "tid": 7, "ts": 1716454223069127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993137, "dur": 11,
    "args": {
      "External id": 109243,
      "cbid": 211, "correlation": 109243
    }
  },
  {
    "ph": "s", "id": 109243, "pid": 76337, "tid": -914061504, "ts": 1716454222993137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993224, "dur": 1,
    "args": {
      "External id": 109254,
      "cbid": 251, "correlation": 109254
    }
  },
  {
    "ph": "f", "id": 109254, "pid": 76337, "tid": -914061504, "ts": 1716454222993224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223069193, "dur": 147,
    "args": {
      "External id": 109255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109255,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109255, "pid": 5, "tid": 7, "ts": 1716454223069193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993230, "dur": 13,
    "args": {
      "External id": 109255,
      "cbid": 211, "correlation": 109255
    }
  },
  {
    "ph": "s", "id": 109255, "pid": 76337, "tid": -914061504, "ts": 1716454222993230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993300, "dur": 1,
    "args": {
      "External id": 109266,
      "cbid": 251, "correlation": 109266
    }
  },
  {
    "ph": "f", "id": 109266, "pid": 76337, "tid": -914061504, "ts": 1716454222993300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223069341, "dur": 141,
    "args": {
      "External id": 109267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109267,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109267, "pid": 5, "tid": 7, "ts": 1716454223069341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993303, "dur": 11,
    "args": {
      "External id": 109267,
      "cbid": 211, "correlation": 109267
    }
  },
  {
    "ph": "s", "id": 109267, "pid": 76337, "tid": -914061504, "ts": 1716454222993303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993369, "dur": 1,
    "args": {
      "External id": 109278,
      "cbid": 251, "correlation": 109278
    }
  },
  {
    "ph": "f", "id": 109278, "pid": 76337, "tid": -914061504, "ts": 1716454222993369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223069483, "dur": 144,
    "args": {
      "External id": 109279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109279,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109279, "pid": 5, "tid": 7, "ts": 1716454223069483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993373, "dur": 11,
    "args": {
      "External id": 109279,
      "cbid": 211, "correlation": 109279
    }
  },
  {
    "ph": "s", "id": 109279, "pid": 76337, "tid": -914061504, "ts": 1716454222993373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223069628, "dur": 1910,
    "args": {
      "External id": 109300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109300,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 109300, "pid": 5, "tid": 7, "ts": 1716454223069628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993455, "dur": 12,
    "args": {
      "External id": 109300,
      "cbid": 211, "correlation": 109300
    }
  },
  {
    "ph": "s", "id": 109300, "pid": 76337, "tid": -914061504, "ts": 1716454222993455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993553, "dur": 1,
    "args": {
      "External id": 109318,
      "cbid": 251, "correlation": 109318
    }
  },
  {
    "ph": "f", "id": 109318, "pid": 76337, "tid": -914061504, "ts": 1716454222993553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223071540, "dur": 145,
    "args": {
      "External id": 109320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109320,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 109320, "pid": 5, "tid": 7, "ts": 1716454223071540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993559, "dur": 13,
    "args": {
      "External id": 109320,
      "cbid": 211, "correlation": 109320
    }
  },
  {
    "ph": "s", "id": 109320, "pid": 76337, "tid": -914061504, "ts": 1716454222993559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223071686, "dur": 35,
    "args": {
      "External id": 109328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109328,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109328, "pid": 5, "tid": 7, "ts": 1716454223071686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993629, "dur": 13,
    "args": {
      "External id": 109328,
      "cbid": 211, "correlation": 109328
    }
  },
  {
    "ph": "s", "id": 109328, "pid": 76337, "tid": -914061504, "ts": 1716454222993629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223071723, "dur": 50,
    "args": {
      "External id": 109336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109336,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109336, "pid": 5, "tid": 7, "ts": 1716454223071723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993668, "dur": 8,
    "args": {
      "External id": 109336,
      "cbid": 211, "correlation": 109336
    }
  },
  {
    "ph": "s", "id": 109336, "pid": 76337, "tid": -914061504, "ts": 1716454222993668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223071774, "dur": 30,
    "args": {
      "External id": 109347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109347, "pid": 5, "tid": 7, "ts": 1716454223071774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993739, "dur": 12,
    "args": {
      "External id": 109347,
      "cbid": 211, "correlation": 109347
    }
  },
  {
    "ph": "s", "id": 109347, "pid": 76337, "tid": -914061504, "ts": 1716454222993739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223071806, "dur": 34,
    "args": {
      "External id": 109369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109369,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109369, "pid": 5, "tid": 7, "ts": 1716454223071806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993769, "dur": 8,
    "args": {
      "External id": 109369,
      "cbid": 211, "correlation": 109369
    }
  },
  {
    "ph": "s", "id": 109369, "pid": 76337, "tid": -914061504, "ts": 1716454222993769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993855, "dur": 1,
    "args": {
      "External id": 109380,
      "cbid": 251, "correlation": 109380
    }
  },
  {
    "ph": "f", "id": 109380, "pid": 76337, "tid": -914061504, "ts": 1716454222993855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223071841, "dur": 89,
    "args": {
      "External id": 109381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109381,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109381, "pid": 5, "tid": 7, "ts": 1716454223071841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993860, "dur": 13,
    "args": {
      "External id": 109381,
      "cbid": 211, "correlation": 109381
    }
  },
  {
    "ph": "s", "id": 109381, "pid": 76337, "tid": -914061504, "ts": 1716454222993860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993929, "dur": 1,
    "args": {
      "External id": 109392,
      "cbid": 251, "correlation": 109392
    }
  },
  {
    "ph": "f", "id": 109392, "pid": 76337, "tid": -914061504, "ts": 1716454222993929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993933, "dur": 0,
    "args": {
      "External id": 109393,
      "cbid": 251, "correlation": 109393
    }
  },
  {
    "ph": "f", "id": 109393, "pid": 76337, "tid": -914061504, "ts": 1716454222993933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223071931, "dur": 11,
    "args": {
      "External id": 109394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109394,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 109394, "pid": 5, "tid": 7, "ts": 1716454223071931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993935, "dur": 12,
    "args": {
      "External id": 109394,
      "cbid": 211, "correlation": 109394
    }
  },
  {
    "ph": "s", "id": 109394, "pid": 76337, "tid": -914061504, "ts": 1716454222993935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223071944, "dur": 5,
    "args": {
      "External id": 109396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109396,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 109396, "pid": 5, "tid": 7, "ts": 1716454223071944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222993949, "dur": 6,
    "args": {
      "External id": 109396,
      "cbid": 211, "correlation": 109396
    }
  },
  {
    "ph": "s", "id": 109396, "pid": 76337, "tid": -914061504, "ts": 1716454222993949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994016, "dur": 1,
    "args": {
      "External id": 109407,
      "cbid": 251, "correlation": 109407
    }
  },
  {
    "ph": "f", "id": 109407, "pid": 76337, "tid": -914061504, "ts": 1716454222994016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994020, "dur": 0,
    "args": {
      "External id": 109408,
      "cbid": 251, "correlation": 109408
    }
  },
  {
    "ph": "f", "id": 109408, "pid": 76337, "tid": -914061504, "ts": 1716454222994020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223071950, "dur": 7,
    "args": {
      "External id": 109409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109409,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 109409, "pid": 5, "tid": 7, "ts": 1716454223071950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994022, "dur": 12,
    "args": {
      "External id": 109409,
      "cbid": 211, "correlation": 109409
    }
  },
  {
    "ph": "s", "id": 109409, "pid": 76337, "tid": -914061504, "ts": 1716454222994022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223071958, "dur": 4,
    "args": {
      "External id": 109411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109411,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 109411, "pid": 5, "tid": 7, "ts": 1716454223071958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994036, "dur": 6,
    "args": {
      "External id": 109411,
      "cbid": 211, "correlation": 109411
    }
  },
  {
    "ph": "s", "id": 109411, "pid": 76337, "tid": -914061504, "ts": 1716454222994036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223071963, "dur": 91,
    "args": {
      "External id": 109432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109432,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 109432, "pid": 5, "tid": 7, "ts": 1716454223071963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994109, "dur": 12,
    "args": {
      "External id": 109432,
      "cbid": 211, "correlation": 109432
    }
  },
  {
    "ph": "s", "id": 109432, "pid": 76337, "tid": -914061504, "ts": 1716454222994109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994206, "dur": 1,
    "args": {
      "External id": 109450,
      "cbid": 251, "correlation": 109450
    }
  },
  {
    "ph": "f", "id": 109450, "pid": 76337, "tid": -914061504, "ts": 1716454222994206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223072055, "dur": 97,
    "args": {
      "External id": 109452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109452,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109452, "pid": 5, "tid": 7, "ts": 1716454223072055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994211, "dur": 13,
    "args": {
      "External id": 109452,
      "cbid": 211, "correlation": 109452
    }
  },
  {
    "ph": "s", "id": 109452, "pid": 76337, "tid": -914061504, "ts": 1716454222994211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223072154, "dur": 19,
    "args": {
      "External id": 109460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109460,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109460, "pid": 5, "tid": 7, "ts": 1716454223072154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994280, "dur": 13,
    "args": {
      "External id": 109460,
      "cbid": 211, "correlation": 109460
    }
  },
  {
    "ph": "s", "id": 109460, "pid": 76337, "tid": -914061504, "ts": 1716454222994280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223072175, "dur": 37,
    "args": {
      "External id": 109468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109468,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109468, "pid": 5, "tid": 7, "ts": 1716454223072175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994321, "dur": 9,
    "args": {
      "External id": 109468,
      "cbid": 211, "correlation": 109468
    }
  },
  {
    "ph": "s", "id": 109468, "pid": 76337, "tid": -914061504, "ts": 1716454222994321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223072213, "dur": 33,
    "args": {
      "External id": 109490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109490,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109490, "pid": 5, "tid": 7, "ts": 1716454223072213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994372, "dur": 10,
    "args": {
      "External id": 109490,
      "cbid": 211, "correlation": 109490
    }
  },
  {
    "ph": "s", "id": 109490, "pid": 76337, "tid": -914061504, "ts": 1716454222994372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994460, "dur": 1,
    "args": {
      "External id": 109506,
      "cbid": 251, "correlation": 109506
    }
  },
  {
    "ph": "f", "id": 109506, "pid": 76337, "tid": -914061504, "ts": 1716454222994460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994465, "dur": 0,
    "args": {
      "External id": 109508,
      "cbid": 251, "correlation": 109508
    }
  },
  {
    "ph": "f", "id": 109508, "pid": 76337, "tid": -914061504, "ts": 1716454222994465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223072248, "dur": 532,
    "args": {
      "External id": 109509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109509,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 109509, "pid": 5, "tid": 7, "ts": 1716454223072248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994468, "dur": 13,
    "args": {
      "External id": 109509,
      "cbid": 211, "correlation": 109509
    }
  },
  {
    "ph": "s", "id": 109509, "pid": 76337, "tid": -914061504, "ts": 1716454222994468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223072781, "dur": 123,
    "args": {
      "External id": 109517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109517, "pid": 5, "tid": 7, "ts": 1716454223072781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994532, "dur": 13,
    "args": {
      "External id": 109517,
      "cbid": 211, "correlation": 109517
    }
  },
  {
    "ph": "s", "id": 109517, "pid": 76337, "tid": -914061504, "ts": 1716454222994532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223072906, "dur": 129,
    "args": {
      "External id": 109525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109525,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109525, "pid": 5, "tid": 7, "ts": 1716454223072906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994563, "dur": 8,
    "args": {
      "External id": 109525,
      "cbid": 211, "correlation": 109525
    }
  },
  {
    "ph": "s", "id": 109525, "pid": 76337, "tid": -914061504, "ts": 1716454222994563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994640, "dur": 1,
    "args": {
      "External id": 109541,
      "cbid": 251, "correlation": 109541
    }
  },
  {
    "ph": "f", "id": 109541, "pid": 76337, "tid": -914061504, "ts": 1716454222994640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223073036, "dur": 301,
    "args": {
      "External id": 109543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109543,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109543, "pid": 5, "tid": 7, "ts": 1716454223073036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994645, "dur": 12,
    "args": {
      "External id": 109543,
      "cbid": 211, "correlation": 109543
    }
  },
  {
    "ph": "s", "id": 109543, "pid": 76337, "tid": -914061504, "ts": 1716454222994645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223073339, "dur": 27,
    "args": {
      "External id": 109551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109551,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109551, "pid": 5, "tid": 7, "ts": 1716454223073339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994688, "dur": 10,
    "args": {
      "External id": 109551,
      "cbid": 211, "correlation": 109551
    }
  },
  {
    "ph": "s", "id": 109551, "pid": 76337, "tid": -914061504, "ts": 1716454222994688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223073367, "dur": 80,
    "args": {
      "External id": 109562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109562,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109562, "pid": 5, "tid": 7, "ts": 1716454223073367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994754, "dur": 12,
    "args": {
      "External id": 109562,
      "cbid": 211, "correlation": 109562
    }
  },
  {
    "ph": "s", "id": 109562, "pid": 76337, "tid": -914061504, "ts": 1716454222994754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994816, "dur": 0,
    "args": {
      "External id": 109574,
      "cbid": 317, "correlation": 109574
    }
  },
  {
    "ph": "f", "id": 109574, "pid": 76337, "tid": -914061504, "ts": 1716454222994816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994817, "dur": 0,
    "args": {
      "External id": 109575,
      "cbid": 203, "correlation": 109575
    }
  },
  {
    "ph": "f", "id": 109575, "pid": 76337, "tid": -914061504, "ts": 1716454222994817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994818, "dur": 0,
    "args": {
      "External id": 109576,
      "cbid": 205, "correlation": 109576
    }
  },
  {
    "ph": "f", "id": 109576, "pid": 76337, "tid": -914061504, "ts": 1716454222994818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223073449, "dur": 23,
    "args": {
      "External id": 109580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109580,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109580, "pid": 5, "tid": 7, "ts": 1716454223073449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994832, "dur": 13,
    "args": {
      "External id": 109580,
      "cbid": 211, "correlation": 109580
    }
  },
  {
    "ph": "s", "id": 109580, "pid": 76337, "tid": -914061504, "ts": 1716454222994832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223073473, "dur": 119,
    "args": {
      "External id": 109582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109582,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109582, "pid": 5, "tid": 7, "ts": 1716454223073473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994852, "dur": 6,
    "args": {
      "External id": 109582,
      "cbid": 211, "correlation": 109582
    }
  },
  {
    "ph": "s", "id": 109582, "pid": 76337, "tid": -914061504, "ts": 1716454222994852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223073593, "dur": 22,
    "args": {
      "External id": 109584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109584,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109584, "pid": 5, "tid": 7, "ts": 1716454223073593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994861, "dur": 5,
    "args": {
      "External id": 109584,
      "cbid": 211, "correlation": 109584
    }
  },
  {
    "ph": "s", "id": 109584, "pid": 76337, "tid": -914061504, "ts": 1716454222994861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223073616, "dur": 32,
    "args": {
      "External id": 109590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109590, "pid": 5, "tid": 7, "ts": 1716454223073616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994889, "dur": 9,
    "args": {
      "External id": 109590,
      "cbid": 211, "correlation": 109590
    }
  },
  {
    "ph": "s", "id": 109590, "pid": 76337, "tid": -914061504, "ts": 1716454222994889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223073650, "dur": 28,
    "args": {
      "External id": 109598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109598,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109598, "pid": 5, "tid": 7, "ts": 1716454223073650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994921, "dur": 8,
    "args": {
      "External id": 109598,
      "cbid": 211, "correlation": 109598
    }
  },
  {
    "ph": "s", "id": 109598, "pid": 76337, "tid": -914061504, "ts": 1716454222994921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223073678, "dur": 100,
    "args": {
      "External id": 109609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109609,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109609, "pid": 5, "tid": 7, "ts": 1716454223073678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222994991, "dur": 12,
    "args": {
      "External id": 109609,
      "cbid": 211, "correlation": 109609
    }
  },
  {
    "ph": "s", "id": 109609, "pid": 76337, "tid": -914061504, "ts": 1716454222994991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995047, "dur": 0,
    "args": {
      "External id": 109619,
      "cbid": 317, "correlation": 109619
    }
  },
  {
    "ph": "f", "id": 109619, "pid": 76337, "tid": -914061504, "ts": 1716454222995047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995048, "dur": 0,
    "args": {
      "External id": 109620,
      "cbid": 203, "correlation": 109620
    }
  },
  {
    "ph": "f", "id": 109620, "pid": 76337, "tid": -914061504, "ts": 1716454222995048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995049, "dur": 0,
    "args": {
      "External id": 109621,
      "cbid": 205, "correlation": 109621
    }
  },
  {
    "ph": "f", "id": 109621, "pid": 76337, "tid": -914061504, "ts": 1716454222995049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223073780, "dur": 75,
    "args": {
      "External id": 109625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109625, "pid": 5, "tid": 7, "ts": 1716454223073780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995062, "dur": 11,
    "args": {
      "External id": 109625,
      "cbid": 211, "correlation": 109625
    }
  },
  {
    "ph": "s", "id": 109625, "pid": 76337, "tid": -914061504, "ts": 1716454222995062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223073855, "dur": 43,
    "args": {
      "External id": 109627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109627, "pid": 5, "tid": 7, "ts": 1716454223073855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995076, "dur": 6,
    "args": {
      "External id": 109627,
      "cbid": 211, "correlation": 109627
    }
  },
  {
    "ph": "s", "id": 109627, "pid": 76337, "tid": -914061504, "ts": 1716454222995076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223073900, "dur": 4,
    "args": {
      "External id": 109629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109629,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109629, "pid": 5, "tid": 7, "ts": 1716454223073900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995087, "dur": 6,
    "args": {
      "External id": 109629,
      "cbid": 211, "correlation": 109629
    }
  },
  {
    "ph": "s", "id": 109629, "pid": 76337, "tid": -914061504, "ts": 1716454222995087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995095, "dur": 0,
    "args": {
      "External id": 109630,
      "cbid": 51, "correlation": 109630
    }
  },
  {
    "ph": "s", "id": 109630, "pid": 76337, "tid": -914061504, "ts": 1716454222995095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223073905, "dur": 2228,
    "args": {
      "External id": 109631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109631,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109631, "pid": 5, "tid": 7, "ts": 1716454223073905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995096, "dur": 5,
    "args": {
      "External id": 109631,
      "cbid": 211, "correlation": 109631
    }
  },
  {
    "ph": "s", "id": 109631, "pid": 76337, "tid": -914061504, "ts": 1716454222995096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223076134, "dur": 112,
    "args": {
      "External id": 109636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109636, "pid": 5, "tid": 7, "ts": 1716454223076134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995125, "dur": 9,
    "args": {
      "External id": 109636,
      "cbid": 211, "correlation": 109636
    }
  },
  {
    "ph": "s", "id": 109636, "pid": 76337, "tid": -914061504, "ts": 1716454222995125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223076246, "dur": 162,
    "args": {
      "External id": 109645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109645,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109645, "pid": 5, "tid": 7, "ts": 1716454223076246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995217, "dur": 13,
    "args": {
      "External id": 109645,
      "cbid": 211, "correlation": 109645
    }
  },
  {
    "ph": "s", "id": 109645, "pid": 76337, "tid": -914061504, "ts": 1716454222995217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223076409, "dur": 127,
    "args": {
      "External id": 109665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109665,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 109665, "pid": 5, "tid": 7, "ts": 1716454223076409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995288, "dur": 11,
    "args": {
      "External id": 109665,
      "cbid": 211, "correlation": 109665
    }
  },
  {
    "ph": "s", "id": 109665, "pid": 76337, "tid": -914061504, "ts": 1716454222995288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223076537, "dur": 5,
    "args": {
      "External id": 109677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109677,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 109677, "pid": 5, "tid": 7, "ts": 1716454223076537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995309, "dur": 6,
    "args": {
      "External id": 109677,
      "cbid": 211, "correlation": 109677
    }
  },
  {
    "ph": "s", "id": 109677, "pid": 76337, "tid": -914061504, "ts": 1716454222995309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223076543, "dur": 159,
    "args": {
      "External id": 109680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109680, "pid": 5, "tid": 7, "ts": 1716454223076543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995327, "dur": 7,
    "args": {
      "External id": 109680,
      "cbid": 211, "correlation": 109680
    }
  },
  {
    "ph": "s", "id": 109680, "pid": 76337, "tid": -914061504, "ts": 1716454222995327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223076703, "dur": 101,
    "args": {
      "External id": 109689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109689,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109689, "pid": 5, "tid": 7, "ts": 1716454223076703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995367, "dur": 9,
    "args": {
      "External id": 109689,
      "cbid": 211, "correlation": 109689
    }
  },
  {
    "ph": "s", "id": 109689, "pid": 76337, "tid": -914061504, "ts": 1716454222995367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995420, "dur": 0,
    "args": {
      "External id": 109699,
      "cbid": 317, "correlation": 109699
    }
  },
  {
    "ph": "f", "id": 109699, "pid": 76337, "tid": -914061504, "ts": 1716454222995420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995421, "dur": 0,
    "args": {
      "External id": 109700,
      "cbid": 203, "correlation": 109700
    }
  },
  {
    "ph": "f", "id": 109700, "pid": 76337, "tid": -914061504, "ts": 1716454222995421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995422, "dur": 0,
    "args": {
      "External id": 109701,
      "cbid": 205, "correlation": 109701
    }
  },
  {
    "ph": "f", "id": 109701, "pid": 76337, "tid": -914061504, "ts": 1716454222995422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223076806, "dur": 111,
    "args": {
      "External id": 109705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109705,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109705, "pid": 5, "tid": 7, "ts": 1716454223076806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995439, "dur": 11,
    "args": {
      "External id": 109705,
      "cbid": 211, "correlation": 109705
    }
  },
  {
    "ph": "s", "id": 109705, "pid": 76337, "tid": -914061504, "ts": 1716454222995439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223076918, "dur": 34,
    "args": {
      "External id": 109707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109707,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109707, "pid": 5, "tid": 7, "ts": 1716454223076918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995452, "dur": 5,
    "args": {
      "External id": 109707,
      "cbid": 211, "correlation": 109707
    }
  },
  {
    "ph": "s", "id": 109707, "pid": 76337, "tid": -914061504, "ts": 1716454222995452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223076953, "dur": 4,
    "args": {
      "External id": 109709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109709, "pid": 5, "tid": 7, "ts": 1716454223076953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995462, "dur": 5,
    "args": {
      "External id": 109709,
      "cbid": 211, "correlation": 109709
    }
  },
  {
    "ph": "s", "id": 109709, "pid": 76337, "tid": -914061504, "ts": 1716454222995462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995471, "dur": 0,
    "args": {
      "External id": 109710,
      "cbid": 51, "correlation": 109710
    }
  },
  {
    "ph": "s", "id": 109710, "pid": 76337, "tid": -914061504, "ts": 1716454222995471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223076958, "dur": 1985,
    "args": {
      "External id": 109711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109711,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109711, "pid": 5, "tid": 7, "ts": 1716454223076958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995472, "dur": 5,
    "args": {
      "External id": 109711,
      "cbid": 211, "correlation": 109711
    }
  },
  {
    "ph": "s", "id": 109711, "pid": 76337, "tid": -914061504, "ts": 1716454222995472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223078944, "dur": 59,
    "args": {
      "External id": 109716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109716, "pid": 5, "tid": 7, "ts": 1716454223078944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995501, "dur": 8,
    "args": {
      "External id": 109716,
      "cbid": 211, "correlation": 109716
    }
  },
  {
    "ph": "s", "id": 109716, "pid": 76337, "tid": -914061504, "ts": 1716454222995501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223079004, "dur": 3,
    "args": {
      "External id": 109724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109724,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109724, "pid": 5, "tid": 7, "ts": 1716454223079004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995544, "dur": 9,
    "args": {
      "External id": 109724,
      "cbid": 211, "correlation": 109724
    }
  },
  {
    "ph": "s", "id": 109724, "pid": 76337, "tid": -914061504, "ts": 1716454222995544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995609, "dur": 1,
    "args": {
      "External id": 109740,
      "cbid": 251, "correlation": 109740
    }
  },
  {
    "ph": "f", "id": 109740, "pid": 76337, "tid": -914061504, "ts": 1716454222995609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995614, "dur": 0,
    "args": {
      "External id": 109742,
      "cbid": 251, "correlation": 109742
    }
  },
  {
    "ph": "f", "id": 109742, "pid": 76337, "tid": -914061504, "ts": 1716454222995614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223079009, "dur": 11,
    "args": {
      "External id": 109743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109743,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 109743, "pid": 5, "tid": 7, "ts": 1716454223079009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995616, "dur": 11,
    "args": {
      "External id": 109743,
      "cbid": 211, "correlation": 109743
    }
  },
  {
    "ph": "s", "id": 109743, "pid": 76337, "tid": -914061504, "ts": 1716454222995616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223079021, "dur": 5,
    "args": {
      "External id": 109745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109745,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 109745, "pid": 5, "tid": 7, "ts": 1716454223079021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995629, "dur": 6,
    "args": {
      "External id": 109745,
      "cbid": 211, "correlation": 109745
    }
  },
  {
    "ph": "s", "id": 109745, "pid": 76337, "tid": -914061504, "ts": 1716454222995629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223079027, "dur": 54,
    "args": {
      "External id": 109755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109755,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109755, "pid": 5, "tid": 7, "ts": 1716454223079027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995688, "dur": 11,
    "args": {
      "External id": 109755,
      "cbid": 211, "correlation": 109755
    }
  },
  {
    "ph": "s", "id": 109755, "pid": 76337, "tid": -914061504, "ts": 1716454222995688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223079083, "dur": 50,
    "args": {
      "External id": 109775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109775,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 109775, "pid": 5, "tid": 7, "ts": 1716454223079083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995753, "dur": 11,
    "args": {
      "External id": 109775,
      "cbid": 211, "correlation": 109775
    }
  },
  {
    "ph": "s", "id": 109775, "pid": 76337, "tid": -914061504, "ts": 1716454222995753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223079135, "dur": 4,
    "args": {
      "External id": 109787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109787,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109787, "pid": 5, "tid": 7, "ts": 1716454223079135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995774, "dur": 6,
    "args": {
      "External id": 109787,
      "cbid": 211, "correlation": 109787
    }
  },
  {
    "ph": "s", "id": 109787, "pid": 76337, "tid": -914061504, "ts": 1716454222995774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223079140, "dur": 54,
    "args": {
      "External id": 109790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109790, "pid": 5, "tid": 7, "ts": 1716454223079140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995792, "dur": 6,
    "args": {
      "External id": 109790,
      "cbid": 211, "correlation": 109790
    }
  },
  {
    "ph": "s", "id": 109790, "pid": 76337, "tid": -914061504, "ts": 1716454222995792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223079195, "dur": 37,
    "args": {
      "External id": 109799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109799,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109799, "pid": 5, "tid": 7, "ts": 1716454223079195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995832, "dur": 11,
    "args": {
      "External id": 109799,
      "cbid": 211, "correlation": 109799
    }
  },
  {
    "ph": "s", "id": 109799, "pid": 76337, "tid": -914061504, "ts": 1716454222995832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995896, "dur": 0,
    "args": {
      "External id": 109809,
      "cbid": 317, "correlation": 109809
    }
  },
  {
    "ph": "f", "id": 109809, "pid": 76337, "tid": -914061504, "ts": 1716454222995896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995897, "dur": 0,
    "args": {
      "External id": 109810,
      "cbid": 203, "correlation": 109810
    }
  },
  {
    "ph": "f", "id": 109810, "pid": 76337, "tid": -914061504, "ts": 1716454222995897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995898, "dur": 0,
    "args": {
      "External id": 109811,
      "cbid": 205, "correlation": 109811
    }
  },
  {
    "ph": "f", "id": 109811, "pid": 76337, "tid": -914061504, "ts": 1716454222995898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223079233, "dur": 40,
    "args": {
      "External id": 109815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109815,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109815, "pid": 5, "tid": 7, "ts": 1716454223079233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995915, "dur": 12,
    "args": {
      "External id": 109815,
      "cbid": 211, "correlation": 109815
    }
  },
  {
    "ph": "s", "id": 109815, "pid": 76337, "tid": -914061504, "ts": 1716454222995915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223079274, "dur": 14,
    "args": {
      "External id": 109817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109817,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109817, "pid": 5, "tid": 7, "ts": 1716454223079274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995930, "dur": 5,
    "args": {
      "External id": 109817,
      "cbid": 211, "correlation": 109817
    }
  },
  {
    "ph": "s", "id": 109817, "pid": 76337, "tid": -914061504, "ts": 1716454222995930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223079290, "dur": 3,
    "args": {
      "External id": 109819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109819, "pid": 5, "tid": 7, "ts": 1716454223079290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995940, "dur": 6,
    "args": {
      "External id": 109819,
      "cbid": 211, "correlation": 109819
    }
  },
  {
    "ph": "s", "id": 109819, "pid": 76337, "tid": -914061504, "ts": 1716454222995940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995948, "dur": 0,
    "args": {
      "External id": 109820,
      "cbid": 51, "correlation": 109820
    }
  },
  {
    "ph": "s", "id": 109820, "pid": 76337, "tid": -914061504, "ts": 1716454222995948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223079294, "dur": 688,
    "args": {
      "External id": 109821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109821,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109821, "pid": 5, "tid": 7, "ts": 1716454223079294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995949, "dur": 5,
    "args": {
      "External id": 109821,
      "cbid": 211, "correlation": 109821
    }
  },
  {
    "ph": "s", "id": 109821, "pid": 76337, "tid": -914061504, "ts": 1716454222995949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223079984, "dur": 59,
    "args": {
      "External id": 109826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109826,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109826, "pid": 5, "tid": 7, "ts": 1716454223079984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222995985, "dur": 10,
    "args": {
      "External id": 109826,
      "cbid": 211, "correlation": 109826
    }
  },
  {
    "ph": "s", "id": 109826, "pid": 76337, "tid": -914061504, "ts": 1716454222995985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996044, "dur": 0,
    "args": {
      "External id": 109836,
      "cbid": 317, "correlation": 109836
    }
  },
  {
    "ph": "f", "id": 109836, "pid": 76337, "tid": -914061504, "ts": 1716454222996044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996045, "dur": 0,
    "args": {
      "External id": 109837,
      "cbid": 203, "correlation": 109837
    }
  },
  {
    "ph": "f", "id": 109837, "pid": 76337, "tid": -914061504, "ts": 1716454222996045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996046, "dur": 0,
    "args": {
      "External id": 109838,
      "cbid": 205, "correlation": 109838
    }
  },
  {
    "ph": "f", "id": 109838, "pid": 76337, "tid": -914061504, "ts": 1716454222996046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223080044, "dur": 3,
    "args": {
      "External id": 109842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109842, "pid": 5, "tid": 7, "ts": 1716454223080044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996062, "dur": 11,
    "args": {
      "External id": 109842,
      "cbid": 211, "correlation": 109842
    }
  },
  {
    "ph": "s", "id": 109842, "pid": 76337, "tid": -914061504, "ts": 1716454222996062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996077, "dur": 0,
    "args": {
      "External id": 109843,
      "cbid": 51, "correlation": 109843
    }
  },
  {
    "ph": "s", "id": 109843, "pid": 76337, "tid": -914061504, "ts": 1716454222996077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223080049, "dur": 262,
    "args": {
      "External id": 109844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109844,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109844, "pid": 5, "tid": 7, "ts": 1716454223080049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996079, "dur": 8,
    "args": {
      "External id": 109844,
      "cbid": 211, "correlation": 109844
    }
  },
  {
    "ph": "s", "id": 109844, "pid": 76337, "tid": -914061504, "ts": 1716454222996079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223080312, "dur": 59,
    "args": {
      "External id": 109849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109849, "pid": 5, "tid": 7, "ts": 1716454223080312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996108, "dur": 8,
    "args": {
      "External id": 109849,
      "cbid": 211, "correlation": 109849
    }
  },
  {
    "ph": "s", "id": 109849, "pid": 76337, "tid": -914061504, "ts": 1716454222996108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223080372, "dur": 50,
    "args": {
      "External id": 109857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109857,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109857, "pid": 5, "tid": 7, "ts": 1716454223080372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996137, "dur": 8,
    "args": {
      "External id": 109857,
      "cbid": 211, "correlation": 109857
    }
  },
  {
    "ph": "s", "id": 109857, "pid": 76337, "tid": -914061504, "ts": 1716454222996137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223080423, "dur": 35,
    "args": {
      "External id": 109865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109865,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109865, "pid": 5, "tid": 7, "ts": 1716454223080423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996165, "dur": 8,
    "args": {
      "External id": 109865,
      "cbid": 211, "correlation": 109865
    }
  },
  {
    "ph": "s", "id": 109865, "pid": 76337, "tid": -914061504, "ts": 1716454222996165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223080460, "dur": 50,
    "args": {
      "External id": 109885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109885,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 109885, "pid": 5, "tid": 7, "ts": 1716454223080460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996248, "dur": 12,
    "args": {
      "External id": 109885,
      "cbid": 211, "correlation": 109885
    }
  },
  {
    "ph": "s", "id": 109885, "pid": 76337, "tid": -914061504, "ts": 1716454222996248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223080511, "dur": 4,
    "args": {
      "External id": 109897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109897,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 109897, "pid": 5, "tid": 7, "ts": 1716454223080511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996270, "dur": 6,
    "args": {
      "External id": 109897,
      "cbid": 211, "correlation": 109897
    }
  },
  {
    "ph": "s", "id": 109897, "pid": 76337, "tid": -914061504, "ts": 1716454222996270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223080516, "dur": 55,
    "args": {
      "External id": 109900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109900, "pid": 5, "tid": 7, "ts": 1716454223080516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996288, "dur": 6,
    "args": {
      "External id": 109900,
      "cbid": 211, "correlation": 109900
    }
  },
  {
    "ph": "s", "id": 109900, "pid": 76337, "tid": -914061504, "ts": 1716454222996288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996345, "dur": 0,
    "args": {
      "External id": 109911,
      "cbid": 317, "correlation": 109911
    }
  },
  {
    "ph": "f", "id": 109911, "pid": 76337, "tid": -914061504, "ts": 1716454222996345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996345, "dur": 0,
    "args": {
      "External id": 109912,
      "cbid": 203, "correlation": 109912
    }
  },
  {
    "ph": "f", "id": 109912, "pid": 76337, "tid": -914061504, "ts": 1716454222996345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996346, "dur": 0,
    "args": {
      "External id": 109913,
      "cbid": 205, "correlation": 109913
    }
  },
  {
    "ph": "f", "id": 109913, "pid": 76337, "tid": -914061504, "ts": 1716454222996346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996375, "dur": 2,
    "args": {
      "External id": 109917,
      "cbid": 251, "correlation": 109917
    }
  },
  {
    "ph": "f", "id": 109917, "pid": 76337, "tid": -914061504, "ts": 1716454222996375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996378, "dur": 0,
    "args": {
      "External id": 109918,
      "cbid": 251, "correlation": 109918
    }
  },
  {
    "ph": "f", "id": 109918, "pid": 76337, "tid": -914061504, "ts": 1716454222996378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996380, "dur": 0,
    "args": {
      "External id": 109919,
      "cbid": 251, "correlation": 109919
    }
  },
  {
    "ph": "f", "id": 109919, "pid": 76337, "tid": -914061504, "ts": 1716454222996380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996381, "dur": 1,
    "args": {
      "External id": 109920,
      "cbid": 251, "correlation": 109920
    }
  },
  {
    "ph": "f", "id": 109920, "pid": 76337, "tid": -914061504, "ts": 1716454222996381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996383, "dur": 1,
    "args": {
      "External id": 109921,
      "cbid": 251, "correlation": 109921
    }
  },
  {
    "ph": "f", "id": 109921, "pid": 76337, "tid": -914061504, "ts": 1716454222996383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996385, "dur": 1,
    "args": {
      "External id": 109922,
      "cbid": 251, "correlation": 109922
    }
  },
  {
    "ph": "f", "id": 109922, "pid": 76337, "tid": -914061504, "ts": 1716454222996385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996386, "dur": 1,
    "args": {
      "External id": 109923,
      "cbid": 251, "correlation": 109923
    }
  },
  {
    "ph": "f", "id": 109923, "pid": 76337, "tid": -914061504, "ts": 1716454222996386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996389, "dur": 1,
    "args": {
      "External id": 109924,
      "cbid": 251, "correlation": 109924
    }
  },
  {
    "ph": "f", "id": 109924, "pid": 76337, "tid": -914061504, "ts": 1716454222996389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996391, "dur": 0,
    "args": {
      "External id": 109925,
      "cbid": 251, "correlation": 109925
    }
  },
  {
    "ph": "f", "id": 109925, "pid": 76337, "tid": -914061504, "ts": 1716454222996391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223080572, "dur": 112,
    "args": {
      "External id": 109926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109926,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 109926, "pid": 5, "tid": 7, "ts": 1716454223080572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996394, "dur": 13,
    "args": {
      "External id": 109926,
      "cbid": 211, "correlation": 109926
    }
  },
  {
    "ph": "s", "id": 109926, "pid": 76337, "tid": -914061504, "ts": 1716454222996394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223080685, "dur": 59,
    "args": {
      "External id": 109932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109932, "pid": 5, "tid": 7, "ts": 1716454223080685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996429, "dur": 9,
    "args": {
      "External id": 109932,
      "cbid": 211, "correlation": 109932
    }
  },
  {
    "ph": "s", "id": 109932, "pid": 76337, "tid": -914061504, "ts": 1716454222996429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223080745, "dur": 494,
    "args": {
      "External id": 109941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109941,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109941, "pid": 5, "tid": 7, "ts": 1716454223080745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996512, "dur": 14,
    "args": {
      "External id": 109941,
      "cbid": 211, "correlation": 109941
    }
  },
  {
    "ph": "s", "id": 109941, "pid": 76337, "tid": -914061504, "ts": 1716454222996512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223081240, "dur": 178,
    "args": {
      "External id": 109963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109963,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 109963, "pid": 5, "tid": 7, "ts": 1716454223081240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996570, "dur": 11,
    "args": {
      "External id": 109963,
      "cbid": 211, "correlation": 109963
    }
  },
  {
    "ph": "s", "id": 109963, "pid": 76337, "tid": -914061504, "ts": 1716454222996570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996657, "dur": 1,
    "args": {
      "External id": 109974,
      "cbid": 251, "correlation": 109974
    }
  },
  {
    "ph": "f", "id": 109974, "pid": 76337, "tid": -914061504, "ts": 1716454222996657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223081419, "dur": 192,
    "args": {
      "External id": 109975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109975,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109975, "pid": 5, "tid": 7, "ts": 1716454223081419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996662, "dur": 13,
    "args": {
      "External id": 109975,
      "cbid": 211, "correlation": 109975
    }
  },
  {
    "ph": "s", "id": 109975, "pid": 76337, "tid": -914061504, "ts": 1716454222996662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996732, "dur": 1,
    "args": {
      "External id": 109986,
      "cbid": 251, "correlation": 109986
    }
  },
  {
    "ph": "f", "id": 109986, "pid": 76337, "tid": -914061504, "ts": 1716454222996732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223081612, "dur": 183,
    "args": {
      "External id": 109987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109987,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109987, "pid": 5, "tid": 7, "ts": 1716454223081612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996736, "dur": 11,
    "args": {
      "External id": 109987,
      "cbid": 211, "correlation": 109987
    }
  },
  {
    "ph": "s", "id": 109987, "pid": 76337, "tid": -914061504, "ts": 1716454222996736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996800, "dur": 1,
    "args": {
      "External id": 109998,
      "cbid": 251, "correlation": 109998
    }
  },
  {
    "ph": "f", "id": 109998, "pid": 76337, "tid": -914061504, "ts": 1716454222996800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223081797, "dur": 181,
    "args": {
      "External id": 109999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 109999,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 109999, "pid": 5, "tid": 7, "ts": 1716454223081797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996804, "dur": 11,
    "args": {
      "External id": 109999,
      "cbid": 211, "correlation": 109999
    }
  },
  {
    "ph": "s", "id": 109999, "pid": 76337, "tid": -914061504, "ts": 1716454222996804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223081979, "dur": 18291,
    "args": {
      "External id": 110020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110020,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110020, "pid": 5, "tid": 7, "ts": 1716454223081979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996888, "dur": 13,
    "args": {
      "External id": 110020,
      "cbid": 211, "correlation": 110020
    }
  },
  {
    "ph": "s", "id": 110020, "pid": 76337, "tid": -914061504, "ts": 1716454222996888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996993, "dur": 1,
    "args": {
      "External id": 110038,
      "cbid": 251, "correlation": 110038
    }
  },
  {
    "ph": "f", "id": 110038, "pid": 76337, "tid": -914061504, "ts": 1716454222996993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223100271, "dur": 203,
    "args": {
      "External id": 110040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110040,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110040, "pid": 5, "tid": 7, "ts": 1716454223100271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222996999, "dur": 14,
    "args": {
      "External id": 110040,
      "cbid": 211, "correlation": 110040
    }
  },
  {
    "ph": "s", "id": 110040, "pid": 76337, "tid": -914061504, "ts": 1716454222996999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223100476, "dur": 66,
    "args": {
      "External id": 110048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110048,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110048, "pid": 5, "tid": 7, "ts": 1716454223100476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997070, "dur": 12,
    "args": {
      "External id": 110048,
      "cbid": 211, "correlation": 110048
    }
  },
  {
    "ph": "s", "id": 110048, "pid": 76337, "tid": -914061504, "ts": 1716454222997070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223100544, "dur": 96,
    "args": {
      "External id": 110056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110056,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110056, "pid": 5, "tid": 7, "ts": 1716454223100544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997109, "dur": 9,
    "args": {
      "External id": 110056,
      "cbid": 211, "correlation": 110056
    }
  },
  {
    "ph": "s", "id": 110056, "pid": 76337, "tid": -914061504, "ts": 1716454222997109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223100641, "dur": 55,
    "args": {
      "External id": 110067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110067, "pid": 5, "tid": 7, "ts": 1716454223100641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997181, "dur": 12,
    "args": {
      "External id": 110067,
      "cbid": 211, "correlation": 110067
    }
  },
  {
    "ph": "s", "id": 110067, "pid": 76337, "tid": -914061504, "ts": 1716454222997181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223100697, "dur": 92,
    "args": {
      "External id": 110089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110089,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110089, "pid": 5, "tid": 7, "ts": 1716454223100697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997212, "dur": 7,
    "args": {
      "External id": 110089,
      "cbid": 211, "correlation": 110089
    }
  },
  {
    "ph": "s", "id": 110089, "pid": 76337, "tid": -914061504, "ts": 1716454222997212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997296, "dur": 1,
    "args": {
      "External id": 110100,
      "cbid": 251, "correlation": 110100
    }
  },
  {
    "ph": "f", "id": 110100, "pid": 76337, "tid": -914061504, "ts": 1716454222997296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223100790, "dur": 103,
    "args": {
      "External id": 110101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110101,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110101, "pid": 5, "tid": 7, "ts": 1716454223100790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997301, "dur": 12,
    "args": {
      "External id": 110101,
      "cbid": 211, "correlation": 110101
    }
  },
  {
    "ph": "s", "id": 110101, "pid": 76337, "tid": -914061504, "ts": 1716454222997301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997376, "dur": 1,
    "args": {
      "External id": 110112,
      "cbid": 251, "correlation": 110112
    }
  },
  {
    "ph": "f", "id": 110112, "pid": 76337, "tid": -914061504, "ts": 1716454222997376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997380, "dur": 0,
    "args": {
      "External id": 110113,
      "cbid": 251, "correlation": 110113
    }
  },
  {
    "ph": "f", "id": 110113, "pid": 76337, "tid": -914061504, "ts": 1716454222997380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223100895, "dur": 10,
    "args": {
      "External id": 110114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110114,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 110114, "pid": 5, "tid": 7, "ts": 1716454223100895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997382, "dur": 13,
    "args": {
      "External id": 110114,
      "cbid": 211, "correlation": 110114
    }
  },
  {
    "ph": "s", "id": 110114, "pid": 76337, "tid": -914061504, "ts": 1716454222997382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223100906, "dur": 5,
    "args": {
      "External id": 110116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110116,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 110116, "pid": 5, "tid": 7, "ts": 1716454223100906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997399, "dur": 6,
    "args": {
      "External id": 110116,
      "cbid": 211, "correlation": 110116
    }
  },
  {
    "ph": "s", "id": 110116, "pid": 76337, "tid": -914061504, "ts": 1716454222997399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997459, "dur": 1,
    "args": {
      "External id": 110127,
      "cbid": 251, "correlation": 110127
    }
  },
  {
    "ph": "f", "id": 110127, "pid": 76337, "tid": -914061504, "ts": 1716454222997459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997463, "dur": 0,
    "args": {
      "External id": 110128,
      "cbid": 251, "correlation": 110128
    }
  },
  {
    "ph": "f", "id": 110128, "pid": 76337, "tid": -914061504, "ts": 1716454222997463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223100912, "dur": 6,
    "args": {
      "External id": 110129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110129,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 110129, "pid": 5, "tid": 7, "ts": 1716454223100912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997464, "dur": 12,
    "args": {
      "External id": 110129,
      "cbid": 211, "correlation": 110129
    }
  },
  {
    "ph": "s", "id": 110129, "pid": 76337, "tid": -914061504, "ts": 1716454222997464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223100919, "dur": 3,
    "args": {
      "External id": 110131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110131,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 110131, "pid": 5, "tid": 7, "ts": 1716454223100919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997478, "dur": 5,
    "args": {
      "External id": 110131,
      "cbid": 211, "correlation": 110131
    }
  },
  {
    "ph": "s", "id": 110131, "pid": 76337, "tid": -914061504, "ts": 1716454222997478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223100924, "dur": 156,
    "args": {
      "External id": 110152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110152,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110152, "pid": 5, "tid": 7, "ts": 1716454223100924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997551, "dur": 12,
    "args": {
      "External id": 110152,
      "cbid": 211, "correlation": 110152
    }
  },
  {
    "ph": "s", "id": 110152, "pid": 76337, "tid": -914061504, "ts": 1716454222997551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997648, "dur": 1,
    "args": {
      "External id": 110170,
      "cbid": 251, "correlation": 110170
    }
  },
  {
    "ph": "f", "id": 110170, "pid": 76337, "tid": -914061504, "ts": 1716454222997648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223101082, "dur": 106,
    "args": {
      "External id": 110172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110172,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110172, "pid": 5, "tid": 7, "ts": 1716454223101082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997654, "dur": 14,
    "args": {
      "External id": 110172,
      "cbid": 211, "correlation": 110172
    }
  },
  {
    "ph": "s", "id": 110172, "pid": 76337, "tid": -914061504, "ts": 1716454222997654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223101189, "dur": 35,
    "args": {
      "External id": 110180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110180,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110180, "pid": 5, "tid": 7, "ts": 1716454223101189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997726, "dur": 12,
    "args": {
      "External id": 110180,
      "cbid": 211, "correlation": 110180
    }
  },
  {
    "ph": "s", "id": 110180, "pid": 76337, "tid": -914061504, "ts": 1716454222997726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223101225, "dur": 67,
    "args": {
      "External id": 110188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110188,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110188, "pid": 5, "tid": 7, "ts": 1716454223101225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997766, "dur": 10,
    "args": {
      "External id": 110188,
      "cbid": 211, "correlation": 110188
    }
  },
  {
    "ph": "s", "id": 110188, "pid": 76337, "tid": -914061504, "ts": 1716454222997766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223101293, "dur": 92,
    "args": {
      "External id": 110210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110210,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110210, "pid": 5, "tid": 7, "ts": 1716454223101293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997818, "dur": 11,
    "args": {
      "External id": 110210,
      "cbid": 211, "correlation": 110210
    }
  },
  {
    "ph": "s", "id": 110210, "pid": 76337, "tid": -914061504, "ts": 1716454222997818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997905, "dur": 1,
    "args": {
      "External id": 110226,
      "cbid": 251, "correlation": 110226
    }
  },
  {
    "ph": "f", "id": 110226, "pid": 76337, "tid": -914061504, "ts": 1716454222997905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223101386, "dur": 572,
    "args": {
      "External id": 110228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110228,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110228, "pid": 5, "tid": 7, "ts": 1716454223101386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997911, "dur": 12,
    "args": {
      "External id": 110228,
      "cbid": 211, "correlation": 110228
    }
  },
  {
    "ph": "s", "id": 110228, "pid": 76337, "tid": -914061504, "ts": 1716454222997911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223101960, "dur": 244,
    "args": {
      "External id": 110236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110236, "pid": 5, "tid": 7, "ts": 1716454223101960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222997984, "dur": 13,
    "args": {
      "External id": 110236,
      "cbid": 211, "correlation": 110236
    }
  },
  {
    "ph": "s", "id": 110236, "pid": 76337, "tid": -914061504, "ts": 1716454222997984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223102205, "dur": 251,
    "args": {
      "External id": 110244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110244, "pid": 5, "tid": 7, "ts": 1716454223102205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998017, "dur": 9,
    "args": {
      "External id": 110244,
      "cbid": 211, "correlation": 110244
    }
  },
  {
    "ph": "s", "id": 110244, "pid": 76337, "tid": -914061504, "ts": 1716454222998017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998101, "dur": 1,
    "args": {
      "External id": 110260,
      "cbid": 251, "correlation": 110260
    }
  },
  {
    "ph": "f", "id": 110260, "pid": 76337, "tid": -914061504, "ts": 1716454222998101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998106, "dur": 0,
    "args": {
      "External id": 110262,
      "cbid": 251, "correlation": 110262
    }
  },
  {
    "ph": "f", "id": 110262, "pid": 76337, "tid": -914061504, "ts": 1716454222998106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223102457, "dur": 355,
    "args": {
      "External id": 110263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110263,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 110263, "pid": 5, "tid": 7, "ts": 1716454223102457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998108, "dur": 12,
    "args": {
      "External id": 110263,
      "cbid": 211, "correlation": 110263
    }
  },
  {
    "ph": "s", "id": 110263, "pid": 76337, "tid": -914061504, "ts": 1716454222998108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223102813, "dur": 50,
    "args": {
      "External id": 110271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110271,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110271, "pid": 5, "tid": 7, "ts": 1716454223102813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998151, "dur": 9,
    "args": {
      "External id": 110271,
      "cbid": 211, "correlation": 110271
    }
  },
  {
    "ph": "s", "id": 110271, "pid": 76337, "tid": -914061504, "ts": 1716454222998151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223102865, "dur": 158,
    "args": {
      "External id": 110282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110282, "pid": 5, "tid": 7, "ts": 1716454223102865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998218, "dur": 13,
    "args": {
      "External id": 110282,
      "cbid": 211, "correlation": 110282
    }
  },
  {
    "ph": "s", "id": 110282, "pid": 76337, "tid": -914061504, "ts": 1716454222998218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998283, "dur": 0,
    "args": {
      "External id": 110294,
      "cbid": 317, "correlation": 110294
    }
  },
  {
    "ph": "f", "id": 110294, "pid": 76337, "tid": -914061504, "ts": 1716454222998283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998284, "dur": 0,
    "args": {
      "External id": 110295,
      "cbid": 203, "correlation": 110295
    }
  },
  {
    "ph": "f", "id": 110295, "pid": 76337, "tid": -914061504, "ts": 1716454222998284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998285, "dur": 0,
    "args": {
      "External id": 110296,
      "cbid": 205, "correlation": 110296
    }
  },
  {
    "ph": "f", "id": 110296, "pid": 76337, "tid": -914061504, "ts": 1716454222998285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998308, "dur": 1,
    "args": {
      "External id": 110300,
      "cbid": 251, "correlation": 110300
    }
  },
  {
    "ph": "f", "id": 110300, "pid": 76337, "tid": -914061504, "ts": 1716454222998308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998310, "dur": 0,
    "args": {
      "External id": 110301,
      "cbid": 251, "correlation": 110301
    }
  },
  {
    "ph": "f", "id": 110301, "pid": 76337, "tid": -914061504, "ts": 1716454222998310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998311, "dur": 0,
    "args": {
      "External id": 110302,
      "cbid": 251, "correlation": 110302
    }
  },
  {
    "ph": "f", "id": 110302, "pid": 76337, "tid": -914061504, "ts": 1716454222998311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998312, "dur": 0,
    "args": {
      "External id": 110303,
      "cbid": 251, "correlation": 110303
    }
  },
  {
    "ph": "f", "id": 110303, "pid": 76337, "tid": -914061504, "ts": 1716454222998312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998312, "dur": 0,
    "args": {
      "External id": 110304,
      "cbid": 251, "correlation": 110304
    }
  },
  {
    "ph": "f", "id": 110304, "pid": 76337, "tid": -914061504, "ts": 1716454222998312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998313, "dur": 0,
    "args": {
      "External id": 110305,
      "cbid": 251, "correlation": 110305
    }
  },
  {
    "ph": "f", "id": 110305, "pid": 76337, "tid": -914061504, "ts": 1716454222998313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998314, "dur": 0,
    "args": {
      "External id": 110306,
      "cbid": 251, "correlation": 110306
    }
  },
  {
    "ph": "f", "id": 110306, "pid": 76337, "tid": -914061504, "ts": 1716454222998314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998315, "dur": 0,
    "args": {
      "External id": 110307,
      "cbid": 251, "correlation": 110307
    }
  },
  {
    "ph": "f", "id": 110307, "pid": 76337, "tid": -914061504, "ts": 1716454222998315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998316, "dur": 0,
    "args": {
      "External id": 110308,
      "cbid": 251, "correlation": 110308
    }
  },
  {
    "ph": "f", "id": 110308, "pid": 76337, "tid": -914061504, "ts": 1716454222998316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223103024, "dur": 115,
    "args": {
      "External id": 110309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110309,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110309, "pid": 5, "tid": 7, "ts": 1716454223103024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998318, "dur": 13,
    "args": {
      "External id": 110309,
      "cbid": 211, "correlation": 110309
    }
  },
  {
    "ph": "s", "id": 110309, "pid": 76337, "tid": -914061504, "ts": 1716454222998318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223103140, "dur": 59,
    "args": {
      "External id": 110315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110315, "pid": 5, "tid": 7, "ts": 1716454223103140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998353, "dur": 9,
    "args": {
      "External id": 110315,
      "cbid": 211, "correlation": 110315
    }
  },
  {
    "ph": "s", "id": 110315, "pid": 76337, "tid": -914061504, "ts": 1716454222998353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223103201, "dur": 49,
    "args": {
      "External id": 110323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110323,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110323, "pid": 5, "tid": 7, "ts": 1716454223103201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998386, "dur": 8,
    "args": {
      "External id": 110323,
      "cbid": 211, "correlation": 110323
    }
  },
  {
    "ph": "s", "id": 110323, "pid": 76337, "tid": -914061504, "ts": 1716454222998386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223103252, "dur": 99,
    "args": {
      "External id": 110332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110332,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110332, "pid": 5, "tid": 7, "ts": 1716454223103252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998425, "dur": 10,
    "args": {
      "External id": 110332,
      "cbid": 211, "correlation": 110332
    }
  },
  {
    "ph": "s", "id": 110332, "pid": 76337, "tid": -914061504, "ts": 1716454222998425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223103352, "dur": 91,
    "args": {
      "External id": 110352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110352,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 110352, "pid": 5, "tid": 7, "ts": 1716454223103352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998496, "dur": 11,
    "args": {
      "External id": 110352,
      "cbid": 211, "correlation": 110352
    }
  },
  {
    "ph": "s", "id": 110352, "pid": 76337, "tid": -914061504, "ts": 1716454222998496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223103444, "dur": 5,
    "args": {
      "External id": 110364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110364,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 110364, "pid": 5, "tid": 7, "ts": 1716454223103444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998517, "dur": 7,
    "args": {
      "External id": 110364,
      "cbid": 211, "correlation": 110364
    }
  },
  {
    "ph": "s", "id": 110364, "pid": 76337, "tid": -914061504, "ts": 1716454222998517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223103451, "dur": 109,
    "args": {
      "External id": 110367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110367, "pid": 5, "tid": 7, "ts": 1716454223103451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998536, "dur": 7,
    "args": {
      "External id": 110367,
      "cbid": 211, "correlation": 110367
    }
  },
  {
    "ph": "s", "id": 110367, "pid": 76337, "tid": -914061504, "ts": 1716454222998536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223103561, "dur": 70,
    "args": {
      "External id": 110376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110376,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110376, "pid": 5, "tid": 7, "ts": 1716454223103561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998575, "dur": 10,
    "args": {
      "External id": 110376,
      "cbid": 211, "correlation": 110376
    }
  },
  {
    "ph": "s", "id": 110376, "pid": 76337, "tid": -914061504, "ts": 1716454222998575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998627, "dur": 0,
    "args": {
      "External id": 110386,
      "cbid": 317, "correlation": 110386
    }
  },
  {
    "ph": "f", "id": 110386, "pid": 76337, "tid": -914061504, "ts": 1716454222998627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998628, "dur": 0,
    "args": {
      "External id": 110387,
      "cbid": 203, "correlation": 110387
    }
  },
  {
    "ph": "f", "id": 110387, "pid": 76337, "tid": -914061504, "ts": 1716454222998628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998628, "dur": 0,
    "args": {
      "External id": 110388,
      "cbid": 205, "correlation": 110388
    }
  },
  {
    "ph": "f", "id": 110388, "pid": 76337, "tid": -914061504, "ts": 1716454222998628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223103633, "dur": 75,
    "args": {
      "External id": 110392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110392,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110392, "pid": 5, "tid": 7, "ts": 1716454223103633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998644, "dur": 12,
    "args": {
      "External id": 110392,
      "cbid": 211, "correlation": 110392
    }
  },
  {
    "ph": "s", "id": 110392, "pid": 76337, "tid": -914061504, "ts": 1716454222998644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223103709, "dur": 24,
    "args": {
      "External id": 110394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110394,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110394, "pid": 5, "tid": 7, "ts": 1716454223103709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998658, "dur": 5,
    "args": {
      "External id": 110394,
      "cbid": 211, "correlation": 110394
    }
  },
  {
    "ph": "s", "id": 110394, "pid": 76337, "tid": -914061504, "ts": 1716454222998658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223103735, "dur": 4,
    "args": {
      "External id": 110396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 110396, "pid": 5, "tid": 7, "ts": 1716454223103735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998668, "dur": 7,
    "args": {
      "External id": 110396,
      "cbid": 211, "correlation": 110396
    }
  },
  {
    "ph": "s", "id": 110396, "pid": 76337, "tid": -914061504, "ts": 1716454222998668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998678, "dur": 0,
    "args": {
      "External id": 110397,
      "cbid": 51, "correlation": 110397
    }
  },
  {
    "ph": "s", "id": 110397, "pid": 76337, "tid": -914061504, "ts": 1716454222998678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223103740, "dur": 1360,
    "args": {
      "External id": 110398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110398,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110398, "pid": 5, "tid": 7, "ts": 1716454223103740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998679, "dur": 5,
    "args": {
      "External id": 110398,
      "cbid": 211, "correlation": 110398
    }
  },
  {
    "ph": "s", "id": 110398, "pid": 76337, "tid": -914061504, "ts": 1716454222998679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223105102, "dur": 59,
    "args": {
      "External id": 110403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110403,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110403, "pid": 5, "tid": 7, "ts": 1716454223105102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998706, "dur": 9,
    "args": {
      "External id": 110403,
      "cbid": 211, "correlation": 110403
    }
  },
  {
    "ph": "s", "id": 110403, "pid": 76337, "tid": -914061504, "ts": 1716454222998706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223105162, "dur": 3,
    "args": {
      "External id": 110411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110411,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 110411, "pid": 5, "tid": 7, "ts": 1716454223105162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998750, "dur": 9,
    "args": {
      "External id": 110411,
      "cbid": 211, "correlation": 110411
    }
  },
  {
    "ph": "s", "id": 110411, "pid": 76337, "tid": -914061504, "ts": 1716454222998750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998814, "dur": 1,
    "args": {
      "External id": 110427,
      "cbid": 251, "correlation": 110427
    }
  },
  {
    "ph": "f", "id": 110427, "pid": 76337, "tid": -914061504, "ts": 1716454222998814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998820, "dur": 0,
    "args": {
      "External id": 110429,
      "cbid": 251, "correlation": 110429
    }
  },
  {
    "ph": "f", "id": 110429, "pid": 76337, "tid": -914061504, "ts": 1716454222998820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223105167, "dur": 11,
    "args": {
      "External id": 110430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110430,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 110430, "pid": 5, "tid": 7, "ts": 1716454223105167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998822, "dur": 12,
    "args": {
      "External id": 110430,
      "cbid": 211, "correlation": 110430
    }
  },
  {
    "ph": "s", "id": 110430, "pid": 76337, "tid": -914061504, "ts": 1716454222998822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223105180, "dur": 5,
    "args": {
      "External id": 110432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110432,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 110432, "pid": 5, "tid": 7, "ts": 1716454223105180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998836, "dur": 5,
    "args": {
      "External id": 110432,
      "cbid": 211, "correlation": 110432
    }
  },
  {
    "ph": "s", "id": 110432, "pid": 76337, "tid": -914061504, "ts": 1716454222998836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223105186, "dur": 54,
    "args": {
      "External id": 110442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110442, "pid": 5, "tid": 7, "ts": 1716454223105186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998893, "dur": 12,
    "args": {
      "External id": 110442,
      "cbid": 211, "correlation": 110442
    }
  },
  {
    "ph": "s", "id": 110442, "pid": 76337, "tid": -914061504, "ts": 1716454222998893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223105242, "dur": 51,
    "args": {
      "External id": 110462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110462,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 110462, "pid": 5, "tid": 7, "ts": 1716454223105242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998959, "dur": 10,
    "args": {
      "External id": 110462,
      "cbid": 211, "correlation": 110462
    }
  },
  {
    "ph": "s", "id": 110462, "pid": 76337, "tid": -914061504, "ts": 1716454222998959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223105294, "dur": 4,
    "args": {
      "External id": 110474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110474,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 110474, "pid": 5, "tid": 7, "ts": 1716454223105294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222998988, "dur": 7,
    "args": {
      "External id": 110474,
      "cbid": 211, "correlation": 110474
    }
  },
  {
    "ph": "s", "id": 110474, "pid": 76337, "tid": -914061504, "ts": 1716454222998988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223105299, "dur": 55,
    "args": {
      "External id": 110477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110477, "pid": 5, "tid": 7, "ts": 1716454223105299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999006, "dur": 7,
    "args": {
      "External id": 110477,
      "cbid": 211, "correlation": 110477
    }
  },
  {
    "ph": "s", "id": 110477, "pid": 76337, "tid": -914061504, "ts": 1716454222999006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223105356, "dur": 37,
    "args": {
      "External id": 110486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110486,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110486, "pid": 5, "tid": 7, "ts": 1716454223105356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999048, "dur": 10,
    "args": {
      "External id": 110486,
      "cbid": 211, "correlation": 110486
    }
  },
  {
    "ph": "s", "id": 110486, "pid": 76337, "tid": -914061504, "ts": 1716454222999048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999111, "dur": 0,
    "args": {
      "External id": 110496,
      "cbid": 317, "correlation": 110496
    }
  },
  {
    "ph": "f", "id": 110496, "pid": 76337, "tid": -914061504, "ts": 1716454222999111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999112, "dur": 0,
    "args": {
      "External id": 110497,
      "cbid": 203, "correlation": 110497
    }
  },
  {
    "ph": "f", "id": 110497, "pid": 76337, "tid": -914061504, "ts": 1716454222999112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999112, "dur": 0,
    "args": {
      "External id": 110498,
      "cbid": 205, "correlation": 110498
    }
  },
  {
    "ph": "f", "id": 110498, "pid": 76337, "tid": -914061504, "ts": 1716454222999112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223105394, "dur": 39,
    "args": {
      "External id": 110502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110502,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110502, "pid": 5, "tid": 7, "ts": 1716454223105394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999127, "dur": 12,
    "args": {
      "External id": 110502,
      "cbid": 211, "correlation": 110502
    }
  },
  {
    "ph": "s", "id": 110502, "pid": 76337, "tid": -914061504, "ts": 1716454222999127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223105435, "dur": 15,
    "args": {
      "External id": 110504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110504,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110504, "pid": 5, "tid": 7, "ts": 1716454223105435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999141, "dur": 5,
    "args": {
      "External id": 110504,
      "cbid": 211, "correlation": 110504
    }
  },
  {
    "ph": "s", "id": 110504, "pid": 76337, "tid": -914061504, "ts": 1716454222999141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223105451, "dur": 3,
    "args": {
      "External id": 110506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110506,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 110506, "pid": 5, "tid": 7, "ts": 1716454223105451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999151, "dur": 5,
    "args": {
      "External id": 110506,
      "cbid": 211, "correlation": 110506
    }
  },
  {
    "ph": "s", "id": 110506, "pid": 76337, "tid": -914061504, "ts": 1716454222999151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999159, "dur": 0,
    "args": {
      "External id": 110507,
      "cbid": 51, "correlation": 110507
    }
  },
  {
    "ph": "s", "id": 110507, "pid": 76337, "tid": -914061504, "ts": 1716454222999159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223105455, "dur": 694,
    "args": {
      "External id": 110508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110508,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110508, "pid": 5, "tid": 7, "ts": 1716454223105455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999160, "dur": 5,
    "args": {
      "External id": 110508,
      "cbid": 211, "correlation": 110508
    }
  },
  {
    "ph": "s", "id": 110508, "pid": 76337, "tid": -914061504, "ts": 1716454222999160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223106150, "dur": 59,
    "args": {
      "External id": 110513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110513,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110513, "pid": 5, "tid": 7, "ts": 1716454223106150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999187, "dur": 8,
    "args": {
      "External id": 110513,
      "cbid": 211, "correlation": 110513
    }
  },
  {
    "ph": "s", "id": 110513, "pid": 76337, "tid": -914061504, "ts": 1716454222999187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999244, "dur": 0,
    "args": {
      "External id": 110523,
      "cbid": 317, "correlation": 110523
    }
  },
  {
    "ph": "f", "id": 110523, "pid": 76337, "tid": -914061504, "ts": 1716454222999244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999245, "dur": 0,
    "args": {
      "External id": 110524,
      "cbid": 203, "correlation": 110524
    }
  },
  {
    "ph": "f", "id": 110524, "pid": 76337, "tid": -914061504, "ts": 1716454222999245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999246, "dur": 0,
    "args": {
      "External id": 110525,
      "cbid": 205, "correlation": 110525
    }
  },
  {
    "ph": "f", "id": 110525, "pid": 76337, "tid": -914061504, "ts": 1716454222999246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223106210, "dur": 74,
    "args": {
      "External id": 110529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110529,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110529, "pid": 5, "tid": 7, "ts": 1716454223106210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999258, "dur": 11,
    "args": {
      "External id": 110529,
      "cbid": 211, "correlation": 110529
    }
  },
  {
    "ph": "s", "id": 110529, "pid": 76337, "tid": -914061504, "ts": 1716454222999258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223106286, "dur": 207,
    "args": {
      "External id": 110531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110531,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110531, "pid": 5, "tid": 7, "ts": 1716454223106286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999278, "dur": 8,
    "args": {
      "External id": 110531,
      "cbid": 211, "correlation": 110531
    }
  },
  {
    "ph": "s", "id": 110531, "pid": 76337, "tid": -914061504, "ts": 1716454222999278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223106494, "dur": 38,
    "args": {
      "External id": 110533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110533,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110533, "pid": 5, "tid": 7, "ts": 1716454223106494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999290, "dur": 6,
    "args": {
      "External id": 110533,
      "cbid": 211, "correlation": 110533
    }
  },
  {
    "ph": "s", "id": 110533, "pid": 76337, "tid": -914061504, "ts": 1716454222999290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223106533, "dur": 59,
    "args": {
      "External id": 110539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110539,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110539, "pid": 5, "tid": 7, "ts": 1716454223106533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999316, "dur": 10,
    "args": {
      "External id": 110539,
      "cbid": 211, "correlation": 110539
    }
  },
  {
    "ph": "s", "id": 110539, "pid": 76337, "tid": -914061504, "ts": 1716454222999316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223106594, "dur": 50,
    "args": {
      "External id": 110547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110547,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110547, "pid": 5, "tid": 7, "ts": 1716454223106594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999347, "dur": 8,
    "args": {
      "External id": 110547,
      "cbid": 211, "correlation": 110547
    }
  },
  {
    "ph": "s", "id": 110547, "pid": 76337, "tid": -914061504, "ts": 1716454222999347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223106645, "dur": 36,
    "args": {
      "External id": 110555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110555,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110555, "pid": 5, "tid": 7, "ts": 1716454223106645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999377, "dur": 8,
    "args": {
      "External id": 110555,
      "cbid": 211, "correlation": 110555
    }
  },
  {
    "ph": "s", "id": 110555, "pid": 76337, "tid": -914061504, "ts": 1716454222999377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223106682, "dur": 53,
    "args": {
      "External id": 110575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110575,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 110575, "pid": 5, "tid": 7, "ts": 1716454223106682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999459, "dur": 13,
    "args": {
      "External id": 110575,
      "cbid": 211, "correlation": 110575
    }
  },
  {
    "ph": "s", "id": 110575, "pid": 76337, "tid": -914061504, "ts": 1716454222999459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223106737, "dur": 5,
    "args": {
      "External id": 110587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110587,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 110587, "pid": 5, "tid": 7, "ts": 1716454223106737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999483, "dur": 6,
    "args": {
      "External id": 110587,
      "cbid": 211, "correlation": 110587
    }
  },
  {
    "ph": "s", "id": 110587, "pid": 76337, "tid": -914061504, "ts": 1716454222999483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223106743, "dur": 55,
    "args": {
      "External id": 110590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110590, "pid": 5, "tid": 7, "ts": 1716454223106743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999500, "dur": 6,
    "args": {
      "External id": 110590,
      "cbid": 211, "correlation": 110590
    }
  },
  {
    "ph": "s", "id": 110590, "pid": 76337, "tid": -914061504, "ts": 1716454222999500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999557, "dur": 0,
    "args": {
      "External id": 110601,
      "cbid": 317, "correlation": 110601
    }
  },
  {
    "ph": "f", "id": 110601, "pid": 76337, "tid": -914061504, "ts": 1716454222999557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999557, "dur": 0,
    "args": {
      "External id": 110602,
      "cbid": 203, "correlation": 110602
    }
  },
  {
    "ph": "f", "id": 110602, "pid": 76337, "tid": -914061504, "ts": 1716454222999557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999558, "dur": 0,
    "args": {
      "External id": 110603,
      "cbid": 205, "correlation": 110603
    }
  },
  {
    "ph": "f", "id": 110603, "pid": 76337, "tid": -914061504, "ts": 1716454222999558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999581, "dur": 1,
    "args": {
      "External id": 110607,
      "cbid": 251, "correlation": 110607
    }
  },
  {
    "ph": "f", "id": 110607, "pid": 76337, "tid": -914061504, "ts": 1716454222999581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999583, "dur": 0,
    "args": {
      "External id": 110608,
      "cbid": 251, "correlation": 110608
    }
  },
  {
    "ph": "f", "id": 110608, "pid": 76337, "tid": -914061504, "ts": 1716454222999583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999583, "dur": 0,
    "args": {
      "External id": 110609,
      "cbid": 251, "correlation": 110609
    }
  },
  {
    "ph": "f", "id": 110609, "pid": 76337, "tid": -914061504, "ts": 1716454222999583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999584, "dur": 0,
    "args": {
      "External id": 110610,
      "cbid": 251, "correlation": 110610
    }
  },
  {
    "ph": "f", "id": 110610, "pid": 76337, "tid": -914061504, "ts": 1716454222999584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999585, "dur": 0,
    "args": {
      "External id": 110611,
      "cbid": 251, "correlation": 110611
    }
  },
  {
    "ph": "f", "id": 110611, "pid": 76337, "tid": -914061504, "ts": 1716454222999585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999585, "dur": 0,
    "args": {
      "External id": 110612,
      "cbid": 251, "correlation": 110612
    }
  },
  {
    "ph": "f", "id": 110612, "pid": 76337, "tid": -914061504, "ts": 1716454222999585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999586, "dur": 0,
    "args": {
      "External id": 110613,
      "cbid": 251, "correlation": 110613
    }
  },
  {
    "ph": "f", "id": 110613, "pid": 76337, "tid": -914061504, "ts": 1716454222999586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999587, "dur": 0,
    "args": {
      "External id": 110614,
      "cbid": 251, "correlation": 110614
    }
  },
  {
    "ph": "f", "id": 110614, "pid": 76337, "tid": -914061504, "ts": 1716454222999587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999588, "dur": 0,
    "args": {
      "External id": 110615,
      "cbid": 251, "correlation": 110615
    }
  },
  {
    "ph": "f", "id": 110615, "pid": 76337, "tid": -914061504, "ts": 1716454222999588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223106799, "dur": 111,
    "args": {
      "External id": 110616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110616,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110616, "pid": 5, "tid": 7, "ts": 1716454223106799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999590, "dur": 12,
    "args": {
      "External id": 110616,
      "cbid": 211, "correlation": 110616
    }
  },
  {
    "ph": "s", "id": 110616, "pid": 76337, "tid": -914061504, "ts": 1716454222999590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223106912, "dur": 59,
    "args": {
      "External id": 110622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110622,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110622, "pid": 5, "tid": 7, "ts": 1716454223106912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999626, "dur": 9,
    "args": {
      "External id": 110622,
      "cbid": 211, "correlation": 110622
    }
  },
  {
    "ph": "s", "id": 110622, "pid": 76337, "tid": -914061504, "ts": 1716454222999626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223106972, "dur": 556,
    "args": {
      "External id": 110631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110631,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110631, "pid": 5, "tid": 7, "ts": 1716454223106972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999707, "dur": 14,
    "args": {
      "External id": 110631,
      "cbid": 211, "correlation": 110631
    }
  },
  {
    "ph": "s", "id": 110631, "pid": 76337, "tid": -914061504, "ts": 1716454222999707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223107529, "dur": 181,
    "args": {
      "External id": 110653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110653,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110653, "pid": 5, "tid": 7, "ts": 1716454223107529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999764, "dur": 11,
    "args": {
      "External id": 110653,
      "cbid": 211, "correlation": 110653
    }
  },
  {
    "ph": "s", "id": 110653, "pid": 76337, "tid": -914061504, "ts": 1716454222999764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999851, "dur": 1,
    "args": {
      "External id": 110664,
      "cbid": 251, "correlation": 110664
    }
  },
  {
    "ph": "f", "id": 110664, "pid": 76337, "tid": -914061504, "ts": 1716454222999851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223107712, "dur": 196,
    "args": {
      "External id": 110665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110665,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110665, "pid": 5, "tid": 7, "ts": 1716454223107712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999856, "dur": 13,
    "args": {
      "External id": 110665,
      "cbid": 211, "correlation": 110665
    }
  },
  {
    "ph": "s", "id": 110665, "pid": 76337, "tid": -914061504, "ts": 1716454222999856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999924, "dur": 1,
    "args": {
      "External id": 110676,
      "cbid": 251, "correlation": 110676
    }
  },
  {
    "ph": "f", "id": 110676, "pid": 76337, "tid": -914061504, "ts": 1716454222999924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223107909, "dur": 191,
    "args": {
      "External id": 110677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110677,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110677, "pid": 5, "tid": 7, "ts": 1716454223107909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454222999928, "dur": 11,
    "args": {
      "External id": 110677,
      "cbid": 211, "correlation": 110677
    }
  },
  {
    "ph": "s", "id": 110677, "pid": 76337, "tid": -914061504, "ts": 1716454222999928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000000, "dur": 1,
    "args": {
      "External id": 110688,
      "cbid": 251, "correlation": 110688
    }
  },
  {
    "ph": "f", "id": 110688, "pid": 76337, "tid": -914061504, "ts": 1716454223000000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223108101, "dur": 187,
    "args": {
      "External id": 110689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110689,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110689, "pid": 5, "tid": 7, "ts": 1716454223108101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000004, "dur": 11,
    "args": {
      "External id": 110689,
      "cbid": 211, "correlation": 110689
    }
  },
  {
    "ph": "s", "id": 110689, "pid": 76337, "tid": -914061504, "ts": 1716454223000004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223108289, "dur": 18578,
    "args": {
      "External id": 110710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110710,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110710, "pid": 5, "tid": 7, "ts": 1716454223108289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000087, "dur": 12,
    "args": {
      "External id": 110710,
      "cbid": 211, "correlation": 110710
    }
  },
  {
    "ph": "s", "id": 110710, "pid": 76337, "tid": -914061504, "ts": 1716454223000087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000184, "dur": 1,
    "args": {
      "External id": 110728,
      "cbid": 251, "correlation": 110728
    }
  },
  {
    "ph": "f", "id": 110728, "pid": 76337, "tid": -914061504, "ts": 1716454223000184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223126868, "dur": 204,
    "args": {
      "External id": 110730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110730,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110730, "pid": 5, "tid": 7, "ts": 1716454223126868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000190, "dur": 13,
    "args": {
      "External id": 110730,
      "cbid": 211, "correlation": 110730
    }
  },
  {
    "ph": "s", "id": 110730, "pid": 76337, "tid": -914061504, "ts": 1716454223000190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223127073, "dur": 67,
    "args": {
      "External id": 110738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110738,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110738, "pid": 5, "tid": 7, "ts": 1716454223127073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000259, "dur": 12,
    "args": {
      "External id": 110738,
      "cbid": 211, "correlation": 110738
    }
  },
  {
    "ph": "s", "id": 110738, "pid": 76337, "tid": -914061504, "ts": 1716454223000259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223127141, "dur": 97,
    "args": {
      "External id": 110746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110746,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110746, "pid": 5, "tid": 7, "ts": 1716454223127141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000299, "dur": 8,
    "args": {
      "External id": 110746,
      "cbid": 211, "correlation": 110746
    }
  },
  {
    "ph": "s", "id": 110746, "pid": 76337, "tid": -914061504, "ts": 1716454223000299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223127240, "dur": 55,
    "args": {
      "External id": 110757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110757,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110757, "pid": 5, "tid": 7, "ts": 1716454223127240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000369, "dur": 13,
    "args": {
      "External id": 110757,
      "cbid": 211, "correlation": 110757
    }
  },
  {
    "ph": "s", "id": 110757, "pid": 76337, "tid": -914061504, "ts": 1716454223000369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223127297, "dur": 92,
    "args": {
      "External id": 110779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110779,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110779, "pid": 5, "tid": 7, "ts": 1716454223127297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223000401, "dur": 1219,
    "args": {
      "External id": 110779,
      "cbid": 211, "correlation": 110779
    }
  },
  {
    "ph": "s", "id": 110779, "pid": 76337, "tid": -914061504, "ts": 1716454223000401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001698, "dur": 1,
    "args": {
      "External id": 110790,
      "cbid": 251, "correlation": 110790
    }
  },
  {
    "ph": "f", "id": 110790, "pid": 76337, "tid": -914061504, "ts": 1716454223001698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223127390, "dur": 106,
    "args": {
      "External id": 110791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110791,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110791, "pid": 5, "tid": 7, "ts": 1716454223127390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001704, "dur": 64,
    "args": {
      "External id": 110791,
      "cbid": 211, "correlation": 110791
    }
  },
  {
    "ph": "s", "id": 110791, "pid": 76337, "tid": -914061504, "ts": 1716454223001704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001828, "dur": 1,
    "args": {
      "External id": 110802,
      "cbid": 251, "correlation": 110802
    }
  },
  {
    "ph": "f", "id": 110802, "pid": 76337, "tid": -914061504, "ts": 1716454223001828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001832, "dur": 0,
    "args": {
      "External id": 110803,
      "cbid": 251, "correlation": 110803
    }
  },
  {
    "ph": "f", "id": 110803, "pid": 76337, "tid": -914061504, "ts": 1716454223001832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223127498, "dur": 10,
    "args": {
      "External id": 110804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110804,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 110804, "pid": 5, "tid": 7, "ts": 1716454223127498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001834, "dur": 12,
    "args": {
      "External id": 110804,
      "cbid": 211, "correlation": 110804
    }
  },
  {
    "ph": "s", "id": 110804, "pid": 76337, "tid": -914061504, "ts": 1716454223001834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223127509, "dur": 5,
    "args": {
      "External id": 110806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110806,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 110806, "pid": 5, "tid": 7, "ts": 1716454223127509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001847, "dur": 6,
    "args": {
      "External id": 110806,
      "cbid": 211, "correlation": 110806
    }
  },
  {
    "ph": "s", "id": 110806, "pid": 76337, "tid": -914061504, "ts": 1716454223001847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001908, "dur": 1,
    "args": {
      "External id": 110817,
      "cbid": 251, "correlation": 110817
    }
  },
  {
    "ph": "f", "id": 110817, "pid": 76337, "tid": -914061504, "ts": 1716454223001908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001911, "dur": 0,
    "args": {
      "External id": 110818,
      "cbid": 251, "correlation": 110818
    }
  },
  {
    "ph": "f", "id": 110818, "pid": 76337, "tid": -914061504, "ts": 1716454223001911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223127516, "dur": 6,
    "args": {
      "External id": 110819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110819,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 110819, "pid": 5, "tid": 7, "ts": 1716454223127516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001913, "dur": 12,
    "args": {
      "External id": 110819,
      "cbid": 211, "correlation": 110819
    }
  },
  {
    "ph": "s", "id": 110819, "pid": 76337, "tid": -914061504, "ts": 1716454223001913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223127523, "dur": 3,
    "args": {
      "External id": 110821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110821,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 110821, "pid": 5, "tid": 7, "ts": 1716454223127523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223001926, "dur": 6,
    "args": {
      "External id": 110821,
      "cbid": 211, "correlation": 110821
    }
  },
  {
    "ph": "s", "id": 110821, "pid": 76337, "tid": -914061504, "ts": 1716454223001926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223127528, "dur": 157,
    "args": {
      "External id": 110842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110842,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110842, "pid": 5, "tid": 7, "ts": 1716454223127528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002009, "dur": 13,
    "args": {
      "External id": 110842,
      "cbid": 211, "correlation": 110842
    }
  },
  {
    "ph": "s", "id": 110842, "pid": 76337, "tid": -914061504, "ts": 1716454223002009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002108, "dur": 1,
    "args": {
      "External id": 110860,
      "cbid": 251, "correlation": 110860
    }
  },
  {
    "ph": "f", "id": 110860, "pid": 76337, "tid": -914061504, "ts": 1716454223002108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223127686, "dur": 106,
    "args": {
      "External id": 110862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110862,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110862, "pid": 5, "tid": 7, "ts": 1716454223127686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002114, "dur": 14,
    "args": {
      "External id": 110862,
      "cbid": 211, "correlation": 110862
    }
  },
  {
    "ph": "s", "id": 110862, "pid": 76337, "tid": -914061504, "ts": 1716454223002114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223127793, "dur": 34,
    "args": {
      "External id": 110870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110870,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110870, "pid": 5, "tid": 7, "ts": 1716454223127793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002185, "dur": 12,
    "args": {
      "External id": 110870,
      "cbid": 211, "correlation": 110870
    }
  },
  {
    "ph": "s", "id": 110870, "pid": 76337, "tid": -914061504, "ts": 1716454223002185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223127829, "dur": 67,
    "args": {
      "External id": 110878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110878, "pid": 5, "tid": 7, "ts": 1716454223127829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002225, "dur": 9,
    "args": {
      "External id": 110878,
      "cbid": 211, "correlation": 110878
    }
  },
  {
    "ph": "s", "id": 110878, "pid": 76337, "tid": -914061504, "ts": 1716454223002225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223127897, "dur": 92,
    "args": {
      "External id": 110900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110900,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110900, "pid": 5, "tid": 7, "ts": 1716454223127897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002275, "dur": 11,
    "args": {
      "External id": 110900,
      "cbid": 211, "correlation": 110900
    }
  },
  {
    "ph": "s", "id": 110900, "pid": 76337, "tid": -914061504, "ts": 1716454223002275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002361, "dur": 1,
    "args": {
      "External id": 110916,
      "cbid": 251, "correlation": 110916
    }
  },
  {
    "ph": "f", "id": 110916, "pid": 76337, "tid": -914061504, "ts": 1716454223002361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223127990, "dur": 573,
    "args": {
      "External id": 110918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110918,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 110918, "pid": 5, "tid": 7, "ts": 1716454223127990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002366, "dur": 13,
    "args": {
      "External id": 110918,
      "cbid": 211, "correlation": 110918
    }
  },
  {
    "ph": "s", "id": 110918, "pid": 76337, "tid": -914061504, "ts": 1716454223002366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223128565, "dur": 244,
    "args": {
      "External id": 110926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110926, "pid": 5, "tid": 7, "ts": 1716454223128565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002432, "dur": 12,
    "args": {
      "External id": 110926,
      "cbid": 211, "correlation": 110926
    }
  },
  {
    "ph": "s", "id": 110926, "pid": 76337, "tid": -914061504, "ts": 1716454223002432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223128810, "dur": 252,
    "args": {
      "External id": 110934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110934, "pid": 5, "tid": 7, "ts": 1716454223128810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002462, "dur": 8,
    "args": {
      "External id": 110934,
      "cbid": 211, "correlation": 110934
    }
  },
  {
    "ph": "s", "id": 110934, "pid": 76337, "tid": -914061504, "ts": 1716454223002462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002545, "dur": 1,
    "args": {
      "External id": 110950,
      "cbid": 251, "correlation": 110950
    }
  },
  {
    "ph": "f", "id": 110950, "pid": 76337, "tid": -914061504, "ts": 1716454223002545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002550, "dur": 0,
    "args": {
      "External id": 110952,
      "cbid": 251, "correlation": 110952
    }
  },
  {
    "ph": "f", "id": 110952, "pid": 76337, "tid": -914061504, "ts": 1716454223002550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223129063, "dur": 359,
    "args": {
      "External id": 110953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110953,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 110953, "pid": 5, "tid": 7, "ts": 1716454223129063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002553, "dur": 13,
    "args": {
      "External id": 110953,
      "cbid": 211, "correlation": 110953
    }
  },
  {
    "ph": "s", "id": 110953, "pid": 76337, "tid": -914061504, "ts": 1716454223002553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223129423, "dur": 50,
    "args": {
      "External id": 110961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110961,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110961, "pid": 5, "tid": 7, "ts": 1716454223129423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002595, "dur": 10,
    "args": {
      "External id": 110961,
      "cbid": 211, "correlation": 110961
    }
  },
  {
    "ph": "s", "id": 110961, "pid": 76337, "tid": -914061504, "ts": 1716454223002595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223129475, "dur": 156,
    "args": {
      "External id": 110972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 110972, "pid": 5, "tid": 7, "ts": 1716454223129475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002661, "dur": 200,
    "args": {
      "External id": 110972,
      "cbid": 211, "correlation": 110972
    }
  },
  {
    "ph": "s", "id": 110972, "pid": 76337, "tid": -914061504, "ts": 1716454223002661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002915, "dur": 0,
    "args": {
      "External id": 110984,
      "cbid": 317, "correlation": 110984
    }
  },
  {
    "ph": "f", "id": 110984, "pid": 76337, "tid": -914061504, "ts": 1716454223002915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002916, "dur": 0,
    "args": {
      "External id": 110985,
      "cbid": 203, "correlation": 110985
    }
  },
  {
    "ph": "f", "id": 110985, "pid": 76337, "tid": -914061504, "ts": 1716454223002916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002917, "dur": 0,
    "args": {
      "External id": 110986,
      "cbid": 205, "correlation": 110986
    }
  },
  {
    "ph": "f", "id": 110986, "pid": 76337, "tid": -914061504, "ts": 1716454223002917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002939, "dur": 1,
    "args": {
      "External id": 110990,
      "cbid": 251, "correlation": 110990
    }
  },
  {
    "ph": "f", "id": 110990, "pid": 76337, "tid": -914061504, "ts": 1716454223002939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002941, "dur": 0,
    "args": {
      "External id": 110991,
      "cbid": 251, "correlation": 110991
    }
  },
  {
    "ph": "f", "id": 110991, "pid": 76337, "tid": -914061504, "ts": 1716454223002941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002942, "dur": 0,
    "args": {
      "External id": 110992,
      "cbid": 251, "correlation": 110992
    }
  },
  {
    "ph": "f", "id": 110992, "pid": 76337, "tid": -914061504, "ts": 1716454223002942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002943, "dur": 0,
    "args": {
      "External id": 110993,
      "cbid": 251, "correlation": 110993
    }
  },
  {
    "ph": "f", "id": 110993, "pid": 76337, "tid": -914061504, "ts": 1716454223002943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002943, "dur": 0,
    "args": {
      "External id": 110994,
      "cbid": 251, "correlation": 110994
    }
  },
  {
    "ph": "f", "id": 110994, "pid": 76337, "tid": -914061504, "ts": 1716454223002943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002944, "dur": 0,
    "args": {
      "External id": 110995,
      "cbid": 251, "correlation": 110995
    }
  },
  {
    "ph": "f", "id": 110995, "pid": 76337, "tid": -914061504, "ts": 1716454223002944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002945, "dur": 0,
    "args": {
      "External id": 110996,
      "cbid": 251, "correlation": 110996
    }
  },
  {
    "ph": "f", "id": 110996, "pid": 76337, "tid": -914061504, "ts": 1716454223002945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002946, "dur": 0,
    "args": {
      "External id": 110997,
      "cbid": 251, "correlation": 110997
    }
  },
  {
    "ph": "f", "id": 110997, "pid": 76337, "tid": -914061504, "ts": 1716454223002946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002947, "dur": 0,
    "args": {
      "External id": 110998,
      "cbid": 251, "correlation": 110998
    }
  },
  {
    "ph": "f", "id": 110998, "pid": 76337, "tid": -914061504, "ts": 1716454223002947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223129632, "dur": 115,
    "args": {
      "External id": 110999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 110999,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 110999, "pid": 5, "tid": 7, "ts": 1716454223129632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223002949, "dur": 38,
    "args": {
      "External id": 110999,
      "cbid": 211, "correlation": 110999
    }
  },
  {
    "ph": "s", "id": 110999, "pid": 76337, "tid": -914061504, "ts": 1716454223002949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223129748, "dur": 59,
    "args": {
      "External id": 111005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111005, "pid": 5, "tid": 7, "ts": 1716454223129748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003011, "dur": 102,
    "args": {
      "External id": 111005,
      "cbid": 211, "correlation": 111005
    }
  },
  {
    "ph": "s", "id": 111005, "pid": 76337, "tid": -914061504, "ts": 1716454223003011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223129809, "dur": 51,
    "args": {
      "External id": 111013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111013,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111013, "pid": 5, "tid": 7, "ts": 1716454223129809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003137, "dur": 277,
    "args": {
      "External id": 111013,
      "cbid": 211, "correlation": 111013
    }
  },
  {
    "ph": "s", "id": 111013, "pid": 76337, "tid": -914061504, "ts": 1716454223003137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223129861, "dur": 99,
    "args": {
      "External id": 111022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111022,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111022, "pid": 5, "tid": 7, "ts": 1716454223129861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003444, "dur": 11,
    "args": {
      "External id": 111022,
      "cbid": 211, "correlation": 111022
    }
  },
  {
    "ph": "s", "id": 111022, "pid": 76337, "tid": -914061504, "ts": 1716454223003444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223129961, "dur": 92,
    "args": {
      "External id": 111042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111042,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 111042, "pid": 5, "tid": 7, "ts": 1716454223129961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003516, "dur": 12,
    "args": {
      "External id": 111042,
      "cbid": 211, "correlation": 111042
    }
  },
  {
    "ph": "s", "id": 111042, "pid": 76337, "tid": -914061504, "ts": 1716454223003516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223130054, "dur": 5,
    "args": {
      "External id": 111054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111054,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 111054, "pid": 5, "tid": 7, "ts": 1716454223130054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003537, "dur": 10,
    "args": {
      "External id": 111054,
      "cbid": 211, "correlation": 111054
    }
  },
  {
    "ph": "s", "id": 111054, "pid": 76337, "tid": -914061504, "ts": 1716454223003537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223130060, "dur": 108,
    "args": {
      "External id": 111057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111057, "pid": 5, "tid": 7, "ts": 1716454223130060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003560, "dur": 107,
    "args": {
      "External id": 111057,
      "cbid": 211, "correlation": 111057
    }
  },
  {
    "ph": "s", "id": 111057, "pid": 76337, "tid": -914061504, "ts": 1716454223003560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223130169, "dur": 69,
    "args": {
      "External id": 111066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111066,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111066, "pid": 5, "tid": 7, "ts": 1716454223130169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003700, "dur": 10,
    "args": {
      "External id": 111066,
      "cbid": 211, "correlation": 111066
    }
  },
  {
    "ph": "s", "id": 111066, "pid": 76337, "tid": -914061504, "ts": 1716454223003700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003752, "dur": 0,
    "args": {
      "External id": 111076,
      "cbid": 317, "correlation": 111076
    }
  },
  {
    "ph": "f", "id": 111076, "pid": 76337, "tid": -914061504, "ts": 1716454223003752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003753, "dur": 0,
    "args": {
      "External id": 111077,
      "cbid": 203, "correlation": 111077
    }
  },
  {
    "ph": "f", "id": 111077, "pid": 76337, "tid": -914061504, "ts": 1716454223003753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003753, "dur": 0,
    "args": {
      "External id": 111078,
      "cbid": 205, "correlation": 111078
    }
  },
  {
    "ph": "f", "id": 111078, "pid": 76337, "tid": -914061504, "ts": 1716454223003753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223130240, "dur": 75,
    "args": {
      "External id": 111082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111082, "pid": 5, "tid": 7, "ts": 1716454223130240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003768, "dur": 12,
    "args": {
      "External id": 111082,
      "cbid": 211, "correlation": 111082
    }
  },
  {
    "ph": "s", "id": 111082, "pid": 76337, "tid": -914061504, "ts": 1716454223003768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223130316, "dur": 24,
    "args": {
      "External id": 111084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111084,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111084, "pid": 5, "tid": 7, "ts": 1716454223130316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003782, "dur": 5,
    "args": {
      "External id": 111084,
      "cbid": 211, "correlation": 111084
    }
  },
  {
    "ph": "s", "id": 111084, "pid": 76337, "tid": -914061504, "ts": 1716454223003782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223130341, "dur": 3,
    "args": {
      "External id": 111086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111086, "pid": 5, "tid": 7, "ts": 1716454223130341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003791, "dur": 5,
    "args": {
      "External id": 111086,
      "cbid": 211, "correlation": 111086
    }
  },
  {
    "ph": "s", "id": 111086, "pid": 76337, "tid": -914061504, "ts": 1716454223003791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003800, "dur": 0,
    "args": {
      "External id": 111087,
      "cbid": 51, "correlation": 111087
    }
  },
  {
    "ph": "s", "id": 111087, "pid": 76337, "tid": -914061504, "ts": 1716454223003800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223130346, "dur": 1359,
    "args": {
      "External id": 111088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111088,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111088, "pid": 5, "tid": 7, "ts": 1716454223130346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003801, "dur": 5,
    "args": {
      "External id": 111088,
      "cbid": 211, "correlation": 111088
    }
  },
  {
    "ph": "s", "id": 111088, "pid": 76337, "tid": -914061504, "ts": 1716454223003801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223131706, "dur": 59,
    "args": {
      "External id": 111093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111093, "pid": 5, "tid": 7, "ts": 1716454223131706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003828, "dur": 9,
    "args": {
      "External id": 111093,
      "cbid": 211, "correlation": 111093
    }
  },
  {
    "ph": "s", "id": 111093, "pid": 76337, "tid": -914061504, "ts": 1716454223003828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223131766, "dur": 3,
    "args": {
      "External id": 111101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111101,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111101, "pid": 5, "tid": 7, "ts": 1716454223131766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003872, "dur": 10,
    "args": {
      "External id": 111101,
      "cbid": 211, "correlation": 111101
    }
  },
  {
    "ph": "s", "id": 111101, "pid": 76337, "tid": -914061504, "ts": 1716454223003872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003937, "dur": 1,
    "args": {
      "External id": 111117,
      "cbid": 251, "correlation": 111117
    }
  },
  {
    "ph": "f", "id": 111117, "pid": 76337, "tid": -914061504, "ts": 1716454223003937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003942, "dur": 0,
    "args": {
      "External id": 111119,
      "cbid": 251, "correlation": 111119
    }
  },
  {
    "ph": "f", "id": 111119, "pid": 76337, "tid": -914061504, "ts": 1716454223003942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223131771, "dur": 11,
    "args": {
      "External id": 111120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111120,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 111120, "pid": 5, "tid": 7, "ts": 1716454223131771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003944, "dur": 11,
    "args": {
      "External id": 111120,
      "cbid": 211, "correlation": 111120
    }
  },
  {
    "ph": "s", "id": 111120, "pid": 76337, "tid": -914061504, "ts": 1716454223003944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223131783, "dur": 5,
    "args": {
      "External id": 111122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111122,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 111122, "pid": 5, "tid": 7, "ts": 1716454223131783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223003957, "dur": 7,
    "args": {
      "External id": 111122,
      "cbid": 211, "correlation": 111122
    }
  },
  {
    "ph": "s", "id": 111122, "pid": 76337, "tid": -914061504, "ts": 1716454223003957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223131790, "dur": 54,
    "args": {
      "External id": 111132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111132, "pid": 5, "tid": 7, "ts": 1716454223131790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004024, "dur": 538,
    "args": {
      "External id": 111132,
      "cbid": 211, "correlation": 111132
    }
  },
  {
    "ph": "s", "id": 111132, "pid": 76337, "tid": -914061504, "ts": 1716454223004024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223131845, "dur": 52,
    "args": {
      "External id": 111152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111152,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 111152, "pid": 5, "tid": 7, "ts": 1716454223131845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004616, "dur": 12,
    "args": {
      "External id": 111152,
      "cbid": 211, "correlation": 111152
    }
  },
  {
    "ph": "s", "id": 111152, "pid": 76337, "tid": -914061504, "ts": 1716454223004616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223131898, "dur": 4,
    "args": {
      "External id": 111164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111164,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111164, "pid": 5, "tid": 7, "ts": 1716454223131898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004638, "dur": 6,
    "args": {
      "External id": 111164,
      "cbid": 211, "correlation": 111164
    }
  },
  {
    "ph": "s", "id": 111164, "pid": 76337, "tid": -914061504, "ts": 1716454223004638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223131903, "dur": 55,
    "args": {
      "External id": 111167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111167, "pid": 5, "tid": 7, "ts": 1716454223131903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004656, "dur": 7,
    "args": {
      "External id": 111167,
      "cbid": 211, "correlation": 111167
    }
  },
  {
    "ph": "s", "id": 111167, "pid": 76337, "tid": -914061504, "ts": 1716454223004656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223131960, "dur": 36,
    "args": {
      "External id": 111176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111176,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111176, "pid": 5, "tid": 7, "ts": 1716454223131960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004697, "dur": 10,
    "args": {
      "External id": 111176,
      "cbid": 211, "correlation": 111176
    }
  },
  {
    "ph": "s", "id": 111176, "pid": 76337, "tid": -914061504, "ts": 1716454223004697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004761, "dur": 0,
    "args": {
      "External id": 111186,
      "cbid": 317, "correlation": 111186
    }
  },
  {
    "ph": "f", "id": 111186, "pid": 76337, "tid": -914061504, "ts": 1716454223004761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004762, "dur": 0,
    "args": {
      "External id": 111187,
      "cbid": 203, "correlation": 111187
    }
  },
  {
    "ph": "f", "id": 111187, "pid": 76337, "tid": -914061504, "ts": 1716454223004762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004763, "dur": 0,
    "args": {
      "External id": 111188,
      "cbid": 205, "correlation": 111188
    }
  },
  {
    "ph": "f", "id": 111188, "pid": 76337, "tid": -914061504, "ts": 1716454223004763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223131998, "dur": 40,
    "args": {
      "External id": 111192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111192,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111192, "pid": 5, "tid": 7, "ts": 1716454223131998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004777, "dur": 12,
    "args": {
      "External id": 111192,
      "cbid": 211, "correlation": 111192
    }
  },
  {
    "ph": "s", "id": 111192, "pid": 76337, "tid": -914061504, "ts": 1716454223004777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223132039, "dur": 14,
    "args": {
      "External id": 111194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111194,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111194, "pid": 5, "tid": 7, "ts": 1716454223132039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004792, "dur": 5,
    "args": {
      "External id": 111194,
      "cbid": 211, "correlation": 111194
    }
  },
  {
    "ph": "s", "id": 111194, "pid": 76337, "tid": -914061504, "ts": 1716454223004792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223132054, "dur": 3,
    "args": {
      "External id": 111196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111196,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111196, "pid": 5, "tid": 7, "ts": 1716454223132054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004801, "dur": 5,
    "args": {
      "External id": 111196,
      "cbid": 211, "correlation": 111196
    }
  },
  {
    "ph": "s", "id": 111196, "pid": 76337, "tid": -914061504, "ts": 1716454223004801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004810, "dur": 0,
    "args": {
      "External id": 111197,
      "cbid": 51, "correlation": 111197
    }
  },
  {
    "ph": "s", "id": 111197, "pid": 76337, "tid": -914061504, "ts": 1716454223004810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223132059, "dur": 694,
    "args": {
      "External id": 111198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111198,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111198, "pid": 5, "tid": 7, "ts": 1716454223132059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004810, "dur": 5,
    "args": {
      "External id": 111198,
      "cbid": 211, "correlation": 111198
    }
  },
  {
    "ph": "s", "id": 111198, "pid": 76337, "tid": -914061504, "ts": 1716454223004810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223132755, "dur": 59,
    "args": {
      "External id": 111203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111203,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111203, "pid": 5, "tid": 7, "ts": 1716454223132755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004838, "dur": 8,
    "args": {
      "External id": 111203,
      "cbid": 211, "correlation": 111203
    }
  },
  {
    "ph": "s", "id": 111203, "pid": 76337, "tid": -914061504, "ts": 1716454223004838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004895, "dur": 0,
    "args": {
      "External id": 111213,
      "cbid": 317, "correlation": 111213
    }
  },
  {
    "ph": "f", "id": 111213, "pid": 76337, "tid": -914061504, "ts": 1716454223004895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004896, "dur": 0,
    "args": {
      "External id": 111214,
      "cbid": 203, "correlation": 111214
    }
  },
  {
    "ph": "f", "id": 111214, "pid": 76337, "tid": -914061504, "ts": 1716454223004896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004897, "dur": 0,
    "args": {
      "External id": 111215,
      "cbid": 205, "correlation": 111215
    }
  },
  {
    "ph": "f", "id": 111215, "pid": 76337, "tid": -914061504, "ts": 1716454223004897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223132815, "dur": 75,
    "args": {
      "External id": 111219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111219,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111219, "pid": 5, "tid": 7, "ts": 1716454223132815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004908, "dur": 12,
    "args": {
      "External id": 111219,
      "cbid": 211, "correlation": 111219
    }
  },
  {
    "ph": "s", "id": 111219, "pid": 76337, "tid": -914061504, "ts": 1716454223004908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223132892, "dur": 211,
    "args": {
      "External id": 111221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111221,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111221, "pid": 5, "tid": 7, "ts": 1716454223132892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004926, "dur": 7,
    "args": {
      "External id": 111221,
      "cbid": 211, "correlation": 111221
    }
  },
  {
    "ph": "s", "id": 111221, "pid": 76337, "tid": -914061504, "ts": 1716454223004926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223133104, "dur": 39,
    "args": {
      "External id": 111223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111223,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111223, "pid": 5, "tid": 7, "ts": 1716454223133104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004938, "dur": 5,
    "args": {
      "External id": 111223,
      "cbid": 211, "correlation": 111223
    }
  },
  {
    "ph": "s", "id": 111223, "pid": 76337, "tid": -914061504, "ts": 1716454223004938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223133144, "dur": 60,
    "args": {
      "External id": 111229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111229,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111229, "pid": 5, "tid": 7, "ts": 1716454223133144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223004964, "dur": 508,
    "args": {
      "External id": 111229,
      "cbid": 211, "correlation": 111229
    }
  },
  {
    "ph": "s", "id": 111229, "pid": 76337, "tid": -914061504, "ts": 1716454223004964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223133205, "dur": 51,
    "args": {
      "External id": 111237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111237,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111237, "pid": 5, "tid": 7, "ts": 1716454223133205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005495, "dur": 9,
    "args": {
      "External id": 111237,
      "cbid": 211, "correlation": 111237
    }
  },
  {
    "ph": "s", "id": 111237, "pid": 76337, "tid": -914061504, "ts": 1716454223005495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223133257, "dur": 35,
    "args": {
      "External id": 111245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111245,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111245, "pid": 5, "tid": 7, "ts": 1716454223133257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005524, "dur": 8,
    "args": {
      "External id": 111245,
      "cbid": 211, "correlation": 111245
    }
  },
  {
    "ph": "s", "id": 111245, "pid": 76337, "tid": -914061504, "ts": 1716454223005524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223133294, "dur": 51,
    "args": {
      "External id": 111265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111265,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 111265, "pid": 5, "tid": 7, "ts": 1716454223133294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005607, "dur": 12,
    "args": {
      "External id": 111265,
      "cbid": 211, "correlation": 111265
    }
  },
  {
    "ph": "s", "id": 111265, "pid": 76337, "tid": -914061504, "ts": 1716454223005607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223133346, "dur": 5,
    "args": {
      "External id": 111277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111277,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111277, "pid": 5, "tid": 7, "ts": 1716454223133346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005629, "dur": 6,
    "args": {
      "External id": 111277,
      "cbid": 211, "correlation": 111277
    }
  },
  {
    "ph": "s", "id": 111277, "pid": 76337, "tid": -914061504, "ts": 1716454223005629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223133352, "dur": 55,
    "args": {
      "External id": 111280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111280, "pid": 5, "tid": 7, "ts": 1716454223133352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005647, "dur": 6,
    "args": {
      "External id": 111280,
      "cbid": 211, "correlation": 111280
    }
  },
  {
    "ph": "s", "id": 111280, "pid": 76337, "tid": -914061504, "ts": 1716454223005647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005704, "dur": 0,
    "args": {
      "External id": 111291,
      "cbid": 317, "correlation": 111291
    }
  },
  {
    "ph": "f", "id": 111291, "pid": 76337, "tid": -914061504, "ts": 1716454223005704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005705, "dur": 0,
    "args": {
      "External id": 111292,
      "cbid": 203, "correlation": 111292
    }
  },
  {
    "ph": "f", "id": 111292, "pid": 76337, "tid": -914061504, "ts": 1716454223005705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005706, "dur": 0,
    "args": {
      "External id": 111293,
      "cbid": 205, "correlation": 111293
    }
  },
  {
    "ph": "f", "id": 111293, "pid": 76337, "tid": -914061504, "ts": 1716454223005706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005727, "dur": 1,
    "args": {
      "External id": 111297,
      "cbid": 251, "correlation": 111297
    }
  },
  {
    "ph": "f", "id": 111297, "pid": 76337, "tid": -914061504, "ts": 1716454223005727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005729, "dur": 0,
    "args": {
      "External id": 111298,
      "cbid": 251, "correlation": 111298
    }
  },
  {
    "ph": "f", "id": 111298, "pid": 76337, "tid": -914061504, "ts": 1716454223005729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005730, "dur": 0,
    "args": {
      "External id": 111299,
      "cbid": 251, "correlation": 111299
    }
  },
  {
    "ph": "f", "id": 111299, "pid": 76337, "tid": -914061504, "ts": 1716454223005730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005730, "dur": 0,
    "args": {
      "External id": 111300,
      "cbid": 251, "correlation": 111300
    }
  },
  {
    "ph": "f", "id": 111300, "pid": 76337, "tid": -914061504, "ts": 1716454223005730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005731, "dur": 0,
    "args": {
      "External id": 111301,
      "cbid": 251, "correlation": 111301
    }
  },
  {
    "ph": "f", "id": 111301, "pid": 76337, "tid": -914061504, "ts": 1716454223005731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005732, "dur": 0,
    "args": {
      "External id": 111302,
      "cbid": 251, "correlation": 111302
    }
  },
  {
    "ph": "f", "id": 111302, "pid": 76337, "tid": -914061504, "ts": 1716454223005732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005733, "dur": 0,
    "args": {
      "External id": 111303,
      "cbid": 251, "correlation": 111303
    }
  },
  {
    "ph": "f", "id": 111303, "pid": 76337, "tid": -914061504, "ts": 1716454223005733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005734, "dur": 0,
    "args": {
      "External id": 111304,
      "cbid": 251, "correlation": 111304
    }
  },
  {
    "ph": "f", "id": 111304, "pid": 76337, "tid": -914061504, "ts": 1716454223005734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005735, "dur": 0,
    "args": {
      "External id": 111305,
      "cbid": 251, "correlation": 111305
    }
  },
  {
    "ph": "f", "id": 111305, "pid": 76337, "tid": -914061504, "ts": 1716454223005735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223133408, "dur": 115,
    "args": {
      "External id": 111306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111306,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 111306, "pid": 5, "tid": 7, "ts": 1716454223133408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005737, "dur": 14,
    "args": {
      "External id": 111306,
      "cbid": 211, "correlation": 111306
    }
  },
  {
    "ph": "s", "id": 111306, "pid": 76337, "tid": -914061504, "ts": 1716454223005737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223133525, "dur": 59,
    "args": {
      "External id": 111312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111312, "pid": 5, "tid": 7, "ts": 1716454223133525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005773, "dur": 8,
    "args": {
      "External id": 111312,
      "cbid": 211, "correlation": 111312
    }
  },
  {
    "ph": "s", "id": 111312, "pid": 76337, "tid": -914061504, "ts": 1716454223005773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223133585, "dur": 583,
    "args": {
      "External id": 111321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111321,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111321, "pid": 5, "tid": 7, "ts": 1716454223133585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005857, "dur": 14,
    "args": {
      "External id": 111321,
      "cbid": 211, "correlation": 111321
    }
  },
  {
    "ph": "s", "id": 111321, "pid": 76337, "tid": -914061504, "ts": 1716454223005857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223134169, "dur": 180,
    "args": {
      "External id": 111343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111343,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111343, "pid": 5, "tid": 7, "ts": 1716454223134169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223005914, "dur": 10,
    "args": {
      "External id": 111343,
      "cbid": 211, "correlation": 111343
    }
  },
  {
    "ph": "s", "id": 111343, "pid": 76337, "tid": -914061504, "ts": 1716454223005914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006009, "dur": 1,
    "args": {
      "External id": 111354,
      "cbid": 251, "correlation": 111354
    }
  },
  {
    "ph": "f", "id": 111354, "pid": 76337, "tid": -914061504, "ts": 1716454223006009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223134351, "dur": 197,
    "args": {
      "External id": 111355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111355,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111355, "pid": 5, "tid": 7, "ts": 1716454223134351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006014, "dur": 13,
    "args": {
      "External id": 111355,
      "cbid": 211, "correlation": 111355
    }
  },
  {
    "ph": "s", "id": 111355, "pid": 76337, "tid": -914061504, "ts": 1716454223006014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006084, "dur": 1,
    "args": {
      "External id": 111366,
      "cbid": 251, "correlation": 111366
    }
  },
  {
    "ph": "f", "id": 111366, "pid": 76337, "tid": -914061504, "ts": 1716454223006084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223134549, "dur": 188,
    "args": {
      "External id": 111367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111367,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111367, "pid": 5, "tid": 7, "ts": 1716454223134549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006087, "dur": 12,
    "args": {
      "External id": 111367,
      "cbid": 211, "correlation": 111367
    }
  },
  {
    "ph": "s", "id": 111367, "pid": 76337, "tid": -914061504, "ts": 1716454223006087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006151, "dur": 1,
    "args": {
      "External id": 111378,
      "cbid": 251, "correlation": 111378
    }
  },
  {
    "ph": "f", "id": 111378, "pid": 76337, "tid": -914061504, "ts": 1716454223006151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223134739, "dur": 189,
    "args": {
      "External id": 111379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111379,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111379, "pid": 5, "tid": 7, "ts": 1716454223134739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006155, "dur": 11,
    "args": {
      "External id": 111379,
      "cbid": 211, "correlation": 111379
    }
  },
  {
    "ph": "s", "id": 111379, "pid": 76337, "tid": -914061504, "ts": 1716454223006155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223134929, "dur": 18543,
    "args": {
      "External id": 111400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111400,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 111400, "pid": 5, "tid": 7, "ts": 1716454223134929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006235, "dur": 14,
    "args": {
      "External id": 111400,
      "cbid": 211, "correlation": 111400
    }
  },
  {
    "ph": "s", "id": 111400, "pid": 76337, "tid": -914061504, "ts": 1716454223006235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006332, "dur": 1,
    "args": {
      "External id": 111418,
      "cbid": 251, "correlation": 111418
    }
  },
  {
    "ph": "f", "id": 111418, "pid": 76337, "tid": -914061504, "ts": 1716454223006332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223153474, "dur": 204,
    "args": {
      "External id": 111420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111420,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111420, "pid": 5, "tid": 7, "ts": 1716454223153474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006338, "dur": 14,
    "args": {
      "External id": 111420,
      "cbid": 211, "correlation": 111420
    }
  },
  {
    "ph": "s", "id": 111420, "pid": 76337, "tid": -914061504, "ts": 1716454223006338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223153679, "dur": 66,
    "args": {
      "External id": 111428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111428,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111428, "pid": 5, "tid": 7, "ts": 1716454223153679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006409, "dur": 12,
    "args": {
      "External id": 111428,
      "cbid": 211, "correlation": 111428
    }
  },
  {
    "ph": "s", "id": 111428, "pid": 76337, "tid": -914061504, "ts": 1716454223006409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223153747, "dur": 97,
    "args": {
      "External id": 111436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111436,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111436, "pid": 5, "tid": 7, "ts": 1716454223153747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006450, "dur": 77,
    "args": {
      "External id": 111436,
      "cbid": 211, "correlation": 111436
    }
  },
  {
    "ph": "s", "id": 111436, "pid": 76337, "tid": -914061504, "ts": 1716454223006450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223153845, "dur": 54,
    "args": {
      "External id": 111447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111447, "pid": 5, "tid": 7, "ts": 1716454223153845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223006590, "dur": 1848,
    "args": {
      "External id": 111447,
      "cbid": 211, "correlation": 111447
    }
  },
  {
    "ph": "s", "id": 111447, "pid": 76337, "tid": -914061504, "ts": 1716454223006590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223153901, "dur": 92,
    "args": {
      "External id": 111469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111469,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111469, "pid": 5, "tid": 7, "ts": 1716454223153901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008458, "dur": 124,
    "args": {
      "External id": 111469,
      "cbid": 211, "correlation": 111469
    }
  },
  {
    "ph": "s", "id": 111469, "pid": 76337, "tid": -914061504, "ts": 1716454223008458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008660, "dur": 1,
    "args": {
      "External id": 111480,
      "cbid": 251, "correlation": 111480
    }
  },
  {
    "ph": "f", "id": 111480, "pid": 76337, "tid": -914061504, "ts": 1716454223008660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223153994, "dur": 104,
    "args": {
      "External id": 111481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111481,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111481, "pid": 5, "tid": 7, "ts": 1716454223153994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008666, "dur": 13,
    "args": {
      "External id": 111481,
      "cbid": 211, "correlation": 111481
    }
  },
  {
    "ph": "s", "id": 111481, "pid": 76337, "tid": -914061504, "ts": 1716454223008666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008739, "dur": 1,
    "args": {
      "External id": 111492,
      "cbid": 251, "correlation": 111492
    }
  },
  {
    "ph": "f", "id": 111492, "pid": 76337, "tid": -914061504, "ts": 1716454223008739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008743, "dur": 0,
    "args": {
      "External id": 111493,
      "cbid": 251, "correlation": 111493
    }
  },
  {
    "ph": "f", "id": 111493, "pid": 76337, "tid": -914061504, "ts": 1716454223008743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223154099, "dur": 10,
    "args": {
      "External id": 111494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111494,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111494, "pid": 5, "tid": 7, "ts": 1716454223154099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008745, "dur": 12,
    "args": {
      "External id": 111494,
      "cbid": 211, "correlation": 111494
    }
  },
  {
    "ph": "s", "id": 111494, "pid": 76337, "tid": -914061504, "ts": 1716454223008745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223154111, "dur": 5,
    "args": {
      "External id": 111496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111496,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 111496, "pid": 5, "tid": 7, "ts": 1716454223154111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008759, "dur": 6,
    "args": {
      "External id": 111496,
      "cbid": 211, "correlation": 111496
    }
  },
  {
    "ph": "s", "id": 111496, "pid": 76337, "tid": -914061504, "ts": 1716454223008759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008819, "dur": 1,
    "args": {
      "External id": 111507,
      "cbid": 251, "correlation": 111507
    }
  },
  {
    "ph": "f", "id": 111507, "pid": 76337, "tid": -914061504, "ts": 1716454223008819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008823, "dur": 0,
    "args": {
      "External id": 111508,
      "cbid": 251, "correlation": 111508
    }
  },
  {
    "ph": "f", "id": 111508, "pid": 76337, "tid": -914061504, "ts": 1716454223008823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223154117, "dur": 6,
    "args": {
      "External id": 111509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111509,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111509, "pid": 5, "tid": 7, "ts": 1716454223154117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008825, "dur": 12,
    "args": {
      "External id": 111509,
      "cbid": 211, "correlation": 111509
    }
  },
  {
    "ph": "s", "id": 111509, "pid": 76337, "tid": -914061504, "ts": 1716454223008825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223154125, "dur": 3,
    "args": {
      "External id": 111511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111511,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 111511, "pid": 5, "tid": 7, "ts": 1716454223154125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008838, "dur": 7,
    "args": {
      "External id": 111511,
      "cbid": 211, "correlation": 111511
    }
  },
  {
    "ph": "s", "id": 111511, "pid": 76337, "tid": -914061504, "ts": 1716454223008838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223154130, "dur": 157,
    "args": {
      "External id": 111532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111532,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 111532, "pid": 5, "tid": 7, "ts": 1716454223154130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223008914, "dur": 12,
    "args": {
      "External id": 111532,
      "cbid": 211, "correlation": 111532
    }
  },
  {
    "ph": "s", "id": 111532, "pid": 76337, "tid": -914061504, "ts": 1716454223008914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009018, "dur": 1,
    "args": {
      "External id": 111550,
      "cbid": 251, "correlation": 111550
    }
  },
  {
    "ph": "f", "id": 111550, "pid": 76337, "tid": -914061504, "ts": 1716454223009018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223154288, "dur": 106,
    "args": {
      "External id": 111552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111552,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 111552, "pid": 5, "tid": 7, "ts": 1716454223154288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009024, "dur": 14,
    "args": {
      "External id": 111552,
      "cbid": 211, "correlation": 111552
    }
  },
  {
    "ph": "s", "id": 111552, "pid": 76337, "tid": -914061504, "ts": 1716454223009024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223154395, "dur": 35,
    "args": {
      "External id": 111560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111560,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111560, "pid": 5, "tid": 7, "ts": 1716454223154395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009096, "dur": 12,
    "args": {
      "External id": 111560,
      "cbid": 211, "correlation": 111560
    }
  },
  {
    "ph": "s", "id": 111560, "pid": 76337, "tid": -914061504, "ts": 1716454223009096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223154431, "dur": 67,
    "args": {
      "External id": 111568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111568, "pid": 5, "tid": 7, "ts": 1716454223154431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009136, "dur": 10,
    "args": {
      "External id": 111568,
      "cbid": 211, "correlation": 111568
    }
  },
  {
    "ph": "s", "id": 111568, "pid": 76337, "tid": -914061504, "ts": 1716454223009136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223154500, "dur": 92,
    "args": {
      "External id": 111590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111590,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111590, "pid": 5, "tid": 7, "ts": 1716454223154500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009187, "dur": 11,
    "args": {
      "External id": 111590,
      "cbid": 211, "correlation": 111590
    }
  },
  {
    "ph": "s", "id": 111590, "pid": 76337, "tid": -914061504, "ts": 1716454223009187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009273, "dur": 1,
    "args": {
      "External id": 111606,
      "cbid": 251, "correlation": 111606
    }
  },
  {
    "ph": "f", "id": 111606, "pid": 76337, "tid": -914061504, "ts": 1716454223009273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223154593, "dur": 569,
    "args": {
      "External id": 111608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111608,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 111608, "pid": 5, "tid": 7, "ts": 1716454223154593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009279, "dur": 13,
    "args": {
      "External id": 111608,
      "cbid": 211, "correlation": 111608
    }
  },
  {
    "ph": "s", "id": 111608, "pid": 76337, "tid": -914061504, "ts": 1716454223009279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223155163, "dur": 244,
    "args": {
      "External id": 111616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111616,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111616, "pid": 5, "tid": 7, "ts": 1716454223155163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009344, "dur": 12,
    "args": {
      "External id": 111616,
      "cbid": 211, "correlation": 111616
    }
  },
  {
    "ph": "s", "id": 111616, "pid": 76337, "tid": -914061504, "ts": 1716454223009344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223155409, "dur": 253,
    "args": {
      "External id": 111624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111624, "pid": 5, "tid": 7, "ts": 1716454223155409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009374, "dur": 8,
    "args": {
      "External id": 111624,
      "cbid": 211, "correlation": 111624
    }
  },
  {
    "ph": "s", "id": 111624, "pid": 76337, "tid": -914061504, "ts": 1716454223009374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009455, "dur": 1,
    "args": {
      "External id": 111640,
      "cbid": 251, "correlation": 111640
    }
  },
  {
    "ph": "f", "id": 111640, "pid": 76337, "tid": -914061504, "ts": 1716454223009455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009460, "dur": 0,
    "args": {
      "External id": 111642,
      "cbid": 251, "correlation": 111642
    }
  },
  {
    "ph": "f", "id": 111642, "pid": 76337, "tid": -914061504, "ts": 1716454223009460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223155663, "dur": 359,
    "args": {
      "External id": 111643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111643,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111643, "pid": 5, "tid": 7, "ts": 1716454223155663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009463, "dur": 12,
    "args": {
      "External id": 111643,
      "cbid": 211, "correlation": 111643
    }
  },
  {
    "ph": "s", "id": 111643, "pid": 76337, "tid": -914061504, "ts": 1716454223009463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223156024, "dur": 50,
    "args": {
      "External id": 111651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111651,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111651, "pid": 5, "tid": 7, "ts": 1716454223156024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009506, "dur": 167,
    "args": {
      "External id": 111651,
      "cbid": 211, "correlation": 111651
    }
  },
  {
    "ph": "s", "id": 111651, "pid": 76337, "tid": -914061504, "ts": 1716454223009506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223156075, "dur": 157,
    "args": {
      "External id": 111662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111662, "pid": 5, "tid": 7, "ts": 1716454223156075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009731, "dur": 68,
    "args": {
      "External id": 111662,
      "cbid": 211, "correlation": 111662
    }
  },
  {
    "ph": "s", "id": 111662, "pid": 76337, "tid": -914061504, "ts": 1716454223009731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009852, "dur": 0,
    "args": {
      "External id": 111674,
      "cbid": 317, "correlation": 111674
    }
  },
  {
    "ph": "f", "id": 111674, "pid": 76337, "tid": -914061504, "ts": 1716454223009852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009853, "dur": 0,
    "args": {
      "External id": 111675,
      "cbid": 203, "correlation": 111675
    }
  },
  {
    "ph": "f", "id": 111675, "pid": 76337, "tid": -914061504, "ts": 1716454223009853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009854, "dur": 0,
    "args": {
      "External id": 111676,
      "cbid": 205, "correlation": 111676
    }
  },
  {
    "ph": "f", "id": 111676, "pid": 76337, "tid": -914061504, "ts": 1716454223009854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009877, "dur": 1,
    "args": {
      "External id": 111680,
      "cbid": 251, "correlation": 111680
    }
  },
  {
    "ph": "f", "id": 111680, "pid": 76337, "tid": -914061504, "ts": 1716454223009877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009879, "dur": 0,
    "args": {
      "External id": 111681,
      "cbid": 251, "correlation": 111681
    }
  },
  {
    "ph": "f", "id": 111681, "pid": 76337, "tid": -914061504, "ts": 1716454223009879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009880, "dur": 0,
    "args": {
      "External id": 111682,
      "cbid": 251, "correlation": 111682
    }
  },
  {
    "ph": "f", "id": 111682, "pid": 76337, "tid": -914061504, "ts": 1716454223009880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009881, "dur": 0,
    "args": {
      "External id": 111683,
      "cbid": 251, "correlation": 111683
    }
  },
  {
    "ph": "f", "id": 111683, "pid": 76337, "tid": -914061504, "ts": 1716454223009881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009881, "dur": 0,
    "args": {
      "External id": 111684,
      "cbid": 251, "correlation": 111684
    }
  },
  {
    "ph": "f", "id": 111684, "pid": 76337, "tid": -914061504, "ts": 1716454223009881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009882, "dur": 0,
    "args": {
      "External id": 111685,
      "cbid": 251, "correlation": 111685
    }
  },
  {
    "ph": "f", "id": 111685, "pid": 76337, "tid": -914061504, "ts": 1716454223009882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009883, "dur": 0,
    "args": {
      "External id": 111686,
      "cbid": 251, "correlation": 111686
    }
  },
  {
    "ph": "f", "id": 111686, "pid": 76337, "tid": -914061504, "ts": 1716454223009883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009884, "dur": 0,
    "args": {
      "External id": 111687,
      "cbid": 251, "correlation": 111687
    }
  },
  {
    "ph": "f", "id": 111687, "pid": 76337, "tid": -914061504, "ts": 1716454223009884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009885, "dur": 0,
    "args": {
      "External id": 111688,
      "cbid": 251, "correlation": 111688
    }
  },
  {
    "ph": "f", "id": 111688, "pid": 76337, "tid": -914061504, "ts": 1716454223009885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223156234, "dur": 114,
    "args": {
      "External id": 111689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111689,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 111689, "pid": 5, "tid": 7, "ts": 1716454223156234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009887, "dur": 42,
    "args": {
      "External id": 111689,
      "cbid": 211, "correlation": 111689
    }
  },
  {
    "ph": "s", "id": 111689, "pid": 76337, "tid": -914061504, "ts": 1716454223009887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223156349, "dur": 60,
    "args": {
      "External id": 111695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111695,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111695, "pid": 5, "tid": 7, "ts": 1716454223156349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223009952, "dur": 279,
    "args": {
      "External id": 111695,
      "cbid": 211, "correlation": 111695
    }
  },
  {
    "ph": "s", "id": 111695, "pid": 76337, "tid": -914061504, "ts": 1716454223009952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223156410, "dur": 50,
    "args": {
      "External id": 111703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111703,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111703, "pid": 5, "tid": 7, "ts": 1716454223156410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010258, "dur": 10,
    "args": {
      "External id": 111703,
      "cbid": 211, "correlation": 111703
    }
  },
  {
    "ph": "s", "id": 111703, "pid": 76337, "tid": -914061504, "ts": 1716454223010258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223156462, "dur": 53,
    "args": {
      "External id": 111723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111723,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 111723, "pid": 5, "tid": 7, "ts": 1716454223156462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010340, "dur": 13,
    "args": {
      "External id": 111723,
      "cbid": 211, "correlation": 111723
    }
  },
  {
    "ph": "s", "id": 111723, "pid": 76337, "tid": -914061504, "ts": 1716454223010340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223156516, "dur": 5,
    "args": {
      "External id": 111735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111735,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111735, "pid": 5, "tid": 7, "ts": 1716454223156516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010363, "dur": 6,
    "args": {
      "External id": 111735,
      "cbid": 211, "correlation": 111735
    }
  },
  {
    "ph": "s", "id": 111735, "pid": 76337, "tid": -914061504, "ts": 1716454223010363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223156522, "dur": 57,
    "args": {
      "External id": 111738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111738,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111738, "pid": 5, "tid": 7, "ts": 1716454223156522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010381, "dur": 101,
    "args": {
      "External id": 111738,
      "cbid": 211, "correlation": 111738
    }
  },
  {
    "ph": "s", "id": 111738, "pid": 76337, "tid": -914061504, "ts": 1716454223010381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223156580, "dur": 37,
    "args": {
      "External id": 111747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111747,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111747, "pid": 5, "tid": 7, "ts": 1716454223156580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010522, "dur": 10,
    "args": {
      "External id": 111747,
      "cbid": 211, "correlation": 111747
    }
  },
  {
    "ph": "s", "id": 111747, "pid": 76337, "tid": -914061504, "ts": 1716454223010522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010578, "dur": 0,
    "args": {
      "External id": 111757,
      "cbid": 317, "correlation": 111757
    }
  },
  {
    "ph": "f", "id": 111757, "pid": 76337, "tid": -914061504, "ts": 1716454223010578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010579, "dur": 0,
    "args": {
      "External id": 111758,
      "cbid": 203, "correlation": 111758
    }
  },
  {
    "ph": "f", "id": 111758, "pid": 76337, "tid": -914061504, "ts": 1716454223010579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010580, "dur": 0,
    "args": {
      "External id": 111759,
      "cbid": 205, "correlation": 111759
    }
  },
  {
    "ph": "f", "id": 111759, "pid": 76337, "tid": -914061504, "ts": 1716454223010580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223156619, "dur": 40,
    "args": {
      "External id": 111763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111763,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111763, "pid": 5, "tid": 7, "ts": 1716454223156619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010595, "dur": 12,
    "args": {
      "External id": 111763,
      "cbid": 211, "correlation": 111763
    }
  },
  {
    "ph": "s", "id": 111763, "pid": 76337, "tid": -914061504, "ts": 1716454223010595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223156660, "dur": 3,
    "args": {
      "External id": 111765,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111765,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.413793103448276
    }
  },
  {
    "ph": "f", "id": 111765, "pid": 5, "tid": 7, "ts": 1716454223156660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010610, "dur": 17,
    "args": {
      "External id": 111765,
      "cbid": 51, "correlation": 111765
    }
  },
  {
    "ph": "s", "id": 111765, "pid": 76337, "tid": -914061504, "ts": 1716454223010610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010632, "dur": 2,
    "args": {
      "External id": 111767,
      "cbid": 200, "correlation": 111767
    }
  },
  {
    "ph": "f", "id": 111767, "pid": 76337, "tid": -914061504, "ts": 1716454223010632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010635, "dur": 0,
    "args": {
      "External id": 111768,
      "cbid": 200, "correlation": 111768
    }
  },
  {
    "ph": "f", "id": 111768, "pid": 76337, "tid": -914061504, "ts": 1716454223010635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010635, "dur": 0,
    "args": {
      "External id": 111769,
      "cbid": 200, "correlation": 111769
    }
  },
  {
    "ph": "f", "id": 111769, "pid": 76337, "tid": -914061504, "ts": 1716454223010635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010636, "dur": 0,
    "args": {
      "External id": 111770,
      "cbid": 200, "correlation": 111770
    }
  },
  {
    "ph": "f", "id": 111770, "pid": 76337, "tid": -914061504, "ts": 1716454223010636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010637, "dur": 4,
    "args": {
      "External id": 111771,
      "cbid": 15, "correlation": 111771
    }
  },
  {
    "ph": "f", "id": 111771, "pid": 76337, "tid": -914061504, "ts": 1716454223010637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010641, "dur": 1,
    "args": {
      "External id": 111772,
      "cbid": 251, "correlation": 111772
    }
  },
  {
    "ph": "f", "id": 111772, "pid": 76337, "tid": -914061504, "ts": 1716454223010641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223156665, "dur": 24,
    "args": {
      "External id": 111773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111773,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111773, "pid": 5, "tid": 7, "ts": 1716454223156665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010644, "dur": 8,
    "args": {
      "External id": 111773,
      "cbid": 211, "correlation": 111773
    }
  },
  {
    "ph": "s", "id": 111773, "pid": 76337, "tid": -914061504, "ts": 1716454223010644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223156690, "dur": 4,
    "args": {
      "External id": 111775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 111775, "pid": 5, "tid": 7, "ts": 1716454223156690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010658, "dur": 7,
    "args": {
      "External id": 111775,
      "cbid": 211, "correlation": 111775
    }
  },
  {
    "ph": "s", "id": 111775, "pid": 76337, "tid": -914061504, "ts": 1716454223010658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010669, "dur": 0,
    "args": {
      "External id": 111776,
      "cbid": 51, "correlation": 111776
    }
  },
  {
    "ph": "s", "id": 111776, "pid": 76337, "tid": -914061504, "ts": 1716454223010669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223156696, "dur": 187,
    "args": {
      "External id": 111777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111777,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111777, "pid": 5, "tid": 7, "ts": 1716454223156696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010670, "dur": 202,
    "args": {
      "External id": 111777,
      "cbid": 211, "correlation": 111777
    }
  },
  {
    "ph": "s", "id": 111777, "pid": 76337, "tid": -914061504, "ts": 1716454223010670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223156884, "dur": 6,
    "args": {
      "External id": 111778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111778, "pid": 5, "tid": 7, "ts": 1716454223156884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010875, "dur": 6,
    "args": {
      "External id": 111778,
      "cbid": 211, "correlation": 111778
    }
  },
  {
    "ph": "s", "id": 111778, "pid": 76337, "tid": -914061504, "ts": 1716454223010875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223156891, "dur": 5,
    "args": {
      "External id": 111784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 111784, "pid": 5, "tid": 7, "ts": 1716454223156891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223010906, "dur": 9,
    "args": {
      "External id": 111784,
      "cbid": 211, "correlation": 111784
    }
  },
  {
    "ph": "s", "id": 111784, "pid": 76337, "tid": -914061504, "ts": 1716454223010906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223156897, "dur": 3,
    "args": {
      "External id": 111792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111792,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111792, "pid": 5, "tid": 7, "ts": 1716454223156897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223012585, "dur": 14,
    "args": {
      "External id": 111792,
      "cbid": 211, "correlation": 111792
    }
  },
  {
    "ph": "s", "id": 111792, "pid": 76337, "tid": -914061504, "ts": 1716454223012585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223156901, "dur": 3,
    "args": {
      "External id": 111800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111800,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111800, "pid": 5, "tid": 7, "ts": 1716454223156901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223012625, "dur": 10,
    "args": {
      "External id": 111800,
      "cbid": 211, "correlation": 111800
    }
  },
  {
    "ph": "s", "id": 111800, "pid": 76337, "tid": -914061504, "ts": 1716454223012625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223156906, "dur": 3,
    "args": {
      "External id": 111808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111808,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111808, "pid": 5, "tid": 7, "ts": 1716454223156906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223012651, "dur": 8,
    "args": {
      "External id": 111808,
      "cbid": 211, "correlation": 111808
    }
  },
  {
    "ph": "s", "id": 111808, "pid": 76337, "tid": -914061504, "ts": 1716454223012651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223156910, "dur": 3,
    "args": {
      "External id": 111817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111817,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111817, "pid": 5, "tid": 7, "ts": 1716454223156910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223012825, "dur": 14,
    "args": {
      "External id": 111817,
      "cbid": 211, "correlation": 111817
    }
  },
  {
    "ph": "s", "id": 111817, "pid": 76337, "tid": -914061504, "ts": 1716454223012825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223156914, "dur": 3,
    "args": {
      "External id": 111826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111826,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111826, "pid": 5, "tid": 7, "ts": 1716454223156914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223012856, "dur": 7,
    "args": {
      "External id": 111826,
      "cbid": 211, "correlation": 111826
    }
  },
  {
    "ph": "s", "id": 111826, "pid": 76337, "tid": -914061504, "ts": 1716454223012856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223156919, "dur": 3,
    "args": {
      "External id": 111834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111834,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111834, "pid": 5, "tid": 7, "ts": 1716454223156919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223012881, "dur": 8,
    "args": {
      "External id": 111834,
      "cbid": 211, "correlation": 111834
    }
  },
  {
    "ph": "s", "id": 111834, "pid": 76337, "tid": -914061504, "ts": 1716454223012881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223156923, "dur": 3,
    "args": {
      "External id": 111842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111842,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111842, "pid": 5, "tid": 7, "ts": 1716454223156923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223013152, "dur": 16,
    "args": {
      "External id": 111842,
      "cbid": 211, "correlation": 111842
    }
  },
  {
    "ph": "s", "id": 111842, "pid": 76337, "tid": -914061504, "ts": 1716454223013152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223156927, "dur": 3,
    "args": {
      "External id": 111850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111850,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 111850, "pid": 5, "tid": 7, "ts": 1716454223156927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223013184, "dur": 8,
    "args": {
      "External id": 111850,
      "cbid": 211, "correlation": 111850
    }
  },
  {
    "ph": "s", "id": 111850, "pid": 76337, "tid": -914061504, "ts": 1716454223013184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223156932, "dur": 1,
    "args": {
      "External id": 111860,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111860,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 111860, "pid": 5, "tid": 7, "ts": 1716454223156932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223013248, "dur": 35,
    "args": {
      "External id": 111860,
      "cbid": 41, "correlation": 111860
    }
  },
  {
    "ph": "s", "id": 111860, "pid": 76337, "tid": -914061504, "ts": 1716454223013248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223013284, "dur": 143664,
    "args": {
      "External id": 111861,
      "cbid": 131, "correlation": 111861
    }
  },
  {
    "ph": "f", "id": 111861, "pid": 76337, "tid": -914061504, "ts": 1716454223013284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223157100, "dur": 3,
    "args": {
      "External id": 111869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111869, "pid": 5, "tid": 7, "ts": 1716454223157100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157077, "dur": 24,
    "args": {
      "External id": 111869,
      "cbid": 211, "correlation": 111869
    }
  },
  {
    "ph": "s", "id": 111869, "pid": 76337, "tid": -914061504, "ts": 1716454223157077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157189, "dur": 3,
    "args": {
      "External id": 111878,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111878,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 111878, "pid": 5, "tid": 7, "ts": 1716454223157189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157161, "dur": 28,
    "args": {
      "External id": 111878,
      "cbid": 41, "correlation": 111878
    }
  },
  {
    "ph": "s", "id": 111878, "pid": 76337, "tid": -914061504, "ts": 1716454223157161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223157282, "dur": 4,
    "args": {
      "External id": 111888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111888,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 111888, "pid": 5, "tid": 7, "ts": 1716454223157282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157266, "dur": 17,
    "args": {
      "External id": 111888,
      "cbid": 211, "correlation": 111888
    }
  },
  {
    "ph": "s", "id": 111888, "pid": 76337, "tid": -914061504, "ts": 1716454223157266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157351, "dur": 1,
    "args": {
      "External id": 111898,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111898,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 111898, "pid": 5, "tid": 7, "ts": 1716454223157351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157333, "dur": 16,
    "args": {
      "External id": 111898,
      "cbid": 41, "correlation": 111898
    }
  },
  {
    "ph": "s", "id": 111898, "pid": 76337, "tid": -914061504, "ts": 1716454223157333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157350, "dur": 8,
    "args": {
      "External id": 111899,
      "cbid": 131, "correlation": 111899
    }
  },
  {
    "ph": "f", "id": 111899, "pid": 76337, "tid": -914061504, "ts": 1716454223157350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157419, "dur": 3,
    "args": {
      "External id": 111906,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111906,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.72
    }
  },
  {
    "ph": "f", "id": 111906, "pid": 5, "tid": 7, "ts": 1716454223157419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157397, "dur": 22,
    "args": {
      "External id": 111906,
      "cbid": 41, "correlation": 111906
    }
  },
  {
    "ph": "s", "id": 111906, "pid": 76337, "tid": -914061504, "ts": 1716454223157397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157508, "dur": 3,
    "args": {
      "External id": 111925,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111925,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005320917858330562
    }
  },
  {
    "ph": "f", "id": 111925, "pid": 5, "tid": 7, "ts": 1716454223157508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157489, "dur": 17,
    "args": {
      "External id": 111925,
      "cbid": 41, "correlation": 111925
    }
  },
  {
    "ph": "s", "id": 111925, "pid": 76337, "tid": -914061504, "ts": 1716454223157489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223157547, "dur": 3,
    "args": {
      "External id": 111931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111931, "pid": 5, "tid": 7, "ts": 1716454223157547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157535, "dur": 12,
    "args": {
      "External id": 111931,
      "cbid": 211, "correlation": 111931
    }
  },
  {
    "ph": "s", "id": 111931, "pid": 76337, "tid": -914061504, "ts": 1716454223157535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223157561, "dur": 6,
    "args": {
      "External id": 111933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111933,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 111933, "pid": 5, "tid": 7, "ts": 1716454223157561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157551, "dur": 9,
    "args": {
      "External id": 111933,
      "cbid": 211, "correlation": 111933
    }
  },
  {
    "ph": "s", "id": 111933, "pid": 76337, "tid": -914061504, "ts": 1716454223157551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454223157570, "dur": 3,
    "args": {
      "External id": 111935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111935, "pid": 5, "tid": 7, "ts": 1716454223157570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157562, "dur": 6,
    "args": {
      "External id": 111935,
      "cbid": 211, "correlation": 111935
    }
  },
  {
    "ph": "s", "id": 111935, "pid": 76337, "tid": -914061504, "ts": 1716454223157562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157605, "dur": 2,
    "args": {
      "External id": 111943,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111943,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002777777777777778
    }
  },
  {
    "ph": "f", "id": 111943, "pid": 5, "tid": 7, "ts": 1716454223157605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157590, "dur": 14,
    "args": {
      "External id": 111943,
      "cbid": 41, "correlation": 111943
    }
  },
  {
    "ph": "s", "id": 111943, "pid": 76337, "tid": -914061504, "ts": 1716454223157590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223157653, "dur": 3,
    "args": {
      "External id": 111957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111957, "pid": 5, "tid": 7, "ts": 1716454223157653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157642, "dur": 12,
    "args": {
      "External id": 111957,
      "cbid": 211, "correlation": 111957
    }
  },
  {
    "ph": "s", "id": 111957, "pid": 76337, "tid": -914061504, "ts": 1716454223157642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223157673, "dur": 2,
    "args": {
      "External id": 111971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111971, "pid": 5, "tid": 7, "ts": 1716454223157673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157665, "dur": 7,
    "args": {
      "External id": 111971,
      "cbid": 211, "correlation": 111971
    }
  },
  {
    "ph": "s", "id": 111971, "pid": 76337, "tid": -914061504, "ts": 1716454223157665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223157707, "dur": 6,
    "args": {
      "External id": 111978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111978,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111978, "pid": 5, "tid": 7, "ts": 1716454223157707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157696, "dur": 11,
    "args": {
      "External id": 111978,
      "cbid": 211, "correlation": 111978
    }
  },
  {
    "ph": "s", "id": 111978, "pid": 76337, "tid": -914061504, "ts": 1716454223157696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223157718, "dur": 6,
    "args": {
      "External id": 111981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111981,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111981, "pid": 5, "tid": 7, "ts": 1716454223157718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157710, "dur": 7,
    "args": {
      "External id": 111981,
      "cbid": 211, "correlation": 111981
    }
  },
  {
    "ph": "s", "id": 111981, "pid": 76337, "tid": -914061504, "ts": 1716454223157710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223157727, "dur": 3,
    "args": {
      "External id": 111983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 111983,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 111983, "pid": 5, "tid": 7, "ts": 1716454223157727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157720, "dur": 7,
    "args": {
      "External id": 111983,
      "cbid": 211, "correlation": 111983
    }
  },
  {
    "ph": "s", "id": 111983, "pid": 76337, "tid": -914061504, "ts": 1716454223157720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157747, "dur": 2,
    "args": {
      "External id": 111986,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 111986,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002777777777777778
    }
  },
  {
    "ph": "f", "id": 111986, "pid": 5, "tid": 7, "ts": 1716454223157747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157735, "dur": 12,
    "args": {
      "External id": 111986,
      "cbid": 41, "correlation": 111986
    }
  },
  {
    "ph": "s", "id": 111986, "pid": 76337, "tid": -914061504, "ts": 1716454223157735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223157799, "dur": 4,
    "args": {
      "External id": 112002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112002,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112002, "pid": 5, "tid": 7, "ts": 1716454223157799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157787, "dur": 13,
    "args": {
      "External id": 112002,
      "cbid": 211, "correlation": 112002
    }
  },
  {
    "ph": "s", "id": 112002, "pid": 76337, "tid": -914061504, "ts": 1716454223157787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223157821, "dur": 3,
    "args": {
      "External id": 112007,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112007,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003155569580309246
    }
  },
  {
    "ph": "f", "id": 112007, "pid": 5, "tid": 7, "ts": 1716454223157821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157805, "dur": 15,
    "args": {
      "External id": 112007,
      "cbid": 41, "correlation": 112007
    }
  },
  {
    "ph": "s", "id": 112007, "pid": 76337, "tid": -914061504, "ts": 1716454223157805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223157848, "dur": 1,
    "args": {
      "External id": 112013,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112013,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 112013, "pid": 5, "tid": 7, "ts": 1716454223157848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157830, "dur": 27,
    "args": {
      "External id": 112013,
      "cbid": 41, "correlation": 112013
    }
  },
  {
    "ph": "s", "id": 112013, "pid": 76337, "tid": -914061504, "ts": 1716454223157830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157858, "dur": 4,
    "args": {
      "External id": 112014,
      "cbid": 131, "correlation": 112014
    }
  },
  {
    "ph": "f", "id": 112014, "pid": 76337, "tid": -914061504, "ts": 1716454223157858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223157908, "dur": 3,
    "args": {
      "External id": 112022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112022,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112022, "pid": 5, "tid": 7, "ts": 1716454223157908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157896, "dur": 12,
    "args": {
      "External id": 112022,
      "cbid": 211, "correlation": 112022
    }
  },
  {
    "ph": "s", "id": 112022, "pid": 76337, "tid": -914061504, "ts": 1716454223157896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223157938, "dur": 3,
    "args": {
      "External id": 112032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112032,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112032, "pid": 5, "tid": 7, "ts": 1716454223157938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157929, "dur": 8,
    "args": {
      "External id": 112032,
      "cbid": 211, "correlation": 112032
    }
  },
  {
    "ph": "s", "id": 112032, "pid": 76337, "tid": -914061504, "ts": 1716454223157929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223157964, "dur": 3,
    "args": {
      "External id": 112041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112041,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112041, "pid": 5, "tid": 7, "ts": 1716454223157964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223157953, "dur": 11,
    "args": {
      "External id": 112041,
      "cbid": 211, "correlation": 112041
    }
  },
  {
    "ph": "s", "id": 112041, "pid": 76337, "tid": -914061504, "ts": 1716454223157953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223158086, "dur": 12,
    "args": {
      "External id": 112051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112051,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112051, "pid": 5, "tid": 7, "ts": 1716454223158086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158072, "dur": 15,
    "args": {
      "External id": 112051,
      "cbid": 211, "correlation": 112051
    }
  },
  {
    "ph": "s", "id": 112051, "pid": 76337, "tid": -914061504, "ts": 1716454223158072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223158128, "dur": 3,
    "args": {
      "External id": 112059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112059,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112059, "pid": 5, "tid": 7, "ts": 1716454223158128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158119, "dur": 8,
    "args": {
      "External id": 112059,
      "cbid": 211, "correlation": 112059
    }
  },
  {
    "ph": "s", "id": 112059, "pid": 76337, "tid": -914061504, "ts": 1716454223158119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223158172, "dur": 11,
    "args": {
      "External id": 112069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112069,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112069, "pid": 5, "tid": 7, "ts": 1716454223158172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158161, "dur": 11,
    "args": {
      "External id": 112069,
      "cbid": 211, "correlation": 112069
    }
  },
  {
    "ph": "s", "id": 112069, "pid": 76337, "tid": -914061504, "ts": 1716454223158161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223158202, "dur": 10,
    "args": {
      "External id": 112077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112077,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112077, "pid": 5, "tid": 7, "ts": 1716454223158202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158192, "dur": 9,
    "args": {
      "External id": 112077,
      "cbid": 211, "correlation": 112077
    }
  },
  {
    "ph": "s", "id": 112077, "pid": 76337, "tid": -914061504, "ts": 1716454223158192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223158229, "dur": 3,
    "args": {
      "External id": 112086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112086, "pid": 5, "tid": 7, "ts": 1716454223158229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158220, "dur": 9,
    "args": {
      "External id": 112086,
      "cbid": 211, "correlation": 112086
    }
  },
  {
    "ph": "s", "id": 112086, "pid": 76337, "tid": -914061504, "ts": 1716454223158220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223158253, "dur": 5,
    "args": {
      "External id": 112095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112095,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112095, "pid": 5, "tid": 7, "ts": 1716454223158253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158244, "dur": 8,
    "args": {
      "External id": 112095,
      "cbid": 211, "correlation": 112095
    }
  },
  {
    "ph": "s", "id": 112095, "pid": 76337, "tid": -914061504, "ts": 1716454223158244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223158291, "dur": 8,
    "args": {
      "External id": 112105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112105,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112105, "pid": 5, "tid": 7, "ts": 1716454223158291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158281, "dur": 10,
    "args": {
      "External id": 112105,
      "cbid": 211, "correlation": 112105
    }
  },
  {
    "ph": "s", "id": 112105, "pid": 76337, "tid": -914061504, "ts": 1716454223158281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223158606, "dur": 3,
    "args": {
      "External id": 112114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112114,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112114, "pid": 5, "tid": 7, "ts": 1716454223158606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158591, "dur": 16,
    "args": {
      "External id": 112114,
      "cbid": 211, "correlation": 112114
    }
  },
  {
    "ph": "s", "id": 112114, "pid": 76337, "tid": -914061504, "ts": 1716454223158591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223158635, "dur": 3,
    "args": {
      "External id": 112122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112122,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112122, "pid": 5, "tid": 7, "ts": 1716454223158635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158625, "dur": 10,
    "args": {
      "External id": 112122,
      "cbid": 211, "correlation": 112122
    }
  },
  {
    "ph": "s", "id": 112122, "pid": 76337, "tid": -914061504, "ts": 1716454223158625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223158686, "dur": 1,
    "args": {
      "External id": 112132,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112132,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 112132, "pid": 5, "tid": 7, "ts": 1716454223158686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158671, "dur": 13,
    "args": {
      "External id": 112132,
      "cbid": 41, "correlation": 112132
    }
  },
  {
    "ph": "s", "id": 112132, "pid": 76337, "tid": -914061504, "ts": 1716454223158671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158685, "dur": 8,
    "args": {
      "External id": 112133,
      "cbid": 131, "correlation": 112133
    }
  },
  {
    "ph": "f", "id": 112133, "pid": 76337, "tid": -914061504, "ts": 1716454223158685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223158775, "dur": 2,
    "args": {
      "External id": 112141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112141, "pid": 5, "tid": 7, "ts": 1716454223158775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158761, "dur": 14,
    "args": {
      "External id": 112141,
      "cbid": 211, "correlation": 112141
    }
  },
  {
    "ph": "s", "id": 112141, "pid": 76337, "tid": -914061504, "ts": 1716454223158761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223158847, "dur": 3,
    "args": {
      "External id": 112150,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112150,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 112150, "pid": 5, "tid": 7, "ts": 1716454223158847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158830, "dur": 17,
    "args": {
      "External id": 112150,
      "cbid": 41, "correlation": 112150
    }
  },
  {
    "ph": "s", "id": 112150, "pid": 76337, "tid": -914061504, "ts": 1716454223158830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223158918, "dur": 3,
    "args": {
      "External id": 112160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112160,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112160, "pid": 5, "tid": 7, "ts": 1716454223158918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158904, "dur": 15,
    "args": {
      "External id": 112160,
      "cbid": 211, "correlation": 112160
    }
  },
  {
    "ph": "s", "id": 112160, "pid": 76337, "tid": -914061504, "ts": 1716454223158904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223158970, "dur": 1,
    "args": {
      "External id": 112170,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112170,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 112170, "pid": 5, "tid": 7, "ts": 1716454223158970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158956, "dur": 12,
    "args": {
      "External id": 112170,
      "cbid": 41, "correlation": 112170
    }
  },
  {
    "ph": "s", "id": 112170, "pid": 76337, "tid": -914061504, "ts": 1716454223158956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223158969, "dur": 14,
    "args": {
      "External id": 112171,
      "cbid": 131, "correlation": 112171
    }
  },
  {
    "ph": "f", "id": 112171, "pid": 76337, "tid": -914061504, "ts": 1716454223158969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223159038, "dur": 3,
    "args": {
      "External id": 112178,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112178,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 112178, "pid": 5, "tid": 7, "ts": 1716454223159038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159019, "dur": 19,
    "args": {
      "External id": 112178,
      "cbid": 41, "correlation": 112178
    }
  },
  {
    "ph": "s", "id": 112178, "pid": 76337, "tid": -914061504, "ts": 1716454223159019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223159087, "dur": 1,
    "args": {
      "External id": 112189,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112189,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 112189, "pid": 5, "tid": 7, "ts": 1716454223159087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159074, "dur": 10,
    "args": {
      "External id": 112189,
      "cbid": 41, "correlation": 112189
    }
  },
  {
    "ph": "s", "id": 112189, "pid": 76337, "tid": -914061504, "ts": 1716454223159074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159086, "dur": 8,
    "args": {
      "External id": 112190,
      "cbid": 131, "correlation": 112190
    }
  },
  {
    "ph": "f", "id": 112190, "pid": 76337, "tid": -914061504, "ts": 1716454223159086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223159136, "dur": 3,
    "args": {
      "External id": 112198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112198,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112198, "pid": 5, "tid": 7, "ts": 1716454223159136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159123, "dur": 12,
    "args": {
      "External id": 112198,
      "cbid": 211, "correlation": 112198
    }
  },
  {
    "ph": "s", "id": 112198, "pid": 76337, "tid": -914061504, "ts": 1716454223159123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223159164, "dur": 3,
    "args": {
      "External id": 112208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112208,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112208, "pid": 5, "tid": 7, "ts": 1716454223159164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159155, "dur": 8,
    "args": {
      "External id": 112208,
      "cbid": 211, "correlation": 112208
    }
  },
  {
    "ph": "s", "id": 112208, "pid": 76337, "tid": -914061504, "ts": 1716454223159155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223159186, "dur": 3,
    "args": {
      "External id": 112217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112217,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112217, "pid": 5, "tid": 7, "ts": 1716454223159186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159178, "dur": 7,
    "args": {
      "External id": 112217,
      "cbid": 211, "correlation": 112217
    }
  },
  {
    "ph": "s", "id": 112217, "pid": 76337, "tid": -914061504, "ts": 1716454223159178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223159254, "dur": 5,
    "args": {
      "External id": 112225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112225,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112225, "pid": 5, "tid": 7, "ts": 1716454223159254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159241, "dur": 14,
    "args": {
      "External id": 112225,
      "cbid": 211, "correlation": 112225
    }
  },
  {
    "ph": "s", "id": 112225, "pid": 76337, "tid": -914061504, "ts": 1716454223159241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223159294, "dur": 3,
    "args": {
      "External id": 112234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112234,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112234, "pid": 5, "tid": 7, "ts": 1716454223159294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159284, "dur": 9,
    "args": {
      "External id": 112234,
      "cbid": 211, "correlation": 112234
    }
  },
  {
    "ph": "s", "id": 112234, "pid": 76337, "tid": -914061504, "ts": 1716454223159284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223159318, "dur": 3,
    "args": {
      "External id": 112243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112243,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112243, "pid": 5, "tid": 7, "ts": 1716454223159318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159309, "dur": 7,
    "args": {
      "External id": 112243,
      "cbid": 211, "correlation": 112243
    }
  },
  {
    "ph": "s", "id": 112243, "pid": 76337, "tid": -914061504, "ts": 1716454223159309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223159379, "dur": 3,
    "args": {
      "External id": 112251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112251,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112251, "pid": 5, "tid": 7, "ts": 1716454223159379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159369, "dur": 10,
    "args": {
      "External id": 112251,
      "cbid": 211, "correlation": 112251
    }
  },
  {
    "ph": "s", "id": 112251, "pid": 76337, "tid": -914061504, "ts": 1716454223159369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223159438, "dur": 1,
    "args": {
      "External id": 112259,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112259,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0043859649122807015
    }
  },
  {
    "ph": "f", "id": 112259, "pid": 5, "tid": 7, "ts": 1716454223159438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159422, "dur": 25,
    "args": {
      "External id": 112259,
      "cbid": 41, "correlation": 112259
    }
  },
  {
    "ph": "s", "id": 112259, "pid": 76337, "tid": -914061504, "ts": 1716454223159422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159448, "dur": 4,
    "args": {
      "External id": 112260,
      "cbid": 131, "correlation": 112260
    }
  },
  {
    "ph": "f", "id": 112260, "pid": 76337, "tid": -914061504, "ts": 1716454223159448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223159510, "dur": 1,
    "args": {
      "External id": 112270,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112270,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 112270, "pid": 5, "tid": 7, "ts": 1716454223159510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159497, "dur": 11,
    "args": {
      "External id": 112270,
      "cbid": 41, "correlation": 112270
    }
  },
  {
    "ph": "s", "id": 112270, "pid": 76337, "tid": -914061504, "ts": 1716454223159497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159509, "dur": 8,
    "args": {
      "External id": 112271,
      "cbid": 131, "correlation": 112271
    }
  },
  {
    "ph": "f", "id": 112271, "pid": 76337, "tid": -914061504, "ts": 1716454223159509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223159567, "dur": 1,
    "args": {
      "External id": 112280,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112280,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 112280, "pid": 5, "tid": 7, "ts": 1716454223159567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159556, "dur": 9,
    "args": {
      "External id": 112280,
      "cbid": 41, "correlation": 112280
    }
  },
  {
    "ph": "s", "id": 112280, "pid": 76337, "tid": -914061504, "ts": 1716454223159556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159565, "dur": 8,
    "args": {
      "External id": 112281,
      "cbid": 131, "correlation": 112281
    }
  },
  {
    "ph": "f", "id": 112281, "pid": 76337, "tid": -914061504, "ts": 1716454223159565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223159640, "dur": 4,
    "args": {
      "External id": 112288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112288, "pid": 5, "tid": 7, "ts": 1716454223159640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159623, "dur": 17,
    "args": {
      "External id": 112288,
      "cbid": 211, "correlation": 112288
    }
  },
  {
    "ph": "s", "id": 112288, "pid": 76337, "tid": -914061504, "ts": 1716454223159623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454223159676, "dur": 4,
    "args": {
      "External id": 112308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112308,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112308, "pid": 5, "tid": 7, "ts": 1716454223159676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159665, "dur": 11,
    "args": {
      "External id": 112308,
      "cbid": 211, "correlation": 112308
    }
  },
  {
    "ph": "s", "id": 112308, "pid": 76337, "tid": -914061504, "ts": 1716454223159665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159677, "dur": 0,
    "args": {
      "External id": 112309,
      "cbid": 11, "correlation": 112309
    }
  },
  {
    "ph": "f", "id": 112309, "pid": 76337, "tid": -914061504, "ts": 1716454223159677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159678, "dur": 0,
    "args": {
      "External id": 112310,
      "cbid": 11, "correlation": 112310
    }
  },
  {
    "ph": "f", "id": 112310, "pid": 76337, "tid": -914061504, "ts": 1716454223159678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223159691, "dur": 1,
    "args": {
      "External id": 112313,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112313,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 112313, "pid": 5, "tid": 7, "ts": 1716454223159691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159679, "dur": 21,
    "args": {
      "External id": 112313,
      "cbid": 41, "correlation": 112313
    }
  },
  {
    "ph": "s", "id": 112313, "pid": 76337, "tid": -914061504, "ts": 1716454223159679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159700, "dur": 3,
    "args": {
      "External id": 112314,
      "cbid": 131, "correlation": 112314
    }
  },
  {
    "ph": "f", "id": 112314, "pid": 76337, "tid": -914061504, "ts": 1716454223159700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223159728, "dur": 3,
    "args": {
      "External id": 112338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112338, "pid": 5, "tid": 7, "ts": 1716454223159728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159719, "dur": 9,
    "args": {
      "External id": 112338,
      "cbid": 211, "correlation": 112338
    }
  },
  {
    "ph": "s", "id": 112338, "pid": 76337, "tid": -914061504, "ts": 1716454223159719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159728, "dur": 0,
    "args": {
      "External id": 112339,
      "cbid": 11, "correlation": 112339
    }
  },
  {
    "ph": "f", "id": 112339, "pid": 76337, "tid": -914061504, "ts": 1716454223159728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159729, "dur": 0,
    "args": {
      "External id": 112340,
      "cbid": 11, "correlation": 112340
    }
  },
  {
    "ph": "f", "id": 112340, "pid": 76337, "tid": -914061504, "ts": 1716454223159729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159730, "dur": 1,
    "args": {
      "External id": 112342,
      "cbid": 200, "correlation": 112342
    }
  },
  {
    "ph": "f", "id": 112342, "pid": 76337, "tid": -914061504, "ts": 1716454223159730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223159741, "dur": 4,
    "args": {
      "External id": 112344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112344,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112344, "pid": 5, "tid": 7, "ts": 1716454223159741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159733, "dur": 8,
    "args": {
      "External id": 112344,
      "cbid": 211, "correlation": 112344
    }
  },
  {
    "ph": "s", "id": 112344, "pid": 76337, "tid": -914061504, "ts": 1716454223159733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159741, "dur": 0,
    "args": {
      "External id": 112345,
      "cbid": 11, "correlation": 112345
    }
  },
  {
    "ph": "f", "id": 112345, "pid": 76337, "tid": -914061504, "ts": 1716454223159741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159742, "dur": 0,
    "args": {
      "External id": 112346,
      "cbid": 11, "correlation": 112346
    }
  },
  {
    "ph": "f", "id": 112346, "pid": 76337, "tid": -914061504, "ts": 1716454223159742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223159780, "dur": 1,
    "args": {
      "External id": 112353,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112353,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004901960784313725
    }
  },
  {
    "ph": "f", "id": 112353, "pid": 5, "tid": 7, "ts": 1716454223159780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159768, "dur": 20,
    "args": {
      "External id": 112353,
      "cbid": 41, "correlation": 112353
    }
  },
  {
    "ph": "s", "id": 112353, "pid": 76337, "tid": -914061504, "ts": 1716454223159768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159789, "dur": 3,
    "args": {
      "External id": 112354,
      "cbid": 131, "correlation": 112354
    }
  },
  {
    "ph": "f", "id": 112354, "pid": 76337, "tid": -914061504, "ts": 1716454223159789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223159840, "dur": 1,
    "args": {
      "External id": 112364,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 112364,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 112364, "pid": 5, "tid": 7, "ts": 1716454223159840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159827, "dur": 11,
    "args": {
      "External id": 112364,
      "cbid": 41, "correlation": 112364
    }
  },
  {
    "ph": "s", "id": 112364, "pid": 76337, "tid": -914061504, "ts": 1716454223159827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159839, "dur": 8,
    "args": {
      "External id": 112365,
      "cbid": 131, "correlation": 112365
    }
  },
  {
    "ph": "f", "id": 112365, "pid": 76337, "tid": -914061504, "ts": 1716454223159839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223159909, "dur": 5,
    "args": {
      "External id": 112372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112372,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112372, "pid": 5, "tid": 7, "ts": 1716454223159909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159894, "dur": 15,
    "args": {
      "External id": 112372,
      "cbid": 211, "correlation": 112372
    }
  },
  {
    "ph": "s", "id": 112372, "pid": 76337, "tid": -914061504, "ts": 1716454223159894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223159986, "dur": 3,
    "args": {
      "External id": 112381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112381,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112381, "pid": 5, "tid": 7, "ts": 1716454223159986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223159966, "dur": 21,
    "args": {
      "External id": 112381,
      "cbid": 211, "correlation": 112381
    }
  },
  {
    "ph": "s", "id": 112381, "pid": 76337, "tid": -914061504, "ts": 1716454223159966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160025, "dur": 3,
    "args": {
      "External id": 112389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112389,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112389, "pid": 5, "tid": 7, "ts": 1716454223160025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160014, "dur": 10,
    "args": {
      "External id": 112389,
      "cbid": 211, "correlation": 112389
    }
  },
  {
    "ph": "s", "id": 112389, "pid": 76337, "tid": -914061504, "ts": 1716454223160014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160059, "dur": 4,
    "args": {
      "External id": 112397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112397,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112397, "pid": 5, "tid": 7, "ts": 1716454223160059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160047, "dur": 12,
    "args": {
      "External id": 112397,
      "cbid": 211, "correlation": 112397
    }
  },
  {
    "ph": "s", "id": 112397, "pid": 76337, "tid": -914061504, "ts": 1716454223160047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160090, "dur": 4,
    "args": {
      "External id": 112405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112405,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112405, "pid": 5, "tid": 7, "ts": 1716454223160090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160078, "dur": 11,
    "args": {
      "External id": 112405,
      "cbid": 211, "correlation": 112405
    }
  },
  {
    "ph": "s", "id": 112405, "pid": 76337, "tid": -914061504, "ts": 1716454223160078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160116, "dur": 3,
    "args": {
      "External id": 112413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112413,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112413, "pid": 5, "tid": 7, "ts": 1716454223160116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160107, "dur": 8,
    "args": {
      "External id": 112413,
      "cbid": 211, "correlation": 112413
    }
  },
  {
    "ph": "s", "id": 112413, "pid": 76337, "tid": -914061504, "ts": 1716454223160107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223160142, "dur": 3,
    "args": {
      "External id": 112421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112421,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112421, "pid": 5, "tid": 7, "ts": 1716454223160142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160132, "dur": 9,
    "args": {
      "External id": 112421,
      "cbid": 211, "correlation": 112421
    }
  },
  {
    "ph": "s", "id": 112421, "pid": 76337, "tid": -914061504, "ts": 1716454223160132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223160163, "dur": 4,
    "args": {
      "External id": 112429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112429, "pid": 5, "tid": 7, "ts": 1716454223160163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160155, "dur": 7,
    "args": {
      "External id": 112429,
      "cbid": 211, "correlation": 112429
    }
  },
  {
    "ph": "s", "id": 112429, "pid": 76337, "tid": -914061504, "ts": 1716454223160155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223160181, "dur": 4,
    "args": {
      "External id": 112437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112437, "pid": 5, "tid": 7, "ts": 1716454223160181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160173, "dur": 6,
    "args": {
      "External id": 112437,
      "cbid": 211, "correlation": 112437
    }
  },
  {
    "ph": "s", "id": 112437, "pid": 76337, "tid": -914061504, "ts": 1716454223160173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223160201, "dur": 3,
    "args": {
      "External id": 112445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112445,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112445, "pid": 5, "tid": 7, "ts": 1716454223160201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160193, "dur": 7,
    "args": {
      "External id": 112445,
      "cbid": 211, "correlation": 112445
    }
  },
  {
    "ph": "s", "id": 112445, "pid": 76337, "tid": -914061504, "ts": 1716454223160193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160258, "dur": 3,
    "args": {
      "External id": 112453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112453,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112453, "pid": 5, "tid": 7, "ts": 1716454223160258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160247, "dur": 11,
    "args": {
      "External id": 112453,
      "cbid": 211, "correlation": 112453
    }
  },
  {
    "ph": "s", "id": 112453, "pid": 76337, "tid": -914061504, "ts": 1716454223160247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223160284, "dur": 4,
    "args": {
      "External id": 112461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112461,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112461, "pid": 5, "tid": 7, "ts": 1716454223160284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160274, "dur": 9,
    "args": {
      "External id": 112461,
      "cbid": 211, "correlation": 112461
    }
  },
  {
    "ph": "s", "id": 112461, "pid": 76337, "tid": -914061504, "ts": 1716454223160274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223160307, "dur": 4,
    "args": {
      "External id": 112469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112469,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112469, "pid": 5, "tid": 7, "ts": 1716454223160307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160299, "dur": 7,
    "args": {
      "External id": 112469,
      "cbid": 211, "correlation": 112469
    }
  },
  {
    "ph": "s", "id": 112469, "pid": 76337, "tid": -914061504, "ts": 1716454223160299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223160325, "dur": 3,
    "args": {
      "External id": 112477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112477,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 112477, "pid": 5, "tid": 7, "ts": 1716454223160325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160318, "dur": 6,
    "args": {
      "External id": 112477,
      "cbid": 211, "correlation": 112477
    }
  },
  {
    "ph": "s", "id": 112477, "pid": 76337, "tid": -914061504, "ts": 1716454223160318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223160732, "dur": 5,
    "args": {
      "External id": 112486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112486,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112486, "pid": 5, "tid": 7, "ts": 1716454223160732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160715, "dur": 17,
    "args": {
      "External id": 112486,
      "cbid": 211, "correlation": 112486
    }
  },
  {
    "ph": "s", "id": 112486, "pid": 76337, "tid": -914061504, "ts": 1716454223160715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223160768, "dur": 5,
    "args": {
      "External id": 112495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112495,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112495, "pid": 5, "tid": 7, "ts": 1716454223160768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160758, "dur": 9,
    "args": {
      "External id": 112495,
      "cbid": 211, "correlation": 112495
    }
  },
  {
    "ph": "s", "id": 112495, "pid": 76337, "tid": -914061504, "ts": 1716454223160758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223160896, "dur": 3,
    "args": {
      "External id": 112511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112511, "pid": 5, "tid": 7, "ts": 1716454223160896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160882, "dur": 14,
    "args": {
      "External id": 112511,
      "cbid": 211, "correlation": 112511
    }
  },
  {
    "ph": "s", "id": 112511, "pid": 76337, "tid": -914061504, "ts": 1716454223160882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160930, "dur": 3,
    "args": {
      "External id": 112519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112519,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112519, "pid": 5, "tid": 7, "ts": 1716454223160930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160919, "dur": 10,
    "args": {
      "External id": 112519,
      "cbid": 211, "correlation": 112519
    }
  },
  {
    "ph": "s", "id": 112519, "pid": 76337, "tid": -914061504, "ts": 1716454223160919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223160961, "dur": 3,
    "args": {
      "External id": 112527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112527,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112527, "pid": 5, "tid": 7, "ts": 1716454223160961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160952, "dur": 8,
    "args": {
      "External id": 112527,
      "cbid": 211, "correlation": 112527
    }
  },
  {
    "ph": "s", "id": 112527, "pid": 76337, "tid": -914061504, "ts": 1716454223160952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223161003, "dur": 4,
    "args": {
      "External id": 112535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112535,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112535, "pid": 5, "tid": 7, "ts": 1716454223161003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223160993, "dur": 10,
    "args": {
      "External id": 112535,
      "cbid": 211, "correlation": 112535
    }
  },
  {
    "ph": "s", "id": 112535, "pid": 76337, "tid": -914061504, "ts": 1716454223160993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223161059, "dur": 4,
    "args": {
      "External id": 112547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112547,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112547, "pid": 5, "tid": 7, "ts": 1716454223161059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161046, "dur": 13,
    "args": {
      "External id": 112547,
      "cbid": 211, "correlation": 112547
    }
  },
  {
    "ph": "s", "id": 112547, "pid": 76337, "tid": -914061504, "ts": 1716454223161046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223161105, "dur": 4,
    "args": {
      "External id": 112558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112558, "pid": 5, "tid": 7, "ts": 1716454223161105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161093, "dur": 11,
    "args": {
      "External id": 112558,
      "cbid": 211, "correlation": 112558
    }
  },
  {
    "ph": "s", "id": 112558, "pid": 76337, "tid": -914061504, "ts": 1716454223161093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223161136, "dur": 3,
    "args": {
      "External id": 112566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112566, "pid": 5, "tid": 7, "ts": 1716454223161136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161127, "dur": 8,
    "args": {
      "External id": 112566,
      "cbid": 211, "correlation": 112566
    }
  },
  {
    "ph": "s", "id": 112566, "pid": 76337, "tid": -914061504, "ts": 1716454223161127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223161169, "dur": 5,
    "args": {
      "External id": 112574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112574,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112574, "pid": 5, "tid": 7, "ts": 1716454223161169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161158, "dur": 11,
    "args": {
      "External id": 112574,
      "cbid": 211, "correlation": 112574
    }
  },
  {
    "ph": "s", "id": 112574, "pid": 76337, "tid": -914061504, "ts": 1716454223161158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223161199, "dur": 5,
    "args": {
      "External id": 112582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112582,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112582, "pid": 5, "tid": 7, "ts": 1716454223161199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161189, "dur": 9,
    "args": {
      "External id": 112582,
      "cbid": 211, "correlation": 112582
    }
  },
  {
    "ph": "s", "id": 112582, "pid": 76337, "tid": -914061504, "ts": 1716454223161189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223161231, "dur": 4,
    "args": {
      "External id": 112591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112591,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112591, "pid": 5, "tid": 7, "ts": 1716454223161231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161220, "dur": 10,
    "args": {
      "External id": 112591,
      "cbid": 211, "correlation": 112591
    }
  },
  {
    "ph": "s", "id": 112591, "pid": 76337, "tid": -914061504, "ts": 1716454223161220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223161291, "dur": 5,
    "args": {
      "External id": 112604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112604,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112604, "pid": 5, "tid": 7, "ts": 1716454223161291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161279, "dur": 13,
    "args": {
      "External id": 112604,
      "cbid": 211, "correlation": 112604
    }
  },
  {
    "ph": "s", "id": 112604, "pid": 76337, "tid": -914061504, "ts": 1716454223161279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223161332, "dur": 5,
    "args": {
      "External id": 112614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112614,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 112614, "pid": 5, "tid": 7, "ts": 1716454223161332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161321, "dur": 10,
    "args": {
      "External id": 112614,
      "cbid": 211, "correlation": 112614
    }
  },
  {
    "ph": "s", "id": 112614, "pid": 76337, "tid": -914061504, "ts": 1716454223161321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161460, "dur": 4,
    "args": {
      "External id": 112631,
      "cbid": 251, "correlation": 112631
    }
  },
  {
    "ph": "f", "id": 112631, "pid": 76337, "tid": -914061504, "ts": 1716454223161460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454223161489, "dur": 12,
    "args": {
      "External id": 112633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112633,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112633, "pid": 5, "tid": 7, "ts": 1716454223161489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161472, "dur": 17,
    "args": {
      "External id": 112633,
      "cbid": 211, "correlation": 112633
    }
  },
  {
    "ph": "s", "id": 112633, "pid": 76337, "tid": -914061504, "ts": 1716454223161472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223161550, "dur": 4,
    "args": {
      "External id": 112641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112641,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112641, "pid": 5, "tid": 7, "ts": 1716454223161550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161538, "dur": 12,
    "args": {
      "External id": 112641,
      "cbid": 211, "correlation": 112641
    }
  },
  {
    "ph": "s", "id": 112641, "pid": 76337, "tid": -914061504, "ts": 1716454223161538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161609, "dur": 2,
    "args": {
      "External id": 112657,
      "cbid": 251, "correlation": 112657
    }
  },
  {
    "ph": "f", "id": 112657, "pid": 76337, "tid": -914061504, "ts": 1716454223161609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161615, "dur": 0,
    "args": {
      "External id": 112659,
      "cbid": 251, "correlation": 112659
    }
  },
  {
    "ph": "f", "id": 112659, "pid": 76337, "tid": -914061504, "ts": 1716454223161615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223161631, "dur": 13,
    "args": {
      "External id": 112660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112660,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112660, "pid": 5, "tid": 7, "ts": 1716454223161631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161617, "dur": 13,
    "args": {
      "External id": 112660,
      "cbid": 211, "correlation": 112660
    }
  },
  {
    "ph": "s", "id": 112660, "pid": 76337, "tid": -914061504, "ts": 1716454223161617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223161645, "dur": 5,
    "args": {
      "External id": 112662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112662,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112662, "pid": 5, "tid": 7, "ts": 1716454223161645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161635, "dur": 9,
    "args": {
      "External id": 112662,
      "cbid": 211, "correlation": 112662
    }
  },
  {
    "ph": "s", "id": 112662, "pid": 76337, "tid": -914061504, "ts": 1716454223161635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161742, "dur": 1,
    "args": {
      "External id": 112672,
      "cbid": 317, "correlation": 112672
    }
  },
  {
    "ph": "f", "id": 112672, "pid": 76337, "tid": -914061504, "ts": 1716454223161742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161744, "dur": 1,
    "args": {
      "External id": 112673,
      "cbid": 203, "correlation": 112673
    }
  },
  {
    "ph": "f", "id": 112673, "pid": 76337, "tid": -914061504, "ts": 1716454223161744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161746, "dur": 1,
    "args": {
      "External id": 112674,
      "cbid": 205, "correlation": 112674
    }
  },
  {
    "ph": "f", "id": 112674, "pid": 76337, "tid": -914061504, "ts": 1716454223161746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223161802, "dur": 7,
    "args": {
      "External id": 112678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112678, "pid": 5, "tid": 7, "ts": 1716454223161802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161787, "dur": 14,
    "args": {
      "External id": 112678,
      "cbid": 211, "correlation": 112678
    }
  },
  {
    "ph": "s", "id": 112678, "pid": 76337, "tid": -914061504, "ts": 1716454223161787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223161812, "dur": 4,
    "args": {
      "External id": 112680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112680,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 112680, "pid": 5, "tid": 7, "ts": 1716454223161812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161805, "dur": 6,
    "args": {
      "External id": 112680,
      "cbid": 211, "correlation": 112680
    }
  },
  {
    "ph": "s", "id": 112680, "pid": 76337, "tid": -914061504, "ts": 1716454223161805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223161832, "dur": 3,
    "args": {
      "External id": 112682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112682, "pid": 5, "tid": 7, "ts": 1716454223161832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161822, "dur": 8,
    "args": {
      "External id": 112682,
      "cbid": 211, "correlation": 112682
    }
  },
  {
    "ph": "s", "id": 112682, "pid": 76337, "tid": -914061504, "ts": 1716454223161822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161837, "dur": 1,
    "args": {
      "External id": 112683,
      "cbid": 51, "correlation": 112683
    }
  },
  {
    "ph": "s", "id": 112683, "pid": 76337, "tid": -914061504, "ts": 1716454223161837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223161848, "dur": 85,
    "args": {
      "External id": 112684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112684,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112684, "pid": 5, "tid": 7, "ts": 1716454223161848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161839, "dur": 8,
    "args": {
      "External id": 112684,
      "cbid": 211, "correlation": 112684
    }
  },
  {
    "ph": "s", "id": 112684, "pid": 76337, "tid": -914061504, "ts": 1716454223161839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223161934, "dur": 60,
    "args": {
      "External id": 112689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112689, "pid": 5, "tid": 7, "ts": 1716454223161934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223161875, "dur": 11,
    "args": {
      "External id": 112689,
      "cbid": 211, "correlation": 112689
    }
  },
  {
    "ph": "s", "id": 112689, "pid": 76337, "tid": -914061504, "ts": 1716454223161875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223163670, "dur": 52,
    "args": {
      "External id": 112709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112709,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 112709, "pid": 5, "tid": 7, "ts": 1716454223163670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163654, "dur": 15,
    "args": {
      "External id": 112709,
      "cbid": 211, "correlation": 112709
    }
  },
  {
    "ph": "s", "id": 112709, "pid": 76337, "tid": -914061504, "ts": 1716454223163654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223163723, "dur": 4,
    "args": {
      "External id": 112721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112721,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112721, "pid": 5, "tid": 7, "ts": 1716454223163723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163682, "dur": 8,
    "args": {
      "External id": 112721,
      "cbid": 211, "correlation": 112721
    }
  },
  {
    "ph": "s", "id": 112721, "pid": 76337, "tid": -914061504, "ts": 1716454223163682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223163729, "dur": 57,
    "args": {
      "External id": 112724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112724,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112724, "pid": 5, "tid": 7, "ts": 1716454223163729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163706, "dur": 8,
    "args": {
      "External id": 112724,
      "cbid": 211, "correlation": 112724
    }
  },
  {
    "ph": "s", "id": 112724, "pid": 76337, "tid": -914061504, "ts": 1716454223163706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223163787, "dur": 37,
    "args": {
      "External id": 112733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112733,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112733, "pid": 5, "tid": 7, "ts": 1716454223163787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163751, "dur": 11,
    "args": {
      "External id": 112733,
      "cbid": 211, "correlation": 112733
    }
  },
  {
    "ph": "s", "id": 112733, "pid": 76337, "tid": -914061504, "ts": 1716454223163751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163808, "dur": 0,
    "args": {
      "External id": 112743,
      "cbid": 317, "correlation": 112743
    }
  },
  {
    "ph": "f", "id": 112743, "pid": 76337, "tid": -914061504, "ts": 1716454223163808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163809, "dur": 0,
    "args": {
      "External id": 112744,
      "cbid": 203, "correlation": 112744
    }
  },
  {
    "ph": "f", "id": 112744, "pid": 76337, "tid": -914061504, "ts": 1716454223163809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163809, "dur": 0,
    "args": {
      "External id": 112745,
      "cbid": 205, "correlation": 112745
    }
  },
  {
    "ph": "f", "id": 112745, "pid": 76337, "tid": -914061504, "ts": 1716454223163809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223163842, "dur": 40,
    "args": {
      "External id": 112749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112749,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112749, "pid": 5, "tid": 7, "ts": 1716454223163842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163828, "dur": 13,
    "args": {
      "External id": 112749,
      "cbid": 211, "correlation": 112749
    }
  },
  {
    "ph": "s", "id": 112749, "pid": 76337, "tid": -914061504, "ts": 1716454223163828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223163883, "dur": 14,
    "args": {
      "External id": 112751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112751,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112751, "pid": 5, "tid": 7, "ts": 1716454223163883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163844, "dur": 6,
    "args": {
      "External id": 112751,
      "cbid": 211, "correlation": 112751
    }
  },
  {
    "ph": "s", "id": 112751, "pid": 76337, "tid": -914061504, "ts": 1716454223163844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223163899, "dur": 3,
    "args": {
      "External id": 112753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112753,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112753, "pid": 5, "tid": 7, "ts": 1716454223163899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163854, "dur": 6,
    "args": {
      "External id": 112753,
      "cbid": 211, "correlation": 112753
    }
  },
  {
    "ph": "s", "id": 112753, "pid": 76337, "tid": -914061504, "ts": 1716454223163854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163864, "dur": 0,
    "args": {
      "External id": 112754,
      "cbid": 51, "correlation": 112754
    }
  },
  {
    "ph": "s", "id": 112754, "pid": 76337, "tid": -914061504, "ts": 1716454223163864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223163903, "dur": 699,
    "args": {
      "External id": 112755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112755,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112755, "pid": 5, "tid": 7, "ts": 1716454223163903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163865, "dur": 7,
    "args": {
      "External id": 112755,
      "cbid": 211, "correlation": 112755
    }
  },
  {
    "ph": "s", "id": 112755, "pid": 76337, "tid": -914061504, "ts": 1716454223163865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223164603, "dur": 59,
    "args": {
      "External id": 112760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112760,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112760, "pid": 5, "tid": 7, "ts": 1716454223164603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163896, "dur": 9,
    "args": {
      "External id": 112760,
      "cbid": 211, "correlation": 112760
    }
  },
  {
    "ph": "s", "id": 112760, "pid": 76337, "tid": -914061504, "ts": 1716454223163896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223164663, "dur": 5,
    "args": {
      "External id": 112768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112768,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112768, "pid": 5, "tid": 7, "ts": 1716454223164663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223163939, "dur": 9,
    "args": {
      "External id": 112768,
      "cbid": 211, "correlation": 112768
    }
  },
  {
    "ph": "s", "id": 112768, "pid": 76337, "tid": -914061504, "ts": 1716454223163939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164011, "dur": 2,
    "args": {
      "External id": 112784,
      "cbid": 251, "correlation": 112784
    }
  },
  {
    "ph": "f", "id": 112784, "pid": 76337, "tid": -914061504, "ts": 1716454223164011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164017, "dur": 0,
    "args": {
      "External id": 112786,
      "cbid": 251, "correlation": 112786
    }
  },
  {
    "ph": "f", "id": 112786, "pid": 76337, "tid": -914061504, "ts": 1716454223164017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223164669, "dur": 9,
    "args": {
      "External id": 112787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112787,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 112787, "pid": 5, "tid": 7, "ts": 1716454223164669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164019, "dur": 13,
    "args": {
      "External id": 112787,
      "cbid": 211, "correlation": 112787
    }
  },
  {
    "ph": "s", "id": 112787, "pid": 76337, "tid": -914061504, "ts": 1716454223164019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223164680, "dur": 4,
    "args": {
      "External id": 112789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112789,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 112789, "pid": 5, "tid": 7, "ts": 1716454223164680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164034, "dur": 6,
    "args": {
      "External id": 112789,
      "cbid": 211, "correlation": 112789
    }
  },
  {
    "ph": "s", "id": 112789, "pid": 76337, "tid": -914061504, "ts": 1716454223164034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223164685, "dur": 55,
    "args": {
      "External id": 112799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112799, "pid": 5, "tid": 7, "ts": 1716454223164685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164094, "dur": 12,
    "args": {
      "External id": 112799,
      "cbid": 211, "correlation": 112799
    }
  },
  {
    "ph": "s", "id": 112799, "pid": 76337, "tid": -914061504, "ts": 1716454223164094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223164741, "dur": 51,
    "args": {
      "External id": 112819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112819,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 112819, "pid": 5, "tid": 7, "ts": 1716454223164741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164160, "dur": 11,
    "args": {
      "External id": 112819,
      "cbid": 211, "correlation": 112819
    }
  },
  {
    "ph": "s", "id": 112819, "pid": 76337, "tid": -914061504, "ts": 1716454223164160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223164794, "dur": 4,
    "args": {
      "External id": 112831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112831,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112831, "pid": 5, "tid": 7, "ts": 1716454223164794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164181, "dur": 7,
    "args": {
      "External id": 112831,
      "cbid": 211, "correlation": 112831
    }
  },
  {
    "ph": "s", "id": 112831, "pid": 76337, "tid": -914061504, "ts": 1716454223164181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223164799, "dur": 55,
    "args": {
      "External id": 112834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112834,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112834, "pid": 5, "tid": 7, "ts": 1716454223164799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164201, "dur": 6,
    "args": {
      "External id": 112834,
      "cbid": 211, "correlation": 112834
    }
  },
  {
    "ph": "s", "id": 112834, "pid": 76337, "tid": -914061504, "ts": 1716454223164201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223164855, "dur": 36,
    "args": {
      "External id": 112843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112843,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112843, "pid": 5, "tid": 7, "ts": 1716454223164855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164241, "dur": 10,
    "args": {
      "External id": 112843,
      "cbid": 211, "correlation": 112843
    }
  },
  {
    "ph": "s", "id": 112843, "pid": 76337, "tid": -914061504, "ts": 1716454223164241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164312, "dur": 0,
    "args": {
      "External id": 112853,
      "cbid": 317, "correlation": 112853
    }
  },
  {
    "ph": "f", "id": 112853, "pid": 76337, "tid": -914061504, "ts": 1716454223164312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164312, "dur": 0,
    "args": {
      "External id": 112854,
      "cbid": 203, "correlation": 112854
    }
  },
  {
    "ph": "f", "id": 112854, "pid": 76337, "tid": -914061504, "ts": 1716454223164312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164313, "dur": 0,
    "args": {
      "External id": 112855,
      "cbid": 205, "correlation": 112855
    }
  },
  {
    "ph": "f", "id": 112855, "pid": 76337, "tid": -914061504, "ts": 1716454223164313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223164893, "dur": 40,
    "args": {
      "External id": 112859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112859,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112859, "pid": 5, "tid": 7, "ts": 1716454223164893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164327, "dur": 13,
    "args": {
      "External id": 112859,
      "cbid": 211, "correlation": 112859
    }
  },
  {
    "ph": "s", "id": 112859, "pid": 76337, "tid": -914061504, "ts": 1716454223164327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223164935, "dur": 14,
    "args": {
      "External id": 112861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112861,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112861, "pid": 5, "tid": 7, "ts": 1716454223164935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164343, "dur": 5,
    "args": {
      "External id": 112861,
      "cbid": 211, "correlation": 112861
    }
  },
  {
    "ph": "s", "id": 112861, "pid": 76337, "tid": -914061504, "ts": 1716454223164343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223164950, "dur": 3,
    "args": {
      "External id": 112863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112863, "pid": 5, "tid": 7, "ts": 1716454223164950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164352, "dur": 5,
    "args": {
      "External id": 112863,
      "cbid": 211, "correlation": 112863
    }
  },
  {
    "ph": "s", "id": 112863, "pid": 76337, "tid": -914061504, "ts": 1716454223164352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164360, "dur": 0,
    "args": {
      "External id": 112864,
      "cbid": 51, "correlation": 112864
    }
  },
  {
    "ph": "s", "id": 112864, "pid": 76337, "tid": -914061504, "ts": 1716454223164360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223164955, "dur": 693,
    "args": {
      "External id": 112865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112865,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112865, "pid": 5, "tid": 7, "ts": 1716454223164955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164361, "dur": 5,
    "args": {
      "External id": 112865,
      "cbid": 211, "correlation": 112865
    }
  },
  {
    "ph": "s", "id": 112865, "pid": 76337, "tid": -914061504, "ts": 1716454223164361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223165649, "dur": 58,
    "args": {
      "External id": 112870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112870, "pid": 5, "tid": 7, "ts": 1716454223165649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164389, "dur": 8,
    "args": {
      "External id": 112870,
      "cbid": 211, "correlation": 112870
    }
  },
  {
    "ph": "s", "id": 112870, "pid": 76337, "tid": -914061504, "ts": 1716454223164389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223165709, "dur": 51,
    "args": {
      "External id": 112878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112878,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112878, "pid": 5, "tid": 7, "ts": 1716454223165709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164421, "dur": 9,
    "args": {
      "External id": 112878,
      "cbid": 211, "correlation": 112878
    }
  },
  {
    "ph": "s", "id": 112878, "pid": 76337, "tid": -914061504, "ts": 1716454223164421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223165761, "dur": 35,
    "args": {
      "External id": 112886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112886,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112886, "pid": 5, "tid": 7, "ts": 1716454223165761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164452, "dur": 9,
    "args": {
      "External id": 112886,
      "cbid": 211, "correlation": 112886
    }
  },
  {
    "ph": "s", "id": 112886, "pid": 76337, "tid": -914061504, "ts": 1716454223164452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223165798, "dur": 52,
    "args": {
      "External id": 112906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112906,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 112906, "pid": 5, "tid": 7, "ts": 1716454223165798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164535, "dur": 12,
    "args": {
      "External id": 112906,
      "cbid": 211, "correlation": 112906
    }
  },
  {
    "ph": "s", "id": 112906, "pid": 76337, "tid": -914061504, "ts": 1716454223164535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223165851, "dur": 4,
    "args": {
      "External id": 112918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112918,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 112918, "pid": 5, "tid": 7, "ts": 1716454223165851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164557, "dur": 6,
    "args": {
      "External id": 112918,
      "cbid": 211, "correlation": 112918
    }
  },
  {
    "ph": "s", "id": 112918, "pid": 76337, "tid": -914061504, "ts": 1716454223164557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223165857, "dur": 55,
    "args": {
      "External id": 112921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112921, "pid": 5, "tid": 7, "ts": 1716454223165857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164574, "dur": 6,
    "args": {
      "External id": 112921,
      "cbid": 211, "correlation": 112921
    }
  },
  {
    "ph": "s", "id": 112921, "pid": 76337, "tid": -914061504, "ts": 1716454223164574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164632, "dur": 0,
    "args": {
      "External id": 112932,
      "cbid": 317, "correlation": 112932
    }
  },
  {
    "ph": "f", "id": 112932, "pid": 76337, "tid": -914061504, "ts": 1716454223164632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164633, "dur": 0,
    "args": {
      "External id": 112933,
      "cbid": 203, "correlation": 112933
    }
  },
  {
    "ph": "f", "id": 112933, "pid": 76337, "tid": -914061504, "ts": 1716454223164633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164634, "dur": 0,
    "args": {
      "External id": 112934,
      "cbid": 205, "correlation": 112934
    }
  },
  {
    "ph": "f", "id": 112934, "pid": 76337, "tid": -914061504, "ts": 1716454223164634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164667, "dur": 2,
    "args": {
      "External id": 112938,
      "cbid": 251, "correlation": 112938
    }
  },
  {
    "ph": "f", "id": 112938, "pid": 76337, "tid": -914061504, "ts": 1716454223164667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164671, "dur": 1,
    "args": {
      "External id": 112939,
      "cbid": 251, "correlation": 112939
    }
  },
  {
    "ph": "f", "id": 112939, "pid": 76337, "tid": -914061504, "ts": 1716454223164671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164673, "dur": 1,
    "args": {
      "External id": 112940,
      "cbid": 251, "correlation": 112940
    }
  },
  {
    "ph": "f", "id": 112940, "pid": 76337, "tid": -914061504, "ts": 1716454223164673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164676, "dur": 1,
    "args": {
      "External id": 112941,
      "cbid": 251, "correlation": 112941
    }
  },
  {
    "ph": "f", "id": 112941, "pid": 76337, "tid": -914061504, "ts": 1716454223164676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164678, "dur": 1,
    "args": {
      "External id": 112942,
      "cbid": 251, "correlation": 112942
    }
  },
  {
    "ph": "f", "id": 112942, "pid": 76337, "tid": -914061504, "ts": 1716454223164678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164680, "dur": 1,
    "args": {
      "External id": 112943,
      "cbid": 251, "correlation": 112943
    }
  },
  {
    "ph": "f", "id": 112943, "pid": 76337, "tid": -914061504, "ts": 1716454223164680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164682, "dur": 1,
    "args": {
      "External id": 112944,
      "cbid": 251, "correlation": 112944
    }
  },
  {
    "ph": "f", "id": 112944, "pid": 76337, "tid": -914061504, "ts": 1716454223164682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164684, "dur": 1,
    "args": {
      "External id": 112945,
      "cbid": 251, "correlation": 112945
    }
  },
  {
    "ph": "f", "id": 112945, "pid": 76337, "tid": -914061504, "ts": 1716454223164684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164687, "dur": 0,
    "args": {
      "External id": 112946,
      "cbid": 251, "correlation": 112946
    }
  },
  {
    "ph": "f", "id": 112946, "pid": 76337, "tid": -914061504, "ts": 1716454223164687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223165913, "dur": 116,
    "args": {
      "External id": 112947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112947,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 112947, "pid": 5, "tid": 7, "ts": 1716454223165913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164691, "dur": 15,
    "args": {
      "External id": 112947,
      "cbid": 211, "correlation": 112947
    }
  },
  {
    "ph": "s", "id": 112947, "pid": 76337, "tid": -914061504, "ts": 1716454223164691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223166030, "dur": 60,
    "args": {
      "External id": 112953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112953,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112953, "pid": 5, "tid": 7, "ts": 1716454223166030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164729, "dur": 9,
    "args": {
      "External id": 112953,
      "cbid": 211, "correlation": 112953
    }
  },
  {
    "ph": "s", "id": 112953, "pid": 76337, "tid": -914061504, "ts": 1716454223164729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223166092, "dur": 566,
    "args": {
      "External id": 112962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112962,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112962, "pid": 5, "tid": 7, "ts": 1716454223166092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164822, "dur": 16,
    "args": {
      "External id": 112962,
      "cbid": 211, "correlation": 112962
    }
  },
  {
    "ph": "s", "id": 112962, "pid": 76337, "tid": -914061504, "ts": 1716454223164822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223166659, "dur": 181,
    "args": {
      "External id": 112984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112984,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 112984, "pid": 5, "tid": 7, "ts": 1716454223166659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223164897, "dur": 12,
    "args": {
      "External id": 112984,
      "cbid": 211, "correlation": 112984
    }
  },
  {
    "ph": "s", "id": 112984, "pid": 76337, "tid": -914061504, "ts": 1716454223164897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165124, "dur": 2,
    "args": {
      "External id": 112995,
      "cbid": 251, "correlation": 112995
    }
  },
  {
    "ph": "f", "id": 112995, "pid": 76337, "tid": -914061504, "ts": 1716454223165124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223166841, "dur": 194,
    "args": {
      "External id": 112996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 112996,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 112996, "pid": 5, "tid": 7, "ts": 1716454223166841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165132, "dur": 15,
    "args": {
      "External id": 112996,
      "cbid": 211, "correlation": 112996
    }
  },
  {
    "ph": "s", "id": 112996, "pid": 76337, "tid": -914061504, "ts": 1716454223165132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165209, "dur": 1,
    "args": {
      "External id": 113007,
      "cbid": 251, "correlation": 113007
    }
  },
  {
    "ph": "f", "id": 113007, "pid": 76337, "tid": -914061504, "ts": 1716454223165209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223167037, "dur": 189,
    "args": {
      "External id": 113008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113008,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113008, "pid": 5, "tid": 7, "ts": 1716454223167037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165213, "dur": 12,
    "args": {
      "External id": 113008,
      "cbid": 211, "correlation": 113008
    }
  },
  {
    "ph": "s", "id": 113008, "pid": 76337, "tid": -914061504, "ts": 1716454223165213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165278, "dur": 1,
    "args": {
      "External id": 113019,
      "cbid": 251, "correlation": 113019
    }
  },
  {
    "ph": "f", "id": 113019, "pid": 76337, "tid": -914061504, "ts": 1716454223165278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223167227, "dur": 188,
    "args": {
      "External id": 113020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113020,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113020, "pid": 5, "tid": 7, "ts": 1716454223167227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165282, "dur": 11,
    "args": {
      "External id": 113020,
      "cbid": 211, "correlation": 113020
    }
  },
  {
    "ph": "s", "id": 113020, "pid": 76337, "tid": -914061504, "ts": 1716454223165282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223167417, "dur": 18597,
    "args": {
      "External id": 113041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113041,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113041, "pid": 5, "tid": 7, "ts": 1716454223167417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165391, "dur": 15,
    "args": {
      "External id": 113041,
      "cbid": 211, "correlation": 113041
    }
  },
  {
    "ph": "s", "id": 113041, "pid": 76337, "tid": -914061504, "ts": 1716454223165391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165506, "dur": 2,
    "args": {
      "External id": 113059,
      "cbid": 251, "correlation": 113059
    }
  },
  {
    "ph": "f", "id": 113059, "pid": 76337, "tid": -914061504, "ts": 1716454223165506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223186015, "dur": 202,
    "args": {
      "External id": 113061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113061,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113061, "pid": 5, "tid": 7, "ts": 1716454223186015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165513, "dur": 13,
    "args": {
      "External id": 113061,
      "cbid": 211, "correlation": 113061
    }
  },
  {
    "ph": "s", "id": 113061, "pid": 76337, "tid": -914061504, "ts": 1716454223165513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223186218, "dur": 66,
    "args": {
      "External id": 113069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113069,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113069, "pid": 5, "tid": 7, "ts": 1716454223186218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165586, "dur": 12,
    "args": {
      "External id": 113069,
      "cbid": 211, "correlation": 113069
    }
  },
  {
    "ph": "s", "id": 113069, "pid": 76337, "tid": -914061504, "ts": 1716454223165586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223186286, "dur": 97,
    "args": {
      "External id": 113077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113077,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113077, "pid": 5, "tid": 7, "ts": 1716454223186286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165627, "dur": 9,
    "args": {
      "External id": 113077,
      "cbid": 211, "correlation": 113077
    }
  },
  {
    "ph": "s", "id": 113077, "pid": 76337, "tid": -914061504, "ts": 1716454223165627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223186384, "dur": 54,
    "args": {
      "External id": 113088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113088,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113088, "pid": 5, "tid": 7, "ts": 1716454223186384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165714, "dur": 14,
    "args": {
      "External id": 113088,
      "cbid": 211, "correlation": 113088
    }
  },
  {
    "ph": "s", "id": 113088, "pid": 76337, "tid": -914061504, "ts": 1716454223165714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223186440, "dur": 92,
    "args": {
      "External id": 113110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113110,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113110, "pid": 5, "tid": 7, "ts": 1716454223186440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165747, "dur": 8,
    "args": {
      "External id": 113110,
      "cbid": 211, "correlation": 113110
    }
  },
  {
    "ph": "s", "id": 113110, "pid": 76337, "tid": -914061504, "ts": 1716454223165747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165833, "dur": 1,
    "args": {
      "External id": 113121,
      "cbid": 251, "correlation": 113121
    }
  },
  {
    "ph": "f", "id": 113121, "pid": 76337, "tid": -914061504, "ts": 1716454223165833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223186534, "dur": 103,
    "args": {
      "External id": 113122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113122,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113122, "pid": 5, "tid": 7, "ts": 1716454223186534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165838, "dur": 13,
    "args": {
      "External id": 113122,
      "cbid": 211, "correlation": 113122
    }
  },
  {
    "ph": "s", "id": 113122, "pid": 76337, "tid": -914061504, "ts": 1716454223165838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165922, "dur": 1,
    "args": {
      "External id": 113133,
      "cbid": 251, "correlation": 113133
    }
  },
  {
    "ph": "f", "id": 113133, "pid": 76337, "tid": -914061504, "ts": 1716454223165922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165926, "dur": 0,
    "args": {
      "External id": 113134,
      "cbid": 251, "correlation": 113134
    }
  },
  {
    "ph": "f", "id": 113134, "pid": 76337, "tid": -914061504, "ts": 1716454223165926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223186638, "dur": 10,
    "args": {
      "External id": 113135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113135,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 113135, "pid": 5, "tid": 7, "ts": 1716454223186638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165928, "dur": 14,
    "args": {
      "External id": 113135,
      "cbid": 211, "correlation": 113135
    }
  },
  {
    "ph": "s", "id": 113135, "pid": 76337, "tid": -914061504, "ts": 1716454223165928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223186650, "dur": 5,
    "args": {
      "External id": 113137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113137,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 113137, "pid": 5, "tid": 7, "ts": 1716454223186650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223165946, "dur": 7,
    "args": {
      "External id": 113137,
      "cbid": 211, "correlation": 113137
    }
  },
  {
    "ph": "s", "id": 113137, "pid": 76337, "tid": -914061504, "ts": 1716454223165946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166019, "dur": 1,
    "args": {
      "External id": 113148,
      "cbid": 251, "correlation": 113148
    }
  },
  {
    "ph": "f", "id": 113148, "pid": 76337, "tid": -914061504, "ts": 1716454223166019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166023, "dur": 0,
    "args": {
      "External id": 113149,
      "cbid": 251, "correlation": 113149
    }
  },
  {
    "ph": "f", "id": 113149, "pid": 76337, "tid": -914061504, "ts": 1716454223166023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223186656, "dur": 6,
    "args": {
      "External id": 113150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113150,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 113150, "pid": 5, "tid": 7, "ts": 1716454223186656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166024, "dur": 12,
    "args": {
      "External id": 113150,
      "cbid": 211, "correlation": 113150
    }
  },
  {
    "ph": "s", "id": 113150, "pid": 76337, "tid": -914061504, "ts": 1716454223166024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223186663, "dur": 3,
    "args": {
      "External id": 113152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113152,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 113152, "pid": 5, "tid": 7, "ts": 1716454223186663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166038, "dur": 5,
    "args": {
      "External id": 113152,
      "cbid": 211, "correlation": 113152
    }
  },
  {
    "ph": "s", "id": 113152, "pid": 76337, "tid": -914061504, "ts": 1716454223166038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223186668, "dur": 154,
    "args": {
      "External id": 113173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113173,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113173, "pid": 5, "tid": 7, "ts": 1716454223186668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166113, "dur": 12,
    "args": {
      "External id": 113173,
      "cbid": 211, "correlation": 113173
    }
  },
  {
    "ph": "s", "id": 113173, "pid": 76337, "tid": -914061504, "ts": 1716454223166113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166212, "dur": 2,
    "args": {
      "External id": 113191,
      "cbid": 251, "correlation": 113191
    }
  },
  {
    "ph": "f", "id": 113191, "pid": 76337, "tid": -914061504, "ts": 1716454223166212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223186824, "dur": 107,
    "args": {
      "External id": 113193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113193,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113193, "pid": 5, "tid": 7, "ts": 1716454223186824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166218, "dur": 14,
    "args": {
      "External id": 113193,
      "cbid": 211, "correlation": 113193
    }
  },
  {
    "ph": "s", "id": 113193, "pid": 76337, "tid": -914061504, "ts": 1716454223166218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223186932, "dur": 35,
    "args": {
      "External id": 113201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113201,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113201, "pid": 5, "tid": 7, "ts": 1716454223186932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166289, "dur": 12,
    "args": {
      "External id": 113201,
      "cbid": 211, "correlation": 113201
    }
  },
  {
    "ph": "s", "id": 113201, "pid": 76337, "tid": -914061504, "ts": 1716454223166289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223186969, "dur": 68,
    "args": {
      "External id": 113209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113209, "pid": 5, "tid": 7, "ts": 1716454223186969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166330, "dur": 9,
    "args": {
      "External id": 113209,
      "cbid": 211, "correlation": 113209
    }
  },
  {
    "ph": "s", "id": 113209, "pid": 76337, "tid": -914061504, "ts": 1716454223166330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223187039, "dur": 92,
    "args": {
      "External id": 113231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113231,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113231, "pid": 5, "tid": 7, "ts": 1716454223187039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166382, "dur": 10,
    "args": {
      "External id": 113231,
      "cbid": 211, "correlation": 113231
    }
  },
  {
    "ph": "s", "id": 113231, "pid": 76337, "tid": -914061504, "ts": 1716454223166382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166472, "dur": 1,
    "args": {
      "External id": 113247,
      "cbid": 251, "correlation": 113247
    }
  },
  {
    "ph": "f", "id": 113247, "pid": 76337, "tid": -914061504, "ts": 1716454223166472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223187132, "dur": 574,
    "args": {
      "External id": 113249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113249,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113249, "pid": 5, "tid": 7, "ts": 1716454223187132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166477, "dur": 13,
    "args": {
      "External id": 113249,
      "cbid": 211, "correlation": 113249
    }
  },
  {
    "ph": "s", "id": 113249, "pid": 76337, "tid": -914061504, "ts": 1716454223166477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223187708, "dur": 244,
    "args": {
      "External id": 113257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113257, "pid": 5, "tid": 7, "ts": 1716454223187708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166559, "dur": 15,
    "args": {
      "External id": 113257,
      "cbid": 211, "correlation": 113257
    }
  },
  {
    "ph": "s", "id": 113257, "pid": 76337, "tid": -914061504, "ts": 1716454223166559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223187953, "dur": 251,
    "args": {
      "External id": 113265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113265, "pid": 5, "tid": 7, "ts": 1716454223187953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166595, "dur": 9,
    "args": {
      "External id": 113265,
      "cbid": 211, "correlation": 113265
    }
  },
  {
    "ph": "s", "id": 113265, "pid": 76337, "tid": -914061504, "ts": 1716454223166595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166681, "dur": 2,
    "args": {
      "External id": 113281,
      "cbid": 251, "correlation": 113281
    }
  },
  {
    "ph": "f", "id": 113281, "pid": 76337, "tid": -914061504, "ts": 1716454223166681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166686, "dur": 0,
    "args": {
      "External id": 113283,
      "cbid": 251, "correlation": 113283
    }
  },
  {
    "ph": "f", "id": 113283, "pid": 76337, "tid": -914061504, "ts": 1716454223166686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223188206, "dur": 362,
    "args": {
      "External id": 113284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113284,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 113284, "pid": 5, "tid": 7, "ts": 1716454223188206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166691, "dur": 14,
    "args": {
      "External id": 113284,
      "cbid": 211, "correlation": 113284
    }
  },
  {
    "ph": "s", "id": 113284, "pid": 76337, "tid": -914061504, "ts": 1716454223166691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223188569, "dur": 50,
    "args": {
      "External id": 113292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113292,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113292, "pid": 5, "tid": 7, "ts": 1716454223188569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166735, "dur": 10,
    "args": {
      "External id": 113292,
      "cbid": 211, "correlation": 113292
    }
  },
  {
    "ph": "s", "id": 113292, "pid": 76337, "tid": -914061504, "ts": 1716454223166735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223188620, "dur": 158,
    "args": {
      "External id": 113303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113303,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113303, "pid": 5, "tid": 7, "ts": 1716454223188620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166806, "dur": 13,
    "args": {
      "External id": 113303,
      "cbid": 211, "correlation": 113303
    }
  },
  {
    "ph": "s", "id": 113303, "pid": 76337, "tid": -914061504, "ts": 1716454223166806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166872, "dur": 0,
    "args": {
      "External id": 113315,
      "cbid": 317, "correlation": 113315
    }
  },
  {
    "ph": "f", "id": 113315, "pid": 76337, "tid": -914061504, "ts": 1716454223166872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166873, "dur": 0,
    "args": {
      "External id": 113316,
      "cbid": 203, "correlation": 113316
    }
  },
  {
    "ph": "f", "id": 113316, "pid": 76337, "tid": -914061504, "ts": 1716454223166873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166874, "dur": 0,
    "args": {
      "External id": 113317,
      "cbid": 205, "correlation": 113317
    }
  },
  {
    "ph": "f", "id": 113317, "pid": 76337, "tid": -914061504, "ts": 1716454223166874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166899, "dur": 1,
    "args": {
      "External id": 113321,
      "cbid": 251, "correlation": 113321
    }
  },
  {
    "ph": "f", "id": 113321, "pid": 76337, "tid": -914061504, "ts": 1716454223166899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166901, "dur": 0,
    "args": {
      "External id": 113322,
      "cbid": 251, "correlation": 113322
    }
  },
  {
    "ph": "f", "id": 113322, "pid": 76337, "tid": -914061504, "ts": 1716454223166901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166902, "dur": 0,
    "args": {
      "External id": 113323,
      "cbid": 251, "correlation": 113323
    }
  },
  {
    "ph": "f", "id": 113323, "pid": 76337, "tid": -914061504, "ts": 1716454223166902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166903, "dur": 0,
    "args": {
      "External id": 113324,
      "cbid": 251, "correlation": 113324
    }
  },
  {
    "ph": "f", "id": 113324, "pid": 76337, "tid": -914061504, "ts": 1716454223166903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166903, "dur": 1,
    "args": {
      "External id": 113325,
      "cbid": 251, "correlation": 113325
    }
  },
  {
    "ph": "f", "id": 113325, "pid": 76337, "tid": -914061504, "ts": 1716454223166903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166905, "dur": 0,
    "args": {
      "External id": 113326,
      "cbid": 251, "correlation": 113326
    }
  },
  {
    "ph": "f", "id": 113326, "pid": 76337, "tid": -914061504, "ts": 1716454223166905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166906, "dur": 0,
    "args": {
      "External id": 113327,
      "cbid": 251, "correlation": 113327
    }
  },
  {
    "ph": "f", "id": 113327, "pid": 76337, "tid": -914061504, "ts": 1716454223166906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166907, "dur": 1,
    "args": {
      "External id": 113328,
      "cbid": 251, "correlation": 113328
    }
  },
  {
    "ph": "f", "id": 113328, "pid": 76337, "tid": -914061504, "ts": 1716454223166907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166909, "dur": 0,
    "args": {
      "External id": 113329,
      "cbid": 251, "correlation": 113329
    }
  },
  {
    "ph": "f", "id": 113329, "pid": 76337, "tid": -914061504, "ts": 1716454223166909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223188780, "dur": 117,
    "args": {
      "External id": 113330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113330,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113330, "pid": 5, "tid": 7, "ts": 1716454223188780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166911, "dur": 12,
    "args": {
      "External id": 113330,
      "cbid": 211, "correlation": 113330
    }
  },
  {
    "ph": "s", "id": 113330, "pid": 76337, "tid": -914061504, "ts": 1716454223166911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223188898, "dur": 60,
    "args": {
      "External id": 113336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113336,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113336, "pid": 5, "tid": 7, "ts": 1716454223188898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166946, "dur": 9,
    "args": {
      "External id": 113336,
      "cbid": 211, "correlation": 113336
    }
  },
  {
    "ph": "s", "id": 113336, "pid": 76337, "tid": -914061504, "ts": 1716454223166946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223188959, "dur": 51,
    "args": {
      "External id": 113344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113344,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113344, "pid": 5, "tid": 7, "ts": 1716454223188959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223166987, "dur": 9,
    "args": {
      "External id": 113344,
      "cbid": 211, "correlation": 113344
    }
  },
  {
    "ph": "s", "id": 113344, "pid": 76337, "tid": -914061504, "ts": 1716454223166987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223189012, "dur": 51,
    "args": {
      "External id": 113364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113364,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 113364, "pid": 5, "tid": 7, "ts": 1716454223189012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167062, "dur": 12,
    "args": {
      "External id": 113364,
      "cbid": 211, "correlation": 113364
    }
  },
  {
    "ph": "s", "id": 113364, "pid": 76337, "tid": -914061504, "ts": 1716454223167062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223189064, "dur": 4,
    "args": {
      "External id": 113376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113376,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 113376, "pid": 5, "tid": 7, "ts": 1716454223189064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167083, "dur": 6,
    "args": {
      "External id": 113376,
      "cbid": 211, "correlation": 113376
    }
  },
  {
    "ph": "s", "id": 113376, "pid": 76337, "tid": -914061504, "ts": 1716454223167083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223189070, "dur": 56,
    "args": {
      "External id": 113379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113379, "pid": 5, "tid": 7, "ts": 1716454223189070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167101, "dur": 7,
    "args": {
      "External id": 113379,
      "cbid": 211, "correlation": 113379
    }
  },
  {
    "ph": "s", "id": 113379, "pid": 76337, "tid": -914061504, "ts": 1716454223167101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223189127, "dur": 36,
    "args": {
      "External id": 113388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113388,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113388, "pid": 5, "tid": 7, "ts": 1716454223189127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167141, "dur": 10,
    "args": {
      "External id": 113388,
      "cbid": 211, "correlation": 113388
    }
  },
  {
    "ph": "s", "id": 113388, "pid": 76337, "tid": -914061504, "ts": 1716454223167141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167192, "dur": 0,
    "args": {
      "External id": 113398,
      "cbid": 317, "correlation": 113398
    }
  },
  {
    "ph": "f", "id": 113398, "pid": 76337, "tid": -914061504, "ts": 1716454223167192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167193, "dur": 0,
    "args": {
      "External id": 113399,
      "cbid": 203, "correlation": 113399
    }
  },
  {
    "ph": "f", "id": 113399, "pid": 76337, "tid": -914061504, "ts": 1716454223167193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167194, "dur": 0,
    "args": {
      "External id": 113400,
      "cbid": 205, "correlation": 113400
    }
  },
  {
    "ph": "f", "id": 113400, "pid": 76337, "tid": -914061504, "ts": 1716454223167194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223189165, "dur": 41,
    "args": {
      "External id": 113404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113404, "pid": 5, "tid": 7, "ts": 1716454223189165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167209, "dur": 14,
    "args": {
      "External id": 113404,
      "cbid": 211, "correlation": 113404
    }
  },
  {
    "ph": "s", "id": 113404, "pid": 76337, "tid": -914061504, "ts": 1716454223167209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223189207, "dur": 15,
    "args": {
      "External id": 113406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113406,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113406, "pid": 5, "tid": 7, "ts": 1716454223189207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167225, "dur": 6,
    "args": {
      "External id": 113406,
      "cbid": 211, "correlation": 113406
    }
  },
  {
    "ph": "s", "id": 113406, "pid": 76337, "tid": -914061504, "ts": 1716454223167225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223189222, "dur": 3,
    "args": {
      "External id": 113408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 113408, "pid": 5, "tid": 7, "ts": 1716454223189222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167235, "dur": 6,
    "args": {
      "External id": 113408,
      "cbid": 211, "correlation": 113408
    }
  },
  {
    "ph": "s", "id": 113408, "pid": 76337, "tid": -914061504, "ts": 1716454223167235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167244, "dur": 0,
    "args": {
      "External id": 113409,
      "cbid": 51, "correlation": 113409
    }
  },
  {
    "ph": "s", "id": 113409, "pid": 76337, "tid": -914061504, "ts": 1716454223167244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223189227, "dur": 701,
    "args": {
      "External id": 113410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113410,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113410, "pid": 5, "tid": 7, "ts": 1716454223189227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167245, "dur": 5,
    "args": {
      "External id": 113410,
      "cbid": 211, "correlation": 113410
    }
  },
  {
    "ph": "s", "id": 113410, "pid": 76337, "tid": -914061504, "ts": 1716454223167245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223189930, "dur": 59,
    "args": {
      "External id": 113415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113415, "pid": 5, "tid": 7, "ts": 1716454223189930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167273, "dur": 9,
    "args": {
      "External id": 113415,
      "cbid": 211, "correlation": 113415
    }
  },
  {
    "ph": "s", "id": 113415, "pid": 76337, "tid": -914061504, "ts": 1716454223167273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223189990, "dur": 4,
    "args": {
      "External id": 113423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113423,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 113423, "pid": 5, "tid": 7, "ts": 1716454223189990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167315, "dur": 9,
    "args": {
      "External id": 113423,
      "cbid": 211, "correlation": 113423
    }
  },
  {
    "ph": "s", "id": 113423, "pid": 76337, "tid": -914061504, "ts": 1716454223167315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167382, "dur": 1,
    "args": {
      "External id": 113439,
      "cbid": 251, "correlation": 113439
    }
  },
  {
    "ph": "f", "id": 113439, "pid": 76337, "tid": -914061504, "ts": 1716454223167382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167387, "dur": 0,
    "args": {
      "External id": 113441,
      "cbid": 251, "correlation": 113441
    }
  },
  {
    "ph": "f", "id": 113441, "pid": 76337, "tid": -914061504, "ts": 1716454223167387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223189996, "dur": 11,
    "args": {
      "External id": 113442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113442,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 113442, "pid": 5, "tid": 7, "ts": 1716454223189996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167389, "dur": 11,
    "args": {
      "External id": 113442,
      "cbid": 211, "correlation": 113442
    }
  },
  {
    "ph": "s", "id": 113442, "pid": 76337, "tid": -914061504, "ts": 1716454223167389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223190008, "dur": 5,
    "args": {
      "External id": 113444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113444,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 113444, "pid": 5, "tid": 7, "ts": 1716454223190008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167402, "dur": 6,
    "args": {
      "External id": 113444,
      "cbid": 211, "correlation": 113444
    }
  },
  {
    "ph": "s", "id": 113444, "pid": 76337, "tid": -914061504, "ts": 1716454223167402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223190015, "dur": 52,
    "args": {
      "External id": 113454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113454, "pid": 5, "tid": 7, "ts": 1716454223190015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167458, "dur": 13,
    "args": {
      "External id": 113454,
      "cbid": 211, "correlation": 113454
    }
  },
  {
    "ph": "s", "id": 113454, "pid": 76337, "tid": -914061504, "ts": 1716454223167458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223190068, "dur": 51,
    "args": {
      "External id": 113474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113474,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 113474, "pid": 5, "tid": 7, "ts": 1716454223190068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167525, "dur": 11,
    "args": {
      "External id": 113474,
      "cbid": 211, "correlation": 113474
    }
  },
  {
    "ph": "s", "id": 113474, "pid": 76337, "tid": -914061504, "ts": 1716454223167525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223190121, "dur": 4,
    "args": {
      "External id": 113486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113486,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 113486, "pid": 5, "tid": 7, "ts": 1716454223190121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167545, "dur": 6,
    "args": {
      "External id": 113486,
      "cbid": 211, "correlation": 113486
    }
  },
  {
    "ph": "s", "id": 113486, "pid": 76337, "tid": -914061504, "ts": 1716454223167545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223190126, "dur": 56,
    "args": {
      "External id": 113489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113489,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113489, "pid": 5, "tid": 7, "ts": 1716454223190126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167565, "dur": 7,
    "args": {
      "External id": 113489,
      "cbid": 211, "correlation": 113489
    }
  },
  {
    "ph": "s", "id": 113489, "pid": 76337, "tid": -914061504, "ts": 1716454223167565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223190183, "dur": 37,
    "args": {
      "External id": 113498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113498,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113498, "pid": 5, "tid": 7, "ts": 1716454223190183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167605, "dur": 11,
    "args": {
      "External id": 113498,
      "cbid": 211, "correlation": 113498
    }
  },
  {
    "ph": "s", "id": 113498, "pid": 76337, "tid": -914061504, "ts": 1716454223167605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167668, "dur": 0,
    "args": {
      "External id": 113508,
      "cbid": 317, "correlation": 113508
    }
  },
  {
    "ph": "f", "id": 113508, "pid": 76337, "tid": -914061504, "ts": 1716454223167668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167669, "dur": 0,
    "args": {
      "External id": 113509,
      "cbid": 203, "correlation": 113509
    }
  },
  {
    "ph": "f", "id": 113509, "pid": 76337, "tid": -914061504, "ts": 1716454223167669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167670, "dur": 0,
    "args": {
      "External id": 113510,
      "cbid": 205, "correlation": 113510
    }
  },
  {
    "ph": "f", "id": 113510, "pid": 76337, "tid": -914061504, "ts": 1716454223167670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223190221, "dur": 40,
    "args": {
      "External id": 113514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113514,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113514, "pid": 5, "tid": 7, "ts": 1716454223190221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167684, "dur": 12,
    "args": {
      "External id": 113514,
      "cbid": 211, "correlation": 113514
    }
  },
  {
    "ph": "s", "id": 113514, "pid": 76337, "tid": -914061504, "ts": 1716454223167684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223190262, "dur": 14,
    "args": {
      "External id": 113516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113516,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113516, "pid": 5, "tid": 7, "ts": 1716454223190262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167698, "dur": 5,
    "args": {
      "External id": 113516,
      "cbid": 211, "correlation": 113516
    }
  },
  {
    "ph": "s", "id": 113516, "pid": 76337, "tid": -914061504, "ts": 1716454223167698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223190278, "dur": 3,
    "args": {
      "External id": 113518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113518,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 113518, "pid": 5, "tid": 7, "ts": 1716454223190278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167707, "dur": 6,
    "args": {
      "External id": 113518,
      "cbid": 211, "correlation": 113518
    }
  },
  {
    "ph": "s", "id": 113518, "pid": 76337, "tid": -914061504, "ts": 1716454223167707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167716, "dur": 0,
    "args": {
      "External id": 113519,
      "cbid": 51, "correlation": 113519
    }
  },
  {
    "ph": "s", "id": 113519, "pid": 76337, "tid": -914061504, "ts": 1716454223167716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223190282, "dur": 693,
    "args": {
      "External id": 113520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113520,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113520, "pid": 5, "tid": 7, "ts": 1716454223190282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167717, "dur": 5,
    "args": {
      "External id": 113520,
      "cbid": 211, "correlation": 113520
    }
  },
  {
    "ph": "s", "id": 113520, "pid": 76337, "tid": -914061504, "ts": 1716454223167717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223190976, "dur": 59,
    "args": {
      "External id": 113525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113525,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113525, "pid": 5, "tid": 7, "ts": 1716454223190976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167744, "dur": 8,
    "args": {
      "External id": 113525,
      "cbid": 211, "correlation": 113525
    }
  },
  {
    "ph": "s", "id": 113525, "pid": 76337, "tid": -914061504, "ts": 1716454223167744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223191036, "dur": 50,
    "args": {
      "External id": 113533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113533,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113533, "pid": 5, "tid": 7, "ts": 1716454223191036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167776, "dur": 8,
    "args": {
      "External id": 113533,
      "cbid": 211, "correlation": 113533
    }
  },
  {
    "ph": "s", "id": 113533, "pid": 76337, "tid": -914061504, "ts": 1716454223167776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223191087, "dur": 35,
    "args": {
      "External id": 113541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113541,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113541, "pid": 5, "tid": 7, "ts": 1716454223191087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167806, "dur": 9,
    "args": {
      "External id": 113541,
      "cbid": 211, "correlation": 113541
    }
  },
  {
    "ph": "s", "id": 113541, "pid": 76337, "tid": -914061504, "ts": 1716454223167806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223191124, "dur": 51,
    "args": {
      "External id": 113561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113561,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 113561, "pid": 5, "tid": 7, "ts": 1716454223191124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167885, "dur": 12,
    "args": {
      "External id": 113561,
      "cbid": 211, "correlation": 113561
    }
  },
  {
    "ph": "s", "id": 113561, "pid": 76337, "tid": -914061504, "ts": 1716454223167885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223191177, "dur": 4,
    "args": {
      "External id": 113573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113573,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 113573, "pid": 5, "tid": 7, "ts": 1716454223191177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167906, "dur": 6,
    "args": {
      "External id": 113573,
      "cbid": 211, "correlation": 113573
    }
  },
  {
    "ph": "s", "id": 113573, "pid": 76337, "tid": -914061504, "ts": 1716454223167906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223191182, "dur": 56,
    "args": {
      "External id": 113576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113576, "pid": 5, "tid": 7, "ts": 1716454223191182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167924, "dur": 7,
    "args": {
      "External id": 113576,
      "cbid": 211, "correlation": 113576
    }
  },
  {
    "ph": "s", "id": 113576, "pid": 76337, "tid": -914061504, "ts": 1716454223167924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167989, "dur": 0,
    "args": {
      "External id": 113587,
      "cbid": 317, "correlation": 113587
    }
  },
  {
    "ph": "f", "id": 113587, "pid": 76337, "tid": -914061504, "ts": 1716454223167989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167989, "dur": 0,
    "args": {
      "External id": 113588,
      "cbid": 203, "correlation": 113588
    }
  },
  {
    "ph": "f", "id": 113588, "pid": 76337, "tid": -914061504, "ts": 1716454223167989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223167990, "dur": 0,
    "args": {
      "External id": 113589,
      "cbid": 205, "correlation": 113589
    }
  },
  {
    "ph": "f", "id": 113589, "pid": 76337, "tid": -914061504, "ts": 1716454223167990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168011, "dur": 1,
    "args": {
      "External id": 113593,
      "cbid": 251, "correlation": 113593
    }
  },
  {
    "ph": "f", "id": 113593, "pid": 76337, "tid": -914061504, "ts": 1716454223168011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168014, "dur": 0,
    "args": {
      "External id": 113594,
      "cbid": 251, "correlation": 113594
    }
  },
  {
    "ph": "f", "id": 113594, "pid": 76337, "tid": -914061504, "ts": 1716454223168014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168015, "dur": 0,
    "args": {
      "External id": 113595,
      "cbid": 251, "correlation": 113595
    }
  },
  {
    "ph": "f", "id": 113595, "pid": 76337, "tid": -914061504, "ts": 1716454223168015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168016, "dur": 0,
    "args": {
      "External id": 113596,
      "cbid": 251, "correlation": 113596
    }
  },
  {
    "ph": "f", "id": 113596, "pid": 76337, "tid": -914061504, "ts": 1716454223168016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168016, "dur": 0,
    "args": {
      "External id": 113597,
      "cbid": 251, "correlation": 113597
    }
  },
  {
    "ph": "f", "id": 113597, "pid": 76337, "tid": -914061504, "ts": 1716454223168016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168017, "dur": 0,
    "args": {
      "External id": 113598,
      "cbid": 251, "correlation": 113598
    }
  },
  {
    "ph": "f", "id": 113598, "pid": 76337, "tid": -914061504, "ts": 1716454223168017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168018, "dur": 0,
    "args": {
      "External id": 113599,
      "cbid": 251, "correlation": 113599
    }
  },
  {
    "ph": "f", "id": 113599, "pid": 76337, "tid": -914061504, "ts": 1716454223168018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168018, "dur": 0,
    "args": {
      "External id": 113600,
      "cbid": 251, "correlation": 113600
    }
  },
  {
    "ph": "f", "id": 113600, "pid": 76337, "tid": -914061504, "ts": 1716454223168018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168020, "dur": 0,
    "args": {
      "External id": 113601,
      "cbid": 251, "correlation": 113601
    }
  },
  {
    "ph": "f", "id": 113601, "pid": 76337, "tid": -914061504, "ts": 1716454223168020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223191239, "dur": 111,
    "args": {
      "External id": 113602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113602,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113602, "pid": 5, "tid": 7, "ts": 1716454223191239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168022, "dur": 12,
    "args": {
      "External id": 113602,
      "cbid": 211, "correlation": 113602
    }
  },
  {
    "ph": "s", "id": 113602, "pid": 76337, "tid": -914061504, "ts": 1716454223168022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223191352, "dur": 59,
    "args": {
      "External id": 113608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113608, "pid": 5, "tid": 7, "ts": 1716454223191352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168057, "dur": 10,
    "args": {
      "External id": 113608,
      "cbid": 211, "correlation": 113608
    }
  },
  {
    "ph": "s", "id": 113608, "pid": 76337, "tid": -914061504, "ts": 1716454223168057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223191412, "dur": 670,
    "args": {
      "External id": 113617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113617,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113617, "pid": 5, "tid": 7, "ts": 1716454223191412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168140, "dur": 14,
    "args": {
      "External id": 113617,
      "cbid": 211, "correlation": 113617
    }
  },
  {
    "ph": "s", "id": 113617, "pid": 76337, "tid": -914061504, "ts": 1716454223168140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223192083, "dur": 181,
    "args": {
      "External id": 113639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113639,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113639, "pid": 5, "tid": 7, "ts": 1716454223192083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168198, "dur": 11,
    "args": {
      "External id": 113639,
      "cbid": 211, "correlation": 113639
    }
  },
  {
    "ph": "s", "id": 113639, "pid": 76337, "tid": -914061504, "ts": 1716454223168198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168284, "dur": 1,
    "args": {
      "External id": 113650,
      "cbid": 251, "correlation": 113650
    }
  },
  {
    "ph": "f", "id": 113650, "pid": 76337, "tid": -914061504, "ts": 1716454223168284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223192266, "dur": 195,
    "args": {
      "External id": 113651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113651,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113651, "pid": 5, "tid": 7, "ts": 1716454223192266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168289, "dur": 13,
    "args": {
      "External id": 113651,
      "cbid": 211, "correlation": 113651
    }
  },
  {
    "ph": "s", "id": 113651, "pid": 76337, "tid": -914061504, "ts": 1716454223168289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168357, "dur": 1,
    "args": {
      "External id": 113662,
      "cbid": 251, "correlation": 113662
    }
  },
  {
    "ph": "f", "id": 113662, "pid": 76337, "tid": -914061504, "ts": 1716454223168357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223192462, "dur": 185,
    "args": {
      "External id": 113663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113663,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113663, "pid": 5, "tid": 7, "ts": 1716454223192462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168362, "dur": 12,
    "args": {
      "External id": 113663,
      "cbid": 211, "correlation": 113663
    }
  },
  {
    "ph": "s", "id": 113663, "pid": 76337, "tid": -914061504, "ts": 1716454223168362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168425, "dur": 1,
    "args": {
      "External id": 113674,
      "cbid": 251, "correlation": 113674
    }
  },
  {
    "ph": "f", "id": 113674, "pid": 76337, "tid": -914061504, "ts": 1716454223168425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223192648, "dur": 188,
    "args": {
      "External id": 113675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113675,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113675, "pid": 5, "tid": 7, "ts": 1716454223192648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168429, "dur": 11,
    "args": {
      "External id": 113675,
      "cbid": 211, "correlation": 113675
    }
  },
  {
    "ph": "s", "id": 113675, "pid": 76337, "tid": -914061504, "ts": 1716454223168429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223192837, "dur": 18620,
    "args": {
      "External id": 113696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113696,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113696, "pid": 5, "tid": 7, "ts": 1716454223192837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168510, "dur": 13,
    "args": {
      "External id": 113696,
      "cbid": 211, "correlation": 113696
    }
  },
  {
    "ph": "s", "id": 113696, "pid": 76337, "tid": -914061504, "ts": 1716454223168510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168606, "dur": 1,
    "args": {
      "External id": 113714,
      "cbid": 251, "correlation": 113714
    }
  },
  {
    "ph": "f", "id": 113714, "pid": 76337, "tid": -914061504, "ts": 1716454223168606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223211458, "dur": 202,
    "args": {
      "External id": 113716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113716,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113716, "pid": 5, "tid": 7, "ts": 1716454223211458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168612, "dur": 14,
    "args": {
      "External id": 113716,
      "cbid": 211, "correlation": 113716
    }
  },
  {
    "ph": "s", "id": 113716, "pid": 76337, "tid": -914061504, "ts": 1716454223168612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223211662, "dur": 66,
    "args": {
      "External id": 113724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113724,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113724, "pid": 5, "tid": 7, "ts": 1716454223211662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168683, "dur": 12,
    "args": {
      "External id": 113724,
      "cbid": 211, "correlation": 113724
    }
  },
  {
    "ph": "s", "id": 113724, "pid": 76337, "tid": -914061504, "ts": 1716454223168683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223211729, "dur": 97,
    "args": {
      "External id": 113732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113732,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113732, "pid": 5, "tid": 7, "ts": 1716454223211729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168722, "dur": 9,
    "args": {
      "External id": 113732,
      "cbid": 211, "correlation": 113732
    }
  },
  {
    "ph": "s", "id": 113732, "pid": 76337, "tid": -914061504, "ts": 1716454223168722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223211827, "dur": 55,
    "args": {
      "External id": 113743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113743, "pid": 5, "tid": 7, "ts": 1716454223211827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168793, "dur": 12,
    "args": {
      "External id": 113743,
      "cbid": 211, "correlation": 113743
    }
  },
  {
    "ph": "s", "id": 113743, "pid": 76337, "tid": -914061504, "ts": 1716454223168793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223211884, "dur": 92,
    "args": {
      "External id": 113765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113765,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113765, "pid": 5, "tid": 7, "ts": 1716454223211884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168825, "dur": 8,
    "args": {
      "External id": 113765,
      "cbid": 211, "correlation": 113765
    }
  },
  {
    "ph": "s", "id": 113765, "pid": 76337, "tid": -914061504, "ts": 1716454223168825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168908, "dur": 1,
    "args": {
      "External id": 113776,
      "cbid": 251, "correlation": 113776
    }
  },
  {
    "ph": "f", "id": 113776, "pid": 76337, "tid": -914061504, "ts": 1716454223168908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223211977, "dur": 102,
    "args": {
      "External id": 113777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113777,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113777, "pid": 5, "tid": 7, "ts": 1716454223211977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168913, "dur": 13,
    "args": {
      "External id": 113777,
      "cbid": 211, "correlation": 113777
    }
  },
  {
    "ph": "s", "id": 113777, "pid": 76337, "tid": -914061504, "ts": 1716454223168913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168990, "dur": 1,
    "args": {
      "External id": 113788,
      "cbid": 251, "correlation": 113788
    }
  },
  {
    "ph": "f", "id": 113788, "pid": 76337, "tid": -914061504, "ts": 1716454223168990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168994, "dur": 0,
    "args": {
      "External id": 113789,
      "cbid": 251, "correlation": 113789
    }
  },
  {
    "ph": "f", "id": 113789, "pid": 76337, "tid": -914061504, "ts": 1716454223168994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223212080, "dur": 10,
    "args": {
      "External id": 113790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113790,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 113790, "pid": 5, "tid": 7, "ts": 1716454223212080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223168996, "dur": 12,
    "args": {
      "External id": 113790,
      "cbid": 211, "correlation": 113790
    }
  },
  {
    "ph": "s", "id": 113790, "pid": 76337, "tid": -914061504, "ts": 1716454223168996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223212092, "dur": 5,
    "args": {
      "External id": 113792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113792,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 113792, "pid": 5, "tid": 7, "ts": 1716454223212092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169010, "dur": 6,
    "args": {
      "External id": 113792,
      "cbid": 211, "correlation": 113792
    }
  },
  {
    "ph": "s", "id": 113792, "pid": 76337, "tid": -914061504, "ts": 1716454223169010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169073, "dur": 1,
    "args": {
      "External id": 113803,
      "cbid": 251, "correlation": 113803
    }
  },
  {
    "ph": "f", "id": 113803, "pid": 76337, "tid": -914061504, "ts": 1716454223169073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169077, "dur": 0,
    "args": {
      "External id": 113804,
      "cbid": 251, "correlation": 113804
    }
  },
  {
    "ph": "f", "id": 113804, "pid": 76337, "tid": -914061504, "ts": 1716454223169077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223212098, "dur": 6,
    "args": {
      "External id": 113805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113805,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 113805, "pid": 5, "tid": 7, "ts": 1716454223212098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169079, "dur": 11,
    "args": {
      "External id": 113805,
      "cbid": 211, "correlation": 113805
    }
  },
  {
    "ph": "s", "id": 113805, "pid": 76337, "tid": -914061504, "ts": 1716454223169079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223212105, "dur": 3,
    "args": {
      "External id": 113807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113807,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 113807, "pid": 5, "tid": 7, "ts": 1716454223212105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169092, "dur": 5,
    "args": {
      "External id": 113807,
      "cbid": 211, "correlation": 113807
    }
  },
  {
    "ph": "s", "id": 113807, "pid": 76337, "tid": -914061504, "ts": 1716454223169092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223212110, "dur": 155,
    "args": {
      "External id": 113828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113828,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113828, "pid": 5, "tid": 7, "ts": 1716454223212110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169165, "dur": 13,
    "args": {
      "External id": 113828,
      "cbid": 211, "correlation": 113828
    }
  },
  {
    "ph": "s", "id": 113828, "pid": 76337, "tid": -914061504, "ts": 1716454223169165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169262, "dur": 1,
    "args": {
      "External id": 113846,
      "cbid": 251, "correlation": 113846
    }
  },
  {
    "ph": "f", "id": 113846, "pid": 76337, "tid": -914061504, "ts": 1716454223169262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223212267, "dur": 105,
    "args": {
      "External id": 113848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113848,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113848, "pid": 5, "tid": 7, "ts": 1716454223212267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169268, "dur": 14,
    "args": {
      "External id": 113848,
      "cbid": 211, "correlation": 113848
    }
  },
  {
    "ph": "s", "id": 113848, "pid": 76337, "tid": -914061504, "ts": 1716454223169268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223212374, "dur": 35,
    "args": {
      "External id": 113856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113856,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113856, "pid": 5, "tid": 7, "ts": 1716454223212374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169338, "dur": 12,
    "args": {
      "External id": 113856,
      "cbid": 211, "correlation": 113856
    }
  },
  {
    "ph": "s", "id": 113856, "pid": 76337, "tid": -914061504, "ts": 1716454223169338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223212410, "dur": 67,
    "args": {
      "External id": 113864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113864, "pid": 5, "tid": 7, "ts": 1716454223212410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169379, "dur": 9,
    "args": {
      "External id": 113864,
      "cbid": 211, "correlation": 113864
    }
  },
  {
    "ph": "s", "id": 113864, "pid": 76337, "tid": -914061504, "ts": 1716454223169379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223212478, "dur": 92,
    "args": {
      "External id": 113886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113886,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113886, "pid": 5, "tid": 7, "ts": 1716454223212478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169430, "dur": 10,
    "args": {
      "External id": 113886,
      "cbid": 211, "correlation": 113886
    }
  },
  {
    "ph": "s", "id": 113886, "pid": 76337, "tid": -914061504, "ts": 1716454223169430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169515, "dur": 1,
    "args": {
      "External id": 113902,
      "cbid": 251, "correlation": 113902
    }
  },
  {
    "ph": "f", "id": 113902, "pid": 76337, "tid": -914061504, "ts": 1716454223169515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223212572, "dur": 573,
    "args": {
      "External id": 113904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113904,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 113904, "pid": 5, "tid": 7, "ts": 1716454223212572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169520, "dur": 13,
    "args": {
      "External id": 113904,
      "cbid": 211, "correlation": 113904
    }
  },
  {
    "ph": "s", "id": 113904, "pid": 76337, "tid": -914061504, "ts": 1716454223169520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223213146, "dur": 240,
    "args": {
      "External id": 113912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113912, "pid": 5, "tid": 7, "ts": 1716454223213146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169586, "dur": 12,
    "args": {
      "External id": 113912,
      "cbid": 211, "correlation": 113912
    }
  },
  {
    "ph": "s", "id": 113912, "pid": 76337, "tid": -914061504, "ts": 1716454223169586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223213388, "dur": 253,
    "args": {
      "External id": 113920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113920,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113920, "pid": 5, "tid": 7, "ts": 1716454223213388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169616, "dur": 8,
    "args": {
      "External id": 113920,
      "cbid": 211, "correlation": 113920
    }
  },
  {
    "ph": "s", "id": 113920, "pid": 76337, "tid": -914061504, "ts": 1716454223169616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169697, "dur": 1,
    "args": {
      "External id": 113936,
      "cbid": 251, "correlation": 113936
    }
  },
  {
    "ph": "f", "id": 113936, "pid": 76337, "tid": -914061504, "ts": 1716454223169697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169702, "dur": 0,
    "args": {
      "External id": 113938,
      "cbid": 251, "correlation": 113938
    }
  },
  {
    "ph": "f", "id": 113938, "pid": 76337, "tid": -914061504, "ts": 1716454223169702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223213642, "dur": 358,
    "args": {
      "External id": 113939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113939,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 113939, "pid": 5, "tid": 7, "ts": 1716454223213642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169705, "dur": 13,
    "args": {
      "External id": 113939,
      "cbid": 211, "correlation": 113939
    }
  },
  {
    "ph": "s", "id": 113939, "pid": 76337, "tid": -914061504, "ts": 1716454223169705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223214001, "dur": 50,
    "args": {
      "External id": 113947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113947,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113947, "pid": 5, "tid": 7, "ts": 1716454223214001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169748, "dur": 9,
    "args": {
      "External id": 113947,
      "cbid": 211, "correlation": 113947
    }
  },
  {
    "ph": "s", "id": 113947, "pid": 76337, "tid": -914061504, "ts": 1716454223169748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223214052, "dur": 157,
    "args": {
      "External id": 113958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113958, "pid": 5, "tid": 7, "ts": 1716454223214052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169814, "dur": 12,
    "args": {
      "External id": 113958,
      "cbid": 211, "correlation": 113958
    }
  },
  {
    "ph": "s", "id": 113958, "pid": 76337, "tid": -914061504, "ts": 1716454223169814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169878, "dur": 0,
    "args": {
      "External id": 113970,
      "cbid": 317, "correlation": 113970
    }
  },
  {
    "ph": "f", "id": 113970, "pid": 76337, "tid": -914061504, "ts": 1716454223169878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169879, "dur": 0,
    "args": {
      "External id": 113971,
      "cbid": 203, "correlation": 113971
    }
  },
  {
    "ph": "f", "id": 113971, "pid": 76337, "tid": -914061504, "ts": 1716454223169879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169880, "dur": 0,
    "args": {
      "External id": 113972,
      "cbid": 205, "correlation": 113972
    }
  },
  {
    "ph": "f", "id": 113972, "pid": 76337, "tid": -914061504, "ts": 1716454223169880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169902, "dur": 1,
    "args": {
      "External id": 113976,
      "cbid": 251, "correlation": 113976
    }
  },
  {
    "ph": "f", "id": 113976, "pid": 76337, "tid": -914061504, "ts": 1716454223169902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169904, "dur": 0,
    "args": {
      "External id": 113977,
      "cbid": 251, "correlation": 113977
    }
  },
  {
    "ph": "f", "id": 113977, "pid": 76337, "tid": -914061504, "ts": 1716454223169904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169904, "dur": 0,
    "args": {
      "External id": 113978,
      "cbid": 251, "correlation": 113978
    }
  },
  {
    "ph": "f", "id": 113978, "pid": 76337, "tid": -914061504, "ts": 1716454223169904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169905, "dur": 0,
    "args": {
      "External id": 113979,
      "cbid": 251, "correlation": 113979
    }
  },
  {
    "ph": "f", "id": 113979, "pid": 76337, "tid": -914061504, "ts": 1716454223169905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169906, "dur": 0,
    "args": {
      "External id": 113980,
      "cbid": 251, "correlation": 113980
    }
  },
  {
    "ph": "f", "id": 113980, "pid": 76337, "tid": -914061504, "ts": 1716454223169906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169906, "dur": 0,
    "args": {
      "External id": 113981,
      "cbid": 251, "correlation": 113981
    }
  },
  {
    "ph": "f", "id": 113981, "pid": 76337, "tid": -914061504, "ts": 1716454223169906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169907, "dur": 0,
    "args": {
      "External id": 113982,
      "cbid": 251, "correlation": 113982
    }
  },
  {
    "ph": "f", "id": 113982, "pid": 76337, "tid": -914061504, "ts": 1716454223169907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169908, "dur": 0,
    "args": {
      "External id": 113983,
      "cbid": 251, "correlation": 113983
    }
  },
  {
    "ph": "f", "id": 113983, "pid": 76337, "tid": -914061504, "ts": 1716454223169908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169909, "dur": 0,
    "args": {
      "External id": 113984,
      "cbid": 251, "correlation": 113984
    }
  },
  {
    "ph": "f", "id": 113984, "pid": 76337, "tid": -914061504, "ts": 1716454223169909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223214211, "dur": 115,
    "args": {
      "External id": 113985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113985,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 113985, "pid": 5, "tid": 7, "ts": 1716454223214211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169911, "dur": 12,
    "args": {
      "External id": 113985,
      "cbid": 211, "correlation": 113985
    }
  },
  {
    "ph": "s", "id": 113985, "pid": 76337, "tid": -914061504, "ts": 1716454223169911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223214327, "dur": 59,
    "args": {
      "External id": 113991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113991, "pid": 5, "tid": 7, "ts": 1716454223214327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169946, "dur": 8,
    "args": {
      "External id": 113991,
      "cbid": 211, "correlation": 113991
    }
  },
  {
    "ph": "s", "id": 113991, "pid": 76337, "tid": -914061504, "ts": 1716454223169946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223214388, "dur": 50,
    "args": {
      "External id": 113999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 113999,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 113999, "pid": 5, "tid": 7, "ts": 1716454223214388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223169985, "dur": 9,
    "args": {
      "External id": 113999,
      "cbid": 211, "correlation": 113999
    }
  },
  {
    "ph": "s", "id": 113999, "pid": 76337, "tid": -914061504, "ts": 1716454223169985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170060, "dur": 0,
    "args": {
      "External id": 114009,
      "cbid": 317, "correlation": 114009
    }
  },
  {
    "ph": "f", "id": 114009, "pid": 76337, "tid": -914061504, "ts": 1716454223170060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170061, "dur": 0,
    "args": {
      "External id": 114010,
      "cbid": 203, "correlation": 114010
    }
  },
  {
    "ph": "f", "id": 114010, "pid": 76337, "tid": -914061504, "ts": 1716454223170061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170061, "dur": 0,
    "args": {
      "External id": 114011,
      "cbid": 205, "correlation": 114011
    }
  },
  {
    "ph": "f", "id": 114011, "pid": 76337, "tid": -914061504, "ts": 1716454223170061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223214439, "dur": 41,
    "args": {
      "External id": 114015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114015,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114015, "pid": 5, "tid": 7, "ts": 1716454223214439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170078, "dur": 13,
    "args": {
      "External id": 114015,
      "cbid": 211, "correlation": 114015
    }
  },
  {
    "ph": "s", "id": 114015, "pid": 76337, "tid": -914061504, "ts": 1716454223170078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223214482, "dur": 14,
    "args": {
      "External id": 114017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114017,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114017, "pid": 5, "tid": 7, "ts": 1716454223214482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170093, "dur": 5,
    "args": {
      "External id": 114017,
      "cbid": 211, "correlation": 114017
    }
  },
  {
    "ph": "s", "id": 114017, "pid": 76337, "tid": -914061504, "ts": 1716454223170093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223214498, "dur": 1,
    "args": {
      "External id": 114019,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 114019,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 114019, "pid": 5, "tid": 7, "ts": 1716454223214498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170112, "dur": 17,
    "args": {
      "External id": 114019,
      "cbid": 51, "correlation": 114019
    }
  },
  {
    "ph": "s", "id": 114019, "pid": 76337, "tid": -914061504, "ts": 1716454223170112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223214502, "dur": 361,
    "args": {
      "External id": 114020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114020,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114020, "pid": 5, "tid": 7, "ts": 1716454223214502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170132, "dur": 9,
    "args": {
      "External id": 114020,
      "cbid": 211, "correlation": 114020
    }
  },
  {
    "ph": "s", "id": 114020, "pid": 76337, "tid": -914061504, "ts": 1716454223170132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223214865, "dur": 14,
    "args": {
      "External id": 114022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114022,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114022, "pid": 5, "tid": 7, "ts": 1716454223214865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170149, "dur": 7,
    "args": {
      "External id": 114022,
      "cbid": 211, "correlation": 114022
    }
  },
  {
    "ph": "s", "id": 114022, "pid": 76337, "tid": -914061504, "ts": 1716454223170149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223214880, "dur": 14,
    "args": {
      "External id": 114028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114028, "pid": 5, "tid": 7, "ts": 1716454223214880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170180, "dur": 8,
    "args": {
      "External id": 114028,
      "cbid": 211, "correlation": 114028
    }
  },
  {
    "ph": "s", "id": 114028, "pid": 76337, "tid": -914061504, "ts": 1716454223170180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223214896, "dur": 19,
    "args": {
      "External id": 114048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114048,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 114048, "pid": 5, "tid": 7, "ts": 1716454223214896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170271, "dur": 13,
    "args": {
      "External id": 114048,
      "cbid": 211, "correlation": 114048
    }
  },
  {
    "ph": "s", "id": 114048, "pid": 76337, "tid": -914061504, "ts": 1716454223170271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223214916, "dur": 4,
    "args": {
      "External id": 114060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114060,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 114060, "pid": 5, "tid": 7, "ts": 1716454223214916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170293, "dur": 6,
    "args": {
      "External id": 114060,
      "cbid": 211, "correlation": 114060
    }
  },
  {
    "ph": "s", "id": 114060, "pid": 76337, "tid": -914061504, "ts": 1716454223170293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223214921, "dur": 18,
    "args": {
      "External id": 114063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114063,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114063, "pid": 5, "tid": 7, "ts": 1716454223214921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170311, "dur": 7,
    "args": {
      "External id": 114063,
      "cbid": 211, "correlation": 114063
    }
  },
  {
    "ph": "s", "id": 114063, "pid": 76337, "tid": -914061504, "ts": 1716454223170311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223214940, "dur": 11,
    "args": {
      "External id": 114072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114072,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114072, "pid": 5, "tid": 7, "ts": 1716454223214940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170351, "dur": 10,
    "args": {
      "External id": 114072,
      "cbid": 211, "correlation": 114072
    }
  },
  {
    "ph": "s", "id": 114072, "pid": 76337, "tid": -914061504, "ts": 1716454223170351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170407, "dur": 0,
    "args": {
      "External id": 114082,
      "cbid": 317, "correlation": 114082
    }
  },
  {
    "ph": "f", "id": 114082, "pid": 76337, "tid": -914061504, "ts": 1716454223170407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170408, "dur": 0,
    "args": {
      "External id": 114083,
      "cbid": 203, "correlation": 114083
    }
  },
  {
    "ph": "f", "id": 114083, "pid": 76337, "tid": -914061504, "ts": 1716454223170408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170409, "dur": 0,
    "args": {
      "External id": 114084,
      "cbid": 205, "correlation": 114084
    }
  },
  {
    "ph": "f", "id": 114084, "pid": 76337, "tid": -914061504, "ts": 1716454223170409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223214953, "dur": 11,
    "args": {
      "External id": 114088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114088, "pid": 5, "tid": 7, "ts": 1716454223214953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170424, "dur": 11,
    "args": {
      "External id": 114088,
      "cbid": 211, "correlation": 114088
    }
  },
  {
    "ph": "s", "id": 114088, "pid": 76337, "tid": -914061504, "ts": 1716454223170424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223214965, "dur": 25,
    "args": {
      "External id": 114090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114090, "pid": 5, "tid": 7, "ts": 1716454223214965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170438, "dur": 5,
    "args": {
      "External id": 114090,
      "cbid": 211, "correlation": 114090
    }
  },
  {
    "ph": "s", "id": 114090, "pid": 76337, "tid": -914061504, "ts": 1716454223170438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223214991, "dur": 4,
    "args": {
      "External id": 114092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 114092, "pid": 5, "tid": 7, "ts": 1716454223214991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170449, "dur": 5,
    "args": {
      "External id": 114092,
      "cbid": 211, "correlation": 114092
    }
  },
  {
    "ph": "s", "id": 114092, "pid": 76337, "tid": -914061504, "ts": 1716454223170449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170459, "dur": 0,
    "args": {
      "External id": 114093,
      "cbid": 51, "correlation": 114093
    }
  },
  {
    "ph": "s", "id": 114093, "pid": 76337, "tid": -914061504, "ts": 1716454223170459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223214996, "dur": 355,
    "args": {
      "External id": 114094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114094,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114094, "pid": 5, "tid": 7, "ts": 1716454223214996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170460, "dur": 8,
    "args": {
      "External id": 114094,
      "cbid": 211, "correlation": 114094
    }
  },
  {
    "ph": "s", "id": 114094, "pid": 76337, "tid": -914061504, "ts": 1716454223170460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223215353, "dur": 20,
    "args": {
      "External id": 114095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114095,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114095, "pid": 5, "tid": 7, "ts": 1716454223215353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170471, "dur": 5,
    "args": {
      "External id": 114095,
      "cbid": 211, "correlation": 114095
    }
  },
  {
    "ph": "s", "id": 114095, "pid": 76337, "tid": -914061504, "ts": 1716454223170471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223215374, "dur": 33,
    "args": {
      "External id": 114101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114101, "pid": 5, "tid": 7, "ts": 1716454223215374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170499, "dur": 9,
    "args": {
      "External id": 114101,
      "cbid": 211, "correlation": 114101
    }
  },
  {
    "ph": "s", "id": 114101, "pid": 76337, "tid": -914061504, "ts": 1716454223170499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223215408, "dur": 5,
    "args": {
      "External id": 114109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114109,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 114109, "pid": 5, "tid": 7, "ts": 1716454223215408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170543, "dur": 9,
    "args": {
      "External id": 114109,
      "cbid": 211, "correlation": 114109
    }
  },
  {
    "ph": "s", "id": 114109, "pid": 76337, "tid": -914061504, "ts": 1716454223170543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170608, "dur": 1,
    "args": {
      "External id": 114125,
      "cbid": 251, "correlation": 114125
    }
  },
  {
    "ph": "f", "id": 114125, "pid": 76337, "tid": -914061504, "ts": 1716454223170608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170613, "dur": 0,
    "args": {
      "External id": 114127,
      "cbid": 251, "correlation": 114127
    }
  },
  {
    "ph": "f", "id": 114127, "pid": 76337, "tid": -914061504, "ts": 1716454223170613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223215414, "dur": 13,
    "args": {
      "External id": 114128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114128,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 114128, "pid": 5, "tid": 7, "ts": 1716454223215414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170615, "dur": 12,
    "args": {
      "External id": 114128,
      "cbid": 211, "correlation": 114128
    }
  },
  {
    "ph": "s", "id": 114128, "pid": 76337, "tid": -914061504, "ts": 1716454223170615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223215428, "dur": 5,
    "args": {
      "External id": 114130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114130,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 114130, "pid": 5, "tid": 7, "ts": 1716454223215428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170629, "dur": 6,
    "args": {
      "External id": 114130,
      "cbid": 211, "correlation": 114130
    }
  },
  {
    "ph": "s", "id": 114130, "pid": 76337, "tid": -914061504, "ts": 1716454223170629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223215435, "dur": 30,
    "args": {
      "External id": 114140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114140, "pid": 5, "tid": 7, "ts": 1716454223215435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170687, "dur": 11,
    "args": {
      "External id": 114140,
      "cbid": 211, "correlation": 114140
    }
  },
  {
    "ph": "s", "id": 114140, "pid": 76337, "tid": -914061504, "ts": 1716454223170687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223215466, "dur": 31,
    "args": {
      "External id": 114160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114160,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 114160, "pid": 5, "tid": 7, "ts": 1716454223215466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170752, "dur": 11,
    "args": {
      "External id": 114160,
      "cbid": 211, "correlation": 114160
    }
  },
  {
    "ph": "s", "id": 114160, "pid": 76337, "tid": -914061504, "ts": 1716454223170752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223215498, "dur": 4,
    "args": {
      "External id": 114172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114172,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 114172, "pid": 5, "tid": 7, "ts": 1716454223215498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170773, "dur": 7,
    "args": {
      "External id": 114172,
      "cbid": 211, "correlation": 114172
    }
  },
  {
    "ph": "s", "id": 114172, "pid": 76337, "tid": -914061504, "ts": 1716454223170773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223215504, "dur": 29,
    "args": {
      "External id": 114175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114175, "pid": 5, "tid": 7, "ts": 1716454223215504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170792, "dur": 7,
    "args": {
      "External id": 114175,
      "cbid": 211, "correlation": 114175
    }
  },
  {
    "ph": "s", "id": 114175, "pid": 76337, "tid": -914061504, "ts": 1716454223170792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223215534, "dur": 21,
    "args": {
      "External id": 114184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114184,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114184, "pid": 5, "tid": 7, "ts": 1716454223215534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170833, "dur": 9,
    "args": {
      "External id": 114184,
      "cbid": 211, "correlation": 114184
    }
  },
  {
    "ph": "s", "id": 114184, "pid": 76337, "tid": -914061504, "ts": 1716454223170833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170896, "dur": 0,
    "args": {
      "External id": 114194,
      "cbid": 317, "correlation": 114194
    }
  },
  {
    "ph": "f", "id": 114194, "pid": 76337, "tid": -914061504, "ts": 1716454223170896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170897, "dur": 0,
    "args": {
      "External id": 114195,
      "cbid": 203, "correlation": 114195
    }
  },
  {
    "ph": "f", "id": 114195, "pid": 76337, "tid": -914061504, "ts": 1716454223170897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170898, "dur": 0,
    "args": {
      "External id": 114196,
      "cbid": 205, "correlation": 114196
    }
  },
  {
    "ph": "f", "id": 114196, "pid": 76337, "tid": -914061504, "ts": 1716454223170898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223215557, "dur": 23,
    "args": {
      "External id": 114200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114200,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114200, "pid": 5, "tid": 7, "ts": 1716454223215557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170917, "dur": 12,
    "args": {
      "External id": 114200,
      "cbid": 211, "correlation": 114200
    }
  },
  {
    "ph": "s", "id": 114200, "pid": 76337, "tid": -914061504, "ts": 1716454223170917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223215581, "dur": 44,
    "args": {
      "External id": 114202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114202, "pid": 5, "tid": 7, "ts": 1716454223215581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170932, "dur": 5,
    "args": {
      "External id": 114202,
      "cbid": 211, "correlation": 114202
    }
  },
  {
    "ph": "s", "id": 114202, "pid": 76337, "tid": -914061504, "ts": 1716454223170932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223215627, "dur": 651,
    "args": {
      "External id": 114204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114204,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114204, "pid": 5, "tid": 7, "ts": 1716454223215627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170946, "dur": 10,
    "args": {
      "External id": 114204,
      "cbid": 211, "correlation": 114204
    }
  },
  {
    "ph": "s", "id": 114204, "pid": 76337, "tid": -914061504, "ts": 1716454223170946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223216279, "dur": 21,
    "args": {
      "External id": 114206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114206,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114206, "pid": 5, "tid": 7, "ts": 1716454223216279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223170960, "dur": 5,
    "args": {
      "External id": 114206,
      "cbid": 211, "correlation": 114206
    }
  },
  {
    "ph": "s", "id": 114206, "pid": 76337, "tid": -914061504, "ts": 1716454223170960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223216301, "dur": 33,
    "args": {
      "External id": 114212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114212, "pid": 5, "tid": 7, "ts": 1716454223216301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171005, "dur": 9,
    "args": {
      "External id": 114212,
      "cbid": 211, "correlation": 114212
    }
  },
  {
    "ph": "s", "id": 114212, "pid": 76337, "tid": -914061504, "ts": 1716454223171005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171066, "dur": 0,
    "args": {
      "External id": 114222,
      "cbid": 317, "correlation": 114222
    }
  },
  {
    "ph": "f", "id": 114222, "pid": 76337, "tid": -914061504, "ts": 1716454223171066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171067, "dur": 0,
    "args": {
      "External id": 114223,
      "cbid": 203, "correlation": 114223
    }
  },
  {
    "ph": "f", "id": 114223, "pid": 76337, "tid": -914061504, "ts": 1716454223171067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171068, "dur": 0,
    "args": {
      "External id": 114224,
      "cbid": 205, "correlation": 114224
    }
  },
  {
    "ph": "f", "id": 114224, "pid": 76337, "tid": -914061504, "ts": 1716454223171068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171090, "dur": 1,
    "args": {
      "External id": 114228,
      "cbid": 251, "correlation": 114228
    }
  },
  {
    "ph": "f", "id": 114228, "pid": 76337, "tid": -914061504, "ts": 1716454223171090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171092, "dur": 0,
    "args": {
      "External id": 114229,
      "cbid": 251, "correlation": 114229
    }
  },
  {
    "ph": "f", "id": 114229, "pid": 76337, "tid": -914061504, "ts": 1716454223171092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171093, "dur": 0,
    "args": {
      "External id": 114230,
      "cbid": 251, "correlation": 114230
    }
  },
  {
    "ph": "f", "id": 114230, "pid": 76337, "tid": -914061504, "ts": 1716454223171093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171093, "dur": 0,
    "args": {
      "External id": 114231,
      "cbid": 251, "correlation": 114231
    }
  },
  {
    "ph": "f", "id": 114231, "pid": 76337, "tid": -914061504, "ts": 1716454223171093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171094, "dur": 0,
    "args": {
      "External id": 114232,
      "cbid": 251, "correlation": 114232
    }
  },
  {
    "ph": "f", "id": 114232, "pid": 76337, "tid": -914061504, "ts": 1716454223171094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171095, "dur": 0,
    "args": {
      "External id": 114233,
      "cbid": 251, "correlation": 114233
    }
  },
  {
    "ph": "f", "id": 114233, "pid": 76337, "tid": -914061504, "ts": 1716454223171095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171095, "dur": 0,
    "args": {
      "External id": 114234,
      "cbid": 251, "correlation": 114234
    }
  },
  {
    "ph": "f", "id": 114234, "pid": 76337, "tid": -914061504, "ts": 1716454223171095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171096, "dur": 0,
    "args": {
      "External id": 114235,
      "cbid": 251, "correlation": 114235
    }
  },
  {
    "ph": "f", "id": 114235, "pid": 76337, "tid": -914061504, "ts": 1716454223171096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171097, "dur": 0,
    "args": {
      "External id": 114236,
      "cbid": 251, "correlation": 114236
    }
  },
  {
    "ph": "f", "id": 114236, "pid": 76337, "tid": -914061504, "ts": 1716454223171097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223216335, "dur": 52,
    "args": {
      "External id": 114237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114237,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 114237, "pid": 5, "tid": 7, "ts": 1716454223216335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171099, "dur": 12,
    "args": {
      "External id": 114237,
      "cbid": 211, "correlation": 114237
    }
  },
  {
    "ph": "s", "id": 114237, "pid": 76337, "tid": -914061504, "ts": 1716454223171099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223216388, "dur": 32,
    "args": {
      "External id": 114243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114243, "pid": 5, "tid": 7, "ts": 1716454223216388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171132, "dur": 8,
    "args": {
      "External id": 114243,
      "cbid": 211, "correlation": 114243
    }
  },
  {
    "ph": "s", "id": 114243, "pid": 76337, "tid": -914061504, "ts": 1716454223171132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223216422, "dur": 27,
    "args": {
      "External id": 114251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114251,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114251, "pid": 5, "tid": 7, "ts": 1716454223216422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171161, "dur": 8,
    "args": {
      "External id": 114251,
      "cbid": 211, "correlation": 114251
    }
  },
  {
    "ph": "s", "id": 114251, "pid": 76337, "tid": -914061504, "ts": 1716454223171161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223216450, "dur": 20,
    "args": {
      "External id": 114259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114259,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114259, "pid": 5, "tid": 7, "ts": 1716454223216450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171191, "dur": 8,
    "args": {
      "External id": 114259,
      "cbid": 211, "correlation": 114259
    }
  },
  {
    "ph": "s", "id": 114259, "pid": 76337, "tid": -914061504, "ts": 1716454223171191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223216470, "dur": 30,
    "args": {
      "External id": 114279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114279,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 114279, "pid": 5, "tid": 7, "ts": 1716454223216470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171272, "dur": 13,
    "args": {
      "External id": 114279,
      "cbid": 211, "correlation": 114279
    }
  },
  {
    "ph": "s", "id": 114279, "pid": 76337, "tid": -914061504, "ts": 1716454223171272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223216502, "dur": 4,
    "args": {
      "External id": 114291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114291,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 114291, "pid": 5, "tid": 7, "ts": 1716454223216502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171295, "dur": 6,
    "args": {
      "External id": 114291,
      "cbid": 211, "correlation": 114291
    }
  },
  {
    "ph": "s", "id": 114291, "pid": 76337, "tid": -914061504, "ts": 1716454223171295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223216508, "dur": 30,
    "args": {
      "External id": 114294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114294, "pid": 5, "tid": 7, "ts": 1716454223216508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171313, "dur": 6,
    "args": {
      "External id": 114294,
      "cbid": 211, "correlation": 114294
    }
  },
  {
    "ph": "s", "id": 114294, "pid": 76337, "tid": -914061504, "ts": 1716454223171313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171371, "dur": 0,
    "args": {
      "External id": 114305,
      "cbid": 317, "correlation": 114305
    }
  },
  {
    "ph": "f", "id": 114305, "pid": 76337, "tid": -914061504, "ts": 1716454223171371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171372, "dur": 0,
    "args": {
      "External id": 114306,
      "cbid": 203, "correlation": 114306
    }
  },
  {
    "ph": "f", "id": 114306, "pid": 76337, "tid": -914061504, "ts": 1716454223171372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171372, "dur": 0,
    "args": {
      "External id": 114307,
      "cbid": 205, "correlation": 114307
    }
  },
  {
    "ph": "f", "id": 114307, "pid": 76337, "tid": -914061504, "ts": 1716454223171372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223216539, "dur": 22,
    "args": {
      "External id": 114311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114311,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114311, "pid": 5, "tid": 7, "ts": 1716454223216539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171389, "dur": 12,
    "args": {
      "External id": 114311,
      "cbid": 211, "correlation": 114311
    }
  },
  {
    "ph": "s", "id": 114311, "pid": 76337, "tid": -914061504, "ts": 1716454223171389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223216562, "dur": 121,
    "args": {
      "External id": 114313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114313,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114313, "pid": 5, "tid": 7, "ts": 1716454223216562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171409, "dur": 8,
    "args": {
      "External id": 114313,
      "cbid": 211, "correlation": 114313
    }
  },
  {
    "ph": "s", "id": 114313, "pid": 76337, "tid": -914061504, "ts": 1716454223171409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223216684, "dur": 21,
    "args": {
      "External id": 114315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114315,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114315, "pid": 5, "tid": 7, "ts": 1716454223216684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171422, "dur": 6,
    "args": {
      "External id": 114315,
      "cbid": 211, "correlation": 114315
    }
  },
  {
    "ph": "s", "id": 114315, "pid": 76337, "tid": -914061504, "ts": 1716454223171422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223216706, "dur": 32,
    "args": {
      "External id": 114321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114321, "pid": 5, "tid": 7, "ts": 1716454223216706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171450, "dur": 8,
    "args": {
      "External id": 114321,
      "cbid": 211, "correlation": 114321
    }
  },
  {
    "ph": "s", "id": 114321, "pid": 76337, "tid": -914061504, "ts": 1716454223171450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223216740, "dur": 188,
    "args": {
      "External id": 114330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114330,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114330, "pid": 5, "tid": 7, "ts": 1716454223216740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171533, "dur": 14,
    "args": {
      "External id": 114330,
      "cbid": 211, "correlation": 114330
    }
  },
  {
    "ph": "s", "id": 114330, "pid": 76337, "tid": -914061504, "ts": 1716454223171533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223216929, "dur": 66,
    "args": {
      "External id": 114352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114352,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114352, "pid": 5, "tid": 7, "ts": 1716454223216929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171591, "dur": 10,
    "args": {
      "External id": 114352,
      "cbid": 211, "correlation": 114352
    }
  },
  {
    "ph": "s", "id": 114352, "pid": 76337, "tid": -914061504, "ts": 1716454223171591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171683, "dur": 1,
    "args": {
      "External id": 114363,
      "cbid": 251, "correlation": 114363
    }
  },
  {
    "ph": "f", "id": 114363, "pid": 76337, "tid": -914061504, "ts": 1716454223171683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223216996, "dur": 155,
    "args": {
      "External id": 114364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114364,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114364, "pid": 5, "tid": 7, "ts": 1716454223216996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171688, "dur": 13,
    "args": {
      "External id": 114364,
      "cbid": 211, "correlation": 114364
    }
  },
  {
    "ph": "s", "id": 114364, "pid": 76337, "tid": -914061504, "ts": 1716454223171688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171758, "dur": 1,
    "args": {
      "External id": 114375,
      "cbid": 251, "correlation": 114375
    }
  },
  {
    "ph": "f", "id": 114375, "pid": 76337, "tid": -914061504, "ts": 1716454223171758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223217152, "dur": 148,
    "args": {
      "External id": 114376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114376,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114376, "pid": 5, "tid": 7, "ts": 1716454223217152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171762, "dur": 11,
    "args": {
      "External id": 114376,
      "cbid": 211, "correlation": 114376
    }
  },
  {
    "ph": "s", "id": 114376, "pid": 76337, "tid": -914061504, "ts": 1716454223171762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171827, "dur": 1,
    "args": {
      "External id": 114387,
      "cbid": 251, "correlation": 114387
    }
  },
  {
    "ph": "f", "id": 114387, "pid": 76337, "tid": -914061504, "ts": 1716454223171827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223217302, "dur": 149,
    "args": {
      "External id": 114388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114388,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114388, "pid": 5, "tid": 7, "ts": 1716454223217302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171831, "dur": 11,
    "args": {
      "External id": 114388,
      "cbid": 211, "correlation": 114388
    }
  },
  {
    "ph": "s", "id": 114388, "pid": 76337, "tid": -914061504, "ts": 1716454223171831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223217452, "dur": 1947,
    "args": {
      "External id": 114409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114409,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 114409, "pid": 5, "tid": 7, "ts": 1716454223217452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223171914, "dur": 13,
    "args": {
      "External id": 114409,
      "cbid": 211, "correlation": 114409
    }
  },
  {
    "ph": "s", "id": 114409, "pid": 76337, "tid": -914061504, "ts": 1716454223171914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172021, "dur": 1,
    "args": {
      "External id": 114427,
      "cbid": 251, "correlation": 114427
    }
  },
  {
    "ph": "f", "id": 114427, "pid": 76337, "tid": -914061504, "ts": 1716454223172021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223219400, "dur": 147,
    "args": {
      "External id": 114429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114429,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 114429, "pid": 5, "tid": 7, "ts": 1716454223219400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172026, "dur": 14,
    "args": {
      "External id": 114429,
      "cbid": 211, "correlation": 114429
    }
  },
  {
    "ph": "s", "id": 114429, "pid": 76337, "tid": -914061504, "ts": 1716454223172026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223219549, "dur": 35,
    "args": {
      "External id": 114437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114437,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114437, "pid": 5, "tid": 7, "ts": 1716454223219549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172099, "dur": 12,
    "args": {
      "External id": 114437,
      "cbid": 211, "correlation": 114437
    }
  },
  {
    "ph": "s", "id": 114437, "pid": 76337, "tid": -914061504, "ts": 1716454223172099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223219585, "dur": 50,
    "args": {
      "External id": 114445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114445,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114445, "pid": 5, "tid": 7, "ts": 1716454223219585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172138, "dur": 9,
    "args": {
      "External id": 114445,
      "cbid": 211, "correlation": 114445
    }
  },
  {
    "ph": "s", "id": 114445, "pid": 76337, "tid": -914061504, "ts": 1716454223172138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223219637, "dur": 31,
    "args": {
      "External id": 114456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114456, "pid": 5, "tid": 7, "ts": 1716454223219637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172211, "dur": 13,
    "args": {
      "External id": 114456,
      "cbid": 211, "correlation": 114456
    }
  },
  {
    "ph": "s", "id": 114456, "pid": 76337, "tid": -914061504, "ts": 1716454223172211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223219669, "dur": 35,
    "args": {
      "External id": 114478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114478,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114478, "pid": 5, "tid": 7, "ts": 1716454223219669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172244, "dur": 7,
    "args": {
      "External id": 114478,
      "cbid": 211, "correlation": 114478
    }
  },
  {
    "ph": "s", "id": 114478, "pid": 76337, "tid": -914061504, "ts": 1716454223172244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172327, "dur": 1,
    "args": {
      "External id": 114489,
      "cbid": 251, "correlation": 114489
    }
  },
  {
    "ph": "f", "id": 114489, "pid": 76337, "tid": -914061504, "ts": 1716454223172327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223219705, "dur": 90,
    "args": {
      "External id": 114490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114490,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114490, "pid": 5, "tid": 7, "ts": 1716454223219705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172332, "dur": 14,
    "args": {
      "External id": 114490,
      "cbid": 211, "correlation": 114490
    }
  },
  {
    "ph": "s", "id": 114490, "pid": 76337, "tid": -914061504, "ts": 1716454223172332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172403, "dur": 1,
    "args": {
      "External id": 114501,
      "cbid": 251, "correlation": 114501
    }
  },
  {
    "ph": "f", "id": 114501, "pid": 76337, "tid": -914061504, "ts": 1716454223172403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172406, "dur": 0,
    "args": {
      "External id": 114502,
      "cbid": 251, "correlation": 114502
    }
  },
  {
    "ph": "f", "id": 114502, "pid": 76337, "tid": -914061504, "ts": 1716454223172406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223219797, "dur": 11,
    "args": {
      "External id": 114503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114503,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 114503, "pid": 5, "tid": 7, "ts": 1716454223219797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172408, "dur": 12,
    "args": {
      "External id": 114503,
      "cbid": 211, "correlation": 114503
    }
  },
  {
    "ph": "s", "id": 114503, "pid": 76337, "tid": -914061504, "ts": 1716454223172408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223219809, "dur": 5,
    "args": {
      "External id": 114505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114505,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 114505, "pid": 5, "tid": 7, "ts": 1716454223219809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172423, "dur": 7,
    "args": {
      "External id": 114505,
      "cbid": 211, "correlation": 114505
    }
  },
  {
    "ph": "s", "id": 114505, "pid": 76337, "tid": -914061504, "ts": 1716454223172423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172482, "dur": 1,
    "args": {
      "External id": 114516,
      "cbid": 251, "correlation": 114516
    }
  },
  {
    "ph": "f", "id": 114516, "pid": 76337, "tid": -914061504, "ts": 1716454223172482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172486, "dur": 0,
    "args": {
      "External id": 114517,
      "cbid": 251, "correlation": 114517
    }
  },
  {
    "ph": "f", "id": 114517, "pid": 76337, "tid": -914061504, "ts": 1716454223172486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223219815, "dur": 7,
    "args": {
      "External id": 114518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114518,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 114518, "pid": 5, "tid": 7, "ts": 1716454223219815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172487, "dur": 12,
    "args": {
      "External id": 114518,
      "cbid": 211, "correlation": 114518
    }
  },
  {
    "ph": "s", "id": 114518, "pid": 76337, "tid": -914061504, "ts": 1716454223172487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223219824, "dur": 3,
    "args": {
      "External id": 114520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114520,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 114520, "pid": 5, "tid": 7, "ts": 1716454223219824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172501, "dur": 5,
    "args": {
      "External id": 114520,
      "cbid": 211, "correlation": 114520
    }
  },
  {
    "ph": "s", "id": 114520, "pid": 76337, "tid": -914061504, "ts": 1716454223172501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223219828, "dur": 91,
    "args": {
      "External id": 114541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114541,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 114541, "pid": 5, "tid": 7, "ts": 1716454223219828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172574, "dur": 13,
    "args": {
      "External id": 114541,
      "cbid": 211, "correlation": 114541
    }
  },
  {
    "ph": "s", "id": 114541, "pid": 76337, "tid": -914061504, "ts": 1716454223172574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172670, "dur": 1,
    "args": {
      "External id": 114559,
      "cbid": 251, "correlation": 114559
    }
  },
  {
    "ph": "f", "id": 114559, "pid": 76337, "tid": -914061504, "ts": 1716454223172670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223219920, "dur": 101,
    "args": {
      "External id": 114561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114561,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114561, "pid": 5, "tid": 7, "ts": 1716454223219920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172676, "dur": 13,
    "args": {
      "External id": 114561,
      "cbid": 211, "correlation": 114561
    }
  },
  {
    "ph": "s", "id": 114561, "pid": 76337, "tid": -914061504, "ts": 1716454223172676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223220023, "dur": 19,
    "args": {
      "External id": 114569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114569,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114569, "pid": 5, "tid": 7, "ts": 1716454223220023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172745, "dur": 12,
    "args": {
      "External id": 114569,
      "cbid": 211, "correlation": 114569
    }
  },
  {
    "ph": "s", "id": 114569, "pid": 76337, "tid": -914061504, "ts": 1716454223172745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223220043, "dur": 37,
    "args": {
      "External id": 114577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114577, "pid": 5, "tid": 7, "ts": 1716454223220043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172786, "dur": 10,
    "args": {
      "External id": 114577,
      "cbid": 211, "correlation": 114577
    }
  },
  {
    "ph": "s", "id": 114577, "pid": 76337, "tid": -914061504, "ts": 1716454223172786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223220082, "dur": 34,
    "args": {
      "External id": 114599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114599,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114599, "pid": 5, "tid": 7, "ts": 1716454223220082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172837, "dur": 11,
    "args": {
      "External id": 114599,
      "cbid": 211, "correlation": 114599
    }
  },
  {
    "ph": "s", "id": 114599, "pid": 76337, "tid": -914061504, "ts": 1716454223172837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172927, "dur": 1,
    "args": {
      "External id": 114615,
      "cbid": 251, "correlation": 114615
    }
  },
  {
    "ph": "f", "id": 114615, "pid": 76337, "tid": -914061504, "ts": 1716454223172927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172932, "dur": 0,
    "args": {
      "External id": 114617,
      "cbid": 251, "correlation": 114617
    }
  },
  {
    "ph": "f", "id": 114617, "pid": 76337, "tid": -914061504, "ts": 1716454223172932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223220117, "dur": 542,
    "args": {
      "External id": 114618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114618,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 114618, "pid": 5, "tid": 7, "ts": 1716454223220117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223172935, "dur": 14,
    "args": {
      "External id": 114618,
      "cbid": 211, "correlation": 114618
    }
  },
  {
    "ph": "s", "id": 114618, "pid": 76337, "tid": -914061504, "ts": 1716454223172935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223220661, "dur": 125,
    "args": {
      "External id": 114626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114626, "pid": 5, "tid": 7, "ts": 1716454223220661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173010, "dur": 13,
    "args": {
      "External id": 114626,
      "cbid": 211, "correlation": 114626
    }
  },
  {
    "ph": "s", "id": 114626, "pid": 76337, "tid": -914061504, "ts": 1716454223173010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223220787, "dur": 127,
    "args": {
      "External id": 114634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114634, "pid": 5, "tid": 7, "ts": 1716454223220787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173040, "dur": 9,
    "args": {
      "External id": 114634,
      "cbid": 211, "correlation": 114634
    }
  },
  {
    "ph": "s", "id": 114634, "pid": 76337, "tid": -914061504, "ts": 1716454223173040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173120, "dur": 1,
    "args": {
      "External id": 114650,
      "cbid": 251, "correlation": 114650
    }
  },
  {
    "ph": "f", "id": 114650, "pid": 76337, "tid": -914061504, "ts": 1716454223173120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223220915, "dur": 304,
    "args": {
      "External id": 114652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114652,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114652, "pid": 5, "tid": 7, "ts": 1716454223220915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173126, "dur": 12,
    "args": {
      "External id": 114652,
      "cbid": 211, "correlation": 114652
    }
  },
  {
    "ph": "s", "id": 114652, "pid": 76337, "tid": -914061504, "ts": 1716454223173126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223221221, "dur": 27,
    "args": {
      "External id": 114660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114660,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114660, "pid": 5, "tid": 7, "ts": 1716454223221221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173169, "dur": 10,
    "args": {
      "External id": 114660,
      "cbid": 211, "correlation": 114660
    }
  },
  {
    "ph": "s", "id": 114660, "pid": 76337, "tid": -914061504, "ts": 1716454223173169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223221249, "dur": 82,
    "args": {
      "External id": 114671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114671, "pid": 5, "tid": 7, "ts": 1716454223221249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173237, "dur": 13,
    "args": {
      "External id": 114671,
      "cbid": 211, "correlation": 114671
    }
  },
  {
    "ph": "s", "id": 114671, "pid": 76337, "tid": -914061504, "ts": 1716454223173237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173301, "dur": 0,
    "args": {
      "External id": 114683,
      "cbid": 317, "correlation": 114683
    }
  },
  {
    "ph": "f", "id": 114683, "pid": 76337, "tid": -914061504, "ts": 1716454223173301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173302, "dur": 0,
    "args": {
      "External id": 114684,
      "cbid": 203, "correlation": 114684
    }
  },
  {
    "ph": "f", "id": 114684, "pid": 76337, "tid": -914061504, "ts": 1716454223173302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173303, "dur": 0,
    "args": {
      "External id": 114685,
      "cbid": 205, "correlation": 114685
    }
  },
  {
    "ph": "f", "id": 114685, "pid": 76337, "tid": -914061504, "ts": 1716454223173303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223221332, "dur": 23,
    "args": {
      "External id": 114689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114689,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114689, "pid": 5, "tid": 7, "ts": 1716454223221332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173318, "dur": 12,
    "args": {
      "External id": 114689,
      "cbid": 211, "correlation": 114689
    }
  },
  {
    "ph": "s", "id": 114689, "pid": 76337, "tid": -914061504, "ts": 1716454223173318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223221356, "dur": 120,
    "args": {
      "External id": 114691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114691,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114691, "pid": 5, "tid": 7, "ts": 1716454223221356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173337, "dur": 7,
    "args": {
      "External id": 114691,
      "cbid": 211, "correlation": 114691
    }
  },
  {
    "ph": "s", "id": 114691, "pid": 76337, "tid": -914061504, "ts": 1716454223173337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223221478, "dur": 22,
    "args": {
      "External id": 114693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114693,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114693, "pid": 5, "tid": 7, "ts": 1716454223221478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173349, "dur": 5,
    "args": {
      "External id": 114693,
      "cbid": 211, "correlation": 114693
    }
  },
  {
    "ph": "s", "id": 114693, "pid": 76337, "tid": -914061504, "ts": 1716454223173349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223221501, "dur": 32,
    "args": {
      "External id": 114699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114699, "pid": 5, "tid": 7, "ts": 1716454223221501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173376, "dur": 8,
    "args": {
      "External id": 114699,
      "cbid": 211, "correlation": 114699
    }
  },
  {
    "ph": "s", "id": 114699, "pid": 76337, "tid": -914061504, "ts": 1716454223173376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223221535, "dur": 27,
    "args": {
      "External id": 114707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114707,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114707, "pid": 5, "tid": 7, "ts": 1716454223221535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173408, "dur": 8,
    "args": {
      "External id": 114707,
      "cbid": 211, "correlation": 114707
    }
  },
  {
    "ph": "s", "id": 114707, "pid": 76337, "tid": -914061504, "ts": 1716454223173408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223221563, "dur": 32,
    "args": {
      "External id": 114727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114727,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 114727, "pid": 5, "tid": 7, "ts": 1716454223221563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173480, "dur": 12,
    "args": {
      "External id": 114727,
      "cbid": 211, "correlation": 114727
    }
  },
  {
    "ph": "s", "id": 114727, "pid": 76337, "tid": -914061504, "ts": 1716454223173480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223221596, "dur": 5,
    "args": {
      "External id": 114739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114739,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 114739, "pid": 5, "tid": 7, "ts": 1716454223221596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173503, "dur": 6,
    "args": {
      "External id": 114739,
      "cbid": 211, "correlation": 114739
    }
  },
  {
    "ph": "s", "id": 114739, "pid": 76337, "tid": -914061504, "ts": 1716454223173503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223221602, "dur": 32,
    "args": {
      "External id": 114742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114742, "pid": 5, "tid": 7, "ts": 1716454223221602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173521, "dur": 6,
    "args": {
      "External id": 114742,
      "cbid": 211, "correlation": 114742
    }
  },
  {
    "ph": "s", "id": 114742, "pid": 76337, "tid": -914061504, "ts": 1716454223173521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223221635, "dur": 21,
    "args": {
      "External id": 114751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114751,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114751, "pid": 5, "tid": 7, "ts": 1716454223221635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173560, "dur": 10,
    "args": {
      "External id": 114751,
      "cbid": 211, "correlation": 114751
    }
  },
  {
    "ph": "s", "id": 114751, "pid": 76337, "tid": -914061504, "ts": 1716454223173560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173611, "dur": 0,
    "args": {
      "External id": 114761,
      "cbid": 317, "correlation": 114761
    }
  },
  {
    "ph": "f", "id": 114761, "pid": 76337, "tid": -914061504, "ts": 1716454223173611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173612, "dur": 0,
    "args": {
      "External id": 114762,
      "cbid": 203, "correlation": 114762
    }
  },
  {
    "ph": "f", "id": 114762, "pid": 76337, "tid": -914061504, "ts": 1716454223173612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173613, "dur": 0,
    "args": {
      "External id": 114763,
      "cbid": 205, "correlation": 114763
    }
  },
  {
    "ph": "f", "id": 114763, "pid": 76337, "tid": -914061504, "ts": 1716454223173613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223221658, "dur": 24,
    "args": {
      "External id": 114767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114767,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114767, "pid": 5, "tid": 7, "ts": 1716454223221658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173625, "dur": 11,
    "args": {
      "External id": 114767,
      "cbid": 211, "correlation": 114767
    }
  },
  {
    "ph": "s", "id": 114767, "pid": 76337, "tid": -914061504, "ts": 1716454223173625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223221683, "dur": 44,
    "args": {
      "External id": 114769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114769,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114769, "pid": 5, "tid": 7, "ts": 1716454223221683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173639, "dur": 5,
    "args": {
      "External id": 114769,
      "cbid": 211, "correlation": 114769
    }
  },
  {
    "ph": "s", "id": 114769, "pid": 76337, "tid": -914061504, "ts": 1716454223173639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223221729, "dur": 653,
    "args": {
      "External id": 114771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114771,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114771, "pid": 5, "tid": 7, "ts": 1716454223221729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173651, "dur": 6,
    "args": {
      "External id": 114771,
      "cbid": 211, "correlation": 114771
    }
  },
  {
    "ph": "s", "id": 114771, "pid": 76337, "tid": -914061504, "ts": 1716454223173651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223222383, "dur": 22,
    "args": {
      "External id": 114773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114773, "pid": 5, "tid": 7, "ts": 1716454223222383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173661, "dur": 5,
    "args": {
      "External id": 114773,
      "cbid": 211, "correlation": 114773
    }
  },
  {
    "ph": "s", "id": 114773, "pid": 76337, "tid": -914061504, "ts": 1716454223173661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223222407, "dur": 33,
    "args": {
      "External id": 114779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114779, "pid": 5, "tid": 7, "ts": 1716454223222407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173687, "dur": 9,
    "args": {
      "External id": 114779,
      "cbid": 211, "correlation": 114779
    }
  },
  {
    "ph": "s", "id": 114779, "pid": 76337, "tid": -914061504, "ts": 1716454223173687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223222441, "dur": 4,
    "args": {
      "External id": 114787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114787,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 114787, "pid": 5, "tid": 7, "ts": 1716454223222441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173731, "dur": 9,
    "args": {
      "External id": 114787,
      "cbid": 211, "correlation": 114787
    }
  },
  {
    "ph": "s", "id": 114787, "pid": 76337, "tid": -914061504, "ts": 1716454223173731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173795, "dur": 1,
    "args": {
      "External id": 114803,
      "cbid": 251, "correlation": 114803
    }
  },
  {
    "ph": "f", "id": 114803, "pid": 76337, "tid": -914061504, "ts": 1716454223173795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173801, "dur": 0,
    "args": {
      "External id": 114805,
      "cbid": 251, "correlation": 114805
    }
  },
  {
    "ph": "f", "id": 114805, "pid": 76337, "tid": -914061504, "ts": 1716454223173801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223222446, "dur": 12,
    "args": {
      "External id": 114806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114806,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 114806, "pid": 5, "tid": 7, "ts": 1716454223222446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173803, "dur": 11,
    "args": {
      "External id": 114806,
      "cbid": 211, "correlation": 114806
    }
  },
  {
    "ph": "s", "id": 114806, "pid": 76337, "tid": -914061504, "ts": 1716454223173803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223222460, "dur": 5,
    "args": {
      "External id": 114808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114808,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 114808, "pid": 5, "tid": 7, "ts": 1716454223222460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173817, "dur": 5,
    "args": {
      "External id": 114808,
      "cbid": 211, "correlation": 114808
    }
  },
  {
    "ph": "s", "id": 114808, "pid": 76337, "tid": -914061504, "ts": 1716454223173817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223222466, "dur": 29,
    "args": {
      "External id": 114818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114818, "pid": 5, "tid": 7, "ts": 1716454223222466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173873, "dur": 12,
    "args": {
      "External id": 114818,
      "cbid": 211, "correlation": 114818
    }
  },
  {
    "ph": "s", "id": 114818, "pid": 76337, "tid": -914061504, "ts": 1716454223173873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223222496, "dur": 31,
    "args": {
      "External id": 114838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114838,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 114838, "pid": 5, "tid": 7, "ts": 1716454223222496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173938, "dur": 11,
    "args": {
      "External id": 114838,
      "cbid": 211, "correlation": 114838
    }
  },
  {
    "ph": "s", "id": 114838, "pid": 76337, "tid": -914061504, "ts": 1716454223173938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223222529, "dur": 4,
    "args": {
      "External id": 114850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114850,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 114850, "pid": 5, "tid": 7, "ts": 1716454223222529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173960, "dur": 6,
    "args": {
      "External id": 114850,
      "cbid": 211, "correlation": 114850
    }
  },
  {
    "ph": "s", "id": 114850, "pid": 76337, "tid": -914061504, "ts": 1716454223173960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223222534, "dur": 30,
    "args": {
      "External id": 114853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114853, "pid": 5, "tid": 7, "ts": 1716454223222534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223173987, "dur": 7,
    "args": {
      "External id": 114853,
      "cbid": 211, "correlation": 114853
    }
  },
  {
    "ph": "s", "id": 114853, "pid": 76337, "tid": -914061504, "ts": 1716454223173987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223222565, "dur": 21,
    "args": {
      "External id": 114862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114862,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114862, "pid": 5, "tid": 7, "ts": 1716454223222565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174030, "dur": 9,
    "args": {
      "External id": 114862,
      "cbid": 211, "correlation": 114862
    }
  },
  {
    "ph": "s", "id": 114862, "pid": 76337, "tid": -914061504, "ts": 1716454223174030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174092, "dur": 0,
    "args": {
      "External id": 114872,
      "cbid": 317, "correlation": 114872
    }
  },
  {
    "ph": "f", "id": 114872, "pid": 76337, "tid": -914061504, "ts": 1716454223174092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174093, "dur": 0,
    "args": {
      "External id": 114873,
      "cbid": 203, "correlation": 114873
    }
  },
  {
    "ph": "f", "id": 114873, "pid": 76337, "tid": -914061504, "ts": 1716454223174093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174093, "dur": 0,
    "args": {
      "External id": 114874,
      "cbid": 205, "correlation": 114874
    }
  },
  {
    "ph": "f", "id": 114874, "pid": 76337, "tid": -914061504, "ts": 1716454223174093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223222588, "dur": 24,
    "args": {
      "External id": 114878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114878,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114878, "pid": 5, "tid": 7, "ts": 1716454223222588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174107, "dur": 12,
    "args": {
      "External id": 114878,
      "cbid": 211, "correlation": 114878
    }
  },
  {
    "ph": "s", "id": 114878, "pid": 76337, "tid": -914061504, "ts": 1716454223174107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223222613, "dur": 44,
    "args": {
      "External id": 114880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114880, "pid": 5, "tid": 7, "ts": 1716454223222613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174122, "dur": 5,
    "args": {
      "External id": 114880,
      "cbid": 211, "correlation": 114880
    }
  },
  {
    "ph": "s", "id": 114880, "pid": 76337, "tid": -914061504, "ts": 1716454223174122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223222658, "dur": 643,
    "args": {
      "External id": 114882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114882,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114882, "pid": 5, "tid": 7, "ts": 1716454223222658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174133, "dur": 6,
    "args": {
      "External id": 114882,
      "cbid": 211, "correlation": 114882
    }
  },
  {
    "ph": "s", "id": 114882, "pid": 76337, "tid": -914061504, "ts": 1716454223174133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223223302, "dur": 22,
    "args": {
      "External id": 114884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114884,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114884, "pid": 5, "tid": 7, "ts": 1716454223223302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174142, "dur": 5,
    "args": {
      "External id": 114884,
      "cbid": 211, "correlation": 114884
    }
  },
  {
    "ph": "s", "id": 114884, "pid": 76337, "tid": -914061504, "ts": 1716454223174142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223223326, "dur": 34,
    "args": {
      "External id": 114890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114890, "pid": 5, "tid": 7, "ts": 1716454223223326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174170, "dur": 8,
    "args": {
      "External id": 114890,
      "cbid": 211, "correlation": 114890
    }
  },
  {
    "ph": "s", "id": 114890, "pid": 76337, "tid": -914061504, "ts": 1716454223174170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223223361, "dur": 27,
    "args": {
      "External id": 114898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114898,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114898, "pid": 5, "tid": 7, "ts": 1716454223223361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174202, "dur": 9,
    "args": {
      "External id": 114898,
      "cbid": 211, "correlation": 114898
    }
  },
  {
    "ph": "s", "id": 114898, "pid": 76337, "tid": -914061504, "ts": 1716454223174202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223223389, "dur": 20,
    "args": {
      "External id": 114906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114906,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114906, "pid": 5, "tid": 7, "ts": 1716454223223389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174231, "dur": 8,
    "args": {
      "External id": 114906,
      "cbid": 211, "correlation": 114906
    }
  },
  {
    "ph": "s", "id": 114906, "pid": 76337, "tid": -914061504, "ts": 1716454223174231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223223410, "dur": 29,
    "args": {
      "External id": 114926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114926,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 114926, "pid": 5, "tid": 7, "ts": 1716454223223410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174311, "dur": 12,
    "args": {
      "External id": 114926,
      "cbid": 211, "correlation": 114926
    }
  },
  {
    "ph": "s", "id": 114926, "pid": 76337, "tid": -914061504, "ts": 1716454223174311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223223441, "dur": 4,
    "args": {
      "External id": 114938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114938,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 114938, "pid": 5, "tid": 7, "ts": 1716454223223441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174333, "dur": 6,
    "args": {
      "External id": 114938,
      "cbid": 211, "correlation": 114938
    }
  },
  {
    "ph": "s", "id": 114938, "pid": 76337, "tid": -914061504, "ts": 1716454223174333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223223446, "dur": 29,
    "args": {
      "External id": 114941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114941, "pid": 5, "tid": 7, "ts": 1716454223223446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174350, "dur": 6,
    "args": {
      "External id": 114941,
      "cbid": 211, "correlation": 114941
    }
  },
  {
    "ph": "s", "id": 114941, "pid": 76337, "tid": -914061504, "ts": 1716454223174350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174408, "dur": 0,
    "args": {
      "External id": 114952,
      "cbid": 317, "correlation": 114952
    }
  },
  {
    "ph": "f", "id": 114952, "pid": 76337, "tid": -914061504, "ts": 1716454223174408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174409, "dur": 0,
    "args": {
      "External id": 114953,
      "cbid": 203, "correlation": 114953
    }
  },
  {
    "ph": "f", "id": 114953, "pid": 76337, "tid": -914061504, "ts": 1716454223174409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174410, "dur": 0,
    "args": {
      "External id": 114954,
      "cbid": 205, "correlation": 114954
    }
  },
  {
    "ph": "f", "id": 114954, "pid": 76337, "tid": -914061504, "ts": 1716454223174410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223223477, "dur": 22,
    "args": {
      "External id": 114958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114958,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114958, "pid": 5, "tid": 7, "ts": 1716454223223477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174424, "dur": 12,
    "args": {
      "External id": 114958,
      "cbid": 211, "correlation": 114958
    }
  },
  {
    "ph": "s", "id": 114958, "pid": 76337, "tid": -914061504, "ts": 1716454223174424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223223500, "dur": 116,
    "args": {
      "External id": 114960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114960,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 114960, "pid": 5, "tid": 7, "ts": 1716454223223500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174441, "dur": 6,
    "args": {
      "External id": 114960,
      "cbid": 211, "correlation": 114960
    }
  },
  {
    "ph": "s", "id": 114960, "pid": 76337, "tid": -914061504, "ts": 1716454223174441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223223617, "dur": 21,
    "args": {
      "External id": 114962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114962,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114962, "pid": 5, "tid": 7, "ts": 1716454223223617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174451, "dur": 5,
    "args": {
      "External id": 114962,
      "cbid": 211, "correlation": 114962
    }
  },
  {
    "ph": "s", "id": 114962, "pid": 76337, "tid": -914061504, "ts": 1716454223174451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223223640, "dur": 32,
    "args": {
      "External id": 114968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114968, "pid": 5, "tid": 7, "ts": 1716454223223640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174478, "dur": 8,
    "args": {
      "External id": 114968,
      "cbid": 211, "correlation": 114968
    }
  },
  {
    "ph": "s", "id": 114968, "pid": 76337, "tid": -914061504, "ts": 1716454223174478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223223673, "dur": 169,
    "args": {
      "External id": 114977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114977,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114977, "pid": 5, "tid": 7, "ts": 1716454223223673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174560, "dur": 14,
    "args": {
      "External id": 114977,
      "cbid": 211, "correlation": 114977
    }
  },
  {
    "ph": "s", "id": 114977, "pid": 76337, "tid": -914061504, "ts": 1716454223174560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223223844, "dur": 65,
    "args": {
      "External id": 114999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 114999,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 114999, "pid": 5, "tid": 7, "ts": 1716454223223844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174617, "dur": 10,
    "args": {
      "External id": 114999,
      "cbid": 211, "correlation": 114999
    }
  },
  {
    "ph": "s", "id": 114999, "pid": 76337, "tid": -914061504, "ts": 1716454223174617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174703, "dur": 1,
    "args": {
      "External id": 115010,
      "cbid": 251, "correlation": 115010
    }
  },
  {
    "ph": "f", "id": 115010, "pid": 76337, "tid": -914061504, "ts": 1716454223174703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223223910, "dur": 156,
    "args": {
      "External id": 115011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115011,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115011, "pid": 5, "tid": 7, "ts": 1716454223223910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174709, "dur": 14,
    "args": {
      "External id": 115011,
      "cbid": 211, "correlation": 115011
    }
  },
  {
    "ph": "s", "id": 115011, "pid": 76337, "tid": -914061504, "ts": 1716454223174709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174779, "dur": 1,
    "args": {
      "External id": 115022,
      "cbid": 251, "correlation": 115022
    }
  },
  {
    "ph": "f", "id": 115022, "pid": 76337, "tid": -914061504, "ts": 1716454223174779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223224067, "dur": 146,
    "args": {
      "External id": 115023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115023,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115023, "pid": 5, "tid": 7, "ts": 1716454223224067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174783, "dur": 11,
    "args": {
      "External id": 115023,
      "cbid": 211, "correlation": 115023
    }
  },
  {
    "ph": "s", "id": 115023, "pid": 76337, "tid": -914061504, "ts": 1716454223174783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174848, "dur": 1,
    "args": {
      "External id": 115034,
      "cbid": 251, "correlation": 115034
    }
  },
  {
    "ph": "f", "id": 115034, "pid": 76337, "tid": -914061504, "ts": 1716454223174848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223224215, "dur": 144,
    "args": {
      "External id": 115035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115035,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115035, "pid": 5, "tid": 7, "ts": 1716454223224215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174852, "dur": 11,
    "args": {
      "External id": 115035,
      "cbid": 211, "correlation": 115035
    }
  },
  {
    "ph": "s", "id": 115035, "pid": 76337, "tid": -914061504, "ts": 1716454223174852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223224360, "dur": 1946,
    "args": {
      "External id": 115056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115056,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 115056, "pid": 5, "tid": 7, "ts": 1716454223224360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223174934, "dur": 12,
    "args": {
      "External id": 115056,
      "cbid": 211, "correlation": 115056
    }
  },
  {
    "ph": "s", "id": 115056, "pid": 76337, "tid": -914061504, "ts": 1716454223174934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175040, "dur": 1,
    "args": {
      "External id": 115074,
      "cbid": 251, "correlation": 115074
    }
  },
  {
    "ph": "f", "id": 115074, "pid": 76337, "tid": -914061504, "ts": 1716454223175040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223226307, "dur": 147,
    "args": {
      "External id": 115076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115076,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 115076, "pid": 5, "tid": 7, "ts": 1716454223226307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175045, "dur": 14,
    "args": {
      "External id": 115076,
      "cbid": 211, "correlation": 115076
    }
  },
  {
    "ph": "s", "id": 115076, "pid": 76337, "tid": -914061504, "ts": 1716454223175045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223226456, "dur": 35,
    "args": {
      "External id": 115084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115084,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115084, "pid": 5, "tid": 7, "ts": 1716454223226456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175116, "dur": 12,
    "args": {
      "External id": 115084,
      "cbid": 211, "correlation": 115084
    }
  },
  {
    "ph": "s", "id": 115084, "pid": 76337, "tid": -914061504, "ts": 1716454223175116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223226492, "dur": 50,
    "args": {
      "External id": 115092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115092,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115092, "pid": 5, "tid": 7, "ts": 1716454223226492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175155, "dur": 9,
    "args": {
      "External id": 115092,
      "cbid": 211, "correlation": 115092
    }
  },
  {
    "ph": "s", "id": 115092, "pid": 76337, "tid": -914061504, "ts": 1716454223175155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223226544, "dur": 30,
    "args": {
      "External id": 115103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115103,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115103, "pid": 5, "tid": 7, "ts": 1716454223226544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175227, "dur": 12,
    "args": {
      "External id": 115103,
      "cbid": 211, "correlation": 115103
    }
  },
  {
    "ph": "s", "id": 115103, "pid": 76337, "tid": -914061504, "ts": 1716454223175227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223226575, "dur": 35,
    "args": {
      "External id": 115125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115125,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115125, "pid": 5, "tid": 7, "ts": 1716454223226575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175256, "dur": 8,
    "args": {
      "External id": 115125,
      "cbid": 211, "correlation": 115125
    }
  },
  {
    "ph": "s", "id": 115125, "pid": 76337, "tid": -914061504, "ts": 1716454223175256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175342, "dur": 1,
    "args": {
      "External id": 115136,
      "cbid": 251, "correlation": 115136
    }
  },
  {
    "ph": "f", "id": 115136, "pid": 76337, "tid": -914061504, "ts": 1716454223175342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223226611, "dur": 89,
    "args": {
      "External id": 115137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115137,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115137, "pid": 5, "tid": 7, "ts": 1716454223226611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175347, "dur": 13,
    "args": {
      "External id": 115137,
      "cbid": 211, "correlation": 115137
    }
  },
  {
    "ph": "s", "id": 115137, "pid": 76337, "tid": -914061504, "ts": 1716454223175347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175417, "dur": 1,
    "args": {
      "External id": 115148,
      "cbid": 251, "correlation": 115148
    }
  },
  {
    "ph": "f", "id": 115148, "pid": 76337, "tid": -914061504, "ts": 1716454223175417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175420, "dur": 0,
    "args": {
      "External id": 115149,
      "cbid": 251, "correlation": 115149
    }
  },
  {
    "ph": "f", "id": 115149, "pid": 76337, "tid": -914061504, "ts": 1716454223175420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223226702, "dur": 12,
    "args": {
      "External id": 115150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115150,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 115150, "pid": 5, "tid": 7, "ts": 1716454223226702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175422, "dur": 12,
    "args": {
      "External id": 115150,
      "cbid": 211, "correlation": 115150
    }
  },
  {
    "ph": "s", "id": 115150, "pid": 76337, "tid": -914061504, "ts": 1716454223175422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223226715, "dur": 5,
    "args": {
      "External id": 115152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115152,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 115152, "pid": 5, "tid": 7, "ts": 1716454223226715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175435, "dur": 6,
    "args": {
      "External id": 115152,
      "cbid": 211, "correlation": 115152
    }
  },
  {
    "ph": "s", "id": 115152, "pid": 76337, "tid": -914061504, "ts": 1716454223175435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175493, "dur": 1,
    "args": {
      "External id": 115163,
      "cbid": 251, "correlation": 115163
    }
  },
  {
    "ph": "f", "id": 115163, "pid": 76337, "tid": -914061504, "ts": 1716454223175493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175497, "dur": 0,
    "args": {
      "External id": 115164,
      "cbid": 251, "correlation": 115164
    }
  },
  {
    "ph": "f", "id": 115164, "pid": 76337, "tid": -914061504, "ts": 1716454223175497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223226721, "dur": 7,
    "args": {
      "External id": 115165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115165,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 115165, "pid": 5, "tid": 7, "ts": 1716454223226721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175498, "dur": 12,
    "args": {
      "External id": 115165,
      "cbid": 211, "correlation": 115165
    }
  },
  {
    "ph": "s", "id": 115165, "pid": 76337, "tid": -914061504, "ts": 1716454223175498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223226730, "dur": 3,
    "args": {
      "External id": 115167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115167,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 115167, "pid": 5, "tid": 7, "ts": 1716454223226730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175511, "dur": 6,
    "args": {
      "External id": 115167,
      "cbid": 211, "correlation": 115167
    }
  },
  {
    "ph": "s", "id": 115167, "pid": 76337, "tid": -914061504, "ts": 1716454223175511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223226734, "dur": 92,
    "args": {
      "External id": 115188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115188,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 115188, "pid": 5, "tid": 7, "ts": 1716454223226734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175586, "dur": 12,
    "args": {
      "External id": 115188,
      "cbid": 211, "correlation": 115188
    }
  },
  {
    "ph": "s", "id": 115188, "pid": 76337, "tid": -914061504, "ts": 1716454223175586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175683, "dur": 1,
    "args": {
      "External id": 115206,
      "cbid": 251, "correlation": 115206
    }
  },
  {
    "ph": "f", "id": 115206, "pid": 76337, "tid": -914061504, "ts": 1716454223175683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223226827, "dur": 99,
    "args": {
      "External id": 115208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115208,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115208, "pid": 5, "tid": 7, "ts": 1716454223226827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175688, "dur": 13,
    "args": {
      "External id": 115208,
      "cbid": 211, "correlation": 115208
    }
  },
  {
    "ph": "s", "id": 115208, "pid": 76337, "tid": -914061504, "ts": 1716454223175688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223226928, "dur": 19,
    "args": {
      "External id": 115216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115216,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115216, "pid": 5, "tid": 7, "ts": 1716454223226928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175757, "dur": 13,
    "args": {
      "External id": 115216,
      "cbid": 211, "correlation": 115216
    }
  },
  {
    "ph": "s", "id": 115216, "pid": 76337, "tid": -914061504, "ts": 1716454223175757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223226948, "dur": 37,
    "args": {
      "External id": 115224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115224,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115224, "pid": 5, "tid": 7, "ts": 1716454223226948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175800, "dur": 9,
    "args": {
      "External id": 115224,
      "cbid": 211, "correlation": 115224
    }
  },
  {
    "ph": "s", "id": 115224, "pid": 76337, "tid": -914061504, "ts": 1716454223175800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223226986, "dur": 35,
    "args": {
      "External id": 115246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115246,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115246, "pid": 5, "tid": 7, "ts": 1716454223226986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175851, "dur": 10,
    "args": {
      "External id": 115246,
      "cbid": 211, "correlation": 115246
    }
  },
  {
    "ph": "s", "id": 115246, "pid": 76337, "tid": -914061504, "ts": 1716454223175851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175939, "dur": 1,
    "args": {
      "External id": 115262,
      "cbid": 251, "correlation": 115262
    }
  },
  {
    "ph": "f", "id": 115262, "pid": 76337, "tid": -914061504, "ts": 1716454223175939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175944, "dur": 0,
    "args": {
      "External id": 115264,
      "cbid": 251, "correlation": 115264
    }
  },
  {
    "ph": "f", "id": 115264, "pid": 76337, "tid": -914061504, "ts": 1716454223175944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223227023, "dur": 541,
    "args": {
      "External id": 115265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115265,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 115265, "pid": 5, "tid": 7, "ts": 1716454223227023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223175948, "dur": 12,
    "args": {
      "External id": 115265,
      "cbid": 211, "correlation": 115265
    }
  },
  {
    "ph": "s", "id": 115265, "pid": 76337, "tid": -914061504, "ts": 1716454223175948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223227565, "dur": 125,
    "args": {
      "External id": 115273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115273,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115273, "pid": 5, "tid": 7, "ts": 1716454223227565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176019, "dur": 14,
    "args": {
      "External id": 115273,
      "cbid": 211, "correlation": 115273
    }
  },
  {
    "ph": "s", "id": 115273, "pid": 76337, "tid": -914061504, "ts": 1716454223176019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223227691, "dur": 127,
    "args": {
      "External id": 115281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115281, "pid": 5, "tid": 7, "ts": 1716454223227691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176051, "dur": 8,
    "args": {
      "External id": 115281,
      "cbid": 211, "correlation": 115281
    }
  },
  {
    "ph": "s", "id": 115281, "pid": 76337, "tid": -914061504, "ts": 1716454223176051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176130, "dur": 1,
    "args": {
      "External id": 115297,
      "cbid": 251, "correlation": 115297
    }
  },
  {
    "ph": "f", "id": 115297, "pid": 76337, "tid": -914061504, "ts": 1716454223176130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223227819, "dur": 302,
    "args": {
      "External id": 115299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115299,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115299, "pid": 5, "tid": 7, "ts": 1716454223227819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176136, "dur": 12,
    "args": {
      "External id": 115299,
      "cbid": 211, "correlation": 115299
    }
  },
  {
    "ph": "s", "id": 115299, "pid": 76337, "tid": -914061504, "ts": 1716454223176136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223228123, "dur": 27,
    "args": {
      "External id": 115307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115307,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115307, "pid": 5, "tid": 7, "ts": 1716454223228123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176177, "dur": 10,
    "args": {
      "External id": 115307,
      "cbid": 211, "correlation": 115307
    }
  },
  {
    "ph": "s", "id": 115307, "pid": 76337, "tid": -914061504, "ts": 1716454223176177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223228151, "dur": 81,
    "args": {
      "External id": 115318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115318, "pid": 5, "tid": 7, "ts": 1716454223228151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176244, "dur": 12,
    "args": {
      "External id": 115318,
      "cbid": 211, "correlation": 115318
    }
  },
  {
    "ph": "s", "id": 115318, "pid": 76337, "tid": -914061504, "ts": 1716454223176244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176307, "dur": 0,
    "args": {
      "External id": 115330,
      "cbid": 317, "correlation": 115330
    }
  },
  {
    "ph": "f", "id": 115330, "pid": 76337, "tid": -914061504, "ts": 1716454223176307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176308, "dur": 0,
    "args": {
      "External id": 115331,
      "cbid": 203, "correlation": 115331
    }
  },
  {
    "ph": "f", "id": 115331, "pid": 76337, "tid": -914061504, "ts": 1716454223176308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176309, "dur": 0,
    "args": {
      "External id": 115332,
      "cbid": 205, "correlation": 115332
    }
  },
  {
    "ph": "f", "id": 115332, "pid": 76337, "tid": -914061504, "ts": 1716454223176309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228233, "dur": 24,
    "args": {
      "External id": 115336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115336, "pid": 5, "tid": 7, "ts": 1716454223228233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176324, "dur": 13,
    "args": {
      "External id": 115336,
      "cbid": 211, "correlation": 115336
    }
  },
  {
    "ph": "s", "id": 115336, "pid": 76337, "tid": -914061504, "ts": 1716454223176324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223228258, "dur": 122,
    "args": {
      "External id": 115338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115338,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115338, "pid": 5, "tid": 7, "ts": 1716454223228258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176343, "dur": 6,
    "args": {
      "External id": 115338,
      "cbid": 211, "correlation": 115338
    }
  },
  {
    "ph": "s", "id": 115338, "pid": 76337, "tid": -914061504, "ts": 1716454223176343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228382, "dur": 22,
    "args": {
      "External id": 115340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115340,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115340, "pid": 5, "tid": 7, "ts": 1716454223228382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176353, "dur": 5,
    "args": {
      "External id": 115340,
      "cbid": 211, "correlation": 115340
    }
  },
  {
    "ph": "s", "id": 115340, "pid": 76337, "tid": -914061504, "ts": 1716454223176353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223228405, "dur": 33,
    "args": {
      "External id": 115346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115346, "pid": 5, "tid": 7, "ts": 1716454223228405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176380, "dur": 8,
    "args": {
      "External id": 115346,
      "cbid": 211, "correlation": 115346
    }
  },
  {
    "ph": "s", "id": 115346, "pid": 76337, "tid": -914061504, "ts": 1716454223176380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223228439, "dur": 27,
    "args": {
      "External id": 115354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115354,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115354, "pid": 5, "tid": 7, "ts": 1716454223228439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176412, "dur": 8,
    "args": {
      "External id": 115354,
      "cbid": 211, "correlation": 115354
    }
  },
  {
    "ph": "s", "id": 115354, "pid": 76337, "tid": -914061504, "ts": 1716454223176412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176483, "dur": 0,
    "args": {
      "External id": 115364,
      "cbid": 317, "correlation": 115364
    }
  },
  {
    "ph": "f", "id": 115364, "pid": 76337, "tid": -914061504, "ts": 1716454223176483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176484, "dur": 0,
    "args": {
      "External id": 115365,
      "cbid": 203, "correlation": 115365
    }
  },
  {
    "ph": "f", "id": 115365, "pid": 76337, "tid": -914061504, "ts": 1716454223176484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176485, "dur": 0,
    "args": {
      "External id": 115366,
      "cbid": 205, "correlation": 115366
    }
  },
  {
    "ph": "f", "id": 115366, "pid": 76337, "tid": -914061504, "ts": 1716454223176485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228467, "dur": 24,
    "args": {
      "External id": 115370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115370,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115370, "pid": 5, "tid": 7, "ts": 1716454223228467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176499, "dur": 12,
    "args": {
      "External id": 115370,
      "cbid": 211, "correlation": 115370
    }
  },
  {
    "ph": "s", "id": 115370, "pid": 76337, "tid": -914061504, "ts": 1716454223176499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228492, "dur": 44,
    "args": {
      "External id": 115372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115372,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115372, "pid": 5, "tid": 7, "ts": 1716454223228492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176514, "dur": 5,
    "args": {
      "External id": 115372,
      "cbid": 211, "correlation": 115372
    }
  },
  {
    "ph": "s", "id": 115372, "pid": 76337, "tid": -914061504, "ts": 1716454223176514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223228537, "dur": 234,
    "args": {
      "External id": 115374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115374,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 115374, "pid": 5, "tid": 7, "ts": 1716454223228537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176526, "dur": 7,
    "args": {
      "External id": 115374,
      "cbid": 211, "correlation": 115374
    }
  },
  {
    "ph": "s", "id": 115374, "pid": 76337, "tid": -914061504, "ts": 1716454223176526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228772, "dur": 6,
    "args": {
      "External id": 115376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115376,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115376, "pid": 5, "tid": 7, "ts": 1716454223228772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176536, "dur": 5,
    "args": {
      "External id": 115376,
      "cbid": 211, "correlation": 115376
    }
  },
  {
    "ph": "s", "id": 115376, "pid": 76337, "tid": -914061504, "ts": 1716454223176536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223228780, "dur": 9,
    "args": {
      "External id": 115382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115382, "pid": 5, "tid": 7, "ts": 1716454223228780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176562, "dur": 8,
    "args": {
      "External id": 115382,
      "cbid": 211, "correlation": 115382
    }
  },
  {
    "ph": "s", "id": 115382, "pid": 76337, "tid": -914061504, "ts": 1716454223176562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228790, "dur": 12,
    "args": {
      "External id": 115402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115402,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 115402, "pid": 5, "tid": 7, "ts": 1716454223228790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176653, "dur": 12,
    "args": {
      "External id": 115402,
      "cbid": 211, "correlation": 115402
    }
  },
  {
    "ph": "s", "id": 115402, "pid": 76337, "tid": -914061504, "ts": 1716454223176653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223228803, "dur": 4,
    "args": {
      "External id": 115414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115414,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 115414, "pid": 5, "tid": 7, "ts": 1716454223228803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176675, "dur": 7,
    "args": {
      "External id": 115414,
      "cbid": 211, "correlation": 115414
    }
  },
  {
    "ph": "s", "id": 115414, "pid": 76337, "tid": -914061504, "ts": 1716454223176675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223228808, "dur": 12,
    "args": {
      "External id": 115417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115417,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115417, "pid": 5, "tid": 7, "ts": 1716454223228808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176693, "dur": 6,
    "args": {
      "External id": 115417,
      "cbid": 211, "correlation": 115417
    }
  },
  {
    "ph": "s", "id": 115417, "pid": 76337, "tid": -914061504, "ts": 1716454223176693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223228822, "dur": 7,
    "args": {
      "External id": 115426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115426,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115426, "pid": 5, "tid": 7, "ts": 1716454223228822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176732, "dur": 10,
    "args": {
      "External id": 115426,
      "cbid": 211, "correlation": 115426
    }
  },
  {
    "ph": "s", "id": 115426, "pid": 76337, "tid": -914061504, "ts": 1716454223176732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176784, "dur": 0,
    "args": {
      "External id": 115436,
      "cbid": 317, "correlation": 115436
    }
  },
  {
    "ph": "f", "id": 115436, "pid": 76337, "tid": -914061504, "ts": 1716454223176784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176785, "dur": 0,
    "args": {
      "External id": 115437,
      "cbid": 203, "correlation": 115437
    }
  },
  {
    "ph": "f", "id": 115437, "pid": 76337, "tid": -914061504, "ts": 1716454223176785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176786, "dur": 0,
    "args": {
      "External id": 115438,
      "cbid": 205, "correlation": 115438
    }
  },
  {
    "ph": "f", "id": 115438, "pid": 76337, "tid": -914061504, "ts": 1716454223176786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228830, "dur": 6,
    "args": {
      "External id": 115442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115442,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115442, "pid": 5, "tid": 7, "ts": 1716454223228830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176801, "dur": 11,
    "args": {
      "External id": 115442,
      "cbid": 211, "correlation": 115442
    }
  },
  {
    "ph": "s", "id": 115442, "pid": 76337, "tid": -914061504, "ts": 1716454223176801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223228836, "dur": 83,
    "args": {
      "External id": 115444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115444, "pid": 5, "tid": 7, "ts": 1716454223228836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176815, "dur": 5,
    "args": {
      "External id": 115444,
      "cbid": 211, "correlation": 115444
    }
  },
  {
    "ph": "s", "id": 115444, "pid": 76337, "tid": -914061504, "ts": 1716454223176815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223228922, "dur": 1,
    "args": {
      "External id": 115446,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 115446,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 115446, "pid": 5, "tid": 7, "ts": 1716454223228922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176828, "dur": 8,
    "args": {
      "External id": 115446,
      "cbid": 51, "correlation": 115446
    }
  },
  {
    "ph": "s", "id": 115446, "pid": 76337, "tid": -914061504, "ts": 1716454223176828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223228926, "dur": 539,
    "args": {
      "External id": 115447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115447,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115447, "pid": 5, "tid": 7, "ts": 1716454223228926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176839, "dur": 8,
    "args": {
      "External id": 115447,
      "cbid": 211, "correlation": 115447
    }
  },
  {
    "ph": "s", "id": 115447, "pid": 76337, "tid": -914061504, "ts": 1716454223176839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223229466, "dur": 12,
    "args": {
      "External id": 115449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115449, "pid": 5, "tid": 7, "ts": 1716454223229466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176851, "dur": 5,
    "args": {
      "External id": 115449,
      "cbid": 211, "correlation": 115449
    }
  },
  {
    "ph": "s", "id": 115449, "pid": 76337, "tid": -914061504, "ts": 1716454223176851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223229479, "dur": 14,
    "args": {
      "External id": 115455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115455, "pid": 5, "tid": 7, "ts": 1716454223229479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176879, "dur": 9,
    "args": {
      "External id": 115455,
      "cbid": 211, "correlation": 115455
    }
  },
  {
    "ph": "s", "id": 115455, "pid": 76337, "tid": -914061504, "ts": 1716454223176879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223229495, "dur": 4,
    "args": {
      "External id": 115463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115463,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 115463, "pid": 5, "tid": 7, "ts": 1716454223229495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176923, "dur": 10,
    "args": {
      "External id": 115463,
      "cbid": 211, "correlation": 115463
    }
  },
  {
    "ph": "s", "id": 115463, "pid": 76337, "tid": -914061504, "ts": 1716454223176923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223176995, "dur": 1,
    "args": {
      "External id": 115479,
      "cbid": 251, "correlation": 115479
    }
  },
  {
    "ph": "f", "id": 115479, "pid": 76337, "tid": -914061504, "ts": 1716454223176995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177000, "dur": 0,
    "args": {
      "External id": 115481,
      "cbid": 251, "correlation": 115481
    }
  },
  {
    "ph": "f", "id": 115481, "pid": 76337, "tid": -914061504, "ts": 1716454223177000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223229500, "dur": 12,
    "args": {
      "External id": 115482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115482,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115482, "pid": 5, "tid": 7, "ts": 1716454223229500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177002, "dur": 11,
    "args": {
      "External id": 115482,
      "cbid": 211, "correlation": 115482
    }
  },
  {
    "ph": "s", "id": 115482, "pid": 76337, "tid": -914061504, "ts": 1716454223177002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223229514, "dur": 5,
    "args": {
      "External id": 115484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115484,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115484, "pid": 5, "tid": 7, "ts": 1716454223229514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177016, "dur": 5,
    "args": {
      "External id": 115484,
      "cbid": 211, "correlation": 115484
    }
  },
  {
    "ph": "s", "id": 115484, "pid": 76337, "tid": -914061504, "ts": 1716454223177016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223229521, "dur": 17,
    "args": {
      "External id": 115494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115494,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115494, "pid": 5, "tid": 7, "ts": 1716454223229521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177074, "dur": 13,
    "args": {
      "External id": 115494,
      "cbid": 211, "correlation": 115494
    }
  },
  {
    "ph": "s", "id": 115494, "pid": 76337, "tid": -914061504, "ts": 1716454223177074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223229539, "dur": 19,
    "args": {
      "External id": 115514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115514,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 115514, "pid": 5, "tid": 7, "ts": 1716454223229539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177142, "dur": 11,
    "args": {
      "External id": 115514,
      "cbid": 211, "correlation": 115514
    }
  },
  {
    "ph": "s", "id": 115514, "pid": 76337, "tid": -914061504, "ts": 1716454223177142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223229560, "dur": 5,
    "args": {
      "External id": 115526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115526,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 115526, "pid": 5, "tid": 7, "ts": 1716454223229560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177163, "dur": 6,
    "args": {
      "External id": 115526,
      "cbid": 211, "correlation": 115526
    }
  },
  {
    "ph": "s", "id": 115526, "pid": 76337, "tid": -914061504, "ts": 1716454223177163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223229566, "dur": 17,
    "args": {
      "External id": 115529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115529,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115529, "pid": 5, "tid": 7, "ts": 1716454223229566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177181, "dur": 7,
    "args": {
      "External id": 115529,
      "cbid": 211, "correlation": 115529
    }
  },
  {
    "ph": "s", "id": 115529, "pid": 76337, "tid": -914061504, "ts": 1716454223177181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223229584, "dur": 11,
    "args": {
      "External id": 115538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115538,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115538, "pid": 5, "tid": 7, "ts": 1716454223229584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177222, "dur": 10,
    "args": {
      "External id": 115538,
      "cbid": 211, "correlation": 115538
    }
  },
  {
    "ph": "s", "id": 115538, "pid": 76337, "tid": -914061504, "ts": 1716454223177222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177284, "dur": 0,
    "args": {
      "External id": 115548,
      "cbid": 317, "correlation": 115548
    }
  },
  {
    "ph": "f", "id": 115548, "pid": 76337, "tid": -914061504, "ts": 1716454223177284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177285, "dur": 0,
    "args": {
      "External id": 115549,
      "cbid": 203, "correlation": 115549
    }
  },
  {
    "ph": "f", "id": 115549, "pid": 76337, "tid": -914061504, "ts": 1716454223177285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177286, "dur": 0,
    "args": {
      "External id": 115550,
      "cbid": 205, "correlation": 115550
    }
  },
  {
    "ph": "f", "id": 115550, "pid": 76337, "tid": -914061504, "ts": 1716454223177286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223229597, "dur": 11,
    "args": {
      "External id": 115554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115554,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115554, "pid": 5, "tid": 7, "ts": 1716454223229597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177303, "dur": 12,
    "args": {
      "External id": 115554,
      "cbid": 211, "correlation": 115554
    }
  },
  {
    "ph": "s", "id": 115554, "pid": 76337, "tid": -914061504, "ts": 1716454223177303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223229609, "dur": 163,
    "args": {
      "External id": 115556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115556,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115556, "pid": 5, "tid": 7, "ts": 1716454223229609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177318, "dur": 5,
    "args": {
      "External id": 115556,
      "cbid": 211, "correlation": 115556
    }
  },
  {
    "ph": "s", "id": 115556, "pid": 76337, "tid": -914061504, "ts": 1716454223177318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223229773, "dur": 1,
    "args": {
      "External id": 115558,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 115558,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 115558, "pid": 5, "tid": 7, "ts": 1716454223229773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177329, "dur": 7,
    "args": {
      "External id": 115558,
      "cbid": 51, "correlation": 115558
    }
  },
  {
    "ph": "s", "id": 115558, "pid": 76337, "tid": -914061504, "ts": 1716454223177329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223229777, "dur": 662,
    "args": {
      "External id": 115559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115559,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115559, "pid": 5, "tid": 7, "ts": 1716454223229777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177338, "dur": 6,
    "args": {
      "External id": 115559,
      "cbid": 211, "correlation": 115559
    }
  },
  {
    "ph": "s", "id": 115559, "pid": 76337, "tid": -914061504, "ts": 1716454223177338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223230441, "dur": 13,
    "args": {
      "External id": 115561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115561,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115561, "pid": 5, "tid": 7, "ts": 1716454223230441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177348, "dur": 5,
    "args": {
      "External id": 115561,
      "cbid": 211, "correlation": 115561
    }
  },
  {
    "ph": "s", "id": 115561, "pid": 76337, "tid": -914061504, "ts": 1716454223177348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223230455, "dur": 15,
    "args": {
      "External id": 115567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115567,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115567, "pid": 5, "tid": 7, "ts": 1716454223230455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177377, "dur": 9,
    "args": {
      "External id": 115567,
      "cbid": 211, "correlation": 115567
    }
  },
  {
    "ph": "s", "id": 115567, "pid": 76337, "tid": -914061504, "ts": 1716454223177377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177435, "dur": 0,
    "args": {
      "External id": 115577,
      "cbid": 317, "correlation": 115577
    }
  },
  {
    "ph": "f", "id": 115577, "pid": 76337, "tid": -914061504, "ts": 1716454223177435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177436, "dur": 0,
    "args": {
      "External id": 115578,
      "cbid": 203, "correlation": 115578
    }
  },
  {
    "ph": "f", "id": 115578, "pid": 76337, "tid": -914061504, "ts": 1716454223177436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177437, "dur": 0,
    "args": {
      "External id": 115579,
      "cbid": 205, "correlation": 115579
    }
  },
  {
    "ph": "f", "id": 115579, "pid": 76337, "tid": -914061504, "ts": 1716454223177437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223230471, "dur": 8,
    "args": {
      "External id": 115583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115583,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115583, "pid": 5, "tid": 7, "ts": 1716454223230471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177451, "dur": 12,
    "args": {
      "External id": 115583,
      "cbid": 211, "correlation": 115583
    }
  },
  {
    "ph": "s", "id": 115583, "pid": 76337, "tid": -914061504, "ts": 1716454223177451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223230481, "dur": 3,
    "args": {
      "External id": 115585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115585,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 115585, "pid": 5, "tid": 7, "ts": 1716454223230481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177468, "dur": 6,
    "args": {
      "External id": 115585,
      "cbid": 211, "correlation": 115585
    }
  },
  {
    "ph": "s", "id": 115585, "pid": 76337, "tid": -914061504, "ts": 1716454223177468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177478, "dur": 0,
    "args": {
      "External id": 115586,
      "cbid": 51, "correlation": 115586
    }
  },
  {
    "ph": "s", "id": 115586, "pid": 76337, "tid": -914061504, "ts": 1716454223177478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223230486, "dur": 57,
    "args": {
      "External id": 115587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115587,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 115587, "pid": 5, "tid": 7, "ts": 1716454223230486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177479, "dur": 6,
    "args": {
      "External id": 115587,
      "cbid": 211, "correlation": 115587
    }
  },
  {
    "ph": "s", "id": 115587, "pid": 76337, "tid": -914061504, "ts": 1716454223177479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223230545, "dur": 14,
    "args": {
      "External id": 115592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115592, "pid": 5, "tid": 7, "ts": 1716454223230545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177504, "dur": 9,
    "args": {
      "External id": 115592,
      "cbid": 211, "correlation": 115592
    }
  },
  {
    "ph": "s", "id": 115592, "pid": 76337, "tid": -914061504, "ts": 1716454223177504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223230560, "dur": 11,
    "args": {
      "External id": 115600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115600,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115600, "pid": 5, "tid": 7, "ts": 1716454223230560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177535, "dur": 8,
    "args": {
      "External id": 115600,
      "cbid": 211, "correlation": 115600
    }
  },
  {
    "ph": "s", "id": 115600, "pid": 76337, "tid": -914061504, "ts": 1716454223177535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223230572, "dur": 10,
    "args": {
      "External id": 115608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115608,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115608, "pid": 5, "tid": 7, "ts": 1716454223230572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177564, "dur": 8,
    "args": {
      "External id": 115608,
      "cbid": 211, "correlation": 115608
    }
  },
  {
    "ph": "s", "id": 115608, "pid": 76337, "tid": -914061504, "ts": 1716454223177564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223230584, "dur": 18,
    "args": {
      "External id": 115628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115628,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 115628, "pid": 5, "tid": 7, "ts": 1716454223230584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177645, "dur": 12,
    "args": {
      "External id": 115628,
      "cbid": 211, "correlation": 115628
    }
  },
  {
    "ph": "s", "id": 115628, "pid": 76337, "tid": -914061504, "ts": 1716454223177645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223230603, "dur": 5,
    "args": {
      "External id": 115640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115640,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 115640, "pid": 5, "tid": 7, "ts": 1716454223230603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177667, "dur": 6,
    "args": {
      "External id": 115640,
      "cbid": 211, "correlation": 115640
    }
  },
  {
    "ph": "s", "id": 115640, "pid": 76337, "tid": -914061504, "ts": 1716454223177667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223230609, "dur": 17,
    "args": {
      "External id": 115643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115643, "pid": 5, "tid": 7, "ts": 1716454223230609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177686, "dur": 6,
    "args": {
      "External id": 115643,
      "cbid": 211, "correlation": 115643
    }
  },
  {
    "ph": "s", "id": 115643, "pid": 76337, "tid": -914061504, "ts": 1716454223177686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177743, "dur": 0,
    "args": {
      "External id": 115654,
      "cbid": 317, "correlation": 115654
    }
  },
  {
    "ph": "f", "id": 115654, "pid": 76337, "tid": -914061504, "ts": 1716454223177743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177744, "dur": 0,
    "args": {
      "External id": 115655,
      "cbid": 203, "correlation": 115655
    }
  },
  {
    "ph": "f", "id": 115655, "pid": 76337, "tid": -914061504, "ts": 1716454223177744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177745, "dur": 0,
    "args": {
      "External id": 115656,
      "cbid": 205, "correlation": 115656
    }
  },
  {
    "ph": "f", "id": 115656, "pid": 76337, "tid": -914061504, "ts": 1716454223177745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223230627, "dur": 12,
    "args": {
      "External id": 115660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115660, "pid": 5, "tid": 7, "ts": 1716454223230627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177760, "dur": 11,
    "args": {
      "External id": 115660,
      "cbid": 211, "correlation": 115660
    }
  },
  {
    "ph": "s", "id": 115660, "pid": 76337, "tid": -914061504, "ts": 1716454223177760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223230640, "dur": 4,
    "args": {
      "External id": 115662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 115662, "pid": 5, "tid": 7, "ts": 1716454223230640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177777, "dur": 7,
    "args": {
      "External id": 115662,
      "cbid": 211, "correlation": 115662
    }
  },
  {
    "ph": "s", "id": 115662, "pid": 76337, "tid": -914061504, "ts": 1716454223177777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177788, "dur": 0,
    "args": {
      "External id": 115663,
      "cbid": 51, "correlation": 115663
    }
  },
  {
    "ph": "s", "id": 115663, "pid": 76337, "tid": -914061504, "ts": 1716454223177788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223230645, "dur": 97,
    "args": {
      "External id": 115664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115664,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 115664, "pid": 5, "tid": 7, "ts": 1716454223230645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177789, "dur": 6,
    "args": {
      "External id": 115664,
      "cbid": 211, "correlation": 115664
    }
  },
  {
    "ph": "s", "id": 115664, "pid": 76337, "tid": -914061504, "ts": 1716454223177789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223230744, "dur": 15,
    "args": {
      "External id": 115669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115669, "pid": 5, "tid": 7, "ts": 1716454223230744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177816, "dur": 8,
    "args": {
      "External id": 115669,
      "cbid": 211, "correlation": 115669
    }
  },
  {
    "ph": "s", "id": 115669, "pid": 76337, "tid": -914061504, "ts": 1716454223177816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223230761, "dur": 84,
    "args": {
      "External id": 115678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115678,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115678, "pid": 5, "tid": 7, "ts": 1716454223230761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177898, "dur": 13,
    "args": {
      "External id": 115678,
      "cbid": 211, "correlation": 115678
    }
  },
  {
    "ph": "s", "id": 115678, "pid": 76337, "tid": -914061504, "ts": 1716454223177898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223230846, "dur": 31,
    "args": {
      "External id": 115700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115700,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115700, "pid": 5, "tid": 7, "ts": 1716454223230846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223177953, "dur": 10,
    "args": {
      "External id": 115700,
      "cbid": 211, "correlation": 115700
    }
  },
  {
    "ph": "s", "id": 115700, "pid": 76337, "tid": -914061504, "ts": 1716454223177953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178049, "dur": 1,
    "args": {
      "External id": 115711,
      "cbid": 251, "correlation": 115711
    }
  },
  {
    "ph": "f", "id": 115711, "pid": 76337, "tid": -914061504, "ts": 1716454223178049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223230879, "dur": 165,
    "args": {
      "External id": 115712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115712,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115712, "pid": 5, "tid": 7, "ts": 1716454223230879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178054, "dur": 13,
    "args": {
      "External id": 115712,
      "cbid": 211, "correlation": 115712
    }
  },
  {
    "ph": "s", "id": 115712, "pid": 76337, "tid": -914061504, "ts": 1716454223178054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178126, "dur": 1,
    "args": {
      "External id": 115723,
      "cbid": 251, "correlation": 115723
    }
  },
  {
    "ph": "f", "id": 115723, "pid": 76337, "tid": -914061504, "ts": 1716454223178126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223231045, "dur": 159,
    "args": {
      "External id": 115724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115724,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115724, "pid": 5, "tid": 7, "ts": 1716454223231045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178130, "dur": 12,
    "args": {
      "External id": 115724,
      "cbid": 211, "correlation": 115724
    }
  },
  {
    "ph": "s", "id": 115724, "pid": 76337, "tid": -914061504, "ts": 1716454223178130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178196, "dur": 1,
    "args": {
      "External id": 115735,
      "cbid": 251, "correlation": 115735
    }
  },
  {
    "ph": "f", "id": 115735, "pid": 76337, "tid": -914061504, "ts": 1716454223178196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223231206, "dur": 160,
    "args": {
      "External id": 115736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115736,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115736, "pid": 5, "tid": 7, "ts": 1716454223231206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178200, "dur": 11,
    "args": {
      "External id": 115736,
      "cbid": 211, "correlation": 115736
    }
  },
  {
    "ph": "s", "id": 115736, "pid": 76337, "tid": -914061504, "ts": 1716454223178200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223231367, "dur": 337,
    "args": {
      "External id": 115761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115761,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115761, "pid": 5, "tid": 7, "ts": 1716454223231367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178288, "dur": 13,
    "args": {
      "External id": 115761,
      "cbid": 211, "correlation": 115761
    }
  },
  {
    "ph": "s", "id": 115761, "pid": 76337, "tid": -914061504, "ts": 1716454223178288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178389, "dur": 1,
    "args": {
      "External id": 115779,
      "cbid": 251, "correlation": 115779
    }
  },
  {
    "ph": "f", "id": 115779, "pid": 76337, "tid": -914061504, "ts": 1716454223178389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223231705, "dur": 166,
    "args": {
      "External id": 115781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115781,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115781, "pid": 5, "tid": 7, "ts": 1716454223231705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178395, "dur": 13,
    "args": {
      "External id": 115781,
      "cbid": 211, "correlation": 115781
    }
  },
  {
    "ph": "s", "id": 115781, "pid": 76337, "tid": -914061504, "ts": 1716454223178395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223231872, "dur": 19,
    "args": {
      "External id": 115789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115789,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115789, "pid": 5, "tid": 7, "ts": 1716454223231872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178465, "dur": 12,
    "args": {
      "External id": 115789,
      "cbid": 211, "correlation": 115789
    }
  },
  {
    "ph": "s", "id": 115789, "pid": 76337, "tid": -914061504, "ts": 1716454223178465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223231892, "dur": 27,
    "args": {
      "External id": 115797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115797,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115797, "pid": 5, "tid": 7, "ts": 1716454223231892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178504, "dur": 8,
    "args": {
      "External id": 115797,
      "cbid": 211, "correlation": 115797
    }
  },
  {
    "ph": "s", "id": 115797, "pid": 76337, "tid": -914061504, "ts": 1716454223178504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223231921, "dur": 17,
    "args": {
      "External id": 115808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115808, "pid": 5, "tid": 7, "ts": 1716454223231921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178575, "dur": 12,
    "args": {
      "External id": 115808,
      "cbid": 211, "correlation": 115808
    }
  },
  {
    "ph": "s", "id": 115808, "pid": 76337, "tid": -914061504, "ts": 1716454223178575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223231939, "dur": 16,
    "args": {
      "External id": 115830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115830,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115830, "pid": 5, "tid": 7, "ts": 1716454223231939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178606, "dur": 8,
    "args": {
      "External id": 115830,
      "cbid": 211, "correlation": 115830
    }
  },
  {
    "ph": "s", "id": 115830, "pid": 76337, "tid": -914061504, "ts": 1716454223178606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178691, "dur": 2,
    "args": {
      "External id": 115841,
      "cbid": 251, "correlation": 115841
    }
  },
  {
    "ph": "f", "id": 115841, "pid": 76337, "tid": -914061504, "ts": 1716454223178691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223231957, "dur": 90,
    "args": {
      "External id": 115842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115842,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 115842, "pid": 5, "tid": 7, "ts": 1716454223231957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178699, "dur": 14,
    "args": {
      "External id": 115842,
      "cbid": 211, "correlation": 115842
    }
  },
  {
    "ph": "s", "id": 115842, "pid": 76337, "tid": -914061504, "ts": 1716454223178699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178769, "dur": 1,
    "args": {
      "External id": 115853,
      "cbid": 251, "correlation": 115853
    }
  },
  {
    "ph": "f", "id": 115853, "pid": 76337, "tid": -914061504, "ts": 1716454223178769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178772, "dur": 0,
    "args": {
      "External id": 115854,
      "cbid": 251, "correlation": 115854
    }
  },
  {
    "ph": "f", "id": 115854, "pid": 76337, "tid": -914061504, "ts": 1716454223178772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223232048, "dur": 13,
    "args": {
      "External id": 115855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115855,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115855, "pid": 5, "tid": 7, "ts": 1716454223232048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178774, "dur": 12,
    "args": {
      "External id": 115855,
      "cbid": 211, "correlation": 115855
    }
  },
  {
    "ph": "s", "id": 115855, "pid": 76337, "tid": -914061504, "ts": 1716454223178774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223232062, "dur": 6,
    "args": {
      "External id": 115857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115857,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115857, "pid": 5, "tid": 7, "ts": 1716454223232062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178787, "dur": 6,
    "args": {
      "External id": 115857,
      "cbid": 211, "correlation": 115857
    }
  },
  {
    "ph": "s", "id": 115857, "pid": 76337, "tid": -914061504, "ts": 1716454223178787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178846, "dur": 1,
    "args": {
      "External id": 115868,
      "cbid": 251, "correlation": 115868
    }
  },
  {
    "ph": "f", "id": 115868, "pid": 76337, "tid": -914061504, "ts": 1716454223178846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178850, "dur": 0,
    "args": {
      "External id": 115869,
      "cbid": 251, "correlation": 115869
    }
  },
  {
    "ph": "f", "id": 115869, "pid": 76337, "tid": -914061504, "ts": 1716454223178850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223232069, "dur": 9,
    "args": {
      "External id": 115870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115870,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115870, "pid": 5, "tid": 7, "ts": 1716454223232069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178851, "dur": 11,
    "args": {
      "External id": 115870,
      "cbid": 211, "correlation": 115870
    }
  },
  {
    "ph": "s", "id": 115870, "pid": 76337, "tid": -914061504, "ts": 1716454223178851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223232079, "dur": 3,
    "args": {
      "External id": 115872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115872,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115872, "pid": 5, "tid": 7, "ts": 1716454223232079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178864, "dur": 5,
    "args": {
      "External id": 115872,
      "cbid": 211, "correlation": 115872
    }
  },
  {
    "ph": "s", "id": 115872, "pid": 76337, "tid": -914061504, "ts": 1716454223178864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223232083, "dur": 56,
    "args": {
      "External id": 115897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115897,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115897, "pid": 5, "tid": 7, "ts": 1716454223232083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223178940, "dur": 13,
    "args": {
      "External id": 115897,
      "cbid": 211, "correlation": 115897
    }
  },
  {
    "ph": "s", "id": 115897, "pid": 76337, "tid": -914061504, "ts": 1716454223178940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179048, "dur": 2,
    "args": {
      "External id": 115915,
      "cbid": 251, "correlation": 115915
    }
  },
  {
    "ph": "f", "id": 115915, "pid": 76337, "tid": -914061504, "ts": 1716454223179048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223232140, "dur": 92,
    "args": {
      "External id": 115917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115917,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 115917, "pid": 5, "tid": 7, "ts": 1716454223232140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179055, "dur": 14,
    "args": {
      "External id": 115917,
      "cbid": 211, "correlation": 115917
    }
  },
  {
    "ph": "s", "id": 115917, "pid": 76337, "tid": -914061504, "ts": 1716454223179055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223232234, "dur": 10,
    "args": {
      "External id": 115925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115925,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115925, "pid": 5, "tid": 7, "ts": 1716454223232234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179126, "dur": 12,
    "args": {
      "External id": 115925,
      "cbid": 211, "correlation": 115925
    }
  },
  {
    "ph": "s", "id": 115925, "pid": 76337, "tid": -914061504, "ts": 1716454223179126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223232244, "dur": 21,
    "args": {
      "External id": 115933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115933,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115933, "pid": 5, "tid": 7, "ts": 1716454223232244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179168, "dur": 9,
    "args": {
      "External id": 115933,
      "cbid": 211, "correlation": 115933
    }
  },
  {
    "ph": "s", "id": 115933, "pid": 76337, "tid": -914061504, "ts": 1716454223179168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223232266, "dur": 17,
    "args": {
      "External id": 115955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115955,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115955, "pid": 5, "tid": 7, "ts": 1716454223232266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179220, "dur": 11,
    "args": {
      "External id": 115955,
      "cbid": 211, "correlation": 115955
    }
  },
  {
    "ph": "s", "id": 115955, "pid": 76337, "tid": -914061504, "ts": 1716454223179220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179312, "dur": 2,
    "args": {
      "External id": 115971,
      "cbid": 251, "correlation": 115971
    }
  },
  {
    "ph": "f", "id": 115971, "pid": 76337, "tid": -914061504, "ts": 1716454223179312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179318, "dur": 0,
    "args": {
      "External id": 115973,
      "cbid": 251, "correlation": 115973
    }
  },
  {
    "ph": "f", "id": 115973, "pid": 76337, "tid": -914061504, "ts": 1716454223179318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223232285, "dur": 495,
    "args": {
      "External id": 115974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115974,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 115974, "pid": 5, "tid": 7, "ts": 1716454223232285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179320, "dur": 14,
    "args": {
      "External id": 115974,
      "cbid": 211, "correlation": 115974
    }
  },
  {
    "ph": "s", "id": 115974, "pid": 76337, "tid": -914061504, "ts": 1716454223179320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223232782, "dur": 65,
    "args": {
      "External id": 115982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115982,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115982, "pid": 5, "tid": 7, "ts": 1716454223232782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179387, "dur": 13,
    "args": {
      "External id": 115982,
      "cbid": 211, "correlation": 115982
    }
  },
  {
    "ph": "s", "id": 115982, "pid": 76337, "tid": -914061504, "ts": 1716454223179387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223232848, "dur": 69,
    "args": {
      "External id": 115990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 115990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 115990, "pid": 5, "tid": 7, "ts": 1716454223232848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179418, "dur": 8,
    "args": {
      "External id": 115990,
      "cbid": 211, "correlation": 115990
    }
  },
  {
    "ph": "s", "id": 115990, "pid": 76337, "tid": -914061504, "ts": 1716454223179418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179499, "dur": 1,
    "args": {
      "External id": 116006,
      "cbid": 251, "correlation": 116006
    }
  },
  {
    "ph": "f", "id": 116006, "pid": 76337, "tid": -914061504, "ts": 1716454223179499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223232919, "dur": 1,
    "args": {
      "External id": 116008,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116008,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1563517915309446
    }
  },
  {
    "ph": "f", "id": 116008, "pid": 5, "tid": 7, "ts": 1716454223232919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179505, "dur": 11,
    "args": {
      "External id": 116008,
      "cbid": 51, "correlation": 116008
    }
  },
  {
    "ph": "s", "id": 116008, "pid": 76337, "tid": -914061504, "ts": 1716454223179505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223232923, "dur": 268,
    "args": {
      "External id": 116009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116009,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116009, "pid": 5, "tid": 7, "ts": 1716454223232923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179518, "dur": 11,
    "args": {
      "External id": 116009,
      "cbid": 211, "correlation": 116009
    }
  },
  {
    "ph": "s", "id": 116009, "pid": 76337, "tid": -914061504, "ts": 1716454223179518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223233192, "dur": 14,
    "args": {
      "External id": 116017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116017,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116017, "pid": 5, "tid": 7, "ts": 1716454223233192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179560, "dur": 10,
    "args": {
      "External id": 116017,
      "cbid": 211, "correlation": 116017
    }
  },
  {
    "ph": "s", "id": 116017, "pid": 76337, "tid": -914061504, "ts": 1716454223179560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223233208, "dur": 38,
    "args": {
      "External id": 116028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116028, "pid": 5, "tid": 7, "ts": 1716454223233208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179627, "dur": 12,
    "args": {
      "External id": 116028,
      "cbid": 211, "correlation": 116028
    }
  },
  {
    "ph": "s", "id": 116028, "pid": 76337, "tid": -914061504, "ts": 1716454223179627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179692, "dur": 0,
    "args": {
      "External id": 116040,
      "cbid": 317, "correlation": 116040
    }
  },
  {
    "ph": "f", "id": 116040, "pid": 76337, "tid": -914061504, "ts": 1716454223179692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179692, "dur": 0,
    "args": {
      "External id": 116041,
      "cbid": 203, "correlation": 116041
    }
  },
  {
    "ph": "f", "id": 116041, "pid": 76337, "tid": -914061504, "ts": 1716454223179692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179693, "dur": 0,
    "args": {
      "External id": 116042,
      "cbid": 205, "correlation": 116042
    }
  },
  {
    "ph": "f", "id": 116042, "pid": 76337, "tid": -914061504, "ts": 1716454223179693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223233247, "dur": 13,
    "args": {
      "External id": 116046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116046,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116046, "pid": 5, "tid": 7, "ts": 1716454223233247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179708, "dur": 12,
    "args": {
      "External id": 116046,
      "cbid": 211, "correlation": 116046
    }
  },
  {
    "ph": "s", "id": 116046, "pid": 76337, "tid": -914061504, "ts": 1716454223179708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223233261, "dur": 4,
    "args": {
      "External id": 116048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116048,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116048, "pid": 5, "tid": 7, "ts": 1716454223233261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179725, "dur": 6,
    "args": {
      "External id": 116048,
      "cbid": 211, "correlation": 116048
    }
  },
  {
    "ph": "s", "id": 116048, "pid": 76337, "tid": -914061504, "ts": 1716454223179725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179734, "dur": 0,
    "args": {
      "External id": 116049,
      "cbid": 51, "correlation": 116049
    }
  },
  {
    "ph": "s", "id": 116049, "pid": 76337, "tid": -914061504, "ts": 1716454223179734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223233266, "dur": 98,
    "args": {
      "External id": 116050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116050,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 116050, "pid": 5, "tid": 7, "ts": 1716454223233266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179734, "dur": 5,
    "args": {
      "External id": 116050,
      "cbid": 211, "correlation": 116050
    }
  },
  {
    "ph": "s", "id": 116050, "pid": 76337, "tid": -914061504, "ts": 1716454223179734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223233365, "dur": 16,
    "args": {
      "External id": 116055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116055, "pid": 5, "tid": 7, "ts": 1716454223233365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179762, "dur": 8,
    "args": {
      "External id": 116055,
      "cbid": 211, "correlation": 116055
    }
  },
  {
    "ph": "s", "id": 116055, "pid": 76337, "tid": -914061504, "ts": 1716454223179762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223233383, "dur": 13,
    "args": {
      "External id": 116063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116063,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116063, "pid": 5, "tid": 7, "ts": 1716454223233383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179793, "dur": 8,
    "args": {
      "External id": 116063,
      "cbid": 211, "correlation": 116063
    }
  },
  {
    "ph": "s", "id": 116063, "pid": 76337, "tid": -914061504, "ts": 1716454223179793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223233398, "dur": 19,
    "args": {
      "External id": 116083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116083,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 116083, "pid": 5, "tid": 7, "ts": 1716454223233398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179866, "dur": 11,
    "args": {
      "External id": 116083,
      "cbid": 211, "correlation": 116083
    }
  },
  {
    "ph": "s", "id": 116083, "pid": 76337, "tid": -914061504, "ts": 1716454223179866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223233418, "dur": 5,
    "args": {
      "External id": 116095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116095,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 116095, "pid": 5, "tid": 7, "ts": 1716454223233418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179888, "dur": 6,
    "args": {
      "External id": 116095,
      "cbid": 211, "correlation": 116095
    }
  },
  {
    "ph": "s", "id": 116095, "pid": 76337, "tid": -914061504, "ts": 1716454223179888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223233424, "dur": 18,
    "args": {
      "External id": 116098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116098, "pid": 5, "tid": 7, "ts": 1716454223233424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179906, "dur": 7,
    "args": {
      "External id": 116098,
      "cbid": 211, "correlation": 116098
    }
  },
  {
    "ph": "s", "id": 116098, "pid": 76337, "tid": -914061504, "ts": 1716454223179906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223233444, "dur": 12,
    "args": {
      "External id": 116107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116107,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116107, "pid": 5, "tid": 7, "ts": 1716454223233444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223179945, "dur": 10,
    "args": {
      "External id": 116107,
      "cbid": 211, "correlation": 116107
    }
  },
  {
    "ph": "s", "id": 116107, "pid": 76337, "tid": -914061504, "ts": 1716454223179945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180004, "dur": 0,
    "args": {
      "External id": 116117,
      "cbid": 317, "correlation": 116117
    }
  },
  {
    "ph": "f", "id": 116117, "pid": 76337, "tid": -914061504, "ts": 1716454223180004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180005, "dur": 0,
    "args": {
      "External id": 116118,
      "cbid": 203, "correlation": 116118
    }
  },
  {
    "ph": "f", "id": 116118, "pid": 76337, "tid": -914061504, "ts": 1716454223180005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180005, "dur": 0,
    "args": {
      "External id": 116119,
      "cbid": 205, "correlation": 116119
    }
  },
  {
    "ph": "f", "id": 116119, "pid": 76337, "tid": -914061504, "ts": 1716454223180005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223233456, "dur": 11,
    "args": {
      "External id": 116123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116123, "pid": 5, "tid": 7, "ts": 1716454223233456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180020, "dur": 12,
    "args": {
      "External id": 116123,
      "cbid": 211, "correlation": 116123
    }
  },
  {
    "ph": "s", "id": 116123, "pid": 76337, "tid": -914061504, "ts": 1716454223180020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223233469, "dur": 162,
    "args": {
      "External id": 116125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116125,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116125, "pid": 5, "tid": 7, "ts": 1716454223233469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180034, "dur": 5,
    "args": {
      "External id": 116125,
      "cbid": 211, "correlation": 116125
    }
  },
  {
    "ph": "s", "id": 116125, "pid": 76337, "tid": -914061504, "ts": 1716454223180034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223233633, "dur": 1,
    "args": {
      "External id": 116127,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116127,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 116127, "pid": 5, "tid": 7, "ts": 1716454223233633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180045, "dur": 7,
    "args": {
      "External id": 116127,
      "cbid": 51, "correlation": 116127
    }
  },
  {
    "ph": "s", "id": 116127, "pid": 76337, "tid": -914061504, "ts": 1716454223180045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223233637, "dur": 661,
    "args": {
      "External id": 116128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116128,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116128, "pid": 5, "tid": 7, "ts": 1716454223233637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180053, "dur": 7,
    "args": {
      "External id": 116128,
      "cbid": 211, "correlation": 116128
    }
  },
  {
    "ph": "s", "id": 116128, "pid": 76337, "tid": -914061504, "ts": 1716454223180053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223234299, "dur": 14,
    "args": {
      "External id": 116130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116130, "pid": 5, "tid": 7, "ts": 1716454223234299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180065, "dur": 5,
    "args": {
      "External id": 116130,
      "cbid": 211, "correlation": 116130
    }
  },
  {
    "ph": "s", "id": 116130, "pid": 76337, "tid": -914061504, "ts": 1716454223180065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223234314, "dur": 14,
    "args": {
      "External id": 116136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116136, "pid": 5, "tid": 7, "ts": 1716454223234314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180093, "dur": 8,
    "args": {
      "External id": 116136,
      "cbid": 211, "correlation": 116136
    }
  },
  {
    "ph": "s", "id": 116136, "pid": 76337, "tid": -914061504, "ts": 1716454223180093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223234330, "dur": 4,
    "args": {
      "External id": 116144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116144,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 116144, "pid": 5, "tid": 7, "ts": 1716454223234330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180137, "dur": 9,
    "args": {
      "External id": 116144,
      "cbid": 211, "correlation": 116144
    }
  },
  {
    "ph": "s", "id": 116144, "pid": 76337, "tid": -914061504, "ts": 1716454223180137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180202, "dur": 1,
    "args": {
      "External id": 116160,
      "cbid": 251, "correlation": 116160
    }
  },
  {
    "ph": "f", "id": 116160, "pid": 76337, "tid": -914061504, "ts": 1716454223180202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180208, "dur": 0,
    "args": {
      "External id": 116162,
      "cbid": 251, "correlation": 116162
    }
  },
  {
    "ph": "f", "id": 116162, "pid": 76337, "tid": -914061504, "ts": 1716454223180208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223234335, "dur": 13,
    "args": {
      "External id": 116163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116163,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116163, "pid": 5, "tid": 7, "ts": 1716454223234335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180210, "dur": 11,
    "args": {
      "External id": 116163,
      "cbid": 211, "correlation": 116163
    }
  },
  {
    "ph": "s", "id": 116163, "pid": 76337, "tid": -914061504, "ts": 1716454223180210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223234350, "dur": 5,
    "args": {
      "External id": 116165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116165,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116165, "pid": 5, "tid": 7, "ts": 1716454223234350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180223, "dur": 5,
    "args": {
      "External id": 116165,
      "cbid": 211, "correlation": 116165
    }
  },
  {
    "ph": "s", "id": 116165, "pid": 76337, "tid": -914061504, "ts": 1716454223180223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223234356, "dur": 17,
    "args": {
      "External id": 116175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116175, "pid": 5, "tid": 7, "ts": 1716454223234356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180279, "dur": 12,
    "args": {
      "External id": 116175,
      "cbid": 211, "correlation": 116175
    }
  },
  {
    "ph": "s", "id": 116175, "pid": 76337, "tid": -914061504, "ts": 1716454223180279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223234375, "dur": 18,
    "args": {
      "External id": 116195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116195,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 116195, "pid": 5, "tid": 7, "ts": 1716454223234375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180345, "dur": 10,
    "args": {
      "External id": 116195,
      "cbid": 211, "correlation": 116195
    }
  },
  {
    "ph": "s", "id": 116195, "pid": 76337, "tid": -914061504, "ts": 1716454223180345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223234394, "dur": 4,
    "args": {
      "External id": 116207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116207,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 116207, "pid": 5, "tid": 7, "ts": 1716454223234394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180367, "dur": 6,
    "args": {
      "External id": 116207,
      "cbid": 211, "correlation": 116207
    }
  },
  {
    "ph": "s", "id": 116207, "pid": 76337, "tid": -914061504, "ts": 1716454223180367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223234400, "dur": 19,
    "args": {
      "External id": 116210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116210, "pid": 5, "tid": 7, "ts": 1716454223234400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180386, "dur": 6,
    "args": {
      "External id": 116210,
      "cbid": 211, "correlation": 116210
    }
  },
  {
    "ph": "s", "id": 116210, "pid": 76337, "tid": -914061504, "ts": 1716454223180386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223234420, "dur": 12,
    "args": {
      "External id": 116219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116219,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116219, "pid": 5, "tid": 7, "ts": 1716454223234420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180427, "dur": 9,
    "args": {
      "External id": 116219,
      "cbid": 211, "correlation": 116219
    }
  },
  {
    "ph": "s", "id": 116219, "pid": 76337, "tid": -914061504, "ts": 1716454223180427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180489, "dur": 0,
    "args": {
      "External id": 116229,
      "cbid": 317, "correlation": 116229
    }
  },
  {
    "ph": "f", "id": 116229, "pid": 76337, "tid": -914061504, "ts": 1716454223180489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180489, "dur": 0,
    "args": {
      "External id": 116230,
      "cbid": 203, "correlation": 116230
    }
  },
  {
    "ph": "f", "id": 116230, "pid": 76337, "tid": -914061504, "ts": 1716454223180489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180490, "dur": 0,
    "args": {
      "External id": 116231,
      "cbid": 205, "correlation": 116231
    }
  },
  {
    "ph": "f", "id": 116231, "pid": 76337, "tid": -914061504, "ts": 1716454223180490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223234433, "dur": 11,
    "args": {
      "External id": 116235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116235, "pid": 5, "tid": 7, "ts": 1716454223234433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180504, "dur": 12,
    "args": {
      "External id": 116235,
      "cbid": 211, "correlation": 116235
    }
  },
  {
    "ph": "s", "id": 116235, "pid": 76337, "tid": -914061504, "ts": 1716454223180504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223234445, "dur": 163,
    "args": {
      "External id": 116237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116237, "pid": 5, "tid": 7, "ts": 1716454223234445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180519, "dur": 5,
    "args": {
      "External id": 116237,
      "cbid": 211, "correlation": 116237
    }
  },
  {
    "ph": "s", "id": 116237, "pid": 76337, "tid": -914061504, "ts": 1716454223180519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223234610, "dur": 1,
    "args": {
      "External id": 116239,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116239,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 116239, "pid": 5, "tid": 7, "ts": 1716454223234610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180530, "dur": 6,
    "args": {
      "External id": 116239,
      "cbid": 51, "correlation": 116239
    }
  },
  {
    "ph": "s", "id": 116239, "pid": 76337, "tid": -914061504, "ts": 1716454223180530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223234614, "dur": 648,
    "args": {
      "External id": 116240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116240,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116240, "pid": 5, "tid": 7, "ts": 1716454223234614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180537, "dur": 6,
    "args": {
      "External id": 116240,
      "cbid": 211, "correlation": 116240
    }
  },
  {
    "ph": "s", "id": 116240, "pid": 76337, "tid": -914061504, "ts": 1716454223180537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223235263, "dur": 13,
    "args": {
      "External id": 116242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116242,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116242, "pid": 5, "tid": 7, "ts": 1716454223235263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180547, "dur": 5,
    "args": {
      "External id": 116242,
      "cbid": 211, "correlation": 116242
    }
  },
  {
    "ph": "s", "id": 116242, "pid": 76337, "tid": -914061504, "ts": 1716454223180547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223235278, "dur": 15,
    "args": {
      "External id": 116248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116248, "pid": 5, "tid": 7, "ts": 1716454223235278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180576, "dur": 9,
    "args": {
      "External id": 116248,
      "cbid": 211, "correlation": 116248
    }
  },
  {
    "ph": "s", "id": 116248, "pid": 76337, "tid": -914061504, "ts": 1716454223180576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223235294, "dur": 12,
    "args": {
      "External id": 116256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116256,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116256, "pid": 5, "tid": 7, "ts": 1716454223235294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180607, "dur": 9,
    "args": {
      "External id": 116256,
      "cbid": 211, "correlation": 116256
    }
  },
  {
    "ph": "s", "id": 116256, "pid": 76337, "tid": -914061504, "ts": 1716454223180607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223235307, "dur": 11,
    "args": {
      "External id": 116264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116264,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116264, "pid": 5, "tid": 7, "ts": 1716454223235307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180636, "dur": 8,
    "args": {
      "External id": 116264,
      "cbid": 211, "correlation": 116264
    }
  },
  {
    "ph": "s", "id": 116264, "pid": 76337, "tid": -914061504, "ts": 1716454223180636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223235319, "dur": 19,
    "args": {
      "External id": 116284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116284,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 116284, "pid": 5, "tid": 7, "ts": 1716454223235319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180715, "dur": 13,
    "args": {
      "External id": 116284,
      "cbid": 211, "correlation": 116284
    }
  },
  {
    "ph": "s", "id": 116284, "pid": 76337, "tid": -914061504, "ts": 1716454223180715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223235339, "dur": 4,
    "args": {
      "External id": 116296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116296,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 116296, "pid": 5, "tid": 7, "ts": 1716454223235339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180737, "dur": 6,
    "args": {
      "External id": 116296,
      "cbid": 211, "correlation": 116296
    }
  },
  {
    "ph": "s", "id": 116296, "pid": 76337, "tid": -914061504, "ts": 1716454223180737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223235345, "dur": 17,
    "args": {
      "External id": 116299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116299, "pid": 5, "tid": 7, "ts": 1716454223235345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180754, "dur": 7,
    "args": {
      "External id": 116299,
      "cbid": 211, "correlation": 116299
    }
  },
  {
    "ph": "s", "id": 116299, "pid": 76337, "tid": -914061504, "ts": 1716454223180754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180812, "dur": 0,
    "args": {
      "External id": 116310,
      "cbid": 317, "correlation": 116310
    }
  },
  {
    "ph": "f", "id": 116310, "pid": 76337, "tid": -914061504, "ts": 1716454223180812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180812, "dur": 0,
    "args": {
      "External id": 116311,
      "cbid": 203, "correlation": 116311
    }
  },
  {
    "ph": "f", "id": 116311, "pid": 76337, "tid": -914061504, "ts": 1716454223180812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180813, "dur": 0,
    "args": {
      "External id": 116312,
      "cbid": 205, "correlation": 116312
    }
  },
  {
    "ph": "f", "id": 116312, "pid": 76337, "tid": -914061504, "ts": 1716454223180813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223235363, "dur": 12,
    "args": {
      "External id": 116316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116316,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116316, "pid": 5, "tid": 7, "ts": 1716454223235363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180827, "dur": 12,
    "args": {
      "External id": 116316,
      "cbid": 211, "correlation": 116316
    }
  },
  {
    "ph": "s", "id": 116316, "pid": 76337, "tid": -914061504, "ts": 1716454223180827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223235376, "dur": 4,
    "args": {
      "External id": 116318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116318, "pid": 5, "tid": 7, "ts": 1716454223235376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180843, "dur": 6,
    "args": {
      "External id": 116318,
      "cbid": 211, "correlation": 116318
    }
  },
  {
    "ph": "s", "id": 116318, "pid": 76337, "tid": -914061504, "ts": 1716454223180843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180851, "dur": 0,
    "args": {
      "External id": 116319,
      "cbid": 51, "correlation": 116319
    }
  },
  {
    "ph": "s", "id": 116319, "pid": 76337, "tid": -914061504, "ts": 1716454223180851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223235381, "dur": 96,
    "args": {
      "External id": 116320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116320,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 116320, "pid": 5, "tid": 7, "ts": 1716454223235381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180852, "dur": 5,
    "args": {
      "External id": 116320,
      "cbid": 211, "correlation": 116320
    }
  },
  {
    "ph": "s", "id": 116320, "pid": 76337, "tid": -914061504, "ts": 1716454223180852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223235478, "dur": 15,
    "args": {
      "External id": 116325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116325, "pid": 5, "tid": 7, "ts": 1716454223235478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180879, "dur": 8,
    "args": {
      "External id": 116325,
      "cbid": 211, "correlation": 116325
    }
  },
  {
    "ph": "s", "id": 116325, "pid": 76337, "tid": -914061504, "ts": 1716454223180879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223235495, "dur": 83,
    "args": {
      "External id": 116334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116334,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116334, "pid": 5, "tid": 7, "ts": 1716454223235495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223180961, "dur": 23,
    "args": {
      "External id": 116334,
      "cbid": 211, "correlation": 116334
    }
  },
  {
    "ph": "s", "id": 116334, "pid": 76337, "tid": -914061504, "ts": 1716454223180961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223235579, "dur": 29,
    "args": {
      "External id": 116356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116356,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116356, "pid": 5, "tid": 7, "ts": 1716454223235579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181028, "dur": 11,
    "args": {
      "External id": 116356,
      "cbid": 211, "correlation": 116356
    }
  },
  {
    "ph": "s", "id": 116356, "pid": 76337, "tid": -914061504, "ts": 1716454223181028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181116, "dur": 1,
    "args": {
      "External id": 116367,
      "cbid": 251, "correlation": 116367
    }
  },
  {
    "ph": "f", "id": 116367, "pid": 76337, "tid": -914061504, "ts": 1716454223181116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223235610, "dur": 165,
    "args": {
      "External id": 116368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116368,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116368, "pid": 5, "tid": 7, "ts": 1716454223235610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181122, "dur": 13,
    "args": {
      "External id": 116368,
      "cbid": 211, "correlation": 116368
    }
  },
  {
    "ph": "s", "id": 116368, "pid": 76337, "tid": -914061504, "ts": 1716454223181122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181191, "dur": 1,
    "args": {
      "External id": 116379,
      "cbid": 251, "correlation": 116379
    }
  },
  {
    "ph": "f", "id": 116379, "pid": 76337, "tid": -914061504, "ts": 1716454223181191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223235777, "dur": 160,
    "args": {
      "External id": 116380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116380,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116380, "pid": 5, "tid": 7, "ts": 1716454223235777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181195, "dur": 11,
    "args": {
      "External id": 116380,
      "cbid": 211, "correlation": 116380
    }
  },
  {
    "ph": "s", "id": 116380, "pid": 76337, "tid": -914061504, "ts": 1716454223181195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181259, "dur": 1,
    "args": {
      "External id": 116391,
      "cbid": 251, "correlation": 116391
    }
  },
  {
    "ph": "f", "id": 116391, "pid": 76337, "tid": -914061504, "ts": 1716454223181259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223235938, "dur": 159,
    "args": {
      "External id": 116392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116392,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116392, "pid": 5, "tid": 7, "ts": 1716454223235938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181263, "dur": 12,
    "args": {
      "External id": 116392,
      "cbid": 211, "correlation": 116392
    }
  },
  {
    "ph": "s", "id": 116392, "pid": 76337, "tid": -914061504, "ts": 1716454223181263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223236098, "dur": 338,
    "args": {
      "External id": 116417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116417,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116417, "pid": 5, "tid": 7, "ts": 1716454223236098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181348, "dur": 13,
    "args": {
      "External id": 116417,
      "cbid": 211, "correlation": 116417
    }
  },
  {
    "ph": "s", "id": 116417, "pid": 76337, "tid": -914061504, "ts": 1716454223181348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181447, "dur": 1,
    "args": {
      "External id": 116435,
      "cbid": 251, "correlation": 116435
    }
  },
  {
    "ph": "f", "id": 116435, "pid": 76337, "tid": -914061504, "ts": 1716454223181447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223236438, "dur": 167,
    "args": {
      "External id": 116437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116437,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116437, "pid": 5, "tid": 7, "ts": 1716454223236438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181453, "dur": 13,
    "args": {
      "External id": 116437,
      "cbid": 211, "correlation": 116437
    }
  },
  {
    "ph": "s", "id": 116437, "pid": 76337, "tid": -914061504, "ts": 1716454223181453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223236606, "dur": 20,
    "args": {
      "External id": 116445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116445,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116445, "pid": 5, "tid": 7, "ts": 1716454223236606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181523, "dur": 12,
    "args": {
      "External id": 116445,
      "cbid": 211, "correlation": 116445
    }
  },
  {
    "ph": "s", "id": 116445, "pid": 76337, "tid": -914061504, "ts": 1716454223181523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223236627, "dur": 27,
    "args": {
      "External id": 116453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116453,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116453, "pid": 5, "tid": 7, "ts": 1716454223236627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181561, "dur": 9,
    "args": {
      "External id": 116453,
      "cbid": 211, "correlation": 116453
    }
  },
  {
    "ph": "s", "id": 116453, "pid": 76337, "tid": -914061504, "ts": 1716454223181561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223236656, "dur": 18,
    "args": {
      "External id": 116464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116464, "pid": 5, "tid": 7, "ts": 1716454223236656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181632, "dur": 12,
    "args": {
      "External id": 116464,
      "cbid": 211, "correlation": 116464
    }
  },
  {
    "ph": "s", "id": 116464, "pid": 76337, "tid": -914061504, "ts": 1716454223181632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223236675, "dur": 16,
    "args": {
      "External id": 116486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116486,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116486, "pid": 5, "tid": 7, "ts": 1716454223236675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181663, "dur": 7,
    "args": {
      "External id": 116486,
      "cbid": 211, "correlation": 116486
    }
  },
  {
    "ph": "s", "id": 116486, "pid": 76337, "tid": -914061504, "ts": 1716454223181663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181748, "dur": 1,
    "args": {
      "External id": 116497,
      "cbid": 251, "correlation": 116497
    }
  },
  {
    "ph": "f", "id": 116497, "pid": 76337, "tid": -914061504, "ts": 1716454223181748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223236693, "dur": 89,
    "args": {
      "External id": 116498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116498,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116498, "pid": 5, "tid": 7, "ts": 1716454223236693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181753, "dur": 13,
    "args": {
      "External id": 116498,
      "cbid": 211, "correlation": 116498
    }
  },
  {
    "ph": "s", "id": 116498, "pid": 76337, "tid": -914061504, "ts": 1716454223181753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181821, "dur": 1,
    "args": {
      "External id": 116509,
      "cbid": 251, "correlation": 116509
    }
  },
  {
    "ph": "f", "id": 116509, "pid": 76337, "tid": -914061504, "ts": 1716454223181821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181825, "dur": 0,
    "args": {
      "External id": 116510,
      "cbid": 251, "correlation": 116510
    }
  },
  {
    "ph": "f", "id": 116510, "pid": 76337, "tid": -914061504, "ts": 1716454223181825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223236784, "dur": 12,
    "args": {
      "External id": 116511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116511,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116511, "pid": 5, "tid": 7, "ts": 1716454223236784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181827, "dur": 12,
    "args": {
      "External id": 116511,
      "cbid": 211, "correlation": 116511
    }
  },
  {
    "ph": "s", "id": 116511, "pid": 76337, "tid": -914061504, "ts": 1716454223181827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223236797, "dur": 6,
    "args": {
      "External id": 116513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116513,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116513, "pid": 5, "tid": 7, "ts": 1716454223236797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181841, "dur": 6,
    "args": {
      "External id": 116513,
      "cbid": 211, "correlation": 116513
    }
  },
  {
    "ph": "s", "id": 116513, "pid": 76337, "tid": -914061504, "ts": 1716454223181841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181898, "dur": 1,
    "args": {
      "External id": 116524,
      "cbid": 251, "correlation": 116524
    }
  },
  {
    "ph": "f", "id": 116524, "pid": 76337, "tid": -914061504, "ts": 1716454223181898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181902, "dur": 0,
    "args": {
      "External id": 116525,
      "cbid": 251, "correlation": 116525
    }
  },
  {
    "ph": "f", "id": 116525, "pid": 76337, "tid": -914061504, "ts": 1716454223181902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223236805, "dur": 8,
    "args": {
      "External id": 116526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116526,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116526, "pid": 5, "tid": 7, "ts": 1716454223236805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181903, "dur": 12,
    "args": {
      "External id": 116526,
      "cbid": 211, "correlation": 116526
    }
  },
  {
    "ph": "s", "id": 116526, "pid": 76337, "tid": -914061504, "ts": 1716454223181903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223236814, "dur": 4,
    "args": {
      "External id": 116528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116528,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116528, "pid": 5, "tid": 7, "ts": 1716454223236814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223181916, "dur": 6,
    "args": {
      "External id": 116528,
      "cbid": 211, "correlation": 116528
    }
  },
  {
    "ph": "s", "id": 116528, "pid": 76337, "tid": -914061504, "ts": 1716454223181916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223236819, "dur": 55,
    "args": {
      "External id": 116553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116553,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116553, "pid": 5, "tid": 7, "ts": 1716454223236819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182001, "dur": 12,
    "args": {
      "External id": 116553,
      "cbid": 211, "correlation": 116553
    }
  },
  {
    "ph": "s", "id": 116553, "pid": 76337, "tid": -914061504, "ts": 1716454223182001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182100, "dur": 1,
    "args": {
      "External id": 116571,
      "cbid": 251, "correlation": 116571
    }
  },
  {
    "ph": "f", "id": 116571, "pid": 76337, "tid": -914061504, "ts": 1716454223182100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223236875, "dur": 91,
    "args": {
      "External id": 116573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116573,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116573, "pid": 5, "tid": 7, "ts": 1716454223236875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182105, "dur": 13,
    "args": {
      "External id": 116573,
      "cbid": 211, "correlation": 116573
    }
  },
  {
    "ph": "s", "id": 116573, "pid": 76337, "tid": -914061504, "ts": 1716454223182105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223236968, "dur": 9,
    "args": {
      "External id": 116581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116581,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116581, "pid": 5, "tid": 7, "ts": 1716454223236968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182175, "dur": 13,
    "args": {
      "External id": 116581,
      "cbid": 211, "correlation": 116581
    }
  },
  {
    "ph": "s", "id": 116581, "pid": 76337, "tid": -914061504, "ts": 1716454223182175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223236978, "dur": 21,
    "args": {
      "External id": 116589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116589, "pid": 5, "tid": 7, "ts": 1716454223236978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182217, "dur": 9,
    "args": {
      "External id": 116589,
      "cbid": 211, "correlation": 116589
    }
  },
  {
    "ph": "s", "id": 116589, "pid": 76337, "tid": -914061504, "ts": 1716454223182217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223237001, "dur": 18,
    "args": {
      "External id": 116611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116611,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116611, "pid": 5, "tid": 7, "ts": 1716454223237001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182267, "dur": 10,
    "args": {
      "External id": 116611,
      "cbid": 211, "correlation": 116611
    }
  },
  {
    "ph": "s", "id": 116611, "pid": 76337, "tid": -914061504, "ts": 1716454223182267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182353, "dur": 1,
    "args": {
      "External id": 116627,
      "cbid": 251, "correlation": 116627
    }
  },
  {
    "ph": "f", "id": 116627, "pid": 76337, "tid": -914061504, "ts": 1716454223182353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182357, "dur": 0,
    "args": {
      "External id": 116629,
      "cbid": 251, "correlation": 116629
    }
  },
  {
    "ph": "f", "id": 116629, "pid": 76337, "tid": -914061504, "ts": 1716454223182357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223237020, "dur": 495,
    "args": {
      "External id": 116630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116630,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116630, "pid": 5, "tid": 7, "ts": 1716454223237020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182359, "dur": 12,
    "args": {
      "External id": 116630,
      "cbid": 211, "correlation": 116630
    }
  },
  {
    "ph": "s", "id": 116630, "pid": 76337, "tid": -914061504, "ts": 1716454223182359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223237516, "dur": 66,
    "args": {
      "External id": 116638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116638, "pid": 5, "tid": 7, "ts": 1716454223237516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182423, "dur": 13,
    "args": {
      "External id": 116638,
      "cbid": 211, "correlation": 116638
    }
  },
  {
    "ph": "s", "id": 116638, "pid": 76337, "tid": -914061504, "ts": 1716454223182423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223237583, "dur": 68,
    "args": {
      "External id": 116646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116646, "pid": 5, "tid": 7, "ts": 1716454223237583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182454, "dur": 8,
    "args": {
      "External id": 116646,
      "cbid": 211, "correlation": 116646
    }
  },
  {
    "ph": "s", "id": 116646, "pid": 76337, "tid": -914061504, "ts": 1716454223182454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182533, "dur": 1,
    "args": {
      "External id": 116662,
      "cbid": 251, "correlation": 116662
    }
  },
  {
    "ph": "f", "id": 116662, "pid": 76337, "tid": -914061504, "ts": 1716454223182533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223237653, "dur": 1,
    "args": {
      "External id": 116664,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116664,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 116664, "pid": 5, "tid": 7, "ts": 1716454223237653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182538, "dur": 9,
    "args": {
      "External id": 116664,
      "cbid": 51, "correlation": 116664
    }
  },
  {
    "ph": "s", "id": 116664, "pid": 76337, "tid": -914061504, "ts": 1716454223182538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223237657, "dur": 272,
    "args": {
      "External id": 116665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116665,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116665, "pid": 5, "tid": 7, "ts": 1716454223237657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182548, "dur": 11,
    "args": {
      "External id": 116665,
      "cbid": 211, "correlation": 116665
    }
  },
  {
    "ph": "s", "id": 116665, "pid": 76337, "tid": -914061504, "ts": 1716454223182548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223237930, "dur": 14,
    "args": {
      "External id": 116673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116673,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116673, "pid": 5, "tid": 7, "ts": 1716454223237930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182590, "dur": 10,
    "args": {
      "External id": 116673,
      "cbid": 211, "correlation": 116673
    }
  },
  {
    "ph": "s", "id": 116673, "pid": 76337, "tid": -914061504, "ts": 1716454223182590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223237945, "dur": 38,
    "args": {
      "External id": 116684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116684, "pid": 5, "tid": 7, "ts": 1716454223237945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182658, "dur": 12,
    "args": {
      "External id": 116684,
      "cbid": 211, "correlation": 116684
    }
  },
  {
    "ph": "s", "id": 116684, "pid": 76337, "tid": -914061504, "ts": 1716454223182658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182722, "dur": 0,
    "args": {
      "External id": 116696,
      "cbid": 317, "correlation": 116696
    }
  },
  {
    "ph": "f", "id": 116696, "pid": 76337, "tid": -914061504, "ts": 1716454223182722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182723, "dur": 0,
    "args": {
      "External id": 116697,
      "cbid": 203, "correlation": 116697
    }
  },
  {
    "ph": "f", "id": 116697, "pid": 76337, "tid": -914061504, "ts": 1716454223182723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182724, "dur": 0,
    "args": {
      "External id": 116698,
      "cbid": 205, "correlation": 116698
    }
  },
  {
    "ph": "f", "id": 116698, "pid": 76337, "tid": -914061504, "ts": 1716454223182724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223237984, "dur": 13,
    "args": {
      "External id": 116702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116702,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116702, "pid": 5, "tid": 7, "ts": 1716454223237984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182739, "dur": 12,
    "args": {
      "External id": 116702,
      "cbid": 211, "correlation": 116702
    }
  },
  {
    "ph": "s", "id": 116702, "pid": 76337, "tid": -914061504, "ts": 1716454223182739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223237998, "dur": 4,
    "args": {
      "External id": 116704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116704,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 116704, "pid": 5, "tid": 7, "ts": 1716454223237998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182756, "dur": 6,
    "args": {
      "External id": 116704,
      "cbid": 211, "correlation": 116704
    }
  },
  {
    "ph": "s", "id": 116704, "pid": 76337, "tid": -914061504, "ts": 1716454223182756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182764, "dur": 0,
    "args": {
      "External id": 116705,
      "cbid": 51, "correlation": 116705
    }
  },
  {
    "ph": "s", "id": 116705, "pid": 76337, "tid": -914061504, "ts": 1716454223182764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223238004, "dur": 98,
    "args": {
      "External id": 116706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116706,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 116706, "pid": 5, "tid": 7, "ts": 1716454223238004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182765, "dur": 5,
    "args": {
      "External id": 116706,
      "cbid": 211, "correlation": 116706
    }
  },
  {
    "ph": "s", "id": 116706, "pid": 76337, "tid": -914061504, "ts": 1716454223182765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223238103, "dur": 17,
    "args": {
      "External id": 116711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116711, "pid": 5, "tid": 7, "ts": 1716454223238103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182792, "dur": 9,
    "args": {
      "External id": 116711,
      "cbid": 211, "correlation": 116711
    }
  },
  {
    "ph": "s", "id": 116711, "pid": 76337, "tid": -914061504, "ts": 1716454223182792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223238121, "dur": 13,
    "args": {
      "External id": 116719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116719,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116719, "pid": 5, "tid": 7, "ts": 1716454223238121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182824, "dur": 9,
    "args": {
      "External id": 116719,
      "cbid": 211, "correlation": 116719
    }
  },
  {
    "ph": "s", "id": 116719, "pid": 76337, "tid": -914061504, "ts": 1716454223182824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182893, "dur": 0,
    "args": {
      "External id": 116729,
      "cbid": 317, "correlation": 116729
    }
  },
  {
    "ph": "f", "id": 116729, "pid": 76337, "tid": -914061504, "ts": 1716454223182893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182894, "dur": 0,
    "args": {
      "External id": 116730,
      "cbid": 203, "correlation": 116730
    }
  },
  {
    "ph": "f", "id": 116730, "pid": 76337, "tid": -914061504, "ts": 1716454223182894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182895, "dur": 0,
    "args": {
      "External id": 116731,
      "cbid": 205, "correlation": 116731
    }
  },
  {
    "ph": "f", "id": 116731, "pid": 76337, "tid": -914061504, "ts": 1716454223182895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223238135, "dur": 12,
    "args": {
      "External id": 116735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116735,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116735, "pid": 5, "tid": 7, "ts": 1716454223238135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182909, "dur": 12,
    "args": {
      "External id": 116735,
      "cbid": 211, "correlation": 116735
    }
  },
  {
    "ph": "s", "id": 116735, "pid": 76337, "tid": -914061504, "ts": 1716454223182909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223238149, "dur": 162,
    "args": {
      "External id": 116737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116737,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116737, "pid": 5, "tid": 7, "ts": 1716454223238149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182923, "dur": 5,
    "args": {
      "External id": 116737,
      "cbid": 211, "correlation": 116737
    }
  },
  {
    "ph": "s", "id": 116737, "pid": 76337, "tid": -914061504, "ts": 1716454223182923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223238313, "dur": 1,
    "args": {
      "External id": 116739,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116739,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 116739, "pid": 5, "tid": 7, "ts": 1716454223238313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182935, "dur": 6,
    "args": {
      "External id": 116739,
      "cbid": 51, "correlation": 116739
    }
  },
  {
    "ph": "s", "id": 116739, "pid": 76337, "tid": -914061504, "ts": 1716454223182935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223238317, "dur": 198,
    "args": {
      "External id": 116740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116740,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 116740, "pid": 5, "tid": 7, "ts": 1716454223238317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182943, "dur": 8,
    "args": {
      "External id": 116740,
      "cbid": 211, "correlation": 116740
    }
  },
  {
    "ph": "s", "id": 116740, "pid": 76337, "tid": -914061504, "ts": 1716454223182943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223238516, "dur": 6,
    "args": {
      "External id": 116742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116742, "pid": 5, "tid": 7, "ts": 1716454223238516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182955, "dur": 5,
    "args": {
      "External id": 116742,
      "cbid": 211, "correlation": 116742
    }
  },
  {
    "ph": "s", "id": 116742, "pid": 76337, "tid": -914061504, "ts": 1716454223182955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223238523, "dur": 6,
    "args": {
      "External id": 116748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116748,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116748, "pid": 5, "tid": 7, "ts": 1716454223238523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223182992, "dur": 9,
    "args": {
      "External id": 116748,
      "cbid": 211, "correlation": 116748
    }
  },
  {
    "ph": "s", "id": 116748, "pid": 76337, "tid": -914061504, "ts": 1716454223182992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223238531, "dur": 11,
    "args": {
      "External id": 116768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116768,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 116768, "pid": 5, "tid": 7, "ts": 1716454223238531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183087, "dur": 12,
    "args": {
      "External id": 116768,
      "cbid": 211, "correlation": 116768
    }
  },
  {
    "ph": "s", "id": 116768, "pid": 76337, "tid": -914061504, "ts": 1716454223183087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223238543, "dur": 4,
    "args": {
      "External id": 116780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116780,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 116780, "pid": 5, "tid": 7, "ts": 1716454223238543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183110, "dur": 7,
    "args": {
      "External id": 116780,
      "cbid": 211, "correlation": 116780
    }
  },
  {
    "ph": "s", "id": 116780, "pid": 76337, "tid": -914061504, "ts": 1716454223183110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223238548, "dur": 8,
    "args": {
      "External id": 116783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116783, "pid": 5, "tid": 7, "ts": 1716454223238548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183129, "dur": 7,
    "args": {
      "External id": 116783,
      "cbid": 211, "correlation": 116783
    }
  },
  {
    "ph": "s", "id": 116783, "pid": 76337, "tid": -914061504, "ts": 1716454223183129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223238558, "dur": 6,
    "args": {
      "External id": 116792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116792,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116792, "pid": 5, "tid": 7, "ts": 1716454223238558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183169, "dur": 10,
    "args": {
      "External id": 116792,
      "cbid": 211, "correlation": 116792
    }
  },
  {
    "ph": "s", "id": 116792, "pid": 76337, "tid": -914061504, "ts": 1716454223183169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183222, "dur": 0,
    "args": {
      "External id": 116802,
      "cbid": 317, "correlation": 116802
    }
  },
  {
    "ph": "f", "id": 116802, "pid": 76337, "tid": -914061504, "ts": 1716454223183222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183223, "dur": 0,
    "args": {
      "External id": 116803,
      "cbid": 203, "correlation": 116803
    }
  },
  {
    "ph": "f", "id": 116803, "pid": 76337, "tid": -914061504, "ts": 1716454223183223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183223, "dur": 0,
    "args": {
      "External id": 116804,
      "cbid": 205, "correlation": 116804
    }
  },
  {
    "ph": "f", "id": 116804, "pid": 76337, "tid": -914061504, "ts": 1716454223183223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223238565, "dur": 5,
    "args": {
      "External id": 116808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116808,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116808, "pid": 5, "tid": 7, "ts": 1716454223238565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183239, "dur": 11,
    "args": {
      "External id": 116808,
      "cbid": 211, "correlation": 116808
    }
  },
  {
    "ph": "s", "id": 116808, "pid": 76337, "tid": -914061504, "ts": 1716454223183239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223238571, "dur": 162,
    "args": {
      "External id": 116810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116810, "pid": 5, "tid": 7, "ts": 1716454223238571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183252, "dur": 5,
    "args": {
      "External id": 116810,
      "cbid": 211, "correlation": 116810
    }
  },
  {
    "ph": "s", "id": 116810, "pid": 76337, "tid": -914061504, "ts": 1716454223183252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223238736, "dur": 1,
    "args": {
      "External id": 116812,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116812,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 116812, "pid": 5, "tid": 7, "ts": 1716454223238736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183263, "dur": 6,
    "args": {
      "External id": 116812,
      "cbid": 51, "correlation": 116812
    }
  },
  {
    "ph": "s", "id": 116812, "pid": 76337, "tid": -914061504, "ts": 1716454223183263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223238740, "dur": 270,
    "args": {
      "External id": 116813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116813,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116813, "pid": 5, "tid": 7, "ts": 1716454223238740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183270, "dur": 6,
    "args": {
      "External id": 116813,
      "cbid": 211, "correlation": 116813
    }
  },
  {
    "ph": "s", "id": 116813, "pid": 76337, "tid": -914061504, "ts": 1716454223183270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239011, "dur": 6,
    "args": {
      "External id": 116815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116815,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116815, "pid": 5, "tid": 7, "ts": 1716454223239011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183281, "dur": 6,
    "args": {
      "External id": 116815,
      "cbid": 211, "correlation": 116815
    }
  },
  {
    "ph": "s", "id": 116815, "pid": 76337, "tid": -914061504, "ts": 1716454223183281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223239018, "dur": 7,
    "args": {
      "External id": 116821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116821,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116821, "pid": 5, "tid": 7, "ts": 1716454223239018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183309, "dur": 8,
    "args": {
      "External id": 116821,
      "cbid": 211, "correlation": 116821
    }
  },
  {
    "ph": "s", "id": 116821, "pid": 76337, "tid": -914061504, "ts": 1716454223183309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223239026, "dur": 4,
    "args": {
      "External id": 116829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116829,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 116829, "pid": 5, "tid": 7, "ts": 1716454223239026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183353, "dur": 9,
    "args": {
      "External id": 116829,
      "cbid": 211, "correlation": 116829
    }
  },
  {
    "ph": "s", "id": 116829, "pid": 76337, "tid": -914061504, "ts": 1716454223183353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183418, "dur": 1,
    "args": {
      "External id": 116845,
      "cbid": 251, "correlation": 116845
    }
  },
  {
    "ph": "f", "id": 116845, "pid": 76337, "tid": -914061504, "ts": 1716454223183418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183423, "dur": 0,
    "args": {
      "External id": 116847,
      "cbid": 251, "correlation": 116847
    }
  },
  {
    "ph": "f", "id": 116847, "pid": 76337, "tid": -914061504, "ts": 1716454223183423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223239031, "dur": 13,
    "args": {
      "External id": 116848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116848,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116848, "pid": 5, "tid": 7, "ts": 1716454223239031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183425, "dur": 12,
    "args": {
      "External id": 116848,
      "cbid": 211, "correlation": 116848
    }
  },
  {
    "ph": "s", "id": 116848, "pid": 76337, "tid": -914061504, "ts": 1716454223183425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223239046, "dur": 5,
    "args": {
      "External id": 116850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116850,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116850, "pid": 5, "tid": 7, "ts": 1716454223239046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183439, "dur": 5,
    "args": {
      "External id": 116850,
      "cbid": 211, "correlation": 116850
    }
  },
  {
    "ph": "s", "id": 116850, "pid": 76337, "tid": -914061504, "ts": 1716454223183439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223239052, "dur": 6,
    "args": {
      "External id": 116860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116860,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116860, "pid": 5, "tid": 7, "ts": 1716454223239052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183496, "dur": 12,
    "args": {
      "External id": 116860,
      "cbid": 211, "correlation": 116860
    }
  },
  {
    "ph": "s", "id": 116860, "pid": 76337, "tid": -914061504, "ts": 1716454223183496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239059, "dur": 10,
    "args": {
      "External id": 116880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116880,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 116880, "pid": 5, "tid": 7, "ts": 1716454223239059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183562, "dur": 11,
    "args": {
      "External id": 116880,
      "cbid": 211, "correlation": 116880
    }
  },
  {
    "ph": "s", "id": 116880, "pid": 76337, "tid": -914061504, "ts": 1716454223183562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223239070, "dur": 4,
    "args": {
      "External id": 116892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116892,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 116892, "pid": 5, "tid": 7, "ts": 1716454223239070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183583, "dur": 7,
    "args": {
      "External id": 116892,
      "cbid": 211, "correlation": 116892
    }
  },
  {
    "ph": "s", "id": 116892, "pid": 76337, "tid": -914061504, "ts": 1716454223183583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223239076, "dur": 7,
    "args": {
      "External id": 116895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116895,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116895, "pid": 5, "tid": 7, "ts": 1716454223239076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183602, "dur": 6,
    "args": {
      "External id": 116895,
      "cbid": 211, "correlation": 116895
    }
  },
  {
    "ph": "s", "id": 116895, "pid": 76337, "tid": -914061504, "ts": 1716454223183602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223239084, "dur": 5,
    "args": {
      "External id": 116904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116904,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116904, "pid": 5, "tid": 7, "ts": 1716454223239084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183643, "dur": 10,
    "args": {
      "External id": 116904,
      "cbid": 211, "correlation": 116904
    }
  },
  {
    "ph": "s", "id": 116904, "pid": 76337, "tid": -914061504, "ts": 1716454223183643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183705, "dur": 0,
    "args": {
      "External id": 116914,
      "cbid": 317, "correlation": 116914
    }
  },
  {
    "ph": "f", "id": 116914, "pid": 76337, "tid": -914061504, "ts": 1716454223183705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183706, "dur": 0,
    "args": {
      "External id": 116915,
      "cbid": 203, "correlation": 116915
    }
  },
  {
    "ph": "f", "id": 116915, "pid": 76337, "tid": -914061504, "ts": 1716454223183706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183707, "dur": 0,
    "args": {
      "External id": 116916,
      "cbid": 205, "correlation": 116916
    }
  },
  {
    "ph": "f", "id": 116916, "pid": 76337, "tid": -914061504, "ts": 1716454223183707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239090, "dur": 5,
    "args": {
      "External id": 116920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116920,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116920, "pid": 5, "tid": 7, "ts": 1716454223239090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183721, "dur": 12,
    "args": {
      "External id": 116920,
      "cbid": 211, "correlation": 116920
    }
  },
  {
    "ph": "s", "id": 116920, "pid": 76337, "tid": -914061504, "ts": 1716454223183721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239096, "dur": 163,
    "args": {
      "External id": 116922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116922,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116922, "pid": 5, "tid": 7, "ts": 1716454223239096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183736, "dur": 6,
    "args": {
      "External id": 116922,
      "cbid": 211, "correlation": 116922
    }
  },
  {
    "ph": "s", "id": 116922, "pid": 76337, "tid": -914061504, "ts": 1716454223183736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223239261, "dur": 1,
    "args": {
      "External id": 116924,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 116924,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 116924, "pid": 5, "tid": 7, "ts": 1716454223239261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183747, "dur": 6,
    "args": {
      "External id": 116924,
      "cbid": 51, "correlation": 116924
    }
  },
  {
    "ph": "s", "id": 116924, "pid": 76337, "tid": -914061504, "ts": 1716454223183747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223239265, "dur": 259,
    "args": {
      "External id": 116925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116925,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 116925, "pid": 5, "tid": 7, "ts": 1716454223239265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183754, "dur": 6,
    "args": {
      "External id": 116925,
      "cbid": 211, "correlation": 116925
    }
  },
  {
    "ph": "s", "id": 116925, "pid": 76337, "tid": -914061504, "ts": 1716454223183754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239525, "dur": 6,
    "args": {
      "External id": 116927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116927,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 116927, "pid": 5, "tid": 7, "ts": 1716454223239525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183764, "dur": 5,
    "args": {
      "External id": 116927,
      "cbid": 211, "correlation": 116927
    }
  },
  {
    "ph": "s", "id": 116927, "pid": 76337, "tid": -914061504, "ts": 1716454223183764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223239532, "dur": 6,
    "args": {
      "External id": 116933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116933,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116933, "pid": 5, "tid": 7, "ts": 1716454223239532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183793, "dur": 9,
    "args": {
      "External id": 116933,
      "cbid": 211, "correlation": 116933
    }
  },
  {
    "ph": "s", "id": 116933, "pid": 76337, "tid": -914061504, "ts": 1716454223183793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223239540, "dur": 5,
    "args": {
      "External id": 116941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116941,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116941, "pid": 5, "tid": 7, "ts": 1716454223239540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183825, "dur": 8,
    "args": {
      "External id": 116941,
      "cbid": 211, "correlation": 116941
    }
  },
  {
    "ph": "s", "id": 116941, "pid": 76337, "tid": -914061504, "ts": 1716454223183825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223239546, "dur": 4,
    "args": {
      "External id": 116949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116949,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116949, "pid": 5, "tid": 7, "ts": 1716454223239546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183853, "dur": 8,
    "args": {
      "External id": 116949,
      "cbid": 211, "correlation": 116949
    }
  },
  {
    "ph": "s", "id": 116949, "pid": 76337, "tid": -914061504, "ts": 1716454223183853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239552, "dur": 10,
    "args": {
      "External id": 116969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116969,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 116969, "pid": 5, "tid": 7, "ts": 1716454223239552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183927, "dur": 12,
    "args": {
      "External id": 116969,
      "cbid": 211, "correlation": 116969
    }
  },
  {
    "ph": "s", "id": 116969, "pid": 76337, "tid": -914061504, "ts": 1716454223183927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223239563, "dur": 4,
    "args": {
      "External id": 116981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116981,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 116981, "pid": 5, "tid": 7, "ts": 1716454223239563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183949, "dur": 6,
    "args": {
      "External id": 116981,
      "cbid": 211, "correlation": 116981
    }
  },
  {
    "ph": "s", "id": 116981, "pid": 76337, "tid": -914061504, "ts": 1716454223183949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223239568, "dur": 7,
    "args": {
      "External id": 116984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116984, "pid": 5, "tid": 7, "ts": 1716454223239568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223183967, "dur": 14,
    "args": {
      "External id": 116984,
      "cbid": 211, "correlation": 116984
    }
  },
  {
    "ph": "s", "id": 116984, "pid": 76337, "tid": -914061504, "ts": 1716454223183967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223239576, "dur": 5,
    "args": {
      "External id": 116993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 116993,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 116993, "pid": 5, "tid": 7, "ts": 1716454223239576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184014, "dur": 11,
    "args": {
      "External id": 116993,
      "cbid": 211, "correlation": 116993
    }
  },
  {
    "ph": "s", "id": 116993, "pid": 76337, "tid": -914061504, "ts": 1716454223184014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184067, "dur": 0,
    "args": {
      "External id": 117003,
      "cbid": 317, "correlation": 117003
    }
  },
  {
    "ph": "f", "id": 117003, "pid": 76337, "tid": -914061504, "ts": 1716454223184067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184068, "dur": 0,
    "args": {
      "External id": 117004,
      "cbid": 203, "correlation": 117004
    }
  },
  {
    "ph": "f", "id": 117004, "pid": 76337, "tid": -914061504, "ts": 1716454223184068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184069, "dur": 0,
    "args": {
      "External id": 117005,
      "cbid": 205, "correlation": 117005
    }
  },
  {
    "ph": "f", "id": 117005, "pid": 76337, "tid": -914061504, "ts": 1716454223184069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239582, "dur": 5,
    "args": {
      "External id": 117009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117009,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117009, "pid": 5, "tid": 7, "ts": 1716454223239582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184082, "dur": 12,
    "args": {
      "External id": 117009,
      "cbid": 211, "correlation": 117009
    }
  },
  {
    "ph": "s", "id": 117009, "pid": 76337, "tid": -914061504, "ts": 1716454223184082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223239588, "dur": 161,
    "args": {
      "External id": 117011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117011,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117011, "pid": 5, "tid": 7, "ts": 1716454223239588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184096, "dur": 5,
    "args": {
      "External id": 117011,
      "cbid": 211, "correlation": 117011
    }
  },
  {
    "ph": "s", "id": 117011, "pid": 76337, "tid": -914061504, "ts": 1716454223184096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223239752, "dur": 1,
    "args": {
      "External id": 117013,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117013,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 117013, "pid": 5, "tid": 7, "ts": 1716454223239752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184107, "dur": 6,
    "args": {
      "External id": 117013,
      "cbid": 51, "correlation": 117013
    }
  },
  {
    "ph": "s", "id": 117013, "pid": 76337, "tid": -914061504, "ts": 1716454223184107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223239756, "dur": 257,
    "args": {
      "External id": 117014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117014,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117014, "pid": 5, "tid": 7, "ts": 1716454223239756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184114, "dur": 6,
    "args": {
      "External id": 117014,
      "cbid": 211, "correlation": 117014
    }
  },
  {
    "ph": "s", "id": 117014, "pid": 76337, "tid": -914061504, "ts": 1716454223184114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240014, "dur": 6,
    "args": {
      "External id": 117016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117016,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117016, "pid": 5, "tid": 7, "ts": 1716454223240014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184124, "dur": 5,
    "args": {
      "External id": 117016,
      "cbid": 211, "correlation": 117016
    }
  },
  {
    "ph": "s", "id": 117016, "pid": 76337, "tid": -914061504, "ts": 1716454223184124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223240022, "dur": 6,
    "args": {
      "External id": 117022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117022,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117022, "pid": 5, "tid": 7, "ts": 1716454223240022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184152, "dur": 8,
    "args": {
      "External id": 117022,
      "cbid": 211, "correlation": 117022
    }
  },
  {
    "ph": "s", "id": 117022, "pid": 76337, "tid": -914061504, "ts": 1716454223184152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223240029, "dur": 4,
    "args": {
      "External id": 117030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117030,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 117030, "pid": 5, "tid": 7, "ts": 1716454223240029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184196, "dur": 9,
    "args": {
      "External id": 117030,
      "cbid": 211, "correlation": 117030
    }
  },
  {
    "ph": "s", "id": 117030, "pid": 76337, "tid": -914061504, "ts": 1716454223184196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184258, "dur": 1,
    "args": {
      "External id": 117046,
      "cbid": 251, "correlation": 117046
    }
  },
  {
    "ph": "f", "id": 117046, "pid": 76337, "tid": -914061504, "ts": 1716454223184258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184263, "dur": 0,
    "args": {
      "External id": 117048,
      "cbid": 251, "correlation": 117048
    }
  },
  {
    "ph": "f", "id": 117048, "pid": 76337, "tid": -914061504, "ts": 1716454223184263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223240034, "dur": 10,
    "args": {
      "External id": 117049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117049,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117049, "pid": 5, "tid": 7, "ts": 1716454223240034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184265, "dur": 11,
    "args": {
      "External id": 117049,
      "cbid": 211, "correlation": 117049
    }
  },
  {
    "ph": "s", "id": 117049, "pid": 76337, "tid": -914061504, "ts": 1716454223184265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223240045, "dur": 4,
    "args": {
      "External id": 117051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117051,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117051, "pid": 5, "tid": 7, "ts": 1716454223240045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184278, "dur": 5,
    "args": {
      "External id": 117051,
      "cbid": 211, "correlation": 117051
    }
  },
  {
    "ph": "s", "id": 117051, "pid": 76337, "tid": -914061504, "ts": 1716454223184278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223240050, "dur": 6,
    "args": {
      "External id": 117061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117061, "pid": 5, "tid": 7, "ts": 1716454223240050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184334, "dur": 13,
    "args": {
      "External id": 117061,
      "cbid": 211, "correlation": 117061
    }
  },
  {
    "ph": "s", "id": 117061, "pid": 76337, "tid": -914061504, "ts": 1716454223184334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240057, "dur": 10,
    "args": {
      "External id": 117081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117081,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 117081, "pid": 5, "tid": 7, "ts": 1716454223240057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184401, "dur": 10,
    "args": {
      "External id": 117081,
      "cbid": 211, "correlation": 117081
    }
  },
  {
    "ph": "s", "id": 117081, "pid": 76337, "tid": -914061504, "ts": 1716454223184401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223240068, "dur": 4,
    "args": {
      "External id": 117093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117093,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 117093, "pid": 5, "tid": 7, "ts": 1716454223240068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184421, "dur": 6,
    "args": {
      "External id": 117093,
      "cbid": 211, "correlation": 117093
    }
  },
  {
    "ph": "s", "id": 117093, "pid": 76337, "tid": -914061504, "ts": 1716454223184421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223240073, "dur": 7,
    "args": {
      "External id": 117096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117096, "pid": 5, "tid": 7, "ts": 1716454223240073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184439, "dur": 7,
    "args": {
      "External id": 117096,
      "cbid": 211, "correlation": 117096
    }
  },
  {
    "ph": "s", "id": 117096, "pid": 76337, "tid": -914061504, "ts": 1716454223184439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223240082, "dur": 5,
    "args": {
      "External id": 117105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117105,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117105, "pid": 5, "tid": 7, "ts": 1716454223240082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184479, "dur": 10,
    "args": {
      "External id": 117105,
      "cbid": 211, "correlation": 117105
    }
  },
  {
    "ph": "s", "id": 117105, "pid": 76337, "tid": -914061504, "ts": 1716454223184479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184541, "dur": 0,
    "args": {
      "External id": 117115,
      "cbid": 317, "correlation": 117115
    }
  },
  {
    "ph": "f", "id": 117115, "pid": 76337, "tid": -914061504, "ts": 1716454223184541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184541, "dur": 0,
    "args": {
      "External id": 117116,
      "cbid": 203, "correlation": 117116
    }
  },
  {
    "ph": "f", "id": 117116, "pid": 76337, "tid": -914061504, "ts": 1716454223184541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184542, "dur": 0,
    "args": {
      "External id": 117117,
      "cbid": 205, "correlation": 117117
    }
  },
  {
    "ph": "f", "id": 117117, "pid": 76337, "tid": -914061504, "ts": 1716454223184542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240087, "dur": 5,
    "args": {
      "External id": 117121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117121,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117121, "pid": 5, "tid": 7, "ts": 1716454223240087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184557, "dur": 12,
    "args": {
      "External id": 117121,
      "cbid": 211, "correlation": 117121
    }
  },
  {
    "ph": "s", "id": 117121, "pid": 76337, "tid": -914061504, "ts": 1716454223184557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240094, "dur": 162,
    "args": {
      "External id": 117123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117123,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117123, "pid": 5, "tid": 7, "ts": 1716454223240094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184571, "dur": 5,
    "args": {
      "External id": 117123,
      "cbid": 211, "correlation": 117123
    }
  },
  {
    "ph": "s", "id": 117123, "pid": 76337, "tid": -914061504, "ts": 1716454223184571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223240258, "dur": 1,
    "args": {
      "External id": 117125,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117125,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 117125, "pid": 5, "tid": 7, "ts": 1716454223240258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184581, "dur": 6,
    "args": {
      "External id": 117125,
      "cbid": 51, "correlation": 117125
    }
  },
  {
    "ph": "s", "id": 117125, "pid": 76337, "tid": -914061504, "ts": 1716454223184581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223240261, "dur": 260,
    "args": {
      "External id": 117126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117126,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117126, "pid": 5, "tid": 7, "ts": 1716454223240261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184589, "dur": 6,
    "args": {
      "External id": 117126,
      "cbid": 211, "correlation": 117126
    }
  },
  {
    "ph": "s", "id": 117126, "pid": 76337, "tid": -914061504, "ts": 1716454223184589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240523, "dur": 6,
    "args": {
      "External id": 117128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117128,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117128, "pid": 5, "tid": 7, "ts": 1716454223240523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184600, "dur": 5,
    "args": {
      "External id": 117128,
      "cbid": 211, "correlation": 117128
    }
  },
  {
    "ph": "s", "id": 117128, "pid": 76337, "tid": -914061504, "ts": 1716454223184600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223240530, "dur": 6,
    "args": {
      "External id": 117134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117134, "pid": 5, "tid": 7, "ts": 1716454223240530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184627, "dur": 9,
    "args": {
      "External id": 117134,
      "cbid": 211, "correlation": 117134
    }
  },
  {
    "ph": "s", "id": 117134, "pid": 76337, "tid": -914061504, "ts": 1716454223184627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223240537, "dur": 5,
    "args": {
      "External id": 117142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117142,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117142, "pid": 5, "tid": 7, "ts": 1716454223240537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184661, "dur": 8,
    "args": {
      "External id": 117142,
      "cbid": 211, "correlation": 117142
    }
  },
  {
    "ph": "s", "id": 117142, "pid": 76337, "tid": -914061504, "ts": 1716454223184661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223240543, "dur": 4,
    "args": {
      "External id": 117150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117150,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117150, "pid": 5, "tid": 7, "ts": 1716454223240543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184690, "dur": 9,
    "args": {
      "External id": 117150,
      "cbid": 211, "correlation": 117150
    }
  },
  {
    "ph": "s", "id": 117150, "pid": 76337, "tid": -914061504, "ts": 1716454223184690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240549, "dur": 9,
    "args": {
      "External id": 117170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117170,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 117170, "pid": 5, "tid": 7, "ts": 1716454223240549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184794, "dur": 13,
    "args": {
      "External id": 117170,
      "cbid": 211, "correlation": 117170
    }
  },
  {
    "ph": "s", "id": 117170, "pid": 76337, "tid": -914061504, "ts": 1716454223184794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223240560, "dur": 4,
    "args": {
      "External id": 117182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117182,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 117182, "pid": 5, "tid": 7, "ts": 1716454223240560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184817, "dur": 6,
    "args": {
      "External id": 117182,
      "cbid": 211, "correlation": 117182
    }
  },
  {
    "ph": "s", "id": 117182, "pid": 76337, "tid": -914061504, "ts": 1716454223184817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223240565, "dur": 7,
    "args": {
      "External id": 117185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117185, "pid": 5, "tid": 7, "ts": 1716454223240565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184834, "dur": 7,
    "args": {
      "External id": 117185,
      "cbid": 211, "correlation": 117185
    }
  },
  {
    "ph": "s", "id": 117185, "pid": 76337, "tid": -914061504, "ts": 1716454223184834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223240573, "dur": 5,
    "args": {
      "External id": 117194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117194,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117194, "pid": 5, "tid": 7, "ts": 1716454223240573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184872, "dur": 9,
    "args": {
      "External id": 117194,
      "cbid": 211, "correlation": 117194
    }
  },
  {
    "ph": "s", "id": 117194, "pid": 76337, "tid": -914061504, "ts": 1716454223184872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184924, "dur": 0,
    "args": {
      "External id": 117204,
      "cbid": 317, "correlation": 117204
    }
  },
  {
    "ph": "f", "id": 117204, "pid": 76337, "tid": -914061504, "ts": 1716454223184924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184925, "dur": 0,
    "args": {
      "External id": 117205,
      "cbid": 203, "correlation": 117205
    }
  },
  {
    "ph": "f", "id": 117205, "pid": 76337, "tid": -914061504, "ts": 1716454223184925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184926, "dur": 0,
    "args": {
      "External id": 117206,
      "cbid": 205, "correlation": 117206
    }
  },
  {
    "ph": "f", "id": 117206, "pid": 76337, "tid": -914061504, "ts": 1716454223184926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240579, "dur": 5,
    "args": {
      "External id": 117210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117210,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117210, "pid": 5, "tid": 7, "ts": 1716454223240579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184939, "dur": 12,
    "args": {
      "External id": 117210,
      "cbid": 211, "correlation": 117210
    }
  },
  {
    "ph": "s", "id": 117210, "pid": 76337, "tid": -914061504, "ts": 1716454223184939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223240585, "dur": 162,
    "args": {
      "External id": 117212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117212,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117212, "pid": 5, "tid": 7, "ts": 1716454223240585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184953, "dur": 6,
    "args": {
      "External id": 117212,
      "cbid": 211, "correlation": 117212
    }
  },
  {
    "ph": "s", "id": 117212, "pid": 76337, "tid": -914061504, "ts": 1716454223184953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223240749, "dur": 1,
    "args": {
      "External id": 117214,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117214,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 117214, "pid": 5, "tid": 7, "ts": 1716454223240749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184964, "dur": 6,
    "args": {
      "External id": 117214,
      "cbid": 51, "correlation": 117214
    }
  },
  {
    "ph": "s", "id": 117214, "pid": 76337, "tid": -914061504, "ts": 1716454223184964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223240753, "dur": 258,
    "args": {
      "External id": 117215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117215,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117215, "pid": 5, "tid": 7, "ts": 1716454223240753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184971, "dur": 14,
    "args": {
      "External id": 117215,
      "cbid": 211, "correlation": 117215
    }
  },
  {
    "ph": "s", "id": 117215, "pid": 76337, "tid": -914061504, "ts": 1716454223184971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241012, "dur": 6,
    "args": {
      "External id": 117217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117217,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117217, "pid": 5, "tid": 7, "ts": 1716454223241012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223184990, "dur": 5,
    "args": {
      "External id": 117217,
      "cbid": 211, "correlation": 117217
    }
  },
  {
    "ph": "s", "id": 117217, "pid": 76337, "tid": -914061504, "ts": 1716454223184990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241019, "dur": 6,
    "args": {
      "External id": 117223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117223, "pid": 5, "tid": 7, "ts": 1716454223241019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185019, "dur": 8,
    "args": {
      "External id": 117223,
      "cbid": 211, "correlation": 117223
    }
  },
  {
    "ph": "s", "id": 117223, "pid": 76337, "tid": -914061504, "ts": 1716454223185019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223241027, "dur": 4,
    "args": {
      "External id": 117231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117231,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 117231, "pid": 5, "tid": 7, "ts": 1716454223241027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185063, "dur": 9,
    "args": {
      "External id": 117231,
      "cbid": 211, "correlation": 117231
    }
  },
  {
    "ph": "s", "id": 117231, "pid": 76337, "tid": -914061504, "ts": 1716454223185063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185126, "dur": 1,
    "args": {
      "External id": 117247,
      "cbid": 251, "correlation": 117247
    }
  },
  {
    "ph": "f", "id": 117247, "pid": 76337, "tid": -914061504, "ts": 1716454223185126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185131, "dur": 0,
    "args": {
      "External id": 117249,
      "cbid": 251, "correlation": 117249
    }
  },
  {
    "ph": "f", "id": 117249, "pid": 76337, "tid": -914061504, "ts": 1716454223185131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223241032, "dur": 11,
    "args": {
      "External id": 117250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117250,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117250, "pid": 5, "tid": 7, "ts": 1716454223241032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185133, "dur": 11,
    "args": {
      "External id": 117250,
      "cbid": 211, "correlation": 117250
    }
  },
  {
    "ph": "s", "id": 117250, "pid": 76337, "tid": -914061504, "ts": 1716454223185133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223241044, "dur": 4,
    "args": {
      "External id": 117252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117252,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117252, "pid": 5, "tid": 7, "ts": 1716454223241044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185146, "dur": 5,
    "args": {
      "External id": 117252,
      "cbid": 211, "correlation": 117252
    }
  },
  {
    "ph": "s", "id": 117252, "pid": 76337, "tid": -914061504, "ts": 1716454223185146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241049, "dur": 6,
    "args": {
      "External id": 117262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117262,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117262, "pid": 5, "tid": 7, "ts": 1716454223241049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185201, "dur": 12,
    "args": {
      "External id": 117262,
      "cbid": 211, "correlation": 117262
    }
  },
  {
    "ph": "s", "id": 117262, "pid": 76337, "tid": -914061504, "ts": 1716454223185201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241056, "dur": 10,
    "args": {
      "External id": 117282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117282,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 117282, "pid": 5, "tid": 7, "ts": 1716454223241056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185268, "dur": 11,
    "args": {
      "External id": 117282,
      "cbid": 211, "correlation": 117282
    }
  },
  {
    "ph": "s", "id": 117282, "pid": 76337, "tid": -914061504, "ts": 1716454223185268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223241067, "dur": 4,
    "args": {
      "External id": 117294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117294,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 117294, "pid": 5, "tid": 7, "ts": 1716454223241067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185288, "dur": 6,
    "args": {
      "External id": 117294,
      "cbid": 211, "correlation": 117294
    }
  },
  {
    "ph": "s", "id": 117294, "pid": 76337, "tid": -914061504, "ts": 1716454223185288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241072, "dur": 7,
    "args": {
      "External id": 117297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117297,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117297, "pid": 5, "tid": 7, "ts": 1716454223241072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185308, "dur": 6,
    "args": {
      "External id": 117297,
      "cbid": 211, "correlation": 117297
    }
  },
  {
    "ph": "s", "id": 117297, "pid": 76337, "tid": -914061504, "ts": 1716454223185308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223241080, "dur": 5,
    "args": {
      "External id": 117306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117306,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117306, "pid": 5, "tid": 7, "ts": 1716454223241080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185348, "dur": 10,
    "args": {
      "External id": 117306,
      "cbid": 211, "correlation": 117306
    }
  },
  {
    "ph": "s", "id": 117306, "pid": 76337, "tid": -914061504, "ts": 1716454223185348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185411, "dur": 0,
    "args": {
      "External id": 117316,
      "cbid": 317, "correlation": 117316
    }
  },
  {
    "ph": "f", "id": 117316, "pid": 76337, "tid": -914061504, "ts": 1716454223185411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185412, "dur": 0,
    "args": {
      "External id": 117317,
      "cbid": 203, "correlation": 117317
    }
  },
  {
    "ph": "f", "id": 117317, "pid": 76337, "tid": -914061504, "ts": 1716454223185412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185413, "dur": 0,
    "args": {
      "External id": 117318,
      "cbid": 205, "correlation": 117318
    }
  },
  {
    "ph": "f", "id": 117318, "pid": 76337, "tid": -914061504, "ts": 1716454223185413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241086, "dur": 5,
    "args": {
      "External id": 117322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117322,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117322, "pid": 5, "tid": 7, "ts": 1716454223241086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185427, "dur": 12,
    "args": {
      "External id": 117322,
      "cbid": 211, "correlation": 117322
    }
  },
  {
    "ph": "s", "id": 117322, "pid": 76337, "tid": -914061504, "ts": 1716454223185427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241092, "dur": 162,
    "args": {
      "External id": 117324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117324,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117324, "pid": 5, "tid": 7, "ts": 1716454223241092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185441, "dur": 5,
    "args": {
      "External id": 117324,
      "cbid": 211, "correlation": 117324
    }
  },
  {
    "ph": "s", "id": 117324, "pid": 76337, "tid": -914061504, "ts": 1716454223185441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223241256, "dur": 1,
    "args": {
      "External id": 117326,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117326,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 117326, "pid": 5, "tid": 7, "ts": 1716454223241256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185452, "dur": 7,
    "args": {
      "External id": 117326,
      "cbid": 51, "correlation": 117326
    }
  },
  {
    "ph": "s", "id": 117326, "pid": 76337, "tid": -914061504, "ts": 1716454223185452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223241260, "dur": 258,
    "args": {
      "External id": 117327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117327,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117327, "pid": 5, "tid": 7, "ts": 1716454223241260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185460, "dur": 6,
    "args": {
      "External id": 117327,
      "cbid": 211, "correlation": 117327
    }
  },
  {
    "ph": "s", "id": 117327, "pid": 76337, "tid": -914061504, "ts": 1716454223185460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241520, "dur": 6,
    "args": {
      "External id": 117329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117329,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117329, "pid": 5, "tid": 7, "ts": 1716454223241520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185469, "dur": 5,
    "args": {
      "External id": 117329,
      "cbid": 211, "correlation": 117329
    }
  },
  {
    "ph": "s", "id": 117329, "pid": 76337, "tid": -914061504, "ts": 1716454223185469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241527, "dur": 6,
    "args": {
      "External id": 117335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117335,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117335, "pid": 5, "tid": 7, "ts": 1716454223241527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185497, "dur": 9,
    "args": {
      "External id": 117335,
      "cbid": 211, "correlation": 117335
    }
  },
  {
    "ph": "s", "id": 117335, "pid": 76337, "tid": -914061504, "ts": 1716454223185497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223241535, "dur": 5,
    "args": {
      "External id": 117343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117343,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117343, "pid": 5, "tid": 7, "ts": 1716454223241535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185530, "dur": 8,
    "args": {
      "External id": 117343,
      "cbid": 211, "correlation": 117343
    }
  },
  {
    "ph": "s", "id": 117343, "pid": 76337, "tid": -914061504, "ts": 1716454223185530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223241541, "dur": 4,
    "args": {
      "External id": 117351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117351,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117351, "pid": 5, "tid": 7, "ts": 1716454223241541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185560, "dur": 8,
    "args": {
      "External id": 117351,
      "cbid": 211, "correlation": 117351
    }
  },
  {
    "ph": "s", "id": 117351, "pid": 76337, "tid": -914061504, "ts": 1716454223185560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241547, "dur": 10,
    "args": {
      "External id": 117371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117371,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 117371, "pid": 5, "tid": 7, "ts": 1716454223241547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185702, "dur": 14,
    "args": {
      "External id": 117371,
      "cbid": 211, "correlation": 117371
    }
  },
  {
    "ph": "s", "id": 117371, "pid": 76337, "tid": -914061504, "ts": 1716454223185702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223241558, "dur": 4,
    "args": {
      "External id": 117383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117383,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 117383, "pid": 5, "tid": 7, "ts": 1716454223241558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185725, "dur": 6,
    "args": {
      "External id": 117383,
      "cbid": 211, "correlation": 117383
    }
  },
  {
    "ph": "s", "id": 117383, "pid": 76337, "tid": -914061504, "ts": 1716454223185725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241563, "dur": 7,
    "args": {
      "External id": 117386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117386, "pid": 5, "tid": 7, "ts": 1716454223241563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185744, "dur": 7,
    "args": {
      "External id": 117386,
      "cbid": 211, "correlation": 117386
    }
  },
  {
    "ph": "s", "id": 117386, "pid": 76337, "tid": -914061504, "ts": 1716454223185744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185803, "dur": 0,
    "args": {
      "External id": 117397,
      "cbid": 317, "correlation": 117397
    }
  },
  {
    "ph": "f", "id": 117397, "pid": 76337, "tid": -914061504, "ts": 1716454223185803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185804, "dur": 0,
    "args": {
      "External id": 117398,
      "cbid": 203, "correlation": 117398
    }
  },
  {
    "ph": "f", "id": 117398, "pid": 76337, "tid": -914061504, "ts": 1716454223185804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185805, "dur": 0,
    "args": {
      "External id": 117399,
      "cbid": 205, "correlation": 117399
    }
  },
  {
    "ph": "f", "id": 117399, "pid": 76337, "tid": -914061504, "ts": 1716454223185805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241571, "dur": 5,
    "args": {
      "External id": 117403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117403,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117403, "pid": 5, "tid": 7, "ts": 1716454223241571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185823, "dur": 12,
    "args": {
      "External id": 117403,
      "cbid": 211, "correlation": 117403
    }
  },
  {
    "ph": "s", "id": 117403, "pid": 76337, "tid": -914061504, "ts": 1716454223185823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223241577, "dur": 37,
    "args": {
      "External id": 117405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117405,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 117405, "pid": 5, "tid": 7, "ts": 1716454223241577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185843, "dur": 10,
    "args": {
      "External id": 117405,
      "cbid": 211, "correlation": 117405
    }
  },
  {
    "ph": "s", "id": 117405, "pid": 76337, "tid": -914061504, "ts": 1716454223185843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241615, "dur": 5,
    "args": {
      "External id": 117407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117407,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117407, "pid": 5, "tid": 7, "ts": 1716454223241615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185857, "dur": 5,
    "args": {
      "External id": 117407,
      "cbid": 211, "correlation": 117407
    }
  },
  {
    "ph": "s", "id": 117407, "pid": 76337, "tid": -914061504, "ts": 1716454223185857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241622, "dur": 6,
    "args": {
      "External id": 117413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117413, "pid": 5, "tid": 7, "ts": 1716454223241622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185884, "dur": 8,
    "args": {
      "External id": 117413,
      "cbid": 211, "correlation": 117413
    }
  },
  {
    "ph": "s", "id": 117413, "pid": 76337, "tid": -914061504, "ts": 1716454223185884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223241629, "dur": 21,
    "args": {
      "External id": 117422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117422,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117422, "pid": 5, "tid": 7, "ts": 1716454223241629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223185968, "dur": 22,
    "args": {
      "External id": 117422,
      "cbid": 211, "correlation": 117422
    }
  },
  {
    "ph": "s", "id": 117422, "pid": 76337, "tid": -914061504, "ts": 1716454223185968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241651, "dur": 11,
    "args": {
      "External id": 117444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117444,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 117444, "pid": 5, "tid": 7, "ts": 1716454223241651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186034, "dur": 10,
    "args": {
      "External id": 117444,
      "cbid": 211, "correlation": 117444
    }
  },
  {
    "ph": "s", "id": 117444, "pid": 76337, "tid": -914061504, "ts": 1716454223186034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186125, "dur": 2,
    "args": {
      "External id": 117455,
      "cbid": 251, "correlation": 117455
    }
  },
  {
    "ph": "f", "id": 117455, "pid": 76337, "tid": -914061504, "ts": 1716454223186125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186129, "dur": 0,
    "args": {
      "External id": 117456,
      "cbid": 251, "correlation": 117456
    }
  },
  {
    "ph": "f", "id": 117456, "pid": 76337, "tid": -914061504, "ts": 1716454223186129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223241663, "dur": 55,
    "args": {
      "External id": 117457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117457,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 117457, "pid": 5, "tid": 7, "ts": 1716454223241663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186132, "dur": 13,
    "args": {
      "External id": 117457,
      "cbid": 211, "correlation": 117457
    }
  },
  {
    "ph": "s", "id": 117457, "pid": 76337, "tid": -914061504, "ts": 1716454223186132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186203, "dur": 1,
    "args": {
      "External id": 117468,
      "cbid": 251, "correlation": 117468
    }
  },
  {
    "ph": "f", "id": 117468, "pid": 76337, "tid": -914061504, "ts": 1716454223186203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186206, "dur": 0,
    "args": {
      "External id": 117469,
      "cbid": 251, "correlation": 117469
    }
  },
  {
    "ph": "f", "id": 117469, "pid": 76337, "tid": -914061504, "ts": 1716454223186206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223241719, "dur": 54,
    "args": {
      "External id": 117470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117470,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 117470, "pid": 5, "tid": 7, "ts": 1716454223241719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186209, "dur": 11,
    "args": {
      "External id": 117470,
      "cbid": 211, "correlation": 117470
    }
  },
  {
    "ph": "s", "id": 117470, "pid": 76337, "tid": -914061504, "ts": 1716454223186209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186273, "dur": 1,
    "args": {
      "External id": 117481,
      "cbid": 251, "correlation": 117481
    }
  },
  {
    "ph": "f", "id": 117481, "pid": 76337, "tid": -914061504, "ts": 1716454223186273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186276, "dur": 0,
    "args": {
      "External id": 117482,
      "cbid": 251, "correlation": 117482
    }
  },
  {
    "ph": "f", "id": 117482, "pid": 76337, "tid": -914061504, "ts": 1716454223186276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223241775, "dur": 53,
    "args": {
      "External id": 117483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117483,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 117483, "pid": 5, "tid": 7, "ts": 1716454223241775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186278, "dur": 11,
    "args": {
      "External id": 117483,
      "cbid": 211, "correlation": 117483
    }
  },
  {
    "ph": "s", "id": 117483, "pid": 76337, "tid": -914061504, "ts": 1716454223186278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223241829, "dur": 56,
    "args": {
      "External id": 117508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117508,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117508, "pid": 5, "tid": 7, "ts": 1716454223241829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186364, "dur": 13,
    "args": {
      "External id": 117508,
      "cbid": 211, "correlation": 117508
    }
  },
  {
    "ph": "s", "id": 117508, "pid": 76337, "tid": -914061504, "ts": 1716454223186364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186465, "dur": 1,
    "args": {
      "External id": 117526,
      "cbid": 251, "correlation": 117526
    }
  },
  {
    "ph": "f", "id": 117526, "pid": 76337, "tid": -914061504, "ts": 1716454223186465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223241887, "dur": 65,
    "args": {
      "External id": 117528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117528,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 117528, "pid": 5, "tid": 7, "ts": 1716454223241887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186471, "dur": 13,
    "args": {
      "External id": 117528,
      "cbid": 211, "correlation": 117528
    }
  },
  {
    "ph": "s", "id": 117528, "pid": 76337, "tid": -914061504, "ts": 1716454223186471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223241953, "dur": 6,
    "args": {
      "External id": 117536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117536,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117536, "pid": 5, "tid": 7, "ts": 1716454223241953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186540, "dur": 12,
    "args": {
      "External id": 117536,
      "cbid": 211, "correlation": 117536
    }
  },
  {
    "ph": "s", "id": 117536, "pid": 76337, "tid": -914061504, "ts": 1716454223186540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223241960, "dur": 8,
    "args": {
      "External id": 117544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117544,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117544, "pid": 5, "tid": 7, "ts": 1716454223241960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186579, "dur": 8,
    "args": {
      "External id": 117544,
      "cbid": 211, "correlation": 117544
    }
  },
  {
    "ph": "s", "id": 117544, "pid": 76337, "tid": -914061504, "ts": 1716454223186579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223241969, "dur": 7,
    "args": {
      "External id": 117555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117555, "pid": 5, "tid": 7, "ts": 1716454223241969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186651, "dur": 14,
    "args": {
      "External id": 117555,
      "cbid": 211, "correlation": 117555
    }
  },
  {
    "ph": "s", "id": 117555, "pid": 76337, "tid": -914061504, "ts": 1716454223186651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223241977, "dur": 9,
    "args": {
      "External id": 117577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117577,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 117577, "pid": 5, "tid": 7, "ts": 1716454223241977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186684, "dur": 8,
    "args": {
      "External id": 117577,
      "cbid": 211, "correlation": 117577
    }
  },
  {
    "ph": "s", "id": 117577, "pid": 76337, "tid": -914061504, "ts": 1716454223186684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186770, "dur": 2,
    "args": {
      "External id": 117588,
      "cbid": 251, "correlation": 117588
    }
  },
  {
    "ph": "f", "id": 117588, "pid": 76337, "tid": -914061504, "ts": 1716454223186770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223241989, "dur": 1,
    "args": {
      "External id": 117589,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117589,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 117589, "pid": 5, "tid": 7, "ts": 1716454223241989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186776, "dur": 10,
    "args": {
      "External id": 117589,
      "cbid": 51, "correlation": 117589
    }
  },
  {
    "ph": "s", "id": 117589, "pid": 76337, "tid": -914061504, "ts": 1716454223186776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223241992, "dur": 36,
    "args": {
      "External id": 117590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117590,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 117590, "pid": 5, "tid": 7, "ts": 1716454223241992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186788, "dur": 12,
    "args": {
      "External id": 117590,
      "cbid": 211, "correlation": 117590
    }
  },
  {
    "ph": "s", "id": 117590, "pid": 76337, "tid": -914061504, "ts": 1716454223186788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186858, "dur": 1,
    "args": {
      "External id": 117601,
      "cbid": 251, "correlation": 117601
    }
  },
  {
    "ph": "f", "id": 117601, "pid": 76337, "tid": -914061504, "ts": 1716454223186858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186863, "dur": 0,
    "args": {
      "External id": 117602,
      "cbid": 251, "correlation": 117602
    }
  },
  {
    "ph": "f", "id": 117602, "pid": 76337, "tid": -914061504, "ts": 1716454223186863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223242030, "dur": 11,
    "args": {
      "External id": 117603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117603,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117603, "pid": 5, "tid": 7, "ts": 1716454223242030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186864, "dur": 13,
    "args": {
      "External id": 117603,
      "cbid": 211, "correlation": 117603
    }
  },
  {
    "ph": "s", "id": 117603, "pid": 76337, "tid": -914061504, "ts": 1716454223186864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223242042, "dur": 5,
    "args": {
      "External id": 117605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117605,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117605, "pid": 5, "tid": 7, "ts": 1716454223242042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186879, "dur": 6,
    "args": {
      "External id": 117605,
      "cbid": 211, "correlation": 117605
    }
  },
  {
    "ph": "s", "id": 117605, "pid": 76337, "tid": -914061504, "ts": 1716454223186879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186936, "dur": 1,
    "args": {
      "External id": 117616,
      "cbid": 251, "correlation": 117616
    }
  },
  {
    "ph": "f", "id": 117616, "pid": 76337, "tid": -914061504, "ts": 1716454223186936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186939, "dur": 0,
    "args": {
      "External id": 117617,
      "cbid": 251, "correlation": 117617
    }
  },
  {
    "ph": "f", "id": 117617, "pid": 76337, "tid": -914061504, "ts": 1716454223186939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223242049, "dur": 8,
    "args": {
      "External id": 117618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117618,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117618, "pid": 5, "tid": 7, "ts": 1716454223242049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186941, "dur": 12,
    "args": {
      "External id": 117618,
      "cbid": 211, "correlation": 117618
    }
  },
  {
    "ph": "s", "id": 117618, "pid": 76337, "tid": -914061504, "ts": 1716454223186941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223242059, "dur": 4,
    "args": {
      "External id": 117620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117620,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117620, "pid": 5, "tid": 7, "ts": 1716454223242059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223186954, "dur": 5,
    "args": {
      "External id": 117620,
      "cbid": 211, "correlation": 117620
    }
  },
  {
    "ph": "s", "id": 117620, "pid": 76337, "tid": -914061504, "ts": 1716454223186954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223242063, "dur": 19,
    "args": {
      "External id": 117645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117645,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 117645, "pid": 5, "tid": 7, "ts": 1716454223242063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187040, "dur": 13,
    "args": {
      "External id": 117645,
      "cbid": 211, "correlation": 117645
    }
  },
  {
    "ph": "s", "id": 117645, "pid": 76337, "tid": -914061504, "ts": 1716454223187040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187141, "dur": 2,
    "args": {
      "External id": 117663,
      "cbid": 251, "correlation": 117663
    }
  },
  {
    "ph": "f", "id": 117663, "pid": 76337, "tid": -914061504, "ts": 1716454223187141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223242085, "dur": 1,
    "args": {
      "External id": 117665,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117665,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3125
    }
  },
  {
    "ph": "f", "id": 117665, "pid": 5, "tid": 7, "ts": 1716454223242085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187147, "dur": 10,
    "args": {
      "External id": 117665,
      "cbid": 51, "correlation": 117665
    }
  },
  {
    "ph": "s", "id": 117665, "pid": 76337, "tid": -914061504, "ts": 1716454223187147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223242089, "dur": 37,
    "args": {
      "External id": 117666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117666,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 117666, "pid": 5, "tid": 7, "ts": 1716454223242089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187159, "dur": 14,
    "args": {
      "External id": 117666,
      "cbid": 211, "correlation": 117666
    }
  },
  {
    "ph": "s", "id": 117666, "pid": 76337, "tid": -914061504, "ts": 1716454223187159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223242127, "dur": 4,
    "args": {
      "External id": 117674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117674,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117674, "pid": 5, "tid": 7, "ts": 1716454223242127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187232, "dur": 12,
    "args": {
      "External id": 117674,
      "cbid": 211, "correlation": 117674
    }
  },
  {
    "ph": "s", "id": 117674, "pid": 76337, "tid": -914061504, "ts": 1716454223187232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223242132, "dur": 8,
    "args": {
      "External id": 117682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117682, "pid": 5, "tid": 7, "ts": 1716454223242132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187274, "dur": 9,
    "args": {
      "External id": 117682,
      "cbid": 211, "correlation": 117682
    }
  },
  {
    "ph": "s", "id": 117682, "pid": 76337, "tid": -914061504, "ts": 1716454223187274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223242142, "dur": 8,
    "args": {
      "External id": 117704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117704,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 117704, "pid": 5, "tid": 7, "ts": 1716454223242142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187325, "dur": 10,
    "args": {
      "External id": 117704,
      "cbid": 211, "correlation": 117704
    }
  },
  {
    "ph": "s", "id": 117704, "pid": 76337, "tid": -914061504, "ts": 1716454223187325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187415, "dur": 1,
    "args": {
      "External id": 117720,
      "cbid": 251, "correlation": 117720
    }
  },
  {
    "ph": "f", "id": 117720, "pid": 76337, "tid": -914061504, "ts": 1716454223187415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187420, "dur": 0,
    "args": {
      "External id": 117722,
      "cbid": 251, "correlation": 117722
    }
  },
  {
    "ph": "f", "id": 117722, "pid": 76337, "tid": -914061504, "ts": 1716454223187420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223242151, "dur": 187,
    "args": {
      "External id": 117723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117723,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117723, "pid": 5, "tid": 7, "ts": 1716454223242151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187422, "dur": 13,
    "args": {
      "External id": 117723,
      "cbid": 211, "correlation": 117723
    }
  },
  {
    "ph": "s", "id": 117723, "pid": 76337, "tid": -914061504, "ts": 1716454223187422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223242340, "dur": 22,
    "args": {
      "External id": 117731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117731, "pid": 5, "tid": 7, "ts": 1716454223242340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187488, "dur": 13,
    "args": {
      "External id": 117731,
      "cbid": 211, "correlation": 117731
    }
  },
  {
    "ph": "s", "id": 117731, "pid": 76337, "tid": -914061504, "ts": 1716454223187488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223242364, "dur": 23,
    "args": {
      "External id": 117739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117739, "pid": 5, "tid": 7, "ts": 1716454223242364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187519, "dur": 9,
    "args": {
      "External id": 117739,
      "cbid": 211, "correlation": 117739
    }
  },
  {
    "ph": "s", "id": 117739, "pid": 76337, "tid": -914061504, "ts": 1716454223187519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187602, "dur": 1,
    "args": {
      "External id": 117755,
      "cbid": 251, "correlation": 117755
    }
  },
  {
    "ph": "f", "id": 117755, "pid": 76337, "tid": -914061504, "ts": 1716454223187602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223242389, "dur": 1,
    "args": {
      "External id": 117757,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117757,
      "bytes": 120, "memory bandwidth (GB/s)": 0.07653061224489796
    }
  },
  {
    "ph": "f", "id": 117757, "pid": 5, "tid": 7, "ts": 1716454223242389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187607, "dur": 9,
    "args": {
      "External id": 117757,
      "cbid": 51, "correlation": 117757
    }
  },
  {
    "ph": "s", "id": 117757, "pid": 76337, "tid": -914061504, "ts": 1716454223187607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223242392, "dur": 109,
    "args": {
      "External id": 117758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117758,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 117758, "pid": 5, "tid": 7, "ts": 1716454223242392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187617, "dur": 13,
    "args": {
      "External id": 117758,
      "cbid": 211, "correlation": 117758
    }
  },
  {
    "ph": "s", "id": 117758, "pid": 76337, "tid": -914061504, "ts": 1716454223187617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223242503, "dur": 5,
    "args": {
      "External id": 117766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117766,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117766, "pid": 5, "tid": 7, "ts": 1716454223242503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187660, "dur": 11,
    "args": {
      "External id": 117766,
      "cbid": 211, "correlation": 117766
    }
  },
  {
    "ph": "s", "id": 117766, "pid": 76337, "tid": -914061504, "ts": 1716454223187660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223242509, "dur": 10,
    "args": {
      "External id": 117777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117777, "pid": 5, "tid": 7, "ts": 1716454223242509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187729, "dur": 12,
    "args": {
      "External id": 117777,
      "cbid": 211, "correlation": 117777
    }
  },
  {
    "ph": "s", "id": 117777, "pid": 76337, "tid": -914061504, "ts": 1716454223187729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187795, "dur": 0,
    "args": {
      "External id": 117789,
      "cbid": 317, "correlation": 117789
    }
  },
  {
    "ph": "f", "id": 117789, "pid": 76337, "tid": -914061504, "ts": 1716454223187795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187796, "dur": 0,
    "args": {
      "External id": 117790,
      "cbid": 203, "correlation": 117790
    }
  },
  {
    "ph": "f", "id": 117790, "pid": 76337, "tid": -914061504, "ts": 1716454223187796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187796, "dur": 0,
    "args": {
      "External id": 117791,
      "cbid": 205, "correlation": 117791
    }
  },
  {
    "ph": "f", "id": 117791, "pid": 76337, "tid": -914061504, "ts": 1716454223187796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223242520, "dur": 6,
    "args": {
      "External id": 117795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117795,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117795, "pid": 5, "tid": 7, "ts": 1716454223242520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187811, "dur": 13,
    "args": {
      "External id": 117795,
      "cbid": 211, "correlation": 117795
    }
  },
  {
    "ph": "s", "id": 117795, "pid": 76337, "tid": -914061504, "ts": 1716454223187811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223242527, "dur": 37,
    "args": {
      "External id": 117797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117797,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 117797, "pid": 5, "tid": 7, "ts": 1716454223242527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187831, "dur": 7,
    "args": {
      "External id": 117797,
      "cbid": 211, "correlation": 117797
    }
  },
  {
    "ph": "s", "id": 117797, "pid": 76337, "tid": -914061504, "ts": 1716454223187831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223242565, "dur": 6,
    "args": {
      "External id": 117799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117799,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117799, "pid": 5, "tid": 7, "ts": 1716454223242565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187841, "dur": 5,
    "args": {
      "External id": 117799,
      "cbid": 211, "correlation": 117799
    }
  },
  {
    "ph": "s", "id": 117799, "pid": 76337, "tid": -914061504, "ts": 1716454223187841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223242573, "dur": 7,
    "args": {
      "External id": 117805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117805, "pid": 5, "tid": 7, "ts": 1716454223242573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187868, "dur": 10,
    "args": {
      "External id": 117805,
      "cbid": 211, "correlation": 117805
    }
  },
  {
    "ph": "s", "id": 117805, "pid": 76337, "tid": -914061504, "ts": 1716454223187868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223242581, "dur": 5,
    "args": {
      "External id": 117813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117813,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117813, "pid": 5, "tid": 7, "ts": 1716454223242581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187900, "dur": 8,
    "args": {
      "External id": 117813,
      "cbid": 211, "correlation": 117813
    }
  },
  {
    "ph": "s", "id": 117813, "pid": 76337, "tid": -914061504, "ts": 1716454223187900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223242587, "dur": 11,
    "args": {
      "External id": 117833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117833,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 117833, "pid": 5, "tid": 7, "ts": 1716454223242587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223187972, "dur": 20,
    "args": {
      "External id": 117833,
      "cbid": 211, "correlation": 117833
    }
  },
  {
    "ph": "s", "id": 117833, "pid": 76337, "tid": -914061504, "ts": 1716454223187972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223242600, "dur": 4,
    "args": {
      "External id": 117845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117845,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 117845, "pid": 5, "tid": 7, "ts": 1716454223242600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188002, "dur": 6,
    "args": {
      "External id": 117845,
      "cbid": 211, "correlation": 117845
    }
  },
  {
    "ph": "s", "id": 117845, "pid": 76337, "tid": -914061504, "ts": 1716454223188002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223242605, "dur": 8,
    "args": {
      "External id": 117848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117848,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117848, "pid": 5, "tid": 7, "ts": 1716454223242605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188021, "dur": 7,
    "args": {
      "External id": 117848,
      "cbid": 211, "correlation": 117848
    }
  },
  {
    "ph": "s", "id": 117848, "pid": 76337, "tid": -914061504, "ts": 1716454223188021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223242615, "dur": 5,
    "args": {
      "External id": 117857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117857,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117857, "pid": 5, "tid": 7, "ts": 1716454223242615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188061, "dur": 10,
    "args": {
      "External id": 117857,
      "cbid": 211, "correlation": 117857
    }
  },
  {
    "ph": "s", "id": 117857, "pid": 76337, "tid": -914061504, "ts": 1716454223188061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188112, "dur": 0,
    "args": {
      "External id": 117867,
      "cbid": 317, "correlation": 117867
    }
  },
  {
    "ph": "f", "id": 117867, "pid": 76337, "tid": -914061504, "ts": 1716454223188112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188113, "dur": 0,
    "args": {
      "External id": 117868,
      "cbid": 203, "correlation": 117868
    }
  },
  {
    "ph": "f", "id": 117868, "pid": 76337, "tid": -914061504, "ts": 1716454223188113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188114, "dur": 0,
    "args": {
      "External id": 117869,
      "cbid": 205, "correlation": 117869
    }
  },
  {
    "ph": "f", "id": 117869, "pid": 76337, "tid": -914061504, "ts": 1716454223188114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223242621, "dur": 5,
    "args": {
      "External id": 117873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117873,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117873, "pid": 5, "tid": 7, "ts": 1716454223242621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188128, "dur": 11,
    "args": {
      "External id": 117873,
      "cbid": 211, "correlation": 117873
    }
  },
  {
    "ph": "s", "id": 117873, "pid": 76337, "tid": -914061504, "ts": 1716454223188128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223242628, "dur": 162,
    "args": {
      "External id": 117875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117875, "pid": 5, "tid": 7, "ts": 1716454223242628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188142, "dur": 5,
    "args": {
      "External id": 117875,
      "cbid": 211, "correlation": 117875
    }
  },
  {
    "ph": "s", "id": 117875, "pid": 76337, "tid": -914061504, "ts": 1716454223188142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223242791, "dur": 1,
    "args": {
      "External id": 117877,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117877,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 117877, "pid": 5, "tid": 7, "ts": 1716454223242791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188152, "dur": 6,
    "args": {
      "External id": 117877,
      "cbid": 51, "correlation": 117877
    }
  },
  {
    "ph": "s", "id": 117877, "pid": 76337, "tid": -914061504, "ts": 1716454223188152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223242795, "dur": 268,
    "args": {
      "External id": 117878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117878,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117878, "pid": 5, "tid": 7, "ts": 1716454223242795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188159, "dur": 6,
    "args": {
      "External id": 117878,
      "cbid": 211, "correlation": 117878
    }
  },
  {
    "ph": "s", "id": 117878, "pid": 76337, "tid": -914061504, "ts": 1716454223188159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243065, "dur": 6,
    "args": {
      "External id": 117880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117880, "pid": 5, "tid": 7, "ts": 1716454223243065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188169, "dur": 6,
    "args": {
      "External id": 117880,
      "cbid": 211, "correlation": 117880
    }
  },
  {
    "ph": "s", "id": 117880, "pid": 76337, "tid": -914061504, "ts": 1716454223188169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223243072, "dur": 6,
    "args": {
      "External id": 117886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117886, "pid": 5, "tid": 7, "ts": 1716454223243072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188198, "dur": 8,
    "args": {
      "External id": 117886,
      "cbid": 211, "correlation": 117886
    }
  },
  {
    "ph": "s", "id": 117886, "pid": 76337, "tid": -914061504, "ts": 1716454223188198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223243079, "dur": 4,
    "args": {
      "External id": 117894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117894,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 117894, "pid": 5, "tid": 7, "ts": 1716454223243079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188242, "dur": 9,
    "args": {
      "External id": 117894,
      "cbid": 211, "correlation": 117894
    }
  },
  {
    "ph": "s", "id": 117894, "pid": 76337, "tid": -914061504, "ts": 1716454223188242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188308, "dur": 1,
    "args": {
      "External id": 117910,
      "cbid": 251, "correlation": 117910
    }
  },
  {
    "ph": "f", "id": 117910, "pid": 76337, "tid": -914061504, "ts": 1716454223188308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188313, "dur": 0,
    "args": {
      "External id": 117912,
      "cbid": 251, "correlation": 117912
    }
  },
  {
    "ph": "f", "id": 117912, "pid": 76337, "tid": -914061504, "ts": 1716454223188313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223243085, "dur": 12,
    "args": {
      "External id": 117913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117913,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117913, "pid": 5, "tid": 7, "ts": 1716454223243085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188315, "dur": 11,
    "args": {
      "External id": 117913,
      "cbid": 211, "correlation": 117913
    }
  },
  {
    "ph": "s", "id": 117913, "pid": 76337, "tid": -914061504, "ts": 1716454223188315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223243099, "dur": 5,
    "args": {
      "External id": 117915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117915,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117915, "pid": 5, "tid": 7, "ts": 1716454223243099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188328, "dur": 5,
    "args": {
      "External id": 117915,
      "cbid": 211, "correlation": 117915
    }
  },
  {
    "ph": "s", "id": 117915, "pid": 76337, "tid": -914061504, "ts": 1716454223188328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223243105, "dur": 6,
    "args": {
      "External id": 117925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117925, "pid": 5, "tid": 7, "ts": 1716454223243105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188386, "dur": 12,
    "args": {
      "External id": 117925,
      "cbid": 211, "correlation": 117925
    }
  },
  {
    "ph": "s", "id": 117925, "pid": 76337, "tid": -914061504, "ts": 1716454223188386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243112, "dur": 10,
    "args": {
      "External id": 117945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117945,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 117945, "pid": 5, "tid": 7, "ts": 1716454223243112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188452, "dur": 10,
    "args": {
      "External id": 117945,
      "cbid": 211, "correlation": 117945
    }
  },
  {
    "ph": "s", "id": 117945, "pid": 76337, "tid": -914061504, "ts": 1716454223188452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223243123, "dur": 4,
    "args": {
      "External id": 117957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117957,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 117957, "pid": 5, "tid": 7, "ts": 1716454223243123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188472, "dur": 7,
    "args": {
      "External id": 117957,
      "cbid": 211, "correlation": 117957
    }
  },
  {
    "ph": "s", "id": 117957, "pid": 76337, "tid": -914061504, "ts": 1716454223188472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223243128, "dur": 7,
    "args": {
      "External id": 117960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117960, "pid": 5, "tid": 7, "ts": 1716454223243128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188492, "dur": 6,
    "args": {
      "External id": 117960,
      "cbid": 211, "correlation": 117960
    }
  },
  {
    "ph": "s", "id": 117960, "pid": 76337, "tid": -914061504, "ts": 1716454223188492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223243136, "dur": 5,
    "args": {
      "External id": 117969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117969,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117969, "pid": 5, "tid": 7, "ts": 1716454223243136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188532, "dur": 10,
    "args": {
      "External id": 117969,
      "cbid": 211, "correlation": 117969
    }
  },
  {
    "ph": "s", "id": 117969, "pid": 76337, "tid": -914061504, "ts": 1716454223188532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188595, "dur": 0,
    "args": {
      "External id": 117979,
      "cbid": 317, "correlation": 117979
    }
  },
  {
    "ph": "f", "id": 117979, "pid": 76337, "tid": -914061504, "ts": 1716454223188595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188596, "dur": 0,
    "args": {
      "External id": 117980,
      "cbid": 203, "correlation": 117980
    }
  },
  {
    "ph": "f", "id": 117980, "pid": 76337, "tid": -914061504, "ts": 1716454223188596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188597, "dur": 0,
    "args": {
      "External id": 117981,
      "cbid": 205, "correlation": 117981
    }
  },
  {
    "ph": "f", "id": 117981, "pid": 76337, "tid": -914061504, "ts": 1716454223188597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243142, "dur": 5,
    "args": {
      "External id": 117985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117985,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117985, "pid": 5, "tid": 7, "ts": 1716454223243142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188610, "dur": 12,
    "args": {
      "External id": 117985,
      "cbid": 211, "correlation": 117985
    }
  },
  {
    "ph": "s", "id": 117985, "pid": 76337, "tid": -914061504, "ts": 1716454223188610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243149, "dur": 163,
    "args": {
      "External id": 117987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117987, "pid": 5, "tid": 7, "ts": 1716454223243149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188624, "dur": 6,
    "args": {
      "External id": 117987,
      "cbid": 211, "correlation": 117987
    }
  },
  {
    "ph": "s", "id": 117987, "pid": 76337, "tid": -914061504, "ts": 1716454223188624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223243314, "dur": 1,
    "args": {
      "External id": 117989,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 117989,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 117989, "pid": 5, "tid": 7, "ts": 1716454223243314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188636, "dur": 6,
    "args": {
      "External id": 117989,
      "cbid": 51, "correlation": 117989
    }
  },
  {
    "ph": "s", "id": 117989, "pid": 76337, "tid": -914061504, "ts": 1716454223188636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223243317, "dur": 259,
    "args": {
      "External id": 117990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117990,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 117990, "pid": 5, "tid": 7, "ts": 1716454223243317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188643, "dur": 6,
    "args": {
      "External id": 117990,
      "cbid": 211, "correlation": 117990
    }
  },
  {
    "ph": "s", "id": 117990, "pid": 76337, "tid": -914061504, "ts": 1716454223188643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243578, "dur": 6,
    "args": {
      "External id": 117992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 117992, "pid": 5, "tid": 7, "ts": 1716454223243578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188653, "dur": 5,
    "args": {
      "External id": 117992,
      "cbid": 211, "correlation": 117992
    }
  },
  {
    "ph": "s", "id": 117992, "pid": 76337, "tid": -914061504, "ts": 1716454223188653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223243585, "dur": 6,
    "args": {
      "External id": 117998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 117998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 117998, "pid": 5, "tid": 7, "ts": 1716454223243585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188682, "dur": 9,
    "args": {
      "External id": 117998,
      "cbid": 211, "correlation": 117998
    }
  },
  {
    "ph": "s", "id": 117998, "pid": 76337, "tid": -914061504, "ts": 1716454223188682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223243592, "dur": 5,
    "args": {
      "External id": 118006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118006,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118006, "pid": 5, "tid": 7, "ts": 1716454223243592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188714, "dur": 8,
    "args": {
      "External id": 118006,
      "cbid": 211, "correlation": 118006
    }
  },
  {
    "ph": "s", "id": 118006, "pid": 76337, "tid": -914061504, "ts": 1716454223188714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223243599, "dur": 4,
    "args": {
      "External id": 118014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118014,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118014, "pid": 5, "tid": 7, "ts": 1716454223243599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188744, "dur": 8,
    "args": {
      "External id": 118014,
      "cbid": 211, "correlation": 118014
    }
  },
  {
    "ph": "s", "id": 118014, "pid": 76337, "tid": -914061504, "ts": 1716454223188744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223243605, "dur": 12,
    "args": {
      "External id": 118023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118023,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118023, "pid": 5, "tid": 7, "ts": 1716454223243605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188832, "dur": 13,
    "args": {
      "External id": 118023,
      "cbid": 211, "correlation": 118023
    }
  },
  {
    "ph": "s", "id": 118023, "pid": 76337, "tid": -914061504, "ts": 1716454223188832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243618, "dur": 12,
    "args": {
      "External id": 118043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118043,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118043, "pid": 5, "tid": 7, "ts": 1716454223243618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188900, "dur": 11,
    "args": {
      "External id": 118043,
      "cbid": 211, "correlation": 118043
    }
  },
  {
    "ph": "s", "id": 118043, "pid": 76337, "tid": -914061504, "ts": 1716454223188900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223243631, "dur": 4,
    "args": {
      "External id": 118055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118055,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118055, "pid": 5, "tid": 7, "ts": 1716454223243631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188921, "dur": 6,
    "args": {
      "External id": 118055,
      "cbid": 211, "correlation": 118055
    }
  },
  {
    "ph": "s", "id": 118055, "pid": 76337, "tid": -914061504, "ts": 1716454223188921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223243637, "dur": 10,
    "args": {
      "External id": 118058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118058,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118058, "pid": 5, "tid": 7, "ts": 1716454223243637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188940, "dur": 7,
    "args": {
      "External id": 118058,
      "cbid": 211, "correlation": 118058
    }
  },
  {
    "ph": "s", "id": 118058, "pid": 76337, "tid": -914061504, "ts": 1716454223188940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223243648, "dur": 6,
    "args": {
      "External id": 118067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118067,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118067, "pid": 5, "tid": 7, "ts": 1716454223243648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223188986, "dur": 10,
    "args": {
      "External id": 118067,
      "cbid": 211, "correlation": 118067
    }
  },
  {
    "ph": "s", "id": 118067, "pid": 76337, "tid": -914061504, "ts": 1716454223188986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189041, "dur": 0,
    "args": {
      "External id": 118077,
      "cbid": 317, "correlation": 118077
    }
  },
  {
    "ph": "f", "id": 118077, "pid": 76337, "tid": -914061504, "ts": 1716454223189041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189042, "dur": 0,
    "args": {
      "External id": 118078,
      "cbid": 203, "correlation": 118078
    }
  },
  {
    "ph": "f", "id": 118078, "pid": 76337, "tid": -914061504, "ts": 1716454223189042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189043, "dur": 0,
    "args": {
      "External id": 118079,
      "cbid": 205, "correlation": 118079
    }
  },
  {
    "ph": "f", "id": 118079, "pid": 76337, "tid": -914061504, "ts": 1716454223189043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243656, "dur": 7,
    "args": {
      "External id": 118083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118083,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118083, "pid": 5, "tid": 7, "ts": 1716454223243656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189058, "dur": 11,
    "args": {
      "External id": 118083,
      "cbid": 211, "correlation": 118083
    }
  },
  {
    "ph": "s", "id": 118083, "pid": 76337, "tid": -914061504, "ts": 1716454223189058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223243663, "dur": 321,
    "args": {
      "External id": 118085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118085,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118085, "pid": 5, "tid": 7, "ts": 1716454223243663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189072, "dur": 5,
    "args": {
      "External id": 118085,
      "cbid": 211, "correlation": 118085
    }
  },
  {
    "ph": "s", "id": 118085, "pid": 76337, "tid": -914061504, "ts": 1716454223189072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223243986, "dur": 1,
    "args": {
      "External id": 118087,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118087,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 118087, "pid": 5, "tid": 7, "ts": 1716454223243986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189083, "dur": 8,
    "args": {
      "External id": 118087,
      "cbid": 51, "correlation": 118087
    }
  },
  {
    "ph": "s", "id": 118087, "pid": 76337, "tid": -914061504, "ts": 1716454223189083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223243990, "dur": 494,
    "args": {
      "External id": 118088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118088,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118088, "pid": 5, "tid": 7, "ts": 1716454223243990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189092, "dur": 6,
    "args": {
      "External id": 118088,
      "cbid": 211, "correlation": 118088
    }
  },
  {
    "ph": "s", "id": 118088, "pid": 76337, "tid": -914061504, "ts": 1716454223189092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223244486, "dur": 5,
    "args": {
      "External id": 118090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118090, "pid": 5, "tid": 7, "ts": 1716454223244486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189102, "dur": 5,
    "args": {
      "External id": 118090,
      "cbid": 211, "correlation": 118090
    }
  },
  {
    "ph": "s", "id": 118090, "pid": 76337, "tid": -914061504, "ts": 1716454223189102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223244492, "dur": 6,
    "args": {
      "External id": 118096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118096, "pid": 5, "tid": 7, "ts": 1716454223244492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189132, "dur": 9,
    "args": {
      "External id": 118096,
      "cbid": 211, "correlation": 118096
    }
  },
  {
    "ph": "s", "id": 118096, "pid": 76337, "tid": -914061504, "ts": 1716454223189132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223244500, "dur": 4,
    "args": {
      "External id": 118104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118104,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 118104, "pid": 5, "tid": 7, "ts": 1716454223244500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189175, "dur": 10,
    "args": {
      "External id": 118104,
      "cbid": 211, "correlation": 118104
    }
  },
  {
    "ph": "s", "id": 118104, "pid": 76337, "tid": -914061504, "ts": 1716454223189175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189240, "dur": 1,
    "args": {
      "External id": 118120,
      "cbid": 251, "correlation": 118120
    }
  },
  {
    "ph": "f", "id": 118120, "pid": 76337, "tid": -914061504, "ts": 1716454223189240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189245, "dur": 0,
    "args": {
      "External id": 118122,
      "cbid": 251, "correlation": 118122
    }
  },
  {
    "ph": "f", "id": 118122, "pid": 76337, "tid": -914061504, "ts": 1716454223189245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223244505, "dur": 12,
    "args": {
      "External id": 118123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118123,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118123, "pid": 5, "tid": 7, "ts": 1716454223244505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189247, "dur": 11,
    "args": {
      "External id": 118123,
      "cbid": 211, "correlation": 118123
    }
  },
  {
    "ph": "s", "id": 118123, "pid": 76337, "tid": -914061504, "ts": 1716454223189247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223244518, "dur": 5,
    "args": {
      "External id": 118125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118125,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118125, "pid": 5, "tid": 7, "ts": 1716454223244518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189259, "dur": 5,
    "args": {
      "External id": 118125,
      "cbid": 211, "correlation": 118125
    }
  },
  {
    "ph": "s", "id": 118125, "pid": 76337, "tid": -914061504, "ts": 1716454223189259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223244524, "dur": 6,
    "args": {
      "External id": 118135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118135, "pid": 5, "tid": 7, "ts": 1716454223244524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189316, "dur": 12,
    "args": {
      "External id": 118135,
      "cbid": 211, "correlation": 118135
    }
  },
  {
    "ph": "s", "id": 118135, "pid": 76337, "tid": -914061504, "ts": 1716454223189316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223244531, "dur": 10,
    "args": {
      "External id": 118155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118155,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118155, "pid": 5, "tid": 7, "ts": 1716454223244531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189381, "dur": 11,
    "args": {
      "External id": 118155,
      "cbid": 211, "correlation": 118155
    }
  },
  {
    "ph": "s", "id": 118155, "pid": 76337, "tid": -914061504, "ts": 1716454223189381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223244542, "dur": 4,
    "args": {
      "External id": 118167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118167,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 118167, "pid": 5, "tid": 7, "ts": 1716454223244542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189402, "dur": 6,
    "args": {
      "External id": 118167,
      "cbid": 211, "correlation": 118167
    }
  },
  {
    "ph": "s", "id": 118167, "pid": 76337, "tid": -914061504, "ts": 1716454223189402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223244547, "dur": 7,
    "args": {
      "External id": 118170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118170,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118170, "pid": 5, "tid": 7, "ts": 1716454223244547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189421, "dur": 6,
    "args": {
      "External id": 118170,
      "cbid": 211, "correlation": 118170
    }
  },
  {
    "ph": "s", "id": 118170, "pid": 76337, "tid": -914061504, "ts": 1716454223189421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223244555, "dur": 5,
    "args": {
      "External id": 118179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118179,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118179, "pid": 5, "tid": 7, "ts": 1716454223244555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189461, "dur": 10,
    "args": {
      "External id": 118179,
      "cbid": 211, "correlation": 118179
    }
  },
  {
    "ph": "s", "id": 118179, "pid": 76337, "tid": -914061504, "ts": 1716454223189461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189523, "dur": 0,
    "args": {
      "External id": 118189,
      "cbid": 317, "correlation": 118189
    }
  },
  {
    "ph": "f", "id": 118189, "pid": 76337, "tid": -914061504, "ts": 1716454223189523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189524, "dur": 0,
    "args": {
      "External id": 118190,
      "cbid": 203, "correlation": 118190
    }
  },
  {
    "ph": "f", "id": 118190, "pid": 76337, "tid": -914061504, "ts": 1716454223189524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189525, "dur": 0,
    "args": {
      "External id": 118191,
      "cbid": 205, "correlation": 118191
    }
  },
  {
    "ph": "f", "id": 118191, "pid": 76337, "tid": -914061504, "ts": 1716454223189525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223244561, "dur": 5,
    "args": {
      "External id": 118195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118195,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118195, "pid": 5, "tid": 7, "ts": 1716454223244561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189539, "dur": 12,
    "args": {
      "External id": 118195,
      "cbid": 211, "correlation": 118195
    }
  },
  {
    "ph": "s", "id": 118195, "pid": 76337, "tid": -914061504, "ts": 1716454223189539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223244567, "dur": 163,
    "args": {
      "External id": 118197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118197,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118197, "pid": 5, "tid": 7, "ts": 1716454223244567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189553, "dur": 5,
    "args": {
      "External id": 118197,
      "cbid": 211, "correlation": 118197
    }
  },
  {
    "ph": "s", "id": 118197, "pid": 76337, "tid": -914061504, "ts": 1716454223189553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223244733, "dur": 1,
    "args": {
      "External id": 118199,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118199,
      "bytes": 240, "memory bandwidth (GB/s)": 0.14705882352941177
    }
  },
  {
    "ph": "f", "id": 118199, "pid": 5, "tid": 7, "ts": 1716454223244733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189564, "dur": 6,
    "args": {
      "External id": 118199,
      "cbid": 51, "correlation": 118199
    }
  },
  {
    "ph": "s", "id": 118199, "pid": 76337, "tid": -914061504, "ts": 1716454223189564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223244737, "dur": 258,
    "args": {
      "External id": 118200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118200,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118200, "pid": 5, "tid": 7, "ts": 1716454223244737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189571, "dur": 6,
    "args": {
      "External id": 118200,
      "cbid": 211, "correlation": 118200
    }
  },
  {
    "ph": "s", "id": 118200, "pid": 76337, "tid": -914061504, "ts": 1716454223189571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223244996, "dur": 6,
    "args": {
      "External id": 118202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118202, "pid": 5, "tid": 7, "ts": 1716454223244996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189581, "dur": 6,
    "args": {
      "External id": 118202,
      "cbid": 211, "correlation": 118202
    }
  },
  {
    "ph": "s", "id": 118202, "pid": 76337, "tid": -914061504, "ts": 1716454223189581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223245003, "dur": 6,
    "args": {
      "External id": 118208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118208,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118208, "pid": 5, "tid": 7, "ts": 1716454223245003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189610, "dur": 9,
    "args": {
      "External id": 118208,
      "cbid": 211, "correlation": 118208
    }
  },
  {
    "ph": "s", "id": 118208, "pid": 76337, "tid": -914061504, "ts": 1716454223189610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189669, "dur": 0,
    "args": {
      "External id": 118218,
      "cbid": 317, "correlation": 118218
    }
  },
  {
    "ph": "f", "id": 118218, "pid": 76337, "tid": -914061504, "ts": 1716454223189669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189670, "dur": 0,
    "args": {
      "External id": 118219,
      "cbid": 203, "correlation": 118219
    }
  },
  {
    "ph": "f", "id": 118219, "pid": 76337, "tid": -914061504, "ts": 1716454223189670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189671, "dur": 0,
    "args": {
      "External id": 118220,
      "cbid": 205, "correlation": 118220
    }
  },
  {
    "ph": "f", "id": 118220, "pid": 76337, "tid": -914061504, "ts": 1716454223189671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223245011, "dur": 7,
    "args": {
      "External id": 118224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118224, "pid": 5, "tid": 7, "ts": 1716454223245011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189686, "dur": 12,
    "args": {
      "External id": 118224,
      "cbid": 211, "correlation": 118224
    }
  },
  {
    "ph": "s", "id": 118224, "pid": 76337, "tid": -914061504, "ts": 1716454223189686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223245020, "dur": 3,
    "args": {
      "External id": 118226,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118226,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.499531396438613
    }
  },
  {
    "ph": "f", "id": 118226, "pid": 5, "tid": 7, "ts": 1716454223245020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189704, "dur": 14,
    "args": {
      "External id": 118226,
      "cbid": 51, "correlation": 118226
    }
  },
  {
    "ph": "s", "id": 118226, "pid": 76337, "tid": -914061504, "ts": 1716454223189704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223245024, "dur": 95,
    "args": {
      "External id": 118227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118227,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 118227, "pid": 5, "tid": 7, "ts": 1716454223245024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189720, "dur": 6,
    "args": {
      "External id": 118227,
      "cbid": 211, "correlation": 118227
    }
  },
  {
    "ph": "s", "id": 118227, "pid": 76337, "tid": -914061504, "ts": 1716454223189720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223245121, "dur": 5,
    "args": {
      "External id": 118229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118229, "pid": 5, "tid": 7, "ts": 1716454223245121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189730, "dur": 5,
    "args": {
      "External id": 118229,
      "cbid": 211, "correlation": 118229
    }
  },
  {
    "ph": "s", "id": 118229, "pid": 76337, "tid": -914061504, "ts": 1716454223189730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223245127, "dur": 6,
    "args": {
      "External id": 118235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118235,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118235, "pid": 5, "tid": 7, "ts": 1716454223245127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189757, "dur": 8,
    "args": {
      "External id": 118235,
      "cbid": 211, "correlation": 118235
    }
  },
  {
    "ph": "s", "id": 118235, "pid": 76337, "tid": -914061504, "ts": 1716454223189757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223245135, "dur": 5,
    "args": {
      "External id": 118243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118243,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118243, "pid": 5, "tid": 7, "ts": 1716454223245135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189786, "dur": 8,
    "args": {
      "External id": 118243,
      "cbid": 211, "correlation": 118243
    }
  },
  {
    "ph": "s", "id": 118243, "pid": 76337, "tid": -914061504, "ts": 1716454223189786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223245142, "dur": 4,
    "args": {
      "External id": 118251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118251,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118251, "pid": 5, "tid": 7, "ts": 1716454223245142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189814, "dur": 8,
    "args": {
      "External id": 118251,
      "cbid": 211, "correlation": 118251
    }
  },
  {
    "ph": "s", "id": 118251, "pid": 76337, "tid": -914061504, "ts": 1716454223189814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223245147, "dur": 11,
    "args": {
      "External id": 118260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118260,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118260, "pid": 5, "tid": 7, "ts": 1716454223245147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189858, "dur": 10,
    "args": {
      "External id": 118260,
      "cbid": 211, "correlation": 118260
    }
  },
  {
    "ph": "s", "id": 118260, "pid": 76337, "tid": -914061504, "ts": 1716454223189858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223245160, "dur": 12,
    "args": {
      "External id": 118280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118280,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118280, "pid": 5, "tid": 7, "ts": 1716454223245160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189928, "dur": 12,
    "args": {
      "External id": 118280,
      "cbid": 211, "correlation": 118280
    }
  },
  {
    "ph": "s", "id": 118280, "pid": 76337, "tid": -914061504, "ts": 1716454223189928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223245173, "dur": 4,
    "args": {
      "External id": 118292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118292,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118292, "pid": 5, "tid": 7, "ts": 1716454223245173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189950, "dur": 6,
    "args": {
      "External id": 118292,
      "cbid": 211, "correlation": 118292
    }
  },
  {
    "ph": "s", "id": 118292, "pid": 76337, "tid": -914061504, "ts": 1716454223189950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223245179, "dur": 10,
    "args": {
      "External id": 118295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118295,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118295, "pid": 5, "tid": 7, "ts": 1716454223245179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223189967, "dur": 15,
    "args": {
      "External id": 118295,
      "cbid": 211, "correlation": 118295
    }
  },
  {
    "ph": "s", "id": 118295, "pid": 76337, "tid": -914061504, "ts": 1716454223189967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223245190, "dur": 7,
    "args": {
      "External id": 118304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118304,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118304, "pid": 5, "tid": 7, "ts": 1716454223245190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190016, "dur": 10,
    "args": {
      "External id": 118304,
      "cbid": 211, "correlation": 118304
    }
  },
  {
    "ph": "s", "id": 118304, "pid": 76337, "tid": -914061504, "ts": 1716454223190016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190070, "dur": 0,
    "args": {
      "External id": 118314,
      "cbid": 317, "correlation": 118314
    }
  },
  {
    "ph": "f", "id": 118314, "pid": 76337, "tid": -914061504, "ts": 1716454223190070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190070, "dur": 0,
    "args": {
      "External id": 118315,
      "cbid": 203, "correlation": 118315
    }
  },
  {
    "ph": "f", "id": 118315, "pid": 76337, "tid": -914061504, "ts": 1716454223190070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190071, "dur": 0,
    "args": {
      "External id": 118316,
      "cbid": 205, "correlation": 118316
    }
  },
  {
    "ph": "f", "id": 118316, "pid": 76337, "tid": -914061504, "ts": 1716454223190071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223245198, "dur": 6,
    "args": {
      "External id": 118320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118320,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118320, "pid": 5, "tid": 7, "ts": 1716454223245198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190084, "dur": 12,
    "args": {
      "External id": 118320,
      "cbid": 211, "correlation": 118320
    }
  },
  {
    "ph": "s", "id": 118320, "pid": 76337, "tid": -914061504, "ts": 1716454223190084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223245206, "dur": 320,
    "args": {
      "External id": 118322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118322,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118322, "pid": 5, "tid": 7, "ts": 1716454223245206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190099, "dur": 5,
    "args": {
      "External id": 118322,
      "cbid": 211, "correlation": 118322
    }
  },
  {
    "ph": "s", "id": 118322, "pid": 76337, "tid": -914061504, "ts": 1716454223190099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223245528, "dur": 1,
    "args": {
      "External id": 118324,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118324,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 118324, "pid": 5, "tid": 7, "ts": 1716454223245528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190110, "dur": 6,
    "args": {
      "External id": 118324,
      "cbid": 51, "correlation": 118324
    }
  },
  {
    "ph": "s", "id": 118324, "pid": 76337, "tid": -914061504, "ts": 1716454223190110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223245532, "dur": 497,
    "args": {
      "External id": 118325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118325,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118325, "pid": 5, "tid": 7, "ts": 1716454223245532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190118, "dur": 6,
    "args": {
      "External id": 118325,
      "cbid": 211, "correlation": 118325
    }
  },
  {
    "ph": "s", "id": 118325, "pid": 76337, "tid": -914061504, "ts": 1716454223190118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246031, "dur": 5,
    "args": {
      "External id": 118327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118327,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118327, "pid": 5, "tid": 7, "ts": 1716454223246031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190128, "dur": 5,
    "args": {
      "External id": 118327,
      "cbid": 211, "correlation": 118327
    }
  },
  {
    "ph": "s", "id": 118327, "pid": 76337, "tid": -914061504, "ts": 1716454223190128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223246037, "dur": 7,
    "args": {
      "External id": 118333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118333, "pid": 5, "tid": 7, "ts": 1716454223246037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190157, "dur": 9,
    "args": {
      "External id": 118333,
      "cbid": 211, "correlation": 118333
    }
  },
  {
    "ph": "s", "id": 118333, "pid": 76337, "tid": -914061504, "ts": 1716454223190157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223246045, "dur": 4,
    "args": {
      "External id": 118341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118341,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 118341, "pid": 5, "tid": 7, "ts": 1716454223246045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190201, "dur": 9,
    "args": {
      "External id": 118341,
      "cbid": 211, "correlation": 118341
    }
  },
  {
    "ph": "s", "id": 118341, "pid": 76337, "tid": -914061504, "ts": 1716454223190201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190262, "dur": 1,
    "args": {
      "External id": 118357,
      "cbid": 251, "correlation": 118357
    }
  },
  {
    "ph": "f", "id": 118357, "pid": 76337, "tid": -914061504, "ts": 1716454223190262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190268, "dur": 0,
    "args": {
      "External id": 118359,
      "cbid": 251, "correlation": 118359
    }
  },
  {
    "ph": "f", "id": 118359, "pid": 76337, "tid": -914061504, "ts": 1716454223190268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223246050, "dur": 13,
    "args": {
      "External id": 118360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118360,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118360, "pid": 5, "tid": 7, "ts": 1716454223246050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190270, "dur": 11,
    "args": {
      "External id": 118360,
      "cbid": 211, "correlation": 118360
    }
  },
  {
    "ph": "s", "id": 118360, "pid": 76337, "tid": -914061504, "ts": 1716454223190270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223246065, "dur": 5,
    "args": {
      "External id": 118362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118362,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118362, "pid": 5, "tid": 7, "ts": 1716454223246065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190282, "dur": 5,
    "args": {
      "External id": 118362,
      "cbid": 211, "correlation": 118362
    }
  },
  {
    "ph": "s", "id": 118362, "pid": 76337, "tid": -914061504, "ts": 1716454223190282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223246071, "dur": 6,
    "args": {
      "External id": 118372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118372,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118372, "pid": 5, "tid": 7, "ts": 1716454223246071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190338, "dur": 13,
    "args": {
      "External id": 118372,
      "cbid": 211, "correlation": 118372
    }
  },
  {
    "ph": "s", "id": 118372, "pid": 76337, "tid": -914061504, "ts": 1716454223190338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246078, "dur": 10,
    "args": {
      "External id": 118392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118392,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118392, "pid": 5, "tid": 7, "ts": 1716454223246078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190406, "dur": 11,
    "args": {
      "External id": 118392,
      "cbid": 211, "correlation": 118392
    }
  },
  {
    "ph": "s", "id": 118392, "pid": 76337, "tid": -914061504, "ts": 1716454223190406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223246089, "dur": 4,
    "args": {
      "External id": 118404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118404,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 118404, "pid": 5, "tid": 7, "ts": 1716454223246089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190426, "dur": 6,
    "args": {
      "External id": 118404,
      "cbid": 211, "correlation": 118404
    }
  },
  {
    "ph": "s", "id": 118404, "pid": 76337, "tid": -914061504, "ts": 1716454223190426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223246095, "dur": 7,
    "args": {
      "External id": 118407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118407,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118407, "pid": 5, "tid": 7, "ts": 1716454223246095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190445, "dur": 7,
    "args": {
      "External id": 118407,
      "cbid": 211, "correlation": 118407
    }
  },
  {
    "ph": "s", "id": 118407, "pid": 76337, "tid": -914061504, "ts": 1716454223190445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223246103, "dur": 5,
    "args": {
      "External id": 118416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118416,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118416, "pid": 5, "tid": 7, "ts": 1716454223246103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190484, "dur": 10,
    "args": {
      "External id": 118416,
      "cbid": 211, "correlation": 118416
    }
  },
  {
    "ph": "s", "id": 118416, "pid": 76337, "tid": -914061504, "ts": 1716454223190484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190547, "dur": 0,
    "args": {
      "External id": 118426,
      "cbid": 317, "correlation": 118426
    }
  },
  {
    "ph": "f", "id": 118426, "pid": 76337, "tid": -914061504, "ts": 1716454223190547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190548, "dur": 0,
    "args": {
      "External id": 118427,
      "cbid": 203, "correlation": 118427
    }
  },
  {
    "ph": "f", "id": 118427, "pid": 76337, "tid": -914061504, "ts": 1716454223190548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190549, "dur": 0,
    "args": {
      "External id": 118428,
      "cbid": 205, "correlation": 118428
    }
  },
  {
    "ph": "f", "id": 118428, "pid": 76337, "tid": -914061504, "ts": 1716454223190549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246109, "dur": 5,
    "args": {
      "External id": 118432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118432, "pid": 5, "tid": 7, "ts": 1716454223246109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190563, "dur": 12,
    "args": {
      "External id": 118432,
      "cbid": 211, "correlation": 118432
    }
  },
  {
    "ph": "s", "id": 118432, "pid": 76337, "tid": -914061504, "ts": 1716454223190563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246115, "dur": 161,
    "args": {
      "External id": 118434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118434,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118434, "pid": 5, "tid": 7, "ts": 1716454223246115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190578, "dur": 5,
    "args": {
      "External id": 118434,
      "cbid": 211, "correlation": 118434
    }
  },
  {
    "ph": "s", "id": 118434, "pid": 76337, "tid": -914061504, "ts": 1716454223190578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223246279, "dur": 1,
    "args": {
      "External id": 118436,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118436,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 118436, "pid": 5, "tid": 7, "ts": 1716454223246279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190589, "dur": 6,
    "args": {
      "External id": 118436,
      "cbid": 51, "correlation": 118436
    }
  },
  {
    "ph": "s", "id": 118436, "pid": 76337, "tid": -914061504, "ts": 1716454223190589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223246282, "dur": 259,
    "args": {
      "External id": 118437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118437,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118437, "pid": 5, "tid": 7, "ts": 1716454223246282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190596, "dur": 7,
    "args": {
      "External id": 118437,
      "cbid": 211, "correlation": 118437
    }
  },
  {
    "ph": "s", "id": 118437, "pid": 76337, "tid": -914061504, "ts": 1716454223190596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246542, "dur": 6,
    "args": {
      "External id": 118439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118439,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118439, "pid": 5, "tid": 7, "ts": 1716454223246542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190607, "dur": 5,
    "args": {
      "External id": 118439,
      "cbid": 211, "correlation": 118439
    }
  },
  {
    "ph": "s", "id": 118439, "pid": 76337, "tid": -914061504, "ts": 1716454223190607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223246550, "dur": 6,
    "args": {
      "External id": 118445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118445, "pid": 5, "tid": 7, "ts": 1716454223246550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190635, "dur": 8,
    "args": {
      "External id": 118445,
      "cbid": 211, "correlation": 118445
    }
  },
  {
    "ph": "s", "id": 118445, "pid": 76337, "tid": -914061504, "ts": 1716454223190635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190693, "dur": 0,
    "args": {
      "External id": 118455,
      "cbid": 317, "correlation": 118455
    }
  },
  {
    "ph": "f", "id": 118455, "pid": 76337, "tid": -914061504, "ts": 1716454223190693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190694, "dur": 0,
    "args": {
      "External id": 118456,
      "cbid": 203, "correlation": 118456
    }
  },
  {
    "ph": "f", "id": 118456, "pid": 76337, "tid": -914061504, "ts": 1716454223190694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190694, "dur": 0,
    "args": {
      "External id": 118457,
      "cbid": 205, "correlation": 118457
    }
  },
  {
    "ph": "f", "id": 118457, "pid": 76337, "tid": -914061504, "ts": 1716454223190694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246557, "dur": 8,
    "args": {
      "External id": 118461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118461,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118461, "pid": 5, "tid": 7, "ts": 1716454223246557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190707, "dur": 12,
    "args": {
      "External id": 118461,
      "cbid": 211, "correlation": 118461
    }
  },
  {
    "ph": "s", "id": 118461, "pid": 76337, "tid": -914061504, "ts": 1716454223190707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223246566, "dur": 3,
    "args": {
      "External id": 118463,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118463,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 118463, "pid": 5, "tid": 7, "ts": 1716454223246566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190724, "dur": 9,
    "args": {
      "External id": 118463,
      "cbid": 51, "correlation": 118463
    }
  },
  {
    "ph": "s", "id": 118463, "pid": 76337, "tid": -914061504, "ts": 1716454223190724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223246570, "dur": 95,
    "args": {
      "External id": 118464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118464,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 118464, "pid": 5, "tid": 7, "ts": 1716454223246570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190734, "dur": 5,
    "args": {
      "External id": 118464,
      "cbid": 211, "correlation": 118464
    }
  },
  {
    "ph": "s", "id": 118464, "pid": 76337, "tid": -914061504, "ts": 1716454223190734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246666, "dur": 5,
    "args": {
      "External id": 118466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118466,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118466, "pid": 5, "tid": 7, "ts": 1716454223246666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190743, "dur": 5,
    "args": {
      "External id": 118466,
      "cbid": 211, "correlation": 118466
    }
  },
  {
    "ph": "s", "id": 118466, "pid": 76337, "tid": -914061504, "ts": 1716454223190743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223246673, "dur": 6,
    "args": {
      "External id": 118472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118472,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118472, "pid": 5, "tid": 7, "ts": 1716454223246673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190769, "dur": 8,
    "args": {
      "External id": 118472,
      "cbid": 211, "correlation": 118472
    }
  },
  {
    "ph": "s", "id": 118472, "pid": 76337, "tid": -914061504, "ts": 1716454223190769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223246681, "dur": 5,
    "args": {
      "External id": 118480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118480,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118480, "pid": 5, "tid": 7, "ts": 1716454223246681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190798, "dur": 9,
    "args": {
      "External id": 118480,
      "cbid": 211, "correlation": 118480
    }
  },
  {
    "ph": "s", "id": 118480, "pid": 76337, "tid": -914061504, "ts": 1716454223190798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223246687, "dur": 4,
    "args": {
      "External id": 118488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118488,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118488, "pid": 5, "tid": 7, "ts": 1716454223246687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190827, "dur": 8,
    "args": {
      "External id": 118488,
      "cbid": 211, "correlation": 118488
    }
  },
  {
    "ph": "s", "id": 118488, "pid": 76337, "tid": -914061504, "ts": 1716454223190827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223246693, "dur": 11,
    "args": {
      "External id": 118497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118497,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118497, "pid": 5, "tid": 7, "ts": 1716454223246693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190871, "dur": 10,
    "args": {
      "External id": 118497,
      "cbid": 211, "correlation": 118497
    }
  },
  {
    "ph": "s", "id": 118497, "pid": 76337, "tid": -914061504, "ts": 1716454223190871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246705, "dur": 13,
    "args": {
      "External id": 118517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118517,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118517, "pid": 5, "tid": 7, "ts": 1716454223246705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190940, "dur": 11,
    "args": {
      "External id": 118517,
      "cbid": 211, "correlation": 118517
    }
  },
  {
    "ph": "s", "id": 118517, "pid": 76337, "tid": -914061504, "ts": 1716454223190940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223246719, "dur": 4,
    "args": {
      "External id": 118529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118529,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118529, "pid": 5, "tid": 7, "ts": 1716454223246719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190962, "dur": 6,
    "args": {
      "External id": 118529,
      "cbid": 211, "correlation": 118529
    }
  },
  {
    "ph": "s", "id": 118529, "pid": 76337, "tid": -914061504, "ts": 1716454223190962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223246724, "dur": 10,
    "args": {
      "External id": 118532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118532, "pid": 5, "tid": 7, "ts": 1716454223246724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223190987, "dur": 7,
    "args": {
      "External id": 118532,
      "cbid": 211, "correlation": 118532
    }
  },
  {
    "ph": "s", "id": 118532, "pid": 76337, "tid": -914061504, "ts": 1716454223190987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223246736, "dur": 7,
    "args": {
      "External id": 118541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118541,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118541, "pid": 5, "tid": 7, "ts": 1716454223246736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191027, "dur": 10,
    "args": {
      "External id": 118541,
      "cbid": 211, "correlation": 118541
    }
  },
  {
    "ph": "s", "id": 118541, "pid": 76337, "tid": -914061504, "ts": 1716454223191027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191079, "dur": 0,
    "args": {
      "External id": 118551,
      "cbid": 317, "correlation": 118551
    }
  },
  {
    "ph": "f", "id": 118551, "pid": 76337, "tid": -914061504, "ts": 1716454223191079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191080, "dur": 0,
    "args": {
      "External id": 118552,
      "cbid": 203, "correlation": 118552
    }
  },
  {
    "ph": "f", "id": 118552, "pid": 76337, "tid": -914061504, "ts": 1716454223191080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191081, "dur": 0,
    "args": {
      "External id": 118553,
      "cbid": 205, "correlation": 118553
    }
  },
  {
    "ph": "f", "id": 118553, "pid": 76337, "tid": -914061504, "ts": 1716454223191081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246744, "dur": 7,
    "args": {
      "External id": 118557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118557,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118557, "pid": 5, "tid": 7, "ts": 1716454223246744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191094, "dur": 11,
    "args": {
      "External id": 118557,
      "cbid": 211, "correlation": 118557
    }
  },
  {
    "ph": "s", "id": 118557, "pid": 76337, "tid": -914061504, "ts": 1716454223191094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223246751, "dur": 320,
    "args": {
      "External id": 118559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118559, "pid": 5, "tid": 7, "ts": 1716454223246751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191108, "dur": 6,
    "args": {
      "External id": 118559,
      "cbid": 211, "correlation": 118559
    }
  },
  {
    "ph": "s", "id": 118559, "pid": 76337, "tid": -914061504, "ts": 1716454223191108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223247073, "dur": 1,
    "args": {
      "External id": 118561,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118561,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 118561, "pid": 5, "tid": 7, "ts": 1716454223247073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191119, "dur": 6,
    "args": {
      "External id": 118561,
      "cbid": 51, "correlation": 118561
    }
  },
  {
    "ph": "s", "id": 118561, "pid": 76337, "tid": -914061504, "ts": 1716454223191119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223247077, "dur": 495,
    "args": {
      "External id": 118562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118562,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118562, "pid": 5, "tid": 7, "ts": 1716454223247077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191126, "dur": 6,
    "args": {
      "External id": 118562,
      "cbid": 211, "correlation": 118562
    }
  },
  {
    "ph": "s", "id": 118562, "pid": 76337, "tid": -914061504, "ts": 1716454223191126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223247573, "dur": 5,
    "args": {
      "External id": 118564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118564,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118564, "pid": 5, "tid": 7, "ts": 1716454223247573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191136, "dur": 5,
    "args": {
      "External id": 118564,
      "cbid": 211, "correlation": 118564
    }
  },
  {
    "ph": "s", "id": 118564, "pid": 76337, "tid": -914061504, "ts": 1716454223191136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223247580, "dur": 6,
    "args": {
      "External id": 118570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118570, "pid": 5, "tid": 7, "ts": 1716454223247580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191165, "dur": 9,
    "args": {
      "External id": 118570,
      "cbid": 211, "correlation": 118570
    }
  },
  {
    "ph": "s", "id": 118570, "pid": 76337, "tid": -914061504, "ts": 1716454223191165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223247588, "dur": 4,
    "args": {
      "External id": 118578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118578,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 118578, "pid": 5, "tid": 7, "ts": 1716454223247588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191208, "dur": 10,
    "args": {
      "External id": 118578,
      "cbid": 211, "correlation": 118578
    }
  },
  {
    "ph": "s", "id": 118578, "pid": 76337, "tid": -914061504, "ts": 1716454223191208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191270, "dur": 1,
    "args": {
      "External id": 118594,
      "cbid": 251, "correlation": 118594
    }
  },
  {
    "ph": "f", "id": 118594, "pid": 76337, "tid": -914061504, "ts": 1716454223191270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191275, "dur": 0,
    "args": {
      "External id": 118596,
      "cbid": 251, "correlation": 118596
    }
  },
  {
    "ph": "f", "id": 118596, "pid": 76337, "tid": -914061504, "ts": 1716454223191275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223247593, "dur": 13,
    "args": {
      "External id": 118597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118597,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118597, "pid": 5, "tid": 7, "ts": 1716454223247593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191277, "dur": 11,
    "args": {
      "External id": 118597,
      "cbid": 211, "correlation": 118597
    }
  },
  {
    "ph": "s", "id": 118597, "pid": 76337, "tid": -914061504, "ts": 1716454223191277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223247607, "dur": 5,
    "args": {
      "External id": 118599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118599,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118599, "pid": 5, "tid": 7, "ts": 1716454223247607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191290, "dur": 5,
    "args": {
      "External id": 118599,
      "cbid": 211, "correlation": 118599
    }
  },
  {
    "ph": "s", "id": 118599, "pid": 76337, "tid": -914061504, "ts": 1716454223191290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223247613, "dur": 6,
    "args": {
      "External id": 118609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118609, "pid": 5, "tid": 7, "ts": 1716454223247613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191347, "dur": 13,
    "args": {
      "External id": 118609,
      "cbid": 211, "correlation": 118609
    }
  },
  {
    "ph": "s", "id": 118609, "pid": 76337, "tid": -914061504, "ts": 1716454223191347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223247620, "dur": 9,
    "args": {
      "External id": 118629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118629,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118629, "pid": 5, "tid": 7, "ts": 1716454223247620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191413, "dur": 11,
    "args": {
      "External id": 118629,
      "cbid": 211, "correlation": 118629
    }
  },
  {
    "ph": "s", "id": 118629, "pid": 76337, "tid": -914061504, "ts": 1716454223191413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223247631, "dur": 4,
    "args": {
      "External id": 118641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118641,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 118641, "pid": 5, "tid": 7, "ts": 1716454223247631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191433, "dur": 6,
    "args": {
      "External id": 118641,
      "cbid": 211, "correlation": 118641
    }
  },
  {
    "ph": "s", "id": 118641, "pid": 76337, "tid": -914061504, "ts": 1716454223191433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223247636, "dur": 7,
    "args": {
      "External id": 118644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118644,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118644, "pid": 5, "tid": 7, "ts": 1716454223247636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191452, "dur": 6,
    "args": {
      "External id": 118644,
      "cbid": 211, "correlation": 118644
    }
  },
  {
    "ph": "s", "id": 118644, "pid": 76337, "tid": -914061504, "ts": 1716454223191452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223247644, "dur": 5,
    "args": {
      "External id": 118653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118653,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118653, "pid": 5, "tid": 7, "ts": 1716454223247644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191492, "dur": 10,
    "args": {
      "External id": 118653,
      "cbid": 211, "correlation": 118653
    }
  },
  {
    "ph": "s", "id": 118653, "pid": 76337, "tid": -914061504, "ts": 1716454223191492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191555, "dur": 0,
    "args": {
      "External id": 118663,
      "cbid": 317, "correlation": 118663
    }
  },
  {
    "ph": "f", "id": 118663, "pid": 76337, "tid": -914061504, "ts": 1716454223191555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191556, "dur": 0,
    "args": {
      "External id": 118664,
      "cbid": 203, "correlation": 118664
    }
  },
  {
    "ph": "f", "id": 118664, "pid": 76337, "tid": -914061504, "ts": 1716454223191556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191556, "dur": 0,
    "args": {
      "External id": 118665,
      "cbid": 205, "correlation": 118665
    }
  },
  {
    "ph": "f", "id": 118665, "pid": 76337, "tid": -914061504, "ts": 1716454223191556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223247650, "dur": 5,
    "args": {
      "External id": 118669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118669,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118669, "pid": 5, "tid": 7, "ts": 1716454223247650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191571, "dur": 12,
    "args": {
      "External id": 118669,
      "cbid": 211, "correlation": 118669
    }
  },
  {
    "ph": "s", "id": 118669, "pid": 76337, "tid": -914061504, "ts": 1716454223191571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223247656, "dur": 162,
    "args": {
      "External id": 118671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118671, "pid": 5, "tid": 7, "ts": 1716454223247656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191585, "dur": 5,
    "args": {
      "External id": 118671,
      "cbid": 211, "correlation": 118671
    }
  },
  {
    "ph": "s", "id": 118671, "pid": 76337, "tid": -914061504, "ts": 1716454223191585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223247821, "dur": 1,
    "args": {
      "External id": 118673,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118673,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 118673, "pid": 5, "tid": 7, "ts": 1716454223247821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191596, "dur": 8,
    "args": {
      "External id": 118673,
      "cbid": 51, "correlation": 118673
    }
  },
  {
    "ph": "s", "id": 118673, "pid": 76337, "tid": -914061504, "ts": 1716454223191596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223247824, "dur": 258,
    "args": {
      "External id": 118674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118674,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118674, "pid": 5, "tid": 7, "ts": 1716454223247824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191605, "dur": 6,
    "args": {
      "External id": 118674,
      "cbid": 211, "correlation": 118674
    }
  },
  {
    "ph": "s", "id": 118674, "pid": 76337, "tid": -914061504, "ts": 1716454223191605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223248084, "dur": 6,
    "args": {
      "External id": 118676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118676,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118676, "pid": 5, "tid": 7, "ts": 1716454223248084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191615, "dur": 5,
    "args": {
      "External id": 118676,
      "cbid": 211, "correlation": 118676
    }
  },
  {
    "ph": "s", "id": 118676, "pid": 76337, "tid": -914061504, "ts": 1716454223191615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223248091, "dur": 6,
    "args": {
      "External id": 118682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118682, "pid": 5, "tid": 7, "ts": 1716454223248091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191644, "dur": 8,
    "args": {
      "External id": 118682,
      "cbid": 211, "correlation": 118682
    }
  },
  {
    "ph": "s", "id": 118682, "pid": 76337, "tid": -914061504, "ts": 1716454223191644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191702, "dur": 0,
    "args": {
      "External id": 118692,
      "cbid": 317, "correlation": 118692
    }
  },
  {
    "ph": "f", "id": 118692, "pid": 76337, "tid": -914061504, "ts": 1716454223191702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191703, "dur": 0,
    "args": {
      "External id": 118693,
      "cbid": 203, "correlation": 118693
    }
  },
  {
    "ph": "f", "id": 118693, "pid": 76337, "tid": -914061504, "ts": 1716454223191703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191704, "dur": 0,
    "args": {
      "External id": 118694,
      "cbid": 205, "correlation": 118694
    }
  },
  {
    "ph": "f", "id": 118694, "pid": 76337, "tid": -914061504, "ts": 1716454223191704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223248099, "dur": 8,
    "args": {
      "External id": 118698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118698,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118698, "pid": 5, "tid": 7, "ts": 1716454223248099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191716, "dur": 11,
    "args": {
      "External id": 118698,
      "cbid": 211, "correlation": 118698
    }
  },
  {
    "ph": "s", "id": 118698, "pid": 76337, "tid": -914061504, "ts": 1716454223191716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223248108, "dur": 3,
    "args": {
      "External id": 118700,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118700,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 118700, "pid": 5, "tid": 7, "ts": 1716454223248108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191732, "dur": 9,
    "args": {
      "External id": 118700,
      "cbid": 51, "correlation": 118700
    }
  },
  {
    "ph": "s", "id": 118700, "pid": 76337, "tid": -914061504, "ts": 1716454223191732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223248112, "dur": 94,
    "args": {
      "External id": 118701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118701,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 118701, "pid": 5, "tid": 7, "ts": 1716454223248112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191742, "dur": 6,
    "args": {
      "External id": 118701,
      "cbid": 211, "correlation": 118701
    }
  },
  {
    "ph": "s", "id": 118701, "pid": 76337, "tid": -914061504, "ts": 1716454223191742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223248207, "dur": 6,
    "args": {
      "External id": 118703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118703,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118703, "pid": 5, "tid": 7, "ts": 1716454223248207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191751, "dur": 5,
    "args": {
      "External id": 118703,
      "cbid": 211, "correlation": 118703
    }
  },
  {
    "ph": "s", "id": 118703, "pid": 76337, "tid": -914061504, "ts": 1716454223191751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223248214, "dur": 6,
    "args": {
      "External id": 118709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118709, "pid": 5, "tid": 7, "ts": 1716454223248214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191778, "dur": 8,
    "args": {
      "External id": 118709,
      "cbid": 211, "correlation": 118709
    }
  },
  {
    "ph": "s", "id": 118709, "pid": 76337, "tid": -914061504, "ts": 1716454223191778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223248222, "dur": 5,
    "args": {
      "External id": 118717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118717,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118717, "pid": 5, "tid": 7, "ts": 1716454223248222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191807, "dur": 8,
    "args": {
      "External id": 118717,
      "cbid": 211, "correlation": 118717
    }
  },
  {
    "ph": "s", "id": 118717, "pid": 76337, "tid": -914061504, "ts": 1716454223191807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223248228, "dur": 4,
    "args": {
      "External id": 118725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118725,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 118725, "pid": 5, "tid": 7, "ts": 1716454223248228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191836, "dur": 8,
    "args": {
      "External id": 118725,
      "cbid": 211, "correlation": 118725
    }
  },
  {
    "ph": "s", "id": 118725, "pid": 76337, "tid": -914061504, "ts": 1716454223191836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223248234, "dur": 14,
    "args": {
      "External id": 118736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118736,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118736, "pid": 5, "tid": 7, "ts": 1716454223248234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191917, "dur": 13,
    "args": {
      "External id": 118736,
      "cbid": 211, "correlation": 118736
    }
  },
  {
    "ph": "s", "id": 118736, "pid": 76337, "tid": -914061504, "ts": 1716454223191917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191982, "dur": 0,
    "args": {
      "External id": 118746,
      "cbid": 317, "correlation": 118746
    }
  },
  {
    "ph": "f", "id": 118746, "pid": 76337, "tid": -914061504, "ts": 1716454223191982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191983, "dur": 0,
    "args": {
      "External id": 118747,
      "cbid": 203, "correlation": 118747
    }
  },
  {
    "ph": "f", "id": 118747, "pid": 76337, "tid": -914061504, "ts": 1716454223191983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191984, "dur": 0,
    "args": {
      "External id": 118748,
      "cbid": 205, "correlation": 118748
    }
  },
  {
    "ph": "f", "id": 118748, "pid": 76337, "tid": -914061504, "ts": 1716454223191984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223248249, "dur": 9,
    "args": {
      "External id": 118752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118752,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118752, "pid": 5, "tid": 7, "ts": 1716454223248249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223191997, "dur": 12,
    "args": {
      "External id": 118752,
      "cbid": 211, "correlation": 118752
    }
  },
  {
    "ph": "s", "id": 118752, "pid": 76337, "tid": -914061504, "ts": 1716454223191997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223248260, "dur": 163,
    "args": {
      "External id": 118754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118754,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118754, "pid": 5, "tid": 7, "ts": 1716454223248260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192011, "dur": 5,
    "args": {
      "External id": 118754,
      "cbid": 211, "correlation": 118754
    }
  },
  {
    "ph": "s", "id": 118754, "pid": 76337, "tid": -914061504, "ts": 1716454223192011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223248425, "dur": 1,
    "args": {
      "External id": 118756,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118756,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 118756, "pid": 5, "tid": 7, "ts": 1716454223248425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192022, "dur": 6,
    "args": {
      "External id": 118756,
      "cbid": 51, "correlation": 118756
    }
  },
  {
    "ph": "s", "id": 118756, "pid": 76337, "tid": -914061504, "ts": 1716454223192022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223248429, "dur": 649,
    "args": {
      "External id": 118757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118757,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118757, "pid": 5, "tid": 7, "ts": 1716454223248429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192029, "dur": 6,
    "args": {
      "External id": 118757,
      "cbid": 211, "correlation": 118757
    }
  },
  {
    "ph": "s", "id": 118757, "pid": 76337, "tid": -914061504, "ts": 1716454223192029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223249079, "dur": 13,
    "args": {
      "External id": 118759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118759,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118759, "pid": 5, "tid": 7, "ts": 1716454223249079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192039, "dur": 5,
    "args": {
      "External id": 118759,
      "cbid": 211, "correlation": 118759
    }
  },
  {
    "ph": "s", "id": 118759, "pid": 76337, "tid": -914061504, "ts": 1716454223192039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223249093, "dur": 15,
    "args": {
      "External id": 118765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118765, "pid": 5, "tid": 7, "ts": 1716454223249093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192068, "dur": 9,
    "args": {
      "External id": 118765,
      "cbid": 211, "correlation": 118765
    }
  },
  {
    "ph": "s", "id": 118765, "pid": 76337, "tid": -914061504, "ts": 1716454223192068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223249109, "dur": 30,
    "args": {
      "External id": 118774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118774,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118774, "pid": 5, "tid": 7, "ts": 1716454223249109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192158, "dur": 13,
    "args": {
      "External id": 118774,
      "cbid": 211, "correlation": 118774
    }
  },
  {
    "ph": "s", "id": 118774, "pid": 76337, "tid": -914061504, "ts": 1716454223192158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223249140, "dur": 30,
    "args": {
      "External id": 118794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118794,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118794, "pid": 5, "tid": 7, "ts": 1716454223249140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192228, "dur": 11,
    "args": {
      "External id": 118794,
      "cbid": 211, "correlation": 118794
    }
  },
  {
    "ph": "s", "id": 118794, "pid": 76337, "tid": -914061504, "ts": 1716454223192228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223249172, "dur": 4,
    "args": {
      "External id": 118806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118806,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118806, "pid": 5, "tid": 7, "ts": 1716454223249172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192248, "dur": 6,
    "args": {
      "External id": 118806,
      "cbid": 211, "correlation": 118806
    }
  },
  {
    "ph": "s", "id": 118806, "pid": 76337, "tid": -914061504, "ts": 1716454223192248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223249177, "dur": 31,
    "args": {
      "External id": 118809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118809, "pid": 5, "tid": 7, "ts": 1716454223249177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192266, "dur": 7,
    "args": {
      "External id": 118809,
      "cbid": 211, "correlation": 118809
    }
  },
  {
    "ph": "s", "id": 118809, "pid": 76337, "tid": -914061504, "ts": 1716454223192266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223249209, "dur": 22,
    "args": {
      "External id": 118818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118818,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118818, "pid": 5, "tid": 7, "ts": 1716454223249209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192305, "dur": 10,
    "args": {
      "External id": 118818,
      "cbid": 211, "correlation": 118818
    }
  },
  {
    "ph": "s", "id": 118818, "pid": 76337, "tid": -914061504, "ts": 1716454223192305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192356, "dur": 0,
    "args": {
      "External id": 118828,
      "cbid": 317, "correlation": 118828
    }
  },
  {
    "ph": "f", "id": 118828, "pid": 76337, "tid": -914061504, "ts": 1716454223192356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192357, "dur": 0,
    "args": {
      "External id": 118829,
      "cbid": 203, "correlation": 118829
    }
  },
  {
    "ph": "f", "id": 118829, "pid": 76337, "tid": -914061504, "ts": 1716454223192357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192358, "dur": 0,
    "args": {
      "External id": 118830,
      "cbid": 205, "correlation": 118830
    }
  },
  {
    "ph": "f", "id": 118830, "pid": 76337, "tid": -914061504, "ts": 1716454223192358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223249232, "dur": 21,
    "args": {
      "External id": 118834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118834,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118834, "pid": 5, "tid": 7, "ts": 1716454223249232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192373, "dur": 11,
    "args": {
      "External id": 118834,
      "cbid": 211, "correlation": 118834
    }
  },
  {
    "ph": "s", "id": 118834, "pid": 76337, "tid": -914061504, "ts": 1716454223192373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223249255, "dur": 320,
    "args": {
      "External id": 118836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118836,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118836, "pid": 5, "tid": 7, "ts": 1716454223249255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192387, "dur": 5,
    "args": {
      "External id": 118836,
      "cbid": 211, "correlation": 118836
    }
  },
  {
    "ph": "s", "id": 118836, "pid": 76337, "tid": -914061504, "ts": 1716454223192387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223249577, "dur": 1,
    "args": {
      "External id": 118838,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118838,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 118838, "pid": 5, "tid": 7, "ts": 1716454223249577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192398, "dur": 6,
    "args": {
      "External id": 118838,
      "cbid": 51, "correlation": 118838
    }
  },
  {
    "ph": "s", "id": 118838, "pid": 76337, "tid": -914061504, "ts": 1716454223192398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223249581, "dur": 1236,
    "args": {
      "External id": 118839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118839,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118839, "pid": 5, "tid": 7, "ts": 1716454223249581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192405, "dur": 6,
    "args": {
      "External id": 118839,
      "cbid": 211, "correlation": 118839
    }
  },
  {
    "ph": "s", "id": 118839, "pid": 76337, "tid": -914061504, "ts": 1716454223192405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223250819, "dur": 12,
    "args": {
      "External id": 118841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118841,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118841, "pid": 5, "tid": 7, "ts": 1716454223250819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192417, "dur": 6,
    "args": {
      "External id": 118841,
      "cbid": 211, "correlation": 118841
    }
  },
  {
    "ph": "s", "id": 118841, "pid": 76337, "tid": -914061504, "ts": 1716454223192417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223250832, "dur": 15,
    "args": {
      "External id": 118847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118847, "pid": 5, "tid": 7, "ts": 1716454223250832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192445, "dur": 8,
    "args": {
      "External id": 118847,
      "cbid": 211, "correlation": 118847
    }
  },
  {
    "ph": "s", "id": 118847, "pid": 76337, "tid": -914061504, "ts": 1716454223192445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223250849, "dur": 4,
    "args": {
      "External id": 118855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118855,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 118855, "pid": 5, "tid": 7, "ts": 1716454223250849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192490, "dur": 9,
    "args": {
      "External id": 118855,
      "cbid": 211, "correlation": 118855
    }
  },
  {
    "ph": "s", "id": 118855, "pid": 76337, "tid": -914061504, "ts": 1716454223192490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192554, "dur": 1,
    "args": {
      "External id": 118871,
      "cbid": 251, "correlation": 118871
    }
  },
  {
    "ph": "f", "id": 118871, "pid": 76337, "tid": -914061504, "ts": 1716454223192554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192559, "dur": 0,
    "args": {
      "External id": 118873,
      "cbid": 251, "correlation": 118873
    }
  },
  {
    "ph": "f", "id": 118873, "pid": 76337, "tid": -914061504, "ts": 1716454223192559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223250854, "dur": 12,
    "args": {
      "External id": 118874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118874,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118874, "pid": 5, "tid": 7, "ts": 1716454223250854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192560, "dur": 12,
    "args": {
      "External id": 118874,
      "cbid": 211, "correlation": 118874
    }
  },
  {
    "ph": "s", "id": 118874, "pid": 76337, "tid": -914061504, "ts": 1716454223192560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223250868, "dur": 5,
    "args": {
      "External id": 118876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118876,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118876, "pid": 5, "tid": 7, "ts": 1716454223250868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192574, "dur": 5,
    "args": {
      "External id": 118876,
      "cbid": 211, "correlation": 118876
    }
  },
  {
    "ph": "s", "id": 118876, "pid": 76337, "tid": -914061504, "ts": 1716454223192574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223250874, "dur": 16,
    "args": {
      "External id": 118886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118886, "pid": 5, "tid": 7, "ts": 1716454223250874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192630, "dur": 12,
    "args": {
      "External id": 118886,
      "cbid": 211, "correlation": 118886
    }
  },
  {
    "ph": "s", "id": 118886, "pid": 76337, "tid": -914061504, "ts": 1716454223192630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223250892, "dur": 17,
    "args": {
      "External id": 118906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118906,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 118906, "pid": 5, "tid": 7, "ts": 1716454223250892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192697, "dur": 10,
    "args": {
      "External id": 118906,
      "cbid": 211, "correlation": 118906
    }
  },
  {
    "ph": "s", "id": 118906, "pid": 76337, "tid": -914061504, "ts": 1716454223192697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223250911, "dur": 4,
    "args": {
      "External id": 118918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118918,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 118918, "pid": 5, "tid": 7, "ts": 1716454223250911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192717, "dur": 7,
    "args": {
      "External id": 118918,
      "cbid": 211, "correlation": 118918
    }
  },
  {
    "ph": "s", "id": 118918, "pid": 76337, "tid": -914061504, "ts": 1716454223192717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223250916, "dur": 17,
    "args": {
      "External id": 118921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118921, "pid": 5, "tid": 7, "ts": 1716454223250916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192736, "dur": 6,
    "args": {
      "External id": 118921,
      "cbid": 211, "correlation": 118921
    }
  },
  {
    "ph": "s", "id": 118921, "pid": 76337, "tid": -914061504, "ts": 1716454223192736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223250934, "dur": 11,
    "args": {
      "External id": 118930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118930,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118930, "pid": 5, "tid": 7, "ts": 1716454223250934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192776, "dur": 9,
    "args": {
      "External id": 118930,
      "cbid": 211, "correlation": 118930
    }
  },
  {
    "ph": "s", "id": 118930, "pid": 76337, "tid": -914061504, "ts": 1716454223192776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192838, "dur": 0,
    "args": {
      "External id": 118940,
      "cbid": 317, "correlation": 118940
    }
  },
  {
    "ph": "f", "id": 118940, "pid": 76337, "tid": -914061504, "ts": 1716454223192838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192839, "dur": 0,
    "args": {
      "External id": 118941,
      "cbid": 203, "correlation": 118941
    }
  },
  {
    "ph": "f", "id": 118941, "pid": 76337, "tid": -914061504, "ts": 1716454223192839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192840, "dur": 0,
    "args": {
      "External id": 118942,
      "cbid": 205, "correlation": 118942
    }
  },
  {
    "ph": "f", "id": 118942, "pid": 76337, "tid": -914061504, "ts": 1716454223192840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223250946, "dur": 11,
    "args": {
      "External id": 118946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118946, "pid": 5, "tid": 7, "ts": 1716454223250946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192853, "dur": 12,
    "args": {
      "External id": 118946,
      "cbid": 211, "correlation": 118946
    }
  },
  {
    "ph": "s", "id": 118946, "pid": 76337, "tid": -914061504, "ts": 1716454223192853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223250958, "dur": 162,
    "args": {
      "External id": 118948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118948, "pid": 5, "tid": 7, "ts": 1716454223250958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192867, "dur": 5,
    "args": {
      "External id": 118948,
      "cbid": 211, "correlation": 118948
    }
  },
  {
    "ph": "s", "id": 118948, "pid": 76337, "tid": -914061504, "ts": 1716454223192867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223251123, "dur": 1,
    "args": {
      "External id": 118950,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 118950,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 118950, "pid": 5, "tid": 7, "ts": 1716454223251123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192879, "dur": 6,
    "args": {
      "External id": 118950,
      "cbid": 51, "correlation": 118950
    }
  },
  {
    "ph": "s", "id": 118950, "pid": 76337, "tid": -914061504, "ts": 1716454223192879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223251127, "dur": 648,
    "args": {
      "External id": 118951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118951,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 118951, "pid": 5, "tid": 7, "ts": 1716454223251127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192886, "dur": 6,
    "args": {
      "External id": 118951,
      "cbid": 211, "correlation": 118951
    }
  },
  {
    "ph": "s", "id": 118951, "pid": 76337, "tid": -914061504, "ts": 1716454223192886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223251776, "dur": 12,
    "args": {
      "External id": 118953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118953,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118953, "pid": 5, "tid": 7, "ts": 1716454223251776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192895, "dur": 5,
    "args": {
      "External id": 118953,
      "cbid": 211, "correlation": 118953
    }
  },
  {
    "ph": "s", "id": 118953, "pid": 76337, "tid": -914061504, "ts": 1716454223192895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223251789, "dur": 15,
    "args": {
      "External id": 118959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118959, "pid": 5, "tid": 7, "ts": 1716454223251789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192923, "dur": 9,
    "args": {
      "External id": 118959,
      "cbid": 211, "correlation": 118959
    }
  },
  {
    "ph": "s", "id": 118959, "pid": 76337, "tid": -914061504, "ts": 1716454223192923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192989, "dur": 0,
    "args": {
      "External id": 118969,
      "cbid": 317, "correlation": 118969
    }
  },
  {
    "ph": "f", "id": 118969, "pid": 76337, "tid": -914061504, "ts": 1716454223192989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192990, "dur": 0,
    "args": {
      "External id": 118970,
      "cbid": 203, "correlation": 118970
    }
  },
  {
    "ph": "f", "id": 118970, "pid": 76337, "tid": -914061504, "ts": 1716454223192990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223192991, "dur": 0,
    "args": {
      "External id": 118971,
      "cbid": 205, "correlation": 118971
    }
  },
  {
    "ph": "f", "id": 118971, "pid": 76337, "tid": -914061504, "ts": 1716454223192991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223251805, "dur": 21,
    "args": {
      "External id": 118975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118975,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118975, "pid": 5, "tid": 7, "ts": 1716454223251805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193004, "dur": 12,
    "args": {
      "External id": 118975,
      "cbid": 211, "correlation": 118975
    }
  },
  {
    "ph": "s", "id": 118975, "pid": 76337, "tid": -914061504, "ts": 1716454223193004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223251827, "dur": 4,
    "args": {
      "External id": 118977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 118977, "pid": 5, "tid": 7, "ts": 1716454223251827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193022, "dur": 7,
    "args": {
      "External id": 118977,
      "cbid": 211, "correlation": 118977
    }
  },
  {
    "ph": "s", "id": 118977, "pid": 76337, "tid": -914061504, "ts": 1716454223193022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193032, "dur": 0,
    "args": {
      "External id": 118978,
      "cbid": 51, "correlation": 118978
    }
  },
  {
    "ph": "s", "id": 118978, "pid": 76337, "tid": -914061504, "ts": 1716454223193032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223251832, "dur": 174,
    "args": {
      "External id": 118979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118979,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 118979, "pid": 5, "tid": 7, "ts": 1716454223251832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193032, "dur": 5,
    "args": {
      "External id": 118979,
      "cbid": 211, "correlation": 118979
    }
  },
  {
    "ph": "s", "id": 118979, "pid": 76337, "tid": -914061504, "ts": 1716454223193032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223252007, "dur": 16,
    "args": {
      "External id": 118984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118984, "pid": 5, "tid": 7, "ts": 1716454223252007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193058, "dur": 8,
    "args": {
      "External id": 118984,
      "cbid": 211, "correlation": 118984
    }
  },
  {
    "ph": "s", "id": 118984, "pid": 76337, "tid": -914061504, "ts": 1716454223193058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223252025, "dur": 11,
    "args": {
      "External id": 118992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 118992,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 118992, "pid": 5, "tid": 7, "ts": 1716454223252025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193087, "dur": 8,
    "args": {
      "External id": 118992,
      "cbid": 211, "correlation": 118992
    }
  },
  {
    "ph": "s", "id": 118992, "pid": 76337, "tid": -914061504, "ts": 1716454223193087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223252037, "dur": 10,
    "args": {
      "External id": 119000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119000,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119000, "pid": 5, "tid": 7, "ts": 1716454223252037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193116, "dur": 8,
    "args": {
      "External id": 119000,
      "cbid": 211, "correlation": 119000
    }
  },
  {
    "ph": "s", "id": 119000, "pid": 76337, "tid": -914061504, "ts": 1716454223193116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223252049, "dur": 19,
    "args": {
      "External id": 119020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119020,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 119020, "pid": 5, "tid": 7, "ts": 1716454223252049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193198, "dur": 12,
    "args": {
      "External id": 119020,
      "cbid": 211, "correlation": 119020
    }
  },
  {
    "ph": "s", "id": 119020, "pid": 76337, "tid": -914061504, "ts": 1716454223193198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223252070, "dur": 5,
    "args": {
      "External id": 119032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119032,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 119032, "pid": 5, "tid": 7, "ts": 1716454223252070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193220, "dur": 6,
    "args": {
      "External id": 119032,
      "cbid": 211, "correlation": 119032
    }
  },
  {
    "ph": "s", "id": 119032, "pid": 76337, "tid": -914061504, "ts": 1716454223193220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223252076, "dur": 17,
    "args": {
      "External id": 119035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119035,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119035, "pid": 5, "tid": 7, "ts": 1716454223252076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193238, "dur": 6,
    "args": {
      "External id": 119035,
      "cbid": 211, "correlation": 119035
    }
  },
  {
    "ph": "s", "id": 119035, "pid": 76337, "tid": -914061504, "ts": 1716454223193238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193295, "dur": 0,
    "args": {
      "External id": 119046,
      "cbid": 317, "correlation": 119046
    }
  },
  {
    "ph": "f", "id": 119046, "pid": 76337, "tid": -914061504, "ts": 1716454223193295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193296, "dur": 0,
    "args": {
      "External id": 119047,
      "cbid": 203, "correlation": 119047
    }
  },
  {
    "ph": "f", "id": 119047, "pid": 76337, "tid": -914061504, "ts": 1716454223193296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193297, "dur": 0,
    "args": {
      "External id": 119048,
      "cbid": 205, "correlation": 119048
    }
  },
  {
    "ph": "f", "id": 119048, "pid": 76337, "tid": -914061504, "ts": 1716454223193297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223252094, "dur": 12,
    "args": {
      "External id": 119052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119052,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119052, "pid": 5, "tid": 7, "ts": 1716454223252094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193310, "dur": 12,
    "args": {
      "External id": 119052,
      "cbid": 211, "correlation": 119052
    }
  },
  {
    "ph": "s", "id": 119052, "pid": 76337, "tid": -914061504, "ts": 1716454223193310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223252107, "dur": 4,
    "args": {
      "External id": 119054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119054,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119054, "pid": 5, "tid": 7, "ts": 1716454223252107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193327, "dur": 6,
    "args": {
      "External id": 119054,
      "cbid": 211, "correlation": 119054
    }
  },
  {
    "ph": "s", "id": 119054, "pid": 76337, "tid": -914061504, "ts": 1716454223193327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193336, "dur": 0,
    "args": {
      "External id": 119055,
      "cbid": 51, "correlation": 119055
    }
  },
  {
    "ph": "s", "id": 119055, "pid": 76337, "tid": -914061504, "ts": 1716454223193336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223252112, "dur": 91,
    "args": {
      "External id": 119056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119056,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 119056, "pid": 5, "tid": 7, "ts": 1716454223252112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193337, "dur": 5,
    "args": {
      "External id": 119056,
      "cbid": 211, "correlation": 119056
    }
  },
  {
    "ph": "s", "id": 119056, "pid": 76337, "tid": -914061504, "ts": 1716454223193337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223252204, "dur": 16,
    "args": {
      "External id": 119061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119061, "pid": 5, "tid": 7, "ts": 1716454223252204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193363, "dur": 9,
    "args": {
      "External id": 119061,
      "cbid": 211, "correlation": 119061
    }
  },
  {
    "ph": "s", "id": 119061, "pid": 76337, "tid": -914061504, "ts": 1716454223193363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223252221, "dur": 83,
    "args": {
      "External id": 119070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119070,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119070, "pid": 5, "tid": 7, "ts": 1716454223252221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193445, "dur": 14,
    "args": {
      "External id": 119070,
      "cbid": 211, "correlation": 119070
    }
  },
  {
    "ph": "s", "id": 119070, "pid": 76337, "tid": -914061504, "ts": 1716454223193445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223252306, "dur": 31,
    "args": {
      "External id": 119092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119092,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119092, "pid": 5, "tid": 7, "ts": 1716454223252306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193503, "dur": 10,
    "args": {
      "External id": 119092,
      "cbid": 211, "correlation": 119092
    }
  },
  {
    "ph": "s", "id": 119092, "pid": 76337, "tid": -914061504, "ts": 1716454223193503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193597, "dur": 1,
    "args": {
      "External id": 119103,
      "cbid": 251, "correlation": 119103
    }
  },
  {
    "ph": "f", "id": 119103, "pid": 76337, "tid": -914061504, "ts": 1716454223193597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223252338, "dur": 140,
    "args": {
      "External id": 119104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119104,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119104, "pid": 5, "tid": 7, "ts": 1716454223252338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193603, "dur": 13,
    "args": {
      "External id": 119104,
      "cbid": 211, "correlation": 119104
    }
  },
  {
    "ph": "s", "id": 119104, "pid": 76337, "tid": -914061504, "ts": 1716454223193603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193673, "dur": 1,
    "args": {
      "External id": 119115,
      "cbid": 251, "correlation": 119115
    }
  },
  {
    "ph": "f", "id": 119115, "pid": 76337, "tid": -914061504, "ts": 1716454223193673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223252479, "dur": 155,
    "args": {
      "External id": 119116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119116,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119116, "pid": 5, "tid": 7, "ts": 1716454223252479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193678, "dur": 12,
    "args": {
      "External id": 119116,
      "cbid": 211, "correlation": 119116
    }
  },
  {
    "ph": "s", "id": 119116, "pid": 76337, "tid": -914061504, "ts": 1716454223193678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193742, "dur": 1,
    "args": {
      "External id": 119127,
      "cbid": 251, "correlation": 119127
    }
  },
  {
    "ph": "f", "id": 119127, "pid": 76337, "tid": -914061504, "ts": 1716454223193742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223252635, "dur": 157,
    "args": {
      "External id": 119128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119128,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119128, "pid": 5, "tid": 7, "ts": 1716454223252635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193746, "dur": 11,
    "args": {
      "External id": 119128,
      "cbid": 211, "correlation": 119128
    }
  },
  {
    "ph": "s", "id": 119128, "pid": 76337, "tid": -914061504, "ts": 1716454223193746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223252794, "dur": 334,
    "args": {
      "External id": 119153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119153,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119153, "pid": 5, "tid": 7, "ts": 1716454223252794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193832, "dur": 13,
    "args": {
      "External id": 119153,
      "cbid": 211, "correlation": 119153
    }
  },
  {
    "ph": "s", "id": 119153, "pid": 76337, "tid": -914061504, "ts": 1716454223193832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193934, "dur": 1,
    "args": {
      "External id": 119171,
      "cbid": 251, "correlation": 119171
    }
  },
  {
    "ph": "f", "id": 119171, "pid": 76337, "tid": -914061504, "ts": 1716454223193934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223253129, "dur": 169,
    "args": {
      "External id": 119173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119173,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119173, "pid": 5, "tid": 7, "ts": 1716454223253129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223193940, "dur": 13,
    "args": {
      "External id": 119173,
      "cbid": 211, "correlation": 119173
    }
  },
  {
    "ph": "s", "id": 119173, "pid": 76337, "tid": -914061504, "ts": 1716454223193940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223253299, "dur": 19,
    "args": {
      "External id": 119181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119181,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119181, "pid": 5, "tid": 7, "ts": 1716454223253299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194019, "dur": 13,
    "args": {
      "External id": 119181,
      "cbid": 211, "correlation": 119181
    }
  },
  {
    "ph": "s", "id": 119181, "pid": 76337, "tid": -914061504, "ts": 1716454223194019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223253320, "dur": 28,
    "args": {
      "External id": 119189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119189,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119189, "pid": 5, "tid": 7, "ts": 1716454223253320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194060, "dur": 8,
    "args": {
      "External id": 119189,
      "cbid": 211, "correlation": 119189
    }
  },
  {
    "ph": "s", "id": 119189, "pid": 76337, "tid": -914061504, "ts": 1716454223194060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223253349, "dur": 19,
    "args": {
      "External id": 119200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119200, "pid": 5, "tid": 7, "ts": 1716454223253349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194133, "dur": 14,
    "args": {
      "External id": 119200,
      "cbid": 211, "correlation": 119200
    }
  },
  {
    "ph": "s", "id": 119200, "pid": 76337, "tid": -914061504, "ts": 1716454223194133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223253369, "dur": 16,
    "args": {
      "External id": 119222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119222,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119222, "pid": 5, "tid": 7, "ts": 1716454223253369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194165, "dur": 7,
    "args": {
      "External id": 119222,
      "cbid": 211, "correlation": 119222
    }
  },
  {
    "ph": "s", "id": 119222, "pid": 76337, "tid": -914061504, "ts": 1716454223194165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194251, "dur": 1,
    "args": {
      "External id": 119233,
      "cbid": 251, "correlation": 119233
    }
  },
  {
    "ph": "f", "id": 119233, "pid": 76337, "tid": -914061504, "ts": 1716454223194251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223253386, "dur": 89,
    "args": {
      "External id": 119234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119234,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119234, "pid": 5, "tid": 7, "ts": 1716454223253386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194256, "dur": 13,
    "args": {
      "External id": 119234,
      "cbid": 211, "correlation": 119234
    }
  },
  {
    "ph": "s", "id": 119234, "pid": 76337, "tid": -914061504, "ts": 1716454223194256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194328, "dur": 1,
    "args": {
      "External id": 119245,
      "cbid": 251, "correlation": 119245
    }
  },
  {
    "ph": "f", "id": 119245, "pid": 76337, "tid": -914061504, "ts": 1716454223194328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194332, "dur": 0,
    "args": {
      "External id": 119246,
      "cbid": 251, "correlation": 119246
    }
  },
  {
    "ph": "f", "id": 119246, "pid": 76337, "tid": -914061504, "ts": 1716454223194332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223253476, "dur": 13,
    "args": {
      "External id": 119247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119247,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119247, "pid": 5, "tid": 7, "ts": 1716454223253476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194334, "dur": 13,
    "args": {
      "External id": 119247,
      "cbid": 211, "correlation": 119247
    }
  },
  {
    "ph": "s", "id": 119247, "pid": 76337, "tid": -914061504, "ts": 1716454223194334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223253490, "dur": 6,
    "args": {
      "External id": 119249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119249,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119249, "pid": 5, "tid": 7, "ts": 1716454223253490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194348, "dur": 6,
    "args": {
      "External id": 119249,
      "cbid": 211, "correlation": 119249
    }
  },
  {
    "ph": "s", "id": 119249, "pid": 76337, "tid": -914061504, "ts": 1716454223194348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194406, "dur": 1,
    "args": {
      "External id": 119260,
      "cbid": 251, "correlation": 119260
    }
  },
  {
    "ph": "f", "id": 119260, "pid": 76337, "tid": -914061504, "ts": 1716454223194406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194410, "dur": 0,
    "args": {
      "External id": 119261,
      "cbid": 251, "correlation": 119261
    }
  },
  {
    "ph": "f", "id": 119261, "pid": 76337, "tid": -914061504, "ts": 1716454223194410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223253498, "dur": 8,
    "args": {
      "External id": 119262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119262,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119262, "pid": 5, "tid": 7, "ts": 1716454223253498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194411, "dur": 11,
    "args": {
      "External id": 119262,
      "cbid": 211, "correlation": 119262
    }
  },
  {
    "ph": "s", "id": 119262, "pid": 76337, "tid": -914061504, "ts": 1716454223194411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223253507, "dur": 4,
    "args": {
      "External id": 119264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119264,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119264, "pid": 5, "tid": 7, "ts": 1716454223253507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194424, "dur": 5,
    "args": {
      "External id": 119264,
      "cbid": 211, "correlation": 119264
    }
  },
  {
    "ph": "s", "id": 119264, "pid": 76337, "tid": -914061504, "ts": 1716454223194424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223253512, "dur": 56,
    "args": {
      "External id": 119289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119289,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119289, "pid": 5, "tid": 7, "ts": 1716454223253512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194501, "dur": 12,
    "args": {
      "External id": 119289,
      "cbid": 211, "correlation": 119289
    }
  },
  {
    "ph": "s", "id": 119289, "pid": 76337, "tid": -914061504, "ts": 1716454223194501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194599, "dur": 2,
    "args": {
      "External id": 119307,
      "cbid": 251, "correlation": 119307
    }
  },
  {
    "ph": "f", "id": 119307, "pid": 76337, "tid": -914061504, "ts": 1716454223194599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223253569, "dur": 93,
    "args": {
      "External id": 119309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119309,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119309, "pid": 5, "tid": 7, "ts": 1716454223253569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194605, "dur": 14,
    "args": {
      "External id": 119309,
      "cbid": 211, "correlation": 119309
    }
  },
  {
    "ph": "s", "id": 119309, "pid": 76337, "tid": -914061504, "ts": 1716454223194605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223253663, "dur": 9,
    "args": {
      "External id": 119317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119317,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119317, "pid": 5, "tid": 7, "ts": 1716454223253663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194675, "dur": 12,
    "args": {
      "External id": 119317,
      "cbid": 211, "correlation": 119317
    }
  },
  {
    "ph": "s", "id": 119317, "pid": 76337, "tid": -914061504, "ts": 1716454223194675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223253674, "dur": 21,
    "args": {
      "External id": 119325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119325, "pid": 5, "tid": 7, "ts": 1716454223253674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194717, "dur": 9,
    "args": {
      "External id": 119325,
      "cbid": 211, "correlation": 119325
    }
  },
  {
    "ph": "s", "id": 119325, "pid": 76337, "tid": -914061504, "ts": 1716454223194717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223253696, "dur": 17,
    "args": {
      "External id": 119347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119347,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119347, "pid": 5, "tid": 7, "ts": 1716454223253696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194768, "dur": 10,
    "args": {
      "External id": 119347,
      "cbid": 211, "correlation": 119347
    }
  },
  {
    "ph": "s", "id": 119347, "pid": 76337, "tid": -914061504, "ts": 1716454223194768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194856, "dur": 1,
    "args": {
      "External id": 119363,
      "cbid": 251, "correlation": 119363
    }
  },
  {
    "ph": "f", "id": 119363, "pid": 76337, "tid": -914061504, "ts": 1716454223194856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194861, "dur": 0,
    "args": {
      "External id": 119365,
      "cbid": 251, "correlation": 119365
    }
  },
  {
    "ph": "f", "id": 119365, "pid": 76337, "tid": -914061504, "ts": 1716454223194861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223253715, "dur": 494,
    "args": {
      "External id": 119366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119366,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119366, "pid": 5, "tid": 7, "ts": 1716454223253715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194863, "dur": 13,
    "args": {
      "External id": 119366,
      "cbid": 211, "correlation": 119366
    }
  },
  {
    "ph": "s", "id": 119366, "pid": 76337, "tid": -914061504, "ts": 1716454223194863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223254210, "dur": 66,
    "args": {
      "External id": 119374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119374, "pid": 5, "tid": 7, "ts": 1716454223254210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194929, "dur": 13,
    "args": {
      "External id": 119374,
      "cbid": 211, "correlation": 119374
    }
  },
  {
    "ph": "s", "id": 119374, "pid": 76337, "tid": -914061504, "ts": 1716454223194929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223254277, "dur": 66,
    "args": {
      "External id": 119382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119382, "pid": 5, "tid": 7, "ts": 1716454223254277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223194961, "dur": 8,
    "args": {
      "External id": 119382,
      "cbid": 211, "correlation": 119382
    }
  },
  {
    "ph": "s", "id": 119382, "pid": 76337, "tid": -914061504, "ts": 1716454223194961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195050, "dur": 1,
    "args": {
      "External id": 119398,
      "cbid": 251, "correlation": 119398
    }
  },
  {
    "ph": "f", "id": 119398, "pid": 76337, "tid": -914061504, "ts": 1716454223195050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223254345, "dur": 1,
    "args": {
      "External id": 119400,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 119400,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 119400, "pid": 5, "tid": 7, "ts": 1716454223254345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195055, "dur": 10,
    "args": {
      "External id": 119400,
      "cbid": 51, "correlation": 119400
    }
  },
  {
    "ph": "s", "id": 119400, "pid": 76337, "tid": -914061504, "ts": 1716454223195055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223254348, "dur": 274,
    "args": {
      "External id": 119401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119401,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119401, "pid": 5, "tid": 7, "ts": 1716454223254348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195066, "dur": 11,
    "args": {
      "External id": 119401,
      "cbid": 211, "correlation": 119401
    }
  },
  {
    "ph": "s", "id": 119401, "pid": 76337, "tid": -914061504, "ts": 1716454223195066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223254624, "dur": 14,
    "args": {
      "External id": 119409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119409,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119409, "pid": 5, "tid": 7, "ts": 1716454223254624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195109, "dur": 10,
    "args": {
      "External id": 119409,
      "cbid": 211, "correlation": 119409
    }
  },
  {
    "ph": "s", "id": 119409, "pid": 76337, "tid": -914061504, "ts": 1716454223195109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223254639, "dur": 38,
    "args": {
      "External id": 119420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119420, "pid": 5, "tid": 7, "ts": 1716454223254639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195177, "dur": 13,
    "args": {
      "External id": 119420,
      "cbid": 211, "correlation": 119420
    }
  },
  {
    "ph": "s", "id": 119420, "pid": 76337, "tid": -914061504, "ts": 1716454223195177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195243, "dur": 0,
    "args": {
      "External id": 119432,
      "cbid": 317, "correlation": 119432
    }
  },
  {
    "ph": "f", "id": 119432, "pid": 76337, "tid": -914061504, "ts": 1716454223195243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195244, "dur": 0,
    "args": {
      "External id": 119433,
      "cbid": 203, "correlation": 119433
    }
  },
  {
    "ph": "f", "id": 119433, "pid": 76337, "tid": -914061504, "ts": 1716454223195244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195245, "dur": 0,
    "args": {
      "External id": 119434,
      "cbid": 205, "correlation": 119434
    }
  },
  {
    "ph": "f", "id": 119434, "pid": 76337, "tid": -914061504, "ts": 1716454223195245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223254678, "dur": 13,
    "args": {
      "External id": 119438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119438,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119438, "pid": 5, "tid": 7, "ts": 1716454223254678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195259, "dur": 12,
    "args": {
      "External id": 119438,
      "cbid": 211, "correlation": 119438
    }
  },
  {
    "ph": "s", "id": 119438, "pid": 76337, "tid": -914061504, "ts": 1716454223195259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223254692, "dur": 4,
    "args": {
      "External id": 119440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119440,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119440, "pid": 5, "tid": 7, "ts": 1716454223254692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195276, "dur": 6,
    "args": {
      "External id": 119440,
      "cbid": 211, "correlation": 119440
    }
  },
  {
    "ph": "s", "id": 119440, "pid": 76337, "tid": -914061504, "ts": 1716454223195276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195284, "dur": 0,
    "args": {
      "External id": 119441,
      "cbid": 51, "correlation": 119441
    }
  },
  {
    "ph": "s", "id": 119441, "pid": 76337, "tid": -914061504, "ts": 1716454223195284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223254698, "dur": 98,
    "args": {
      "External id": 119442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119442,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 119442, "pid": 5, "tid": 7, "ts": 1716454223254698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195285, "dur": 6,
    "args": {
      "External id": 119442,
      "cbid": 211, "correlation": 119442
    }
  },
  {
    "ph": "s", "id": 119442, "pid": 76337, "tid": -914061504, "ts": 1716454223195285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223254797, "dur": 17,
    "args": {
      "External id": 119447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119447,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119447, "pid": 5, "tid": 7, "ts": 1716454223254797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195313, "dur": 8,
    "args": {
      "External id": 119447,
      "cbid": 211, "correlation": 119447
    }
  },
  {
    "ph": "s", "id": 119447, "pid": 76337, "tid": -914061504, "ts": 1716454223195313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223254815, "dur": 12,
    "args": {
      "External id": 119455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119455,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119455, "pid": 5, "tid": 7, "ts": 1716454223254815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195345, "dur": 8,
    "args": {
      "External id": 119455,
      "cbid": 211, "correlation": 119455
    }
  },
  {
    "ph": "s", "id": 119455, "pid": 76337, "tid": -914061504, "ts": 1716454223195345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223254828, "dur": 31,
    "args": {
      "External id": 119464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119464,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119464, "pid": 5, "tid": 7, "ts": 1716454223254828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195383, "dur": 11,
    "args": {
      "External id": 119464,
      "cbid": 211, "correlation": 119464
    }
  },
  {
    "ph": "s", "id": 119464, "pid": 76337, "tid": -914061504, "ts": 1716454223195383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223254860, "dur": 31,
    "args": {
      "External id": 119484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119484,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 119484, "pid": 5, "tid": 7, "ts": 1716454223254860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195455, "dur": 11,
    "args": {
      "External id": 119484,
      "cbid": 211, "correlation": 119484
    }
  },
  {
    "ph": "s", "id": 119484, "pid": 76337, "tid": -914061504, "ts": 1716454223195455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223254893, "dur": 5,
    "args": {
      "External id": 119496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119496,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119496, "pid": 5, "tid": 7, "ts": 1716454223254893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195476, "dur": 6,
    "args": {
      "External id": 119496,
      "cbid": 211, "correlation": 119496
    }
  },
  {
    "ph": "s", "id": 119496, "pid": 76337, "tid": -914061504, "ts": 1716454223195476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223254899, "dur": 31,
    "args": {
      "External id": 119499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119499, "pid": 5, "tid": 7, "ts": 1716454223254899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195495, "dur": 6,
    "args": {
      "External id": 119499,
      "cbid": 211, "correlation": 119499
    }
  },
  {
    "ph": "s", "id": 119499, "pid": 76337, "tid": -914061504, "ts": 1716454223195495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223254931, "dur": 21,
    "args": {
      "External id": 119508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119508,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119508, "pid": 5, "tid": 7, "ts": 1716454223254931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195534, "dur": 11,
    "args": {
      "External id": 119508,
      "cbid": 211, "correlation": 119508
    }
  },
  {
    "ph": "s", "id": 119508, "pid": 76337, "tid": -914061504, "ts": 1716454223195534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195586, "dur": 0,
    "args": {
      "External id": 119518,
      "cbid": 317, "correlation": 119518
    }
  },
  {
    "ph": "f", "id": 119518, "pid": 76337, "tid": -914061504, "ts": 1716454223195586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195587, "dur": 0,
    "args": {
      "External id": 119519,
      "cbid": 203, "correlation": 119519
    }
  },
  {
    "ph": "f", "id": 119519, "pid": 76337, "tid": -914061504, "ts": 1716454223195587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195588, "dur": 0,
    "args": {
      "External id": 119520,
      "cbid": 205, "correlation": 119520
    }
  },
  {
    "ph": "f", "id": 119520, "pid": 76337, "tid": -914061504, "ts": 1716454223195588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223254954, "dur": 22,
    "args": {
      "External id": 119524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119524, "pid": 5, "tid": 7, "ts": 1716454223254954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195602, "dur": 11,
    "args": {
      "External id": 119524,
      "cbid": 211, "correlation": 119524
    }
  },
  {
    "ph": "s", "id": 119524, "pid": 76337, "tid": -914061504, "ts": 1716454223195602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223254977, "dur": 320,
    "args": {
      "External id": 119526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119526,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119526, "pid": 5, "tid": 7, "ts": 1716454223254977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195616, "dur": 5,
    "args": {
      "External id": 119526,
      "cbid": 211, "correlation": 119526
    }
  },
  {
    "ph": "s", "id": 119526, "pid": 76337, "tid": -914061504, "ts": 1716454223195616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223255299, "dur": 1,
    "args": {
      "External id": 119528,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 119528,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 119528, "pid": 5, "tid": 7, "ts": 1716454223255299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195627, "dur": 6,
    "args": {
      "External id": 119528,
      "cbid": 51, "correlation": 119528
    }
  },
  {
    "ph": "s", "id": 119528, "pid": 76337, "tid": -914061504, "ts": 1716454223195627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223255303, "dur": 1254,
    "args": {
      "External id": 119529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119529,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119529, "pid": 5, "tid": 7, "ts": 1716454223255303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195634, "dur": 6,
    "args": {
      "External id": 119529,
      "cbid": 211, "correlation": 119529
    }
  },
  {
    "ph": "s", "id": 119529, "pid": 76337, "tid": -914061504, "ts": 1716454223195634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223256559, "dur": 13,
    "args": {
      "External id": 119531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119531,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119531, "pid": 5, "tid": 7, "ts": 1716454223256559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195645, "dur": 5,
    "args": {
      "External id": 119531,
      "cbid": 211, "correlation": 119531
    }
  },
  {
    "ph": "s", "id": 119531, "pid": 76337, "tid": -914061504, "ts": 1716454223195645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223256573, "dur": 15,
    "args": {
      "External id": 119537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119537,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119537, "pid": 5, "tid": 7, "ts": 1716454223256573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195673, "dur": 8,
    "args": {
      "External id": 119537,
      "cbid": 211, "correlation": 119537
    }
  },
  {
    "ph": "s", "id": 119537, "pid": 76337, "tid": -914061504, "ts": 1716454223195673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223256589, "dur": 4,
    "args": {
      "External id": 119545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119545,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 119545, "pid": 5, "tid": 7, "ts": 1716454223256589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195717, "dur": 10,
    "args": {
      "External id": 119545,
      "cbid": 211, "correlation": 119545
    }
  },
  {
    "ph": "s", "id": 119545, "pid": 76337, "tid": -914061504, "ts": 1716454223195717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195782, "dur": 1,
    "args": {
      "External id": 119561,
      "cbid": 251, "correlation": 119561
    }
  },
  {
    "ph": "f", "id": 119561, "pid": 76337, "tid": -914061504, "ts": 1716454223195782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195787, "dur": 0,
    "args": {
      "External id": 119563,
      "cbid": 251, "correlation": 119563
    }
  },
  {
    "ph": "f", "id": 119563, "pid": 76337, "tid": -914061504, "ts": 1716454223195787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223256595, "dur": 13,
    "args": {
      "External id": 119564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119564,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119564, "pid": 5, "tid": 7, "ts": 1716454223256595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195789, "dur": 12,
    "args": {
      "External id": 119564,
      "cbid": 211, "correlation": 119564
    }
  },
  {
    "ph": "s", "id": 119564, "pid": 76337, "tid": -914061504, "ts": 1716454223195789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223256609, "dur": 5,
    "args": {
      "External id": 119566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119566,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119566, "pid": 5, "tid": 7, "ts": 1716454223256609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195803, "dur": 5,
    "args": {
      "External id": 119566,
      "cbid": 211, "correlation": 119566
    }
  },
  {
    "ph": "s", "id": 119566, "pid": 76337, "tid": -914061504, "ts": 1716454223195803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223256615, "dur": 17,
    "args": {
      "External id": 119576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119576, "pid": 5, "tid": 7, "ts": 1716454223256615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195861, "dur": 12,
    "args": {
      "External id": 119576,
      "cbid": 211, "correlation": 119576
    }
  },
  {
    "ph": "s", "id": 119576, "pid": 76337, "tid": -914061504, "ts": 1716454223195861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223256634, "dur": 17,
    "args": {
      "External id": 119596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119596,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 119596, "pid": 5, "tid": 7, "ts": 1716454223256634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195927, "dur": 11,
    "args": {
      "External id": 119596,
      "cbid": 211, "correlation": 119596
    }
  },
  {
    "ph": "s", "id": 119596, "pid": 76337, "tid": -914061504, "ts": 1716454223195927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223256653, "dur": 4,
    "args": {
      "External id": 119608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119608,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 119608, "pid": 5, "tid": 7, "ts": 1716454223256653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195949, "dur": 6,
    "args": {
      "External id": 119608,
      "cbid": 211, "correlation": 119608
    }
  },
  {
    "ph": "s", "id": 119608, "pid": 76337, "tid": -914061504, "ts": 1716454223195949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223256658, "dur": 17,
    "args": {
      "External id": 119611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119611, "pid": 5, "tid": 7, "ts": 1716454223256658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223195967, "dur": 14,
    "args": {
      "External id": 119611,
      "cbid": 211, "correlation": 119611
    }
  },
  {
    "ph": "s", "id": 119611, "pid": 76337, "tid": -914061504, "ts": 1716454223195967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223256677, "dur": 12,
    "args": {
      "External id": 119620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119620,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119620, "pid": 5, "tid": 7, "ts": 1716454223256677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196017, "dur": 10,
    "args": {
      "External id": 119620,
      "cbid": 211, "correlation": 119620
    }
  },
  {
    "ph": "s", "id": 119620, "pid": 76337, "tid": -914061504, "ts": 1716454223196017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196080, "dur": 0,
    "args": {
      "External id": 119630,
      "cbid": 317, "correlation": 119630
    }
  },
  {
    "ph": "f", "id": 119630, "pid": 76337, "tid": -914061504, "ts": 1716454223196080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196081, "dur": 0,
    "args": {
      "External id": 119631,
      "cbid": 203, "correlation": 119631
    }
  },
  {
    "ph": "f", "id": 119631, "pid": 76337, "tid": -914061504, "ts": 1716454223196081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196081, "dur": 0,
    "args": {
      "External id": 119632,
      "cbid": 205, "correlation": 119632
    }
  },
  {
    "ph": "f", "id": 119632, "pid": 76337, "tid": -914061504, "ts": 1716454223196081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223256690, "dur": 12,
    "args": {
      "External id": 119636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119636,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119636, "pid": 5, "tid": 7, "ts": 1716454223256690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196095, "dur": 12,
    "args": {
      "External id": 119636,
      "cbid": 211, "correlation": 119636
    }
  },
  {
    "ph": "s", "id": 119636, "pid": 76337, "tid": -914061504, "ts": 1716454223196095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223256703, "dur": 162,
    "args": {
      "External id": 119638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119638,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119638, "pid": 5, "tid": 7, "ts": 1716454223256703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196110, "dur": 5,
    "args": {
      "External id": 119638,
      "cbid": 211, "correlation": 119638
    }
  },
  {
    "ph": "s", "id": 119638, "pid": 76337, "tid": -914061504, "ts": 1716454223196110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223256868, "dur": 1,
    "args": {
      "External id": 119640,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 119640,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 119640, "pid": 5, "tid": 7, "ts": 1716454223256868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196121, "dur": 6,
    "args": {
      "External id": 119640,
      "cbid": 51, "correlation": 119640
    }
  },
  {
    "ph": "s", "id": 119640, "pid": 76337, "tid": -914061504, "ts": 1716454223196121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223256871, "dur": 649,
    "args": {
      "External id": 119641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119641,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119641, "pid": 5, "tid": 7, "ts": 1716454223256871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196128, "dur": 6,
    "args": {
      "External id": 119641,
      "cbid": 211, "correlation": 119641
    }
  },
  {
    "ph": "s", "id": 119641, "pid": 76337, "tid": -914061504, "ts": 1716454223196128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223257521, "dur": 12,
    "args": {
      "External id": 119643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119643,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119643, "pid": 5, "tid": 7, "ts": 1716454223257521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196139, "dur": 5,
    "args": {
      "External id": 119643,
      "cbid": 211, "correlation": 119643
    }
  },
  {
    "ph": "s", "id": 119643, "pid": 76337, "tid": -914061504, "ts": 1716454223196139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223257535, "dur": 15,
    "args": {
      "External id": 119649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119649,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119649, "pid": 5, "tid": 7, "ts": 1716454223257535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196168, "dur": 8,
    "args": {
      "External id": 119649,
      "cbid": 211, "correlation": 119649
    }
  },
  {
    "ph": "s", "id": 119649, "pid": 76337, "tid": -914061504, "ts": 1716454223196168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196226, "dur": 0,
    "args": {
      "External id": 119659,
      "cbid": 317, "correlation": 119659
    }
  },
  {
    "ph": "f", "id": 119659, "pid": 76337, "tid": -914061504, "ts": 1716454223196226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196227, "dur": 0,
    "args": {
      "External id": 119660,
      "cbid": 203, "correlation": 119660
    }
  },
  {
    "ph": "f", "id": 119660, "pid": 76337, "tid": -914061504, "ts": 1716454223196227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196228, "dur": 0,
    "args": {
      "External id": 119661,
      "cbid": 205, "correlation": 119661
    }
  },
  {
    "ph": "f", "id": 119661, "pid": 76337, "tid": -914061504, "ts": 1716454223196228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223257552, "dur": 21,
    "args": {
      "External id": 119665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119665,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119665, "pid": 5, "tid": 7, "ts": 1716454223257552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196239, "dur": 12,
    "args": {
      "External id": 119665,
      "cbid": 211, "correlation": 119665
    }
  },
  {
    "ph": "s", "id": 119665, "pid": 76337, "tid": -914061504, "ts": 1716454223196239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223257574, "dur": 4,
    "args": {
      "External id": 119667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119667, "pid": 5, "tid": 7, "ts": 1716454223257574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196256, "dur": 6,
    "args": {
      "External id": 119667,
      "cbid": 211, "correlation": 119667
    }
  },
  {
    "ph": "s", "id": 119667, "pid": 76337, "tid": -914061504, "ts": 1716454223196256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196264, "dur": 0,
    "args": {
      "External id": 119668,
      "cbid": 51, "correlation": 119668
    }
  },
  {
    "ph": "s", "id": 119668, "pid": 76337, "tid": -914061504, "ts": 1716454223196264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223257579, "dur": 170,
    "args": {
      "External id": 119669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119669,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 119669, "pid": 5, "tid": 7, "ts": 1716454223257579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196265, "dur": 5,
    "args": {
      "External id": 119669,
      "cbid": 211, "correlation": 119669
    }
  },
  {
    "ph": "s", "id": 119669, "pid": 76337, "tid": -914061504, "ts": 1716454223196265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223257751, "dur": 15,
    "args": {
      "External id": 119674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119674, "pid": 5, "tid": 7, "ts": 1716454223257751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196291, "dur": 8,
    "args": {
      "External id": 119674,
      "cbid": 211, "correlation": 119674
    }
  },
  {
    "ph": "s", "id": 119674, "pid": 76337, "tid": -914061504, "ts": 1716454223196291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223257767, "dur": 13,
    "args": {
      "External id": 119682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119682,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119682, "pid": 5, "tid": 7, "ts": 1716454223257767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196320, "dur": 8,
    "args": {
      "External id": 119682,
      "cbid": 211, "correlation": 119682
    }
  },
  {
    "ph": "s", "id": 119682, "pid": 76337, "tid": -914061504, "ts": 1716454223196320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223257781, "dur": 10,
    "args": {
      "External id": 119690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119690,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119690, "pid": 5, "tid": 7, "ts": 1716454223257781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196347, "dur": 9,
    "args": {
      "External id": 119690,
      "cbid": 211, "correlation": 119690
    }
  },
  {
    "ph": "s", "id": 119690, "pid": 76337, "tid": -914061504, "ts": 1716454223196347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223257793, "dur": 19,
    "args": {
      "External id": 119710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119710,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 119710, "pid": 5, "tid": 7, "ts": 1716454223257793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196430, "dur": 12,
    "args": {
      "External id": 119710,
      "cbid": 211, "correlation": 119710
    }
  },
  {
    "ph": "s", "id": 119710, "pid": 76337, "tid": -914061504, "ts": 1716454223196430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223257813, "dur": 4,
    "args": {
      "External id": 119722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119722,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 119722, "pid": 5, "tid": 7, "ts": 1716454223257813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196452, "dur": 7,
    "args": {
      "External id": 119722,
      "cbid": 211, "correlation": 119722
    }
  },
  {
    "ph": "s", "id": 119722, "pid": 76337, "tid": -914061504, "ts": 1716454223196452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223257819, "dur": 17,
    "args": {
      "External id": 119725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119725,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119725, "pid": 5, "tid": 7, "ts": 1716454223257819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196470, "dur": 6,
    "args": {
      "External id": 119725,
      "cbid": 211, "correlation": 119725
    }
  },
  {
    "ph": "s", "id": 119725, "pid": 76337, "tid": -914061504, "ts": 1716454223196470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196528, "dur": 0,
    "args": {
      "External id": 119736,
      "cbid": 317, "correlation": 119736
    }
  },
  {
    "ph": "f", "id": 119736, "pid": 76337, "tid": -914061504, "ts": 1716454223196528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196528, "dur": 0,
    "args": {
      "External id": 119737,
      "cbid": 203, "correlation": 119737
    }
  },
  {
    "ph": "f", "id": 119737, "pid": 76337, "tid": -914061504, "ts": 1716454223196528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196529, "dur": 0,
    "args": {
      "External id": 119738,
      "cbid": 205, "correlation": 119738
    }
  },
  {
    "ph": "f", "id": 119738, "pid": 76337, "tid": -914061504, "ts": 1716454223196529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223257837, "dur": 11,
    "args": {
      "External id": 119742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119742, "pid": 5, "tid": 7, "ts": 1716454223257837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196542, "dur": 12,
    "args": {
      "External id": 119742,
      "cbid": 211, "correlation": 119742
    }
  },
  {
    "ph": "s", "id": 119742, "pid": 76337, "tid": -914061504, "ts": 1716454223196542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223257850, "dur": 3,
    "args": {
      "External id": 119744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119744, "pid": 5, "tid": 7, "ts": 1716454223257850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196559, "dur": 6,
    "args": {
      "External id": 119744,
      "cbid": 211, "correlation": 119744
    }
  },
  {
    "ph": "s", "id": 119744, "pid": 76337, "tid": -914061504, "ts": 1716454223196559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196567, "dur": 0,
    "args": {
      "External id": 119745,
      "cbid": 51, "correlation": 119745
    }
  },
  {
    "ph": "s", "id": 119745, "pid": 76337, "tid": -914061504, "ts": 1716454223196567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223257855, "dur": 92,
    "args": {
      "External id": 119746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119746,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 119746, "pid": 5, "tid": 7, "ts": 1716454223257855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196568, "dur": 5,
    "args": {
      "External id": 119746,
      "cbid": 211, "correlation": 119746
    }
  },
  {
    "ph": "s", "id": 119746, "pid": 76337, "tid": -914061504, "ts": 1716454223196568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223257948, "dur": 16,
    "args": {
      "External id": 119751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119751, "pid": 5, "tid": 7, "ts": 1716454223257948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196594, "dur": 9,
    "args": {
      "External id": 119751,
      "cbid": 211, "correlation": 119751
    }
  },
  {
    "ph": "s", "id": 119751, "pid": 76337, "tid": -914061504, "ts": 1716454223196594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223257964, "dur": 83,
    "args": {
      "External id": 119760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119760,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119760, "pid": 5, "tid": 7, "ts": 1716454223257964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196674, "dur": 14,
    "args": {
      "External id": 119760,
      "cbid": 211, "correlation": 119760
    }
  },
  {
    "ph": "s", "id": 119760, "pid": 76337, "tid": -914061504, "ts": 1716454223196674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223258049, "dur": 30,
    "args": {
      "External id": 119782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119782,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119782, "pid": 5, "tid": 7, "ts": 1716454223258049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196731, "dur": 10,
    "args": {
      "External id": 119782,
      "cbid": 211, "correlation": 119782
    }
  },
  {
    "ph": "s", "id": 119782, "pid": 76337, "tid": -914061504, "ts": 1716454223196731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196819, "dur": 1,
    "args": {
      "External id": 119793,
      "cbid": 251, "correlation": 119793
    }
  },
  {
    "ph": "f", "id": 119793, "pid": 76337, "tid": -914061504, "ts": 1716454223196819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223258080, "dur": 164,
    "args": {
      "External id": 119794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119794,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119794, "pid": 5, "tid": 7, "ts": 1716454223258080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196824, "dur": 13,
    "args": {
      "External id": 119794,
      "cbid": 211, "correlation": 119794
    }
  },
  {
    "ph": "s", "id": 119794, "pid": 76337, "tid": -914061504, "ts": 1716454223196824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196895, "dur": 1,
    "args": {
      "External id": 119805,
      "cbid": 251, "correlation": 119805
    }
  },
  {
    "ph": "f", "id": 119805, "pid": 76337, "tid": -914061504, "ts": 1716454223196895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223258245, "dur": 160,
    "args": {
      "External id": 119806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119806,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119806, "pid": 5, "tid": 7, "ts": 1716454223258245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196899, "dur": 12,
    "args": {
      "External id": 119806,
      "cbid": 211, "correlation": 119806
    }
  },
  {
    "ph": "s", "id": 119806, "pid": 76337, "tid": -914061504, "ts": 1716454223196899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196966, "dur": 1,
    "args": {
      "External id": 119817,
      "cbid": 251, "correlation": 119817
    }
  },
  {
    "ph": "f", "id": 119817, "pid": 76337, "tid": -914061504, "ts": 1716454223196966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223258407, "dur": 157,
    "args": {
      "External id": 119818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119818,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119818, "pid": 5, "tid": 7, "ts": 1716454223258407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223196970, "dur": 18,
    "args": {
      "External id": 119818,
      "cbid": 211, "correlation": 119818
    }
  },
  {
    "ph": "s", "id": 119818, "pid": 76337, "tid": -914061504, "ts": 1716454223196970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223258565, "dur": 338,
    "args": {
      "External id": 119843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119843,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119843, "pid": 5, "tid": 7, "ts": 1716454223258565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197062, "dur": 12,
    "args": {
      "External id": 119843,
      "cbid": 211, "correlation": 119843
    }
  },
  {
    "ph": "s", "id": 119843, "pid": 76337, "tid": -914061504, "ts": 1716454223197062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197160, "dur": 1,
    "args": {
      "External id": 119861,
      "cbid": 251, "correlation": 119861
    }
  },
  {
    "ph": "f", "id": 119861, "pid": 76337, "tid": -914061504, "ts": 1716454223197160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223258904, "dur": 167,
    "args": {
      "External id": 119863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119863,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119863, "pid": 5, "tid": 7, "ts": 1716454223258904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197167, "dur": 13,
    "args": {
      "External id": 119863,
      "cbid": 211, "correlation": 119863
    }
  },
  {
    "ph": "s", "id": 119863, "pid": 76337, "tid": -914061504, "ts": 1716454223197167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223259073, "dur": 19,
    "args": {
      "External id": 119871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119871,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119871, "pid": 5, "tid": 7, "ts": 1716454223259073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197237, "dur": 12,
    "args": {
      "External id": 119871,
      "cbid": 211, "correlation": 119871
    }
  },
  {
    "ph": "s", "id": 119871, "pid": 76337, "tid": -914061504, "ts": 1716454223197237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223259094, "dur": 27,
    "args": {
      "External id": 119879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119879,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119879, "pid": 5, "tid": 7, "ts": 1716454223259094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197277, "dur": 8,
    "args": {
      "External id": 119879,
      "cbid": 211, "correlation": 119879
    }
  },
  {
    "ph": "s", "id": 119879, "pid": 76337, "tid": -914061504, "ts": 1716454223197277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223259122, "dur": 19,
    "args": {
      "External id": 119890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119890, "pid": 5, "tid": 7, "ts": 1716454223259122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197348, "dur": 12,
    "args": {
      "External id": 119890,
      "cbid": 211, "correlation": 119890
    }
  },
  {
    "ph": "s", "id": 119890, "pid": 76337, "tid": -914061504, "ts": 1716454223197348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223259143, "dur": 16,
    "args": {
      "External id": 119912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119912,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 119912, "pid": 5, "tid": 7, "ts": 1716454223259143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197380, "dur": 8,
    "args": {
      "External id": 119912,
      "cbid": 211, "correlation": 119912
    }
  },
  {
    "ph": "s", "id": 119912, "pid": 76337, "tid": -914061504, "ts": 1716454223197380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197465, "dur": 1,
    "args": {
      "External id": 119923,
      "cbid": 251, "correlation": 119923
    }
  },
  {
    "ph": "f", "id": 119923, "pid": 76337, "tid": -914061504, "ts": 1716454223197465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223259160, "dur": 89,
    "args": {
      "External id": 119924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119924,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119924, "pid": 5, "tid": 7, "ts": 1716454223259160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197470, "dur": 13,
    "args": {
      "External id": 119924,
      "cbid": 211, "correlation": 119924
    }
  },
  {
    "ph": "s", "id": 119924, "pid": 76337, "tid": -914061504, "ts": 1716454223197470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197539, "dur": 1,
    "args": {
      "External id": 119935,
      "cbid": 251, "correlation": 119935
    }
  },
  {
    "ph": "f", "id": 119935, "pid": 76337, "tid": -914061504, "ts": 1716454223197539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197543, "dur": 0,
    "args": {
      "External id": 119936,
      "cbid": 251, "correlation": 119936
    }
  },
  {
    "ph": "f", "id": 119936, "pid": 76337, "tid": -914061504, "ts": 1716454223197543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223259251, "dur": 12,
    "args": {
      "External id": 119937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119937,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119937, "pid": 5, "tid": 7, "ts": 1716454223259251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197545, "dur": 12,
    "args": {
      "External id": 119937,
      "cbid": 211, "correlation": 119937
    }
  },
  {
    "ph": "s", "id": 119937, "pid": 76337, "tid": -914061504, "ts": 1716454223197545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223259264, "dur": 6,
    "args": {
      "External id": 119939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119939,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119939, "pid": 5, "tid": 7, "ts": 1716454223259264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197558, "dur": 6,
    "args": {
      "External id": 119939,
      "cbid": 211, "correlation": 119939
    }
  },
  {
    "ph": "s", "id": 119939, "pid": 76337, "tid": -914061504, "ts": 1716454223197558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197615, "dur": 1,
    "args": {
      "External id": 119950,
      "cbid": 251, "correlation": 119950
    }
  },
  {
    "ph": "f", "id": 119950, "pid": 76337, "tid": -914061504, "ts": 1716454223197615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197619, "dur": 0,
    "args": {
      "External id": 119951,
      "cbid": 251, "correlation": 119951
    }
  },
  {
    "ph": "f", "id": 119951, "pid": 76337, "tid": -914061504, "ts": 1716454223197619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223259271, "dur": 9,
    "args": {
      "External id": 119952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119952,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119952, "pid": 5, "tid": 7, "ts": 1716454223259271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197621, "dur": 12,
    "args": {
      "External id": 119952,
      "cbid": 211, "correlation": 119952
    }
  },
  {
    "ph": "s", "id": 119952, "pid": 76337, "tid": -914061504, "ts": 1716454223197621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223259281, "dur": 4,
    "args": {
      "External id": 119954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119954,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119954, "pid": 5, "tid": 7, "ts": 1716454223259281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197634, "dur": 5,
    "args": {
      "External id": 119954,
      "cbid": 211, "correlation": 119954
    }
  },
  {
    "ph": "s", "id": 119954, "pid": 76337, "tid": -914061504, "ts": 1716454223197634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223259286, "dur": 55,
    "args": {
      "External id": 119979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119979,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 119979, "pid": 5, "tid": 7, "ts": 1716454223259286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197711, "dur": 13,
    "args": {
      "External id": 119979,
      "cbid": 211, "correlation": 119979
    }
  },
  {
    "ph": "s", "id": 119979, "pid": 76337, "tid": -914061504, "ts": 1716454223197711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197809, "dur": 1,
    "args": {
      "External id": 119997,
      "cbid": 251, "correlation": 119997
    }
  },
  {
    "ph": "f", "id": 119997, "pid": 76337, "tid": -914061504, "ts": 1716454223197809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223259342, "dur": 91,
    "args": {
      "External id": 119999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 119999,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 119999, "pid": 5, "tid": 7, "ts": 1716454223259342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197815, "dur": 14,
    "args": {
      "External id": 119999,
      "cbid": 211, "correlation": 119999
    }
  },
  {
    "ph": "s", "id": 119999, "pid": 76337, "tid": -914061504, "ts": 1716454223197815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223259434, "dur": 9,
    "args": {
      "External id": 120007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120007,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120007, "pid": 5, "tid": 7, "ts": 1716454223259434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197886, "dur": 12,
    "args": {
      "External id": 120007,
      "cbid": 211, "correlation": 120007
    }
  },
  {
    "ph": "s", "id": 120007, "pid": 76337, "tid": -914061504, "ts": 1716454223197886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223259445, "dur": 20,
    "args": {
      "External id": 120015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120015,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120015, "pid": 5, "tid": 7, "ts": 1716454223259445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197926, "dur": 10,
    "args": {
      "External id": 120015,
      "cbid": 211, "correlation": 120015
    }
  },
  {
    "ph": "s", "id": 120015, "pid": 76337, "tid": -914061504, "ts": 1716454223197926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223259467, "dur": 18,
    "args": {
      "External id": 120037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120037,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120037, "pid": 5, "tid": 7, "ts": 1716454223259467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223197996, "dur": 11,
    "args": {
      "External id": 120037,
      "cbid": 211, "correlation": 120037
    }
  },
  {
    "ph": "s", "id": 120037, "pid": 76337, "tid": -914061504, "ts": 1716454223197996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198086, "dur": 1,
    "args": {
      "External id": 120053,
      "cbid": 251, "correlation": 120053
    }
  },
  {
    "ph": "f", "id": 120053, "pid": 76337, "tid": -914061504, "ts": 1716454223198086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198091, "dur": 0,
    "args": {
      "External id": 120055,
      "cbid": 251, "correlation": 120055
    }
  },
  {
    "ph": "f", "id": 120055, "pid": 76337, "tid": -914061504, "ts": 1716454223198091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223259486, "dur": 497,
    "args": {
      "External id": 120056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120056,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120056, "pid": 5, "tid": 7, "ts": 1716454223259486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198093, "dur": 12,
    "args": {
      "External id": 120056,
      "cbid": 211, "correlation": 120056
    }
  },
  {
    "ph": "s", "id": 120056, "pid": 76337, "tid": -914061504, "ts": 1716454223198093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223259984, "dur": 66,
    "args": {
      "External id": 120064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120064, "pid": 5, "tid": 7, "ts": 1716454223259984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198158, "dur": 12,
    "args": {
      "External id": 120064,
      "cbid": 211, "correlation": 120064
    }
  },
  {
    "ph": "s", "id": 120064, "pid": 76337, "tid": -914061504, "ts": 1716454223198158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223260052, "dur": 68,
    "args": {
      "External id": 120072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120072, "pid": 5, "tid": 7, "ts": 1716454223260052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198189, "dur": 8,
    "args": {
      "External id": 120072,
      "cbid": 211, "correlation": 120072
    }
  },
  {
    "ph": "s", "id": 120072, "pid": 76337, "tid": -914061504, "ts": 1716454223198189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198268, "dur": 1,
    "args": {
      "External id": 120088,
      "cbid": 251, "correlation": 120088
    }
  },
  {
    "ph": "f", "id": 120088, "pid": 76337, "tid": -914061504, "ts": 1716454223198268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223260122, "dur": 1,
    "args": {
      "External id": 120090,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 120090,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 120090, "pid": 5, "tid": 7, "ts": 1716454223260122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198273, "dur": 10,
    "args": {
      "External id": 120090,
      "cbid": 51, "correlation": 120090
    }
  },
  {
    "ph": "s", "id": 120090, "pid": 76337, "tid": -914061504, "ts": 1716454223198273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223260126, "dur": 271,
    "args": {
      "External id": 120091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120091,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120091, "pid": 5, "tid": 7, "ts": 1716454223260126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198285, "dur": 11,
    "args": {
      "External id": 120091,
      "cbid": 211, "correlation": 120091
    }
  },
  {
    "ph": "s", "id": 120091, "pid": 76337, "tid": -914061504, "ts": 1716454223198285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223260398, "dur": 14,
    "args": {
      "External id": 120099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120099,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120099, "pid": 5, "tid": 7, "ts": 1716454223260398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198326, "dur": 11,
    "args": {
      "External id": 120099,
      "cbid": 211, "correlation": 120099
    }
  },
  {
    "ph": "s", "id": 120099, "pid": 76337, "tid": -914061504, "ts": 1716454223198326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223260413, "dur": 38,
    "args": {
      "External id": 120110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120110,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120110, "pid": 5, "tid": 7, "ts": 1716454223260413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198395, "dur": 12,
    "args": {
      "External id": 120110,
      "cbid": 211, "correlation": 120110
    }
  },
  {
    "ph": "s", "id": 120110, "pid": 76337, "tid": -914061504, "ts": 1716454223198395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198459, "dur": 0,
    "args": {
      "External id": 120122,
      "cbid": 317, "correlation": 120122
    }
  },
  {
    "ph": "f", "id": 120122, "pid": 76337, "tid": -914061504, "ts": 1716454223198459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198460, "dur": 0,
    "args": {
      "External id": 120123,
      "cbid": 203, "correlation": 120123
    }
  },
  {
    "ph": "f", "id": 120123, "pid": 76337, "tid": -914061504, "ts": 1716454223198460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198460, "dur": 0,
    "args": {
      "External id": 120124,
      "cbid": 205, "correlation": 120124
    }
  },
  {
    "ph": "f", "id": 120124, "pid": 76337, "tid": -914061504, "ts": 1716454223198460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223260453, "dur": 15,
    "args": {
      "External id": 120128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120128,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120128, "pid": 5, "tid": 7, "ts": 1716454223260453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198475, "dur": 13,
    "args": {
      "External id": 120128,
      "cbid": 211, "correlation": 120128
    }
  },
  {
    "ph": "s", "id": 120128, "pid": 76337, "tid": -914061504, "ts": 1716454223198475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223260469, "dur": 4,
    "args": {
      "External id": 120130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120130, "pid": 5, "tid": 7, "ts": 1716454223260469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198493, "dur": 6,
    "args": {
      "External id": 120130,
      "cbid": 211, "correlation": 120130
    }
  },
  {
    "ph": "s", "id": 120130, "pid": 76337, "tid": -914061504, "ts": 1716454223198493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198502, "dur": 0,
    "args": {
      "External id": 120131,
      "cbid": 51, "correlation": 120131
    }
  },
  {
    "ph": "s", "id": 120131, "pid": 76337, "tid": -914061504, "ts": 1716454223198502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223260474, "dur": 97,
    "args": {
      "External id": 120132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120132,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 120132, "pid": 5, "tid": 7, "ts": 1716454223260474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198503, "dur": 5,
    "args": {
      "External id": 120132,
      "cbid": 211, "correlation": 120132
    }
  },
  {
    "ph": "s", "id": 120132, "pid": 76337, "tid": -914061504, "ts": 1716454223198503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223260572, "dur": 16,
    "args": {
      "External id": 120137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120137,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120137, "pid": 5, "tid": 7, "ts": 1716454223260572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198530, "dur": 9,
    "args": {
      "External id": 120137,
      "cbid": 211, "correlation": 120137
    }
  },
  {
    "ph": "s", "id": 120137, "pid": 76337, "tid": -914061504, "ts": 1716454223198530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223260590, "dur": 11,
    "args": {
      "External id": 120145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120145,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120145, "pid": 5, "tid": 7, "ts": 1716454223260590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198563, "dur": 8,
    "args": {
      "External id": 120145,
      "cbid": 211, "correlation": 120145
    }
  },
  {
    "ph": "s", "id": 120145, "pid": 76337, "tid": -914061504, "ts": 1716454223198563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223260603, "dur": 25,
    "args": {
      "External id": 120154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120154,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120154, "pid": 5, "tid": 7, "ts": 1716454223260603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198601, "dur": 10,
    "args": {
      "External id": 120154,
      "cbid": 211, "correlation": 120154
    }
  },
  {
    "ph": "s", "id": 120154, "pid": 76337, "tid": -914061504, "ts": 1716454223198601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223260629, "dur": 23,
    "args": {
      "External id": 120174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120174,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 120174, "pid": 5, "tid": 7, "ts": 1716454223260629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198672, "dur": 13,
    "args": {
      "External id": 120174,
      "cbid": 211, "correlation": 120174
    }
  },
  {
    "ph": "s", "id": 120174, "pid": 76337, "tid": -914061504, "ts": 1716454223198672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223260654, "dur": 5,
    "args": {
      "External id": 120186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120186,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 120186, "pid": 5, "tid": 7, "ts": 1716454223260654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198694, "dur": 7,
    "args": {
      "External id": 120186,
      "cbid": 211, "correlation": 120186
    }
  },
  {
    "ph": "s", "id": 120186, "pid": 76337, "tid": -914061504, "ts": 1716454223198694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223260660, "dur": 25,
    "args": {
      "External id": 120189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120189, "pid": 5, "tid": 7, "ts": 1716454223260660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198712, "dur": 7,
    "args": {
      "External id": 120189,
      "cbid": 211, "correlation": 120189
    }
  },
  {
    "ph": "s", "id": 120189, "pid": 76337, "tid": -914061504, "ts": 1716454223198712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223260686, "dur": 17,
    "args": {
      "External id": 120198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120198,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120198, "pid": 5, "tid": 7, "ts": 1716454223260686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198752, "dur": 9,
    "args": {
      "External id": 120198,
      "cbid": 211, "correlation": 120198
    }
  },
  {
    "ph": "s", "id": 120198, "pid": 76337, "tid": -914061504, "ts": 1716454223198752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198803, "dur": 0,
    "args": {
      "External id": 120208,
      "cbid": 317, "correlation": 120208
    }
  },
  {
    "ph": "f", "id": 120208, "pid": 76337, "tid": -914061504, "ts": 1716454223198803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198803, "dur": 0,
    "args": {
      "External id": 120209,
      "cbid": 203, "correlation": 120209
    }
  },
  {
    "ph": "f", "id": 120209, "pid": 76337, "tid": -914061504, "ts": 1716454223198803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198804, "dur": 0,
    "args": {
      "External id": 120210,
      "cbid": 205, "correlation": 120210
    }
  },
  {
    "ph": "f", "id": 120210, "pid": 76337, "tid": -914061504, "ts": 1716454223198804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223260704, "dur": 17,
    "args": {
      "External id": 120214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120214,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120214, "pid": 5, "tid": 7, "ts": 1716454223260704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198818, "dur": 11,
    "args": {
      "External id": 120214,
      "cbid": 211, "correlation": 120214
    }
  },
  {
    "ph": "s", "id": 120214, "pid": 76337, "tid": -914061504, "ts": 1716454223198818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223260723, "dur": 241,
    "args": {
      "External id": 120216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120216,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120216, "pid": 5, "tid": 7, "ts": 1716454223260723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198832, "dur": 6,
    "args": {
      "External id": 120216,
      "cbid": 211, "correlation": 120216
    }
  },
  {
    "ph": "s", "id": 120216, "pid": 76337, "tid": -914061504, "ts": 1716454223198832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223260967, "dur": 1,
    "args": {
      "External id": 120218,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 120218,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 120218, "pid": 5, "tid": 7, "ts": 1716454223260967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198844, "dur": 7,
    "args": {
      "External id": 120218,
      "cbid": 51, "correlation": 120218
    }
  },
  {
    "ph": "s", "id": 120218, "pid": 76337, "tid": -914061504, "ts": 1716454223198844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223260970, "dur": 808,
    "args": {
      "External id": 120219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120219,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120219, "pid": 5, "tid": 7, "ts": 1716454223260970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198853, "dur": 6,
    "args": {
      "External id": 120219,
      "cbid": 211, "correlation": 120219
    }
  },
  {
    "ph": "s", "id": 120219, "pid": 76337, "tid": -914061504, "ts": 1716454223198853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223261780, "dur": 14,
    "args": {
      "External id": 120221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120221,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120221, "pid": 5, "tid": 7, "ts": 1716454223261780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198863, "dur": 5,
    "args": {
      "External id": 120221,
      "cbid": 211, "correlation": 120221
    }
  },
  {
    "ph": "s", "id": 120221, "pid": 76337, "tid": -914061504, "ts": 1716454223198863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223261795, "dur": 14,
    "args": {
      "External id": 120227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120227,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120227, "pid": 5, "tid": 7, "ts": 1716454223261795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198892, "dur": 8,
    "args": {
      "External id": 120227,
      "cbid": 211, "correlation": 120227
    }
  },
  {
    "ph": "s", "id": 120227, "pid": 76337, "tid": -914061504, "ts": 1716454223198892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223261811, "dur": 5,
    "args": {
      "External id": 120235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120235,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 120235, "pid": 5, "tid": 7, "ts": 1716454223261811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223198935, "dur": 10,
    "args": {
      "External id": 120235,
      "cbid": 211, "correlation": 120235
    }
  },
  {
    "ph": "s", "id": 120235, "pid": 76337, "tid": -914061504, "ts": 1716454223198935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199008, "dur": 1,
    "args": {
      "External id": 120251,
      "cbid": 251, "correlation": 120251
    }
  },
  {
    "ph": "f", "id": 120251, "pid": 76337, "tid": -914061504, "ts": 1716454223199008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199013, "dur": 0,
    "args": {
      "External id": 120253,
      "cbid": 251, "correlation": 120253
    }
  },
  {
    "ph": "f", "id": 120253, "pid": 76337, "tid": -914061504, "ts": 1716454223199013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223261817, "dur": 13,
    "args": {
      "External id": 120254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120254,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120254, "pid": 5, "tid": 7, "ts": 1716454223261817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199014, "dur": 12,
    "args": {
      "External id": 120254,
      "cbid": 211, "correlation": 120254
    }
  },
  {
    "ph": "s", "id": 120254, "pid": 76337, "tid": -914061504, "ts": 1716454223199014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223261832, "dur": 5,
    "args": {
      "External id": 120256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120256,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120256, "pid": 5, "tid": 7, "ts": 1716454223261832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199028, "dur": 5,
    "args": {
      "External id": 120256,
      "cbid": 211, "correlation": 120256
    }
  },
  {
    "ph": "s", "id": 120256, "pid": 76337, "tid": -914061504, "ts": 1716454223199028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223261838, "dur": 19,
    "args": {
      "External id": 120266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120266,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120266, "pid": 5, "tid": 7, "ts": 1716454223261838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199086, "dur": 12,
    "args": {
      "External id": 120266,
      "cbid": 211, "correlation": 120266
    }
  },
  {
    "ph": "s", "id": 120266, "pid": 76337, "tid": -914061504, "ts": 1716454223199086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223261858, "dur": 18,
    "args": {
      "External id": 120286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120286,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 120286, "pid": 5, "tid": 7, "ts": 1716454223261858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199153, "dur": 10,
    "args": {
      "External id": 120286,
      "cbid": 211, "correlation": 120286
    }
  },
  {
    "ph": "s", "id": 120286, "pid": 76337, "tid": -914061504, "ts": 1716454223199153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223261877, "dur": 4,
    "args": {
      "External id": 120298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120298,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 120298, "pid": 5, "tid": 7, "ts": 1716454223261877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199173, "dur": 6,
    "args": {
      "External id": 120298,
      "cbid": 211, "correlation": 120298
    }
  },
  {
    "ph": "s", "id": 120298, "pid": 76337, "tid": -914061504, "ts": 1716454223199173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223261883, "dur": 16,
    "args": {
      "External id": 120301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120301, "pid": 5, "tid": 7, "ts": 1716454223261883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199192, "dur": 6,
    "args": {
      "External id": 120301,
      "cbid": 211, "correlation": 120301
    }
  },
  {
    "ph": "s", "id": 120301, "pid": 76337, "tid": -914061504, "ts": 1716454223199192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223261901, "dur": 11,
    "args": {
      "External id": 120310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120310,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120310, "pid": 5, "tid": 7, "ts": 1716454223261901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199231, "dur": 10,
    "args": {
      "External id": 120310,
      "cbid": 211, "correlation": 120310
    }
  },
  {
    "ph": "s", "id": 120310, "pid": 76337, "tid": -914061504, "ts": 1716454223199231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199294, "dur": 0,
    "args": {
      "External id": 120320,
      "cbid": 317, "correlation": 120320
    }
  },
  {
    "ph": "f", "id": 120320, "pid": 76337, "tid": -914061504, "ts": 1716454223199294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199294, "dur": 0,
    "args": {
      "External id": 120321,
      "cbid": 203, "correlation": 120321
    }
  },
  {
    "ph": "f", "id": 120321, "pid": 76337, "tid": -914061504, "ts": 1716454223199294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199295, "dur": 0,
    "args": {
      "External id": 120322,
      "cbid": 205, "correlation": 120322
    }
  },
  {
    "ph": "f", "id": 120322, "pid": 76337, "tid": -914061504, "ts": 1716454223199295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223261913, "dur": 12,
    "args": {
      "External id": 120326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120326,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120326, "pid": 5, "tid": 7, "ts": 1716454223261913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199308, "dur": 12,
    "args": {
      "External id": 120326,
      "cbid": 211, "correlation": 120326
    }
  },
  {
    "ph": "s", "id": 120326, "pid": 76337, "tid": -914061504, "ts": 1716454223199308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223261926, "dur": 163,
    "args": {
      "External id": 120328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120328,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120328, "pid": 5, "tid": 7, "ts": 1716454223261926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199323, "dur": 5,
    "args": {
      "External id": 120328,
      "cbid": 211, "correlation": 120328
    }
  },
  {
    "ph": "s", "id": 120328, "pid": 76337, "tid": -914061504, "ts": 1716454223199323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223262091, "dur": 1,
    "args": {
      "External id": 120330,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 120330,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 120330, "pid": 5, "tid": 7, "ts": 1716454223262091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199333, "dur": 7,
    "args": {
      "External id": 120330,
      "cbid": 51, "correlation": 120330
    }
  },
  {
    "ph": "s", "id": 120330, "pid": 76337, "tid": -914061504, "ts": 1716454223199333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223262095, "dur": 649,
    "args": {
      "External id": 120331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120331,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120331, "pid": 5, "tid": 7, "ts": 1716454223262095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199342, "dur": 6,
    "args": {
      "External id": 120331,
      "cbid": 211, "correlation": 120331
    }
  },
  {
    "ph": "s", "id": 120331, "pid": 76337, "tid": -914061504, "ts": 1716454223199342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223262746, "dur": 12,
    "args": {
      "External id": 120333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120333,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120333, "pid": 5, "tid": 7, "ts": 1716454223262746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199352, "dur": 5,
    "args": {
      "External id": 120333,
      "cbid": 211, "correlation": 120333
    }
  },
  {
    "ph": "s", "id": 120333, "pid": 76337, "tid": -914061504, "ts": 1716454223199352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223262759, "dur": 15,
    "args": {
      "External id": 120339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120339,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120339, "pid": 5, "tid": 7, "ts": 1716454223262759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199381, "dur": 10,
    "args": {
      "External id": 120339,
      "cbid": 211, "correlation": 120339
    }
  },
  {
    "ph": "s", "id": 120339, "pid": 76337, "tid": -914061504, "ts": 1716454223199381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199442, "dur": 0,
    "args": {
      "External id": 120349,
      "cbid": 317, "correlation": 120349
    }
  },
  {
    "ph": "f", "id": 120349, "pid": 76337, "tid": -914061504, "ts": 1716454223199442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199442, "dur": 0,
    "args": {
      "External id": 120350,
      "cbid": 203, "correlation": 120350
    }
  },
  {
    "ph": "f", "id": 120350, "pid": 76337, "tid": -914061504, "ts": 1716454223199442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199443, "dur": 0,
    "args": {
      "External id": 120351,
      "cbid": 205, "correlation": 120351
    }
  },
  {
    "ph": "f", "id": 120351, "pid": 76337, "tid": -914061504, "ts": 1716454223199443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223262775, "dur": 17,
    "args": {
      "External id": 120355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120355,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120355, "pid": 5, "tid": 7, "ts": 1716454223262775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199457, "dur": 12,
    "args": {
      "External id": 120355,
      "cbid": 211, "correlation": 120355
    }
  },
  {
    "ph": "s", "id": 120355, "pid": 76337, "tid": -914061504, "ts": 1716454223199457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223262794, "dur": 4,
    "args": {
      "External id": 120357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120357, "pid": 5, "tid": 7, "ts": 1716454223262794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199473, "dur": 6,
    "args": {
      "External id": 120357,
      "cbid": 211, "correlation": 120357
    }
  },
  {
    "ph": "s", "id": 120357, "pid": 76337, "tid": -914061504, "ts": 1716454223199473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199482, "dur": 0,
    "args": {
      "External id": 120358,
      "cbid": 51, "correlation": 120358
    }
  },
  {
    "ph": "s", "id": 120358, "pid": 76337, "tid": -914061504, "ts": 1716454223199482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223262799, "dur": 132,
    "args": {
      "External id": 120359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120359,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 120359, "pid": 5, "tid": 7, "ts": 1716454223262799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199483, "dur": 6,
    "args": {
      "External id": 120359,
      "cbid": 211, "correlation": 120359
    }
  },
  {
    "ph": "s", "id": 120359, "pid": 76337, "tid": -914061504, "ts": 1716454223199483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223262933, "dur": 15,
    "args": {
      "External id": 120364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120364, "pid": 5, "tid": 7, "ts": 1716454223262933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199509, "dur": 8,
    "args": {
      "External id": 120364,
      "cbid": 211, "correlation": 120364
    }
  },
  {
    "ph": "s", "id": 120364, "pid": 76337, "tid": -914061504, "ts": 1716454223199509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223262949, "dur": 12,
    "args": {
      "External id": 120372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120372,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120372, "pid": 5, "tid": 7, "ts": 1716454223262949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199538, "dur": 8,
    "args": {
      "External id": 120372,
      "cbid": 211, "correlation": 120372
    }
  },
  {
    "ph": "s", "id": 120372, "pid": 76337, "tid": -914061504, "ts": 1716454223199538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223262963, "dur": 11,
    "args": {
      "External id": 120380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120380,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120380, "pid": 5, "tid": 7, "ts": 1716454223262963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199566, "dur": 8,
    "args": {
      "External id": 120380,
      "cbid": 211, "correlation": 120380
    }
  },
  {
    "ph": "s", "id": 120380, "pid": 76337, "tid": -914061504, "ts": 1716454223199566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223262975, "dur": 19,
    "args": {
      "External id": 120400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120400,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 120400, "pid": 5, "tid": 7, "ts": 1716454223262975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199650, "dur": 12,
    "args": {
      "External id": 120400,
      "cbid": 211, "correlation": 120400
    }
  },
  {
    "ph": "s", "id": 120400, "pid": 76337, "tid": -914061504, "ts": 1716454223199650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223262995, "dur": 4,
    "args": {
      "External id": 120412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120412,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 120412, "pid": 5, "tid": 7, "ts": 1716454223262995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199671, "dur": 6,
    "args": {
      "External id": 120412,
      "cbid": 211, "correlation": 120412
    }
  },
  {
    "ph": "s", "id": 120412, "pid": 76337, "tid": -914061504, "ts": 1716454223199671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223263001, "dur": 17,
    "args": {
      "External id": 120415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120415, "pid": 5, "tid": 7, "ts": 1716454223263001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199689, "dur": 7,
    "args": {
      "External id": 120415,
      "cbid": 211, "correlation": 120415
    }
  },
  {
    "ph": "s", "id": 120415, "pid": 76337, "tid": -914061504, "ts": 1716454223199689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199747, "dur": 0,
    "args": {
      "External id": 120426,
      "cbid": 317, "correlation": 120426
    }
  },
  {
    "ph": "f", "id": 120426, "pid": 76337, "tid": -914061504, "ts": 1716454223199747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199747, "dur": 0,
    "args": {
      "External id": 120427,
      "cbid": 203, "correlation": 120427
    }
  },
  {
    "ph": "f", "id": 120427, "pid": 76337, "tid": -914061504, "ts": 1716454223199747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199748, "dur": 0,
    "args": {
      "External id": 120428,
      "cbid": 205, "correlation": 120428
    }
  },
  {
    "ph": "f", "id": 120428, "pid": 76337, "tid": -914061504, "ts": 1716454223199748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223263019, "dur": 12,
    "args": {
      "External id": 120432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120432, "pid": 5, "tid": 7, "ts": 1716454223263019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199761, "dur": 11,
    "args": {
      "External id": 120432,
      "cbid": 211, "correlation": 120432
    }
  },
  {
    "ph": "s", "id": 120432, "pid": 76337, "tid": -914061504, "ts": 1716454223199761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223263032, "dur": 3,
    "args": {
      "External id": 120434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120434,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120434, "pid": 5, "tid": 7, "ts": 1716454223263032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199777, "dur": 6,
    "args": {
      "External id": 120434,
      "cbid": 211, "correlation": 120434
    }
  },
  {
    "ph": "s", "id": 120434, "pid": 76337, "tid": -914061504, "ts": 1716454223199777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199785, "dur": 0,
    "args": {
      "External id": 120435,
      "cbid": 51, "correlation": 120435
    }
  },
  {
    "ph": "s", "id": 120435, "pid": 76337, "tid": -914061504, "ts": 1716454223199785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223263037, "dur": 91,
    "args": {
      "External id": 120436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120436,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 120436, "pid": 5, "tid": 7, "ts": 1716454223263037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199786, "dur": 5,
    "args": {
      "External id": 120436,
      "cbid": 211, "correlation": 120436
    }
  },
  {
    "ph": "s", "id": 120436, "pid": 76337, "tid": -914061504, "ts": 1716454223199786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223263130, "dur": 15,
    "args": {
      "External id": 120441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120441,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120441, "pid": 5, "tid": 7, "ts": 1716454223263130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199812, "dur": 9,
    "args": {
      "External id": 120441,
      "cbid": 211, "correlation": 120441
    }
  },
  {
    "ph": "s", "id": 120441, "pid": 76337, "tid": -914061504, "ts": 1716454223199812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223263146, "dur": 85,
    "args": {
      "External id": 120450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120450,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120450, "pid": 5, "tid": 7, "ts": 1716454223263146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199892, "dur": 15,
    "args": {
      "External id": 120450,
      "cbid": 211, "correlation": 120450
    }
  },
  {
    "ph": "s", "id": 120450, "pid": 76337, "tid": -914061504, "ts": 1716454223199892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223263233, "dur": 31,
    "args": {
      "External id": 120472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120472,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120472, "pid": 5, "tid": 7, "ts": 1716454223263233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223199951, "dur": 10,
    "args": {
      "External id": 120472,
      "cbid": 211, "correlation": 120472
    }
  },
  {
    "ph": "s", "id": 120472, "pid": 76337, "tid": -914061504, "ts": 1716454223199951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200049, "dur": 1,
    "args": {
      "External id": 120483,
      "cbid": 251, "correlation": 120483
    }
  },
  {
    "ph": "f", "id": 120483, "pid": 76337, "tid": -914061504, "ts": 1716454223200049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223263265, "dur": 165,
    "args": {
      "External id": 120484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120484,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120484, "pid": 5, "tid": 7, "ts": 1716454223263265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200055, "dur": 14,
    "args": {
      "External id": 120484,
      "cbid": 211, "correlation": 120484
    }
  },
  {
    "ph": "s", "id": 120484, "pid": 76337, "tid": -914061504, "ts": 1716454223200055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200126, "dur": 1,
    "args": {
      "External id": 120495,
      "cbid": 251, "correlation": 120495
    }
  },
  {
    "ph": "f", "id": 120495, "pid": 76337, "tid": -914061504, "ts": 1716454223200126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223263432, "dur": 157,
    "args": {
      "External id": 120496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120496,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120496, "pid": 5, "tid": 7, "ts": 1716454223263432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200130, "dur": 11,
    "args": {
      "External id": 120496,
      "cbid": 211, "correlation": 120496
    }
  },
  {
    "ph": "s", "id": 120496, "pid": 76337, "tid": -914061504, "ts": 1716454223200130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200195, "dur": 1,
    "args": {
      "External id": 120507,
      "cbid": 251, "correlation": 120507
    }
  },
  {
    "ph": "f", "id": 120507, "pid": 76337, "tid": -914061504, "ts": 1716454223200195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223263589, "dur": 158,
    "args": {
      "External id": 120508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120508,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120508, "pid": 5, "tid": 7, "ts": 1716454223263589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200200, "dur": 11,
    "args": {
      "External id": 120508,
      "cbid": 211, "correlation": 120508
    }
  },
  {
    "ph": "s", "id": 120508, "pid": 76337, "tid": -914061504, "ts": 1716454223200200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223263749, "dur": 337,
    "args": {
      "External id": 120533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120533,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120533, "pid": 5, "tid": 7, "ts": 1716454223263749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200283, "dur": 12,
    "args": {
      "External id": 120533,
      "cbid": 211, "correlation": 120533
    }
  },
  {
    "ph": "s", "id": 120533, "pid": 76337, "tid": -914061504, "ts": 1716454223200283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200384, "dur": 1,
    "args": {
      "External id": 120551,
      "cbid": 251, "correlation": 120551
    }
  },
  {
    "ph": "f", "id": 120551, "pid": 76337, "tid": -914061504, "ts": 1716454223200384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223264088, "dur": 144,
    "args": {
      "External id": 120553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120553,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120553, "pid": 5, "tid": 7, "ts": 1716454223264088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200389, "dur": 14,
    "args": {
      "External id": 120553,
      "cbid": 211, "correlation": 120553
    }
  },
  {
    "ph": "s", "id": 120553, "pid": 76337, "tid": -914061504, "ts": 1716454223200389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223264233, "dur": 20,
    "args": {
      "External id": 120561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120561,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120561, "pid": 5, "tid": 7, "ts": 1716454223264233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200461, "dur": 12,
    "args": {
      "External id": 120561,
      "cbid": 211, "correlation": 120561
    }
  },
  {
    "ph": "s", "id": 120561, "pid": 76337, "tid": -914061504, "ts": 1716454223200461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223264254, "dur": 28,
    "args": {
      "External id": 120569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120569,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120569, "pid": 5, "tid": 7, "ts": 1716454223264254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200499, "dur": 9,
    "args": {
      "External id": 120569,
      "cbid": 211, "correlation": 120569
    }
  },
  {
    "ph": "s", "id": 120569, "pid": 76337, "tid": -914061504, "ts": 1716454223200499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223264283, "dur": 19,
    "args": {
      "External id": 120580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120580,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120580, "pid": 5, "tid": 7, "ts": 1716454223264283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200571, "dur": 12,
    "args": {
      "External id": 120580,
      "cbid": 211, "correlation": 120580
    }
  },
  {
    "ph": "s", "id": 120580, "pid": 76337, "tid": -914061504, "ts": 1716454223200571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223264303, "dur": 16,
    "args": {
      "External id": 120602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120602,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120602, "pid": 5, "tid": 7, "ts": 1716454223264303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200602, "dur": 8,
    "args": {
      "External id": 120602,
      "cbid": 211, "correlation": 120602
    }
  },
  {
    "ph": "s", "id": 120602, "pid": 76337, "tid": -914061504, "ts": 1716454223200602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200686, "dur": 1,
    "args": {
      "External id": 120613,
      "cbid": 251, "correlation": 120613
    }
  },
  {
    "ph": "f", "id": 120613, "pid": 76337, "tid": -914061504, "ts": 1716454223200686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223264321, "dur": 90,
    "args": {
      "External id": 120614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120614,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120614, "pid": 5, "tid": 7, "ts": 1716454223264321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200692, "dur": 14,
    "args": {
      "External id": 120614,
      "cbid": 211, "correlation": 120614
    }
  },
  {
    "ph": "s", "id": 120614, "pid": 76337, "tid": -914061504, "ts": 1716454223200692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200761, "dur": 1,
    "args": {
      "External id": 120625,
      "cbid": 251, "correlation": 120625
    }
  },
  {
    "ph": "f", "id": 120625, "pid": 76337, "tid": -914061504, "ts": 1716454223200761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200765, "dur": 0,
    "args": {
      "External id": 120626,
      "cbid": 251, "correlation": 120626
    }
  },
  {
    "ph": "f", "id": 120626, "pid": 76337, "tid": -914061504, "ts": 1716454223200765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223264412, "dur": 12,
    "args": {
      "External id": 120627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120627,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120627, "pid": 5, "tid": 7, "ts": 1716454223264412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200766, "dur": 12,
    "args": {
      "External id": 120627,
      "cbid": 211, "correlation": 120627
    }
  },
  {
    "ph": "s", "id": 120627, "pid": 76337, "tid": -914061504, "ts": 1716454223200766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223264426, "dur": 6,
    "args": {
      "External id": 120629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120629,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120629, "pid": 5, "tid": 7, "ts": 1716454223264426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200780, "dur": 6,
    "args": {
      "External id": 120629,
      "cbid": 211, "correlation": 120629
    }
  },
  {
    "ph": "s", "id": 120629, "pid": 76337, "tid": -914061504, "ts": 1716454223200780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200837, "dur": 1,
    "args": {
      "External id": 120640,
      "cbid": 251, "correlation": 120640
    }
  },
  {
    "ph": "f", "id": 120640, "pid": 76337, "tid": -914061504, "ts": 1716454223200837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200840, "dur": 0,
    "args": {
      "External id": 120641,
      "cbid": 251, "correlation": 120641
    }
  },
  {
    "ph": "f", "id": 120641, "pid": 76337, "tid": -914061504, "ts": 1716454223200840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223264433, "dur": 8,
    "args": {
      "External id": 120642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120642,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120642, "pid": 5, "tid": 7, "ts": 1716454223264433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200841, "dur": 11,
    "args": {
      "External id": 120642,
      "cbid": 211, "correlation": 120642
    }
  },
  {
    "ph": "s", "id": 120642, "pid": 76337, "tid": -914061504, "ts": 1716454223200841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223264443, "dur": 4,
    "args": {
      "External id": 120644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120644,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120644, "pid": 5, "tid": 7, "ts": 1716454223264443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200855, "dur": 6,
    "args": {
      "External id": 120644,
      "cbid": 211, "correlation": 120644
    }
  },
  {
    "ph": "s", "id": 120644, "pid": 76337, "tid": -914061504, "ts": 1716454223200855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223264447, "dur": 55,
    "args": {
      "External id": 120669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120669,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120669, "pid": 5, "tid": 7, "ts": 1716454223264447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223200931, "dur": 12,
    "args": {
      "External id": 120669,
      "cbid": 211, "correlation": 120669
    }
  },
  {
    "ph": "s", "id": 120669, "pid": 76337, "tid": -914061504, "ts": 1716454223200931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201038, "dur": 1,
    "args": {
      "External id": 120687,
      "cbid": 251, "correlation": 120687
    }
  },
  {
    "ph": "f", "id": 120687, "pid": 76337, "tid": -914061504, "ts": 1716454223201038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223264504, "dur": 93,
    "args": {
      "External id": 120689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120689,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120689, "pid": 5, "tid": 7, "ts": 1716454223264504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201044, "dur": 15,
    "args": {
      "External id": 120689,
      "cbid": 211, "correlation": 120689
    }
  },
  {
    "ph": "s", "id": 120689, "pid": 76337, "tid": -914061504, "ts": 1716454223201044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223264598, "dur": 10,
    "args": {
      "External id": 120697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120697,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120697, "pid": 5, "tid": 7, "ts": 1716454223264598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201116, "dur": 12,
    "args": {
      "External id": 120697,
      "cbid": 211, "correlation": 120697
    }
  },
  {
    "ph": "s", "id": 120697, "pid": 76337, "tid": -914061504, "ts": 1716454223201116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223264609, "dur": 20,
    "args": {
      "External id": 120705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120705, "pid": 5, "tid": 7, "ts": 1716454223264609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201156, "dur": 10,
    "args": {
      "External id": 120705,
      "cbid": 211, "correlation": 120705
    }
  },
  {
    "ph": "s", "id": 120705, "pid": 76337, "tid": -914061504, "ts": 1716454223201156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223264631, "dur": 17,
    "args": {
      "External id": 120727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120727,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120727, "pid": 5, "tid": 7, "ts": 1716454223264631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201207, "dur": 10,
    "args": {
      "External id": 120727,
      "cbid": 211, "correlation": 120727
    }
  },
  {
    "ph": "s", "id": 120727, "pid": 76337, "tid": -914061504, "ts": 1716454223201207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201294, "dur": 1,
    "args": {
      "External id": 120743,
      "cbid": 251, "correlation": 120743
    }
  },
  {
    "ph": "f", "id": 120743, "pid": 76337, "tid": -914061504, "ts": 1716454223201294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201298, "dur": 0,
    "args": {
      "External id": 120745,
      "cbid": 251, "correlation": 120745
    }
  },
  {
    "ph": "f", "id": 120745, "pid": 76337, "tid": -914061504, "ts": 1716454223201298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223264649, "dur": 497,
    "args": {
      "External id": 120746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120746,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120746, "pid": 5, "tid": 7, "ts": 1716454223264649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201300, "dur": 13,
    "args": {
      "External id": 120746,
      "cbid": 211, "correlation": 120746
    }
  },
  {
    "ph": "s", "id": 120746, "pid": 76337, "tid": -914061504, "ts": 1716454223201300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223265148, "dur": 66,
    "args": {
      "External id": 120754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120754, "pid": 5, "tid": 7, "ts": 1716454223265148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201367, "dur": 12,
    "args": {
      "External id": 120754,
      "cbid": 211, "correlation": 120754
    }
  },
  {
    "ph": "s", "id": 120754, "pid": 76337, "tid": -914061504, "ts": 1716454223201367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223265215, "dur": 67,
    "args": {
      "External id": 120762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120762,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120762, "pid": 5, "tid": 7, "ts": 1716454223265215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201397, "dur": 9,
    "args": {
      "External id": 120762,
      "cbid": 211, "correlation": 120762
    }
  },
  {
    "ph": "s", "id": 120762, "pid": 76337, "tid": -914061504, "ts": 1716454223201397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201476, "dur": 1,
    "args": {
      "External id": 120778,
      "cbid": 251, "correlation": 120778
    }
  },
  {
    "ph": "f", "id": 120778, "pid": 76337, "tid": -914061504, "ts": 1716454223201476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223265284, "dur": 1,
    "args": {
      "External id": 120780,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 120780,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 120780, "pid": 5, "tid": 7, "ts": 1716454223265284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201481, "dur": 9,
    "args": {
      "External id": 120780,
      "cbid": 51, "correlation": 120780
    }
  },
  {
    "ph": "s", "id": 120780, "pid": 76337, "tid": -914061504, "ts": 1716454223201481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223265287, "dur": 271,
    "args": {
      "External id": 120781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120781,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120781, "pid": 5, "tid": 7, "ts": 1716454223265287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201492, "dur": 11,
    "args": {
      "External id": 120781,
      "cbid": 211, "correlation": 120781
    }
  },
  {
    "ph": "s", "id": 120781, "pid": 76337, "tid": -914061504, "ts": 1716454223201492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223265560, "dur": 13,
    "args": {
      "External id": 120789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120789, "pid": 5, "tid": 7, "ts": 1716454223265560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201534, "dur": 10,
    "args": {
      "External id": 120789,
      "cbid": 211, "correlation": 120789
    }
  },
  {
    "ph": "s", "id": 120789, "pid": 76337, "tid": -914061504, "ts": 1716454223201534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223265575, "dur": 38,
    "args": {
      "External id": 120800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120800, "pid": 5, "tid": 7, "ts": 1716454223265575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201601, "dur": 13,
    "args": {
      "External id": 120800,
      "cbid": 211, "correlation": 120800
    }
  },
  {
    "ph": "s", "id": 120800, "pid": 76337, "tid": -914061504, "ts": 1716454223201601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201665, "dur": 0,
    "args": {
      "External id": 120812,
      "cbid": 317, "correlation": 120812
    }
  },
  {
    "ph": "f", "id": 120812, "pid": 76337, "tid": -914061504, "ts": 1716454223201665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201666, "dur": 0,
    "args": {
      "External id": 120813,
      "cbid": 203, "correlation": 120813
    }
  },
  {
    "ph": "f", "id": 120813, "pid": 76337, "tid": -914061504, "ts": 1716454223201666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201667, "dur": 0,
    "args": {
      "External id": 120814,
      "cbid": 205, "correlation": 120814
    }
  },
  {
    "ph": "f", "id": 120814, "pid": 76337, "tid": -914061504, "ts": 1716454223201667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223265614, "dur": 13,
    "args": {
      "External id": 120818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120818,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120818, "pid": 5, "tid": 7, "ts": 1716454223265614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201683, "dur": 12,
    "args": {
      "External id": 120818,
      "cbid": 211, "correlation": 120818
    }
  },
  {
    "ph": "s", "id": 120818, "pid": 76337, "tid": -914061504, "ts": 1716454223201683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223265628, "dur": 4,
    "args": {
      "External id": 120820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 120820, "pid": 5, "tid": 7, "ts": 1716454223265628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201699, "dur": 6,
    "args": {
      "External id": 120820,
      "cbid": 211, "correlation": 120820
    }
  },
  {
    "ph": "s", "id": 120820, "pid": 76337, "tid": -914061504, "ts": 1716454223201699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201707, "dur": 0,
    "args": {
      "External id": 120821,
      "cbid": 51, "correlation": 120821
    }
  },
  {
    "ph": "s", "id": 120821, "pid": 76337, "tid": -914061504, "ts": 1716454223201707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223265633, "dur": 98,
    "args": {
      "External id": 120822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120822,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 120822, "pid": 5, "tid": 7, "ts": 1716454223265633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201708, "dur": 6,
    "args": {
      "External id": 120822,
      "cbid": 211, "correlation": 120822
    }
  },
  {
    "ph": "s", "id": 120822, "pid": 76337, "tid": -914061504, "ts": 1716454223201708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223265733, "dur": 16,
    "args": {
      "External id": 120827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120827,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120827, "pid": 5, "tid": 7, "ts": 1716454223265733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201735, "dur": 8,
    "args": {
      "External id": 120827,
      "cbid": 211, "correlation": 120827
    }
  },
  {
    "ph": "s", "id": 120827, "pid": 76337, "tid": -914061504, "ts": 1716454223201735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223265750, "dur": 13,
    "args": {
      "External id": 120835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120835,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120835, "pid": 5, "tid": 7, "ts": 1716454223265750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201767, "dur": 8,
    "args": {
      "External id": 120835,
      "cbid": 211, "correlation": 120835
    }
  },
  {
    "ph": "s", "id": 120835, "pid": 76337, "tid": -914061504, "ts": 1716454223201767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223265764, "dur": 57,
    "args": {
      "External id": 120846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120846,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120846, "pid": 5, "tid": 7, "ts": 1716454223265764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201828, "dur": 12,
    "args": {
      "External id": 120846,
      "cbid": 211, "correlation": 120846
    }
  },
  {
    "ph": "s", "id": 120846, "pid": 76337, "tid": -914061504, "ts": 1716454223201828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201882, "dur": 0,
    "args": {
      "External id": 120856,
      "cbid": 317, "correlation": 120856
    }
  },
  {
    "ph": "f", "id": 120856, "pid": 76337, "tid": -914061504, "ts": 1716454223201882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201883, "dur": 0,
    "args": {
      "External id": 120857,
      "cbid": 203, "correlation": 120857
    }
  },
  {
    "ph": "f", "id": 120857, "pid": 76337, "tid": -914061504, "ts": 1716454223201883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201883, "dur": 0,
    "args": {
      "External id": 120858,
      "cbid": 205, "correlation": 120858
    }
  },
  {
    "ph": "f", "id": 120858, "pid": 76337, "tid": -914061504, "ts": 1716454223201883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223265822, "dur": 39,
    "args": {
      "External id": 120862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120862,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120862, "pid": 5, "tid": 7, "ts": 1716454223265822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201899, "dur": 12,
    "args": {
      "External id": 120862,
      "cbid": 211, "correlation": 120862
    }
  },
  {
    "ph": "s", "id": 120862, "pid": 76337, "tid": -914061504, "ts": 1716454223201899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223265863, "dur": 163,
    "args": {
      "External id": 120864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120864,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120864, "pid": 5, "tid": 7, "ts": 1716454223265863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201914, "dur": 5,
    "args": {
      "External id": 120864,
      "cbid": 211, "correlation": 120864
    }
  },
  {
    "ph": "s", "id": 120864, "pid": 76337, "tid": -914061504, "ts": 1716454223201914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223266027, "dur": 1960,
    "args": {
      "External id": 120866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120866,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120866, "pid": 5, "tid": 7, "ts": 1716454223266027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201925, "dur": 8,
    "args": {
      "External id": 120866,
      "cbid": 211, "correlation": 120866
    }
  },
  {
    "ph": "s", "id": 120866, "pid": 76337, "tid": -914061504, "ts": 1716454223201925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223267988, "dur": 38,
    "args": {
      "External id": 120868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120868, "pid": 5, "tid": 7, "ts": 1716454223267988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201937, "dur": 5,
    "args": {
      "External id": 120868,
      "cbid": 211, "correlation": 120868
    }
  },
  {
    "ph": "s", "id": 120868, "pid": 76337, "tid": -914061504, "ts": 1716454223201937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223268028, "dur": 60,
    "args": {
      "External id": 120874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120874,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120874, "pid": 5, "tid": 7, "ts": 1716454223268028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223201965, "dur": 16,
    "args": {
      "External id": 120874,
      "cbid": 211, "correlation": 120874
    }
  },
  {
    "ph": "s", "id": 120874, "pid": 76337, "tid": -914061504, "ts": 1716454223201965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223268089, "dur": 89,
    "args": {
      "External id": 120883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120883,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120883, "pid": 5, "tid": 7, "ts": 1716454223268089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202062, "dur": 14,
    "args": {
      "External id": 120883,
      "cbid": 211, "correlation": 120883
    }
  },
  {
    "ph": "s", "id": 120883, "pid": 76337, "tid": -914061504, "ts": 1716454223202062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223268179, "dur": 72,
    "args": {
      "External id": 120903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120903,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 120903, "pid": 5, "tid": 7, "ts": 1716454223268179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202131, "dur": 11,
    "args": {
      "External id": 120903,
      "cbid": 211, "correlation": 120903
    }
  },
  {
    "ph": "s", "id": 120903, "pid": 76337, "tid": -914061504, "ts": 1716454223202131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223268253, "dur": 4,
    "args": {
      "External id": 120915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120915,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 120915, "pid": 5, "tid": 7, "ts": 1716454223268253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202152, "dur": 6,
    "args": {
      "External id": 120915,
      "cbid": 211, "correlation": 120915
    }
  },
  {
    "ph": "s", "id": 120915, "pid": 76337, "tid": -914061504, "ts": 1716454223202152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223268259, "dur": 82,
    "args": {
      "External id": 120918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120918,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120918, "pid": 5, "tid": 7, "ts": 1716454223268259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202171, "dur": 6,
    "args": {
      "External id": 120918,
      "cbid": 211, "correlation": 120918
    }
  },
  {
    "ph": "s", "id": 120918, "pid": 76337, "tid": -914061504, "ts": 1716454223202171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223268342, "dur": 53,
    "args": {
      "External id": 120927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120927,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120927, "pid": 5, "tid": 7, "ts": 1716454223268342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202211, "dur": 10,
    "args": {
      "External id": 120927,
      "cbid": 211, "correlation": 120927
    }
  },
  {
    "ph": "s", "id": 120927, "pid": 76337, "tid": -914061504, "ts": 1716454223202211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202263, "dur": 0,
    "args": {
      "External id": 120937,
      "cbid": 317, "correlation": 120937
    }
  },
  {
    "ph": "f", "id": 120937, "pid": 76337, "tid": -914061504, "ts": 1716454223202263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202264, "dur": 0,
    "args": {
      "External id": 120938,
      "cbid": 203, "correlation": 120938
    }
  },
  {
    "ph": "f", "id": 120938, "pid": 76337, "tid": -914061504, "ts": 1716454223202264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202265, "dur": 0,
    "args": {
      "External id": 120939,
      "cbid": 205, "correlation": 120939
    }
  },
  {
    "ph": "f", "id": 120939, "pid": 76337, "tid": -914061504, "ts": 1716454223202265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223268396, "dur": 57,
    "args": {
      "External id": 120943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120943,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120943, "pid": 5, "tid": 7, "ts": 1716454223268396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202281, "dur": 11,
    "args": {
      "External id": 120943,
      "cbid": 211, "correlation": 120943
    }
  },
  {
    "ph": "s", "id": 120943, "pid": 76337, "tid": -914061504, "ts": 1716454223202281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223268454, "dur": 122,
    "args": {
      "External id": 120945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120945,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120945, "pid": 5, "tid": 7, "ts": 1716454223268454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202295, "dur": 5,
    "args": {
      "External id": 120945,
      "cbid": 211, "correlation": 120945
    }
  },
  {
    "ph": "s", "id": 120945, "pid": 76337, "tid": -914061504, "ts": 1716454223202295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223268577, "dur": 1893,
    "args": {
      "External id": 120947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120947,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 120947, "pid": 5, "tid": 7, "ts": 1716454223268577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202306, "dur": 6,
    "args": {
      "External id": 120947,
      "cbid": 211, "correlation": 120947
    }
  },
  {
    "ph": "s", "id": 120947, "pid": 76337, "tid": -914061504, "ts": 1716454223202306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223270471, "dur": 19,
    "args": {
      "External id": 120949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120949, "pid": 5, "tid": 7, "ts": 1716454223270471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202316, "dur": 5,
    "args": {
      "External id": 120949,
      "cbid": 211, "correlation": 120949
    }
  },
  {
    "ph": "s", "id": 120949, "pid": 76337, "tid": -914061504, "ts": 1716454223202316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223270492, "dur": 33,
    "args": {
      "External id": 120955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120955, "pid": 5, "tid": 7, "ts": 1716454223270492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202344, "dur": 8,
    "args": {
      "External id": 120955,
      "cbid": 211, "correlation": 120955
    }
  },
  {
    "ph": "s", "id": 120955, "pid": 76337, "tid": -914061504, "ts": 1716454223202344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223270526, "dur": 4,
    "args": {
      "External id": 120963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120963,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 120963, "pid": 5, "tid": 7, "ts": 1716454223270526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202388, "dur": 9,
    "args": {
      "External id": 120963,
      "cbid": 211, "correlation": 120963
    }
  },
  {
    "ph": "s", "id": 120963, "pid": 76337, "tid": -914061504, "ts": 1716454223202388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202452, "dur": 1,
    "args": {
      "External id": 120979,
      "cbid": 251, "correlation": 120979
    }
  },
  {
    "ph": "f", "id": 120979, "pid": 76337, "tid": -914061504, "ts": 1716454223202452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202457, "dur": 0,
    "args": {
      "External id": 120981,
      "cbid": 251, "correlation": 120981
    }
  },
  {
    "ph": "f", "id": 120981, "pid": 76337, "tid": -914061504, "ts": 1716454223202457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223270532, "dur": 12,
    "args": {
      "External id": 120982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120982,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 120982, "pid": 5, "tid": 7, "ts": 1716454223270532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202459, "dur": 12,
    "args": {
      "External id": 120982,
      "cbid": 211, "correlation": 120982
    }
  },
  {
    "ph": "s", "id": 120982, "pid": 76337, "tid": -914061504, "ts": 1716454223202459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223270545, "dur": 5,
    "args": {
      "External id": 120984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120984,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 120984, "pid": 5, "tid": 7, "ts": 1716454223270545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202473, "dur": 5,
    "args": {
      "External id": 120984,
      "cbid": 211, "correlation": 120984
    }
  },
  {
    "ph": "s", "id": 120984, "pid": 76337, "tid": -914061504, "ts": 1716454223202473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223270551, "dur": 29,
    "args": {
      "External id": 120994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 120994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 120994, "pid": 5, "tid": 7, "ts": 1716454223270551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202530, "dur": 12,
    "args": {
      "External id": 120994,
      "cbid": 211, "correlation": 120994
    }
  },
  {
    "ph": "s", "id": 120994, "pid": 76337, "tid": -914061504, "ts": 1716454223202530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223270582, "dur": 31,
    "args": {
      "External id": 121014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121014,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 121014, "pid": 5, "tid": 7, "ts": 1716454223270582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202595, "dur": 11,
    "args": {
      "External id": 121014,
      "cbid": 211, "correlation": 121014
    }
  },
  {
    "ph": "s", "id": 121014, "pid": 76337, "tid": -914061504, "ts": 1716454223202595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223270614, "dur": 4,
    "args": {
      "External id": 121026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121026,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 121026, "pid": 5, "tid": 7, "ts": 1716454223270614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202615, "dur": 7,
    "args": {
      "External id": 121026,
      "cbid": 211, "correlation": 121026
    }
  },
  {
    "ph": "s", "id": 121026, "pid": 76337, "tid": -914061504, "ts": 1716454223202615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223270619, "dur": 30,
    "args": {
      "External id": 121029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121029,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121029, "pid": 5, "tid": 7, "ts": 1716454223270619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202633, "dur": 6,
    "args": {
      "External id": 121029,
      "cbid": 211, "correlation": 121029
    }
  },
  {
    "ph": "s", "id": 121029, "pid": 76337, "tid": -914061504, "ts": 1716454223202633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223270651, "dur": 21,
    "args": {
      "External id": 121038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121038,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121038, "pid": 5, "tid": 7, "ts": 1716454223270651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202673, "dur": 10,
    "args": {
      "External id": 121038,
      "cbid": 211, "correlation": 121038
    }
  },
  {
    "ph": "s", "id": 121038, "pid": 76337, "tid": -914061504, "ts": 1716454223202673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202736, "dur": 0,
    "args": {
      "External id": 121048,
      "cbid": 317, "correlation": 121048
    }
  },
  {
    "ph": "f", "id": 121048, "pid": 76337, "tid": -914061504, "ts": 1716454223202736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202737, "dur": 0,
    "args": {
      "External id": 121049,
      "cbid": 203, "correlation": 121049
    }
  },
  {
    "ph": "f", "id": 121049, "pid": 76337, "tid": -914061504, "ts": 1716454223202737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202738, "dur": 0,
    "args": {
      "External id": 121050,
      "cbid": 205, "correlation": 121050
    }
  },
  {
    "ph": "f", "id": 121050, "pid": 76337, "tid": -914061504, "ts": 1716454223202738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223270673, "dur": 23,
    "args": {
      "External id": 121054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121054, "pid": 5, "tid": 7, "ts": 1716454223270673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202753, "dur": 12,
    "args": {
      "External id": 121054,
      "cbid": 211, "correlation": 121054
    }
  },
  {
    "ph": "s", "id": 121054, "pid": 76337, "tid": -914061504, "ts": 1716454223202753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223270697, "dur": 44,
    "args": {
      "External id": 121056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121056, "pid": 5, "tid": 7, "ts": 1716454223270697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202768, "dur": 5,
    "args": {
      "External id": 121056,
      "cbid": 211, "correlation": 121056
    }
  },
  {
    "ph": "s", "id": 121056, "pid": 76337, "tid": -914061504, "ts": 1716454223202768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223270742, "dur": 647,
    "args": {
      "External id": 121058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121058,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121058, "pid": 5, "tid": 7, "ts": 1716454223270742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202781, "dur": 6,
    "args": {
      "External id": 121058,
      "cbid": 211, "correlation": 121058
    }
  },
  {
    "ph": "s", "id": 121058, "pid": 76337, "tid": -914061504, "ts": 1716454223202781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223271390, "dur": 24,
    "args": {
      "External id": 121060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121060,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121060, "pid": 5, "tid": 7, "ts": 1716454223271390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202790, "dur": 5,
    "args": {
      "External id": 121060,
      "cbid": 211, "correlation": 121060
    }
  },
  {
    "ph": "s", "id": 121060, "pid": 76337, "tid": -914061504, "ts": 1716454223202790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223271415, "dur": 33,
    "args": {
      "External id": 121066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121066, "pid": 5, "tid": 7, "ts": 1716454223271415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202818, "dur": 9,
    "args": {
      "External id": 121066,
      "cbid": 211, "correlation": 121066
    }
  },
  {
    "ph": "s", "id": 121066, "pid": 76337, "tid": -914061504, "ts": 1716454223202818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202876, "dur": 0,
    "args": {
      "External id": 121076,
      "cbid": 317, "correlation": 121076
    }
  },
  {
    "ph": "f", "id": 121076, "pid": 76337, "tid": -914061504, "ts": 1716454223202876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202877, "dur": 0,
    "args": {
      "External id": 121077,
      "cbid": 203, "correlation": 121077
    }
  },
  {
    "ph": "f", "id": 121077, "pid": 76337, "tid": -914061504, "ts": 1716454223202877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202877, "dur": 0,
    "args": {
      "External id": 121078,
      "cbid": 205, "correlation": 121078
    }
  },
  {
    "ph": "f", "id": 121078, "pid": 76337, "tid": -914061504, "ts": 1716454223202877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223271449, "dur": 57,
    "args": {
      "External id": 121082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121082, "pid": 5, "tid": 7, "ts": 1716454223271449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202890, "dur": 12,
    "args": {
      "External id": 121082,
      "cbid": 211, "correlation": 121082
    }
  },
  {
    "ph": "s", "id": 121082, "pid": 76337, "tid": -914061504, "ts": 1716454223202890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223271507, "dur": 270,
    "args": {
      "External id": 121084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121084,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121084, "pid": 5, "tid": 7, "ts": 1716454223271507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202908, "dur": 9,
    "args": {
      "External id": 121084,
      "cbid": 211, "correlation": 121084
    }
  },
  {
    "ph": "s", "id": 121084, "pid": 76337, "tid": -914061504, "ts": 1716454223202908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223271778, "dur": 22,
    "args": {
      "External id": 121086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121086,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121086, "pid": 5, "tid": 7, "ts": 1716454223271778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202921, "dur": 5,
    "args": {
      "External id": 121086,
      "cbid": 211, "correlation": 121086
    }
  },
  {
    "ph": "s", "id": 121086, "pid": 76337, "tid": -914061504, "ts": 1716454223202921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223271801, "dur": 32,
    "args": {
      "External id": 121092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121092,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121092, "pid": 5, "tid": 7, "ts": 1716454223271801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202946, "dur": 8,
    "args": {
      "External id": 121092,
      "cbid": 211, "correlation": 121092
    }
  },
  {
    "ph": "s", "id": 121092, "pid": 76337, "tid": -914061504, "ts": 1716454223202946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223271835, "dur": 27,
    "args": {
      "External id": 121100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121100,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121100, "pid": 5, "tid": 7, "ts": 1716454223271835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223202983, "dur": 8,
    "args": {
      "External id": 121100,
      "cbid": 211, "correlation": 121100
    }
  },
  {
    "ph": "s", "id": 121100, "pid": 76337, "tid": -914061504, "ts": 1716454223202983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223271863, "dur": 20,
    "args": {
      "External id": 121108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121108,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121108, "pid": 5, "tid": 7, "ts": 1716454223271863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203013, "dur": 9,
    "args": {
      "External id": 121108,
      "cbid": 211, "correlation": 121108
    }
  },
  {
    "ph": "s", "id": 121108, "pid": 76337, "tid": -914061504, "ts": 1716454223203013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223271885, "dur": 30,
    "args": {
      "External id": 121128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121128,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 121128, "pid": 5, "tid": 7, "ts": 1716454223271885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203095, "dur": 12,
    "args": {
      "External id": 121128,
      "cbid": 211, "correlation": 121128
    }
  },
  {
    "ph": "s", "id": 121128, "pid": 76337, "tid": -914061504, "ts": 1716454223203095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223271916, "dur": 4,
    "args": {
      "External id": 121140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121140,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 121140, "pid": 5, "tid": 7, "ts": 1716454223271916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203117, "dur": 7,
    "args": {
      "External id": 121140,
      "cbid": 211, "correlation": 121140
    }
  },
  {
    "ph": "s", "id": 121140, "pid": 76337, "tid": -914061504, "ts": 1716454223203117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223271922, "dur": 31,
    "args": {
      "External id": 121143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121143,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121143, "pid": 5, "tid": 7, "ts": 1716454223271922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203135, "dur": 6,
    "args": {
      "External id": 121143,
      "cbid": 211, "correlation": 121143
    }
  },
  {
    "ph": "s", "id": 121143, "pid": 76337, "tid": -914061504, "ts": 1716454223203135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203192, "dur": 0,
    "args": {
      "External id": 121154,
      "cbid": 317, "correlation": 121154
    }
  },
  {
    "ph": "f", "id": 121154, "pid": 76337, "tid": -914061504, "ts": 1716454223203192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203193, "dur": 0,
    "args": {
      "External id": 121155,
      "cbid": 203, "correlation": 121155
    }
  },
  {
    "ph": "f", "id": 121155, "pid": 76337, "tid": -914061504, "ts": 1716454223203193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203194, "dur": 0,
    "args": {
      "External id": 121156,
      "cbid": 205, "correlation": 121156
    }
  },
  {
    "ph": "f", "id": 121156, "pid": 76337, "tid": -914061504, "ts": 1716454223203194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223271954, "dur": 22,
    "args": {
      "External id": 121160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121160,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121160, "pid": 5, "tid": 7, "ts": 1716454223271954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203208, "dur": 12,
    "args": {
      "External id": 121160,
      "cbid": 211, "correlation": 121160
    }
  },
  {
    "ph": "s", "id": 121160, "pid": 76337, "tid": -914061504, "ts": 1716454223203208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223271978, "dur": 105,
    "args": {
      "External id": 121162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121162,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121162, "pid": 5, "tid": 7, "ts": 1716454223271978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203226, "dur": 6,
    "args": {
      "External id": 121162,
      "cbid": 211, "correlation": 121162
    }
  },
  {
    "ph": "s", "id": 121162, "pid": 76337, "tid": -914061504, "ts": 1716454223203226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223272084, "dur": 23,
    "args": {
      "External id": 121164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121164,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121164, "pid": 5, "tid": 7, "ts": 1716454223272084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203236, "dur": 5,
    "args": {
      "External id": 121164,
      "cbid": 211, "correlation": 121164
    }
  },
  {
    "ph": "s", "id": 121164, "pid": 76337, "tid": -914061504, "ts": 1716454223203236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223272107, "dur": 32,
    "args": {
      "External id": 121170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121170,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121170, "pid": 5, "tid": 7, "ts": 1716454223272107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203263, "dur": 9,
    "args": {
      "External id": 121170,
      "cbid": 211, "correlation": 121170
    }
  },
  {
    "ph": "s", "id": 121170, "pid": 76337, "tid": -914061504, "ts": 1716454223203263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223272141, "dur": 190,
    "args": {
      "External id": 121179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121179,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121179, "pid": 5, "tid": 7, "ts": 1716454223272141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203344, "dur": 14,
    "args": {
      "External id": 121179,
      "cbid": 211, "correlation": 121179
    }
  },
  {
    "ph": "s", "id": 121179, "pid": 76337, "tid": -914061504, "ts": 1716454223203344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223272333, "dur": 65,
    "args": {
      "External id": 121201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121201,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121201, "pid": 5, "tid": 7, "ts": 1716454223272333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203402, "dur": 10,
    "args": {
      "External id": 121201,
      "cbid": 211, "correlation": 121201
    }
  },
  {
    "ph": "s", "id": 121201, "pid": 76337, "tid": -914061504, "ts": 1716454223203402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203491, "dur": 2,
    "args": {
      "External id": 121212,
      "cbid": 251, "correlation": 121212
    }
  },
  {
    "ph": "f", "id": 121212, "pid": 76337, "tid": -914061504, "ts": 1716454223203491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223272399, "dur": 154,
    "args": {
      "External id": 121213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121213,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121213, "pid": 5, "tid": 7, "ts": 1716454223272399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203496, "dur": 13,
    "args": {
      "External id": 121213,
      "cbid": 211, "correlation": 121213
    }
  },
  {
    "ph": "s", "id": 121213, "pid": 76337, "tid": -914061504, "ts": 1716454223203496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203566, "dur": 1,
    "args": {
      "External id": 121224,
      "cbid": 251, "correlation": 121224
    }
  },
  {
    "ph": "f", "id": 121224, "pid": 76337, "tid": -914061504, "ts": 1716454223203566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223272555, "dur": 148,
    "args": {
      "External id": 121225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121225,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121225, "pid": 5, "tid": 7, "ts": 1716454223272555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203570, "dur": 12,
    "args": {
      "External id": 121225,
      "cbid": 211, "correlation": 121225
    }
  },
  {
    "ph": "s", "id": 121225, "pid": 76337, "tid": -914061504, "ts": 1716454223203570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203635, "dur": 1,
    "args": {
      "External id": 121236,
      "cbid": 251, "correlation": 121236
    }
  },
  {
    "ph": "f", "id": 121236, "pid": 76337, "tid": -914061504, "ts": 1716454223203635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223272704, "dur": 147,
    "args": {
      "External id": 121237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121237,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121237, "pid": 5, "tid": 7, "ts": 1716454223272704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203639, "dur": 11,
    "args": {
      "External id": 121237,
      "cbid": 211, "correlation": 121237
    }
  },
  {
    "ph": "s", "id": 121237, "pid": 76337, "tid": -914061504, "ts": 1716454223203639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223272852, "dur": 1949,
    "args": {
      "External id": 121258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121258,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 121258, "pid": 5, "tid": 7, "ts": 1716454223272852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203721, "dur": 13,
    "args": {
      "External id": 121258,
      "cbid": 211, "correlation": 121258
    }
  },
  {
    "ph": "s", "id": 121258, "pid": 76337, "tid": -914061504, "ts": 1716454223203721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203819, "dur": 2,
    "args": {
      "External id": 121276,
      "cbid": 251, "correlation": 121276
    }
  },
  {
    "ph": "f", "id": 121276, "pid": 76337, "tid": -914061504, "ts": 1716454223203819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223274802, "dur": 149,
    "args": {
      "External id": 121278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121278,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 121278, "pid": 5, "tid": 7, "ts": 1716454223274802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203826, "dur": 14,
    "args": {
      "External id": 121278,
      "cbid": 211, "correlation": 121278
    }
  },
  {
    "ph": "s", "id": 121278, "pid": 76337, "tid": -914061504, "ts": 1716454223203826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223274953, "dur": 36,
    "args": {
      "External id": 121286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121286,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121286, "pid": 5, "tid": 7, "ts": 1716454223274953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203897, "dur": 12,
    "args": {
      "External id": 121286,
      "cbid": 211, "correlation": 121286
    }
  },
  {
    "ph": "s", "id": 121286, "pid": 76337, "tid": -914061504, "ts": 1716454223203897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223274990, "dur": 50,
    "args": {
      "External id": 121294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121294,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121294, "pid": 5, "tid": 7, "ts": 1716454223274990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223203937, "dur": 9,
    "args": {
      "External id": 121294,
      "cbid": 211, "correlation": 121294
    }
  },
  {
    "ph": "s", "id": 121294, "pid": 76337, "tid": -914061504, "ts": 1716454223203937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223275041, "dur": 30,
    "args": {
      "External id": 121305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121305,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121305, "pid": 5, "tid": 7, "ts": 1716454223275041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204017, "dur": 14,
    "args": {
      "External id": 121305,
      "cbid": 211, "correlation": 121305
    }
  },
  {
    "ph": "s", "id": 121305, "pid": 76337, "tid": -914061504, "ts": 1716454223204017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223275073, "dur": 35,
    "args": {
      "External id": 121327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121327,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121327, "pid": 5, "tid": 7, "ts": 1716454223275073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204049, "dur": 8,
    "args": {
      "External id": 121327,
      "cbid": 211, "correlation": 121327
    }
  },
  {
    "ph": "s", "id": 121327, "pid": 76337, "tid": -914061504, "ts": 1716454223204049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204134, "dur": 1,
    "args": {
      "External id": 121338,
      "cbid": 251, "correlation": 121338
    }
  },
  {
    "ph": "f", "id": 121338, "pid": 76337, "tid": -914061504, "ts": 1716454223204134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223275109, "dur": 91,
    "args": {
      "External id": 121339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121339,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121339, "pid": 5, "tid": 7, "ts": 1716454223275109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204139, "dur": 13,
    "args": {
      "External id": 121339,
      "cbid": 211, "correlation": 121339
    }
  },
  {
    "ph": "s", "id": 121339, "pid": 76337, "tid": -914061504, "ts": 1716454223204139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204208, "dur": 1,
    "args": {
      "External id": 121350,
      "cbid": 251, "correlation": 121350
    }
  },
  {
    "ph": "f", "id": 121350, "pid": 76337, "tid": -914061504, "ts": 1716454223204208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204212, "dur": 0,
    "args": {
      "External id": 121351,
      "cbid": 251, "correlation": 121351
    }
  },
  {
    "ph": "f", "id": 121351, "pid": 76337, "tid": -914061504, "ts": 1716454223204212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223275201, "dur": 11,
    "args": {
      "External id": 121352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121352,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 121352, "pid": 5, "tid": 7, "ts": 1716454223275201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204214, "dur": 12,
    "args": {
      "External id": 121352,
      "cbid": 211, "correlation": 121352
    }
  },
  {
    "ph": "s", "id": 121352, "pid": 76337, "tid": -914061504, "ts": 1716454223204214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223275214, "dur": 5,
    "args": {
      "External id": 121354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121354,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 121354, "pid": 5, "tid": 7, "ts": 1716454223275214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204228, "dur": 6,
    "args": {
      "External id": 121354,
      "cbid": 211, "correlation": 121354
    }
  },
  {
    "ph": "s", "id": 121354, "pid": 76337, "tid": -914061504, "ts": 1716454223204228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204285, "dur": 1,
    "args": {
      "External id": 121365,
      "cbid": 251, "correlation": 121365
    }
  },
  {
    "ph": "f", "id": 121365, "pid": 76337, "tid": -914061504, "ts": 1716454223204285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204288, "dur": 0,
    "args": {
      "External id": 121366,
      "cbid": 251, "correlation": 121366
    }
  },
  {
    "ph": "f", "id": 121366, "pid": 76337, "tid": -914061504, "ts": 1716454223204288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223275220, "dur": 7,
    "args": {
      "External id": 121367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121367,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 121367, "pid": 5, "tid": 7, "ts": 1716454223275220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204290, "dur": 11,
    "args": {
      "External id": 121367,
      "cbid": 211, "correlation": 121367
    }
  },
  {
    "ph": "s", "id": 121367, "pid": 76337, "tid": -914061504, "ts": 1716454223204290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223275228, "dur": 3,
    "args": {
      "External id": 121369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121369,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 121369, "pid": 5, "tid": 7, "ts": 1716454223275228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204303, "dur": 6,
    "args": {
      "External id": 121369,
      "cbid": 211, "correlation": 121369
    }
  },
  {
    "ph": "s", "id": 121369, "pid": 76337, "tid": -914061504, "ts": 1716454223204303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223275233, "dur": 93,
    "args": {
      "External id": 121390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121390,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 121390, "pid": 5, "tid": 7, "ts": 1716454223275233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204376, "dur": 13,
    "args": {
      "External id": 121390,
      "cbid": 211, "correlation": 121390
    }
  },
  {
    "ph": "s", "id": 121390, "pid": 76337, "tid": -914061504, "ts": 1716454223204376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204472, "dur": 1,
    "args": {
      "External id": 121408,
      "cbid": 251, "correlation": 121408
    }
  },
  {
    "ph": "f", "id": 121408, "pid": 76337, "tid": -914061504, "ts": 1716454223204472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223275327, "dur": 99,
    "args": {
      "External id": 121410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121410,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121410, "pid": 5, "tid": 7, "ts": 1716454223275327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204478, "dur": 13,
    "args": {
      "External id": 121410,
      "cbid": 211, "correlation": 121410
    }
  },
  {
    "ph": "s", "id": 121410, "pid": 76337, "tid": -914061504, "ts": 1716454223204478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223275427, "dur": 19,
    "args": {
      "External id": 121418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121418,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121418, "pid": 5, "tid": 7, "ts": 1716454223275427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204547, "dur": 12,
    "args": {
      "External id": 121418,
      "cbid": 211, "correlation": 121418
    }
  },
  {
    "ph": "s", "id": 121418, "pid": 76337, "tid": -914061504, "ts": 1716454223204547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223275448, "dur": 37,
    "args": {
      "External id": 121426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121426, "pid": 5, "tid": 7, "ts": 1716454223275448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204588, "dur": 9,
    "args": {
      "External id": 121426,
      "cbid": 211, "correlation": 121426
    }
  },
  {
    "ph": "s", "id": 121426, "pid": 76337, "tid": -914061504, "ts": 1716454223204588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223275486, "dur": 34,
    "args": {
      "External id": 121448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121448,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121448, "pid": 5, "tid": 7, "ts": 1716454223275486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204641, "dur": 9,
    "args": {
      "External id": 121448,
      "cbid": 211, "correlation": 121448
    }
  },
  {
    "ph": "s", "id": 121448, "pid": 76337, "tid": -914061504, "ts": 1716454223204641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204728, "dur": 1,
    "args": {
      "External id": 121464,
      "cbid": 251, "correlation": 121464
    }
  },
  {
    "ph": "f", "id": 121464, "pid": 76337, "tid": -914061504, "ts": 1716454223204728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204733, "dur": 0,
    "args": {
      "External id": 121466,
      "cbid": 251, "correlation": 121466
    }
  },
  {
    "ph": "f", "id": 121466, "pid": 76337, "tid": -914061504, "ts": 1716454223204733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223275522, "dur": 541,
    "args": {
      "External id": 121467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121467,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 121467, "pid": 5, "tid": 7, "ts": 1716454223275522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204739, "dur": 12,
    "args": {
      "External id": 121467,
      "cbid": 211, "correlation": 121467
    }
  },
  {
    "ph": "s", "id": 121467, "pid": 76337, "tid": -914061504, "ts": 1716454223204739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223276064, "dur": 126,
    "args": {
      "External id": 121475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121475,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121475, "pid": 5, "tid": 7, "ts": 1716454223276064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204803, "dur": 12,
    "args": {
      "External id": 121475,
      "cbid": 211, "correlation": 121475
    }
  },
  {
    "ph": "s", "id": 121475, "pid": 76337, "tid": -914061504, "ts": 1716454223204803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223276191, "dur": 130,
    "args": {
      "External id": 121483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121483, "pid": 5, "tid": 7, "ts": 1716454223276191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204834, "dur": 9,
    "args": {
      "External id": 121483,
      "cbid": 211, "correlation": 121483
    }
  },
  {
    "ph": "s", "id": 121483, "pid": 76337, "tid": -914061504, "ts": 1716454223204834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204911, "dur": 1,
    "args": {
      "External id": 121499,
      "cbid": 251, "correlation": 121499
    }
  },
  {
    "ph": "f", "id": 121499, "pid": 76337, "tid": -914061504, "ts": 1716454223204911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223276323, "dur": 305,
    "args": {
      "External id": 121501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121501,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121501, "pid": 5, "tid": 7, "ts": 1716454223276323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204917, "dur": 12,
    "args": {
      "External id": 121501,
      "cbid": 211, "correlation": 121501
    }
  },
  {
    "ph": "s", "id": 121501, "pid": 76337, "tid": -914061504, "ts": 1716454223204917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223276629, "dur": 27,
    "args": {
      "External id": 121509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121509,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121509, "pid": 5, "tid": 7, "ts": 1716454223276629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223204959, "dur": 9,
    "args": {
      "External id": 121509,
      "cbid": 211, "correlation": 121509
    }
  },
  {
    "ph": "s", "id": 121509, "pid": 76337, "tid": -914061504, "ts": 1716454223204959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223276657, "dur": 81,
    "args": {
      "External id": 121520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121520, "pid": 5, "tid": 7, "ts": 1716454223276657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205035, "dur": 13,
    "args": {
      "External id": 121520,
      "cbid": 211, "correlation": 121520
    }
  },
  {
    "ph": "s", "id": 121520, "pid": 76337, "tid": -914061504, "ts": 1716454223205035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205101, "dur": 0,
    "args": {
      "External id": 121532,
      "cbid": 317, "correlation": 121532
    }
  },
  {
    "ph": "f", "id": 121532, "pid": 76337, "tid": -914061504, "ts": 1716454223205101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205102, "dur": 0,
    "args": {
      "External id": 121533,
      "cbid": 203, "correlation": 121533
    }
  },
  {
    "ph": "f", "id": 121533, "pid": 76337, "tid": -914061504, "ts": 1716454223205102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205102, "dur": 0,
    "args": {
      "External id": 121534,
      "cbid": 205, "correlation": 121534
    }
  },
  {
    "ph": "f", "id": 121534, "pid": 76337, "tid": -914061504, "ts": 1716454223205102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223276739, "dur": 24,
    "args": {
      "External id": 121538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121538,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121538, "pid": 5, "tid": 7, "ts": 1716454223276739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205117, "dur": 12,
    "args": {
      "External id": 121538,
      "cbid": 211, "correlation": 121538
    }
  },
  {
    "ph": "s", "id": 121538, "pid": 76337, "tid": -914061504, "ts": 1716454223205117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223276764, "dur": 120,
    "args": {
      "External id": 121540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121540,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121540, "pid": 5, "tid": 7, "ts": 1716454223276764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205136, "dur": 7,
    "args": {
      "External id": 121540,
      "cbid": 211, "correlation": 121540
    }
  },
  {
    "ph": "s", "id": 121540, "pid": 76337, "tid": -914061504, "ts": 1716454223205136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223276885, "dur": 24,
    "args": {
      "External id": 121542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121542, "pid": 5, "tid": 7, "ts": 1716454223276885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205146, "dur": 5,
    "args": {
      "External id": 121542,
      "cbid": 211, "correlation": 121542
    }
  },
  {
    "ph": "s", "id": 121542, "pid": 76337, "tid": -914061504, "ts": 1716454223205146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223276911, "dur": 32,
    "args": {
      "External id": 121548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121548, "pid": 5, "tid": 7, "ts": 1716454223276911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205174, "dur": 9,
    "args": {
      "External id": 121548,
      "cbid": 211, "correlation": 121548
    }
  },
  {
    "ph": "s", "id": 121548, "pid": 76337, "tid": -914061504, "ts": 1716454223205174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223276945, "dur": 26,
    "args": {
      "External id": 121556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121556,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121556, "pid": 5, "tid": 7, "ts": 1716454223276945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205206, "dur": 8,
    "args": {
      "External id": 121556,
      "cbid": 211, "correlation": 121556
    }
  },
  {
    "ph": "s", "id": 121556, "pid": 76337, "tid": -914061504, "ts": 1716454223205206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223276972, "dur": 54,
    "args": {
      "External id": 121565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121565,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121565, "pid": 5, "tid": 7, "ts": 1716454223276972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205245, "dur": 10,
    "args": {
      "External id": 121565,
      "cbid": 211, "correlation": 121565
    }
  },
  {
    "ph": "s", "id": 121565, "pid": 76337, "tid": -914061504, "ts": 1716454223205245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223277028, "dur": 52,
    "args": {
      "External id": 121585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121585,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 121585, "pid": 5, "tid": 7, "ts": 1716454223277028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205315, "dur": 11,
    "args": {
      "External id": 121585,
      "cbid": 211, "correlation": 121585
    }
  },
  {
    "ph": "s", "id": 121585, "pid": 76337, "tid": -914061504, "ts": 1716454223205315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223277081, "dur": 4,
    "args": {
      "External id": 121597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121597,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 121597, "pid": 5, "tid": 7, "ts": 1716454223277081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205336, "dur": 7,
    "args": {
      "External id": 121597,
      "cbid": 211, "correlation": 121597
    }
  },
  {
    "ph": "s", "id": 121597, "pid": 76337, "tid": -914061504, "ts": 1716454223205336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223277087, "dur": 56,
    "args": {
      "External id": 121600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121600, "pid": 5, "tid": 7, "ts": 1716454223277087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205355, "dur": 7,
    "args": {
      "External id": 121600,
      "cbid": 211, "correlation": 121600
    }
  },
  {
    "ph": "s", "id": 121600, "pid": 76337, "tid": -914061504, "ts": 1716454223205355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223277144, "dur": 37,
    "args": {
      "External id": 121609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121609,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121609, "pid": 5, "tid": 7, "ts": 1716454223277144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205394, "dur": 10,
    "args": {
      "External id": 121609,
      "cbid": 211, "correlation": 121609
    }
  },
  {
    "ph": "s", "id": 121609, "pid": 76337, "tid": -914061504, "ts": 1716454223205394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205447, "dur": 0,
    "args": {
      "External id": 121619,
      "cbid": 317, "correlation": 121619
    }
  },
  {
    "ph": "f", "id": 121619, "pid": 76337, "tid": -914061504, "ts": 1716454223205447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205448, "dur": 0,
    "args": {
      "External id": 121620,
      "cbid": 203, "correlation": 121620
    }
  },
  {
    "ph": "f", "id": 121620, "pid": 76337, "tid": -914061504, "ts": 1716454223205448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205449, "dur": 0,
    "args": {
      "External id": 121621,
      "cbid": 205, "correlation": 121621
    }
  },
  {
    "ph": "f", "id": 121621, "pid": 76337, "tid": -914061504, "ts": 1716454223205449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223277182, "dur": 40,
    "args": {
      "External id": 121625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121625, "pid": 5, "tid": 7, "ts": 1716454223277182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205464, "dur": 11,
    "args": {
      "External id": 121625,
      "cbid": 211, "correlation": 121625
    }
  },
  {
    "ph": "s", "id": 121625, "pid": 76337, "tid": -914061504, "ts": 1716454223205464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223277223, "dur": 83,
    "args": {
      "External id": 121627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121627, "pid": 5, "tid": 7, "ts": 1716454223277223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205477, "dur": 5,
    "args": {
      "External id": 121627,
      "cbid": 211, "correlation": 121627
    }
  },
  {
    "ph": "s", "id": 121627, "pid": 76337, "tid": -914061504, "ts": 1716454223205477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223277308, "dur": 1279,
    "args": {
      "External id": 121629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121629,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121629, "pid": 5, "tid": 7, "ts": 1716454223277308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205489, "dur": 7,
    "args": {
      "External id": 121629,
      "cbid": 211, "correlation": 121629
    }
  },
  {
    "ph": "s", "id": 121629, "pid": 76337, "tid": -914061504, "ts": 1716454223205489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223278588, "dur": 22,
    "args": {
      "External id": 121631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121631, "pid": 5, "tid": 7, "ts": 1716454223278588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205500, "dur": 5,
    "args": {
      "External id": 121631,
      "cbid": 211, "correlation": 121631
    }
  },
  {
    "ph": "s", "id": 121631, "pid": 76337, "tid": -914061504, "ts": 1716454223205500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223278611, "dur": 33,
    "args": {
      "External id": 121637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121637, "pid": 5, "tid": 7, "ts": 1716454223278611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205527, "dur": 8,
    "args": {
      "External id": 121637,
      "cbid": 211, "correlation": 121637
    }
  },
  {
    "ph": "s", "id": 121637, "pid": 76337, "tid": -914061504, "ts": 1716454223205527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223278646, "dur": 4,
    "args": {
      "External id": 121645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121645,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 121645, "pid": 5, "tid": 7, "ts": 1716454223278646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205570, "dur": 9,
    "args": {
      "External id": 121645,
      "cbid": 211, "correlation": 121645
    }
  },
  {
    "ph": "s", "id": 121645, "pid": 76337, "tid": -914061504, "ts": 1716454223205570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205634, "dur": 1,
    "args": {
      "External id": 121661,
      "cbid": 251, "correlation": 121661
    }
  },
  {
    "ph": "f", "id": 121661, "pid": 76337, "tid": -914061504, "ts": 1716454223205634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205640, "dur": 0,
    "args": {
      "External id": 121663,
      "cbid": 251, "correlation": 121663
    }
  },
  {
    "ph": "f", "id": 121663, "pid": 76337, "tid": -914061504, "ts": 1716454223205640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223278652, "dur": 12,
    "args": {
      "External id": 121664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121664,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 121664, "pid": 5, "tid": 7, "ts": 1716454223278652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205642, "dur": 11,
    "args": {
      "External id": 121664,
      "cbid": 211, "correlation": 121664
    }
  },
  {
    "ph": "s", "id": 121664, "pid": 76337, "tid": -914061504, "ts": 1716454223205642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223278665, "dur": 5,
    "args": {
      "External id": 121666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121666,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 121666, "pid": 5, "tid": 7, "ts": 1716454223278665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205655, "dur": 5,
    "args": {
      "External id": 121666,
      "cbid": 211, "correlation": 121666
    }
  },
  {
    "ph": "s", "id": 121666, "pid": 76337, "tid": -914061504, "ts": 1716454223205655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223278672, "dur": 29,
    "args": {
      "External id": 121676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121676, "pid": 5, "tid": 7, "ts": 1716454223278672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205711, "dur": 12,
    "args": {
      "External id": 121676,
      "cbid": 211, "correlation": 121676
    }
  },
  {
    "ph": "s", "id": 121676, "pid": 76337, "tid": -914061504, "ts": 1716454223205711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223278702, "dur": 31,
    "args": {
      "External id": 121696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121696,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 121696, "pid": 5, "tid": 7, "ts": 1716454223278702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205777, "dur": 11,
    "args": {
      "External id": 121696,
      "cbid": 211, "correlation": 121696
    }
  },
  {
    "ph": "s", "id": 121696, "pid": 76337, "tid": -914061504, "ts": 1716454223205777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223278735, "dur": 4,
    "args": {
      "External id": 121708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121708,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 121708, "pid": 5, "tid": 7, "ts": 1716454223278735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205798, "dur": 6,
    "args": {
      "External id": 121708,
      "cbid": 211, "correlation": 121708
    }
  },
  {
    "ph": "s", "id": 121708, "pid": 76337, "tid": -914061504, "ts": 1716454223205798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223278741, "dur": 30,
    "args": {
      "External id": 121711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121711, "pid": 5, "tid": 7, "ts": 1716454223278741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205817, "dur": 6,
    "args": {
      "External id": 121711,
      "cbid": 211, "correlation": 121711
    }
  },
  {
    "ph": "s", "id": 121711, "pid": 76337, "tid": -914061504, "ts": 1716454223205817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223278772, "dur": 20,
    "args": {
      "External id": 121720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121720,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121720, "pid": 5, "tid": 7, "ts": 1716454223278772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205857, "dur": 12,
    "args": {
      "External id": 121720,
      "cbid": 211, "correlation": 121720
    }
  },
  {
    "ph": "s", "id": 121720, "pid": 76337, "tid": -914061504, "ts": 1716454223205857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205920, "dur": 0,
    "args": {
      "External id": 121730,
      "cbid": 317, "correlation": 121730
    }
  },
  {
    "ph": "f", "id": 121730, "pid": 76337, "tid": -914061504, "ts": 1716454223205920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205921, "dur": 0,
    "args": {
      "External id": 121731,
      "cbid": 203, "correlation": 121731
    }
  },
  {
    "ph": "f", "id": 121731, "pid": 76337, "tid": -914061504, "ts": 1716454223205921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205922, "dur": 0,
    "args": {
      "External id": 121732,
      "cbid": 205, "correlation": 121732
    }
  },
  {
    "ph": "f", "id": 121732, "pid": 76337, "tid": -914061504, "ts": 1716454223205922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223278793, "dur": 23,
    "args": {
      "External id": 121736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121736,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121736, "pid": 5, "tid": 7, "ts": 1716454223278793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205936, "dur": 13,
    "args": {
      "External id": 121736,
      "cbid": 211, "correlation": 121736
    }
  },
  {
    "ph": "s", "id": 121736, "pid": 76337, "tid": -914061504, "ts": 1716454223205936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223278817, "dur": 44,
    "args": {
      "External id": 121738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121738, "pid": 5, "tid": 7, "ts": 1716454223278817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205952, "dur": 5,
    "args": {
      "External id": 121738,
      "cbid": 211, "correlation": 121738
    }
  },
  {
    "ph": "s", "id": 121738, "pid": 76337, "tid": -914061504, "ts": 1716454223205952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223278863, "dur": 643,
    "args": {
      "External id": 121740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121740,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121740, "pid": 5, "tid": 7, "ts": 1716454223278863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205963, "dur": 6,
    "args": {
      "External id": 121740,
      "cbid": 211, "correlation": 121740
    }
  },
  {
    "ph": "s", "id": 121740, "pid": 76337, "tid": -914061504, "ts": 1716454223205963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223279508, "dur": 21,
    "args": {
      "External id": 121742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121742, "pid": 5, "tid": 7, "ts": 1716454223279508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223205972, "dur": 13,
    "args": {
      "External id": 121742,
      "cbid": 211, "correlation": 121742
    }
  },
  {
    "ph": "s", "id": 121742, "pid": 76337, "tid": -914061504, "ts": 1716454223205972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223279530, "dur": 33,
    "args": {
      "External id": 121748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121748,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121748, "pid": 5, "tid": 7, "ts": 1716454223279530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206009, "dur": 9,
    "args": {
      "External id": 121748,
      "cbid": 211, "correlation": 121748
    }
  },
  {
    "ph": "s", "id": 121748, "pid": 76337, "tid": -914061504, "ts": 1716454223206009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206067, "dur": 0,
    "args": {
      "External id": 121758,
      "cbid": 317, "correlation": 121758
    }
  },
  {
    "ph": "f", "id": 121758, "pid": 76337, "tid": -914061504, "ts": 1716454223206067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206068, "dur": 0,
    "args": {
      "External id": 121759,
      "cbid": 203, "correlation": 121759
    }
  },
  {
    "ph": "f", "id": 121759, "pid": 76337, "tid": -914061504, "ts": 1716454223206068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206069, "dur": 0,
    "args": {
      "External id": 121760,
      "cbid": 205, "correlation": 121760
    }
  },
  {
    "ph": "f", "id": 121760, "pid": 76337, "tid": -914061504, "ts": 1716454223206069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223279564, "dur": 38,
    "args": {
      "External id": 121764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121764,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121764, "pid": 5, "tid": 7, "ts": 1716454223279564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206081, "dur": 11,
    "args": {
      "External id": 121764,
      "cbid": 211, "correlation": 121764
    }
  },
  {
    "ph": "s", "id": 121764, "pid": 76337, "tid": -914061504, "ts": 1716454223206081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223279604, "dur": 191,
    "args": {
      "External id": 121766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121766,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121766, "pid": 5, "tid": 7, "ts": 1716454223279604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206099, "dur": 6,
    "args": {
      "External id": 121766,
      "cbid": 211, "correlation": 121766
    }
  },
  {
    "ph": "s", "id": 121766, "pid": 76337, "tid": -914061504, "ts": 1716454223206099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223279796, "dur": 22,
    "args": {
      "External id": 121768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121768, "pid": 5, "tid": 7, "ts": 1716454223279796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206108, "dur": 5,
    "args": {
      "External id": 121768,
      "cbid": 211, "correlation": 121768
    }
  },
  {
    "ph": "s", "id": 121768, "pid": 76337, "tid": -914061504, "ts": 1716454223206108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223279820, "dur": 33,
    "args": {
      "External id": 121774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121774, "pid": 5, "tid": 7, "ts": 1716454223279820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206133, "dur": 8,
    "args": {
      "External id": 121774,
      "cbid": 211, "correlation": 121774
    }
  },
  {
    "ph": "s", "id": 121774, "pid": 76337, "tid": -914061504, "ts": 1716454223206133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223279854, "dur": 27,
    "args": {
      "External id": 121782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121782,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121782, "pid": 5, "tid": 7, "ts": 1716454223279854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206163, "dur": 7,
    "args": {
      "External id": 121782,
      "cbid": 211, "correlation": 121782
    }
  },
  {
    "ph": "s", "id": 121782, "pid": 76337, "tid": -914061504, "ts": 1716454223206163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223279882, "dur": 20,
    "args": {
      "External id": 121790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121790,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121790, "pid": 5, "tid": 7, "ts": 1716454223279882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206191, "dur": 9,
    "args": {
      "External id": 121790,
      "cbid": 211, "correlation": 121790
    }
  },
  {
    "ph": "s", "id": 121790, "pid": 76337, "tid": -914061504, "ts": 1716454223206191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223279904, "dur": 30,
    "args": {
      "External id": 121810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121810,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 121810, "pid": 5, "tid": 7, "ts": 1716454223279904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206274, "dur": 12,
    "args": {
      "External id": 121810,
      "cbid": 211, "correlation": 121810
    }
  },
  {
    "ph": "s", "id": 121810, "pid": 76337, "tid": -914061504, "ts": 1716454223206274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223279935, "dur": 4,
    "args": {
      "External id": 121822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121822,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 121822, "pid": 5, "tid": 7, "ts": 1716454223279935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206296, "dur": 7,
    "args": {
      "External id": 121822,
      "cbid": 211, "correlation": 121822
    }
  },
  {
    "ph": "s", "id": 121822, "pid": 76337, "tid": -914061504, "ts": 1716454223206296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223279940, "dur": 30,
    "args": {
      "External id": 121825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121825,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121825, "pid": 5, "tid": 7, "ts": 1716454223279940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206315, "dur": 6,
    "args": {
      "External id": 121825,
      "cbid": 211, "correlation": 121825
    }
  },
  {
    "ph": "s", "id": 121825, "pid": 76337, "tid": -914061504, "ts": 1716454223206315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206371, "dur": 0,
    "args": {
      "External id": 121836,
      "cbid": 317, "correlation": 121836
    }
  },
  {
    "ph": "f", "id": 121836, "pid": 76337, "tid": -914061504, "ts": 1716454223206371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206372, "dur": 0,
    "args": {
      "External id": 121837,
      "cbid": 203, "correlation": 121837
    }
  },
  {
    "ph": "f", "id": 121837, "pid": 76337, "tid": -914061504, "ts": 1716454223206372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206373, "dur": 0,
    "args": {
      "External id": 121838,
      "cbid": 205, "correlation": 121838
    }
  },
  {
    "ph": "f", "id": 121838, "pid": 76337, "tid": -914061504, "ts": 1716454223206373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223279971, "dur": 22,
    "args": {
      "External id": 121842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121842,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121842, "pid": 5, "tid": 7, "ts": 1716454223279971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206387, "dur": 11,
    "args": {
      "External id": 121842,
      "cbid": 211, "correlation": 121842
    }
  },
  {
    "ph": "s", "id": 121842, "pid": 76337, "tid": -914061504, "ts": 1716454223206387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223279995, "dur": 105,
    "args": {
      "External id": 121844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121844,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121844, "pid": 5, "tid": 7, "ts": 1716454223279995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206405, "dur": 6,
    "args": {
      "External id": 121844,
      "cbid": 211, "correlation": 121844
    }
  },
  {
    "ph": "s", "id": 121844, "pid": 76337, "tid": -914061504, "ts": 1716454223206405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223280101, "dur": 21,
    "args": {
      "External id": 121846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121846, "pid": 5, "tid": 7, "ts": 1716454223280101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206414, "dur": 5,
    "args": {
      "External id": 121846,
      "cbid": 211, "correlation": 121846
    }
  },
  {
    "ph": "s", "id": 121846, "pid": 76337, "tid": -914061504, "ts": 1716454223206414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223280124, "dur": 32,
    "args": {
      "External id": 121852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121852, "pid": 5, "tid": 7, "ts": 1716454223280124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206441, "dur": 8,
    "args": {
      "External id": 121852,
      "cbid": 211, "correlation": 121852
    }
  },
  {
    "ph": "s", "id": 121852, "pid": 76337, "tid": -914061504, "ts": 1716454223206441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223280158, "dur": 191,
    "args": {
      "External id": 121861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121861,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121861, "pid": 5, "tid": 7, "ts": 1716454223280158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206521, "dur": 14,
    "args": {
      "External id": 121861,
      "cbid": 211, "correlation": 121861
    }
  },
  {
    "ph": "s", "id": 121861, "pid": 76337, "tid": -914061504, "ts": 1716454223206521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223280350, "dur": 65,
    "args": {
      "External id": 121883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121883,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121883, "pid": 5, "tid": 7, "ts": 1716454223280350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206578, "dur": 10,
    "args": {
      "External id": 121883,
      "cbid": 211, "correlation": 121883
    }
  },
  {
    "ph": "s", "id": 121883, "pid": 76337, "tid": -914061504, "ts": 1716454223206578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206666, "dur": 1,
    "args": {
      "External id": 121894,
      "cbid": 251, "correlation": 121894
    }
  },
  {
    "ph": "f", "id": 121894, "pid": 76337, "tid": -914061504, "ts": 1716454223206666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223280416, "dur": 154,
    "args": {
      "External id": 121895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121895,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121895, "pid": 5, "tid": 7, "ts": 1716454223280416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206671, "dur": 13,
    "args": {
      "External id": 121895,
      "cbid": 211, "correlation": 121895
    }
  },
  {
    "ph": "s", "id": 121895, "pid": 76337, "tid": -914061504, "ts": 1716454223206671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206742, "dur": 1,
    "args": {
      "External id": 121906,
      "cbid": 251, "correlation": 121906
    }
  },
  {
    "ph": "f", "id": 121906, "pid": 76337, "tid": -914061504, "ts": 1716454223206742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223280571, "dur": 145,
    "args": {
      "External id": 121907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121907,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121907, "pid": 5, "tid": 7, "ts": 1716454223280571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206746, "dur": 12,
    "args": {
      "External id": 121907,
      "cbid": 211, "correlation": 121907
    }
  },
  {
    "ph": "s", "id": 121907, "pid": 76337, "tid": -914061504, "ts": 1716454223206746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206813, "dur": 1,
    "args": {
      "External id": 121918,
      "cbid": 251, "correlation": 121918
    }
  },
  {
    "ph": "f", "id": 121918, "pid": 76337, "tid": -914061504, "ts": 1716454223206813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223280718, "dur": 147,
    "args": {
      "External id": 121919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121919,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 121919, "pid": 5, "tid": 7, "ts": 1716454223280718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206817, "dur": 11,
    "args": {
      "External id": 121919,
      "cbid": 211, "correlation": 121919
    }
  },
  {
    "ph": "s", "id": 121919, "pid": 76337, "tid": -914061504, "ts": 1716454223206817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223280866, "dur": 1946,
    "args": {
      "External id": 121940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121940,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 121940, "pid": 5, "tid": 7, "ts": 1716454223280866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223206896, "dur": 13,
    "args": {
      "External id": 121940,
      "cbid": 211, "correlation": 121940
    }
  },
  {
    "ph": "s", "id": 121940, "pid": 76337, "tid": -914061504, "ts": 1716454223206896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207001, "dur": 1,
    "args": {
      "External id": 121958,
      "cbid": 251, "correlation": 121958
    }
  },
  {
    "ph": "f", "id": 121958, "pid": 76337, "tid": -914061504, "ts": 1716454223207001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223282814, "dur": 149,
    "args": {
      "External id": 121960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121960,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 121960, "pid": 5, "tid": 7, "ts": 1716454223282814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207007, "dur": 14,
    "args": {
      "External id": 121960,
      "cbid": 211, "correlation": 121960
    }
  },
  {
    "ph": "s", "id": 121960, "pid": 76337, "tid": -914061504, "ts": 1716454223207007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223282964, "dur": 35,
    "args": {
      "External id": 121968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121968,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121968, "pid": 5, "tid": 7, "ts": 1716454223282964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207079, "dur": 12,
    "args": {
      "External id": 121968,
      "cbid": 211, "correlation": 121968
    }
  },
  {
    "ph": "s", "id": 121968, "pid": 76337, "tid": -914061504, "ts": 1716454223207079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223283001, "dur": 51,
    "args": {
      "External id": 121976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121976,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121976, "pid": 5, "tid": 7, "ts": 1716454223283001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207118, "dur": 9,
    "args": {
      "External id": 121976,
      "cbid": 211, "correlation": 121976
    }
  },
  {
    "ph": "s", "id": 121976, "pid": 76337, "tid": -914061504, "ts": 1716454223207118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223283053, "dur": 31,
    "args": {
      "External id": 121987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 121987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 121987, "pid": 5, "tid": 7, "ts": 1716454223283053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207190, "dur": 12,
    "args": {
      "External id": 121987,
      "cbid": 211, "correlation": 121987
    }
  },
  {
    "ph": "s", "id": 121987, "pid": 76337, "tid": -914061504, "ts": 1716454223207190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223283085, "dur": 35,
    "args": {
      "External id": 122009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122009,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122009, "pid": 5, "tid": 7, "ts": 1716454223283085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207222, "dur": 7,
    "args": {
      "External id": 122009,
      "cbid": 211, "correlation": 122009
    }
  },
  {
    "ph": "s", "id": 122009, "pid": 76337, "tid": -914061504, "ts": 1716454223207222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207306, "dur": 1,
    "args": {
      "External id": 122020,
      "cbid": 251, "correlation": 122020
    }
  },
  {
    "ph": "f", "id": 122020, "pid": 76337, "tid": -914061504, "ts": 1716454223207306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223283121, "dur": 92,
    "args": {
      "External id": 122021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122021,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122021, "pid": 5, "tid": 7, "ts": 1716454223283121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207311, "dur": 13,
    "args": {
      "External id": 122021,
      "cbid": 211, "correlation": 122021
    }
  },
  {
    "ph": "s", "id": 122021, "pid": 76337, "tid": -914061504, "ts": 1716454223207311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207379, "dur": 1,
    "args": {
      "External id": 122032,
      "cbid": 251, "correlation": 122032
    }
  },
  {
    "ph": "f", "id": 122032, "pid": 76337, "tid": -914061504, "ts": 1716454223207379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207382, "dur": 0,
    "args": {
      "External id": 122033,
      "cbid": 251, "correlation": 122033
    }
  },
  {
    "ph": "f", "id": 122033, "pid": 76337, "tid": -914061504, "ts": 1716454223207382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223283214, "dur": 11,
    "args": {
      "External id": 122034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122034,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 122034, "pid": 5, "tid": 7, "ts": 1716454223283214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207384, "dur": 12,
    "args": {
      "External id": 122034,
      "cbid": 211, "correlation": 122034
    }
  },
  {
    "ph": "s", "id": 122034, "pid": 76337, "tid": -914061504, "ts": 1716454223207384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223283227, "dur": 5,
    "args": {
      "External id": 122036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122036,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 122036, "pid": 5, "tid": 7, "ts": 1716454223283227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207397, "dur": 6,
    "args": {
      "External id": 122036,
      "cbid": 211, "correlation": 122036
    }
  },
  {
    "ph": "s", "id": 122036, "pid": 76337, "tid": -914061504, "ts": 1716454223207397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207454, "dur": 1,
    "args": {
      "External id": 122047,
      "cbid": 251, "correlation": 122047
    }
  },
  {
    "ph": "f", "id": 122047, "pid": 76337, "tid": -914061504, "ts": 1716454223207454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207457, "dur": 0,
    "args": {
      "External id": 122048,
      "cbid": 251, "correlation": 122048
    }
  },
  {
    "ph": "f", "id": 122048, "pid": 76337, "tid": -914061504, "ts": 1716454223207457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223283233, "dur": 7,
    "args": {
      "External id": 122049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122049,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 122049, "pid": 5, "tid": 7, "ts": 1716454223283233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207459, "dur": 12,
    "args": {
      "External id": 122049,
      "cbid": 211, "correlation": 122049
    }
  },
  {
    "ph": "s", "id": 122049, "pid": 76337, "tid": -914061504, "ts": 1716454223207459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223283242, "dur": 3,
    "args": {
      "External id": 122051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122051,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 122051, "pid": 5, "tid": 7, "ts": 1716454223283242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207472, "dur": 6,
    "args": {
      "External id": 122051,
      "cbid": 211, "correlation": 122051
    }
  },
  {
    "ph": "s", "id": 122051, "pid": 76337, "tid": -914061504, "ts": 1716454223207472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223283246, "dur": 91,
    "args": {
      "External id": 122072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122072,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 122072, "pid": 5, "tid": 7, "ts": 1716454223283246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207545, "dur": 12,
    "args": {
      "External id": 122072,
      "cbid": 211, "correlation": 122072
    }
  },
  {
    "ph": "s", "id": 122072, "pid": 76337, "tid": -914061504, "ts": 1716454223207545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207643, "dur": 1,
    "args": {
      "External id": 122090,
      "cbid": 251, "correlation": 122090
    }
  },
  {
    "ph": "f", "id": 122090, "pid": 76337, "tid": -914061504, "ts": 1716454223207643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223283339, "dur": 98,
    "args": {
      "External id": 122092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122092,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122092, "pid": 5, "tid": 7, "ts": 1716454223283339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207648, "dur": 13,
    "args": {
      "External id": 122092,
      "cbid": 211, "correlation": 122092
    }
  },
  {
    "ph": "s", "id": 122092, "pid": 76337, "tid": -914061504, "ts": 1716454223207648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223283438, "dur": 19,
    "args": {
      "External id": 122100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122100,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122100, "pid": 5, "tid": 7, "ts": 1716454223283438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207716, "dur": 13,
    "args": {
      "External id": 122100,
      "cbid": 211, "correlation": 122100
    }
  },
  {
    "ph": "s", "id": 122100, "pid": 76337, "tid": -914061504, "ts": 1716454223207716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223283458, "dur": 38,
    "args": {
      "External id": 122108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122108, "pid": 5, "tid": 7, "ts": 1716454223283458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207758, "dur": 9,
    "args": {
      "External id": 122108,
      "cbid": 211, "correlation": 122108
    }
  },
  {
    "ph": "s", "id": 122108, "pid": 76337, "tid": -914061504, "ts": 1716454223207758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223283497, "dur": 35,
    "args": {
      "External id": 122130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122130,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122130, "pid": 5, "tid": 7, "ts": 1716454223283497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207808, "dur": 10,
    "args": {
      "External id": 122130,
      "cbid": 211, "correlation": 122130
    }
  },
  {
    "ph": "s", "id": 122130, "pid": 76337, "tid": -914061504, "ts": 1716454223207808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207897, "dur": 1,
    "args": {
      "External id": 122146,
      "cbid": 251, "correlation": 122146
    }
  },
  {
    "ph": "f", "id": 122146, "pid": 76337, "tid": -914061504, "ts": 1716454223207897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207902, "dur": 0,
    "args": {
      "External id": 122148,
      "cbid": 251, "correlation": 122148
    }
  },
  {
    "ph": "f", "id": 122148, "pid": 76337, "tid": -914061504, "ts": 1716454223207902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223283533, "dur": 539,
    "args": {
      "External id": 122149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122149,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 122149, "pid": 5, "tid": 7, "ts": 1716454223283533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207905, "dur": 13,
    "args": {
      "External id": 122149,
      "cbid": 211, "correlation": 122149
    }
  },
  {
    "ph": "s", "id": 122149, "pid": 76337, "tid": -914061504, "ts": 1716454223207905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223284073, "dur": 126,
    "args": {
      "External id": 122157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122157,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122157, "pid": 5, "tid": 7, "ts": 1716454223284073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223207969, "dur": 21,
    "args": {
      "External id": 122157,
      "cbid": 211, "correlation": 122157
    }
  },
  {
    "ph": "s", "id": 122157, "pid": 76337, "tid": -914061504, "ts": 1716454223207969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223284201, "dur": 130,
    "args": {
      "External id": 122165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122165, "pid": 5, "tid": 7, "ts": 1716454223284201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208008, "dur": 8,
    "args": {
      "External id": 122165,
      "cbid": 211, "correlation": 122165
    }
  },
  {
    "ph": "s", "id": 122165, "pid": 76337, "tid": -914061504, "ts": 1716454223208008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208084, "dur": 1,
    "args": {
      "External id": 122181,
      "cbid": 251, "correlation": 122181
    }
  },
  {
    "ph": "f", "id": 122181, "pid": 76337, "tid": -914061504, "ts": 1716454223208084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223284332, "dur": 303,
    "args": {
      "External id": 122183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122183,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122183, "pid": 5, "tid": 7, "ts": 1716454223284332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208090, "dur": 12,
    "args": {
      "External id": 122183,
      "cbid": 211, "correlation": 122183
    }
  },
  {
    "ph": "s", "id": 122183, "pid": 76337, "tid": -914061504, "ts": 1716454223208090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223284636, "dur": 27,
    "args": {
      "External id": 122191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122191,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122191, "pid": 5, "tid": 7, "ts": 1716454223284636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208133, "dur": 10,
    "args": {
      "External id": 122191,
      "cbid": 211, "correlation": 122191
    }
  },
  {
    "ph": "s", "id": 122191, "pid": 76337, "tid": -914061504, "ts": 1716454223208133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223284664, "dur": 81,
    "args": {
      "External id": 122202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122202, "pid": 5, "tid": 7, "ts": 1716454223284664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208200, "dur": 12,
    "args": {
      "External id": 122202,
      "cbid": 211, "correlation": 122202
    }
  },
  {
    "ph": "s", "id": 122202, "pid": 76337, "tid": -914061504, "ts": 1716454223208200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208262, "dur": 0,
    "args": {
      "External id": 122214,
      "cbid": 317, "correlation": 122214
    }
  },
  {
    "ph": "f", "id": 122214, "pid": 76337, "tid": -914061504, "ts": 1716454223208262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208263, "dur": 0,
    "args": {
      "External id": 122215,
      "cbid": 203, "correlation": 122215
    }
  },
  {
    "ph": "f", "id": 122215, "pid": 76337, "tid": -914061504, "ts": 1716454223208263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208264, "dur": 0,
    "args": {
      "External id": 122216,
      "cbid": 205, "correlation": 122216
    }
  },
  {
    "ph": "f", "id": 122216, "pid": 76337, "tid": -914061504, "ts": 1716454223208264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223284747, "dur": 24,
    "args": {
      "External id": 122220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122220,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122220, "pid": 5, "tid": 7, "ts": 1716454223284747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208279, "dur": 12,
    "args": {
      "External id": 122220,
      "cbid": 211, "correlation": 122220
    }
  },
  {
    "ph": "s", "id": 122220, "pid": 76337, "tid": -914061504, "ts": 1716454223208279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223284772, "dur": 120,
    "args": {
      "External id": 122222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122222,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122222, "pid": 5, "tid": 7, "ts": 1716454223284772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208298, "dur": 6,
    "args": {
      "External id": 122222,
      "cbid": 211, "correlation": 122222
    }
  },
  {
    "ph": "s", "id": 122222, "pid": 76337, "tid": -914061504, "ts": 1716454223208298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223284893, "dur": 23,
    "args": {
      "External id": 122224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122224, "pid": 5, "tid": 7, "ts": 1716454223284893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208307, "dur": 5,
    "args": {
      "External id": 122224,
      "cbid": 211, "correlation": 122224
    }
  },
  {
    "ph": "s", "id": 122224, "pid": 76337, "tid": -914061504, "ts": 1716454223208307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223284918, "dur": 32,
    "args": {
      "External id": 122230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122230,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122230, "pid": 5, "tid": 7, "ts": 1716454223284918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208335, "dur": 8,
    "args": {
      "External id": 122230,
      "cbid": 211, "correlation": 122230
    }
  },
  {
    "ph": "s", "id": 122230, "pid": 76337, "tid": -914061504, "ts": 1716454223208335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223284952, "dur": 27,
    "args": {
      "External id": 122238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122238,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122238, "pid": 5, "tid": 7, "ts": 1716454223284952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208366, "dur": 8,
    "args": {
      "External id": 122238,
      "cbid": 211, "correlation": 122238
    }
  },
  {
    "ph": "s", "id": 122238, "pid": 76337, "tid": -914061504, "ts": 1716454223208366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223284980, "dur": 46,
    "args": {
      "External id": 122247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122247,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122247, "pid": 5, "tid": 7, "ts": 1716454223284980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208404, "dur": 10,
    "args": {
      "External id": 122247,
      "cbid": 211, "correlation": 122247
    }
  },
  {
    "ph": "s", "id": 122247, "pid": 76337, "tid": -914061504, "ts": 1716454223208404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223285027, "dur": 42,
    "args": {
      "External id": 122267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122267,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 122267, "pid": 5, "tid": 7, "ts": 1716454223285027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208475, "dur": 12,
    "args": {
      "External id": 122267,
      "cbid": 211, "correlation": 122267
    }
  },
  {
    "ph": "s", "id": 122267, "pid": 76337, "tid": -914061504, "ts": 1716454223208475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223285070, "dur": 5,
    "args": {
      "External id": 122279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122279,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 122279, "pid": 5, "tid": 7, "ts": 1716454223285070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208496, "dur": 6,
    "args": {
      "External id": 122279,
      "cbid": 211, "correlation": 122279
    }
  },
  {
    "ph": "s", "id": 122279, "pid": 76337, "tid": -914061504, "ts": 1716454223208496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223285076, "dur": 43,
    "args": {
      "External id": 122282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122282, "pid": 5, "tid": 7, "ts": 1716454223285076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208515, "dur": 7,
    "args": {
      "External id": 122282,
      "cbid": 211, "correlation": 122282
    }
  },
  {
    "ph": "s", "id": 122282, "pid": 76337, "tid": -914061504, "ts": 1716454223208515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223285120, "dur": 29,
    "args": {
      "External id": 122291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122291,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122291, "pid": 5, "tid": 7, "ts": 1716454223285120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208555, "dur": 10,
    "args": {
      "External id": 122291,
      "cbid": 211, "correlation": 122291
    }
  },
  {
    "ph": "s", "id": 122291, "pid": 76337, "tid": -914061504, "ts": 1716454223208555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208606, "dur": 0,
    "args": {
      "External id": 122301,
      "cbid": 317, "correlation": 122301
    }
  },
  {
    "ph": "f", "id": 122301, "pid": 76337, "tid": -914061504, "ts": 1716454223208606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208607, "dur": 0,
    "args": {
      "External id": 122302,
      "cbid": 203, "correlation": 122302
    }
  },
  {
    "ph": "f", "id": 122302, "pid": 76337, "tid": -914061504, "ts": 1716454223208607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208608, "dur": 0,
    "args": {
      "External id": 122303,
      "cbid": 205, "correlation": 122303
    }
  },
  {
    "ph": "f", "id": 122303, "pid": 76337, "tid": -914061504, "ts": 1716454223208608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223285151, "dur": 31,
    "args": {
      "External id": 122307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122307,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122307, "pid": 5, "tid": 7, "ts": 1716454223285151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208622, "dur": 12,
    "args": {
      "External id": 122307,
      "cbid": 211, "correlation": 122307
    }
  },
  {
    "ph": "s", "id": 122307, "pid": 76337, "tid": -914061504, "ts": 1716454223208622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223285183, "dur": 63,
    "args": {
      "External id": 122309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122309, "pid": 5, "tid": 7, "ts": 1716454223285183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208637, "dur": 5,
    "args": {
      "External id": 122309,
      "cbid": 211, "correlation": 122309
    }
  },
  {
    "ph": "s", "id": 122309, "pid": 76337, "tid": -914061504, "ts": 1716454223208637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223285247, "dur": 963,
    "args": {
      "External id": 122311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122311,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122311, "pid": 5, "tid": 7, "ts": 1716454223285247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208648, "dur": 6,
    "args": {
      "External id": 122311,
      "cbid": 211, "correlation": 122311
    }
  },
  {
    "ph": "s", "id": 122311, "pid": 76337, "tid": -914061504, "ts": 1716454223208648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223286212, "dur": 23,
    "args": {
      "External id": 122313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122313, "pid": 5, "tid": 7, "ts": 1716454223286212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208658, "dur": 5,
    "args": {
      "External id": 122313,
      "cbid": 211, "correlation": 122313
    }
  },
  {
    "ph": "s", "id": 122313, "pid": 76337, "tid": -914061504, "ts": 1716454223208658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223286236, "dur": 33,
    "args": {
      "External id": 122319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122319,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122319, "pid": 5, "tid": 7, "ts": 1716454223286236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208687, "dur": 8,
    "args": {
      "External id": 122319,
      "cbid": 211, "correlation": 122319
    }
  },
  {
    "ph": "s", "id": 122319, "pid": 76337, "tid": -914061504, "ts": 1716454223208687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223286270, "dur": 4,
    "args": {
      "External id": 122327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122327,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 122327, "pid": 5, "tid": 7, "ts": 1716454223286270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208730, "dur": 10,
    "args": {
      "External id": 122327,
      "cbid": 211, "correlation": 122327
    }
  },
  {
    "ph": "s", "id": 122327, "pid": 76337, "tid": -914061504, "ts": 1716454223208730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208795, "dur": 1,
    "args": {
      "External id": 122343,
      "cbid": 251, "correlation": 122343
    }
  },
  {
    "ph": "f", "id": 122343, "pid": 76337, "tid": -914061504, "ts": 1716454223208795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208800, "dur": 0,
    "args": {
      "External id": 122345,
      "cbid": 251, "correlation": 122345
    }
  },
  {
    "ph": "f", "id": 122345, "pid": 76337, "tid": -914061504, "ts": 1716454223208800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223286276, "dur": 11,
    "args": {
      "External id": 122346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122346,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 122346, "pid": 5, "tid": 7, "ts": 1716454223286276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208802, "dur": 11,
    "args": {
      "External id": 122346,
      "cbid": 211, "correlation": 122346
    }
  },
  {
    "ph": "s", "id": 122346, "pid": 76337, "tid": -914061504, "ts": 1716454223208802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223286289, "dur": 5,
    "args": {
      "External id": 122348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122348,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 122348, "pid": 5, "tid": 7, "ts": 1716454223286289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208814, "dur": 5,
    "args": {
      "External id": 122348,
      "cbid": 211, "correlation": 122348
    }
  },
  {
    "ph": "s", "id": 122348, "pid": 76337, "tid": -914061504, "ts": 1716454223208814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223286295, "dur": 29,
    "args": {
      "External id": 122358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122358, "pid": 5, "tid": 7, "ts": 1716454223286295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208872, "dur": 12,
    "args": {
      "External id": 122358,
      "cbid": 211, "correlation": 122358
    }
  },
  {
    "ph": "s", "id": 122358, "pid": 76337, "tid": -914061504, "ts": 1716454223208872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223286325, "dur": 31,
    "args": {
      "External id": 122378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122378,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 122378, "pid": 5, "tid": 7, "ts": 1716454223286325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208938, "dur": 11,
    "args": {
      "External id": 122378,
      "cbid": 211, "correlation": 122378
    }
  },
  {
    "ph": "s", "id": 122378, "pid": 76337, "tid": -914061504, "ts": 1716454223208938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223286357, "dur": 4,
    "args": {
      "External id": 122390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122390,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 122390, "pid": 5, "tid": 7, "ts": 1716454223286357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208957, "dur": 6,
    "args": {
      "External id": 122390,
      "cbid": 211, "correlation": 122390
    }
  },
  {
    "ph": "s", "id": 122390, "pid": 76337, "tid": -914061504, "ts": 1716454223208957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223286363, "dur": 30,
    "args": {
      "External id": 122393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122393,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122393, "pid": 5, "tid": 7, "ts": 1716454223286363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223208984, "dur": 8,
    "args": {
      "External id": 122393,
      "cbid": 211, "correlation": 122393
    }
  },
  {
    "ph": "s", "id": 122393, "pid": 76337, "tid": -914061504, "ts": 1716454223208984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223286394, "dur": 21,
    "args": {
      "External id": 122402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122402,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122402, "pid": 5, "tid": 7, "ts": 1716454223286394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209027, "dur": 10,
    "args": {
      "External id": 122402,
      "cbid": 211, "correlation": 122402
    }
  },
  {
    "ph": "s", "id": 122402, "pid": 76337, "tid": -914061504, "ts": 1716454223209027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209090, "dur": 0,
    "args": {
      "External id": 122412,
      "cbid": 317, "correlation": 122412
    }
  },
  {
    "ph": "f", "id": 122412, "pid": 76337, "tid": -914061504, "ts": 1716454223209090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209091, "dur": 0,
    "args": {
      "External id": 122413,
      "cbid": 203, "correlation": 122413
    }
  },
  {
    "ph": "f", "id": 122413, "pid": 76337, "tid": -914061504, "ts": 1716454223209091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209092, "dur": 0,
    "args": {
      "External id": 122414,
      "cbid": 205, "correlation": 122414
    }
  },
  {
    "ph": "f", "id": 122414, "pid": 76337, "tid": -914061504, "ts": 1716454223209092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223286416, "dur": 22,
    "args": {
      "External id": 122418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122418,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122418, "pid": 5, "tid": 7, "ts": 1716454223286416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209105, "dur": 12,
    "args": {
      "External id": 122418,
      "cbid": 211, "correlation": 122418
    }
  },
  {
    "ph": "s", "id": 122418, "pid": 76337, "tid": -914061504, "ts": 1716454223209105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223286439, "dur": 44,
    "args": {
      "External id": 122420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122420, "pid": 5, "tid": 7, "ts": 1716454223286439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209120, "dur": 5,
    "args": {
      "External id": 122420,
      "cbid": 211, "correlation": 122420
    }
  },
  {
    "ph": "s", "id": 122420, "pid": 76337, "tid": -914061504, "ts": 1716454223209120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223286485, "dur": 644,
    "args": {
      "External id": 122422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122422,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122422, "pid": 5, "tid": 7, "ts": 1716454223286485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209131, "dur": 6,
    "args": {
      "External id": 122422,
      "cbid": 211, "correlation": 122422
    }
  },
  {
    "ph": "s", "id": 122422, "pid": 76337, "tid": -914061504, "ts": 1716454223209131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287130, "dur": 21,
    "args": {
      "External id": 122424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122424,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122424, "pid": 5, "tid": 7, "ts": 1716454223287130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209142, "dur": 5,
    "args": {
      "External id": 122424,
      "cbid": 211, "correlation": 122424
    }
  },
  {
    "ph": "s", "id": 122424, "pid": 76337, "tid": -914061504, "ts": 1716454223209142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223287152, "dur": 32,
    "args": {
      "External id": 122430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122430, "pid": 5, "tid": 7, "ts": 1716454223287152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209169, "dur": 9,
    "args": {
      "External id": 122430,
      "cbid": 211, "correlation": 122430
    }
  },
  {
    "ph": "s", "id": 122430, "pid": 76337, "tid": -914061504, "ts": 1716454223209169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209227, "dur": 0,
    "args": {
      "External id": 122440,
      "cbid": 317, "correlation": 122440
    }
  },
  {
    "ph": "f", "id": 122440, "pid": 76337, "tid": -914061504, "ts": 1716454223209227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209227, "dur": 0,
    "args": {
      "External id": 122441,
      "cbid": 203, "correlation": 122441
    }
  },
  {
    "ph": "f", "id": 122441, "pid": 76337, "tid": -914061504, "ts": 1716454223209227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209228, "dur": 0,
    "args": {
      "External id": 122442,
      "cbid": 205, "correlation": 122442
    }
  },
  {
    "ph": "f", "id": 122442, "pid": 76337, "tid": -914061504, "ts": 1716454223209228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287186, "dur": 31,
    "args": {
      "External id": 122446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122446, "pid": 5, "tid": 7, "ts": 1716454223287186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209242, "dur": 11,
    "args": {
      "External id": 122446,
      "cbid": 211, "correlation": 122446
    }
  },
  {
    "ph": "s", "id": 122446, "pid": 76337, "tid": -914061504, "ts": 1716454223209242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223287218, "dur": 152,
    "args": {
      "External id": 122448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122448,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122448, "pid": 5, "tid": 7, "ts": 1716454223287218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209260, "dur": 6,
    "args": {
      "External id": 122448,
      "cbid": 211, "correlation": 122448
    }
  },
  {
    "ph": "s", "id": 122448, "pid": 76337, "tid": -914061504, "ts": 1716454223209260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287372, "dur": 23,
    "args": {
      "External id": 122450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122450,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122450, "pid": 5, "tid": 7, "ts": 1716454223287372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209270, "dur": 5,
    "args": {
      "External id": 122450,
      "cbid": 211, "correlation": 122450
    }
  },
  {
    "ph": "s", "id": 122450, "pid": 76337, "tid": -914061504, "ts": 1716454223209270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223287396, "dur": 32,
    "args": {
      "External id": 122456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122456, "pid": 5, "tid": 7, "ts": 1716454223287396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209296, "dur": 8,
    "args": {
      "External id": 122456,
      "cbid": 211, "correlation": 122456
    }
  },
  {
    "ph": "s", "id": 122456, "pid": 76337, "tid": -914061504, "ts": 1716454223209296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223287430, "dur": 27,
    "args": {
      "External id": 122464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122464,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122464, "pid": 5, "tid": 7, "ts": 1716454223287430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209324, "dur": 8,
    "args": {
      "External id": 122464,
      "cbid": 211, "correlation": 122464
    }
  },
  {
    "ph": "s", "id": 122464, "pid": 76337, "tid": -914061504, "ts": 1716454223209324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223287458, "dur": 20,
    "args": {
      "External id": 122472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122472,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122472, "pid": 5, "tid": 7, "ts": 1716454223287458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209352, "dur": 8,
    "args": {
      "External id": 122472,
      "cbid": 211, "correlation": 122472
    }
  },
  {
    "ph": "s", "id": 122472, "pid": 76337, "tid": -914061504, "ts": 1716454223209352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287479, "dur": 29,
    "args": {
      "External id": 122492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122492,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 122492, "pid": 5, "tid": 7, "ts": 1716454223287479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209434, "dur": 13,
    "args": {
      "External id": 122492,
      "cbid": 211, "correlation": 122492
    }
  },
  {
    "ph": "s", "id": 122492, "pid": 76337, "tid": -914061504, "ts": 1716454223209434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223287510, "dur": 4,
    "args": {
      "External id": 122504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122504,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 122504, "pid": 5, "tid": 7, "ts": 1716454223287510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209457, "dur": 6,
    "args": {
      "External id": 122504,
      "cbid": 211, "correlation": 122504
    }
  },
  {
    "ph": "s", "id": 122504, "pid": 76337, "tid": -914061504, "ts": 1716454223209457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223287516, "dur": 32,
    "args": {
      "External id": 122507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122507, "pid": 5, "tid": 7, "ts": 1716454223287516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209474, "dur": 6,
    "args": {
      "External id": 122507,
      "cbid": 211, "correlation": 122507
    }
  },
  {
    "ph": "s", "id": 122507, "pid": 76337, "tid": -914061504, "ts": 1716454223209474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209530, "dur": 0,
    "args": {
      "External id": 122518,
      "cbid": 317, "correlation": 122518
    }
  },
  {
    "ph": "f", "id": 122518, "pid": 76337, "tid": -914061504, "ts": 1716454223209530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209531, "dur": 0,
    "args": {
      "External id": 122519,
      "cbid": 203, "correlation": 122519
    }
  },
  {
    "ph": "f", "id": 122519, "pid": 76337, "tid": -914061504, "ts": 1716454223209531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209531, "dur": 0,
    "args": {
      "External id": 122520,
      "cbid": 205, "correlation": 122520
    }
  },
  {
    "ph": "f", "id": 122520, "pid": 76337, "tid": -914061504, "ts": 1716454223209531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287549, "dur": 21,
    "args": {
      "External id": 122524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122524, "pid": 5, "tid": 7, "ts": 1716454223287549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209545, "dur": 11,
    "args": {
      "External id": 122524,
      "cbid": 211, "correlation": 122524
    }
  },
  {
    "ph": "s", "id": 122524, "pid": 76337, "tid": -914061504, "ts": 1716454223209545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223287572, "dur": 105,
    "args": {
      "External id": 122526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122526,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122526, "pid": 5, "tid": 7, "ts": 1716454223287572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209562, "dur": 6,
    "args": {
      "External id": 122526,
      "cbid": 211, "correlation": 122526
    }
  },
  {
    "ph": "s", "id": 122526, "pid": 76337, "tid": -914061504, "ts": 1716454223209562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287678, "dur": 22,
    "args": {
      "External id": 122528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122528,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122528, "pid": 5, "tid": 7, "ts": 1716454223287678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209571, "dur": 5,
    "args": {
      "External id": 122528,
      "cbid": 211, "correlation": 122528
    }
  },
  {
    "ph": "s", "id": 122528, "pid": 76337, "tid": -914061504, "ts": 1716454223209571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223287701, "dur": 33,
    "args": {
      "External id": 122534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122534, "pid": 5, "tid": 7, "ts": 1716454223287701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209599, "dur": 8,
    "args": {
      "External id": 122534,
      "cbid": 211, "correlation": 122534
    }
  },
  {
    "ph": "s", "id": 122534, "pid": 76337, "tid": -914061504, "ts": 1716454223209599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223287735, "dur": 194,
    "args": {
      "External id": 122543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122543,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122543, "pid": 5, "tid": 7, "ts": 1716454223287735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209679, "dur": 14,
    "args": {
      "External id": 122543,
      "cbid": 211, "correlation": 122543
    }
  },
  {
    "ph": "s", "id": 122543, "pid": 76337, "tid": -914061504, "ts": 1716454223209679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223287930, "dur": 65,
    "args": {
      "External id": 122565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122565,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122565, "pid": 5, "tid": 7, "ts": 1716454223287930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209736, "dur": 11,
    "args": {
      "External id": 122565,
      "cbid": 211, "correlation": 122565
    }
  },
  {
    "ph": "s", "id": 122565, "pid": 76337, "tid": -914061504, "ts": 1716454223209736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209822, "dur": 1,
    "args": {
      "External id": 122576,
      "cbid": 251, "correlation": 122576
    }
  },
  {
    "ph": "f", "id": 122576, "pid": 76337, "tid": -914061504, "ts": 1716454223209822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223287997, "dur": 150,
    "args": {
      "External id": 122577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122577,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122577, "pid": 5, "tid": 7, "ts": 1716454223287997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209827, "dur": 13,
    "args": {
      "External id": 122577,
      "cbid": 211, "correlation": 122577
    }
  },
  {
    "ph": "s", "id": 122577, "pid": 76337, "tid": -914061504, "ts": 1716454223209827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209898, "dur": 1,
    "args": {
      "External id": 122588,
      "cbid": 251, "correlation": 122588
    }
  },
  {
    "ph": "f", "id": 122588, "pid": 76337, "tid": -914061504, "ts": 1716454223209898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223288149, "dur": 145,
    "args": {
      "External id": 122589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122589,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122589, "pid": 5, "tid": 7, "ts": 1716454223288149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209902, "dur": 11,
    "args": {
      "External id": 122589,
      "cbid": 211, "correlation": 122589
    }
  },
  {
    "ph": "s", "id": 122589, "pid": 76337, "tid": -914061504, "ts": 1716454223209902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209967, "dur": 1,
    "args": {
      "External id": 122600,
      "cbid": 251, "correlation": 122600
    }
  },
  {
    "ph": "f", "id": 122600, "pid": 76337, "tid": -914061504, "ts": 1716454223209967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223288295, "dur": 145,
    "args": {
      "External id": 122601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122601,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122601, "pid": 5, "tid": 7, "ts": 1716454223288295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223209971, "dur": 19,
    "args": {
      "External id": 122601,
      "cbid": 211, "correlation": 122601
    }
  },
  {
    "ph": "s", "id": 122601, "pid": 76337, "tid": -914061504, "ts": 1716454223209971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223288442, "dur": 1943,
    "args": {
      "External id": 122622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122622,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 122622, "pid": 5, "tid": 7, "ts": 1716454223288442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210061, "dur": 12,
    "args": {
      "External id": 122622,
      "cbid": 211, "correlation": 122622
    }
  },
  {
    "ph": "s", "id": 122622, "pid": 76337, "tid": -914061504, "ts": 1716454223210061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210158, "dur": 1,
    "args": {
      "External id": 122640,
      "cbid": 251, "correlation": 122640
    }
  },
  {
    "ph": "f", "id": 122640, "pid": 76337, "tid": -914061504, "ts": 1716454223210158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223290386, "dur": 148,
    "args": {
      "External id": 122642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122642,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 122642, "pid": 5, "tid": 7, "ts": 1716454223290386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210164, "dur": 13,
    "args": {
      "External id": 122642,
      "cbid": 211, "correlation": 122642
    }
  },
  {
    "ph": "s", "id": 122642, "pid": 76337, "tid": -914061504, "ts": 1716454223210164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223290535, "dur": 36,
    "args": {
      "External id": 122650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122650,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122650, "pid": 5, "tid": 7, "ts": 1716454223290535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210234, "dur": 13,
    "args": {
      "External id": 122650,
      "cbid": 211, "correlation": 122650
    }
  },
  {
    "ph": "s", "id": 122650, "pid": 76337, "tid": -914061504, "ts": 1716454223210234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223290573, "dur": 51,
    "args": {
      "External id": 122658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122658,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122658, "pid": 5, "tid": 7, "ts": 1716454223290573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210273, "dur": 9,
    "args": {
      "External id": 122658,
      "cbid": 211, "correlation": 122658
    }
  },
  {
    "ph": "s", "id": 122658, "pid": 76337, "tid": -914061504, "ts": 1716454223210273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223290625, "dur": 30,
    "args": {
      "External id": 122669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122669, "pid": 5, "tid": 7, "ts": 1716454223290625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210345, "dur": 12,
    "args": {
      "External id": 122669,
      "cbid": 211, "correlation": 122669
    }
  },
  {
    "ph": "s", "id": 122669, "pid": 76337, "tid": -914061504, "ts": 1716454223210345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223290657, "dur": 34,
    "args": {
      "External id": 122691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122691,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122691, "pid": 5, "tid": 7, "ts": 1716454223290657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210375, "dur": 8,
    "args": {
      "External id": 122691,
      "cbid": 211, "correlation": 122691
    }
  },
  {
    "ph": "s", "id": 122691, "pid": 76337, "tid": -914061504, "ts": 1716454223210375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210459, "dur": 1,
    "args": {
      "External id": 122702,
      "cbid": 251, "correlation": 122702
    }
  },
  {
    "ph": "f", "id": 122702, "pid": 76337, "tid": -914061504, "ts": 1716454223210459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223290692, "dur": 90,
    "args": {
      "External id": 122703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122703,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122703, "pid": 5, "tid": 7, "ts": 1716454223290692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210464, "dur": 13,
    "args": {
      "External id": 122703,
      "cbid": 211, "correlation": 122703
    }
  },
  {
    "ph": "s", "id": 122703, "pid": 76337, "tid": -914061504, "ts": 1716454223210464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210534, "dur": 1,
    "args": {
      "External id": 122714,
      "cbid": 251, "correlation": 122714
    }
  },
  {
    "ph": "f", "id": 122714, "pid": 76337, "tid": -914061504, "ts": 1716454223210534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210537, "dur": 0,
    "args": {
      "External id": 122715,
      "cbid": 251, "correlation": 122715
    }
  },
  {
    "ph": "f", "id": 122715, "pid": 76337, "tid": -914061504, "ts": 1716454223210537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223290784, "dur": 11,
    "args": {
      "External id": 122716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122716,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 122716, "pid": 5, "tid": 7, "ts": 1716454223290784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210539, "dur": 12,
    "args": {
      "External id": 122716,
      "cbid": 211, "correlation": 122716
    }
  },
  {
    "ph": "s", "id": 122716, "pid": 76337, "tid": -914061504, "ts": 1716454223210539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223290796, "dur": 5,
    "args": {
      "External id": 122718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122718,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 122718, "pid": 5, "tid": 7, "ts": 1716454223290796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210553, "dur": 6,
    "args": {
      "External id": 122718,
      "cbid": 211, "correlation": 122718
    }
  },
  {
    "ph": "s", "id": 122718, "pid": 76337, "tid": -914061504, "ts": 1716454223210553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210610, "dur": 1,
    "args": {
      "External id": 122729,
      "cbid": 251, "correlation": 122729
    }
  },
  {
    "ph": "f", "id": 122729, "pid": 76337, "tid": -914061504, "ts": 1716454223210610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210613, "dur": 0,
    "args": {
      "External id": 122730,
      "cbid": 251, "correlation": 122730
    }
  },
  {
    "ph": "f", "id": 122730, "pid": 76337, "tid": -914061504, "ts": 1716454223210613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223290802, "dur": 8,
    "args": {
      "External id": 122731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122731,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 122731, "pid": 5, "tid": 7, "ts": 1716454223290802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210615, "dur": 11,
    "args": {
      "External id": 122731,
      "cbid": 211, "correlation": 122731
    }
  },
  {
    "ph": "s", "id": 122731, "pid": 76337, "tid": -914061504, "ts": 1716454223210615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223290811, "dur": 3,
    "args": {
      "External id": 122733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122733,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 122733, "pid": 5, "tid": 7, "ts": 1716454223290811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210628, "dur": 5,
    "args": {
      "External id": 122733,
      "cbid": 211, "correlation": 122733
    }
  },
  {
    "ph": "s", "id": 122733, "pid": 76337, "tid": -914061504, "ts": 1716454223210628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223290816, "dur": 91,
    "args": {
      "External id": 122754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122754,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 122754, "pid": 5, "tid": 7, "ts": 1716454223290816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210702, "dur": 13,
    "args": {
      "External id": 122754,
      "cbid": 211, "correlation": 122754
    }
  },
  {
    "ph": "s", "id": 122754, "pid": 76337, "tid": -914061504, "ts": 1716454223210702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210799, "dur": 1,
    "args": {
      "External id": 122772,
      "cbid": 251, "correlation": 122772
    }
  },
  {
    "ph": "f", "id": 122772, "pid": 76337, "tid": -914061504, "ts": 1716454223210799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223290908, "dur": 85,
    "args": {
      "External id": 122774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122774,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122774, "pid": 5, "tid": 7, "ts": 1716454223290908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210805, "dur": 13,
    "args": {
      "External id": 122774,
      "cbid": 211, "correlation": 122774
    }
  },
  {
    "ph": "s", "id": 122774, "pid": 76337, "tid": -914061504, "ts": 1716454223210805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223290994, "dur": 19,
    "args": {
      "External id": 122782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122782,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122782, "pid": 5, "tid": 7, "ts": 1716454223290994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210874, "dur": 12,
    "args": {
      "External id": 122782,
      "cbid": 211, "correlation": 122782
    }
  },
  {
    "ph": "s", "id": 122782, "pid": 76337, "tid": -914061504, "ts": 1716454223210874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223291015, "dur": 38,
    "args": {
      "External id": 122790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122790,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122790, "pid": 5, "tid": 7, "ts": 1716454223291015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210916, "dur": 9,
    "args": {
      "External id": 122790,
      "cbid": 211, "correlation": 122790
    }
  },
  {
    "ph": "s", "id": 122790, "pid": 76337, "tid": -914061504, "ts": 1716454223210916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223291054, "dur": 34,
    "args": {
      "External id": 122812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122812,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122812, "pid": 5, "tid": 7, "ts": 1716454223291054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223210967, "dur": 17,
    "args": {
      "External id": 122812,
      "cbid": 211, "correlation": 122812
    }
  },
  {
    "ph": "s", "id": 122812, "pid": 76337, "tid": -914061504, "ts": 1716454223210967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211064, "dur": 1,
    "args": {
      "External id": 122828,
      "cbid": 251, "correlation": 122828
    }
  },
  {
    "ph": "f", "id": 122828, "pid": 76337, "tid": -914061504, "ts": 1716454223211064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211069, "dur": 0,
    "args": {
      "External id": 122830,
      "cbid": 251, "correlation": 122830
    }
  },
  {
    "ph": "f", "id": 122830, "pid": 76337, "tid": -914061504, "ts": 1716454223211069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223291090, "dur": 540,
    "args": {
      "External id": 122831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122831,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 122831, "pid": 5, "tid": 7, "ts": 1716454223291090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211072, "dur": 13,
    "args": {
      "External id": 122831,
      "cbid": 211, "correlation": 122831
    }
  },
  {
    "ph": "s", "id": 122831, "pid": 76337, "tid": -914061504, "ts": 1716454223211072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223291631, "dur": 126,
    "args": {
      "External id": 122839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122839, "pid": 5, "tid": 7, "ts": 1716454223291631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211137, "dur": 12,
    "args": {
      "External id": 122839,
      "cbid": 211, "correlation": 122839
    }
  },
  {
    "ph": "s", "id": 122839, "pid": 76337, "tid": -914061504, "ts": 1716454223211137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223291758, "dur": 129,
    "args": {
      "External id": 122847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122847, "pid": 5, "tid": 7, "ts": 1716454223291758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211167, "dur": 8,
    "args": {
      "External id": 122847,
      "cbid": 211, "correlation": 122847
    }
  },
  {
    "ph": "s", "id": 122847, "pid": 76337, "tid": -914061504, "ts": 1716454223211167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211243, "dur": 1,
    "args": {
      "External id": 122863,
      "cbid": 251, "correlation": 122863
    }
  },
  {
    "ph": "f", "id": 122863, "pid": 76337, "tid": -914061504, "ts": 1716454223211243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223291889, "dur": 308,
    "args": {
      "External id": 122865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122865,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122865, "pid": 5, "tid": 7, "ts": 1716454223291889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211249, "dur": 13,
    "args": {
      "External id": 122865,
      "cbid": 211, "correlation": 122865
    }
  },
  {
    "ph": "s", "id": 122865, "pid": 76337, "tid": -914061504, "ts": 1716454223211249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223292198, "dur": 27,
    "args": {
      "External id": 122873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122873,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122873, "pid": 5, "tid": 7, "ts": 1716454223292198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211291, "dur": 10,
    "args": {
      "External id": 122873,
      "cbid": 211, "correlation": 122873
    }
  },
  {
    "ph": "s", "id": 122873, "pid": 76337, "tid": -914061504, "ts": 1716454223211291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223292227, "dur": 81,
    "args": {
      "External id": 122884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122884,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122884, "pid": 5, "tid": 7, "ts": 1716454223292227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211357, "dur": 12,
    "args": {
      "External id": 122884,
      "cbid": 211, "correlation": 122884
    }
  },
  {
    "ph": "s", "id": 122884, "pid": 76337, "tid": -914061504, "ts": 1716454223211357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211421, "dur": 0,
    "args": {
      "External id": 122896,
      "cbid": 317, "correlation": 122896
    }
  },
  {
    "ph": "f", "id": 122896, "pid": 76337, "tid": -914061504, "ts": 1716454223211421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211421, "dur": 0,
    "args": {
      "External id": 122897,
      "cbid": 203, "correlation": 122897
    }
  },
  {
    "ph": "f", "id": 122897, "pid": 76337, "tid": -914061504, "ts": 1716454223211421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211422, "dur": 0,
    "args": {
      "External id": 122898,
      "cbid": 205, "correlation": 122898
    }
  },
  {
    "ph": "f", "id": 122898, "pid": 76337, "tid": -914061504, "ts": 1716454223211422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223292309, "dur": 22,
    "args": {
      "External id": 122902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122902,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122902, "pid": 5, "tid": 7, "ts": 1716454223292309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211437, "dur": 12,
    "args": {
      "External id": 122902,
      "cbid": 211, "correlation": 122902
    }
  },
  {
    "ph": "s", "id": 122902, "pid": 76337, "tid": -914061504, "ts": 1716454223211437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223292333, "dur": 121,
    "args": {
      "External id": 122904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122904,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122904, "pid": 5, "tid": 7, "ts": 1716454223292333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211456, "dur": 7,
    "args": {
      "External id": 122904,
      "cbid": 211, "correlation": 122904
    }
  },
  {
    "ph": "s", "id": 122904, "pid": 76337, "tid": -914061504, "ts": 1716454223211456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223292456, "dur": 23,
    "args": {
      "External id": 122906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122906,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122906, "pid": 5, "tid": 7, "ts": 1716454223292456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211467, "dur": 5,
    "args": {
      "External id": 122906,
      "cbid": 211, "correlation": 122906
    }
  },
  {
    "ph": "s", "id": 122906, "pid": 76337, "tid": -914061504, "ts": 1716454223211467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223292480, "dur": 33,
    "args": {
      "External id": 122912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122912, "pid": 5, "tid": 7, "ts": 1716454223292480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211494, "dur": 8,
    "args": {
      "External id": 122912,
      "cbid": 211, "correlation": 122912
    }
  },
  {
    "ph": "s", "id": 122912, "pid": 76337, "tid": -914061504, "ts": 1716454223211494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223292514, "dur": 26,
    "args": {
      "External id": 122920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122920,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122920, "pid": 5, "tid": 7, "ts": 1716454223292514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211525, "dur": 8,
    "args": {
      "External id": 122920,
      "cbid": 211, "correlation": 122920
    }
  },
  {
    "ph": "s", "id": 122920, "pid": 76337, "tid": -914061504, "ts": 1716454223211525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223292542, "dur": 102,
    "args": {
      "External id": 122931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122931,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122931, "pid": 5, "tid": 7, "ts": 1716454223292542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211586, "dur": 11,
    "args": {
      "External id": 122931,
      "cbid": 211, "correlation": 122931
    }
  },
  {
    "ph": "s", "id": 122931, "pid": 76337, "tid": -914061504, "ts": 1716454223211586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211641, "dur": 0,
    "args": {
      "External id": 122941,
      "cbid": 317, "correlation": 122941
    }
  },
  {
    "ph": "f", "id": 122941, "pid": 76337, "tid": -914061504, "ts": 1716454223211641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211641, "dur": 0,
    "args": {
      "External id": 122942,
      "cbid": 203, "correlation": 122942
    }
  },
  {
    "ph": "f", "id": 122942, "pid": 76337, "tid": -914061504, "ts": 1716454223211641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211642, "dur": 0,
    "args": {
      "External id": 122943,
      "cbid": 205, "correlation": 122943
    }
  },
  {
    "ph": "f", "id": 122943, "pid": 76337, "tid": -914061504, "ts": 1716454223211642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223292645, "dur": 76,
    "args": {
      "External id": 122947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122947, "pid": 5, "tid": 7, "ts": 1716454223292645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211656, "dur": 13,
    "args": {
      "External id": 122947,
      "cbid": 211, "correlation": 122947
    }
  },
  {
    "ph": "s", "id": 122947, "pid": 76337, "tid": -914061504, "ts": 1716454223211656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223292722, "dur": 44,
    "args": {
      "External id": 122949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122949, "pid": 5, "tid": 7, "ts": 1716454223292722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211671, "dur": 5,
    "args": {
      "External id": 122949,
      "cbid": 211, "correlation": 122949
    }
  },
  {
    "ph": "s", "id": 122949, "pid": 76337, "tid": -914061504, "ts": 1716454223211671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223292767, "dur": 4,
    "args": {
      "External id": 122951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122951,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 122951, "pid": 5, "tid": 7, "ts": 1716454223292767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211681, "dur": 6,
    "args": {
      "External id": 122951,
      "cbid": 211, "correlation": 122951
    }
  },
  {
    "ph": "s", "id": 122951, "pid": 76337, "tid": -914061504, "ts": 1716454223211681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211689, "dur": 0,
    "args": {
      "External id": 122952,
      "cbid": 51, "correlation": 122952
    }
  },
  {
    "ph": "s", "id": 122952, "pid": 76337, "tid": -914061504, "ts": 1716454223211689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223292773, "dur": 2225,
    "args": {
      "External id": 122953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122953,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 122953, "pid": 5, "tid": 7, "ts": 1716454223292773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211690, "dur": 5,
    "args": {
      "External id": 122953,
      "cbid": 211, "correlation": 122953
    }
  },
  {
    "ph": "s", "id": 122953, "pid": 76337, "tid": -914061504, "ts": 1716454223211690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223294999, "dur": 113,
    "args": {
      "External id": 122958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122958, "pid": 5, "tid": 7, "ts": 1716454223294999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211718, "dur": 8,
    "args": {
      "External id": 122958,
      "cbid": 211, "correlation": 122958
    }
  },
  {
    "ph": "s", "id": 122958, "pid": 76337, "tid": -914061504, "ts": 1716454223211718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223295114, "dur": 165,
    "args": {
      "External id": 122967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122967,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 122967, "pid": 5, "tid": 7, "ts": 1716454223295114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211808, "dur": 13,
    "args": {
      "External id": 122967,
      "cbid": 211, "correlation": 122967
    }
  },
  {
    "ph": "s", "id": 122967, "pid": 76337, "tid": -914061504, "ts": 1716454223211808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223295281, "dur": 128,
    "args": {
      "External id": 122987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122987,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 122987, "pid": 5, "tid": 7, "ts": 1716454223295281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211879, "dur": 11,
    "args": {
      "External id": 122987,
      "cbid": 211, "correlation": 122987
    }
  },
  {
    "ph": "s", "id": 122987, "pid": 76337, "tid": -914061504, "ts": 1716454223211879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223295410, "dur": 4,
    "args": {
      "External id": 122999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 122999,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 122999, "pid": 5, "tid": 7, "ts": 1716454223295410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211899, "dur": 6,
    "args": {
      "External id": 122999,
      "cbid": 211, "correlation": 122999
    }
  },
  {
    "ph": "s", "id": 122999, "pid": 76337, "tid": -914061504, "ts": 1716454223211899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223295416, "dur": 160,
    "args": {
      "External id": 123002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123002, "pid": 5, "tid": 7, "ts": 1716454223295416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211919, "dur": 6,
    "args": {
      "External id": 123002,
      "cbid": 211, "correlation": 123002
    }
  },
  {
    "ph": "s", "id": 123002, "pid": 76337, "tid": -914061504, "ts": 1716454223211919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223295578, "dur": 101,
    "args": {
      "External id": 123011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123011,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123011, "pid": 5, "tid": 7, "ts": 1716454223295578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223211958, "dur": 11,
    "args": {
      "External id": 123011,
      "cbid": 211, "correlation": 123011
    }
  },
  {
    "ph": "s", "id": 123011, "pid": 76337, "tid": -914061504, "ts": 1716454223211958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212021, "dur": 0,
    "args": {
      "External id": 123021,
      "cbid": 317, "correlation": 123021
    }
  },
  {
    "ph": "f", "id": 123021, "pid": 76337, "tid": -914061504, "ts": 1716454223212021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212022, "dur": 0,
    "args": {
      "External id": 123022,
      "cbid": 203, "correlation": 123022
    }
  },
  {
    "ph": "f", "id": 123022, "pid": 76337, "tid": -914061504, "ts": 1716454223212022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212023, "dur": 0,
    "args": {
      "External id": 123023,
      "cbid": 205, "correlation": 123023
    }
  },
  {
    "ph": "f", "id": 123023, "pid": 76337, "tid": -914061504, "ts": 1716454223212023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223295680, "dur": 111,
    "args": {
      "External id": 123027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123027,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123027, "pid": 5, "tid": 7, "ts": 1716454223295680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212039, "dur": 12,
    "args": {
      "External id": 123027,
      "cbid": 211, "correlation": 123027
    }
  },
  {
    "ph": "s", "id": 123027, "pid": 76337, "tid": -914061504, "ts": 1716454223212039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223295792, "dur": 34,
    "args": {
      "External id": 123029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123029,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123029, "pid": 5, "tid": 7, "ts": 1716454223295792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212053, "dur": 5,
    "args": {
      "External id": 123029,
      "cbid": 211, "correlation": 123029
    }
  },
  {
    "ph": "s", "id": 123029, "pid": 76337, "tid": -914061504, "ts": 1716454223212053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223295827, "dur": 4,
    "args": {
      "External id": 123031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123031, "pid": 5, "tid": 7, "ts": 1716454223295827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212064, "dur": 6,
    "args": {
      "External id": 123031,
      "cbid": 211, "correlation": 123031
    }
  },
  {
    "ph": "s", "id": 123031, "pid": 76337, "tid": -914061504, "ts": 1716454223212064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212072, "dur": 0,
    "args": {
      "External id": 123032,
      "cbid": 51, "correlation": 123032
    }
  },
  {
    "ph": "s", "id": 123032, "pid": 76337, "tid": -914061504, "ts": 1716454223212072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223295832, "dur": 2007,
    "args": {
      "External id": 123033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123033,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123033, "pid": 5, "tid": 7, "ts": 1716454223295832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212073, "dur": 6,
    "args": {
      "External id": 123033,
      "cbid": 211, "correlation": 123033
    }
  },
  {
    "ph": "s", "id": 123033, "pid": 76337, "tid": -914061504, "ts": 1716454223212073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223297840, "dur": 59,
    "args": {
      "External id": 123038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123038, "pid": 5, "tid": 7, "ts": 1716454223297840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212102, "dur": 10,
    "args": {
      "External id": 123038,
      "cbid": 211, "correlation": 123038
    }
  },
  {
    "ph": "s", "id": 123038, "pid": 76337, "tid": -914061504, "ts": 1716454223212102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223297901, "dur": 4,
    "args": {
      "External id": 123046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123046,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123046, "pid": 5, "tid": 7, "ts": 1716454223297901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212146, "dur": 9,
    "args": {
      "External id": 123046,
      "cbid": 211, "correlation": 123046
    }
  },
  {
    "ph": "s", "id": 123046, "pid": 76337, "tid": -914061504, "ts": 1716454223212146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212213, "dur": 2,
    "args": {
      "External id": 123062,
      "cbid": 251, "correlation": 123062
    }
  },
  {
    "ph": "f", "id": 123062, "pid": 76337, "tid": -914061504, "ts": 1716454223212213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212219, "dur": 0,
    "args": {
      "External id": 123064,
      "cbid": 251, "correlation": 123064
    }
  },
  {
    "ph": "f", "id": 123064, "pid": 76337, "tid": -914061504, "ts": 1716454223212219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223297906, "dur": 11,
    "args": {
      "External id": 123065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123065,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 123065, "pid": 5, "tid": 7, "ts": 1716454223297906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212221, "dur": 11,
    "args": {
      "External id": 123065,
      "cbid": 211, "correlation": 123065
    }
  },
  {
    "ph": "s", "id": 123065, "pid": 76337, "tid": -914061504, "ts": 1716454223212221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223297919, "dur": 5,
    "args": {
      "External id": 123067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123067,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 123067, "pid": 5, "tid": 7, "ts": 1716454223297919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212234, "dur": 5,
    "args": {
      "External id": 123067,
      "cbid": 211, "correlation": 123067
    }
  },
  {
    "ph": "s", "id": 123067, "pid": 76337, "tid": -914061504, "ts": 1716454223212234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223297925, "dur": 54,
    "args": {
      "External id": 123077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123077,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123077, "pid": 5, "tid": 7, "ts": 1716454223297925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212291, "dur": 12,
    "args": {
      "External id": 123077,
      "cbid": 211, "correlation": 123077
    }
  },
  {
    "ph": "s", "id": 123077, "pid": 76337, "tid": -914061504, "ts": 1716454223212291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223297981, "dur": 50,
    "args": {
      "External id": 123097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123097,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 123097, "pid": 5, "tid": 7, "ts": 1716454223297981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212358, "dur": 11,
    "args": {
      "External id": 123097,
      "cbid": 211, "correlation": 123097
    }
  },
  {
    "ph": "s", "id": 123097, "pid": 76337, "tid": -914061504, "ts": 1716454223212358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223298033, "dur": 4,
    "args": {
      "External id": 123109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123109,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123109, "pid": 5, "tid": 7, "ts": 1716454223298033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212380, "dur": 6,
    "args": {
      "External id": 123109,
      "cbid": 211, "correlation": 123109
    }
  },
  {
    "ph": "s", "id": 123109, "pid": 76337, "tid": -914061504, "ts": 1716454223212380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223298038, "dur": 56,
    "args": {
      "External id": 123112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123112, "pid": 5, "tid": 7, "ts": 1716454223298038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212398, "dur": 6,
    "args": {
      "External id": 123112,
      "cbid": 211, "correlation": 123112
    }
  },
  {
    "ph": "s", "id": 123112, "pid": 76337, "tid": -914061504, "ts": 1716454223212398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223298095, "dur": 37,
    "args": {
      "External id": 123121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123121,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123121, "pid": 5, "tid": 7, "ts": 1716454223298095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212438, "dur": 10,
    "args": {
      "External id": 123121,
      "cbid": 211, "correlation": 123121
    }
  },
  {
    "ph": "s", "id": 123121, "pid": 76337, "tid": -914061504, "ts": 1716454223212438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212502, "dur": 0,
    "args": {
      "External id": 123131,
      "cbid": 317, "correlation": 123131
    }
  },
  {
    "ph": "f", "id": 123131, "pid": 76337, "tid": -914061504, "ts": 1716454223212502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212503, "dur": 0,
    "args": {
      "External id": 123132,
      "cbid": 203, "correlation": 123132
    }
  },
  {
    "ph": "f", "id": 123132, "pid": 76337, "tid": -914061504, "ts": 1716454223212503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212503, "dur": 0,
    "args": {
      "External id": 123133,
      "cbid": 205, "correlation": 123133
    }
  },
  {
    "ph": "f", "id": 123133, "pid": 76337, "tid": -914061504, "ts": 1716454223212503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223298133, "dur": 41,
    "args": {
      "External id": 123137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123137,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123137, "pid": 5, "tid": 7, "ts": 1716454223298133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212520, "dur": 12,
    "args": {
      "External id": 123137,
      "cbid": 211, "correlation": 123137
    }
  },
  {
    "ph": "s", "id": 123137, "pid": 76337, "tid": -914061504, "ts": 1716454223212520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223298176, "dur": 14,
    "args": {
      "External id": 123139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123139,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123139, "pid": 5, "tid": 7, "ts": 1716454223298176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212534, "dur": 5,
    "args": {
      "External id": 123139,
      "cbid": 211, "correlation": 123139
    }
  },
  {
    "ph": "s", "id": 123139, "pid": 76337, "tid": -914061504, "ts": 1716454223212534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223298191, "dur": 3,
    "args": {
      "External id": 123141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123141, "pid": 5, "tid": 7, "ts": 1716454223298191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212543, "dur": 5,
    "args": {
      "External id": 123141,
      "cbid": 211, "correlation": 123141
    }
  },
  {
    "ph": "s", "id": 123141, "pid": 76337, "tid": -914061504, "ts": 1716454223212543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212551, "dur": 0,
    "args": {
      "External id": 123142,
      "cbid": 51, "correlation": 123142
    }
  },
  {
    "ph": "s", "id": 123142, "pid": 76337, "tid": -914061504, "ts": 1716454223212551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223298196, "dur": 700,
    "args": {
      "External id": 123143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123143,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123143, "pid": 5, "tid": 7, "ts": 1716454223298196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212552, "dur": 5,
    "args": {
      "External id": 123143,
      "cbid": 211, "correlation": 123143
    }
  },
  {
    "ph": "s", "id": 123143, "pid": 76337, "tid": -914061504, "ts": 1716454223212552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223298897, "dur": 60,
    "args": {
      "External id": 123148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123148,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123148, "pid": 5, "tid": 7, "ts": 1716454223298897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212580, "dur": 9,
    "args": {
      "External id": 123148,
      "cbid": 211, "correlation": 123148
    }
  },
  {
    "ph": "s", "id": 123148, "pid": 76337, "tid": -914061504, "ts": 1716454223212580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212638, "dur": 0,
    "args": {
      "External id": 123158,
      "cbid": 317, "correlation": 123158
    }
  },
  {
    "ph": "f", "id": 123158, "pid": 76337, "tid": -914061504, "ts": 1716454223212638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212639, "dur": 0,
    "args": {
      "External id": 123159,
      "cbid": 203, "correlation": 123159
    }
  },
  {
    "ph": "f", "id": 123159, "pid": 76337, "tid": -914061504, "ts": 1716454223212639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212639, "dur": 0,
    "args": {
      "External id": 123160,
      "cbid": 205, "correlation": 123160
    }
  },
  {
    "ph": "f", "id": 123160, "pid": 76337, "tid": -914061504, "ts": 1716454223212639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223298958, "dur": 3,
    "args": {
      "External id": 123164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123164, "pid": 5, "tid": 7, "ts": 1716454223298958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212655, "dur": 13,
    "args": {
      "External id": 123164,
      "cbid": 211, "correlation": 123164
    }
  },
  {
    "ph": "s", "id": 123164, "pid": 76337, "tid": -914061504, "ts": 1716454223212655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212672, "dur": 0,
    "args": {
      "External id": 123165,
      "cbid": 51, "correlation": 123165
    }
  },
  {
    "ph": "s", "id": 123165, "pid": 76337, "tid": -914061504, "ts": 1716454223212672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223298963, "dur": 266,
    "args": {
      "External id": 123166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123166,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123166, "pid": 5, "tid": 7, "ts": 1716454223298963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212673, "dur": 7,
    "args": {
      "External id": 123166,
      "cbid": 211, "correlation": 123166
    }
  },
  {
    "ph": "s", "id": 123166, "pid": 76337, "tid": -914061504, "ts": 1716454223212673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223299230, "dur": 60,
    "args": {
      "External id": 123171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123171,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123171, "pid": 5, "tid": 7, "ts": 1716454223299230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212700, "dur": 8,
    "args": {
      "External id": 123171,
      "cbid": 211, "correlation": 123171
    }
  },
  {
    "ph": "s", "id": 123171, "pid": 76337, "tid": -914061504, "ts": 1716454223212700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223299291, "dur": 50,
    "args": {
      "External id": 123179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123179,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123179, "pid": 5, "tid": 7, "ts": 1716454223299291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212729, "dur": 8,
    "args": {
      "External id": 123179,
      "cbid": 211, "correlation": 123179
    }
  },
  {
    "ph": "s", "id": 123179, "pid": 76337, "tid": -914061504, "ts": 1716454223212729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223299342, "dur": 35,
    "args": {
      "External id": 123187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123187,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123187, "pid": 5, "tid": 7, "ts": 1716454223299342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212756, "dur": 9,
    "args": {
      "External id": 123187,
      "cbid": 211, "correlation": 123187
    }
  },
  {
    "ph": "s", "id": 123187, "pid": 76337, "tid": -914061504, "ts": 1716454223212756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223299378, "dur": 51,
    "args": {
      "External id": 123207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123207,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 123207, "pid": 5, "tid": 7, "ts": 1716454223299378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212838, "dur": 12,
    "args": {
      "External id": 123207,
      "cbid": 211, "correlation": 123207
    }
  },
  {
    "ph": "s", "id": 123207, "pid": 76337, "tid": -914061504, "ts": 1716454223212838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223299431, "dur": 4,
    "args": {
      "External id": 123219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123219,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123219, "pid": 5, "tid": 7, "ts": 1716454223299431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212860, "dur": 6,
    "args": {
      "External id": 123219,
      "cbid": 211, "correlation": 123219
    }
  },
  {
    "ph": "s", "id": 123219, "pid": 76337, "tid": -914061504, "ts": 1716454223212860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223299437, "dur": 54,
    "args": {
      "External id": 123222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123222, "pid": 5, "tid": 7, "ts": 1716454223299437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212877, "dur": 7,
    "args": {
      "External id": 123222,
      "cbid": 211, "correlation": 123222
    }
  },
  {
    "ph": "s", "id": 123222, "pid": 76337, "tid": -914061504, "ts": 1716454223212877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212935, "dur": 0,
    "args": {
      "External id": 123233,
      "cbid": 317, "correlation": 123233
    }
  },
  {
    "ph": "f", "id": 123233, "pid": 76337, "tid": -914061504, "ts": 1716454223212935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212935, "dur": 0,
    "args": {
      "External id": 123234,
      "cbid": 203, "correlation": 123234
    }
  },
  {
    "ph": "f", "id": 123234, "pid": 76337, "tid": -914061504, "ts": 1716454223212935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212936, "dur": 0,
    "args": {
      "External id": 123235,
      "cbid": 205, "correlation": 123235
    }
  },
  {
    "ph": "f", "id": 123235, "pid": 76337, "tid": -914061504, "ts": 1716454223212936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212966, "dur": 2,
    "args": {
      "External id": 123239,
      "cbid": 251, "correlation": 123239
    }
  },
  {
    "ph": "f", "id": 123239, "pid": 76337, "tid": -914061504, "ts": 1716454223212966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212969, "dur": 1,
    "args": {
      "External id": 123240,
      "cbid": 251, "correlation": 123240
    }
  },
  {
    "ph": "f", "id": 123240, "pid": 76337, "tid": -914061504, "ts": 1716454223212969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212971, "dur": 1,
    "args": {
      "External id": 123241,
      "cbid": 251, "correlation": 123241
    }
  },
  {
    "ph": "f", "id": 123241, "pid": 76337, "tid": -914061504, "ts": 1716454223212971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212973, "dur": 8,
    "args": {
      "External id": 123242,
      "cbid": 251, "correlation": 123242
    }
  },
  {
    "ph": "f", "id": 123242, "pid": 76337, "tid": -914061504, "ts": 1716454223212973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212982, "dur": 0,
    "args": {
      "External id": 123243,
      "cbid": 251, "correlation": 123243
    }
  },
  {
    "ph": "f", "id": 123243, "pid": 76337, "tid": -914061504, "ts": 1716454223212982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212983, "dur": 1,
    "args": {
      "External id": 123244,
      "cbid": 251, "correlation": 123244
    }
  },
  {
    "ph": "f", "id": 123244, "pid": 76337, "tid": -914061504, "ts": 1716454223212983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212985, "dur": 1,
    "args": {
      "External id": 123245,
      "cbid": 251, "correlation": 123245
    }
  },
  {
    "ph": "f", "id": 123245, "pid": 76337, "tid": -914061504, "ts": 1716454223212985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212986, "dur": 0,
    "args": {
      "External id": 123246,
      "cbid": 251, "correlation": 123246
    }
  },
  {
    "ph": "f", "id": 123246, "pid": 76337, "tid": -914061504, "ts": 1716454223212986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212988, "dur": 0,
    "args": {
      "External id": 123247,
      "cbid": 251, "correlation": 123247
    }
  },
  {
    "ph": "f", "id": 123247, "pid": 76337, "tid": -914061504, "ts": 1716454223212988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223299492, "dur": 116,
    "args": {
      "External id": 123248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123248,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 123248, "pid": 5, "tid": 7, "ts": 1716454223299492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223212991, "dur": 13,
    "args": {
      "External id": 123248,
      "cbid": 211, "correlation": 123248
    }
  },
  {
    "ph": "s", "id": 123248, "pid": 76337, "tid": -914061504, "ts": 1716454223212991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223299610, "dur": 60,
    "args": {
      "External id": 123254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123254, "pid": 5, "tid": 7, "ts": 1716454223299610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213028, "dur": 9,
    "args": {
      "External id": 123254,
      "cbid": 211, "correlation": 123254
    }
  },
  {
    "ph": "s", "id": 123254, "pid": 76337, "tid": -914061504, "ts": 1716454223213028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223299671, "dur": 532,
    "args": {
      "External id": 123263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123263,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123263, "pid": 5, "tid": 7, "ts": 1716454223299671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213109, "dur": 14,
    "args": {
      "External id": 123263,
      "cbid": 211, "correlation": 123263
    }
  },
  {
    "ph": "s", "id": 123263, "pid": 76337, "tid": -914061504, "ts": 1716454223213109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223300204, "dur": 181,
    "args": {
      "External id": 123285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123285,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123285, "pid": 5, "tid": 7, "ts": 1716454223300204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213168, "dur": 10,
    "args": {
      "External id": 123285,
      "cbid": 211, "correlation": 123285
    }
  },
  {
    "ph": "s", "id": 123285, "pid": 76337, "tid": -914061504, "ts": 1716454223213168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213255, "dur": 2,
    "args": {
      "External id": 123296,
      "cbid": 251, "correlation": 123296
    }
  },
  {
    "ph": "f", "id": 123296, "pid": 76337, "tid": -914061504, "ts": 1716454223213255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223300386, "dur": 200,
    "args": {
      "External id": 123297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123297,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123297, "pid": 5, "tid": 7, "ts": 1716454223300386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213260, "dur": 14,
    "args": {
      "External id": 123297,
      "cbid": 211, "correlation": 123297
    }
  },
  {
    "ph": "s", "id": 123297, "pid": 76337, "tid": -914061504, "ts": 1716454223213260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213330, "dur": 1,
    "args": {
      "External id": 123308,
      "cbid": 251, "correlation": 123308
    }
  },
  {
    "ph": "f", "id": 123308, "pid": 76337, "tid": -914061504, "ts": 1716454223213330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223300587, "dur": 185,
    "args": {
      "External id": 123309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123309,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123309, "pid": 5, "tid": 7, "ts": 1716454223300587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213334, "dur": 11,
    "args": {
      "External id": 123309,
      "cbid": 211, "correlation": 123309
    }
  },
  {
    "ph": "s", "id": 123309, "pid": 76337, "tid": -914061504, "ts": 1716454223213334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213397, "dur": 1,
    "args": {
      "External id": 123320,
      "cbid": 251, "correlation": 123320
    }
  },
  {
    "ph": "f", "id": 123320, "pid": 76337, "tid": -914061504, "ts": 1716454223213397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223300774, "dur": 183,
    "args": {
      "External id": 123321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123321,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123321, "pid": 5, "tid": 7, "ts": 1716454223300774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213401, "dur": 12,
    "args": {
      "External id": 123321,
      "cbid": 211, "correlation": 123321
    }
  },
  {
    "ph": "s", "id": 123321, "pid": 76337, "tid": -914061504, "ts": 1716454223213401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223300959, "dur": 18570,
    "args": {
      "External id": 123342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123342,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 123342, "pid": 5, "tid": 7, "ts": 1716454223300959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213484, "dur": 13,
    "args": {
      "External id": 123342,
      "cbid": 211, "correlation": 123342
    }
  },
  {
    "ph": "s", "id": 123342, "pid": 76337, "tid": -914061504, "ts": 1716454223213484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213581, "dur": 1,
    "args": {
      "External id": 123360,
      "cbid": 251, "correlation": 123360
    }
  },
  {
    "ph": "f", "id": 123360, "pid": 76337, "tid": -914061504, "ts": 1716454223213581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223319530, "dur": 200,
    "args": {
      "External id": 123362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123362,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123362, "pid": 5, "tid": 7, "ts": 1716454223319530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213586, "dur": 13,
    "args": {
      "External id": 123362,
      "cbid": 211, "correlation": 123362
    }
  },
  {
    "ph": "s", "id": 123362, "pid": 76337, "tid": -914061504, "ts": 1716454223213586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223319731, "dur": 66,
    "args": {
      "External id": 123370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123370,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123370, "pid": 5, "tid": 7, "ts": 1716454223319731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213656, "dur": 13,
    "args": {
      "External id": 123370,
      "cbid": 211, "correlation": 123370
    }
  },
  {
    "ph": "s", "id": 123370, "pid": 76337, "tid": -914061504, "ts": 1716454223213656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223319799, "dur": 97,
    "args": {
      "External id": 123378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123378,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123378, "pid": 5, "tid": 7, "ts": 1716454223319799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213696, "dur": 8,
    "args": {
      "External id": 123378,
      "cbid": 211, "correlation": 123378
    }
  },
  {
    "ph": "s", "id": 123378, "pid": 76337, "tid": -914061504, "ts": 1716454223213696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223319897, "dur": 54,
    "args": {
      "External id": 123389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123389, "pid": 5, "tid": 7, "ts": 1716454223319897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213766, "dur": 13,
    "args": {
      "External id": 123389,
      "cbid": 211, "correlation": 123389
    }
  },
  {
    "ph": "s", "id": 123389, "pid": 76337, "tid": -914061504, "ts": 1716454223213766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223319952, "dur": 92,
    "args": {
      "External id": 123411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123411,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123411, "pid": 5, "tid": 7, "ts": 1716454223319952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213797, "dur": 7,
    "args": {
      "External id": 123411,
      "cbid": 211, "correlation": 123411
    }
  },
  {
    "ph": "s", "id": 123411, "pid": 76337, "tid": -914061504, "ts": 1716454223213797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213880, "dur": 1,
    "args": {
      "External id": 123422,
      "cbid": 251, "correlation": 123422
    }
  },
  {
    "ph": "f", "id": 123422, "pid": 76337, "tid": -914061504, "ts": 1716454223213880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223320045, "dur": 102,
    "args": {
      "External id": 123423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123423,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123423, "pid": 5, "tid": 7, "ts": 1716454223320045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213885, "dur": 13,
    "args": {
      "External id": 123423,
      "cbid": 211, "correlation": 123423
    }
  },
  {
    "ph": "s", "id": 123423, "pid": 76337, "tid": -914061504, "ts": 1716454223213885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213959, "dur": 1,
    "args": {
      "External id": 123434,
      "cbid": 251, "correlation": 123434
    }
  },
  {
    "ph": "f", "id": 123434, "pid": 76337, "tid": -914061504, "ts": 1716454223213959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213963, "dur": 0,
    "args": {
      "External id": 123435,
      "cbid": 251, "correlation": 123435
    }
  },
  {
    "ph": "f", "id": 123435, "pid": 76337, "tid": -914061504, "ts": 1716454223213963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223320148, "dur": 10,
    "args": {
      "External id": 123436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123436,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 123436, "pid": 5, "tid": 7, "ts": 1716454223320148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213965, "dur": 21,
    "args": {
      "External id": 123436,
      "cbid": 211, "correlation": 123436
    }
  },
  {
    "ph": "s", "id": 123436, "pid": 76337, "tid": -914061504, "ts": 1716454223213965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223320160, "dur": 5,
    "args": {
      "External id": 123438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123438,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 123438, "pid": 5, "tid": 7, "ts": 1716454223320160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223213990, "dur": 7,
    "args": {
      "External id": 123438,
      "cbid": 211, "correlation": 123438
    }
  },
  {
    "ph": "s", "id": 123438, "pid": 76337, "tid": -914061504, "ts": 1716454223213990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214053, "dur": 1,
    "args": {
      "External id": 123449,
      "cbid": 251, "correlation": 123449
    }
  },
  {
    "ph": "f", "id": 123449, "pid": 76337, "tid": -914061504, "ts": 1716454223214053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214056, "dur": 0,
    "args": {
      "External id": 123450,
      "cbid": 251, "correlation": 123450
    }
  },
  {
    "ph": "f", "id": 123450, "pid": 76337, "tid": -914061504, "ts": 1716454223214056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223320166, "dur": 6,
    "args": {
      "External id": 123451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123451,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 123451, "pid": 5, "tid": 7, "ts": 1716454223320166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214058, "dur": 12,
    "args": {
      "External id": 123451,
      "cbid": 211, "correlation": 123451
    }
  },
  {
    "ph": "s", "id": 123451, "pid": 76337, "tid": -914061504, "ts": 1716454223214058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223320174, "dur": 3,
    "args": {
      "External id": 123453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123453,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 123453, "pid": 5, "tid": 7, "ts": 1716454223320174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214072, "dur": 6,
    "args": {
      "External id": 123453,
      "cbid": 211, "correlation": 123453
    }
  },
  {
    "ph": "s", "id": 123453, "pid": 76337, "tid": -914061504, "ts": 1716454223214072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223320179, "dur": 156,
    "args": {
      "External id": 123474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123474,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 123474, "pid": 5, "tid": 7, "ts": 1716454223320179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214145, "dur": 12,
    "args": {
      "External id": 123474,
      "cbid": 211, "correlation": 123474
    }
  },
  {
    "ph": "s", "id": 123474, "pid": 76337, "tid": -914061504, "ts": 1716454223214145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214240, "dur": 1,
    "args": {
      "External id": 123492,
      "cbid": 251, "correlation": 123492
    }
  },
  {
    "ph": "f", "id": 123492, "pid": 76337, "tid": -914061504, "ts": 1716454223214240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223320336, "dur": 107,
    "args": {
      "External id": 123494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123494,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 123494, "pid": 5, "tid": 7, "ts": 1716454223320336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214247, "dur": 13,
    "args": {
      "External id": 123494,
      "cbid": 211, "correlation": 123494
    }
  },
  {
    "ph": "s", "id": 123494, "pid": 76337, "tid": -914061504, "ts": 1716454223214247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223320444, "dur": 35,
    "args": {
      "External id": 123502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123502,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123502, "pid": 5, "tid": 7, "ts": 1716454223320444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214317, "dur": 12,
    "args": {
      "External id": 123502,
      "cbid": 211, "correlation": 123502
    }
  },
  {
    "ph": "s", "id": 123502, "pid": 76337, "tid": -914061504, "ts": 1716454223214317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223320480, "dur": 68,
    "args": {
      "External id": 123510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123510, "pid": 5, "tid": 7, "ts": 1716454223320480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214357, "dur": 9,
    "args": {
      "External id": 123510,
      "cbid": 211, "correlation": 123510
    }
  },
  {
    "ph": "s", "id": 123510, "pid": 76337, "tid": -914061504, "ts": 1716454223214357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223320550, "dur": 91,
    "args": {
      "External id": 123532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123532,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123532, "pid": 5, "tid": 7, "ts": 1716454223320550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214408, "dur": 10,
    "args": {
      "External id": 123532,
      "cbid": 211, "correlation": 123532
    }
  },
  {
    "ph": "s", "id": 123532, "pid": 76337, "tid": -914061504, "ts": 1716454223214408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214496, "dur": 1,
    "args": {
      "External id": 123548,
      "cbid": 251, "correlation": 123548
    }
  },
  {
    "ph": "f", "id": 123548, "pid": 76337, "tid": -914061504, "ts": 1716454223214496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223320643, "dur": 571,
    "args": {
      "External id": 123550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123550,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123550, "pid": 5, "tid": 7, "ts": 1716454223320643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214502, "dur": 12,
    "args": {
      "External id": 123550,
      "cbid": 211, "correlation": 123550
    }
  },
  {
    "ph": "s", "id": 123550, "pid": 76337, "tid": -914061504, "ts": 1716454223214502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223321215, "dur": 244,
    "args": {
      "External id": 123558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123558, "pid": 5, "tid": 7, "ts": 1716454223321215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214567, "dur": 13,
    "args": {
      "External id": 123558,
      "cbid": 211, "correlation": 123558
    }
  },
  {
    "ph": "s", "id": 123558, "pid": 76337, "tid": -914061504, "ts": 1716454223214567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223321460, "dur": 251,
    "args": {
      "External id": 123566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123566, "pid": 5, "tid": 7, "ts": 1716454223321460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214600, "dur": 8,
    "args": {
      "External id": 123566,
      "cbid": 211, "correlation": 123566
    }
  },
  {
    "ph": "s", "id": 123566, "pid": 76337, "tid": -914061504, "ts": 1716454223214600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214682, "dur": 1,
    "args": {
      "External id": 123582,
      "cbid": 251, "correlation": 123582
    }
  },
  {
    "ph": "f", "id": 123582, "pid": 76337, "tid": -914061504, "ts": 1716454223214682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214687, "dur": 0,
    "args": {
      "External id": 123584,
      "cbid": 251, "correlation": 123584
    }
  },
  {
    "ph": "f", "id": 123584, "pid": 76337, "tid": -914061504, "ts": 1716454223214687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223321713, "dur": 352,
    "args": {
      "External id": 123585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123585,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 123585, "pid": 5, "tid": 7, "ts": 1716454223321713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214690, "dur": 13,
    "args": {
      "External id": 123585,
      "cbid": 211, "correlation": 123585
    }
  },
  {
    "ph": "s", "id": 123585, "pid": 76337, "tid": -914061504, "ts": 1716454223214690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223322066, "dur": 50,
    "args": {
      "External id": 123593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123593,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123593, "pid": 5, "tid": 7, "ts": 1716454223322066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214732, "dur": 10,
    "args": {
      "External id": 123593,
      "cbid": 211, "correlation": 123593
    }
  },
  {
    "ph": "s", "id": 123593, "pid": 76337, "tid": -914061504, "ts": 1716454223214732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223322117, "dur": 157,
    "args": {
      "External id": 123604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123604, "pid": 5, "tid": 7, "ts": 1716454223322117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214800, "dur": 12,
    "args": {
      "External id": 123604,
      "cbid": 211, "correlation": 123604
    }
  },
  {
    "ph": "s", "id": 123604, "pid": 76337, "tid": -914061504, "ts": 1716454223214800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214864, "dur": 0,
    "args": {
      "External id": 123616,
      "cbid": 317, "correlation": 123616
    }
  },
  {
    "ph": "f", "id": 123616, "pid": 76337, "tid": -914061504, "ts": 1716454223214864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214865, "dur": 0,
    "args": {
      "External id": 123617,
      "cbid": 203, "correlation": 123617
    }
  },
  {
    "ph": "f", "id": 123617, "pid": 76337, "tid": -914061504, "ts": 1716454223214865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214866, "dur": 0,
    "args": {
      "External id": 123618,
      "cbid": 205, "correlation": 123618
    }
  },
  {
    "ph": "f", "id": 123618, "pid": 76337, "tid": -914061504, "ts": 1716454223214866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214890, "dur": 1,
    "args": {
      "External id": 123622,
      "cbid": 251, "correlation": 123622
    }
  },
  {
    "ph": "f", "id": 123622, "pid": 76337, "tid": -914061504, "ts": 1716454223214890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214892, "dur": 0,
    "args": {
      "External id": 123623,
      "cbid": 251, "correlation": 123623
    }
  },
  {
    "ph": "f", "id": 123623, "pid": 76337, "tid": -914061504, "ts": 1716454223214892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214893, "dur": 0,
    "args": {
      "External id": 123624,
      "cbid": 251, "correlation": 123624
    }
  },
  {
    "ph": "f", "id": 123624, "pid": 76337, "tid": -914061504, "ts": 1716454223214893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214894, "dur": 0,
    "args": {
      "External id": 123625,
      "cbid": 251, "correlation": 123625
    }
  },
  {
    "ph": "f", "id": 123625, "pid": 76337, "tid": -914061504, "ts": 1716454223214894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214895, "dur": 0,
    "args": {
      "External id": 123626,
      "cbid": 251, "correlation": 123626
    }
  },
  {
    "ph": "f", "id": 123626, "pid": 76337, "tid": -914061504, "ts": 1716454223214895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214895, "dur": 0,
    "args": {
      "External id": 123627,
      "cbid": 251, "correlation": 123627
    }
  },
  {
    "ph": "f", "id": 123627, "pid": 76337, "tid": -914061504, "ts": 1716454223214895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214896, "dur": 0,
    "args": {
      "External id": 123628,
      "cbid": 251, "correlation": 123628
    }
  },
  {
    "ph": "f", "id": 123628, "pid": 76337, "tid": -914061504, "ts": 1716454223214896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214897, "dur": 0,
    "args": {
      "External id": 123629,
      "cbid": 251, "correlation": 123629
    }
  },
  {
    "ph": "f", "id": 123629, "pid": 76337, "tid": -914061504, "ts": 1716454223214897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214898, "dur": 0,
    "args": {
      "External id": 123630,
      "cbid": 251, "correlation": 123630
    }
  },
  {
    "ph": "f", "id": 123630, "pid": 76337, "tid": -914061504, "ts": 1716454223214898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223322276, "dur": 115,
    "args": {
      "External id": 123631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123631,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 123631, "pid": 5, "tid": 7, "ts": 1716454223322276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214900, "dur": 12,
    "args": {
      "External id": 123631,
      "cbid": 211, "correlation": 123631
    }
  },
  {
    "ph": "s", "id": 123631, "pid": 76337, "tid": -914061504, "ts": 1716454223214900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223322392, "dur": 60,
    "args": {
      "External id": 123637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123637, "pid": 5, "tid": 7, "ts": 1716454223322392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214936, "dur": 9,
    "args": {
      "External id": 123637,
      "cbid": 211, "correlation": 123637
    }
  },
  {
    "ph": "s", "id": 123637, "pid": 76337, "tid": -914061504, "ts": 1716454223214936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223322453, "dur": 50,
    "args": {
      "External id": 123645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123645,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123645, "pid": 5, "tid": 7, "ts": 1716454223322453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223214968, "dur": 18,
    "args": {
      "External id": 123645,
      "cbid": 211, "correlation": 123645
    }
  },
  {
    "ph": "s", "id": 123645, "pid": 76337, "tid": -914061504, "ts": 1716454223214968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223322505, "dur": 101,
    "args": {
      "External id": 123654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123654,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123654, "pid": 5, "tid": 7, "ts": 1716454223322505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215017, "dur": 11,
    "args": {
      "External id": 123654,
      "cbid": 211, "correlation": 123654
    }
  },
  {
    "ph": "s", "id": 123654, "pid": 76337, "tid": -914061504, "ts": 1716454223215017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223322606, "dur": 92,
    "args": {
      "External id": 123674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123674,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 123674, "pid": 5, "tid": 7, "ts": 1716454223322606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215090, "dur": 12,
    "args": {
      "External id": 123674,
      "cbid": 211, "correlation": 123674
    }
  },
  {
    "ph": "s", "id": 123674, "pid": 76337, "tid": -914061504, "ts": 1716454223215090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223322700, "dur": 4,
    "args": {
      "External id": 123686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123686,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 123686, "pid": 5, "tid": 7, "ts": 1716454223322700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215111, "dur": 7,
    "args": {
      "External id": 123686,
      "cbid": 211, "correlation": 123686
    }
  },
  {
    "ph": "s", "id": 123686, "pid": 76337, "tid": -914061504, "ts": 1716454223215111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223322706, "dur": 109,
    "args": {
      "External id": 123689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123689, "pid": 5, "tid": 7, "ts": 1716454223322706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215130, "dur": 6,
    "args": {
      "External id": 123689,
      "cbid": 211, "correlation": 123689
    }
  },
  {
    "ph": "s", "id": 123689, "pid": 76337, "tid": -914061504, "ts": 1716454223215130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223322816, "dur": 69,
    "args": {
      "External id": 123698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123698,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123698, "pid": 5, "tid": 7, "ts": 1716454223322816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215169, "dur": 11,
    "args": {
      "External id": 123698,
      "cbid": 211, "correlation": 123698
    }
  },
  {
    "ph": "s", "id": 123698, "pid": 76337, "tid": -914061504, "ts": 1716454223215169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215221, "dur": 0,
    "args": {
      "External id": 123708,
      "cbid": 317, "correlation": 123708
    }
  },
  {
    "ph": "f", "id": 123708, "pid": 76337, "tid": -914061504, "ts": 1716454223215221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215222, "dur": 0,
    "args": {
      "External id": 123709,
      "cbid": 203, "correlation": 123709
    }
  },
  {
    "ph": "f", "id": 123709, "pid": 76337, "tid": -914061504, "ts": 1716454223215222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215223, "dur": 0,
    "args": {
      "External id": 123710,
      "cbid": 205, "correlation": 123710
    }
  },
  {
    "ph": "f", "id": 123710, "pid": 76337, "tid": -914061504, "ts": 1716454223215223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223322886, "dur": 76,
    "args": {
      "External id": 123714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123714,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123714, "pid": 5, "tid": 7, "ts": 1716454223322886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215238, "dur": 11,
    "args": {
      "External id": 123714,
      "cbid": 211, "correlation": 123714
    }
  },
  {
    "ph": "s", "id": 123714, "pid": 76337, "tid": -914061504, "ts": 1716454223215238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223322964, "dur": 24,
    "args": {
      "External id": 123716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123716,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123716, "pid": 5, "tid": 7, "ts": 1716454223322964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215252, "dur": 5,
    "args": {
      "External id": 123716,
      "cbid": 211, "correlation": 123716
    }
  },
  {
    "ph": "s", "id": 123716, "pid": 76337, "tid": -914061504, "ts": 1716454223215252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223322989, "dur": 4,
    "args": {
      "External id": 123718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123718, "pid": 5, "tid": 7, "ts": 1716454223322989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215262, "dur": 6,
    "args": {
      "External id": 123718,
      "cbid": 211, "correlation": 123718
    }
  },
  {
    "ph": "s", "id": 123718, "pid": 76337, "tid": -914061504, "ts": 1716454223215262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215271, "dur": 0,
    "args": {
      "External id": 123719,
      "cbid": 51, "correlation": 123719
    }
  },
  {
    "ph": "s", "id": 123719, "pid": 76337, "tid": -914061504, "ts": 1716454223215271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223322995, "dur": 1364,
    "args": {
      "External id": 123720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123720,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123720, "pid": 5, "tid": 7, "ts": 1716454223322995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215272, "dur": 6,
    "args": {
      "External id": 123720,
      "cbid": 211, "correlation": 123720
    }
  },
  {
    "ph": "s", "id": 123720, "pid": 76337, "tid": -914061504, "ts": 1716454223215272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223324360, "dur": 60,
    "args": {
      "External id": 123725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123725,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123725, "pid": 5, "tid": 7, "ts": 1716454223324360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215300, "dur": 8,
    "args": {
      "External id": 123725,
      "cbid": 211, "correlation": 123725
    }
  },
  {
    "ph": "s", "id": 123725, "pid": 76337, "tid": -914061504, "ts": 1716454223215300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223324421, "dur": 5,
    "args": {
      "External id": 123733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123733,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123733, "pid": 5, "tid": 7, "ts": 1716454223324421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215343, "dur": 9,
    "args": {
      "External id": 123733,
      "cbid": 211, "correlation": 123733
    }
  },
  {
    "ph": "s", "id": 123733, "pid": 76337, "tid": -914061504, "ts": 1716454223215343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215410, "dur": 1,
    "args": {
      "External id": 123749,
      "cbid": 251, "correlation": 123749
    }
  },
  {
    "ph": "f", "id": 123749, "pid": 76337, "tid": -914061504, "ts": 1716454223215410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215415, "dur": 0,
    "args": {
      "External id": 123751,
      "cbid": 251, "correlation": 123751
    }
  },
  {
    "ph": "f", "id": 123751, "pid": 76337, "tid": -914061504, "ts": 1716454223215415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223324427, "dur": 11,
    "args": {
      "External id": 123752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123752,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 123752, "pid": 5, "tid": 7, "ts": 1716454223324427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215417, "dur": 12,
    "args": {
      "External id": 123752,
      "cbid": 211, "correlation": 123752
    }
  },
  {
    "ph": "s", "id": 123752, "pid": 76337, "tid": -914061504, "ts": 1716454223215417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223324440, "dur": 5,
    "args": {
      "External id": 123754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123754,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 123754, "pid": 5, "tid": 7, "ts": 1716454223324440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215431, "dur": 6,
    "args": {
      "External id": 123754,
      "cbid": 211, "correlation": 123754
    }
  },
  {
    "ph": "s", "id": 123754, "pid": 76337, "tid": -914061504, "ts": 1716454223215431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223324446, "dur": 53,
    "args": {
      "External id": 123764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123764,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123764, "pid": 5, "tid": 7, "ts": 1716454223324446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215489, "dur": 12,
    "args": {
      "External id": 123764,
      "cbid": 211, "correlation": 123764
    }
  },
  {
    "ph": "s", "id": 123764, "pid": 76337, "tid": -914061504, "ts": 1716454223215489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223324501, "dur": 53,
    "args": {
      "External id": 123784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123784,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 123784, "pid": 5, "tid": 7, "ts": 1716454223324501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215555, "dur": 11,
    "args": {
      "External id": 123784,
      "cbid": 211, "correlation": 123784
    }
  },
  {
    "ph": "s", "id": 123784, "pid": 76337, "tid": -914061504, "ts": 1716454223215555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223324555, "dur": 4,
    "args": {
      "External id": 123796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123796,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123796, "pid": 5, "tid": 7, "ts": 1716454223324555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215576, "dur": 7,
    "args": {
      "External id": 123796,
      "cbid": 211, "correlation": 123796
    }
  },
  {
    "ph": "s", "id": 123796, "pid": 76337, "tid": -914061504, "ts": 1716454223215576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223324560, "dur": 55,
    "args": {
      "External id": 123799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123799, "pid": 5, "tid": 7, "ts": 1716454223324560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215595, "dur": 6,
    "args": {
      "External id": 123799,
      "cbid": 211, "correlation": 123799
    }
  },
  {
    "ph": "s", "id": 123799, "pid": 76337, "tid": -914061504, "ts": 1716454223215595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223324617, "dur": 36,
    "args": {
      "External id": 123808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123808,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123808, "pid": 5, "tid": 7, "ts": 1716454223324617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215637, "dur": 10,
    "args": {
      "External id": 123808,
      "cbid": 211, "correlation": 123808
    }
  },
  {
    "ph": "s", "id": 123808, "pid": 76337, "tid": -914061504, "ts": 1716454223215637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215700, "dur": 0,
    "args": {
      "External id": 123818,
      "cbid": 317, "correlation": 123818
    }
  },
  {
    "ph": "f", "id": 123818, "pid": 76337, "tid": -914061504, "ts": 1716454223215700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215701, "dur": 0,
    "args": {
      "External id": 123819,
      "cbid": 203, "correlation": 123819
    }
  },
  {
    "ph": "f", "id": 123819, "pid": 76337, "tid": -914061504, "ts": 1716454223215701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215702, "dur": 0,
    "args": {
      "External id": 123820,
      "cbid": 205, "correlation": 123820
    }
  },
  {
    "ph": "f", "id": 123820, "pid": 76337, "tid": -914061504, "ts": 1716454223215702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223324654, "dur": 41,
    "args": {
      "External id": 123824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123824,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123824, "pid": 5, "tid": 7, "ts": 1716454223324654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215716, "dur": 12,
    "args": {
      "External id": 123824,
      "cbid": 211, "correlation": 123824
    }
  },
  {
    "ph": "s", "id": 123824, "pid": 76337, "tid": -914061504, "ts": 1716454223215716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223324696, "dur": 14,
    "args": {
      "External id": 123826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123826,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123826, "pid": 5, "tid": 7, "ts": 1716454223324696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215731, "dur": 6,
    "args": {
      "External id": 123826,
      "cbid": 211, "correlation": 123826
    }
  },
  {
    "ph": "s", "id": 123826, "pid": 76337, "tid": -914061504, "ts": 1716454223215731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223324712, "dur": 3,
    "args": {
      "External id": 123828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123828,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123828, "pid": 5, "tid": 7, "ts": 1716454223324712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215741, "dur": 5,
    "args": {
      "External id": 123828,
      "cbid": 211, "correlation": 123828
    }
  },
  {
    "ph": "s", "id": 123828, "pid": 76337, "tid": -914061504, "ts": 1716454223215741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215749, "dur": 0,
    "args": {
      "External id": 123829,
      "cbid": 51, "correlation": 123829
    }
  },
  {
    "ph": "s", "id": 123829, "pid": 76337, "tid": -914061504, "ts": 1716454223215749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223324717, "dur": 698,
    "args": {
      "External id": 123830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123830,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123830, "pid": 5, "tid": 7, "ts": 1716454223324717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215750, "dur": 5,
    "args": {
      "External id": 123830,
      "cbid": 211, "correlation": 123830
    }
  },
  {
    "ph": "s", "id": 123830, "pid": 76337, "tid": -914061504, "ts": 1716454223215750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223325416, "dur": 60,
    "args": {
      "External id": 123835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123835,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123835, "pid": 5, "tid": 7, "ts": 1716454223325416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215776, "dur": 10,
    "args": {
      "External id": 123835,
      "cbid": 211, "correlation": 123835
    }
  },
  {
    "ph": "s", "id": 123835, "pid": 76337, "tid": -914061504, "ts": 1716454223215776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215834, "dur": 0,
    "args": {
      "External id": 123845,
      "cbid": 317, "correlation": 123845
    }
  },
  {
    "ph": "f", "id": 123845, "pid": 76337, "tid": -914061504, "ts": 1716454223215834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215835, "dur": 0,
    "args": {
      "External id": 123846,
      "cbid": 203, "correlation": 123846
    }
  },
  {
    "ph": "f", "id": 123846, "pid": 76337, "tid": -914061504, "ts": 1716454223215835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215836, "dur": 0,
    "args": {
      "External id": 123847,
      "cbid": 205, "correlation": 123847
    }
  },
  {
    "ph": "f", "id": 123847, "pid": 76337, "tid": -914061504, "ts": 1716454223215836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223325477, "dur": 75,
    "args": {
      "External id": 123851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123851,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123851, "pid": 5, "tid": 7, "ts": 1716454223325477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215849, "dur": 12,
    "args": {
      "External id": 123851,
      "cbid": 211, "correlation": 123851
    }
  },
  {
    "ph": "s", "id": 123851, "pid": 76337, "tid": -914061504, "ts": 1716454223215849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223325553, "dur": 208,
    "args": {
      "External id": 123853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123853,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123853, "pid": 5, "tid": 7, "ts": 1716454223325553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215868, "dur": 8,
    "args": {
      "External id": 123853,
      "cbid": 211, "correlation": 123853
    }
  },
  {
    "ph": "s", "id": 123853, "pid": 76337, "tid": -914061504, "ts": 1716454223215868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223325763, "dur": 40,
    "args": {
      "External id": 123855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123855,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123855, "pid": 5, "tid": 7, "ts": 1716454223325763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223215881, "dur": 406,
    "args": {
      "External id": 123855,
      "cbid": 211, "correlation": 123855
    }
  },
  {
    "ph": "s", "id": 123855, "pid": 76337, "tid": -914061504, "ts": 1716454223215881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223325804, "dur": 58,
    "args": {
      "External id": 123861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123861, "pid": 5, "tid": 7, "ts": 1716454223325804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216308, "dur": 9,
    "args": {
      "External id": 123861,
      "cbid": 211, "correlation": 123861
    }
  },
  {
    "ph": "s", "id": 123861, "pid": 76337, "tid": -914061504, "ts": 1716454223216308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223325864, "dur": 50,
    "args": {
      "External id": 123869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123869,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123869, "pid": 5, "tid": 7, "ts": 1716454223325864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216339, "dur": 8,
    "args": {
      "External id": 123869,
      "cbid": 211, "correlation": 123869
    }
  },
  {
    "ph": "s", "id": 123869, "pid": 76337, "tid": -914061504, "ts": 1716454223216339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223325915, "dur": 35,
    "args": {
      "External id": 123877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123877,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123877, "pid": 5, "tid": 7, "ts": 1716454223325915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216368, "dur": 28,
    "args": {
      "External id": 123877,
      "cbid": 211, "correlation": 123877
    }
  },
  {
    "ph": "s", "id": 123877, "pid": 76337, "tid": -914061504, "ts": 1716454223216368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223325951, "dur": 50,
    "args": {
      "External id": 123897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123897,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 123897, "pid": 5, "tid": 7, "ts": 1716454223325951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216473, "dur": 14,
    "args": {
      "External id": 123897,
      "cbid": 211, "correlation": 123897
    }
  },
  {
    "ph": "s", "id": 123897, "pid": 76337, "tid": -914061504, "ts": 1716454223216473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223326003, "dur": 4,
    "args": {
      "External id": 123909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123909,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 123909, "pid": 5, "tid": 7, "ts": 1716454223326003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216496, "dur": 6,
    "args": {
      "External id": 123909,
      "cbid": 211, "correlation": 123909
    }
  },
  {
    "ph": "s", "id": 123909, "pid": 76337, "tid": -914061504, "ts": 1716454223216496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223326009, "dur": 55,
    "args": {
      "External id": 123912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123912, "pid": 5, "tid": 7, "ts": 1716454223326009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216514, "dur": 6,
    "args": {
      "External id": 123912,
      "cbid": 211, "correlation": 123912
    }
  },
  {
    "ph": "s", "id": 123912, "pid": 76337, "tid": -914061504, "ts": 1716454223216514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216571, "dur": 0,
    "args": {
      "External id": 123923,
      "cbid": 317, "correlation": 123923
    }
  },
  {
    "ph": "f", "id": 123923, "pid": 76337, "tid": -914061504, "ts": 1716454223216571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216572, "dur": 0,
    "args": {
      "External id": 123924,
      "cbid": 203, "correlation": 123924
    }
  },
  {
    "ph": "f", "id": 123924, "pid": 76337, "tid": -914061504, "ts": 1716454223216572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216573, "dur": 0,
    "args": {
      "External id": 123925,
      "cbid": 205, "correlation": 123925
    }
  },
  {
    "ph": "f", "id": 123925, "pid": 76337, "tid": -914061504, "ts": 1716454223216573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216595, "dur": 1,
    "args": {
      "External id": 123929,
      "cbid": 251, "correlation": 123929
    }
  },
  {
    "ph": "f", "id": 123929, "pid": 76337, "tid": -914061504, "ts": 1716454223216595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216597, "dur": 0,
    "args": {
      "External id": 123930,
      "cbid": 251, "correlation": 123930
    }
  },
  {
    "ph": "f", "id": 123930, "pid": 76337, "tid": -914061504, "ts": 1716454223216597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216598, "dur": 0,
    "args": {
      "External id": 123931,
      "cbid": 251, "correlation": 123931
    }
  },
  {
    "ph": "f", "id": 123931, "pid": 76337, "tid": -914061504, "ts": 1716454223216598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216599, "dur": 0,
    "args": {
      "External id": 123932,
      "cbid": 251, "correlation": 123932
    }
  },
  {
    "ph": "f", "id": 123932, "pid": 76337, "tid": -914061504, "ts": 1716454223216599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216599, "dur": 0,
    "args": {
      "External id": 123933,
      "cbid": 251, "correlation": 123933
    }
  },
  {
    "ph": "f", "id": 123933, "pid": 76337, "tid": -914061504, "ts": 1716454223216599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216600, "dur": 0,
    "args": {
      "External id": 123934,
      "cbid": 251, "correlation": 123934
    }
  },
  {
    "ph": "f", "id": 123934, "pid": 76337, "tid": -914061504, "ts": 1716454223216600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216601, "dur": 0,
    "args": {
      "External id": 123935,
      "cbid": 251, "correlation": 123935
    }
  },
  {
    "ph": "f", "id": 123935, "pid": 76337, "tid": -914061504, "ts": 1716454223216601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216602, "dur": 0,
    "args": {
      "External id": 123936,
      "cbid": 251, "correlation": 123936
    }
  },
  {
    "ph": "f", "id": 123936, "pid": 76337, "tid": -914061504, "ts": 1716454223216602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216603, "dur": 0,
    "args": {
      "External id": 123937,
      "cbid": 251, "correlation": 123937
    }
  },
  {
    "ph": "f", "id": 123937, "pid": 76337, "tid": -914061504, "ts": 1716454223216603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223326065, "dur": 113,
    "args": {
      "External id": 123938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123938,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 123938, "pid": 5, "tid": 7, "ts": 1716454223326065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216605, "dur": 13,
    "args": {
      "External id": 123938,
      "cbid": 211, "correlation": 123938
    }
  },
  {
    "ph": "s", "id": 123938, "pid": 76337, "tid": -914061504, "ts": 1716454223216605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223326179, "dur": 60,
    "args": {
      "External id": 123944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123944,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123944, "pid": 5, "tid": 7, "ts": 1716454223326179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216642, "dur": 9,
    "args": {
      "External id": 123944,
      "cbid": 211, "correlation": 123944
    }
  },
  {
    "ph": "s", "id": 123944, "pid": 76337, "tid": -914061504, "ts": 1716454223216642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223326240, "dur": 476,
    "args": {
      "External id": 123953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123953,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123953, "pid": 5, "tid": 7, "ts": 1716454223326240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216722, "dur": 15,
    "args": {
      "External id": 123953,
      "cbid": 211, "correlation": 123953
    }
  },
  {
    "ph": "s", "id": 123953, "pid": 76337, "tid": -914061504, "ts": 1716454223216722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223326718, "dur": 181,
    "args": {
      "External id": 123975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123975,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 123975, "pid": 5, "tid": 7, "ts": 1716454223326718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216780, "dur": 11,
    "args": {
      "External id": 123975,
      "cbid": 211, "correlation": 123975
    }
  },
  {
    "ph": "s", "id": 123975, "pid": 76337, "tid": -914061504, "ts": 1716454223216780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216867, "dur": 1,
    "args": {
      "External id": 123986,
      "cbid": 251, "correlation": 123986
    }
  },
  {
    "ph": "f", "id": 123986, "pid": 76337, "tid": -914061504, "ts": 1716454223216867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223326901, "dur": 199,
    "args": {
      "External id": 123987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123987,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123987, "pid": 5, "tid": 7, "ts": 1716454223326901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216872, "dur": 13,
    "args": {
      "External id": 123987,
      "cbid": 211, "correlation": 123987
    }
  },
  {
    "ph": "s", "id": 123987, "pid": 76337, "tid": -914061504, "ts": 1716454223216872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216942, "dur": 1,
    "args": {
      "External id": 123998,
      "cbid": 251, "correlation": 123998
    }
  },
  {
    "ph": "f", "id": 123998, "pid": 76337, "tid": -914061504, "ts": 1716454223216942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223327101, "dur": 194,
    "args": {
      "External id": 123999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 123999,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 123999, "pid": 5, "tid": 7, "ts": 1716454223327101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223216946, "dur": 11,
    "args": {
      "External id": 123999,
      "cbid": 211, "correlation": 123999
    }
  },
  {
    "ph": "s", "id": 123999, "pid": 76337, "tid": -914061504, "ts": 1716454223216946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217019, "dur": 1,
    "args": {
      "External id": 124010,
      "cbid": 251, "correlation": 124010
    }
  },
  {
    "ph": "f", "id": 124010, "pid": 76337, "tid": -914061504, "ts": 1716454223217019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223327296, "dur": 189,
    "args": {
      "External id": 124011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124011,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124011, "pid": 5, "tid": 7, "ts": 1716454223327296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217023, "dur": 11,
    "args": {
      "External id": 124011,
      "cbid": 211, "correlation": 124011
    }
  },
  {
    "ph": "s", "id": 124011, "pid": 76337, "tid": -914061504, "ts": 1716454223217023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223327487, "dur": 18589,
    "args": {
      "External id": 124032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124032,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124032, "pid": 5, "tid": 7, "ts": 1716454223327487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217104, "dur": 13,
    "args": {
      "External id": 124032,
      "cbid": 211, "correlation": 124032
    }
  },
  {
    "ph": "s", "id": 124032, "pid": 76337, "tid": -914061504, "ts": 1716454223217104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217200, "dur": 1,
    "args": {
      "External id": 124050,
      "cbid": 251, "correlation": 124050
    }
  },
  {
    "ph": "f", "id": 124050, "pid": 76337, "tid": -914061504, "ts": 1716454223217200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223346078, "dur": 202,
    "args": {
      "External id": 124052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124052,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124052, "pid": 5, "tid": 7, "ts": 1716454223346078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217206, "dur": 13,
    "args": {
      "External id": 124052,
      "cbid": 211, "correlation": 124052
    }
  },
  {
    "ph": "s", "id": 124052, "pid": 76337, "tid": -914061504, "ts": 1716454223217206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223346281, "dur": 66,
    "args": {
      "External id": 124060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124060,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124060, "pid": 5, "tid": 7, "ts": 1716454223346281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217277, "dur": 12,
    "args": {
      "External id": 124060,
      "cbid": 211, "correlation": 124060
    }
  },
  {
    "ph": "s", "id": 124060, "pid": 76337, "tid": -914061504, "ts": 1716454223217277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223346348, "dur": 96,
    "args": {
      "External id": 124068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124068,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124068, "pid": 5, "tid": 7, "ts": 1716454223346348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217316, "dur": 8,
    "args": {
      "External id": 124068,
      "cbid": 211, "correlation": 124068
    }
  },
  {
    "ph": "s", "id": 124068, "pid": 76337, "tid": -914061504, "ts": 1716454223217316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223346446, "dur": 54,
    "args": {
      "External id": 124079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124079,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124079, "pid": 5, "tid": 7, "ts": 1716454223346446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217387, "dur": 74,
    "args": {
      "External id": 124079,
      "cbid": 211, "correlation": 124079
    }
  },
  {
    "ph": "s", "id": 124079, "pid": 76337, "tid": -914061504, "ts": 1716454223217387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223346501, "dur": 92,
    "args": {
      "External id": 124101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124101,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124101, "pid": 5, "tid": 7, "ts": 1716454223346501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223217480, "dur": 1929,
    "args": {
      "External id": 124101,
      "cbid": 211, "correlation": 124101
    }
  },
  {
    "ph": "s", "id": 124101, "pid": 76337, "tid": -914061504, "ts": 1716454223217480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219486, "dur": 1,
    "args": {
      "External id": 124112,
      "cbid": 251, "correlation": 124112
    }
  },
  {
    "ph": "f", "id": 124112, "pid": 76337, "tid": -914061504, "ts": 1716454223219486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223346595, "dur": 102,
    "args": {
      "External id": 124113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124113,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124113, "pid": 5, "tid": 7, "ts": 1716454223346595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219492, "dur": 67,
    "args": {
      "External id": 124113,
      "cbid": 211, "correlation": 124113
    }
  },
  {
    "ph": "s", "id": 124113, "pid": 76337, "tid": -914061504, "ts": 1716454223219492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219617, "dur": 1,
    "args": {
      "External id": 124124,
      "cbid": 251, "correlation": 124124
    }
  },
  {
    "ph": "f", "id": 124124, "pid": 76337, "tid": -914061504, "ts": 1716454223219617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219621, "dur": 0,
    "args": {
      "External id": 124125,
      "cbid": 251, "correlation": 124125
    }
  },
  {
    "ph": "f", "id": 124125, "pid": 76337, "tid": -914061504, "ts": 1716454223219621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223346698, "dur": 10,
    "args": {
      "External id": 124126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124126,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 124126, "pid": 5, "tid": 7, "ts": 1716454223346698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219623, "dur": 13,
    "args": {
      "External id": 124126,
      "cbid": 211, "correlation": 124126
    }
  },
  {
    "ph": "s", "id": 124126, "pid": 76337, "tid": -914061504, "ts": 1716454223219623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223346710, "dur": 5,
    "args": {
      "External id": 124128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124128,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 124128, "pid": 5, "tid": 7, "ts": 1716454223346710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219637, "dur": 6,
    "args": {
      "External id": 124128,
      "cbid": 211, "correlation": 124128
    }
  },
  {
    "ph": "s", "id": 124128, "pid": 76337, "tid": -914061504, "ts": 1716454223219637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219698, "dur": 1,
    "args": {
      "External id": 124139,
      "cbid": 251, "correlation": 124139
    }
  },
  {
    "ph": "f", "id": 124139, "pid": 76337, "tid": -914061504, "ts": 1716454223219698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219701, "dur": 0,
    "args": {
      "External id": 124140,
      "cbid": 251, "correlation": 124140
    }
  },
  {
    "ph": "f", "id": 124140, "pid": 76337, "tid": -914061504, "ts": 1716454223219701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223346716, "dur": 6,
    "args": {
      "External id": 124141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124141,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 124141, "pid": 5, "tid": 7, "ts": 1716454223346716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219703, "dur": 12,
    "args": {
      "External id": 124141,
      "cbid": 211, "correlation": 124141
    }
  },
  {
    "ph": "s", "id": 124141, "pid": 76337, "tid": -914061504, "ts": 1716454223219703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223346724, "dur": 3,
    "args": {
      "External id": 124143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124143,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 124143, "pid": 5, "tid": 7, "ts": 1716454223346724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219717, "dur": 5,
    "args": {
      "External id": 124143,
      "cbid": 211, "correlation": 124143
    }
  },
  {
    "ph": "s", "id": 124143, "pid": 76337, "tid": -914061504, "ts": 1716454223219717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223346729, "dur": 155,
    "args": {
      "External id": 124164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124164,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124164, "pid": 5, "tid": 7, "ts": 1716454223346729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219789, "dur": 16,
    "args": {
      "External id": 124164,
      "cbid": 211, "correlation": 124164
    }
  },
  {
    "ph": "s", "id": 124164, "pid": 76337, "tid": -914061504, "ts": 1716454223219789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219891, "dur": 1,
    "args": {
      "External id": 124182,
      "cbid": 251, "correlation": 124182
    }
  },
  {
    "ph": "f", "id": 124182, "pid": 76337, "tid": -914061504, "ts": 1716454223219891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223346885, "dur": 108,
    "args": {
      "External id": 124184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124184,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124184, "pid": 5, "tid": 7, "ts": 1716454223346885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219897, "dur": 13,
    "args": {
      "External id": 124184,
      "cbid": 211, "correlation": 124184
    }
  },
  {
    "ph": "s", "id": 124184, "pid": 76337, "tid": -914061504, "ts": 1716454223219897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223346994, "dur": 35,
    "args": {
      "External id": 124192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124192,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124192, "pid": 5, "tid": 7, "ts": 1716454223346994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223219967, "dur": 19,
    "args": {
      "External id": 124192,
      "cbid": 211, "correlation": 124192
    }
  },
  {
    "ph": "s", "id": 124192, "pid": 76337, "tid": -914061504, "ts": 1716454223219967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223347030, "dur": 67,
    "args": {
      "External id": 124200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124200, "pid": 5, "tid": 7, "ts": 1716454223347030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220017, "dur": 10,
    "args": {
      "External id": 124200,
      "cbid": 211, "correlation": 124200
    }
  },
  {
    "ph": "s", "id": 124200, "pid": 76337, "tid": -914061504, "ts": 1716454223220017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223347098, "dur": 91,
    "args": {
      "External id": 124222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124222,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124222, "pid": 5, "tid": 7, "ts": 1716454223347098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220069, "dur": 10,
    "args": {
      "External id": 124222,
      "cbid": 211, "correlation": 124222
    }
  },
  {
    "ph": "s", "id": 124222, "pid": 76337, "tid": -914061504, "ts": 1716454223220069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220154, "dur": 1,
    "args": {
      "External id": 124238,
      "cbid": 251, "correlation": 124238
    }
  },
  {
    "ph": "f", "id": 124238, "pid": 76337, "tid": -914061504, "ts": 1716454223220154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223347191, "dur": 574,
    "args": {
      "External id": 124240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124240,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124240, "pid": 5, "tid": 7, "ts": 1716454223347191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220159, "dur": 14,
    "args": {
      "External id": 124240,
      "cbid": 211, "correlation": 124240
    }
  },
  {
    "ph": "s", "id": 124240, "pid": 76337, "tid": -914061504, "ts": 1716454223220159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223347766, "dur": 244,
    "args": {
      "External id": 124248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124248, "pid": 5, "tid": 7, "ts": 1716454223347766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220226, "dur": 12,
    "args": {
      "External id": 124248,
      "cbid": 211, "correlation": 124248
    }
  },
  {
    "ph": "s", "id": 124248, "pid": 76337, "tid": -914061504, "ts": 1716454223220226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223348011, "dur": 253,
    "args": {
      "External id": 124256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124256, "pid": 5, "tid": 7, "ts": 1716454223348011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220256, "dur": 9,
    "args": {
      "External id": 124256,
      "cbid": 211, "correlation": 124256
    }
  },
  {
    "ph": "s", "id": 124256, "pid": 76337, "tid": -914061504, "ts": 1716454223220256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220337, "dur": 1,
    "args": {
      "External id": 124272,
      "cbid": 251, "correlation": 124272
    }
  },
  {
    "ph": "f", "id": 124272, "pid": 76337, "tid": -914061504, "ts": 1716454223220337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220342, "dur": 0,
    "args": {
      "External id": 124274,
      "cbid": 251, "correlation": 124274
    }
  },
  {
    "ph": "f", "id": 124274, "pid": 76337, "tid": -914061504, "ts": 1716454223220342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223348265, "dur": 357,
    "args": {
      "External id": 124275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124275,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 124275, "pid": 5, "tid": 7, "ts": 1716454223348265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220345, "dur": 12,
    "args": {
      "External id": 124275,
      "cbid": 211, "correlation": 124275
    }
  },
  {
    "ph": "s", "id": 124275, "pid": 76337, "tid": -914061504, "ts": 1716454223220345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223348624, "dur": 50,
    "args": {
      "External id": 124283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124283,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124283, "pid": 5, "tid": 7, "ts": 1716454223348624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220387, "dur": 10,
    "args": {
      "External id": 124283,
      "cbid": 211, "correlation": 124283
    }
  },
  {
    "ph": "s", "id": 124283, "pid": 76337, "tid": -914061504, "ts": 1716454223220387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223348676, "dur": 157,
    "args": {
      "External id": 124294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124294, "pid": 5, "tid": 7, "ts": 1716454223348676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220452, "dur": 217,
    "args": {
      "External id": 124294,
      "cbid": 211, "correlation": 124294
    }
  },
  {
    "ph": "s", "id": 124294, "pid": 76337, "tid": -914061504, "ts": 1716454223220452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220722, "dur": 0,
    "args": {
      "External id": 124306,
      "cbid": 317, "correlation": 124306
    }
  },
  {
    "ph": "f", "id": 124306, "pid": 76337, "tid": -914061504, "ts": 1716454223220722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220723, "dur": 0,
    "args": {
      "External id": 124307,
      "cbid": 203, "correlation": 124307
    }
  },
  {
    "ph": "f", "id": 124307, "pid": 76337, "tid": -914061504, "ts": 1716454223220723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220724, "dur": 0,
    "args": {
      "External id": 124308,
      "cbid": 205, "correlation": 124308
    }
  },
  {
    "ph": "f", "id": 124308, "pid": 76337, "tid": -914061504, "ts": 1716454223220724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220747, "dur": 1,
    "args": {
      "External id": 124312,
      "cbid": 251, "correlation": 124312
    }
  },
  {
    "ph": "f", "id": 124312, "pid": 76337, "tid": -914061504, "ts": 1716454223220747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220749, "dur": 0,
    "args": {
      "External id": 124313,
      "cbid": 251, "correlation": 124313
    }
  },
  {
    "ph": "f", "id": 124313, "pid": 76337, "tid": -914061504, "ts": 1716454223220749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220749, "dur": 0,
    "args": {
      "External id": 124314,
      "cbid": 251, "correlation": 124314
    }
  },
  {
    "ph": "f", "id": 124314, "pid": 76337, "tid": -914061504, "ts": 1716454223220749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220750, "dur": 0,
    "args": {
      "External id": 124315,
      "cbid": 251, "correlation": 124315
    }
  },
  {
    "ph": "f", "id": 124315, "pid": 76337, "tid": -914061504, "ts": 1716454223220750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220751, "dur": 0,
    "args": {
      "External id": 124316,
      "cbid": 251, "correlation": 124316
    }
  },
  {
    "ph": "f", "id": 124316, "pid": 76337, "tid": -914061504, "ts": 1716454223220751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220751, "dur": 0,
    "args": {
      "External id": 124317,
      "cbid": 251, "correlation": 124317
    }
  },
  {
    "ph": "f", "id": 124317, "pid": 76337, "tid": -914061504, "ts": 1716454223220751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220752, "dur": 0,
    "args": {
      "External id": 124318,
      "cbid": 251, "correlation": 124318
    }
  },
  {
    "ph": "f", "id": 124318, "pid": 76337, "tid": -914061504, "ts": 1716454223220752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220753, "dur": 0,
    "args": {
      "External id": 124319,
      "cbid": 251, "correlation": 124319
    }
  },
  {
    "ph": "f", "id": 124319, "pid": 76337, "tid": -914061504, "ts": 1716454223220753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220754, "dur": 0,
    "args": {
      "External id": 124320,
      "cbid": 251, "correlation": 124320
    }
  },
  {
    "ph": "f", "id": 124320, "pid": 76337, "tid": -914061504, "ts": 1716454223220754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223348834, "dur": 115,
    "args": {
      "External id": 124321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124321,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124321, "pid": 5, "tid": 7, "ts": 1716454223348834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220756, "dur": 40,
    "args": {
      "External id": 124321,
      "cbid": 211, "correlation": 124321
    }
  },
  {
    "ph": "s", "id": 124321, "pid": 76337, "tid": -914061504, "ts": 1716454223220756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223348950, "dur": 59,
    "args": {
      "External id": 124327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124327,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124327, "pid": 5, "tid": 7, "ts": 1716454223348950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220820, "dur": 103,
    "args": {
      "External id": 124327,
      "cbid": 211, "correlation": 124327
    }
  },
  {
    "ph": "s", "id": 124327, "pid": 76337, "tid": -914061504, "ts": 1716454223220820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223349010, "dur": 49,
    "args": {
      "External id": 124335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124335,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124335, "pid": 5, "tid": 7, "ts": 1716454223349010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223220946, "dur": 284,
    "args": {
      "External id": 124335,
      "cbid": 211, "correlation": 124335
    }
  },
  {
    "ph": "s", "id": 124335, "pid": 76337, "tid": -914061504, "ts": 1716454223220946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223349061, "dur": 98,
    "args": {
      "External id": 124344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124344,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124344, "pid": 5, "tid": 7, "ts": 1716454223349061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221261, "dur": 11,
    "args": {
      "External id": 124344,
      "cbid": 211, "correlation": 124344
    }
  },
  {
    "ph": "s", "id": 124344, "pid": 76337, "tid": -914061504, "ts": 1716454223221261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223349161, "dur": 92,
    "args": {
      "External id": 124364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124364,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 124364, "pid": 5, "tid": 7, "ts": 1716454223349161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221334, "dur": 12,
    "args": {
      "External id": 124364,
      "cbid": 211, "correlation": 124364
    }
  },
  {
    "ph": "s", "id": 124364, "pid": 76337, "tid": -914061504, "ts": 1716454223221334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223349254, "dur": 4,
    "args": {
      "External id": 124376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124376,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 124376, "pid": 5, "tid": 7, "ts": 1716454223349254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221356, "dur": 6,
    "args": {
      "External id": 124376,
      "cbid": 211, "correlation": 124376
    }
  },
  {
    "ph": "s", "id": 124376, "pid": 76337, "tid": -914061504, "ts": 1716454223221356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223349260, "dur": 107,
    "args": {
      "External id": 124379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124379, "pid": 5, "tid": 7, "ts": 1716454223349260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221374, "dur": 111,
    "args": {
      "External id": 124379,
      "cbid": 211, "correlation": 124379
    }
  },
  {
    "ph": "s", "id": 124379, "pid": 76337, "tid": -914061504, "ts": 1716454223221374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223349368, "dur": 69,
    "args": {
      "External id": 124388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124388,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124388, "pid": 5, "tid": 7, "ts": 1716454223349368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221519, "dur": 10,
    "args": {
      "External id": 124388,
      "cbid": 211, "correlation": 124388
    }
  },
  {
    "ph": "s", "id": 124388, "pid": 76337, "tid": -914061504, "ts": 1716454223221519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221571, "dur": 0,
    "args": {
      "External id": 124398,
      "cbid": 317, "correlation": 124398
    }
  },
  {
    "ph": "f", "id": 124398, "pid": 76337, "tid": -914061504, "ts": 1716454223221571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221572, "dur": 0,
    "args": {
      "External id": 124399,
      "cbid": 203, "correlation": 124399
    }
  },
  {
    "ph": "f", "id": 124399, "pid": 76337, "tid": -914061504, "ts": 1716454223221572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221573, "dur": 0,
    "args": {
      "External id": 124400,
      "cbid": 205, "correlation": 124400
    }
  },
  {
    "ph": "f", "id": 124400, "pid": 76337, "tid": -914061504, "ts": 1716454223221573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223349438, "dur": 75,
    "args": {
      "External id": 124404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124404, "pid": 5, "tid": 7, "ts": 1716454223349438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221587, "dur": 12,
    "args": {
      "External id": 124404,
      "cbid": 211, "correlation": 124404
    }
  },
  {
    "ph": "s", "id": 124404, "pid": 76337, "tid": -914061504, "ts": 1716454223221587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223349515, "dur": 24,
    "args": {
      "External id": 124406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124406,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124406, "pid": 5, "tid": 7, "ts": 1716454223349515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221601, "dur": 5,
    "args": {
      "External id": 124406,
      "cbid": 211, "correlation": 124406
    }
  },
  {
    "ph": "s", "id": 124406, "pid": 76337, "tid": -914061504, "ts": 1716454223221601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223349540, "dur": 4,
    "args": {
      "External id": 124408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 124408, "pid": 5, "tid": 7, "ts": 1716454223349540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221611, "dur": 6,
    "args": {
      "External id": 124408,
      "cbid": 211, "correlation": 124408
    }
  },
  {
    "ph": "s", "id": 124408, "pid": 76337, "tid": -914061504, "ts": 1716454223221611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221620, "dur": 0,
    "args": {
      "External id": 124409,
      "cbid": 51, "correlation": 124409
    }
  },
  {
    "ph": "s", "id": 124409, "pid": 76337, "tid": -914061504, "ts": 1716454223221620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223349545, "dur": 1359,
    "args": {
      "External id": 124410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124410,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124410, "pid": 5, "tid": 7, "ts": 1716454223349545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221621, "dur": 5,
    "args": {
      "External id": 124410,
      "cbid": 211, "correlation": 124410
    }
  },
  {
    "ph": "s", "id": 124410, "pid": 76337, "tid": -914061504, "ts": 1716454223221621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223350905, "dur": 59,
    "args": {
      "External id": 124415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124415, "pid": 5, "tid": 7, "ts": 1716454223350905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221648, "dur": 9,
    "args": {
      "External id": 124415,
      "cbid": 211, "correlation": 124415
    }
  },
  {
    "ph": "s", "id": 124415, "pid": 76337, "tid": -914061504, "ts": 1716454223221648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223350965, "dur": 4,
    "args": {
      "External id": 124423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124423,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 124423, "pid": 5, "tid": 7, "ts": 1716454223350965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221693, "dur": 10,
    "args": {
      "External id": 124423,
      "cbid": 211, "correlation": 124423
    }
  },
  {
    "ph": "s", "id": 124423, "pid": 76337, "tid": -914061504, "ts": 1716454223221693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221760, "dur": 2,
    "args": {
      "External id": 124439,
      "cbid": 251, "correlation": 124439
    }
  },
  {
    "ph": "f", "id": 124439, "pid": 76337, "tid": -914061504, "ts": 1716454223221760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221766, "dur": 0,
    "args": {
      "External id": 124441,
      "cbid": 251, "correlation": 124441
    }
  },
  {
    "ph": "f", "id": 124441, "pid": 76337, "tid": -914061504, "ts": 1716454223221766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223350971, "dur": 11,
    "args": {
      "External id": 124442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124442,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 124442, "pid": 5, "tid": 7, "ts": 1716454223350971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221768, "dur": 12,
    "args": {
      "External id": 124442,
      "cbid": 211, "correlation": 124442
    }
  },
  {
    "ph": "s", "id": 124442, "pid": 76337, "tid": -914061504, "ts": 1716454223221768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223350983, "dur": 5,
    "args": {
      "External id": 124444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124444,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 124444, "pid": 5, "tid": 7, "ts": 1716454223350983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221782, "dur": 6,
    "args": {
      "External id": 124444,
      "cbid": 211, "correlation": 124444
    }
  },
  {
    "ph": "s", "id": 124444, "pid": 76337, "tid": -914061504, "ts": 1716454223221782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223350989, "dur": 54,
    "args": {
      "External id": 124454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124454, "pid": 5, "tid": 7, "ts": 1716454223350989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223221841, "dur": 553,
    "args": {
      "External id": 124454,
      "cbid": 211, "correlation": 124454
    }
  },
  {
    "ph": "s", "id": 124454, "pid": 76337, "tid": -914061504, "ts": 1716454223221841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223351045, "dur": 51,
    "args": {
      "External id": 124474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124474,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 124474, "pid": 5, "tid": 7, "ts": 1716454223351045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222450, "dur": 12,
    "args": {
      "External id": 124474,
      "cbid": 211, "correlation": 124474
    }
  },
  {
    "ph": "s", "id": 124474, "pid": 76337, "tid": -914061504, "ts": 1716454223222450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223351098, "dur": 4,
    "args": {
      "External id": 124486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124486,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 124486, "pid": 5, "tid": 7, "ts": 1716454223351098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222471, "dur": 7,
    "args": {
      "External id": 124486,
      "cbid": 211, "correlation": 124486
    }
  },
  {
    "ph": "s", "id": 124486, "pid": 76337, "tid": -914061504, "ts": 1716454223222471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223351103, "dur": 55,
    "args": {
      "External id": 124489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124489,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124489, "pid": 5, "tid": 7, "ts": 1716454223351103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222492, "dur": 7,
    "args": {
      "External id": 124489,
      "cbid": 211, "correlation": 124489
    }
  },
  {
    "ph": "s", "id": 124489, "pid": 76337, "tid": -914061504, "ts": 1716454223222492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223351159, "dur": 36,
    "args": {
      "External id": 124498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124498,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124498, "pid": 5, "tid": 7, "ts": 1716454223351159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222533, "dur": 10,
    "args": {
      "External id": 124498,
      "cbid": 211, "correlation": 124498
    }
  },
  {
    "ph": "s", "id": 124498, "pid": 76337, "tid": -914061504, "ts": 1716454223222533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222595, "dur": 0,
    "args": {
      "External id": 124508,
      "cbid": 317, "correlation": 124508
    }
  },
  {
    "ph": "f", "id": 124508, "pid": 76337, "tid": -914061504, "ts": 1716454223222595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222596, "dur": 0,
    "args": {
      "External id": 124509,
      "cbid": 203, "correlation": 124509
    }
  },
  {
    "ph": "f", "id": 124509, "pid": 76337, "tid": -914061504, "ts": 1716454223222596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222597, "dur": 0,
    "args": {
      "External id": 124510,
      "cbid": 205, "correlation": 124510
    }
  },
  {
    "ph": "f", "id": 124510, "pid": 76337, "tid": -914061504, "ts": 1716454223222597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223351197, "dur": 39,
    "args": {
      "External id": 124514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124514,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124514, "pid": 5, "tid": 7, "ts": 1716454223351197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222611, "dur": 12,
    "args": {
      "External id": 124514,
      "cbid": 211, "correlation": 124514
    }
  },
  {
    "ph": "s", "id": 124514, "pid": 76337, "tid": -914061504, "ts": 1716454223222611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223351238, "dur": 14,
    "args": {
      "External id": 124516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124516,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124516, "pid": 5, "tid": 7, "ts": 1716454223351238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222625, "dur": 6,
    "args": {
      "External id": 124516,
      "cbid": 211, "correlation": 124516
    }
  },
  {
    "ph": "s", "id": 124516, "pid": 76337, "tid": -914061504, "ts": 1716454223222625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223351253, "dur": 3,
    "args": {
      "External id": 124518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124518,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 124518, "pid": 5, "tid": 7, "ts": 1716454223351253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222635, "dur": 5,
    "args": {
      "External id": 124518,
      "cbid": 211, "correlation": 124518
    }
  },
  {
    "ph": "s", "id": 124518, "pid": 76337, "tid": -914061504, "ts": 1716454223222635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222643, "dur": 0,
    "args": {
      "External id": 124519,
      "cbid": 51, "correlation": 124519
    }
  },
  {
    "ph": "s", "id": 124519, "pid": 76337, "tid": -914061504, "ts": 1716454223222643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223351258, "dur": 694,
    "args": {
      "External id": 124520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124520,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124520, "pid": 5, "tid": 7, "ts": 1716454223351258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222644, "dur": 5,
    "args": {
      "External id": 124520,
      "cbid": 211, "correlation": 124520
    }
  },
  {
    "ph": "s", "id": 124520, "pid": 76337, "tid": -914061504, "ts": 1716454223222644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223351954, "dur": 59,
    "args": {
      "External id": 124525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124525,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124525, "pid": 5, "tid": 7, "ts": 1716454223351954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222671, "dur": 9,
    "args": {
      "External id": 124525,
      "cbid": 211, "correlation": 124525
    }
  },
  {
    "ph": "s", "id": 124525, "pid": 76337, "tid": -914061504, "ts": 1716454223222671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222728, "dur": 0,
    "args": {
      "External id": 124535,
      "cbid": 317, "correlation": 124535
    }
  },
  {
    "ph": "f", "id": 124535, "pid": 76337, "tid": -914061504, "ts": 1716454223222728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222729, "dur": 0,
    "args": {
      "External id": 124536,
      "cbid": 203, "correlation": 124536
    }
  },
  {
    "ph": "f", "id": 124536, "pid": 76337, "tid": -914061504, "ts": 1716454223222729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222730, "dur": 0,
    "args": {
      "External id": 124537,
      "cbid": 205, "correlation": 124537
    }
  },
  {
    "ph": "f", "id": 124537, "pid": 76337, "tid": -914061504, "ts": 1716454223222730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223352014, "dur": 74,
    "args": {
      "External id": 124541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124541,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124541, "pid": 5, "tid": 7, "ts": 1716454223352014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222742, "dur": 12,
    "args": {
      "External id": 124541,
      "cbid": 211, "correlation": 124541
    }
  },
  {
    "ph": "s", "id": 124541, "pid": 76337, "tid": -914061504, "ts": 1716454223222742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223352090, "dur": 205,
    "args": {
      "External id": 124543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124543,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124543, "pid": 5, "tid": 7, "ts": 1716454223352090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222760, "dur": 6,
    "args": {
      "External id": 124543,
      "cbid": 211, "correlation": 124543
    }
  },
  {
    "ph": "s", "id": 124543, "pid": 76337, "tid": -914061504, "ts": 1716454223222760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223352296, "dur": 38,
    "args": {
      "External id": 124545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124545,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124545, "pid": 5, "tid": 7, "ts": 1716454223352296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222771, "dur": 5,
    "args": {
      "External id": 124545,
      "cbid": 211, "correlation": 124545
    }
  },
  {
    "ph": "s", "id": 124545, "pid": 76337, "tid": -914061504, "ts": 1716454223222771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223352335, "dur": 60,
    "args": {
      "External id": 124551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124551,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124551, "pid": 5, "tid": 7, "ts": 1716454223352335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223222798, "dur": 513,
    "args": {
      "External id": 124551,
      "cbid": 211, "correlation": 124551
    }
  },
  {
    "ph": "s", "id": 124551, "pid": 76337, "tid": -914061504, "ts": 1716454223222798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223352396, "dur": 50,
    "args": {
      "External id": 124559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124559,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124559, "pid": 5, "tid": 7, "ts": 1716454223352396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223332, "dur": 9,
    "args": {
      "External id": 124559,
      "cbid": 211, "correlation": 124559
    }
  },
  {
    "ph": "s", "id": 124559, "pid": 76337, "tid": -914061504, "ts": 1716454223223332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223352448, "dur": 35,
    "args": {
      "External id": 124567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124567,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124567, "pid": 5, "tid": 7, "ts": 1716454223352448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223363, "dur": 8,
    "args": {
      "External id": 124567,
      "cbid": 211, "correlation": 124567
    }
  },
  {
    "ph": "s", "id": 124567, "pid": 76337, "tid": -914061504, "ts": 1716454223223363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223352484, "dur": 53,
    "args": {
      "External id": 124587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124587,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 124587, "pid": 5, "tid": 7, "ts": 1716454223352484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223445, "dur": 13,
    "args": {
      "External id": 124587,
      "cbid": 211, "correlation": 124587
    }
  },
  {
    "ph": "s", "id": 124587, "pid": 76337, "tid": -914061504, "ts": 1716454223223445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223352539, "dur": 4,
    "args": {
      "External id": 124599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124599,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 124599, "pid": 5, "tid": 7, "ts": 1716454223352539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223468, "dur": 6,
    "args": {
      "External id": 124599,
      "cbid": 211, "correlation": 124599
    }
  },
  {
    "ph": "s", "id": 124599, "pid": 76337, "tid": -914061504, "ts": 1716454223223468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223352544, "dur": 54,
    "args": {
      "External id": 124602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124602, "pid": 5, "tid": 7, "ts": 1716454223352544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223486, "dur": 7,
    "args": {
      "External id": 124602,
      "cbid": 211, "correlation": 124602
    }
  },
  {
    "ph": "s", "id": 124602, "pid": 76337, "tid": -914061504, "ts": 1716454223223486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223543, "dur": 0,
    "args": {
      "External id": 124613,
      "cbid": 317, "correlation": 124613
    }
  },
  {
    "ph": "f", "id": 124613, "pid": 76337, "tid": -914061504, "ts": 1716454223223543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223544, "dur": 0,
    "args": {
      "External id": 124614,
      "cbid": 203, "correlation": 124614
    }
  },
  {
    "ph": "f", "id": 124614, "pid": 76337, "tid": -914061504, "ts": 1716454223223544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223545, "dur": 0,
    "args": {
      "External id": 124615,
      "cbid": 205, "correlation": 124615
    }
  },
  {
    "ph": "f", "id": 124615, "pid": 76337, "tid": -914061504, "ts": 1716454223223545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223567, "dur": 1,
    "args": {
      "External id": 124619,
      "cbid": 251, "correlation": 124619
    }
  },
  {
    "ph": "f", "id": 124619, "pid": 76337, "tid": -914061504, "ts": 1716454223223567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223570, "dur": 0,
    "args": {
      "External id": 124620,
      "cbid": 251, "correlation": 124620
    }
  },
  {
    "ph": "f", "id": 124620, "pid": 76337, "tid": -914061504, "ts": 1716454223223570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223570, "dur": 0,
    "args": {
      "External id": 124621,
      "cbid": 251, "correlation": 124621
    }
  },
  {
    "ph": "f", "id": 124621, "pid": 76337, "tid": -914061504, "ts": 1716454223223570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223571, "dur": 0,
    "args": {
      "External id": 124622,
      "cbid": 251, "correlation": 124622
    }
  },
  {
    "ph": "f", "id": 124622, "pid": 76337, "tid": -914061504, "ts": 1716454223223571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223572, "dur": 0,
    "args": {
      "External id": 124623,
      "cbid": 251, "correlation": 124623
    }
  },
  {
    "ph": "f", "id": 124623, "pid": 76337, "tid": -914061504, "ts": 1716454223223572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223572, "dur": 0,
    "args": {
      "External id": 124624,
      "cbid": 251, "correlation": 124624
    }
  },
  {
    "ph": "f", "id": 124624, "pid": 76337, "tid": -914061504, "ts": 1716454223223572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223573, "dur": 0,
    "args": {
      "External id": 124625,
      "cbid": 251, "correlation": 124625
    }
  },
  {
    "ph": "f", "id": 124625, "pid": 76337, "tid": -914061504, "ts": 1716454223223573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223574, "dur": 0,
    "args": {
      "External id": 124626,
      "cbid": 251, "correlation": 124626
    }
  },
  {
    "ph": "f", "id": 124626, "pid": 76337, "tid": -914061504, "ts": 1716454223223574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223575, "dur": 0,
    "args": {
      "External id": 124627,
      "cbid": 251, "correlation": 124627
    }
  },
  {
    "ph": "f", "id": 124627, "pid": 76337, "tid": -914061504, "ts": 1716454223223575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223352600, "dur": 113,
    "args": {
      "External id": 124628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124628,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124628, "pid": 5, "tid": 7, "ts": 1716454223352600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223577, "dur": 12,
    "args": {
      "External id": 124628,
      "cbid": 211, "correlation": 124628
    }
  },
  {
    "ph": "s", "id": 124628, "pid": 76337, "tid": -914061504, "ts": 1716454223223577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223352714, "dur": 59,
    "args": {
      "External id": 124634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124634, "pid": 5, "tid": 7, "ts": 1716454223352714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223612, "dur": 9,
    "args": {
      "External id": 124634,
      "cbid": 211, "correlation": 124634
    }
  },
  {
    "ph": "s", "id": 124634, "pid": 76337, "tid": -914061504, "ts": 1716454223223612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223352775, "dur": 591,
    "args": {
      "External id": 124643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124643,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124643, "pid": 5, "tid": 7, "ts": 1716454223352775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223695, "dur": 14,
    "args": {
      "External id": 124643,
      "cbid": 211, "correlation": 124643
    }
  },
  {
    "ph": "s", "id": 124643, "pid": 76337, "tid": -914061504, "ts": 1716454223223695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223353367, "dur": 180,
    "args": {
      "External id": 124665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124665,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124665, "pid": 5, "tid": 7, "ts": 1716454223353367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223753, "dur": 11,
    "args": {
      "External id": 124665,
      "cbid": 211, "correlation": 124665
    }
  },
  {
    "ph": "s", "id": 124665, "pid": 76337, "tid": -914061504, "ts": 1716454223223753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223841, "dur": 1,
    "args": {
      "External id": 124676,
      "cbid": 251, "correlation": 124676
    }
  },
  {
    "ph": "f", "id": 124676, "pid": 76337, "tid": -914061504, "ts": 1716454223223841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223353549, "dur": 196,
    "args": {
      "External id": 124677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124677,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124677, "pid": 5, "tid": 7, "ts": 1716454223353549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223846, "dur": 14,
    "args": {
      "External id": 124677,
      "cbid": 211, "correlation": 124677
    }
  },
  {
    "ph": "s", "id": 124677, "pid": 76337, "tid": -914061504, "ts": 1716454223223846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223915, "dur": 1,
    "args": {
      "External id": 124688,
      "cbid": 251, "correlation": 124688
    }
  },
  {
    "ph": "f", "id": 124688, "pid": 76337, "tid": -914061504, "ts": 1716454223223915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223353746, "dur": 189,
    "args": {
      "External id": 124689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124689,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124689, "pid": 5, "tid": 7, "ts": 1716454223353746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223919, "dur": 11,
    "args": {
      "External id": 124689,
      "cbid": 211, "correlation": 124689
    }
  },
  {
    "ph": "s", "id": 124689, "pid": 76337, "tid": -914061504, "ts": 1716454223223919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223989, "dur": 1,
    "args": {
      "External id": 124700,
      "cbid": 251, "correlation": 124700
    }
  },
  {
    "ph": "f", "id": 124700, "pid": 76337, "tid": -914061504, "ts": 1716454223223989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223353936, "dur": 187,
    "args": {
      "External id": 124701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124701,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124701, "pid": 5, "tid": 7, "ts": 1716454223353936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223223994, "dur": 13,
    "args": {
      "External id": 124701,
      "cbid": 211, "correlation": 124701
    }
  },
  {
    "ph": "s", "id": 124701, "pid": 76337, "tid": -914061504, "ts": 1716454223223994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223354125, "dur": 18569,
    "args": {
      "External id": 124722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124722,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124722, "pid": 5, "tid": 7, "ts": 1716454223354125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223224076, "dur": 12,
    "args": {
      "External id": 124722,
      "cbid": 211, "correlation": 124722
    }
  },
  {
    "ph": "s", "id": 124722, "pid": 76337, "tid": -914061504, "ts": 1716454223224076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223224174, "dur": 1,
    "args": {
      "External id": 124740,
      "cbid": 251, "correlation": 124740
    }
  },
  {
    "ph": "f", "id": 124740, "pid": 76337, "tid": -914061504, "ts": 1716454223224174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223372695, "dur": 205,
    "args": {
      "External id": 124742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124742,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124742, "pid": 5, "tid": 7, "ts": 1716454223372695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223224180, "dur": 13,
    "args": {
      "External id": 124742,
      "cbid": 211, "correlation": 124742
    }
  },
  {
    "ph": "s", "id": 124742, "pid": 76337, "tid": -914061504, "ts": 1716454223224180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223372902, "dur": 66,
    "args": {
      "External id": 124750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124750,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124750, "pid": 5, "tid": 7, "ts": 1716454223372902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223224250, "dur": 13,
    "args": {
      "External id": 124750,
      "cbid": 211, "correlation": 124750
    }
  },
  {
    "ph": "s", "id": 124750, "pid": 76337, "tid": -914061504, "ts": 1716454223224250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223372970, "dur": 97,
    "args": {
      "External id": 124758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124758,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124758, "pid": 5, "tid": 7, "ts": 1716454223372970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223224290, "dur": 78,
    "args": {
      "External id": 124758,
      "cbid": 211, "correlation": 124758
    }
  },
  {
    "ph": "s", "id": 124758, "pid": 76337, "tid": -914061504, "ts": 1716454223224290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223373068, "dur": 54,
    "args": {
      "External id": 124769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124769,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124769, "pid": 5, "tid": 7, "ts": 1716454223373068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223224431, "dur": 1887,
    "args": {
      "External id": 124769,
      "cbid": 211, "correlation": 124769
    }
  },
  {
    "ph": "s", "id": 124769, "pid": 76337, "tid": -914061504, "ts": 1716454223224431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223373123, "dur": 91,
    "args": {
      "External id": 124791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124791,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124791, "pid": 5, "tid": 7, "ts": 1716454223373123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226337, "dur": 126,
    "args": {
      "External id": 124791,
      "cbid": 211, "correlation": 124791
    }
  },
  {
    "ph": "s", "id": 124791, "pid": 76337, "tid": -914061504, "ts": 1716454223226337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226540, "dur": 1,
    "args": {
      "External id": 124802,
      "cbid": 251, "correlation": 124802
    }
  },
  {
    "ph": "f", "id": 124802, "pid": 76337, "tid": -914061504, "ts": 1716454223226540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223373215, "dur": 102,
    "args": {
      "External id": 124803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124803,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124803, "pid": 5, "tid": 7, "ts": 1716454223373215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226546, "dur": 13,
    "args": {
      "External id": 124803,
      "cbid": 211, "correlation": 124803
    }
  },
  {
    "ph": "s", "id": 124803, "pid": 76337, "tid": -914061504, "ts": 1716454223226546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226617, "dur": 1,
    "args": {
      "External id": 124814,
      "cbid": 251, "correlation": 124814
    }
  },
  {
    "ph": "f", "id": 124814, "pid": 76337, "tid": -914061504, "ts": 1716454223226617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226621, "dur": 0,
    "args": {
      "External id": 124815,
      "cbid": 251, "correlation": 124815
    }
  },
  {
    "ph": "f", "id": 124815, "pid": 76337, "tid": -914061504, "ts": 1716454223226621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223373319, "dur": 10,
    "args": {
      "External id": 124816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124816,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 124816, "pid": 5, "tid": 7, "ts": 1716454223373319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226622, "dur": 13,
    "args": {
      "External id": 124816,
      "cbid": 211, "correlation": 124816
    }
  },
  {
    "ph": "s", "id": 124816, "pid": 76337, "tid": -914061504, "ts": 1716454223226622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223373330, "dur": 5,
    "args": {
      "External id": 124818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124818,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 124818, "pid": 5, "tid": 7, "ts": 1716454223373330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226637, "dur": 6,
    "args": {
      "External id": 124818,
      "cbid": 211, "correlation": 124818
    }
  },
  {
    "ph": "s", "id": 124818, "pid": 76337, "tid": -914061504, "ts": 1716454223226637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226699, "dur": 1,
    "args": {
      "External id": 124829,
      "cbid": 251, "correlation": 124829
    }
  },
  {
    "ph": "f", "id": 124829, "pid": 76337, "tid": -914061504, "ts": 1716454223226699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226702, "dur": 0,
    "args": {
      "External id": 124830,
      "cbid": 251, "correlation": 124830
    }
  },
  {
    "ph": "f", "id": 124830, "pid": 76337, "tid": -914061504, "ts": 1716454223226702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223373337, "dur": 6,
    "args": {
      "External id": 124831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124831,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 124831, "pid": 5, "tid": 7, "ts": 1716454223373337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226703, "dur": 12,
    "args": {
      "External id": 124831,
      "cbid": 211, "correlation": 124831
    }
  },
  {
    "ph": "s", "id": 124831, "pid": 76337, "tid": -914061504, "ts": 1716454223226703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223373344, "dur": 3,
    "args": {
      "External id": 124833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124833,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 124833, "pid": 5, "tid": 7, "ts": 1716454223373344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226717, "dur": 6,
    "args": {
      "External id": 124833,
      "cbid": 211, "correlation": 124833
    }
  },
  {
    "ph": "s", "id": 124833, "pid": 76337, "tid": -914061504, "ts": 1716454223226717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223373349, "dur": 155,
    "args": {
      "External id": 124854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124854,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124854, "pid": 5, "tid": 7, "ts": 1716454223373349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226791, "dur": 12,
    "args": {
      "External id": 124854,
      "cbid": 211, "correlation": 124854
    }
  },
  {
    "ph": "s", "id": 124854, "pid": 76337, "tid": -914061504, "ts": 1716454223226791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226888, "dur": 1,
    "args": {
      "External id": 124872,
      "cbid": 251, "correlation": 124872
    }
  },
  {
    "ph": "f", "id": 124872, "pid": 76337, "tid": -914061504, "ts": 1716454223226888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223373506, "dur": 108,
    "args": {
      "External id": 124874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124874,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 124874, "pid": 5, "tid": 7, "ts": 1716454223373506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226894, "dur": 13,
    "args": {
      "External id": 124874,
      "cbid": 211, "correlation": 124874
    }
  },
  {
    "ph": "s", "id": 124874, "pid": 76337, "tid": -914061504, "ts": 1716454223226894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223373615, "dur": 35,
    "args": {
      "External id": 124882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124882,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124882, "pid": 5, "tid": 7, "ts": 1716454223373615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223226963, "dur": 22,
    "args": {
      "External id": 124882,
      "cbid": 211, "correlation": 124882
    }
  },
  {
    "ph": "s", "id": 124882, "pid": 76337, "tid": -914061504, "ts": 1716454223226963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223373651, "dur": 67,
    "args": {
      "External id": 124890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124890, "pid": 5, "tid": 7, "ts": 1716454223373651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227014, "dur": 10,
    "args": {
      "External id": 124890,
      "cbid": 211, "correlation": 124890
    }
  },
  {
    "ph": "s", "id": 124890, "pid": 76337, "tid": -914061504, "ts": 1716454223227014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223373719, "dur": 92,
    "args": {
      "External id": 124912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124912,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124912, "pid": 5, "tid": 7, "ts": 1716454223373719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227066, "dur": 10,
    "args": {
      "External id": 124912,
      "cbid": 211, "correlation": 124912
    }
  },
  {
    "ph": "s", "id": 124912, "pid": 76337, "tid": -914061504, "ts": 1716454223227066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227151, "dur": 1,
    "args": {
      "External id": 124928,
      "cbid": 251, "correlation": 124928
    }
  },
  {
    "ph": "f", "id": 124928, "pid": 76337, "tid": -914061504, "ts": 1716454223227151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223373812, "dur": 570,
    "args": {
      "External id": 124930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124930,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 124930, "pid": 5, "tid": 7, "ts": 1716454223373812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227157, "dur": 13,
    "args": {
      "External id": 124930,
      "cbid": 211, "correlation": 124930
    }
  },
  {
    "ph": "s", "id": 124930, "pid": 76337, "tid": -914061504, "ts": 1716454223227157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223374383, "dur": 242,
    "args": {
      "External id": 124938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124938,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124938, "pid": 5, "tid": 7, "ts": 1716454223374383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227222, "dur": 13,
    "args": {
      "External id": 124938,
      "cbid": 211, "correlation": 124938
    }
  },
  {
    "ph": "s", "id": 124938, "pid": 76337, "tid": -914061504, "ts": 1716454223227222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223374627, "dur": 251,
    "args": {
      "External id": 124946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124946, "pid": 5, "tid": 7, "ts": 1716454223374627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227254, "dur": 8,
    "args": {
      "External id": 124946,
      "cbid": 211, "correlation": 124946
    }
  },
  {
    "ph": "s", "id": 124946, "pid": 76337, "tid": -914061504, "ts": 1716454223227254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227335, "dur": 1,
    "args": {
      "External id": 124962,
      "cbid": 251, "correlation": 124962
    }
  },
  {
    "ph": "f", "id": 124962, "pid": 76337, "tid": -914061504, "ts": 1716454223227335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227340, "dur": 0,
    "args": {
      "External id": 124964,
      "cbid": 251, "correlation": 124964
    }
  },
  {
    "ph": "f", "id": 124964, "pid": 76337, "tid": -914061504, "ts": 1716454223227340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223374879, "dur": 358,
    "args": {
      "External id": 124965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124965,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 124965, "pid": 5, "tid": 7, "ts": 1716454223374879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227343, "dur": 12,
    "args": {
      "External id": 124965,
      "cbid": 211, "correlation": 124965
    }
  },
  {
    "ph": "s", "id": 124965, "pid": 76337, "tid": -914061504, "ts": 1716454223227343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223375238, "dur": 50,
    "args": {
      "External id": 124973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124973,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124973, "pid": 5, "tid": 7, "ts": 1716454223375238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227385, "dur": 188,
    "args": {
      "External id": 124973,
      "cbid": 211, "correlation": 124973
    }
  },
  {
    "ph": "s", "id": 124973, "pid": 76337, "tid": -914061504, "ts": 1716454223227385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223375289, "dur": 158,
    "args": {
      "External id": 124984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 124984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 124984, "pid": 5, "tid": 7, "ts": 1716454223375289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227631, "dur": 69,
    "args": {
      "External id": 124984,
      "cbid": 211, "correlation": 124984
    }
  },
  {
    "ph": "s", "id": 124984, "pid": 76337, "tid": -914061504, "ts": 1716454223227631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227753, "dur": 0,
    "args": {
      "External id": 124996,
      "cbid": 317, "correlation": 124996
    }
  },
  {
    "ph": "f", "id": 124996, "pid": 76337, "tid": -914061504, "ts": 1716454223227753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227754, "dur": 0,
    "args": {
      "External id": 124997,
      "cbid": 203, "correlation": 124997
    }
  },
  {
    "ph": "f", "id": 124997, "pid": 76337, "tid": -914061504, "ts": 1716454223227754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227754, "dur": 0,
    "args": {
      "External id": 124998,
      "cbid": 205, "correlation": 124998
    }
  },
  {
    "ph": "f", "id": 124998, "pid": 76337, "tid": -914061504, "ts": 1716454223227754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227778, "dur": 1,
    "args": {
      "External id": 125002,
      "cbid": 251, "correlation": 125002
    }
  },
  {
    "ph": "f", "id": 125002, "pid": 76337, "tid": -914061504, "ts": 1716454223227778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227780, "dur": 0,
    "args": {
      "External id": 125003,
      "cbid": 251, "correlation": 125003
    }
  },
  {
    "ph": "f", "id": 125003, "pid": 76337, "tid": -914061504, "ts": 1716454223227780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227781, "dur": 0,
    "args": {
      "External id": 125004,
      "cbid": 251, "correlation": 125004
    }
  },
  {
    "ph": "f", "id": 125004, "pid": 76337, "tid": -914061504, "ts": 1716454223227781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227781, "dur": 0,
    "args": {
      "External id": 125005,
      "cbid": 251, "correlation": 125005
    }
  },
  {
    "ph": "f", "id": 125005, "pid": 76337, "tid": -914061504, "ts": 1716454223227781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227782, "dur": 0,
    "args": {
      "External id": 125006,
      "cbid": 251, "correlation": 125006
    }
  },
  {
    "ph": "f", "id": 125006, "pid": 76337, "tid": -914061504, "ts": 1716454223227782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227783, "dur": 0,
    "args": {
      "External id": 125007,
      "cbid": 251, "correlation": 125007
    }
  },
  {
    "ph": "f", "id": 125007, "pid": 76337, "tid": -914061504, "ts": 1716454223227783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227784, "dur": 0,
    "args": {
      "External id": 125008,
      "cbid": 251, "correlation": 125008
    }
  },
  {
    "ph": "f", "id": 125008, "pid": 76337, "tid": -914061504, "ts": 1716454223227784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227784, "dur": 0,
    "args": {
      "External id": 125009,
      "cbid": 251, "correlation": 125009
    }
  },
  {
    "ph": "f", "id": 125009, "pid": 76337, "tid": -914061504, "ts": 1716454223227784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227785, "dur": 0,
    "args": {
      "External id": 125010,
      "cbid": 251, "correlation": 125010
    }
  },
  {
    "ph": "f", "id": 125010, "pid": 76337, "tid": -914061504, "ts": 1716454223227785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223375448, "dur": 114,
    "args": {
      "External id": 125011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125011,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 125011, "pid": 5, "tid": 7, "ts": 1716454223375448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227787, "dur": 41,
    "args": {
      "External id": 125011,
      "cbid": 211, "correlation": 125011
    }
  },
  {
    "ph": "s", "id": 125011, "pid": 76337, "tid": -914061504, "ts": 1716454223227787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223375564, "dur": 60,
    "args": {
      "External id": 125017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125017, "pid": 5, "tid": 7, "ts": 1716454223375564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223227851, "dur": 280,
    "args": {
      "External id": 125017,
      "cbid": 211, "correlation": 125017
    }
  },
  {
    "ph": "s", "id": 125017, "pid": 76337, "tid": -914061504, "ts": 1716454223227851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223375626, "dur": 50,
    "args": {
      "External id": 125025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125025,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125025, "pid": 5, "tid": 7, "ts": 1716454223375626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228155, "dur": 9,
    "args": {
      "External id": 125025,
      "cbid": 211, "correlation": 125025
    }
  },
  {
    "ph": "s", "id": 125025, "pid": 76337, "tid": -914061504, "ts": 1716454223228155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223375677, "dur": 53,
    "args": {
      "External id": 125045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125045,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 125045, "pid": 5, "tid": 7, "ts": 1716454223375677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228236, "dur": 12,
    "args": {
      "External id": 125045,
      "cbid": 211, "correlation": 125045
    }
  },
  {
    "ph": "s", "id": 125045, "pid": 76337, "tid": -914061504, "ts": 1716454223228236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223375731, "dur": 4,
    "args": {
      "External id": 125057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125057,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 125057, "pid": 5, "tid": 7, "ts": 1716454223375731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228258, "dur": 7,
    "args": {
      "External id": 125057,
      "cbid": 211, "correlation": 125057
    }
  },
  {
    "ph": "s", "id": 125057, "pid": 76337, "tid": -914061504, "ts": 1716454223228258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223375737, "dur": 56,
    "args": {
      "External id": 125060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125060,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125060, "pid": 5, "tid": 7, "ts": 1716454223375737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228277, "dur": 112,
    "args": {
      "External id": 125060,
      "cbid": 211, "correlation": 125060
    }
  },
  {
    "ph": "s", "id": 125060, "pid": 76337, "tid": -914061504, "ts": 1716454223228277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223375794, "dur": 37,
    "args": {
      "External id": 125069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125069,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125069, "pid": 5, "tid": 7, "ts": 1716454223375794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228426, "dur": 11,
    "args": {
      "External id": 125069,
      "cbid": 211, "correlation": 125069
    }
  },
  {
    "ph": "s", "id": 125069, "pid": 76337, "tid": -914061504, "ts": 1716454223228426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228482, "dur": 0,
    "args": {
      "External id": 125079,
      "cbid": 317, "correlation": 125079
    }
  },
  {
    "ph": "f", "id": 125079, "pid": 76337, "tid": -914061504, "ts": 1716454223228482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228483, "dur": 0,
    "args": {
      "External id": 125080,
      "cbid": 203, "correlation": 125080
    }
  },
  {
    "ph": "f", "id": 125080, "pid": 76337, "tid": -914061504, "ts": 1716454223228483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228484, "dur": 0,
    "args": {
      "External id": 125081,
      "cbid": 205, "correlation": 125081
    }
  },
  {
    "ph": "f", "id": 125081, "pid": 76337, "tid": -914061504, "ts": 1716454223228484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223375832, "dur": 42,
    "args": {
      "External id": 125085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125085,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125085, "pid": 5, "tid": 7, "ts": 1716454223375832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228500, "dur": 12,
    "args": {
      "External id": 125085,
      "cbid": 211, "correlation": 125085
    }
  },
  {
    "ph": "s", "id": 125085, "pid": 76337, "tid": -914061504, "ts": 1716454223228500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223375875, "dur": 3,
    "args": {
      "External id": 125087,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125087,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.80327868852459
    }
  },
  {
    "ph": "f", "id": 125087, "pid": 5, "tid": 7, "ts": 1716454223375875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228515, "dur": 17,
    "args": {
      "External id": 125087,
      "cbid": 51, "correlation": 125087
    }
  },
  {
    "ph": "s", "id": 125087, "pid": 76337, "tid": -914061504, "ts": 1716454223228515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228538, "dur": 2,
    "args": {
      "External id": 125089,
      "cbid": 200, "correlation": 125089
    }
  },
  {
    "ph": "f", "id": 125089, "pid": 76337, "tid": -914061504, "ts": 1716454223228538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228540, "dur": 0,
    "args": {
      "External id": 125090,
      "cbid": 200, "correlation": 125090
    }
  },
  {
    "ph": "f", "id": 125090, "pid": 76337, "tid": -914061504, "ts": 1716454223228540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228541, "dur": 0,
    "args": {
      "External id": 125091,
      "cbid": 200, "correlation": 125091
    }
  },
  {
    "ph": "f", "id": 125091, "pid": 76337, "tid": -914061504, "ts": 1716454223228541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228541, "dur": 0,
    "args": {
      "External id": 125092,
      "cbid": 200, "correlation": 125092
    }
  },
  {
    "ph": "f", "id": 125092, "pid": 76337, "tid": -914061504, "ts": 1716454223228541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228543, "dur": 4,
    "args": {
      "External id": 125093,
      "cbid": 15, "correlation": 125093
    }
  },
  {
    "ph": "f", "id": 125093, "pid": 76337, "tid": -914061504, "ts": 1716454223228543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228547, "dur": 1,
    "args": {
      "External id": 125094,
      "cbid": 251, "correlation": 125094
    }
  },
  {
    "ph": "f", "id": 125094, "pid": 76337, "tid": -914061504, "ts": 1716454223228547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223375880, "dur": 25,
    "args": {
      "External id": 125095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125095,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125095, "pid": 5, "tid": 7, "ts": 1716454223375880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228550, "dur": 8,
    "args": {
      "External id": 125095,
      "cbid": 211, "correlation": 125095
    }
  },
  {
    "ph": "s", "id": 125095, "pid": 76337, "tid": -914061504, "ts": 1716454223228550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223375906, "dur": 4,
    "args": {
      "External id": 125097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 125097, "pid": 5, "tid": 7, "ts": 1716454223375906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228564, "dur": 6,
    "args": {
      "External id": 125097,
      "cbid": 211, "correlation": 125097
    }
  },
  {
    "ph": "s", "id": 125097, "pid": 76337, "tid": -914061504, "ts": 1716454223228564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228575, "dur": 0,
    "args": {
      "External id": 125098,
      "cbid": 51, "correlation": 125098
    }
  },
  {
    "ph": "s", "id": 125098, "pid": 76337, "tid": -914061504, "ts": 1716454223228575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223375912, "dur": 187,
    "args": {
      "External id": 125099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125099,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125099, "pid": 5, "tid": 7, "ts": 1716454223375912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228576, "dur": 203,
    "args": {
      "External id": 125099,
      "cbid": 211, "correlation": 125099
    }
  },
  {
    "ph": "s", "id": 125099, "pid": 76337, "tid": -914061504, "ts": 1716454223228576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223376101, "dur": 6,
    "args": {
      "External id": 125100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125100, "pid": 5, "tid": 7, "ts": 1716454223376101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228782, "dur": 6,
    "args": {
      "External id": 125100,
      "cbid": 211, "correlation": 125100
    }
  },
  {
    "ph": "s", "id": 125100, "pid": 76337, "tid": -914061504, "ts": 1716454223228782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223376108, "dur": 5,
    "args": {
      "External id": 125106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 125106, "pid": 5, "tid": 7, "ts": 1716454223376108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223228812, "dur": 9,
    "args": {
      "External id": 125106,
      "cbid": 211, "correlation": 125106
    }
  },
  {
    "ph": "s", "id": 125106, "pid": 76337, "tid": -914061504, "ts": 1716454223228812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223376114, "dur": 3,
    "args": {
      "External id": 125114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125114,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125114, "pid": 5, "tid": 7, "ts": 1716454223376114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223230504, "dur": 15,
    "args": {
      "External id": 125114,
      "cbid": 211, "correlation": 125114
    }
  },
  {
    "ph": "s", "id": 125114, "pid": 76337, "tid": -914061504, "ts": 1716454223230504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223376118, "dur": 3,
    "args": {
      "External id": 125122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125122,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125122, "pid": 5, "tid": 7, "ts": 1716454223376118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223230545, "dur": 10,
    "args": {
      "External id": 125122,
      "cbid": 211, "correlation": 125122
    }
  },
  {
    "ph": "s", "id": 125122, "pid": 76337, "tid": -914061504, "ts": 1716454223230545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223376123, "dur": 3,
    "args": {
      "External id": 125130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125130,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125130, "pid": 5, "tid": 7, "ts": 1716454223376123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223230572, "dur": 9,
    "args": {
      "External id": 125130,
      "cbid": 211, "correlation": 125130
    }
  },
  {
    "ph": "s", "id": 125130, "pid": 76337, "tid": -914061504, "ts": 1716454223230572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223376127, "dur": 3,
    "args": {
      "External id": 125139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125139,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125139, "pid": 5, "tid": 7, "ts": 1716454223376127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223230745, "dur": 14,
    "args": {
      "External id": 125139,
      "cbid": 211, "correlation": 125139
    }
  },
  {
    "ph": "s", "id": 125139, "pid": 76337, "tid": -914061504, "ts": 1716454223230745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223376131, "dur": 3,
    "args": {
      "External id": 125148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125148,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125148, "pid": 5, "tid": 7, "ts": 1716454223376131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223230775, "dur": 7,
    "args": {
      "External id": 125148,
      "cbid": 211, "correlation": 125148
    }
  },
  {
    "ph": "s", "id": 125148, "pid": 76337, "tid": -914061504, "ts": 1716454223230775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223376135, "dur": 3,
    "args": {
      "External id": 125156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125156,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125156, "pid": 5, "tid": 7, "ts": 1716454223376135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223230799, "dur": 8,
    "args": {
      "External id": 125156,
      "cbid": 211, "correlation": 125156
    }
  },
  {
    "ph": "s", "id": 125156, "pid": 76337, "tid": -914061504, "ts": 1716454223230799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223376139, "dur": 3,
    "args": {
      "External id": 125164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125164,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125164, "pid": 5, "tid": 7, "ts": 1716454223376139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223231066, "dur": 16,
    "args": {
      "External id": 125164,
      "cbid": 211, "correlation": 125164
    }
  },
  {
    "ph": "s", "id": 125164, "pid": 76337, "tid": -914061504, "ts": 1716454223231066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223376143, "dur": 3,
    "args": {
      "External id": 125172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125172,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125172, "pid": 5, "tid": 7, "ts": 1716454223376143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223231098, "dur": 8,
    "args": {
      "External id": 125172,
      "cbid": 211, "correlation": 125172
    }
  },
  {
    "ph": "s", "id": 125172, "pid": 76337, "tid": -914061504, "ts": 1716454223231098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376148, "dur": 1,
    "args": {
      "External id": 125182,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125182,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 125182, "pid": 5, "tid": 7, "ts": 1716454223376148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223231164, "dur": 32,
    "args": {
      "External id": 125182,
      "cbid": 41, "correlation": 125182
    }
  },
  {
    "ph": "s", "id": 125182, "pid": 76337, "tid": -914061504, "ts": 1716454223231164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223231197, "dur": 144968,
    "args": {
      "External id": 125183,
      "cbid": 131, "correlation": 125183
    }
  },
  {
    "ph": "f", "id": 125183, "pid": 76337, "tid": -914061504, "ts": 1716454223231197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223376322, "dur": 3,
    "args": {
      "External id": 125191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125191, "pid": 5, "tid": 7, "ts": 1716454223376322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376299, "dur": 24,
    "args": {
      "External id": 125191,
      "cbid": 211, "correlation": 125191
    }
  },
  {
    "ph": "s", "id": 125191, "pid": 76337, "tid": -914061504, "ts": 1716454223376299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376413, "dur": 3,
    "args": {
      "External id": 125200,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125200,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 125200, "pid": 5, "tid": 7, "ts": 1716454223376413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376384, "dur": 29,
    "args": {
      "External id": 125200,
      "cbid": 41, "correlation": 125200
    }
  },
  {
    "ph": "s", "id": 125200, "pid": 76337, "tid": -914061504, "ts": 1716454223376384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223376512, "dur": 4,
    "args": {
      "External id": 125210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125210,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125210, "pid": 5, "tid": 7, "ts": 1716454223376512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376496, "dur": 17,
    "args": {
      "External id": 125210,
      "cbid": 211, "correlation": 125210
    }
  },
  {
    "ph": "s", "id": 125210, "pid": 76337, "tid": -914061504, "ts": 1716454223376496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376585, "dur": 1,
    "args": {
      "External id": 125220,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125220,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 125220, "pid": 5, "tid": 7, "ts": 1716454223376585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376566, "dur": 17,
    "args": {
      "External id": 125220,
      "cbid": 41, "correlation": 125220
    }
  },
  {
    "ph": "s", "id": 125220, "pid": 76337, "tid": -914061504, "ts": 1716454223376566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376584, "dur": 8,
    "args": {
      "External id": 125221,
      "cbid": 131, "correlation": 125221
    }
  },
  {
    "ph": "f", "id": 125221, "pid": 76337, "tid": -914061504, "ts": 1716454223376584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376654, "dur": 3,
    "args": {
      "External id": 125228,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125228,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.72
    }
  },
  {
    "ph": "f", "id": 125228, "pid": 5, "tid": 7, "ts": 1716454223376654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376632, "dur": 22,
    "args": {
      "External id": 125228,
      "cbid": 41, "correlation": 125228
    }
  },
  {
    "ph": "s", "id": 125228, "pid": 76337, "tid": -914061504, "ts": 1716454223376632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376740, "dur": 3,
    "args": {
      "External id": 125247,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125247,
      "bytes": 16, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 125247, "pid": 5, "tid": 7, "ts": 1716454223376740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376722, "dur": 18,
    "args": {
      "External id": 125247,
      "cbid": 41, "correlation": 125247
    }
  },
  {
    "ph": "s", "id": 125247, "pid": 76337, "tid": -914061504, "ts": 1716454223376722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223376777, "dur": 3,
    "args": {
      "External id": 125253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125253, "pid": 5, "tid": 7, "ts": 1716454223376777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376766, "dur": 11,
    "args": {
      "External id": 125253,
      "cbid": 211, "correlation": 125253
    }
  },
  {
    "ph": "s", "id": 125253, "pid": 76337, "tid": -914061504, "ts": 1716454223376766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223376791, "dur": 6,
    "args": {
      "External id": 125255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125255,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 125255, "pid": 5, "tid": 7, "ts": 1716454223376791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376781, "dur": 10,
    "args": {
      "External id": 125255,
      "cbid": 211, "correlation": 125255
    }
  },
  {
    "ph": "s", "id": 125255, "pid": 76337, "tid": -914061504, "ts": 1716454223376781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454223376800, "dur": 3,
    "args": {
      "External id": 125257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125257, "pid": 5, "tid": 7, "ts": 1716454223376800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376792, "dur": 7,
    "args": {
      "External id": 125257,
      "cbid": 211, "correlation": 125257
    }
  },
  {
    "ph": "s", "id": 125257, "pid": 76337, "tid": -914061504, "ts": 1716454223376792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376835, "dur": 2,
    "args": {
      "External id": 125265,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125265,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002777777777777778
    }
  },
  {
    "ph": "f", "id": 125265, "pid": 5, "tid": 7, "ts": 1716454223376835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376821, "dur": 13,
    "args": {
      "External id": 125265,
      "cbid": 41, "correlation": 125265
    }
  },
  {
    "ph": "s", "id": 125265, "pid": 76337, "tid": -914061504, "ts": 1716454223376821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223376883, "dur": 3,
    "args": {
      "External id": 125279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125279, "pid": 5, "tid": 7, "ts": 1716454223376883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376872, "dur": 12,
    "args": {
      "External id": 125279,
      "cbid": 211, "correlation": 125279
    }
  },
  {
    "ph": "s", "id": 125279, "pid": 76337, "tid": -914061504, "ts": 1716454223376872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223376903, "dur": 2,
    "args": {
      "External id": 125293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125293,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125293, "pid": 5, "tid": 7, "ts": 1716454223376903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376895, "dur": 6,
    "args": {
      "External id": 125293,
      "cbid": 211, "correlation": 125293
    }
  },
  {
    "ph": "s", "id": 125293, "pid": 76337, "tid": -914061504, "ts": 1716454223376895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223376939, "dur": 6,
    "args": {
      "External id": 125300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125300,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125300, "pid": 5, "tid": 7, "ts": 1716454223376939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376929, "dur": 10,
    "args": {
      "External id": 125300,
      "cbid": 211, "correlation": 125300
    }
  },
  {
    "ph": "s", "id": 125300, "pid": 76337, "tid": -914061504, "ts": 1716454223376929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223376949, "dur": 6,
    "args": {
      "External id": 125303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125303,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125303, "pid": 5, "tid": 7, "ts": 1716454223376949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376942, "dur": 7,
    "args": {
      "External id": 125303,
      "cbid": 211, "correlation": 125303
    }
  },
  {
    "ph": "s", "id": 125303, "pid": 76337, "tid": -914061504, "ts": 1716454223376942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223376958, "dur": 3,
    "args": {
      "External id": 125305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125305,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125305, "pid": 5, "tid": 7, "ts": 1716454223376958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376950, "dur": 7,
    "args": {
      "External id": 125305,
      "cbid": 211, "correlation": 125305
    }
  },
  {
    "ph": "s", "id": 125305, "pid": 76337, "tid": -914061504, "ts": 1716454223376950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223376984, "dur": 2,
    "args": {
      "External id": 125308,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125308,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002777777777777778
    }
  },
  {
    "ph": "f", "id": 125308, "pid": 5, "tid": 7, "ts": 1716454223376984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223376965, "dur": 19,
    "args": {
      "External id": 125308,
      "cbid": 41, "correlation": 125308
    }
  },
  {
    "ph": "s", "id": 125308, "pid": 76337, "tid": -914061504, "ts": 1716454223376965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223377039, "dur": 4,
    "args": {
      "External id": 125324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125324,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125324, "pid": 5, "tid": 7, "ts": 1716454223377039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377026, "dur": 14,
    "args": {
      "External id": 125324,
      "cbid": 211, "correlation": 125324
    }
  },
  {
    "ph": "s", "id": 125324, "pid": 76337, "tid": -914061504, "ts": 1716454223377026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223377060, "dur": 3,
    "args": {
      "External id": 125329,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125329,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003156565656565657
    }
  },
  {
    "ph": "f", "id": 125329, "pid": 5, "tid": 7, "ts": 1716454223377060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377045, "dur": 15,
    "args": {
      "External id": 125329,
      "cbid": 41, "correlation": 125329
    }
  },
  {
    "ph": "s", "id": 125329, "pid": 76337, "tid": -914061504, "ts": 1716454223377045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223377088, "dur": 1,
    "args": {
      "External id": 125335,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125335,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 125335, "pid": 5, "tid": 7, "ts": 1716454223377088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377068, "dur": 29,
    "args": {
      "External id": 125335,
      "cbid": 41, "correlation": 125335
    }
  },
  {
    "ph": "s", "id": 125335, "pid": 76337, "tid": -914061504, "ts": 1716454223377068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377098, "dur": 4,
    "args": {
      "External id": 125336,
      "cbid": 131, "correlation": 125336
    }
  },
  {
    "ph": "f", "id": 125336, "pid": 76337, "tid": -914061504, "ts": 1716454223377098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223377151, "dur": 3,
    "args": {
      "External id": 125344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125344,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125344, "pid": 5, "tid": 7, "ts": 1716454223377151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377137, "dur": 14,
    "args": {
      "External id": 125344,
      "cbid": 211, "correlation": 125344
    }
  },
  {
    "ph": "s", "id": 125344, "pid": 76337, "tid": -914061504, "ts": 1716454223377137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223377182, "dur": 3,
    "args": {
      "External id": 125354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125354,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125354, "pid": 5, "tid": 7, "ts": 1716454223377182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377172, "dur": 9,
    "args": {
      "External id": 125354,
      "cbid": 211, "correlation": 125354
    }
  },
  {
    "ph": "s", "id": 125354, "pid": 76337, "tid": -914061504, "ts": 1716454223377172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223377212, "dur": 3,
    "args": {
      "External id": 125363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125363,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125363, "pid": 5, "tid": 7, "ts": 1716454223377212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377198, "dur": 13,
    "args": {
      "External id": 125363,
      "cbid": 211, "correlation": 125363
    }
  },
  {
    "ph": "s", "id": 125363, "pid": 76337, "tid": -914061504, "ts": 1716454223377198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223377329, "dur": 12,
    "args": {
      "External id": 125373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125373,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125373, "pid": 5, "tid": 7, "ts": 1716454223377329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377315, "dur": 15,
    "args": {
      "External id": 125373,
      "cbid": 211, "correlation": 125373
    }
  },
  {
    "ph": "s", "id": 125373, "pid": 76337, "tid": -914061504, "ts": 1716454223377315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223377368, "dur": 3,
    "args": {
      "External id": 125381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125381,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125381, "pid": 5, "tid": 7, "ts": 1716454223377368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377359, "dur": 8,
    "args": {
      "External id": 125381,
      "cbid": 211, "correlation": 125381
    }
  },
  {
    "ph": "s", "id": 125381, "pid": 76337, "tid": -914061504, "ts": 1716454223377359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223377413, "dur": 12,
    "args": {
      "External id": 125391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125391,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125391, "pid": 5, "tid": 7, "ts": 1716454223377413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377402, "dur": 11,
    "args": {
      "External id": 125391,
      "cbid": 211, "correlation": 125391
    }
  },
  {
    "ph": "s", "id": 125391, "pid": 76337, "tid": -914061504, "ts": 1716454223377402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223377445, "dur": 10,
    "args": {
      "External id": 125399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125399,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125399, "pid": 5, "tid": 7, "ts": 1716454223377445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377436, "dur": 9,
    "args": {
      "External id": 125399,
      "cbid": 211, "correlation": 125399
    }
  },
  {
    "ph": "s", "id": 125399, "pid": 76337, "tid": -914061504, "ts": 1716454223377436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223377472, "dur": 3,
    "args": {
      "External id": 125408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125408, "pid": 5, "tid": 7, "ts": 1716454223377472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377463, "dur": 9,
    "args": {
      "External id": 125408,
      "cbid": 211, "correlation": 125408
    }
  },
  {
    "ph": "s", "id": 125408, "pid": 76337, "tid": -914061504, "ts": 1716454223377463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223377495, "dur": 5,
    "args": {
      "External id": 125417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125417,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125417, "pid": 5, "tid": 7, "ts": 1716454223377495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377487, "dur": 7,
    "args": {
      "External id": 125417,
      "cbid": 211, "correlation": 125417
    }
  },
  {
    "ph": "s", "id": 125417, "pid": 76337, "tid": -914061504, "ts": 1716454223377487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223377534, "dur": 8,
    "args": {
      "External id": 125427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125427,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125427, "pid": 5, "tid": 7, "ts": 1716454223377534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377523, "dur": 11,
    "args": {
      "External id": 125427,
      "cbid": 211, "correlation": 125427
    }
  },
  {
    "ph": "s", "id": 125427, "pid": 76337, "tid": -914061504, "ts": 1716454223377523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223377849, "dur": 3,
    "args": {
      "External id": 125436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125436,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125436, "pid": 5, "tid": 7, "ts": 1716454223377849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377834, "dur": 15,
    "args": {
      "External id": 125436,
      "cbid": 211, "correlation": 125436
    }
  },
  {
    "ph": "s", "id": 125436, "pid": 76337, "tid": -914061504, "ts": 1716454223377834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223377878, "dur": 3,
    "args": {
      "External id": 125444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125444,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125444, "pid": 5, "tid": 7, "ts": 1716454223377878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377867, "dur": 10,
    "args": {
      "External id": 125444,
      "cbid": 211, "correlation": 125444
    }
  },
  {
    "ph": "s", "id": 125444, "pid": 76337, "tid": -914061504, "ts": 1716454223377867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223377930, "dur": 1,
    "args": {
      "External id": 125454,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125454,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 125454, "pid": 5, "tid": 7, "ts": 1716454223377930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377914, "dur": 14,
    "args": {
      "External id": 125454,
      "cbid": 41, "correlation": 125454
    }
  },
  {
    "ph": "s", "id": 125454, "pid": 76337, "tid": -914061504, "ts": 1716454223377914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223377929, "dur": 8,
    "args": {
      "External id": 125455,
      "cbid": 131, "correlation": 125455
    }
  },
  {
    "ph": "f", "id": 125455, "pid": 76337, "tid": -914061504, "ts": 1716454223377929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223378028, "dur": 2,
    "args": {
      "External id": 125463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125463, "pid": 5, "tid": 7, "ts": 1716454223378028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378012, "dur": 16,
    "args": {
      "External id": 125463,
      "cbid": 211, "correlation": 125463
    }
  },
  {
    "ph": "s", "id": 125463, "pid": 76337, "tid": -914061504, "ts": 1716454223378012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223378101, "dur": 3,
    "args": {
      "External id": 125472,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125472,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 125472, "pid": 5, "tid": 7, "ts": 1716454223378101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378083, "dur": 18,
    "args": {
      "External id": 125472,
      "cbid": 41, "correlation": 125472
    }
  },
  {
    "ph": "s", "id": 125472, "pid": 76337, "tid": -914061504, "ts": 1716454223378083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223378172, "dur": 3,
    "args": {
      "External id": 125482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125482,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125482, "pid": 5, "tid": 7, "ts": 1716454223378172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378158, "dur": 15,
    "args": {
      "External id": 125482,
      "cbid": 211, "correlation": 125482
    }
  },
  {
    "ph": "s", "id": 125482, "pid": 76337, "tid": -914061504, "ts": 1716454223378158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223378225, "dur": 1,
    "args": {
      "External id": 125492,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125492,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 125492, "pid": 5, "tid": 7, "ts": 1716454223378225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378210, "dur": 13,
    "args": {
      "External id": 125492,
      "cbid": 41, "correlation": 125492
    }
  },
  {
    "ph": "s", "id": 125492, "pid": 76337, "tid": -914061504, "ts": 1716454223378210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378224, "dur": 7,
    "args": {
      "External id": 125493,
      "cbid": 131, "correlation": 125493
    }
  },
  {
    "ph": "f", "id": 125493, "pid": 76337, "tid": -914061504, "ts": 1716454223378224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223378285, "dur": 3,
    "args": {
      "External id": 125500,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125500,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.670103092783506
    }
  },
  {
    "ph": "f", "id": 125500, "pid": 5, "tid": 7, "ts": 1716454223378285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378264, "dur": 20,
    "args": {
      "External id": 125500,
      "cbid": 41, "correlation": 125500
    }
  },
  {
    "ph": "s", "id": 125500, "pid": 76337, "tid": -914061504, "ts": 1716454223378264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223378332, "dur": 1,
    "args": {
      "External id": 125511,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125511,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 125511, "pid": 5, "tid": 7, "ts": 1716454223378332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378319, "dur": 11,
    "args": {
      "External id": 125511,
      "cbid": 41, "correlation": 125511
    }
  },
  {
    "ph": "s", "id": 125511, "pid": 76337, "tid": -914061504, "ts": 1716454223378319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378331, "dur": 8,
    "args": {
      "External id": 125512,
      "cbid": 131, "correlation": 125512
    }
  },
  {
    "ph": "f", "id": 125512, "pid": 76337, "tid": -914061504, "ts": 1716454223378331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223378381, "dur": 3,
    "args": {
      "External id": 125520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125520,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125520, "pid": 5, "tid": 7, "ts": 1716454223378381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378367, "dur": 14,
    "args": {
      "External id": 125520,
      "cbid": 211, "correlation": 125520
    }
  },
  {
    "ph": "s", "id": 125520, "pid": 76337, "tid": -914061504, "ts": 1716454223378367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223378410, "dur": 3,
    "args": {
      "External id": 125530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125530,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125530, "pid": 5, "tid": 7, "ts": 1716454223378410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378401, "dur": 8,
    "args": {
      "External id": 125530,
      "cbid": 211, "correlation": 125530
    }
  },
  {
    "ph": "s", "id": 125530, "pid": 76337, "tid": -914061504, "ts": 1716454223378401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223378433, "dur": 3,
    "args": {
      "External id": 125539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125539,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125539, "pid": 5, "tid": 7, "ts": 1716454223378433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378424, "dur": 8,
    "args": {
      "External id": 125539,
      "cbid": 211, "correlation": 125539
    }
  },
  {
    "ph": "s", "id": 125539, "pid": 76337, "tid": -914061504, "ts": 1716454223378424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223378500, "dur": 5,
    "args": {
      "External id": 125547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125547,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125547, "pid": 5, "tid": 7, "ts": 1716454223378500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378487, "dur": 13,
    "args": {
      "External id": 125547,
      "cbid": 211, "correlation": 125547
    }
  },
  {
    "ph": "s", "id": 125547, "pid": 76337, "tid": -914061504, "ts": 1716454223378487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223378541, "dur": 3,
    "args": {
      "External id": 125556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125556,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125556, "pid": 5, "tid": 7, "ts": 1716454223378541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378531, "dur": 9,
    "args": {
      "External id": 125556,
      "cbid": 211, "correlation": 125556
    }
  },
  {
    "ph": "s", "id": 125556, "pid": 76337, "tid": -914061504, "ts": 1716454223378531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223378564, "dur": 3,
    "args": {
      "External id": 125565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125565,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125565, "pid": 5, "tid": 7, "ts": 1716454223378564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378555, "dur": 7,
    "args": {
      "External id": 125565,
      "cbid": 211, "correlation": 125565
    }
  },
  {
    "ph": "s", "id": 125565, "pid": 76337, "tid": -914061504, "ts": 1716454223378555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223378625, "dur": 3,
    "args": {
      "External id": 125573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125573,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125573, "pid": 5, "tid": 7, "ts": 1716454223378625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378614, "dur": 10,
    "args": {
      "External id": 125573,
      "cbid": 211, "correlation": 125573
    }
  },
  {
    "ph": "s", "id": 125573, "pid": 76337, "tid": -914061504, "ts": 1716454223378614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223378683, "dur": 1,
    "args": {
      "External id": 125581,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125581,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 125581, "pid": 5, "tid": 7, "ts": 1716454223378683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378667, "dur": 26,
    "args": {
      "External id": 125581,
      "cbid": 41, "correlation": 125581
    }
  },
  {
    "ph": "s", "id": 125581, "pid": 76337, "tid": -914061504, "ts": 1716454223378667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378694, "dur": 3,
    "args": {
      "External id": 125582,
      "cbid": 131, "correlation": 125582
    }
  },
  {
    "ph": "f", "id": 125582, "pid": 76337, "tid": -914061504, "ts": 1716454223378694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223378755, "dur": 1,
    "args": {
      "External id": 125592,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125592,
      "bytes": 42, "memory bandwidth (GB/s)": 0.027925531914893616
    }
  },
  {
    "ph": "f", "id": 125592, "pid": 5, "tid": 7, "ts": 1716454223378755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378742, "dur": 10,
    "args": {
      "External id": 125592,
      "cbid": 41, "correlation": 125592
    }
  },
  {
    "ph": "s", "id": 125592, "pid": 76337, "tid": -914061504, "ts": 1716454223378742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378754, "dur": 8,
    "args": {
      "External id": 125593,
      "cbid": 131, "correlation": 125593
    }
  },
  {
    "ph": "f", "id": 125593, "pid": 76337, "tid": -914061504, "ts": 1716454223378754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223378809, "dur": 1,
    "args": {
      "External id": 125602,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125602,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 125602, "pid": 5, "tid": 7, "ts": 1716454223378809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378798, "dur": 8,
    "args": {
      "External id": 125602,
      "cbid": 41, "correlation": 125602
    }
  },
  {
    "ph": "s", "id": 125602, "pid": 76337, "tid": -914061504, "ts": 1716454223378798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378807, "dur": 8,
    "args": {
      "External id": 125603,
      "cbid": 131, "correlation": 125603
    }
  },
  {
    "ph": "f", "id": 125603, "pid": 76337, "tid": -914061504, "ts": 1716454223378807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223378880, "dur": 4,
    "args": {
      "External id": 125610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125610,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125610, "pid": 5, "tid": 7, "ts": 1716454223378880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378864, "dur": 17,
    "args": {
      "External id": 125610,
      "cbid": 211, "correlation": 125610
    }
  },
  {
    "ph": "s", "id": 125610, "pid": 76337, "tid": -914061504, "ts": 1716454223378864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454223378918, "dur": 4,
    "args": {
      "External id": 125630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125630,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125630, "pid": 5, "tid": 7, "ts": 1716454223378918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378906, "dur": 13,
    "args": {
      "External id": 125630,
      "cbid": 211, "correlation": 125630
    }
  },
  {
    "ph": "s", "id": 125630, "pid": 76337, "tid": -914061504, "ts": 1716454223378906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378920, "dur": 0,
    "args": {
      "External id": 125631,
      "cbid": 11, "correlation": 125631
    }
  },
  {
    "ph": "f", "id": 125631, "pid": 76337, "tid": -914061504, "ts": 1716454223378920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378921, "dur": 0,
    "args": {
      "External id": 125632,
      "cbid": 11, "correlation": 125632
    }
  },
  {
    "ph": "f", "id": 125632, "pid": 76337, "tid": -914061504, "ts": 1716454223378921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223378934, "dur": 1,
    "args": {
      "External id": 125635,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125635,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 125635, "pid": 5, "tid": 7, "ts": 1716454223378934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378922, "dur": 21,
    "args": {
      "External id": 125635,
      "cbid": 41, "correlation": 125635
    }
  },
  {
    "ph": "s", "id": 125635, "pid": 76337, "tid": -914061504, "ts": 1716454223378922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378944, "dur": 3,
    "args": {
      "External id": 125636,
      "cbid": 131, "correlation": 125636
    }
  },
  {
    "ph": "f", "id": 125636, "pid": 76337, "tid": -914061504, "ts": 1716454223378944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223378971, "dur": 3,
    "args": {
      "External id": 125660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125660, "pid": 5, "tid": 7, "ts": 1716454223378971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378961, "dur": 9,
    "args": {
      "External id": 125660,
      "cbid": 211, "correlation": 125660
    }
  },
  {
    "ph": "s", "id": 125660, "pid": 76337, "tid": -914061504, "ts": 1716454223378961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378971, "dur": 0,
    "args": {
      "External id": 125661,
      "cbid": 11, "correlation": 125661
    }
  },
  {
    "ph": "f", "id": 125661, "pid": 76337, "tid": -914061504, "ts": 1716454223378971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378972, "dur": 0,
    "args": {
      "External id": 125662,
      "cbid": 11, "correlation": 125662
    }
  },
  {
    "ph": "f", "id": 125662, "pid": 76337, "tid": -914061504, "ts": 1716454223378972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378980, "dur": 1,
    "args": {
      "External id": 125664,
      "cbid": 200, "correlation": 125664
    }
  },
  {
    "ph": "f", "id": 125664, "pid": 76337, "tid": -914061504, "ts": 1716454223378980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223378991, "dur": 4,
    "args": {
      "External id": 125666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125666,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125666, "pid": 5, "tid": 7, "ts": 1716454223378991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378983, "dur": 8,
    "args": {
      "External id": 125666,
      "cbid": 211, "correlation": 125666
    }
  },
  {
    "ph": "s", "id": 125666, "pid": 76337, "tid": -914061504, "ts": 1716454223378983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378992, "dur": 0,
    "args": {
      "External id": 125667,
      "cbid": 11, "correlation": 125667
    }
  },
  {
    "ph": "f", "id": 125667, "pid": 76337, "tid": -914061504, "ts": 1716454223378992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223378993, "dur": 0,
    "args": {
      "External id": 125668,
      "cbid": 11, "correlation": 125668
    }
  },
  {
    "ph": "f", "id": 125668, "pid": 76337, "tid": -914061504, "ts": 1716454223378993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223379032, "dur": 1,
    "args": {
      "External id": 125675,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125675,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004901960784313725
    }
  },
  {
    "ph": "f", "id": 125675, "pid": 5, "tid": 7, "ts": 1716454223379032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379020, "dur": 20,
    "args": {
      "External id": 125675,
      "cbid": 41, "correlation": 125675
    }
  },
  {
    "ph": "s", "id": 125675, "pid": 76337, "tid": -914061504, "ts": 1716454223379020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379041, "dur": 3,
    "args": {
      "External id": 125676,
      "cbid": 131, "correlation": 125676
    }
  },
  {
    "ph": "f", "id": 125676, "pid": 76337, "tid": -914061504, "ts": 1716454223379041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223379092, "dur": 1,
    "args": {
      "External id": 125686,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 125686,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 125686, "pid": 5, "tid": 7, "ts": 1716454223379092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379080, "dur": 10,
    "args": {
      "External id": 125686,
      "cbid": 41, "correlation": 125686
    }
  },
  {
    "ph": "s", "id": 125686, "pid": 76337, "tid": -914061504, "ts": 1716454223379080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379090, "dur": 7,
    "args": {
      "External id": 125687,
      "cbid": 131, "correlation": 125687
    }
  },
  {
    "ph": "f", "id": 125687, "pid": 76337, "tid": -914061504, "ts": 1716454223379090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223379161, "dur": 5,
    "args": {
      "External id": 125694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125694,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125694, "pid": 5, "tid": 7, "ts": 1716454223379161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379145, "dur": 17,
    "args": {
      "External id": 125694,
      "cbid": 211, "correlation": 125694
    }
  },
  {
    "ph": "s", "id": 125694, "pid": 76337, "tid": -914061504, "ts": 1716454223379145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223379231, "dur": 3,
    "args": {
      "External id": 125703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125703,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125703, "pid": 5, "tid": 7, "ts": 1716454223379231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379218, "dur": 13,
    "args": {
      "External id": 125703,
      "cbid": 211, "correlation": 125703
    }
  },
  {
    "ph": "s", "id": 125703, "pid": 76337, "tid": -914061504, "ts": 1716454223379218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223379266, "dur": 3,
    "args": {
      "External id": 125711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125711,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125711, "pid": 5, "tid": 7, "ts": 1716454223379266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379256, "dur": 10,
    "args": {
      "External id": 125711,
      "cbid": 211, "correlation": 125711
    }
  },
  {
    "ph": "s", "id": 125711, "pid": 76337, "tid": -914061504, "ts": 1716454223379256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223379300, "dur": 4,
    "args": {
      "External id": 125719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125719,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125719, "pid": 5, "tid": 7, "ts": 1716454223379300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379287, "dur": 13,
    "args": {
      "External id": 125719,
      "cbid": 211, "correlation": 125719
    }
  },
  {
    "ph": "s", "id": 125719, "pid": 76337, "tid": -914061504, "ts": 1716454223379287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223379330, "dur": 4,
    "args": {
      "External id": 125727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125727,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125727, "pid": 5, "tid": 7, "ts": 1716454223379330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379319, "dur": 9,
    "args": {
      "External id": 125727,
      "cbid": 211, "correlation": 125727
    }
  },
  {
    "ph": "s", "id": 125727, "pid": 76337, "tid": -914061504, "ts": 1716454223379319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223379355, "dur": 3,
    "args": {
      "External id": 125735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125735,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125735, "pid": 5, "tid": 7, "ts": 1716454223379355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379346, "dur": 8,
    "args": {
      "External id": 125735,
      "cbid": 211, "correlation": 125735
    }
  },
  {
    "ph": "s", "id": 125735, "pid": 76337, "tid": -914061504, "ts": 1716454223379346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223379380, "dur": 3,
    "args": {
      "External id": 125743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125743,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125743, "pid": 5, "tid": 7, "ts": 1716454223379380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379371, "dur": 8,
    "args": {
      "External id": 125743,
      "cbid": 211, "correlation": 125743
    }
  },
  {
    "ph": "s", "id": 125743, "pid": 76337, "tid": -914061504, "ts": 1716454223379371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223379402, "dur": 4,
    "args": {
      "External id": 125751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125751, "pid": 5, "tid": 7, "ts": 1716454223379402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379393, "dur": 8,
    "args": {
      "External id": 125751,
      "cbid": 211, "correlation": 125751
    }
  },
  {
    "ph": "s", "id": 125751, "pid": 76337, "tid": -914061504, "ts": 1716454223379393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223379421, "dur": 5,
    "args": {
      "External id": 125759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125759, "pid": 5, "tid": 7, "ts": 1716454223379421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379413, "dur": 6,
    "args": {
      "External id": 125759,
      "cbid": 211, "correlation": 125759
    }
  },
  {
    "ph": "s", "id": 125759, "pid": 76337, "tid": -914061504, "ts": 1716454223379413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223379440, "dur": 3,
    "args": {
      "External id": 125767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125767,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125767, "pid": 5, "tid": 7, "ts": 1716454223379440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379432, "dur": 6,
    "args": {
      "External id": 125767,
      "cbid": 211, "correlation": 125767
    }
  },
  {
    "ph": "s", "id": 125767, "pid": 76337, "tid": -914061504, "ts": 1716454223379432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223379499, "dur": 3,
    "args": {
      "External id": 125775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125775,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 125775, "pid": 5, "tid": 7, "ts": 1716454223379499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379487, "dur": 12,
    "args": {
      "External id": 125775,
      "cbid": 211, "correlation": 125775
    }
  },
  {
    "ph": "s", "id": 125775, "pid": 76337, "tid": -914061504, "ts": 1716454223379487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223379524, "dur": 4,
    "args": {
      "External id": 125783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125783, "pid": 5, "tid": 7, "ts": 1716454223379524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379515, "dur": 8,
    "args": {
      "External id": 125783,
      "cbid": 211, "correlation": 125783
    }
  },
  {
    "ph": "s", "id": 125783, "pid": 76337, "tid": -914061504, "ts": 1716454223379515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223379546, "dur": 4,
    "args": {
      "External id": 125791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125791,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125791, "pid": 5, "tid": 7, "ts": 1716454223379546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379537, "dur": 8,
    "args": {
      "External id": 125791,
      "cbid": 211, "correlation": 125791
    }
  },
  {
    "ph": "s", "id": 125791, "pid": 76337, "tid": -914061504, "ts": 1716454223379537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223379566, "dur": 3,
    "args": {
      "External id": 125799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125799,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 125799, "pid": 5, "tid": 7, "ts": 1716454223379566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379558, "dur": 6,
    "args": {
      "External id": 125799,
      "cbid": 211, "correlation": 125799
    }
  },
  {
    "ph": "s", "id": 125799, "pid": 76337, "tid": -914061504, "ts": 1716454223379558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223379963, "dur": 5,
    "args": {
      "External id": 125808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125808,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125808, "pid": 5, "tid": 7, "ts": 1716454223379963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379947, "dur": 17,
    "args": {
      "External id": 125808,
      "cbid": 211, "correlation": 125808
    }
  },
  {
    "ph": "s", "id": 125808, "pid": 76337, "tid": -914061504, "ts": 1716454223379947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223380008, "dur": 5,
    "args": {
      "External id": 125817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125817,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125817, "pid": 5, "tid": 7, "ts": 1716454223380008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223379997, "dur": 10,
    "args": {
      "External id": 125817,
      "cbid": 211, "correlation": 125817
    }
  },
  {
    "ph": "s", "id": 125817, "pid": 76337, "tid": -914061504, "ts": 1716454223379997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223380135, "dur": 3,
    "args": {
      "External id": 125833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125833, "pid": 5, "tid": 7, "ts": 1716454223380135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380121, "dur": 14,
    "args": {
      "External id": 125833,
      "cbid": 211, "correlation": 125833
    }
  },
  {
    "ph": "s", "id": 125833, "pid": 76337, "tid": -914061504, "ts": 1716454223380121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380168, "dur": 3,
    "args": {
      "External id": 125841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125841,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125841, "pid": 5, "tid": 7, "ts": 1716454223380168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380159, "dur": 9,
    "args": {
      "External id": 125841,
      "cbid": 211, "correlation": 125841
    }
  },
  {
    "ph": "s", "id": 125841, "pid": 76337, "tid": -914061504, "ts": 1716454223380159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380200, "dur": 3,
    "args": {
      "External id": 125849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125849,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125849, "pid": 5, "tid": 7, "ts": 1716454223380200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380190, "dur": 8,
    "args": {
      "External id": 125849,
      "cbid": 211, "correlation": 125849
    }
  },
  {
    "ph": "s", "id": 125849, "pid": 76337, "tid": -914061504, "ts": 1716454223380190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380231, "dur": 4,
    "args": {
      "External id": 125857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125857,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125857, "pid": 5, "tid": 7, "ts": 1716454223380231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380221, "dur": 9,
    "args": {
      "External id": 125857,
      "cbid": 211, "correlation": 125857
    }
  },
  {
    "ph": "s", "id": 125857, "pid": 76337, "tid": -914061504, "ts": 1716454223380221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223380287, "dur": 4,
    "args": {
      "External id": 125869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125869,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125869, "pid": 5, "tid": 7, "ts": 1716454223380287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380274, "dur": 13,
    "args": {
      "External id": 125869,
      "cbid": 211, "correlation": 125869
    }
  },
  {
    "ph": "s", "id": 125869, "pid": 76337, "tid": -914061504, "ts": 1716454223380274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223380332, "dur": 4,
    "args": {
      "External id": 125880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125880, "pid": 5, "tid": 7, "ts": 1716454223380332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380320, "dur": 12,
    "args": {
      "External id": 125880,
      "cbid": 211, "correlation": 125880
    }
  },
  {
    "ph": "s", "id": 125880, "pid": 76337, "tid": -914061504, "ts": 1716454223380320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380363, "dur": 3,
    "args": {
      "External id": 125888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125888, "pid": 5, "tid": 7, "ts": 1716454223380363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380353, "dur": 8,
    "args": {
      "External id": 125888,
      "cbid": 211, "correlation": 125888
    }
  },
  {
    "ph": "s", "id": 125888, "pid": 76337, "tid": -914061504, "ts": 1716454223380353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380397, "dur": 5,
    "args": {
      "External id": 125896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125896,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125896, "pid": 5, "tid": 7, "ts": 1716454223380397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380387, "dur": 10,
    "args": {
      "External id": 125896,
      "cbid": 211, "correlation": 125896
    }
  },
  {
    "ph": "s", "id": 125896, "pid": 76337, "tid": -914061504, "ts": 1716454223380387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380427, "dur": 5,
    "args": {
      "External id": 125904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125904,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125904, "pid": 5, "tid": 7, "ts": 1716454223380427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380416, "dur": 10,
    "args": {
      "External id": 125904,
      "cbid": 211, "correlation": 125904
    }
  },
  {
    "ph": "s", "id": 125904, "pid": 76337, "tid": -914061504, "ts": 1716454223380416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223380457, "dur": 4,
    "args": {
      "External id": 125913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125913,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125913, "pid": 5, "tid": 7, "ts": 1716454223380457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380447, "dur": 9,
    "args": {
      "External id": 125913,
      "cbid": 211, "correlation": 125913
    }
  },
  {
    "ph": "s", "id": 125913, "pid": 76337, "tid": -914061504, "ts": 1716454223380447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223380518, "dur": 5,
    "args": {
      "External id": 125926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125926,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 125926, "pid": 5, "tid": 7, "ts": 1716454223380518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380505, "dur": 14,
    "args": {
      "External id": 125926,
      "cbid": 211, "correlation": 125926
    }
  },
  {
    "ph": "s", "id": 125926, "pid": 76337, "tid": -914061504, "ts": 1716454223380505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223380559, "dur": 5,
    "args": {
      "External id": 125936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125936,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 125936, "pid": 5, "tid": 7, "ts": 1716454223380559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380548, "dur": 10,
    "args": {
      "External id": 125936,
      "cbid": 211, "correlation": 125936
    }
  },
  {
    "ph": "s", "id": 125936, "pid": 76337, "tid": -914061504, "ts": 1716454223380548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380686, "dur": 5,
    "args": {
      "External id": 125953,
      "cbid": 251, "correlation": 125953
    }
  },
  {
    "ph": "f", "id": 125953, "pid": 76337, "tid": -914061504, "ts": 1716454223380686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454223380717, "dur": 12,
    "args": {
      "External id": 125955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125955,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 125955, "pid": 5, "tid": 7, "ts": 1716454223380717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380700, "dur": 18,
    "args": {
      "External id": 125955,
      "cbid": 211, "correlation": 125955
    }
  },
  {
    "ph": "s", "id": 125955, "pid": 76337, "tid": -914061504, "ts": 1716454223380700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223380776, "dur": 3,
    "args": {
      "External id": 125963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125963,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 125963, "pid": 5, "tid": 7, "ts": 1716454223380776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380764, "dur": 12,
    "args": {
      "External id": 125963,
      "cbid": 211, "correlation": 125963
    }
  },
  {
    "ph": "s", "id": 125963, "pid": 76337, "tid": -914061504, "ts": 1716454223380764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380835, "dur": 2,
    "args": {
      "External id": 125979,
      "cbid": 251, "correlation": 125979
    }
  },
  {
    "ph": "f", "id": 125979, "pid": 76337, "tid": -914061504, "ts": 1716454223380835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380841, "dur": 0,
    "args": {
      "External id": 125981,
      "cbid": 251, "correlation": 125981
    }
  },
  {
    "ph": "f", "id": 125981, "pid": 76337, "tid": -914061504, "ts": 1716454223380841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223380857, "dur": 14,
    "args": {
      "External id": 125982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125982,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 125982, "pid": 5, "tid": 7, "ts": 1716454223380857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380844, "dur": 14,
    "args": {
      "External id": 125982,
      "cbid": 211, "correlation": 125982
    }
  },
  {
    "ph": "s", "id": 125982, "pid": 76337, "tid": -914061504, "ts": 1716454223380844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223380873, "dur": 5,
    "args": {
      "External id": 125984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 125984,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 125984, "pid": 5, "tid": 7, "ts": 1716454223380873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380862, "dur": 9,
    "args": {
      "External id": 125984,
      "cbid": 211, "correlation": 125984
    }
  },
  {
    "ph": "s", "id": 125984, "pid": 76337, "tid": -914061504, "ts": 1716454223380862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380982, "dur": 1,
    "args": {
      "External id": 125994,
      "cbid": 317, "correlation": 125994
    }
  },
  {
    "ph": "f", "id": 125994, "pid": 76337, "tid": -914061504, "ts": 1716454223380982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380983, "dur": 1,
    "args": {
      "External id": 125995,
      "cbid": 203, "correlation": 125995
    }
  },
  {
    "ph": "f", "id": 125995, "pid": 76337, "tid": -914061504, "ts": 1716454223380983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223380985, "dur": 1,
    "args": {
      "External id": 125996,
      "cbid": 205, "correlation": 125996
    }
  },
  {
    "ph": "f", "id": 125996, "pid": 76337, "tid": -914061504, "ts": 1716454223380985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223381043, "dur": 6,
    "args": {
      "External id": 126000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126000,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126000, "pid": 5, "tid": 7, "ts": 1716454223381043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223381027, "dur": 16,
    "args": {
      "External id": 126000,
      "cbid": 211, "correlation": 126000
    }
  },
  {
    "ph": "s", "id": 126000, "pid": 76337, "tid": -914061504, "ts": 1716454223381027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223381055, "dur": 4,
    "args": {
      "External id": 126002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126002,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 126002, "pid": 5, "tid": 7, "ts": 1716454223381055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223381047, "dur": 6,
    "args": {
      "External id": 126002,
      "cbid": 211, "correlation": 126002
    }
  },
  {
    "ph": "s", "id": 126002, "pid": 76337, "tid": -914061504, "ts": 1716454223381047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223381074, "dur": 3,
    "args": {
      "External id": 126004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126004, "pid": 5, "tid": 7, "ts": 1716454223381074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223381065, "dur": 8,
    "args": {
      "External id": 126004,
      "cbid": 211, "correlation": 126004
    }
  },
  {
    "ph": "s", "id": 126004, "pid": 76337, "tid": -914061504, "ts": 1716454223381065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223381080, "dur": 1,
    "args": {
      "External id": 126005,
      "cbid": 51, "correlation": 126005
    }
  },
  {
    "ph": "s", "id": 126005, "pid": 76337, "tid": -914061504, "ts": 1716454223381080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223381091, "dur": 84,
    "args": {
      "External id": 126006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126006,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126006, "pid": 5, "tid": 7, "ts": 1716454223381091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223381082, "dur": 8,
    "args": {
      "External id": 126006,
      "cbid": 211, "correlation": 126006
    }
  },
  {
    "ph": "s", "id": 126006, "pid": 76337, "tid": -914061504, "ts": 1716454223381082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223381177, "dur": 60,
    "args": {
      "External id": 126011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126011, "pid": 5, "tid": 7, "ts": 1716454223381177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223381120, "dur": 11,
    "args": {
      "External id": 126011,
      "cbid": 211, "correlation": 126011
    }
  },
  {
    "ph": "s", "id": 126011, "pid": 76337, "tid": -914061504, "ts": 1716454223381120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223382959, "dur": 51,
    "args": {
      "External id": 126031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126031,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 126031, "pid": 5, "tid": 7, "ts": 1716454223382959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223382943, "dur": 16,
    "args": {
      "External id": 126031,
      "cbid": 211, "correlation": 126031
    }
  },
  {
    "ph": "s", "id": 126031, "pid": 76337, "tid": -914061504, "ts": 1716454223382943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223383011, "dur": 5,
    "args": {
      "External id": 126043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126043,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126043, "pid": 5, "tid": 7, "ts": 1716454223383011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223382970, "dur": 16,
    "args": {
      "External id": 126043,
      "cbid": 211, "correlation": 126043
    }
  },
  {
    "ph": "s", "id": 126043, "pid": 76337, "tid": -914061504, "ts": 1716454223382970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223383017, "dur": 58,
    "args": {
      "External id": 126046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126046,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126046, "pid": 5, "tid": 7, "ts": 1716454223383017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383002, "dur": 7,
    "args": {
      "External id": 126046,
      "cbid": 211, "correlation": 126046
    }
  },
  {
    "ph": "s", "id": 126046, "pid": 76337, "tid": -914061504, "ts": 1716454223383002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223383076, "dur": 36,
    "args": {
      "External id": 126055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126055,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126055, "pid": 5, "tid": 7, "ts": 1716454223383076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383047, "dur": 10,
    "args": {
      "External id": 126055,
      "cbid": 211, "correlation": 126055
    }
  },
  {
    "ph": "s", "id": 126055, "pid": 76337, "tid": -914061504, "ts": 1716454223383047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383105, "dur": 0,
    "args": {
      "External id": 126065,
      "cbid": 317, "correlation": 126065
    }
  },
  {
    "ph": "f", "id": 126065, "pid": 76337, "tid": -914061504, "ts": 1716454223383105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383105, "dur": 0,
    "args": {
      "External id": 126066,
      "cbid": 203, "correlation": 126066
    }
  },
  {
    "ph": "f", "id": 126066, "pid": 76337, "tid": -914061504, "ts": 1716454223383105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383106, "dur": 0,
    "args": {
      "External id": 126067,
      "cbid": 205, "correlation": 126067
    }
  },
  {
    "ph": "f", "id": 126067, "pid": 76337, "tid": -914061504, "ts": 1716454223383106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223383139, "dur": 40,
    "args": {
      "External id": 126071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126071,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126071, "pid": 5, "tid": 7, "ts": 1716454223383139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383124, "dur": 13,
    "args": {
      "External id": 126071,
      "cbid": 211, "correlation": 126071
    }
  },
  {
    "ph": "s", "id": 126071, "pid": 76337, "tid": -914061504, "ts": 1716454223383124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223383180, "dur": 14,
    "args": {
      "External id": 126073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126073,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126073, "pid": 5, "tid": 7, "ts": 1716454223383180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383140, "dur": 6,
    "args": {
      "External id": 126073,
      "cbid": 211, "correlation": 126073
    }
  },
  {
    "ph": "s", "id": 126073, "pid": 76337, "tid": -914061504, "ts": 1716454223383140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223383195, "dur": 3,
    "args": {
      "External id": 126075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126075,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126075, "pid": 5, "tid": 7, "ts": 1716454223383195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383152, "dur": 6,
    "args": {
      "External id": 126075,
      "cbid": 211, "correlation": 126075
    }
  },
  {
    "ph": "s", "id": 126075, "pid": 76337, "tid": -914061504, "ts": 1716454223383152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383162, "dur": 0,
    "args": {
      "External id": 126076,
      "cbid": 51, "correlation": 126076
    }
  },
  {
    "ph": "s", "id": 126076, "pid": 76337, "tid": -914061504, "ts": 1716454223383162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223383200, "dur": 698,
    "args": {
      "External id": 126077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126077,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126077, "pid": 5, "tid": 7, "ts": 1716454223383200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383163, "dur": 7,
    "args": {
      "External id": 126077,
      "cbid": 211, "correlation": 126077
    }
  },
  {
    "ph": "s", "id": 126077, "pid": 76337, "tid": -914061504, "ts": 1716454223383163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223383900, "dur": 58,
    "args": {
      "External id": 126082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126082,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126082, "pid": 5, "tid": 7, "ts": 1716454223383900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383194, "dur": 8,
    "args": {
      "External id": 126082,
      "cbid": 211, "correlation": 126082
    }
  },
  {
    "ph": "s", "id": 126082, "pid": 76337, "tid": -914061504, "ts": 1716454223383194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223383960, "dur": 3,
    "args": {
      "External id": 126090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126090,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126090, "pid": 5, "tid": 7, "ts": 1716454223383960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383238, "dur": 10,
    "args": {
      "External id": 126090,
      "cbid": 211, "correlation": 126090
    }
  },
  {
    "ph": "s", "id": 126090, "pid": 76337, "tid": -914061504, "ts": 1716454223383238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383304, "dur": 2,
    "args": {
      "External id": 126106,
      "cbid": 251, "correlation": 126106
    }
  },
  {
    "ph": "f", "id": 126106, "pid": 76337, "tid": -914061504, "ts": 1716454223383304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383310, "dur": 0,
    "args": {
      "External id": 126108,
      "cbid": 251, "correlation": 126108
    }
  },
  {
    "ph": "f", "id": 126108, "pid": 76337, "tid": -914061504, "ts": 1716454223383310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223383964, "dur": 9,
    "args": {
      "External id": 126109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126109,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 126109, "pid": 5, "tid": 7, "ts": 1716454223383964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383312, "dur": 13,
    "args": {
      "External id": 126109,
      "cbid": 211, "correlation": 126109
    }
  },
  {
    "ph": "s", "id": 126109, "pid": 76337, "tid": -914061504, "ts": 1716454223383312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223383975, "dur": 4,
    "args": {
      "External id": 126111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126111,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 126111, "pid": 5, "tid": 7, "ts": 1716454223383975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383326, "dur": 6,
    "args": {
      "External id": 126111,
      "cbid": 211, "correlation": 126111
    }
  },
  {
    "ph": "s", "id": 126111, "pid": 76337, "tid": -914061504, "ts": 1716454223383326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223383980, "dur": 53,
    "args": {
      "External id": 126121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126121, "pid": 5, "tid": 7, "ts": 1716454223383980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383386, "dur": 11,
    "args": {
      "External id": 126121,
      "cbid": 211, "correlation": 126121
    }
  },
  {
    "ph": "s", "id": 126121, "pid": 76337, "tid": -914061504, "ts": 1716454223383386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223384035, "dur": 51,
    "args": {
      "External id": 126141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126141,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 126141, "pid": 5, "tid": 7, "ts": 1716454223384035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383452, "dur": 11,
    "args": {
      "External id": 126141,
      "cbid": 211, "correlation": 126141
    }
  },
  {
    "ph": "s", "id": 126141, "pid": 76337, "tid": -914061504, "ts": 1716454223383452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223384087, "dur": 4,
    "args": {
      "External id": 126153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126153,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126153, "pid": 5, "tid": 7, "ts": 1716454223384087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383473, "dur": 7,
    "args": {
      "External id": 126153,
      "cbid": 211, "correlation": 126153
    }
  },
  {
    "ph": "s", "id": 126153, "pid": 76337, "tid": -914061504, "ts": 1716454223383473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223384092, "dur": 55,
    "args": {
      "External id": 126156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126156,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126156, "pid": 5, "tid": 7, "ts": 1716454223384092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383492, "dur": 6,
    "args": {
      "External id": 126156,
      "cbid": 211, "correlation": 126156
    }
  },
  {
    "ph": "s", "id": 126156, "pid": 76337, "tid": -914061504, "ts": 1716454223383492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223384149, "dur": 38,
    "args": {
      "External id": 126165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126165,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126165, "pid": 5, "tid": 7, "ts": 1716454223384149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383534, "dur": 10,
    "args": {
      "External id": 126165,
      "cbid": 211, "correlation": 126165
    }
  },
  {
    "ph": "s", "id": 126165, "pid": 76337, "tid": -914061504, "ts": 1716454223383534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383604, "dur": 0,
    "args": {
      "External id": 126175,
      "cbid": 317, "correlation": 126175
    }
  },
  {
    "ph": "f", "id": 126175, "pid": 76337, "tid": -914061504, "ts": 1716454223383604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383605, "dur": 0,
    "args": {
      "External id": 126176,
      "cbid": 203, "correlation": 126176
    }
  },
  {
    "ph": "f", "id": 126176, "pid": 76337, "tid": -914061504, "ts": 1716454223383605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383606, "dur": 0,
    "args": {
      "External id": 126177,
      "cbid": 205, "correlation": 126177
    }
  },
  {
    "ph": "f", "id": 126177, "pid": 76337, "tid": -914061504, "ts": 1716454223383606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223384188, "dur": 40,
    "args": {
      "External id": 126181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126181,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126181, "pid": 5, "tid": 7, "ts": 1716454223384188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383619, "dur": 13,
    "args": {
      "External id": 126181,
      "cbid": 211, "correlation": 126181
    }
  },
  {
    "ph": "s", "id": 126181, "pid": 76337, "tid": -914061504, "ts": 1716454223383619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223384229, "dur": 14,
    "args": {
      "External id": 126183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126183,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126183, "pid": 5, "tid": 7, "ts": 1716454223384229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383634, "dur": 5,
    "args": {
      "External id": 126183,
      "cbid": 211, "correlation": 126183
    }
  },
  {
    "ph": "s", "id": 126183, "pid": 76337, "tid": -914061504, "ts": 1716454223383634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223384244, "dur": 3,
    "args": {
      "External id": 126185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126185, "pid": 5, "tid": 7, "ts": 1716454223384244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383643, "dur": 5,
    "args": {
      "External id": 126185,
      "cbid": 211, "correlation": 126185
    }
  },
  {
    "ph": "s", "id": 126185, "pid": 76337, "tid": -914061504, "ts": 1716454223383643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383652, "dur": 0,
    "args": {
      "External id": 126186,
      "cbid": 51, "correlation": 126186
    }
  },
  {
    "ph": "s", "id": 126186, "pid": 76337, "tid": -914061504, "ts": 1716454223383652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223384249, "dur": 693,
    "args": {
      "External id": 126187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126187,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126187, "pid": 5, "tid": 7, "ts": 1716454223384249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383653, "dur": 5,
    "args": {
      "External id": 126187,
      "cbid": 211, "correlation": 126187
    }
  },
  {
    "ph": "s", "id": 126187, "pid": 76337, "tid": -914061504, "ts": 1716454223383653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223384944, "dur": 59,
    "args": {
      "External id": 126192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126192, "pid": 5, "tid": 7, "ts": 1716454223384944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383681, "dur": 9,
    "args": {
      "External id": 126192,
      "cbid": 211, "correlation": 126192
    }
  },
  {
    "ph": "s", "id": 126192, "pid": 76337, "tid": -914061504, "ts": 1716454223383681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223385004, "dur": 50,
    "args": {
      "External id": 126200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126200,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126200, "pid": 5, "tid": 7, "ts": 1716454223385004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383713, "dur": 8,
    "args": {
      "External id": 126200,
      "cbid": 211, "correlation": 126200
    }
  },
  {
    "ph": "s", "id": 126200, "pid": 76337, "tid": -914061504, "ts": 1716454223383713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223385055, "dur": 35,
    "args": {
      "External id": 126208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126208,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126208, "pid": 5, "tid": 7, "ts": 1716454223385055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383745, "dur": 10,
    "args": {
      "External id": 126208,
      "cbid": 211, "correlation": 126208
    }
  },
  {
    "ph": "s", "id": 126208, "pid": 76337, "tid": -914061504, "ts": 1716454223383745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223385092, "dur": 52,
    "args": {
      "External id": 126228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126228,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 126228, "pid": 5, "tid": 7, "ts": 1716454223385092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383827, "dur": 13,
    "args": {
      "External id": 126228,
      "cbid": 211, "correlation": 126228
    }
  },
  {
    "ph": "s", "id": 126228, "pid": 76337, "tid": -914061504, "ts": 1716454223383827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223385145, "dur": 4,
    "args": {
      "External id": 126240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126240,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126240, "pid": 5, "tid": 7, "ts": 1716454223385145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383849, "dur": 6,
    "args": {
      "External id": 126240,
      "cbid": 211, "correlation": 126240
    }
  },
  {
    "ph": "s", "id": 126240, "pid": 76337, "tid": -914061504, "ts": 1716454223383849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223385150, "dur": 55,
    "args": {
      "External id": 126243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126243, "pid": 5, "tid": 7, "ts": 1716454223385150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383866, "dur": 6,
    "args": {
      "External id": 126243,
      "cbid": 211, "correlation": 126243
    }
  },
  {
    "ph": "s", "id": 126243, "pid": 76337, "tid": -914061504, "ts": 1716454223383866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383924, "dur": 0,
    "args": {
      "External id": 126254,
      "cbid": 317, "correlation": 126254
    }
  },
  {
    "ph": "f", "id": 126254, "pid": 76337, "tid": -914061504, "ts": 1716454223383924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383924, "dur": 0,
    "args": {
      "External id": 126255,
      "cbid": 203, "correlation": 126255
    }
  },
  {
    "ph": "f", "id": 126255, "pid": 76337, "tid": -914061504, "ts": 1716454223383924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383925, "dur": 0,
    "args": {
      "External id": 126256,
      "cbid": 205, "correlation": 126256
    }
  },
  {
    "ph": "f", "id": 126256, "pid": 76337, "tid": -914061504, "ts": 1716454223383925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383961, "dur": 2,
    "args": {
      "External id": 126260,
      "cbid": 251, "correlation": 126260
    }
  },
  {
    "ph": "f", "id": 126260, "pid": 76337, "tid": -914061504, "ts": 1716454223383961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383964, "dur": 1,
    "args": {
      "External id": 126261,
      "cbid": 251, "correlation": 126261
    }
  },
  {
    "ph": "f", "id": 126261, "pid": 76337, "tid": -914061504, "ts": 1716454223383964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383966, "dur": 1,
    "args": {
      "External id": 126262,
      "cbid": 251, "correlation": 126262
    }
  },
  {
    "ph": "f", "id": 126262, "pid": 76337, "tid": -914061504, "ts": 1716454223383966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383969, "dur": 1,
    "args": {
      "External id": 126263,
      "cbid": 251, "correlation": 126263
    }
  },
  {
    "ph": "f", "id": 126263, "pid": 76337, "tid": -914061504, "ts": 1716454223383969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383971, "dur": 1,
    "args": {
      "External id": 126264,
      "cbid": 251, "correlation": 126264
    }
  },
  {
    "ph": "f", "id": 126264, "pid": 76337, "tid": -914061504, "ts": 1716454223383971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383982, "dur": 1,
    "args": {
      "External id": 126265,
      "cbid": 251, "correlation": 126265
    }
  },
  {
    "ph": "f", "id": 126265, "pid": 76337, "tid": -914061504, "ts": 1716454223383982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383983, "dur": 1,
    "args": {
      "External id": 126266,
      "cbid": 251, "correlation": 126266
    }
  },
  {
    "ph": "f", "id": 126266, "pid": 76337, "tid": -914061504, "ts": 1716454223383983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383985, "dur": 1,
    "args": {
      "External id": 126267,
      "cbid": 251, "correlation": 126267
    }
  },
  {
    "ph": "f", "id": 126267, "pid": 76337, "tid": -914061504, "ts": 1716454223383985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383988, "dur": 0,
    "args": {
      "External id": 126268,
      "cbid": 251, "correlation": 126268
    }
  },
  {
    "ph": "f", "id": 126268, "pid": 76337, "tid": -914061504, "ts": 1716454223383988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223385206, "dur": 113,
    "args": {
      "External id": 126269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126269,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 126269, "pid": 5, "tid": 7, "ts": 1716454223385206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223383992, "dur": 15,
    "args": {
      "External id": 126269,
      "cbid": 211, "correlation": 126269
    }
  },
  {
    "ph": "s", "id": 126269, "pid": 76337, "tid": -914061504, "ts": 1716454223383992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223385321, "dur": 60,
    "args": {
      "External id": 126275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126275,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126275, "pid": 5, "tid": 7, "ts": 1716454223385321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384031, "dur": 9,
    "args": {
      "External id": 126275,
      "cbid": 211, "correlation": 126275
    }
  },
  {
    "ph": "s", "id": 126275, "pid": 76337, "tid": -914061504, "ts": 1716454223384031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223385382, "dur": 623,
    "args": {
      "External id": 126284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126284,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126284, "pid": 5, "tid": 7, "ts": 1716454223385382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384124, "dur": 17,
    "args": {
      "External id": 126284,
      "cbid": 211, "correlation": 126284
    }
  },
  {
    "ph": "s", "id": 126284, "pid": 76337, "tid": -914061504, "ts": 1716454223384124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223386007, "dur": 181,
    "args": {
      "External id": 126306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126306,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126306, "pid": 5, "tid": 7, "ts": 1716454223386007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384197, "dur": 11,
    "args": {
      "External id": 126306,
      "cbid": 211, "correlation": 126306
    }
  },
  {
    "ph": "s", "id": 126306, "pid": 76337, "tid": -914061504, "ts": 1716454223384197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384314, "dur": 2,
    "args": {
      "External id": 126317,
      "cbid": 251, "correlation": 126317
    }
  },
  {
    "ph": "f", "id": 126317, "pid": 76337, "tid": -914061504, "ts": 1716454223384314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223386189, "dur": 196,
    "args": {
      "External id": 126318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126318,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126318, "pid": 5, "tid": 7, "ts": 1716454223386189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384321, "dur": 14,
    "args": {
      "External id": 126318,
      "cbid": 211, "correlation": 126318
    }
  },
  {
    "ph": "s", "id": 126318, "pid": 76337, "tid": -914061504, "ts": 1716454223384321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384394, "dur": 1,
    "args": {
      "External id": 126329,
      "cbid": 251, "correlation": 126329
    }
  },
  {
    "ph": "f", "id": 126329, "pid": 76337, "tid": -914061504, "ts": 1716454223384394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223386386, "dur": 188,
    "args": {
      "External id": 126330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126330,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126330, "pid": 5, "tid": 7, "ts": 1716454223386386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384399, "dur": 12,
    "args": {
      "External id": 126330,
      "cbid": 211, "correlation": 126330
    }
  },
  {
    "ph": "s", "id": 126330, "pid": 76337, "tid": -914061504, "ts": 1716454223384399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384465, "dur": 1,
    "args": {
      "External id": 126341,
      "cbid": 251, "correlation": 126341
    }
  },
  {
    "ph": "f", "id": 126341, "pid": 76337, "tid": -914061504, "ts": 1716454223384465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223386576, "dur": 186,
    "args": {
      "External id": 126342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126342,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126342, "pid": 5, "tid": 7, "ts": 1716454223386576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384469, "dur": 11,
    "args": {
      "External id": 126342,
      "cbid": 211, "correlation": 126342
    }
  },
  {
    "ph": "s", "id": 126342, "pid": 76337, "tid": -914061504, "ts": 1716454223384469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223386763, "dur": 18594,
    "args": {
      "External id": 126363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126363,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 126363, "pid": 5, "tid": 7, "ts": 1716454223386763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384578, "dur": 14,
    "args": {
      "External id": 126363,
      "cbid": 211, "correlation": 126363
    }
  },
  {
    "ph": "s", "id": 126363, "pid": 76337, "tid": -914061504, "ts": 1716454223384578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384690, "dur": 2,
    "args": {
      "External id": 126381,
      "cbid": 251, "correlation": 126381
    }
  },
  {
    "ph": "f", "id": 126381, "pid": 76337, "tid": -914061504, "ts": 1716454223384690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223405358, "dur": 203,
    "args": {
      "External id": 126383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126383,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126383, "pid": 5, "tid": 7, "ts": 1716454223405358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384697, "dur": 14,
    "args": {
      "External id": 126383,
      "cbid": 211, "correlation": 126383
    }
  },
  {
    "ph": "s", "id": 126383, "pid": 76337, "tid": -914061504, "ts": 1716454223384697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223405562, "dur": 66,
    "args": {
      "External id": 126391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126391,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126391, "pid": 5, "tid": 7, "ts": 1716454223405562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384770, "dur": 12,
    "args": {
      "External id": 126391,
      "cbid": 211, "correlation": 126391
    }
  },
  {
    "ph": "s", "id": 126391, "pid": 76337, "tid": -914061504, "ts": 1716454223384770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223405630, "dur": 98,
    "args": {
      "External id": 126399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126399,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126399, "pid": 5, "tid": 7, "ts": 1716454223405630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384811, "dur": 8,
    "args": {
      "External id": 126399,
      "cbid": 211, "correlation": 126399
    }
  },
  {
    "ph": "s", "id": 126399, "pid": 76337, "tid": -914061504, "ts": 1716454223384811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223405729, "dur": 53,
    "args": {
      "External id": 126410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126410,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126410, "pid": 5, "tid": 7, "ts": 1716454223405729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384895, "dur": 15,
    "args": {
      "External id": 126410,
      "cbid": 211, "correlation": 126410
    }
  },
  {
    "ph": "s", "id": 126410, "pid": 76337, "tid": -914061504, "ts": 1716454223384895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223405784, "dur": 92,
    "args": {
      "External id": 126432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126432,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126432, "pid": 5, "tid": 7, "ts": 1716454223405784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223384928, "dur": 8,
    "args": {
      "External id": 126432,
      "cbid": 211, "correlation": 126432
    }
  },
  {
    "ph": "s", "id": 126432, "pid": 76337, "tid": -914061504, "ts": 1716454223384928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385020, "dur": 1,
    "args": {
      "External id": 126443,
      "cbid": 251, "correlation": 126443
    }
  },
  {
    "ph": "f", "id": 126443, "pid": 76337, "tid": -914061504, "ts": 1716454223385020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223405877, "dur": 102,
    "args": {
      "External id": 126444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126444,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126444, "pid": 5, "tid": 7, "ts": 1716454223405877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385026, "dur": 13,
    "args": {
      "External id": 126444,
      "cbid": 211, "correlation": 126444
    }
  },
  {
    "ph": "s", "id": 126444, "pid": 76337, "tid": -914061504, "ts": 1716454223385026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385112, "dur": 1,
    "args": {
      "External id": 126455,
      "cbid": 251, "correlation": 126455
    }
  },
  {
    "ph": "f", "id": 126455, "pid": 76337, "tid": -914061504, "ts": 1716454223385112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385116, "dur": 0,
    "args": {
      "External id": 126456,
      "cbid": 251, "correlation": 126456
    }
  },
  {
    "ph": "f", "id": 126456, "pid": 76337, "tid": -914061504, "ts": 1716454223385116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223405980, "dur": 11,
    "args": {
      "External id": 126457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126457,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 126457, "pid": 5, "tid": 7, "ts": 1716454223405980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385118, "dur": 13,
    "args": {
      "External id": 126457,
      "cbid": 211, "correlation": 126457
    }
  },
  {
    "ph": "s", "id": 126457, "pid": 76337, "tid": -914061504, "ts": 1716454223385118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223405992, "dur": 5,
    "args": {
      "External id": 126459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126459,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 126459, "pid": 5, "tid": 7, "ts": 1716454223405992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385134, "dur": 8,
    "args": {
      "External id": 126459,
      "cbid": 211, "correlation": 126459
    }
  },
  {
    "ph": "s", "id": 126459, "pid": 76337, "tid": -914061504, "ts": 1716454223385134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385198, "dur": 1,
    "args": {
      "External id": 126470,
      "cbid": 251, "correlation": 126470
    }
  },
  {
    "ph": "f", "id": 126470, "pid": 76337, "tid": -914061504, "ts": 1716454223385198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385201, "dur": 0,
    "args": {
      "External id": 126471,
      "cbid": 251, "correlation": 126471
    }
  },
  {
    "ph": "f", "id": 126471, "pid": 76337, "tid": -914061504, "ts": 1716454223385201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223405998, "dur": 6,
    "args": {
      "External id": 126472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126472,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 126472, "pid": 5, "tid": 7, "ts": 1716454223405998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385203, "dur": 12,
    "args": {
      "External id": 126472,
      "cbid": 211, "correlation": 126472
    }
  },
  {
    "ph": "s", "id": 126472, "pid": 76337, "tid": -914061504, "ts": 1716454223385203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223406006, "dur": 3,
    "args": {
      "External id": 126474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126474,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 126474, "pid": 5, "tid": 7, "ts": 1716454223406006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385216, "dur": 5,
    "args": {
      "External id": 126474,
      "cbid": 211, "correlation": 126474
    }
  },
  {
    "ph": "s", "id": 126474, "pid": 76337, "tid": -914061504, "ts": 1716454223385216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223406010, "dur": 156,
    "args": {
      "External id": 126495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126495,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 126495, "pid": 5, "tid": 7, "ts": 1716454223406010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385289, "dur": 12,
    "args": {
      "External id": 126495,
      "cbid": 211, "correlation": 126495
    }
  },
  {
    "ph": "s", "id": 126495, "pid": 76337, "tid": -914061504, "ts": 1716454223385289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385386, "dur": 2,
    "args": {
      "External id": 126513,
      "cbid": 251, "correlation": 126513
    }
  },
  {
    "ph": "f", "id": 126513, "pid": 76337, "tid": -914061504, "ts": 1716454223385386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223406167, "dur": 105,
    "args": {
      "External id": 126515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126515,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 126515, "pid": 5, "tid": 7, "ts": 1716454223406167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385394, "dur": 13,
    "args": {
      "External id": 126515,
      "cbid": 211, "correlation": 126515
    }
  },
  {
    "ph": "s", "id": 126515, "pid": 76337, "tid": -914061504, "ts": 1716454223385394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223406274, "dur": 35,
    "args": {
      "External id": 126523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126523,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126523, "pid": 5, "tid": 7, "ts": 1716454223406274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385464, "dur": 12,
    "args": {
      "External id": 126523,
      "cbid": 211, "correlation": 126523
    }
  },
  {
    "ph": "s", "id": 126523, "pid": 76337, "tid": -914061504, "ts": 1716454223385464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223406310, "dur": 66,
    "args": {
      "External id": 126531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126531, "pid": 5, "tid": 7, "ts": 1716454223406310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385504, "dur": 9,
    "args": {
      "External id": 126531,
      "cbid": 211, "correlation": 126531
    }
  },
  {
    "ph": "s", "id": 126531, "pid": 76337, "tid": -914061504, "ts": 1716454223385504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223406378, "dur": 92,
    "args": {
      "External id": 126553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126553,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126553, "pid": 5, "tid": 7, "ts": 1716454223406378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385555, "dur": 10,
    "args": {
      "External id": 126553,
      "cbid": 211, "correlation": 126553
    }
  },
  {
    "ph": "s", "id": 126553, "pid": 76337, "tid": -914061504, "ts": 1716454223385555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385645, "dur": 1,
    "args": {
      "External id": 126569,
      "cbid": 251, "correlation": 126569
    }
  },
  {
    "ph": "f", "id": 126569, "pid": 76337, "tid": -914061504, "ts": 1716454223385645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223406471, "dur": 576,
    "args": {
      "External id": 126571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126571,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126571, "pid": 5, "tid": 7, "ts": 1716454223406471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385651, "dur": 12,
    "args": {
      "External id": 126571,
      "cbid": 211, "correlation": 126571
    }
  },
  {
    "ph": "s", "id": 126571, "pid": 76337, "tid": -914061504, "ts": 1716454223385651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223407049, "dur": 246,
    "args": {
      "External id": 126579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126579,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126579, "pid": 5, "tid": 7, "ts": 1716454223407049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385728, "dur": 15,
    "args": {
      "External id": 126579,
      "cbid": 211, "correlation": 126579
    }
  },
  {
    "ph": "s", "id": 126579, "pid": 76337, "tid": -914061504, "ts": 1716454223385728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223407296, "dur": 252,
    "args": {
      "External id": 126587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126587,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126587, "pid": 5, "tid": 7, "ts": 1716454223407296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385763, "dur": 9,
    "args": {
      "External id": 126587,
      "cbid": 211, "correlation": 126587
    }
  },
  {
    "ph": "s", "id": 126587, "pid": 76337, "tid": -914061504, "ts": 1716454223385763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385849, "dur": 2,
    "args": {
      "External id": 126603,
      "cbid": 251, "correlation": 126603
    }
  },
  {
    "ph": "f", "id": 126603, "pid": 76337, "tid": -914061504, "ts": 1716454223385849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385854, "dur": 0,
    "args": {
      "External id": 126605,
      "cbid": 251, "correlation": 126605
    }
  },
  {
    "ph": "f", "id": 126605, "pid": 76337, "tid": -914061504, "ts": 1716454223385854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223407550, "dur": 362,
    "args": {
      "External id": 126606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126606,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 126606, "pid": 5, "tid": 7, "ts": 1716454223407550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385860, "dur": 13,
    "args": {
      "External id": 126606,
      "cbid": 211, "correlation": 126606
    }
  },
  {
    "ph": "s", "id": 126606, "pid": 76337, "tid": -914061504, "ts": 1716454223385860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223407913, "dur": 50,
    "args": {
      "External id": 126614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126614,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126614, "pid": 5, "tid": 7, "ts": 1716454223407913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385903, "dur": 9,
    "args": {
      "External id": 126614,
      "cbid": 211, "correlation": 126614
    }
  },
  {
    "ph": "s", "id": 126614, "pid": 76337, "tid": -914061504, "ts": 1716454223385903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223407964, "dur": 158,
    "args": {
      "External id": 126625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126625,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126625, "pid": 5, "tid": 7, "ts": 1716454223407964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223385981, "dur": 14,
    "args": {
      "External id": 126625,
      "cbid": 211, "correlation": 126625
    }
  },
  {
    "ph": "s", "id": 126625, "pid": 76337, "tid": -914061504, "ts": 1716454223385981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386049, "dur": 0,
    "args": {
      "External id": 126637,
      "cbid": 317, "correlation": 126637
    }
  },
  {
    "ph": "f", "id": 126637, "pid": 76337, "tid": -914061504, "ts": 1716454223386049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386050, "dur": 0,
    "args": {
      "External id": 126638,
      "cbid": 203, "correlation": 126638
    }
  },
  {
    "ph": "f", "id": 126638, "pid": 76337, "tid": -914061504, "ts": 1716454223386050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386051, "dur": 0,
    "args": {
      "External id": 126639,
      "cbid": 205, "correlation": 126639
    }
  },
  {
    "ph": "f", "id": 126639, "pid": 76337, "tid": -914061504, "ts": 1716454223386051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386075, "dur": 1,
    "args": {
      "External id": 126643,
      "cbid": 251, "correlation": 126643
    }
  },
  {
    "ph": "f", "id": 126643, "pid": 76337, "tid": -914061504, "ts": 1716454223386075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386077, "dur": 0,
    "args": {
      "External id": 126644,
      "cbid": 251, "correlation": 126644
    }
  },
  {
    "ph": "f", "id": 126644, "pid": 76337, "tid": -914061504, "ts": 1716454223386077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386078, "dur": 0,
    "args": {
      "External id": 126645,
      "cbid": 251, "correlation": 126645
    }
  },
  {
    "ph": "f", "id": 126645, "pid": 76337, "tid": -914061504, "ts": 1716454223386078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386078, "dur": 0,
    "args": {
      "External id": 126646,
      "cbid": 251, "correlation": 126646
    }
  },
  {
    "ph": "f", "id": 126646, "pid": 76337, "tid": -914061504, "ts": 1716454223386078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386079, "dur": 0,
    "args": {
      "External id": 126647,
      "cbid": 251, "correlation": 126647
    }
  },
  {
    "ph": "f", "id": 126647, "pid": 76337, "tid": -914061504, "ts": 1716454223386079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386080, "dur": 0,
    "args": {
      "External id": 126648,
      "cbid": 251, "correlation": 126648
    }
  },
  {
    "ph": "f", "id": 126648, "pid": 76337, "tid": -914061504, "ts": 1716454223386080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386081, "dur": 0,
    "args": {
      "External id": 126649,
      "cbid": 251, "correlation": 126649
    }
  },
  {
    "ph": "f", "id": 126649, "pid": 76337, "tid": -914061504, "ts": 1716454223386081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386081, "dur": 0,
    "args": {
      "External id": 126650,
      "cbid": 251, "correlation": 126650
    }
  },
  {
    "ph": "f", "id": 126650, "pid": 76337, "tid": -914061504, "ts": 1716454223386081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386083, "dur": 0,
    "args": {
      "External id": 126651,
      "cbid": 251, "correlation": 126651
    }
  },
  {
    "ph": "f", "id": 126651, "pid": 76337, "tid": -914061504, "ts": 1716454223386083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223408124, "dur": 114,
    "args": {
      "External id": 126652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126652,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 126652, "pid": 5, "tid": 7, "ts": 1716454223408124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386085, "dur": 13,
    "args": {
      "External id": 126652,
      "cbid": 211, "correlation": 126652
    }
  },
  {
    "ph": "s", "id": 126652, "pid": 76337, "tid": -914061504, "ts": 1716454223386085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223408239, "dur": 59,
    "args": {
      "External id": 126658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126658,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126658, "pid": 5, "tid": 7, "ts": 1716454223408239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386120, "dur": 9,
    "args": {
      "External id": 126658,
      "cbid": 211, "correlation": 126658
    }
  },
  {
    "ph": "s", "id": 126658, "pid": 76337, "tid": -914061504, "ts": 1716454223386120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223408299, "dur": 50,
    "args": {
      "External id": 126666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126666,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126666, "pid": 5, "tid": 7, "ts": 1716454223408299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386153, "dur": 8,
    "args": {
      "External id": 126666,
      "cbid": 211, "correlation": 126666
    }
  },
  {
    "ph": "s", "id": 126666, "pid": 76337, "tid": -914061504, "ts": 1716454223386153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223408351, "dur": 52,
    "args": {
      "External id": 126686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126686,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 126686, "pid": 5, "tid": 7, "ts": 1716454223408351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386227, "dur": 12,
    "args": {
      "External id": 126686,
      "cbid": 211, "correlation": 126686
    }
  },
  {
    "ph": "s", "id": 126686, "pid": 76337, "tid": -914061504, "ts": 1716454223386227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223408404, "dur": 5,
    "args": {
      "External id": 126698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126698,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126698, "pid": 5, "tid": 7, "ts": 1716454223408404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386250, "dur": 7,
    "args": {
      "External id": 126698,
      "cbid": 211, "correlation": 126698
    }
  },
  {
    "ph": "s", "id": 126698, "pid": 76337, "tid": -914061504, "ts": 1716454223386250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223408410, "dur": 56,
    "args": {
      "External id": 126701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126701, "pid": 5, "tid": 7, "ts": 1716454223408410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386268, "dur": 7,
    "args": {
      "External id": 126701,
      "cbid": 211, "correlation": 126701
    }
  },
  {
    "ph": "s", "id": 126701, "pid": 76337, "tid": -914061504, "ts": 1716454223386268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223408467, "dur": 36,
    "args": {
      "External id": 126710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126710,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126710, "pid": 5, "tid": 7, "ts": 1716454223408467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386308, "dur": 10,
    "args": {
      "External id": 126710,
      "cbid": 211, "correlation": 126710
    }
  },
  {
    "ph": "s", "id": 126710, "pid": 76337, "tid": -914061504, "ts": 1716454223386308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386359, "dur": 0,
    "args": {
      "External id": 126720,
      "cbid": 317, "correlation": 126720
    }
  },
  {
    "ph": "f", "id": 126720, "pid": 76337, "tid": -914061504, "ts": 1716454223386359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386360, "dur": 0,
    "args": {
      "External id": 126721,
      "cbid": 203, "correlation": 126721
    }
  },
  {
    "ph": "f", "id": 126721, "pid": 76337, "tid": -914061504, "ts": 1716454223386360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386361, "dur": 0,
    "args": {
      "External id": 126722,
      "cbid": 205, "correlation": 126722
    }
  },
  {
    "ph": "f", "id": 126722, "pid": 76337, "tid": -914061504, "ts": 1716454223386361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223408505, "dur": 41,
    "args": {
      "External id": 126726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126726, "pid": 5, "tid": 7, "ts": 1716454223408505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386375, "dur": 11,
    "args": {
      "External id": 126726,
      "cbid": 211, "correlation": 126726
    }
  },
  {
    "ph": "s", "id": 126726, "pid": 76337, "tid": -914061504, "ts": 1716454223386375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223408548, "dur": 14,
    "args": {
      "External id": 126728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126728,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126728, "pid": 5, "tid": 7, "ts": 1716454223408548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386389, "dur": 6,
    "args": {
      "External id": 126728,
      "cbid": 211, "correlation": 126728
    }
  },
  {
    "ph": "s", "id": 126728, "pid": 76337, "tid": -914061504, "ts": 1716454223386389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223408564, "dur": 4,
    "args": {
      "External id": 126730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126730, "pid": 5, "tid": 7, "ts": 1716454223408564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386400, "dur": 5,
    "args": {
      "External id": 126730,
      "cbid": 211, "correlation": 126730
    }
  },
  {
    "ph": "s", "id": 126730, "pid": 76337, "tid": -914061504, "ts": 1716454223386400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386408, "dur": 0,
    "args": {
      "External id": 126731,
      "cbid": 51, "correlation": 126731
    }
  },
  {
    "ph": "s", "id": 126731, "pid": 76337, "tid": -914061504, "ts": 1716454223386408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223408569, "dur": 699,
    "args": {
      "External id": 126732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126732,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126732, "pid": 5, "tid": 7, "ts": 1716454223408569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386409, "dur": 6,
    "args": {
      "External id": 126732,
      "cbid": 211, "correlation": 126732
    }
  },
  {
    "ph": "s", "id": 126732, "pid": 76337, "tid": -914061504, "ts": 1716454223386409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223409270, "dur": 59,
    "args": {
      "External id": 126737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126737, "pid": 5, "tid": 7, "ts": 1716454223409270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386436, "dur": 9,
    "args": {
      "External id": 126737,
      "cbid": 211, "correlation": 126737
    }
  },
  {
    "ph": "s", "id": 126737, "pid": 76337, "tid": -914061504, "ts": 1716454223386436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223409330, "dur": 4,
    "args": {
      "External id": 126745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126745,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126745, "pid": 5, "tid": 7, "ts": 1716454223409330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386480, "dur": 9,
    "args": {
      "External id": 126745,
      "cbid": 211, "correlation": 126745
    }
  },
  {
    "ph": "s", "id": 126745, "pid": 76337, "tid": -914061504, "ts": 1716454223386480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386546, "dur": 1,
    "args": {
      "External id": 126761,
      "cbid": 251, "correlation": 126761
    }
  },
  {
    "ph": "f", "id": 126761, "pid": 76337, "tid": -914061504, "ts": 1716454223386546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386551, "dur": 0,
    "args": {
      "External id": 126763,
      "cbid": 251, "correlation": 126763
    }
  },
  {
    "ph": "f", "id": 126763, "pid": 76337, "tid": -914061504, "ts": 1716454223386551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223409335, "dur": 11,
    "args": {
      "External id": 126764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126764,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 126764, "pid": 5, "tid": 7, "ts": 1716454223409335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386553, "dur": 11,
    "args": {
      "External id": 126764,
      "cbid": 211, "correlation": 126764
    }
  },
  {
    "ph": "s", "id": 126764, "pid": 76337, "tid": -914061504, "ts": 1716454223386553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223409347, "dur": 5,
    "args": {
      "External id": 126766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126766,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 126766, "pid": 5, "tid": 7, "ts": 1716454223409347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386566, "dur": 5,
    "args": {
      "External id": 126766,
      "cbid": 211, "correlation": 126766
    }
  },
  {
    "ph": "s", "id": 126766, "pid": 76337, "tid": -914061504, "ts": 1716454223386566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223409354, "dur": 53,
    "args": {
      "External id": 126776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126776, "pid": 5, "tid": 7, "ts": 1716454223409354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386622, "dur": 12,
    "args": {
      "External id": 126776,
      "cbid": 211, "correlation": 126776
    }
  },
  {
    "ph": "s", "id": 126776, "pid": 76337, "tid": -914061504, "ts": 1716454223386622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223409409, "dur": 50,
    "args": {
      "External id": 126796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126796,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 126796, "pid": 5, "tid": 7, "ts": 1716454223409409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386689, "dur": 11,
    "args": {
      "External id": 126796,
      "cbid": 211, "correlation": 126796
    }
  },
  {
    "ph": "s", "id": 126796, "pid": 76337, "tid": -914061504, "ts": 1716454223386689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223409460, "dur": 4,
    "args": {
      "External id": 126808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126808,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126808, "pid": 5, "tid": 7, "ts": 1716454223409460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386710, "dur": 6,
    "args": {
      "External id": 126808,
      "cbid": 211, "correlation": 126808
    }
  },
  {
    "ph": "s", "id": 126808, "pid": 76337, "tid": -914061504, "ts": 1716454223386710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223409465, "dur": 55,
    "args": {
      "External id": 126811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126811,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126811, "pid": 5, "tid": 7, "ts": 1716454223409465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386728, "dur": 6,
    "args": {
      "External id": 126811,
      "cbid": 211, "correlation": 126811
    }
  },
  {
    "ph": "s", "id": 126811, "pid": 76337, "tid": -914061504, "ts": 1716454223386728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223409522, "dur": 36,
    "args": {
      "External id": 126820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126820,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126820, "pid": 5, "tid": 7, "ts": 1716454223409522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386769, "dur": 9,
    "args": {
      "External id": 126820,
      "cbid": 211, "correlation": 126820
    }
  },
  {
    "ph": "s", "id": 126820, "pid": 76337, "tid": -914061504, "ts": 1716454223386769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386831, "dur": 0,
    "args": {
      "External id": 126830,
      "cbid": 317, "correlation": 126830
    }
  },
  {
    "ph": "f", "id": 126830, "pid": 76337, "tid": -914061504, "ts": 1716454223386831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386832, "dur": 0,
    "args": {
      "External id": 126831,
      "cbid": 203, "correlation": 126831
    }
  },
  {
    "ph": "f", "id": 126831, "pid": 76337, "tid": -914061504, "ts": 1716454223386832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386833, "dur": 0,
    "args": {
      "External id": 126832,
      "cbid": 205, "correlation": 126832
    }
  },
  {
    "ph": "f", "id": 126832, "pid": 76337, "tid": -914061504, "ts": 1716454223386833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223409560, "dur": 41,
    "args": {
      "External id": 126836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126836,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126836, "pid": 5, "tid": 7, "ts": 1716454223409560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386846, "dur": 13,
    "args": {
      "External id": 126836,
      "cbid": 211, "correlation": 126836
    }
  },
  {
    "ph": "s", "id": 126836, "pid": 76337, "tid": -914061504, "ts": 1716454223386846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223409602, "dur": 14,
    "args": {
      "External id": 126838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126838, "pid": 5, "tid": 7, "ts": 1716454223409602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386861, "dur": 5,
    "args": {
      "External id": 126838,
      "cbid": 211, "correlation": 126838
    }
  },
  {
    "ph": "s", "id": 126838, "pid": 76337, "tid": -914061504, "ts": 1716454223386861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223409617, "dur": 3,
    "args": {
      "External id": 126840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126840,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126840, "pid": 5, "tid": 7, "ts": 1716454223409617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386870, "dur": 5,
    "args": {
      "External id": 126840,
      "cbid": 211, "correlation": 126840
    }
  },
  {
    "ph": "s", "id": 126840, "pid": 76337, "tid": -914061504, "ts": 1716454223386870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386878, "dur": 0,
    "args": {
      "External id": 126841,
      "cbid": 51, "correlation": 126841
    }
  },
  {
    "ph": "s", "id": 126841, "pid": 76337, "tid": -914061504, "ts": 1716454223386878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223409622, "dur": 692,
    "args": {
      "External id": 126842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126842,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126842, "pid": 5, "tid": 7, "ts": 1716454223409622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386879, "dur": 5,
    "args": {
      "External id": 126842,
      "cbid": 211, "correlation": 126842
    }
  },
  {
    "ph": "s", "id": 126842, "pid": 76337, "tid": -914061504, "ts": 1716454223386879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223410315, "dur": 59,
    "args": {
      "External id": 126847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126847, "pid": 5, "tid": 7, "ts": 1716454223410315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386907, "dur": 8,
    "args": {
      "External id": 126847,
      "cbid": 211, "correlation": 126847
    }
  },
  {
    "ph": "s", "id": 126847, "pid": 76337, "tid": -914061504, "ts": 1716454223386907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223410375, "dur": 50,
    "args": {
      "External id": 126855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126855,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126855, "pid": 5, "tid": 7, "ts": 1716454223410375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386938, "dur": 9,
    "args": {
      "External id": 126855,
      "cbid": 211, "correlation": 126855
    }
  },
  {
    "ph": "s", "id": 126855, "pid": 76337, "tid": -914061504, "ts": 1716454223386938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223410427, "dur": 35,
    "args": {
      "External id": 126863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126863,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126863, "pid": 5, "tid": 7, "ts": 1716454223410427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223386968, "dur": 16,
    "args": {
      "External id": 126863,
      "cbid": 211, "correlation": 126863
    }
  },
  {
    "ph": "s", "id": 126863, "pid": 76337, "tid": -914061504, "ts": 1716454223386968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223410463, "dur": 52,
    "args": {
      "External id": 126883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126883,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 126883, "pid": 5, "tid": 7, "ts": 1716454223410463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387058, "dur": 12,
    "args": {
      "External id": 126883,
      "cbid": 211, "correlation": 126883
    }
  },
  {
    "ph": "s", "id": 126883, "pid": 76337, "tid": -914061504, "ts": 1716454223387058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223410516, "dur": 4,
    "args": {
      "External id": 126895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126895,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 126895, "pid": 5, "tid": 7, "ts": 1716454223410516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387079, "dur": 6,
    "args": {
      "External id": 126895,
      "cbid": 211, "correlation": 126895
    }
  },
  {
    "ph": "s", "id": 126895, "pid": 76337, "tid": -914061504, "ts": 1716454223387079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223410521, "dur": 55,
    "args": {
      "External id": 126898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126898, "pid": 5, "tid": 7, "ts": 1716454223410521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387097, "dur": 7,
    "args": {
      "External id": 126898,
      "cbid": 211, "correlation": 126898
    }
  },
  {
    "ph": "s", "id": 126898, "pid": 76337, "tid": -914061504, "ts": 1716454223387097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387155, "dur": 0,
    "args": {
      "External id": 126909,
      "cbid": 317, "correlation": 126909
    }
  },
  {
    "ph": "f", "id": 126909, "pid": 76337, "tid": -914061504, "ts": 1716454223387155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387155, "dur": 0,
    "args": {
      "External id": 126910,
      "cbid": 203, "correlation": 126910
    }
  },
  {
    "ph": "f", "id": 126910, "pid": 76337, "tid": -914061504, "ts": 1716454223387155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387156, "dur": 0,
    "args": {
      "External id": 126911,
      "cbid": 205, "correlation": 126911
    }
  },
  {
    "ph": "f", "id": 126911, "pid": 76337, "tid": -914061504, "ts": 1716454223387156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387178, "dur": 1,
    "args": {
      "External id": 126915,
      "cbid": 251, "correlation": 126915
    }
  },
  {
    "ph": "f", "id": 126915, "pid": 76337, "tid": -914061504, "ts": 1716454223387178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387180, "dur": 0,
    "args": {
      "External id": 126916,
      "cbid": 251, "correlation": 126916
    }
  },
  {
    "ph": "f", "id": 126916, "pid": 76337, "tid": -914061504, "ts": 1716454223387180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387181, "dur": 0,
    "args": {
      "External id": 126917,
      "cbid": 251, "correlation": 126917
    }
  },
  {
    "ph": "f", "id": 126917, "pid": 76337, "tid": -914061504, "ts": 1716454223387181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387181, "dur": 0,
    "args": {
      "External id": 126918,
      "cbid": 251, "correlation": 126918
    }
  },
  {
    "ph": "f", "id": 126918, "pid": 76337, "tid": -914061504, "ts": 1716454223387181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387182, "dur": 0,
    "args": {
      "External id": 126919,
      "cbid": 251, "correlation": 126919
    }
  },
  {
    "ph": "f", "id": 126919, "pid": 76337, "tid": -914061504, "ts": 1716454223387182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387183, "dur": 0,
    "args": {
      "External id": 126920,
      "cbid": 251, "correlation": 126920
    }
  },
  {
    "ph": "f", "id": 126920, "pid": 76337, "tid": -914061504, "ts": 1716454223387183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387183, "dur": 0,
    "args": {
      "External id": 126921,
      "cbid": 251, "correlation": 126921
    }
  },
  {
    "ph": "f", "id": 126921, "pid": 76337, "tid": -914061504, "ts": 1716454223387183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387184, "dur": 0,
    "args": {
      "External id": 126922,
      "cbid": 251, "correlation": 126922
    }
  },
  {
    "ph": "f", "id": 126922, "pid": 76337, "tid": -914061504, "ts": 1716454223387184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387185, "dur": 0,
    "args": {
      "External id": 126923,
      "cbid": 251, "correlation": 126923
    }
  },
  {
    "ph": "f", "id": 126923, "pid": 76337, "tid": -914061504, "ts": 1716454223387185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223410577, "dur": 112,
    "args": {
      "External id": 126924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126924,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 126924, "pid": 5, "tid": 7, "ts": 1716454223410577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387187, "dur": 13,
    "args": {
      "External id": 126924,
      "cbid": 211, "correlation": 126924
    }
  },
  {
    "ph": "s", "id": 126924, "pid": 76337, "tid": -914061504, "ts": 1716454223387187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223410691, "dur": 60,
    "args": {
      "External id": 126930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126930,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126930, "pid": 5, "tid": 7, "ts": 1716454223410691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387223, "dur": 9,
    "args": {
      "External id": 126930,
      "cbid": 211, "correlation": 126930
    }
  },
  {
    "ph": "s", "id": 126930, "pid": 76337, "tid": -914061504, "ts": 1716454223387223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223410751, "dur": 659,
    "args": {
      "External id": 126939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126939,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126939, "pid": 5, "tid": 7, "ts": 1716454223410751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387307, "dur": 14,
    "args": {
      "External id": 126939,
      "cbid": 211, "correlation": 126939
    }
  },
  {
    "ph": "s", "id": 126939, "pid": 76337, "tid": -914061504, "ts": 1716454223387307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223411412, "dur": 180,
    "args": {
      "External id": 126961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126961,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 126961, "pid": 5, "tid": 7, "ts": 1716454223411412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387365, "dur": 10,
    "args": {
      "External id": 126961,
      "cbid": 211, "correlation": 126961
    }
  },
  {
    "ph": "s", "id": 126961, "pid": 76337, "tid": -914061504, "ts": 1716454223387365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387451, "dur": 1,
    "args": {
      "External id": 126972,
      "cbid": 251, "correlation": 126972
    }
  },
  {
    "ph": "f", "id": 126972, "pid": 76337, "tid": -914061504, "ts": 1716454223387451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223411593, "dur": 192,
    "args": {
      "External id": 126973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126973,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126973, "pid": 5, "tid": 7, "ts": 1716454223411593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387456, "dur": 13,
    "args": {
      "External id": 126973,
      "cbid": 211, "correlation": 126973
    }
  },
  {
    "ph": "s", "id": 126973, "pid": 76337, "tid": -914061504, "ts": 1716454223387456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387524, "dur": 1,
    "args": {
      "External id": 126984,
      "cbid": 251, "correlation": 126984
    }
  },
  {
    "ph": "f", "id": 126984, "pid": 76337, "tid": -914061504, "ts": 1716454223387524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223411787, "dur": 189,
    "args": {
      "External id": 126985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126985,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126985, "pid": 5, "tid": 7, "ts": 1716454223411787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387528, "dur": 12,
    "args": {
      "External id": 126985,
      "cbid": 211, "correlation": 126985
    }
  },
  {
    "ph": "s", "id": 126985, "pid": 76337, "tid": -914061504, "ts": 1716454223387528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387590, "dur": 1,
    "args": {
      "External id": 126996,
      "cbid": 251, "correlation": 126996
    }
  },
  {
    "ph": "f", "id": 126996, "pid": 76337, "tid": -914061504, "ts": 1716454223387590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223411977, "dur": 185,
    "args": {
      "External id": 126997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 126997,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 126997, "pid": 5, "tid": 7, "ts": 1716454223411977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387594, "dur": 12,
    "args": {
      "External id": 126997,
      "cbid": 211, "correlation": 126997
    }
  },
  {
    "ph": "s", "id": 126997, "pid": 76337, "tid": -914061504, "ts": 1716454223387594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223412164, "dur": 18627,
    "args": {
      "External id": 127018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127018,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127018, "pid": 5, "tid": 7, "ts": 1716454223412164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387674, "dur": 12,
    "args": {
      "External id": 127018,
      "cbid": 211, "correlation": 127018
    }
  },
  {
    "ph": "s", "id": 127018, "pid": 76337, "tid": -914061504, "ts": 1716454223387674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387771, "dur": 1,
    "args": {
      "External id": 127036,
      "cbid": 251, "correlation": 127036
    }
  },
  {
    "ph": "f", "id": 127036, "pid": 76337, "tid": -914061504, "ts": 1716454223387771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223430792, "dur": 199,
    "args": {
      "External id": 127038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127038,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127038, "pid": 5, "tid": 7, "ts": 1716454223430792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387777, "dur": 13,
    "args": {
      "External id": 127038,
      "cbid": 211, "correlation": 127038
    }
  },
  {
    "ph": "s", "id": 127038, "pid": 76337, "tid": -914061504, "ts": 1716454223387777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223430993, "dur": 67,
    "args": {
      "External id": 127046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127046,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127046, "pid": 5, "tid": 7, "ts": 1716454223430993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387848, "dur": 13,
    "args": {
      "External id": 127046,
      "cbid": 211, "correlation": 127046
    }
  },
  {
    "ph": "s", "id": 127046, "pid": 76337, "tid": -914061504, "ts": 1716454223387848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223431061, "dur": 96,
    "args": {
      "External id": 127054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127054,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127054, "pid": 5, "tid": 7, "ts": 1716454223431061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387888, "dur": 8,
    "args": {
      "External id": 127054,
      "cbid": 211, "correlation": 127054
    }
  },
  {
    "ph": "s", "id": 127054, "pid": 76337, "tid": -914061504, "ts": 1716454223387888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223431159, "dur": 54,
    "args": {
      "External id": 127065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127065, "pid": 5, "tid": 7, "ts": 1716454223431159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387958, "dur": 13,
    "args": {
      "External id": 127065,
      "cbid": 211, "correlation": 127065
    }
  },
  {
    "ph": "s", "id": 127065, "pid": 76337, "tid": -914061504, "ts": 1716454223387958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223431215, "dur": 91,
    "args": {
      "External id": 127087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127087,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127087, "pid": 5, "tid": 7, "ts": 1716454223431215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223387998, "dur": 9,
    "args": {
      "External id": 127087,
      "cbid": 211, "correlation": 127087
    }
  },
  {
    "ph": "s", "id": 127087, "pid": 76337, "tid": -914061504, "ts": 1716454223387998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388084, "dur": 1,
    "args": {
      "External id": 127098,
      "cbid": 251, "correlation": 127098
    }
  },
  {
    "ph": "f", "id": 127098, "pid": 76337, "tid": -914061504, "ts": 1716454223388084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223431308, "dur": 103,
    "args": {
      "External id": 127099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127099,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127099, "pid": 5, "tid": 7, "ts": 1716454223431308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388089, "dur": 13,
    "args": {
      "External id": 127099,
      "cbid": 211, "correlation": 127099
    }
  },
  {
    "ph": "s", "id": 127099, "pid": 76337, "tid": -914061504, "ts": 1716454223388089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388161, "dur": 1,
    "args": {
      "External id": 127110,
      "cbid": 251, "correlation": 127110
    }
  },
  {
    "ph": "f", "id": 127110, "pid": 76337, "tid": -914061504, "ts": 1716454223388161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388165, "dur": 0,
    "args": {
      "External id": 127111,
      "cbid": 251, "correlation": 127111
    }
  },
  {
    "ph": "f", "id": 127111, "pid": 76337, "tid": -914061504, "ts": 1716454223388165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223431412, "dur": 10,
    "args": {
      "External id": 127112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127112,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 127112, "pid": 5, "tid": 7, "ts": 1716454223431412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388166, "dur": 12,
    "args": {
      "External id": 127112,
      "cbid": 211, "correlation": 127112
    }
  },
  {
    "ph": "s", "id": 127112, "pid": 76337, "tid": -914061504, "ts": 1716454223388166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223431423, "dur": 5,
    "args": {
      "External id": 127114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127114,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 127114, "pid": 5, "tid": 7, "ts": 1716454223431423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388180, "dur": 6,
    "args": {
      "External id": 127114,
      "cbid": 211, "correlation": 127114
    }
  },
  {
    "ph": "s", "id": 127114, "pid": 76337, "tid": -914061504, "ts": 1716454223388180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388241, "dur": 1,
    "args": {
      "External id": 127125,
      "cbid": 251, "correlation": 127125
    }
  },
  {
    "ph": "f", "id": 127125, "pid": 76337, "tid": -914061504, "ts": 1716454223388241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388244, "dur": 0,
    "args": {
      "External id": 127126,
      "cbid": 251, "correlation": 127126
    }
  },
  {
    "ph": "f", "id": 127126, "pid": 76337, "tid": -914061504, "ts": 1716454223388244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223431429, "dur": 6,
    "args": {
      "External id": 127127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127127,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 127127, "pid": 5, "tid": 7, "ts": 1716454223431429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388245, "dur": 12,
    "args": {
      "External id": 127127,
      "cbid": 211, "correlation": 127127
    }
  },
  {
    "ph": "s", "id": 127127, "pid": 76337, "tid": -914061504, "ts": 1716454223388245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223431437, "dur": 3,
    "args": {
      "External id": 127129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127129,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 127129, "pid": 5, "tid": 7, "ts": 1716454223431437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388259, "dur": 5,
    "args": {
      "External id": 127129,
      "cbid": 211, "correlation": 127129
    }
  },
  {
    "ph": "s", "id": 127129, "pid": 76337, "tid": -914061504, "ts": 1716454223388259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223431442, "dur": 152,
    "args": {
      "External id": 127150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127150,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127150, "pid": 5, "tid": 7, "ts": 1716454223431442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388332, "dur": 12,
    "args": {
      "External id": 127150,
      "cbid": 211, "correlation": 127150
    }
  },
  {
    "ph": "s", "id": 127150, "pid": 76337, "tid": -914061504, "ts": 1716454223388332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388428, "dur": 1,
    "args": {
      "External id": 127168,
      "cbid": 251, "correlation": 127168
    }
  },
  {
    "ph": "f", "id": 127168, "pid": 76337, "tid": -914061504, "ts": 1716454223388428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223431595, "dur": 107,
    "args": {
      "External id": 127170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127170,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127170, "pid": 5, "tid": 7, "ts": 1716454223431595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388434, "dur": 13,
    "args": {
      "External id": 127170,
      "cbid": 211, "correlation": 127170
    }
  },
  {
    "ph": "s", "id": 127170, "pid": 76337, "tid": -914061504, "ts": 1716454223388434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223431704, "dur": 35,
    "args": {
      "External id": 127178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127178,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127178, "pid": 5, "tid": 7, "ts": 1716454223431704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388502, "dur": 13,
    "args": {
      "External id": 127178,
      "cbid": 211, "correlation": 127178
    }
  },
  {
    "ph": "s", "id": 127178, "pid": 76337, "tid": -914061504, "ts": 1716454223388502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223431740, "dur": 66,
    "args": {
      "External id": 127186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127186, "pid": 5, "tid": 7, "ts": 1716454223431740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388543, "dur": 9,
    "args": {
      "External id": 127186,
      "cbid": 211, "correlation": 127186
    }
  },
  {
    "ph": "s", "id": 127186, "pid": 76337, "tid": -914061504, "ts": 1716454223388543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223431808, "dur": 91,
    "args": {
      "External id": 127208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127208,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127208, "pid": 5, "tid": 7, "ts": 1716454223431808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388593, "dur": 10,
    "args": {
      "External id": 127208,
      "cbid": 211, "correlation": 127208
    }
  },
  {
    "ph": "s", "id": 127208, "pid": 76337, "tid": -914061504, "ts": 1716454223388593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388679, "dur": 1,
    "args": {
      "External id": 127224,
      "cbid": 251, "correlation": 127224
    }
  },
  {
    "ph": "f", "id": 127224, "pid": 76337, "tid": -914061504, "ts": 1716454223388679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223431901, "dur": 572,
    "args": {
      "External id": 127226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127226,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127226, "pid": 5, "tid": 7, "ts": 1716454223431901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388685, "dur": 12,
    "args": {
      "External id": 127226,
      "cbid": 211, "correlation": 127226
    }
  },
  {
    "ph": "s", "id": 127226, "pid": 76337, "tid": -914061504, "ts": 1716454223388685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223432474, "dur": 244,
    "args": {
      "External id": 127234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127234, "pid": 5, "tid": 7, "ts": 1716454223432474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388749, "dur": 13,
    "args": {
      "External id": 127234,
      "cbid": 211, "correlation": 127234
    }
  },
  {
    "ph": "s", "id": 127234, "pid": 76337, "tid": -914061504, "ts": 1716454223388749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223432719, "dur": 250,
    "args": {
      "External id": 127242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127242,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127242, "pid": 5, "tid": 7, "ts": 1716454223432719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388780, "dur": 8,
    "args": {
      "External id": 127242,
      "cbid": 211, "correlation": 127242
    }
  },
  {
    "ph": "s", "id": 127242, "pid": 76337, "tid": -914061504, "ts": 1716454223388780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388862, "dur": 1,
    "args": {
      "External id": 127258,
      "cbid": 251, "correlation": 127258
    }
  },
  {
    "ph": "f", "id": 127258, "pid": 76337, "tid": -914061504, "ts": 1716454223388862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388867, "dur": 0,
    "args": {
      "External id": 127260,
      "cbid": 251, "correlation": 127260
    }
  },
  {
    "ph": "f", "id": 127260, "pid": 76337, "tid": -914061504, "ts": 1716454223388867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223432971, "dur": 359,
    "args": {
      "External id": 127261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127261,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 127261, "pid": 5, "tid": 7, "ts": 1716454223432971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388870, "dur": 12,
    "args": {
      "External id": 127261,
      "cbid": 211, "correlation": 127261
    }
  },
  {
    "ph": "s", "id": 127261, "pid": 76337, "tid": -914061504, "ts": 1716454223388870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223433331, "dur": 50,
    "args": {
      "External id": 127269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127269,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127269, "pid": 5, "tid": 7, "ts": 1716454223433331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388913, "dur": 10,
    "args": {
      "External id": 127269,
      "cbid": 211, "correlation": 127269
    }
  },
  {
    "ph": "s", "id": 127269, "pid": 76337, "tid": -914061504, "ts": 1716454223388913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223433382, "dur": 157,
    "args": {
      "External id": 127280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127280, "pid": 5, "tid": 7, "ts": 1716454223433382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223388987, "dur": 13,
    "args": {
      "External id": 127280,
      "cbid": 211, "correlation": 127280
    }
  },
  {
    "ph": "s", "id": 127280, "pid": 76337, "tid": -914061504, "ts": 1716454223388987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389052, "dur": 0,
    "args": {
      "External id": 127292,
      "cbid": 317, "correlation": 127292
    }
  },
  {
    "ph": "f", "id": 127292, "pid": 76337, "tid": -914061504, "ts": 1716454223389052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389053, "dur": 0,
    "args": {
      "External id": 127293,
      "cbid": 203, "correlation": 127293
    }
  },
  {
    "ph": "f", "id": 127293, "pid": 76337, "tid": -914061504, "ts": 1716454223389053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389054, "dur": 0,
    "args": {
      "External id": 127294,
      "cbid": 205, "correlation": 127294
    }
  },
  {
    "ph": "f", "id": 127294, "pid": 76337, "tid": -914061504, "ts": 1716454223389054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389076, "dur": 1,
    "args": {
      "External id": 127298,
      "cbid": 251, "correlation": 127298
    }
  },
  {
    "ph": "f", "id": 127298, "pid": 76337, "tid": -914061504, "ts": 1716454223389076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389078, "dur": 0,
    "args": {
      "External id": 127299,
      "cbid": 251, "correlation": 127299
    }
  },
  {
    "ph": "f", "id": 127299, "pid": 76337, "tid": -914061504, "ts": 1716454223389078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389079, "dur": 0,
    "args": {
      "External id": 127300,
      "cbid": 251, "correlation": 127300
    }
  },
  {
    "ph": "f", "id": 127300, "pid": 76337, "tid": -914061504, "ts": 1716454223389079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389080, "dur": 0,
    "args": {
      "External id": 127301,
      "cbid": 251, "correlation": 127301
    }
  },
  {
    "ph": "f", "id": 127301, "pid": 76337, "tid": -914061504, "ts": 1716454223389080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389080, "dur": 0,
    "args": {
      "External id": 127302,
      "cbid": 251, "correlation": 127302
    }
  },
  {
    "ph": "f", "id": 127302, "pid": 76337, "tid": -914061504, "ts": 1716454223389080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389081, "dur": 0,
    "args": {
      "External id": 127303,
      "cbid": 251, "correlation": 127303
    }
  },
  {
    "ph": "f", "id": 127303, "pid": 76337, "tid": -914061504, "ts": 1716454223389081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389082, "dur": 0,
    "args": {
      "External id": 127304,
      "cbid": 251, "correlation": 127304
    }
  },
  {
    "ph": "f", "id": 127304, "pid": 76337, "tid": -914061504, "ts": 1716454223389082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389083, "dur": 0,
    "args": {
      "External id": 127305,
      "cbid": 251, "correlation": 127305
    }
  },
  {
    "ph": "f", "id": 127305, "pid": 76337, "tid": -914061504, "ts": 1716454223389083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389084, "dur": 0,
    "args": {
      "External id": 127306,
      "cbid": 251, "correlation": 127306
    }
  },
  {
    "ph": "f", "id": 127306, "pid": 76337, "tid": -914061504, "ts": 1716454223389084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223433540, "dur": 115,
    "args": {
      "External id": 127307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127307,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127307, "pid": 5, "tid": 7, "ts": 1716454223433540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389086, "dur": 12,
    "args": {
      "External id": 127307,
      "cbid": 211, "correlation": 127307
    }
  },
  {
    "ph": "s", "id": 127307, "pid": 76337, "tid": -914061504, "ts": 1716454223389086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223433657, "dur": 59,
    "args": {
      "External id": 127313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127313, "pid": 5, "tid": 7, "ts": 1716454223433657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389121, "dur": 9,
    "args": {
      "External id": 127313,
      "cbid": 211, "correlation": 127313
    }
  },
  {
    "ph": "s", "id": 127313, "pid": 76337, "tid": -914061504, "ts": 1716454223389121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223433717, "dur": 50,
    "args": {
      "External id": 127321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127321,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127321, "pid": 5, "tid": 7, "ts": 1716454223433717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389152, "dur": 9,
    "args": {
      "External id": 127321,
      "cbid": 211, "correlation": 127321
    }
  },
  {
    "ph": "s", "id": 127321, "pid": 76337, "tid": -914061504, "ts": 1716454223389152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389224, "dur": 0,
    "args": {
      "External id": 127331,
      "cbid": 317, "correlation": 127331
    }
  },
  {
    "ph": "f", "id": 127331, "pid": 76337, "tid": -914061504, "ts": 1716454223389224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389225, "dur": 0,
    "args": {
      "External id": 127332,
      "cbid": 203, "correlation": 127332
    }
  },
  {
    "ph": "f", "id": 127332, "pid": 76337, "tid": -914061504, "ts": 1716454223389225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389226, "dur": 0,
    "args": {
      "External id": 127333,
      "cbid": 205, "correlation": 127333
    }
  },
  {
    "ph": "f", "id": 127333, "pid": 76337, "tid": -914061504, "ts": 1716454223389226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223433768, "dur": 40,
    "args": {
      "External id": 127337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127337, "pid": 5, "tid": 7, "ts": 1716454223433768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389243, "dur": 13,
    "args": {
      "External id": 127337,
      "cbid": 211, "correlation": 127337
    }
  },
  {
    "ph": "s", "id": 127337, "pid": 76337, "tid": -914061504, "ts": 1716454223389243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223433810, "dur": 14,
    "args": {
      "External id": 127339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127339,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127339, "pid": 5, "tid": 7, "ts": 1716454223433810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389258, "dur": 5,
    "args": {
      "External id": 127339,
      "cbid": 211, "correlation": 127339
    }
  },
  {
    "ph": "s", "id": 127339, "pid": 76337, "tid": -914061504, "ts": 1716454223389258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223433827, "dur": 1,
    "args": {
      "External id": 127341,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 127341,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 127341, "pid": 5, "tid": 7, "ts": 1716454223433827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389277, "dur": 17,
    "args": {
      "External id": 127341,
      "cbid": 51, "correlation": 127341
    }
  },
  {
    "ph": "s", "id": 127341, "pid": 76337, "tid": -914061504, "ts": 1716454223389277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223433830, "dur": 360,
    "args": {
      "External id": 127342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127342,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127342, "pid": 5, "tid": 7, "ts": 1716454223433830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389296, "dur": 10,
    "args": {
      "External id": 127342,
      "cbid": 211, "correlation": 127342
    }
  },
  {
    "ph": "s", "id": 127342, "pid": 76337, "tid": -914061504, "ts": 1716454223389296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434192, "dur": 14,
    "args": {
      "External id": 127344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127344, "pid": 5, "tid": 7, "ts": 1716454223434192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389314, "dur": 7,
    "args": {
      "External id": 127344,
      "cbid": 211, "correlation": 127344
    }
  },
  {
    "ph": "s", "id": 127344, "pid": 76337, "tid": -914061504, "ts": 1716454223389314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223434206, "dur": 15,
    "args": {
      "External id": 127350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127350, "pid": 5, "tid": 7, "ts": 1716454223434206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389343, "dur": 9,
    "args": {
      "External id": 127350,
      "cbid": 211, "correlation": 127350
    }
  },
  {
    "ph": "s", "id": 127350, "pid": 76337, "tid": -914061504, "ts": 1716454223389343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434223, "dur": 18,
    "args": {
      "External id": 127370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127370,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 127370, "pid": 5, "tid": 7, "ts": 1716454223434223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389435, "dur": 12,
    "args": {
      "External id": 127370,
      "cbid": 211, "correlation": 127370
    }
  },
  {
    "ph": "s", "id": 127370, "pid": 76337, "tid": -914061504, "ts": 1716454223389435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223434243, "dur": 4,
    "args": {
      "External id": 127382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127382,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 127382, "pid": 5, "tid": 7, "ts": 1716454223434243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389459, "dur": 6,
    "args": {
      "External id": 127382,
      "cbid": 211, "correlation": 127382
    }
  },
  {
    "ph": "s", "id": 127382, "pid": 76337, "tid": -914061504, "ts": 1716454223389459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223434249, "dur": 19,
    "args": {
      "External id": 127385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127385,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127385, "pid": 5, "tid": 7, "ts": 1716454223434249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389477, "dur": 6,
    "args": {
      "External id": 127385,
      "cbid": 211, "correlation": 127385
    }
  },
  {
    "ph": "s", "id": 127385, "pid": 76337, "tid": -914061504, "ts": 1716454223389477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223434269, "dur": 11,
    "args": {
      "External id": 127394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127394,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127394, "pid": 5, "tid": 7, "ts": 1716454223434269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389516, "dur": 9,
    "args": {
      "External id": 127394,
      "cbid": 211, "correlation": 127394
    }
  },
  {
    "ph": "s", "id": 127394, "pid": 76337, "tid": -914061504, "ts": 1716454223389516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389571, "dur": 0,
    "args": {
      "External id": 127404,
      "cbid": 317, "correlation": 127404
    }
  },
  {
    "ph": "f", "id": 127404, "pid": 76337, "tid": -914061504, "ts": 1716454223389571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389572, "dur": 0,
    "args": {
      "External id": 127405,
      "cbid": 203, "correlation": 127405
    }
  },
  {
    "ph": "f", "id": 127405, "pid": 76337, "tid": -914061504, "ts": 1716454223389572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389573, "dur": 0,
    "args": {
      "External id": 127406,
      "cbid": 205, "correlation": 127406
    }
  },
  {
    "ph": "f", "id": 127406, "pid": 76337, "tid": -914061504, "ts": 1716454223389573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434281, "dur": 11,
    "args": {
      "External id": 127410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127410, "pid": 5, "tid": 7, "ts": 1716454223434281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389589, "dur": 11,
    "args": {
      "External id": 127410,
      "cbid": 211, "correlation": 127410
    }
  },
  {
    "ph": "s", "id": 127410, "pid": 76337, "tid": -914061504, "ts": 1716454223389589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434293, "dur": 24,
    "args": {
      "External id": 127412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127412, "pid": 5, "tid": 7, "ts": 1716454223434293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389603, "dur": 6,
    "args": {
      "External id": 127412,
      "cbid": 211, "correlation": 127412
    }
  },
  {
    "ph": "s", "id": 127412, "pid": 76337, "tid": -914061504, "ts": 1716454223389603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223434319, "dur": 4,
    "args": {
      "External id": 127414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127414,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 127414, "pid": 5, "tid": 7, "ts": 1716454223434319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389614, "dur": 5,
    "args": {
      "External id": 127414,
      "cbid": 211, "correlation": 127414
    }
  },
  {
    "ph": "s", "id": 127414, "pid": 76337, "tid": -914061504, "ts": 1716454223389614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389623, "dur": 0,
    "args": {
      "External id": 127415,
      "cbid": 51, "correlation": 127415
    }
  },
  {
    "ph": "s", "id": 127415, "pid": 76337, "tid": -914061504, "ts": 1716454223389623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223434324, "dur": 355,
    "args": {
      "External id": 127416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127416,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127416, "pid": 5, "tid": 7, "ts": 1716454223434324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389624, "dur": 7,
    "args": {
      "External id": 127416,
      "cbid": 211, "correlation": 127416
    }
  },
  {
    "ph": "s", "id": 127416, "pid": 76337, "tid": -914061504, "ts": 1716454223389624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434680, "dur": 20,
    "args": {
      "External id": 127417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127417,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127417, "pid": 5, "tid": 7, "ts": 1716454223434680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389634, "dur": 5,
    "args": {
      "External id": 127417,
      "cbid": 211, "correlation": 127417
    }
  },
  {
    "ph": "s", "id": 127417, "pid": 76337, "tid": -914061504, "ts": 1716454223389634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223434701, "dur": 33,
    "args": {
      "External id": 127423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127423, "pid": 5, "tid": 7, "ts": 1716454223434701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389662, "dur": 8,
    "args": {
      "External id": 127423,
      "cbid": 211, "correlation": 127423
    }
  },
  {
    "ph": "s", "id": 127423, "pid": 76337, "tid": -914061504, "ts": 1716454223389662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223434735, "dur": 3,
    "args": {
      "External id": 127431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127431,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 127431, "pid": 5, "tid": 7, "ts": 1716454223434735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389704, "dur": 9,
    "args": {
      "External id": 127431,
      "cbid": 211, "correlation": 127431
    }
  },
  {
    "ph": "s", "id": 127431, "pid": 76337, "tid": -914061504, "ts": 1716454223389704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389768, "dur": 1,
    "args": {
      "External id": 127447,
      "cbid": 251, "correlation": 127447
    }
  },
  {
    "ph": "f", "id": 127447, "pid": 76337, "tid": -914061504, "ts": 1716454223389768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389773, "dur": 0,
    "args": {
      "External id": 127449,
      "cbid": 251, "correlation": 127449
    }
  },
  {
    "ph": "f", "id": 127449, "pid": 76337, "tid": -914061504, "ts": 1716454223389773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223434740, "dur": 12,
    "args": {
      "External id": 127450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127450,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 127450, "pid": 5, "tid": 7, "ts": 1716454223434740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389775, "dur": 11,
    "args": {
      "External id": 127450,
      "cbid": 211, "correlation": 127450
    }
  },
  {
    "ph": "s", "id": 127450, "pid": 76337, "tid": -914061504, "ts": 1716454223389775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223434753, "dur": 5,
    "args": {
      "External id": 127452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127452,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 127452, "pid": 5, "tid": 7, "ts": 1716454223434753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389788, "dur": 5,
    "args": {
      "External id": 127452,
      "cbid": 211, "correlation": 127452
    }
  },
  {
    "ph": "s", "id": 127452, "pid": 76337, "tid": -914061504, "ts": 1716454223389788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223434759, "dur": 29,
    "args": {
      "External id": 127462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127462, "pid": 5, "tid": 7, "ts": 1716454223434759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389847, "dur": 13,
    "args": {
      "External id": 127462,
      "cbid": 211, "correlation": 127462
    }
  },
  {
    "ph": "s", "id": 127462, "pid": 76337, "tid": -914061504, "ts": 1716454223389847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434790, "dur": 32,
    "args": {
      "External id": 127482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127482,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 127482, "pid": 5, "tid": 7, "ts": 1716454223434790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389914, "dur": 11,
    "args": {
      "External id": 127482,
      "cbid": 211, "correlation": 127482
    }
  },
  {
    "ph": "s", "id": 127482, "pid": 76337, "tid": -914061504, "ts": 1716454223389914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223434823, "dur": 4,
    "args": {
      "External id": 127494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127494,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 127494, "pid": 5, "tid": 7, "ts": 1716454223434823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389935, "dur": 6,
    "args": {
      "External id": 127494,
      "cbid": 211, "correlation": 127494
    }
  },
  {
    "ph": "s", "id": 127494, "pid": 76337, "tid": -914061504, "ts": 1716454223389935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223434828, "dur": 30,
    "args": {
      "External id": 127497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127497, "pid": 5, "tid": 7, "ts": 1716454223434828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223389954, "dur": 6,
    "args": {
      "External id": 127497,
      "cbid": 211, "correlation": 127497
    }
  },
  {
    "ph": "s", "id": 127497, "pid": 76337, "tid": -914061504, "ts": 1716454223389954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223434859, "dur": 20,
    "args": {
      "External id": 127506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127506,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127506, "pid": 5, "tid": 7, "ts": 1716454223434859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390005, "dur": 10,
    "args": {
      "External id": 127506,
      "cbid": 211, "correlation": 127506
    }
  },
  {
    "ph": "s", "id": 127506, "pid": 76337, "tid": -914061504, "ts": 1716454223390005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390069, "dur": 0,
    "args": {
      "External id": 127516,
      "cbid": 317, "correlation": 127516
    }
  },
  {
    "ph": "f", "id": 127516, "pid": 76337, "tid": -914061504, "ts": 1716454223390069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390070, "dur": 0,
    "args": {
      "External id": 127517,
      "cbid": 203, "correlation": 127517
    }
  },
  {
    "ph": "f", "id": 127517, "pid": 76337, "tid": -914061504, "ts": 1716454223390070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390071, "dur": 0,
    "args": {
      "External id": 127518,
      "cbid": 205, "correlation": 127518
    }
  },
  {
    "ph": "f", "id": 127518, "pid": 76337, "tid": -914061504, "ts": 1716454223390071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434880, "dur": 22,
    "args": {
      "External id": 127522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127522,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127522, "pid": 5, "tid": 7, "ts": 1716454223434880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390088, "dur": 13,
    "args": {
      "External id": 127522,
      "cbid": 211, "correlation": 127522
    }
  },
  {
    "ph": "s", "id": 127522, "pid": 76337, "tid": -914061504, "ts": 1716454223390088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223434904, "dur": 43,
    "args": {
      "External id": 127524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127524, "pid": 5, "tid": 7, "ts": 1716454223434904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390104, "dur": 5,
    "args": {
      "External id": 127524,
      "cbid": 211, "correlation": 127524
    }
  },
  {
    "ph": "s", "id": 127524, "pid": 76337, "tid": -914061504, "ts": 1716454223390104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223434949, "dur": 649,
    "args": {
      "External id": 127526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127526,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127526, "pid": 5, "tid": 7, "ts": 1716454223434949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390118, "dur": 9,
    "args": {
      "External id": 127526,
      "cbid": 211, "correlation": 127526
    }
  },
  {
    "ph": "s", "id": 127526, "pid": 76337, "tid": -914061504, "ts": 1716454223390118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223435599, "dur": 22,
    "args": {
      "External id": 127528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127528,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127528, "pid": 5, "tid": 7, "ts": 1716454223435599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390131, "dur": 5,
    "args": {
      "External id": 127528,
      "cbid": 211, "correlation": 127528
    }
  },
  {
    "ph": "s", "id": 127528, "pid": 76337, "tid": -914061504, "ts": 1716454223390131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223435623, "dur": 33,
    "args": {
      "External id": 127534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127534, "pid": 5, "tid": 7, "ts": 1716454223435623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390159, "dur": 8,
    "args": {
      "External id": 127534,
      "cbid": 211, "correlation": 127534
    }
  },
  {
    "ph": "s", "id": 127534, "pid": 76337, "tid": -914061504, "ts": 1716454223390159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390216, "dur": 0,
    "args": {
      "External id": 127544,
      "cbid": 317, "correlation": 127544
    }
  },
  {
    "ph": "f", "id": 127544, "pid": 76337, "tid": -914061504, "ts": 1716454223390216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390217, "dur": 0,
    "args": {
      "External id": 127545,
      "cbid": 203, "correlation": 127545
    }
  },
  {
    "ph": "f", "id": 127545, "pid": 76337, "tid": -914061504, "ts": 1716454223390217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390218, "dur": 0,
    "args": {
      "External id": 127546,
      "cbid": 205, "correlation": 127546
    }
  },
  {
    "ph": "f", "id": 127546, "pid": 76337, "tid": -914061504, "ts": 1716454223390218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390239, "dur": 1,
    "args": {
      "External id": 127550,
      "cbid": 251, "correlation": 127550
    }
  },
  {
    "ph": "f", "id": 127550, "pid": 76337, "tid": -914061504, "ts": 1716454223390239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390241, "dur": 0,
    "args": {
      "External id": 127551,
      "cbid": 251, "correlation": 127551
    }
  },
  {
    "ph": "f", "id": 127551, "pid": 76337, "tid": -914061504, "ts": 1716454223390241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390242, "dur": 0,
    "args": {
      "External id": 127552,
      "cbid": 251, "correlation": 127552
    }
  },
  {
    "ph": "f", "id": 127552, "pid": 76337, "tid": -914061504, "ts": 1716454223390242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390243, "dur": 0,
    "args": {
      "External id": 127553,
      "cbid": 251, "correlation": 127553
    }
  },
  {
    "ph": "f", "id": 127553, "pid": 76337, "tid": -914061504, "ts": 1716454223390243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390243, "dur": 0,
    "args": {
      "External id": 127554,
      "cbid": 251, "correlation": 127554
    }
  },
  {
    "ph": "f", "id": 127554, "pid": 76337, "tid": -914061504, "ts": 1716454223390243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390244, "dur": 0,
    "args": {
      "External id": 127555,
      "cbid": 251, "correlation": 127555
    }
  },
  {
    "ph": "f", "id": 127555, "pid": 76337, "tid": -914061504, "ts": 1716454223390244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390245, "dur": 0,
    "args": {
      "External id": 127556,
      "cbid": 251, "correlation": 127556
    }
  },
  {
    "ph": "f", "id": 127556, "pid": 76337, "tid": -914061504, "ts": 1716454223390245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390245, "dur": 0,
    "args": {
      "External id": 127557,
      "cbid": 251, "correlation": 127557
    }
  },
  {
    "ph": "f", "id": 127557, "pid": 76337, "tid": -914061504, "ts": 1716454223390245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390247, "dur": 0,
    "args": {
      "External id": 127558,
      "cbid": 251, "correlation": 127558
    }
  },
  {
    "ph": "f", "id": 127558, "pid": 76337, "tid": -914061504, "ts": 1716454223390247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223435658, "dur": 51,
    "args": {
      "External id": 127559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127559,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127559, "pid": 5, "tid": 7, "ts": 1716454223435658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390249, "dur": 12,
    "args": {
      "External id": 127559,
      "cbid": 211, "correlation": 127559
    }
  },
  {
    "ph": "s", "id": 127559, "pid": 76337, "tid": -914061504, "ts": 1716454223390249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223435710, "dur": 32,
    "args": {
      "External id": 127565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127565, "pid": 5, "tid": 7, "ts": 1716454223435710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390283, "dur": 9,
    "args": {
      "External id": 127565,
      "cbid": 211, "correlation": 127565
    }
  },
  {
    "ph": "s", "id": 127565, "pid": 76337, "tid": -914061504, "ts": 1716454223390283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223435743, "dur": 27,
    "args": {
      "External id": 127573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127573,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127573, "pid": 5, "tid": 7, "ts": 1716454223435743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390313, "dur": 8,
    "args": {
      "External id": 127573,
      "cbid": 211, "correlation": 127573
    }
  },
  {
    "ph": "s", "id": 127573, "pid": 76337, "tid": -914061504, "ts": 1716454223390313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223435771, "dur": 19,
    "args": {
      "External id": 127581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127581,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127581, "pid": 5, "tid": 7, "ts": 1716454223435771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390342, "dur": 9,
    "args": {
      "External id": 127581,
      "cbid": 211, "correlation": 127581
    }
  },
  {
    "ph": "s", "id": 127581, "pid": 76337, "tid": -914061504, "ts": 1716454223390342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223435792, "dur": 29,
    "args": {
      "External id": 127601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127601,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 127601, "pid": 5, "tid": 7, "ts": 1716454223435792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390424, "dur": 12,
    "args": {
      "External id": 127601,
      "cbid": 211, "correlation": 127601
    }
  },
  {
    "ph": "s", "id": 127601, "pid": 76337, "tid": -914061504, "ts": 1716454223390424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223435823, "dur": 4,
    "args": {
      "External id": 127613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127613,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 127613, "pid": 5, "tid": 7, "ts": 1716454223435823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390446, "dur": 7,
    "args": {
      "External id": 127613,
      "cbid": 211, "correlation": 127613
    }
  },
  {
    "ph": "s", "id": 127613, "pid": 76337, "tid": -914061504, "ts": 1716454223390446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223435828, "dur": 30,
    "args": {
      "External id": 127616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127616,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127616, "pid": 5, "tid": 7, "ts": 1716454223435828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390464, "dur": 7,
    "args": {
      "External id": 127616,
      "cbid": 211, "correlation": 127616
    }
  },
  {
    "ph": "s", "id": 127616, "pid": 76337, "tid": -914061504, "ts": 1716454223390464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390521, "dur": 0,
    "args": {
      "External id": 127627,
      "cbid": 317, "correlation": 127627
    }
  },
  {
    "ph": "f", "id": 127627, "pid": 76337, "tid": -914061504, "ts": 1716454223390521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390522, "dur": 0,
    "args": {
      "External id": 127628,
      "cbid": 203, "correlation": 127628
    }
  },
  {
    "ph": "f", "id": 127628, "pid": 76337, "tid": -914061504, "ts": 1716454223390522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390523, "dur": 0,
    "args": {
      "External id": 127629,
      "cbid": 205, "correlation": 127629
    }
  },
  {
    "ph": "f", "id": 127629, "pid": 76337, "tid": -914061504, "ts": 1716454223390523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223435859, "dur": 21,
    "args": {
      "External id": 127633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127633,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127633, "pid": 5, "tid": 7, "ts": 1716454223435859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390538, "dur": 12,
    "args": {
      "External id": 127633,
      "cbid": 211, "correlation": 127633
    }
  },
  {
    "ph": "s", "id": 127633, "pid": 76337, "tid": -914061504, "ts": 1716454223390538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223435882, "dur": 120,
    "args": {
      "External id": 127635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127635,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127635, "pid": 5, "tid": 7, "ts": 1716454223435882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390558, "dur": 8,
    "args": {
      "External id": 127635,
      "cbid": 211, "correlation": 127635
    }
  },
  {
    "ph": "s", "id": 127635, "pid": 76337, "tid": -914061504, "ts": 1716454223390558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223436003, "dur": 21,
    "args": {
      "External id": 127637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127637,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127637, "pid": 5, "tid": 7, "ts": 1716454223436003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390570, "dur": 5,
    "args": {
      "External id": 127637,
      "cbid": 211, "correlation": 127637
    }
  },
  {
    "ph": "s", "id": 127637, "pid": 76337, "tid": -914061504, "ts": 1716454223390570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223436026, "dur": 33,
    "args": {
      "External id": 127643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127643, "pid": 5, "tid": 7, "ts": 1716454223436026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390597, "dur": 9,
    "args": {
      "External id": 127643,
      "cbid": 211, "correlation": 127643
    }
  },
  {
    "ph": "s", "id": 127643, "pid": 76337, "tid": -914061504, "ts": 1716454223390597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223436060, "dur": 165,
    "args": {
      "External id": 127652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127652,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127652, "pid": 5, "tid": 7, "ts": 1716454223436060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390679, "dur": 14,
    "args": {
      "External id": 127652,
      "cbid": 211, "correlation": 127652
    }
  },
  {
    "ph": "s", "id": 127652, "pid": 76337, "tid": -914061504, "ts": 1716454223390679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223436226, "dur": 64,
    "args": {
      "External id": 127674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127674,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127674, "pid": 5, "tid": 7, "ts": 1716454223436226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390737, "dur": 10,
    "args": {
      "External id": 127674,
      "cbid": 211, "correlation": 127674
    }
  },
  {
    "ph": "s", "id": 127674, "pid": 76337, "tid": -914061504, "ts": 1716454223390737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390827, "dur": 1,
    "args": {
      "External id": 127685,
      "cbid": 251, "correlation": 127685
    }
  },
  {
    "ph": "f", "id": 127685, "pid": 76337, "tid": -914061504, "ts": 1716454223390827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223436291, "dur": 155,
    "args": {
      "External id": 127686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127686,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127686, "pid": 5, "tid": 7, "ts": 1716454223436291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390832, "dur": 13,
    "args": {
      "External id": 127686,
      "cbid": 211, "correlation": 127686
    }
  },
  {
    "ph": "s", "id": 127686, "pid": 76337, "tid": -914061504, "ts": 1716454223390832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390902, "dur": 1,
    "args": {
      "External id": 127697,
      "cbid": 251, "correlation": 127697
    }
  },
  {
    "ph": "f", "id": 127697, "pid": 76337, "tid": -914061504, "ts": 1716454223390902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223436448, "dur": 147,
    "args": {
      "External id": 127698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127698,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127698, "pid": 5, "tid": 7, "ts": 1716454223436448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390907, "dur": 11,
    "args": {
      "External id": 127698,
      "cbid": 211, "correlation": 127698
    }
  },
  {
    "ph": "s", "id": 127698, "pid": 76337, "tid": -914061504, "ts": 1716454223390907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390971, "dur": 1,
    "args": {
      "External id": 127709,
      "cbid": 251, "correlation": 127709
    }
  },
  {
    "ph": "f", "id": 127709, "pid": 76337, "tid": -914061504, "ts": 1716454223390971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223436596, "dur": 144,
    "args": {
      "External id": 127710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127710,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127710, "pid": 5, "tid": 7, "ts": 1716454223436596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223390982, "dur": 12,
    "args": {
      "External id": 127710,
      "cbid": 211, "correlation": 127710
    }
  },
  {
    "ph": "s", "id": 127710, "pid": 76337, "tid": -914061504, "ts": 1716454223390982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223436741, "dur": 1935,
    "args": {
      "External id": 127731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127731,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127731, "pid": 5, "tid": 7, "ts": 1716454223436741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391068, "dur": 13,
    "args": {
      "External id": 127731,
      "cbid": 211, "correlation": 127731
    }
  },
  {
    "ph": "s", "id": 127731, "pid": 76337, "tid": -914061504, "ts": 1716454223391068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391168, "dur": 1,
    "args": {
      "External id": 127749,
      "cbid": 251, "correlation": 127749
    }
  },
  {
    "ph": "f", "id": 127749, "pid": 76337, "tid": -914061504, "ts": 1716454223391168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223438677, "dur": 146,
    "args": {
      "External id": 127751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127751,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 127751, "pid": 5, "tid": 7, "ts": 1716454223438677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391174, "dur": 13,
    "args": {
      "External id": 127751,
      "cbid": 211, "correlation": 127751
    }
  },
  {
    "ph": "s", "id": 127751, "pid": 76337, "tid": -914061504, "ts": 1716454223391174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223438824, "dur": 35,
    "args": {
      "External id": 127759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127759,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127759, "pid": 5, "tid": 7, "ts": 1716454223438824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391245, "dur": 13,
    "args": {
      "External id": 127759,
      "cbid": 211, "correlation": 127759
    }
  },
  {
    "ph": "s", "id": 127759, "pid": 76337, "tid": -914061504, "ts": 1716454223391245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223438861, "dur": 51,
    "args": {
      "External id": 127767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127767,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127767, "pid": 5, "tid": 7, "ts": 1716454223438861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391285, "dur": 8,
    "args": {
      "External id": 127767,
      "cbid": 211, "correlation": 127767
    }
  },
  {
    "ph": "s", "id": 127767, "pid": 76337, "tid": -914061504, "ts": 1716454223391285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223438913, "dur": 30,
    "args": {
      "External id": 127778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127778, "pid": 5, "tid": 7, "ts": 1716454223438913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391358, "dur": 13,
    "args": {
      "External id": 127778,
      "cbid": 211, "correlation": 127778
    }
  },
  {
    "ph": "s", "id": 127778, "pid": 76337, "tid": -914061504, "ts": 1716454223391358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223438945, "dur": 34,
    "args": {
      "External id": 127800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127800,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127800, "pid": 5, "tid": 7, "ts": 1716454223438945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391389, "dur": 7,
    "args": {
      "External id": 127800,
      "cbid": 211, "correlation": 127800
    }
  },
  {
    "ph": "s", "id": 127800, "pid": 76337, "tid": -914061504, "ts": 1716454223391389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391474, "dur": 1,
    "args": {
      "External id": 127811,
      "cbid": 251, "correlation": 127811
    }
  },
  {
    "ph": "f", "id": 127811, "pid": 76337, "tid": -914061504, "ts": 1716454223391474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223438980, "dur": 90,
    "args": {
      "External id": 127812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127812,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127812, "pid": 5, "tid": 7, "ts": 1716454223438980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391479, "dur": 12,
    "args": {
      "External id": 127812,
      "cbid": 211, "correlation": 127812
    }
  },
  {
    "ph": "s", "id": 127812, "pid": 76337, "tid": -914061504, "ts": 1716454223391479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391550, "dur": 1,
    "args": {
      "External id": 127823,
      "cbid": 251, "correlation": 127823
    }
  },
  {
    "ph": "f", "id": 127823, "pid": 76337, "tid": -914061504, "ts": 1716454223391550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391554, "dur": 0,
    "args": {
      "External id": 127824,
      "cbid": 251, "correlation": 127824
    }
  },
  {
    "ph": "f", "id": 127824, "pid": 76337, "tid": -914061504, "ts": 1716454223391554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223439071, "dur": 11,
    "args": {
      "External id": 127825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127825,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 127825, "pid": 5, "tid": 7, "ts": 1716454223439071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391556, "dur": 12,
    "args": {
      "External id": 127825,
      "cbid": 211, "correlation": 127825
    }
  },
  {
    "ph": "s", "id": 127825, "pid": 76337, "tid": -914061504, "ts": 1716454223391556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223439084, "dur": 5,
    "args": {
      "External id": 127827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127827,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 127827, "pid": 5, "tid": 7, "ts": 1716454223439084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391570, "dur": 7,
    "args": {
      "External id": 127827,
      "cbid": 211, "correlation": 127827
    }
  },
  {
    "ph": "s", "id": 127827, "pid": 76337, "tid": -914061504, "ts": 1716454223391570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391629, "dur": 1,
    "args": {
      "External id": 127838,
      "cbid": 251, "correlation": 127838
    }
  },
  {
    "ph": "f", "id": 127838, "pid": 76337, "tid": -914061504, "ts": 1716454223391629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391632, "dur": 0,
    "args": {
      "External id": 127839,
      "cbid": 251, "correlation": 127839
    }
  },
  {
    "ph": "f", "id": 127839, "pid": 76337, "tid": -914061504, "ts": 1716454223391632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223439091, "dur": 7,
    "args": {
      "External id": 127840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127840,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 127840, "pid": 5, "tid": 7, "ts": 1716454223439091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391633, "dur": 11,
    "args": {
      "External id": 127840,
      "cbid": 211, "correlation": 127840
    }
  },
  {
    "ph": "s", "id": 127840, "pid": 76337, "tid": -914061504, "ts": 1716454223391633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223439099, "dur": 3,
    "args": {
      "External id": 127842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127842,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 127842, "pid": 5, "tid": 7, "ts": 1716454223439099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391646, "dur": 6,
    "args": {
      "External id": 127842,
      "cbid": 211, "correlation": 127842
    }
  },
  {
    "ph": "s", "id": 127842, "pid": 76337, "tid": -914061504, "ts": 1716454223391646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223439104, "dur": 90,
    "args": {
      "External id": 127863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127863,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 127863, "pid": 5, "tid": 7, "ts": 1716454223439104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391720, "dur": 12,
    "args": {
      "External id": 127863,
      "cbid": 211, "correlation": 127863
    }
  },
  {
    "ph": "s", "id": 127863, "pid": 76337, "tid": -914061504, "ts": 1716454223391720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391817, "dur": 1,
    "args": {
      "External id": 127881,
      "cbid": 251, "correlation": 127881
    }
  },
  {
    "ph": "f", "id": 127881, "pid": 76337, "tid": -914061504, "ts": 1716454223391817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223439195, "dur": 96,
    "args": {
      "External id": 127883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127883,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127883, "pid": 5, "tid": 7, "ts": 1716454223439195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391823, "dur": 13,
    "args": {
      "External id": 127883,
      "cbid": 211, "correlation": 127883
    }
  },
  {
    "ph": "s", "id": 127883, "pid": 76337, "tid": -914061504, "ts": 1716454223391823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223439292, "dur": 19,
    "args": {
      "External id": 127891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127891,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127891, "pid": 5, "tid": 7, "ts": 1716454223439292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391891, "dur": 13,
    "args": {
      "External id": 127891,
      "cbid": 211, "correlation": 127891
    }
  },
  {
    "ph": "s", "id": 127891, "pid": 76337, "tid": -914061504, "ts": 1716454223391891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223439313, "dur": 37,
    "args": {
      "External id": 127899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127899, "pid": 5, "tid": 7, "ts": 1716454223439313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391933, "dur": 9,
    "args": {
      "External id": 127899,
      "cbid": 211, "correlation": 127899
    }
  },
  {
    "ph": "s", "id": 127899, "pid": 76337, "tid": -914061504, "ts": 1716454223391933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223439351, "dur": 35,
    "args": {
      "External id": 127921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127921,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127921, "pid": 5, "tid": 7, "ts": 1716454223439351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223391991, "dur": 11,
    "args": {
      "External id": 127921,
      "cbid": 211, "correlation": 127921
    }
  },
  {
    "ph": "s", "id": 127921, "pid": 76337, "tid": -914061504, "ts": 1716454223391991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392081, "dur": 1,
    "args": {
      "External id": 127937,
      "cbid": 251, "correlation": 127937
    }
  },
  {
    "ph": "f", "id": 127937, "pid": 76337, "tid": -914061504, "ts": 1716454223392081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392086, "dur": 0,
    "args": {
      "External id": 127939,
      "cbid": 251, "correlation": 127939
    }
  },
  {
    "ph": "f", "id": 127939, "pid": 76337, "tid": -914061504, "ts": 1716454223392086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223439387, "dur": 537,
    "args": {
      "External id": 127940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127940,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 127940, "pid": 5, "tid": 7, "ts": 1716454223439387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392090, "dur": 13,
    "args": {
      "External id": 127940,
      "cbid": 211, "correlation": 127940
    }
  },
  {
    "ph": "s", "id": 127940, "pid": 76337, "tid": -914061504, "ts": 1716454223392090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223439926, "dur": 126,
    "args": {
      "External id": 127948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127948, "pid": 5, "tid": 7, "ts": 1716454223439926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392157, "dur": 12,
    "args": {
      "External id": 127948,
      "cbid": 211, "correlation": 127948
    }
  },
  {
    "ph": "s", "id": 127948, "pid": 76337, "tid": -914061504, "ts": 1716454223392157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223440053, "dur": 126,
    "args": {
      "External id": 127956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127956, "pid": 5, "tid": 7, "ts": 1716454223440053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392186, "dur": 8,
    "args": {
      "External id": 127956,
      "cbid": 211, "correlation": 127956
    }
  },
  {
    "ph": "s", "id": 127956, "pid": 76337, "tid": -914061504, "ts": 1716454223392186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392265, "dur": 1,
    "args": {
      "External id": 127972,
      "cbid": 251, "correlation": 127972
    }
  },
  {
    "ph": "f", "id": 127972, "pid": 76337, "tid": -914061504, "ts": 1716454223392265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223440181, "dur": 299,
    "args": {
      "External id": 127974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127974,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 127974, "pid": 5, "tid": 7, "ts": 1716454223440181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392271, "dur": 13,
    "args": {
      "External id": 127974,
      "cbid": 211, "correlation": 127974
    }
  },
  {
    "ph": "s", "id": 127974, "pid": 76337, "tid": -914061504, "ts": 1716454223392271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223440481, "dur": 28,
    "args": {
      "External id": 127982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127982,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127982, "pid": 5, "tid": 7, "ts": 1716454223440481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392314, "dur": 10,
    "args": {
      "External id": 127982,
      "cbid": 211, "correlation": 127982
    }
  },
  {
    "ph": "s", "id": 127982, "pid": 76337, "tid": -914061504, "ts": 1716454223392314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223440510, "dur": 81,
    "args": {
      "External id": 127993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 127993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 127993, "pid": 5, "tid": 7, "ts": 1716454223440510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392383, "dur": 12,
    "args": {
      "External id": 127993,
      "cbid": 211, "correlation": 127993
    }
  },
  {
    "ph": "s", "id": 127993, "pid": 76337, "tid": -914061504, "ts": 1716454223392383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392446, "dur": 0,
    "args": {
      "External id": 128005,
      "cbid": 317, "correlation": 128005
    }
  },
  {
    "ph": "f", "id": 128005, "pid": 76337, "tid": -914061504, "ts": 1716454223392446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392447, "dur": 0,
    "args": {
      "External id": 128006,
      "cbid": 203, "correlation": 128006
    }
  },
  {
    "ph": "f", "id": 128006, "pid": 76337, "tid": -914061504, "ts": 1716454223392447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392448, "dur": 0,
    "args": {
      "External id": 128007,
      "cbid": 205, "correlation": 128007
    }
  },
  {
    "ph": "f", "id": 128007, "pid": 76337, "tid": -914061504, "ts": 1716454223392448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223440592, "dur": 24,
    "args": {
      "External id": 128011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128011,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128011, "pid": 5, "tid": 7, "ts": 1716454223440592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392463, "dur": 12,
    "args": {
      "External id": 128011,
      "cbid": 211, "correlation": 128011
    }
  },
  {
    "ph": "s", "id": 128011, "pid": 76337, "tid": -914061504, "ts": 1716454223392463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223440617, "dur": 119,
    "args": {
      "External id": 128013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128013,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128013, "pid": 5, "tid": 7, "ts": 1716454223440617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392481, "dur": 7,
    "args": {
      "External id": 128013,
      "cbid": 211, "correlation": 128013
    }
  },
  {
    "ph": "s", "id": 128013, "pid": 76337, "tid": -914061504, "ts": 1716454223392481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223440737, "dur": 22,
    "args": {
      "External id": 128015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128015,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128015, "pid": 5, "tid": 7, "ts": 1716454223440737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392492, "dur": 5,
    "args": {
      "External id": 128015,
      "cbid": 211, "correlation": 128015
    }
  },
  {
    "ph": "s", "id": 128015, "pid": 76337, "tid": -914061504, "ts": 1716454223392492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223440761, "dur": 32,
    "args": {
      "External id": 128021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128021,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128021, "pid": 5, "tid": 7, "ts": 1716454223440761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392520, "dur": 8,
    "args": {
      "External id": 128021,
      "cbid": 211, "correlation": 128021
    }
  },
  {
    "ph": "s", "id": 128021, "pid": 76337, "tid": -914061504, "ts": 1716454223392520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223440795, "dur": 27,
    "args": {
      "External id": 128029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128029,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128029, "pid": 5, "tid": 7, "ts": 1716454223440795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392551, "dur": 8,
    "args": {
      "External id": 128029,
      "cbid": 211, "correlation": 128029
    }
  },
  {
    "ph": "s", "id": 128029, "pid": 76337, "tid": -914061504, "ts": 1716454223392551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223440823, "dur": 30,
    "args": {
      "External id": 128049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128049,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 128049, "pid": 5, "tid": 7, "ts": 1716454223440823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392623, "dur": 12,
    "args": {
      "External id": 128049,
      "cbid": 211, "correlation": 128049
    }
  },
  {
    "ph": "s", "id": 128049, "pid": 76337, "tid": -914061504, "ts": 1716454223392623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223440854, "dur": 5,
    "args": {
      "External id": 128061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128061,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 128061, "pid": 5, "tid": 7, "ts": 1716454223440854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392646, "dur": 7,
    "args": {
      "External id": 128061,
      "cbid": 211, "correlation": 128061
    }
  },
  {
    "ph": "s", "id": 128061, "pid": 76337, "tid": -914061504, "ts": 1716454223392646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223440860, "dur": 31,
    "args": {
      "External id": 128064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128064, "pid": 5, "tid": 7, "ts": 1716454223440860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392664, "dur": 8,
    "args": {
      "External id": 128064,
      "cbid": 211, "correlation": 128064
    }
  },
  {
    "ph": "s", "id": 128064, "pid": 76337, "tid": -914061504, "ts": 1716454223392664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223440893, "dur": 22,
    "args": {
      "External id": 128073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128073,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128073, "pid": 5, "tid": 7, "ts": 1716454223440893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392704, "dur": 9,
    "args": {
      "External id": 128073,
      "cbid": 211, "correlation": 128073
    }
  },
  {
    "ph": "s", "id": 128073, "pid": 76337, "tid": -914061504, "ts": 1716454223392704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392755, "dur": 0,
    "args": {
      "External id": 128083,
      "cbid": 317, "correlation": 128083
    }
  },
  {
    "ph": "f", "id": 128083, "pid": 76337, "tid": -914061504, "ts": 1716454223392755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392756, "dur": 0,
    "args": {
      "External id": 128084,
      "cbid": 203, "correlation": 128084
    }
  },
  {
    "ph": "f", "id": 128084, "pid": 76337, "tid": -914061504, "ts": 1716454223392756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392757, "dur": 0,
    "args": {
      "External id": 128085,
      "cbid": 205, "correlation": 128085
    }
  },
  {
    "ph": "f", "id": 128085, "pid": 76337, "tid": -914061504, "ts": 1716454223392757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223440916, "dur": 23,
    "args": {
      "External id": 128089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128089,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128089, "pid": 5, "tid": 7, "ts": 1716454223440916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392770, "dur": 11,
    "args": {
      "External id": 128089,
      "cbid": 211, "correlation": 128089
    }
  },
  {
    "ph": "s", "id": 128089, "pid": 76337, "tid": -914061504, "ts": 1716454223392770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223440940, "dur": 43,
    "args": {
      "External id": 128091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128091,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128091, "pid": 5, "tid": 7, "ts": 1716454223440940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392784, "dur": 5,
    "args": {
      "External id": 128091,
      "cbid": 211, "correlation": 128091
    }
  },
  {
    "ph": "s", "id": 128091, "pid": 76337, "tid": -914061504, "ts": 1716454223392784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223440985, "dur": 654,
    "args": {
      "External id": 128093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128093,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128093, "pid": 5, "tid": 7, "ts": 1716454223440985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392795, "dur": 6,
    "args": {
      "External id": 128093,
      "cbid": 211, "correlation": 128093
    }
  },
  {
    "ph": "s", "id": 128093, "pid": 76337, "tid": -914061504, "ts": 1716454223392795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223441640, "dur": 23,
    "args": {
      "External id": 128095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128095,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128095, "pid": 5, "tid": 7, "ts": 1716454223441640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392805, "dur": 5,
    "args": {
      "External id": 128095,
      "cbid": 211, "correlation": 128095
    }
  },
  {
    "ph": "s", "id": 128095, "pid": 76337, "tid": -914061504, "ts": 1716454223392805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223441664, "dur": 33,
    "args": {
      "External id": 128101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128101, "pid": 5, "tid": 7, "ts": 1716454223441664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392832, "dur": 9,
    "args": {
      "External id": 128101,
      "cbid": 211, "correlation": 128101
    }
  },
  {
    "ph": "s", "id": 128101, "pid": 76337, "tid": -914061504, "ts": 1716454223392832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223441698, "dur": 3,
    "args": {
      "External id": 128109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128109,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 128109, "pid": 5, "tid": 7, "ts": 1716454223441698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392876, "dur": 9,
    "args": {
      "External id": 128109,
      "cbid": 211, "correlation": 128109
    }
  },
  {
    "ph": "s", "id": 128109, "pid": 76337, "tid": -914061504, "ts": 1716454223392876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392941, "dur": 1,
    "args": {
      "External id": 128125,
      "cbid": 251, "correlation": 128125
    }
  },
  {
    "ph": "f", "id": 128125, "pid": 76337, "tid": -914061504, "ts": 1716454223392941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392946, "dur": 0,
    "args": {
      "External id": 128127,
      "cbid": 251, "correlation": 128127
    }
  },
  {
    "ph": "f", "id": 128127, "pid": 76337, "tid": -914061504, "ts": 1716454223392946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223441703, "dur": 11,
    "args": {
      "External id": 128128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128128,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 128128, "pid": 5, "tid": 7, "ts": 1716454223441703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392948, "dur": 11,
    "args": {
      "External id": 128128,
      "cbid": 211, "correlation": 128128
    }
  },
  {
    "ph": "s", "id": 128128, "pid": 76337, "tid": -914061504, "ts": 1716454223392948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223441716, "dur": 5,
    "args": {
      "External id": 128130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128130,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 128130, "pid": 5, "tid": 7, "ts": 1716454223441716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223392961, "dur": 5,
    "args": {
      "External id": 128130,
      "cbid": 211, "correlation": 128130
    }
  },
  {
    "ph": "s", "id": 128130, "pid": 76337, "tid": -914061504, "ts": 1716454223392961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223441722, "dur": 29,
    "args": {
      "External id": 128140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128140, "pid": 5, "tid": 7, "ts": 1716454223441722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393027, "dur": 13,
    "args": {
      "External id": 128140,
      "cbid": 211, "correlation": 128140
    }
  },
  {
    "ph": "s", "id": 128140, "pid": 76337, "tid": -914061504, "ts": 1716454223393027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223441753, "dur": 31,
    "args": {
      "External id": 128160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128160,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 128160, "pid": 5, "tid": 7, "ts": 1716454223441753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393093, "dur": 12,
    "args": {
      "External id": 128160,
      "cbid": 211, "correlation": 128160
    }
  },
  {
    "ph": "s", "id": 128160, "pid": 76337, "tid": -914061504, "ts": 1716454223393093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223441785, "dur": 4,
    "args": {
      "External id": 128172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128172,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 128172, "pid": 5, "tid": 7, "ts": 1716454223441785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393115, "dur": 6,
    "args": {
      "External id": 128172,
      "cbid": 211, "correlation": 128172
    }
  },
  {
    "ph": "s", "id": 128172, "pid": 76337, "tid": -914061504, "ts": 1716454223393115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223441790, "dur": 29,
    "args": {
      "External id": 128175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128175, "pid": 5, "tid": 7, "ts": 1716454223441790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393133, "dur": 7,
    "args": {
      "External id": 128175,
      "cbid": 211, "correlation": 128175
    }
  },
  {
    "ph": "s", "id": 128175, "pid": 76337, "tid": -914061504, "ts": 1716454223393133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223441820, "dur": 19,
    "args": {
      "External id": 128184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128184,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128184, "pid": 5, "tid": 7, "ts": 1716454223441820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393175, "dur": 10,
    "args": {
      "External id": 128184,
      "cbid": 211, "correlation": 128184
    }
  },
  {
    "ph": "s", "id": 128184, "pid": 76337, "tid": -914061504, "ts": 1716454223393175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393237, "dur": 0,
    "args": {
      "External id": 128194,
      "cbid": 317, "correlation": 128194
    }
  },
  {
    "ph": "f", "id": 128194, "pid": 76337, "tid": -914061504, "ts": 1716454223393237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393238, "dur": 0,
    "args": {
      "External id": 128195,
      "cbid": 203, "correlation": 128195
    }
  },
  {
    "ph": "f", "id": 128195, "pid": 76337, "tid": -914061504, "ts": 1716454223393238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393239, "dur": 0,
    "args": {
      "External id": 128196,
      "cbid": 205, "correlation": 128196
    }
  },
  {
    "ph": "f", "id": 128196, "pid": 76337, "tid": -914061504, "ts": 1716454223393239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223441841, "dur": 24,
    "args": {
      "External id": 128200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128200,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128200, "pid": 5, "tid": 7, "ts": 1716454223441841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393252, "dur": 12,
    "args": {
      "External id": 128200,
      "cbid": 211, "correlation": 128200
    }
  },
  {
    "ph": "s", "id": 128200, "pid": 76337, "tid": -914061504, "ts": 1716454223393252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223441867, "dur": 44,
    "args": {
      "External id": 128202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128202, "pid": 5, "tid": 7, "ts": 1716454223441867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393267, "dur": 5,
    "args": {
      "External id": 128202,
      "cbid": 211, "correlation": 128202
    }
  },
  {
    "ph": "s", "id": 128202, "pid": 76337, "tid": -914061504, "ts": 1716454223393267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223441912, "dur": 644,
    "args": {
      "External id": 128204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128204,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128204, "pid": 5, "tid": 7, "ts": 1716454223441912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393278, "dur": 6,
    "args": {
      "External id": 128204,
      "cbid": 211, "correlation": 128204
    }
  },
  {
    "ph": "s", "id": 128204, "pid": 76337, "tid": -914061504, "ts": 1716454223393278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223442557, "dur": 22,
    "args": {
      "External id": 128206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128206,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128206, "pid": 5, "tid": 7, "ts": 1716454223442557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393288, "dur": 5,
    "args": {
      "External id": 128206,
      "cbid": 211, "correlation": 128206
    }
  },
  {
    "ph": "s", "id": 128206, "pid": 76337, "tid": -914061504, "ts": 1716454223393288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223442580, "dur": 33,
    "args": {
      "External id": 128212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128212, "pid": 5, "tid": 7, "ts": 1716454223442580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393316, "dur": 8,
    "args": {
      "External id": 128212,
      "cbid": 211, "correlation": 128212
    }
  },
  {
    "ph": "s", "id": 128212, "pid": 76337, "tid": -914061504, "ts": 1716454223393316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223442615, "dur": 28,
    "args": {
      "External id": 128220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128220,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128220, "pid": 5, "tid": 7, "ts": 1716454223442615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393348, "dur": 9,
    "args": {
      "External id": 128220,
      "cbid": 211, "correlation": 128220
    }
  },
  {
    "ph": "s", "id": 128220, "pid": 76337, "tid": -914061504, "ts": 1716454223393348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223442644, "dur": 20,
    "args": {
      "External id": 128228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128228,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128228, "pid": 5, "tid": 7, "ts": 1716454223442644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393377, "dur": 8,
    "args": {
      "External id": 128228,
      "cbid": 211, "correlation": 128228
    }
  },
  {
    "ph": "s", "id": 128228, "pid": 76337, "tid": -914061504, "ts": 1716454223393377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223442665, "dur": 30,
    "args": {
      "External id": 128248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128248,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 128248, "pid": 5, "tid": 7, "ts": 1716454223442665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393456, "dur": 13,
    "args": {
      "External id": 128248,
      "cbid": 211, "correlation": 128248
    }
  },
  {
    "ph": "s", "id": 128248, "pid": 76337, "tid": -914061504, "ts": 1716454223393456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223442696, "dur": 4,
    "args": {
      "External id": 128260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128260,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 128260, "pid": 5, "tid": 7, "ts": 1716454223442696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393478, "dur": 6,
    "args": {
      "External id": 128260,
      "cbid": 211, "correlation": 128260
    }
  },
  {
    "ph": "s", "id": 128260, "pid": 76337, "tid": -914061504, "ts": 1716454223393478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223442702, "dur": 30,
    "args": {
      "External id": 128263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128263, "pid": 5, "tid": 7, "ts": 1716454223442702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393495, "dur": 7,
    "args": {
      "External id": 128263,
      "cbid": 211, "correlation": 128263
    }
  },
  {
    "ph": "s", "id": 128263, "pid": 76337, "tid": -914061504, "ts": 1716454223393495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393553, "dur": 0,
    "args": {
      "External id": 128274,
      "cbid": 317, "correlation": 128274
    }
  },
  {
    "ph": "f", "id": 128274, "pid": 76337, "tid": -914061504, "ts": 1716454223393553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393554, "dur": 0,
    "args": {
      "External id": 128275,
      "cbid": 203, "correlation": 128275
    }
  },
  {
    "ph": "f", "id": 128275, "pid": 76337, "tid": -914061504, "ts": 1716454223393554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393555, "dur": 0,
    "args": {
      "External id": 128276,
      "cbid": 205, "correlation": 128276
    }
  },
  {
    "ph": "f", "id": 128276, "pid": 76337, "tid": -914061504, "ts": 1716454223393555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223442733, "dur": 22,
    "args": {
      "External id": 128280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128280,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128280, "pid": 5, "tid": 7, "ts": 1716454223442733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393568, "dur": 12,
    "args": {
      "External id": 128280,
      "cbid": 211, "correlation": 128280
    }
  },
  {
    "ph": "s", "id": 128280, "pid": 76337, "tid": -914061504, "ts": 1716454223393568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223442757, "dur": 116,
    "args": {
      "External id": 128282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128282,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128282, "pid": 5, "tid": 7, "ts": 1716454223442757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393586, "dur": 6,
    "args": {
      "External id": 128282,
      "cbid": 211, "correlation": 128282
    }
  },
  {
    "ph": "s", "id": 128282, "pid": 76337, "tid": -914061504, "ts": 1716454223393586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223442874, "dur": 23,
    "args": {
      "External id": 128284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128284,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128284, "pid": 5, "tid": 7, "ts": 1716454223442874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393595, "dur": 5,
    "args": {
      "External id": 128284,
      "cbid": 211, "correlation": 128284
    }
  },
  {
    "ph": "s", "id": 128284, "pid": 76337, "tid": -914061504, "ts": 1716454223393595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223442898, "dur": 33,
    "args": {
      "External id": 128290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128290, "pid": 5, "tid": 7, "ts": 1716454223442898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393623, "dur": 8,
    "args": {
      "External id": 128290,
      "cbid": 211, "correlation": 128290
    }
  },
  {
    "ph": "s", "id": 128290, "pid": 76337, "tid": -914061504, "ts": 1716454223393623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223442933, "dur": 163,
    "args": {
      "External id": 128299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128299,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128299, "pid": 5, "tid": 7, "ts": 1716454223442933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393706, "dur": 14,
    "args": {
      "External id": 128299,
      "cbid": 211, "correlation": 128299
    }
  },
  {
    "ph": "s", "id": 128299, "pid": 76337, "tid": -914061504, "ts": 1716454223393706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223443097, "dur": 64,
    "args": {
      "External id": 128321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128321,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128321, "pid": 5, "tid": 7, "ts": 1716454223443097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393764, "dur": 10,
    "args": {
      "External id": 128321,
      "cbid": 211, "correlation": 128321
    }
  },
  {
    "ph": "s", "id": 128321, "pid": 76337, "tid": -914061504, "ts": 1716454223393764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393851, "dur": 1,
    "args": {
      "External id": 128332,
      "cbid": 251, "correlation": 128332
    }
  },
  {
    "ph": "f", "id": 128332, "pid": 76337, "tid": -914061504, "ts": 1716454223393851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223443163, "dur": 155,
    "args": {
      "External id": 128333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128333,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128333, "pid": 5, "tid": 7, "ts": 1716454223443163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393856, "dur": 14,
    "args": {
      "External id": 128333,
      "cbid": 211, "correlation": 128333
    }
  },
  {
    "ph": "s", "id": 128333, "pid": 76337, "tid": -914061504, "ts": 1716454223393856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393927, "dur": 1,
    "args": {
      "External id": 128344,
      "cbid": 251, "correlation": 128344
    }
  },
  {
    "ph": "f", "id": 128344, "pid": 76337, "tid": -914061504, "ts": 1716454223393927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223443319, "dur": 145,
    "args": {
      "External id": 128345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128345,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128345, "pid": 5, "tid": 7, "ts": 1716454223443319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223393931, "dur": 11,
    "args": {
      "External id": 128345,
      "cbid": 211, "correlation": 128345
    }
  },
  {
    "ph": "s", "id": 128345, "pid": 76337, "tid": -914061504, "ts": 1716454223393931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394004, "dur": 1,
    "args": {
      "External id": 128356,
      "cbid": 251, "correlation": 128356
    }
  },
  {
    "ph": "f", "id": 128356, "pid": 76337, "tid": -914061504, "ts": 1716454223394004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223443465, "dur": 141,
    "args": {
      "External id": 128357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128357,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128357, "pid": 5, "tid": 7, "ts": 1716454223443465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394008, "dur": 12,
    "args": {
      "External id": 128357,
      "cbid": 211, "correlation": 128357
    }
  },
  {
    "ph": "s", "id": 128357, "pid": 76337, "tid": -914061504, "ts": 1716454223394008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223443607, "dur": 1937,
    "args": {
      "External id": 128378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128378,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 128378, "pid": 5, "tid": 7, "ts": 1716454223443607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394090, "dur": 13,
    "args": {
      "External id": 128378,
      "cbid": 211, "correlation": 128378
    }
  },
  {
    "ph": "s", "id": 128378, "pid": 76337, "tid": -914061504, "ts": 1716454223394090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394189, "dur": 1,
    "args": {
      "External id": 128396,
      "cbid": 251, "correlation": 128396
    }
  },
  {
    "ph": "f", "id": 128396, "pid": 76337, "tid": -914061504, "ts": 1716454223394189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223445546, "dur": 147,
    "args": {
      "External id": 128398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128398,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 128398, "pid": 5, "tid": 7, "ts": 1716454223445546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394194, "dur": 13,
    "args": {
      "External id": 128398,
      "cbid": 211, "correlation": 128398
    }
  },
  {
    "ph": "s", "id": 128398, "pid": 76337, "tid": -914061504, "ts": 1716454223394194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223445694, "dur": 36,
    "args": {
      "External id": 128406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128406,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128406, "pid": 5, "tid": 7, "ts": 1716454223445694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394266, "dur": 12,
    "args": {
      "External id": 128406,
      "cbid": 211, "correlation": 128406
    }
  },
  {
    "ph": "s", "id": 128406, "pid": 76337, "tid": -914061504, "ts": 1716454223394266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223445731, "dur": 50,
    "args": {
      "External id": 128414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128414,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128414, "pid": 5, "tid": 7, "ts": 1716454223445731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394304, "dur": 9,
    "args": {
      "External id": 128414,
      "cbid": 211, "correlation": 128414
    }
  },
  {
    "ph": "s", "id": 128414, "pid": 76337, "tid": -914061504, "ts": 1716454223394304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223445783, "dur": 30,
    "args": {
      "External id": 128425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128425,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128425, "pid": 5, "tid": 7, "ts": 1716454223445783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394376, "dur": 12,
    "args": {
      "External id": 128425,
      "cbid": 211, "correlation": 128425
    }
  },
  {
    "ph": "s", "id": 128425, "pid": 76337, "tid": -914061504, "ts": 1716454223394376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223445814, "dur": 34,
    "args": {
      "External id": 128447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128447,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128447, "pid": 5, "tid": 7, "ts": 1716454223445814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394406, "dur": 8,
    "args": {
      "External id": 128447,
      "cbid": 211, "correlation": 128447
    }
  },
  {
    "ph": "s", "id": 128447, "pid": 76337, "tid": -914061504, "ts": 1716454223394406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394491, "dur": 1,
    "args": {
      "External id": 128458,
      "cbid": 251, "correlation": 128458
    }
  },
  {
    "ph": "f", "id": 128458, "pid": 76337, "tid": -914061504, "ts": 1716454223394491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223445849, "dur": 91,
    "args": {
      "External id": 128459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128459,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128459, "pid": 5, "tid": 7, "ts": 1716454223445849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394496, "dur": 13,
    "args": {
      "External id": 128459,
      "cbid": 211, "correlation": 128459
    }
  },
  {
    "ph": "s", "id": 128459, "pid": 76337, "tid": -914061504, "ts": 1716454223394496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394565, "dur": 1,
    "args": {
      "External id": 128470,
      "cbid": 251, "correlation": 128470
    }
  },
  {
    "ph": "f", "id": 128470, "pid": 76337, "tid": -914061504, "ts": 1716454223394565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394568, "dur": 0,
    "args": {
      "External id": 128471,
      "cbid": 251, "correlation": 128471
    }
  },
  {
    "ph": "f", "id": 128471, "pid": 76337, "tid": -914061504, "ts": 1716454223394568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223445942, "dur": 11,
    "args": {
      "External id": 128472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128472,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 128472, "pid": 5, "tid": 7, "ts": 1716454223445942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394570, "dur": 12,
    "args": {
      "External id": 128472,
      "cbid": 211, "correlation": 128472
    }
  },
  {
    "ph": "s", "id": 128472, "pid": 76337, "tid": -914061504, "ts": 1716454223394570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223445954, "dur": 5,
    "args": {
      "External id": 128474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128474,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 128474, "pid": 5, "tid": 7, "ts": 1716454223445954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394583, "dur": 6,
    "args": {
      "External id": 128474,
      "cbid": 211, "correlation": 128474
    }
  },
  {
    "ph": "s", "id": 128474, "pid": 76337, "tid": -914061504, "ts": 1716454223394583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394641, "dur": 1,
    "args": {
      "External id": 128485,
      "cbid": 251, "correlation": 128485
    }
  },
  {
    "ph": "f", "id": 128485, "pid": 76337, "tid": -914061504, "ts": 1716454223394641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394645, "dur": 0,
    "args": {
      "External id": 128486,
      "cbid": 251, "correlation": 128486
    }
  },
  {
    "ph": "f", "id": 128486, "pid": 76337, "tid": -914061504, "ts": 1716454223394645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223445960, "dur": 7,
    "args": {
      "External id": 128487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128487,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 128487, "pid": 5, "tid": 7, "ts": 1716454223445960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394646, "dur": 12,
    "args": {
      "External id": 128487,
      "cbid": 211, "correlation": 128487
    }
  },
  {
    "ph": "s", "id": 128487, "pid": 76337, "tid": -914061504, "ts": 1716454223394646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223445969, "dur": 3,
    "args": {
      "External id": 128489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128489,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 128489, "pid": 5, "tid": 7, "ts": 1716454223445969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394660, "dur": 6,
    "args": {
      "External id": 128489,
      "cbid": 211, "correlation": 128489
    }
  },
  {
    "ph": "s", "id": 128489, "pid": 76337, "tid": -914061504, "ts": 1716454223394660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223445973, "dur": 92,
    "args": {
      "External id": 128510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128510,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 128510, "pid": 5, "tid": 7, "ts": 1716454223445973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394732, "dur": 12,
    "args": {
      "External id": 128510,
      "cbid": 211, "correlation": 128510
    }
  },
  {
    "ph": "s", "id": 128510, "pid": 76337, "tid": -914061504, "ts": 1716454223394732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394828, "dur": 1,
    "args": {
      "External id": 128528,
      "cbid": 251, "correlation": 128528
    }
  },
  {
    "ph": "f", "id": 128528, "pid": 76337, "tid": -914061504, "ts": 1716454223394828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223446067, "dur": 98,
    "args": {
      "External id": 128530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128530,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128530, "pid": 5, "tid": 7, "ts": 1716454223446067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394833, "dur": 13,
    "args": {
      "External id": 128530,
      "cbid": 211, "correlation": 128530
    }
  },
  {
    "ph": "s", "id": 128530, "pid": 76337, "tid": -914061504, "ts": 1716454223394833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223446165, "dur": 20,
    "args": {
      "External id": 128538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128538,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128538, "pid": 5, "tid": 7, "ts": 1716454223446165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394903, "dur": 13,
    "args": {
      "External id": 128538,
      "cbid": 211, "correlation": 128538
    }
  },
  {
    "ph": "s", "id": 128538, "pid": 76337, "tid": -914061504, "ts": 1716454223394903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223446186, "dur": 38,
    "args": {
      "External id": 128546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128546, "pid": 5, "tid": 7, "ts": 1716454223446186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223394945, "dur": 9,
    "args": {
      "External id": 128546,
      "cbid": 211, "correlation": 128546
    }
  },
  {
    "ph": "s", "id": 128546, "pid": 76337, "tid": -914061504, "ts": 1716454223394945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223446226, "dur": 34,
    "args": {
      "External id": 128568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128568,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128568, "pid": 5, "tid": 7, "ts": 1716454223446226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395004, "dur": 11,
    "args": {
      "External id": 128568,
      "cbid": 211, "correlation": 128568
    }
  },
  {
    "ph": "s", "id": 128568, "pid": 76337, "tid": -914061504, "ts": 1716454223395004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395094, "dur": 1,
    "args": {
      "External id": 128584,
      "cbid": 251, "correlation": 128584
    }
  },
  {
    "ph": "f", "id": 128584, "pid": 76337, "tid": -914061504, "ts": 1716454223395094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395099, "dur": 0,
    "args": {
      "External id": 128586,
      "cbid": 251, "correlation": 128586
    }
  },
  {
    "ph": "f", "id": 128586, "pid": 76337, "tid": -914061504, "ts": 1716454223395099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223446261, "dur": 535,
    "args": {
      "External id": 128587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128587,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 128587, "pid": 5, "tid": 7, "ts": 1716454223446261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395102, "dur": 14,
    "args": {
      "External id": 128587,
      "cbid": 211, "correlation": 128587
    }
  },
  {
    "ph": "s", "id": 128587, "pid": 76337, "tid": -914061504, "ts": 1716454223395102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223446797, "dur": 125,
    "args": {
      "External id": 128595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128595,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128595, "pid": 5, "tid": 7, "ts": 1716454223446797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395168, "dur": 12,
    "args": {
      "External id": 128595,
      "cbid": 211, "correlation": 128595
    }
  },
  {
    "ph": "s", "id": 128595, "pid": 76337, "tid": -914061504, "ts": 1716454223395168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223446924, "dur": 128,
    "args": {
      "External id": 128603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128603, "pid": 5, "tid": 7, "ts": 1716454223446924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395198, "dur": 8,
    "args": {
      "External id": 128603,
      "cbid": 211, "correlation": 128603
    }
  },
  {
    "ph": "s", "id": 128603, "pid": 76337, "tid": -914061504, "ts": 1716454223395198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395274, "dur": 1,
    "args": {
      "External id": 128619,
      "cbid": 251, "correlation": 128619
    }
  },
  {
    "ph": "f", "id": 128619, "pid": 76337, "tid": -914061504, "ts": 1716454223395274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223447053, "dur": 301,
    "args": {
      "External id": 128621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128621,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128621, "pid": 5, "tid": 7, "ts": 1716454223447053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395280, "dur": 12,
    "args": {
      "External id": 128621,
      "cbid": 211, "correlation": 128621
    }
  },
  {
    "ph": "s", "id": 128621, "pid": 76337, "tid": -914061504, "ts": 1716454223395280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223447356, "dur": 27,
    "args": {
      "External id": 128629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128629,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128629, "pid": 5, "tid": 7, "ts": 1716454223447356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395323, "dur": 9,
    "args": {
      "External id": 128629,
      "cbid": 211, "correlation": 128629
    }
  },
  {
    "ph": "s", "id": 128629, "pid": 76337, "tid": -914061504, "ts": 1716454223395323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223447385, "dur": 81,
    "args": {
      "External id": 128640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128640, "pid": 5, "tid": 7, "ts": 1716454223447385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395390, "dur": 12,
    "args": {
      "External id": 128640,
      "cbid": 211, "correlation": 128640
    }
  },
  {
    "ph": "s", "id": 128640, "pid": 76337, "tid": -914061504, "ts": 1716454223395390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395454, "dur": 0,
    "args": {
      "External id": 128652,
      "cbid": 317, "correlation": 128652
    }
  },
  {
    "ph": "f", "id": 128652, "pid": 76337, "tid": -914061504, "ts": 1716454223395454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395455, "dur": 0,
    "args": {
      "External id": 128653,
      "cbid": 203, "correlation": 128653
    }
  },
  {
    "ph": "f", "id": 128653, "pid": 76337, "tid": -914061504, "ts": 1716454223395455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395455, "dur": 0,
    "args": {
      "External id": 128654,
      "cbid": 205, "correlation": 128654
    }
  },
  {
    "ph": "f", "id": 128654, "pid": 76337, "tid": -914061504, "ts": 1716454223395455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223447467, "dur": 23,
    "args": {
      "External id": 128658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128658, "pid": 5, "tid": 7, "ts": 1716454223447467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395471, "dur": 12,
    "args": {
      "External id": 128658,
      "cbid": 211, "correlation": 128658
    }
  },
  {
    "ph": "s", "id": 128658, "pid": 76337, "tid": -914061504, "ts": 1716454223395471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223447491, "dur": 121,
    "args": {
      "External id": 128660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128660,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128660, "pid": 5, "tid": 7, "ts": 1716454223447491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395489, "dur": 6,
    "args": {
      "External id": 128660,
      "cbid": 211, "correlation": 128660
    }
  },
  {
    "ph": "s", "id": 128660, "pid": 76337, "tid": -914061504, "ts": 1716454223395489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223447613, "dur": 23,
    "args": {
      "External id": 128662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128662,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128662, "pid": 5, "tid": 7, "ts": 1716454223447613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395499, "dur": 5,
    "args": {
      "External id": 128662,
      "cbid": 211, "correlation": 128662
    }
  },
  {
    "ph": "s", "id": 128662, "pid": 76337, "tid": -914061504, "ts": 1716454223395499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223447638, "dur": 32,
    "args": {
      "External id": 128668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128668, "pid": 5, "tid": 7, "ts": 1716454223447638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395527, "dur": 8,
    "args": {
      "External id": 128668,
      "cbid": 211, "correlation": 128668
    }
  },
  {
    "ph": "s", "id": 128668, "pid": 76337, "tid": -914061504, "ts": 1716454223395527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223447671, "dur": 27,
    "args": {
      "External id": 128676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128676,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128676, "pid": 5, "tid": 7, "ts": 1716454223447671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395558, "dur": 9,
    "args": {
      "External id": 128676,
      "cbid": 211, "correlation": 128676
    }
  },
  {
    "ph": "s", "id": 128676, "pid": 76337, "tid": -914061504, "ts": 1716454223395558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395629, "dur": 0,
    "args": {
      "External id": 128686,
      "cbid": 317, "correlation": 128686
    }
  },
  {
    "ph": "f", "id": 128686, "pid": 76337, "tid": -914061504, "ts": 1716454223395629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395630, "dur": 0,
    "args": {
      "External id": 128687,
      "cbid": 203, "correlation": 128687
    }
  },
  {
    "ph": "f", "id": 128687, "pid": 76337, "tid": -914061504, "ts": 1716454223395630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395631, "dur": 0,
    "args": {
      "External id": 128688,
      "cbid": 205, "correlation": 128688
    }
  },
  {
    "ph": "f", "id": 128688, "pid": 76337, "tid": -914061504, "ts": 1716454223395631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223447700, "dur": 24,
    "args": {
      "External id": 128692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128692, "pid": 5, "tid": 7, "ts": 1716454223447700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395645, "dur": 12,
    "args": {
      "External id": 128692,
      "cbid": 211, "correlation": 128692
    }
  },
  {
    "ph": "s", "id": 128692, "pid": 76337, "tid": -914061504, "ts": 1716454223395645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223447724, "dur": 44,
    "args": {
      "External id": 128694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128694,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128694, "pid": 5, "tid": 7, "ts": 1716454223447724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395659, "dur": 6,
    "args": {
      "External id": 128694,
      "cbid": 211, "correlation": 128694
    }
  },
  {
    "ph": "s", "id": 128694, "pid": 76337, "tid": -914061504, "ts": 1716454223395659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223447770, "dur": 235,
    "args": {
      "External id": 128696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128696,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 128696, "pid": 5, "tid": 7, "ts": 1716454223447770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395672, "dur": 6,
    "args": {
      "External id": 128696,
      "cbid": 211, "correlation": 128696
    }
  },
  {
    "ph": "s", "id": 128696, "pid": 76337, "tid": -914061504, "ts": 1716454223395672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448006, "dur": 7,
    "args": {
      "External id": 128698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128698,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128698, "pid": 5, "tid": 7, "ts": 1716454223448006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395682, "dur": 5,
    "args": {
      "External id": 128698,
      "cbid": 211, "correlation": 128698
    }
  },
  {
    "ph": "s", "id": 128698, "pid": 76337, "tid": -914061504, "ts": 1716454223395682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223448014, "dur": 9,
    "args": {
      "External id": 128704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128704,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128704, "pid": 5, "tid": 7, "ts": 1716454223448014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395708, "dur": 9,
    "args": {
      "External id": 128704,
      "cbid": 211, "correlation": 128704
    }
  },
  {
    "ph": "s", "id": 128704, "pid": 76337, "tid": -914061504, "ts": 1716454223395708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448024, "dur": 12,
    "args": {
      "External id": 128724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128724,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 128724, "pid": 5, "tid": 7, "ts": 1716454223448024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395801, "dur": 12,
    "args": {
      "External id": 128724,
      "cbid": 211, "correlation": 128724
    }
  },
  {
    "ph": "s", "id": 128724, "pid": 76337, "tid": -914061504, "ts": 1716454223395801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223448037, "dur": 4,
    "args": {
      "External id": 128736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128736,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 128736, "pid": 5, "tid": 7, "ts": 1716454223448037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395824, "dur": 6,
    "args": {
      "External id": 128736,
      "cbid": 211, "correlation": 128736
    }
  },
  {
    "ph": "s", "id": 128736, "pid": 76337, "tid": -914061504, "ts": 1716454223395824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223448042, "dur": 12,
    "args": {
      "External id": 128739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128739,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128739, "pid": 5, "tid": 7, "ts": 1716454223448042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395842, "dur": 7,
    "args": {
      "External id": 128739,
      "cbid": 211, "correlation": 128739
    }
  },
  {
    "ph": "s", "id": 128739, "pid": 76337, "tid": -914061504, "ts": 1716454223395842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223448056, "dur": 7,
    "args": {
      "External id": 128748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128748,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128748, "pid": 5, "tid": 7, "ts": 1716454223448056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395881, "dur": 9,
    "args": {
      "External id": 128748,
      "cbid": 211, "correlation": 128748
    }
  },
  {
    "ph": "s", "id": 128748, "pid": 76337, "tid": -914061504, "ts": 1716454223395881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395934, "dur": 0,
    "args": {
      "External id": 128758,
      "cbid": 317, "correlation": 128758
    }
  },
  {
    "ph": "f", "id": 128758, "pid": 76337, "tid": -914061504, "ts": 1716454223395934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395935, "dur": 0,
    "args": {
      "External id": 128759,
      "cbid": 203, "correlation": 128759
    }
  },
  {
    "ph": "f", "id": 128759, "pid": 76337, "tid": -914061504, "ts": 1716454223395935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395935, "dur": 0,
    "args": {
      "External id": 128760,
      "cbid": 205, "correlation": 128760
    }
  },
  {
    "ph": "f", "id": 128760, "pid": 76337, "tid": -914061504, "ts": 1716454223395935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448064, "dur": 6,
    "args": {
      "External id": 128764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128764,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128764, "pid": 5, "tid": 7, "ts": 1716454223448064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395951, "dur": 11,
    "args": {
      "External id": 128764,
      "cbid": 211, "correlation": 128764
    }
  },
  {
    "ph": "s", "id": 128764, "pid": 76337, "tid": -914061504, "ts": 1716454223395951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448071, "dur": 83,
    "args": {
      "External id": 128766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128766, "pid": 5, "tid": 7, "ts": 1716454223448071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395965, "dur": 6,
    "args": {
      "External id": 128766,
      "cbid": 211, "correlation": 128766
    }
  },
  {
    "ph": "s", "id": 128766, "pid": 76337, "tid": -914061504, "ts": 1716454223395965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223448156, "dur": 1,
    "args": {
      "External id": 128768,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 128768,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 128768, "pid": 5, "tid": 7, "ts": 1716454223448156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395986, "dur": 9,
    "args": {
      "External id": 128768,
      "cbid": 51, "correlation": 128768
    }
  },
  {
    "ph": "s", "id": 128768, "pid": 76337, "tid": -914061504, "ts": 1716454223395986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223448160, "dur": 537,
    "args": {
      "External id": 128769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128769,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128769, "pid": 5, "tid": 7, "ts": 1716454223448160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223395997, "dur": 8,
    "args": {
      "External id": 128769,
      "cbid": 211, "correlation": 128769
    }
  },
  {
    "ph": "s", "id": 128769, "pid": 76337, "tid": -914061504, "ts": 1716454223395997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448698, "dur": 11,
    "args": {
      "External id": 128771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128771,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128771, "pid": 5, "tid": 7, "ts": 1716454223448698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396009, "dur": 5,
    "args": {
      "External id": 128771,
      "cbid": 211, "correlation": 128771
    }
  },
  {
    "ph": "s", "id": 128771, "pid": 76337, "tid": -914061504, "ts": 1716454223396009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223448710, "dur": 14,
    "args": {
      "External id": 128777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128777, "pid": 5, "tid": 7, "ts": 1716454223448710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396038, "dur": 8,
    "args": {
      "External id": 128777,
      "cbid": 211, "correlation": 128777
    }
  },
  {
    "ph": "s", "id": 128777, "pid": 76337, "tid": -914061504, "ts": 1716454223396038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223448726, "dur": 3,
    "args": {
      "External id": 128785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128785,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 128785, "pid": 5, "tid": 7, "ts": 1716454223448726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396082, "dur": 9,
    "args": {
      "External id": 128785,
      "cbid": 211, "correlation": 128785
    }
  },
  {
    "ph": "s", "id": 128785, "pid": 76337, "tid": -914061504, "ts": 1716454223396082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396146, "dur": 1,
    "args": {
      "External id": 128801,
      "cbid": 251, "correlation": 128801
    }
  },
  {
    "ph": "f", "id": 128801, "pid": 76337, "tid": -914061504, "ts": 1716454223396146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396152, "dur": 0,
    "args": {
      "External id": 128803,
      "cbid": 251, "correlation": 128803
    }
  },
  {
    "ph": "f", "id": 128803, "pid": 76337, "tid": -914061504, "ts": 1716454223396152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223448731, "dur": 13,
    "args": {
      "External id": 128804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128804,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128804, "pid": 5, "tid": 7, "ts": 1716454223448731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396153, "dur": 11,
    "args": {
      "External id": 128804,
      "cbid": 211, "correlation": 128804
    }
  },
  {
    "ph": "s", "id": 128804, "pid": 76337, "tid": -914061504, "ts": 1716454223396153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223448745, "dur": 5,
    "args": {
      "External id": 128806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128806,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128806, "pid": 5, "tid": 7, "ts": 1716454223448745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396166, "dur": 7,
    "args": {
      "External id": 128806,
      "cbid": 211, "correlation": 128806
    }
  },
  {
    "ph": "s", "id": 128806, "pid": 76337, "tid": -914061504, "ts": 1716454223396166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223448751, "dur": 17,
    "args": {
      "External id": 128816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128816,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128816, "pid": 5, "tid": 7, "ts": 1716454223448751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396225, "dur": 12,
    "args": {
      "External id": 128816,
      "cbid": 211, "correlation": 128816
    }
  },
  {
    "ph": "s", "id": 128816, "pid": 76337, "tid": -914061504, "ts": 1716454223396225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448770, "dur": 19,
    "args": {
      "External id": 128836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128836,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 128836, "pid": 5, "tid": 7, "ts": 1716454223448770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396291, "dur": 11,
    "args": {
      "External id": 128836,
      "cbid": 211, "correlation": 128836
    }
  },
  {
    "ph": "s", "id": 128836, "pid": 76337, "tid": -914061504, "ts": 1716454223396291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223448790, "dur": 5,
    "args": {
      "External id": 128848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128848,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 128848, "pid": 5, "tid": 7, "ts": 1716454223448790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396312, "dur": 6,
    "args": {
      "External id": 128848,
      "cbid": 211, "correlation": 128848
    }
  },
  {
    "ph": "s", "id": 128848, "pid": 76337, "tid": -914061504, "ts": 1716454223396312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223448796, "dur": 17,
    "args": {
      "External id": 128851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128851,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128851, "pid": 5, "tid": 7, "ts": 1716454223448796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396332, "dur": 6,
    "args": {
      "External id": 128851,
      "cbid": 211, "correlation": 128851
    }
  },
  {
    "ph": "s", "id": 128851, "pid": 76337, "tid": -914061504, "ts": 1716454223396332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223448815, "dur": 10,
    "args": {
      "External id": 128860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128860,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128860, "pid": 5, "tid": 7, "ts": 1716454223448815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396372, "dur": 9,
    "args": {
      "External id": 128860,
      "cbid": 211, "correlation": 128860
    }
  },
  {
    "ph": "s", "id": 128860, "pid": 76337, "tid": -914061504, "ts": 1716454223396372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396434, "dur": 0,
    "args": {
      "External id": 128870,
      "cbid": 317, "correlation": 128870
    }
  },
  {
    "ph": "f", "id": 128870, "pid": 76337, "tid": -914061504, "ts": 1716454223396434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396435, "dur": 0,
    "args": {
      "External id": 128871,
      "cbid": 203, "correlation": 128871
    }
  },
  {
    "ph": "f", "id": 128871, "pid": 76337, "tid": -914061504, "ts": 1716454223396435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396435, "dur": 0,
    "args": {
      "External id": 128872,
      "cbid": 205, "correlation": 128872
    }
  },
  {
    "ph": "f", "id": 128872, "pid": 76337, "tid": -914061504, "ts": 1716454223396435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448826, "dur": 11,
    "args": {
      "External id": 128876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128876,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128876, "pid": 5, "tid": 7, "ts": 1716454223448826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396453, "dur": 12,
    "args": {
      "External id": 128876,
      "cbid": 211, "correlation": 128876
    }
  },
  {
    "ph": "s", "id": 128876, "pid": 76337, "tid": -914061504, "ts": 1716454223396453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223448838, "dur": 162,
    "args": {
      "External id": 128878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128878,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128878, "pid": 5, "tid": 7, "ts": 1716454223448838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396467, "dur": 6,
    "args": {
      "External id": 128878,
      "cbid": 211, "correlation": 128878
    }
  },
  {
    "ph": "s", "id": 128878, "pid": 76337, "tid": -914061504, "ts": 1716454223396467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223449002, "dur": 1,
    "args": {
      "External id": 128880,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 128880,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 128880, "pid": 5, "tid": 7, "ts": 1716454223449002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396480, "dur": 7,
    "args": {
      "External id": 128880,
      "cbid": 51, "correlation": 128880
    }
  },
  {
    "ph": "s", "id": 128880, "pid": 76337, "tid": -914061504, "ts": 1716454223396480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223449006, "dur": 659,
    "args": {
      "External id": 128881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128881,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 128881, "pid": 5, "tid": 7, "ts": 1716454223449006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396488, "dur": 6,
    "args": {
      "External id": 128881,
      "cbid": 211, "correlation": 128881
    }
  },
  {
    "ph": "s", "id": 128881, "pid": 76337, "tid": -914061504, "ts": 1716454223396488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223449666, "dur": 12,
    "args": {
      "External id": 128883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128883,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128883, "pid": 5, "tid": 7, "ts": 1716454223449666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396498, "dur": 5,
    "args": {
      "External id": 128883,
      "cbid": 211, "correlation": 128883
    }
  },
  {
    "ph": "s", "id": 128883, "pid": 76337, "tid": -914061504, "ts": 1716454223396498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223449680, "dur": 15,
    "args": {
      "External id": 128889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128889,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128889, "pid": 5, "tid": 7, "ts": 1716454223449680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396528, "dur": 9,
    "args": {
      "External id": 128889,
      "cbid": 211, "correlation": 128889
    }
  },
  {
    "ph": "s", "id": 128889, "pid": 76337, "tid": -914061504, "ts": 1716454223396528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396585, "dur": 0,
    "args": {
      "External id": 128899,
      "cbid": 317, "correlation": 128899
    }
  },
  {
    "ph": "f", "id": 128899, "pid": 76337, "tid": -914061504, "ts": 1716454223396585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396586, "dur": 0,
    "args": {
      "External id": 128900,
      "cbid": 203, "correlation": 128900
    }
  },
  {
    "ph": "f", "id": 128900, "pid": 76337, "tid": -914061504, "ts": 1716454223396586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396587, "dur": 0,
    "args": {
      "External id": 128901,
      "cbid": 205, "correlation": 128901
    }
  },
  {
    "ph": "f", "id": 128901, "pid": 76337, "tid": -914061504, "ts": 1716454223396587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223449696, "dur": 8,
    "args": {
      "External id": 128905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128905,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128905, "pid": 5, "tid": 7, "ts": 1716454223449696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396602, "dur": 12,
    "args": {
      "External id": 128905,
      "cbid": 211, "correlation": 128905
    }
  },
  {
    "ph": "s", "id": 128905, "pid": 76337, "tid": -914061504, "ts": 1716454223396602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223449705, "dur": 4,
    "args": {
      "External id": 128907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128907,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 128907, "pid": 5, "tid": 7, "ts": 1716454223449705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396619, "dur": 6,
    "args": {
      "External id": 128907,
      "cbid": 211, "correlation": 128907
    }
  },
  {
    "ph": "s", "id": 128907, "pid": 76337, "tid": -914061504, "ts": 1716454223396619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396628, "dur": 0,
    "args": {
      "External id": 128908,
      "cbid": 51, "correlation": 128908
    }
  },
  {
    "ph": "s", "id": 128908, "pid": 76337, "tid": -914061504, "ts": 1716454223396628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223449710, "dur": 57,
    "args": {
      "External id": 128909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128909,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 128909, "pid": 5, "tid": 7, "ts": 1716454223449710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396629, "dur": 5,
    "args": {
      "External id": 128909,
      "cbid": 211, "correlation": 128909
    }
  },
  {
    "ph": "s", "id": 128909, "pid": 76337, "tid": -914061504, "ts": 1716454223396629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223449768, "dur": 14,
    "args": {
      "External id": 128914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128914, "pid": 5, "tid": 7, "ts": 1716454223449768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396654, "dur": 8,
    "args": {
      "External id": 128914,
      "cbid": 211, "correlation": 128914
    }
  },
  {
    "ph": "s", "id": 128914, "pid": 76337, "tid": -914061504, "ts": 1716454223396654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223449783, "dur": 12,
    "args": {
      "External id": 128922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128922,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128922, "pid": 5, "tid": 7, "ts": 1716454223449783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396683, "dur": 8,
    "args": {
      "External id": 128922,
      "cbid": 211, "correlation": 128922
    }
  },
  {
    "ph": "s", "id": 128922, "pid": 76337, "tid": -914061504, "ts": 1716454223396683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223449797, "dur": 10,
    "args": {
      "External id": 128930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128930,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128930, "pid": 5, "tid": 7, "ts": 1716454223449797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396712, "dur": 9,
    "args": {
      "External id": 128930,
      "cbid": 211, "correlation": 128930
    }
  },
  {
    "ph": "s", "id": 128930, "pid": 76337, "tid": -914061504, "ts": 1716454223396712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223449809, "dur": 18,
    "args": {
      "External id": 128950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128950,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 128950, "pid": 5, "tid": 7, "ts": 1716454223449809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396792, "dur": 12,
    "args": {
      "External id": 128950,
      "cbid": 211, "correlation": 128950
    }
  },
  {
    "ph": "s", "id": 128950, "pid": 76337, "tid": -914061504, "ts": 1716454223396792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223449829, "dur": 5,
    "args": {
      "External id": 128962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128962,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 128962, "pid": 5, "tid": 7, "ts": 1716454223449829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396814, "dur": 7,
    "args": {
      "External id": 128962,
      "cbid": 211, "correlation": 128962
    }
  },
  {
    "ph": "s", "id": 128962, "pid": 76337, "tid": -914061504, "ts": 1716454223396814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223449835, "dur": 17,
    "args": {
      "External id": 128965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128965, "pid": 5, "tid": 7, "ts": 1716454223449835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396832, "dur": 6,
    "args": {
      "External id": 128965,
      "cbid": 211, "correlation": 128965
    }
  },
  {
    "ph": "s", "id": 128965, "pid": 76337, "tid": -914061504, "ts": 1716454223396832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396887, "dur": 0,
    "args": {
      "External id": 128976,
      "cbid": 317, "correlation": 128976
    }
  },
  {
    "ph": "f", "id": 128976, "pid": 76337, "tid": -914061504, "ts": 1716454223396887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396888, "dur": 0,
    "args": {
      "External id": 128977,
      "cbid": 203, "correlation": 128977
    }
  },
  {
    "ph": "f", "id": 128977, "pid": 76337, "tid": -914061504, "ts": 1716454223396888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396889, "dur": 0,
    "args": {
      "External id": 128978,
      "cbid": 205, "correlation": 128978
    }
  },
  {
    "ph": "f", "id": 128978, "pid": 76337, "tid": -914061504, "ts": 1716454223396889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223449853, "dur": 12,
    "args": {
      "External id": 128982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128982, "pid": 5, "tid": 7, "ts": 1716454223449853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396904, "dur": 12,
    "args": {
      "External id": 128982,
      "cbid": 211, "correlation": 128982
    }
  },
  {
    "ph": "s", "id": 128982, "pid": 76337, "tid": -914061504, "ts": 1716454223396904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223449866, "dur": 3,
    "args": {
      "External id": 128984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 128984, "pid": 5, "tid": 7, "ts": 1716454223449866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396921, "dur": 6,
    "args": {
      "External id": 128984,
      "cbid": 211, "correlation": 128984
    }
  },
  {
    "ph": "s", "id": 128984, "pid": 76337, "tid": -914061504, "ts": 1716454223396921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396930, "dur": 0,
    "args": {
      "External id": 128985,
      "cbid": 51, "correlation": 128985
    }
  },
  {
    "ph": "s", "id": 128985, "pid": 76337, "tid": -914061504, "ts": 1716454223396930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223449871, "dur": 97,
    "args": {
      "External id": 128986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128986,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 128986, "pid": 5, "tid": 7, "ts": 1716454223449871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396931, "dur": 6,
    "args": {
      "External id": 128986,
      "cbid": 211, "correlation": 128986
    }
  },
  {
    "ph": "s", "id": 128986, "pid": 76337, "tid": -914061504, "ts": 1716454223396931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223449969, "dur": 15,
    "args": {
      "External id": 128991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 128991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 128991, "pid": 5, "tid": 7, "ts": 1716454223449969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223396959, "dur": 9,
    "args": {
      "External id": 128991,
      "cbid": 211, "correlation": 128991
    }
  },
  {
    "ph": "s", "id": 128991, "pid": 76337, "tid": -914061504, "ts": 1716454223396959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223449986, "dur": 84,
    "args": {
      "External id": 129000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129000,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129000, "pid": 5, "tid": 7, "ts": 1716454223449986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397049, "dur": 14,
    "args": {
      "External id": 129000,
      "cbid": 211, "correlation": 129000
    }
  },
  {
    "ph": "s", "id": 129000, "pid": 76337, "tid": -914061504, "ts": 1716454223397049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223450071, "dur": 30,
    "args": {
      "External id": 129022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129022,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129022, "pid": 5, "tid": 7, "ts": 1716454223450071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397105, "dur": 10,
    "args": {
      "External id": 129022,
      "cbid": 211, "correlation": 129022
    }
  },
  {
    "ph": "s", "id": 129022, "pid": 76337, "tid": -914061504, "ts": 1716454223397105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397197, "dur": 1,
    "args": {
      "External id": 129033,
      "cbid": 251, "correlation": 129033
    }
  },
  {
    "ph": "f", "id": 129033, "pid": 76337, "tid": -914061504, "ts": 1716454223397197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223450102, "dur": 162,
    "args": {
      "External id": 129034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129034,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129034, "pid": 5, "tid": 7, "ts": 1716454223450102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397202, "dur": 13,
    "args": {
      "External id": 129034,
      "cbid": 211, "correlation": 129034
    }
  },
  {
    "ph": "s", "id": 129034, "pid": 76337, "tid": -914061504, "ts": 1716454223397202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397273, "dur": 1,
    "args": {
      "External id": 129045,
      "cbid": 251, "correlation": 129045
    }
  },
  {
    "ph": "f", "id": 129045, "pid": 76337, "tid": -914061504, "ts": 1716454223397273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223450266, "dur": 159,
    "args": {
      "External id": 129046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129046,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129046, "pid": 5, "tid": 7, "ts": 1716454223450266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397277, "dur": 11,
    "args": {
      "External id": 129046,
      "cbid": 211, "correlation": 129046
    }
  },
  {
    "ph": "s", "id": 129046, "pid": 76337, "tid": -914061504, "ts": 1716454223397277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397341, "dur": 1,
    "args": {
      "External id": 129057,
      "cbid": 251, "correlation": 129057
    }
  },
  {
    "ph": "f", "id": 129057, "pid": 76337, "tid": -914061504, "ts": 1716454223397341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223450426, "dur": 159,
    "args": {
      "External id": 129058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129058,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129058, "pid": 5, "tid": 7, "ts": 1716454223450426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397345, "dur": 11,
    "args": {
      "External id": 129058,
      "cbid": 211, "correlation": 129058
    }
  },
  {
    "ph": "s", "id": 129058, "pid": 76337, "tid": -914061504, "ts": 1716454223397345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223450587, "dur": 338,
    "args": {
      "External id": 129083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129083,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129083, "pid": 5, "tid": 7, "ts": 1716454223450587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397430, "dur": 13,
    "args": {
      "External id": 129083,
      "cbid": 211, "correlation": 129083
    }
  },
  {
    "ph": "s", "id": 129083, "pid": 76337, "tid": -914061504, "ts": 1716454223397430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397532, "dur": 1,
    "args": {
      "External id": 129101,
      "cbid": 251, "correlation": 129101
    }
  },
  {
    "ph": "f", "id": 129101, "pid": 76337, "tid": -914061504, "ts": 1716454223397532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223450926, "dur": 166,
    "args": {
      "External id": 129103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129103,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129103, "pid": 5, "tid": 7, "ts": 1716454223450926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397538, "dur": 13,
    "args": {
      "External id": 129103,
      "cbid": 211, "correlation": 129103
    }
  },
  {
    "ph": "s", "id": 129103, "pid": 76337, "tid": -914061504, "ts": 1716454223397538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223451093, "dur": 19,
    "args": {
      "External id": 129111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129111,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129111, "pid": 5, "tid": 7, "ts": 1716454223451093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397608, "dur": 13,
    "args": {
      "External id": 129111,
      "cbid": 211, "correlation": 129111
    }
  },
  {
    "ph": "s", "id": 129111, "pid": 76337, "tid": -914061504, "ts": 1716454223397608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223451113, "dur": 28,
    "args": {
      "External id": 129119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129119,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129119, "pid": 5, "tid": 7, "ts": 1716454223451113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397648, "dur": 8,
    "args": {
      "External id": 129119,
      "cbid": 211, "correlation": 129119
    }
  },
  {
    "ph": "s", "id": 129119, "pid": 76337, "tid": -914061504, "ts": 1716454223397648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223451143, "dur": 18,
    "args": {
      "External id": 129130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129130, "pid": 5, "tid": 7, "ts": 1716454223451143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397720, "dur": 12,
    "args": {
      "External id": 129130,
      "cbid": 211, "correlation": 129130
    }
  },
  {
    "ph": "s", "id": 129130, "pid": 76337, "tid": -914061504, "ts": 1716454223397720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223451162, "dur": 16,
    "args": {
      "External id": 129152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129152,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129152, "pid": 5, "tid": 7, "ts": 1716454223451162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397751, "dur": 7,
    "args": {
      "External id": 129152,
      "cbid": 211, "correlation": 129152
    }
  },
  {
    "ph": "s", "id": 129152, "pid": 76337, "tid": -914061504, "ts": 1716454223397751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397837, "dur": 1,
    "args": {
      "External id": 129163,
      "cbid": 251, "correlation": 129163
    }
  },
  {
    "ph": "f", "id": 129163, "pid": 76337, "tid": -914061504, "ts": 1716454223397837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223451180, "dur": 89,
    "args": {
      "External id": 129164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129164,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129164, "pid": 5, "tid": 7, "ts": 1716454223451180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397844, "dur": 15,
    "args": {
      "External id": 129164,
      "cbid": 211, "correlation": 129164
    }
  },
  {
    "ph": "s", "id": 129164, "pid": 76337, "tid": -914061504, "ts": 1716454223397844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397917, "dur": 1,
    "args": {
      "External id": 129175,
      "cbid": 251, "correlation": 129175
    }
  },
  {
    "ph": "f", "id": 129175, "pid": 76337, "tid": -914061504, "ts": 1716454223397917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397921, "dur": 0,
    "args": {
      "External id": 129176,
      "cbid": 251, "correlation": 129176
    }
  },
  {
    "ph": "f", "id": 129176, "pid": 76337, "tid": -914061504, "ts": 1716454223397921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223451270, "dur": 12,
    "args": {
      "External id": 129177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129177,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129177, "pid": 5, "tid": 7, "ts": 1716454223451270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397923, "dur": 12,
    "args": {
      "External id": 129177,
      "cbid": 211, "correlation": 129177
    }
  },
  {
    "ph": "s", "id": 129177, "pid": 76337, "tid": -914061504, "ts": 1716454223397923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223451283, "dur": 6,
    "args": {
      "External id": 129179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129179,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129179, "pid": 5, "tid": 7, "ts": 1716454223451283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223397936, "dur": 6,
    "args": {
      "External id": 129179,
      "cbid": 211, "correlation": 129179
    }
  },
  {
    "ph": "s", "id": 129179, "pid": 76337, "tid": -914061504, "ts": 1716454223397936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398003, "dur": 1,
    "args": {
      "External id": 129190,
      "cbid": 251, "correlation": 129190
    }
  },
  {
    "ph": "f", "id": 129190, "pid": 76337, "tid": -914061504, "ts": 1716454223398003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398006, "dur": 0,
    "args": {
      "External id": 129191,
      "cbid": 251, "correlation": 129191
    }
  },
  {
    "ph": "f", "id": 129191, "pid": 76337, "tid": -914061504, "ts": 1716454223398006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223451290, "dur": 8,
    "args": {
      "External id": 129192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129192,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129192, "pid": 5, "tid": 7, "ts": 1716454223451290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398008, "dur": 13,
    "args": {
      "External id": 129192,
      "cbid": 211, "correlation": 129192
    }
  },
  {
    "ph": "s", "id": 129192, "pid": 76337, "tid": -914061504, "ts": 1716454223398008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223451299, "dur": 3,
    "args": {
      "External id": 129194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129194,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129194, "pid": 5, "tid": 7, "ts": 1716454223451299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398022, "dur": 5,
    "args": {
      "External id": 129194,
      "cbid": 211, "correlation": 129194
    }
  },
  {
    "ph": "s", "id": 129194, "pid": 76337, "tid": -914061504, "ts": 1716454223398022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223451304, "dur": 55,
    "args": {
      "External id": 129219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129219,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129219, "pid": 5, "tid": 7, "ts": 1716454223451304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398099, "dur": 13,
    "args": {
      "External id": 129219,
      "cbid": 211, "correlation": 129219
    }
  },
  {
    "ph": "s", "id": 129219, "pid": 76337, "tid": -914061504, "ts": 1716454223398099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398200, "dur": 2,
    "args": {
      "External id": 129237,
      "cbid": 251, "correlation": 129237
    }
  },
  {
    "ph": "f", "id": 129237, "pid": 76337, "tid": -914061504, "ts": 1716454223398200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223451360, "dur": 91,
    "args": {
      "External id": 129239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129239,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129239, "pid": 5, "tid": 7, "ts": 1716454223451360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398206, "dur": 16,
    "args": {
      "External id": 129239,
      "cbid": 211, "correlation": 129239
    }
  },
  {
    "ph": "s", "id": 129239, "pid": 76337, "tid": -914061504, "ts": 1716454223398206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223451452, "dur": 10,
    "args": {
      "External id": 129247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129247,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129247, "pid": 5, "tid": 7, "ts": 1716454223451452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398278, "dur": 12,
    "args": {
      "External id": 129247,
      "cbid": 211, "correlation": 129247
    }
  },
  {
    "ph": "s", "id": 129247, "pid": 76337, "tid": -914061504, "ts": 1716454223398278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223451463, "dur": 21,
    "args": {
      "External id": 129255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129255, "pid": 5, "tid": 7, "ts": 1716454223451463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398320, "dur": 9,
    "args": {
      "External id": 129255,
      "cbid": 211, "correlation": 129255
    }
  },
  {
    "ph": "s", "id": 129255, "pid": 76337, "tid": -914061504, "ts": 1716454223398320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223451485, "dur": 19,
    "args": {
      "External id": 129277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129277,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129277, "pid": 5, "tid": 7, "ts": 1716454223451485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398371, "dur": 10,
    "args": {
      "External id": 129277,
      "cbid": 211, "correlation": 129277
    }
  },
  {
    "ph": "s", "id": 129277, "pid": 76337, "tid": -914061504, "ts": 1716454223398371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398460, "dur": 2,
    "args": {
      "External id": 129293,
      "cbid": 251, "correlation": 129293
    }
  },
  {
    "ph": "f", "id": 129293, "pid": 76337, "tid": -914061504, "ts": 1716454223398460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398466, "dur": 0,
    "args": {
      "External id": 129295,
      "cbid": 251, "correlation": 129295
    }
  },
  {
    "ph": "f", "id": 129295, "pid": 76337, "tid": -914061504, "ts": 1716454223398466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223451505, "dur": 495,
    "args": {
      "External id": 129296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129296,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129296, "pid": 5, "tid": 7, "ts": 1716454223451505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398469, "dur": 15,
    "args": {
      "External id": 129296,
      "cbid": 211, "correlation": 129296
    }
  },
  {
    "ph": "s", "id": 129296, "pid": 76337, "tid": -914061504, "ts": 1716454223398469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223452001, "dur": 65,
    "args": {
      "External id": 129304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129304, "pid": 5, "tid": 7, "ts": 1716454223452001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398537, "dur": 12,
    "args": {
      "External id": 129304,
      "cbid": 211, "correlation": 129304
    }
  },
  {
    "ph": "s", "id": 129304, "pid": 76337, "tid": -914061504, "ts": 1716454223398537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223452067, "dur": 69,
    "args": {
      "External id": 129312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129312, "pid": 5, "tid": 7, "ts": 1716454223452067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398568, "dur": 8,
    "args": {
      "External id": 129312,
      "cbid": 211, "correlation": 129312
    }
  },
  {
    "ph": "s", "id": 129312, "pid": 76337, "tid": -914061504, "ts": 1716454223398568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398648, "dur": 1,
    "args": {
      "External id": 129328,
      "cbid": 251, "correlation": 129328
    }
  },
  {
    "ph": "f", "id": 129328, "pid": 76337, "tid": -914061504, "ts": 1716454223398648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223452139, "dur": 1,
    "args": {
      "External id": 129330,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 129330,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1388888888888889
    }
  },
  {
    "ph": "f", "id": 129330, "pid": 5, "tid": 7, "ts": 1716454223452139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398653, "dur": 13,
    "args": {
      "External id": 129330,
      "cbid": 51, "correlation": 129330
    }
  },
  {
    "ph": "s", "id": 129330, "pid": 76337, "tid": -914061504, "ts": 1716454223398653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223452143, "dur": 269,
    "args": {
      "External id": 129331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129331,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129331, "pid": 5, "tid": 7, "ts": 1716454223452143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398667, "dur": 11,
    "args": {
      "External id": 129331,
      "cbid": 211, "correlation": 129331
    }
  },
  {
    "ph": "s", "id": 129331, "pid": 76337, "tid": -914061504, "ts": 1716454223398667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223452413, "dur": 14,
    "args": {
      "External id": 129339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129339,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129339, "pid": 5, "tid": 7, "ts": 1716454223452413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398708, "dur": 11,
    "args": {
      "External id": 129339,
      "cbid": 211, "correlation": 129339
    }
  },
  {
    "ph": "s", "id": 129339, "pid": 76337, "tid": -914061504, "ts": 1716454223398708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223452428, "dur": 38,
    "args": {
      "External id": 129350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129350, "pid": 5, "tid": 7, "ts": 1716454223452428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398777, "dur": 12,
    "args": {
      "External id": 129350,
      "cbid": 211, "correlation": 129350
    }
  },
  {
    "ph": "s", "id": 129350, "pid": 76337, "tid": -914061504, "ts": 1716454223398777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398843, "dur": 0,
    "args": {
      "External id": 129362,
      "cbid": 317, "correlation": 129362
    }
  },
  {
    "ph": "f", "id": 129362, "pid": 76337, "tid": -914061504, "ts": 1716454223398843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398844, "dur": 0,
    "args": {
      "External id": 129363,
      "cbid": 203, "correlation": 129363
    }
  },
  {
    "ph": "f", "id": 129363, "pid": 76337, "tid": -914061504, "ts": 1716454223398844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398844, "dur": 0,
    "args": {
      "External id": 129364,
      "cbid": 205, "correlation": 129364
    }
  },
  {
    "ph": "f", "id": 129364, "pid": 76337, "tid": -914061504, "ts": 1716454223398844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223452467, "dur": 12,
    "args": {
      "External id": 129368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129368,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129368, "pid": 5, "tid": 7, "ts": 1716454223452467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398860, "dur": 13,
    "args": {
      "External id": 129368,
      "cbid": 211, "correlation": 129368
    }
  },
  {
    "ph": "s", "id": 129368, "pid": 76337, "tid": -914061504, "ts": 1716454223398860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223452480, "dur": 4,
    "args": {
      "External id": 129370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129370,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129370, "pid": 5, "tid": 7, "ts": 1716454223452480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398877, "dur": 6,
    "args": {
      "External id": 129370,
      "cbid": 211, "correlation": 129370
    }
  },
  {
    "ph": "s", "id": 129370, "pid": 76337, "tid": -914061504, "ts": 1716454223398877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398886, "dur": 0,
    "args": {
      "External id": 129371,
      "cbid": 51, "correlation": 129371
    }
  },
  {
    "ph": "s", "id": 129371, "pid": 76337, "tid": -914061504, "ts": 1716454223398886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223452486, "dur": 97,
    "args": {
      "External id": 129372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129372,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 129372, "pid": 5, "tid": 7, "ts": 1716454223452486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398887, "dur": 5,
    "args": {
      "External id": 129372,
      "cbid": 211, "correlation": 129372
    }
  },
  {
    "ph": "s", "id": 129372, "pid": 76337, "tid": -914061504, "ts": 1716454223398887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223452585, "dur": 16,
    "args": {
      "External id": 129377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129377,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129377, "pid": 5, "tid": 7, "ts": 1716454223452585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398914, "dur": 8,
    "args": {
      "External id": 129377,
      "cbid": 211, "correlation": 129377
    }
  },
  {
    "ph": "s", "id": 129377, "pid": 76337, "tid": -914061504, "ts": 1716454223398914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223452602, "dur": 12,
    "args": {
      "External id": 129385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129385,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129385, "pid": 5, "tid": 7, "ts": 1716454223452602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223398945, "dur": 8,
    "args": {
      "External id": 129385,
      "cbid": 211, "correlation": 129385
    }
  },
  {
    "ph": "s", "id": 129385, "pid": 76337, "tid": -914061504, "ts": 1716454223398945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223452615, "dur": 18,
    "args": {
      "External id": 129405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129405,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 129405, "pid": 5, "tid": 7, "ts": 1716454223452615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399028, "dur": 12,
    "args": {
      "External id": 129405,
      "cbid": 211, "correlation": 129405
    }
  },
  {
    "ph": "s", "id": 129405, "pid": 76337, "tid": -914061504, "ts": 1716454223399028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223452635, "dur": 5,
    "args": {
      "External id": 129417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129417,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 129417, "pid": 5, "tid": 7, "ts": 1716454223452635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399051, "dur": 6,
    "args": {
      "External id": 129417,
      "cbid": 211, "correlation": 129417
    }
  },
  {
    "ph": "s", "id": 129417, "pid": 76337, "tid": -914061504, "ts": 1716454223399051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223452641, "dur": 18,
    "args": {
      "External id": 129420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129420, "pid": 5, "tid": 7, "ts": 1716454223452641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399069, "dur": 6,
    "args": {
      "External id": 129420,
      "cbid": 211, "correlation": 129420
    }
  },
  {
    "ph": "s", "id": 129420, "pid": 76337, "tid": -914061504, "ts": 1716454223399069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223452660, "dur": 11,
    "args": {
      "External id": 129429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129429,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129429, "pid": 5, "tid": 7, "ts": 1716454223452660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399107, "dur": 11,
    "args": {
      "External id": 129429,
      "cbid": 211, "correlation": 129429
    }
  },
  {
    "ph": "s", "id": 129429, "pid": 76337, "tid": -914061504, "ts": 1716454223399107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399159, "dur": 0,
    "args": {
      "External id": 129439,
      "cbid": 317, "correlation": 129439
    }
  },
  {
    "ph": "f", "id": 129439, "pid": 76337, "tid": -914061504, "ts": 1716454223399159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399160, "dur": 0,
    "args": {
      "External id": 129440,
      "cbid": 203, "correlation": 129440
    }
  },
  {
    "ph": "f", "id": 129440, "pid": 76337, "tid": -914061504, "ts": 1716454223399160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399161, "dur": 0,
    "args": {
      "External id": 129441,
      "cbid": 205, "correlation": 129441
    }
  },
  {
    "ph": "f", "id": 129441, "pid": 76337, "tid": -914061504, "ts": 1716454223399161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223452672, "dur": 13,
    "args": {
      "External id": 129445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129445,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129445, "pid": 5, "tid": 7, "ts": 1716454223452672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399175, "dur": 11,
    "args": {
      "External id": 129445,
      "cbid": 211, "correlation": 129445
    }
  },
  {
    "ph": "s", "id": 129445, "pid": 76337, "tid": -914061504, "ts": 1716454223399175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223452687, "dur": 161,
    "args": {
      "External id": 129447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129447,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129447, "pid": 5, "tid": 7, "ts": 1716454223452687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399189, "dur": 5,
    "args": {
      "External id": 129447,
      "cbid": 211, "correlation": 129447
    }
  },
  {
    "ph": "s", "id": 129447, "pid": 76337, "tid": -914061504, "ts": 1716454223399189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223452850, "dur": 1,
    "args": {
      "External id": 129449,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 129449,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 129449, "pid": 5, "tid": 7, "ts": 1716454223452850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399200, "dur": 7,
    "args": {
      "External id": 129449,
      "cbid": 51, "correlation": 129449
    }
  },
  {
    "ph": "s", "id": 129449, "pid": 76337, "tid": -914061504, "ts": 1716454223399200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223452853, "dur": 660,
    "args": {
      "External id": 129450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129450,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129450, "pid": 5, "tid": 7, "ts": 1716454223452853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399207, "dur": 7,
    "args": {
      "External id": 129450,
      "cbid": 211, "correlation": 129450
    }
  },
  {
    "ph": "s", "id": 129450, "pid": 76337, "tid": -914061504, "ts": 1716454223399207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223453514, "dur": 13,
    "args": {
      "External id": 129452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129452, "pid": 5, "tid": 7, "ts": 1716454223453514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399218, "dur": 5,
    "args": {
      "External id": 129452,
      "cbid": 211, "correlation": 129452
    }
  },
  {
    "ph": "s", "id": 129452, "pid": 76337, "tid": -914061504, "ts": 1716454223399218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223453529, "dur": 15,
    "args": {
      "External id": 129458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129458, "pid": 5, "tid": 7, "ts": 1716454223453529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399246, "dur": 8,
    "args": {
      "External id": 129458,
      "cbid": 211, "correlation": 129458
    }
  },
  {
    "ph": "s", "id": 129458, "pid": 76337, "tid": -914061504, "ts": 1716454223399246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223453545, "dur": 3,
    "args": {
      "External id": 129466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129466,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 129466, "pid": 5, "tid": 7, "ts": 1716454223453545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399290, "dur": 10,
    "args": {
      "External id": 129466,
      "cbid": 211, "correlation": 129466
    }
  },
  {
    "ph": "s", "id": 129466, "pid": 76337, "tid": -914061504, "ts": 1716454223399290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399358, "dur": 1,
    "args": {
      "External id": 129482,
      "cbid": 251, "correlation": 129482
    }
  },
  {
    "ph": "f", "id": 129482, "pid": 76337, "tid": -914061504, "ts": 1716454223399358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399364, "dur": 0,
    "args": {
      "External id": 129484,
      "cbid": 251, "correlation": 129484
    }
  },
  {
    "ph": "f", "id": 129484, "pid": 76337, "tid": -914061504, "ts": 1716454223399364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223453550, "dur": 13,
    "args": {
      "External id": 129485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129485,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129485, "pid": 5, "tid": 7, "ts": 1716454223453550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399366, "dur": 11,
    "args": {
      "External id": 129485,
      "cbid": 211, "correlation": 129485
    }
  },
  {
    "ph": "s", "id": 129485, "pid": 76337, "tid": -914061504, "ts": 1716454223399366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223453565, "dur": 5,
    "args": {
      "External id": 129487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129487,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129487, "pid": 5, "tid": 7, "ts": 1716454223453565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399379, "dur": 5,
    "args": {
      "External id": 129487,
      "cbid": 211, "correlation": 129487
    }
  },
  {
    "ph": "s", "id": 129487, "pid": 76337, "tid": -914061504, "ts": 1716454223399379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223453571, "dur": 17,
    "args": {
      "External id": 129497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129497, "pid": 5, "tid": 7, "ts": 1716454223453571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399438, "dur": 12,
    "args": {
      "External id": 129497,
      "cbid": 211, "correlation": 129497
    }
  },
  {
    "ph": "s", "id": 129497, "pid": 76337, "tid": -914061504, "ts": 1716454223399438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223453589, "dur": 18,
    "args": {
      "External id": 129517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129517,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 129517, "pid": 5, "tid": 7, "ts": 1716454223453589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399504, "dur": 11,
    "args": {
      "External id": 129517,
      "cbid": 211, "correlation": 129517
    }
  },
  {
    "ph": "s", "id": 129517, "pid": 76337, "tid": -914061504, "ts": 1716454223399504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223453609, "dur": 4,
    "args": {
      "External id": 129529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129529,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 129529, "pid": 5, "tid": 7, "ts": 1716454223453609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399525, "dur": 6,
    "args": {
      "External id": 129529,
      "cbid": 211, "correlation": 129529
    }
  },
  {
    "ph": "s", "id": 129529, "pid": 76337, "tid": -914061504, "ts": 1716454223399525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223453614, "dur": 16,
    "args": {
      "External id": 129532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129532, "pid": 5, "tid": 7, "ts": 1716454223453614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399544, "dur": 6,
    "args": {
      "External id": 129532,
      "cbid": 211, "correlation": 129532
    }
  },
  {
    "ph": "s", "id": 129532, "pid": 76337, "tid": -914061504, "ts": 1716454223399544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223453632, "dur": 11,
    "args": {
      "External id": 129541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129541,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129541, "pid": 5, "tid": 7, "ts": 1716454223453632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399584, "dur": 10,
    "args": {
      "External id": 129541,
      "cbid": 211, "correlation": 129541
    }
  },
  {
    "ph": "s", "id": 129541, "pid": 76337, "tid": -914061504, "ts": 1716454223399584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399646, "dur": 0,
    "args": {
      "External id": 129551,
      "cbid": 317, "correlation": 129551
    }
  },
  {
    "ph": "f", "id": 129551, "pid": 76337, "tid": -914061504, "ts": 1716454223399646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399646, "dur": 0,
    "args": {
      "External id": 129552,
      "cbid": 203, "correlation": 129552
    }
  },
  {
    "ph": "f", "id": 129552, "pid": 76337, "tid": -914061504, "ts": 1716454223399646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399647, "dur": 0,
    "args": {
      "External id": 129553,
      "cbid": 205, "correlation": 129553
    }
  },
  {
    "ph": "f", "id": 129553, "pid": 76337, "tid": -914061504, "ts": 1716454223399647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223453645, "dur": 11,
    "args": {
      "External id": 129557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129557,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129557, "pid": 5, "tid": 7, "ts": 1716454223453645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399662, "dur": 12,
    "args": {
      "External id": 129557,
      "cbid": 211, "correlation": 129557
    }
  },
  {
    "ph": "s", "id": 129557, "pid": 76337, "tid": -914061504, "ts": 1716454223399662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223453656, "dur": 162,
    "args": {
      "External id": 129559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129559, "pid": 5, "tid": 7, "ts": 1716454223453656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399676, "dur": 5,
    "args": {
      "External id": 129559,
      "cbid": 211, "correlation": 129559
    }
  },
  {
    "ph": "s", "id": 129559, "pid": 76337, "tid": -914061504, "ts": 1716454223399676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223453820, "dur": 1,
    "args": {
      "External id": 129561,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 129561,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 129561, "pid": 5, "tid": 7, "ts": 1716454223453820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399687, "dur": 6,
    "args": {
      "External id": 129561,
      "cbid": 51, "correlation": 129561
    }
  },
  {
    "ph": "s", "id": 129561, "pid": 76337, "tid": -914061504, "ts": 1716454223399687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223453824, "dur": 648,
    "args": {
      "External id": 129562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129562,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129562, "pid": 5, "tid": 7, "ts": 1716454223453824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399694, "dur": 6,
    "args": {
      "External id": 129562,
      "cbid": 211, "correlation": 129562
    }
  },
  {
    "ph": "s", "id": 129562, "pid": 76337, "tid": -914061504, "ts": 1716454223399694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223454473, "dur": 13,
    "args": {
      "External id": 129564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129564,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129564, "pid": 5, "tid": 7, "ts": 1716454223454473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399704, "dur": 5,
    "args": {
      "External id": 129564,
      "cbid": 211, "correlation": 129564
    }
  },
  {
    "ph": "s", "id": 129564, "pid": 76337, "tid": -914061504, "ts": 1716454223399704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223454487, "dur": 15,
    "args": {
      "External id": 129570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129570, "pid": 5, "tid": 7, "ts": 1716454223454487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399732, "dur": 9,
    "args": {
      "External id": 129570,
      "cbid": 211, "correlation": 129570
    }
  },
  {
    "ph": "s", "id": 129570, "pid": 76337, "tid": -914061504, "ts": 1716454223399732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223454503, "dur": 12,
    "args": {
      "External id": 129578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129578,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129578, "pid": 5, "tid": 7, "ts": 1716454223454503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399765, "dur": 8,
    "args": {
      "External id": 129578,
      "cbid": 211, "correlation": 129578
    }
  },
  {
    "ph": "s", "id": 129578, "pid": 76337, "tid": -914061504, "ts": 1716454223399765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223454516, "dur": 10,
    "args": {
      "External id": 129586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129586,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129586, "pid": 5, "tid": 7, "ts": 1716454223454516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399793, "dur": 8,
    "args": {
      "External id": 129586,
      "cbid": 211, "correlation": 129586
    }
  },
  {
    "ph": "s", "id": 129586, "pid": 76337, "tid": -914061504, "ts": 1716454223399793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223454527, "dur": 19,
    "args": {
      "External id": 129606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129606,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 129606, "pid": 5, "tid": 7, "ts": 1716454223454527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399873, "dur": 12,
    "args": {
      "External id": 129606,
      "cbid": 211, "correlation": 129606
    }
  },
  {
    "ph": "s", "id": 129606, "pid": 76337, "tid": -914061504, "ts": 1716454223399873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223454547, "dur": 4,
    "args": {
      "External id": 129618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129618,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 129618, "pid": 5, "tid": 7, "ts": 1716454223454547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399895, "dur": 6,
    "args": {
      "External id": 129618,
      "cbid": 211, "correlation": 129618
    }
  },
  {
    "ph": "s", "id": 129618, "pid": 76337, "tid": -914061504, "ts": 1716454223399895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223454552, "dur": 17,
    "args": {
      "External id": 129621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129621, "pid": 5, "tid": 7, "ts": 1716454223454552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399913, "dur": 7,
    "args": {
      "External id": 129621,
      "cbid": 211, "correlation": 129621
    }
  },
  {
    "ph": "s", "id": 129621, "pid": 76337, "tid": -914061504, "ts": 1716454223399913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399969, "dur": 0,
    "args": {
      "External id": 129632,
      "cbid": 317, "correlation": 129632
    }
  },
  {
    "ph": "f", "id": 129632, "pid": 76337, "tid": -914061504, "ts": 1716454223399969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399970, "dur": 0,
    "args": {
      "External id": 129633,
      "cbid": 203, "correlation": 129633
    }
  },
  {
    "ph": "f", "id": 129633, "pid": 76337, "tid": -914061504, "ts": 1716454223399970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399971, "dur": 0,
    "args": {
      "External id": 129634,
      "cbid": 205, "correlation": 129634
    }
  },
  {
    "ph": "f", "id": 129634, "pid": 76337, "tid": -914061504, "ts": 1716454223399971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223454571, "dur": 11,
    "args": {
      "External id": 129638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129638,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129638, "pid": 5, "tid": 7, "ts": 1716454223454571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223399992, "dur": 12,
    "args": {
      "External id": 129638,
      "cbid": 211, "correlation": 129638
    }
  },
  {
    "ph": "s", "id": 129638, "pid": 76337, "tid": -914061504, "ts": 1716454223399992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223454583, "dur": 4,
    "args": {
      "External id": 129640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129640, "pid": 5, "tid": 7, "ts": 1716454223454583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400009, "dur": 6,
    "args": {
      "External id": 129640,
      "cbid": 211, "correlation": 129640
    }
  },
  {
    "ph": "s", "id": 129640, "pid": 76337, "tid": -914061504, "ts": 1716454223400009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400018, "dur": 0,
    "args": {
      "External id": 129641,
      "cbid": 51, "correlation": 129641
    }
  },
  {
    "ph": "s", "id": 129641, "pid": 76337, "tid": -914061504, "ts": 1716454223400018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223454588, "dur": 94,
    "args": {
      "External id": 129642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129642,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 129642, "pid": 5, "tid": 7, "ts": 1716454223454588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400019, "dur": 5,
    "args": {
      "External id": 129642,
      "cbid": 211, "correlation": 129642
    }
  },
  {
    "ph": "s", "id": 129642, "pid": 76337, "tid": -914061504, "ts": 1716454223400019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223454684, "dur": 15,
    "args": {
      "External id": 129647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129647, "pid": 5, "tid": 7, "ts": 1716454223454684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400046, "dur": 8,
    "args": {
      "External id": 129647,
      "cbid": 211, "correlation": 129647
    }
  },
  {
    "ph": "s", "id": 129647, "pid": 76337, "tid": -914061504, "ts": 1716454223400046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223454700, "dur": 83,
    "args": {
      "External id": 129656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129656,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129656, "pid": 5, "tid": 7, "ts": 1716454223454700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400127, "dur": 14,
    "args": {
      "External id": 129656,
      "cbid": 211, "correlation": 129656
    }
  },
  {
    "ph": "s", "id": 129656, "pid": 76337, "tid": -914061504, "ts": 1716454223400127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223454785, "dur": 30,
    "args": {
      "External id": 129678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129678,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129678, "pid": 5, "tid": 7, "ts": 1716454223454785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400183, "dur": 10,
    "args": {
      "External id": 129678,
      "cbid": 211, "correlation": 129678
    }
  },
  {
    "ph": "s", "id": 129678, "pid": 76337, "tid": -914061504, "ts": 1716454223400183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400271, "dur": 1,
    "args": {
      "External id": 129689,
      "cbid": 251, "correlation": 129689
    }
  },
  {
    "ph": "f", "id": 129689, "pid": 76337, "tid": -914061504, "ts": 1716454223400271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223454816, "dur": 162,
    "args": {
      "External id": 129690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129690,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129690, "pid": 5, "tid": 7, "ts": 1716454223454816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400276, "dur": 13,
    "args": {
      "External id": 129690,
      "cbid": 211, "correlation": 129690
    }
  },
  {
    "ph": "s", "id": 129690, "pid": 76337, "tid": -914061504, "ts": 1716454223400276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400345, "dur": 1,
    "args": {
      "External id": 129701,
      "cbid": 251, "correlation": 129701
    }
  },
  {
    "ph": "f", "id": 129701, "pid": 76337, "tid": -914061504, "ts": 1716454223400345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223454979, "dur": 160,
    "args": {
      "External id": 129702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129702,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129702, "pid": 5, "tid": 7, "ts": 1716454223454979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400349, "dur": 11,
    "args": {
      "External id": 129702,
      "cbid": 211, "correlation": 129702
    }
  },
  {
    "ph": "s", "id": 129702, "pid": 76337, "tid": -914061504, "ts": 1716454223400349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400415, "dur": 1,
    "args": {
      "External id": 129713,
      "cbid": 251, "correlation": 129713
    }
  },
  {
    "ph": "f", "id": 129713, "pid": 76337, "tid": -914061504, "ts": 1716454223400415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223455140, "dur": 158,
    "args": {
      "External id": 129714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129714,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129714, "pid": 5, "tid": 7, "ts": 1716454223455140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400419, "dur": 11,
    "args": {
      "External id": 129714,
      "cbid": 211, "correlation": 129714
    }
  },
  {
    "ph": "s", "id": 129714, "pid": 76337, "tid": -914061504, "ts": 1716454223400419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223455299, "dur": 334,
    "args": {
      "External id": 129739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129739,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129739, "pid": 5, "tid": 7, "ts": 1716454223455299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400501, "dur": 13,
    "args": {
      "External id": 129739,
      "cbid": 211, "correlation": 129739
    }
  },
  {
    "ph": "s", "id": 129739, "pid": 76337, "tid": -914061504, "ts": 1716454223400501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400599, "dur": 1,
    "args": {
      "External id": 129757,
      "cbid": 251, "correlation": 129757
    }
  },
  {
    "ph": "f", "id": 129757, "pid": 76337, "tid": -914061504, "ts": 1716454223400599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223455635, "dur": 167,
    "args": {
      "External id": 129759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129759,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129759, "pid": 5, "tid": 7, "ts": 1716454223455635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400605, "dur": 14,
    "args": {
      "External id": 129759,
      "cbid": 211, "correlation": 129759
    }
  },
  {
    "ph": "s", "id": 129759, "pid": 76337, "tid": -914061504, "ts": 1716454223400605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223455803, "dur": 19,
    "args": {
      "External id": 129767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129767,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129767, "pid": 5, "tid": 7, "ts": 1716454223455803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400675, "dur": 12,
    "args": {
      "External id": 129767,
      "cbid": 211, "correlation": 129767
    }
  },
  {
    "ph": "s", "id": 129767, "pid": 76337, "tid": -914061504, "ts": 1716454223400675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223455823, "dur": 27,
    "args": {
      "External id": 129775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129775,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129775, "pid": 5, "tid": 7, "ts": 1716454223455823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400713, "dur": 8,
    "args": {
      "External id": 129775,
      "cbid": 211, "correlation": 129775
    }
  },
  {
    "ph": "s", "id": 129775, "pid": 76337, "tid": -914061504, "ts": 1716454223400713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223455852, "dur": 18,
    "args": {
      "External id": 129786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129786, "pid": 5, "tid": 7, "ts": 1716454223455852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400782, "dur": 13,
    "args": {
      "External id": 129786,
      "cbid": 211, "correlation": 129786
    }
  },
  {
    "ph": "s", "id": 129786, "pid": 76337, "tid": -914061504, "ts": 1716454223400782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223455871, "dur": 17,
    "args": {
      "External id": 129808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129808,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129808, "pid": 5, "tid": 7, "ts": 1716454223455871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400814, "dur": 8,
    "args": {
      "External id": 129808,
      "cbid": 211, "correlation": 129808
    }
  },
  {
    "ph": "s", "id": 129808, "pid": 76337, "tid": -914061504, "ts": 1716454223400814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400897, "dur": 1,
    "args": {
      "External id": 129819,
      "cbid": 251, "correlation": 129819
    }
  },
  {
    "ph": "f", "id": 129819, "pid": 76337, "tid": -914061504, "ts": 1716454223400897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223455889, "dur": 88,
    "args": {
      "External id": 129820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129820,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129820, "pid": 5, "tid": 7, "ts": 1716454223455889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400902, "dur": 13,
    "args": {
      "External id": 129820,
      "cbid": 211, "correlation": 129820
    }
  },
  {
    "ph": "s", "id": 129820, "pid": 76337, "tid": -914061504, "ts": 1716454223400902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400970, "dur": 1,
    "args": {
      "External id": 129831,
      "cbid": 251, "correlation": 129831
    }
  },
  {
    "ph": "f", "id": 129831, "pid": 76337, "tid": -914061504, "ts": 1716454223400970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400982, "dur": 0,
    "args": {
      "External id": 129832,
      "cbid": 251, "correlation": 129832
    }
  },
  {
    "ph": "f", "id": 129832, "pid": 76337, "tid": -914061504, "ts": 1716454223400982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223455979, "dur": 12,
    "args": {
      "External id": 129833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129833,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129833, "pid": 5, "tid": 7, "ts": 1716454223455979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400984, "dur": 13,
    "args": {
      "External id": 129833,
      "cbid": 211, "correlation": 129833
    }
  },
  {
    "ph": "s", "id": 129833, "pid": 76337, "tid": -914061504, "ts": 1716454223400984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223455993, "dur": 6,
    "args": {
      "External id": 129835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129835,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129835, "pid": 5, "tid": 7, "ts": 1716454223455993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223400999, "dur": 6,
    "args": {
      "External id": 129835,
      "cbid": 211, "correlation": 129835
    }
  },
  {
    "ph": "s", "id": 129835, "pid": 76337, "tid": -914061504, "ts": 1716454223400999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401059, "dur": 1,
    "args": {
      "External id": 129846,
      "cbid": 251, "correlation": 129846
    }
  },
  {
    "ph": "f", "id": 129846, "pid": 76337, "tid": -914061504, "ts": 1716454223401059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401062, "dur": 0,
    "args": {
      "External id": 129847,
      "cbid": 251, "correlation": 129847
    }
  },
  {
    "ph": "f", "id": 129847, "pid": 76337, "tid": -914061504, "ts": 1716454223401062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223456000, "dur": 8,
    "args": {
      "External id": 129848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129848,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129848, "pid": 5, "tid": 7, "ts": 1716454223456000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401064, "dur": 11,
    "args": {
      "External id": 129848,
      "cbid": 211, "correlation": 129848
    }
  },
  {
    "ph": "s", "id": 129848, "pid": 76337, "tid": -914061504, "ts": 1716454223401064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223456009, "dur": 3,
    "args": {
      "External id": 129850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129850,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129850, "pid": 5, "tid": 7, "ts": 1716454223456009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401077, "dur": 6,
    "args": {
      "External id": 129850,
      "cbid": 211, "correlation": 129850
    }
  },
  {
    "ph": "s", "id": 129850, "pid": 76337, "tid": -914061504, "ts": 1716454223401077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223456013, "dur": 55,
    "args": {
      "External id": 129875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129875,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129875, "pid": 5, "tid": 7, "ts": 1716454223456013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401153, "dur": 13,
    "args": {
      "External id": 129875,
      "cbid": 211, "correlation": 129875
    }
  },
  {
    "ph": "s", "id": 129875, "pid": 76337, "tid": -914061504, "ts": 1716454223401153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401251, "dur": 1,
    "args": {
      "External id": 129893,
      "cbid": 251, "correlation": 129893
    }
  },
  {
    "ph": "f", "id": 129893, "pid": 76337, "tid": -914061504, "ts": 1716454223401251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223456069, "dur": 89,
    "args": {
      "External id": 129895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129895,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129895, "pid": 5, "tid": 7, "ts": 1716454223456069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401257, "dur": 15,
    "args": {
      "External id": 129895,
      "cbid": 211, "correlation": 129895
    }
  },
  {
    "ph": "s", "id": 129895, "pid": 76337, "tid": -914061504, "ts": 1716454223401257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223456160, "dur": 10,
    "args": {
      "External id": 129903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129903,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129903, "pid": 5, "tid": 7, "ts": 1716454223456160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401327, "dur": 12,
    "args": {
      "External id": 129903,
      "cbid": 211, "correlation": 129903
    }
  },
  {
    "ph": "s", "id": 129903, "pid": 76337, "tid": -914061504, "ts": 1716454223401327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223456171, "dur": 20,
    "args": {
      "External id": 129911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129911,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129911, "pid": 5, "tid": 7, "ts": 1716454223456171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401368, "dur": 9,
    "args": {
      "External id": 129911,
      "cbid": 211, "correlation": 129911
    }
  },
  {
    "ph": "s", "id": 129911, "pid": 76337, "tid": -914061504, "ts": 1716454223401368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223456193, "dur": 17,
    "args": {
      "External id": 129933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129933,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129933, "pid": 5, "tid": 7, "ts": 1716454223456193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401419, "dur": 10,
    "args": {
      "External id": 129933,
      "cbid": 211, "correlation": 129933
    }
  },
  {
    "ph": "s", "id": 129933, "pid": 76337, "tid": -914061504, "ts": 1716454223401419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401505, "dur": 1,
    "args": {
      "External id": 129949,
      "cbid": 251, "correlation": 129949
    }
  },
  {
    "ph": "f", "id": 129949, "pid": 76337, "tid": -914061504, "ts": 1716454223401505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401510, "dur": 0,
    "args": {
      "External id": 129951,
      "cbid": 251, "correlation": 129951
    }
  },
  {
    "ph": "f", "id": 129951, "pid": 76337, "tid": -914061504, "ts": 1716454223401510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223456211, "dur": 495,
    "args": {
      "External id": 129952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129952,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 129952, "pid": 5, "tid": 7, "ts": 1716454223456211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401512, "dur": 13,
    "args": {
      "External id": 129952,
      "cbid": 211, "correlation": 129952
    }
  },
  {
    "ph": "s", "id": 129952, "pid": 76337, "tid": -914061504, "ts": 1716454223401512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223456707, "dur": 65,
    "args": {
      "External id": 129960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129960, "pid": 5, "tid": 7, "ts": 1716454223456707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401577, "dur": 12,
    "args": {
      "External id": 129960,
      "cbid": 211, "correlation": 129960
    }
  },
  {
    "ph": "s", "id": 129960, "pid": 76337, "tid": -914061504, "ts": 1716454223401577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223456774, "dur": 68,
    "args": {
      "External id": 129968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129968, "pid": 5, "tid": 7, "ts": 1716454223456774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401606, "dur": 9,
    "args": {
      "External id": 129968,
      "cbid": 211, "correlation": 129968
    }
  },
  {
    "ph": "s", "id": 129968, "pid": 76337, "tid": -914061504, "ts": 1716454223401606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401686, "dur": 1,
    "args": {
      "External id": 129984,
      "cbid": 251, "correlation": 129984
    }
  },
  {
    "ph": "f", "id": 129984, "pid": 76337, "tid": -914061504, "ts": 1716454223401686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223456845, "dur": 1,
    "args": {
      "External id": 129986,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 129986,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 129986, "pid": 5, "tid": 7, "ts": 1716454223456845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401691, "dur": 9,
    "args": {
      "External id": 129986,
      "cbid": 51, "correlation": 129986
    }
  },
  {
    "ph": "s", "id": 129986, "pid": 76337, "tid": -914061504, "ts": 1716454223401691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223456848, "dur": 269,
    "args": {
      "External id": 129987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129987,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 129987, "pid": 5, "tid": 7, "ts": 1716454223456848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401701, "dur": 11,
    "args": {
      "External id": 129987,
      "cbid": 211, "correlation": 129987
    }
  },
  {
    "ph": "s", "id": 129987, "pid": 76337, "tid": -914061504, "ts": 1716454223401701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223457119, "dur": 14,
    "args": {
      "External id": 129995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 129995,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 129995, "pid": 5, "tid": 7, "ts": 1716454223457119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401744, "dur": 10,
    "args": {
      "External id": 129995,
      "cbid": 211, "correlation": 129995
    }
  },
  {
    "ph": "s", "id": 129995, "pid": 76337, "tid": -914061504, "ts": 1716454223401744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223457134, "dur": 37,
    "args": {
      "External id": 130006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130006, "pid": 5, "tid": 7, "ts": 1716454223457134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401811, "dur": 13,
    "args": {
      "External id": 130006,
      "cbid": 211, "correlation": 130006
    }
  },
  {
    "ph": "s", "id": 130006, "pid": 76337, "tid": -914061504, "ts": 1716454223401811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401875, "dur": 0,
    "args": {
      "External id": 130018,
      "cbid": 317, "correlation": 130018
    }
  },
  {
    "ph": "f", "id": 130018, "pid": 76337, "tid": -914061504, "ts": 1716454223401875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401876, "dur": 0,
    "args": {
      "External id": 130019,
      "cbid": 203, "correlation": 130019
    }
  },
  {
    "ph": "f", "id": 130019, "pid": 76337, "tid": -914061504, "ts": 1716454223401876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401877, "dur": 0,
    "args": {
      "External id": 130020,
      "cbid": 205, "correlation": 130020
    }
  },
  {
    "ph": "f", "id": 130020, "pid": 76337, "tid": -914061504, "ts": 1716454223401877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457172, "dur": 13,
    "args": {
      "External id": 130024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130024,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130024, "pid": 5, "tid": 7, "ts": 1716454223457172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401891, "dur": 12,
    "args": {
      "External id": 130024,
      "cbid": 211, "correlation": 130024
    }
  },
  {
    "ph": "s", "id": 130024, "pid": 76337, "tid": -914061504, "ts": 1716454223401891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223457187, "dur": 4,
    "args": {
      "External id": 130026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 130026, "pid": 5, "tid": 7, "ts": 1716454223457187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401907, "dur": 6,
    "args": {
      "External id": 130026,
      "cbid": 211, "correlation": 130026
    }
  },
  {
    "ph": "s", "id": 130026, "pid": 76337, "tid": -914061504, "ts": 1716454223401907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401917, "dur": 0,
    "args": {
      "External id": 130027,
      "cbid": 51, "correlation": 130027
    }
  },
  {
    "ph": "s", "id": 130027, "pid": 76337, "tid": -914061504, "ts": 1716454223401917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223457192, "dur": 97,
    "args": {
      "External id": 130028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130028,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 130028, "pid": 5, "tid": 7, "ts": 1716454223457192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401917, "dur": 5,
    "args": {
      "External id": 130028,
      "cbid": 211, "correlation": 130028
    }
  },
  {
    "ph": "s", "id": 130028, "pid": 76337, "tid": -914061504, "ts": 1716454223401917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223457291, "dur": 17,
    "args": {
      "External id": 130033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130033, "pid": 5, "tid": 7, "ts": 1716454223457291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401944, "dur": 9,
    "args": {
      "External id": 130033,
      "cbid": 211, "correlation": 130033
    }
  },
  {
    "ph": "s", "id": 130033, "pid": 76337, "tid": -914061504, "ts": 1716454223401944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223457309, "dur": 12,
    "args": {
      "External id": 130041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130041,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130041, "pid": 5, "tid": 7, "ts": 1716454223457309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223401985, "dur": 9,
    "args": {
      "External id": 130041,
      "cbid": 211, "correlation": 130041
    }
  },
  {
    "ph": "s", "id": 130041, "pid": 76337, "tid": -914061504, "ts": 1716454223401985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402055, "dur": 0,
    "args": {
      "External id": 130051,
      "cbid": 317, "correlation": 130051
    }
  },
  {
    "ph": "f", "id": 130051, "pid": 76337, "tid": -914061504, "ts": 1716454223402055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402056, "dur": 0,
    "args": {
      "External id": 130052,
      "cbid": 203, "correlation": 130052
    }
  },
  {
    "ph": "f", "id": 130052, "pid": 76337, "tid": -914061504, "ts": 1716454223402056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402056, "dur": 0,
    "args": {
      "External id": 130053,
      "cbid": 205, "correlation": 130053
    }
  },
  {
    "ph": "f", "id": 130053, "pid": 76337, "tid": -914061504, "ts": 1716454223402056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457322, "dur": 12,
    "args": {
      "External id": 130057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130057,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130057, "pid": 5, "tid": 7, "ts": 1716454223457322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402071, "dur": 12,
    "args": {
      "External id": 130057,
      "cbid": 211, "correlation": 130057
    }
  },
  {
    "ph": "s", "id": 130057, "pid": 76337, "tid": -914061504, "ts": 1716454223402071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457335, "dur": 162,
    "args": {
      "External id": 130059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130059,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130059, "pid": 5, "tid": 7, "ts": 1716454223457335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402086, "dur": 5,
    "args": {
      "External id": 130059,
      "cbid": 211, "correlation": 130059
    }
  },
  {
    "ph": "s", "id": 130059, "pid": 76337, "tid": -914061504, "ts": 1716454223402086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223457500, "dur": 1,
    "args": {
      "External id": 130061,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130061,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 130061, "pid": 5, "tid": 7, "ts": 1716454223457500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402098, "dur": 6,
    "args": {
      "External id": 130061,
      "cbid": 51, "correlation": 130061
    }
  },
  {
    "ph": "s", "id": 130061, "pid": 76337, "tid": -914061504, "ts": 1716454223402098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223457504, "dur": 197,
    "args": {
      "External id": 130062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130062,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 130062, "pid": 5, "tid": 7, "ts": 1716454223457504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402105, "dur": 9,
    "args": {
      "External id": 130062,
      "cbid": 211, "correlation": 130062
    }
  },
  {
    "ph": "s", "id": 130062, "pid": 76337, "tid": -914061504, "ts": 1716454223402105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457702, "dur": 6,
    "args": {
      "External id": 130064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130064,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130064, "pid": 5, "tid": 7, "ts": 1716454223457702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402120, "dur": 5,
    "args": {
      "External id": 130064,
      "cbid": 211, "correlation": 130064
    }
  },
  {
    "ph": "s", "id": 130064, "pid": 76337, "tid": -914061504, "ts": 1716454223402120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223457709, "dur": 6,
    "args": {
      "External id": 130070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130070,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130070, "pid": 5, "tid": 7, "ts": 1716454223457709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402147, "dur": 8,
    "args": {
      "External id": 130070,
      "cbid": 211, "correlation": 130070
    }
  },
  {
    "ph": "s", "id": 130070, "pid": 76337, "tid": -914061504, "ts": 1716454223402147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457716, "dur": 10,
    "args": {
      "External id": 130090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130090,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130090, "pid": 5, "tid": 7, "ts": 1716454223457716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402241, "dur": 12,
    "args": {
      "External id": 130090,
      "cbid": 211, "correlation": 130090
    }
  },
  {
    "ph": "s", "id": 130090, "pid": 76337, "tid": -914061504, "ts": 1716454223402241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223457728, "dur": 4,
    "args": {
      "External id": 130102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130102,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130102, "pid": 5, "tid": 7, "ts": 1716454223457728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402264, "dur": 6,
    "args": {
      "External id": 130102,
      "cbid": 211, "correlation": 130102
    }
  },
  {
    "ph": "s", "id": 130102, "pid": 76337, "tid": -914061504, "ts": 1716454223402264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223457733, "dur": 9,
    "args": {
      "External id": 130105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130105,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130105, "pid": 5, "tid": 7, "ts": 1716454223457733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402283, "dur": 6,
    "args": {
      "External id": 130105,
      "cbid": 211, "correlation": 130105
    }
  },
  {
    "ph": "s", "id": 130105, "pid": 76337, "tid": -914061504, "ts": 1716454223402283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223457743, "dur": 5,
    "args": {
      "External id": 130114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130114,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130114, "pid": 5, "tid": 7, "ts": 1716454223457743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402323, "dur": 9,
    "args": {
      "External id": 130114,
      "cbid": 211, "correlation": 130114
    }
  },
  {
    "ph": "s", "id": 130114, "pid": 76337, "tid": -914061504, "ts": 1716454223402323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402374, "dur": 0,
    "args": {
      "External id": 130124,
      "cbid": 317, "correlation": 130124
    }
  },
  {
    "ph": "f", "id": 130124, "pid": 76337, "tid": -914061504, "ts": 1716454223402374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402375, "dur": 0,
    "args": {
      "External id": 130125,
      "cbid": 203, "correlation": 130125
    }
  },
  {
    "ph": "f", "id": 130125, "pid": 76337, "tid": -914061504, "ts": 1716454223402375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402376, "dur": 0,
    "args": {
      "External id": 130126,
      "cbid": 205, "correlation": 130126
    }
  },
  {
    "ph": "f", "id": 130126, "pid": 76337, "tid": -914061504, "ts": 1716454223402376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457750, "dur": 5,
    "args": {
      "External id": 130130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130130, "pid": 5, "tid": 7, "ts": 1716454223457750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402391, "dur": 11,
    "args": {
      "External id": 130130,
      "cbid": 211, "correlation": 130130
    }
  },
  {
    "ph": "s", "id": 130130, "pid": 76337, "tid": -914061504, "ts": 1716454223402391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223457756, "dur": 161,
    "args": {
      "External id": 130132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130132, "pid": 5, "tid": 7, "ts": 1716454223457756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402405, "dur": 5,
    "args": {
      "External id": 130132,
      "cbid": 211, "correlation": 130132
    }
  },
  {
    "ph": "s", "id": 130132, "pid": 76337, "tid": -914061504, "ts": 1716454223402405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223457919, "dur": 1,
    "args": {
      "External id": 130134,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130134,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 130134, "pid": 5, "tid": 7, "ts": 1716454223457919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402416, "dur": 6,
    "args": {
      "External id": 130134,
      "cbid": 51, "correlation": 130134
    }
  },
  {
    "ph": "s", "id": 130134, "pid": 76337, "tid": -914061504, "ts": 1716454223402416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223457923, "dur": 268,
    "args": {
      "External id": 130135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130135,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130135, "pid": 5, "tid": 7, "ts": 1716454223457923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402423, "dur": 6,
    "args": {
      "External id": 130135,
      "cbid": 211, "correlation": 130135
    }
  },
  {
    "ph": "s", "id": 130135, "pid": 76337, "tid": -914061504, "ts": 1716454223402423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458192, "dur": 6,
    "args": {
      "External id": 130137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130137,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130137, "pid": 5, "tid": 7, "ts": 1716454223458192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402433, "dur": 5,
    "args": {
      "External id": 130137,
      "cbid": 211, "correlation": 130137
    }
  },
  {
    "ph": "s", "id": 130137, "pid": 76337, "tid": -914061504, "ts": 1716454223402433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223458199, "dur": 6,
    "args": {
      "External id": 130143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130143,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130143, "pid": 5, "tid": 7, "ts": 1716454223458199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402462, "dur": 9,
    "args": {
      "External id": 130143,
      "cbid": 211, "correlation": 130143
    }
  },
  {
    "ph": "s", "id": 130143, "pid": 76337, "tid": -914061504, "ts": 1716454223402462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223458207, "dur": 3,
    "args": {
      "External id": 130151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130151,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 130151, "pid": 5, "tid": 7, "ts": 1716454223458207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402505, "dur": 10,
    "args": {
      "External id": 130151,
      "cbid": 211, "correlation": 130151
    }
  },
  {
    "ph": "s", "id": 130151, "pid": 76337, "tid": -914061504, "ts": 1716454223402505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402571, "dur": 1,
    "args": {
      "External id": 130167,
      "cbid": 251, "correlation": 130167
    }
  },
  {
    "ph": "f", "id": 130167, "pid": 76337, "tid": -914061504, "ts": 1716454223402571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402576, "dur": 0,
    "args": {
      "External id": 130169,
      "cbid": 251, "correlation": 130169
    }
  },
  {
    "ph": "f", "id": 130169, "pid": 76337, "tid": -914061504, "ts": 1716454223402576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223458211, "dur": 13,
    "args": {
      "External id": 130170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130170,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130170, "pid": 5, "tid": 7, "ts": 1716454223458211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402578, "dur": 11,
    "args": {
      "External id": 130170,
      "cbid": 211, "correlation": 130170
    }
  },
  {
    "ph": "s", "id": 130170, "pid": 76337, "tid": -914061504, "ts": 1716454223402578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223458225, "dur": 5,
    "args": {
      "External id": 130172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130172,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130172, "pid": 5, "tid": 7, "ts": 1716454223458225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402591, "dur": 5,
    "args": {
      "External id": 130172,
      "cbid": 211, "correlation": 130172
    }
  },
  {
    "ph": "s", "id": 130172, "pid": 76337, "tid": -914061504, "ts": 1716454223402591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223458232, "dur": 6,
    "args": {
      "External id": 130182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130182,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130182, "pid": 5, "tid": 7, "ts": 1716454223458232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402648, "dur": 12,
    "args": {
      "External id": 130182,
      "cbid": 211, "correlation": 130182
    }
  },
  {
    "ph": "s", "id": 130182, "pid": 76337, "tid": -914061504, "ts": 1716454223402648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458239, "dur": 10,
    "args": {
      "External id": 130202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130202,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130202, "pid": 5, "tid": 7, "ts": 1716454223458239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402716, "dur": 11,
    "args": {
      "External id": 130202,
      "cbid": 211, "correlation": 130202
    }
  },
  {
    "ph": "s", "id": 130202, "pid": 76337, "tid": -914061504, "ts": 1716454223402716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223458250, "dur": 4,
    "args": {
      "External id": 130214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130214,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130214, "pid": 5, "tid": 7, "ts": 1716454223458250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402736, "dur": 6,
    "args": {
      "External id": 130214,
      "cbid": 211, "correlation": 130214
    }
  },
  {
    "ph": "s", "id": 130214, "pid": 76337, "tid": -914061504, "ts": 1716454223402736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223458255, "dur": 7,
    "args": {
      "External id": 130217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130217,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130217, "pid": 5, "tid": 7, "ts": 1716454223458255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402756, "dur": 7,
    "args": {
      "External id": 130217,
      "cbid": 211, "correlation": 130217
    }
  },
  {
    "ph": "s", "id": 130217, "pid": 76337, "tid": -914061504, "ts": 1716454223402756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223458263, "dur": 5,
    "args": {
      "External id": 130226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130226,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130226, "pid": 5, "tid": 7, "ts": 1716454223458263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402796, "dur": 10,
    "args": {
      "External id": 130226,
      "cbid": 211, "correlation": 130226
    }
  },
  {
    "ph": "s", "id": 130226, "pid": 76337, "tid": -914061504, "ts": 1716454223402796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402860, "dur": 0,
    "args": {
      "External id": 130236,
      "cbid": 317, "correlation": 130236
    }
  },
  {
    "ph": "f", "id": 130236, "pid": 76337, "tid": -914061504, "ts": 1716454223402860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402860, "dur": 0,
    "args": {
      "External id": 130237,
      "cbid": 203, "correlation": 130237
    }
  },
  {
    "ph": "f", "id": 130237, "pid": 76337, "tid": -914061504, "ts": 1716454223402860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402861, "dur": 0,
    "args": {
      "External id": 130238,
      "cbid": 205, "correlation": 130238
    }
  },
  {
    "ph": "f", "id": 130238, "pid": 76337, "tid": -914061504, "ts": 1716454223402861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458269, "dur": 5,
    "args": {
      "External id": 130242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130242,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130242, "pid": 5, "tid": 7, "ts": 1716454223458269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402876, "dur": 12,
    "args": {
      "External id": 130242,
      "cbid": 211, "correlation": 130242
    }
  },
  {
    "ph": "s", "id": 130242, "pid": 76337, "tid": -914061504, "ts": 1716454223402876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458276, "dur": 162,
    "args": {
      "External id": 130244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130244,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130244, "pid": 5, "tid": 7, "ts": 1716454223458276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402890, "dur": 5,
    "args": {
      "External id": 130244,
      "cbid": 211, "correlation": 130244
    }
  },
  {
    "ph": "s", "id": 130244, "pid": 76337, "tid": -914061504, "ts": 1716454223402890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223458440, "dur": 1,
    "args": {
      "External id": 130246,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130246,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 130246, "pid": 5, "tid": 7, "ts": 1716454223458440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402900, "dur": 6,
    "args": {
      "External id": 130246,
      "cbid": 51, "correlation": 130246
    }
  },
  {
    "ph": "s", "id": 130246, "pid": 76337, "tid": -914061504, "ts": 1716454223402900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223458443, "dur": 259,
    "args": {
      "External id": 130247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130247,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130247, "pid": 5, "tid": 7, "ts": 1716454223458443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402908, "dur": 7,
    "args": {
      "External id": 130247,
      "cbid": 211, "correlation": 130247
    }
  },
  {
    "ph": "s", "id": 130247, "pid": 76337, "tid": -914061504, "ts": 1716454223402908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458704, "dur": 6,
    "args": {
      "External id": 130249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130249,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130249, "pid": 5, "tid": 7, "ts": 1716454223458704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402919, "dur": 5,
    "args": {
      "External id": 130249,
      "cbid": 211, "correlation": 130249
    }
  },
  {
    "ph": "s", "id": 130249, "pid": 76337, "tid": -914061504, "ts": 1716454223402919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223458711, "dur": 6,
    "args": {
      "External id": 130255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130255, "pid": 5, "tid": 7, "ts": 1716454223458711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402947, "dur": 9,
    "args": {
      "External id": 130255,
      "cbid": 211, "correlation": 130255
    }
  },
  {
    "ph": "s", "id": 130255, "pid": 76337, "tid": -914061504, "ts": 1716454223402947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223458718, "dur": 5,
    "args": {
      "External id": 130263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130263,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130263, "pid": 5, "tid": 7, "ts": 1716454223458718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223402988, "dur": 9,
    "args": {
      "External id": 130263,
      "cbid": 211, "correlation": 130263
    }
  },
  {
    "ph": "s", "id": 130263, "pid": 76337, "tid": -914061504, "ts": 1716454223402988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223458725, "dur": 4,
    "args": {
      "External id": 130271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130271,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130271, "pid": 5, "tid": 7, "ts": 1716454223458725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403020, "dur": 8,
    "args": {
      "External id": 130271,
      "cbid": 211, "correlation": 130271
    }
  },
  {
    "ph": "s", "id": 130271, "pid": 76337, "tid": -914061504, "ts": 1716454223403020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458731, "dur": 9,
    "args": {
      "External id": 130291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130291,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130291, "pid": 5, "tid": 7, "ts": 1716454223458731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403094, "dur": 12,
    "args": {
      "External id": 130291,
      "cbid": 211, "correlation": 130291
    }
  },
  {
    "ph": "s", "id": 130291, "pid": 76337, "tid": -914061504, "ts": 1716454223403094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223458741, "dur": 4,
    "args": {
      "External id": 130303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130303,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130303, "pid": 5, "tid": 7, "ts": 1716454223458741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403117, "dur": 6,
    "args": {
      "External id": 130303,
      "cbid": 211, "correlation": 130303
    }
  },
  {
    "ph": "s", "id": 130303, "pid": 76337, "tid": -914061504, "ts": 1716454223403117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223458746, "dur": 7,
    "args": {
      "External id": 130306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130306, "pid": 5, "tid": 7, "ts": 1716454223458746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403134, "dur": 7,
    "args": {
      "External id": 130306,
      "cbid": 211, "correlation": 130306
    }
  },
  {
    "ph": "s", "id": 130306, "pid": 76337, "tid": -914061504, "ts": 1716454223403134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223458754, "dur": 5,
    "args": {
      "External id": 130315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130315,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130315, "pid": 5, "tid": 7, "ts": 1716454223458754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403172, "dur": 9,
    "args": {
      "External id": 130315,
      "cbid": 211, "correlation": 130315
    }
  },
  {
    "ph": "s", "id": 130315, "pid": 76337, "tid": -914061504, "ts": 1716454223403172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403223, "dur": 0,
    "args": {
      "External id": 130325,
      "cbid": 317, "correlation": 130325
    }
  },
  {
    "ph": "f", "id": 130325, "pid": 76337, "tid": -914061504, "ts": 1716454223403223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403224, "dur": 0,
    "args": {
      "External id": 130326,
      "cbid": 203, "correlation": 130326
    }
  },
  {
    "ph": "f", "id": 130326, "pid": 76337, "tid": -914061504, "ts": 1716454223403224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403225, "dur": 0,
    "args": {
      "External id": 130327,
      "cbid": 205, "correlation": 130327
    }
  },
  {
    "ph": "f", "id": 130327, "pid": 76337, "tid": -914061504, "ts": 1716454223403225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458760, "dur": 5,
    "args": {
      "External id": 130331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130331,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130331, "pid": 5, "tid": 7, "ts": 1716454223458760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403238, "dur": 11,
    "args": {
      "External id": 130331,
      "cbid": 211, "correlation": 130331
    }
  },
  {
    "ph": "s", "id": 130331, "pid": 76337, "tid": -914061504, "ts": 1716454223403238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223458766, "dur": 160,
    "args": {
      "External id": 130333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130333,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130333, "pid": 5, "tid": 7, "ts": 1716454223458766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403252, "dur": 5,
    "args": {
      "External id": 130333,
      "cbid": 211, "correlation": 130333
    }
  },
  {
    "ph": "s", "id": 130333, "pid": 76337, "tid": -914061504, "ts": 1716454223403252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223458929, "dur": 1,
    "args": {
      "External id": 130335,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130335,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 130335, "pid": 5, "tid": 7, "ts": 1716454223458929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403262, "dur": 7,
    "args": {
      "External id": 130335,
      "cbid": 51, "correlation": 130335
    }
  },
  {
    "ph": "s", "id": 130335, "pid": 76337, "tid": -914061504, "ts": 1716454223403262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223458932, "dur": 256,
    "args": {
      "External id": 130336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130336,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130336, "pid": 5, "tid": 7, "ts": 1716454223458932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403270, "dur": 6,
    "args": {
      "External id": 130336,
      "cbid": 211, "correlation": 130336
    }
  },
  {
    "ph": "s", "id": 130336, "pid": 76337, "tid": -914061504, "ts": 1716454223403270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459190, "dur": 6,
    "args": {
      "External id": 130338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130338,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130338, "pid": 5, "tid": 7, "ts": 1716454223459190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403279, "dur": 5,
    "args": {
      "External id": 130338,
      "cbid": 211, "correlation": 130338
    }
  },
  {
    "ph": "s", "id": 130338, "pid": 76337, "tid": -914061504, "ts": 1716454223403279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223459197, "dur": 6,
    "args": {
      "External id": 130344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130344,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130344, "pid": 5, "tid": 7, "ts": 1716454223459197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403307, "dur": 9,
    "args": {
      "External id": 130344,
      "cbid": 211, "correlation": 130344
    }
  },
  {
    "ph": "s", "id": 130344, "pid": 76337, "tid": -914061504, "ts": 1716454223403307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223459205, "dur": 3,
    "args": {
      "External id": 130352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130352,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 130352, "pid": 5, "tid": 7, "ts": 1716454223459205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403351, "dur": 9,
    "args": {
      "External id": 130352,
      "cbid": 211, "correlation": 130352
    }
  },
  {
    "ph": "s", "id": 130352, "pid": 76337, "tid": -914061504, "ts": 1716454223403351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403414, "dur": 1,
    "args": {
      "External id": 130368,
      "cbid": 251, "correlation": 130368
    }
  },
  {
    "ph": "f", "id": 130368, "pid": 76337, "tid": -914061504, "ts": 1716454223403414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403419, "dur": 0,
    "args": {
      "External id": 130370,
      "cbid": 251, "correlation": 130370
    }
  },
  {
    "ph": "f", "id": 130370, "pid": 76337, "tid": -914061504, "ts": 1716454223403419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223459209, "dur": 10,
    "args": {
      "External id": 130371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130371,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130371, "pid": 5, "tid": 7, "ts": 1716454223459209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403421, "dur": 11,
    "args": {
      "External id": 130371,
      "cbid": 211, "correlation": 130371
    }
  },
  {
    "ph": "s", "id": 130371, "pid": 76337, "tid": -914061504, "ts": 1716454223403421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223459220, "dur": 3,
    "args": {
      "External id": 130373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130373,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130373, "pid": 5, "tid": 7, "ts": 1716454223459220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403433, "dur": 5,
    "args": {
      "External id": 130373,
      "cbid": 211, "correlation": 130373
    }
  },
  {
    "ph": "s", "id": 130373, "pid": 76337, "tid": -914061504, "ts": 1716454223403433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223459225, "dur": 6,
    "args": {
      "External id": 130383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130383, "pid": 5, "tid": 7, "ts": 1716454223459225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403490, "dur": 12,
    "args": {
      "External id": 130383,
      "cbid": 211, "correlation": 130383
    }
  },
  {
    "ph": "s", "id": 130383, "pid": 76337, "tid": -914061504, "ts": 1716454223403490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459232, "dur": 9,
    "args": {
      "External id": 130403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130403,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130403, "pid": 5, "tid": 7, "ts": 1716454223459232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403555, "dur": 11,
    "args": {
      "External id": 130403,
      "cbid": 211, "correlation": 130403
    }
  },
  {
    "ph": "s", "id": 130403, "pid": 76337, "tid": -914061504, "ts": 1716454223403555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223459243, "dur": 4,
    "args": {
      "External id": 130415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130415,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130415, "pid": 5, "tid": 7, "ts": 1716454223459243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403576, "dur": 6,
    "args": {
      "External id": 130415,
      "cbid": 211, "correlation": 130415
    }
  },
  {
    "ph": "s", "id": 130415, "pid": 76337, "tid": -914061504, "ts": 1716454223403576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223459248, "dur": 6,
    "args": {
      "External id": 130418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130418, "pid": 5, "tid": 7, "ts": 1716454223459248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403594, "dur": 6,
    "args": {
      "External id": 130418,
      "cbid": 211, "correlation": 130418
    }
  },
  {
    "ph": "s", "id": 130418, "pid": 76337, "tid": -914061504, "ts": 1716454223403594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223459256, "dur": 5,
    "args": {
      "External id": 130427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130427,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130427, "pid": 5, "tid": 7, "ts": 1716454223459256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403634, "dur": 10,
    "args": {
      "External id": 130427,
      "cbid": 211, "correlation": 130427
    }
  },
  {
    "ph": "s", "id": 130427, "pid": 76337, "tid": -914061504, "ts": 1716454223403634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403696, "dur": 0,
    "args": {
      "External id": 130437,
      "cbid": 317, "correlation": 130437
    }
  },
  {
    "ph": "f", "id": 130437, "pid": 76337, "tid": -914061504, "ts": 1716454223403696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403697, "dur": 0,
    "args": {
      "External id": 130438,
      "cbid": 203, "correlation": 130438
    }
  },
  {
    "ph": "f", "id": 130438, "pid": 76337, "tid": -914061504, "ts": 1716454223403697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403698, "dur": 0,
    "args": {
      "External id": 130439,
      "cbid": 205, "correlation": 130439
    }
  },
  {
    "ph": "f", "id": 130439, "pid": 76337, "tid": -914061504, "ts": 1716454223403698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459262, "dur": 5,
    "args": {
      "External id": 130443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130443,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130443, "pid": 5, "tid": 7, "ts": 1716454223459262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403712, "dur": 13,
    "args": {
      "External id": 130443,
      "cbid": 211, "correlation": 130443
    }
  },
  {
    "ph": "s", "id": 130443, "pid": 76337, "tid": -914061504, "ts": 1716454223403712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459268, "dur": 161,
    "args": {
      "External id": 130445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130445,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130445, "pid": 5, "tid": 7, "ts": 1716454223459268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403727, "dur": 5,
    "args": {
      "External id": 130445,
      "cbid": 211, "correlation": 130445
    }
  },
  {
    "ph": "s", "id": 130445, "pid": 76337, "tid": -914061504, "ts": 1716454223403727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223459431, "dur": 1,
    "args": {
      "External id": 130447,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130447,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 130447, "pid": 5, "tid": 7, "ts": 1716454223459431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403738, "dur": 7,
    "args": {
      "External id": 130447,
      "cbid": 51, "correlation": 130447
    }
  },
  {
    "ph": "s", "id": 130447, "pid": 76337, "tid": -914061504, "ts": 1716454223403738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223459435, "dur": 258,
    "args": {
      "External id": 130448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130448,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130448, "pid": 5, "tid": 7, "ts": 1716454223459435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403746, "dur": 6,
    "args": {
      "External id": 130448,
      "cbid": 211, "correlation": 130448
    }
  },
  {
    "ph": "s", "id": 130448, "pid": 76337, "tid": -914061504, "ts": 1716454223403746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459695, "dur": 6,
    "args": {
      "External id": 130450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130450,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130450, "pid": 5, "tid": 7, "ts": 1716454223459695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403755, "dur": 5,
    "args": {
      "External id": 130450,
      "cbid": 211, "correlation": 130450
    }
  },
  {
    "ph": "s", "id": 130450, "pid": 76337, "tid": -914061504, "ts": 1716454223403755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223459702, "dur": 6,
    "args": {
      "External id": 130456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130456, "pid": 5, "tid": 7, "ts": 1716454223459702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403784, "dur": 9,
    "args": {
      "External id": 130456,
      "cbid": 211, "correlation": 130456
    }
  },
  {
    "ph": "s", "id": 130456, "pid": 76337, "tid": -914061504, "ts": 1716454223403784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223459709, "dur": 5,
    "args": {
      "External id": 130464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130464,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130464, "pid": 5, "tid": 7, "ts": 1716454223459709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403818, "dur": 9,
    "args": {
      "External id": 130464,
      "cbid": 211, "correlation": 130464
    }
  },
  {
    "ph": "s", "id": 130464, "pid": 76337, "tid": -914061504, "ts": 1716454223403818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223459715, "dur": 4,
    "args": {
      "External id": 130472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130472,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130472, "pid": 5, "tid": 7, "ts": 1716454223459715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403847, "dur": 8,
    "args": {
      "External id": 130472,
      "cbid": 211, "correlation": 130472
    }
  },
  {
    "ph": "s", "id": 130472, "pid": 76337, "tid": -914061504, "ts": 1716454223403847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459721, "dur": 9,
    "args": {
      "External id": 130492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130492,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130492, "pid": 5, "tid": 7, "ts": 1716454223459721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403950, "dur": 13,
    "args": {
      "External id": 130492,
      "cbid": 211, "correlation": 130492
    }
  },
  {
    "ph": "s", "id": 130492, "pid": 76337, "tid": -914061504, "ts": 1716454223403950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223459732, "dur": 4,
    "args": {
      "External id": 130504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130504,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130504, "pid": 5, "tid": 7, "ts": 1716454223459732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223403982, "dur": 7,
    "args": {
      "External id": 130504,
      "cbid": 211, "correlation": 130504
    }
  },
  {
    "ph": "s", "id": 130504, "pid": 76337, "tid": -914061504, "ts": 1716454223403982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223459737, "dur": 7,
    "args": {
      "External id": 130507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130507, "pid": 5, "tid": 7, "ts": 1716454223459737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404001, "dur": 7,
    "args": {
      "External id": 130507,
      "cbid": 211, "correlation": 130507
    }
  },
  {
    "ph": "s", "id": 130507, "pid": 76337, "tid": -914061504, "ts": 1716454223404001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223459744, "dur": 5,
    "args": {
      "External id": 130516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130516,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130516, "pid": 5, "tid": 7, "ts": 1716454223459744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404041, "dur": 9,
    "args": {
      "External id": 130516,
      "cbid": 211, "correlation": 130516
    }
  },
  {
    "ph": "s", "id": 130516, "pid": 76337, "tid": -914061504, "ts": 1716454223404041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404094, "dur": 0,
    "args": {
      "External id": 130526,
      "cbid": 317, "correlation": 130526
    }
  },
  {
    "ph": "f", "id": 130526, "pid": 76337, "tid": -914061504, "ts": 1716454223404094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404095, "dur": 0,
    "args": {
      "External id": 130527,
      "cbid": 203, "correlation": 130527
    }
  },
  {
    "ph": "f", "id": 130527, "pid": 76337, "tid": -914061504, "ts": 1716454223404095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404096, "dur": 0,
    "args": {
      "External id": 130528,
      "cbid": 205, "correlation": 130528
    }
  },
  {
    "ph": "f", "id": 130528, "pid": 76337, "tid": -914061504, "ts": 1716454223404096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459750, "dur": 5,
    "args": {
      "External id": 130532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130532,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130532, "pid": 5, "tid": 7, "ts": 1716454223459750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404109, "dur": 12,
    "args": {
      "External id": 130532,
      "cbid": 211, "correlation": 130532
    }
  },
  {
    "ph": "s", "id": 130532, "pid": 76337, "tid": -914061504, "ts": 1716454223404109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223459756, "dur": 161,
    "args": {
      "External id": 130534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130534,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130534, "pid": 5, "tid": 7, "ts": 1716454223459756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404123, "dur": 5,
    "args": {
      "External id": 130534,
      "cbid": 211, "correlation": 130534
    }
  },
  {
    "ph": "s", "id": 130534, "pid": 76337, "tid": -914061504, "ts": 1716454223404123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223459920, "dur": 1,
    "args": {
      "External id": 130536,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130536,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1388888888888889
    }
  },
  {
    "ph": "f", "id": 130536, "pid": 5, "tid": 7, "ts": 1716454223459920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404134, "dur": 6,
    "args": {
      "External id": 130536,
      "cbid": 51, "correlation": 130536
    }
  },
  {
    "ph": "s", "id": 130536, "pid": 76337, "tid": -914061504, "ts": 1716454223404134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223459924, "dur": 256,
    "args": {
      "External id": 130537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130537,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130537, "pid": 5, "tid": 7, "ts": 1716454223459924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404141, "dur": 6,
    "args": {
      "External id": 130537,
      "cbid": 211, "correlation": 130537
    }
  },
  {
    "ph": "s", "id": 130537, "pid": 76337, "tid": -914061504, "ts": 1716454223404141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460181, "dur": 6,
    "args": {
      "External id": 130539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130539,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130539, "pid": 5, "tid": 7, "ts": 1716454223460181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404150, "dur": 5,
    "args": {
      "External id": 130539,
      "cbid": 211, "correlation": 130539
    }
  },
  {
    "ph": "s", "id": 130539, "pid": 76337, "tid": -914061504, "ts": 1716454223404150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223460189, "dur": 6,
    "args": {
      "External id": 130545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130545, "pid": 5, "tid": 7, "ts": 1716454223460189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404179, "dur": 8,
    "args": {
      "External id": 130545,
      "cbid": 211, "correlation": 130545
    }
  },
  {
    "ph": "s", "id": 130545, "pid": 76337, "tid": -914061504, "ts": 1716454223404179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223460196, "dur": 3,
    "args": {
      "External id": 130553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130553,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 130553, "pid": 5, "tid": 7, "ts": 1716454223460196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404223, "dur": 9,
    "args": {
      "External id": 130553,
      "cbid": 211, "correlation": 130553
    }
  },
  {
    "ph": "s", "id": 130553, "pid": 76337, "tid": -914061504, "ts": 1716454223404223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404285, "dur": 1,
    "args": {
      "External id": 130569,
      "cbid": 251, "correlation": 130569
    }
  },
  {
    "ph": "f", "id": 130569, "pid": 76337, "tid": -914061504, "ts": 1716454223404285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404291, "dur": 0,
    "args": {
      "External id": 130571,
      "cbid": 251, "correlation": 130571
    }
  },
  {
    "ph": "f", "id": 130571, "pid": 76337, "tid": -914061504, "ts": 1716454223404291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223460200, "dur": 10,
    "args": {
      "External id": 130572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130572,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130572, "pid": 5, "tid": 7, "ts": 1716454223460200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404293, "dur": 11,
    "args": {
      "External id": 130572,
      "cbid": 211, "correlation": 130572
    }
  },
  {
    "ph": "s", "id": 130572, "pid": 76337, "tid": -914061504, "ts": 1716454223404293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223460212, "dur": 4,
    "args": {
      "External id": 130574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130574,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130574, "pid": 5, "tid": 7, "ts": 1716454223460212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404306, "dur": 5,
    "args": {
      "External id": 130574,
      "cbid": 211, "correlation": 130574
    }
  },
  {
    "ph": "s", "id": 130574, "pid": 76337, "tid": -914061504, "ts": 1716454223404306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223460217, "dur": 6,
    "args": {
      "External id": 130584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130584,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130584, "pid": 5, "tid": 7, "ts": 1716454223460217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404362, "dur": 13,
    "args": {
      "External id": 130584,
      "cbid": 211, "correlation": 130584
    }
  },
  {
    "ph": "s", "id": 130584, "pid": 76337, "tid": -914061504, "ts": 1716454223404362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460224, "dur": 10,
    "args": {
      "External id": 130604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130604,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130604, "pid": 5, "tid": 7, "ts": 1716454223460224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404429, "dur": 10,
    "args": {
      "External id": 130604,
      "cbid": 211, "correlation": 130604
    }
  },
  {
    "ph": "s", "id": 130604, "pid": 76337, "tid": -914061504, "ts": 1716454223404429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223460235, "dur": 4,
    "args": {
      "External id": 130616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130616,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130616, "pid": 5, "tid": 7, "ts": 1716454223460235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404448, "dur": 6,
    "args": {
      "External id": 130616,
      "cbid": 211, "correlation": 130616
    }
  },
  {
    "ph": "s", "id": 130616, "pid": 76337, "tid": -914061504, "ts": 1716454223404448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223460240, "dur": 7,
    "args": {
      "External id": 130619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130619,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130619, "pid": 5, "tid": 7, "ts": 1716454223460240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404466, "dur": 7,
    "args": {
      "External id": 130619,
      "cbid": 211, "correlation": 130619
    }
  },
  {
    "ph": "s", "id": 130619, "pid": 76337, "tid": -914061504, "ts": 1716454223404466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223460248, "dur": 5,
    "args": {
      "External id": 130628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130628,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130628, "pid": 5, "tid": 7, "ts": 1716454223460248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404507, "dur": 10,
    "args": {
      "External id": 130628,
      "cbid": 211, "correlation": 130628
    }
  },
  {
    "ph": "s", "id": 130628, "pid": 76337, "tid": -914061504, "ts": 1716454223404507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404570, "dur": 0,
    "args": {
      "External id": 130638,
      "cbid": 317, "correlation": 130638
    }
  },
  {
    "ph": "f", "id": 130638, "pid": 76337, "tid": -914061504, "ts": 1716454223404570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404571, "dur": 0,
    "args": {
      "External id": 130639,
      "cbid": 203, "correlation": 130639
    }
  },
  {
    "ph": "f", "id": 130639, "pid": 76337, "tid": -914061504, "ts": 1716454223404571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404572, "dur": 0,
    "args": {
      "External id": 130640,
      "cbid": 205, "correlation": 130640
    }
  },
  {
    "ph": "f", "id": 130640, "pid": 76337, "tid": -914061504, "ts": 1716454223404572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460254, "dur": 5,
    "args": {
      "External id": 130644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130644,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130644, "pid": 5, "tid": 7, "ts": 1716454223460254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404586, "dur": 12,
    "args": {
      "External id": 130644,
      "cbid": 211, "correlation": 130644
    }
  },
  {
    "ph": "s", "id": 130644, "pid": 76337, "tid": -914061504, "ts": 1716454223404586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460260, "dur": 161,
    "args": {
      "External id": 130646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130646,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130646, "pid": 5, "tid": 7, "ts": 1716454223460260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404600, "dur": 5,
    "args": {
      "External id": 130646,
      "cbid": 211, "correlation": 130646
    }
  },
  {
    "ph": "s", "id": 130646, "pid": 76337, "tid": -914061504, "ts": 1716454223404600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223460423, "dur": 1,
    "args": {
      "External id": 130648,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130648,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 130648, "pid": 5, "tid": 7, "ts": 1716454223460423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404611, "dur": 6,
    "args": {
      "External id": 130648,
      "cbid": 51, "correlation": 130648
    }
  },
  {
    "ph": "s", "id": 130648, "pid": 76337, "tid": -914061504, "ts": 1716454223404611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223460427, "dur": 257,
    "args": {
      "External id": 130649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130649,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130649, "pid": 5, "tid": 7, "ts": 1716454223460427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404618, "dur": 7,
    "args": {
      "External id": 130649,
      "cbid": 211, "correlation": 130649
    }
  },
  {
    "ph": "s", "id": 130649, "pid": 76337, "tid": -914061504, "ts": 1716454223404618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460685, "dur": 6,
    "args": {
      "External id": 130651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130651,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130651, "pid": 5, "tid": 7, "ts": 1716454223460685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404628, "dur": 5,
    "args": {
      "External id": 130651,
      "cbid": 211, "correlation": 130651
    }
  },
  {
    "ph": "s", "id": 130651, "pid": 76337, "tid": -914061504, "ts": 1716454223404628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223460692, "dur": 6,
    "args": {
      "External id": 130657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130657,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130657, "pid": 5, "tid": 7, "ts": 1716454223460692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404656, "dur": 8,
    "args": {
      "External id": 130657,
      "cbid": 211, "correlation": 130657
    }
  },
  {
    "ph": "s", "id": 130657, "pid": 76337, "tid": -914061504, "ts": 1716454223404656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223460699, "dur": 5,
    "args": {
      "External id": 130665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130665,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130665, "pid": 5, "tid": 7, "ts": 1716454223460699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404689, "dur": 8,
    "args": {
      "External id": 130665,
      "cbid": 211, "correlation": 130665
    }
  },
  {
    "ph": "s", "id": 130665, "pid": 76337, "tid": -914061504, "ts": 1716454223404689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223460706, "dur": 4,
    "args": {
      "External id": 130673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130673,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130673, "pid": 5, "tid": 7, "ts": 1716454223460706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404718, "dur": 9,
    "args": {
      "External id": 130673,
      "cbid": 211, "correlation": 130673
    }
  },
  {
    "ph": "s", "id": 130673, "pid": 76337, "tid": -914061504, "ts": 1716454223404718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460712, "dur": 9,
    "args": {
      "External id": 130693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130693,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 130693, "pid": 5, "tid": 7, "ts": 1716454223460712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404861, "dur": 14,
    "args": {
      "External id": 130693,
      "cbid": 211, "correlation": 130693
    }
  },
  {
    "ph": "s", "id": 130693, "pid": 76337, "tid": -914061504, "ts": 1716454223404861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223460722, "dur": 4,
    "args": {
      "External id": 130705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130705,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 130705, "pid": 5, "tid": 7, "ts": 1716454223460722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404884, "dur": 6,
    "args": {
      "External id": 130705,
      "cbid": 211, "correlation": 130705
    }
  },
  {
    "ph": "s", "id": 130705, "pid": 76337, "tid": -914061504, "ts": 1716454223404884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223460727, "dur": 7,
    "args": {
      "External id": 130708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130708, "pid": 5, "tid": 7, "ts": 1716454223460727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404903, "dur": 6,
    "args": {
      "External id": 130708,
      "cbid": 211, "correlation": 130708
    }
  },
  {
    "ph": "s", "id": 130708, "pid": 76337, "tid": -914061504, "ts": 1716454223404903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404961, "dur": 0,
    "args": {
      "External id": 130719,
      "cbid": 317, "correlation": 130719
    }
  },
  {
    "ph": "f", "id": 130719, "pid": 76337, "tid": -914061504, "ts": 1716454223404961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404961, "dur": 0,
    "args": {
      "External id": 130720,
      "cbid": 203, "correlation": 130720
    }
  },
  {
    "ph": "f", "id": 130720, "pid": 76337, "tid": -914061504, "ts": 1716454223404961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404962, "dur": 0,
    "args": {
      "External id": 130721,
      "cbid": 205, "correlation": 130721
    }
  },
  {
    "ph": "f", "id": 130721, "pid": 76337, "tid": -914061504, "ts": 1716454223404962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460735, "dur": 5,
    "args": {
      "External id": 130725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130725,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130725, "pid": 5, "tid": 7, "ts": 1716454223460735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223404989, "dur": 12,
    "args": {
      "External id": 130725,
      "cbid": 211, "correlation": 130725
    }
  },
  {
    "ph": "s", "id": 130725, "pid": 76337, "tid": -914061504, "ts": 1716454223404989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223460742, "dur": 37,
    "args": {
      "External id": 130727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130727,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 130727, "pid": 5, "tid": 7, "ts": 1716454223460742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405008, "dur": 10,
    "args": {
      "External id": 130727,
      "cbid": 211, "correlation": 130727
    }
  },
  {
    "ph": "s", "id": 130727, "pid": 76337, "tid": -914061504, "ts": 1716454223405008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460780, "dur": 5,
    "args": {
      "External id": 130729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130729,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130729, "pid": 5, "tid": 7, "ts": 1716454223460780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405022, "dur": 5,
    "args": {
      "External id": 130729,
      "cbid": 211, "correlation": 130729
    }
  },
  {
    "ph": "s", "id": 130729, "pid": 76337, "tid": -914061504, "ts": 1716454223405022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223460786, "dur": 6,
    "args": {
      "External id": 130735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130735, "pid": 5, "tid": 7, "ts": 1716454223460786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405050, "dur": 8,
    "args": {
      "External id": 130735,
      "cbid": 211, "correlation": 130735
    }
  },
  {
    "ph": "s", "id": 130735, "pid": 76337, "tid": -914061504, "ts": 1716454223405050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223460793, "dur": 20,
    "args": {
      "External id": 130744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130744,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130744, "pid": 5, "tid": 7, "ts": 1716454223460793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405133, "dur": 14,
    "args": {
      "External id": 130744,
      "cbid": 211, "correlation": 130744
    }
  },
  {
    "ph": "s", "id": 130744, "pid": 76337, "tid": -914061504, "ts": 1716454223405133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223460815, "dur": 11,
    "args": {
      "External id": 130766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130766,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 130766, "pid": 5, "tid": 7, "ts": 1716454223460815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405190, "dur": 10,
    "args": {
      "External id": 130766,
      "cbid": 211, "correlation": 130766
    }
  },
  {
    "ph": "s", "id": 130766, "pid": 76337, "tid": -914061504, "ts": 1716454223405190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405279, "dur": 2,
    "args": {
      "External id": 130777,
      "cbid": 251, "correlation": 130777
    }
  },
  {
    "ph": "f", "id": 130777, "pid": 76337, "tid": -914061504, "ts": 1716454223405279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405284, "dur": 0,
    "args": {
      "External id": 130778,
      "cbid": 251, "correlation": 130778
    }
  },
  {
    "ph": "f", "id": 130778, "pid": 76337, "tid": -914061504, "ts": 1716454223405284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223460827, "dur": 53,
    "args": {
      "External id": 130779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130779,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 130779, "pid": 5, "tid": 7, "ts": 1716454223460827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405287, "dur": 14,
    "args": {
      "External id": 130779,
      "cbid": 211, "correlation": 130779
    }
  },
  {
    "ph": "s", "id": 130779, "pid": 76337, "tid": -914061504, "ts": 1716454223405287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405358, "dur": 1,
    "args": {
      "External id": 130790,
      "cbid": 251, "correlation": 130790
    }
  },
  {
    "ph": "f", "id": 130790, "pid": 76337, "tid": -914061504, "ts": 1716454223405358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405362, "dur": 0,
    "args": {
      "External id": 130791,
      "cbid": 251, "correlation": 130791
    }
  },
  {
    "ph": "f", "id": 130791, "pid": 76337, "tid": -914061504, "ts": 1716454223405362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223460882, "dur": 52,
    "args": {
      "External id": 130792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130792,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 130792, "pid": 5, "tid": 7, "ts": 1716454223460882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405364, "dur": 12,
    "args": {
      "External id": 130792,
      "cbid": 211, "correlation": 130792
    }
  },
  {
    "ph": "s", "id": 130792, "pid": 76337, "tid": -914061504, "ts": 1716454223405364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405429, "dur": 1,
    "args": {
      "External id": 130803,
      "cbid": 251, "correlation": 130803
    }
  },
  {
    "ph": "f", "id": 130803, "pid": 76337, "tid": -914061504, "ts": 1716454223405429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405432, "dur": 0,
    "args": {
      "External id": 130804,
      "cbid": 251, "correlation": 130804
    }
  },
  {
    "ph": "f", "id": 130804, "pid": 76337, "tid": -914061504, "ts": 1716454223405432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223460935, "dur": 53,
    "args": {
      "External id": 130805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130805,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 130805, "pid": 5, "tid": 7, "ts": 1716454223460935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405434, "dur": 11,
    "args": {
      "External id": 130805,
      "cbid": 211, "correlation": 130805
    }
  },
  {
    "ph": "s", "id": 130805, "pid": 76337, "tid": -914061504, "ts": 1716454223405434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223460990, "dur": 55,
    "args": {
      "External id": 130830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130830,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130830, "pid": 5, "tid": 7, "ts": 1716454223460990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405517, "dur": 13,
    "args": {
      "External id": 130830,
      "cbid": 211, "correlation": 130830
    }
  },
  {
    "ph": "s", "id": 130830, "pid": 76337, "tid": -914061504, "ts": 1716454223405517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405616, "dur": 1,
    "args": {
      "External id": 130848,
      "cbid": 251, "correlation": 130848
    }
  },
  {
    "ph": "f", "id": 130848, "pid": 76337, "tid": -914061504, "ts": 1716454223405616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223461047, "dur": 62,
    "args": {
      "External id": 130850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130850,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 130850, "pid": 5, "tid": 7, "ts": 1716454223461047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405622, "dur": 13,
    "args": {
      "External id": 130850,
      "cbid": 211, "correlation": 130850
    }
  },
  {
    "ph": "s", "id": 130850, "pid": 76337, "tid": -914061504, "ts": 1716454223405622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223461110, "dur": 6,
    "args": {
      "External id": 130858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130858,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130858, "pid": 5, "tid": 7, "ts": 1716454223461110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405692, "dur": 12,
    "args": {
      "External id": 130858,
      "cbid": 211, "correlation": 130858
    }
  },
  {
    "ph": "s", "id": 130858, "pid": 76337, "tid": -914061504, "ts": 1716454223405692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223461118, "dur": 7,
    "args": {
      "External id": 130866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130866,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 130866, "pid": 5, "tid": 7, "ts": 1716454223461118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405730, "dur": 8,
    "args": {
      "External id": 130866,
      "cbid": 211, "correlation": 130866
    }
  },
  {
    "ph": "s", "id": 130866, "pid": 76337, "tid": -914061504, "ts": 1716454223405730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461126, "dur": 7,
    "args": {
      "External id": 130877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130877,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130877, "pid": 5, "tid": 7, "ts": 1716454223461126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405802, "dur": 12,
    "args": {
      "External id": 130877,
      "cbid": 211, "correlation": 130877
    }
  },
  {
    "ph": "s", "id": 130877, "pid": 76337, "tid": -914061504, "ts": 1716454223405802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461135, "dur": 8,
    "args": {
      "External id": 130899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130899,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 130899, "pid": 5, "tid": 7, "ts": 1716454223461135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405835, "dur": 7,
    "args": {
      "External id": 130899,
      "cbid": 211, "correlation": 130899
    }
  },
  {
    "ph": "s", "id": 130899, "pid": 76337, "tid": -914061504, "ts": 1716454223405835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405921, "dur": 2,
    "args": {
      "External id": 130910,
      "cbid": 251, "correlation": 130910
    }
  },
  {
    "ph": "f", "id": 130910, "pid": 76337, "tid": -914061504, "ts": 1716454223405921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223461145, "dur": 1,
    "args": {
      "External id": 130911,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130911,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 130911, "pid": 5, "tid": 7, "ts": 1716454223461145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405927, "dur": 11,
    "args": {
      "External id": 130911,
      "cbid": 51, "correlation": 130911
    }
  },
  {
    "ph": "s", "id": 130911, "pid": 76337, "tid": -914061504, "ts": 1716454223405927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223461149, "dur": 36,
    "args": {
      "External id": 130912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130912,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 130912, "pid": 5, "tid": 7, "ts": 1716454223461149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223405939, "dur": 12,
    "args": {
      "External id": 130912,
      "cbid": 211, "correlation": 130912
    }
  },
  {
    "ph": "s", "id": 130912, "pid": 76337, "tid": -914061504, "ts": 1716454223405939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406019, "dur": 1,
    "args": {
      "External id": 130923,
      "cbid": 251, "correlation": 130923
    }
  },
  {
    "ph": "f", "id": 130923, "pid": 76337, "tid": -914061504, "ts": 1716454223406019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406023, "dur": 0,
    "args": {
      "External id": 130924,
      "cbid": 251, "correlation": 130924
    }
  },
  {
    "ph": "f", "id": 130924, "pid": 76337, "tid": -914061504, "ts": 1716454223406023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223461187, "dur": 12,
    "args": {
      "External id": 130925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130925,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130925, "pid": 5, "tid": 7, "ts": 1716454223461187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406025, "dur": 13,
    "args": {
      "External id": 130925,
      "cbid": 211, "correlation": 130925
    }
  },
  {
    "ph": "s", "id": 130925, "pid": 76337, "tid": -914061504, "ts": 1716454223406025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223461200, "dur": 5,
    "args": {
      "External id": 130927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130927,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130927, "pid": 5, "tid": 7, "ts": 1716454223461200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406041, "dur": 6,
    "args": {
      "External id": 130927,
      "cbid": 211, "correlation": 130927
    }
  },
  {
    "ph": "s", "id": 130927, "pid": 76337, "tid": -914061504, "ts": 1716454223406041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406099, "dur": 1,
    "args": {
      "External id": 130938,
      "cbid": 251, "correlation": 130938
    }
  },
  {
    "ph": "f", "id": 130938, "pid": 76337, "tid": -914061504, "ts": 1716454223406099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406102, "dur": 0,
    "args": {
      "External id": 130939,
      "cbid": 251, "correlation": 130939
    }
  },
  {
    "ph": "f", "id": 130939, "pid": 76337, "tid": -914061504, "ts": 1716454223406102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223461206, "dur": 8,
    "args": {
      "External id": 130940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130940,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130940, "pid": 5, "tid": 7, "ts": 1716454223461206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406104, "dur": 11,
    "args": {
      "External id": 130940,
      "cbid": 211, "correlation": 130940
    }
  },
  {
    "ph": "s", "id": 130940, "pid": 76337, "tid": -914061504, "ts": 1716454223406104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223461216, "dur": 4,
    "args": {
      "External id": 130942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130942,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 130942, "pid": 5, "tid": 7, "ts": 1716454223461216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406117, "dur": 5,
    "args": {
      "External id": 130942,
      "cbid": 211, "correlation": 130942
    }
  },
  {
    "ph": "s", "id": 130942, "pid": 76337, "tid": -914061504, "ts": 1716454223406117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223461221, "dur": 20,
    "args": {
      "External id": 130967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130967,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 130967, "pid": 5, "tid": 7, "ts": 1716454223461221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406196, "dur": 12,
    "args": {
      "External id": 130967,
      "cbid": 211, "correlation": 130967
    }
  },
  {
    "ph": "s", "id": 130967, "pid": 76337, "tid": -914061504, "ts": 1716454223406196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406296, "dur": 2,
    "args": {
      "External id": 130985,
      "cbid": 251, "correlation": 130985
    }
  },
  {
    "ph": "f", "id": 130985, "pid": 76337, "tid": -914061504, "ts": 1716454223406296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223461243, "dur": 1,
    "args": {
      "External id": 130987,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 130987,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 130987, "pid": 5, "tid": 7, "ts": 1716454223461243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406302, "dur": 10,
    "args": {
      "External id": 130987,
      "cbid": 51, "correlation": 130987
    }
  },
  {
    "ph": "s", "id": 130987, "pid": 76337, "tid": -914061504, "ts": 1716454223406302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223461246, "dur": 36,
    "args": {
      "External id": 130988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130988,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 130988, "pid": 5, "tid": 7, "ts": 1716454223461246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406314, "dur": 13,
    "args": {
      "External id": 130988,
      "cbid": 211, "correlation": 130988
    }
  },
  {
    "ph": "s", "id": 130988, "pid": 76337, "tid": -914061504, "ts": 1716454223406314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223461284, "dur": 4,
    "args": {
      "External id": 130996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 130996,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 130996, "pid": 5, "tid": 7, "ts": 1716454223461284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406386, "dur": 12,
    "args": {
      "External id": 130996,
      "cbid": 211, "correlation": 130996
    }
  },
  {
    "ph": "s", "id": 130996, "pid": 76337, "tid": -914061504, "ts": 1716454223406386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461290, "dur": 8,
    "args": {
      "External id": 131004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131004, "pid": 5, "tid": 7, "ts": 1716454223461290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406428, "dur": 9,
    "args": {
      "External id": 131004,
      "cbid": 211, "correlation": 131004
    }
  },
  {
    "ph": "s", "id": 131004, "pid": 76337, "tid": -914061504, "ts": 1716454223406428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461299, "dur": 8,
    "args": {
      "External id": 131026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131026,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 131026, "pid": 5, "tid": 7, "ts": 1716454223461299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406479, "dur": 10,
    "args": {
      "External id": 131026,
      "cbid": 211, "correlation": 131026
    }
  },
  {
    "ph": "s", "id": 131026, "pid": 76337, "tid": -914061504, "ts": 1716454223406479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406569, "dur": 1,
    "args": {
      "External id": 131042,
      "cbid": 251, "correlation": 131042
    }
  },
  {
    "ph": "f", "id": 131042, "pid": 76337, "tid": -914061504, "ts": 1716454223406569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406575, "dur": 0,
    "args": {
      "External id": 131044,
      "cbid": 251, "correlation": 131044
    }
  },
  {
    "ph": "f", "id": 131044, "pid": 76337, "tid": -914061504, "ts": 1716454223406575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223461308, "dur": 190,
    "args": {
      "External id": 131045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131045,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131045, "pid": 5, "tid": 7, "ts": 1716454223461308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406577, "dur": 13,
    "args": {
      "External id": 131045,
      "cbid": 211, "correlation": 131045
    }
  },
  {
    "ph": "s", "id": 131045, "pid": 76337, "tid": -914061504, "ts": 1716454223406577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461500, "dur": 21,
    "args": {
      "External id": 131053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131053,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131053, "pid": 5, "tid": 7, "ts": 1716454223461500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406644, "dur": 13,
    "args": {
      "External id": 131053,
      "cbid": 211, "correlation": 131053
    }
  },
  {
    "ph": "s", "id": 131053, "pid": 76337, "tid": -914061504, "ts": 1716454223406644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461522, "dur": 21,
    "args": {
      "External id": 131061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131061, "pid": 5, "tid": 7, "ts": 1716454223461522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406675, "dur": 9,
    "args": {
      "External id": 131061,
      "cbid": 211, "correlation": 131061
    }
  },
  {
    "ph": "s", "id": 131061, "pid": 76337, "tid": -914061504, "ts": 1716454223406675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406756, "dur": 1,
    "args": {
      "External id": 131077,
      "cbid": 251, "correlation": 131077
    }
  },
  {
    "ph": "f", "id": 131077, "pid": 76337, "tid": -914061504, "ts": 1716454223406756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223461545, "dur": 1,
    "args": {
      "External id": 131079,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131079,
      "bytes": 120, "memory bandwidth (GB/s)": 0.0797872340425532
    }
  },
  {
    "ph": "f", "id": 131079, "pid": 5, "tid": 7, "ts": 1716454223461545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406761, "dur": 8,
    "args": {
      "External id": 131079,
      "cbid": 51, "correlation": 131079
    }
  },
  {
    "ph": "s", "id": 131079, "pid": 76337, "tid": -914061504, "ts": 1716454223406761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223461549, "dur": 109,
    "args": {
      "External id": 131080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131080,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 131080, "pid": 5, "tid": 7, "ts": 1716454223461549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406771, "dur": 12,
    "args": {
      "External id": 131080,
      "cbid": 211, "correlation": 131080
    }
  },
  {
    "ph": "s", "id": 131080, "pid": 76337, "tid": -914061504, "ts": 1716454223406771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223461659, "dur": 5,
    "args": {
      "External id": 131088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131088,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131088, "pid": 5, "tid": 7, "ts": 1716454223461659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406814, "dur": 10,
    "args": {
      "External id": 131088,
      "cbid": 211, "correlation": 131088
    }
  },
  {
    "ph": "s", "id": 131088, "pid": 76337, "tid": -914061504, "ts": 1716454223406814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461666, "dur": 9,
    "args": {
      "External id": 131099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131099, "pid": 5, "tid": 7, "ts": 1716454223461666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406883, "dur": 12,
    "args": {
      "External id": 131099,
      "cbid": 211, "correlation": 131099
    }
  },
  {
    "ph": "s", "id": 131099, "pid": 76337, "tid": -914061504, "ts": 1716454223406883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406947, "dur": 0,
    "args": {
      "External id": 131111,
      "cbid": 317, "correlation": 131111
    }
  },
  {
    "ph": "f", "id": 131111, "pid": 76337, "tid": -914061504, "ts": 1716454223406947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406947, "dur": 0,
    "args": {
      "External id": 131112,
      "cbid": 203, "correlation": 131112
    }
  },
  {
    "ph": "f", "id": 131112, "pid": 76337, "tid": -914061504, "ts": 1716454223406947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406948, "dur": 0,
    "args": {
      "External id": 131113,
      "cbid": 205, "correlation": 131113
    }
  },
  {
    "ph": "f", "id": 131113, "pid": 76337, "tid": -914061504, "ts": 1716454223406948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461676, "dur": 6,
    "args": {
      "External id": 131117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131117,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131117, "pid": 5, "tid": 7, "ts": 1716454223461676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406963, "dur": 20,
    "args": {
      "External id": 131117,
      "cbid": 211, "correlation": 131117
    }
  },
  {
    "ph": "s", "id": 131117, "pid": 76337, "tid": -914061504, "ts": 1716454223406963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223461683, "dur": 36,
    "args": {
      "External id": 131119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131119,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 131119, "pid": 5, "tid": 7, "ts": 1716454223461683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223406990, "dur": 7,
    "args": {
      "External id": 131119,
      "cbid": 211, "correlation": 131119
    }
  },
  {
    "ph": "s", "id": 131119, "pid": 76337, "tid": -914061504, "ts": 1716454223406990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461721, "dur": 6,
    "args": {
      "External id": 131121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131121,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131121, "pid": 5, "tid": 7, "ts": 1716454223461721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407001, "dur": 5,
    "args": {
      "External id": 131121,
      "cbid": 211, "correlation": 131121
    }
  },
  {
    "ph": "s", "id": 131121, "pid": 76337, "tid": -914061504, "ts": 1716454223407001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461728, "dur": 7,
    "args": {
      "External id": 131127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131127, "pid": 5, "tid": 7, "ts": 1716454223461728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407030, "dur": 9,
    "args": {
      "External id": 131127,
      "cbid": 211, "correlation": 131127
    }
  },
  {
    "ph": "s", "id": 131127, "pid": 76337, "tid": -914061504, "ts": 1716454223407030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223461737, "dur": 5,
    "args": {
      "External id": 131135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131135,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131135, "pid": 5, "tid": 7, "ts": 1716454223461737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407061, "dur": 8,
    "args": {
      "External id": 131135,
      "cbid": 211, "correlation": 131135
    }
  },
  {
    "ph": "s", "id": 131135, "pid": 76337, "tid": -914061504, "ts": 1716454223407061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461743, "dur": 11,
    "args": {
      "External id": 131155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131155,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131155, "pid": 5, "tid": 7, "ts": 1716454223461743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407135, "dur": 11,
    "args": {
      "External id": 131155,
      "cbid": 211, "correlation": 131155
    }
  },
  {
    "ph": "s", "id": 131155, "pid": 76337, "tid": -914061504, "ts": 1716454223407135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223461755, "dur": 4,
    "args": {
      "External id": 131167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131167,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 131167, "pid": 5, "tid": 7, "ts": 1716454223461755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407156, "dur": 6,
    "args": {
      "External id": 131167,
      "cbid": 211, "correlation": 131167
    }
  },
  {
    "ph": "s", "id": 131167, "pid": 76337, "tid": -914061504, "ts": 1716454223407156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223461761, "dur": 9,
    "args": {
      "External id": 131170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131170,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131170, "pid": 5, "tid": 7, "ts": 1716454223461761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407175, "dur": 7,
    "args": {
      "External id": 131170,
      "cbid": 211, "correlation": 131170
    }
  },
  {
    "ph": "s", "id": 131170, "pid": 76337, "tid": -914061504, "ts": 1716454223407175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223461771, "dur": 5,
    "args": {
      "External id": 131179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131179,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131179, "pid": 5, "tid": 7, "ts": 1716454223461771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407214, "dur": 10,
    "args": {
      "External id": 131179,
      "cbid": 211, "correlation": 131179
    }
  },
  {
    "ph": "s", "id": 131179, "pid": 76337, "tid": -914061504, "ts": 1716454223407214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407266, "dur": 0,
    "args": {
      "External id": 131189,
      "cbid": 317, "correlation": 131189
    }
  },
  {
    "ph": "f", "id": 131189, "pid": 76337, "tid": -914061504, "ts": 1716454223407266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407266, "dur": 0,
    "args": {
      "External id": 131190,
      "cbid": 203, "correlation": 131190
    }
  },
  {
    "ph": "f", "id": 131190, "pid": 76337, "tid": -914061504, "ts": 1716454223407266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407267, "dur": 0,
    "args": {
      "External id": 131191,
      "cbid": 205, "correlation": 131191
    }
  },
  {
    "ph": "f", "id": 131191, "pid": 76337, "tid": -914061504, "ts": 1716454223407267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461777, "dur": 5,
    "args": {
      "External id": 131195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131195,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131195, "pid": 5, "tid": 7, "ts": 1716454223461777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407281, "dur": 12,
    "args": {
      "External id": 131195,
      "cbid": 211, "correlation": 131195
    }
  },
  {
    "ph": "s", "id": 131195, "pid": 76337, "tid": -914061504, "ts": 1716454223407281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223461784, "dur": 161,
    "args": {
      "External id": 131197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131197,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131197, "pid": 5, "tid": 7, "ts": 1716454223461784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407295, "dur": 5,
    "args": {
      "External id": 131197,
      "cbid": 211, "correlation": 131197
    }
  },
  {
    "ph": "s", "id": 131197, "pid": 76337, "tid": -914061504, "ts": 1716454223407295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223461947, "dur": 1,
    "args": {
      "External id": 131199,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131199,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 131199, "pid": 5, "tid": 7, "ts": 1716454223461947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407307, "dur": 6,
    "args": {
      "External id": 131199,
      "cbid": 51, "correlation": 131199
    }
  },
  {
    "ph": "s", "id": 131199, "pid": 76337, "tid": -914061504, "ts": 1716454223407307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223461951, "dur": 268,
    "args": {
      "External id": 131200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131200,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131200, "pid": 5, "tid": 7, "ts": 1716454223461951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407314, "dur": 6,
    "args": {
      "External id": 131200,
      "cbid": 211, "correlation": 131200
    }
  },
  {
    "ph": "s", "id": 131200, "pid": 76337, "tid": -914061504, "ts": 1716454223407314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462220, "dur": 6,
    "args": {
      "External id": 131202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131202, "pid": 5, "tid": 7, "ts": 1716454223462220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407324, "dur": 5,
    "args": {
      "External id": 131202,
      "cbid": 211, "correlation": 131202
    }
  },
  {
    "ph": "s", "id": 131202, "pid": 76337, "tid": -914061504, "ts": 1716454223407324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223462227, "dur": 6,
    "args": {
      "External id": 131208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131208,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131208, "pid": 5, "tid": 7, "ts": 1716454223462227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407352, "dur": 8,
    "args": {
      "External id": 131208,
      "cbid": 211, "correlation": 131208
    }
  },
  {
    "ph": "s", "id": 131208, "pid": 76337, "tid": -914061504, "ts": 1716454223407352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223462235, "dur": 3,
    "args": {
      "External id": 131216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131216,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 131216, "pid": 5, "tid": 7, "ts": 1716454223462235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407396, "dur": 9,
    "args": {
      "External id": 131216,
      "cbid": 211, "correlation": 131216
    }
  },
  {
    "ph": "s", "id": 131216, "pid": 76337, "tid": -914061504, "ts": 1716454223407396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407460, "dur": 1,
    "args": {
      "External id": 131232,
      "cbid": 251, "correlation": 131232
    }
  },
  {
    "ph": "f", "id": 131232, "pid": 76337, "tid": -914061504, "ts": 1716454223407460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407465, "dur": 0,
    "args": {
      "External id": 131234,
      "cbid": 251, "correlation": 131234
    }
  },
  {
    "ph": "f", "id": 131234, "pid": 76337, "tid": -914061504, "ts": 1716454223407465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223462239, "dur": 13,
    "args": {
      "External id": 131235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131235,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131235, "pid": 5, "tid": 7, "ts": 1716454223462239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407467, "dur": 11,
    "args": {
      "External id": 131235,
      "cbid": 211, "correlation": 131235
    }
  },
  {
    "ph": "s", "id": 131235, "pid": 76337, "tid": -914061504, "ts": 1716454223407467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223462253, "dur": 5,
    "args": {
      "External id": 131237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131237,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131237, "pid": 5, "tid": 7, "ts": 1716454223462253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407480, "dur": 6,
    "args": {
      "External id": 131237,
      "cbid": 211, "correlation": 131237
    }
  },
  {
    "ph": "s", "id": 131237, "pid": 76337, "tid": -914061504, "ts": 1716454223407480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223462260, "dur": 6,
    "args": {
      "External id": 131247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131247, "pid": 5, "tid": 7, "ts": 1716454223462260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407538, "dur": 12,
    "args": {
      "External id": 131247,
      "cbid": 211, "correlation": 131247
    }
  },
  {
    "ph": "s", "id": 131247, "pid": 76337, "tid": -914061504, "ts": 1716454223407538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462266, "dur": 10,
    "args": {
      "External id": 131267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131267,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131267, "pid": 5, "tid": 7, "ts": 1716454223462266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407603, "dur": 11,
    "args": {
      "External id": 131267,
      "cbid": 211, "correlation": 131267
    }
  },
  {
    "ph": "s", "id": 131267, "pid": 76337, "tid": -914061504, "ts": 1716454223407603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223462278, "dur": 4,
    "args": {
      "External id": 131279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131279,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 131279, "pid": 5, "tid": 7, "ts": 1716454223462278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407624, "dur": 6,
    "args": {
      "External id": 131279,
      "cbid": 211, "correlation": 131279
    }
  },
  {
    "ph": "s", "id": 131279, "pid": 76337, "tid": -914061504, "ts": 1716454223407624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223462283, "dur": 7,
    "args": {
      "External id": 131282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131282, "pid": 5, "tid": 7, "ts": 1716454223462283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407643, "dur": 6,
    "args": {
      "External id": 131282,
      "cbid": 211, "correlation": 131282
    }
  },
  {
    "ph": "s", "id": 131282, "pid": 76337, "tid": -914061504, "ts": 1716454223407643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223462291, "dur": 5,
    "args": {
      "External id": 131291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131291,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131291, "pid": 5, "tid": 7, "ts": 1716454223462291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407683, "dur": 10,
    "args": {
      "External id": 131291,
      "cbid": 211, "correlation": 131291
    }
  },
  {
    "ph": "s", "id": 131291, "pid": 76337, "tid": -914061504, "ts": 1716454223407683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407745, "dur": 0,
    "args": {
      "External id": 131301,
      "cbid": 317, "correlation": 131301
    }
  },
  {
    "ph": "f", "id": 131301, "pid": 76337, "tid": -914061504, "ts": 1716454223407745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407746, "dur": 0,
    "args": {
      "External id": 131302,
      "cbid": 203, "correlation": 131302
    }
  },
  {
    "ph": "f", "id": 131302, "pid": 76337, "tid": -914061504, "ts": 1716454223407746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407747, "dur": 0,
    "args": {
      "External id": 131303,
      "cbid": 205, "correlation": 131303
    }
  },
  {
    "ph": "f", "id": 131303, "pid": 76337, "tid": -914061504, "ts": 1716454223407747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462296, "dur": 5,
    "args": {
      "External id": 131307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131307,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131307, "pid": 5, "tid": 7, "ts": 1716454223462296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407760, "dur": 12,
    "args": {
      "External id": 131307,
      "cbid": 211, "correlation": 131307
    }
  },
  {
    "ph": "s", "id": 131307, "pid": 76337, "tid": -914061504, "ts": 1716454223407760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462303, "dur": 161,
    "args": {
      "External id": 131309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131309, "pid": 5, "tid": 7, "ts": 1716454223462303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407775, "dur": 5,
    "args": {
      "External id": 131309,
      "cbid": 211, "correlation": 131309
    }
  },
  {
    "ph": "s", "id": 131309, "pid": 76337, "tid": -914061504, "ts": 1716454223407775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223462466, "dur": 1,
    "args": {
      "External id": 131311,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131311,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 131311, "pid": 5, "tid": 7, "ts": 1716454223462466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407786, "dur": 6,
    "args": {
      "External id": 131311,
      "cbid": 51, "correlation": 131311
    }
  },
  {
    "ph": "s", "id": 131311, "pid": 76337, "tid": -914061504, "ts": 1716454223407786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223462470, "dur": 257,
    "args": {
      "External id": 131312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131312,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131312, "pid": 5, "tid": 7, "ts": 1716454223462470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407793, "dur": 6,
    "args": {
      "External id": 131312,
      "cbid": 211, "correlation": 131312
    }
  },
  {
    "ph": "s", "id": 131312, "pid": 76337, "tid": -914061504, "ts": 1716454223407793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462728, "dur": 6,
    "args": {
      "External id": 131314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131314,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131314, "pid": 5, "tid": 7, "ts": 1716454223462728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407803, "dur": 5,
    "args": {
      "External id": 131314,
      "cbid": 211, "correlation": 131314
    }
  },
  {
    "ph": "s", "id": 131314, "pid": 76337, "tid": -914061504, "ts": 1716454223407803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223462736, "dur": 6,
    "args": {
      "External id": 131320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131320, "pid": 5, "tid": 7, "ts": 1716454223462736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407831, "dur": 9,
    "args": {
      "External id": 131320,
      "cbid": 211, "correlation": 131320
    }
  },
  {
    "ph": "s", "id": 131320, "pid": 76337, "tid": -914061504, "ts": 1716454223407831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223462743, "dur": 5,
    "args": {
      "External id": 131328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131328,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131328, "pid": 5, "tid": 7, "ts": 1716454223462743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407864, "dur": 8,
    "args": {
      "External id": 131328,
      "cbid": 211, "correlation": 131328
    }
  },
  {
    "ph": "s", "id": 131328, "pid": 76337, "tid": -914061504, "ts": 1716454223407864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223462749, "dur": 4,
    "args": {
      "External id": 131336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131336,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131336, "pid": 5, "tid": 7, "ts": 1716454223462749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407894, "dur": 8,
    "args": {
      "External id": 131336,
      "cbid": 211, "correlation": 131336
    }
  },
  {
    "ph": "s", "id": 131336, "pid": 76337, "tid": -914061504, "ts": 1716454223407894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223462755, "dur": 12,
    "args": {
      "External id": 131345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131345,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131345, "pid": 5, "tid": 7, "ts": 1716454223462755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223407989, "dur": 14,
    "args": {
      "External id": 131345,
      "cbid": 211, "correlation": 131345
    }
  },
  {
    "ph": "s", "id": 131345, "pid": 76337, "tid": -914061504, "ts": 1716454223407989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462768, "dur": 12,
    "args": {
      "External id": 131365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131365,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131365, "pid": 5, "tid": 7, "ts": 1716454223462768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408058, "dur": 11,
    "args": {
      "External id": 131365,
      "cbid": 211, "correlation": 131365
    }
  },
  {
    "ph": "s", "id": 131365, "pid": 76337, "tid": -914061504, "ts": 1716454223408058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223462781, "dur": 4,
    "args": {
      "External id": 131377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131377,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131377, "pid": 5, "tid": 7, "ts": 1716454223462781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408079, "dur": 7,
    "args": {
      "External id": 131377,
      "cbid": 211, "correlation": 131377
    }
  },
  {
    "ph": "s", "id": 131377, "pid": 76337, "tid": -914061504, "ts": 1716454223408079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223462786, "dur": 10,
    "args": {
      "External id": 131380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131380, "pid": 5, "tid": 7, "ts": 1716454223462786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408098, "dur": 7,
    "args": {
      "External id": 131380,
      "cbid": 211, "correlation": 131380
    }
  },
  {
    "ph": "s", "id": 131380, "pid": 76337, "tid": -914061504, "ts": 1716454223408098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223462797, "dur": 6,
    "args": {
      "External id": 131389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131389,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131389, "pid": 5, "tid": 7, "ts": 1716454223462797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408136, "dur": 9,
    "args": {
      "External id": 131389,
      "cbid": 211, "correlation": 131389
    }
  },
  {
    "ph": "s", "id": 131389, "pid": 76337, "tid": -914061504, "ts": 1716454223408136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408189, "dur": 0,
    "args": {
      "External id": 131399,
      "cbid": 317, "correlation": 131399
    }
  },
  {
    "ph": "f", "id": 131399, "pid": 76337, "tid": -914061504, "ts": 1716454223408189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408190, "dur": 0,
    "args": {
      "External id": 131400,
      "cbid": 203, "correlation": 131400
    }
  },
  {
    "ph": "f", "id": 131400, "pid": 76337, "tid": -914061504, "ts": 1716454223408190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408191, "dur": 0,
    "args": {
      "External id": 131401,
      "cbid": 205, "correlation": 131401
    }
  },
  {
    "ph": "f", "id": 131401, "pid": 76337, "tid": -914061504, "ts": 1716454223408191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462804, "dur": 6,
    "args": {
      "External id": 131405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131405,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131405, "pid": 5, "tid": 7, "ts": 1716454223462804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408207, "dur": 11,
    "args": {
      "External id": 131405,
      "cbid": 211, "correlation": 131405
    }
  },
  {
    "ph": "s", "id": 131405, "pid": 76337, "tid": -914061504, "ts": 1716454223408207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223462812, "dur": 318,
    "args": {
      "External id": 131407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131407,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131407, "pid": 5, "tid": 7, "ts": 1716454223462812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408220, "dur": 5,
    "args": {
      "External id": 131407,
      "cbid": 211, "correlation": 131407
    }
  },
  {
    "ph": "s", "id": 131407, "pid": 76337, "tid": -914061504, "ts": 1716454223408220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223463133, "dur": 1,
    "args": {
      "External id": 131409,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131409,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 131409, "pid": 5, "tid": 7, "ts": 1716454223463133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408231, "dur": 9,
    "args": {
      "External id": 131409,
      "cbid": 51, "correlation": 131409
    }
  },
  {
    "ph": "s", "id": 131409, "pid": 76337, "tid": -914061504, "ts": 1716454223408231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223463136, "dur": 492,
    "args": {
      "External id": 131410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131410,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131410, "pid": 5, "tid": 7, "ts": 1716454223463136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408241, "dur": 6,
    "args": {
      "External id": 131410,
      "cbid": 211, "correlation": 131410
    }
  },
  {
    "ph": "s", "id": 131410, "pid": 76337, "tid": -914061504, "ts": 1716454223408241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223463630, "dur": 6,
    "args": {
      "External id": 131412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131412, "pid": 5, "tid": 7, "ts": 1716454223463630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408251, "dur": 5,
    "args": {
      "External id": 131412,
      "cbid": 211, "correlation": 131412
    }
  },
  {
    "ph": "s", "id": 131412, "pid": 76337, "tid": -914061504, "ts": 1716454223408251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223463637, "dur": 6,
    "args": {
      "External id": 131418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131418, "pid": 5, "tid": 7, "ts": 1716454223463637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408280, "dur": 9,
    "args": {
      "External id": 131418,
      "cbid": 211, "correlation": 131418
    }
  },
  {
    "ph": "s", "id": 131418, "pid": 76337, "tid": -914061504, "ts": 1716454223408280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223463644, "dur": 3,
    "args": {
      "External id": 131426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131426,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 131426, "pid": 5, "tid": 7, "ts": 1716454223463644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408323, "dur": 11,
    "args": {
      "External id": 131426,
      "cbid": 211, "correlation": 131426
    }
  },
  {
    "ph": "s", "id": 131426, "pid": 76337, "tid": -914061504, "ts": 1716454223408323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408388, "dur": 1,
    "args": {
      "External id": 131442,
      "cbid": 251, "correlation": 131442
    }
  },
  {
    "ph": "f", "id": 131442, "pid": 76337, "tid": -914061504, "ts": 1716454223408388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408393, "dur": 0,
    "args": {
      "External id": 131444,
      "cbid": 251, "correlation": 131444
    }
  },
  {
    "ph": "f", "id": 131444, "pid": 76337, "tid": -914061504, "ts": 1716454223408393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223463649, "dur": 11,
    "args": {
      "External id": 131445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131445,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131445, "pid": 5, "tid": 7, "ts": 1716454223463649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408394, "dur": 11,
    "args": {
      "External id": 131445,
      "cbid": 211, "correlation": 131445
    }
  },
  {
    "ph": "s", "id": 131445, "pid": 76337, "tid": -914061504, "ts": 1716454223408394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223463661, "dur": 4,
    "args": {
      "External id": 131447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131447,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131447, "pid": 5, "tid": 7, "ts": 1716454223463661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408407, "dur": 6,
    "args": {
      "External id": 131447,
      "cbid": 211, "correlation": 131447
    }
  },
  {
    "ph": "s", "id": 131447, "pid": 76337, "tid": -914061504, "ts": 1716454223408407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223463667, "dur": 6,
    "args": {
      "External id": 131457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131457, "pid": 5, "tid": 7, "ts": 1716454223463667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408464, "dur": 12,
    "args": {
      "External id": 131457,
      "cbid": 211, "correlation": 131457
    }
  },
  {
    "ph": "s", "id": 131457, "pid": 76337, "tid": -914061504, "ts": 1716454223408464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223463674, "dur": 9,
    "args": {
      "External id": 131477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131477,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131477, "pid": 5, "tid": 7, "ts": 1716454223463674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408529, "dur": 11,
    "args": {
      "External id": 131477,
      "cbid": 211, "correlation": 131477
    }
  },
  {
    "ph": "s", "id": 131477, "pid": 76337, "tid": -914061504, "ts": 1716454223408529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223463685, "dur": 4,
    "args": {
      "External id": 131489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131489,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 131489, "pid": 5, "tid": 7, "ts": 1716454223463685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408551, "dur": 6,
    "args": {
      "External id": 131489,
      "cbid": 211, "correlation": 131489
    }
  },
  {
    "ph": "s", "id": 131489, "pid": 76337, "tid": -914061504, "ts": 1716454223408551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223463690, "dur": 6,
    "args": {
      "External id": 131492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131492,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131492, "pid": 5, "tid": 7, "ts": 1716454223463690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408569, "dur": 6,
    "args": {
      "External id": 131492,
      "cbid": 211, "correlation": 131492
    }
  },
  {
    "ph": "s", "id": 131492, "pid": 76337, "tid": -914061504, "ts": 1716454223408569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223463698, "dur": 5,
    "args": {
      "External id": 131501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131501,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131501, "pid": 5, "tid": 7, "ts": 1716454223463698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408610, "dur": 10,
    "args": {
      "External id": 131501,
      "cbid": 211, "correlation": 131501
    }
  },
  {
    "ph": "s", "id": 131501, "pid": 76337, "tid": -914061504, "ts": 1716454223408610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408672, "dur": 0,
    "args": {
      "External id": 131511,
      "cbid": 317, "correlation": 131511
    }
  },
  {
    "ph": "f", "id": 131511, "pid": 76337, "tid": -914061504, "ts": 1716454223408672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408673, "dur": 0,
    "args": {
      "External id": 131512,
      "cbid": 203, "correlation": 131512
    }
  },
  {
    "ph": "f", "id": 131512, "pid": 76337, "tid": -914061504, "ts": 1716454223408673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408674, "dur": 0,
    "args": {
      "External id": 131513,
      "cbid": 205, "correlation": 131513
    }
  },
  {
    "ph": "f", "id": 131513, "pid": 76337, "tid": -914061504, "ts": 1716454223408674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223463704, "dur": 5,
    "args": {
      "External id": 131517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131517,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131517, "pid": 5, "tid": 7, "ts": 1716454223463704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408689, "dur": 12,
    "args": {
      "External id": 131517,
      "cbid": 211, "correlation": 131517
    }
  },
  {
    "ph": "s", "id": 131517, "pid": 76337, "tid": -914061504, "ts": 1716454223408689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223463710, "dur": 162,
    "args": {
      "External id": 131519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131519,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131519, "pid": 5, "tid": 7, "ts": 1716454223463710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408703, "dur": 6,
    "args": {
      "External id": 131519,
      "cbid": 211, "correlation": 131519
    }
  },
  {
    "ph": "s", "id": 131519, "pid": 76337, "tid": -914061504, "ts": 1716454223408703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223463874, "dur": 1,
    "args": {
      "External id": 131521,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131521,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 131521, "pid": 5, "tid": 7, "ts": 1716454223463874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408714, "dur": 6,
    "args": {
      "External id": 131521,
      "cbid": 51, "correlation": 131521
    }
  },
  {
    "ph": "s", "id": 131521, "pid": 76337, "tid": -914061504, "ts": 1716454223408714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223463878, "dur": 257,
    "args": {
      "External id": 131522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131522,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131522, "pid": 5, "tid": 7, "ts": 1716454223463878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408721, "dur": 6,
    "args": {
      "External id": 131522,
      "cbid": 211, "correlation": 131522
    }
  },
  {
    "ph": "s", "id": 131522, "pid": 76337, "tid": -914061504, "ts": 1716454223408721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223464136, "dur": 6,
    "args": {
      "External id": 131524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131524, "pid": 5, "tid": 7, "ts": 1716454223464136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408731, "dur": 5,
    "args": {
      "External id": 131524,
      "cbid": 211, "correlation": 131524
    }
  },
  {
    "ph": "s", "id": 131524, "pid": 76337, "tid": -914061504, "ts": 1716454223408731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223464144, "dur": 6,
    "args": {
      "External id": 131530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131530, "pid": 5, "tid": 7, "ts": 1716454223464144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408759, "dur": 8,
    "args": {
      "External id": 131530,
      "cbid": 211, "correlation": 131530
    }
  },
  {
    "ph": "s", "id": 131530, "pid": 76337, "tid": -914061504, "ts": 1716454223408759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408817, "dur": 0,
    "args": {
      "External id": 131540,
      "cbid": 317, "correlation": 131540
    }
  },
  {
    "ph": "f", "id": 131540, "pid": 76337, "tid": -914061504, "ts": 1716454223408817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408818, "dur": 0,
    "args": {
      "External id": 131541,
      "cbid": 203, "correlation": 131541
    }
  },
  {
    "ph": "f", "id": 131541, "pid": 76337, "tid": -914061504, "ts": 1716454223408818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408819, "dur": 0,
    "args": {
      "External id": 131542,
      "cbid": 205, "correlation": 131542
    }
  },
  {
    "ph": "f", "id": 131542, "pid": 76337, "tid": -914061504, "ts": 1716454223408819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223464151, "dur": 8,
    "args": {
      "External id": 131546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131546, "pid": 5, "tid": 7, "ts": 1716454223464151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408833, "dur": 13,
    "args": {
      "External id": 131546,
      "cbid": 211, "correlation": 131546
    }
  },
  {
    "ph": "s", "id": 131546, "pid": 76337, "tid": -914061504, "ts": 1716454223408833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223464160, "dur": 3,
    "args": {
      "External id": 131548,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131548,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 131548, "pid": 5, "tid": 7, "ts": 1716454223464160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408852, "dur": 14,
    "args": {
      "External id": 131548,
      "cbid": 51, "correlation": 131548
    }
  },
  {
    "ph": "s", "id": 131548, "pid": 76337, "tid": -914061504, "ts": 1716454223408852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223464164, "dur": 97,
    "args": {
      "External id": 131549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131549,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 131549, "pid": 5, "tid": 7, "ts": 1716454223464164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408867, "dur": 6,
    "args": {
      "External id": 131549,
      "cbid": 211, "correlation": 131549
    }
  },
  {
    "ph": "s", "id": 131549, "pid": 76337, "tid": -914061504, "ts": 1716454223408867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223464263, "dur": 6,
    "args": {
      "External id": 131551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131551,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131551, "pid": 5, "tid": 7, "ts": 1716454223464263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408877, "dur": 5,
    "args": {
      "External id": 131551,
      "cbid": 211, "correlation": 131551
    }
  },
  {
    "ph": "s", "id": 131551, "pid": 76337, "tid": -914061504, "ts": 1716454223408877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223464270, "dur": 6,
    "args": {
      "External id": 131557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131557, "pid": 5, "tid": 7, "ts": 1716454223464270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408904, "dur": 8,
    "args": {
      "External id": 131557,
      "cbid": 211, "correlation": 131557
    }
  },
  {
    "ph": "s", "id": 131557, "pid": 76337, "tid": -914061504, "ts": 1716454223408904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223464277, "dur": 5,
    "args": {
      "External id": 131565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131565,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131565, "pid": 5, "tid": 7, "ts": 1716454223464277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408932, "dur": 9,
    "args": {
      "External id": 131565,
      "cbid": 211, "correlation": 131565
    }
  },
  {
    "ph": "s", "id": 131565, "pid": 76337, "tid": -914061504, "ts": 1716454223408932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223464283, "dur": 4,
    "args": {
      "External id": 131573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131573,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131573, "pid": 5, "tid": 7, "ts": 1716454223464283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223408961, "dur": 8,
    "args": {
      "External id": 131573,
      "cbid": 211, "correlation": 131573
    }
  },
  {
    "ph": "s", "id": 131573, "pid": 76337, "tid": -914061504, "ts": 1716454223408961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223464289, "dur": 11,
    "args": {
      "External id": 131582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131582,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131582, "pid": 5, "tid": 7, "ts": 1716454223464289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409015, "dur": 11,
    "args": {
      "External id": 131582,
      "cbid": 211, "correlation": 131582
    }
  },
  {
    "ph": "s", "id": 131582, "pid": 76337, "tid": -914061504, "ts": 1716454223409015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223464301, "dur": 12,
    "args": {
      "External id": 131602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131602,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131602, "pid": 5, "tid": 7, "ts": 1716454223464301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409085, "dur": 12,
    "args": {
      "External id": 131602,
      "cbid": 211, "correlation": 131602
    }
  },
  {
    "ph": "s", "id": 131602, "pid": 76337, "tid": -914061504, "ts": 1716454223409085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223464315, "dur": 4,
    "args": {
      "External id": 131614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131614,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131614, "pid": 5, "tid": 7, "ts": 1716454223464315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409107, "dur": 6,
    "args": {
      "External id": 131614,
      "cbid": 211, "correlation": 131614
    }
  },
  {
    "ph": "s", "id": 131614, "pid": 76337, "tid": -914061504, "ts": 1716454223409107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223464320, "dur": 10,
    "args": {
      "External id": 131617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131617,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131617, "pid": 5, "tid": 7, "ts": 1716454223464320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409125, "dur": 6,
    "args": {
      "External id": 131617,
      "cbid": 211, "correlation": 131617
    }
  },
  {
    "ph": "s", "id": 131617, "pid": 76337, "tid": -914061504, "ts": 1716454223409125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223464332, "dur": 6,
    "args": {
      "External id": 131626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131626,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131626, "pid": 5, "tid": 7, "ts": 1716454223464332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409163, "dur": 9,
    "args": {
      "External id": 131626,
      "cbid": 211, "correlation": 131626
    }
  },
  {
    "ph": "s", "id": 131626, "pid": 76337, "tid": -914061504, "ts": 1716454223409163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409214, "dur": 0,
    "args": {
      "External id": 131636,
      "cbid": 317, "correlation": 131636
    }
  },
  {
    "ph": "f", "id": 131636, "pid": 76337, "tid": -914061504, "ts": 1716454223409214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409215, "dur": 0,
    "args": {
      "External id": 131637,
      "cbid": 203, "correlation": 131637
    }
  },
  {
    "ph": "f", "id": 131637, "pid": 76337, "tid": -914061504, "ts": 1716454223409215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409216, "dur": 0,
    "args": {
      "External id": 131638,
      "cbid": 205, "correlation": 131638
    }
  },
  {
    "ph": "f", "id": 131638, "pid": 76337, "tid": -914061504, "ts": 1716454223409216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223464340, "dur": 7,
    "args": {
      "External id": 131642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131642,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131642, "pid": 5, "tid": 7, "ts": 1716454223464340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409229, "dur": 12,
    "args": {
      "External id": 131642,
      "cbid": 211, "correlation": 131642
    }
  },
  {
    "ph": "s", "id": 131642, "pid": 76337, "tid": -914061504, "ts": 1716454223409229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223464348, "dur": 318,
    "args": {
      "External id": 131644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131644,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131644, "pid": 5, "tid": 7, "ts": 1716454223464348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409244, "dur": 6,
    "args": {
      "External id": 131644,
      "cbid": 211, "correlation": 131644
    }
  },
  {
    "ph": "s", "id": 131644, "pid": 76337, "tid": -914061504, "ts": 1716454223409244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223464667, "dur": 1,
    "args": {
      "External id": 131646,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131646,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 131646, "pid": 5, "tid": 7, "ts": 1716454223464667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409255, "dur": 6,
    "args": {
      "External id": 131646,
      "cbid": 51, "correlation": 131646
    }
  },
  {
    "ph": "s", "id": 131646, "pid": 76337, "tid": -914061504, "ts": 1716454223409255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223464671, "dur": 492,
    "args": {
      "External id": 131647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131647,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131647, "pid": 5, "tid": 7, "ts": 1716454223464671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409262, "dur": 6,
    "args": {
      "External id": 131647,
      "cbid": 211, "correlation": 131647
    }
  },
  {
    "ph": "s", "id": 131647, "pid": 76337, "tid": -914061504, "ts": 1716454223409262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465165, "dur": 6,
    "args": {
      "External id": 131649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131649,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131649, "pid": 5, "tid": 7, "ts": 1716454223465165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409271, "dur": 5,
    "args": {
      "External id": 131649,
      "cbid": 211, "correlation": 131649
    }
  },
  {
    "ph": "s", "id": 131649, "pid": 76337, "tid": -914061504, "ts": 1716454223409271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223465172, "dur": 6,
    "args": {
      "External id": 131655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131655, "pid": 5, "tid": 7, "ts": 1716454223465172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409300, "dur": 8,
    "args": {
      "External id": 131655,
      "cbid": 211, "correlation": 131655
    }
  },
  {
    "ph": "s", "id": 131655, "pid": 76337, "tid": -914061504, "ts": 1716454223409300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223465179, "dur": 3,
    "args": {
      "External id": 131663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131663,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 131663, "pid": 5, "tid": 7, "ts": 1716454223465179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409344, "dur": 9,
    "args": {
      "External id": 131663,
      "cbid": 211, "correlation": 131663
    }
  },
  {
    "ph": "s", "id": 131663, "pid": 76337, "tid": -914061504, "ts": 1716454223409344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409405, "dur": 1,
    "args": {
      "External id": 131679,
      "cbid": 251, "correlation": 131679
    }
  },
  {
    "ph": "f", "id": 131679, "pid": 76337, "tid": -914061504, "ts": 1716454223409405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409410, "dur": 0,
    "args": {
      "External id": 131681,
      "cbid": 251, "correlation": 131681
    }
  },
  {
    "ph": "f", "id": 131681, "pid": 76337, "tid": -914061504, "ts": 1716454223409410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223465184, "dur": 13,
    "args": {
      "External id": 131682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131682,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131682, "pid": 5, "tid": 7, "ts": 1716454223465184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409412, "dur": 11,
    "args": {
      "External id": 131682,
      "cbid": 211, "correlation": 131682
    }
  },
  {
    "ph": "s", "id": 131682, "pid": 76337, "tid": -914061504, "ts": 1716454223409412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223465198, "dur": 5,
    "args": {
      "External id": 131684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131684,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131684, "pid": 5, "tid": 7, "ts": 1716454223465198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409425, "dur": 5,
    "args": {
      "External id": 131684,
      "cbid": 211, "correlation": 131684
    }
  },
  {
    "ph": "s", "id": 131684, "pid": 76337, "tid": -914061504, "ts": 1716454223409425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223465204, "dur": 6,
    "args": {
      "External id": 131694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131694, "pid": 5, "tid": 7, "ts": 1716454223465204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409482, "dur": 13,
    "args": {
      "External id": 131694,
      "cbid": 211, "correlation": 131694
    }
  },
  {
    "ph": "s", "id": 131694, "pid": 76337, "tid": -914061504, "ts": 1716454223409482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465212, "dur": 9,
    "args": {
      "External id": 131714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131714,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131714, "pid": 5, "tid": 7, "ts": 1716454223465212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409549, "dur": 11,
    "args": {
      "External id": 131714,
      "cbid": 211, "correlation": 131714
    }
  },
  {
    "ph": "s", "id": 131714, "pid": 76337, "tid": -914061504, "ts": 1716454223409549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223465222, "dur": 3,
    "args": {
      "External id": 131726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131726,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 131726, "pid": 5, "tid": 7, "ts": 1716454223465222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409569, "dur": 6,
    "args": {
      "External id": 131726,
      "cbid": 211, "correlation": 131726
    }
  },
  {
    "ph": "s", "id": 131726, "pid": 76337, "tid": -914061504, "ts": 1716454223409569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223465227, "dur": 7,
    "args": {
      "External id": 131729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131729,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131729, "pid": 5, "tid": 7, "ts": 1716454223465227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409588, "dur": 7,
    "args": {
      "External id": 131729,
      "cbid": 211, "correlation": 131729
    }
  },
  {
    "ph": "s", "id": 131729, "pid": 76337, "tid": -914061504, "ts": 1716454223409588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223465235, "dur": 5,
    "args": {
      "External id": 131738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131738,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131738, "pid": 5, "tid": 7, "ts": 1716454223465235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409628, "dur": 10,
    "args": {
      "External id": 131738,
      "cbid": 211, "correlation": 131738
    }
  },
  {
    "ph": "s", "id": 131738, "pid": 76337, "tid": -914061504, "ts": 1716454223409628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409691, "dur": 0,
    "args": {
      "External id": 131748,
      "cbid": 317, "correlation": 131748
    }
  },
  {
    "ph": "f", "id": 131748, "pid": 76337, "tid": -914061504, "ts": 1716454223409691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409692, "dur": 0,
    "args": {
      "External id": 131749,
      "cbid": 203, "correlation": 131749
    }
  },
  {
    "ph": "f", "id": 131749, "pid": 76337, "tid": -914061504, "ts": 1716454223409692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409693, "dur": 0,
    "args": {
      "External id": 131750,
      "cbid": 205, "correlation": 131750
    }
  },
  {
    "ph": "f", "id": 131750, "pid": 76337, "tid": -914061504, "ts": 1716454223409693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465241, "dur": 5,
    "args": {
      "External id": 131754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131754,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131754, "pid": 5, "tid": 7, "ts": 1716454223465241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409707, "dur": 12,
    "args": {
      "External id": 131754,
      "cbid": 211, "correlation": 131754
    }
  },
  {
    "ph": "s", "id": 131754, "pid": 76337, "tid": -914061504, "ts": 1716454223409707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465247, "dur": 161,
    "args": {
      "External id": 131756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131756,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131756, "pid": 5, "tid": 7, "ts": 1716454223465247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409721, "dur": 5,
    "args": {
      "External id": 131756,
      "cbid": 211, "correlation": 131756
    }
  },
  {
    "ph": "s", "id": 131756, "pid": 76337, "tid": -914061504, "ts": 1716454223409721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223465410, "dur": 1,
    "args": {
      "External id": 131758,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131758,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 131758, "pid": 5, "tid": 7, "ts": 1716454223465410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409731, "dur": 7,
    "args": {
      "External id": 131758,
      "cbid": 51, "correlation": 131758
    }
  },
  {
    "ph": "s", "id": 131758, "pid": 76337, "tid": -914061504, "ts": 1716454223409731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223465414, "dur": 257,
    "args": {
      "External id": 131759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131759,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131759, "pid": 5, "tid": 7, "ts": 1716454223465414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409739, "dur": 7,
    "args": {
      "External id": 131759,
      "cbid": 211, "correlation": 131759
    }
  },
  {
    "ph": "s", "id": 131759, "pid": 76337, "tid": -914061504, "ts": 1716454223409739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465673, "dur": 6,
    "args": {
      "External id": 131761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131761, "pid": 5, "tid": 7, "ts": 1716454223465673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409750, "dur": 5,
    "args": {
      "External id": 131761,
      "cbid": 211, "correlation": 131761
    }
  },
  {
    "ph": "s", "id": 131761, "pid": 76337, "tid": -914061504, "ts": 1716454223409750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223465680, "dur": 6,
    "args": {
      "External id": 131767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131767, "pid": 5, "tid": 7, "ts": 1716454223465680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409778, "dur": 8,
    "args": {
      "External id": 131767,
      "cbid": 211, "correlation": 131767
    }
  },
  {
    "ph": "s", "id": 131767, "pid": 76337, "tid": -914061504, "ts": 1716454223409778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409836, "dur": 0,
    "args": {
      "External id": 131777,
      "cbid": 317, "correlation": 131777
    }
  },
  {
    "ph": "f", "id": 131777, "pid": 76337, "tid": -914061504, "ts": 1716454223409836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409837, "dur": 0,
    "args": {
      "External id": 131778,
      "cbid": 203, "correlation": 131778
    }
  },
  {
    "ph": "f", "id": 131778, "pid": 76337, "tid": -914061504, "ts": 1716454223409837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409838, "dur": 0,
    "args": {
      "External id": 131779,
      "cbid": 205, "correlation": 131779
    }
  },
  {
    "ph": "f", "id": 131779, "pid": 76337, "tid": -914061504, "ts": 1716454223409838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465687, "dur": 8,
    "args": {
      "External id": 131783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131783,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131783, "pid": 5, "tid": 7, "ts": 1716454223465687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409850, "dur": 12,
    "args": {
      "External id": 131783,
      "cbid": 211, "correlation": 131783
    }
  },
  {
    "ph": "s", "id": 131783, "pid": 76337, "tid": -914061504, "ts": 1716454223409850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223465696, "dur": 3,
    "args": {
      "External id": 131785,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131785,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4705882352941178
    }
  },
  {
    "ph": "f", "id": 131785, "pid": 5, "tid": 7, "ts": 1716454223465696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409867, "dur": 9,
    "args": {
      "External id": 131785,
      "cbid": 51, "correlation": 131785
    }
  },
  {
    "ph": "s", "id": 131785, "pid": 76337, "tid": -914061504, "ts": 1716454223409867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223465700, "dur": 95,
    "args": {
      "External id": 131786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131786,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 131786, "pid": 5, "tid": 7, "ts": 1716454223465700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409877, "dur": 6,
    "args": {
      "External id": 131786,
      "cbid": 211, "correlation": 131786
    }
  },
  {
    "ph": "s", "id": 131786, "pid": 76337, "tid": -914061504, "ts": 1716454223409877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465797, "dur": 5,
    "args": {
      "External id": 131788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131788,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131788, "pid": 5, "tid": 7, "ts": 1716454223465797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409886, "dur": 5,
    "args": {
      "External id": 131788,
      "cbid": 211, "correlation": 131788
    }
  },
  {
    "ph": "s", "id": 131788, "pid": 76337, "tid": -914061504, "ts": 1716454223409886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223465804, "dur": 6,
    "args": {
      "External id": 131794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131794,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131794, "pid": 5, "tid": 7, "ts": 1716454223465804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409913, "dur": 8,
    "args": {
      "External id": 131794,
      "cbid": 211, "correlation": 131794
    }
  },
  {
    "ph": "s", "id": 131794, "pid": 76337, "tid": -914061504, "ts": 1716454223409913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223465811, "dur": 5,
    "args": {
      "External id": 131802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131802,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131802, "pid": 5, "tid": 7, "ts": 1716454223465811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409942, "dur": 9,
    "args": {
      "External id": 131802,
      "cbid": 211, "correlation": 131802
    }
  },
  {
    "ph": "s", "id": 131802, "pid": 76337, "tid": -914061504, "ts": 1716454223409942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223465817, "dur": 4,
    "args": {
      "External id": 131810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131810,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131810, "pid": 5, "tid": 7, "ts": 1716454223465817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223409971, "dur": 17,
    "args": {
      "External id": 131810,
      "cbid": 211, "correlation": 131810
    }
  },
  {
    "ph": "s", "id": 131810, "pid": 76337, "tid": -914061504, "ts": 1716454223409971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223465823, "dur": 11,
    "args": {
      "External id": 131819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131819,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131819, "pid": 5, "tid": 7, "ts": 1716454223465823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410025, "dur": 10,
    "args": {
      "External id": 131819,
      "cbid": 211, "correlation": 131819
    }
  },
  {
    "ph": "s", "id": 131819, "pid": 76337, "tid": -914061504, "ts": 1716454223410025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465835, "dur": 13,
    "args": {
      "External id": 131839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131839,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131839, "pid": 5, "tid": 7, "ts": 1716454223465835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410095, "dur": 12,
    "args": {
      "External id": 131839,
      "cbid": 211, "correlation": 131839
    }
  },
  {
    "ph": "s", "id": 131839, "pid": 76337, "tid": -914061504, "ts": 1716454223410095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223465849, "dur": 4,
    "args": {
      "External id": 131851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131851,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131851, "pid": 5, "tid": 7, "ts": 1716454223465849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410117, "dur": 6,
    "args": {
      "External id": 131851,
      "cbid": 211, "correlation": 131851
    }
  },
  {
    "ph": "s", "id": 131851, "pid": 76337, "tid": -914061504, "ts": 1716454223410117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223465854, "dur": 11,
    "args": {
      "External id": 131854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131854, "pid": 5, "tid": 7, "ts": 1716454223465854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410135, "dur": 7,
    "args": {
      "External id": 131854,
      "cbid": 211, "correlation": 131854
    }
  },
  {
    "ph": "s", "id": 131854, "pid": 76337, "tid": -914061504, "ts": 1716454223410135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223465866, "dur": 6,
    "args": {
      "External id": 131863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131863,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131863, "pid": 5, "tid": 7, "ts": 1716454223465866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410173, "dur": 9,
    "args": {
      "External id": 131863,
      "cbid": 211, "correlation": 131863
    }
  },
  {
    "ph": "s", "id": 131863, "pid": 76337, "tid": -914061504, "ts": 1716454223410173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410224, "dur": 0,
    "args": {
      "External id": 131873,
      "cbid": 317, "correlation": 131873
    }
  },
  {
    "ph": "f", "id": 131873, "pid": 76337, "tid": -914061504, "ts": 1716454223410224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410225, "dur": 0,
    "args": {
      "External id": 131874,
      "cbid": 203, "correlation": 131874
    }
  },
  {
    "ph": "f", "id": 131874, "pid": 76337, "tid": -914061504, "ts": 1716454223410225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410226, "dur": 0,
    "args": {
      "External id": 131875,
      "cbid": 205, "correlation": 131875
    }
  },
  {
    "ph": "f", "id": 131875, "pid": 76337, "tid": -914061504, "ts": 1716454223410226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465874, "dur": 7,
    "args": {
      "External id": 131879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131879,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131879, "pid": 5, "tid": 7, "ts": 1716454223465874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410239, "dur": 12,
    "args": {
      "External id": 131879,
      "cbid": 211, "correlation": 131879
    }
  },
  {
    "ph": "s", "id": 131879, "pid": 76337, "tid": -914061504, "ts": 1716454223410239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223465882, "dur": 318,
    "args": {
      "External id": 131881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131881, "pid": 5, "tid": 7, "ts": 1716454223465882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410254, "dur": 6,
    "args": {
      "External id": 131881,
      "cbid": 211, "correlation": 131881
    }
  },
  {
    "ph": "s", "id": 131881, "pid": 76337, "tid": -914061504, "ts": 1716454223410254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223466202, "dur": 1,
    "args": {
      "External id": 131883,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131883,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 131883, "pid": 5, "tid": 7, "ts": 1716454223466202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410265, "dur": 6,
    "args": {
      "External id": 131883,
      "cbid": 51, "correlation": 131883
    }
  },
  {
    "ph": "s", "id": 131883, "pid": 76337, "tid": -914061504, "ts": 1716454223410265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223466205, "dur": 492,
    "args": {
      "External id": 131884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131884,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131884, "pid": 5, "tid": 7, "ts": 1716454223466205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410272, "dur": 6,
    "args": {
      "External id": 131884,
      "cbid": 211, "correlation": 131884
    }
  },
  {
    "ph": "s", "id": 131884, "pid": 76337, "tid": -914061504, "ts": 1716454223410272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223466699, "dur": 5,
    "args": {
      "External id": 131886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131886,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131886, "pid": 5, "tid": 7, "ts": 1716454223466699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410282, "dur": 5,
    "args": {
      "External id": 131886,
      "cbid": 211, "correlation": 131886
    }
  },
  {
    "ph": "s", "id": 131886, "pid": 76337, "tid": -914061504, "ts": 1716454223410282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223466705, "dur": 7,
    "args": {
      "External id": 131892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131892, "pid": 5, "tid": 7, "ts": 1716454223466705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410310, "dur": 9,
    "args": {
      "External id": 131892,
      "cbid": 211, "correlation": 131892
    }
  },
  {
    "ph": "s", "id": 131892, "pid": 76337, "tid": -914061504, "ts": 1716454223410310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223466713, "dur": 3,
    "args": {
      "External id": 131900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131900,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 131900, "pid": 5, "tid": 7, "ts": 1716454223466713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410354, "dur": 9,
    "args": {
      "External id": 131900,
      "cbid": 211, "correlation": 131900
    }
  },
  {
    "ph": "s", "id": 131900, "pid": 76337, "tid": -914061504, "ts": 1716454223410354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410415, "dur": 1,
    "args": {
      "External id": 131916,
      "cbid": 251, "correlation": 131916
    }
  },
  {
    "ph": "f", "id": 131916, "pid": 76337, "tid": -914061504, "ts": 1716454223410415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410421, "dur": 0,
    "args": {
      "External id": 131918,
      "cbid": 251, "correlation": 131918
    }
  },
  {
    "ph": "f", "id": 131918, "pid": 76337, "tid": -914061504, "ts": 1716454223410421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223466718, "dur": 12,
    "args": {
      "External id": 131919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131919,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131919, "pid": 5, "tid": 7, "ts": 1716454223466718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410422, "dur": 11,
    "args": {
      "External id": 131919,
      "cbid": 211, "correlation": 131919
    }
  },
  {
    "ph": "s", "id": 131919, "pid": 76337, "tid": -914061504, "ts": 1716454223410422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223466732, "dur": 5,
    "args": {
      "External id": 131921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131921,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131921, "pid": 5, "tid": 7, "ts": 1716454223466732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410435, "dur": 5,
    "args": {
      "External id": 131921,
      "cbid": 211, "correlation": 131921
    }
  },
  {
    "ph": "s", "id": 131921, "pid": 76337, "tid": -914061504, "ts": 1716454223410435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223466738, "dur": 6,
    "args": {
      "External id": 131931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131931, "pid": 5, "tid": 7, "ts": 1716454223466738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410491, "dur": 13,
    "args": {
      "External id": 131931,
      "cbid": 211, "correlation": 131931
    }
  },
  {
    "ph": "s", "id": 131931, "pid": 76337, "tid": -914061504, "ts": 1716454223410491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223466745, "dur": 9,
    "args": {
      "External id": 131951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131951,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 131951, "pid": 5, "tid": 7, "ts": 1716454223466745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410558, "dur": 11,
    "args": {
      "External id": 131951,
      "cbid": 211, "correlation": 131951
    }
  },
  {
    "ph": "s", "id": 131951, "pid": 76337, "tid": -914061504, "ts": 1716454223410558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223466756, "dur": 4,
    "args": {
      "External id": 131963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131963,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 131963, "pid": 5, "tid": 7, "ts": 1716454223466756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410578, "dur": 6,
    "args": {
      "External id": 131963,
      "cbid": 211, "correlation": 131963
    }
  },
  {
    "ph": "s", "id": 131963, "pid": 76337, "tid": -914061504, "ts": 1716454223410578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223466761, "dur": 7,
    "args": {
      "External id": 131966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131966,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131966, "pid": 5, "tid": 7, "ts": 1716454223466761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410598, "dur": 7,
    "args": {
      "External id": 131966,
      "cbid": 211, "correlation": 131966
    }
  },
  {
    "ph": "s", "id": 131966, "pid": 76337, "tid": -914061504, "ts": 1716454223410598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223466768, "dur": 5,
    "args": {
      "External id": 131975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131975,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 131975, "pid": 5, "tid": 7, "ts": 1716454223466768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410637, "dur": 10,
    "args": {
      "External id": 131975,
      "cbid": 211, "correlation": 131975
    }
  },
  {
    "ph": "s", "id": 131975, "pid": 76337, "tid": -914061504, "ts": 1716454223410637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410700, "dur": 0,
    "args": {
      "External id": 131985,
      "cbid": 317, "correlation": 131985
    }
  },
  {
    "ph": "f", "id": 131985, "pid": 76337, "tid": -914061504, "ts": 1716454223410700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410702, "dur": 0,
    "args": {
      "External id": 131986,
      "cbid": 203, "correlation": 131986
    }
  },
  {
    "ph": "f", "id": 131986, "pid": 76337, "tid": -914061504, "ts": 1716454223410702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410702, "dur": 0,
    "args": {
      "External id": 131987,
      "cbid": 205, "correlation": 131987
    }
  },
  {
    "ph": "f", "id": 131987, "pid": 76337, "tid": -914061504, "ts": 1716454223410702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223466774, "dur": 5,
    "args": {
      "External id": 131991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131991,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131991, "pid": 5, "tid": 7, "ts": 1716454223466774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410716, "dur": 12,
    "args": {
      "External id": 131991,
      "cbid": 211, "correlation": 131991
    }
  },
  {
    "ph": "s", "id": 131991, "pid": 76337, "tid": -914061504, "ts": 1716454223410716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223466781, "dur": 162,
    "args": {
      "External id": 131993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131993,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131993, "pid": 5, "tid": 7, "ts": 1716454223466781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410731, "dur": 5,
    "args": {
      "External id": 131993,
      "cbid": 211, "correlation": 131993
    }
  },
  {
    "ph": "s", "id": 131993, "pid": 76337, "tid": -914061504, "ts": 1716454223410731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223466945, "dur": 1,
    "args": {
      "External id": 131995,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 131995,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 131995, "pid": 5, "tid": 7, "ts": 1716454223466945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410741, "dur": 7,
    "args": {
      "External id": 131995,
      "cbid": 51, "correlation": 131995
    }
  },
  {
    "ph": "s", "id": 131995, "pid": 76337, "tid": -914061504, "ts": 1716454223410741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223466948, "dur": 257,
    "args": {
      "External id": 131996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131996,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 131996, "pid": 5, "tid": 7, "ts": 1716454223466948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410750, "dur": 6,
    "args": {
      "External id": 131996,
      "cbid": 211, "correlation": 131996
    }
  },
  {
    "ph": "s", "id": 131996, "pid": 76337, "tid": -914061504, "ts": 1716454223410750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223467207, "dur": 6,
    "args": {
      "External id": 131998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 131998,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 131998, "pid": 5, "tid": 7, "ts": 1716454223467207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410759, "dur": 5,
    "args": {
      "External id": 131998,
      "cbid": 211, "correlation": 131998
    }
  },
  {
    "ph": "s", "id": 131998, "pid": 76337, "tid": -914061504, "ts": 1716454223410759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223467214, "dur": 6,
    "args": {
      "External id": 132004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 132004, "pid": 5, "tid": 7, "ts": 1716454223467214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410788, "dur": 9,
    "args": {
      "External id": 132004,
      "cbid": 211, "correlation": 132004
    }
  },
  {
    "ph": "s", "id": 132004, "pid": 76337, "tid": -914061504, "ts": 1716454223410788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410846, "dur": 0,
    "args": {
      "External id": 132014,
      "cbid": 317, "correlation": 132014
    }
  },
  {
    "ph": "f", "id": 132014, "pid": 76337, "tid": -914061504, "ts": 1716454223410846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410847, "dur": 0,
    "args": {
      "External id": 132015,
      "cbid": 203, "correlation": 132015
    }
  },
  {
    "ph": "f", "id": 132015, "pid": 76337, "tid": -914061504, "ts": 1716454223410847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410848, "dur": 0,
    "args": {
      "External id": 132016,
      "cbid": 205, "correlation": 132016
    }
  },
  {
    "ph": "f", "id": 132016, "pid": 76337, "tid": -914061504, "ts": 1716454223410848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223467221, "dur": 8,
    "args": {
      "External id": 132020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132020, "pid": 5, "tid": 7, "ts": 1716454223467221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410860, "dur": 12,
    "args": {
      "External id": 132020,
      "cbid": 211, "correlation": 132020
    }
  },
  {
    "ph": "s", "id": 132020, "pid": 76337, "tid": -914061504, "ts": 1716454223410860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223467231, "dur": 3,
    "args": {
      "External id": 132022,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132022,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5151515151515151
    }
  },
  {
    "ph": "f", "id": 132022, "pid": 5, "tid": 7, "ts": 1716454223467231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410877, "dur": 9,
    "args": {
      "External id": 132022,
      "cbid": 51, "correlation": 132022
    }
  },
  {
    "ph": "s", "id": 132022, "pid": 76337, "tid": -914061504, "ts": 1716454223410877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223467235, "dur": 94,
    "args": {
      "External id": 132023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132023,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 132023, "pid": 5, "tid": 7, "ts": 1716454223467235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410887, "dur": 6,
    "args": {
      "External id": 132023,
      "cbid": 211, "correlation": 132023
    }
  },
  {
    "ph": "s", "id": 132023, "pid": 76337, "tid": -914061504, "ts": 1716454223410887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223467330, "dur": 5,
    "args": {
      "External id": 132025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132025,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132025, "pid": 5, "tid": 7, "ts": 1716454223467330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410896, "dur": 6,
    "args": {
      "External id": 132025,
      "cbid": 211, "correlation": 132025
    }
  },
  {
    "ph": "s", "id": 132025, "pid": 76337, "tid": -914061504, "ts": 1716454223410896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223467337, "dur": 6,
    "args": {
      "External id": 132031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 132031, "pid": 5, "tid": 7, "ts": 1716454223467337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410924, "dur": 8,
    "args": {
      "External id": 132031,
      "cbid": 211, "correlation": 132031
    }
  },
  {
    "ph": "s", "id": 132031, "pid": 76337, "tid": -914061504, "ts": 1716454223410924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223467344, "dur": 5,
    "args": {
      "External id": 132039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132039,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 132039, "pid": 5, "tid": 7, "ts": 1716454223467344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410954, "dur": 7,
    "args": {
      "External id": 132039,
      "cbid": 211, "correlation": 132039
    }
  },
  {
    "ph": "s", "id": 132039, "pid": 76337, "tid": -914061504, "ts": 1716454223410954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223467350, "dur": 4,
    "args": {
      "External id": 132047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132047,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 132047, "pid": 5, "tid": 7, "ts": 1716454223467350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223410990, "dur": 9,
    "args": {
      "External id": 132047,
      "cbid": 211, "correlation": 132047
    }
  },
  {
    "ph": "s", "id": 132047, "pid": 76337, "tid": -914061504, "ts": 1716454223410990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223467356, "dur": 14,
    "args": {
      "External id": 132058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132058,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132058, "pid": 5, "tid": 7, "ts": 1716454223467356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411072, "dur": 13,
    "args": {
      "External id": 132058,
      "cbid": 211, "correlation": 132058
    }
  },
  {
    "ph": "s", "id": 132058, "pid": 76337, "tid": -914061504, "ts": 1716454223411072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411127, "dur": 0,
    "args": {
      "External id": 132068,
      "cbid": 317, "correlation": 132068
    }
  },
  {
    "ph": "f", "id": 132068, "pid": 76337, "tid": -914061504, "ts": 1716454223411127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411128, "dur": 0,
    "args": {
      "External id": 132069,
      "cbid": 203, "correlation": 132069
    }
  },
  {
    "ph": "f", "id": 132069, "pid": 76337, "tid": -914061504, "ts": 1716454223411128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411129, "dur": 0,
    "args": {
      "External id": 132070,
      "cbid": 205, "correlation": 132070
    }
  },
  {
    "ph": "f", "id": 132070, "pid": 76337, "tid": -914061504, "ts": 1716454223411129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223467371, "dur": 9,
    "args": {
      "External id": 132074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132074,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132074, "pid": 5, "tid": 7, "ts": 1716454223467371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411142, "dur": 12,
    "args": {
      "External id": 132074,
      "cbid": 211, "correlation": 132074
    }
  },
  {
    "ph": "s", "id": 132074, "pid": 76337, "tid": -914061504, "ts": 1716454223411142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223467381, "dur": 162,
    "args": {
      "External id": 132076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132076, "pid": 5, "tid": 7, "ts": 1716454223467381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411157, "dur": 5,
    "args": {
      "External id": 132076,
      "cbid": 211, "correlation": 132076
    }
  },
  {
    "ph": "s", "id": 132076, "pid": 76337, "tid": -914061504, "ts": 1716454223411157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223467546, "dur": 1,
    "args": {
      "External id": 132078,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132078,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 132078, "pid": 5, "tid": 7, "ts": 1716454223467546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411167, "dur": 6,
    "args": {
      "External id": 132078,
      "cbid": 51, "correlation": 132078
    }
  },
  {
    "ph": "s", "id": 132078, "pid": 76337, "tid": -914061504, "ts": 1716454223411167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223467550, "dur": 647,
    "args": {
      "External id": 132079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132079,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132079, "pid": 5, "tid": 7, "ts": 1716454223467550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411174, "dur": 6,
    "args": {
      "External id": 132079,
      "cbid": 211, "correlation": 132079
    }
  },
  {
    "ph": "s", "id": 132079, "pid": 76337, "tid": -914061504, "ts": 1716454223411174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223468198, "dur": 12,
    "args": {
      "External id": 132081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132081,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132081, "pid": 5, "tid": 7, "ts": 1716454223468198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411184, "dur": 5,
    "args": {
      "External id": 132081,
      "cbid": 211, "correlation": 132081
    }
  },
  {
    "ph": "s", "id": 132081, "pid": 76337, "tid": -914061504, "ts": 1716454223411184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223468211, "dur": 14,
    "args": {
      "External id": 132087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132087, "pid": 5, "tid": 7, "ts": 1716454223468211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411213, "dur": 8,
    "args": {
      "External id": 132087,
      "cbid": 211, "correlation": 132087
    }
  },
  {
    "ph": "s", "id": 132087, "pid": 76337, "tid": -914061504, "ts": 1716454223411213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223468227, "dur": 31,
    "args": {
      "External id": 132096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132096, "pid": 5, "tid": 7, "ts": 1716454223468227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411303, "dur": 13,
    "args": {
      "External id": 132096,
      "cbid": 211, "correlation": 132096
    }
  },
  {
    "ph": "s", "id": 132096, "pid": 76337, "tid": -914061504, "ts": 1716454223411303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223468259, "dur": 30,
    "args": {
      "External id": 132116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132116,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 132116, "pid": 5, "tid": 7, "ts": 1716454223468259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411370, "dur": 11,
    "args": {
      "External id": 132116,
      "cbid": 211, "correlation": 132116
    }
  },
  {
    "ph": "s", "id": 132116, "pid": 76337, "tid": -914061504, "ts": 1716454223411370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223468291, "dur": 4,
    "args": {
      "External id": 132128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132128,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132128, "pid": 5, "tid": 7, "ts": 1716454223468291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411391, "dur": 6,
    "args": {
      "External id": 132128,
      "cbid": 211, "correlation": 132128
    }
  },
  {
    "ph": "s", "id": 132128, "pid": 76337, "tid": -914061504, "ts": 1716454223411391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223468296, "dur": 30,
    "args": {
      "External id": 132131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132131, "pid": 5, "tid": 7, "ts": 1716454223468296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411409, "dur": 6,
    "args": {
      "External id": 132131,
      "cbid": 211, "correlation": 132131
    }
  },
  {
    "ph": "s", "id": 132131, "pid": 76337, "tid": -914061504, "ts": 1716454223411409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223468327, "dur": 21,
    "args": {
      "External id": 132140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132140,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132140, "pid": 5, "tid": 7, "ts": 1716454223468327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411447, "dur": 10,
    "args": {
      "External id": 132140,
      "cbid": 211, "correlation": 132140
    }
  },
  {
    "ph": "s", "id": 132140, "pid": 76337, "tid": -914061504, "ts": 1716454223411447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411499, "dur": 0,
    "args": {
      "External id": 132150,
      "cbid": 317, "correlation": 132150
    }
  },
  {
    "ph": "f", "id": 132150, "pid": 76337, "tid": -914061504, "ts": 1716454223411499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411500, "dur": 0,
    "args": {
      "External id": 132151,
      "cbid": 203, "correlation": 132151
    }
  },
  {
    "ph": "f", "id": 132151, "pid": 76337, "tid": -914061504, "ts": 1716454223411500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411501, "dur": 0,
    "args": {
      "External id": 132152,
      "cbid": 205, "correlation": 132152
    }
  },
  {
    "ph": "f", "id": 132152, "pid": 76337, "tid": -914061504, "ts": 1716454223411501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223468350, "dur": 22,
    "args": {
      "External id": 132156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132156, "pid": 5, "tid": 7, "ts": 1716454223468350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411516, "dur": 11,
    "args": {
      "External id": 132156,
      "cbid": 211, "correlation": 132156
    }
  },
  {
    "ph": "s", "id": 132156, "pid": 76337, "tid": -914061504, "ts": 1716454223411516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223468373, "dur": 318,
    "args": {
      "External id": 132158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132158,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132158, "pid": 5, "tid": 7, "ts": 1716454223468373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411530, "dur": 5,
    "args": {
      "External id": 132158,
      "cbid": 211, "correlation": 132158
    }
  },
  {
    "ph": "s", "id": 132158, "pid": 76337, "tid": -914061504, "ts": 1716454223411530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223468693, "dur": 1,
    "args": {
      "External id": 132160,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132160,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 132160, "pid": 5, "tid": 7, "ts": 1716454223468693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411541, "dur": 6,
    "args": {
      "External id": 132160,
      "cbid": 51, "correlation": 132160
    }
  },
  {
    "ph": "s", "id": 132160, "pid": 76337, "tid": -914061504, "ts": 1716454223411541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223468697, "dur": 1231,
    "args": {
      "External id": 132161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132161,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132161, "pid": 5, "tid": 7, "ts": 1716454223468697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411548, "dur": 7,
    "args": {
      "External id": 132161,
      "cbid": 211, "correlation": 132161
    }
  },
  {
    "ph": "s", "id": 132161, "pid": 76337, "tid": -914061504, "ts": 1716454223411548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223469929, "dur": 13,
    "args": {
      "External id": 132163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132163,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132163, "pid": 5, "tid": 7, "ts": 1716454223469929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411558, "dur": 5,
    "args": {
      "External id": 132163,
      "cbid": 211, "correlation": 132163
    }
  },
  {
    "ph": "s", "id": 132163, "pid": 76337, "tid": -914061504, "ts": 1716454223411558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223469943, "dur": 14,
    "args": {
      "External id": 132169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132169, "pid": 5, "tid": 7, "ts": 1716454223469943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411586, "dur": 8,
    "args": {
      "External id": 132169,
      "cbid": 211, "correlation": 132169
    }
  },
  {
    "ph": "s", "id": 132169, "pid": 76337, "tid": -914061504, "ts": 1716454223411586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223469959, "dur": 3,
    "args": {
      "External id": 132177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132177,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 132177, "pid": 5, "tid": 7, "ts": 1716454223469959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411631, "dur": 9,
    "args": {
      "External id": 132177,
      "cbid": 211, "correlation": 132177
    }
  },
  {
    "ph": "s", "id": 132177, "pid": 76337, "tid": -914061504, "ts": 1716454223411631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411693, "dur": 1,
    "args": {
      "External id": 132193,
      "cbid": 251, "correlation": 132193
    }
  },
  {
    "ph": "f", "id": 132193, "pid": 76337, "tid": -914061504, "ts": 1716454223411693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411698, "dur": 0,
    "args": {
      "External id": 132195,
      "cbid": 251, "correlation": 132195
    }
  },
  {
    "ph": "f", "id": 132195, "pid": 76337, "tid": -914061504, "ts": 1716454223411698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223469964, "dur": 12,
    "args": {
      "External id": 132196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132196,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132196, "pid": 5, "tid": 7, "ts": 1716454223469964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411700, "dur": 12,
    "args": {
      "External id": 132196,
      "cbid": 211, "correlation": 132196
    }
  },
  {
    "ph": "s", "id": 132196, "pid": 76337, "tid": -914061504, "ts": 1716454223411700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223469977, "dur": 5,
    "args": {
      "External id": 132198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132198,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132198, "pid": 5, "tid": 7, "ts": 1716454223469977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411714, "dur": 5,
    "args": {
      "External id": 132198,
      "cbid": 211, "correlation": 132198
    }
  },
  {
    "ph": "s", "id": 132198, "pid": 76337, "tid": -914061504, "ts": 1716454223411714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223469983, "dur": 17,
    "args": {
      "External id": 132208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132208,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132208, "pid": 5, "tid": 7, "ts": 1716454223469983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411770, "dur": 12,
    "args": {
      "External id": 132208,
      "cbid": 211, "correlation": 132208
    }
  },
  {
    "ph": "s", "id": 132208, "pid": 76337, "tid": -914061504, "ts": 1716454223411770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223470001, "dur": 17,
    "args": {
      "External id": 132228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132228,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 132228, "pid": 5, "tid": 7, "ts": 1716454223470001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411836, "dur": 10,
    "args": {
      "External id": 132228,
      "cbid": 211, "correlation": 132228
    }
  },
  {
    "ph": "s", "id": 132228, "pid": 76337, "tid": -914061504, "ts": 1716454223411836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223470020, "dur": 5,
    "args": {
      "External id": 132240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132240,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 132240, "pid": 5, "tid": 7, "ts": 1716454223470020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411856, "dur": 7,
    "args": {
      "External id": 132240,
      "cbid": 211, "correlation": 132240
    }
  },
  {
    "ph": "s", "id": 132240, "pid": 76337, "tid": -914061504, "ts": 1716454223411856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223470026, "dur": 16,
    "args": {
      "External id": 132243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132243, "pid": 5, "tid": 7, "ts": 1716454223470026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411875, "dur": 6,
    "args": {
      "External id": 132243,
      "cbid": 211, "correlation": 132243
    }
  },
  {
    "ph": "s", "id": 132243, "pid": 76337, "tid": -914061504, "ts": 1716454223411875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223470043, "dur": 12,
    "args": {
      "External id": 132252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132252,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132252, "pid": 5, "tid": 7, "ts": 1716454223470043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411915, "dur": 9,
    "args": {
      "External id": 132252,
      "cbid": 211, "correlation": 132252
    }
  },
  {
    "ph": "s", "id": 132252, "pid": 76337, "tid": -914061504, "ts": 1716454223411915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411985, "dur": 0,
    "args": {
      "External id": 132262,
      "cbid": 317, "correlation": 132262
    }
  },
  {
    "ph": "f", "id": 132262, "pid": 76337, "tid": -914061504, "ts": 1716454223411985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411986, "dur": 0,
    "args": {
      "External id": 132263,
      "cbid": 203, "correlation": 132263
    }
  },
  {
    "ph": "f", "id": 132263, "pid": 76337, "tid": -914061504, "ts": 1716454223411986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223411986, "dur": 0,
    "args": {
      "External id": 132264,
      "cbid": 205, "correlation": 132264
    }
  },
  {
    "ph": "f", "id": 132264, "pid": 76337, "tid": -914061504, "ts": 1716454223411986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223470056, "dur": 11,
    "args": {
      "External id": 132268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132268, "pid": 5, "tid": 7, "ts": 1716454223470056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412000, "dur": 13,
    "args": {
      "External id": 132268,
      "cbid": 211, "correlation": 132268
    }
  },
  {
    "ph": "s", "id": 132268, "pid": 76337, "tid": -914061504, "ts": 1716454223412000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223470069, "dur": 162,
    "args": {
      "External id": 132270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132270, "pid": 5, "tid": 7, "ts": 1716454223470069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412015, "dur": 5,
    "args": {
      "External id": 132270,
      "cbid": 211, "correlation": 132270
    }
  },
  {
    "ph": "s", "id": 132270, "pid": 76337, "tid": -914061504, "ts": 1716454223412015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223470233, "dur": 1,
    "args": {
      "External id": 132272,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132272,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 132272, "pid": 5, "tid": 7, "ts": 1716454223470233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412026, "dur": 8,
    "args": {
      "External id": 132272,
      "cbid": 51, "correlation": 132272
    }
  },
  {
    "ph": "s", "id": 132272, "pid": 76337, "tid": -914061504, "ts": 1716454223412026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223470237, "dur": 646,
    "args": {
      "External id": 132273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132273,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132273, "pid": 5, "tid": 7, "ts": 1716454223470237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412035, "dur": 6,
    "args": {
      "External id": 132273,
      "cbid": 211, "correlation": 132273
    }
  },
  {
    "ph": "s", "id": 132273, "pid": 76337, "tid": -914061504, "ts": 1716454223412035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223470884, "dur": 13,
    "args": {
      "External id": 132275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132275,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132275, "pid": 5, "tid": 7, "ts": 1716454223470884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412045, "dur": 5,
    "args": {
      "External id": 132275,
      "cbid": 211, "correlation": 132275
    }
  },
  {
    "ph": "s", "id": 132275, "pid": 76337, "tid": -914061504, "ts": 1716454223412045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223470898, "dur": 15,
    "args": {
      "External id": 132281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132281, "pid": 5, "tid": 7, "ts": 1716454223470898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412073, "dur": 8,
    "args": {
      "External id": 132281,
      "cbid": 211, "correlation": 132281
    }
  },
  {
    "ph": "s", "id": 132281, "pid": 76337, "tid": -914061504, "ts": 1716454223412073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412132, "dur": 0,
    "args": {
      "External id": 132291,
      "cbid": 317, "correlation": 132291
    }
  },
  {
    "ph": "f", "id": 132291, "pid": 76337, "tid": -914061504, "ts": 1716454223412132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412133, "dur": 0,
    "args": {
      "External id": 132292,
      "cbid": 203, "correlation": 132292
    }
  },
  {
    "ph": "f", "id": 132292, "pid": 76337, "tid": -914061504, "ts": 1716454223412133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412134, "dur": 0,
    "args": {
      "External id": 132293,
      "cbid": 205, "correlation": 132293
    }
  },
  {
    "ph": "f", "id": 132293, "pid": 76337, "tid": -914061504, "ts": 1716454223412134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223470914, "dur": 20,
    "args": {
      "External id": 132297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132297,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132297, "pid": 5, "tid": 7, "ts": 1716454223470914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412147, "dur": 12,
    "args": {
      "External id": 132297,
      "cbid": 211, "correlation": 132297
    }
  },
  {
    "ph": "s", "id": 132297, "pid": 76337, "tid": -914061504, "ts": 1716454223412147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223470936, "dur": 4,
    "args": {
      "External id": 132299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132299, "pid": 5, "tid": 7, "ts": 1716454223470936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412165, "dur": 6,
    "args": {
      "External id": 132299,
      "cbid": 211, "correlation": 132299
    }
  },
  {
    "ph": "s", "id": 132299, "pid": 76337, "tid": -914061504, "ts": 1716454223412165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412174, "dur": 0,
    "args": {
      "External id": 132300,
      "cbid": 51, "correlation": 132300
    }
  },
  {
    "ph": "s", "id": 132300, "pid": 76337, "tid": -914061504, "ts": 1716454223412174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223470941, "dur": 174,
    "args": {
      "External id": 132301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132301,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 132301, "pid": 5, "tid": 7, "ts": 1716454223470941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412175, "dur": 5,
    "args": {
      "External id": 132301,
      "cbid": 211, "correlation": 132301
    }
  },
  {
    "ph": "s", "id": 132301, "pid": 76337, "tid": -914061504, "ts": 1716454223412175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223471116, "dur": 17,
    "args": {
      "External id": 132306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132306, "pid": 5, "tid": 7, "ts": 1716454223471116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412200, "dur": 8,
    "args": {
      "External id": 132306,
      "cbid": 211, "correlation": 132306
    }
  },
  {
    "ph": "s", "id": 132306, "pid": 76337, "tid": -914061504, "ts": 1716454223412200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223471134, "dur": 12,
    "args": {
      "External id": 132314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132314,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132314, "pid": 5, "tid": 7, "ts": 1716454223471134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412228, "dur": 8,
    "args": {
      "External id": 132314,
      "cbid": 211, "correlation": 132314
    }
  },
  {
    "ph": "s", "id": 132314, "pid": 76337, "tid": -914061504, "ts": 1716454223412228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223471148, "dur": 10,
    "args": {
      "External id": 132322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132322,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132322, "pid": 5, "tid": 7, "ts": 1716454223471148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412256, "dur": 9,
    "args": {
      "External id": 132322,
      "cbid": 211, "correlation": 132322
    }
  },
  {
    "ph": "s", "id": 132322, "pid": 76337, "tid": -914061504, "ts": 1716454223412256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223471159, "dur": 18,
    "args": {
      "External id": 132342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132342,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 132342, "pid": 5, "tid": 7, "ts": 1716454223471159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412338, "dur": 12,
    "args": {
      "External id": 132342,
      "cbid": 211, "correlation": 132342
    }
  },
  {
    "ph": "s", "id": 132342, "pid": 76337, "tid": -914061504, "ts": 1716454223412338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223471178, "dur": 4,
    "args": {
      "External id": 132354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132354,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 132354, "pid": 5, "tid": 7, "ts": 1716454223471178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412361, "dur": 7,
    "args": {
      "External id": 132354,
      "cbid": 211, "correlation": 132354
    }
  },
  {
    "ph": "s", "id": 132354, "pid": 76337, "tid": -914061504, "ts": 1716454223412361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223471184, "dur": 17,
    "args": {
      "External id": 132357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132357, "pid": 5, "tid": 7, "ts": 1716454223471184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412379, "dur": 6,
    "args": {
      "External id": 132357,
      "cbid": 211, "correlation": 132357
    }
  },
  {
    "ph": "s", "id": 132357, "pid": 76337, "tid": -914061504, "ts": 1716454223412379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412436, "dur": 0,
    "args": {
      "External id": 132368,
      "cbid": 317, "correlation": 132368
    }
  },
  {
    "ph": "f", "id": 132368, "pid": 76337, "tid": -914061504, "ts": 1716454223412436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412437, "dur": 0,
    "args": {
      "External id": 132369,
      "cbid": 203, "correlation": 132369
    }
  },
  {
    "ph": "f", "id": 132369, "pid": 76337, "tid": -914061504, "ts": 1716454223412437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412437, "dur": 0,
    "args": {
      "External id": 132370,
      "cbid": 205, "correlation": 132370
    }
  },
  {
    "ph": "f", "id": 132370, "pid": 76337, "tid": -914061504, "ts": 1716454223412437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223471202, "dur": 11,
    "args": {
      "External id": 132374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132374,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132374, "pid": 5, "tid": 7, "ts": 1716454223471202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412451, "dur": 12,
    "args": {
      "External id": 132374,
      "cbid": 211, "correlation": 132374
    }
  },
  {
    "ph": "s", "id": 132374, "pid": 76337, "tid": -914061504, "ts": 1716454223412451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223471214, "dur": 3,
    "args": {
      "External id": 132376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132376,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132376, "pid": 5, "tid": 7, "ts": 1716454223471214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412468, "dur": 6,
    "args": {
      "External id": 132376,
      "cbid": 211, "correlation": 132376
    }
  },
  {
    "ph": "s", "id": 132376, "pid": 76337, "tid": -914061504, "ts": 1716454223412468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412476, "dur": 0,
    "args": {
      "External id": 132377,
      "cbid": 51, "correlation": 132377
    }
  },
  {
    "ph": "s", "id": 132377, "pid": 76337, "tid": -914061504, "ts": 1716454223412476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223471218, "dur": 90,
    "args": {
      "External id": 132378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132378,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 132378, "pid": 5, "tid": 7, "ts": 1716454223471218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412477, "dur": 5,
    "args": {
      "External id": 132378,
      "cbid": 211, "correlation": 132378
    }
  },
  {
    "ph": "s", "id": 132378, "pid": 76337, "tid": -914061504, "ts": 1716454223412477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223471309, "dur": 15,
    "args": {
      "External id": 132383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132383, "pid": 5, "tid": 7, "ts": 1716454223471309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412503, "dur": 9,
    "args": {
      "External id": 132383,
      "cbid": 211, "correlation": 132383
    }
  },
  {
    "ph": "s", "id": 132383, "pid": 76337, "tid": -914061504, "ts": 1716454223412503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223471326, "dur": 84,
    "args": {
      "External id": 132392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132392,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132392, "pid": 5, "tid": 7, "ts": 1716454223471326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412585, "dur": 15,
    "args": {
      "External id": 132392,
      "cbid": 211, "correlation": 132392
    }
  },
  {
    "ph": "s", "id": 132392, "pid": 76337, "tid": -914061504, "ts": 1716454223412585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223471411, "dur": 31,
    "args": {
      "External id": 132414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132414,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132414, "pid": 5, "tid": 7, "ts": 1716454223471411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412643, "dur": 10,
    "args": {
      "External id": 132414,
      "cbid": 211, "correlation": 132414
    }
  },
  {
    "ph": "s", "id": 132414, "pid": 76337, "tid": -914061504, "ts": 1716454223412643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412736, "dur": 1,
    "args": {
      "External id": 132425,
      "cbid": 251, "correlation": 132425
    }
  },
  {
    "ph": "f", "id": 132425, "pid": 76337, "tid": -914061504, "ts": 1716454223412736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223471443, "dur": 162,
    "args": {
      "External id": 132426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132426,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132426, "pid": 5, "tid": 7, "ts": 1716454223471443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412742, "dur": 14,
    "args": {
      "External id": 132426,
      "cbid": 211, "correlation": 132426
    }
  },
  {
    "ph": "s", "id": 132426, "pid": 76337, "tid": -914061504, "ts": 1716454223412742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412814, "dur": 1,
    "args": {
      "External id": 132437,
      "cbid": 251, "correlation": 132437
    }
  },
  {
    "ph": "f", "id": 132437, "pid": 76337, "tid": -914061504, "ts": 1716454223412814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223471606, "dur": 156,
    "args": {
      "External id": 132438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132438,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132438, "pid": 5, "tid": 7, "ts": 1716454223471606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412818, "dur": 11,
    "args": {
      "External id": 132438,
      "cbid": 211, "correlation": 132438
    }
  },
  {
    "ph": "s", "id": 132438, "pid": 76337, "tid": -914061504, "ts": 1716454223412818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412884, "dur": 1,
    "args": {
      "External id": 132449,
      "cbid": 251, "correlation": 132449
    }
  },
  {
    "ph": "f", "id": 132449, "pid": 76337, "tid": -914061504, "ts": 1716454223412884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223471763, "dur": 157,
    "args": {
      "External id": 132450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132450,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132450, "pid": 5, "tid": 7, "ts": 1716454223471763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412888, "dur": 11,
    "args": {
      "External id": 132450,
      "cbid": 211, "correlation": 132450
    }
  },
  {
    "ph": "s", "id": 132450, "pid": 76337, "tid": -914061504, "ts": 1716454223412888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223471922, "dur": 336,
    "args": {
      "External id": 132475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132475,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132475, "pid": 5, "tid": 7, "ts": 1716454223471922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223412983, "dur": 13,
    "args": {
      "External id": 132475,
      "cbid": 211, "correlation": 132475
    }
  },
  {
    "ph": "s", "id": 132475, "pid": 76337, "tid": -914061504, "ts": 1716454223412983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413087, "dur": 1,
    "args": {
      "External id": 132493,
      "cbid": 251, "correlation": 132493
    }
  },
  {
    "ph": "f", "id": 132493, "pid": 76337, "tid": -914061504, "ts": 1716454223413087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223472259, "dur": 166,
    "args": {
      "External id": 132495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132495,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132495, "pid": 5, "tid": 7, "ts": 1716454223472259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413093, "dur": 13,
    "args": {
      "External id": 132495,
      "cbid": 211, "correlation": 132495
    }
  },
  {
    "ph": "s", "id": 132495, "pid": 76337, "tid": -914061504, "ts": 1716454223413093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223472427, "dur": 19,
    "args": {
      "External id": 132503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132503,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132503, "pid": 5, "tid": 7, "ts": 1716454223472427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413163, "dur": 13,
    "args": {
      "External id": 132503,
      "cbid": 211, "correlation": 132503
    }
  },
  {
    "ph": "s", "id": 132503, "pid": 76337, "tid": -914061504, "ts": 1716454223413163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223472447, "dur": 28,
    "args": {
      "External id": 132511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132511,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132511, "pid": 5, "tid": 7, "ts": 1716454223472447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413203, "dur": 8,
    "args": {
      "External id": 132511,
      "cbid": 211, "correlation": 132511
    }
  },
  {
    "ph": "s", "id": 132511, "pid": 76337, "tid": -914061504, "ts": 1716454223413203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223472476, "dur": 18,
    "args": {
      "External id": 132522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132522, "pid": 5, "tid": 7, "ts": 1716454223472476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413277, "dur": 13,
    "args": {
      "External id": 132522,
      "cbid": 211, "correlation": 132522
    }
  },
  {
    "ph": "s", "id": 132522, "pid": 76337, "tid": -914061504, "ts": 1716454223413277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223472496, "dur": 16,
    "args": {
      "External id": 132544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132544,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132544, "pid": 5, "tid": 7, "ts": 1716454223472496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413309, "dur": 8,
    "args": {
      "External id": 132544,
      "cbid": 211, "correlation": 132544
    }
  },
  {
    "ph": "s", "id": 132544, "pid": 76337, "tid": -914061504, "ts": 1716454223413309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413395, "dur": 2,
    "args": {
      "External id": 132555,
      "cbid": 251, "correlation": 132555
    }
  },
  {
    "ph": "f", "id": 132555, "pid": 76337, "tid": -914061504, "ts": 1716454223413395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223472513, "dur": 89,
    "args": {
      "External id": 132556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132556,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132556, "pid": 5, "tid": 7, "ts": 1716454223472513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413400, "dur": 13,
    "args": {
      "External id": 132556,
      "cbid": 211, "correlation": 132556
    }
  },
  {
    "ph": "s", "id": 132556, "pid": 76337, "tid": -914061504, "ts": 1716454223413400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413471, "dur": 1,
    "args": {
      "External id": 132567,
      "cbid": 251, "correlation": 132567
    }
  },
  {
    "ph": "f", "id": 132567, "pid": 76337, "tid": -914061504, "ts": 1716454223413471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413475, "dur": 0,
    "args": {
      "External id": 132568,
      "cbid": 251, "correlation": 132568
    }
  },
  {
    "ph": "f", "id": 132568, "pid": 76337, "tid": -914061504, "ts": 1716454223413475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223472604, "dur": 12,
    "args": {
      "External id": 132569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132569,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132569, "pid": 5, "tid": 7, "ts": 1716454223472604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413477, "dur": 12,
    "args": {
      "External id": 132569,
      "cbid": 211, "correlation": 132569
    }
  },
  {
    "ph": "s", "id": 132569, "pid": 76337, "tid": -914061504, "ts": 1716454223413477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223472617, "dur": 6,
    "args": {
      "External id": 132571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132571,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132571, "pid": 5, "tid": 7, "ts": 1716454223472617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413491, "dur": 6,
    "args": {
      "External id": 132571,
      "cbid": 211, "correlation": 132571
    }
  },
  {
    "ph": "s", "id": 132571, "pid": 76337, "tid": -914061504, "ts": 1716454223413491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413548, "dur": 1,
    "args": {
      "External id": 132582,
      "cbid": 251, "correlation": 132582
    }
  },
  {
    "ph": "f", "id": 132582, "pid": 76337, "tid": -914061504, "ts": 1716454223413548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413552, "dur": 0,
    "args": {
      "External id": 132583,
      "cbid": 251, "correlation": 132583
    }
  },
  {
    "ph": "f", "id": 132583, "pid": 76337, "tid": -914061504, "ts": 1716454223413552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223472624, "dur": 9,
    "args": {
      "External id": 132584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132584,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132584, "pid": 5, "tid": 7, "ts": 1716454223472624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413553, "dur": 12,
    "args": {
      "External id": 132584,
      "cbid": 211, "correlation": 132584
    }
  },
  {
    "ph": "s", "id": 132584, "pid": 76337, "tid": -914061504, "ts": 1716454223413553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223472634, "dur": 3,
    "args": {
      "External id": 132586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132586,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132586, "pid": 5, "tid": 7, "ts": 1716454223472634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413567, "dur": 5,
    "args": {
      "External id": 132586,
      "cbid": 211, "correlation": 132586
    }
  },
  {
    "ph": "s", "id": 132586, "pid": 76337, "tid": -914061504, "ts": 1716454223413567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223472638, "dur": 55,
    "args": {
      "External id": 132611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132611,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132611, "pid": 5, "tid": 7, "ts": 1716454223472638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413643, "dur": 12,
    "args": {
      "External id": 132611,
      "cbid": 211, "correlation": 132611
    }
  },
  {
    "ph": "s", "id": 132611, "pid": 76337, "tid": -914061504, "ts": 1716454223413643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413742, "dur": 2,
    "args": {
      "External id": 132629,
      "cbid": 251, "correlation": 132629
    }
  },
  {
    "ph": "f", "id": 132629, "pid": 76337, "tid": -914061504, "ts": 1716454223413742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223472694, "dur": 90,
    "args": {
      "External id": 132631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132631,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132631, "pid": 5, "tid": 7, "ts": 1716454223472694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413748, "dur": 14,
    "args": {
      "External id": 132631,
      "cbid": 211, "correlation": 132631
    }
  },
  {
    "ph": "s", "id": 132631, "pid": 76337, "tid": -914061504, "ts": 1716454223413748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223472785, "dur": 9,
    "args": {
      "External id": 132639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132639,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132639, "pid": 5, "tid": 7, "ts": 1716454223472785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413818, "dur": 12,
    "args": {
      "External id": 132639,
      "cbid": 211, "correlation": 132639
    }
  },
  {
    "ph": "s", "id": 132639, "pid": 76337, "tid": -914061504, "ts": 1716454223413818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223472796, "dur": 20,
    "args": {
      "External id": 132647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132647, "pid": 5, "tid": 7, "ts": 1716454223472796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413858, "dur": 10,
    "args": {
      "External id": 132647,
      "cbid": 211, "correlation": 132647
    }
  },
  {
    "ph": "s", "id": 132647, "pid": 76337, "tid": -914061504, "ts": 1716454223413858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223472817, "dur": 17,
    "args": {
      "External id": 132669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132669,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132669, "pid": 5, "tid": 7, "ts": 1716454223472817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223413909, "dur": 11,
    "args": {
      "External id": 132669,
      "cbid": 211, "correlation": 132669
    }
  },
  {
    "ph": "s", "id": 132669, "pid": 76337, "tid": -914061504, "ts": 1716454223413909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414006, "dur": 1,
    "args": {
      "External id": 132685,
      "cbid": 251, "correlation": 132685
    }
  },
  {
    "ph": "f", "id": 132685, "pid": 76337, "tid": -914061504, "ts": 1716454223414006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414011, "dur": 0,
    "args": {
      "External id": 132687,
      "cbid": 251, "correlation": 132687
    }
  },
  {
    "ph": "f", "id": 132687, "pid": 76337, "tid": -914061504, "ts": 1716454223414011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223472835, "dur": 495,
    "args": {
      "External id": 132688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132688,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132688, "pid": 5, "tid": 7, "ts": 1716454223472835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414013, "dur": 13,
    "args": {
      "External id": 132688,
      "cbid": 211, "correlation": 132688
    }
  },
  {
    "ph": "s", "id": 132688, "pid": 76337, "tid": -914061504, "ts": 1716454223414013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223473331, "dur": 66,
    "args": {
      "External id": 132696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132696, "pid": 5, "tid": 7, "ts": 1716454223473331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414081, "dur": 13,
    "args": {
      "External id": 132696,
      "cbid": 211, "correlation": 132696
    }
  },
  {
    "ph": "s", "id": 132696, "pid": 76337, "tid": -914061504, "ts": 1716454223414081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223473399, "dur": 66,
    "args": {
      "External id": 132704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132704,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132704, "pid": 5, "tid": 7, "ts": 1716454223473399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414113, "dur": 8,
    "args": {
      "External id": 132704,
      "cbid": 211, "correlation": 132704
    }
  },
  {
    "ph": "s", "id": 132704, "pid": 76337, "tid": -914061504, "ts": 1716454223414113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414193, "dur": 1,
    "args": {
      "External id": 132720,
      "cbid": 251, "correlation": 132720
    }
  },
  {
    "ph": "f", "id": 132720, "pid": 76337, "tid": -914061504, "ts": 1716454223414193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223473468, "dur": 1,
    "args": {
      "External id": 132722,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132722,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1595744680851064
    }
  },
  {
    "ph": "f", "id": 132722, "pid": 5, "tid": 7, "ts": 1716454223473468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414198, "dur": 11,
    "args": {
      "External id": 132722,
      "cbid": 51, "correlation": 132722
    }
  },
  {
    "ph": "s", "id": 132722, "pid": 76337, "tid": -914061504, "ts": 1716454223414198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223473471, "dur": 272,
    "args": {
      "External id": 132723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132723,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132723, "pid": 5, "tid": 7, "ts": 1716454223473471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414211, "dur": 11,
    "args": {
      "External id": 132723,
      "cbid": 211, "correlation": 132723
    }
  },
  {
    "ph": "s", "id": 132723, "pid": 76337, "tid": -914061504, "ts": 1716454223414211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223473745, "dur": 14,
    "args": {
      "External id": 132731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132731,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132731, "pid": 5, "tid": 7, "ts": 1716454223473745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414251, "dur": 11,
    "args": {
      "External id": 132731,
      "cbid": 211, "correlation": 132731
    }
  },
  {
    "ph": "s", "id": 132731, "pid": 76337, "tid": -914061504, "ts": 1716454223414251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223473760, "dur": 38,
    "args": {
      "External id": 132742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132742, "pid": 5, "tid": 7, "ts": 1716454223473760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414321, "dur": 12,
    "args": {
      "External id": 132742,
      "cbid": 211, "correlation": 132742
    }
  },
  {
    "ph": "s", "id": 132742, "pid": 76337, "tid": -914061504, "ts": 1716454223414321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414385, "dur": 0,
    "args": {
      "External id": 132754,
      "cbid": 317, "correlation": 132754
    }
  },
  {
    "ph": "f", "id": 132754, "pid": 76337, "tid": -914061504, "ts": 1716454223414385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414386, "dur": 0,
    "args": {
      "External id": 132755,
      "cbid": 203, "correlation": 132755
    }
  },
  {
    "ph": "f", "id": 132755, "pid": 76337, "tid": -914061504, "ts": 1716454223414386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414387, "dur": 0,
    "args": {
      "External id": 132756,
      "cbid": 205, "correlation": 132756
    }
  },
  {
    "ph": "f", "id": 132756, "pid": 76337, "tid": -914061504, "ts": 1716454223414387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223473799, "dur": 13,
    "args": {
      "External id": 132760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132760,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132760, "pid": 5, "tid": 7, "ts": 1716454223473799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414403, "dur": 13,
    "args": {
      "External id": 132760,
      "cbid": 211, "correlation": 132760
    }
  },
  {
    "ph": "s", "id": 132760, "pid": 76337, "tid": -914061504, "ts": 1716454223414403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223473814, "dur": 4,
    "args": {
      "External id": 132762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132762,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132762, "pid": 5, "tid": 7, "ts": 1716454223473814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414420, "dur": 6,
    "args": {
      "External id": 132762,
      "cbid": 211, "correlation": 132762
    }
  },
  {
    "ph": "s", "id": 132762, "pid": 76337, "tid": -914061504, "ts": 1716454223414420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414429, "dur": 0,
    "args": {
      "External id": 132763,
      "cbid": 51, "correlation": 132763
    }
  },
  {
    "ph": "s", "id": 132763, "pid": 76337, "tid": -914061504, "ts": 1716454223414429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223473819, "dur": 97,
    "args": {
      "External id": 132764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132764,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 132764, "pid": 5, "tid": 7, "ts": 1716454223473819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414429, "dur": 5,
    "args": {
      "External id": 132764,
      "cbid": 211, "correlation": 132764
    }
  },
  {
    "ph": "s", "id": 132764, "pid": 76337, "tid": -914061504, "ts": 1716454223414429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223473916, "dur": 16,
    "args": {
      "External id": 132769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132769,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132769, "pid": 5, "tid": 7, "ts": 1716454223473916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414457, "dur": 9,
    "args": {
      "External id": 132769,
      "cbid": 211, "correlation": 132769
    }
  },
  {
    "ph": "s", "id": 132769, "pid": 76337, "tid": -914061504, "ts": 1716454223414457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223473934, "dur": 12,
    "args": {
      "External id": 132777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132777,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132777, "pid": 5, "tid": 7, "ts": 1716454223473934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414489, "dur": 8,
    "args": {
      "External id": 132777,
      "cbid": 211, "correlation": 132777
    }
  },
  {
    "ph": "s", "id": 132777, "pid": 76337, "tid": -914061504, "ts": 1716454223414489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223473947, "dur": 30,
    "args": {
      "External id": 132786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132786,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132786, "pid": 5, "tid": 7, "ts": 1716454223473947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414528, "dur": 10,
    "args": {
      "External id": 132786,
      "cbid": 211, "correlation": 132786
    }
  },
  {
    "ph": "s", "id": 132786, "pid": 76337, "tid": -914061504, "ts": 1716454223414528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223473978, "dur": 30,
    "args": {
      "External id": 132806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132806,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 132806, "pid": 5, "tid": 7, "ts": 1716454223473978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414598, "dur": 11,
    "args": {
      "External id": 132806,
      "cbid": 211, "correlation": 132806
    }
  },
  {
    "ph": "s", "id": 132806, "pid": 76337, "tid": -914061504, "ts": 1716454223414598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223474010, "dur": 5,
    "args": {
      "External id": 132818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132818,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132818, "pid": 5, "tid": 7, "ts": 1716454223474010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414620, "dur": 6,
    "args": {
      "External id": 132818,
      "cbid": 211, "correlation": 132818
    }
  },
  {
    "ph": "s", "id": 132818, "pid": 76337, "tid": -914061504, "ts": 1716454223414620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223474016, "dur": 31,
    "args": {
      "External id": 132821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132821,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132821, "pid": 5, "tid": 7, "ts": 1716454223474016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414638, "dur": 6,
    "args": {
      "External id": 132821,
      "cbid": 211, "correlation": 132821
    }
  },
  {
    "ph": "s", "id": 132821, "pid": 76337, "tid": -914061504, "ts": 1716454223414638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223474048, "dur": 21,
    "args": {
      "External id": 132830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132830,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132830, "pid": 5, "tid": 7, "ts": 1716454223474048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414677, "dur": 9,
    "args": {
      "External id": 132830,
      "cbid": 211, "correlation": 132830
    }
  },
  {
    "ph": "s", "id": 132830, "pid": 76337, "tid": -914061504, "ts": 1716454223414677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414729, "dur": 0,
    "args": {
      "External id": 132840,
      "cbid": 317, "correlation": 132840
    }
  },
  {
    "ph": "f", "id": 132840, "pid": 76337, "tid": -914061504, "ts": 1716454223414729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414730, "dur": 0,
    "args": {
      "External id": 132841,
      "cbid": 203, "correlation": 132841
    }
  },
  {
    "ph": "f", "id": 132841, "pid": 76337, "tid": -914061504, "ts": 1716454223414730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414730, "dur": 0,
    "args": {
      "External id": 132842,
      "cbid": 205, "correlation": 132842
    }
  },
  {
    "ph": "f", "id": 132842, "pid": 76337, "tid": -914061504, "ts": 1716454223414730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223474070, "dur": 22,
    "args": {
      "External id": 132846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132846, "pid": 5, "tid": 7, "ts": 1716454223474070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414743, "dur": 11,
    "args": {
      "External id": 132846,
      "cbid": 211, "correlation": 132846
    }
  },
  {
    "ph": "s", "id": 132846, "pid": 76337, "tid": -914061504, "ts": 1716454223414743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223474094, "dur": 318,
    "args": {
      "External id": 132848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132848,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132848, "pid": 5, "tid": 7, "ts": 1716454223474094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414757, "dur": 5,
    "args": {
      "External id": 132848,
      "cbid": 211, "correlation": 132848
    }
  },
  {
    "ph": "s", "id": 132848, "pid": 76337, "tid": -914061504, "ts": 1716454223414757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223474414, "dur": 1,
    "args": {
      "External id": 132850,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132850,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 132850, "pid": 5, "tid": 7, "ts": 1716454223474414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414769, "dur": 6,
    "args": {
      "External id": 132850,
      "cbid": 51, "correlation": 132850
    }
  },
  {
    "ph": "s", "id": 132850, "pid": 76337, "tid": -914061504, "ts": 1716454223414769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223474418, "dur": 1250,
    "args": {
      "External id": 132851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132851,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132851, "pid": 5, "tid": 7, "ts": 1716454223474418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414776, "dur": 6,
    "args": {
      "External id": 132851,
      "cbid": 211, "correlation": 132851
    }
  },
  {
    "ph": "s", "id": 132851, "pid": 76337, "tid": -914061504, "ts": 1716454223414776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223475669, "dur": 13,
    "args": {
      "External id": 132853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132853,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132853, "pid": 5, "tid": 7, "ts": 1716454223475669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414786, "dur": 5,
    "args": {
      "External id": 132853,
      "cbid": 211, "correlation": 132853
    }
  },
  {
    "ph": "s", "id": 132853, "pid": 76337, "tid": -914061504, "ts": 1716454223414786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223475683, "dur": 15,
    "args": {
      "External id": 132859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132859,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132859, "pid": 5, "tid": 7, "ts": 1716454223475683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414815, "dur": 8,
    "args": {
      "External id": 132859,
      "cbid": 211, "correlation": 132859
    }
  },
  {
    "ph": "s", "id": 132859, "pid": 76337, "tid": -914061504, "ts": 1716454223414815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223475699, "dur": 4,
    "args": {
      "External id": 132867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132867,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 132867, "pid": 5, "tid": 7, "ts": 1716454223475699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414858, "dur": 10,
    "args": {
      "External id": 132867,
      "cbid": 211, "correlation": 132867
    }
  },
  {
    "ph": "s", "id": 132867, "pid": 76337, "tid": -914061504, "ts": 1716454223414858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414924, "dur": 1,
    "args": {
      "External id": 132883,
      "cbid": 251, "correlation": 132883
    }
  },
  {
    "ph": "f", "id": 132883, "pid": 76337, "tid": -914061504, "ts": 1716454223414924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414929, "dur": 0,
    "args": {
      "External id": 132885,
      "cbid": 251, "correlation": 132885
    }
  },
  {
    "ph": "f", "id": 132885, "pid": 76337, "tid": -914061504, "ts": 1716454223414929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223475704, "dur": 13,
    "args": {
      "External id": 132886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132886,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132886, "pid": 5, "tid": 7, "ts": 1716454223475704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414931, "dur": 11,
    "args": {
      "External id": 132886,
      "cbid": 211, "correlation": 132886
    }
  },
  {
    "ph": "s", "id": 132886, "pid": 76337, "tid": -914061504, "ts": 1716454223414931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223475719, "dur": 5,
    "args": {
      "External id": 132888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132888,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132888, "pid": 5, "tid": 7, "ts": 1716454223475719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223414944, "dur": 5,
    "args": {
      "External id": 132888,
      "cbid": 211, "correlation": 132888
    }
  },
  {
    "ph": "s", "id": 132888, "pid": 76337, "tid": -914061504, "ts": 1716454223414944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223475725, "dur": 18,
    "args": {
      "External id": 132898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132898, "pid": 5, "tid": 7, "ts": 1716454223475725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415010, "dur": 13,
    "args": {
      "External id": 132898,
      "cbid": 211, "correlation": 132898
    }
  },
  {
    "ph": "s", "id": 132898, "pid": 76337, "tid": -914061504, "ts": 1716454223415010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223475744, "dur": 17,
    "args": {
      "External id": 132918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132918,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 132918, "pid": 5, "tid": 7, "ts": 1716454223475744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415078, "dur": 11,
    "args": {
      "External id": 132918,
      "cbid": 211, "correlation": 132918
    }
  },
  {
    "ph": "s", "id": 132918, "pid": 76337, "tid": -914061504, "ts": 1716454223415078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223475762, "dur": 4,
    "args": {
      "External id": 132930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132930,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 132930, "pid": 5, "tid": 7, "ts": 1716454223475762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415099, "dur": 6,
    "args": {
      "External id": 132930,
      "cbid": 211, "correlation": 132930
    }
  },
  {
    "ph": "s", "id": 132930, "pid": 76337, "tid": -914061504, "ts": 1716454223415099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223475768, "dur": 16,
    "args": {
      "External id": 132933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132933,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132933, "pid": 5, "tid": 7, "ts": 1716454223475768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415117, "dur": 7,
    "args": {
      "External id": 132933,
      "cbid": 211, "correlation": 132933
    }
  },
  {
    "ph": "s", "id": 132933, "pid": 76337, "tid": -914061504, "ts": 1716454223415117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223475785, "dur": 11,
    "args": {
      "External id": 132942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132942,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132942, "pid": 5, "tid": 7, "ts": 1716454223475785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415157, "dur": 11,
    "args": {
      "External id": 132942,
      "cbid": 211, "correlation": 132942
    }
  },
  {
    "ph": "s", "id": 132942, "pid": 76337, "tid": -914061504, "ts": 1716454223415157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415220, "dur": 0,
    "args": {
      "External id": 132952,
      "cbid": 317, "correlation": 132952
    }
  },
  {
    "ph": "f", "id": 132952, "pid": 76337, "tid": -914061504, "ts": 1716454223415220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415221, "dur": 0,
    "args": {
      "External id": 132953,
      "cbid": 203, "correlation": 132953
    }
  },
  {
    "ph": "f", "id": 132953, "pid": 76337, "tid": -914061504, "ts": 1716454223415221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415222, "dur": 0,
    "args": {
      "External id": 132954,
      "cbid": 205, "correlation": 132954
    }
  },
  {
    "ph": "f", "id": 132954, "pid": 76337, "tid": -914061504, "ts": 1716454223415222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223475798, "dur": 12,
    "args": {
      "External id": 132958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132958,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132958, "pid": 5, "tid": 7, "ts": 1716454223475798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415236, "dur": 12,
    "args": {
      "External id": 132958,
      "cbid": 211, "correlation": 132958
    }
  },
  {
    "ph": "s", "id": 132958, "pid": 76337, "tid": -914061504, "ts": 1716454223415236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223475811, "dur": 162,
    "args": {
      "External id": 132960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132960,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132960, "pid": 5, "tid": 7, "ts": 1716454223475811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415250, "dur": 5,
    "args": {
      "External id": 132960,
      "cbid": 211, "correlation": 132960
    }
  },
  {
    "ph": "s", "id": 132960, "pid": 76337, "tid": -914061504, "ts": 1716454223415250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223475976, "dur": 1,
    "args": {
      "External id": 132962,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 132962,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 132962, "pid": 5, "tid": 7, "ts": 1716454223475976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415261, "dur": 7,
    "args": {
      "External id": 132962,
      "cbid": 51, "correlation": 132962
    }
  },
  {
    "ph": "s", "id": 132962, "pid": 76337, "tid": -914061504, "ts": 1716454223415261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223475979, "dur": 646,
    "args": {
      "External id": 132963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132963,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 132963, "pid": 5, "tid": 7, "ts": 1716454223475979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415269, "dur": 6,
    "args": {
      "External id": 132963,
      "cbid": 211, "correlation": 132963
    }
  },
  {
    "ph": "s", "id": 132963, "pid": 76337, "tid": -914061504, "ts": 1716454223415269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223476627, "dur": 13,
    "args": {
      "External id": 132965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132965,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132965, "pid": 5, "tid": 7, "ts": 1716454223476627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415279, "dur": 5,
    "args": {
      "External id": 132965,
      "cbid": 211, "correlation": 132965
    }
  },
  {
    "ph": "s", "id": 132965, "pid": 76337, "tid": -914061504, "ts": 1716454223415279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223476641, "dur": 15,
    "args": {
      "External id": 132971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132971, "pid": 5, "tid": 7, "ts": 1716454223476641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415306, "dur": 9,
    "args": {
      "External id": 132971,
      "cbid": 211, "correlation": 132971
    }
  },
  {
    "ph": "s", "id": 132971, "pid": 76337, "tid": -914061504, "ts": 1716454223415306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415366, "dur": 0,
    "args": {
      "External id": 132981,
      "cbid": 317, "correlation": 132981
    }
  },
  {
    "ph": "f", "id": 132981, "pid": 76337, "tid": -914061504, "ts": 1716454223415366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415366, "dur": 0,
    "args": {
      "External id": 132982,
      "cbid": 203, "correlation": 132982
    }
  },
  {
    "ph": "f", "id": 132982, "pid": 76337, "tid": -914061504, "ts": 1716454223415366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415367, "dur": 0,
    "args": {
      "External id": 132983,
      "cbid": 205, "correlation": 132983
    }
  },
  {
    "ph": "f", "id": 132983, "pid": 76337, "tid": -914061504, "ts": 1716454223415367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223476657, "dur": 21,
    "args": {
      "External id": 132987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132987,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132987, "pid": 5, "tid": 7, "ts": 1716454223476657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415378, "dur": 11,
    "args": {
      "External id": 132987,
      "cbid": 211, "correlation": 132987
    }
  },
  {
    "ph": "s", "id": 132987, "pid": 76337, "tid": -914061504, "ts": 1716454223415378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223476679, "dur": 4,
    "args": {
      "External id": 132989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 132989, "pid": 5, "tid": 7, "ts": 1716454223476679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415393, "dur": 6,
    "args": {
      "External id": 132989,
      "cbid": 211, "correlation": 132989
    }
  },
  {
    "ph": "s", "id": 132989, "pid": 76337, "tid": -914061504, "ts": 1716454223415393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415402, "dur": 0,
    "args": {
      "External id": 132990,
      "cbid": 51, "correlation": 132990
    }
  },
  {
    "ph": "s", "id": 132990, "pid": 76337, "tid": -914061504, "ts": 1716454223415402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223476684, "dur": 170,
    "args": {
      "External id": 132991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132991,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 132991, "pid": 5, "tid": 7, "ts": 1716454223476684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415402, "dur": 5,
    "args": {
      "External id": 132991,
      "cbid": 211, "correlation": 132991
    }
  },
  {
    "ph": "s", "id": 132991, "pid": 76337, "tid": -914061504, "ts": 1716454223415402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223476855, "dur": 15,
    "args": {
      "External id": 132996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 132996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 132996, "pid": 5, "tid": 7, "ts": 1716454223476855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415428, "dur": 8,
    "args": {
      "External id": 132996,
      "cbid": 211, "correlation": 132996
    }
  },
  {
    "ph": "s", "id": 132996, "pid": 76337, "tid": -914061504, "ts": 1716454223415428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223476871, "dur": 13,
    "args": {
      "External id": 133004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133004,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133004, "pid": 5, "tid": 7, "ts": 1716454223476871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415457, "dur": 8,
    "args": {
      "External id": 133004,
      "cbid": 211, "correlation": 133004
    }
  },
  {
    "ph": "s", "id": 133004, "pid": 76337, "tid": -914061504, "ts": 1716454223415457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223476886, "dur": 10,
    "args": {
      "External id": 133012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133012,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133012, "pid": 5, "tid": 7, "ts": 1716454223476886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415486, "dur": 8,
    "args": {
      "External id": 133012,
      "cbid": 211, "correlation": 133012
    }
  },
  {
    "ph": "s", "id": 133012, "pid": 76337, "tid": -914061504, "ts": 1716454223415486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223476898, "dur": 18,
    "args": {
      "External id": 133032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133032,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 133032, "pid": 5, "tid": 7, "ts": 1716454223476898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415568, "dur": 12,
    "args": {
      "External id": 133032,
      "cbid": 211, "correlation": 133032
    }
  },
  {
    "ph": "s", "id": 133032, "pid": 76337, "tid": -914061504, "ts": 1716454223415568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223476917, "dur": 4,
    "args": {
      "External id": 133044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133044,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 133044, "pid": 5, "tid": 7, "ts": 1716454223476917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415589, "dur": 6,
    "args": {
      "External id": 133044,
      "cbid": 211, "correlation": 133044
    }
  },
  {
    "ph": "s", "id": 133044, "pid": 76337, "tid": -914061504, "ts": 1716454223415589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223476923, "dur": 17,
    "args": {
      "External id": 133047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133047, "pid": 5, "tid": 7, "ts": 1716454223476923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415607, "dur": 7,
    "args": {
      "External id": 133047,
      "cbid": 211, "correlation": 133047
    }
  },
  {
    "ph": "s", "id": 133047, "pid": 76337, "tid": -914061504, "ts": 1716454223415607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415664, "dur": 0,
    "args": {
      "External id": 133058,
      "cbid": 317, "correlation": 133058
    }
  },
  {
    "ph": "f", "id": 133058, "pid": 76337, "tid": -914061504, "ts": 1716454223415664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415665, "dur": 0,
    "args": {
      "External id": 133059,
      "cbid": 203, "correlation": 133059
    }
  },
  {
    "ph": "f", "id": 133059, "pid": 76337, "tid": -914061504, "ts": 1716454223415665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415666, "dur": 0,
    "args": {
      "External id": 133060,
      "cbid": 205, "correlation": 133060
    }
  },
  {
    "ph": "f", "id": 133060, "pid": 76337, "tid": -914061504, "ts": 1716454223415666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223476941, "dur": 11,
    "args": {
      "External id": 133064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133064,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133064, "pid": 5, "tid": 7, "ts": 1716454223476941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415679, "dur": 12,
    "args": {
      "External id": 133064,
      "cbid": 211, "correlation": 133064
    }
  },
  {
    "ph": "s", "id": 133064, "pid": 76337, "tid": -914061504, "ts": 1716454223415679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223476953, "dur": 3,
    "args": {
      "External id": 133066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133066, "pid": 5, "tid": 7, "ts": 1716454223476953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415695, "dur": 5,
    "args": {
      "External id": 133066,
      "cbid": 211, "correlation": 133066
    }
  },
  {
    "ph": "s", "id": 133066, "pid": 76337, "tid": -914061504, "ts": 1716454223415695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415703, "dur": 0,
    "args": {
      "External id": 133067,
      "cbid": 51, "correlation": 133067
    }
  },
  {
    "ph": "s", "id": 133067, "pid": 76337, "tid": -914061504, "ts": 1716454223415703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223476958, "dur": 90,
    "args": {
      "External id": 133068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133068,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 133068, "pid": 5, "tid": 7, "ts": 1716454223476958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415703, "dur": 5,
    "args": {
      "External id": 133068,
      "cbid": 211, "correlation": 133068
    }
  },
  {
    "ph": "s", "id": 133068, "pid": 76337, "tid": -914061504, "ts": 1716454223415703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223477049, "dur": 16,
    "args": {
      "External id": 133073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133073, "pid": 5, "tid": 7, "ts": 1716454223477049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415730, "dur": 8,
    "args": {
      "External id": 133073,
      "cbid": 211, "correlation": 133073
    }
  },
  {
    "ph": "s", "id": 133073, "pid": 76337, "tid": -914061504, "ts": 1716454223415730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223477066, "dur": 84,
    "args": {
      "External id": 133082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133082,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133082, "pid": 5, "tid": 7, "ts": 1716454223477066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415811, "dur": 14,
    "args": {
      "External id": 133082,
      "cbid": 211, "correlation": 133082
    }
  },
  {
    "ph": "s", "id": 133082, "pid": 76337, "tid": -914061504, "ts": 1716454223415811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223477151, "dur": 30,
    "args": {
      "External id": 133104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133104,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133104, "pid": 5, "tid": 7, "ts": 1716454223477151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415866, "dur": 11,
    "args": {
      "External id": 133104,
      "cbid": 211, "correlation": 133104
    }
  },
  {
    "ph": "s", "id": 133104, "pid": 76337, "tid": -914061504, "ts": 1716454223415866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415954, "dur": 1,
    "args": {
      "External id": 133115,
      "cbid": 251, "correlation": 133115
    }
  },
  {
    "ph": "f", "id": 133115, "pid": 76337, "tid": -914061504, "ts": 1716454223415954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223477182, "dur": 163,
    "args": {
      "External id": 133116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133116,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133116, "pid": 5, "tid": 7, "ts": 1716454223477182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223415960, "dur": 22,
    "args": {
      "External id": 133116,
      "cbid": 211, "correlation": 133116
    }
  },
  {
    "ph": "s", "id": 133116, "pid": 76337, "tid": -914061504, "ts": 1716454223415960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416041, "dur": 1,
    "args": {
      "External id": 133127,
      "cbid": 251, "correlation": 133127
    }
  },
  {
    "ph": "f", "id": 133127, "pid": 76337, "tid": -914061504, "ts": 1716454223416041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223477347, "dur": 160,
    "args": {
      "External id": 133128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133128,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133128, "pid": 5, "tid": 7, "ts": 1716454223477347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416045, "dur": 12,
    "args": {
      "External id": 133128,
      "cbid": 211, "correlation": 133128
    }
  },
  {
    "ph": "s", "id": 133128, "pid": 76337, "tid": -914061504, "ts": 1716454223416045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416111, "dur": 1,
    "args": {
      "External id": 133139,
      "cbid": 251, "correlation": 133139
    }
  },
  {
    "ph": "f", "id": 133139, "pid": 76337, "tid": -914061504, "ts": 1716454223416111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223477508, "dur": 157,
    "args": {
      "External id": 133140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133140,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133140, "pid": 5, "tid": 7, "ts": 1716454223477508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416115, "dur": 12,
    "args": {
      "External id": 133140,
      "cbid": 211, "correlation": 133140
    }
  },
  {
    "ph": "s", "id": 133140, "pid": 76337, "tid": -914061504, "ts": 1716454223416115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223477667, "dur": 333,
    "args": {
      "External id": 133165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133165,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133165, "pid": 5, "tid": 7, "ts": 1716454223477667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416198, "dur": 12,
    "args": {
      "External id": 133165,
      "cbid": 211, "correlation": 133165
    }
  },
  {
    "ph": "s", "id": 133165, "pid": 76337, "tid": -914061504, "ts": 1716454223416198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416297, "dur": 1,
    "args": {
      "External id": 133183,
      "cbid": 251, "correlation": 133183
    }
  },
  {
    "ph": "f", "id": 133183, "pid": 76337, "tid": -914061504, "ts": 1716454223416297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223478001, "dur": 168,
    "args": {
      "External id": 133185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133185,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133185, "pid": 5, "tid": 7, "ts": 1716454223478001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416302, "dur": 13,
    "args": {
      "External id": 133185,
      "cbid": 211, "correlation": 133185
    }
  },
  {
    "ph": "s", "id": 133185, "pid": 76337, "tid": -914061504, "ts": 1716454223416302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223478171, "dur": 19,
    "args": {
      "External id": 133193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133193,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133193, "pid": 5, "tid": 7, "ts": 1716454223478171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416373, "dur": 12,
    "args": {
      "External id": 133193,
      "cbid": 211, "correlation": 133193
    }
  },
  {
    "ph": "s", "id": 133193, "pid": 76337, "tid": -914061504, "ts": 1716454223416373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223478191, "dur": 29,
    "args": {
      "External id": 133201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133201,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133201, "pid": 5, "tid": 7, "ts": 1716454223478191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416412, "dur": 9,
    "args": {
      "External id": 133201,
      "cbid": 211, "correlation": 133201
    }
  },
  {
    "ph": "s", "id": 133201, "pid": 76337, "tid": -914061504, "ts": 1716454223416412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223478221, "dur": 19,
    "args": {
      "External id": 133212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133212, "pid": 5, "tid": 7, "ts": 1716454223478221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416484, "dur": 12,
    "args": {
      "External id": 133212,
      "cbid": 211, "correlation": 133212
    }
  },
  {
    "ph": "s", "id": 133212, "pid": 76337, "tid": -914061504, "ts": 1716454223416484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223478241, "dur": 16,
    "args": {
      "External id": 133234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133234,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133234, "pid": 5, "tid": 7, "ts": 1716454223478241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416514, "dur": 8,
    "args": {
      "External id": 133234,
      "cbid": 211, "correlation": 133234
    }
  },
  {
    "ph": "s", "id": 133234, "pid": 76337, "tid": -914061504, "ts": 1716454223416514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416599, "dur": 1,
    "args": {
      "External id": 133245,
      "cbid": 251, "correlation": 133245
    }
  },
  {
    "ph": "f", "id": 133245, "pid": 76337, "tid": -914061504, "ts": 1716454223416599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223478258, "dur": 89,
    "args": {
      "External id": 133246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133246,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133246, "pid": 5, "tid": 7, "ts": 1716454223478258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416604, "dur": 13,
    "args": {
      "External id": 133246,
      "cbid": 211, "correlation": 133246
    }
  },
  {
    "ph": "s", "id": 133246, "pid": 76337, "tid": -914061504, "ts": 1716454223416604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416674, "dur": 1,
    "args": {
      "External id": 133257,
      "cbid": 251, "correlation": 133257
    }
  },
  {
    "ph": "f", "id": 133257, "pid": 76337, "tid": -914061504, "ts": 1716454223416674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416678, "dur": 0,
    "args": {
      "External id": 133258,
      "cbid": 251, "correlation": 133258
    }
  },
  {
    "ph": "f", "id": 133258, "pid": 76337, "tid": -914061504, "ts": 1716454223416678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223478349, "dur": 13,
    "args": {
      "External id": 133259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133259,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133259, "pid": 5, "tid": 7, "ts": 1716454223478349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416680, "dur": 11,
    "args": {
      "External id": 133259,
      "cbid": 211, "correlation": 133259
    }
  },
  {
    "ph": "s", "id": 133259, "pid": 76337, "tid": -914061504, "ts": 1716454223416680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223478363, "dur": 6,
    "args": {
      "External id": 133261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133261,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133261, "pid": 5, "tid": 7, "ts": 1716454223478363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416693, "dur": 6,
    "args": {
      "External id": 133261,
      "cbid": 211, "correlation": 133261
    }
  },
  {
    "ph": "s", "id": 133261, "pid": 76337, "tid": -914061504, "ts": 1716454223416693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416750, "dur": 1,
    "args": {
      "External id": 133272,
      "cbid": 251, "correlation": 133272
    }
  },
  {
    "ph": "f", "id": 133272, "pid": 76337, "tid": -914061504, "ts": 1716454223416750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416753, "dur": 0,
    "args": {
      "External id": 133273,
      "cbid": 251, "correlation": 133273
    }
  },
  {
    "ph": "f", "id": 133273, "pid": 76337, "tid": -914061504, "ts": 1716454223416753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223478369, "dur": 8,
    "args": {
      "External id": 133274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133274,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133274, "pid": 5, "tid": 7, "ts": 1716454223478369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416755, "dur": 12,
    "args": {
      "External id": 133274,
      "cbid": 211, "correlation": 133274
    }
  },
  {
    "ph": "s", "id": 133274, "pid": 76337, "tid": -914061504, "ts": 1716454223416755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223478379, "dur": 3,
    "args": {
      "External id": 133276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133276,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133276, "pid": 5, "tid": 7, "ts": 1716454223478379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416768, "dur": 6,
    "args": {
      "External id": 133276,
      "cbid": 211, "correlation": 133276
    }
  },
  {
    "ph": "s", "id": 133276, "pid": 76337, "tid": -914061504, "ts": 1716454223416768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223478383, "dur": 54,
    "args": {
      "External id": 133301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133301,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133301, "pid": 5, "tid": 7, "ts": 1716454223478383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416844, "dur": 12,
    "args": {
      "External id": 133301,
      "cbid": 211, "correlation": 133301
    }
  },
  {
    "ph": "s", "id": 133301, "pid": 76337, "tid": -914061504, "ts": 1716454223416844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416944, "dur": 1,
    "args": {
      "External id": 133319,
      "cbid": 251, "correlation": 133319
    }
  },
  {
    "ph": "f", "id": 133319, "pid": 76337, "tid": -914061504, "ts": 1716454223416944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223478438, "dur": 91,
    "args": {
      "External id": 133321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133321,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133321, "pid": 5, "tid": 7, "ts": 1716454223478438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223416950, "dur": 14,
    "args": {
      "External id": 133321,
      "cbid": 211, "correlation": 133321
    }
  },
  {
    "ph": "s", "id": 133321, "pid": 76337, "tid": -914061504, "ts": 1716454223416950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223478530, "dur": 9,
    "args": {
      "External id": 133329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133329,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133329, "pid": 5, "tid": 7, "ts": 1716454223478530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417029, "dur": 13,
    "args": {
      "External id": 133329,
      "cbid": 211, "correlation": 133329
    }
  },
  {
    "ph": "s", "id": 133329, "pid": 76337, "tid": -914061504, "ts": 1716454223417029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223478541, "dur": 21,
    "args": {
      "External id": 133337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133337,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133337, "pid": 5, "tid": 7, "ts": 1716454223478541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417071, "dur": 10,
    "args": {
      "External id": 133337,
      "cbid": 211, "correlation": 133337
    }
  },
  {
    "ph": "s", "id": 133337, "pid": 76337, "tid": -914061504, "ts": 1716454223417071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223478564, "dur": 18,
    "args": {
      "External id": 133359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133359,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133359, "pid": 5, "tid": 7, "ts": 1716454223478564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417123, "dur": 10,
    "args": {
      "External id": 133359,
      "cbid": 211, "correlation": 133359
    }
  },
  {
    "ph": "s", "id": 133359, "pid": 76337, "tid": -914061504, "ts": 1716454223417123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417208, "dur": 1,
    "args": {
      "External id": 133375,
      "cbid": 251, "correlation": 133375
    }
  },
  {
    "ph": "f", "id": 133375, "pid": 76337, "tid": -914061504, "ts": 1716454223417208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417214, "dur": 0,
    "args": {
      "External id": 133377,
      "cbid": 251, "correlation": 133377
    }
  },
  {
    "ph": "f", "id": 133377, "pid": 76337, "tid": -914061504, "ts": 1716454223417214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223478583, "dur": 495,
    "args": {
      "External id": 133378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133378,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133378, "pid": 5, "tid": 7, "ts": 1716454223478583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417215, "dur": 14,
    "args": {
      "External id": 133378,
      "cbid": 211, "correlation": 133378
    }
  },
  {
    "ph": "s", "id": 133378, "pid": 76337, "tid": -914061504, "ts": 1716454223417215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223479079, "dur": 65,
    "args": {
      "External id": 133386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133386, "pid": 5, "tid": 7, "ts": 1716454223479079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417282, "dur": 12,
    "args": {
      "External id": 133386,
      "cbid": 211, "correlation": 133386
    }
  },
  {
    "ph": "s", "id": 133386, "pid": 76337, "tid": -914061504, "ts": 1716454223417282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223479146, "dur": 67,
    "args": {
      "External id": 133394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133394, "pid": 5, "tid": 7, "ts": 1716454223479146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417312, "dur": 9,
    "args": {
      "External id": 133394,
      "cbid": 211, "correlation": 133394
    }
  },
  {
    "ph": "s", "id": 133394, "pid": 76337, "tid": -914061504, "ts": 1716454223417312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417392, "dur": 1,
    "args": {
      "External id": 133410,
      "cbid": 251, "correlation": 133410
    }
  },
  {
    "ph": "f", "id": 133410, "pid": 76337, "tid": -914061504, "ts": 1716454223417392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223479215, "dur": 1,
    "args": {
      "External id": 133412,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 133412,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 133412, "pid": 5, "tid": 7, "ts": 1716454223479215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417397, "dur": 9,
    "args": {
      "External id": 133412,
      "cbid": 51, "correlation": 133412
    }
  },
  {
    "ph": "s", "id": 133412, "pid": 76337, "tid": -914061504, "ts": 1716454223417397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223479218, "dur": 268,
    "args": {
      "External id": 133413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133413,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133413, "pid": 5, "tid": 7, "ts": 1716454223479218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417408, "dur": 11,
    "args": {
      "External id": 133413,
      "cbid": 211, "correlation": 133413
    }
  },
  {
    "ph": "s", "id": 133413, "pid": 76337, "tid": -914061504, "ts": 1716454223417408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223479488, "dur": 14,
    "args": {
      "External id": 133421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133421,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133421, "pid": 5, "tid": 7, "ts": 1716454223479488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417450, "dur": 10,
    "args": {
      "External id": 133421,
      "cbid": 211, "correlation": 133421
    }
  },
  {
    "ph": "s", "id": 133421, "pid": 76337, "tid": -914061504, "ts": 1716454223417450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223479503, "dur": 37,
    "args": {
      "External id": 133432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133432,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133432, "pid": 5, "tid": 7, "ts": 1716454223479503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417518, "dur": 13,
    "args": {
      "External id": 133432,
      "cbid": 211, "correlation": 133432
    }
  },
  {
    "ph": "s", "id": 133432, "pid": 76337, "tid": -914061504, "ts": 1716454223417518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417583, "dur": 0,
    "args": {
      "External id": 133444,
      "cbid": 317, "correlation": 133444
    }
  },
  {
    "ph": "f", "id": 133444, "pid": 76337, "tid": -914061504, "ts": 1716454223417583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417584, "dur": 0,
    "args": {
      "External id": 133445,
      "cbid": 203, "correlation": 133445
    }
  },
  {
    "ph": "f", "id": 133445, "pid": 76337, "tid": -914061504, "ts": 1716454223417584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417585, "dur": 0,
    "args": {
      "External id": 133446,
      "cbid": 205, "correlation": 133446
    }
  },
  {
    "ph": "f", "id": 133446, "pid": 76337, "tid": -914061504, "ts": 1716454223417585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223479541, "dur": 12,
    "args": {
      "External id": 133450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133450,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133450, "pid": 5, "tid": 7, "ts": 1716454223479541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417599, "dur": 12,
    "args": {
      "External id": 133450,
      "cbid": 211, "correlation": 133450
    }
  },
  {
    "ph": "s", "id": 133450, "pid": 76337, "tid": -914061504, "ts": 1716454223417599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223479555, "dur": 4,
    "args": {
      "External id": 133452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133452, "pid": 5, "tid": 7, "ts": 1716454223479555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417615, "dur": 6,
    "args": {
      "External id": 133452,
      "cbid": 211, "correlation": 133452
    }
  },
  {
    "ph": "s", "id": 133452, "pid": 76337, "tid": -914061504, "ts": 1716454223417615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417624, "dur": 0,
    "args": {
      "External id": 133453,
      "cbid": 51, "correlation": 133453
    }
  },
  {
    "ph": "s", "id": 133453, "pid": 76337, "tid": -914061504, "ts": 1716454223417624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223479560, "dur": 95,
    "args": {
      "External id": 133454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133454,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 133454, "pid": 5, "tid": 7, "ts": 1716454223479560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417625, "dur": 5,
    "args": {
      "External id": 133454,
      "cbid": 211, "correlation": 133454
    }
  },
  {
    "ph": "s", "id": 133454, "pid": 76337, "tid": -914061504, "ts": 1716454223417625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223479657, "dur": 17,
    "args": {
      "External id": 133459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133459,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133459, "pid": 5, "tid": 7, "ts": 1716454223479657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417652, "dur": 9,
    "args": {
      "External id": 133459,
      "cbid": 211, "correlation": 133459
    }
  },
  {
    "ph": "s", "id": 133459, "pid": 76337, "tid": -914061504, "ts": 1716454223417652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223479675, "dur": 12,
    "args": {
      "External id": 133467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133467,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133467, "pid": 5, "tid": 7, "ts": 1716454223479675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417684, "dur": 8,
    "args": {
      "External id": 133467,
      "cbid": 211, "correlation": 133467
    }
  },
  {
    "ph": "s", "id": 133467, "pid": 76337, "tid": -914061504, "ts": 1716454223417684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223479688, "dur": 25,
    "args": {
      "External id": 133476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133476,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133476, "pid": 5, "tid": 7, "ts": 1716454223479688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417723, "dur": 10,
    "args": {
      "External id": 133476,
      "cbid": 211, "correlation": 133476
    }
  },
  {
    "ph": "s", "id": 133476, "pid": 76337, "tid": -914061504, "ts": 1716454223417723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223479715, "dur": 24,
    "args": {
      "External id": 133496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133496,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 133496, "pid": 5, "tid": 7, "ts": 1716454223479715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417793, "dur": 11,
    "args": {
      "External id": 133496,
      "cbid": 211, "correlation": 133496
    }
  },
  {
    "ph": "s", "id": 133496, "pid": 76337, "tid": -914061504, "ts": 1716454223417793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223479740, "dur": 5,
    "args": {
      "External id": 133508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133508,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 133508, "pid": 5, "tid": 7, "ts": 1716454223479740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417814, "dur": 6,
    "args": {
      "External id": 133508,
      "cbid": 211, "correlation": 133508
    }
  },
  {
    "ph": "s", "id": 133508, "pid": 76337, "tid": -914061504, "ts": 1716454223417814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223479746, "dur": 26,
    "args": {
      "External id": 133511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133511, "pid": 5, "tid": 7, "ts": 1716454223479746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417833, "dur": 7,
    "args": {
      "External id": 133511,
      "cbid": 211, "correlation": 133511
    }
  },
  {
    "ph": "s", "id": 133511, "pid": 76337, "tid": -914061504, "ts": 1716454223417833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223479773, "dur": 17,
    "args": {
      "External id": 133520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133520,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133520, "pid": 5, "tid": 7, "ts": 1716454223479773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417872, "dur": 10,
    "args": {
      "External id": 133520,
      "cbid": 211, "correlation": 133520
    }
  },
  {
    "ph": "s", "id": 133520, "pid": 76337, "tid": -914061504, "ts": 1716454223417872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417924, "dur": 0,
    "args": {
      "External id": 133530,
      "cbid": 317, "correlation": 133530
    }
  },
  {
    "ph": "f", "id": 133530, "pid": 76337, "tid": -914061504, "ts": 1716454223417924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417925, "dur": 0,
    "args": {
      "External id": 133531,
      "cbid": 203, "correlation": 133531
    }
  },
  {
    "ph": "f", "id": 133531, "pid": 76337, "tid": -914061504, "ts": 1716454223417925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417926, "dur": 0,
    "args": {
      "External id": 133532,
      "cbid": 205, "correlation": 133532
    }
  },
  {
    "ph": "f", "id": 133532, "pid": 76337, "tid": -914061504, "ts": 1716454223417926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223479791, "dur": 17,
    "args": {
      "External id": 133536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133536,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133536, "pid": 5, "tid": 7, "ts": 1716454223479791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417940, "dur": 11,
    "args": {
      "External id": 133536,
      "cbid": 211, "correlation": 133536
    }
  },
  {
    "ph": "s", "id": 133536, "pid": 76337, "tid": -914061504, "ts": 1716454223417940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223479809, "dur": 241,
    "args": {
      "External id": 133538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133538,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133538, "pid": 5, "tid": 7, "ts": 1716454223479809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417954, "dur": 5,
    "args": {
      "External id": 133538,
      "cbid": 211, "correlation": 133538
    }
  },
  {
    "ph": "s", "id": 133538, "pid": 76337, "tid": -914061504, "ts": 1716454223417954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223480053, "dur": 1,
    "args": {
      "External id": 133540,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 133540,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 133540, "pid": 5, "tid": 7, "ts": 1716454223480053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417966, "dur": 16,
    "args": {
      "External id": 133540,
      "cbid": 51, "correlation": 133540
    }
  },
  {
    "ph": "s", "id": 133540, "pid": 76337, "tid": -914061504, "ts": 1716454223417966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223480056, "dur": 808,
    "args": {
      "External id": 133541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133541,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133541, "pid": 5, "tid": 7, "ts": 1716454223480056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417983, "dur": 7,
    "args": {
      "External id": 133541,
      "cbid": 211, "correlation": 133541
    }
  },
  {
    "ph": "s", "id": 133541, "pid": 76337, "tid": -914061504, "ts": 1716454223417983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223480866, "dur": 13,
    "args": {
      "External id": 133543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133543,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133543, "pid": 5, "tid": 7, "ts": 1716454223480866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223417994, "dur": 5,
    "args": {
      "External id": 133543,
      "cbid": 211, "correlation": 133543
    }
  },
  {
    "ph": "s", "id": 133543, "pid": 76337, "tid": -914061504, "ts": 1716454223417994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223480880, "dur": 15,
    "args": {
      "External id": 133549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133549,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133549, "pid": 5, "tid": 7, "ts": 1716454223480880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418023, "dur": 8,
    "args": {
      "External id": 133549,
      "cbid": 211, "correlation": 133549
    }
  },
  {
    "ph": "s", "id": 133549, "pid": 76337, "tid": -914061504, "ts": 1716454223418023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223480897, "dur": 3,
    "args": {
      "External id": 133557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133557,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 133557, "pid": 5, "tid": 7, "ts": 1716454223480897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418067, "dur": 10,
    "args": {
      "External id": 133557,
      "cbid": 211, "correlation": 133557
    }
  },
  {
    "ph": "s", "id": 133557, "pid": 76337, "tid": -914061504, "ts": 1716454223418067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418133, "dur": 1,
    "args": {
      "External id": 133573,
      "cbid": 251, "correlation": 133573
    }
  },
  {
    "ph": "f", "id": 133573, "pid": 76337, "tid": -914061504, "ts": 1716454223418133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418138, "dur": 0,
    "args": {
      "External id": 133575,
      "cbid": 251, "correlation": 133575
    }
  },
  {
    "ph": "f", "id": 133575, "pid": 76337, "tid": -914061504, "ts": 1716454223418138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223480901, "dur": 13,
    "args": {
      "External id": 133576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133576,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133576, "pid": 5, "tid": 7, "ts": 1716454223480901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418140, "dur": 11,
    "args": {
      "External id": 133576,
      "cbid": 211, "correlation": 133576
    }
  },
  {
    "ph": "s", "id": 133576, "pid": 76337, "tid": -914061504, "ts": 1716454223418140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223480916, "dur": 5,
    "args": {
      "External id": 133578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133578,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133578, "pid": 5, "tid": 7, "ts": 1716454223480916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418153, "dur": 5,
    "args": {
      "External id": 133578,
      "cbid": 211, "correlation": 133578
    }
  },
  {
    "ph": "s", "id": 133578, "pid": 76337, "tid": -914061504, "ts": 1716454223418153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223480922, "dur": 17,
    "args": {
      "External id": 133588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133588,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133588, "pid": 5, "tid": 7, "ts": 1716454223480922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418211, "dur": 13,
    "args": {
      "External id": 133588,
      "cbid": 211, "correlation": 133588
    }
  },
  {
    "ph": "s", "id": 133588, "pid": 76337, "tid": -914061504, "ts": 1716454223418211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223480940, "dur": 18,
    "args": {
      "External id": 133608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133608,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 133608, "pid": 5, "tid": 7, "ts": 1716454223480940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418278, "dur": 10,
    "args": {
      "External id": 133608,
      "cbid": 211, "correlation": 133608
    }
  },
  {
    "ph": "s", "id": 133608, "pid": 76337, "tid": -914061504, "ts": 1716454223418278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223480959, "dur": 4,
    "args": {
      "External id": 133620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133620,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 133620, "pid": 5, "tid": 7, "ts": 1716454223480959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418298, "dur": 6,
    "args": {
      "External id": 133620,
      "cbid": 211, "correlation": 133620
    }
  },
  {
    "ph": "s", "id": 133620, "pid": 76337, "tid": -914061504, "ts": 1716454223418298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223480965, "dur": 16,
    "args": {
      "External id": 133623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133623,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133623, "pid": 5, "tid": 7, "ts": 1716454223480965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418317, "dur": 7,
    "args": {
      "External id": 133623,
      "cbid": 211, "correlation": 133623
    }
  },
  {
    "ph": "s", "id": 133623, "pid": 76337, "tid": -914061504, "ts": 1716454223418317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223480982, "dur": 11,
    "args": {
      "External id": 133632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133632,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133632, "pid": 5, "tid": 7, "ts": 1716454223480982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418357, "dur": 10,
    "args": {
      "External id": 133632,
      "cbid": 211, "correlation": 133632
    }
  },
  {
    "ph": "s", "id": 133632, "pid": 76337, "tid": -914061504, "ts": 1716454223418357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418419, "dur": 0,
    "args": {
      "External id": 133642,
      "cbid": 317, "correlation": 133642
    }
  },
  {
    "ph": "f", "id": 133642, "pid": 76337, "tid": -914061504, "ts": 1716454223418419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418420, "dur": 0,
    "args": {
      "External id": 133643,
      "cbid": 203, "correlation": 133643
    }
  },
  {
    "ph": "f", "id": 133643, "pid": 76337, "tid": -914061504, "ts": 1716454223418420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418421, "dur": 0,
    "args": {
      "External id": 133644,
      "cbid": 205, "correlation": 133644
    }
  },
  {
    "ph": "f", "id": 133644, "pid": 76337, "tid": -914061504, "ts": 1716454223418421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223480994, "dur": 11,
    "args": {
      "External id": 133648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133648,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133648, "pid": 5, "tid": 7, "ts": 1716454223480994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418434, "dur": 12,
    "args": {
      "External id": 133648,
      "cbid": 211, "correlation": 133648
    }
  },
  {
    "ph": "s", "id": 133648, "pid": 76337, "tid": -914061504, "ts": 1716454223418434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223481007, "dur": 163,
    "args": {
      "External id": 133650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133650,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133650, "pid": 5, "tid": 7, "ts": 1716454223481007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418448, "dur": 5,
    "args": {
      "External id": 133650,
      "cbid": 211, "correlation": 133650
    }
  },
  {
    "ph": "s", "id": 133650, "pid": 76337, "tid": -914061504, "ts": 1716454223418448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223481172, "dur": 1,
    "args": {
      "External id": 133652,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 133652,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 133652, "pid": 5, "tid": 7, "ts": 1716454223481172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418459, "dur": 7,
    "args": {
      "External id": 133652,
      "cbid": 51, "correlation": 133652
    }
  },
  {
    "ph": "s", "id": 133652, "pid": 76337, "tid": -914061504, "ts": 1716454223418459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223481176, "dur": 645,
    "args": {
      "External id": 133653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133653,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133653, "pid": 5, "tid": 7, "ts": 1716454223481176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418467, "dur": 6,
    "args": {
      "External id": 133653,
      "cbid": 211, "correlation": 133653
    }
  },
  {
    "ph": "s", "id": 133653, "pid": 76337, "tid": -914061504, "ts": 1716454223418467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223481822, "dur": 13,
    "args": {
      "External id": 133655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133655, "pid": 5, "tid": 7, "ts": 1716454223481822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418478, "dur": 5,
    "args": {
      "External id": 133655,
      "cbid": 211, "correlation": 133655
    }
  },
  {
    "ph": "s", "id": 133655, "pid": 76337, "tid": -914061504, "ts": 1716454223418478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223481837, "dur": 14,
    "args": {
      "External id": 133661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133661, "pid": 5, "tid": 7, "ts": 1716454223481837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418506, "dur": 9,
    "args": {
      "External id": 133661,
      "cbid": 211, "correlation": 133661
    }
  },
  {
    "ph": "s", "id": 133661, "pid": 76337, "tid": -914061504, "ts": 1716454223418506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418565, "dur": 0,
    "args": {
      "External id": 133671,
      "cbid": 317, "correlation": 133671
    }
  },
  {
    "ph": "f", "id": 133671, "pid": 76337, "tid": -914061504, "ts": 1716454223418565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418566, "dur": 0,
    "args": {
      "External id": 133672,
      "cbid": 203, "correlation": 133672
    }
  },
  {
    "ph": "f", "id": 133672, "pid": 76337, "tid": -914061504, "ts": 1716454223418566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418566, "dur": 0,
    "args": {
      "External id": 133673,
      "cbid": 205, "correlation": 133673
    }
  },
  {
    "ph": "f", "id": 133673, "pid": 76337, "tid": -914061504, "ts": 1716454223418566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223481852, "dur": 17,
    "args": {
      "External id": 133677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133677,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133677, "pid": 5, "tid": 7, "ts": 1716454223481852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418580, "dur": 12,
    "args": {
      "External id": 133677,
      "cbid": 211, "correlation": 133677
    }
  },
  {
    "ph": "s", "id": 133677, "pid": 76337, "tid": -914061504, "ts": 1716454223418580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223481871, "dur": 4,
    "args": {
      "External id": 133679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133679,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133679, "pid": 5, "tid": 7, "ts": 1716454223481871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418596, "dur": 6,
    "args": {
      "External id": 133679,
      "cbid": 211, "correlation": 133679
    }
  },
  {
    "ph": "s", "id": 133679, "pid": 76337, "tid": -914061504, "ts": 1716454223418596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418605, "dur": 0,
    "args": {
      "External id": 133680,
      "cbid": 51, "correlation": 133680
    }
  },
  {
    "ph": "s", "id": 133680, "pid": 76337, "tid": -914061504, "ts": 1716454223418605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223481876, "dur": 131,
    "args": {
      "External id": 133681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133681,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 133681, "pid": 5, "tid": 7, "ts": 1716454223481876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418606, "dur": 5,
    "args": {
      "External id": 133681,
      "cbid": 211, "correlation": 133681
    }
  },
  {
    "ph": "s", "id": 133681, "pid": 76337, "tid": -914061504, "ts": 1716454223418606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223482008, "dur": 15,
    "args": {
      "External id": 133686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133686,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133686, "pid": 5, "tid": 7, "ts": 1716454223482008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418632, "dur": 8,
    "args": {
      "External id": 133686,
      "cbid": 211, "correlation": 133686
    }
  },
  {
    "ph": "s", "id": 133686, "pid": 76337, "tid": -914061504, "ts": 1716454223418632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223482024, "dur": 12,
    "args": {
      "External id": 133694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133694,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133694, "pid": 5, "tid": 7, "ts": 1716454223482024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418660, "dur": 8,
    "args": {
      "External id": 133694,
      "cbid": 211, "correlation": 133694
    }
  },
  {
    "ph": "s", "id": 133694, "pid": 76337, "tid": -914061504, "ts": 1716454223418660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223482037, "dur": 10,
    "args": {
      "External id": 133702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133702,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133702, "pid": 5, "tid": 7, "ts": 1716454223482037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418689, "dur": 8,
    "args": {
      "External id": 133702,
      "cbid": 211, "correlation": 133702
    }
  },
  {
    "ph": "s", "id": 133702, "pid": 76337, "tid": -914061504, "ts": 1716454223418689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223482049, "dur": 18,
    "args": {
      "External id": 133722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133722,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 133722, "pid": 5, "tid": 7, "ts": 1716454223482049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418770, "dur": 13,
    "args": {
      "External id": 133722,
      "cbid": 211, "correlation": 133722
    }
  },
  {
    "ph": "s", "id": 133722, "pid": 76337, "tid": -914061504, "ts": 1716454223418770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223482068, "dur": 4,
    "args": {
      "External id": 133734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133734,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 133734, "pid": 5, "tid": 7, "ts": 1716454223482068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418793, "dur": 6,
    "args": {
      "External id": 133734,
      "cbid": 211, "correlation": 133734
    }
  },
  {
    "ph": "s", "id": 133734, "pid": 76337, "tid": -914061504, "ts": 1716454223418793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223482074, "dur": 17,
    "args": {
      "External id": 133737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133737, "pid": 5, "tid": 7, "ts": 1716454223482074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418811, "dur": 7,
    "args": {
      "External id": 133737,
      "cbid": 211, "correlation": 133737
    }
  },
  {
    "ph": "s", "id": 133737, "pid": 76337, "tid": -914061504, "ts": 1716454223418811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418867, "dur": 0,
    "args": {
      "External id": 133748,
      "cbid": 317, "correlation": 133748
    }
  },
  {
    "ph": "f", "id": 133748, "pid": 76337, "tid": -914061504, "ts": 1716454223418867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418868, "dur": 0,
    "args": {
      "External id": 133749,
      "cbid": 203, "correlation": 133749
    }
  },
  {
    "ph": "f", "id": 133749, "pid": 76337, "tid": -914061504, "ts": 1716454223418868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418869, "dur": 0,
    "args": {
      "External id": 133750,
      "cbid": 205, "correlation": 133750
    }
  },
  {
    "ph": "f", "id": 133750, "pid": 76337, "tid": -914061504, "ts": 1716454223418869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223482092, "dur": 11,
    "args": {
      "External id": 133754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133754,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133754, "pid": 5, "tid": 7, "ts": 1716454223482092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418882, "dur": 12,
    "args": {
      "External id": 133754,
      "cbid": 211, "correlation": 133754
    }
  },
  {
    "ph": "s", "id": 133754, "pid": 76337, "tid": -914061504, "ts": 1716454223418882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223482104, "dur": 3,
    "args": {
      "External id": 133756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133756,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133756, "pid": 5, "tid": 7, "ts": 1716454223482104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418898, "dur": 5,
    "args": {
      "External id": 133756,
      "cbid": 211, "correlation": 133756
    }
  },
  {
    "ph": "s", "id": 133756, "pid": 76337, "tid": -914061504, "ts": 1716454223418898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418906, "dur": 0,
    "args": {
      "External id": 133757,
      "cbid": 51, "correlation": 133757
    }
  },
  {
    "ph": "s", "id": 133757, "pid": 76337, "tid": -914061504, "ts": 1716454223418906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223482109, "dur": 90,
    "args": {
      "External id": 133758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133758,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 133758, "pid": 5, "tid": 7, "ts": 1716454223482109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418907, "dur": 5,
    "args": {
      "External id": 133758,
      "cbid": 211, "correlation": 133758
    }
  },
  {
    "ph": "s", "id": 133758, "pid": 76337, "tid": -914061504, "ts": 1716454223418907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223482200, "dur": 15,
    "args": {
      "External id": 133763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133763, "pid": 5, "tid": 7, "ts": 1716454223482200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223418933, "dur": 8,
    "args": {
      "External id": 133763,
      "cbid": 211, "correlation": 133763
    }
  },
  {
    "ph": "s", "id": 133763, "pid": 76337, "tid": -914061504, "ts": 1716454223418933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223482217, "dur": 83,
    "args": {
      "External id": 133772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133772,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133772, "pid": 5, "tid": 7, "ts": 1716454223482217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419023, "dur": 14,
    "args": {
      "External id": 133772,
      "cbid": 211, "correlation": 133772
    }
  },
  {
    "ph": "s", "id": 133772, "pid": 76337, "tid": -914061504, "ts": 1716454223419023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223482301, "dur": 30,
    "args": {
      "External id": 133794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133794,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133794, "pid": 5, "tid": 7, "ts": 1716454223482301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419081, "dur": 10,
    "args": {
      "External id": 133794,
      "cbid": 211, "correlation": 133794
    }
  },
  {
    "ph": "s", "id": 133794, "pid": 76337, "tid": -914061504, "ts": 1716454223419081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419169, "dur": 1,
    "args": {
      "External id": 133805,
      "cbid": 251, "correlation": 133805
    }
  },
  {
    "ph": "f", "id": 133805, "pid": 76337, "tid": -914061504, "ts": 1716454223419169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223482332, "dur": 164,
    "args": {
      "External id": 133806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133806,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133806, "pid": 5, "tid": 7, "ts": 1716454223482332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419175, "dur": 13,
    "args": {
      "External id": 133806,
      "cbid": 211, "correlation": 133806
    }
  },
  {
    "ph": "s", "id": 133806, "pid": 76337, "tid": -914061504, "ts": 1716454223419175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419244, "dur": 1,
    "args": {
      "External id": 133817,
      "cbid": 251, "correlation": 133817
    }
  },
  {
    "ph": "f", "id": 133817, "pid": 76337, "tid": -914061504, "ts": 1716454223419244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223482498, "dur": 156,
    "args": {
      "External id": 133818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133818,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133818, "pid": 5, "tid": 7, "ts": 1716454223482498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419248, "dur": 11,
    "args": {
      "External id": 133818,
      "cbid": 211, "correlation": 133818
    }
  },
  {
    "ph": "s", "id": 133818, "pid": 76337, "tid": -914061504, "ts": 1716454223419248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419314, "dur": 1,
    "args": {
      "External id": 133829,
      "cbid": 251, "correlation": 133829
    }
  },
  {
    "ph": "f", "id": 133829, "pid": 76337, "tid": -914061504, "ts": 1716454223419314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223482656, "dur": 158,
    "args": {
      "External id": 133830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133830,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133830, "pid": 5, "tid": 7, "ts": 1716454223482656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419318, "dur": 12,
    "args": {
      "External id": 133830,
      "cbid": 211, "correlation": 133830
    }
  },
  {
    "ph": "s", "id": 133830, "pid": 76337, "tid": -914061504, "ts": 1716454223419318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223482815, "dur": 335,
    "args": {
      "External id": 133855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133855,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133855, "pid": 5, "tid": 7, "ts": 1716454223482815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419401, "dur": 12,
    "args": {
      "External id": 133855,
      "cbid": 211, "correlation": 133855
    }
  },
  {
    "ph": "s", "id": 133855, "pid": 76337, "tid": -914061504, "ts": 1716454223419401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419501, "dur": 1,
    "args": {
      "External id": 133873,
      "cbid": 251, "correlation": 133873
    }
  },
  {
    "ph": "f", "id": 133873, "pid": 76337, "tid": -914061504, "ts": 1716454223419501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223483152, "dur": 164,
    "args": {
      "External id": 133875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133875,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133875, "pid": 5, "tid": 7, "ts": 1716454223483152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419507, "dur": 13,
    "args": {
      "External id": 133875,
      "cbid": 211, "correlation": 133875
    }
  },
  {
    "ph": "s", "id": 133875, "pid": 76337, "tid": -914061504, "ts": 1716454223419507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223483318, "dur": 20,
    "args": {
      "External id": 133883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133883,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133883, "pid": 5, "tid": 7, "ts": 1716454223483318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419578, "dur": 12,
    "args": {
      "External id": 133883,
      "cbid": 211, "correlation": 133883
    }
  },
  {
    "ph": "s", "id": 133883, "pid": 76337, "tid": -914061504, "ts": 1716454223419578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223483338, "dur": 28,
    "args": {
      "External id": 133891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133891,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133891, "pid": 5, "tid": 7, "ts": 1716454223483338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419616, "dur": 9,
    "args": {
      "External id": 133891,
      "cbid": 211, "correlation": 133891
    }
  },
  {
    "ph": "s", "id": 133891, "pid": 76337, "tid": -914061504, "ts": 1716454223419616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223483368, "dur": 19,
    "args": {
      "External id": 133902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133902,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133902, "pid": 5, "tid": 7, "ts": 1716454223483368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419689, "dur": 13,
    "args": {
      "External id": 133902,
      "cbid": 211, "correlation": 133902
    }
  },
  {
    "ph": "s", "id": 133902, "pid": 76337, "tid": -914061504, "ts": 1716454223419689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223483388, "dur": 16,
    "args": {
      "External id": 133924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133924,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 133924, "pid": 5, "tid": 7, "ts": 1716454223483388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419719, "dur": 8,
    "args": {
      "External id": 133924,
      "cbid": 211, "correlation": 133924
    }
  },
  {
    "ph": "s", "id": 133924, "pid": 76337, "tid": -914061504, "ts": 1716454223419719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419805, "dur": 1,
    "args": {
      "External id": 133935,
      "cbid": 251, "correlation": 133935
    }
  },
  {
    "ph": "f", "id": 133935, "pid": 76337, "tid": -914061504, "ts": 1716454223419805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223483405, "dur": 88,
    "args": {
      "External id": 133936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133936,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 133936, "pid": 5, "tid": 7, "ts": 1716454223483405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419810, "dur": 13,
    "args": {
      "External id": 133936,
      "cbid": 211, "correlation": 133936
    }
  },
  {
    "ph": "s", "id": 133936, "pid": 76337, "tid": -914061504, "ts": 1716454223419810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419880, "dur": 1,
    "args": {
      "External id": 133947,
      "cbid": 251, "correlation": 133947
    }
  },
  {
    "ph": "f", "id": 133947, "pid": 76337, "tid": -914061504, "ts": 1716454223419880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419883, "dur": 0,
    "args": {
      "External id": 133948,
      "cbid": 251, "correlation": 133948
    }
  },
  {
    "ph": "f", "id": 133948, "pid": 76337, "tid": -914061504, "ts": 1716454223419883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223483495, "dur": 12,
    "args": {
      "External id": 133949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133949,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133949, "pid": 5, "tid": 7, "ts": 1716454223483495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419885, "dur": 12,
    "args": {
      "External id": 133949,
      "cbid": 211, "correlation": 133949
    }
  },
  {
    "ph": "s", "id": 133949, "pid": 76337, "tid": -914061504, "ts": 1716454223419885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223483508, "dur": 5,
    "args": {
      "External id": 133951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133951,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133951, "pid": 5, "tid": 7, "ts": 1716454223483508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419899, "dur": 6,
    "args": {
      "External id": 133951,
      "cbid": 211, "correlation": 133951
    }
  },
  {
    "ph": "s", "id": 133951, "pid": 76337, "tid": -914061504, "ts": 1716454223419899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419955, "dur": 1,
    "args": {
      "External id": 133962,
      "cbid": 251, "correlation": 133962
    }
  },
  {
    "ph": "f", "id": 133962, "pid": 76337, "tid": -914061504, "ts": 1716454223419955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419958, "dur": 0,
    "args": {
      "External id": 133963,
      "cbid": 251, "correlation": 133963
    }
  },
  {
    "ph": "f", "id": 133963, "pid": 76337, "tid": -914061504, "ts": 1716454223419958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223483515, "dur": 8,
    "args": {
      "External id": 133964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133964,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133964, "pid": 5, "tid": 7, "ts": 1716454223483515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419960, "dur": 11,
    "args": {
      "External id": 133964,
      "cbid": 211, "correlation": 133964
    }
  },
  {
    "ph": "s", "id": 133964, "pid": 76337, "tid": -914061504, "ts": 1716454223419960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223483524, "dur": 3,
    "args": {
      "External id": 133966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133966,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133966, "pid": 5, "tid": 7, "ts": 1716454223483524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223419981, "dur": 6,
    "args": {
      "External id": 133966,
      "cbid": 211, "correlation": 133966
    }
  },
  {
    "ph": "s", "id": 133966, "pid": 76337, "tid": -914061504, "ts": 1716454223419981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223483529, "dur": 55,
    "args": {
      "External id": 133991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 133991,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 133991, "pid": 5, "tid": 7, "ts": 1716454223483529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420058, "dur": 12,
    "args": {
      "External id": 133991,
      "cbid": 211, "correlation": 133991
    }
  },
  {
    "ph": "s", "id": 133991, "pid": 76337, "tid": -914061504, "ts": 1716454223420058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420157, "dur": 1,
    "args": {
      "External id": 134009,
      "cbid": 251, "correlation": 134009
    }
  },
  {
    "ph": "f", "id": 134009, "pid": 76337, "tid": -914061504, "ts": 1716454223420157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223483584, "dur": 90,
    "args": {
      "External id": 134011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134011,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 134011, "pid": 5, "tid": 7, "ts": 1716454223483584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420162, "dur": 13,
    "args": {
      "External id": 134011,
      "cbid": 211, "correlation": 134011
    }
  },
  {
    "ph": "s", "id": 134011, "pid": 76337, "tid": -914061504, "ts": 1716454223420162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223483675, "dur": 9,
    "args": {
      "External id": 134019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134019,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134019, "pid": 5, "tid": 7, "ts": 1716454223483675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420232, "dur": 12,
    "args": {
      "External id": 134019,
      "cbid": 211, "correlation": 134019
    }
  },
  {
    "ph": "s", "id": 134019, "pid": 76337, "tid": -914061504, "ts": 1716454223420232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223483686, "dur": 21,
    "args": {
      "External id": 134027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134027, "pid": 5, "tid": 7, "ts": 1716454223483686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420274, "dur": 10,
    "args": {
      "External id": 134027,
      "cbid": 211, "correlation": 134027
    }
  },
  {
    "ph": "s", "id": 134027, "pid": 76337, "tid": -914061504, "ts": 1716454223420274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223483708, "dur": 18,
    "args": {
      "External id": 134049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134049,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134049, "pid": 5, "tid": 7, "ts": 1716454223483708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420326, "dur": 10,
    "args": {
      "External id": 134049,
      "cbid": 211, "correlation": 134049
    }
  },
  {
    "ph": "s", "id": 134049, "pid": 76337, "tid": -914061504, "ts": 1716454223420326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420413, "dur": 1,
    "args": {
      "External id": 134065,
      "cbid": 251, "correlation": 134065
    }
  },
  {
    "ph": "f", "id": 134065, "pid": 76337, "tid": -914061504, "ts": 1716454223420413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420418, "dur": 0,
    "args": {
      "External id": 134067,
      "cbid": 251, "correlation": 134067
    }
  },
  {
    "ph": "f", "id": 134067, "pid": 76337, "tid": -914061504, "ts": 1716454223420418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223483727, "dur": 496,
    "args": {
      "External id": 134068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134068,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134068, "pid": 5, "tid": 7, "ts": 1716454223483727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420420, "dur": 13,
    "args": {
      "External id": 134068,
      "cbid": 211, "correlation": 134068
    }
  },
  {
    "ph": "s", "id": 134068, "pid": 76337, "tid": -914061504, "ts": 1716454223420420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223484224, "dur": 66,
    "args": {
      "External id": 134076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134076,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134076, "pid": 5, "tid": 7, "ts": 1716454223484224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420487, "dur": 12,
    "args": {
      "External id": 134076,
      "cbid": 211, "correlation": 134076
    }
  },
  {
    "ph": "s", "id": 134076, "pid": 76337, "tid": -914061504, "ts": 1716454223420487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223484291, "dur": 68,
    "args": {
      "External id": 134084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134084,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134084, "pid": 5, "tid": 7, "ts": 1716454223484291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420517, "dur": 8,
    "args": {
      "External id": 134084,
      "cbid": 211, "correlation": 134084
    }
  },
  {
    "ph": "s", "id": 134084, "pid": 76337, "tid": -914061504, "ts": 1716454223420517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420597, "dur": 1,
    "args": {
      "External id": 134100,
      "cbid": 251, "correlation": 134100
    }
  },
  {
    "ph": "f", "id": 134100, "pid": 76337, "tid": -914061504, "ts": 1716454223420597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223484361, "dur": 1,
    "args": {
      "External id": 134102,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 134102,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 134102, "pid": 5, "tid": 7, "ts": 1716454223484361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420602, "dur": 9,
    "args": {
      "External id": 134102,
      "cbid": 51, "correlation": 134102
    }
  },
  {
    "ph": "s", "id": 134102, "pid": 76337, "tid": -914061504, "ts": 1716454223420602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223484365, "dur": 269,
    "args": {
      "External id": 134103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134103,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 134103, "pid": 5, "tid": 7, "ts": 1716454223484365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420612, "dur": 11,
    "args": {
      "External id": 134103,
      "cbid": 211, "correlation": 134103
    }
  },
  {
    "ph": "s", "id": 134103, "pid": 76337, "tid": -914061504, "ts": 1716454223420612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223484636, "dur": 14,
    "args": {
      "External id": 134111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134111,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134111, "pid": 5, "tid": 7, "ts": 1716454223484636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420656, "dur": 10,
    "args": {
      "External id": 134111,
      "cbid": 211, "correlation": 134111
    }
  },
  {
    "ph": "s", "id": 134111, "pid": 76337, "tid": -914061504, "ts": 1716454223420656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223484651, "dur": 37,
    "args": {
      "External id": 134122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134122,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134122, "pid": 5, "tid": 7, "ts": 1716454223484651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420724, "dur": 14,
    "args": {
      "External id": 134122,
      "cbid": 211, "correlation": 134122
    }
  },
  {
    "ph": "s", "id": 134122, "pid": 76337, "tid": -914061504, "ts": 1716454223420724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420790, "dur": 0,
    "args": {
      "External id": 134134,
      "cbid": 317, "correlation": 134134
    }
  },
  {
    "ph": "f", "id": 134134, "pid": 76337, "tid": -914061504, "ts": 1716454223420790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420791, "dur": 0,
    "args": {
      "External id": 134135,
      "cbid": 203, "correlation": 134135
    }
  },
  {
    "ph": "f", "id": 134135, "pid": 76337, "tid": -914061504, "ts": 1716454223420791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420791, "dur": 0,
    "args": {
      "External id": 134136,
      "cbid": 205, "correlation": 134136
    }
  },
  {
    "ph": "f", "id": 134136, "pid": 76337, "tid": -914061504, "ts": 1716454223420791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223484689, "dur": 13,
    "args": {
      "External id": 134140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134140,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134140, "pid": 5, "tid": 7, "ts": 1716454223484689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420806, "dur": 12,
    "args": {
      "External id": 134140,
      "cbid": 211, "correlation": 134140
    }
  },
  {
    "ph": "s", "id": 134140, "pid": 76337, "tid": -914061504, "ts": 1716454223420806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223484704, "dur": 4,
    "args": {
      "External id": 134142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 134142, "pid": 5, "tid": 7, "ts": 1716454223484704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420823, "dur": 6,
    "args": {
      "External id": 134142,
      "cbid": 211, "correlation": 134142
    }
  },
  {
    "ph": "s", "id": 134142, "pid": 76337, "tid": -914061504, "ts": 1716454223420823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420832, "dur": 0,
    "args": {
      "External id": 134143,
      "cbid": 51, "correlation": 134143
    }
  },
  {
    "ph": "s", "id": 134143, "pid": 76337, "tid": -914061504, "ts": 1716454223420832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223484709, "dur": 96,
    "args": {
      "External id": 134144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134144,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 134144, "pid": 5, "tid": 7, "ts": 1716454223484709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420833, "dur": 5,
    "args": {
      "External id": 134144,
      "cbid": 211, "correlation": 134144
    }
  },
  {
    "ph": "s", "id": 134144, "pid": 76337, "tid": -914061504, "ts": 1716454223420833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223484806, "dur": 16,
    "args": {
      "External id": 134149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134149,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134149, "pid": 5, "tid": 7, "ts": 1716454223484806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420860, "dur": 9,
    "args": {
      "External id": 134149,
      "cbid": 211, "correlation": 134149
    }
  },
  {
    "ph": "s", "id": 134149, "pid": 76337, "tid": -914061504, "ts": 1716454223420860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223484824, "dur": 12,
    "args": {
      "External id": 134157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134157,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134157, "pid": 5, "tid": 7, "ts": 1716454223484824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420892, "dur": 8,
    "args": {
      "External id": 134157,
      "cbid": 211, "correlation": 134157
    }
  },
  {
    "ph": "s", "id": 134157, "pid": 76337, "tid": -914061504, "ts": 1716454223420892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223484838, "dur": 56,
    "args": {
      "External id": 134168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134168,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134168, "pid": 5, "tid": 7, "ts": 1716454223484838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223420955, "dur": 12,
    "args": {
      "External id": 134168,
      "cbid": 211, "correlation": 134168
    }
  },
  {
    "ph": "s", "id": 134168, "pid": 76337, "tid": -914061504, "ts": 1716454223420955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421019, "dur": 0,
    "args": {
      "External id": 134178,
      "cbid": 317, "correlation": 134178
    }
  },
  {
    "ph": "f", "id": 134178, "pid": 76337, "tid": -914061504, "ts": 1716454223421019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421020, "dur": 0,
    "args": {
      "External id": 134179,
      "cbid": 203, "correlation": 134179
    }
  },
  {
    "ph": "f", "id": 134179, "pid": 76337, "tid": -914061504, "ts": 1716454223421020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421020, "dur": 0,
    "args": {
      "External id": 134180,
      "cbid": 205, "correlation": 134180
    }
  },
  {
    "ph": "f", "id": 134180, "pid": 76337, "tid": -914061504, "ts": 1716454223421020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223484895, "dur": 38,
    "args": {
      "External id": 134184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134184,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134184, "pid": 5, "tid": 7, "ts": 1716454223484895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421037, "dur": 12,
    "args": {
      "External id": 134184,
      "cbid": 211, "correlation": 134184
    }
  },
  {
    "ph": "s", "id": 134184, "pid": 76337, "tid": -914061504, "ts": 1716454223421037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223484934, "dur": 162,
    "args": {
      "External id": 134186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134186, "pid": 5, "tid": 7, "ts": 1716454223484934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421051, "dur": 5,
    "args": {
      "External id": 134186,
      "cbid": 211, "correlation": 134186
    }
  },
  {
    "ph": "s", "id": 134186, "pid": 76337, "tid": -914061504, "ts": 1716454223421051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223485097, "dur": 1984,
    "args": {
      "External id": 134188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134188,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134188, "pid": 5, "tid": 7, "ts": 1716454223485097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421063, "dur": 8,
    "args": {
      "External id": 134188,
      "cbid": 211, "correlation": 134188
    }
  },
  {
    "ph": "s", "id": 134188, "pid": 76337, "tid": -914061504, "ts": 1716454223421063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223487082, "dur": 38,
    "args": {
      "External id": 134190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134190, "pid": 5, "tid": 7, "ts": 1716454223487082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421074, "dur": 6,
    "args": {
      "External id": 134190,
      "cbid": 211, "correlation": 134190
    }
  },
  {
    "ph": "s", "id": 134190, "pid": 76337, "tid": -914061504, "ts": 1716454223421074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223487122, "dur": 60,
    "args": {
      "External id": 134196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134196,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134196, "pid": 5, "tid": 7, "ts": 1716454223487122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421104, "dur": 8,
    "args": {
      "External id": 134196,
      "cbid": 211, "correlation": 134196
    }
  },
  {
    "ph": "s", "id": 134196, "pid": 76337, "tid": -914061504, "ts": 1716454223421104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223487183, "dur": 84,
    "args": {
      "External id": 134205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134205,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134205, "pid": 5, "tid": 7, "ts": 1716454223487183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421193, "dur": 13,
    "args": {
      "External id": 134205,
      "cbid": 211, "correlation": 134205
    }
  },
  {
    "ph": "s", "id": 134205, "pid": 76337, "tid": -914061504, "ts": 1716454223421193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223487268, "dur": 72,
    "args": {
      "External id": 134225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134225,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 134225, "pid": 5, "tid": 7, "ts": 1716454223487268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421263, "dur": 11,
    "args": {
      "External id": 134225,
      "cbid": 211, "correlation": 134225
    }
  },
  {
    "ph": "s", "id": 134225, "pid": 76337, "tid": -914061504, "ts": 1716454223421263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223487342, "dur": 5,
    "args": {
      "External id": 134237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134237,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 134237, "pid": 5, "tid": 7, "ts": 1716454223487342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421285, "dur": 6,
    "args": {
      "External id": 134237,
      "cbid": 211, "correlation": 134237
    }
  },
  {
    "ph": "s", "id": 134237, "pid": 76337, "tid": -914061504, "ts": 1716454223421285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223487348, "dur": 80,
    "args": {
      "External id": 134240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134240,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134240, "pid": 5, "tid": 7, "ts": 1716454223487348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421304, "dur": 7,
    "args": {
      "External id": 134240,
      "cbid": 211, "correlation": 134240
    }
  },
  {
    "ph": "s", "id": 134240, "pid": 76337, "tid": -914061504, "ts": 1716454223421304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223487429, "dur": 53,
    "args": {
      "External id": 134249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134249,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134249, "pid": 5, "tid": 7, "ts": 1716454223487429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421345, "dur": 10,
    "args": {
      "External id": 134249,
      "cbid": 211, "correlation": 134249
    }
  },
  {
    "ph": "s", "id": 134249, "pid": 76337, "tid": -914061504, "ts": 1716454223421345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421396, "dur": 0,
    "args": {
      "External id": 134259,
      "cbid": 317, "correlation": 134259
    }
  },
  {
    "ph": "f", "id": 134259, "pid": 76337, "tid": -914061504, "ts": 1716454223421396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421397, "dur": 0,
    "args": {
      "External id": 134260,
      "cbid": 203, "correlation": 134260
    }
  },
  {
    "ph": "f", "id": 134260, "pid": 76337, "tid": -914061504, "ts": 1716454223421397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421397, "dur": 0,
    "args": {
      "External id": 134261,
      "cbid": 205, "correlation": 134261
    }
  },
  {
    "ph": "f", "id": 134261, "pid": 76337, "tid": -914061504, "ts": 1716454223421397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223487484, "dur": 57,
    "args": {
      "External id": 134265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134265,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134265, "pid": 5, "tid": 7, "ts": 1716454223487484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421413, "dur": 12,
    "args": {
      "External id": 134265,
      "cbid": 211, "correlation": 134265
    }
  },
  {
    "ph": "s", "id": 134265, "pid": 76337, "tid": -914061504, "ts": 1716454223421413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223487542, "dur": 122,
    "args": {
      "External id": 134267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134267,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134267, "pid": 5, "tid": 7, "ts": 1716454223487542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421428, "dur": 5,
    "args": {
      "External id": 134267,
      "cbid": 211, "correlation": 134267
    }
  },
  {
    "ph": "s", "id": 134267, "pid": 76337, "tid": -914061504, "ts": 1716454223421428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223487665, "dur": 1889,
    "args": {
      "External id": 134269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134269,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134269, "pid": 5, "tid": 7, "ts": 1716454223487665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421439, "dur": 6,
    "args": {
      "External id": 134269,
      "cbid": 211, "correlation": 134269
    }
  },
  {
    "ph": "s", "id": 134269, "pid": 76337, "tid": -914061504, "ts": 1716454223421439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223489555, "dur": 20,
    "args": {
      "External id": 134271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134271, "pid": 5, "tid": 7, "ts": 1716454223489555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421449, "dur": 5,
    "args": {
      "External id": 134271,
      "cbid": 211, "correlation": 134271
    }
  },
  {
    "ph": "s", "id": 134271, "pid": 76337, "tid": -914061504, "ts": 1716454223421449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223489577, "dur": 32,
    "args": {
      "External id": 134277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134277, "pid": 5, "tid": 7, "ts": 1716454223489577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421477, "dur": 8,
    "args": {
      "External id": 134277,
      "cbid": 211, "correlation": 134277
    }
  },
  {
    "ph": "s", "id": 134277, "pid": 76337, "tid": -914061504, "ts": 1716454223421477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223489611, "dur": 3,
    "args": {
      "External id": 134285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134285,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 134285, "pid": 5, "tid": 7, "ts": 1716454223489611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421520, "dur": 10,
    "args": {
      "External id": 134285,
      "cbid": 211, "correlation": 134285
    }
  },
  {
    "ph": "s", "id": 134285, "pid": 76337, "tid": -914061504, "ts": 1716454223421520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421587, "dur": 1,
    "args": {
      "External id": 134301,
      "cbid": 251, "correlation": 134301
    }
  },
  {
    "ph": "f", "id": 134301, "pid": 76337, "tid": -914061504, "ts": 1716454223421587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421592, "dur": 0,
    "args": {
      "External id": 134303,
      "cbid": 251, "correlation": 134303
    }
  },
  {
    "ph": "f", "id": 134303, "pid": 76337, "tid": -914061504, "ts": 1716454223421592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223489615, "dur": 12,
    "args": {
      "External id": 134304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134304,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 134304, "pid": 5, "tid": 7, "ts": 1716454223489615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421594, "dur": 11,
    "args": {
      "External id": 134304,
      "cbid": 211, "correlation": 134304
    }
  },
  {
    "ph": "s", "id": 134304, "pid": 76337, "tid": -914061504, "ts": 1716454223421594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223489629, "dur": 5,
    "args": {
      "External id": 134306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134306,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 134306, "pid": 5, "tid": 7, "ts": 1716454223489629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421607, "dur": 6,
    "args": {
      "External id": 134306,
      "cbid": 211, "correlation": 134306
    }
  },
  {
    "ph": "s", "id": 134306, "pid": 76337, "tid": -914061504, "ts": 1716454223421607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223489635, "dur": 28,
    "args": {
      "External id": 134316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134316, "pid": 5, "tid": 7, "ts": 1716454223489635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421665, "dur": 13,
    "args": {
      "External id": 134316,
      "cbid": 211, "correlation": 134316
    }
  },
  {
    "ph": "s", "id": 134316, "pid": 76337, "tid": -914061504, "ts": 1716454223421665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223489665, "dur": 30,
    "args": {
      "External id": 134336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134336,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 134336, "pid": 5, "tid": 7, "ts": 1716454223489665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421733, "dur": 10,
    "args": {
      "External id": 134336,
      "cbid": 211, "correlation": 134336
    }
  },
  {
    "ph": "s", "id": 134336, "pid": 76337, "tid": -914061504, "ts": 1716454223421733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223489696, "dur": 4,
    "args": {
      "External id": 134348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134348,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 134348, "pid": 5, "tid": 7, "ts": 1716454223489696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421753, "dur": 6,
    "args": {
      "External id": 134348,
      "cbid": 211, "correlation": 134348
    }
  },
  {
    "ph": "s", "id": 134348, "pid": 76337, "tid": -914061504, "ts": 1716454223421753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223489701, "dur": 30,
    "args": {
      "External id": 134351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134351,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134351, "pid": 5, "tid": 7, "ts": 1716454223489701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421771, "dur": 8,
    "args": {
      "External id": 134351,
      "cbid": 211, "correlation": 134351
    }
  },
  {
    "ph": "s", "id": 134351, "pid": 76337, "tid": -914061504, "ts": 1716454223421771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223489732, "dur": 20,
    "args": {
      "External id": 134360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134360,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134360, "pid": 5, "tid": 7, "ts": 1716454223489732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421812, "dur": 10,
    "args": {
      "External id": 134360,
      "cbid": 211, "correlation": 134360
    }
  },
  {
    "ph": "s", "id": 134360, "pid": 76337, "tid": -914061504, "ts": 1716454223421812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421875, "dur": 0,
    "args": {
      "External id": 134370,
      "cbid": 317, "correlation": 134370
    }
  },
  {
    "ph": "f", "id": 134370, "pid": 76337, "tid": -914061504, "ts": 1716454223421875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421876, "dur": 0,
    "args": {
      "External id": 134371,
      "cbid": 203, "correlation": 134371
    }
  },
  {
    "ph": "f", "id": 134371, "pid": 76337, "tid": -914061504, "ts": 1716454223421876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421877, "dur": 0,
    "args": {
      "External id": 134372,
      "cbid": 205, "correlation": 134372
    }
  },
  {
    "ph": "f", "id": 134372, "pid": 76337, "tid": -914061504, "ts": 1716454223421877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223489754, "dur": 21,
    "args": {
      "External id": 134376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134376,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134376, "pid": 5, "tid": 7, "ts": 1716454223489754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421892, "dur": 12,
    "args": {
      "External id": 134376,
      "cbid": 211, "correlation": 134376
    }
  },
  {
    "ph": "s", "id": 134376, "pid": 76337, "tid": -914061504, "ts": 1716454223421892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223489777, "dur": 43,
    "args": {
      "External id": 134378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134378, "pid": 5, "tid": 7, "ts": 1716454223489777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421907, "dur": 5,
    "args": {
      "External id": 134378,
      "cbid": 211, "correlation": 134378
    }
  },
  {
    "ph": "s", "id": 134378, "pid": 76337, "tid": -914061504, "ts": 1716454223421907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223489821, "dur": 645,
    "args": {
      "External id": 134380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134380,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134380, "pid": 5, "tid": 7, "ts": 1716454223489821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421919, "dur": 6,
    "args": {
      "External id": 134380,
      "cbid": 211, "correlation": 134380
    }
  },
  {
    "ph": "s", "id": 134380, "pid": 76337, "tid": -914061504, "ts": 1716454223421919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223490468, "dur": 22,
    "args": {
      "External id": 134382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134382,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134382, "pid": 5, "tid": 7, "ts": 1716454223490468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421929, "dur": 6,
    "args": {
      "External id": 134382,
      "cbid": 211, "correlation": 134382
    }
  },
  {
    "ph": "s", "id": 134382, "pid": 76337, "tid": -914061504, "ts": 1716454223421929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223490492, "dur": 32,
    "args": {
      "External id": 134388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134388,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134388, "pid": 5, "tid": 7, "ts": 1716454223490492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223421957, "dur": 8,
    "args": {
      "External id": 134388,
      "cbid": 211, "correlation": 134388
    }
  },
  {
    "ph": "s", "id": 134388, "pid": 76337, "tid": -914061504, "ts": 1716454223421957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422025, "dur": 0,
    "args": {
      "External id": 134398,
      "cbid": 317, "correlation": 134398
    }
  },
  {
    "ph": "f", "id": 134398, "pid": 76337, "tid": -914061504, "ts": 1716454223422025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422026, "dur": 0,
    "args": {
      "External id": 134399,
      "cbid": 203, "correlation": 134399
    }
  },
  {
    "ph": "f", "id": 134399, "pid": 76337, "tid": -914061504, "ts": 1716454223422026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422026, "dur": 0,
    "args": {
      "External id": 134400,
      "cbid": 205, "correlation": 134400
    }
  },
  {
    "ph": "f", "id": 134400, "pid": 76337, "tid": -914061504, "ts": 1716454223422026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223490525, "dur": 56,
    "args": {
      "External id": 134404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134404, "pid": 5, "tid": 7, "ts": 1716454223490525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422040, "dur": 12,
    "args": {
      "External id": 134404,
      "cbid": 211, "correlation": 134404
    }
  },
  {
    "ph": "s", "id": 134404, "pid": 76337, "tid": -914061504, "ts": 1716454223422040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223490583, "dur": 268,
    "args": {
      "External id": 134406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134406,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134406, "pid": 5, "tid": 7, "ts": 1716454223490583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422058, "dur": 7,
    "args": {
      "External id": 134406,
      "cbid": 211, "correlation": 134406
    }
  },
  {
    "ph": "s", "id": 134406, "pid": 76337, "tid": -914061504, "ts": 1716454223422058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223490852, "dur": 21,
    "args": {
      "External id": 134408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134408,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134408, "pid": 5, "tid": 7, "ts": 1716454223490852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422069, "dur": 5,
    "args": {
      "External id": 134408,
      "cbid": 211, "correlation": 134408
    }
  },
  {
    "ph": "s", "id": 134408, "pid": 76337, "tid": -914061504, "ts": 1716454223422069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223490874, "dur": 33,
    "args": {
      "External id": 134414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134414,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134414, "pid": 5, "tid": 7, "ts": 1716454223490874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422096, "dur": 8,
    "args": {
      "External id": 134414,
      "cbid": 211, "correlation": 134414
    }
  },
  {
    "ph": "s", "id": 134414, "pid": 76337, "tid": -914061504, "ts": 1716454223422096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223490909, "dur": 28,
    "args": {
      "External id": 134422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134422,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134422, "pid": 5, "tid": 7, "ts": 1716454223490909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422124, "dur": 8,
    "args": {
      "External id": 134422,
      "cbid": 211, "correlation": 134422
    }
  },
  {
    "ph": "s", "id": 134422, "pid": 76337, "tid": -914061504, "ts": 1716454223422124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223490937, "dur": 20,
    "args": {
      "External id": 134430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134430,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134430, "pid": 5, "tid": 7, "ts": 1716454223490937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422152, "dur": 8,
    "args": {
      "External id": 134430,
      "cbid": 211, "correlation": 134430
    }
  },
  {
    "ph": "s", "id": 134430, "pid": 76337, "tid": -914061504, "ts": 1716454223422152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223490959, "dur": 29,
    "args": {
      "External id": 134450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134450,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 134450, "pid": 5, "tid": 7, "ts": 1716454223490959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422234, "dur": 12,
    "args": {
      "External id": 134450,
      "cbid": 211, "correlation": 134450
    }
  },
  {
    "ph": "s", "id": 134450, "pid": 76337, "tid": -914061504, "ts": 1716454223422234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223490990, "dur": 4,
    "args": {
      "External id": 134462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134462,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 134462, "pid": 5, "tid": 7, "ts": 1716454223490990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422255, "dur": 6,
    "args": {
      "External id": 134462,
      "cbid": 211, "correlation": 134462
    }
  },
  {
    "ph": "s", "id": 134462, "pid": 76337, "tid": -914061504, "ts": 1716454223422255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223490995, "dur": 32,
    "args": {
      "External id": 134465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134465,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134465, "pid": 5, "tid": 7, "ts": 1716454223490995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422273, "dur": 7,
    "args": {
      "External id": 134465,
      "cbid": 211, "correlation": 134465
    }
  },
  {
    "ph": "s", "id": 134465, "pid": 76337, "tid": -914061504, "ts": 1716454223422273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422330, "dur": 0,
    "args": {
      "External id": 134476,
      "cbid": 317, "correlation": 134476
    }
  },
  {
    "ph": "f", "id": 134476, "pid": 76337, "tid": -914061504, "ts": 1716454223422330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422331, "dur": 0,
    "args": {
      "External id": 134477,
      "cbid": 203, "correlation": 134477
    }
  },
  {
    "ph": "f", "id": 134477, "pid": 76337, "tid": -914061504, "ts": 1716454223422331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422332, "dur": 0,
    "args": {
      "External id": 134478,
      "cbid": 205, "correlation": 134478
    }
  },
  {
    "ph": "f", "id": 134478, "pid": 76337, "tid": -914061504, "ts": 1716454223422332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223491029, "dur": 21,
    "args": {
      "External id": 134482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134482,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134482, "pid": 5, "tid": 7, "ts": 1716454223491029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422346, "dur": 12,
    "args": {
      "External id": 134482,
      "cbid": 211, "correlation": 134482
    }
  },
  {
    "ph": "s", "id": 134482, "pid": 76337, "tid": -914061504, "ts": 1716454223422346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223491051, "dur": 105,
    "args": {
      "External id": 134484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134484,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134484, "pid": 5, "tid": 7, "ts": 1716454223491051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422364, "dur": 6,
    "args": {
      "External id": 134484,
      "cbid": 211, "correlation": 134484
    }
  },
  {
    "ph": "s", "id": 134484, "pid": 76337, "tid": -914061504, "ts": 1716454223422364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223491157, "dur": 22,
    "args": {
      "External id": 134486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134486,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134486, "pid": 5, "tid": 7, "ts": 1716454223491157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422374, "dur": 5,
    "args": {
      "External id": 134486,
      "cbid": 211, "correlation": 134486
    }
  },
  {
    "ph": "s", "id": 134486, "pid": 76337, "tid": -914061504, "ts": 1716454223422374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223491180, "dur": 32,
    "args": {
      "External id": 134492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134492,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134492, "pid": 5, "tid": 7, "ts": 1716454223491180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422401, "dur": 8,
    "args": {
      "External id": 134492,
      "cbid": 211, "correlation": 134492
    }
  },
  {
    "ph": "s", "id": 134492, "pid": 76337, "tid": -914061504, "ts": 1716454223422401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223491214, "dur": 178,
    "args": {
      "External id": 134501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134501,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134501, "pid": 5, "tid": 7, "ts": 1716454223491214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422483, "dur": 13,
    "args": {
      "External id": 134501,
      "cbid": 211, "correlation": 134501
    }
  },
  {
    "ph": "s", "id": 134501, "pid": 76337, "tid": -914061504, "ts": 1716454223422483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223491394, "dur": 64,
    "args": {
      "External id": 134523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134523,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134523, "pid": 5, "tid": 7, "ts": 1716454223491394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422541, "dur": 10,
    "args": {
      "External id": 134523,
      "cbid": 211, "correlation": 134523
    }
  },
  {
    "ph": "s", "id": 134523, "pid": 76337, "tid": -914061504, "ts": 1716454223422541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422630, "dur": 2,
    "args": {
      "External id": 134534,
      "cbid": 251, "correlation": 134534
    }
  },
  {
    "ph": "f", "id": 134534, "pid": 76337, "tid": -914061504, "ts": 1716454223422630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223491459, "dur": 155,
    "args": {
      "External id": 134535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134535,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134535, "pid": 5, "tid": 7, "ts": 1716454223491459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422636, "dur": 13,
    "args": {
      "External id": 134535,
      "cbid": 211, "correlation": 134535
    }
  },
  {
    "ph": "s", "id": 134535, "pid": 76337, "tid": -914061504, "ts": 1716454223422636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422705, "dur": 1,
    "args": {
      "External id": 134546,
      "cbid": 251, "correlation": 134546
    }
  },
  {
    "ph": "f", "id": 134546, "pid": 76337, "tid": -914061504, "ts": 1716454223422705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223491615, "dur": 148,
    "args": {
      "External id": 134547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134547,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134547, "pid": 5, "tid": 7, "ts": 1716454223491615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422709, "dur": 11,
    "args": {
      "External id": 134547,
      "cbid": 211, "correlation": 134547
    }
  },
  {
    "ph": "s", "id": 134547, "pid": 76337, "tid": -914061504, "ts": 1716454223422709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422773, "dur": 1,
    "args": {
      "External id": 134558,
      "cbid": 251, "correlation": 134558
    }
  },
  {
    "ph": "f", "id": 134558, "pid": 76337, "tid": -914061504, "ts": 1716454223422773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223491765, "dur": 143,
    "args": {
      "External id": 134559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134559,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134559, "pid": 5, "tid": 7, "ts": 1716454223491765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422777, "dur": 12,
    "args": {
      "External id": 134559,
      "cbid": 211, "correlation": 134559
    }
  },
  {
    "ph": "s", "id": 134559, "pid": 76337, "tid": -914061504, "ts": 1716454223422777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223491909, "dur": 1919,
    "args": {
      "External id": 134580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134580,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 134580, "pid": 5, "tid": 7, "ts": 1716454223491909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422859, "dur": 13,
    "args": {
      "External id": 134580,
      "cbid": 211, "correlation": 134580
    }
  },
  {
    "ph": "s", "id": 134580, "pid": 76337, "tid": -914061504, "ts": 1716454223422859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422959, "dur": 2,
    "args": {
      "External id": 134598,
      "cbid": 251, "correlation": 134598
    }
  },
  {
    "ph": "f", "id": 134598, "pid": 76337, "tid": -914061504, "ts": 1716454223422959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223493829, "dur": 147,
    "args": {
      "External id": 134600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134600,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 134600, "pid": 5, "tid": 7, "ts": 1716454223493829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223422965, "dur": 28,
    "args": {
      "External id": 134600,
      "cbid": 211, "correlation": 134600
    }
  },
  {
    "ph": "s", "id": 134600, "pid": 76337, "tid": -914061504, "ts": 1716454223422965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223493977, "dur": 36,
    "args": {
      "External id": 134608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134608,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134608, "pid": 5, "tid": 7, "ts": 1716454223493977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423051, "dur": 12,
    "args": {
      "External id": 134608,
      "cbid": 211, "correlation": 134608
    }
  },
  {
    "ph": "s", "id": 134608, "pid": 76337, "tid": -914061504, "ts": 1716454223423051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223494014, "dur": 51,
    "args": {
      "External id": 134616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134616,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134616, "pid": 5, "tid": 7, "ts": 1716454223494014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423091, "dur": 8,
    "args": {
      "External id": 134616,
      "cbid": 211, "correlation": 134616
    }
  },
  {
    "ph": "s", "id": 134616, "pid": 76337, "tid": -914061504, "ts": 1716454223423091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223494066, "dur": 30,
    "args": {
      "External id": 134627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134627,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134627, "pid": 5, "tid": 7, "ts": 1716454223494066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423162, "dur": 13,
    "args": {
      "External id": 134627,
      "cbid": 211, "correlation": 134627
    }
  },
  {
    "ph": "s", "id": 134627, "pid": 76337, "tid": -914061504, "ts": 1716454223423162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223494097, "dur": 33,
    "args": {
      "External id": 134649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134649,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134649, "pid": 5, "tid": 7, "ts": 1716454223494097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423194, "dur": 8,
    "args": {
      "External id": 134649,
      "cbid": 211, "correlation": 134649
    }
  },
  {
    "ph": "s", "id": 134649, "pid": 76337, "tid": -914061504, "ts": 1716454223423194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423279, "dur": 1,
    "args": {
      "External id": 134660,
      "cbid": 251, "correlation": 134660
    }
  },
  {
    "ph": "f", "id": 134660, "pid": 76337, "tid": -914061504, "ts": 1716454223423279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223494132, "dur": 75,
    "args": {
      "External id": 134661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134661,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134661, "pid": 5, "tid": 7, "ts": 1716454223494132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423284, "dur": 13,
    "args": {
      "External id": 134661,
      "cbid": 211, "correlation": 134661
    }
  },
  {
    "ph": "s", "id": 134661, "pid": 76337, "tid": -914061504, "ts": 1716454223423284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423355, "dur": 1,
    "args": {
      "External id": 134672,
      "cbid": 251, "correlation": 134672
    }
  },
  {
    "ph": "f", "id": 134672, "pid": 76337, "tid": -914061504, "ts": 1716454223423355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423359, "dur": 0,
    "args": {
      "External id": 134673,
      "cbid": 251, "correlation": 134673
    }
  },
  {
    "ph": "f", "id": 134673, "pid": 76337, "tid": -914061504, "ts": 1716454223423359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223494208, "dur": 12,
    "args": {
      "External id": 134674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134674,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 134674, "pid": 5, "tid": 7, "ts": 1716454223494208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423360, "dur": 12,
    "args": {
      "External id": 134674,
      "cbid": 211, "correlation": 134674
    }
  },
  {
    "ph": "s", "id": 134674, "pid": 76337, "tid": -914061504, "ts": 1716454223423360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223494222, "dur": 5,
    "args": {
      "External id": 134676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134676,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 134676, "pid": 5, "tid": 7, "ts": 1716454223494222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423374, "dur": 6,
    "args": {
      "External id": 134676,
      "cbid": 211, "correlation": 134676
    }
  },
  {
    "ph": "s", "id": 134676, "pid": 76337, "tid": -914061504, "ts": 1716454223423374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423431, "dur": 1,
    "args": {
      "External id": 134687,
      "cbid": 251, "correlation": 134687
    }
  },
  {
    "ph": "f", "id": 134687, "pid": 76337, "tid": -914061504, "ts": 1716454223423431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423434, "dur": 0,
    "args": {
      "External id": 134688,
      "cbid": 251, "correlation": 134688
    }
  },
  {
    "ph": "f", "id": 134688, "pid": 76337, "tid": -914061504, "ts": 1716454223423434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223494228, "dur": 7,
    "args": {
      "External id": 134689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134689,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 134689, "pid": 5, "tid": 7, "ts": 1716454223494228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423436, "dur": 12,
    "args": {
      "External id": 134689,
      "cbid": 211, "correlation": 134689
    }
  },
  {
    "ph": "s", "id": 134689, "pid": 76337, "tid": -914061504, "ts": 1716454223423436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223494237, "dur": 3,
    "args": {
      "External id": 134691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134691,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 134691, "pid": 5, "tid": 7, "ts": 1716454223494237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423450, "dur": 5,
    "args": {
      "External id": 134691,
      "cbid": 211, "correlation": 134691
    }
  },
  {
    "ph": "s", "id": 134691, "pid": 76337, "tid": -914061504, "ts": 1716454223423450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223494241, "dur": 90,
    "args": {
      "External id": 134712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134712,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 134712, "pid": 5, "tid": 7, "ts": 1716454223494241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423522, "dur": 13,
    "args": {
      "External id": 134712,
      "cbid": 211, "correlation": 134712
    }
  },
  {
    "ph": "s", "id": 134712, "pid": 76337, "tid": -914061504, "ts": 1716454223423522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423619, "dur": 1,
    "args": {
      "External id": 134730,
      "cbid": 251, "correlation": 134730
    }
  },
  {
    "ph": "f", "id": 134730, "pid": 76337, "tid": -914061504, "ts": 1716454223423619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223494333, "dur": 96,
    "args": {
      "External id": 134732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134732,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134732, "pid": 5, "tid": 7, "ts": 1716454223494333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423624, "dur": 13,
    "args": {
      "External id": 134732,
      "cbid": 211, "correlation": 134732
    }
  },
  {
    "ph": "s", "id": 134732, "pid": 76337, "tid": -914061504, "ts": 1716454223423624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223494430, "dur": 19,
    "args": {
      "External id": 134740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134740,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134740, "pid": 5, "tid": 7, "ts": 1716454223494430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423694, "dur": 13,
    "args": {
      "External id": 134740,
      "cbid": 211, "correlation": 134740
    }
  },
  {
    "ph": "s", "id": 134740, "pid": 76337, "tid": -914061504, "ts": 1716454223423694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223494450, "dur": 39,
    "args": {
      "External id": 134748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134748,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134748, "pid": 5, "tid": 7, "ts": 1716454223494450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423736, "dur": 10,
    "args": {
      "External id": 134748,
      "cbid": 211, "correlation": 134748
    }
  },
  {
    "ph": "s", "id": 134748, "pid": 76337, "tid": -914061504, "ts": 1716454223423736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223494489, "dur": 34,
    "args": {
      "External id": 134770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134770,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134770, "pid": 5, "tid": 7, "ts": 1716454223494489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423787, "dur": 11,
    "args": {
      "External id": 134770,
      "cbid": 211, "correlation": 134770
    }
  },
  {
    "ph": "s", "id": 134770, "pid": 76337, "tid": -914061504, "ts": 1716454223423787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423875, "dur": 1,
    "args": {
      "External id": 134786,
      "cbid": 251, "correlation": 134786
    }
  },
  {
    "ph": "f", "id": 134786, "pid": 76337, "tid": -914061504, "ts": 1716454223423875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423880, "dur": 0,
    "args": {
      "External id": 134788,
      "cbid": 251, "correlation": 134788
    }
  },
  {
    "ph": "f", "id": 134788, "pid": 76337, "tid": -914061504, "ts": 1716454223423880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223494525, "dur": 536,
    "args": {
      "External id": 134789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134789,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 134789, "pid": 5, "tid": 7, "ts": 1716454223494525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423885, "dur": 14,
    "args": {
      "External id": 134789,
      "cbid": 211, "correlation": 134789
    }
  },
  {
    "ph": "s", "id": 134789, "pid": 76337, "tid": -914061504, "ts": 1716454223423885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223495063, "dur": 125,
    "args": {
      "External id": 134797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134797,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134797, "pid": 5, "tid": 7, "ts": 1716454223495063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423951, "dur": 12,
    "args": {
      "External id": 134797,
      "cbid": 211, "correlation": 134797
    }
  },
  {
    "ph": "s", "id": 134797, "pid": 76337, "tid": -914061504, "ts": 1716454223423951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223495189, "dur": 127,
    "args": {
      "External id": 134805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134805, "pid": 5, "tid": 7, "ts": 1716454223495189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223423990, "dur": 10,
    "args": {
      "External id": 134805,
      "cbid": 211, "correlation": 134805
    }
  },
  {
    "ph": "s", "id": 134805, "pid": 76337, "tid": -914061504, "ts": 1716454223423990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424070, "dur": 1,
    "args": {
      "External id": 134821,
      "cbid": 251, "correlation": 134821
    }
  },
  {
    "ph": "f", "id": 134821, "pid": 76337, "tid": -914061504, "ts": 1716454223424070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223495317, "dur": 310,
    "args": {
      "External id": 134823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134823,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134823, "pid": 5, "tid": 7, "ts": 1716454223495317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424076, "dur": 12,
    "args": {
      "External id": 134823,
      "cbid": 211, "correlation": 134823
    }
  },
  {
    "ph": "s", "id": 134823, "pid": 76337, "tid": -914061504, "ts": 1716454223424076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223495629, "dur": 27,
    "args": {
      "External id": 134831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134831,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134831, "pid": 5, "tid": 7, "ts": 1716454223495629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424118, "dur": 10,
    "args": {
      "External id": 134831,
      "cbid": 211, "correlation": 134831
    }
  },
  {
    "ph": "s", "id": 134831, "pid": 76337, "tid": -914061504, "ts": 1716454223424118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223495658, "dur": 80,
    "args": {
      "External id": 134842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134842,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134842, "pid": 5, "tid": 7, "ts": 1716454223495658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424185, "dur": 13,
    "args": {
      "External id": 134842,
      "cbid": 211, "correlation": 134842
    }
  },
  {
    "ph": "s", "id": 134842, "pid": 76337, "tid": -914061504, "ts": 1716454223424185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424251, "dur": 0,
    "args": {
      "External id": 134854,
      "cbid": 317, "correlation": 134854
    }
  },
  {
    "ph": "f", "id": 134854, "pid": 76337, "tid": -914061504, "ts": 1716454223424251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424251, "dur": 0,
    "args": {
      "External id": 134855,
      "cbid": 203, "correlation": 134855
    }
  },
  {
    "ph": "f", "id": 134855, "pid": 76337, "tid": -914061504, "ts": 1716454223424251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424252, "dur": 0,
    "args": {
      "External id": 134856,
      "cbid": 205, "correlation": 134856
    }
  },
  {
    "ph": "f", "id": 134856, "pid": 76337, "tid": -914061504, "ts": 1716454223424252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223495739, "dur": 22,
    "args": {
      "External id": 134860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134860,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134860, "pid": 5, "tid": 7, "ts": 1716454223495739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424268, "dur": 12,
    "args": {
      "External id": 134860,
      "cbid": 211, "correlation": 134860
    }
  },
  {
    "ph": "s", "id": 134860, "pid": 76337, "tid": -914061504, "ts": 1716454223424268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223495763, "dur": 119,
    "args": {
      "External id": 134862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134862,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134862, "pid": 5, "tid": 7, "ts": 1716454223495763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424286, "dur": 7,
    "args": {
      "External id": 134862,
      "cbid": 211, "correlation": 134862
    }
  },
  {
    "ph": "s", "id": 134862, "pid": 76337, "tid": -914061504, "ts": 1716454223424286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223495883, "dur": 23,
    "args": {
      "External id": 134864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134864,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134864, "pid": 5, "tid": 7, "ts": 1716454223495883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424298, "dur": 5,
    "args": {
      "External id": 134864,
      "cbid": 211, "correlation": 134864
    }
  },
  {
    "ph": "s", "id": 134864, "pid": 76337, "tid": -914061504, "ts": 1716454223424298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223495908, "dur": 33,
    "args": {
      "External id": 134870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134870, "pid": 5, "tid": 7, "ts": 1716454223495908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424325, "dur": 8,
    "args": {
      "External id": 134870,
      "cbid": 211, "correlation": 134870
    }
  },
  {
    "ph": "s", "id": 134870, "pid": 76337, "tid": -914061504, "ts": 1716454223424325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223495942, "dur": 27,
    "args": {
      "External id": 134878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134878,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134878, "pid": 5, "tid": 7, "ts": 1716454223495942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424357, "dur": 8,
    "args": {
      "External id": 134878,
      "cbid": 211, "correlation": 134878
    }
  },
  {
    "ph": "s", "id": 134878, "pid": 76337, "tid": -914061504, "ts": 1716454223424357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223495969, "dur": 53,
    "args": {
      "External id": 134887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134887,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134887, "pid": 5, "tid": 7, "ts": 1716454223495969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424397, "dur": 10,
    "args": {
      "External id": 134887,
      "cbid": 211, "correlation": 134887
    }
  },
  {
    "ph": "s", "id": 134887, "pid": 76337, "tid": -914061504, "ts": 1716454223424397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223496024, "dur": 52,
    "args": {
      "External id": 134907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134907,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 134907, "pid": 5, "tid": 7, "ts": 1716454223496024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424468, "dur": 11,
    "args": {
      "External id": 134907,
      "cbid": 211, "correlation": 134907
    }
  },
  {
    "ph": "s", "id": 134907, "pid": 76337, "tid": -914061504, "ts": 1716454223424468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223496077, "dur": 4,
    "args": {
      "External id": 134919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134919,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 134919, "pid": 5, "tid": 7, "ts": 1716454223496077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424489, "dur": 7,
    "args": {
      "External id": 134919,
      "cbid": 211, "correlation": 134919
    }
  },
  {
    "ph": "s", "id": 134919, "pid": 76337, "tid": -914061504, "ts": 1716454223424489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223496083, "dur": 55,
    "args": {
      "External id": 134922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134922, "pid": 5, "tid": 7, "ts": 1716454223496083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424507, "dur": 6,
    "args": {
      "External id": 134922,
      "cbid": 211, "correlation": 134922
    }
  },
  {
    "ph": "s", "id": 134922, "pid": 76337, "tid": -914061504, "ts": 1716454223424507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223496139, "dur": 37,
    "args": {
      "External id": 134931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134931,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134931, "pid": 5, "tid": 7, "ts": 1716454223496139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424547, "dur": 10,
    "args": {
      "External id": 134931,
      "cbid": 211, "correlation": 134931
    }
  },
  {
    "ph": "s", "id": 134931, "pid": 76337, "tid": -914061504, "ts": 1716454223424547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424599, "dur": 0,
    "args": {
      "External id": 134941,
      "cbid": 317, "correlation": 134941
    }
  },
  {
    "ph": "f", "id": 134941, "pid": 76337, "tid": -914061504, "ts": 1716454223424599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424600, "dur": 0,
    "args": {
      "External id": 134942,
      "cbid": 203, "correlation": 134942
    }
  },
  {
    "ph": "f", "id": 134942, "pid": 76337, "tid": -914061504, "ts": 1716454223424600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424601, "dur": 0,
    "args": {
      "External id": 134943,
      "cbid": 205, "correlation": 134943
    }
  },
  {
    "ph": "f", "id": 134943, "pid": 76337, "tid": -914061504, "ts": 1716454223424601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223496178, "dur": 39,
    "args": {
      "External id": 134947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134947, "pid": 5, "tid": 7, "ts": 1716454223496178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424615, "dur": 12,
    "args": {
      "External id": 134947,
      "cbid": 211, "correlation": 134947
    }
  },
  {
    "ph": "s", "id": 134947, "pid": 76337, "tid": -914061504, "ts": 1716454223424615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223496218, "dur": 83,
    "args": {
      "External id": 134949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134949, "pid": 5, "tid": 7, "ts": 1716454223496218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424629, "dur": 5,
    "args": {
      "External id": 134949,
      "cbid": 211, "correlation": 134949
    }
  },
  {
    "ph": "s", "id": 134949, "pid": 76337, "tid": -914061504, "ts": 1716454223424629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223496302, "dur": 1266,
    "args": {
      "External id": 134951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134951,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 134951, "pid": 5, "tid": 7, "ts": 1716454223496302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424640, "dur": 7,
    "args": {
      "External id": 134951,
      "cbid": 211, "correlation": 134951
    }
  },
  {
    "ph": "s", "id": 134951, "pid": 76337, "tid": -914061504, "ts": 1716454223424640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223497569, "dur": 21,
    "args": {
      "External id": 134953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134953,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134953, "pid": 5, "tid": 7, "ts": 1716454223497569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424651, "dur": 5,
    "args": {
      "External id": 134953,
      "cbid": 211, "correlation": 134953
    }
  },
  {
    "ph": "s", "id": 134953, "pid": 76337, "tid": -914061504, "ts": 1716454223424651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223497592, "dur": 33,
    "args": {
      "External id": 134959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134959, "pid": 5, "tid": 7, "ts": 1716454223497592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424678, "dur": 8,
    "args": {
      "External id": 134959,
      "cbid": 211, "correlation": 134959
    }
  },
  {
    "ph": "s", "id": 134959, "pid": 76337, "tid": -914061504, "ts": 1716454223424678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223497626, "dur": 3,
    "args": {
      "External id": 134967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134967,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 134967, "pid": 5, "tid": 7, "ts": 1716454223497626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424721, "dur": 9,
    "args": {
      "External id": 134967,
      "cbid": 211, "correlation": 134967
    }
  },
  {
    "ph": "s", "id": 134967, "pid": 76337, "tid": -914061504, "ts": 1716454223424721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424785, "dur": 1,
    "args": {
      "External id": 134983,
      "cbid": 251, "correlation": 134983
    }
  },
  {
    "ph": "f", "id": 134983, "pid": 76337, "tid": -914061504, "ts": 1716454223424785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424790, "dur": 0,
    "args": {
      "External id": 134985,
      "cbid": 251, "correlation": 134985
    }
  },
  {
    "ph": "f", "id": 134985, "pid": 76337, "tid": -914061504, "ts": 1716454223424790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223497630, "dur": 12,
    "args": {
      "External id": 134986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134986,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 134986, "pid": 5, "tid": 7, "ts": 1716454223497630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424793, "dur": 11,
    "args": {
      "External id": 134986,
      "cbid": 211, "correlation": 134986
    }
  },
  {
    "ph": "s", "id": 134986, "pid": 76337, "tid": -914061504, "ts": 1716454223424793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223497643, "dur": 5,
    "args": {
      "External id": 134988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134988,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 134988, "pid": 5, "tid": 7, "ts": 1716454223497643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424806, "dur": 5,
    "args": {
      "External id": 134988,
      "cbid": 211, "correlation": 134988
    }
  },
  {
    "ph": "s", "id": 134988, "pid": 76337, "tid": -914061504, "ts": 1716454223424806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223497650, "dur": 28,
    "args": {
      "External id": 134998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 134998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 134998, "pid": 5, "tid": 7, "ts": 1716454223497650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424862, "dur": 12,
    "args": {
      "External id": 134998,
      "cbid": 211, "correlation": 134998
    }
  },
  {
    "ph": "s", "id": 134998, "pid": 76337, "tid": -914061504, "ts": 1716454223424862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223497679, "dur": 31,
    "args": {
      "External id": 135018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135018,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 135018, "pid": 5, "tid": 7, "ts": 1716454223497679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424928, "dur": 11,
    "args": {
      "External id": 135018,
      "cbid": 211, "correlation": 135018
    }
  },
  {
    "ph": "s", "id": 135018, "pid": 76337, "tid": -914061504, "ts": 1716454223424928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223497712, "dur": 4,
    "args": {
      "External id": 135030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135030,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 135030, "pid": 5, "tid": 7, "ts": 1716454223497712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424949, "dur": 6,
    "args": {
      "External id": 135030,
      "cbid": 211, "correlation": 135030
    }
  },
  {
    "ph": "s", "id": 135030, "pid": 76337, "tid": -914061504, "ts": 1716454223424949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223497717, "dur": 29,
    "args": {
      "External id": 135033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135033, "pid": 5, "tid": 7, "ts": 1716454223497717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223424968, "dur": 15,
    "args": {
      "External id": 135033,
      "cbid": 211, "correlation": 135033
    }
  },
  {
    "ph": "s", "id": 135033, "pid": 76337, "tid": -914061504, "ts": 1716454223424968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223497747, "dur": 20,
    "args": {
      "External id": 135042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135042,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135042, "pid": 5, "tid": 7, "ts": 1716454223497747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425019, "dur": 11,
    "args": {
      "External id": 135042,
      "cbid": 211, "correlation": 135042
    }
  },
  {
    "ph": "s", "id": 135042, "pid": 76337, "tid": -914061504, "ts": 1716454223425019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425083, "dur": 0,
    "args": {
      "External id": 135052,
      "cbid": 317, "correlation": 135052
    }
  },
  {
    "ph": "f", "id": 135052, "pid": 76337, "tid": -914061504, "ts": 1716454223425083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425083, "dur": 0,
    "args": {
      "External id": 135053,
      "cbid": 203, "correlation": 135053
    }
  },
  {
    "ph": "f", "id": 135053, "pid": 76337, "tid": -914061504, "ts": 1716454223425083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425084, "dur": 0,
    "args": {
      "External id": 135054,
      "cbid": 205, "correlation": 135054
    }
  },
  {
    "ph": "f", "id": 135054, "pid": 76337, "tid": -914061504, "ts": 1716454223425084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223497769, "dur": 21,
    "args": {
      "External id": 135058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135058,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135058, "pid": 5, "tid": 7, "ts": 1716454223497769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425098, "dur": 12,
    "args": {
      "External id": 135058,
      "cbid": 211, "correlation": 135058
    }
  },
  {
    "ph": "s", "id": 135058, "pid": 76337, "tid": -914061504, "ts": 1716454223425098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223497791, "dur": 43,
    "args": {
      "External id": 135060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135060,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135060, "pid": 5, "tid": 7, "ts": 1716454223497791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425113, "dur": 5,
    "args": {
      "External id": 135060,
      "cbid": 211, "correlation": 135060
    }
  },
  {
    "ph": "s", "id": 135060, "pid": 76337, "tid": -914061504, "ts": 1716454223425113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223497836, "dur": 639,
    "args": {
      "External id": 135062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135062,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135062, "pid": 5, "tid": 7, "ts": 1716454223497836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425124, "dur": 6,
    "args": {
      "External id": 135062,
      "cbid": 211, "correlation": 135062
    }
  },
  {
    "ph": "s", "id": 135062, "pid": 76337, "tid": -914061504, "ts": 1716454223425124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223498476, "dur": 21,
    "args": {
      "External id": 135064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135064,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135064, "pid": 5, "tid": 7, "ts": 1716454223498476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425133, "dur": 5,
    "args": {
      "External id": 135064,
      "cbid": 211, "correlation": 135064
    }
  },
  {
    "ph": "s", "id": 135064, "pid": 76337, "tid": -914061504, "ts": 1716454223425133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223498498, "dur": 32,
    "args": {
      "External id": 135070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135070,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135070, "pid": 5, "tid": 7, "ts": 1716454223498498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425162, "dur": 8,
    "args": {
      "External id": 135070,
      "cbid": 211, "correlation": 135070
    }
  },
  {
    "ph": "s", "id": 135070, "pid": 76337, "tid": -914061504, "ts": 1716454223425162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425220, "dur": 0,
    "args": {
      "External id": 135080,
      "cbid": 317, "correlation": 135080
    }
  },
  {
    "ph": "f", "id": 135080, "pid": 76337, "tid": -914061504, "ts": 1716454223425220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425220, "dur": 0,
    "args": {
      "External id": 135081,
      "cbid": 203, "correlation": 135081
    }
  },
  {
    "ph": "f", "id": 135081, "pid": 76337, "tid": -914061504, "ts": 1716454223425220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425221, "dur": 0,
    "args": {
      "External id": 135082,
      "cbid": 205, "correlation": 135082
    }
  },
  {
    "ph": "f", "id": 135082, "pid": 76337, "tid": -914061504, "ts": 1716454223425221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223498531, "dur": 38,
    "args": {
      "External id": 135086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135086,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135086, "pid": 5, "tid": 7, "ts": 1716454223498531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425233, "dur": 11,
    "args": {
      "External id": 135086,
      "cbid": 211, "correlation": 135086
    }
  },
  {
    "ph": "s", "id": 135086, "pid": 76337, "tid": -914061504, "ts": 1716454223425233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223498571, "dur": 187,
    "args": {
      "External id": 135088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135088,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135088, "pid": 5, "tid": 7, "ts": 1716454223498571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425250, "dur": 6,
    "args": {
      "External id": 135088,
      "cbid": 211, "correlation": 135088
    }
  },
  {
    "ph": "s", "id": 135088, "pid": 76337, "tid": -914061504, "ts": 1716454223425250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223498759, "dur": 22,
    "args": {
      "External id": 135090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135090, "pid": 5, "tid": 7, "ts": 1716454223498759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425260, "dur": 5,
    "args": {
      "External id": 135090,
      "cbid": 211, "correlation": 135090
    }
  },
  {
    "ph": "s", "id": 135090, "pid": 76337, "tid": -914061504, "ts": 1716454223425260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223498782, "dur": 32,
    "args": {
      "External id": 135096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135096, "pid": 5, "tid": 7, "ts": 1716454223498782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425285, "dur": 8,
    "args": {
      "External id": 135096,
      "cbid": 211, "correlation": 135096
    }
  },
  {
    "ph": "s", "id": 135096, "pid": 76337, "tid": -914061504, "ts": 1716454223425285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223498815, "dur": 27,
    "args": {
      "External id": 135104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135104,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135104, "pid": 5, "tid": 7, "ts": 1716454223498815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425313, "dur": 8,
    "args": {
      "External id": 135104,
      "cbid": 211, "correlation": 135104
    }
  },
  {
    "ph": "s", "id": 135104, "pid": 76337, "tid": -914061504, "ts": 1716454223425313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223498843, "dur": 20,
    "args": {
      "External id": 135112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135112,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135112, "pid": 5, "tid": 7, "ts": 1716454223498843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425341, "dur": 9,
    "args": {
      "External id": 135112,
      "cbid": 211, "correlation": 135112
    }
  },
  {
    "ph": "s", "id": 135112, "pid": 76337, "tid": -914061504, "ts": 1716454223425341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223498864, "dur": 30,
    "args": {
      "External id": 135132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135132,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 135132, "pid": 5, "tid": 7, "ts": 1716454223498864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425425, "dur": 12,
    "args": {
      "External id": 135132,
      "cbid": 211, "correlation": 135132
    }
  },
  {
    "ph": "s", "id": 135132, "pid": 76337, "tid": -914061504, "ts": 1716454223425425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223498895, "dur": 4,
    "args": {
      "External id": 135144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135144,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 135144, "pid": 5, "tid": 7, "ts": 1716454223498895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425447, "dur": 7,
    "args": {
      "External id": 135144,
      "cbid": 211, "correlation": 135144
    }
  },
  {
    "ph": "s", "id": 135144, "pid": 76337, "tid": -914061504, "ts": 1716454223425447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223498900, "dur": 31,
    "args": {
      "External id": 135147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135147,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135147, "pid": 5, "tid": 7, "ts": 1716454223498900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425465, "dur": 6,
    "args": {
      "External id": 135147,
      "cbid": 211, "correlation": 135147
    }
  },
  {
    "ph": "s", "id": 135147, "pid": 76337, "tid": -914061504, "ts": 1716454223425465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425520, "dur": 0,
    "args": {
      "External id": 135158,
      "cbid": 317, "correlation": 135158
    }
  },
  {
    "ph": "f", "id": 135158, "pid": 76337, "tid": -914061504, "ts": 1716454223425520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425521, "dur": 0,
    "args": {
      "External id": 135159,
      "cbid": 203, "correlation": 135159
    }
  },
  {
    "ph": "f", "id": 135159, "pid": 76337, "tid": -914061504, "ts": 1716454223425521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425522, "dur": 0,
    "args": {
      "External id": 135160,
      "cbid": 205, "correlation": 135160
    }
  },
  {
    "ph": "f", "id": 135160, "pid": 76337, "tid": -914061504, "ts": 1716454223425522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223498932, "dur": 22,
    "args": {
      "External id": 135164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135164,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135164, "pid": 5, "tid": 7, "ts": 1716454223498932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425535, "dur": 12,
    "args": {
      "External id": 135164,
      "cbid": 211, "correlation": 135164
    }
  },
  {
    "ph": "s", "id": 135164, "pid": 76337, "tid": -914061504, "ts": 1716454223425535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223498956, "dur": 103,
    "args": {
      "External id": 135166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135166,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135166, "pid": 5, "tid": 7, "ts": 1716454223498956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425553, "dur": 6,
    "args": {
      "External id": 135166,
      "cbid": 211, "correlation": 135166
    }
  },
  {
    "ph": "s", "id": 135166, "pid": 76337, "tid": -914061504, "ts": 1716454223425553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223499060, "dur": 23,
    "args": {
      "External id": 135168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135168, "pid": 5, "tid": 7, "ts": 1716454223499060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425563, "dur": 5,
    "args": {
      "External id": 135168,
      "cbid": 211, "correlation": 135168
    }
  },
  {
    "ph": "s", "id": 135168, "pid": 76337, "tid": -914061504, "ts": 1716454223425563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223499084, "dur": 32,
    "args": {
      "External id": 135174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135174, "pid": 5, "tid": 7, "ts": 1716454223499084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425590, "dur": 9,
    "args": {
      "External id": 135174,
      "cbid": 211, "correlation": 135174
    }
  },
  {
    "ph": "s", "id": 135174, "pid": 76337, "tid": -914061504, "ts": 1716454223425590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223499118, "dur": 194,
    "args": {
      "External id": 135183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135183,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135183, "pid": 5, "tid": 7, "ts": 1716454223499118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425673, "dur": 14,
    "args": {
      "External id": 135183,
      "cbid": 211, "correlation": 135183
    }
  },
  {
    "ph": "s", "id": 135183, "pid": 76337, "tid": -914061504, "ts": 1716454223425673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223499313, "dur": 65,
    "args": {
      "External id": 135205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135205,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135205, "pid": 5, "tid": 7, "ts": 1716454223499313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425731, "dur": 10,
    "args": {
      "External id": 135205,
      "cbid": 211, "correlation": 135205
    }
  },
  {
    "ph": "s", "id": 135205, "pid": 76337, "tid": -914061504, "ts": 1716454223425731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425819, "dur": 1,
    "args": {
      "External id": 135216,
      "cbid": 251, "correlation": 135216
    }
  },
  {
    "ph": "f", "id": 135216, "pid": 76337, "tid": -914061504, "ts": 1716454223425819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223499379, "dur": 152,
    "args": {
      "External id": 135217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135217,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135217, "pid": 5, "tid": 7, "ts": 1716454223499379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425825, "dur": 13,
    "args": {
      "External id": 135217,
      "cbid": 211, "correlation": 135217
    }
  },
  {
    "ph": "s", "id": 135217, "pid": 76337, "tid": -914061504, "ts": 1716454223425825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425894, "dur": 1,
    "args": {
      "External id": 135228,
      "cbid": 251, "correlation": 135228
    }
  },
  {
    "ph": "f", "id": 135228, "pid": 76337, "tid": -914061504, "ts": 1716454223425894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223499532, "dur": 146,
    "args": {
      "External id": 135229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135229,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135229, "pid": 5, "tid": 7, "ts": 1716454223499532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425898, "dur": 12,
    "args": {
      "External id": 135229,
      "cbid": 211, "correlation": 135229
    }
  },
  {
    "ph": "s", "id": 135229, "pid": 76337, "tid": -914061504, "ts": 1716454223425898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425964, "dur": 1,
    "args": {
      "External id": 135240,
      "cbid": 251, "correlation": 135240
    }
  },
  {
    "ph": "f", "id": 135240, "pid": 76337, "tid": -914061504, "ts": 1716454223425964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223499680, "dur": 145,
    "args": {
      "External id": 135241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135241,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135241, "pid": 5, "tid": 7, "ts": 1716454223499680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223425968, "dur": 20,
    "args": {
      "External id": 135241,
      "cbid": 211, "correlation": 135241
    }
  },
  {
    "ph": "s", "id": 135241, "pid": 76337, "tid": -914061504, "ts": 1716454223425968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223499826, "dur": 1910,
    "args": {
      "External id": 135262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135262,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 135262, "pid": 5, "tid": 7, "ts": 1716454223499826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426059, "dur": 12,
    "args": {
      "External id": 135262,
      "cbid": 211, "correlation": 135262
    }
  },
  {
    "ph": "s", "id": 135262, "pid": 76337, "tid": -914061504, "ts": 1716454223426059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426156, "dur": 1,
    "args": {
      "External id": 135280,
      "cbid": 251, "correlation": 135280
    }
  },
  {
    "ph": "f", "id": 135280, "pid": 76337, "tid": -914061504, "ts": 1716454223426156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223501737, "dur": 147,
    "args": {
      "External id": 135282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135282,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 135282, "pid": 5, "tid": 7, "ts": 1716454223501737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426162, "dur": 13,
    "args": {
      "External id": 135282,
      "cbid": 211, "correlation": 135282
    }
  },
  {
    "ph": "s", "id": 135282, "pid": 76337, "tid": -914061504, "ts": 1716454223426162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223501886, "dur": 35,
    "args": {
      "External id": 135290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135290,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135290, "pid": 5, "tid": 7, "ts": 1716454223501886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426232, "dur": 12,
    "args": {
      "External id": 135290,
      "cbid": 211, "correlation": 135290
    }
  },
  {
    "ph": "s", "id": 135290, "pid": 76337, "tid": -914061504, "ts": 1716454223426232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223501922, "dur": 51,
    "args": {
      "External id": 135298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135298,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135298, "pid": 5, "tid": 7, "ts": 1716454223501922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426271, "dur": 9,
    "args": {
      "External id": 135298,
      "cbid": 211, "correlation": 135298
    }
  },
  {
    "ph": "s", "id": 135298, "pid": 76337, "tid": -914061504, "ts": 1716454223426271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223501974, "dur": 30,
    "args": {
      "External id": 135309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135309, "pid": 5, "tid": 7, "ts": 1716454223501974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426342, "dur": 13,
    "args": {
      "External id": 135309,
      "cbid": 211, "correlation": 135309
    }
  },
  {
    "ph": "s", "id": 135309, "pid": 76337, "tid": -914061504, "ts": 1716454223426342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223502005, "dur": 34,
    "args": {
      "External id": 135331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135331,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135331, "pid": 5, "tid": 7, "ts": 1716454223502005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426374, "dur": 8,
    "args": {
      "External id": 135331,
      "cbid": 211, "correlation": 135331
    }
  },
  {
    "ph": "s", "id": 135331, "pid": 76337, "tid": -914061504, "ts": 1716454223426374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426458, "dur": 1,
    "args": {
      "External id": 135342,
      "cbid": 251, "correlation": 135342
    }
  },
  {
    "ph": "f", "id": 135342, "pid": 76337, "tid": -914061504, "ts": 1716454223426458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223502041, "dur": 89,
    "args": {
      "External id": 135343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135343,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135343, "pid": 5, "tid": 7, "ts": 1716454223502041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426463, "dur": 13,
    "args": {
      "External id": 135343,
      "cbid": 211, "correlation": 135343
    }
  },
  {
    "ph": "s", "id": 135343, "pid": 76337, "tid": -914061504, "ts": 1716454223426463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426532, "dur": 1,
    "args": {
      "External id": 135354,
      "cbid": 251, "correlation": 135354
    }
  },
  {
    "ph": "f", "id": 135354, "pid": 76337, "tid": -914061504, "ts": 1716454223426532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426536, "dur": 0,
    "args": {
      "External id": 135355,
      "cbid": 251, "correlation": 135355
    }
  },
  {
    "ph": "f", "id": 135355, "pid": 76337, "tid": -914061504, "ts": 1716454223426536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223502131, "dur": 11,
    "args": {
      "External id": 135356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135356,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 135356, "pid": 5, "tid": 7, "ts": 1716454223502131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426537, "dur": 12,
    "args": {
      "External id": 135356,
      "cbid": 211, "correlation": 135356
    }
  },
  {
    "ph": "s", "id": 135356, "pid": 76337, "tid": -914061504, "ts": 1716454223426537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223502143, "dur": 5,
    "args": {
      "External id": 135358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135358,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 135358, "pid": 5, "tid": 7, "ts": 1716454223502143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426551, "dur": 5,
    "args": {
      "External id": 135358,
      "cbid": 211, "correlation": 135358
    }
  },
  {
    "ph": "s", "id": 135358, "pid": 76337, "tid": -914061504, "ts": 1716454223426551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426608, "dur": 1,
    "args": {
      "External id": 135369,
      "cbid": 251, "correlation": 135369
    }
  },
  {
    "ph": "f", "id": 135369, "pid": 76337, "tid": -914061504, "ts": 1716454223426608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426612, "dur": 0,
    "args": {
      "External id": 135370,
      "cbid": 251, "correlation": 135370
    }
  },
  {
    "ph": "f", "id": 135370, "pid": 76337, "tid": -914061504, "ts": 1716454223426612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223502149, "dur": 7,
    "args": {
      "External id": 135371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135371,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 135371, "pid": 5, "tid": 7, "ts": 1716454223502149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426613, "dur": 12,
    "args": {
      "External id": 135371,
      "cbid": 211, "correlation": 135371
    }
  },
  {
    "ph": "s", "id": 135371, "pid": 76337, "tid": -914061504, "ts": 1716454223426613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223502157, "dur": 3,
    "args": {
      "External id": 135373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135373,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 135373, "pid": 5, "tid": 7, "ts": 1716454223502157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426627, "dur": 5,
    "args": {
      "External id": 135373,
      "cbid": 211, "correlation": 135373
    }
  },
  {
    "ph": "s", "id": 135373, "pid": 76337, "tid": -914061504, "ts": 1716454223426627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223502162, "dur": 89,
    "args": {
      "External id": 135394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135394,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 135394, "pid": 5, "tid": 7, "ts": 1716454223502162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426700, "dur": 13,
    "args": {
      "External id": 135394,
      "cbid": 211, "correlation": 135394
    }
  },
  {
    "ph": "s", "id": 135394, "pid": 76337, "tid": -914061504, "ts": 1716454223426700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426797, "dur": 1,
    "args": {
      "External id": 135412,
      "cbid": 251, "correlation": 135412
    }
  },
  {
    "ph": "f", "id": 135412, "pid": 76337, "tid": -914061504, "ts": 1716454223426797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223502253, "dur": 97,
    "args": {
      "External id": 135414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135414,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135414, "pid": 5, "tid": 7, "ts": 1716454223502253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426803, "dur": 13,
    "args": {
      "External id": 135414,
      "cbid": 211, "correlation": 135414
    }
  },
  {
    "ph": "s", "id": 135414, "pid": 76337, "tid": -914061504, "ts": 1716454223426803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223502351, "dur": 19,
    "args": {
      "External id": 135422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135422,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135422, "pid": 5, "tid": 7, "ts": 1716454223502351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426872, "dur": 12,
    "args": {
      "External id": 135422,
      "cbid": 211, "correlation": 135422
    }
  },
  {
    "ph": "s", "id": 135422, "pid": 76337, "tid": -914061504, "ts": 1716454223426872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223502371, "dur": 37,
    "args": {
      "External id": 135430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135430, "pid": 5, "tid": 7, "ts": 1716454223502371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426913, "dur": 10,
    "args": {
      "External id": 135430,
      "cbid": 211, "correlation": 135430
    }
  },
  {
    "ph": "s", "id": 135430, "pid": 76337, "tid": -914061504, "ts": 1716454223426913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223502410, "dur": 34,
    "args": {
      "External id": 135452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135452,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135452, "pid": 5, "tid": 7, "ts": 1716454223502410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223426965, "dur": 19,
    "args": {
      "External id": 135452,
      "cbid": 211, "correlation": 135452
    }
  },
  {
    "ph": "s", "id": 135452, "pid": 76337, "tid": -914061504, "ts": 1716454223426965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427065, "dur": 1,
    "args": {
      "External id": 135468,
      "cbid": 251, "correlation": 135468
    }
  },
  {
    "ph": "f", "id": 135468, "pid": 76337, "tid": -914061504, "ts": 1716454223427065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427070, "dur": 0,
    "args": {
      "External id": 135470,
      "cbid": 251, "correlation": 135470
    }
  },
  {
    "ph": "f", "id": 135470, "pid": 76337, "tid": -914061504, "ts": 1716454223427070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223502445, "dur": 529,
    "args": {
      "External id": 135471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135471,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 135471, "pid": 5, "tid": 7, "ts": 1716454223502445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427074, "dur": 13,
    "args": {
      "External id": 135471,
      "cbid": 211, "correlation": 135471
    }
  },
  {
    "ph": "s", "id": 135471, "pid": 76337, "tid": -914061504, "ts": 1716454223427074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223502976, "dur": 123,
    "args": {
      "External id": 135479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135479, "pid": 5, "tid": 7, "ts": 1716454223502976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427140, "dur": 13,
    "args": {
      "External id": 135479,
      "cbid": 211, "correlation": 135479
    }
  },
  {
    "ph": "s", "id": 135479, "pid": 76337, "tid": -914061504, "ts": 1716454223427140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223503100, "dur": 128,
    "args": {
      "External id": 135487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135487,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135487, "pid": 5, "tid": 7, "ts": 1716454223503100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427169, "dur": 9,
    "args": {
      "External id": 135487,
      "cbid": 211, "correlation": 135487
    }
  },
  {
    "ph": "s", "id": 135487, "pid": 76337, "tid": -914061504, "ts": 1716454223427169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427246, "dur": 1,
    "args": {
      "External id": 135503,
      "cbid": 251, "correlation": 135503
    }
  },
  {
    "ph": "f", "id": 135503, "pid": 76337, "tid": -914061504, "ts": 1716454223427246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223503229, "dur": 303,
    "args": {
      "External id": 135505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135505,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135505, "pid": 5, "tid": 7, "ts": 1716454223503229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427252, "dur": 12,
    "args": {
      "External id": 135505,
      "cbid": 211, "correlation": 135505
    }
  },
  {
    "ph": "s", "id": 135505, "pid": 76337, "tid": -914061504, "ts": 1716454223427252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223503533, "dur": 27,
    "args": {
      "External id": 135513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135513,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135513, "pid": 5, "tid": 7, "ts": 1716454223503533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427294, "dur": 9,
    "args": {
      "External id": 135513,
      "cbid": 211, "correlation": 135513
    }
  },
  {
    "ph": "s", "id": 135513, "pid": 76337, "tid": -914061504, "ts": 1716454223427294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223503561, "dur": 80,
    "args": {
      "External id": 135524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135524, "pid": 5, "tid": 7, "ts": 1716454223503561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427360, "dur": 13,
    "args": {
      "External id": 135524,
      "cbid": 211, "correlation": 135524
    }
  },
  {
    "ph": "s", "id": 135524, "pid": 76337, "tid": -914061504, "ts": 1716454223427360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427425, "dur": 0,
    "args": {
      "External id": 135536,
      "cbid": 317, "correlation": 135536
    }
  },
  {
    "ph": "f", "id": 135536, "pid": 76337, "tid": -914061504, "ts": 1716454223427425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427426, "dur": 0,
    "args": {
      "External id": 135537,
      "cbid": 203, "correlation": 135537
    }
  },
  {
    "ph": "f", "id": 135537, "pid": 76337, "tid": -914061504, "ts": 1716454223427426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427427, "dur": 0,
    "args": {
      "External id": 135538,
      "cbid": 205, "correlation": 135538
    }
  },
  {
    "ph": "f", "id": 135538, "pid": 76337, "tid": -914061504, "ts": 1716454223427427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223503642, "dur": 23,
    "args": {
      "External id": 135542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135542,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135542, "pid": 5, "tid": 7, "ts": 1716454223503642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427441, "dur": 12,
    "args": {
      "External id": 135542,
      "cbid": 211, "correlation": 135542
    }
  },
  {
    "ph": "s", "id": 135542, "pid": 76337, "tid": -914061504, "ts": 1716454223427441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223503667, "dur": 119,
    "args": {
      "External id": 135544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135544,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135544, "pid": 5, "tid": 7, "ts": 1716454223503667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427459, "dur": 6,
    "args": {
      "External id": 135544,
      "cbid": 211, "correlation": 135544
    }
  },
  {
    "ph": "s", "id": 135544, "pid": 76337, "tid": -914061504, "ts": 1716454223427459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223503787, "dur": 23,
    "args": {
      "External id": 135546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135546, "pid": 5, "tid": 7, "ts": 1716454223503787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427470, "dur": 6,
    "args": {
      "External id": 135546,
      "cbid": 211, "correlation": 135546
    }
  },
  {
    "ph": "s", "id": 135546, "pid": 76337, "tid": -914061504, "ts": 1716454223427470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223503811, "dur": 33,
    "args": {
      "External id": 135552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135552,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135552, "pid": 5, "tid": 7, "ts": 1716454223503811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427498, "dur": 8,
    "args": {
      "External id": 135552,
      "cbid": 211, "correlation": 135552
    }
  },
  {
    "ph": "s", "id": 135552, "pid": 76337, "tid": -914061504, "ts": 1716454223427498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223503845, "dur": 27,
    "args": {
      "External id": 135560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135560,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135560, "pid": 5, "tid": 7, "ts": 1716454223503845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427530, "dur": 8,
    "args": {
      "External id": 135560,
      "cbid": 211, "correlation": 135560
    }
  },
  {
    "ph": "s", "id": 135560, "pid": 76337, "tid": -914061504, "ts": 1716454223427530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223503873, "dur": 45,
    "args": {
      "External id": 135569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135569,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135569, "pid": 5, "tid": 7, "ts": 1716454223503873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427567, "dur": 11,
    "args": {
      "External id": 135569,
      "cbid": 211, "correlation": 135569
    }
  },
  {
    "ph": "s", "id": 135569, "pid": 76337, "tid": -914061504, "ts": 1716454223427567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223503919, "dur": 42,
    "args": {
      "External id": 135589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135589,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 135589, "pid": 5, "tid": 7, "ts": 1716454223503919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427638, "dur": 11,
    "args": {
      "External id": 135589,
      "cbid": 211, "correlation": 135589
    }
  },
  {
    "ph": "s", "id": 135589, "pid": 76337, "tid": -914061504, "ts": 1716454223427638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223503962, "dur": 5,
    "args": {
      "External id": 135601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135601,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 135601, "pid": 5, "tid": 7, "ts": 1716454223503962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427659, "dur": 6,
    "args": {
      "External id": 135601,
      "cbid": 211, "correlation": 135601
    }
  },
  {
    "ph": "s", "id": 135601, "pid": 76337, "tid": -914061504, "ts": 1716454223427659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223503968, "dur": 43,
    "args": {
      "External id": 135604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135604, "pid": 5, "tid": 7, "ts": 1716454223503968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427679, "dur": 6,
    "args": {
      "External id": 135604,
      "cbid": 211, "correlation": 135604
    }
  },
  {
    "ph": "s", "id": 135604, "pid": 76337, "tid": -914061504, "ts": 1716454223427679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223504013, "dur": 29,
    "args": {
      "External id": 135613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135613,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135613, "pid": 5, "tid": 7, "ts": 1716454223504013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427717, "dur": 11,
    "args": {
      "External id": 135613,
      "cbid": 211, "correlation": 135613
    }
  },
  {
    "ph": "s", "id": 135613, "pid": 76337, "tid": -914061504, "ts": 1716454223427717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427769, "dur": 0,
    "args": {
      "External id": 135623,
      "cbid": 317, "correlation": 135623
    }
  },
  {
    "ph": "f", "id": 135623, "pid": 76337, "tid": -914061504, "ts": 1716454223427769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427770, "dur": 0,
    "args": {
      "External id": 135624,
      "cbid": 203, "correlation": 135624
    }
  },
  {
    "ph": "f", "id": 135624, "pid": 76337, "tid": -914061504, "ts": 1716454223427770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427771, "dur": 0,
    "args": {
      "External id": 135625,
      "cbid": 205, "correlation": 135625
    }
  },
  {
    "ph": "f", "id": 135625, "pid": 76337, "tid": -914061504, "ts": 1716454223427771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223504043, "dur": 30,
    "args": {
      "External id": 135629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135629,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135629, "pid": 5, "tid": 7, "ts": 1716454223504043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427786, "dur": 12,
    "args": {
      "External id": 135629,
      "cbid": 211, "correlation": 135629
    }
  },
  {
    "ph": "s", "id": 135629, "pid": 76337, "tid": -914061504, "ts": 1716454223427786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223504074, "dur": 63,
    "args": {
      "External id": 135631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135631, "pid": 5, "tid": 7, "ts": 1716454223504074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427800, "dur": 5,
    "args": {
      "External id": 135631,
      "cbid": 211, "correlation": 135631
    }
  },
  {
    "ph": "s", "id": 135631, "pid": 76337, "tid": -914061504, "ts": 1716454223427800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223504138, "dur": 958,
    "args": {
      "External id": 135633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135633,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135633, "pid": 5, "tid": 7, "ts": 1716454223504138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427811, "dur": 6,
    "args": {
      "External id": 135633,
      "cbid": 211, "correlation": 135633
    }
  },
  {
    "ph": "s", "id": 135633, "pid": 76337, "tid": -914061504, "ts": 1716454223427811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223505097, "dur": 20,
    "args": {
      "External id": 135635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135635,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135635, "pid": 5, "tid": 7, "ts": 1716454223505097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427821, "dur": 6,
    "args": {
      "External id": 135635,
      "cbid": 211, "correlation": 135635
    }
  },
  {
    "ph": "s", "id": 135635, "pid": 76337, "tid": -914061504, "ts": 1716454223427821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223505119, "dur": 32,
    "args": {
      "External id": 135641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135641,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135641, "pid": 5, "tid": 7, "ts": 1716454223505119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427849, "dur": 8,
    "args": {
      "External id": 135641,
      "cbid": 211, "correlation": 135641
    }
  },
  {
    "ph": "s", "id": 135641, "pid": 76337, "tid": -914061504, "ts": 1716454223427849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223505153, "dur": 3,
    "args": {
      "External id": 135649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135649,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 135649, "pid": 5, "tid": 7, "ts": 1716454223505153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427892, "dur": 9,
    "args": {
      "External id": 135649,
      "cbid": 211, "correlation": 135649
    }
  },
  {
    "ph": "s", "id": 135649, "pid": 76337, "tid": -914061504, "ts": 1716454223427892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427956, "dur": 1,
    "args": {
      "External id": 135665,
      "cbid": 251, "correlation": 135665
    }
  },
  {
    "ph": "f", "id": 135665, "pid": 76337, "tid": -914061504, "ts": 1716454223427956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427962, "dur": 0,
    "args": {
      "External id": 135667,
      "cbid": 251, "correlation": 135667
    }
  },
  {
    "ph": "f", "id": 135667, "pid": 76337, "tid": -914061504, "ts": 1716454223427962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223505157, "dur": 12,
    "args": {
      "External id": 135668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135668,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 135668, "pid": 5, "tid": 7, "ts": 1716454223505157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427963, "dur": 19,
    "args": {
      "External id": 135668,
      "cbid": 211, "correlation": 135668
    }
  },
  {
    "ph": "s", "id": 135668, "pid": 76337, "tid": -914061504, "ts": 1716454223427963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223505171, "dur": 5,
    "args": {
      "External id": 135670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135670,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 135670, "pid": 5, "tid": 7, "ts": 1716454223505171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223427985, "dur": 6,
    "args": {
      "External id": 135670,
      "cbid": 211, "correlation": 135670
    }
  },
  {
    "ph": "s", "id": 135670, "pid": 76337, "tid": -914061504, "ts": 1716454223427985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223505177, "dur": 29,
    "args": {
      "External id": 135680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135680, "pid": 5, "tid": 7, "ts": 1716454223505177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428043, "dur": 12,
    "args": {
      "External id": 135680,
      "cbid": 211, "correlation": 135680
    }
  },
  {
    "ph": "s", "id": 135680, "pid": 76337, "tid": -914061504, "ts": 1716454223428043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223505207, "dur": 31,
    "args": {
      "External id": 135700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135700,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 135700, "pid": 5, "tid": 7, "ts": 1716454223505207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428108, "dur": 11,
    "args": {
      "External id": 135700,
      "cbid": 211, "correlation": 135700
    }
  },
  {
    "ph": "s", "id": 135700, "pid": 76337, "tid": -914061504, "ts": 1716454223428108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223505239, "dur": 4,
    "args": {
      "External id": 135712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135712,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 135712, "pid": 5, "tid": 7, "ts": 1716454223505239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428128, "dur": 7,
    "args": {
      "External id": 135712,
      "cbid": 211, "correlation": 135712
    }
  },
  {
    "ph": "s", "id": 135712, "pid": 76337, "tid": -914061504, "ts": 1716454223428128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223505244, "dur": 29,
    "args": {
      "External id": 135715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135715,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135715, "pid": 5, "tid": 7, "ts": 1716454223505244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428147, "dur": 6,
    "args": {
      "External id": 135715,
      "cbid": 211, "correlation": 135715
    }
  },
  {
    "ph": "s", "id": 135715, "pid": 76337, "tid": -914061504, "ts": 1716454223428147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223505275, "dur": 20,
    "args": {
      "External id": 135724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135724,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135724, "pid": 5, "tid": 7, "ts": 1716454223505275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428188, "dur": 9,
    "args": {
      "External id": 135724,
      "cbid": 211, "correlation": 135724
    }
  },
  {
    "ph": "s", "id": 135724, "pid": 76337, "tid": -914061504, "ts": 1716454223428188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428250, "dur": 0,
    "args": {
      "External id": 135734,
      "cbid": 317, "correlation": 135734
    }
  },
  {
    "ph": "f", "id": 135734, "pid": 76337, "tid": -914061504, "ts": 1716454223428250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428251, "dur": 0,
    "args": {
      "External id": 135735,
      "cbid": 203, "correlation": 135735
    }
  },
  {
    "ph": "f", "id": 135735, "pid": 76337, "tid": -914061504, "ts": 1716454223428251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428251, "dur": 0,
    "args": {
      "External id": 135736,
      "cbid": 205, "correlation": 135736
    }
  },
  {
    "ph": "f", "id": 135736, "pid": 76337, "tid": -914061504, "ts": 1716454223428251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223505296, "dur": 23,
    "args": {
      "External id": 135740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135740,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135740, "pid": 5, "tid": 7, "ts": 1716454223505296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428265, "dur": 12,
    "args": {
      "External id": 135740,
      "cbid": 211, "correlation": 135740
    }
  },
  {
    "ph": "s", "id": 135740, "pid": 76337, "tid": -914061504, "ts": 1716454223428265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223505320, "dur": 43,
    "args": {
      "External id": 135742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135742, "pid": 5, "tid": 7, "ts": 1716454223505320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428279, "dur": 6,
    "args": {
      "External id": 135742,
      "cbid": 211, "correlation": 135742
    }
  },
  {
    "ph": "s", "id": 135742, "pid": 76337, "tid": -914061504, "ts": 1716454223428279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223505365, "dur": 638,
    "args": {
      "External id": 135744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135744,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135744, "pid": 5, "tid": 7, "ts": 1716454223505365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428291, "dur": 6,
    "args": {
      "External id": 135744,
      "cbid": 211, "correlation": 135744
    }
  },
  {
    "ph": "s", "id": 135744, "pid": 76337, "tid": -914061504, "ts": 1716454223428291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506004, "dur": 21,
    "args": {
      "External id": 135746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135746,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135746, "pid": 5, "tid": 7, "ts": 1716454223506004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428301, "dur": 5,
    "args": {
      "External id": 135746,
      "cbid": 211, "correlation": 135746
    }
  },
  {
    "ph": "s", "id": 135746, "pid": 76337, "tid": -914061504, "ts": 1716454223428301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223506026, "dur": 33,
    "args": {
      "External id": 135752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135752, "pid": 5, "tid": 7, "ts": 1716454223506026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428329, "dur": 9,
    "args": {
      "External id": 135752,
      "cbid": 211, "correlation": 135752
    }
  },
  {
    "ph": "s", "id": 135752, "pid": 76337, "tid": -914061504, "ts": 1716454223428329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428387, "dur": 0,
    "args": {
      "External id": 135762,
      "cbid": 317, "correlation": 135762
    }
  },
  {
    "ph": "f", "id": 135762, "pid": 76337, "tid": -914061504, "ts": 1716454223428387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428387, "dur": 0,
    "args": {
      "External id": 135763,
      "cbid": 203, "correlation": 135763
    }
  },
  {
    "ph": "f", "id": 135763, "pid": 76337, "tid": -914061504, "ts": 1716454223428387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428388, "dur": 0,
    "args": {
      "External id": 135764,
      "cbid": 205, "correlation": 135764
    }
  },
  {
    "ph": "f", "id": 135764, "pid": 76337, "tid": -914061504, "ts": 1716454223428388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506061, "dur": 30,
    "args": {
      "External id": 135768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135768, "pid": 5, "tid": 7, "ts": 1716454223506061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428402, "dur": 12,
    "args": {
      "External id": 135768,
      "cbid": 211, "correlation": 135768
    }
  },
  {
    "ph": "s", "id": 135768, "pid": 76337, "tid": -914061504, "ts": 1716454223428402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223506092, "dur": 150,
    "args": {
      "External id": 135770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135770,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135770, "pid": 5, "tid": 7, "ts": 1716454223506092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428420, "dur": 6,
    "args": {
      "External id": 135770,
      "cbid": 211, "correlation": 135770
    }
  },
  {
    "ph": "s", "id": 135770, "pid": 76337, "tid": -914061504, "ts": 1716454223428420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506243, "dur": 20,
    "args": {
      "External id": 135772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135772, "pid": 5, "tid": 7, "ts": 1716454223506243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428430, "dur": 6,
    "args": {
      "External id": 135772,
      "cbid": 211, "correlation": 135772
    }
  },
  {
    "ph": "s", "id": 135772, "pid": 76337, "tid": -914061504, "ts": 1716454223428430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223506264, "dur": 32,
    "args": {
      "External id": 135778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135778, "pid": 5, "tid": 7, "ts": 1716454223506264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428456, "dur": 8,
    "args": {
      "External id": 135778,
      "cbid": 211, "correlation": 135778
    }
  },
  {
    "ph": "s", "id": 135778, "pid": 76337, "tid": -914061504, "ts": 1716454223428456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223506297, "dur": 27,
    "args": {
      "External id": 135786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135786,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135786, "pid": 5, "tid": 7, "ts": 1716454223506297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428484, "dur": 8,
    "args": {
      "External id": 135786,
      "cbid": 211, "correlation": 135786
    }
  },
  {
    "ph": "s", "id": 135786, "pid": 76337, "tid": -914061504, "ts": 1716454223428484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223506326, "dur": 20,
    "args": {
      "External id": 135794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135794,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135794, "pid": 5, "tid": 7, "ts": 1716454223506326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428512, "dur": 8,
    "args": {
      "External id": 135794,
      "cbid": 211, "correlation": 135794
    }
  },
  {
    "ph": "s", "id": 135794, "pid": 76337, "tid": -914061504, "ts": 1716454223428512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506347, "dur": 29,
    "args": {
      "External id": 135814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135814,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 135814, "pid": 5, "tid": 7, "ts": 1716454223506347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428595, "dur": 12,
    "args": {
      "External id": 135814,
      "cbid": 211, "correlation": 135814
    }
  },
  {
    "ph": "s", "id": 135814, "pid": 76337, "tid": -914061504, "ts": 1716454223428595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223506378, "dur": 5,
    "args": {
      "External id": 135826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135826,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 135826, "pid": 5, "tid": 7, "ts": 1716454223506378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428617, "dur": 6,
    "args": {
      "External id": 135826,
      "cbid": 211, "correlation": 135826
    }
  },
  {
    "ph": "s", "id": 135826, "pid": 76337, "tid": -914061504, "ts": 1716454223428617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223506384, "dur": 30,
    "args": {
      "External id": 135829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135829, "pid": 5, "tid": 7, "ts": 1716454223506384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428635, "dur": 6,
    "args": {
      "External id": 135829,
      "cbid": 211, "correlation": 135829
    }
  },
  {
    "ph": "s", "id": 135829, "pid": 76337, "tid": -914061504, "ts": 1716454223428635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428691, "dur": 0,
    "args": {
      "External id": 135840,
      "cbid": 317, "correlation": 135840
    }
  },
  {
    "ph": "f", "id": 135840, "pid": 76337, "tid": -914061504, "ts": 1716454223428691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428692, "dur": 0,
    "args": {
      "External id": 135841,
      "cbid": 203, "correlation": 135841
    }
  },
  {
    "ph": "f", "id": 135841, "pid": 76337, "tid": -914061504, "ts": 1716454223428692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428693, "dur": 0,
    "args": {
      "External id": 135842,
      "cbid": 205, "correlation": 135842
    }
  },
  {
    "ph": "f", "id": 135842, "pid": 76337, "tid": -914061504, "ts": 1716454223428693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506415, "dur": 22,
    "args": {
      "External id": 135846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135846, "pid": 5, "tid": 7, "ts": 1716454223506415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428706, "dur": 12,
    "args": {
      "External id": 135846,
      "cbid": 211, "correlation": 135846
    }
  },
  {
    "ph": "s", "id": 135846, "pid": 76337, "tid": -914061504, "ts": 1716454223428706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223506438, "dur": 103,
    "args": {
      "External id": 135848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135848,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135848, "pid": 5, "tid": 7, "ts": 1716454223506438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428724, "dur": 6,
    "args": {
      "External id": 135848,
      "cbid": 211, "correlation": 135848
    }
  },
  {
    "ph": "s", "id": 135848, "pid": 76337, "tid": -914061504, "ts": 1716454223428724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506543, "dur": 21,
    "args": {
      "External id": 135850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135850,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135850, "pid": 5, "tid": 7, "ts": 1716454223506543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428733, "dur": 6,
    "args": {
      "External id": 135850,
      "cbid": 211, "correlation": 135850
    }
  },
  {
    "ph": "s", "id": 135850, "pid": 76337, "tid": -914061504, "ts": 1716454223428733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223506565, "dur": 32,
    "args": {
      "External id": 135856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135856, "pid": 5, "tid": 7, "ts": 1716454223506565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428761, "dur": 8,
    "args": {
      "External id": 135856,
      "cbid": 211, "correlation": 135856
    }
  },
  {
    "ph": "s", "id": 135856, "pid": 76337, "tid": -914061504, "ts": 1716454223428761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223506598, "dur": 184,
    "args": {
      "External id": 135865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135865,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135865, "pid": 5, "tid": 7, "ts": 1716454223506598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428844, "dur": 14,
    "args": {
      "External id": 135865,
      "cbid": 211, "correlation": 135865
    }
  },
  {
    "ph": "s", "id": 135865, "pid": 76337, "tid": -914061504, "ts": 1716454223428844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223506784, "dur": 64,
    "args": {
      "External id": 135887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135887,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135887, "pid": 5, "tid": 7, "ts": 1716454223506784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428901, "dur": 10,
    "args": {
      "External id": 135887,
      "cbid": 211, "correlation": 135887
    }
  },
  {
    "ph": "s", "id": 135887, "pid": 76337, "tid": -914061504, "ts": 1716454223428901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223428995, "dur": 1,
    "args": {
      "External id": 135898,
      "cbid": 251, "correlation": 135898
    }
  },
  {
    "ph": "f", "id": 135898, "pid": 76337, "tid": -914061504, "ts": 1716454223428995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223506849, "dur": 151,
    "args": {
      "External id": 135899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135899,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135899, "pid": 5, "tid": 7, "ts": 1716454223506849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429001, "dur": 13,
    "args": {
      "External id": 135899,
      "cbid": 211, "correlation": 135899
    }
  },
  {
    "ph": "s", "id": 135899, "pid": 76337, "tid": -914061504, "ts": 1716454223429001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429072, "dur": 1,
    "args": {
      "External id": 135910,
      "cbid": 251, "correlation": 135910
    }
  },
  {
    "ph": "f", "id": 135910, "pid": 76337, "tid": -914061504, "ts": 1716454223429072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223507002, "dur": 142,
    "args": {
      "External id": 135911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135911,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135911, "pid": 5, "tid": 7, "ts": 1716454223507002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429077, "dur": 12,
    "args": {
      "External id": 135911,
      "cbid": 211, "correlation": 135911
    }
  },
  {
    "ph": "s", "id": 135911, "pid": 76337, "tid": -914061504, "ts": 1716454223429077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429143, "dur": 1,
    "args": {
      "External id": 135922,
      "cbid": 251, "correlation": 135922
    }
  },
  {
    "ph": "f", "id": 135922, "pid": 76337, "tid": -914061504, "ts": 1716454223429143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223507146, "dur": 144,
    "args": {
      "External id": 135923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135923,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 135923, "pid": 5, "tid": 7, "ts": 1716454223507146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429147, "dur": 11,
    "args": {
      "External id": 135923,
      "cbid": 211, "correlation": 135923
    }
  },
  {
    "ph": "s", "id": 135923, "pid": 76337, "tid": -914061504, "ts": 1716454223429147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223507292, "dur": 1908,
    "args": {
      "External id": 135944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135944,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 135944, "pid": 5, "tid": 7, "ts": 1716454223507292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429227, "dur": 13,
    "args": {
      "External id": 135944,
      "cbid": 211, "correlation": 135944
    }
  },
  {
    "ph": "s", "id": 135944, "pid": 76337, "tid": -914061504, "ts": 1716454223429227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429324, "dur": 1,
    "args": {
      "External id": 135962,
      "cbid": 251, "correlation": 135962
    }
  },
  {
    "ph": "f", "id": 135962, "pid": 76337, "tid": -914061504, "ts": 1716454223429324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223509200, "dur": 147,
    "args": {
      "External id": 135964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135964,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 135964, "pid": 5, "tid": 7, "ts": 1716454223509200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429330, "dur": 14,
    "args": {
      "External id": 135964,
      "cbid": 211, "correlation": 135964
    }
  },
  {
    "ph": "s", "id": 135964, "pid": 76337, "tid": -914061504, "ts": 1716454223429330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223509349, "dur": 35,
    "args": {
      "External id": 135972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135972,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135972, "pid": 5, "tid": 7, "ts": 1716454223509349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429401, "dur": 12,
    "args": {
      "External id": 135972,
      "cbid": 211, "correlation": 135972
    }
  },
  {
    "ph": "s", "id": 135972, "pid": 76337, "tid": -914061504, "ts": 1716454223429401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223509385, "dur": 50,
    "args": {
      "External id": 135980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135980,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135980, "pid": 5, "tid": 7, "ts": 1716454223509385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429440, "dur": 9,
    "args": {
      "External id": 135980,
      "cbid": 211, "correlation": 135980
    }
  },
  {
    "ph": "s", "id": 135980, "pid": 76337, "tid": -914061504, "ts": 1716454223429440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223509437, "dur": 29,
    "args": {
      "External id": 135991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 135991,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 135991, "pid": 5, "tid": 7, "ts": 1716454223509437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429510, "dur": 13,
    "args": {
      "External id": 135991,
      "cbid": 211, "correlation": 135991
    }
  },
  {
    "ph": "s", "id": 135991, "pid": 76337, "tid": -914061504, "ts": 1716454223429510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223509468, "dur": 35,
    "args": {
      "External id": 136013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136013,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136013, "pid": 5, "tid": 7, "ts": 1716454223509468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429541, "dur": 8,
    "args": {
      "External id": 136013,
      "cbid": 211, "correlation": 136013
    }
  },
  {
    "ph": "s", "id": 136013, "pid": 76337, "tid": -914061504, "ts": 1716454223429541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429626, "dur": 1,
    "args": {
      "External id": 136024,
      "cbid": 251, "correlation": 136024
    }
  },
  {
    "ph": "f", "id": 136024, "pid": 76337, "tid": -914061504, "ts": 1716454223429626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223509504, "dur": 87,
    "args": {
      "External id": 136025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136025,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136025, "pid": 5, "tid": 7, "ts": 1716454223509504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429631, "dur": 13,
    "args": {
      "External id": 136025,
      "cbid": 211, "correlation": 136025
    }
  },
  {
    "ph": "s", "id": 136025, "pid": 76337, "tid": -914061504, "ts": 1716454223429631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429699, "dur": 1,
    "args": {
      "External id": 136036,
      "cbid": 251, "correlation": 136036
    }
  },
  {
    "ph": "f", "id": 136036, "pid": 76337, "tid": -914061504, "ts": 1716454223429699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429703, "dur": 0,
    "args": {
      "External id": 136037,
      "cbid": 251, "correlation": 136037
    }
  },
  {
    "ph": "f", "id": 136037, "pid": 76337, "tid": -914061504, "ts": 1716454223429703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223509592, "dur": 11,
    "args": {
      "External id": 136038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136038,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 136038, "pid": 5, "tid": 7, "ts": 1716454223509592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429705, "dur": 12,
    "args": {
      "External id": 136038,
      "cbid": 211, "correlation": 136038
    }
  },
  {
    "ph": "s", "id": 136038, "pid": 76337, "tid": -914061504, "ts": 1716454223429705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223509605, "dur": 5,
    "args": {
      "External id": 136040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136040,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 136040, "pid": 5, "tid": 7, "ts": 1716454223509605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429718, "dur": 6,
    "args": {
      "External id": 136040,
      "cbid": 211, "correlation": 136040
    }
  },
  {
    "ph": "s", "id": 136040, "pid": 76337, "tid": -914061504, "ts": 1716454223429718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429776, "dur": 1,
    "args": {
      "External id": 136051,
      "cbid": 251, "correlation": 136051
    }
  },
  {
    "ph": "f", "id": 136051, "pid": 76337, "tid": -914061504, "ts": 1716454223429776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429779, "dur": 0,
    "args": {
      "External id": 136052,
      "cbid": 251, "correlation": 136052
    }
  },
  {
    "ph": "f", "id": 136052, "pid": 76337, "tid": -914061504, "ts": 1716454223429779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223509611, "dur": 7,
    "args": {
      "External id": 136053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136053,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 136053, "pid": 5, "tid": 7, "ts": 1716454223509611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429780, "dur": 12,
    "args": {
      "External id": 136053,
      "cbid": 211, "correlation": 136053
    }
  },
  {
    "ph": "s", "id": 136053, "pid": 76337, "tid": -914061504, "ts": 1716454223429780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223509619, "dur": 4,
    "args": {
      "External id": 136055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136055,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 136055, "pid": 5, "tid": 7, "ts": 1716454223509619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429794, "dur": 5,
    "args": {
      "External id": 136055,
      "cbid": 211, "correlation": 136055
    }
  },
  {
    "ph": "s", "id": 136055, "pid": 76337, "tid": -914061504, "ts": 1716454223429794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223509624, "dur": 90,
    "args": {
      "External id": 136076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136076,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 136076, "pid": 5, "tid": 7, "ts": 1716454223509624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429866, "dur": 12,
    "args": {
      "External id": 136076,
      "cbid": 211, "correlation": 136076
    }
  },
  {
    "ph": "s", "id": 136076, "pid": 76337, "tid": -914061504, "ts": 1716454223429866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429963, "dur": 1,
    "args": {
      "External id": 136094,
      "cbid": 251, "correlation": 136094
    }
  },
  {
    "ph": "f", "id": 136094, "pid": 76337, "tid": -914061504, "ts": 1716454223429963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223509716, "dur": 97,
    "args": {
      "External id": 136096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136096,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136096, "pid": 5, "tid": 7, "ts": 1716454223509716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223429969, "dur": 20,
    "args": {
      "External id": 136096,
      "cbid": 211, "correlation": 136096
    }
  },
  {
    "ph": "s", "id": 136096, "pid": 76337, "tid": -914061504, "ts": 1716454223429969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223509814, "dur": 19,
    "args": {
      "External id": 136104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136104,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136104, "pid": 5, "tid": 7, "ts": 1716454223509814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430045, "dur": 12,
    "args": {
      "External id": 136104,
      "cbid": 211, "correlation": 136104
    }
  },
  {
    "ph": "s", "id": 136104, "pid": 76337, "tid": -914061504, "ts": 1716454223430045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223509834, "dur": 38,
    "args": {
      "External id": 136112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136112,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136112, "pid": 5, "tid": 7, "ts": 1716454223509834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430087, "dur": 10,
    "args": {
      "External id": 136112,
      "cbid": 211, "correlation": 136112
    }
  },
  {
    "ph": "s", "id": 136112, "pid": 76337, "tid": -914061504, "ts": 1716454223430087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223509874, "dur": 34,
    "args": {
      "External id": 136134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136134,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136134, "pid": 5, "tid": 7, "ts": 1716454223509874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430137, "dur": 11,
    "args": {
      "External id": 136134,
      "cbid": 211, "correlation": 136134
    }
  },
  {
    "ph": "s", "id": 136134, "pid": 76337, "tid": -914061504, "ts": 1716454223430137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430226, "dur": 1,
    "args": {
      "External id": 136150,
      "cbid": 251, "correlation": 136150
    }
  },
  {
    "ph": "f", "id": 136150, "pid": 76337, "tid": -914061504, "ts": 1716454223430226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430231, "dur": 0,
    "args": {
      "External id": 136152,
      "cbid": 251, "correlation": 136152
    }
  },
  {
    "ph": "f", "id": 136152, "pid": 76337, "tid": -914061504, "ts": 1716454223430231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223509909, "dur": 532,
    "args": {
      "External id": 136153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136153,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 136153, "pid": 5, "tid": 7, "ts": 1716454223509909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430234, "dur": 13,
    "args": {
      "External id": 136153,
      "cbid": 211, "correlation": 136153
    }
  },
  {
    "ph": "s", "id": 136153, "pid": 76337, "tid": -914061504, "ts": 1716454223430234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223510443, "dur": 124,
    "args": {
      "External id": 136161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136161, "pid": 5, "tid": 7, "ts": 1716454223510443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430299, "dur": 12,
    "args": {
      "External id": 136161,
      "cbid": 211, "correlation": 136161
    }
  },
  {
    "ph": "s", "id": 136161, "pid": 76337, "tid": -914061504, "ts": 1716454223430299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223510569, "dur": 129,
    "args": {
      "External id": 136169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136169, "pid": 5, "tid": 7, "ts": 1716454223510569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430328, "dur": 8,
    "args": {
      "External id": 136169,
      "cbid": 211, "correlation": 136169
    }
  },
  {
    "ph": "s", "id": 136169, "pid": 76337, "tid": -914061504, "ts": 1716454223430328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430405, "dur": 1,
    "args": {
      "External id": 136185,
      "cbid": 251, "correlation": 136185
    }
  },
  {
    "ph": "f", "id": 136185, "pid": 76337, "tid": -914061504, "ts": 1716454223430405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223510700, "dur": 303,
    "args": {
      "External id": 136187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136187,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136187, "pid": 5, "tid": 7, "ts": 1716454223510700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430411, "dur": 12,
    "args": {
      "External id": 136187,
      "cbid": 211, "correlation": 136187
    }
  },
  {
    "ph": "s", "id": 136187, "pid": 76337, "tid": -914061504, "ts": 1716454223430411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223511004, "dur": 27,
    "args": {
      "External id": 136195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136195,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136195, "pid": 5, "tid": 7, "ts": 1716454223511004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430452, "dur": 10,
    "args": {
      "External id": 136195,
      "cbid": 211, "correlation": 136195
    }
  },
  {
    "ph": "s", "id": 136195, "pid": 76337, "tid": -914061504, "ts": 1716454223430452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223511032, "dur": 80,
    "args": {
      "External id": 136206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136206,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136206, "pid": 5, "tid": 7, "ts": 1716454223511032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430519, "dur": 12,
    "args": {
      "External id": 136206,
      "cbid": 211, "correlation": 136206
    }
  },
  {
    "ph": "s", "id": 136206, "pid": 76337, "tid": -914061504, "ts": 1716454223430519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430582, "dur": 0,
    "args": {
      "External id": 136218,
      "cbid": 317, "correlation": 136218
    }
  },
  {
    "ph": "f", "id": 136218, "pid": 76337, "tid": -914061504, "ts": 1716454223430582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430583, "dur": 0,
    "args": {
      "External id": 136219,
      "cbid": 203, "correlation": 136219
    }
  },
  {
    "ph": "f", "id": 136219, "pid": 76337, "tid": -914061504, "ts": 1716454223430583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430584, "dur": 0,
    "args": {
      "External id": 136220,
      "cbid": 205, "correlation": 136220
    }
  },
  {
    "ph": "f", "id": 136220, "pid": 76337, "tid": -914061504, "ts": 1716454223430584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223511113, "dur": 24,
    "args": {
      "External id": 136224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136224,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136224, "pid": 5, "tid": 7, "ts": 1716454223511113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430599, "dur": 12,
    "args": {
      "External id": 136224,
      "cbid": 211, "correlation": 136224
    }
  },
  {
    "ph": "s", "id": 136224, "pid": 76337, "tid": -914061504, "ts": 1716454223430599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223511138, "dur": 119,
    "args": {
      "External id": 136226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136226,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136226, "pid": 5, "tid": 7, "ts": 1716454223511138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430618, "dur": 6,
    "args": {
      "External id": 136226,
      "cbid": 211, "correlation": 136226
    }
  },
  {
    "ph": "s", "id": 136226, "pid": 76337, "tid": -914061504, "ts": 1716454223430618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223511258, "dur": 23,
    "args": {
      "External id": 136228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136228,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136228, "pid": 5, "tid": 7, "ts": 1716454223511258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430628, "dur": 5,
    "args": {
      "External id": 136228,
      "cbid": 211, "correlation": 136228
    }
  },
  {
    "ph": "s", "id": 136228, "pid": 76337, "tid": -914061504, "ts": 1716454223430628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223511283, "dur": 33,
    "args": {
      "External id": 136234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136234, "pid": 5, "tid": 7, "ts": 1716454223511283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430656, "dur": 9,
    "args": {
      "External id": 136234,
      "cbid": 211, "correlation": 136234
    }
  },
  {
    "ph": "s", "id": 136234, "pid": 76337, "tid": -914061504, "ts": 1716454223430656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223511316, "dur": 27,
    "args": {
      "External id": 136242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136242,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136242, "pid": 5, "tid": 7, "ts": 1716454223511316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430687, "dur": 9,
    "args": {
      "External id": 136242,
      "cbid": 211, "correlation": 136242
    }
  },
  {
    "ph": "s", "id": 136242, "pid": 76337, "tid": -914061504, "ts": 1716454223430687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223511344, "dur": 101,
    "args": {
      "External id": 136253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136253,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136253, "pid": 5, "tid": 7, "ts": 1716454223511344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430749, "dur": 11,
    "args": {
      "External id": 136253,
      "cbid": 211, "correlation": 136253
    }
  },
  {
    "ph": "s", "id": 136253, "pid": 76337, "tid": -914061504, "ts": 1716454223430749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430805, "dur": 0,
    "args": {
      "External id": 136263,
      "cbid": 317, "correlation": 136263
    }
  },
  {
    "ph": "f", "id": 136263, "pid": 76337, "tid": -914061504, "ts": 1716454223430805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430805, "dur": 0,
    "args": {
      "External id": 136264,
      "cbid": 203, "correlation": 136264
    }
  },
  {
    "ph": "f", "id": 136264, "pid": 76337, "tid": -914061504, "ts": 1716454223430805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430806, "dur": 0,
    "args": {
      "External id": 136265,
      "cbid": 205, "correlation": 136265
    }
  },
  {
    "ph": "f", "id": 136265, "pid": 76337, "tid": -914061504, "ts": 1716454223430806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223511446, "dur": 74,
    "args": {
      "External id": 136269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136269,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136269, "pid": 5, "tid": 7, "ts": 1716454223511446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430820, "dur": 11,
    "args": {
      "External id": 136269,
      "cbid": 211, "correlation": 136269
    }
  },
  {
    "ph": "s", "id": 136269, "pid": 76337, "tid": -914061504, "ts": 1716454223430820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223511522, "dur": 43,
    "args": {
      "External id": 136271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136271, "pid": 5, "tid": 7, "ts": 1716454223511522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430834, "dur": 5,
    "args": {
      "External id": 136271,
      "cbid": 211, "correlation": 136271
    }
  },
  {
    "ph": "s", "id": 136271, "pid": 76337, "tid": -914061504, "ts": 1716454223430834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223511567, "dur": 4,
    "args": {
      "External id": 136273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136273,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136273, "pid": 5, "tid": 7, "ts": 1716454223511567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430845, "dur": 6,
    "args": {
      "External id": 136273,
      "cbid": 211, "correlation": 136273
    }
  },
  {
    "ph": "s", "id": 136273, "pid": 76337, "tid": -914061504, "ts": 1716454223430845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430854, "dur": 0,
    "args": {
      "External id": 136274,
      "cbid": 51, "correlation": 136274
    }
  },
  {
    "ph": "s", "id": 136274, "pid": 76337, "tid": -914061504, "ts": 1716454223430854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223511572, "dur": 2226,
    "args": {
      "External id": 136275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136275,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136275, "pid": 5, "tid": 7, "ts": 1716454223511572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430855, "dur": 5,
    "args": {
      "External id": 136275,
      "cbid": 211, "correlation": 136275
    }
  },
  {
    "ph": "s", "id": 136275, "pid": 76337, "tid": -914061504, "ts": 1716454223430855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223513799, "dur": 112,
    "args": {
      "External id": 136280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136280,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136280, "pid": 5, "tid": 7, "ts": 1716454223513799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430883, "dur": 8,
    "args": {
      "External id": 136280,
      "cbid": 211, "correlation": 136280
    }
  },
  {
    "ph": "s", "id": 136280, "pid": 76337, "tid": -914061504, "ts": 1716454223430883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223513912, "dur": 168,
    "args": {
      "External id": 136289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136289,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136289, "pid": 5, "tid": 7, "ts": 1716454223513912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223430984, "dur": 15,
    "args": {
      "External id": 136289,
      "cbid": 211, "correlation": 136289
    }
  },
  {
    "ph": "s", "id": 136289, "pid": 76337, "tid": -914061504, "ts": 1716454223430984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223514082, "dur": 129,
    "args": {
      "External id": 136309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136309,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 136309, "pid": 5, "tid": 7, "ts": 1716454223514082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431057, "dur": 11,
    "args": {
      "External id": 136309,
      "cbid": 211, "correlation": 136309
    }
  },
  {
    "ph": "s", "id": 136309, "pid": 76337, "tid": -914061504, "ts": 1716454223431057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223514212, "dur": 5,
    "args": {
      "External id": 136321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136321,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 136321, "pid": 5, "tid": 7, "ts": 1716454223514212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431078, "dur": 6,
    "args": {
      "External id": 136321,
      "cbid": 211, "correlation": 136321
    }
  },
  {
    "ph": "s", "id": 136321, "pid": 76337, "tid": -914061504, "ts": 1716454223431078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223514218, "dur": 156,
    "args": {
      "External id": 136324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136324, "pid": 5, "tid": 7, "ts": 1716454223514218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431097, "dur": 7,
    "args": {
      "External id": 136324,
      "cbid": 211, "correlation": 136324
    }
  },
  {
    "ph": "s", "id": 136324, "pid": 76337, "tid": -914061504, "ts": 1716454223431097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223514376, "dur": 101,
    "args": {
      "External id": 136333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136333,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136333, "pid": 5, "tid": 7, "ts": 1716454223514376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431136, "dur": 10,
    "args": {
      "External id": 136333,
      "cbid": 211, "correlation": 136333
    }
  },
  {
    "ph": "s", "id": 136333, "pid": 76337, "tid": -914061504, "ts": 1716454223431136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431189, "dur": 0,
    "args": {
      "External id": 136343,
      "cbid": 317, "correlation": 136343
    }
  },
  {
    "ph": "f", "id": 136343, "pid": 76337, "tid": -914061504, "ts": 1716454223431189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431190, "dur": 0,
    "args": {
      "External id": 136344,
      "cbid": 203, "correlation": 136344
    }
  },
  {
    "ph": "f", "id": 136344, "pid": 76337, "tid": -914061504, "ts": 1716454223431190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431191, "dur": 0,
    "args": {
      "External id": 136345,
      "cbid": 205, "correlation": 136345
    }
  },
  {
    "ph": "f", "id": 136345, "pid": 76337, "tid": -914061504, "ts": 1716454223431191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223514478, "dur": 110,
    "args": {
      "External id": 136349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136349, "pid": 5, "tid": 7, "ts": 1716454223514478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431207, "dur": 11,
    "args": {
      "External id": 136349,
      "cbid": 211, "correlation": 136349
    }
  },
  {
    "ph": "s", "id": 136349, "pid": 76337, "tid": -914061504, "ts": 1716454223431207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223514589, "dur": 34,
    "args": {
      "External id": 136351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136351,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136351, "pid": 5, "tid": 7, "ts": 1716454223514589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431221, "dur": 5,
    "args": {
      "External id": 136351,
      "cbid": 211, "correlation": 136351
    }
  },
  {
    "ph": "s", "id": 136351, "pid": 76337, "tid": -914061504, "ts": 1716454223431221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223514624, "dur": 4,
    "args": {
      "External id": 136353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136353, "pid": 5, "tid": 7, "ts": 1716454223514624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431231, "dur": 6,
    "args": {
      "External id": 136353,
      "cbid": 211, "correlation": 136353
    }
  },
  {
    "ph": "s", "id": 136353, "pid": 76337, "tid": -914061504, "ts": 1716454223431231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431240, "dur": 0,
    "args": {
      "External id": 136354,
      "cbid": 51, "correlation": 136354
    }
  },
  {
    "ph": "s", "id": 136354, "pid": 76337, "tid": -914061504, "ts": 1716454223431240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223514629, "dur": 1987,
    "args": {
      "External id": 136355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136355,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136355, "pid": 5, "tid": 7, "ts": 1716454223514629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431241, "dur": 6,
    "args": {
      "External id": 136355,
      "cbid": 211, "correlation": 136355
    }
  },
  {
    "ph": "s", "id": 136355, "pid": 76337, "tid": -914061504, "ts": 1716454223431241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223516617, "dur": 59,
    "args": {
      "External id": 136360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136360, "pid": 5, "tid": 7, "ts": 1716454223516617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431270, "dur": 8,
    "args": {
      "External id": 136360,
      "cbid": 211, "correlation": 136360
    }
  },
  {
    "ph": "s", "id": 136360, "pid": 76337, "tid": -914061504, "ts": 1716454223431270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223516677, "dur": 3,
    "args": {
      "External id": 136368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136368,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136368, "pid": 5, "tid": 7, "ts": 1716454223516677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431314, "dur": 10,
    "args": {
      "External id": 136368,
      "cbid": 211, "correlation": 136368
    }
  },
  {
    "ph": "s", "id": 136368, "pid": 76337, "tid": -914061504, "ts": 1716454223431314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431379, "dur": 1,
    "args": {
      "External id": 136384,
      "cbid": 251, "correlation": 136384
    }
  },
  {
    "ph": "f", "id": 136384, "pid": 76337, "tid": -914061504, "ts": 1716454223431379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431384, "dur": 0,
    "args": {
      "External id": 136386,
      "cbid": 251, "correlation": 136386
    }
  },
  {
    "ph": "f", "id": 136386, "pid": 76337, "tid": -914061504, "ts": 1716454223431384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223516682, "dur": 11,
    "args": {
      "External id": 136387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136387,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 136387, "pid": 5, "tid": 7, "ts": 1716454223516682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431386, "dur": 12,
    "args": {
      "External id": 136387,
      "cbid": 211, "correlation": 136387
    }
  },
  {
    "ph": "s", "id": 136387, "pid": 76337, "tid": -914061504, "ts": 1716454223431386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223516694, "dur": 5,
    "args": {
      "External id": 136389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136389,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 136389, "pid": 5, "tid": 7, "ts": 1716454223516694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431400, "dur": 6,
    "args": {
      "External id": 136389,
      "cbid": 211, "correlation": 136389
    }
  },
  {
    "ph": "s", "id": 136389, "pid": 76337, "tid": -914061504, "ts": 1716454223431400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223516701, "dur": 54,
    "args": {
      "External id": 136399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136399,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136399, "pid": 5, "tid": 7, "ts": 1716454223516701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431458, "dur": 12,
    "args": {
      "External id": 136399,
      "cbid": 211, "correlation": 136399
    }
  },
  {
    "ph": "s", "id": 136399, "pid": 76337, "tid": -914061504, "ts": 1716454223431458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223516756, "dur": 49,
    "args": {
      "External id": 136419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136419,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 136419, "pid": 5, "tid": 7, "ts": 1716454223516756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431524, "dur": 11,
    "args": {
      "External id": 136419,
      "cbid": 211, "correlation": 136419
    }
  },
  {
    "ph": "s", "id": 136419, "pid": 76337, "tid": -914061504, "ts": 1716454223431524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223516807, "dur": 4,
    "args": {
      "External id": 136431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136431,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136431, "pid": 5, "tid": 7, "ts": 1716454223516807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431544, "dur": 7,
    "args": {
      "External id": 136431,
      "cbid": 211, "correlation": 136431
    }
  },
  {
    "ph": "s", "id": 136431, "pid": 76337, "tid": -914061504, "ts": 1716454223431544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223516812, "dur": 55,
    "args": {
      "External id": 136434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136434,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136434, "pid": 5, "tid": 7, "ts": 1716454223516812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431563, "dur": 6,
    "args": {
      "External id": 136434,
      "cbid": 211, "correlation": 136434
    }
  },
  {
    "ph": "s", "id": 136434, "pid": 76337, "tid": -914061504, "ts": 1716454223431563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223516869, "dur": 36,
    "args": {
      "External id": 136443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136443,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136443, "pid": 5, "tid": 7, "ts": 1716454223516869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431604, "dur": 10,
    "args": {
      "External id": 136443,
      "cbid": 211, "correlation": 136443
    }
  },
  {
    "ph": "s", "id": 136443, "pid": 76337, "tid": -914061504, "ts": 1716454223431604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431666, "dur": 0,
    "args": {
      "External id": 136453,
      "cbid": 317, "correlation": 136453
    }
  },
  {
    "ph": "f", "id": 136453, "pid": 76337, "tid": -914061504, "ts": 1716454223431666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431667, "dur": 0,
    "args": {
      "External id": 136454,
      "cbid": 203, "correlation": 136454
    }
  },
  {
    "ph": "f", "id": 136454, "pid": 76337, "tid": -914061504, "ts": 1716454223431667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431668, "dur": 0,
    "args": {
      "External id": 136455,
      "cbid": 205, "correlation": 136455
    }
  },
  {
    "ph": "f", "id": 136455, "pid": 76337, "tid": -914061504, "ts": 1716454223431668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223516906, "dur": 40,
    "args": {
      "External id": 136459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136459,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136459, "pid": 5, "tid": 7, "ts": 1716454223516906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431684, "dur": 12,
    "args": {
      "External id": 136459,
      "cbid": 211, "correlation": 136459
    }
  },
  {
    "ph": "s", "id": 136459, "pid": 76337, "tid": -914061504, "ts": 1716454223431684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223516947, "dur": 14,
    "args": {
      "External id": 136461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136461,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136461, "pid": 5, "tid": 7, "ts": 1716454223516947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431699, "dur": 6,
    "args": {
      "External id": 136461,
      "cbid": 211, "correlation": 136461
    }
  },
  {
    "ph": "s", "id": 136461, "pid": 76337, "tid": -914061504, "ts": 1716454223431699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223516963, "dur": 3,
    "args": {
      "External id": 136463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136463, "pid": 5, "tid": 7, "ts": 1716454223516963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431709, "dur": 5,
    "args": {
      "External id": 136463,
      "cbid": 211, "correlation": 136463
    }
  },
  {
    "ph": "s", "id": 136463, "pid": 76337, "tid": -914061504, "ts": 1716454223431709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431718, "dur": 0,
    "args": {
      "External id": 136464,
      "cbid": 51, "correlation": 136464
    }
  },
  {
    "ph": "s", "id": 136464, "pid": 76337, "tid": -914061504, "ts": 1716454223431718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223516968, "dur": 692,
    "args": {
      "External id": 136465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136465,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136465, "pid": 5, "tid": 7, "ts": 1716454223516968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431719, "dur": 5,
    "args": {
      "External id": 136465,
      "cbid": 211, "correlation": 136465
    }
  },
  {
    "ph": "s", "id": 136465, "pid": 76337, "tid": -914061504, "ts": 1716454223431719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223517661, "dur": 59,
    "args": {
      "External id": 136470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136470,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136470, "pid": 5, "tid": 7, "ts": 1716454223517661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431746, "dur": 10,
    "args": {
      "External id": 136470,
      "cbid": 211, "correlation": 136470
    }
  },
  {
    "ph": "s", "id": 136470, "pid": 76337, "tid": -914061504, "ts": 1716454223431746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431805, "dur": 0,
    "args": {
      "External id": 136480,
      "cbid": 317, "correlation": 136480
    }
  },
  {
    "ph": "f", "id": 136480, "pid": 76337, "tid": -914061504, "ts": 1716454223431805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431806, "dur": 0,
    "args": {
      "External id": 136481,
      "cbid": 203, "correlation": 136481
    }
  },
  {
    "ph": "f", "id": 136481, "pid": 76337, "tid": -914061504, "ts": 1716454223431806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431806, "dur": 0,
    "args": {
      "External id": 136482,
      "cbid": 205, "correlation": 136482
    }
  },
  {
    "ph": "f", "id": 136482, "pid": 76337, "tid": -914061504, "ts": 1716454223431806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223517721, "dur": 3,
    "args": {
      "External id": 136486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136486, "pid": 5, "tid": 7, "ts": 1716454223517721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431821, "dur": 12,
    "args": {
      "External id": 136486,
      "cbid": 211, "correlation": 136486
    }
  },
  {
    "ph": "s", "id": 136486, "pid": 76337, "tid": -914061504, "ts": 1716454223431821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431837, "dur": 0,
    "args": {
      "External id": 136487,
      "cbid": 51, "correlation": 136487
    }
  },
  {
    "ph": "s", "id": 136487, "pid": 76337, "tid": -914061504, "ts": 1716454223431837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223517725, "dur": 262,
    "args": {
      "External id": 136488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136488,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136488, "pid": 5, "tid": 7, "ts": 1716454223517725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431838, "dur": 7,
    "args": {
      "External id": 136488,
      "cbid": 211, "correlation": 136488
    }
  },
  {
    "ph": "s", "id": 136488, "pid": 76337, "tid": -914061504, "ts": 1716454223431838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223517989, "dur": 58,
    "args": {
      "External id": 136493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136493,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136493, "pid": 5, "tid": 7, "ts": 1716454223517989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431866, "dur": 8,
    "args": {
      "External id": 136493,
      "cbid": 211, "correlation": 136493
    }
  },
  {
    "ph": "s", "id": 136493, "pid": 76337, "tid": -914061504, "ts": 1716454223431866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223518049, "dur": 50,
    "args": {
      "External id": 136501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136501,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136501, "pid": 5, "tid": 7, "ts": 1716454223518049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431894, "dur": 8,
    "args": {
      "External id": 136501,
      "cbid": 211, "correlation": 136501
    }
  },
  {
    "ph": "s", "id": 136501, "pid": 76337, "tid": -914061504, "ts": 1716454223431894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223518099, "dur": 35,
    "args": {
      "External id": 136509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136509,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136509, "pid": 5, "tid": 7, "ts": 1716454223518099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223431923, "dur": 8,
    "args": {
      "External id": 136509,
      "cbid": 211, "correlation": 136509
    }
  },
  {
    "ph": "s", "id": 136509, "pid": 76337, "tid": -914061504, "ts": 1716454223431923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223518136, "dur": 53,
    "args": {
      "External id": 136529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136529,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 136529, "pid": 5, "tid": 7, "ts": 1716454223518136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432018, "dur": 13,
    "args": {
      "External id": 136529,
      "cbid": 211, "correlation": 136529
    }
  },
  {
    "ph": "s", "id": 136529, "pid": 76337, "tid": -914061504, "ts": 1716454223432018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223518190, "dur": 4,
    "args": {
      "External id": 136541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136541,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 136541, "pid": 5, "tid": 7, "ts": 1716454223518190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432040, "dur": 6,
    "args": {
      "External id": 136541,
      "cbid": 211, "correlation": 136541
    }
  },
  {
    "ph": "s", "id": 136541, "pid": 76337, "tid": -914061504, "ts": 1716454223432040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223518195, "dur": 54,
    "args": {
      "External id": 136544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136544,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136544, "pid": 5, "tid": 7, "ts": 1716454223518195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432059, "dur": 7,
    "args": {
      "External id": 136544,
      "cbid": 211, "correlation": 136544
    }
  },
  {
    "ph": "s", "id": 136544, "pid": 76337, "tid": -914061504, "ts": 1716454223432059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432117, "dur": 0,
    "args": {
      "External id": 136555,
      "cbid": 317, "correlation": 136555
    }
  },
  {
    "ph": "f", "id": 136555, "pid": 76337, "tid": -914061504, "ts": 1716454223432117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432117, "dur": 0,
    "args": {
      "External id": 136556,
      "cbid": 203, "correlation": 136556
    }
  },
  {
    "ph": "f", "id": 136556, "pid": 76337, "tid": -914061504, "ts": 1716454223432117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432118, "dur": 0,
    "args": {
      "External id": 136557,
      "cbid": 205, "correlation": 136557
    }
  },
  {
    "ph": "f", "id": 136557, "pid": 76337, "tid": -914061504, "ts": 1716454223432118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432146, "dur": 2,
    "args": {
      "External id": 136561,
      "cbid": 251, "correlation": 136561
    }
  },
  {
    "ph": "f", "id": 136561, "pid": 76337, "tid": -914061504, "ts": 1716454223432146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432149, "dur": 1,
    "args": {
      "External id": 136562,
      "cbid": 251, "correlation": 136562
    }
  },
  {
    "ph": "f", "id": 136562, "pid": 76337, "tid": -914061504, "ts": 1716454223432149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432151, "dur": 1,
    "args": {
      "External id": 136563,
      "cbid": 251, "correlation": 136563
    }
  },
  {
    "ph": "f", "id": 136563, "pid": 76337, "tid": -914061504, "ts": 1716454223432151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432153, "dur": 1,
    "args": {
      "External id": 136564,
      "cbid": 251, "correlation": 136564
    }
  },
  {
    "ph": "f", "id": 136564, "pid": 76337, "tid": -914061504, "ts": 1716454223432153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432155, "dur": 1,
    "args": {
      "External id": 136565,
      "cbid": 251, "correlation": 136565
    }
  },
  {
    "ph": "f", "id": 136565, "pid": 76337, "tid": -914061504, "ts": 1716454223432155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432156, "dur": 1,
    "args": {
      "External id": 136566,
      "cbid": 251, "correlation": 136566
    }
  },
  {
    "ph": "f", "id": 136566, "pid": 76337, "tid": -914061504, "ts": 1716454223432156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432158, "dur": 0,
    "args": {
      "External id": 136567,
      "cbid": 251, "correlation": 136567
    }
  },
  {
    "ph": "f", "id": 136567, "pid": 76337, "tid": -914061504, "ts": 1716454223432158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432159, "dur": 1,
    "args": {
      "External id": 136568,
      "cbid": 251, "correlation": 136568
    }
  },
  {
    "ph": "f", "id": 136568, "pid": 76337, "tid": -914061504, "ts": 1716454223432159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432161, "dur": 0,
    "args": {
      "External id": 136569,
      "cbid": 251, "correlation": 136569
    }
  },
  {
    "ph": "f", "id": 136569, "pid": 76337, "tid": -914061504, "ts": 1716454223432161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223518251, "dur": 113,
    "args": {
      "External id": 136570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136570,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 136570, "pid": 5, "tid": 7, "ts": 1716454223518251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432163, "dur": 12,
    "args": {
      "External id": 136570,
      "cbid": 211, "correlation": 136570
    }
  },
  {
    "ph": "s", "id": 136570, "pid": 76337, "tid": -914061504, "ts": 1716454223432163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223518365, "dur": 59,
    "args": {
      "External id": 136576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136576, "pid": 5, "tid": 7, "ts": 1716454223518365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432198, "dur": 10,
    "args": {
      "External id": 136576,
      "cbid": 211, "correlation": 136576
    }
  },
  {
    "ph": "s", "id": 136576, "pid": 76337, "tid": -914061504, "ts": 1716454223432198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223518425, "dur": 609,
    "args": {
      "External id": 136585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136585,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136585, "pid": 5, "tid": 7, "ts": 1716454223518425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432281, "dur": 13,
    "args": {
      "External id": 136585,
      "cbid": 211, "correlation": 136585
    }
  },
  {
    "ph": "s", "id": 136585, "pid": 76337, "tid": -914061504, "ts": 1716454223432281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223519035, "dur": 179,
    "args": {
      "External id": 136607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136607,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136607, "pid": 5, "tid": 7, "ts": 1716454223519035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432339, "dur": 10,
    "args": {
      "External id": 136607,
      "cbid": 211, "correlation": 136607
    }
  },
  {
    "ph": "s", "id": 136607, "pid": 76337, "tid": -914061504, "ts": 1716454223432339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432428, "dur": 1,
    "args": {
      "External id": 136618,
      "cbid": 251, "correlation": 136618
    }
  },
  {
    "ph": "f", "id": 136618, "pid": 76337, "tid": -914061504, "ts": 1716454223432428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223519215, "dur": 196,
    "args": {
      "External id": 136619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136619,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136619, "pid": 5, "tid": 7, "ts": 1716454223519215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432433, "dur": 12,
    "args": {
      "External id": 136619,
      "cbid": 211, "correlation": 136619
    }
  },
  {
    "ph": "s", "id": 136619, "pid": 76337, "tid": -914061504, "ts": 1716454223432433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432502, "dur": 1,
    "args": {
      "External id": 136630,
      "cbid": 251, "correlation": 136630
    }
  },
  {
    "ph": "f", "id": 136630, "pid": 76337, "tid": -914061504, "ts": 1716454223432502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223519412, "dur": 185,
    "args": {
      "External id": 136631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136631,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136631, "pid": 5, "tid": 7, "ts": 1716454223519412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432506, "dur": 12,
    "args": {
      "External id": 136631,
      "cbid": 211, "correlation": 136631
    }
  },
  {
    "ph": "s", "id": 136631, "pid": 76337, "tid": -914061504, "ts": 1716454223432506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432569, "dur": 1,
    "args": {
      "External id": 136642,
      "cbid": 251, "correlation": 136642
    }
  },
  {
    "ph": "f", "id": 136642, "pid": 76337, "tid": -914061504, "ts": 1716454223432569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223519599, "dur": 187,
    "args": {
      "External id": 136643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136643,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136643, "pid": 5, "tid": 7, "ts": 1716454223519599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432573, "dur": 12,
    "args": {
      "External id": 136643,
      "cbid": 211, "correlation": 136643
    }
  },
  {
    "ph": "s", "id": 136643, "pid": 76337, "tid": -914061504, "ts": 1716454223432573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223519787, "dur": 18362,
    "args": {
      "External id": 136664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136664,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 136664, "pid": 5, "tid": 7, "ts": 1716454223519787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432655, "dur": 14,
    "args": {
      "External id": 136664,
      "cbid": 211, "correlation": 136664
    }
  },
  {
    "ph": "s", "id": 136664, "pid": 76337, "tid": -914061504, "ts": 1716454223432655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432753, "dur": 1,
    "args": {
      "External id": 136682,
      "cbid": 251, "correlation": 136682
    }
  },
  {
    "ph": "f", "id": 136682, "pid": 76337, "tid": -914061504, "ts": 1716454223432753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223538150, "dur": 200,
    "args": {
      "External id": 136684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136684,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136684, "pid": 5, "tid": 7, "ts": 1716454223538150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432758, "dur": 14,
    "args": {
      "External id": 136684,
      "cbid": 211, "correlation": 136684
    }
  },
  {
    "ph": "s", "id": 136684, "pid": 76337, "tid": -914061504, "ts": 1716454223432758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223538352, "dur": 67,
    "args": {
      "External id": 136692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136692,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136692, "pid": 5, "tid": 7, "ts": 1716454223538352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432829, "dur": 12,
    "args": {
      "External id": 136692,
      "cbid": 211, "correlation": 136692
    }
  },
  {
    "ph": "s", "id": 136692, "pid": 76337, "tid": -914061504, "ts": 1716454223432829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223538420, "dur": 97,
    "args": {
      "External id": 136700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136700,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136700, "pid": 5, "tid": 7, "ts": 1716454223538420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432868, "dur": 8,
    "args": {
      "External id": 136700,
      "cbid": 211, "correlation": 136700
    }
  },
  {
    "ph": "s", "id": 136700, "pid": 76337, "tid": -914061504, "ts": 1716454223432868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223538518, "dur": 55,
    "args": {
      "External id": 136711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136711, "pid": 5, "tid": 7, "ts": 1716454223538518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432939, "dur": 12,
    "args": {
      "External id": 136711,
      "cbid": 211, "correlation": 136711
    }
  },
  {
    "ph": "s", "id": 136711, "pid": 76337, "tid": -914061504, "ts": 1716454223432939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223538574, "dur": 91,
    "args": {
      "External id": 136733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136733,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136733, "pid": 5, "tid": 7, "ts": 1716454223538574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223432971, "dur": 18,
    "args": {
      "External id": 136733,
      "cbid": 211, "correlation": 136733
    }
  },
  {
    "ph": "s", "id": 136733, "pid": 76337, "tid": -914061504, "ts": 1716454223432971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433065, "dur": 1,
    "args": {
      "External id": 136744,
      "cbid": 251, "correlation": 136744
    }
  },
  {
    "ph": "f", "id": 136744, "pid": 76337, "tid": -914061504, "ts": 1716454223433065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223538666, "dur": 105,
    "args": {
      "External id": 136745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136745,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136745, "pid": 5, "tid": 7, "ts": 1716454223538666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433070, "dur": 13,
    "args": {
      "External id": 136745,
      "cbid": 211, "correlation": 136745
    }
  },
  {
    "ph": "s", "id": 136745, "pid": 76337, "tid": -914061504, "ts": 1716454223433070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433144, "dur": 1,
    "args": {
      "External id": 136756,
      "cbid": 251, "correlation": 136756
    }
  },
  {
    "ph": "f", "id": 136756, "pid": 76337, "tid": -914061504, "ts": 1716454223433144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433148, "dur": 0,
    "args": {
      "External id": 136757,
      "cbid": 251, "correlation": 136757
    }
  },
  {
    "ph": "f", "id": 136757, "pid": 76337, "tid": -914061504, "ts": 1716454223433148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223538772, "dur": 11,
    "args": {
      "External id": 136758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136758,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 136758, "pid": 5, "tid": 7, "ts": 1716454223538772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433150, "dur": 13,
    "args": {
      "External id": 136758,
      "cbid": 211, "correlation": 136758
    }
  },
  {
    "ph": "s", "id": 136758, "pid": 76337, "tid": -914061504, "ts": 1716454223433150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223538784, "dur": 5,
    "args": {
      "External id": 136760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136760,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 136760, "pid": 5, "tid": 7, "ts": 1716454223538784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433166, "dur": 8,
    "args": {
      "External id": 136760,
      "cbid": 211, "correlation": 136760
    }
  },
  {
    "ph": "s", "id": 136760, "pid": 76337, "tid": -914061504, "ts": 1716454223433166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433228, "dur": 1,
    "args": {
      "External id": 136771,
      "cbid": 251, "correlation": 136771
    }
  },
  {
    "ph": "f", "id": 136771, "pid": 76337, "tid": -914061504, "ts": 1716454223433228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433232, "dur": 0,
    "args": {
      "External id": 136772,
      "cbid": 251, "correlation": 136772
    }
  },
  {
    "ph": "f", "id": 136772, "pid": 76337, "tid": -914061504, "ts": 1716454223433232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223538791, "dur": 6,
    "args": {
      "External id": 136773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136773,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 136773, "pid": 5, "tid": 7, "ts": 1716454223538791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433233, "dur": 11,
    "args": {
      "External id": 136773,
      "cbid": 211, "correlation": 136773
    }
  },
  {
    "ph": "s", "id": 136773, "pid": 76337, "tid": -914061504, "ts": 1716454223433233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223538798, "dur": 3,
    "args": {
      "External id": 136775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136775,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 136775, "pid": 5, "tid": 7, "ts": 1716454223538798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433246, "dur": 5,
    "args": {
      "External id": 136775,
      "cbid": 211, "correlation": 136775
    }
  },
  {
    "ph": "s", "id": 136775, "pid": 76337, "tid": -914061504, "ts": 1716454223433246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223538803, "dur": 155,
    "args": {
      "External id": 136796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136796,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 136796, "pid": 5, "tid": 7, "ts": 1716454223538803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433320, "dur": 13,
    "args": {
      "External id": 136796,
      "cbid": 211, "correlation": 136796
    }
  },
  {
    "ph": "s", "id": 136796, "pid": 76337, "tid": -914061504, "ts": 1716454223433320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433414, "dur": 1,
    "args": {
      "External id": 136814,
      "cbid": 251, "correlation": 136814
    }
  },
  {
    "ph": "f", "id": 136814, "pid": 76337, "tid": -914061504, "ts": 1716454223433414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223538959, "dur": 104,
    "args": {
      "External id": 136816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136816,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 136816, "pid": 5, "tid": 7, "ts": 1716454223538959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433420, "dur": 14,
    "args": {
      "External id": 136816,
      "cbid": 211, "correlation": 136816
    }
  },
  {
    "ph": "s", "id": 136816, "pid": 76337, "tid": -914061504, "ts": 1716454223433420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223539065, "dur": 35,
    "args": {
      "External id": 136824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136824,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136824, "pid": 5, "tid": 7, "ts": 1716454223539065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433490, "dur": 12,
    "args": {
      "External id": 136824,
      "cbid": 211, "correlation": 136824
    }
  },
  {
    "ph": "s", "id": 136824, "pid": 76337, "tid": -914061504, "ts": 1716454223433490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223539101, "dur": 68,
    "args": {
      "External id": 136832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136832, "pid": 5, "tid": 7, "ts": 1716454223539101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433530, "dur": 9,
    "args": {
      "External id": 136832,
      "cbid": 211, "correlation": 136832
    }
  },
  {
    "ph": "s", "id": 136832, "pid": 76337, "tid": -914061504, "ts": 1716454223433530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223539170, "dur": 90,
    "args": {
      "External id": 136854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136854,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136854, "pid": 5, "tid": 7, "ts": 1716454223539170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433582, "dur": 10,
    "args": {
      "External id": 136854,
      "cbid": 211, "correlation": 136854
    }
  },
  {
    "ph": "s", "id": 136854, "pid": 76337, "tid": -914061504, "ts": 1716454223433582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433668, "dur": 1,
    "args": {
      "External id": 136870,
      "cbid": 251, "correlation": 136870
    }
  },
  {
    "ph": "f", "id": 136870, "pid": 76337, "tid": -914061504, "ts": 1716454223433668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223539262, "dur": 567,
    "args": {
      "External id": 136872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136872,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 136872, "pid": 5, "tid": 7, "ts": 1716454223539262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433674, "dur": 13,
    "args": {
      "External id": 136872,
      "cbid": 211, "correlation": 136872
    }
  },
  {
    "ph": "s", "id": 136872, "pid": 76337, "tid": -914061504, "ts": 1716454223433674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223539830, "dur": 242,
    "args": {
      "External id": 136880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136880, "pid": 5, "tid": 7, "ts": 1716454223539830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433740, "dur": 13,
    "args": {
      "External id": 136880,
      "cbid": 211, "correlation": 136880
    }
  },
  {
    "ph": "s", "id": 136880, "pid": 76337, "tid": -914061504, "ts": 1716454223433740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223540074, "dur": 249,
    "args": {
      "External id": 136888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136888, "pid": 5, "tid": 7, "ts": 1716454223540074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433771, "dur": 9,
    "args": {
      "External id": 136888,
      "cbid": 211, "correlation": 136888
    }
  },
  {
    "ph": "s", "id": 136888, "pid": 76337, "tid": -914061504, "ts": 1716454223433771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433853, "dur": 1,
    "args": {
      "External id": 136904,
      "cbid": 251, "correlation": 136904
    }
  },
  {
    "ph": "f", "id": 136904, "pid": 76337, "tid": -914061504, "ts": 1716454223433853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433858, "dur": 0,
    "args": {
      "External id": 136906,
      "cbid": 251, "correlation": 136906
    }
  },
  {
    "ph": "f", "id": 136906, "pid": 76337, "tid": -914061504, "ts": 1716454223433858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223540324, "dur": 356,
    "args": {
      "External id": 136907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136907,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 136907, "pid": 5, "tid": 7, "ts": 1716454223540324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433861, "dur": 13,
    "args": {
      "External id": 136907,
      "cbid": 211, "correlation": 136907
    }
  },
  {
    "ph": "s", "id": 136907, "pid": 76337, "tid": -914061504, "ts": 1716454223433861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223540682, "dur": 50,
    "args": {
      "External id": 136915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136915,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136915, "pid": 5, "tid": 7, "ts": 1716454223540682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433903, "dur": 10,
    "args": {
      "External id": 136915,
      "cbid": 211, "correlation": 136915
    }
  },
  {
    "ph": "s", "id": 136915, "pid": 76337, "tid": -914061504, "ts": 1716454223433903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223540733, "dur": 157,
    "args": {
      "External id": 136926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136926, "pid": 5, "tid": 7, "ts": 1716454223540733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223433970, "dur": 21,
    "args": {
      "External id": 136926,
      "cbid": 211, "correlation": 136926
    }
  },
  {
    "ph": "s", "id": 136926, "pid": 76337, "tid": -914061504, "ts": 1716454223433970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434044, "dur": 0,
    "args": {
      "External id": 136938,
      "cbid": 317, "correlation": 136938
    }
  },
  {
    "ph": "f", "id": 136938, "pid": 76337, "tid": -914061504, "ts": 1716454223434044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434045, "dur": 0,
    "args": {
      "External id": 136939,
      "cbid": 203, "correlation": 136939
    }
  },
  {
    "ph": "f", "id": 136939, "pid": 76337, "tid": -914061504, "ts": 1716454223434045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434045, "dur": 0,
    "args": {
      "External id": 136940,
      "cbid": 205, "correlation": 136940
    }
  },
  {
    "ph": "f", "id": 136940, "pid": 76337, "tid": -914061504, "ts": 1716454223434045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434069, "dur": 1,
    "args": {
      "External id": 136944,
      "cbid": 251, "correlation": 136944
    }
  },
  {
    "ph": "f", "id": 136944, "pid": 76337, "tid": -914061504, "ts": 1716454223434069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434071, "dur": 0,
    "args": {
      "External id": 136945,
      "cbid": 251, "correlation": 136945
    }
  },
  {
    "ph": "f", "id": 136945, "pid": 76337, "tid": -914061504, "ts": 1716454223434071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434072, "dur": 0,
    "args": {
      "External id": 136946,
      "cbid": 251, "correlation": 136946
    }
  },
  {
    "ph": "f", "id": 136946, "pid": 76337, "tid": -914061504, "ts": 1716454223434072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434073, "dur": 0,
    "args": {
      "External id": 136947,
      "cbid": 251, "correlation": 136947
    }
  },
  {
    "ph": "f", "id": 136947, "pid": 76337, "tid": -914061504, "ts": 1716454223434073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434074, "dur": 0,
    "args": {
      "External id": 136948,
      "cbid": 251, "correlation": 136948
    }
  },
  {
    "ph": "f", "id": 136948, "pid": 76337, "tid": -914061504, "ts": 1716454223434074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434075, "dur": 0,
    "args": {
      "External id": 136949,
      "cbid": 251, "correlation": 136949
    }
  },
  {
    "ph": "f", "id": 136949, "pid": 76337, "tid": -914061504, "ts": 1716454223434075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434076, "dur": 0,
    "args": {
      "External id": 136950,
      "cbid": 251, "correlation": 136950
    }
  },
  {
    "ph": "f", "id": 136950, "pid": 76337, "tid": -914061504, "ts": 1716454223434076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434077, "dur": 0,
    "args": {
      "External id": 136951,
      "cbid": 251, "correlation": 136951
    }
  },
  {
    "ph": "f", "id": 136951, "pid": 76337, "tid": -914061504, "ts": 1716454223434077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434078, "dur": 0,
    "args": {
      "External id": 136952,
      "cbid": 251, "correlation": 136952
    }
  },
  {
    "ph": "f", "id": 136952, "pid": 76337, "tid": -914061504, "ts": 1716454223434078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223540891, "dur": 114,
    "args": {
      "External id": 136953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136953,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 136953, "pid": 5, "tid": 7, "ts": 1716454223540891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434081, "dur": 13,
    "args": {
      "External id": 136953,
      "cbid": 211, "correlation": 136953
    }
  },
  {
    "ph": "s", "id": 136953, "pid": 76337, "tid": -914061504, "ts": 1716454223434081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223541006, "dur": 59,
    "args": {
      "External id": 136959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136959, "pid": 5, "tid": 7, "ts": 1716454223541006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434116, "dur": 9,
    "args": {
      "External id": 136959,
      "cbid": 211, "correlation": 136959
    }
  },
  {
    "ph": "s", "id": 136959, "pid": 76337, "tid": -914061504, "ts": 1716454223434116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223541067, "dur": 49,
    "args": {
      "External id": 136967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136967,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136967, "pid": 5, "tid": 7, "ts": 1716454223541067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434147, "dur": 8,
    "args": {
      "External id": 136967,
      "cbid": 211, "correlation": 136967
    }
  },
  {
    "ph": "s", "id": 136967, "pid": 76337, "tid": -914061504, "ts": 1716454223434147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223541117, "dur": 97,
    "args": {
      "External id": 136976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136976,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 136976, "pid": 5, "tid": 7, "ts": 1716454223541117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434187, "dur": 10,
    "args": {
      "External id": 136976,
      "cbid": 211, "correlation": 136976
    }
  },
  {
    "ph": "s", "id": 136976, "pid": 76337, "tid": -914061504, "ts": 1716454223434187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223541216, "dur": 92,
    "args": {
      "External id": 136996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 136996,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 136996, "pid": 5, "tid": 7, "ts": 1716454223541216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434258, "dur": 11,
    "args": {
      "External id": 136996,
      "cbid": 211, "correlation": 136996
    }
  },
  {
    "ph": "s", "id": 136996, "pid": 76337, "tid": -914061504, "ts": 1716454223434258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223541309, "dur": 4,
    "args": {
      "External id": 137008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137008,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 137008, "pid": 5, "tid": 7, "ts": 1716454223541309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434280, "dur": 7,
    "args": {
      "External id": 137008,
      "cbid": 211, "correlation": 137008
    }
  },
  {
    "ph": "s", "id": 137008, "pid": 76337, "tid": -914061504, "ts": 1716454223434280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223541315, "dur": 107,
    "args": {
      "External id": 137011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137011, "pid": 5, "tid": 7, "ts": 1716454223541315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434298, "dur": 6,
    "args": {
      "External id": 137011,
      "cbid": 211, "correlation": 137011
    }
  },
  {
    "ph": "s", "id": 137011, "pid": 76337, "tid": -914061504, "ts": 1716454223434298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223541423, "dur": 69,
    "args": {
      "External id": 137020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137020,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137020, "pid": 5, "tid": 7, "ts": 1716454223541423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434338, "dur": 10,
    "args": {
      "External id": 137020,
      "cbid": 211, "correlation": 137020
    }
  },
  {
    "ph": "s", "id": 137020, "pid": 76337, "tid": -914061504, "ts": 1716454223434338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434390, "dur": 0,
    "args": {
      "External id": 137030,
      "cbid": 317, "correlation": 137030
    }
  },
  {
    "ph": "f", "id": 137030, "pid": 76337, "tid": -914061504, "ts": 1716454223434390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434391, "dur": 0,
    "args": {
      "External id": 137031,
      "cbid": 203, "correlation": 137031
    }
  },
  {
    "ph": "f", "id": 137031, "pid": 76337, "tid": -914061504, "ts": 1716454223434391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434391, "dur": 0,
    "args": {
      "External id": 137032,
      "cbid": 205, "correlation": 137032
    }
  },
  {
    "ph": "f", "id": 137032, "pid": 76337, "tid": -914061504, "ts": 1716454223434391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223541493, "dur": 75,
    "args": {
      "External id": 137036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137036,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137036, "pid": 5, "tid": 7, "ts": 1716454223541493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434407, "dur": 12,
    "args": {
      "External id": 137036,
      "cbid": 211, "correlation": 137036
    }
  },
  {
    "ph": "s", "id": 137036, "pid": 76337, "tid": -914061504, "ts": 1716454223434407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223541569, "dur": 23,
    "args": {
      "External id": 137038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137038,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137038, "pid": 5, "tid": 7, "ts": 1716454223541569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434421, "dur": 5,
    "args": {
      "External id": 137038,
      "cbid": 211, "correlation": 137038
    }
  },
  {
    "ph": "s", "id": 137038, "pid": 76337, "tid": -914061504, "ts": 1716454223434421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223541593, "dur": 4,
    "args": {
      "External id": 137040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137040, "pid": 5, "tid": 7, "ts": 1716454223541593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434431, "dur": 6,
    "args": {
      "External id": 137040,
      "cbid": 211, "correlation": 137040
    }
  },
  {
    "ph": "s", "id": 137040, "pid": 76337, "tid": -914061504, "ts": 1716454223434431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434440, "dur": 0,
    "args": {
      "External id": 137041,
      "cbid": 51, "correlation": 137041
    }
  },
  {
    "ph": "s", "id": 137041, "pid": 76337, "tid": -914061504, "ts": 1716454223434440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223541599, "dur": 1349,
    "args": {
      "External id": 137042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137042,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137042, "pid": 5, "tid": 7, "ts": 1716454223541599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434441, "dur": 5,
    "args": {
      "External id": 137042,
      "cbid": 211, "correlation": 137042
    }
  },
  {
    "ph": "s", "id": 137042, "pid": 76337, "tid": -914061504, "ts": 1716454223434441,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223542949, "dur": 60,
    "args": {
      "External id": 137047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137047, "pid": 5, "tid": 7, "ts": 1716454223542949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434468, "dur": 8,
    "args": {
      "External id": 137047,
      "cbid": 211, "correlation": 137047
    }
  },
  {
    "ph": "s", "id": 137047, "pid": 76337, "tid": -914061504, "ts": 1716454223434468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223543010, "dur": 4,
    "args": {
      "External id": 137055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137055,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137055, "pid": 5, "tid": 7, "ts": 1716454223543010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434512, "dur": 10,
    "args": {
      "External id": 137055,
      "cbid": 211, "correlation": 137055
    }
  },
  {
    "ph": "s", "id": 137055, "pid": 76337, "tid": -914061504, "ts": 1716454223434512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434577, "dur": 1,
    "args": {
      "External id": 137071,
      "cbid": 251, "correlation": 137071
    }
  },
  {
    "ph": "f", "id": 137071, "pid": 76337, "tid": -914061504, "ts": 1716454223434577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434582, "dur": 0,
    "args": {
      "External id": 137073,
      "cbid": 251, "correlation": 137073
    }
  },
  {
    "ph": "f", "id": 137073, "pid": 76337, "tid": -914061504, "ts": 1716454223434582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223543015, "dur": 10,
    "args": {
      "External id": 137074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137074,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 137074, "pid": 5, "tid": 7, "ts": 1716454223543015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434585, "dur": 12,
    "args": {
      "External id": 137074,
      "cbid": 211, "correlation": 137074
    }
  },
  {
    "ph": "s", "id": 137074, "pid": 76337, "tid": -914061504, "ts": 1716454223434585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223543026, "dur": 5,
    "args": {
      "External id": 137076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137076,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 137076, "pid": 5, "tid": 7, "ts": 1716454223543026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434598, "dur": 88,
    "args": {
      "External id": 137076,
      "cbid": 211, "correlation": 137076
    }
  },
  {
    "ph": "s", "id": 137076, "pid": 76337, "tid": -914061504, "ts": 1716454223434598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223543033, "dur": 55,
    "args": {
      "External id": 137086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137086,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137086, "pid": 5, "tid": 7, "ts": 1716454223543033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434739, "dur": 13,
    "args": {
      "External id": 137086,
      "cbid": 211, "correlation": 137086
    }
  },
  {
    "ph": "s", "id": 137086, "pid": 76337, "tid": -914061504, "ts": 1716454223434739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223543089, "dur": 51,
    "args": {
      "External id": 137106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137106,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 137106, "pid": 5, "tid": 7, "ts": 1716454223543089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434806, "dur": 11,
    "args": {
      "External id": 137106,
      "cbid": 211, "correlation": 137106
    }
  },
  {
    "ph": "s", "id": 137106, "pid": 76337, "tid": -914061504, "ts": 1716454223434806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223543141, "dur": 4,
    "args": {
      "External id": 137118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137118,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137118, "pid": 5, "tid": 7, "ts": 1716454223543141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434827, "dur": 6,
    "args": {
      "External id": 137118,
      "cbid": 211, "correlation": 137118
    }
  },
  {
    "ph": "s", "id": 137118, "pid": 76337, "tid": -914061504, "ts": 1716454223434827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223543147, "dur": 56,
    "args": {
      "External id": 137121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137121, "pid": 5, "tid": 7, "ts": 1716454223543147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434845, "dur": 7,
    "args": {
      "External id": 137121,
      "cbid": 211, "correlation": 137121
    }
  },
  {
    "ph": "s", "id": 137121, "pid": 76337, "tid": -914061504, "ts": 1716454223434845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223543204, "dur": 37,
    "args": {
      "External id": 137130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137130,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137130, "pid": 5, "tid": 7, "ts": 1716454223543204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434886, "dur": 10,
    "args": {
      "External id": 137130,
      "cbid": 211, "correlation": 137130
    }
  },
  {
    "ph": "s", "id": 137130, "pid": 76337, "tid": -914061504, "ts": 1716454223434886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434950, "dur": 0,
    "args": {
      "External id": 137140,
      "cbid": 317, "correlation": 137140
    }
  },
  {
    "ph": "f", "id": 137140, "pid": 76337, "tid": -914061504, "ts": 1716454223434950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434951, "dur": 0,
    "args": {
      "External id": 137141,
      "cbid": 203, "correlation": 137141
    }
  },
  {
    "ph": "f", "id": 137141, "pid": 76337, "tid": -914061504, "ts": 1716454223434951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434952, "dur": 0,
    "args": {
      "External id": 137142,
      "cbid": 205, "correlation": 137142
    }
  },
  {
    "ph": "f", "id": 137142, "pid": 76337, "tid": -914061504, "ts": 1716454223434952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223543242, "dur": 40,
    "args": {
      "External id": 137146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137146,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137146, "pid": 5, "tid": 7, "ts": 1716454223543242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434966, "dur": 20,
    "args": {
      "External id": 137146,
      "cbid": 211, "correlation": 137146
    }
  },
  {
    "ph": "s", "id": 137146, "pid": 76337, "tid": -914061504, "ts": 1716454223434966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223543283, "dur": 14,
    "args": {
      "External id": 137148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137148,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137148, "pid": 5, "tid": 7, "ts": 1716454223543283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434989, "dur": 6,
    "args": {
      "External id": 137148,
      "cbid": 211, "correlation": 137148
    }
  },
  {
    "ph": "s", "id": 137148, "pid": 76337, "tid": -914061504, "ts": 1716454223434989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223543299, "dur": 3,
    "args": {
      "External id": 137150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137150,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137150, "pid": 5, "tid": 7, "ts": 1716454223543299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223434999, "dur": 6,
    "args": {
      "External id": 137150,
      "cbid": 211, "correlation": 137150
    }
  },
  {
    "ph": "s", "id": 137150, "pid": 76337, "tid": -914061504, "ts": 1716454223434999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435007, "dur": 0,
    "args": {
      "External id": 137151,
      "cbid": 51, "correlation": 137151
    }
  },
  {
    "ph": "s", "id": 137151, "pid": 76337, "tid": -914061504, "ts": 1716454223435007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223543303, "dur": 686,
    "args": {
      "External id": 137152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137152,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137152, "pid": 5, "tid": 7, "ts": 1716454223543303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435008, "dur": 5,
    "args": {
      "External id": 137152,
      "cbid": 211, "correlation": 137152
    }
  },
  {
    "ph": "s", "id": 137152, "pid": 76337, "tid": -914061504, "ts": 1716454223435008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223543991, "dur": 60,
    "args": {
      "External id": 137157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137157,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137157, "pid": 5, "tid": 7, "ts": 1716454223543991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435035, "dur": 8,
    "args": {
      "External id": 137157,
      "cbid": 211, "correlation": 137157
    }
  },
  {
    "ph": "s", "id": 137157, "pid": 76337, "tid": -914061504, "ts": 1716454223435035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435093, "dur": 0,
    "args": {
      "External id": 137167,
      "cbid": 317, "correlation": 137167
    }
  },
  {
    "ph": "f", "id": 137167, "pid": 76337, "tid": -914061504, "ts": 1716454223435093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435094, "dur": 0,
    "args": {
      "External id": 137168,
      "cbid": 203, "correlation": 137168
    }
  },
  {
    "ph": "f", "id": 137168, "pid": 76337, "tid": -914061504, "ts": 1716454223435094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435094, "dur": 0,
    "args": {
      "External id": 137169,
      "cbid": 205, "correlation": 137169
    }
  },
  {
    "ph": "f", "id": 137169, "pid": 76337, "tid": -914061504, "ts": 1716454223435094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223544052, "dur": 75,
    "args": {
      "External id": 137173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137173,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137173, "pid": 5, "tid": 7, "ts": 1716454223544052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435108, "dur": 12,
    "args": {
      "External id": 137173,
      "cbid": 211, "correlation": 137173
    }
  },
  {
    "ph": "s", "id": 137173, "pid": 76337, "tid": -914061504, "ts": 1716454223435108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223544128, "dur": 207,
    "args": {
      "External id": 137175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137175,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137175, "pid": 5, "tid": 7, "ts": 1716454223544128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435128, "dur": 8,
    "args": {
      "External id": 137175,
      "cbid": 211, "correlation": 137175
    }
  },
  {
    "ph": "s", "id": 137175, "pid": 76337, "tid": -914061504, "ts": 1716454223435128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223544337, "dur": 39,
    "args": {
      "External id": 137177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137177,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137177, "pid": 5, "tid": 7, "ts": 1716454223544337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435140, "dur": 466,
    "args": {
      "External id": 137177,
      "cbid": 211, "correlation": 137177
    }
  },
  {
    "ph": "s", "id": 137177, "pid": 76337, "tid": -914061504, "ts": 1716454223435140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223544377, "dur": 58,
    "args": {
      "External id": 137183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137183,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137183, "pid": 5, "tid": 7, "ts": 1716454223544377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435627, "dur": 9,
    "args": {
      "External id": 137183,
      "cbid": 211, "correlation": 137183
    }
  },
  {
    "ph": "s", "id": 137183, "pid": 76337, "tid": -914061504, "ts": 1716454223435627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223544437, "dur": 50,
    "args": {
      "External id": 137191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137191,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137191, "pid": 5, "tid": 7, "ts": 1716454223544437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435657, "dur": 8,
    "args": {
      "External id": 137191,
      "cbid": 211, "correlation": 137191
    }
  },
  {
    "ph": "s", "id": 137191, "pid": 76337, "tid": -914061504, "ts": 1716454223435657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223544488, "dur": 35,
    "args": {
      "External id": 137199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137199,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137199, "pid": 5, "tid": 7, "ts": 1716454223544488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435686, "dur": 32,
    "args": {
      "External id": 137199,
      "cbid": 211, "correlation": 137199
    }
  },
  {
    "ph": "s", "id": 137199, "pid": 76337, "tid": -914061504, "ts": 1716454223435686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223544524, "dur": 51,
    "args": {
      "External id": 137219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137219,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 137219, "pid": 5, "tid": 7, "ts": 1716454223544524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435794, "dur": 12,
    "args": {
      "External id": 137219,
      "cbid": 211, "correlation": 137219
    }
  },
  {
    "ph": "s", "id": 137219, "pid": 76337, "tid": -914061504, "ts": 1716454223435794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223544576, "dur": 4,
    "args": {
      "External id": 137231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137231,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137231, "pid": 5, "tid": 7, "ts": 1716454223544576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435816, "dur": 6,
    "args": {
      "External id": 137231,
      "cbid": 211, "correlation": 137231
    }
  },
  {
    "ph": "s", "id": 137231, "pid": 76337, "tid": -914061504, "ts": 1716454223435816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223544582, "dur": 54,
    "args": {
      "External id": 137234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137234, "pid": 5, "tid": 7, "ts": 1716454223544582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435834, "dur": 6,
    "args": {
      "External id": 137234,
      "cbid": 211, "correlation": 137234
    }
  },
  {
    "ph": "s", "id": 137234, "pid": 76337, "tid": -914061504, "ts": 1716454223435834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435891, "dur": 0,
    "args": {
      "External id": 137245,
      "cbid": 317, "correlation": 137245
    }
  },
  {
    "ph": "f", "id": 137245, "pid": 76337, "tid": -914061504, "ts": 1716454223435891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435892, "dur": 0,
    "args": {
      "External id": 137246,
      "cbid": 203, "correlation": 137246
    }
  },
  {
    "ph": "f", "id": 137246, "pid": 76337, "tid": -914061504, "ts": 1716454223435892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435892, "dur": 0,
    "args": {
      "External id": 137247,
      "cbid": 205, "correlation": 137247
    }
  },
  {
    "ph": "f", "id": 137247, "pid": 76337, "tid": -914061504, "ts": 1716454223435892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435914, "dur": 1,
    "args": {
      "External id": 137251,
      "cbid": 251, "correlation": 137251
    }
  },
  {
    "ph": "f", "id": 137251, "pid": 76337, "tid": -914061504, "ts": 1716454223435914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435916, "dur": 0,
    "args": {
      "External id": 137252,
      "cbid": 251, "correlation": 137252
    }
  },
  {
    "ph": "f", "id": 137252, "pid": 76337, "tid": -914061504, "ts": 1716454223435916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435917, "dur": 0,
    "args": {
      "External id": 137253,
      "cbid": 251, "correlation": 137253
    }
  },
  {
    "ph": "f", "id": 137253, "pid": 76337, "tid": -914061504, "ts": 1716454223435917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435918, "dur": 0,
    "args": {
      "External id": 137254,
      "cbid": 251, "correlation": 137254
    }
  },
  {
    "ph": "f", "id": 137254, "pid": 76337, "tid": -914061504, "ts": 1716454223435918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435919, "dur": 0,
    "args": {
      "External id": 137255,
      "cbid": 251, "correlation": 137255
    }
  },
  {
    "ph": "f", "id": 137255, "pid": 76337, "tid": -914061504, "ts": 1716454223435919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435920, "dur": 0,
    "args": {
      "External id": 137256,
      "cbid": 251, "correlation": 137256
    }
  },
  {
    "ph": "f", "id": 137256, "pid": 76337, "tid": -914061504, "ts": 1716454223435920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435920, "dur": 0,
    "args": {
      "External id": 137257,
      "cbid": 251, "correlation": 137257
    }
  },
  {
    "ph": "f", "id": 137257, "pid": 76337, "tid": -914061504, "ts": 1716454223435920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435921, "dur": 0,
    "args": {
      "External id": 137258,
      "cbid": 251, "correlation": 137258
    }
  },
  {
    "ph": "f", "id": 137258, "pid": 76337, "tid": -914061504, "ts": 1716454223435921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435923, "dur": 0,
    "args": {
      "External id": 137259,
      "cbid": 251, "correlation": 137259
    }
  },
  {
    "ph": "f", "id": 137259, "pid": 76337, "tid": -914061504, "ts": 1716454223435923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223544637, "dur": 112,
    "args": {
      "External id": 137260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137260,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 137260, "pid": 5, "tid": 7, "ts": 1716454223544637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435925, "dur": 13,
    "args": {
      "External id": 137260,
      "cbid": 211, "correlation": 137260
    }
  },
  {
    "ph": "s", "id": 137260, "pid": 76337, "tid": -914061504, "ts": 1716454223435925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223544750, "dur": 59,
    "args": {
      "External id": 137266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137266,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137266, "pid": 5, "tid": 7, "ts": 1716454223544750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223435960, "dur": 9,
    "args": {
      "External id": 137266,
      "cbid": 211, "correlation": 137266
    }
  },
  {
    "ph": "s", "id": 137266, "pid": 76337, "tid": -914061504, "ts": 1716454223435960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223544810, "dur": 662,
    "args": {
      "External id": 137275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137275,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137275, "pid": 5, "tid": 7, "ts": 1716454223544810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436050, "dur": 14,
    "args": {
      "External id": 137275,
      "cbid": 211, "correlation": 137275
    }
  },
  {
    "ph": "s", "id": 137275, "pid": 76337, "tid": -914061504, "ts": 1716454223436050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223545473, "dur": 179,
    "args": {
      "External id": 137297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137297,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137297, "pid": 5, "tid": 7, "ts": 1716454223545473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436107, "dur": 10,
    "args": {
      "External id": 137297,
      "cbid": 211, "correlation": 137297
    }
  },
  {
    "ph": "s", "id": 137297, "pid": 76337, "tid": -914061504, "ts": 1716454223436107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436194, "dur": 1,
    "args": {
      "External id": 137308,
      "cbid": 251, "correlation": 137308
    }
  },
  {
    "ph": "f", "id": 137308, "pid": 76337, "tid": -914061504, "ts": 1716454223436194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223545654, "dur": 197,
    "args": {
      "External id": 137309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137309,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137309, "pid": 5, "tid": 7, "ts": 1716454223545654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436199, "dur": 13,
    "args": {
      "External id": 137309,
      "cbid": 211, "correlation": 137309
    }
  },
  {
    "ph": "s", "id": 137309, "pid": 76337, "tid": -914061504, "ts": 1716454223436199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436266, "dur": 1,
    "args": {
      "External id": 137320,
      "cbid": 251, "correlation": 137320
    }
  },
  {
    "ph": "f", "id": 137320, "pid": 76337, "tid": -914061504, "ts": 1716454223436266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223545853, "dur": 184,
    "args": {
      "External id": 137321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137321,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137321, "pid": 5, "tid": 7, "ts": 1716454223545853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436271, "dur": 12,
    "args": {
      "External id": 137321,
      "cbid": 211, "correlation": 137321
    }
  },
  {
    "ph": "s", "id": 137321, "pid": 76337, "tid": -914061504, "ts": 1716454223436271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436335, "dur": 1,
    "args": {
      "External id": 137332,
      "cbid": 251, "correlation": 137332
    }
  },
  {
    "ph": "f", "id": 137332, "pid": 76337, "tid": -914061504, "ts": 1716454223436335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223546038, "dur": 184,
    "args": {
      "External id": 137333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137333,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137333, "pid": 5, "tid": 7, "ts": 1716454223546038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436339, "dur": 11,
    "args": {
      "External id": 137333,
      "cbid": 211, "correlation": 137333
    }
  },
  {
    "ph": "s", "id": 137333, "pid": 76337, "tid": -914061504, "ts": 1716454223436339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223546224, "dur": 18358,
    "args": {
      "External id": 137354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137354,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 137354, "pid": 5, "tid": 7, "ts": 1716454223546224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436418, "dur": 12,
    "args": {
      "External id": 137354,
      "cbid": 211, "correlation": 137354
    }
  },
  {
    "ph": "s", "id": 137354, "pid": 76337, "tid": -914061504, "ts": 1716454223436418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436515, "dur": 1,
    "args": {
      "External id": 137372,
      "cbid": 251, "correlation": 137372
    }
  },
  {
    "ph": "f", "id": 137372, "pid": 76337, "tid": -914061504, "ts": 1716454223436515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223564583, "dur": 198,
    "args": {
      "External id": 137374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137374,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137374, "pid": 5, "tid": 7, "ts": 1716454223564583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436521, "dur": 13,
    "args": {
      "External id": 137374,
      "cbid": 211, "correlation": 137374
    }
  },
  {
    "ph": "s", "id": 137374, "pid": 76337, "tid": -914061504, "ts": 1716454223436521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223564783, "dur": 67,
    "args": {
      "External id": 137382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137382,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137382, "pid": 5, "tid": 7, "ts": 1716454223564783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436590, "dur": 12,
    "args": {
      "External id": 137382,
      "cbid": 211, "correlation": 137382
    }
  },
  {
    "ph": "s", "id": 137382, "pid": 76337, "tid": -914061504, "ts": 1716454223436590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223564851, "dur": 97,
    "args": {
      "External id": 137390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137390,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137390, "pid": 5, "tid": 7, "ts": 1716454223564851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436628, "dur": 8,
    "args": {
      "External id": 137390,
      "cbid": 211, "correlation": 137390
    }
  },
  {
    "ph": "s", "id": 137390, "pid": 76337, "tid": -914061504, "ts": 1716454223436628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223564949, "dur": 53,
    "args": {
      "External id": 137401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137401, "pid": 5, "tid": 7, "ts": 1716454223564949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436700, "dur": 51,
    "args": {
      "External id": 137401,
      "cbid": 211, "correlation": 137401
    }
  },
  {
    "ph": "s", "id": 137401, "pid": 76337, "tid": -914061504, "ts": 1716454223436700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223565003, "dur": 91,
    "args": {
      "External id": 137423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137423,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137423, "pid": 5, "tid": 7, "ts": 1716454223565003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223436769, "dur": 1917,
    "args": {
      "External id": 137423,
      "cbid": 211, "correlation": 137423
    }
  },
  {
    "ph": "s", "id": 137423, "pid": 76337, "tid": -914061504, "ts": 1716454223436769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438762, "dur": 1,
    "args": {
      "External id": 137434,
      "cbid": 251, "correlation": 137434
    }
  },
  {
    "ph": "f", "id": 137434, "pid": 76337, "tid": -914061504, "ts": 1716454223438762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223565095, "dur": 103,
    "args": {
      "External id": 137435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137435,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137435, "pid": 5, "tid": 7, "ts": 1716454223565095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438767, "dur": 67,
    "args": {
      "External id": 137435,
      "cbid": 211, "correlation": 137435
    }
  },
  {
    "ph": "s", "id": 137435, "pid": 76337, "tid": -914061504, "ts": 1716454223438767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438893, "dur": 1,
    "args": {
      "External id": 137446,
      "cbid": 251, "correlation": 137446
    }
  },
  {
    "ph": "f", "id": 137446, "pid": 76337, "tid": -914061504, "ts": 1716454223438893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438897, "dur": 0,
    "args": {
      "External id": 137447,
      "cbid": 251, "correlation": 137447
    }
  },
  {
    "ph": "f", "id": 137447, "pid": 76337, "tid": -914061504, "ts": 1716454223438897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223565199, "dur": 11,
    "args": {
      "External id": 137448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137448,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 137448, "pid": 5, "tid": 7, "ts": 1716454223565199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438899, "dur": 13,
    "args": {
      "External id": 137448,
      "cbid": 211, "correlation": 137448
    }
  },
  {
    "ph": "s", "id": 137448, "pid": 76337, "tid": -914061504, "ts": 1716454223438899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223565211, "dur": 5,
    "args": {
      "External id": 137450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137450,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 137450, "pid": 5, "tid": 7, "ts": 1716454223565211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438914, "dur": 6,
    "args": {
      "External id": 137450,
      "cbid": 211, "correlation": 137450
    }
  },
  {
    "ph": "s", "id": 137450, "pid": 76337, "tid": -914061504, "ts": 1716454223438914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438983, "dur": 1,
    "args": {
      "External id": 137461,
      "cbid": 251, "correlation": 137461
    }
  },
  {
    "ph": "f", "id": 137461, "pid": 76337, "tid": -914061504, "ts": 1716454223438983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438988, "dur": 0,
    "args": {
      "External id": 137462,
      "cbid": 251, "correlation": 137462
    }
  },
  {
    "ph": "f", "id": 137462, "pid": 76337, "tid": -914061504, "ts": 1716454223438988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223565217, "dur": 6,
    "args": {
      "External id": 137463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137463,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 137463, "pid": 5, "tid": 7, "ts": 1716454223565217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223438989, "dur": 12,
    "args": {
      "External id": 137463,
      "cbid": 211, "correlation": 137463
    }
  },
  {
    "ph": "s", "id": 137463, "pid": 76337, "tid": -914061504, "ts": 1716454223438989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223565225, "dur": 3,
    "args": {
      "External id": 137465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137465,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 137465, "pid": 5, "tid": 7, "ts": 1716454223565225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439003, "dur": 5,
    "args": {
      "External id": 137465,
      "cbid": 211, "correlation": 137465
    }
  },
  {
    "ph": "s", "id": 137465, "pid": 76337, "tid": -914061504, "ts": 1716454223439003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223565230, "dur": 156,
    "args": {
      "External id": 137486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137486,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 137486, "pid": 5, "tid": 7, "ts": 1716454223565230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439076, "dur": 13,
    "args": {
      "External id": 137486,
      "cbid": 211, "correlation": 137486
    }
  },
  {
    "ph": "s", "id": 137486, "pid": 76337, "tid": -914061504, "ts": 1716454223439076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439173, "dur": 1,
    "args": {
      "External id": 137504,
      "cbid": 251, "correlation": 137504
    }
  },
  {
    "ph": "f", "id": 137504, "pid": 76337, "tid": -914061504, "ts": 1716454223439173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223565387, "dur": 106,
    "args": {
      "External id": 137506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137506,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 137506, "pid": 5, "tid": 7, "ts": 1716454223565387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439179, "dur": 14,
    "args": {
      "External id": 137506,
      "cbid": 211, "correlation": 137506
    }
  },
  {
    "ph": "s", "id": 137506, "pid": 76337, "tid": -914061504, "ts": 1716454223439179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223565495, "dur": 35,
    "args": {
      "External id": 137514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137514,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137514, "pid": 5, "tid": 7, "ts": 1716454223565495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439249, "dur": 12,
    "args": {
      "External id": 137514,
      "cbid": 211, "correlation": 137514
    }
  },
  {
    "ph": "s", "id": 137514, "pid": 76337, "tid": -914061504, "ts": 1716454223439249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223565531, "dur": 68,
    "args": {
      "External id": 137522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137522, "pid": 5, "tid": 7, "ts": 1716454223565531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439290, "dur": 9,
    "args": {
      "External id": 137522,
      "cbid": 211, "correlation": 137522
    }
  },
  {
    "ph": "s", "id": 137522, "pid": 76337, "tid": -914061504, "ts": 1716454223439290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223565600, "dur": 91,
    "args": {
      "External id": 137544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137544,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137544, "pid": 5, "tid": 7, "ts": 1716454223565600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439341, "dur": 10,
    "args": {
      "External id": 137544,
      "cbid": 211, "correlation": 137544
    }
  },
  {
    "ph": "s", "id": 137544, "pid": 76337, "tid": -914061504, "ts": 1716454223439341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439425, "dur": 1,
    "args": {
      "External id": 137560,
      "cbid": 251, "correlation": 137560
    }
  },
  {
    "ph": "f", "id": 137560, "pid": 76337, "tid": -914061504, "ts": 1716454223439425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223565692, "dur": 566,
    "args": {
      "External id": 137562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137562,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137562, "pid": 5, "tid": 7, "ts": 1716454223565692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439431, "dur": 14,
    "args": {
      "External id": 137562,
      "cbid": 211, "correlation": 137562
    }
  },
  {
    "ph": "s", "id": 137562, "pid": 76337, "tid": -914061504, "ts": 1716454223439431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223566260, "dur": 242,
    "args": {
      "External id": 137570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137570, "pid": 5, "tid": 7, "ts": 1716454223566260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439498, "dur": 12,
    "args": {
      "External id": 137570,
      "cbid": 211, "correlation": 137570
    }
  },
  {
    "ph": "s", "id": 137570, "pid": 76337, "tid": -914061504, "ts": 1716454223439498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223566503, "dur": 252,
    "args": {
      "External id": 137578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137578, "pid": 5, "tid": 7, "ts": 1716454223566503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439528, "dur": 8,
    "args": {
      "External id": 137578,
      "cbid": 211, "correlation": 137578
    }
  },
  {
    "ph": "s", "id": 137578, "pid": 76337, "tid": -914061504, "ts": 1716454223439528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439609, "dur": 1,
    "args": {
      "External id": 137594,
      "cbid": 251, "correlation": 137594
    }
  },
  {
    "ph": "f", "id": 137594, "pid": 76337, "tid": -914061504, "ts": 1716454223439609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439614, "dur": 0,
    "args": {
      "External id": 137596,
      "cbid": 251, "correlation": 137596
    }
  },
  {
    "ph": "f", "id": 137596, "pid": 76337, "tid": -914061504, "ts": 1716454223439614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223566756, "dur": 356,
    "args": {
      "External id": 137597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137597,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 137597, "pid": 5, "tid": 7, "ts": 1716454223566756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439617, "dur": 13,
    "args": {
      "External id": 137597,
      "cbid": 211, "correlation": 137597
    }
  },
  {
    "ph": "s", "id": 137597, "pid": 76337, "tid": -914061504, "ts": 1716454223439617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223567114, "dur": 50,
    "args": {
      "External id": 137605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137605,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137605, "pid": 5, "tid": 7, "ts": 1716454223567114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439659, "dur": 10,
    "args": {
      "External id": 137605,
      "cbid": 211, "correlation": 137605
    }
  },
  {
    "ph": "s", "id": 137605, "pid": 76337, "tid": -914061504, "ts": 1716454223439659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223567165, "dur": 156,
    "args": {
      "External id": 137616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137616,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137616, "pid": 5, "tid": 7, "ts": 1716454223567165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439727, "dur": 209,
    "args": {
      "External id": 137616,
      "cbid": 211, "correlation": 137616
    }
  },
  {
    "ph": "s", "id": 137616, "pid": 76337, "tid": -914061504, "ts": 1716454223439727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439995, "dur": 0,
    "args": {
      "External id": 137628,
      "cbid": 317, "correlation": 137628
    }
  },
  {
    "ph": "f", "id": 137628, "pid": 76337, "tid": -914061504, "ts": 1716454223439995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439996, "dur": 0,
    "args": {
      "External id": 137629,
      "cbid": 203, "correlation": 137629
    }
  },
  {
    "ph": "f", "id": 137629, "pid": 76337, "tid": -914061504, "ts": 1716454223439996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223439997, "dur": 0,
    "args": {
      "External id": 137630,
      "cbid": 205, "correlation": 137630
    }
  },
  {
    "ph": "f", "id": 137630, "pid": 76337, "tid": -914061504, "ts": 1716454223439997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440019, "dur": 1,
    "args": {
      "External id": 137634,
      "cbid": 251, "correlation": 137634
    }
  },
  {
    "ph": "f", "id": 137634, "pid": 76337, "tid": -914061504, "ts": 1716454223440019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440021, "dur": 0,
    "args": {
      "External id": 137635,
      "cbid": 251, "correlation": 137635
    }
  },
  {
    "ph": "f", "id": 137635, "pid": 76337, "tid": -914061504, "ts": 1716454223440021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440022, "dur": 0,
    "args": {
      "External id": 137636,
      "cbid": 251, "correlation": 137636
    }
  },
  {
    "ph": "f", "id": 137636, "pid": 76337, "tid": -914061504, "ts": 1716454223440022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440023, "dur": 0,
    "args": {
      "External id": 137637,
      "cbid": 251, "correlation": 137637
    }
  },
  {
    "ph": "f", "id": 137637, "pid": 76337, "tid": -914061504, "ts": 1716454223440023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440024, "dur": 0,
    "args": {
      "External id": 137638,
      "cbid": 251, "correlation": 137638
    }
  },
  {
    "ph": "f", "id": 137638, "pid": 76337, "tid": -914061504, "ts": 1716454223440024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440025, "dur": 0,
    "args": {
      "External id": 137639,
      "cbid": 251, "correlation": 137639
    }
  },
  {
    "ph": "f", "id": 137639, "pid": 76337, "tid": -914061504, "ts": 1716454223440025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440026, "dur": 0,
    "args": {
      "External id": 137640,
      "cbid": 251, "correlation": 137640
    }
  },
  {
    "ph": "f", "id": 137640, "pid": 76337, "tid": -914061504, "ts": 1716454223440026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440027, "dur": 0,
    "args": {
      "External id": 137641,
      "cbid": 251, "correlation": 137641
    }
  },
  {
    "ph": "f", "id": 137641, "pid": 76337, "tid": -914061504, "ts": 1716454223440027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440029, "dur": 0,
    "args": {
      "External id": 137642,
      "cbid": 251, "correlation": 137642
    }
  },
  {
    "ph": "f", "id": 137642, "pid": 76337, "tid": -914061504, "ts": 1716454223440029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223567322, "dur": 113,
    "args": {
      "External id": 137643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137643,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 137643, "pid": 5, "tid": 7, "ts": 1716454223567322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440031, "dur": 31,
    "args": {
      "External id": 137643,
      "cbid": 211, "correlation": 137643
    }
  },
  {
    "ph": "s", "id": 137643, "pid": 76337, "tid": -914061504, "ts": 1716454223440031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223567437, "dur": 58,
    "args": {
      "External id": 137649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137649,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137649, "pid": 5, "tid": 7, "ts": 1716454223567437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440086, "dur": 102,
    "args": {
      "External id": 137649,
      "cbid": 211, "correlation": 137649
    }
  },
  {
    "ph": "s", "id": 137649, "pid": 76337, "tid": -914061504, "ts": 1716454223440086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223567496, "dur": 50,
    "args": {
      "External id": 137657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137657,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137657, "pid": 5, "tid": 7, "ts": 1716454223567496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440212, "dur": 277,
    "args": {
      "External id": 137657,
      "cbid": 211, "correlation": 137657
    }
  },
  {
    "ph": "s", "id": 137657, "pid": 76337, "tid": -914061504, "ts": 1716454223440212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223567547, "dur": 97,
    "args": {
      "External id": 137666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137666,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137666, "pid": 5, "tid": 7, "ts": 1716454223567547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440518, "dur": 11,
    "args": {
      "External id": 137666,
      "cbid": 211, "correlation": 137666
    }
  },
  {
    "ph": "s", "id": 137666, "pid": 76337, "tid": -914061504, "ts": 1716454223440518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223567645, "dur": 91,
    "args": {
      "External id": 137686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137686,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 137686, "pid": 5, "tid": 7, "ts": 1716454223567645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440591, "dur": 11,
    "args": {
      "External id": 137686,
      "cbid": 211, "correlation": 137686
    }
  },
  {
    "ph": "s", "id": 137686, "pid": 76337, "tid": -914061504, "ts": 1716454223440591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223567738, "dur": 4,
    "args": {
      "External id": 137698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137698,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 137698, "pid": 5, "tid": 7, "ts": 1716454223567738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440612, "dur": 12,
    "args": {
      "External id": 137698,
      "cbid": 211, "correlation": 137698
    }
  },
  {
    "ph": "s", "id": 137698, "pid": 76337, "tid": -914061504, "ts": 1716454223440612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223567743, "dur": 106,
    "args": {
      "External id": 137701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137701, "pid": 5, "tid": 7, "ts": 1716454223567743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440636, "dur": 108,
    "args": {
      "External id": 137701,
      "cbid": 211, "correlation": 137701
    }
  },
  {
    "ph": "s", "id": 137701, "pid": 76337, "tid": -914061504, "ts": 1716454223440636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223567851, "dur": 69,
    "args": {
      "External id": 137710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137710,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137710, "pid": 5, "tid": 7, "ts": 1716454223567851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440778, "dur": 10,
    "args": {
      "External id": 137710,
      "cbid": 211, "correlation": 137710
    }
  },
  {
    "ph": "s", "id": 137710, "pid": 76337, "tid": -914061504, "ts": 1716454223440778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440830, "dur": 0,
    "args": {
      "External id": 137720,
      "cbid": 317, "correlation": 137720
    }
  },
  {
    "ph": "f", "id": 137720, "pid": 76337, "tid": -914061504, "ts": 1716454223440830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440831, "dur": 0,
    "args": {
      "External id": 137721,
      "cbid": 203, "correlation": 137721
    }
  },
  {
    "ph": "f", "id": 137721, "pid": 76337, "tid": -914061504, "ts": 1716454223440831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440832, "dur": 0,
    "args": {
      "External id": 137722,
      "cbid": 205, "correlation": 137722
    }
  },
  {
    "ph": "f", "id": 137722, "pid": 76337, "tid": -914061504, "ts": 1716454223440832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223567921, "dur": 77,
    "args": {
      "External id": 137726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137726, "pid": 5, "tid": 7, "ts": 1716454223567921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440846, "dur": 13,
    "args": {
      "External id": 137726,
      "cbid": 211, "correlation": 137726
    }
  },
  {
    "ph": "s", "id": 137726, "pid": 76337, "tid": -914061504, "ts": 1716454223440846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223567999, "dur": 24,
    "args": {
      "External id": 137728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137728,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137728, "pid": 5, "tid": 7, "ts": 1716454223567999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440861, "dur": 5,
    "args": {
      "External id": 137728,
      "cbid": 211, "correlation": 137728
    }
  },
  {
    "ph": "s", "id": 137728, "pid": 76337, "tid": -914061504, "ts": 1716454223440861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223568025, "dur": 3,
    "args": {
      "External id": 137730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137730, "pid": 5, "tid": 7, "ts": 1716454223568025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440871, "dur": 5,
    "args": {
      "External id": 137730,
      "cbid": 211, "correlation": 137730
    }
  },
  {
    "ph": "s", "id": 137730, "pid": 76337, "tid": -914061504, "ts": 1716454223440871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440879, "dur": 0,
    "args": {
      "External id": 137731,
      "cbid": 51, "correlation": 137731
    }
  },
  {
    "ph": "s", "id": 137731, "pid": 76337, "tid": -914061504, "ts": 1716454223440879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223568029, "dur": 1346,
    "args": {
      "External id": 137732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137732,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137732, "pid": 5, "tid": 7, "ts": 1716454223568029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440880, "dur": 5,
    "args": {
      "External id": 137732,
      "cbid": 211, "correlation": 137732
    }
  },
  {
    "ph": "s", "id": 137732, "pid": 76337, "tid": -914061504, "ts": 1716454223440880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223569377, "dur": 59,
    "args": {
      "External id": 137737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137737, "pid": 5, "tid": 7, "ts": 1716454223569377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440907, "dur": 8,
    "args": {
      "External id": 137737,
      "cbid": 211, "correlation": 137737
    }
  },
  {
    "ph": "s", "id": 137737, "pid": 76337, "tid": -914061504, "ts": 1716454223440907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223569438, "dur": 4,
    "args": {
      "External id": 137745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137745,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137745, "pid": 5, "tid": 7, "ts": 1716454223569438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223440951, "dur": 10,
    "args": {
      "External id": 137745,
      "cbid": 211, "correlation": 137745
    }
  },
  {
    "ph": "s", "id": 137745, "pid": 76337, "tid": -914061504, "ts": 1716454223440951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441023, "dur": 1,
    "args": {
      "External id": 137761,
      "cbid": 251, "correlation": 137761
    }
  },
  {
    "ph": "f", "id": 137761, "pid": 76337, "tid": -914061504, "ts": 1716454223441023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441028, "dur": 0,
    "args": {
      "External id": 137763,
      "cbid": 251, "correlation": 137763
    }
  },
  {
    "ph": "f", "id": 137763, "pid": 76337, "tid": -914061504, "ts": 1716454223441028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223569443, "dur": 11,
    "args": {
      "External id": 137764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137764,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 137764, "pid": 5, "tid": 7, "ts": 1716454223569443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441030, "dur": 12,
    "args": {
      "External id": 137764,
      "cbid": 211, "correlation": 137764
    }
  },
  {
    "ph": "s", "id": 137764, "pid": 76337, "tid": -914061504, "ts": 1716454223441030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223569455, "dur": 5,
    "args": {
      "External id": 137766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137766,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 137766, "pid": 5, "tid": 7, "ts": 1716454223569455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441044, "dur": 6,
    "args": {
      "External id": 137766,
      "cbid": 211, "correlation": 137766
    }
  },
  {
    "ph": "s", "id": 137766, "pid": 76337, "tid": -914061504, "ts": 1716454223441044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223569461, "dur": 55,
    "args": {
      "External id": 137776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137776,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137776, "pid": 5, "tid": 7, "ts": 1716454223569461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441102, "dur": 547,
    "args": {
      "External id": 137776,
      "cbid": 211, "correlation": 137776
    }
  },
  {
    "ph": "s", "id": 137776, "pid": 76337, "tid": -914061504, "ts": 1716454223441102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223569518, "dur": 51,
    "args": {
      "External id": 137796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137796,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 137796, "pid": 5, "tid": 7, "ts": 1716454223569518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441704, "dur": 11,
    "args": {
      "External id": 137796,
      "cbid": 211, "correlation": 137796
    }
  },
  {
    "ph": "s", "id": 137796, "pid": 76337, "tid": -914061504, "ts": 1716454223441704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223569570, "dur": 4,
    "args": {
      "External id": 137808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137808,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137808, "pid": 5, "tid": 7, "ts": 1716454223569570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441724, "dur": 7,
    "args": {
      "External id": 137808,
      "cbid": 211, "correlation": 137808
    }
  },
  {
    "ph": "s", "id": 137808, "pid": 76337, "tid": -914061504, "ts": 1716454223441724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223569575, "dur": 55,
    "args": {
      "External id": 137811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137811,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137811, "pid": 5, "tid": 7, "ts": 1716454223569575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441743, "dur": 7,
    "args": {
      "External id": 137811,
      "cbid": 211, "correlation": 137811
    }
  },
  {
    "ph": "s", "id": 137811, "pid": 76337, "tid": -914061504, "ts": 1716454223441743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223569632, "dur": 37,
    "args": {
      "External id": 137820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137820,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137820, "pid": 5, "tid": 7, "ts": 1716454223569632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441784, "dur": 10,
    "args": {
      "External id": 137820,
      "cbid": 211, "correlation": 137820
    }
  },
  {
    "ph": "s", "id": 137820, "pid": 76337, "tid": -914061504, "ts": 1716454223441784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441846, "dur": 0,
    "args": {
      "External id": 137830,
      "cbid": 317, "correlation": 137830
    }
  },
  {
    "ph": "f", "id": 137830, "pid": 76337, "tid": -914061504, "ts": 1716454223441846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441847, "dur": 0,
    "args": {
      "External id": 137831,
      "cbid": 203, "correlation": 137831
    }
  },
  {
    "ph": "f", "id": 137831, "pid": 76337, "tid": -914061504, "ts": 1716454223441847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441848, "dur": 0,
    "args": {
      "External id": 137832,
      "cbid": 205, "correlation": 137832
    }
  },
  {
    "ph": "f", "id": 137832, "pid": 76337, "tid": -914061504, "ts": 1716454223441848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223569671, "dur": 40,
    "args": {
      "External id": 137836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137836,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137836, "pid": 5, "tid": 7, "ts": 1716454223569671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441862, "dur": 12,
    "args": {
      "External id": 137836,
      "cbid": 211, "correlation": 137836
    }
  },
  {
    "ph": "s", "id": 137836, "pid": 76337, "tid": -914061504, "ts": 1716454223441862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223569712, "dur": 14,
    "args": {
      "External id": 137838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137838, "pid": 5, "tid": 7, "ts": 1716454223569712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441876, "dur": 5,
    "args": {
      "External id": 137838,
      "cbid": 211, "correlation": 137838
    }
  },
  {
    "ph": "s", "id": 137838, "pid": 76337, "tid": -914061504, "ts": 1716454223441876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223569727, "dur": 3,
    "args": {
      "External id": 137840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137840,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137840, "pid": 5, "tid": 7, "ts": 1716454223569727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441886, "dur": 6,
    "args": {
      "External id": 137840,
      "cbid": 211, "correlation": 137840
    }
  },
  {
    "ph": "s", "id": 137840, "pid": 76337, "tid": -914061504, "ts": 1716454223441886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441894, "dur": 0,
    "args": {
      "External id": 137841,
      "cbid": 51, "correlation": 137841
    }
  },
  {
    "ph": "s", "id": 137841, "pid": 76337, "tid": -914061504, "ts": 1716454223441894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223569732, "dur": 703,
    "args": {
      "External id": 137842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137842,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137842, "pid": 5, "tid": 7, "ts": 1716454223569732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441895, "dur": 5,
    "args": {
      "External id": 137842,
      "cbid": 211, "correlation": 137842
    }
  },
  {
    "ph": "s", "id": 137842, "pid": 76337, "tid": -914061504, "ts": 1716454223441895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223570436, "dur": 60,
    "args": {
      "External id": 137847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137847,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137847, "pid": 5, "tid": 7, "ts": 1716454223570436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441922, "dur": 9,
    "args": {
      "External id": 137847,
      "cbid": 211, "correlation": 137847
    }
  },
  {
    "ph": "s", "id": 137847, "pid": 76337, "tid": -914061504, "ts": 1716454223441922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441987, "dur": 0,
    "args": {
      "External id": 137857,
      "cbid": 317, "correlation": 137857
    }
  },
  {
    "ph": "f", "id": 137857, "pid": 76337, "tid": -914061504, "ts": 1716454223441987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441988, "dur": 0,
    "args": {
      "External id": 137858,
      "cbid": 203, "correlation": 137858
    }
  },
  {
    "ph": "f", "id": 137858, "pid": 76337, "tid": -914061504, "ts": 1716454223441988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223441988, "dur": 0,
    "args": {
      "External id": 137859,
      "cbid": 205, "correlation": 137859
    }
  },
  {
    "ph": "f", "id": 137859, "pid": 76337, "tid": -914061504, "ts": 1716454223441988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223570497, "dur": 75,
    "args": {
      "External id": 137863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137863,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137863, "pid": 5, "tid": 7, "ts": 1716454223570497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442000, "dur": 12,
    "args": {
      "External id": 137863,
      "cbid": 211, "correlation": 137863
    }
  },
  {
    "ph": "s", "id": 137863, "pid": 76337, "tid": -914061504, "ts": 1716454223442000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223570574, "dur": 207,
    "args": {
      "External id": 137865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137865,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137865, "pid": 5, "tid": 7, "ts": 1716454223570574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442018, "dur": 6,
    "args": {
      "External id": 137865,
      "cbid": 211, "correlation": 137865
    }
  },
  {
    "ph": "s", "id": 137865, "pid": 76337, "tid": -914061504, "ts": 1716454223442018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223570782, "dur": 39,
    "args": {
      "External id": 137867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137867,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137867, "pid": 5, "tid": 7, "ts": 1716454223570782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442029, "dur": 6,
    "args": {
      "External id": 137867,
      "cbid": 211, "correlation": 137867
    }
  },
  {
    "ph": "s", "id": 137867, "pid": 76337, "tid": -914061504, "ts": 1716454223442029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223570822, "dur": 60,
    "args": {
      "External id": 137873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137873,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137873, "pid": 5, "tid": 7, "ts": 1716454223570822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442056, "dur": 509,
    "args": {
      "External id": 137873,
      "cbid": 211, "correlation": 137873
    }
  },
  {
    "ph": "s", "id": 137873, "pid": 76337, "tid": -914061504, "ts": 1716454223442056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223570883, "dur": 50,
    "args": {
      "External id": 137881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137881,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137881, "pid": 5, "tid": 7, "ts": 1716454223570883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442586, "dur": 8,
    "args": {
      "External id": 137881,
      "cbid": 211, "correlation": 137881
    }
  },
  {
    "ph": "s", "id": 137881, "pid": 76337, "tid": -914061504, "ts": 1716454223442586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223570934, "dur": 35,
    "args": {
      "External id": 137889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137889,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137889, "pid": 5, "tid": 7, "ts": 1716454223570934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442616, "dur": 9,
    "args": {
      "External id": 137889,
      "cbid": 211, "correlation": 137889
    }
  },
  {
    "ph": "s", "id": 137889, "pid": 76337, "tid": -914061504, "ts": 1716454223442616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223570971, "dur": 53,
    "args": {
      "External id": 137909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137909,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 137909, "pid": 5, "tid": 7, "ts": 1716454223570971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442700, "dur": 13,
    "args": {
      "External id": 137909,
      "cbid": 211, "correlation": 137909
    }
  },
  {
    "ph": "s", "id": 137909, "pid": 76337, "tid": -914061504, "ts": 1716454223442700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223571025, "dur": 4,
    "args": {
      "External id": 137921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137921,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 137921, "pid": 5, "tid": 7, "ts": 1716454223571025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442723, "dur": 6,
    "args": {
      "External id": 137921,
      "cbid": 211, "correlation": 137921
    }
  },
  {
    "ph": "s", "id": 137921, "pid": 76337, "tid": -914061504, "ts": 1716454223442723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223571030, "dur": 55,
    "args": {
      "External id": 137924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137924, "pid": 5, "tid": 7, "ts": 1716454223571030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442740, "dur": 7,
    "args": {
      "External id": 137924,
      "cbid": 211, "correlation": 137924
    }
  },
  {
    "ph": "s", "id": 137924, "pid": 76337, "tid": -914061504, "ts": 1716454223442740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442797, "dur": 0,
    "args": {
      "External id": 137935,
      "cbid": 317, "correlation": 137935
    }
  },
  {
    "ph": "f", "id": 137935, "pid": 76337, "tid": -914061504, "ts": 1716454223442797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442798, "dur": 0,
    "args": {
      "External id": 137936,
      "cbid": 203, "correlation": 137936
    }
  },
  {
    "ph": "f", "id": 137936, "pid": 76337, "tid": -914061504, "ts": 1716454223442798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442799, "dur": 0,
    "args": {
      "External id": 137937,
      "cbid": 205, "correlation": 137937
    }
  },
  {
    "ph": "f", "id": 137937, "pid": 76337, "tid": -914061504, "ts": 1716454223442799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442822, "dur": 1,
    "args": {
      "External id": 137941,
      "cbid": 251, "correlation": 137941
    }
  },
  {
    "ph": "f", "id": 137941, "pid": 76337, "tid": -914061504, "ts": 1716454223442822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442824, "dur": 0,
    "args": {
      "External id": 137942,
      "cbid": 251, "correlation": 137942
    }
  },
  {
    "ph": "f", "id": 137942, "pid": 76337, "tid": -914061504, "ts": 1716454223442824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442824, "dur": 0,
    "args": {
      "External id": 137943,
      "cbid": 251, "correlation": 137943
    }
  },
  {
    "ph": "f", "id": 137943, "pid": 76337, "tid": -914061504, "ts": 1716454223442824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442825, "dur": 0,
    "args": {
      "External id": 137944,
      "cbid": 251, "correlation": 137944
    }
  },
  {
    "ph": "f", "id": 137944, "pid": 76337, "tid": -914061504, "ts": 1716454223442825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442826, "dur": 0,
    "args": {
      "External id": 137945,
      "cbid": 251, "correlation": 137945
    }
  },
  {
    "ph": "f", "id": 137945, "pid": 76337, "tid": -914061504, "ts": 1716454223442826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442827, "dur": 0,
    "args": {
      "External id": 137946,
      "cbid": 251, "correlation": 137946
    }
  },
  {
    "ph": "f", "id": 137946, "pid": 76337, "tid": -914061504, "ts": 1716454223442827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442827, "dur": 0,
    "args": {
      "External id": 137947,
      "cbid": 251, "correlation": 137947
    }
  },
  {
    "ph": "f", "id": 137947, "pid": 76337, "tid": -914061504, "ts": 1716454223442827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442828, "dur": 0,
    "args": {
      "External id": 137948,
      "cbid": 251, "correlation": 137948
    }
  },
  {
    "ph": "f", "id": 137948, "pid": 76337, "tid": -914061504, "ts": 1716454223442828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442829, "dur": 0,
    "args": {
      "External id": 137949,
      "cbid": 251, "correlation": 137949
    }
  },
  {
    "ph": "f", "id": 137949, "pid": 76337, "tid": -914061504, "ts": 1716454223442829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223571087, "dur": 116,
    "args": {
      "External id": 137950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137950,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 137950, "pid": 5, "tid": 7, "ts": 1716454223571087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442831, "dur": 12,
    "args": {
      "External id": 137950,
      "cbid": 211, "correlation": 137950
    }
  },
  {
    "ph": "s", "id": 137950, "pid": 76337, "tid": -914061504, "ts": 1716454223442831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223571204, "dur": 59,
    "args": {
      "External id": 137956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137956, "pid": 5, "tid": 7, "ts": 1716454223571204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442866, "dur": 9,
    "args": {
      "External id": 137956,
      "cbid": 211, "correlation": 137956
    }
  },
  {
    "ph": "s", "id": 137956, "pid": 76337, "tid": -914061504, "ts": 1716454223442866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223571265, "dur": 598,
    "args": {
      "External id": 137965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137965,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137965, "pid": 5, "tid": 7, "ts": 1716454223571265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223442949, "dur": 14,
    "args": {
      "External id": 137965,
      "cbid": 211, "correlation": 137965
    }
  },
  {
    "ph": "s", "id": 137965, "pid": 76337, "tid": -914061504, "ts": 1716454223442949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223571865, "dur": 184,
    "args": {
      "External id": 137987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137987,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 137987, "pid": 5, "tid": 7, "ts": 1716454223571865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443017, "dur": 12,
    "args": {
      "External id": 137987,
      "cbid": 211, "correlation": 137987
    }
  },
  {
    "ph": "s", "id": 137987, "pid": 76337, "tid": -914061504, "ts": 1716454223443017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443103, "dur": 1,
    "args": {
      "External id": 137998,
      "cbid": 251, "correlation": 137998
    }
  },
  {
    "ph": "f", "id": 137998, "pid": 76337, "tid": -914061504, "ts": 1716454223443103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223572051, "dur": 198,
    "args": {
      "External id": 137999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 137999,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 137999, "pid": 5, "tid": 7, "ts": 1716454223572051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443108, "dur": 14,
    "args": {
      "External id": 137999,
      "cbid": 211, "correlation": 137999
    }
  },
  {
    "ph": "s", "id": 137999, "pid": 76337, "tid": -914061504, "ts": 1716454223443108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443177, "dur": 1,
    "args": {
      "External id": 138010,
      "cbid": 251, "correlation": 138010
    }
  },
  {
    "ph": "f", "id": 138010, "pid": 76337, "tid": -914061504, "ts": 1716454223443177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223572250, "dur": 190,
    "args": {
      "External id": 138011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138011,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 138011, "pid": 5, "tid": 7, "ts": 1716454223572250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443181, "dur": 11,
    "args": {
      "External id": 138011,
      "cbid": 211, "correlation": 138011
    }
  },
  {
    "ph": "s", "id": 138011, "pid": 76337, "tid": -914061504, "ts": 1716454223443181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443244, "dur": 1,
    "args": {
      "External id": 138022,
      "cbid": 251, "correlation": 138022
    }
  },
  {
    "ph": "f", "id": 138022, "pid": 76337, "tid": -914061504, "ts": 1716454223443244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223572442, "dur": 187,
    "args": {
      "External id": 138023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138023,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 138023, "pid": 5, "tid": 7, "ts": 1716454223572442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443247, "dur": 11,
    "args": {
      "External id": 138023,
      "cbid": 211, "correlation": 138023
    }
  },
  {
    "ph": "s", "id": 138023, "pid": 76337, "tid": -914061504, "ts": 1716454223443247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223572630, "dur": 18954,
    "args": {
      "External id": 138044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138044,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 138044, "pid": 5, "tid": 7, "ts": 1716454223572630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443328, "dur": 12,
    "args": {
      "External id": 138044,
      "cbid": 211, "correlation": 138044
    }
  },
  {
    "ph": "s", "id": 138044, "pid": 76337, "tid": -914061504, "ts": 1716454223443328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443423, "dur": 1,
    "args": {
      "External id": 138062,
      "cbid": 251, "correlation": 138062
    }
  },
  {
    "ph": "f", "id": 138062, "pid": 76337, "tid": -914061504, "ts": 1716454223443423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223591586, "dur": 205,
    "args": {
      "External id": 138064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138064,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 138064, "pid": 5, "tid": 7, "ts": 1716454223591586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443429, "dur": 13,
    "args": {
      "External id": 138064,
      "cbid": 211, "correlation": 138064
    }
  },
  {
    "ph": "s", "id": 138064, "pid": 76337, "tid": -914061504, "ts": 1716454223443429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223591792, "dur": 66,
    "args": {
      "External id": 138072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138072,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138072, "pid": 5, "tid": 7, "ts": 1716454223591792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443498, "dur": 12,
    "args": {
      "External id": 138072,
      "cbid": 211, "correlation": 138072
    }
  },
  {
    "ph": "s", "id": 138072, "pid": 76337, "tid": -914061504, "ts": 1716454223443498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223591860, "dur": 97,
    "args": {
      "External id": 138080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138080,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138080, "pid": 5, "tid": 7, "ts": 1716454223591860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443537, "dur": 78,
    "args": {
      "External id": 138080,
      "cbid": 211, "correlation": 138080
    }
  },
  {
    "ph": "s", "id": 138080, "pid": 76337, "tid": -914061504, "ts": 1716454223443537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223591958, "dur": 55,
    "args": {
      "External id": 138091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138091, "pid": 5, "tid": 7, "ts": 1716454223591958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223443679, "dur": 1877,
    "args": {
      "External id": 138091,
      "cbid": 211, "correlation": 138091
    }
  },
  {
    "ph": "s", "id": 138091, "pid": 76337, "tid": -914061504, "ts": 1716454223443679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223592014, "dur": 94,
    "args": {
      "External id": 138113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138113,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138113, "pid": 5, "tid": 7, "ts": 1716454223592014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445576, "dur": 126,
    "args": {
      "External id": 138113,
      "cbid": 211, "correlation": 138113
    }
  },
  {
    "ph": "s", "id": 138113, "pid": 76337, "tid": -914061504, "ts": 1716454223445576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445778, "dur": 1,
    "args": {
      "External id": 138124,
      "cbid": 251, "correlation": 138124
    }
  },
  {
    "ph": "f", "id": 138124, "pid": 76337, "tid": -914061504, "ts": 1716454223445778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223592109, "dur": 109,
    "args": {
      "External id": 138125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138125,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 138125, "pid": 5, "tid": 7, "ts": 1716454223592109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445783, "dur": 14,
    "args": {
      "External id": 138125,
      "cbid": 211, "correlation": 138125
    }
  },
  {
    "ph": "s", "id": 138125, "pid": 76337, "tid": -914061504, "ts": 1716454223445783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445856, "dur": 1,
    "args": {
      "External id": 138136,
      "cbid": 251, "correlation": 138136
    }
  },
  {
    "ph": "f", "id": 138136, "pid": 76337, "tid": -914061504, "ts": 1716454223445856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445859, "dur": 0,
    "args": {
      "External id": 138137,
      "cbid": 251, "correlation": 138137
    }
  },
  {
    "ph": "f", "id": 138137, "pid": 76337, "tid": -914061504, "ts": 1716454223445859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223592219, "dur": 11,
    "args": {
      "External id": 138138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138138,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138138, "pid": 5, "tid": 7, "ts": 1716454223592219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445861, "dur": 12,
    "args": {
      "External id": 138138,
      "cbid": 211, "correlation": 138138
    }
  },
  {
    "ph": "s", "id": 138138, "pid": 76337, "tid": -914061504, "ts": 1716454223445861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223592232, "dur": 5,
    "args": {
      "External id": 138140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138140,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 138140, "pid": 5, "tid": 7, "ts": 1716454223592232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445875, "dur": 6,
    "args": {
      "External id": 138140,
      "cbid": 211, "correlation": 138140
    }
  },
  {
    "ph": "s", "id": 138140, "pid": 76337, "tid": -914061504, "ts": 1716454223445875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445936, "dur": 1,
    "args": {
      "External id": 138151,
      "cbid": 251, "correlation": 138151
    }
  },
  {
    "ph": "f", "id": 138151, "pid": 76337, "tid": -914061504, "ts": 1716454223445936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445939, "dur": 0,
    "args": {
      "External id": 138152,
      "cbid": 251, "correlation": 138152
    }
  },
  {
    "ph": "f", "id": 138152, "pid": 76337, "tid": -914061504, "ts": 1716454223445939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223592238, "dur": 6,
    "args": {
      "External id": 138153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138153,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138153, "pid": 5, "tid": 7, "ts": 1716454223592238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445941, "dur": 12,
    "args": {
      "External id": 138153,
      "cbid": 211, "correlation": 138153
    }
  },
  {
    "ph": "s", "id": 138153, "pid": 76337, "tid": -914061504, "ts": 1716454223445941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223592246, "dur": 4,
    "args": {
      "External id": 138155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138155,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 138155, "pid": 5, "tid": 7, "ts": 1716454223592246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223445955, "dur": 6,
    "args": {
      "External id": 138155,
      "cbid": 211, "correlation": 138155
    }
  },
  {
    "ph": "s", "id": 138155, "pid": 76337, "tid": -914061504, "ts": 1716454223445955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223592251, "dur": 159,
    "args": {
      "External id": 138176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138176,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 138176, "pid": 5, "tid": 7, "ts": 1716454223592251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446036, "dur": 13,
    "args": {
      "External id": 138176,
      "cbid": 211, "correlation": 138176
    }
  },
  {
    "ph": "s", "id": 138176, "pid": 76337, "tid": -914061504, "ts": 1716454223446036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446131, "dur": 1,
    "args": {
      "External id": 138194,
      "cbid": 251, "correlation": 138194
    }
  },
  {
    "ph": "f", "id": 138194, "pid": 76337, "tid": -914061504, "ts": 1716454223446131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223592411, "dur": 108,
    "args": {
      "External id": 138196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138196,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 138196, "pid": 5, "tid": 7, "ts": 1716454223592411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446137, "dur": 14,
    "args": {
      "External id": 138196,
      "cbid": 211, "correlation": 138196
    }
  },
  {
    "ph": "s", "id": 138196, "pid": 76337, "tid": -914061504, "ts": 1716454223446137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223592520, "dur": 34,
    "args": {
      "External id": 138204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138204,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138204, "pid": 5, "tid": 7, "ts": 1716454223592520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446207, "dur": 12,
    "args": {
      "External id": 138204,
      "cbid": 211, "correlation": 138204
    }
  },
  {
    "ph": "s", "id": 138204, "pid": 76337, "tid": -914061504, "ts": 1716454223446207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223592556, "dur": 68,
    "args": {
      "External id": 138212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138212, "pid": 5, "tid": 7, "ts": 1716454223592556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446248, "dur": 9,
    "args": {
      "External id": 138212,
      "cbid": 211, "correlation": 138212
    }
  },
  {
    "ph": "s", "id": 138212, "pid": 76337, "tid": -914061504, "ts": 1716454223446248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223592626, "dur": 94,
    "args": {
      "External id": 138234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138234,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138234, "pid": 5, "tid": 7, "ts": 1716454223592626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446299, "dur": 10,
    "args": {
      "External id": 138234,
      "cbid": 211, "correlation": 138234
    }
  },
  {
    "ph": "s", "id": 138234, "pid": 76337, "tid": -914061504, "ts": 1716454223446299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446384, "dur": 1,
    "args": {
      "External id": 138250,
      "cbid": 251, "correlation": 138250
    }
  },
  {
    "ph": "f", "id": 138250, "pid": 76337, "tid": -914061504, "ts": 1716454223446384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223592721, "dur": 584,
    "args": {
      "External id": 138252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138252,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 138252, "pid": 5, "tid": 7, "ts": 1716454223592721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446390, "dur": 13,
    "args": {
      "External id": 138252,
      "cbid": 211, "correlation": 138252
    }
  },
  {
    "ph": "s", "id": 138252, "pid": 76337, "tid": -914061504, "ts": 1716454223446390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223593306, "dur": 247,
    "args": {
      "External id": 138260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138260,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138260, "pid": 5, "tid": 7, "ts": 1716454223593306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446456, "dur": 12,
    "args": {
      "External id": 138260,
      "cbid": 211, "correlation": 138260
    }
  },
  {
    "ph": "s", "id": 138260, "pid": 76337, "tid": -914061504, "ts": 1716454223446456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223593555, "dur": 253,
    "args": {
      "External id": 138268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138268,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138268, "pid": 5, "tid": 7, "ts": 1716454223593555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446485, "dur": 8,
    "args": {
      "External id": 138268,
      "cbid": 211, "correlation": 138268
    }
  },
  {
    "ph": "s", "id": 138268, "pid": 76337, "tid": -914061504, "ts": 1716454223446485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446566, "dur": 1,
    "args": {
      "External id": 138284,
      "cbid": 251, "correlation": 138284
    }
  },
  {
    "ph": "f", "id": 138284, "pid": 76337, "tid": -914061504, "ts": 1716454223446566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446571, "dur": 0,
    "args": {
      "External id": 138286,
      "cbid": 251, "correlation": 138286
    }
  },
  {
    "ph": "f", "id": 138286, "pid": 76337, "tid": -914061504, "ts": 1716454223446571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223593810, "dur": 362,
    "args": {
      "External id": 138287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138287,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138287, "pid": 5, "tid": 7, "ts": 1716454223593810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446574, "dur": 12,
    "args": {
      "External id": 138287,
      "cbid": 211, "correlation": 138287
    }
  },
  {
    "ph": "s", "id": 138287, "pid": 76337, "tid": -914061504, "ts": 1716454223446574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223594174, "dur": 50,
    "args": {
      "External id": 138295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138295,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138295, "pid": 5, "tid": 7, "ts": 1716454223594174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446616, "dur": 189,
    "args": {
      "External id": 138295,
      "cbid": 211, "correlation": 138295
    }
  },
  {
    "ph": "s", "id": 138295, "pid": 76337, "tid": -914061504, "ts": 1716454223446616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223594225, "dur": 161,
    "args": {
      "External id": 138306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138306, "pid": 5, "tid": 7, "ts": 1716454223594225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446864, "dur": 69,
    "args": {
      "External id": 138306,
      "cbid": 211, "correlation": 138306
    }
  },
  {
    "ph": "s", "id": 138306, "pid": 76337, "tid": -914061504, "ts": 1716454223446864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446996, "dur": 0,
    "args": {
      "External id": 138318,
      "cbid": 317, "correlation": 138318
    }
  },
  {
    "ph": "f", "id": 138318, "pid": 76337, "tid": -914061504, "ts": 1716454223446996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446996, "dur": 0,
    "args": {
      "External id": 138319,
      "cbid": 203, "correlation": 138319
    }
  },
  {
    "ph": "f", "id": 138319, "pid": 76337, "tid": -914061504, "ts": 1716454223446996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223446997, "dur": 0,
    "args": {
      "External id": 138320,
      "cbid": 205, "correlation": 138320
    }
  },
  {
    "ph": "f", "id": 138320, "pid": 76337, "tid": -914061504, "ts": 1716454223446997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447020, "dur": 1,
    "args": {
      "External id": 138324,
      "cbid": 251, "correlation": 138324
    }
  },
  {
    "ph": "f", "id": 138324, "pid": 76337, "tid": -914061504, "ts": 1716454223447020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447022, "dur": 0,
    "args": {
      "External id": 138325,
      "cbid": 251, "correlation": 138325
    }
  },
  {
    "ph": "f", "id": 138325, "pid": 76337, "tid": -914061504, "ts": 1716454223447022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447022, "dur": 0,
    "args": {
      "External id": 138326,
      "cbid": 251, "correlation": 138326
    }
  },
  {
    "ph": "f", "id": 138326, "pid": 76337, "tid": -914061504, "ts": 1716454223447022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447023, "dur": 0,
    "args": {
      "External id": 138327,
      "cbid": 251, "correlation": 138327
    }
  },
  {
    "ph": "f", "id": 138327, "pid": 76337, "tid": -914061504, "ts": 1716454223447023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447024, "dur": 0,
    "args": {
      "External id": 138328,
      "cbid": 251, "correlation": 138328
    }
  },
  {
    "ph": "f", "id": 138328, "pid": 76337, "tid": -914061504, "ts": 1716454223447024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447025, "dur": 0,
    "args": {
      "External id": 138329,
      "cbid": 251, "correlation": 138329
    }
  },
  {
    "ph": "f", "id": 138329, "pid": 76337, "tid": -914061504, "ts": 1716454223447025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447026, "dur": 0,
    "args": {
      "External id": 138330,
      "cbid": 251, "correlation": 138330
    }
  },
  {
    "ph": "f", "id": 138330, "pid": 76337, "tid": -914061504, "ts": 1716454223447026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447027, "dur": 0,
    "args": {
      "External id": 138331,
      "cbid": 251, "correlation": 138331
    }
  },
  {
    "ph": "f", "id": 138331, "pid": 76337, "tid": -914061504, "ts": 1716454223447027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447028, "dur": 0,
    "args": {
      "External id": 138332,
      "cbid": 251, "correlation": 138332
    }
  },
  {
    "ph": "f", "id": 138332, "pid": 76337, "tid": -914061504, "ts": 1716454223447028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223594388, "dur": 118,
    "args": {
      "External id": 138333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138333,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 138333, "pid": 5, "tid": 7, "ts": 1716454223594388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447030, "dur": 33,
    "args": {
      "External id": 138333,
      "cbid": 211, "correlation": 138333
    }
  },
  {
    "ph": "s", "id": 138333, "pid": 76337, "tid": -914061504, "ts": 1716454223447030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223594508, "dur": 60,
    "args": {
      "External id": 138339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138339,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138339, "pid": 5, "tid": 7, "ts": 1716454223594508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447085, "dur": 279,
    "args": {
      "External id": 138339,
      "cbid": 211, "correlation": 138339
    }
  },
  {
    "ph": "s", "id": 138339, "pid": 76337, "tid": -914061504, "ts": 1716454223447085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223594569, "dur": 50,
    "args": {
      "External id": 138347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138347,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138347, "pid": 5, "tid": 7, "ts": 1716454223594569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447388, "dur": 8,
    "args": {
      "External id": 138347,
      "cbid": 211, "correlation": 138347
    }
  },
  {
    "ph": "s", "id": 138347, "pid": 76337, "tid": -914061504, "ts": 1716454223447388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223594620, "dur": 53,
    "args": {
      "External id": 138367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138367,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 138367, "pid": 5, "tid": 7, "ts": 1716454223594620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447467, "dur": 12,
    "args": {
      "External id": 138367,
      "cbid": 211, "correlation": 138367
    }
  },
  {
    "ph": "s", "id": 138367, "pid": 76337, "tid": -914061504, "ts": 1716454223447467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223594674, "dur": 4,
    "args": {
      "External id": 138379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138379,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 138379, "pid": 5, "tid": 7, "ts": 1716454223594674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447490, "dur": 11,
    "args": {
      "External id": 138379,
      "cbid": 211, "correlation": 138379
    }
  },
  {
    "ph": "s", "id": 138379, "pid": 76337, "tid": -914061504, "ts": 1716454223447490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223594680, "dur": 55,
    "args": {
      "External id": 138382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138382,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138382, "pid": 5, "tid": 7, "ts": 1716454223594680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447512, "dur": 108,
    "args": {
      "External id": 138382,
      "cbid": 211, "correlation": 138382
    }
  },
  {
    "ph": "s", "id": 138382, "pid": 76337, "tid": -914061504, "ts": 1716454223447512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223594737, "dur": 37,
    "args": {
      "External id": 138391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138391,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138391, "pid": 5, "tid": 7, "ts": 1716454223594737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447659, "dur": 10,
    "args": {
      "External id": 138391,
      "cbid": 211, "correlation": 138391
    }
  },
  {
    "ph": "s", "id": 138391, "pid": 76337, "tid": -914061504, "ts": 1716454223447659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447714, "dur": 0,
    "args": {
      "External id": 138401,
      "cbid": 317, "correlation": 138401
    }
  },
  {
    "ph": "f", "id": 138401, "pid": 76337, "tid": -914061504, "ts": 1716454223447714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447715, "dur": 0,
    "args": {
      "External id": 138402,
      "cbid": 203, "correlation": 138402
    }
  },
  {
    "ph": "f", "id": 138402, "pid": 76337, "tid": -914061504, "ts": 1716454223447715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447716, "dur": 0,
    "args": {
      "External id": 138403,
      "cbid": 205, "correlation": 138403
    }
  },
  {
    "ph": "f", "id": 138403, "pid": 76337, "tid": -914061504, "ts": 1716454223447716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223594775, "dur": 41,
    "args": {
      "External id": 138407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138407,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138407, "pid": 5, "tid": 7, "ts": 1716454223594775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447731, "dur": 12,
    "args": {
      "External id": 138407,
      "cbid": 211, "correlation": 138407
    }
  },
  {
    "ph": "s", "id": 138407, "pid": 76337, "tid": -914061504, "ts": 1716454223447731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223594818, "dur": 3,
    "args": {
      "External id": 138409,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138409,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.203389830508474
    }
  },
  {
    "ph": "f", "id": 138409, "pid": 5, "tid": 7, "ts": 1716454223594818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447747, "dur": 17,
    "args": {
      "External id": 138409,
      "cbid": 51, "correlation": 138409
    }
  },
  {
    "ph": "s", "id": 138409, "pid": 76337, "tid": -914061504, "ts": 1716454223447747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447769, "dur": 1,
    "args": {
      "External id": 138411,
      "cbid": 200, "correlation": 138411
    }
  },
  {
    "ph": "f", "id": 138411, "pid": 76337, "tid": -914061504, "ts": 1716454223447769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447771, "dur": 0,
    "args": {
      "External id": 138412,
      "cbid": 200, "correlation": 138412
    }
  },
  {
    "ph": "f", "id": 138412, "pid": 76337, "tid": -914061504, "ts": 1716454223447771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447772, "dur": 0,
    "args": {
      "External id": 138413,
      "cbid": 200, "correlation": 138413
    }
  },
  {
    "ph": "f", "id": 138413, "pid": 76337, "tid": -914061504, "ts": 1716454223447772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447772, "dur": 0,
    "args": {
      "External id": 138414,
      "cbid": 200, "correlation": 138414
    }
  },
  {
    "ph": "f", "id": 138414, "pid": 76337, "tid": -914061504, "ts": 1716454223447772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447773, "dur": 4,
    "args": {
      "External id": 138415,
      "cbid": 15, "correlation": 138415
    }
  },
  {
    "ph": "f", "id": 138415, "pid": 76337, "tid": -914061504, "ts": 1716454223447773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447778, "dur": 1,
    "args": {
      "External id": 138416,
      "cbid": 251, "correlation": 138416
    }
  },
  {
    "ph": "f", "id": 138416, "pid": 76337, "tid": -914061504, "ts": 1716454223447778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223594823, "dur": 24,
    "args": {
      "External id": 138417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138417,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138417, "pid": 5, "tid": 7, "ts": 1716454223594823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447781, "dur": 8,
    "args": {
      "External id": 138417,
      "cbid": 211, "correlation": 138417
    }
  },
  {
    "ph": "s", "id": 138417, "pid": 76337, "tid": -914061504, "ts": 1716454223447781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223594848, "dur": 4,
    "args": {
      "External id": 138419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 138419, "pid": 5, "tid": 7, "ts": 1716454223594848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447795, "dur": 6,
    "args": {
      "External id": 138419,
      "cbid": 211, "correlation": 138419
    }
  },
  {
    "ph": "s", "id": 138419, "pid": 76337, "tid": -914061504, "ts": 1716454223447795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447804, "dur": 0,
    "args": {
      "External id": 138420,
      "cbid": 51, "correlation": 138420
    }
  },
  {
    "ph": "s", "id": 138420, "pid": 76337, "tid": -914061504, "ts": 1716454223447804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223594853, "dur": 191,
    "args": {
      "External id": 138421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138421,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138421, "pid": 5, "tid": 7, "ts": 1716454223594853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223447805, "dur": 209,
    "args": {
      "External id": 138421,
      "cbid": 211, "correlation": 138421
    }
  },
  {
    "ph": "s", "id": 138421, "pid": 76337, "tid": -914061504, "ts": 1716454223447805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223595045, "dur": 6,
    "args": {
      "External id": 138422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138422,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138422, "pid": 5, "tid": 7, "ts": 1716454223595045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223448018, "dur": 6,
    "args": {
      "External id": 138422,
      "cbid": 211, "correlation": 138422
    }
  },
  {
    "ph": "s", "id": 138422, "pid": 76337, "tid": -914061504, "ts": 1716454223448018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223595053, "dur": 5,
    "args": {
      "External id": 138428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 138428, "pid": 5, "tid": 7, "ts": 1716454223595053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223448049, "dur": 9,
    "args": {
      "External id": 138428,
      "cbid": 211, "correlation": 138428
    }
  },
  {
    "ph": "s", "id": 138428, "pid": 76337, "tid": -914061504, "ts": 1716454223448049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223595060, "dur": 3,
    "args": {
      "External id": 138436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138436,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138436, "pid": 5, "tid": 7, "ts": 1716454223595060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223449717, "dur": 21,
    "args": {
      "External id": 138436,
      "cbid": 211, "correlation": 138436
    }
  },
  {
    "ph": "s", "id": 138436, "pid": 76337, "tid": -914061504, "ts": 1716454223449717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223595064, "dur": 3,
    "args": {
      "External id": 138444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138444,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138444, "pid": 5, "tid": 7, "ts": 1716454223595064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223449766, "dur": 12,
    "args": {
      "External id": 138444,
      "cbid": 211, "correlation": 138444
    }
  },
  {
    "ph": "s", "id": 138444, "pid": 76337, "tid": -914061504, "ts": 1716454223449766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223595068, "dur": 3,
    "args": {
      "External id": 138452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138452,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138452, "pid": 5, "tid": 7, "ts": 1716454223595068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223449797, "dur": 8,
    "args": {
      "External id": 138452,
      "cbid": 211, "correlation": 138452
    }
  },
  {
    "ph": "s", "id": 138452, "pid": 76337, "tid": -914061504, "ts": 1716454223449797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223595072, "dur": 3,
    "args": {
      "External id": 138461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138461,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138461, "pid": 5, "tid": 7, "ts": 1716454223595072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223449987, "dur": 14,
    "args": {
      "External id": 138461,
      "cbid": 211, "correlation": 138461
    }
  },
  {
    "ph": "s", "id": 138461, "pid": 76337, "tid": -914061504, "ts": 1716454223449987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223595076, "dur": 3,
    "args": {
      "External id": 138470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138470,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138470, "pid": 5, "tid": 7, "ts": 1716454223595076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223450017, "dur": 7,
    "args": {
      "External id": 138470,
      "cbid": 211, "correlation": 138470
    }
  },
  {
    "ph": "s", "id": 138470, "pid": 76337, "tid": -914061504, "ts": 1716454223450017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223595080, "dur": 3,
    "args": {
      "External id": 138478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138478,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138478, "pid": 5, "tid": 7, "ts": 1716454223595080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223450045, "dur": 9,
    "args": {
      "External id": 138478,
      "cbid": 211, "correlation": 138478
    }
  },
  {
    "ph": "s", "id": 138478, "pid": 76337, "tid": -914061504, "ts": 1716454223450045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223595085, "dur": 3,
    "args": {
      "External id": 138486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138486,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138486, "pid": 5, "tid": 7, "ts": 1716454223595085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223450306, "dur": 15,
    "args": {
      "External id": 138486,
      "cbid": 211, "correlation": 138486
    }
  },
  {
    "ph": "s", "id": 138486, "pid": 76337, "tid": -914061504, "ts": 1716454223450306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223595089, "dur": 3,
    "args": {
      "External id": 138494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138494,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138494, "pid": 5, "tid": 7, "ts": 1716454223595089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223450336, "dur": 8,
    "args": {
      "External id": 138494,
      "cbid": 211, "correlation": 138494
    }
  },
  {
    "ph": "s", "id": 138494, "pid": 76337, "tid": -914061504, "ts": 1716454223450336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223595094, "dur": 1,
    "args": {
      "External id": 138504,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138504,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 138504, "pid": 5, "tid": 7, "ts": 1716454223595094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223450405, "dur": 36,
    "args": {
      "External id": 138504,
      "cbid": 41, "correlation": 138504
    }
  },
  {
    "ph": "s", "id": 138504, "pid": 76337, "tid": -914061504, "ts": 1716454223450405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223450442, "dur": 144670,
    "args": {
      "External id": 138505,
      "cbid": 131, "correlation": 138505
    }
  },
  {
    "ph": "f", "id": 138505, "pid": 76337, "tid": -914061504, "ts": 1716454223450442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223595305, "dur": 3,
    "args": {
      "External id": 138513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138513,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138513, "pid": 5, "tid": 7, "ts": 1716454223595305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595273, "dur": 35,
    "args": {
      "External id": 138513,
      "cbid": 211, "correlation": 138513
    }
  },
  {
    "ph": "s", "id": 138513, "pid": 76337, "tid": -914061504, "ts": 1716454223595273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223595409, "dur": 3,
    "args": {
      "External id": 138522,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138522,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002380952380952381
    }
  },
  {
    "ph": "f", "id": 138522, "pid": 5, "tid": 7, "ts": 1716454223595409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595372, "dur": 38,
    "args": {
      "External id": 138522,
      "cbid": 41, "correlation": 138522
    }
  },
  {
    "ph": "s", "id": 138522, "pid": 76337, "tid": -914061504, "ts": 1716454223595372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223595511, "dur": 4,
    "args": {
      "External id": 138532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138532,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138532, "pid": 5, "tid": 7, "ts": 1716454223595511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595495, "dur": 17,
    "args": {
      "External id": 138532,
      "cbid": 211, "correlation": 138532
    }
  },
  {
    "ph": "s", "id": 138532, "pid": 76337, "tid": -914061504, "ts": 1716454223595495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223595594, "dur": 1,
    "args": {
      "External id": 138542,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138542,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 138542, "pid": 5, "tid": 7, "ts": 1716454223595594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595574, "dur": 18,
    "args": {
      "External id": 138542,
      "cbid": 41, "correlation": 138542
    }
  },
  {
    "ph": "s", "id": 138542, "pid": 76337, "tid": -914061504, "ts": 1716454223595574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595593, "dur": 8,
    "args": {
      "External id": 138543,
      "cbid": 131, "correlation": 138543
    }
  },
  {
    "ph": "f", "id": 138543, "pid": 76337, "tid": -914061504, "ts": 1716454223595593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223595672, "dur": 3,
    "args": {
      "External id": 138550,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138550,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.11764705882353
    }
  },
  {
    "ph": "f", "id": 138550, "pid": 5, "tid": 7, "ts": 1716454223595672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595645, "dur": 26,
    "args": {
      "External id": 138550,
      "cbid": 41, "correlation": 138550
    }
  },
  {
    "ph": "s", "id": 138550, "pid": 76337, "tid": -914061504, "ts": 1716454223595645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223595764, "dur": 3,
    "args": {
      "External id": 138569,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138569,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005263157894736842
    }
  },
  {
    "ph": "f", "id": 138569, "pid": 5, "tid": 7, "ts": 1716454223595764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595745, "dur": 18,
    "args": {
      "External id": 138569,
      "cbid": 41, "correlation": 138569
    }
  },
  {
    "ph": "s", "id": 138569, "pid": 76337, "tid": -914061504, "ts": 1716454223595745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223595804, "dur": 3,
    "args": {
      "External id": 138575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138575, "pid": 5, "tid": 7, "ts": 1716454223595804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595793, "dur": 12,
    "args": {
      "External id": 138575,
      "cbid": 211, "correlation": 138575
    }
  },
  {
    "ph": "s", "id": 138575, "pid": 76337, "tid": -914061504, "ts": 1716454223595793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223595819, "dur": 6,
    "args": {
      "External id": 138577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138577,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 138577, "pid": 5, "tid": 7, "ts": 1716454223595819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595809, "dur": 9,
    "args": {
      "External id": 138577,
      "cbid": 211, "correlation": 138577
    }
  },
  {
    "ph": "s", "id": 138577, "pid": 76337, "tid": -914061504, "ts": 1716454223595809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454223595827, "dur": 3,
    "args": {
      "External id": 138579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138579,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138579, "pid": 5, "tid": 7, "ts": 1716454223595827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595819, "dur": 7,
    "args": {
      "External id": 138579,
      "cbid": 211, "correlation": 138579
    }
  },
  {
    "ph": "s", "id": 138579, "pid": 76337, "tid": -914061504, "ts": 1716454223595819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223595862, "dur": 2,
    "args": {
      "External id": 138587,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138587,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 138587, "pid": 5, "tid": 7, "ts": 1716454223595862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595848, "dur": 13,
    "args": {
      "External id": 138587,
      "cbid": 41, "correlation": 138587
    }
  },
  {
    "ph": "s", "id": 138587, "pid": 76337, "tid": -914061504, "ts": 1716454223595848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223595912, "dur": 3,
    "args": {
      "External id": 138601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138601,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138601, "pid": 5, "tid": 7, "ts": 1716454223595912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595901, "dur": 12,
    "args": {
      "External id": 138601,
      "cbid": 211, "correlation": 138601
    }
  },
  {
    "ph": "s", "id": 138601, "pid": 76337, "tid": -914061504, "ts": 1716454223595901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223595931, "dur": 2,
    "args": {
      "External id": 138615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138615,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138615, "pid": 5, "tid": 7, "ts": 1716454223595931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595924, "dur": 6,
    "args": {
      "External id": 138615,
      "cbid": 211, "correlation": 138615
    }
  },
  {
    "ph": "s", "id": 138615, "pid": 76337, "tid": -914061504, "ts": 1716454223595924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223595969, "dur": 6,
    "args": {
      "External id": 138622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138622,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138622, "pid": 5, "tid": 7, "ts": 1716454223595969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595959, "dur": 11,
    "args": {
      "External id": 138622,
      "cbid": 211, "correlation": 138622
    }
  },
  {
    "ph": "s", "id": 138622, "pid": 76337, "tid": -914061504, "ts": 1716454223595959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223595988, "dur": 6,
    "args": {
      "External id": 138625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138625,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138625, "pid": 5, "tid": 7, "ts": 1716454223595988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595980, "dur": 8,
    "args": {
      "External id": 138625,
      "cbid": 211, "correlation": 138625
    }
  },
  {
    "ph": "s", "id": 138625, "pid": 76337, "tid": -914061504, "ts": 1716454223595980,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223595998, "dur": 3,
    "args": {
      "External id": 138627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138627,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138627, "pid": 5, "tid": 7, "ts": 1716454223595998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223595990, "dur": 7,
    "args": {
      "External id": 138627,
      "cbid": 211, "correlation": 138627
    }
  },
  {
    "ph": "s", "id": 138627, "pid": 76337, "tid": -914061504, "ts": 1716454223595990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223596019, "dur": 2,
    "args": {
      "External id": 138630,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138630,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002717391304347826
    }
  },
  {
    "ph": "f", "id": 138630, "pid": 5, "tid": 7, "ts": 1716454223596019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596006, "dur": 12,
    "args": {
      "External id": 138630,
      "cbid": 41, "correlation": 138630
    }
  },
  {
    "ph": "s", "id": 138630, "pid": 76337, "tid": -914061504, "ts": 1716454223596006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223596072, "dur": 4,
    "args": {
      "External id": 138646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138646,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138646, "pid": 5, "tid": 7, "ts": 1716454223596072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596059, "dur": 13,
    "args": {
      "External id": 138646,
      "cbid": 211, "correlation": 138646
    }
  },
  {
    "ph": "s", "id": 138646, "pid": 76337, "tid": -914061504, "ts": 1716454223596059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223596092, "dur": 3,
    "args": {
      "External id": 138651,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138651,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003094059405940594
    }
  },
  {
    "ph": "f", "id": 138651, "pid": 5, "tid": 7, "ts": 1716454223596092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596077, "dur": 14,
    "args": {
      "External id": 138651,
      "cbid": 41, "correlation": 138651
    }
  },
  {
    "ph": "s", "id": 138651, "pid": 76337, "tid": -914061504, "ts": 1716454223596077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223596119, "dur": 1,
    "args": {
      "External id": 138657,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138657,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 138657, "pid": 5, "tid": 7, "ts": 1716454223596119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596101, "dur": 27,
    "args": {
      "External id": 138657,
      "cbid": 41, "correlation": 138657
    }
  },
  {
    "ph": "s", "id": 138657, "pid": 76337, "tid": -914061504, "ts": 1716454223596101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596129, "dur": 4,
    "args": {
      "External id": 138658,
      "cbid": 131, "correlation": 138658
    }
  },
  {
    "ph": "f", "id": 138658, "pid": 76337, "tid": -914061504, "ts": 1716454223596129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223596181, "dur": 3,
    "args": {
      "External id": 138666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138666,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138666, "pid": 5, "tid": 7, "ts": 1716454223596181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596168, "dur": 13,
    "args": {
      "External id": 138666,
      "cbid": 211, "correlation": 138666
    }
  },
  {
    "ph": "s", "id": 138666, "pid": 76337, "tid": -914061504, "ts": 1716454223596168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223596212, "dur": 3,
    "args": {
      "External id": 138676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138676,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138676, "pid": 5, "tid": 7, "ts": 1716454223596212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596203, "dur": 8,
    "args": {
      "External id": 138676,
      "cbid": 211, "correlation": 138676
    }
  },
  {
    "ph": "s", "id": 138676, "pid": 76337, "tid": -914061504, "ts": 1716454223596203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223596243, "dur": 3,
    "args": {
      "External id": 138685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138685,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138685, "pid": 5, "tid": 7, "ts": 1716454223596243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596229, "dur": 13,
    "args": {
      "External id": 138685,
      "cbid": 211, "correlation": 138685
    }
  },
  {
    "ph": "s", "id": 138685, "pid": 76337, "tid": -914061504, "ts": 1716454223596229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223596363, "dur": 12,
    "args": {
      "External id": 138695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138695,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138695, "pid": 5, "tid": 7, "ts": 1716454223596363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596349, "dur": 14,
    "args": {
      "External id": 138695,
      "cbid": 211, "correlation": 138695
    }
  },
  {
    "ph": "s", "id": 138695, "pid": 76337, "tid": -914061504, "ts": 1716454223596349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223596402, "dur": 3,
    "args": {
      "External id": 138703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138703,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138703, "pid": 5, "tid": 7, "ts": 1716454223596402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596392, "dur": 9,
    "args": {
      "External id": 138703,
      "cbid": 211, "correlation": 138703
    }
  },
  {
    "ph": "s", "id": 138703, "pid": 76337, "tid": -914061504, "ts": 1716454223596392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223596448, "dur": 11,
    "args": {
      "External id": 138713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138713,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138713, "pid": 5, "tid": 7, "ts": 1716454223596448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596437, "dur": 11,
    "args": {
      "External id": 138713,
      "cbid": 211, "correlation": 138713
    }
  },
  {
    "ph": "s", "id": 138713, "pid": 76337, "tid": -914061504, "ts": 1716454223596437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223596481, "dur": 10,
    "args": {
      "External id": 138721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138721,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138721, "pid": 5, "tid": 7, "ts": 1716454223596481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596472, "dur": 9,
    "args": {
      "External id": 138721,
      "cbid": 211, "correlation": 138721
    }
  },
  {
    "ph": "s", "id": 138721, "pid": 76337, "tid": -914061504, "ts": 1716454223596472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223596508, "dur": 3,
    "args": {
      "External id": 138730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138730, "pid": 5, "tid": 7, "ts": 1716454223596508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596500, "dur": 9,
    "args": {
      "External id": 138730,
      "cbid": 211, "correlation": 138730
    }
  },
  {
    "ph": "s", "id": 138730, "pid": 76337, "tid": -914061504, "ts": 1716454223596500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223596533, "dur": 5,
    "args": {
      "External id": 138739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138739,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138739, "pid": 5, "tid": 7, "ts": 1716454223596533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596524, "dur": 8,
    "args": {
      "External id": 138739,
      "cbid": 211, "correlation": 138739
    }
  },
  {
    "ph": "s", "id": 138739, "pid": 76337, "tid": -914061504, "ts": 1716454223596524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223596571, "dur": 8,
    "args": {
      "External id": 138749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138749,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138749, "pid": 5, "tid": 7, "ts": 1716454223596571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596561, "dur": 10,
    "args": {
      "External id": 138749,
      "cbid": 211, "correlation": 138749
    }
  },
  {
    "ph": "s", "id": 138749, "pid": 76337, "tid": -914061504, "ts": 1716454223596561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223596898, "dur": 3,
    "args": {
      "External id": 138758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138758,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138758, "pid": 5, "tid": 7, "ts": 1716454223596898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596883, "dur": 15,
    "args": {
      "External id": 138758,
      "cbid": 211, "correlation": 138758
    }
  },
  {
    "ph": "s", "id": 138758, "pid": 76337, "tid": -914061504, "ts": 1716454223596883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223596928, "dur": 3,
    "args": {
      "External id": 138766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138766,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138766, "pid": 5, "tid": 7, "ts": 1716454223596928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596918, "dur": 9,
    "args": {
      "External id": 138766,
      "cbid": 211, "correlation": 138766
    }
  },
  {
    "ph": "s", "id": 138766, "pid": 76337, "tid": -914061504, "ts": 1716454223596918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223596987, "dur": 1,
    "args": {
      "External id": 138776,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138776,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 138776, "pid": 5, "tid": 7, "ts": 1716454223596987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596964, "dur": 21,
    "args": {
      "External id": 138776,
      "cbid": 41, "correlation": 138776
    }
  },
  {
    "ph": "s", "id": 138776, "pid": 76337, "tid": -914061504, "ts": 1716454223596964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223596986, "dur": 7,
    "args": {
      "External id": 138777,
      "cbid": 131, "correlation": 138777
    }
  },
  {
    "ph": "f", "id": 138777, "pid": 76337, "tid": -914061504, "ts": 1716454223596986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223597075, "dur": 2,
    "args": {
      "External id": 138785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138785, "pid": 5, "tid": 7, "ts": 1716454223597075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597061, "dur": 14,
    "args": {
      "External id": 138785,
      "cbid": 211, "correlation": 138785
    }
  },
  {
    "ph": "s", "id": 138785, "pid": 76337, "tid": -914061504, "ts": 1716454223597061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223597146, "dur": 3,
    "args": {
      "External id": 138794,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138794,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002577319587628866
    }
  },
  {
    "ph": "f", "id": 138794, "pid": 5, "tid": 7, "ts": 1716454223597146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597129, "dur": 17,
    "args": {
      "External id": 138794,
      "cbid": 41, "correlation": 138794
    }
  },
  {
    "ph": "s", "id": 138794, "pid": 76337, "tid": -914061504, "ts": 1716454223597129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223597216, "dur": 3,
    "args": {
      "External id": 138804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138804,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 138804, "pid": 5, "tid": 7, "ts": 1716454223597216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597202, "dur": 14,
    "args": {
      "External id": 138804,
      "cbid": 211, "correlation": 138804
    }
  },
  {
    "ph": "s", "id": 138804, "pid": 76337, "tid": -914061504, "ts": 1716454223597202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223597267, "dur": 1,
    "args": {
      "External id": 138814,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138814,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 138814, "pid": 5, "tid": 7, "ts": 1716454223597267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597253, "dur": 12,
    "args": {
      "External id": 138814,
      "cbid": 41, "correlation": 138814
    }
  },
  {
    "ph": "s", "id": 138814, "pid": 76337, "tid": -914061504, "ts": 1716454223597253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597266, "dur": 8,
    "args": {
      "External id": 138815,
      "cbid": 131, "correlation": 138815
    }
  },
  {
    "ph": "f", "id": 138815, "pid": 76337, "tid": -914061504, "ts": 1716454223597266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223597367, "dur": 3,
    "args": {
      "External id": 138822,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138822,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 138822, "pid": 5, "tid": 7, "ts": 1716454223597367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597307, "dur": 60,
    "args": {
      "External id": 138822,
      "cbid": 41, "correlation": 138822
    }
  },
  {
    "ph": "s", "id": 138822, "pid": 76337, "tid": -914061504, "ts": 1716454223597307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223597419, "dur": 1,
    "args": {
      "External id": 138833,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138833,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 138833, "pid": 5, "tid": 7, "ts": 1716454223597419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597406, "dur": 12,
    "args": {
      "External id": 138833,
      "cbid": 41, "correlation": 138833
    }
  },
  {
    "ph": "s", "id": 138833, "pid": 76337, "tid": -914061504, "ts": 1716454223597406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597418, "dur": 8,
    "args": {
      "External id": 138834,
      "cbid": 131, "correlation": 138834
    }
  },
  {
    "ph": "f", "id": 138834, "pid": 76337, "tid": -914061504, "ts": 1716454223597418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223597469, "dur": 3,
    "args": {
      "External id": 138842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138842,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138842, "pid": 5, "tid": 7, "ts": 1716454223597469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597455, "dur": 13,
    "args": {
      "External id": 138842,
      "cbid": 211, "correlation": 138842
    }
  },
  {
    "ph": "s", "id": 138842, "pid": 76337, "tid": -914061504, "ts": 1716454223597455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223597497, "dur": 3,
    "args": {
      "External id": 138852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138852,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138852, "pid": 5, "tid": 7, "ts": 1716454223597497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597488, "dur": 8,
    "args": {
      "External id": 138852,
      "cbid": 211, "correlation": 138852
    }
  },
  {
    "ph": "s", "id": 138852, "pid": 76337, "tid": -914061504, "ts": 1716454223597488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223597521, "dur": 3,
    "args": {
      "External id": 138861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138861,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138861, "pid": 5, "tid": 7, "ts": 1716454223597521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597512, "dur": 8,
    "args": {
      "External id": 138861,
      "cbid": 211, "correlation": 138861
    }
  },
  {
    "ph": "s", "id": 138861, "pid": 76337, "tid": -914061504, "ts": 1716454223597512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223597590, "dur": 6,
    "args": {
      "External id": 138869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138869,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138869, "pid": 5, "tid": 7, "ts": 1716454223597590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597577, "dur": 13,
    "args": {
      "External id": 138869,
      "cbid": 211, "correlation": 138869
    }
  },
  {
    "ph": "s", "id": 138869, "pid": 76337, "tid": -914061504, "ts": 1716454223597577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223597630, "dur": 3,
    "args": {
      "External id": 138878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138878,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138878, "pid": 5, "tid": 7, "ts": 1716454223597630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597620, "dur": 9,
    "args": {
      "External id": 138878,
      "cbid": 211, "correlation": 138878
    }
  },
  {
    "ph": "s", "id": 138878, "pid": 76337, "tid": -914061504, "ts": 1716454223597620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223597652, "dur": 3,
    "args": {
      "External id": 138887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138887,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138887, "pid": 5, "tid": 7, "ts": 1716454223597652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597644, "dur": 7,
    "args": {
      "External id": 138887,
      "cbid": 211, "correlation": 138887
    }
  },
  {
    "ph": "s", "id": 138887, "pid": 76337, "tid": -914061504, "ts": 1716454223597644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223597715, "dur": 3,
    "args": {
      "External id": 138895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138895,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 138895, "pid": 5, "tid": 7, "ts": 1716454223597715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597704, "dur": 10,
    "args": {
      "External id": 138895,
      "cbid": 211, "correlation": 138895
    }
  },
  {
    "ph": "s", "id": 138895, "pid": 76337, "tid": -914061504, "ts": 1716454223597704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223597775, "dur": 1,
    "args": {
      "External id": 138903,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138903,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0043859649122807015
    }
  },
  {
    "ph": "f", "id": 138903, "pid": 5, "tid": 7, "ts": 1716454223597775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597759, "dur": 26,
    "args": {
      "External id": 138903,
      "cbid": 41, "correlation": 138903
    }
  },
  {
    "ph": "s", "id": 138903, "pid": 76337, "tid": -914061504, "ts": 1716454223597759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597785, "dur": 3,
    "args": {
      "External id": 138904,
      "cbid": 131, "correlation": 138904
    }
  },
  {
    "ph": "f", "id": 138904, "pid": 76337, "tid": -914061504, "ts": 1716454223597785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223597846, "dur": 1,
    "args": {
      "External id": 138914,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138914,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 138914, "pid": 5, "tid": 7, "ts": 1716454223597846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597834, "dur": 10,
    "args": {
      "External id": 138914,
      "cbid": 41, "correlation": 138914
    }
  },
  {
    "ph": "s", "id": 138914, "pid": 76337, "tid": -914061504, "ts": 1716454223597834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597845, "dur": 8,
    "args": {
      "External id": 138915,
      "cbid": 131, "correlation": 138915
    }
  },
  {
    "ph": "f", "id": 138915, "pid": 76337, "tid": -914061504, "ts": 1716454223597845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223597904, "dur": 1,
    "args": {
      "External id": 138924,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138924,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 138924, "pid": 5, "tid": 7, "ts": 1716454223597904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597893, "dur": 8,
    "args": {
      "External id": 138924,
      "cbid": 41, "correlation": 138924
    }
  },
  {
    "ph": "s", "id": 138924, "pid": 76337, "tid": -914061504, "ts": 1716454223597893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597903, "dur": 7,
    "args": {
      "External id": 138925,
      "cbid": 131, "correlation": 138925
    }
  },
  {
    "ph": "f", "id": 138925, "pid": 76337, "tid": -914061504, "ts": 1716454223597903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223597984, "dur": 4,
    "args": {
      "External id": 138932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138932, "pid": 5, "tid": 7, "ts": 1716454223597984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223597961, "dur": 25,
    "args": {
      "External id": 138932,
      "cbid": 211, "correlation": 138932
    }
  },
  {
    "ph": "s", "id": 138932, "pid": 76337, "tid": -914061504, "ts": 1716454223597961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454223598023, "dur": 4,
    "args": {
      "External id": 138952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138952,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138952, "pid": 5, "tid": 7, "ts": 1716454223598023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598011, "dur": 13,
    "args": {
      "External id": 138952,
      "cbid": 211, "correlation": 138952
    }
  },
  {
    "ph": "s", "id": 138952, "pid": 76337, "tid": -914061504, "ts": 1716454223598011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598025, "dur": 0,
    "args": {
      "External id": 138953,
      "cbid": 11, "correlation": 138953
    }
  },
  {
    "ph": "f", "id": 138953, "pid": 76337, "tid": -914061504, "ts": 1716454223598025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598026, "dur": 0,
    "args": {
      "External id": 138954,
      "cbid": 11, "correlation": 138954
    }
  },
  {
    "ph": "f", "id": 138954, "pid": 76337, "tid": -914061504, "ts": 1716454223598026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223598040, "dur": 1,
    "args": {
      "External id": 138957,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138957,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 138957, "pid": 5, "tid": 7, "ts": 1716454223598040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598027, "dur": 22,
    "args": {
      "External id": 138957,
      "cbid": 41, "correlation": 138957
    }
  },
  {
    "ph": "s", "id": 138957, "pid": 76337, "tid": -914061504, "ts": 1716454223598027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598050, "dur": 3,
    "args": {
      "External id": 138958,
      "cbid": 131, "correlation": 138958
    }
  },
  {
    "ph": "f", "id": 138958, "pid": 76337, "tid": -914061504, "ts": 1716454223598050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223598077, "dur": 3,
    "args": {
      "External id": 138982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138982,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138982, "pid": 5, "tid": 7, "ts": 1716454223598077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598068, "dur": 9,
    "args": {
      "External id": 138982,
      "cbid": 211, "correlation": 138982
    }
  },
  {
    "ph": "s", "id": 138982, "pid": 76337, "tid": -914061504, "ts": 1716454223598068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598078, "dur": 0,
    "args": {
      "External id": 138983,
      "cbid": 11, "correlation": 138983
    }
  },
  {
    "ph": "f", "id": 138983, "pid": 76337, "tid": -914061504, "ts": 1716454223598078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598078, "dur": 0,
    "args": {
      "External id": 138984,
      "cbid": 11, "correlation": 138984
    }
  },
  {
    "ph": "f", "id": 138984, "pid": 76337, "tid": -914061504, "ts": 1716454223598078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598080, "dur": 1,
    "args": {
      "External id": 138986,
      "cbid": 200, "correlation": 138986
    }
  },
  {
    "ph": "f", "id": 138986, "pid": 76337, "tid": -914061504, "ts": 1716454223598080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223598090, "dur": 4,
    "args": {
      "External id": 138988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 138988,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 138988, "pid": 5, "tid": 7, "ts": 1716454223598090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598083, "dur": 8,
    "args": {
      "External id": 138988,
      "cbid": 211, "correlation": 138988
    }
  },
  {
    "ph": "s", "id": 138988, "pid": 76337, "tid": -914061504, "ts": 1716454223598083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598091, "dur": 0,
    "args": {
      "External id": 138989,
      "cbid": 11, "correlation": 138989
    }
  },
  {
    "ph": "f", "id": 138989, "pid": 76337, "tid": -914061504, "ts": 1716454223598091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598092, "dur": 0,
    "args": {
      "External id": 138990,
      "cbid": 11, "correlation": 138990
    }
  },
  {
    "ph": "f", "id": 138990, "pid": 76337, "tid": -914061504, "ts": 1716454223598092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223598128, "dur": 1,
    "args": {
      "External id": 138997,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 138997,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 138997, "pid": 5, "tid": 7, "ts": 1716454223598128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598117, "dur": 20,
    "args": {
      "External id": 138997,
      "cbid": 41, "correlation": 138997
    }
  },
  {
    "ph": "s", "id": 138997, "pid": 76337, "tid": -914061504, "ts": 1716454223598117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598137, "dur": 3,
    "args": {
      "External id": 138998,
      "cbid": 131, "correlation": 138998
    }
  },
  {
    "ph": "f", "id": 138998, "pid": 76337, "tid": -914061504, "ts": 1716454223598137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223598188, "dur": 1,
    "args": {
      "External id": 139008,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 139008,
      "bytes": 8, "memory bandwidth (GB/s)": 0.00510204081632653
    }
  },
  {
    "ph": "f", "id": 139008, "pid": 5, "tid": 7, "ts": 1716454223598188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598176, "dur": 9,
    "args": {
      "External id": 139008,
      "cbid": 41, "correlation": 139008
    }
  },
  {
    "ph": "s", "id": 139008, "pid": 76337, "tid": -914061504, "ts": 1716454223598176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598186, "dur": 8,
    "args": {
      "External id": 139009,
      "cbid": 131, "correlation": 139009
    }
  },
  {
    "ph": "f", "id": 139009, "pid": 76337, "tid": -914061504, "ts": 1716454223598186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223598259, "dur": 5,
    "args": {
      "External id": 139016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139016,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139016, "pid": 5, "tid": 7, "ts": 1716454223598259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598243, "dur": 16,
    "args": {
      "External id": 139016,
      "cbid": 211, "correlation": 139016
    }
  },
  {
    "ph": "s", "id": 139016, "pid": 76337, "tid": -914061504, "ts": 1716454223598243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223598329, "dur": 3,
    "args": {
      "External id": 139025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139025,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139025, "pid": 5, "tid": 7, "ts": 1716454223598329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598317, "dur": 12,
    "args": {
      "External id": 139025,
      "cbid": 211, "correlation": 139025
    }
  },
  {
    "ph": "s", "id": 139025, "pid": 76337, "tid": -914061504, "ts": 1716454223598317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223598366, "dur": 3,
    "args": {
      "External id": 139033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139033,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139033, "pid": 5, "tid": 7, "ts": 1716454223598366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598356, "dur": 10,
    "args": {
      "External id": 139033,
      "cbid": 211, "correlation": 139033
    }
  },
  {
    "ph": "s", "id": 139033, "pid": 76337, "tid": -914061504, "ts": 1716454223598356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223598399, "dur": 4,
    "args": {
      "External id": 139041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139041,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139041, "pid": 5, "tid": 7, "ts": 1716454223598399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598387, "dur": 12,
    "args": {
      "External id": 139041,
      "cbid": 211, "correlation": 139041
    }
  },
  {
    "ph": "s", "id": 139041, "pid": 76337, "tid": -914061504, "ts": 1716454223598387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223598428, "dur": 4,
    "args": {
      "External id": 139049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139049,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139049, "pid": 5, "tid": 7, "ts": 1716454223598428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598418, "dur": 9,
    "args": {
      "External id": 139049,
      "cbid": 211, "correlation": 139049
    }
  },
  {
    "ph": "s", "id": 139049, "pid": 76337, "tid": -914061504, "ts": 1716454223598418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223598455, "dur": 3,
    "args": {
      "External id": 139057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139057,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139057, "pid": 5, "tid": 7, "ts": 1716454223598455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598446, "dur": 8,
    "args": {
      "External id": 139057,
      "cbid": 211, "correlation": 139057
    }
  },
  {
    "ph": "s", "id": 139057, "pid": 76337, "tid": -914061504, "ts": 1716454223598446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223598480, "dur": 3,
    "args": {
      "External id": 139065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139065,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139065, "pid": 5, "tid": 7, "ts": 1716454223598480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598471, "dur": 9,
    "args": {
      "External id": 139065,
      "cbid": 211, "correlation": 139065
    }
  },
  {
    "ph": "s", "id": 139065, "pid": 76337, "tid": -914061504, "ts": 1716454223598471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223598504, "dur": 4,
    "args": {
      "External id": 139073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139073, "pid": 5, "tid": 7, "ts": 1716454223598504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598495, "dur": 7,
    "args": {
      "External id": 139073,
      "cbid": 211, "correlation": 139073
    }
  },
  {
    "ph": "s", "id": 139073, "pid": 76337, "tid": -914061504, "ts": 1716454223598495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223598522, "dur": 5,
    "args": {
      "External id": 139081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139081,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139081, "pid": 5, "tid": 7, "ts": 1716454223598522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598514, "dur": 7,
    "args": {
      "External id": 139081,
      "cbid": 211, "correlation": 139081
    }
  },
  {
    "ph": "s", "id": 139081, "pid": 76337, "tid": -914061504, "ts": 1716454223598514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223598542, "dur": 3,
    "args": {
      "External id": 139089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139089,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139089, "pid": 5, "tid": 7, "ts": 1716454223598542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598533, "dur": 7,
    "args": {
      "External id": 139089,
      "cbid": 211, "correlation": 139089
    }
  },
  {
    "ph": "s", "id": 139089, "pid": 76337, "tid": -914061504, "ts": 1716454223598533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223598599, "dur": 3,
    "args": {
      "External id": 139097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139097,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139097, "pid": 5, "tid": 7, "ts": 1716454223598599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598587, "dur": 11,
    "args": {
      "External id": 139097,
      "cbid": 211, "correlation": 139097
    }
  },
  {
    "ph": "s", "id": 139097, "pid": 76337, "tid": -914061504, "ts": 1716454223598587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223598624, "dur": 4,
    "args": {
      "External id": 139105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139105,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139105, "pid": 5, "tid": 7, "ts": 1716454223598624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598615, "dur": 8,
    "args": {
      "External id": 139105,
      "cbid": 211, "correlation": 139105
    }
  },
  {
    "ph": "s", "id": 139105, "pid": 76337, "tid": -914061504, "ts": 1716454223598615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223598647, "dur": 4,
    "args": {
      "External id": 139113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139113,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139113, "pid": 5, "tid": 7, "ts": 1716454223598647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598638, "dur": 8,
    "args": {
      "External id": 139113,
      "cbid": 211, "correlation": 139113
    }
  },
  {
    "ph": "s", "id": 139113, "pid": 76337, "tid": -914061504, "ts": 1716454223598638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223598666, "dur": 3,
    "args": {
      "External id": 139121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139121,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139121, "pid": 5, "tid": 7, "ts": 1716454223598666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223598658, "dur": 6,
    "args": {
      "External id": 139121,
      "cbid": 211, "correlation": 139121
    }
  },
  {
    "ph": "s", "id": 139121, "pid": 76337, "tid": -914061504, "ts": 1716454223598658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223599069, "dur": 5,
    "args": {
      "External id": 139130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139130,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139130, "pid": 5, "tid": 7, "ts": 1716454223599069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599053, "dur": 16,
    "args": {
      "External id": 139130,
      "cbid": 211, "correlation": 139130
    }
  },
  {
    "ph": "s", "id": 139130, "pid": 76337, "tid": -914061504, "ts": 1716454223599053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223599107, "dur": 5,
    "args": {
      "External id": 139139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139139,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139139, "pid": 5, "tid": 7, "ts": 1716454223599107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599096, "dur": 10,
    "args": {
      "External id": 139139,
      "cbid": 211, "correlation": 139139
    }
  },
  {
    "ph": "s", "id": 139139, "pid": 76337, "tid": -914061504, "ts": 1716454223599096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223599237, "dur": 3,
    "args": {
      "External id": 139155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139155, "pid": 5, "tid": 7, "ts": 1716454223599237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599223, "dur": 14,
    "args": {
      "External id": 139155,
      "cbid": 211, "correlation": 139155
    }
  },
  {
    "ph": "s", "id": 139155, "pid": 76337, "tid": -914061504, "ts": 1716454223599223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599270, "dur": 3,
    "args": {
      "External id": 139163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139163,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139163, "pid": 5, "tid": 7, "ts": 1716454223599270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599260, "dur": 9,
    "args": {
      "External id": 139163,
      "cbid": 211, "correlation": 139163
    }
  },
  {
    "ph": "s", "id": 139163, "pid": 76337, "tid": -914061504, "ts": 1716454223599260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599301, "dur": 3,
    "args": {
      "External id": 139171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139171,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139171, "pid": 5, "tid": 7, "ts": 1716454223599301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599291, "dur": 8,
    "args": {
      "External id": 139171,
      "cbid": 211, "correlation": 139171
    }
  },
  {
    "ph": "s", "id": 139171, "pid": 76337, "tid": -914061504, "ts": 1716454223599291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599331, "dur": 4,
    "args": {
      "External id": 139179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139179,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139179, "pid": 5, "tid": 7, "ts": 1716454223599331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599321, "dur": 8,
    "args": {
      "External id": 139179,
      "cbid": 211, "correlation": 139179
    }
  },
  {
    "ph": "s", "id": 139179, "pid": 76337, "tid": -914061504, "ts": 1716454223599321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223599386, "dur": 4,
    "args": {
      "External id": 139191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139191,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139191, "pid": 5, "tid": 7, "ts": 1716454223599386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599373, "dur": 13,
    "args": {
      "External id": 139191,
      "cbid": 211, "correlation": 139191
    }
  },
  {
    "ph": "s", "id": 139191, "pid": 76337, "tid": -914061504, "ts": 1716454223599373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223599432, "dur": 4,
    "args": {
      "External id": 139202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139202, "pid": 5, "tid": 7, "ts": 1716454223599432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599420, "dur": 11,
    "args": {
      "External id": 139202,
      "cbid": 211, "correlation": 139202
    }
  },
  {
    "ph": "s", "id": 139202, "pid": 76337, "tid": -914061504, "ts": 1716454223599420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599464, "dur": 3,
    "args": {
      "External id": 139210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139210, "pid": 5, "tid": 7, "ts": 1716454223599464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599454, "dur": 9,
    "args": {
      "External id": 139210,
      "cbid": 211, "correlation": 139210
    }
  },
  {
    "ph": "s", "id": 139210, "pid": 76337, "tid": -914061504, "ts": 1716454223599454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599498, "dur": 5,
    "args": {
      "External id": 139218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139218,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139218, "pid": 5, "tid": 7, "ts": 1716454223599498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599488, "dur": 10,
    "args": {
      "External id": 139218,
      "cbid": 211, "correlation": 139218
    }
  },
  {
    "ph": "s", "id": 139218, "pid": 76337, "tid": -914061504, "ts": 1716454223599488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599527, "dur": 5,
    "args": {
      "External id": 139226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139226,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139226, "pid": 5, "tid": 7, "ts": 1716454223599527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599517, "dur": 9,
    "args": {
      "External id": 139226,
      "cbid": 211, "correlation": 139226
    }
  },
  {
    "ph": "s", "id": 139226, "pid": 76337, "tid": -914061504, "ts": 1716454223599517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223599558, "dur": 4,
    "args": {
      "External id": 139235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139235,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139235, "pid": 5, "tid": 7, "ts": 1716454223599558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599548, "dur": 10,
    "args": {
      "External id": 139235,
      "cbid": 211, "correlation": 139235
    }
  },
  {
    "ph": "s", "id": 139235, "pid": 76337, "tid": -914061504, "ts": 1716454223599548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223599619, "dur": 5,
    "args": {
      "External id": 139248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139248,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139248, "pid": 5, "tid": 7, "ts": 1716454223599619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599606, "dur": 13,
    "args": {
      "External id": 139248,
      "cbid": 211, "correlation": 139248
    }
  },
  {
    "ph": "s", "id": 139248, "pid": 76337, "tid": -914061504, "ts": 1716454223599606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223599659, "dur": 5,
    "args": {
      "External id": 139258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139258,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139258, "pid": 5, "tid": 7, "ts": 1716454223599659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599648, "dur": 11,
    "args": {
      "External id": 139258,
      "cbid": 211, "correlation": 139258
    }
  },
  {
    "ph": "s", "id": 139258, "pid": 76337, "tid": -914061504, "ts": 1716454223599648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599788, "dur": 5,
    "args": {
      "External id": 139275,
      "cbid": 251, "correlation": 139275
    }
  },
  {
    "ph": "f", "id": 139275, "pid": 76337, "tid": -914061504, "ts": 1716454223599788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454223599817, "dur": 11,
    "args": {
      "External id": 139277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139277,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139277, "pid": 5, "tid": 7, "ts": 1716454223599817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599802, "dur": 17,
    "args": {
      "External id": 139277,
      "cbid": 211, "correlation": 139277
    }
  },
  {
    "ph": "s", "id": 139277, "pid": 76337, "tid": -914061504, "ts": 1716454223599802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223599880, "dur": 3,
    "args": {
      "External id": 139285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139285,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139285, "pid": 5, "tid": 7, "ts": 1716454223599880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599867, "dur": 12,
    "args": {
      "External id": 139285,
      "cbid": 211, "correlation": 139285
    }
  },
  {
    "ph": "s", "id": 139285, "pid": 76337, "tid": -914061504, "ts": 1716454223599867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599938, "dur": 2,
    "args": {
      "External id": 139301,
      "cbid": 251, "correlation": 139301
    }
  },
  {
    "ph": "f", "id": 139301, "pid": 76337, "tid": -914061504, "ts": 1716454223599938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599944, "dur": 0,
    "args": {
      "External id": 139303,
      "cbid": 251, "correlation": 139303
    }
  },
  {
    "ph": "f", "id": 139303, "pid": 76337, "tid": -914061504, "ts": 1716454223599944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223599960, "dur": 14,
    "args": {
      "External id": 139304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139304,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139304, "pid": 5, "tid": 7, "ts": 1716454223599960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599946, "dur": 14,
    "args": {
      "External id": 139304,
      "cbid": 211, "correlation": 139304
    }
  },
  {
    "ph": "s", "id": 139304, "pid": 76337, "tid": -914061504, "ts": 1716454223599946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223599975, "dur": 5,
    "args": {
      "External id": 139306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139306,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139306, "pid": 5, "tid": 7, "ts": 1716454223599975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223599965, "dur": 17,
    "args": {
      "External id": 139306,
      "cbid": 211, "correlation": 139306
    }
  },
  {
    "ph": "s", "id": 139306, "pid": 76337, "tid": -914061504, "ts": 1716454223599965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600087, "dur": 1,
    "args": {
      "External id": 139316,
      "cbid": 317, "correlation": 139316
    }
  },
  {
    "ph": "f", "id": 139316, "pid": 76337, "tid": -914061504, "ts": 1716454223600087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600089, "dur": 1,
    "args": {
      "External id": 139317,
      "cbid": 203, "correlation": 139317
    }
  },
  {
    "ph": "f", "id": 139317, "pid": 76337, "tid": -914061504, "ts": 1716454223600089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600091, "dur": 1,
    "args": {
      "External id": 139318,
      "cbid": 205, "correlation": 139318
    }
  },
  {
    "ph": "f", "id": 139318, "pid": 76337, "tid": -914061504, "ts": 1716454223600091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223600149, "dur": 7,
    "args": {
      "External id": 139322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139322,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139322, "pid": 5, "tid": 7, "ts": 1716454223600149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600134, "dur": 15,
    "args": {
      "External id": 139322,
      "cbid": 211, "correlation": 139322
    }
  },
  {
    "ph": "s", "id": 139322, "pid": 76337, "tid": -914061504, "ts": 1716454223600134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223600160, "dur": 4,
    "args": {
      "External id": 139324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139324,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 139324, "pid": 5, "tid": 7, "ts": 1716454223600160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600153, "dur": 6,
    "args": {
      "External id": 139324,
      "cbid": 211, "correlation": 139324
    }
  },
  {
    "ph": "s", "id": 139324, "pid": 76337, "tid": -914061504, "ts": 1716454223600153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223600180, "dur": 4,
    "args": {
      "External id": 139326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139326, "pid": 5, "tid": 7, "ts": 1716454223600180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600170, "dur": 9,
    "args": {
      "External id": 139326,
      "cbid": 211, "correlation": 139326
    }
  },
  {
    "ph": "s", "id": 139326, "pid": 76337, "tid": -914061504, "ts": 1716454223600170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600186, "dur": 0,
    "args": {
      "External id": 139327,
      "cbid": 51, "correlation": 139327
    }
  },
  {
    "ph": "s", "id": 139327, "pid": 76337, "tid": -914061504, "ts": 1716454223600186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223600198, "dur": 87,
    "args": {
      "External id": 139328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139328,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139328, "pid": 5, "tid": 7, "ts": 1716454223600198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600188, "dur": 9,
    "args": {
      "External id": 139328,
      "cbid": 211, "correlation": 139328
    }
  },
  {
    "ph": "s", "id": 139328, "pid": 76337, "tid": -914061504, "ts": 1716454223600188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223600286, "dur": 61,
    "args": {
      "External id": 139333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139333, "pid": 5, "tid": 7, "ts": 1716454223600286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223600228, "dur": 10,
    "args": {
      "External id": 139333,
      "cbid": 211, "correlation": 139333
    }
  },
  {
    "ph": "s", "id": 139333, "pid": 76337, "tid": -914061504, "ts": 1716454223600228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223602055, "dur": 51,
    "args": {
      "External id": 139353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139353,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 139353, "pid": 5, "tid": 7, "ts": 1716454223602055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602038, "dur": 17,
    "args": {
      "External id": 139353,
      "cbid": 211, "correlation": 139353
    }
  },
  {
    "ph": "s", "id": 139353, "pid": 76337, "tid": -914061504, "ts": 1716454223602038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223602108, "dur": 4,
    "args": {
      "External id": 139365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139365,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139365, "pid": 5, "tid": 7, "ts": 1716454223602108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602066, "dur": 8,
    "args": {
      "External id": 139365,
      "cbid": 211, "correlation": 139365
    }
  },
  {
    "ph": "s", "id": 139365, "pid": 76337, "tid": -914061504, "ts": 1716454223602066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223602114, "dur": 58,
    "args": {
      "External id": 139368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139368,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139368, "pid": 5, "tid": 7, "ts": 1716454223602114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602090, "dur": 8,
    "args": {
      "External id": 139368,
      "cbid": 211, "correlation": 139368
    }
  },
  {
    "ph": "s", "id": 139368, "pid": 76337, "tid": -914061504, "ts": 1716454223602090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223602173, "dur": 37,
    "args": {
      "External id": 139377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139377,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139377, "pid": 5, "tid": 7, "ts": 1716454223602173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602135, "dur": 11,
    "args": {
      "External id": 139377,
      "cbid": 211, "correlation": 139377
    }
  },
  {
    "ph": "s", "id": 139377, "pid": 76337, "tid": -914061504, "ts": 1716454223602135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602194, "dur": 0,
    "args": {
      "External id": 139387,
      "cbid": 317, "correlation": 139387
    }
  },
  {
    "ph": "f", "id": 139387, "pid": 76337, "tid": -914061504, "ts": 1716454223602194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602194, "dur": 0,
    "args": {
      "External id": 139388,
      "cbid": 203, "correlation": 139388
    }
  },
  {
    "ph": "f", "id": 139388, "pid": 76337, "tid": -914061504, "ts": 1716454223602194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602195, "dur": 0,
    "args": {
      "External id": 139389,
      "cbid": 205, "correlation": 139389
    }
  },
  {
    "ph": "f", "id": 139389, "pid": 76337, "tid": -914061504, "ts": 1716454223602195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223602227, "dur": 40,
    "args": {
      "External id": 139393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139393,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139393, "pid": 5, "tid": 7, "ts": 1716454223602227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602214, "dur": 12,
    "args": {
      "External id": 139393,
      "cbid": 211, "correlation": 139393
    }
  },
  {
    "ph": "s", "id": 139393, "pid": 76337, "tid": -914061504, "ts": 1716454223602214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223602269, "dur": 14,
    "args": {
      "External id": 139395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139395,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139395, "pid": 5, "tid": 7, "ts": 1716454223602269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602229, "dur": 6,
    "args": {
      "External id": 139395,
      "cbid": 211, "correlation": 139395
    }
  },
  {
    "ph": "s", "id": 139395, "pid": 76337, "tid": -914061504, "ts": 1716454223602229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223602285, "dur": 3,
    "args": {
      "External id": 139397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139397, "pid": 5, "tid": 7, "ts": 1716454223602285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602240, "dur": 6,
    "args": {
      "External id": 139397,
      "cbid": 211, "correlation": 139397
    }
  },
  {
    "ph": "s", "id": 139397, "pid": 76337, "tid": -914061504, "ts": 1716454223602240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602250, "dur": 0,
    "args": {
      "External id": 139398,
      "cbid": 51, "correlation": 139398
    }
  },
  {
    "ph": "s", "id": 139398, "pid": 76337, "tid": -914061504, "ts": 1716454223602250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223602289, "dur": 714,
    "args": {
      "External id": 139399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139399,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139399, "pid": 5, "tid": 7, "ts": 1716454223602289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602251, "dur": 7,
    "args": {
      "External id": 139399,
      "cbid": 211, "correlation": 139399
    }
  },
  {
    "ph": "s", "id": 139399, "pid": 76337, "tid": -914061504, "ts": 1716454223602251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223603004, "dur": 60,
    "args": {
      "External id": 139404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139404,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139404, "pid": 5, "tid": 7, "ts": 1716454223603004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602281, "dur": 10,
    "args": {
      "External id": 139404,
      "cbid": 211, "correlation": 139404
    }
  },
  {
    "ph": "s", "id": 139404, "pid": 76337, "tid": -914061504, "ts": 1716454223602281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223603066, "dur": 5,
    "args": {
      "External id": 139412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139412,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139412, "pid": 5, "tid": 7, "ts": 1716454223603066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602325, "dur": 10,
    "args": {
      "External id": 139412,
      "cbid": 211, "correlation": 139412
    }
  },
  {
    "ph": "s", "id": 139412, "pid": 76337, "tid": -914061504, "ts": 1716454223602325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602393, "dur": 2,
    "args": {
      "External id": 139428,
      "cbid": 251, "correlation": 139428
    }
  },
  {
    "ph": "f", "id": 139428, "pid": 76337, "tid": -914061504, "ts": 1716454223602393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602399, "dur": 0,
    "args": {
      "External id": 139430,
      "cbid": 251, "correlation": 139430
    }
  },
  {
    "ph": "f", "id": 139430, "pid": 76337, "tid": -914061504, "ts": 1716454223602399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223603072, "dur": 9,
    "args": {
      "External id": 139431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139431,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 139431, "pid": 5, "tid": 7, "ts": 1716454223603072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602400, "dur": 12,
    "args": {
      "External id": 139431,
      "cbid": 211, "correlation": 139431
    }
  },
  {
    "ph": "s", "id": 139431, "pid": 76337, "tid": -914061504, "ts": 1716454223602400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223603082, "dur": 4,
    "args": {
      "External id": 139433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139433,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 139433, "pid": 5, "tid": 7, "ts": 1716454223603082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602414, "dur": 6,
    "args": {
      "External id": 139433,
      "cbid": 211, "correlation": 139433
    }
  },
  {
    "ph": "s", "id": 139433, "pid": 76337, "tid": -914061504, "ts": 1716454223602414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223603087, "dur": 55,
    "args": {
      "External id": 139443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139443, "pid": 5, "tid": 7, "ts": 1716454223603087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602473, "dur": 13,
    "args": {
      "External id": 139443,
      "cbid": 211, "correlation": 139443
    }
  },
  {
    "ph": "s", "id": 139443, "pid": 76337, "tid": -914061504, "ts": 1716454223602473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223603144, "dur": 52,
    "args": {
      "External id": 139463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139463,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 139463, "pid": 5, "tid": 7, "ts": 1716454223603144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602542, "dur": 10,
    "args": {
      "External id": 139463,
      "cbid": 211, "correlation": 139463
    }
  },
  {
    "ph": "s", "id": 139463, "pid": 76337, "tid": -914061504, "ts": 1716454223602542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223603197, "dur": 4,
    "args": {
      "External id": 139475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139475,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139475, "pid": 5, "tid": 7, "ts": 1716454223603197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602561, "dur": 7,
    "args": {
      "External id": 139475,
      "cbid": 211, "correlation": 139475
    }
  },
  {
    "ph": "s", "id": 139475, "pid": 76337, "tid": -914061504, "ts": 1716454223602561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223603202, "dur": 56,
    "args": {
      "External id": 139478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139478,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139478, "pid": 5, "tid": 7, "ts": 1716454223603202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602579, "dur": 7,
    "args": {
      "External id": 139478,
      "cbid": 211, "correlation": 139478
    }
  },
  {
    "ph": "s", "id": 139478, "pid": 76337, "tid": -914061504, "ts": 1716454223602579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223603259, "dur": 37,
    "args": {
      "External id": 139487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139487,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139487, "pid": 5, "tid": 7, "ts": 1716454223603259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602620, "dur": 10,
    "args": {
      "External id": 139487,
      "cbid": 211, "correlation": 139487
    }
  },
  {
    "ph": "s", "id": 139487, "pid": 76337, "tid": -914061504, "ts": 1716454223602620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602693, "dur": 0,
    "args": {
      "External id": 139497,
      "cbid": 317, "correlation": 139497
    }
  },
  {
    "ph": "f", "id": 139497, "pid": 76337, "tid": -914061504, "ts": 1716454223602693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602694, "dur": 0,
    "args": {
      "External id": 139498,
      "cbid": 203, "correlation": 139498
    }
  },
  {
    "ph": "f", "id": 139498, "pid": 76337, "tid": -914061504, "ts": 1716454223602694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602694, "dur": 0,
    "args": {
      "External id": 139499,
      "cbid": 205, "correlation": 139499
    }
  },
  {
    "ph": "f", "id": 139499, "pid": 76337, "tid": -914061504, "ts": 1716454223602694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223603297, "dur": 40,
    "args": {
      "External id": 139503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139503,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139503, "pid": 5, "tid": 7, "ts": 1716454223603297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602708, "dur": 12,
    "args": {
      "External id": 139503,
      "cbid": 211, "correlation": 139503
    }
  },
  {
    "ph": "s", "id": 139503, "pid": 76337, "tid": -914061504, "ts": 1716454223602708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223603339, "dur": 14,
    "args": {
      "External id": 139505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139505,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139505, "pid": 5, "tid": 7, "ts": 1716454223603339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602723, "dur": 5,
    "args": {
      "External id": 139505,
      "cbid": 211, "correlation": 139505
    }
  },
  {
    "ph": "s", "id": 139505, "pid": 76337, "tid": -914061504, "ts": 1716454223602723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223603355, "dur": 3,
    "args": {
      "External id": 139507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139507,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139507, "pid": 5, "tid": 7, "ts": 1716454223603355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602732, "dur": 6,
    "args": {
      "External id": 139507,
      "cbid": 211, "correlation": 139507
    }
  },
  {
    "ph": "s", "id": 139507, "pid": 76337, "tid": -914061504, "ts": 1716454223602732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602742, "dur": 0,
    "args": {
      "External id": 139508,
      "cbid": 51, "correlation": 139508
    }
  },
  {
    "ph": "s", "id": 139508, "pid": 76337, "tid": -914061504, "ts": 1716454223602742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223603360, "dur": 707,
    "args": {
      "External id": 139509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139509,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139509, "pid": 5, "tid": 7, "ts": 1716454223603360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602742, "dur": 5,
    "args": {
      "External id": 139509,
      "cbid": 211, "correlation": 139509
    }
  },
  {
    "ph": "s", "id": 139509, "pid": 76337, "tid": -914061504, "ts": 1716454223602742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223604068, "dur": 60,
    "args": {
      "External id": 139514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139514, "pid": 5, "tid": 7, "ts": 1716454223604068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602770, "dur": 9,
    "args": {
      "External id": 139514,
      "cbid": 211, "correlation": 139514
    }
  },
  {
    "ph": "s", "id": 139514, "pid": 76337, "tid": -914061504, "ts": 1716454223602770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223604130, "dur": 50,
    "args": {
      "External id": 139522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139522,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139522, "pid": 5, "tid": 7, "ts": 1716454223604130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602803, "dur": 9,
    "args": {
      "External id": 139522,
      "cbid": 211, "correlation": 139522
    }
  },
  {
    "ph": "s", "id": 139522, "pid": 76337, "tid": -914061504, "ts": 1716454223602803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223604181, "dur": 35,
    "args": {
      "External id": 139530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139530,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139530, "pid": 5, "tid": 7, "ts": 1716454223604181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602834, "dur": 10,
    "args": {
      "External id": 139530,
      "cbid": 211, "correlation": 139530
    }
  },
  {
    "ph": "s", "id": 139530, "pid": 76337, "tid": -914061504, "ts": 1716454223602834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223604218, "dur": 52,
    "args": {
      "External id": 139550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139550,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 139550, "pid": 5, "tid": 7, "ts": 1716454223604218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602917, "dur": 12,
    "args": {
      "External id": 139550,
      "cbid": 211, "correlation": 139550
    }
  },
  {
    "ph": "s", "id": 139550, "pid": 76337, "tid": -914061504, "ts": 1716454223602917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223604271, "dur": 4,
    "args": {
      "External id": 139562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139562,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 139562, "pid": 5, "tid": 7, "ts": 1716454223604271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602939, "dur": 7,
    "args": {
      "External id": 139562,
      "cbid": 211, "correlation": 139562
    }
  },
  {
    "ph": "s", "id": 139562, "pid": 76337, "tid": -914061504, "ts": 1716454223602939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223604276, "dur": 58,
    "args": {
      "External id": 139565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139565, "pid": 5, "tid": 7, "ts": 1716454223604276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223602957, "dur": 8,
    "args": {
      "External id": 139565,
      "cbid": 211, "correlation": 139565
    }
  },
  {
    "ph": "s", "id": 139565, "pid": 76337, "tid": -914061504, "ts": 1716454223602957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603024, "dur": 0,
    "args": {
      "External id": 139576,
      "cbid": 317, "correlation": 139576
    }
  },
  {
    "ph": "f", "id": 139576, "pid": 76337, "tid": -914061504, "ts": 1716454223603024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603025, "dur": 0,
    "args": {
      "External id": 139577,
      "cbid": 203, "correlation": 139577
    }
  },
  {
    "ph": "f", "id": 139577, "pid": 76337, "tid": -914061504, "ts": 1716454223603025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603026, "dur": 0,
    "args": {
      "External id": 139578,
      "cbid": 205, "correlation": 139578
    }
  },
  {
    "ph": "f", "id": 139578, "pid": 76337, "tid": -914061504, "ts": 1716454223603026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603060, "dur": 2,
    "args": {
      "External id": 139582,
      "cbid": 251, "correlation": 139582
    }
  },
  {
    "ph": "f", "id": 139582, "pid": 76337, "tid": -914061504, "ts": 1716454223603060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603063, "dur": 1,
    "args": {
      "External id": 139583,
      "cbid": 251, "correlation": 139583
    }
  },
  {
    "ph": "f", "id": 139583, "pid": 76337, "tid": -914061504, "ts": 1716454223603063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603065, "dur": 1,
    "args": {
      "External id": 139584,
      "cbid": 251, "correlation": 139584
    }
  },
  {
    "ph": "f", "id": 139584, "pid": 76337, "tid": -914061504, "ts": 1716454223603065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603068, "dur": 1,
    "args": {
      "External id": 139585,
      "cbid": 251, "correlation": 139585
    }
  },
  {
    "ph": "f", "id": 139585, "pid": 76337, "tid": -914061504, "ts": 1716454223603068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603069, "dur": 1,
    "args": {
      "External id": 139586,
      "cbid": 251, "correlation": 139586
    }
  },
  {
    "ph": "f", "id": 139586, "pid": 76337, "tid": -914061504, "ts": 1716454223603069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603071, "dur": 1,
    "args": {
      "External id": 139587,
      "cbid": 251, "correlation": 139587
    }
  },
  {
    "ph": "f", "id": 139587, "pid": 76337, "tid": -914061504, "ts": 1716454223603071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603074, "dur": 1,
    "args": {
      "External id": 139588,
      "cbid": 251, "correlation": 139588
    }
  },
  {
    "ph": "f", "id": 139588, "pid": 76337, "tid": -914061504, "ts": 1716454223603074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603076, "dur": 1,
    "args": {
      "External id": 139589,
      "cbid": 251, "correlation": 139589
    }
  },
  {
    "ph": "f", "id": 139589, "pid": 76337, "tid": -914061504, "ts": 1716454223603076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603079, "dur": 0,
    "args": {
      "External id": 139590,
      "cbid": 251, "correlation": 139590
    }
  },
  {
    "ph": "f", "id": 139590, "pid": 76337, "tid": -914061504, "ts": 1716454223603079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223604335, "dur": 119,
    "args": {
      "External id": 139591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139591,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 139591, "pid": 5, "tid": 7, "ts": 1716454223604335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603083, "dur": 15,
    "args": {
      "External id": 139591,
      "cbid": 211, "correlation": 139591
    }
  },
  {
    "ph": "s", "id": 139591, "pid": 76337, "tid": -914061504, "ts": 1716454223603083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223604455, "dur": 60,
    "args": {
      "External id": 139597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139597,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139597, "pid": 5, "tid": 7, "ts": 1716454223604455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603121, "dur": 9,
    "args": {
      "External id": 139597,
      "cbid": 211, "correlation": 139597
    }
  },
  {
    "ph": "s", "id": 139597, "pid": 76337, "tid": -914061504, "ts": 1716454223603121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223604517, "dur": 644,
    "args": {
      "External id": 139606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139606,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139606, "pid": 5, "tid": 7, "ts": 1716454223604517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603216, "dur": 16,
    "args": {
      "External id": 139606,
      "cbid": 211, "correlation": 139606
    }
  },
  {
    "ph": "s", "id": 139606, "pid": 76337, "tid": -914061504, "ts": 1716454223603216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223605162, "dur": 184,
    "args": {
      "External id": 139628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139628,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139628, "pid": 5, "tid": 7, "ts": 1716454223605162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603289, "dur": 12,
    "args": {
      "External id": 139628,
      "cbid": 211, "correlation": 139628
    }
  },
  {
    "ph": "s", "id": 139628, "pid": 76337, "tid": -914061504, "ts": 1716454223603289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603404, "dur": 2,
    "args": {
      "External id": 139639,
      "cbid": 251, "correlation": 139639
    }
  },
  {
    "ph": "f", "id": 139639, "pid": 76337, "tid": -914061504, "ts": 1716454223603404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223605348, "dur": 197,
    "args": {
      "External id": 139640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139640,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139640, "pid": 5, "tid": 7, "ts": 1716454223605348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603412, "dur": 14,
    "args": {
      "External id": 139640,
      "cbid": 211, "correlation": 139640
    }
  },
  {
    "ph": "s", "id": 139640, "pid": 76337, "tid": -914061504, "ts": 1716454223603412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603485, "dur": 1,
    "args": {
      "External id": 139651,
      "cbid": 251, "correlation": 139651
    }
  },
  {
    "ph": "f", "id": 139651, "pid": 76337, "tid": -914061504, "ts": 1716454223603485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223605546, "dur": 194,
    "args": {
      "External id": 139652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139652,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139652, "pid": 5, "tid": 7, "ts": 1716454223605546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603489, "dur": 11,
    "args": {
      "External id": 139652,
      "cbid": 211, "correlation": 139652
    }
  },
  {
    "ph": "s", "id": 139652, "pid": 76337, "tid": -914061504, "ts": 1716454223603489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603555, "dur": 1,
    "args": {
      "External id": 139663,
      "cbid": 251, "correlation": 139663
    }
  },
  {
    "ph": "f", "id": 139663, "pid": 76337, "tid": -914061504, "ts": 1716454223603555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223605742, "dur": 192,
    "args": {
      "External id": 139664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139664,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139664, "pid": 5, "tid": 7, "ts": 1716454223605742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603559, "dur": 11,
    "args": {
      "External id": 139664,
      "cbid": 211, "correlation": 139664
    }
  },
  {
    "ph": "s", "id": 139664, "pid": 76337, "tid": -914061504, "ts": 1716454223603559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223605935, "dur": 19011,
    "args": {
      "External id": 139685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139685,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 139685, "pid": 5, "tid": 7, "ts": 1716454223605935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603666, "dur": 16,
    "args": {
      "External id": 139685,
      "cbid": 211, "correlation": 139685
    }
  },
  {
    "ph": "s", "id": 139685, "pid": 76337, "tid": -914061504, "ts": 1716454223603666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603780, "dur": 2,
    "args": {
      "External id": 139703,
      "cbid": 251, "correlation": 139703
    }
  },
  {
    "ph": "f", "id": 139703, "pid": 76337, "tid": -914061504, "ts": 1716454223603780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223624947, "dur": 206,
    "args": {
      "External id": 139705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139705,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139705, "pid": 5, "tid": 7, "ts": 1716454223624947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603787, "dur": 14,
    "args": {
      "External id": 139705,
      "cbid": 211, "correlation": 139705
    }
  },
  {
    "ph": "s", "id": 139705, "pid": 76337, "tid": -914061504, "ts": 1716454223603787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223625155, "dur": 66,
    "args": {
      "External id": 139713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139713,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139713, "pid": 5, "tid": 7, "ts": 1716454223625155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603860, "dur": 13,
    "args": {
      "External id": 139713,
      "cbid": 211, "correlation": 139713
    }
  },
  {
    "ph": "s", "id": 139713, "pid": 76337, "tid": -914061504, "ts": 1716454223603860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223625222, "dur": 98,
    "args": {
      "External id": 139721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139721,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139721, "pid": 5, "tid": 7, "ts": 1716454223625222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603901, "dur": 9,
    "args": {
      "External id": 139721,
      "cbid": 211, "correlation": 139721
    }
  },
  {
    "ph": "s", "id": 139721, "pid": 76337, "tid": -914061504, "ts": 1716454223603901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223625322, "dur": 55,
    "args": {
      "External id": 139732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139732,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139732, "pid": 5, "tid": 7, "ts": 1716454223625322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223603996, "dur": 15,
    "args": {
      "External id": 139732,
      "cbid": 211, "correlation": 139732
    }
  },
  {
    "ph": "s", "id": 139732, "pid": 76337, "tid": -914061504, "ts": 1716454223603996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223625378, "dur": 94,
    "args": {
      "External id": 139754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139754,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139754, "pid": 5, "tid": 7, "ts": 1716454223625378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604031, "dur": 7,
    "args": {
      "External id": 139754,
      "cbid": 211, "correlation": 139754
    }
  },
  {
    "ph": "s", "id": 139754, "pid": 76337, "tid": -914061504, "ts": 1716454223604031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604116, "dur": 1,
    "args": {
      "External id": 139765,
      "cbid": 251, "correlation": 139765
    }
  },
  {
    "ph": "f", "id": 139765, "pid": 76337, "tid": -914061504, "ts": 1716454223604116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223625474, "dur": 106,
    "args": {
      "External id": 139766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139766,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139766, "pid": 5, "tid": 7, "ts": 1716454223625474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604121, "dur": 13,
    "args": {
      "External id": 139766,
      "cbid": 211, "correlation": 139766
    }
  },
  {
    "ph": "s", "id": 139766, "pid": 76337, "tid": -914061504, "ts": 1716454223604121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604203, "dur": 1,
    "args": {
      "External id": 139777,
      "cbid": 251, "correlation": 139777
    }
  },
  {
    "ph": "f", "id": 139777, "pid": 76337, "tid": -914061504, "ts": 1716454223604203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604207, "dur": 0,
    "args": {
      "External id": 139778,
      "cbid": 251, "correlation": 139778
    }
  },
  {
    "ph": "f", "id": 139778, "pid": 76337, "tid": -914061504, "ts": 1716454223604207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223625582, "dur": 10,
    "args": {
      "External id": 139779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139779,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139779, "pid": 5, "tid": 7, "ts": 1716454223625582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604209, "dur": 14,
    "args": {
      "External id": 139779,
      "cbid": 211, "correlation": 139779
    }
  },
  {
    "ph": "s", "id": 139779, "pid": 76337, "tid": -914061504, "ts": 1716454223604209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223625593, "dur": 5,
    "args": {
      "External id": 139781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139781,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 139781, "pid": 5, "tid": 7, "ts": 1716454223625593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604226, "dur": 8,
    "args": {
      "External id": 139781,
      "cbid": 211, "correlation": 139781
    }
  },
  {
    "ph": "s", "id": 139781, "pid": 76337, "tid": -914061504, "ts": 1716454223604226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604289, "dur": 1,
    "args": {
      "External id": 139792,
      "cbid": 251, "correlation": 139792
    }
  },
  {
    "ph": "f", "id": 139792, "pid": 76337, "tid": -914061504, "ts": 1716454223604289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604293, "dur": 0,
    "args": {
      "External id": 139793,
      "cbid": 251, "correlation": 139793
    }
  },
  {
    "ph": "f", "id": 139793, "pid": 76337, "tid": -914061504, "ts": 1716454223604293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223625600, "dur": 6,
    "args": {
      "External id": 139794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139794,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 139794, "pid": 5, "tid": 7, "ts": 1716454223625600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604294, "dur": 12,
    "args": {
      "External id": 139794,
      "cbid": 211, "correlation": 139794
    }
  },
  {
    "ph": "s", "id": 139794, "pid": 76337, "tid": -914061504, "ts": 1716454223604294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223625607, "dur": 4,
    "args": {
      "External id": 139796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139796,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 139796, "pid": 5, "tid": 7, "ts": 1716454223625607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604308, "dur": 5,
    "args": {
      "External id": 139796,
      "cbid": 211, "correlation": 139796
    }
  },
  {
    "ph": "s", "id": 139796, "pid": 76337, "tid": -914061504, "ts": 1716454223604308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223625612, "dur": 158,
    "args": {
      "External id": 139817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139817,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 139817, "pid": 5, "tid": 7, "ts": 1716454223625612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604381, "dur": 13,
    "args": {
      "External id": 139817,
      "cbid": 211, "correlation": 139817
    }
  },
  {
    "ph": "s", "id": 139817, "pid": 76337, "tid": -914061504, "ts": 1716454223604381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604479, "dur": 2,
    "args": {
      "External id": 139835,
      "cbid": 251, "correlation": 139835
    }
  },
  {
    "ph": "f", "id": 139835, "pid": 76337, "tid": -914061504, "ts": 1716454223604479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223625771, "dur": 107,
    "args": {
      "External id": 139837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139837,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 139837, "pid": 5, "tid": 7, "ts": 1716454223625771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604486, "dur": 14,
    "args": {
      "External id": 139837,
      "cbid": 211, "correlation": 139837
    }
  },
  {
    "ph": "s", "id": 139837, "pid": 76337, "tid": -914061504, "ts": 1716454223604486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223625880, "dur": 35,
    "args": {
      "External id": 139845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139845,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139845, "pid": 5, "tid": 7, "ts": 1716454223625880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604557, "dur": 13,
    "args": {
      "External id": 139845,
      "cbid": 211, "correlation": 139845
    }
  },
  {
    "ph": "s", "id": 139845, "pid": 76337, "tid": -914061504, "ts": 1716454223604557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223625916, "dur": 69,
    "args": {
      "External id": 139853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139853, "pid": 5, "tid": 7, "ts": 1716454223625916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604598, "dur": 9,
    "args": {
      "External id": 139853,
      "cbid": 211, "correlation": 139853
    }
  },
  {
    "ph": "s", "id": 139853, "pid": 76337, "tid": -914061504, "ts": 1716454223604598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223625987, "dur": 94,
    "args": {
      "External id": 139875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139875,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139875, "pid": 5, "tid": 7, "ts": 1716454223625987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604650, "dur": 10,
    "args": {
      "External id": 139875,
      "cbid": 211, "correlation": 139875
    }
  },
  {
    "ph": "s", "id": 139875, "pid": 76337, "tid": -914061504, "ts": 1716454223604650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604742, "dur": 1,
    "args": {
      "External id": 139891,
      "cbid": 251, "correlation": 139891
    }
  },
  {
    "ph": "f", "id": 139891, "pid": 76337, "tid": -914061504, "ts": 1716454223604742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223626082, "dur": 587,
    "args": {
      "External id": 139893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139893,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 139893, "pid": 5, "tid": 7, "ts": 1716454223626082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604748, "dur": 13,
    "args": {
      "External id": 139893,
      "cbid": 211, "correlation": 139893
    }
  },
  {
    "ph": "s", "id": 139893, "pid": 76337, "tid": -914061504, "ts": 1716454223604748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223626671, "dur": 248,
    "args": {
      "External id": 139901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139901, "pid": 5, "tid": 7, "ts": 1716454223626671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604827, "dur": 15,
    "args": {
      "External id": 139901,
      "cbid": 211, "correlation": 139901
    }
  },
  {
    "ph": "s", "id": 139901, "pid": 76337, "tid": -914061504, "ts": 1716454223604827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223626920, "dur": 255,
    "args": {
      "External id": 139909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139909, "pid": 5, "tid": 7, "ts": 1716454223626920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604863, "dur": 9,
    "args": {
      "External id": 139909,
      "cbid": 211, "correlation": 139909
    }
  },
  {
    "ph": "s", "id": 139909, "pid": 76337, "tid": -914061504, "ts": 1716454223604863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604947, "dur": 2,
    "args": {
      "External id": 139925,
      "cbid": 251, "correlation": 139925
    }
  },
  {
    "ph": "f", "id": 139925, "pid": 76337, "tid": -914061504, "ts": 1716454223604947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604953, "dur": 0,
    "args": {
      "External id": 139927,
      "cbid": 251, "correlation": 139927
    }
  },
  {
    "ph": "f", "id": 139927, "pid": 76337, "tid": -914061504, "ts": 1716454223604953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223627177, "dur": 363,
    "args": {
      "External id": 139928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139928,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 139928, "pid": 5, "tid": 7, "ts": 1716454223627177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223604957, "dur": 15,
    "args": {
      "External id": 139928,
      "cbid": 211, "correlation": 139928
    }
  },
  {
    "ph": "s", "id": 139928, "pid": 76337, "tid": -914061504, "ts": 1716454223604957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223627541, "dur": 50,
    "args": {
      "External id": 139936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139936,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139936, "pid": 5, "tid": 7, "ts": 1716454223627541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605009, "dur": 12,
    "args": {
      "External id": 139936,
      "cbid": 211, "correlation": 139936
    }
  },
  {
    "ph": "s", "id": 139936, "pid": 76337, "tid": -914061504, "ts": 1716454223605009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223627592, "dur": 160,
    "args": {
      "External id": 139947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139947,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139947, "pid": 5, "tid": 7, "ts": 1716454223627592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605082, "dur": 12,
    "args": {
      "External id": 139947,
      "cbid": 211, "correlation": 139947
    }
  },
  {
    "ph": "s", "id": 139947, "pid": 76337, "tid": -914061504, "ts": 1716454223605082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605148, "dur": 0,
    "args": {
      "External id": 139959,
      "cbid": 317, "correlation": 139959
    }
  },
  {
    "ph": "f", "id": 139959, "pid": 76337, "tid": -914061504, "ts": 1716454223605148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605149, "dur": 0,
    "args": {
      "External id": 139960,
      "cbid": 203, "correlation": 139960
    }
  },
  {
    "ph": "f", "id": 139960, "pid": 76337, "tid": -914061504, "ts": 1716454223605149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605150, "dur": 0,
    "args": {
      "External id": 139961,
      "cbid": 205, "correlation": 139961
    }
  },
  {
    "ph": "f", "id": 139961, "pid": 76337, "tid": -914061504, "ts": 1716454223605150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605175, "dur": 1,
    "args": {
      "External id": 139965,
      "cbid": 251, "correlation": 139965
    }
  },
  {
    "ph": "f", "id": 139965, "pid": 76337, "tid": -914061504, "ts": 1716454223605175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605177, "dur": 0,
    "args": {
      "External id": 139966,
      "cbid": 251, "correlation": 139966
    }
  },
  {
    "ph": "f", "id": 139966, "pid": 76337, "tid": -914061504, "ts": 1716454223605177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605177, "dur": 0,
    "args": {
      "External id": 139967,
      "cbid": 251, "correlation": 139967
    }
  },
  {
    "ph": "f", "id": 139967, "pid": 76337, "tid": -914061504, "ts": 1716454223605177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605178, "dur": 0,
    "args": {
      "External id": 139968,
      "cbid": 251, "correlation": 139968
    }
  },
  {
    "ph": "f", "id": 139968, "pid": 76337, "tid": -914061504, "ts": 1716454223605178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605179, "dur": 1,
    "args": {
      "External id": 139969,
      "cbid": 251, "correlation": 139969
    }
  },
  {
    "ph": "f", "id": 139969, "pid": 76337, "tid": -914061504, "ts": 1716454223605179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605180, "dur": 0,
    "args": {
      "External id": 139970,
      "cbid": 251, "correlation": 139970
    }
  },
  {
    "ph": "f", "id": 139970, "pid": 76337, "tid": -914061504, "ts": 1716454223605180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605181, "dur": 0,
    "args": {
      "External id": 139971,
      "cbid": 251, "correlation": 139971
    }
  },
  {
    "ph": "f", "id": 139971, "pid": 76337, "tid": -914061504, "ts": 1716454223605181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605182, "dur": 0,
    "args": {
      "External id": 139972,
      "cbid": 251, "correlation": 139972
    }
  },
  {
    "ph": "f", "id": 139972, "pid": 76337, "tid": -914061504, "ts": 1716454223605182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605183, "dur": 0,
    "args": {
      "External id": 139973,
      "cbid": 251, "correlation": 139973
    }
  },
  {
    "ph": "f", "id": 139973, "pid": 76337, "tid": -914061504, "ts": 1716454223605183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223627754, "dur": 117,
    "args": {
      "External id": 139974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139974,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 139974, "pid": 5, "tid": 7, "ts": 1716454223627754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605185, "dur": 12,
    "args": {
      "External id": 139974,
      "cbid": 211, "correlation": 139974
    }
  },
  {
    "ph": "s", "id": 139974, "pid": 76337, "tid": -914061504, "ts": 1716454223605185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223627873, "dur": 61,
    "args": {
      "External id": 139980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139980,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139980, "pid": 5, "tid": 7, "ts": 1716454223627873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605220, "dur": 9,
    "args": {
      "External id": 139980,
      "cbid": 211, "correlation": 139980
    }
  },
  {
    "ph": "s", "id": 139980, "pid": 76337, "tid": -914061504, "ts": 1716454223605220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223627935, "dur": 50,
    "args": {
      "External id": 139988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 139988,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 139988, "pid": 5, "tid": 7, "ts": 1716454223627935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605253, "dur": 8,
    "args": {
      "External id": 139988,
      "cbid": 211, "correlation": 139988
    }
  },
  {
    "ph": "s", "id": 139988, "pid": 76337, "tid": -914061504, "ts": 1716454223605253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223627986, "dur": 52,
    "args": {
      "External id": 140008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140008,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 140008, "pid": 5, "tid": 7, "ts": 1716454223627986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605328, "dur": 11,
    "args": {
      "External id": 140008,
      "cbid": 211, "correlation": 140008
    }
  },
  {
    "ph": "s", "id": 140008, "pid": 76337, "tid": -914061504, "ts": 1716454223605328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223628039, "dur": 5,
    "args": {
      "External id": 140020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140020,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140020, "pid": 5, "tid": 7, "ts": 1716454223628039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605349, "dur": 6,
    "args": {
      "External id": 140020,
      "cbid": 211, "correlation": 140020
    }
  },
  {
    "ph": "s", "id": 140020, "pid": 76337, "tid": -914061504, "ts": 1716454223605349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223628045, "dur": 58,
    "args": {
      "External id": 140023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140023, "pid": 5, "tid": 7, "ts": 1716454223628045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605368, "dur": 6,
    "args": {
      "External id": 140023,
      "cbid": 211, "correlation": 140023
    }
  },
  {
    "ph": "s", "id": 140023, "pid": 76337, "tid": -914061504, "ts": 1716454223605368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223628104, "dur": 37,
    "args": {
      "External id": 140032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140032,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140032, "pid": 5, "tid": 7, "ts": 1716454223628104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605407, "dur": 11,
    "args": {
      "External id": 140032,
      "cbid": 211, "correlation": 140032
    }
  },
  {
    "ph": "s", "id": 140032, "pid": 76337, "tid": -914061504, "ts": 1716454223605407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605460, "dur": 0,
    "args": {
      "External id": 140042,
      "cbid": 317, "correlation": 140042
    }
  },
  {
    "ph": "f", "id": 140042, "pid": 76337, "tid": -914061504, "ts": 1716454223605460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605460, "dur": 0,
    "args": {
      "External id": 140043,
      "cbid": 203, "correlation": 140043
    }
  },
  {
    "ph": "f", "id": 140043, "pid": 76337, "tid": -914061504, "ts": 1716454223605460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605461, "dur": 0,
    "args": {
      "External id": 140044,
      "cbid": 205, "correlation": 140044
    }
  },
  {
    "ph": "f", "id": 140044, "pid": 76337, "tid": -914061504, "ts": 1716454223605461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223628142, "dur": 40,
    "args": {
      "External id": 140048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140048, "pid": 5, "tid": 7, "ts": 1716454223628142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605476, "dur": 12,
    "args": {
      "External id": 140048,
      "cbid": 211, "correlation": 140048
    }
  },
  {
    "ph": "s", "id": 140048, "pid": 76337, "tid": -914061504, "ts": 1716454223605476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223628184, "dur": 15,
    "args": {
      "External id": 140050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140050, "pid": 5, "tid": 7, "ts": 1716454223628184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605490, "dur": 5,
    "args": {
      "External id": 140050,
      "cbid": 211, "correlation": 140050
    }
  },
  {
    "ph": "s", "id": 140050, "pid": 76337, "tid": -914061504, "ts": 1716454223605490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223628200, "dur": 4,
    "args": {
      "External id": 140052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140052, "pid": 5, "tid": 7, "ts": 1716454223628200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605499, "dur": 5,
    "args": {
      "External id": 140052,
      "cbid": 211, "correlation": 140052
    }
  },
  {
    "ph": "s", "id": 140052, "pid": 76337, "tid": -914061504, "ts": 1716454223605499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605508, "dur": 0,
    "args": {
      "External id": 140053,
      "cbid": 51, "correlation": 140053
    }
  },
  {
    "ph": "s", "id": 140053, "pid": 76337, "tid": -914061504, "ts": 1716454223605508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223628205, "dur": 715,
    "args": {
      "External id": 140054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140054,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140054, "pid": 5, "tid": 7, "ts": 1716454223628205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605509, "dur": 6,
    "args": {
      "External id": 140054,
      "cbid": 211, "correlation": 140054
    }
  },
  {
    "ph": "s", "id": 140054, "pid": 76337, "tid": -914061504, "ts": 1716454223605509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223628921, "dur": 60,
    "args": {
      "External id": 140059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140059, "pid": 5, "tid": 7, "ts": 1716454223628921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605537, "dur": 8,
    "args": {
      "External id": 140059,
      "cbid": 211, "correlation": 140059
    }
  },
  {
    "ph": "s", "id": 140059, "pid": 76337, "tid": -914061504, "ts": 1716454223605537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223628982, "dur": 4,
    "args": {
      "External id": 140067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140067,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140067, "pid": 5, "tid": 7, "ts": 1716454223628982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605580, "dur": 10,
    "args": {
      "External id": 140067,
      "cbid": 211, "correlation": 140067
    }
  },
  {
    "ph": "s", "id": 140067, "pid": 76337, "tid": -914061504, "ts": 1716454223605580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605645, "dur": 1,
    "args": {
      "External id": 140083,
      "cbid": 251, "correlation": 140083
    }
  },
  {
    "ph": "f", "id": 140083, "pid": 76337, "tid": -914061504, "ts": 1716454223605645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605650, "dur": 0,
    "args": {
      "External id": 140085,
      "cbid": 251, "correlation": 140085
    }
  },
  {
    "ph": "f", "id": 140085, "pid": 76337, "tid": -914061504, "ts": 1716454223605650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223628987, "dur": 11,
    "args": {
      "External id": 140086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140086,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 140086, "pid": 5, "tid": 7, "ts": 1716454223628987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605652, "dur": 12,
    "args": {
      "External id": 140086,
      "cbid": 211, "correlation": 140086
    }
  },
  {
    "ph": "s", "id": 140086, "pid": 76337, "tid": -914061504, "ts": 1716454223605652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223629000, "dur": 5,
    "args": {
      "External id": 140088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140088,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 140088, "pid": 5, "tid": 7, "ts": 1716454223629000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605666, "dur": 6,
    "args": {
      "External id": 140088,
      "cbid": 211, "correlation": 140088
    }
  },
  {
    "ph": "s", "id": 140088, "pid": 76337, "tid": -914061504, "ts": 1716454223605666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223629007, "dur": 55,
    "args": {
      "External id": 140098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140098, "pid": 5, "tid": 7, "ts": 1716454223629007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605723, "dur": 12,
    "args": {
      "External id": 140098,
      "cbid": 211, "correlation": 140098
    }
  },
  {
    "ph": "s", "id": 140098, "pid": 76337, "tid": -914061504, "ts": 1716454223605723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223629063, "dur": 51,
    "args": {
      "External id": 140118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140118,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 140118, "pid": 5, "tid": 7, "ts": 1716454223629063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605791, "dur": 11,
    "args": {
      "External id": 140118,
      "cbid": 211, "correlation": 140118
    }
  },
  {
    "ph": "s", "id": 140118, "pid": 76337, "tid": -914061504, "ts": 1716454223605791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223629115, "dur": 4,
    "args": {
      "External id": 140130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140130,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140130, "pid": 5, "tid": 7, "ts": 1716454223629115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605811, "dur": 7,
    "args": {
      "External id": 140130,
      "cbid": 211, "correlation": 140130
    }
  },
  {
    "ph": "s", "id": 140130, "pid": 76337, "tid": -914061504, "ts": 1716454223605811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223629120, "dur": 56,
    "args": {
      "External id": 140133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140133, "pid": 5, "tid": 7, "ts": 1716454223629120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605830, "dur": 7,
    "args": {
      "External id": 140133,
      "cbid": 211, "correlation": 140133
    }
  },
  {
    "ph": "s", "id": 140133, "pid": 76337, "tid": -914061504, "ts": 1716454223605830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223629177, "dur": 37,
    "args": {
      "External id": 140142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140142,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140142, "pid": 5, "tid": 7, "ts": 1716454223629177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605870, "dur": 10,
    "args": {
      "External id": 140142,
      "cbid": 211, "correlation": 140142
    }
  },
  {
    "ph": "s", "id": 140142, "pid": 76337, "tid": -914061504, "ts": 1716454223605870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605934, "dur": 0,
    "args": {
      "External id": 140152,
      "cbid": 317, "correlation": 140152
    }
  },
  {
    "ph": "f", "id": 140152, "pid": 76337, "tid": -914061504, "ts": 1716454223605934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605935, "dur": 0,
    "args": {
      "External id": 140153,
      "cbid": 203, "correlation": 140153
    }
  },
  {
    "ph": "f", "id": 140153, "pid": 76337, "tid": -914061504, "ts": 1716454223605935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605936, "dur": 0,
    "args": {
      "External id": 140154,
      "cbid": 205, "correlation": 140154
    }
  },
  {
    "ph": "f", "id": 140154, "pid": 76337, "tid": -914061504, "ts": 1716454223605936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223629216, "dur": 40,
    "args": {
      "External id": 140158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140158,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140158, "pid": 5, "tid": 7, "ts": 1716454223629216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605949, "dur": 12,
    "args": {
      "External id": 140158,
      "cbid": 211, "correlation": 140158
    }
  },
  {
    "ph": "s", "id": 140158, "pid": 76337, "tid": -914061504, "ts": 1716454223605949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223629257, "dur": 15,
    "args": {
      "External id": 140160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140160,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140160, "pid": 5, "tid": 7, "ts": 1716454223629257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605963, "dur": 5,
    "args": {
      "External id": 140160,
      "cbid": 211, "correlation": 140160
    }
  },
  {
    "ph": "s", "id": 140160, "pid": 76337, "tid": -914061504, "ts": 1716454223605963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223629273, "dur": 3,
    "args": {
      "External id": 140162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140162, "pid": 5, "tid": 7, "ts": 1716454223629273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605981, "dur": 6,
    "args": {
      "External id": 140162,
      "cbid": 211, "correlation": 140162
    }
  },
  {
    "ph": "s", "id": 140162, "pid": 76337, "tid": -914061504, "ts": 1716454223605981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605990, "dur": 0,
    "args": {
      "External id": 140163,
      "cbid": 51, "correlation": 140163
    }
  },
  {
    "ph": "s", "id": 140163, "pid": 76337, "tid": -914061504, "ts": 1716454223605990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223629278, "dur": 708,
    "args": {
      "External id": 140164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140164,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140164, "pid": 5, "tid": 7, "ts": 1716454223629278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223605991, "dur": 5,
    "args": {
      "External id": 140164,
      "cbid": 211, "correlation": 140164
    }
  },
  {
    "ph": "s", "id": 140164, "pid": 76337, "tid": -914061504, "ts": 1716454223605991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223629988, "dur": 60,
    "args": {
      "External id": 140169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140169, "pid": 5, "tid": 7, "ts": 1716454223629988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606019, "dur": 9,
    "args": {
      "External id": 140169,
      "cbid": 211, "correlation": 140169
    }
  },
  {
    "ph": "s", "id": 140169, "pid": 76337, "tid": -914061504, "ts": 1716454223606019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223630049, "dur": 50,
    "args": {
      "External id": 140177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140177,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140177, "pid": 5, "tid": 7, "ts": 1716454223630049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606052, "dur": 8,
    "args": {
      "External id": 140177,
      "cbid": 211, "correlation": 140177
    }
  },
  {
    "ph": "s", "id": 140177, "pid": 76337, "tid": -914061504, "ts": 1716454223606052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223630101, "dur": 36,
    "args": {
      "External id": 140185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140185,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140185, "pid": 5, "tid": 7, "ts": 1716454223630101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606081, "dur": 8,
    "args": {
      "External id": 140185,
      "cbid": 211, "correlation": 140185
    }
  },
  {
    "ph": "s", "id": 140185, "pid": 76337, "tid": -914061504, "ts": 1716454223606081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223630138, "dur": 51,
    "args": {
      "External id": 140205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140205,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 140205, "pid": 5, "tid": 7, "ts": 1716454223630138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606159, "dur": 13,
    "args": {
      "External id": 140205,
      "cbid": 211, "correlation": 140205
    }
  },
  {
    "ph": "s", "id": 140205, "pid": 76337, "tid": -914061504, "ts": 1716454223606159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223630191, "dur": 4,
    "args": {
      "External id": 140217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140217,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140217, "pid": 5, "tid": 7, "ts": 1716454223630191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606182, "dur": 6,
    "args": {
      "External id": 140217,
      "cbid": 211, "correlation": 140217
    }
  },
  {
    "ph": "s", "id": 140217, "pid": 76337, "tid": -914061504, "ts": 1716454223606182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223630196, "dur": 56,
    "args": {
      "External id": 140220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140220,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140220, "pid": 5, "tid": 7, "ts": 1716454223630196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606199, "dur": 7,
    "args": {
      "External id": 140220,
      "cbid": 211, "correlation": 140220
    }
  },
  {
    "ph": "s", "id": 140220, "pid": 76337, "tid": -914061504, "ts": 1716454223606199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606257, "dur": 0,
    "args": {
      "External id": 140231,
      "cbid": 317, "correlation": 140231
    }
  },
  {
    "ph": "f", "id": 140231, "pid": 76337, "tid": -914061504, "ts": 1716454223606257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606258, "dur": 0,
    "args": {
      "External id": 140232,
      "cbid": 203, "correlation": 140232
    }
  },
  {
    "ph": "f", "id": 140232, "pid": 76337, "tid": -914061504, "ts": 1716454223606258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606258, "dur": 0,
    "args": {
      "External id": 140233,
      "cbid": 205, "correlation": 140233
    }
  },
  {
    "ph": "f", "id": 140233, "pid": 76337, "tid": -914061504, "ts": 1716454223606258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606281, "dur": 1,
    "args": {
      "External id": 140237,
      "cbid": 251, "correlation": 140237
    }
  },
  {
    "ph": "f", "id": 140237, "pid": 76337, "tid": -914061504, "ts": 1716454223606281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606283, "dur": 0,
    "args": {
      "External id": 140238,
      "cbid": 251, "correlation": 140238
    }
  },
  {
    "ph": "f", "id": 140238, "pid": 76337, "tid": -914061504, "ts": 1716454223606283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606284, "dur": 0,
    "args": {
      "External id": 140239,
      "cbid": 251, "correlation": 140239
    }
  },
  {
    "ph": "f", "id": 140239, "pid": 76337, "tid": -914061504, "ts": 1716454223606284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606284, "dur": 0,
    "args": {
      "External id": 140240,
      "cbid": 251, "correlation": 140240
    }
  },
  {
    "ph": "f", "id": 140240, "pid": 76337, "tid": -914061504, "ts": 1716454223606284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606285, "dur": 0,
    "args": {
      "External id": 140241,
      "cbid": 251, "correlation": 140241
    }
  },
  {
    "ph": "f", "id": 140241, "pid": 76337, "tid": -914061504, "ts": 1716454223606285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606286, "dur": 0,
    "args": {
      "External id": 140242,
      "cbid": 251, "correlation": 140242
    }
  },
  {
    "ph": "f", "id": 140242, "pid": 76337, "tid": -914061504, "ts": 1716454223606286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606287, "dur": 0,
    "args": {
      "External id": 140243,
      "cbid": 251, "correlation": 140243
    }
  },
  {
    "ph": "f", "id": 140243, "pid": 76337, "tid": -914061504, "ts": 1716454223606287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606287, "dur": 0,
    "args": {
      "External id": 140244,
      "cbid": 251, "correlation": 140244
    }
  },
  {
    "ph": "f", "id": 140244, "pid": 76337, "tid": -914061504, "ts": 1716454223606287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606289, "dur": 0,
    "args": {
      "External id": 140245,
      "cbid": 251, "correlation": 140245
    }
  },
  {
    "ph": "f", "id": 140245, "pid": 76337, "tid": -914061504, "ts": 1716454223606289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223630253, "dur": 113,
    "args": {
      "External id": 140246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140246,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 140246, "pid": 5, "tid": 7, "ts": 1716454223630253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606291, "dur": 12,
    "args": {
      "External id": 140246,
      "cbid": 211, "correlation": 140246
    }
  },
  {
    "ph": "s", "id": 140246, "pid": 76337, "tid": -914061504, "ts": 1716454223606291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223630368, "dur": 59,
    "args": {
      "External id": 140252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140252,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140252, "pid": 5, "tid": 7, "ts": 1716454223630368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606326, "dur": 9,
    "args": {
      "External id": 140252,
      "cbid": 211, "correlation": 140252
    }
  },
  {
    "ph": "s", "id": 140252, "pid": 76337, "tid": -914061504, "ts": 1716454223606326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223630429, "dur": 525,
    "args": {
      "External id": 140261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140261,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140261, "pid": 5, "tid": 7, "ts": 1716454223630429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606408, "dur": 15,
    "args": {
      "External id": 140261,
      "cbid": 211, "correlation": 140261
    }
  },
  {
    "ph": "s", "id": 140261, "pid": 76337, "tid": -914061504, "ts": 1716454223606408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223630955, "dur": 184,
    "args": {
      "External id": 140283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140283,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140283, "pid": 5, "tid": 7, "ts": 1716454223630955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606465, "dur": 11,
    "args": {
      "External id": 140283,
      "cbid": 211, "correlation": 140283
    }
  },
  {
    "ph": "s", "id": 140283, "pid": 76337, "tid": -914061504, "ts": 1716454223606465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606549, "dur": 1,
    "args": {
      "External id": 140294,
      "cbid": 251, "correlation": 140294
    }
  },
  {
    "ph": "f", "id": 140294, "pid": 76337, "tid": -914061504, "ts": 1716454223606549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223631141, "dur": 198,
    "args": {
      "External id": 140295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140295,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140295, "pid": 5, "tid": 7, "ts": 1716454223631141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606554, "dur": 13,
    "args": {
      "External id": 140295,
      "cbid": 211, "correlation": 140295
    }
  },
  {
    "ph": "s", "id": 140295, "pid": 76337, "tid": -914061504, "ts": 1716454223606554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606623, "dur": 1,
    "args": {
      "External id": 140306,
      "cbid": 251, "correlation": 140306
    }
  },
  {
    "ph": "f", "id": 140306, "pid": 76337, "tid": -914061504, "ts": 1716454223606623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223631339, "dur": 187,
    "args": {
      "External id": 140307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140307,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140307, "pid": 5, "tid": 7, "ts": 1716454223631339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606627, "dur": 11,
    "args": {
      "External id": 140307,
      "cbid": 211, "correlation": 140307
    }
  },
  {
    "ph": "s", "id": 140307, "pid": 76337, "tid": -914061504, "ts": 1716454223606627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606688, "dur": 1,
    "args": {
      "External id": 140318,
      "cbid": 251, "correlation": 140318
    }
  },
  {
    "ph": "f", "id": 140318, "pid": 76337, "tid": -914061504, "ts": 1716454223606688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223631527, "dur": 188,
    "args": {
      "External id": 140319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140319,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140319, "pid": 5, "tid": 7, "ts": 1716454223631527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606692, "dur": 11,
    "args": {
      "External id": 140319,
      "cbid": 211, "correlation": 140319
    }
  },
  {
    "ph": "s", "id": 140319, "pid": 76337, "tid": -914061504, "ts": 1716454223606692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223631717, "dur": 19029,
    "args": {
      "External id": 140340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140340,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 140340, "pid": 5, "tid": 7, "ts": 1716454223631717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606774, "dur": 12,
    "args": {
      "External id": 140340,
      "cbid": 211, "correlation": 140340
    }
  },
  {
    "ph": "s", "id": 140340, "pid": 76337, "tid": -914061504, "ts": 1716454223606774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606870, "dur": 1,
    "args": {
      "External id": 140358,
      "cbid": 251, "correlation": 140358
    }
  },
  {
    "ph": "f", "id": 140358, "pid": 76337, "tid": -914061504, "ts": 1716454223606870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223650747, "dur": 202,
    "args": {
      "External id": 140360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140360,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140360, "pid": 5, "tid": 7, "ts": 1716454223650747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606877, "dur": 13,
    "args": {
      "External id": 140360,
      "cbid": 211, "correlation": 140360
    }
  },
  {
    "ph": "s", "id": 140360, "pid": 76337, "tid": -914061504, "ts": 1716454223606877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223650950, "dur": 66,
    "args": {
      "External id": 140368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140368,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140368, "pid": 5, "tid": 7, "ts": 1716454223650950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606946, "dur": 12,
    "args": {
      "External id": 140368,
      "cbid": 211, "correlation": 140368
    }
  },
  {
    "ph": "s", "id": 140368, "pid": 76337, "tid": -914061504, "ts": 1716454223606946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223651018, "dur": 97,
    "args": {
      "External id": 140376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140376,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140376, "pid": 5, "tid": 7, "ts": 1716454223651018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223606993, "dur": 10,
    "args": {
      "External id": 140376,
      "cbid": 211, "correlation": 140376
    }
  },
  {
    "ph": "s", "id": 140376, "pid": 76337, "tid": -914061504, "ts": 1716454223606993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223651116, "dur": 55,
    "args": {
      "External id": 140387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140387, "pid": 5, "tid": 7, "ts": 1716454223651116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607066, "dur": 13,
    "args": {
      "External id": 140387,
      "cbid": 211, "correlation": 140387
    }
  },
  {
    "ph": "s", "id": 140387, "pid": 76337, "tid": -914061504, "ts": 1716454223607066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223651172, "dur": 93,
    "args": {
      "External id": 140409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140409,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140409, "pid": 5, "tid": 7, "ts": 1716454223651172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607098, "dur": 7,
    "args": {
      "External id": 140409,
      "cbid": 211, "correlation": 140409
    }
  },
  {
    "ph": "s", "id": 140409, "pid": 76337, "tid": -914061504, "ts": 1716454223607098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607182, "dur": 1,
    "args": {
      "External id": 140420,
      "cbid": 251, "correlation": 140420
    }
  },
  {
    "ph": "f", "id": 140420, "pid": 76337, "tid": -914061504, "ts": 1716454223607182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223651267, "dur": 106,
    "args": {
      "External id": 140421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140421,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140421, "pid": 5, "tid": 7, "ts": 1716454223651267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607187, "dur": 13,
    "args": {
      "External id": 140421,
      "cbid": 211, "correlation": 140421
    }
  },
  {
    "ph": "s", "id": 140421, "pid": 76337, "tid": -914061504, "ts": 1716454223607187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607257, "dur": 1,
    "args": {
      "External id": 140432,
      "cbid": 251, "correlation": 140432
    }
  },
  {
    "ph": "f", "id": 140432, "pid": 76337, "tid": -914061504, "ts": 1716454223607257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607260, "dur": 0,
    "args": {
      "External id": 140433,
      "cbid": 251, "correlation": 140433
    }
  },
  {
    "ph": "f", "id": 140433, "pid": 76337, "tid": -914061504, "ts": 1716454223607260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223651374, "dur": 10,
    "args": {
      "External id": 140434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140434,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 140434, "pid": 5, "tid": 7, "ts": 1716454223651374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607262, "dur": 12,
    "args": {
      "External id": 140434,
      "cbid": 211, "correlation": 140434
    }
  },
  {
    "ph": "s", "id": 140434, "pid": 76337, "tid": -914061504, "ts": 1716454223607262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223651385, "dur": 5,
    "args": {
      "External id": 140436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140436,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 140436, "pid": 5, "tid": 7, "ts": 1716454223651385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607276, "dur": 6,
    "args": {
      "External id": 140436,
      "cbid": 211, "correlation": 140436
    }
  },
  {
    "ph": "s", "id": 140436, "pid": 76337, "tid": -914061504, "ts": 1716454223607276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607336, "dur": 1,
    "args": {
      "External id": 140447,
      "cbid": 251, "correlation": 140447
    }
  },
  {
    "ph": "f", "id": 140447, "pid": 76337, "tid": -914061504, "ts": 1716454223607336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607339, "dur": 0,
    "args": {
      "External id": 140448,
      "cbid": 251, "correlation": 140448
    }
  },
  {
    "ph": "f", "id": 140448, "pid": 76337, "tid": -914061504, "ts": 1716454223607339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223651391, "dur": 6,
    "args": {
      "External id": 140449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140449,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 140449, "pid": 5, "tid": 7, "ts": 1716454223651391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607341, "dur": 12,
    "args": {
      "External id": 140449,
      "cbid": 211, "correlation": 140449
    }
  },
  {
    "ph": "s", "id": 140449, "pid": 76337, "tid": -914061504, "ts": 1716454223607341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223651399, "dur": 3,
    "args": {
      "External id": 140451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140451,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 140451, "pid": 5, "tid": 7, "ts": 1716454223651399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607354, "dur": 6,
    "args": {
      "External id": 140451,
      "cbid": 211, "correlation": 140451
    }
  },
  {
    "ph": "s", "id": 140451, "pid": 76337, "tid": -914061504, "ts": 1716454223607354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223651404, "dur": 158,
    "args": {
      "External id": 140472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140472,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 140472, "pid": 5, "tid": 7, "ts": 1716454223651404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607429, "dur": 12,
    "args": {
      "External id": 140472,
      "cbid": 211, "correlation": 140472
    }
  },
  {
    "ph": "s", "id": 140472, "pid": 76337, "tid": -914061504, "ts": 1716454223607429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607525, "dur": 1,
    "args": {
      "External id": 140490,
      "cbid": 251, "correlation": 140490
    }
  },
  {
    "ph": "f", "id": 140490, "pid": 76337, "tid": -914061504, "ts": 1716454223607525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223651563, "dur": 107,
    "args": {
      "External id": 140492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140492,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 140492, "pid": 5, "tid": 7, "ts": 1716454223651563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607531, "dur": 13,
    "args": {
      "External id": 140492,
      "cbid": 211, "correlation": 140492
    }
  },
  {
    "ph": "s", "id": 140492, "pid": 76337, "tid": -914061504, "ts": 1716454223607531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223651671, "dur": 35,
    "args": {
      "External id": 140500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140500,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140500, "pid": 5, "tid": 7, "ts": 1716454223651671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607600, "dur": 12,
    "args": {
      "External id": 140500,
      "cbid": 211, "correlation": 140500
    }
  },
  {
    "ph": "s", "id": 140500, "pid": 76337, "tid": -914061504, "ts": 1716454223607600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223651708, "dur": 67,
    "args": {
      "External id": 140508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140508, "pid": 5, "tid": 7, "ts": 1716454223651708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607641, "dur": 9,
    "args": {
      "External id": 140508,
      "cbid": 211, "correlation": 140508
    }
  },
  {
    "ph": "s", "id": 140508, "pid": 76337, "tid": -914061504, "ts": 1716454223607641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223651777, "dur": 94,
    "args": {
      "External id": 140530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140530,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140530, "pid": 5, "tid": 7, "ts": 1716454223651777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607693, "dur": 10,
    "args": {
      "External id": 140530,
      "cbid": 211, "correlation": 140530
    }
  },
  {
    "ph": "s", "id": 140530, "pid": 76337, "tid": -914061504, "ts": 1716454223607693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607777, "dur": 1,
    "args": {
      "External id": 140546,
      "cbid": 251, "correlation": 140546
    }
  },
  {
    "ph": "f", "id": 140546, "pid": 76337, "tid": -914061504, "ts": 1716454223607777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223651872, "dur": 580,
    "args": {
      "External id": 140548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140548,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140548, "pid": 5, "tid": 7, "ts": 1716454223651872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607784, "dur": 12,
    "args": {
      "External id": 140548,
      "cbid": 211, "correlation": 140548
    }
  },
  {
    "ph": "s", "id": 140548, "pid": 76337, "tid": -914061504, "ts": 1716454223607784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223652453, "dur": 246,
    "args": {
      "External id": 140556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140556, "pid": 5, "tid": 7, "ts": 1716454223652453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607848, "dur": 12,
    "args": {
      "External id": 140556,
      "cbid": 211, "correlation": 140556
    }
  },
  {
    "ph": "s", "id": 140556, "pid": 76337, "tid": -914061504, "ts": 1716454223607848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223652700, "dur": 252,
    "args": {
      "External id": 140564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140564,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140564, "pid": 5, "tid": 7, "ts": 1716454223652700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607877, "dur": 9,
    "args": {
      "External id": 140564,
      "cbid": 211, "correlation": 140564
    }
  },
  {
    "ph": "s", "id": 140564, "pid": 76337, "tid": -914061504, "ts": 1716454223607877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607958, "dur": 1,
    "args": {
      "External id": 140580,
      "cbid": 251, "correlation": 140580
    }
  },
  {
    "ph": "f", "id": 140580, "pid": 76337, "tid": -914061504, "ts": 1716454223607958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607963, "dur": 0,
    "args": {
      "External id": 140582,
      "cbid": 251, "correlation": 140582
    }
  },
  {
    "ph": "f", "id": 140582, "pid": 76337, "tid": -914061504, "ts": 1716454223607963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223652953, "dur": 360,
    "args": {
      "External id": 140583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140583,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 140583, "pid": 5, "tid": 7, "ts": 1716454223652953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223607966, "dur": 21,
    "args": {
      "External id": 140583,
      "cbid": 211, "correlation": 140583
    }
  },
  {
    "ph": "s", "id": 140583, "pid": 76337, "tid": -914061504, "ts": 1716454223607966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223653315, "dur": 49,
    "args": {
      "External id": 140591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140591,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140591, "pid": 5, "tid": 7, "ts": 1716454223653315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608017, "dur": 10,
    "args": {
      "External id": 140591,
      "cbid": 211, "correlation": 140591
    }
  },
  {
    "ph": "s", "id": 140591, "pid": 76337, "tid": -914061504, "ts": 1716454223608017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223653365, "dur": 157,
    "args": {
      "External id": 140602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140602, "pid": 5, "tid": 7, "ts": 1716454223653365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608084, "dur": 12,
    "args": {
      "External id": 140602,
      "cbid": 211, "correlation": 140602
    }
  },
  {
    "ph": "s", "id": 140602, "pid": 76337, "tid": -914061504, "ts": 1716454223608084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608148, "dur": 0,
    "args": {
      "External id": 140614,
      "cbid": 317, "correlation": 140614
    }
  },
  {
    "ph": "f", "id": 140614, "pid": 76337, "tid": -914061504, "ts": 1716454223608148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608149, "dur": 0,
    "args": {
      "External id": 140615,
      "cbid": 203, "correlation": 140615
    }
  },
  {
    "ph": "f", "id": 140615, "pid": 76337, "tid": -914061504, "ts": 1716454223608149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608149, "dur": 0,
    "args": {
      "External id": 140616,
      "cbid": 205, "correlation": 140616
    }
  },
  {
    "ph": "f", "id": 140616, "pid": 76337, "tid": -914061504, "ts": 1716454223608149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608172, "dur": 1,
    "args": {
      "External id": 140620,
      "cbid": 251, "correlation": 140620
    }
  },
  {
    "ph": "f", "id": 140620, "pid": 76337, "tid": -914061504, "ts": 1716454223608172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608173, "dur": 0,
    "args": {
      "External id": 140621,
      "cbid": 251, "correlation": 140621
    }
  },
  {
    "ph": "f", "id": 140621, "pid": 76337, "tid": -914061504, "ts": 1716454223608173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608174, "dur": 0,
    "args": {
      "External id": 140622,
      "cbid": 251, "correlation": 140622
    }
  },
  {
    "ph": "f", "id": 140622, "pid": 76337, "tid": -914061504, "ts": 1716454223608174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608175, "dur": 0,
    "args": {
      "External id": 140623,
      "cbid": 251, "correlation": 140623
    }
  },
  {
    "ph": "f", "id": 140623, "pid": 76337, "tid": -914061504, "ts": 1716454223608175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608176, "dur": 0,
    "args": {
      "External id": 140624,
      "cbid": 251, "correlation": 140624
    }
  },
  {
    "ph": "f", "id": 140624, "pid": 76337, "tid": -914061504, "ts": 1716454223608176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608176, "dur": 0,
    "args": {
      "External id": 140625,
      "cbid": 251, "correlation": 140625
    }
  },
  {
    "ph": "f", "id": 140625, "pid": 76337, "tid": -914061504, "ts": 1716454223608176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608177, "dur": 0,
    "args": {
      "External id": 140626,
      "cbid": 251, "correlation": 140626
    }
  },
  {
    "ph": "f", "id": 140626, "pid": 76337, "tid": -914061504, "ts": 1716454223608177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608178, "dur": 0,
    "args": {
      "External id": 140627,
      "cbid": 251, "correlation": 140627
    }
  },
  {
    "ph": "f", "id": 140627, "pid": 76337, "tid": -914061504, "ts": 1716454223608178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608179, "dur": 0,
    "args": {
      "External id": 140628,
      "cbid": 251, "correlation": 140628
    }
  },
  {
    "ph": "f", "id": 140628, "pid": 76337, "tid": -914061504, "ts": 1716454223608179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223653524, "dur": 115,
    "args": {
      "External id": 140629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140629,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 140629, "pid": 5, "tid": 7, "ts": 1716454223653524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608181, "dur": 13,
    "args": {
      "External id": 140629,
      "cbid": 211, "correlation": 140629
    }
  },
  {
    "ph": "s", "id": 140629, "pid": 76337, "tid": -914061504, "ts": 1716454223608181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223653641, "dur": 59,
    "args": {
      "External id": 140635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140635, "pid": 5, "tid": 7, "ts": 1716454223653641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608216, "dur": 9,
    "args": {
      "External id": 140635,
      "cbid": 211, "correlation": 140635
    }
  },
  {
    "ph": "s", "id": 140635, "pid": 76337, "tid": -914061504, "ts": 1716454223608216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223653702, "dur": 50,
    "args": {
      "External id": 140643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140643,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140643, "pid": 5, "tid": 7, "ts": 1716454223653702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608248, "dur": 8,
    "args": {
      "External id": 140643,
      "cbid": 211, "correlation": 140643
    }
  },
  {
    "ph": "s", "id": 140643, "pid": 76337, "tid": -914061504, "ts": 1716454223608248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608319, "dur": 0,
    "args": {
      "External id": 140653,
      "cbid": 317, "correlation": 140653
    }
  },
  {
    "ph": "f", "id": 140653, "pid": 76337, "tid": -914061504, "ts": 1716454223608319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608320, "dur": 0,
    "args": {
      "External id": 140654,
      "cbid": 203, "correlation": 140654
    }
  },
  {
    "ph": "f", "id": 140654, "pid": 76337, "tid": -914061504, "ts": 1716454223608320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608321, "dur": 0,
    "args": {
      "External id": 140655,
      "cbid": 205, "correlation": 140655
    }
  },
  {
    "ph": "f", "id": 140655, "pid": 76337, "tid": -914061504, "ts": 1716454223608321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223653753, "dur": 41,
    "args": {
      "External id": 140659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140659, "pid": 5, "tid": 7, "ts": 1716454223653753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608337, "dur": 12,
    "args": {
      "External id": 140659,
      "cbid": 211, "correlation": 140659
    }
  },
  {
    "ph": "s", "id": 140659, "pid": 76337, "tid": -914061504, "ts": 1716454223608337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223653795, "dur": 15,
    "args": {
      "External id": 140661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140661, "pid": 5, "tid": 7, "ts": 1716454223653795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608352, "dur": 5,
    "args": {
      "External id": 140661,
      "cbid": 211, "correlation": 140661
    }
  },
  {
    "ph": "s", "id": 140661, "pid": 76337, "tid": -914061504, "ts": 1716454223608352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223653812, "dur": 1,
    "args": {
      "External id": 140663,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 140663,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 140663, "pid": 5, "tid": 7, "ts": 1716454223653812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608372, "dur": 17,
    "args": {
      "External id": 140663,
      "cbid": 51, "correlation": 140663
    }
  },
  {
    "ph": "s", "id": 140663, "pid": 76337, "tid": -914061504, "ts": 1716454223608372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223653815, "dur": 364,
    "args": {
      "External id": 140664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140664,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140664, "pid": 5, "tid": 7, "ts": 1716454223653815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608390, "dur": 10,
    "args": {
      "External id": 140664,
      "cbid": 211, "correlation": 140664
    }
  },
  {
    "ph": "s", "id": 140664, "pid": 76337, "tid": -914061504, "ts": 1716454223608390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654181, "dur": 13,
    "args": {
      "External id": 140666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140666, "pid": 5, "tid": 7, "ts": 1716454223654181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608408, "dur": 7,
    "args": {
      "External id": 140666,
      "cbid": 211, "correlation": 140666
    }
  },
  {
    "ph": "s", "id": 140666, "pid": 76337, "tid": -914061504, "ts": 1716454223608408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223654195, "dur": 15,
    "args": {
      "External id": 140672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140672, "pid": 5, "tid": 7, "ts": 1716454223654195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608440, "dur": 9,
    "args": {
      "External id": 140672,
      "cbid": 211, "correlation": 140672
    }
  },
  {
    "ph": "s", "id": 140672, "pid": 76337, "tid": -914061504, "ts": 1716454223608440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654211, "dur": 19,
    "args": {
      "External id": 140692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140692,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 140692, "pid": 5, "tid": 7, "ts": 1716454223654211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608534, "dur": 12,
    "args": {
      "External id": 140692,
      "cbid": 211, "correlation": 140692
    }
  },
  {
    "ph": "s", "id": 140692, "pid": 76337, "tid": -914061504, "ts": 1716454223608534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223654231, "dur": 4,
    "args": {
      "External id": 140704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140704,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140704, "pid": 5, "tid": 7, "ts": 1716454223654231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608556, "dur": 6,
    "args": {
      "External id": 140704,
      "cbid": 211, "correlation": 140704
    }
  },
  {
    "ph": "s", "id": 140704, "pid": 76337, "tid": -914061504, "ts": 1716454223608556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223654237, "dur": 17,
    "args": {
      "External id": 140707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140707,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140707, "pid": 5, "tid": 7, "ts": 1716454223654237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608574, "dur": 8,
    "args": {
      "External id": 140707,
      "cbid": 211, "correlation": 140707
    }
  },
  {
    "ph": "s", "id": 140707, "pid": 76337, "tid": -914061504, "ts": 1716454223608574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223654256, "dur": 11,
    "args": {
      "External id": 140716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140716,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140716, "pid": 5, "tid": 7, "ts": 1716454223654256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608614, "dur": 10,
    "args": {
      "External id": 140716,
      "cbid": 211, "correlation": 140716
    }
  },
  {
    "ph": "s", "id": 140716, "pid": 76337, "tid": -914061504, "ts": 1716454223608614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608668, "dur": 0,
    "args": {
      "External id": 140726,
      "cbid": 317, "correlation": 140726
    }
  },
  {
    "ph": "f", "id": 140726, "pid": 76337, "tid": -914061504, "ts": 1716454223608668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608669, "dur": 0,
    "args": {
      "External id": 140727,
      "cbid": 203, "correlation": 140727
    }
  },
  {
    "ph": "f", "id": 140727, "pid": 76337, "tid": -914061504, "ts": 1716454223608669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608670, "dur": 0,
    "args": {
      "External id": 140728,
      "cbid": 205, "correlation": 140728
    }
  },
  {
    "ph": "f", "id": 140728, "pid": 76337, "tid": -914061504, "ts": 1716454223608670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654268, "dur": 11,
    "args": {
      "External id": 140732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140732, "pid": 5, "tid": 7, "ts": 1716454223654268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608685, "dur": 11,
    "args": {
      "External id": 140732,
      "cbid": 211, "correlation": 140732
    }
  },
  {
    "ph": "s", "id": 140732, "pid": 76337, "tid": -914061504, "ts": 1716454223608685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654281, "dur": 25,
    "args": {
      "External id": 140734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140734, "pid": 5, "tid": 7, "ts": 1716454223654281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608699, "dur": 5,
    "args": {
      "External id": 140734,
      "cbid": 211, "correlation": 140734
    }
  },
  {
    "ph": "s", "id": 140734, "pid": 76337, "tid": -914061504, "ts": 1716454223608699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223654307, "dur": 3,
    "args": {
      "External id": 140736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140736,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 140736, "pid": 5, "tid": 7, "ts": 1716454223654307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608710, "dur": 5,
    "args": {
      "External id": 140736,
      "cbid": 211, "correlation": 140736
    }
  },
  {
    "ph": "s", "id": 140736, "pid": 76337, "tid": -914061504, "ts": 1716454223608710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608719, "dur": 0,
    "args": {
      "External id": 140737,
      "cbid": 51, "correlation": 140737
    }
  },
  {
    "ph": "s", "id": 140737, "pid": 76337, "tid": -914061504, "ts": 1716454223608719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223654311, "dur": 357,
    "args": {
      "External id": 140738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140738,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140738, "pid": 5, "tid": 7, "ts": 1716454223654311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608720, "dur": 9,
    "args": {
      "External id": 140738,
      "cbid": 211, "correlation": 140738
    }
  },
  {
    "ph": "s", "id": 140738, "pid": 76337, "tid": -914061504, "ts": 1716454223608720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654669, "dur": 22,
    "args": {
      "External id": 140739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140739, "pid": 5, "tid": 7, "ts": 1716454223654669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608732, "dur": 5,
    "args": {
      "External id": 140739,
      "cbid": 211, "correlation": 140739
    }
  },
  {
    "ph": "s", "id": 140739, "pid": 76337, "tid": -914061504, "ts": 1716454223608732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223654692, "dur": 32,
    "args": {
      "External id": 140745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140745, "pid": 5, "tid": 7, "ts": 1716454223654692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608759, "dur": 8,
    "args": {
      "External id": 140745,
      "cbid": 211, "correlation": 140745
    }
  },
  {
    "ph": "s", "id": 140745, "pid": 76337, "tid": -914061504, "ts": 1716454223608759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223654726, "dur": 4,
    "args": {
      "External id": 140753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140753,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 140753, "pid": 5, "tid": 7, "ts": 1716454223654726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608802, "dur": 9,
    "args": {
      "External id": 140753,
      "cbid": 211, "correlation": 140753
    }
  },
  {
    "ph": "s", "id": 140753, "pid": 76337, "tid": -914061504, "ts": 1716454223608802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608867, "dur": 1,
    "args": {
      "External id": 140769,
      "cbid": 251, "correlation": 140769
    }
  },
  {
    "ph": "f", "id": 140769, "pid": 76337, "tid": -914061504, "ts": 1716454223608867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608872, "dur": 0,
    "args": {
      "External id": 140771,
      "cbid": 251, "correlation": 140771
    }
  },
  {
    "ph": "f", "id": 140771, "pid": 76337, "tid": -914061504, "ts": 1716454223608872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223654732, "dur": 13,
    "args": {
      "External id": 140772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140772,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 140772, "pid": 5, "tid": 7, "ts": 1716454223654732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608874, "dur": 12,
    "args": {
      "External id": 140772,
      "cbid": 211, "correlation": 140772
    }
  },
  {
    "ph": "s", "id": 140772, "pid": 76337, "tid": -914061504, "ts": 1716454223608874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223654745, "dur": 5,
    "args": {
      "External id": 140774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140774,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 140774, "pid": 5, "tid": 7, "ts": 1716454223654745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608888, "dur": 5,
    "args": {
      "External id": 140774,
      "cbid": 211, "correlation": 140774
    }
  },
  {
    "ph": "s", "id": 140774, "pid": 76337, "tid": -914061504, "ts": 1716454223608888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223654752, "dur": 30,
    "args": {
      "External id": 140784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140784, "pid": 5, "tid": 7, "ts": 1716454223654752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223608946, "dur": 12,
    "args": {
      "External id": 140784,
      "cbid": 211, "correlation": 140784
    }
  },
  {
    "ph": "s", "id": 140784, "pid": 76337, "tid": -914061504, "ts": 1716454223608946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654783, "dur": 31,
    "args": {
      "External id": 140804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140804,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 140804, "pid": 5, "tid": 7, "ts": 1716454223654783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609021, "dur": 12,
    "args": {
      "External id": 140804,
      "cbid": 211, "correlation": 140804
    }
  },
  {
    "ph": "s", "id": 140804, "pid": 76337, "tid": -914061504, "ts": 1716454223609021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223654815, "dur": 4,
    "args": {
      "External id": 140816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140816,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 140816, "pid": 5, "tid": 7, "ts": 1716454223654815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609043, "dur": 6,
    "args": {
      "External id": 140816,
      "cbid": 211, "correlation": 140816
    }
  },
  {
    "ph": "s", "id": 140816, "pid": 76337, "tid": -914061504, "ts": 1716454223609043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223654820, "dur": 30,
    "args": {
      "External id": 140819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140819, "pid": 5, "tid": 7, "ts": 1716454223654820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609061, "dur": 7,
    "args": {
      "External id": 140819,
      "cbid": 211, "correlation": 140819
    }
  },
  {
    "ph": "s", "id": 140819, "pid": 76337, "tid": -914061504, "ts": 1716454223609061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223654852, "dur": 21,
    "args": {
      "External id": 140828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140828,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140828, "pid": 5, "tid": 7, "ts": 1716454223654852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609103, "dur": 9,
    "args": {
      "External id": 140828,
      "cbid": 211, "correlation": 140828
    }
  },
  {
    "ph": "s", "id": 140828, "pid": 76337, "tid": -914061504, "ts": 1716454223609103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609166, "dur": 0,
    "args": {
      "External id": 140838,
      "cbid": 317, "correlation": 140838
    }
  },
  {
    "ph": "f", "id": 140838, "pid": 76337, "tid": -914061504, "ts": 1716454223609166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609167, "dur": 0,
    "args": {
      "External id": 140839,
      "cbid": 203, "correlation": 140839
    }
  },
  {
    "ph": "f", "id": 140839, "pid": 76337, "tid": -914061504, "ts": 1716454223609167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609167, "dur": 0,
    "args": {
      "External id": 140840,
      "cbid": 205, "correlation": 140840
    }
  },
  {
    "ph": "f", "id": 140840, "pid": 76337, "tid": -914061504, "ts": 1716454223609167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654874, "dur": 22,
    "args": {
      "External id": 140844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140844,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140844, "pid": 5, "tid": 7, "ts": 1716454223654874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609185, "dur": 12,
    "args": {
      "External id": 140844,
      "cbid": 211, "correlation": 140844
    }
  },
  {
    "ph": "s", "id": 140844, "pid": 76337, "tid": -914061504, "ts": 1716454223609185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223654897, "dur": 44,
    "args": {
      "External id": 140846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140846, "pid": 5, "tid": 7, "ts": 1716454223654897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609200, "dur": 5,
    "args": {
      "External id": 140846,
      "cbid": 211, "correlation": 140846
    }
  },
  {
    "ph": "s", "id": 140846, "pid": 76337, "tid": -914061504, "ts": 1716454223609200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223654942, "dur": 659,
    "args": {
      "External id": 140848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140848,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140848, "pid": 5, "tid": 7, "ts": 1716454223654942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609214, "dur": 10,
    "args": {
      "External id": 140848,
      "cbid": 211, "correlation": 140848
    }
  },
  {
    "ph": "s", "id": 140848, "pid": 76337, "tid": -914061504, "ts": 1716454223609214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223655603, "dur": 22,
    "args": {
      "External id": 140850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140850,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140850, "pid": 5, "tid": 7, "ts": 1716454223655603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609228, "dur": 5,
    "args": {
      "External id": 140850,
      "cbid": 211, "correlation": 140850
    }
  },
  {
    "ph": "s", "id": 140850, "pid": 76337, "tid": -914061504, "ts": 1716454223609228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223655626, "dur": 33,
    "args": {
      "External id": 140856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140856, "pid": 5, "tid": 7, "ts": 1716454223655626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609256, "dur": 8,
    "args": {
      "External id": 140856,
      "cbid": 211, "correlation": 140856
    }
  },
  {
    "ph": "s", "id": 140856, "pid": 76337, "tid": -914061504, "ts": 1716454223609256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609315, "dur": 0,
    "args": {
      "External id": 140866,
      "cbid": 317, "correlation": 140866
    }
  },
  {
    "ph": "f", "id": 140866, "pid": 76337, "tid": -914061504, "ts": 1716454223609315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609316, "dur": 0,
    "args": {
      "External id": 140867,
      "cbid": 203, "correlation": 140867
    }
  },
  {
    "ph": "f", "id": 140867, "pid": 76337, "tid": -914061504, "ts": 1716454223609316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609316, "dur": 0,
    "args": {
      "External id": 140868,
      "cbid": 205, "correlation": 140868
    }
  },
  {
    "ph": "f", "id": 140868, "pid": 76337, "tid": -914061504, "ts": 1716454223609316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609339, "dur": 1,
    "args": {
      "External id": 140872,
      "cbid": 251, "correlation": 140872
    }
  },
  {
    "ph": "f", "id": 140872, "pid": 76337, "tid": -914061504, "ts": 1716454223609339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609341, "dur": 0,
    "args": {
      "External id": 140873,
      "cbid": 251, "correlation": 140873
    }
  },
  {
    "ph": "f", "id": 140873, "pid": 76337, "tid": -914061504, "ts": 1716454223609341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609342, "dur": 0,
    "args": {
      "External id": 140874,
      "cbid": 251, "correlation": 140874
    }
  },
  {
    "ph": "f", "id": 140874, "pid": 76337, "tid": -914061504, "ts": 1716454223609342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609343, "dur": 0,
    "args": {
      "External id": 140875,
      "cbid": 251, "correlation": 140875
    }
  },
  {
    "ph": "f", "id": 140875, "pid": 76337, "tid": -914061504, "ts": 1716454223609343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609343, "dur": 0,
    "args": {
      "External id": 140876,
      "cbid": 251, "correlation": 140876
    }
  },
  {
    "ph": "f", "id": 140876, "pid": 76337, "tid": -914061504, "ts": 1716454223609343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609344, "dur": 0,
    "args": {
      "External id": 140877,
      "cbid": 251, "correlation": 140877
    }
  },
  {
    "ph": "f", "id": 140877, "pid": 76337, "tid": -914061504, "ts": 1716454223609344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609345, "dur": 0,
    "args": {
      "External id": 140878,
      "cbid": 251, "correlation": 140878
    }
  },
  {
    "ph": "f", "id": 140878, "pid": 76337, "tid": -914061504, "ts": 1716454223609345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609345, "dur": 0,
    "args": {
      "External id": 140879,
      "cbid": 251, "correlation": 140879
    }
  },
  {
    "ph": "f", "id": 140879, "pid": 76337, "tid": -914061504, "ts": 1716454223609345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609347, "dur": 0,
    "args": {
      "External id": 140880,
      "cbid": 251, "correlation": 140880
    }
  },
  {
    "ph": "f", "id": 140880, "pid": 76337, "tid": -914061504, "ts": 1716454223609347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223655661, "dur": 52,
    "args": {
      "External id": 140881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140881,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 140881, "pid": 5, "tid": 7, "ts": 1716454223655661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609349, "dur": 12,
    "args": {
      "External id": 140881,
      "cbid": 211, "correlation": 140881
    }
  },
  {
    "ph": "s", "id": 140881, "pid": 76337, "tid": -914061504, "ts": 1716454223609349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223655714, "dur": 32,
    "args": {
      "External id": 140887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140887, "pid": 5, "tid": 7, "ts": 1716454223655714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609381, "dur": 8,
    "args": {
      "External id": 140887,
      "cbid": 211, "correlation": 140887
    }
  },
  {
    "ph": "s", "id": 140887, "pid": 76337, "tid": -914061504, "ts": 1716454223609381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223655748, "dur": 27,
    "args": {
      "External id": 140895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140895,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140895, "pid": 5, "tid": 7, "ts": 1716454223655748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609410, "dur": 8,
    "args": {
      "External id": 140895,
      "cbid": 211, "correlation": 140895
    }
  },
  {
    "ph": "s", "id": 140895, "pid": 76337, "tid": -914061504, "ts": 1716454223609410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223655776, "dur": 19,
    "args": {
      "External id": 140903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140903,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140903, "pid": 5, "tid": 7, "ts": 1716454223655776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609439, "dur": 8,
    "args": {
      "External id": 140903,
      "cbid": 211, "correlation": 140903
    }
  },
  {
    "ph": "s", "id": 140903, "pid": 76337, "tid": -914061504, "ts": 1716454223609439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223655796, "dur": 30,
    "args": {
      "External id": 140923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140923,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 140923, "pid": 5, "tid": 7, "ts": 1716454223655796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609520, "dur": 13,
    "args": {
      "External id": 140923,
      "cbid": 211, "correlation": 140923
    }
  },
  {
    "ph": "s", "id": 140923, "pid": 76337, "tid": -914061504, "ts": 1716454223609520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223655828, "dur": 4,
    "args": {
      "External id": 140935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140935,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 140935, "pid": 5, "tid": 7, "ts": 1716454223655828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609542, "dur": 6,
    "args": {
      "External id": 140935,
      "cbid": 211, "correlation": 140935
    }
  },
  {
    "ph": "s", "id": 140935, "pid": 76337, "tid": -914061504, "ts": 1716454223609542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223655833, "dur": 30,
    "args": {
      "External id": 140938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140938,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140938, "pid": 5, "tid": 7, "ts": 1716454223655833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609560, "dur": 6,
    "args": {
      "External id": 140938,
      "cbid": 211, "correlation": 140938
    }
  },
  {
    "ph": "s", "id": 140938, "pid": 76337, "tid": -914061504, "ts": 1716454223609560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609618, "dur": 0,
    "args": {
      "External id": 140949,
      "cbid": 317, "correlation": 140949
    }
  },
  {
    "ph": "f", "id": 140949, "pid": 76337, "tid": -914061504, "ts": 1716454223609618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609618, "dur": 0,
    "args": {
      "External id": 140950,
      "cbid": 203, "correlation": 140950
    }
  },
  {
    "ph": "f", "id": 140950, "pid": 76337, "tid": -914061504, "ts": 1716454223609618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609619, "dur": 0,
    "args": {
      "External id": 140951,
      "cbid": 205, "correlation": 140951
    }
  },
  {
    "ph": "f", "id": 140951, "pid": 76337, "tid": -914061504, "ts": 1716454223609619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223655864, "dur": 22,
    "args": {
      "External id": 140955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140955,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140955, "pid": 5, "tid": 7, "ts": 1716454223655864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609635, "dur": 11,
    "args": {
      "External id": 140955,
      "cbid": 211, "correlation": 140955
    }
  },
  {
    "ph": "s", "id": 140955, "pid": 76337, "tid": -914061504, "ts": 1716454223609635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223655887, "dur": 121,
    "args": {
      "External id": 140957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140957,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 140957, "pid": 5, "tid": 7, "ts": 1716454223655887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609655, "dur": 8,
    "args": {
      "External id": 140957,
      "cbid": 211, "correlation": 140957
    }
  },
  {
    "ph": "s", "id": 140957, "pid": 76337, "tid": -914061504, "ts": 1716454223609655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223656010, "dur": 22,
    "args": {
      "External id": 140959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140959,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140959, "pid": 5, "tid": 7, "ts": 1716454223656010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609666, "dur": 5,
    "args": {
      "External id": 140959,
      "cbid": 211, "correlation": 140959
    }
  },
  {
    "ph": "s", "id": 140959, "pid": 76337, "tid": -914061504, "ts": 1716454223609666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223656033, "dur": 32,
    "args": {
      "External id": 140965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140965, "pid": 5, "tid": 7, "ts": 1716454223656033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609694, "dur": 8,
    "args": {
      "External id": 140965,
      "cbid": 211, "correlation": 140965
    }
  },
  {
    "ph": "s", "id": 140965, "pid": 76337, "tid": -914061504, "ts": 1716454223609694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223656066, "dur": 191,
    "args": {
      "External id": 140974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140974,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140974, "pid": 5, "tid": 7, "ts": 1716454223656066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609776, "dur": 14,
    "args": {
      "External id": 140974,
      "cbid": 211, "correlation": 140974
    }
  },
  {
    "ph": "s", "id": 140974, "pid": 76337, "tid": -914061504, "ts": 1716454223609776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223656259, "dur": 65,
    "args": {
      "External id": 140996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 140996,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 140996, "pid": 5, "tid": 7, "ts": 1716454223656259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609834, "dur": 10,
    "args": {
      "External id": 140996,
      "cbid": 211, "correlation": 140996
    }
  },
  {
    "ph": "s", "id": 140996, "pid": 76337, "tid": -914061504, "ts": 1716454223609834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609925, "dur": 1,
    "args": {
      "External id": 141007,
      "cbid": 251, "correlation": 141007
    }
  },
  {
    "ph": "f", "id": 141007, "pid": 76337, "tid": -914061504, "ts": 1716454223609925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223656325, "dur": 156,
    "args": {
      "External id": 141008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141008,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141008, "pid": 5, "tid": 7, "ts": 1716454223656325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223609931, "dur": 13,
    "args": {
      "External id": 141008,
      "cbid": 211, "correlation": 141008
    }
  },
  {
    "ph": "s", "id": 141008, "pid": 76337, "tid": -914061504, "ts": 1716454223609931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610009, "dur": 1,
    "args": {
      "External id": 141019,
      "cbid": 251, "correlation": 141019
    }
  },
  {
    "ph": "f", "id": 141019, "pid": 76337, "tid": -914061504, "ts": 1716454223610009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223656483, "dur": 147,
    "args": {
      "External id": 141020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141020,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141020, "pid": 5, "tid": 7, "ts": 1716454223656483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610013, "dur": 12,
    "args": {
      "External id": 141020,
      "cbid": 211, "correlation": 141020
    }
  },
  {
    "ph": "s", "id": 141020, "pid": 76337, "tid": -914061504, "ts": 1716454223610013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610079, "dur": 1,
    "args": {
      "External id": 141031,
      "cbid": 251, "correlation": 141031
    }
  },
  {
    "ph": "f", "id": 141031, "pid": 76337, "tid": -914061504, "ts": 1716454223610079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223656631, "dur": 144,
    "args": {
      "External id": 141032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141032,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141032, "pid": 5, "tid": 7, "ts": 1716454223656631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610083, "dur": 11,
    "args": {
      "External id": 141032,
      "cbid": 211, "correlation": 141032
    }
  },
  {
    "ph": "s", "id": 141032, "pid": 76337, "tid": -914061504, "ts": 1716454223610083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223656777, "dur": 1954,
    "args": {
      "External id": 141053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141053,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 141053, "pid": 5, "tid": 7, "ts": 1716454223656777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610167, "dur": 14,
    "args": {
      "External id": 141053,
      "cbid": 211, "correlation": 141053
    }
  },
  {
    "ph": "s", "id": 141053, "pid": 76337, "tid": -914061504, "ts": 1716454223610167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610267, "dur": 1,
    "args": {
      "External id": 141071,
      "cbid": 251, "correlation": 141071
    }
  },
  {
    "ph": "f", "id": 141071, "pid": 76337, "tid": -914061504, "ts": 1716454223610267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223658732, "dur": 147,
    "args": {
      "External id": 141073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141073,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 141073, "pid": 5, "tid": 7, "ts": 1716454223658732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610273, "dur": 14,
    "args": {
      "External id": 141073,
      "cbid": 211, "correlation": 141073
    }
  },
  {
    "ph": "s", "id": 141073, "pid": 76337, "tid": -914061504, "ts": 1716454223610273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223658881, "dur": 36,
    "args": {
      "External id": 141081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141081,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141081, "pid": 5, "tid": 7, "ts": 1716454223658881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610345, "dur": 12,
    "args": {
      "External id": 141081,
      "cbid": 211, "correlation": 141081
    }
  },
  {
    "ph": "s", "id": 141081, "pid": 76337, "tid": -914061504, "ts": 1716454223610345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223658918, "dur": 51,
    "args": {
      "External id": 141089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141089,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141089, "pid": 5, "tid": 7, "ts": 1716454223658918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610385, "dur": 8,
    "args": {
      "External id": 141089,
      "cbid": 211, "correlation": 141089
    }
  },
  {
    "ph": "s", "id": 141089, "pid": 76337, "tid": -914061504, "ts": 1716454223610385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223658970, "dur": 30,
    "args": {
      "External id": 141100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141100, "pid": 5, "tid": 7, "ts": 1716454223658970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610457, "dur": 13,
    "args": {
      "External id": 141100,
      "cbid": 211, "correlation": 141100
    }
  },
  {
    "ph": "s", "id": 141100, "pid": 76337, "tid": -914061504, "ts": 1716454223610457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223659001, "dur": 34,
    "args": {
      "External id": 141122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141122,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141122, "pid": 5, "tid": 7, "ts": 1716454223659001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610489, "dur": 8,
    "args": {
      "External id": 141122,
      "cbid": 211, "correlation": 141122
    }
  },
  {
    "ph": "s", "id": 141122, "pid": 76337, "tid": -914061504, "ts": 1716454223610489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610573, "dur": 1,
    "args": {
      "External id": 141133,
      "cbid": 251, "correlation": 141133
    }
  },
  {
    "ph": "f", "id": 141133, "pid": 76337, "tid": -914061504, "ts": 1716454223610573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223659037, "dur": 91,
    "args": {
      "External id": 141134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141134,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141134, "pid": 5, "tid": 7, "ts": 1716454223659037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610579, "dur": 14,
    "args": {
      "External id": 141134,
      "cbid": 211, "correlation": 141134
    }
  },
  {
    "ph": "s", "id": 141134, "pid": 76337, "tid": -914061504, "ts": 1716454223610579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610648, "dur": 1,
    "args": {
      "External id": 141145,
      "cbid": 251, "correlation": 141145
    }
  },
  {
    "ph": "f", "id": 141145, "pid": 76337, "tid": -914061504, "ts": 1716454223610648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610652, "dur": 0,
    "args": {
      "External id": 141146,
      "cbid": 251, "correlation": 141146
    }
  },
  {
    "ph": "f", "id": 141146, "pid": 76337, "tid": -914061504, "ts": 1716454223610652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223659129, "dur": 11,
    "args": {
      "External id": 141147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141147,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 141147, "pid": 5, "tid": 7, "ts": 1716454223659129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610654, "dur": 12,
    "args": {
      "External id": 141147,
      "cbid": 211, "correlation": 141147
    }
  },
  {
    "ph": "s", "id": 141147, "pid": 76337, "tid": -914061504, "ts": 1716454223610654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223659141, "dur": 5,
    "args": {
      "External id": 141149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141149,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 141149, "pid": 5, "tid": 7, "ts": 1716454223659141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610669, "dur": 7,
    "args": {
      "External id": 141149,
      "cbid": 211, "correlation": 141149
    }
  },
  {
    "ph": "s", "id": 141149, "pid": 76337, "tid": -914061504, "ts": 1716454223610669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610727, "dur": 1,
    "args": {
      "External id": 141160,
      "cbid": 251, "correlation": 141160
    }
  },
  {
    "ph": "f", "id": 141160, "pid": 76337, "tid": -914061504, "ts": 1716454223610727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610731, "dur": 0,
    "args": {
      "External id": 141161,
      "cbid": 251, "correlation": 141161
    }
  },
  {
    "ph": "f", "id": 141161, "pid": 76337, "tid": -914061504, "ts": 1716454223610731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223659148, "dur": 7,
    "args": {
      "External id": 141162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141162,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 141162, "pid": 5, "tid": 7, "ts": 1716454223659148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610732, "dur": 12,
    "args": {
      "External id": 141162,
      "cbid": 211, "correlation": 141162
    }
  },
  {
    "ph": "s", "id": 141162, "pid": 76337, "tid": -914061504, "ts": 1716454223610732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223659156, "dur": 3,
    "args": {
      "External id": 141164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141164,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 141164, "pid": 5, "tid": 7, "ts": 1716454223659156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610746, "dur": 5,
    "args": {
      "External id": 141164,
      "cbid": 211, "correlation": 141164
    }
  },
  {
    "ph": "s", "id": 141164, "pid": 76337, "tid": -914061504, "ts": 1716454223610746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223659160, "dur": 92,
    "args": {
      "External id": 141185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141185,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 141185, "pid": 5, "tid": 7, "ts": 1716454223659160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610819, "dur": 12,
    "args": {
      "External id": 141185,
      "cbid": 211, "correlation": 141185
    }
  },
  {
    "ph": "s", "id": 141185, "pid": 76337, "tid": -914061504, "ts": 1716454223610819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610914, "dur": 1,
    "args": {
      "External id": 141203,
      "cbid": 251, "correlation": 141203
    }
  },
  {
    "ph": "f", "id": 141203, "pid": 76337, "tid": -914061504, "ts": 1716454223610914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223659253, "dur": 99,
    "args": {
      "External id": 141205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141205,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141205, "pid": 5, "tid": 7, "ts": 1716454223659253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610919, "dur": 13,
    "args": {
      "External id": 141205,
      "cbid": 211, "correlation": 141205
    }
  },
  {
    "ph": "s", "id": 141205, "pid": 76337, "tid": -914061504, "ts": 1716454223610919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223659353, "dur": 19,
    "args": {
      "External id": 141213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141213,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141213, "pid": 5, "tid": 7, "ts": 1716454223659353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223610997, "dur": 12,
    "args": {
      "External id": 141213,
      "cbid": 211, "correlation": 141213
    }
  },
  {
    "ph": "s", "id": 141213, "pid": 76337, "tid": -914061504, "ts": 1716454223610997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223659373, "dur": 37,
    "args": {
      "External id": 141221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141221,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141221, "pid": 5, "tid": 7, "ts": 1716454223659373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611038, "dur": 10,
    "args": {
      "External id": 141221,
      "cbid": 211, "correlation": 141221
    }
  },
  {
    "ph": "s", "id": 141221, "pid": 76337, "tid": -914061504, "ts": 1716454223611038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223659412, "dur": 35,
    "args": {
      "External id": 141243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141243,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141243, "pid": 5, "tid": 7, "ts": 1716454223659412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611090, "dur": 10,
    "args": {
      "External id": 141243,
      "cbid": 211, "correlation": 141243
    }
  },
  {
    "ph": "s", "id": 141243, "pid": 76337, "tid": -914061504, "ts": 1716454223611090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611178, "dur": 1,
    "args": {
      "External id": 141259,
      "cbid": 251, "correlation": 141259
    }
  },
  {
    "ph": "f", "id": 141259, "pid": 76337, "tid": -914061504, "ts": 1716454223611178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611184, "dur": 0,
    "args": {
      "External id": 141261,
      "cbid": 251, "correlation": 141261
    }
  },
  {
    "ph": "f", "id": 141261, "pid": 76337, "tid": -914061504, "ts": 1716454223611184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223659448, "dur": 543,
    "args": {
      "External id": 141262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141262,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 141262, "pid": 5, "tid": 7, "ts": 1716454223659448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611187, "dur": 14,
    "args": {
      "External id": 141262,
      "cbid": 211, "correlation": 141262
    }
  },
  {
    "ph": "s", "id": 141262, "pid": 76337, "tid": -914061504, "ts": 1716454223611187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223659993, "dur": 125,
    "args": {
      "External id": 141270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141270, "pid": 5, "tid": 7, "ts": 1716454223659993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611254, "dur": 12,
    "args": {
      "External id": 141270,
      "cbid": 211, "correlation": 141270
    }
  },
  {
    "ph": "s", "id": 141270, "pid": 76337, "tid": -914061504, "ts": 1716454223611254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223660119, "dur": 127,
    "args": {
      "External id": 141278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141278, "pid": 5, "tid": 7, "ts": 1716454223660119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611284, "dur": 9,
    "args": {
      "External id": 141278,
      "cbid": 211, "correlation": 141278
    }
  },
  {
    "ph": "s", "id": 141278, "pid": 76337, "tid": -914061504, "ts": 1716454223611284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611364, "dur": 1,
    "args": {
      "External id": 141294,
      "cbid": 251, "correlation": 141294
    }
  },
  {
    "ph": "f", "id": 141294, "pid": 76337, "tid": -914061504, "ts": 1716454223611364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223660248, "dur": 311,
    "args": {
      "External id": 141296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141296,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141296, "pid": 5, "tid": 7, "ts": 1716454223660248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611370, "dur": 13,
    "args": {
      "External id": 141296,
      "cbid": 211, "correlation": 141296
    }
  },
  {
    "ph": "s", "id": 141296, "pid": 76337, "tid": -914061504, "ts": 1716454223611370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223660560, "dur": 27,
    "args": {
      "External id": 141304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141304,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141304, "pid": 5, "tid": 7, "ts": 1716454223660560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611412, "dur": 10,
    "args": {
      "External id": 141304,
      "cbid": 211, "correlation": 141304
    }
  },
  {
    "ph": "s", "id": 141304, "pid": 76337, "tid": -914061504, "ts": 1716454223611412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223660588, "dur": 82,
    "args": {
      "External id": 141315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141315, "pid": 5, "tid": 7, "ts": 1716454223660588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611480, "dur": 13,
    "args": {
      "External id": 141315,
      "cbid": 211, "correlation": 141315
    }
  },
  {
    "ph": "s", "id": 141315, "pid": 76337, "tid": -914061504, "ts": 1716454223611480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611543, "dur": 0,
    "args": {
      "External id": 141327,
      "cbid": 317, "correlation": 141327
    }
  },
  {
    "ph": "f", "id": 141327, "pid": 76337, "tid": -914061504, "ts": 1716454223611543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611544, "dur": 0,
    "args": {
      "External id": 141328,
      "cbid": 203, "correlation": 141328
    }
  },
  {
    "ph": "f", "id": 141328, "pid": 76337, "tid": -914061504, "ts": 1716454223611544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611545, "dur": 0,
    "args": {
      "External id": 141329,
      "cbid": 205, "correlation": 141329
    }
  },
  {
    "ph": "f", "id": 141329, "pid": 76337, "tid": -914061504, "ts": 1716454223611545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223660671, "dur": 22,
    "args": {
      "External id": 141333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141333,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141333, "pid": 5, "tid": 7, "ts": 1716454223660671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611560, "dur": 12,
    "args": {
      "External id": 141333,
      "cbid": 211, "correlation": 141333
    }
  },
  {
    "ph": "s", "id": 141333, "pid": 76337, "tid": -914061504, "ts": 1716454223611560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223660695, "dur": 121,
    "args": {
      "External id": 141335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141335,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141335, "pid": 5, "tid": 7, "ts": 1716454223660695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611579, "dur": 6,
    "args": {
      "External id": 141335,
      "cbid": 211, "correlation": 141335
    }
  },
  {
    "ph": "s", "id": 141335, "pid": 76337, "tid": -914061504, "ts": 1716454223611579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223660818, "dur": 23,
    "args": {
      "External id": 141337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141337, "pid": 5, "tid": 7, "ts": 1716454223660818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611590, "dur": 5,
    "args": {
      "External id": 141337,
      "cbid": 211, "correlation": 141337
    }
  },
  {
    "ph": "s", "id": 141337, "pid": 76337, "tid": -914061504, "ts": 1716454223611590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223660842, "dur": 32,
    "args": {
      "External id": 141343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141343,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141343, "pid": 5, "tid": 7, "ts": 1716454223660842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611617, "dur": 8,
    "args": {
      "External id": 141343,
      "cbid": 211, "correlation": 141343
    }
  },
  {
    "ph": "s", "id": 141343, "pid": 76337, "tid": -914061504, "ts": 1716454223611617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223660876, "dur": 27,
    "args": {
      "External id": 141351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141351,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141351, "pid": 5, "tid": 7, "ts": 1716454223660876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611650, "dur": 8,
    "args": {
      "External id": 141351,
      "cbid": 211, "correlation": 141351
    }
  },
  {
    "ph": "s", "id": 141351, "pid": 76337, "tid": -914061504, "ts": 1716454223611650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223660904, "dur": 30,
    "args": {
      "External id": 141371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141371,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 141371, "pid": 5, "tid": 7, "ts": 1716454223660904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611722, "dur": 11,
    "args": {
      "External id": 141371,
      "cbid": 211, "correlation": 141371
    }
  },
  {
    "ph": "s", "id": 141371, "pid": 76337, "tid": -914061504, "ts": 1716454223611722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223660935, "dur": 4,
    "args": {
      "External id": 141383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141383,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 141383, "pid": 5, "tid": 7, "ts": 1716454223660935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611743, "dur": 6,
    "args": {
      "External id": 141383,
      "cbid": 211, "correlation": 141383
    }
  },
  {
    "ph": "s", "id": 141383, "pid": 76337, "tid": -914061504, "ts": 1716454223611743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223660941, "dur": 31,
    "args": {
      "External id": 141386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141386, "pid": 5, "tid": 7, "ts": 1716454223660941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611761, "dur": 6,
    "args": {
      "External id": 141386,
      "cbid": 211, "correlation": 141386
    }
  },
  {
    "ph": "s", "id": 141386, "pid": 76337, "tid": -914061504, "ts": 1716454223611761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223660973, "dur": 21,
    "args": {
      "External id": 141395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141395,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141395, "pid": 5, "tid": 7, "ts": 1716454223660973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611800, "dur": 9,
    "args": {
      "External id": 141395,
      "cbid": 211, "correlation": 141395
    }
  },
  {
    "ph": "s", "id": 141395, "pid": 76337, "tid": -914061504, "ts": 1716454223611800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611851, "dur": 0,
    "args": {
      "External id": 141405,
      "cbid": 317, "correlation": 141405
    }
  },
  {
    "ph": "f", "id": 141405, "pid": 76337, "tid": -914061504, "ts": 1716454223611851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611852, "dur": 0,
    "args": {
      "External id": 141406,
      "cbid": 203, "correlation": 141406
    }
  },
  {
    "ph": "f", "id": 141406, "pid": 76337, "tid": -914061504, "ts": 1716454223611852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611852, "dur": 0,
    "args": {
      "External id": 141407,
      "cbid": 205, "correlation": 141407
    }
  },
  {
    "ph": "f", "id": 141407, "pid": 76337, "tid": -914061504, "ts": 1716454223611852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223660995, "dur": 23,
    "args": {
      "External id": 141411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141411,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141411, "pid": 5, "tid": 7, "ts": 1716454223660995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611865, "dur": 11,
    "args": {
      "External id": 141411,
      "cbid": 211, "correlation": 141411
    }
  },
  {
    "ph": "s", "id": 141411, "pid": 76337, "tid": -914061504, "ts": 1716454223611865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223661020, "dur": 44,
    "args": {
      "External id": 141413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141413,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141413, "pid": 5, "tid": 7, "ts": 1716454223661020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611878, "dur": 5,
    "args": {
      "External id": 141413,
      "cbid": 211, "correlation": 141413
    }
  },
  {
    "ph": "s", "id": 141413, "pid": 76337, "tid": -914061504, "ts": 1716454223611878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223661065, "dur": 655,
    "args": {
      "External id": 141415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141415,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141415, "pid": 5, "tid": 7, "ts": 1716454223661065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611889, "dur": 6,
    "args": {
      "External id": 141415,
      "cbid": 211, "correlation": 141415
    }
  },
  {
    "ph": "s", "id": 141415, "pid": 76337, "tid": -914061504, "ts": 1716454223611889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223661722, "dur": 20,
    "args": {
      "External id": 141417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141417,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141417, "pid": 5, "tid": 7, "ts": 1716454223661722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611900, "dur": 5,
    "args": {
      "External id": 141417,
      "cbid": 211, "correlation": 141417
    }
  },
  {
    "ph": "s", "id": 141417, "pid": 76337, "tid": -914061504, "ts": 1716454223611900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223661743, "dur": 33,
    "args": {
      "External id": 141423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141423, "pid": 5, "tid": 7, "ts": 1716454223661743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611926, "dur": 8,
    "args": {
      "External id": 141423,
      "cbid": 211, "correlation": 141423
    }
  },
  {
    "ph": "s", "id": 141423, "pid": 76337, "tid": -914061504, "ts": 1716454223611926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223661777, "dur": 4,
    "args": {
      "External id": 141431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141431,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 141431, "pid": 5, "tid": 7, "ts": 1716454223661777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223611970, "dur": 23,
    "args": {
      "External id": 141431,
      "cbid": 211, "correlation": 141431
    }
  },
  {
    "ph": "s", "id": 141431, "pid": 76337, "tid": -914061504, "ts": 1716454223611970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612049, "dur": 1,
    "args": {
      "External id": 141447,
      "cbid": 251, "correlation": 141447
    }
  },
  {
    "ph": "f", "id": 141447, "pid": 76337, "tid": -914061504, "ts": 1716454223612049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612054, "dur": 0,
    "args": {
      "External id": 141449,
      "cbid": 251, "correlation": 141449
    }
  },
  {
    "ph": "f", "id": 141449, "pid": 76337, "tid": -914061504, "ts": 1716454223612054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223661783, "dur": 12,
    "args": {
      "External id": 141450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141450,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 141450, "pid": 5, "tid": 7, "ts": 1716454223661783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612056, "dur": 11,
    "args": {
      "External id": 141450,
      "cbid": 211, "correlation": 141450
    }
  },
  {
    "ph": "s", "id": 141450, "pid": 76337, "tid": -914061504, "ts": 1716454223612056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223661796, "dur": 5,
    "args": {
      "External id": 141452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141452,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 141452, "pid": 5, "tid": 7, "ts": 1716454223661796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612069, "dur": 5,
    "args": {
      "External id": 141452,
      "cbid": 211, "correlation": 141452
    }
  },
  {
    "ph": "s", "id": 141452, "pid": 76337, "tid": -914061504, "ts": 1716454223612069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223661803, "dur": 29,
    "args": {
      "External id": 141462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141462, "pid": 5, "tid": 7, "ts": 1716454223661803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612127, "dur": 12,
    "args": {
      "External id": 141462,
      "cbid": 211, "correlation": 141462
    }
  },
  {
    "ph": "s", "id": 141462, "pid": 76337, "tid": -914061504, "ts": 1716454223612127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223661833, "dur": 31,
    "args": {
      "External id": 141482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141482,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 141482, "pid": 5, "tid": 7, "ts": 1716454223661833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612193, "dur": 11,
    "args": {
      "External id": 141482,
      "cbid": 211, "correlation": 141482
    }
  },
  {
    "ph": "s", "id": 141482, "pid": 76337, "tid": -914061504, "ts": 1716454223612193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223661865, "dur": 4,
    "args": {
      "External id": 141494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141494,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 141494, "pid": 5, "tid": 7, "ts": 1716454223661865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612213, "dur": 6,
    "args": {
      "External id": 141494,
      "cbid": 211, "correlation": 141494
    }
  },
  {
    "ph": "s", "id": 141494, "pid": 76337, "tid": -914061504, "ts": 1716454223612213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223661871, "dur": 29,
    "args": {
      "External id": 141497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141497, "pid": 5, "tid": 7, "ts": 1716454223661871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612232, "dur": 6,
    "args": {
      "External id": 141497,
      "cbid": 211, "correlation": 141497
    }
  },
  {
    "ph": "s", "id": 141497, "pid": 76337, "tid": -914061504, "ts": 1716454223612232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223661901, "dur": 20,
    "args": {
      "External id": 141506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141506,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141506, "pid": 5, "tid": 7, "ts": 1716454223661901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612273, "dur": 10,
    "args": {
      "External id": 141506,
      "cbid": 211, "correlation": 141506
    }
  },
  {
    "ph": "s", "id": 141506, "pid": 76337, "tid": -914061504, "ts": 1716454223612273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612336, "dur": 0,
    "args": {
      "External id": 141516,
      "cbid": 317, "correlation": 141516
    }
  },
  {
    "ph": "f", "id": 141516, "pid": 76337, "tid": -914061504, "ts": 1716454223612336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612337, "dur": 0,
    "args": {
      "External id": 141517,
      "cbid": 203, "correlation": 141517
    }
  },
  {
    "ph": "f", "id": 141517, "pid": 76337, "tid": -914061504, "ts": 1716454223612337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612337, "dur": 0,
    "args": {
      "External id": 141518,
      "cbid": 205, "correlation": 141518
    }
  },
  {
    "ph": "f", "id": 141518, "pid": 76337, "tid": -914061504, "ts": 1716454223612337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223661923, "dur": 23,
    "args": {
      "External id": 141522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141522,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141522, "pid": 5, "tid": 7, "ts": 1716454223661923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612351, "dur": 12,
    "args": {
      "External id": 141522,
      "cbid": 211, "correlation": 141522
    }
  },
  {
    "ph": "s", "id": 141522, "pid": 76337, "tid": -914061504, "ts": 1716454223612351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223661947, "dur": 44,
    "args": {
      "External id": 141524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141524, "pid": 5, "tid": 7, "ts": 1716454223661947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612366, "dur": 5,
    "args": {
      "External id": 141524,
      "cbid": 211, "correlation": 141524
    }
  },
  {
    "ph": "s", "id": 141524, "pid": 76337, "tid": -914061504, "ts": 1716454223612366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223661993, "dur": 645,
    "args": {
      "External id": 141526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141526,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141526, "pid": 5, "tid": 7, "ts": 1716454223661993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612377, "dur": 6,
    "args": {
      "External id": 141526,
      "cbid": 211, "correlation": 141526
    }
  },
  {
    "ph": "s", "id": 141526, "pid": 76337, "tid": -914061504, "ts": 1716454223612377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223662639, "dur": 22,
    "args": {
      "External id": 141528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141528,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141528, "pid": 5, "tid": 7, "ts": 1716454223662639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612386, "dur": 5,
    "args": {
      "External id": 141528,
      "cbid": 211, "correlation": 141528
    }
  },
  {
    "ph": "s", "id": 141528, "pid": 76337, "tid": -914061504, "ts": 1716454223612386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223662663, "dur": 33,
    "args": {
      "External id": 141534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141534, "pid": 5, "tid": 7, "ts": 1716454223662663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612414, "dur": 8,
    "args": {
      "External id": 141534,
      "cbid": 211, "correlation": 141534
    }
  },
  {
    "ph": "s", "id": 141534, "pid": 76337, "tid": -914061504, "ts": 1716454223612414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223662697, "dur": 27,
    "args": {
      "External id": 141542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141542,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141542, "pid": 5, "tid": 7, "ts": 1716454223662697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612447, "dur": 9,
    "args": {
      "External id": 141542,
      "cbid": 211, "correlation": 141542
    }
  },
  {
    "ph": "s", "id": 141542, "pid": 76337, "tid": -914061504, "ts": 1716454223612447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223662725, "dur": 20,
    "args": {
      "External id": 141550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141550,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141550, "pid": 5, "tid": 7, "ts": 1716454223662725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612476, "dur": 8,
    "args": {
      "External id": 141550,
      "cbid": 211, "correlation": 141550
    }
  },
  {
    "ph": "s", "id": 141550, "pid": 76337, "tid": -914061504, "ts": 1716454223612476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223662746, "dur": 30,
    "args": {
      "External id": 141570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141570,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 141570, "pid": 5, "tid": 7, "ts": 1716454223662746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612557, "dur": 13,
    "args": {
      "External id": 141570,
      "cbid": 211, "correlation": 141570
    }
  },
  {
    "ph": "s", "id": 141570, "pid": 76337, "tid": -914061504, "ts": 1716454223612557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223662777, "dur": 4,
    "args": {
      "External id": 141582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141582,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 141582, "pid": 5, "tid": 7, "ts": 1716454223662777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612579, "dur": 6,
    "args": {
      "External id": 141582,
      "cbid": 211, "correlation": 141582
    }
  },
  {
    "ph": "s", "id": 141582, "pid": 76337, "tid": -914061504, "ts": 1716454223612579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223662782, "dur": 30,
    "args": {
      "External id": 141585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141585,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141585, "pid": 5, "tid": 7, "ts": 1716454223662782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612597, "dur": 7,
    "args": {
      "External id": 141585,
      "cbid": 211, "correlation": 141585
    }
  },
  {
    "ph": "s", "id": 141585, "pid": 76337, "tid": -914061504, "ts": 1716454223612597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612654, "dur": 0,
    "args": {
      "External id": 141596,
      "cbid": 317, "correlation": 141596
    }
  },
  {
    "ph": "f", "id": 141596, "pid": 76337, "tid": -914061504, "ts": 1716454223612654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612655, "dur": 0,
    "args": {
      "External id": 141597,
      "cbid": 203, "correlation": 141597
    }
  },
  {
    "ph": "f", "id": 141597, "pid": 76337, "tid": -914061504, "ts": 1716454223612655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612656, "dur": 0,
    "args": {
      "External id": 141598,
      "cbid": 205, "correlation": 141598
    }
  },
  {
    "ph": "f", "id": 141598, "pid": 76337, "tid": -914061504, "ts": 1716454223612656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223662814, "dur": 22,
    "args": {
      "External id": 141602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141602, "pid": 5, "tid": 7, "ts": 1716454223662814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612669, "dur": 12,
    "args": {
      "External id": 141602,
      "cbid": 211, "correlation": 141602
    }
  },
  {
    "ph": "s", "id": 141602, "pid": 76337, "tid": -914061504, "ts": 1716454223612669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223662837, "dur": 117,
    "args": {
      "External id": 141604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141604,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141604, "pid": 5, "tid": 7, "ts": 1716454223662837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612687, "dur": 6,
    "args": {
      "External id": 141604,
      "cbid": 211, "correlation": 141604
    }
  },
  {
    "ph": "s", "id": 141604, "pid": 76337, "tid": -914061504, "ts": 1716454223612687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223662955, "dur": 23,
    "args": {
      "External id": 141606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141606,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141606, "pid": 5, "tid": 7, "ts": 1716454223662955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612696, "dur": 6,
    "args": {
      "External id": 141606,
      "cbid": 211, "correlation": 141606
    }
  },
  {
    "ph": "s", "id": 141606, "pid": 76337, "tid": -914061504, "ts": 1716454223612696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223662980, "dur": 33,
    "args": {
      "External id": 141612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141612, "pid": 5, "tid": 7, "ts": 1716454223662980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612724, "dur": 8,
    "args": {
      "External id": 141612,
      "cbid": 211, "correlation": 141612
    }
  },
  {
    "ph": "s", "id": 141612, "pid": 76337, "tid": -914061504, "ts": 1716454223612724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223663014, "dur": 207,
    "args": {
      "External id": 141621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141621,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141621, "pid": 5, "tid": 7, "ts": 1716454223663014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612806, "dur": 13,
    "args": {
      "External id": 141621,
      "cbid": 211, "correlation": 141621
    }
  },
  {
    "ph": "s", "id": 141621, "pid": 76337, "tid": -914061504, "ts": 1716454223612806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223663222, "dur": 66,
    "args": {
      "External id": 141643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141643,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141643, "pid": 5, "tid": 7, "ts": 1716454223663222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612863, "dur": 10,
    "args": {
      "External id": 141643,
      "cbid": 211, "correlation": 141643
    }
  },
  {
    "ph": "s", "id": 141643, "pid": 76337, "tid": -914061504, "ts": 1716454223612863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612951, "dur": 1,
    "args": {
      "External id": 141654,
      "cbid": 251, "correlation": 141654
    }
  },
  {
    "ph": "f", "id": 141654, "pid": 76337, "tid": -914061504, "ts": 1716454223612951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223663290, "dur": 154,
    "args": {
      "External id": 141655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141655,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141655, "pid": 5, "tid": 7, "ts": 1716454223663290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223612957, "dur": 13,
    "args": {
      "External id": 141655,
      "cbid": 211, "correlation": 141655
    }
  },
  {
    "ph": "s", "id": 141655, "pid": 76337, "tid": -914061504, "ts": 1716454223612957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613036, "dur": 1,
    "args": {
      "External id": 141666,
      "cbid": 251, "correlation": 141666
    }
  },
  {
    "ph": "f", "id": 141666, "pid": 76337, "tid": -914061504, "ts": 1716454223613036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223663445, "dur": 148,
    "args": {
      "External id": 141667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141667,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141667, "pid": 5, "tid": 7, "ts": 1716454223663445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613040, "dur": 12,
    "args": {
      "External id": 141667,
      "cbid": 211, "correlation": 141667
    }
  },
  {
    "ph": "s", "id": 141667, "pid": 76337, "tid": -914061504, "ts": 1716454223613040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613105, "dur": 1,
    "args": {
      "External id": 141678,
      "cbid": 251, "correlation": 141678
    }
  },
  {
    "ph": "f", "id": 141678, "pid": 76337, "tid": -914061504, "ts": 1716454223613105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223663595, "dur": 148,
    "args": {
      "External id": 141679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141679,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141679, "pid": 5, "tid": 7, "ts": 1716454223663595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613110, "dur": 11,
    "args": {
      "External id": 141679,
      "cbid": 211, "correlation": 141679
    }
  },
  {
    "ph": "s", "id": 141679, "pid": 76337, "tid": -914061504, "ts": 1716454223613110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223663744, "dur": 1956,
    "args": {
      "External id": 141700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141700,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 141700, "pid": 5, "tid": 7, "ts": 1716454223663744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613189, "dur": 12,
    "args": {
      "External id": 141700,
      "cbid": 211, "correlation": 141700
    }
  },
  {
    "ph": "s", "id": 141700, "pid": 76337, "tid": -914061504, "ts": 1716454223613189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613287, "dur": 1,
    "args": {
      "External id": 141718,
      "cbid": 251, "correlation": 141718
    }
  },
  {
    "ph": "f", "id": 141718, "pid": 76337, "tid": -914061504, "ts": 1716454223613287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223665701, "dur": 149,
    "args": {
      "External id": 141720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141720,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 141720, "pid": 5, "tid": 7, "ts": 1716454223665701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613293, "dur": 13,
    "args": {
      "External id": 141720,
      "cbid": 211, "correlation": 141720
    }
  },
  {
    "ph": "s", "id": 141720, "pid": 76337, "tid": -914061504, "ts": 1716454223613293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223665851, "dur": 35,
    "args": {
      "External id": 141728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141728,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141728, "pid": 5, "tid": 7, "ts": 1716454223665851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613364, "dur": 12,
    "args": {
      "External id": 141728,
      "cbid": 211, "correlation": 141728
    }
  },
  {
    "ph": "s", "id": 141728, "pid": 76337, "tid": -914061504, "ts": 1716454223613364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223665887, "dur": 50,
    "args": {
      "External id": 141736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141736,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141736, "pid": 5, "tid": 7, "ts": 1716454223665887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613402, "dur": 9,
    "args": {
      "External id": 141736,
      "cbid": 211, "correlation": 141736
    }
  },
  {
    "ph": "s", "id": 141736, "pid": 76337, "tid": -914061504, "ts": 1716454223613402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223665939, "dur": 30,
    "args": {
      "External id": 141747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141747, "pid": 5, "tid": 7, "ts": 1716454223665939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613473, "dur": 12,
    "args": {
      "External id": 141747,
      "cbid": 211, "correlation": 141747
    }
  },
  {
    "ph": "s", "id": 141747, "pid": 76337, "tid": -914061504, "ts": 1716454223613473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223665971, "dur": 35,
    "args": {
      "External id": 141769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141769,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141769, "pid": 5, "tid": 7, "ts": 1716454223665971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613503, "dur": 8,
    "args": {
      "External id": 141769,
      "cbid": 211, "correlation": 141769
    }
  },
  {
    "ph": "s", "id": 141769, "pid": 76337, "tid": -914061504, "ts": 1716454223613503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613588, "dur": 1,
    "args": {
      "External id": 141780,
      "cbid": 251, "correlation": 141780
    }
  },
  {
    "ph": "f", "id": 141780, "pid": 76337, "tid": -914061504, "ts": 1716454223613588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223666006, "dur": 90,
    "args": {
      "External id": 141781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141781,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141781, "pid": 5, "tid": 7, "ts": 1716454223666006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613594, "dur": 13,
    "args": {
      "External id": 141781,
      "cbid": 211, "correlation": 141781
    }
  },
  {
    "ph": "s", "id": 141781, "pid": 76337, "tid": -914061504, "ts": 1716454223613594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613664, "dur": 1,
    "args": {
      "External id": 141792,
      "cbid": 251, "correlation": 141792
    }
  },
  {
    "ph": "f", "id": 141792, "pid": 76337, "tid": -914061504, "ts": 1716454223613664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613668, "dur": 0,
    "args": {
      "External id": 141793,
      "cbid": 251, "correlation": 141793
    }
  },
  {
    "ph": "f", "id": 141793, "pid": 76337, "tid": -914061504, "ts": 1716454223613668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223666098, "dur": 11,
    "args": {
      "External id": 141794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141794,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 141794, "pid": 5, "tid": 7, "ts": 1716454223666098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613669, "dur": 12,
    "args": {
      "External id": 141794,
      "cbid": 211, "correlation": 141794
    }
  },
  {
    "ph": "s", "id": 141794, "pid": 76337, "tid": -914061504, "ts": 1716454223613669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223666110, "dur": 5,
    "args": {
      "External id": 141796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141796,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 141796, "pid": 5, "tid": 7, "ts": 1716454223666110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613683, "dur": 6,
    "args": {
      "External id": 141796,
      "cbid": 211, "correlation": 141796
    }
  },
  {
    "ph": "s", "id": 141796, "pid": 76337, "tid": -914061504, "ts": 1716454223613683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613740, "dur": 1,
    "args": {
      "External id": 141807,
      "cbid": 251, "correlation": 141807
    }
  },
  {
    "ph": "f", "id": 141807, "pid": 76337, "tid": -914061504, "ts": 1716454223613740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613743, "dur": 0,
    "args": {
      "External id": 141808,
      "cbid": 251, "correlation": 141808
    }
  },
  {
    "ph": "f", "id": 141808, "pid": 76337, "tid": -914061504, "ts": 1716454223613743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223666117, "dur": 7,
    "args": {
      "External id": 141809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141809,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 141809, "pid": 5, "tid": 7, "ts": 1716454223666117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613745, "dur": 12,
    "args": {
      "External id": 141809,
      "cbid": 211, "correlation": 141809
    }
  },
  {
    "ph": "s", "id": 141809, "pid": 76337, "tid": -914061504, "ts": 1716454223613745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223666125, "dur": 3,
    "args": {
      "External id": 141811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141811,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 141811, "pid": 5, "tid": 7, "ts": 1716454223666125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613758, "dur": 6,
    "args": {
      "External id": 141811,
      "cbid": 211, "correlation": 141811
    }
  },
  {
    "ph": "s", "id": 141811, "pid": 76337, "tid": -914061504, "ts": 1716454223613758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223666129, "dur": 92,
    "args": {
      "External id": 141832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141832,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 141832, "pid": 5, "tid": 7, "ts": 1716454223666129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613832, "dur": 12,
    "args": {
      "External id": 141832,
      "cbid": 211, "correlation": 141832
    }
  },
  {
    "ph": "s", "id": 141832, "pid": 76337, "tid": -914061504, "ts": 1716454223613832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613928, "dur": 1,
    "args": {
      "External id": 141850,
      "cbid": 251, "correlation": 141850
    }
  },
  {
    "ph": "f", "id": 141850, "pid": 76337, "tid": -914061504, "ts": 1716454223613928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223666223, "dur": 100,
    "args": {
      "External id": 141852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141852,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141852, "pid": 5, "tid": 7, "ts": 1716454223666223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223613934, "dur": 13,
    "args": {
      "External id": 141852,
      "cbid": 211, "correlation": 141852
    }
  },
  {
    "ph": "s", "id": 141852, "pid": 76337, "tid": -914061504, "ts": 1716454223613934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223666324, "dur": 19,
    "args": {
      "External id": 141860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141860,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141860, "pid": 5, "tid": 7, "ts": 1716454223666324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614011, "dur": 13,
    "args": {
      "External id": 141860,
      "cbid": 211, "correlation": 141860
    }
  },
  {
    "ph": "s", "id": 141860, "pid": 76337, "tid": -914061504, "ts": 1716454223614011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223666344, "dur": 37,
    "args": {
      "External id": 141868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141868,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141868, "pid": 5, "tid": 7, "ts": 1716454223666344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614054, "dur": 10,
    "args": {
      "External id": 141868,
      "cbid": 211, "correlation": 141868
    }
  },
  {
    "ph": "s", "id": 141868, "pid": 76337, "tid": -914061504, "ts": 1716454223614054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223666383, "dur": 34,
    "args": {
      "External id": 141890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141890,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141890, "pid": 5, "tid": 7, "ts": 1716454223666383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614105, "dur": 11,
    "args": {
      "External id": 141890,
      "cbid": 211, "correlation": 141890
    }
  },
  {
    "ph": "s", "id": 141890, "pid": 76337, "tid": -914061504, "ts": 1716454223614105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614194, "dur": 1,
    "args": {
      "External id": 141906,
      "cbid": 251, "correlation": 141906
    }
  },
  {
    "ph": "f", "id": 141906, "pid": 76337, "tid": -914061504, "ts": 1716454223614194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614199, "dur": 0,
    "args": {
      "External id": 141908,
      "cbid": 251, "correlation": 141908
    }
  },
  {
    "ph": "f", "id": 141908, "pid": 76337, "tid": -914061504, "ts": 1716454223614199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223666418, "dur": 541,
    "args": {
      "External id": 141909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141909,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 141909, "pid": 5, "tid": 7, "ts": 1716454223666418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614202, "dur": 13,
    "args": {
      "External id": 141909,
      "cbid": 211, "correlation": 141909
    }
  },
  {
    "ph": "s", "id": 141909, "pid": 76337, "tid": -914061504, "ts": 1716454223614202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223666961, "dur": 127,
    "args": {
      "External id": 141917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141917,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141917, "pid": 5, "tid": 7, "ts": 1716454223666961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614267, "dur": 12,
    "args": {
      "External id": 141917,
      "cbid": 211, "correlation": 141917
    }
  },
  {
    "ph": "s", "id": 141917, "pid": 76337, "tid": -914061504, "ts": 1716454223614267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223667089, "dur": 128,
    "args": {
      "External id": 141925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141925, "pid": 5, "tid": 7, "ts": 1716454223667089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614296, "dur": 8,
    "args": {
      "External id": 141925,
      "cbid": 211, "correlation": 141925
    }
  },
  {
    "ph": "s", "id": 141925, "pid": 76337, "tid": -914061504, "ts": 1716454223614296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614373, "dur": 1,
    "args": {
      "External id": 141941,
      "cbid": 251, "correlation": 141941
    }
  },
  {
    "ph": "f", "id": 141941, "pid": 76337, "tid": -914061504, "ts": 1716454223614373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223667218, "dur": 304,
    "args": {
      "External id": 141943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141943,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141943, "pid": 5, "tid": 7, "ts": 1716454223667218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614378, "dur": 12,
    "args": {
      "External id": 141943,
      "cbid": 211, "correlation": 141943
    }
  },
  {
    "ph": "s", "id": 141943, "pid": 76337, "tid": -914061504, "ts": 1716454223614378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223667524, "dur": 28,
    "args": {
      "External id": 141951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141951,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141951, "pid": 5, "tid": 7, "ts": 1716454223667524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614421, "dur": 9,
    "args": {
      "External id": 141951,
      "cbid": 211, "correlation": 141951
    }
  },
  {
    "ph": "s", "id": 141951, "pid": 76337, "tid": -914061504, "ts": 1716454223614421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223667553, "dur": 82,
    "args": {
      "External id": 141962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141962, "pid": 5, "tid": 7, "ts": 1716454223667553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614488, "dur": 12,
    "args": {
      "External id": 141962,
      "cbid": 211, "correlation": 141962
    }
  },
  {
    "ph": "s", "id": 141962, "pid": 76337, "tid": -914061504, "ts": 1716454223614488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614550, "dur": 0,
    "args": {
      "External id": 141974,
      "cbid": 317, "correlation": 141974
    }
  },
  {
    "ph": "f", "id": 141974, "pid": 76337, "tid": -914061504, "ts": 1716454223614550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614551, "dur": 0,
    "args": {
      "External id": 141975,
      "cbid": 203, "correlation": 141975
    }
  },
  {
    "ph": "f", "id": 141975, "pid": 76337, "tid": -914061504, "ts": 1716454223614551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614552, "dur": 0,
    "args": {
      "External id": 141976,
      "cbid": 205, "correlation": 141976
    }
  },
  {
    "ph": "f", "id": 141976, "pid": 76337, "tid": -914061504, "ts": 1716454223614552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223667636, "dur": 24,
    "args": {
      "External id": 141980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141980, "pid": 5, "tid": 7, "ts": 1716454223667636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614567, "dur": 12,
    "args": {
      "External id": 141980,
      "cbid": 211, "correlation": 141980
    }
  },
  {
    "ph": "s", "id": 141980, "pid": 76337, "tid": -914061504, "ts": 1716454223614567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223667661, "dur": 122,
    "args": {
      "External id": 141982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141982,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 141982, "pid": 5, "tid": 7, "ts": 1716454223667661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614585, "dur": 6,
    "args": {
      "External id": 141982,
      "cbid": 211, "correlation": 141982
    }
  },
  {
    "ph": "s", "id": 141982, "pid": 76337, "tid": -914061504, "ts": 1716454223614585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223667784, "dur": 23,
    "args": {
      "External id": 141984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141984,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141984, "pid": 5, "tid": 7, "ts": 1716454223667784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614595, "dur": 5,
    "args": {
      "External id": 141984,
      "cbid": 211, "correlation": 141984
    }
  },
  {
    "ph": "s", "id": 141984, "pid": 76337, "tid": -914061504, "ts": 1716454223614595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223667809, "dur": 33,
    "args": {
      "External id": 141990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141990, "pid": 5, "tid": 7, "ts": 1716454223667809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614623, "dur": 8,
    "args": {
      "External id": 141990,
      "cbid": 211, "correlation": 141990
    }
  },
  {
    "ph": "s", "id": 141990, "pid": 76337, "tid": -914061504, "ts": 1716454223614623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223667843, "dur": 27,
    "args": {
      "External id": 141998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 141998,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 141998, "pid": 5, "tid": 7, "ts": 1716454223667843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614654, "dur": 8,
    "args": {
      "External id": 141998,
      "cbid": 211, "correlation": 141998
    }
  },
  {
    "ph": "s", "id": 141998, "pid": 76337, "tid": -914061504, "ts": 1716454223614654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614726, "dur": 0,
    "args": {
      "External id": 142008,
      "cbid": 317, "correlation": 142008
    }
  },
  {
    "ph": "f", "id": 142008, "pid": 76337, "tid": -914061504, "ts": 1716454223614726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614726, "dur": 0,
    "args": {
      "External id": 142009,
      "cbid": 203, "correlation": 142009
    }
  },
  {
    "ph": "f", "id": 142009, "pid": 76337, "tid": -914061504, "ts": 1716454223614726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614727, "dur": 0,
    "args": {
      "External id": 142010,
      "cbid": 205, "correlation": 142010
    }
  },
  {
    "ph": "f", "id": 142010, "pid": 76337, "tid": -914061504, "ts": 1716454223614727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223667871, "dur": 22,
    "args": {
      "External id": 142014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142014, "pid": 5, "tid": 7, "ts": 1716454223667871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614742, "dur": 12,
    "args": {
      "External id": 142014,
      "cbid": 211, "correlation": 142014
    }
  },
  {
    "ph": "s", "id": 142014, "pid": 76337, "tid": -914061504, "ts": 1716454223614742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223667894, "dur": 44,
    "args": {
      "External id": 142016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142016,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142016, "pid": 5, "tid": 7, "ts": 1716454223667894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614757, "dur": 5,
    "args": {
      "External id": 142016,
      "cbid": 211, "correlation": 142016
    }
  },
  {
    "ph": "s", "id": 142016, "pid": 76337, "tid": -914061504, "ts": 1716454223614757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223667939, "dur": 234,
    "args": {
      "External id": 142018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142018,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 142018, "pid": 5, "tid": 7, "ts": 1716454223667939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614769, "dur": 7,
    "args": {
      "External id": 142018,
      "cbid": 211, "correlation": 142018
    }
  },
  {
    "ph": "s", "id": 142018, "pid": 76337, "tid": -914061504, "ts": 1716454223614769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223668174, "dur": 7,
    "args": {
      "External id": 142020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142020, "pid": 5, "tid": 7, "ts": 1716454223668174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614778, "dur": 5,
    "args": {
      "External id": 142020,
      "cbid": 211, "correlation": 142020
    }
  },
  {
    "ph": "s", "id": 142020, "pid": 76337, "tid": -914061504, "ts": 1716454223614778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223668182, "dur": 9,
    "args": {
      "External id": 142026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142026, "pid": 5, "tid": 7, "ts": 1716454223668182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614804, "dur": 8,
    "args": {
      "External id": 142026,
      "cbid": 211, "correlation": 142026
    }
  },
  {
    "ph": "s", "id": 142026, "pid": 76337, "tid": -914061504, "ts": 1716454223614804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223668192, "dur": 12,
    "args": {
      "External id": 142046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142046,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 142046, "pid": 5, "tid": 7, "ts": 1716454223668192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614896, "dur": 12,
    "args": {
      "External id": 142046,
      "cbid": 211, "correlation": 142046
    }
  },
  {
    "ph": "s", "id": 142046, "pid": 76337, "tid": -914061504, "ts": 1716454223614896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223668205, "dur": 4,
    "args": {
      "External id": 142058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142058,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 142058, "pid": 5, "tid": 7, "ts": 1716454223668205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614918, "dur": 7,
    "args": {
      "External id": 142058,
      "cbid": 211, "correlation": 142058
    }
  },
  {
    "ph": "s", "id": 142058, "pid": 76337, "tid": -914061504, "ts": 1716454223614918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223668211, "dur": 12,
    "args": {
      "External id": 142061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142061,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142061, "pid": 5, "tid": 7, "ts": 1716454223668211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614936, "dur": 6,
    "args": {
      "External id": 142061,
      "cbid": 211, "correlation": 142061
    }
  },
  {
    "ph": "s", "id": 142061, "pid": 76337, "tid": -914061504, "ts": 1716454223614936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223668224, "dur": 7,
    "args": {
      "External id": 142070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142070,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142070, "pid": 5, "tid": 7, "ts": 1716454223668224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223614983, "dur": 10,
    "args": {
      "External id": 142070,
      "cbid": 211, "correlation": 142070
    }
  },
  {
    "ph": "s", "id": 142070, "pid": 76337, "tid": -914061504, "ts": 1716454223614983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615036, "dur": 0,
    "args": {
      "External id": 142080,
      "cbid": 317, "correlation": 142080
    }
  },
  {
    "ph": "f", "id": 142080, "pid": 76337, "tid": -914061504, "ts": 1716454223615036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615037, "dur": 0,
    "args": {
      "External id": 142081,
      "cbid": 203, "correlation": 142081
    }
  },
  {
    "ph": "f", "id": 142081, "pid": 76337, "tid": -914061504, "ts": 1716454223615037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615038, "dur": 0,
    "args": {
      "External id": 142082,
      "cbid": 205, "correlation": 142082
    }
  },
  {
    "ph": "f", "id": 142082, "pid": 76337, "tid": -914061504, "ts": 1716454223615038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223668233, "dur": 6,
    "args": {
      "External id": 142086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142086,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142086, "pid": 5, "tid": 7, "ts": 1716454223668233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615053, "dur": 11,
    "args": {
      "External id": 142086,
      "cbid": 211, "correlation": 142086
    }
  },
  {
    "ph": "s", "id": 142086, "pid": 76337, "tid": -914061504, "ts": 1716454223615053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223668239, "dur": 84,
    "args": {
      "External id": 142088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142088, "pid": 5, "tid": 7, "ts": 1716454223668239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615067, "dur": 6,
    "args": {
      "External id": 142088,
      "cbid": 211, "correlation": 142088
    }
  },
  {
    "ph": "s", "id": 142088, "pid": 76337, "tid": -914061504, "ts": 1716454223615067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223668326, "dur": 1,
    "args": {
      "External id": 142090,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 142090,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 142090, "pid": 5, "tid": 7, "ts": 1716454223668326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615080, "dur": 9,
    "args": {
      "External id": 142090,
      "cbid": 51, "correlation": 142090
    }
  },
  {
    "ph": "s", "id": 142090, "pid": 76337, "tid": -914061504, "ts": 1716454223615080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223668329, "dur": 543,
    "args": {
      "External id": 142091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142091,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142091, "pid": 5, "tid": 7, "ts": 1716454223668329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615091, "dur": 8,
    "args": {
      "External id": 142091,
      "cbid": 211, "correlation": 142091
    }
  },
  {
    "ph": "s", "id": 142091, "pid": 76337, "tid": -914061504, "ts": 1716454223615091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223668873, "dur": 11,
    "args": {
      "External id": 142093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142093,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142093, "pid": 5, "tid": 7, "ts": 1716454223668873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615102, "dur": 5,
    "args": {
      "External id": 142093,
      "cbid": 211, "correlation": 142093
    }
  },
  {
    "ph": "s", "id": 142093, "pid": 76337, "tid": -914061504, "ts": 1716454223615102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223668886, "dur": 14,
    "args": {
      "External id": 142099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142099, "pid": 5, "tid": 7, "ts": 1716454223668886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615130, "dur": 8,
    "args": {
      "External id": 142099,
      "cbid": 211, "correlation": 142099
    }
  },
  {
    "ph": "s", "id": 142099, "pid": 76337, "tid": -914061504, "ts": 1716454223615130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223668902, "dur": 4,
    "args": {
      "External id": 142107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142107,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 142107, "pid": 5, "tid": 7, "ts": 1716454223668902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615174, "dur": 9,
    "args": {
      "External id": 142107,
      "cbid": 211, "correlation": 142107
    }
  },
  {
    "ph": "s", "id": 142107, "pid": 76337, "tid": -914061504, "ts": 1716454223615174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615239, "dur": 1,
    "args": {
      "External id": 142123,
      "cbid": 251, "correlation": 142123
    }
  },
  {
    "ph": "f", "id": 142123, "pid": 76337, "tid": -914061504, "ts": 1716454223615239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615244, "dur": 0,
    "args": {
      "External id": 142125,
      "cbid": 251, "correlation": 142125
    }
  },
  {
    "ph": "f", "id": 142125, "pid": 76337, "tid": -914061504, "ts": 1716454223615244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223668907, "dur": 13,
    "args": {
      "External id": 142126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142126,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142126, "pid": 5, "tid": 7, "ts": 1716454223668907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615246, "dur": 11,
    "args": {
      "External id": 142126,
      "cbid": 211, "correlation": 142126
    }
  },
  {
    "ph": "s", "id": 142126, "pid": 76337, "tid": -914061504, "ts": 1716454223615246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223668922, "dur": 5,
    "args": {
      "External id": 142128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142128,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142128, "pid": 5, "tid": 7, "ts": 1716454223668922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615258, "dur": 5,
    "args": {
      "External id": 142128,
      "cbid": 211, "correlation": 142128
    }
  },
  {
    "ph": "s", "id": 142128, "pid": 76337, "tid": -914061504, "ts": 1716454223615258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223668929, "dur": 16,
    "args": {
      "External id": 142138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142138, "pid": 5, "tid": 7, "ts": 1716454223668929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615315, "dur": 13,
    "args": {
      "External id": 142138,
      "cbid": 211, "correlation": 142138
    }
  },
  {
    "ph": "s", "id": 142138, "pid": 76337, "tid": -914061504, "ts": 1716454223615315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223668946, "dur": 19,
    "args": {
      "External id": 142158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142158,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 142158, "pid": 5, "tid": 7, "ts": 1716454223668946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615383, "dur": 10,
    "args": {
      "External id": 142158,
      "cbid": 211, "correlation": 142158
    }
  },
  {
    "ph": "s", "id": 142158, "pid": 76337, "tid": -914061504, "ts": 1716454223615383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223668967, "dur": 5,
    "args": {
      "External id": 142170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142170,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 142170, "pid": 5, "tid": 7, "ts": 1716454223668967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615404, "dur": 6,
    "args": {
      "External id": 142170,
      "cbid": 211, "correlation": 142170
    }
  },
  {
    "ph": "s", "id": 142170, "pid": 76337, "tid": -914061504, "ts": 1716454223615404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223668973, "dur": 17,
    "args": {
      "External id": 142173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142173,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142173, "pid": 5, "tid": 7, "ts": 1716454223668973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615423, "dur": 7,
    "args": {
      "External id": 142173,
      "cbid": 211, "correlation": 142173
    }
  },
  {
    "ph": "s", "id": 142173, "pid": 76337, "tid": -914061504, "ts": 1716454223615423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223668991, "dur": 11,
    "args": {
      "External id": 142182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142182,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142182, "pid": 5, "tid": 7, "ts": 1716454223668991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615462, "dur": 10,
    "args": {
      "External id": 142182,
      "cbid": 211, "correlation": 142182
    }
  },
  {
    "ph": "s", "id": 142182, "pid": 76337, "tid": -914061504, "ts": 1716454223615462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615524, "dur": 0,
    "args": {
      "External id": 142192,
      "cbid": 317, "correlation": 142192
    }
  },
  {
    "ph": "f", "id": 142192, "pid": 76337, "tid": -914061504, "ts": 1716454223615524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615525, "dur": 0,
    "args": {
      "External id": 142193,
      "cbid": 203, "correlation": 142193
    }
  },
  {
    "ph": "f", "id": 142193, "pid": 76337, "tid": -914061504, "ts": 1716454223615525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615526, "dur": 0,
    "args": {
      "External id": 142194,
      "cbid": 205, "correlation": 142194
    }
  },
  {
    "ph": "f", "id": 142194, "pid": 76337, "tid": -914061504, "ts": 1716454223615526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223669003, "dur": 11,
    "args": {
      "External id": 142198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142198,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142198, "pid": 5, "tid": 7, "ts": 1716454223669003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615543, "dur": 12,
    "args": {
      "External id": 142198,
      "cbid": 211, "correlation": 142198
    }
  },
  {
    "ph": "s", "id": 142198, "pid": 76337, "tid": -914061504, "ts": 1716454223615543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223669016, "dur": 163,
    "args": {
      "External id": 142200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142200,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142200, "pid": 5, "tid": 7, "ts": 1716454223669016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615557, "dur": 5,
    "args": {
      "External id": 142200,
      "cbid": 211, "correlation": 142200
    }
  },
  {
    "ph": "s", "id": 142200, "pid": 76337, "tid": -914061504, "ts": 1716454223615557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223669181, "dur": 1,
    "args": {
      "External id": 142202,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 142202,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 142202, "pid": 5, "tid": 7, "ts": 1716454223669181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615569, "dur": 8,
    "args": {
      "External id": 142202,
      "cbid": 51, "correlation": 142202
    }
  },
  {
    "ph": "s", "id": 142202, "pid": 76337, "tid": -914061504, "ts": 1716454223615569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223669184, "dur": 663,
    "args": {
      "External id": 142203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142203,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142203, "pid": 5, "tid": 7, "ts": 1716454223669184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615578, "dur": 6,
    "args": {
      "External id": 142203,
      "cbid": 211, "correlation": 142203
    }
  },
  {
    "ph": "s", "id": 142203, "pid": 76337, "tid": -914061504, "ts": 1716454223615578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223669848, "dur": 14,
    "args": {
      "External id": 142205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142205,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142205, "pid": 5, "tid": 7, "ts": 1716454223669848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615589, "dur": 5,
    "args": {
      "External id": 142205,
      "cbid": 211, "correlation": 142205
    }
  },
  {
    "ph": "s", "id": 142205, "pid": 76337, "tid": -914061504, "ts": 1716454223615589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223669863, "dur": 15,
    "args": {
      "External id": 142211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142211, "pid": 5, "tid": 7, "ts": 1716454223669863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615617, "dur": 10,
    "args": {
      "External id": 142211,
      "cbid": 211, "correlation": 142211
    }
  },
  {
    "ph": "s", "id": 142211, "pid": 76337, "tid": -914061504, "ts": 1716454223615617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615676, "dur": 0,
    "args": {
      "External id": 142221,
      "cbid": 317, "correlation": 142221
    }
  },
  {
    "ph": "f", "id": 142221, "pid": 76337, "tid": -914061504, "ts": 1716454223615676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615677, "dur": 0,
    "args": {
      "External id": 142222,
      "cbid": 203, "correlation": 142222
    }
  },
  {
    "ph": "f", "id": 142222, "pid": 76337, "tid": -914061504, "ts": 1716454223615677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615677, "dur": 0,
    "args": {
      "External id": 142223,
      "cbid": 205, "correlation": 142223
    }
  },
  {
    "ph": "f", "id": 142223, "pid": 76337, "tid": -914061504, "ts": 1716454223615677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223669879, "dur": 8,
    "args": {
      "External id": 142227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142227,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142227, "pid": 5, "tid": 7, "ts": 1716454223669879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615692, "dur": 11,
    "args": {
      "External id": 142227,
      "cbid": 211, "correlation": 142227
    }
  },
  {
    "ph": "s", "id": 142227, "pid": 76337, "tid": -914061504, "ts": 1716454223615692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223669889, "dur": 3,
    "args": {
      "External id": 142229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142229,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142229, "pid": 5, "tid": 7, "ts": 1716454223669889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615709, "dur": 6,
    "args": {
      "External id": 142229,
      "cbid": 211, "correlation": 142229
    }
  },
  {
    "ph": "s", "id": 142229, "pid": 76337, "tid": -914061504, "ts": 1716454223615709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615719, "dur": 0,
    "args": {
      "External id": 142230,
      "cbid": 51, "correlation": 142230
    }
  },
  {
    "ph": "s", "id": 142230, "pid": 76337, "tid": -914061504, "ts": 1716454223615719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223669894, "dur": 58,
    "args": {
      "External id": 142231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142231,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 142231, "pid": 5, "tid": 7, "ts": 1716454223669894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615720, "dur": 5,
    "args": {
      "External id": 142231,
      "cbid": 211, "correlation": 142231
    }
  },
  {
    "ph": "s", "id": 142231, "pid": 76337, "tid": -914061504, "ts": 1716454223615720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223669953, "dur": 14,
    "args": {
      "External id": 142236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142236, "pid": 5, "tid": 7, "ts": 1716454223669953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615746, "dur": 8,
    "args": {
      "External id": 142236,
      "cbid": 211, "correlation": 142236
    }
  },
  {
    "ph": "s", "id": 142236, "pid": 76337, "tid": -914061504, "ts": 1716454223615746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223669968, "dur": 11,
    "args": {
      "External id": 142244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142244,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142244, "pid": 5, "tid": 7, "ts": 1716454223669968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615775, "dur": 8,
    "args": {
      "External id": 142244,
      "cbid": 211, "correlation": 142244
    }
  },
  {
    "ph": "s", "id": 142244, "pid": 76337, "tid": -914061504, "ts": 1716454223615775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223669981, "dur": 10,
    "args": {
      "External id": 142252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142252,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142252, "pid": 5, "tid": 7, "ts": 1716454223669981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615803, "dur": 8,
    "args": {
      "External id": 142252,
      "cbid": 211, "correlation": 142252
    }
  },
  {
    "ph": "s", "id": 142252, "pid": 76337, "tid": -914061504, "ts": 1716454223615803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223669992, "dur": 19,
    "args": {
      "External id": 142272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142272,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 142272, "pid": 5, "tid": 7, "ts": 1716454223669992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615885, "dur": 12,
    "args": {
      "External id": 142272,
      "cbid": 211, "correlation": 142272
    }
  },
  {
    "ph": "s", "id": 142272, "pid": 76337, "tid": -914061504, "ts": 1716454223615885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223670013, "dur": 5,
    "args": {
      "External id": 142284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142284,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 142284, "pid": 5, "tid": 7, "ts": 1716454223670013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615907, "dur": 6,
    "args": {
      "External id": 142284,
      "cbid": 211, "correlation": 142284
    }
  },
  {
    "ph": "s", "id": 142284, "pid": 76337, "tid": -914061504, "ts": 1716454223615907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223670019, "dur": 17,
    "args": {
      "External id": 142287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142287, "pid": 5, "tid": 7, "ts": 1716454223670019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615925, "dur": 6,
    "args": {
      "External id": 142287,
      "cbid": 211, "correlation": 142287
    }
  },
  {
    "ph": "s", "id": 142287, "pid": 76337, "tid": -914061504, "ts": 1716454223615925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615989, "dur": 0,
    "args": {
      "External id": 142298,
      "cbid": 317, "correlation": 142298
    }
  },
  {
    "ph": "f", "id": 142298, "pid": 76337, "tid": -914061504, "ts": 1716454223615989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615990, "dur": 0,
    "args": {
      "External id": 142299,
      "cbid": 203, "correlation": 142299
    }
  },
  {
    "ph": "f", "id": 142299, "pid": 76337, "tid": -914061504, "ts": 1716454223615990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223615991, "dur": 0,
    "args": {
      "External id": 142300,
      "cbid": 205, "correlation": 142300
    }
  },
  {
    "ph": "f", "id": 142300, "pid": 76337, "tid": -914061504, "ts": 1716454223615991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223670037, "dur": 12,
    "args": {
      "External id": 142304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142304, "pid": 5, "tid": 7, "ts": 1716454223670037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616006, "dur": 12,
    "args": {
      "External id": 142304,
      "cbid": 211, "correlation": 142304
    }
  },
  {
    "ph": "s", "id": 142304, "pid": 76337, "tid": -914061504, "ts": 1716454223616006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223670050, "dur": 3,
    "args": {
      "External id": 142306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142306, "pid": 5, "tid": 7, "ts": 1716454223670050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616024, "dur": 5,
    "args": {
      "External id": 142306,
      "cbid": 211, "correlation": 142306
    }
  },
  {
    "ph": "s", "id": 142306, "pid": 76337, "tid": -914061504, "ts": 1716454223616024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616033, "dur": 0,
    "args": {
      "External id": 142307,
      "cbid": 51, "correlation": 142307
    }
  },
  {
    "ph": "s", "id": 142307, "pid": 76337, "tid": -914061504, "ts": 1716454223616033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223670054, "dur": 97,
    "args": {
      "External id": 142308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142308,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 142308, "pid": 5, "tid": 7, "ts": 1716454223670054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616034, "dur": 5,
    "args": {
      "External id": 142308,
      "cbid": 211, "correlation": 142308
    }
  },
  {
    "ph": "s", "id": 142308, "pid": 76337, "tid": -914061504, "ts": 1716454223616034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223670153, "dur": 15,
    "args": {
      "External id": 142313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142313, "pid": 5, "tid": 7, "ts": 1716454223670153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616061, "dur": 9,
    "args": {
      "External id": 142313,
      "cbid": 211, "correlation": 142313
    }
  },
  {
    "ph": "s", "id": 142313, "pid": 76337, "tid": -914061504, "ts": 1716454223616061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223670170, "dur": 84,
    "args": {
      "External id": 142322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142322,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142322, "pid": 5, "tid": 7, "ts": 1716454223670170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616143, "dur": 14,
    "args": {
      "External id": 142322,
      "cbid": 211, "correlation": 142322
    }
  },
  {
    "ph": "s", "id": 142322, "pid": 76337, "tid": -914061504, "ts": 1716454223616143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223670255, "dur": 31,
    "args": {
      "External id": 142344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142344,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142344, "pid": 5, "tid": 7, "ts": 1716454223670255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616200, "dur": 10,
    "args": {
      "External id": 142344,
      "cbid": 211, "correlation": 142344
    }
  },
  {
    "ph": "s", "id": 142344, "pid": 76337, "tid": -914061504, "ts": 1716454223616200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616289, "dur": 1,
    "args": {
      "External id": 142355,
      "cbid": 251, "correlation": 142355
    }
  },
  {
    "ph": "f", "id": 142355, "pid": 76337, "tid": -914061504, "ts": 1716454223616289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223670287, "dur": 165,
    "args": {
      "External id": 142356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142356,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142356, "pid": 5, "tid": 7, "ts": 1716454223670287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616295, "dur": 13,
    "args": {
      "External id": 142356,
      "cbid": 211, "correlation": 142356
    }
  },
  {
    "ph": "s", "id": 142356, "pid": 76337, "tid": -914061504, "ts": 1716454223616295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616365, "dur": 1,
    "args": {
      "External id": 142367,
      "cbid": 251, "correlation": 142367
    }
  },
  {
    "ph": "f", "id": 142367, "pid": 76337, "tid": -914061504, "ts": 1716454223616365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223670453, "dur": 159,
    "args": {
      "External id": 142368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142368,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142368, "pid": 5, "tid": 7, "ts": 1716454223670453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616369, "dur": 12,
    "args": {
      "External id": 142368,
      "cbid": 211, "correlation": 142368
    }
  },
  {
    "ph": "s", "id": 142368, "pid": 76337, "tid": -914061504, "ts": 1716454223616369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616434, "dur": 1,
    "args": {
      "External id": 142379,
      "cbid": 251, "correlation": 142379
    }
  },
  {
    "ph": "f", "id": 142379, "pid": 76337, "tid": -914061504, "ts": 1716454223616434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223670613, "dur": 158,
    "args": {
      "External id": 142380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142380,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142380, "pid": 5, "tid": 7, "ts": 1716454223670613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616438, "dur": 11,
    "args": {
      "External id": 142380,
      "cbid": 211, "correlation": 142380
    }
  },
  {
    "ph": "s", "id": 142380, "pid": 76337, "tid": -914061504, "ts": 1716454223616438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223670773, "dur": 337,
    "args": {
      "External id": 142405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142405,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142405, "pid": 5, "tid": 7, "ts": 1716454223670773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616524, "dur": 13,
    "args": {
      "External id": 142405,
      "cbid": 211, "correlation": 142405
    }
  },
  {
    "ph": "s", "id": 142405, "pid": 76337, "tid": -914061504, "ts": 1716454223616524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616625, "dur": 1,
    "args": {
      "External id": 142423,
      "cbid": 251, "correlation": 142423
    }
  },
  {
    "ph": "f", "id": 142423, "pid": 76337, "tid": -914061504, "ts": 1716454223616625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223671111, "dur": 168,
    "args": {
      "External id": 142425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142425,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142425, "pid": 5, "tid": 7, "ts": 1716454223671111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616632, "dur": 13,
    "args": {
      "External id": 142425,
      "cbid": 211, "correlation": 142425
    }
  },
  {
    "ph": "s", "id": 142425, "pid": 76337, "tid": -914061504, "ts": 1716454223616632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223671281, "dur": 20,
    "args": {
      "External id": 142433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142433,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142433, "pid": 5, "tid": 7, "ts": 1716454223671281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616701, "dur": 12,
    "args": {
      "External id": 142433,
      "cbid": 211, "correlation": 142433
    }
  },
  {
    "ph": "s", "id": 142433, "pid": 76337, "tid": -914061504, "ts": 1716454223616701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223671302, "dur": 28,
    "args": {
      "External id": 142441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142441,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142441, "pid": 5, "tid": 7, "ts": 1716454223671302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616740, "dur": 9,
    "args": {
      "External id": 142441,
      "cbid": 211, "correlation": 142441
    }
  },
  {
    "ph": "s", "id": 142441, "pid": 76337, "tid": -914061504, "ts": 1716454223616740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223671332, "dur": 18,
    "args": {
      "External id": 142452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142452, "pid": 5, "tid": 7, "ts": 1716454223671332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616812, "dur": 12,
    "args": {
      "External id": 142452,
      "cbid": 211, "correlation": 142452
    }
  },
  {
    "ph": "s", "id": 142452, "pid": 76337, "tid": -914061504, "ts": 1716454223616812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223671351, "dur": 16,
    "args": {
      "External id": 142474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142474,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142474, "pid": 5, "tid": 7, "ts": 1716454223671351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616844, "dur": 7,
    "args": {
      "External id": 142474,
      "cbid": 211, "correlation": 142474
    }
  },
  {
    "ph": "s", "id": 142474, "pid": 76337, "tid": -914061504, "ts": 1716454223616844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616929, "dur": 1,
    "args": {
      "External id": 142485,
      "cbid": 251, "correlation": 142485
    }
  },
  {
    "ph": "f", "id": 142485, "pid": 76337, "tid": -914061504, "ts": 1716454223616929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223671368, "dur": 89,
    "args": {
      "External id": 142486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142486,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142486, "pid": 5, "tid": 7, "ts": 1716454223671368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223616936, "dur": 15,
    "args": {
      "External id": 142486,
      "cbid": 211, "correlation": 142486
    }
  },
  {
    "ph": "s", "id": 142486, "pid": 76337, "tid": -914061504, "ts": 1716454223616936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617015, "dur": 1,
    "args": {
      "External id": 142497,
      "cbid": 251, "correlation": 142497
    }
  },
  {
    "ph": "f", "id": 142497, "pid": 76337, "tid": -914061504, "ts": 1716454223617015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617019, "dur": 0,
    "args": {
      "External id": 142498,
      "cbid": 251, "correlation": 142498
    }
  },
  {
    "ph": "f", "id": 142498, "pid": 76337, "tid": -914061504, "ts": 1716454223617019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223671459, "dur": 12,
    "args": {
      "External id": 142499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142499,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142499, "pid": 5, "tid": 7, "ts": 1716454223671459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617021, "dur": 13,
    "args": {
      "External id": 142499,
      "cbid": 211, "correlation": 142499
    }
  },
  {
    "ph": "s", "id": 142499, "pid": 76337, "tid": -914061504, "ts": 1716454223617021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223671472, "dur": 5,
    "args": {
      "External id": 142501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142501,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142501, "pid": 5, "tid": 7, "ts": 1716454223671472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617035, "dur": 6,
    "args": {
      "External id": 142501,
      "cbid": 211, "correlation": 142501
    }
  },
  {
    "ph": "s", "id": 142501, "pid": 76337, "tid": -914061504, "ts": 1716454223617035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617093, "dur": 1,
    "args": {
      "External id": 142512,
      "cbid": 251, "correlation": 142512
    }
  },
  {
    "ph": "f", "id": 142512, "pid": 76337, "tid": -914061504, "ts": 1716454223617093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617096, "dur": 0,
    "args": {
      "External id": 142513,
      "cbid": 251, "correlation": 142513
    }
  },
  {
    "ph": "f", "id": 142513, "pid": 76337, "tid": -914061504, "ts": 1716454223617096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223671479, "dur": 8,
    "args": {
      "External id": 142514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142514,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142514, "pid": 5, "tid": 7, "ts": 1716454223671479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617098, "dur": 11,
    "args": {
      "External id": 142514,
      "cbid": 211, "correlation": 142514
    }
  },
  {
    "ph": "s", "id": 142514, "pid": 76337, "tid": -914061504, "ts": 1716454223617098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223671488, "dur": 3,
    "args": {
      "External id": 142516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142516,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142516, "pid": 5, "tid": 7, "ts": 1716454223671488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617111, "dur": 5,
    "args": {
      "External id": 142516,
      "cbid": 211, "correlation": 142516
    }
  },
  {
    "ph": "s", "id": 142516, "pid": 76337, "tid": -914061504, "ts": 1716454223617111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223671493, "dur": 56,
    "args": {
      "External id": 142541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142541,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142541, "pid": 5, "tid": 7, "ts": 1716454223671493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617188, "dur": 12,
    "args": {
      "External id": 142541,
      "cbid": 211, "correlation": 142541
    }
  },
  {
    "ph": "s", "id": 142541, "pid": 76337, "tid": -914061504, "ts": 1716454223617188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617286, "dur": 2,
    "args": {
      "External id": 142559,
      "cbid": 251, "correlation": 142559
    }
  },
  {
    "ph": "f", "id": 142559, "pid": 76337, "tid": -914061504, "ts": 1716454223617286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223671550, "dur": 91,
    "args": {
      "External id": 142561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142561,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142561, "pid": 5, "tid": 7, "ts": 1716454223671550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617293, "dur": 14,
    "args": {
      "External id": 142561,
      "cbid": 211, "correlation": 142561
    }
  },
  {
    "ph": "s", "id": 142561, "pid": 76337, "tid": -914061504, "ts": 1716454223617293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223671643, "dur": 10,
    "args": {
      "External id": 142569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142569,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142569, "pid": 5, "tid": 7, "ts": 1716454223671643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617363, "dur": 12,
    "args": {
      "External id": 142569,
      "cbid": 211, "correlation": 142569
    }
  },
  {
    "ph": "s", "id": 142569, "pid": 76337, "tid": -914061504, "ts": 1716454223617363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223671654, "dur": 20,
    "args": {
      "External id": 142577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142577, "pid": 5, "tid": 7, "ts": 1716454223671654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617405, "dur": 9,
    "args": {
      "External id": 142577,
      "cbid": 211, "correlation": 142577
    }
  },
  {
    "ph": "s", "id": 142577, "pid": 76337, "tid": -914061504, "ts": 1716454223617405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223671675, "dur": 17,
    "args": {
      "External id": 142599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142599,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142599, "pid": 5, "tid": 7, "ts": 1716454223671675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617456, "dur": 10,
    "args": {
      "External id": 142599,
      "cbid": 211, "correlation": 142599
    }
  },
  {
    "ph": "s", "id": 142599, "pid": 76337, "tid": -914061504, "ts": 1716454223617456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617547, "dur": 2,
    "args": {
      "External id": 142615,
      "cbid": 251, "correlation": 142615
    }
  },
  {
    "ph": "f", "id": 142615, "pid": 76337, "tid": -914061504, "ts": 1716454223617547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617552, "dur": 0,
    "args": {
      "External id": 142617,
      "cbid": 251, "correlation": 142617
    }
  },
  {
    "ph": "f", "id": 142617, "pid": 76337, "tid": -914061504, "ts": 1716454223617552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223671694, "dur": 496,
    "args": {
      "External id": 142618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142618,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142618, "pid": 5, "tid": 7, "ts": 1716454223671694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617555, "dur": 15,
    "args": {
      "External id": 142618,
      "cbid": 211, "correlation": 142618
    }
  },
  {
    "ph": "s", "id": 142618, "pid": 76337, "tid": -914061504, "ts": 1716454223617555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223672191, "dur": 67,
    "args": {
      "External id": 142626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142626, "pid": 5, "tid": 7, "ts": 1716454223672191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617622, "dur": 12,
    "args": {
      "External id": 142626,
      "cbid": 211, "correlation": 142626
    }
  },
  {
    "ph": "s", "id": 142626, "pid": 76337, "tid": -914061504, "ts": 1716454223617622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223672259, "dur": 68,
    "args": {
      "External id": 142634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142634, "pid": 5, "tid": 7, "ts": 1716454223672259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617654, "dur": 9,
    "args": {
      "External id": 142634,
      "cbid": 211, "correlation": 142634
    }
  },
  {
    "ph": "s", "id": 142634, "pid": 76337, "tid": -914061504, "ts": 1716454223617654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617734, "dur": 1,
    "args": {
      "External id": 142650,
      "cbid": 251, "correlation": 142650
    }
  },
  {
    "ph": "f", "id": 142650, "pid": 76337, "tid": -914061504, "ts": 1716454223617734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223672329, "dur": 1,
    "args": {
      "External id": 142652,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 142652,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 142652, "pid": 5, "tid": 7, "ts": 1716454223672329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617740, "dur": 12,
    "args": {
      "External id": 142652,
      "cbid": 51, "correlation": 142652
    }
  },
  {
    "ph": "s", "id": 142652, "pid": 76337, "tid": -914061504, "ts": 1716454223617740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223672333, "dur": 272,
    "args": {
      "External id": 142653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142653,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142653, "pid": 5, "tid": 7, "ts": 1716454223672333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617754, "dur": 11,
    "args": {
      "External id": 142653,
      "cbid": 211, "correlation": 142653
    }
  },
  {
    "ph": "s", "id": 142653, "pid": 76337, "tid": -914061504, "ts": 1716454223617754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223672606, "dur": 14,
    "args": {
      "External id": 142661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142661,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142661, "pid": 5, "tid": 7, "ts": 1716454223672606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617796, "dur": 10,
    "args": {
      "External id": 142661,
      "cbid": 211, "correlation": 142661
    }
  },
  {
    "ph": "s", "id": 142661, "pid": 76337, "tid": -914061504, "ts": 1716454223617796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223672621, "dur": 38,
    "args": {
      "External id": 142672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142672,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142672, "pid": 5, "tid": 7, "ts": 1716454223672621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617862, "dur": 12,
    "args": {
      "External id": 142672,
      "cbid": 211, "correlation": 142672
    }
  },
  {
    "ph": "s", "id": 142672, "pid": 76337, "tid": -914061504, "ts": 1716454223617862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617928, "dur": 0,
    "args": {
      "External id": 142684,
      "cbid": 317, "correlation": 142684
    }
  },
  {
    "ph": "f", "id": 142684, "pid": 76337, "tid": -914061504, "ts": 1716454223617928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617929, "dur": 0,
    "args": {
      "External id": 142685,
      "cbid": 203, "correlation": 142685
    }
  },
  {
    "ph": "f", "id": 142685, "pid": 76337, "tid": -914061504, "ts": 1716454223617929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617930, "dur": 0,
    "args": {
      "External id": 142686,
      "cbid": 205, "correlation": 142686
    }
  },
  {
    "ph": "f", "id": 142686, "pid": 76337, "tid": -914061504, "ts": 1716454223617930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223672661, "dur": 13,
    "args": {
      "External id": 142690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142690,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142690, "pid": 5, "tid": 7, "ts": 1716454223672661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617946, "dur": 12,
    "args": {
      "External id": 142690,
      "cbid": 211, "correlation": 142690
    }
  },
  {
    "ph": "s", "id": 142690, "pid": 76337, "tid": -914061504, "ts": 1716454223617946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223672675, "dur": 4,
    "args": {
      "External id": 142692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142692, "pid": 5, "tid": 7, "ts": 1716454223672675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617962, "dur": 6,
    "args": {
      "External id": 142692,
      "cbid": 211, "correlation": 142692
    }
  },
  {
    "ph": "s", "id": 142692, "pid": 76337, "tid": -914061504, "ts": 1716454223617962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617970, "dur": 0,
    "args": {
      "External id": 142693,
      "cbid": 51, "correlation": 142693
    }
  },
  {
    "ph": "s", "id": 142693, "pid": 76337, "tid": -914061504, "ts": 1716454223617970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223672680, "dur": 97,
    "args": {
      "External id": 142694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142694,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 142694, "pid": 5, "tid": 7, "ts": 1716454223672680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223617971, "dur": 14,
    "args": {
      "External id": 142694,
      "cbid": 211, "correlation": 142694
    }
  },
  {
    "ph": "s", "id": 142694, "pid": 76337, "tid": -914061504, "ts": 1716454223617971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223672778, "dur": 16,
    "args": {
      "External id": 142699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142699, "pid": 5, "tid": 7, "ts": 1716454223672778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618008, "dur": 9,
    "args": {
      "External id": 142699,
      "cbid": 211, "correlation": 142699
    }
  },
  {
    "ph": "s", "id": 142699, "pid": 76337, "tid": -914061504, "ts": 1716454223618008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223672796, "dur": 12,
    "args": {
      "External id": 142707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142707,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142707, "pid": 5, "tid": 7, "ts": 1716454223672796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618041, "dur": 9,
    "args": {
      "External id": 142707,
      "cbid": 211, "correlation": 142707
    }
  },
  {
    "ph": "s", "id": 142707, "pid": 76337, "tid": -914061504, "ts": 1716454223618041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223672809, "dur": 18,
    "args": {
      "External id": 142727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142727,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 142727, "pid": 5, "tid": 7, "ts": 1716454223672809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618113, "dur": 11,
    "args": {
      "External id": 142727,
      "cbid": 211, "correlation": 142727
    }
  },
  {
    "ph": "s", "id": 142727, "pid": 76337, "tid": -914061504, "ts": 1716454223618113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223672828, "dur": 5,
    "args": {
      "External id": 142739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142739,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 142739, "pid": 5, "tid": 7, "ts": 1716454223672828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618135, "dur": 7,
    "args": {
      "External id": 142739,
      "cbid": 211, "correlation": 142739
    }
  },
  {
    "ph": "s", "id": 142739, "pid": 76337, "tid": -914061504, "ts": 1716454223618135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223672835, "dur": 18,
    "args": {
      "External id": 142742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142742, "pid": 5, "tid": 7, "ts": 1716454223672835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618153, "dur": 7,
    "args": {
      "External id": 142742,
      "cbid": 211, "correlation": 142742
    }
  },
  {
    "ph": "s", "id": 142742, "pid": 76337, "tid": -914061504, "ts": 1716454223618153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223672854, "dur": 13,
    "args": {
      "External id": 142751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142751,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142751, "pid": 5, "tid": 7, "ts": 1716454223672854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618193, "dur": 9,
    "args": {
      "External id": 142751,
      "cbid": 211, "correlation": 142751
    }
  },
  {
    "ph": "s", "id": 142751, "pid": 76337, "tid": -914061504, "ts": 1716454223618193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618244, "dur": 0,
    "args": {
      "External id": 142761,
      "cbid": 317, "correlation": 142761
    }
  },
  {
    "ph": "f", "id": 142761, "pid": 76337, "tid": -914061504, "ts": 1716454223618244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618245, "dur": 0,
    "args": {
      "External id": 142762,
      "cbid": 203, "correlation": 142762
    }
  },
  {
    "ph": "f", "id": 142762, "pid": 76337, "tid": -914061504, "ts": 1716454223618245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618246, "dur": 0,
    "args": {
      "External id": 142763,
      "cbid": 205, "correlation": 142763
    }
  },
  {
    "ph": "f", "id": 142763, "pid": 76337, "tid": -914061504, "ts": 1716454223618246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223672868, "dur": 11,
    "args": {
      "External id": 142767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142767,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142767, "pid": 5, "tid": 7, "ts": 1716454223672868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618260, "dur": 11,
    "args": {
      "External id": 142767,
      "cbid": 211, "correlation": 142767
    }
  },
  {
    "ph": "s", "id": 142767, "pid": 76337, "tid": -914061504, "ts": 1716454223618260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223672881, "dur": 163,
    "args": {
      "External id": 142769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142769,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142769, "pid": 5, "tid": 7, "ts": 1716454223672881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618273, "dur": 5,
    "args": {
      "External id": 142769,
      "cbid": 211, "correlation": 142769
    }
  },
  {
    "ph": "s", "id": 142769, "pid": 76337, "tid": -914061504, "ts": 1716454223618273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223673046, "dur": 1,
    "args": {
      "External id": 142771,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 142771,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 142771, "pid": 5, "tid": 7, "ts": 1716454223673046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618284, "dur": 7,
    "args": {
      "External id": 142771,
      "cbid": 51, "correlation": 142771
    }
  },
  {
    "ph": "s", "id": 142771, "pid": 76337, "tid": -914061504, "ts": 1716454223618284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223673050, "dur": 664,
    "args": {
      "External id": 142772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142772,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142772, "pid": 5, "tid": 7, "ts": 1716454223673050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618292, "dur": 6,
    "args": {
      "External id": 142772,
      "cbid": 211, "correlation": 142772
    }
  },
  {
    "ph": "s", "id": 142772, "pid": 76337, "tid": -914061504, "ts": 1716454223618292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223673715, "dur": 13,
    "args": {
      "External id": 142774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142774,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142774, "pid": 5, "tid": 7, "ts": 1716454223673715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618302, "dur": 5,
    "args": {
      "External id": 142774,
      "cbid": 211, "correlation": 142774
    }
  },
  {
    "ph": "s", "id": 142774, "pid": 76337, "tid": -914061504, "ts": 1716454223618302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223673730, "dur": 14,
    "args": {
      "External id": 142780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142780,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142780, "pid": 5, "tid": 7, "ts": 1716454223673730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618330, "dur": 9,
    "args": {
      "External id": 142780,
      "cbid": 211, "correlation": 142780
    }
  },
  {
    "ph": "s", "id": 142780, "pid": 76337, "tid": -914061504, "ts": 1716454223618330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223673746, "dur": 4,
    "args": {
      "External id": 142788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142788,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 142788, "pid": 5, "tid": 7, "ts": 1716454223673746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618374, "dur": 9,
    "args": {
      "External id": 142788,
      "cbid": 211, "correlation": 142788
    }
  },
  {
    "ph": "s", "id": 142788, "pid": 76337, "tid": -914061504, "ts": 1716454223618374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618438, "dur": 1,
    "args": {
      "External id": 142804,
      "cbid": 251, "correlation": 142804
    }
  },
  {
    "ph": "f", "id": 142804, "pid": 76337, "tid": -914061504, "ts": 1716454223618438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618444, "dur": 0,
    "args": {
      "External id": 142806,
      "cbid": 251, "correlation": 142806
    }
  },
  {
    "ph": "f", "id": 142806, "pid": 76337, "tid": -914061504, "ts": 1716454223618444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223673751, "dur": 13,
    "args": {
      "External id": 142807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142807,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142807, "pid": 5, "tid": 7, "ts": 1716454223673751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618446, "dur": 11,
    "args": {
      "External id": 142807,
      "cbid": 211, "correlation": 142807
    }
  },
  {
    "ph": "s", "id": 142807, "pid": 76337, "tid": -914061504, "ts": 1716454223618446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223673766, "dur": 5,
    "args": {
      "External id": 142809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142809,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142809, "pid": 5, "tid": 7, "ts": 1716454223673766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618459, "dur": 5,
    "args": {
      "External id": 142809,
      "cbid": 211, "correlation": 142809
    }
  },
  {
    "ph": "s", "id": 142809, "pid": 76337, "tid": -914061504, "ts": 1716454223618459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223673772, "dur": 17,
    "args": {
      "External id": 142819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142819, "pid": 5, "tid": 7, "ts": 1716454223673772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618517, "dur": 12,
    "args": {
      "External id": 142819,
      "cbid": 211, "correlation": 142819
    }
  },
  {
    "ph": "s", "id": 142819, "pid": 76337, "tid": -914061504, "ts": 1716454223618517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223673790, "dur": 17,
    "args": {
      "External id": 142839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142839,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 142839, "pid": 5, "tid": 7, "ts": 1716454223673790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618583, "dur": 11,
    "args": {
      "External id": 142839,
      "cbid": 211, "correlation": 142839
    }
  },
  {
    "ph": "s", "id": 142839, "pid": 76337, "tid": -914061504, "ts": 1716454223618583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223673809, "dur": 4,
    "args": {
      "External id": 142851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142851,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 142851, "pid": 5, "tid": 7, "ts": 1716454223673809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618604, "dur": 6,
    "args": {
      "External id": 142851,
      "cbid": 211, "correlation": 142851
    }
  },
  {
    "ph": "s", "id": 142851, "pid": 76337, "tid": -914061504, "ts": 1716454223618604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223673815, "dur": 17,
    "args": {
      "External id": 142854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142854, "pid": 5, "tid": 7, "ts": 1716454223673815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618622, "dur": 6,
    "args": {
      "External id": 142854,
      "cbid": 211, "correlation": 142854
    }
  },
  {
    "ph": "s", "id": 142854, "pid": 76337, "tid": -914061504, "ts": 1716454223618622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223673833, "dur": 11,
    "args": {
      "External id": 142863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142863,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142863, "pid": 5, "tid": 7, "ts": 1716454223673833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618663, "dur": 9,
    "args": {
      "External id": 142863,
      "cbid": 211, "correlation": 142863
    }
  },
  {
    "ph": "s", "id": 142863, "pid": 76337, "tid": -914061504, "ts": 1716454223618663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618724, "dur": 0,
    "args": {
      "External id": 142873,
      "cbid": 317, "correlation": 142873
    }
  },
  {
    "ph": "f", "id": 142873, "pid": 76337, "tid": -914061504, "ts": 1716454223618724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618725, "dur": 0,
    "args": {
      "External id": 142874,
      "cbid": 203, "correlation": 142874
    }
  },
  {
    "ph": "f", "id": 142874, "pid": 76337, "tid": -914061504, "ts": 1716454223618725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618726, "dur": 0,
    "args": {
      "External id": 142875,
      "cbid": 205, "correlation": 142875
    }
  },
  {
    "ph": "f", "id": 142875, "pid": 76337, "tid": -914061504, "ts": 1716454223618726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223673846, "dur": 11,
    "args": {
      "External id": 142879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142879,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142879, "pid": 5, "tid": 7, "ts": 1716454223673846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618739, "dur": 13,
    "args": {
      "External id": 142879,
      "cbid": 211, "correlation": 142879
    }
  },
  {
    "ph": "s", "id": 142879, "pid": 76337, "tid": -914061504, "ts": 1716454223618739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223673858, "dur": 164,
    "args": {
      "External id": 142881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142881, "pid": 5, "tid": 7, "ts": 1716454223673858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618754, "dur": 5,
    "args": {
      "External id": 142881,
      "cbid": 211, "correlation": 142881
    }
  },
  {
    "ph": "s", "id": 142881, "pid": 76337, "tid": -914061504, "ts": 1716454223618754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223674024, "dur": 1,
    "args": {
      "External id": 142883,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 142883,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 142883, "pid": 5, "tid": 7, "ts": 1716454223674024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618765, "dur": 6,
    "args": {
      "External id": 142883,
      "cbid": 51, "correlation": 142883
    }
  },
  {
    "ph": "s", "id": 142883, "pid": 76337, "tid": -914061504, "ts": 1716454223618765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223674028, "dur": 650,
    "args": {
      "External id": 142884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142884,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 142884, "pid": 5, "tid": 7, "ts": 1716454223674028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618773, "dur": 6,
    "args": {
      "External id": 142884,
      "cbid": 211, "correlation": 142884
    }
  },
  {
    "ph": "s", "id": 142884, "pid": 76337, "tid": -914061504, "ts": 1716454223618773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223674679, "dur": 13,
    "args": {
      "External id": 142886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142886,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142886, "pid": 5, "tid": 7, "ts": 1716454223674679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618782, "dur": 5,
    "args": {
      "External id": 142886,
      "cbid": 211, "correlation": 142886
    }
  },
  {
    "ph": "s", "id": 142886, "pid": 76337, "tid": -914061504, "ts": 1716454223618782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223674694, "dur": 15,
    "args": {
      "External id": 142892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142892, "pid": 5, "tid": 7, "ts": 1716454223674694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618811, "dur": 8,
    "args": {
      "External id": 142892,
      "cbid": 211, "correlation": 142892
    }
  },
  {
    "ph": "s", "id": 142892, "pid": 76337, "tid": -914061504, "ts": 1716454223618811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223674710, "dur": 12,
    "args": {
      "External id": 142900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142900,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142900, "pid": 5, "tid": 7, "ts": 1716454223674710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618842, "dur": 9,
    "args": {
      "External id": 142900,
      "cbid": 211, "correlation": 142900
    }
  },
  {
    "ph": "s", "id": 142900, "pid": 76337, "tid": -914061504, "ts": 1716454223618842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223674723, "dur": 10,
    "args": {
      "External id": 142908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142908,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142908, "pid": 5, "tid": 7, "ts": 1716454223674723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618872, "dur": 8,
    "args": {
      "External id": 142908,
      "cbid": 211, "correlation": 142908
    }
  },
  {
    "ph": "s", "id": 142908, "pid": 76337, "tid": -914061504, "ts": 1716454223618872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223674735, "dur": 19,
    "args": {
      "External id": 142928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142928,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 142928, "pid": 5, "tid": 7, "ts": 1716454223674735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618951, "dur": 12,
    "args": {
      "External id": 142928,
      "cbid": 211, "correlation": 142928
    }
  },
  {
    "ph": "s", "id": 142928, "pid": 76337, "tid": -914061504, "ts": 1716454223618951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223674755, "dur": 4,
    "args": {
      "External id": 142940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142940,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 142940, "pid": 5, "tid": 7, "ts": 1716454223674755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223618981, "dur": 7,
    "args": {
      "External id": 142940,
      "cbid": 211, "correlation": 142940
    }
  },
  {
    "ph": "s", "id": 142940, "pid": 76337, "tid": -914061504, "ts": 1716454223618981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223674760, "dur": 16,
    "args": {
      "External id": 142943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142943, "pid": 5, "tid": 7, "ts": 1716454223674760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619000, "dur": 7,
    "args": {
      "External id": 142943,
      "cbid": 211, "correlation": 142943
    }
  },
  {
    "ph": "s", "id": 142943, "pid": 76337, "tid": -914061504, "ts": 1716454223619000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619058, "dur": 0,
    "args": {
      "External id": 142954,
      "cbid": 317, "correlation": 142954
    }
  },
  {
    "ph": "f", "id": 142954, "pid": 76337, "tid": -914061504, "ts": 1716454223619058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619058, "dur": 0,
    "args": {
      "External id": 142955,
      "cbid": 203, "correlation": 142955
    }
  },
  {
    "ph": "f", "id": 142955, "pid": 76337, "tid": -914061504, "ts": 1716454223619058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619059, "dur": 0,
    "args": {
      "External id": 142956,
      "cbid": 205, "correlation": 142956
    }
  },
  {
    "ph": "f", "id": 142956, "pid": 76337, "tid": -914061504, "ts": 1716454223619059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223674778, "dur": 11,
    "args": {
      "External id": 142960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142960,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142960, "pid": 5, "tid": 7, "ts": 1716454223674778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619072, "dur": 12,
    "args": {
      "External id": 142960,
      "cbid": 211, "correlation": 142960
    }
  },
  {
    "ph": "s", "id": 142960, "pid": 76337, "tid": -914061504, "ts": 1716454223619072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223674790, "dur": 4,
    "args": {
      "External id": 142962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 142962, "pid": 5, "tid": 7, "ts": 1716454223674790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619088, "dur": 6,
    "args": {
      "External id": 142962,
      "cbid": 211, "correlation": 142962
    }
  },
  {
    "ph": "s", "id": 142962, "pid": 76337, "tid": -914061504, "ts": 1716454223619088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619097, "dur": 0,
    "args": {
      "External id": 142963,
      "cbid": 51, "correlation": 142963
    }
  },
  {
    "ph": "s", "id": 142963, "pid": 76337, "tid": -914061504, "ts": 1716454223619097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223674795, "dur": 94,
    "args": {
      "External id": 142964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142964,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 142964, "pid": 5, "tid": 7, "ts": 1716454223674795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619098, "dur": 6,
    "args": {
      "External id": 142964,
      "cbid": 211, "correlation": 142964
    }
  },
  {
    "ph": "s", "id": 142964, "pid": 76337, "tid": -914061504, "ts": 1716454223619098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223674891, "dur": 15,
    "args": {
      "External id": 142969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142969,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142969, "pid": 5, "tid": 7, "ts": 1716454223674891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619125, "dur": 8,
    "args": {
      "External id": 142969,
      "cbid": 211, "correlation": 142969
    }
  },
  {
    "ph": "s", "id": 142969, "pid": 76337, "tid": -914061504, "ts": 1716454223619125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223674907, "dur": 85,
    "args": {
      "External id": 142978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 142978,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 142978, "pid": 5, "tid": 7, "ts": 1716454223674907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619207, "dur": 14,
    "args": {
      "External id": 142978,
      "cbid": 211, "correlation": 142978
    }
  },
  {
    "ph": "s", "id": 142978, "pid": 76337, "tid": -914061504, "ts": 1716454223619207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223674993, "dur": 31,
    "args": {
      "External id": 143000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143000,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143000, "pid": 5, "tid": 7, "ts": 1716454223674993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619264, "dur": 10,
    "args": {
      "External id": 143000,
      "cbid": 211, "correlation": 143000
    }
  },
  {
    "ph": "s", "id": 143000, "pid": 76337, "tid": -914061504, "ts": 1716454223619264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619351, "dur": 1,
    "args": {
      "External id": 143011,
      "cbid": 251, "correlation": 143011
    }
  },
  {
    "ph": "f", "id": 143011, "pid": 76337, "tid": -914061504, "ts": 1716454223619351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223675025, "dur": 163,
    "args": {
      "External id": 143012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143012,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143012, "pid": 5, "tid": 7, "ts": 1716454223675025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619356, "dur": 13,
    "args": {
      "External id": 143012,
      "cbid": 211, "correlation": 143012
    }
  },
  {
    "ph": "s", "id": 143012, "pid": 76337, "tid": -914061504, "ts": 1716454223619356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619425, "dur": 1,
    "args": {
      "External id": 143023,
      "cbid": 251, "correlation": 143023
    }
  },
  {
    "ph": "f", "id": 143023, "pid": 76337, "tid": -914061504, "ts": 1716454223619425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223675189, "dur": 159,
    "args": {
      "External id": 143024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143024,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143024, "pid": 5, "tid": 7, "ts": 1716454223675189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619429, "dur": 11,
    "args": {
      "External id": 143024,
      "cbid": 211, "correlation": 143024
    }
  },
  {
    "ph": "s", "id": 143024, "pid": 76337, "tid": -914061504, "ts": 1716454223619429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619492, "dur": 1,
    "args": {
      "External id": 143035,
      "cbid": 251, "correlation": 143035
    }
  },
  {
    "ph": "f", "id": 143035, "pid": 76337, "tid": -914061504, "ts": 1716454223619492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223675350, "dur": 160,
    "args": {
      "External id": 143036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143036,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143036, "pid": 5, "tid": 7, "ts": 1716454223675350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619497, "dur": 12,
    "args": {
      "External id": 143036,
      "cbid": 211, "correlation": 143036
    }
  },
  {
    "ph": "s", "id": 143036, "pid": 76337, "tid": -914061504, "ts": 1716454223619497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223675511, "dur": 339,
    "args": {
      "External id": 143061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143061,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143061, "pid": 5, "tid": 7, "ts": 1716454223675511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619580, "dur": 12,
    "args": {
      "External id": 143061,
      "cbid": 211, "correlation": 143061
    }
  },
  {
    "ph": "s", "id": 143061, "pid": 76337, "tid": -914061504, "ts": 1716454223619580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619680, "dur": 1,
    "args": {
      "External id": 143079,
      "cbid": 251, "correlation": 143079
    }
  },
  {
    "ph": "f", "id": 143079, "pid": 76337, "tid": -914061504, "ts": 1716454223619680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223675851, "dur": 166,
    "args": {
      "External id": 143081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143081,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143081, "pid": 5, "tid": 7, "ts": 1716454223675851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619685, "dur": 13,
    "args": {
      "External id": 143081,
      "cbid": 211, "correlation": 143081
    }
  },
  {
    "ph": "s", "id": 143081, "pid": 76337, "tid": -914061504, "ts": 1716454223619685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223676019, "dur": 19,
    "args": {
      "External id": 143089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143089,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143089, "pid": 5, "tid": 7, "ts": 1716454223676019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619757, "dur": 12,
    "args": {
      "External id": 143089,
      "cbid": 211, "correlation": 143089
    }
  },
  {
    "ph": "s", "id": 143089, "pid": 76337, "tid": -914061504, "ts": 1716454223619757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223676040, "dur": 28,
    "args": {
      "External id": 143097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143097,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143097, "pid": 5, "tid": 7, "ts": 1716454223676040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619795, "dur": 9,
    "args": {
      "External id": 143097,
      "cbid": 211, "correlation": 143097
    }
  },
  {
    "ph": "s", "id": 143097, "pid": 76337, "tid": -914061504, "ts": 1716454223619795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223676069, "dur": 18,
    "args": {
      "External id": 143108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143108, "pid": 5, "tid": 7, "ts": 1716454223676069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619866, "dur": 12,
    "args": {
      "External id": 143108,
      "cbid": 211, "correlation": 143108
    }
  },
  {
    "ph": "s", "id": 143108, "pid": 76337, "tid": -914061504, "ts": 1716454223619866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223676088, "dur": 16,
    "args": {
      "External id": 143130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143130,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143130, "pid": 5, "tid": 7, "ts": 1716454223676088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619897, "dur": 8,
    "args": {
      "External id": 143130,
      "cbid": 211, "correlation": 143130
    }
  },
  {
    "ph": "s", "id": 143130, "pid": 76337, "tid": -914061504, "ts": 1716454223619897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619989, "dur": 1,
    "args": {
      "External id": 143141,
      "cbid": 251, "correlation": 143141
    }
  },
  {
    "ph": "f", "id": 143141, "pid": 76337, "tid": -914061504, "ts": 1716454223619989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223676105, "dur": 88,
    "args": {
      "External id": 143142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143142,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 143142, "pid": 5, "tid": 7, "ts": 1716454223676105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223619994, "dur": 14,
    "args": {
      "External id": 143142,
      "cbid": 211, "correlation": 143142
    }
  },
  {
    "ph": "s", "id": 143142, "pid": 76337, "tid": -914061504, "ts": 1716454223619994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620064, "dur": 1,
    "args": {
      "External id": 143153,
      "cbid": 251, "correlation": 143153
    }
  },
  {
    "ph": "f", "id": 143153, "pid": 76337, "tid": -914061504, "ts": 1716454223620064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620068, "dur": 0,
    "args": {
      "External id": 143154,
      "cbid": 251, "correlation": 143154
    }
  },
  {
    "ph": "f", "id": 143154, "pid": 76337, "tid": -914061504, "ts": 1716454223620068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223676195, "dur": 12,
    "args": {
      "External id": 143155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143155,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143155, "pid": 5, "tid": 7, "ts": 1716454223676195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620069, "dur": 12,
    "args": {
      "External id": 143155,
      "cbid": 211, "correlation": 143155
    }
  },
  {
    "ph": "s", "id": 143155, "pid": 76337, "tid": -914061504, "ts": 1716454223620069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223676209, "dur": 6,
    "args": {
      "External id": 143157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143157,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143157, "pid": 5, "tid": 7, "ts": 1716454223676209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620083, "dur": 6,
    "args": {
      "External id": 143157,
      "cbid": 211, "correlation": 143157
    }
  },
  {
    "ph": "s", "id": 143157, "pid": 76337, "tid": -914061504, "ts": 1716454223620083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620140, "dur": 1,
    "args": {
      "External id": 143168,
      "cbid": 251, "correlation": 143168
    }
  },
  {
    "ph": "f", "id": 143168, "pid": 76337, "tid": -914061504, "ts": 1716454223620140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620144, "dur": 0,
    "args": {
      "External id": 143169,
      "cbid": 251, "correlation": 143169
    }
  },
  {
    "ph": "f", "id": 143169, "pid": 76337, "tid": -914061504, "ts": 1716454223620144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223676215, "dur": 8,
    "args": {
      "External id": 143170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143170,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143170, "pid": 5, "tid": 7, "ts": 1716454223676215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620145, "dur": 12,
    "args": {
      "External id": 143170,
      "cbid": 211, "correlation": 143170
    }
  },
  {
    "ph": "s", "id": 143170, "pid": 76337, "tid": -914061504, "ts": 1716454223620145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223676225, "dur": 4,
    "args": {
      "External id": 143172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143172,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143172, "pid": 5, "tid": 7, "ts": 1716454223676225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620159, "dur": 5,
    "args": {
      "External id": 143172,
      "cbid": 211, "correlation": 143172
    }
  },
  {
    "ph": "s", "id": 143172, "pid": 76337, "tid": -914061504, "ts": 1716454223620159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223676230, "dur": 55,
    "args": {
      "External id": 143197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143197,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143197, "pid": 5, "tid": 7, "ts": 1716454223676230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620236, "dur": 12,
    "args": {
      "External id": 143197,
      "cbid": 211, "correlation": 143197
    }
  },
  {
    "ph": "s", "id": 143197, "pid": 76337, "tid": -914061504, "ts": 1716454223620236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620333, "dur": 1,
    "args": {
      "External id": 143215,
      "cbid": 251, "correlation": 143215
    }
  },
  {
    "ph": "f", "id": 143215, "pid": 76337, "tid": -914061504, "ts": 1716454223620333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223676287, "dur": 91,
    "args": {
      "External id": 143217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143217,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 143217, "pid": 5, "tid": 7, "ts": 1716454223676287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620339, "dur": 13,
    "args": {
      "External id": 143217,
      "cbid": 211, "correlation": 143217
    }
  },
  {
    "ph": "s", "id": 143217, "pid": 76337, "tid": -914061504, "ts": 1716454223620339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223676380, "dur": 10,
    "args": {
      "External id": 143225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143225,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143225, "pid": 5, "tid": 7, "ts": 1716454223676380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620408, "dur": 12,
    "args": {
      "External id": 143225,
      "cbid": 211, "correlation": 143225
    }
  },
  {
    "ph": "s", "id": 143225, "pid": 76337, "tid": -914061504, "ts": 1716454223620408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223676391, "dur": 21,
    "args": {
      "External id": 143233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143233,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143233, "pid": 5, "tid": 7, "ts": 1716454223676391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620449, "dur": 10,
    "args": {
      "External id": 143233,
      "cbid": 211, "correlation": 143233
    }
  },
  {
    "ph": "s", "id": 143233, "pid": 76337, "tid": -914061504, "ts": 1716454223620449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223676413, "dur": 17,
    "args": {
      "External id": 143255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143255,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143255, "pid": 5, "tid": 7, "ts": 1716454223676413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620500, "dur": 11,
    "args": {
      "External id": 143255,
      "cbid": 211, "correlation": 143255
    }
  },
  {
    "ph": "s", "id": 143255, "pid": 76337, "tid": -914061504, "ts": 1716454223620500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620587, "dur": 1,
    "args": {
      "External id": 143271,
      "cbid": 251, "correlation": 143271
    }
  },
  {
    "ph": "f", "id": 143271, "pid": 76337, "tid": -914061504, "ts": 1716454223620587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620592, "dur": 0,
    "args": {
      "External id": 143273,
      "cbid": 251, "correlation": 143273
    }
  },
  {
    "ph": "f", "id": 143273, "pid": 76337, "tid": -914061504, "ts": 1716454223620592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223676432, "dur": 494,
    "args": {
      "External id": 143274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143274,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143274, "pid": 5, "tid": 7, "ts": 1716454223676432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620594, "dur": 13,
    "args": {
      "External id": 143274,
      "cbid": 211, "correlation": 143274
    }
  },
  {
    "ph": "s", "id": 143274, "pid": 76337, "tid": -914061504, "ts": 1716454223620594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223676927, "dur": 67,
    "args": {
      "External id": 143282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143282, "pid": 5, "tid": 7, "ts": 1716454223676927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620658, "dur": 13,
    "args": {
      "External id": 143282,
      "cbid": 211, "correlation": 143282
    }
  },
  {
    "ph": "s", "id": 143282, "pid": 76337, "tid": -914061504, "ts": 1716454223620658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223676995, "dur": 67,
    "args": {
      "External id": 143290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143290, "pid": 5, "tid": 7, "ts": 1716454223676995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620688, "dur": 8,
    "args": {
      "External id": 143290,
      "cbid": 211, "correlation": 143290
    }
  },
  {
    "ph": "s", "id": 143290, "pid": 76337, "tid": -914061504, "ts": 1716454223620688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620767, "dur": 1,
    "args": {
      "External id": 143306,
      "cbid": 251, "correlation": 143306
    }
  },
  {
    "ph": "f", "id": 143306, "pid": 76337, "tid": -914061504, "ts": 1716454223620767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223677064, "dur": 1,
    "args": {
      "External id": 143308,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143308,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15315890236119975
    }
  },
  {
    "ph": "f", "id": 143308, "pid": 5, "tid": 7, "ts": 1716454223677064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620772, "dur": 9,
    "args": {
      "External id": 143308,
      "cbid": 51, "correlation": 143308
    }
  },
  {
    "ph": "s", "id": 143308, "pid": 76337, "tid": -914061504, "ts": 1716454223620772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223677067, "dur": 272,
    "args": {
      "External id": 143309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143309,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 143309, "pid": 5, "tid": 7, "ts": 1716454223677067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620782, "dur": 11,
    "args": {
      "External id": 143309,
      "cbid": 211, "correlation": 143309
    }
  },
  {
    "ph": "s", "id": 143309, "pid": 76337, "tid": -914061504, "ts": 1716454223620782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223677340, "dur": 14,
    "args": {
      "External id": 143317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143317,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143317, "pid": 5, "tid": 7, "ts": 1716454223677340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620825, "dur": 10,
    "args": {
      "External id": 143317,
      "cbid": 211, "correlation": 143317
    }
  },
  {
    "ph": "s", "id": 143317, "pid": 76337, "tid": -914061504, "ts": 1716454223620825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223677356, "dur": 37,
    "args": {
      "External id": 143328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143328, "pid": 5, "tid": 7, "ts": 1716454223677356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620892, "dur": 12,
    "args": {
      "External id": 143328,
      "cbid": 211, "correlation": 143328
    }
  },
  {
    "ph": "s", "id": 143328, "pid": 76337, "tid": -914061504, "ts": 1716454223620892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620956, "dur": 0,
    "args": {
      "External id": 143340,
      "cbid": 317, "correlation": 143340
    }
  },
  {
    "ph": "f", "id": 143340, "pid": 76337, "tid": -914061504, "ts": 1716454223620956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620957, "dur": 0,
    "args": {
      "External id": 143341,
      "cbid": 203, "correlation": 143341
    }
  },
  {
    "ph": "f", "id": 143341, "pid": 76337, "tid": -914061504, "ts": 1716454223620957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620958, "dur": 0,
    "args": {
      "External id": 143342,
      "cbid": 205, "correlation": 143342
    }
  },
  {
    "ph": "f", "id": 143342, "pid": 76337, "tid": -914061504, "ts": 1716454223620958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677394, "dur": 12,
    "args": {
      "External id": 143346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143346,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143346, "pid": 5, "tid": 7, "ts": 1716454223677394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620981, "dur": 13,
    "args": {
      "External id": 143346,
      "cbid": 211, "correlation": 143346
    }
  },
  {
    "ph": "s", "id": 143346, "pid": 76337, "tid": -914061504, "ts": 1716454223620981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223677408, "dur": 4,
    "args": {
      "External id": 143348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143348,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 143348, "pid": 5, "tid": 7, "ts": 1716454223677408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223620998, "dur": 6,
    "args": {
      "External id": 143348,
      "cbid": 211, "correlation": 143348
    }
  },
  {
    "ph": "s", "id": 143348, "pid": 76337, "tid": -914061504, "ts": 1716454223620998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621007, "dur": 0,
    "args": {
      "External id": 143349,
      "cbid": 51, "correlation": 143349
    }
  },
  {
    "ph": "s", "id": 143349, "pid": 76337, "tid": -914061504, "ts": 1716454223621007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223677413, "dur": 98,
    "args": {
      "External id": 143350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143350,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 143350, "pid": 5, "tid": 7, "ts": 1716454223677413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621008, "dur": 6,
    "args": {
      "External id": 143350,
      "cbid": 211, "correlation": 143350
    }
  },
  {
    "ph": "s", "id": 143350, "pid": 76337, "tid": -914061504, "ts": 1716454223621008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223677512, "dur": 17,
    "args": {
      "External id": 143355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143355, "pid": 5, "tid": 7, "ts": 1716454223677512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621036, "dur": 8,
    "args": {
      "External id": 143355,
      "cbid": 211, "correlation": 143355
    }
  },
  {
    "ph": "s", "id": 143355, "pid": 76337, "tid": -914061504, "ts": 1716454223621036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223677531, "dur": 12,
    "args": {
      "External id": 143363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143363,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143363, "pid": 5, "tid": 7, "ts": 1716454223677531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621069, "dur": 8,
    "args": {
      "External id": 143363,
      "cbid": 211, "correlation": 143363
    }
  },
  {
    "ph": "s", "id": 143363, "pid": 76337, "tid": -914061504, "ts": 1716454223621069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621138, "dur": 0,
    "args": {
      "External id": 143373,
      "cbid": 317, "correlation": 143373
    }
  },
  {
    "ph": "f", "id": 143373, "pid": 76337, "tid": -914061504, "ts": 1716454223621138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621139, "dur": 0,
    "args": {
      "External id": 143374,
      "cbid": 203, "correlation": 143374
    }
  },
  {
    "ph": "f", "id": 143374, "pid": 76337, "tid": -914061504, "ts": 1716454223621139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621140, "dur": 0,
    "args": {
      "External id": 143375,
      "cbid": 205, "correlation": 143375
    }
  },
  {
    "ph": "f", "id": 143375, "pid": 76337, "tid": -914061504, "ts": 1716454223621140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677544, "dur": 12,
    "args": {
      "External id": 143379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143379,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143379, "pid": 5, "tid": 7, "ts": 1716454223677544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621154, "dur": 14,
    "args": {
      "External id": 143379,
      "cbid": 211, "correlation": 143379
    }
  },
  {
    "ph": "s", "id": 143379, "pid": 76337, "tid": -914061504, "ts": 1716454223621154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677558, "dur": 163,
    "args": {
      "External id": 143381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143381,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143381, "pid": 5, "tid": 7, "ts": 1716454223677558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621170, "dur": 5,
    "args": {
      "External id": 143381,
      "cbid": 211, "correlation": 143381
    }
  },
  {
    "ph": "s", "id": 143381, "pid": 76337, "tid": -914061504, "ts": 1716454223621170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223677723, "dur": 1,
    "args": {
      "External id": 143383,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143383,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 143383, "pid": 5, "tid": 7, "ts": 1716454223677723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621181, "dur": 6,
    "args": {
      "External id": 143383,
      "cbid": 51, "correlation": 143383
    }
  },
  {
    "ph": "s", "id": 143383, "pid": 76337, "tid": -914061504, "ts": 1716454223621181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223677726, "dur": 198,
    "args": {
      "External id": 143384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143384,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 143384, "pid": 5, "tid": 7, "ts": 1716454223677726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621189, "dur": 8,
    "args": {
      "External id": 143384,
      "cbid": 211, "correlation": 143384
    }
  },
  {
    "ph": "s", "id": 143384, "pid": 76337, "tid": -914061504, "ts": 1716454223621189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677925, "dur": 7,
    "args": {
      "External id": 143386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143386,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143386, "pid": 5, "tid": 7, "ts": 1716454223677925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621201, "dur": 5,
    "args": {
      "External id": 143386,
      "cbid": 211, "correlation": 143386
    }
  },
  {
    "ph": "s", "id": 143386, "pid": 76337, "tid": -914061504, "ts": 1716454223621201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223677933, "dur": 6,
    "args": {
      "External id": 143392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143392,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143392, "pid": 5, "tid": 7, "ts": 1716454223677933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621230, "dur": 8,
    "args": {
      "External id": 143392,
      "cbid": 211, "correlation": 143392
    }
  },
  {
    "ph": "s", "id": 143392, "pid": 76337, "tid": -914061504, "ts": 1716454223621230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677941, "dur": 11,
    "args": {
      "External id": 143412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143412,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 143412, "pid": 5, "tid": 7, "ts": 1716454223677941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621323, "dur": 12,
    "args": {
      "External id": 143412,
      "cbid": 211, "correlation": 143412
    }
  },
  {
    "ph": "s", "id": 143412, "pid": 76337, "tid": -914061504, "ts": 1716454223621323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223677953, "dur": 4,
    "args": {
      "External id": 143424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143424,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 143424, "pid": 5, "tid": 7, "ts": 1716454223677953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621345, "dur": 6,
    "args": {
      "External id": 143424,
      "cbid": 211, "correlation": 143424
    }
  },
  {
    "ph": "s", "id": 143424, "pid": 76337, "tid": -914061504, "ts": 1716454223621345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223677959, "dur": 9,
    "args": {
      "External id": 143427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143427,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143427, "pid": 5, "tid": 7, "ts": 1716454223677959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621365, "dur": 6,
    "args": {
      "External id": 143427,
      "cbid": 211, "correlation": 143427
    }
  },
  {
    "ph": "s", "id": 143427, "pid": 76337, "tid": -914061504, "ts": 1716454223621365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223677969, "dur": 5,
    "args": {
      "External id": 143436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143436,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143436, "pid": 5, "tid": 7, "ts": 1716454223677969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621404, "dur": 10,
    "args": {
      "External id": 143436,
      "cbid": 211, "correlation": 143436
    }
  },
  {
    "ph": "s", "id": 143436, "pid": 76337, "tid": -914061504, "ts": 1716454223621404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621455, "dur": 0,
    "args": {
      "External id": 143446,
      "cbid": 317, "correlation": 143446
    }
  },
  {
    "ph": "f", "id": 143446, "pid": 76337, "tid": -914061504, "ts": 1716454223621455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621456, "dur": 0,
    "args": {
      "External id": 143447,
      "cbid": 203, "correlation": 143447
    }
  },
  {
    "ph": "f", "id": 143447, "pid": 76337, "tid": -914061504, "ts": 1716454223621456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621457, "dur": 0,
    "args": {
      "External id": 143448,
      "cbid": 205, "correlation": 143448
    }
  },
  {
    "ph": "f", "id": 143448, "pid": 76337, "tid": -914061504, "ts": 1716454223621457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677975, "dur": 5,
    "args": {
      "External id": 143452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143452, "pid": 5, "tid": 7, "ts": 1716454223677975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621472, "dur": 11,
    "args": {
      "External id": 143452,
      "cbid": 211, "correlation": 143452
    }
  },
  {
    "ph": "s", "id": 143452, "pid": 76337, "tid": -914061504, "ts": 1716454223621472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223677982, "dur": 163,
    "args": {
      "External id": 143454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143454, "pid": 5, "tid": 7, "ts": 1716454223677982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621486, "dur": 5,
    "args": {
      "External id": 143454,
      "cbid": 211, "correlation": 143454
    }
  },
  {
    "ph": "s", "id": 143454, "pid": 76337, "tid": -914061504, "ts": 1716454223621486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223678147, "dur": 1,
    "args": {
      "External id": 143456,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143456,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 143456, "pid": 5, "tid": 7, "ts": 1716454223678147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621497, "dur": 6,
    "args": {
      "External id": 143456,
      "cbid": 51, "correlation": 143456
    }
  },
  {
    "ph": "s", "id": 143456, "pid": 76337, "tid": -914061504, "ts": 1716454223621497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223678151, "dur": 269,
    "args": {
      "External id": 143457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143457,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143457, "pid": 5, "tid": 7, "ts": 1716454223678151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621504, "dur": 6,
    "args": {
      "External id": 143457,
      "cbid": 211, "correlation": 143457
    }
  },
  {
    "ph": "s", "id": 143457, "pid": 76337, "tid": -914061504, "ts": 1716454223621504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678422, "dur": 6,
    "args": {
      "External id": 143459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143459,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143459, "pid": 5, "tid": 7, "ts": 1716454223678422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621516, "dur": 5,
    "args": {
      "External id": 143459,
      "cbid": 211, "correlation": 143459
    }
  },
  {
    "ph": "s", "id": 143459, "pid": 76337, "tid": -914061504, "ts": 1716454223621516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223678429, "dur": 6,
    "args": {
      "External id": 143465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143465,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143465, "pid": 5, "tid": 7, "ts": 1716454223678429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621544, "dur": 8,
    "args": {
      "External id": 143465,
      "cbid": 211, "correlation": 143465
    }
  },
  {
    "ph": "s", "id": 143465, "pid": 76337, "tid": -914061504, "ts": 1716454223621544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223678436, "dur": 4,
    "args": {
      "External id": 143473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143473,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 143473, "pid": 5, "tid": 7, "ts": 1716454223678436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621587, "dur": 9,
    "args": {
      "External id": 143473,
      "cbid": 211, "correlation": 143473
    }
  },
  {
    "ph": "s", "id": 143473, "pid": 76337, "tid": -914061504, "ts": 1716454223621587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621653, "dur": 1,
    "args": {
      "External id": 143489,
      "cbid": 251, "correlation": 143489
    }
  },
  {
    "ph": "f", "id": 143489, "pid": 76337, "tid": -914061504, "ts": 1716454223621653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621659, "dur": 0,
    "args": {
      "External id": 143491,
      "cbid": 251, "correlation": 143491
    }
  },
  {
    "ph": "f", "id": 143491, "pid": 76337, "tid": -914061504, "ts": 1716454223621659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223678441, "dur": 13,
    "args": {
      "External id": 143492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143492,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143492, "pid": 5, "tid": 7, "ts": 1716454223678441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621661, "dur": 11,
    "args": {
      "External id": 143492,
      "cbid": 211, "correlation": 143492
    }
  },
  {
    "ph": "s", "id": 143492, "pid": 76337, "tid": -914061504, "ts": 1716454223621661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223678456, "dur": 5,
    "args": {
      "External id": 143494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143494,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143494, "pid": 5, "tid": 7, "ts": 1716454223678456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621674, "dur": 5,
    "args": {
      "External id": 143494,
      "cbid": 211, "correlation": 143494
    }
  },
  {
    "ph": "s", "id": 143494, "pid": 76337, "tid": -914061504, "ts": 1716454223621674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223678462, "dur": 6,
    "args": {
      "External id": 143504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143504,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143504, "pid": 5, "tid": 7, "ts": 1716454223678462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621730, "dur": 12,
    "args": {
      "External id": 143504,
      "cbid": 211, "correlation": 143504
    }
  },
  {
    "ph": "s", "id": 143504, "pid": 76337, "tid": -914061504, "ts": 1716454223621730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678469, "dur": 10,
    "args": {
      "External id": 143524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143524,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 143524, "pid": 5, "tid": 7, "ts": 1716454223678469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621796, "dur": 11,
    "args": {
      "External id": 143524,
      "cbid": 211, "correlation": 143524
    }
  },
  {
    "ph": "s", "id": 143524, "pid": 76337, "tid": -914061504, "ts": 1716454223621796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223678480, "dur": 4,
    "args": {
      "External id": 143536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143536,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 143536, "pid": 5, "tid": 7, "ts": 1716454223678480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621817, "dur": 6,
    "args": {
      "External id": 143536,
      "cbid": 211, "correlation": 143536
    }
  },
  {
    "ph": "s", "id": 143536, "pid": 76337, "tid": -914061504, "ts": 1716454223621817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223678485, "dur": 7,
    "args": {
      "External id": 143539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143539,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143539, "pid": 5, "tid": 7, "ts": 1716454223678485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621837, "dur": 6,
    "args": {
      "External id": 143539,
      "cbid": 211, "correlation": 143539
    }
  },
  {
    "ph": "s", "id": 143539, "pid": 76337, "tid": -914061504, "ts": 1716454223621837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223678494, "dur": 5,
    "args": {
      "External id": 143548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143548,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143548, "pid": 5, "tid": 7, "ts": 1716454223678494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621877, "dur": 10,
    "args": {
      "External id": 143548,
      "cbid": 211, "correlation": 143548
    }
  },
  {
    "ph": "s", "id": 143548, "pid": 76337, "tid": -914061504, "ts": 1716454223621877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621940, "dur": 0,
    "args": {
      "External id": 143558,
      "cbid": 317, "correlation": 143558
    }
  },
  {
    "ph": "f", "id": 143558, "pid": 76337, "tid": -914061504, "ts": 1716454223621940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621941, "dur": 0,
    "args": {
      "External id": 143559,
      "cbid": 203, "correlation": 143559
    }
  },
  {
    "ph": "f", "id": 143559, "pid": 76337, "tid": -914061504, "ts": 1716454223621941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621942, "dur": 0,
    "args": {
      "External id": 143560,
      "cbid": 205, "correlation": 143560
    }
  },
  {
    "ph": "f", "id": 143560, "pid": 76337, "tid": -914061504, "ts": 1716454223621942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678500, "dur": 5,
    "args": {
      "External id": 143564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143564,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143564, "pid": 5, "tid": 7, "ts": 1716454223678500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621955, "dur": 13,
    "args": {
      "External id": 143564,
      "cbid": 211, "correlation": 143564
    }
  },
  {
    "ph": "s", "id": 143564, "pid": 76337, "tid": -914061504, "ts": 1716454223621955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678506, "dur": 163,
    "args": {
      "External id": 143566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143566,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143566, "pid": 5, "tid": 7, "ts": 1716454223678506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621971, "dur": 14,
    "args": {
      "External id": 143566,
      "cbid": 211, "correlation": 143566
    }
  },
  {
    "ph": "s", "id": 143566, "pid": 76337, "tid": -914061504, "ts": 1716454223621971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223678671, "dur": 1,
    "args": {
      "External id": 143568,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143568,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 143568, "pid": 5, "tid": 7, "ts": 1716454223678671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621991, "dur": 6,
    "args": {
      "External id": 143568,
      "cbid": 51, "correlation": 143568
    }
  },
  {
    "ph": "s", "id": 143568, "pid": 76337, "tid": -914061504, "ts": 1716454223621991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223678675, "dur": 260,
    "args": {
      "External id": 143569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143569,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143569, "pid": 5, "tid": 7, "ts": 1716454223678675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223621998, "dur": 6,
    "args": {
      "External id": 143569,
      "cbid": 211, "correlation": 143569
    }
  },
  {
    "ph": "s", "id": 143569, "pid": 76337, "tid": -914061504, "ts": 1716454223621998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678937, "dur": 6,
    "args": {
      "External id": 143571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143571,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143571, "pid": 5, "tid": 7, "ts": 1716454223678937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622008, "dur": 5,
    "args": {
      "External id": 143571,
      "cbid": 211, "correlation": 143571
    }
  },
  {
    "ph": "s", "id": 143571, "pid": 76337, "tid": -914061504, "ts": 1716454223622008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223678944, "dur": 6,
    "args": {
      "External id": 143577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143577, "pid": 5, "tid": 7, "ts": 1716454223678944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622037, "dur": 9,
    "args": {
      "External id": 143577,
      "cbid": 211, "correlation": 143577
    }
  },
  {
    "ph": "s", "id": 143577, "pid": 76337, "tid": -914061504, "ts": 1716454223622037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223678952, "dur": 5,
    "args": {
      "External id": 143585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143585,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143585, "pid": 5, "tid": 7, "ts": 1716454223678952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622072, "dur": 8,
    "args": {
      "External id": 143585,
      "cbid": 211, "correlation": 143585
    }
  },
  {
    "ph": "s", "id": 143585, "pid": 76337, "tid": -914061504, "ts": 1716454223622072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223678958, "dur": 4,
    "args": {
      "External id": 143593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143593,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143593, "pid": 5, "tid": 7, "ts": 1716454223678958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622101, "dur": 8,
    "args": {
      "External id": 143593,
      "cbid": 211, "correlation": 143593
    }
  },
  {
    "ph": "s", "id": 143593, "pid": 76337, "tid": -914061504, "ts": 1716454223622101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678964, "dur": 10,
    "args": {
      "External id": 143613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143613,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 143613, "pid": 5, "tid": 7, "ts": 1716454223678964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622178, "dur": 12,
    "args": {
      "External id": 143613,
      "cbid": 211, "correlation": 143613
    }
  },
  {
    "ph": "s", "id": 143613, "pid": 76337, "tid": -914061504, "ts": 1716454223622178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223678975, "dur": 3,
    "args": {
      "External id": 143625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143625,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 143625, "pid": 5, "tid": 7, "ts": 1716454223678975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622199, "dur": 6,
    "args": {
      "External id": 143625,
      "cbid": 211, "correlation": 143625
    }
  },
  {
    "ph": "s", "id": 143625, "pid": 76337, "tid": -914061504, "ts": 1716454223622199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223678980, "dur": 6,
    "args": {
      "External id": 143628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143628, "pid": 5, "tid": 7, "ts": 1716454223678980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622217, "dur": 7,
    "args": {
      "External id": 143628,
      "cbid": 211, "correlation": 143628
    }
  },
  {
    "ph": "s", "id": 143628, "pid": 76337, "tid": -914061504, "ts": 1716454223622217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223678988, "dur": 5,
    "args": {
      "External id": 143637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143637,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143637, "pid": 5, "tid": 7, "ts": 1716454223678988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622256, "dur": 9,
    "args": {
      "External id": 143637,
      "cbid": 211, "correlation": 143637
    }
  },
  {
    "ph": "s", "id": 143637, "pid": 76337, "tid": -914061504, "ts": 1716454223622256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622309, "dur": 0,
    "args": {
      "External id": 143647,
      "cbid": 317, "correlation": 143647
    }
  },
  {
    "ph": "f", "id": 143647, "pid": 76337, "tid": -914061504, "ts": 1716454223622309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622310, "dur": 0,
    "args": {
      "External id": 143648,
      "cbid": 203, "correlation": 143648
    }
  },
  {
    "ph": "f", "id": 143648, "pid": 76337, "tid": -914061504, "ts": 1716454223622310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622310, "dur": 0,
    "args": {
      "External id": 143649,
      "cbid": 205, "correlation": 143649
    }
  },
  {
    "ph": "f", "id": 143649, "pid": 76337, "tid": -914061504, "ts": 1716454223622310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223678993, "dur": 5,
    "args": {
      "External id": 143653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143653,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143653, "pid": 5, "tid": 7, "ts": 1716454223678993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622324, "dur": 11,
    "args": {
      "External id": 143653,
      "cbid": 211, "correlation": 143653
    }
  },
  {
    "ph": "s", "id": 143653, "pid": 76337, "tid": -914061504, "ts": 1716454223622324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679000, "dur": 163,
    "args": {
      "External id": 143655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143655, "pid": 5, "tid": 7, "ts": 1716454223679000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622338, "dur": 5,
    "args": {
      "External id": 143655,
      "cbid": 211, "correlation": 143655
    }
  },
  {
    "ph": "s", "id": 143655, "pid": 76337, "tid": -914061504, "ts": 1716454223622338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223679165, "dur": 1,
    "args": {
      "External id": 143657,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143657,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 143657, "pid": 5, "tid": 7, "ts": 1716454223679165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622349, "dur": 6,
    "args": {
      "External id": 143657,
      "cbid": 51, "correlation": 143657
    }
  },
  {
    "ph": "s", "id": 143657, "pid": 76337, "tid": -914061504, "ts": 1716454223622349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223679169, "dur": 259,
    "args": {
      "External id": 143658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143658,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143658, "pid": 5, "tid": 7, "ts": 1716454223679169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622356, "dur": 6,
    "args": {
      "External id": 143658,
      "cbid": 211, "correlation": 143658
    }
  },
  {
    "ph": "s", "id": 143658, "pid": 76337, "tid": -914061504, "ts": 1716454223622356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679429, "dur": 6,
    "args": {
      "External id": 143660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143660, "pid": 5, "tid": 7, "ts": 1716454223679429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622365, "dur": 5,
    "args": {
      "External id": 143660,
      "cbid": 211, "correlation": 143660
    }
  },
  {
    "ph": "s", "id": 143660, "pid": 76337, "tid": -914061504, "ts": 1716454223622365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223679436, "dur": 6,
    "args": {
      "External id": 143666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143666,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143666, "pid": 5, "tid": 7, "ts": 1716454223679436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622393, "dur": 8,
    "args": {
      "External id": 143666,
      "cbid": 211, "correlation": 143666
    }
  },
  {
    "ph": "s", "id": 143666, "pid": 76337, "tid": -914061504, "ts": 1716454223622393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223679443, "dur": 4,
    "args": {
      "External id": 143674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143674,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 143674, "pid": 5, "tid": 7, "ts": 1716454223679443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622437, "dur": 9,
    "args": {
      "External id": 143674,
      "cbid": 211, "correlation": 143674
    }
  },
  {
    "ph": "s", "id": 143674, "pid": 76337, "tid": -914061504, "ts": 1716454223622437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622499, "dur": 1,
    "args": {
      "External id": 143690,
      "cbid": 251, "correlation": 143690
    }
  },
  {
    "ph": "f", "id": 143690, "pid": 76337, "tid": -914061504, "ts": 1716454223622499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622504, "dur": 0,
    "args": {
      "External id": 143692,
      "cbid": 251, "correlation": 143692
    }
  },
  {
    "ph": "f", "id": 143692, "pid": 76337, "tid": -914061504, "ts": 1716454223622504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223679448, "dur": 10,
    "args": {
      "External id": 143693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143693,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143693, "pid": 5, "tid": 7, "ts": 1716454223679448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622506, "dur": 11,
    "args": {
      "External id": 143693,
      "cbid": 211, "correlation": 143693
    }
  },
  {
    "ph": "s", "id": 143693, "pid": 76337, "tid": -914061504, "ts": 1716454223622506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223679459, "dur": 4,
    "args": {
      "External id": 143695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143695,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143695, "pid": 5, "tid": 7, "ts": 1716454223679459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622520, "dur": 6,
    "args": {
      "External id": 143695,
      "cbid": 211, "correlation": 143695
    }
  },
  {
    "ph": "s", "id": 143695, "pid": 76337, "tid": -914061504, "ts": 1716454223622520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223679464, "dur": 6,
    "args": {
      "External id": 143705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143705, "pid": 5, "tid": 7, "ts": 1716454223679464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622576, "dur": 12,
    "args": {
      "External id": 143705,
      "cbid": 211, "correlation": 143705
    }
  },
  {
    "ph": "s", "id": 143705, "pid": 76337, "tid": -914061504, "ts": 1716454223622576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679471, "dur": 10,
    "args": {
      "External id": 143725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143725,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 143725, "pid": 5, "tid": 7, "ts": 1716454223679471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622641, "dur": 10,
    "args": {
      "External id": 143725,
      "cbid": 211, "correlation": 143725
    }
  },
  {
    "ph": "s", "id": 143725, "pid": 76337, "tid": -914061504, "ts": 1716454223622641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223679482, "dur": 4,
    "args": {
      "External id": 143737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143737,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 143737, "pid": 5, "tid": 7, "ts": 1716454223679482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622662, "dur": 6,
    "args": {
      "External id": 143737,
      "cbid": 211, "correlation": 143737
    }
  },
  {
    "ph": "s", "id": 143737, "pid": 76337, "tid": -914061504, "ts": 1716454223622662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223679488, "dur": 7,
    "args": {
      "External id": 143740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143740, "pid": 5, "tid": 7, "ts": 1716454223679488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622681, "dur": 6,
    "args": {
      "External id": 143740,
      "cbid": 211, "correlation": 143740
    }
  },
  {
    "ph": "s", "id": 143740, "pid": 76337, "tid": -914061504, "ts": 1716454223622681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223679496, "dur": 5,
    "args": {
      "External id": 143749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143749,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143749, "pid": 5, "tid": 7, "ts": 1716454223679496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622721, "dur": 11,
    "args": {
      "External id": 143749,
      "cbid": 211, "correlation": 143749
    }
  },
  {
    "ph": "s", "id": 143749, "pid": 76337, "tid": -914061504, "ts": 1716454223622721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622785, "dur": 0,
    "args": {
      "External id": 143759,
      "cbid": 317, "correlation": 143759
    }
  },
  {
    "ph": "f", "id": 143759, "pid": 76337, "tid": -914061504, "ts": 1716454223622785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622786, "dur": 0,
    "args": {
      "External id": 143760,
      "cbid": 203, "correlation": 143760
    }
  },
  {
    "ph": "f", "id": 143760, "pid": 76337, "tid": -914061504, "ts": 1716454223622786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622787, "dur": 0,
    "args": {
      "External id": 143761,
      "cbid": 205, "correlation": 143761
    }
  },
  {
    "ph": "f", "id": 143761, "pid": 76337, "tid": -914061504, "ts": 1716454223622787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679502, "dur": 5,
    "args": {
      "External id": 143765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143765,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143765, "pid": 5, "tid": 7, "ts": 1716454223679502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622800, "dur": 12,
    "args": {
      "External id": 143765,
      "cbid": 211, "correlation": 143765
    }
  },
  {
    "ph": "s", "id": 143765, "pid": 76337, "tid": -914061504, "ts": 1716454223622800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679508, "dur": 163,
    "args": {
      "External id": 143767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143767,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143767, "pid": 5, "tid": 7, "ts": 1716454223679508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622814, "dur": 5,
    "args": {
      "External id": 143767,
      "cbid": 211, "correlation": 143767
    }
  },
  {
    "ph": "s", "id": 143767, "pid": 76337, "tid": -914061504, "ts": 1716454223622814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223679673, "dur": 1,
    "args": {
      "External id": 143769,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143769,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 143769, "pid": 5, "tid": 7, "ts": 1716454223679673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622825, "dur": 7,
    "args": {
      "External id": 143769,
      "cbid": 51, "correlation": 143769
    }
  },
  {
    "ph": "s", "id": 143769, "pid": 76337, "tid": -914061504, "ts": 1716454223622825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223679677, "dur": 260,
    "args": {
      "External id": 143770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143770,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143770, "pid": 5, "tid": 7, "ts": 1716454223679677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622833, "dur": 6,
    "args": {
      "External id": 143770,
      "cbid": 211, "correlation": 143770
    }
  },
  {
    "ph": "s", "id": 143770, "pid": 76337, "tid": -914061504, "ts": 1716454223622833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679938, "dur": 6,
    "args": {
      "External id": 143772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143772, "pid": 5, "tid": 7, "ts": 1716454223679938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622843, "dur": 5,
    "args": {
      "External id": 143772,
      "cbid": 211, "correlation": 143772
    }
  },
  {
    "ph": "s", "id": 143772, "pid": 76337, "tid": -914061504, "ts": 1716454223622843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223679945, "dur": 6,
    "args": {
      "External id": 143778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143778,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143778, "pid": 5, "tid": 7, "ts": 1716454223679945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622871, "dur": 9,
    "args": {
      "External id": 143778,
      "cbid": 211, "correlation": 143778
    }
  },
  {
    "ph": "s", "id": 143778, "pid": 76337, "tid": -914061504, "ts": 1716454223622871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223679953, "dur": 5,
    "args": {
      "External id": 143786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143786,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143786, "pid": 5, "tid": 7, "ts": 1716454223679953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622904, "dur": 8,
    "args": {
      "External id": 143786,
      "cbid": 211, "correlation": 143786
    }
  },
  {
    "ph": "s", "id": 143786, "pid": 76337, "tid": -914061504, "ts": 1716454223622904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223679959, "dur": 4,
    "args": {
      "External id": 143794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143794,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143794, "pid": 5, "tid": 7, "ts": 1716454223679959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223622934, "dur": 8,
    "args": {
      "External id": 143794,
      "cbid": 211, "correlation": 143794
    }
  },
  {
    "ph": "s", "id": 143794, "pid": 76337, "tid": -914061504, "ts": 1716454223622934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679965, "dur": 10,
    "args": {
      "External id": 143814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143814,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 143814, "pid": 5, "tid": 7, "ts": 1716454223679965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623046, "dur": 13,
    "args": {
      "External id": 143814,
      "cbid": 211, "correlation": 143814
    }
  },
  {
    "ph": "s", "id": 143814, "pid": 76337, "tid": -914061504, "ts": 1716454223623046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223679976, "dur": 4,
    "args": {
      "External id": 143826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143826,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 143826, "pid": 5, "tid": 7, "ts": 1716454223679976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623068, "dur": 6,
    "args": {
      "External id": 143826,
      "cbid": 211, "correlation": 143826
    }
  },
  {
    "ph": "s", "id": 143826, "pid": 76337, "tid": -914061504, "ts": 1716454223623068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223679981, "dur": 7,
    "args": {
      "External id": 143829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143829, "pid": 5, "tid": 7, "ts": 1716454223679981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623086, "dur": 6,
    "args": {
      "External id": 143829,
      "cbid": 211, "correlation": 143829
    }
  },
  {
    "ph": "s", "id": 143829, "pid": 76337, "tid": -914061504, "ts": 1716454223623086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223679989, "dur": 5,
    "args": {
      "External id": 143838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143838,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143838, "pid": 5, "tid": 7, "ts": 1716454223679989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623124, "dur": 10,
    "args": {
      "External id": 143838,
      "cbid": 211, "correlation": 143838
    }
  },
  {
    "ph": "s", "id": 143838, "pid": 76337, "tid": -914061504, "ts": 1716454223623124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623177, "dur": 0,
    "args": {
      "External id": 143848,
      "cbid": 317, "correlation": 143848
    }
  },
  {
    "ph": "f", "id": 143848, "pid": 76337, "tid": -914061504, "ts": 1716454223623177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623177, "dur": 0,
    "args": {
      "External id": 143849,
      "cbid": 203, "correlation": 143849
    }
  },
  {
    "ph": "f", "id": 143849, "pid": 76337, "tid": -914061504, "ts": 1716454223623177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623178, "dur": 0,
    "args": {
      "External id": 143850,
      "cbid": 205, "correlation": 143850
    }
  },
  {
    "ph": "f", "id": 143850, "pid": 76337, "tid": -914061504, "ts": 1716454223623178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223679995, "dur": 5,
    "args": {
      "External id": 143854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143854,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143854, "pid": 5, "tid": 7, "ts": 1716454223679995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623192, "dur": 11,
    "args": {
      "External id": 143854,
      "cbid": 211, "correlation": 143854
    }
  },
  {
    "ph": "s", "id": 143854, "pid": 76337, "tid": -914061504, "ts": 1716454223623192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680001, "dur": 162,
    "args": {
      "External id": 143856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143856,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143856, "pid": 5, "tid": 7, "ts": 1716454223680001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623206, "dur": 5,
    "args": {
      "External id": 143856,
      "cbid": 211, "correlation": 143856
    }
  },
  {
    "ph": "s", "id": 143856, "pid": 76337, "tid": -914061504, "ts": 1716454223623206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223680165, "dur": 1,
    "args": {
      "External id": 143858,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143858,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 143858, "pid": 5, "tid": 7, "ts": 1716454223680165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623216, "dur": 7,
    "args": {
      "External id": 143858,
      "cbid": 51, "correlation": 143858
    }
  },
  {
    "ph": "s", "id": 143858, "pid": 76337, "tid": -914061504, "ts": 1716454223623216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223680169, "dur": 260,
    "args": {
      "External id": 143859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143859,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143859, "pid": 5, "tid": 7, "ts": 1716454223680169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623224, "dur": 6,
    "args": {
      "External id": 143859,
      "cbid": 211, "correlation": 143859
    }
  },
  {
    "ph": "s", "id": 143859, "pid": 76337, "tid": -914061504, "ts": 1716454223623224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680430, "dur": 6,
    "args": {
      "External id": 143861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143861,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143861, "pid": 5, "tid": 7, "ts": 1716454223680430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623235, "dur": 5,
    "args": {
      "External id": 143861,
      "cbid": 211, "correlation": 143861
    }
  },
  {
    "ph": "s", "id": 143861, "pid": 76337, "tid": -914061504, "ts": 1716454223623235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223680438, "dur": 6,
    "args": {
      "External id": 143867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143867, "pid": 5, "tid": 7, "ts": 1716454223680438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623264, "dur": 9,
    "args": {
      "External id": 143867,
      "cbid": 211, "correlation": 143867
    }
  },
  {
    "ph": "s", "id": 143867, "pid": 76337, "tid": -914061504, "ts": 1716454223623264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223680445, "dur": 3,
    "args": {
      "External id": 143875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143875,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 143875, "pid": 5, "tid": 7, "ts": 1716454223680445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623307, "dur": 9,
    "args": {
      "External id": 143875,
      "cbid": 211, "correlation": 143875
    }
  },
  {
    "ph": "s", "id": 143875, "pid": 76337, "tid": -914061504, "ts": 1716454223623307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623369, "dur": 1,
    "args": {
      "External id": 143891,
      "cbid": 251, "correlation": 143891
    }
  },
  {
    "ph": "f", "id": 143891, "pid": 76337, "tid": -914061504, "ts": 1716454223623369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623374, "dur": 0,
    "args": {
      "External id": 143893,
      "cbid": 251, "correlation": 143893
    }
  },
  {
    "ph": "f", "id": 143893, "pid": 76337, "tid": -914061504, "ts": 1716454223623374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223680450, "dur": 10,
    "args": {
      "External id": 143894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143894,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143894, "pid": 5, "tid": 7, "ts": 1716454223680450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623376, "dur": 11,
    "args": {
      "External id": 143894,
      "cbid": 211, "correlation": 143894
    }
  },
  {
    "ph": "s", "id": 143894, "pid": 76337, "tid": -914061504, "ts": 1716454223623376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223680461, "dur": 4,
    "args": {
      "External id": 143896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143896,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143896, "pid": 5, "tid": 7, "ts": 1716454223680461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623389, "dur": 5,
    "args": {
      "External id": 143896,
      "cbid": 211, "correlation": 143896
    }
  },
  {
    "ph": "s", "id": 143896, "pid": 76337, "tid": -914061504, "ts": 1716454223623389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223680466, "dur": 6,
    "args": {
      "External id": 143906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143906,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143906, "pid": 5, "tid": 7, "ts": 1716454223680466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623445, "dur": 12,
    "args": {
      "External id": 143906,
      "cbid": 211, "correlation": 143906
    }
  },
  {
    "ph": "s", "id": 143906, "pid": 76337, "tid": -914061504, "ts": 1716454223623445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680473, "dur": 10,
    "args": {
      "External id": 143926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143926,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 143926, "pid": 5, "tid": 7, "ts": 1716454223680473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623510, "dur": 11,
    "args": {
      "External id": 143926,
      "cbid": 211, "correlation": 143926
    }
  },
  {
    "ph": "s", "id": 143926, "pid": 76337, "tid": -914061504, "ts": 1716454223623510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223680484, "dur": 4,
    "args": {
      "External id": 143938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143938,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 143938, "pid": 5, "tid": 7, "ts": 1716454223680484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623531, "dur": 7,
    "args": {
      "External id": 143938,
      "cbid": 211, "correlation": 143938
    }
  },
  {
    "ph": "s", "id": 143938, "pid": 76337, "tid": -914061504, "ts": 1716454223623531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223680489, "dur": 7,
    "args": {
      "External id": 143941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143941,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143941, "pid": 5, "tid": 7, "ts": 1716454223680489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623551, "dur": 6,
    "args": {
      "External id": 143941,
      "cbid": 211, "correlation": 143941
    }
  },
  {
    "ph": "s", "id": 143941, "pid": 76337, "tid": -914061504, "ts": 1716454223623551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223680497, "dur": 5,
    "args": {
      "External id": 143950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143950,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143950, "pid": 5, "tid": 7, "ts": 1716454223680497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623591, "dur": 9,
    "args": {
      "External id": 143950,
      "cbid": 211, "correlation": 143950
    }
  },
  {
    "ph": "s", "id": 143950, "pid": 76337, "tid": -914061504, "ts": 1716454223623591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623654, "dur": 0,
    "args": {
      "External id": 143960,
      "cbid": 317, "correlation": 143960
    }
  },
  {
    "ph": "f", "id": 143960, "pid": 76337, "tid": -914061504, "ts": 1716454223623654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623654, "dur": 0,
    "args": {
      "External id": 143961,
      "cbid": 203, "correlation": 143961
    }
  },
  {
    "ph": "f", "id": 143961, "pid": 76337, "tid": -914061504, "ts": 1716454223623654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623655, "dur": 0,
    "args": {
      "External id": 143962,
      "cbid": 205, "correlation": 143962
    }
  },
  {
    "ph": "f", "id": 143962, "pid": 76337, "tid": -914061504, "ts": 1716454223623655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680503, "dur": 5,
    "args": {
      "External id": 143966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143966,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143966, "pid": 5, "tid": 7, "ts": 1716454223680503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623669, "dur": 12,
    "args": {
      "External id": 143966,
      "cbid": 211, "correlation": 143966
    }
  },
  {
    "ph": "s", "id": 143966, "pid": 76337, "tid": -914061504, "ts": 1716454223623669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680509, "dur": 163,
    "args": {
      "External id": 143968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143968,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143968, "pid": 5, "tid": 7, "ts": 1716454223680509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623683, "dur": 6,
    "args": {
      "External id": 143968,
      "cbid": 211, "correlation": 143968
    }
  },
  {
    "ph": "s", "id": 143968, "pid": 76337, "tid": -914061504, "ts": 1716454223623683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223680675, "dur": 1,
    "args": {
      "External id": 143970,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 143970,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 143970, "pid": 5, "tid": 7, "ts": 1716454223680675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623694, "dur": 6,
    "args": {
      "External id": 143970,
      "cbid": 51, "correlation": 143970
    }
  },
  {
    "ph": "s", "id": 143970, "pid": 76337, "tid": -914061504, "ts": 1716454223623694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223680678, "dur": 260,
    "args": {
      "External id": 143971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143971,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 143971, "pid": 5, "tid": 7, "ts": 1716454223680678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623702, "dur": 6,
    "args": {
      "External id": 143971,
      "cbid": 211, "correlation": 143971
    }
  },
  {
    "ph": "s", "id": 143971, "pid": 76337, "tid": -914061504, "ts": 1716454223623702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680939, "dur": 6,
    "args": {
      "External id": 143973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143973,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 143973, "pid": 5, "tid": 7, "ts": 1716454223680939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623711, "dur": 5,
    "args": {
      "External id": 143973,
      "cbid": 211, "correlation": 143973
    }
  },
  {
    "ph": "s", "id": 143973, "pid": 76337, "tid": -914061504, "ts": 1716454223623711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223680947, "dur": 6,
    "args": {
      "External id": 143979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143979, "pid": 5, "tid": 7, "ts": 1716454223680947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623740, "dur": 8,
    "args": {
      "External id": 143979,
      "cbid": 211, "correlation": 143979
    }
  },
  {
    "ph": "s", "id": 143979, "pid": 76337, "tid": -914061504, "ts": 1716454223623740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223680955, "dur": 5,
    "args": {
      "External id": 143987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143987,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143987, "pid": 5, "tid": 7, "ts": 1716454223680955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623772, "dur": 8,
    "args": {
      "External id": 143987,
      "cbid": 211, "correlation": 143987
    }
  },
  {
    "ph": "s", "id": 143987, "pid": 76337, "tid": -914061504, "ts": 1716454223623772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223680961, "dur": 4,
    "args": {
      "External id": 143995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 143995,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 143995, "pid": 5, "tid": 7, "ts": 1716454223680961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623801, "dur": 8,
    "args": {
      "External id": 143995,
      "cbid": 211, "correlation": 143995
    }
  },
  {
    "ph": "s", "id": 143995, "pid": 76337, "tid": -914061504, "ts": 1716454223623801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680967, "dur": 9,
    "args": {
      "External id": 144015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144015,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 144015, "pid": 5, "tid": 7, "ts": 1716454223680967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623939, "dur": 14,
    "args": {
      "External id": 144015,
      "cbid": 211, "correlation": 144015
    }
  },
  {
    "ph": "s", "id": 144015, "pid": 76337, "tid": -914061504, "ts": 1716454223623939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223680978, "dur": 4,
    "args": {
      "External id": 144027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144027,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 144027, "pid": 5, "tid": 7, "ts": 1716454223680978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623962, "dur": 7,
    "args": {
      "External id": 144027,
      "cbid": 211, "correlation": 144027
    }
  },
  {
    "ph": "s", "id": 144027, "pid": 76337, "tid": -914061504, "ts": 1716454223623962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223680983, "dur": 7,
    "args": {
      "External id": 144030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144030, "pid": 5, "tid": 7, "ts": 1716454223680983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223623989, "dur": 8,
    "args": {
      "External id": 144030,
      "cbid": 211, "correlation": 144030
    }
  },
  {
    "ph": "s", "id": 144030, "pid": 76337, "tid": -914061504, "ts": 1716454223623989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624050, "dur": 0,
    "args": {
      "External id": 144041,
      "cbid": 317, "correlation": 144041
    }
  },
  {
    "ph": "f", "id": 144041, "pid": 76337, "tid": -914061504, "ts": 1716454223624050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624051, "dur": 0,
    "args": {
      "External id": 144042,
      "cbid": 203, "correlation": 144042
    }
  },
  {
    "ph": "f", "id": 144042, "pid": 76337, "tid": -914061504, "ts": 1716454223624051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624052, "dur": 0,
    "args": {
      "External id": 144043,
      "cbid": 205, "correlation": 144043
    }
  },
  {
    "ph": "f", "id": 144043, "pid": 76337, "tid": -914061504, "ts": 1716454223624052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223680991, "dur": 5,
    "args": {
      "External id": 144047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144047,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144047, "pid": 5, "tid": 7, "ts": 1716454223680991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624069, "dur": 12,
    "args": {
      "External id": 144047,
      "cbid": 211, "correlation": 144047
    }
  },
  {
    "ph": "s", "id": 144047, "pid": 76337, "tid": -914061504, "ts": 1716454223624069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223680997, "dur": 37,
    "args": {
      "External id": 144049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144049,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 144049, "pid": 5, "tid": 7, "ts": 1716454223680997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624088, "dur": 9,
    "args": {
      "External id": 144049,
      "cbid": 211, "correlation": 144049
    }
  },
  {
    "ph": "s", "id": 144049, "pid": 76337, "tid": -914061504, "ts": 1716454223624088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223681036, "dur": 5,
    "args": {
      "External id": 144051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144051,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144051, "pid": 5, "tid": 7, "ts": 1716454223681036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624101, "dur": 5,
    "args": {
      "External id": 144051,
      "cbid": 211, "correlation": 144051
    }
  },
  {
    "ph": "s", "id": 144051, "pid": 76337, "tid": -914061504, "ts": 1716454223624101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681042, "dur": 6,
    "args": {
      "External id": 144057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144057, "pid": 5, "tid": 7, "ts": 1716454223681042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624128, "dur": 9,
    "args": {
      "External id": 144057,
      "cbid": 211, "correlation": 144057
    }
  },
  {
    "ph": "s", "id": 144057, "pid": 76337, "tid": -914061504, "ts": 1716454223624128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223681049, "dur": 20,
    "args": {
      "External id": 144066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144066,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144066, "pid": 5, "tid": 7, "ts": 1716454223681049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624211, "dur": 14,
    "args": {
      "External id": 144066,
      "cbid": 211, "correlation": 144066
    }
  },
  {
    "ph": "s", "id": 144066, "pid": 76337, "tid": -914061504, "ts": 1716454223624211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223681071, "dur": 11,
    "args": {
      "External id": 144088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144088,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 144088, "pid": 5, "tid": 7, "ts": 1716454223681071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624268, "dur": 10,
    "args": {
      "External id": 144088,
      "cbid": 211, "correlation": 144088
    }
  },
  {
    "ph": "s", "id": 144088, "pid": 76337, "tid": -914061504, "ts": 1716454223624268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624361, "dur": 2,
    "args": {
      "External id": 144099,
      "cbid": 251, "correlation": 144099
    }
  },
  {
    "ph": "f", "id": 144099, "pid": 76337, "tid": -914061504, "ts": 1716454223624361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624366, "dur": 0,
    "args": {
      "External id": 144100,
      "cbid": 251, "correlation": 144100
    }
  },
  {
    "ph": "f", "id": 144100, "pid": 76337, "tid": -914061504, "ts": 1716454223624366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681083, "dur": 54,
    "args": {
      "External id": 144101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144101,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 144101, "pid": 5, "tid": 7, "ts": 1716454223681083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624369, "dur": 14,
    "args": {
      "External id": 144101,
      "cbid": 211, "correlation": 144101
    }
  },
  {
    "ph": "s", "id": 144101, "pid": 76337, "tid": -914061504, "ts": 1716454223624369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624440, "dur": 1,
    "args": {
      "External id": 144112,
      "cbid": 251, "correlation": 144112
    }
  },
  {
    "ph": "f", "id": 144112, "pid": 76337, "tid": -914061504, "ts": 1716454223624440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624444, "dur": 0,
    "args": {
      "External id": 144113,
      "cbid": 251, "correlation": 144113
    }
  },
  {
    "ph": "f", "id": 144113, "pid": 76337, "tid": -914061504, "ts": 1716454223624444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681138, "dur": 54,
    "args": {
      "External id": 144114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144114,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 144114, "pid": 5, "tid": 7, "ts": 1716454223681138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624446, "dur": 11,
    "args": {
      "External id": 144114,
      "cbid": 211, "correlation": 144114
    }
  },
  {
    "ph": "s", "id": 144114, "pid": 76337, "tid": -914061504, "ts": 1716454223624446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624508, "dur": 1,
    "args": {
      "External id": 144125,
      "cbid": 251, "correlation": 144125
    }
  },
  {
    "ph": "f", "id": 144125, "pid": 76337, "tid": -914061504, "ts": 1716454223624508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624511, "dur": 0,
    "args": {
      "External id": 144126,
      "cbid": 251, "correlation": 144126
    }
  },
  {
    "ph": "f", "id": 144126, "pid": 76337, "tid": -914061504, "ts": 1716454223624511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681193, "dur": 54,
    "args": {
      "External id": 144127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144127,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 144127, "pid": 5, "tid": 7, "ts": 1716454223681193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624513, "dur": 11,
    "args": {
      "External id": 144127,
      "cbid": 211, "correlation": 144127
    }
  },
  {
    "ph": "s", "id": 144127, "pid": 76337, "tid": -914061504, "ts": 1716454223624513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223681248, "dur": 58,
    "args": {
      "External id": 144152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144152,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144152, "pid": 5, "tid": 7, "ts": 1716454223681248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624598, "dur": 13,
    "args": {
      "External id": 144152,
      "cbid": 211, "correlation": 144152
    }
  },
  {
    "ph": "s", "id": 144152, "pid": 76337, "tid": -914061504, "ts": 1716454223624598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624698, "dur": 1,
    "args": {
      "External id": 144170,
      "cbid": 251, "correlation": 144170
    }
  },
  {
    "ph": "f", "id": 144170, "pid": 76337, "tid": -914061504, "ts": 1716454223624698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681307, "dur": 64,
    "args": {
      "External id": 144172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144172,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 144172, "pid": 5, "tid": 7, "ts": 1716454223681307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624703, "dur": 13,
    "args": {
      "External id": 144172,
      "cbid": 211, "correlation": 144172
    }
  },
  {
    "ph": "s", "id": 144172, "pid": 76337, "tid": -914061504, "ts": 1716454223624703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223681372, "dur": 6,
    "args": {
      "External id": 144180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144180,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144180, "pid": 5, "tid": 7, "ts": 1716454223681372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624773, "dur": 13,
    "args": {
      "External id": 144180,
      "cbid": 211, "correlation": 144180
    }
  },
  {
    "ph": "s", "id": 144180, "pid": 76337, "tid": -914061504, "ts": 1716454223624773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223681379, "dur": 7,
    "args": {
      "External id": 144188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144188,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144188, "pid": 5, "tid": 7, "ts": 1716454223681379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624812, "dur": 8,
    "args": {
      "External id": 144188,
      "cbid": 211, "correlation": 144188
    }
  },
  {
    "ph": "s", "id": 144188, "pid": 76337, "tid": -914061504, "ts": 1716454223624812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681387, "dur": 8,
    "args": {
      "External id": 144199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144199, "pid": 5, "tid": 7, "ts": 1716454223681387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624883, "dur": 14,
    "args": {
      "External id": 144199,
      "cbid": 211, "correlation": 144199
    }
  },
  {
    "ph": "s", "id": 144199, "pid": 76337, "tid": -914061504, "ts": 1716454223624883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223681396, "dur": 8,
    "args": {
      "External id": 144221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144221,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 144221, "pid": 5, "tid": 7, "ts": 1716454223681396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223624915, "dur": 8,
    "args": {
      "External id": 144221,
      "cbid": 211, "correlation": 144221
    }
  },
  {
    "ph": "s", "id": 144221, "pid": 76337, "tid": -914061504, "ts": 1716454223624915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625011, "dur": 2,
    "args": {
      "External id": 144232,
      "cbid": 251, "correlation": 144232
    }
  },
  {
    "ph": "f", "id": 144232, "pid": 76337, "tid": -914061504, "ts": 1716454223625011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223681406, "dur": 1,
    "args": {
      "External id": 144233,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144233,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 144233, "pid": 5, "tid": 7, "ts": 1716454223681406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625016, "dur": 11,
    "args": {
      "External id": 144233,
      "cbid": 51, "correlation": 144233
    }
  },
  {
    "ph": "s", "id": 144233, "pid": 76337, "tid": -914061504, "ts": 1716454223625016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681410, "dur": 37,
    "args": {
      "External id": 144234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144234,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 144234, "pid": 5, "tid": 7, "ts": 1716454223681410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625028, "dur": 13,
    "args": {
      "External id": 144234,
      "cbid": 211, "correlation": 144234
    }
  },
  {
    "ph": "s", "id": 144234, "pid": 76337, "tid": -914061504, "ts": 1716454223625028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625101, "dur": 1,
    "args": {
      "External id": 144245,
      "cbid": 251, "correlation": 144245
    }
  },
  {
    "ph": "f", "id": 144245, "pid": 76337, "tid": -914061504, "ts": 1716454223625101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625105, "dur": 0,
    "args": {
      "External id": 144246,
      "cbid": 251, "correlation": 144246
    }
  },
  {
    "ph": "f", "id": 144246, "pid": 76337, "tid": -914061504, "ts": 1716454223625105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223681448, "dur": 12,
    "args": {
      "External id": 144247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144247,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144247, "pid": 5, "tid": 7, "ts": 1716454223681448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625106, "dur": 12,
    "args": {
      "External id": 144247,
      "cbid": 211, "correlation": 144247
    }
  },
  {
    "ph": "s", "id": 144247, "pid": 76337, "tid": -914061504, "ts": 1716454223625106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223681461, "dur": 5,
    "args": {
      "External id": 144249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144249,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144249, "pid": 5, "tid": 7, "ts": 1716454223681461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625121, "dur": 6,
    "args": {
      "External id": 144249,
      "cbid": 211, "correlation": 144249
    }
  },
  {
    "ph": "s", "id": 144249, "pid": 76337, "tid": -914061504, "ts": 1716454223625121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625178, "dur": 1,
    "args": {
      "External id": 144260,
      "cbid": 251, "correlation": 144260
    }
  },
  {
    "ph": "f", "id": 144260, "pid": 76337, "tid": -914061504, "ts": 1716454223625178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625182, "dur": 0,
    "args": {
      "External id": 144261,
      "cbid": 251, "correlation": 144261
    }
  },
  {
    "ph": "f", "id": 144261, "pid": 76337, "tid": -914061504, "ts": 1716454223625182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223681467, "dur": 8,
    "args": {
      "External id": 144262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144262,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144262, "pid": 5, "tid": 7, "ts": 1716454223681467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625183, "dur": 12,
    "args": {
      "External id": 144262,
      "cbid": 211, "correlation": 144262
    }
  },
  {
    "ph": "s", "id": 144262, "pid": 76337, "tid": -914061504, "ts": 1716454223625183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223681476, "dur": 4,
    "args": {
      "External id": 144264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144264,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144264, "pid": 5, "tid": 7, "ts": 1716454223681476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625197, "dur": 5,
    "args": {
      "External id": 144264,
      "cbid": 211, "correlation": 144264
    }
  },
  {
    "ph": "s", "id": 144264, "pid": 76337, "tid": -914061504, "ts": 1716454223625197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223681481, "dur": 20,
    "args": {
      "External id": 144289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144289,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 144289, "pid": 5, "tid": 7, "ts": 1716454223681481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625274, "dur": 13,
    "args": {
      "External id": 144289,
      "cbid": 211, "correlation": 144289
    }
  },
  {
    "ph": "s", "id": 144289, "pid": 76337, "tid": -914061504, "ts": 1716454223625274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625374, "dur": 2,
    "args": {
      "External id": 144307,
      "cbid": 251, "correlation": 144307
    }
  },
  {
    "ph": "f", "id": 144307, "pid": 76337, "tid": -914061504, "ts": 1716454223625374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223681503, "dur": 1,
    "args": {
      "External id": 144309,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144309,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 144309, "pid": 5, "tid": 7, "ts": 1716454223681503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625380, "dur": 12,
    "args": {
      "External id": 144309,
      "cbid": 51, "correlation": 144309
    }
  },
  {
    "ph": "s", "id": 144309, "pid": 76337, "tid": -914061504, "ts": 1716454223625380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681507, "dur": 36,
    "args": {
      "External id": 144310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144310,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 144310, "pid": 5, "tid": 7, "ts": 1716454223681507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625393, "dur": 12,
    "args": {
      "External id": 144310,
      "cbid": 211, "correlation": 144310
    }
  },
  {
    "ph": "s", "id": 144310, "pid": 76337, "tid": -914061504, "ts": 1716454223625393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223681545, "dur": 4,
    "args": {
      "External id": 144318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144318,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144318, "pid": 5, "tid": 7, "ts": 1716454223681545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625464, "dur": 12,
    "args": {
      "External id": 144318,
      "cbid": 211, "correlation": 144318
    }
  },
  {
    "ph": "s", "id": 144318, "pid": 76337, "tid": -914061504, "ts": 1716454223625464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681550, "dur": 8,
    "args": {
      "External id": 144326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144326, "pid": 5, "tid": 7, "ts": 1716454223681550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625505, "dur": 9,
    "args": {
      "External id": 144326,
      "cbid": 211, "correlation": 144326
    }
  },
  {
    "ph": "s", "id": 144326, "pid": 76337, "tid": -914061504, "ts": 1716454223625505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223681559, "dur": 8,
    "args": {
      "External id": 144348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144348,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 144348, "pid": 5, "tid": 7, "ts": 1716454223681559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625557, "dur": 10,
    "args": {
      "External id": 144348,
      "cbid": 211, "correlation": 144348
    }
  },
  {
    "ph": "s", "id": 144348, "pid": 76337, "tid": -914061504, "ts": 1716454223625557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625650, "dur": 1,
    "args": {
      "External id": 144364,
      "cbid": 251, "correlation": 144364
    }
  },
  {
    "ph": "f", "id": 144364, "pid": 76337, "tid": -914061504, "ts": 1716454223625650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625655, "dur": 0,
    "args": {
      "External id": 144366,
      "cbid": 251, "correlation": 144366
    }
  },
  {
    "ph": "f", "id": 144366, "pid": 76337, "tid": -914061504, "ts": 1716454223625655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223681569, "dur": 192,
    "args": {
      "External id": 144367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144367,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144367, "pid": 5, "tid": 7, "ts": 1716454223681569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625657, "dur": 13,
    "args": {
      "External id": 144367,
      "cbid": 211, "correlation": 144367
    }
  },
  {
    "ph": "s", "id": 144367, "pid": 76337, "tid": -914061504, "ts": 1716454223625657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681762, "dur": 20,
    "args": {
      "External id": 144375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144375,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144375, "pid": 5, "tid": 7, "ts": 1716454223681762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625723, "dur": 12,
    "args": {
      "External id": 144375,
      "cbid": 211, "correlation": 144375
    }
  },
  {
    "ph": "s", "id": 144375, "pid": 76337, "tid": -914061504, "ts": 1716454223625723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681784, "dur": 22,
    "args": {
      "External id": 144383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144383, "pid": 5, "tid": 7, "ts": 1716454223681784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625754, "dur": 8,
    "args": {
      "External id": 144383,
      "cbid": 211, "correlation": 144383
    }
  },
  {
    "ph": "s", "id": 144383, "pid": 76337, "tid": -914061504, "ts": 1716454223625754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625835, "dur": 1,
    "args": {
      "External id": 144399,
      "cbid": 251, "correlation": 144399
    }
  },
  {
    "ph": "f", "id": 144399, "pid": 76337, "tid": -914061504, "ts": 1716454223625835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223681808, "dur": 1,
    "args": {
      "External id": 144401,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144401,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 144401, "pid": 5, "tid": 7, "ts": 1716454223681808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625840, "dur": 10,
    "args": {
      "External id": 144401,
      "cbid": 51, "correlation": 144401
    }
  },
  {
    "ph": "s", "id": 144401, "pid": 76337, "tid": -914061504, "ts": 1716454223625840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223681812, "dur": 109,
    "args": {
      "External id": 144402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144402,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 144402, "pid": 5, "tid": 7, "ts": 1716454223681812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625851, "dur": 12,
    "args": {
      "External id": 144402,
      "cbid": 211, "correlation": 144402
    }
  },
  {
    "ph": "s", "id": 144402, "pid": 76337, "tid": -914061504, "ts": 1716454223625851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223681922, "dur": 5,
    "args": {
      "External id": 144410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144410,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144410, "pid": 5, "tid": 7, "ts": 1716454223681922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625894, "dur": 11,
    "args": {
      "External id": 144410,
      "cbid": 211, "correlation": 144410
    }
  },
  {
    "ph": "s", "id": 144410, "pid": 76337, "tid": -914061504, "ts": 1716454223625894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681929, "dur": 10,
    "args": {
      "External id": 144421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144421, "pid": 5, "tid": 7, "ts": 1716454223681929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223625964, "dur": 20,
    "args": {
      "External id": 144421,
      "cbid": 211, "correlation": 144421
    }
  },
  {
    "ph": "s", "id": 144421, "pid": 76337, "tid": -914061504, "ts": 1716454223625964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626039, "dur": 0,
    "args": {
      "External id": 144433,
      "cbid": 317, "correlation": 144433
    }
  },
  {
    "ph": "f", "id": 144433, "pid": 76337, "tid": -914061504, "ts": 1716454223626039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626040, "dur": 0,
    "args": {
      "External id": 144434,
      "cbid": 203, "correlation": 144434
    }
  },
  {
    "ph": "f", "id": 144434, "pid": 76337, "tid": -914061504, "ts": 1716454223626040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626041, "dur": 0,
    "args": {
      "External id": 144435,
      "cbid": 205, "correlation": 144435
    }
  },
  {
    "ph": "f", "id": 144435, "pid": 76337, "tid": -914061504, "ts": 1716454223626041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223681939, "dur": 6,
    "args": {
      "External id": 144439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144439,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144439, "pid": 5, "tid": 7, "ts": 1716454223681939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626056, "dur": 12,
    "args": {
      "External id": 144439,
      "cbid": 211, "correlation": 144439
    }
  },
  {
    "ph": "s", "id": 144439, "pid": 76337, "tid": -914061504, "ts": 1716454223626056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223681947, "dur": 37,
    "args": {
      "External id": 144441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144441,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 144441, "pid": 5, "tid": 7, "ts": 1716454223681947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626075, "dur": 7,
    "args": {
      "External id": 144441,
      "cbid": 211, "correlation": 144441
    }
  },
  {
    "ph": "s", "id": 144441, "pid": 76337, "tid": -914061504, "ts": 1716454223626075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223681985, "dur": 6,
    "args": {
      "External id": 144443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144443,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144443, "pid": 5, "tid": 7, "ts": 1716454223681985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626086, "dur": 5,
    "args": {
      "External id": 144443,
      "cbid": 211, "correlation": 144443
    }
  },
  {
    "ph": "s", "id": 144443, "pid": 76337, "tid": -914061504, "ts": 1716454223626086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223681992, "dur": 7,
    "args": {
      "External id": 144449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144449, "pid": 5, "tid": 7, "ts": 1716454223681992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626114, "dur": 9,
    "args": {
      "External id": 144449,
      "cbid": 211, "correlation": 144449
    }
  },
  {
    "ph": "s", "id": 144449, "pid": 76337, "tid": -914061504, "ts": 1716454223626114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223682001, "dur": 5,
    "args": {
      "External id": 144457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144457,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144457, "pid": 5, "tid": 7, "ts": 1716454223682001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626145, "dur": 9,
    "args": {
      "External id": 144457,
      "cbid": 211, "correlation": 144457
    }
  },
  {
    "ph": "s", "id": 144457, "pid": 76337, "tid": -914061504, "ts": 1716454223626145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682007, "dur": 11,
    "args": {
      "External id": 144477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144477,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 144477, "pid": 5, "tid": 7, "ts": 1716454223682007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626217, "dur": 12,
    "args": {
      "External id": 144477,
      "cbid": 211, "correlation": 144477
    }
  },
  {
    "ph": "s", "id": 144477, "pid": 76337, "tid": -914061504, "ts": 1716454223626217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223682019, "dur": 4,
    "args": {
      "External id": 144489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144489,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 144489, "pid": 5, "tid": 7, "ts": 1716454223682019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626239, "dur": 6,
    "args": {
      "External id": 144489,
      "cbid": 211, "correlation": 144489
    }
  },
  {
    "ph": "s", "id": 144489, "pid": 76337, "tid": -914061504, "ts": 1716454223626239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223682025, "dur": 8,
    "args": {
      "External id": 144492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144492,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144492, "pid": 5, "tid": 7, "ts": 1716454223682025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626258, "dur": 6,
    "args": {
      "External id": 144492,
      "cbid": 211, "correlation": 144492
    }
  },
  {
    "ph": "s", "id": 144492, "pid": 76337, "tid": -914061504, "ts": 1716454223626258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223682035, "dur": 5,
    "args": {
      "External id": 144501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144501,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144501, "pid": 5, "tid": 7, "ts": 1716454223682035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626296, "dur": 11,
    "args": {
      "External id": 144501,
      "cbid": 211, "correlation": 144501
    }
  },
  {
    "ph": "s", "id": 144501, "pid": 76337, "tid": -914061504, "ts": 1716454223626296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626349, "dur": 0,
    "args": {
      "External id": 144511,
      "cbid": 317, "correlation": 144511
    }
  },
  {
    "ph": "f", "id": 144511, "pid": 76337, "tid": -914061504, "ts": 1716454223626349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626349, "dur": 0,
    "args": {
      "External id": 144512,
      "cbid": 203, "correlation": 144512
    }
  },
  {
    "ph": "f", "id": 144512, "pid": 76337, "tid": -914061504, "ts": 1716454223626349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626350, "dur": 0,
    "args": {
      "External id": 144513,
      "cbid": 205, "correlation": 144513
    }
  },
  {
    "ph": "f", "id": 144513, "pid": 76337, "tid": -914061504, "ts": 1716454223626350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682041, "dur": 5,
    "args": {
      "External id": 144517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144517,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144517, "pid": 5, "tid": 7, "ts": 1716454223682041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626364, "dur": 11,
    "args": {
      "External id": 144517,
      "cbid": 211, "correlation": 144517
    }
  },
  {
    "ph": "s", "id": 144517, "pid": 76337, "tid": -914061504, "ts": 1716454223626364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682047, "dur": 163,
    "args": {
      "External id": 144519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144519,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144519, "pid": 5, "tid": 7, "ts": 1716454223682047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626378, "dur": 5,
    "args": {
      "External id": 144519,
      "cbid": 211, "correlation": 144519
    }
  },
  {
    "ph": "s", "id": 144519, "pid": 76337, "tid": -914061504, "ts": 1716454223626378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223682213, "dur": 1,
    "args": {
      "External id": 144521,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144521,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 144521, "pid": 5, "tid": 7, "ts": 1716454223682213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626388, "dur": 6,
    "args": {
      "External id": 144521,
      "cbid": 51, "correlation": 144521
    }
  },
  {
    "ph": "s", "id": 144521, "pid": 76337, "tid": -914061504, "ts": 1716454223626388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223682217, "dur": 270,
    "args": {
      "External id": 144522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144522,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144522, "pid": 5, "tid": 7, "ts": 1716454223682217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626395, "dur": 7,
    "args": {
      "External id": 144522,
      "cbid": 211, "correlation": 144522
    }
  },
  {
    "ph": "s", "id": 144522, "pid": 76337, "tid": -914061504, "ts": 1716454223626395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682488, "dur": 6,
    "args": {
      "External id": 144524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144524,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144524, "pid": 5, "tid": 7, "ts": 1716454223682488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626406, "dur": 5,
    "args": {
      "External id": 144524,
      "cbid": 211, "correlation": 144524
    }
  },
  {
    "ph": "s", "id": 144524, "pid": 76337, "tid": -914061504, "ts": 1716454223626406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223682495, "dur": 6,
    "args": {
      "External id": 144530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144530, "pid": 5, "tid": 7, "ts": 1716454223682495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626433, "dur": 8,
    "args": {
      "External id": 144530,
      "cbid": 211, "correlation": 144530
    }
  },
  {
    "ph": "s", "id": 144530, "pid": 76337, "tid": -914061504, "ts": 1716454223626433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223682503, "dur": 4,
    "args": {
      "External id": 144538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144538,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 144538, "pid": 5, "tid": 7, "ts": 1716454223682503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626478, "dur": 9,
    "args": {
      "External id": 144538,
      "cbid": 211, "correlation": 144538
    }
  },
  {
    "ph": "s", "id": 144538, "pid": 76337, "tid": -914061504, "ts": 1716454223626478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626543, "dur": 1,
    "args": {
      "External id": 144554,
      "cbid": 251, "correlation": 144554
    }
  },
  {
    "ph": "f", "id": 144554, "pid": 76337, "tid": -914061504, "ts": 1716454223626543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626549, "dur": 0,
    "args": {
      "External id": 144556,
      "cbid": 251, "correlation": 144556
    }
  },
  {
    "ph": "f", "id": 144556, "pid": 76337, "tid": -914061504, "ts": 1716454223626549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223682508, "dur": 13,
    "args": {
      "External id": 144557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144557,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144557, "pid": 5, "tid": 7, "ts": 1716454223682508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626551, "dur": 11,
    "args": {
      "External id": 144557,
      "cbid": 211, "correlation": 144557
    }
  },
  {
    "ph": "s", "id": 144557, "pid": 76337, "tid": -914061504, "ts": 1716454223626551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223682522, "dur": 5,
    "args": {
      "External id": 144559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144559,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144559, "pid": 5, "tid": 7, "ts": 1716454223682522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626564, "dur": 5,
    "args": {
      "External id": 144559,
      "cbid": 211, "correlation": 144559
    }
  },
  {
    "ph": "s", "id": 144559, "pid": 76337, "tid": -914061504, "ts": 1716454223626564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223682528, "dur": 6,
    "args": {
      "External id": 144569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144569, "pid": 5, "tid": 7, "ts": 1716454223682528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626621, "dur": 12,
    "args": {
      "External id": 144569,
      "cbid": 211, "correlation": 144569
    }
  },
  {
    "ph": "s", "id": 144569, "pid": 76337, "tid": -914061504, "ts": 1716454223626621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682535, "dur": 10,
    "args": {
      "External id": 144589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144589,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 144589, "pid": 5, "tid": 7, "ts": 1716454223682535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626687, "dur": 10,
    "args": {
      "External id": 144589,
      "cbid": 211, "correlation": 144589
    }
  },
  {
    "ph": "s", "id": 144589, "pid": 76337, "tid": -914061504, "ts": 1716454223626687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223682547, "dur": 4,
    "args": {
      "External id": 144601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144601,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 144601, "pid": 5, "tid": 7, "ts": 1716454223682547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626708, "dur": 6,
    "args": {
      "External id": 144601,
      "cbid": 211, "correlation": 144601
    }
  },
  {
    "ph": "s", "id": 144601, "pid": 76337, "tid": -914061504, "ts": 1716454223626708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223682552, "dur": 7,
    "args": {
      "External id": 144604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144604, "pid": 5, "tid": 7, "ts": 1716454223682552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626727, "dur": 6,
    "args": {
      "External id": 144604,
      "cbid": 211, "correlation": 144604
    }
  },
  {
    "ph": "s", "id": 144604, "pid": 76337, "tid": -914061504, "ts": 1716454223626727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223682560, "dur": 5,
    "args": {
      "External id": 144613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144613,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144613, "pid": 5, "tid": 7, "ts": 1716454223682560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626768, "dur": 9,
    "args": {
      "External id": 144613,
      "cbid": 211, "correlation": 144613
    }
  },
  {
    "ph": "s", "id": 144613, "pid": 76337, "tid": -914061504, "ts": 1716454223626768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626829, "dur": 0,
    "args": {
      "External id": 144623,
      "cbid": 317, "correlation": 144623
    }
  },
  {
    "ph": "f", "id": 144623, "pid": 76337, "tid": -914061504, "ts": 1716454223626829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626830, "dur": 0,
    "args": {
      "External id": 144624,
      "cbid": 203, "correlation": 144624
    }
  },
  {
    "ph": "f", "id": 144624, "pid": 76337, "tid": -914061504, "ts": 1716454223626830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626831, "dur": 0,
    "args": {
      "External id": 144625,
      "cbid": 205, "correlation": 144625
    }
  },
  {
    "ph": "f", "id": 144625, "pid": 76337, "tid": -914061504, "ts": 1716454223626831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682566, "dur": 5,
    "args": {
      "External id": 144629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144629,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144629, "pid": 5, "tid": 7, "ts": 1716454223682566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626845, "dur": 13,
    "args": {
      "External id": 144629,
      "cbid": 211, "correlation": 144629
    }
  },
  {
    "ph": "s", "id": 144629, "pid": 76337, "tid": -914061504, "ts": 1716454223626845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223682572, "dur": 164,
    "args": {
      "External id": 144631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144631, "pid": 5, "tid": 7, "ts": 1716454223682572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626860, "dur": 5,
    "args": {
      "External id": 144631,
      "cbid": 211, "correlation": 144631
    }
  },
  {
    "ph": "s", "id": 144631, "pid": 76337, "tid": -914061504, "ts": 1716454223626860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223682738, "dur": 1,
    "args": {
      "External id": 144633,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144633,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 144633, "pid": 5, "tid": 7, "ts": 1716454223682738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626871, "dur": 6,
    "args": {
      "External id": 144633,
      "cbid": 51, "correlation": 144633
    }
  },
  {
    "ph": "s", "id": 144633, "pid": 76337, "tid": -914061504, "ts": 1716454223626871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223682742, "dur": 260,
    "args": {
      "External id": 144634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144634,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144634, "pid": 5, "tid": 7, "ts": 1716454223682742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626878, "dur": 6,
    "args": {
      "External id": 144634,
      "cbid": 211, "correlation": 144634
    }
  },
  {
    "ph": "s", "id": 144634, "pid": 76337, "tid": -914061504, "ts": 1716454223626878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683003, "dur": 6,
    "args": {
      "External id": 144636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144636,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144636, "pid": 5, "tid": 7, "ts": 1716454223683003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626888, "dur": 5,
    "args": {
      "External id": 144636,
      "cbid": 211, "correlation": 144636
    }
  },
  {
    "ph": "s", "id": 144636, "pid": 76337, "tid": -914061504, "ts": 1716454223626888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223683011, "dur": 6,
    "args": {
      "External id": 144642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144642,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144642, "pid": 5, "tid": 7, "ts": 1716454223683011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626917, "dur": 8,
    "args": {
      "External id": 144642,
      "cbid": 211, "correlation": 144642
    }
  },
  {
    "ph": "s", "id": 144642, "pid": 76337, "tid": -914061504, "ts": 1716454223626917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223683018, "dur": 5,
    "args": {
      "External id": 144650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144650,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144650, "pid": 5, "tid": 7, "ts": 1716454223683018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626949, "dur": 9,
    "args": {
      "External id": 144650,
      "cbid": 211, "correlation": 144650
    }
  },
  {
    "ph": "s", "id": 144650, "pid": 76337, "tid": -914061504, "ts": 1716454223626949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223683024, "dur": 4,
    "args": {
      "External id": 144658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144658,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144658, "pid": 5, "tid": 7, "ts": 1716454223683024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223626988, "dur": 9,
    "args": {
      "External id": 144658,
      "cbid": 211, "correlation": 144658
    }
  },
  {
    "ph": "s", "id": 144658, "pid": 76337, "tid": -914061504, "ts": 1716454223626988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223683030, "dur": 12,
    "args": {
      "External id": 144667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144667,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144667, "pid": 5, "tid": 7, "ts": 1716454223683030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223627079, "dur": 14,
    "args": {
      "External id": 144667,
      "cbid": 211, "correlation": 144667
    }
  },
  {
    "ph": "s", "id": 144667, "pid": 76337, "tid": -914061504, "ts": 1716454223627079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683043, "dur": 12,
    "args": {
      "External id": 144687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144687,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 144687, "pid": 5, "tid": 7, "ts": 1716454223683043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223627147, "dur": 12,
    "args": {
      "External id": 144687,
      "cbid": 211, "correlation": 144687
    }
  },
  {
    "ph": "s", "id": 144687, "pid": 76337, "tid": -914061504, "ts": 1716454223627147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223683057, "dur": 4,
    "args": {
      "External id": 144699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144699,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144699, "pid": 5, "tid": 7, "ts": 1716454223683057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223627169, "dur": 6,
    "args": {
      "External id": 144699,
      "cbid": 211, "correlation": 144699
    }
  },
  {
    "ph": "s", "id": 144699, "pid": 76337, "tid": -914061504, "ts": 1716454223627169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223683062, "dur": 10,
    "args": {
      "External id": 144702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144702, "pid": 5, "tid": 7, "ts": 1716454223683062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223627187, "dur": 7,
    "args": {
      "External id": 144702,
      "cbid": 211, "correlation": 144702
    }
  },
  {
    "ph": "s", "id": 144702, "pid": 76337, "tid": -914061504, "ts": 1716454223627187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223683074, "dur": 6,
    "args": {
      "External id": 144711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144711,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144711, "pid": 5, "tid": 7, "ts": 1716454223683074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630514, "dur": 37,
    "args": {
      "External id": 144711,
      "cbid": 211, "correlation": 144711
    }
  },
  {
    "ph": "s", "id": 144711, "pid": 76337, "tid": -914061504, "ts": 1716454223630514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630680, "dur": 1,
    "args": {
      "External id": 144721,
      "cbid": 317, "correlation": 144721
    }
  },
  {
    "ph": "f", "id": 144721, "pid": 76337, "tid": -914061504, "ts": 1716454223630680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630681, "dur": 1,
    "args": {
      "External id": 144722,
      "cbid": 203, "correlation": 144722
    }
  },
  {
    "ph": "f", "id": 144722, "pid": 76337, "tid": -914061504, "ts": 1716454223630681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630683, "dur": 1,
    "args": {
      "External id": 144723,
      "cbid": 205, "correlation": 144723
    }
  },
  {
    "ph": "f", "id": 144723, "pid": 76337, "tid": -914061504, "ts": 1716454223630683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683081, "dur": 6,
    "args": {
      "External id": 144727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144727,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144727, "pid": 5, "tid": 7, "ts": 1716454223683081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630751, "dur": 16,
    "args": {
      "External id": 144727,
      "cbid": 211, "correlation": 144727
    }
  },
  {
    "ph": "s", "id": 144727, "pid": 76337, "tid": -914061504, "ts": 1716454223630751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683089, "dur": 320,
    "args": {
      "External id": 144729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144729,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144729, "pid": 5, "tid": 7, "ts": 1716454223683089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630770, "dur": 5,
    "args": {
      "External id": 144729,
      "cbid": 211, "correlation": 144729
    }
  },
  {
    "ph": "s", "id": 144729, "pid": 76337, "tid": -914061504, "ts": 1716454223630770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223683411, "dur": 1,
    "args": {
      "External id": 144731,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144731,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 144731, "pid": 5, "tid": 7, "ts": 1716454223683411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630788, "dur": 16,
    "args": {
      "External id": 144731,
      "cbid": 51, "correlation": 144731
    }
  },
  {
    "ph": "s", "id": 144731, "pid": 76337, "tid": -914061504, "ts": 1716454223630788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223683415, "dur": 496,
    "args": {
      "External id": 144732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144732,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144732, "pid": 5, "tid": 7, "ts": 1716454223683415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630806, "dur": 9,
    "args": {
      "External id": 144732,
      "cbid": 211, "correlation": 144732
    }
  },
  {
    "ph": "s", "id": 144732, "pid": 76337, "tid": -914061504, "ts": 1716454223630806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683912, "dur": 5,
    "args": {
      "External id": 144734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144734, "pid": 5, "tid": 7, "ts": 1716454223683912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630822, "dur": 7,
    "args": {
      "External id": 144734,
      "cbid": 211, "correlation": 144734
    }
  },
  {
    "ph": "s", "id": 144734, "pid": 76337, "tid": -914061504, "ts": 1716454223630822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223683919, "dur": 6,
    "args": {
      "External id": 144740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144740, "pid": 5, "tid": 7, "ts": 1716454223683919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630912, "dur": 12,
    "args": {
      "External id": 144740,
      "cbid": 211, "correlation": 144740
    }
  },
  {
    "ph": "s", "id": 144740, "pid": 76337, "tid": -914061504, "ts": 1716454223630912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223683926, "dur": 3,
    "args": {
      "External id": 144748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144748,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 144748, "pid": 5, "tid": 7, "ts": 1716454223683926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223630982, "dur": 11,
    "args": {
      "External id": 144748,
      "cbid": 211, "correlation": 144748
    }
  },
  {
    "ph": "s", "id": 144748, "pid": 76337, "tid": -914061504, "ts": 1716454223630982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631119, "dur": 4,
    "args": {
      "External id": 144764,
      "cbid": 251, "correlation": 144764
    }
  },
  {
    "ph": "f", "id": 144764, "pid": 76337, "tid": -914061504, "ts": 1716454223631119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631128, "dur": 0,
    "args": {
      "External id": 144766,
      "cbid": 251, "correlation": 144766
    }
  },
  {
    "ph": "f", "id": 144766, "pid": 76337, "tid": -914061504, "ts": 1716454223631128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223683931, "dur": 10,
    "args": {
      "External id": 144767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144767,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144767, "pid": 5, "tid": 7, "ts": 1716454223683931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631132, "dur": 16,
    "args": {
      "External id": 144767,
      "cbid": 211, "correlation": 144767
    }
  },
  {
    "ph": "s", "id": 144767, "pid": 76337, "tid": -914061504, "ts": 1716454223631132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223683943, "dur": 5,
    "args": {
      "External id": 144769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144769,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144769, "pid": 5, "tid": 7, "ts": 1716454223683943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631153, "dur": 8,
    "args": {
      "External id": 144769,
      "cbid": 211, "correlation": 144769
    }
  },
  {
    "ph": "s", "id": 144769, "pid": 76337, "tid": -914061504, "ts": 1716454223631153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223683949, "dur": 6,
    "args": {
      "External id": 144779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144779, "pid": 5, "tid": 7, "ts": 1716454223683949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631241, "dur": 12,
    "args": {
      "External id": 144779,
      "cbid": 211, "correlation": 144779
    }
  },
  {
    "ph": "s", "id": 144779, "pid": 76337, "tid": -914061504, "ts": 1716454223631241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683956, "dur": 9,
    "args": {
      "External id": 144799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144799,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 144799, "pid": 5, "tid": 7, "ts": 1716454223683956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631349, "dur": 13,
    "args": {
      "External id": 144799,
      "cbid": 211, "correlation": 144799
    }
  },
  {
    "ph": "s", "id": 144799, "pid": 76337, "tid": -914061504, "ts": 1716454223631349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223683966, "dur": 4,
    "args": {
      "External id": 144811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144811,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 144811, "pid": 5, "tid": 7, "ts": 1716454223683966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631374, "dur": 8,
    "args": {
      "External id": 144811,
      "cbid": 211, "correlation": 144811
    }
  },
  {
    "ph": "s", "id": 144811, "pid": 76337, "tid": -914061504, "ts": 1716454223631374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223683972, "dur": 7,
    "args": {
      "External id": 144814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144814,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144814, "pid": 5, "tid": 7, "ts": 1716454223683972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631413, "dur": 7,
    "args": {
      "External id": 144814,
      "cbid": 211, "correlation": 144814
    }
  },
  {
    "ph": "s", "id": 144814, "pid": 76337, "tid": -914061504, "ts": 1716454223631413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223683980, "dur": 5,
    "args": {
      "External id": 144823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144823,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144823, "pid": 5, "tid": 7, "ts": 1716454223683980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631457, "dur": 9,
    "args": {
      "External id": 144823,
      "cbid": 211, "correlation": 144823
    }
  },
  {
    "ph": "s", "id": 144823, "pid": 76337, "tid": -914061504, "ts": 1716454223631457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631535, "dur": 0,
    "args": {
      "External id": 144833,
      "cbid": 317, "correlation": 144833
    }
  },
  {
    "ph": "f", "id": 144833, "pid": 76337, "tid": -914061504, "ts": 1716454223631535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631536, "dur": 1,
    "args": {
      "External id": 144834,
      "cbid": 203, "correlation": 144834
    }
  },
  {
    "ph": "f", "id": 144834, "pid": 76337, "tid": -914061504, "ts": 1716454223631536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631538, "dur": 0,
    "args": {
      "External id": 144835,
      "cbid": 205, "correlation": 144835
    }
  },
  {
    "ph": "f", "id": 144835, "pid": 76337, "tid": -914061504, "ts": 1716454223631538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683986, "dur": 5,
    "args": {
      "External id": 144839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144839,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144839, "pid": 5, "tid": 7, "ts": 1716454223683986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631704, "dur": 13,
    "args": {
      "External id": 144839,
      "cbid": 211, "correlation": 144839
    }
  },
  {
    "ph": "s", "id": 144839, "pid": 76337, "tid": -914061504, "ts": 1716454223631704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223683992, "dur": 163,
    "args": {
      "External id": 144841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144841,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144841, "pid": 5, "tid": 7, "ts": 1716454223683992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631719, "dur": 5,
    "args": {
      "External id": 144841,
      "cbid": 211, "correlation": 144841
    }
  },
  {
    "ph": "s", "id": 144841, "pid": 76337, "tid": -914061504, "ts": 1716454223631719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223684157, "dur": 1,
    "args": {
      "External id": 144843,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144843,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 144843, "pid": 5, "tid": 7, "ts": 1716454223684157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631733, "dur": 10,
    "args": {
      "External id": 144843,
      "cbid": 51, "correlation": 144843
    }
  },
  {
    "ph": "s", "id": 144843, "pid": 76337, "tid": -914061504, "ts": 1716454223631733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223684160, "dur": 260,
    "args": {
      "External id": 144844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144844,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144844, "pid": 5, "tid": 7, "ts": 1716454223684160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631744, "dur": 7,
    "args": {
      "External id": 144844,
      "cbid": 211, "correlation": 144844
    }
  },
  {
    "ph": "s", "id": 144844, "pid": 76337, "tid": -914061504, "ts": 1716454223631744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223684422, "dur": 5,
    "args": {
      "External id": 144846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144846, "pid": 5, "tid": 7, "ts": 1716454223684422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631756, "dur": 5,
    "args": {
      "External id": 144846,
      "cbid": 211, "correlation": 144846
    }
  },
  {
    "ph": "s", "id": 144846, "pid": 76337, "tid": -914061504, "ts": 1716454223631756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223684428, "dur": 6,
    "args": {
      "External id": 144852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144852, "pid": 5, "tid": 7, "ts": 1716454223684428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631787, "dur": 8,
    "args": {
      "External id": 144852,
      "cbid": 211, "correlation": 144852
    }
  },
  {
    "ph": "s", "id": 144852, "pid": 76337, "tid": -914061504, "ts": 1716454223631787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631848, "dur": 0,
    "args": {
      "External id": 144862,
      "cbid": 317, "correlation": 144862
    }
  },
  {
    "ph": "f", "id": 144862, "pid": 76337, "tid": -914061504, "ts": 1716454223631848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631849, "dur": 0,
    "args": {
      "External id": 144863,
      "cbid": 203, "correlation": 144863
    }
  },
  {
    "ph": "f", "id": 144863, "pid": 76337, "tid": -914061504, "ts": 1716454223631849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631849, "dur": 0,
    "args": {
      "External id": 144864,
      "cbid": 205, "correlation": 144864
    }
  },
  {
    "ph": "f", "id": 144864, "pid": 76337, "tid": -914061504, "ts": 1716454223631849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223684436, "dur": 8,
    "args": {
      "External id": 144868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144868, "pid": 5, "tid": 7, "ts": 1716454223684436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631920, "dur": 12,
    "args": {
      "External id": 144868,
      "cbid": 211, "correlation": 144868
    }
  },
  {
    "ph": "s", "id": 144868, "pid": 76337, "tid": -914061504, "ts": 1716454223631920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223684445, "dur": 3,
    "args": {
      "External id": 144870,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144870,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 144870, "pid": 5, "tid": 7, "ts": 1716454223684445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631939, "dur": 14,
    "args": {
      "External id": 144870,
      "cbid": 51, "correlation": 144870
    }
  },
  {
    "ph": "s", "id": 144870, "pid": 76337, "tid": -914061504, "ts": 1716454223631939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223684449, "dur": 97,
    "args": {
      "External id": 144871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144871,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 144871, "pid": 5, "tid": 7, "ts": 1716454223684449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631955, "dur": 9,
    "args": {
      "External id": 144871,
      "cbid": 211, "correlation": 144871
    }
  },
  {
    "ph": "s", "id": 144871, "pid": 76337, "tid": -914061504, "ts": 1716454223631955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223684548, "dur": 6,
    "args": {
      "External id": 144873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144873,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144873, "pid": 5, "tid": 7, "ts": 1716454223684548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223631967, "dur": 18,
    "args": {
      "External id": 144873,
      "cbid": 211, "correlation": 144873
    }
  },
  {
    "ph": "s", "id": 144873, "pid": 76337, "tid": -914061504, "ts": 1716454223631967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223684555, "dur": 6,
    "args": {
      "External id": 144879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144879,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144879, "pid": 5, "tid": 7, "ts": 1716454223684555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632009, "dur": 9,
    "args": {
      "External id": 144879,
      "cbid": 211, "correlation": 144879
    }
  },
  {
    "ph": "s", "id": 144879, "pid": 76337, "tid": -914061504, "ts": 1716454223632009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223684562, "dur": 5,
    "args": {
      "External id": 144887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144887,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144887, "pid": 5, "tid": 7, "ts": 1716454223684562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632042, "dur": 9,
    "args": {
      "External id": 144887,
      "cbid": 211, "correlation": 144887
    }
  },
  {
    "ph": "s", "id": 144887, "pid": 76337, "tid": -914061504, "ts": 1716454223632042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223684569, "dur": 4,
    "args": {
      "External id": 144895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144895,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144895, "pid": 5, "tid": 7, "ts": 1716454223684569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632092, "dur": 10,
    "args": {
      "External id": 144895,
      "cbid": 211, "correlation": 144895
    }
  },
  {
    "ph": "s", "id": 144895, "pid": 76337, "tid": -914061504, "ts": 1716454223632092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223684574, "dur": 11,
    "args": {
      "External id": 144904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144904,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144904, "pid": 5, "tid": 7, "ts": 1716454223684574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632166, "dur": 14,
    "args": {
      "External id": 144904,
      "cbid": 211, "correlation": 144904
    }
  },
  {
    "ph": "s", "id": 144904, "pid": 76337, "tid": -914061504, "ts": 1716454223632166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223684587, "dur": 12,
    "args": {
      "External id": 144924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144924,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 144924, "pid": 5, "tid": 7, "ts": 1716454223684587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632248, "dur": 11,
    "args": {
      "External id": 144924,
      "cbid": 211, "correlation": 144924
    }
  },
  {
    "ph": "s", "id": 144924, "pid": 76337, "tid": -914061504, "ts": 1716454223632248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223684600, "dur": 4,
    "args": {
      "External id": 144936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144936,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144936, "pid": 5, "tid": 7, "ts": 1716454223684600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632269, "dur": 7,
    "args": {
      "External id": 144936,
      "cbid": 211, "correlation": 144936
    }
  },
  {
    "ph": "s", "id": 144936, "pid": 76337, "tid": -914061504, "ts": 1716454223632269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223684606, "dur": 11,
    "args": {
      "External id": 144939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144939,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144939, "pid": 5, "tid": 7, "ts": 1716454223684606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632290, "dur": 6,
    "args": {
      "External id": 144939,
      "cbid": 211, "correlation": 144939
    }
  },
  {
    "ph": "s", "id": 144939, "pid": 76337, "tid": -914061504, "ts": 1716454223632290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223684617, "dur": 6,
    "args": {
      "External id": 144948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144948,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144948, "pid": 5, "tid": 7, "ts": 1716454223684617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632331, "dur": 9,
    "args": {
      "External id": 144948,
      "cbid": 211, "correlation": 144948
    }
  },
  {
    "ph": "s", "id": 144948, "pid": 76337, "tid": -914061504, "ts": 1716454223632331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632385, "dur": 0,
    "args": {
      "External id": 144958,
      "cbid": 317, "correlation": 144958
    }
  },
  {
    "ph": "f", "id": 144958, "pid": 76337, "tid": -914061504, "ts": 1716454223632385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632386, "dur": 0,
    "args": {
      "External id": 144959,
      "cbid": 203, "correlation": 144959
    }
  },
  {
    "ph": "f", "id": 144959, "pid": 76337, "tid": -914061504, "ts": 1716454223632386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632387, "dur": 0,
    "args": {
      "External id": 144960,
      "cbid": 205, "correlation": 144960
    }
  },
  {
    "ph": "f", "id": 144960, "pid": 76337, "tid": -914061504, "ts": 1716454223632387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223684625, "dur": 7,
    "args": {
      "External id": 144964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144964,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144964, "pid": 5, "tid": 7, "ts": 1716454223684625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632410, "dur": 12,
    "args": {
      "External id": 144964,
      "cbid": 211, "correlation": 144964
    }
  },
  {
    "ph": "s", "id": 144964, "pid": 76337, "tid": -914061504, "ts": 1716454223632410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223684633, "dur": 321,
    "args": {
      "External id": 144966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144966,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144966, "pid": 5, "tid": 7, "ts": 1716454223684633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632425, "dur": 6,
    "args": {
      "External id": 144966,
      "cbid": 211, "correlation": 144966
    }
  },
  {
    "ph": "s", "id": 144966, "pid": 76337, "tid": -914061504, "ts": 1716454223632425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223684956, "dur": 1,
    "args": {
      "External id": 144968,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 144968,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 144968, "pid": 5, "tid": 7, "ts": 1716454223684956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632437, "dur": 7,
    "args": {
      "External id": 144968,
      "cbid": 51, "correlation": 144968
    }
  },
  {
    "ph": "s", "id": 144968, "pid": 76337, "tid": -914061504, "ts": 1716454223632437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223684960, "dur": 499,
    "args": {
      "External id": 144969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144969,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 144969, "pid": 5, "tid": 7, "ts": 1716454223684960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632445, "dur": 6,
    "args": {
      "External id": 144969,
      "cbid": 211, "correlation": 144969
    }
  },
  {
    "ph": "s", "id": 144969, "pid": 76337, "tid": -914061504, "ts": 1716454223632445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223685460, "dur": 5,
    "args": {
      "External id": 144971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144971,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 144971, "pid": 5, "tid": 7, "ts": 1716454223685460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632454, "dur": 5,
    "args": {
      "External id": 144971,
      "cbid": 211, "correlation": 144971
    }
  },
  {
    "ph": "s", "id": 144971, "pid": 76337, "tid": -914061504, "ts": 1716454223632454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223685467, "dur": 6,
    "args": {
      "External id": 144977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 144977, "pid": 5, "tid": 7, "ts": 1716454223685467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632483, "dur": 8,
    "args": {
      "External id": 144977,
      "cbid": 211, "correlation": 144977
    }
  },
  {
    "ph": "s", "id": 144977, "pid": 76337, "tid": -914061504, "ts": 1716454223632483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223685474, "dur": 4,
    "args": {
      "External id": 144985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 144985,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 144985, "pid": 5, "tid": 7, "ts": 1716454223685474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632526, "dur": 9,
    "args": {
      "External id": 144985,
      "cbid": 211, "correlation": 144985
    }
  },
  {
    "ph": "s", "id": 144985, "pid": 76337, "tid": -914061504, "ts": 1716454223632526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632608, "dur": 2,
    "args": {
      "External id": 145001,
      "cbid": 251, "correlation": 145001
    }
  },
  {
    "ph": "f", "id": 145001, "pid": 76337, "tid": -914061504, "ts": 1716454223632608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632614, "dur": 0,
    "args": {
      "External id": 145003,
      "cbid": 251, "correlation": 145003
    }
  },
  {
    "ph": "f", "id": 145003, "pid": 76337, "tid": -914061504, "ts": 1716454223632614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223685480, "dur": 12,
    "args": {
      "External id": 145004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145004,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145004, "pid": 5, "tid": 7, "ts": 1716454223685480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632616, "dur": 14,
    "args": {
      "External id": 145004,
      "cbid": 211, "correlation": 145004
    }
  },
  {
    "ph": "s", "id": 145004, "pid": 76337, "tid": -914061504, "ts": 1716454223632616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223685494, "dur": 5,
    "args": {
      "External id": 145006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145006,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145006, "pid": 5, "tid": 7, "ts": 1716454223685494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632633, "dur": 6,
    "args": {
      "External id": 145006,
      "cbid": 211, "correlation": 145006
    }
  },
  {
    "ph": "s", "id": 145006, "pid": 76337, "tid": -914061504, "ts": 1716454223632633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223685500, "dur": 6,
    "args": {
      "External id": 145016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145016, "pid": 5, "tid": 7, "ts": 1716454223685500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632693, "dur": 12,
    "args": {
      "External id": 145016,
      "cbid": 211, "correlation": 145016
    }
  },
  {
    "ph": "s", "id": 145016, "pid": 76337, "tid": -914061504, "ts": 1716454223632693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223685507, "dur": 9,
    "args": {
      "External id": 145036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145036,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 145036, "pid": 5, "tid": 7, "ts": 1716454223685507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632759, "dur": 11,
    "args": {
      "External id": 145036,
      "cbid": 211, "correlation": 145036
    }
  },
  {
    "ph": "s", "id": 145036, "pid": 76337, "tid": -914061504, "ts": 1716454223632759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223685518, "dur": 4,
    "args": {
      "External id": 145048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145048,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 145048, "pid": 5, "tid": 7, "ts": 1716454223685518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632781, "dur": 6,
    "args": {
      "External id": 145048,
      "cbid": 211, "correlation": 145048
    }
  },
  {
    "ph": "s", "id": 145048, "pid": 76337, "tid": -914061504, "ts": 1716454223632781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223685524, "dur": 7,
    "args": {
      "External id": 145051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145051, "pid": 5, "tid": 7, "ts": 1716454223685524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632798, "dur": 6,
    "args": {
      "External id": 145051,
      "cbid": 211, "correlation": 145051
    }
  },
  {
    "ph": "s", "id": 145051, "pid": 76337, "tid": -914061504, "ts": 1716454223632798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223685532, "dur": 5,
    "args": {
      "External id": 145060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145060,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145060, "pid": 5, "tid": 7, "ts": 1716454223685532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632838, "dur": 9,
    "args": {
      "External id": 145060,
      "cbid": 211, "correlation": 145060
    }
  },
  {
    "ph": "s", "id": 145060, "pid": 76337, "tid": -914061504, "ts": 1716454223632838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632903, "dur": 0,
    "args": {
      "External id": 145070,
      "cbid": 317, "correlation": 145070
    }
  },
  {
    "ph": "f", "id": 145070, "pid": 76337, "tid": -914061504, "ts": 1716454223632903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632904, "dur": 0,
    "args": {
      "External id": 145071,
      "cbid": 203, "correlation": 145071
    }
  },
  {
    "ph": "f", "id": 145071, "pid": 76337, "tid": -914061504, "ts": 1716454223632904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632905, "dur": 0,
    "args": {
      "External id": 145072,
      "cbid": 205, "correlation": 145072
    }
  },
  {
    "ph": "f", "id": 145072, "pid": 76337, "tid": -914061504, "ts": 1716454223632905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223685537, "dur": 5,
    "args": {
      "External id": 145076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145076, "pid": 5, "tid": 7, "ts": 1716454223685537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632920, "dur": 13,
    "args": {
      "External id": 145076,
      "cbid": 211, "correlation": 145076
    }
  },
  {
    "ph": "s", "id": 145076, "pid": 76337, "tid": -914061504, "ts": 1716454223632920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223685544, "dur": 162,
    "args": {
      "External id": 145078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145078,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145078, "pid": 5, "tid": 7, "ts": 1716454223685544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632935, "dur": 5,
    "args": {
      "External id": 145078,
      "cbid": 211, "correlation": 145078
    }
  },
  {
    "ph": "s", "id": 145078, "pid": 76337, "tid": -914061504, "ts": 1716454223632935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223685708, "dur": 1,
    "args": {
      "External id": 145080,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145080,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 145080, "pid": 5, "tid": 7, "ts": 1716454223685708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632945, "dur": 6,
    "args": {
      "External id": 145080,
      "cbid": 51, "correlation": 145080
    }
  },
  {
    "ph": "s", "id": 145080, "pid": 76337, "tid": -914061504, "ts": 1716454223632945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223685712, "dur": 260,
    "args": {
      "External id": 145081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145081,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145081, "pid": 5, "tid": 7, "ts": 1716454223685712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632952, "dur": 6,
    "args": {
      "External id": 145081,
      "cbid": 211, "correlation": 145081
    }
  },
  {
    "ph": "s", "id": 145081, "pid": 76337, "tid": -914061504, "ts": 1716454223632952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223685973, "dur": 6,
    "args": {
      "External id": 145083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145083,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145083, "pid": 5, "tid": 7, "ts": 1716454223685973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223632962, "dur": 5,
    "args": {
      "External id": 145083,
      "cbid": 211, "correlation": 145083
    }
  },
  {
    "ph": "s", "id": 145083, "pid": 76337, "tid": -914061504, "ts": 1716454223632962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223685980, "dur": 6,
    "args": {
      "External id": 145089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145089, "pid": 5, "tid": 7, "ts": 1716454223685980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633000, "dur": 9,
    "args": {
      "External id": 145089,
      "cbid": 211, "correlation": 145089
    }
  },
  {
    "ph": "s", "id": 145089, "pid": 76337, "tid": -914061504, "ts": 1716454223633000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633061, "dur": 0,
    "args": {
      "External id": 145099,
      "cbid": 317, "correlation": 145099
    }
  },
  {
    "ph": "f", "id": 145099, "pid": 76337, "tid": -914061504, "ts": 1716454223633061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633061, "dur": 0,
    "args": {
      "External id": 145100,
      "cbid": 203, "correlation": 145100
    }
  },
  {
    "ph": "f", "id": 145100, "pid": 76337, "tid": -914061504, "ts": 1716454223633061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633062, "dur": 0,
    "args": {
      "External id": 145101,
      "cbid": 205, "correlation": 145101
    }
  },
  {
    "ph": "f", "id": 145101, "pid": 76337, "tid": -914061504, "ts": 1716454223633062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223685987, "dur": 8,
    "args": {
      "External id": 145105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145105,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145105, "pid": 5, "tid": 7, "ts": 1716454223685987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633077, "dur": 11,
    "args": {
      "External id": 145105,
      "cbid": 211, "correlation": 145105
    }
  },
  {
    "ph": "s", "id": 145105, "pid": 76337, "tid": -914061504, "ts": 1716454223633077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223685997, "dur": 3,
    "args": {
      "External id": 145107,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145107,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 145107, "pid": 5, "tid": 7, "ts": 1716454223685997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633093, "dur": 10,
    "args": {
      "External id": 145107,
      "cbid": 51, "correlation": 145107
    }
  },
  {
    "ph": "s", "id": 145107, "pid": 76337, "tid": -914061504, "ts": 1716454223633093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223686001, "dur": 95,
    "args": {
      "External id": 145108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145108,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 145108, "pid": 5, "tid": 7, "ts": 1716454223686001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633104, "dur": 6,
    "args": {
      "External id": 145108,
      "cbid": 211, "correlation": 145108
    }
  },
  {
    "ph": "s", "id": 145108, "pid": 76337, "tid": -914061504, "ts": 1716454223633104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223686098, "dur": 5,
    "args": {
      "External id": 145110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145110,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145110, "pid": 5, "tid": 7, "ts": 1716454223686098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633113, "dur": 5,
    "args": {
      "External id": 145110,
      "cbid": 211, "correlation": 145110
    }
  },
  {
    "ph": "s", "id": 145110, "pid": 76337, "tid": -914061504, "ts": 1716454223633113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223686105, "dur": 6,
    "args": {
      "External id": 145116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145116,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145116, "pid": 5, "tid": 7, "ts": 1716454223686105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633140, "dur": 8,
    "args": {
      "External id": 145116,
      "cbid": 211, "correlation": 145116
    }
  },
  {
    "ph": "s", "id": 145116, "pid": 76337, "tid": -914061504, "ts": 1716454223633140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223686112, "dur": 5,
    "args": {
      "External id": 145124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145124,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145124, "pid": 5, "tid": 7, "ts": 1716454223686112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633169, "dur": 9,
    "args": {
      "External id": 145124,
      "cbid": 211, "correlation": 145124
    }
  },
  {
    "ph": "s", "id": 145124, "pid": 76337, "tid": -914061504, "ts": 1716454223633169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223686118, "dur": 4,
    "args": {
      "External id": 145132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145132,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145132, "pid": 5, "tid": 7, "ts": 1716454223686118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633199, "dur": 8,
    "args": {
      "External id": 145132,
      "cbid": 211, "correlation": 145132
    }
  },
  {
    "ph": "s", "id": 145132, "pid": 76337, "tid": -914061504, "ts": 1716454223633199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223686124, "dur": 11,
    "args": {
      "External id": 145141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145141,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145141, "pid": 5, "tid": 7, "ts": 1716454223686124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633243, "dur": 10,
    "args": {
      "External id": 145141,
      "cbid": 211, "correlation": 145141
    }
  },
  {
    "ph": "s", "id": 145141, "pid": 76337, "tid": -914061504, "ts": 1716454223633243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223686137, "dur": 13,
    "args": {
      "External id": 145161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145161,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 145161, "pid": 5, "tid": 7, "ts": 1716454223686137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633313, "dur": 11,
    "args": {
      "External id": 145161,
      "cbid": 211, "correlation": 145161
    }
  },
  {
    "ph": "s", "id": 145161, "pid": 76337, "tid": -914061504, "ts": 1716454223633313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223686150, "dur": 4,
    "args": {
      "External id": 145173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145173,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145173, "pid": 5, "tid": 7, "ts": 1716454223686150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633334, "dur": 6,
    "args": {
      "External id": 145173,
      "cbid": 211, "correlation": 145173
    }
  },
  {
    "ph": "s", "id": 145173, "pid": 76337, "tid": -914061504, "ts": 1716454223633334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223686156, "dur": 10,
    "args": {
      "External id": 145176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145176, "pid": 5, "tid": 7, "ts": 1716454223686156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633352, "dur": 7,
    "args": {
      "External id": 145176,
      "cbid": 211, "correlation": 145176
    }
  },
  {
    "ph": "s", "id": 145176, "pid": 76337, "tid": -914061504, "ts": 1716454223633352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223686168, "dur": 6,
    "args": {
      "External id": 145185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145185,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145185, "pid": 5, "tid": 7, "ts": 1716454223686168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633390, "dur": 9,
    "args": {
      "External id": 145185,
      "cbid": 211, "correlation": 145185
    }
  },
  {
    "ph": "s", "id": 145185, "pid": 76337, "tid": -914061504, "ts": 1716454223633390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633442, "dur": 0,
    "args": {
      "External id": 145195,
      "cbid": 317, "correlation": 145195
    }
  },
  {
    "ph": "f", "id": 145195, "pid": 76337, "tid": -914061504, "ts": 1716454223633442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633443, "dur": 0,
    "args": {
      "External id": 145196,
      "cbid": 203, "correlation": 145196
    }
  },
  {
    "ph": "f", "id": 145196, "pid": 76337, "tid": -914061504, "ts": 1716454223633443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633443, "dur": 0,
    "args": {
      "External id": 145197,
      "cbid": 205, "correlation": 145197
    }
  },
  {
    "ph": "f", "id": 145197, "pid": 76337, "tid": -914061504, "ts": 1716454223633443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223686175, "dur": 7,
    "args": {
      "External id": 145201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145201,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145201, "pid": 5, "tid": 7, "ts": 1716454223686175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633458, "dur": 11,
    "args": {
      "External id": 145201,
      "cbid": 211, "correlation": 145201
    }
  },
  {
    "ph": "s", "id": 145201, "pid": 76337, "tid": -914061504, "ts": 1716454223633458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223686183, "dur": 319,
    "args": {
      "External id": 145203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145203,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145203, "pid": 5, "tid": 7, "ts": 1716454223686183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633471, "dur": 6,
    "args": {
      "External id": 145203,
      "cbid": 211, "correlation": 145203
    }
  },
  {
    "ph": "s", "id": 145203, "pid": 76337, "tid": -914061504, "ts": 1716454223633471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223686504, "dur": 1,
    "args": {
      "External id": 145205,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145205,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 145205, "pid": 5, "tid": 7, "ts": 1716454223686504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633483, "dur": 6,
    "args": {
      "External id": 145205,
      "cbid": 51, "correlation": 145205
    }
  },
  {
    "ph": "s", "id": 145205, "pid": 76337, "tid": -914061504, "ts": 1716454223633483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223686508, "dur": 496,
    "args": {
      "External id": 145206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145206,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145206, "pid": 5, "tid": 7, "ts": 1716454223686508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633490, "dur": 6,
    "args": {
      "External id": 145206,
      "cbid": 211, "correlation": 145206
    }
  },
  {
    "ph": "s", "id": 145206, "pid": 76337, "tid": -914061504, "ts": 1716454223633490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687006, "dur": 5,
    "args": {
      "External id": 145208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145208,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145208, "pid": 5, "tid": 7, "ts": 1716454223687006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633500, "dur": 5,
    "args": {
      "External id": 145208,
      "cbid": 211, "correlation": 145208
    }
  },
  {
    "ph": "s", "id": 145208, "pid": 76337, "tid": -914061504, "ts": 1716454223633500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223687013, "dur": 7,
    "args": {
      "External id": 145214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145214, "pid": 5, "tid": 7, "ts": 1716454223687013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633528, "dur": 9,
    "args": {
      "External id": 145214,
      "cbid": 211, "correlation": 145214
    }
  },
  {
    "ph": "s", "id": 145214, "pid": 76337, "tid": -914061504, "ts": 1716454223633528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223687021, "dur": 4,
    "args": {
      "External id": 145222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145222,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 145222, "pid": 5, "tid": 7, "ts": 1716454223687021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633572, "dur": 9,
    "args": {
      "External id": 145222,
      "cbid": 211, "correlation": 145222
    }
  },
  {
    "ph": "s", "id": 145222, "pid": 76337, "tid": -914061504, "ts": 1716454223633572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633635, "dur": 1,
    "args": {
      "External id": 145238,
      "cbid": 251, "correlation": 145238
    }
  },
  {
    "ph": "f", "id": 145238, "pid": 76337, "tid": -914061504, "ts": 1716454223633635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633640, "dur": 0,
    "args": {
      "External id": 145240,
      "cbid": 251, "correlation": 145240
    }
  },
  {
    "ph": "f", "id": 145240, "pid": 76337, "tid": -914061504, "ts": 1716454223633640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223687026, "dur": 13,
    "args": {
      "External id": 145241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145241,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145241, "pid": 5, "tid": 7, "ts": 1716454223687026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633642, "dur": 12,
    "args": {
      "External id": 145241,
      "cbid": 211, "correlation": 145241
    }
  },
  {
    "ph": "s", "id": 145241, "pid": 76337, "tid": -914061504, "ts": 1716454223633642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223687040, "dur": 5,
    "args": {
      "External id": 145243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145243,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145243, "pid": 5, "tid": 7, "ts": 1716454223687040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633656, "dur": 5,
    "args": {
      "External id": 145243,
      "cbid": 211, "correlation": 145243
    }
  },
  {
    "ph": "s", "id": 145243, "pid": 76337, "tid": -914061504, "ts": 1716454223633656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223687047, "dur": 6,
    "args": {
      "External id": 145253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145253, "pid": 5, "tid": 7, "ts": 1716454223687047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633712, "dur": 12,
    "args": {
      "External id": 145253,
      "cbid": 211, "correlation": 145253
    }
  },
  {
    "ph": "s", "id": 145253, "pid": 76337, "tid": -914061504, "ts": 1716454223633712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687054, "dur": 9,
    "args": {
      "External id": 145273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145273,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 145273, "pid": 5, "tid": 7, "ts": 1716454223687054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633777, "dur": 11,
    "args": {
      "External id": 145273,
      "cbid": 211, "correlation": 145273
    }
  },
  {
    "ph": "s", "id": 145273, "pid": 76337, "tid": -914061504, "ts": 1716454223633777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223687064, "dur": 4,
    "args": {
      "External id": 145285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145285,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 145285, "pid": 5, "tid": 7, "ts": 1716454223687064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633798, "dur": 6,
    "args": {
      "External id": 145285,
      "cbid": 211, "correlation": 145285
    }
  },
  {
    "ph": "s", "id": 145285, "pid": 76337, "tid": -914061504, "ts": 1716454223633798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223687070, "dur": 7,
    "args": {
      "External id": 145288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145288,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145288, "pid": 5, "tid": 7, "ts": 1716454223687070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633816, "dur": 6,
    "args": {
      "External id": 145288,
      "cbid": 211, "correlation": 145288
    }
  },
  {
    "ph": "s", "id": 145288, "pid": 76337, "tid": -914061504, "ts": 1716454223633816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223687078, "dur": 5,
    "args": {
      "External id": 145297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145297,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145297, "pid": 5, "tid": 7, "ts": 1716454223687078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633856, "dur": 9,
    "args": {
      "External id": 145297,
      "cbid": 211, "correlation": 145297
    }
  },
  {
    "ph": "s", "id": 145297, "pid": 76337, "tid": -914061504, "ts": 1716454223633856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633919, "dur": 0,
    "args": {
      "External id": 145307,
      "cbid": 317, "correlation": 145307
    }
  },
  {
    "ph": "f", "id": 145307, "pid": 76337, "tid": -914061504, "ts": 1716454223633919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633920, "dur": 0,
    "args": {
      "External id": 145308,
      "cbid": 203, "correlation": 145308
    }
  },
  {
    "ph": "f", "id": 145308, "pid": 76337, "tid": -914061504, "ts": 1716454223633920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633921, "dur": 0,
    "args": {
      "External id": 145309,
      "cbid": 205, "correlation": 145309
    }
  },
  {
    "ph": "f", "id": 145309, "pid": 76337, "tid": -914061504, "ts": 1716454223633921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687084, "dur": 5,
    "args": {
      "External id": 145313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145313, "pid": 5, "tid": 7, "ts": 1716454223687084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633934, "dur": 13,
    "args": {
      "External id": 145313,
      "cbid": 211, "correlation": 145313
    }
  },
  {
    "ph": "s", "id": 145313, "pid": 76337, "tid": -914061504, "ts": 1716454223633934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687090, "dur": 162,
    "args": {
      "External id": 145315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145315,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145315, "pid": 5, "tid": 7, "ts": 1716454223687090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633949, "dur": 5,
    "args": {
      "External id": 145315,
      "cbid": 211, "correlation": 145315
    }
  },
  {
    "ph": "s", "id": 145315, "pid": 76337, "tid": -914061504, "ts": 1716454223633949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223687255, "dur": 1,
    "args": {
      "External id": 145317,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145317,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 145317, "pid": 5, "tid": 7, "ts": 1716454223687255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633960, "dur": 6,
    "args": {
      "External id": 145317,
      "cbid": 51, "correlation": 145317
    }
  },
  {
    "ph": "s", "id": 145317, "pid": 76337, "tid": -914061504, "ts": 1716454223633960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223687258, "dur": 260,
    "args": {
      "External id": 145318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145318,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145318, "pid": 5, "tid": 7, "ts": 1716454223687258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633967, "dur": 14,
    "args": {
      "External id": 145318,
      "cbid": 211, "correlation": 145318
    }
  },
  {
    "ph": "s", "id": 145318, "pid": 76337, "tid": -914061504, "ts": 1716454223633967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687520, "dur": 6,
    "args": {
      "External id": 145320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145320,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145320, "pid": 5, "tid": 7, "ts": 1716454223687520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223633985, "dur": 6,
    "args": {
      "External id": 145320,
      "cbid": 211, "correlation": 145320
    }
  },
  {
    "ph": "s", "id": 145320, "pid": 76337, "tid": -914061504, "ts": 1716454223633985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223687527, "dur": 6,
    "args": {
      "External id": 145326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145326, "pid": 5, "tid": 7, "ts": 1716454223687527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634015, "dur": 8,
    "args": {
      "External id": 145326,
      "cbid": 211, "correlation": 145326
    }
  },
  {
    "ph": "s", "id": 145326, "pid": 76337, "tid": -914061504, "ts": 1716454223634015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634074, "dur": 0,
    "args": {
      "External id": 145336,
      "cbid": 317, "correlation": 145336
    }
  },
  {
    "ph": "f", "id": 145336, "pid": 76337, "tid": -914061504, "ts": 1716454223634074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634074, "dur": 0,
    "args": {
      "External id": 145337,
      "cbid": 203, "correlation": 145337
    }
  },
  {
    "ph": "f", "id": 145337, "pid": 76337, "tid": -914061504, "ts": 1716454223634074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634075, "dur": 0,
    "args": {
      "External id": 145338,
      "cbid": 205, "correlation": 145338
    }
  },
  {
    "ph": "f", "id": 145338, "pid": 76337, "tid": -914061504, "ts": 1716454223634075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687535, "dur": 8,
    "args": {
      "External id": 145342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145342,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145342, "pid": 5, "tid": 7, "ts": 1716454223687535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634087, "dur": 13,
    "args": {
      "External id": 145342,
      "cbid": 211, "correlation": 145342
    }
  },
  {
    "ph": "s", "id": 145342, "pid": 76337, "tid": -914061504, "ts": 1716454223634087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223687544, "dur": 3,
    "args": {
      "External id": 145344,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145344,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 145344, "pid": 5, "tid": 7, "ts": 1716454223687544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634105, "dur": 9,
    "args": {
      "External id": 145344,
      "cbid": 51, "correlation": 145344
    }
  },
  {
    "ph": "s", "id": 145344, "pid": 76337, "tid": -914061504, "ts": 1716454223634105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223687548, "dur": 93,
    "args": {
      "External id": 145345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145345,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 145345, "pid": 5, "tid": 7, "ts": 1716454223687548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634115, "dur": 6,
    "args": {
      "External id": 145345,
      "cbid": 211, "correlation": 145345
    }
  },
  {
    "ph": "s", "id": 145345, "pid": 76337, "tid": -914061504, "ts": 1716454223634115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687642, "dur": 5,
    "args": {
      "External id": 145347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145347,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145347, "pid": 5, "tid": 7, "ts": 1716454223687642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634124, "dur": 5,
    "args": {
      "External id": 145347,
      "cbid": 211, "correlation": 145347
    }
  },
  {
    "ph": "s", "id": 145347, "pid": 76337, "tid": -914061504, "ts": 1716454223634124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223687649, "dur": 6,
    "args": {
      "External id": 145353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145353, "pid": 5, "tid": 7, "ts": 1716454223687649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634152, "dur": 8,
    "args": {
      "External id": 145353,
      "cbid": 211, "correlation": 145353
    }
  },
  {
    "ph": "s", "id": 145353, "pid": 76337, "tid": -914061504, "ts": 1716454223634152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223687656, "dur": 5,
    "args": {
      "External id": 145361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145361,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145361, "pid": 5, "tid": 7, "ts": 1716454223687656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634180, "dur": 9,
    "args": {
      "External id": 145361,
      "cbid": 211, "correlation": 145361
    }
  },
  {
    "ph": "s", "id": 145361, "pid": 76337, "tid": -914061504, "ts": 1716454223634180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223687663, "dur": 4,
    "args": {
      "External id": 145369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145369,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 145369, "pid": 5, "tid": 7, "ts": 1716454223687663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634209, "dur": 8,
    "args": {
      "External id": 145369,
      "cbid": 211, "correlation": 145369
    }
  },
  {
    "ph": "s", "id": 145369, "pid": 76337, "tid": -914061504, "ts": 1716454223634209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223687669, "dur": 14,
    "args": {
      "External id": 145380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145380,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145380, "pid": 5, "tid": 7, "ts": 1716454223687669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634300, "dur": 12,
    "args": {
      "External id": 145380,
      "cbid": 211, "correlation": 145380
    }
  },
  {
    "ph": "s", "id": 145380, "pid": 76337, "tid": -914061504, "ts": 1716454223634300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634356, "dur": 0,
    "args": {
      "External id": 145390,
      "cbid": 317, "correlation": 145390
    }
  },
  {
    "ph": "f", "id": 145390, "pid": 76337, "tid": -914061504, "ts": 1716454223634356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634357, "dur": 0,
    "args": {
      "External id": 145391,
      "cbid": 203, "correlation": 145391
    }
  },
  {
    "ph": "f", "id": 145391, "pid": 76337, "tid": -914061504, "ts": 1716454223634357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634358, "dur": 0,
    "args": {
      "External id": 145392,
      "cbid": 205, "correlation": 145392
    }
  },
  {
    "ph": "f", "id": 145392, "pid": 76337, "tid": -914061504, "ts": 1716454223634358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687684, "dur": 9,
    "args": {
      "External id": 145396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145396,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145396, "pid": 5, "tid": 7, "ts": 1716454223687684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634373, "dur": 12,
    "args": {
      "External id": 145396,
      "cbid": 211, "correlation": 145396
    }
  },
  {
    "ph": "s", "id": 145396, "pid": 76337, "tid": -914061504, "ts": 1716454223634373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223687694, "dur": 164,
    "args": {
      "External id": 145398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145398,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145398, "pid": 5, "tid": 7, "ts": 1716454223687694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634388, "dur": 5,
    "args": {
      "External id": 145398,
      "cbid": 211, "correlation": 145398
    }
  },
  {
    "ph": "s", "id": 145398, "pid": 76337, "tid": -914061504, "ts": 1716454223634388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223687860, "dur": 1,
    "args": {
      "External id": 145400,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145400,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 145400, "pid": 5, "tid": 7, "ts": 1716454223687860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634399, "dur": 6,
    "args": {
      "External id": 145400,
      "cbid": 51, "correlation": 145400
    }
  },
  {
    "ph": "s", "id": 145400, "pid": 76337, "tid": -914061504, "ts": 1716454223634399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223687863, "dur": 650,
    "args": {
      "External id": 145401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145401,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145401, "pid": 5, "tid": 7, "ts": 1716454223687863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634407, "dur": 6,
    "args": {
      "External id": 145401,
      "cbid": 211, "correlation": 145401
    }
  },
  {
    "ph": "s", "id": 145401, "pid": 76337, "tid": -914061504, "ts": 1716454223634407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223688514, "dur": 12,
    "args": {
      "External id": 145403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145403,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145403, "pid": 5, "tid": 7, "ts": 1716454223688514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634417, "dur": 5,
    "args": {
      "External id": 145403,
      "cbid": 211, "correlation": 145403
    }
  },
  {
    "ph": "s", "id": 145403, "pid": 76337, "tid": -914061504, "ts": 1716454223634417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223688528, "dur": 14,
    "args": {
      "External id": 145409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145409, "pid": 5, "tid": 7, "ts": 1716454223688528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634445, "dur": 8,
    "args": {
      "External id": 145409,
      "cbid": 211, "correlation": 145409
    }
  },
  {
    "ph": "s", "id": 145409, "pid": 76337, "tid": -914061504, "ts": 1716454223634445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223688543, "dur": 31,
    "args": {
      "External id": 145418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145418,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145418, "pid": 5, "tid": 7, "ts": 1716454223688543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634557, "dur": 12,
    "args": {
      "External id": 145418,
      "cbid": 211, "correlation": 145418
    }
  },
  {
    "ph": "s", "id": 145418, "pid": 76337, "tid": -914061504, "ts": 1716454223634557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223688575, "dur": 30,
    "args": {
      "External id": 145438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145438,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 145438, "pid": 5, "tid": 7, "ts": 1716454223688575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634627, "dur": 13,
    "args": {
      "External id": 145438,
      "cbid": 211, "correlation": 145438
    }
  },
  {
    "ph": "s", "id": 145438, "pid": 76337, "tid": -914061504, "ts": 1716454223634627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223688607, "dur": 4,
    "args": {
      "External id": 145450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145450,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145450, "pid": 5, "tid": 7, "ts": 1716454223688607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634649, "dur": 6,
    "args": {
      "External id": 145450,
      "cbid": 211, "correlation": 145450
    }
  },
  {
    "ph": "s", "id": 145450, "pid": 76337, "tid": -914061504, "ts": 1716454223634649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223688612, "dur": 30,
    "args": {
      "External id": 145453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145453, "pid": 5, "tid": 7, "ts": 1716454223688612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634668, "dur": 6,
    "args": {
      "External id": 145453,
      "cbid": 211, "correlation": 145453
    }
  },
  {
    "ph": "s", "id": 145453, "pid": 76337, "tid": -914061504, "ts": 1716454223634668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223688644, "dur": 20,
    "args": {
      "External id": 145462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145462,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145462, "pid": 5, "tid": 7, "ts": 1716454223688644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634706, "dur": 9,
    "args": {
      "External id": 145462,
      "cbid": 211, "correlation": 145462
    }
  },
  {
    "ph": "s", "id": 145462, "pid": 76337, "tid": -914061504, "ts": 1716454223634706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634758, "dur": 0,
    "args": {
      "External id": 145472,
      "cbid": 317, "correlation": 145472
    }
  },
  {
    "ph": "f", "id": 145472, "pid": 76337, "tid": -914061504, "ts": 1716454223634758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634759, "dur": 0,
    "args": {
      "External id": 145473,
      "cbid": 203, "correlation": 145473
    }
  },
  {
    "ph": "f", "id": 145473, "pid": 76337, "tid": -914061504, "ts": 1716454223634759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634760, "dur": 0,
    "args": {
      "External id": 145474,
      "cbid": 205, "correlation": 145474
    }
  },
  {
    "ph": "f", "id": 145474, "pid": 76337, "tid": -914061504, "ts": 1716454223634760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223688665, "dur": 22,
    "args": {
      "External id": 145478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145478, "pid": 5, "tid": 7, "ts": 1716454223688665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634774, "dur": 11,
    "args": {
      "External id": 145478,
      "cbid": 211, "correlation": 145478
    }
  },
  {
    "ph": "s", "id": 145478, "pid": 76337, "tid": -914061504, "ts": 1716454223634774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223688688, "dur": 321,
    "args": {
      "External id": 145480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145480,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145480, "pid": 5, "tid": 7, "ts": 1716454223688688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634788, "dur": 5,
    "args": {
      "External id": 145480,
      "cbid": 211, "correlation": 145480
    }
  },
  {
    "ph": "s", "id": 145480, "pid": 76337, "tid": -914061504, "ts": 1716454223634788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223689011, "dur": 1,
    "args": {
      "External id": 145482,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145482,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 145482, "pid": 5, "tid": 7, "ts": 1716454223689011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634799, "dur": 6,
    "args": {
      "External id": 145482,
      "cbid": 51, "correlation": 145482
    }
  },
  {
    "ph": "s", "id": 145482, "pid": 76337, "tid": -914061504, "ts": 1716454223634799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223689015, "dur": 1244,
    "args": {
      "External id": 145483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145483,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145483, "pid": 5, "tid": 7, "ts": 1716454223689015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634806, "dur": 6,
    "args": {
      "External id": 145483,
      "cbid": 211, "correlation": 145483
    }
  },
  {
    "ph": "s", "id": 145483, "pid": 76337, "tid": -914061504, "ts": 1716454223634806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223690260, "dur": 12,
    "args": {
      "External id": 145485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145485,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145485, "pid": 5, "tid": 7, "ts": 1716454223690260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634815, "dur": 5,
    "args": {
      "External id": 145485,
      "cbid": 211, "correlation": 145485
    }
  },
  {
    "ph": "s", "id": 145485, "pid": 76337, "tid": -914061504, "ts": 1716454223634815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223690273, "dur": 15,
    "args": {
      "External id": 145491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145491, "pid": 5, "tid": 7, "ts": 1716454223690273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634844, "dur": 9,
    "args": {
      "External id": 145491,
      "cbid": 211, "correlation": 145491
    }
  },
  {
    "ph": "s", "id": 145491, "pid": 76337, "tid": -914061504, "ts": 1716454223634844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223690289, "dur": 4,
    "args": {
      "External id": 145499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145499,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 145499, "pid": 5, "tid": 7, "ts": 1716454223690289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634889, "dur": 10,
    "args": {
      "External id": 145499,
      "cbid": 211, "correlation": 145499
    }
  },
  {
    "ph": "s", "id": 145499, "pid": 76337, "tid": -914061504, "ts": 1716454223634889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634951, "dur": 1,
    "args": {
      "External id": 145515,
      "cbid": 251, "correlation": 145515
    }
  },
  {
    "ph": "f", "id": 145515, "pid": 76337, "tid": -914061504, "ts": 1716454223634951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634957, "dur": 0,
    "args": {
      "External id": 145517,
      "cbid": 251, "correlation": 145517
    }
  },
  {
    "ph": "f", "id": 145517, "pid": 76337, "tid": -914061504, "ts": 1716454223634957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223690295, "dur": 12,
    "args": {
      "External id": 145518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145518,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145518, "pid": 5, "tid": 7, "ts": 1716454223690295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634959, "dur": 11,
    "args": {
      "External id": 145518,
      "cbid": 211, "correlation": 145518
    }
  },
  {
    "ph": "s", "id": 145518, "pid": 76337, "tid": -914061504, "ts": 1716454223634959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223690308, "dur": 5,
    "args": {
      "External id": 145520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145520,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145520, "pid": 5, "tid": 7, "ts": 1716454223690308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223634971, "dur": 14,
    "args": {
      "External id": 145520,
      "cbid": 211, "correlation": 145520
    }
  },
  {
    "ph": "s", "id": 145520, "pid": 76337, "tid": -914061504, "ts": 1716454223634971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223690314, "dur": 17,
    "args": {
      "External id": 145530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145530, "pid": 5, "tid": 7, "ts": 1716454223690314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635037, "dur": 13,
    "args": {
      "External id": 145530,
      "cbid": 211, "correlation": 145530
    }
  },
  {
    "ph": "s", "id": 145530, "pid": 76337, "tid": -914061504, "ts": 1716454223635037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223690333, "dur": 19,
    "args": {
      "External id": 145550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145550,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 145550, "pid": 5, "tid": 7, "ts": 1716454223690333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635103, "dur": 10,
    "args": {
      "External id": 145550,
      "cbid": 211, "correlation": 145550
    }
  },
  {
    "ph": "s", "id": 145550, "pid": 76337, "tid": -914061504, "ts": 1716454223635103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223690353, "dur": 4,
    "args": {
      "External id": 145562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145562,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 145562, "pid": 5, "tid": 7, "ts": 1716454223690353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635124, "dur": 6,
    "args": {
      "External id": 145562,
      "cbid": 211, "correlation": 145562
    }
  },
  {
    "ph": "s", "id": 145562, "pid": 76337, "tid": -914061504, "ts": 1716454223635124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223690358, "dur": 17,
    "args": {
      "External id": 145565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145565, "pid": 5, "tid": 7, "ts": 1716454223690358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635142, "dur": 6,
    "args": {
      "External id": 145565,
      "cbid": 211, "correlation": 145565
    }
  },
  {
    "ph": "s", "id": 145565, "pid": 76337, "tid": -914061504, "ts": 1716454223635142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223690377, "dur": 11,
    "args": {
      "External id": 145574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145574,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145574, "pid": 5, "tid": 7, "ts": 1716454223690377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635182, "dur": 11,
    "args": {
      "External id": 145574,
      "cbid": 211, "correlation": 145574
    }
  },
  {
    "ph": "s", "id": 145574, "pid": 76337, "tid": -914061504, "ts": 1716454223635182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635246, "dur": 0,
    "args": {
      "External id": 145584,
      "cbid": 317, "correlation": 145584
    }
  },
  {
    "ph": "f", "id": 145584, "pid": 76337, "tid": -914061504, "ts": 1716454223635246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635247, "dur": 0,
    "args": {
      "External id": 145585,
      "cbid": 203, "correlation": 145585
    }
  },
  {
    "ph": "f", "id": 145585, "pid": 76337, "tid": -914061504, "ts": 1716454223635247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635248, "dur": 0,
    "args": {
      "External id": 145586,
      "cbid": 205, "correlation": 145586
    }
  },
  {
    "ph": "f", "id": 145586, "pid": 76337, "tid": -914061504, "ts": 1716454223635248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223690389, "dur": 12,
    "args": {
      "External id": 145590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145590,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145590, "pid": 5, "tid": 7, "ts": 1716454223690389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635262, "dur": 12,
    "args": {
      "External id": 145590,
      "cbid": 211, "correlation": 145590
    }
  },
  {
    "ph": "s", "id": 145590, "pid": 76337, "tid": -914061504, "ts": 1716454223635262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223690402, "dur": 163,
    "args": {
      "External id": 145592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145592, "pid": 5, "tid": 7, "ts": 1716454223690402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635276, "dur": 5,
    "args": {
      "External id": 145592,
      "cbid": 211, "correlation": 145592
    }
  },
  {
    "ph": "s", "id": 145592, "pid": 76337, "tid": -914061504, "ts": 1716454223635276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223690567, "dur": 1,
    "args": {
      "External id": 145594,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 145594,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 145594, "pid": 5, "tid": 7, "ts": 1716454223690567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635287, "dur": 7,
    "args": {
      "External id": 145594,
      "cbid": 51, "correlation": 145594
    }
  },
  {
    "ph": "s", "id": 145594, "pid": 76337, "tid": -914061504, "ts": 1716454223635287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223690571, "dur": 650,
    "args": {
      "External id": 145595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145595,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145595, "pid": 5, "tid": 7, "ts": 1716454223690571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635295, "dur": 6,
    "args": {
      "External id": 145595,
      "cbid": 211, "correlation": 145595
    }
  },
  {
    "ph": "s", "id": 145595, "pid": 76337, "tid": -914061504, "ts": 1716454223635295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223691222, "dur": 13,
    "args": {
      "External id": 145597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145597,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145597, "pid": 5, "tid": 7, "ts": 1716454223691222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635305, "dur": 5,
    "args": {
      "External id": 145597,
      "cbid": 211, "correlation": 145597
    }
  },
  {
    "ph": "s", "id": 145597, "pid": 76337, "tid": -914061504, "ts": 1716454223635305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223691236, "dur": 15,
    "args": {
      "External id": 145603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145603, "pid": 5, "tid": 7, "ts": 1716454223691236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635332, "dur": 9,
    "args": {
      "External id": 145603,
      "cbid": 211, "correlation": 145603
    }
  },
  {
    "ph": "s", "id": 145603, "pid": 76337, "tid": -914061504, "ts": 1716454223635332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635391, "dur": 0,
    "args": {
      "External id": 145613,
      "cbid": 317, "correlation": 145613
    }
  },
  {
    "ph": "f", "id": 145613, "pid": 76337, "tid": -914061504, "ts": 1716454223635391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635392, "dur": 0,
    "args": {
      "External id": 145614,
      "cbid": 203, "correlation": 145614
    }
  },
  {
    "ph": "f", "id": 145614, "pid": 76337, "tid": -914061504, "ts": 1716454223635392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635393, "dur": 0,
    "args": {
      "External id": 145615,
      "cbid": 205, "correlation": 145615
    }
  },
  {
    "ph": "f", "id": 145615, "pid": 76337, "tid": -914061504, "ts": 1716454223635393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223691253, "dur": 22,
    "args": {
      "External id": 145619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145619,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145619, "pid": 5, "tid": 7, "ts": 1716454223691253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635408, "dur": 11,
    "args": {
      "External id": 145619,
      "cbid": 211, "correlation": 145619
    }
  },
  {
    "ph": "s", "id": 145619, "pid": 76337, "tid": -914061504, "ts": 1716454223635408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223691276, "dur": 4,
    "args": {
      "External id": 145621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 145621, "pid": 5, "tid": 7, "ts": 1716454223691276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635427, "dur": 9,
    "args": {
      "External id": 145621,
      "cbid": 211, "correlation": 145621
    }
  },
  {
    "ph": "s", "id": 145621, "pid": 76337, "tid": -914061504, "ts": 1716454223635427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635442, "dur": 0,
    "args": {
      "External id": 145622,
      "cbid": 51, "correlation": 145622
    }
  },
  {
    "ph": "s", "id": 145622, "pid": 76337, "tid": -914061504, "ts": 1716454223635442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223691281, "dur": 176,
    "args": {
      "External id": 145623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145623,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 145623, "pid": 5, "tid": 7, "ts": 1716454223691281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635443, "dur": 8,
    "args": {
      "External id": 145623,
      "cbid": 211, "correlation": 145623
    }
  },
  {
    "ph": "s", "id": 145623, "pid": 76337, "tid": -914061504, "ts": 1716454223635443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223691458, "dur": 16,
    "args": {
      "External id": 145628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145628, "pid": 5, "tid": 7, "ts": 1716454223691458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635472, "dur": 8,
    "args": {
      "External id": 145628,
      "cbid": 211, "correlation": 145628
    }
  },
  {
    "ph": "s", "id": 145628, "pid": 76337, "tid": -914061504, "ts": 1716454223635472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223691476, "dur": 12,
    "args": {
      "External id": 145636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145636,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145636, "pid": 5, "tid": 7, "ts": 1716454223691476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635500, "dur": 7,
    "args": {
      "External id": 145636,
      "cbid": 211, "correlation": 145636
    }
  },
  {
    "ph": "s", "id": 145636, "pid": 76337, "tid": -914061504, "ts": 1716454223635500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223691490, "dur": 10,
    "args": {
      "External id": 145644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145644,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145644, "pid": 5, "tid": 7, "ts": 1716454223691490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635529, "dur": 9,
    "args": {
      "External id": 145644,
      "cbid": 211, "correlation": 145644
    }
  },
  {
    "ph": "s", "id": 145644, "pid": 76337, "tid": -914061504, "ts": 1716454223635529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223691501, "dur": 19,
    "args": {
      "External id": 145664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145664,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 145664, "pid": 5, "tid": 7, "ts": 1716454223691501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635615, "dur": 12,
    "args": {
      "External id": 145664,
      "cbid": 211, "correlation": 145664
    }
  },
  {
    "ph": "s", "id": 145664, "pid": 76337, "tid": -914061504, "ts": 1716454223635615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223691522, "dur": 5,
    "args": {
      "External id": 145676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145676,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 145676, "pid": 5, "tid": 7, "ts": 1716454223691522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635636, "dur": 7,
    "args": {
      "External id": 145676,
      "cbid": 211, "correlation": 145676
    }
  },
  {
    "ph": "s", "id": 145676, "pid": 76337, "tid": -914061504, "ts": 1716454223635636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223691527, "dur": 17,
    "args": {
      "External id": 145679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145679,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145679, "pid": 5, "tid": 7, "ts": 1716454223691527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635654, "dur": 6,
    "args": {
      "External id": 145679,
      "cbid": 211, "correlation": 145679
    }
  },
  {
    "ph": "s", "id": 145679, "pid": 76337, "tid": -914061504, "ts": 1716454223635654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635712, "dur": 0,
    "args": {
      "External id": 145690,
      "cbid": 317, "correlation": 145690
    }
  },
  {
    "ph": "f", "id": 145690, "pid": 76337, "tid": -914061504, "ts": 1716454223635712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635713, "dur": 0,
    "args": {
      "External id": 145691,
      "cbid": 203, "correlation": 145691
    }
  },
  {
    "ph": "f", "id": 145691, "pid": 76337, "tid": -914061504, "ts": 1716454223635713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635714, "dur": 0,
    "args": {
      "External id": 145692,
      "cbid": 205, "correlation": 145692
    }
  },
  {
    "ph": "f", "id": 145692, "pid": 76337, "tid": -914061504, "ts": 1716454223635714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223691546, "dur": 12,
    "args": {
      "External id": 145696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145696,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145696, "pid": 5, "tid": 7, "ts": 1716454223691546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635730, "dur": 12,
    "args": {
      "External id": 145696,
      "cbid": 211, "correlation": 145696
    }
  },
  {
    "ph": "s", "id": 145696, "pid": 76337, "tid": -914061504, "ts": 1716454223635730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223691559, "dur": 3,
    "args": {
      "External id": 145698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 145698, "pid": 5, "tid": 7, "ts": 1716454223691559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635747, "dur": 7,
    "args": {
      "External id": 145698,
      "cbid": 211, "correlation": 145698
    }
  },
  {
    "ph": "s", "id": 145698, "pid": 76337, "tid": -914061504, "ts": 1716454223635747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635756, "dur": 0,
    "args": {
      "External id": 145699,
      "cbid": 51, "correlation": 145699
    }
  },
  {
    "ph": "s", "id": 145699, "pid": 76337, "tid": -914061504, "ts": 1716454223635756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223691564, "dur": 91,
    "args": {
      "External id": 145700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145700,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 145700, "pid": 5, "tid": 7, "ts": 1716454223691564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635757, "dur": 5,
    "args": {
      "External id": 145700,
      "cbid": 211, "correlation": 145700
    }
  },
  {
    "ph": "s", "id": 145700, "pid": 76337, "tid": -914061504, "ts": 1716454223635757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223691656, "dur": 16,
    "args": {
      "External id": 145705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145705, "pid": 5, "tid": 7, "ts": 1716454223691656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635784, "dur": 9,
    "args": {
      "External id": 145705,
      "cbid": 211, "correlation": 145705
    }
  },
  {
    "ph": "s", "id": 145705, "pid": 76337, "tid": -914061504, "ts": 1716454223635784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223691673, "dur": 84,
    "args": {
      "External id": 145714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145714,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145714, "pid": 5, "tid": 7, "ts": 1716454223691673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635883, "dur": 16,
    "args": {
      "External id": 145714,
      "cbid": 211, "correlation": 145714
    }
  },
  {
    "ph": "s", "id": 145714, "pid": 76337, "tid": -914061504, "ts": 1716454223635883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223691759, "dur": 30,
    "args": {
      "External id": 145736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145736,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145736, "pid": 5, "tid": 7, "ts": 1716454223691759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223635956, "dur": 12,
    "args": {
      "External id": 145736,
      "cbid": 211, "correlation": 145736
    }
  },
  {
    "ph": "s", "id": 145736, "pid": 76337, "tid": -914061504, "ts": 1716454223635956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636091, "dur": 2,
    "args": {
      "External id": 145747,
      "cbid": 251, "correlation": 145747
    }
  },
  {
    "ph": "f", "id": 145747, "pid": 76337, "tid": -914061504, "ts": 1716454223636091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223691790, "dur": 164,
    "args": {
      "External id": 145748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145748,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145748, "pid": 5, "tid": 7, "ts": 1716454223691790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636100, "dur": 15,
    "args": {
      "External id": 145748,
      "cbid": 211, "correlation": 145748
    }
  },
  {
    "ph": "s", "id": 145748, "pid": 76337, "tid": -914061504, "ts": 1716454223636100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636179, "dur": 1,
    "args": {
      "External id": 145759,
      "cbid": 251, "correlation": 145759
    }
  },
  {
    "ph": "f", "id": 145759, "pid": 76337, "tid": -914061504, "ts": 1716454223636179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223691956, "dur": 160,
    "args": {
      "External id": 145760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145760,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145760, "pid": 5, "tid": 7, "ts": 1716454223691956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636183, "dur": 11,
    "args": {
      "External id": 145760,
      "cbid": 211, "correlation": 145760
    }
  },
  {
    "ph": "s", "id": 145760, "pid": 76337, "tid": -914061504, "ts": 1716454223636183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636249, "dur": 1,
    "args": {
      "External id": 145771,
      "cbid": 251, "correlation": 145771
    }
  },
  {
    "ph": "f", "id": 145771, "pid": 76337, "tid": -914061504, "ts": 1716454223636249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223692117, "dur": 161,
    "args": {
      "External id": 145772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145772,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145772, "pid": 5, "tid": 7, "ts": 1716454223692117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636253, "dur": 11,
    "args": {
      "External id": 145772,
      "cbid": 211, "correlation": 145772
    }
  },
  {
    "ph": "s", "id": 145772, "pid": 76337, "tid": -914061504, "ts": 1716454223636253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223692280, "dur": 339,
    "args": {
      "External id": 145797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145797,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145797, "pid": 5, "tid": 7, "ts": 1716454223692280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636365, "dur": 16,
    "args": {
      "External id": 145797,
      "cbid": 211, "correlation": 145797
    }
  },
  {
    "ph": "s", "id": 145797, "pid": 76337, "tid": -914061504, "ts": 1716454223636365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636486, "dur": 2,
    "args": {
      "External id": 145815,
      "cbid": 251, "correlation": 145815
    }
  },
  {
    "ph": "f", "id": 145815, "pid": 76337, "tid": -914061504, "ts": 1716454223636486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223692620, "dur": 167,
    "args": {
      "External id": 145817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145817,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145817, "pid": 5, "tid": 7, "ts": 1716454223692620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636494, "dur": 13,
    "args": {
      "External id": 145817,
      "cbid": 211, "correlation": 145817
    }
  },
  {
    "ph": "s", "id": 145817, "pid": 76337, "tid": -914061504, "ts": 1716454223636494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223692788, "dur": 19,
    "args": {
      "External id": 145825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145825,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145825, "pid": 5, "tid": 7, "ts": 1716454223692788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636564, "dur": 13,
    "args": {
      "External id": 145825,
      "cbid": 211, "correlation": 145825
    }
  },
  {
    "ph": "s", "id": 145825, "pid": 76337, "tid": -914061504, "ts": 1716454223636564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223692809, "dur": 28,
    "args": {
      "External id": 145833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145833,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145833, "pid": 5, "tid": 7, "ts": 1716454223692809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636603, "dur": 8,
    "args": {
      "External id": 145833,
      "cbid": 211, "correlation": 145833
    }
  },
  {
    "ph": "s", "id": 145833, "pid": 76337, "tid": -914061504, "ts": 1716454223636603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223692838, "dur": 18,
    "args": {
      "External id": 145844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145844, "pid": 5, "tid": 7, "ts": 1716454223692838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636698, "dur": 15,
    "args": {
      "External id": 145844,
      "cbid": 211, "correlation": 145844
    }
  },
  {
    "ph": "s", "id": 145844, "pid": 76337, "tid": -914061504, "ts": 1716454223636698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223692857, "dur": 16,
    "args": {
      "External id": 145866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145866,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145866, "pid": 5, "tid": 7, "ts": 1716454223692857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636732, "dur": 7,
    "args": {
      "External id": 145866,
      "cbid": 211, "correlation": 145866
    }
  },
  {
    "ph": "s", "id": 145866, "pid": 76337, "tid": -914061504, "ts": 1716454223636732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636819, "dur": 2,
    "args": {
      "External id": 145877,
      "cbid": 251, "correlation": 145877
    }
  },
  {
    "ph": "f", "id": 145877, "pid": 76337, "tid": -914061504, "ts": 1716454223636819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223692875, "dur": 90,
    "args": {
      "External id": 145878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145878,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 145878, "pid": 5, "tid": 7, "ts": 1716454223692875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636826, "dur": 13,
    "args": {
      "External id": 145878,
      "cbid": 211, "correlation": 145878
    }
  },
  {
    "ph": "s", "id": 145878, "pid": 76337, "tid": -914061504, "ts": 1716454223636826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636904, "dur": 1,
    "args": {
      "External id": 145889,
      "cbid": 251, "correlation": 145889
    }
  },
  {
    "ph": "f", "id": 145889, "pid": 76337, "tid": -914061504, "ts": 1716454223636904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636907, "dur": 0,
    "args": {
      "External id": 145890,
      "cbid": 251, "correlation": 145890
    }
  },
  {
    "ph": "f", "id": 145890, "pid": 76337, "tid": -914061504, "ts": 1716454223636907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223692966, "dur": 12,
    "args": {
      "External id": 145891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145891,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145891, "pid": 5, "tid": 7, "ts": 1716454223692966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636909, "dur": 12,
    "args": {
      "External id": 145891,
      "cbid": 211, "correlation": 145891
    }
  },
  {
    "ph": "s", "id": 145891, "pid": 76337, "tid": -914061504, "ts": 1716454223636909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223692980, "dur": 5,
    "args": {
      "External id": 145893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145893,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145893, "pid": 5, "tid": 7, "ts": 1716454223692980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636924, "dur": 6,
    "args": {
      "External id": 145893,
      "cbid": 211, "correlation": 145893
    }
  },
  {
    "ph": "s", "id": 145893, "pid": 76337, "tid": -914061504, "ts": 1716454223636924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636992, "dur": 1,
    "args": {
      "External id": 145904,
      "cbid": 251, "correlation": 145904
    }
  },
  {
    "ph": "f", "id": 145904, "pid": 76337, "tid": -914061504, "ts": 1716454223636992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636996, "dur": 0,
    "args": {
      "External id": 145905,
      "cbid": 251, "correlation": 145905
    }
  },
  {
    "ph": "f", "id": 145905, "pid": 76337, "tid": -914061504, "ts": 1716454223636996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223692986, "dur": 9,
    "args": {
      "External id": 145906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145906,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145906, "pid": 5, "tid": 7, "ts": 1716454223692986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223636997, "dur": 13,
    "args": {
      "External id": 145906,
      "cbid": 211, "correlation": 145906
    }
  },
  {
    "ph": "s", "id": 145906, "pid": 76337, "tid": -914061504, "ts": 1716454223636997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223692996, "dur": 4,
    "args": {
      "External id": 145908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145908,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145908, "pid": 5, "tid": 7, "ts": 1716454223692996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637012, "dur": 6,
    "args": {
      "External id": 145908,
      "cbid": 211, "correlation": 145908
    }
  },
  {
    "ph": "s", "id": 145908, "pid": 76337, "tid": -914061504, "ts": 1716454223637012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223693001, "dur": 54,
    "args": {
      "External id": 145933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145933,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 145933, "pid": 5, "tid": 7, "ts": 1716454223693001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637090, "dur": 12,
    "args": {
      "External id": 145933,
      "cbid": 211, "correlation": 145933
    }
  },
  {
    "ph": "s", "id": 145933, "pid": 76337, "tid": -914061504, "ts": 1716454223637090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637189, "dur": 2,
    "args": {
      "External id": 145951,
      "cbid": 251, "correlation": 145951
    }
  },
  {
    "ph": "f", "id": 145951, "pid": 76337, "tid": -914061504, "ts": 1716454223637189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223693057, "dur": 91,
    "args": {
      "External id": 145953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145953,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 145953, "pid": 5, "tid": 7, "ts": 1716454223693057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637196, "dur": 14,
    "args": {
      "External id": 145953,
      "cbid": 211, "correlation": 145953
    }
  },
  {
    "ph": "s", "id": 145953, "pid": 76337, "tid": -914061504, "ts": 1716454223637196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223693149, "dur": 10,
    "args": {
      "External id": 145961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145961,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145961, "pid": 5, "tid": 7, "ts": 1716454223693149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637267, "dur": 12,
    "args": {
      "External id": 145961,
      "cbid": 211, "correlation": 145961
    }
  },
  {
    "ph": "s", "id": 145961, "pid": 76337, "tid": -914061504, "ts": 1716454223637267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223693160, "dur": 20,
    "args": {
      "External id": 145969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145969,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145969, "pid": 5, "tid": 7, "ts": 1716454223693160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637309, "dur": 9,
    "args": {
      "External id": 145969,
      "cbid": 211, "correlation": 145969
    }
  },
  {
    "ph": "s", "id": 145969, "pid": 76337, "tid": -914061504, "ts": 1716454223637309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223693182, "dur": 17,
    "args": {
      "External id": 145991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 145991,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 145991, "pid": 5, "tid": 7, "ts": 1716454223693182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637361, "dur": 10,
    "args": {
      "External id": 145991,
      "cbid": 211, "correlation": 145991
    }
  },
  {
    "ph": "s", "id": 145991, "pid": 76337, "tid": -914061504, "ts": 1716454223637361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637458, "dur": 2,
    "args": {
      "External id": 146007,
      "cbid": 251, "correlation": 146007
    }
  },
  {
    "ph": "f", "id": 146007, "pid": 76337, "tid": -914061504, "ts": 1716454223637458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637465, "dur": 0,
    "args": {
      "External id": 146009,
      "cbid": 251, "correlation": 146009
    }
  },
  {
    "ph": "f", "id": 146009, "pid": 76337, "tid": -914061504, "ts": 1716454223637465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223693201, "dur": 494,
    "args": {
      "External id": 146010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146010,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146010, "pid": 5, "tid": 7, "ts": 1716454223693201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637468, "dur": 15,
    "args": {
      "External id": 146010,
      "cbid": 211, "correlation": 146010
    }
  },
  {
    "ph": "s", "id": 146010, "pid": 76337, "tid": -914061504, "ts": 1716454223637468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223693697, "dur": 66,
    "args": {
      "External id": 146018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146018, "pid": 5, "tid": 7, "ts": 1716454223693697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637556, "dur": 14,
    "args": {
      "External id": 146018,
      "cbid": 211, "correlation": 146018
    }
  },
  {
    "ph": "s", "id": 146018, "pid": 76337, "tid": -914061504, "ts": 1716454223637556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223693764, "dur": 67,
    "args": {
      "External id": 146026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146026, "pid": 5, "tid": 7, "ts": 1716454223693764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637597, "dur": 11,
    "args": {
      "External id": 146026,
      "cbid": 211, "correlation": 146026
    }
  },
  {
    "ph": "s", "id": 146026, "pid": 76337, "tid": -914061504, "ts": 1716454223637597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637681, "dur": 1,
    "args": {
      "External id": 146042,
      "cbid": 251, "correlation": 146042
    }
  },
  {
    "ph": "f", "id": 146042, "pid": 76337, "tid": -914061504, "ts": 1716454223637681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223693833, "dur": 1,
    "args": {
      "External id": 146044,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 146044,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 146044, "pid": 5, "tid": 7, "ts": 1716454223693833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637686, "dur": 11,
    "args": {
      "External id": 146044,
      "cbid": 51, "correlation": 146044
    }
  },
  {
    "ph": "s", "id": 146044, "pid": 76337, "tid": -914061504, "ts": 1716454223637686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223693836, "dur": 271,
    "args": {
      "External id": 146045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146045,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146045, "pid": 5, "tid": 7, "ts": 1716454223693836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637699, "dur": 11,
    "args": {
      "External id": 146045,
      "cbid": 211, "correlation": 146045
    }
  },
  {
    "ph": "s", "id": 146045, "pid": 76337, "tid": -914061504, "ts": 1716454223637699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223694109, "dur": 14,
    "args": {
      "External id": 146053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146053,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146053, "pid": 5, "tid": 7, "ts": 1716454223694109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637740, "dur": 10,
    "args": {
      "External id": 146053,
      "cbid": 211, "correlation": 146053
    }
  },
  {
    "ph": "s", "id": 146053, "pid": 76337, "tid": -914061504, "ts": 1716454223637740,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223694125, "dur": 38,
    "args": {
      "External id": 146064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146064, "pid": 5, "tid": 7, "ts": 1716454223694125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637813, "dur": 12,
    "args": {
      "External id": 146064,
      "cbid": 211, "correlation": 146064
    }
  },
  {
    "ph": "s", "id": 146064, "pid": 76337, "tid": -914061504, "ts": 1716454223637813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637878, "dur": 0,
    "args": {
      "External id": 146076,
      "cbid": 317, "correlation": 146076
    }
  },
  {
    "ph": "f", "id": 146076, "pid": 76337, "tid": -914061504, "ts": 1716454223637878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637879, "dur": 0,
    "args": {
      "External id": 146077,
      "cbid": 203, "correlation": 146077
    }
  },
  {
    "ph": "f", "id": 146077, "pid": 76337, "tid": -914061504, "ts": 1716454223637879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637880, "dur": 0,
    "args": {
      "External id": 146078,
      "cbid": 205, "correlation": 146078
    }
  },
  {
    "ph": "f", "id": 146078, "pid": 76337, "tid": -914061504, "ts": 1716454223637880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223694164, "dur": 13,
    "args": {
      "External id": 146082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146082,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146082, "pid": 5, "tid": 7, "ts": 1716454223694164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637895, "dur": 12,
    "args": {
      "External id": 146082,
      "cbid": 211, "correlation": 146082
    }
  },
  {
    "ph": "s", "id": 146082, "pid": 76337, "tid": -914061504, "ts": 1716454223637895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223694179, "dur": 4,
    "args": {
      "External id": 146084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146084,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146084, "pid": 5, "tid": 7, "ts": 1716454223694179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637912, "dur": 6,
    "args": {
      "External id": 146084,
      "cbid": 211, "correlation": 146084
    }
  },
  {
    "ph": "s", "id": 146084, "pid": 76337, "tid": -914061504, "ts": 1716454223637912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637921, "dur": 0,
    "args": {
      "External id": 146085,
      "cbid": 51, "correlation": 146085
    }
  },
  {
    "ph": "s", "id": 146085, "pid": 76337, "tid": -914061504, "ts": 1716454223637921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223694184, "dur": 97,
    "args": {
      "External id": 146086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146086,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 146086, "pid": 5, "tid": 7, "ts": 1716454223694184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637921, "dur": 7,
    "args": {
      "External id": 146086,
      "cbid": 211, "correlation": 146086
    }
  },
  {
    "ph": "s", "id": 146086, "pid": 76337, "tid": -914061504, "ts": 1716454223637921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223694282, "dur": 17,
    "args": {
      "External id": 146091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146091,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146091, "pid": 5, "tid": 7, "ts": 1716454223694282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637950, "dur": 9,
    "args": {
      "External id": 146091,
      "cbid": 211, "correlation": 146091
    }
  },
  {
    "ph": "s", "id": 146091, "pid": 76337, "tid": -914061504, "ts": 1716454223637950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223694300, "dur": 12,
    "args": {
      "External id": 146099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146099,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146099, "pid": 5, "tid": 7, "ts": 1716454223694300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223637990, "dur": 9,
    "args": {
      "External id": 146099,
      "cbid": 211, "correlation": 146099
    }
  },
  {
    "ph": "s", "id": 146099, "pid": 76337, "tid": -914061504, "ts": 1716454223637990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223694313, "dur": 31,
    "args": {
      "External id": 146108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146108,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146108, "pid": 5, "tid": 7, "ts": 1716454223694313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638031, "dur": 10,
    "args": {
      "External id": 146108,
      "cbid": 211, "correlation": 146108
    }
  },
  {
    "ph": "s", "id": 146108, "pid": 76337, "tid": -914061504, "ts": 1716454223638031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223694345, "dur": 31,
    "args": {
      "External id": 146128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146128,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 146128, "pid": 5, "tid": 7, "ts": 1716454223694345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638105, "dur": 12,
    "args": {
      "External id": 146128,
      "cbid": 211, "correlation": 146128
    }
  },
  {
    "ph": "s", "id": 146128, "pid": 76337, "tid": -914061504, "ts": 1716454223638105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223694378, "dur": 5,
    "args": {
      "External id": 146140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146140,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146140, "pid": 5, "tid": 7, "ts": 1716454223694378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638127, "dur": 6,
    "args": {
      "External id": 146140,
      "cbid": 211, "correlation": 146140
    }
  },
  {
    "ph": "s", "id": 146140, "pid": 76337, "tid": -914061504, "ts": 1716454223638127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223694383, "dur": 31,
    "args": {
      "External id": 146143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146143,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146143, "pid": 5, "tid": 7, "ts": 1716454223694383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638145, "dur": 6,
    "args": {
      "External id": 146143,
      "cbid": 211, "correlation": 146143
    }
  },
  {
    "ph": "s", "id": 146143, "pid": 76337, "tid": -914061504, "ts": 1716454223638145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223694415, "dur": 20,
    "args": {
      "External id": 146152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146152,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146152, "pid": 5, "tid": 7, "ts": 1716454223694415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638184, "dur": 9,
    "args": {
      "External id": 146152,
      "cbid": 211, "correlation": 146152
    }
  },
  {
    "ph": "s", "id": 146152, "pid": 76337, "tid": -914061504, "ts": 1716454223638184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638235, "dur": 0,
    "args": {
      "External id": 146162,
      "cbid": 317, "correlation": 146162
    }
  },
  {
    "ph": "f", "id": 146162, "pid": 76337, "tid": -914061504, "ts": 1716454223638235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638236, "dur": 0,
    "args": {
      "External id": 146163,
      "cbid": 203, "correlation": 146163
    }
  },
  {
    "ph": "f", "id": 146163, "pid": 76337, "tid": -914061504, "ts": 1716454223638236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638237, "dur": 0,
    "args": {
      "External id": 146164,
      "cbid": 205, "correlation": 146164
    }
  },
  {
    "ph": "f", "id": 146164, "pid": 76337, "tid": -914061504, "ts": 1716454223638237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223694437, "dur": 23,
    "args": {
      "External id": 146168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146168, "pid": 5, "tid": 7, "ts": 1716454223694437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638251, "dur": 13,
    "args": {
      "External id": 146168,
      "cbid": 211, "correlation": 146168
    }
  },
  {
    "ph": "s", "id": 146168, "pid": 76337, "tid": -914061504, "ts": 1716454223638251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223694461, "dur": 322,
    "args": {
      "External id": 146170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146170,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146170, "pid": 5, "tid": 7, "ts": 1716454223694461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638266, "dur": 5,
    "args": {
      "External id": 146170,
      "cbid": 211, "correlation": 146170
    }
  },
  {
    "ph": "s", "id": 146170, "pid": 76337, "tid": -914061504, "ts": 1716454223638266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223694785, "dur": 1,
    "args": {
      "External id": 146172,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 146172,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 146172, "pid": 5, "tid": 7, "ts": 1716454223694785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638277, "dur": 6,
    "args": {
      "External id": 146172,
      "cbid": 51, "correlation": 146172
    }
  },
  {
    "ph": "s", "id": 146172, "pid": 76337, "tid": -914061504, "ts": 1716454223638277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223694789, "dur": 1260,
    "args": {
      "External id": 146173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146173,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146173, "pid": 5, "tid": 7, "ts": 1716454223694789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638284, "dur": 6,
    "args": {
      "External id": 146173,
      "cbid": 211, "correlation": 146173
    }
  },
  {
    "ph": "s", "id": 146173, "pid": 76337, "tid": -914061504, "ts": 1716454223638284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223696051, "dur": 13,
    "args": {
      "External id": 146175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146175,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146175, "pid": 5, "tid": 7, "ts": 1716454223696051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638294, "dur": 5,
    "args": {
      "External id": 146175,
      "cbid": 211, "correlation": 146175
    }
  },
  {
    "ph": "s", "id": 146175, "pid": 76337, "tid": -914061504, "ts": 1716454223638294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223696065, "dur": 15,
    "args": {
      "External id": 146181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146181,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146181, "pid": 5, "tid": 7, "ts": 1716454223696065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638323, "dur": 8,
    "args": {
      "External id": 146181,
      "cbid": 211, "correlation": 146181
    }
  },
  {
    "ph": "s", "id": 146181, "pid": 76337, "tid": -914061504, "ts": 1716454223638323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223696081, "dur": 4,
    "args": {
      "External id": 146189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146189,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 146189, "pid": 5, "tid": 7, "ts": 1716454223696081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638367, "dur": 9,
    "args": {
      "External id": 146189,
      "cbid": 211, "correlation": 146189
    }
  },
  {
    "ph": "s", "id": 146189, "pid": 76337, "tid": -914061504, "ts": 1716454223638367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638431, "dur": 1,
    "args": {
      "External id": 146205,
      "cbid": 251, "correlation": 146205
    }
  },
  {
    "ph": "f", "id": 146205, "pid": 76337, "tid": -914061504, "ts": 1716454223638431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638436, "dur": 0,
    "args": {
      "External id": 146207,
      "cbid": 251, "correlation": 146207
    }
  },
  {
    "ph": "f", "id": 146207, "pid": 76337, "tid": -914061504, "ts": 1716454223638436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223696087, "dur": 13,
    "args": {
      "External id": 146208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146208,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146208, "pid": 5, "tid": 7, "ts": 1716454223696087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638438, "dur": 11,
    "args": {
      "External id": 146208,
      "cbid": 211, "correlation": 146208
    }
  },
  {
    "ph": "s", "id": 146208, "pid": 76337, "tid": -914061504, "ts": 1716454223638438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223696101, "dur": 5,
    "args": {
      "External id": 146210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146210,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146210, "pid": 5, "tid": 7, "ts": 1716454223696101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638451, "dur": 5,
    "args": {
      "External id": 146210,
      "cbid": 211, "correlation": 146210
    }
  },
  {
    "ph": "s", "id": 146210, "pid": 76337, "tid": -914061504, "ts": 1716454223638451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223696108, "dur": 17,
    "args": {
      "External id": 146220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146220,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146220, "pid": 5, "tid": 7, "ts": 1716454223696108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638510, "dur": 13,
    "args": {
      "External id": 146220,
      "cbid": 211, "correlation": 146220
    }
  },
  {
    "ph": "s", "id": 146220, "pid": 76337, "tid": -914061504, "ts": 1716454223638510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223696126, "dur": 18,
    "args": {
      "External id": 146240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146240,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 146240, "pid": 5, "tid": 7, "ts": 1716454223696126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638576, "dur": 11,
    "args": {
      "External id": 146240,
      "cbid": 211, "correlation": 146240
    }
  },
  {
    "ph": "s", "id": 146240, "pid": 76337, "tid": -914061504, "ts": 1716454223638576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223696146, "dur": 4,
    "args": {
      "External id": 146252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146252,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 146252, "pid": 5, "tid": 7, "ts": 1716454223696146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638596, "dur": 6,
    "args": {
      "External id": 146252,
      "cbid": 211, "correlation": 146252
    }
  },
  {
    "ph": "s", "id": 146252, "pid": 76337, "tid": -914061504, "ts": 1716454223638596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223696152, "dur": 16,
    "args": {
      "External id": 146255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146255,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146255, "pid": 5, "tid": 7, "ts": 1716454223696152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638615, "dur": 6,
    "args": {
      "External id": 146255,
      "cbid": 211, "correlation": 146255
    }
  },
  {
    "ph": "s", "id": 146255, "pid": 76337, "tid": -914061504, "ts": 1716454223638615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223696169, "dur": 11,
    "args": {
      "External id": 146264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146264,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146264, "pid": 5, "tid": 7, "ts": 1716454223696169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638656, "dur": 11,
    "args": {
      "External id": 146264,
      "cbid": 211, "correlation": 146264
    }
  },
  {
    "ph": "s", "id": 146264, "pid": 76337, "tid": -914061504, "ts": 1716454223638656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638720, "dur": 0,
    "args": {
      "External id": 146274,
      "cbid": 317, "correlation": 146274
    }
  },
  {
    "ph": "f", "id": 146274, "pid": 76337, "tid": -914061504, "ts": 1716454223638720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638721, "dur": 0,
    "args": {
      "External id": 146275,
      "cbid": 203, "correlation": 146275
    }
  },
  {
    "ph": "f", "id": 146275, "pid": 76337, "tid": -914061504, "ts": 1716454223638721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638721, "dur": 0,
    "args": {
      "External id": 146276,
      "cbid": 205, "correlation": 146276
    }
  },
  {
    "ph": "f", "id": 146276, "pid": 76337, "tid": -914061504, "ts": 1716454223638721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223696182, "dur": 12,
    "args": {
      "External id": 146280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146280,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146280, "pid": 5, "tid": 7, "ts": 1716454223696182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638737, "dur": 12,
    "args": {
      "External id": 146280,
      "cbid": 211, "correlation": 146280
    }
  },
  {
    "ph": "s", "id": 146280, "pid": 76337, "tid": -914061504, "ts": 1716454223638737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223696195, "dur": 164,
    "args": {
      "External id": 146282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146282,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146282, "pid": 5, "tid": 7, "ts": 1716454223696195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638752, "dur": 5,
    "args": {
      "External id": 146282,
      "cbid": 211, "correlation": 146282
    }
  },
  {
    "ph": "s", "id": 146282, "pid": 76337, "tid": -914061504, "ts": 1716454223638752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223696361, "dur": 1,
    "args": {
      "External id": 146284,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 146284,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 146284, "pid": 5, "tid": 7, "ts": 1716454223696361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638763, "dur": 6,
    "args": {
      "External id": 146284,
      "cbid": 51, "correlation": 146284
    }
  },
  {
    "ph": "s", "id": 146284, "pid": 76337, "tid": -914061504, "ts": 1716454223638763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223696364, "dur": 651,
    "args": {
      "External id": 146285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146285,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146285, "pid": 5, "tid": 7, "ts": 1716454223696364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638770, "dur": 6,
    "args": {
      "External id": 146285,
      "cbid": 211, "correlation": 146285
    }
  },
  {
    "ph": "s", "id": 146285, "pid": 76337, "tid": -914061504, "ts": 1716454223638770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223697016, "dur": 13,
    "args": {
      "External id": 146287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146287,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146287, "pid": 5, "tid": 7, "ts": 1716454223697016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638781, "dur": 5,
    "args": {
      "External id": 146287,
      "cbid": 211, "correlation": 146287
    }
  },
  {
    "ph": "s", "id": 146287, "pid": 76337, "tid": -914061504, "ts": 1716454223638781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223697030, "dur": 15,
    "args": {
      "External id": 146293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146293,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146293, "pid": 5, "tid": 7, "ts": 1716454223697030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638809, "dur": 9,
    "args": {
      "External id": 146293,
      "cbid": 211, "correlation": 146293
    }
  },
  {
    "ph": "s", "id": 146293, "pid": 76337, "tid": -914061504, "ts": 1716454223638809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638868, "dur": 0,
    "args": {
      "External id": 146303,
      "cbid": 317, "correlation": 146303
    }
  },
  {
    "ph": "f", "id": 146303, "pid": 76337, "tid": -914061504, "ts": 1716454223638868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638869, "dur": 0,
    "args": {
      "External id": 146304,
      "cbid": 203, "correlation": 146304
    }
  },
  {
    "ph": "f", "id": 146304, "pid": 76337, "tid": -914061504, "ts": 1716454223638869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638870, "dur": 0,
    "args": {
      "External id": 146305,
      "cbid": 205, "correlation": 146305
    }
  },
  {
    "ph": "f", "id": 146305, "pid": 76337, "tid": -914061504, "ts": 1716454223638870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223697046, "dur": 20,
    "args": {
      "External id": 146309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146309,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146309, "pid": 5, "tid": 7, "ts": 1716454223697046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638882, "dur": 11,
    "args": {
      "External id": 146309,
      "cbid": 211, "correlation": 146309
    }
  },
  {
    "ph": "s", "id": 146309, "pid": 76337, "tid": -914061504, "ts": 1716454223638882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223697068, "dur": 4,
    "args": {
      "External id": 146311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146311, "pid": 5, "tid": 7, "ts": 1716454223697068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638897, "dur": 5,
    "args": {
      "External id": 146311,
      "cbid": 211, "correlation": 146311
    }
  },
  {
    "ph": "s", "id": 146311, "pid": 76337, "tid": -914061504, "ts": 1716454223638897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638905, "dur": 0,
    "args": {
      "External id": 146312,
      "cbid": 51, "correlation": 146312
    }
  },
  {
    "ph": "s", "id": 146312, "pid": 76337, "tid": -914061504, "ts": 1716454223638905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223697073, "dur": 172,
    "args": {
      "External id": 146313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146313,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 146313, "pid": 5, "tid": 7, "ts": 1716454223697073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638906, "dur": 5,
    "args": {
      "External id": 146313,
      "cbid": 211, "correlation": 146313
    }
  },
  {
    "ph": "s", "id": 146313, "pid": 76337, "tid": -914061504, "ts": 1716454223638906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223697247, "dur": 16,
    "args": {
      "External id": 146318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146318, "pid": 5, "tid": 7, "ts": 1716454223697247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638932, "dur": 8,
    "args": {
      "External id": 146318,
      "cbid": 211, "correlation": 146318
    }
  },
  {
    "ph": "s", "id": 146318, "pid": 76337, "tid": -914061504, "ts": 1716454223638932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223697264, "dur": 12,
    "args": {
      "External id": 146326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146326,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146326, "pid": 5, "tid": 7, "ts": 1716454223697264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638959, "dur": 9,
    "args": {
      "External id": 146326,
      "cbid": 211, "correlation": 146326
    }
  },
  {
    "ph": "s", "id": 146326, "pid": 76337, "tid": -914061504, "ts": 1716454223638959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223697277, "dur": 10,
    "args": {
      "External id": 146334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146334,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146334, "pid": 5, "tid": 7, "ts": 1716454223697277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223638998, "dur": 9,
    "args": {
      "External id": 146334,
      "cbid": 211, "correlation": 146334
    }
  },
  {
    "ph": "s", "id": 146334, "pid": 76337, "tid": -914061504, "ts": 1716454223638998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223697288, "dur": 20,
    "args": {
      "External id": 146354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146354,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 146354, "pid": 5, "tid": 7, "ts": 1716454223697288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639082, "dur": 12,
    "args": {
      "External id": 146354,
      "cbid": 211, "correlation": 146354
    }
  },
  {
    "ph": "s", "id": 146354, "pid": 76337, "tid": -914061504, "ts": 1716454223639082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223697309, "dur": 4,
    "args": {
      "External id": 146366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146366,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 146366, "pid": 5, "tid": 7, "ts": 1716454223697309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639104, "dur": 6,
    "args": {
      "External id": 146366,
      "cbid": 211, "correlation": 146366
    }
  },
  {
    "ph": "s", "id": 146366, "pid": 76337, "tid": -914061504, "ts": 1716454223639104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223697315, "dur": 17,
    "args": {
      "External id": 146369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146369,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146369, "pid": 5, "tid": 7, "ts": 1716454223697315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639122, "dur": 6,
    "args": {
      "External id": 146369,
      "cbid": 211, "correlation": 146369
    }
  },
  {
    "ph": "s", "id": 146369, "pid": 76337, "tid": -914061504, "ts": 1716454223639122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639179, "dur": 0,
    "args": {
      "External id": 146380,
      "cbid": 317, "correlation": 146380
    }
  },
  {
    "ph": "f", "id": 146380, "pid": 76337, "tid": -914061504, "ts": 1716454223639179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639180, "dur": 0,
    "args": {
      "External id": 146381,
      "cbid": 203, "correlation": 146381
    }
  },
  {
    "ph": "f", "id": 146381, "pid": 76337, "tid": -914061504, "ts": 1716454223639180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639180, "dur": 0,
    "args": {
      "External id": 146382,
      "cbid": 205, "correlation": 146382
    }
  },
  {
    "ph": "f", "id": 146382, "pid": 76337, "tid": -914061504, "ts": 1716454223639180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223697333, "dur": 12,
    "args": {
      "External id": 146386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146386,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146386, "pid": 5, "tid": 7, "ts": 1716454223697333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639194, "dur": 11,
    "args": {
      "External id": 146386,
      "cbid": 211, "correlation": 146386
    }
  },
  {
    "ph": "s", "id": 146386, "pid": 76337, "tid": -914061504, "ts": 1716454223639194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223697346, "dur": 3,
    "args": {
      "External id": 146388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146388,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146388, "pid": 5, "tid": 7, "ts": 1716454223697346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639209, "dur": 6,
    "args": {
      "External id": 146388,
      "cbid": 211, "correlation": 146388
    }
  },
  {
    "ph": "s", "id": 146388, "pid": 76337, "tid": -914061504, "ts": 1716454223639209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639219, "dur": 0,
    "args": {
      "External id": 146389,
      "cbid": 51, "correlation": 146389
    }
  },
  {
    "ph": "s", "id": 146389, "pid": 76337, "tid": -914061504, "ts": 1716454223639219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223697351, "dur": 90,
    "args": {
      "External id": 146390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146390,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 146390, "pid": 5, "tid": 7, "ts": 1716454223697351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639219, "dur": 5,
    "args": {
      "External id": 146390,
      "cbid": 211, "correlation": 146390
    }
  },
  {
    "ph": "s", "id": 146390, "pid": 76337, "tid": -914061504, "ts": 1716454223639219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223697443, "dur": 16,
    "args": {
      "External id": 146395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146395, "pid": 5, "tid": 7, "ts": 1716454223697443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639245, "dur": 8,
    "args": {
      "External id": 146395,
      "cbid": 211, "correlation": 146395
    }
  },
  {
    "ph": "s", "id": 146395, "pid": 76337, "tid": -914061504, "ts": 1716454223639245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223697460, "dur": 85,
    "args": {
      "External id": 146404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146404,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146404, "pid": 5, "tid": 7, "ts": 1716454223697460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639327, "dur": 14,
    "args": {
      "External id": 146404,
      "cbid": 211, "correlation": 146404
    }
  },
  {
    "ph": "s", "id": 146404, "pid": 76337, "tid": -914061504, "ts": 1716454223639327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223697546, "dur": 30,
    "args": {
      "External id": 146426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146426,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146426, "pid": 5, "tid": 7, "ts": 1716454223697546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639383, "dur": 10,
    "args": {
      "External id": 146426,
      "cbid": 211, "correlation": 146426
    }
  },
  {
    "ph": "s", "id": 146426, "pid": 76337, "tid": -914061504, "ts": 1716454223639383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639473, "dur": 2,
    "args": {
      "External id": 146437,
      "cbid": 251, "correlation": 146437
    }
  },
  {
    "ph": "f", "id": 146437, "pid": 76337, "tid": -914061504, "ts": 1716454223639473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223697577, "dur": 164,
    "args": {
      "External id": 146438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146438,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146438, "pid": 5, "tid": 7, "ts": 1716454223697577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639479, "dur": 13,
    "args": {
      "External id": 146438,
      "cbid": 211, "correlation": 146438
    }
  },
  {
    "ph": "s", "id": 146438, "pid": 76337, "tid": -914061504, "ts": 1716454223639479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639548, "dur": 1,
    "args": {
      "External id": 146449,
      "cbid": 251, "correlation": 146449
    }
  },
  {
    "ph": "f", "id": 146449, "pid": 76337, "tid": -914061504, "ts": 1716454223639548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223697742, "dur": 160,
    "args": {
      "External id": 146450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146450,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146450, "pid": 5, "tid": 7, "ts": 1716454223697742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639552, "dur": 11,
    "args": {
      "External id": 146450,
      "cbid": 211, "correlation": 146450
    }
  },
  {
    "ph": "s", "id": 146450, "pid": 76337, "tid": -914061504, "ts": 1716454223639552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639619, "dur": 1,
    "args": {
      "External id": 146461,
      "cbid": 251, "correlation": 146461
    }
  },
  {
    "ph": "f", "id": 146461, "pid": 76337, "tid": -914061504, "ts": 1716454223639619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223697904, "dur": 161,
    "args": {
      "External id": 146462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146462,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146462, "pid": 5, "tid": 7, "ts": 1716454223697904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639623, "dur": 11,
    "args": {
      "External id": 146462,
      "cbid": 211, "correlation": 146462
    }
  },
  {
    "ph": "s", "id": 146462, "pid": 76337, "tid": -914061504, "ts": 1716454223639623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223698067, "dur": 335,
    "args": {
      "External id": 146487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146487,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146487, "pid": 5, "tid": 7, "ts": 1716454223698067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639708, "dur": 13,
    "args": {
      "External id": 146487,
      "cbid": 211, "correlation": 146487
    }
  },
  {
    "ph": "s", "id": 146487, "pid": 76337, "tid": -914061504, "ts": 1716454223639708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639810, "dur": 1,
    "args": {
      "External id": 146505,
      "cbid": 251, "correlation": 146505
    }
  },
  {
    "ph": "f", "id": 146505, "pid": 76337, "tid": -914061504, "ts": 1716454223639810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223698403, "dur": 169,
    "args": {
      "External id": 146507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146507,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146507, "pid": 5, "tid": 7, "ts": 1716454223698403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639816, "dur": 14,
    "args": {
      "External id": 146507,
      "cbid": 211, "correlation": 146507
    }
  },
  {
    "ph": "s", "id": 146507, "pid": 76337, "tid": -914061504, "ts": 1716454223639816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223698574, "dur": 19,
    "args": {
      "External id": 146515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146515,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146515, "pid": 5, "tid": 7, "ts": 1716454223698574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639887, "dur": 12,
    "args": {
      "External id": 146515,
      "cbid": 211, "correlation": 146515
    }
  },
  {
    "ph": "s", "id": 146515, "pid": 76337, "tid": -914061504, "ts": 1716454223639887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223698594, "dur": 27,
    "args": {
      "External id": 146523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146523,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146523, "pid": 5, "tid": 7, "ts": 1716454223698594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223639927, "dur": 8,
    "args": {
      "External id": 146523,
      "cbid": 211, "correlation": 146523
    }
  },
  {
    "ph": "s", "id": 146523, "pid": 76337, "tid": -914061504, "ts": 1716454223639927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223698623, "dur": 18,
    "args": {
      "External id": 146534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146534, "pid": 5, "tid": 7, "ts": 1716454223698623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640007, "dur": 14,
    "args": {
      "External id": 146534,
      "cbid": 211, "correlation": 146534
    }
  },
  {
    "ph": "s", "id": 146534, "pid": 76337, "tid": -914061504, "ts": 1716454223640007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223698642, "dur": 16,
    "args": {
      "External id": 146556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146556,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146556, "pid": 5, "tid": 7, "ts": 1716454223698642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640039, "dur": 8,
    "args": {
      "External id": 146556,
      "cbid": 211, "correlation": 146556
    }
  },
  {
    "ph": "s", "id": 146556, "pid": 76337, "tid": -914061504, "ts": 1716454223640039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640127, "dur": 1,
    "args": {
      "External id": 146567,
      "cbid": 251, "correlation": 146567
    }
  },
  {
    "ph": "f", "id": 146567, "pid": 76337, "tid": -914061504, "ts": 1716454223640127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223698660, "dur": 89,
    "args": {
      "External id": 146568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146568,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146568, "pid": 5, "tid": 7, "ts": 1716454223698660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640132, "dur": 13,
    "args": {
      "External id": 146568,
      "cbid": 211, "correlation": 146568
    }
  },
  {
    "ph": "s", "id": 146568, "pid": 76337, "tid": -914061504, "ts": 1716454223640132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640200, "dur": 1,
    "args": {
      "External id": 146579,
      "cbid": 251, "correlation": 146579
    }
  },
  {
    "ph": "f", "id": 146579, "pid": 76337, "tid": -914061504, "ts": 1716454223640200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640204, "dur": 0,
    "args": {
      "External id": 146580,
      "cbid": 251, "correlation": 146580
    }
  },
  {
    "ph": "f", "id": 146580, "pid": 76337, "tid": -914061504, "ts": 1716454223640204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223698750, "dur": 12,
    "args": {
      "External id": 146581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146581,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146581, "pid": 5, "tid": 7, "ts": 1716454223698750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640206, "dur": 12,
    "args": {
      "External id": 146581,
      "cbid": 211, "correlation": 146581
    }
  },
  {
    "ph": "s", "id": 146581, "pid": 76337, "tid": -914061504, "ts": 1716454223640206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223698764, "dur": 5,
    "args": {
      "External id": 146583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146583,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146583, "pid": 5, "tid": 7, "ts": 1716454223698764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640219, "dur": 6,
    "args": {
      "External id": 146583,
      "cbid": 211, "correlation": 146583
    }
  },
  {
    "ph": "s", "id": 146583, "pid": 76337, "tid": -914061504, "ts": 1716454223640219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640278, "dur": 1,
    "args": {
      "External id": 146594,
      "cbid": 251, "correlation": 146594
    }
  },
  {
    "ph": "f", "id": 146594, "pid": 76337, "tid": -914061504, "ts": 1716454223640278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640282, "dur": 0,
    "args": {
      "External id": 146595,
      "cbid": 251, "correlation": 146595
    }
  },
  {
    "ph": "f", "id": 146595, "pid": 76337, "tid": -914061504, "ts": 1716454223640282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223698771, "dur": 8,
    "args": {
      "External id": 146596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146596,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146596, "pid": 5, "tid": 7, "ts": 1716454223698771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640283, "dur": 11,
    "args": {
      "External id": 146596,
      "cbid": 211, "correlation": 146596
    }
  },
  {
    "ph": "s", "id": 146596, "pid": 76337, "tid": -914061504, "ts": 1716454223640283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223698780, "dur": 4,
    "args": {
      "External id": 146598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146598,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146598, "pid": 5, "tid": 7, "ts": 1716454223698780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640296, "dur": 6,
    "args": {
      "External id": 146598,
      "cbid": 211, "correlation": 146598
    }
  },
  {
    "ph": "s", "id": 146598, "pid": 76337, "tid": -914061504, "ts": 1716454223640296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223698785, "dur": 55,
    "args": {
      "External id": 146623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146623,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146623, "pid": 5, "tid": 7, "ts": 1716454223698785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640373, "dur": 13,
    "args": {
      "External id": 146623,
      "cbid": 211, "correlation": 146623
    }
  },
  {
    "ph": "s", "id": 146623, "pid": 76337, "tid": -914061504, "ts": 1716454223640373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640472, "dur": 1,
    "args": {
      "External id": 146641,
      "cbid": 251, "correlation": 146641
    }
  },
  {
    "ph": "f", "id": 146641, "pid": 76337, "tid": -914061504, "ts": 1716454223640472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223698842, "dur": 91,
    "args": {
      "External id": 146643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146643,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146643, "pid": 5, "tid": 7, "ts": 1716454223698842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640478, "dur": 14,
    "args": {
      "External id": 146643,
      "cbid": 211, "correlation": 146643
    }
  },
  {
    "ph": "s", "id": 146643, "pid": 76337, "tid": -914061504, "ts": 1716454223640478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223698935, "dur": 10,
    "args": {
      "External id": 146651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146651,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146651, "pid": 5, "tid": 7, "ts": 1716454223698935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640549, "dur": 12,
    "args": {
      "External id": 146651,
      "cbid": 211, "correlation": 146651
    }
  },
  {
    "ph": "s", "id": 146651, "pid": 76337, "tid": -914061504, "ts": 1716454223640549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223698946, "dur": 21,
    "args": {
      "External id": 146659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146659,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146659, "pid": 5, "tid": 7, "ts": 1716454223698946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640591, "dur": 9,
    "args": {
      "External id": 146659,
      "cbid": 211, "correlation": 146659
    }
  },
  {
    "ph": "s", "id": 146659, "pid": 76337, "tid": -914061504, "ts": 1716454223640591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223698968, "dur": 18,
    "args": {
      "External id": 146681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146681,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146681, "pid": 5, "tid": 7, "ts": 1716454223698968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640643, "dur": 10,
    "args": {
      "External id": 146681,
      "cbid": 211, "correlation": 146681
    }
  },
  {
    "ph": "s", "id": 146681, "pid": 76337, "tid": -914061504, "ts": 1716454223640643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640729, "dur": 1,
    "args": {
      "External id": 146697,
      "cbid": 251, "correlation": 146697
    }
  },
  {
    "ph": "f", "id": 146697, "pid": 76337, "tid": -914061504, "ts": 1716454223640729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640735, "dur": 0,
    "args": {
      "External id": 146699,
      "cbid": 251, "correlation": 146699
    }
  },
  {
    "ph": "f", "id": 146699, "pid": 76337, "tid": -914061504, "ts": 1716454223640735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223698987, "dur": 496,
    "args": {
      "External id": 146700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146700,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146700, "pid": 5, "tid": 7, "ts": 1716454223698987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640737, "dur": 12,
    "args": {
      "External id": 146700,
      "cbid": 211, "correlation": 146700
    }
  },
  {
    "ph": "s", "id": 146700, "pid": 76337, "tid": -914061504, "ts": 1716454223640737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223699484, "dur": 67,
    "args": {
      "External id": 146708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146708, "pid": 5, "tid": 7, "ts": 1716454223699484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640801, "dur": 13,
    "args": {
      "External id": 146708,
      "cbid": 211, "correlation": 146708
    }
  },
  {
    "ph": "s", "id": 146708, "pid": 76337, "tid": -914061504, "ts": 1716454223640801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223699553, "dur": 67,
    "args": {
      "External id": 146716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146716, "pid": 5, "tid": 7, "ts": 1716454223699553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640833, "dur": 8,
    "args": {
      "External id": 146716,
      "cbid": 211, "correlation": 146716
    }
  },
  {
    "ph": "s", "id": 146716, "pid": 76337, "tid": -914061504, "ts": 1716454223640833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640914, "dur": 1,
    "args": {
      "External id": 146732,
      "cbid": 251, "correlation": 146732
    }
  },
  {
    "ph": "f", "id": 146732, "pid": 76337, "tid": -914061504, "ts": 1716454223640914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223699622, "dur": 1,
    "args": {
      "External id": 146734,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 146734,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 146734, "pid": 5, "tid": 7, "ts": 1716454223699622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640918, "dur": 9,
    "args": {
      "External id": 146734,
      "cbid": 51, "correlation": 146734
    }
  },
  {
    "ph": "s", "id": 146734, "pid": 76337, "tid": -914061504, "ts": 1716454223640918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223699626, "dur": 271,
    "args": {
      "External id": 146735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146735,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146735, "pid": 5, "tid": 7, "ts": 1716454223699626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640929, "dur": 12,
    "args": {
      "External id": 146735,
      "cbid": 211, "correlation": 146735
    }
  },
  {
    "ph": "s", "id": 146735, "pid": 76337, "tid": -914061504, "ts": 1716454223640929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223699898, "dur": 14,
    "args": {
      "External id": 146743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146743,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146743, "pid": 5, "tid": 7, "ts": 1716454223699898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223640972, "dur": 20,
    "args": {
      "External id": 146743,
      "cbid": 211, "correlation": 146743
    }
  },
  {
    "ph": "s", "id": 146743, "pid": 76337, "tid": -914061504, "ts": 1716454223640972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223699913, "dur": 37,
    "args": {
      "External id": 146754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146754,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146754, "pid": 5, "tid": 7, "ts": 1716454223699913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641051, "dur": 12,
    "args": {
      "External id": 146754,
      "cbid": 211, "correlation": 146754
    }
  },
  {
    "ph": "s", "id": 146754, "pid": 76337, "tid": -914061504, "ts": 1716454223641051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641116, "dur": 0,
    "args": {
      "External id": 146766,
      "cbid": 317, "correlation": 146766
    }
  },
  {
    "ph": "f", "id": 146766, "pid": 76337, "tid": -914061504, "ts": 1716454223641116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641117, "dur": 0,
    "args": {
      "External id": 146767,
      "cbid": 203, "correlation": 146767
    }
  },
  {
    "ph": "f", "id": 146767, "pid": 76337, "tid": -914061504, "ts": 1716454223641117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641117, "dur": 0,
    "args": {
      "External id": 146768,
      "cbid": 205, "correlation": 146768
    }
  },
  {
    "ph": "f", "id": 146768, "pid": 76337, "tid": -914061504, "ts": 1716454223641117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223699951, "dur": 14,
    "args": {
      "External id": 146772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146772,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146772, "pid": 5, "tid": 7, "ts": 1716454223699951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641133, "dur": 12,
    "args": {
      "External id": 146772,
      "cbid": 211, "correlation": 146772
    }
  },
  {
    "ph": "s", "id": 146772, "pid": 76337, "tid": -914061504, "ts": 1716454223641133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223699967, "dur": 4,
    "args": {
      "External id": 146774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 146774, "pid": 5, "tid": 7, "ts": 1716454223699967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641150, "dur": 6,
    "args": {
      "External id": 146774,
      "cbid": 211, "correlation": 146774
    }
  },
  {
    "ph": "s", "id": 146774, "pid": 76337, "tid": -914061504, "ts": 1716454223641150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641158, "dur": 0,
    "args": {
      "External id": 146775,
      "cbid": 51, "correlation": 146775
    }
  },
  {
    "ph": "s", "id": 146775, "pid": 76337, "tid": -914061504, "ts": 1716454223641158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223699973, "dur": 98,
    "args": {
      "External id": 146776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146776,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 146776, "pid": 5, "tid": 7, "ts": 1716454223699973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641159, "dur": 5,
    "args": {
      "External id": 146776,
      "cbid": 211, "correlation": 146776
    }
  },
  {
    "ph": "s", "id": 146776, "pid": 76337, "tid": -914061504, "ts": 1716454223641159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223700072, "dur": 16,
    "args": {
      "External id": 146781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146781,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146781, "pid": 5, "tid": 7, "ts": 1716454223700072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641187, "dur": 8,
    "args": {
      "External id": 146781,
      "cbid": 211, "correlation": 146781
    }
  },
  {
    "ph": "s", "id": 146781, "pid": 76337, "tid": -914061504, "ts": 1716454223641187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223700090, "dur": 12,
    "args": {
      "External id": 146789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146789, "pid": 5, "tid": 7, "ts": 1716454223700090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641218, "dur": 8,
    "args": {
      "External id": 146789,
      "cbid": 211, "correlation": 146789
    }
  },
  {
    "ph": "s", "id": 146789, "pid": 76337, "tid": -914061504, "ts": 1716454223641218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223700103, "dur": 25,
    "args": {
      "External id": 146798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146798,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146798, "pid": 5, "tid": 7, "ts": 1716454223700103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641256, "dur": 10,
    "args": {
      "External id": 146798,
      "cbid": 211, "correlation": 146798
    }
  },
  {
    "ph": "s", "id": 146798, "pid": 76337, "tid": -914061504, "ts": 1716454223641256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223700129, "dur": 25,
    "args": {
      "External id": 146818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146818,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 146818, "pid": 5, "tid": 7, "ts": 1716454223700129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641327, "dur": 12,
    "args": {
      "External id": 146818,
      "cbid": 211, "correlation": 146818
    }
  },
  {
    "ph": "s", "id": 146818, "pid": 76337, "tid": -914061504, "ts": 1716454223641327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223700155, "dur": 5,
    "args": {
      "External id": 146830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146830,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 146830, "pid": 5, "tid": 7, "ts": 1716454223700155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641349, "dur": 6,
    "args": {
      "External id": 146830,
      "cbid": 211, "correlation": 146830
    }
  },
  {
    "ph": "s", "id": 146830, "pid": 76337, "tid": -914061504, "ts": 1716454223641349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223700161, "dur": 25,
    "args": {
      "External id": 146833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146833, "pid": 5, "tid": 7, "ts": 1716454223700161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641367, "dur": 6,
    "args": {
      "External id": 146833,
      "cbid": 211, "correlation": 146833
    }
  },
  {
    "ph": "s", "id": 146833, "pid": 76337, "tid": -914061504, "ts": 1716454223641367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223700187, "dur": 17,
    "args": {
      "External id": 146842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146842,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146842, "pid": 5, "tid": 7, "ts": 1716454223700187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641406, "dur": 9,
    "args": {
      "External id": 146842,
      "cbid": 211, "correlation": 146842
    }
  },
  {
    "ph": "s", "id": 146842, "pid": 76337, "tid": -914061504, "ts": 1716454223641406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641457, "dur": 0,
    "args": {
      "External id": 146852,
      "cbid": 317, "correlation": 146852
    }
  },
  {
    "ph": "f", "id": 146852, "pid": 76337, "tid": -914061504, "ts": 1716454223641457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641458, "dur": 0,
    "args": {
      "External id": 146853,
      "cbid": 203, "correlation": 146853
    }
  },
  {
    "ph": "f", "id": 146853, "pid": 76337, "tid": -914061504, "ts": 1716454223641458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641459, "dur": 0,
    "args": {
      "External id": 146854,
      "cbid": 205, "correlation": 146854
    }
  },
  {
    "ph": "f", "id": 146854, "pid": 76337, "tid": -914061504, "ts": 1716454223641459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223700206, "dur": 18,
    "args": {
      "External id": 146858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146858,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146858, "pid": 5, "tid": 7, "ts": 1716454223700206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641474, "dur": 12,
    "args": {
      "External id": 146858,
      "cbid": 211, "correlation": 146858
    }
  },
  {
    "ph": "s", "id": 146858, "pid": 76337, "tid": -914061504, "ts": 1716454223641474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223700225, "dur": 243,
    "args": {
      "External id": 146860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146860,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146860, "pid": 5, "tid": 7, "ts": 1716454223700225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641489, "dur": 5,
    "args": {
      "External id": 146860,
      "cbid": 211, "correlation": 146860
    }
  },
  {
    "ph": "s", "id": 146860, "pid": 76337, "tid": -914061504, "ts": 1716454223641489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223700470, "dur": 1,
    "args": {
      "External id": 146862,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 146862,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 146862, "pid": 5, "tid": 7, "ts": 1716454223700470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641500, "dur": 8,
    "args": {
      "External id": 146862,
      "cbid": 51, "correlation": 146862
    }
  },
  {
    "ph": "s", "id": 146862, "pid": 76337, "tid": -914061504, "ts": 1716454223641500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223700474, "dur": 814,
    "args": {
      "External id": 146863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146863,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146863, "pid": 5, "tid": 7, "ts": 1716454223700474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641509, "dur": 6,
    "args": {
      "External id": 146863,
      "cbid": 211, "correlation": 146863
    }
  },
  {
    "ph": "s", "id": 146863, "pid": 76337, "tid": -914061504, "ts": 1716454223641509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223701289, "dur": 14,
    "args": {
      "External id": 146865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146865,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146865, "pid": 5, "tid": 7, "ts": 1716454223701289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641519, "dur": 6,
    "args": {
      "External id": 146865,
      "cbid": 211, "correlation": 146865
    }
  },
  {
    "ph": "s", "id": 146865, "pid": 76337, "tid": -914061504, "ts": 1716454223641519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223701303, "dur": 15,
    "args": {
      "External id": 146871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146871,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146871, "pid": 5, "tid": 7, "ts": 1716454223701303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641548, "dur": 8,
    "args": {
      "External id": 146871,
      "cbid": 211, "correlation": 146871
    }
  },
  {
    "ph": "s", "id": 146871, "pid": 76337, "tid": -914061504, "ts": 1716454223641548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223701319, "dur": 4,
    "args": {
      "External id": 146879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146879,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 146879, "pid": 5, "tid": 7, "ts": 1716454223701319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641592, "dur": 9,
    "args": {
      "External id": 146879,
      "cbid": 211, "correlation": 146879
    }
  },
  {
    "ph": "s", "id": 146879, "pid": 76337, "tid": -914061504, "ts": 1716454223641592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641656, "dur": 1,
    "args": {
      "External id": 146895,
      "cbid": 251, "correlation": 146895
    }
  },
  {
    "ph": "f", "id": 146895, "pid": 76337, "tid": -914061504, "ts": 1716454223641656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641661, "dur": 0,
    "args": {
      "External id": 146897,
      "cbid": 251, "correlation": 146897
    }
  },
  {
    "ph": "f", "id": 146897, "pid": 76337, "tid": -914061504, "ts": 1716454223641661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223701325, "dur": 13,
    "args": {
      "External id": 146898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146898,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146898, "pid": 5, "tid": 7, "ts": 1716454223701325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641663, "dur": 11,
    "args": {
      "External id": 146898,
      "cbid": 211, "correlation": 146898
    }
  },
  {
    "ph": "s", "id": 146898, "pid": 76337, "tid": -914061504, "ts": 1716454223641663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223701339, "dur": 5,
    "args": {
      "External id": 146900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146900,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146900, "pid": 5, "tid": 7, "ts": 1716454223701339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641676, "dur": 5,
    "args": {
      "External id": 146900,
      "cbid": 211, "correlation": 146900
    }
  },
  {
    "ph": "s", "id": 146900, "pid": 76337, "tid": -914061504, "ts": 1716454223641676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223701346, "dur": 17,
    "args": {
      "External id": 146910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146910,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146910, "pid": 5, "tid": 7, "ts": 1716454223701346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641734, "dur": 12,
    "args": {
      "External id": 146910,
      "cbid": 211, "correlation": 146910
    }
  },
  {
    "ph": "s", "id": 146910, "pid": 76337, "tid": -914061504, "ts": 1716454223641734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223701364, "dur": 17,
    "args": {
      "External id": 146930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146930,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 146930, "pid": 5, "tid": 7, "ts": 1716454223701364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641799, "dur": 11,
    "args": {
      "External id": 146930,
      "cbid": 211, "correlation": 146930
    }
  },
  {
    "ph": "s", "id": 146930, "pid": 76337, "tid": -914061504, "ts": 1716454223641799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223701383, "dur": 5,
    "args": {
      "External id": 146942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146942,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 146942, "pid": 5, "tid": 7, "ts": 1716454223701383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641820, "dur": 6,
    "args": {
      "External id": 146942,
      "cbid": 211, "correlation": 146942
    }
  },
  {
    "ph": "s", "id": 146942, "pid": 76337, "tid": -914061504, "ts": 1716454223641820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223701389, "dur": 17,
    "args": {
      "External id": 146945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146945, "pid": 5, "tid": 7, "ts": 1716454223701389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641838, "dur": 7,
    "args": {
      "External id": 146945,
      "cbid": 211, "correlation": 146945
    }
  },
  {
    "ph": "s", "id": 146945, "pid": 76337, "tid": -914061504, "ts": 1716454223641838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223701407, "dur": 11,
    "args": {
      "External id": 146954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146954,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146954, "pid": 5, "tid": 7, "ts": 1716454223701407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641879, "dur": 10,
    "args": {
      "External id": 146954,
      "cbid": 211, "correlation": 146954
    }
  },
  {
    "ph": "s", "id": 146954, "pid": 76337, "tid": -914061504, "ts": 1716454223641879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641942, "dur": 0,
    "args": {
      "External id": 146964,
      "cbid": 317, "correlation": 146964
    }
  },
  {
    "ph": "f", "id": 146964, "pid": 76337, "tid": -914061504, "ts": 1716454223641942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641943, "dur": 0,
    "args": {
      "External id": 146965,
      "cbid": 203, "correlation": 146965
    }
  },
  {
    "ph": "f", "id": 146965, "pid": 76337, "tid": -914061504, "ts": 1716454223641943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641943, "dur": 0,
    "args": {
      "External id": 146966,
      "cbid": 205, "correlation": 146966
    }
  },
  {
    "ph": "f", "id": 146966, "pid": 76337, "tid": -914061504, "ts": 1716454223641943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223701419, "dur": 11,
    "args": {
      "External id": 146970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146970,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146970, "pid": 5, "tid": 7, "ts": 1716454223701419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641959, "dur": 12,
    "args": {
      "External id": 146970,
      "cbid": 211, "correlation": 146970
    }
  },
  {
    "ph": "s", "id": 146970, "pid": 76337, "tid": -914061504, "ts": 1716454223641959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223701432, "dur": 163,
    "args": {
      "External id": 146972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146972,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146972, "pid": 5, "tid": 7, "ts": 1716454223701432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641983, "dur": 6,
    "args": {
      "External id": 146972,
      "cbid": 211, "correlation": 146972
    }
  },
  {
    "ph": "s", "id": 146972, "pid": 76337, "tid": -914061504, "ts": 1716454223641983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223701597, "dur": 1,
    "args": {
      "External id": 146974,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 146974,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 146974, "pid": 5, "tid": 7, "ts": 1716454223701597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223641995, "dur": 6,
    "args": {
      "External id": 146974,
      "cbid": 51, "correlation": 146974
    }
  },
  {
    "ph": "s", "id": 146974, "pid": 76337, "tid": -914061504, "ts": 1716454223641995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223701600, "dur": 650,
    "args": {
      "External id": 146975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146975,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 146975, "pid": 5, "tid": 7, "ts": 1716454223701600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642003, "dur": 6,
    "args": {
      "External id": 146975,
      "cbid": 211, "correlation": 146975
    }
  },
  {
    "ph": "s", "id": 146975, "pid": 76337, "tid": -914061504, "ts": 1716454223642003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223702252, "dur": 12,
    "args": {
      "External id": 146977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146977,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146977, "pid": 5, "tid": 7, "ts": 1716454223702252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642013, "dur": 5,
    "args": {
      "External id": 146977,
      "cbid": 211, "correlation": 146977
    }
  },
  {
    "ph": "s", "id": 146977, "pid": 76337, "tid": -914061504, "ts": 1716454223642013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223702266, "dur": 15,
    "args": {
      "External id": 146983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146983, "pid": 5, "tid": 7, "ts": 1716454223702266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642042, "dur": 8,
    "args": {
      "External id": 146983,
      "cbid": 211, "correlation": 146983
    }
  },
  {
    "ph": "s", "id": 146983, "pid": 76337, "tid": -914061504, "ts": 1716454223642042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642101, "dur": 0,
    "args": {
      "External id": 146993,
      "cbid": 317, "correlation": 146993
    }
  },
  {
    "ph": "f", "id": 146993, "pid": 76337, "tid": -914061504, "ts": 1716454223642101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642102, "dur": 0,
    "args": {
      "External id": 146994,
      "cbid": 203, "correlation": 146994
    }
  },
  {
    "ph": "f", "id": 146994, "pid": 76337, "tid": -914061504, "ts": 1716454223642102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642102, "dur": 0,
    "args": {
      "External id": 146995,
      "cbid": 205, "correlation": 146995
    }
  },
  {
    "ph": "f", "id": 146995, "pid": 76337, "tid": -914061504, "ts": 1716454223642102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223702282, "dur": 17,
    "args": {
      "External id": 146999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 146999,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 146999, "pid": 5, "tid": 7, "ts": 1716454223702282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642116, "dur": 11,
    "args": {
      "External id": 146999,
      "cbid": 211, "correlation": 146999
    }
  },
  {
    "ph": "s", "id": 146999, "pid": 76337, "tid": -914061504, "ts": 1716454223642116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223702300, "dur": 4,
    "args": {
      "External id": 147001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147001,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147001, "pid": 5, "tid": 7, "ts": 1716454223702300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642132, "dur": 7,
    "args": {
      "External id": 147001,
      "cbid": 211, "correlation": 147001
    }
  },
  {
    "ph": "s", "id": 147001, "pid": 76337, "tid": -914061504, "ts": 1716454223642132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642142, "dur": 0,
    "args": {
      "External id": 147002,
      "cbid": 51, "correlation": 147002
    }
  },
  {
    "ph": "s", "id": 147002, "pid": 76337, "tid": -914061504, "ts": 1716454223642142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223702305, "dur": 132,
    "args": {
      "External id": 147003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147003,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 147003, "pid": 5, "tid": 7, "ts": 1716454223702305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642142, "dur": 5,
    "args": {
      "External id": 147003,
      "cbid": 211, "correlation": 147003
    }
  },
  {
    "ph": "s", "id": 147003, "pid": 76337, "tid": -914061504, "ts": 1716454223642142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223702439, "dur": 15,
    "args": {
      "External id": 147008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147008,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147008, "pid": 5, "tid": 7, "ts": 1716454223702439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642167, "dur": 8,
    "args": {
      "External id": 147008,
      "cbid": 211, "correlation": 147008
    }
  },
  {
    "ph": "s", "id": 147008, "pid": 76337, "tid": -914061504, "ts": 1716454223642167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223702455, "dur": 14,
    "args": {
      "External id": 147016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147016,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147016, "pid": 5, "tid": 7, "ts": 1716454223702455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642196, "dur": 8,
    "args": {
      "External id": 147016,
      "cbid": 211, "correlation": 147016
    }
  },
  {
    "ph": "s", "id": 147016, "pid": 76337, "tid": -914061504, "ts": 1716454223642196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223702470, "dur": 10,
    "args": {
      "External id": 147024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147024,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147024, "pid": 5, "tid": 7, "ts": 1716454223702470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642225, "dur": 8,
    "args": {
      "External id": 147024,
      "cbid": 211, "correlation": 147024
    }
  },
  {
    "ph": "s", "id": 147024, "pid": 76337, "tid": -914061504, "ts": 1716454223642225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223702482, "dur": 19,
    "args": {
      "External id": 147044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147044,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 147044, "pid": 5, "tid": 7, "ts": 1716454223702482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642308, "dur": 12,
    "args": {
      "External id": 147044,
      "cbid": 211, "correlation": 147044
    }
  },
  {
    "ph": "s", "id": 147044, "pid": 76337, "tid": -914061504, "ts": 1716454223642308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223702502, "dur": 4,
    "args": {
      "External id": 147056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147056,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 147056, "pid": 5, "tid": 7, "ts": 1716454223702502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642330, "dur": 6,
    "args": {
      "External id": 147056,
      "cbid": 211, "correlation": 147056
    }
  },
  {
    "ph": "s", "id": 147056, "pid": 76337, "tid": -914061504, "ts": 1716454223642330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223702507, "dur": 18,
    "args": {
      "External id": 147059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147059, "pid": 5, "tid": 7, "ts": 1716454223702507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642348, "dur": 6,
    "args": {
      "External id": 147059,
      "cbid": 211, "correlation": 147059
    }
  },
  {
    "ph": "s", "id": 147059, "pid": 76337, "tid": -914061504, "ts": 1716454223642348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642404, "dur": 0,
    "args": {
      "External id": 147070,
      "cbid": 317, "correlation": 147070
    }
  },
  {
    "ph": "f", "id": 147070, "pid": 76337, "tid": -914061504, "ts": 1716454223642404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642404, "dur": 0,
    "args": {
      "External id": 147071,
      "cbid": 203, "correlation": 147071
    }
  },
  {
    "ph": "f", "id": 147071, "pid": 76337, "tid": -914061504, "ts": 1716454223642404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642405, "dur": 0,
    "args": {
      "External id": 147072,
      "cbid": 205, "correlation": 147072
    }
  },
  {
    "ph": "f", "id": 147072, "pid": 76337, "tid": -914061504, "ts": 1716454223642405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223702526, "dur": 11,
    "args": {
      "External id": 147076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147076, "pid": 5, "tid": 7, "ts": 1716454223702526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642418, "dur": 11,
    "args": {
      "External id": 147076,
      "cbid": 211, "correlation": 147076
    }
  },
  {
    "ph": "s", "id": 147076, "pid": 76337, "tid": -914061504, "ts": 1716454223642418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223702539, "dur": 3,
    "args": {
      "External id": 147078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147078,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147078, "pid": 5, "tid": 7, "ts": 1716454223702539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642434, "dur": 7,
    "args": {
      "External id": 147078,
      "cbid": 211, "correlation": 147078
    }
  },
  {
    "ph": "s", "id": 147078, "pid": 76337, "tid": -914061504, "ts": 1716454223642434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642444, "dur": 0,
    "args": {
      "External id": 147079,
      "cbid": 51, "correlation": 147079
    }
  },
  {
    "ph": "s", "id": 147079, "pid": 76337, "tid": -914061504, "ts": 1716454223642444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223702544, "dur": 90,
    "args": {
      "External id": 147080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147080,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 147080, "pid": 5, "tid": 7, "ts": 1716454223702544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642444, "dur": 5,
    "args": {
      "External id": 147080,
      "cbid": 211, "correlation": 147080
    }
  },
  {
    "ph": "s", "id": 147080, "pid": 76337, "tid": -914061504, "ts": 1716454223642444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223702635, "dur": 16,
    "args": {
      "External id": 147085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147085, "pid": 5, "tid": 7, "ts": 1716454223702635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642470, "dur": 8,
    "args": {
      "External id": 147085,
      "cbid": 211, "correlation": 147085
    }
  },
  {
    "ph": "s", "id": 147085, "pid": 76337, "tid": -914061504, "ts": 1716454223642470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223702652, "dur": 85,
    "args": {
      "External id": 147094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147094,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147094, "pid": 5, "tid": 7, "ts": 1716454223702652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642550, "dur": 14,
    "args": {
      "External id": 147094,
      "cbid": 211, "correlation": 147094
    }
  },
  {
    "ph": "s", "id": 147094, "pid": 76337, "tid": -914061504, "ts": 1716454223642550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223702738, "dur": 30,
    "args": {
      "External id": 147116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147116,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147116, "pid": 5, "tid": 7, "ts": 1716454223702738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642606, "dur": 10,
    "args": {
      "External id": 147116,
      "cbid": 211, "correlation": 147116
    }
  },
  {
    "ph": "s", "id": 147116, "pid": 76337, "tid": -914061504, "ts": 1716454223642606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642696, "dur": 1,
    "args": {
      "External id": 147127,
      "cbid": 251, "correlation": 147127
    }
  },
  {
    "ph": "f", "id": 147127, "pid": 76337, "tid": -914061504, "ts": 1716454223642696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223702770, "dur": 165,
    "args": {
      "External id": 147128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147128,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147128, "pid": 5, "tid": 7, "ts": 1716454223702770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642701, "dur": 13,
    "args": {
      "External id": 147128,
      "cbid": 211, "correlation": 147128
    }
  },
  {
    "ph": "s", "id": 147128, "pid": 76337, "tid": -914061504, "ts": 1716454223642701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642771, "dur": 1,
    "args": {
      "External id": 147139,
      "cbid": 251, "correlation": 147139
    }
  },
  {
    "ph": "f", "id": 147139, "pid": 76337, "tid": -914061504, "ts": 1716454223642771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223702936, "dur": 158,
    "args": {
      "External id": 147140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147140,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147140, "pid": 5, "tid": 7, "ts": 1716454223702936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642775, "dur": 12,
    "args": {
      "External id": 147140,
      "cbid": 211, "correlation": 147140
    }
  },
  {
    "ph": "s", "id": 147140, "pid": 76337, "tid": -914061504, "ts": 1716454223642775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642840, "dur": 1,
    "args": {
      "External id": 147151,
      "cbid": 251, "correlation": 147151
    }
  },
  {
    "ph": "f", "id": 147151, "pid": 76337, "tid": -914061504, "ts": 1716454223642840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223703095, "dur": 158,
    "args": {
      "External id": 147152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147152,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147152, "pid": 5, "tid": 7, "ts": 1716454223703095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642844, "dur": 11,
    "args": {
      "External id": 147152,
      "cbid": 211, "correlation": 147152
    }
  },
  {
    "ph": "s", "id": 147152, "pid": 76337, "tid": -914061504, "ts": 1716454223642844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223703255, "dur": 334,
    "args": {
      "External id": 147177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147177,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147177, "pid": 5, "tid": 7, "ts": 1716454223703255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223642926, "dur": 13,
    "args": {
      "External id": 147177,
      "cbid": 211, "correlation": 147177
    }
  },
  {
    "ph": "s", "id": 147177, "pid": 76337, "tid": -914061504, "ts": 1716454223642926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643035, "dur": 1,
    "args": {
      "External id": 147195,
      "cbid": 251, "correlation": 147195
    }
  },
  {
    "ph": "f", "id": 147195, "pid": 76337, "tid": -914061504, "ts": 1716454223643035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223703590, "dur": 164,
    "args": {
      "External id": 147197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147197,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147197, "pid": 5, "tid": 7, "ts": 1716454223703590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643041, "dur": 14,
    "args": {
      "External id": 147197,
      "cbid": 211, "correlation": 147197
    }
  },
  {
    "ph": "s", "id": 147197, "pid": 76337, "tid": -914061504, "ts": 1716454223643041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223703755, "dur": 20,
    "args": {
      "External id": 147205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147205,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147205, "pid": 5, "tid": 7, "ts": 1716454223703755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643113, "dur": 12,
    "args": {
      "External id": 147205,
      "cbid": 211, "correlation": 147205
    }
  },
  {
    "ph": "s", "id": 147205, "pid": 76337, "tid": -914061504, "ts": 1716454223643113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223703776, "dur": 28,
    "args": {
      "External id": 147213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147213,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147213, "pid": 5, "tid": 7, "ts": 1716454223703776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643153, "dur": 8,
    "args": {
      "External id": 147213,
      "cbid": 211, "correlation": 147213
    }
  },
  {
    "ph": "s", "id": 147213, "pid": 76337, "tid": -914061504, "ts": 1716454223643153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223703805, "dur": 19,
    "args": {
      "External id": 147224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147224,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147224, "pid": 5, "tid": 7, "ts": 1716454223703805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643224, "dur": 12,
    "args": {
      "External id": 147224,
      "cbid": 211, "correlation": 147224
    }
  },
  {
    "ph": "s", "id": 147224, "pid": 76337, "tid": -914061504, "ts": 1716454223643224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223703825, "dur": 16,
    "args": {
      "External id": 147246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147246,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147246, "pid": 5, "tid": 7, "ts": 1716454223703825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643256, "dur": 7,
    "args": {
      "External id": 147246,
      "cbid": 211, "correlation": 147246
    }
  },
  {
    "ph": "s", "id": 147246, "pid": 76337, "tid": -914061504, "ts": 1716454223643256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643341, "dur": 1,
    "args": {
      "External id": 147257,
      "cbid": 251, "correlation": 147257
    }
  },
  {
    "ph": "f", "id": 147257, "pid": 76337, "tid": -914061504, "ts": 1716454223643341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223703843, "dur": 90,
    "args": {
      "External id": 147258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147258,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147258, "pid": 5, "tid": 7, "ts": 1716454223703843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643346, "dur": 13,
    "args": {
      "External id": 147258,
      "cbid": 211, "correlation": 147258
    }
  },
  {
    "ph": "s", "id": 147258, "pid": 76337, "tid": -914061504, "ts": 1716454223643346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643415, "dur": 1,
    "args": {
      "External id": 147269,
      "cbid": 251, "correlation": 147269
    }
  },
  {
    "ph": "f", "id": 147269, "pid": 76337, "tid": -914061504, "ts": 1716454223643415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643419, "dur": 0,
    "args": {
      "External id": 147270,
      "cbid": 251, "correlation": 147270
    }
  },
  {
    "ph": "f", "id": 147270, "pid": 76337, "tid": -914061504, "ts": 1716454223643419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223703934, "dur": 12,
    "args": {
      "External id": 147271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147271,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147271, "pid": 5, "tid": 7, "ts": 1716454223703934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643420, "dur": 11,
    "args": {
      "External id": 147271,
      "cbid": 211, "correlation": 147271
    }
  },
  {
    "ph": "s", "id": 147271, "pid": 76337, "tid": -914061504, "ts": 1716454223643420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223703948, "dur": 5,
    "args": {
      "External id": 147273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147273,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147273, "pid": 5, "tid": 7, "ts": 1716454223703948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643433, "dur": 7,
    "args": {
      "External id": 147273,
      "cbid": 211, "correlation": 147273
    }
  },
  {
    "ph": "s", "id": 147273, "pid": 76337, "tid": -914061504, "ts": 1716454223643433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643492, "dur": 1,
    "args": {
      "External id": 147284,
      "cbid": 251, "correlation": 147284
    }
  },
  {
    "ph": "f", "id": 147284, "pid": 76337, "tid": -914061504, "ts": 1716454223643492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643495, "dur": 0,
    "args": {
      "External id": 147285,
      "cbid": 251, "correlation": 147285
    }
  },
  {
    "ph": "f", "id": 147285, "pid": 76337, "tid": -914061504, "ts": 1716454223643495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223703954, "dur": 8,
    "args": {
      "External id": 147286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147286,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147286, "pid": 5, "tid": 7, "ts": 1716454223703954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643497, "dur": 12,
    "args": {
      "External id": 147286,
      "cbid": 211, "correlation": 147286
    }
  },
  {
    "ph": "s", "id": 147286, "pid": 76337, "tid": -914061504, "ts": 1716454223643497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223703964, "dur": 4,
    "args": {
      "External id": 147288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147288,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147288, "pid": 5, "tid": 7, "ts": 1716454223703964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643510, "dur": 5,
    "args": {
      "External id": 147288,
      "cbid": 211, "correlation": 147288
    }
  },
  {
    "ph": "s", "id": 147288, "pid": 76337, "tid": -914061504, "ts": 1716454223643510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223703969, "dur": 56,
    "args": {
      "External id": 147313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147313,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147313, "pid": 5, "tid": 7, "ts": 1716454223703969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643585, "dur": 12,
    "args": {
      "External id": 147313,
      "cbid": 211, "correlation": 147313
    }
  },
  {
    "ph": "s", "id": 147313, "pid": 76337, "tid": -914061504, "ts": 1716454223643585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643683, "dur": 1,
    "args": {
      "External id": 147331,
      "cbid": 251, "correlation": 147331
    }
  },
  {
    "ph": "f", "id": 147331, "pid": 76337, "tid": -914061504, "ts": 1716454223643683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223704026, "dur": 91,
    "args": {
      "External id": 147333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147333,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147333, "pid": 5, "tid": 7, "ts": 1716454223704026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643689, "dur": 13,
    "args": {
      "External id": 147333,
      "cbid": 211, "correlation": 147333
    }
  },
  {
    "ph": "s", "id": 147333, "pid": 76337, "tid": -914061504, "ts": 1716454223643689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223704118, "dur": 9,
    "args": {
      "External id": 147341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147341,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147341, "pid": 5, "tid": 7, "ts": 1716454223704118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643760, "dur": 12,
    "args": {
      "External id": 147341,
      "cbid": 211, "correlation": 147341
    }
  },
  {
    "ph": "s", "id": 147341, "pid": 76337, "tid": -914061504, "ts": 1716454223643760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223704129, "dur": 22,
    "args": {
      "External id": 147349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147349, "pid": 5, "tid": 7, "ts": 1716454223704129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643801, "dur": 10,
    "args": {
      "External id": 147349,
      "cbid": 211, "correlation": 147349
    }
  },
  {
    "ph": "s", "id": 147349, "pid": 76337, "tid": -914061504, "ts": 1716454223643801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223704152, "dur": 18,
    "args": {
      "External id": 147371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147371,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147371, "pid": 5, "tid": 7, "ts": 1716454223704152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643853, "dur": 11,
    "args": {
      "External id": 147371,
      "cbid": 211, "correlation": 147371
    }
  },
  {
    "ph": "s", "id": 147371, "pid": 76337, "tid": -914061504, "ts": 1716454223643853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643940, "dur": 1,
    "args": {
      "External id": 147387,
      "cbid": 251, "correlation": 147387
    }
  },
  {
    "ph": "f", "id": 147387, "pid": 76337, "tid": -914061504, "ts": 1716454223643940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643945, "dur": 0,
    "args": {
      "External id": 147389,
      "cbid": 251, "correlation": 147389
    }
  },
  {
    "ph": "f", "id": 147389, "pid": 76337, "tid": -914061504, "ts": 1716454223643945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223704171, "dur": 496,
    "args": {
      "External id": 147390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147390,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147390, "pid": 5, "tid": 7, "ts": 1716454223704171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223643947, "dur": 13,
    "args": {
      "External id": 147390,
      "cbid": 211, "correlation": 147390
    }
  },
  {
    "ph": "s", "id": 147390, "pid": 76337, "tid": -914061504, "ts": 1716454223643947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223704669, "dur": 67,
    "args": {
      "External id": 147398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147398,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147398, "pid": 5, "tid": 7, "ts": 1716454223704669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644022, "dur": 13,
    "args": {
      "External id": 147398,
      "cbid": 211, "correlation": 147398
    }
  },
  {
    "ph": "s", "id": 147398, "pid": 76337, "tid": -914061504, "ts": 1716454223644022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223704737, "dur": 66,
    "args": {
      "External id": 147406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147406,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147406, "pid": 5, "tid": 7, "ts": 1716454223704737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644053, "dur": 9,
    "args": {
      "External id": 147406,
      "cbid": 211, "correlation": 147406
    }
  },
  {
    "ph": "s", "id": 147406, "pid": 76337, "tid": -914061504, "ts": 1716454223644053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644132, "dur": 1,
    "args": {
      "External id": 147422,
      "cbid": 251, "correlation": 147422
    }
  },
  {
    "ph": "f", "id": 147422, "pid": 76337, "tid": -914061504, "ts": 1716454223644132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223704806, "dur": 1,
    "args": {
      "External id": 147424,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 147424,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 147424, "pid": 5, "tid": 7, "ts": 1716454223704806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644137, "dur": 9,
    "args": {
      "External id": 147424,
      "cbid": 51, "correlation": 147424
    }
  },
  {
    "ph": "s", "id": 147424, "pid": 76337, "tid": -914061504, "ts": 1716454223644137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223704809, "dur": 270,
    "args": {
      "External id": 147425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147425,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147425, "pid": 5, "tid": 7, "ts": 1716454223704809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644148, "dur": 12,
    "args": {
      "External id": 147425,
      "cbid": 211, "correlation": 147425
    }
  },
  {
    "ph": "s", "id": 147425, "pid": 76337, "tid": -914061504, "ts": 1716454223644148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223705081, "dur": 14,
    "args": {
      "External id": 147433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147433,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147433, "pid": 5, "tid": 7, "ts": 1716454223705081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644191, "dur": 10,
    "args": {
      "External id": 147433,
      "cbid": 211, "correlation": 147433
    }
  },
  {
    "ph": "s", "id": 147433, "pid": 76337, "tid": -914061504, "ts": 1716454223644191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223705096, "dur": 38,
    "args": {
      "External id": 147444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147444,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147444, "pid": 5, "tid": 7, "ts": 1716454223705096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644259, "dur": 13,
    "args": {
      "External id": 147444,
      "cbid": 211, "correlation": 147444
    }
  },
  {
    "ph": "s", "id": 147444, "pid": 76337, "tid": -914061504, "ts": 1716454223644259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644324, "dur": 0,
    "args": {
      "External id": 147456,
      "cbid": 317, "correlation": 147456
    }
  },
  {
    "ph": "f", "id": 147456, "pid": 76337, "tid": -914061504, "ts": 1716454223644324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644324, "dur": 0,
    "args": {
      "External id": 147457,
      "cbid": 203, "correlation": 147457
    }
  },
  {
    "ph": "f", "id": 147457, "pid": 76337, "tid": -914061504, "ts": 1716454223644324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644325, "dur": 0,
    "args": {
      "External id": 147458,
      "cbid": 205, "correlation": 147458
    }
  },
  {
    "ph": "f", "id": 147458, "pid": 76337, "tid": -914061504, "ts": 1716454223644325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223705136, "dur": 13,
    "args": {
      "External id": 147462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147462,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147462, "pid": 5, "tid": 7, "ts": 1716454223705136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644340, "dur": 12,
    "args": {
      "External id": 147462,
      "cbid": 211, "correlation": 147462
    }
  },
  {
    "ph": "s", "id": 147462, "pid": 76337, "tid": -914061504, "ts": 1716454223644340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223705150, "dur": 4,
    "args": {
      "External id": 147464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147464, "pid": 5, "tid": 7, "ts": 1716454223705150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644356, "dur": 6,
    "args": {
      "External id": 147464,
      "cbid": 211, "correlation": 147464
    }
  },
  {
    "ph": "s", "id": 147464, "pid": 76337, "tid": -914061504, "ts": 1716454223644356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644365, "dur": 0,
    "args": {
      "External id": 147465,
      "cbid": 51, "correlation": 147465
    }
  },
  {
    "ph": "s", "id": 147465, "pid": 76337, "tid": -914061504, "ts": 1716454223644365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223705155, "dur": 98,
    "args": {
      "External id": 147466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147466,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 147466, "pid": 5, "tid": 7, "ts": 1716454223705155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644366, "dur": 5,
    "args": {
      "External id": 147466,
      "cbid": 211, "correlation": 147466
    }
  },
  {
    "ph": "s", "id": 147466, "pid": 76337, "tid": -914061504, "ts": 1716454223644366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223705254, "dur": 16,
    "args": {
      "External id": 147471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147471,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147471, "pid": 5, "tid": 7, "ts": 1716454223705254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644393, "dur": 8,
    "args": {
      "External id": 147471,
      "cbid": 211, "correlation": 147471
    }
  },
  {
    "ph": "s", "id": 147471, "pid": 76337, "tid": -914061504, "ts": 1716454223644393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223705272, "dur": 12,
    "args": {
      "External id": 147479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147479,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147479, "pid": 5, "tid": 7, "ts": 1716454223705272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644425, "dur": 8,
    "args": {
      "External id": 147479,
      "cbid": 211, "correlation": 147479
    }
  },
  {
    "ph": "s", "id": 147479, "pid": 76337, "tid": -914061504, "ts": 1716454223644425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223705285, "dur": 56,
    "args": {
      "External id": 147490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147490,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147490, "pid": 5, "tid": 7, "ts": 1716454223705285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644488, "dur": 12,
    "args": {
      "External id": 147490,
      "cbid": 211, "correlation": 147490
    }
  },
  {
    "ph": "s", "id": 147490, "pid": 76337, "tid": -914061504, "ts": 1716454223644488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644543, "dur": 0,
    "args": {
      "External id": 147500,
      "cbid": 317, "correlation": 147500
    }
  },
  {
    "ph": "f", "id": 147500, "pid": 76337, "tid": -914061504, "ts": 1716454223644543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644544, "dur": 0,
    "args": {
      "External id": 147501,
      "cbid": 203, "correlation": 147501
    }
  },
  {
    "ph": "f", "id": 147501, "pid": 76337, "tid": -914061504, "ts": 1716454223644544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644544, "dur": 0,
    "args": {
      "External id": 147502,
      "cbid": 205, "correlation": 147502
    }
  },
  {
    "ph": "f", "id": 147502, "pid": 76337, "tid": -914061504, "ts": 1716454223644544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223705342, "dur": 40,
    "args": {
      "External id": 147506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147506,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147506, "pid": 5, "tid": 7, "ts": 1716454223705342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644562, "dur": 11,
    "args": {
      "External id": 147506,
      "cbid": 211, "correlation": 147506
    }
  },
  {
    "ph": "s", "id": 147506, "pid": 76337, "tid": -914061504, "ts": 1716454223644562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223705384, "dur": 164,
    "args": {
      "External id": 147508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147508,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147508, "pid": 5, "tid": 7, "ts": 1716454223705384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644575, "dur": 5,
    "args": {
      "External id": 147508,
      "cbid": 211, "correlation": 147508
    }
  },
  {
    "ph": "s", "id": 147508, "pid": 76337, "tid": -914061504, "ts": 1716454223644575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223705549, "dur": 1973,
    "args": {
      "External id": 147510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147510,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147510, "pid": 5, "tid": 7, "ts": 1716454223705549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644588, "dur": 10,
    "args": {
      "External id": 147510,
      "cbid": 211, "correlation": 147510
    }
  },
  {
    "ph": "s", "id": 147510, "pid": 76337, "tid": -914061504, "ts": 1716454223644588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223707524, "dur": 40,
    "args": {
      "External id": 147512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147512,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147512, "pid": 5, "tid": 7, "ts": 1716454223707524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644602, "dur": 6,
    "args": {
      "External id": 147512,
      "cbid": 211, "correlation": 147512
    }
  },
  {
    "ph": "s", "id": 147512, "pid": 76337, "tid": -914061504, "ts": 1716454223644602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223707565, "dur": 60,
    "args": {
      "External id": 147518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147518,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147518, "pid": 5, "tid": 7, "ts": 1716454223707565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644630, "dur": 9,
    "args": {
      "External id": 147518,
      "cbid": 211, "correlation": 147518
    }
  },
  {
    "ph": "s", "id": 147518, "pid": 76337, "tid": -914061504, "ts": 1716454223644630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223707626, "dur": 83,
    "args": {
      "External id": 147527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147527,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147527, "pid": 5, "tid": 7, "ts": 1716454223707626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644722, "dur": 13,
    "args": {
      "External id": 147527,
      "cbid": 211, "correlation": 147527
    }
  },
  {
    "ph": "s", "id": 147527, "pid": 76337, "tid": -914061504, "ts": 1716454223644722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223707711, "dur": 73,
    "args": {
      "External id": 147547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147547,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 147547, "pid": 5, "tid": 7, "ts": 1716454223707711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644792, "dur": 11,
    "args": {
      "External id": 147547,
      "cbid": 211, "correlation": 147547
    }
  },
  {
    "ph": "s", "id": 147547, "pid": 76337, "tid": -914061504, "ts": 1716454223644792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223707785, "dur": 5,
    "args": {
      "External id": 147559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147559,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 147559, "pid": 5, "tid": 7, "ts": 1716454223707785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644815, "dur": 6,
    "args": {
      "External id": 147559,
      "cbid": 211, "correlation": 147559
    }
  },
  {
    "ph": "s", "id": 147559, "pid": 76337, "tid": -914061504, "ts": 1716454223644815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223707791, "dur": 81,
    "args": {
      "External id": 147562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147562,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147562, "pid": 5, "tid": 7, "ts": 1716454223707791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644833, "dur": 6,
    "args": {
      "External id": 147562,
      "cbid": 211, "correlation": 147562
    }
  },
  {
    "ph": "s", "id": 147562, "pid": 76337, "tid": -914061504, "ts": 1716454223644833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223707873, "dur": 53,
    "args": {
      "External id": 147571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147571,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147571, "pid": 5, "tid": 7, "ts": 1716454223707873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644874, "dur": 9,
    "args": {
      "External id": 147571,
      "cbid": 211, "correlation": 147571
    }
  },
  {
    "ph": "s", "id": 147571, "pid": 76337, "tid": -914061504, "ts": 1716454223644874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644925, "dur": 0,
    "args": {
      "External id": 147581,
      "cbid": 317, "correlation": 147581
    }
  },
  {
    "ph": "f", "id": 147581, "pid": 76337, "tid": -914061504, "ts": 1716454223644925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644926, "dur": 0,
    "args": {
      "External id": 147582,
      "cbid": 203, "correlation": 147582
    }
  },
  {
    "ph": "f", "id": 147582, "pid": 76337, "tid": -914061504, "ts": 1716454223644926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644926, "dur": 0,
    "args": {
      "External id": 147583,
      "cbid": 205, "correlation": 147583
    }
  },
  {
    "ph": "f", "id": 147583, "pid": 76337, "tid": -914061504, "ts": 1716454223644926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223707928, "dur": 58,
    "args": {
      "External id": 147587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147587, "pid": 5, "tid": 7, "ts": 1716454223707928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644944, "dur": 11,
    "args": {
      "External id": 147587,
      "cbid": 211, "correlation": 147587
    }
  },
  {
    "ph": "s", "id": 147587, "pid": 76337, "tid": -914061504, "ts": 1716454223644944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223707987, "dur": 123,
    "args": {
      "External id": 147589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147589,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147589, "pid": 5, "tid": 7, "ts": 1716454223707987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644958, "dur": 6,
    "args": {
      "External id": 147589,
      "cbid": 211, "correlation": 147589
    }
  },
  {
    "ph": "s", "id": 147589, "pid": 76337, "tid": -914061504, "ts": 1716454223644958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223708111, "dur": 1902,
    "args": {
      "External id": 147591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147591,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147591, "pid": 5, "tid": 7, "ts": 1716454223708111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644970, "dur": 14,
    "args": {
      "External id": 147591,
      "cbid": 211, "correlation": 147591
    }
  },
  {
    "ph": "s", "id": 147591, "pid": 76337, "tid": -914061504, "ts": 1716454223644970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223710015, "dur": 20,
    "args": {
      "External id": 147593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147593,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147593, "pid": 5, "tid": 7, "ts": 1716454223710015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223644988, "dur": 5,
    "args": {
      "External id": 147593,
      "cbid": 211, "correlation": 147593
    }
  },
  {
    "ph": "s", "id": 147593, "pid": 76337, "tid": -914061504, "ts": 1716454223644988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223710036, "dur": 33,
    "args": {
      "External id": 147599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147599, "pid": 5, "tid": 7, "ts": 1716454223710036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645017, "dur": 8,
    "args": {
      "External id": 147599,
      "cbid": 211, "correlation": 147599
    }
  },
  {
    "ph": "s", "id": 147599, "pid": 76337, "tid": -914061504, "ts": 1716454223645017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223710071, "dur": 4,
    "args": {
      "External id": 147607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147607,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 147607, "pid": 5, "tid": 7, "ts": 1716454223710071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645062, "dur": 9,
    "args": {
      "External id": 147607,
      "cbid": 211, "correlation": 147607
    }
  },
  {
    "ph": "s", "id": 147607, "pid": 76337, "tid": -914061504, "ts": 1716454223645062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645125, "dur": 1,
    "args": {
      "External id": 147623,
      "cbid": 251, "correlation": 147623
    }
  },
  {
    "ph": "f", "id": 147623, "pid": 76337, "tid": -914061504, "ts": 1716454223645125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645131, "dur": 0,
    "args": {
      "External id": 147625,
      "cbid": 251, "correlation": 147625
    }
  },
  {
    "ph": "f", "id": 147625, "pid": 76337, "tid": -914061504, "ts": 1716454223645131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223710076, "dur": 12,
    "args": {
      "External id": 147626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147626,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 147626, "pid": 5, "tid": 7, "ts": 1716454223710076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645133, "dur": 11,
    "args": {
      "External id": 147626,
      "cbid": 211, "correlation": 147626
    }
  },
  {
    "ph": "s", "id": 147626, "pid": 76337, "tid": -914061504, "ts": 1716454223645133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223710090, "dur": 5,
    "args": {
      "External id": 147628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147628,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 147628, "pid": 5, "tid": 7, "ts": 1716454223710090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645146, "dur": 6,
    "args": {
      "External id": 147628,
      "cbid": 211, "correlation": 147628
    }
  },
  {
    "ph": "s", "id": 147628, "pid": 76337, "tid": -914061504, "ts": 1716454223645146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223710096, "dur": 29,
    "args": {
      "External id": 147638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147638, "pid": 5, "tid": 7, "ts": 1716454223710096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645205, "dur": 13,
    "args": {
      "External id": 147638,
      "cbid": 211, "correlation": 147638
    }
  },
  {
    "ph": "s", "id": 147638, "pid": 76337, "tid": -914061504, "ts": 1716454223645205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223710127, "dur": 31,
    "args": {
      "External id": 147658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147658,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 147658, "pid": 5, "tid": 7, "ts": 1716454223710127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645271, "dur": 10,
    "args": {
      "External id": 147658,
      "cbid": 211, "correlation": 147658
    }
  },
  {
    "ph": "s", "id": 147658, "pid": 76337, "tid": -914061504, "ts": 1716454223645271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223710159, "dur": 4,
    "args": {
      "External id": 147670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147670,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 147670, "pid": 5, "tid": 7, "ts": 1716454223710159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645292, "dur": 6,
    "args": {
      "External id": 147670,
      "cbid": 211, "correlation": 147670
    }
  },
  {
    "ph": "s", "id": 147670, "pid": 76337, "tid": -914061504, "ts": 1716454223645292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223710165, "dur": 30,
    "args": {
      "External id": 147673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147673,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147673, "pid": 5, "tid": 7, "ts": 1716454223710165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645311, "dur": 6,
    "args": {
      "External id": 147673,
      "cbid": 211, "correlation": 147673
    }
  },
  {
    "ph": "s", "id": 147673, "pid": 76337, "tid": -914061504, "ts": 1716454223645311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223710197, "dur": 21,
    "args": {
      "External id": 147682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147682,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147682, "pid": 5, "tid": 7, "ts": 1716454223710197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645350, "dur": 10,
    "args": {
      "External id": 147682,
      "cbid": 211, "correlation": 147682
    }
  },
  {
    "ph": "s", "id": 147682, "pid": 76337, "tid": -914061504, "ts": 1716454223645350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645412, "dur": 0,
    "args": {
      "External id": 147692,
      "cbid": 317, "correlation": 147692
    }
  },
  {
    "ph": "f", "id": 147692, "pid": 76337, "tid": -914061504, "ts": 1716454223645412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645413, "dur": 0,
    "args": {
      "External id": 147693,
      "cbid": 203, "correlation": 147693
    }
  },
  {
    "ph": "f", "id": 147693, "pid": 76337, "tid": -914061504, "ts": 1716454223645413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645414, "dur": 0,
    "args": {
      "External id": 147694,
      "cbid": 205, "correlation": 147694
    }
  },
  {
    "ph": "f", "id": 147694, "pid": 76337, "tid": -914061504, "ts": 1716454223645414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223710219, "dur": 22,
    "args": {
      "External id": 147698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147698,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147698, "pid": 5, "tid": 7, "ts": 1716454223710219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645431, "dur": 12,
    "args": {
      "External id": 147698,
      "cbid": 211, "correlation": 147698
    }
  },
  {
    "ph": "s", "id": 147698, "pid": 76337, "tid": -914061504, "ts": 1716454223645431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223710242, "dur": 45,
    "args": {
      "External id": 147700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147700, "pid": 5, "tid": 7, "ts": 1716454223710242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645445, "dur": 5,
    "args": {
      "External id": 147700,
      "cbid": 211, "correlation": 147700
    }
  },
  {
    "ph": "s", "id": 147700, "pid": 76337, "tid": -914061504, "ts": 1716454223645445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223710288, "dur": 648,
    "args": {
      "External id": 147702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147702,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147702, "pid": 5, "tid": 7, "ts": 1716454223710288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645457, "dur": 8,
    "args": {
      "External id": 147702,
      "cbid": 211, "correlation": 147702
    }
  },
  {
    "ph": "s", "id": 147702, "pid": 76337, "tid": -914061504, "ts": 1716454223645457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223710938, "dur": 23,
    "args": {
      "External id": 147704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147704, "pid": 5, "tid": 7, "ts": 1716454223710938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645468, "dur": 5,
    "args": {
      "External id": 147704,
      "cbid": 211, "correlation": 147704
    }
  },
  {
    "ph": "s", "id": 147704, "pid": 76337, "tid": -914061504, "ts": 1716454223645468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223710962, "dur": 33,
    "args": {
      "External id": 147710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147710, "pid": 5, "tid": 7, "ts": 1716454223710962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645495, "dur": 8,
    "args": {
      "External id": 147710,
      "cbid": 211, "correlation": 147710
    }
  },
  {
    "ph": "s", "id": 147710, "pid": 76337, "tid": -914061504, "ts": 1716454223645495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645552, "dur": 0,
    "args": {
      "External id": 147720,
      "cbid": 317, "correlation": 147720
    }
  },
  {
    "ph": "f", "id": 147720, "pid": 76337, "tid": -914061504, "ts": 1716454223645552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645553, "dur": 0,
    "args": {
      "External id": 147721,
      "cbid": 203, "correlation": 147721
    }
  },
  {
    "ph": "f", "id": 147721, "pid": 76337, "tid": -914061504, "ts": 1716454223645553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645554, "dur": 0,
    "args": {
      "External id": 147722,
      "cbid": 205, "correlation": 147722
    }
  },
  {
    "ph": "f", "id": 147722, "pid": 76337, "tid": -914061504, "ts": 1716454223645554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223710996, "dur": 56,
    "args": {
      "External id": 147726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147726, "pid": 5, "tid": 7, "ts": 1716454223710996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645567, "dur": 11,
    "args": {
      "External id": 147726,
      "cbid": 211, "correlation": 147726
    }
  },
  {
    "ph": "s", "id": 147726, "pid": 76337, "tid": -914061504, "ts": 1716454223645567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223711054, "dur": 271,
    "args": {
      "External id": 147728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147728,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147728, "pid": 5, "tid": 7, "ts": 1716454223711054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645585, "dur": 8,
    "args": {
      "External id": 147728,
      "cbid": 211, "correlation": 147728
    }
  },
  {
    "ph": "s", "id": 147728, "pid": 76337, "tid": -914061504, "ts": 1716454223645585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223711326, "dur": 21,
    "args": {
      "External id": 147730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147730,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147730, "pid": 5, "tid": 7, "ts": 1716454223711326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645597, "dur": 5,
    "args": {
      "External id": 147730,
      "cbid": 211, "correlation": 147730
    }
  },
  {
    "ph": "s", "id": 147730, "pid": 76337, "tid": -914061504, "ts": 1716454223645597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223711349, "dur": 33,
    "args": {
      "External id": 147736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147736,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147736, "pid": 5, "tid": 7, "ts": 1716454223711349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645622, "dur": 8,
    "args": {
      "External id": 147736,
      "cbid": 211, "correlation": 147736
    }
  },
  {
    "ph": "s", "id": 147736, "pid": 76337, "tid": -914061504, "ts": 1716454223645622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223711383, "dur": 27,
    "args": {
      "External id": 147744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147744,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147744, "pid": 5, "tid": 7, "ts": 1716454223711383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645650, "dur": 8,
    "args": {
      "External id": 147744,
      "cbid": 211, "correlation": 147744
    }
  },
  {
    "ph": "s", "id": 147744, "pid": 76337, "tid": -914061504, "ts": 1716454223645650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223711411, "dur": 19,
    "args": {
      "External id": 147752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147752,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147752, "pid": 5, "tid": 7, "ts": 1716454223711411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645679, "dur": 9,
    "args": {
      "External id": 147752,
      "cbid": 211, "correlation": 147752
    }
  },
  {
    "ph": "s", "id": 147752, "pid": 76337, "tid": -914061504, "ts": 1716454223645679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223711432, "dur": 30,
    "args": {
      "External id": 147772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147772,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 147772, "pid": 5, "tid": 7, "ts": 1716454223711432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645761, "dur": 12,
    "args": {
      "External id": 147772,
      "cbid": 211, "correlation": 147772
    }
  },
  {
    "ph": "s", "id": 147772, "pid": 76337, "tid": -914061504, "ts": 1716454223645761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223711464, "dur": 4,
    "args": {
      "External id": 147784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147784,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 147784, "pid": 5, "tid": 7, "ts": 1716454223711464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645782, "dur": 6,
    "args": {
      "External id": 147784,
      "cbid": 211, "correlation": 147784
    }
  },
  {
    "ph": "s", "id": 147784, "pid": 76337, "tid": -914061504, "ts": 1716454223645782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223711469, "dur": 31,
    "args": {
      "External id": 147787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147787,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147787, "pid": 5, "tid": 7, "ts": 1716454223711469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645800, "dur": 6,
    "args": {
      "External id": 147787,
      "cbid": 211, "correlation": 147787
    }
  },
  {
    "ph": "s", "id": 147787, "pid": 76337, "tid": -914061504, "ts": 1716454223645800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645855, "dur": 0,
    "args": {
      "External id": 147798,
      "cbid": 317, "correlation": 147798
    }
  },
  {
    "ph": "f", "id": 147798, "pid": 76337, "tid": -914061504, "ts": 1716454223645855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645856, "dur": 0,
    "args": {
      "External id": 147799,
      "cbid": 203, "correlation": 147799
    }
  },
  {
    "ph": "f", "id": 147799, "pid": 76337, "tid": -914061504, "ts": 1716454223645856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645857, "dur": 0,
    "args": {
      "External id": 147800,
      "cbid": 205, "correlation": 147800
    }
  },
  {
    "ph": "f", "id": 147800, "pid": 76337, "tid": -914061504, "ts": 1716454223645857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223711502, "dur": 22,
    "args": {
      "External id": 147804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147804,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147804, "pid": 5, "tid": 7, "ts": 1716454223711502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645873, "dur": 11,
    "args": {
      "External id": 147804,
      "cbid": 211, "correlation": 147804
    }
  },
  {
    "ph": "s", "id": 147804, "pid": 76337, "tid": -914061504, "ts": 1716454223645873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223711525, "dur": 106,
    "args": {
      "External id": 147806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147806,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147806, "pid": 5, "tid": 7, "ts": 1716454223711525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645890, "dur": 6,
    "args": {
      "External id": 147806,
      "cbid": 211, "correlation": 147806
    }
  },
  {
    "ph": "s", "id": 147806, "pid": 76337, "tid": -914061504, "ts": 1716454223645890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223711632, "dur": 23,
    "args": {
      "External id": 147808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147808,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147808, "pid": 5, "tid": 7, "ts": 1716454223711632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645900, "dur": 5,
    "args": {
      "External id": 147808,
      "cbid": 211, "correlation": 147808
    }
  },
  {
    "ph": "s", "id": 147808, "pid": 76337, "tid": -914061504, "ts": 1716454223645900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223711656, "dur": 32,
    "args": {
      "External id": 147814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147814,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147814, "pid": 5, "tid": 7, "ts": 1716454223711656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223645927, "dur": 8,
    "args": {
      "External id": 147814,
      "cbid": 211, "correlation": 147814
    }
  },
  {
    "ph": "s", "id": 147814, "pid": 76337, "tid": -914061504, "ts": 1716454223645927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223711690, "dur": 195,
    "args": {
      "External id": 147823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147823,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147823, "pid": 5, "tid": 7, "ts": 1716454223711690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646020, "dur": 15,
    "args": {
      "External id": 147823,
      "cbid": 211, "correlation": 147823
    }
  },
  {
    "ph": "s", "id": 147823, "pid": 76337, "tid": -914061504, "ts": 1716454223646020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223711886, "dur": 64,
    "args": {
      "External id": 147845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147845,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147845, "pid": 5, "tid": 7, "ts": 1716454223711886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646079, "dur": 10,
    "args": {
      "External id": 147845,
      "cbid": 211, "correlation": 147845
    }
  },
  {
    "ph": "s", "id": 147845, "pid": 76337, "tid": -914061504, "ts": 1716454223646079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646170, "dur": 1,
    "args": {
      "External id": 147856,
      "cbid": 251, "correlation": 147856
    }
  },
  {
    "ph": "f", "id": 147856, "pid": 76337, "tid": -914061504, "ts": 1716454223646170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223711951, "dur": 155,
    "args": {
      "External id": 147857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147857,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147857, "pid": 5, "tid": 7, "ts": 1716454223711951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646176, "dur": 13,
    "args": {
      "External id": 147857,
      "cbid": 211, "correlation": 147857
    }
  },
  {
    "ph": "s", "id": 147857, "pid": 76337, "tid": -914061504, "ts": 1716454223646176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646246, "dur": 1,
    "args": {
      "External id": 147868,
      "cbid": 251, "correlation": 147868
    }
  },
  {
    "ph": "f", "id": 147868, "pid": 76337, "tid": -914061504, "ts": 1716454223646246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223712108, "dur": 148,
    "args": {
      "External id": 147869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147869,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147869, "pid": 5, "tid": 7, "ts": 1716454223712108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646250, "dur": 12,
    "args": {
      "External id": 147869,
      "cbid": 211, "correlation": 147869
    }
  },
  {
    "ph": "s", "id": 147869, "pid": 76337, "tid": -914061504, "ts": 1716454223646250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646316, "dur": 1,
    "args": {
      "External id": 147880,
      "cbid": 251, "correlation": 147880
    }
  },
  {
    "ph": "f", "id": 147880, "pid": 76337, "tid": -914061504, "ts": 1716454223646316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223712258, "dur": 144,
    "args": {
      "External id": 147881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147881,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147881, "pid": 5, "tid": 7, "ts": 1716454223712258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646320, "dur": 11,
    "args": {
      "External id": 147881,
      "cbid": 211, "correlation": 147881
    }
  },
  {
    "ph": "s", "id": 147881, "pid": 76337, "tid": -914061504, "ts": 1716454223646320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223712403, "dur": 1955,
    "args": {
      "External id": 147902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147902,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 147902, "pid": 5, "tid": 7, "ts": 1716454223712403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646402, "dur": 14,
    "args": {
      "External id": 147902,
      "cbid": 211, "correlation": 147902
    }
  },
  {
    "ph": "s", "id": 147902, "pid": 76337, "tid": -914061504, "ts": 1716454223646402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646502, "dur": 2,
    "args": {
      "External id": 147920,
      "cbid": 251, "correlation": 147920
    }
  },
  {
    "ph": "f", "id": 147920, "pid": 76337, "tid": -914061504, "ts": 1716454223646502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223714360, "dur": 149,
    "args": {
      "External id": 147922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147922,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 147922, "pid": 5, "tid": 7, "ts": 1716454223714360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646510, "dur": 14,
    "args": {
      "External id": 147922,
      "cbid": 211, "correlation": 147922
    }
  },
  {
    "ph": "s", "id": 147922, "pid": 76337, "tid": -914061504, "ts": 1716454223646510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223714510, "dur": 36,
    "args": {
      "External id": 147930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147930,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147930, "pid": 5, "tid": 7, "ts": 1716454223714510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646582, "dur": 12,
    "args": {
      "External id": 147930,
      "cbid": 211, "correlation": 147930
    }
  },
  {
    "ph": "s", "id": 147930, "pid": 76337, "tid": -914061504, "ts": 1716454223646582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223714547, "dur": 51,
    "args": {
      "External id": 147938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147938,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147938, "pid": 5, "tid": 7, "ts": 1716454223714547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646621, "dur": 8,
    "args": {
      "External id": 147938,
      "cbid": 211, "correlation": 147938
    }
  },
  {
    "ph": "s", "id": 147938, "pid": 76337, "tid": -914061504, "ts": 1716454223646621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223714599, "dur": 30,
    "args": {
      "External id": 147949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147949,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147949, "pid": 5, "tid": 7, "ts": 1716454223714599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646692, "dur": 12,
    "args": {
      "External id": 147949,
      "cbid": 211, "correlation": 147949
    }
  },
  {
    "ph": "s", "id": 147949, "pid": 76337, "tid": -914061504, "ts": 1716454223646692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223714630, "dur": 35,
    "args": {
      "External id": 147971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147971,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 147971, "pid": 5, "tid": 7, "ts": 1716454223714630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646725, "dur": 7,
    "args": {
      "External id": 147971,
      "cbid": 211, "correlation": 147971
    }
  },
  {
    "ph": "s", "id": 147971, "pid": 76337, "tid": -914061504, "ts": 1716454223646725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646809, "dur": 1,
    "args": {
      "External id": 147982,
      "cbid": 251, "correlation": 147982
    }
  },
  {
    "ph": "f", "id": 147982, "pid": 76337, "tid": -914061504, "ts": 1716454223646809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223714666, "dur": 90,
    "args": {
      "External id": 147983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147983,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 147983, "pid": 5, "tid": 7, "ts": 1716454223714666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646815, "dur": 13,
    "args": {
      "External id": 147983,
      "cbid": 211, "correlation": 147983
    }
  },
  {
    "ph": "s", "id": 147983, "pid": 76337, "tid": -914061504, "ts": 1716454223646815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646884, "dur": 1,
    "args": {
      "External id": 147994,
      "cbid": 251, "correlation": 147994
    }
  },
  {
    "ph": "f", "id": 147994, "pid": 76337, "tid": -914061504, "ts": 1716454223646884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646888, "dur": 0,
    "args": {
      "External id": 147995,
      "cbid": 251, "correlation": 147995
    }
  },
  {
    "ph": "f", "id": 147995, "pid": 76337, "tid": -914061504, "ts": 1716454223646888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223714758, "dur": 11,
    "args": {
      "External id": 147996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147996,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 147996, "pid": 5, "tid": 7, "ts": 1716454223714758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646889, "dur": 12,
    "args": {
      "External id": 147996,
      "cbid": 211, "correlation": 147996
    }
  },
  {
    "ph": "s", "id": 147996, "pid": 76337, "tid": -914061504, "ts": 1716454223646889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223714770, "dur": 5,
    "args": {
      "External id": 147998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 147998,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 147998, "pid": 5, "tid": 7, "ts": 1716454223714770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646902, "dur": 5,
    "args": {
      "External id": 147998,
      "cbid": 211, "correlation": 147998
    }
  },
  {
    "ph": "s", "id": 147998, "pid": 76337, "tid": -914061504, "ts": 1716454223646902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646958, "dur": 1,
    "args": {
      "External id": 148009,
      "cbid": 251, "correlation": 148009
    }
  },
  {
    "ph": "f", "id": 148009, "pid": 76337, "tid": -914061504, "ts": 1716454223646958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646962, "dur": 0,
    "args": {
      "External id": 148010,
      "cbid": 251, "correlation": 148010
    }
  },
  {
    "ph": "f", "id": 148010, "pid": 76337, "tid": -914061504, "ts": 1716454223646962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223714776, "dur": 8,
    "args": {
      "External id": 148011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148011,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 148011, "pid": 5, "tid": 7, "ts": 1716454223714776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646963, "dur": 24,
    "args": {
      "External id": 148011,
      "cbid": 211, "correlation": 148011
    }
  },
  {
    "ph": "s", "id": 148011, "pid": 76337, "tid": -914061504, "ts": 1716454223646963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223714785, "dur": 3,
    "args": {
      "External id": 148013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148013,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 148013, "pid": 5, "tid": 7, "ts": 1716454223714785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223646989, "dur": 6,
    "args": {
      "External id": 148013,
      "cbid": 211, "correlation": 148013
    }
  },
  {
    "ph": "s", "id": 148013, "pid": 76337, "tid": -914061504, "ts": 1716454223646989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223714790, "dur": 93,
    "args": {
      "External id": 148034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148034,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 148034, "pid": 5, "tid": 7, "ts": 1716454223714790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647064, "dur": 13,
    "args": {
      "External id": 148034,
      "cbid": 211, "correlation": 148034
    }
  },
  {
    "ph": "s", "id": 148034, "pid": 76337, "tid": -914061504, "ts": 1716454223647064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647161, "dur": 1,
    "args": {
      "External id": 148052,
      "cbid": 251, "correlation": 148052
    }
  },
  {
    "ph": "f", "id": 148052, "pid": 76337, "tid": -914061504, "ts": 1716454223647161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223714884, "dur": 98,
    "args": {
      "External id": 148054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148054,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148054, "pid": 5, "tid": 7, "ts": 1716454223714884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647167, "dur": 14,
    "args": {
      "External id": 148054,
      "cbid": 211, "correlation": 148054
    }
  },
  {
    "ph": "s", "id": 148054, "pid": 76337, "tid": -914061504, "ts": 1716454223647167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223714984, "dur": 19,
    "args": {
      "External id": 148062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148062,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148062, "pid": 5, "tid": 7, "ts": 1716454223714984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647237, "dur": 12,
    "args": {
      "External id": 148062,
      "cbid": 211, "correlation": 148062
    }
  },
  {
    "ph": "s", "id": 148062, "pid": 76337, "tid": -914061504, "ts": 1716454223647237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223715004, "dur": 37,
    "args": {
      "External id": 148070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148070,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148070, "pid": 5, "tid": 7, "ts": 1716454223715004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647278, "dur": 9,
    "args": {
      "External id": 148070,
      "cbid": 211, "correlation": 148070
    }
  },
  {
    "ph": "s", "id": 148070, "pid": 76337, "tid": -914061504, "ts": 1716454223647278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223715042, "dur": 35,
    "args": {
      "External id": 148092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148092,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148092, "pid": 5, "tid": 7, "ts": 1716454223715042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647330, "dur": 10,
    "args": {
      "External id": 148092,
      "cbid": 211, "correlation": 148092
    }
  },
  {
    "ph": "s", "id": 148092, "pid": 76337, "tid": -914061504, "ts": 1716454223647330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647418, "dur": 1,
    "args": {
      "External id": 148108,
      "cbid": 251, "correlation": 148108
    }
  },
  {
    "ph": "f", "id": 148108, "pid": 76337, "tid": -914061504, "ts": 1716454223647418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647424, "dur": 0,
    "args": {
      "External id": 148110,
      "cbid": 251, "correlation": 148110
    }
  },
  {
    "ph": "f", "id": 148110, "pid": 76337, "tid": -914061504, "ts": 1716454223647424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223715078, "dur": 543,
    "args": {
      "External id": 148111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148111,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 148111, "pid": 5, "tid": 7, "ts": 1716454223715078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647429, "dur": 12,
    "args": {
      "External id": 148111,
      "cbid": 211, "correlation": 148111
    }
  },
  {
    "ph": "s", "id": 148111, "pid": 76337, "tid": -914061504, "ts": 1716454223647429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223715623, "dur": 127,
    "args": {
      "External id": 148119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148119,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148119, "pid": 5, "tid": 7, "ts": 1716454223715623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647493, "dur": 12,
    "args": {
      "External id": 148119,
      "cbid": 211, "correlation": 148119
    }
  },
  {
    "ph": "s", "id": 148119, "pid": 76337, "tid": -914061504, "ts": 1716454223647493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223715751, "dur": 129,
    "args": {
      "External id": 148127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148127, "pid": 5, "tid": 7, "ts": 1716454223715751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647526, "dur": 8,
    "args": {
      "External id": 148127,
      "cbid": 211, "correlation": 148127
    }
  },
  {
    "ph": "s", "id": 148127, "pid": 76337, "tid": -914061504, "ts": 1716454223647526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647603, "dur": 1,
    "args": {
      "External id": 148143,
      "cbid": 251, "correlation": 148143
    }
  },
  {
    "ph": "f", "id": 148143, "pid": 76337, "tid": -914061504, "ts": 1716454223647603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223715882, "dur": 314,
    "args": {
      "External id": 148145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148145,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148145, "pid": 5, "tid": 7, "ts": 1716454223715882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647608, "dur": 13,
    "args": {
      "External id": 148145,
      "cbid": 211, "correlation": 148145
    }
  },
  {
    "ph": "s", "id": 148145, "pid": 76337, "tid": -914061504, "ts": 1716454223647608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223716197, "dur": 27,
    "args": {
      "External id": 148153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148153,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148153, "pid": 5, "tid": 7, "ts": 1716454223716197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647650, "dur": 10,
    "args": {
      "External id": 148153,
      "cbid": 211, "correlation": 148153
    }
  },
  {
    "ph": "s", "id": 148153, "pid": 76337, "tid": -914061504, "ts": 1716454223647650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223716226, "dur": 81,
    "args": {
      "External id": 148164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148164,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148164, "pid": 5, "tid": 7, "ts": 1716454223716226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647718, "dur": 13,
    "args": {
      "External id": 148164,
      "cbid": 211, "correlation": 148164
    }
  },
  {
    "ph": "s", "id": 148164, "pid": 76337, "tid": -914061504, "ts": 1716454223647718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647782, "dur": 0,
    "args": {
      "External id": 148176,
      "cbid": 317, "correlation": 148176
    }
  },
  {
    "ph": "f", "id": 148176, "pid": 76337, "tid": -914061504, "ts": 1716454223647782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647783, "dur": 0,
    "args": {
      "External id": 148177,
      "cbid": 203, "correlation": 148177
    }
  },
  {
    "ph": "f", "id": 148177, "pid": 76337, "tid": -914061504, "ts": 1716454223647783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647784, "dur": 0,
    "args": {
      "External id": 148178,
      "cbid": 205, "correlation": 148178
    }
  },
  {
    "ph": "f", "id": 148178, "pid": 76337, "tid": -914061504, "ts": 1716454223647784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223716308, "dur": 23,
    "args": {
      "External id": 148182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148182,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148182, "pid": 5, "tid": 7, "ts": 1716454223716308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647799, "dur": 12,
    "args": {
      "External id": 148182,
      "cbid": 211, "correlation": 148182
    }
  },
  {
    "ph": "s", "id": 148182, "pid": 76337, "tid": -914061504, "ts": 1716454223647799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223716332, "dur": 120,
    "args": {
      "External id": 148184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148184,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148184, "pid": 5, "tid": 7, "ts": 1716454223716332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647817, "dur": 7,
    "args": {
      "External id": 148184,
      "cbid": 211, "correlation": 148184
    }
  },
  {
    "ph": "s", "id": 148184, "pid": 76337, "tid": -914061504, "ts": 1716454223647817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223716454, "dur": 22,
    "args": {
      "External id": 148186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148186, "pid": 5, "tid": 7, "ts": 1716454223716454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647828, "dur": 5,
    "args": {
      "External id": 148186,
      "cbid": 211, "correlation": 148186
    }
  },
  {
    "ph": "s", "id": 148186, "pid": 76337, "tid": -914061504, "ts": 1716454223647828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223716477, "dur": 33,
    "args": {
      "External id": 148192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148192,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148192, "pid": 5, "tid": 7, "ts": 1716454223716477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647855, "dur": 8,
    "args": {
      "External id": 148192,
      "cbid": 211, "correlation": 148192
    }
  },
  {
    "ph": "s", "id": 148192, "pid": 76337, "tid": -914061504, "ts": 1716454223647855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223716512, "dur": 26,
    "args": {
      "External id": 148200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148200,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148200, "pid": 5, "tid": 7, "ts": 1716454223716512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647887, "dur": 8,
    "args": {
      "External id": 148200,
      "cbid": 211, "correlation": 148200
    }
  },
  {
    "ph": "s", "id": 148200, "pid": 76337, "tid": -914061504, "ts": 1716454223647887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223716540, "dur": 53,
    "args": {
      "External id": 148209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148209,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148209, "pid": 5, "tid": 7, "ts": 1716454223716540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223647925, "dur": 11,
    "args": {
      "External id": 148209,
      "cbid": 211, "correlation": 148209
    }
  },
  {
    "ph": "s", "id": 148209, "pid": 76337, "tid": -914061504, "ts": 1716454223647925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223716594, "dur": 51,
    "args": {
      "External id": 148229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148229,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 148229, "pid": 5, "tid": 7, "ts": 1716454223716594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648006, "dur": 12,
    "args": {
      "External id": 148229,
      "cbid": 211, "correlation": 148229
    }
  },
  {
    "ph": "s", "id": 148229, "pid": 76337, "tid": -914061504, "ts": 1716454223648006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223716646, "dur": 5,
    "args": {
      "External id": 148241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148241,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 148241, "pid": 5, "tid": 7, "ts": 1716454223716646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648029, "dur": 6,
    "args": {
      "External id": 148241,
      "cbid": 211, "correlation": 148241
    }
  },
  {
    "ph": "s", "id": 148241, "pid": 76337, "tid": -914061504, "ts": 1716454223648029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223716652, "dur": 56,
    "args": {
      "External id": 148244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148244, "pid": 5, "tid": 7, "ts": 1716454223716652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648047, "dur": 6,
    "args": {
      "External id": 148244,
      "cbid": 211, "correlation": 148244
    }
  },
  {
    "ph": "s", "id": 148244, "pid": 76337, "tid": -914061504, "ts": 1716454223648047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223716709, "dur": 37,
    "args": {
      "External id": 148253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148253,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148253, "pid": 5, "tid": 7, "ts": 1716454223716709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648087, "dur": 9,
    "args": {
      "External id": 148253,
      "cbid": 211, "correlation": 148253
    }
  },
  {
    "ph": "s", "id": 148253, "pid": 76337, "tid": -914061504, "ts": 1716454223648087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648138, "dur": 0,
    "args": {
      "External id": 148263,
      "cbid": 317, "correlation": 148263
    }
  },
  {
    "ph": "f", "id": 148263, "pid": 76337, "tid": -914061504, "ts": 1716454223648138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648138, "dur": 0,
    "args": {
      "External id": 148264,
      "cbid": 203, "correlation": 148264
    }
  },
  {
    "ph": "f", "id": 148264, "pid": 76337, "tid": -914061504, "ts": 1716454223648138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648139, "dur": 0,
    "args": {
      "External id": 148265,
      "cbid": 205, "correlation": 148265
    }
  },
  {
    "ph": "f", "id": 148265, "pid": 76337, "tid": -914061504, "ts": 1716454223648139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223716748, "dur": 40,
    "args": {
      "External id": 148269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148269,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148269, "pid": 5, "tid": 7, "ts": 1716454223716748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648154, "dur": 11,
    "args": {
      "External id": 148269,
      "cbid": 211, "correlation": 148269
    }
  },
  {
    "ph": "s", "id": 148269, "pid": 76337, "tid": -914061504, "ts": 1716454223648154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223716789, "dur": 84,
    "args": {
      "External id": 148271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148271, "pid": 5, "tid": 7, "ts": 1716454223716789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648168, "dur": 5,
    "args": {
      "External id": 148271,
      "cbid": 211, "correlation": 148271
    }
  },
  {
    "ph": "s", "id": 148271, "pid": 76337, "tid": -914061504, "ts": 1716454223648168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223716874, "dur": 1282,
    "args": {
      "External id": 148273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148273,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148273, "pid": 5, "tid": 7, "ts": 1716454223716874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648179, "dur": 6,
    "args": {
      "External id": 148273,
      "cbid": 211, "correlation": 148273
    }
  },
  {
    "ph": "s", "id": 148273, "pid": 76337, "tid": -914061504, "ts": 1716454223648179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223718157, "dur": 20,
    "args": {
      "External id": 148275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148275,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148275, "pid": 5, "tid": 7, "ts": 1716454223718157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648189, "dur": 6,
    "args": {
      "External id": 148275,
      "cbid": 211, "correlation": 148275
    }
  },
  {
    "ph": "s", "id": 148275, "pid": 76337, "tid": -914061504, "ts": 1716454223648189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223718179, "dur": 33,
    "args": {
      "External id": 148281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148281, "pid": 5, "tid": 7, "ts": 1716454223718179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648218, "dur": 8,
    "args": {
      "External id": 148281,
      "cbid": 211, "correlation": 148281
    }
  },
  {
    "ph": "s", "id": 148281, "pid": 76337, "tid": -914061504, "ts": 1716454223648218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223718213, "dur": 4,
    "args": {
      "External id": 148289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148289,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 148289, "pid": 5, "tid": 7, "ts": 1716454223718213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648260, "dur": 9,
    "args": {
      "External id": 148289,
      "cbid": 211, "correlation": 148289
    }
  },
  {
    "ph": "s", "id": 148289, "pid": 76337, "tid": -914061504, "ts": 1716454223648260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648323, "dur": 1,
    "args": {
      "External id": 148305,
      "cbid": 251, "correlation": 148305
    }
  },
  {
    "ph": "f", "id": 148305, "pid": 76337, "tid": -914061504, "ts": 1716454223648323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648329, "dur": 0,
    "args": {
      "External id": 148307,
      "cbid": 251, "correlation": 148307
    }
  },
  {
    "ph": "f", "id": 148307, "pid": 76337, "tid": -914061504, "ts": 1716454223648329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223718218, "dur": 12,
    "args": {
      "External id": 148308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148308,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 148308, "pid": 5, "tid": 7, "ts": 1716454223718218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648331, "dur": 11,
    "args": {
      "External id": 148308,
      "cbid": 211, "correlation": 148308
    }
  },
  {
    "ph": "s", "id": 148308, "pid": 76337, "tid": -914061504, "ts": 1716454223648331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223718232, "dur": 5,
    "args": {
      "External id": 148310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148310,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 148310, "pid": 5, "tid": 7, "ts": 1716454223718232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648344, "dur": 5,
    "args": {
      "External id": 148310,
      "cbid": 211, "correlation": 148310
    }
  },
  {
    "ph": "s", "id": 148310, "pid": 76337, "tid": -914061504, "ts": 1716454223648344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223718238, "dur": 30,
    "args": {
      "External id": 148320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148320, "pid": 5, "tid": 7, "ts": 1716454223718238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648402, "dur": 12,
    "args": {
      "External id": 148320,
      "cbid": 211, "correlation": 148320
    }
  },
  {
    "ph": "s", "id": 148320, "pid": 76337, "tid": -914061504, "ts": 1716454223648402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223718269, "dur": 31,
    "args": {
      "External id": 148340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148340,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 148340, "pid": 5, "tid": 7, "ts": 1716454223718269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648468, "dur": 12,
    "args": {
      "External id": 148340,
      "cbid": 211, "correlation": 148340
    }
  },
  {
    "ph": "s", "id": 148340, "pid": 76337, "tid": -914061504, "ts": 1716454223648468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223718302, "dur": 4,
    "args": {
      "External id": 148352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148352,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 148352, "pid": 5, "tid": 7, "ts": 1716454223718302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648489, "dur": 6,
    "args": {
      "External id": 148352,
      "cbid": 211, "correlation": 148352
    }
  },
  {
    "ph": "s", "id": 148352, "pid": 76337, "tid": -914061504, "ts": 1716454223648489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223718307, "dur": 30,
    "args": {
      "External id": 148355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148355, "pid": 5, "tid": 7, "ts": 1716454223718307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648506, "dur": 7,
    "args": {
      "External id": 148355,
      "cbid": 211, "correlation": 148355
    }
  },
  {
    "ph": "s", "id": 148355, "pid": 76337, "tid": -914061504, "ts": 1716454223648506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223718338, "dur": 20,
    "args": {
      "External id": 148364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148364,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148364, "pid": 5, "tid": 7, "ts": 1716454223718338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648548, "dur": 11,
    "args": {
      "External id": 148364,
      "cbid": 211, "correlation": 148364
    }
  },
  {
    "ph": "s", "id": 148364, "pid": 76337, "tid": -914061504, "ts": 1716454223648548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648610, "dur": 0,
    "args": {
      "External id": 148374,
      "cbid": 317, "correlation": 148374
    }
  },
  {
    "ph": "f", "id": 148374, "pid": 76337, "tid": -914061504, "ts": 1716454223648610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648611, "dur": 0,
    "args": {
      "External id": 148375,
      "cbid": 203, "correlation": 148375
    }
  },
  {
    "ph": "f", "id": 148375, "pid": 76337, "tid": -914061504, "ts": 1716454223648611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648612, "dur": 0,
    "args": {
      "External id": 148376,
      "cbid": 205, "correlation": 148376
    }
  },
  {
    "ph": "f", "id": 148376, "pid": 76337, "tid": -914061504, "ts": 1716454223648612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223718360, "dur": 22,
    "args": {
      "External id": 148380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148380,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148380, "pid": 5, "tid": 7, "ts": 1716454223718360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648629, "dur": 12,
    "args": {
      "External id": 148380,
      "cbid": 211, "correlation": 148380
    }
  },
  {
    "ph": "s", "id": 148380, "pid": 76337, "tid": -914061504, "ts": 1716454223648629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223718383, "dur": 44,
    "args": {
      "External id": 148382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148382,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148382, "pid": 5, "tid": 7, "ts": 1716454223718383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648644, "dur": 5,
    "args": {
      "External id": 148382,
      "cbid": 211, "correlation": 148382
    }
  },
  {
    "ph": "s", "id": 148382, "pid": 76337, "tid": -914061504, "ts": 1716454223648644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223718428, "dur": 645,
    "args": {
      "External id": 148384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148384,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148384, "pid": 5, "tid": 7, "ts": 1716454223718428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648654, "dur": 6,
    "args": {
      "External id": 148384,
      "cbid": 211, "correlation": 148384
    }
  },
  {
    "ph": "s", "id": 148384, "pid": 76337, "tid": -914061504, "ts": 1716454223648654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719075, "dur": 22,
    "args": {
      "External id": 148386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148386,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148386, "pid": 5, "tid": 7, "ts": 1716454223719075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648664, "dur": 5,
    "args": {
      "External id": 148386,
      "cbid": 211, "correlation": 148386
    }
  },
  {
    "ph": "s", "id": 148386, "pid": 76337, "tid": -914061504, "ts": 1716454223648664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223719098, "dur": 33,
    "args": {
      "External id": 148392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148392,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148392, "pid": 5, "tid": 7, "ts": 1716454223719098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648692, "dur": 8,
    "args": {
      "External id": 148392,
      "cbid": 211, "correlation": 148392
    }
  },
  {
    "ph": "s", "id": 148392, "pid": 76337, "tid": -914061504, "ts": 1716454223648692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648750, "dur": 0,
    "args": {
      "External id": 148402,
      "cbid": 317, "correlation": 148402
    }
  },
  {
    "ph": "f", "id": 148402, "pid": 76337, "tid": -914061504, "ts": 1716454223648750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648751, "dur": 0,
    "args": {
      "External id": 148403,
      "cbid": 203, "correlation": 148403
    }
  },
  {
    "ph": "f", "id": 148403, "pid": 76337, "tid": -914061504, "ts": 1716454223648751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648752, "dur": 0,
    "args": {
      "External id": 148404,
      "cbid": 205, "correlation": 148404
    }
  },
  {
    "ph": "f", "id": 148404, "pid": 76337, "tid": -914061504, "ts": 1716454223648752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719132, "dur": 38,
    "args": {
      "External id": 148408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148408,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148408, "pid": 5, "tid": 7, "ts": 1716454223719132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648764, "dur": 12,
    "args": {
      "External id": 148408,
      "cbid": 211, "correlation": 148408
    }
  },
  {
    "ph": "s", "id": 148408, "pid": 76337, "tid": -914061504, "ts": 1716454223648764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223719172, "dur": 189,
    "args": {
      "External id": 148410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148410,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148410, "pid": 5, "tid": 7, "ts": 1716454223719172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648782, "dur": 7,
    "args": {
      "External id": 148410,
      "cbid": 211, "correlation": 148410
    }
  },
  {
    "ph": "s", "id": 148410, "pid": 76337, "tid": -914061504, "ts": 1716454223648782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719363, "dur": 22,
    "args": {
      "External id": 148412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148412, "pid": 5, "tid": 7, "ts": 1716454223719363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648793, "dur": 5,
    "args": {
      "External id": 148412,
      "cbid": 211, "correlation": 148412
    }
  },
  {
    "ph": "s", "id": 148412, "pid": 76337, "tid": -914061504, "ts": 1716454223648793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223719386, "dur": 32,
    "args": {
      "External id": 148418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148418, "pid": 5, "tid": 7, "ts": 1716454223719386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648818, "dur": 8,
    "args": {
      "External id": 148418,
      "cbid": 211, "correlation": 148418
    }
  },
  {
    "ph": "s", "id": 148418, "pid": 76337, "tid": -914061504, "ts": 1716454223648818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223719420, "dur": 27,
    "args": {
      "External id": 148426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148426,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148426, "pid": 5, "tid": 7, "ts": 1716454223719420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648847, "dur": 8,
    "args": {
      "External id": 148426,
      "cbid": 211, "correlation": 148426
    }
  },
  {
    "ph": "s", "id": 148426, "pid": 76337, "tid": -914061504, "ts": 1716454223648847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223719449, "dur": 20,
    "args": {
      "External id": 148434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148434,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148434, "pid": 5, "tid": 7, "ts": 1716454223719449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648875, "dur": 9,
    "args": {
      "External id": 148434,
      "cbid": 211, "correlation": 148434
    }
  },
  {
    "ph": "s", "id": 148434, "pid": 76337, "tid": -914061504, "ts": 1716454223648875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719470, "dur": 31,
    "args": {
      "External id": 148454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148454,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 148454, "pid": 5, "tid": 7, "ts": 1716454223719470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648958, "dur": 12,
    "args": {
      "External id": 148454,
      "cbid": 211, "correlation": 148454
    }
  },
  {
    "ph": "s", "id": 148454, "pid": 76337, "tid": -914061504, "ts": 1716454223648958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223719502, "dur": 5,
    "args": {
      "External id": 148466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148466,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 148466, "pid": 5, "tid": 7, "ts": 1716454223719502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223648988, "dur": 7,
    "args": {
      "External id": 148466,
      "cbid": 211, "correlation": 148466
    }
  },
  {
    "ph": "s", "id": 148466, "pid": 76337, "tid": -914061504, "ts": 1716454223648988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223719508, "dur": 30,
    "args": {
      "External id": 148469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148469,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148469, "pid": 5, "tid": 7, "ts": 1716454223719508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649007, "dur": 7,
    "args": {
      "External id": 148469,
      "cbid": 211, "correlation": 148469
    }
  },
  {
    "ph": "s", "id": 148469, "pid": 76337, "tid": -914061504, "ts": 1716454223649007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649065, "dur": 0,
    "args": {
      "External id": 148480,
      "cbid": 317, "correlation": 148480
    }
  },
  {
    "ph": "f", "id": 148480, "pid": 76337, "tid": -914061504, "ts": 1716454223649065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649066, "dur": 0,
    "args": {
      "External id": 148481,
      "cbid": 203, "correlation": 148481
    }
  },
  {
    "ph": "f", "id": 148481, "pid": 76337, "tid": -914061504, "ts": 1716454223649066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649066, "dur": 0,
    "args": {
      "External id": 148482,
      "cbid": 205, "correlation": 148482
    }
  },
  {
    "ph": "f", "id": 148482, "pid": 76337, "tid": -914061504, "ts": 1716454223649066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719539, "dur": 24,
    "args": {
      "External id": 148486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148486,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148486, "pid": 5, "tid": 7, "ts": 1716454223719539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649079, "dur": 12,
    "args": {
      "External id": 148486,
      "cbid": 211, "correlation": 148486
    }
  },
  {
    "ph": "s", "id": 148486, "pid": 76337, "tid": -914061504, "ts": 1716454223649079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223719564, "dur": 104,
    "args": {
      "External id": 148488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148488,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148488, "pid": 5, "tid": 7, "ts": 1716454223719564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649097, "dur": 7,
    "args": {
      "External id": 148488,
      "cbid": 211, "correlation": 148488
    }
  },
  {
    "ph": "s", "id": 148488, "pid": 76337, "tid": -914061504, "ts": 1716454223649097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719670, "dur": 23,
    "args": {
      "External id": 148490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148490, "pid": 5, "tid": 7, "ts": 1716454223719670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649108, "dur": 5,
    "args": {
      "External id": 148490,
      "cbid": 211, "correlation": 148490
    }
  },
  {
    "ph": "s", "id": 148490, "pid": 76337, "tid": -914061504, "ts": 1716454223649108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223719694, "dur": 33,
    "args": {
      "External id": 148496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148496, "pid": 5, "tid": 7, "ts": 1716454223719694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649135, "dur": 8,
    "args": {
      "External id": 148496,
      "cbid": 211, "correlation": 148496
    }
  },
  {
    "ph": "s", "id": 148496, "pid": 76337, "tid": -914061504, "ts": 1716454223649135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223719728, "dur": 179,
    "args": {
      "External id": 148505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148505,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148505, "pid": 5, "tid": 7, "ts": 1716454223719728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649216, "dur": 15,
    "args": {
      "External id": 148505,
      "cbid": 211, "correlation": 148505
    }
  },
  {
    "ph": "s", "id": 148505, "pid": 76337, "tid": -914061504, "ts": 1716454223649216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223719909, "dur": 65,
    "args": {
      "External id": 148527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148527,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148527, "pid": 5, "tid": 7, "ts": 1716454223719909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649274, "dur": 11,
    "args": {
      "External id": 148527,
      "cbid": 211, "correlation": 148527
    }
  },
  {
    "ph": "s", "id": 148527, "pid": 76337, "tid": -914061504, "ts": 1716454223649274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649363, "dur": 2,
    "args": {
      "External id": 148538,
      "cbid": 251, "correlation": 148538
    }
  },
  {
    "ph": "f", "id": 148538, "pid": 76337, "tid": -914061504, "ts": 1716454223649363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223719975, "dur": 156,
    "args": {
      "External id": 148539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148539,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148539, "pid": 5, "tid": 7, "ts": 1716454223719975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649368, "dur": 13,
    "args": {
      "External id": 148539,
      "cbid": 211, "correlation": 148539
    }
  },
  {
    "ph": "s", "id": 148539, "pid": 76337, "tid": -914061504, "ts": 1716454223649368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649438, "dur": 1,
    "args": {
      "External id": 148550,
      "cbid": 251, "correlation": 148550
    }
  },
  {
    "ph": "f", "id": 148550, "pid": 76337, "tid": -914061504, "ts": 1716454223649438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223720132, "dur": 146,
    "args": {
      "External id": 148551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148551,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148551, "pid": 5, "tid": 7, "ts": 1716454223720132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649442, "dur": 11,
    "args": {
      "External id": 148551,
      "cbid": 211, "correlation": 148551
    }
  },
  {
    "ph": "s", "id": 148551, "pid": 76337, "tid": -914061504, "ts": 1716454223649442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649508, "dur": 1,
    "args": {
      "External id": 148562,
      "cbid": 251, "correlation": 148562
    }
  },
  {
    "ph": "f", "id": 148562, "pid": 76337, "tid": -914061504, "ts": 1716454223649508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223720279, "dur": 146,
    "args": {
      "External id": 148563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148563,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148563, "pid": 5, "tid": 7, "ts": 1716454223720279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649512, "dur": 11,
    "args": {
      "External id": 148563,
      "cbid": 211, "correlation": 148563
    }
  },
  {
    "ph": "s", "id": 148563, "pid": 76337, "tid": -914061504, "ts": 1716454223649512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223720426, "dur": 1944,
    "args": {
      "External id": 148584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148584,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 148584, "pid": 5, "tid": 7, "ts": 1716454223720426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649593, "dur": 13,
    "args": {
      "External id": 148584,
      "cbid": 211, "correlation": 148584
    }
  },
  {
    "ph": "s", "id": 148584, "pid": 76337, "tid": -914061504, "ts": 1716454223649593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649697, "dur": 1,
    "args": {
      "External id": 148602,
      "cbid": 251, "correlation": 148602
    }
  },
  {
    "ph": "f", "id": 148602, "pid": 76337, "tid": -914061504, "ts": 1716454223649697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223722371, "dur": 148,
    "args": {
      "External id": 148604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148604,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 148604, "pid": 5, "tid": 7, "ts": 1716454223722371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649703, "dur": 13,
    "args": {
      "External id": 148604,
      "cbid": 211, "correlation": 148604
    }
  },
  {
    "ph": "s", "id": 148604, "pid": 76337, "tid": -914061504, "ts": 1716454223649703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223722521, "dur": 36,
    "args": {
      "External id": 148612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148612,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148612, "pid": 5, "tid": 7, "ts": 1716454223722521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649776, "dur": 13,
    "args": {
      "External id": 148612,
      "cbid": 211, "correlation": 148612
    }
  },
  {
    "ph": "s", "id": 148612, "pid": 76337, "tid": -914061504, "ts": 1716454223649776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223722558, "dur": 51,
    "args": {
      "External id": 148620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148620,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148620, "pid": 5, "tid": 7, "ts": 1716454223722558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649816, "dur": 9,
    "args": {
      "External id": 148620,
      "cbid": 211, "correlation": 148620
    }
  },
  {
    "ph": "s", "id": 148620, "pid": 76337, "tid": -914061504, "ts": 1716454223649816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223722610, "dur": 29,
    "args": {
      "External id": 148631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148631,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148631, "pid": 5, "tid": 7, "ts": 1716454223722610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649896, "dur": 13,
    "args": {
      "External id": 148631,
      "cbid": 211, "correlation": 148631
    }
  },
  {
    "ph": "s", "id": 148631, "pid": 76337, "tid": -914061504, "ts": 1716454223649896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223722641, "dur": 34,
    "args": {
      "External id": 148653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148653,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148653, "pid": 5, "tid": 7, "ts": 1716454223722641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223649928, "dur": 9,
    "args": {
      "External id": 148653,
      "cbid": 211, "correlation": 148653
    }
  },
  {
    "ph": "s", "id": 148653, "pid": 76337, "tid": -914061504, "ts": 1716454223649928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650028, "dur": 1,
    "args": {
      "External id": 148664,
      "cbid": 251, "correlation": 148664
    }
  },
  {
    "ph": "f", "id": 148664, "pid": 76337, "tid": -914061504, "ts": 1716454223650028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223722676, "dur": 89,
    "args": {
      "External id": 148665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148665,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148665, "pid": 5, "tid": 7, "ts": 1716454223722676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650033, "dur": 15,
    "args": {
      "External id": 148665,
      "cbid": 211, "correlation": 148665
    }
  },
  {
    "ph": "s", "id": 148665, "pid": 76337, "tid": -914061504, "ts": 1716454223650033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650109, "dur": 1,
    "args": {
      "External id": 148676,
      "cbid": 251, "correlation": 148676
    }
  },
  {
    "ph": "f", "id": 148676, "pid": 76337, "tid": -914061504, "ts": 1716454223650109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650113, "dur": 0,
    "args": {
      "External id": 148677,
      "cbid": 251, "correlation": 148677
    }
  },
  {
    "ph": "f", "id": 148677, "pid": 76337, "tid": -914061504, "ts": 1716454223650113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223722767, "dur": 11,
    "args": {
      "External id": 148678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148678,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 148678, "pid": 5, "tid": 7, "ts": 1716454223722767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650115, "dur": 12,
    "args": {
      "External id": 148678,
      "cbid": 211, "correlation": 148678
    }
  },
  {
    "ph": "s", "id": 148678, "pid": 76337, "tid": -914061504, "ts": 1716454223650115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223722780, "dur": 5,
    "args": {
      "External id": 148680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148680,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 148680, "pid": 5, "tid": 7, "ts": 1716454223722780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650131, "dur": 7,
    "args": {
      "External id": 148680,
      "cbid": 211, "correlation": 148680
    }
  },
  {
    "ph": "s", "id": 148680, "pid": 76337, "tid": -914061504, "ts": 1716454223650131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650189, "dur": 1,
    "args": {
      "External id": 148691,
      "cbid": 251, "correlation": 148691
    }
  },
  {
    "ph": "f", "id": 148691, "pid": 76337, "tid": -914061504, "ts": 1716454223650189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650192, "dur": 0,
    "args": {
      "External id": 148692,
      "cbid": 251, "correlation": 148692
    }
  },
  {
    "ph": "f", "id": 148692, "pid": 76337, "tid": -914061504, "ts": 1716454223650192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223722786, "dur": 8,
    "args": {
      "External id": 148693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148693,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 148693, "pid": 5, "tid": 7, "ts": 1716454223722786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650194, "dur": 12,
    "args": {
      "External id": 148693,
      "cbid": 211, "correlation": 148693
    }
  },
  {
    "ph": "s", "id": 148693, "pid": 76337, "tid": -914061504, "ts": 1716454223650194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223722795, "dur": 3,
    "args": {
      "External id": 148695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148695,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 148695, "pid": 5, "tid": 7, "ts": 1716454223722795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650207, "dur": 5,
    "args": {
      "External id": 148695,
      "cbid": 211, "correlation": 148695
    }
  },
  {
    "ph": "s", "id": 148695, "pid": 76337, "tid": -914061504, "ts": 1716454223650207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223722800, "dur": 91,
    "args": {
      "External id": 148716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148716,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 148716, "pid": 5, "tid": 7, "ts": 1716454223722800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650285, "dur": 13,
    "args": {
      "External id": 148716,
      "cbid": 211, "correlation": 148716
    }
  },
  {
    "ph": "s", "id": 148716, "pid": 76337, "tid": -914061504, "ts": 1716454223650285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650386, "dur": 1,
    "args": {
      "External id": 148734,
      "cbid": 251, "correlation": 148734
    }
  },
  {
    "ph": "f", "id": 148734, "pid": 76337, "tid": -914061504, "ts": 1716454223650386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223722892, "dur": 97,
    "args": {
      "External id": 148736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148736,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148736, "pid": 5, "tid": 7, "ts": 1716454223722892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650392, "dur": 13,
    "args": {
      "External id": 148736,
      "cbid": 211, "correlation": 148736
    }
  },
  {
    "ph": "s", "id": 148736, "pid": 76337, "tid": -914061504, "ts": 1716454223650392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223722991, "dur": 19,
    "args": {
      "External id": 148744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148744,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148744, "pid": 5, "tid": 7, "ts": 1716454223722991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650461, "dur": 12,
    "args": {
      "External id": 148744,
      "cbid": 211, "correlation": 148744
    }
  },
  {
    "ph": "s", "id": 148744, "pid": 76337, "tid": -914061504, "ts": 1716454223650461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223723011, "dur": 38,
    "args": {
      "External id": 148752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148752, "pid": 5, "tid": 7, "ts": 1716454223723011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650504, "dur": 9,
    "args": {
      "External id": 148752,
      "cbid": 211, "correlation": 148752
    }
  },
  {
    "ph": "s", "id": 148752, "pid": 76337, "tid": -914061504, "ts": 1716454223650504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223723050, "dur": 35,
    "args": {
      "External id": 148774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148774,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148774, "pid": 5, "tid": 7, "ts": 1716454223723050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650555, "dur": 10,
    "args": {
      "External id": 148774,
      "cbid": 211, "correlation": 148774
    }
  },
  {
    "ph": "s", "id": 148774, "pid": 76337, "tid": -914061504, "ts": 1716454223650555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650650, "dur": 1,
    "args": {
      "External id": 148790,
      "cbid": 251, "correlation": 148790
    }
  },
  {
    "ph": "f", "id": 148790, "pid": 76337, "tid": -914061504, "ts": 1716454223650650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650655, "dur": 0,
    "args": {
      "External id": 148792,
      "cbid": 251, "correlation": 148792
    }
  },
  {
    "ph": "f", "id": 148792, "pid": 76337, "tid": -914061504, "ts": 1716454223650655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223723086, "dur": 540,
    "args": {
      "External id": 148793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148793,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 148793, "pid": 5, "tid": 7, "ts": 1716454223723086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650659, "dur": 12,
    "args": {
      "External id": 148793,
      "cbid": 211, "correlation": 148793
    }
  },
  {
    "ph": "s", "id": 148793, "pid": 76337, "tid": -914061504, "ts": 1716454223650659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223723627, "dur": 124,
    "args": {
      "External id": 148801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148801, "pid": 5, "tid": 7, "ts": 1716454223723627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650733, "dur": 14,
    "args": {
      "External id": 148801,
      "cbid": 211, "correlation": 148801
    }
  },
  {
    "ph": "s", "id": 148801, "pid": 76337, "tid": -914061504, "ts": 1716454223650733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223723753, "dur": 128,
    "args": {
      "External id": 148809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148809, "pid": 5, "tid": 7, "ts": 1716454223723753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650768, "dur": 10,
    "args": {
      "External id": 148809,
      "cbid": 211, "correlation": 148809
    }
  },
  {
    "ph": "s", "id": 148809, "pid": 76337, "tid": -914061504, "ts": 1716454223650768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650847, "dur": 1,
    "args": {
      "External id": 148825,
      "cbid": 251, "correlation": 148825
    }
  },
  {
    "ph": "f", "id": 148825, "pid": 76337, "tid": -914061504, "ts": 1716454223650847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223723882, "dur": 306,
    "args": {
      "External id": 148827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148827,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148827, "pid": 5, "tid": 7, "ts": 1716454223723882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650853, "dur": 12,
    "args": {
      "External id": 148827,
      "cbid": 211, "correlation": 148827
    }
  },
  {
    "ph": "s", "id": 148827, "pid": 76337, "tid": -914061504, "ts": 1716454223650853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223724190, "dur": 27,
    "args": {
      "External id": 148835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148835,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148835, "pid": 5, "tid": 7, "ts": 1716454223724190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650895, "dur": 10,
    "args": {
      "External id": 148835,
      "cbid": 211, "correlation": 148835
    }
  },
  {
    "ph": "s", "id": 148835, "pid": 76337, "tid": -914061504, "ts": 1716454223650895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223724218, "dur": 81,
    "args": {
      "External id": 148846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148846, "pid": 5, "tid": 7, "ts": 1716454223724218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223650967, "dur": 21,
    "args": {
      "External id": 148846,
      "cbid": 211, "correlation": 148846
    }
  },
  {
    "ph": "s", "id": 148846, "pid": 76337, "tid": -914061504, "ts": 1716454223650967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651051, "dur": 0,
    "args": {
      "External id": 148858,
      "cbid": 317, "correlation": 148858
    }
  },
  {
    "ph": "f", "id": 148858, "pid": 76337, "tid": -914061504, "ts": 1716454223651051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651052, "dur": 1,
    "args": {
      "External id": 148859,
      "cbid": 203, "correlation": 148859
    }
  },
  {
    "ph": "f", "id": 148859, "pid": 76337, "tid": -914061504, "ts": 1716454223651052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651054, "dur": 0,
    "args": {
      "External id": 148860,
      "cbid": 205, "correlation": 148860
    }
  },
  {
    "ph": "f", "id": 148860, "pid": 76337, "tid": -914061504, "ts": 1716454223651054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223724301, "dur": 24,
    "args": {
      "External id": 148864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148864,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148864, "pid": 5, "tid": 7, "ts": 1716454223724301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651077, "dur": 13,
    "args": {
      "External id": 148864,
      "cbid": 211, "correlation": 148864
    }
  },
  {
    "ph": "s", "id": 148864, "pid": 76337, "tid": -914061504, "ts": 1716454223651077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223724326, "dur": 121,
    "args": {
      "External id": 148866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148866,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148866, "pid": 5, "tid": 7, "ts": 1716454223724326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651100, "dur": 7,
    "args": {
      "External id": 148866,
      "cbid": 211, "correlation": 148866
    }
  },
  {
    "ph": "s", "id": 148866, "pid": 76337, "tid": -914061504, "ts": 1716454223651100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223724448, "dur": 24,
    "args": {
      "External id": 148868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148868, "pid": 5, "tid": 7, "ts": 1716454223724448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651112, "dur": 9,
    "args": {
      "External id": 148868,
      "cbid": 211, "correlation": 148868
    }
  },
  {
    "ph": "s", "id": 148868, "pid": 76337, "tid": -914061504, "ts": 1716454223651112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223724473, "dur": 33,
    "args": {
      "External id": 148874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148874,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148874, "pid": 5, "tid": 7, "ts": 1716454223724473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651145, "dur": 8,
    "args": {
      "External id": 148874,
      "cbid": 211, "correlation": 148874
    }
  },
  {
    "ph": "s", "id": 148874, "pid": 76337, "tid": -914061504, "ts": 1716454223651145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223724507, "dur": 27,
    "args": {
      "External id": 148882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148882,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148882, "pid": 5, "tid": 7, "ts": 1716454223724507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651177, "dur": 8,
    "args": {
      "External id": 148882,
      "cbid": 211, "correlation": 148882
    }
  },
  {
    "ph": "s", "id": 148882, "pid": 76337, "tid": -914061504, "ts": 1716454223651177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223724535, "dur": 45,
    "args": {
      "External id": 148891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148891,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148891, "pid": 5, "tid": 7, "ts": 1716454223724535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651224, "dur": 11,
    "args": {
      "External id": 148891,
      "cbid": 211, "correlation": 148891
    }
  },
  {
    "ph": "s", "id": 148891, "pid": 76337, "tid": -914061504, "ts": 1716454223651224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223724581, "dur": 42,
    "args": {
      "External id": 148911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148911,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 148911, "pid": 5, "tid": 7, "ts": 1716454223724581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651309, "dur": 12,
    "args": {
      "External id": 148911,
      "cbid": 211, "correlation": 148911
    }
  },
  {
    "ph": "s", "id": 148911, "pid": 76337, "tid": -914061504, "ts": 1716454223651309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223724625, "dur": 5,
    "args": {
      "External id": 148923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148923,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 148923, "pid": 5, "tid": 7, "ts": 1716454223724625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651332, "dur": 7,
    "args": {
      "External id": 148923,
      "cbid": 211, "correlation": 148923
    }
  },
  {
    "ph": "s", "id": 148923, "pid": 76337, "tid": -914061504, "ts": 1716454223651332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223724631, "dur": 44,
    "args": {
      "External id": 148926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148926, "pid": 5, "tid": 7, "ts": 1716454223724631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651353, "dur": 6,
    "args": {
      "External id": 148926,
      "cbid": 211, "correlation": 148926
    }
  },
  {
    "ph": "s", "id": 148926, "pid": 76337, "tid": -914061504, "ts": 1716454223651353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223724676, "dur": 30,
    "args": {
      "External id": 148935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148935,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148935, "pid": 5, "tid": 7, "ts": 1716454223724676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651398, "dur": 10,
    "args": {
      "External id": 148935,
      "cbid": 211, "correlation": 148935
    }
  },
  {
    "ph": "s", "id": 148935, "pid": 76337, "tid": -914061504, "ts": 1716454223651398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651452, "dur": 0,
    "args": {
      "External id": 148945,
      "cbid": 317, "correlation": 148945
    }
  },
  {
    "ph": "f", "id": 148945, "pid": 76337, "tid": -914061504, "ts": 1716454223651452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651452, "dur": 0,
    "args": {
      "External id": 148946,
      "cbid": 203, "correlation": 148946
    }
  },
  {
    "ph": "f", "id": 148946, "pid": 76337, "tid": -914061504, "ts": 1716454223651452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651453, "dur": 0,
    "args": {
      "External id": 148947,
      "cbid": 205, "correlation": 148947
    }
  },
  {
    "ph": "f", "id": 148947, "pid": 76337, "tid": -914061504, "ts": 1716454223651453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223724706, "dur": 30,
    "args": {
      "External id": 148951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148951,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148951, "pid": 5, "tid": 7, "ts": 1716454223724706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651471, "dur": 12,
    "args": {
      "External id": 148951,
      "cbid": 211, "correlation": 148951
    }
  },
  {
    "ph": "s", "id": 148951, "pid": 76337, "tid": -914061504, "ts": 1716454223651471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223724738, "dur": 64,
    "args": {
      "External id": 148953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148953,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148953, "pid": 5, "tid": 7, "ts": 1716454223724738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651485, "dur": 5,
    "args": {
      "External id": 148953,
      "cbid": 211, "correlation": 148953
    }
  },
  {
    "ph": "s", "id": 148953, "pid": 76337, "tid": -914061504, "ts": 1716454223651485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223724803, "dur": 967,
    "args": {
      "External id": 148955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148955,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 148955, "pid": 5, "tid": 7, "ts": 1716454223724803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651496, "dur": 8,
    "args": {
      "External id": 148955,
      "cbid": 211, "correlation": 148955
    }
  },
  {
    "ph": "s", "id": 148955, "pid": 76337, "tid": -914061504, "ts": 1716454223651496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223725771, "dur": 21,
    "args": {
      "External id": 148957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148957,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148957, "pid": 5, "tid": 7, "ts": 1716454223725771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651508, "dur": 5,
    "args": {
      "External id": 148957,
      "cbid": 211, "correlation": 148957
    }
  },
  {
    "ph": "s", "id": 148957, "pid": 76337, "tid": -914061504, "ts": 1716454223651508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223725794, "dur": 33,
    "args": {
      "External id": 148963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148963,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 148963, "pid": 5, "tid": 7, "ts": 1716454223725794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651536, "dur": 8,
    "args": {
      "External id": 148963,
      "cbid": 211, "correlation": 148963
    }
  },
  {
    "ph": "s", "id": 148963, "pid": 76337, "tid": -914061504, "ts": 1716454223651536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223725828, "dur": 4,
    "args": {
      "External id": 148971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148971,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 148971, "pid": 5, "tid": 7, "ts": 1716454223725828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651579, "dur": 9,
    "args": {
      "External id": 148971,
      "cbid": 211, "correlation": 148971
    }
  },
  {
    "ph": "s", "id": 148971, "pid": 76337, "tid": -914061504, "ts": 1716454223651579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651647, "dur": 1,
    "args": {
      "External id": 148987,
      "cbid": 251, "correlation": 148987
    }
  },
  {
    "ph": "f", "id": 148987, "pid": 76337, "tid": -914061504, "ts": 1716454223651647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651652, "dur": 0,
    "args": {
      "External id": 148989,
      "cbid": 251, "correlation": 148989
    }
  },
  {
    "ph": "f", "id": 148989, "pid": 76337, "tid": -914061504, "ts": 1716454223651652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223725834, "dur": 12,
    "args": {
      "External id": 148990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148990,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 148990, "pid": 5, "tid": 7, "ts": 1716454223725834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651654, "dur": 12,
    "args": {
      "External id": 148990,
      "cbid": 211, "correlation": 148990
    }
  },
  {
    "ph": "s", "id": 148990, "pid": 76337, "tid": -914061504, "ts": 1716454223651654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223725847, "dur": 5,
    "args": {
      "External id": 148992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 148992,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 148992, "pid": 5, "tid": 7, "ts": 1716454223725847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651668, "dur": 6,
    "args": {
      "External id": 148992,
      "cbid": 211, "correlation": 148992
    }
  },
  {
    "ph": "s", "id": 148992, "pid": 76337, "tid": -914061504, "ts": 1716454223651668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223725854, "dur": 29,
    "args": {
      "External id": 149002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149002, "pid": 5, "tid": 7, "ts": 1716454223725854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651730, "dur": 12,
    "args": {
      "External id": 149002,
      "cbid": 211, "correlation": 149002
    }
  },
  {
    "ph": "s", "id": 149002, "pid": 76337, "tid": -914061504, "ts": 1716454223651730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223725884, "dur": 30,
    "args": {
      "External id": 149022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149022,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 149022, "pid": 5, "tid": 7, "ts": 1716454223725884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651796, "dur": 10,
    "args": {
      "External id": 149022,
      "cbid": 211, "correlation": 149022
    }
  },
  {
    "ph": "s", "id": 149022, "pid": 76337, "tid": -914061504, "ts": 1716454223651796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223725915, "dur": 4,
    "args": {
      "External id": 149034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149034,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 149034, "pid": 5, "tid": 7, "ts": 1716454223725915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651816, "dur": 7,
    "args": {
      "External id": 149034,
      "cbid": 211, "correlation": 149034
    }
  },
  {
    "ph": "s", "id": 149034, "pid": 76337, "tid": -914061504, "ts": 1716454223651816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223725921, "dur": 30,
    "args": {
      "External id": 149037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149037,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149037, "pid": 5, "tid": 7, "ts": 1716454223725921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651835, "dur": 6,
    "args": {
      "External id": 149037,
      "cbid": 211, "correlation": 149037
    }
  },
  {
    "ph": "s", "id": 149037, "pid": 76337, "tid": -914061504, "ts": 1716454223651835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223725952, "dur": 20,
    "args": {
      "External id": 149046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149046,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149046, "pid": 5, "tid": 7, "ts": 1716454223725952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651876, "dur": 9,
    "args": {
      "External id": 149046,
      "cbid": 211, "correlation": 149046
    }
  },
  {
    "ph": "s", "id": 149046, "pid": 76337, "tid": -914061504, "ts": 1716454223651876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651938, "dur": 0,
    "args": {
      "External id": 149056,
      "cbid": 317, "correlation": 149056
    }
  },
  {
    "ph": "f", "id": 149056, "pid": 76337, "tid": -914061504, "ts": 1716454223651938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651939, "dur": 0,
    "args": {
      "External id": 149057,
      "cbid": 203, "correlation": 149057
    }
  },
  {
    "ph": "f", "id": 149057, "pid": 76337, "tid": -914061504, "ts": 1716454223651939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651940, "dur": 0,
    "args": {
      "External id": 149058,
      "cbid": 205, "correlation": 149058
    }
  },
  {
    "ph": "f", "id": 149058, "pid": 76337, "tid": -914061504, "ts": 1716454223651940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223725973, "dur": 22,
    "args": {
      "External id": 149062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149062,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149062, "pid": 5, "tid": 7, "ts": 1716454223725973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651957, "dur": 12,
    "args": {
      "External id": 149062,
      "cbid": 211, "correlation": 149062
    }
  },
  {
    "ph": "s", "id": 149062, "pid": 76337, "tid": -914061504, "ts": 1716454223651957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223725997, "dur": 44,
    "args": {
      "External id": 149064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149064,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149064, "pid": 5, "tid": 7, "ts": 1716454223725997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651972, "dur": 14,
    "args": {
      "External id": 149064,
      "cbid": 211, "correlation": 149064
    }
  },
  {
    "ph": "s", "id": 149064, "pid": 76337, "tid": -914061504, "ts": 1716454223651972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223726042, "dur": 644,
    "args": {
      "External id": 149066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149066,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149066, "pid": 5, "tid": 7, "ts": 1716454223726042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223651992, "dur": 7,
    "args": {
      "External id": 149066,
      "cbid": 211, "correlation": 149066
    }
  },
  {
    "ph": "s", "id": 149066, "pid": 76337, "tid": -914061504, "ts": 1716454223651992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223726687, "dur": 22,
    "args": {
      "External id": 149068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149068,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149068, "pid": 5, "tid": 7, "ts": 1716454223726687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652003, "dur": 5,
    "args": {
      "External id": 149068,
      "cbid": 211, "correlation": 149068
    }
  },
  {
    "ph": "s", "id": 149068, "pid": 76337, "tid": -914061504, "ts": 1716454223652003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223726711, "dur": 33,
    "args": {
      "External id": 149074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149074, "pid": 5, "tid": 7, "ts": 1716454223726711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652032, "dur": 8,
    "args": {
      "External id": 149074,
      "cbid": 211, "correlation": 149074
    }
  },
  {
    "ph": "s", "id": 149074, "pid": 76337, "tid": -914061504, "ts": 1716454223652032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652090, "dur": 0,
    "args": {
      "External id": 149084,
      "cbid": 317, "correlation": 149084
    }
  },
  {
    "ph": "f", "id": 149084, "pid": 76337, "tid": -914061504, "ts": 1716454223652090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652090, "dur": 0,
    "args": {
      "External id": 149085,
      "cbid": 203, "correlation": 149085
    }
  },
  {
    "ph": "f", "id": 149085, "pid": 76337, "tid": -914061504, "ts": 1716454223652090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652091, "dur": 0,
    "args": {
      "External id": 149086,
      "cbid": 205, "correlation": 149086
    }
  },
  {
    "ph": "f", "id": 149086, "pid": 76337, "tid": -914061504, "ts": 1716454223652091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223726745, "dur": 30,
    "args": {
      "External id": 149090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149090, "pid": 5, "tid": 7, "ts": 1716454223726745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652105, "dur": 12,
    "args": {
      "External id": 149090,
      "cbid": 211, "correlation": 149090
    }
  },
  {
    "ph": "s", "id": 149090, "pid": 76337, "tid": -914061504, "ts": 1716454223652105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223726776, "dur": 154,
    "args": {
      "External id": 149092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149092,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149092, "pid": 5, "tid": 7, "ts": 1716454223726776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652123, "dur": 6,
    "args": {
      "External id": 149092,
      "cbid": 211, "correlation": 149092
    }
  },
  {
    "ph": "s", "id": 149092, "pid": 76337, "tid": -914061504, "ts": 1716454223652123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223726931, "dur": 24,
    "args": {
      "External id": 149094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149094,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149094, "pid": 5, "tid": 7, "ts": 1716454223726931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652133, "dur": 5,
    "args": {
      "External id": 149094,
      "cbid": 211, "correlation": 149094
    }
  },
  {
    "ph": "s", "id": 149094, "pid": 76337, "tid": -914061504, "ts": 1716454223652133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223726956, "dur": 32,
    "args": {
      "External id": 149100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149100, "pid": 5, "tid": 7, "ts": 1716454223726956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652158, "dur": 8,
    "args": {
      "External id": 149100,
      "cbid": 211, "correlation": 149100
    }
  },
  {
    "ph": "s", "id": 149100, "pid": 76337, "tid": -914061504, "ts": 1716454223652158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223726990, "dur": 27,
    "args": {
      "External id": 149108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149108,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149108, "pid": 5, "tid": 7, "ts": 1716454223726990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652186, "dur": 7,
    "args": {
      "External id": 149108,
      "cbid": 211, "correlation": 149108
    }
  },
  {
    "ph": "s", "id": 149108, "pid": 76337, "tid": -914061504, "ts": 1716454223652186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223727018, "dur": 20,
    "args": {
      "External id": 149116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149116,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149116, "pid": 5, "tid": 7, "ts": 1716454223727018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652215, "dur": 11,
    "args": {
      "External id": 149116,
      "cbid": 211, "correlation": 149116
    }
  },
  {
    "ph": "s", "id": 149116, "pid": 76337, "tid": -914061504, "ts": 1716454223652215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223727039, "dur": 29,
    "args": {
      "External id": 149136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149136,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 149136, "pid": 5, "tid": 7, "ts": 1716454223727039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652300, "dur": 12,
    "args": {
      "External id": 149136,
      "cbid": 211, "correlation": 149136
    }
  },
  {
    "ph": "s", "id": 149136, "pid": 76337, "tid": -914061504, "ts": 1716454223652300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223727070, "dur": 5,
    "args": {
      "External id": 149148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149148,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 149148, "pid": 5, "tid": 7, "ts": 1716454223727070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652322, "dur": 6,
    "args": {
      "External id": 149148,
      "cbid": 211, "correlation": 149148
    }
  },
  {
    "ph": "s", "id": 149148, "pid": 76337, "tid": -914061504, "ts": 1716454223652322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223727076, "dur": 31,
    "args": {
      "External id": 149151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149151, "pid": 5, "tid": 7, "ts": 1716454223727076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652340, "dur": 6,
    "args": {
      "External id": 149151,
      "cbid": 211, "correlation": 149151
    }
  },
  {
    "ph": "s", "id": 149151, "pid": 76337, "tid": -914061504, "ts": 1716454223652340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652395, "dur": 0,
    "args": {
      "External id": 149162,
      "cbid": 317, "correlation": 149162
    }
  },
  {
    "ph": "f", "id": 149162, "pid": 76337, "tid": -914061504, "ts": 1716454223652395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652396, "dur": 0,
    "args": {
      "External id": 149163,
      "cbid": 203, "correlation": 149163
    }
  },
  {
    "ph": "f", "id": 149163, "pid": 76337, "tid": -914061504, "ts": 1716454223652396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652397, "dur": 0,
    "args": {
      "External id": 149164,
      "cbid": 205, "correlation": 149164
    }
  },
  {
    "ph": "f", "id": 149164, "pid": 76337, "tid": -914061504, "ts": 1716454223652397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223727108, "dur": 22,
    "args": {
      "External id": 149168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149168, "pid": 5, "tid": 7, "ts": 1716454223727108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652410, "dur": 12,
    "args": {
      "External id": 149168,
      "cbid": 211, "correlation": 149168
    }
  },
  {
    "ph": "s", "id": 149168, "pid": 76337, "tid": -914061504, "ts": 1716454223652410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223727131, "dur": 105,
    "args": {
      "External id": 149170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149170,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149170, "pid": 5, "tid": 7, "ts": 1716454223727131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652428, "dur": 6,
    "args": {
      "External id": 149170,
      "cbid": 211, "correlation": 149170
    }
  },
  {
    "ph": "s", "id": 149170, "pid": 76337, "tid": -914061504, "ts": 1716454223652428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223727237, "dur": 22,
    "args": {
      "External id": 149172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149172,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149172, "pid": 5, "tid": 7, "ts": 1716454223727237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652437, "dur": 5,
    "args": {
      "External id": 149172,
      "cbid": 211, "correlation": 149172
    }
  },
  {
    "ph": "s", "id": 149172, "pid": 76337, "tid": -914061504, "ts": 1716454223652437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223727260, "dur": 32,
    "args": {
      "External id": 149178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149178, "pid": 5, "tid": 7, "ts": 1716454223727260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652464, "dur": 8,
    "args": {
      "External id": 149178,
      "cbid": 211, "correlation": 149178
    }
  },
  {
    "ph": "s", "id": 149178, "pid": 76337, "tid": -914061504, "ts": 1716454223652464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223727293, "dur": 193,
    "args": {
      "External id": 149187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149187,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149187, "pid": 5, "tid": 7, "ts": 1716454223727293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652549, "dur": 16,
    "args": {
      "External id": 149187,
      "cbid": 211, "correlation": 149187
    }
  },
  {
    "ph": "s", "id": 149187, "pid": 76337, "tid": -914061504, "ts": 1716454223652549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223727487, "dur": 65,
    "args": {
      "External id": 149209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149209,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149209, "pid": 5, "tid": 7, "ts": 1716454223727487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652610, "dur": 10,
    "args": {
      "External id": 149209,
      "cbid": 211, "correlation": 149209
    }
  },
  {
    "ph": "s", "id": 149209, "pid": 76337, "tid": -914061504, "ts": 1716454223652610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652701, "dur": 2,
    "args": {
      "External id": 149220,
      "cbid": 251, "correlation": 149220
    }
  },
  {
    "ph": "f", "id": 149220, "pid": 76337, "tid": -914061504, "ts": 1716454223652701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223727553, "dur": 149,
    "args": {
      "External id": 149221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149221,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149221, "pid": 5, "tid": 7, "ts": 1716454223727553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652708, "dur": 13,
    "args": {
      "External id": 149221,
      "cbid": 211, "correlation": 149221
    }
  },
  {
    "ph": "s", "id": 149221, "pid": 76337, "tid": -914061504, "ts": 1716454223652708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652780, "dur": 1,
    "args": {
      "External id": 149232,
      "cbid": 251, "correlation": 149232
    }
  },
  {
    "ph": "f", "id": 149232, "pid": 76337, "tid": -914061504, "ts": 1716454223652780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223727704, "dur": 146,
    "args": {
      "External id": 149233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149233,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149233, "pid": 5, "tid": 7, "ts": 1716454223727704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652784, "dur": 11,
    "args": {
      "External id": 149233,
      "cbid": 211, "correlation": 149233
    }
  },
  {
    "ph": "s", "id": 149233, "pid": 76337, "tid": -914061504, "ts": 1716454223652784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652850, "dur": 1,
    "args": {
      "External id": 149244,
      "cbid": 251, "correlation": 149244
    }
  },
  {
    "ph": "f", "id": 149244, "pid": 76337, "tid": -914061504, "ts": 1716454223652850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223727851, "dur": 144,
    "args": {
      "External id": 149245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149245,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149245, "pid": 5, "tid": 7, "ts": 1716454223727851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652854, "dur": 11,
    "args": {
      "External id": 149245,
      "cbid": 211, "correlation": 149245
    }
  },
  {
    "ph": "s", "id": 149245, "pid": 76337, "tid": -914061504, "ts": 1716454223652854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223727996, "dur": 1948,
    "args": {
      "External id": 149266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149266,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 149266, "pid": 5, "tid": 7, "ts": 1716454223727996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223652937, "dur": 13,
    "args": {
      "External id": 149266,
      "cbid": 211, "correlation": 149266
    }
  },
  {
    "ph": "s", "id": 149266, "pid": 76337, "tid": -914061504, "ts": 1716454223652937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653047, "dur": 1,
    "args": {
      "External id": 149284,
      "cbid": 251, "correlation": 149284
    }
  },
  {
    "ph": "f", "id": 149284, "pid": 76337, "tid": -914061504, "ts": 1716454223653047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223729945, "dur": 151,
    "args": {
      "External id": 149286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149286,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 149286, "pid": 5, "tid": 7, "ts": 1716454223729945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653052, "dur": 13,
    "args": {
      "External id": 149286,
      "cbid": 211, "correlation": 149286
    }
  },
  {
    "ph": "s", "id": 149286, "pid": 76337, "tid": -914061504, "ts": 1716454223653052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223730097, "dur": 36,
    "args": {
      "External id": 149294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149294,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149294, "pid": 5, "tid": 7, "ts": 1716454223730097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653123, "dur": 13,
    "args": {
      "External id": 149294,
      "cbid": 211, "correlation": 149294
    }
  },
  {
    "ph": "s", "id": 149294, "pid": 76337, "tid": -914061504, "ts": 1716454223653123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223730134, "dur": 51,
    "args": {
      "External id": 149302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149302,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149302, "pid": 5, "tid": 7, "ts": 1716454223730134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653162, "dur": 9,
    "args": {
      "External id": 149302,
      "cbid": 211, "correlation": 149302
    }
  },
  {
    "ph": "s", "id": 149302, "pid": 76337, "tid": -914061504, "ts": 1716454223653162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223730186, "dur": 29,
    "args": {
      "External id": 149313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149313,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149313, "pid": 5, "tid": 7, "ts": 1716454223730186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653237, "dur": 13,
    "args": {
      "External id": 149313,
      "cbid": 211, "correlation": 149313
    }
  },
  {
    "ph": "s", "id": 149313, "pid": 76337, "tid": -914061504, "ts": 1716454223653237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223730216, "dur": 35,
    "args": {
      "External id": 149335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149335,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149335, "pid": 5, "tid": 7, "ts": 1716454223730216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653268, "dur": 7,
    "args": {
      "External id": 149335,
      "cbid": 211, "correlation": 149335
    }
  },
  {
    "ph": "s", "id": 149335, "pid": 76337, "tid": -914061504, "ts": 1716454223653268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653352, "dur": 1,
    "args": {
      "External id": 149346,
      "cbid": 251, "correlation": 149346
    }
  },
  {
    "ph": "f", "id": 149346, "pid": 76337, "tid": -914061504, "ts": 1716454223653352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223730252, "dur": 89,
    "args": {
      "External id": 149347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149347,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149347, "pid": 5, "tid": 7, "ts": 1716454223730252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653357, "dur": 12,
    "args": {
      "External id": 149347,
      "cbid": 211, "correlation": 149347
    }
  },
  {
    "ph": "s", "id": 149347, "pid": 76337, "tid": -914061504, "ts": 1716454223653357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653427, "dur": 1,
    "args": {
      "External id": 149358,
      "cbid": 251, "correlation": 149358
    }
  },
  {
    "ph": "f", "id": 149358, "pid": 76337, "tid": -914061504, "ts": 1716454223653427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653431, "dur": 0,
    "args": {
      "External id": 149359,
      "cbid": 251, "correlation": 149359
    }
  },
  {
    "ph": "f", "id": 149359, "pid": 76337, "tid": -914061504, "ts": 1716454223653431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223730342, "dur": 12,
    "args": {
      "External id": 149360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149360,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 149360, "pid": 5, "tid": 7, "ts": 1716454223730342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653433, "dur": 11,
    "args": {
      "External id": 149360,
      "cbid": 211, "correlation": 149360
    }
  },
  {
    "ph": "s", "id": 149360, "pid": 76337, "tid": -914061504, "ts": 1716454223653433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223730355, "dur": 5,
    "args": {
      "External id": 149362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149362,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 149362, "pid": 5, "tid": 7, "ts": 1716454223730355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653446, "dur": 5,
    "args": {
      "External id": 149362,
      "cbid": 211, "correlation": 149362
    }
  },
  {
    "ph": "s", "id": 149362, "pid": 76337, "tid": -914061504, "ts": 1716454223653446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653504, "dur": 1,
    "args": {
      "External id": 149373,
      "cbid": 251, "correlation": 149373
    }
  },
  {
    "ph": "f", "id": 149373, "pid": 76337, "tid": -914061504, "ts": 1716454223653504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653507, "dur": 0,
    "args": {
      "External id": 149374,
      "cbid": 251, "correlation": 149374
    }
  },
  {
    "ph": "f", "id": 149374, "pid": 76337, "tid": -914061504, "ts": 1716454223653507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223730362, "dur": 7,
    "args": {
      "External id": 149375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149375,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 149375, "pid": 5, "tid": 7, "ts": 1716454223730362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653508, "dur": 11,
    "args": {
      "External id": 149375,
      "cbid": 211, "correlation": 149375
    }
  },
  {
    "ph": "s", "id": 149375, "pid": 76337, "tid": -914061504, "ts": 1716454223653508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223730371, "dur": 3,
    "args": {
      "External id": 149377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149377,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 149377, "pid": 5, "tid": 7, "ts": 1716454223730371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653521, "dur": 6,
    "args": {
      "External id": 149377,
      "cbid": 211, "correlation": 149377
    }
  },
  {
    "ph": "s", "id": 149377, "pid": 76337, "tid": -914061504, "ts": 1716454223653521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223730375, "dur": 91,
    "args": {
      "External id": 149398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149398,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 149398, "pid": 5, "tid": 7, "ts": 1716454223730375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653595, "dur": 12,
    "args": {
      "External id": 149398,
      "cbid": 211, "correlation": 149398
    }
  },
  {
    "ph": "s", "id": 149398, "pid": 76337, "tid": -914061504, "ts": 1716454223653595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653694, "dur": 1,
    "args": {
      "External id": 149416,
      "cbid": 251, "correlation": 149416
    }
  },
  {
    "ph": "f", "id": 149416, "pid": 76337, "tid": -914061504, "ts": 1716454223653694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223730467, "dur": 98,
    "args": {
      "External id": 149418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149418,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149418, "pid": 5, "tid": 7, "ts": 1716454223730467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653699, "dur": 13,
    "args": {
      "External id": 149418,
      "cbid": 211, "correlation": 149418
    }
  },
  {
    "ph": "s", "id": 149418, "pid": 76337, "tid": -914061504, "ts": 1716454223653699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223730567, "dur": 20,
    "args": {
      "External id": 149426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149426,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149426, "pid": 5, "tid": 7, "ts": 1716454223730567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653767, "dur": 13,
    "args": {
      "External id": 149426,
      "cbid": 211, "correlation": 149426
    }
  },
  {
    "ph": "s", "id": 149426, "pid": 76337, "tid": -914061504, "ts": 1716454223653767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223730588, "dur": 38,
    "args": {
      "External id": 149434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149434,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149434, "pid": 5, "tid": 7, "ts": 1716454223730588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653809, "dur": 9,
    "args": {
      "External id": 149434,
      "cbid": 211, "correlation": 149434
    }
  },
  {
    "ph": "s", "id": 149434, "pid": 76337, "tid": -914061504, "ts": 1716454223653809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223730627, "dur": 34,
    "args": {
      "External id": 149456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149456,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149456, "pid": 5, "tid": 7, "ts": 1716454223730627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653861, "dur": 10,
    "args": {
      "External id": 149456,
      "cbid": 211, "correlation": 149456
    }
  },
  {
    "ph": "s", "id": 149456, "pid": 76337, "tid": -914061504, "ts": 1716454223653861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653951, "dur": 1,
    "args": {
      "External id": 149472,
      "cbid": 251, "correlation": 149472
    }
  },
  {
    "ph": "f", "id": 149472, "pid": 76337, "tid": -914061504, "ts": 1716454223653951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653956, "dur": 0,
    "args": {
      "External id": 149474,
      "cbid": 251, "correlation": 149474
    }
  },
  {
    "ph": "f", "id": 149474, "pid": 76337, "tid": -914061504, "ts": 1716454223653956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223730663, "dur": 538,
    "args": {
      "External id": 149475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149475,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 149475, "pid": 5, "tid": 7, "ts": 1716454223730663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223653959, "dur": 12,
    "args": {
      "External id": 149475,
      "cbid": 211, "correlation": 149475
    }
  },
  {
    "ph": "s", "id": 149475, "pid": 76337, "tid": -914061504, "ts": 1716454223653959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223731202, "dur": 127,
    "args": {
      "External id": 149483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149483, "pid": 5, "tid": 7, "ts": 1716454223731202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654037, "dur": 14,
    "args": {
      "External id": 149483,
      "cbid": 211, "correlation": 149483
    }
  },
  {
    "ph": "s", "id": 149483, "pid": 76337, "tid": -914061504, "ts": 1716454223654037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223731331, "dur": 128,
    "args": {
      "External id": 149491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149491, "pid": 5, "tid": 7, "ts": 1716454223731331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654070, "dur": 9,
    "args": {
      "External id": 149491,
      "cbid": 211, "correlation": 149491
    }
  },
  {
    "ph": "s", "id": 149491, "pid": 76337, "tid": -914061504, "ts": 1716454223654070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654148, "dur": 1,
    "args": {
      "External id": 149507,
      "cbid": 251, "correlation": 149507
    }
  },
  {
    "ph": "f", "id": 149507, "pid": 76337, "tid": -914061504, "ts": 1716454223654148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223731460, "dur": 307,
    "args": {
      "External id": 149509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149509,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149509, "pid": 5, "tid": 7, "ts": 1716454223731460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654154, "dur": 12,
    "args": {
      "External id": 149509,
      "cbid": 211, "correlation": 149509
    }
  },
  {
    "ph": "s", "id": 149509, "pid": 76337, "tid": -914061504, "ts": 1716454223654154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223731768, "dur": 27,
    "args": {
      "External id": 149517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149517,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149517, "pid": 5, "tid": 7, "ts": 1716454223731768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654196, "dur": 10,
    "args": {
      "External id": 149517,
      "cbid": 211, "correlation": 149517
    }
  },
  {
    "ph": "s", "id": 149517, "pid": 76337, "tid": -914061504, "ts": 1716454223654196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223731797, "dur": 81,
    "args": {
      "External id": 149528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149528,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149528, "pid": 5, "tid": 7, "ts": 1716454223731797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654263, "dur": 13,
    "args": {
      "External id": 149528,
      "cbid": 211, "correlation": 149528
    }
  },
  {
    "ph": "s", "id": 149528, "pid": 76337, "tid": -914061504, "ts": 1716454223654263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654327, "dur": 0,
    "args": {
      "External id": 149540,
      "cbid": 317, "correlation": 149540
    }
  },
  {
    "ph": "f", "id": 149540, "pid": 76337, "tid": -914061504, "ts": 1716454223654327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654328, "dur": 0,
    "args": {
      "External id": 149541,
      "cbid": 203, "correlation": 149541
    }
  },
  {
    "ph": "f", "id": 149541, "pid": 76337, "tid": -914061504, "ts": 1716454223654328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654329, "dur": 0,
    "args": {
      "External id": 149542,
      "cbid": 205, "correlation": 149542
    }
  },
  {
    "ph": "f", "id": 149542, "pid": 76337, "tid": -914061504, "ts": 1716454223654329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223731880, "dur": 22,
    "args": {
      "External id": 149546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149546,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149546, "pid": 5, "tid": 7, "ts": 1716454223731880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654344, "dur": 12,
    "args": {
      "External id": 149546,
      "cbid": 211, "correlation": 149546
    }
  },
  {
    "ph": "s", "id": 149546, "pid": 76337, "tid": -914061504, "ts": 1716454223654344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223731903, "dur": 119,
    "args": {
      "External id": 149548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149548,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149548, "pid": 5, "tid": 7, "ts": 1716454223731903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654362, "dur": 7,
    "args": {
      "External id": 149548,
      "cbid": 211, "correlation": 149548
    }
  },
  {
    "ph": "s", "id": 149548, "pid": 76337, "tid": -914061504, "ts": 1716454223654362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223732023, "dur": 23,
    "args": {
      "External id": 149550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149550,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149550, "pid": 5, "tid": 7, "ts": 1716454223732023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654373, "dur": 5,
    "args": {
      "External id": 149550,
      "cbid": 211, "correlation": 149550
    }
  },
  {
    "ph": "s", "id": 149550, "pid": 76337, "tid": -914061504, "ts": 1716454223654373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223732047, "dur": 33,
    "args": {
      "External id": 149556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149556, "pid": 5, "tid": 7, "ts": 1716454223732047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654401, "dur": 8,
    "args": {
      "External id": 149556,
      "cbid": 211, "correlation": 149556
    }
  },
  {
    "ph": "s", "id": 149556, "pid": 76337, "tid": -914061504, "ts": 1716454223654401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223732082, "dur": 27,
    "args": {
      "External id": 149564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149564,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149564, "pid": 5, "tid": 7, "ts": 1716454223732082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654432, "dur": 8,
    "args": {
      "External id": 149564,
      "cbid": 211, "correlation": 149564
    }
  },
  {
    "ph": "s", "id": 149564, "pid": 76337, "tid": -914061504, "ts": 1716454223654432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223732109, "dur": 101,
    "args": {
      "External id": 149575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149575,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149575, "pid": 5, "tid": 7, "ts": 1716454223732109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654499, "dur": 12,
    "args": {
      "External id": 149575,
      "cbid": 211, "correlation": 149575
    }
  },
  {
    "ph": "s", "id": 149575, "pid": 76337, "tid": -914061504, "ts": 1716454223654499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654555, "dur": 0,
    "args": {
      "External id": 149585,
      "cbid": 317, "correlation": 149585
    }
  },
  {
    "ph": "f", "id": 149585, "pid": 76337, "tid": -914061504, "ts": 1716454223654555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654556, "dur": 0,
    "args": {
      "External id": 149586,
      "cbid": 203, "correlation": 149586
    }
  },
  {
    "ph": "f", "id": 149586, "pid": 76337, "tid": -914061504, "ts": 1716454223654556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654557, "dur": 0,
    "args": {
      "External id": 149587,
      "cbid": 205, "correlation": 149587
    }
  },
  {
    "ph": "f", "id": 149587, "pid": 76337, "tid": -914061504, "ts": 1716454223654557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223732212, "dur": 74,
    "args": {
      "External id": 149591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149591, "pid": 5, "tid": 7, "ts": 1716454223732212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654573, "dur": 11,
    "args": {
      "External id": 149591,
      "cbid": 211, "correlation": 149591
    }
  },
  {
    "ph": "s", "id": 149591, "pid": 76337, "tid": -914061504, "ts": 1716454223654573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223732287, "dur": 44,
    "args": {
      "External id": 149593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149593,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149593, "pid": 5, "tid": 7, "ts": 1716454223732287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654587, "dur": 5,
    "args": {
      "External id": 149593,
      "cbid": 211, "correlation": 149593
    }
  },
  {
    "ph": "s", "id": 149593, "pid": 76337, "tid": -914061504, "ts": 1716454223654587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223732333, "dur": 4,
    "args": {
      "External id": 149595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149595,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149595, "pid": 5, "tid": 7, "ts": 1716454223732333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654598, "dur": 7,
    "args": {
      "External id": 149595,
      "cbid": 211, "correlation": 149595
    }
  },
  {
    "ph": "s", "id": 149595, "pid": 76337, "tid": -914061504, "ts": 1716454223654598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654609, "dur": 0,
    "args": {
      "External id": 149596,
      "cbid": 51, "correlation": 149596
    }
  },
  {
    "ph": "s", "id": 149596, "pid": 76337, "tid": -914061504, "ts": 1716454223654609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223732338, "dur": 2236,
    "args": {
      "External id": 149597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149597,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149597, "pid": 5, "tid": 7, "ts": 1716454223732338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654611, "dur": 7,
    "args": {
      "External id": 149597,
      "cbid": 211, "correlation": 149597
    }
  },
  {
    "ph": "s", "id": 149597, "pid": 76337, "tid": -914061504, "ts": 1716454223654611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223734576, "dur": 113,
    "args": {
      "External id": 149602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149602,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149602, "pid": 5, "tid": 7, "ts": 1716454223734576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654640, "dur": 8,
    "args": {
      "External id": 149602,
      "cbid": 211, "correlation": 149602
    }
  },
  {
    "ph": "s", "id": 149602, "pid": 76337, "tid": -914061504, "ts": 1716454223654640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223734690, "dur": 165,
    "args": {
      "External id": 149611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149611,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149611, "pid": 5, "tid": 7, "ts": 1716454223734690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654738, "dur": 13,
    "args": {
      "External id": 149611,
      "cbid": 211, "correlation": 149611
    }
  },
  {
    "ph": "s", "id": 149611, "pid": 76337, "tid": -914061504, "ts": 1716454223654738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223734857, "dur": 127,
    "args": {
      "External id": 149631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149631,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 149631, "pid": 5, "tid": 7, "ts": 1716454223734857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654809, "dur": 12,
    "args": {
      "External id": 149631,
      "cbid": 211, "correlation": 149631
    }
  },
  {
    "ph": "s", "id": 149631, "pid": 76337, "tid": -914061504, "ts": 1716454223654809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223734985, "dur": 4,
    "args": {
      "External id": 149643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149643,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 149643, "pid": 5, "tid": 7, "ts": 1716454223734985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654831, "dur": 8,
    "args": {
      "External id": 149643,
      "cbid": 211, "correlation": 149643
    }
  },
  {
    "ph": "s", "id": 149643, "pid": 76337, "tid": -914061504, "ts": 1716454223654831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223734991, "dur": 158,
    "args": {
      "External id": 149646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149646, "pid": 5, "tid": 7, "ts": 1716454223734991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654851, "dur": 6,
    "args": {
      "External id": 149646,
      "cbid": 211, "correlation": 149646
    }
  },
  {
    "ph": "s", "id": 149646, "pid": 76337, "tid": -914061504, "ts": 1716454223654851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223735151, "dur": 101,
    "args": {
      "External id": 149655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149655,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149655, "pid": 5, "tid": 7, "ts": 1716454223735151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654891, "dur": 9,
    "args": {
      "External id": 149655,
      "cbid": 211, "correlation": 149655
    }
  },
  {
    "ph": "s", "id": 149655, "pid": 76337, "tid": -914061504, "ts": 1716454223654891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654943, "dur": 0,
    "args": {
      "External id": 149665,
      "cbid": 317, "correlation": 149665
    }
  },
  {
    "ph": "f", "id": 149665, "pid": 76337, "tid": -914061504, "ts": 1716454223654943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654944, "dur": 0,
    "args": {
      "External id": 149666,
      "cbid": 203, "correlation": 149666
    }
  },
  {
    "ph": "f", "id": 149666, "pid": 76337, "tid": -914061504, "ts": 1716454223654944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654945, "dur": 0,
    "args": {
      "External id": 149667,
      "cbid": 205, "correlation": 149667
    }
  },
  {
    "ph": "f", "id": 149667, "pid": 76337, "tid": -914061504, "ts": 1716454223654945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223735253, "dur": 111,
    "args": {
      "External id": 149671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149671,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149671, "pid": 5, "tid": 7, "ts": 1716454223735253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654965, "dur": 22,
    "args": {
      "External id": 149671,
      "cbid": 211, "correlation": 149671
    }
  },
  {
    "ph": "s", "id": 149671, "pid": 76337, "tid": -914061504, "ts": 1716454223654965,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223735365, "dur": 34,
    "args": {
      "External id": 149673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149673,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149673, "pid": 5, "tid": 7, "ts": 1716454223735365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223654989, "dur": 5,
    "args": {
      "External id": 149673,
      "cbid": 211, "correlation": 149673
    }
  },
  {
    "ph": "s", "id": 149673, "pid": 76337, "tid": -914061504, "ts": 1716454223654989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223735400, "dur": 4,
    "args": {
      "External id": 149675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149675, "pid": 5, "tid": 7, "ts": 1716454223735400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655000, "dur": 5,
    "args": {
      "External id": 149675,
      "cbid": 211, "correlation": 149675
    }
  },
  {
    "ph": "s", "id": 149675, "pid": 76337, "tid": -914061504, "ts": 1716454223655000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655009, "dur": 0,
    "args": {
      "External id": 149676,
      "cbid": 51, "correlation": 149676
    }
  },
  {
    "ph": "s", "id": 149676, "pid": 76337, "tid": -914061504, "ts": 1716454223655009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223735405, "dur": 2003,
    "args": {
      "External id": 149677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149677,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149677, "pid": 5, "tid": 7, "ts": 1716454223735405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655010, "dur": 7,
    "args": {
      "External id": 149677,
      "cbid": 211, "correlation": 149677
    }
  },
  {
    "ph": "s", "id": 149677, "pid": 76337, "tid": -914061504, "ts": 1716454223655010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223737409, "dur": 59,
    "args": {
      "External id": 149682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149682, "pid": 5, "tid": 7, "ts": 1716454223737409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655041, "dur": 8,
    "args": {
      "External id": 149682,
      "cbid": 211, "correlation": 149682
    }
  },
  {
    "ph": "s", "id": 149682, "pid": 76337, "tid": -914061504, "ts": 1716454223655041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223737470, "dur": 4,
    "args": {
      "External id": 149690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149690,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149690, "pid": 5, "tid": 7, "ts": 1716454223737470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655084, "dur": 10,
    "args": {
      "External id": 149690,
      "cbid": 211, "correlation": 149690
    }
  },
  {
    "ph": "s", "id": 149690, "pid": 76337, "tid": -914061504, "ts": 1716454223655084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655150, "dur": 1,
    "args": {
      "External id": 149706,
      "cbid": 251, "correlation": 149706
    }
  },
  {
    "ph": "f", "id": 149706, "pid": 76337, "tid": -914061504, "ts": 1716454223655150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655155, "dur": 0,
    "args": {
      "External id": 149708,
      "cbid": 251, "correlation": 149708
    }
  },
  {
    "ph": "f", "id": 149708, "pid": 76337, "tid": -914061504, "ts": 1716454223655155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223737475, "dur": 11,
    "args": {
      "External id": 149709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149709,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 149709, "pid": 5, "tid": 7, "ts": 1716454223737475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655157, "dur": 12,
    "args": {
      "External id": 149709,
      "cbid": 211, "correlation": 149709
    }
  },
  {
    "ph": "s", "id": 149709, "pid": 76337, "tid": -914061504, "ts": 1716454223655157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223737488, "dur": 5,
    "args": {
      "External id": 149711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149711,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 149711, "pid": 5, "tid": 7, "ts": 1716454223737488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655171, "dur": 6,
    "args": {
      "External id": 149711,
      "cbid": 211, "correlation": 149711
    }
  },
  {
    "ph": "s", "id": 149711, "pid": 76337, "tid": -914061504, "ts": 1716454223655171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223737494, "dur": 53,
    "args": {
      "External id": 149721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149721,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149721, "pid": 5, "tid": 7, "ts": 1716454223737494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655229, "dur": 12,
    "args": {
      "External id": 149721,
      "cbid": 211, "correlation": 149721
    }
  },
  {
    "ph": "s", "id": 149721, "pid": 76337, "tid": -914061504, "ts": 1716454223655229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223737549, "dur": 51,
    "args": {
      "External id": 149741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149741,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 149741, "pid": 5, "tid": 7, "ts": 1716454223737549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655296, "dur": 10,
    "args": {
      "External id": 149741,
      "cbid": 211, "correlation": 149741
    }
  },
  {
    "ph": "s", "id": 149741, "pid": 76337, "tid": -914061504, "ts": 1716454223655296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223737601, "dur": 4,
    "args": {
      "External id": 149753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149753,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149753, "pid": 5, "tid": 7, "ts": 1716454223737601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655317, "dur": 7,
    "args": {
      "External id": 149753,
      "cbid": 211, "correlation": 149753
    }
  },
  {
    "ph": "s", "id": 149753, "pid": 76337, "tid": -914061504, "ts": 1716454223655317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223737606, "dur": 55,
    "args": {
      "External id": 149756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149756,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149756, "pid": 5, "tid": 7, "ts": 1716454223737606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655335, "dur": 6,
    "args": {
      "External id": 149756,
      "cbid": 211, "correlation": 149756
    }
  },
  {
    "ph": "s", "id": 149756, "pid": 76337, "tid": -914061504, "ts": 1716454223655335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223737663, "dur": 36,
    "args": {
      "External id": 149765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149765,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149765, "pid": 5, "tid": 7, "ts": 1716454223737663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655376, "dur": 9,
    "args": {
      "External id": 149765,
      "cbid": 211, "correlation": 149765
    }
  },
  {
    "ph": "s", "id": 149765, "pid": 76337, "tid": -914061504, "ts": 1716454223655376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655439, "dur": 0,
    "args": {
      "External id": 149775,
      "cbid": 317, "correlation": 149775
    }
  },
  {
    "ph": "f", "id": 149775, "pid": 76337, "tid": -914061504, "ts": 1716454223655439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655440, "dur": 0,
    "args": {
      "External id": 149776,
      "cbid": 203, "correlation": 149776
    }
  },
  {
    "ph": "f", "id": 149776, "pid": 76337, "tid": -914061504, "ts": 1716454223655440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655441, "dur": 0,
    "args": {
      "External id": 149777,
      "cbid": 205, "correlation": 149777
    }
  },
  {
    "ph": "f", "id": 149777, "pid": 76337, "tid": -914061504, "ts": 1716454223655441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223737699, "dur": 39,
    "args": {
      "External id": 149781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149781,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149781, "pid": 5, "tid": 7, "ts": 1716454223737699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655458, "dur": 14,
    "args": {
      "External id": 149781,
      "cbid": 211, "correlation": 149781
    }
  },
  {
    "ph": "s", "id": 149781, "pid": 76337, "tid": -914061504, "ts": 1716454223655458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223737740, "dur": 14,
    "args": {
      "External id": 149783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149783,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149783, "pid": 5, "tid": 7, "ts": 1716454223737740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655475, "dur": 5,
    "args": {
      "External id": 149783,
      "cbid": 211, "correlation": 149783
    }
  },
  {
    "ph": "s", "id": 149783, "pid": 76337, "tid": -914061504, "ts": 1716454223655475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223737755, "dur": 4,
    "args": {
      "External id": 149785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149785, "pid": 5, "tid": 7, "ts": 1716454223737755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655485, "dur": 5,
    "args": {
      "External id": 149785,
      "cbid": 211, "correlation": 149785
    }
  },
  {
    "ph": "s", "id": 149785, "pid": 76337, "tid": -914061504, "ts": 1716454223655485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655493, "dur": 0,
    "args": {
      "External id": 149786,
      "cbid": 51, "correlation": 149786
    }
  },
  {
    "ph": "s", "id": 149786, "pid": 76337, "tid": -914061504, "ts": 1716454223655493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223737760, "dur": 696,
    "args": {
      "External id": 149787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149787,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149787, "pid": 5, "tid": 7, "ts": 1716454223737760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655494, "dur": 5,
    "args": {
      "External id": 149787,
      "cbid": 211, "correlation": 149787
    }
  },
  {
    "ph": "s", "id": 149787, "pid": 76337, "tid": -914061504, "ts": 1716454223655494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223738457, "dur": 59,
    "args": {
      "External id": 149792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149792,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149792, "pid": 5, "tid": 7, "ts": 1716454223738457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655522, "dur": 8,
    "args": {
      "External id": 149792,
      "cbid": 211, "correlation": 149792
    }
  },
  {
    "ph": "s", "id": 149792, "pid": 76337, "tid": -914061504, "ts": 1716454223655522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655579, "dur": 0,
    "args": {
      "External id": 149802,
      "cbid": 317, "correlation": 149802
    }
  },
  {
    "ph": "f", "id": 149802, "pid": 76337, "tid": -914061504, "ts": 1716454223655579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655580, "dur": 0,
    "args": {
      "External id": 149803,
      "cbid": 203, "correlation": 149803
    }
  },
  {
    "ph": "f", "id": 149803, "pid": 76337, "tid": -914061504, "ts": 1716454223655580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655581, "dur": 0,
    "args": {
      "External id": 149804,
      "cbid": 205, "correlation": 149804
    }
  },
  {
    "ph": "f", "id": 149804, "pid": 76337, "tid": -914061504, "ts": 1716454223655581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223738518, "dur": 4,
    "args": {
      "External id": 149808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149808, "pid": 5, "tid": 7, "ts": 1716454223738518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655598, "dur": 11,
    "args": {
      "External id": 149808,
      "cbid": 211, "correlation": 149808
    }
  },
  {
    "ph": "s", "id": 149808, "pid": 76337, "tid": -914061504, "ts": 1716454223655598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655613, "dur": 0,
    "args": {
      "External id": 149809,
      "cbid": 51, "correlation": 149809
    }
  },
  {
    "ph": "s", "id": 149809, "pid": 76337, "tid": -914061504, "ts": 1716454223655613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223738523, "dur": 265,
    "args": {
      "External id": 149810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149810,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149810, "pid": 5, "tid": 7, "ts": 1716454223738523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655614, "dur": 8,
    "args": {
      "External id": 149810,
      "cbid": 211, "correlation": 149810
    }
  },
  {
    "ph": "s", "id": 149810, "pid": 76337, "tid": -914061504, "ts": 1716454223655614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223738789, "dur": 59,
    "args": {
      "External id": 149815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149815, "pid": 5, "tid": 7, "ts": 1716454223738789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655642, "dur": 8,
    "args": {
      "External id": 149815,
      "cbid": 211, "correlation": 149815
    }
  },
  {
    "ph": "s", "id": 149815, "pid": 76337, "tid": -914061504, "ts": 1716454223655642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223738849, "dur": 50,
    "args": {
      "External id": 149823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149823,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149823, "pid": 5, "tid": 7, "ts": 1716454223738849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655671, "dur": 9,
    "args": {
      "External id": 149823,
      "cbid": 211, "correlation": 149823
    }
  },
  {
    "ph": "s", "id": 149823, "pid": 76337, "tid": -914061504, "ts": 1716454223655671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223738900, "dur": 35,
    "args": {
      "External id": 149831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149831,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149831, "pid": 5, "tid": 7, "ts": 1716454223738900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655701, "dur": 8,
    "args": {
      "External id": 149831,
      "cbid": 211, "correlation": 149831
    }
  },
  {
    "ph": "s", "id": 149831, "pid": 76337, "tid": -914061504, "ts": 1716454223655701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223738936, "dur": 50,
    "args": {
      "External id": 149851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149851,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 149851, "pid": 5, "tid": 7, "ts": 1716454223738936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655782, "dur": 12,
    "args": {
      "External id": 149851,
      "cbid": 211, "correlation": 149851
    }
  },
  {
    "ph": "s", "id": 149851, "pid": 76337, "tid": -914061504, "ts": 1716454223655782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223738988, "dur": 4,
    "args": {
      "External id": 149863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149863,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 149863, "pid": 5, "tid": 7, "ts": 1716454223738988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655804, "dur": 6,
    "args": {
      "External id": 149863,
      "cbid": 211, "correlation": 149863
    }
  },
  {
    "ph": "s", "id": 149863, "pid": 76337, "tid": -914061504, "ts": 1716454223655804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223738993, "dur": 55,
    "args": {
      "External id": 149866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149866, "pid": 5, "tid": 7, "ts": 1716454223738993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655822, "dur": 6,
    "args": {
      "External id": 149866,
      "cbid": 211, "correlation": 149866
    }
  },
  {
    "ph": "s", "id": 149866, "pid": 76337, "tid": -914061504, "ts": 1716454223655822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655878, "dur": 0,
    "args": {
      "External id": 149877,
      "cbid": 317, "correlation": 149877
    }
  },
  {
    "ph": "f", "id": 149877, "pid": 76337, "tid": -914061504, "ts": 1716454223655878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655879, "dur": 0,
    "args": {
      "External id": 149878,
      "cbid": 203, "correlation": 149878
    }
  },
  {
    "ph": "f", "id": 149878, "pid": 76337, "tid": -914061504, "ts": 1716454223655879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655880, "dur": 0,
    "args": {
      "External id": 149879,
      "cbid": 205, "correlation": 149879
    }
  },
  {
    "ph": "f", "id": 149879, "pid": 76337, "tid": -914061504, "ts": 1716454223655880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655910, "dur": 2,
    "args": {
      "External id": 149883,
      "cbid": 251, "correlation": 149883
    }
  },
  {
    "ph": "f", "id": 149883, "pid": 76337, "tid": -914061504, "ts": 1716454223655910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655914, "dur": 1,
    "args": {
      "External id": 149884,
      "cbid": 251, "correlation": 149884
    }
  },
  {
    "ph": "f", "id": 149884, "pid": 76337, "tid": -914061504, "ts": 1716454223655914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655916, "dur": 1,
    "args": {
      "External id": 149885,
      "cbid": 251, "correlation": 149885
    }
  },
  {
    "ph": "f", "id": 149885, "pid": 76337, "tid": -914061504, "ts": 1716454223655916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655918, "dur": 1,
    "args": {
      "External id": 149886,
      "cbid": 251, "correlation": 149886
    }
  },
  {
    "ph": "f", "id": 149886, "pid": 76337, "tid": -914061504, "ts": 1716454223655918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655920, "dur": 1,
    "args": {
      "External id": 149887,
      "cbid": 251, "correlation": 149887
    }
  },
  {
    "ph": "f", "id": 149887, "pid": 76337, "tid": -914061504, "ts": 1716454223655920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655922, "dur": 1,
    "args": {
      "External id": 149888,
      "cbid": 251, "correlation": 149888
    }
  },
  {
    "ph": "f", "id": 149888, "pid": 76337, "tid": -914061504, "ts": 1716454223655922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655924, "dur": 1,
    "args": {
      "External id": 149889,
      "cbid": 251, "correlation": 149889
    }
  },
  {
    "ph": "f", "id": 149889, "pid": 76337, "tid": -914061504, "ts": 1716454223655924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655926, "dur": 1,
    "args": {
      "External id": 149890,
      "cbid": 251, "correlation": 149890
    }
  },
  {
    "ph": "f", "id": 149890, "pid": 76337, "tid": -914061504, "ts": 1716454223655926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655929, "dur": 0,
    "args": {
      "External id": 149891,
      "cbid": 251, "correlation": 149891
    }
  },
  {
    "ph": "f", "id": 149891, "pid": 76337, "tid": -914061504, "ts": 1716454223655929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223739049, "dur": 115,
    "args": {
      "External id": 149892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149892,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 149892, "pid": 5, "tid": 7, "ts": 1716454223739049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655931, "dur": 14,
    "args": {
      "External id": 149892,
      "cbid": 211, "correlation": 149892
    }
  },
  {
    "ph": "s", "id": 149892, "pid": 76337, "tid": -914061504, "ts": 1716454223655931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223739165, "dur": 59,
    "args": {
      "External id": 149898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149898, "pid": 5, "tid": 7, "ts": 1716454223739165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223655968, "dur": 17,
    "args": {
      "External id": 149898,
      "cbid": 211, "correlation": 149898
    }
  },
  {
    "ph": "s", "id": 149898, "pid": 76337, "tid": -914061504, "ts": 1716454223655968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223739226, "dur": 608,
    "args": {
      "External id": 149907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149907,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149907, "pid": 5, "tid": 7, "ts": 1716454223739226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656059, "dur": 15,
    "args": {
      "External id": 149907,
      "cbid": 211, "correlation": 149907
    }
  },
  {
    "ph": "s", "id": 149907, "pid": 76337, "tid": -914061504, "ts": 1716454223656059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223739835, "dur": 181,
    "args": {
      "External id": 149929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149929,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 149929, "pid": 5, "tid": 7, "ts": 1716454223739835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656117, "dur": 10,
    "args": {
      "External id": 149929,
      "cbid": 211, "correlation": 149929
    }
  },
  {
    "ph": "s", "id": 149929, "pid": 76337, "tid": -914061504, "ts": 1716454223656117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656203, "dur": 1,
    "args": {
      "External id": 149940,
      "cbid": 251, "correlation": 149940
    }
  },
  {
    "ph": "f", "id": 149940, "pid": 76337, "tid": -914061504, "ts": 1716454223656203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223740017, "dur": 199,
    "args": {
      "External id": 149941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149941,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149941, "pid": 5, "tid": 7, "ts": 1716454223740017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656208, "dur": 14,
    "args": {
      "External id": 149941,
      "cbid": 211, "correlation": 149941
    }
  },
  {
    "ph": "s", "id": 149941, "pid": 76337, "tid": -914061504, "ts": 1716454223656208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656277, "dur": 1,
    "args": {
      "External id": 149952,
      "cbid": 251, "correlation": 149952
    }
  },
  {
    "ph": "f", "id": 149952, "pid": 76337, "tid": -914061504, "ts": 1716454223656277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223740217, "dur": 191,
    "args": {
      "External id": 149953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149953,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149953, "pid": 5, "tid": 7, "ts": 1716454223740217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656281, "dur": 11,
    "args": {
      "External id": 149953,
      "cbid": 211, "correlation": 149953
    }
  },
  {
    "ph": "s", "id": 149953, "pid": 76337, "tid": -914061504, "ts": 1716454223656281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656344, "dur": 1,
    "args": {
      "External id": 149964,
      "cbid": 251, "correlation": 149964
    }
  },
  {
    "ph": "f", "id": 149964, "pid": 76337, "tid": -914061504, "ts": 1716454223656344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223740409, "dur": 187,
    "args": {
      "External id": 149965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149965,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 149965, "pid": 5, "tid": 7, "ts": 1716454223740409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656348, "dur": 12,
    "args": {
      "External id": 149965,
      "cbid": 211, "correlation": 149965
    }
  },
  {
    "ph": "s", "id": 149965, "pid": 76337, "tid": -914061504, "ts": 1716454223656348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223740598, "dur": 18557,
    "args": {
      "External id": 149986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 149986,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 149986, "pid": 5, "tid": 7, "ts": 1716454223740598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656430, "dur": 14,
    "args": {
      "External id": 149986,
      "cbid": 211, "correlation": 149986
    }
  },
  {
    "ph": "s", "id": 149986, "pid": 76337, "tid": -914061504, "ts": 1716454223656430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656529, "dur": 1,
    "args": {
      "External id": 150004,
      "cbid": 251, "correlation": 150004
    }
  },
  {
    "ph": "f", "id": 150004, "pid": 76337, "tid": -914061504, "ts": 1716454223656529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223759156, "dur": 202,
    "args": {
      "External id": 150006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150006,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150006, "pid": 5, "tid": 7, "ts": 1716454223759156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656535, "dur": 13,
    "args": {
      "External id": 150006,
      "cbid": 211, "correlation": 150006
    }
  },
  {
    "ph": "s", "id": 150006, "pid": 76337, "tid": -914061504, "ts": 1716454223656535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223759360, "dur": 67,
    "args": {
      "External id": 150014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150014,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150014, "pid": 5, "tid": 7, "ts": 1716454223759360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656604, "dur": 13,
    "args": {
      "External id": 150014,
      "cbid": 211, "correlation": 150014
    }
  },
  {
    "ph": "s", "id": 150014, "pid": 76337, "tid": -914061504, "ts": 1716454223656604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223759428, "dur": 96,
    "args": {
      "External id": 150022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150022,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150022, "pid": 5, "tid": 7, "ts": 1716454223759428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656643, "dur": 9,
    "args": {
      "External id": 150022,
      "cbid": 211, "correlation": 150022
    }
  },
  {
    "ph": "s", "id": 150022, "pid": 76337, "tid": -914061504, "ts": 1716454223656643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223759526, "dur": 55,
    "args": {
      "External id": 150033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150033, "pid": 5, "tid": 7, "ts": 1716454223759526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656716, "dur": 12,
    "args": {
      "External id": 150033,
      "cbid": 211, "correlation": 150033
    }
  },
  {
    "ph": "s", "id": 150033, "pid": 76337, "tid": -914061504, "ts": 1716454223656716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223759582, "dur": 91,
    "args": {
      "External id": 150055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150055,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150055, "pid": 5, "tid": 7, "ts": 1716454223759582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656747, "dur": 7,
    "args": {
      "External id": 150055,
      "cbid": 211, "correlation": 150055
    }
  },
  {
    "ph": "s", "id": 150055, "pid": 76337, "tid": -914061504, "ts": 1716454223656747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656830, "dur": 1,
    "args": {
      "External id": 150066,
      "cbid": 251, "correlation": 150066
    }
  },
  {
    "ph": "f", "id": 150066, "pid": 76337, "tid": -914061504, "ts": 1716454223656830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223759675, "dur": 103,
    "args": {
      "External id": 150067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150067,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150067, "pid": 5, "tid": 7, "ts": 1716454223759675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656835, "dur": 12,
    "args": {
      "External id": 150067,
      "cbid": 211, "correlation": 150067
    }
  },
  {
    "ph": "s", "id": 150067, "pid": 76337, "tid": -914061504, "ts": 1716454223656835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656908, "dur": 2,
    "args": {
      "External id": 150078,
      "cbid": 251, "correlation": 150078
    }
  },
  {
    "ph": "f", "id": 150078, "pid": 76337, "tid": -914061504, "ts": 1716454223656908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656912, "dur": 0,
    "args": {
      "External id": 150079,
      "cbid": 251, "correlation": 150079
    }
  },
  {
    "ph": "f", "id": 150079, "pid": 76337, "tid": -914061504, "ts": 1716454223656912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223759779, "dur": 10,
    "args": {
      "External id": 150080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150080,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 150080, "pid": 5, "tid": 7, "ts": 1716454223759779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656915, "dur": 13,
    "args": {
      "External id": 150080,
      "cbid": 211, "correlation": 150080
    }
  },
  {
    "ph": "s", "id": 150080, "pid": 76337, "tid": -914061504, "ts": 1716454223656915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223759790, "dur": 5,
    "args": {
      "External id": 150082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150082,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 150082, "pid": 5, "tid": 7, "ts": 1716454223759790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223656931, "dur": 7,
    "args": {
      "External id": 150082,
      "cbid": 211, "correlation": 150082
    }
  },
  {
    "ph": "s", "id": 150082, "pid": 76337, "tid": -914061504, "ts": 1716454223656931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657002, "dur": 1,
    "args": {
      "External id": 150093,
      "cbid": 251, "correlation": 150093
    }
  },
  {
    "ph": "f", "id": 150093, "pid": 76337, "tid": -914061504, "ts": 1716454223657002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657006, "dur": 0,
    "args": {
      "External id": 150094,
      "cbid": 251, "correlation": 150094
    }
  },
  {
    "ph": "f", "id": 150094, "pid": 76337, "tid": -914061504, "ts": 1716454223657006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223759796, "dur": 6,
    "args": {
      "External id": 150095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150095,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 150095, "pid": 5, "tid": 7, "ts": 1716454223759796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657007, "dur": 13,
    "args": {
      "External id": 150095,
      "cbid": 211, "correlation": 150095
    }
  },
  {
    "ph": "s", "id": 150095, "pid": 76337, "tid": -914061504, "ts": 1716454223657007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223759804, "dur": 3,
    "args": {
      "External id": 150097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150097,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 150097, "pid": 5, "tid": 7, "ts": 1716454223759804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657022, "dur": 5,
    "args": {
      "External id": 150097,
      "cbid": 211, "correlation": 150097
    }
  },
  {
    "ph": "s", "id": 150097, "pid": 76337, "tid": -914061504, "ts": 1716454223657022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223759808, "dur": 155,
    "args": {
      "External id": 150118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150118,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150118, "pid": 5, "tid": 7, "ts": 1716454223759808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657096, "dur": 12,
    "args": {
      "External id": 150118,
      "cbid": 211, "correlation": 150118
    }
  },
  {
    "ph": "s", "id": 150118, "pid": 76337, "tid": -914061504, "ts": 1716454223657096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657193, "dur": 2,
    "args": {
      "External id": 150136,
      "cbid": 251, "correlation": 150136
    }
  },
  {
    "ph": "f", "id": 150136, "pid": 76337, "tid": -914061504, "ts": 1716454223657193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223759965, "dur": 107,
    "args": {
      "External id": 150138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150138,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150138, "pid": 5, "tid": 7, "ts": 1716454223759965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657199, "dur": 14,
    "args": {
      "External id": 150138,
      "cbid": 211, "correlation": 150138
    }
  },
  {
    "ph": "s", "id": 150138, "pid": 76337, "tid": -914061504, "ts": 1716454223657199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223760074, "dur": 35,
    "args": {
      "External id": 150146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150146,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150146, "pid": 5, "tid": 7, "ts": 1716454223760074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657271, "dur": 12,
    "args": {
      "External id": 150146,
      "cbid": 211, "correlation": 150146
    }
  },
  {
    "ph": "s", "id": 150146, "pid": 76337, "tid": -914061504, "ts": 1716454223657271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223760110, "dur": 66,
    "args": {
      "External id": 150154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150154,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150154, "pid": 5, "tid": 7, "ts": 1716454223760110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657311, "dur": 10,
    "args": {
      "External id": 150154,
      "cbid": 211, "correlation": 150154
    }
  },
  {
    "ph": "s", "id": 150154, "pid": 76337, "tid": -914061504, "ts": 1716454223657311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223760177, "dur": 92,
    "args": {
      "External id": 150176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150176,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150176, "pid": 5, "tid": 7, "ts": 1716454223760177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657362, "dur": 10,
    "args": {
      "External id": 150176,
      "cbid": 211, "correlation": 150176
    }
  },
  {
    "ph": "s", "id": 150176, "pid": 76337, "tid": -914061504, "ts": 1716454223657362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657449, "dur": 1,
    "args": {
      "External id": 150192,
      "cbid": 251, "correlation": 150192
    }
  },
  {
    "ph": "f", "id": 150192, "pid": 76337, "tid": -914061504, "ts": 1716454223657449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223760270, "dur": 572,
    "args": {
      "External id": 150194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150194,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150194, "pid": 5, "tid": 7, "ts": 1716454223760270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657455, "dur": 12,
    "args": {
      "External id": 150194,
      "cbid": 211, "correlation": 150194
    }
  },
  {
    "ph": "s", "id": 150194, "pid": 76337, "tid": -914061504, "ts": 1716454223657455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223760844, "dur": 244,
    "args": {
      "External id": 150202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150202, "pid": 5, "tid": 7, "ts": 1716454223760844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657520, "dur": 12,
    "args": {
      "External id": 150202,
      "cbid": 211, "correlation": 150202
    }
  },
  {
    "ph": "s", "id": 150202, "pid": 76337, "tid": -914061504, "ts": 1716454223657520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223761089, "dur": 252,
    "args": {
      "External id": 150210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150210, "pid": 5, "tid": 7, "ts": 1716454223761089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657552, "dur": 8,
    "args": {
      "External id": 150210,
      "cbid": 211, "correlation": 150210
    }
  },
  {
    "ph": "s", "id": 150210, "pid": 76337, "tid": -914061504, "ts": 1716454223657552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657634, "dur": 1,
    "args": {
      "External id": 150226,
      "cbid": 251, "correlation": 150226
    }
  },
  {
    "ph": "f", "id": 150226, "pid": 76337, "tid": -914061504, "ts": 1716454223657634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657638, "dur": 0,
    "args": {
      "External id": 150228,
      "cbid": 251, "correlation": 150228
    }
  },
  {
    "ph": "f", "id": 150228, "pid": 76337, "tid": -914061504, "ts": 1716454223657638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223761342, "dur": 357,
    "args": {
      "External id": 150229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150229,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 150229, "pid": 5, "tid": 7, "ts": 1716454223761342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657641, "dur": 12,
    "args": {
      "External id": 150229,
      "cbid": 211, "correlation": 150229
    }
  },
  {
    "ph": "s", "id": 150229, "pid": 76337, "tid": -914061504, "ts": 1716454223657641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223761700, "dur": 50,
    "args": {
      "External id": 150237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150237,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150237, "pid": 5, "tid": 7, "ts": 1716454223761700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657684, "dur": 10,
    "args": {
      "External id": 150237,
      "cbid": 211, "correlation": 150237
    }
  },
  {
    "ph": "s", "id": 150237, "pid": 76337, "tid": -914061504, "ts": 1716454223657684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223761752, "dur": 157,
    "args": {
      "External id": 150248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150248, "pid": 5, "tid": 7, "ts": 1716454223761752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657751, "dur": 12,
    "args": {
      "External id": 150248,
      "cbid": 211, "correlation": 150248
    }
  },
  {
    "ph": "s", "id": 150248, "pid": 76337, "tid": -914061504, "ts": 1716454223657751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657815, "dur": 0,
    "args": {
      "External id": 150260,
      "cbid": 317, "correlation": 150260
    }
  },
  {
    "ph": "f", "id": 150260, "pid": 76337, "tid": -914061504, "ts": 1716454223657815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657816, "dur": 0,
    "args": {
      "External id": 150261,
      "cbid": 203, "correlation": 150261
    }
  },
  {
    "ph": "f", "id": 150261, "pid": 76337, "tid": -914061504, "ts": 1716454223657816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657817, "dur": 0,
    "args": {
      "External id": 150262,
      "cbid": 205, "correlation": 150262
    }
  },
  {
    "ph": "f", "id": 150262, "pid": 76337, "tid": -914061504, "ts": 1716454223657817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657842, "dur": 1,
    "args": {
      "External id": 150266,
      "cbid": 251, "correlation": 150266
    }
  },
  {
    "ph": "f", "id": 150266, "pid": 76337, "tid": -914061504, "ts": 1716454223657842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657844, "dur": 0,
    "args": {
      "External id": 150267,
      "cbid": 251, "correlation": 150267
    }
  },
  {
    "ph": "f", "id": 150267, "pid": 76337, "tid": -914061504, "ts": 1716454223657844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657845, "dur": 0,
    "args": {
      "External id": 150268,
      "cbid": 251, "correlation": 150268
    }
  },
  {
    "ph": "f", "id": 150268, "pid": 76337, "tid": -914061504, "ts": 1716454223657845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657846, "dur": 0,
    "args": {
      "External id": 150269,
      "cbid": 251, "correlation": 150269
    }
  },
  {
    "ph": "f", "id": 150269, "pid": 76337, "tid": -914061504, "ts": 1716454223657846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657846, "dur": 0,
    "args": {
      "External id": 150270,
      "cbid": 251, "correlation": 150270
    }
  },
  {
    "ph": "f", "id": 150270, "pid": 76337, "tid": -914061504, "ts": 1716454223657846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657847, "dur": 0,
    "args": {
      "External id": 150271,
      "cbid": 251, "correlation": 150271
    }
  },
  {
    "ph": "f", "id": 150271, "pid": 76337, "tid": -914061504, "ts": 1716454223657847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657848, "dur": 0,
    "args": {
      "External id": 150272,
      "cbid": 251, "correlation": 150272
    }
  },
  {
    "ph": "f", "id": 150272, "pid": 76337, "tid": -914061504, "ts": 1716454223657848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657849, "dur": 0,
    "args": {
      "External id": 150273,
      "cbid": 251, "correlation": 150273
    }
  },
  {
    "ph": "f", "id": 150273, "pid": 76337, "tid": -914061504, "ts": 1716454223657849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657850, "dur": 0,
    "args": {
      "External id": 150274,
      "cbid": 251, "correlation": 150274
    }
  },
  {
    "ph": "f", "id": 150274, "pid": 76337, "tid": -914061504, "ts": 1716454223657850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223761910, "dur": 114,
    "args": {
      "External id": 150275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150275,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150275, "pid": 5, "tid": 7, "ts": 1716454223761910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657852, "dur": 12,
    "args": {
      "External id": 150275,
      "cbid": 211, "correlation": 150275
    }
  },
  {
    "ph": "s", "id": 150275, "pid": 76337, "tid": -914061504, "ts": 1716454223657852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223762026, "dur": 60,
    "args": {
      "External id": 150281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150281, "pid": 5, "tid": 7, "ts": 1716454223762026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657887, "dur": 9,
    "args": {
      "External id": 150281,
      "cbid": 211, "correlation": 150281
    }
  },
  {
    "ph": "s", "id": 150281, "pid": 76337, "tid": -914061504, "ts": 1716454223657887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223762087, "dur": 50,
    "args": {
      "External id": 150289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150289,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150289, "pid": 5, "tid": 7, "ts": 1716454223762087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657918, "dur": 9,
    "args": {
      "External id": 150289,
      "cbid": 211, "correlation": 150289
    }
  },
  {
    "ph": "s", "id": 150289, "pid": 76337, "tid": -914061504, "ts": 1716454223657918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223762138, "dur": 98,
    "args": {
      "External id": 150298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150298,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150298, "pid": 5, "tid": 7, "ts": 1716454223762138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223657958, "dur": 10,
    "args": {
      "External id": 150298,
      "cbid": 211, "correlation": 150298
    }
  },
  {
    "ph": "s", "id": 150298, "pid": 76337, "tid": -914061504, "ts": 1716454223657958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223762237, "dur": 92,
    "args": {
      "External id": 150318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150318,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 150318, "pid": 5, "tid": 7, "ts": 1716454223762237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658039, "dur": 12,
    "args": {
      "External id": 150318,
      "cbid": 211, "correlation": 150318
    }
  },
  {
    "ph": "s", "id": 150318, "pid": 76337, "tid": -914061504, "ts": 1716454223658039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223762330, "dur": 5,
    "args": {
      "External id": 150330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150330,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 150330, "pid": 5, "tid": 7, "ts": 1716454223762330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658061, "dur": 7,
    "args": {
      "External id": 150330,
      "cbid": 211, "correlation": 150330
    }
  },
  {
    "ph": "s", "id": 150330, "pid": 76337, "tid": -914061504, "ts": 1716454223658061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223762337, "dur": 111,
    "args": {
      "External id": 150333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150333,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150333, "pid": 5, "tid": 7, "ts": 1716454223762337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658081, "dur": 6,
    "args": {
      "External id": 150333,
      "cbid": 211, "correlation": 150333
    }
  },
  {
    "ph": "s", "id": 150333, "pid": 76337, "tid": -914061504, "ts": 1716454223658081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223762449, "dur": 69,
    "args": {
      "External id": 150342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150342,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150342, "pid": 5, "tid": 7, "ts": 1716454223762449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658120, "dur": 10,
    "args": {
      "External id": 150342,
      "cbid": 211, "correlation": 150342
    }
  },
  {
    "ph": "s", "id": 150342, "pid": 76337, "tid": -914061504, "ts": 1716454223658120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658170, "dur": 0,
    "args": {
      "External id": 150352,
      "cbid": 317, "correlation": 150352
    }
  },
  {
    "ph": "f", "id": 150352, "pid": 76337, "tid": -914061504, "ts": 1716454223658170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658171, "dur": 0,
    "args": {
      "External id": 150353,
      "cbid": 203, "correlation": 150353
    }
  },
  {
    "ph": "f", "id": 150353, "pid": 76337, "tid": -914061504, "ts": 1716454223658171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658172, "dur": 0,
    "args": {
      "External id": 150354,
      "cbid": 205, "correlation": 150354
    }
  },
  {
    "ph": "f", "id": 150354, "pid": 76337, "tid": -914061504, "ts": 1716454223658172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223762519, "dur": 77,
    "args": {
      "External id": 150358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150358,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150358, "pid": 5, "tid": 7, "ts": 1716454223762519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658188, "dur": 11,
    "args": {
      "External id": 150358,
      "cbid": 211, "correlation": 150358
    }
  },
  {
    "ph": "s", "id": 150358, "pid": 76337, "tid": -914061504, "ts": 1716454223658188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223762598, "dur": 24,
    "args": {
      "External id": 150360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150360,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150360, "pid": 5, "tid": 7, "ts": 1716454223762598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658202, "dur": 5,
    "args": {
      "External id": 150360,
      "cbid": 211, "correlation": 150360
    }
  },
  {
    "ph": "s", "id": 150360, "pid": 76337, "tid": -914061504, "ts": 1716454223658202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223762623, "dur": 4,
    "args": {
      "External id": 150362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 150362, "pid": 5, "tid": 7, "ts": 1716454223762623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658213, "dur": 5,
    "args": {
      "External id": 150362,
      "cbid": 211, "correlation": 150362
    }
  },
  {
    "ph": "s", "id": 150362, "pid": 76337, "tid": -914061504, "ts": 1716454223658213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658222, "dur": 0,
    "args": {
      "External id": 150363,
      "cbid": 51, "correlation": 150363
    }
  },
  {
    "ph": "s", "id": 150363, "pid": 76337, "tid": -914061504, "ts": 1716454223658222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223762629, "dur": 1361,
    "args": {
      "External id": 150364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150364,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150364, "pid": 5, "tid": 7, "ts": 1716454223762629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658223, "dur": 5,
    "args": {
      "External id": 150364,
      "cbid": 211, "correlation": 150364
    }
  },
  {
    "ph": "s", "id": 150364, "pid": 76337, "tid": -914061504, "ts": 1716454223658223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223763991, "dur": 59,
    "args": {
      "External id": 150369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150369,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150369, "pid": 5, "tid": 7, "ts": 1716454223763991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658249, "dur": 8,
    "args": {
      "External id": 150369,
      "cbid": 211, "correlation": 150369
    }
  },
  {
    "ph": "s", "id": 150369, "pid": 76337, "tid": -914061504, "ts": 1716454223658249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223764051, "dur": 4,
    "args": {
      "External id": 150377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150377,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 150377, "pid": 5, "tid": 7, "ts": 1716454223764051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658293, "dur": 10,
    "args": {
      "External id": 150377,
      "cbid": 211, "correlation": 150377
    }
  },
  {
    "ph": "s", "id": 150377, "pid": 76337, "tid": -914061504, "ts": 1716454223658293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658358, "dur": 1,
    "args": {
      "External id": 150393,
      "cbid": 251, "correlation": 150393
    }
  },
  {
    "ph": "f", "id": 150393, "pid": 76337, "tid": -914061504, "ts": 1716454223658358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658363, "dur": 0,
    "args": {
      "External id": 150395,
      "cbid": 251, "correlation": 150395
    }
  },
  {
    "ph": "f", "id": 150395, "pid": 76337, "tid": -914061504, "ts": 1716454223658363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223764057, "dur": 11,
    "args": {
      "External id": 150396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150396,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 150396, "pid": 5, "tid": 7, "ts": 1716454223764057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658365, "dur": 12,
    "args": {
      "External id": 150396,
      "cbid": 211, "correlation": 150396
    }
  },
  {
    "ph": "s", "id": 150396, "pid": 76337, "tid": -914061504, "ts": 1716454223658365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223764069, "dur": 5,
    "args": {
      "External id": 150398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150398,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 150398, "pid": 5, "tid": 7, "ts": 1716454223764069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658379, "dur": 5,
    "args": {
      "External id": 150398,
      "cbid": 211, "correlation": 150398
    }
  },
  {
    "ph": "s", "id": 150398, "pid": 76337, "tid": -914061504, "ts": 1716454223658379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223764076, "dur": 55,
    "args": {
      "External id": 150408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150408,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150408, "pid": 5, "tid": 7, "ts": 1716454223764076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658437, "dur": 12,
    "args": {
      "External id": 150408,
      "cbid": 211, "correlation": 150408
    }
  },
  {
    "ph": "s", "id": 150408, "pid": 76337, "tid": -914061504, "ts": 1716454223658437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223764132, "dur": 52,
    "args": {
      "External id": 150428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150428,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 150428, "pid": 5, "tid": 7, "ts": 1716454223764132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658503, "dur": 10,
    "args": {
      "External id": 150428,
      "cbid": 211, "correlation": 150428
    }
  },
  {
    "ph": "s", "id": 150428, "pid": 76337, "tid": -914061504, "ts": 1716454223658503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223764186, "dur": 4,
    "args": {
      "External id": 150440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150440,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 150440, "pid": 5, "tid": 7, "ts": 1716454223764186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658523, "dur": 6,
    "args": {
      "External id": 150440,
      "cbid": 211, "correlation": 150440
    }
  },
  {
    "ph": "s", "id": 150440, "pid": 76337, "tid": -914061504, "ts": 1716454223658523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223764191, "dur": 55,
    "args": {
      "External id": 150443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150443, "pid": 5, "tid": 7, "ts": 1716454223764191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658541, "dur": 6,
    "args": {
      "External id": 150443,
      "cbid": 211, "correlation": 150443
    }
  },
  {
    "ph": "s", "id": 150443, "pid": 76337, "tid": -914061504, "ts": 1716454223658541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223764247, "dur": 36,
    "args": {
      "External id": 150452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150452,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150452, "pid": 5, "tid": 7, "ts": 1716454223764247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658581, "dur": 10,
    "args": {
      "External id": 150452,
      "cbid": 211, "correlation": 150452
    }
  },
  {
    "ph": "s", "id": 150452, "pid": 76337, "tid": -914061504, "ts": 1716454223658581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658642, "dur": 0,
    "args": {
      "External id": 150462,
      "cbid": 317, "correlation": 150462
    }
  },
  {
    "ph": "f", "id": 150462, "pid": 76337, "tid": -914061504, "ts": 1716454223658642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658643, "dur": 0,
    "args": {
      "External id": 150463,
      "cbid": 203, "correlation": 150463
    }
  },
  {
    "ph": "f", "id": 150463, "pid": 76337, "tid": -914061504, "ts": 1716454223658643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658644, "dur": 0,
    "args": {
      "External id": 150464,
      "cbid": 205, "correlation": 150464
    }
  },
  {
    "ph": "f", "id": 150464, "pid": 76337, "tid": -914061504, "ts": 1716454223658644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223764285, "dur": 40,
    "args": {
      "External id": 150468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150468,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150468, "pid": 5, "tid": 7, "ts": 1716454223764285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658660, "dur": 12,
    "args": {
      "External id": 150468,
      "cbid": 211, "correlation": 150468
    }
  },
  {
    "ph": "s", "id": 150468, "pid": 76337, "tid": -914061504, "ts": 1716454223658660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223764327, "dur": 14,
    "args": {
      "External id": 150470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150470,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150470, "pid": 5, "tid": 7, "ts": 1716454223764327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658675, "dur": 6,
    "args": {
      "External id": 150470,
      "cbid": 211, "correlation": 150470
    }
  },
  {
    "ph": "s", "id": 150470, "pid": 76337, "tid": -914061504, "ts": 1716454223658675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223764342, "dur": 3,
    "args": {
      "External id": 150472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150472,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 150472, "pid": 5, "tid": 7, "ts": 1716454223764342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658685, "dur": 5,
    "args": {
      "External id": 150472,
      "cbid": 211, "correlation": 150472
    }
  },
  {
    "ph": "s", "id": 150472, "pid": 76337, "tid": -914061504, "ts": 1716454223658685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658693, "dur": 0,
    "args": {
      "External id": 150473,
      "cbid": 51, "correlation": 150473
    }
  },
  {
    "ph": "s", "id": 150473, "pid": 76337, "tid": -914061504, "ts": 1716454223658693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223764347, "dur": 696,
    "args": {
      "External id": 150474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150474,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150474, "pid": 5, "tid": 7, "ts": 1716454223764347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658693, "dur": 5,
    "args": {
      "External id": 150474,
      "cbid": 211, "correlation": 150474
    }
  },
  {
    "ph": "s", "id": 150474, "pid": 76337, "tid": -914061504, "ts": 1716454223658693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223765045, "dur": 59,
    "args": {
      "External id": 150479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150479, "pid": 5, "tid": 7, "ts": 1716454223765045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658720, "dur": 9,
    "args": {
      "External id": 150479,
      "cbid": 211, "correlation": 150479
    }
  },
  {
    "ph": "s", "id": 150479, "pid": 76337, "tid": -914061504, "ts": 1716454223658720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658778, "dur": 0,
    "args": {
      "External id": 150489,
      "cbid": 317, "correlation": 150489
    }
  },
  {
    "ph": "f", "id": 150489, "pid": 76337, "tid": -914061504, "ts": 1716454223658778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658779, "dur": 0,
    "args": {
      "External id": 150490,
      "cbid": 203, "correlation": 150490
    }
  },
  {
    "ph": "f", "id": 150490, "pid": 76337, "tid": -914061504, "ts": 1716454223658779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658780, "dur": 0,
    "args": {
      "External id": 150491,
      "cbid": 205, "correlation": 150491
    }
  },
  {
    "ph": "f", "id": 150491, "pid": 76337, "tid": -914061504, "ts": 1716454223658780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223765106, "dur": 76,
    "args": {
      "External id": 150495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150495,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150495, "pid": 5, "tid": 7, "ts": 1716454223765106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658793, "dur": 11,
    "args": {
      "External id": 150495,
      "cbid": 211, "correlation": 150495
    }
  },
  {
    "ph": "s", "id": 150495, "pid": 76337, "tid": -914061504, "ts": 1716454223658793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223765183, "dur": 207,
    "args": {
      "External id": 150497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150497,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150497, "pid": 5, "tid": 7, "ts": 1716454223765183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658812, "dur": 8,
    "args": {
      "External id": 150497,
      "cbid": 211, "correlation": 150497
    }
  },
  {
    "ph": "s", "id": 150497, "pid": 76337, "tid": -914061504, "ts": 1716454223658812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223765392, "dur": 39,
    "args": {
      "External id": 150499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150499,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150499, "pid": 5, "tid": 7, "ts": 1716454223765392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658824, "dur": 6,
    "args": {
      "External id": 150499,
      "cbid": 211, "correlation": 150499
    }
  },
  {
    "ph": "s", "id": 150499, "pid": 76337, "tid": -914061504, "ts": 1716454223658824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223765433, "dur": 59,
    "args": {
      "External id": 150505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150505,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150505, "pid": 5, "tid": 7, "ts": 1716454223765433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658851, "dur": 9,
    "args": {
      "External id": 150505,
      "cbid": 211, "correlation": 150505
    }
  },
  {
    "ph": "s", "id": 150505, "pid": 76337, "tid": -914061504, "ts": 1716454223658851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223765493, "dur": 50,
    "args": {
      "External id": 150513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150513,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150513, "pid": 5, "tid": 7, "ts": 1716454223765493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658881, "dur": 8,
    "args": {
      "External id": 150513,
      "cbid": 211, "correlation": 150513
    }
  },
  {
    "ph": "s", "id": 150513, "pid": 76337, "tid": -914061504, "ts": 1716454223658881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223765545, "dur": 35,
    "args": {
      "External id": 150521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150521,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150521, "pid": 5, "tid": 7, "ts": 1716454223765545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223658910, "dur": 8,
    "args": {
      "External id": 150521,
      "cbid": 211, "correlation": 150521
    }
  },
  {
    "ph": "s", "id": 150521, "pid": 76337, "tid": -914061504, "ts": 1716454223658910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223765581, "dur": 52,
    "args": {
      "External id": 150541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150541,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 150541, "pid": 5, "tid": 7, "ts": 1716454223765581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659000, "dur": 12,
    "args": {
      "External id": 150541,
      "cbid": 211, "correlation": 150541
    }
  },
  {
    "ph": "s", "id": 150541, "pid": 76337, "tid": -914061504, "ts": 1716454223659000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223765635, "dur": 5,
    "args": {
      "External id": 150553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150553,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 150553, "pid": 5, "tid": 7, "ts": 1716454223765635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659022, "dur": 6,
    "args": {
      "External id": 150553,
      "cbid": 211, "correlation": 150553
    }
  },
  {
    "ph": "s", "id": 150553, "pid": 76337, "tid": -914061504, "ts": 1716454223659022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223765641, "dur": 55,
    "args": {
      "External id": 150556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150556, "pid": 5, "tid": 7, "ts": 1716454223765641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659041, "dur": 6,
    "args": {
      "External id": 150556,
      "cbid": 211, "correlation": 150556
    }
  },
  {
    "ph": "s", "id": 150556, "pid": 76337, "tid": -914061504, "ts": 1716454223659041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659098, "dur": 0,
    "args": {
      "External id": 150567,
      "cbid": 317, "correlation": 150567
    }
  },
  {
    "ph": "f", "id": 150567, "pid": 76337, "tid": -914061504, "ts": 1716454223659098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659099, "dur": 0,
    "args": {
      "External id": 150568,
      "cbid": 203, "correlation": 150568
    }
  },
  {
    "ph": "f", "id": 150568, "pid": 76337, "tid": -914061504, "ts": 1716454223659099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659100, "dur": 0,
    "args": {
      "External id": 150569,
      "cbid": 205, "correlation": 150569
    }
  },
  {
    "ph": "f", "id": 150569, "pid": 76337, "tid": -914061504, "ts": 1716454223659100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659122, "dur": 1,
    "args": {
      "External id": 150573,
      "cbid": 251, "correlation": 150573
    }
  },
  {
    "ph": "f", "id": 150573, "pid": 76337, "tid": -914061504, "ts": 1716454223659122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659124, "dur": 0,
    "args": {
      "External id": 150574,
      "cbid": 251, "correlation": 150574
    }
  },
  {
    "ph": "f", "id": 150574, "pid": 76337, "tid": -914061504, "ts": 1716454223659124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659124, "dur": 0,
    "args": {
      "External id": 150575,
      "cbid": 251, "correlation": 150575
    }
  },
  {
    "ph": "f", "id": 150575, "pid": 76337, "tid": -914061504, "ts": 1716454223659124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659125, "dur": 0,
    "args": {
      "External id": 150576,
      "cbid": 251, "correlation": 150576
    }
  },
  {
    "ph": "f", "id": 150576, "pid": 76337, "tid": -914061504, "ts": 1716454223659125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659126, "dur": 0,
    "args": {
      "External id": 150577,
      "cbid": 251, "correlation": 150577
    }
  },
  {
    "ph": "f", "id": 150577, "pid": 76337, "tid": -914061504, "ts": 1716454223659126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659127, "dur": 0,
    "args": {
      "External id": 150578,
      "cbid": 251, "correlation": 150578
    }
  },
  {
    "ph": "f", "id": 150578, "pid": 76337, "tid": -914061504, "ts": 1716454223659127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659128, "dur": 0,
    "args": {
      "External id": 150579,
      "cbid": 251, "correlation": 150579
    }
  },
  {
    "ph": "f", "id": 150579, "pid": 76337, "tid": -914061504, "ts": 1716454223659128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659129, "dur": 0,
    "args": {
      "External id": 150580,
      "cbid": 251, "correlation": 150580
    }
  },
  {
    "ph": "f", "id": 150580, "pid": 76337, "tid": -914061504, "ts": 1716454223659129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659130, "dur": 0,
    "args": {
      "External id": 150581,
      "cbid": 251, "correlation": 150581
    }
  },
  {
    "ph": "f", "id": 150581, "pid": 76337, "tid": -914061504, "ts": 1716454223659130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223765697, "dur": 114,
    "args": {
      "External id": 150582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150582,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150582, "pid": 5, "tid": 7, "ts": 1716454223765697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659133, "dur": 13,
    "args": {
      "External id": 150582,
      "cbid": 211, "correlation": 150582
    }
  },
  {
    "ph": "s", "id": 150582, "pid": 76337, "tid": -914061504, "ts": 1716454223659133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223765813, "dur": 60,
    "args": {
      "External id": 150588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150588,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150588, "pid": 5, "tid": 7, "ts": 1716454223765813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659168, "dur": 9,
    "args": {
      "External id": 150588,
      "cbid": 211, "correlation": 150588
    }
  },
  {
    "ph": "s", "id": 150588, "pid": 76337, "tid": -914061504, "ts": 1716454223659168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223765874, "dur": 506,
    "args": {
      "External id": 150597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150597,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150597, "pid": 5, "tid": 7, "ts": 1716454223765874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659251, "dur": 13,
    "args": {
      "External id": 150597,
      "cbid": 211, "correlation": 150597
    }
  },
  {
    "ph": "s", "id": 150597, "pid": 76337, "tid": -914061504, "ts": 1716454223659251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223766380, "dur": 180,
    "args": {
      "External id": 150619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150619,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150619, "pid": 5, "tid": 7, "ts": 1716454223766380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659309, "dur": 9,
    "args": {
      "External id": 150619,
      "cbid": 211, "correlation": 150619
    }
  },
  {
    "ph": "s", "id": 150619, "pid": 76337, "tid": -914061504, "ts": 1716454223659309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659397, "dur": 1,
    "args": {
      "External id": 150630,
      "cbid": 251, "correlation": 150630
    }
  },
  {
    "ph": "f", "id": 150630, "pid": 76337, "tid": -914061504, "ts": 1716454223659397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223766562, "dur": 198,
    "args": {
      "External id": 150631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150631,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150631, "pid": 5, "tid": 7, "ts": 1716454223766562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659402, "dur": 12,
    "args": {
      "External id": 150631,
      "cbid": 211, "correlation": 150631
    }
  },
  {
    "ph": "s", "id": 150631, "pid": 76337, "tid": -914061504, "ts": 1716454223659402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659469, "dur": 1,
    "args": {
      "External id": 150642,
      "cbid": 251, "correlation": 150642
    }
  },
  {
    "ph": "f", "id": 150642, "pid": 76337, "tid": -914061504, "ts": 1716454223659469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223766761, "dur": 189,
    "args": {
      "External id": 150643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150643,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150643, "pid": 5, "tid": 7, "ts": 1716454223766761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659473, "dur": 12,
    "args": {
      "External id": 150643,
      "cbid": 211, "correlation": 150643
    }
  },
  {
    "ph": "s", "id": 150643, "pid": 76337, "tid": -914061504, "ts": 1716454223659473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659537, "dur": 1,
    "args": {
      "External id": 150654,
      "cbid": 251, "correlation": 150654
    }
  },
  {
    "ph": "f", "id": 150654, "pid": 76337, "tid": -914061504, "ts": 1716454223659537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223766951, "dur": 192,
    "args": {
      "External id": 150655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150655,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150655, "pid": 5, "tid": 7, "ts": 1716454223766951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659541, "dur": 11,
    "args": {
      "External id": 150655,
      "cbid": 211, "correlation": 150655
    }
  },
  {
    "ph": "s", "id": 150655, "pid": 76337, "tid": -914061504, "ts": 1716454223659541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223767145, "dur": 18618,
    "args": {
      "External id": 150676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150676,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150676, "pid": 5, "tid": 7, "ts": 1716454223767145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659621, "dur": 12,
    "args": {
      "External id": 150676,
      "cbid": 211, "correlation": 150676
    }
  },
  {
    "ph": "s", "id": 150676, "pid": 76337, "tid": -914061504, "ts": 1716454223659621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659717, "dur": 1,
    "args": {
      "External id": 150694,
      "cbid": 251, "correlation": 150694
    }
  },
  {
    "ph": "f", "id": 150694, "pid": 76337, "tid": -914061504, "ts": 1716454223659717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223785764, "dur": 199,
    "args": {
      "External id": 150696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150696,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150696, "pid": 5, "tid": 7, "ts": 1716454223785764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659723, "dur": 14,
    "args": {
      "External id": 150696,
      "cbid": 211, "correlation": 150696
    }
  },
  {
    "ph": "s", "id": 150696, "pid": 76337, "tid": -914061504, "ts": 1716454223659723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223785965, "dur": 66,
    "args": {
      "External id": 150704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150704,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150704, "pid": 5, "tid": 7, "ts": 1716454223785965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659795, "dur": 12,
    "args": {
      "External id": 150704,
      "cbid": 211, "correlation": 150704
    }
  },
  {
    "ph": "s", "id": 150704, "pid": 76337, "tid": -914061504, "ts": 1716454223659795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223786033, "dur": 96,
    "args": {
      "External id": 150712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150712,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150712, "pid": 5, "tid": 7, "ts": 1716454223786033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659833, "dur": 9,
    "args": {
      "External id": 150712,
      "cbid": 211, "correlation": 150712
    }
  },
  {
    "ph": "s", "id": 150712, "pid": 76337, "tid": -914061504, "ts": 1716454223659833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223786130, "dur": 55,
    "args": {
      "External id": 150723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150723,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150723, "pid": 5, "tid": 7, "ts": 1716454223786130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659905, "dur": 12,
    "args": {
      "External id": 150723,
      "cbid": 211, "correlation": 150723
    }
  },
  {
    "ph": "s", "id": 150723, "pid": 76337, "tid": -914061504, "ts": 1716454223659905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223786186, "dur": 92,
    "args": {
      "External id": 150745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150745,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150745, "pid": 5, "tid": 7, "ts": 1716454223786186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223659935, "dur": 8,
    "args": {
      "External id": 150745,
      "cbid": 211, "correlation": 150745
    }
  },
  {
    "ph": "s", "id": 150745, "pid": 76337, "tid": -914061504, "ts": 1716454223659935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660028, "dur": 1,
    "args": {
      "External id": 150756,
      "cbid": 251, "correlation": 150756
    }
  },
  {
    "ph": "f", "id": 150756, "pid": 76337, "tid": -914061504, "ts": 1716454223660028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223786279, "dur": 105,
    "args": {
      "External id": 150757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150757,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150757, "pid": 5, "tid": 7, "ts": 1716454223786279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660033, "dur": 14,
    "args": {
      "External id": 150757,
      "cbid": 211, "correlation": 150757
    }
  },
  {
    "ph": "s", "id": 150757, "pid": 76337, "tid": -914061504, "ts": 1716454223660033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660104, "dur": 1,
    "args": {
      "External id": 150768,
      "cbid": 251, "correlation": 150768
    }
  },
  {
    "ph": "f", "id": 150768, "pid": 76337, "tid": -914061504, "ts": 1716454223660104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660107, "dur": 0,
    "args": {
      "External id": 150769,
      "cbid": 251, "correlation": 150769
    }
  },
  {
    "ph": "f", "id": 150769, "pid": 76337, "tid": -914061504, "ts": 1716454223660107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223786386, "dur": 11,
    "args": {
      "External id": 150770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150770,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 150770, "pid": 5, "tid": 7, "ts": 1716454223786386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660109, "dur": 12,
    "args": {
      "External id": 150770,
      "cbid": 211, "correlation": 150770
    }
  },
  {
    "ph": "s", "id": 150770, "pid": 76337, "tid": -914061504, "ts": 1716454223660109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223786398, "dur": 5,
    "args": {
      "External id": 150772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150772,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 150772, "pid": 5, "tid": 7, "ts": 1716454223786398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660123, "dur": 6,
    "args": {
      "External id": 150772,
      "cbid": 211, "correlation": 150772
    }
  },
  {
    "ph": "s", "id": 150772, "pid": 76337, "tid": -914061504, "ts": 1716454223660123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660184, "dur": 1,
    "args": {
      "External id": 150783,
      "cbid": 251, "correlation": 150783
    }
  },
  {
    "ph": "f", "id": 150783, "pid": 76337, "tid": -914061504, "ts": 1716454223660184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660187, "dur": 0,
    "args": {
      "External id": 150784,
      "cbid": 251, "correlation": 150784
    }
  },
  {
    "ph": "f", "id": 150784, "pid": 76337, "tid": -914061504, "ts": 1716454223660187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223786404, "dur": 6,
    "args": {
      "External id": 150785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150785,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 150785, "pid": 5, "tid": 7, "ts": 1716454223786404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660189, "dur": 12,
    "args": {
      "External id": 150785,
      "cbid": 211, "correlation": 150785
    }
  },
  {
    "ph": "s", "id": 150785, "pid": 76337, "tid": -914061504, "ts": 1716454223660189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223786411, "dur": 3,
    "args": {
      "External id": 150787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150787,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 150787, "pid": 5, "tid": 7, "ts": 1716454223786411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660202, "dur": 7,
    "args": {
      "External id": 150787,
      "cbid": 211, "correlation": 150787
    }
  },
  {
    "ph": "s", "id": 150787, "pid": 76337, "tid": -914061504, "ts": 1716454223660202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223786416, "dur": 155,
    "args": {
      "External id": 150808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150808,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150808, "pid": 5, "tid": 7, "ts": 1716454223786416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660277, "dur": 13,
    "args": {
      "External id": 150808,
      "cbid": 211, "correlation": 150808
    }
  },
  {
    "ph": "s", "id": 150808, "pid": 76337, "tid": -914061504, "ts": 1716454223660277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660374, "dur": 1,
    "args": {
      "External id": 150826,
      "cbid": 251, "correlation": 150826
    }
  },
  {
    "ph": "f", "id": 150826, "pid": 76337, "tid": -914061504, "ts": 1716454223660374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223786572, "dur": 107,
    "args": {
      "External id": 150828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150828,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150828, "pid": 5, "tid": 7, "ts": 1716454223786572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660380, "dur": 14,
    "args": {
      "External id": 150828,
      "cbid": 211, "correlation": 150828
    }
  },
  {
    "ph": "s", "id": 150828, "pid": 76337, "tid": -914061504, "ts": 1716454223660380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223786680, "dur": 35,
    "args": {
      "External id": 150836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150836,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150836, "pid": 5, "tid": 7, "ts": 1716454223786680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660450, "dur": 12,
    "args": {
      "External id": 150836,
      "cbid": 211, "correlation": 150836
    }
  },
  {
    "ph": "s", "id": 150836, "pid": 76337, "tid": -914061504, "ts": 1716454223660450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223786717, "dur": 67,
    "args": {
      "External id": 150844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150844, "pid": 5, "tid": 7, "ts": 1716454223786717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660490, "dur": 9,
    "args": {
      "External id": 150844,
      "cbid": 211, "correlation": 150844
    }
  },
  {
    "ph": "s", "id": 150844, "pid": 76337, "tid": -914061504, "ts": 1716454223660490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223786785, "dur": 92,
    "args": {
      "External id": 150866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150866,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150866, "pid": 5, "tid": 7, "ts": 1716454223786785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660542, "dur": 10,
    "args": {
      "External id": 150866,
      "cbid": 211, "correlation": 150866
    }
  },
  {
    "ph": "s", "id": 150866, "pid": 76337, "tid": -914061504, "ts": 1716454223660542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660629, "dur": 1,
    "args": {
      "External id": 150882,
      "cbid": 251, "correlation": 150882
    }
  },
  {
    "ph": "f", "id": 150882, "pid": 76337, "tid": -914061504, "ts": 1716454223660629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223786878, "dur": 569,
    "args": {
      "External id": 150884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150884,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 150884, "pid": 5, "tid": 7, "ts": 1716454223786878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660635, "dur": 13,
    "args": {
      "External id": 150884,
      "cbid": 211, "correlation": 150884
    }
  },
  {
    "ph": "s", "id": 150884, "pid": 76337, "tid": -914061504, "ts": 1716454223660635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223787449, "dur": 243,
    "args": {
      "External id": 150892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150892, "pid": 5, "tid": 7, "ts": 1716454223787449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660700, "dur": 13,
    "args": {
      "External id": 150892,
      "cbid": 211, "correlation": 150892
    }
  },
  {
    "ph": "s", "id": 150892, "pid": 76337, "tid": -914061504, "ts": 1716454223660700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223787694, "dur": 252,
    "args": {
      "External id": 150900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150900, "pid": 5, "tid": 7, "ts": 1716454223787694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660732, "dur": 9,
    "args": {
      "External id": 150900,
      "cbid": 211, "correlation": 150900
    }
  },
  {
    "ph": "s", "id": 150900, "pid": 76337, "tid": -914061504, "ts": 1716454223660732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660814, "dur": 1,
    "args": {
      "External id": 150916,
      "cbid": 251, "correlation": 150916
    }
  },
  {
    "ph": "f", "id": 150916, "pid": 76337, "tid": -914061504, "ts": 1716454223660814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660820, "dur": 0,
    "args": {
      "External id": 150918,
      "cbid": 251, "correlation": 150918
    }
  },
  {
    "ph": "f", "id": 150918, "pid": 76337, "tid": -914061504, "ts": 1716454223660820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223787947, "dur": 357,
    "args": {
      "External id": 150919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150919,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 150919, "pid": 5, "tid": 7, "ts": 1716454223787947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660822, "dur": 13,
    "args": {
      "External id": 150919,
      "cbid": 211, "correlation": 150919
    }
  },
  {
    "ph": "s", "id": 150919, "pid": 76337, "tid": -914061504, "ts": 1716454223660822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223788306, "dur": 50,
    "args": {
      "External id": 150927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150927,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150927, "pid": 5, "tid": 7, "ts": 1716454223788306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660865, "dur": 10,
    "args": {
      "External id": 150927,
      "cbid": 211, "correlation": 150927
    }
  },
  {
    "ph": "s", "id": 150927, "pid": 76337, "tid": -914061504, "ts": 1716454223660865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223788357, "dur": 158,
    "args": {
      "External id": 150938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150938,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150938, "pid": 5, "tid": 7, "ts": 1716454223788357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223660932, "dur": 14,
    "args": {
      "External id": 150938,
      "cbid": 211, "correlation": 150938
    }
  },
  {
    "ph": "s", "id": 150938, "pid": 76337, "tid": -914061504, "ts": 1716454223660932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661009, "dur": 0,
    "args": {
      "External id": 150950,
      "cbid": 317, "correlation": 150950
    }
  },
  {
    "ph": "f", "id": 150950, "pid": 76337, "tid": -914061504, "ts": 1716454223661009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661010, "dur": 0,
    "args": {
      "External id": 150951,
      "cbid": 203, "correlation": 150951
    }
  },
  {
    "ph": "f", "id": 150951, "pid": 76337, "tid": -914061504, "ts": 1716454223661010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661011, "dur": 0,
    "args": {
      "External id": 150952,
      "cbid": 205, "correlation": 150952
    }
  },
  {
    "ph": "f", "id": 150952, "pid": 76337, "tid": -914061504, "ts": 1716454223661011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661035, "dur": 1,
    "args": {
      "External id": 150956,
      "cbid": 251, "correlation": 150956
    }
  },
  {
    "ph": "f", "id": 150956, "pid": 76337, "tid": -914061504, "ts": 1716454223661035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661037, "dur": 0,
    "args": {
      "External id": 150957,
      "cbid": 251, "correlation": 150957
    }
  },
  {
    "ph": "f", "id": 150957, "pid": 76337, "tid": -914061504, "ts": 1716454223661037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661038, "dur": 0,
    "args": {
      "External id": 150958,
      "cbid": 251, "correlation": 150958
    }
  },
  {
    "ph": "f", "id": 150958, "pid": 76337, "tid": -914061504, "ts": 1716454223661038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661039, "dur": 0,
    "args": {
      "External id": 150959,
      "cbid": 251, "correlation": 150959
    }
  },
  {
    "ph": "f", "id": 150959, "pid": 76337, "tid": -914061504, "ts": 1716454223661039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661040, "dur": 0,
    "args": {
      "External id": 150960,
      "cbid": 251, "correlation": 150960
    }
  },
  {
    "ph": "f", "id": 150960, "pid": 76337, "tid": -914061504, "ts": 1716454223661040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661040, "dur": 0,
    "args": {
      "External id": 150961,
      "cbid": 251, "correlation": 150961
    }
  },
  {
    "ph": "f", "id": 150961, "pid": 76337, "tid": -914061504, "ts": 1716454223661040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661041, "dur": 0,
    "args": {
      "External id": 150962,
      "cbid": 251, "correlation": 150962
    }
  },
  {
    "ph": "f", "id": 150962, "pid": 76337, "tid": -914061504, "ts": 1716454223661041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661042, "dur": 0,
    "args": {
      "External id": 150963,
      "cbid": 251, "correlation": 150963
    }
  },
  {
    "ph": "f", "id": 150963, "pid": 76337, "tid": -914061504, "ts": 1716454223661042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661044, "dur": 0,
    "args": {
      "External id": 150964,
      "cbid": 251, "correlation": 150964
    }
  },
  {
    "ph": "f", "id": 150964, "pid": 76337, "tid": -914061504, "ts": 1716454223661044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223788517, "dur": 115,
    "args": {
      "External id": 150965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150965,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 150965, "pid": 5, "tid": 7, "ts": 1716454223788517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661046, "dur": 12,
    "args": {
      "External id": 150965,
      "cbid": 211, "correlation": 150965
    }
  },
  {
    "ph": "s", "id": 150965, "pid": 76337, "tid": -914061504, "ts": 1716454223661046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223788633, "dur": 59,
    "args": {
      "External id": 150971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150971,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150971, "pid": 5, "tid": 7, "ts": 1716454223788633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661081, "dur": 9,
    "args": {
      "External id": 150971,
      "cbid": 211, "correlation": 150971
    }
  },
  {
    "ph": "s", "id": 150971, "pid": 76337, "tid": -914061504, "ts": 1716454223661081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223788693, "dur": 50,
    "args": {
      "External id": 150979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150979,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150979, "pid": 5, "tid": 7, "ts": 1716454223788693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661114, "dur": 8,
    "args": {
      "External id": 150979,
      "cbid": 211, "correlation": 150979
    }
  },
  {
    "ph": "s", "id": 150979, "pid": 76337, "tid": -914061504, "ts": 1716454223661114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223788744, "dur": 98,
    "args": {
      "External id": 150988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 150988,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 150988, "pid": 5, "tid": 7, "ts": 1716454223788744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661152, "dur": 10,
    "args": {
      "External id": 150988,
      "cbid": 211, "correlation": 150988
    }
  },
  {
    "ph": "s", "id": 150988, "pid": 76337, "tid": -914061504, "ts": 1716454223661152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223788843, "dur": 92,
    "args": {
      "External id": 151008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151008,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 151008, "pid": 5, "tid": 7, "ts": 1716454223788843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661223, "dur": 11,
    "args": {
      "External id": 151008,
      "cbid": 211, "correlation": 151008
    }
  },
  {
    "ph": "s", "id": 151008, "pid": 76337, "tid": -914061504, "ts": 1716454223661223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223788937, "dur": 5,
    "args": {
      "External id": 151020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151020,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 151020, "pid": 5, "tid": 7, "ts": 1716454223788937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661244, "dur": 7,
    "args": {
      "External id": 151020,
      "cbid": 211, "correlation": 151020
    }
  },
  {
    "ph": "s", "id": 151020, "pid": 76337, "tid": -914061504, "ts": 1716454223661244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223788943, "dur": 109,
    "args": {
      "External id": 151023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151023, "pid": 5, "tid": 7, "ts": 1716454223788943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661263, "dur": 6,
    "args": {
      "External id": 151023,
      "cbid": 211, "correlation": 151023
    }
  },
  {
    "ph": "s", "id": 151023, "pid": 76337, "tid": -914061504, "ts": 1716454223661263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223789053, "dur": 69,
    "args": {
      "External id": 151032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151032,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151032, "pid": 5, "tid": 7, "ts": 1716454223789053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661303, "dur": 10,
    "args": {
      "External id": 151032,
      "cbid": 211, "correlation": 151032
    }
  },
  {
    "ph": "s", "id": 151032, "pid": 76337, "tid": -914061504, "ts": 1716454223661303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661354, "dur": 0,
    "args": {
      "External id": 151042,
      "cbid": 317, "correlation": 151042
    }
  },
  {
    "ph": "f", "id": 151042, "pid": 76337, "tid": -914061504, "ts": 1716454223661354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661355, "dur": 0,
    "args": {
      "External id": 151043,
      "cbid": 203, "correlation": 151043
    }
  },
  {
    "ph": "f", "id": 151043, "pid": 76337, "tid": -914061504, "ts": 1716454223661355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661356, "dur": 0,
    "args": {
      "External id": 151044,
      "cbid": 205, "correlation": 151044
    }
  },
  {
    "ph": "f", "id": 151044, "pid": 76337, "tid": -914061504, "ts": 1716454223661356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223789123, "dur": 77,
    "args": {
      "External id": 151048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151048, "pid": 5, "tid": 7, "ts": 1716454223789123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661370, "dur": 11,
    "args": {
      "External id": 151048,
      "cbid": 211, "correlation": 151048
    }
  },
  {
    "ph": "s", "id": 151048, "pid": 76337, "tid": -914061504, "ts": 1716454223661370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223789201, "dur": 24,
    "args": {
      "External id": 151050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151050,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151050, "pid": 5, "tid": 7, "ts": 1716454223789201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661384, "dur": 5,
    "args": {
      "External id": 151050,
      "cbid": 211, "correlation": 151050
    }
  },
  {
    "ph": "s", "id": 151050, "pid": 76337, "tid": -914061504, "ts": 1716454223661384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223789227, "dur": 4,
    "args": {
      "External id": 151052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151052, "pid": 5, "tid": 7, "ts": 1716454223789227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661394, "dur": 6,
    "args": {
      "External id": 151052,
      "cbid": 211, "correlation": 151052
    }
  },
  {
    "ph": "s", "id": 151052, "pid": 76337, "tid": -914061504, "ts": 1716454223661394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661403, "dur": 0,
    "args": {
      "External id": 151053,
      "cbid": 51, "correlation": 151053
    }
  },
  {
    "ph": "s", "id": 151053, "pid": 76337, "tid": -914061504, "ts": 1716454223661403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223789232, "dur": 1357,
    "args": {
      "External id": 151054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151054,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151054, "pid": 5, "tid": 7, "ts": 1716454223789232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661404, "dur": 5,
    "args": {
      "External id": 151054,
      "cbid": 211, "correlation": 151054
    }
  },
  {
    "ph": "s", "id": 151054, "pid": 76337, "tid": -914061504, "ts": 1716454223661404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223790590, "dur": 59,
    "args": {
      "External id": 151059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151059, "pid": 5, "tid": 7, "ts": 1716454223790590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661430, "dur": 9,
    "args": {
      "External id": 151059,
      "cbid": 211, "correlation": 151059
    }
  },
  {
    "ph": "s", "id": 151059, "pid": 76337, "tid": -914061504, "ts": 1716454223661430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223790651, "dur": 4,
    "args": {
      "External id": 151067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151067,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151067, "pid": 5, "tid": 7, "ts": 1716454223790651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661475, "dur": 9,
    "args": {
      "External id": 151067,
      "cbid": 211, "correlation": 151067
    }
  },
  {
    "ph": "s", "id": 151067, "pid": 76337, "tid": -914061504, "ts": 1716454223661475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661540, "dur": 1,
    "args": {
      "External id": 151083,
      "cbid": 251, "correlation": 151083
    }
  },
  {
    "ph": "f", "id": 151083, "pid": 76337, "tid": -914061504, "ts": 1716454223661540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661546, "dur": 0,
    "args": {
      "External id": 151085,
      "cbid": 251, "correlation": 151085
    }
  },
  {
    "ph": "f", "id": 151085, "pid": 76337, "tid": -914061504, "ts": 1716454223661546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223790656, "dur": 11,
    "args": {
      "External id": 151086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151086,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 151086, "pid": 5, "tid": 7, "ts": 1716454223790656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661548, "dur": 11,
    "args": {
      "External id": 151086,
      "cbid": 211, "correlation": 151086
    }
  },
  {
    "ph": "s", "id": 151086, "pid": 76337, "tid": -914061504, "ts": 1716454223661548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223790668, "dur": 5,
    "args": {
      "External id": 151088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151088,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 151088, "pid": 5, "tid": 7, "ts": 1716454223790668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661561, "dur": 5,
    "args": {
      "External id": 151088,
      "cbid": 211, "correlation": 151088
    }
  },
  {
    "ph": "s", "id": 151088, "pid": 76337, "tid": -914061504, "ts": 1716454223661561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223790675, "dur": 53,
    "args": {
      "External id": 151098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151098,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151098, "pid": 5, "tid": 7, "ts": 1716454223790675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661619, "dur": 113,
    "args": {
      "External id": 151098,
      "cbid": 211, "correlation": 151098
    }
  },
  {
    "ph": "s", "id": 151098, "pid": 76337, "tid": -914061504, "ts": 1716454223661619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223790730, "dur": 50,
    "args": {
      "External id": 151118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151118,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 151118, "pid": 5, "tid": 7, "ts": 1716454223790730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661786, "dur": 11,
    "args": {
      "External id": 151118,
      "cbid": 211, "correlation": 151118
    }
  },
  {
    "ph": "s", "id": 151118, "pid": 76337, "tid": -914061504, "ts": 1716454223661786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223790781, "dur": 4,
    "args": {
      "External id": 151130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151130,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151130, "pid": 5, "tid": 7, "ts": 1716454223790781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661806, "dur": 6,
    "args": {
      "External id": 151130,
      "cbid": 211, "correlation": 151130
    }
  },
  {
    "ph": "s", "id": 151130, "pid": 76337, "tid": -914061504, "ts": 1716454223661806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223790786, "dur": 56,
    "args": {
      "External id": 151133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151133,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151133, "pid": 5, "tid": 7, "ts": 1716454223790786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661825, "dur": 7,
    "args": {
      "External id": 151133,
      "cbid": 211, "correlation": 151133
    }
  },
  {
    "ph": "s", "id": 151133, "pid": 76337, "tid": -914061504, "ts": 1716454223661825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223790843, "dur": 37,
    "args": {
      "External id": 151142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151142,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151142, "pid": 5, "tid": 7, "ts": 1716454223790843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661865, "dur": 10,
    "args": {
      "External id": 151142,
      "cbid": 211, "correlation": 151142
    }
  },
  {
    "ph": "s", "id": 151142, "pid": 76337, "tid": -914061504, "ts": 1716454223661865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661929, "dur": 0,
    "args": {
      "External id": 151152,
      "cbid": 317, "correlation": 151152
    }
  },
  {
    "ph": "f", "id": 151152, "pid": 76337, "tid": -914061504, "ts": 1716454223661929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661930, "dur": 0,
    "args": {
      "External id": 151153,
      "cbid": 203, "correlation": 151153
    }
  },
  {
    "ph": "f", "id": 151153, "pid": 76337, "tid": -914061504, "ts": 1716454223661930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661931, "dur": 0,
    "args": {
      "External id": 151154,
      "cbid": 205, "correlation": 151154
    }
  },
  {
    "ph": "f", "id": 151154, "pid": 76337, "tid": -914061504, "ts": 1716454223661931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223790881, "dur": 40,
    "args": {
      "External id": 151158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151158,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151158, "pid": 5, "tid": 7, "ts": 1716454223790881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661945, "dur": 12,
    "args": {
      "External id": 151158,
      "cbid": 211, "correlation": 151158
    }
  },
  {
    "ph": "s", "id": 151158, "pid": 76337, "tid": -914061504, "ts": 1716454223661945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223790923, "dur": 14,
    "args": {
      "External id": 151160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151160,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151160, "pid": 5, "tid": 7, "ts": 1716454223790923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661959, "dur": 5,
    "args": {
      "External id": 151160,
      "cbid": 211, "correlation": 151160
    }
  },
  {
    "ph": "s", "id": 151160, "pid": 76337, "tid": -914061504, "ts": 1716454223661959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223790939, "dur": 4,
    "args": {
      "External id": 151162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151162, "pid": 5, "tid": 7, "ts": 1716454223790939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661969, "dur": 15,
    "args": {
      "External id": 151162,
      "cbid": 211, "correlation": 151162
    }
  },
  {
    "ph": "s", "id": 151162, "pid": 76337, "tid": -914061504, "ts": 1716454223661969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661987, "dur": 0,
    "args": {
      "External id": 151163,
      "cbid": 51, "correlation": 151163
    }
  },
  {
    "ph": "s", "id": 151163, "pid": 76337, "tid": -914061504, "ts": 1716454223661987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223790944, "dur": 694,
    "args": {
      "External id": 151164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151164,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151164, "pid": 5, "tid": 7, "ts": 1716454223790944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223661988, "dur": 5,
    "args": {
      "External id": 151164,
      "cbid": 211, "correlation": 151164
    }
  },
  {
    "ph": "s", "id": 151164, "pid": 76337, "tid": -914061504, "ts": 1716454223661988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223791639, "dur": 59,
    "args": {
      "External id": 151169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151169,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151169, "pid": 5, "tid": 7, "ts": 1716454223791639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662017, "dur": 9,
    "args": {
      "External id": 151169,
      "cbid": 211, "correlation": 151169
    }
  },
  {
    "ph": "s", "id": 151169, "pid": 76337, "tid": -914061504, "ts": 1716454223662017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662075, "dur": 0,
    "args": {
      "External id": 151179,
      "cbid": 317, "correlation": 151179
    }
  },
  {
    "ph": "f", "id": 151179, "pid": 76337, "tid": -914061504, "ts": 1716454223662075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662076, "dur": 0,
    "args": {
      "External id": 151180,
      "cbid": 203, "correlation": 151180
    }
  },
  {
    "ph": "f", "id": 151180, "pid": 76337, "tid": -914061504, "ts": 1716454223662076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662076, "dur": 0,
    "args": {
      "External id": 151181,
      "cbid": 205, "correlation": 151181
    }
  },
  {
    "ph": "f", "id": 151181, "pid": 76337, "tid": -914061504, "ts": 1716454223662076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223791699, "dur": 75,
    "args": {
      "External id": 151185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151185,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151185, "pid": 5, "tid": 7, "ts": 1716454223791699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662088, "dur": 12,
    "args": {
      "External id": 151185,
      "cbid": 211, "correlation": 151185
    }
  },
  {
    "ph": "s", "id": 151185, "pid": 76337, "tid": -914061504, "ts": 1716454223662088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223791776, "dur": 205,
    "args": {
      "External id": 151187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151187,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151187, "pid": 5, "tid": 7, "ts": 1716454223791776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662106, "dur": 7,
    "args": {
      "External id": 151187,
      "cbid": 211, "correlation": 151187
    }
  },
  {
    "ph": "s", "id": 151187, "pid": 76337, "tid": -914061504, "ts": 1716454223662106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223791983, "dur": 39,
    "args": {
      "External id": 151189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151189,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151189, "pid": 5, "tid": 7, "ts": 1716454223791983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662118, "dur": 5,
    "args": {
      "External id": 151189,
      "cbid": 211, "correlation": 151189
    }
  },
  {
    "ph": "s", "id": 151189, "pid": 76337, "tid": -914061504, "ts": 1716454223662118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223792023, "dur": 59,
    "args": {
      "External id": 151195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151195,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151195, "pid": 5, "tid": 7, "ts": 1716454223792023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662143, "dur": 503,
    "args": {
      "External id": 151195,
      "cbid": 211, "correlation": 151195
    }
  },
  {
    "ph": "s", "id": 151195, "pid": 76337, "tid": -914061504, "ts": 1716454223662143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223792084, "dur": 50,
    "args": {
      "External id": 151203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151203,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151203, "pid": 5, "tid": 7, "ts": 1716454223792084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662667, "dur": 8,
    "args": {
      "External id": 151203,
      "cbid": 211, "correlation": 151203
    }
  },
  {
    "ph": "s", "id": 151203, "pid": 76337, "tid": -914061504, "ts": 1716454223662667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223792136, "dur": 35,
    "args": {
      "External id": 151211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151211,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151211, "pid": 5, "tid": 7, "ts": 1716454223792136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662696, "dur": 8,
    "args": {
      "External id": 151211,
      "cbid": 211, "correlation": 151211
    }
  },
  {
    "ph": "s", "id": 151211, "pid": 76337, "tid": -914061504, "ts": 1716454223662696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223792172, "dur": 52,
    "args": {
      "External id": 151231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151231,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 151231, "pid": 5, "tid": 7, "ts": 1716454223792172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662778, "dur": 13,
    "args": {
      "External id": 151231,
      "cbid": 211, "correlation": 151231
    }
  },
  {
    "ph": "s", "id": 151231, "pid": 76337, "tid": -914061504, "ts": 1716454223662778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223792226, "dur": 4,
    "args": {
      "External id": 151243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151243,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151243, "pid": 5, "tid": 7, "ts": 1716454223792226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662801, "dur": 6,
    "args": {
      "External id": 151243,
      "cbid": 211, "correlation": 151243
    }
  },
  {
    "ph": "s", "id": 151243, "pid": 76337, "tid": -914061504, "ts": 1716454223662801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223792231, "dur": 55,
    "args": {
      "External id": 151246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151246, "pid": 5, "tid": 7, "ts": 1716454223792231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662818, "dur": 6,
    "args": {
      "External id": 151246,
      "cbid": 211, "correlation": 151246
    }
  },
  {
    "ph": "s", "id": 151246, "pid": 76337, "tid": -914061504, "ts": 1716454223662818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662876, "dur": 0,
    "args": {
      "External id": 151257,
      "cbid": 317, "correlation": 151257
    }
  },
  {
    "ph": "f", "id": 151257, "pid": 76337, "tid": -914061504, "ts": 1716454223662876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662876, "dur": 0,
    "args": {
      "External id": 151258,
      "cbid": 203, "correlation": 151258
    }
  },
  {
    "ph": "f", "id": 151258, "pid": 76337, "tid": -914061504, "ts": 1716454223662876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662877, "dur": 0,
    "args": {
      "External id": 151259,
      "cbid": 205, "correlation": 151259
    }
  },
  {
    "ph": "f", "id": 151259, "pid": 76337, "tid": -914061504, "ts": 1716454223662877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662900, "dur": 1,
    "args": {
      "External id": 151263,
      "cbid": 251, "correlation": 151263
    }
  },
  {
    "ph": "f", "id": 151263, "pid": 76337, "tid": -914061504, "ts": 1716454223662900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662902, "dur": 0,
    "args": {
      "External id": 151264,
      "cbid": 251, "correlation": 151264
    }
  },
  {
    "ph": "f", "id": 151264, "pid": 76337, "tid": -914061504, "ts": 1716454223662902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662902, "dur": 0,
    "args": {
      "External id": 151265,
      "cbid": 251, "correlation": 151265
    }
  },
  {
    "ph": "f", "id": 151265, "pid": 76337, "tid": -914061504, "ts": 1716454223662902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662903, "dur": 0,
    "args": {
      "External id": 151266,
      "cbid": 251, "correlation": 151266
    }
  },
  {
    "ph": "f", "id": 151266, "pid": 76337, "tid": -914061504, "ts": 1716454223662903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662904, "dur": 0,
    "args": {
      "External id": 151267,
      "cbid": 251, "correlation": 151267
    }
  },
  {
    "ph": "f", "id": 151267, "pid": 76337, "tid": -914061504, "ts": 1716454223662904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662905, "dur": 0,
    "args": {
      "External id": 151268,
      "cbid": 251, "correlation": 151268
    }
  },
  {
    "ph": "f", "id": 151268, "pid": 76337, "tid": -914061504, "ts": 1716454223662905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662906, "dur": 0,
    "args": {
      "External id": 151269,
      "cbid": 251, "correlation": 151269
    }
  },
  {
    "ph": "f", "id": 151269, "pid": 76337, "tid": -914061504, "ts": 1716454223662906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662907, "dur": 0,
    "args": {
      "External id": 151270,
      "cbid": 251, "correlation": 151270
    }
  },
  {
    "ph": "f", "id": 151270, "pid": 76337, "tid": -914061504, "ts": 1716454223662907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662908, "dur": 0,
    "args": {
      "External id": 151271,
      "cbid": 251, "correlation": 151271
    }
  },
  {
    "ph": "f", "id": 151271, "pid": 76337, "tid": -914061504, "ts": 1716454223662908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223792287, "dur": 115,
    "args": {
      "External id": 151272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151272,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 151272, "pid": 5, "tid": 7, "ts": 1716454223792287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662910, "dur": 13,
    "args": {
      "External id": 151272,
      "cbid": 211, "correlation": 151272
    }
  },
  {
    "ph": "s", "id": 151272, "pid": 76337, "tid": -914061504, "ts": 1716454223662910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223792404, "dur": 60,
    "args": {
      "External id": 151278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151278, "pid": 5, "tid": 7, "ts": 1716454223792404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223662946, "dur": 9,
    "args": {
      "External id": 151278,
      "cbid": 211, "correlation": 151278
    }
  },
  {
    "ph": "s", "id": 151278, "pid": 76337, "tid": -914061504, "ts": 1716454223662946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223792465, "dur": 553,
    "args": {
      "External id": 151287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151287,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151287, "pid": 5, "tid": 7, "ts": 1716454223792465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663038, "dur": 14,
    "args": {
      "External id": 151287,
      "cbid": 211, "correlation": 151287
    }
  },
  {
    "ph": "s", "id": 151287, "pid": 76337, "tid": -914061504, "ts": 1716454223663038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223793019, "dur": 181,
    "args": {
      "External id": 151309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151309,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151309, "pid": 5, "tid": 7, "ts": 1716454223793019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663096, "dur": 10,
    "args": {
      "External id": 151309,
      "cbid": 211, "correlation": 151309
    }
  },
  {
    "ph": "s", "id": 151309, "pid": 76337, "tid": -914061504, "ts": 1716454223663096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663182, "dur": 1,
    "args": {
      "External id": 151320,
      "cbid": 251, "correlation": 151320
    }
  },
  {
    "ph": "f", "id": 151320, "pid": 76337, "tid": -914061504, "ts": 1716454223663182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223793201, "dur": 196,
    "args": {
      "External id": 151321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151321,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151321, "pid": 5, "tid": 7, "ts": 1716454223793201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663187, "dur": 13,
    "args": {
      "External id": 151321,
      "cbid": 211, "correlation": 151321
    }
  },
  {
    "ph": "s", "id": 151321, "pid": 76337, "tid": -914061504, "ts": 1716454223663187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663256, "dur": 1,
    "args": {
      "External id": 151332,
      "cbid": 251, "correlation": 151332
    }
  },
  {
    "ph": "f", "id": 151332, "pid": 76337, "tid": -914061504, "ts": 1716454223663256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223793399, "dur": 188,
    "args": {
      "External id": 151333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151333,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151333, "pid": 5, "tid": 7, "ts": 1716454223793399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663260, "dur": 11,
    "args": {
      "External id": 151333,
      "cbid": 211, "correlation": 151333
    }
  },
  {
    "ph": "s", "id": 151333, "pid": 76337, "tid": -914061504, "ts": 1716454223663260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663322, "dur": 1,
    "args": {
      "External id": 151344,
      "cbid": 251, "correlation": 151344
    }
  },
  {
    "ph": "f", "id": 151344, "pid": 76337, "tid": -914061504, "ts": 1716454223663322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223793588, "dur": 186,
    "args": {
      "External id": 151345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151345,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151345, "pid": 5, "tid": 7, "ts": 1716454223793588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663326, "dur": 12,
    "args": {
      "External id": 151345,
      "cbid": 211, "correlation": 151345
    }
  },
  {
    "ph": "s", "id": 151345, "pid": 76337, "tid": -914061504, "ts": 1716454223663326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223793776, "dur": 18554,
    "args": {
      "External id": 151366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151366,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 151366, "pid": 5, "tid": 7, "ts": 1716454223793776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663408, "dur": 12,
    "args": {
      "External id": 151366,
      "cbid": 211, "correlation": 151366
    }
  },
  {
    "ph": "s", "id": 151366, "pid": 76337, "tid": -914061504, "ts": 1716454223663408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663505, "dur": 1,
    "args": {
      "External id": 151384,
      "cbid": 251, "correlation": 151384
    }
  },
  {
    "ph": "f", "id": 151384, "pid": 76337, "tid": -914061504, "ts": 1716454223663505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223812331, "dur": 202,
    "args": {
      "External id": 151386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151386,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151386, "pid": 5, "tid": 7, "ts": 1716454223812331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663511, "dur": 13,
    "args": {
      "External id": 151386,
      "cbid": 211, "correlation": 151386
    }
  },
  {
    "ph": "s", "id": 151386, "pid": 76337, "tid": -914061504, "ts": 1716454223663511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223812535, "dur": 66,
    "args": {
      "External id": 151394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151394,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151394, "pid": 5, "tid": 7, "ts": 1716454223812535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663582, "dur": 22,
    "args": {
      "External id": 151394,
      "cbid": 211, "correlation": 151394
    }
  },
  {
    "ph": "s", "id": 151394, "pid": 76337, "tid": -914061504, "ts": 1716454223663582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223812602, "dur": 97,
    "args": {
      "External id": 151402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151402,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151402, "pid": 5, "tid": 7, "ts": 1716454223812602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663631, "dur": 120,
    "args": {
      "External id": 151402,
      "cbid": 211, "correlation": 151402
    }
  },
  {
    "ph": "s", "id": 151402, "pid": 76337, "tid": -914061504, "ts": 1716454223663631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223812700, "dur": 53,
    "args": {
      "External id": 151413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151413, "pid": 5, "tid": 7, "ts": 1716454223812700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223663815, "dur": 1896,
    "args": {
      "External id": 151413,
      "cbid": 211, "correlation": 151413
    }
  },
  {
    "ph": "s", "id": 151413, "pid": 76337, "tid": -914061504, "ts": 1716454223663815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223812755, "dur": 92,
    "args": {
      "External id": 151435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151435,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151435, "pid": 5, "tid": 7, "ts": 1716454223812755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223665730, "dur": 128,
    "args": {
      "External id": 151435,
      "cbid": 211, "correlation": 151435
    }
  },
  {
    "ph": "s", "id": 151435, "pid": 76337, "tid": -914061504, "ts": 1716454223665730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223665937, "dur": 1,
    "args": {
      "External id": 151446,
      "cbid": 251, "correlation": 151446
    }
  },
  {
    "ph": "f", "id": 151446, "pid": 76337, "tid": -914061504, "ts": 1716454223665937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223812848, "dur": 102,
    "args": {
      "External id": 151447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151447,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151447, "pid": 5, "tid": 7, "ts": 1716454223812848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223665942, "dur": 13,
    "args": {
      "External id": 151447,
      "cbid": 211, "correlation": 151447
    }
  },
  {
    "ph": "s", "id": 151447, "pid": 76337, "tid": -914061504, "ts": 1716454223665942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666024, "dur": 1,
    "args": {
      "External id": 151458,
      "cbid": 251, "correlation": 151458
    }
  },
  {
    "ph": "f", "id": 151458, "pid": 76337, "tid": -914061504, "ts": 1716454223666024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666028, "dur": 0,
    "args": {
      "External id": 151459,
      "cbid": 251, "correlation": 151459
    }
  },
  {
    "ph": "f", "id": 151459, "pid": 76337, "tid": -914061504, "ts": 1716454223666028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223812952, "dur": 10,
    "args": {
      "External id": 151460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151460,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151460, "pid": 5, "tid": 7, "ts": 1716454223812952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666030, "dur": 13,
    "args": {
      "External id": 151460,
      "cbid": 211, "correlation": 151460
    }
  },
  {
    "ph": "s", "id": 151460, "pid": 76337, "tid": -914061504, "ts": 1716454223666030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223812963, "dur": 5,
    "args": {
      "External id": 151462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151462,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 151462, "pid": 5, "tid": 7, "ts": 1716454223812963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666044, "dur": 6,
    "args": {
      "External id": 151462,
      "cbid": 211, "correlation": 151462
    }
  },
  {
    "ph": "s", "id": 151462, "pid": 76337, "tid": -914061504, "ts": 1716454223666044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666106, "dur": 1,
    "args": {
      "External id": 151473,
      "cbid": 251, "correlation": 151473
    }
  },
  {
    "ph": "f", "id": 151473, "pid": 76337, "tid": -914061504, "ts": 1716454223666106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666109, "dur": 0,
    "args": {
      "External id": 151474,
      "cbid": 251, "correlation": 151474
    }
  },
  {
    "ph": "f", "id": 151474, "pid": 76337, "tid": -914061504, "ts": 1716454223666109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223812970, "dur": 6,
    "args": {
      "External id": 151475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151475,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151475, "pid": 5, "tid": 7, "ts": 1716454223812970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666111, "dur": 13,
    "args": {
      "External id": 151475,
      "cbid": 211, "correlation": 151475
    }
  },
  {
    "ph": "s", "id": 151475, "pid": 76337, "tid": -914061504, "ts": 1716454223666111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223812977, "dur": 3,
    "args": {
      "External id": 151477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151477,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 151477, "pid": 5, "tid": 7, "ts": 1716454223812977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666125, "dur": 6,
    "args": {
      "External id": 151477,
      "cbid": 211, "correlation": 151477
    }
  },
  {
    "ph": "s", "id": 151477, "pid": 76337, "tid": -914061504, "ts": 1716454223666125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223812982, "dur": 156,
    "args": {
      "External id": 151498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151498,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 151498, "pid": 5, "tid": 7, "ts": 1716454223812982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666200, "dur": 12,
    "args": {
      "External id": 151498,
      "cbid": 211, "correlation": 151498
    }
  },
  {
    "ph": "s", "id": 151498, "pid": 76337, "tid": -914061504, "ts": 1716454223666200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666297, "dur": 1,
    "args": {
      "External id": 151516,
      "cbid": 251, "correlation": 151516
    }
  },
  {
    "ph": "f", "id": 151516, "pid": 76337, "tid": -914061504, "ts": 1716454223666297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223813139, "dur": 107,
    "args": {
      "External id": 151518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151518,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 151518, "pid": 5, "tid": 7, "ts": 1716454223813139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666302, "dur": 13,
    "args": {
      "External id": 151518,
      "cbid": 211, "correlation": 151518
    }
  },
  {
    "ph": "s", "id": 151518, "pid": 76337, "tid": -914061504, "ts": 1716454223666302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223813248, "dur": 34,
    "args": {
      "External id": 151526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151526,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151526, "pid": 5, "tid": 7, "ts": 1716454223813248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666371, "dur": 13,
    "args": {
      "External id": 151526,
      "cbid": 211, "correlation": 151526
    }
  },
  {
    "ph": "s", "id": 151526, "pid": 76337, "tid": -914061504, "ts": 1716454223666371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223813284, "dur": 68,
    "args": {
      "External id": 151534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151534, "pid": 5, "tid": 7, "ts": 1716454223813284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666412, "dur": 9,
    "args": {
      "External id": 151534,
      "cbid": 211, "correlation": 151534
    }
  },
  {
    "ph": "s", "id": 151534, "pid": 76337, "tid": -914061504, "ts": 1716454223666412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223813353, "dur": 91,
    "args": {
      "External id": 151556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151556,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151556, "pid": 5, "tid": 7, "ts": 1716454223813353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666463, "dur": 10,
    "args": {
      "External id": 151556,
      "cbid": 211, "correlation": 151556
    }
  },
  {
    "ph": "s", "id": 151556, "pid": 76337, "tid": -914061504, "ts": 1716454223666463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666550, "dur": 1,
    "args": {
      "External id": 151572,
      "cbid": 251, "correlation": 151572
    }
  },
  {
    "ph": "f", "id": 151572, "pid": 76337, "tid": -914061504, "ts": 1716454223666550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223813446, "dur": 575,
    "args": {
      "External id": 151574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151574,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 151574, "pid": 5, "tid": 7, "ts": 1716454223813446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666556, "dur": 12,
    "args": {
      "External id": 151574,
      "cbid": 211, "correlation": 151574
    }
  },
  {
    "ph": "s", "id": 151574, "pid": 76337, "tid": -914061504, "ts": 1716454223666556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223814022, "dur": 242,
    "args": {
      "External id": 151582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151582, "pid": 5, "tid": 7, "ts": 1716454223814022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666620, "dur": 13,
    "args": {
      "External id": 151582,
      "cbid": 211, "correlation": 151582
    }
  },
  {
    "ph": "s", "id": 151582, "pid": 76337, "tid": -914061504, "ts": 1716454223666620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223814265, "dur": 252,
    "args": {
      "External id": 151590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151590, "pid": 5, "tid": 7, "ts": 1716454223814265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666651, "dur": 8,
    "args": {
      "External id": 151590,
      "cbid": 211, "correlation": 151590
    }
  },
  {
    "ph": "s", "id": 151590, "pid": 76337, "tid": -914061504, "ts": 1716454223666651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666733, "dur": 1,
    "args": {
      "External id": 151606,
      "cbid": 251, "correlation": 151606
    }
  },
  {
    "ph": "f", "id": 151606, "pid": 76337, "tid": -914061504, "ts": 1716454223666733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666738, "dur": 0,
    "args": {
      "External id": 151608,
      "cbid": 251, "correlation": 151608
    }
  },
  {
    "ph": "f", "id": 151608, "pid": 76337, "tid": -914061504, "ts": 1716454223666738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223814518, "dur": 361,
    "args": {
      "External id": 151609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151609,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151609, "pid": 5, "tid": 7, "ts": 1716454223814518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666741, "dur": 13,
    "args": {
      "External id": 151609,
      "cbid": 211, "correlation": 151609
    }
  },
  {
    "ph": "s", "id": 151609, "pid": 76337, "tid": -914061504, "ts": 1716454223666741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223814880, "dur": 50,
    "args": {
      "External id": 151617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151617,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151617, "pid": 5, "tid": 7, "ts": 1716454223814880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223666783, "dur": 186,
    "args": {
      "External id": 151617,
      "cbid": 211, "correlation": 151617
    }
  },
  {
    "ph": "s", "id": 151617, "pid": 76337, "tid": -914061504, "ts": 1716454223666783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223814932, "dur": 158,
    "args": {
      "External id": 151628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151628, "pid": 5, "tid": 7, "ts": 1716454223814932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667035, "dur": 64,
    "args": {
      "External id": 151628,
      "cbid": 211, "correlation": 151628
    }
  },
  {
    "ph": "s", "id": 151628, "pid": 76337, "tid": -914061504, "ts": 1716454223667035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667151, "dur": 0,
    "args": {
      "External id": 151640,
      "cbid": 317, "correlation": 151640
    }
  },
  {
    "ph": "f", "id": 151640, "pid": 76337, "tid": -914061504, "ts": 1716454223667151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667152, "dur": 0,
    "args": {
      "External id": 151641,
      "cbid": 203, "correlation": 151641
    }
  },
  {
    "ph": "f", "id": 151641, "pid": 76337, "tid": -914061504, "ts": 1716454223667152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667152, "dur": 0,
    "args": {
      "External id": 151642,
      "cbid": 205, "correlation": 151642
    }
  },
  {
    "ph": "f", "id": 151642, "pid": 76337, "tid": -914061504, "ts": 1716454223667152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667176, "dur": 1,
    "args": {
      "External id": 151646,
      "cbid": 251, "correlation": 151646
    }
  },
  {
    "ph": "f", "id": 151646, "pid": 76337, "tid": -914061504, "ts": 1716454223667176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667178, "dur": 0,
    "args": {
      "External id": 151647,
      "cbid": 251, "correlation": 151647
    }
  },
  {
    "ph": "f", "id": 151647, "pid": 76337, "tid": -914061504, "ts": 1716454223667178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667179, "dur": 0,
    "args": {
      "External id": 151648,
      "cbid": 251, "correlation": 151648
    }
  },
  {
    "ph": "f", "id": 151648, "pid": 76337, "tid": -914061504, "ts": 1716454223667179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667179, "dur": 0,
    "args": {
      "External id": 151649,
      "cbid": 251, "correlation": 151649
    }
  },
  {
    "ph": "f", "id": 151649, "pid": 76337, "tid": -914061504, "ts": 1716454223667179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667180, "dur": 0,
    "args": {
      "External id": 151650,
      "cbid": 251, "correlation": 151650
    }
  },
  {
    "ph": "f", "id": 151650, "pid": 76337, "tid": -914061504, "ts": 1716454223667180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667181, "dur": 0,
    "args": {
      "External id": 151651,
      "cbid": 251, "correlation": 151651
    }
  },
  {
    "ph": "f", "id": 151651, "pid": 76337, "tid": -914061504, "ts": 1716454223667181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667182, "dur": 0,
    "args": {
      "External id": 151652,
      "cbid": 251, "correlation": 151652
    }
  },
  {
    "ph": "f", "id": 151652, "pid": 76337, "tid": -914061504, "ts": 1716454223667182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667183, "dur": 0,
    "args": {
      "External id": 151653,
      "cbid": 251, "correlation": 151653
    }
  },
  {
    "ph": "f", "id": 151653, "pid": 76337, "tid": -914061504, "ts": 1716454223667183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667184, "dur": 0,
    "args": {
      "External id": 151654,
      "cbid": 251, "correlation": 151654
    }
  },
  {
    "ph": "f", "id": 151654, "pid": 76337, "tid": -914061504, "ts": 1716454223667184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223815091, "dur": 114,
    "args": {
      "External id": 151655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151655,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 151655, "pid": 5, "tid": 7, "ts": 1716454223815091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667186, "dur": 42,
    "args": {
      "External id": 151655,
      "cbid": 211, "correlation": 151655
    }
  },
  {
    "ph": "s", "id": 151655, "pid": 76337, "tid": -914061504, "ts": 1716454223667186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223815206, "dur": 61,
    "args": {
      "External id": 151661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151661, "pid": 5, "tid": 7, "ts": 1716454223815206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667251, "dur": 281,
    "args": {
      "External id": 151661,
      "cbid": 211, "correlation": 151661
    }
  },
  {
    "ph": "s", "id": 151661, "pid": 76337, "tid": -914061504, "ts": 1716454223667251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223815269, "dur": 50,
    "args": {
      "External id": 151669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151669,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151669, "pid": 5, "tid": 7, "ts": 1716454223815269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667555, "dur": 9,
    "args": {
      "External id": 151669,
      "cbid": 211, "correlation": 151669
    }
  },
  {
    "ph": "s", "id": 151669, "pid": 76337, "tid": -914061504, "ts": 1716454223667555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223815319, "dur": 52,
    "args": {
      "External id": 151689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151689,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 151689, "pid": 5, "tid": 7, "ts": 1716454223815319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667635, "dur": 12,
    "args": {
      "External id": 151689,
      "cbid": 211, "correlation": 151689
    }
  },
  {
    "ph": "s", "id": 151689, "pid": 76337, "tid": -914061504, "ts": 1716454223667635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223815372, "dur": 5,
    "args": {
      "External id": 151701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151701,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151701, "pid": 5, "tid": 7, "ts": 1716454223815372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667657, "dur": 11,
    "args": {
      "External id": 151701,
      "cbid": 211, "correlation": 151701
    }
  },
  {
    "ph": "s", "id": 151701, "pid": 76337, "tid": -914061504, "ts": 1716454223667657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223815378, "dur": 56,
    "args": {
      "External id": 151704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151704,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151704, "pid": 5, "tid": 7, "ts": 1716454223815378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667680, "dur": 110,
    "args": {
      "External id": 151704,
      "cbid": 211, "correlation": 151704
    }
  },
  {
    "ph": "s", "id": 151704, "pid": 76337, "tid": -914061504, "ts": 1716454223667680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223815435, "dur": 37,
    "args": {
      "External id": 151713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151713,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151713, "pid": 5, "tid": 7, "ts": 1716454223815435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667830, "dur": 11,
    "args": {
      "External id": 151713,
      "cbid": 211, "correlation": 151713
    }
  },
  {
    "ph": "s", "id": 151713, "pid": 76337, "tid": -914061504, "ts": 1716454223667830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667887, "dur": 0,
    "args": {
      "External id": 151723,
      "cbid": 317, "correlation": 151723
    }
  },
  {
    "ph": "f", "id": 151723, "pid": 76337, "tid": -914061504, "ts": 1716454223667887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667888, "dur": 0,
    "args": {
      "External id": 151724,
      "cbid": 203, "correlation": 151724
    }
  },
  {
    "ph": "f", "id": 151724, "pid": 76337, "tid": -914061504, "ts": 1716454223667888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667889, "dur": 0,
    "args": {
      "External id": 151725,
      "cbid": 205, "correlation": 151725
    }
  },
  {
    "ph": "f", "id": 151725, "pid": 76337, "tid": -914061504, "ts": 1716454223667889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223815473, "dur": 41,
    "args": {
      "External id": 151729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151729,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151729, "pid": 5, "tid": 7, "ts": 1716454223815473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667905, "dur": 12,
    "args": {
      "External id": 151729,
      "cbid": 211, "correlation": 151729
    }
  },
  {
    "ph": "s", "id": 151729, "pid": 76337, "tid": -914061504, "ts": 1716454223667905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223815516, "dur": 4,
    "args": {
      "External id": 151731,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151731,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.428571428571429
    }
  },
  {
    "ph": "f", "id": 151731, "pid": 5, "tid": 7, "ts": 1716454223815516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667920, "dur": 23,
    "args": {
      "External id": 151731,
      "cbid": 51, "correlation": 151731
    }
  },
  {
    "ph": "s", "id": 151731, "pid": 76337, "tid": -914061504, "ts": 1716454223667920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667949, "dur": 1,
    "args": {
      "External id": 151733,
      "cbid": 200, "correlation": 151733
    }
  },
  {
    "ph": "f", "id": 151733, "pid": 76337, "tid": -914061504, "ts": 1716454223667949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667951, "dur": 0,
    "args": {
      "External id": 151734,
      "cbid": 200, "correlation": 151734
    }
  },
  {
    "ph": "f", "id": 151734, "pid": 76337, "tid": -914061504, "ts": 1716454223667951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667951, "dur": 0,
    "args": {
      "External id": 151735,
      "cbid": 200, "correlation": 151735
    }
  },
  {
    "ph": "f", "id": 151735, "pid": 76337, "tid": -914061504, "ts": 1716454223667951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667952, "dur": 0,
    "args": {
      "External id": 151736,
      "cbid": 200, "correlation": 151736
    }
  },
  {
    "ph": "f", "id": 151736, "pid": 76337, "tid": -914061504, "ts": 1716454223667952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667953, "dur": 4,
    "args": {
      "External id": 151737,
      "cbid": 15, "correlation": 151737
    }
  },
  {
    "ph": "f", "id": 151737, "pid": 76337, "tid": -914061504, "ts": 1716454223667953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667958, "dur": 1,
    "args": {
      "External id": 151738,
      "cbid": 251, "correlation": 151738
    }
  },
  {
    "ph": "f", "id": 151738, "pid": 76337, "tid": -914061504, "ts": 1716454223667958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223815521, "dur": 25,
    "args": {
      "External id": 151739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151739,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151739, "pid": 5, "tid": 7, "ts": 1716454223815521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667961, "dur": 9,
    "args": {
      "External id": 151739,
      "cbid": 211, "correlation": 151739
    }
  },
  {
    "ph": "s", "id": 151739, "pid": 76337, "tid": -914061504, "ts": 1716454223667961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223815548, "dur": 4,
    "args": {
      "External id": 151741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 151741, "pid": 5, "tid": 7, "ts": 1716454223815548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667983, "dur": 6,
    "args": {
      "External id": 151741,
      "cbid": 211, "correlation": 151741
    }
  },
  {
    "ph": "s", "id": 151741, "pid": 76337, "tid": -914061504, "ts": 1716454223667983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667994, "dur": 0,
    "args": {
      "External id": 151742,
      "cbid": 51, "correlation": 151742
    }
  },
  {
    "ph": "s", "id": 151742, "pid": 76337, "tid": -914061504, "ts": 1716454223667994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223815553, "dur": 187,
    "args": {
      "External id": 151743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151743,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151743, "pid": 5, "tid": 7, "ts": 1716454223815553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223667995, "dur": 187,
    "args": {
      "External id": 151743,
      "cbid": 211, "correlation": 151743
    }
  },
  {
    "ph": "s", "id": 151743, "pid": 76337, "tid": -914061504, "ts": 1716454223667995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223815741, "dur": 6,
    "args": {
      "External id": 151744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151744,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151744, "pid": 5, "tid": 7, "ts": 1716454223815741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223668185, "dur": 6,
    "args": {
      "External id": 151744,
      "cbid": 211, "correlation": 151744
    }
  },
  {
    "ph": "s", "id": 151744, "pid": 76337, "tid": -914061504, "ts": 1716454223668185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223815749, "dur": 5,
    "args": {
      "External id": 151750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 151750, "pid": 5, "tid": 7, "ts": 1716454223815749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223668216, "dur": 9,
    "args": {
      "External id": 151750,
      "cbid": 211, "correlation": 151750
    }
  },
  {
    "ph": "s", "id": 151750, "pid": 76337, "tid": -914061504, "ts": 1716454223668216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223815755, "dur": 3,
    "args": {
      "External id": 151758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151758,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151758, "pid": 5, "tid": 7, "ts": 1716454223815755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223669997, "dur": 17,
    "args": {
      "External id": 151758,
      "cbid": 211, "correlation": 151758
    }
  },
  {
    "ph": "s", "id": 151758, "pid": 76337, "tid": -914061504, "ts": 1716454223669997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223815759, "dur": 3,
    "args": {
      "External id": 151766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151766,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151766, "pid": 5, "tid": 7, "ts": 1716454223815759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670040, "dur": 10,
    "args": {
      "External id": 151766,
      "cbid": 211, "correlation": 151766
    }
  },
  {
    "ph": "s", "id": 151766, "pid": 76337, "tid": -914061504, "ts": 1716454223670040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223815763, "dur": 3,
    "args": {
      "External id": 151774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151774,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151774, "pid": 5, "tid": 7, "ts": 1716454223815763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670067, "dur": 8,
    "args": {
      "External id": 151774,
      "cbid": 211, "correlation": 151774
    }
  },
  {
    "ph": "s", "id": 151774, "pid": 76337, "tid": -914061504, "ts": 1716454223670067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223815767, "dur": 3,
    "args": {
      "External id": 151783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151783,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151783, "pid": 5, "tid": 7, "ts": 1716454223815767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670246, "dur": 14,
    "args": {
      "External id": 151783,
      "cbid": 211, "correlation": 151783
    }
  },
  {
    "ph": "s", "id": 151783, "pid": 76337, "tid": -914061504, "ts": 1716454223670246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223815771, "dur": 3,
    "args": {
      "External id": 151792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151792,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151792, "pid": 5, "tid": 7, "ts": 1716454223815771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670276, "dur": 8,
    "args": {
      "External id": 151792,
      "cbid": 211, "correlation": 151792
    }
  },
  {
    "ph": "s", "id": 151792, "pid": 76337, "tid": -914061504, "ts": 1716454223670276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223815775, "dur": 3,
    "args": {
      "External id": 151800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151800,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151800, "pid": 5, "tid": 7, "ts": 1716454223815775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670301, "dur": 8,
    "args": {
      "External id": 151800,
      "cbid": 211, "correlation": 151800
    }
  },
  {
    "ph": "s", "id": 151800, "pid": 76337, "tid": -914061504, "ts": 1716454223670301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223815780, "dur": 3,
    "args": {
      "External id": 151808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151808,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151808, "pid": 5, "tid": 7, "ts": 1716454223815780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670561, "dur": 15,
    "args": {
      "External id": 151808,
      "cbid": 211, "correlation": 151808
    }
  },
  {
    "ph": "s", "id": 151808, "pid": 76337, "tid": -914061504, "ts": 1716454223670561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223815784, "dur": 3,
    "args": {
      "External id": 151816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151816,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151816, "pid": 5, "tid": 7, "ts": 1716454223815784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670591, "dur": 7,
    "args": {
      "External id": 151816,
      "cbid": 211, "correlation": 151816
    }
  },
  {
    "ph": "s", "id": 151816, "pid": 76337, "tid": -914061504, "ts": 1716454223670591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223815789, "dur": 1,
    "args": {
      "External id": 151826,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151826,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 151826, "pid": 5, "tid": 7, "ts": 1716454223815789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670657, "dur": 35,
    "args": {
      "External id": 151826,
      "cbid": 41, "correlation": 151826
    }
  },
  {
    "ph": "s", "id": 151826, "pid": 76337, "tid": -914061504, "ts": 1716454223670657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223670693, "dur": 145114,
    "args": {
      "External id": 151827,
      "cbid": 131, "correlation": 151827
    }
  },
  {
    "ph": "f", "id": 151827, "pid": 76337, "tid": -914061504, "ts": 1716454223670693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223815961, "dur": 3,
    "args": {
      "External id": 151835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151835,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151835, "pid": 5, "tid": 7, "ts": 1716454223815961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223815939, "dur": 24,
    "args": {
      "External id": 151835,
      "cbid": 211, "correlation": 151835
    }
  },
  {
    "ph": "s", "id": 151835, "pid": 76337, "tid": -914061504, "ts": 1716454223815939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816066, "dur": 3,
    "args": {
      "External id": 151844,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151844,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 151844, "pid": 5, "tid": 7, "ts": 1716454223816066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816035, "dur": 31,
    "args": {
      "External id": 151844,
      "cbid": 41, "correlation": 151844
    }
  },
  {
    "ph": "s", "id": 151844, "pid": 76337, "tid": -914061504, "ts": 1716454223816035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223816165, "dur": 4,
    "args": {
      "External id": 151854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151854,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 151854, "pid": 5, "tid": 7, "ts": 1716454223816165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816149, "dur": 18,
    "args": {
      "External id": 151854,
      "cbid": 211, "correlation": 151854
    }
  },
  {
    "ph": "s", "id": 151854, "pid": 76337, "tid": -914061504, "ts": 1716454223816149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816241, "dur": 1,
    "args": {
      "External id": 151864,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151864,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 151864, "pid": 5, "tid": 7, "ts": 1716454223816241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816222, "dur": 17,
    "args": {
      "External id": 151864,
      "cbid": 41, "correlation": 151864
    }
  },
  {
    "ph": "s", "id": 151864, "pid": 76337, "tid": -914061504, "ts": 1716454223816222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816240, "dur": 8,
    "args": {
      "External id": 151865,
      "cbid": 131, "correlation": 151865
    }
  },
  {
    "ph": "f", "id": 151865, "pid": 76337, "tid": -914061504, "ts": 1716454223816240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816312, "dur": 3,
    "args": {
      "External id": 151872,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151872,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.72
    }
  },
  {
    "ph": "f", "id": 151872, "pid": 5, "tid": 7, "ts": 1716454223816312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816287, "dur": 25,
    "args": {
      "External id": 151872,
      "cbid": 41, "correlation": 151872
    }
  },
  {
    "ph": "s", "id": 151872, "pid": 76337, "tid": -914061504, "ts": 1716454223816287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816401, "dur": 3,
    "args": {
      "External id": 151891,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151891,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 151891, "pid": 5, "tid": 7, "ts": 1716454223816401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816383, "dur": 18,
    "args": {
      "External id": 151891,
      "cbid": 41, "correlation": 151891
    }
  },
  {
    "ph": "s", "id": 151891, "pid": 76337, "tid": -914061504, "ts": 1716454223816383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223816440, "dur": 3,
    "args": {
      "External id": 151897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151897,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151897, "pid": 5, "tid": 7, "ts": 1716454223816440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816428, "dur": 11,
    "args": {
      "External id": 151897,
      "cbid": 211, "correlation": 151897
    }
  },
  {
    "ph": "s", "id": 151897, "pid": 76337, "tid": -914061504, "ts": 1716454223816428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223816454, "dur": 6,
    "args": {
      "External id": 151899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151899,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 151899, "pid": 5, "tid": 7, "ts": 1716454223816454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816443, "dur": 10,
    "args": {
      "External id": 151899,
      "cbid": 211, "correlation": 151899
    }
  },
  {
    "ph": "s", "id": 151899, "pid": 76337, "tid": -914061504, "ts": 1716454223816443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454223816463, "dur": 3,
    "args": {
      "External id": 151901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151901,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151901, "pid": 5, "tid": 7, "ts": 1716454223816463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816455, "dur": 7,
    "args": {
      "External id": 151901,
      "cbid": 211, "correlation": 151901
    }
  },
  {
    "ph": "s", "id": 151901, "pid": 76337, "tid": -914061504, "ts": 1716454223816455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816497, "dur": 2,
    "args": {
      "External id": 151909,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151909,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 151909, "pid": 5, "tid": 7, "ts": 1716454223816497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816483, "dur": 13,
    "args": {
      "External id": 151909,
      "cbid": 41, "correlation": 151909
    }
  },
  {
    "ph": "s", "id": 151909, "pid": 76337, "tid": -914061504, "ts": 1716454223816483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223816545, "dur": 3,
    "args": {
      "External id": 151923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151923, "pid": 5, "tid": 7, "ts": 1716454223816545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816534, "dur": 12,
    "args": {
      "External id": 151923,
      "cbid": 211, "correlation": 151923
    }
  },
  {
    "ph": "s", "id": 151923, "pid": 76337, "tid": -914061504, "ts": 1716454223816534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223816564, "dur": 3,
    "args": {
      "External id": 151937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151937,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151937, "pid": 5, "tid": 7, "ts": 1716454223816564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816557, "dur": 6,
    "args": {
      "External id": 151937,
      "cbid": 211, "correlation": 151937
    }
  },
  {
    "ph": "s", "id": 151937, "pid": 76337, "tid": -914061504, "ts": 1716454223816557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223816600, "dur": 6,
    "args": {
      "External id": 151944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151944,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151944, "pid": 5, "tid": 7, "ts": 1716454223816600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816588, "dur": 12,
    "args": {
      "External id": 151944,
      "cbid": 211, "correlation": 151944
    }
  },
  {
    "ph": "s", "id": 151944, "pid": 76337, "tid": -914061504, "ts": 1716454223816588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454223816610, "dur": 6,
    "args": {
      "External id": 151947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151947,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151947, "pid": 5, "tid": 7, "ts": 1716454223816610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816603, "dur": 6,
    "args": {
      "External id": 151947,
      "cbid": 211, "correlation": 151947
    }
  },
  {
    "ph": "s", "id": 151947, "pid": 76337, "tid": -914061504, "ts": 1716454223816603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223816619, "dur": 3,
    "args": {
      "External id": 151949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151949,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151949, "pid": 5, "tid": 7, "ts": 1716454223816619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816611, "dur": 6,
    "args": {
      "External id": 151949,
      "cbid": 211, "correlation": 151949
    }
  },
  {
    "ph": "s", "id": 151949, "pid": 76337, "tid": -914061504, "ts": 1716454223816611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816638, "dur": 2,
    "args": {
      "External id": 151952,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151952,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002777777777777778
    }
  },
  {
    "ph": "f", "id": 151952, "pid": 5, "tid": 7, "ts": 1716454223816638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816626, "dur": 11,
    "args": {
      "External id": 151952,
      "cbid": 41, "correlation": 151952
    }
  },
  {
    "ph": "s", "id": 151952, "pid": 76337, "tid": -914061504, "ts": 1716454223816626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223816691, "dur": 4,
    "args": {
      "External id": 151968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151968,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 151968, "pid": 5, "tid": 7, "ts": 1716454223816691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816678, "dur": 13,
    "args": {
      "External id": 151968,
      "cbid": 211, "correlation": 151968
    }
  },
  {
    "ph": "s", "id": 151968, "pid": 76337, "tid": -914061504, "ts": 1716454223816678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223816713, "dur": 3,
    "args": {
      "External id": 151973,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151973,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003156565656565657
    }
  },
  {
    "ph": "f", "id": 151973, "pid": 5, "tid": 7, "ts": 1716454223816713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816697, "dur": 14,
    "args": {
      "External id": 151973,
      "cbid": 41, "correlation": 151973
    }
  },
  {
    "ph": "s", "id": 151973, "pid": 76337, "tid": -914061504, "ts": 1716454223816697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223816740, "dur": 1,
    "args": {
      "External id": 151979,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 151979,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005896226415094339
    }
  },
  {
    "ph": "f", "id": 151979, "pid": 5, "tid": 7, "ts": 1716454223816740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816721, "dur": 28,
    "args": {
      "External id": 151979,
      "cbid": 41, "correlation": 151979
    }
  },
  {
    "ph": "s", "id": 151979, "pid": 76337, "tid": -914061504, "ts": 1716454223816721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816750, "dur": 3,
    "args": {
      "External id": 151980,
      "cbid": 131, "correlation": 151980
    }
  },
  {
    "ph": "f", "id": 151980, "pid": 76337, "tid": -914061504, "ts": 1716454223816750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223816800, "dur": 3,
    "args": {
      "External id": 151988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151988,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151988, "pid": 5, "tid": 7, "ts": 1716454223816800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816786, "dur": 13,
    "args": {
      "External id": 151988,
      "cbid": 211, "correlation": 151988
    }
  },
  {
    "ph": "s", "id": 151988, "pid": 76337, "tid": -914061504, "ts": 1716454223816786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223816830, "dur": 3,
    "args": {
      "External id": 151998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 151998,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 151998, "pid": 5, "tid": 7, "ts": 1716454223816830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816821, "dur": 8,
    "args": {
      "External id": 151998,
      "cbid": 211, "correlation": 151998
    }
  },
  {
    "ph": "s", "id": 151998, "pid": 76337, "tid": -914061504, "ts": 1716454223816821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223816855, "dur": 3,
    "args": {
      "External id": 152007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152007,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152007, "pid": 5, "tid": 7, "ts": 1716454223816855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816845, "dur": 9,
    "args": {
      "External id": 152007,
      "cbid": 211, "correlation": 152007
    }
  },
  {
    "ph": "s", "id": 152007, "pid": 76337, "tid": -914061504, "ts": 1716454223816845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223816968, "dur": 12,
    "args": {
      "External id": 152017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152017,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152017, "pid": 5, "tid": 7, "ts": 1716454223816968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223816953, "dur": 16,
    "args": {
      "External id": 152017,
      "cbid": 211, "correlation": 152017
    }
  },
  {
    "ph": "s", "id": 152017, "pid": 76337, "tid": -914061504, "ts": 1716454223816953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223817019, "dur": 3,
    "args": {
      "External id": 152025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152025,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152025, "pid": 5, "tid": 7, "ts": 1716454223817019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817008, "dur": 9,
    "args": {
      "External id": 152025,
      "cbid": 211, "correlation": 152025
    }
  },
  {
    "ph": "s", "id": 152025, "pid": 76337, "tid": -914061504, "ts": 1716454223817008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223817064, "dur": 12,
    "args": {
      "External id": 152035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152035,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152035, "pid": 5, "tid": 7, "ts": 1716454223817064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817052, "dur": 11,
    "args": {
      "External id": 152035,
      "cbid": 211, "correlation": 152035
    }
  },
  {
    "ph": "s", "id": 152035, "pid": 76337, "tid": -914061504, "ts": 1716454223817052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223817095, "dur": 10,
    "args": {
      "External id": 152043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152043,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152043, "pid": 5, "tid": 7, "ts": 1716454223817095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817086, "dur": 9,
    "args": {
      "External id": 152043,
      "cbid": 211, "correlation": 152043
    }
  },
  {
    "ph": "s", "id": 152043, "pid": 76337, "tid": -914061504, "ts": 1716454223817086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223817121, "dur": 3,
    "args": {
      "External id": 152052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152052, "pid": 5, "tid": 7, "ts": 1716454223817121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817112, "dur": 9,
    "args": {
      "External id": 152052,
      "cbid": 211, "correlation": 152052
    }
  },
  {
    "ph": "s", "id": 152052, "pid": 76337, "tid": -914061504, "ts": 1716454223817112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223817148, "dur": 5,
    "args": {
      "External id": 152061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152061,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152061, "pid": 5, "tid": 7, "ts": 1716454223817148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817136, "dur": 10,
    "args": {
      "External id": 152061,
      "cbid": 211, "correlation": 152061
    }
  },
  {
    "ph": "s", "id": 152061, "pid": 76337, "tid": -914061504, "ts": 1716454223817136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223817187, "dur": 8,
    "args": {
      "External id": 152071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152071,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152071, "pid": 5, "tid": 7, "ts": 1716454223817187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817175, "dur": 12,
    "args": {
      "External id": 152071,
      "cbid": 211, "correlation": 152071
    }
  },
  {
    "ph": "s", "id": 152071, "pid": 76337, "tid": -914061504, "ts": 1716454223817175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223817514, "dur": 3,
    "args": {
      "External id": 152080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152080,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152080, "pid": 5, "tid": 7, "ts": 1716454223817514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817498, "dur": 15,
    "args": {
      "External id": 152080,
      "cbid": 211, "correlation": 152080
    }
  },
  {
    "ph": "s", "id": 152080, "pid": 76337, "tid": -914061504, "ts": 1716454223817498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223817542, "dur": 3,
    "args": {
      "External id": 152088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152088,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152088, "pid": 5, "tid": 7, "ts": 1716454223817542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817532, "dur": 9,
    "args": {
      "External id": 152088,
      "cbid": 211, "correlation": 152088
    }
  },
  {
    "ph": "s", "id": 152088, "pid": 76337, "tid": -914061504, "ts": 1716454223817532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223817593, "dur": 1,
    "args": {
      "External id": 152098,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152098,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 152098, "pid": 5, "tid": 7, "ts": 1716454223817593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817578, "dur": 13,
    "args": {
      "External id": 152098,
      "cbid": 41, "correlation": 152098
    }
  },
  {
    "ph": "s", "id": 152098, "pid": 76337, "tid": -914061504, "ts": 1716454223817578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817592, "dur": 8,
    "args": {
      "External id": 152099,
      "cbid": 131, "correlation": 152099
    }
  },
  {
    "ph": "f", "id": 152099, "pid": 76337, "tid": -914061504, "ts": 1716454223817592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454223817683, "dur": 2,
    "args": {
      "External id": 152107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152107, "pid": 5, "tid": 7, "ts": 1716454223817683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817669, "dur": 14,
    "args": {
      "External id": 152107,
      "cbid": 211, "correlation": 152107
    }
  },
  {
    "ph": "s", "id": 152107, "pid": 76337, "tid": -914061504, "ts": 1716454223817669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223817755, "dur": 3,
    "args": {
      "External id": 152116,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152116,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 152116, "pid": 5, "tid": 7, "ts": 1716454223817755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817737, "dur": 17,
    "args": {
      "External id": 152116,
      "cbid": 41, "correlation": 152116
    }
  },
  {
    "ph": "s", "id": 152116, "pid": 76337, "tid": -914061504, "ts": 1716454223817737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223817828, "dur": 3,
    "args": {
      "External id": 152126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152126,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152126, "pid": 5, "tid": 7, "ts": 1716454223817828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817814, "dur": 14,
    "args": {
      "External id": 152126,
      "cbid": 211, "correlation": 152126
    }
  },
  {
    "ph": "s", "id": 152126, "pid": 76337, "tid": -914061504, "ts": 1716454223817814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223817880, "dur": 1,
    "args": {
      "External id": 152136,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152136,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 152136, "pid": 5, "tid": 7, "ts": 1716454223817880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817865, "dur": 12,
    "args": {
      "External id": 152136,
      "cbid": 41, "correlation": 152136
    }
  },
  {
    "ph": "s", "id": 152136, "pid": 76337, "tid": -914061504, "ts": 1716454223817865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817879, "dur": 8,
    "args": {
      "External id": 152137,
      "cbid": 131, "correlation": 152137
    }
  },
  {
    "ph": "f", "id": 152137, "pid": 76337, "tid": -914061504, "ts": 1716454223817879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223817939, "dur": 3,
    "args": {
      "External id": 152144,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152144,
      "bytes": 98304, "memory bandwidth (GB/s)": 32
    }
  },
  {
    "ph": "f", "id": 152144, "pid": 5, "tid": 7, "ts": 1716454223817939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817920, "dur": 19,
    "args": {
      "External id": 152144,
      "cbid": 41, "correlation": 152144
    }
  },
  {
    "ph": "s", "id": 152144, "pid": 76337, "tid": -914061504, "ts": 1716454223817920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223817996, "dur": 1,
    "args": {
      "External id": 152155,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152155,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 152155, "pid": 5, "tid": 7, "ts": 1716454223817996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817983, "dur": 10,
    "args": {
      "External id": 152155,
      "cbid": 41, "correlation": 152155
    }
  },
  {
    "ph": "s", "id": 152155, "pid": 76337, "tid": -914061504, "ts": 1716454223817983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223817995, "dur": 8,
    "args": {
      "External id": 152156,
      "cbid": 131, "correlation": 152156
    }
  },
  {
    "ph": "f", "id": 152156, "pid": 76337, "tid": -914061504, "ts": 1716454223817995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818046, "dur": 3,
    "args": {
      "External id": 152164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152164,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152164, "pid": 5, "tid": 7, "ts": 1716454223818046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818033, "dur": 13,
    "args": {
      "External id": 152164,
      "cbid": 211, "correlation": 152164
    }
  },
  {
    "ph": "s", "id": 152164, "pid": 76337, "tid": -914061504, "ts": 1716454223818033,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818075, "dur": 3,
    "args": {
      "External id": 152174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152174,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152174, "pid": 5, "tid": 7, "ts": 1716454223818075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818066, "dur": 8,
    "args": {
      "External id": 152174,
      "cbid": 211, "correlation": 152174
    }
  },
  {
    "ph": "s", "id": 152174, "pid": 76337, "tid": -914061504, "ts": 1716454223818066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223818097, "dur": 3,
    "args": {
      "External id": 152183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152183,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152183, "pid": 5, "tid": 7, "ts": 1716454223818097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818089, "dur": 7,
    "args": {
      "External id": 152183,
      "cbid": 211, "correlation": 152183
    }
  },
  {
    "ph": "s", "id": 152183, "pid": 76337, "tid": -914061504, "ts": 1716454223818089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223818167, "dur": 6,
    "args": {
      "External id": 152191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152191,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152191, "pid": 5, "tid": 7, "ts": 1716454223818167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818153, "dur": 15,
    "args": {
      "External id": 152191,
      "cbid": 211, "correlation": 152191
    }
  },
  {
    "ph": "s", "id": 152191, "pid": 76337, "tid": -914061504, "ts": 1716454223818153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818207, "dur": 3,
    "args": {
      "External id": 152200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152200,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152200, "pid": 5, "tid": 7, "ts": 1716454223818207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818196, "dur": 10,
    "args": {
      "External id": 152200,
      "cbid": 211, "correlation": 152200
    }
  },
  {
    "ph": "s", "id": 152200, "pid": 76337, "tid": -914061504, "ts": 1716454223818196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223818230, "dur": 3,
    "args": {
      "External id": 152209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152209,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152209, "pid": 5, "tid": 7, "ts": 1716454223818230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818222, "dur": 7,
    "args": {
      "External id": 152209,
      "cbid": 211, "correlation": 152209
    }
  },
  {
    "ph": "s", "id": 152209, "pid": 76337, "tid": -914061504, "ts": 1716454223818222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818291, "dur": 3,
    "args": {
      "External id": 152217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152217,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152217, "pid": 5, "tid": 7, "ts": 1716454223818291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818280, "dur": 9,
    "args": {
      "External id": 152217,
      "cbid": 211, "correlation": 152217
    }
  },
  {
    "ph": "s", "id": 152217, "pid": 76337, "tid": -914061504, "ts": 1716454223818280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223818350, "dur": 1,
    "args": {
      "External id": 152225,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152225,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 152225, "pid": 5, "tid": 7, "ts": 1716454223818350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818333, "dur": 27,
    "args": {
      "External id": 152225,
      "cbid": 41, "correlation": 152225
    }
  },
  {
    "ph": "s", "id": 152225, "pid": 76337, "tid": -914061504, "ts": 1716454223818333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818361, "dur": 3,
    "args": {
      "External id": 152226,
      "cbid": 131, "correlation": 152226
    }
  },
  {
    "ph": "f", "id": 152226, "pid": 76337, "tid": -914061504, "ts": 1716454223818361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223818422, "dur": 1,
    "args": {
      "External id": 152236,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152236,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 152236, "pid": 5, "tid": 7, "ts": 1716454223818422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818410, "dur": 10,
    "args": {
      "External id": 152236,
      "cbid": 41, "correlation": 152236
    }
  },
  {
    "ph": "s", "id": 152236, "pid": 76337, "tid": -914061504, "ts": 1716454223818410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818421, "dur": 8,
    "args": {
      "External id": 152237,
      "cbid": 131, "correlation": 152237
    }
  },
  {
    "ph": "f", "id": 152237, "pid": 76337, "tid": -914061504, "ts": 1716454223818421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223818478, "dur": 1,
    "args": {
      "External id": 152246,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152246,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 152246, "pid": 5, "tid": 7, "ts": 1716454223818478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818467, "dur": 8,
    "args": {
      "External id": 152246,
      "cbid": 41, "correlation": 152246
    }
  },
  {
    "ph": "s", "id": 152246, "pid": 76337, "tid": -914061504, "ts": 1716454223818467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818476, "dur": 8,
    "args": {
      "External id": 152247,
      "cbid": 131, "correlation": 152247
    }
  },
  {
    "ph": "f", "id": 152247, "pid": 76337, "tid": -914061504, "ts": 1716454223818476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223818551, "dur": 4,
    "args": {
      "External id": 152254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152254, "pid": 5, "tid": 7, "ts": 1716454223818551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818533, "dur": 18,
    "args": {
      "External id": 152254,
      "cbid": 211, "correlation": 152254
    }
  },
  {
    "ph": "s", "id": 152254, "pid": 76337, "tid": -914061504, "ts": 1716454223818533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454223818586, "dur": 4,
    "args": {
      "External id": 152274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152274,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152274, "pid": 5, "tid": 7, "ts": 1716454223818586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818575, "dur": 12,
    "args": {
      "External id": 152274,
      "cbid": 211, "correlation": 152274
    }
  },
  {
    "ph": "s", "id": 152274, "pid": 76337, "tid": -914061504, "ts": 1716454223818575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818588, "dur": 0,
    "args": {
      "External id": 152275,
      "cbid": 11, "correlation": 152275
    }
  },
  {
    "ph": "f", "id": 152275, "pid": 76337, "tid": -914061504, "ts": 1716454223818588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818588, "dur": 0,
    "args": {
      "External id": 152276,
      "cbid": 11, "correlation": 152276
    }
  },
  {
    "ph": "f", "id": 152276, "pid": 76337, "tid": -914061504, "ts": 1716454223818588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223818603, "dur": 1,
    "args": {
      "External id": 152279,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152279,
      "bytes": 4, "memory bandwidth (GB/s)": 0.002403846153846154
    }
  },
  {
    "ph": "f", "id": 152279, "pid": 5, "tid": 7, "ts": 1716454223818603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818590, "dur": 22,
    "args": {
      "External id": 152279,
      "cbid": 41, "correlation": 152279
    }
  },
  {
    "ph": "s", "id": 152279, "pid": 76337, "tid": -914061504, "ts": 1716454223818590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818612, "dur": 3,
    "args": {
      "External id": 152280,
      "cbid": 131, "correlation": 152280
    }
  },
  {
    "ph": "f", "id": 152280, "pid": 76337, "tid": -914061504, "ts": 1716454223818612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223818640, "dur": 3,
    "args": {
      "External id": 152304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152304, "pid": 5, "tid": 7, "ts": 1716454223818640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818631, "dur": 9,
    "args": {
      "External id": 152304,
      "cbid": 211, "correlation": 152304
    }
  },
  {
    "ph": "s", "id": 152304, "pid": 76337, "tid": -914061504, "ts": 1716454223818631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818641, "dur": 0,
    "args": {
      "External id": 152305,
      "cbid": 11, "correlation": 152305
    }
  },
  {
    "ph": "f", "id": 152305, "pid": 76337, "tid": -914061504, "ts": 1716454223818641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818642, "dur": 0,
    "args": {
      "External id": 152306,
      "cbid": 11, "correlation": 152306
    }
  },
  {
    "ph": "f", "id": 152306, "pid": 76337, "tid": -914061504, "ts": 1716454223818642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818643, "dur": 1,
    "args": {
      "External id": 152308,
      "cbid": 200, "correlation": 152308
    }
  },
  {
    "ph": "f", "id": 152308, "pid": 76337, "tid": -914061504, "ts": 1716454223818643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454223818654, "dur": 4,
    "args": {
      "External id": 152310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152310,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152310, "pid": 5, "tid": 7, "ts": 1716454223818654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818646, "dur": 8,
    "args": {
      "External id": 152310,
      "cbid": 211, "correlation": 152310
    }
  },
  {
    "ph": "s", "id": 152310, "pid": 76337, "tid": -914061504, "ts": 1716454223818646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818655, "dur": 0,
    "args": {
      "External id": 152311,
      "cbid": 11, "correlation": 152311
    }
  },
  {
    "ph": "f", "id": 152311, "pid": 76337, "tid": -914061504, "ts": 1716454223818655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818655, "dur": 0,
    "args": {
      "External id": 152312,
      "cbid": 11, "correlation": 152312
    }
  },
  {
    "ph": "f", "id": 152312, "pid": 76337, "tid": -914061504, "ts": 1716454223818655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454223818692, "dur": 1,
    "args": {
      "External id": 152319,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152319,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 152319, "pid": 5, "tid": 7, "ts": 1716454223818692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818681, "dur": 20,
    "args": {
      "External id": 152319,
      "cbid": 41, "correlation": 152319
    }
  },
  {
    "ph": "s", "id": 152319, "pid": 76337, "tid": -914061504, "ts": 1716454223818681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818701, "dur": 3,
    "args": {
      "External id": 152320,
      "cbid": 131, "correlation": 152320
    }
  },
  {
    "ph": "f", "id": 152320, "pid": 76337, "tid": -914061504, "ts": 1716454223818701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454223818752, "dur": 1,
    "args": {
      "External id": 152330,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 152330,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 152330, "pid": 5, "tid": 7, "ts": 1716454223818752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818739, "dur": 11,
    "args": {
      "External id": 152330,
      "cbid": 41, "correlation": 152330
    }
  },
  {
    "ph": "s", "id": 152330, "pid": 76337, "tid": -914061504, "ts": 1716454223818739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818750, "dur": 8,
    "args": {
      "External id": 152331,
      "cbid": 131, "correlation": 152331
    }
  },
  {
    "ph": "f", "id": 152331, "pid": 76337, "tid": -914061504, "ts": 1716454223818750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223818821, "dur": 5,
    "args": {
      "External id": 152338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152338,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152338, "pid": 5, "tid": 7, "ts": 1716454223818821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818806, "dur": 15,
    "args": {
      "External id": 152338,
      "cbid": 211, "correlation": 152338
    }
  },
  {
    "ph": "s", "id": 152338, "pid": 76337, "tid": -914061504, "ts": 1716454223818806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818894, "dur": 3,
    "args": {
      "External id": 152347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152347,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152347, "pid": 5, "tid": 7, "ts": 1716454223818894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818880, "dur": 13,
    "args": {
      "External id": 152347,
      "cbid": 211, "correlation": 152347
    }
  },
  {
    "ph": "s", "id": 152347, "pid": 76337, "tid": -914061504, "ts": 1716454223818880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818931, "dur": 3,
    "args": {
      "External id": 152355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152355,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152355, "pid": 5, "tid": 7, "ts": 1716454223818931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818921, "dur": 10,
    "args": {
      "External id": 152355,
      "cbid": 211, "correlation": 152355
    }
  },
  {
    "ph": "s", "id": 152355, "pid": 76337, "tid": -914061504, "ts": 1716454223818921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223818963, "dur": 4,
    "args": {
      "External id": 152363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152363,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152363, "pid": 5, "tid": 7, "ts": 1716454223818963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818953, "dur": 11,
    "args": {
      "External id": 152363,
      "cbid": 211, "correlation": 152363
    }
  },
  {
    "ph": "s", "id": 152363, "pid": 76337, "tid": -914061504, "ts": 1716454223818953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223819002, "dur": 4,
    "args": {
      "External id": 152371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152371,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152371, "pid": 5, "tid": 7, "ts": 1716454223819002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223818991, "dur": 11,
    "args": {
      "External id": 152371,
      "cbid": 211, "correlation": 152371
    }
  },
  {
    "ph": "s", "id": 152371, "pid": 76337, "tid": -914061504, "ts": 1716454223818991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223819030, "dur": 3,
    "args": {
      "External id": 152379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152379,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152379, "pid": 5, "tid": 7, "ts": 1716454223819030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819020, "dur": 9,
    "args": {
      "External id": 152379,
      "cbid": 211, "correlation": 152379
    }
  },
  {
    "ph": "s", "id": 152379, "pid": 76337, "tid": -914061504, "ts": 1716454223819020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223819056, "dur": 3,
    "args": {
      "External id": 152387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152387,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152387, "pid": 5, "tid": 7, "ts": 1716454223819056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819047, "dur": 8,
    "args": {
      "External id": 152387,
      "cbid": 211, "correlation": 152387
    }
  },
  {
    "ph": "s", "id": 152387, "pid": 76337, "tid": -914061504, "ts": 1716454223819047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223819077, "dur": 4,
    "args": {
      "External id": 152395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152395, "pid": 5, "tid": 7, "ts": 1716454223819077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819069, "dur": 7,
    "args": {
      "External id": 152395,
      "cbid": 211, "correlation": 152395
    }
  },
  {
    "ph": "s", "id": 152395, "pid": 76337, "tid": -914061504, "ts": 1716454223819069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223819097, "dur": 4,
    "args": {
      "External id": 152403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152403,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152403, "pid": 5, "tid": 7, "ts": 1716454223819097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819089, "dur": 6,
    "args": {
      "External id": 152403,
      "cbid": 211, "correlation": 152403
    }
  },
  {
    "ph": "s", "id": 152403, "pid": 76337, "tid": -914061504, "ts": 1716454223819089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223819116, "dur": 3,
    "args": {
      "External id": 152411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152411,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152411, "pid": 5, "tid": 7, "ts": 1716454223819116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819108, "dur": 7,
    "args": {
      "External id": 152411,
      "cbid": 211, "correlation": 152411
    }
  },
  {
    "ph": "s", "id": 152411, "pid": 76337, "tid": -914061504, "ts": 1716454223819108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223819172, "dur": 3,
    "args": {
      "External id": 152419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152419,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152419, "pid": 5, "tid": 7, "ts": 1716454223819172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819162, "dur": 10,
    "args": {
      "External id": 152419,
      "cbid": 211, "correlation": 152419
    }
  },
  {
    "ph": "s", "id": 152419, "pid": 76337, "tid": -914061504, "ts": 1716454223819162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223819199, "dur": 4,
    "args": {
      "External id": 152427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152427,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152427, "pid": 5, "tid": 7, "ts": 1716454223819199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819190, "dur": 8,
    "args": {
      "External id": 152427,
      "cbid": 211, "correlation": 152427
    }
  },
  {
    "ph": "s", "id": 152427, "pid": 76337, "tid": -914061504, "ts": 1716454223819190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223819221, "dur": 4,
    "args": {
      "External id": 152435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152435, "pid": 5, "tid": 7, "ts": 1716454223819221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819213, "dur": 7,
    "args": {
      "External id": 152435,
      "cbid": 211, "correlation": 152435
    }
  },
  {
    "ph": "s", "id": 152435, "pid": 76337, "tid": -914061504, "ts": 1716454223819213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223819241, "dur": 3,
    "args": {
      "External id": 152443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152443,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 152443, "pid": 5, "tid": 7, "ts": 1716454223819241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819233, "dur": 7,
    "args": {
      "External id": 152443,
      "cbid": 211, "correlation": 152443
    }
  },
  {
    "ph": "s", "id": 152443, "pid": 76337, "tid": -914061504, "ts": 1716454223819233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223819653, "dur": 5,
    "args": {
      "External id": 152452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152452,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152452, "pid": 5, "tid": 7, "ts": 1716454223819653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819637, "dur": 17,
    "args": {
      "External id": 152452,
      "cbid": 211, "correlation": 152452
    }
  },
  {
    "ph": "s", "id": 152452, "pid": 76337, "tid": -914061504, "ts": 1716454223819637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223819689, "dur": 5,
    "args": {
      "External id": 152461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152461,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152461, "pid": 5, "tid": 7, "ts": 1716454223819689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819678, "dur": 10,
    "args": {
      "External id": 152461,
      "cbid": 211, "correlation": 152461
    }
  },
  {
    "ph": "s", "id": 152461, "pid": 76337, "tid": -914061504, "ts": 1716454223819678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454223819818, "dur": 3,
    "args": {
      "External id": 152477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152477, "pid": 5, "tid": 7, "ts": 1716454223819818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819804, "dur": 15,
    "args": {
      "External id": 152477,
      "cbid": 211, "correlation": 152477
    }
  },
  {
    "ph": "s", "id": 152477, "pid": 76337, "tid": -914061504, "ts": 1716454223819804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223819852, "dur": 3,
    "args": {
      "External id": 152485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152485,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152485, "pid": 5, "tid": 7, "ts": 1716454223819852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819842, "dur": 8,
    "args": {
      "External id": 152485,
      "cbid": 211, "correlation": 152485
    }
  },
  {
    "ph": "s", "id": 152485, "pid": 76337, "tid": -914061504, "ts": 1716454223819842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223819883, "dur": 3,
    "args": {
      "External id": 152493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152493,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152493, "pid": 5, "tid": 7, "ts": 1716454223819883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819872, "dur": 10,
    "args": {
      "External id": 152493,
      "cbid": 211, "correlation": 152493
    }
  },
  {
    "ph": "s", "id": 152493, "pid": 76337, "tid": -914061504, "ts": 1716454223819872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223819913, "dur": 4,
    "args": {
      "External id": 152501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152501,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152501, "pid": 5, "tid": 7, "ts": 1716454223819913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819904, "dur": 8,
    "args": {
      "External id": 152501,
      "cbid": 211, "correlation": 152501
    }
  },
  {
    "ph": "s", "id": 152501, "pid": 76337, "tid": -914061504, "ts": 1716454223819904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454223819967, "dur": 4,
    "args": {
      "External id": 152513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152513,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152513, "pid": 5, "tid": 7, "ts": 1716454223819967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223819955, "dur": 13,
    "args": {
      "External id": 152513,
      "cbid": 211, "correlation": 152513
    }
  },
  {
    "ph": "s", "id": 152513, "pid": 76337, "tid": -914061504, "ts": 1716454223819955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223820023, "dur": 4,
    "args": {
      "External id": 152524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152524, "pid": 5, "tid": 7, "ts": 1716454223820023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820011, "dur": 12,
    "args": {
      "External id": 152524,
      "cbid": 211, "correlation": 152524
    }
  },
  {
    "ph": "s", "id": 152524, "pid": 76337, "tid": -914061504, "ts": 1716454223820011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223820054, "dur": 3,
    "args": {
      "External id": 152532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152532, "pid": 5, "tid": 7, "ts": 1716454223820054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820045, "dur": 8,
    "args": {
      "External id": 152532,
      "cbid": 211, "correlation": 152532
    }
  },
  {
    "ph": "s", "id": 152532, "pid": 76337, "tid": -914061504, "ts": 1716454223820045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223820088, "dur": 5,
    "args": {
      "External id": 152540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152540,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152540, "pid": 5, "tid": 7, "ts": 1716454223820088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820078, "dur": 9,
    "args": {
      "External id": 152540,
      "cbid": 211, "correlation": 152540
    }
  },
  {
    "ph": "s", "id": 152540, "pid": 76337, "tid": -914061504, "ts": 1716454223820078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223820117, "dur": 5,
    "args": {
      "External id": 152548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152548,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152548, "pid": 5, "tid": 7, "ts": 1716454223820117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820107, "dur": 9,
    "args": {
      "External id": 152548,
      "cbid": 211, "correlation": 152548
    }
  },
  {
    "ph": "s", "id": 152548, "pid": 76337, "tid": -914061504, "ts": 1716454223820107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223820147, "dur": 4,
    "args": {
      "External id": 152557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152557,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152557, "pid": 5, "tid": 7, "ts": 1716454223820147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820137, "dur": 9,
    "args": {
      "External id": 152557,
      "cbid": 211, "correlation": 152557
    }
  },
  {
    "ph": "s", "id": 152557, "pid": 76337, "tid": -914061504, "ts": 1716454223820137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223820207, "dur": 5,
    "args": {
      "External id": 152570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152570,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152570, "pid": 5, "tid": 7, "ts": 1716454223820207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820194, "dur": 14,
    "args": {
      "External id": 152570,
      "cbid": 211, "correlation": 152570
    }
  },
  {
    "ph": "s", "id": 152570, "pid": 76337, "tid": -914061504, "ts": 1716454223820194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454223820248, "dur": 5,
    "args": {
      "External id": 152580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152580,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 152580, "pid": 5, "tid": 7, "ts": 1716454223820248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820237, "dur": 10,
    "args": {
      "External id": 152580,
      "cbid": 211, "correlation": 152580
    }
  },
  {
    "ph": "s", "id": 152580, "pid": 76337, "tid": -914061504, "ts": 1716454223820237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820376, "dur": 4,
    "args": {
      "External id": 152597,
      "cbid": 251, "correlation": 152597
    }
  },
  {
    "ph": "f", "id": 152597, "pid": 76337, "tid": -914061504, "ts": 1716454223820376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454223820406, "dur": 12,
    "args": {
      "External id": 152599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152599,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152599, "pid": 5, "tid": 7, "ts": 1716454223820406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820389, "dur": 17,
    "args": {
      "External id": 152599,
      "cbid": 211, "correlation": 152599
    }
  },
  {
    "ph": "s", "id": 152599, "pid": 76337, "tid": -914061504, "ts": 1716454223820389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223820467, "dur": 3,
    "args": {
      "External id": 152607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152607,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152607, "pid": 5, "tid": 7, "ts": 1716454223820467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820455, "dur": 12,
    "args": {
      "External id": 152607,
      "cbid": 211, "correlation": 152607
    }
  },
  {
    "ph": "s", "id": 152607, "pid": 76337, "tid": -914061504, "ts": 1716454223820455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820523, "dur": 2,
    "args": {
      "External id": 152623,
      "cbid": 251, "correlation": 152623
    }
  },
  {
    "ph": "f", "id": 152623, "pid": 76337, "tid": -914061504, "ts": 1716454223820523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820529, "dur": 0,
    "args": {
      "External id": 152625,
      "cbid": 251, "correlation": 152625
    }
  },
  {
    "ph": "f", "id": 152625, "pid": 76337, "tid": -914061504, "ts": 1716454223820529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223820546, "dur": 14,
    "args": {
      "External id": 152626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152626,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152626, "pid": 5, "tid": 7, "ts": 1716454223820546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820532, "dur": 14,
    "args": {
      "External id": 152626,
      "cbid": 211, "correlation": 152626
    }
  },
  {
    "ph": "s", "id": 152626, "pid": 76337, "tid": -914061504, "ts": 1716454223820532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223820561, "dur": 5,
    "args": {
      "External id": 152628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152628,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152628, "pid": 5, "tid": 7, "ts": 1716454223820561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820550, "dur": 9,
    "args": {
      "External id": 152628,
      "cbid": 211, "correlation": 152628
    }
  },
  {
    "ph": "s", "id": 152628, "pid": 76337, "tid": -914061504, "ts": 1716454223820550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820659, "dur": 1,
    "args": {
      "External id": 152638,
      "cbid": 317, "correlation": 152638
    }
  },
  {
    "ph": "f", "id": 152638, "pid": 76337, "tid": -914061504, "ts": 1716454223820659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820661, "dur": 1,
    "args": {
      "External id": 152639,
      "cbid": 203, "correlation": 152639
    }
  },
  {
    "ph": "f", "id": 152639, "pid": 76337, "tid": -914061504, "ts": 1716454223820661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820663, "dur": 1,
    "args": {
      "External id": 152640,
      "cbid": 205, "correlation": 152640
    }
  },
  {
    "ph": "f", "id": 152640, "pid": 76337, "tid": -914061504, "ts": 1716454223820663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223820717, "dur": 7,
    "args": {
      "External id": 152644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152644,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152644, "pid": 5, "tid": 7, "ts": 1716454223820717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820703, "dur": 14,
    "args": {
      "External id": 152644,
      "cbid": 211, "correlation": 152644
    }
  },
  {
    "ph": "s", "id": 152644, "pid": 76337, "tid": -914061504, "ts": 1716454223820703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223820728, "dur": 4,
    "args": {
      "External id": 152646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152646,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 152646, "pid": 5, "tid": 7, "ts": 1716454223820728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820720, "dur": 6,
    "args": {
      "External id": 152646,
      "cbid": 211, "correlation": 152646
    }
  },
  {
    "ph": "s", "id": 152646, "pid": 76337, "tid": -914061504, "ts": 1716454223820720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223820746, "dur": 3,
    "args": {
      "External id": 152648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152648, "pid": 5, "tid": 7, "ts": 1716454223820746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820737, "dur": 8,
    "args": {
      "External id": 152648,
      "cbid": 211, "correlation": 152648
    }
  },
  {
    "ph": "s", "id": 152648, "pid": 76337, "tid": -914061504, "ts": 1716454223820737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820752, "dur": 0,
    "args": {
      "External id": 152649,
      "cbid": 51, "correlation": 152649
    }
  },
  {
    "ph": "s", "id": 152649, "pid": 76337, "tid": -914061504, "ts": 1716454223820752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223820761, "dur": 85,
    "args": {
      "External id": 152650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152650,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152650, "pid": 5, "tid": 7, "ts": 1716454223820761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820753, "dur": 7,
    "args": {
      "External id": 152650,
      "cbid": 211, "correlation": 152650
    }
  },
  {
    "ph": "s", "id": 152650, "pid": 76337, "tid": -914061504, "ts": 1716454223820753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223820848, "dur": 60,
    "args": {
      "External id": 152655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152655, "pid": 5, "tid": 7, "ts": 1716454223820848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223820790, "dur": 10,
    "args": {
      "External id": 152655,
      "cbid": 211, "correlation": 152655
    }
  },
  {
    "ph": "s", "id": 152655, "pid": 76337, "tid": -914061504, "ts": 1716454223820790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223822614, "dur": 53,
    "args": {
      "External id": 152675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152675,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 152675, "pid": 5, "tid": 7, "ts": 1716454223822614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822598, "dur": 16,
    "args": {
      "External id": 152675,
      "cbid": 211, "correlation": 152675
    }
  },
  {
    "ph": "s", "id": 152675, "pid": 76337, "tid": -914061504, "ts": 1716454223822598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223822669, "dur": 5,
    "args": {
      "External id": 152687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152687,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152687, "pid": 5, "tid": 7, "ts": 1716454223822669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822628, "dur": 8,
    "args": {
      "External id": 152687,
      "cbid": 211, "correlation": 152687
    }
  },
  {
    "ph": "s", "id": 152687, "pid": 76337, "tid": -914061504, "ts": 1716454223822628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223822675, "dur": 57,
    "args": {
      "External id": 152690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152690,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152690, "pid": 5, "tid": 7, "ts": 1716454223822675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822652, "dur": 7,
    "args": {
      "External id": 152690,
      "cbid": 211, "correlation": 152690
    }
  },
  {
    "ph": "s", "id": 152690, "pid": 76337, "tid": -914061504, "ts": 1716454223822652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223822734, "dur": 36,
    "args": {
      "External id": 152699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152699,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152699, "pid": 5, "tid": 7, "ts": 1716454223822734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822696, "dur": 10,
    "args": {
      "External id": 152699,
      "cbid": 211, "correlation": 152699
    }
  },
  {
    "ph": "s", "id": 152699, "pid": 76337, "tid": -914061504, "ts": 1716454223822696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822754, "dur": 0,
    "args": {
      "External id": 152709,
      "cbid": 317, "correlation": 152709
    }
  },
  {
    "ph": "f", "id": 152709, "pid": 76337, "tid": -914061504, "ts": 1716454223822754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822755, "dur": 0,
    "args": {
      "External id": 152710,
      "cbid": 203, "correlation": 152710
    }
  },
  {
    "ph": "f", "id": 152710, "pid": 76337, "tid": -914061504, "ts": 1716454223822755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822756, "dur": 0,
    "args": {
      "External id": 152711,
      "cbid": 205, "correlation": 152711
    }
  },
  {
    "ph": "f", "id": 152711, "pid": 76337, "tid": -914061504, "ts": 1716454223822756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223822788, "dur": 39,
    "args": {
      "External id": 152715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152715,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152715, "pid": 5, "tid": 7, "ts": 1716454223822788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822774, "dur": 13,
    "args": {
      "External id": 152715,
      "cbid": 211, "correlation": 152715
    }
  },
  {
    "ph": "s", "id": 152715, "pid": 76337, "tid": -914061504, "ts": 1716454223822774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223822828, "dur": 14,
    "args": {
      "External id": 152717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152717,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152717, "pid": 5, "tid": 7, "ts": 1716454223822828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822790, "dur": 6,
    "args": {
      "External id": 152717,
      "cbid": 211, "correlation": 152717
    }
  },
  {
    "ph": "s", "id": 152717, "pid": 76337, "tid": -914061504, "ts": 1716454223822790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223822844, "dur": 3,
    "args": {
      "External id": 152719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152719,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152719, "pid": 5, "tid": 7, "ts": 1716454223822844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822801, "dur": 6,
    "args": {
      "External id": 152719,
      "cbid": 211, "correlation": 152719
    }
  },
  {
    "ph": "s", "id": 152719, "pid": 76337, "tid": -914061504, "ts": 1716454223822801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822811, "dur": 0,
    "args": {
      "External id": 152720,
      "cbid": 51, "correlation": 152720
    }
  },
  {
    "ph": "s", "id": 152720, "pid": 76337, "tid": -914061504, "ts": 1716454223822811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223822848, "dur": 699,
    "args": {
      "External id": 152721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152721,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152721, "pid": 5, "tid": 7, "ts": 1716454223822848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822812, "dur": 6,
    "args": {
      "External id": 152721,
      "cbid": 211, "correlation": 152721
    }
  },
  {
    "ph": "s", "id": 152721, "pid": 76337, "tid": -914061504, "ts": 1716454223822812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223823548, "dur": 58,
    "args": {
      "External id": 152726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152726,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152726, "pid": 5, "tid": 7, "ts": 1716454223823548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822843, "dur": 8,
    "args": {
      "External id": 152726,
      "cbid": 211, "correlation": 152726
    }
  },
  {
    "ph": "s", "id": 152726, "pid": 76337, "tid": -914061504, "ts": 1716454223822843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223823608, "dur": 4,
    "args": {
      "External id": 152734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152734,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152734, "pid": 5, "tid": 7, "ts": 1716454223823608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822886, "dur": 9,
    "args": {
      "External id": 152734,
      "cbid": 211, "correlation": 152734
    }
  },
  {
    "ph": "s", "id": 152734, "pid": 76337, "tid": -914061504, "ts": 1716454223822886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822952, "dur": 1,
    "args": {
      "External id": 152750,
      "cbid": 251, "correlation": 152750
    }
  },
  {
    "ph": "f", "id": 152750, "pid": 76337, "tid": -914061504, "ts": 1716454223822952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822957, "dur": 0,
    "args": {
      "External id": 152752,
      "cbid": 251, "correlation": 152752
    }
  },
  {
    "ph": "f", "id": 152752, "pid": 76337, "tid": -914061504, "ts": 1716454223822957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223823613, "dur": 9,
    "args": {
      "External id": 152753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152753,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 152753, "pid": 5, "tid": 7, "ts": 1716454223823613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822960, "dur": 11,
    "args": {
      "External id": 152753,
      "cbid": 211, "correlation": 152753
    }
  },
  {
    "ph": "s", "id": 152753, "pid": 76337, "tid": -914061504, "ts": 1716454223822960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223823624, "dur": 4,
    "args": {
      "External id": 152755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152755,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 152755, "pid": 5, "tid": 7, "ts": 1716454223823624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223822981, "dur": 6,
    "args": {
      "External id": 152755,
      "cbid": 211, "correlation": 152755
    }
  },
  {
    "ph": "s", "id": 152755, "pid": 76337, "tid": -914061504, "ts": 1716454223822981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223823629, "dur": 55,
    "args": {
      "External id": 152765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152765, "pid": 5, "tid": 7, "ts": 1716454223823629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823042, "dur": 12,
    "args": {
      "External id": 152765,
      "cbid": 211, "correlation": 152765
    }
  },
  {
    "ph": "s", "id": 152765, "pid": 76337, "tid": -914061504, "ts": 1716454223823042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223823685, "dur": 51,
    "args": {
      "External id": 152785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152785,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 152785, "pid": 5, "tid": 7, "ts": 1716454223823685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823109, "dur": 11,
    "args": {
      "External id": 152785,
      "cbid": 211, "correlation": 152785
    }
  },
  {
    "ph": "s", "id": 152785, "pid": 76337, "tid": -914061504, "ts": 1716454223823109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223823738, "dur": 4,
    "args": {
      "External id": 152797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152797,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152797, "pid": 5, "tid": 7, "ts": 1716454223823738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823132, "dur": 6,
    "args": {
      "External id": 152797,
      "cbid": 211, "correlation": 152797
    }
  },
  {
    "ph": "s", "id": 152797, "pid": 76337, "tid": -914061504, "ts": 1716454223823132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223823742, "dur": 55,
    "args": {
      "External id": 152800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152800,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152800, "pid": 5, "tid": 7, "ts": 1716454223823742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823151, "dur": 6,
    "args": {
      "External id": 152800,
      "cbid": 211, "correlation": 152800
    }
  },
  {
    "ph": "s", "id": 152800, "pid": 76337, "tid": -914061504, "ts": 1716454223823151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223823799, "dur": 37,
    "args": {
      "External id": 152809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152809,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152809, "pid": 5, "tid": 7, "ts": 1716454223823799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823192, "dur": 9,
    "args": {
      "External id": 152809,
      "cbid": 211, "correlation": 152809
    }
  },
  {
    "ph": "s", "id": 152809, "pid": 76337, "tid": -914061504, "ts": 1716454223823192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823261, "dur": 0,
    "args": {
      "External id": 152819,
      "cbid": 317, "correlation": 152819
    }
  },
  {
    "ph": "f", "id": 152819, "pid": 76337, "tid": -914061504, "ts": 1716454223823261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823262, "dur": 0,
    "args": {
      "External id": 152820,
      "cbid": 203, "correlation": 152820
    }
  },
  {
    "ph": "f", "id": 152820, "pid": 76337, "tid": -914061504, "ts": 1716454223823262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823263, "dur": 0,
    "args": {
      "External id": 152821,
      "cbid": 205, "correlation": 152821
    }
  },
  {
    "ph": "f", "id": 152821, "pid": 76337, "tid": -914061504, "ts": 1716454223823263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223823837, "dur": 40,
    "args": {
      "External id": 152825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152825,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152825, "pid": 5, "tid": 7, "ts": 1716454223823837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823279, "dur": 13,
    "args": {
      "External id": 152825,
      "cbid": 211, "correlation": 152825
    }
  },
  {
    "ph": "s", "id": 152825, "pid": 76337, "tid": -914061504, "ts": 1716454223823279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223823879, "dur": 14,
    "args": {
      "External id": 152827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152827,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152827, "pid": 5, "tid": 7, "ts": 1716454223823879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823294, "dur": 5,
    "args": {
      "External id": 152827,
      "cbid": 211, "correlation": 152827
    }
  },
  {
    "ph": "s", "id": 152827, "pid": 76337, "tid": -914061504, "ts": 1716454223823294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223823894, "dur": 3,
    "args": {
      "External id": 152829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152829,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152829, "pid": 5, "tid": 7, "ts": 1716454223823894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823303, "dur": 6,
    "args": {
      "External id": 152829,
      "cbid": 211, "correlation": 152829
    }
  },
  {
    "ph": "s", "id": 152829, "pid": 76337, "tid": -914061504, "ts": 1716454223823303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823312, "dur": 0,
    "args": {
      "External id": 152830,
      "cbid": 51, "correlation": 152830
    }
  },
  {
    "ph": "s", "id": 152830, "pid": 76337, "tid": -914061504, "ts": 1716454223823312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223823898, "dur": 694,
    "args": {
      "External id": 152831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152831,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152831, "pid": 5, "tid": 7, "ts": 1716454223823898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823312, "dur": 5,
    "args": {
      "External id": 152831,
      "cbid": 211, "correlation": 152831
    }
  },
  {
    "ph": "s", "id": 152831, "pid": 76337, "tid": -914061504, "ts": 1716454223823312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223824594, "dur": 59,
    "args": {
      "External id": 152836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152836,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152836, "pid": 5, "tid": 7, "ts": 1716454223824594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823341, "dur": 9,
    "args": {
      "External id": 152836,
      "cbid": 211, "correlation": 152836
    }
  },
  {
    "ph": "s", "id": 152836, "pid": 76337, "tid": -914061504, "ts": 1716454223823341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223824654, "dur": 50,
    "args": {
      "External id": 152844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152844,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152844, "pid": 5, "tid": 7, "ts": 1716454223824654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823373, "dur": 8,
    "args": {
      "External id": 152844,
      "cbid": 211, "correlation": 152844
    }
  },
  {
    "ph": "s", "id": 152844, "pid": 76337, "tid": -914061504, "ts": 1716454223823373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223824705, "dur": 35,
    "args": {
      "External id": 152852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152852,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152852, "pid": 5, "tid": 7, "ts": 1716454223824705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823403, "dur": 9,
    "args": {
      "External id": 152852,
      "cbid": 211, "correlation": 152852
    }
  },
  {
    "ph": "s", "id": 152852, "pid": 76337, "tid": -914061504, "ts": 1716454223823403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223824742, "dur": 50,
    "args": {
      "External id": 152872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152872,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 152872, "pid": 5, "tid": 7, "ts": 1716454223824742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823488, "dur": 12,
    "args": {
      "External id": 152872,
      "cbid": 211, "correlation": 152872
    }
  },
  {
    "ph": "s", "id": 152872, "pid": 76337, "tid": -914061504, "ts": 1716454223823488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223824793, "dur": 4,
    "args": {
      "External id": 152884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152884,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 152884, "pid": 5, "tid": 7, "ts": 1716454223824793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823510, "dur": 6,
    "args": {
      "External id": 152884,
      "cbid": 211, "correlation": 152884
    }
  },
  {
    "ph": "s", "id": 152884, "pid": 76337, "tid": -914061504, "ts": 1716454223823510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223824798, "dur": 56,
    "args": {
      "External id": 152887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152887, "pid": 5, "tid": 7, "ts": 1716454223824798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823528, "dur": 7,
    "args": {
      "External id": 152887,
      "cbid": 211, "correlation": 152887
    }
  },
  {
    "ph": "s", "id": 152887, "pid": 76337, "tid": -914061504, "ts": 1716454223823528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823586, "dur": 0,
    "args": {
      "External id": 152898,
      "cbid": 317, "correlation": 152898
    }
  },
  {
    "ph": "f", "id": 152898, "pid": 76337, "tid": -914061504, "ts": 1716454223823586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823587, "dur": 0,
    "args": {
      "External id": 152899,
      "cbid": 203, "correlation": 152899
    }
  },
  {
    "ph": "f", "id": 152899, "pid": 76337, "tid": -914061504, "ts": 1716454223823587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823588, "dur": 0,
    "args": {
      "External id": 152900,
      "cbid": 205, "correlation": 152900
    }
  },
  {
    "ph": "f", "id": 152900, "pid": 76337, "tid": -914061504, "ts": 1716454223823588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823622, "dur": 3,
    "args": {
      "External id": 152904,
      "cbid": 251, "correlation": 152904
    }
  },
  {
    "ph": "f", "id": 152904, "pid": 76337, "tid": -914061504, "ts": 1716454223823622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823626, "dur": 1,
    "args": {
      "External id": 152905,
      "cbid": 251, "correlation": 152905
    }
  },
  {
    "ph": "f", "id": 152905, "pid": 76337, "tid": -914061504, "ts": 1716454223823626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823628, "dur": 1,
    "args": {
      "External id": 152906,
      "cbid": 251, "correlation": 152906
    }
  },
  {
    "ph": "f", "id": 152906, "pid": 76337, "tid": -914061504, "ts": 1716454223823628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823630, "dur": 1,
    "args": {
      "External id": 152907,
      "cbid": 251, "correlation": 152907
    }
  },
  {
    "ph": "f", "id": 152907, "pid": 76337, "tid": -914061504, "ts": 1716454223823630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823632, "dur": 1,
    "args": {
      "External id": 152908,
      "cbid": 251, "correlation": 152908
    }
  },
  {
    "ph": "f", "id": 152908, "pid": 76337, "tid": -914061504, "ts": 1716454223823632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823635, "dur": 1,
    "args": {
      "External id": 152909,
      "cbid": 251, "correlation": 152909
    }
  },
  {
    "ph": "f", "id": 152909, "pid": 76337, "tid": -914061504, "ts": 1716454223823635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823637, "dur": 1,
    "args": {
      "External id": 152910,
      "cbid": 251, "correlation": 152910
    }
  },
  {
    "ph": "f", "id": 152910, "pid": 76337, "tid": -914061504, "ts": 1716454223823637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823639, "dur": 1,
    "args": {
      "External id": 152911,
      "cbid": 251, "correlation": 152911
    }
  },
  {
    "ph": "f", "id": 152911, "pid": 76337, "tid": -914061504, "ts": 1716454223823639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823641, "dur": 0,
    "args": {
      "External id": 152912,
      "cbid": 251, "correlation": 152912
    }
  },
  {
    "ph": "f", "id": 152912, "pid": 76337, "tid": -914061504, "ts": 1716454223823641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223824856, "dur": 115,
    "args": {
      "External id": 152913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152913,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 152913, "pid": 5, "tid": 7, "ts": 1716454223824856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823645, "dur": 14,
    "args": {
      "External id": 152913,
      "cbid": 211, "correlation": 152913
    }
  },
  {
    "ph": "s", "id": 152913, "pid": 76337, "tid": -914061504, "ts": 1716454223823645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223824972, "dur": 60,
    "args": {
      "External id": 152919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152919,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152919, "pid": 5, "tid": 7, "ts": 1716454223824972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823682, "dur": 9,
    "args": {
      "External id": 152919,
      "cbid": 211, "correlation": 152919
    }
  },
  {
    "ph": "s", "id": 152919, "pid": 76337, "tid": -914061504, "ts": 1716454223823682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223825033, "dur": 654,
    "args": {
      "External id": 152928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152928,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152928, "pid": 5, "tid": 7, "ts": 1716454223825033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823777, "dur": 15,
    "args": {
      "External id": 152928,
      "cbid": 211, "correlation": 152928
    }
  },
  {
    "ph": "s", "id": 152928, "pid": 76337, "tid": -914061504, "ts": 1716454223823777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223825689, "dur": 180,
    "args": {
      "External id": 152950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152950,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 152950, "pid": 5, "tid": 7, "ts": 1716454223825689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823846, "dur": 12,
    "args": {
      "External id": 152950,
      "cbid": 211, "correlation": 152950
    }
  },
  {
    "ph": "s", "id": 152950, "pid": 76337, "tid": -914061504, "ts": 1716454223823846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823960, "dur": 2,
    "args": {
      "External id": 152961,
      "cbid": 251, "correlation": 152961
    }
  },
  {
    "ph": "f", "id": 152961, "pid": 76337, "tid": -914061504, "ts": 1716454223823960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223825870, "dur": 194,
    "args": {
      "External id": 152962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152962,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152962, "pid": 5, "tid": 7, "ts": 1716454223825870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223823967, "dur": 23,
    "args": {
      "External id": 152962,
      "cbid": 211, "correlation": 152962
    }
  },
  {
    "ph": "s", "id": 152962, "pid": 76337, "tid": -914061504, "ts": 1716454223823967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824051, "dur": 1,
    "args": {
      "External id": 152973,
      "cbid": 251, "correlation": 152973
    }
  },
  {
    "ph": "f", "id": 152973, "pid": 76337, "tid": -914061504, "ts": 1716454223824051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223826066, "dur": 188,
    "args": {
      "External id": 152974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152974,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152974, "pid": 5, "tid": 7, "ts": 1716454223826066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824055, "dur": 13,
    "args": {
      "External id": 152974,
      "cbid": 211, "correlation": 152974
    }
  },
  {
    "ph": "s", "id": 152974, "pid": 76337, "tid": -914061504, "ts": 1716454223824055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824121, "dur": 1,
    "args": {
      "External id": 152985,
      "cbid": 251, "correlation": 152985
    }
  },
  {
    "ph": "f", "id": 152985, "pid": 76337, "tid": -914061504, "ts": 1716454223824121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223826255, "dur": 191,
    "args": {
      "External id": 152986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 152986,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 152986, "pid": 5, "tid": 7, "ts": 1716454223826255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824124, "dur": 11,
    "args": {
      "External id": 152986,
      "cbid": 211, "correlation": 152986
    }
  },
  {
    "ph": "s", "id": 152986, "pid": 76337, "tid": -914061504, "ts": 1716454223824124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223826447, "dur": 18561,
    "args": {
      "External id": 153007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153007,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153007, "pid": 5, "tid": 7, "ts": 1716454223826447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824232, "dur": 14,
    "args": {
      "External id": 153007,
      "cbid": 211, "correlation": 153007
    }
  },
  {
    "ph": "s", "id": 153007, "pid": 76337, "tid": -914061504, "ts": 1716454223824232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824343, "dur": 2,
    "args": {
      "External id": 153025,
      "cbid": 251, "correlation": 153025
    }
  },
  {
    "ph": "f", "id": 153025, "pid": 76337, "tid": -914061504, "ts": 1716454223824343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223845009, "dur": 203,
    "args": {
      "External id": 153027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153027,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153027, "pid": 5, "tid": 7, "ts": 1716454223845009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824349, "dur": 14,
    "args": {
      "External id": 153027,
      "cbid": 211, "correlation": 153027
    }
  },
  {
    "ph": "s", "id": 153027, "pid": 76337, "tid": -914061504, "ts": 1716454223824349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223845213, "dur": 66,
    "args": {
      "External id": 153035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153035,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153035, "pid": 5, "tid": 7, "ts": 1716454223845213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824422, "dur": 12,
    "args": {
      "External id": 153035,
      "cbid": 211, "correlation": 153035
    }
  },
  {
    "ph": "s", "id": 153035, "pid": 76337, "tid": -914061504, "ts": 1716454223824422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223845281, "dur": 97,
    "args": {
      "External id": 153043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153043,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153043, "pid": 5, "tid": 7, "ts": 1716454223845281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824463, "dur": 8,
    "args": {
      "External id": 153043,
      "cbid": 211, "correlation": 153043
    }
  },
  {
    "ph": "s", "id": 153043, "pid": 76337, "tid": -914061504, "ts": 1716454223824463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223845379, "dur": 55,
    "args": {
      "External id": 153054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153054,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153054, "pid": 5, "tid": 7, "ts": 1716454223845379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824548, "dur": 16,
    "args": {
      "External id": 153054,
      "cbid": 211, "correlation": 153054
    }
  },
  {
    "ph": "s", "id": 153054, "pid": 76337, "tid": -914061504, "ts": 1716454223824548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223845435, "dur": 91,
    "args": {
      "External id": 153076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153076,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153076, "pid": 5, "tid": 7, "ts": 1716454223845435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824583, "dur": 7,
    "args": {
      "External id": 153076,
      "cbid": 211, "correlation": 153076
    }
  },
  {
    "ph": "s", "id": 153076, "pid": 76337, "tid": -914061504, "ts": 1716454223824583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824666, "dur": 1,
    "args": {
      "External id": 153087,
      "cbid": 251, "correlation": 153087
    }
  },
  {
    "ph": "f", "id": 153087, "pid": 76337, "tid": -914061504, "ts": 1716454223824666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223845528, "dur": 103,
    "args": {
      "External id": 153088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153088,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153088, "pid": 5, "tid": 7, "ts": 1716454223845528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824671, "dur": 13,
    "args": {
      "External id": 153088,
      "cbid": 211, "correlation": 153088
    }
  },
  {
    "ph": "s", "id": 153088, "pid": 76337, "tid": -914061504, "ts": 1716454223824671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824749, "dur": 2,
    "args": {
      "External id": 153099,
      "cbid": 251, "correlation": 153099
    }
  },
  {
    "ph": "f", "id": 153099, "pid": 76337, "tid": -914061504, "ts": 1716454223824749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824754, "dur": 0,
    "args": {
      "External id": 153100,
      "cbid": 251, "correlation": 153100
    }
  },
  {
    "ph": "f", "id": 153100, "pid": 76337, "tid": -914061504, "ts": 1716454223824754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223845633, "dur": 10,
    "args": {
      "External id": 153101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153101,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 153101, "pid": 5, "tid": 7, "ts": 1716454223845633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824756, "dur": 13,
    "args": {
      "External id": 153101,
      "cbid": 211, "correlation": 153101
    }
  },
  {
    "ph": "s", "id": 153101, "pid": 76337, "tid": -914061504, "ts": 1716454223824756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223845644, "dur": 5,
    "args": {
      "External id": 153103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153103,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 153103, "pid": 5, "tid": 7, "ts": 1716454223845644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824773, "dur": 7,
    "args": {
      "External id": 153103,
      "cbid": 211, "correlation": 153103
    }
  },
  {
    "ph": "s", "id": 153103, "pid": 76337, "tid": -914061504, "ts": 1716454223824773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824834, "dur": 1,
    "args": {
      "External id": 153114,
      "cbid": 251, "correlation": 153114
    }
  },
  {
    "ph": "f", "id": 153114, "pid": 76337, "tid": -914061504, "ts": 1716454223824834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824838, "dur": 0,
    "args": {
      "External id": 153115,
      "cbid": 251, "correlation": 153115
    }
  },
  {
    "ph": "f", "id": 153115, "pid": 76337, "tid": -914061504, "ts": 1716454223824838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223845650, "dur": 6,
    "args": {
      "External id": 153116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153116,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 153116, "pid": 5, "tid": 7, "ts": 1716454223845650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824839, "dur": 12,
    "args": {
      "External id": 153116,
      "cbid": 211, "correlation": 153116
    }
  },
  {
    "ph": "s", "id": 153116, "pid": 76337, "tid": -914061504, "ts": 1716454223824839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223845658, "dur": 3,
    "args": {
      "External id": 153118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153118,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 153118, "pid": 5, "tid": 7, "ts": 1716454223845658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824853, "dur": 7,
    "args": {
      "External id": 153118,
      "cbid": 211, "correlation": 153118
    }
  },
  {
    "ph": "s", "id": 153118, "pid": 76337, "tid": -914061504, "ts": 1716454223824853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223845662, "dur": 154,
    "args": {
      "External id": 153139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153139,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153139, "pid": 5, "tid": 7, "ts": 1716454223845662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223824929, "dur": 12,
    "args": {
      "External id": 153139,
      "cbid": 211, "correlation": 153139
    }
  },
  {
    "ph": "s", "id": 153139, "pid": 76337, "tid": -914061504, "ts": 1716454223824929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825036, "dur": 2,
    "args": {
      "External id": 153157,
      "cbid": 251, "correlation": 153157
    }
  },
  {
    "ph": "f", "id": 153157, "pid": 76337, "tid": -914061504, "ts": 1716454223825036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223845818, "dur": 105,
    "args": {
      "External id": 153159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153159,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153159, "pid": 5, "tid": 7, "ts": 1716454223845818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825043, "dur": 15,
    "args": {
      "External id": 153159,
      "cbid": 211, "correlation": 153159
    }
  },
  {
    "ph": "s", "id": 153159, "pid": 76337, "tid": -914061504, "ts": 1716454223825043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223845924, "dur": 35,
    "args": {
      "External id": 153167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153167,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153167, "pid": 5, "tid": 7, "ts": 1716454223845924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825115, "dur": 12,
    "args": {
      "External id": 153167,
      "cbid": 211, "correlation": 153167
    }
  },
  {
    "ph": "s", "id": 153167, "pid": 76337, "tid": -914061504, "ts": 1716454223825115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223845960, "dur": 68,
    "args": {
      "External id": 153175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153175, "pid": 5, "tid": 7, "ts": 1716454223845960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825155, "dur": 9,
    "args": {
      "External id": 153175,
      "cbid": 211, "correlation": 153175
    }
  },
  {
    "ph": "s", "id": 153175, "pid": 76337, "tid": -914061504, "ts": 1716454223825155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223846029, "dur": 92,
    "args": {
      "External id": 153197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153197,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153197, "pid": 5, "tid": 7, "ts": 1716454223846029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825206, "dur": 10,
    "args": {
      "External id": 153197,
      "cbid": 211, "correlation": 153197
    }
  },
  {
    "ph": "s", "id": 153197, "pid": 76337, "tid": -914061504, "ts": 1716454223825206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825297, "dur": 1,
    "args": {
      "External id": 153213,
      "cbid": 251, "correlation": 153213
    }
  },
  {
    "ph": "f", "id": 153213, "pid": 76337, "tid": -914061504, "ts": 1716454223825297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223846122, "dur": 582,
    "args": {
      "External id": 153215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153215,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153215, "pid": 5, "tid": 7, "ts": 1716454223846122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825302, "dur": 13,
    "args": {
      "External id": 153215,
      "cbid": 211, "correlation": 153215
    }
  },
  {
    "ph": "s", "id": 153215, "pid": 76337, "tid": -914061504, "ts": 1716454223825302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223846705, "dur": 244,
    "args": {
      "External id": 153223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153223, "pid": 5, "tid": 7, "ts": 1716454223846705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825380, "dur": 14,
    "args": {
      "External id": 153223,
      "cbid": 211, "correlation": 153223
    }
  },
  {
    "ph": "s", "id": 153223, "pid": 76337, "tid": -914061504, "ts": 1716454223825380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223846950, "dur": 251,
    "args": {
      "External id": 153231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153231, "pid": 5, "tid": 7, "ts": 1716454223846950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825415, "dur": 8,
    "args": {
      "External id": 153231,
      "cbid": 211, "correlation": 153231
    }
  },
  {
    "ph": "s", "id": 153231, "pid": 76337, "tid": -914061504, "ts": 1716454223825415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825497, "dur": 2,
    "args": {
      "External id": 153247,
      "cbid": 251, "correlation": 153247
    }
  },
  {
    "ph": "f", "id": 153247, "pid": 76337, "tid": -914061504, "ts": 1716454223825497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825503, "dur": 0,
    "args": {
      "External id": 153249,
      "cbid": 251, "correlation": 153249
    }
  },
  {
    "ph": "f", "id": 153249, "pid": 76337, "tid": -914061504, "ts": 1716454223825503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223847203, "dur": 360,
    "args": {
      "External id": 153250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153250,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 153250, "pid": 5, "tid": 7, "ts": 1716454223847203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825507, "dur": 13,
    "args": {
      "External id": 153250,
      "cbid": 211, "correlation": 153250
    }
  },
  {
    "ph": "s", "id": 153250, "pid": 76337, "tid": -914061504, "ts": 1716454223825507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223847565, "dur": 50,
    "args": {
      "External id": 153258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153258,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153258, "pid": 5, "tid": 7, "ts": 1716454223847565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825551, "dur": 10,
    "args": {
      "External id": 153258,
      "cbid": 211, "correlation": 153258
    }
  },
  {
    "ph": "s", "id": 153258, "pid": 76337, "tid": -914061504, "ts": 1716454223825551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223847616, "dur": 158,
    "args": {
      "External id": 153269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153269,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153269, "pid": 5, "tid": 7, "ts": 1716454223847616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825620, "dur": 12,
    "args": {
      "External id": 153269,
      "cbid": 211, "correlation": 153269
    }
  },
  {
    "ph": "s", "id": 153269, "pid": 76337, "tid": -914061504, "ts": 1716454223825620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825684, "dur": 0,
    "args": {
      "External id": 153281,
      "cbid": 317, "correlation": 153281
    }
  },
  {
    "ph": "f", "id": 153281, "pid": 76337, "tid": -914061504, "ts": 1716454223825684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825685, "dur": 0,
    "args": {
      "External id": 153282,
      "cbid": 203, "correlation": 153282
    }
  },
  {
    "ph": "f", "id": 153282, "pid": 76337, "tid": -914061504, "ts": 1716454223825685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825686, "dur": 0,
    "args": {
      "External id": 153283,
      "cbid": 205, "correlation": 153283
    }
  },
  {
    "ph": "f", "id": 153283, "pid": 76337, "tid": -914061504, "ts": 1716454223825686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825710, "dur": 1,
    "args": {
      "External id": 153287,
      "cbid": 251, "correlation": 153287
    }
  },
  {
    "ph": "f", "id": 153287, "pid": 76337, "tid": -914061504, "ts": 1716454223825710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825712, "dur": 0,
    "args": {
      "External id": 153288,
      "cbid": 251, "correlation": 153288
    }
  },
  {
    "ph": "f", "id": 153288, "pid": 76337, "tid": -914061504, "ts": 1716454223825712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825712, "dur": 0,
    "args": {
      "External id": 153289,
      "cbid": 251, "correlation": 153289
    }
  },
  {
    "ph": "f", "id": 153289, "pid": 76337, "tid": -914061504, "ts": 1716454223825712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825713, "dur": 0,
    "args": {
      "External id": 153290,
      "cbid": 251, "correlation": 153290
    }
  },
  {
    "ph": "f", "id": 153290, "pid": 76337, "tid": -914061504, "ts": 1716454223825713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825714, "dur": 0,
    "args": {
      "External id": 153291,
      "cbid": 251, "correlation": 153291
    }
  },
  {
    "ph": "f", "id": 153291, "pid": 76337, "tid": -914061504, "ts": 1716454223825714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825715, "dur": 0,
    "args": {
      "External id": 153292,
      "cbid": 251, "correlation": 153292
    }
  },
  {
    "ph": "f", "id": 153292, "pid": 76337, "tid": -914061504, "ts": 1716454223825715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825716, "dur": 0,
    "args": {
      "External id": 153293,
      "cbid": 251, "correlation": 153293
    }
  },
  {
    "ph": "f", "id": 153293, "pid": 76337, "tid": -914061504, "ts": 1716454223825716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825717, "dur": 0,
    "args": {
      "External id": 153294,
      "cbid": 251, "correlation": 153294
    }
  },
  {
    "ph": "f", "id": 153294, "pid": 76337, "tid": -914061504, "ts": 1716454223825717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825718, "dur": 0,
    "args": {
      "External id": 153295,
      "cbid": 251, "correlation": 153295
    }
  },
  {
    "ph": "f", "id": 153295, "pid": 76337, "tid": -914061504, "ts": 1716454223825718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223847775, "dur": 114,
    "args": {
      "External id": 153296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153296,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153296, "pid": 5, "tid": 7, "ts": 1716454223847775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825720, "dur": 12,
    "args": {
      "External id": 153296,
      "cbid": 211, "correlation": 153296
    }
  },
  {
    "ph": "s", "id": 153296, "pid": 76337, "tid": -914061504, "ts": 1716454223825720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223847890, "dur": 60,
    "args": {
      "External id": 153302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153302,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153302, "pid": 5, "tid": 7, "ts": 1716454223847890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825755, "dur": 9,
    "args": {
      "External id": 153302,
      "cbid": 211, "correlation": 153302
    }
  },
  {
    "ph": "s", "id": 153302, "pid": 76337, "tid": -914061504, "ts": 1716454223825755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223847951, "dur": 49,
    "args": {
      "External id": 153310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153310,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153310, "pid": 5, "tid": 7, "ts": 1716454223847951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825788, "dur": 8,
    "args": {
      "External id": 153310,
      "cbid": 211, "correlation": 153310
    }
  },
  {
    "ph": "s", "id": 153310, "pid": 76337, "tid": -914061504, "ts": 1716454223825788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223848001, "dur": 51,
    "args": {
      "External id": 153330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153330,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 153330, "pid": 5, "tid": 7, "ts": 1716454223848001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825862, "dur": 12,
    "args": {
      "External id": 153330,
      "cbid": 211, "correlation": 153330
    }
  },
  {
    "ph": "s", "id": 153330, "pid": 76337, "tid": -914061504, "ts": 1716454223825862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223848054, "dur": 4,
    "args": {
      "External id": 153342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153342,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 153342, "pid": 5, "tid": 7, "ts": 1716454223848054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825883, "dur": 6,
    "args": {
      "External id": 153342,
      "cbid": 211, "correlation": 153342
    }
  },
  {
    "ph": "s", "id": 153342, "pid": 76337, "tid": -914061504, "ts": 1716454223825883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223848060, "dur": 55,
    "args": {
      "External id": 153345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153345,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153345, "pid": 5, "tid": 7, "ts": 1716454223848060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825901, "dur": 8,
    "args": {
      "External id": 153345,
      "cbid": 211, "correlation": 153345
    }
  },
  {
    "ph": "s", "id": 153345, "pid": 76337, "tid": -914061504, "ts": 1716454223825901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223848116, "dur": 36,
    "args": {
      "External id": 153354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153354,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153354, "pid": 5, "tid": 7, "ts": 1716454223848116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223825941, "dur": 10,
    "args": {
      "External id": 153354,
      "cbid": 211, "correlation": 153354
    }
  },
  {
    "ph": "s", "id": 153354, "pid": 76337, "tid": -914061504, "ts": 1716454223825941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826001, "dur": 0,
    "args": {
      "External id": 153364,
      "cbid": 317, "correlation": 153364
    }
  },
  {
    "ph": "f", "id": 153364, "pid": 76337, "tid": -914061504, "ts": 1716454223826001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826002, "dur": 0,
    "args": {
      "External id": 153365,
      "cbid": 203, "correlation": 153365
    }
  },
  {
    "ph": "f", "id": 153365, "pid": 76337, "tid": -914061504, "ts": 1716454223826002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826002, "dur": 0,
    "args": {
      "External id": 153366,
      "cbid": 205, "correlation": 153366
    }
  },
  {
    "ph": "f", "id": 153366, "pid": 76337, "tid": -914061504, "ts": 1716454223826002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223848154, "dur": 41,
    "args": {
      "External id": 153370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153370,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153370, "pid": 5, "tid": 7, "ts": 1716454223848154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826017, "dur": 12,
    "args": {
      "External id": 153370,
      "cbid": 211, "correlation": 153370
    }
  },
  {
    "ph": "s", "id": 153370, "pid": 76337, "tid": -914061504, "ts": 1716454223826017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223848196, "dur": 14,
    "args": {
      "External id": 153372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153372,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153372, "pid": 5, "tid": 7, "ts": 1716454223848196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826031, "dur": 5,
    "args": {
      "External id": 153372,
      "cbid": 211, "correlation": 153372
    }
  },
  {
    "ph": "s", "id": 153372, "pid": 76337, "tid": -914061504, "ts": 1716454223826031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223848212, "dur": 3,
    "args": {
      "External id": 153374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 153374, "pid": 5, "tid": 7, "ts": 1716454223848212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826041, "dur": 6,
    "args": {
      "External id": 153374,
      "cbid": 211, "correlation": 153374
    }
  },
  {
    "ph": "s", "id": 153374, "pid": 76337, "tid": -914061504, "ts": 1716454223826041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826050, "dur": 0,
    "args": {
      "External id": 153375,
      "cbid": 51, "correlation": 153375
    }
  },
  {
    "ph": "s", "id": 153375, "pid": 76337, "tid": -914061504, "ts": 1716454223826050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223848216, "dur": 699,
    "args": {
      "External id": 153376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153376,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153376, "pid": 5, "tid": 7, "ts": 1716454223848216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826050, "dur": 6,
    "args": {
      "External id": 153376,
      "cbid": 211, "correlation": 153376
    }
  },
  {
    "ph": "s", "id": 153376, "pid": 76337, "tid": -914061504, "ts": 1716454223826050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223848916, "dur": 59,
    "args": {
      "External id": 153381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153381, "pid": 5, "tid": 7, "ts": 1716454223848916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826079, "dur": 9,
    "args": {
      "External id": 153381,
      "cbid": 211, "correlation": 153381
    }
  },
  {
    "ph": "s", "id": 153381, "pid": 76337, "tid": -914061504, "ts": 1716454223826079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223848976, "dur": 4,
    "args": {
      "External id": 153389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153389,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 153389, "pid": 5, "tid": 7, "ts": 1716454223848976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826123, "dur": 10,
    "args": {
      "External id": 153389,
      "cbid": 211, "correlation": 153389
    }
  },
  {
    "ph": "s", "id": 153389, "pid": 76337, "tid": -914061504, "ts": 1716454223826123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826188, "dur": 1,
    "args": {
      "External id": 153405,
      "cbid": 251, "correlation": 153405
    }
  },
  {
    "ph": "f", "id": 153405, "pid": 76337, "tid": -914061504, "ts": 1716454223826188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826192, "dur": 0,
    "args": {
      "External id": 153407,
      "cbid": 251, "correlation": 153407
    }
  },
  {
    "ph": "f", "id": 153407, "pid": 76337, "tid": -914061504, "ts": 1716454223826192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223848981, "dur": 11,
    "args": {
      "External id": 153408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153408,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 153408, "pid": 5, "tid": 7, "ts": 1716454223848981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826194, "dur": 12,
    "args": {
      "External id": 153408,
      "cbid": 211, "correlation": 153408
    }
  },
  {
    "ph": "s", "id": 153408, "pid": 76337, "tid": -914061504, "ts": 1716454223826194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223848993, "dur": 5,
    "args": {
      "External id": 153410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153410,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 153410, "pid": 5, "tid": 7, "ts": 1716454223848993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826208, "dur": 6,
    "args": {
      "External id": 153410,
      "cbid": 211, "correlation": 153410
    }
  },
  {
    "ph": "s", "id": 153410, "pid": 76337, "tid": -914061504, "ts": 1716454223826208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223849000, "dur": 51,
    "args": {
      "External id": 153420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153420, "pid": 5, "tid": 7, "ts": 1716454223849000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826265, "dur": 12,
    "args": {
      "External id": 153420,
      "cbid": 211, "correlation": 153420
    }
  },
  {
    "ph": "s", "id": 153420, "pid": 76337, "tid": -914061504, "ts": 1716454223826265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223849053, "dur": 51,
    "args": {
      "External id": 153440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153440,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 153440, "pid": 5, "tid": 7, "ts": 1716454223849053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826331, "dur": 10,
    "args": {
      "External id": 153440,
      "cbid": 211, "correlation": 153440
    }
  },
  {
    "ph": "s", "id": 153440, "pid": 76337, "tid": -914061504, "ts": 1716454223826331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223849105, "dur": 4,
    "args": {
      "External id": 153452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153452,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 153452, "pid": 5, "tid": 7, "ts": 1716454223849105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826352, "dur": 6,
    "args": {
      "External id": 153452,
      "cbid": 211, "correlation": 153452
    }
  },
  {
    "ph": "s", "id": 153452, "pid": 76337, "tid": -914061504, "ts": 1716454223826352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223849111, "dur": 55,
    "args": {
      "External id": 153455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153455, "pid": 5, "tid": 7, "ts": 1716454223849111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826371, "dur": 7,
    "args": {
      "External id": 153455,
      "cbid": 211, "correlation": 153455
    }
  },
  {
    "ph": "s", "id": 153455, "pid": 76337, "tid": -914061504, "ts": 1716454223826371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223849167, "dur": 36,
    "args": {
      "External id": 153464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153464,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153464, "pid": 5, "tid": 7, "ts": 1716454223849167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826411, "dur": 10,
    "args": {
      "External id": 153464,
      "cbid": 211, "correlation": 153464
    }
  },
  {
    "ph": "s", "id": 153464, "pid": 76337, "tid": -914061504, "ts": 1716454223826411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826474, "dur": 0,
    "args": {
      "External id": 153474,
      "cbid": 317, "correlation": 153474
    }
  },
  {
    "ph": "f", "id": 153474, "pid": 76337, "tid": -914061504, "ts": 1716454223826474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826474, "dur": 0,
    "args": {
      "External id": 153475,
      "cbid": 203, "correlation": 153475
    }
  },
  {
    "ph": "f", "id": 153475, "pid": 76337, "tid": -914061504, "ts": 1716454223826474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826475, "dur": 0,
    "args": {
      "External id": 153476,
      "cbid": 205, "correlation": 153476
    }
  },
  {
    "ph": "f", "id": 153476, "pid": 76337, "tid": -914061504, "ts": 1716454223826475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223849204, "dur": 40,
    "args": {
      "External id": 153480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153480,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153480, "pid": 5, "tid": 7, "ts": 1716454223849204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826489, "dur": 12,
    "args": {
      "External id": 153480,
      "cbid": 211, "correlation": 153480
    }
  },
  {
    "ph": "s", "id": 153480, "pid": 76337, "tid": -914061504, "ts": 1716454223826489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223849245, "dur": 14,
    "args": {
      "External id": 153482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153482,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153482, "pid": 5, "tid": 7, "ts": 1716454223849245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826503, "dur": 6,
    "args": {
      "External id": 153482,
      "cbid": 211, "correlation": 153482
    }
  },
  {
    "ph": "s", "id": 153482, "pid": 76337, "tid": -914061504, "ts": 1716454223826503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223849260, "dur": 3,
    "args": {
      "External id": 153484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 153484, "pid": 5, "tid": 7, "ts": 1716454223849260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826513, "dur": 5,
    "args": {
      "External id": 153484,
      "cbid": 211, "correlation": 153484
    }
  },
  {
    "ph": "s", "id": 153484, "pid": 76337, "tid": -914061504, "ts": 1716454223826513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826522, "dur": 0,
    "args": {
      "External id": 153485,
      "cbid": 51, "correlation": 153485
    }
  },
  {
    "ph": "s", "id": 153485, "pid": 76337, "tid": -914061504, "ts": 1716454223826522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223849265, "dur": 693,
    "args": {
      "External id": 153486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153486,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153486, "pid": 5, "tid": 7, "ts": 1716454223849265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826523, "dur": 5,
    "args": {
      "External id": 153486,
      "cbid": 211, "correlation": 153486
    }
  },
  {
    "ph": "s", "id": 153486, "pid": 76337, "tid": -914061504, "ts": 1716454223826523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223849960, "dur": 58,
    "args": {
      "External id": 153491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153491, "pid": 5, "tid": 7, "ts": 1716454223849960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826550, "dur": 9,
    "args": {
      "External id": 153491,
      "cbid": 211, "correlation": 153491
    }
  },
  {
    "ph": "s", "id": 153491, "pid": 76337, "tid": -914061504, "ts": 1716454223826550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223850019, "dur": 50,
    "args": {
      "External id": 153499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153499,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153499, "pid": 5, "tid": 7, "ts": 1716454223850019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826583, "dur": 8,
    "args": {
      "External id": 153499,
      "cbid": 211, "correlation": 153499
    }
  },
  {
    "ph": "s", "id": 153499, "pid": 76337, "tid": -914061504, "ts": 1716454223826583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223850070, "dur": 35,
    "args": {
      "External id": 153507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153507,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153507, "pid": 5, "tid": 7, "ts": 1716454223850070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826612, "dur": 8,
    "args": {
      "External id": 153507,
      "cbid": 211, "correlation": 153507
    }
  },
  {
    "ph": "s", "id": 153507, "pid": 76337, "tid": -914061504, "ts": 1716454223826612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223850106, "dur": 51,
    "args": {
      "External id": 153527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153527,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 153527, "pid": 5, "tid": 7, "ts": 1716454223850106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826690, "dur": 12,
    "args": {
      "External id": 153527,
      "cbid": 211, "correlation": 153527
    }
  },
  {
    "ph": "s", "id": 153527, "pid": 76337, "tid": -914061504, "ts": 1716454223826690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223850158, "dur": 4,
    "args": {
      "External id": 153539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153539,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 153539, "pid": 5, "tid": 7, "ts": 1716454223850158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826712, "dur": 7,
    "args": {
      "External id": 153539,
      "cbid": 211, "correlation": 153539
    }
  },
  {
    "ph": "s", "id": 153539, "pid": 76337, "tid": -914061504, "ts": 1716454223826712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223850163, "dur": 55,
    "args": {
      "External id": 153542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153542, "pid": 5, "tid": 7, "ts": 1716454223850163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826731, "dur": 6,
    "args": {
      "External id": 153542,
      "cbid": 211, "correlation": 153542
    }
  },
  {
    "ph": "s", "id": 153542, "pid": 76337, "tid": -914061504, "ts": 1716454223826731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826788, "dur": 0,
    "args": {
      "External id": 153553,
      "cbid": 317, "correlation": 153553
    }
  },
  {
    "ph": "f", "id": 153553, "pid": 76337, "tid": -914061504, "ts": 1716454223826788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826789, "dur": 0,
    "args": {
      "External id": 153554,
      "cbid": 203, "correlation": 153554
    }
  },
  {
    "ph": "f", "id": 153554, "pid": 76337, "tid": -914061504, "ts": 1716454223826789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826789, "dur": 0,
    "args": {
      "External id": 153555,
      "cbid": 205, "correlation": 153555
    }
  },
  {
    "ph": "f", "id": 153555, "pid": 76337, "tid": -914061504, "ts": 1716454223826789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826811, "dur": 1,
    "args": {
      "External id": 153559,
      "cbid": 251, "correlation": 153559
    }
  },
  {
    "ph": "f", "id": 153559, "pid": 76337, "tid": -914061504, "ts": 1716454223826811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826813, "dur": 0,
    "args": {
      "External id": 153560,
      "cbid": 251, "correlation": 153560
    }
  },
  {
    "ph": "f", "id": 153560, "pid": 76337, "tid": -914061504, "ts": 1716454223826813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826814, "dur": 0,
    "args": {
      "External id": 153561,
      "cbid": 251, "correlation": 153561
    }
  },
  {
    "ph": "f", "id": 153561, "pid": 76337, "tid": -914061504, "ts": 1716454223826814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826815, "dur": 0,
    "args": {
      "External id": 153562,
      "cbid": 251, "correlation": 153562
    }
  },
  {
    "ph": "f", "id": 153562, "pid": 76337, "tid": -914061504, "ts": 1716454223826815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826815, "dur": 0,
    "args": {
      "External id": 153563,
      "cbid": 251, "correlation": 153563
    }
  },
  {
    "ph": "f", "id": 153563, "pid": 76337, "tid": -914061504, "ts": 1716454223826815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826816, "dur": 0,
    "args": {
      "External id": 153564,
      "cbid": 251, "correlation": 153564
    }
  },
  {
    "ph": "f", "id": 153564, "pid": 76337, "tid": -914061504, "ts": 1716454223826816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826817, "dur": 0,
    "args": {
      "External id": 153565,
      "cbid": 251, "correlation": 153565
    }
  },
  {
    "ph": "f", "id": 153565, "pid": 76337, "tid": -914061504, "ts": 1716454223826817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826818, "dur": 0,
    "args": {
      "External id": 153566,
      "cbid": 251, "correlation": 153566
    }
  },
  {
    "ph": "f", "id": 153566, "pid": 76337, "tid": -914061504, "ts": 1716454223826818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826819, "dur": 0,
    "args": {
      "External id": 153567,
      "cbid": 251, "correlation": 153567
    }
  },
  {
    "ph": "f", "id": 153567, "pid": 76337, "tid": -914061504, "ts": 1716454223826819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223850220, "dur": 114,
    "args": {
      "External id": 153568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153568,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153568, "pid": 5, "tid": 7, "ts": 1716454223850220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826822, "dur": 12,
    "args": {
      "External id": 153568,
      "cbid": 211, "correlation": 153568
    }
  },
  {
    "ph": "s", "id": 153568, "pid": 76337, "tid": -914061504, "ts": 1716454223826822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223850335, "dur": 60,
    "args": {
      "External id": 153574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153574,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153574, "pid": 5, "tid": 7, "ts": 1716454223850335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826856, "dur": 9,
    "args": {
      "External id": 153574,
      "cbid": 211, "correlation": 153574
    }
  },
  {
    "ph": "s", "id": 153574, "pid": 76337, "tid": -914061504, "ts": 1716454223826856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223850396, "dur": 676,
    "args": {
      "External id": 153583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153583,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153583, "pid": 5, "tid": 7, "ts": 1716454223850396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223826939, "dur": 13,
    "args": {
      "External id": 153583,
      "cbid": 211, "correlation": 153583
    }
  },
  {
    "ph": "s", "id": 153583, "pid": 76337, "tid": -914061504, "ts": 1716454223826939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223851073, "dur": 180,
    "args": {
      "External id": 153605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153605,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153605, "pid": 5, "tid": 7, "ts": 1716454223851073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827003, "dur": 10,
    "args": {
      "External id": 153605,
      "cbid": 211, "correlation": 153605
    }
  },
  {
    "ph": "s", "id": 153605, "pid": 76337, "tid": -914061504, "ts": 1716454223827003,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827090, "dur": 1,
    "args": {
      "External id": 153616,
      "cbid": 251, "correlation": 153616
    }
  },
  {
    "ph": "f", "id": 153616, "pid": 76337, "tid": -914061504, "ts": 1716454223827090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223851255, "dur": 196,
    "args": {
      "External id": 153617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153617,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153617, "pid": 5, "tid": 7, "ts": 1716454223851255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827095, "dur": 14,
    "args": {
      "External id": 153617,
      "cbid": 211, "correlation": 153617
    }
  },
  {
    "ph": "s", "id": 153617, "pid": 76337, "tid": -914061504, "ts": 1716454223827095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827163, "dur": 1,
    "args": {
      "External id": 153628,
      "cbid": 251, "correlation": 153628
    }
  },
  {
    "ph": "f", "id": 153628, "pid": 76337, "tid": -914061504, "ts": 1716454223827163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223851453, "dur": 188,
    "args": {
      "External id": 153629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153629,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153629, "pid": 5, "tid": 7, "ts": 1716454223851453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827168, "dur": 12,
    "args": {
      "External id": 153629,
      "cbid": 211, "correlation": 153629
    }
  },
  {
    "ph": "s", "id": 153629, "pid": 76337, "tid": -914061504, "ts": 1716454223827168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827232, "dur": 1,
    "args": {
      "External id": 153640,
      "cbid": 251, "correlation": 153640
    }
  },
  {
    "ph": "f", "id": 153640, "pid": 76337, "tid": -914061504, "ts": 1716454223827232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223851642, "dur": 190,
    "args": {
      "External id": 153641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153641,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153641, "pid": 5, "tid": 7, "ts": 1716454223851642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827236, "dur": 11,
    "args": {
      "External id": 153641,
      "cbid": 211, "correlation": 153641
    }
  },
  {
    "ph": "s", "id": 153641, "pid": 76337, "tid": -914061504, "ts": 1716454223827236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223851833, "dur": 18577,
    "args": {
      "External id": 153662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153662,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153662, "pid": 5, "tid": 7, "ts": 1716454223851833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827316, "dur": 13,
    "args": {
      "External id": 153662,
      "cbid": 211, "correlation": 153662
    }
  },
  {
    "ph": "s", "id": 153662, "pid": 76337, "tid": -914061504, "ts": 1716454223827316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827413, "dur": 1,
    "args": {
      "External id": 153680,
      "cbid": 251, "correlation": 153680
    }
  },
  {
    "ph": "f", "id": 153680, "pid": 76337, "tid": -914061504, "ts": 1716454223827413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223870411, "dur": 198,
    "args": {
      "External id": 153682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153682,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153682, "pid": 5, "tid": 7, "ts": 1716454223870411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827419, "dur": 13,
    "args": {
      "External id": 153682,
      "cbid": 211, "correlation": 153682
    }
  },
  {
    "ph": "s", "id": 153682, "pid": 76337, "tid": -914061504, "ts": 1716454223827419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223870611, "dur": 66,
    "args": {
      "External id": 153690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153690,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153690, "pid": 5, "tid": 7, "ts": 1716454223870611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827489, "dur": 12,
    "args": {
      "External id": 153690,
      "cbid": 211, "correlation": 153690
    }
  },
  {
    "ph": "s", "id": 153690, "pid": 76337, "tid": -914061504, "ts": 1716454223827489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223870678, "dur": 96,
    "args": {
      "External id": 153698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153698,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153698, "pid": 5, "tid": 7, "ts": 1716454223870678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827528, "dur": 9,
    "args": {
      "External id": 153698,
      "cbid": 211, "correlation": 153698
    }
  },
  {
    "ph": "s", "id": 153698, "pid": 76337, "tid": -914061504, "ts": 1716454223827528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223870776, "dur": 54,
    "args": {
      "External id": 153709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153709, "pid": 5, "tid": 7, "ts": 1716454223870776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827600, "dur": 12,
    "args": {
      "External id": 153709,
      "cbid": 211, "correlation": 153709
    }
  },
  {
    "ph": "s", "id": 153709, "pid": 76337, "tid": -914061504, "ts": 1716454223827600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223870831, "dur": 92,
    "args": {
      "External id": 153731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153731,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153731, "pid": 5, "tid": 7, "ts": 1716454223870831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827632, "dur": 8,
    "args": {
      "External id": 153731,
      "cbid": 211, "correlation": 153731
    }
  },
  {
    "ph": "s", "id": 153731, "pid": 76337, "tid": -914061504, "ts": 1716454223827632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827714, "dur": 1,
    "args": {
      "External id": 153742,
      "cbid": 251, "correlation": 153742
    }
  },
  {
    "ph": "f", "id": 153742, "pid": 76337, "tid": -914061504, "ts": 1716454223827714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223870924, "dur": 106,
    "args": {
      "External id": 153743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153743,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153743, "pid": 5, "tid": 7, "ts": 1716454223870924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827720, "dur": 13,
    "args": {
      "External id": 153743,
      "cbid": 211, "correlation": 153743
    }
  },
  {
    "ph": "s", "id": 153743, "pid": 76337, "tid": -914061504, "ts": 1716454223827720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827791, "dur": 1,
    "args": {
      "External id": 153754,
      "cbid": 251, "correlation": 153754
    }
  },
  {
    "ph": "f", "id": 153754, "pid": 76337, "tid": -914061504, "ts": 1716454223827791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827795, "dur": 0,
    "args": {
      "External id": 153755,
      "cbid": 251, "correlation": 153755
    }
  },
  {
    "ph": "f", "id": 153755, "pid": 76337, "tid": -914061504, "ts": 1716454223827795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223871031, "dur": 10,
    "args": {
      "External id": 153756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153756,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 153756, "pid": 5, "tid": 7, "ts": 1716454223871031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827797, "dur": 12,
    "args": {
      "External id": 153756,
      "cbid": 211, "correlation": 153756
    }
  },
  {
    "ph": "s", "id": 153756, "pid": 76337, "tid": -914061504, "ts": 1716454223827797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223871043, "dur": 5,
    "args": {
      "External id": 153758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153758,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 153758, "pid": 5, "tid": 7, "ts": 1716454223871043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827811, "dur": 5,
    "args": {
      "External id": 153758,
      "cbid": 211, "correlation": 153758
    }
  },
  {
    "ph": "s", "id": 153758, "pid": 76337, "tid": -914061504, "ts": 1716454223827811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827871, "dur": 1,
    "args": {
      "External id": 153769,
      "cbid": 251, "correlation": 153769
    }
  },
  {
    "ph": "f", "id": 153769, "pid": 76337, "tid": -914061504, "ts": 1716454223827871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827874, "dur": 0,
    "args": {
      "External id": 153770,
      "cbid": 251, "correlation": 153770
    }
  },
  {
    "ph": "f", "id": 153770, "pid": 76337, "tid": -914061504, "ts": 1716454223827874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223871049, "dur": 6,
    "args": {
      "External id": 153771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153771,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 153771, "pid": 5, "tid": 7, "ts": 1716454223871049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827876, "dur": 12,
    "args": {
      "External id": 153771,
      "cbid": 211, "correlation": 153771
    }
  },
  {
    "ph": "s", "id": 153771, "pid": 76337, "tid": -914061504, "ts": 1716454223827876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223871057, "dur": 3,
    "args": {
      "External id": 153773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153773,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 153773, "pid": 5, "tid": 7, "ts": 1716454223871057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827890, "dur": 5,
    "args": {
      "External id": 153773,
      "cbid": 211, "correlation": 153773
    }
  },
  {
    "ph": "s", "id": 153773, "pid": 76337, "tid": -914061504, "ts": 1716454223827890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223871062, "dur": 155,
    "args": {
      "External id": 153794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153794,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153794, "pid": 5, "tid": 7, "ts": 1716454223871062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223827963, "dur": 21,
    "args": {
      "External id": 153794,
      "cbid": 211, "correlation": 153794
    }
  },
  {
    "ph": "s", "id": 153794, "pid": 76337, "tid": -914061504, "ts": 1716454223827963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828070, "dur": 1,
    "args": {
      "External id": 153812,
      "cbid": 251, "correlation": 153812
    }
  },
  {
    "ph": "f", "id": 153812, "pid": 76337, "tid": -914061504, "ts": 1716454223828070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223871218, "dur": 107,
    "args": {
      "External id": 153814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153814,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153814, "pid": 5, "tid": 7, "ts": 1716454223871218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828076, "dur": 13,
    "args": {
      "External id": 153814,
      "cbid": 211, "correlation": 153814
    }
  },
  {
    "ph": "s", "id": 153814, "pid": 76337, "tid": -914061504, "ts": 1716454223828076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223871327, "dur": 35,
    "args": {
      "External id": 153822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153822,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153822, "pid": 5, "tid": 7, "ts": 1716454223871327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828145, "dur": 12,
    "args": {
      "External id": 153822,
      "cbid": 211, "correlation": 153822
    }
  },
  {
    "ph": "s", "id": 153822, "pid": 76337, "tid": -914061504, "ts": 1716454223828145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223871363, "dur": 68,
    "args": {
      "External id": 153830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153830, "pid": 5, "tid": 7, "ts": 1716454223871363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828186, "dur": 9,
    "args": {
      "External id": 153830,
      "cbid": 211, "correlation": 153830
    }
  },
  {
    "ph": "s", "id": 153830, "pid": 76337, "tid": -914061504, "ts": 1716454223828186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223871433, "dur": 92,
    "args": {
      "External id": 153852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153852,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153852, "pid": 5, "tid": 7, "ts": 1716454223871433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828237, "dur": 10,
    "args": {
      "External id": 153852,
      "cbid": 211, "correlation": 153852
    }
  },
  {
    "ph": "s", "id": 153852, "pid": 76337, "tid": -914061504, "ts": 1716454223828237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828324, "dur": 1,
    "args": {
      "External id": 153868,
      "cbid": 251, "correlation": 153868
    }
  },
  {
    "ph": "f", "id": 153868, "pid": 76337, "tid": -914061504, "ts": 1716454223828324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223871526, "dur": 569,
    "args": {
      "External id": 153870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153870,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153870, "pid": 5, "tid": 7, "ts": 1716454223871526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828330, "dur": 12,
    "args": {
      "External id": 153870,
      "cbid": 211, "correlation": 153870
    }
  },
  {
    "ph": "s", "id": 153870, "pid": 76337, "tid": -914061504, "ts": 1716454223828330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223872096, "dur": 244,
    "args": {
      "External id": 153878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153878, "pid": 5, "tid": 7, "ts": 1716454223872096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828394, "dur": 12,
    "args": {
      "External id": 153878,
      "cbid": 211, "correlation": 153878
    }
  },
  {
    "ph": "s", "id": 153878, "pid": 76337, "tid": -914061504, "ts": 1716454223828394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223872341, "dur": 253,
    "args": {
      "External id": 153886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153886,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153886, "pid": 5, "tid": 7, "ts": 1716454223872341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828424, "dur": 9,
    "args": {
      "External id": 153886,
      "cbid": 211, "correlation": 153886
    }
  },
  {
    "ph": "s", "id": 153886, "pid": 76337, "tid": -914061504, "ts": 1716454223828424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828506, "dur": 1,
    "args": {
      "External id": 153902,
      "cbid": 251, "correlation": 153902
    }
  },
  {
    "ph": "f", "id": 153902, "pid": 76337, "tid": -914061504, "ts": 1716454223828506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828510, "dur": 0,
    "args": {
      "External id": 153904,
      "cbid": 251, "correlation": 153904
    }
  },
  {
    "ph": "f", "id": 153904, "pid": 76337, "tid": -914061504, "ts": 1716454223828510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223872595, "dur": 362,
    "args": {
      "External id": 153905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153905,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 153905, "pid": 5, "tid": 7, "ts": 1716454223872595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828513, "dur": 13,
    "args": {
      "External id": 153905,
      "cbid": 211, "correlation": 153905
    }
  },
  {
    "ph": "s", "id": 153905, "pid": 76337, "tid": -914061504, "ts": 1716454223828513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223872958, "dur": 49,
    "args": {
      "External id": 153913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153913,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153913, "pid": 5, "tid": 7, "ts": 1716454223872958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828555, "dur": 9,
    "args": {
      "External id": 153913,
      "cbid": 211, "correlation": 153913
    }
  },
  {
    "ph": "s", "id": 153913, "pid": 76337, "tid": -914061504, "ts": 1716454223828555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223873008, "dur": 158,
    "args": {
      "External id": 153924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153924, "pid": 5, "tid": 7, "ts": 1716454223873008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828623, "dur": 12,
    "args": {
      "External id": 153924,
      "cbid": 211, "correlation": 153924
    }
  },
  {
    "ph": "s", "id": 153924, "pid": 76337, "tid": -914061504, "ts": 1716454223828623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828686, "dur": 0,
    "args": {
      "External id": 153936,
      "cbid": 317, "correlation": 153936
    }
  },
  {
    "ph": "f", "id": 153936, "pid": 76337, "tid": -914061504, "ts": 1716454223828686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828687, "dur": 0,
    "args": {
      "External id": 153937,
      "cbid": 203, "correlation": 153937
    }
  },
  {
    "ph": "f", "id": 153937, "pid": 76337, "tid": -914061504, "ts": 1716454223828687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828687, "dur": 0,
    "args": {
      "External id": 153938,
      "cbid": 205, "correlation": 153938
    }
  },
  {
    "ph": "f", "id": 153938, "pid": 76337, "tid": -914061504, "ts": 1716454223828687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828710, "dur": 1,
    "args": {
      "External id": 153942,
      "cbid": 251, "correlation": 153942
    }
  },
  {
    "ph": "f", "id": 153942, "pid": 76337, "tid": -914061504, "ts": 1716454223828710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828712, "dur": 0,
    "args": {
      "External id": 153943,
      "cbid": 251, "correlation": 153943
    }
  },
  {
    "ph": "f", "id": 153943, "pid": 76337, "tid": -914061504, "ts": 1716454223828712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828712, "dur": 0,
    "args": {
      "External id": 153944,
      "cbid": 251, "correlation": 153944
    }
  },
  {
    "ph": "f", "id": 153944, "pid": 76337, "tid": -914061504, "ts": 1716454223828712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828713, "dur": 0,
    "args": {
      "External id": 153945,
      "cbid": 251, "correlation": 153945
    }
  },
  {
    "ph": "f", "id": 153945, "pid": 76337, "tid": -914061504, "ts": 1716454223828713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828714, "dur": 0,
    "args": {
      "External id": 153946,
      "cbid": 251, "correlation": 153946
    }
  },
  {
    "ph": "f", "id": 153946, "pid": 76337, "tid": -914061504, "ts": 1716454223828714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828715, "dur": 0,
    "args": {
      "External id": 153947,
      "cbid": 251, "correlation": 153947
    }
  },
  {
    "ph": "f", "id": 153947, "pid": 76337, "tid": -914061504, "ts": 1716454223828715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828715, "dur": 0,
    "args": {
      "External id": 153948,
      "cbid": 251, "correlation": 153948
    }
  },
  {
    "ph": "f", "id": 153948, "pid": 76337, "tid": -914061504, "ts": 1716454223828715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828716, "dur": 0,
    "args": {
      "External id": 153949,
      "cbid": 251, "correlation": 153949
    }
  },
  {
    "ph": "f", "id": 153949, "pid": 76337, "tid": -914061504, "ts": 1716454223828716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828718, "dur": 0,
    "args": {
      "External id": 153950,
      "cbid": 251, "correlation": 153950
    }
  },
  {
    "ph": "f", "id": 153950, "pid": 76337, "tid": -914061504, "ts": 1716454223828718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223873167, "dur": 115,
    "args": {
      "External id": 153951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153951,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 153951, "pid": 5, "tid": 7, "ts": 1716454223873167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828720, "dur": 13,
    "args": {
      "External id": 153951,
      "cbid": 211, "correlation": 153951
    }
  },
  {
    "ph": "s", "id": 153951, "pid": 76337, "tid": -914061504, "ts": 1716454223828720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223873284, "dur": 59,
    "args": {
      "External id": 153957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153957, "pid": 5, "tid": 7, "ts": 1716454223873284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828755, "dur": 9,
    "args": {
      "External id": 153957,
      "cbid": 211, "correlation": 153957
    }
  },
  {
    "ph": "s", "id": 153957, "pid": 76337, "tid": -914061504, "ts": 1716454223828755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223873345, "dur": 49,
    "args": {
      "External id": 153965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153965,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153965, "pid": 5, "tid": 7, "ts": 1716454223873345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828787, "dur": 8,
    "args": {
      "External id": 153965,
      "cbid": 211, "correlation": 153965
    }
  },
  {
    "ph": "s", "id": 153965, "pid": 76337, "tid": -914061504, "ts": 1716454223828787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828859, "dur": 0,
    "args": {
      "External id": 153975,
      "cbid": 317, "correlation": 153975
    }
  },
  {
    "ph": "f", "id": 153975, "pid": 76337, "tid": -914061504, "ts": 1716454223828859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828860, "dur": 0,
    "args": {
      "External id": 153976,
      "cbid": 203, "correlation": 153976
    }
  },
  {
    "ph": "f", "id": 153976, "pid": 76337, "tid": -914061504, "ts": 1716454223828860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828860, "dur": 0,
    "args": {
      "External id": 153977,
      "cbid": 205, "correlation": 153977
    }
  },
  {
    "ph": "f", "id": 153977, "pid": 76337, "tid": -914061504, "ts": 1716454223828860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223873396, "dur": 41,
    "args": {
      "External id": 153981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153981, "pid": 5, "tid": 7, "ts": 1716454223873396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828877, "dur": 12,
    "args": {
      "External id": 153981,
      "cbid": 211, "correlation": 153981
    }
  },
  {
    "ph": "s", "id": 153981, "pid": 76337, "tid": -914061504, "ts": 1716454223828877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223873438, "dur": 14,
    "args": {
      "External id": 153983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153983,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153983, "pid": 5, "tid": 7, "ts": 1716454223873438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828892, "dur": 5,
    "args": {
      "External id": 153983,
      "cbid": 211, "correlation": 153983
    }
  },
  {
    "ph": "s", "id": 153983, "pid": 76337, "tid": -914061504, "ts": 1716454223828892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223873454, "dur": 1,
    "args": {
      "External id": 153985,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 153985,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 153985, "pid": 5, "tid": 7, "ts": 1716454223873454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828911, "dur": 19,
    "args": {
      "External id": 153985,
      "cbid": 51, "correlation": 153985
    }
  },
  {
    "ph": "s", "id": 153985, "pid": 76337, "tid": -914061504, "ts": 1716454223828911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223873458, "dur": 361,
    "args": {
      "External id": 153986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153986,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 153986, "pid": 5, "tid": 7, "ts": 1716454223873458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828932, "dur": 9,
    "args": {
      "External id": 153986,
      "cbid": 211, "correlation": 153986
    }
  },
  {
    "ph": "s", "id": 153986, "pid": 76337, "tid": -914061504, "ts": 1716454223828932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223873820, "dur": 14,
    "args": {
      "External id": 153988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153988, "pid": 5, "tid": 7, "ts": 1716454223873820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828950, "dur": 8,
    "args": {
      "External id": 153988,
      "cbid": 211, "correlation": 153988
    }
  },
  {
    "ph": "s", "id": 153988, "pid": 76337, "tid": -914061504, "ts": 1716454223828950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223873836, "dur": 15,
    "args": {
      "External id": 153994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 153994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 153994, "pid": 5, "tid": 7, "ts": 1716454223873836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223828989, "dur": 9,
    "args": {
      "External id": 153994,
      "cbid": 211, "correlation": 153994
    }
  },
  {
    "ph": "s", "id": 153994, "pid": 76337, "tid": -914061504, "ts": 1716454223828989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223873852, "dur": 19,
    "args": {
      "External id": 154014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154014,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 154014, "pid": 5, "tid": 7, "ts": 1716454223873852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829083, "dur": 12,
    "args": {
      "External id": 154014,
      "cbid": 211, "correlation": 154014
    }
  },
  {
    "ph": "s", "id": 154014, "pid": 76337, "tid": -914061504, "ts": 1716454223829083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223873871, "dur": 4,
    "args": {
      "External id": 154026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154026,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 154026, "pid": 5, "tid": 7, "ts": 1716454223873871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829105, "dur": 7,
    "args": {
      "External id": 154026,
      "cbid": 211, "correlation": 154026
    }
  },
  {
    "ph": "s", "id": 154026, "pid": 76337, "tid": -914061504, "ts": 1716454223829105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223873877, "dur": 18,
    "args": {
      "External id": 154029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154029,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154029, "pid": 5, "tid": 7, "ts": 1716454223873877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829124, "dur": 7,
    "args": {
      "External id": 154029,
      "cbid": 211, "correlation": 154029
    }
  },
  {
    "ph": "s", "id": 154029, "pid": 76337, "tid": -914061504, "ts": 1716454223829124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223873897, "dur": 13,
    "args": {
      "External id": 154038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154038,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154038, "pid": 5, "tid": 7, "ts": 1716454223873897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829163, "dur": 10,
    "args": {
      "External id": 154038,
      "cbid": 211, "correlation": 154038
    }
  },
  {
    "ph": "s", "id": 154038, "pid": 76337, "tid": -914061504, "ts": 1716454223829163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829218, "dur": 0,
    "args": {
      "External id": 154048,
      "cbid": 317, "correlation": 154048
    }
  },
  {
    "ph": "f", "id": 154048, "pid": 76337, "tid": -914061504, "ts": 1716454223829218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829219, "dur": 0,
    "args": {
      "External id": 154049,
      "cbid": 203, "correlation": 154049
    }
  },
  {
    "ph": "f", "id": 154049, "pid": 76337, "tid": -914061504, "ts": 1716454223829219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829220, "dur": 0,
    "args": {
      "External id": 154050,
      "cbid": 205, "correlation": 154050
    }
  },
  {
    "ph": "f", "id": 154050, "pid": 76337, "tid": -914061504, "ts": 1716454223829220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223873911, "dur": 11,
    "args": {
      "External id": 154054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154054, "pid": 5, "tid": 7, "ts": 1716454223873911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829235, "dur": 11,
    "args": {
      "External id": 154054,
      "cbid": 211, "correlation": 154054
    }
  },
  {
    "ph": "s", "id": 154054, "pid": 76337, "tid": -914061504, "ts": 1716454223829235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223873923, "dur": 24,
    "args": {
      "External id": 154056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154056, "pid": 5, "tid": 7, "ts": 1716454223873923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829249, "dur": 6,
    "args": {
      "External id": 154056,
      "cbid": 211, "correlation": 154056
    }
  },
  {
    "ph": "s", "id": 154056, "pid": 76337, "tid": -914061504, "ts": 1716454223829249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223873949, "dur": 3,
    "args": {
      "External id": 154058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154058,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 154058, "pid": 5, "tid": 7, "ts": 1716454223873949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829261, "dur": 5,
    "args": {
      "External id": 154058,
      "cbid": 211, "correlation": 154058
    }
  },
  {
    "ph": "s", "id": 154058, "pid": 76337, "tid": -914061504, "ts": 1716454223829261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829270, "dur": 0,
    "args": {
      "External id": 154059,
      "cbid": 51, "correlation": 154059
    }
  },
  {
    "ph": "s", "id": 154059, "pid": 76337, "tid": -914061504, "ts": 1716454223829270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223873953, "dur": 357,
    "args": {
      "External id": 154060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154060,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154060, "pid": 5, "tid": 7, "ts": 1716454223873953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829271, "dur": 7,
    "args": {
      "External id": 154060,
      "cbid": 211, "correlation": 154060
    }
  },
  {
    "ph": "s", "id": 154060, "pid": 76337, "tid": -914061504, "ts": 1716454223829271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223874312, "dur": 20,
    "args": {
      "External id": 154061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154061,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154061, "pid": 5, "tid": 7, "ts": 1716454223874312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829281, "dur": 5,
    "args": {
      "External id": 154061,
      "cbid": 211, "correlation": 154061
    }
  },
  {
    "ph": "s", "id": 154061, "pid": 76337, "tid": -914061504, "ts": 1716454223829281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223874333, "dur": 33,
    "args": {
      "External id": 154067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154067, "pid": 5, "tid": 7, "ts": 1716454223874333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829308, "dur": 9,
    "args": {
      "External id": 154067,
      "cbid": 211, "correlation": 154067
    }
  },
  {
    "ph": "s", "id": 154067, "pid": 76337, "tid": -914061504, "ts": 1716454223829308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223874368, "dur": 3,
    "args": {
      "External id": 154075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154075,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 154075, "pid": 5, "tid": 7, "ts": 1716454223874368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829351, "dur": 10,
    "args": {
      "External id": 154075,
      "cbid": 211, "correlation": 154075
    }
  },
  {
    "ph": "s", "id": 154075, "pid": 76337, "tid": -914061504, "ts": 1716454223829351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829416, "dur": 1,
    "args": {
      "External id": 154091,
      "cbid": 251, "correlation": 154091
    }
  },
  {
    "ph": "f", "id": 154091, "pid": 76337, "tid": -914061504, "ts": 1716454223829416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829421, "dur": 0,
    "args": {
      "External id": 154093,
      "cbid": 251, "correlation": 154093
    }
  },
  {
    "ph": "f", "id": 154093, "pid": 76337, "tid": -914061504, "ts": 1716454223829421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223874372, "dur": 13,
    "args": {
      "External id": 154094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154094,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 154094, "pid": 5, "tid": 7, "ts": 1716454223874372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829423, "dur": 11,
    "args": {
      "External id": 154094,
      "cbid": 211, "correlation": 154094
    }
  },
  {
    "ph": "s", "id": 154094, "pid": 76337, "tid": -914061504, "ts": 1716454223829423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223874386, "dur": 5,
    "args": {
      "External id": 154096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 154096, "pid": 5, "tid": 7, "ts": 1716454223874386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829436, "dur": 6,
    "args": {
      "External id": 154096,
      "cbid": 211, "correlation": 154096
    }
  },
  {
    "ph": "s", "id": 154096, "pid": 76337, "tid": -914061504, "ts": 1716454223829436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223874392, "dur": 29,
    "args": {
      "External id": 154106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154106, "pid": 5, "tid": 7, "ts": 1716454223874392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829494, "dur": 13,
    "args": {
      "External id": 154106,
      "cbid": 211, "correlation": 154106
    }
  },
  {
    "ph": "s", "id": 154106, "pid": 76337, "tid": -914061504, "ts": 1716454223829494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223874423, "dur": 30,
    "args": {
      "External id": 154126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154126,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 154126, "pid": 5, "tid": 7, "ts": 1716454223874423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829563, "dur": 11,
    "args": {
      "External id": 154126,
      "cbid": 211, "correlation": 154126
    }
  },
  {
    "ph": "s", "id": 154126, "pid": 76337, "tid": -914061504, "ts": 1716454223829563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223874454, "dur": 4,
    "args": {
      "External id": 154138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154138,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 154138, "pid": 5, "tid": 7, "ts": 1716454223874454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829584, "dur": 6,
    "args": {
      "External id": 154138,
      "cbid": 211, "correlation": 154138
    }
  },
  {
    "ph": "s", "id": 154138, "pid": 76337, "tid": -914061504, "ts": 1716454223829584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223874460, "dur": 29,
    "args": {
      "External id": 154141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154141, "pid": 5, "tid": 7, "ts": 1716454223874460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829603, "dur": 8,
    "args": {
      "External id": 154141,
      "cbid": 211, "correlation": 154141
    }
  },
  {
    "ph": "s", "id": 154141, "pid": 76337, "tid": -914061504, "ts": 1716454223829603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223874490, "dur": 20,
    "args": {
      "External id": 154150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154150,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154150, "pid": 5, "tid": 7, "ts": 1716454223874490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829644, "dur": 10,
    "args": {
      "External id": 154150,
      "cbid": 211, "correlation": 154150
    }
  },
  {
    "ph": "s", "id": 154150, "pid": 76337, "tid": -914061504, "ts": 1716454223829644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829708, "dur": 0,
    "args": {
      "External id": 154160,
      "cbid": 317, "correlation": 154160
    }
  },
  {
    "ph": "f", "id": 154160, "pid": 76337, "tid": -914061504, "ts": 1716454223829708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829709, "dur": 0,
    "args": {
      "External id": 154161,
      "cbid": 203, "correlation": 154161
    }
  },
  {
    "ph": "f", "id": 154161, "pid": 76337, "tid": -914061504, "ts": 1716454223829709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829710, "dur": 0,
    "args": {
      "External id": 154162,
      "cbid": 205, "correlation": 154162
    }
  },
  {
    "ph": "f", "id": 154162, "pid": 76337, "tid": -914061504, "ts": 1716454223829710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223874512, "dur": 23,
    "args": {
      "External id": 154166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154166,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154166, "pid": 5, "tid": 7, "ts": 1716454223874512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829727, "dur": 12,
    "args": {
      "External id": 154166,
      "cbid": 211, "correlation": 154166
    }
  },
  {
    "ph": "s", "id": 154166, "pid": 76337, "tid": -914061504, "ts": 1716454223829727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223874536, "dur": 44,
    "args": {
      "External id": 154168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154168, "pid": 5, "tid": 7, "ts": 1716454223874536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829742, "dur": 5,
    "args": {
      "External id": 154168,
      "cbid": 211, "correlation": 154168
    }
  },
  {
    "ph": "s", "id": 154168, "pid": 76337, "tid": -914061504, "ts": 1716454223829742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223874581, "dur": 655,
    "args": {
      "External id": 154170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154170,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154170, "pid": 5, "tid": 7, "ts": 1716454223874581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829756, "dur": 9,
    "args": {
      "External id": 154170,
      "cbid": 211, "correlation": 154170
    }
  },
  {
    "ph": "s", "id": 154170, "pid": 76337, "tid": -914061504, "ts": 1716454223829756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223875237, "dur": 21,
    "args": {
      "External id": 154172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154172,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154172, "pid": 5, "tid": 7, "ts": 1716454223875237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829769, "dur": 5,
    "args": {
      "External id": 154172,
      "cbid": 211, "correlation": 154172
    }
  },
  {
    "ph": "s", "id": 154172, "pid": 76337, "tid": -914061504, "ts": 1716454223829769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223875259, "dur": 33,
    "args": {
      "External id": 154178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154178, "pid": 5, "tid": 7, "ts": 1716454223875259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829797, "dur": 9,
    "args": {
      "External id": 154178,
      "cbid": 211, "correlation": 154178
    }
  },
  {
    "ph": "s", "id": 154178, "pid": 76337, "tid": -914061504, "ts": 1716454223829797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829856, "dur": 0,
    "args": {
      "External id": 154188,
      "cbid": 317, "correlation": 154188
    }
  },
  {
    "ph": "f", "id": 154188, "pid": 76337, "tid": -914061504, "ts": 1716454223829856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829857, "dur": 0,
    "args": {
      "External id": 154189,
      "cbid": 203, "correlation": 154189
    }
  },
  {
    "ph": "f", "id": 154189, "pid": 76337, "tid": -914061504, "ts": 1716454223829857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829858, "dur": 0,
    "args": {
      "External id": 154190,
      "cbid": 205, "correlation": 154190
    }
  },
  {
    "ph": "f", "id": 154190, "pid": 76337, "tid": -914061504, "ts": 1716454223829858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829880, "dur": 1,
    "args": {
      "External id": 154194,
      "cbid": 251, "correlation": 154194
    }
  },
  {
    "ph": "f", "id": 154194, "pid": 76337, "tid": -914061504, "ts": 1716454223829880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829882, "dur": 0,
    "args": {
      "External id": 154195,
      "cbid": 251, "correlation": 154195
    }
  },
  {
    "ph": "f", "id": 154195, "pid": 76337, "tid": -914061504, "ts": 1716454223829882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829882, "dur": 0,
    "args": {
      "External id": 154196,
      "cbid": 251, "correlation": 154196
    }
  },
  {
    "ph": "f", "id": 154196, "pid": 76337, "tid": -914061504, "ts": 1716454223829882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829883, "dur": 0,
    "args": {
      "External id": 154197,
      "cbid": 251, "correlation": 154197
    }
  },
  {
    "ph": "f", "id": 154197, "pid": 76337, "tid": -914061504, "ts": 1716454223829883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829884, "dur": 0,
    "args": {
      "External id": 154198,
      "cbid": 251, "correlation": 154198
    }
  },
  {
    "ph": "f", "id": 154198, "pid": 76337, "tid": -914061504, "ts": 1716454223829884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829884, "dur": 0,
    "args": {
      "External id": 154199,
      "cbid": 251, "correlation": 154199
    }
  },
  {
    "ph": "f", "id": 154199, "pid": 76337, "tid": -914061504, "ts": 1716454223829884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829885, "dur": 0,
    "args": {
      "External id": 154200,
      "cbid": 251, "correlation": 154200
    }
  },
  {
    "ph": "f", "id": 154200, "pid": 76337, "tid": -914061504, "ts": 1716454223829885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829886, "dur": 0,
    "args": {
      "External id": 154201,
      "cbid": 251, "correlation": 154201
    }
  },
  {
    "ph": "f", "id": 154201, "pid": 76337, "tid": -914061504, "ts": 1716454223829886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829887, "dur": 0,
    "args": {
      "External id": 154202,
      "cbid": 251, "correlation": 154202
    }
  },
  {
    "ph": "f", "id": 154202, "pid": 76337, "tid": -914061504, "ts": 1716454223829887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223875293, "dur": 52,
    "args": {
      "External id": 154203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154203,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 154203, "pid": 5, "tid": 7, "ts": 1716454223875293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829889, "dur": 12,
    "args": {
      "External id": 154203,
      "cbid": 211, "correlation": 154203
    }
  },
  {
    "ph": "s", "id": 154203, "pid": 76337, "tid": -914061504, "ts": 1716454223829889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223875347, "dur": 32,
    "args": {
      "External id": 154209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154209, "pid": 5, "tid": 7, "ts": 1716454223875347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829922, "dur": 8,
    "args": {
      "External id": 154209,
      "cbid": 211, "correlation": 154209
    }
  },
  {
    "ph": "s", "id": 154209, "pid": 76337, "tid": -914061504, "ts": 1716454223829922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223875380, "dur": 27,
    "args": {
      "External id": 154217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154217,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154217, "pid": 5, "tid": 7, "ts": 1716454223875380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829951, "dur": 8,
    "args": {
      "External id": 154217,
      "cbid": 211, "correlation": 154217
    }
  },
  {
    "ph": "s", "id": 154217, "pid": 76337, "tid": -914061504, "ts": 1716454223829951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223875408, "dur": 20,
    "args": {
      "External id": 154225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154225,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154225, "pid": 5, "tid": 7, "ts": 1716454223875408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223829988, "dur": 9,
    "args": {
      "External id": 154225,
      "cbid": 211, "correlation": 154225
    }
  },
  {
    "ph": "s", "id": 154225, "pid": 76337, "tid": -914061504, "ts": 1716454223829988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223875429, "dur": 31,
    "args": {
      "External id": 154245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154245,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 154245, "pid": 5, "tid": 7, "ts": 1716454223875429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830071, "dur": 12,
    "args": {
      "External id": 154245,
      "cbid": 211, "correlation": 154245
    }
  },
  {
    "ph": "s", "id": 154245, "pid": 76337, "tid": -914061504, "ts": 1716454223830071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223875461, "dur": 4,
    "args": {
      "External id": 154257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154257,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 154257, "pid": 5, "tid": 7, "ts": 1716454223875461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830093, "dur": 6,
    "args": {
      "External id": 154257,
      "cbid": 211, "correlation": 154257
    }
  },
  {
    "ph": "s", "id": 154257, "pid": 76337, "tid": -914061504, "ts": 1716454223830093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223875466, "dur": 31,
    "args": {
      "External id": 154260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154260,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154260, "pid": 5, "tid": 7, "ts": 1716454223875466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830112, "dur": 6,
    "args": {
      "External id": 154260,
      "cbid": 211, "correlation": 154260
    }
  },
  {
    "ph": "s", "id": 154260, "pid": 76337, "tid": -914061504, "ts": 1716454223830112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830169, "dur": 0,
    "args": {
      "External id": 154271,
      "cbid": 317, "correlation": 154271
    }
  },
  {
    "ph": "f", "id": 154271, "pid": 76337, "tid": -914061504, "ts": 1716454223830169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830170, "dur": 0,
    "args": {
      "External id": 154272,
      "cbid": 203, "correlation": 154272
    }
  },
  {
    "ph": "f", "id": 154272, "pid": 76337, "tid": -914061504, "ts": 1716454223830170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830171, "dur": 0,
    "args": {
      "External id": 154273,
      "cbid": 205, "correlation": 154273
    }
  },
  {
    "ph": "f", "id": 154273, "pid": 76337, "tid": -914061504, "ts": 1716454223830171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223875499, "dur": 21,
    "args": {
      "External id": 154277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154277,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154277, "pid": 5, "tid": 7, "ts": 1716454223875499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830186, "dur": 11,
    "args": {
      "External id": 154277,
      "cbid": 211, "correlation": 154277
    }
  },
  {
    "ph": "s", "id": 154277, "pid": 76337, "tid": -914061504, "ts": 1716454223830186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223875521, "dur": 120,
    "args": {
      "External id": 154279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154279,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154279, "pid": 5, "tid": 7, "ts": 1716454223875521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830206, "dur": 8,
    "args": {
      "External id": 154279,
      "cbid": 211, "correlation": 154279
    }
  },
  {
    "ph": "s", "id": 154279, "pid": 76337, "tid": -914061504, "ts": 1716454223830206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223875642, "dur": 21,
    "args": {
      "External id": 154281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154281,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154281, "pid": 5, "tid": 7, "ts": 1716454223875642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830218, "dur": 5,
    "args": {
      "External id": 154281,
      "cbid": 211, "correlation": 154281
    }
  },
  {
    "ph": "s", "id": 154281, "pid": 76337, "tid": -914061504, "ts": 1716454223830218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223875664, "dur": 32,
    "args": {
      "External id": 154287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154287,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154287, "pid": 5, "tid": 7, "ts": 1716454223875664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830245, "dur": 9,
    "args": {
      "External id": 154287,
      "cbid": 211, "correlation": 154287
    }
  },
  {
    "ph": "s", "id": 154287, "pid": 76337, "tid": -914061504, "ts": 1716454223830245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223875698, "dur": 180,
    "args": {
      "External id": 154296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154296,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154296, "pid": 5, "tid": 7, "ts": 1716454223875698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830327, "dur": 13,
    "args": {
      "External id": 154296,
      "cbid": 211, "correlation": 154296
    }
  },
  {
    "ph": "s", "id": 154296, "pid": 76337, "tid": -914061504, "ts": 1716454223830327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223875879, "dur": 65,
    "args": {
      "External id": 154318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154318,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154318, "pid": 5, "tid": 7, "ts": 1716454223875879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830384, "dur": 10,
    "args": {
      "External id": 154318,
      "cbid": 211, "correlation": 154318
    }
  },
  {
    "ph": "s", "id": 154318, "pid": 76337, "tid": -914061504, "ts": 1716454223830384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830476, "dur": 1,
    "args": {
      "External id": 154329,
      "cbid": 251, "correlation": 154329
    }
  },
  {
    "ph": "f", "id": 154329, "pid": 76337, "tid": -914061504, "ts": 1716454223830476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223875946, "dur": 156,
    "args": {
      "External id": 154330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154330,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154330, "pid": 5, "tid": 7, "ts": 1716454223875946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830481, "dur": 13,
    "args": {
      "External id": 154330,
      "cbid": 211, "correlation": 154330
    }
  },
  {
    "ph": "s", "id": 154330, "pid": 76337, "tid": -914061504, "ts": 1716454223830481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830551, "dur": 1,
    "args": {
      "External id": 154341,
      "cbid": 251, "correlation": 154341
    }
  },
  {
    "ph": "f", "id": 154341, "pid": 76337, "tid": -914061504, "ts": 1716454223830551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223876103, "dur": 146,
    "args": {
      "External id": 154342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154342,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154342, "pid": 5, "tid": 7, "ts": 1716454223876103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830555, "dur": 11,
    "args": {
      "External id": 154342,
      "cbid": 211, "correlation": 154342
    }
  },
  {
    "ph": "s", "id": 154342, "pid": 76337, "tid": -914061504, "ts": 1716454223830555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830620, "dur": 1,
    "args": {
      "External id": 154353,
      "cbid": 251, "correlation": 154353
    }
  },
  {
    "ph": "f", "id": 154353, "pid": 76337, "tid": -914061504, "ts": 1716454223830620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223876250, "dur": 144,
    "args": {
      "External id": 154354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154354,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154354, "pid": 5, "tid": 7, "ts": 1716454223876250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830624, "dur": 11,
    "args": {
      "External id": 154354,
      "cbid": 211, "correlation": 154354
    }
  },
  {
    "ph": "s", "id": 154354, "pid": 76337, "tid": -914061504, "ts": 1716454223830624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223876395, "dur": 1945,
    "args": {
      "External id": 154375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154375,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 154375, "pid": 5, "tid": 7, "ts": 1716454223876395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830707, "dur": 13,
    "args": {
      "External id": 154375,
      "cbid": 211, "correlation": 154375
    }
  },
  {
    "ph": "s", "id": 154375, "pid": 76337, "tid": -914061504, "ts": 1716454223830707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830809, "dur": 1,
    "args": {
      "External id": 154393,
      "cbid": 251, "correlation": 154393
    }
  },
  {
    "ph": "f", "id": 154393, "pid": 76337, "tid": -914061504, "ts": 1716454223830809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223878342, "dur": 146,
    "args": {
      "External id": 154395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154395,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 154395, "pid": 5, "tid": 7, "ts": 1716454223878342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830815, "dur": 13,
    "args": {
      "External id": 154395,
      "cbid": 211, "correlation": 154395
    }
  },
  {
    "ph": "s", "id": 154395, "pid": 76337, "tid": -914061504, "ts": 1716454223830815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223878489, "dur": 35,
    "args": {
      "External id": 154403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154403,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154403, "pid": 5, "tid": 7, "ts": 1716454223878489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830886, "dur": 12,
    "args": {
      "External id": 154403,
      "cbid": 211, "correlation": 154403
    }
  },
  {
    "ph": "s", "id": 154403, "pid": 76337, "tid": -914061504, "ts": 1716454223830886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223878526, "dur": 50,
    "args": {
      "External id": 154411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154411,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154411, "pid": 5, "tid": 7, "ts": 1716454223878526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223830926, "dur": 9,
    "args": {
      "External id": 154411,
      "cbid": 211, "correlation": 154411
    }
  },
  {
    "ph": "s", "id": 154411, "pid": 76337, "tid": -914061504, "ts": 1716454223830926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223878577, "dur": 30,
    "args": {
      "External id": 154422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154422, "pid": 5, "tid": 7, "ts": 1716454223878577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831008, "dur": 13,
    "args": {
      "External id": 154422,
      "cbid": 211, "correlation": 154422
    }
  },
  {
    "ph": "s", "id": 154422, "pid": 76337, "tid": -914061504, "ts": 1716454223831008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223878608, "dur": 34,
    "args": {
      "External id": 154444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154444,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154444, "pid": 5, "tid": 7, "ts": 1716454223878608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831039, "dur": 8,
    "args": {
      "External id": 154444,
      "cbid": 211, "correlation": 154444
    }
  },
  {
    "ph": "s", "id": 154444, "pid": 76337, "tid": -914061504, "ts": 1716454223831039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831124, "dur": 1,
    "args": {
      "External id": 154455,
      "cbid": 251, "correlation": 154455
    }
  },
  {
    "ph": "f", "id": 154455, "pid": 76337, "tid": -914061504, "ts": 1716454223831124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223878644, "dur": 90,
    "args": {
      "External id": 154456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154456,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154456, "pid": 5, "tid": 7, "ts": 1716454223878644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831129, "dur": 13,
    "args": {
      "External id": 154456,
      "cbid": 211, "correlation": 154456
    }
  },
  {
    "ph": "s", "id": 154456, "pid": 76337, "tid": -914061504, "ts": 1716454223831129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831199, "dur": 1,
    "args": {
      "External id": 154467,
      "cbid": 251, "correlation": 154467
    }
  },
  {
    "ph": "f", "id": 154467, "pid": 76337, "tid": -914061504, "ts": 1716454223831199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831203, "dur": 0,
    "args": {
      "External id": 154468,
      "cbid": 251, "correlation": 154468
    }
  },
  {
    "ph": "f", "id": 154468, "pid": 76337, "tid": -914061504, "ts": 1716454223831203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223878736, "dur": 11,
    "args": {
      "External id": 154469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154469,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 154469, "pid": 5, "tid": 7, "ts": 1716454223878736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831204, "dur": 12,
    "args": {
      "External id": 154469,
      "cbid": 211, "correlation": 154469
    }
  },
  {
    "ph": "s", "id": 154469, "pid": 76337, "tid": -914061504, "ts": 1716454223831204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223878748, "dur": 5,
    "args": {
      "External id": 154471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154471,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 154471, "pid": 5, "tid": 7, "ts": 1716454223878748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831219, "dur": 8,
    "args": {
      "External id": 154471,
      "cbid": 211, "correlation": 154471
    }
  },
  {
    "ph": "s", "id": 154471, "pid": 76337, "tid": -914061504, "ts": 1716454223831219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831278, "dur": 1,
    "args": {
      "External id": 154482,
      "cbid": 251, "correlation": 154482
    }
  },
  {
    "ph": "f", "id": 154482, "pid": 76337, "tid": -914061504, "ts": 1716454223831278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831282, "dur": 0,
    "args": {
      "External id": 154483,
      "cbid": 251, "correlation": 154483
    }
  },
  {
    "ph": "f", "id": 154483, "pid": 76337, "tid": -914061504, "ts": 1716454223831282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223878754, "dur": 7,
    "args": {
      "External id": 154484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154484,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 154484, "pid": 5, "tid": 7, "ts": 1716454223878754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831283, "dur": 12,
    "args": {
      "External id": 154484,
      "cbid": 211, "correlation": 154484
    }
  },
  {
    "ph": "s", "id": 154484, "pid": 76337, "tid": -914061504, "ts": 1716454223831283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223878763, "dur": 3,
    "args": {
      "External id": 154486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154486,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 154486, "pid": 5, "tid": 7, "ts": 1716454223878763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831297, "dur": 6,
    "args": {
      "External id": 154486,
      "cbid": 211, "correlation": 154486
    }
  },
  {
    "ph": "s", "id": 154486, "pid": 76337, "tid": -914061504, "ts": 1716454223831297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223878767, "dur": 92,
    "args": {
      "External id": 154507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154507,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 154507, "pid": 5, "tid": 7, "ts": 1716454223878767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831370, "dur": 13,
    "args": {
      "External id": 154507,
      "cbid": 211, "correlation": 154507
    }
  },
  {
    "ph": "s", "id": 154507, "pid": 76337, "tid": -914061504, "ts": 1716454223831370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831469, "dur": 1,
    "args": {
      "External id": 154525,
      "cbid": 251, "correlation": 154525
    }
  },
  {
    "ph": "f", "id": 154525, "pid": 76337, "tid": -914061504, "ts": 1716454223831469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223878861, "dur": 100,
    "args": {
      "External id": 154527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154527,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154527, "pid": 5, "tid": 7, "ts": 1716454223878861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831475, "dur": 13,
    "args": {
      "External id": 154527,
      "cbid": 211, "correlation": 154527
    }
  },
  {
    "ph": "s", "id": 154527, "pid": 76337, "tid": -914061504, "ts": 1716454223831475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223878962, "dur": 19,
    "args": {
      "External id": 154535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154535,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154535, "pid": 5, "tid": 7, "ts": 1716454223878962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831542, "dur": 12,
    "args": {
      "External id": 154535,
      "cbid": 211, "correlation": 154535
    }
  },
  {
    "ph": "s", "id": 154535, "pid": 76337, "tid": -914061504, "ts": 1716454223831542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223878982, "dur": 37,
    "args": {
      "External id": 154543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154543,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154543, "pid": 5, "tid": 7, "ts": 1716454223878982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831584, "dur": 9,
    "args": {
      "External id": 154543,
      "cbid": 211, "correlation": 154543
    }
  },
  {
    "ph": "s", "id": 154543, "pid": 76337, "tid": -914061504, "ts": 1716454223831584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223879020, "dur": 34,
    "args": {
      "External id": 154565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154565,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154565, "pid": 5, "tid": 7, "ts": 1716454223879020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831635, "dur": 11,
    "args": {
      "External id": 154565,
      "cbid": 211, "correlation": 154565
    }
  },
  {
    "ph": "s", "id": 154565, "pid": 76337, "tid": -914061504, "ts": 1716454223831635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831725, "dur": 1,
    "args": {
      "External id": 154581,
      "cbid": 251, "correlation": 154581
    }
  },
  {
    "ph": "f", "id": 154581, "pid": 76337, "tid": -914061504, "ts": 1716454223831725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831730, "dur": 0,
    "args": {
      "External id": 154583,
      "cbid": 251, "correlation": 154583
    }
  },
  {
    "ph": "f", "id": 154583, "pid": 76337, "tid": -914061504, "ts": 1716454223831730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223879055, "dur": 540,
    "args": {
      "External id": 154584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154584,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 154584, "pid": 5, "tid": 7, "ts": 1716454223879055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831733, "dur": 13,
    "args": {
      "External id": 154584,
      "cbid": 211, "correlation": 154584
    }
  },
  {
    "ph": "s", "id": 154584, "pid": 76337, "tid": -914061504, "ts": 1716454223831733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223879596, "dur": 125,
    "args": {
      "External id": 154592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154592, "pid": 5, "tid": 7, "ts": 1716454223879596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831799, "dur": 12,
    "args": {
      "External id": 154592,
      "cbid": 211, "correlation": 154592
    }
  },
  {
    "ph": "s", "id": 154592, "pid": 76337, "tid": -914061504, "ts": 1716454223831799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223879723, "dur": 127,
    "args": {
      "External id": 154600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154600, "pid": 5, "tid": 7, "ts": 1716454223879723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831829, "dur": 8,
    "args": {
      "External id": 154600,
      "cbid": 211, "correlation": 154600
    }
  },
  {
    "ph": "s", "id": 154600, "pid": 76337, "tid": -914061504, "ts": 1716454223831829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831907, "dur": 1,
    "args": {
      "External id": 154616,
      "cbid": 251, "correlation": 154616
    }
  },
  {
    "ph": "f", "id": 154616, "pid": 76337, "tid": -914061504, "ts": 1716454223831907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223879852, "dur": 305,
    "args": {
      "External id": 154618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154618,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154618, "pid": 5, "tid": 7, "ts": 1716454223879852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831913, "dur": 12,
    "args": {
      "External id": 154618,
      "cbid": 211, "correlation": 154618
    }
  },
  {
    "ph": "s", "id": 154618, "pid": 76337, "tid": -914061504, "ts": 1716454223831913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223880158, "dur": 27,
    "args": {
      "External id": 154626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154626,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154626, "pid": 5, "tid": 7, "ts": 1716454223880158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223831955, "dur": 10,
    "args": {
      "External id": 154626,
      "cbid": 211, "correlation": 154626
    }
  },
  {
    "ph": "s", "id": 154626, "pid": 76337, "tid": -914061504, "ts": 1716454223831955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223880186, "dur": 82,
    "args": {
      "External id": 154637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154637, "pid": 5, "tid": 7, "ts": 1716454223880186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832034, "dur": 13,
    "args": {
      "External id": 154637,
      "cbid": 211, "correlation": 154637
    }
  },
  {
    "ph": "s", "id": 154637, "pid": 76337, "tid": -914061504, "ts": 1716454223832034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832099, "dur": 0,
    "args": {
      "External id": 154649,
      "cbid": 317, "correlation": 154649
    }
  },
  {
    "ph": "f", "id": 154649, "pid": 76337, "tid": -914061504, "ts": 1716454223832099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832100, "dur": 0,
    "args": {
      "External id": 154650,
      "cbid": 203, "correlation": 154650
    }
  },
  {
    "ph": "f", "id": 154650, "pid": 76337, "tid": -914061504, "ts": 1716454223832100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832101, "dur": 0,
    "args": {
      "External id": 154651,
      "cbid": 205, "correlation": 154651
    }
  },
  {
    "ph": "f", "id": 154651, "pid": 76337, "tid": -914061504, "ts": 1716454223832101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223880269, "dur": 23,
    "args": {
      "External id": 154655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154655,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154655, "pid": 5, "tid": 7, "ts": 1716454223880269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832116, "dur": 12,
    "args": {
      "External id": 154655,
      "cbid": 211, "correlation": 154655
    }
  },
  {
    "ph": "s", "id": 154655, "pid": 76337, "tid": -914061504, "ts": 1716454223832116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223880294, "dur": 120,
    "args": {
      "External id": 154657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154657,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154657, "pid": 5, "tid": 7, "ts": 1716454223880294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832135, "dur": 7,
    "args": {
      "External id": 154657,
      "cbid": 211, "correlation": 154657
    }
  },
  {
    "ph": "s", "id": 154657, "pid": 76337, "tid": -914061504, "ts": 1716454223832135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223880415, "dur": 22,
    "args": {
      "External id": 154659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154659,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154659, "pid": 5, "tid": 7, "ts": 1716454223880415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832146, "dur": 5,
    "args": {
      "External id": 154659,
      "cbid": 211, "correlation": 154659
    }
  },
  {
    "ph": "s", "id": 154659, "pid": 76337, "tid": -914061504, "ts": 1716454223832146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223880438, "dur": 33,
    "args": {
      "External id": 154665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154665, "pid": 5, "tid": 7, "ts": 1716454223880438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832173, "dur": 8,
    "args": {
      "External id": 154665,
      "cbid": 211, "correlation": 154665
    }
  },
  {
    "ph": "s", "id": 154665, "pid": 76337, "tid": -914061504, "ts": 1716454223832173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223880473, "dur": 27,
    "args": {
      "External id": 154673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154673,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154673, "pid": 5, "tid": 7, "ts": 1716454223880473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832205, "dur": 8,
    "args": {
      "External id": 154673,
      "cbid": 211, "correlation": 154673
    }
  },
  {
    "ph": "s", "id": 154673, "pid": 76337, "tid": -914061504, "ts": 1716454223832205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223880501, "dur": 31,
    "args": {
      "External id": 154693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154693,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 154693, "pid": 5, "tid": 7, "ts": 1716454223880501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832277, "dur": 12,
    "args": {
      "External id": 154693,
      "cbid": 211, "correlation": 154693
    }
  },
  {
    "ph": "s", "id": 154693, "pid": 76337, "tid": -914061504, "ts": 1716454223832277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223880533, "dur": 4,
    "args": {
      "External id": 154705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154705,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 154705, "pid": 5, "tid": 7, "ts": 1716454223880533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832300, "dur": 6,
    "args": {
      "External id": 154705,
      "cbid": 211, "correlation": 154705
    }
  },
  {
    "ph": "s", "id": 154705, "pid": 76337, "tid": -914061504, "ts": 1716454223832300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223880539, "dur": 31,
    "args": {
      "External id": 154708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154708, "pid": 5, "tid": 7, "ts": 1716454223880539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832318, "dur": 7,
    "args": {
      "External id": 154708,
      "cbid": 211, "correlation": 154708
    }
  },
  {
    "ph": "s", "id": 154708, "pid": 76337, "tid": -914061504, "ts": 1716454223832318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223880571, "dur": 21,
    "args": {
      "External id": 154717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154717,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154717, "pid": 5, "tid": 7, "ts": 1716454223880571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832357, "dur": 9,
    "args": {
      "External id": 154717,
      "cbid": 211, "correlation": 154717
    }
  },
  {
    "ph": "s", "id": 154717, "pid": 76337, "tid": -914061504, "ts": 1716454223832357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832408, "dur": 0,
    "args": {
      "External id": 154727,
      "cbid": 317, "correlation": 154727
    }
  },
  {
    "ph": "f", "id": 154727, "pid": 76337, "tid": -914061504, "ts": 1716454223832408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832408, "dur": 0,
    "args": {
      "External id": 154728,
      "cbid": 203, "correlation": 154728
    }
  },
  {
    "ph": "f", "id": 154728, "pid": 76337, "tid": -914061504, "ts": 1716454223832408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832409, "dur": 0,
    "args": {
      "External id": 154729,
      "cbid": 205, "correlation": 154729
    }
  },
  {
    "ph": "f", "id": 154729, "pid": 76337, "tid": -914061504, "ts": 1716454223832409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223880594, "dur": 22,
    "args": {
      "External id": 154733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154733,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154733, "pid": 5, "tid": 7, "ts": 1716454223880594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832423, "dur": 11,
    "args": {
      "External id": 154733,
      "cbid": 211, "correlation": 154733
    }
  },
  {
    "ph": "s", "id": 154733, "pid": 76337, "tid": -914061504, "ts": 1716454223832423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223880618, "dur": 44,
    "args": {
      "External id": 154735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154735,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154735, "pid": 5, "tid": 7, "ts": 1716454223880618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832437, "dur": 5,
    "args": {
      "External id": 154735,
      "cbid": 211, "correlation": 154735
    }
  },
  {
    "ph": "s", "id": 154735, "pid": 76337, "tid": -914061504, "ts": 1716454223832437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223880663, "dur": 654,
    "args": {
      "External id": 154737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154737,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154737, "pid": 5, "tid": 7, "ts": 1716454223880663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832448, "dur": 7,
    "args": {
      "External id": 154737,
      "cbid": 211, "correlation": 154737
    }
  },
  {
    "ph": "s", "id": 154737, "pid": 76337, "tid": -914061504, "ts": 1716454223832448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223881318, "dur": 22,
    "args": {
      "External id": 154739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154739,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154739, "pid": 5, "tid": 7, "ts": 1716454223881318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832458, "dur": 5,
    "args": {
      "External id": 154739,
      "cbid": 211, "correlation": 154739
    }
  },
  {
    "ph": "s", "id": 154739, "pid": 76337, "tid": -914061504, "ts": 1716454223832458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223881341, "dur": 33,
    "args": {
      "External id": 154745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154745, "pid": 5, "tid": 7, "ts": 1716454223881341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832485, "dur": 8,
    "args": {
      "External id": 154745,
      "cbid": 211, "correlation": 154745
    }
  },
  {
    "ph": "s", "id": 154745, "pid": 76337, "tid": -914061504, "ts": 1716454223832485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223881375, "dur": 3,
    "args": {
      "External id": 154753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154753,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 154753, "pid": 5, "tid": 7, "ts": 1716454223881375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832528, "dur": 9,
    "args": {
      "External id": 154753,
      "cbid": 211, "correlation": 154753
    }
  },
  {
    "ph": "s", "id": 154753, "pid": 76337, "tid": -914061504, "ts": 1716454223832528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832594, "dur": 2,
    "args": {
      "External id": 154769,
      "cbid": 251, "correlation": 154769
    }
  },
  {
    "ph": "f", "id": 154769, "pid": 76337, "tid": -914061504, "ts": 1716454223832594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832600, "dur": 0,
    "args": {
      "External id": 154771,
      "cbid": 251, "correlation": 154771
    }
  },
  {
    "ph": "f", "id": 154771, "pid": 76337, "tid": -914061504, "ts": 1716454223832600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223881380, "dur": 12,
    "args": {
      "External id": 154772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154772,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 154772, "pid": 5, "tid": 7, "ts": 1716454223881380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832602, "dur": 11,
    "args": {
      "External id": 154772,
      "cbid": 211, "correlation": 154772
    }
  },
  {
    "ph": "s", "id": 154772, "pid": 76337, "tid": -914061504, "ts": 1716454223832602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223881393, "dur": 5,
    "args": {
      "External id": 154774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154774,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 154774, "pid": 5, "tid": 7, "ts": 1716454223881393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832615, "dur": 5,
    "args": {
      "External id": 154774,
      "cbid": 211, "correlation": 154774
    }
  },
  {
    "ph": "s", "id": 154774, "pid": 76337, "tid": -914061504, "ts": 1716454223832615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223881399, "dur": 30,
    "args": {
      "External id": 154784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154784, "pid": 5, "tid": 7, "ts": 1716454223881399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832672, "dur": 12,
    "args": {
      "External id": 154784,
      "cbid": 211, "correlation": 154784
    }
  },
  {
    "ph": "s", "id": 154784, "pid": 76337, "tid": -914061504, "ts": 1716454223832672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223881430, "dur": 30,
    "args": {
      "External id": 154804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154804,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 154804, "pid": 5, "tid": 7, "ts": 1716454223881430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832738, "dur": 11,
    "args": {
      "External id": 154804,
      "cbid": 211, "correlation": 154804
    }
  },
  {
    "ph": "s", "id": 154804, "pid": 76337, "tid": -914061504, "ts": 1716454223832738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223881461, "dur": 4,
    "args": {
      "External id": 154816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154816,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 154816, "pid": 5, "tid": 7, "ts": 1716454223881461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832758, "dur": 6,
    "args": {
      "External id": 154816,
      "cbid": 211, "correlation": 154816
    }
  },
  {
    "ph": "s", "id": 154816, "pid": 76337, "tid": -914061504, "ts": 1716454223832758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223881466, "dur": 30,
    "args": {
      "External id": 154819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154819, "pid": 5, "tid": 7, "ts": 1716454223881466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832776, "dur": 6,
    "args": {
      "External id": 154819,
      "cbid": 211, "correlation": 154819
    }
  },
  {
    "ph": "s", "id": 154819, "pid": 76337, "tid": -914061504, "ts": 1716454223832776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223881498, "dur": 20,
    "args": {
      "External id": 154828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154828,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154828, "pid": 5, "tid": 7, "ts": 1716454223881498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832817, "dur": 9,
    "args": {
      "External id": 154828,
      "cbid": 211, "correlation": 154828
    }
  },
  {
    "ph": "s", "id": 154828, "pid": 76337, "tid": -914061504, "ts": 1716454223832817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832879, "dur": 0,
    "args": {
      "External id": 154838,
      "cbid": 317, "correlation": 154838
    }
  },
  {
    "ph": "f", "id": 154838, "pid": 76337, "tid": -914061504, "ts": 1716454223832879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832880, "dur": 0,
    "args": {
      "External id": 154839,
      "cbid": 203, "correlation": 154839
    }
  },
  {
    "ph": "f", "id": 154839, "pid": 76337, "tid": -914061504, "ts": 1716454223832880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832881, "dur": 0,
    "args": {
      "External id": 154840,
      "cbid": 205, "correlation": 154840
    }
  },
  {
    "ph": "f", "id": 154840, "pid": 76337, "tid": -914061504, "ts": 1716454223832881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223881520, "dur": 22,
    "args": {
      "External id": 154844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154844,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154844, "pid": 5, "tid": 7, "ts": 1716454223881520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832894, "dur": 13,
    "args": {
      "External id": 154844,
      "cbid": 211, "correlation": 154844
    }
  },
  {
    "ph": "s", "id": 154844, "pid": 76337, "tid": -914061504, "ts": 1716454223832894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223881543, "dur": 44,
    "args": {
      "External id": 154846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154846, "pid": 5, "tid": 7, "ts": 1716454223881543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832909, "dur": 5,
    "args": {
      "External id": 154846,
      "cbid": 211, "correlation": 154846
    }
  },
  {
    "ph": "s", "id": 154846, "pid": 76337, "tid": -914061504, "ts": 1716454223832909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223881588, "dur": 645,
    "args": {
      "External id": 154848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154848,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154848, "pid": 5, "tid": 7, "ts": 1716454223881588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832920, "dur": 6,
    "args": {
      "External id": 154848,
      "cbid": 211, "correlation": 154848
    }
  },
  {
    "ph": "s", "id": 154848, "pid": 76337, "tid": -914061504, "ts": 1716454223832920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223882234, "dur": 22,
    "args": {
      "External id": 154850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154850,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154850, "pid": 5, "tid": 7, "ts": 1716454223882234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832929, "dur": 5,
    "args": {
      "External id": 154850,
      "cbid": 211, "correlation": 154850
    }
  },
  {
    "ph": "s", "id": 154850, "pid": 76337, "tid": -914061504, "ts": 1716454223832929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223882257, "dur": 35,
    "args": {
      "External id": 154856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154856, "pid": 5, "tid": 7, "ts": 1716454223882257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832957, "dur": 8,
    "args": {
      "External id": 154856,
      "cbid": 211, "correlation": 154856
    }
  },
  {
    "ph": "s", "id": 154856, "pid": 76337, "tid": -914061504, "ts": 1716454223832957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223882293, "dur": 26,
    "args": {
      "External id": 154864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154864,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154864, "pid": 5, "tid": 7, "ts": 1716454223882293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223832999, "dur": 9,
    "args": {
      "External id": 154864,
      "cbid": 211, "correlation": 154864
    }
  },
  {
    "ph": "s", "id": 154864, "pid": 76337, "tid": -914061504, "ts": 1716454223832999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223882321, "dur": 20,
    "args": {
      "External id": 154872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154872,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154872, "pid": 5, "tid": 7, "ts": 1716454223882321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833029, "dur": 9,
    "args": {
      "External id": 154872,
      "cbid": 211, "correlation": 154872
    }
  },
  {
    "ph": "s", "id": 154872, "pid": 76337, "tid": -914061504, "ts": 1716454223833029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223882342, "dur": 30,
    "args": {
      "External id": 154892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154892,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 154892, "pid": 5, "tid": 7, "ts": 1716454223882342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833109, "dur": 12,
    "args": {
      "External id": 154892,
      "cbid": 211, "correlation": 154892
    }
  },
  {
    "ph": "s", "id": 154892, "pid": 76337, "tid": -914061504, "ts": 1716454223833109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223882374, "dur": 4,
    "args": {
      "External id": 154904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154904,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 154904, "pid": 5, "tid": 7, "ts": 1716454223882374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833132, "dur": 6,
    "args": {
      "External id": 154904,
      "cbid": 211, "correlation": 154904
    }
  },
  {
    "ph": "s", "id": 154904, "pid": 76337, "tid": -914061504, "ts": 1716454223833132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223882379, "dur": 29,
    "args": {
      "External id": 154907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154907,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154907, "pid": 5, "tid": 7, "ts": 1716454223882379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833149, "dur": 7,
    "args": {
      "External id": 154907,
      "cbid": 211, "correlation": 154907
    }
  },
  {
    "ph": "s", "id": 154907, "pid": 76337, "tid": -914061504, "ts": 1716454223833149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833207, "dur": 0,
    "args": {
      "External id": 154918,
      "cbid": 317, "correlation": 154918
    }
  },
  {
    "ph": "f", "id": 154918, "pid": 76337, "tid": -914061504, "ts": 1716454223833207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833208, "dur": 0,
    "args": {
      "External id": 154919,
      "cbid": 203, "correlation": 154919
    }
  },
  {
    "ph": "f", "id": 154919, "pid": 76337, "tid": -914061504, "ts": 1716454223833208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833209, "dur": 0,
    "args": {
      "External id": 154920,
      "cbid": 205, "correlation": 154920
    }
  },
  {
    "ph": "f", "id": 154920, "pid": 76337, "tid": -914061504, "ts": 1716454223833209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223882410, "dur": 22,
    "args": {
      "External id": 154924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154924,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154924, "pid": 5, "tid": 7, "ts": 1716454223882410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833222, "dur": 12,
    "args": {
      "External id": 154924,
      "cbid": 211, "correlation": 154924
    }
  },
  {
    "ph": "s", "id": 154924, "pid": 76337, "tid": -914061504, "ts": 1716454223833222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223882433, "dur": 118,
    "args": {
      "External id": 154926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154926,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154926, "pid": 5, "tid": 7, "ts": 1716454223882433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833239, "dur": 7,
    "args": {
      "External id": 154926,
      "cbid": 211, "correlation": 154926
    }
  },
  {
    "ph": "s", "id": 154926, "pid": 76337, "tid": -914061504, "ts": 1716454223833239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223882552, "dur": 22,
    "args": {
      "External id": 154928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154928,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154928, "pid": 5, "tid": 7, "ts": 1716454223882552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833250, "dur": 5,
    "args": {
      "External id": 154928,
      "cbid": 211, "correlation": 154928
    }
  },
  {
    "ph": "s", "id": 154928, "pid": 76337, "tid": -914061504, "ts": 1716454223833250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223882575, "dur": 32,
    "args": {
      "External id": 154934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154934, "pid": 5, "tid": 7, "ts": 1716454223882575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833277, "dur": 8,
    "args": {
      "External id": 154934,
      "cbid": 211, "correlation": 154934
    }
  },
  {
    "ph": "s", "id": 154934, "pid": 76337, "tid": -914061504, "ts": 1716454223833277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223882609, "dur": 200,
    "args": {
      "External id": 154943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154943,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154943, "pid": 5, "tid": 7, "ts": 1716454223882609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833358, "dur": 14,
    "args": {
      "External id": 154943,
      "cbid": 211, "correlation": 154943
    }
  },
  {
    "ph": "s", "id": 154943, "pid": 76337, "tid": -914061504, "ts": 1716454223833358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223882810, "dur": 65,
    "args": {
      "External id": 154965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154965,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 154965, "pid": 5, "tid": 7, "ts": 1716454223882810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833413, "dur": 10,
    "args": {
      "External id": 154965,
      "cbid": 211, "correlation": 154965
    }
  },
  {
    "ph": "s", "id": 154965, "pid": 76337, "tid": -914061504, "ts": 1716454223833413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833499, "dur": 1,
    "args": {
      "External id": 154976,
      "cbid": 251, "correlation": 154976
    }
  },
  {
    "ph": "f", "id": 154976, "pid": 76337, "tid": -914061504, "ts": 1716454223833499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223882876, "dur": 154,
    "args": {
      "External id": 154977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154977,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154977, "pid": 5, "tid": 7, "ts": 1716454223882876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833504, "dur": 13,
    "args": {
      "External id": 154977,
      "cbid": 211, "correlation": 154977
    }
  },
  {
    "ph": "s", "id": 154977, "pid": 76337, "tid": -914061504, "ts": 1716454223833504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833574, "dur": 1,
    "args": {
      "External id": 154988,
      "cbid": 251, "correlation": 154988
    }
  },
  {
    "ph": "f", "id": 154988, "pid": 76337, "tid": -914061504, "ts": 1716454223833574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223883032, "dur": 146,
    "args": {
      "External id": 154989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 154989,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 154989, "pid": 5, "tid": 7, "ts": 1716454223883032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833578, "dur": 12,
    "args": {
      "External id": 154989,
      "cbid": 211, "correlation": 154989
    }
  },
  {
    "ph": "s", "id": 154989, "pid": 76337, "tid": -914061504, "ts": 1716454223833578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833645, "dur": 1,
    "args": {
      "External id": 155000,
      "cbid": 251, "correlation": 155000
    }
  },
  {
    "ph": "f", "id": 155000, "pid": 76337, "tid": -914061504, "ts": 1716454223833645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223883179, "dur": 149,
    "args": {
      "External id": 155001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155001,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155001, "pid": 5, "tid": 7, "ts": 1716454223883179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833649, "dur": 11,
    "args": {
      "External id": 155001,
      "cbid": 211, "correlation": 155001
    }
  },
  {
    "ph": "s", "id": 155001, "pid": 76337, "tid": -914061504, "ts": 1716454223833649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223883329, "dur": 1945,
    "args": {
      "External id": 155022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155022,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 155022, "pid": 5, "tid": 7, "ts": 1716454223883329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833729, "dur": 14,
    "args": {
      "External id": 155022,
      "cbid": 211, "correlation": 155022
    }
  },
  {
    "ph": "s", "id": 155022, "pid": 76337, "tid": -914061504, "ts": 1716454223833729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833829, "dur": 1,
    "args": {
      "External id": 155040,
      "cbid": 251, "correlation": 155040
    }
  },
  {
    "ph": "f", "id": 155040, "pid": 76337, "tid": -914061504, "ts": 1716454223833829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223885276, "dur": 147,
    "args": {
      "External id": 155042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155042,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155042, "pid": 5, "tid": 7, "ts": 1716454223885276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833835, "dur": 14,
    "args": {
      "External id": 155042,
      "cbid": 211, "correlation": 155042
    }
  },
  {
    "ph": "s", "id": 155042, "pid": 76337, "tid": -914061504, "ts": 1716454223833835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223885425, "dur": 36,
    "args": {
      "External id": 155050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155050,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155050, "pid": 5, "tid": 7, "ts": 1716454223885425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833905, "dur": 12,
    "args": {
      "External id": 155050,
      "cbid": 211, "correlation": 155050
    }
  },
  {
    "ph": "s", "id": 155050, "pid": 76337, "tid": -914061504, "ts": 1716454223833905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223885461, "dur": 51,
    "args": {
      "External id": 155058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155058,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155058, "pid": 5, "tid": 7, "ts": 1716454223885461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223833945, "dur": 8,
    "args": {
      "External id": 155058,
      "cbid": 211, "correlation": 155058
    }
  },
  {
    "ph": "s", "id": 155058, "pid": 76337, "tid": -914061504, "ts": 1716454223833945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223885514, "dur": 31,
    "args": {
      "External id": 155069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155069, "pid": 5, "tid": 7, "ts": 1716454223885514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834027, "dur": 14,
    "args": {
      "External id": 155069,
      "cbid": 211, "correlation": 155069
    }
  },
  {
    "ph": "s", "id": 155069, "pid": 76337, "tid": -914061504, "ts": 1716454223834027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223885545, "dur": 35,
    "args": {
      "External id": 155091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155091,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155091, "pid": 5, "tid": 7, "ts": 1716454223885545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834060, "dur": 7,
    "args": {
      "External id": 155091,
      "cbid": 211, "correlation": 155091
    }
  },
  {
    "ph": "s", "id": 155091, "pid": 76337, "tid": -914061504, "ts": 1716454223834060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834144, "dur": 1,
    "args": {
      "External id": 155102,
      "cbid": 251, "correlation": 155102
    }
  },
  {
    "ph": "f", "id": 155102, "pid": 76337, "tid": -914061504, "ts": 1716454223834144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223885582, "dur": 87,
    "args": {
      "External id": 155103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155103,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155103, "pid": 5, "tid": 7, "ts": 1716454223885582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834149, "dur": 13,
    "args": {
      "External id": 155103,
      "cbid": 211, "correlation": 155103
    }
  },
  {
    "ph": "s", "id": 155103, "pid": 76337, "tid": -914061504, "ts": 1716454223834149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834218, "dur": 1,
    "args": {
      "External id": 155114,
      "cbid": 251, "correlation": 155114
    }
  },
  {
    "ph": "f", "id": 155114, "pid": 76337, "tid": -914061504, "ts": 1716454223834218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834221, "dur": 0,
    "args": {
      "External id": 155115,
      "cbid": 251, "correlation": 155115
    }
  },
  {
    "ph": "f", "id": 155115, "pid": 76337, "tid": -914061504, "ts": 1716454223834221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223885670, "dur": 11,
    "args": {
      "External id": 155116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155116,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 155116, "pid": 5, "tid": 7, "ts": 1716454223885670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834223, "dur": 11,
    "args": {
      "External id": 155116,
      "cbid": 211, "correlation": 155116
    }
  },
  {
    "ph": "s", "id": 155116, "pid": 76337, "tid": -914061504, "ts": 1716454223834223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223885683, "dur": 5,
    "args": {
      "External id": 155118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155118,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 155118, "pid": 5, "tid": 7, "ts": 1716454223885683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834236, "dur": 6,
    "args": {
      "External id": 155118,
      "cbid": 211, "correlation": 155118
    }
  },
  {
    "ph": "s", "id": 155118, "pid": 76337, "tid": -914061504, "ts": 1716454223834236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834294, "dur": 1,
    "args": {
      "External id": 155129,
      "cbid": 251, "correlation": 155129
    }
  },
  {
    "ph": "f", "id": 155129, "pid": 76337, "tid": -914061504, "ts": 1716454223834294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834298, "dur": 0,
    "args": {
      "External id": 155130,
      "cbid": 251, "correlation": 155130
    }
  },
  {
    "ph": "f", "id": 155130, "pid": 76337, "tid": -914061504, "ts": 1716454223834298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223885690, "dur": 7,
    "args": {
      "External id": 155131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155131,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 155131, "pid": 5, "tid": 7, "ts": 1716454223885690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834300, "dur": 11,
    "args": {
      "External id": 155131,
      "cbid": 211, "correlation": 155131
    }
  },
  {
    "ph": "s", "id": 155131, "pid": 76337, "tid": -914061504, "ts": 1716454223834300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223885698, "dur": 3,
    "args": {
      "External id": 155133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155133,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 155133, "pid": 5, "tid": 7, "ts": 1716454223885698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834313, "dur": 5,
    "args": {
      "External id": 155133,
      "cbid": 211, "correlation": 155133
    }
  },
  {
    "ph": "s", "id": 155133, "pid": 76337, "tid": -914061504, "ts": 1716454223834313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223885702, "dur": 91,
    "args": {
      "External id": 155154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155154,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 155154, "pid": 5, "tid": 7, "ts": 1716454223885702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834386, "dur": 13,
    "args": {
      "External id": 155154,
      "cbid": 211, "correlation": 155154
    }
  },
  {
    "ph": "s", "id": 155154, "pid": 76337, "tid": -914061504, "ts": 1716454223834386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834484, "dur": 1,
    "args": {
      "External id": 155172,
      "cbid": 251, "correlation": 155172
    }
  },
  {
    "ph": "f", "id": 155172, "pid": 76337, "tid": -914061504, "ts": 1716454223834484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223885795, "dur": 99,
    "args": {
      "External id": 155174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155174,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155174, "pid": 5, "tid": 7, "ts": 1716454223885795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834489, "dur": 14,
    "args": {
      "External id": 155174,
      "cbid": 211, "correlation": 155174
    }
  },
  {
    "ph": "s", "id": 155174, "pid": 76337, "tid": -914061504, "ts": 1716454223834489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223885895, "dur": 19,
    "args": {
      "External id": 155182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155182,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155182, "pid": 5, "tid": 7, "ts": 1716454223885895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834559, "dur": 12,
    "args": {
      "External id": 155182,
      "cbid": 211, "correlation": 155182
    }
  },
  {
    "ph": "s", "id": 155182, "pid": 76337, "tid": -914061504, "ts": 1716454223834559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223885915, "dur": 37,
    "args": {
      "External id": 155190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155190,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155190, "pid": 5, "tid": 7, "ts": 1716454223885915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834601, "dur": 9,
    "args": {
      "External id": 155190,
      "cbid": 211, "correlation": 155190
    }
  },
  {
    "ph": "s", "id": 155190, "pid": 76337, "tid": -914061504, "ts": 1716454223834601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223885953, "dur": 34,
    "args": {
      "External id": 155212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155212,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155212, "pid": 5, "tid": 7, "ts": 1716454223885953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834653, "dur": 10,
    "args": {
      "External id": 155212,
      "cbid": 211, "correlation": 155212
    }
  },
  {
    "ph": "s", "id": 155212, "pid": 76337, "tid": -914061504, "ts": 1716454223834653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834741, "dur": 1,
    "args": {
      "External id": 155228,
      "cbid": 251, "correlation": 155228
    }
  },
  {
    "ph": "f", "id": 155228, "pid": 76337, "tid": -914061504, "ts": 1716454223834741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834746, "dur": 0,
    "args": {
      "External id": 155230,
      "cbid": 251, "correlation": 155230
    }
  },
  {
    "ph": "f", "id": 155230, "pid": 76337, "tid": -914061504, "ts": 1716454223834746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223885989, "dur": 540,
    "args": {
      "External id": 155231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155231,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155231, "pid": 5, "tid": 7, "ts": 1716454223885989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834750, "dur": 13,
    "args": {
      "External id": 155231,
      "cbid": 211, "correlation": 155231
    }
  },
  {
    "ph": "s", "id": 155231, "pid": 76337, "tid": -914061504, "ts": 1716454223834750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223886531, "dur": 126,
    "args": {
      "External id": 155239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155239,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155239, "pid": 5, "tid": 7, "ts": 1716454223886531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834815, "dur": 12,
    "args": {
      "External id": 155239,
      "cbid": 211, "correlation": 155239
    }
  },
  {
    "ph": "s", "id": 155239, "pid": 76337, "tid": -914061504, "ts": 1716454223834815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223886658, "dur": 126,
    "args": {
      "External id": 155247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155247, "pid": 5, "tid": 7, "ts": 1716454223886658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834846, "dur": 8,
    "args": {
      "External id": 155247,
      "cbid": 211, "correlation": 155247
    }
  },
  {
    "ph": "s", "id": 155247, "pid": 76337, "tid": -914061504, "ts": 1716454223834846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834924, "dur": 1,
    "args": {
      "External id": 155263,
      "cbid": 251, "correlation": 155263
    }
  },
  {
    "ph": "f", "id": 155263, "pid": 76337, "tid": -914061504, "ts": 1716454223834924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223886785, "dur": 305,
    "args": {
      "External id": 155265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155265,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155265, "pid": 5, "tid": 7, "ts": 1716454223886785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834929, "dur": 12,
    "args": {
      "External id": 155265,
      "cbid": 211, "correlation": 155265
    }
  },
  {
    "ph": "s", "id": 155265, "pid": 76337, "tid": -914061504, "ts": 1716454223834929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223887091, "dur": 27,
    "args": {
      "External id": 155273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155273,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155273, "pid": 5, "tid": 7, "ts": 1716454223887091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223834972, "dur": 22,
    "args": {
      "External id": 155273,
      "cbid": 211, "correlation": 155273
    }
  },
  {
    "ph": "s", "id": 155273, "pid": 76337, "tid": -914061504, "ts": 1716454223834972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223887120, "dur": 81,
    "args": {
      "External id": 155284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155284, "pid": 5, "tid": 7, "ts": 1716454223887120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835054, "dur": 12,
    "args": {
      "External id": 155284,
      "cbid": 211, "correlation": 155284
    }
  },
  {
    "ph": "s", "id": 155284, "pid": 76337, "tid": -914061504, "ts": 1716454223835054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835118, "dur": 0,
    "args": {
      "External id": 155296,
      "cbid": 317, "correlation": 155296
    }
  },
  {
    "ph": "f", "id": 155296, "pid": 76337, "tid": -914061504, "ts": 1716454223835118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835119, "dur": 0,
    "args": {
      "External id": 155297,
      "cbid": 203, "correlation": 155297
    }
  },
  {
    "ph": "f", "id": 155297, "pid": 76337, "tid": -914061504, "ts": 1716454223835119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835120, "dur": 0,
    "args": {
      "External id": 155298,
      "cbid": 205, "correlation": 155298
    }
  },
  {
    "ph": "f", "id": 155298, "pid": 76337, "tid": -914061504, "ts": 1716454223835120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887203, "dur": 23,
    "args": {
      "External id": 155302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155302, "pid": 5, "tid": 7, "ts": 1716454223887203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835136, "dur": 13,
    "args": {
      "External id": 155302,
      "cbid": 211, "correlation": 155302
    }
  },
  {
    "ph": "s", "id": 155302, "pid": 76337, "tid": -914061504, "ts": 1716454223835136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223887227, "dur": 121,
    "args": {
      "External id": 155304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155304,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155304, "pid": 5, "tid": 7, "ts": 1716454223887227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835155, "dur": 6,
    "args": {
      "External id": 155304,
      "cbid": 211, "correlation": 155304
    }
  },
  {
    "ph": "s", "id": 155304, "pid": 76337, "tid": -914061504, "ts": 1716454223835155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887349, "dur": 24,
    "args": {
      "External id": 155306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155306,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155306, "pid": 5, "tid": 7, "ts": 1716454223887349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835165, "dur": 5,
    "args": {
      "External id": 155306,
      "cbid": 211, "correlation": 155306
    }
  },
  {
    "ph": "s", "id": 155306, "pid": 76337, "tid": -914061504, "ts": 1716454223835165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223887374, "dur": 32,
    "args": {
      "External id": 155312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155312, "pid": 5, "tid": 7, "ts": 1716454223887374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835193, "dur": 9,
    "args": {
      "External id": 155312,
      "cbid": 211, "correlation": 155312
    }
  },
  {
    "ph": "s", "id": 155312, "pid": 76337, "tid": -914061504, "ts": 1716454223835193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223887408, "dur": 27,
    "args": {
      "External id": 155320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155320,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155320, "pid": 5, "tid": 7, "ts": 1716454223887408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835225, "dur": 8,
    "args": {
      "External id": 155320,
      "cbid": 211, "correlation": 155320
    }
  },
  {
    "ph": "s", "id": 155320, "pid": 76337, "tid": -914061504, "ts": 1716454223835225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835296, "dur": 0,
    "args": {
      "External id": 155330,
      "cbid": 317, "correlation": 155330
    }
  },
  {
    "ph": "f", "id": 155330, "pid": 76337, "tid": -914061504, "ts": 1716454223835296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835297, "dur": 0,
    "args": {
      "External id": 155331,
      "cbid": 203, "correlation": 155331
    }
  },
  {
    "ph": "f", "id": 155331, "pid": 76337, "tid": -914061504, "ts": 1716454223835297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835298, "dur": 0,
    "args": {
      "External id": 155332,
      "cbid": 205, "correlation": 155332
    }
  },
  {
    "ph": "f", "id": 155332, "pid": 76337, "tid": -914061504, "ts": 1716454223835298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887436, "dur": 23,
    "args": {
      "External id": 155336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155336, "pid": 5, "tid": 7, "ts": 1716454223887436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835312, "dur": 12,
    "args": {
      "External id": 155336,
      "cbid": 211, "correlation": 155336
    }
  },
  {
    "ph": "s", "id": 155336, "pid": 76337, "tid": -914061504, "ts": 1716454223835312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887460, "dur": 44,
    "args": {
      "External id": 155338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155338,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155338, "pid": 5, "tid": 7, "ts": 1716454223887460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835327, "dur": 5,
    "args": {
      "External id": 155338,
      "cbid": 211, "correlation": 155338
    }
  },
  {
    "ph": "s", "id": 155338, "pid": 76337, "tid": -914061504, "ts": 1716454223835327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223887506, "dur": 235,
    "args": {
      "External id": 155340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155340,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 155340, "pid": 5, "tid": 7, "ts": 1716454223887506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835338, "dur": 8,
    "args": {
      "External id": 155340,
      "cbid": 211, "correlation": 155340
    }
  },
  {
    "ph": "s", "id": 155340, "pid": 76337, "tid": -914061504, "ts": 1716454223835338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887742, "dur": 6,
    "args": {
      "External id": 155342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155342,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155342, "pid": 5, "tid": 7, "ts": 1716454223887742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835350, "dur": 5,
    "args": {
      "External id": 155342,
      "cbid": 211, "correlation": 155342
    }
  },
  {
    "ph": "s", "id": 155342, "pid": 76337, "tid": -914061504, "ts": 1716454223835350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223887750, "dur": 9,
    "args": {
      "External id": 155348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155348,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155348, "pid": 5, "tid": 7, "ts": 1716454223887750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835376, "dur": 8,
    "args": {
      "External id": 155348,
      "cbid": 211, "correlation": 155348
    }
  },
  {
    "ph": "s", "id": 155348, "pid": 76337, "tid": -914061504, "ts": 1716454223835376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887760, "dur": 12,
    "args": {
      "External id": 155368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155368,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 155368, "pid": 5, "tid": 7, "ts": 1716454223887760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835468, "dur": 12,
    "args": {
      "External id": 155368,
      "cbid": 211, "correlation": 155368
    }
  },
  {
    "ph": "s", "id": 155368, "pid": 76337, "tid": -914061504, "ts": 1716454223835468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223887773, "dur": 4,
    "args": {
      "External id": 155380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155380,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 155380, "pid": 5, "tid": 7, "ts": 1716454223887773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835491, "dur": 7,
    "args": {
      "External id": 155380,
      "cbid": 211, "correlation": 155380
    }
  },
  {
    "ph": "s", "id": 155380, "pid": 76337, "tid": -914061504, "ts": 1716454223835491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223887779, "dur": 12,
    "args": {
      "External id": 155383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155383,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155383, "pid": 5, "tid": 7, "ts": 1716454223887779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835509, "dur": 6,
    "args": {
      "External id": 155383,
      "cbid": 211, "correlation": 155383
    }
  },
  {
    "ph": "s", "id": 155383, "pid": 76337, "tid": -914061504, "ts": 1716454223835509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223887792, "dur": 7,
    "args": {
      "External id": 155392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155392,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155392, "pid": 5, "tid": 7, "ts": 1716454223887792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835549, "dur": 9,
    "args": {
      "External id": 155392,
      "cbid": 211, "correlation": 155392
    }
  },
  {
    "ph": "s", "id": 155392, "pid": 76337, "tid": -914061504, "ts": 1716454223835549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835601, "dur": 0,
    "args": {
      "External id": 155402,
      "cbid": 317, "correlation": 155402
    }
  },
  {
    "ph": "f", "id": 155402, "pid": 76337, "tid": -914061504, "ts": 1716454223835601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835602, "dur": 0,
    "args": {
      "External id": 155403,
      "cbid": 203, "correlation": 155403
    }
  },
  {
    "ph": "f", "id": 155403, "pid": 76337, "tid": -914061504, "ts": 1716454223835602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835603, "dur": 0,
    "args": {
      "External id": 155404,
      "cbid": 205, "correlation": 155404
    }
  },
  {
    "ph": "f", "id": 155404, "pid": 76337, "tid": -914061504, "ts": 1716454223835603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887801, "dur": 6,
    "args": {
      "External id": 155408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155408,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155408, "pid": 5, "tid": 7, "ts": 1716454223887801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835618, "dur": 11,
    "args": {
      "External id": 155408,
      "cbid": 211, "correlation": 155408
    }
  },
  {
    "ph": "s", "id": 155408, "pid": 76337, "tid": -914061504, "ts": 1716454223835618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223887807, "dur": 84,
    "args": {
      "External id": 155410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155410,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155410, "pid": 5, "tid": 7, "ts": 1716454223887807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835632, "dur": 5,
    "args": {
      "External id": 155410,
      "cbid": 211, "correlation": 155410
    }
  },
  {
    "ph": "s", "id": 155410, "pid": 76337, "tid": -914061504, "ts": 1716454223835632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223887893, "dur": 1,
    "args": {
      "External id": 155412,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 155412,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 155412, "pid": 5, "tid": 7, "ts": 1716454223887893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835645, "dur": 9,
    "args": {
      "External id": 155412,
      "cbid": 51, "correlation": 155412
    }
  },
  {
    "ph": "s", "id": 155412, "pid": 76337, "tid": -914061504, "ts": 1716454223835645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223887897, "dur": 537,
    "args": {
      "External id": 155413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155413,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155413, "pid": 5, "tid": 7, "ts": 1716454223887897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835656, "dur": 9,
    "args": {
      "External id": 155413,
      "cbid": 211, "correlation": 155413
    }
  },
  {
    "ph": "s", "id": 155413, "pid": 76337, "tid": -914061504, "ts": 1716454223835656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223888435, "dur": 12,
    "args": {
      "External id": 155415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155415,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155415, "pid": 5, "tid": 7, "ts": 1716454223888435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835668, "dur": 5,
    "args": {
      "External id": 155415,
      "cbid": 211, "correlation": 155415
    }
  },
  {
    "ph": "s", "id": 155415, "pid": 76337, "tid": -914061504, "ts": 1716454223835668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223888449, "dur": 14,
    "args": {
      "External id": 155421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155421, "pid": 5, "tid": 7, "ts": 1716454223888449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835697, "dur": 8,
    "args": {
      "External id": 155421,
      "cbid": 211, "correlation": 155421
    }
  },
  {
    "ph": "s", "id": 155421, "pid": 76337, "tid": -914061504, "ts": 1716454223835697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223888465, "dur": 3,
    "args": {
      "External id": 155429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155429,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 155429, "pid": 5, "tid": 7, "ts": 1716454223888465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835741, "dur": 9,
    "args": {
      "External id": 155429,
      "cbid": 211, "correlation": 155429
    }
  },
  {
    "ph": "s", "id": 155429, "pid": 76337, "tid": -914061504, "ts": 1716454223835741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835805, "dur": 1,
    "args": {
      "External id": 155445,
      "cbid": 251, "correlation": 155445
    }
  },
  {
    "ph": "f", "id": 155445, "pid": 76337, "tid": -914061504, "ts": 1716454223835805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835810, "dur": 0,
    "args": {
      "External id": 155447,
      "cbid": 251, "correlation": 155447
    }
  },
  {
    "ph": "f", "id": 155447, "pid": 76337, "tid": -914061504, "ts": 1716454223835810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223888470, "dur": 13,
    "args": {
      "External id": 155448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155448,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155448, "pid": 5, "tid": 7, "ts": 1716454223888470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835812, "dur": 11,
    "args": {
      "External id": 155448,
      "cbid": 211, "correlation": 155448
    }
  },
  {
    "ph": "s", "id": 155448, "pid": 76337, "tid": -914061504, "ts": 1716454223835812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223888484, "dur": 5,
    "args": {
      "External id": 155450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155450,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155450, "pid": 5, "tid": 7, "ts": 1716454223888484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835825, "dur": 5,
    "args": {
      "External id": 155450,
      "cbid": 211, "correlation": 155450
    }
  },
  {
    "ph": "s", "id": 155450, "pid": 76337, "tid": -914061504, "ts": 1716454223835825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223888490, "dur": 17,
    "args": {
      "External id": 155460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155460,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155460, "pid": 5, "tid": 7, "ts": 1716454223888490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835883, "dur": 13,
    "args": {
      "External id": 155460,
      "cbid": 211, "correlation": 155460
    }
  },
  {
    "ph": "s", "id": 155460, "pid": 76337, "tid": -914061504, "ts": 1716454223835883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223888509, "dur": 18,
    "args": {
      "External id": 155480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155480,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 155480, "pid": 5, "tid": 7, "ts": 1716454223888509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835951, "dur": 10,
    "args": {
      "External id": 155480,
      "cbid": 211, "correlation": 155480
    }
  },
  {
    "ph": "s", "id": 155480, "pid": 76337, "tid": -914061504, "ts": 1716454223835951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223888528, "dur": 5,
    "args": {
      "External id": 155492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155492,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 155492, "pid": 5, "tid": 7, "ts": 1716454223888528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223835972, "dur": 16,
    "args": {
      "External id": 155492,
      "cbid": 211, "correlation": 155492
    }
  },
  {
    "ph": "s", "id": 155492, "pid": 76337, "tid": -914061504, "ts": 1716454223835972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223888534, "dur": 16,
    "args": {
      "External id": 155495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155495,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155495, "pid": 5, "tid": 7, "ts": 1716454223888534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836001, "dur": 7,
    "args": {
      "External id": 155495,
      "cbid": 211, "correlation": 155495
    }
  },
  {
    "ph": "s", "id": 155495, "pid": 76337, "tid": -914061504, "ts": 1716454223836001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223888552, "dur": 11,
    "args": {
      "External id": 155504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155504,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155504, "pid": 5, "tid": 7, "ts": 1716454223888552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836043, "dur": 10,
    "args": {
      "External id": 155504,
      "cbid": 211, "correlation": 155504
    }
  },
  {
    "ph": "s", "id": 155504, "pid": 76337, "tid": -914061504, "ts": 1716454223836043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836105, "dur": 0,
    "args": {
      "External id": 155514,
      "cbid": 317, "correlation": 155514
    }
  },
  {
    "ph": "f", "id": 155514, "pid": 76337, "tid": -914061504, "ts": 1716454223836105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836105, "dur": 0,
    "args": {
      "External id": 155515,
      "cbid": 203, "correlation": 155515
    }
  },
  {
    "ph": "f", "id": 155515, "pid": 76337, "tid": -914061504, "ts": 1716454223836105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836106, "dur": 0,
    "args": {
      "External id": 155516,
      "cbid": 205, "correlation": 155516
    }
  },
  {
    "ph": "f", "id": 155516, "pid": 76337, "tid": -914061504, "ts": 1716454223836106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223888564, "dur": 11,
    "args": {
      "External id": 155520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155520,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155520, "pid": 5, "tid": 7, "ts": 1716454223888564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836123, "dur": 12,
    "args": {
      "External id": 155520,
      "cbid": 211, "correlation": 155520
    }
  },
  {
    "ph": "s", "id": 155520, "pid": 76337, "tid": -914061504, "ts": 1716454223836123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223888576, "dur": 163,
    "args": {
      "External id": 155522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155522,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155522, "pid": 5, "tid": 7, "ts": 1716454223888576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836138, "dur": 5,
    "args": {
      "External id": 155522,
      "cbid": 211, "correlation": 155522
    }
  },
  {
    "ph": "s", "id": 155522, "pid": 76337, "tid": -914061504, "ts": 1716454223836138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223888741, "dur": 1,
    "args": {
      "External id": 155524,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 155524,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 155524, "pid": 5, "tid": 7, "ts": 1716454223888741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836151, "dur": 6,
    "args": {
      "External id": 155524,
      "cbid": 51, "correlation": 155524
    }
  },
  {
    "ph": "s", "id": 155524, "pid": 76337, "tid": -914061504, "ts": 1716454223836151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223888745, "dur": 662,
    "args": {
      "External id": 155525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155525,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155525, "pid": 5, "tid": 7, "ts": 1716454223888745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836159, "dur": 7,
    "args": {
      "External id": 155525,
      "cbid": 211, "correlation": 155525
    }
  },
  {
    "ph": "s", "id": 155525, "pid": 76337, "tid": -914061504, "ts": 1716454223836159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223889409, "dur": 13,
    "args": {
      "External id": 155527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155527,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155527, "pid": 5, "tid": 7, "ts": 1716454223889409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836169, "dur": 5,
    "args": {
      "External id": 155527,
      "cbid": 211, "correlation": 155527
    }
  },
  {
    "ph": "s", "id": 155527, "pid": 76337, "tid": -914061504, "ts": 1716454223836169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223889423, "dur": 15,
    "args": {
      "External id": 155533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155533,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155533, "pid": 5, "tid": 7, "ts": 1716454223889423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836198, "dur": 8,
    "args": {
      "External id": 155533,
      "cbid": 211, "correlation": 155533
    }
  },
  {
    "ph": "s", "id": 155533, "pid": 76337, "tid": -914061504, "ts": 1716454223836198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836256, "dur": 0,
    "args": {
      "External id": 155543,
      "cbid": 317, "correlation": 155543
    }
  },
  {
    "ph": "f", "id": 155543, "pid": 76337, "tid": -914061504, "ts": 1716454223836256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836257, "dur": 0,
    "args": {
      "External id": 155544,
      "cbid": 203, "correlation": 155544
    }
  },
  {
    "ph": "f", "id": 155544, "pid": 76337, "tid": -914061504, "ts": 1716454223836257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836257, "dur": 0,
    "args": {
      "External id": 155545,
      "cbid": 205, "correlation": 155545
    }
  },
  {
    "ph": "f", "id": 155545, "pid": 76337, "tid": -914061504, "ts": 1716454223836257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223889439, "dur": 9,
    "args": {
      "External id": 155549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155549,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155549, "pid": 5, "tid": 7, "ts": 1716454223889439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836272, "dur": 11,
    "args": {
      "External id": 155549,
      "cbid": 211, "correlation": 155549
    }
  },
  {
    "ph": "s", "id": 155549, "pid": 76337, "tid": -914061504, "ts": 1716454223836272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223889449, "dur": 4,
    "args": {
      "External id": 155551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155551,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155551, "pid": 5, "tid": 7, "ts": 1716454223889449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836289, "dur": 6,
    "args": {
      "External id": 155551,
      "cbid": 211, "correlation": 155551
    }
  },
  {
    "ph": "s", "id": 155551, "pid": 76337, "tid": -914061504, "ts": 1716454223836289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836298, "dur": 0,
    "args": {
      "External id": 155552,
      "cbid": 51, "correlation": 155552
    }
  },
  {
    "ph": "s", "id": 155552, "pid": 76337, "tid": -914061504, "ts": 1716454223836298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223889454, "dur": 56,
    "args": {
      "External id": 155553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155553,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 155553, "pid": 5, "tid": 7, "ts": 1716454223889454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836299, "dur": 5,
    "args": {
      "External id": 155553,
      "cbid": 211, "correlation": 155553
    }
  },
  {
    "ph": "s", "id": 155553, "pid": 76337, "tid": -914061504, "ts": 1716454223836299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223889511, "dur": 14,
    "args": {
      "External id": 155558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155558, "pid": 5, "tid": 7, "ts": 1716454223889511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836325, "dur": 9,
    "args": {
      "External id": 155558,
      "cbid": 211, "correlation": 155558
    }
  },
  {
    "ph": "s", "id": 155558, "pid": 76337, "tid": -914061504, "ts": 1716454223836325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223889527, "dur": 12,
    "args": {
      "External id": 155566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155566,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155566, "pid": 5, "tid": 7, "ts": 1716454223889527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836354, "dur": 8,
    "args": {
      "External id": 155566,
      "cbid": 211, "correlation": 155566
    }
  },
  {
    "ph": "s", "id": 155566, "pid": 76337, "tid": -914061504, "ts": 1716454223836354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223889540, "dur": 10,
    "args": {
      "External id": 155574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155574,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155574, "pid": 5, "tid": 7, "ts": 1716454223889540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836383, "dur": 9,
    "args": {
      "External id": 155574,
      "cbid": 211, "correlation": 155574
    }
  },
  {
    "ph": "s", "id": 155574, "pid": 76337, "tid": -914061504, "ts": 1716454223836383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223889552, "dur": 18,
    "args": {
      "External id": 155594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155594,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 155594, "pid": 5, "tid": 7, "ts": 1716454223889552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836463, "dur": 12,
    "args": {
      "External id": 155594,
      "cbid": 211, "correlation": 155594
    }
  },
  {
    "ph": "s", "id": 155594, "pid": 76337, "tid": -914061504, "ts": 1716454223836463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223889571, "dur": 4,
    "args": {
      "External id": 155606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155606,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 155606, "pid": 5, "tid": 7, "ts": 1716454223889571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836485, "dur": 7,
    "args": {
      "External id": 155606,
      "cbid": 211, "correlation": 155606
    }
  },
  {
    "ph": "s", "id": 155606, "pid": 76337, "tid": -914061504, "ts": 1716454223836485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223889577, "dur": 17,
    "args": {
      "External id": 155609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155609, "pid": 5, "tid": 7, "ts": 1716454223889577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836504, "dur": 6,
    "args": {
      "External id": 155609,
      "cbid": 211, "correlation": 155609
    }
  },
  {
    "ph": "s", "id": 155609, "pid": 76337, "tid": -914061504, "ts": 1716454223836504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836561, "dur": 0,
    "args": {
      "External id": 155620,
      "cbid": 317, "correlation": 155620
    }
  },
  {
    "ph": "f", "id": 155620, "pid": 76337, "tid": -914061504, "ts": 1716454223836561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836562, "dur": 0,
    "args": {
      "External id": 155621,
      "cbid": 203, "correlation": 155621
    }
  },
  {
    "ph": "f", "id": 155621, "pid": 76337, "tid": -914061504, "ts": 1716454223836562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836563, "dur": 0,
    "args": {
      "External id": 155622,
      "cbid": 205, "correlation": 155622
    }
  },
  {
    "ph": "f", "id": 155622, "pid": 76337, "tid": -914061504, "ts": 1716454223836563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223889595, "dur": 11,
    "args": {
      "External id": 155626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155626, "pid": 5, "tid": 7, "ts": 1716454223889595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836578, "dur": 13,
    "args": {
      "External id": 155626,
      "cbid": 211, "correlation": 155626
    }
  },
  {
    "ph": "s", "id": 155626, "pid": 76337, "tid": -914061504, "ts": 1716454223836578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223889607, "dur": 3,
    "args": {
      "External id": 155628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155628, "pid": 5, "tid": 7, "ts": 1716454223889607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836596, "dur": 6,
    "args": {
      "External id": 155628,
      "cbid": 211, "correlation": 155628
    }
  },
  {
    "ph": "s", "id": 155628, "pid": 76337, "tid": -914061504, "ts": 1716454223836596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836604, "dur": 0,
    "args": {
      "External id": 155629,
      "cbid": 51, "correlation": 155629
    }
  },
  {
    "ph": "s", "id": 155629, "pid": 76337, "tid": -914061504, "ts": 1716454223836604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223889612, "dur": 98,
    "args": {
      "External id": 155630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155630,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 155630, "pid": 5, "tid": 7, "ts": 1716454223889612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836605, "dur": 6,
    "args": {
      "External id": 155630,
      "cbid": 211, "correlation": 155630
    }
  },
  {
    "ph": "s", "id": 155630, "pid": 76337, "tid": -914061504, "ts": 1716454223836605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223889711, "dur": 16,
    "args": {
      "External id": 155635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155635, "pid": 5, "tid": 7, "ts": 1716454223889711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836634, "dur": 9,
    "args": {
      "External id": 155635,
      "cbid": 211, "correlation": 155635
    }
  },
  {
    "ph": "s", "id": 155635, "pid": 76337, "tid": -914061504, "ts": 1716454223836634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223889728, "dur": 84,
    "args": {
      "External id": 155644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155644,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155644, "pid": 5, "tid": 7, "ts": 1716454223889728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836717, "dur": 14,
    "args": {
      "External id": 155644,
      "cbid": 211, "correlation": 155644
    }
  },
  {
    "ph": "s", "id": 155644, "pid": 76337, "tid": -914061504, "ts": 1716454223836717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223889814, "dur": 30,
    "args": {
      "External id": 155666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155666,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155666, "pid": 5, "tid": 7, "ts": 1716454223889814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836774, "dur": 10,
    "args": {
      "External id": 155666,
      "cbid": 211, "correlation": 155666
    }
  },
  {
    "ph": "s", "id": 155666, "pid": 76337, "tid": -914061504, "ts": 1716454223836774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836865, "dur": 1,
    "args": {
      "External id": 155677,
      "cbid": 251, "correlation": 155677
    }
  },
  {
    "ph": "f", "id": 155677, "pid": 76337, "tid": -914061504, "ts": 1716454223836865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223889845, "dur": 163,
    "args": {
      "External id": 155678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155678,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155678, "pid": 5, "tid": 7, "ts": 1716454223889845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836871, "dur": 13,
    "args": {
      "External id": 155678,
      "cbid": 211, "correlation": 155678
    }
  },
  {
    "ph": "s", "id": 155678, "pid": 76337, "tid": -914061504, "ts": 1716454223836871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836944, "dur": 1,
    "args": {
      "External id": 155689,
      "cbid": 251, "correlation": 155689
    }
  },
  {
    "ph": "f", "id": 155689, "pid": 76337, "tid": -914061504, "ts": 1716454223836944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223890009, "dur": 157,
    "args": {
      "External id": 155690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155690,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155690, "pid": 5, "tid": 7, "ts": 1716454223890009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223836948, "dur": 11,
    "args": {
      "External id": 155690,
      "cbid": 211, "correlation": 155690
    }
  },
  {
    "ph": "s", "id": 155690, "pid": 76337, "tid": -914061504, "ts": 1716454223836948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837021, "dur": 1,
    "args": {
      "External id": 155701,
      "cbid": 251, "correlation": 155701
    }
  },
  {
    "ph": "f", "id": 155701, "pid": 76337, "tid": -914061504, "ts": 1716454223837021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223890167, "dur": 160,
    "args": {
      "External id": 155702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155702,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155702, "pid": 5, "tid": 7, "ts": 1716454223890167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837026, "dur": 13,
    "args": {
      "External id": 155702,
      "cbid": 211, "correlation": 155702
    }
  },
  {
    "ph": "s", "id": 155702, "pid": 76337, "tid": -914061504, "ts": 1716454223837026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223890329, "dur": 340,
    "args": {
      "External id": 155727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155727,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155727, "pid": 5, "tid": 7, "ts": 1716454223890329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837114, "dur": 13,
    "args": {
      "External id": 155727,
      "cbid": 211, "correlation": 155727
    }
  },
  {
    "ph": "s", "id": 155727, "pid": 76337, "tid": -914061504, "ts": 1716454223837114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837217, "dur": 1,
    "args": {
      "External id": 155745,
      "cbid": 251, "correlation": 155745
    }
  },
  {
    "ph": "f", "id": 155745, "pid": 76337, "tid": -914061504, "ts": 1716454223837217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223890670, "dur": 166,
    "args": {
      "External id": 155747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155747,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155747, "pid": 5, "tid": 7, "ts": 1716454223890670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837223, "dur": 14,
    "args": {
      "External id": 155747,
      "cbid": 211, "correlation": 155747
    }
  },
  {
    "ph": "s", "id": 155747, "pid": 76337, "tid": -914061504, "ts": 1716454223837223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223890837, "dur": 19,
    "args": {
      "External id": 155755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155755,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155755, "pid": 5, "tid": 7, "ts": 1716454223890837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837293, "dur": 12,
    "args": {
      "External id": 155755,
      "cbid": 211, "correlation": 155755
    }
  },
  {
    "ph": "s", "id": 155755, "pid": 76337, "tid": -914061504, "ts": 1716454223837293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223890858, "dur": 28,
    "args": {
      "External id": 155763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155763,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155763, "pid": 5, "tid": 7, "ts": 1716454223890858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837332, "dur": 9,
    "args": {
      "External id": 155763,
      "cbid": 211, "correlation": 155763
    }
  },
  {
    "ph": "s", "id": 155763, "pid": 76337, "tid": -914061504, "ts": 1716454223837332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223890887, "dur": 20,
    "args": {
      "External id": 155774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155774, "pid": 5, "tid": 7, "ts": 1716454223890887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837403, "dur": 12,
    "args": {
      "External id": 155774,
      "cbid": 211, "correlation": 155774
    }
  },
  {
    "ph": "s", "id": 155774, "pid": 76337, "tid": -914061504, "ts": 1716454223837403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223890908, "dur": 16,
    "args": {
      "External id": 155796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155796,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155796, "pid": 5, "tid": 7, "ts": 1716454223890908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837434, "dur": 8,
    "args": {
      "External id": 155796,
      "cbid": 211, "correlation": 155796
    }
  },
  {
    "ph": "s", "id": 155796, "pid": 76337, "tid": -914061504, "ts": 1716454223837434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837519, "dur": 1,
    "args": {
      "External id": 155807,
      "cbid": 251, "correlation": 155807
    }
  },
  {
    "ph": "f", "id": 155807, "pid": 76337, "tid": -914061504, "ts": 1716454223837519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223890925, "dur": 91,
    "args": {
      "External id": 155808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155808,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155808, "pid": 5, "tid": 7, "ts": 1716454223890925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837526, "dur": 15,
    "args": {
      "External id": 155808,
      "cbid": 211, "correlation": 155808
    }
  },
  {
    "ph": "s", "id": 155808, "pid": 76337, "tid": -914061504, "ts": 1716454223837526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837599, "dur": 1,
    "args": {
      "External id": 155819,
      "cbid": 251, "correlation": 155819
    }
  },
  {
    "ph": "f", "id": 155819, "pid": 76337, "tid": -914061504, "ts": 1716454223837599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837603, "dur": 0,
    "args": {
      "External id": 155820,
      "cbid": 251, "correlation": 155820
    }
  },
  {
    "ph": "f", "id": 155820, "pid": 76337, "tid": -914061504, "ts": 1716454223837603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223891017, "dur": 12,
    "args": {
      "External id": 155821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155821,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155821, "pid": 5, "tid": 7, "ts": 1716454223891017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837605, "dur": 12,
    "args": {
      "External id": 155821,
      "cbid": 211, "correlation": 155821
    }
  },
  {
    "ph": "s", "id": 155821, "pid": 76337, "tid": -914061504, "ts": 1716454223837605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223891031, "dur": 6,
    "args": {
      "External id": 155823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155823,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155823, "pid": 5, "tid": 7, "ts": 1716454223891031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837618, "dur": 6,
    "args": {
      "External id": 155823,
      "cbid": 211, "correlation": 155823
    }
  },
  {
    "ph": "s", "id": 155823, "pid": 76337, "tid": -914061504, "ts": 1716454223837618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837675, "dur": 1,
    "args": {
      "External id": 155834,
      "cbid": 251, "correlation": 155834
    }
  },
  {
    "ph": "f", "id": 155834, "pid": 76337, "tid": -914061504, "ts": 1716454223837675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837678, "dur": 0,
    "args": {
      "External id": 155835,
      "cbid": 251, "correlation": 155835
    }
  },
  {
    "ph": "f", "id": 155835, "pid": 76337, "tid": -914061504, "ts": 1716454223837678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223891038, "dur": 8,
    "args": {
      "External id": 155836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155836,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155836, "pid": 5, "tid": 7, "ts": 1716454223891038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837680, "dur": 12,
    "args": {
      "External id": 155836,
      "cbid": 211, "correlation": 155836
    }
  },
  {
    "ph": "s", "id": 155836, "pid": 76337, "tid": -914061504, "ts": 1716454223837680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223891047, "dur": 3,
    "args": {
      "External id": 155838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155838,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155838, "pid": 5, "tid": 7, "ts": 1716454223891047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837693, "dur": 5,
    "args": {
      "External id": 155838,
      "cbid": 211, "correlation": 155838
    }
  },
  {
    "ph": "s", "id": 155838, "pid": 76337, "tid": -914061504, "ts": 1716454223837693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223891052, "dur": 54,
    "args": {
      "External id": 155863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155863,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155863, "pid": 5, "tid": 7, "ts": 1716454223891052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837770, "dur": 12,
    "args": {
      "External id": 155863,
      "cbid": 211, "correlation": 155863
    }
  },
  {
    "ph": "s", "id": 155863, "pid": 76337, "tid": -914061504, "ts": 1716454223837770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837869, "dur": 2,
    "args": {
      "External id": 155881,
      "cbid": 251, "correlation": 155881
    }
  },
  {
    "ph": "f", "id": 155881, "pid": 76337, "tid": -914061504, "ts": 1716454223837869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223891107, "dur": 92,
    "args": {
      "External id": 155883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155883,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155883, "pid": 5, "tid": 7, "ts": 1716454223891107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837876, "dur": 16,
    "args": {
      "External id": 155883,
      "cbid": 211, "correlation": 155883
    }
  },
  {
    "ph": "s", "id": 155883, "pid": 76337, "tid": -914061504, "ts": 1716454223837876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223891201, "dur": 10,
    "args": {
      "External id": 155891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155891,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155891, "pid": 5, "tid": 7, "ts": 1716454223891201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223837948, "dur": 12,
    "args": {
      "External id": 155891,
      "cbid": 211, "correlation": 155891
    }
  },
  {
    "ph": "s", "id": 155891, "pid": 76337, "tid": -914061504, "ts": 1716454223837948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223891212, "dur": 21,
    "args": {
      "External id": 155899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155899, "pid": 5, "tid": 7, "ts": 1716454223891212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838001, "dur": 10,
    "args": {
      "External id": 155899,
      "cbid": 211, "correlation": 155899
    }
  },
  {
    "ph": "s", "id": 155899, "pid": 76337, "tid": -914061504, "ts": 1716454223838001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223891234, "dur": 17,
    "args": {
      "External id": 155921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155921,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155921, "pid": 5, "tid": 7, "ts": 1716454223891234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838054, "dur": 10,
    "args": {
      "External id": 155921,
      "cbid": 211, "correlation": 155921
    }
  },
  {
    "ph": "s", "id": 155921, "pid": 76337, "tid": -914061504, "ts": 1716454223838054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838143, "dur": 2,
    "args": {
      "External id": 155937,
      "cbid": 251, "correlation": 155937
    }
  },
  {
    "ph": "f", "id": 155937, "pid": 76337, "tid": -914061504, "ts": 1716454223838143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838148, "dur": 0,
    "args": {
      "External id": 155939,
      "cbid": 251, "correlation": 155939
    }
  },
  {
    "ph": "f", "id": 155939, "pid": 76337, "tid": -914061504, "ts": 1716454223838148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223891252, "dur": 494,
    "args": {
      "External id": 155940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155940,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 155940, "pid": 5, "tid": 7, "ts": 1716454223891252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838151, "dur": 14,
    "args": {
      "External id": 155940,
      "cbid": 211, "correlation": 155940
    }
  },
  {
    "ph": "s", "id": 155940, "pid": 76337, "tid": -914061504, "ts": 1716454223838151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223891747, "dur": 66,
    "args": {
      "External id": 155948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155948, "pid": 5, "tid": 7, "ts": 1716454223891747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838218, "dur": 12,
    "args": {
      "External id": 155948,
      "cbid": 211, "correlation": 155948
    }
  },
  {
    "ph": "s", "id": 155948, "pid": 76337, "tid": -914061504, "ts": 1716454223838218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223891815, "dur": 68,
    "args": {
      "External id": 155956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155956, "pid": 5, "tid": 7, "ts": 1716454223891815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838249, "dur": 8,
    "args": {
      "External id": 155956,
      "cbid": 211, "correlation": 155956
    }
  },
  {
    "ph": "s", "id": 155956, "pid": 76337, "tid": -914061504, "ts": 1716454223838249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838330, "dur": 1,
    "args": {
      "External id": 155972,
      "cbid": 251, "correlation": 155972
    }
  },
  {
    "ph": "f", "id": 155972, "pid": 76337, "tid": -914061504, "ts": 1716454223838330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223891885, "dur": 1,
    "args": {
      "External id": 155974,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 155974,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 155974, "pid": 5, "tid": 7, "ts": 1716454223891885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838336, "dur": 11,
    "args": {
      "External id": 155974,
      "cbid": 51, "correlation": 155974
    }
  },
  {
    "ph": "s", "id": 155974, "pid": 76337, "tid": -914061504, "ts": 1716454223838336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223891888, "dur": 270,
    "args": {
      "External id": 155975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155975,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 155975, "pid": 5, "tid": 7, "ts": 1716454223891888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838349, "dur": 11,
    "args": {
      "External id": 155975,
      "cbid": 211, "correlation": 155975
    }
  },
  {
    "ph": "s", "id": 155975, "pid": 76337, "tid": -914061504, "ts": 1716454223838349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223892159, "dur": 14,
    "args": {
      "External id": 155983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155983,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155983, "pid": 5, "tid": 7, "ts": 1716454223892159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838390, "dur": 10,
    "args": {
      "External id": 155983,
      "cbid": 211, "correlation": 155983
    }
  },
  {
    "ph": "s", "id": 155983, "pid": 76337, "tid": -914061504, "ts": 1716454223838390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223892174, "dur": 37,
    "args": {
      "External id": 155994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 155994,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 155994, "pid": 5, "tid": 7, "ts": 1716454223892174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838458, "dur": 12,
    "args": {
      "External id": 155994,
      "cbid": 211, "correlation": 155994
    }
  },
  {
    "ph": "s", "id": 155994, "pid": 76337, "tid": -914061504, "ts": 1716454223838458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838522, "dur": 0,
    "args": {
      "External id": 156006,
      "cbid": 317, "correlation": 156006
    }
  },
  {
    "ph": "f", "id": 156006, "pid": 76337, "tid": -914061504, "ts": 1716454223838522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838522, "dur": 0,
    "args": {
      "External id": 156007,
      "cbid": 203, "correlation": 156007
    }
  },
  {
    "ph": "f", "id": 156007, "pid": 76337, "tid": -914061504, "ts": 1716454223838522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838523, "dur": 0,
    "args": {
      "External id": 156008,
      "cbid": 205, "correlation": 156008
    }
  },
  {
    "ph": "f", "id": 156008, "pid": 76337, "tid": -914061504, "ts": 1716454223838523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223892213, "dur": 13,
    "args": {
      "External id": 156012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156012,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156012, "pid": 5, "tid": 7, "ts": 1716454223892213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838540, "dur": 12,
    "args": {
      "External id": 156012,
      "cbid": 211, "correlation": 156012
    }
  },
  {
    "ph": "s", "id": 156012, "pid": 76337, "tid": -914061504, "ts": 1716454223838540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223892227, "dur": 4,
    "args": {
      "External id": 156014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156014,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 156014, "pid": 5, "tid": 7, "ts": 1716454223892227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838556, "dur": 6,
    "args": {
      "External id": 156014,
      "cbid": 211, "correlation": 156014
    }
  },
  {
    "ph": "s", "id": 156014, "pid": 76337, "tid": -914061504, "ts": 1716454223838556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838565, "dur": 0,
    "args": {
      "External id": 156015,
      "cbid": 51, "correlation": 156015
    }
  },
  {
    "ph": "s", "id": 156015, "pid": 76337, "tid": -914061504, "ts": 1716454223838565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223892233, "dur": 97,
    "args": {
      "External id": 156016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156016,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 156016, "pid": 5, "tid": 7, "ts": 1716454223892233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838566, "dur": 5,
    "args": {
      "External id": 156016,
      "cbid": 211, "correlation": 156016
    }
  },
  {
    "ph": "s", "id": 156016, "pid": 76337, "tid": -914061504, "ts": 1716454223838566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223892332, "dur": 16,
    "args": {
      "External id": 156021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156021,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156021, "pid": 5, "tid": 7, "ts": 1716454223892332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838593, "dur": 8,
    "args": {
      "External id": 156021,
      "cbid": 211, "correlation": 156021
    }
  },
  {
    "ph": "s", "id": 156021, "pid": 76337, "tid": -914061504, "ts": 1716454223838593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223892350, "dur": 13,
    "args": {
      "External id": 156029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156029,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156029, "pid": 5, "tid": 7, "ts": 1716454223892350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838624, "dur": 9,
    "args": {
      "External id": 156029,
      "cbid": 211, "correlation": 156029
    }
  },
  {
    "ph": "s", "id": 156029, "pid": 76337, "tid": -914061504, "ts": 1716454223838624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223892363, "dur": 18,
    "args": {
      "External id": 156049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156049,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 156049, "pid": 5, "tid": 7, "ts": 1716454223892363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838697, "dur": 11,
    "args": {
      "External id": 156049,
      "cbid": 211, "correlation": 156049
    }
  },
  {
    "ph": "s", "id": 156049, "pid": 76337, "tid": -914061504, "ts": 1716454223838697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223892382, "dur": 5,
    "args": {
      "External id": 156061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156061,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 156061, "pid": 5, "tid": 7, "ts": 1716454223892382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838718, "dur": 6,
    "args": {
      "External id": 156061,
      "cbid": 211, "correlation": 156061
    }
  },
  {
    "ph": "s", "id": 156061, "pid": 76337, "tid": -914061504, "ts": 1716454223838718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223892388, "dur": 17,
    "args": {
      "External id": 156064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156064, "pid": 5, "tid": 7, "ts": 1716454223892388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838737, "dur": 6,
    "args": {
      "External id": 156064,
      "cbid": 211, "correlation": 156064
    }
  },
  {
    "ph": "s", "id": 156064, "pid": 76337, "tid": -914061504, "ts": 1716454223838737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223892407, "dur": 13,
    "args": {
      "External id": 156073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156073,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156073, "pid": 5, "tid": 7, "ts": 1716454223892407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838775, "dur": 10,
    "args": {
      "External id": 156073,
      "cbid": 211, "correlation": 156073
    }
  },
  {
    "ph": "s", "id": 156073, "pid": 76337, "tid": -914061504, "ts": 1716454223838775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838825, "dur": 0,
    "args": {
      "External id": 156083,
      "cbid": 317, "correlation": 156083
    }
  },
  {
    "ph": "f", "id": 156083, "pid": 76337, "tid": -914061504, "ts": 1716454223838825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838826, "dur": 0,
    "args": {
      "External id": 156084,
      "cbid": 203, "correlation": 156084
    }
  },
  {
    "ph": "f", "id": 156084, "pid": 76337, "tid": -914061504, "ts": 1716454223838826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838827, "dur": 0,
    "args": {
      "External id": 156085,
      "cbid": 205, "correlation": 156085
    }
  },
  {
    "ph": "f", "id": 156085, "pid": 76337, "tid": -914061504, "ts": 1716454223838827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223892421, "dur": 11,
    "args": {
      "External id": 156089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156089,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156089, "pid": 5, "tid": 7, "ts": 1716454223892421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838841, "dur": 11,
    "args": {
      "External id": 156089,
      "cbid": 211, "correlation": 156089
    }
  },
  {
    "ph": "s", "id": 156089, "pid": 76337, "tid": -914061504, "ts": 1716454223838841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223892434, "dur": 162,
    "args": {
      "External id": 156091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156091,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156091, "pid": 5, "tid": 7, "ts": 1716454223892434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838855, "dur": 5,
    "args": {
      "External id": 156091,
      "cbid": 211, "correlation": 156091
    }
  },
  {
    "ph": "s", "id": 156091, "pid": 76337, "tid": -914061504, "ts": 1716454223838855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223892598, "dur": 1,
    "args": {
      "External id": 156093,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156093,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 156093, "pid": 5, "tid": 7, "ts": 1716454223892598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838866, "dur": 6,
    "args": {
      "External id": 156093,
      "cbid": 51, "correlation": 156093
    }
  },
  {
    "ph": "s", "id": 156093, "pid": 76337, "tid": -914061504, "ts": 1716454223838866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223892602, "dur": 661,
    "args": {
      "External id": 156094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156094,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156094, "pid": 5, "tid": 7, "ts": 1716454223892602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838873, "dur": 6,
    "args": {
      "External id": 156094,
      "cbid": 211, "correlation": 156094
    }
  },
  {
    "ph": "s", "id": 156094, "pid": 76337, "tid": -914061504, "ts": 1716454223838873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223893264, "dur": 13,
    "args": {
      "External id": 156096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156096, "pid": 5, "tid": 7, "ts": 1716454223893264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838884, "dur": 5,
    "args": {
      "External id": 156096,
      "cbid": 211, "correlation": 156096
    }
  },
  {
    "ph": "s", "id": 156096, "pid": 76337, "tid": -914061504, "ts": 1716454223838884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223893278, "dur": 15,
    "args": {
      "External id": 156102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156102, "pid": 5, "tid": 7, "ts": 1716454223893278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838912, "dur": 8,
    "args": {
      "External id": 156102,
      "cbid": 211, "correlation": 156102
    }
  },
  {
    "ph": "s", "id": 156102, "pid": 76337, "tid": -914061504, "ts": 1716454223838912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223893294, "dur": 3,
    "args": {
      "External id": 156110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156110,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 156110, "pid": 5, "tid": 7, "ts": 1716454223893294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223838956, "dur": 9,
    "args": {
      "External id": 156110,
      "cbid": 211, "correlation": 156110
    }
  },
  {
    "ph": "s", "id": 156110, "pid": 76337, "tid": -914061504, "ts": 1716454223838956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839028, "dur": 1,
    "args": {
      "External id": 156126,
      "cbid": 251, "correlation": 156126
    }
  },
  {
    "ph": "f", "id": 156126, "pid": 76337, "tid": -914061504, "ts": 1716454223839028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839033, "dur": 0,
    "args": {
      "External id": 156128,
      "cbid": 251, "correlation": 156128
    }
  },
  {
    "ph": "f", "id": 156128, "pid": 76337, "tid": -914061504, "ts": 1716454223839033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223893299, "dur": 14,
    "args": {
      "External id": 156129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156129,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156129, "pid": 5, "tid": 7, "ts": 1716454223893299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839035, "dur": 12,
    "args": {
      "External id": 156129,
      "cbid": 211, "correlation": 156129
    }
  },
  {
    "ph": "s", "id": 156129, "pid": 76337, "tid": -914061504, "ts": 1716454223839035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223893314, "dur": 5,
    "args": {
      "External id": 156131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156131,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156131, "pid": 5, "tid": 7, "ts": 1716454223893314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839049, "dur": 5,
    "args": {
      "External id": 156131,
      "cbid": 211, "correlation": 156131
    }
  },
  {
    "ph": "s", "id": 156131, "pid": 76337, "tid": -914061504, "ts": 1716454223839049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223893321, "dur": 17,
    "args": {
      "External id": 156141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156141, "pid": 5, "tid": 7, "ts": 1716454223893321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839107, "dur": 12,
    "args": {
      "External id": 156141,
      "cbid": 211, "correlation": 156141
    }
  },
  {
    "ph": "s", "id": 156141, "pid": 76337, "tid": -914061504, "ts": 1716454223839107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223893338, "dur": 17,
    "args": {
      "External id": 156161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156161,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 156161, "pid": 5, "tid": 7, "ts": 1716454223893338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839172, "dur": 11,
    "args": {
      "External id": 156161,
      "cbid": 211, "correlation": 156161
    }
  },
  {
    "ph": "s", "id": 156161, "pid": 76337, "tid": -914061504, "ts": 1716454223839172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223893357, "dur": 4,
    "args": {
      "External id": 156173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156173,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 156173, "pid": 5, "tid": 7, "ts": 1716454223893357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839193, "dur": 6,
    "args": {
      "External id": 156173,
      "cbid": 211, "correlation": 156173
    }
  },
  {
    "ph": "s", "id": 156173, "pid": 76337, "tid": -914061504, "ts": 1716454223839193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223893362, "dur": 17,
    "args": {
      "External id": 156176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156176, "pid": 5, "tid": 7, "ts": 1716454223893362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839211, "dur": 7,
    "args": {
      "External id": 156176,
      "cbid": 211, "correlation": 156176
    }
  },
  {
    "ph": "s", "id": 156176, "pid": 76337, "tid": -914061504, "ts": 1716454223839211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223893380, "dur": 12,
    "args": {
      "External id": 156185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156185,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156185, "pid": 5, "tid": 7, "ts": 1716454223893380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839252, "dur": 10,
    "args": {
      "External id": 156185,
      "cbid": 211, "correlation": 156185
    }
  },
  {
    "ph": "s", "id": 156185, "pid": 76337, "tid": -914061504, "ts": 1716454223839252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839313, "dur": 0,
    "args": {
      "External id": 156195,
      "cbid": 317, "correlation": 156195
    }
  },
  {
    "ph": "f", "id": 156195, "pid": 76337, "tid": -914061504, "ts": 1716454223839313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839314, "dur": 0,
    "args": {
      "External id": 156196,
      "cbid": 203, "correlation": 156196
    }
  },
  {
    "ph": "f", "id": 156196, "pid": 76337, "tid": -914061504, "ts": 1716454223839314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839315, "dur": 0,
    "args": {
      "External id": 156197,
      "cbid": 205, "correlation": 156197
    }
  },
  {
    "ph": "f", "id": 156197, "pid": 76337, "tid": -914061504, "ts": 1716454223839315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223893394, "dur": 11,
    "args": {
      "External id": 156201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156201,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156201, "pid": 5, "tid": 7, "ts": 1716454223893394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839329, "dur": 12,
    "args": {
      "External id": 156201,
      "cbid": 211, "correlation": 156201
    }
  },
  {
    "ph": "s", "id": 156201, "pid": 76337, "tid": -914061504, "ts": 1716454223839329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223893406, "dur": 163,
    "args": {
      "External id": 156203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156203,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156203, "pid": 5, "tid": 7, "ts": 1716454223893406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839343, "dur": 5,
    "args": {
      "External id": 156203,
      "cbid": 211, "correlation": 156203
    }
  },
  {
    "ph": "s", "id": 156203, "pid": 76337, "tid": -914061504, "ts": 1716454223839343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223893571, "dur": 1,
    "args": {
      "External id": 156205,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156205,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 156205, "pid": 5, "tid": 7, "ts": 1716454223893571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839354, "dur": 6,
    "args": {
      "External id": 156205,
      "cbid": 51, "correlation": 156205
    }
  },
  {
    "ph": "s", "id": 156205, "pid": 76337, "tid": -914061504, "ts": 1716454223839354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223893575, "dur": 646,
    "args": {
      "External id": 156206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156206,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156206, "pid": 5, "tid": 7, "ts": 1716454223893575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839361, "dur": 6,
    "args": {
      "External id": 156206,
      "cbid": 211, "correlation": 156206
    }
  },
  {
    "ph": "s", "id": 156206, "pid": 76337, "tid": -914061504, "ts": 1716454223839361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223894222, "dur": 12,
    "args": {
      "External id": 156208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156208,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156208, "pid": 5, "tid": 7, "ts": 1716454223894222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839373, "dur": 6,
    "args": {
      "External id": 156208,
      "cbid": 211, "correlation": 156208
    }
  },
  {
    "ph": "s", "id": 156208, "pid": 76337, "tid": -914061504, "ts": 1716454223839373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223894235, "dur": 15,
    "args": {
      "External id": 156214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156214, "pid": 5, "tid": 7, "ts": 1716454223894235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839402, "dur": 9,
    "args": {
      "External id": 156214,
      "cbid": 211, "correlation": 156214
    }
  },
  {
    "ph": "s", "id": 156214, "pid": 76337, "tid": -914061504, "ts": 1716454223839402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223894251, "dur": 12,
    "args": {
      "External id": 156222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156222,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156222, "pid": 5, "tid": 7, "ts": 1716454223894251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839434, "dur": 8,
    "args": {
      "External id": 156222,
      "cbid": 211, "correlation": 156222
    }
  },
  {
    "ph": "s", "id": 156222, "pid": 76337, "tid": -914061504, "ts": 1716454223839434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223894264, "dur": 10,
    "args": {
      "External id": 156230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156230,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156230, "pid": 5, "tid": 7, "ts": 1716454223894264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839462, "dur": 9,
    "args": {
      "External id": 156230,
      "cbid": 211, "correlation": 156230
    }
  },
  {
    "ph": "s", "id": 156230, "pid": 76337, "tid": -914061504, "ts": 1716454223839462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223894276, "dur": 20,
    "args": {
      "External id": 156250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156250,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 156250, "pid": 5, "tid": 7, "ts": 1716454223894276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839540, "dur": 12,
    "args": {
      "External id": 156250,
      "cbid": 211, "correlation": 156250
    }
  },
  {
    "ph": "s", "id": 156250, "pid": 76337, "tid": -914061504, "ts": 1716454223839540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223894297, "dur": 4,
    "args": {
      "External id": 156262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156262,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 156262, "pid": 5, "tid": 7, "ts": 1716454223894297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839562, "dur": 6,
    "args": {
      "External id": 156262,
      "cbid": 211, "correlation": 156262
    }
  },
  {
    "ph": "s", "id": 156262, "pid": 76337, "tid": -914061504, "ts": 1716454223839562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223894303, "dur": 16,
    "args": {
      "External id": 156265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156265, "pid": 5, "tid": 7, "ts": 1716454223894303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839581, "dur": 6,
    "args": {
      "External id": 156265,
      "cbid": 211, "correlation": 156265
    }
  },
  {
    "ph": "s", "id": 156265, "pid": 76337, "tid": -914061504, "ts": 1716454223839581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839637, "dur": 0,
    "args": {
      "External id": 156276,
      "cbid": 317, "correlation": 156276
    }
  },
  {
    "ph": "f", "id": 156276, "pid": 76337, "tid": -914061504, "ts": 1716454223839637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839638, "dur": 0,
    "args": {
      "External id": 156277,
      "cbid": 203, "correlation": 156277
    }
  },
  {
    "ph": "f", "id": 156277, "pid": 76337, "tid": -914061504, "ts": 1716454223839638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839639, "dur": 0,
    "args": {
      "External id": 156278,
      "cbid": 205, "correlation": 156278
    }
  },
  {
    "ph": "f", "id": 156278, "pid": 76337, "tid": -914061504, "ts": 1716454223839639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223894320, "dur": 12,
    "args": {
      "External id": 156282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156282,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156282, "pid": 5, "tid": 7, "ts": 1716454223894320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839652, "dur": 11,
    "args": {
      "External id": 156282,
      "cbid": 211, "correlation": 156282
    }
  },
  {
    "ph": "s", "id": 156282, "pid": 76337, "tid": -914061504, "ts": 1716454223839652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223894334, "dur": 4,
    "args": {
      "External id": 156284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 156284, "pid": 5, "tid": 7, "ts": 1716454223894334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839668, "dur": 7,
    "args": {
      "External id": 156284,
      "cbid": 211, "correlation": 156284
    }
  },
  {
    "ph": "s", "id": 156284, "pid": 76337, "tid": -914061504, "ts": 1716454223839668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839677, "dur": 0,
    "args": {
      "External id": 156285,
      "cbid": 51, "correlation": 156285
    }
  },
  {
    "ph": "s", "id": 156285, "pid": 76337, "tid": -914061504, "ts": 1716454223839677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223894339, "dur": 95,
    "args": {
      "External id": 156286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156286,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 156286, "pid": 5, "tid": 7, "ts": 1716454223894339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839678, "dur": 5,
    "args": {
      "External id": 156286,
      "cbid": 211, "correlation": 156286
    }
  },
  {
    "ph": "s", "id": 156286, "pid": 76337, "tid": -914061504, "ts": 1716454223839678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223894434, "dur": 16,
    "args": {
      "External id": 156291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156291,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156291, "pid": 5, "tid": 7, "ts": 1716454223894434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839704, "dur": 8,
    "args": {
      "External id": 156291,
      "cbid": 211, "correlation": 156291
    }
  },
  {
    "ph": "s", "id": 156291, "pid": 76337, "tid": -914061504, "ts": 1716454223839704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223894451, "dur": 84,
    "args": {
      "External id": 156300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156300,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156300, "pid": 5, "tid": 7, "ts": 1716454223894451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839787, "dur": 13,
    "args": {
      "External id": 156300,
      "cbid": 211, "correlation": 156300
    }
  },
  {
    "ph": "s", "id": 156300, "pid": 76337, "tid": -914061504, "ts": 1716454223839787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223894536, "dur": 30,
    "args": {
      "External id": 156322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156322,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156322, "pid": 5, "tid": 7, "ts": 1716454223894536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839842, "dur": 10,
    "args": {
      "External id": 156322,
      "cbid": 211, "correlation": 156322
    }
  },
  {
    "ph": "s", "id": 156322, "pid": 76337, "tid": -914061504, "ts": 1716454223839842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839930, "dur": 1,
    "args": {
      "External id": 156333,
      "cbid": 251, "correlation": 156333
    }
  },
  {
    "ph": "f", "id": 156333, "pid": 76337, "tid": -914061504, "ts": 1716454223839930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223894567, "dur": 142,
    "args": {
      "External id": 156334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156334,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156334, "pid": 5, "tid": 7, "ts": 1716454223894567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223839935, "dur": 13,
    "args": {
      "External id": 156334,
      "cbid": 211, "correlation": 156334
    }
  },
  {
    "ph": "s", "id": 156334, "pid": 76337, "tid": -914061504, "ts": 1716454223839935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840013, "dur": 1,
    "args": {
      "External id": 156345,
      "cbid": 251, "correlation": 156345
    }
  },
  {
    "ph": "f", "id": 156345, "pid": 76337, "tid": -914061504, "ts": 1716454223840013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223894711, "dur": 159,
    "args": {
      "External id": 156346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156346,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156346, "pid": 5, "tid": 7, "ts": 1716454223894711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840017, "dur": 13,
    "args": {
      "External id": 156346,
      "cbid": 211, "correlation": 156346
    }
  },
  {
    "ph": "s", "id": 156346, "pid": 76337, "tid": -914061504, "ts": 1716454223840017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840085, "dur": 1,
    "args": {
      "External id": 156357,
      "cbid": 251, "correlation": 156357
    }
  },
  {
    "ph": "f", "id": 156357, "pid": 76337, "tid": -914061504, "ts": 1716454223840085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223894872, "dur": 158,
    "args": {
      "External id": 156358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156358,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156358, "pid": 5, "tid": 7, "ts": 1716454223894872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840088, "dur": 11,
    "args": {
      "External id": 156358,
      "cbid": 211, "correlation": 156358
    }
  },
  {
    "ph": "s", "id": 156358, "pid": 76337, "tid": -914061504, "ts": 1716454223840088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223895031, "dur": 339,
    "args": {
      "External id": 156383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156383,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156383, "pid": 5, "tid": 7, "ts": 1716454223895031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840171, "dur": 13,
    "args": {
      "External id": 156383,
      "cbid": 211, "correlation": 156383
    }
  },
  {
    "ph": "s", "id": 156383, "pid": 76337, "tid": -914061504, "ts": 1716454223840171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840271, "dur": 1,
    "args": {
      "External id": 156401,
      "cbid": 251, "correlation": 156401
    }
  },
  {
    "ph": "f", "id": 156401, "pid": 76337, "tid": -914061504, "ts": 1716454223840271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223895371, "dur": 167,
    "args": {
      "External id": 156403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156403,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156403, "pid": 5, "tid": 7, "ts": 1716454223895371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840277, "dur": 14,
    "args": {
      "External id": 156403,
      "cbid": 211, "correlation": 156403
    }
  },
  {
    "ph": "s", "id": 156403, "pid": 76337, "tid": -914061504, "ts": 1716454223840277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223895540, "dur": 19,
    "args": {
      "External id": 156411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156411,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156411, "pid": 5, "tid": 7, "ts": 1716454223895540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840347, "dur": 12,
    "args": {
      "External id": 156411,
      "cbid": 211, "correlation": 156411
    }
  },
  {
    "ph": "s", "id": 156411, "pid": 76337, "tid": -914061504, "ts": 1716454223840347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223895560, "dur": 28,
    "args": {
      "External id": 156419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156419,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156419, "pid": 5, "tid": 7, "ts": 1716454223895560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840385, "dur": 8,
    "args": {
      "External id": 156419,
      "cbid": 211, "correlation": 156419
    }
  },
  {
    "ph": "s", "id": 156419, "pid": 76337, "tid": -914061504, "ts": 1716454223840385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223895589, "dur": 18,
    "args": {
      "External id": 156430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156430, "pid": 5, "tid": 7, "ts": 1716454223895589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840455, "dur": 12,
    "args": {
      "External id": 156430,
      "cbid": 211, "correlation": 156430
    }
  },
  {
    "ph": "s", "id": 156430, "pid": 76337, "tid": -914061504, "ts": 1716454223840455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223895609, "dur": 16,
    "args": {
      "External id": 156452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156452,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156452, "pid": 5, "tid": 7, "ts": 1716454223895609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840487, "dur": 7,
    "args": {
      "External id": 156452,
      "cbid": 211, "correlation": 156452
    }
  },
  {
    "ph": "s", "id": 156452, "pid": 76337, "tid": -914061504, "ts": 1716454223840487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840572, "dur": 1,
    "args": {
      "External id": 156463,
      "cbid": 251, "correlation": 156463
    }
  },
  {
    "ph": "f", "id": 156463, "pid": 76337, "tid": -914061504, "ts": 1716454223840572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223895626, "dur": 89,
    "args": {
      "External id": 156464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156464,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 156464, "pid": 5, "tid": 7, "ts": 1716454223895626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840578, "dur": 14,
    "args": {
      "External id": 156464,
      "cbid": 211, "correlation": 156464
    }
  },
  {
    "ph": "s", "id": 156464, "pid": 76337, "tid": -914061504, "ts": 1716454223840578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840648, "dur": 1,
    "args": {
      "External id": 156475,
      "cbid": 251, "correlation": 156475
    }
  },
  {
    "ph": "f", "id": 156475, "pid": 76337, "tid": -914061504, "ts": 1716454223840648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840651, "dur": 0,
    "args": {
      "External id": 156476,
      "cbid": 251, "correlation": 156476
    }
  },
  {
    "ph": "f", "id": 156476, "pid": 76337, "tid": -914061504, "ts": 1716454223840651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223895716, "dur": 13,
    "args": {
      "External id": 156477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156477,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156477, "pid": 5, "tid": 7, "ts": 1716454223895716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840653, "dur": 12,
    "args": {
      "External id": 156477,
      "cbid": 211, "correlation": 156477
    }
  },
  {
    "ph": "s", "id": 156477, "pid": 76337, "tid": -914061504, "ts": 1716454223840653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223895730, "dur": 5,
    "args": {
      "External id": 156479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156479,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156479, "pid": 5, "tid": 7, "ts": 1716454223895730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840667, "dur": 6,
    "args": {
      "External id": 156479,
      "cbid": 211, "correlation": 156479
    }
  },
  {
    "ph": "s", "id": 156479, "pid": 76337, "tid": -914061504, "ts": 1716454223840667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840724, "dur": 1,
    "args": {
      "External id": 156490,
      "cbid": 251, "correlation": 156490
    }
  },
  {
    "ph": "f", "id": 156490, "pid": 76337, "tid": -914061504, "ts": 1716454223840724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840728, "dur": 0,
    "args": {
      "External id": 156491,
      "cbid": 251, "correlation": 156491
    }
  },
  {
    "ph": "f", "id": 156491, "pid": 76337, "tid": -914061504, "ts": 1716454223840728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223895737, "dur": 8,
    "args": {
      "External id": 156492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156492,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156492, "pid": 5, "tid": 7, "ts": 1716454223895737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840730, "dur": 12,
    "args": {
      "External id": 156492,
      "cbid": 211, "correlation": 156492
    }
  },
  {
    "ph": "s", "id": 156492, "pid": 76337, "tid": -914061504, "ts": 1716454223840730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223895746, "dur": 3,
    "args": {
      "External id": 156494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156494,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156494, "pid": 5, "tid": 7, "ts": 1716454223895746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840743, "dur": 5,
    "args": {
      "External id": 156494,
      "cbid": 211, "correlation": 156494
    }
  },
  {
    "ph": "s", "id": 156494, "pid": 76337, "tid": -914061504, "ts": 1716454223840743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223895751, "dur": 54,
    "args": {
      "External id": 156519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156519,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156519, "pid": 5, "tid": 7, "ts": 1716454223895751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840819, "dur": 13,
    "args": {
      "External id": 156519,
      "cbid": 211, "correlation": 156519
    }
  },
  {
    "ph": "s", "id": 156519, "pid": 76337, "tid": -914061504, "ts": 1716454223840819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840918, "dur": 1,
    "args": {
      "External id": 156537,
      "cbid": 251, "correlation": 156537
    }
  },
  {
    "ph": "f", "id": 156537, "pid": 76337, "tid": -914061504, "ts": 1716454223840918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223895806, "dur": 91,
    "args": {
      "External id": 156539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156539,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 156539, "pid": 5, "tid": 7, "ts": 1716454223895806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223840924, "dur": 14,
    "args": {
      "External id": 156539,
      "cbid": 211, "correlation": 156539
    }
  },
  {
    "ph": "s", "id": 156539, "pid": 76337, "tid": -914061504, "ts": 1716454223840924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223895898, "dur": 10,
    "args": {
      "External id": 156547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156547,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156547, "pid": 5, "tid": 7, "ts": 1716454223895898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841004, "dur": 12,
    "args": {
      "External id": 156547,
      "cbid": 211, "correlation": 156547
    }
  },
  {
    "ph": "s", "id": 156547, "pid": 76337, "tid": -914061504, "ts": 1716454223841004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223895909, "dur": 21,
    "args": {
      "External id": 156555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156555, "pid": 5, "tid": 7, "ts": 1716454223895909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841047, "dur": 9,
    "args": {
      "External id": 156555,
      "cbid": 211, "correlation": 156555
    }
  },
  {
    "ph": "s", "id": 156555, "pid": 76337, "tid": -914061504, "ts": 1716454223841047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223895932, "dur": 18,
    "args": {
      "External id": 156577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156577,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156577, "pid": 5, "tid": 7, "ts": 1716454223895932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841098, "dur": 10,
    "args": {
      "External id": 156577,
      "cbid": 211, "correlation": 156577
    }
  },
  {
    "ph": "s", "id": 156577, "pid": 76337, "tid": -914061504, "ts": 1716454223841098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841186, "dur": 1,
    "args": {
      "External id": 156593,
      "cbid": 251, "correlation": 156593
    }
  },
  {
    "ph": "f", "id": 156593, "pid": 76337, "tid": -914061504, "ts": 1716454223841186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841191, "dur": 0,
    "args": {
      "External id": 156595,
      "cbid": 251, "correlation": 156595
    }
  },
  {
    "ph": "f", "id": 156595, "pid": 76337, "tid": -914061504, "ts": 1716454223841191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223895951, "dur": 495,
    "args": {
      "External id": 156596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156596,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156596, "pid": 5, "tid": 7, "ts": 1716454223895951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841193, "dur": 12,
    "args": {
      "External id": 156596,
      "cbid": 211, "correlation": 156596
    }
  },
  {
    "ph": "s", "id": 156596, "pid": 76337, "tid": -914061504, "ts": 1716454223841193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223896447, "dur": 66,
    "args": {
      "External id": 156604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156604, "pid": 5, "tid": 7, "ts": 1716454223896447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841257, "dur": 12,
    "args": {
      "External id": 156604,
      "cbid": 211, "correlation": 156604
    }
  },
  {
    "ph": "s", "id": 156604, "pid": 76337, "tid": -914061504, "ts": 1716454223841257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223896514, "dur": 69,
    "args": {
      "External id": 156612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156612, "pid": 5, "tid": 7, "ts": 1716454223896514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841288, "dur": 8,
    "args": {
      "External id": 156612,
      "cbid": 211, "correlation": 156612
    }
  },
  {
    "ph": "s", "id": 156612, "pid": 76337, "tid": -914061504, "ts": 1716454223841288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841366, "dur": 1,
    "args": {
      "External id": 156628,
      "cbid": 251, "correlation": 156628
    }
  },
  {
    "ph": "f", "id": 156628, "pid": 76337, "tid": -914061504, "ts": 1716454223841366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223896585, "dur": 1,
    "args": {
      "External id": 156630,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156630,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 156630, "pid": 5, "tid": 7, "ts": 1716454223896585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841371, "dur": 10,
    "args": {
      "External id": 156630,
      "cbid": 51, "correlation": 156630
    }
  },
  {
    "ph": "s", "id": 156630, "pid": 76337, "tid": -914061504, "ts": 1716454223841371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223896589, "dur": 271,
    "args": {
      "External id": 156631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156631,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 156631, "pid": 5, "tid": 7, "ts": 1716454223896589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841384, "dur": 11,
    "args": {
      "External id": 156631,
      "cbid": 211, "correlation": 156631
    }
  },
  {
    "ph": "s", "id": 156631, "pid": 76337, "tid": -914061504, "ts": 1716454223841384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223896862, "dur": 14,
    "args": {
      "External id": 156639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156639,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156639, "pid": 5, "tid": 7, "ts": 1716454223896862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841426, "dur": 11,
    "args": {
      "External id": 156639,
      "cbid": 211, "correlation": 156639
    }
  },
  {
    "ph": "s", "id": 156639, "pid": 76337, "tid": -914061504, "ts": 1716454223841426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223896877, "dur": 37,
    "args": {
      "External id": 156650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156650, "pid": 5, "tid": 7, "ts": 1716454223896877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841494, "dur": 12,
    "args": {
      "External id": 156650,
      "cbid": 211, "correlation": 156650
    }
  },
  {
    "ph": "s", "id": 156650, "pid": 76337, "tid": -914061504, "ts": 1716454223841494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841558, "dur": 0,
    "args": {
      "External id": 156662,
      "cbid": 317, "correlation": 156662
    }
  },
  {
    "ph": "f", "id": 156662, "pid": 76337, "tid": -914061504, "ts": 1716454223841558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841559, "dur": 0,
    "args": {
      "External id": 156663,
      "cbid": 203, "correlation": 156663
    }
  },
  {
    "ph": "f", "id": 156663, "pid": 76337, "tid": -914061504, "ts": 1716454223841559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841559, "dur": 0,
    "args": {
      "External id": 156664,
      "cbid": 205, "correlation": 156664
    }
  },
  {
    "ph": "f", "id": 156664, "pid": 76337, "tid": -914061504, "ts": 1716454223841559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223896915, "dur": 13,
    "args": {
      "External id": 156668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156668,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156668, "pid": 5, "tid": 7, "ts": 1716454223896915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841575, "dur": 13,
    "args": {
      "External id": 156668,
      "cbid": 211, "correlation": 156668
    }
  },
  {
    "ph": "s", "id": 156668, "pid": 76337, "tid": -914061504, "ts": 1716454223841575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223896929, "dur": 4,
    "args": {
      "External id": 156670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156670,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 156670, "pid": 5, "tid": 7, "ts": 1716454223896929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841593, "dur": 5,
    "args": {
      "External id": 156670,
      "cbid": 211, "correlation": 156670
    }
  },
  {
    "ph": "s", "id": 156670, "pid": 76337, "tid": -914061504, "ts": 1716454223841593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841601, "dur": 0,
    "args": {
      "External id": 156671,
      "cbid": 51, "correlation": 156671
    }
  },
  {
    "ph": "s", "id": 156671, "pid": 76337, "tid": -914061504, "ts": 1716454223841601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223896935, "dur": 97,
    "args": {
      "External id": 156672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156672,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 156672, "pid": 5, "tid": 7, "ts": 1716454223896935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841602, "dur": 5,
    "args": {
      "External id": 156672,
      "cbid": 211, "correlation": 156672
    }
  },
  {
    "ph": "s", "id": 156672, "pid": 76337, "tid": -914061504, "ts": 1716454223841602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223897033, "dur": 17,
    "args": {
      "External id": 156677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156677, "pid": 5, "tid": 7, "ts": 1716454223897033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841629, "dur": 9,
    "args": {
      "External id": 156677,
      "cbid": 211, "correlation": 156677
    }
  },
  {
    "ph": "s", "id": 156677, "pid": 76337, "tid": -914061504, "ts": 1716454223841629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223897051, "dur": 12,
    "args": {
      "External id": 156685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156685,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156685, "pid": 5, "tid": 7, "ts": 1716454223897051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841660, "dur": 8,
    "args": {
      "External id": 156685,
      "cbid": 211, "correlation": 156685
    }
  },
  {
    "ph": "s", "id": 156685, "pid": 76337, "tid": -914061504, "ts": 1716454223841660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841730, "dur": 0,
    "args": {
      "External id": 156695,
      "cbid": 317, "correlation": 156695
    }
  },
  {
    "ph": "f", "id": 156695, "pid": 76337, "tid": -914061504, "ts": 1716454223841730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841731, "dur": 0,
    "args": {
      "External id": 156696,
      "cbid": 203, "correlation": 156696
    }
  },
  {
    "ph": "f", "id": 156696, "pid": 76337, "tid": -914061504, "ts": 1716454223841731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841732, "dur": 0,
    "args": {
      "External id": 156697,
      "cbid": 205, "correlation": 156697
    }
  },
  {
    "ph": "f", "id": 156697, "pid": 76337, "tid": -914061504, "ts": 1716454223841732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897064, "dur": 11,
    "args": {
      "External id": 156701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156701,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156701, "pid": 5, "tid": 7, "ts": 1716454223897064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841746, "dur": 12,
    "args": {
      "External id": 156701,
      "cbid": 211, "correlation": 156701
    }
  },
  {
    "ph": "s", "id": 156701, "pid": 76337, "tid": -914061504, "ts": 1716454223841746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897077, "dur": 162,
    "args": {
      "External id": 156703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156703,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156703, "pid": 5, "tid": 7, "ts": 1716454223897077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841761, "dur": 5,
    "args": {
      "External id": 156703,
      "cbid": 211, "correlation": 156703
    }
  },
  {
    "ph": "s", "id": 156703, "pid": 76337, "tid": -914061504, "ts": 1716454223841761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223897241, "dur": 1,
    "args": {
      "External id": 156705,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156705,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 156705, "pid": 5, "tid": 7, "ts": 1716454223897241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841772, "dur": 7,
    "args": {
      "External id": 156705,
      "cbid": 51, "correlation": 156705
    }
  },
  {
    "ph": "s", "id": 156705, "pid": 76337, "tid": -914061504, "ts": 1716454223841772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223897245, "dur": 197,
    "args": {
      "External id": 156706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156706,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 156706, "pid": 5, "tid": 7, "ts": 1716454223897245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841781, "dur": 8,
    "args": {
      "External id": 156706,
      "cbid": 211, "correlation": 156706
    }
  },
  {
    "ph": "s", "id": 156706, "pid": 76337, "tid": -914061504, "ts": 1716454223841781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897444, "dur": 6,
    "args": {
      "External id": 156708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156708,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156708, "pid": 5, "tid": 7, "ts": 1716454223897444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841793, "dur": 5,
    "args": {
      "External id": 156708,
      "cbid": 211, "correlation": 156708
    }
  },
  {
    "ph": "s", "id": 156708, "pid": 76337, "tid": -914061504, "ts": 1716454223841793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223897451, "dur": 6,
    "args": {
      "External id": 156714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156714,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156714, "pid": 5, "tid": 7, "ts": 1716454223897451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841820, "dur": 9,
    "args": {
      "External id": 156714,
      "cbid": 211, "correlation": 156714
    }
  },
  {
    "ph": "s", "id": 156714, "pid": 76337, "tid": -914061504, "ts": 1716454223841820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897459, "dur": 11,
    "args": {
      "External id": 156734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156734,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 156734, "pid": 5, "tid": 7, "ts": 1716454223897459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841913, "dur": 12,
    "args": {
      "External id": 156734,
      "cbid": 211, "correlation": 156734
    }
  },
  {
    "ph": "s", "id": 156734, "pid": 76337, "tid": -914061504, "ts": 1716454223841913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223897471, "dur": 4,
    "args": {
      "External id": 156746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156746,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 156746, "pid": 5, "tid": 7, "ts": 1716454223897471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841937, "dur": 6,
    "args": {
      "External id": 156746,
      "cbid": 211, "correlation": 156746
    }
  },
  {
    "ph": "s", "id": 156746, "pid": 76337, "tid": -914061504, "ts": 1716454223841937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223897477, "dur": 9,
    "args": {
      "External id": 156749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156749,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156749, "pid": 5, "tid": 7, "ts": 1716454223897477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223841956, "dur": 7,
    "args": {
      "External id": 156749,
      "cbid": 211, "correlation": 156749
    }
  },
  {
    "ph": "s", "id": 156749, "pid": 76337, "tid": -914061504, "ts": 1716454223841956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223897487, "dur": 5,
    "args": {
      "External id": 156758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156758,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156758, "pid": 5, "tid": 7, "ts": 1716454223897487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842006, "dur": 10,
    "args": {
      "External id": 156758,
      "cbid": 211, "correlation": 156758
    }
  },
  {
    "ph": "s", "id": 156758, "pid": 76337, "tid": -914061504, "ts": 1716454223842006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842059, "dur": 0,
    "args": {
      "External id": 156768,
      "cbid": 317, "correlation": 156768
    }
  },
  {
    "ph": "f", "id": 156768, "pid": 76337, "tid": -914061504, "ts": 1716454223842059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842060, "dur": 0,
    "args": {
      "External id": 156769,
      "cbid": 203, "correlation": 156769
    }
  },
  {
    "ph": "f", "id": 156769, "pid": 76337, "tid": -914061504, "ts": 1716454223842060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842061, "dur": 0,
    "args": {
      "External id": 156770,
      "cbid": 205, "correlation": 156770
    }
  },
  {
    "ph": "f", "id": 156770, "pid": 76337, "tid": -914061504, "ts": 1716454223842061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897493, "dur": 5,
    "args": {
      "External id": 156774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156774,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156774, "pid": 5, "tid": 7, "ts": 1716454223897493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842076, "dur": 12,
    "args": {
      "External id": 156774,
      "cbid": 211, "correlation": 156774
    }
  },
  {
    "ph": "s", "id": 156774, "pid": 76337, "tid": -914061504, "ts": 1716454223842076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897499, "dur": 162,
    "args": {
      "External id": 156776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156776,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156776, "pid": 5, "tid": 7, "ts": 1716454223897499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842091, "dur": 5,
    "args": {
      "External id": 156776,
      "cbid": 211, "correlation": 156776
    }
  },
  {
    "ph": "s", "id": 156776, "pid": 76337, "tid": -914061504, "ts": 1716454223842091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223897663, "dur": 1,
    "args": {
      "External id": 156778,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156778,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 156778, "pid": 5, "tid": 7, "ts": 1716454223897663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842102, "dur": 6,
    "args": {
      "External id": 156778,
      "cbid": 51, "correlation": 156778
    }
  },
  {
    "ph": "s", "id": 156778, "pid": 76337, "tid": -914061504, "ts": 1716454223842102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223897667, "dur": 270,
    "args": {
      "External id": 156779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156779,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156779, "pid": 5, "tid": 7, "ts": 1716454223897667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842109, "dur": 6,
    "args": {
      "External id": 156779,
      "cbid": 211, "correlation": 156779
    }
  },
  {
    "ph": "s", "id": 156779, "pid": 76337, "tid": -914061504, "ts": 1716454223842109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897938, "dur": 6,
    "args": {
      "External id": 156781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156781,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156781, "pid": 5, "tid": 7, "ts": 1716454223897938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842120, "dur": 5,
    "args": {
      "External id": 156781,
      "cbid": 211, "correlation": 156781
    }
  },
  {
    "ph": "s", "id": 156781, "pid": 76337, "tid": -914061504, "ts": 1716454223842120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223897946, "dur": 6,
    "args": {
      "External id": 156787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156787,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156787, "pid": 5, "tid": 7, "ts": 1716454223897946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842149, "dur": 8,
    "args": {
      "External id": 156787,
      "cbid": 211, "correlation": 156787
    }
  },
  {
    "ph": "s", "id": 156787, "pid": 76337, "tid": -914061504, "ts": 1716454223842149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223897953, "dur": 3,
    "args": {
      "External id": 156795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156795,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 156795, "pid": 5, "tid": 7, "ts": 1716454223897953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842192, "dur": 9,
    "args": {
      "External id": 156795,
      "cbid": 211, "correlation": 156795
    }
  },
  {
    "ph": "s", "id": 156795, "pid": 76337, "tid": -914061504, "ts": 1716454223842192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842257, "dur": 1,
    "args": {
      "External id": 156811,
      "cbid": 251, "correlation": 156811
    }
  },
  {
    "ph": "f", "id": 156811, "pid": 76337, "tid": -914061504, "ts": 1716454223842257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842262, "dur": 0,
    "args": {
      "External id": 156813,
      "cbid": 251, "correlation": 156813
    }
  },
  {
    "ph": "f", "id": 156813, "pid": 76337, "tid": -914061504, "ts": 1716454223842262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223897958, "dur": 14,
    "args": {
      "External id": 156814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156814,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156814, "pid": 5, "tid": 7, "ts": 1716454223897958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842264, "dur": 12,
    "args": {
      "External id": 156814,
      "cbid": 211, "correlation": 156814
    }
  },
  {
    "ph": "s", "id": 156814, "pid": 76337, "tid": -914061504, "ts": 1716454223842264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223897972, "dur": 5,
    "args": {
      "External id": 156816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156816,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156816, "pid": 5, "tid": 7, "ts": 1716454223897972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842278, "dur": 5,
    "args": {
      "External id": 156816,
      "cbid": 211, "correlation": 156816
    }
  },
  {
    "ph": "s", "id": 156816, "pid": 76337, "tid": -914061504, "ts": 1716454223842278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223897979, "dur": 6,
    "args": {
      "External id": 156826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156826,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156826, "pid": 5, "tid": 7, "ts": 1716454223897979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842338, "dur": 12,
    "args": {
      "External id": 156826,
      "cbid": 211, "correlation": 156826
    }
  },
  {
    "ph": "s", "id": 156826, "pid": 76337, "tid": -914061504, "ts": 1716454223842338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223897986, "dur": 10,
    "args": {
      "External id": 156846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156846,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 156846, "pid": 5, "tid": 7, "ts": 1716454223897986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842405, "dur": 11,
    "args": {
      "External id": 156846,
      "cbid": 211, "correlation": 156846
    }
  },
  {
    "ph": "s", "id": 156846, "pid": 76337, "tid": -914061504, "ts": 1716454223842405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223897997, "dur": 4,
    "args": {
      "External id": 156858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156858,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 156858, "pid": 5, "tid": 7, "ts": 1716454223897997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842426, "dur": 6,
    "args": {
      "External id": 156858,
      "cbid": 211, "correlation": 156858
    }
  },
  {
    "ph": "s", "id": 156858, "pid": 76337, "tid": -914061504, "ts": 1716454223842426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223898002, "dur": 7,
    "args": {
      "External id": 156861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156861,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156861, "pid": 5, "tid": 7, "ts": 1716454223898002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842445, "dur": 6,
    "args": {
      "External id": 156861,
      "cbid": 211, "correlation": 156861
    }
  },
  {
    "ph": "s", "id": 156861, "pid": 76337, "tid": -914061504, "ts": 1716454223842445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223898011, "dur": 5,
    "args": {
      "External id": 156870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156870,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156870, "pid": 5, "tid": 7, "ts": 1716454223898011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842485, "dur": 10,
    "args": {
      "External id": 156870,
      "cbid": 211, "correlation": 156870
    }
  },
  {
    "ph": "s", "id": 156870, "pid": 76337, "tid": -914061504, "ts": 1716454223842485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842548, "dur": 0,
    "args": {
      "External id": 156880,
      "cbid": 317, "correlation": 156880
    }
  },
  {
    "ph": "f", "id": 156880, "pid": 76337, "tid": -914061504, "ts": 1716454223842548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842549, "dur": 0,
    "args": {
      "External id": 156881,
      "cbid": 203, "correlation": 156881
    }
  },
  {
    "ph": "f", "id": 156881, "pid": 76337, "tid": -914061504, "ts": 1716454223842549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842550, "dur": 0,
    "args": {
      "External id": 156882,
      "cbid": 205, "correlation": 156882
    }
  },
  {
    "ph": "f", "id": 156882, "pid": 76337, "tid": -914061504, "ts": 1716454223842550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898017, "dur": 5,
    "args": {
      "External id": 156886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156886,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156886, "pid": 5, "tid": 7, "ts": 1716454223898017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842564, "dur": 13,
    "args": {
      "External id": 156886,
      "cbid": 211, "correlation": 156886
    }
  },
  {
    "ph": "s", "id": 156886, "pid": 76337, "tid": -914061504, "ts": 1716454223842564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898023, "dur": 163,
    "args": {
      "External id": 156888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156888,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156888, "pid": 5, "tid": 7, "ts": 1716454223898023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842579, "dur": 5,
    "args": {
      "External id": 156888,
      "cbid": 211, "correlation": 156888
    }
  },
  {
    "ph": "s", "id": 156888, "pid": 76337, "tid": -914061504, "ts": 1716454223842579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223898188, "dur": 1,
    "args": {
      "External id": 156890,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156890,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 156890, "pid": 5, "tid": 7, "ts": 1716454223898188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842590, "dur": 6,
    "args": {
      "External id": 156890,
      "cbid": 51, "correlation": 156890
    }
  },
  {
    "ph": "s", "id": 156890, "pid": 76337, "tid": -914061504, "ts": 1716454223842590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223898192, "dur": 259,
    "args": {
      "External id": 156891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156891,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156891, "pid": 5, "tid": 7, "ts": 1716454223898192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842597, "dur": 6,
    "args": {
      "External id": 156891,
      "cbid": 211, "correlation": 156891
    }
  },
  {
    "ph": "s", "id": 156891, "pid": 76337, "tid": -914061504, "ts": 1716454223842597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898452, "dur": 6,
    "args": {
      "External id": 156893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156893,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156893, "pid": 5, "tid": 7, "ts": 1716454223898452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842607, "dur": 5,
    "args": {
      "External id": 156893,
      "cbid": 211, "correlation": 156893
    }
  },
  {
    "ph": "s", "id": 156893, "pid": 76337, "tid": -914061504, "ts": 1716454223842607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223898459, "dur": 6,
    "args": {
      "External id": 156899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156899, "pid": 5, "tid": 7, "ts": 1716454223898459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842636, "dur": 8,
    "args": {
      "External id": 156899,
      "cbid": 211, "correlation": 156899
    }
  },
  {
    "ph": "s", "id": 156899, "pid": 76337, "tid": -914061504, "ts": 1716454223842636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223898467, "dur": 5,
    "args": {
      "External id": 156907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156907,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156907, "pid": 5, "tid": 7, "ts": 1716454223898467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842668, "dur": 9,
    "args": {
      "External id": 156907,
      "cbid": 211, "correlation": 156907
    }
  },
  {
    "ph": "s", "id": 156907, "pid": 76337, "tid": -914061504, "ts": 1716454223842668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223898473, "dur": 4,
    "args": {
      "External id": 156915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156915,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156915, "pid": 5, "tid": 7, "ts": 1716454223898473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842697, "dur": 8,
    "args": {
      "External id": 156915,
      "cbid": 211, "correlation": 156915
    }
  },
  {
    "ph": "s", "id": 156915, "pid": 76337, "tid": -914061504, "ts": 1716454223842697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898479, "dur": 10,
    "args": {
      "External id": 156935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156935,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 156935, "pid": 5, "tid": 7, "ts": 1716454223898479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842770, "dur": 12,
    "args": {
      "External id": 156935,
      "cbid": 211, "correlation": 156935
    }
  },
  {
    "ph": "s", "id": 156935, "pid": 76337, "tid": -914061504, "ts": 1716454223842770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223898490, "dur": 4,
    "args": {
      "External id": 156947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156947,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 156947, "pid": 5, "tid": 7, "ts": 1716454223898490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842792, "dur": 6,
    "args": {
      "External id": 156947,
      "cbid": 211, "correlation": 156947
    }
  },
  {
    "ph": "s", "id": 156947, "pid": 76337, "tid": -914061504, "ts": 1716454223842792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223898495, "dur": 6,
    "args": {
      "External id": 156950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156950, "pid": 5, "tid": 7, "ts": 1716454223898495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842809, "dur": 7,
    "args": {
      "External id": 156950,
      "cbid": 211, "correlation": 156950
    }
  },
  {
    "ph": "s", "id": 156950, "pid": 76337, "tid": -914061504, "ts": 1716454223842809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223898502, "dur": 5,
    "args": {
      "External id": 156959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156959,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156959, "pid": 5, "tid": 7, "ts": 1716454223898502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842848, "dur": 9,
    "args": {
      "External id": 156959,
      "cbid": 211, "correlation": 156959
    }
  },
  {
    "ph": "s", "id": 156959, "pid": 76337, "tid": -914061504, "ts": 1716454223842848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842900, "dur": 0,
    "args": {
      "External id": 156969,
      "cbid": 317, "correlation": 156969
    }
  },
  {
    "ph": "f", "id": 156969, "pid": 76337, "tid": -914061504, "ts": 1716454223842900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842901, "dur": 0,
    "args": {
      "External id": 156970,
      "cbid": 203, "correlation": 156970
    }
  },
  {
    "ph": "f", "id": 156970, "pid": 76337, "tid": -914061504, "ts": 1716454223842901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842902, "dur": 0,
    "args": {
      "External id": 156971,
      "cbid": 205, "correlation": 156971
    }
  },
  {
    "ph": "f", "id": 156971, "pid": 76337, "tid": -914061504, "ts": 1716454223842902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898508, "dur": 5,
    "args": {
      "External id": 156975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156975,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156975, "pid": 5, "tid": 7, "ts": 1716454223898508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842915, "dur": 11,
    "args": {
      "External id": 156975,
      "cbid": 211, "correlation": 156975
    }
  },
  {
    "ph": "s", "id": 156975, "pid": 76337, "tid": -914061504, "ts": 1716454223842915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898514, "dur": 162,
    "args": {
      "External id": 156977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156977,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156977, "pid": 5, "tid": 7, "ts": 1716454223898514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842929, "dur": 6,
    "args": {
      "External id": 156977,
      "cbid": 211, "correlation": 156977
    }
  },
  {
    "ph": "s", "id": 156977, "pid": 76337, "tid": -914061504, "ts": 1716454223842929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223898679, "dur": 1,
    "args": {
      "External id": 156979,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 156979,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 156979, "pid": 5, "tid": 7, "ts": 1716454223898679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842940, "dur": 6,
    "args": {
      "External id": 156979,
      "cbid": 51, "correlation": 156979
    }
  },
  {
    "ph": "s", "id": 156979, "pid": 76337, "tid": -914061504, "ts": 1716454223842940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223898683, "dur": 258,
    "args": {
      "External id": 156980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156980,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 156980, "pid": 5, "tid": 7, "ts": 1716454223898683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842947, "dur": 6,
    "args": {
      "External id": 156980,
      "cbid": 211, "correlation": 156980
    }
  },
  {
    "ph": "s", "id": 156980, "pid": 76337, "tid": -914061504, "ts": 1716454223842947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898942, "dur": 6,
    "args": {
      "External id": 156982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 156982, "pid": 5, "tid": 7, "ts": 1716454223898942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842956, "dur": 5,
    "args": {
      "External id": 156982,
      "cbid": 211, "correlation": 156982
    }
  },
  {
    "ph": "s", "id": 156982, "pid": 76337, "tid": -914061504, "ts": 1716454223842956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223898949, "dur": 6,
    "args": {
      "External id": 156988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156988,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 156988, "pid": 5, "tid": 7, "ts": 1716454223898949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223842993, "dur": 9,
    "args": {
      "External id": 156988,
      "cbid": 211, "correlation": 156988
    }
  },
  {
    "ph": "s", "id": 156988, "pid": 76337, "tid": -914061504, "ts": 1716454223842993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223898956, "dur": 3,
    "args": {
      "External id": 156996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 156996,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 156996, "pid": 5, "tid": 7, "ts": 1716454223898956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843039, "dur": 9,
    "args": {
      "External id": 156996,
      "cbid": 211, "correlation": 156996
    }
  },
  {
    "ph": "s", "id": 156996, "pid": 76337, "tid": -914061504, "ts": 1716454223843039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843102, "dur": 1,
    "args": {
      "External id": 157012,
      "cbid": 251, "correlation": 157012
    }
  },
  {
    "ph": "f", "id": 157012, "pid": 76337, "tid": -914061504, "ts": 1716454223843102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843107, "dur": 0,
    "args": {
      "External id": 157014,
      "cbid": 251, "correlation": 157014
    }
  },
  {
    "ph": "f", "id": 157014, "pid": 76337, "tid": -914061504, "ts": 1716454223843107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223898961, "dur": 10,
    "args": {
      "External id": 157015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157015,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157015, "pid": 5, "tid": 7, "ts": 1716454223898961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843108, "dur": 11,
    "args": {
      "External id": 157015,
      "cbid": 211, "correlation": 157015
    }
  },
  {
    "ph": "s", "id": 157015, "pid": 76337, "tid": -914061504, "ts": 1716454223843108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223898972, "dur": 4,
    "args": {
      "External id": 157017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157017,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157017, "pid": 5, "tid": 7, "ts": 1716454223898972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843121, "dur": 6,
    "args": {
      "External id": 157017,
      "cbid": 211, "correlation": 157017
    }
  },
  {
    "ph": "s", "id": 157017, "pid": 76337, "tid": -914061504, "ts": 1716454223843121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223898977, "dur": 6,
    "args": {
      "External id": 157027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157027, "pid": 5, "tid": 7, "ts": 1716454223898977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843178, "dur": 12,
    "args": {
      "External id": 157027,
      "cbid": 211, "correlation": 157027
    }
  },
  {
    "ph": "s", "id": 157027, "pid": 76337, "tid": -914061504, "ts": 1716454223843178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223898984, "dur": 9,
    "args": {
      "External id": 157047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157047,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 157047, "pid": 5, "tid": 7, "ts": 1716454223898984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843243, "dur": 10,
    "args": {
      "External id": 157047,
      "cbid": 211, "correlation": 157047
    }
  },
  {
    "ph": "s", "id": 157047, "pid": 76337, "tid": -914061504, "ts": 1716454223843243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223898995, "dur": 4,
    "args": {
      "External id": 157059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157059,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 157059, "pid": 5, "tid": 7, "ts": 1716454223898995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843264, "dur": 6,
    "args": {
      "External id": 157059,
      "cbid": 211, "correlation": 157059
    }
  },
  {
    "ph": "s", "id": 157059, "pid": 76337, "tid": -914061504, "ts": 1716454223843264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223899000, "dur": 7,
    "args": {
      "External id": 157062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157062, "pid": 5, "tid": 7, "ts": 1716454223899000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843283, "dur": 6,
    "args": {
      "External id": 157062,
      "cbid": 211, "correlation": 157062
    }
  },
  {
    "ph": "s", "id": 157062, "pid": 76337, "tid": -914061504, "ts": 1716454223843283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223899008, "dur": 5,
    "args": {
      "External id": 157071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157071,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157071, "pid": 5, "tid": 7, "ts": 1716454223899008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843322, "dur": 10,
    "args": {
      "External id": 157071,
      "cbid": 211, "correlation": 157071
    }
  },
  {
    "ph": "s", "id": 157071, "pid": 76337, "tid": -914061504, "ts": 1716454223843322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843385, "dur": 0,
    "args": {
      "External id": 157081,
      "cbid": 317, "correlation": 157081
    }
  },
  {
    "ph": "f", "id": 157081, "pid": 76337, "tid": -914061504, "ts": 1716454223843385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843385, "dur": 0,
    "args": {
      "External id": 157082,
      "cbid": 203, "correlation": 157082
    }
  },
  {
    "ph": "f", "id": 157082, "pid": 76337, "tid": -914061504, "ts": 1716454223843385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843386, "dur": 0,
    "args": {
      "External id": 157083,
      "cbid": 205, "correlation": 157083
    }
  },
  {
    "ph": "f", "id": 157083, "pid": 76337, "tid": -914061504, "ts": 1716454223843386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899014, "dur": 5,
    "args": {
      "External id": 157087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157087,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157087, "pid": 5, "tid": 7, "ts": 1716454223899014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843400, "dur": 12,
    "args": {
      "External id": 157087,
      "cbid": 211, "correlation": 157087
    }
  },
  {
    "ph": "s", "id": 157087, "pid": 76337, "tid": -914061504, "ts": 1716454223843400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899021, "dur": 162,
    "args": {
      "External id": 157089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157089,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157089, "pid": 5, "tid": 7, "ts": 1716454223899021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843414, "dur": 5,
    "args": {
      "External id": 157089,
      "cbid": 211, "correlation": 157089
    }
  },
  {
    "ph": "s", "id": 157089, "pid": 76337, "tid": -914061504, "ts": 1716454223843414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223899185, "dur": 1,
    "args": {
      "External id": 157091,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157091,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15
    }
  },
  {
    "ph": "f", "id": 157091, "pid": 5, "tid": 7, "ts": 1716454223899185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843425, "dur": 6,
    "args": {
      "External id": 157091,
      "cbid": 51, "correlation": 157091
    }
  },
  {
    "ph": "s", "id": 157091, "pid": 76337, "tid": -914061504, "ts": 1716454223843425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223899188, "dur": 258,
    "args": {
      "External id": 157092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157092,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157092, "pid": 5, "tid": 7, "ts": 1716454223899188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843432, "dur": 6,
    "args": {
      "External id": 157092,
      "cbid": 211, "correlation": 157092
    }
  },
  {
    "ph": "s", "id": 157092, "pid": 76337, "tid": -914061504, "ts": 1716454223843432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899447, "dur": 6,
    "args": {
      "External id": 157094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157094,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157094, "pid": 5, "tid": 7, "ts": 1716454223899447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843442, "dur": 5,
    "args": {
      "External id": 157094,
      "cbid": 211, "correlation": 157094
    }
  },
  {
    "ph": "s", "id": 157094, "pid": 76337, "tid": -914061504, "ts": 1716454223843442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223899455, "dur": 6,
    "args": {
      "External id": 157100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157100,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157100, "pid": 5, "tid": 7, "ts": 1716454223899455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843470, "dur": 10,
    "args": {
      "External id": 157100,
      "cbid": 211, "correlation": 157100
    }
  },
  {
    "ph": "s", "id": 157100, "pid": 76337, "tid": -914061504, "ts": 1716454223843470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223899462, "dur": 5,
    "args": {
      "External id": 157108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157108,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157108, "pid": 5, "tid": 7, "ts": 1716454223899462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843503, "dur": 8,
    "args": {
      "External id": 157108,
      "cbid": 211, "correlation": 157108
    }
  },
  {
    "ph": "s", "id": 157108, "pid": 76337, "tid": -914061504, "ts": 1716454223843503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223899468, "dur": 4,
    "args": {
      "External id": 157116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157116,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157116, "pid": 5, "tid": 7, "ts": 1716454223899468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843532, "dur": 8,
    "args": {
      "External id": 157116,
      "cbid": 211, "correlation": 157116
    }
  },
  {
    "ph": "s", "id": 157116, "pid": 76337, "tid": -914061504, "ts": 1716454223843532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899474, "dur": 10,
    "args": {
      "External id": 157136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157136,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 157136, "pid": 5, "tid": 7, "ts": 1716454223899474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843637, "dur": 13,
    "args": {
      "External id": 157136,
      "cbid": 211, "correlation": 157136
    }
  },
  {
    "ph": "s", "id": 157136, "pid": 76337, "tid": -914061504, "ts": 1716454223843637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223899485, "dur": 4,
    "args": {
      "External id": 157148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157148,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 157148, "pid": 5, "tid": 7, "ts": 1716454223899485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843659, "dur": 6,
    "args": {
      "External id": 157148,
      "cbid": 211, "correlation": 157148
    }
  },
  {
    "ph": "s", "id": 157148, "pid": 76337, "tid": -914061504, "ts": 1716454223843659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223899490, "dur": 7,
    "args": {
      "External id": 157151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157151, "pid": 5, "tid": 7, "ts": 1716454223899490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843678, "dur": 7,
    "args": {
      "External id": 157151,
      "cbid": 211, "correlation": 157151
    }
  },
  {
    "ph": "s", "id": 157151, "pid": 76337, "tid": -914061504, "ts": 1716454223843678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223899498, "dur": 5,
    "args": {
      "External id": 157160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157160,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157160, "pid": 5, "tid": 7, "ts": 1716454223899498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843715, "dur": 10,
    "args": {
      "External id": 157160,
      "cbid": 211, "correlation": 157160
    }
  },
  {
    "ph": "s", "id": 157160, "pid": 76337, "tid": -914061504, "ts": 1716454223843715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843768, "dur": 0,
    "args": {
      "External id": 157170,
      "cbid": 317, "correlation": 157170
    }
  },
  {
    "ph": "f", "id": 157170, "pid": 76337, "tid": -914061504, "ts": 1716454223843768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843769, "dur": 0,
    "args": {
      "External id": 157171,
      "cbid": 203, "correlation": 157171
    }
  },
  {
    "ph": "f", "id": 157171, "pid": 76337, "tid": -914061504, "ts": 1716454223843769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843770, "dur": 0,
    "args": {
      "External id": 157172,
      "cbid": 205, "correlation": 157172
    }
  },
  {
    "ph": "f", "id": 157172, "pid": 76337, "tid": -914061504, "ts": 1716454223843770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899504, "dur": 5,
    "args": {
      "External id": 157176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157176,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157176, "pid": 5, "tid": 7, "ts": 1716454223899504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843783, "dur": 11,
    "args": {
      "External id": 157176,
      "cbid": 211, "correlation": 157176
    }
  },
  {
    "ph": "s", "id": 157176, "pid": 76337, "tid": -914061504, "ts": 1716454223843783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899510, "dur": 162,
    "args": {
      "External id": 157178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157178,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157178, "pid": 5, "tid": 7, "ts": 1716454223899510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843797, "dur": 5,
    "args": {
      "External id": 157178,
      "cbid": 211, "correlation": 157178
    }
  },
  {
    "ph": "s", "id": 157178, "pid": 76337, "tid": -914061504, "ts": 1716454223843797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223899674, "dur": 1,
    "args": {
      "External id": 157180,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157180,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 157180, "pid": 5, "tid": 7, "ts": 1716454223899674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843808, "dur": 6,
    "args": {
      "External id": 157180,
      "cbid": 51, "correlation": 157180
    }
  },
  {
    "ph": "s", "id": 157180, "pid": 76337, "tid": -914061504, "ts": 1716454223843808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223899678, "dur": 258,
    "args": {
      "External id": 157181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157181,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157181, "pid": 5, "tid": 7, "ts": 1716454223899678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843815, "dur": 6,
    "args": {
      "External id": 157181,
      "cbid": 211, "correlation": 157181
    }
  },
  {
    "ph": "s", "id": 157181, "pid": 76337, "tid": -914061504, "ts": 1716454223843815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899937, "dur": 6,
    "args": {
      "External id": 157183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157183,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157183, "pid": 5, "tid": 7, "ts": 1716454223899937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843826, "dur": 5,
    "args": {
      "External id": 157183,
      "cbid": 211, "correlation": 157183
    }
  },
  {
    "ph": "s", "id": 157183, "pid": 76337, "tid": -914061504, "ts": 1716454223843826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223899944, "dur": 6,
    "args": {
      "External id": 157189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157189, "pid": 5, "tid": 7, "ts": 1716454223899944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843854, "dur": 9,
    "args": {
      "External id": 157189,
      "cbid": 211, "correlation": 157189
    }
  },
  {
    "ph": "s", "id": 157189, "pid": 76337, "tid": -914061504, "ts": 1716454223843854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223899952, "dur": 3,
    "args": {
      "External id": 157197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157197,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 157197, "pid": 5, "tid": 7, "ts": 1716454223899952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843898, "dur": 9,
    "args": {
      "External id": 157197,
      "cbid": 211, "correlation": 157197
    }
  },
  {
    "ph": "s", "id": 157197, "pid": 76337, "tid": -914061504, "ts": 1716454223843898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843959, "dur": 1,
    "args": {
      "External id": 157213,
      "cbid": 251, "correlation": 157213
    }
  },
  {
    "ph": "f", "id": 157213, "pid": 76337, "tid": -914061504, "ts": 1716454223843959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843965, "dur": 0,
    "args": {
      "External id": 157215,
      "cbid": 251, "correlation": 157215
    }
  },
  {
    "ph": "f", "id": 157215, "pid": 76337, "tid": -914061504, "ts": 1716454223843965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223899956, "dur": 11,
    "args": {
      "External id": 157216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157216,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157216, "pid": 5, "tid": 7, "ts": 1716454223899956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843967, "dur": 20,
    "args": {
      "External id": 157216,
      "cbid": 211, "correlation": 157216
    }
  },
  {
    "ph": "s", "id": 157216, "pid": 76337, "tid": -914061504, "ts": 1716454223843967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223899968, "dur": 4,
    "args": {
      "External id": 157218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157218,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157218, "pid": 5, "tid": 7, "ts": 1716454223899968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223843989, "dur": 6,
    "args": {
      "External id": 157218,
      "cbid": 211, "correlation": 157218
    }
  },
  {
    "ph": "s", "id": 157218, "pid": 76337, "tid": -914061504, "ts": 1716454223843989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223899973, "dur": 6,
    "args": {
      "External id": 157228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157228,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157228, "pid": 5, "tid": 7, "ts": 1716454223899973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844047, "dur": 12,
    "args": {
      "External id": 157228,
      "cbid": 211, "correlation": 157228
    }
  },
  {
    "ph": "s", "id": 157228, "pid": 76337, "tid": -914061504, "ts": 1716454223844047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223899980, "dur": 10,
    "args": {
      "External id": 157248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157248,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 157248, "pid": 5, "tid": 7, "ts": 1716454223899980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844114, "dur": 12,
    "args": {
      "External id": 157248,
      "cbid": 211, "correlation": 157248
    }
  },
  {
    "ph": "s", "id": 157248, "pid": 76337, "tid": -914061504, "ts": 1716454223844114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223899991, "dur": 4,
    "args": {
      "External id": 157260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157260,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 157260, "pid": 5, "tid": 7, "ts": 1716454223899991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844136, "dur": 6,
    "args": {
      "External id": 157260,
      "cbid": 211, "correlation": 157260
    }
  },
  {
    "ph": "s", "id": 157260, "pid": 76337, "tid": -914061504, "ts": 1716454223844136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223899996, "dur": 7,
    "args": {
      "External id": 157263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157263,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157263, "pid": 5, "tid": 7, "ts": 1716454223899996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844153, "dur": 6,
    "args": {
      "External id": 157263,
      "cbid": 211, "correlation": 157263
    }
  },
  {
    "ph": "s", "id": 157263, "pid": 76337, "tid": -914061504, "ts": 1716454223844153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223900004, "dur": 5,
    "args": {
      "External id": 157272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157272,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157272, "pid": 5, "tid": 7, "ts": 1716454223900004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844194, "dur": 10,
    "args": {
      "External id": 157272,
      "cbid": 211, "correlation": 157272
    }
  },
  {
    "ph": "s", "id": 157272, "pid": 76337, "tid": -914061504, "ts": 1716454223844194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844257, "dur": 0,
    "args": {
      "External id": 157282,
      "cbid": 317, "correlation": 157282
    }
  },
  {
    "ph": "f", "id": 157282, "pid": 76337, "tid": -914061504, "ts": 1716454223844257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844257, "dur": 0,
    "args": {
      "External id": 157283,
      "cbid": 203, "correlation": 157283
    }
  },
  {
    "ph": "f", "id": 157283, "pid": 76337, "tid": -914061504, "ts": 1716454223844257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844258, "dur": 0,
    "args": {
      "External id": 157284,
      "cbid": 205, "correlation": 157284
    }
  },
  {
    "ph": "f", "id": 157284, "pid": 76337, "tid": -914061504, "ts": 1716454223844258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900010, "dur": 5,
    "args": {
      "External id": 157288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157288, "pid": 5, "tid": 7, "ts": 1716454223900010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844273, "dur": 12,
    "args": {
      "External id": 157288,
      "cbid": 211, "correlation": 157288
    }
  },
  {
    "ph": "s", "id": 157288, "pid": 76337, "tid": -914061504, "ts": 1716454223844273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900016, "dur": 162,
    "args": {
      "External id": 157290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157290,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157290, "pid": 5, "tid": 7, "ts": 1716454223900016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844288, "dur": 5,
    "args": {
      "External id": 157290,
      "cbid": 211, "correlation": 157290
    }
  },
  {
    "ph": "s", "id": 157290, "pid": 76337, "tid": -914061504, "ts": 1716454223844288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223900180, "dur": 1,
    "args": {
      "External id": 157292,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157292,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 157292, "pid": 5, "tid": 7, "ts": 1716454223900180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844298, "dur": 6,
    "args": {
      "External id": 157292,
      "cbid": 51, "correlation": 157292
    }
  },
  {
    "ph": "s", "id": 157292, "pid": 76337, "tid": -914061504, "ts": 1716454223844298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223900184, "dur": 258,
    "args": {
      "External id": 157293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157293,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157293, "pid": 5, "tid": 7, "ts": 1716454223900184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844305, "dur": 6,
    "args": {
      "External id": 157293,
      "cbid": 211, "correlation": 157293
    }
  },
  {
    "ph": "s", "id": 157293, "pid": 76337, "tid": -914061504, "ts": 1716454223844305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900443, "dur": 6,
    "args": {
      "External id": 157295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157295,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157295, "pid": 5, "tid": 7, "ts": 1716454223900443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844315, "dur": 5,
    "args": {
      "External id": 157295,
      "cbid": 211, "correlation": 157295
    }
  },
  {
    "ph": "s", "id": 157295, "pid": 76337, "tid": -914061504, "ts": 1716454223844315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223900450, "dur": 6,
    "args": {
      "External id": 157301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157301,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157301, "pid": 5, "tid": 7, "ts": 1716454223900450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844343, "dur": 9,
    "args": {
      "External id": 157301,
      "cbid": 211, "correlation": 157301
    }
  },
  {
    "ph": "s", "id": 157301, "pid": 76337, "tid": -914061504, "ts": 1716454223844343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223900458, "dur": 5,
    "args": {
      "External id": 157309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157309,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157309, "pid": 5, "tid": 7, "ts": 1716454223900458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844376, "dur": 8,
    "args": {
      "External id": 157309,
      "cbid": 211, "correlation": 157309
    }
  },
  {
    "ph": "s", "id": 157309, "pid": 76337, "tid": -914061504, "ts": 1716454223844376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223900464, "dur": 4,
    "args": {
      "External id": 157317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157317,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157317, "pid": 5, "tid": 7, "ts": 1716454223900464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844406, "dur": 8,
    "args": {
      "External id": 157317,
      "cbid": 211, "correlation": 157317
    }
  },
  {
    "ph": "s", "id": 157317, "pid": 76337, "tid": -914061504, "ts": 1716454223844406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900470, "dur": 10,
    "args": {
      "External id": 157337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157337,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 157337, "pid": 5, "tid": 7, "ts": 1716454223900470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844548, "dur": 13,
    "args": {
      "External id": 157337,
      "cbid": 211, "correlation": 157337
    }
  },
  {
    "ph": "s", "id": 157337, "pid": 76337, "tid": -914061504, "ts": 1716454223844548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223900481, "dur": 4,
    "args": {
      "External id": 157349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157349,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 157349, "pid": 5, "tid": 7, "ts": 1716454223900481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844571, "dur": 7,
    "args": {
      "External id": 157349,
      "cbid": 211, "correlation": 157349
    }
  },
  {
    "ph": "s", "id": 157349, "pid": 76337, "tid": -914061504, "ts": 1716454223844571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223900486, "dur": 6,
    "args": {
      "External id": 157352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157352, "pid": 5, "tid": 7, "ts": 1716454223900486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844590, "dur": 7,
    "args": {
      "External id": 157352,
      "cbid": 211, "correlation": 157352
    }
  },
  {
    "ph": "s", "id": 157352, "pid": 76337, "tid": -914061504, "ts": 1716454223844590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844649, "dur": 0,
    "args": {
      "External id": 157363,
      "cbid": 317, "correlation": 157363
    }
  },
  {
    "ph": "f", "id": 157363, "pid": 76337, "tid": -914061504, "ts": 1716454223844649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844650, "dur": 0,
    "args": {
      "External id": 157364,
      "cbid": 203, "correlation": 157364
    }
  },
  {
    "ph": "f", "id": 157364, "pid": 76337, "tid": -914061504, "ts": 1716454223844650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844651, "dur": 0,
    "args": {
      "External id": 157365,
      "cbid": 205, "correlation": 157365
    }
  },
  {
    "ph": "f", "id": 157365, "pid": 76337, "tid": -914061504, "ts": 1716454223844651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900494, "dur": 5,
    "args": {
      "External id": 157369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157369,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157369, "pid": 5, "tid": 7, "ts": 1716454223900494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844669, "dur": 13,
    "args": {
      "External id": 157369,
      "cbid": 211, "correlation": 157369
    }
  },
  {
    "ph": "s", "id": 157369, "pid": 76337, "tid": -914061504, "ts": 1716454223844669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223900500, "dur": 37,
    "args": {
      "External id": 157371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157371,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 157371, "pid": 5, "tid": 7, "ts": 1716454223900500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844688, "dur": 9,
    "args": {
      "External id": 157371,
      "cbid": 211, "correlation": 157371
    }
  },
  {
    "ph": "s", "id": 157371, "pid": 76337, "tid": -914061504, "ts": 1716454223844688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900538, "dur": 5,
    "args": {
      "External id": 157373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157373,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157373, "pid": 5, "tid": 7, "ts": 1716454223900538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844701, "dur": 5,
    "args": {
      "External id": 157373,
      "cbid": 211, "correlation": 157373
    }
  },
  {
    "ph": "s", "id": 157373, "pid": 76337, "tid": -914061504, "ts": 1716454223844701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223900545, "dur": 6,
    "args": {
      "External id": 157379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157379, "pid": 5, "tid": 7, "ts": 1716454223900545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844728, "dur": 9,
    "args": {
      "External id": 157379,
      "cbid": 211, "correlation": 157379
    }
  },
  {
    "ph": "s", "id": 157379, "pid": 76337, "tid": -914061504, "ts": 1716454223844728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223900552, "dur": 20,
    "args": {
      "External id": 157388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157388,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157388, "pid": 5, "tid": 7, "ts": 1716454223900552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844811, "dur": 14,
    "args": {
      "External id": 157388,
      "cbid": 211, "correlation": 157388
    }
  },
  {
    "ph": "s", "id": 157388, "pid": 76337, "tid": -914061504, "ts": 1716454223844811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900574, "dur": 11,
    "args": {
      "External id": 157410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157410,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 157410, "pid": 5, "tid": 7, "ts": 1716454223900574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844867, "dur": 11,
    "args": {
      "External id": 157410,
      "cbid": 211, "correlation": 157410
    }
  },
  {
    "ph": "s", "id": 157410, "pid": 76337, "tid": -914061504, "ts": 1716454223844867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844960, "dur": 2,
    "args": {
      "External id": 157421,
      "cbid": 251, "correlation": 157421
    }
  },
  {
    "ph": "f", "id": 157421, "pid": 76337, "tid": -914061504, "ts": 1716454223844960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844965, "dur": 0,
    "args": {
      "External id": 157422,
      "cbid": 251, "correlation": 157422
    }
  },
  {
    "ph": "f", "id": 157422, "pid": 76337, "tid": -914061504, "ts": 1716454223844965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223900586, "dur": 54,
    "args": {
      "External id": 157423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157423,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 157423, "pid": 5, "tid": 7, "ts": 1716454223900586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223844969, "dur": 22,
    "args": {
      "External id": 157423,
      "cbid": 211, "correlation": 157423
    }
  },
  {
    "ph": "s", "id": 157423, "pid": 76337, "tid": -914061504, "ts": 1716454223844969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845050, "dur": 1,
    "args": {
      "External id": 157434,
      "cbid": 251, "correlation": 157434
    }
  },
  {
    "ph": "f", "id": 157434, "pid": 76337, "tid": -914061504, "ts": 1716454223845050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845054, "dur": 0,
    "args": {
      "External id": 157435,
      "cbid": 251, "correlation": 157435
    }
  },
  {
    "ph": "f", "id": 157435, "pid": 76337, "tid": -914061504, "ts": 1716454223845054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223900641, "dur": 53,
    "args": {
      "External id": 157436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157436,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 157436, "pid": 5, "tid": 7, "ts": 1716454223900641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845056, "dur": 12,
    "args": {
      "External id": 157436,
      "cbid": 211, "correlation": 157436
    }
  },
  {
    "ph": "s", "id": 157436, "pid": 76337, "tid": -914061504, "ts": 1716454223845056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845123, "dur": 1,
    "args": {
      "External id": 157447,
      "cbid": 251, "correlation": 157447
    }
  },
  {
    "ph": "f", "id": 157447, "pid": 76337, "tid": -914061504, "ts": 1716454223845123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845126, "dur": 0,
    "args": {
      "External id": 157448,
      "cbid": 251, "correlation": 157448
    }
  },
  {
    "ph": "f", "id": 157448, "pid": 76337, "tid": -914061504, "ts": 1716454223845126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223900695, "dur": 53,
    "args": {
      "External id": 157449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157449,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 157449, "pid": 5, "tid": 7, "ts": 1716454223900695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845128, "dur": 11,
    "args": {
      "External id": 157449,
      "cbid": 211, "correlation": 157449
    }
  },
  {
    "ph": "s", "id": 157449, "pid": 76337, "tid": -914061504, "ts": 1716454223845128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223900750, "dur": 55,
    "args": {
      "External id": 157474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157474,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157474, "pid": 5, "tid": 7, "ts": 1716454223900750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845213, "dur": 14,
    "args": {
      "External id": 157474,
      "cbid": 211, "correlation": 157474
    }
  },
  {
    "ph": "s", "id": 157474, "pid": 76337, "tid": -914061504, "ts": 1716454223845213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845313, "dur": 1,
    "args": {
      "External id": 157492,
      "cbid": 251, "correlation": 157492
    }
  },
  {
    "ph": "f", "id": 157492, "pid": 76337, "tid": -914061504, "ts": 1716454223845313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223900806, "dur": 63,
    "args": {
      "External id": 157494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157494,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 157494, "pid": 5, "tid": 7, "ts": 1716454223900806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845320, "dur": 14,
    "args": {
      "External id": 157494,
      "cbid": 211, "correlation": 157494
    }
  },
  {
    "ph": "s", "id": 157494, "pid": 76337, "tid": -914061504, "ts": 1716454223845320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223900871, "dur": 6,
    "args": {
      "External id": 157502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157502,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157502, "pid": 5, "tid": 7, "ts": 1716454223900871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845390, "dur": 12,
    "args": {
      "External id": 157502,
      "cbid": 211, "correlation": 157502
    }
  },
  {
    "ph": "s", "id": 157502, "pid": 76337, "tid": -914061504, "ts": 1716454223845390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223900878, "dur": 7,
    "args": {
      "External id": 157510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157510,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157510, "pid": 5, "tid": 7, "ts": 1716454223900878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845428, "dur": 8,
    "args": {
      "External id": 157510,
      "cbid": 211, "correlation": 157510
    }
  },
  {
    "ph": "s", "id": 157510, "pid": 76337, "tid": -914061504, "ts": 1716454223845428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223900886, "dur": 8,
    "args": {
      "External id": 157521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157521, "pid": 5, "tid": 7, "ts": 1716454223900886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845500, "dur": 13,
    "args": {
      "External id": 157521,
      "cbid": 211, "correlation": 157521
    }
  },
  {
    "ph": "s", "id": 157521, "pid": 76337, "tid": -914061504, "ts": 1716454223845500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223900895, "dur": 9,
    "args": {
      "External id": 157543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157543,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 157543, "pid": 5, "tid": 7, "ts": 1716454223900895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845533, "dur": 7,
    "args": {
      "External id": 157543,
      "cbid": 211, "correlation": 157543
    }
  },
  {
    "ph": "s", "id": 157543, "pid": 76337, "tid": -914061504, "ts": 1716454223845533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845618, "dur": 3,
    "args": {
      "External id": 157554,
      "cbid": 251, "correlation": 157554
    }
  },
  {
    "ph": "f", "id": 157554, "pid": 76337, "tid": -914061504, "ts": 1716454223845618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223900906, "dur": 1,
    "args": {
      "External id": 157555,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157555,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 157555, "pid": 5, "tid": 7, "ts": 1716454223900906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845625, "dur": 10,
    "args": {
      "External id": 157555,
      "cbid": 51, "correlation": 157555
    }
  },
  {
    "ph": "s", "id": 157555, "pid": 76337, "tid": -914061504, "ts": 1716454223845625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223900910, "dur": 36,
    "args": {
      "External id": 157556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157556,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 157556, "pid": 5, "tid": 7, "ts": 1716454223900910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845636, "dur": 12,
    "args": {
      "External id": 157556,
      "cbid": 211, "correlation": 157556
    }
  },
  {
    "ph": "s", "id": 157556, "pid": 76337, "tid": -914061504, "ts": 1716454223845636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845707, "dur": 1,
    "args": {
      "External id": 157567,
      "cbid": 251, "correlation": 157567
    }
  },
  {
    "ph": "f", "id": 157567, "pid": 76337, "tid": -914061504, "ts": 1716454223845707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845711, "dur": 0,
    "args": {
      "External id": 157568,
      "cbid": 251, "correlation": 157568
    }
  },
  {
    "ph": "f", "id": 157568, "pid": 76337, "tid": -914061504, "ts": 1716454223845711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223900947, "dur": 12,
    "args": {
      "External id": 157569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157569,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157569, "pid": 5, "tid": 7, "ts": 1716454223900947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845713, "dur": 13,
    "args": {
      "External id": 157569,
      "cbid": 211, "correlation": 157569
    }
  },
  {
    "ph": "s", "id": 157569, "pid": 76337, "tid": -914061504, "ts": 1716454223845713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223900960, "dur": 5,
    "args": {
      "External id": 157571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157571,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157571, "pid": 5, "tid": 7, "ts": 1716454223900960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845727, "dur": 6,
    "args": {
      "External id": 157571,
      "cbid": 211, "correlation": 157571
    }
  },
  {
    "ph": "s", "id": 157571, "pid": 76337, "tid": -914061504, "ts": 1716454223845727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845783, "dur": 1,
    "args": {
      "External id": 157582,
      "cbid": 251, "correlation": 157582
    }
  },
  {
    "ph": "f", "id": 157582, "pid": 76337, "tid": -914061504, "ts": 1716454223845783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845787, "dur": 0,
    "args": {
      "External id": 157583,
      "cbid": 251, "correlation": 157583
    }
  },
  {
    "ph": "f", "id": 157583, "pid": 76337, "tid": -914061504, "ts": 1716454223845787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223900967, "dur": 8,
    "args": {
      "External id": 157584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157584,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157584, "pid": 5, "tid": 7, "ts": 1716454223900967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845788, "dur": 11,
    "args": {
      "External id": 157584,
      "cbid": 211, "correlation": 157584
    }
  },
  {
    "ph": "s", "id": 157584, "pid": 76337, "tid": -914061504, "ts": 1716454223845788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223900976, "dur": 4,
    "args": {
      "External id": 157586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157586,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157586, "pid": 5, "tid": 7, "ts": 1716454223900976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845801, "dur": 5,
    "args": {
      "External id": 157586,
      "cbid": 211, "correlation": 157586
    }
  },
  {
    "ph": "s", "id": 157586, "pid": 76337, "tid": -914061504, "ts": 1716454223845801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223900981, "dur": 20,
    "args": {
      "External id": 157611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157611,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 157611, "pid": 5, "tid": 7, "ts": 1716454223900981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845880, "dur": 12,
    "args": {
      "External id": 157611,
      "cbid": 211, "correlation": 157611
    }
  },
  {
    "ph": "s", "id": 157611, "pid": 76337, "tid": -914061504, "ts": 1716454223845880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845987, "dur": 2,
    "args": {
      "External id": 157629,
      "cbid": 251, "correlation": 157629
    }
  },
  {
    "ph": "f", "id": 157629, "pid": 76337, "tid": -914061504, "ts": 1716454223845987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223901003, "dur": 1,
    "args": {
      "External id": 157631,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157631,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 157631, "pid": 5, "tid": 7, "ts": 1716454223901003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223845993, "dur": 10,
    "args": {
      "External id": 157631,
      "cbid": 51, "correlation": 157631
    }
  },
  {
    "ph": "s", "id": 157631, "pid": 76337, "tid": -914061504, "ts": 1716454223845993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223901007, "dur": 36,
    "args": {
      "External id": 157632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157632,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 157632, "pid": 5, "tid": 7, "ts": 1716454223901007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846004, "dur": 13,
    "args": {
      "External id": 157632,
      "cbid": 211, "correlation": 157632
    }
  },
  {
    "ph": "s", "id": 157632, "pid": 76337, "tid": -914061504, "ts": 1716454223846004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223901045, "dur": 4,
    "args": {
      "External id": 157640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157640,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157640, "pid": 5, "tid": 7, "ts": 1716454223901045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846077, "dur": 12,
    "args": {
      "External id": 157640,
      "cbid": 211, "correlation": 157640
    }
  },
  {
    "ph": "s", "id": 157640, "pid": 76337, "tid": -914061504, "ts": 1716454223846077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901050, "dur": 8,
    "args": {
      "External id": 157648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157648, "pid": 5, "tid": 7, "ts": 1716454223901050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846118, "dur": 10,
    "args": {
      "External id": 157648,
      "cbid": 211, "correlation": 157648
    }
  },
  {
    "ph": "s", "id": 157648, "pid": 76337, "tid": -914061504, "ts": 1716454223846118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901060, "dur": 8,
    "args": {
      "External id": 157670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157670,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 157670, "pid": 5, "tid": 7, "ts": 1716454223901060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846170, "dur": 10,
    "args": {
      "External id": 157670,
      "cbid": 211, "correlation": 157670
    }
  },
  {
    "ph": "s", "id": 157670, "pid": 76337, "tid": -914061504, "ts": 1716454223846170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846261, "dur": 1,
    "args": {
      "External id": 157686,
      "cbid": 251, "correlation": 157686
    }
  },
  {
    "ph": "f", "id": 157686, "pid": 76337, "tid": -914061504, "ts": 1716454223846261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846266, "dur": 0,
    "args": {
      "External id": 157688,
      "cbid": 251, "correlation": 157688
    }
  },
  {
    "ph": "f", "id": 157688, "pid": 76337, "tid": -914061504, "ts": 1716454223846266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223901069, "dur": 190,
    "args": {
      "External id": 157689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157689,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157689, "pid": 5, "tid": 7, "ts": 1716454223901069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846268, "dur": 14,
    "args": {
      "External id": 157689,
      "cbid": 211, "correlation": 157689
    }
  },
  {
    "ph": "s", "id": 157689, "pid": 76337, "tid": -914061504, "ts": 1716454223846268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901260, "dur": 21,
    "args": {
      "External id": 157697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157697,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157697, "pid": 5, "tid": 7, "ts": 1716454223901260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846335, "dur": 13,
    "args": {
      "External id": 157697,
      "cbid": 211, "correlation": 157697
    }
  },
  {
    "ph": "s", "id": 157697, "pid": 76337, "tid": -914061504, "ts": 1716454223846335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901282, "dur": 22,
    "args": {
      "External id": 157705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157705, "pid": 5, "tid": 7, "ts": 1716454223901282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846366, "dur": 9,
    "args": {
      "External id": 157705,
      "cbid": 211, "correlation": 157705
    }
  },
  {
    "ph": "s", "id": 157705, "pid": 76337, "tid": -914061504, "ts": 1716454223846366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846449, "dur": 1,
    "args": {
      "External id": 157721,
      "cbid": 251, "correlation": 157721
    }
  },
  {
    "ph": "f", "id": 157721, "pid": 76337, "tid": -914061504, "ts": 1716454223846449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223901306, "dur": 1,
    "args": {
      "External id": 157723,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157723,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 157723, "pid": 5, "tid": 7, "ts": 1716454223901306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846454, "dur": 8,
    "args": {
      "External id": 157723,
      "cbid": 51, "correlation": 157723
    }
  },
  {
    "ph": "s", "id": 157723, "pid": 76337, "tid": -914061504, "ts": 1716454223846454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223901310, "dur": 108,
    "args": {
      "External id": 157724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157724,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 157724, "pid": 5, "tid": 7, "ts": 1716454223901310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846464, "dur": 13,
    "args": {
      "External id": 157724,
      "cbid": 211, "correlation": 157724
    }
  },
  {
    "ph": "s", "id": 157724, "pid": 76337, "tid": -914061504, "ts": 1716454223846464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223901420, "dur": 5,
    "args": {
      "External id": 157732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157732,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157732, "pid": 5, "tid": 7, "ts": 1716454223901420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846506, "dur": 10,
    "args": {
      "External id": 157732,
      "cbid": 211, "correlation": 157732
    }
  },
  {
    "ph": "s", "id": 157732, "pid": 76337, "tid": -914061504, "ts": 1716454223846506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901426, "dur": 10,
    "args": {
      "External id": 157743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157743, "pid": 5, "tid": 7, "ts": 1716454223901426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846577, "dur": 12,
    "args": {
      "External id": 157743,
      "cbid": 211, "correlation": 157743
    }
  },
  {
    "ph": "s", "id": 157743, "pid": 76337, "tid": -914061504, "ts": 1716454223846577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846640, "dur": 0,
    "args": {
      "External id": 157755,
      "cbid": 317, "correlation": 157755
    }
  },
  {
    "ph": "f", "id": 157755, "pid": 76337, "tid": -914061504, "ts": 1716454223846640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846641, "dur": 0,
    "args": {
      "External id": 157756,
      "cbid": 203, "correlation": 157756
    }
  },
  {
    "ph": "f", "id": 157756, "pid": 76337, "tid": -914061504, "ts": 1716454223846641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846642, "dur": 0,
    "args": {
      "External id": 157757,
      "cbid": 205, "correlation": 157757
    }
  },
  {
    "ph": "f", "id": 157757, "pid": 76337, "tid": -914061504, "ts": 1716454223846642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901437, "dur": 6,
    "args": {
      "External id": 157761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157761,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157761, "pid": 5, "tid": 7, "ts": 1716454223901437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846658, "dur": 13,
    "args": {
      "External id": 157761,
      "cbid": 211, "correlation": 157761
    }
  },
  {
    "ph": "s", "id": 157761, "pid": 76337, "tid": -914061504, "ts": 1716454223846658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223901444, "dur": 38,
    "args": {
      "External id": 157763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157763,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 157763, "pid": 5, "tid": 7, "ts": 1716454223901444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846677, "dur": 7,
    "args": {
      "External id": 157763,
      "cbid": 211, "correlation": 157763
    }
  },
  {
    "ph": "s", "id": 157763, "pid": 76337, "tid": -914061504, "ts": 1716454223846677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901483, "dur": 6,
    "args": {
      "External id": 157765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157765,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157765, "pid": 5, "tid": 7, "ts": 1716454223901483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846688, "dur": 5,
    "args": {
      "External id": 157765,
      "cbid": 211, "correlation": 157765
    }
  },
  {
    "ph": "s", "id": 157765, "pid": 76337, "tid": -914061504, "ts": 1716454223846688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901490, "dur": 7,
    "args": {
      "External id": 157771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157771, "pid": 5, "tid": 7, "ts": 1716454223901490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846715, "dur": 10,
    "args": {
      "External id": 157771,
      "cbid": 211, "correlation": 157771
    }
  },
  {
    "ph": "s", "id": 157771, "pid": 76337, "tid": -914061504, "ts": 1716454223846715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223901499, "dur": 5,
    "args": {
      "External id": 157779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157779,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157779, "pid": 5, "tid": 7, "ts": 1716454223901499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846747, "dur": 8,
    "args": {
      "External id": 157779,
      "cbid": 211, "correlation": 157779
    }
  },
  {
    "ph": "s", "id": 157779, "pid": 76337, "tid": -914061504, "ts": 1716454223846747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901505, "dur": 11,
    "args": {
      "External id": 157799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157799,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 157799, "pid": 5, "tid": 7, "ts": 1716454223901505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846820, "dur": 12,
    "args": {
      "External id": 157799,
      "cbid": 211, "correlation": 157799
    }
  },
  {
    "ph": "s", "id": 157799, "pid": 76337, "tid": -914061504, "ts": 1716454223846820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223901518, "dur": 4,
    "args": {
      "External id": 157811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157811,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 157811, "pid": 5, "tid": 7, "ts": 1716454223901518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846842, "dur": 6,
    "args": {
      "External id": 157811,
      "cbid": 211, "correlation": 157811
    }
  },
  {
    "ph": "s", "id": 157811, "pid": 76337, "tid": -914061504, "ts": 1716454223846842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901524, "dur": 9,
    "args": {
      "External id": 157814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157814,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157814, "pid": 5, "tid": 7, "ts": 1716454223901524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846860, "dur": 7,
    "args": {
      "External id": 157814,
      "cbid": 211, "correlation": 157814
    }
  },
  {
    "ph": "s", "id": 157814, "pid": 76337, "tid": -914061504, "ts": 1716454223846860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223901534, "dur": 5,
    "args": {
      "External id": 157823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157823,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157823, "pid": 5, "tid": 7, "ts": 1716454223901534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846898, "dur": 10,
    "args": {
      "External id": 157823,
      "cbid": 211, "correlation": 157823
    }
  },
  {
    "ph": "s", "id": 157823, "pid": 76337, "tid": -914061504, "ts": 1716454223846898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846949, "dur": 0,
    "args": {
      "External id": 157833,
      "cbid": 317, "correlation": 157833
    }
  },
  {
    "ph": "f", "id": 157833, "pid": 76337, "tid": -914061504, "ts": 1716454223846949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846950, "dur": 0,
    "args": {
      "External id": 157834,
      "cbid": 203, "correlation": 157834
    }
  },
  {
    "ph": "f", "id": 157834, "pid": 76337, "tid": -914061504, "ts": 1716454223846950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846951, "dur": 0,
    "args": {
      "External id": 157835,
      "cbid": 205, "correlation": 157835
    }
  },
  {
    "ph": "f", "id": 157835, "pid": 76337, "tid": -914061504, "ts": 1716454223846951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901541, "dur": 5,
    "args": {
      "External id": 157839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157839,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157839, "pid": 5, "tid": 7, "ts": 1716454223901541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846964, "dur": 20,
    "args": {
      "External id": 157839,
      "cbid": 211, "correlation": 157839
    }
  },
  {
    "ph": "s", "id": 157839, "pid": 76337, "tid": -914061504, "ts": 1716454223846964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901547, "dur": 162,
    "args": {
      "External id": 157841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157841,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157841, "pid": 5, "tid": 7, "ts": 1716454223901547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846987, "dur": 5,
    "args": {
      "External id": 157841,
      "cbid": 211, "correlation": 157841
    }
  },
  {
    "ph": "s", "id": 157841, "pid": 76337, "tid": -914061504, "ts": 1716454223846987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223901711, "dur": 1,
    "args": {
      "External id": 157843,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157843,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 157843, "pid": 5, "tid": 7, "ts": 1716454223901711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223846998, "dur": 7,
    "args": {
      "External id": 157843,
      "cbid": 51, "correlation": 157843
    }
  },
  {
    "ph": "s", "id": 157843, "pid": 76337, "tid": -914061504, "ts": 1716454223846998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223901715, "dur": 269,
    "args": {
      "External id": 157844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157844,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157844, "pid": 5, "tid": 7, "ts": 1716454223901715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847007, "dur": 6,
    "args": {
      "External id": 157844,
      "cbid": 211, "correlation": 157844
    }
  },
  {
    "ph": "s", "id": 157844, "pid": 76337, "tid": -914061504, "ts": 1716454223847007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223901985, "dur": 6,
    "args": {
      "External id": 157846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157846,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157846, "pid": 5, "tid": 7, "ts": 1716454223901985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847017, "dur": 6,
    "args": {
      "External id": 157846,
      "cbid": 211, "correlation": 157846
    }
  },
  {
    "ph": "s", "id": 157846, "pid": 76337, "tid": -914061504, "ts": 1716454223847017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223901992, "dur": 6,
    "args": {
      "External id": 157852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157852, "pid": 5, "tid": 7, "ts": 1716454223901992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847046, "dur": 8,
    "args": {
      "External id": 157852,
      "cbid": 211, "correlation": 157852
    }
  },
  {
    "ph": "s", "id": 157852, "pid": 76337, "tid": -914061504, "ts": 1716454223847046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223901999, "dur": 3,
    "args": {
      "External id": 157860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157860,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 157860, "pid": 5, "tid": 7, "ts": 1716454223901999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847091, "dur": 9,
    "args": {
      "External id": 157860,
      "cbid": 211, "correlation": 157860
    }
  },
  {
    "ph": "s", "id": 157860, "pid": 76337, "tid": -914061504, "ts": 1716454223847091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847156, "dur": 1,
    "args": {
      "External id": 157876,
      "cbid": 251, "correlation": 157876
    }
  },
  {
    "ph": "f", "id": 157876, "pid": 76337, "tid": -914061504, "ts": 1716454223847156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847161, "dur": 0,
    "args": {
      "External id": 157878,
      "cbid": 251, "correlation": 157878
    }
  },
  {
    "ph": "f", "id": 157878, "pid": 76337, "tid": -914061504, "ts": 1716454223847161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223902004, "dur": 12,
    "args": {
      "External id": 157879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157879,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157879, "pid": 5, "tid": 7, "ts": 1716454223902004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847163, "dur": 12,
    "args": {
      "External id": 157879,
      "cbid": 211, "correlation": 157879
    }
  },
  {
    "ph": "s", "id": 157879, "pid": 76337, "tid": -914061504, "ts": 1716454223847163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223902017, "dur": 5,
    "args": {
      "External id": 157881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157881,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157881, "pid": 5, "tid": 7, "ts": 1716454223902017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847177, "dur": 6,
    "args": {
      "External id": 157881,
      "cbid": 211, "correlation": 157881
    }
  },
  {
    "ph": "s", "id": 157881, "pid": 76337, "tid": -914061504, "ts": 1716454223847177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223902024, "dur": 6,
    "args": {
      "External id": 157891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157891, "pid": 5, "tid": 7, "ts": 1716454223902024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847234, "dur": 12,
    "args": {
      "External id": 157891,
      "cbid": 211, "correlation": 157891
    }
  },
  {
    "ph": "s", "id": 157891, "pid": 76337, "tid": -914061504, "ts": 1716454223847234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902031, "dur": 10,
    "args": {
      "External id": 157911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157911,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 157911, "pid": 5, "tid": 7, "ts": 1716454223902031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847299, "dur": 11,
    "args": {
      "External id": 157911,
      "cbid": 211, "correlation": 157911
    }
  },
  {
    "ph": "s", "id": 157911, "pid": 76337, "tid": -914061504, "ts": 1716454223847299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223902042, "dur": 4,
    "args": {
      "External id": 157923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157923,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 157923, "pid": 5, "tid": 7, "ts": 1716454223902042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847320, "dur": 7,
    "args": {
      "External id": 157923,
      "cbid": 211, "correlation": 157923
    }
  },
  {
    "ph": "s", "id": 157923, "pid": 76337, "tid": -914061504, "ts": 1716454223847320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223902047, "dur": 7,
    "args": {
      "External id": 157926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157926, "pid": 5, "tid": 7, "ts": 1716454223902047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847339, "dur": 6,
    "args": {
      "External id": 157926,
      "cbid": 211, "correlation": 157926
    }
  },
  {
    "ph": "s", "id": 157926, "pid": 76337, "tid": -914061504, "ts": 1716454223847339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223902055, "dur": 5,
    "args": {
      "External id": 157935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157935,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157935, "pid": 5, "tid": 7, "ts": 1716454223902055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847379, "dur": 10,
    "args": {
      "External id": 157935,
      "cbid": 211, "correlation": 157935
    }
  },
  {
    "ph": "s", "id": 157935, "pid": 76337, "tid": -914061504, "ts": 1716454223847379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847442, "dur": 0,
    "args": {
      "External id": 157945,
      "cbid": 317, "correlation": 157945
    }
  },
  {
    "ph": "f", "id": 157945, "pid": 76337, "tid": -914061504, "ts": 1716454223847442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847443, "dur": 0,
    "args": {
      "External id": 157946,
      "cbid": 203, "correlation": 157946
    }
  },
  {
    "ph": "f", "id": 157946, "pid": 76337, "tid": -914061504, "ts": 1716454223847443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847444, "dur": 0,
    "args": {
      "External id": 157947,
      "cbid": 205, "correlation": 157947
    }
  },
  {
    "ph": "f", "id": 157947, "pid": 76337, "tid": -914061504, "ts": 1716454223847444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902061, "dur": 5,
    "args": {
      "External id": 157951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157951,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157951, "pid": 5, "tid": 7, "ts": 1716454223902061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847457, "dur": 12,
    "args": {
      "External id": 157951,
      "cbid": 211, "correlation": 157951
    }
  },
  {
    "ph": "s", "id": 157951, "pid": 76337, "tid": -914061504, "ts": 1716454223847457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902067, "dur": 161,
    "args": {
      "External id": 157953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157953,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157953, "pid": 5, "tid": 7, "ts": 1716454223902067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847472, "dur": 5,
    "args": {
      "External id": 157953,
      "cbid": 211, "correlation": 157953
    }
  },
  {
    "ph": "s", "id": 157953, "pid": 76337, "tid": -914061504, "ts": 1716454223847472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223902230, "dur": 1,
    "args": {
      "External id": 157955,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 157955,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 157955, "pid": 5, "tid": 7, "ts": 1716454223902230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847483, "dur": 6,
    "args": {
      "External id": 157955,
      "cbid": 51, "correlation": 157955
    }
  },
  {
    "ph": "s", "id": 157955, "pid": 76337, "tid": -914061504, "ts": 1716454223847483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223902234, "dur": 258,
    "args": {
      "External id": 157956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157956,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 157956, "pid": 5, "tid": 7, "ts": 1716454223902234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847490, "dur": 6,
    "args": {
      "External id": 157956,
      "cbid": 211, "correlation": 157956
    }
  },
  {
    "ph": "s", "id": 157956, "pid": 76337, "tid": -914061504, "ts": 1716454223847490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902494, "dur": 6,
    "args": {
      "External id": 157958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157958,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157958, "pid": 5, "tid": 7, "ts": 1716454223902494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847499, "dur": 5,
    "args": {
      "External id": 157958,
      "cbid": 211, "correlation": 157958
    }
  },
  {
    "ph": "s", "id": 157958, "pid": 76337, "tid": -914061504, "ts": 1716454223847499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223902501, "dur": 6,
    "args": {
      "External id": 157964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157964,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157964, "pid": 5, "tid": 7, "ts": 1716454223902501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847528, "dur": 8,
    "args": {
      "External id": 157964,
      "cbid": 211, "correlation": 157964
    }
  },
  {
    "ph": "s", "id": 157964, "pid": 76337, "tid": -914061504, "ts": 1716454223847528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223902509, "dur": 5,
    "args": {
      "External id": 157972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157972,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157972, "pid": 5, "tid": 7, "ts": 1716454223902509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847560, "dur": 9,
    "args": {
      "External id": 157972,
      "cbid": 211, "correlation": 157972
    }
  },
  {
    "ph": "s", "id": 157972, "pid": 76337, "tid": -914061504, "ts": 1716454223847560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223902515, "dur": 4,
    "args": {
      "External id": 157980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157980,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 157980, "pid": 5, "tid": 7, "ts": 1716454223902515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847590, "dur": 8,
    "args": {
      "External id": 157980,
      "cbid": 211, "correlation": 157980
    }
  },
  {
    "ph": "s", "id": 157980, "pid": 76337, "tid": -914061504, "ts": 1716454223847590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223902521, "dur": 11,
    "args": {
      "External id": 157989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 157989,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 157989, "pid": 5, "tid": 7, "ts": 1716454223902521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847678, "dur": 13,
    "args": {
      "External id": 157989,
      "cbid": 211, "correlation": 157989
    }
  },
  {
    "ph": "s", "id": 157989, "pid": 76337, "tid": -914061504, "ts": 1716454223847678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902533, "dur": 12,
    "args": {
      "External id": 158009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158009,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158009, "pid": 5, "tid": 7, "ts": 1716454223902533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847746, "dur": 11,
    "args": {
      "External id": 158009,
      "cbid": 211, "correlation": 158009
    }
  },
  {
    "ph": "s", "id": 158009, "pid": 76337, "tid": -914061504, "ts": 1716454223847746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223902547, "dur": 4,
    "args": {
      "External id": 158021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158021,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158021, "pid": 5, "tid": 7, "ts": 1716454223902547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847767, "dur": 7,
    "args": {
      "External id": 158021,
      "cbid": 211, "correlation": 158021
    }
  },
  {
    "ph": "s", "id": 158021, "pid": 76337, "tid": -914061504, "ts": 1716454223847767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223902552, "dur": 10,
    "args": {
      "External id": 158024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158024,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158024, "pid": 5, "tid": 7, "ts": 1716454223902552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847786, "dur": 6,
    "args": {
      "External id": 158024,
      "cbid": 211, "correlation": 158024
    }
  },
  {
    "ph": "s", "id": 158024, "pid": 76337, "tid": -914061504, "ts": 1716454223847786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223902563, "dur": 6,
    "args": {
      "External id": 158033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158033,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158033, "pid": 5, "tid": 7, "ts": 1716454223902563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847824, "dur": 10,
    "args": {
      "External id": 158033,
      "cbid": 211, "correlation": 158033
    }
  },
  {
    "ph": "s", "id": 158033, "pid": 76337, "tid": -914061504, "ts": 1716454223847824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847877, "dur": 0,
    "args": {
      "External id": 158043,
      "cbid": 317, "correlation": 158043
    }
  },
  {
    "ph": "f", "id": 158043, "pid": 76337, "tid": -914061504, "ts": 1716454223847877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847878, "dur": 0,
    "args": {
      "External id": 158044,
      "cbid": 203, "correlation": 158044
    }
  },
  {
    "ph": "f", "id": 158044, "pid": 76337, "tid": -914061504, "ts": 1716454223847878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847879, "dur": 0,
    "args": {
      "External id": 158045,
      "cbid": 205, "correlation": 158045
    }
  },
  {
    "ph": "f", "id": 158045, "pid": 76337, "tid": -914061504, "ts": 1716454223847879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902571, "dur": 7,
    "args": {
      "External id": 158049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158049,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158049, "pid": 5, "tid": 7, "ts": 1716454223902571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847894, "dur": 11,
    "args": {
      "External id": 158049,
      "cbid": 211, "correlation": 158049
    }
  },
  {
    "ph": "s", "id": 158049, "pid": 76337, "tid": -914061504, "ts": 1716454223847894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223902579, "dur": 319,
    "args": {
      "External id": 158051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158051,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158051, "pid": 5, "tid": 7, "ts": 1716454223902579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847908, "dur": 5,
    "args": {
      "External id": 158051,
      "cbid": 211, "correlation": 158051
    }
  },
  {
    "ph": "s", "id": 158051, "pid": 76337, "tid": -914061504, "ts": 1716454223847908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223902900, "dur": 1,
    "args": {
      "External id": 158053,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158053,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 158053, "pid": 5, "tid": 7, "ts": 1716454223902900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847919, "dur": 8,
    "args": {
      "External id": 158053,
      "cbid": 51, "correlation": 158053
    }
  },
  {
    "ph": "s", "id": 158053, "pid": 76337, "tid": -914061504, "ts": 1716454223847919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223902903, "dur": 494,
    "args": {
      "External id": 158054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158054,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158054, "pid": 5, "tid": 7, "ts": 1716454223902903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847928, "dur": 6,
    "args": {
      "External id": 158054,
      "cbid": 211, "correlation": 158054
    }
  },
  {
    "ph": "s", "id": 158054, "pid": 76337, "tid": -914061504, "ts": 1716454223847928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223903398, "dur": 5,
    "args": {
      "External id": 158056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158056, "pid": 5, "tid": 7, "ts": 1716454223903398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847938, "dur": 5,
    "args": {
      "External id": 158056,
      "cbid": 211, "correlation": 158056
    }
  },
  {
    "ph": "s", "id": 158056, "pid": 76337, "tid": -914061504, "ts": 1716454223847938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223903405, "dur": 6,
    "args": {
      "External id": 158062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158062, "pid": 5, "tid": 7, "ts": 1716454223903405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223847966, "dur": 18,
    "args": {
      "External id": 158062,
      "cbid": 211, "correlation": 158062
    }
  },
  {
    "ph": "s", "id": 158062, "pid": 76337, "tid": -914061504, "ts": 1716454223847966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223903413, "dur": 3,
    "args": {
      "External id": 158070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158070,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 158070, "pid": 5, "tid": 7, "ts": 1716454223903413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848019, "dur": 11,
    "args": {
      "External id": 158070,
      "cbid": 211, "correlation": 158070
    }
  },
  {
    "ph": "s", "id": 158070, "pid": 76337, "tid": -914061504, "ts": 1716454223848019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848084, "dur": 1,
    "args": {
      "External id": 158086,
      "cbid": 251, "correlation": 158086
    }
  },
  {
    "ph": "f", "id": 158086, "pid": 76337, "tid": -914061504, "ts": 1716454223848084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848089, "dur": 0,
    "args": {
      "External id": 158088,
      "cbid": 251, "correlation": 158088
    }
  },
  {
    "ph": "f", "id": 158088, "pid": 76337, "tid": -914061504, "ts": 1716454223848089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223903417, "dur": 11,
    "args": {
      "External id": 158089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158089,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158089, "pid": 5, "tid": 7, "ts": 1716454223903417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848091, "dur": 11,
    "args": {
      "External id": 158089,
      "cbid": 211, "correlation": 158089
    }
  },
  {
    "ph": "s", "id": 158089, "pid": 76337, "tid": -914061504, "ts": 1716454223848091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223903430, "dur": 5,
    "args": {
      "External id": 158091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158091,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158091, "pid": 5, "tid": 7, "ts": 1716454223903430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848104, "dur": 6,
    "args": {
      "External id": 158091,
      "cbid": 211, "correlation": 158091
    }
  },
  {
    "ph": "s", "id": 158091, "pid": 76337, "tid": -914061504, "ts": 1716454223848104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223903436, "dur": 6,
    "args": {
      "External id": 158101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158101, "pid": 5, "tid": 7, "ts": 1716454223903436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848161, "dur": 13,
    "args": {
      "External id": 158101,
      "cbid": 211, "correlation": 158101
    }
  },
  {
    "ph": "s", "id": 158101, "pid": 76337, "tid": -914061504, "ts": 1716454223848161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223903443, "dur": 9,
    "args": {
      "External id": 158121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158121,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158121, "pid": 5, "tid": 7, "ts": 1716454223903443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848229, "dur": 11,
    "args": {
      "External id": 158121,
      "cbid": 211, "correlation": 158121
    }
  },
  {
    "ph": "s", "id": 158121, "pid": 76337, "tid": -914061504, "ts": 1716454223848229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223903454, "dur": 4,
    "args": {
      "External id": 158133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158133,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 158133, "pid": 5, "tid": 7, "ts": 1716454223903454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848250, "dur": 6,
    "args": {
      "External id": 158133,
      "cbid": 211, "correlation": 158133
    }
  },
  {
    "ph": "s", "id": 158133, "pid": 76337, "tid": -914061504, "ts": 1716454223848250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223903459, "dur": 7,
    "args": {
      "External id": 158136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158136, "pid": 5, "tid": 7, "ts": 1716454223903459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848268, "dur": 7,
    "args": {
      "External id": 158136,
      "cbid": 211, "correlation": 158136
    }
  },
  {
    "ph": "s", "id": 158136, "pid": 76337, "tid": -914061504, "ts": 1716454223848268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223903467, "dur": 5,
    "args": {
      "External id": 158145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158145,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158145, "pid": 5, "tid": 7, "ts": 1716454223903467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848308, "dur": 10,
    "args": {
      "External id": 158145,
      "cbid": 211, "correlation": 158145
    }
  },
  {
    "ph": "s", "id": 158145, "pid": 76337, "tid": -914061504, "ts": 1716454223848308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848371, "dur": 0,
    "args": {
      "External id": 158155,
      "cbid": 317, "correlation": 158155
    }
  },
  {
    "ph": "f", "id": 158155, "pid": 76337, "tid": -914061504, "ts": 1716454223848371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848372, "dur": 0,
    "args": {
      "External id": 158156,
      "cbid": 203, "correlation": 158156
    }
  },
  {
    "ph": "f", "id": 158156, "pid": 76337, "tid": -914061504, "ts": 1716454223848372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848372, "dur": 0,
    "args": {
      "External id": 158157,
      "cbid": 205, "correlation": 158157
    }
  },
  {
    "ph": "f", "id": 158157, "pid": 76337, "tid": -914061504, "ts": 1716454223848372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223903473, "dur": 5,
    "args": {
      "External id": 158161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158161,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158161, "pid": 5, "tid": 7, "ts": 1716454223903473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848387, "dur": 12,
    "args": {
      "External id": 158161,
      "cbid": 211, "correlation": 158161
    }
  },
  {
    "ph": "s", "id": 158161, "pid": 76337, "tid": -914061504, "ts": 1716454223848387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223903480, "dur": 162,
    "args": {
      "External id": 158163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158163,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158163, "pid": 5, "tid": 7, "ts": 1716454223903480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848401, "dur": 5,
    "args": {
      "External id": 158163,
      "cbid": 211, "correlation": 158163
    }
  },
  {
    "ph": "s", "id": 158163, "pid": 76337, "tid": -914061504, "ts": 1716454223848401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223903644, "dur": 1,
    "args": {
      "External id": 158165,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158165,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 158165, "pid": 5, "tid": 7, "ts": 1716454223903644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848412, "dur": 6,
    "args": {
      "External id": 158165,
      "cbid": 51, "correlation": 158165
    }
  },
  {
    "ph": "s", "id": 158165, "pid": 76337, "tid": -914061504, "ts": 1716454223848412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223903648, "dur": 258,
    "args": {
      "External id": 158166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158166,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158166, "pid": 5, "tid": 7, "ts": 1716454223903648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848419, "dur": 6,
    "args": {
      "External id": 158166,
      "cbid": 211, "correlation": 158166
    }
  },
  {
    "ph": "s", "id": 158166, "pid": 76337, "tid": -914061504, "ts": 1716454223848419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223903907, "dur": 6,
    "args": {
      "External id": 158168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158168, "pid": 5, "tid": 7, "ts": 1716454223903907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848430, "dur": 5,
    "args": {
      "External id": 158168,
      "cbid": 211, "correlation": 158168
    }
  },
  {
    "ph": "s", "id": 158168, "pid": 76337, "tid": -914061504, "ts": 1716454223848430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223903915, "dur": 6,
    "args": {
      "External id": 158174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158174, "pid": 5, "tid": 7, "ts": 1716454223903915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848458, "dur": 8,
    "args": {
      "External id": 158174,
      "cbid": 211, "correlation": 158174
    }
  },
  {
    "ph": "s", "id": 158174, "pid": 76337, "tid": -914061504, "ts": 1716454223848458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848516, "dur": 0,
    "args": {
      "External id": 158184,
      "cbid": 317, "correlation": 158184
    }
  },
  {
    "ph": "f", "id": 158184, "pid": 76337, "tid": -914061504, "ts": 1716454223848516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848517, "dur": 0,
    "args": {
      "External id": 158185,
      "cbid": 203, "correlation": 158185
    }
  },
  {
    "ph": "f", "id": 158185, "pid": 76337, "tid": -914061504, "ts": 1716454223848517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848518, "dur": 0,
    "args": {
      "External id": 158186,
      "cbid": 205, "correlation": 158186
    }
  },
  {
    "ph": "f", "id": 158186, "pid": 76337, "tid": -914061504, "ts": 1716454223848518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223903922, "dur": 8,
    "args": {
      "External id": 158190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158190, "pid": 5, "tid": 7, "ts": 1716454223903922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848534, "dur": 11,
    "args": {
      "External id": 158190,
      "cbid": 211, "correlation": 158190
    }
  },
  {
    "ph": "s", "id": 158190, "pid": 76337, "tid": -914061504, "ts": 1716454223848534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223903931, "dur": 3,
    "args": {
      "External id": 158192,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158192,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4150943396226414
    }
  },
  {
    "ph": "f", "id": 158192, "pid": 5, "tid": 7, "ts": 1716454223903931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848551, "dur": 14,
    "args": {
      "External id": 158192,
      "cbid": 51, "correlation": 158192
    }
  },
  {
    "ph": "s", "id": 158192, "pid": 76337, "tid": -914061504, "ts": 1716454223848551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223903936, "dur": 96,
    "args": {
      "External id": 158193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158193,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 158193, "pid": 5, "tid": 7, "ts": 1716454223903936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848567, "dur": 7,
    "args": {
      "External id": 158193,
      "cbid": 211, "correlation": 158193
    }
  },
  {
    "ph": "s", "id": 158193, "pid": 76337, "tid": -914061504, "ts": 1716454223848567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223904033, "dur": 5,
    "args": {
      "External id": 158195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158195,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158195, "pid": 5, "tid": 7, "ts": 1716454223904033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848578, "dur": 5,
    "args": {
      "External id": 158195,
      "cbid": 211, "correlation": 158195
    }
  },
  {
    "ph": "s", "id": 158195, "pid": 76337, "tid": -914061504, "ts": 1716454223848578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223904040, "dur": 6,
    "args": {
      "External id": 158201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158201,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158201, "pid": 5, "tid": 7, "ts": 1716454223904040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848605, "dur": 8,
    "args": {
      "External id": 158201,
      "cbid": 211, "correlation": 158201
    }
  },
  {
    "ph": "s", "id": 158201, "pid": 76337, "tid": -914061504, "ts": 1716454223848605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223904047, "dur": 5,
    "args": {
      "External id": 158209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158209,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158209, "pid": 5, "tid": 7, "ts": 1716454223904047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848635, "dur": 7,
    "args": {
      "External id": 158209,
      "cbid": 211, "correlation": 158209
    }
  },
  {
    "ph": "s", "id": 158209, "pid": 76337, "tid": -914061504, "ts": 1716454223848635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223904053, "dur": 4,
    "args": {
      "External id": 158217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158217,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158217, "pid": 5, "tid": 7, "ts": 1716454223904053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848663, "dur": 8,
    "args": {
      "External id": 158217,
      "cbid": 211, "correlation": 158217
    }
  },
  {
    "ph": "s", "id": 158217, "pid": 76337, "tid": -914061504, "ts": 1716454223848663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223904059, "dur": 11,
    "args": {
      "External id": 158226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158226,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158226, "pid": 5, "tid": 7, "ts": 1716454223904059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848707, "dur": 10,
    "args": {
      "External id": 158226,
      "cbid": 211, "correlation": 158226
    }
  },
  {
    "ph": "s", "id": 158226, "pid": 76337, "tid": -914061504, "ts": 1716454223848707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223904072, "dur": 12,
    "args": {
      "External id": 158246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158246,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158246, "pid": 5, "tid": 7, "ts": 1716454223904072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848775, "dur": 12,
    "args": {
      "External id": 158246,
      "cbid": 211, "correlation": 158246
    }
  },
  {
    "ph": "s", "id": 158246, "pid": 76337, "tid": -914061504, "ts": 1716454223848775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223904085, "dur": 4,
    "args": {
      "External id": 158258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158258,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158258, "pid": 5, "tid": 7, "ts": 1716454223904085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848796, "dur": 6,
    "args": {
      "External id": 158258,
      "cbid": 211, "correlation": 158258
    }
  },
  {
    "ph": "s", "id": 158258, "pid": 76337, "tid": -914061504, "ts": 1716454223848796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223904090, "dur": 10,
    "args": {
      "External id": 158261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158261,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158261, "pid": 5, "tid": 7, "ts": 1716454223904090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848814, "dur": 6,
    "args": {
      "External id": 158261,
      "cbid": 211, "correlation": 158261
    }
  },
  {
    "ph": "s", "id": 158261, "pid": 76337, "tid": -914061504, "ts": 1716454223848814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223904102, "dur": 7,
    "args": {
      "External id": 158270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158270,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158270, "pid": 5, "tid": 7, "ts": 1716454223904102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848852, "dur": 9,
    "args": {
      "External id": 158270,
      "cbid": 211, "correlation": 158270
    }
  },
  {
    "ph": "s", "id": 158270, "pid": 76337, "tid": -914061504, "ts": 1716454223848852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848904, "dur": 0,
    "args": {
      "External id": 158280,
      "cbid": 317, "correlation": 158280
    }
  },
  {
    "ph": "f", "id": 158280, "pid": 76337, "tid": -914061504, "ts": 1716454223848904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848905, "dur": 0,
    "args": {
      "External id": 158281,
      "cbid": 203, "correlation": 158281
    }
  },
  {
    "ph": "f", "id": 158281, "pid": 76337, "tid": -914061504, "ts": 1716454223848905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848906, "dur": 0,
    "args": {
      "External id": 158282,
      "cbid": 205, "correlation": 158282
    }
  },
  {
    "ph": "f", "id": 158282, "pid": 76337, "tid": -914061504, "ts": 1716454223848906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223904110, "dur": 7,
    "args": {
      "External id": 158286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158286,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158286, "pid": 5, "tid": 7, "ts": 1716454223904110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848920, "dur": 12,
    "args": {
      "External id": 158286,
      "cbid": 211, "correlation": 158286
    }
  },
  {
    "ph": "s", "id": 158286, "pid": 76337, "tid": -914061504, "ts": 1716454223848920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223904118, "dur": 319,
    "args": {
      "External id": 158288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158288, "pid": 5, "tid": 7, "ts": 1716454223904118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848934, "dur": 5,
    "args": {
      "External id": 158288,
      "cbid": 211, "correlation": 158288
    }
  },
  {
    "ph": "s", "id": 158288, "pid": 76337, "tid": -914061504, "ts": 1716454223848934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223904440, "dur": 1,
    "args": {
      "External id": 158290,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158290,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 158290, "pid": 5, "tid": 7, "ts": 1716454223904440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848945, "dur": 6,
    "args": {
      "External id": 158290,
      "cbid": 51, "correlation": 158290
    }
  },
  {
    "ph": "s", "id": 158290, "pid": 76337, "tid": -914061504, "ts": 1716454223848945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223904443, "dur": 495,
    "args": {
      "External id": 158291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158291,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158291, "pid": 5, "tid": 7, "ts": 1716454223904443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848952, "dur": 6,
    "args": {
      "External id": 158291,
      "cbid": 211, "correlation": 158291
    }
  },
  {
    "ph": "s", "id": 158291, "pid": 76337, "tid": -914061504, "ts": 1716454223848952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223904940, "dur": 6,
    "args": {
      "External id": 158293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158293,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158293, "pid": 5, "tid": 7, "ts": 1716454223904940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848961, "dur": 5,
    "args": {
      "External id": 158293,
      "cbid": 211, "correlation": 158293
    }
  },
  {
    "ph": "s", "id": 158293, "pid": 76337, "tid": -914061504, "ts": 1716454223848961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223904947, "dur": 6,
    "args": {
      "External id": 158299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158299, "pid": 5, "tid": 7, "ts": 1716454223904947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223848997, "dur": 9,
    "args": {
      "External id": 158299,
      "cbid": 211, "correlation": 158299
    }
  },
  {
    "ph": "s", "id": 158299, "pid": 76337, "tid": -914061504, "ts": 1716454223848997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223904955, "dur": 3,
    "args": {
      "External id": 158307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158307,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 158307, "pid": 5, "tid": 7, "ts": 1716454223904955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849042, "dur": 9,
    "args": {
      "External id": 158307,
      "cbid": 211, "correlation": 158307
    }
  },
  {
    "ph": "s", "id": 158307, "pid": 76337, "tid": -914061504, "ts": 1716454223849042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849104, "dur": 1,
    "args": {
      "External id": 158323,
      "cbid": 251, "correlation": 158323
    }
  },
  {
    "ph": "f", "id": 158323, "pid": 76337, "tid": -914061504, "ts": 1716454223849104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849108, "dur": 0,
    "args": {
      "External id": 158325,
      "cbid": 251, "correlation": 158325
    }
  },
  {
    "ph": "f", "id": 158325, "pid": 76337, "tid": -914061504, "ts": 1716454223849108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223904959, "dur": 12,
    "args": {
      "External id": 158326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158326,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158326, "pid": 5, "tid": 7, "ts": 1716454223904959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849110, "dur": 11,
    "args": {
      "External id": 158326,
      "cbid": 211, "correlation": 158326
    }
  },
  {
    "ph": "s", "id": 158326, "pid": 76337, "tid": -914061504, "ts": 1716454223849110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223904972, "dur": 5,
    "args": {
      "External id": 158328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158328,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158328, "pid": 5, "tid": 7, "ts": 1716454223904972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849123, "dur": 6,
    "args": {
      "External id": 158328,
      "cbid": 211, "correlation": 158328
    }
  },
  {
    "ph": "s", "id": 158328, "pid": 76337, "tid": -914061504, "ts": 1716454223849123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223904979, "dur": 6,
    "args": {
      "External id": 158338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158338,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158338, "pid": 5, "tid": 7, "ts": 1716454223904979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849181, "dur": 12,
    "args": {
      "External id": 158338,
      "cbid": 211, "correlation": 158338
    }
  },
  {
    "ph": "s", "id": 158338, "pid": 76337, "tid": -914061504, "ts": 1716454223849181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223904986, "dur": 9,
    "args": {
      "External id": 158358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158358,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158358, "pid": 5, "tid": 7, "ts": 1716454223904986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849246, "dur": 11,
    "args": {
      "External id": 158358,
      "cbid": 211, "correlation": 158358
    }
  },
  {
    "ph": "s", "id": 158358, "pid": 76337, "tid": -914061504, "ts": 1716454223849246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223904997, "dur": 4,
    "args": {
      "External id": 158370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158370,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 158370, "pid": 5, "tid": 7, "ts": 1716454223904997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849267, "dur": 6,
    "args": {
      "External id": 158370,
      "cbid": 211, "correlation": 158370
    }
  },
  {
    "ph": "s", "id": 158370, "pid": 76337, "tid": -914061504, "ts": 1716454223849267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223905002, "dur": 7,
    "args": {
      "External id": 158373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158373, "pid": 5, "tid": 7, "ts": 1716454223905002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849287, "dur": 6,
    "args": {
      "External id": 158373,
      "cbid": 211, "correlation": 158373
    }
  },
  {
    "ph": "s", "id": 158373, "pid": 76337, "tid": -914061504, "ts": 1716454223849287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223905010, "dur": 5,
    "args": {
      "External id": 158382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158382,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158382, "pid": 5, "tid": 7, "ts": 1716454223905010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849328, "dur": 10,
    "args": {
      "External id": 158382,
      "cbid": 211, "correlation": 158382
    }
  },
  {
    "ph": "s", "id": 158382, "pid": 76337, "tid": -914061504, "ts": 1716454223849328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849392, "dur": 0,
    "args": {
      "External id": 158392,
      "cbid": 317, "correlation": 158392
    }
  },
  {
    "ph": "f", "id": 158392, "pid": 76337, "tid": -914061504, "ts": 1716454223849392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849393, "dur": 0,
    "args": {
      "External id": 158393,
      "cbid": 203, "correlation": 158393
    }
  },
  {
    "ph": "f", "id": 158393, "pid": 76337, "tid": -914061504, "ts": 1716454223849393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849394, "dur": 0,
    "args": {
      "External id": 158394,
      "cbid": 205, "correlation": 158394
    }
  },
  {
    "ph": "f", "id": 158394, "pid": 76337, "tid": -914061504, "ts": 1716454223849394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905016, "dur": 5,
    "args": {
      "External id": 158398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158398,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158398, "pid": 5, "tid": 7, "ts": 1716454223905016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849409, "dur": 12,
    "args": {
      "External id": 158398,
      "cbid": 211, "correlation": 158398
    }
  },
  {
    "ph": "s", "id": 158398, "pid": 76337, "tid": -914061504, "ts": 1716454223849409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905022, "dur": 162,
    "args": {
      "External id": 158400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158400,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158400, "pid": 5, "tid": 7, "ts": 1716454223905022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849425, "dur": 6,
    "args": {
      "External id": 158400,
      "cbid": 211, "correlation": 158400
    }
  },
  {
    "ph": "s", "id": 158400, "pid": 76337, "tid": -914061504, "ts": 1716454223849425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223905186, "dur": 1,
    "args": {
      "External id": 158402,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158402,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 158402, "pid": 5, "tid": 7, "ts": 1716454223905186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849436, "dur": 7,
    "args": {
      "External id": 158402,
      "cbid": 51, "correlation": 158402
    }
  },
  {
    "ph": "s", "id": 158402, "pid": 76337, "tid": -914061504, "ts": 1716454223849436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223905190, "dur": 258,
    "args": {
      "External id": 158403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158403,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158403, "pid": 5, "tid": 7, "ts": 1716454223905190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849445, "dur": 6,
    "args": {
      "External id": 158403,
      "cbid": 211, "correlation": 158403
    }
  },
  {
    "ph": "s", "id": 158403, "pid": 76337, "tid": -914061504, "ts": 1716454223849445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905449, "dur": 6,
    "args": {
      "External id": 158405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158405,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158405, "pid": 5, "tid": 7, "ts": 1716454223905449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849456, "dur": 5,
    "args": {
      "External id": 158405,
      "cbid": 211, "correlation": 158405
    }
  },
  {
    "ph": "s", "id": 158405, "pid": 76337, "tid": -914061504, "ts": 1716454223849456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223905456, "dur": 6,
    "args": {
      "External id": 158411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158411, "pid": 5, "tid": 7, "ts": 1716454223905456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849486, "dur": 9,
    "args": {
      "External id": 158411,
      "cbid": 211, "correlation": 158411
    }
  },
  {
    "ph": "s", "id": 158411, "pid": 76337, "tid": -914061504, "ts": 1716454223849486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849545, "dur": 0,
    "args": {
      "External id": 158421,
      "cbid": 317, "correlation": 158421
    }
  },
  {
    "ph": "f", "id": 158421, "pid": 76337, "tid": -914061504, "ts": 1716454223849545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849546, "dur": 0,
    "args": {
      "External id": 158422,
      "cbid": 203, "correlation": 158422
    }
  },
  {
    "ph": "f", "id": 158422, "pid": 76337, "tid": -914061504, "ts": 1716454223849546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849546, "dur": 0,
    "args": {
      "External id": 158423,
      "cbid": 205, "correlation": 158423
    }
  },
  {
    "ph": "f", "id": 158423, "pid": 76337, "tid": -914061504, "ts": 1716454223849546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905464, "dur": 8,
    "args": {
      "External id": 158427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158427,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158427, "pid": 5, "tid": 7, "ts": 1716454223905464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849559, "dur": 12,
    "args": {
      "External id": 158427,
      "cbid": 211, "correlation": 158427
    }
  },
  {
    "ph": "s", "id": 158427, "pid": 76337, "tid": -914061504, "ts": 1716454223849559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223905473, "dur": 3,
    "args": {
      "External id": 158429,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158429,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 158429, "pid": 5, "tid": 7, "ts": 1716454223905473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849576, "dur": 11,
    "args": {
      "External id": 158429,
      "cbid": 51, "correlation": 158429
    }
  },
  {
    "ph": "s", "id": 158429, "pid": 76337, "tid": -914061504, "ts": 1716454223849576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223905477, "dur": 96,
    "args": {
      "External id": 158430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158430,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 158430, "pid": 5, "tid": 7, "ts": 1716454223905477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849589, "dur": 7,
    "args": {
      "External id": 158430,
      "cbid": 211, "correlation": 158430
    }
  },
  {
    "ph": "s", "id": 158430, "pid": 76337, "tid": -914061504, "ts": 1716454223849589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905575, "dur": 6,
    "args": {
      "External id": 158432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158432,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158432, "pid": 5, "tid": 7, "ts": 1716454223905575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849599, "dur": 5,
    "args": {
      "External id": 158432,
      "cbid": 211, "correlation": 158432
    }
  },
  {
    "ph": "s", "id": 158432, "pid": 76337, "tid": -914061504, "ts": 1716454223849599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223905582, "dur": 6,
    "args": {
      "External id": 158438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158438,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158438, "pid": 5, "tid": 7, "ts": 1716454223905582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849626, "dur": 9,
    "args": {
      "External id": 158438,
      "cbid": 211, "correlation": 158438
    }
  },
  {
    "ph": "s", "id": 158438, "pid": 76337, "tid": -914061504, "ts": 1716454223849626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223905589, "dur": 5,
    "args": {
      "External id": 158446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158446,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158446, "pid": 5, "tid": 7, "ts": 1716454223905589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849656, "dur": 8,
    "args": {
      "External id": 158446,
      "cbid": 211, "correlation": 158446
    }
  },
  {
    "ph": "s", "id": 158446, "pid": 76337, "tid": -914061504, "ts": 1716454223849656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223905595, "dur": 4,
    "args": {
      "External id": 158454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158454,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158454, "pid": 5, "tid": 7, "ts": 1716454223905595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849688, "dur": 8,
    "args": {
      "External id": 158454,
      "cbid": 211, "correlation": 158454
    }
  },
  {
    "ph": "s", "id": 158454, "pid": 76337, "tid": -914061504, "ts": 1716454223849688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223905601, "dur": 11,
    "args": {
      "External id": 158463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158463,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158463, "pid": 5, "tid": 7, "ts": 1716454223905601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849738, "dur": 11,
    "args": {
      "External id": 158463,
      "cbid": 211, "correlation": 158463
    }
  },
  {
    "ph": "s", "id": 158463, "pid": 76337, "tid": -914061504, "ts": 1716454223849738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905614, "dur": 12,
    "args": {
      "External id": 158483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158483,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158483, "pid": 5, "tid": 7, "ts": 1716454223905614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849814, "dur": 12,
    "args": {
      "External id": 158483,
      "cbid": 211, "correlation": 158483
    }
  },
  {
    "ph": "s", "id": 158483, "pid": 76337, "tid": -914061504, "ts": 1716454223849814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223905628, "dur": 4,
    "args": {
      "External id": 158495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158495,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158495, "pid": 5, "tid": 7, "ts": 1716454223905628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849836, "dur": 6,
    "args": {
      "External id": 158495,
      "cbid": 211, "correlation": 158495
    }
  },
  {
    "ph": "s", "id": 158495, "pid": 76337, "tid": -914061504, "ts": 1716454223849836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223905633, "dur": 11,
    "args": {
      "External id": 158498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158498, "pid": 5, "tid": 7, "ts": 1716454223905633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849855, "dur": 7,
    "args": {
      "External id": 158498,
      "cbid": 211, "correlation": 158498
    }
  },
  {
    "ph": "s", "id": 158498, "pid": 76337, "tid": -914061504, "ts": 1716454223849855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223905645, "dur": 6,
    "args": {
      "External id": 158507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158507,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158507, "pid": 5, "tid": 7, "ts": 1716454223905645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849895, "dur": 10,
    "args": {
      "External id": 158507,
      "cbid": 211, "correlation": 158507
    }
  },
  {
    "ph": "s", "id": 158507, "pid": 76337, "tid": -914061504, "ts": 1716454223849895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849947, "dur": 0,
    "args": {
      "External id": 158517,
      "cbid": 317, "correlation": 158517
    }
  },
  {
    "ph": "f", "id": 158517, "pid": 76337, "tid": -914061504, "ts": 1716454223849947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849948, "dur": 0,
    "args": {
      "External id": 158518,
      "cbid": 203, "correlation": 158518
    }
  },
  {
    "ph": "f", "id": 158518, "pid": 76337, "tid": -914061504, "ts": 1716454223849948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849949, "dur": 0,
    "args": {
      "External id": 158519,
      "cbid": 205, "correlation": 158519
    }
  },
  {
    "ph": "f", "id": 158519, "pid": 76337, "tid": -914061504, "ts": 1716454223849949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905653, "dur": 7,
    "args": {
      "External id": 158523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158523,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158523, "pid": 5, "tid": 7, "ts": 1716454223905653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849963, "dur": 22,
    "args": {
      "External id": 158523,
      "cbid": 211, "correlation": 158523
    }
  },
  {
    "ph": "s", "id": 158523, "pid": 76337, "tid": -914061504, "ts": 1716454223849963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223905661, "dur": 319,
    "args": {
      "External id": 158525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158525,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158525, "pid": 5, "tid": 7, "ts": 1716454223905661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849987, "dur": 5,
    "args": {
      "External id": 158525,
      "cbid": 211, "correlation": 158525
    }
  },
  {
    "ph": "s", "id": 158525, "pid": 76337, "tid": -914061504, "ts": 1716454223849987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223905981, "dur": 1,
    "args": {
      "External id": 158527,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158527,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 158527, "pid": 5, "tid": 7, "ts": 1716454223905981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223849998, "dur": 7,
    "args": {
      "External id": 158527,
      "cbid": 51, "correlation": 158527
    }
  },
  {
    "ph": "s", "id": 158527, "pid": 76337, "tid": -914061504, "ts": 1716454223849998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223905985, "dur": 496,
    "args": {
      "External id": 158528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158528,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158528, "pid": 5, "tid": 7, "ts": 1716454223905985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850006, "dur": 6,
    "args": {
      "External id": 158528,
      "cbid": 211, "correlation": 158528
    }
  },
  {
    "ph": "s", "id": 158528, "pid": 76337, "tid": -914061504, "ts": 1716454223850006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223906482, "dur": 5,
    "args": {
      "External id": 158530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158530,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158530, "pid": 5, "tid": 7, "ts": 1716454223906482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850016, "dur": 5,
    "args": {
      "External id": 158530,
      "cbid": 211, "correlation": 158530
    }
  },
  {
    "ph": "s", "id": 158530, "pid": 76337, "tid": -914061504, "ts": 1716454223850016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223906489, "dur": 6,
    "args": {
      "External id": 158536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158536, "pid": 5, "tid": 7, "ts": 1716454223906489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850046, "dur": 9,
    "args": {
      "External id": 158536,
      "cbid": 211, "correlation": 158536
    }
  },
  {
    "ph": "s", "id": 158536, "pid": 76337, "tid": -914061504, "ts": 1716454223850046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223906496, "dur": 3,
    "args": {
      "External id": 158544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158544,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 158544, "pid": 5, "tid": 7, "ts": 1716454223906496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850090, "dur": 9,
    "args": {
      "External id": 158544,
      "cbid": 211, "correlation": 158544
    }
  },
  {
    "ph": "s", "id": 158544, "pid": 76337, "tid": -914061504, "ts": 1716454223850090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850167, "dur": 1,
    "args": {
      "External id": 158560,
      "cbid": 251, "correlation": 158560
    }
  },
  {
    "ph": "f", "id": 158560, "pid": 76337, "tid": -914061504, "ts": 1716454223850167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850174, "dur": 0,
    "args": {
      "External id": 158562,
      "cbid": 251, "correlation": 158562
    }
  },
  {
    "ph": "f", "id": 158562, "pid": 76337, "tid": -914061504, "ts": 1716454223850174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223906500, "dur": 12,
    "args": {
      "External id": 158563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158563,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158563, "pid": 5, "tid": 7, "ts": 1716454223906500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850176, "dur": 12,
    "args": {
      "External id": 158563,
      "cbid": 211, "correlation": 158563
    }
  },
  {
    "ph": "s", "id": 158563, "pid": 76337, "tid": -914061504, "ts": 1716454223850176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223906514, "dur": 5,
    "args": {
      "External id": 158565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158565,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158565, "pid": 5, "tid": 7, "ts": 1716454223906514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850191, "dur": 6,
    "args": {
      "External id": 158565,
      "cbid": 211, "correlation": 158565
    }
  },
  {
    "ph": "s", "id": 158565, "pid": 76337, "tid": -914061504, "ts": 1716454223850191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223906521, "dur": 6,
    "args": {
      "External id": 158575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158575, "pid": 5, "tid": 7, "ts": 1716454223906521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850249, "dur": 12,
    "args": {
      "External id": 158575,
      "cbid": 211, "correlation": 158575
    }
  },
  {
    "ph": "s", "id": 158575, "pid": 76337, "tid": -914061504, "ts": 1716454223850249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223906528, "dur": 9,
    "args": {
      "External id": 158595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158595,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158595, "pid": 5, "tid": 7, "ts": 1716454223906528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850315, "dur": 11,
    "args": {
      "External id": 158595,
      "cbid": 211, "correlation": 158595
    }
  },
  {
    "ph": "s", "id": 158595, "pid": 76337, "tid": -914061504, "ts": 1716454223850315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223906538, "dur": 4,
    "args": {
      "External id": 158607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158607,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 158607, "pid": 5, "tid": 7, "ts": 1716454223906538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850336, "dur": 6,
    "args": {
      "External id": 158607,
      "cbid": 211, "correlation": 158607
    }
  },
  {
    "ph": "s", "id": 158607, "pid": 76337, "tid": -914061504, "ts": 1716454223850336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223906543, "dur": 7,
    "args": {
      "External id": 158610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158610,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158610, "pid": 5, "tid": 7, "ts": 1716454223906543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850355, "dur": 6,
    "args": {
      "External id": 158610,
      "cbid": 211, "correlation": 158610
    }
  },
  {
    "ph": "s", "id": 158610, "pid": 76337, "tid": -914061504, "ts": 1716454223850355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223906552, "dur": 5,
    "args": {
      "External id": 158619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158619,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158619, "pid": 5, "tid": 7, "ts": 1716454223906552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850396, "dur": 10,
    "args": {
      "External id": 158619,
      "cbid": 211, "correlation": 158619
    }
  },
  {
    "ph": "s", "id": 158619, "pid": 76337, "tid": -914061504, "ts": 1716454223850396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850459, "dur": 0,
    "args": {
      "External id": 158629,
      "cbid": 317, "correlation": 158629
    }
  },
  {
    "ph": "f", "id": 158629, "pid": 76337, "tid": -914061504, "ts": 1716454223850459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850459, "dur": 0,
    "args": {
      "External id": 158630,
      "cbid": 203, "correlation": 158630
    }
  },
  {
    "ph": "f", "id": 158630, "pid": 76337, "tid": -914061504, "ts": 1716454223850459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850460, "dur": 0,
    "args": {
      "External id": 158631,
      "cbid": 205, "correlation": 158631
    }
  },
  {
    "ph": "f", "id": 158631, "pid": 76337, "tid": -914061504, "ts": 1716454223850460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223906558, "dur": 5,
    "args": {
      "External id": 158635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158635,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158635, "pid": 5, "tid": 7, "ts": 1716454223906558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850475, "dur": 12,
    "args": {
      "External id": 158635,
      "cbid": 211, "correlation": 158635
    }
  },
  {
    "ph": "s", "id": 158635, "pid": 76337, "tid": -914061504, "ts": 1716454223850475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223906564, "dur": 162,
    "args": {
      "External id": 158637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158637,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158637, "pid": 5, "tid": 7, "ts": 1716454223906564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850489, "dur": 5,
    "args": {
      "External id": 158637,
      "cbid": 211, "correlation": 158637
    }
  },
  {
    "ph": "s", "id": 158637, "pid": 76337, "tid": -914061504, "ts": 1716454223850489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223906729, "dur": 1,
    "args": {
      "External id": 158639,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158639,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 158639, "pid": 5, "tid": 7, "ts": 1716454223906729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850500, "dur": 6,
    "args": {
      "External id": 158639,
      "cbid": 51, "correlation": 158639
    }
  },
  {
    "ph": "s", "id": 158639, "pid": 76337, "tid": -914061504, "ts": 1716454223850500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223906732, "dur": 258,
    "args": {
      "External id": 158640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158640,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158640, "pid": 5, "tid": 7, "ts": 1716454223906732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850507, "dur": 6,
    "args": {
      "External id": 158640,
      "cbid": 211, "correlation": 158640
    }
  },
  {
    "ph": "s", "id": 158640, "pid": 76337, "tid": -914061504, "ts": 1716454223850507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223906992, "dur": 6,
    "args": {
      "External id": 158642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158642,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158642, "pid": 5, "tid": 7, "ts": 1716454223906992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850518, "dur": 5,
    "args": {
      "External id": 158642,
      "cbid": 211, "correlation": 158642
    }
  },
  {
    "ph": "s", "id": 158642, "pid": 76337, "tid": -914061504, "ts": 1716454223850518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223906999, "dur": 6,
    "args": {
      "External id": 158648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158648, "pid": 5, "tid": 7, "ts": 1716454223906999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850547, "dur": 8,
    "args": {
      "External id": 158648,
      "cbid": 211, "correlation": 158648
    }
  },
  {
    "ph": "s", "id": 158648, "pid": 76337, "tid": -914061504, "ts": 1716454223850547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850605, "dur": 0,
    "args": {
      "External id": 158658,
      "cbid": 317, "correlation": 158658
    }
  },
  {
    "ph": "f", "id": 158658, "pid": 76337, "tid": -914061504, "ts": 1716454223850605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850606, "dur": 0,
    "args": {
      "External id": 158659,
      "cbid": 203, "correlation": 158659
    }
  },
  {
    "ph": "f", "id": 158659, "pid": 76337, "tid": -914061504, "ts": 1716454223850606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850606, "dur": 0,
    "args": {
      "External id": 158660,
      "cbid": 205, "correlation": 158660
    }
  },
  {
    "ph": "f", "id": 158660, "pid": 76337, "tid": -914061504, "ts": 1716454223850606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223907007, "dur": 8,
    "args": {
      "External id": 158664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158664,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158664, "pid": 5, "tid": 7, "ts": 1716454223907007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850618, "dur": 13,
    "args": {
      "External id": 158664,
      "cbid": 211, "correlation": 158664
    }
  },
  {
    "ph": "s", "id": 158664, "pid": 76337, "tid": -914061504, "ts": 1716454223850618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223907016, "dur": 3,
    "args": {
      "External id": 158666,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158666,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5151515151515151
    }
  },
  {
    "ph": "f", "id": 158666, "pid": 5, "tid": 7, "ts": 1716454223907016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850636, "dur": 9,
    "args": {
      "External id": 158666,
      "cbid": 51, "correlation": 158666
    }
  },
  {
    "ph": "s", "id": 158666, "pid": 76337, "tid": -914061504, "ts": 1716454223850636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223907020, "dur": 94,
    "args": {
      "External id": 158667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158667,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 158667, "pid": 5, "tid": 7, "ts": 1716454223907020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850646, "dur": 6,
    "args": {
      "External id": 158667,
      "cbid": 211, "correlation": 158667
    }
  },
  {
    "ph": "s", "id": 158667, "pid": 76337, "tid": -914061504, "ts": 1716454223850646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223907115, "dur": 5,
    "args": {
      "External id": 158669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158669,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158669, "pid": 5, "tid": 7, "ts": 1716454223907115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850655, "dur": 5,
    "args": {
      "External id": 158669,
      "cbid": 211, "correlation": 158669
    }
  },
  {
    "ph": "s", "id": 158669, "pid": 76337, "tid": -914061504, "ts": 1716454223850655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223907122, "dur": 6,
    "args": {
      "External id": 158675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158675, "pid": 5, "tid": 7, "ts": 1716454223907122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850683, "dur": 8,
    "args": {
      "External id": 158675,
      "cbid": 211, "correlation": 158675
    }
  },
  {
    "ph": "s", "id": 158675, "pid": 76337, "tid": -914061504, "ts": 1716454223850683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223907129, "dur": 5,
    "args": {
      "External id": 158683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158683,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158683, "pid": 5, "tid": 7, "ts": 1716454223907129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850712, "dur": 8,
    "args": {
      "External id": 158683,
      "cbid": 211, "correlation": 158683
    }
  },
  {
    "ph": "s", "id": 158683, "pid": 76337, "tid": -914061504, "ts": 1716454223850712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223907135, "dur": 4,
    "args": {
      "External id": 158691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158691,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 158691, "pid": 5, "tid": 7, "ts": 1716454223907135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850741, "dur": 8,
    "args": {
      "External id": 158691,
      "cbid": 211, "correlation": 158691
    }
  },
  {
    "ph": "s", "id": 158691, "pid": 76337, "tid": -914061504, "ts": 1716454223850741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223907141, "dur": 14,
    "args": {
      "External id": 158702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158702,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158702, "pid": 5, "tid": 7, "ts": 1716454223907141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850827, "dur": 12,
    "args": {
      "External id": 158702,
      "cbid": 211, "correlation": 158702
    }
  },
  {
    "ph": "s", "id": 158702, "pid": 76337, "tid": -914061504, "ts": 1716454223850827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850883, "dur": 0,
    "args": {
      "External id": 158712,
      "cbid": 317, "correlation": 158712
    }
  },
  {
    "ph": "f", "id": 158712, "pid": 76337, "tid": -914061504, "ts": 1716454223850883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850884, "dur": 0,
    "args": {
      "External id": 158713,
      "cbid": 203, "correlation": 158713
    }
  },
  {
    "ph": "f", "id": 158713, "pid": 76337, "tid": -914061504, "ts": 1716454223850884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850885, "dur": 0,
    "args": {
      "External id": 158714,
      "cbid": 205, "correlation": 158714
    }
  },
  {
    "ph": "f", "id": 158714, "pid": 76337, "tid": -914061504, "ts": 1716454223850885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223907157, "dur": 8,
    "args": {
      "External id": 158718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158718,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158718, "pid": 5, "tid": 7, "ts": 1716454223907157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850899, "dur": 12,
    "args": {
      "External id": 158718,
      "cbid": 211, "correlation": 158718
    }
  },
  {
    "ph": "s", "id": 158718, "pid": 76337, "tid": -914061504, "ts": 1716454223850899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223907166, "dur": 163,
    "args": {
      "External id": 158720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158720,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158720, "pid": 5, "tid": 7, "ts": 1716454223907166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850913, "dur": 5,
    "args": {
      "External id": 158720,
      "cbid": 211, "correlation": 158720
    }
  },
  {
    "ph": "s", "id": 158720, "pid": 76337, "tid": -914061504, "ts": 1716454223850913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223907331, "dur": 1,
    "args": {
      "External id": 158722,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158722,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 158722, "pid": 5, "tid": 7, "ts": 1716454223907331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850924, "dur": 7,
    "args": {
      "External id": 158722,
      "cbid": 51, "correlation": 158722
    }
  },
  {
    "ph": "s", "id": 158722, "pid": 76337, "tid": -914061504, "ts": 1716454223850924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223907335, "dur": 648,
    "args": {
      "External id": 158723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158723,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158723, "pid": 5, "tid": 7, "ts": 1716454223907335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850932, "dur": 6,
    "args": {
      "External id": 158723,
      "cbid": 211, "correlation": 158723
    }
  },
  {
    "ph": "s", "id": 158723, "pid": 76337, "tid": -914061504, "ts": 1716454223850932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223907984, "dur": 12,
    "args": {
      "External id": 158725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158725,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158725, "pid": 5, "tid": 7, "ts": 1716454223907984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850941, "dur": 5,
    "args": {
      "External id": 158725,
      "cbid": 211, "correlation": 158725
    }
  },
  {
    "ph": "s", "id": 158725, "pid": 76337, "tid": -914061504, "ts": 1716454223850941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223907997, "dur": 15,
    "args": {
      "External id": 158731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158731, "pid": 5, "tid": 7, "ts": 1716454223907997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223850970, "dur": 20,
    "args": {
      "External id": 158731,
      "cbid": 211, "correlation": 158731
    }
  },
  {
    "ph": "s", "id": 158731, "pid": 76337, "tid": -914061504, "ts": 1716454223850970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223908014, "dur": 32,
    "args": {
      "External id": 158740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158740,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158740, "pid": 5, "tid": 7, "ts": 1716454223908014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851081, "dur": 13,
    "args": {
      "External id": 158740,
      "cbid": 211, "correlation": 158740
    }
  },
  {
    "ph": "s", "id": 158740, "pid": 76337, "tid": -914061504, "ts": 1716454223851081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223908047, "dur": 31,
    "args": {
      "External id": 158760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158760,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158760, "pid": 5, "tid": 7, "ts": 1716454223908047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851150, "dur": 11,
    "args": {
      "External id": 158760,
      "cbid": 211, "correlation": 158760
    }
  },
  {
    "ph": "s", "id": 158760, "pid": 76337, "tid": -914061504, "ts": 1716454223851150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223908079, "dur": 5,
    "args": {
      "External id": 158772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158772,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158772, "pid": 5, "tid": 7, "ts": 1716454223908079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851170, "dur": 6,
    "args": {
      "External id": 158772,
      "cbid": 211, "correlation": 158772
    }
  },
  {
    "ph": "s", "id": 158772, "pid": 76337, "tid": -914061504, "ts": 1716454223851170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223908085, "dur": 30,
    "args": {
      "External id": 158775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158775, "pid": 5, "tid": 7, "ts": 1716454223908085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851189, "dur": 7,
    "args": {
      "External id": 158775,
      "cbid": 211, "correlation": 158775
    }
  },
  {
    "ph": "s", "id": 158775, "pid": 76337, "tid": -914061504, "ts": 1716454223851189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223908117, "dur": 22,
    "args": {
      "External id": 158784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158784,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158784, "pid": 5, "tid": 7, "ts": 1716454223908117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851229, "dur": 10,
    "args": {
      "External id": 158784,
      "cbid": 211, "correlation": 158784
    }
  },
  {
    "ph": "s", "id": 158784, "pid": 76337, "tid": -914061504, "ts": 1716454223851229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851281, "dur": 0,
    "args": {
      "External id": 158794,
      "cbid": 317, "correlation": 158794
    }
  },
  {
    "ph": "f", "id": 158794, "pid": 76337, "tid": -914061504, "ts": 1716454223851281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851282, "dur": 0,
    "args": {
      "External id": 158795,
      "cbid": 203, "correlation": 158795
    }
  },
  {
    "ph": "f", "id": 158795, "pid": 76337, "tid": -914061504, "ts": 1716454223851282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851283, "dur": 0,
    "args": {
      "External id": 158796,
      "cbid": 205, "correlation": 158796
    }
  },
  {
    "ph": "f", "id": 158796, "pid": 76337, "tid": -914061504, "ts": 1716454223851283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223908140, "dur": 22,
    "args": {
      "External id": 158800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158800, "pid": 5, "tid": 7, "ts": 1716454223908140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851299, "dur": 11,
    "args": {
      "External id": 158800,
      "cbid": 211, "correlation": 158800
    }
  },
  {
    "ph": "s", "id": 158800, "pid": 76337, "tid": -914061504, "ts": 1716454223851299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223908163, "dur": 320,
    "args": {
      "External id": 158802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158802, "pid": 5, "tid": 7, "ts": 1716454223908163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851312, "dur": 5,
    "args": {
      "External id": 158802,
      "cbid": 211, "correlation": 158802
    }
  },
  {
    "ph": "s", "id": 158802, "pid": 76337, "tid": -914061504, "ts": 1716454223851312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223908486, "dur": 1,
    "args": {
      "External id": 158804,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158804,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 158804, "pid": 5, "tid": 7, "ts": 1716454223908486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851323, "dur": 8,
    "args": {
      "External id": 158804,
      "cbid": 51, "correlation": 158804
    }
  },
  {
    "ph": "s", "id": 158804, "pid": 76337, "tid": -914061504, "ts": 1716454223851323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223908489, "dur": 1237,
    "args": {
      "External id": 158805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158805,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158805, "pid": 5, "tid": 7, "ts": 1716454223908489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851332, "dur": 6,
    "args": {
      "External id": 158805,
      "cbid": 211, "correlation": 158805
    }
  },
  {
    "ph": "s", "id": 158805, "pid": 76337, "tid": -914061504, "ts": 1716454223851332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223909728, "dur": 12,
    "args": {
      "External id": 158807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158807,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158807, "pid": 5, "tid": 7, "ts": 1716454223909728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851342, "dur": 5,
    "args": {
      "External id": 158807,
      "cbid": 211, "correlation": 158807
    }
  },
  {
    "ph": "s", "id": 158807, "pid": 76337, "tid": -914061504, "ts": 1716454223851342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223909741, "dur": 14,
    "args": {
      "External id": 158813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158813, "pid": 5, "tid": 7, "ts": 1716454223909741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851370, "dur": 10,
    "args": {
      "External id": 158813,
      "cbid": 211, "correlation": 158813
    }
  },
  {
    "ph": "s", "id": 158813, "pid": 76337, "tid": -914061504, "ts": 1716454223851370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223909757, "dur": 3,
    "args": {
      "External id": 158821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158821,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 158821, "pid": 5, "tid": 7, "ts": 1716454223909757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851415, "dur": 10,
    "args": {
      "External id": 158821,
      "cbid": 211, "correlation": 158821
    }
  },
  {
    "ph": "s", "id": 158821, "pid": 76337, "tid": -914061504, "ts": 1716454223851415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851483, "dur": 1,
    "args": {
      "External id": 158837,
      "cbid": 251, "correlation": 158837
    }
  },
  {
    "ph": "f", "id": 158837, "pid": 76337, "tid": -914061504, "ts": 1716454223851483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851488, "dur": 0,
    "args": {
      "External id": 158839,
      "cbid": 251, "correlation": 158839
    }
  },
  {
    "ph": "f", "id": 158839, "pid": 76337, "tid": -914061504, "ts": 1716454223851488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223909762, "dur": 12,
    "args": {
      "External id": 158840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158840,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158840, "pid": 5, "tid": 7, "ts": 1716454223909762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851490, "dur": 11,
    "args": {
      "External id": 158840,
      "cbid": 211, "correlation": 158840
    }
  },
  {
    "ph": "s", "id": 158840, "pid": 76337, "tid": -914061504, "ts": 1716454223851490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223909775, "dur": 5,
    "args": {
      "External id": 158842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158842,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158842, "pid": 5, "tid": 7, "ts": 1716454223909775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851503, "dur": 6,
    "args": {
      "External id": 158842,
      "cbid": 211, "correlation": 158842
    }
  },
  {
    "ph": "s", "id": 158842, "pid": 76337, "tid": -914061504, "ts": 1716454223851503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223909782, "dur": 17,
    "args": {
      "External id": 158852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158852, "pid": 5, "tid": 7, "ts": 1716454223909782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851561, "dur": 13,
    "args": {
      "External id": 158852,
      "cbid": 211, "correlation": 158852
    }
  },
  {
    "ph": "s", "id": 158852, "pid": 76337, "tid": -914061504, "ts": 1716454223851561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223909800, "dur": 17,
    "args": {
      "External id": 158872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158872,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158872, "pid": 5, "tid": 7, "ts": 1716454223909800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851628, "dur": 11,
    "args": {
      "External id": 158872,
      "cbid": 211, "correlation": 158872
    }
  },
  {
    "ph": "s", "id": 158872, "pid": 76337, "tid": -914061504, "ts": 1716454223851628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223909818, "dur": 4,
    "args": {
      "External id": 158884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158884,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 158884, "pid": 5, "tid": 7, "ts": 1716454223909818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851649, "dur": 6,
    "args": {
      "External id": 158884,
      "cbid": 211, "correlation": 158884
    }
  },
  {
    "ph": "s", "id": 158884, "pid": 76337, "tid": -914061504, "ts": 1716454223851649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223909824, "dur": 16,
    "args": {
      "External id": 158887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158887, "pid": 5, "tid": 7, "ts": 1716454223909824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851668, "dur": 7,
    "args": {
      "External id": 158887,
      "cbid": 211, "correlation": 158887
    }
  },
  {
    "ph": "s", "id": 158887, "pid": 76337, "tid": -914061504, "ts": 1716454223851668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223909842, "dur": 11,
    "args": {
      "External id": 158896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158896,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158896, "pid": 5, "tid": 7, "ts": 1716454223909842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851708, "dur": 10,
    "args": {
      "External id": 158896,
      "cbid": 211, "correlation": 158896
    }
  },
  {
    "ph": "s", "id": 158896, "pid": 76337, "tid": -914061504, "ts": 1716454223851708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851770, "dur": 0,
    "args": {
      "External id": 158906,
      "cbid": 317, "correlation": 158906
    }
  },
  {
    "ph": "f", "id": 158906, "pid": 76337, "tid": -914061504, "ts": 1716454223851770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851771, "dur": 0,
    "args": {
      "External id": 158907,
      "cbid": 203, "correlation": 158907
    }
  },
  {
    "ph": "f", "id": 158907, "pid": 76337, "tid": -914061504, "ts": 1716454223851771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851771, "dur": 0,
    "args": {
      "External id": 158908,
      "cbid": 205, "correlation": 158908
    }
  },
  {
    "ph": "f", "id": 158908, "pid": 76337, "tid": -914061504, "ts": 1716454223851771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223909854, "dur": 11,
    "args": {
      "External id": 158912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158912,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158912, "pid": 5, "tid": 7, "ts": 1716454223909854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851786, "dur": 12,
    "args": {
      "External id": 158912,
      "cbid": 211, "correlation": 158912
    }
  },
  {
    "ph": "s", "id": 158912, "pid": 76337, "tid": -914061504, "ts": 1716454223851786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223909867, "dur": 162,
    "args": {
      "External id": 158914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158914, "pid": 5, "tid": 7, "ts": 1716454223909867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851800, "dur": 5,
    "args": {
      "External id": 158914,
      "cbid": 211, "correlation": 158914
    }
  },
  {
    "ph": "s", "id": 158914, "pid": 76337, "tid": -914061504, "ts": 1716454223851800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223910031, "dur": 1,
    "args": {
      "External id": 158916,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 158916,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 158916, "pid": 5, "tid": 7, "ts": 1716454223910031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851811, "dur": 6,
    "args": {
      "External id": 158916,
      "cbid": 51, "correlation": 158916
    }
  },
  {
    "ph": "s", "id": 158916, "pid": 76337, "tid": -914061504, "ts": 1716454223851811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223910034, "dur": 646,
    "args": {
      "External id": 158917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158917,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 158917, "pid": 5, "tid": 7, "ts": 1716454223910034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851818, "dur": 6,
    "args": {
      "External id": 158917,
      "cbid": 211, "correlation": 158917
    }
  },
  {
    "ph": "s", "id": 158917, "pid": 76337, "tid": -914061504, "ts": 1716454223851818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223910682, "dur": 12,
    "args": {
      "External id": 158919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158919,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158919, "pid": 5, "tid": 7, "ts": 1716454223910682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851829, "dur": 5,
    "args": {
      "External id": 158919,
      "cbid": 211, "correlation": 158919
    }
  },
  {
    "ph": "s", "id": 158919, "pid": 76337, "tid": -914061504, "ts": 1716454223851829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223910696, "dur": 15,
    "args": {
      "External id": 158925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158925, "pid": 5, "tid": 7, "ts": 1716454223910696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851857, "dur": 9,
    "args": {
      "External id": 158925,
      "cbid": 211, "correlation": 158925
    }
  },
  {
    "ph": "s", "id": 158925, "pid": 76337, "tid": -914061504, "ts": 1716454223851857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851916, "dur": 0,
    "args": {
      "External id": 158935,
      "cbid": 317, "correlation": 158935
    }
  },
  {
    "ph": "f", "id": 158935, "pid": 76337, "tid": -914061504, "ts": 1716454223851916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851917, "dur": 0,
    "args": {
      "External id": 158936,
      "cbid": 203, "correlation": 158936
    }
  },
  {
    "ph": "f", "id": 158936, "pid": 76337, "tid": -914061504, "ts": 1716454223851917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851917, "dur": 0,
    "args": {
      "External id": 158937,
      "cbid": 205, "correlation": 158937
    }
  },
  {
    "ph": "f", "id": 158937, "pid": 76337, "tid": -914061504, "ts": 1716454223851917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223910712, "dur": 21,
    "args": {
      "External id": 158941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158941,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158941, "pid": 5, "tid": 7, "ts": 1716454223910712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851932, "dur": 11,
    "args": {
      "External id": 158941,
      "cbid": 211, "correlation": 158941
    }
  },
  {
    "ph": "s", "id": 158941, "pid": 76337, "tid": -914061504, "ts": 1716454223851932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223910734, "dur": 4,
    "args": {
      "External id": 158943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 158943, "pid": 5, "tid": 7, "ts": 1716454223910734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851950, "dur": 7,
    "args": {
      "External id": 158943,
      "cbid": 211, "correlation": 158943
    }
  },
  {
    "ph": "s", "id": 158943, "pid": 76337, "tid": -914061504, "ts": 1716454223851950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851962, "dur": 0,
    "args": {
      "External id": 158944,
      "cbid": 51, "correlation": 158944
    }
  },
  {
    "ph": "s", "id": 158944, "pid": 76337, "tid": -914061504, "ts": 1716454223851962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223910739, "dur": 174,
    "args": {
      "External id": 158945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158945,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 158945, "pid": 5, "tid": 7, "ts": 1716454223910739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223851963, "dur": 7,
    "args": {
      "External id": 158945,
      "cbid": 211, "correlation": 158945
    }
  },
  {
    "ph": "s", "id": 158945, "pid": 76337, "tid": -914061504, "ts": 1716454223851963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223910914, "dur": 16,
    "args": {
      "External id": 158950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158950, "pid": 5, "tid": 7, "ts": 1716454223910914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852000, "dur": 9,
    "args": {
      "External id": 158950,
      "cbid": 211, "correlation": 158950
    }
  },
  {
    "ph": "s", "id": 158950, "pid": 76337, "tid": -914061504, "ts": 1716454223852000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223910931, "dur": 12,
    "args": {
      "External id": 158958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158958,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158958, "pid": 5, "tid": 7, "ts": 1716454223910931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852030, "dur": 8,
    "args": {
      "External id": 158958,
      "cbid": 211, "correlation": 158958
    }
  },
  {
    "ph": "s", "id": 158958, "pid": 76337, "tid": -914061504, "ts": 1716454223852030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223910945, "dur": 11,
    "args": {
      "External id": 158966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158966,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 158966, "pid": 5, "tid": 7, "ts": 1716454223910945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852059, "dur": 9,
    "args": {
      "External id": 158966,
      "cbid": 211, "correlation": 158966
    }
  },
  {
    "ph": "s", "id": 158966, "pid": 76337, "tid": -914061504, "ts": 1716454223852059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223910957, "dur": 18,
    "args": {
      "External id": 158986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158986,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 158986, "pid": 5, "tid": 7, "ts": 1716454223910957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852144, "dur": 12,
    "args": {
      "External id": 158986,
      "cbid": 211, "correlation": 158986
    }
  },
  {
    "ph": "s", "id": 158986, "pid": 76337, "tid": -914061504, "ts": 1716454223852144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223910976, "dur": 4,
    "args": {
      "External id": 158998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 158998,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 158998, "pid": 5, "tid": 7, "ts": 1716454223910976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852166, "dur": 6,
    "args": {
      "External id": 158998,
      "cbid": 211, "correlation": 158998
    }
  },
  {
    "ph": "s", "id": 158998, "pid": 76337, "tid": -914061504, "ts": 1716454223852166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223910982, "dur": 17,
    "args": {
      "External id": 159001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159001,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159001, "pid": 5, "tid": 7, "ts": 1716454223910982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852184, "dur": 6,
    "args": {
      "External id": 159001,
      "cbid": 211, "correlation": 159001
    }
  },
  {
    "ph": "s", "id": 159001, "pid": 76337, "tid": -914061504, "ts": 1716454223852184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852242, "dur": 0,
    "args": {
      "External id": 159012,
      "cbid": 317, "correlation": 159012
    }
  },
  {
    "ph": "f", "id": 159012, "pid": 76337, "tid": -914061504, "ts": 1716454223852242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852243, "dur": 0,
    "args": {
      "External id": 159013,
      "cbid": 203, "correlation": 159013
    }
  },
  {
    "ph": "f", "id": 159013, "pid": 76337, "tid": -914061504, "ts": 1716454223852243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852243, "dur": 0,
    "args": {
      "External id": 159014,
      "cbid": 205, "correlation": 159014
    }
  },
  {
    "ph": "f", "id": 159014, "pid": 76337, "tid": -914061504, "ts": 1716454223852243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223911000, "dur": 11,
    "args": {
      "External id": 159018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159018,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159018, "pid": 5, "tid": 7, "ts": 1716454223911000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852258, "dur": 12,
    "args": {
      "External id": 159018,
      "cbid": 211, "correlation": 159018
    }
  },
  {
    "ph": "s", "id": 159018, "pid": 76337, "tid": -914061504, "ts": 1716454223852258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223911012, "dur": 3,
    "args": {
      "External id": 159020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159020,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159020, "pid": 5, "tid": 7, "ts": 1716454223911012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852275, "dur": 6,
    "args": {
      "External id": 159020,
      "cbid": 211, "correlation": 159020
    }
  },
  {
    "ph": "s", "id": 159020, "pid": 76337, "tid": -914061504, "ts": 1716454223852275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852284, "dur": 0,
    "args": {
      "External id": 159021,
      "cbid": 51, "correlation": 159021
    }
  },
  {
    "ph": "s", "id": 159021, "pid": 76337, "tid": -914061504, "ts": 1716454223852284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223911017, "dur": 90,
    "args": {
      "External id": 159022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159022,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 159022, "pid": 5, "tid": 7, "ts": 1716454223911017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852285, "dur": 5,
    "args": {
      "External id": 159022,
      "cbid": 211, "correlation": 159022
    }
  },
  {
    "ph": "s", "id": 159022, "pid": 76337, "tid": -914061504, "ts": 1716454223852285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223911109, "dur": 16,
    "args": {
      "External id": 159027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159027, "pid": 5, "tid": 7, "ts": 1716454223911109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852311, "dur": 8,
    "args": {
      "External id": 159027,
      "cbid": 211, "correlation": 159027
    }
  },
  {
    "ph": "s", "id": 159027, "pid": 76337, "tid": -914061504, "ts": 1716454223852311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223911126, "dur": 83,
    "args": {
      "External id": 159036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159036,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159036, "pid": 5, "tid": 7, "ts": 1716454223911126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852401, "dur": 16,
    "args": {
      "External id": 159036,
      "cbid": 211, "correlation": 159036
    }
  },
  {
    "ph": "s", "id": 159036, "pid": 76337, "tid": -914061504, "ts": 1716454223852401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223911210, "dur": 31,
    "args": {
      "External id": 159058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159058,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159058, "pid": 5, "tid": 7, "ts": 1716454223911210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852471, "dur": 11,
    "args": {
      "External id": 159058,
      "cbid": 211, "correlation": 159058
    }
  },
  {
    "ph": "s", "id": 159058, "pid": 76337, "tid": -914061504, "ts": 1716454223852471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852583, "dur": 2,
    "args": {
      "External id": 159069,
      "cbid": 251, "correlation": 159069
    }
  },
  {
    "ph": "f", "id": 159069, "pid": 76337, "tid": -914061504, "ts": 1716454223852583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223911242, "dur": 161,
    "args": {
      "External id": 159070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159070,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159070, "pid": 5, "tid": 7, "ts": 1716454223911242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852591, "dur": 14,
    "args": {
      "External id": 159070,
      "cbid": 211, "correlation": 159070
    }
  },
  {
    "ph": "s", "id": 159070, "pid": 76337, "tid": -914061504, "ts": 1716454223852591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852666, "dur": 1,
    "args": {
      "External id": 159081,
      "cbid": 251, "correlation": 159081
    }
  },
  {
    "ph": "f", "id": 159081, "pid": 76337, "tid": -914061504, "ts": 1716454223852666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223911405, "dur": 159,
    "args": {
      "External id": 159082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159082,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159082, "pid": 5, "tid": 7, "ts": 1716454223911405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852670, "dur": 12,
    "args": {
      "External id": 159082,
      "cbid": 211, "correlation": 159082
    }
  },
  {
    "ph": "s", "id": 159082, "pid": 76337, "tid": -914061504, "ts": 1716454223852670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852738, "dur": 1,
    "args": {
      "External id": 159093,
      "cbid": 251, "correlation": 159093
    }
  },
  {
    "ph": "f", "id": 159093, "pid": 76337, "tid": -914061504, "ts": 1716454223852738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223911565, "dur": 135,
    "args": {
      "External id": 159094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159094,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159094, "pid": 5, "tid": 7, "ts": 1716454223911565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852742, "dur": 11,
    "args": {
      "External id": 159094,
      "cbid": 211, "correlation": 159094
    }
  },
  {
    "ph": "s", "id": 159094, "pid": 76337, "tid": -914061504, "ts": 1716454223852742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223911701, "dur": 334,
    "args": {
      "External id": 159119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159119,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159119, "pid": 5, "tid": 7, "ts": 1716454223911701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852848, "dur": 16,
    "args": {
      "External id": 159119,
      "cbid": 211, "correlation": 159119
    }
  },
  {
    "ph": "s", "id": 159119, "pid": 76337, "tid": -914061504, "ts": 1716454223852848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852965, "dur": 2,
    "args": {
      "External id": 159137,
      "cbid": 251, "correlation": 159137
    }
  },
  {
    "ph": "f", "id": 159137, "pid": 76337, "tid": -914061504, "ts": 1716454223852965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223912037, "dur": 165,
    "args": {
      "External id": 159139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159139,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159139, "pid": 5, "tid": 7, "ts": 1716454223912037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223852971, "dur": 22,
    "args": {
      "External id": 159139,
      "cbid": 211, "correlation": 159139
    }
  },
  {
    "ph": "s", "id": 159139, "pid": 76337, "tid": -914061504, "ts": 1716454223852971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223912202, "dur": 19,
    "args": {
      "External id": 159147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159147,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159147, "pid": 5, "tid": 7, "ts": 1716454223912202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853051, "dur": 13,
    "args": {
      "External id": 159147,
      "cbid": 211, "correlation": 159147
    }
  },
  {
    "ph": "s", "id": 159147, "pid": 76337, "tid": -914061504, "ts": 1716454223853051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223912223, "dur": 28,
    "args": {
      "External id": 159155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159155,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159155, "pid": 5, "tid": 7, "ts": 1716454223912223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853091, "dur": 9,
    "args": {
      "External id": 159155,
      "cbid": 211, "correlation": 159155
    }
  },
  {
    "ph": "s", "id": 159155, "pid": 76337, "tid": -914061504, "ts": 1716454223853091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223912252, "dur": 18,
    "args": {
      "External id": 159166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159166, "pid": 5, "tid": 7, "ts": 1716454223912252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853179, "dur": 14,
    "args": {
      "External id": 159166,
      "cbid": 211, "correlation": 159166
    }
  },
  {
    "ph": "s", "id": 159166, "pid": 76337, "tid": -914061504, "ts": 1716454223853179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223912272, "dur": 16,
    "args": {
      "External id": 159188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159188,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159188, "pid": 5, "tid": 7, "ts": 1716454223912272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853212, "dur": 8,
    "args": {
      "External id": 159188,
      "cbid": 211, "correlation": 159188
    }
  },
  {
    "ph": "s", "id": 159188, "pid": 76337, "tid": -914061504, "ts": 1716454223853212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853298, "dur": 2,
    "args": {
      "External id": 159199,
      "cbid": 251, "correlation": 159199
    }
  },
  {
    "ph": "f", "id": 159199, "pid": 76337, "tid": -914061504, "ts": 1716454223853298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223912289, "dur": 89,
    "args": {
      "External id": 159200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159200,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159200, "pid": 5, "tid": 7, "ts": 1716454223912289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853305, "dur": 13,
    "args": {
      "External id": 159200,
      "cbid": 211, "correlation": 159200
    }
  },
  {
    "ph": "s", "id": 159200, "pid": 76337, "tid": -914061504, "ts": 1716454223853305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853381, "dur": 1,
    "args": {
      "External id": 159211,
      "cbid": 251, "correlation": 159211
    }
  },
  {
    "ph": "f", "id": 159211, "pid": 76337, "tid": -914061504, "ts": 1716454223853381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853384, "dur": 0,
    "args": {
      "External id": 159212,
      "cbid": 251, "correlation": 159212
    }
  },
  {
    "ph": "f", "id": 159212, "pid": 76337, "tid": -914061504, "ts": 1716454223853384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223912379, "dur": 12,
    "args": {
      "External id": 159213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159213,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159213, "pid": 5, "tid": 7, "ts": 1716454223912379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853386, "dur": 12,
    "args": {
      "External id": 159213,
      "cbid": 211, "correlation": 159213
    }
  },
  {
    "ph": "s", "id": 159213, "pid": 76337, "tid": -914061504, "ts": 1716454223853386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223912393, "dur": 6,
    "args": {
      "External id": 159215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159215,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159215, "pid": 5, "tid": 7, "ts": 1716454223912393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853400, "dur": 8,
    "args": {
      "External id": 159215,
      "cbid": 211, "correlation": 159215
    }
  },
  {
    "ph": "s", "id": 159215, "pid": 76337, "tid": -914061504, "ts": 1716454223853400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853460, "dur": 1,
    "args": {
      "External id": 159226,
      "cbid": 251, "correlation": 159226
    }
  },
  {
    "ph": "f", "id": 159226, "pid": 76337, "tid": -914061504, "ts": 1716454223853460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853463, "dur": 0,
    "args": {
      "External id": 159227,
      "cbid": 251, "correlation": 159227
    }
  },
  {
    "ph": "f", "id": 159227, "pid": 76337, "tid": -914061504, "ts": 1716454223853463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223912400, "dur": 9,
    "args": {
      "External id": 159228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159228,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159228, "pid": 5, "tid": 7, "ts": 1716454223912400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853464, "dur": 11,
    "args": {
      "External id": 159228,
      "cbid": 211, "correlation": 159228
    }
  },
  {
    "ph": "s", "id": 159228, "pid": 76337, "tid": -914061504, "ts": 1716454223853464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223912410, "dur": 3,
    "args": {
      "External id": 159230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159230,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159230, "pid": 5, "tid": 7, "ts": 1716454223912410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853477, "dur": 6,
    "args": {
      "External id": 159230,
      "cbid": 211, "correlation": 159230
    }
  },
  {
    "ph": "s", "id": 159230, "pid": 76337, "tid": -914061504, "ts": 1716454223853477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223912414, "dur": 55,
    "args": {
      "External id": 159255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159255,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159255, "pid": 5, "tid": 7, "ts": 1716454223912414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853557, "dur": 12,
    "args": {
      "External id": 159255,
      "cbid": 211, "correlation": 159255
    }
  },
  {
    "ph": "s", "id": 159255, "pid": 76337, "tid": -914061504, "ts": 1716454223853557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853658, "dur": 2,
    "args": {
      "External id": 159273,
      "cbid": 251, "correlation": 159273
    }
  },
  {
    "ph": "f", "id": 159273, "pid": 76337, "tid": -914061504, "ts": 1716454223853658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223912471, "dur": 90,
    "args": {
      "External id": 159275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159275,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159275, "pid": 5, "tid": 7, "ts": 1716454223912471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853665, "dur": 14,
    "args": {
      "External id": 159275,
      "cbid": 211, "correlation": 159275
    }
  },
  {
    "ph": "s", "id": 159275, "pid": 76337, "tid": -914061504, "ts": 1716454223853665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223912562, "dur": 10,
    "args": {
      "External id": 159283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159283,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159283, "pid": 5, "tid": 7, "ts": 1716454223912562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853735, "dur": 12,
    "args": {
      "External id": 159283,
      "cbid": 211, "correlation": 159283
    }
  },
  {
    "ph": "s", "id": 159283, "pid": 76337, "tid": -914061504, "ts": 1716454223853735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223912573, "dur": 21,
    "args": {
      "External id": 159291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159291,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159291, "pid": 5, "tid": 7, "ts": 1716454223912573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853777, "dur": 9,
    "args": {
      "External id": 159291,
      "cbid": 211, "correlation": 159291
    }
  },
  {
    "ph": "s", "id": 159291, "pid": 76337, "tid": -914061504, "ts": 1716454223853777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223912595, "dur": 17,
    "args": {
      "External id": 159313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159313,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159313, "pid": 5, "tid": 7, "ts": 1716454223912595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853829, "dur": 10,
    "args": {
      "External id": 159313,
      "cbid": 211, "correlation": 159313
    }
  },
  {
    "ph": "s", "id": 159313, "pid": 76337, "tid": -914061504, "ts": 1716454223853829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853926, "dur": 1,
    "args": {
      "External id": 159329,
      "cbid": 251, "correlation": 159329
    }
  },
  {
    "ph": "f", "id": 159329, "pid": 76337, "tid": -914061504, "ts": 1716454223853926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853931, "dur": 0,
    "args": {
      "External id": 159331,
      "cbid": 251, "correlation": 159331
    }
  },
  {
    "ph": "f", "id": 159331, "pid": 76337, "tid": -914061504, "ts": 1716454223853931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223912614, "dur": 493,
    "args": {
      "External id": 159332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159332,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159332, "pid": 5, "tid": 7, "ts": 1716454223912614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223853934, "dur": 15,
    "args": {
      "External id": 159332,
      "cbid": 211, "correlation": 159332
    }
  },
  {
    "ph": "s", "id": 159332, "pid": 76337, "tid": -914061504, "ts": 1716454223853934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223913108, "dur": 65,
    "args": {
      "External id": 159340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159340, "pid": 5, "tid": 7, "ts": 1716454223913108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854023, "dur": 13,
    "args": {
      "External id": 159340,
      "cbid": 211, "correlation": 159340
    }
  },
  {
    "ph": "s", "id": 159340, "pid": 76337, "tid": -914061504, "ts": 1716454223854023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223913174, "dur": 67,
    "args": {
      "External id": 159348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159348,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159348, "pid": 5, "tid": 7, "ts": 1716454223913174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854061, "dur": 9,
    "args": {
      "External id": 159348,
      "cbid": 211, "correlation": 159348
    }
  },
  {
    "ph": "s", "id": 159348, "pid": 76337, "tid": -914061504, "ts": 1716454223854061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854144, "dur": 1,
    "args": {
      "External id": 159364,
      "cbid": 251, "correlation": 159364
    }
  },
  {
    "ph": "f", "id": 159364, "pid": 76337, "tid": -914061504, "ts": 1716454223854144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223913244, "dur": 1,
    "args": {
      "External id": 159366,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 159366,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 159366, "pid": 5, "tid": 7, "ts": 1716454223913244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854149, "dur": 11,
    "args": {
      "External id": 159366,
      "cbid": 51, "correlation": 159366
    }
  },
  {
    "ph": "s", "id": 159366, "pid": 76337, "tid": -914061504, "ts": 1716454223854149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223913247, "dur": 273,
    "args": {
      "External id": 159367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159367,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159367, "pid": 5, "tid": 7, "ts": 1716454223913247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854162, "dur": 11,
    "args": {
      "External id": 159367,
      "cbid": 211, "correlation": 159367
    }
  },
  {
    "ph": "s", "id": 159367, "pid": 76337, "tid": -914061504, "ts": 1716454223854162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223913522, "dur": 13,
    "args": {
      "External id": 159375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159375,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159375, "pid": 5, "tid": 7, "ts": 1716454223913522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854205, "dur": 10,
    "args": {
      "External id": 159375,
      "cbid": 211, "correlation": 159375
    }
  },
  {
    "ph": "s", "id": 159375, "pid": 76337, "tid": -914061504, "ts": 1716454223854205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223913537, "dur": 37,
    "args": {
      "External id": 159386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159386, "pid": 5, "tid": 7, "ts": 1716454223913537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854277, "dur": 12,
    "args": {
      "External id": 159386,
      "cbid": 211, "correlation": 159386
    }
  },
  {
    "ph": "s", "id": 159386, "pid": 76337, "tid": -914061504, "ts": 1716454223854277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854341, "dur": 0,
    "args": {
      "External id": 159398,
      "cbid": 317, "correlation": 159398
    }
  },
  {
    "ph": "f", "id": 159398, "pid": 76337, "tid": -914061504, "ts": 1716454223854341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854343, "dur": 0,
    "args": {
      "External id": 159399,
      "cbid": 203, "correlation": 159399
    }
  },
  {
    "ph": "f", "id": 159399, "pid": 76337, "tid": -914061504, "ts": 1716454223854343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854343, "dur": 0,
    "args": {
      "External id": 159400,
      "cbid": 205, "correlation": 159400
    }
  },
  {
    "ph": "f", "id": 159400, "pid": 76337, "tid": -914061504, "ts": 1716454223854343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223913575, "dur": 13,
    "args": {
      "External id": 159404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159404,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159404, "pid": 5, "tid": 7, "ts": 1716454223913575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854360, "dur": 12,
    "args": {
      "External id": 159404,
      "cbid": 211, "correlation": 159404
    }
  },
  {
    "ph": "s", "id": 159404, "pid": 76337, "tid": -914061504, "ts": 1716454223854360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223913590, "dur": 4,
    "args": {
      "External id": 159406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159406,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159406, "pid": 5, "tid": 7, "ts": 1716454223913590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854376, "dur": 6,
    "args": {
      "External id": 159406,
      "cbid": 211, "correlation": 159406
    }
  },
  {
    "ph": "s", "id": 159406, "pid": 76337, "tid": -914061504, "ts": 1716454223854376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854385, "dur": 0,
    "args": {
      "External id": 159407,
      "cbid": 51, "correlation": 159407
    }
  },
  {
    "ph": "s", "id": 159407, "pid": 76337, "tid": -914061504, "ts": 1716454223854385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223913595, "dur": 97,
    "args": {
      "External id": 159408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159408,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 159408, "pid": 5, "tid": 7, "ts": 1716454223913595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854385, "dur": 5,
    "args": {
      "External id": 159408,
      "cbid": 211, "correlation": 159408
    }
  },
  {
    "ph": "s", "id": 159408, "pid": 76337, "tid": -914061504, "ts": 1716454223854385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223913694, "dur": 16,
    "args": {
      "External id": 159413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159413,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159413, "pid": 5, "tid": 7, "ts": 1716454223913694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854413, "dur": 9,
    "args": {
      "External id": 159413,
      "cbid": 211, "correlation": 159413
    }
  },
  {
    "ph": "s", "id": 159413, "pid": 76337, "tid": -914061504, "ts": 1716454223854413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223913712, "dur": 12,
    "args": {
      "External id": 159421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159421,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159421, "pid": 5, "tid": 7, "ts": 1716454223913712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854445, "dur": 9,
    "args": {
      "External id": 159421,
      "cbid": 211, "correlation": 159421
    }
  },
  {
    "ph": "s", "id": 159421, "pid": 76337, "tid": -914061504, "ts": 1716454223854445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223913725, "dur": 29,
    "args": {
      "External id": 159430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159430,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159430, "pid": 5, "tid": 7, "ts": 1716454223913725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854485, "dur": 10,
    "args": {
      "External id": 159430,
      "cbid": 211, "correlation": 159430
    }
  },
  {
    "ph": "s", "id": 159430, "pid": 76337, "tid": -914061504, "ts": 1716454223854485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223913755, "dur": 31,
    "args": {
      "External id": 159450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159450,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 159450, "pid": 5, "tid": 7, "ts": 1716454223913755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854559, "dur": 12,
    "args": {
      "External id": 159450,
      "cbid": 211, "correlation": 159450
    }
  },
  {
    "ph": "s", "id": 159450, "pid": 76337, "tid": -914061504, "ts": 1716454223854559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223913787, "dur": 6,
    "args": {
      "External id": 159462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159462,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159462, "pid": 5, "tid": 7, "ts": 1716454223913787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854581, "dur": 6,
    "args": {
      "External id": 159462,
      "cbid": 211, "correlation": 159462
    }
  },
  {
    "ph": "s", "id": 159462, "pid": 76337, "tid": -914061504, "ts": 1716454223854581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223913795, "dur": 31,
    "args": {
      "External id": 159465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159465,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159465, "pid": 5, "tid": 7, "ts": 1716454223913795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854599, "dur": 6,
    "args": {
      "External id": 159465,
      "cbid": 211, "correlation": 159465
    }
  },
  {
    "ph": "s", "id": 159465, "pid": 76337, "tid": -914061504, "ts": 1716454223854599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223913826, "dur": 21,
    "args": {
      "External id": 159474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159474,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159474, "pid": 5, "tid": 7, "ts": 1716454223913826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854638, "dur": 10,
    "args": {
      "External id": 159474,
      "cbid": 211, "correlation": 159474
    }
  },
  {
    "ph": "s", "id": 159474, "pid": 76337, "tid": -914061504, "ts": 1716454223854638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854689, "dur": 0,
    "args": {
      "External id": 159484,
      "cbid": 317, "correlation": 159484
    }
  },
  {
    "ph": "f", "id": 159484, "pid": 76337, "tid": -914061504, "ts": 1716454223854689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854690, "dur": 0,
    "args": {
      "External id": 159485,
      "cbid": 203, "correlation": 159485
    }
  },
  {
    "ph": "f", "id": 159485, "pid": 76337, "tid": -914061504, "ts": 1716454223854690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854691, "dur": 0,
    "args": {
      "External id": 159486,
      "cbid": 205, "correlation": 159486
    }
  },
  {
    "ph": "f", "id": 159486, "pid": 76337, "tid": -914061504, "ts": 1716454223854691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223913849, "dur": 21,
    "args": {
      "External id": 159490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159490, "pid": 5, "tid": 7, "ts": 1716454223913849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854705, "dur": 11,
    "args": {
      "External id": 159490,
      "cbid": 211, "correlation": 159490
    }
  },
  {
    "ph": "s", "id": 159490, "pid": 76337, "tid": -914061504, "ts": 1716454223854705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223913872, "dur": 319,
    "args": {
      "External id": 159492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159492,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159492, "pid": 5, "tid": 7, "ts": 1716454223913872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854719, "dur": 5,
    "args": {
      "External id": 159492,
      "cbid": 211, "correlation": 159492
    }
  },
  {
    "ph": "s", "id": 159492, "pid": 76337, "tid": -914061504, "ts": 1716454223854719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223914193, "dur": 1,
    "args": {
      "External id": 159494,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 159494,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5885959534028203
    }
  },
  {
    "ph": "f", "id": 159494, "pid": 5, "tid": 7, "ts": 1716454223914193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854730, "dur": 7,
    "args": {
      "External id": 159494,
      "cbid": 51, "correlation": 159494
    }
  },
  {
    "ph": "s", "id": 159494, "pid": 76337, "tid": -914061504, "ts": 1716454223854730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223914197, "dur": 1254,
    "args": {
      "External id": 159495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159495,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159495, "pid": 5, "tid": 7, "ts": 1716454223914197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854738, "dur": 7,
    "args": {
      "External id": 159495,
      "cbid": 211, "correlation": 159495
    }
  },
  {
    "ph": "s", "id": 159495, "pid": 76337, "tid": -914061504, "ts": 1716454223854738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223915452, "dur": 13,
    "args": {
      "External id": 159497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159497,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159497, "pid": 5, "tid": 7, "ts": 1716454223915452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854750, "dur": 5,
    "args": {
      "External id": 159497,
      "cbid": 211, "correlation": 159497
    }
  },
  {
    "ph": "s", "id": 159497, "pid": 76337, "tid": -914061504, "ts": 1716454223854750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223915466, "dur": 15,
    "args": {
      "External id": 159503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159503,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159503, "pid": 5, "tid": 7, "ts": 1716454223915466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854777, "dur": 8,
    "args": {
      "External id": 159503,
      "cbid": 211, "correlation": 159503
    }
  },
  {
    "ph": "s", "id": 159503, "pid": 76337, "tid": -914061504, "ts": 1716454223854777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223915482, "dur": 4,
    "args": {
      "External id": 159511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159511,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 159511, "pid": 5, "tid": 7, "ts": 1716454223915482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854821, "dur": 9,
    "args": {
      "External id": 159511,
      "cbid": 211, "correlation": 159511
    }
  },
  {
    "ph": "s", "id": 159511, "pid": 76337, "tid": -914061504, "ts": 1716454223854821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854887, "dur": 1,
    "args": {
      "External id": 159527,
      "cbid": 251, "correlation": 159527
    }
  },
  {
    "ph": "f", "id": 159527, "pid": 76337, "tid": -914061504, "ts": 1716454223854887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854893, "dur": 0,
    "args": {
      "External id": 159529,
      "cbid": 251, "correlation": 159529
    }
  },
  {
    "ph": "f", "id": 159529, "pid": 76337, "tid": -914061504, "ts": 1716454223854893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223915487, "dur": 13,
    "args": {
      "External id": 159530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159530,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159530, "pid": 5, "tid": 7, "ts": 1716454223915487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854895, "dur": 12,
    "args": {
      "External id": 159530,
      "cbid": 211, "correlation": 159530
    }
  },
  {
    "ph": "s", "id": 159530, "pid": 76337, "tid": -914061504, "ts": 1716454223854895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223915501, "dur": 5,
    "args": {
      "External id": 159532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159532,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159532, "pid": 5, "tid": 7, "ts": 1716454223915501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854908, "dur": 5,
    "args": {
      "External id": 159532,
      "cbid": 211, "correlation": 159532
    }
  },
  {
    "ph": "s", "id": 159532, "pid": 76337, "tid": -914061504, "ts": 1716454223854908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223915508, "dur": 17,
    "args": {
      "External id": 159542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159542, "pid": 5, "tid": 7, "ts": 1716454223915508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223854966, "dur": 21,
    "args": {
      "External id": 159542,
      "cbid": 211, "correlation": 159542
    }
  },
  {
    "ph": "s", "id": 159542, "pid": 76337, "tid": -914061504, "ts": 1716454223854966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223915526, "dur": 17,
    "args": {
      "External id": 159562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159562,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 159562, "pid": 5, "tid": 7, "ts": 1716454223915526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855041, "dur": 12,
    "args": {
      "External id": 159562,
      "cbid": 211, "correlation": 159562
    }
  },
  {
    "ph": "s", "id": 159562, "pid": 76337, "tid": -914061504, "ts": 1716454223855041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223915544, "dur": 4,
    "args": {
      "External id": 159574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159574,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 159574, "pid": 5, "tid": 7, "ts": 1716454223915544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855064, "dur": 6,
    "args": {
      "External id": 159574,
      "cbid": 211, "correlation": 159574
    }
  },
  {
    "ph": "s", "id": 159574, "pid": 76337, "tid": -914061504, "ts": 1716454223855064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223915550, "dur": 16,
    "args": {
      "External id": 159577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159577,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159577, "pid": 5, "tid": 7, "ts": 1716454223915550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855082, "dur": 6,
    "args": {
      "External id": 159577,
      "cbid": 211, "correlation": 159577
    }
  },
  {
    "ph": "s", "id": 159577, "pid": 76337, "tid": -914061504, "ts": 1716454223855082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223915568, "dur": 11,
    "args": {
      "External id": 159586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159586,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159586, "pid": 5, "tid": 7, "ts": 1716454223915568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855123, "dur": 9,
    "args": {
      "External id": 159586,
      "cbid": 211, "correlation": 159586
    }
  },
  {
    "ph": "s", "id": 159586, "pid": 76337, "tid": -914061504, "ts": 1716454223855123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855184, "dur": 0,
    "args": {
      "External id": 159596,
      "cbid": 317, "correlation": 159596
    }
  },
  {
    "ph": "f", "id": 159596, "pid": 76337, "tid": -914061504, "ts": 1716454223855184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855185, "dur": 0,
    "args": {
      "External id": 159597,
      "cbid": 203, "correlation": 159597
    }
  },
  {
    "ph": "f", "id": 159597, "pid": 76337, "tid": -914061504, "ts": 1716454223855185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855186, "dur": 0,
    "args": {
      "External id": 159598,
      "cbid": 205, "correlation": 159598
    }
  },
  {
    "ph": "f", "id": 159598, "pid": 76337, "tid": -914061504, "ts": 1716454223855186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223915581, "dur": 12,
    "args": {
      "External id": 159602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159602, "pid": 5, "tid": 7, "ts": 1716454223915581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855200, "dur": 12,
    "args": {
      "External id": 159602,
      "cbid": 211, "correlation": 159602
    }
  },
  {
    "ph": "s", "id": 159602, "pid": 76337, "tid": -914061504, "ts": 1716454223855200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223915594, "dur": 162,
    "args": {
      "External id": 159604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159604,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159604, "pid": 5, "tid": 7, "ts": 1716454223915594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855214, "dur": 5,
    "args": {
      "External id": 159604,
      "cbid": 211, "correlation": 159604
    }
  },
  {
    "ph": "s", "id": 159604, "pid": 76337, "tid": -914061504, "ts": 1716454223855214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223915758, "dur": 1,
    "args": {
      "External id": 159606,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 159606,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 159606, "pid": 5, "tid": 7, "ts": 1716454223915758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855226, "dur": 6,
    "args": {
      "External id": 159606,
      "cbid": 51, "correlation": 159606
    }
  },
  {
    "ph": "s", "id": 159606, "pid": 76337, "tid": -914061504, "ts": 1716454223855226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223915762, "dur": 648,
    "args": {
      "External id": 159607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159607,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159607, "pid": 5, "tid": 7, "ts": 1716454223915762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855233, "dur": 6,
    "args": {
      "External id": 159607,
      "cbid": 211, "correlation": 159607
    }
  },
  {
    "ph": "s", "id": 159607, "pid": 76337, "tid": -914061504, "ts": 1716454223855233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223916411, "dur": 12,
    "args": {
      "External id": 159609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159609,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159609, "pid": 5, "tid": 7, "ts": 1716454223916411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855244, "dur": 6,
    "args": {
      "External id": 159609,
      "cbid": 211, "correlation": 159609
    }
  },
  {
    "ph": "s", "id": 159609, "pid": 76337, "tid": -914061504, "ts": 1716454223855244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223916424, "dur": 15,
    "args": {
      "External id": 159615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159615,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159615, "pid": 5, "tid": 7, "ts": 1716454223916424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855273, "dur": 9,
    "args": {
      "External id": 159615,
      "cbid": 211, "correlation": 159615
    }
  },
  {
    "ph": "s", "id": 159615, "pid": 76337, "tid": -914061504, "ts": 1716454223855273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855331, "dur": 0,
    "args": {
      "External id": 159625,
      "cbid": 317, "correlation": 159625
    }
  },
  {
    "ph": "f", "id": 159625, "pid": 76337, "tid": -914061504, "ts": 1716454223855331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855332, "dur": 0,
    "args": {
      "External id": 159626,
      "cbid": 203, "correlation": 159626
    }
  },
  {
    "ph": "f", "id": 159626, "pid": 76337, "tid": -914061504, "ts": 1716454223855332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855332, "dur": 0,
    "args": {
      "External id": 159627,
      "cbid": 205, "correlation": 159627
    }
  },
  {
    "ph": "f", "id": 159627, "pid": 76337, "tid": -914061504, "ts": 1716454223855332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223916441, "dur": 21,
    "args": {
      "External id": 159631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159631,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159631, "pid": 5, "tid": 7, "ts": 1716454223916441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855346, "dur": 11,
    "args": {
      "External id": 159631,
      "cbid": 211, "correlation": 159631
    }
  },
  {
    "ph": "s", "id": 159631, "pid": 76337, "tid": -914061504, "ts": 1716454223855346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223916463, "dur": 4,
    "args": {
      "External id": 159633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159633, "pid": 5, "tid": 7, "ts": 1716454223916463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855361, "dur": 5,
    "args": {
      "External id": 159633,
      "cbid": 211, "correlation": 159633
    }
  },
  {
    "ph": "s", "id": 159633, "pid": 76337, "tid": -914061504, "ts": 1716454223855361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855369, "dur": 0,
    "args": {
      "External id": 159634,
      "cbid": 51, "correlation": 159634
    }
  },
  {
    "ph": "s", "id": 159634, "pid": 76337, "tid": -914061504, "ts": 1716454223855369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223916468, "dur": 171,
    "args": {
      "External id": 159635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159635,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 159635, "pid": 5, "tid": 7, "ts": 1716454223916468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855370, "dur": 5,
    "args": {
      "External id": 159635,
      "cbid": 211, "correlation": 159635
    }
  },
  {
    "ph": "s", "id": 159635, "pid": 76337, "tid": -914061504, "ts": 1716454223855370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223916640, "dur": 16,
    "args": {
      "External id": 159640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159640, "pid": 5, "tid": 7, "ts": 1716454223916640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855396, "dur": 8,
    "args": {
      "External id": 159640,
      "cbid": 211, "correlation": 159640
    }
  },
  {
    "ph": "s", "id": 159640, "pid": 76337, "tid": -914061504, "ts": 1716454223855396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223916658, "dur": 12,
    "args": {
      "External id": 159648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159648,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159648, "pid": 5, "tid": 7, "ts": 1716454223916658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855424, "dur": 8,
    "args": {
      "External id": 159648,
      "cbid": 211, "correlation": 159648
    }
  },
  {
    "ph": "s", "id": 159648, "pid": 76337, "tid": -914061504, "ts": 1716454223855424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223916671, "dur": 10,
    "args": {
      "External id": 159656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159656,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159656, "pid": 5, "tid": 7, "ts": 1716454223916671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855454, "dur": 8,
    "args": {
      "External id": 159656,
      "cbid": 211, "correlation": 159656
    }
  },
  {
    "ph": "s", "id": 159656, "pid": 76337, "tid": -914061504, "ts": 1716454223855454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223916682, "dur": 18,
    "args": {
      "External id": 159676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159676,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 159676, "pid": 5, "tid": 7, "ts": 1716454223916682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855534, "dur": 13,
    "args": {
      "External id": 159676,
      "cbid": 211, "correlation": 159676
    }
  },
  {
    "ph": "s", "id": 159676, "pid": 76337, "tid": -914061504, "ts": 1716454223855534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223916702, "dur": 5,
    "args": {
      "External id": 159688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159688,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 159688, "pid": 5, "tid": 7, "ts": 1716454223916702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855557, "dur": 7,
    "args": {
      "External id": 159688,
      "cbid": 211, "correlation": 159688
    }
  },
  {
    "ph": "s", "id": 159688, "pid": 76337, "tid": -914061504, "ts": 1716454223855557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223916707, "dur": 17,
    "args": {
      "External id": 159691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159691, "pid": 5, "tid": 7, "ts": 1716454223916707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855575, "dur": 6,
    "args": {
      "External id": 159691,
      "cbid": 211, "correlation": 159691
    }
  },
  {
    "ph": "s", "id": 159691, "pid": 76337, "tid": -914061504, "ts": 1716454223855575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855633, "dur": 0,
    "args": {
      "External id": 159702,
      "cbid": 317, "correlation": 159702
    }
  },
  {
    "ph": "f", "id": 159702, "pid": 76337, "tid": -914061504, "ts": 1716454223855633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855634, "dur": 0,
    "args": {
      "External id": 159703,
      "cbid": 203, "correlation": 159703
    }
  },
  {
    "ph": "f", "id": 159703, "pid": 76337, "tid": -914061504, "ts": 1716454223855634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855635, "dur": 0,
    "args": {
      "External id": 159704,
      "cbid": 205, "correlation": 159704
    }
  },
  {
    "ph": "f", "id": 159704, "pid": 76337, "tid": -914061504, "ts": 1716454223855635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223916726, "dur": 11,
    "args": {
      "External id": 159708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159708,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159708, "pid": 5, "tid": 7, "ts": 1716454223916726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855648, "dur": 11,
    "args": {
      "External id": 159708,
      "cbid": 211, "correlation": 159708
    }
  },
  {
    "ph": "s", "id": 159708, "pid": 76337, "tid": -914061504, "ts": 1716454223855648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223916738, "dur": 3,
    "args": {
      "External id": 159710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159710, "pid": 5, "tid": 7, "ts": 1716454223916738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855664, "dur": 5,
    "args": {
      "External id": 159710,
      "cbid": 211, "correlation": 159710
    }
  },
  {
    "ph": "s", "id": 159710, "pid": 76337, "tid": -914061504, "ts": 1716454223855664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855672, "dur": 0,
    "args": {
      "External id": 159711,
      "cbid": 51, "correlation": 159711
    }
  },
  {
    "ph": "s", "id": 159711, "pid": 76337, "tid": -914061504, "ts": 1716454223855672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223916742, "dur": 91,
    "args": {
      "External id": 159712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159712,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 159712, "pid": 5, "tid": 7, "ts": 1716454223916742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855673, "dur": 5,
    "args": {
      "External id": 159712,
      "cbid": 211, "correlation": 159712
    }
  },
  {
    "ph": "s", "id": 159712, "pid": 76337, "tid": -914061504, "ts": 1716454223855673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223916835, "dur": 16,
    "args": {
      "External id": 159717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159717, "pid": 5, "tid": 7, "ts": 1716454223916835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855700, "dur": 8,
    "args": {
      "External id": 159717,
      "cbid": 211, "correlation": 159717
    }
  },
  {
    "ph": "s", "id": 159717, "pid": 76337, "tid": -914061504, "ts": 1716454223855700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223916852, "dur": 84,
    "args": {
      "External id": 159726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159726,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159726, "pid": 5, "tid": 7, "ts": 1716454223916852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855780, "dur": 16,
    "args": {
      "External id": 159726,
      "cbid": 211, "correlation": 159726
    }
  },
  {
    "ph": "s", "id": 159726, "pid": 76337, "tid": -914061504, "ts": 1716454223855780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223916938, "dur": 30,
    "args": {
      "External id": 159748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159748,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159748, "pid": 5, "tid": 7, "ts": 1716454223916938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855839, "dur": 11,
    "args": {
      "External id": 159748,
      "cbid": 211, "correlation": 159748
    }
  },
  {
    "ph": "s", "id": 159748, "pid": 76337, "tid": -914061504, "ts": 1716454223855839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855927, "dur": 2,
    "args": {
      "External id": 159759,
      "cbid": 251, "correlation": 159759
    }
  },
  {
    "ph": "f", "id": 159759, "pid": 76337, "tid": -914061504, "ts": 1716454223855927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223916969, "dur": 164,
    "args": {
      "External id": 159760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159760,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159760, "pid": 5, "tid": 7, "ts": 1716454223916969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223855932, "dur": 14,
    "args": {
      "External id": 159760,
      "cbid": 211, "correlation": 159760
    }
  },
  {
    "ph": "s", "id": 159760, "pid": 76337, "tid": -914061504, "ts": 1716454223855932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856013, "dur": 1,
    "args": {
      "External id": 159771,
      "cbid": 251, "correlation": 159771
    }
  },
  {
    "ph": "f", "id": 159771, "pid": 76337, "tid": -914061504, "ts": 1716454223856013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223917135, "dur": 160,
    "args": {
      "External id": 159772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159772,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159772, "pid": 5, "tid": 7, "ts": 1716454223917135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856017, "dur": 12,
    "args": {
      "External id": 159772,
      "cbid": 211, "correlation": 159772
    }
  },
  {
    "ph": "s", "id": 159772, "pid": 76337, "tid": -914061504, "ts": 1716454223856017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856085, "dur": 1,
    "args": {
      "External id": 159783,
      "cbid": 251, "correlation": 159783
    }
  },
  {
    "ph": "f", "id": 159783, "pid": 76337, "tid": -914061504, "ts": 1716454223856085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223917296, "dur": 160,
    "args": {
      "External id": 159784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159784,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159784, "pid": 5, "tid": 7, "ts": 1716454223917296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856089, "dur": 12,
    "args": {
      "External id": 159784,
      "cbid": 211, "correlation": 159784
    }
  },
  {
    "ph": "s", "id": 159784, "pid": 76337, "tid": -914061504, "ts": 1716454223856089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223917458, "dur": 336,
    "args": {
      "External id": 159809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159809,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159809, "pid": 5, "tid": 7, "ts": 1716454223917458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856174, "dur": 13,
    "args": {
      "External id": 159809,
      "cbid": 211, "correlation": 159809
    }
  },
  {
    "ph": "s", "id": 159809, "pid": 76337, "tid": -914061504, "ts": 1716454223856174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856275, "dur": 1,
    "args": {
      "External id": 159827,
      "cbid": 251, "correlation": 159827
    }
  },
  {
    "ph": "f", "id": 159827, "pid": 76337, "tid": -914061504, "ts": 1716454223856275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223917795, "dur": 168,
    "args": {
      "External id": 159829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159829,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159829, "pid": 5, "tid": 7, "ts": 1716454223917795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856281, "dur": 14,
    "args": {
      "External id": 159829,
      "cbid": 211, "correlation": 159829
    }
  },
  {
    "ph": "s", "id": 159829, "pid": 76337, "tid": -914061504, "ts": 1716454223856281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223917964, "dur": 19,
    "args": {
      "External id": 159837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159837,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159837, "pid": 5, "tid": 7, "ts": 1716454223917964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856351, "dur": 12,
    "args": {
      "External id": 159837,
      "cbid": 211, "correlation": 159837
    }
  },
  {
    "ph": "s", "id": 159837, "pid": 76337, "tid": -914061504, "ts": 1716454223856351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223917985, "dur": 27,
    "args": {
      "External id": 159845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159845,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159845, "pid": 5, "tid": 7, "ts": 1716454223917985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856390, "dur": 9,
    "args": {
      "External id": 159845,
      "cbid": 211, "correlation": 159845
    }
  },
  {
    "ph": "s", "id": 159845, "pid": 76337, "tid": -914061504, "ts": 1716454223856390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223918013, "dur": 19,
    "args": {
      "External id": 159856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159856, "pid": 5, "tid": 7, "ts": 1716454223918013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856462, "dur": 12,
    "args": {
      "External id": 159856,
      "cbid": 211, "correlation": 159856
    }
  },
  {
    "ph": "s", "id": 159856, "pid": 76337, "tid": -914061504, "ts": 1716454223856462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223918033, "dur": 16,
    "args": {
      "External id": 159878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159878,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159878, "pid": 5, "tid": 7, "ts": 1716454223918033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856493, "dur": 8,
    "args": {
      "External id": 159878,
      "cbid": 211, "correlation": 159878
    }
  },
  {
    "ph": "s", "id": 159878, "pid": 76337, "tid": -914061504, "ts": 1716454223856493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856579, "dur": 1,
    "args": {
      "External id": 159889,
      "cbid": 251, "correlation": 159889
    }
  },
  {
    "ph": "f", "id": 159889, "pid": 76337, "tid": -914061504, "ts": 1716454223856579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223918051, "dur": 89,
    "args": {
      "External id": 159890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159890,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159890, "pid": 5, "tid": 7, "ts": 1716454223918051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856584, "dur": 14,
    "args": {
      "External id": 159890,
      "cbid": 211, "correlation": 159890
    }
  },
  {
    "ph": "s", "id": 159890, "pid": 76337, "tid": -914061504, "ts": 1716454223856584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856653, "dur": 1,
    "args": {
      "External id": 159901,
      "cbid": 251, "correlation": 159901
    }
  },
  {
    "ph": "f", "id": 159901, "pid": 76337, "tid": -914061504, "ts": 1716454223856653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856657, "dur": 0,
    "args": {
      "External id": 159902,
      "cbid": 251, "correlation": 159902
    }
  },
  {
    "ph": "f", "id": 159902, "pid": 76337, "tid": -914061504, "ts": 1716454223856657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223918141, "dur": 13,
    "args": {
      "External id": 159903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159903,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159903, "pid": 5, "tid": 7, "ts": 1716454223918141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856659, "dur": 12,
    "args": {
      "External id": 159903,
      "cbid": 211, "correlation": 159903
    }
  },
  {
    "ph": "s", "id": 159903, "pid": 76337, "tid": -914061504, "ts": 1716454223856659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223918155, "dur": 5,
    "args": {
      "External id": 159905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159905,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159905, "pid": 5, "tid": 7, "ts": 1716454223918155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856672, "dur": 6,
    "args": {
      "External id": 159905,
      "cbid": 211, "correlation": 159905
    }
  },
  {
    "ph": "s", "id": 159905, "pid": 76337, "tid": -914061504, "ts": 1716454223856672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856730, "dur": 1,
    "args": {
      "External id": 159916,
      "cbid": 251, "correlation": 159916
    }
  },
  {
    "ph": "f", "id": 159916, "pid": 76337, "tid": -914061504, "ts": 1716454223856730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856733, "dur": 0,
    "args": {
      "External id": 159917,
      "cbid": 251, "correlation": 159917
    }
  },
  {
    "ph": "f", "id": 159917, "pid": 76337, "tid": -914061504, "ts": 1716454223856733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223918162, "dur": 8,
    "args": {
      "External id": 159918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159918,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159918, "pid": 5, "tid": 7, "ts": 1716454223918162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856734, "dur": 13,
    "args": {
      "External id": 159918,
      "cbid": 211, "correlation": 159918
    }
  },
  {
    "ph": "s", "id": 159918, "pid": 76337, "tid": -914061504, "ts": 1716454223856734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223918171, "dur": 3,
    "args": {
      "External id": 159920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159920,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159920, "pid": 5, "tid": 7, "ts": 1716454223918171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856749, "dur": 6,
    "args": {
      "External id": 159920,
      "cbid": 211, "correlation": 159920
    }
  },
  {
    "ph": "s", "id": 159920, "pid": 76337, "tid": -914061504, "ts": 1716454223856749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223918176, "dur": 56,
    "args": {
      "External id": 159945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159945,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 159945, "pid": 5, "tid": 7, "ts": 1716454223918176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856825, "dur": 12,
    "args": {
      "External id": 159945,
      "cbid": 211, "correlation": 159945
    }
  },
  {
    "ph": "s", "id": 159945, "pid": 76337, "tid": -914061504, "ts": 1716454223856825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856923, "dur": 1,
    "args": {
      "External id": 159963,
      "cbid": 251, "correlation": 159963
    }
  },
  {
    "ph": "f", "id": 159963, "pid": 76337, "tid": -914061504, "ts": 1716454223856923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223918233, "dur": 91,
    "args": {
      "External id": 159965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159965,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 159965, "pid": 5, "tid": 7, "ts": 1716454223918233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223856929, "dur": 14,
    "args": {
      "External id": 159965,
      "cbid": 211, "correlation": 159965
    }
  },
  {
    "ph": "s", "id": 159965, "pid": 76337, "tid": -914061504, "ts": 1716454223856929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223918325, "dur": 9,
    "args": {
      "External id": 159973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159973,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159973, "pid": 5, "tid": 7, "ts": 1716454223918325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857009, "dur": 12,
    "args": {
      "External id": 159973,
      "cbid": 211, "correlation": 159973
    }
  },
  {
    "ph": "s", "id": 159973, "pid": 76337, "tid": -914061504, "ts": 1716454223857009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223918336, "dur": 20,
    "args": {
      "External id": 159981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 159981,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 159981, "pid": 5, "tid": 7, "ts": 1716454223918336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857052, "dur": 9,
    "args": {
      "External id": 159981,
      "cbid": 211, "correlation": 159981
    }
  },
  {
    "ph": "s", "id": 159981, "pid": 76337, "tid": -914061504, "ts": 1716454223857052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223918358, "dur": 17,
    "args": {
      "External id": 160003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160003,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160003, "pid": 5, "tid": 7, "ts": 1716454223918358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857104, "dur": 10,
    "args": {
      "External id": 160003,
      "cbid": 211, "correlation": 160003
    }
  },
  {
    "ph": "s", "id": 160003, "pid": 76337, "tid": -914061504, "ts": 1716454223857104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857191, "dur": 1,
    "args": {
      "External id": 160019,
      "cbid": 251, "correlation": 160019
    }
  },
  {
    "ph": "f", "id": 160019, "pid": 76337, "tid": -914061504, "ts": 1716454223857191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857197, "dur": 0,
    "args": {
      "External id": 160021,
      "cbid": 251, "correlation": 160021
    }
  },
  {
    "ph": "f", "id": 160021, "pid": 76337, "tid": -914061504, "ts": 1716454223857197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223918376, "dur": 498,
    "args": {
      "External id": 160022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160022,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160022, "pid": 5, "tid": 7, "ts": 1716454223918376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857199, "dur": 13,
    "args": {
      "External id": 160022,
      "cbid": 211, "correlation": 160022
    }
  },
  {
    "ph": "s", "id": 160022, "pid": 76337, "tid": -914061504, "ts": 1716454223857199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223918876, "dur": 67,
    "args": {
      "External id": 160030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160030, "pid": 5, "tid": 7, "ts": 1716454223918876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857266, "dur": 12,
    "args": {
      "External id": 160030,
      "cbid": 211, "correlation": 160030
    }
  },
  {
    "ph": "s", "id": 160030, "pid": 76337, "tid": -914061504, "ts": 1716454223857266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223918944, "dur": 66,
    "args": {
      "External id": 160038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160038, "pid": 5, "tid": 7, "ts": 1716454223918944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857297, "dur": 9,
    "args": {
      "External id": 160038,
      "cbid": 211, "correlation": 160038
    }
  },
  {
    "ph": "s", "id": 160038, "pid": 76337, "tid": -914061504, "ts": 1716454223857297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857376, "dur": 1,
    "args": {
      "External id": 160054,
      "cbid": 251, "correlation": 160054
    }
  },
  {
    "ph": "f", "id": 160054, "pid": 76337, "tid": -914061504, "ts": 1716454223857376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223919013, "dur": 1,
    "args": {
      "External id": 160056,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 160056,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 160056, "pid": 5, "tid": 7, "ts": 1716454223919013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857381, "dur": 9,
    "args": {
      "External id": 160056,
      "cbid": 51, "correlation": 160056
    }
  },
  {
    "ph": "s", "id": 160056, "pid": 76337, "tid": -914061504, "ts": 1716454223857381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223919016, "dur": 271,
    "args": {
      "External id": 160057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160057,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160057, "pid": 5, "tid": 7, "ts": 1716454223919016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857392, "dur": 11,
    "args": {
      "External id": 160057,
      "cbid": 211, "correlation": 160057
    }
  },
  {
    "ph": "s", "id": 160057, "pid": 76337, "tid": -914061504, "ts": 1716454223857392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223919289, "dur": 14,
    "args": {
      "External id": 160065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160065,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160065, "pid": 5, "tid": 7, "ts": 1716454223919289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857434, "dur": 11,
    "args": {
      "External id": 160065,
      "cbid": 211, "correlation": 160065
    }
  },
  {
    "ph": "s", "id": 160065, "pid": 76337, "tid": -914061504, "ts": 1716454223857434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223919303, "dur": 38,
    "args": {
      "External id": 160076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160076,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160076, "pid": 5, "tid": 7, "ts": 1716454223919303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857503, "dur": 12,
    "args": {
      "External id": 160076,
      "cbid": 211, "correlation": 160076
    }
  },
  {
    "ph": "s", "id": 160076, "pid": 76337, "tid": -914061504, "ts": 1716454223857503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857568, "dur": 0,
    "args": {
      "External id": 160088,
      "cbid": 317, "correlation": 160088
    }
  },
  {
    "ph": "f", "id": 160088, "pid": 76337, "tid": -914061504, "ts": 1716454223857568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857568, "dur": 0,
    "args": {
      "External id": 160089,
      "cbid": 203, "correlation": 160089
    }
  },
  {
    "ph": "f", "id": 160089, "pid": 76337, "tid": -914061504, "ts": 1716454223857568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857569, "dur": 0,
    "args": {
      "External id": 160090,
      "cbid": 205, "correlation": 160090
    }
  },
  {
    "ph": "f", "id": 160090, "pid": 76337, "tid": -914061504, "ts": 1716454223857569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223919342, "dur": 13,
    "args": {
      "External id": 160094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160094,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160094, "pid": 5, "tid": 7, "ts": 1716454223919342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857584, "dur": 13,
    "args": {
      "External id": 160094,
      "cbid": 211, "correlation": 160094
    }
  },
  {
    "ph": "s", "id": 160094, "pid": 76337, "tid": -914061504, "ts": 1716454223857584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223919357, "dur": 4,
    "args": {
      "External id": 160096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160096, "pid": 5, "tid": 7, "ts": 1716454223919357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857601, "dur": 6,
    "args": {
      "External id": 160096,
      "cbid": 211, "correlation": 160096
    }
  },
  {
    "ph": "s", "id": 160096, "pid": 76337, "tid": -914061504, "ts": 1716454223857601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857610, "dur": 0,
    "args": {
      "External id": 160097,
      "cbid": 51, "correlation": 160097
    }
  },
  {
    "ph": "s", "id": 160097, "pid": 76337, "tid": -914061504, "ts": 1716454223857610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223919362, "dur": 96,
    "args": {
      "External id": 160098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160098,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 160098, "pid": 5, "tid": 7, "ts": 1716454223919362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857611, "dur": 5,
    "args": {
      "External id": 160098,
      "cbid": 211, "correlation": 160098
    }
  },
  {
    "ph": "s", "id": 160098, "pid": 76337, "tid": -914061504, "ts": 1716454223857611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223919460, "dur": 16,
    "args": {
      "External id": 160103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160103,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160103, "pid": 5, "tid": 7, "ts": 1716454223919460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857637, "dur": 10,
    "args": {
      "External id": 160103,
      "cbid": 211, "correlation": 160103
    }
  },
  {
    "ph": "s", "id": 160103, "pid": 76337, "tid": -914061504, "ts": 1716454223857637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223919477, "dur": 11,
    "args": {
      "External id": 160111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160111,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160111, "pid": 5, "tid": 7, "ts": 1716454223919477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857670, "dur": 8,
    "args": {
      "External id": 160111,
      "cbid": 211, "correlation": 160111
    }
  },
  {
    "ph": "s", "id": 160111, "pid": 76337, "tid": -914061504, "ts": 1716454223857670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223919490, "dur": 25,
    "args": {
      "External id": 160120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160120,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160120, "pid": 5, "tid": 7, "ts": 1716454223919490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857709, "dur": 10,
    "args": {
      "External id": 160120,
      "cbid": 211, "correlation": 160120
    }
  },
  {
    "ph": "s", "id": 160120, "pid": 76337, "tid": -914061504, "ts": 1716454223857709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223919516, "dur": 24,
    "args": {
      "External id": 160140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160140,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 160140, "pid": 5, "tid": 7, "ts": 1716454223919516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857780, "dur": 12,
    "args": {
      "External id": 160140,
      "cbid": 211, "correlation": 160140
    }
  },
  {
    "ph": "s", "id": 160140, "pid": 76337, "tid": -914061504, "ts": 1716454223857780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223919541, "dur": 5,
    "args": {
      "External id": 160152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160152,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 160152, "pid": 5, "tid": 7, "ts": 1716454223919541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857803, "dur": 6,
    "args": {
      "External id": 160152,
      "cbid": 211, "correlation": 160152
    }
  },
  {
    "ph": "s", "id": 160152, "pid": 76337, "tid": -914061504, "ts": 1716454223857803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223919547, "dur": 24,
    "args": {
      "External id": 160155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160155, "pid": 5, "tid": 7, "ts": 1716454223919547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857821, "dur": 6,
    "args": {
      "External id": 160155,
      "cbid": 211, "correlation": 160155
    }
  },
  {
    "ph": "s", "id": 160155, "pid": 76337, "tid": -914061504, "ts": 1716454223857821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223919572, "dur": 17,
    "args": {
      "External id": 160164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160164,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160164, "pid": 5, "tid": 7, "ts": 1716454223919572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857860, "dur": 9,
    "args": {
      "External id": 160164,
      "cbid": 211, "correlation": 160164
    }
  },
  {
    "ph": "s", "id": 160164, "pid": 76337, "tid": -914061504, "ts": 1716454223857860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857911, "dur": 0,
    "args": {
      "External id": 160174,
      "cbid": 317, "correlation": 160174
    }
  },
  {
    "ph": "f", "id": 160174, "pid": 76337, "tid": -914061504, "ts": 1716454223857911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857912, "dur": 0,
    "args": {
      "External id": 160175,
      "cbid": 203, "correlation": 160175
    }
  },
  {
    "ph": "f", "id": 160175, "pid": 76337, "tid": -914061504, "ts": 1716454223857912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857912, "dur": 0,
    "args": {
      "External id": 160176,
      "cbid": 205, "correlation": 160176
    }
  },
  {
    "ph": "f", "id": 160176, "pid": 76337, "tid": -914061504, "ts": 1716454223857912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223919591, "dur": 18,
    "args": {
      "External id": 160180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160180,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160180, "pid": 5, "tid": 7, "ts": 1716454223919591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857927, "dur": 11,
    "args": {
      "External id": 160180,
      "cbid": 211, "correlation": 160180
    }
  },
  {
    "ph": "s", "id": 160180, "pid": 76337, "tid": -914061504, "ts": 1716454223857927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223919610, "dur": 241,
    "args": {
      "External id": 160182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160182,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160182, "pid": 5, "tid": 7, "ts": 1716454223919610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857940, "dur": 5,
    "args": {
      "External id": 160182,
      "cbid": 211, "correlation": 160182
    }
  },
  {
    "ph": "s", "id": 160182, "pid": 76337, "tid": -914061504, "ts": 1716454223857940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223919852, "dur": 1,
    "args": {
      "External id": 160184,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 160184,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 160184, "pid": 5, "tid": 7, "ts": 1716454223919852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857952, "dur": 8,
    "args": {
      "External id": 160184,
      "cbid": 51, "correlation": 160184
    }
  },
  {
    "ph": "s", "id": 160184, "pid": 76337, "tid": -914061504, "ts": 1716454223857952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223919856, "dur": 810,
    "args": {
      "External id": 160185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160185,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160185, "pid": 5, "tid": 7, "ts": 1716454223919856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857961, "dur": 6,
    "args": {
      "External id": 160185,
      "cbid": 211, "correlation": 160185
    }
  },
  {
    "ph": "s", "id": 160185, "pid": 76337, "tid": -914061504, "ts": 1716454223857961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223920667, "dur": 14,
    "args": {
      "External id": 160187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160187,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160187, "pid": 5, "tid": 7, "ts": 1716454223920667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223857972, "dur": 14,
    "args": {
      "External id": 160187,
      "cbid": 211, "correlation": 160187
    }
  },
  {
    "ph": "s", "id": 160187, "pid": 76337, "tid": -914061504, "ts": 1716454223857972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223920682, "dur": 15,
    "args": {
      "External id": 160193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160193,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160193, "pid": 5, "tid": 7, "ts": 1716454223920682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858009, "dur": 9,
    "args": {
      "External id": 160193,
      "cbid": 211, "correlation": 160193
    }
  },
  {
    "ph": "s", "id": 160193, "pid": 76337, "tid": -914061504, "ts": 1716454223858009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223920699, "dur": 3,
    "args": {
      "External id": 160201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160201,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 160201, "pid": 5, "tid": 7, "ts": 1716454223920699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858054, "dur": 9,
    "args": {
      "External id": 160201,
      "cbid": 211, "correlation": 160201
    }
  },
  {
    "ph": "s", "id": 160201, "pid": 76337, "tid": -914061504, "ts": 1716454223858054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858118, "dur": 1,
    "args": {
      "External id": 160217,
      "cbid": 251, "correlation": 160217
    }
  },
  {
    "ph": "f", "id": 160217, "pid": 76337, "tid": -914061504, "ts": 1716454223858118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858123, "dur": 0,
    "args": {
      "External id": 160219,
      "cbid": 251, "correlation": 160219
    }
  },
  {
    "ph": "f", "id": 160219, "pid": 76337, "tid": -914061504, "ts": 1716454223858123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223920704, "dur": 14,
    "args": {
      "External id": 160220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160220,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160220, "pid": 5, "tid": 7, "ts": 1716454223920704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858125, "dur": 11,
    "args": {
      "External id": 160220,
      "cbid": 211, "correlation": 160220
    }
  },
  {
    "ph": "s", "id": 160220, "pid": 76337, "tid": -914061504, "ts": 1716454223858125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223920719, "dur": 5,
    "args": {
      "External id": 160222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160222,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160222, "pid": 5, "tid": 7, "ts": 1716454223920719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858138, "dur": 6,
    "args": {
      "External id": 160222,
      "cbid": 211, "correlation": 160222
    }
  },
  {
    "ph": "s", "id": 160222, "pid": 76337, "tid": -914061504, "ts": 1716454223858138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223920725, "dur": 16,
    "args": {
      "External id": 160232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160232,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160232, "pid": 5, "tid": 7, "ts": 1716454223920725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858197, "dur": 12,
    "args": {
      "External id": 160232,
      "cbid": 211, "correlation": 160232
    }
  },
  {
    "ph": "s", "id": 160232, "pid": 76337, "tid": -914061504, "ts": 1716454223858197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223920743, "dur": 17,
    "args": {
      "External id": 160252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160252,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 160252, "pid": 5, "tid": 7, "ts": 1716454223920743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858262, "dur": 10,
    "args": {
      "External id": 160252,
      "cbid": 211, "correlation": 160252
    }
  },
  {
    "ph": "s", "id": 160252, "pid": 76337, "tid": -914061504, "ts": 1716454223858262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223920762, "dur": 4,
    "args": {
      "External id": 160264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160264,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 160264, "pid": 5, "tid": 7, "ts": 1716454223920762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858282, "dur": 6,
    "args": {
      "External id": 160264,
      "cbid": 211, "correlation": 160264
    }
  },
  {
    "ph": "s", "id": 160264, "pid": 76337, "tid": -914061504, "ts": 1716454223858282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223920767, "dur": 17,
    "args": {
      "External id": 160267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160267,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160267, "pid": 5, "tid": 7, "ts": 1716454223920767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858301, "dur": 6,
    "args": {
      "External id": 160267,
      "cbid": 211, "correlation": 160267
    }
  },
  {
    "ph": "s", "id": 160267, "pid": 76337, "tid": -914061504, "ts": 1716454223858301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223920785, "dur": 10,
    "args": {
      "External id": 160276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160276,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160276, "pid": 5, "tid": 7, "ts": 1716454223920785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858341, "dur": 11,
    "args": {
      "External id": 160276,
      "cbid": 211, "correlation": 160276
    }
  },
  {
    "ph": "s", "id": 160276, "pid": 76337, "tid": -914061504, "ts": 1716454223858341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858404, "dur": 0,
    "args": {
      "External id": 160286,
      "cbid": 317, "correlation": 160286
    }
  },
  {
    "ph": "f", "id": 160286, "pid": 76337, "tid": -914061504, "ts": 1716454223858404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858405, "dur": 0,
    "args": {
      "External id": 160287,
      "cbid": 203, "correlation": 160287
    }
  },
  {
    "ph": "f", "id": 160287, "pid": 76337, "tid": -914061504, "ts": 1716454223858405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858405, "dur": 0,
    "args": {
      "External id": 160288,
      "cbid": 205, "correlation": 160288
    }
  },
  {
    "ph": "f", "id": 160288, "pid": 76337, "tid": -914061504, "ts": 1716454223858405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223920797, "dur": 12,
    "args": {
      "External id": 160292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160292,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160292, "pid": 5, "tid": 7, "ts": 1716454223920797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858419, "dur": 12,
    "args": {
      "External id": 160292,
      "cbid": 211, "correlation": 160292
    }
  },
  {
    "ph": "s", "id": 160292, "pid": 76337, "tid": -914061504, "ts": 1716454223858419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223920810, "dur": 162,
    "args": {
      "External id": 160294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160294,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160294, "pid": 5, "tid": 7, "ts": 1716454223920810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858433, "dur": 5,
    "args": {
      "External id": 160294,
      "cbid": 211, "correlation": 160294
    }
  },
  {
    "ph": "s", "id": 160294, "pid": 76337, "tid": -914061504, "ts": 1716454223858433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223920974, "dur": 1,
    "args": {
      "External id": 160296,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 160296,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 160296, "pid": 5, "tid": 7, "ts": 1716454223920974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858444, "dur": 6,
    "args": {
      "External id": 160296,
      "cbid": 51, "correlation": 160296
    }
  },
  {
    "ph": "s", "id": 160296, "pid": 76337, "tid": -914061504, "ts": 1716454223858444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223920978, "dur": 646,
    "args": {
      "External id": 160297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160297,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160297, "pid": 5, "tid": 7, "ts": 1716454223920978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858452, "dur": 6,
    "args": {
      "External id": 160297,
      "cbid": 211, "correlation": 160297
    }
  },
  {
    "ph": "s", "id": 160297, "pid": 76337, "tid": -914061504, "ts": 1716454223858452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223921626, "dur": 12,
    "args": {
      "External id": 160299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160299,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160299, "pid": 5, "tid": 7, "ts": 1716454223921626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858463, "dur": 5,
    "args": {
      "External id": 160299,
      "cbid": 211, "correlation": 160299
    }
  },
  {
    "ph": "s", "id": 160299, "pid": 76337, "tid": -914061504, "ts": 1716454223858463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223921639, "dur": 15,
    "args": {
      "External id": 160305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160305,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160305, "pid": 5, "tid": 7, "ts": 1716454223921639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858491, "dur": 9,
    "args": {
      "External id": 160305,
      "cbid": 211, "correlation": 160305
    }
  },
  {
    "ph": "s", "id": 160305, "pid": 76337, "tid": -914061504, "ts": 1716454223858491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858550, "dur": 0,
    "args": {
      "External id": 160315,
      "cbid": 317, "correlation": 160315
    }
  },
  {
    "ph": "f", "id": 160315, "pid": 76337, "tid": -914061504, "ts": 1716454223858550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858551, "dur": 0,
    "args": {
      "External id": 160316,
      "cbid": 203, "correlation": 160316
    }
  },
  {
    "ph": "f", "id": 160316, "pid": 76337, "tid": -914061504, "ts": 1716454223858551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858552, "dur": 0,
    "args": {
      "External id": 160317,
      "cbid": 205, "correlation": 160317
    }
  },
  {
    "ph": "f", "id": 160317, "pid": 76337, "tid": -914061504, "ts": 1716454223858552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223921655, "dur": 17,
    "args": {
      "External id": 160321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160321,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160321, "pid": 5, "tid": 7, "ts": 1716454223921655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858566, "dur": 12,
    "args": {
      "External id": 160321,
      "cbid": 211, "correlation": 160321
    }
  },
  {
    "ph": "s", "id": 160321, "pid": 76337, "tid": -914061504, "ts": 1716454223858566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223921673, "dur": 4,
    "args": {
      "External id": 160323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160323, "pid": 5, "tid": 7, "ts": 1716454223921673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858582, "dur": 6,
    "args": {
      "External id": 160323,
      "cbid": 211, "correlation": 160323
    }
  },
  {
    "ph": "s", "id": 160323, "pid": 76337, "tid": -914061504, "ts": 1716454223858582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858591, "dur": 0,
    "args": {
      "External id": 160324,
      "cbid": 51, "correlation": 160324
    }
  },
  {
    "ph": "s", "id": 160324, "pid": 76337, "tid": -914061504, "ts": 1716454223858591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223921679, "dur": 131,
    "args": {
      "External id": 160325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160325,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 160325, "pid": 5, "tid": 7, "ts": 1716454223921679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858592, "dur": 5,
    "args": {
      "External id": 160325,
      "cbid": 211, "correlation": 160325
    }
  },
  {
    "ph": "s", "id": 160325, "pid": 76337, "tid": -914061504, "ts": 1716454223858592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223921811, "dur": 15,
    "args": {
      "External id": 160330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160330,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160330, "pid": 5, "tid": 7, "ts": 1716454223921811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858618, "dur": 8,
    "args": {
      "External id": 160330,
      "cbid": 211, "correlation": 160330
    }
  },
  {
    "ph": "s", "id": 160330, "pid": 76337, "tid": -914061504, "ts": 1716454223858618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223921828, "dur": 11,
    "args": {
      "External id": 160338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160338,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160338, "pid": 5, "tid": 7, "ts": 1716454223921828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858647, "dur": 8,
    "args": {
      "External id": 160338,
      "cbid": 211, "correlation": 160338
    }
  },
  {
    "ph": "s", "id": 160338, "pid": 76337, "tid": -914061504, "ts": 1716454223858647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223921841, "dur": 10,
    "args": {
      "External id": 160346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160346,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160346, "pid": 5, "tid": 7, "ts": 1716454223921841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858676, "dur": 8,
    "args": {
      "External id": 160346,
      "cbid": 211, "correlation": 160346
    }
  },
  {
    "ph": "s", "id": 160346, "pid": 76337, "tid": -914061504, "ts": 1716454223858676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223921852, "dur": 18,
    "args": {
      "External id": 160366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160366,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 160366, "pid": 5, "tid": 7, "ts": 1716454223921852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858758, "dur": 12,
    "args": {
      "External id": 160366,
      "cbid": 211, "correlation": 160366
    }
  },
  {
    "ph": "s", "id": 160366, "pid": 76337, "tid": -914061504, "ts": 1716454223858758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223921871, "dur": 5,
    "args": {
      "External id": 160378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160378,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 160378, "pid": 5, "tid": 7, "ts": 1716454223921871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858780, "dur": 6,
    "args": {
      "External id": 160378,
      "cbid": 211, "correlation": 160378
    }
  },
  {
    "ph": "s", "id": 160378, "pid": 76337, "tid": -914061504, "ts": 1716454223858780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223921877, "dur": 17,
    "args": {
      "External id": 160381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160381, "pid": 5, "tid": 7, "ts": 1716454223921877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858798, "dur": 7,
    "args": {
      "External id": 160381,
      "cbid": 211, "correlation": 160381
    }
  },
  {
    "ph": "s", "id": 160381, "pid": 76337, "tid": -914061504, "ts": 1716454223858798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858856, "dur": 0,
    "args": {
      "External id": 160392,
      "cbid": 317, "correlation": 160392
    }
  },
  {
    "ph": "f", "id": 160392, "pid": 76337, "tid": -914061504, "ts": 1716454223858856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858857, "dur": 0,
    "args": {
      "External id": 160393,
      "cbid": 203, "correlation": 160393
    }
  },
  {
    "ph": "f", "id": 160393, "pid": 76337, "tid": -914061504, "ts": 1716454223858857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858858, "dur": 0,
    "args": {
      "External id": 160394,
      "cbid": 205, "correlation": 160394
    }
  },
  {
    "ph": "f", "id": 160394, "pid": 76337, "tid": -914061504, "ts": 1716454223858858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223921896, "dur": 12,
    "args": {
      "External id": 160398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160398,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160398, "pid": 5, "tid": 7, "ts": 1716454223921896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858871, "dur": 11,
    "args": {
      "External id": 160398,
      "cbid": 211, "correlation": 160398
    }
  },
  {
    "ph": "s", "id": 160398, "pid": 76337, "tid": -914061504, "ts": 1716454223858871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223921909, "dur": 3,
    "args": {
      "External id": 160400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160400,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160400, "pid": 5, "tid": 7, "ts": 1716454223921909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858887, "dur": 7,
    "args": {
      "External id": 160400,
      "cbid": 211, "correlation": 160400
    }
  },
  {
    "ph": "s", "id": 160400, "pid": 76337, "tid": -914061504, "ts": 1716454223858887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858896, "dur": 0,
    "args": {
      "External id": 160401,
      "cbid": 51, "correlation": 160401
    }
  },
  {
    "ph": "s", "id": 160401, "pid": 76337, "tid": -914061504, "ts": 1716454223858896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223921914, "dur": 91,
    "args": {
      "External id": 160402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160402,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 160402, "pid": 5, "tid": 7, "ts": 1716454223921914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858897, "dur": 5,
    "args": {
      "External id": 160402,
      "cbid": 211, "correlation": 160402
    }
  },
  {
    "ph": "s", "id": 160402, "pid": 76337, "tid": -914061504, "ts": 1716454223858897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223922006, "dur": 16,
    "args": {
      "External id": 160407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160407,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160407, "pid": 5, "tid": 7, "ts": 1716454223922006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223858923, "dur": 9,
    "args": {
      "External id": 160407,
      "cbid": 211, "correlation": 160407
    }
  },
  {
    "ph": "s", "id": 160407, "pid": 76337, "tid": -914061504, "ts": 1716454223858923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223922023, "dur": 84,
    "args": {
      "External id": 160416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160416,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160416, "pid": 5, "tid": 7, "ts": 1716454223922023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859014, "dur": 14,
    "args": {
      "External id": 160416,
      "cbid": 211, "correlation": 160416
    }
  },
  {
    "ph": "s", "id": 160416, "pid": 76337, "tid": -914061504, "ts": 1716454223859014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223922109, "dur": 30,
    "args": {
      "External id": 160438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160438,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160438, "pid": 5, "tid": 7, "ts": 1716454223922109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859071, "dur": 10,
    "args": {
      "External id": 160438,
      "cbid": 211, "correlation": 160438
    }
  },
  {
    "ph": "s", "id": 160438, "pid": 76337, "tid": -914061504, "ts": 1716454223859071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859159, "dur": 1,
    "args": {
      "External id": 160449,
      "cbid": 251, "correlation": 160449
    }
  },
  {
    "ph": "f", "id": 160449, "pid": 76337, "tid": -914061504, "ts": 1716454223859159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223922140, "dur": 164,
    "args": {
      "External id": 160450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160450,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160450, "pid": 5, "tid": 7, "ts": 1716454223922140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859164, "dur": 13,
    "args": {
      "External id": 160450,
      "cbid": 211, "correlation": 160450
    }
  },
  {
    "ph": "s", "id": 160450, "pid": 76337, "tid": -914061504, "ts": 1716454223859164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859234, "dur": 1,
    "args": {
      "External id": 160461,
      "cbid": 251, "correlation": 160461
    }
  },
  {
    "ph": "f", "id": 160461, "pid": 76337, "tid": -914061504, "ts": 1716454223859234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223922305, "dur": 157,
    "args": {
      "External id": 160462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160462,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160462, "pid": 5, "tid": 7, "ts": 1716454223922305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859238, "dur": 12,
    "args": {
      "External id": 160462,
      "cbid": 211, "correlation": 160462
    }
  },
  {
    "ph": "s", "id": 160462, "pid": 76337, "tid": -914061504, "ts": 1716454223859238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859304, "dur": 1,
    "args": {
      "External id": 160473,
      "cbid": 251, "correlation": 160473
    }
  },
  {
    "ph": "f", "id": 160473, "pid": 76337, "tid": -914061504, "ts": 1716454223859304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223922463, "dur": 160,
    "args": {
      "External id": 160474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160474,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160474, "pid": 5, "tid": 7, "ts": 1716454223922463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859308, "dur": 11,
    "args": {
      "External id": 160474,
      "cbid": 211, "correlation": 160474
    }
  },
  {
    "ph": "s", "id": 160474, "pid": 76337, "tid": -914061504, "ts": 1716454223859308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223922624, "dur": 338,
    "args": {
      "External id": 160499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160499,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160499, "pid": 5, "tid": 7, "ts": 1716454223922624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859392, "dur": 12,
    "args": {
      "External id": 160499,
      "cbid": 211, "correlation": 160499
    }
  },
  {
    "ph": "s", "id": 160499, "pid": 76337, "tid": -914061504, "ts": 1716454223859392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859491, "dur": 1,
    "args": {
      "External id": 160517,
      "cbid": 251, "correlation": 160517
    }
  },
  {
    "ph": "f", "id": 160517, "pid": 76337, "tid": -914061504, "ts": 1716454223859491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223922964, "dur": 166,
    "args": {
      "External id": 160519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160519,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160519, "pid": 5, "tid": 7, "ts": 1716454223922964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859497, "dur": 13,
    "args": {
      "External id": 160519,
      "cbid": 211, "correlation": 160519
    }
  },
  {
    "ph": "s", "id": 160519, "pid": 76337, "tid": -914061504, "ts": 1716454223859497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223923132, "dur": 19,
    "args": {
      "External id": 160527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160527,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160527, "pid": 5, "tid": 7, "ts": 1716454223923132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859568, "dur": 12,
    "args": {
      "External id": 160527,
      "cbid": 211, "correlation": 160527
    }
  },
  {
    "ph": "s", "id": 160527, "pid": 76337, "tid": -914061504, "ts": 1716454223859568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223923152, "dur": 28,
    "args": {
      "External id": 160535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160535,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160535, "pid": 5, "tid": 7, "ts": 1716454223923152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859607, "dur": 8,
    "args": {
      "External id": 160535,
      "cbid": 211, "correlation": 160535
    }
  },
  {
    "ph": "s", "id": 160535, "pid": 76337, "tid": -914061504, "ts": 1716454223859607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223923181, "dur": 18,
    "args": {
      "External id": 160546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160546,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160546, "pid": 5, "tid": 7, "ts": 1716454223923181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859677, "dur": 13,
    "args": {
      "External id": 160546,
      "cbid": 211, "correlation": 160546
    }
  },
  {
    "ph": "s", "id": 160546, "pid": 76337, "tid": -914061504, "ts": 1716454223859677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223923200, "dur": 16,
    "args": {
      "External id": 160568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160568,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160568, "pid": 5, "tid": 7, "ts": 1716454223923200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859709, "dur": 7,
    "args": {
      "External id": 160568,
      "cbid": 211, "correlation": 160568
    }
  },
  {
    "ph": "s", "id": 160568, "pid": 76337, "tid": -914061504, "ts": 1716454223859709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859794, "dur": 1,
    "args": {
      "External id": 160579,
      "cbid": 251, "correlation": 160579
    }
  },
  {
    "ph": "f", "id": 160579, "pid": 76337, "tid": -914061504, "ts": 1716454223859794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223923218, "dur": 89,
    "args": {
      "External id": 160580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160580,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160580, "pid": 5, "tid": 7, "ts": 1716454223923218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859799, "dur": 12,
    "args": {
      "External id": 160580,
      "cbid": 211, "correlation": 160580
    }
  },
  {
    "ph": "s", "id": 160580, "pid": 76337, "tid": -914061504, "ts": 1716454223859799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859868, "dur": 1,
    "args": {
      "External id": 160591,
      "cbid": 251, "correlation": 160591
    }
  },
  {
    "ph": "f", "id": 160591, "pid": 76337, "tid": -914061504, "ts": 1716454223859868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859871, "dur": 0,
    "args": {
      "External id": 160592,
      "cbid": 251, "correlation": 160592
    }
  },
  {
    "ph": "f", "id": 160592, "pid": 76337, "tid": -914061504, "ts": 1716454223859871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223923309, "dur": 12,
    "args": {
      "External id": 160593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160593,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160593, "pid": 5, "tid": 7, "ts": 1716454223923309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859873, "dur": 12,
    "args": {
      "External id": 160593,
      "cbid": 211, "correlation": 160593
    }
  },
  {
    "ph": "s", "id": 160593, "pid": 76337, "tid": -914061504, "ts": 1716454223859873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223923322, "dur": 5,
    "args": {
      "External id": 160595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160595,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160595, "pid": 5, "tid": 7, "ts": 1716454223923322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859887, "dur": 5,
    "args": {
      "External id": 160595,
      "cbid": 211, "correlation": 160595
    }
  },
  {
    "ph": "s", "id": 160595, "pid": 76337, "tid": -914061504, "ts": 1716454223859887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859943, "dur": 1,
    "args": {
      "External id": 160606,
      "cbid": 251, "correlation": 160606
    }
  },
  {
    "ph": "f", "id": 160606, "pid": 76337, "tid": -914061504, "ts": 1716454223859943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859947, "dur": 0,
    "args": {
      "External id": 160607,
      "cbid": 251, "correlation": 160607
    }
  },
  {
    "ph": "f", "id": 160607, "pid": 76337, "tid": -914061504, "ts": 1716454223859947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223923329, "dur": 9,
    "args": {
      "External id": 160608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160608,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160608, "pid": 5, "tid": 7, "ts": 1716454223923329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859949, "dur": 11,
    "args": {
      "External id": 160608,
      "cbid": 211, "correlation": 160608
    }
  },
  {
    "ph": "s", "id": 160608, "pid": 76337, "tid": -914061504, "ts": 1716454223859949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223923339, "dur": 3,
    "args": {
      "External id": 160610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160610,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160610, "pid": 5, "tid": 7, "ts": 1716454223923339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223859962, "dur": 5,
    "args": {
      "External id": 160610,
      "cbid": 211, "correlation": 160610
    }
  },
  {
    "ph": "s", "id": 160610, "pid": 76337, "tid": -914061504, "ts": 1716454223859962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223923344, "dur": 56,
    "args": {
      "External id": 160635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160635,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160635, "pid": 5, "tid": 7, "ts": 1716454223923344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860049, "dur": 14,
    "args": {
      "External id": 160635,
      "cbid": 211, "correlation": 160635
    }
  },
  {
    "ph": "s", "id": 160635, "pid": 76337, "tid": -914061504, "ts": 1716454223860049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860149, "dur": 1,
    "args": {
      "External id": 160653,
      "cbid": 251, "correlation": 160653
    }
  },
  {
    "ph": "f", "id": 160653, "pid": 76337, "tid": -914061504, "ts": 1716454223860149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223923401, "dur": 92,
    "args": {
      "External id": 160655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160655,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160655, "pid": 5, "tid": 7, "ts": 1716454223923401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860156, "dur": 13,
    "args": {
      "External id": 160655,
      "cbid": 211, "correlation": 160655
    }
  },
  {
    "ph": "s", "id": 160655, "pid": 76337, "tid": -914061504, "ts": 1716454223860156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223923494, "dur": 10,
    "args": {
      "External id": 160663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160663,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160663, "pid": 5, "tid": 7, "ts": 1716454223923494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860225, "dur": 12,
    "args": {
      "External id": 160663,
      "cbid": 211, "correlation": 160663
    }
  },
  {
    "ph": "s", "id": 160663, "pid": 76337, "tid": -914061504, "ts": 1716454223860225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223923505, "dur": 21,
    "args": {
      "External id": 160671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160671, "pid": 5, "tid": 7, "ts": 1716454223923505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860267, "dur": 9,
    "args": {
      "External id": 160671,
      "cbid": 211, "correlation": 160671
    }
  },
  {
    "ph": "s", "id": 160671, "pid": 76337, "tid": -914061504, "ts": 1716454223860267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223923528, "dur": 18,
    "args": {
      "External id": 160693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160693,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160693, "pid": 5, "tid": 7, "ts": 1716454223923528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860319, "dur": 10,
    "args": {
      "External id": 160693,
      "cbid": 211, "correlation": 160693
    }
  },
  {
    "ph": "s", "id": 160693, "pid": 76337, "tid": -914061504, "ts": 1716454223860319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860406, "dur": 1,
    "args": {
      "External id": 160709,
      "cbid": 251, "correlation": 160709
    }
  },
  {
    "ph": "f", "id": 160709, "pid": 76337, "tid": -914061504, "ts": 1716454223860406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860411, "dur": 0,
    "args": {
      "External id": 160711,
      "cbid": 251, "correlation": 160711
    }
  },
  {
    "ph": "f", "id": 160711, "pid": 76337, "tid": -914061504, "ts": 1716454223860411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223923547, "dur": 498,
    "args": {
      "External id": 160712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160712,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160712, "pid": 5, "tid": 7, "ts": 1716454223923547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860413, "dur": 12,
    "args": {
      "External id": 160712,
      "cbid": 211, "correlation": 160712
    }
  },
  {
    "ph": "s", "id": 160712, "pid": 76337, "tid": -914061504, "ts": 1716454223860413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223924047, "dur": 66,
    "args": {
      "External id": 160720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160720,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160720, "pid": 5, "tid": 7, "ts": 1716454223924047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860478, "dur": 12,
    "args": {
      "External id": 160720,
      "cbid": 211, "correlation": 160720
    }
  },
  {
    "ph": "s", "id": 160720, "pid": 76337, "tid": -914061504, "ts": 1716454223860478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223924114, "dur": 67,
    "args": {
      "External id": 160728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160728,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160728, "pid": 5, "tid": 7, "ts": 1716454223924114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860509, "dur": 8,
    "args": {
      "External id": 160728,
      "cbid": 211, "correlation": 160728
    }
  },
  {
    "ph": "s", "id": 160728, "pid": 76337, "tid": -914061504, "ts": 1716454223860509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860587, "dur": 1,
    "args": {
      "External id": 160744,
      "cbid": 251, "correlation": 160744
    }
  },
  {
    "ph": "f", "id": 160744, "pid": 76337, "tid": -914061504, "ts": 1716454223860587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454223924183, "dur": 1,
    "args": {
      "External id": 160746,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 160746,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 160746, "pid": 5, "tid": 7, "ts": 1716454223924183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860592, "dur": 9,
    "args": {
      "External id": 160746,
      "cbid": 51, "correlation": 160746
    }
  },
  {
    "ph": "s", "id": 160746, "pid": 76337, "tid": -914061504, "ts": 1716454223860592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223924187, "dur": 269,
    "args": {
      "External id": 160747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160747,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160747, "pid": 5, "tid": 7, "ts": 1716454223924187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860604, "dur": 11,
    "args": {
      "External id": 160747,
      "cbid": 211, "correlation": 160747
    }
  },
  {
    "ph": "s", "id": 160747, "pid": 76337, "tid": -914061504, "ts": 1716454223860604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223924458, "dur": 13,
    "args": {
      "External id": 160755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160755,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160755, "pid": 5, "tid": 7, "ts": 1716454223924458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860645, "dur": 12,
    "args": {
      "External id": 160755,
      "cbid": 211, "correlation": 160755
    }
  },
  {
    "ph": "s", "id": 160755, "pid": 76337, "tid": -914061504, "ts": 1716454223860645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223924473, "dur": 38,
    "args": {
      "External id": 160766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160766,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160766, "pid": 5, "tid": 7, "ts": 1716454223924473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860715, "dur": 12,
    "args": {
      "External id": 160766,
      "cbid": 211, "correlation": 160766
    }
  },
  {
    "ph": "s", "id": 160766, "pid": 76337, "tid": -914061504, "ts": 1716454223860715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860779, "dur": 0,
    "args": {
      "External id": 160778,
      "cbid": 317, "correlation": 160778
    }
  },
  {
    "ph": "f", "id": 160778, "pid": 76337, "tid": -914061504, "ts": 1716454223860779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860779, "dur": 0,
    "args": {
      "External id": 160779,
      "cbid": 203, "correlation": 160779
    }
  },
  {
    "ph": "f", "id": 160779, "pid": 76337, "tid": -914061504, "ts": 1716454223860779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860780, "dur": 0,
    "args": {
      "External id": 160780,
      "cbid": 205, "correlation": 160780
    }
  },
  {
    "ph": "f", "id": 160780, "pid": 76337, "tid": -914061504, "ts": 1716454223860780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223924511, "dur": 13,
    "args": {
      "External id": 160784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160784,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160784, "pid": 5, "tid": 7, "ts": 1716454223924511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860796, "dur": 13,
    "args": {
      "External id": 160784,
      "cbid": 211, "correlation": 160784
    }
  },
  {
    "ph": "s", "id": 160784, "pid": 76337, "tid": -914061504, "ts": 1716454223860796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223924526, "dur": 4,
    "args": {
      "External id": 160786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 160786, "pid": 5, "tid": 7, "ts": 1716454223924526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860813, "dur": 6,
    "args": {
      "External id": 160786,
      "cbid": 211, "correlation": 160786
    }
  },
  {
    "ph": "s", "id": 160786, "pid": 76337, "tid": -914061504, "ts": 1716454223860813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860821, "dur": 0,
    "args": {
      "External id": 160787,
      "cbid": 51, "correlation": 160787
    }
  },
  {
    "ph": "s", "id": 160787, "pid": 76337, "tid": -914061504, "ts": 1716454223860821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223924531, "dur": 97,
    "args": {
      "External id": 160788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160788,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 160788, "pid": 5, "tid": 7, "ts": 1716454223924531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860822, "dur": 5,
    "args": {
      "External id": 160788,
      "cbid": 211, "correlation": 160788
    }
  },
  {
    "ph": "s", "id": 160788, "pid": 76337, "tid": -914061504, "ts": 1716454223860822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223924629, "dur": 17,
    "args": {
      "External id": 160793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160793,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160793, "pid": 5, "tid": 7, "ts": 1716454223924629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860849, "dur": 9,
    "args": {
      "External id": 160793,
      "cbid": 211, "correlation": 160793
    }
  },
  {
    "ph": "s", "id": 160793, "pid": 76337, "tid": -914061504, "ts": 1716454223860849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223924648, "dur": 12,
    "args": {
      "External id": 160801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160801,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160801, "pid": 5, "tid": 7, "ts": 1716454223924648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860881, "dur": 8,
    "args": {
      "External id": 160801,
      "cbid": 211, "correlation": 160801
    }
  },
  {
    "ph": "s", "id": 160801, "pid": 76337, "tid": -914061504, "ts": 1716454223860881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223924661, "dur": 57,
    "args": {
      "External id": 160812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160812,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160812, "pid": 5, "tid": 7, "ts": 1716454223924661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223860945, "dur": 13,
    "args": {
      "External id": 160812,
      "cbid": 211, "correlation": 160812
    }
  },
  {
    "ph": "s", "id": 160812, "pid": 76337, "tid": -914061504, "ts": 1716454223860945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861013, "dur": 0,
    "args": {
      "External id": 160822,
      "cbid": 317, "correlation": 160822
    }
  },
  {
    "ph": "f", "id": 160822, "pid": 76337, "tid": -914061504, "ts": 1716454223861013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861013, "dur": 0,
    "args": {
      "External id": 160823,
      "cbid": 203, "correlation": 160823
    }
  },
  {
    "ph": "f", "id": 160823, "pid": 76337, "tid": -914061504, "ts": 1716454223861013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861014, "dur": 0,
    "args": {
      "External id": 160824,
      "cbid": 205, "correlation": 160824
    }
  },
  {
    "ph": "f", "id": 160824, "pid": 76337, "tid": -914061504, "ts": 1716454223861014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223924720, "dur": 38,
    "args": {
      "External id": 160828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160828,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160828, "pid": 5, "tid": 7, "ts": 1716454223924720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861030, "dur": 12,
    "args": {
      "External id": 160828,
      "cbid": 211, "correlation": 160828
    }
  },
  {
    "ph": "s", "id": 160828, "pid": 76337, "tid": -914061504, "ts": 1716454223861030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223924760, "dur": 162,
    "args": {
      "External id": 160830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160830,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160830, "pid": 5, "tid": 7, "ts": 1716454223924760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861045, "dur": 5,
    "args": {
      "External id": 160830,
      "cbid": 211, "correlation": 160830
    }
  },
  {
    "ph": "s", "id": 160830, "pid": 76337, "tid": -914061504, "ts": 1716454223861045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223924923, "dur": 1965,
    "args": {
      "External id": 160832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160832,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160832, "pid": 5, "tid": 7, "ts": 1716454223924923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861057, "dur": 9,
    "args": {
      "External id": 160832,
      "cbid": 211, "correlation": 160832
    }
  },
  {
    "ph": "s", "id": 160832, "pid": 76337, "tid": -914061504, "ts": 1716454223861057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223926890, "dur": 39,
    "args": {
      "External id": 160834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160834,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160834, "pid": 5, "tid": 7, "ts": 1716454223926890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861070, "dur": 5,
    "args": {
      "External id": 160834,
      "cbid": 211, "correlation": 160834
    }
  },
  {
    "ph": "s", "id": 160834, "pid": 76337, "tid": -914061504, "ts": 1716454223861070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223926930, "dur": 59,
    "args": {
      "External id": 160840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160840,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160840, "pid": 5, "tid": 7, "ts": 1716454223926930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861098, "dur": 9,
    "args": {
      "External id": 160840,
      "cbid": 211, "correlation": 160840
    }
  },
  {
    "ph": "s", "id": 160840, "pid": 76337, "tid": -914061504, "ts": 1716454223861098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223926990, "dur": 87,
    "args": {
      "External id": 160849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160849,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160849, "pid": 5, "tid": 7, "ts": 1716454223926990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861190, "dur": 13,
    "args": {
      "External id": 160849,
      "cbid": 211, "correlation": 160849
    }
  },
  {
    "ph": "s", "id": 160849, "pid": 76337, "tid": -914061504, "ts": 1716454223861190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223927078, "dur": 73,
    "args": {
      "External id": 160869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160869,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 160869, "pid": 5, "tid": 7, "ts": 1716454223927078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861262, "dur": 11,
    "args": {
      "External id": 160869,
      "cbid": 211, "correlation": 160869
    }
  },
  {
    "ph": "s", "id": 160869, "pid": 76337, "tid": -914061504, "ts": 1716454223861262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223927153, "dur": 5,
    "args": {
      "External id": 160881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160881,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 160881, "pid": 5, "tid": 7, "ts": 1716454223927153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861283, "dur": 7,
    "args": {
      "External id": 160881,
      "cbid": 211, "correlation": 160881
    }
  },
  {
    "ph": "s", "id": 160881, "pid": 76337, "tid": -914061504, "ts": 1716454223861283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223927159, "dur": 82,
    "args": {
      "External id": 160884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160884,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160884, "pid": 5, "tid": 7, "ts": 1716454223927159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861302, "dur": 8,
    "args": {
      "External id": 160884,
      "cbid": 211, "correlation": 160884
    }
  },
  {
    "ph": "s", "id": 160884, "pid": 76337, "tid": -914061504, "ts": 1716454223861302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223927243, "dur": 53,
    "args": {
      "External id": 160893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160893,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160893, "pid": 5, "tid": 7, "ts": 1716454223927243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861342, "dur": 10,
    "args": {
      "External id": 160893,
      "cbid": 211, "correlation": 160893
    }
  },
  {
    "ph": "s", "id": 160893, "pid": 76337, "tid": -914061504, "ts": 1716454223861342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861393, "dur": 0,
    "args": {
      "External id": 160903,
      "cbid": 317, "correlation": 160903
    }
  },
  {
    "ph": "f", "id": 160903, "pid": 76337, "tid": -914061504, "ts": 1716454223861393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861394, "dur": 0,
    "args": {
      "External id": 160904,
      "cbid": 203, "correlation": 160904
    }
  },
  {
    "ph": "f", "id": 160904, "pid": 76337, "tid": -914061504, "ts": 1716454223861394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861394, "dur": 0,
    "args": {
      "External id": 160905,
      "cbid": 205, "correlation": 160905
    }
  },
  {
    "ph": "f", "id": 160905, "pid": 76337, "tid": -914061504, "ts": 1716454223861394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223927296, "dur": 57,
    "args": {
      "External id": 160909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160909,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160909, "pid": 5, "tid": 7, "ts": 1716454223927296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861411, "dur": 11,
    "args": {
      "External id": 160909,
      "cbid": 211, "correlation": 160909
    }
  },
  {
    "ph": "s", "id": 160909, "pid": 76337, "tid": -914061504, "ts": 1716454223861411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223927355, "dur": 122,
    "args": {
      "External id": 160911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160911,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160911, "pid": 5, "tid": 7, "ts": 1716454223927355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861425, "dur": 5,
    "args": {
      "External id": 160911,
      "cbid": 211, "correlation": 160911
    }
  },
  {
    "ph": "s", "id": 160911, "pid": 76337, "tid": -914061504, "ts": 1716454223861425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223927478, "dur": 1894,
    "args": {
      "External id": 160913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160913,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 160913, "pid": 5, "tid": 7, "ts": 1716454223927478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861436, "dur": 6,
    "args": {
      "External id": 160913,
      "cbid": 211, "correlation": 160913
    }
  },
  {
    "ph": "s", "id": 160913, "pid": 76337, "tid": -914061504, "ts": 1716454223861436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223929373, "dur": 22,
    "args": {
      "External id": 160915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160915, "pid": 5, "tid": 7, "ts": 1716454223929373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861446, "dur": 5,
    "args": {
      "External id": 160915,
      "cbid": 211, "correlation": 160915
    }
  },
  {
    "ph": "s", "id": 160915, "pid": 76337, "tid": -914061504, "ts": 1716454223861446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223929396, "dur": 32,
    "args": {
      "External id": 160921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160921, "pid": 5, "tid": 7, "ts": 1716454223929396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861474, "dur": 9,
    "args": {
      "External id": 160921,
      "cbid": 211, "correlation": 160921
    }
  },
  {
    "ph": "s", "id": 160921, "pid": 76337, "tid": -914061504, "ts": 1716454223861474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223929429, "dur": 3,
    "args": {
      "External id": 160929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160929,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 160929, "pid": 5, "tid": 7, "ts": 1716454223929429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861518, "dur": 10,
    "args": {
      "External id": 160929,
      "cbid": 211, "correlation": 160929
    }
  },
  {
    "ph": "s", "id": 160929, "pid": 76337, "tid": -914061504, "ts": 1716454223861518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861584, "dur": 1,
    "args": {
      "External id": 160945,
      "cbid": 251, "correlation": 160945
    }
  },
  {
    "ph": "f", "id": 160945, "pid": 76337, "tid": -914061504, "ts": 1716454223861584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861589, "dur": 0,
    "args": {
      "External id": 160947,
      "cbid": 251, "correlation": 160947
    }
  },
  {
    "ph": "f", "id": 160947, "pid": 76337, "tid": -914061504, "ts": 1716454223861589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223929434, "dur": 12,
    "args": {
      "External id": 160948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160948,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 160948, "pid": 5, "tid": 7, "ts": 1716454223929434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861591, "dur": 11,
    "args": {
      "External id": 160948,
      "cbid": 211, "correlation": 160948
    }
  },
  {
    "ph": "s", "id": 160948, "pid": 76337, "tid": -914061504, "ts": 1716454223861591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223929447, "dur": 5,
    "args": {
      "External id": 160950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160950,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 160950, "pid": 5, "tid": 7, "ts": 1716454223929447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861604, "dur": 6,
    "args": {
      "External id": 160950,
      "cbid": 211, "correlation": 160950
    }
  },
  {
    "ph": "s", "id": 160950, "pid": 76337, "tid": -914061504, "ts": 1716454223861604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223929454, "dur": 29,
    "args": {
      "External id": 160960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160960, "pid": 5, "tid": 7, "ts": 1716454223929454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861662, "dur": 12,
    "args": {
      "External id": 160960,
      "cbid": 211, "correlation": 160960
    }
  },
  {
    "ph": "s", "id": 160960, "pid": 76337, "tid": -914061504, "ts": 1716454223861662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223929484, "dur": 31,
    "args": {
      "External id": 160980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160980,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 160980, "pid": 5, "tid": 7, "ts": 1716454223929484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861727, "dur": 11,
    "args": {
      "External id": 160980,
      "cbid": 211, "correlation": 160980
    }
  },
  {
    "ph": "s", "id": 160980, "pid": 76337, "tid": -914061504, "ts": 1716454223861727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223929516, "dur": 4,
    "args": {
      "External id": 160992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160992,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 160992, "pid": 5, "tid": 7, "ts": 1716454223929516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861748, "dur": 6,
    "args": {
      "External id": 160992,
      "cbid": 211, "correlation": 160992
    }
  },
  {
    "ph": "s", "id": 160992, "pid": 76337, "tid": -914061504, "ts": 1716454223861748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223929522, "dur": 30,
    "args": {
      "External id": 160995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 160995,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 160995, "pid": 5, "tid": 7, "ts": 1716454223929522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861766, "dur": 6,
    "args": {
      "External id": 160995,
      "cbid": 211, "correlation": 160995
    }
  },
  {
    "ph": "s", "id": 160995, "pid": 76337, "tid": -914061504, "ts": 1716454223861766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223929554, "dur": 20,
    "args": {
      "External id": 161004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161004,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161004, "pid": 5, "tid": 7, "ts": 1716454223929554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861805, "dur": 10,
    "args": {
      "External id": 161004,
      "cbid": 211, "correlation": 161004
    }
  },
  {
    "ph": "s", "id": 161004, "pid": 76337, "tid": -914061504, "ts": 1716454223861805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861868, "dur": 0,
    "args": {
      "External id": 161014,
      "cbid": 317, "correlation": 161014
    }
  },
  {
    "ph": "f", "id": 161014, "pid": 76337, "tid": -914061504, "ts": 1716454223861868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861869, "dur": 0,
    "args": {
      "External id": 161015,
      "cbid": 203, "correlation": 161015
    }
  },
  {
    "ph": "f", "id": 161015, "pid": 76337, "tid": -914061504, "ts": 1716454223861869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861869, "dur": 0,
    "args": {
      "External id": 161016,
      "cbid": 205, "correlation": 161016
    }
  },
  {
    "ph": "f", "id": 161016, "pid": 76337, "tid": -914061504, "ts": 1716454223861869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223929575, "dur": 22,
    "args": {
      "External id": 161020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161020, "pid": 5, "tid": 7, "ts": 1716454223929575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861885, "dur": 12,
    "args": {
      "External id": 161020,
      "cbid": 211, "correlation": 161020
    }
  },
  {
    "ph": "s", "id": 161020, "pid": 76337, "tid": -914061504, "ts": 1716454223861885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223929598, "dur": 44,
    "args": {
      "External id": 161022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161022,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161022, "pid": 5, "tid": 7, "ts": 1716454223929598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861899, "dur": 5,
    "args": {
      "External id": 161022,
      "cbid": 211, "correlation": 161022
    }
  },
  {
    "ph": "s", "id": 161022, "pid": 76337, "tid": -914061504, "ts": 1716454223861899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223929643, "dur": 645,
    "args": {
      "External id": 161024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161024,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161024, "pid": 5, "tid": 7, "ts": 1716454223929643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861912, "dur": 6,
    "args": {
      "External id": 161024,
      "cbid": 211, "correlation": 161024
    }
  },
  {
    "ph": "s", "id": 161024, "pid": 76337, "tid": -914061504, "ts": 1716454223861912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223930290, "dur": 21,
    "args": {
      "External id": 161026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161026, "pid": 5, "tid": 7, "ts": 1716454223930290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861921, "dur": 5,
    "args": {
      "External id": 161026,
      "cbid": 211, "correlation": 161026
    }
  },
  {
    "ph": "s", "id": 161026, "pid": 76337, "tid": -914061504, "ts": 1716454223861921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223930312, "dur": 33,
    "args": {
      "External id": 161032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161032, "pid": 5, "tid": 7, "ts": 1716454223930312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223861949, "dur": 9,
    "args": {
      "External id": 161032,
      "cbid": 211, "correlation": 161032
    }
  },
  {
    "ph": "s", "id": 161032, "pid": 76337, "tid": -914061504, "ts": 1716454223861949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862018, "dur": 0,
    "args": {
      "External id": 161042,
      "cbid": 317, "correlation": 161042
    }
  },
  {
    "ph": "f", "id": 161042, "pid": 76337, "tid": -914061504, "ts": 1716454223862018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862018, "dur": 0,
    "args": {
      "External id": 161043,
      "cbid": 203, "correlation": 161043
    }
  },
  {
    "ph": "f", "id": 161043, "pid": 76337, "tid": -914061504, "ts": 1716454223862018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862019, "dur": 0,
    "args": {
      "External id": 161044,
      "cbid": 205, "correlation": 161044
    }
  },
  {
    "ph": "f", "id": 161044, "pid": 76337, "tid": -914061504, "ts": 1716454223862019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223930347, "dur": 57,
    "args": {
      "External id": 161048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161048, "pid": 5, "tid": 7, "ts": 1716454223930347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862031, "dur": 12,
    "args": {
      "External id": 161048,
      "cbid": 211, "correlation": 161048
    }
  },
  {
    "ph": "s", "id": 161048, "pid": 76337, "tid": -914061504, "ts": 1716454223862031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223930405, "dur": 268,
    "args": {
      "External id": 161050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161050,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161050, "pid": 5, "tid": 7, "ts": 1716454223930405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862050, "dur": 9,
    "args": {
      "External id": 161050,
      "cbid": 211, "correlation": 161050
    }
  },
  {
    "ph": "s", "id": 161050, "pid": 76337, "tid": -914061504, "ts": 1716454223862050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223930674, "dur": 22,
    "args": {
      "External id": 161052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161052,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161052, "pid": 5, "tid": 7, "ts": 1716454223930674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862062, "dur": 5,
    "args": {
      "External id": 161052,
      "cbid": 211, "correlation": 161052
    }
  },
  {
    "ph": "s", "id": 161052, "pid": 76337, "tid": -914061504, "ts": 1716454223862062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223930697, "dur": 32,
    "args": {
      "External id": 161058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161058,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161058, "pid": 5, "tid": 7, "ts": 1716454223930697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862088, "dur": 8,
    "args": {
      "External id": 161058,
      "cbid": 211, "correlation": 161058
    }
  },
  {
    "ph": "s", "id": 161058, "pid": 76337, "tid": -914061504, "ts": 1716454223862088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223930731, "dur": 27,
    "args": {
      "External id": 161066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161066,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161066, "pid": 5, "tid": 7, "ts": 1716454223930731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862117, "dur": 8,
    "args": {
      "External id": 161066,
      "cbid": 211, "correlation": 161066
    }
  },
  {
    "ph": "s", "id": 161066, "pid": 76337, "tid": -914061504, "ts": 1716454223862117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223930759, "dur": 20,
    "args": {
      "External id": 161074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161074,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161074, "pid": 5, "tid": 7, "ts": 1716454223930759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862145, "dur": 9,
    "args": {
      "External id": 161074,
      "cbid": 211, "correlation": 161074
    }
  },
  {
    "ph": "s", "id": 161074, "pid": 76337, "tid": -914061504, "ts": 1716454223862145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223930780, "dur": 30,
    "args": {
      "External id": 161094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161094,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 161094, "pid": 5, "tid": 7, "ts": 1716454223930780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862228, "dur": 12,
    "args": {
      "External id": 161094,
      "cbid": 211, "correlation": 161094
    }
  },
  {
    "ph": "s", "id": 161094, "pid": 76337, "tid": -914061504, "ts": 1716454223862228,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223930812, "dur": 4,
    "args": {
      "External id": 161106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161106,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 161106, "pid": 5, "tid": 7, "ts": 1716454223930812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862249, "dur": 7,
    "args": {
      "External id": 161106,
      "cbid": 211, "correlation": 161106
    }
  },
  {
    "ph": "s", "id": 161106, "pid": 76337, "tid": -914061504, "ts": 1716454223862249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223930817, "dur": 31,
    "args": {
      "External id": 161109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161109,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161109, "pid": 5, "tid": 7, "ts": 1716454223930817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862267, "dur": 8,
    "args": {
      "External id": 161109,
      "cbid": 211, "correlation": 161109
    }
  },
  {
    "ph": "s", "id": 161109, "pid": 76337, "tid": -914061504, "ts": 1716454223862267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862326, "dur": 0,
    "args": {
      "External id": 161120,
      "cbid": 317, "correlation": 161120
    }
  },
  {
    "ph": "f", "id": 161120, "pid": 76337, "tid": -914061504, "ts": 1716454223862326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862327, "dur": 0,
    "args": {
      "External id": 161121,
      "cbid": 203, "correlation": 161121
    }
  },
  {
    "ph": "f", "id": 161121, "pid": 76337, "tid": -914061504, "ts": 1716454223862327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862327, "dur": 0,
    "args": {
      "External id": 161122,
      "cbid": 205, "correlation": 161122
    }
  },
  {
    "ph": "f", "id": 161122, "pid": 76337, "tid": -914061504, "ts": 1716454223862327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223930850, "dur": 22,
    "args": {
      "External id": 161126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161126,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161126, "pid": 5, "tid": 7, "ts": 1716454223930850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862341, "dur": 12,
    "args": {
      "External id": 161126,
      "cbid": 211, "correlation": 161126
    }
  },
  {
    "ph": "s", "id": 161126, "pid": 76337, "tid": -914061504, "ts": 1716454223862341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223930873, "dur": 104,
    "args": {
      "External id": 161128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161128,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161128, "pid": 5, "tid": 7, "ts": 1716454223930873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862360, "dur": 6,
    "args": {
      "External id": 161128,
      "cbid": 211, "correlation": 161128
    }
  },
  {
    "ph": "s", "id": 161128, "pid": 76337, "tid": -914061504, "ts": 1716454223862360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223930978, "dur": 23,
    "args": {
      "External id": 161130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161130, "pid": 5, "tid": 7, "ts": 1716454223930978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862369, "dur": 5,
    "args": {
      "External id": 161130,
      "cbid": 211, "correlation": 161130
    }
  },
  {
    "ph": "s", "id": 161130, "pid": 76337, "tid": -914061504, "ts": 1716454223862369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223931003, "dur": 33,
    "args": {
      "External id": 161136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161136, "pid": 5, "tid": 7, "ts": 1716454223931003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862396, "dur": 9,
    "args": {
      "External id": 161136,
      "cbid": 211, "correlation": 161136
    }
  },
  {
    "ph": "s", "id": 161136, "pid": 76337, "tid": -914061504, "ts": 1716454223862396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223931037, "dur": 204,
    "args": {
      "External id": 161145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161145,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161145, "pid": 5, "tid": 7, "ts": 1716454223931037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862478, "dur": 13,
    "args": {
      "External id": 161145,
      "cbid": 211, "correlation": 161145
    }
  },
  {
    "ph": "s", "id": 161145, "pid": 76337, "tid": -914061504, "ts": 1716454223862478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223931242, "dur": 65,
    "args": {
      "External id": 161167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161167,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161167, "pid": 5, "tid": 7, "ts": 1716454223931242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862536, "dur": 10,
    "args": {
      "External id": 161167,
      "cbid": 211, "correlation": 161167
    }
  },
  {
    "ph": "s", "id": 161167, "pid": 76337, "tid": -914061504, "ts": 1716454223862536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862626, "dur": 1,
    "args": {
      "External id": 161178,
      "cbid": 251, "correlation": 161178
    }
  },
  {
    "ph": "f", "id": 161178, "pid": 76337, "tid": -914061504, "ts": 1716454223862626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223931309, "dur": 154,
    "args": {
      "External id": 161179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161179,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161179, "pid": 5, "tid": 7, "ts": 1716454223931309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862631, "dur": 13,
    "args": {
      "External id": 161179,
      "cbid": 211, "correlation": 161179
    }
  },
  {
    "ph": "s", "id": 161179, "pid": 76337, "tid": -914061504, "ts": 1716454223862631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862704, "dur": 1,
    "args": {
      "External id": 161190,
      "cbid": 251, "correlation": 161190
    }
  },
  {
    "ph": "f", "id": 161190, "pid": 76337, "tid": -914061504, "ts": 1716454223862704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223931465, "dur": 147,
    "args": {
      "External id": 161191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161191,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161191, "pid": 5, "tid": 7, "ts": 1716454223931465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862708, "dur": 11,
    "args": {
      "External id": 161191,
      "cbid": 211, "correlation": 161191
    }
  },
  {
    "ph": "s", "id": 161191, "pid": 76337, "tid": -914061504, "ts": 1716454223862708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862772, "dur": 1,
    "args": {
      "External id": 161202,
      "cbid": 251, "correlation": 161202
    }
  },
  {
    "ph": "f", "id": 161202, "pid": 76337, "tid": -914061504, "ts": 1716454223862772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223931613, "dur": 145,
    "args": {
      "External id": 161203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161203,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161203, "pid": 5, "tid": 7, "ts": 1716454223931613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862776, "dur": 11,
    "args": {
      "External id": 161203,
      "cbid": 211, "correlation": 161203
    }
  },
  {
    "ph": "s", "id": 161203, "pid": 76337, "tid": -914061504, "ts": 1716454223862776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223931760, "dur": 1948,
    "args": {
      "External id": 161224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161224,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 161224, "pid": 5, "tid": 7, "ts": 1716454223931760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862860, "dur": 14,
    "args": {
      "External id": 161224,
      "cbid": 211, "correlation": 161224
    }
  },
  {
    "ph": "s", "id": 161224, "pid": 76337, "tid": -914061504, "ts": 1716454223862860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862964, "dur": 2,
    "args": {
      "External id": 161242,
      "cbid": 251, "correlation": 161242
    }
  },
  {
    "ph": "f", "id": 161242, "pid": 76337, "tid": -914061504, "ts": 1716454223862964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223933709, "dur": 148,
    "args": {
      "External id": 161244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161244,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 161244, "pid": 5, "tid": 7, "ts": 1716454223933709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223862970, "dur": 23,
    "args": {
      "External id": 161244,
      "cbid": 211, "correlation": 161244
    }
  },
  {
    "ph": "s", "id": 161244, "pid": 76337, "tid": -914061504, "ts": 1716454223862970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223933858, "dur": 35,
    "args": {
      "External id": 161252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161252,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161252, "pid": 5, "tid": 7, "ts": 1716454223933858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863052, "dur": 12,
    "args": {
      "External id": 161252,
      "cbid": 211, "correlation": 161252
    }
  },
  {
    "ph": "s", "id": 161252, "pid": 76337, "tid": -914061504, "ts": 1716454223863052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223933895, "dur": 50,
    "args": {
      "External id": 161260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161260,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161260, "pid": 5, "tid": 7, "ts": 1716454223933895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863090, "dur": 9,
    "args": {
      "External id": 161260,
      "cbid": 211, "correlation": 161260
    }
  },
  {
    "ph": "s", "id": 161260, "pid": 76337, "tid": -914061504, "ts": 1716454223863090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223933947, "dur": 30,
    "args": {
      "External id": 161271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161271, "pid": 5, "tid": 7, "ts": 1716454223933947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863163, "dur": 12,
    "args": {
      "External id": 161271,
      "cbid": 211, "correlation": 161271
    }
  },
  {
    "ph": "s", "id": 161271, "pid": 76337, "tid": -914061504, "ts": 1716454223863163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223933979, "dur": 35,
    "args": {
      "External id": 161293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161293,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161293, "pid": 5, "tid": 7, "ts": 1716454223933979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863194, "dur": 8,
    "args": {
      "External id": 161293,
      "cbid": 211, "correlation": 161293
    }
  },
  {
    "ph": "s", "id": 161293, "pid": 76337, "tid": -914061504, "ts": 1716454223863194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863279, "dur": 1,
    "args": {
      "External id": 161304,
      "cbid": 251, "correlation": 161304
    }
  },
  {
    "ph": "f", "id": 161304, "pid": 76337, "tid": -914061504, "ts": 1716454223863279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223934015, "dur": 91,
    "args": {
      "External id": 161305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161305,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161305, "pid": 5, "tid": 7, "ts": 1716454223934015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863284, "dur": 13,
    "args": {
      "External id": 161305,
      "cbid": 211, "correlation": 161305
    }
  },
  {
    "ph": "s", "id": 161305, "pid": 76337, "tid": -914061504, "ts": 1716454223863284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863354, "dur": 1,
    "args": {
      "External id": 161316,
      "cbid": 251, "correlation": 161316
    }
  },
  {
    "ph": "f", "id": 161316, "pid": 76337, "tid": -914061504, "ts": 1716454223863354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863358, "dur": 0,
    "args": {
      "External id": 161317,
      "cbid": 251, "correlation": 161317
    }
  },
  {
    "ph": "f", "id": 161317, "pid": 76337, "tid": -914061504, "ts": 1716454223863358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223934107, "dur": 11,
    "args": {
      "External id": 161318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161318,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 161318, "pid": 5, "tid": 7, "ts": 1716454223934107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863359, "dur": 11,
    "args": {
      "External id": 161318,
      "cbid": 211, "correlation": 161318
    }
  },
  {
    "ph": "s", "id": 161318, "pid": 76337, "tid": -914061504, "ts": 1716454223863359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223934119, "dur": 5,
    "args": {
      "External id": 161320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161320,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 161320, "pid": 5, "tid": 7, "ts": 1716454223934119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863373, "dur": 6,
    "args": {
      "External id": 161320,
      "cbid": 211, "correlation": 161320
    }
  },
  {
    "ph": "s", "id": 161320, "pid": 76337, "tid": -914061504, "ts": 1716454223863373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863430, "dur": 1,
    "args": {
      "External id": 161331,
      "cbid": 251, "correlation": 161331
    }
  },
  {
    "ph": "f", "id": 161331, "pid": 76337, "tid": -914061504, "ts": 1716454223863430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863434, "dur": 0,
    "args": {
      "External id": 161332,
      "cbid": 251, "correlation": 161332
    }
  },
  {
    "ph": "f", "id": 161332, "pid": 76337, "tid": -914061504, "ts": 1716454223863434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223934126, "dur": 7,
    "args": {
      "External id": 161333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161333,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 161333, "pid": 5, "tid": 7, "ts": 1716454223934126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863435, "dur": 12,
    "args": {
      "External id": 161333,
      "cbid": 211, "correlation": 161333
    }
  },
  {
    "ph": "s", "id": 161333, "pid": 76337, "tid": -914061504, "ts": 1716454223863435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223934135, "dur": 4,
    "args": {
      "External id": 161335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161335,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 161335, "pid": 5, "tid": 7, "ts": 1716454223934135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863449, "dur": 6,
    "args": {
      "External id": 161335,
      "cbid": 211, "correlation": 161335
    }
  },
  {
    "ph": "s", "id": 161335, "pid": 76337, "tid": -914061504, "ts": 1716454223863449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223934139, "dur": 91,
    "args": {
      "External id": 161356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161356,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 161356, "pid": 5, "tid": 7, "ts": 1716454223934139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863523, "dur": 12,
    "args": {
      "External id": 161356,
      "cbid": 211, "correlation": 161356
    }
  },
  {
    "ph": "s", "id": 161356, "pid": 76337, "tid": -914061504, "ts": 1716454223863523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863620, "dur": 1,
    "args": {
      "External id": 161374,
      "cbid": 251, "correlation": 161374
    }
  },
  {
    "ph": "f", "id": 161374, "pid": 76337, "tid": -914061504, "ts": 1716454223863620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223934232, "dur": 98,
    "args": {
      "External id": 161376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161376,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161376, "pid": 5, "tid": 7, "ts": 1716454223934232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863626, "dur": 13,
    "args": {
      "External id": 161376,
      "cbid": 211, "correlation": 161376
    }
  },
  {
    "ph": "s", "id": 161376, "pid": 76337, "tid": -914061504, "ts": 1716454223863626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223934332, "dur": 19,
    "args": {
      "External id": 161384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161384,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161384, "pid": 5, "tid": 7, "ts": 1716454223934332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863695, "dur": 13,
    "args": {
      "External id": 161384,
      "cbid": 211, "correlation": 161384
    }
  },
  {
    "ph": "s", "id": 161384, "pid": 76337, "tid": -914061504, "ts": 1716454223863695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223934352, "dur": 38,
    "args": {
      "External id": 161392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161392,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161392, "pid": 5, "tid": 7, "ts": 1716454223934352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863736, "dur": 10,
    "args": {
      "External id": 161392,
      "cbid": 211, "correlation": 161392
    }
  },
  {
    "ph": "s", "id": 161392, "pid": 76337, "tid": -914061504, "ts": 1716454223863736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223934392, "dur": 34,
    "args": {
      "External id": 161414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161414,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161414, "pid": 5, "tid": 7, "ts": 1716454223934392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863788, "dur": 11,
    "args": {
      "External id": 161414,
      "cbid": 211, "correlation": 161414
    }
  },
  {
    "ph": "s", "id": 161414, "pid": 76337, "tid": -914061504, "ts": 1716454223863788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863877, "dur": 1,
    "args": {
      "External id": 161430,
      "cbid": 251, "correlation": 161430
    }
  },
  {
    "ph": "f", "id": 161430, "pid": 76337, "tid": -914061504, "ts": 1716454223863877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863882, "dur": 0,
    "args": {
      "External id": 161432,
      "cbid": 251, "correlation": 161432
    }
  },
  {
    "ph": "f", "id": 161432, "pid": 76337, "tid": -914061504, "ts": 1716454223863882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223934427, "dur": 541,
    "args": {
      "External id": 161433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161433,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 161433, "pid": 5, "tid": 7, "ts": 1716454223934427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863886, "dur": 13,
    "args": {
      "External id": 161433,
      "cbid": 211, "correlation": 161433
    }
  },
  {
    "ph": "s", "id": 161433, "pid": 76337, "tid": -914061504, "ts": 1716454223863886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223934970, "dur": 125,
    "args": {
      "External id": 161441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161441,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161441, "pid": 5, "tid": 7, "ts": 1716454223934970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863953, "dur": 12,
    "args": {
      "External id": 161441,
      "cbid": 211, "correlation": 161441
    }
  },
  {
    "ph": "s", "id": 161441, "pid": 76337, "tid": -914061504, "ts": 1716454223863953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223935096, "dur": 131,
    "args": {
      "External id": 161449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161449, "pid": 5, "tid": 7, "ts": 1716454223935096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223863998, "dur": 10,
    "args": {
      "External id": 161449,
      "cbid": 211, "correlation": 161449
    }
  },
  {
    "ph": "s", "id": 161449, "pid": 76337, "tid": -914061504, "ts": 1716454223863998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864077, "dur": 1,
    "args": {
      "External id": 161465,
      "cbid": 251, "correlation": 161465
    }
  },
  {
    "ph": "f", "id": 161465, "pid": 76337, "tid": -914061504, "ts": 1716454223864077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223935228, "dur": 307,
    "args": {
      "External id": 161467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161467,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161467, "pid": 5, "tid": 7, "ts": 1716454223935228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864083, "dur": 13,
    "args": {
      "External id": 161467,
      "cbid": 211, "correlation": 161467
    }
  },
  {
    "ph": "s", "id": 161467, "pid": 76337, "tid": -914061504, "ts": 1716454223864083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223935537, "dur": 27,
    "args": {
      "External id": 161475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161475,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161475, "pid": 5, "tid": 7, "ts": 1716454223935537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864126, "dur": 9,
    "args": {
      "External id": 161475,
      "cbid": 211, "correlation": 161475
    }
  },
  {
    "ph": "s", "id": 161475, "pid": 76337, "tid": -914061504, "ts": 1716454223864126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223935566, "dur": 81,
    "args": {
      "External id": 161486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161486,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161486, "pid": 5, "tid": 7, "ts": 1716454223935566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864193, "dur": 13,
    "args": {
      "External id": 161486,
      "cbid": 211, "correlation": 161486
    }
  },
  {
    "ph": "s", "id": 161486, "pid": 76337, "tid": -914061504, "ts": 1716454223864193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864258, "dur": 0,
    "args": {
      "External id": 161498,
      "cbid": 317, "correlation": 161498
    }
  },
  {
    "ph": "f", "id": 161498, "pid": 76337, "tid": -914061504, "ts": 1716454223864258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864259, "dur": 0,
    "args": {
      "External id": 161499,
      "cbid": 203, "correlation": 161499
    }
  },
  {
    "ph": "f", "id": 161499, "pid": 76337, "tid": -914061504, "ts": 1716454223864259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864260, "dur": 0,
    "args": {
      "External id": 161500,
      "cbid": 205, "correlation": 161500
    }
  },
  {
    "ph": "f", "id": 161500, "pid": 76337, "tid": -914061504, "ts": 1716454223864260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223935648, "dur": 23,
    "args": {
      "External id": 161504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161504,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161504, "pid": 5, "tid": 7, "ts": 1716454223935648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864275, "dur": 12,
    "args": {
      "External id": 161504,
      "cbid": 211, "correlation": 161504
    }
  },
  {
    "ph": "s", "id": 161504, "pid": 76337, "tid": -914061504, "ts": 1716454223864275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223935672, "dur": 120,
    "args": {
      "External id": 161506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161506,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161506, "pid": 5, "tid": 7, "ts": 1716454223935672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864294, "dur": 7,
    "args": {
      "External id": 161506,
      "cbid": 211, "correlation": 161506
    }
  },
  {
    "ph": "s", "id": 161506, "pid": 76337, "tid": -914061504, "ts": 1716454223864294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223935794, "dur": 23,
    "args": {
      "External id": 161508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161508,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161508, "pid": 5, "tid": 7, "ts": 1716454223935794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864304, "dur": 5,
    "args": {
      "External id": 161508,
      "cbid": 211, "correlation": 161508
    }
  },
  {
    "ph": "s", "id": 161508, "pid": 76337, "tid": -914061504, "ts": 1716454223864304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223935818, "dur": 33,
    "args": {
      "External id": 161514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161514,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161514, "pid": 5, "tid": 7, "ts": 1716454223935818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864332, "dur": 8,
    "args": {
      "External id": 161514,
      "cbid": 211, "correlation": 161514
    }
  },
  {
    "ph": "s", "id": 161514, "pid": 76337, "tid": -914061504, "ts": 1716454223864332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223935852, "dur": 27,
    "args": {
      "External id": 161522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161522,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161522, "pid": 5, "tid": 7, "ts": 1716454223935852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864363, "dur": 8,
    "args": {
      "External id": 161522,
      "cbid": 211, "correlation": 161522
    }
  },
  {
    "ph": "s", "id": 161522, "pid": 76337, "tid": -914061504, "ts": 1716454223864363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223935880, "dur": 53,
    "args": {
      "External id": 161531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161531,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161531, "pid": 5, "tid": 7, "ts": 1716454223935880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864403, "dur": 10,
    "args": {
      "External id": 161531,
      "cbid": 211, "correlation": 161531
    }
  },
  {
    "ph": "s", "id": 161531, "pid": 76337, "tid": -914061504, "ts": 1716454223864403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223935935, "dur": 54,
    "args": {
      "External id": 161551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161551,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 161551, "pid": 5, "tid": 7, "ts": 1716454223935935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864473, "dur": 11,
    "args": {
      "External id": 161551,
      "cbid": 211, "correlation": 161551
    }
  },
  {
    "ph": "s", "id": 161551, "pid": 76337, "tid": -914061504, "ts": 1716454223864473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223935990, "dur": 4,
    "args": {
      "External id": 161563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161563,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 161563, "pid": 5, "tid": 7, "ts": 1716454223935990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864495, "dur": 6,
    "args": {
      "External id": 161563,
      "cbid": 211, "correlation": 161563
    }
  },
  {
    "ph": "s", "id": 161563, "pid": 76337, "tid": -914061504, "ts": 1716454223864495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223935996, "dur": 57,
    "args": {
      "External id": 161566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161566, "pid": 5, "tid": 7, "ts": 1716454223935996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864513, "dur": 6,
    "args": {
      "External id": 161566,
      "cbid": 211, "correlation": 161566
    }
  },
  {
    "ph": "s", "id": 161566, "pid": 76337, "tid": -914061504, "ts": 1716454223864513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223936055, "dur": 38,
    "args": {
      "External id": 161575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161575,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161575, "pid": 5, "tid": 7, "ts": 1716454223936055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864554, "dur": 9,
    "args": {
      "External id": 161575,
      "cbid": 211, "correlation": 161575
    }
  },
  {
    "ph": "s", "id": 161575, "pid": 76337, "tid": -914061504, "ts": 1716454223864554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864607, "dur": 0,
    "args": {
      "External id": 161585,
      "cbid": 317, "correlation": 161585
    }
  },
  {
    "ph": "f", "id": 161585, "pid": 76337, "tid": -914061504, "ts": 1716454223864607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864608, "dur": 0,
    "args": {
      "External id": 161586,
      "cbid": 203, "correlation": 161586
    }
  },
  {
    "ph": "f", "id": 161586, "pid": 76337, "tid": -914061504, "ts": 1716454223864608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864608, "dur": 0,
    "args": {
      "External id": 161587,
      "cbid": 205, "correlation": 161587
    }
  },
  {
    "ph": "f", "id": 161587, "pid": 76337, "tid": -914061504, "ts": 1716454223864608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223936094, "dur": 39,
    "args": {
      "External id": 161591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161591,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161591, "pid": 5, "tid": 7, "ts": 1716454223936094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864622, "dur": 11,
    "args": {
      "External id": 161591,
      "cbid": 211, "correlation": 161591
    }
  },
  {
    "ph": "s", "id": 161591, "pid": 76337, "tid": -914061504, "ts": 1716454223864622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223936134, "dur": 83,
    "args": {
      "External id": 161593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161593,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161593, "pid": 5, "tid": 7, "ts": 1716454223936134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864636, "dur": 5,
    "args": {
      "External id": 161593,
      "cbid": 211, "correlation": 161593
    }
  },
  {
    "ph": "s", "id": 161593, "pid": 76337, "tid": -914061504, "ts": 1716454223864636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223936218, "dur": 1279,
    "args": {
      "External id": 161595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161595,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161595, "pid": 5, "tid": 7, "ts": 1716454223936218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864648, "dur": 7,
    "args": {
      "External id": 161595,
      "cbid": 211, "correlation": 161595
    }
  },
  {
    "ph": "s", "id": 161595, "pid": 76337, "tid": -914061504, "ts": 1716454223864648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223937498, "dur": 23,
    "args": {
      "External id": 161597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161597,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161597, "pid": 5, "tid": 7, "ts": 1716454223937498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864658, "dur": 5,
    "args": {
      "External id": 161597,
      "cbid": 211, "correlation": 161597
    }
  },
  {
    "ph": "s", "id": 161597, "pid": 76337, "tid": -914061504, "ts": 1716454223864658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223937523, "dur": 33,
    "args": {
      "External id": 161603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161603, "pid": 5, "tid": 7, "ts": 1716454223937523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864685, "dur": 9,
    "args": {
      "External id": 161603,
      "cbid": 211, "correlation": 161603
    }
  },
  {
    "ph": "s", "id": 161603, "pid": 76337, "tid": -914061504, "ts": 1716454223864685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223937557, "dur": 3,
    "args": {
      "External id": 161611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161611,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 161611, "pid": 5, "tid": 7, "ts": 1716454223937557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864729, "dur": 9,
    "args": {
      "External id": 161611,
      "cbid": 211, "correlation": 161611
    }
  },
  {
    "ph": "s", "id": 161611, "pid": 76337, "tid": -914061504, "ts": 1716454223864729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864797, "dur": 1,
    "args": {
      "External id": 161627,
      "cbid": 251, "correlation": 161627
    }
  },
  {
    "ph": "f", "id": 161627, "pid": 76337, "tid": -914061504, "ts": 1716454223864797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864802, "dur": 0,
    "args": {
      "External id": 161629,
      "cbid": 251, "correlation": 161629
    }
  },
  {
    "ph": "f", "id": 161629, "pid": 76337, "tid": -914061504, "ts": 1716454223864802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223937562, "dur": 12,
    "args": {
      "External id": 161630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161630,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 161630, "pid": 5, "tid": 7, "ts": 1716454223937562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864804, "dur": 11,
    "args": {
      "External id": 161630,
      "cbid": 211, "correlation": 161630
    }
  },
  {
    "ph": "s", "id": 161630, "pid": 76337, "tid": -914061504, "ts": 1716454223864804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223937575, "dur": 5,
    "args": {
      "External id": 161632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161632,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 161632, "pid": 5, "tid": 7, "ts": 1716454223937575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864817, "dur": 5,
    "args": {
      "External id": 161632,
      "cbid": 211, "correlation": 161632
    }
  },
  {
    "ph": "s", "id": 161632, "pid": 76337, "tid": -914061504, "ts": 1716454223864817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223937581, "dur": 29,
    "args": {
      "External id": 161642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161642,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161642, "pid": 5, "tid": 7, "ts": 1716454223937581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864875, "dur": 12,
    "args": {
      "External id": 161642,
      "cbid": 211, "correlation": 161642
    }
  },
  {
    "ph": "s", "id": 161642, "pid": 76337, "tid": -914061504, "ts": 1716454223864875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223937612, "dur": 31,
    "args": {
      "External id": 161662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161662,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 161662, "pid": 5, "tid": 7, "ts": 1716454223937612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864941, "dur": 11,
    "args": {
      "External id": 161662,
      "cbid": 211, "correlation": 161662
    }
  },
  {
    "ph": "s", "id": 161662, "pid": 76337, "tid": -914061504, "ts": 1716454223864941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223937644, "dur": 4,
    "args": {
      "External id": 161674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161674,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 161674, "pid": 5, "tid": 7, "ts": 1716454223937644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864962, "dur": 6,
    "args": {
      "External id": 161674,
      "cbid": 211, "correlation": 161674
    }
  },
  {
    "ph": "s", "id": 161674, "pid": 76337, "tid": -914061504, "ts": 1716454223864962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223937649, "dur": 30,
    "args": {
      "External id": 161677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161677, "pid": 5, "tid": 7, "ts": 1716454223937649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223864988, "dur": 8,
    "args": {
      "External id": 161677,
      "cbid": 211, "correlation": 161677
    }
  },
  {
    "ph": "s", "id": 161677, "pid": 76337, "tid": -914061504, "ts": 1716454223864988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223937680, "dur": 20,
    "args": {
      "External id": 161686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161686,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161686, "pid": 5, "tid": 7, "ts": 1716454223937680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865031, "dur": 10,
    "args": {
      "External id": 161686,
      "cbid": 211, "correlation": 161686
    }
  },
  {
    "ph": "s", "id": 161686, "pid": 76337, "tid": -914061504, "ts": 1716454223865031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865093, "dur": 0,
    "args": {
      "External id": 161696,
      "cbid": 317, "correlation": 161696
    }
  },
  {
    "ph": "f", "id": 161696, "pid": 76337, "tid": -914061504, "ts": 1716454223865093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865094, "dur": 0,
    "args": {
      "External id": 161697,
      "cbid": 203, "correlation": 161697
    }
  },
  {
    "ph": "f", "id": 161697, "pid": 76337, "tid": -914061504, "ts": 1716454223865094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865095, "dur": 0,
    "args": {
      "External id": 161698,
      "cbid": 205, "correlation": 161698
    }
  },
  {
    "ph": "f", "id": 161698, "pid": 76337, "tid": -914061504, "ts": 1716454223865095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223937701, "dur": 23,
    "args": {
      "External id": 161702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161702,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161702, "pid": 5, "tid": 7, "ts": 1716454223937701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865110, "dur": 12,
    "args": {
      "External id": 161702,
      "cbid": 211, "correlation": 161702
    }
  },
  {
    "ph": "s", "id": 161702, "pid": 76337, "tid": -914061504, "ts": 1716454223865110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223937725, "dur": 44,
    "args": {
      "External id": 161704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161704, "pid": 5, "tid": 7, "ts": 1716454223937725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865124, "dur": 5,
    "args": {
      "External id": 161704,
      "cbid": 211, "correlation": 161704
    }
  },
  {
    "ph": "s", "id": 161704, "pid": 76337, "tid": -914061504, "ts": 1716454223865124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223937770, "dur": 644,
    "args": {
      "External id": 161706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161706,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161706, "pid": 5, "tid": 7, "ts": 1716454223937770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865135, "dur": 6,
    "args": {
      "External id": 161706,
      "cbid": 211, "correlation": 161706
    }
  },
  {
    "ph": "s", "id": 161706, "pid": 76337, "tid": -914061504, "ts": 1716454223865135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223938416, "dur": 21,
    "args": {
      "External id": 161708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161708,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161708, "pid": 5, "tid": 7, "ts": 1716454223938416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865145, "dur": 6,
    "args": {
      "External id": 161708,
      "cbid": 211, "correlation": 161708
    }
  },
  {
    "ph": "s", "id": 161708, "pid": 76337, "tid": -914061504, "ts": 1716454223865145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223938438, "dur": 33,
    "args": {
      "External id": 161714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161714,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161714, "pid": 5, "tid": 7, "ts": 1716454223938438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865173, "dur": 9,
    "args": {
      "External id": 161714,
      "cbid": 211, "correlation": 161714
    }
  },
  {
    "ph": "s", "id": 161714, "pid": 76337, "tid": -914061504, "ts": 1716454223865173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865233, "dur": 0,
    "args": {
      "External id": 161724,
      "cbid": 317, "correlation": 161724
    }
  },
  {
    "ph": "f", "id": 161724, "pid": 76337, "tid": -914061504, "ts": 1716454223865233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865233, "dur": 0,
    "args": {
      "External id": 161725,
      "cbid": 203, "correlation": 161725
    }
  },
  {
    "ph": "f", "id": 161725, "pid": 76337, "tid": -914061504, "ts": 1716454223865233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865234, "dur": 0,
    "args": {
      "External id": 161726,
      "cbid": 205, "correlation": 161726
    }
  },
  {
    "ph": "f", "id": 161726, "pid": 76337, "tid": -914061504, "ts": 1716454223865234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223938473, "dur": 39,
    "args": {
      "External id": 161730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161730,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161730, "pid": 5, "tid": 7, "ts": 1716454223938473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865247, "dur": 14,
    "args": {
      "External id": 161730,
      "cbid": 211, "correlation": 161730
    }
  },
  {
    "ph": "s", "id": 161730, "pid": 76337, "tid": -914061504, "ts": 1716454223865247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223938513, "dur": 192,
    "args": {
      "External id": 161732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161732,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161732, "pid": 5, "tid": 7, "ts": 1716454223938513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865266, "dur": 6,
    "args": {
      "External id": 161732,
      "cbid": 211, "correlation": 161732
    }
  },
  {
    "ph": "s", "id": 161732, "pid": 76337, "tid": -914061504, "ts": 1716454223865266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223938707, "dur": 22,
    "args": {
      "External id": 161734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161734, "pid": 5, "tid": 7, "ts": 1716454223938707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865277, "dur": 5,
    "args": {
      "External id": 161734,
      "cbid": 211, "correlation": 161734
    }
  },
  {
    "ph": "s", "id": 161734, "pid": 76337, "tid": -914061504, "ts": 1716454223865277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223938730, "dur": 33,
    "args": {
      "External id": 161740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161740, "pid": 5, "tid": 7, "ts": 1716454223938730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865303, "dur": 8,
    "args": {
      "External id": 161740,
      "cbid": 211, "correlation": 161740
    }
  },
  {
    "ph": "s", "id": 161740, "pid": 76337, "tid": -914061504, "ts": 1716454223865303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223938764, "dur": 27,
    "args": {
      "External id": 161748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161748,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161748, "pid": 5, "tid": 7, "ts": 1716454223938764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865331, "dur": 8,
    "args": {
      "External id": 161748,
      "cbid": 211, "correlation": 161748
    }
  },
  {
    "ph": "s", "id": 161748, "pid": 76337, "tid": -914061504, "ts": 1716454223865331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223938792, "dur": 19,
    "args": {
      "External id": 161756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161756,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161756, "pid": 5, "tid": 7, "ts": 1716454223938792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865361, "dur": 8,
    "args": {
      "External id": 161756,
      "cbid": 211, "correlation": 161756
    }
  },
  {
    "ph": "s", "id": 161756, "pid": 76337, "tid": -914061504, "ts": 1716454223865361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223938813, "dur": 31,
    "args": {
      "External id": 161776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161776,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 161776, "pid": 5, "tid": 7, "ts": 1716454223938813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865443, "dur": 13,
    "args": {
      "External id": 161776,
      "cbid": 211, "correlation": 161776
    }
  },
  {
    "ph": "s", "id": 161776, "pid": 76337, "tid": -914061504, "ts": 1716454223865443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223938845, "dur": 4,
    "args": {
      "External id": 161788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161788,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 161788, "pid": 5, "tid": 7, "ts": 1716454223938845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865466, "dur": 6,
    "args": {
      "External id": 161788,
      "cbid": 211, "correlation": 161788
    }
  },
  {
    "ph": "s", "id": 161788, "pid": 76337, "tid": -914061504, "ts": 1716454223865466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223938850, "dur": 30,
    "args": {
      "External id": 161791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161791,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161791, "pid": 5, "tid": 7, "ts": 1716454223938850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865483, "dur": 7,
    "args": {
      "External id": 161791,
      "cbid": 211, "correlation": 161791
    }
  },
  {
    "ph": "s", "id": 161791, "pid": 76337, "tid": -914061504, "ts": 1716454223865483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865540, "dur": 0,
    "args": {
      "External id": 161802,
      "cbid": 317, "correlation": 161802
    }
  },
  {
    "ph": "f", "id": 161802, "pid": 76337, "tid": -914061504, "ts": 1716454223865540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865541, "dur": 0,
    "args": {
      "External id": 161803,
      "cbid": 203, "correlation": 161803
    }
  },
  {
    "ph": "f", "id": 161803, "pid": 76337, "tid": -914061504, "ts": 1716454223865541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865541, "dur": 0,
    "args": {
      "External id": 161804,
      "cbid": 205, "correlation": 161804
    }
  },
  {
    "ph": "f", "id": 161804, "pid": 76337, "tid": -914061504, "ts": 1716454223865541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223938882, "dur": 24,
    "args": {
      "External id": 161808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161808,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161808, "pid": 5, "tid": 7, "ts": 1716454223938882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865555, "dur": 12,
    "args": {
      "External id": 161808,
      "cbid": 211, "correlation": 161808
    }
  },
  {
    "ph": "s", "id": 161808, "pid": 76337, "tid": -914061504, "ts": 1716454223865555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223938907, "dur": 105,
    "args": {
      "External id": 161810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161810,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161810, "pid": 5, "tid": 7, "ts": 1716454223938907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865573, "dur": 6,
    "args": {
      "External id": 161810,
      "cbid": 211, "correlation": 161810
    }
  },
  {
    "ph": "s", "id": 161810, "pid": 76337, "tid": -914061504, "ts": 1716454223865573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223939013, "dur": 22,
    "args": {
      "External id": 161812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161812, "pid": 5, "tid": 7, "ts": 1716454223939013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865582, "dur": 5,
    "args": {
      "External id": 161812,
      "cbid": 211, "correlation": 161812
    }
  },
  {
    "ph": "s", "id": 161812, "pid": 76337, "tid": -914061504, "ts": 1716454223865582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223939037, "dur": 33,
    "args": {
      "External id": 161818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161818, "pid": 5, "tid": 7, "ts": 1716454223939037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865609, "dur": 8,
    "args": {
      "External id": 161818,
      "cbid": 211, "correlation": 161818
    }
  },
  {
    "ph": "s", "id": 161818, "pid": 76337, "tid": -914061504, "ts": 1716454223865609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223939071, "dur": 200,
    "args": {
      "External id": 161827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161827,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161827, "pid": 5, "tid": 7, "ts": 1716454223939071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865690, "dur": 15,
    "args": {
      "External id": 161827,
      "cbid": 211, "correlation": 161827
    }
  },
  {
    "ph": "s", "id": 161827, "pid": 76337, "tid": -914061504, "ts": 1716454223865690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223939273, "dur": 65,
    "args": {
      "External id": 161849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161849,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161849, "pid": 5, "tid": 7, "ts": 1716454223939273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865748, "dur": 10,
    "args": {
      "External id": 161849,
      "cbid": 211, "correlation": 161849
    }
  },
  {
    "ph": "s", "id": 161849, "pid": 76337, "tid": -914061504, "ts": 1716454223865748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865836, "dur": 1,
    "args": {
      "External id": 161860,
      "cbid": 251, "correlation": 161860
    }
  },
  {
    "ph": "f", "id": 161860, "pid": 76337, "tid": -914061504, "ts": 1716454223865836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223939339, "dur": 155,
    "args": {
      "External id": 161861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161861,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161861, "pid": 5, "tid": 7, "ts": 1716454223939339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865840, "dur": 14,
    "args": {
      "External id": 161861,
      "cbid": 211, "correlation": 161861
    }
  },
  {
    "ph": "s", "id": 161861, "pid": 76337, "tid": -914061504, "ts": 1716454223865840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865911, "dur": 1,
    "args": {
      "External id": 161872,
      "cbid": 251, "correlation": 161872
    }
  },
  {
    "ph": "f", "id": 161872, "pid": 76337, "tid": -914061504, "ts": 1716454223865911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223939496, "dur": 145,
    "args": {
      "External id": 161873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161873,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161873, "pid": 5, "tid": 7, "ts": 1716454223939496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865915, "dur": 11,
    "args": {
      "External id": 161873,
      "cbid": 211, "correlation": 161873
    }
  },
  {
    "ph": "s", "id": 161873, "pid": 76337, "tid": -914061504, "ts": 1716454223865915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865988, "dur": 1,
    "args": {
      "External id": 161884,
      "cbid": 251, "correlation": 161884
    }
  },
  {
    "ph": "f", "id": 161884, "pid": 76337, "tid": -914061504, "ts": 1716454223865988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223939642, "dur": 145,
    "args": {
      "External id": 161885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161885,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161885, "pid": 5, "tid": 7, "ts": 1716454223939642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223865992, "dur": 13,
    "args": {
      "External id": 161885,
      "cbid": 211, "correlation": 161885
    }
  },
  {
    "ph": "s", "id": 161885, "pid": 76337, "tid": -914061504, "ts": 1716454223865992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223939789, "dur": 1948,
    "args": {
      "External id": 161906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161906,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 161906, "pid": 5, "tid": 7, "ts": 1716454223939789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866075, "dur": 13,
    "args": {
      "External id": 161906,
      "cbid": 211, "correlation": 161906
    }
  },
  {
    "ph": "s", "id": 161906, "pid": 76337, "tid": -914061504, "ts": 1716454223866075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866176, "dur": 1,
    "args": {
      "External id": 161924,
      "cbid": 251, "correlation": 161924
    }
  },
  {
    "ph": "f", "id": 161924, "pid": 76337, "tid": -914061504, "ts": 1716454223866176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223941739, "dur": 148,
    "args": {
      "External id": 161926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161926,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 161926, "pid": 5, "tid": 7, "ts": 1716454223941739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866182, "dur": 13,
    "args": {
      "External id": 161926,
      "cbid": 211, "correlation": 161926
    }
  },
  {
    "ph": "s", "id": 161926, "pid": 76337, "tid": -914061504, "ts": 1716454223866182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223941888, "dur": 35,
    "args": {
      "External id": 161934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161934,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161934, "pid": 5, "tid": 7, "ts": 1716454223941888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866253, "dur": 12,
    "args": {
      "External id": 161934,
      "cbid": 211, "correlation": 161934
    }
  },
  {
    "ph": "s", "id": 161934, "pid": 76337, "tid": -914061504, "ts": 1716454223866253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223941925, "dur": 51,
    "args": {
      "External id": 161942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161942,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161942, "pid": 5, "tid": 7, "ts": 1716454223941925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866291, "dur": 9,
    "args": {
      "External id": 161942,
      "cbid": 211, "correlation": 161942
    }
  },
  {
    "ph": "s", "id": 161942, "pid": 76337, "tid": -914061504, "ts": 1716454223866291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223941978, "dur": 30,
    "args": {
      "External id": 161953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161953,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161953, "pid": 5, "tid": 7, "ts": 1716454223941978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866364, "dur": 13,
    "args": {
      "External id": 161953,
      "cbid": 211, "correlation": 161953
    }
  },
  {
    "ph": "s", "id": 161953, "pid": 76337, "tid": -914061504, "ts": 1716454223866364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223942009, "dur": 34,
    "args": {
      "External id": 161975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161975,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 161975, "pid": 5, "tid": 7, "ts": 1716454223942009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866395, "dur": 8,
    "args": {
      "External id": 161975,
      "cbid": 211, "correlation": 161975
    }
  },
  {
    "ph": "s", "id": 161975, "pid": 76337, "tid": -914061504, "ts": 1716454223866395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866479, "dur": 1,
    "args": {
      "External id": 161986,
      "cbid": 251, "correlation": 161986
    }
  },
  {
    "ph": "f", "id": 161986, "pid": 76337, "tid": -914061504, "ts": 1716454223866479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223942044, "dur": 77,
    "args": {
      "External id": 161987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 161987,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 161987, "pid": 5, "tid": 7, "ts": 1716454223942044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866484, "dur": 13,
    "args": {
      "External id": 161987,
      "cbid": 211, "correlation": 161987
    }
  },
  {
    "ph": "s", "id": 161987, "pid": 76337, "tid": -914061504, "ts": 1716454223866484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866553, "dur": 1,
    "args": {
      "External id": 161998,
      "cbid": 251, "correlation": 161998
    }
  },
  {
    "ph": "f", "id": 161998, "pid": 76337, "tid": -914061504, "ts": 1716454223866553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866557, "dur": 0,
    "args": {
      "External id": 161999,
      "cbid": 251, "correlation": 161999
    }
  },
  {
    "ph": "f", "id": 161999, "pid": 76337, "tid": -914061504, "ts": 1716454223866557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223942122, "dur": 12,
    "args": {
      "External id": 162000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162000,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 162000, "pid": 5, "tid": 7, "ts": 1716454223942122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866558, "dur": 11,
    "args": {
      "External id": 162000,
      "cbid": 211, "correlation": 162000
    }
  },
  {
    "ph": "s", "id": 162000, "pid": 76337, "tid": -914061504, "ts": 1716454223866558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223942135, "dur": 5,
    "args": {
      "External id": 162002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162002,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 162002, "pid": 5, "tid": 7, "ts": 1716454223942135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866571, "dur": 6,
    "args": {
      "External id": 162002,
      "cbid": 211, "correlation": 162002
    }
  },
  {
    "ph": "s", "id": 162002, "pid": 76337, "tid": -914061504, "ts": 1716454223866571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866630, "dur": 1,
    "args": {
      "External id": 162013,
      "cbid": 251, "correlation": 162013
    }
  },
  {
    "ph": "f", "id": 162013, "pid": 76337, "tid": -914061504, "ts": 1716454223866630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866634, "dur": 0,
    "args": {
      "External id": 162014,
      "cbid": 251, "correlation": 162014
    }
  },
  {
    "ph": "f", "id": 162014, "pid": 76337, "tid": -914061504, "ts": 1716454223866634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223942142, "dur": 7,
    "args": {
      "External id": 162015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162015,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 162015, "pid": 5, "tid": 7, "ts": 1716454223942142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866635, "dur": 13,
    "args": {
      "External id": 162015,
      "cbid": 211, "correlation": 162015
    }
  },
  {
    "ph": "s", "id": 162015, "pid": 76337, "tid": -914061504, "ts": 1716454223866635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223942150, "dur": 4,
    "args": {
      "External id": 162017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162017,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 162017, "pid": 5, "tid": 7, "ts": 1716454223942150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866649, "dur": 5,
    "args": {
      "External id": 162017,
      "cbid": 211, "correlation": 162017
    }
  },
  {
    "ph": "s", "id": 162017, "pid": 76337, "tid": -914061504, "ts": 1716454223866649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223942155, "dur": 91,
    "args": {
      "External id": 162038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162038,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 162038, "pid": 5, "tid": 7, "ts": 1716454223942155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866722, "dur": 12,
    "args": {
      "External id": 162038,
      "cbid": 211, "correlation": 162038
    }
  },
  {
    "ph": "s", "id": 162038, "pid": 76337, "tid": -914061504, "ts": 1716454223866722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866820, "dur": 1,
    "args": {
      "External id": 162056,
      "cbid": 251, "correlation": 162056
    }
  },
  {
    "ph": "f", "id": 162056, "pid": 76337, "tid": -914061504, "ts": 1716454223866820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223942248, "dur": 99,
    "args": {
      "External id": 162058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162058,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162058, "pid": 5, "tid": 7, "ts": 1716454223942248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866825, "dur": 13,
    "args": {
      "External id": 162058,
      "cbid": 211, "correlation": 162058
    }
  },
  {
    "ph": "s", "id": 162058, "pid": 76337, "tid": -914061504, "ts": 1716454223866825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223942348, "dur": 19,
    "args": {
      "External id": 162066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162066,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162066, "pid": 5, "tid": 7, "ts": 1716454223942348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866895, "dur": 12,
    "args": {
      "External id": 162066,
      "cbid": 211, "correlation": 162066
    }
  },
  {
    "ph": "s", "id": 162066, "pid": 76337, "tid": -914061504, "ts": 1716454223866895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223942369, "dur": 37,
    "args": {
      "External id": 162074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162074, "pid": 5, "tid": 7, "ts": 1716454223942369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866936, "dur": 10,
    "args": {
      "External id": 162074,
      "cbid": 211, "correlation": 162074
    }
  },
  {
    "ph": "s", "id": 162074, "pid": 76337, "tid": -914061504, "ts": 1716454223866936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223942407, "dur": 35,
    "args": {
      "External id": 162096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162096,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162096, "pid": 5, "tid": 7, "ts": 1716454223942407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223866997, "dur": 11,
    "args": {
      "External id": 162096,
      "cbid": 211, "correlation": 162096
    }
  },
  {
    "ph": "s", "id": 162096, "pid": 76337, "tid": -914061504, "ts": 1716454223866997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867088, "dur": 1,
    "args": {
      "External id": 162112,
      "cbid": 251, "correlation": 162112
    }
  },
  {
    "ph": "f", "id": 162112, "pid": 76337, "tid": -914061504, "ts": 1716454223867088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867093, "dur": 0,
    "args": {
      "External id": 162114,
      "cbid": 251, "correlation": 162114
    }
  },
  {
    "ph": "f", "id": 162114, "pid": 76337, "tid": -914061504, "ts": 1716454223867093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223942443, "dur": 536,
    "args": {
      "External id": 162115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162115,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 162115, "pid": 5, "tid": 7, "ts": 1716454223942443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867097, "dur": 13,
    "args": {
      "External id": 162115,
      "cbid": 211, "correlation": 162115
    }
  },
  {
    "ph": "s", "id": 162115, "pid": 76337, "tid": -914061504, "ts": 1716454223867097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223942981, "dur": 125,
    "args": {
      "External id": 162123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162123, "pid": 5, "tid": 7, "ts": 1716454223942981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867163, "dur": 12,
    "args": {
      "External id": 162123,
      "cbid": 211, "correlation": 162123
    }
  },
  {
    "ph": "s", "id": 162123, "pid": 76337, "tid": -914061504, "ts": 1716454223867163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223943107, "dur": 129,
    "args": {
      "External id": 162131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162131,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162131, "pid": 5, "tid": 7, "ts": 1716454223943107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867194, "dur": 8,
    "args": {
      "External id": 162131,
      "cbid": 211, "correlation": 162131
    }
  },
  {
    "ph": "s", "id": 162131, "pid": 76337, "tid": -914061504, "ts": 1716454223867194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867270, "dur": 1,
    "args": {
      "External id": 162147,
      "cbid": 251, "correlation": 162147
    }
  },
  {
    "ph": "f", "id": 162147, "pid": 76337, "tid": -914061504, "ts": 1716454223867270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223943237, "dur": 306,
    "args": {
      "External id": 162149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162149,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162149, "pid": 5, "tid": 7, "ts": 1716454223943237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867276, "dur": 12,
    "args": {
      "External id": 162149,
      "cbid": 211, "correlation": 162149
    }
  },
  {
    "ph": "s", "id": 162149, "pid": 76337, "tid": -914061504, "ts": 1716454223867276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223943543, "dur": 27,
    "args": {
      "External id": 162157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162157,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162157, "pid": 5, "tid": 7, "ts": 1716454223943543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867318, "dur": 10,
    "args": {
      "External id": 162157,
      "cbid": 211, "correlation": 162157
    }
  },
  {
    "ph": "s", "id": 162157, "pid": 76337, "tid": -914061504, "ts": 1716454223867318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223943572, "dur": 81,
    "args": {
      "External id": 162168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162168, "pid": 5, "tid": 7, "ts": 1716454223943572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867385, "dur": 13,
    "args": {
      "External id": 162168,
      "cbid": 211, "correlation": 162168
    }
  },
  {
    "ph": "s", "id": 162168, "pid": 76337, "tid": -914061504, "ts": 1716454223867385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867449, "dur": 0,
    "args": {
      "External id": 162180,
      "cbid": 317, "correlation": 162180
    }
  },
  {
    "ph": "f", "id": 162180, "pid": 76337, "tid": -914061504, "ts": 1716454223867449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867450, "dur": 0,
    "args": {
      "External id": 162181,
      "cbid": 203, "correlation": 162181
    }
  },
  {
    "ph": "f", "id": 162181, "pid": 76337, "tid": -914061504, "ts": 1716454223867450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867451, "dur": 0,
    "args": {
      "External id": 162182,
      "cbid": 205, "correlation": 162182
    }
  },
  {
    "ph": "f", "id": 162182, "pid": 76337, "tid": -914061504, "ts": 1716454223867451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223943654, "dur": 24,
    "args": {
      "External id": 162186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162186,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162186, "pid": 5, "tid": 7, "ts": 1716454223943654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867465, "dur": 12,
    "args": {
      "External id": 162186,
      "cbid": 211, "correlation": 162186
    }
  },
  {
    "ph": "s", "id": 162186, "pid": 76337, "tid": -914061504, "ts": 1716454223867465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223943680, "dur": 120,
    "args": {
      "External id": 162188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162188,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162188, "pid": 5, "tid": 7, "ts": 1716454223943680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867484, "dur": 7,
    "args": {
      "External id": 162188,
      "cbid": 211, "correlation": 162188
    }
  },
  {
    "ph": "s", "id": 162188, "pid": 76337, "tid": -914061504, "ts": 1716454223867484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223943802, "dur": 23,
    "args": {
      "External id": 162190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162190, "pid": 5, "tid": 7, "ts": 1716454223943802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867495, "dur": 5,
    "args": {
      "External id": 162190,
      "cbid": 211, "correlation": 162190
    }
  },
  {
    "ph": "s", "id": 162190, "pid": 76337, "tid": -914061504, "ts": 1716454223867495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223943826, "dur": 33,
    "args": {
      "External id": 162196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162196,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162196, "pid": 5, "tid": 7, "ts": 1716454223943826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867522, "dur": 8,
    "args": {
      "External id": 162196,
      "cbid": 211, "correlation": 162196
    }
  },
  {
    "ph": "s", "id": 162196, "pid": 76337, "tid": -914061504, "ts": 1716454223867522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223943861, "dur": 26,
    "args": {
      "External id": 162204,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162204,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162204, "pid": 5, "tid": 7, "ts": 1716454223943861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867554, "dur": 8,
    "args": {
      "External id": 162204,
      "cbid": 211, "correlation": 162204
    }
  },
  {
    "ph": "s", "id": 162204, "pid": 76337, "tid": -914061504, "ts": 1716454223867554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223943888, "dur": 48,
    "args": {
      "External id": 162213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162213,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162213, "pid": 5, "tid": 7, "ts": 1716454223943888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867592, "dur": 10,
    "args": {
      "External id": 162213,
      "cbid": 211, "correlation": 162213
    }
  },
  {
    "ph": "s", "id": 162213, "pid": 76337, "tid": -914061504, "ts": 1716454223867592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223943938, "dur": 43,
    "args": {
      "External id": 162233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162233,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 162233, "pid": 5, "tid": 7, "ts": 1716454223943938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867663, "dur": 11,
    "args": {
      "External id": 162233,
      "cbid": 211, "correlation": 162233
    }
  },
  {
    "ph": "s", "id": 162233, "pid": 76337, "tid": -914061504, "ts": 1716454223867663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223943982, "dur": 5,
    "args": {
      "External id": 162245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162245,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 162245, "pid": 5, "tid": 7, "ts": 1716454223943982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867683, "dur": 6,
    "args": {
      "External id": 162245,
      "cbid": 211, "correlation": 162245
    }
  },
  {
    "ph": "s", "id": 162245, "pid": 76337, "tid": -914061504, "ts": 1716454223867683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223943989, "dur": 43,
    "args": {
      "External id": 162248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162248, "pid": 5, "tid": 7, "ts": 1716454223943989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867702, "dur": 6,
    "args": {
      "External id": 162248,
      "cbid": 211, "correlation": 162248
    }
  },
  {
    "ph": "s", "id": 162248, "pid": 76337, "tid": -914061504, "ts": 1716454223867702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223944034, "dur": 29,
    "args": {
      "External id": 162257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162257,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162257, "pid": 5, "tid": 7, "ts": 1716454223944034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867741, "dur": 10,
    "args": {
      "External id": 162257,
      "cbid": 211, "correlation": 162257
    }
  },
  {
    "ph": "s", "id": 162257, "pid": 76337, "tid": -914061504, "ts": 1716454223867741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867792, "dur": 0,
    "args": {
      "External id": 162267,
      "cbid": 317, "correlation": 162267
    }
  },
  {
    "ph": "f", "id": 162267, "pid": 76337, "tid": -914061504, "ts": 1716454223867792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867793, "dur": 0,
    "args": {
      "External id": 162268,
      "cbid": 203, "correlation": 162268
    }
  },
  {
    "ph": "f", "id": 162268, "pid": 76337, "tid": -914061504, "ts": 1716454223867793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867794, "dur": 0,
    "args": {
      "External id": 162269,
      "cbid": 205, "correlation": 162269
    }
  },
  {
    "ph": "f", "id": 162269, "pid": 76337, "tid": -914061504, "ts": 1716454223867794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223944064, "dur": 32,
    "args": {
      "External id": 162273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162273,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162273, "pid": 5, "tid": 7, "ts": 1716454223944064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867808, "dur": 11,
    "args": {
      "External id": 162273,
      "cbid": 211, "correlation": 162273
    }
  },
  {
    "ph": "s", "id": 162273, "pid": 76337, "tid": -914061504, "ts": 1716454223867808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223944097, "dur": 64,
    "args": {
      "External id": 162275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162275,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162275, "pid": 5, "tid": 7, "ts": 1716454223944097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867822, "dur": 5,
    "args": {
      "External id": 162275,
      "cbid": 211, "correlation": 162275
    }
  },
  {
    "ph": "s", "id": 162275, "pid": 76337, "tid": -914061504, "ts": 1716454223867822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223944162, "dur": 968,
    "args": {
      "External id": 162277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162277,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162277, "pid": 5, "tid": 7, "ts": 1716454223944162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867833, "dur": 7,
    "args": {
      "External id": 162277,
      "cbid": 211, "correlation": 162277
    }
  },
  {
    "ph": "s", "id": 162277, "pid": 76337, "tid": -914061504, "ts": 1716454223867833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223945131, "dur": 21,
    "args": {
      "External id": 162279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162279,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162279, "pid": 5, "tid": 7, "ts": 1716454223945131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867845, "dur": 5,
    "args": {
      "External id": 162279,
      "cbid": 211, "correlation": 162279
    }
  },
  {
    "ph": "s", "id": 162279, "pid": 76337, "tid": -914061504, "ts": 1716454223867845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223945154, "dur": 33,
    "args": {
      "External id": 162285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162285,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162285, "pid": 5, "tid": 7, "ts": 1716454223945154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867871, "dur": 8,
    "args": {
      "External id": 162285,
      "cbid": 211, "correlation": 162285
    }
  },
  {
    "ph": "s", "id": 162285, "pid": 76337, "tid": -914061504, "ts": 1716454223867871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223945188, "dur": 3,
    "args": {
      "External id": 162293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162293,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 162293, "pid": 5, "tid": 7, "ts": 1716454223945188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867914, "dur": 9,
    "args": {
      "External id": 162293,
      "cbid": 211, "correlation": 162293
    }
  },
  {
    "ph": "s", "id": 162293, "pid": 76337, "tid": -914061504, "ts": 1716454223867914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867987, "dur": 1,
    "args": {
      "External id": 162309,
      "cbid": 251, "correlation": 162309
    }
  },
  {
    "ph": "f", "id": 162309, "pid": 76337, "tid": -914061504, "ts": 1716454223867987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867993, "dur": 0,
    "args": {
      "External id": 162311,
      "cbid": 251, "correlation": 162311
    }
  },
  {
    "ph": "f", "id": 162311, "pid": 76337, "tid": -914061504, "ts": 1716454223867993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223945193, "dur": 12,
    "args": {
      "External id": 162312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162312,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 162312, "pid": 5, "tid": 7, "ts": 1716454223945193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223867995, "dur": 12,
    "args": {
      "External id": 162312,
      "cbid": 211, "correlation": 162312
    }
  },
  {
    "ph": "s", "id": 162312, "pid": 76337, "tid": -914061504, "ts": 1716454223867995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223945206, "dur": 5,
    "args": {
      "External id": 162314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162314,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 162314, "pid": 5, "tid": 7, "ts": 1716454223945206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868008, "dur": 5,
    "args": {
      "External id": 162314,
      "cbid": 211, "correlation": 162314
    }
  },
  {
    "ph": "s", "id": 162314, "pid": 76337, "tid": -914061504, "ts": 1716454223868008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223945213, "dur": 29,
    "args": {
      "External id": 162324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162324,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162324, "pid": 5, "tid": 7, "ts": 1716454223945213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868066, "dur": 13,
    "args": {
      "External id": 162324,
      "cbid": 211, "correlation": 162324
    }
  },
  {
    "ph": "s", "id": 162324, "pid": 76337, "tid": -914061504, "ts": 1716454223868066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223945243, "dur": 32,
    "args": {
      "External id": 162344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162344,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 162344, "pid": 5, "tid": 7, "ts": 1716454223945243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868133, "dur": 11,
    "args": {
      "External id": 162344,
      "cbid": 211, "correlation": 162344
    }
  },
  {
    "ph": "s", "id": 162344, "pid": 76337, "tid": -914061504, "ts": 1716454223868133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223945276, "dur": 4,
    "args": {
      "External id": 162356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162356,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 162356, "pid": 5, "tid": 7, "ts": 1716454223945276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868153, "dur": 6,
    "args": {
      "External id": 162356,
      "cbid": 211, "correlation": 162356
    }
  },
  {
    "ph": "s", "id": 162356, "pid": 76337, "tid": -914061504, "ts": 1716454223868153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223945281, "dur": 30,
    "args": {
      "External id": 162359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162359,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162359, "pid": 5, "tid": 7, "ts": 1716454223945281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868171, "dur": 7,
    "args": {
      "External id": 162359,
      "cbid": 211, "correlation": 162359
    }
  },
  {
    "ph": "s", "id": 162359, "pid": 76337, "tid": -914061504, "ts": 1716454223868171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223945313, "dur": 22,
    "args": {
      "External id": 162368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162368,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162368, "pid": 5, "tid": 7, "ts": 1716454223945313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868211, "dur": 10,
    "args": {
      "External id": 162368,
      "cbid": 211, "correlation": 162368
    }
  },
  {
    "ph": "s", "id": 162368, "pid": 76337, "tid": -914061504, "ts": 1716454223868211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868274, "dur": 0,
    "args": {
      "External id": 162378,
      "cbid": 317, "correlation": 162378
    }
  },
  {
    "ph": "f", "id": 162378, "pid": 76337, "tid": -914061504, "ts": 1716454223868274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868275, "dur": 0,
    "args": {
      "External id": 162379,
      "cbid": 203, "correlation": 162379
    }
  },
  {
    "ph": "f", "id": 162379, "pid": 76337, "tid": -914061504, "ts": 1716454223868275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868275, "dur": 0,
    "args": {
      "External id": 162380,
      "cbid": 205, "correlation": 162380
    }
  },
  {
    "ph": "f", "id": 162380, "pid": 76337, "tid": -914061504, "ts": 1716454223868275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223945336, "dur": 22,
    "args": {
      "External id": 162384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162384,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162384, "pid": 5, "tid": 7, "ts": 1716454223945336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868290, "dur": 12,
    "args": {
      "External id": 162384,
      "cbid": 211, "correlation": 162384
    }
  },
  {
    "ph": "s", "id": 162384, "pid": 76337, "tid": -914061504, "ts": 1716454223868290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223945359, "dur": 45,
    "args": {
      "External id": 162386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162386,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162386, "pid": 5, "tid": 7, "ts": 1716454223945359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868304, "dur": 5,
    "args": {
      "External id": 162386,
      "cbid": 211, "correlation": 162386
    }
  },
  {
    "ph": "s", "id": 162386, "pid": 76337, "tid": -914061504, "ts": 1716454223868304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223945405, "dur": 646,
    "args": {
      "External id": 162388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162388,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162388, "pid": 5, "tid": 7, "ts": 1716454223945405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868315, "dur": 6,
    "args": {
      "External id": 162388,
      "cbid": 211, "correlation": 162388
    }
  },
  {
    "ph": "s", "id": 162388, "pid": 76337, "tid": -914061504, "ts": 1716454223868315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946053, "dur": 23,
    "args": {
      "External id": 162390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162390,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162390, "pid": 5, "tid": 7, "ts": 1716454223946053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868325, "dur": 5,
    "args": {
      "External id": 162390,
      "cbid": 211, "correlation": 162390
    }
  },
  {
    "ph": "s", "id": 162390, "pid": 76337, "tid": -914061504, "ts": 1716454223868325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223946077, "dur": 32,
    "args": {
      "External id": 162396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162396, "pid": 5, "tid": 7, "ts": 1716454223946077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868353, "dur": 8,
    "args": {
      "External id": 162396,
      "cbid": 211, "correlation": 162396
    }
  },
  {
    "ph": "s", "id": 162396, "pid": 76337, "tid": -914061504, "ts": 1716454223868353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868411, "dur": 0,
    "args": {
      "External id": 162406,
      "cbid": 317, "correlation": 162406
    }
  },
  {
    "ph": "f", "id": 162406, "pid": 76337, "tid": -914061504, "ts": 1716454223868411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868411, "dur": 0,
    "args": {
      "External id": 162407,
      "cbid": 203, "correlation": 162407
    }
  },
  {
    "ph": "f", "id": 162407, "pid": 76337, "tid": -914061504, "ts": 1716454223868411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868412, "dur": 0,
    "args": {
      "External id": 162408,
      "cbid": 205, "correlation": 162408
    }
  },
  {
    "ph": "f", "id": 162408, "pid": 76337, "tid": -914061504, "ts": 1716454223868412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946110, "dur": 30,
    "args": {
      "External id": 162412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162412,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162412, "pid": 5, "tid": 7, "ts": 1716454223946110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868426, "dur": 12,
    "args": {
      "External id": 162412,
      "cbid": 211, "correlation": 162412
    }
  },
  {
    "ph": "s", "id": 162412, "pid": 76337, "tid": -914061504, "ts": 1716454223868426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223946141, "dur": 153,
    "args": {
      "External id": 162414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162414,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162414, "pid": 5, "tid": 7, "ts": 1716454223946141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868445, "dur": 7,
    "args": {
      "External id": 162414,
      "cbid": 211, "correlation": 162414
    }
  },
  {
    "ph": "s", "id": 162414, "pid": 76337, "tid": -914061504, "ts": 1716454223868445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946296, "dur": 22,
    "args": {
      "External id": 162416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162416,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162416, "pid": 5, "tid": 7, "ts": 1716454223946296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868455, "dur": 5,
    "args": {
      "External id": 162416,
      "cbid": 211, "correlation": 162416
    }
  },
  {
    "ph": "s", "id": 162416, "pid": 76337, "tid": -914061504, "ts": 1716454223868455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223946319, "dur": 32,
    "args": {
      "External id": 162422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162422, "pid": 5, "tid": 7, "ts": 1716454223946319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868480, "dur": 9,
    "args": {
      "External id": 162422,
      "cbid": 211, "correlation": 162422
    }
  },
  {
    "ph": "s", "id": 162422, "pid": 76337, "tid": -914061504, "ts": 1716454223868480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223946352, "dur": 27,
    "args": {
      "External id": 162430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162430,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162430, "pid": 5, "tid": 7, "ts": 1716454223946352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868509, "dur": 8,
    "args": {
      "External id": 162430,
      "cbid": 211, "correlation": 162430
    }
  },
  {
    "ph": "s", "id": 162430, "pid": 76337, "tid": -914061504, "ts": 1716454223868509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223946381, "dur": 20,
    "args": {
      "External id": 162438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162438,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162438, "pid": 5, "tid": 7, "ts": 1716454223946381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868537, "dur": 8,
    "args": {
      "External id": 162438,
      "cbid": 211, "correlation": 162438
    }
  },
  {
    "ph": "s", "id": 162438, "pid": 76337, "tid": -914061504, "ts": 1716454223868537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946402, "dur": 30,
    "args": {
      "External id": 162458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162458,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 162458, "pid": 5, "tid": 7, "ts": 1716454223946402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868618, "dur": 12,
    "args": {
      "External id": 162458,
      "cbid": 211, "correlation": 162458
    }
  },
  {
    "ph": "s", "id": 162458, "pid": 76337, "tid": -914061504, "ts": 1716454223868618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223946433, "dur": 4,
    "args": {
      "External id": 162470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162470,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 162470, "pid": 5, "tid": 7, "ts": 1716454223946433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868640, "dur": 7,
    "args": {
      "External id": 162470,
      "cbid": 211, "correlation": 162470
    }
  },
  {
    "ph": "s", "id": 162470, "pid": 76337, "tid": -914061504, "ts": 1716454223868640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223946439, "dur": 31,
    "args": {
      "External id": 162473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162473, "pid": 5, "tid": 7, "ts": 1716454223946439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868659, "dur": 6,
    "args": {
      "External id": 162473,
      "cbid": 211, "correlation": 162473
    }
  },
  {
    "ph": "s", "id": 162473, "pid": 76337, "tid": -914061504, "ts": 1716454223868659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868715, "dur": 0,
    "args": {
      "External id": 162484,
      "cbid": 317, "correlation": 162484
    }
  },
  {
    "ph": "f", "id": 162484, "pid": 76337, "tid": -914061504, "ts": 1716454223868715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868716, "dur": 0,
    "args": {
      "External id": 162485,
      "cbid": 203, "correlation": 162485
    }
  },
  {
    "ph": "f", "id": 162485, "pid": 76337, "tid": -914061504, "ts": 1716454223868716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868717, "dur": 0,
    "args": {
      "External id": 162486,
      "cbid": 205, "correlation": 162486
    }
  },
  {
    "ph": "f", "id": 162486, "pid": 76337, "tid": -914061504, "ts": 1716454223868717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946471, "dur": 21,
    "args": {
      "External id": 162490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162490, "pid": 5, "tid": 7, "ts": 1716454223946471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868730, "dur": 12,
    "args": {
      "External id": 162490,
      "cbid": 211, "correlation": 162490
    }
  },
  {
    "ph": "s", "id": 162490, "pid": 76337, "tid": -914061504, "ts": 1716454223868730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223946493, "dur": 105,
    "args": {
      "External id": 162492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162492,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162492, "pid": 5, "tid": 7, "ts": 1716454223946493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868748, "dur": 6,
    "args": {
      "External id": 162492,
      "cbid": 211, "correlation": 162492
    }
  },
  {
    "ph": "s", "id": 162492, "pid": 76337, "tid": -914061504, "ts": 1716454223868748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946600, "dur": 22,
    "args": {
      "External id": 162494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162494,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162494, "pid": 5, "tid": 7, "ts": 1716454223946600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868757, "dur": 5,
    "args": {
      "External id": 162494,
      "cbid": 211, "correlation": 162494
    }
  },
  {
    "ph": "s", "id": 162494, "pid": 76337, "tid": -914061504, "ts": 1716454223868757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223946624, "dur": 33,
    "args": {
      "External id": 162500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162500, "pid": 5, "tid": 7, "ts": 1716454223946624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868784, "dur": 9,
    "args": {
      "External id": 162500,
      "cbid": 211, "correlation": 162500
    }
  },
  {
    "ph": "s", "id": 162500, "pid": 76337, "tid": -914061504, "ts": 1716454223868784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223946658, "dur": 183,
    "args": {
      "External id": 162509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162509,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162509, "pid": 5, "tid": 7, "ts": 1716454223946658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868866, "dur": 14,
    "args": {
      "External id": 162509,
      "cbid": 211, "correlation": 162509
    }
  },
  {
    "ph": "s", "id": 162509, "pid": 76337, "tid": -914061504, "ts": 1716454223868866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223946842, "dur": 65,
    "args": {
      "External id": 162531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162531,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162531, "pid": 5, "tid": 7, "ts": 1716454223946842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223868922, "dur": 10,
    "args": {
      "External id": 162531,
      "cbid": 211, "correlation": 162531
    }
  },
  {
    "ph": "s", "id": 162531, "pid": 76337, "tid": -914061504, "ts": 1716454223868922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869019, "dur": 1,
    "args": {
      "External id": 162542,
      "cbid": 251, "correlation": 162542
    }
  },
  {
    "ph": "f", "id": 162542, "pid": 76337, "tid": -914061504, "ts": 1716454223869019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223946909, "dur": 151,
    "args": {
      "External id": 162543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162543,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162543, "pid": 5, "tid": 7, "ts": 1716454223946909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869024, "dur": 13,
    "args": {
      "External id": 162543,
      "cbid": 211, "correlation": 162543
    }
  },
  {
    "ph": "s", "id": 162543, "pid": 76337, "tid": -914061504, "ts": 1716454223869024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869095, "dur": 1,
    "args": {
      "External id": 162554,
      "cbid": 251, "correlation": 162554
    }
  },
  {
    "ph": "f", "id": 162554, "pid": 76337, "tid": -914061504, "ts": 1716454223869095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223947060, "dur": 143,
    "args": {
      "External id": 162555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162555,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162555, "pid": 5, "tid": 7, "ts": 1716454223947060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869099, "dur": 11,
    "args": {
      "External id": 162555,
      "cbid": 211, "correlation": 162555
    }
  },
  {
    "ph": "s", "id": 162555, "pid": 76337, "tid": -914061504, "ts": 1716454223869099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869163, "dur": 1,
    "args": {
      "External id": 162566,
      "cbid": 251, "correlation": 162566
    }
  },
  {
    "ph": "f", "id": 162566, "pid": 76337, "tid": -914061504, "ts": 1716454223869163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223947205, "dur": 145,
    "args": {
      "External id": 162567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162567,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162567, "pid": 5, "tid": 7, "ts": 1716454223947205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869167, "dur": 11,
    "args": {
      "External id": 162567,
      "cbid": 211, "correlation": 162567
    }
  },
  {
    "ph": "s", "id": 162567, "pid": 76337, "tid": -914061504, "ts": 1716454223869167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223947352, "dur": 1950,
    "args": {
      "External id": 162588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162588,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 162588, "pid": 5, "tid": 7, "ts": 1716454223947352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869248, "dur": 13,
    "args": {
      "External id": 162588,
      "cbid": 211, "correlation": 162588
    }
  },
  {
    "ph": "s", "id": 162588, "pid": 76337, "tid": -914061504, "ts": 1716454223869248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869349, "dur": 1,
    "args": {
      "External id": 162606,
      "cbid": 251, "correlation": 162606
    }
  },
  {
    "ph": "f", "id": 162606, "pid": 76337, "tid": -914061504, "ts": 1716454223869349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223949303, "dur": 146,
    "args": {
      "External id": 162608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162608,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 162608, "pid": 5, "tid": 7, "ts": 1716454223949303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869355, "dur": 13,
    "args": {
      "External id": 162608,
      "cbid": 211, "correlation": 162608
    }
  },
  {
    "ph": "s", "id": 162608, "pid": 76337, "tid": -914061504, "ts": 1716454223869355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223949451, "dur": 36,
    "args": {
      "External id": 162616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162616,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162616, "pid": 5, "tid": 7, "ts": 1716454223949451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869425, "dur": 12,
    "args": {
      "External id": 162616,
      "cbid": 211, "correlation": 162616
    }
  },
  {
    "ph": "s", "id": 162616, "pid": 76337, "tid": -914061504, "ts": 1716454223869425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223949488, "dur": 50,
    "args": {
      "External id": 162624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162624,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162624, "pid": 5, "tid": 7, "ts": 1716454223949488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869465, "dur": 8,
    "args": {
      "External id": 162624,
      "cbid": 211, "correlation": 162624
    }
  },
  {
    "ph": "s", "id": 162624, "pid": 76337, "tid": -914061504, "ts": 1716454223869465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223949540, "dur": 30,
    "args": {
      "External id": 162635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162635,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162635, "pid": 5, "tid": 7, "ts": 1716454223949540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869535, "dur": 13,
    "args": {
      "External id": 162635,
      "cbid": 211, "correlation": 162635
    }
  },
  {
    "ph": "s", "id": 162635, "pid": 76337, "tid": -914061504, "ts": 1716454223869535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223949571, "dur": 34,
    "args": {
      "External id": 162657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162657,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162657, "pid": 5, "tid": 7, "ts": 1716454223949571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869566, "dur": 7,
    "args": {
      "External id": 162657,
      "cbid": 211, "correlation": 162657
    }
  },
  {
    "ph": "s", "id": 162657, "pid": 76337, "tid": -914061504, "ts": 1716454223869566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869651, "dur": 1,
    "args": {
      "External id": 162668,
      "cbid": 251, "correlation": 162668
    }
  },
  {
    "ph": "f", "id": 162668, "pid": 76337, "tid": -914061504, "ts": 1716454223869651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223949607, "dur": 76,
    "args": {
      "External id": 162669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162669,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162669, "pid": 5, "tid": 7, "ts": 1716454223949607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869656, "dur": 13,
    "args": {
      "External id": 162669,
      "cbid": 211, "correlation": 162669
    }
  },
  {
    "ph": "s", "id": 162669, "pid": 76337, "tid": -914061504, "ts": 1716454223869656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869725, "dur": 1,
    "args": {
      "External id": 162680,
      "cbid": 251, "correlation": 162680
    }
  },
  {
    "ph": "f", "id": 162680, "pid": 76337, "tid": -914061504, "ts": 1716454223869725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869728, "dur": 0,
    "args": {
      "External id": 162681,
      "cbid": 251, "correlation": 162681
    }
  },
  {
    "ph": "f", "id": 162681, "pid": 76337, "tid": -914061504, "ts": 1716454223869728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223949684, "dur": 11,
    "args": {
      "External id": 162682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162682,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 162682, "pid": 5, "tid": 7, "ts": 1716454223949684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869730, "dur": 12,
    "args": {
      "External id": 162682,
      "cbid": 211, "correlation": 162682
    }
  },
  {
    "ph": "s", "id": 162682, "pid": 76337, "tid": -914061504, "ts": 1716454223869730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223949697, "dur": 5,
    "args": {
      "External id": 162684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162684,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 162684, "pid": 5, "tid": 7, "ts": 1716454223949697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869744, "dur": 5,
    "args": {
      "External id": 162684,
      "cbid": 211, "correlation": 162684
    }
  },
  {
    "ph": "s", "id": 162684, "pid": 76337, "tid": -914061504, "ts": 1716454223869744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869801, "dur": 1,
    "args": {
      "External id": 162695,
      "cbid": 251, "correlation": 162695
    }
  },
  {
    "ph": "f", "id": 162695, "pid": 76337, "tid": -914061504, "ts": 1716454223869801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869805, "dur": 0,
    "args": {
      "External id": 162696,
      "cbid": 251, "correlation": 162696
    }
  },
  {
    "ph": "f", "id": 162696, "pid": 76337, "tid": -914061504, "ts": 1716454223869805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223949703, "dur": 7,
    "args": {
      "External id": 162697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162697,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 162697, "pid": 5, "tid": 7, "ts": 1716454223949703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869807, "dur": 12,
    "args": {
      "External id": 162697,
      "cbid": 211, "correlation": 162697
    }
  },
  {
    "ph": "s", "id": 162697, "pid": 76337, "tid": -914061504, "ts": 1716454223869807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223949712, "dur": 4,
    "args": {
      "External id": 162699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162699,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 162699, "pid": 5, "tid": 7, "ts": 1716454223949712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869820, "dur": 5,
    "args": {
      "External id": 162699,
      "cbid": 211, "correlation": 162699
    }
  },
  {
    "ph": "s", "id": 162699, "pid": 76337, "tid": -914061504, "ts": 1716454223869820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223949717, "dur": 92,
    "args": {
      "External id": 162720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162720,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 162720, "pid": 5, "tid": 7, "ts": 1716454223949717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869893, "dur": 13,
    "args": {
      "External id": 162720,
      "cbid": 211, "correlation": 162720
    }
  },
  {
    "ph": "s", "id": 162720, "pid": 76337, "tid": -914061504, "ts": 1716454223869893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223869999, "dur": 1,
    "args": {
      "External id": 162738,
      "cbid": 251, "correlation": 162738
    }
  },
  {
    "ph": "f", "id": 162738, "pid": 76337, "tid": -914061504, "ts": 1716454223869999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223949810, "dur": 100,
    "args": {
      "External id": 162740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162740,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162740, "pid": 5, "tid": 7, "ts": 1716454223949810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870005, "dur": 14,
    "args": {
      "External id": 162740,
      "cbid": 211, "correlation": 162740
    }
  },
  {
    "ph": "s", "id": 162740, "pid": 76337, "tid": -914061504, "ts": 1716454223870005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223949912, "dur": 19,
    "args": {
      "External id": 162748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162748,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162748, "pid": 5, "tid": 7, "ts": 1716454223949912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870076, "dur": 12,
    "args": {
      "External id": 162748,
      "cbid": 211, "correlation": 162748
    }
  },
  {
    "ph": "s", "id": 162748, "pid": 76337, "tid": -914061504, "ts": 1716454223870076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223949932, "dur": 38,
    "args": {
      "External id": 162756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162756,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162756, "pid": 5, "tid": 7, "ts": 1716454223949932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870117, "dur": 9,
    "args": {
      "External id": 162756,
      "cbid": 211, "correlation": 162756
    }
  },
  {
    "ph": "s", "id": 162756, "pid": 76337, "tid": -914061504, "ts": 1716454223870117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223949972, "dur": 34,
    "args": {
      "External id": 162778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162778,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162778, "pid": 5, "tid": 7, "ts": 1716454223949972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870168, "dur": 10,
    "args": {
      "External id": 162778,
      "cbid": 211, "correlation": 162778
    }
  },
  {
    "ph": "s", "id": 162778, "pid": 76337, "tid": -914061504, "ts": 1716454223870168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870258, "dur": 1,
    "args": {
      "External id": 162794,
      "cbid": 251, "correlation": 162794
    }
  },
  {
    "ph": "f", "id": 162794, "pid": 76337, "tid": -914061504, "ts": 1716454223870258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870263, "dur": 0,
    "args": {
      "External id": 162796,
      "cbid": 251, "correlation": 162796
    }
  },
  {
    "ph": "f", "id": 162796, "pid": 76337, "tid": -914061504, "ts": 1716454223870263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223950007, "dur": 539,
    "args": {
      "External id": 162797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162797,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 162797, "pid": 5, "tid": 7, "ts": 1716454223950007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870267, "dur": 13,
    "args": {
      "External id": 162797,
      "cbid": 211, "correlation": 162797
    }
  },
  {
    "ph": "s", "id": 162797, "pid": 76337, "tid": -914061504, "ts": 1716454223870267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223950547, "dur": 127,
    "args": {
      "External id": 162805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162805,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162805, "pid": 5, "tid": 7, "ts": 1716454223950547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870332, "dur": 12,
    "args": {
      "External id": 162805,
      "cbid": 211, "correlation": 162805
    }
  },
  {
    "ph": "s", "id": 162805, "pid": 76337, "tid": -914061504, "ts": 1716454223870332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223950675, "dur": 130,
    "args": {
      "External id": 162813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162813, "pid": 5, "tid": 7, "ts": 1716454223950675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870363, "dur": 8,
    "args": {
      "External id": 162813,
      "cbid": 211, "correlation": 162813
    }
  },
  {
    "ph": "s", "id": 162813, "pid": 76337, "tid": -914061504, "ts": 1716454223870363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870439, "dur": 1,
    "args": {
      "External id": 162829,
      "cbid": 251, "correlation": 162829
    }
  },
  {
    "ph": "f", "id": 162829, "pid": 76337, "tid": -914061504, "ts": 1716454223870439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223950807, "dur": 302,
    "args": {
      "External id": 162831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162831,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162831, "pid": 5, "tid": 7, "ts": 1716454223950807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870444, "dur": 13,
    "args": {
      "External id": 162831,
      "cbid": 211, "correlation": 162831
    }
  },
  {
    "ph": "s", "id": 162831, "pid": 76337, "tid": -914061504, "ts": 1716454223870444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223951110, "dur": 27,
    "args": {
      "External id": 162839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162839,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162839, "pid": 5, "tid": 7, "ts": 1716454223951110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870487, "dur": 10,
    "args": {
      "External id": 162839,
      "cbid": 211, "correlation": 162839
    }
  },
  {
    "ph": "s", "id": 162839, "pid": 76337, "tid": -914061504, "ts": 1716454223870487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223951139, "dur": 81,
    "args": {
      "External id": 162850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162850,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162850, "pid": 5, "tid": 7, "ts": 1716454223951139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870553, "dur": 12,
    "args": {
      "External id": 162850,
      "cbid": 211, "correlation": 162850
    }
  },
  {
    "ph": "s", "id": 162850, "pid": 76337, "tid": -914061504, "ts": 1716454223870553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870616, "dur": 0,
    "args": {
      "External id": 162862,
      "cbid": 317, "correlation": 162862
    }
  },
  {
    "ph": "f", "id": 162862, "pid": 76337, "tid": -914061504, "ts": 1716454223870616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870617, "dur": 0,
    "args": {
      "External id": 162863,
      "cbid": 203, "correlation": 162863
    }
  },
  {
    "ph": "f", "id": 162863, "pid": 76337, "tid": -914061504, "ts": 1716454223870617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870618, "dur": 0,
    "args": {
      "External id": 162864,
      "cbid": 205, "correlation": 162864
    }
  },
  {
    "ph": "f", "id": 162864, "pid": 76337, "tid": -914061504, "ts": 1716454223870618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223951221, "dur": 23,
    "args": {
      "External id": 162868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162868,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162868, "pid": 5, "tid": 7, "ts": 1716454223951221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870632, "dur": 12,
    "args": {
      "External id": 162868,
      "cbid": 211, "correlation": 162868
    }
  },
  {
    "ph": "s", "id": 162868, "pid": 76337, "tid": -914061504, "ts": 1716454223870632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223951246, "dur": 120,
    "args": {
      "External id": 162870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162870,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162870, "pid": 5, "tid": 7, "ts": 1716454223951246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870651, "dur": 7,
    "args": {
      "External id": 162870,
      "cbid": 211, "correlation": 162870
    }
  },
  {
    "ph": "s", "id": 162870, "pid": 76337, "tid": -914061504, "ts": 1716454223870651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223951367, "dur": 24,
    "args": {
      "External id": 162872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162872,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162872, "pid": 5, "tid": 7, "ts": 1716454223951367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870662, "dur": 5,
    "args": {
      "External id": 162872,
      "cbid": 211, "correlation": 162872
    }
  },
  {
    "ph": "s", "id": 162872, "pid": 76337, "tid": -914061504, "ts": 1716454223870662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223951393, "dur": 33,
    "args": {
      "External id": 162878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162878, "pid": 5, "tid": 7, "ts": 1716454223951393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870689, "dur": 9,
    "args": {
      "External id": 162878,
      "cbid": 211, "correlation": 162878
    }
  },
  {
    "ph": "s", "id": 162878, "pid": 76337, "tid": -914061504, "ts": 1716454223870689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223951427, "dur": 27,
    "args": {
      "External id": 162886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162886,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162886, "pid": 5, "tid": 7, "ts": 1716454223951427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870721, "dur": 8,
    "args": {
      "External id": 162886,
      "cbid": 211, "correlation": 162886
    }
  },
  {
    "ph": "s", "id": 162886, "pid": 76337, "tid": -914061504, "ts": 1716454223870721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454223951455, "dur": 101,
    "args": {
      "External id": 162897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162897,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162897, "pid": 5, "tid": 7, "ts": 1716454223951455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870783, "dur": 11,
    "args": {
      "External id": 162897,
      "cbid": 211, "correlation": 162897
    }
  },
  {
    "ph": "s", "id": 162897, "pid": 76337, "tid": -914061504, "ts": 1716454223870783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870837, "dur": 0,
    "args": {
      "External id": 162907,
      "cbid": 317, "correlation": 162907
    }
  },
  {
    "ph": "f", "id": 162907, "pid": 76337, "tid": -914061504, "ts": 1716454223870837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870838, "dur": 0,
    "args": {
      "External id": 162908,
      "cbid": 203, "correlation": 162908
    }
  },
  {
    "ph": "f", "id": 162908, "pid": 76337, "tid": -914061504, "ts": 1716454223870838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870839, "dur": 0,
    "args": {
      "External id": 162909,
      "cbid": 205, "correlation": 162909
    }
  },
  {
    "ph": "f", "id": 162909, "pid": 76337, "tid": -914061504, "ts": 1716454223870839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223951558, "dur": 75,
    "args": {
      "External id": 162913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162913, "pid": 5, "tid": 7, "ts": 1716454223951558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870853, "dur": 12,
    "args": {
      "External id": 162913,
      "cbid": 211, "correlation": 162913
    }
  },
  {
    "ph": "s", "id": 162913, "pid": 76337, "tid": -914061504, "ts": 1716454223870853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223951635, "dur": 43,
    "args": {
      "External id": 162915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162915, "pid": 5, "tid": 7, "ts": 1716454223951635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870867, "dur": 5,
    "args": {
      "External id": 162915,
      "cbid": 211, "correlation": 162915
    }
  },
  {
    "ph": "s", "id": 162915, "pid": 76337, "tid": -914061504, "ts": 1716454223870867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223951679, "dur": 4,
    "args": {
      "External id": 162917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162917,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 162917, "pid": 5, "tid": 7, "ts": 1716454223951679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870878, "dur": 6,
    "args": {
      "External id": 162917,
      "cbid": 211, "correlation": 162917
    }
  },
  {
    "ph": "s", "id": 162917, "pid": 76337, "tid": -914061504, "ts": 1716454223870878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870887, "dur": 0,
    "args": {
      "External id": 162918,
      "cbid": 51, "correlation": 162918
    }
  },
  {
    "ph": "s", "id": 162918, "pid": 76337, "tid": -914061504, "ts": 1716454223870887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223951684, "dur": 2231,
    "args": {
      "External id": 162919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162919,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162919, "pid": 5, "tid": 7, "ts": 1716454223951684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870888, "dur": 6,
    "args": {
      "External id": 162919,
      "cbid": 211, "correlation": 162919
    }
  },
  {
    "ph": "s", "id": 162919, "pid": 76337, "tid": -914061504, "ts": 1716454223870888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223953917, "dur": 114,
    "args": {
      "External id": 162924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162924,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162924, "pid": 5, "tid": 7, "ts": 1716454223953917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223870916, "dur": 8,
    "args": {
      "External id": 162924,
      "cbid": 211, "correlation": 162924
    }
  },
  {
    "ph": "s", "id": 162924, "pid": 76337, "tid": -914061504, "ts": 1716454223870916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223954032, "dur": 164,
    "args": {
      "External id": 162933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162933,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162933, "pid": 5, "tid": 7, "ts": 1716454223954032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871016, "dur": 13,
    "args": {
      "External id": 162933,
      "cbid": 211, "correlation": 162933
    }
  },
  {
    "ph": "s", "id": 162933, "pid": 76337, "tid": -914061504, "ts": 1716454223871016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223954198, "dur": 128,
    "args": {
      "External id": 162953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162953,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 162953, "pid": 5, "tid": 7, "ts": 1716454223954198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871088, "dur": 11,
    "args": {
      "External id": 162953,
      "cbid": 211, "correlation": 162953
    }
  },
  {
    "ph": "s", "id": 162953, "pid": 76337, "tid": -914061504, "ts": 1716454223871088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223954327, "dur": 5,
    "args": {
      "External id": 162965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162965,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 162965, "pid": 5, "tid": 7, "ts": 1716454223954327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871110, "dur": 6,
    "args": {
      "External id": 162965,
      "cbid": 211, "correlation": 162965
    }
  },
  {
    "ph": "s", "id": 162965, "pid": 76337, "tid": -914061504, "ts": 1716454223871110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223954333, "dur": 160,
    "args": {
      "External id": 162968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162968, "pid": 5, "tid": 7, "ts": 1716454223954333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871128, "dur": 6,
    "args": {
      "External id": 162968,
      "cbid": 211, "correlation": 162968
    }
  },
  {
    "ph": "s", "id": 162968, "pid": 76337, "tid": -914061504, "ts": 1716454223871128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223954494, "dur": 102,
    "args": {
      "External id": 162977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162977,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162977, "pid": 5, "tid": 7, "ts": 1716454223954494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871169, "dur": 10,
    "args": {
      "External id": 162977,
      "cbid": 211, "correlation": 162977
    }
  },
  {
    "ph": "s", "id": 162977, "pid": 76337, "tid": -914061504, "ts": 1716454223871169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871221, "dur": 0,
    "args": {
      "External id": 162987,
      "cbid": 317, "correlation": 162987
    }
  },
  {
    "ph": "f", "id": 162987, "pid": 76337, "tid": -914061504, "ts": 1716454223871221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871222, "dur": 0,
    "args": {
      "External id": 162988,
      "cbid": 203, "correlation": 162988
    }
  },
  {
    "ph": "f", "id": 162988, "pid": 76337, "tid": -914061504, "ts": 1716454223871222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871223, "dur": 0,
    "args": {
      "External id": 162989,
      "cbid": 205, "correlation": 162989
    }
  },
  {
    "ph": "f", "id": 162989, "pid": 76337, "tid": -914061504, "ts": 1716454223871223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223954597, "dur": 112,
    "args": {
      "External id": 162993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162993,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162993, "pid": 5, "tid": 7, "ts": 1716454223954597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871239, "dur": 11,
    "args": {
      "External id": 162993,
      "cbid": 211, "correlation": 162993
    }
  },
  {
    "ph": "s", "id": 162993, "pid": 76337, "tid": -914061504, "ts": 1716454223871239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223954710, "dur": 34,
    "args": {
      "External id": 162995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162995,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 162995, "pid": 5, "tid": 7, "ts": 1716454223954710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871253, "dur": 6,
    "args": {
      "External id": 162995,
      "cbid": 211, "correlation": 162995
    }
  },
  {
    "ph": "s", "id": 162995, "pid": 76337, "tid": -914061504, "ts": 1716454223871253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223954746, "dur": 4,
    "args": {
      "External id": 162997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 162997, "pid": 5, "tid": 7, "ts": 1716454223954746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871264, "dur": 6,
    "args": {
      "External id": 162997,
      "cbid": 211, "correlation": 162997
    }
  },
  {
    "ph": "s", "id": 162997, "pid": 76337, "tid": -914061504, "ts": 1716454223871264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871273, "dur": 0,
    "args": {
      "External id": 162998,
      "cbid": 51, "correlation": 162998
    }
  },
  {
    "ph": "s", "id": 162998, "pid": 76337, "tid": -914061504, "ts": 1716454223871273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223954751, "dur": 2008,
    "args": {
      "External id": 162999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 162999,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 162999, "pid": 5, "tid": 7, "ts": 1716454223954751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871274, "dur": 6,
    "args": {
      "External id": 162999,
      "cbid": 211, "correlation": 162999
    }
  },
  {
    "ph": "s", "id": 162999, "pid": 76337, "tid": -914061504, "ts": 1716454223871274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223956760, "dur": 59,
    "args": {
      "External id": 163004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163004, "pid": 5, "tid": 7, "ts": 1716454223956760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871304, "dur": 9,
    "args": {
      "External id": 163004,
      "cbid": 211, "correlation": 163004
    }
  },
  {
    "ph": "s", "id": 163004, "pid": 76337, "tid": -914061504, "ts": 1716454223871304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223956820, "dur": 3,
    "args": {
      "External id": 163012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163012,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163012, "pid": 5, "tid": 7, "ts": 1716454223956820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871347, "dur": 10,
    "args": {
      "External id": 163012,
      "cbid": 211, "correlation": 163012
    }
  },
  {
    "ph": "s", "id": 163012, "pid": 76337, "tid": -914061504, "ts": 1716454223871347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871412, "dur": 1,
    "args": {
      "External id": 163028,
      "cbid": 251, "correlation": 163028
    }
  },
  {
    "ph": "f", "id": 163028, "pid": 76337, "tid": -914061504, "ts": 1716454223871412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871418, "dur": 0,
    "args": {
      "External id": 163030,
      "cbid": 251, "correlation": 163030
    }
  },
  {
    "ph": "f", "id": 163030, "pid": 76337, "tid": -914061504, "ts": 1716454223871418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223956825, "dur": 11,
    "args": {
      "External id": 163031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163031,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 163031, "pid": 5, "tid": 7, "ts": 1716454223956825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871419, "dur": 11,
    "args": {
      "External id": 163031,
      "cbid": 211, "correlation": 163031
    }
  },
  {
    "ph": "s", "id": 163031, "pid": 76337, "tid": -914061504, "ts": 1716454223871419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223956838, "dur": 5,
    "args": {
      "External id": 163033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163033,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 163033, "pid": 5, "tid": 7, "ts": 1716454223956838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871433, "dur": 6,
    "args": {
      "External id": 163033,
      "cbid": 211, "correlation": 163033
    }
  },
  {
    "ph": "s", "id": 163033, "pid": 76337, "tid": -914061504, "ts": 1716454223871433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223956844, "dur": 55,
    "args": {
      "External id": 163043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163043,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163043, "pid": 5, "tid": 7, "ts": 1716454223956844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871491, "dur": 12,
    "args": {
      "External id": 163043,
      "cbid": 211, "correlation": 163043
    }
  },
  {
    "ph": "s", "id": 163043, "pid": 76337, "tid": -914061504, "ts": 1716454223871491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223956901, "dur": 52,
    "args": {
      "External id": 163063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163063,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 163063, "pid": 5, "tid": 7, "ts": 1716454223956901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871558, "dur": 11,
    "args": {
      "External id": 163063,
      "cbid": 211, "correlation": 163063
    }
  },
  {
    "ph": "s", "id": 163063, "pid": 76337, "tid": -914061504, "ts": 1716454223871558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223956954, "dur": 4,
    "args": {
      "External id": 163075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163075,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163075, "pid": 5, "tid": 7, "ts": 1716454223956954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871579, "dur": 6,
    "args": {
      "External id": 163075,
      "cbid": 211, "correlation": 163075
    }
  },
  {
    "ph": "s", "id": 163075, "pid": 76337, "tid": -914061504, "ts": 1716454223871579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223956959, "dur": 55,
    "args": {
      "External id": 163078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163078,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163078, "pid": 5, "tid": 7, "ts": 1716454223956959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871597, "dur": 7,
    "args": {
      "External id": 163078,
      "cbid": 211, "correlation": 163078
    }
  },
  {
    "ph": "s", "id": 163078, "pid": 76337, "tid": -914061504, "ts": 1716454223871597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223957015, "dur": 37,
    "args": {
      "External id": 163087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163087,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163087, "pid": 5, "tid": 7, "ts": 1716454223957015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871637, "dur": 10,
    "args": {
      "External id": 163087,
      "cbid": 211, "correlation": 163087
    }
  },
  {
    "ph": "s", "id": 163087, "pid": 76337, "tid": -914061504, "ts": 1716454223871637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871701, "dur": 0,
    "args": {
      "External id": 163097,
      "cbid": 317, "correlation": 163097
    }
  },
  {
    "ph": "f", "id": 163097, "pid": 76337, "tid": -914061504, "ts": 1716454223871701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871702, "dur": 0,
    "args": {
      "External id": 163098,
      "cbid": 203, "correlation": 163098
    }
  },
  {
    "ph": "f", "id": 163098, "pid": 76337, "tid": -914061504, "ts": 1716454223871702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871703, "dur": 0,
    "args": {
      "External id": 163099,
      "cbid": 205, "correlation": 163099
    }
  },
  {
    "ph": "f", "id": 163099, "pid": 76337, "tid": -914061504, "ts": 1716454223871703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223957054, "dur": 40,
    "args": {
      "External id": 163103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163103,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163103, "pid": 5, "tid": 7, "ts": 1716454223957054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871721, "dur": 12,
    "args": {
      "External id": 163103,
      "cbid": 211, "correlation": 163103
    }
  },
  {
    "ph": "s", "id": 163103, "pid": 76337, "tid": -914061504, "ts": 1716454223871721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223957095, "dur": 14,
    "args": {
      "External id": 163105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163105,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163105, "pid": 5, "tid": 7, "ts": 1716454223957095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871735, "dur": 5,
    "args": {
      "External id": 163105,
      "cbid": 211, "correlation": 163105
    }
  },
  {
    "ph": "s", "id": 163105, "pid": 76337, "tid": -914061504, "ts": 1716454223871735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223957110, "dur": 3,
    "args": {
      "External id": 163107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163107, "pid": 5, "tid": 7, "ts": 1716454223957110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871745, "dur": 6,
    "args": {
      "External id": 163107,
      "cbid": 211, "correlation": 163107
    }
  },
  {
    "ph": "s", "id": 163107, "pid": 76337, "tid": -914061504, "ts": 1716454223871745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871755, "dur": 0,
    "args": {
      "External id": 163108,
      "cbid": 51, "correlation": 163108
    }
  },
  {
    "ph": "s", "id": 163108, "pid": 76337, "tid": -914061504, "ts": 1716454223871755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223957115, "dur": 699,
    "args": {
      "External id": 163109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163109,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163109, "pid": 5, "tid": 7, "ts": 1716454223957115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871755, "dur": 5,
    "args": {
      "External id": 163109,
      "cbid": 211, "correlation": 163109
    }
  },
  {
    "ph": "s", "id": 163109, "pid": 76337, "tid": -914061504, "ts": 1716454223871755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223957816, "dur": 59,
    "args": {
      "External id": 163114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163114,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163114, "pid": 5, "tid": 7, "ts": 1716454223957816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871783, "dur": 8,
    "args": {
      "External id": 163114,
      "cbid": 211, "correlation": 163114
    }
  },
  {
    "ph": "s", "id": 163114, "pid": 76337, "tid": -914061504, "ts": 1716454223871783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871840, "dur": 0,
    "args": {
      "External id": 163124,
      "cbid": 317, "correlation": 163124
    }
  },
  {
    "ph": "f", "id": 163124, "pid": 76337, "tid": -914061504, "ts": 1716454223871840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871841, "dur": 0,
    "args": {
      "External id": 163125,
      "cbid": 203, "correlation": 163125
    }
  },
  {
    "ph": "f", "id": 163125, "pid": 76337, "tid": -914061504, "ts": 1716454223871841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871841, "dur": 0,
    "args": {
      "External id": 163126,
      "cbid": 205, "correlation": 163126
    }
  },
  {
    "ph": "f", "id": 163126, "pid": 76337, "tid": -914061504, "ts": 1716454223871841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223957876, "dur": 4,
    "args": {
      "External id": 163130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163130, "pid": 5, "tid": 7, "ts": 1716454223957876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871858, "dur": 12,
    "args": {
      "External id": 163130,
      "cbid": 211, "correlation": 163130
    }
  },
  {
    "ph": "s", "id": 163130, "pid": 76337, "tid": -914061504, "ts": 1716454223871858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871873, "dur": 0,
    "args": {
      "External id": 163131,
      "cbid": 51, "correlation": 163131
    }
  },
  {
    "ph": "s", "id": 163131, "pid": 76337, "tid": -914061504, "ts": 1716454223871873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223957881, "dur": 266,
    "args": {
      "External id": 163132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163132,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163132, "pid": 5, "tid": 7, "ts": 1716454223957881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871874, "dur": 7,
    "args": {
      "External id": 163132,
      "cbid": 211, "correlation": 163132
    }
  },
  {
    "ph": "s", "id": 163132, "pid": 76337, "tid": -914061504, "ts": 1716454223871874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223958149, "dur": 59,
    "args": {
      "External id": 163137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163137,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163137, "pid": 5, "tid": 7, "ts": 1716454223958149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871903, "dur": 8,
    "args": {
      "External id": 163137,
      "cbid": 211, "correlation": 163137
    }
  },
  {
    "ph": "s", "id": 163137, "pid": 76337, "tid": -914061504, "ts": 1716454223871903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223958209, "dur": 50,
    "args": {
      "External id": 163145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163145,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163145, "pid": 5, "tid": 7, "ts": 1716454223958209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871931, "dur": 8,
    "args": {
      "External id": 163145,
      "cbid": 211, "correlation": 163145
    }
  },
  {
    "ph": "s", "id": 163145, "pid": 76337, "tid": -914061504, "ts": 1716454223871931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223958260, "dur": 35,
    "args": {
      "External id": 163153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163153,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163153, "pid": 5, "tid": 7, "ts": 1716454223958260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223871960, "dur": 8,
    "args": {
      "External id": 163153,
      "cbid": 211, "correlation": 163153
    }
  },
  {
    "ph": "s", "id": 163153, "pid": 76337, "tid": -914061504, "ts": 1716454223871960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223958297, "dur": 53,
    "args": {
      "External id": 163173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163173,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 163173, "pid": 5, "tid": 7, "ts": 1716454223958297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872050, "dur": 13,
    "args": {
      "External id": 163173,
      "cbid": 211, "correlation": 163173
    }
  },
  {
    "ph": "s", "id": 163173, "pid": 76337, "tid": -914061504, "ts": 1716454223872050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223958351, "dur": 4,
    "args": {
      "External id": 163185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163185,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163185, "pid": 5, "tid": 7, "ts": 1716454223958351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872073, "dur": 7,
    "args": {
      "External id": 163185,
      "cbid": 211, "correlation": 163185
    }
  },
  {
    "ph": "s", "id": 163185, "pid": 76337, "tid": -914061504, "ts": 1716454223872073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223958356, "dur": 55,
    "args": {
      "External id": 163188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163188,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163188, "pid": 5, "tid": 7, "ts": 1716454223958356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872091, "dur": 7,
    "args": {
      "External id": 163188,
      "cbid": 211, "correlation": 163188
    }
  },
  {
    "ph": "s", "id": 163188, "pid": 76337, "tid": -914061504, "ts": 1716454223872091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872149, "dur": 0,
    "args": {
      "External id": 163199,
      "cbid": 317, "correlation": 163199
    }
  },
  {
    "ph": "f", "id": 163199, "pid": 76337, "tid": -914061504, "ts": 1716454223872149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872150, "dur": 0,
    "args": {
      "External id": 163200,
      "cbid": 203, "correlation": 163200
    }
  },
  {
    "ph": "f", "id": 163200, "pid": 76337, "tid": -914061504, "ts": 1716454223872150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872151, "dur": 0,
    "args": {
      "External id": 163201,
      "cbid": 205, "correlation": 163201
    }
  },
  {
    "ph": "f", "id": 163201, "pid": 76337, "tid": -914061504, "ts": 1716454223872151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872183, "dur": 2,
    "args": {
      "External id": 163205,
      "cbid": 251, "correlation": 163205
    }
  },
  {
    "ph": "f", "id": 163205, "pid": 76337, "tid": -914061504, "ts": 1716454223872183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872187, "dur": 1,
    "args": {
      "External id": 163206,
      "cbid": 251, "correlation": 163206
    }
  },
  {
    "ph": "f", "id": 163206, "pid": 76337, "tid": -914061504, "ts": 1716454223872187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872189, "dur": 1,
    "args": {
      "External id": 163207,
      "cbid": 251, "correlation": 163207
    }
  },
  {
    "ph": "f", "id": 163207, "pid": 76337, "tid": -914061504, "ts": 1716454223872189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872191, "dur": 1,
    "args": {
      "External id": 163208,
      "cbid": 251, "correlation": 163208
    }
  },
  {
    "ph": "f", "id": 163208, "pid": 76337, "tid": -914061504, "ts": 1716454223872191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872193, "dur": 1,
    "args": {
      "External id": 163209,
      "cbid": 251, "correlation": 163209
    }
  },
  {
    "ph": "f", "id": 163209, "pid": 76337, "tid": -914061504, "ts": 1716454223872193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872195, "dur": 1,
    "args": {
      "External id": 163210,
      "cbid": 251, "correlation": 163210
    }
  },
  {
    "ph": "f", "id": 163210, "pid": 76337, "tid": -914061504, "ts": 1716454223872195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872197, "dur": 1,
    "args": {
      "External id": 163211,
      "cbid": 251, "correlation": 163211
    }
  },
  {
    "ph": "f", "id": 163211, "pid": 76337, "tid": -914061504, "ts": 1716454223872197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872199, "dur": 1,
    "args": {
      "External id": 163212,
      "cbid": 251, "correlation": 163212
    }
  },
  {
    "ph": "f", "id": 163212, "pid": 76337, "tid": -914061504, "ts": 1716454223872199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872202, "dur": 0,
    "args": {
      "External id": 163213,
      "cbid": 251, "correlation": 163213
    }
  },
  {
    "ph": "f", "id": 163213, "pid": 76337, "tid": -914061504, "ts": 1716454223872202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223958412, "dur": 116,
    "args": {
      "External id": 163214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163214,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163214, "pid": 5, "tid": 7, "ts": 1716454223958412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872204, "dur": 13,
    "args": {
      "External id": 163214,
      "cbid": 211, "correlation": 163214
    }
  },
  {
    "ph": "s", "id": 163214, "pid": 76337, "tid": -914061504, "ts": 1716454223872204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223958529, "dur": 60,
    "args": {
      "External id": 163220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163220,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163220, "pid": 5, "tid": 7, "ts": 1716454223958529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872241, "dur": 10,
    "args": {
      "External id": 163220,
      "cbid": 211, "correlation": 163220
    }
  },
  {
    "ph": "s", "id": 163220, "pid": 76337, "tid": -914061504, "ts": 1716454223872241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223958591, "dur": 556,
    "args": {
      "External id": 163229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163229,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163229, "pid": 5, "tid": 7, "ts": 1716454223958591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872324, "dur": 14,
    "args": {
      "External id": 163229,
      "cbid": 211, "correlation": 163229
    }
  },
  {
    "ph": "s", "id": 163229, "pid": 76337, "tid": -914061504, "ts": 1716454223872324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223959149, "dur": 183,
    "args": {
      "External id": 163251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163251,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163251, "pid": 5, "tid": 7, "ts": 1716454223959149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872382, "dur": 10,
    "args": {
      "External id": 163251,
      "cbid": 211, "correlation": 163251
    }
  },
  {
    "ph": "s", "id": 163251, "pid": 76337, "tid": -914061504, "ts": 1716454223872382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872471, "dur": 1,
    "args": {
      "External id": 163262,
      "cbid": 251, "correlation": 163262
    }
  },
  {
    "ph": "f", "id": 163262, "pid": 76337, "tid": -914061504, "ts": 1716454223872471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223959334, "dur": 197,
    "args": {
      "External id": 163263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163263,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163263, "pid": 5, "tid": 7, "ts": 1716454223959334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872476, "dur": 13,
    "args": {
      "External id": 163263,
      "cbid": 211, "correlation": 163263
    }
  },
  {
    "ph": "s", "id": 163263, "pid": 76337, "tid": -914061504, "ts": 1716454223872476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872546, "dur": 1,
    "args": {
      "External id": 163274,
      "cbid": 251, "correlation": 163274
    }
  },
  {
    "ph": "f", "id": 163274, "pid": 76337, "tid": -914061504, "ts": 1716454223872546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223959532, "dur": 189,
    "args": {
      "External id": 163275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163275,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163275, "pid": 5, "tid": 7, "ts": 1716454223959532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872551, "dur": 12,
    "args": {
      "External id": 163275,
      "cbid": 211, "correlation": 163275
    }
  },
  {
    "ph": "s", "id": 163275, "pid": 76337, "tid": -914061504, "ts": 1716454223872551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872614, "dur": 1,
    "args": {
      "External id": 163286,
      "cbid": 251, "correlation": 163286
    }
  },
  {
    "ph": "f", "id": 163286, "pid": 76337, "tid": -914061504, "ts": 1716454223872614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223959723, "dur": 187,
    "args": {
      "External id": 163287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163287,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163287, "pid": 5, "tid": 7, "ts": 1716454223959723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872618, "dur": 12,
    "args": {
      "External id": 163287,
      "cbid": 211, "correlation": 163287
    }
  },
  {
    "ph": "s", "id": 163287, "pid": 76337, "tid": -914061504, "ts": 1716454223872618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223959911, "dur": 18741,
    "args": {
      "External id": 163308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163308,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163308, "pid": 5, "tid": 7, "ts": 1716454223959911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872703, "dur": 13,
    "args": {
      "External id": 163308,
      "cbid": 211, "correlation": 163308
    }
  },
  {
    "ph": "s", "id": 163308, "pid": 76337, "tid": -914061504, "ts": 1716454223872703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872803, "dur": 1,
    "args": {
      "External id": 163326,
      "cbid": 251, "correlation": 163326
    }
  },
  {
    "ph": "f", "id": 163326, "pid": 76337, "tid": -914061504, "ts": 1716454223872803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223978653, "dur": 203,
    "args": {
      "External id": 163328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163328,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163328, "pid": 5, "tid": 7, "ts": 1716454223978653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872808, "dur": 13,
    "args": {
      "External id": 163328,
      "cbid": 211, "correlation": 163328
    }
  },
  {
    "ph": "s", "id": 163328, "pid": 76337, "tid": -914061504, "ts": 1716454223872808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223978858, "dur": 67,
    "args": {
      "External id": 163336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163336,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163336, "pid": 5, "tid": 7, "ts": 1716454223978858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872878, "dur": 12,
    "args": {
      "External id": 163336,
      "cbid": 211, "correlation": 163336
    }
  },
  {
    "ph": "s", "id": 163336, "pid": 76337, "tid": -914061504, "ts": 1716454223872878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223978926, "dur": 97,
    "args": {
      "External id": 163344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163344,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163344, "pid": 5, "tid": 7, "ts": 1716454223978926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223872917, "dur": 9,
    "args": {
      "External id": 163344,
      "cbid": 211, "correlation": 163344
    }
  },
  {
    "ph": "s", "id": 163344, "pid": 76337, "tid": -914061504, "ts": 1716454223872917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223979025, "dur": 55,
    "args": {
      "External id": 163355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163355, "pid": 5, "tid": 7, "ts": 1716454223979025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873002, "dur": 13,
    "args": {
      "External id": 163355,
      "cbid": 211, "correlation": 163355
    }
  },
  {
    "ph": "s", "id": 163355, "pid": 76337, "tid": -914061504, "ts": 1716454223873002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223979081, "dur": 93,
    "args": {
      "External id": 163377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163377,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163377, "pid": 5, "tid": 7, "ts": 1716454223979081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873034, "dur": 7,
    "args": {
      "External id": 163377,
      "cbid": 211, "correlation": 163377
    }
  },
  {
    "ph": "s", "id": 163377, "pid": 76337, "tid": -914061504, "ts": 1716454223873034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873119, "dur": 1,
    "args": {
      "External id": 163388,
      "cbid": 251, "correlation": 163388
    }
  },
  {
    "ph": "f", "id": 163388, "pid": 76337, "tid": -914061504, "ts": 1716454223873119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223979176, "dur": 105,
    "args": {
      "External id": 163389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163389,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163389, "pid": 5, "tid": 7, "ts": 1716454223979176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873124, "dur": 13,
    "args": {
      "External id": 163389,
      "cbid": 211, "correlation": 163389
    }
  },
  {
    "ph": "s", "id": 163389, "pid": 76337, "tid": -914061504, "ts": 1716454223873124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873199, "dur": 2,
    "args": {
      "External id": 163400,
      "cbid": 251, "correlation": 163400
    }
  },
  {
    "ph": "f", "id": 163400, "pid": 76337, "tid": -914061504, "ts": 1716454223873199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873204, "dur": 0,
    "args": {
      "External id": 163401,
      "cbid": 251, "correlation": 163401
    }
  },
  {
    "ph": "f", "id": 163401, "pid": 76337, "tid": -914061504, "ts": 1716454223873204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223979282, "dur": 10,
    "args": {
      "External id": 163402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163402,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 163402, "pid": 5, "tid": 7, "ts": 1716454223979282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873206, "dur": 13,
    "args": {
      "External id": 163402,
      "cbid": 211, "correlation": 163402
    }
  },
  {
    "ph": "s", "id": 163402, "pid": 76337, "tid": -914061504, "ts": 1716454223873206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223979294, "dur": 5,
    "args": {
      "External id": 163404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163404,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 163404, "pid": 5, "tid": 7, "ts": 1716454223979294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873223, "dur": 7,
    "args": {
      "External id": 163404,
      "cbid": 211, "correlation": 163404
    }
  },
  {
    "ph": "s", "id": 163404, "pid": 76337, "tid": -914061504, "ts": 1716454223873223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873285, "dur": 1,
    "args": {
      "External id": 163415,
      "cbid": 251, "correlation": 163415
    }
  },
  {
    "ph": "f", "id": 163415, "pid": 76337, "tid": -914061504, "ts": 1716454223873285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873289, "dur": 0,
    "args": {
      "External id": 163416,
      "cbid": 251, "correlation": 163416
    }
  },
  {
    "ph": "f", "id": 163416, "pid": 76337, "tid": -914061504, "ts": 1716454223873289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223979300, "dur": 6,
    "args": {
      "External id": 163417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163417,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 163417, "pid": 5, "tid": 7, "ts": 1716454223979300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873290, "dur": 13,
    "args": {
      "External id": 163417,
      "cbid": 211, "correlation": 163417
    }
  },
  {
    "ph": "s", "id": 163417, "pid": 76337, "tid": -914061504, "ts": 1716454223873290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223979308, "dur": 3,
    "args": {
      "External id": 163419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163419,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 163419, "pid": 5, "tid": 7, "ts": 1716454223979308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873304, "dur": 5,
    "args": {
      "External id": 163419,
      "cbid": 211, "correlation": 163419
    }
  },
  {
    "ph": "s", "id": 163419, "pid": 76337, "tid": -914061504, "ts": 1716454223873304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223979313, "dur": 158,
    "args": {
      "External id": 163440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163440,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163440, "pid": 5, "tid": 7, "ts": 1716454223979313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873379, "dur": 12,
    "args": {
      "External id": 163440,
      "cbid": 211, "correlation": 163440
    }
  },
  {
    "ph": "s", "id": 163440, "pid": 76337, "tid": -914061504, "ts": 1716454223873379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873475, "dur": 1,
    "args": {
      "External id": 163458,
      "cbid": 251, "correlation": 163458
    }
  },
  {
    "ph": "f", "id": 163458, "pid": 76337, "tid": -914061504, "ts": 1716454223873475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223979472, "dur": 106,
    "args": {
      "External id": 163460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163460,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163460, "pid": 5, "tid": 7, "ts": 1716454223979472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873481, "dur": 14,
    "args": {
      "External id": 163460,
      "cbid": 211, "correlation": 163460
    }
  },
  {
    "ph": "s", "id": 163460, "pid": 76337, "tid": -914061504, "ts": 1716454223873481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223979579, "dur": 35,
    "args": {
      "External id": 163468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163468,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163468, "pid": 5, "tid": 7, "ts": 1716454223979579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873552, "dur": 12,
    "args": {
      "External id": 163468,
      "cbid": 211, "correlation": 163468
    }
  },
  {
    "ph": "s", "id": 163468, "pid": 76337, "tid": -914061504, "ts": 1716454223873552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223979616, "dur": 67,
    "args": {
      "External id": 163476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163476,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163476, "pid": 5, "tid": 7, "ts": 1716454223979616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873593, "dur": 10,
    "args": {
      "External id": 163476,
      "cbid": 211, "correlation": 163476
    }
  },
  {
    "ph": "s", "id": 163476, "pid": 76337, "tid": -914061504, "ts": 1716454223873593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223979684, "dur": 93,
    "args": {
      "External id": 163498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163498,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163498, "pid": 5, "tid": 7, "ts": 1716454223979684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873644, "dur": 10,
    "args": {
      "External id": 163498,
      "cbid": 211, "correlation": 163498
    }
  },
  {
    "ph": "s", "id": 163498, "pid": 76337, "tid": -914061504, "ts": 1716454223873644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873729, "dur": 1,
    "args": {
      "External id": 163514,
      "cbid": 251, "correlation": 163514
    }
  },
  {
    "ph": "f", "id": 163514, "pid": 76337, "tid": -914061504, "ts": 1716454223873729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223979779, "dur": 582,
    "args": {
      "External id": 163516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163516,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163516, "pid": 5, "tid": 7, "ts": 1716454223979779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873734, "dur": 13,
    "args": {
      "External id": 163516,
      "cbid": 211, "correlation": 163516
    }
  },
  {
    "ph": "s", "id": 163516, "pid": 76337, "tid": -914061504, "ts": 1716454223873734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223980363, "dur": 247,
    "args": {
      "External id": 163524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163524, "pid": 5, "tid": 7, "ts": 1716454223980363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873800, "dur": 12,
    "args": {
      "External id": 163524,
      "cbid": 211, "correlation": 163524
    }
  },
  {
    "ph": "s", "id": 163524, "pid": 76337, "tid": -914061504, "ts": 1716454223873800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223980611, "dur": 251,
    "args": {
      "External id": 163532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163532, "pid": 5, "tid": 7, "ts": 1716454223980611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873831, "dur": 9,
    "args": {
      "External id": 163532,
      "cbid": 211, "correlation": 163532
    }
  },
  {
    "ph": "s", "id": 163532, "pid": 76337, "tid": -914061504, "ts": 1716454223873831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873914, "dur": 1,
    "args": {
      "External id": 163548,
      "cbid": 251, "correlation": 163548
    }
  },
  {
    "ph": "f", "id": 163548, "pid": 76337, "tid": -914061504, "ts": 1716454223873914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873918, "dur": 0,
    "args": {
      "External id": 163550,
      "cbid": 251, "correlation": 163550
    }
  },
  {
    "ph": "f", "id": 163550, "pid": 76337, "tid": -914061504, "ts": 1716454223873918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454223980863, "dur": 357,
    "args": {
      "External id": 163551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163551,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 163551, "pid": 5, "tid": 7, "ts": 1716454223980863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873921, "dur": 13,
    "args": {
      "External id": 163551,
      "cbid": 211, "correlation": 163551
    }
  },
  {
    "ph": "s", "id": 163551, "pid": 76337, "tid": -914061504, "ts": 1716454223873921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223981222, "dur": 50,
    "args": {
      "External id": 163559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163559,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163559, "pid": 5, "tid": 7, "ts": 1716454223981222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223873964, "dur": 17,
    "args": {
      "External id": 163559,
      "cbid": 211, "correlation": 163559
    }
  },
  {
    "ph": "s", "id": 163559, "pid": 76337, "tid": -914061504, "ts": 1716454223873964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223981273, "dur": 160,
    "args": {
      "External id": 163570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163570, "pid": 5, "tid": 7, "ts": 1716454223981273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874041, "dur": 14,
    "args": {
      "External id": 163570,
      "cbid": 211, "correlation": 163570
    }
  },
  {
    "ph": "s", "id": 163570, "pid": 76337, "tid": -914061504, "ts": 1716454223874041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874107, "dur": 0,
    "args": {
      "External id": 163582,
      "cbid": 317, "correlation": 163582
    }
  },
  {
    "ph": "f", "id": 163582, "pid": 76337, "tid": -914061504, "ts": 1716454223874107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874108, "dur": 0,
    "args": {
      "External id": 163583,
      "cbid": 203, "correlation": 163583
    }
  },
  {
    "ph": "f", "id": 163583, "pid": 76337, "tid": -914061504, "ts": 1716454223874108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874108, "dur": 0,
    "args": {
      "External id": 163584,
      "cbid": 205, "correlation": 163584
    }
  },
  {
    "ph": "f", "id": 163584, "pid": 76337, "tid": -914061504, "ts": 1716454223874108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874134, "dur": 1,
    "args": {
      "External id": 163588,
      "cbid": 251, "correlation": 163588
    }
  },
  {
    "ph": "f", "id": 163588, "pid": 76337, "tid": -914061504, "ts": 1716454223874134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874136, "dur": 0,
    "args": {
      "External id": 163589,
      "cbid": 251, "correlation": 163589
    }
  },
  {
    "ph": "f", "id": 163589, "pid": 76337, "tid": -914061504, "ts": 1716454223874136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874137, "dur": 0,
    "args": {
      "External id": 163590,
      "cbid": 251, "correlation": 163590
    }
  },
  {
    "ph": "f", "id": 163590, "pid": 76337, "tid": -914061504, "ts": 1716454223874137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874138, "dur": 0,
    "args": {
      "External id": 163591,
      "cbid": 251, "correlation": 163591
    }
  },
  {
    "ph": "f", "id": 163591, "pid": 76337, "tid": -914061504, "ts": 1716454223874138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874139, "dur": 0,
    "args": {
      "External id": 163592,
      "cbid": 251, "correlation": 163592
    }
  },
  {
    "ph": "f", "id": 163592, "pid": 76337, "tid": -914061504, "ts": 1716454223874139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874140, "dur": 0,
    "args": {
      "External id": 163593,
      "cbid": 251, "correlation": 163593
    }
  },
  {
    "ph": "f", "id": 163593, "pid": 76337, "tid": -914061504, "ts": 1716454223874140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874141, "dur": 0,
    "args": {
      "External id": 163594,
      "cbid": 251, "correlation": 163594
    }
  },
  {
    "ph": "f", "id": 163594, "pid": 76337, "tid": -914061504, "ts": 1716454223874141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874142, "dur": 0,
    "args": {
      "External id": 163595,
      "cbid": 251, "correlation": 163595
    }
  },
  {
    "ph": "f", "id": 163595, "pid": 76337, "tid": -914061504, "ts": 1716454223874142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874144, "dur": 0,
    "args": {
      "External id": 163596,
      "cbid": 251, "correlation": 163596
    }
  },
  {
    "ph": "f", "id": 163596, "pid": 76337, "tid": -914061504, "ts": 1716454223874144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223981435, "dur": 116,
    "args": {
      "External id": 163597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163597,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163597, "pid": 5, "tid": 7, "ts": 1716454223981435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874146, "dur": 13,
    "args": {
      "External id": 163597,
      "cbid": 211, "correlation": 163597
    }
  },
  {
    "ph": "s", "id": 163597, "pid": 76337, "tid": -914061504, "ts": 1716454223874146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223981552, "dur": 60,
    "args": {
      "External id": 163603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163603, "pid": 5, "tid": 7, "ts": 1716454223981552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874181, "dur": 9,
    "args": {
      "External id": 163603,
      "cbid": 211, "correlation": 163603
    }
  },
  {
    "ph": "s", "id": 163603, "pid": 76337, "tid": -914061504, "ts": 1716454223874181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223981614, "dur": 50,
    "args": {
      "External id": 163611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163611,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163611, "pid": 5, "tid": 7, "ts": 1716454223981614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874212, "dur": 8,
    "args": {
      "External id": 163611,
      "cbid": 211, "correlation": 163611
    }
  },
  {
    "ph": "s", "id": 163611, "pid": 76337, "tid": -914061504, "ts": 1716454223874212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223981665, "dur": 98,
    "args": {
      "External id": 163620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163620,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163620, "pid": 5, "tid": 7, "ts": 1716454223981665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874251, "dur": 10,
    "args": {
      "External id": 163620,
      "cbid": 211, "correlation": 163620
    }
  },
  {
    "ph": "s", "id": 163620, "pid": 76337, "tid": -914061504, "ts": 1716454223874251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223981764, "dur": 92,
    "args": {
      "External id": 163640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163640,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 163640, "pid": 5, "tid": 7, "ts": 1716454223981764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874323, "dur": 11,
    "args": {
      "External id": 163640,
      "cbid": 211, "correlation": 163640
    }
  },
  {
    "ph": "s", "id": 163640, "pid": 76337, "tid": -914061504, "ts": 1716454223874323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223981858, "dur": 5,
    "args": {
      "External id": 163652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163652,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 163652, "pid": 5, "tid": 7, "ts": 1716454223981858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874344, "dur": 7,
    "args": {
      "External id": 163652,
      "cbid": 211, "correlation": 163652
    }
  },
  {
    "ph": "s", "id": 163652, "pid": 76337, "tid": -914061504, "ts": 1716454223874344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223981864, "dur": 109,
    "args": {
      "External id": 163655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163655,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163655, "pid": 5, "tid": 7, "ts": 1716454223981864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874362, "dur": 7,
    "args": {
      "External id": 163655,
      "cbid": 211, "correlation": 163655
    }
  },
  {
    "ph": "s", "id": 163655, "pid": 76337, "tid": -914061504, "ts": 1716454223874362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223981974, "dur": 69,
    "args": {
      "External id": 163664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163664,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163664, "pid": 5, "tid": 7, "ts": 1716454223981974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874402, "dur": 9,
    "args": {
      "External id": 163664,
      "cbid": 211, "correlation": 163664
    }
  },
  {
    "ph": "s", "id": 163664, "pid": 76337, "tid": -914061504, "ts": 1716454223874402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874454, "dur": 0,
    "args": {
      "External id": 163674,
      "cbid": 317, "correlation": 163674
    }
  },
  {
    "ph": "f", "id": 163674, "pid": 76337, "tid": -914061504, "ts": 1716454223874454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874454, "dur": 0,
    "args": {
      "External id": 163675,
      "cbid": 203, "correlation": 163675
    }
  },
  {
    "ph": "f", "id": 163675, "pid": 76337, "tid": -914061504, "ts": 1716454223874454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874455, "dur": 0,
    "args": {
      "External id": 163676,
      "cbid": 205, "correlation": 163676
    }
  },
  {
    "ph": "f", "id": 163676, "pid": 76337, "tid": -914061504, "ts": 1716454223874455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223982045, "dur": 76,
    "args": {
      "External id": 163680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163680,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163680, "pid": 5, "tid": 7, "ts": 1716454223982045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874470, "dur": 11,
    "args": {
      "External id": 163680,
      "cbid": 211, "correlation": 163680
    }
  },
  {
    "ph": "s", "id": 163680, "pid": 76337, "tid": -914061504, "ts": 1716454223874470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223982122, "dur": 24,
    "args": {
      "External id": 163682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163682,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163682, "pid": 5, "tid": 7, "ts": 1716454223982122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874484, "dur": 5,
    "args": {
      "External id": 163682,
      "cbid": 211, "correlation": 163682
    }
  },
  {
    "ph": "s", "id": 163682, "pid": 76337, "tid": -914061504, "ts": 1716454223874484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223982148, "dur": 4,
    "args": {
      "External id": 163684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163684, "pid": 5, "tid": 7, "ts": 1716454223982148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874494, "dur": 6,
    "args": {
      "External id": 163684,
      "cbid": 211, "correlation": 163684
    }
  },
  {
    "ph": "s", "id": 163684, "pid": 76337, "tid": -914061504, "ts": 1716454223874494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874503, "dur": 0,
    "args": {
      "External id": 163685,
      "cbid": 51, "correlation": 163685
    }
  },
  {
    "ph": "s", "id": 163685, "pid": 76337, "tid": -914061504, "ts": 1716454223874503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223982153, "dur": 1379,
    "args": {
      "External id": 163686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163686,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163686, "pid": 5, "tid": 7, "ts": 1716454223982153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874504, "dur": 5,
    "args": {
      "External id": 163686,
      "cbid": 211, "correlation": 163686
    }
  },
  {
    "ph": "s", "id": 163686, "pid": 76337, "tid": -914061504, "ts": 1716454223874504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223983534, "dur": 59,
    "args": {
      "External id": 163691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163691, "pid": 5, "tid": 7, "ts": 1716454223983534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874531, "dur": 9,
    "args": {
      "External id": 163691,
      "cbid": 211, "correlation": 163691
    }
  },
  {
    "ph": "s", "id": 163691, "pid": 76337, "tid": -914061504, "ts": 1716454223874531,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223983595, "dur": 3,
    "args": {
      "External id": 163699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163699,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163699, "pid": 5, "tid": 7, "ts": 1716454223983595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874575, "dur": 9,
    "args": {
      "External id": 163699,
      "cbid": 211, "correlation": 163699
    }
  },
  {
    "ph": "s", "id": 163699, "pid": 76337, "tid": -914061504, "ts": 1716454223874575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874640, "dur": 2,
    "args": {
      "External id": 163715,
      "cbid": 251, "correlation": 163715
    }
  },
  {
    "ph": "f", "id": 163715, "pid": 76337, "tid": -914061504, "ts": 1716454223874640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874646, "dur": 0,
    "args": {
      "External id": 163717,
      "cbid": 251, "correlation": 163717
    }
  },
  {
    "ph": "f", "id": 163717, "pid": 76337, "tid": -914061504, "ts": 1716454223874646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223983599, "dur": 11,
    "args": {
      "External id": 163718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163718,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 163718, "pid": 5, "tid": 7, "ts": 1716454223983599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874648, "dur": 11,
    "args": {
      "External id": 163718,
      "cbid": 211, "correlation": 163718
    }
  },
  {
    "ph": "s", "id": 163718, "pid": 76337, "tid": -914061504, "ts": 1716454223874648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454223983612, "dur": 5,
    "args": {
      "External id": 163720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163720,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 163720, "pid": 5, "tid": 7, "ts": 1716454223983612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874661, "dur": 5,
    "args": {
      "External id": 163720,
      "cbid": 211, "correlation": 163720
    }
  },
  {
    "ph": "s", "id": 163720, "pid": 76337, "tid": -914061504, "ts": 1716454223874661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223983618, "dur": 55,
    "args": {
      "External id": 163730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163730,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163730, "pid": 5, "tid": 7, "ts": 1716454223983618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874719, "dur": 12,
    "args": {
      "External id": 163730,
      "cbid": 211, "correlation": 163730
    }
  },
  {
    "ph": "s", "id": 163730, "pid": 76337, "tid": -914061504, "ts": 1716454223874719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223983675, "dur": 53,
    "args": {
      "External id": 163750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163750,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 163750, "pid": 5, "tid": 7, "ts": 1716454223983675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874784, "dur": 12,
    "args": {
      "External id": 163750,
      "cbid": 211, "correlation": 163750
    }
  },
  {
    "ph": "s", "id": 163750, "pid": 76337, "tid": -914061504, "ts": 1716454223874784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223983729, "dur": 4,
    "args": {
      "External id": 163762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163762,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163762, "pid": 5, "tid": 7, "ts": 1716454223983729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874805, "dur": 6,
    "args": {
      "External id": 163762,
      "cbid": 211, "correlation": 163762
    }
  },
  {
    "ph": "s", "id": 163762, "pid": 76337, "tid": -914061504, "ts": 1716454223874805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223983734, "dur": 56,
    "args": {
      "External id": 163765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163765, "pid": 5, "tid": 7, "ts": 1716454223983734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874823, "dur": 7,
    "args": {
      "External id": 163765,
      "cbid": 211, "correlation": 163765
    }
  },
  {
    "ph": "s", "id": 163765, "pid": 76337, "tid": -914061504, "ts": 1716454223874823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223983792, "dur": 36,
    "args": {
      "External id": 163774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163774,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163774, "pid": 5, "tid": 7, "ts": 1716454223983792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874864, "dur": 10,
    "args": {
      "External id": 163774,
      "cbid": 211, "correlation": 163774
    }
  },
  {
    "ph": "s", "id": 163774, "pid": 76337, "tid": -914061504, "ts": 1716454223874864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874927, "dur": 0,
    "args": {
      "External id": 163784,
      "cbid": 317, "correlation": 163784
    }
  },
  {
    "ph": "f", "id": 163784, "pid": 76337, "tid": -914061504, "ts": 1716454223874927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874928, "dur": 0,
    "args": {
      "External id": 163785,
      "cbid": 203, "correlation": 163785
    }
  },
  {
    "ph": "f", "id": 163785, "pid": 76337, "tid": -914061504, "ts": 1716454223874928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874929, "dur": 0,
    "args": {
      "External id": 163786,
      "cbid": 205, "correlation": 163786
    }
  },
  {
    "ph": "f", "id": 163786, "pid": 76337, "tid": -914061504, "ts": 1716454223874929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223983829, "dur": 40,
    "args": {
      "External id": 163790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163790,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163790, "pid": 5, "tid": 7, "ts": 1716454223983829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874944, "dur": 13,
    "args": {
      "External id": 163790,
      "cbid": 211, "correlation": 163790
    }
  },
  {
    "ph": "s", "id": 163790, "pid": 76337, "tid": -914061504, "ts": 1716454223874944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223983871, "dur": 14,
    "args": {
      "External id": 163792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163792,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163792, "pid": 5, "tid": 7, "ts": 1716454223983871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874959, "dur": 5,
    "args": {
      "External id": 163792,
      "cbid": 211, "correlation": 163792
    }
  },
  {
    "ph": "s", "id": 163792, "pid": 76337, "tid": -914061504, "ts": 1716454223874959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454223983886, "dur": 3,
    "args": {
      "External id": 163794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163794,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163794, "pid": 5, "tid": 7, "ts": 1716454223983886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874968, "dur": 16,
    "args": {
      "External id": 163794,
      "cbid": 211, "correlation": 163794
    }
  },
  {
    "ph": "s", "id": 163794, "pid": 76337, "tid": -914061504, "ts": 1716454223874968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874987, "dur": 0,
    "args": {
      "External id": 163795,
      "cbid": 51, "correlation": 163795
    }
  },
  {
    "ph": "s", "id": 163795, "pid": 76337, "tid": -914061504, "ts": 1716454223874987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454223983891, "dur": 706,
    "args": {
      "External id": 163796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163796,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163796, "pid": 5, "tid": 7, "ts": 1716454223983891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223874988, "dur": 5,
    "args": {
      "External id": 163796,
      "cbid": 211, "correlation": 163796
    }
  },
  {
    "ph": "s", "id": 163796, "pid": 76337, "tid": -914061504, "ts": 1716454223874988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223984598, "dur": 60,
    "args": {
      "External id": 163801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163801, "pid": 5, "tid": 7, "ts": 1716454223984598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875017, "dur": 9,
    "args": {
      "External id": 163801,
      "cbid": 211, "correlation": 163801
    }
  },
  {
    "ph": "s", "id": 163801, "pid": 76337, "tid": -914061504, "ts": 1716454223875017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875075, "dur": 0,
    "args": {
      "External id": 163811,
      "cbid": 317, "correlation": 163811
    }
  },
  {
    "ph": "f", "id": 163811, "pid": 76337, "tid": -914061504, "ts": 1716454223875075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875075, "dur": 0,
    "args": {
      "External id": 163812,
      "cbid": 203, "correlation": 163812
    }
  },
  {
    "ph": "f", "id": 163812, "pid": 76337, "tid": -914061504, "ts": 1716454223875075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875076, "dur": 0,
    "args": {
      "External id": 163813,
      "cbid": 205, "correlation": 163813
    }
  },
  {
    "ph": "f", "id": 163813, "pid": 76337, "tid": -914061504, "ts": 1716454223875076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223984659, "dur": 74,
    "args": {
      "External id": 163817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163817,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163817, "pid": 5, "tid": 7, "ts": 1716454223984659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875089, "dur": 12,
    "args": {
      "External id": 163817,
      "cbid": 211, "correlation": 163817
    }
  },
  {
    "ph": "s", "id": 163817, "pid": 76337, "tid": -914061504, "ts": 1716454223875089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454223984734, "dur": 208,
    "args": {
      "External id": 163819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163819,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163819, "pid": 5, "tid": 7, "ts": 1716454223984734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875109, "dur": 8,
    "args": {
      "External id": 163819,
      "cbid": 211, "correlation": 163819
    }
  },
  {
    "ph": "s", "id": 163819, "pid": 76337, "tid": -914061504, "ts": 1716454223875109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454223984944, "dur": 39,
    "args": {
      "External id": 163821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163821,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163821, "pid": 5, "tid": 7, "ts": 1716454223984944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875121, "dur": 122,
    "args": {
      "External id": 163821,
      "cbid": 211, "correlation": 163821
    }
  },
  {
    "ph": "s", "id": 163821, "pid": 76337, "tid": -914061504, "ts": 1716454223875121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223984983, "dur": 59,
    "args": {
      "External id": 163827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163827,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163827, "pid": 5, "tid": 7, "ts": 1716454223984983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875267, "dur": 9,
    "args": {
      "External id": 163827,
      "cbid": 211, "correlation": 163827
    }
  },
  {
    "ph": "s", "id": 163827, "pid": 76337, "tid": -914061504, "ts": 1716454223875267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454223985044, "dur": 50,
    "args": {
      "External id": 163835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163835,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163835, "pid": 5, "tid": 7, "ts": 1716454223985044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875297, "dur": 8,
    "args": {
      "External id": 163835,
      "cbid": 211, "correlation": 163835
    }
  },
  {
    "ph": "s", "id": 163835, "pid": 76337, "tid": -914061504, "ts": 1716454223875297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454223985095, "dur": 35,
    "args": {
      "External id": 163843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163843,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163843, "pid": 5, "tid": 7, "ts": 1716454223985095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875326, "dur": 28,
    "args": {
      "External id": 163843,
      "cbid": 211, "correlation": 163843
    }
  },
  {
    "ph": "s", "id": 163843, "pid": 76337, "tid": -914061504, "ts": 1716454223875326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223985131, "dur": 51,
    "args": {
      "External id": 163863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163863,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 163863, "pid": 5, "tid": 7, "ts": 1716454223985131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875430, "dur": 13,
    "args": {
      "External id": 163863,
      "cbid": 211, "correlation": 163863
    }
  },
  {
    "ph": "s", "id": 163863, "pid": 76337, "tid": -914061504, "ts": 1716454223875430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454223985184, "dur": 4,
    "args": {
      "External id": 163875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163875,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 163875, "pid": 5, "tid": 7, "ts": 1716454223985184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875453, "dur": 6,
    "args": {
      "External id": 163875,
      "cbid": 211, "correlation": 163875
    }
  },
  {
    "ph": "s", "id": 163875, "pid": 76337, "tid": -914061504, "ts": 1716454223875453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223985190, "dur": 56,
    "args": {
      "External id": 163878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163878, "pid": 5, "tid": 7, "ts": 1716454223985190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875471, "dur": 6,
    "args": {
      "External id": 163878,
      "cbid": 211, "correlation": 163878
    }
  },
  {
    "ph": "s", "id": 163878, "pid": 76337, "tid": -914061504, "ts": 1716454223875471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875528, "dur": 0,
    "args": {
      "External id": 163889,
      "cbid": 317, "correlation": 163889
    }
  },
  {
    "ph": "f", "id": 163889, "pid": 76337, "tid": -914061504, "ts": 1716454223875528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875529, "dur": 0,
    "args": {
      "External id": 163890,
      "cbid": 203, "correlation": 163890
    }
  },
  {
    "ph": "f", "id": 163890, "pid": 76337, "tid": -914061504, "ts": 1716454223875529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875529, "dur": 0,
    "args": {
      "External id": 163891,
      "cbid": 205, "correlation": 163891
    }
  },
  {
    "ph": "f", "id": 163891, "pid": 76337, "tid": -914061504, "ts": 1716454223875529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875552, "dur": 1,
    "args": {
      "External id": 163895,
      "cbid": 251, "correlation": 163895
    }
  },
  {
    "ph": "f", "id": 163895, "pid": 76337, "tid": -914061504, "ts": 1716454223875552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875555, "dur": 0,
    "args": {
      "External id": 163896,
      "cbid": 251, "correlation": 163896
    }
  },
  {
    "ph": "f", "id": 163896, "pid": 76337, "tid": -914061504, "ts": 1716454223875555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875555, "dur": 0,
    "args": {
      "External id": 163897,
      "cbid": 251, "correlation": 163897
    }
  },
  {
    "ph": "f", "id": 163897, "pid": 76337, "tid": -914061504, "ts": 1716454223875555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875557, "dur": 0,
    "args": {
      "External id": 163898,
      "cbid": 251, "correlation": 163898
    }
  },
  {
    "ph": "f", "id": 163898, "pid": 76337, "tid": -914061504, "ts": 1716454223875557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875557, "dur": 0,
    "args": {
      "External id": 163899,
      "cbid": 251, "correlation": 163899
    }
  },
  {
    "ph": "f", "id": 163899, "pid": 76337, "tid": -914061504, "ts": 1716454223875557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875558, "dur": 0,
    "args": {
      "External id": 163900,
      "cbid": 251, "correlation": 163900
    }
  },
  {
    "ph": "f", "id": 163900, "pid": 76337, "tid": -914061504, "ts": 1716454223875558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875559, "dur": 0,
    "args": {
      "External id": 163901,
      "cbid": 251, "correlation": 163901
    }
  },
  {
    "ph": "f", "id": 163901, "pid": 76337, "tid": -914061504, "ts": 1716454223875559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875560, "dur": 0,
    "args": {
      "External id": 163902,
      "cbid": 251, "correlation": 163902
    }
  },
  {
    "ph": "f", "id": 163902, "pid": 76337, "tid": -914061504, "ts": 1716454223875560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875562, "dur": 0,
    "args": {
      "External id": 163903,
      "cbid": 251, "correlation": 163903
    }
  },
  {
    "ph": "f", "id": 163903, "pid": 76337, "tid": -914061504, "ts": 1716454223875562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454223985247, "dur": 114,
    "args": {
      "External id": 163904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163904,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163904, "pid": 5, "tid": 7, "ts": 1716454223985247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875564, "dur": 12,
    "args": {
      "External id": 163904,
      "cbid": 211, "correlation": 163904
    }
  },
  {
    "ph": "s", "id": 163904, "pid": 76337, "tid": -914061504, "ts": 1716454223875564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454223985362, "dur": 59,
    "args": {
      "External id": 163910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163910,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163910, "pid": 5, "tid": 7, "ts": 1716454223985362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875599, "dur": 9,
    "args": {
      "External id": 163910,
      "cbid": 211, "correlation": 163910
    }
  },
  {
    "ph": "s", "id": 163910, "pid": 76337, "tid": -914061504, "ts": 1716454223875599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454223985423, "dur": 631,
    "args": {
      "External id": 163919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163919,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163919, "pid": 5, "tid": 7, "ts": 1716454223985423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875682, "dur": 14,
    "args": {
      "External id": 163919,
      "cbid": 211, "correlation": 163919
    }
  },
  {
    "ph": "s", "id": 163919, "pid": 76337, "tid": -914061504, "ts": 1716454223875682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454223986056, "dur": 184,
    "args": {
      "External id": 163941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163941,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 163941, "pid": 5, "tid": 7, "ts": 1716454223986056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875739, "dur": 10,
    "args": {
      "External id": 163941,
      "cbid": 211, "correlation": 163941
    }
  },
  {
    "ph": "s", "id": 163941, "pid": 76337, "tid": -914061504, "ts": 1716454223875739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875826, "dur": 1,
    "args": {
      "External id": 163952,
      "cbid": 251, "correlation": 163952
    }
  },
  {
    "ph": "f", "id": 163952, "pid": 76337, "tid": -914061504, "ts": 1716454223875826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223986241, "dur": 199,
    "args": {
      "External id": 163953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163953,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163953, "pid": 5, "tid": 7, "ts": 1716454223986241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875831, "dur": 14,
    "args": {
      "External id": 163953,
      "cbid": 211, "correlation": 163953
    }
  },
  {
    "ph": "s", "id": 163953, "pid": 76337, "tid": -914061504, "ts": 1716454223875831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875901, "dur": 1,
    "args": {
      "External id": 163964,
      "cbid": 251, "correlation": 163964
    }
  },
  {
    "ph": "f", "id": 163964, "pid": 76337, "tid": -914061504, "ts": 1716454223875901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223986441, "dur": 189,
    "args": {
      "External id": 163965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163965,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163965, "pid": 5, "tid": 7, "ts": 1716454223986441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875906, "dur": 12,
    "args": {
      "External id": 163965,
      "cbid": 211, "correlation": 163965
    }
  },
  {
    "ph": "s", "id": 163965, "pid": 76337, "tid": -914061504, "ts": 1716454223875906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223875969, "dur": 1,
    "args": {
      "External id": 163976,
      "cbid": 251, "correlation": 163976
    }
  },
  {
    "ph": "f", "id": 163976, "pid": 76337, "tid": -914061504, "ts": 1716454223875969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454223986632, "dur": 192,
    "args": {
      "External id": 163977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163977,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 163977, "pid": 5, "tid": 7, "ts": 1716454223986632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876080, "dur": 13,
    "args": {
      "External id": 163977,
      "cbid": 211, "correlation": 163977
    }
  },
  {
    "ph": "s", "id": 163977, "pid": 76337, "tid": -914061504, "ts": 1716454223876080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454223986825, "dur": 18867,
    "args": {
      "External id": 163998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 163998,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 163998, "pid": 5, "tid": 7, "ts": 1716454223986825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876173, "dur": 12,
    "args": {
      "External id": 163998,
      "cbid": 211, "correlation": 163998
    }
  },
  {
    "ph": "s", "id": 163998, "pid": 76337, "tid": -914061504, "ts": 1716454223876173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876275, "dur": 1,
    "args": {
      "External id": 164016,
      "cbid": 251, "correlation": 164016
    }
  },
  {
    "ph": "f", "id": 164016, "pid": 76337, "tid": -914061504, "ts": 1716454223876275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224005694, "dur": 203,
    "args": {
      "External id": 164018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164018,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164018, "pid": 5, "tid": 7, "ts": 1716454224005694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876281, "dur": 13,
    "args": {
      "External id": 164018,
      "cbid": 211, "correlation": 164018
    }
  },
  {
    "ph": "s", "id": 164018, "pid": 76337, "tid": -914061504, "ts": 1716454223876281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224005898, "dur": 66,
    "args": {
      "External id": 164026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164026,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164026, "pid": 5, "tid": 7, "ts": 1716454224005898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876352, "dur": 12,
    "args": {
      "External id": 164026,
      "cbid": 211, "correlation": 164026
    }
  },
  {
    "ph": "s", "id": 164026, "pid": 76337, "tid": -914061504, "ts": 1716454223876352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224005966, "dur": 96,
    "args": {
      "External id": 164034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164034,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164034, "pid": 5, "tid": 7, "ts": 1716454224005966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876391, "dur": 10,
    "args": {
      "External id": 164034,
      "cbid": 211, "correlation": 164034
    }
  },
  {
    "ph": "s", "id": 164034, "pid": 76337, "tid": -914061504, "ts": 1716454223876391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224006064, "dur": 55,
    "args": {
      "External id": 164045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164045,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164045, "pid": 5, "tid": 7, "ts": 1716454224006064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876464, "dur": 13,
    "args": {
      "External id": 164045,
      "cbid": 211, "correlation": 164045
    }
  },
  {
    "ph": "s", "id": 164045, "pid": 76337, "tid": -914061504, "ts": 1716454223876464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224006120, "dur": 93,
    "args": {
      "External id": 164067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164067,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164067, "pid": 5, "tid": 7, "ts": 1716454224006120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223876496, "dur": 1854,
    "args": {
      "External id": 164067,
      "cbid": 211, "correlation": 164067
    }
  },
  {
    "ph": "s", "id": 164067, "pid": 76337, "tid": -914061504, "ts": 1716454223876496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878429, "dur": 1,
    "args": {
      "External id": 164078,
      "cbid": 251, "correlation": 164078
    }
  },
  {
    "ph": "f", "id": 164078, "pid": 76337, "tid": -914061504, "ts": 1716454223878429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224006215, "dur": 107,
    "args": {
      "External id": 164079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164079,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164079, "pid": 5, "tid": 7, "ts": 1716454224006215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878434, "dur": 64,
    "args": {
      "External id": 164079,
      "cbid": 211, "correlation": 164079
    }
  },
  {
    "ph": "s", "id": 164079, "pid": 76337, "tid": -914061504, "ts": 1716454223878434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878560, "dur": 1,
    "args": {
      "External id": 164090,
      "cbid": 251, "correlation": 164090
    }
  },
  {
    "ph": "f", "id": 164090, "pid": 76337, "tid": -914061504, "ts": 1716454223878560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878564, "dur": 0,
    "args": {
      "External id": 164091,
      "cbid": 251, "correlation": 164091
    }
  },
  {
    "ph": "f", "id": 164091, "pid": 76337, "tid": -914061504, "ts": 1716454223878564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224006322, "dur": 10,
    "args": {
      "External id": 164092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164092,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 164092, "pid": 5, "tid": 7, "ts": 1716454224006322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878566, "dur": 13,
    "args": {
      "External id": 164092,
      "cbid": 211, "correlation": 164092
    }
  },
  {
    "ph": "s", "id": 164092, "pid": 76337, "tid": -914061504, "ts": 1716454223878566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224006334, "dur": 5,
    "args": {
      "External id": 164094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164094,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 164094, "pid": 5, "tid": 7, "ts": 1716454224006334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878580, "dur": 6,
    "args": {
      "External id": 164094,
      "cbid": 211, "correlation": 164094
    }
  },
  {
    "ph": "s", "id": 164094, "pid": 76337, "tid": -914061504, "ts": 1716454223878580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878641, "dur": 1,
    "args": {
      "External id": 164105,
      "cbid": 251, "correlation": 164105
    }
  },
  {
    "ph": "f", "id": 164105, "pid": 76337, "tid": -914061504, "ts": 1716454223878641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878645, "dur": 0,
    "args": {
      "External id": 164106,
      "cbid": 251, "correlation": 164106
    }
  },
  {
    "ph": "f", "id": 164106, "pid": 76337, "tid": -914061504, "ts": 1716454223878645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224006340, "dur": 6,
    "args": {
      "External id": 164107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164107,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 164107, "pid": 5, "tid": 7, "ts": 1716454224006340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878646, "dur": 12,
    "args": {
      "External id": 164107,
      "cbid": 211, "correlation": 164107
    }
  },
  {
    "ph": "s", "id": 164107, "pid": 76337, "tid": -914061504, "ts": 1716454223878646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224006348, "dur": 4,
    "args": {
      "External id": 164109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164109,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 164109, "pid": 5, "tid": 7, "ts": 1716454224006348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878659, "dur": 6,
    "args": {
      "External id": 164109,
      "cbid": 211, "correlation": 164109
    }
  },
  {
    "ph": "s", "id": 164109, "pid": 76337, "tid": -914061504, "ts": 1716454223878659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224006353, "dur": 158,
    "args": {
      "External id": 164130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164130,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164130, "pid": 5, "tid": 7, "ts": 1716454224006353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878734, "dur": 12,
    "args": {
      "External id": 164130,
      "cbid": 211, "correlation": 164130
    }
  },
  {
    "ph": "s", "id": 164130, "pid": 76337, "tid": -914061504, "ts": 1716454223878734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878831, "dur": 2,
    "args": {
      "External id": 164148,
      "cbid": 251, "correlation": 164148
    }
  },
  {
    "ph": "f", "id": 164148, "pid": 76337, "tid": -914061504, "ts": 1716454223878831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224006512, "dur": 107,
    "args": {
      "External id": 164150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164150,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164150, "pid": 5, "tid": 7, "ts": 1716454224006512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878837, "dur": 13,
    "args": {
      "External id": 164150,
      "cbid": 211, "correlation": 164150
    }
  },
  {
    "ph": "s", "id": 164150, "pid": 76337, "tid": -914061504, "ts": 1716454223878837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224006621, "dur": 35,
    "args": {
      "External id": 164158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164158,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164158, "pid": 5, "tid": 7, "ts": 1716454224006621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878907, "dur": 13,
    "args": {
      "External id": 164158,
      "cbid": 211, "correlation": 164158
    }
  },
  {
    "ph": "s", "id": 164158, "pid": 76337, "tid": -914061504, "ts": 1716454223878907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224006657, "dur": 68,
    "args": {
      "External id": 164166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164166, "pid": 5, "tid": 7, "ts": 1716454224006657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223878949, "dur": 9,
    "args": {
      "External id": 164166,
      "cbid": 211, "correlation": 164166
    }
  },
  {
    "ph": "s", "id": 164166, "pid": 76337, "tid": -914061504, "ts": 1716454223878949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224006726, "dur": 93,
    "args": {
      "External id": 164188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164188,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164188, "pid": 5, "tid": 7, "ts": 1716454224006726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879009, "dur": 11,
    "args": {
      "External id": 164188,
      "cbid": 211, "correlation": 164188
    }
  },
  {
    "ph": "s", "id": 164188, "pid": 76337, "tid": -914061504, "ts": 1716454223879009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879097, "dur": 1,
    "args": {
      "External id": 164204,
      "cbid": 251, "correlation": 164204
    }
  },
  {
    "ph": "f", "id": 164204, "pid": 76337, "tid": -914061504, "ts": 1716454223879097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224006821, "dur": 579,
    "args": {
      "External id": 164206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164206,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164206, "pid": 5, "tid": 7, "ts": 1716454224006821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879103, "dur": 13,
    "args": {
      "External id": 164206,
      "cbid": 211, "correlation": 164206
    }
  },
  {
    "ph": "s", "id": 164206, "pid": 76337, "tid": -914061504, "ts": 1716454223879103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224007401, "dur": 244,
    "args": {
      "External id": 164214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164214, "pid": 5, "tid": 7, "ts": 1716454224007401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879169, "dur": 12,
    "args": {
      "External id": 164214,
      "cbid": 211, "correlation": 164214
    }
  },
  {
    "ph": "s", "id": 164214, "pid": 76337, "tid": -914061504, "ts": 1716454223879169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224007647, "dur": 252,
    "args": {
      "External id": 164222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164222, "pid": 5, "tid": 7, "ts": 1716454224007647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879199, "dur": 9,
    "args": {
      "External id": 164222,
      "cbid": 211, "correlation": 164222
    }
  },
  {
    "ph": "s", "id": 164222, "pid": 76337, "tid": -914061504, "ts": 1716454223879199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879281, "dur": 1,
    "args": {
      "External id": 164238,
      "cbid": 251, "correlation": 164238
    }
  },
  {
    "ph": "f", "id": 164238, "pid": 76337, "tid": -914061504, "ts": 1716454223879281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879286, "dur": 0,
    "args": {
      "External id": 164240,
      "cbid": 251, "correlation": 164240
    }
  },
  {
    "ph": "f", "id": 164240, "pid": 76337, "tid": -914061504, "ts": 1716454223879286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224007900, "dur": 362,
    "args": {
      "External id": 164241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164241,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 164241, "pid": 5, "tid": 7, "ts": 1716454224007900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879289, "dur": 12,
    "args": {
      "External id": 164241,
      "cbid": 211, "correlation": 164241
    }
  },
  {
    "ph": "s", "id": 164241, "pid": 76337, "tid": -914061504, "ts": 1716454223879289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224008263, "dur": 50,
    "args": {
      "External id": 164249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164249,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164249, "pid": 5, "tid": 7, "ts": 1716454224008263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879330, "dur": 10,
    "args": {
      "External id": 164249,
      "cbid": 211, "correlation": 164249
    }
  },
  {
    "ph": "s", "id": 164249, "pid": 76337, "tid": -914061504, "ts": 1716454223879330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224008315, "dur": 159,
    "args": {
      "External id": 164260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164260,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164260, "pid": 5, "tid": 7, "ts": 1716454224008315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879397, "dur": 208,
    "args": {
      "External id": 164260,
      "cbid": 211, "correlation": 164260
    }
  },
  {
    "ph": "s", "id": 164260, "pid": 76337, "tid": -914061504, "ts": 1716454223879397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879659, "dur": 0,
    "args": {
      "External id": 164272,
      "cbid": 317, "correlation": 164272
    }
  },
  {
    "ph": "f", "id": 164272, "pid": 76337, "tid": -914061504, "ts": 1716454223879659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879660, "dur": 0,
    "args": {
      "External id": 164273,
      "cbid": 203, "correlation": 164273
    }
  },
  {
    "ph": "f", "id": 164273, "pid": 76337, "tid": -914061504, "ts": 1716454223879660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879661, "dur": 0,
    "args": {
      "External id": 164274,
      "cbid": 205, "correlation": 164274
    }
  },
  {
    "ph": "f", "id": 164274, "pid": 76337, "tid": -914061504, "ts": 1716454223879661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879684, "dur": 1,
    "args": {
      "External id": 164278,
      "cbid": 251, "correlation": 164278
    }
  },
  {
    "ph": "f", "id": 164278, "pid": 76337, "tid": -914061504, "ts": 1716454223879684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879686, "dur": 0,
    "args": {
      "External id": 164279,
      "cbid": 251, "correlation": 164279
    }
  },
  {
    "ph": "f", "id": 164279, "pid": 76337, "tid": -914061504, "ts": 1716454223879686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879687, "dur": 0,
    "args": {
      "External id": 164280,
      "cbid": 251, "correlation": 164280
    }
  },
  {
    "ph": "f", "id": 164280, "pid": 76337, "tid": -914061504, "ts": 1716454223879687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879688, "dur": 0,
    "args": {
      "External id": 164281,
      "cbid": 251, "correlation": 164281
    }
  },
  {
    "ph": "f", "id": 164281, "pid": 76337, "tid": -914061504, "ts": 1716454223879688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879688, "dur": 0,
    "args": {
      "External id": 164282,
      "cbid": 251, "correlation": 164282
    }
  },
  {
    "ph": "f", "id": 164282, "pid": 76337, "tid": -914061504, "ts": 1716454223879688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879689, "dur": 0,
    "args": {
      "External id": 164283,
      "cbid": 251, "correlation": 164283
    }
  },
  {
    "ph": "f", "id": 164283, "pid": 76337, "tid": -914061504, "ts": 1716454223879689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879690, "dur": 0,
    "args": {
      "External id": 164284,
      "cbid": 251, "correlation": 164284
    }
  },
  {
    "ph": "f", "id": 164284, "pid": 76337, "tid": -914061504, "ts": 1716454223879690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879691, "dur": 0,
    "args": {
      "External id": 164285,
      "cbid": 251, "correlation": 164285
    }
  },
  {
    "ph": "f", "id": 164285, "pid": 76337, "tid": -914061504, "ts": 1716454223879691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879692, "dur": 0,
    "args": {
      "External id": 164286,
      "cbid": 251, "correlation": 164286
    }
  },
  {
    "ph": "f", "id": 164286, "pid": 76337, "tid": -914061504, "ts": 1716454223879692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224008475, "dur": 115,
    "args": {
      "External id": 164287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164287,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164287, "pid": 5, "tid": 7, "ts": 1716454224008475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879694, "dur": 38,
    "args": {
      "External id": 164287,
      "cbid": 211, "correlation": 164287
    }
  },
  {
    "ph": "s", "id": 164287, "pid": 76337, "tid": -914061504, "ts": 1716454223879694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224008592, "dur": 60,
    "args": {
      "External id": 164293,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164293,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164293, "pid": 5, "tid": 7, "ts": 1716454224008592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879755, "dur": 104,
    "args": {
      "External id": 164293,
      "cbid": 211, "correlation": 164293
    }
  },
  {
    "ph": "s", "id": 164293, "pid": 76337, "tid": -914061504, "ts": 1716454223879755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224008653, "dur": 50,
    "args": {
      "External id": 164301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164301,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164301, "pid": 5, "tid": 7, "ts": 1716454224008653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223879883, "dur": 283,
    "args": {
      "External id": 164301,
      "cbid": 211, "correlation": 164301
    }
  },
  {
    "ph": "s", "id": 164301, "pid": 76337, "tid": -914061504, "ts": 1716454223879883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224008704, "dur": 99,
    "args": {
      "External id": 164310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164310,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164310, "pid": 5, "tid": 7, "ts": 1716454224008704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880198, "dur": 10,
    "args": {
      "External id": 164310,
      "cbid": 211, "correlation": 164310
    }
  },
  {
    "ph": "s", "id": 164310, "pid": 76337, "tid": -914061504, "ts": 1716454223880198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224008805, "dur": 92,
    "args": {
      "External id": 164330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164330,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 164330, "pid": 5, "tid": 7, "ts": 1716454224008805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880269, "dur": 12,
    "args": {
      "External id": 164330,
      "cbid": 211, "correlation": 164330
    }
  },
  {
    "ph": "s", "id": 164330, "pid": 76337, "tid": -914061504, "ts": 1716454223880269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224008898, "dur": 5,
    "args": {
      "External id": 164342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164342,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 164342, "pid": 5, "tid": 7, "ts": 1716454224008898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880291, "dur": 10,
    "args": {
      "External id": 164342,
      "cbid": 211, "correlation": 164342
    }
  },
  {
    "ph": "s", "id": 164342, "pid": 76337, "tid": -914061504, "ts": 1716454223880291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224008904, "dur": 112,
    "args": {
      "External id": 164345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164345,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164345, "pid": 5, "tid": 7, "ts": 1716454224008904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880313, "dur": 108,
    "args": {
      "External id": 164345,
      "cbid": 211, "correlation": 164345
    }
  },
  {
    "ph": "s", "id": 164345, "pid": 76337, "tid": -914061504, "ts": 1716454223880313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224009017, "dur": 69,
    "args": {
      "External id": 164354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164354,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164354, "pid": 5, "tid": 7, "ts": 1716454224009017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880456, "dur": 10,
    "args": {
      "External id": 164354,
      "cbid": 211, "correlation": 164354
    }
  },
  {
    "ph": "s", "id": 164354, "pid": 76337, "tid": -914061504, "ts": 1716454223880456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880508, "dur": 0,
    "args": {
      "External id": 164364,
      "cbid": 317, "correlation": 164364
    }
  },
  {
    "ph": "f", "id": 164364, "pid": 76337, "tid": -914061504, "ts": 1716454223880508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880509, "dur": 0,
    "args": {
      "External id": 164365,
      "cbid": 203, "correlation": 164365
    }
  },
  {
    "ph": "f", "id": 164365, "pid": 76337, "tid": -914061504, "ts": 1716454223880509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880509, "dur": 0,
    "args": {
      "External id": 164366,
      "cbid": 205, "correlation": 164366
    }
  },
  {
    "ph": "f", "id": 164366, "pid": 76337, "tid": -914061504, "ts": 1716454223880509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224009088, "dur": 76,
    "args": {
      "External id": 164370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164370,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164370, "pid": 5, "tid": 7, "ts": 1716454224009088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880525, "dur": 12,
    "args": {
      "External id": 164370,
      "cbid": 211, "correlation": 164370
    }
  },
  {
    "ph": "s", "id": 164370, "pid": 76337, "tid": -914061504, "ts": 1716454223880525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224009165, "dur": 25,
    "args": {
      "External id": 164372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164372,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164372, "pid": 5, "tid": 7, "ts": 1716454224009165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880539, "dur": 5,
    "args": {
      "External id": 164372,
      "cbid": 211, "correlation": 164372
    }
  },
  {
    "ph": "s", "id": 164372, "pid": 76337, "tid": -914061504, "ts": 1716454223880539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224009191, "dur": 4,
    "args": {
      "External id": 164374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 164374, "pid": 5, "tid": 7, "ts": 1716454224009191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880549, "dur": 5,
    "args": {
      "External id": 164374,
      "cbid": 211, "correlation": 164374
    }
  },
  {
    "ph": "s", "id": 164374, "pid": 76337, "tid": -914061504, "ts": 1716454223880549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880558, "dur": 0,
    "args": {
      "External id": 164375,
      "cbid": 51, "correlation": 164375
    }
  },
  {
    "ph": "s", "id": 164375, "pid": 76337, "tid": -914061504, "ts": 1716454223880558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224009196, "dur": 1379,
    "args": {
      "External id": 164376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164376,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164376, "pid": 5, "tid": 7, "ts": 1716454224009196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880559, "dur": 5,
    "args": {
      "External id": 164376,
      "cbid": 211, "correlation": 164376
    }
  },
  {
    "ph": "s", "id": 164376, "pid": 76337, "tid": -914061504, "ts": 1716454223880559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224010576, "dur": 60,
    "args": {
      "External id": 164381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164381,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164381, "pid": 5, "tid": 7, "ts": 1716454224010576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880586, "dur": 9,
    "args": {
      "External id": 164381,
      "cbid": 211, "correlation": 164381
    }
  },
  {
    "ph": "s", "id": 164381, "pid": 76337, "tid": -914061504, "ts": 1716454223880586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224010638, "dur": 3,
    "args": {
      "External id": 164389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164389,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 164389, "pid": 5, "tid": 7, "ts": 1716454224010638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880630, "dur": 10,
    "args": {
      "External id": 164389,
      "cbid": 211, "correlation": 164389
    }
  },
  {
    "ph": "s", "id": 164389, "pid": 76337, "tid": -914061504, "ts": 1716454223880630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880697, "dur": 1,
    "args": {
      "External id": 164405,
      "cbid": 251, "correlation": 164405
    }
  },
  {
    "ph": "f", "id": 164405, "pid": 76337, "tid": -914061504, "ts": 1716454223880697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880703, "dur": 0,
    "args": {
      "External id": 164407,
      "cbid": 251, "correlation": 164407
    }
  },
  {
    "ph": "f", "id": 164407, "pid": 76337, "tid": -914061504, "ts": 1716454223880703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224010642, "dur": 11,
    "args": {
      "External id": 164408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164408,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 164408, "pid": 5, "tid": 7, "ts": 1716454224010642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880705, "dur": 12,
    "args": {
      "External id": 164408,
      "cbid": 211, "correlation": 164408
    }
  },
  {
    "ph": "s", "id": 164408, "pid": 76337, "tid": -914061504, "ts": 1716454223880705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224010655, "dur": 5,
    "args": {
      "External id": 164410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164410,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 164410, "pid": 5, "tid": 7, "ts": 1716454224010655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880718, "dur": 6,
    "args": {
      "External id": 164410,
      "cbid": 211, "correlation": 164410
    }
  },
  {
    "ph": "s", "id": 164410, "pid": 76337, "tid": -914061504, "ts": 1716454223880718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224010662, "dur": 54,
    "args": {
      "External id": 164420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164420,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164420, "pid": 5, "tid": 7, "ts": 1716454224010662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223880777, "dur": 551,
    "args": {
      "External id": 164420,
      "cbid": 211, "correlation": 164420
    }
  },
  {
    "ph": "s", "id": 164420, "pid": 76337, "tid": -914061504, "ts": 1716454223880777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224010717, "dur": 53,
    "args": {
      "External id": 164440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164440,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 164440, "pid": 5, "tid": 7, "ts": 1716454224010717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881383, "dur": 11,
    "args": {
      "External id": 164440,
      "cbid": 211, "correlation": 164440
    }
  },
  {
    "ph": "s", "id": 164440, "pid": 76337, "tid": -914061504, "ts": 1716454223881383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224010771, "dur": 4,
    "args": {
      "External id": 164452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164452,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 164452, "pid": 5, "tid": 7, "ts": 1716454224010771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881403, "dur": 7,
    "args": {
      "External id": 164452,
      "cbid": 211, "correlation": 164452
    }
  },
  {
    "ph": "s", "id": 164452, "pid": 76337, "tid": -914061504, "ts": 1716454223881403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224010776, "dur": 56,
    "args": {
      "External id": 164455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164455,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164455, "pid": 5, "tid": 7, "ts": 1716454224010776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881423, "dur": 7,
    "args": {
      "External id": 164455,
      "cbid": 211, "correlation": 164455
    }
  },
  {
    "ph": "s", "id": 164455, "pid": 76337, "tid": -914061504, "ts": 1716454223881423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224010833, "dur": 37,
    "args": {
      "External id": 164464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164464,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164464, "pid": 5, "tid": 7, "ts": 1716454224010833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881463, "dur": 10,
    "args": {
      "External id": 164464,
      "cbid": 211, "correlation": 164464
    }
  },
  {
    "ph": "s", "id": 164464, "pid": 76337, "tid": -914061504, "ts": 1716454223881463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881527, "dur": 0,
    "args": {
      "External id": 164474,
      "cbid": 317, "correlation": 164474
    }
  },
  {
    "ph": "f", "id": 164474, "pid": 76337, "tid": -914061504, "ts": 1716454223881527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881528, "dur": 0,
    "args": {
      "External id": 164475,
      "cbid": 203, "correlation": 164475
    }
  },
  {
    "ph": "f", "id": 164475, "pid": 76337, "tid": -914061504, "ts": 1716454223881528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881529, "dur": 0,
    "args": {
      "External id": 164476,
      "cbid": 205, "correlation": 164476
    }
  },
  {
    "ph": "f", "id": 164476, "pid": 76337, "tid": -914061504, "ts": 1716454223881529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224010871, "dur": 41,
    "args": {
      "External id": 164480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164480,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164480, "pid": 5, "tid": 7, "ts": 1716454224010871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881543, "dur": 12,
    "args": {
      "External id": 164480,
      "cbid": 211, "correlation": 164480
    }
  },
  {
    "ph": "s", "id": 164480, "pid": 76337, "tid": -914061504, "ts": 1716454223881543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224010914, "dur": 14,
    "args": {
      "External id": 164482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164482,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164482, "pid": 5, "tid": 7, "ts": 1716454224010914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881557, "dur": 6,
    "args": {
      "External id": 164482,
      "cbid": 211, "correlation": 164482
    }
  },
  {
    "ph": "s", "id": 164482, "pid": 76337, "tid": -914061504, "ts": 1716454223881557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224010929, "dur": 3,
    "args": {
      "External id": 164484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 164484, "pid": 5, "tid": 7, "ts": 1716454224010929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881567, "dur": 5,
    "args": {
      "External id": 164484,
      "cbid": 211, "correlation": 164484
    }
  },
  {
    "ph": "s", "id": 164484, "pid": 76337, "tid": -914061504, "ts": 1716454223881567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881576, "dur": 0,
    "args": {
      "External id": 164485,
      "cbid": 51, "correlation": 164485
    }
  },
  {
    "ph": "s", "id": 164485, "pid": 76337, "tid": -914061504, "ts": 1716454223881576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224010934, "dur": 703,
    "args": {
      "External id": 164486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164486,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164486, "pid": 5, "tid": 7, "ts": 1716454224010934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881577, "dur": 5,
    "args": {
      "External id": 164486,
      "cbid": 211, "correlation": 164486
    }
  },
  {
    "ph": "s", "id": 164486, "pid": 76337, "tid": -914061504, "ts": 1716454223881577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224011639, "dur": 60,
    "args": {
      "External id": 164491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164491,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164491, "pid": 5, "tid": 7, "ts": 1716454224011639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881604, "dur": 9,
    "args": {
      "External id": 164491,
      "cbid": 211, "correlation": 164491
    }
  },
  {
    "ph": "s", "id": 164491, "pid": 76337, "tid": -914061504, "ts": 1716454223881604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881662, "dur": 0,
    "args": {
      "External id": 164501,
      "cbid": 317, "correlation": 164501
    }
  },
  {
    "ph": "f", "id": 164501, "pid": 76337, "tid": -914061504, "ts": 1716454223881662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881663, "dur": 0,
    "args": {
      "External id": 164502,
      "cbid": 203, "correlation": 164502
    }
  },
  {
    "ph": "f", "id": 164502, "pid": 76337, "tid": -914061504, "ts": 1716454223881663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881663, "dur": 0,
    "args": {
      "External id": 164503,
      "cbid": 205, "correlation": 164503
    }
  },
  {
    "ph": "f", "id": 164503, "pid": 76337, "tid": -914061504, "ts": 1716454223881663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224011700, "dur": 75,
    "args": {
      "External id": 164507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164507,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164507, "pid": 5, "tid": 7, "ts": 1716454224011700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881676, "dur": 12,
    "args": {
      "External id": 164507,
      "cbid": 211, "correlation": 164507
    }
  },
  {
    "ph": "s", "id": 164507, "pid": 76337, "tid": -914061504, "ts": 1716454223881676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224011776, "dur": 211,
    "args": {
      "External id": 164509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164509,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164509, "pid": 5, "tid": 7, "ts": 1716454224011776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881694, "dur": 6,
    "args": {
      "External id": 164509,
      "cbid": 211, "correlation": 164509
    }
  },
  {
    "ph": "s", "id": 164509, "pid": 76337, "tid": -914061504, "ts": 1716454223881694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224011988, "dur": 39,
    "args": {
      "External id": 164511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164511,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164511, "pid": 5, "tid": 7, "ts": 1716454224011988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881705, "dur": 6,
    "args": {
      "External id": 164511,
      "cbid": 211, "correlation": 164511
    }
  },
  {
    "ph": "s", "id": 164511, "pid": 76337, "tid": -914061504, "ts": 1716454223881705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224012029, "dur": 60,
    "args": {
      "External id": 164517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164517, "pid": 5, "tid": 7, "ts": 1716454224012029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223881733, "dur": 510,
    "args": {
      "External id": 164517,
      "cbid": 211, "correlation": 164517
    }
  },
  {
    "ph": "s", "id": 164517, "pid": 76337, "tid": -914061504, "ts": 1716454223881733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224012090, "dur": 50,
    "args": {
      "External id": 164525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164525,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164525, "pid": 5, "tid": 7, "ts": 1716454224012090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882264, "dur": 8,
    "args": {
      "External id": 164525,
      "cbid": 211, "correlation": 164525
    }
  },
  {
    "ph": "s", "id": 164525, "pid": 76337, "tid": -914061504, "ts": 1716454223882264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224012142, "dur": 36,
    "args": {
      "External id": 164533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164533,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164533, "pid": 5, "tid": 7, "ts": 1716454224012142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882294, "dur": 8,
    "args": {
      "External id": 164533,
      "cbid": 211, "correlation": 164533
    }
  },
  {
    "ph": "s", "id": 164533, "pid": 76337, "tid": -914061504, "ts": 1716454223882294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224012179, "dur": 50,
    "args": {
      "External id": 164553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164553,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 164553, "pid": 5, "tid": 7, "ts": 1716454224012179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882377, "dur": 13,
    "args": {
      "External id": 164553,
      "cbid": 211, "correlation": 164553
    }
  },
  {
    "ph": "s", "id": 164553, "pid": 76337, "tid": -914061504, "ts": 1716454223882377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224012230, "dur": 4,
    "args": {
      "External id": 164565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164565,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 164565, "pid": 5, "tid": 7, "ts": 1716454224012230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882400, "dur": 6,
    "args": {
      "External id": 164565,
      "cbid": 211, "correlation": 164565
    }
  },
  {
    "ph": "s", "id": 164565, "pid": 76337, "tid": -914061504, "ts": 1716454223882400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224012236, "dur": 55,
    "args": {
      "External id": 164568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164568, "pid": 5, "tid": 7, "ts": 1716454224012236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882417, "dur": 7,
    "args": {
      "External id": 164568,
      "cbid": 211, "correlation": 164568
    }
  },
  {
    "ph": "s", "id": 164568, "pid": 76337, "tid": -914061504, "ts": 1716454223882417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882474, "dur": 0,
    "args": {
      "External id": 164579,
      "cbid": 317, "correlation": 164579
    }
  },
  {
    "ph": "f", "id": 164579, "pid": 76337, "tid": -914061504, "ts": 1716454223882474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882475, "dur": 0,
    "args": {
      "External id": 164580,
      "cbid": 203, "correlation": 164580
    }
  },
  {
    "ph": "f", "id": 164580, "pid": 76337, "tid": -914061504, "ts": 1716454223882475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882476, "dur": 0,
    "args": {
      "External id": 164581,
      "cbid": 205, "correlation": 164581
    }
  },
  {
    "ph": "f", "id": 164581, "pid": 76337, "tid": -914061504, "ts": 1716454223882476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882498, "dur": 1,
    "args": {
      "External id": 164585,
      "cbid": 251, "correlation": 164585
    }
  },
  {
    "ph": "f", "id": 164585, "pid": 76337, "tid": -914061504, "ts": 1716454223882498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882500, "dur": 0,
    "args": {
      "External id": 164586,
      "cbid": 251, "correlation": 164586
    }
  },
  {
    "ph": "f", "id": 164586, "pid": 76337, "tid": -914061504, "ts": 1716454223882500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882501, "dur": 0,
    "args": {
      "External id": 164587,
      "cbid": 251, "correlation": 164587
    }
  },
  {
    "ph": "f", "id": 164587, "pid": 76337, "tid": -914061504, "ts": 1716454223882501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882502, "dur": 0,
    "args": {
      "External id": 164588,
      "cbid": 251, "correlation": 164588
    }
  },
  {
    "ph": "f", "id": 164588, "pid": 76337, "tid": -914061504, "ts": 1716454223882502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882502, "dur": 0,
    "args": {
      "External id": 164589,
      "cbid": 251, "correlation": 164589
    }
  },
  {
    "ph": "f", "id": 164589, "pid": 76337, "tid": -914061504, "ts": 1716454223882502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882503, "dur": 0,
    "args": {
      "External id": 164590,
      "cbid": 251, "correlation": 164590
    }
  },
  {
    "ph": "f", "id": 164590, "pid": 76337, "tid": -914061504, "ts": 1716454223882503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882504, "dur": 0,
    "args": {
      "External id": 164591,
      "cbid": 251, "correlation": 164591
    }
  },
  {
    "ph": "f", "id": 164591, "pid": 76337, "tid": -914061504, "ts": 1716454223882504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882505, "dur": 0,
    "args": {
      "External id": 164592,
      "cbid": 251, "correlation": 164592
    }
  },
  {
    "ph": "f", "id": 164592, "pid": 76337, "tid": -914061504, "ts": 1716454223882505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882506, "dur": 0,
    "args": {
      "External id": 164593,
      "cbid": 251, "correlation": 164593
    }
  },
  {
    "ph": "f", "id": 164593, "pid": 76337, "tid": -914061504, "ts": 1716454223882506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224012293, "dur": 113,
    "args": {
      "External id": 164594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164594,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164594, "pid": 5, "tid": 7, "ts": 1716454224012293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882508, "dur": 12,
    "args": {
      "External id": 164594,
      "cbid": 211, "correlation": 164594
    }
  },
  {
    "ph": "s", "id": 164594, "pid": 76337, "tid": -914061504, "ts": 1716454223882508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224012407, "dur": 61,
    "args": {
      "External id": 164600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164600, "pid": 5, "tid": 7, "ts": 1716454224012407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882543, "dur": 9,
    "args": {
      "External id": 164600,
      "cbid": 211, "correlation": 164600
    }
  },
  {
    "ph": "s", "id": 164600, "pid": 76337, "tid": -914061504, "ts": 1716454223882543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224012469, "dur": 598,
    "args": {
      "External id": 164609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164609,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164609, "pid": 5, "tid": 7, "ts": 1716454224012469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882625, "dur": 14,
    "args": {
      "External id": 164609,
      "cbid": 211, "correlation": 164609
    }
  },
  {
    "ph": "s", "id": 164609, "pid": 76337, "tid": -914061504, "ts": 1716454223882625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224013068, "dur": 183,
    "args": {
      "External id": 164631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164631,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164631, "pid": 5, "tid": 7, "ts": 1716454224013068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882682, "dur": 10,
    "args": {
      "External id": 164631,
      "cbid": 211, "correlation": 164631
    }
  },
  {
    "ph": "s", "id": 164631, "pid": 76337, "tid": -914061504, "ts": 1716454223882682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882768, "dur": 1,
    "args": {
      "External id": 164642,
      "cbid": 251, "correlation": 164642
    }
  },
  {
    "ph": "f", "id": 164642, "pid": 76337, "tid": -914061504, "ts": 1716454223882768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224013252, "dur": 197,
    "args": {
      "External id": 164643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164643,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164643, "pid": 5, "tid": 7, "ts": 1716454224013252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882773, "dur": 14,
    "args": {
      "External id": 164643,
      "cbid": 211, "correlation": 164643
    }
  },
  {
    "ph": "s", "id": 164643, "pid": 76337, "tid": -914061504, "ts": 1716454223882773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882841, "dur": 1,
    "args": {
      "External id": 164654,
      "cbid": 251, "correlation": 164654
    }
  },
  {
    "ph": "f", "id": 164654, "pid": 76337, "tid": -914061504, "ts": 1716454223882841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224013450, "dur": 191,
    "args": {
      "External id": 164655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164655,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164655, "pid": 5, "tid": 7, "ts": 1716454224013450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882845, "dur": 11,
    "args": {
      "External id": 164655,
      "cbid": 211, "correlation": 164655
    }
  },
  {
    "ph": "s", "id": 164655, "pid": 76337, "tid": -914061504, "ts": 1716454223882845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882908, "dur": 1,
    "args": {
      "External id": 164666,
      "cbid": 251, "correlation": 164666
    }
  },
  {
    "ph": "f", "id": 164666, "pid": 76337, "tid": -914061504, "ts": 1716454223882908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224013643, "dur": 188,
    "args": {
      "External id": 164667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164667,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164667, "pid": 5, "tid": 7, "ts": 1716454224013643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223882912, "dur": 12,
    "args": {
      "External id": 164667,
      "cbid": 211, "correlation": 164667
    }
  },
  {
    "ph": "s", "id": 164667, "pid": 76337, "tid": -914061504, "ts": 1716454223882912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224013832, "dur": 18753,
    "args": {
      "External id": 164688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164688,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164688, "pid": 5, "tid": 7, "ts": 1716454224013832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223883001, "dur": 13,
    "args": {
      "External id": 164688,
      "cbid": 211, "correlation": 164688
    }
  },
  {
    "ph": "s", "id": 164688, "pid": 76337, "tid": -914061504, "ts": 1716454223883001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223883099, "dur": 1,
    "args": {
      "External id": 164706,
      "cbid": 251, "correlation": 164706
    }
  },
  {
    "ph": "f", "id": 164706, "pid": 76337, "tid": -914061504, "ts": 1716454223883099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224032586, "dur": 203,
    "args": {
      "External id": 164708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164708,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164708, "pid": 5, "tid": 7, "ts": 1716454224032586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223883105, "dur": 13,
    "args": {
      "External id": 164708,
      "cbid": 211, "correlation": 164708
    }
  },
  {
    "ph": "s", "id": 164708, "pid": 76337, "tid": -914061504, "ts": 1716454223883105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224032791, "dur": 67,
    "args": {
      "External id": 164716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164716,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164716, "pid": 5, "tid": 7, "ts": 1716454224032791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223883175, "dur": 12,
    "args": {
      "External id": 164716,
      "cbid": 211, "correlation": 164716
    }
  },
  {
    "ph": "s", "id": 164716, "pid": 76337, "tid": -914061504, "ts": 1716454223883175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224032859, "dur": 98,
    "args": {
      "External id": 164724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164724,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164724, "pid": 5, "tid": 7, "ts": 1716454224032859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223883214, "dur": 124,
    "args": {
      "External id": 164724,
      "cbid": 211, "correlation": 164724
    }
  },
  {
    "ph": "s", "id": 164724, "pid": 76337, "tid": -914061504, "ts": 1716454223883214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224032958, "dur": 55,
    "args": {
      "External id": 164735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164735, "pid": 5, "tid": 7, "ts": 1716454224032958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223883401, "dur": 1885,
    "args": {
      "External id": 164735,
      "cbid": 211, "correlation": 164735
    }
  },
  {
    "ph": "s", "id": 164735, "pid": 76337, "tid": -914061504, "ts": 1716454223883401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224033015, "dur": 93,
    "args": {
      "External id": 164757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164757,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164757, "pid": 5, "tid": 7, "ts": 1716454224033015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885306, "dur": 125,
    "args": {
      "External id": 164757,
      "cbid": 211, "correlation": 164757
    }
  },
  {
    "ph": "s", "id": 164757, "pid": 76337, "tid": -914061504, "ts": 1716454223885306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885509, "dur": 1,
    "args": {
      "External id": 164768,
      "cbid": 251, "correlation": 164768
    }
  },
  {
    "ph": "f", "id": 164768, "pid": 76337, "tid": -914061504, "ts": 1716454223885509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224033108, "dur": 104,
    "args": {
      "External id": 164769,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164769,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164769, "pid": 5, "tid": 7, "ts": 1716454224033108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885514, "dur": 13,
    "args": {
      "External id": 164769,
      "cbid": 211, "correlation": 164769
    }
  },
  {
    "ph": "s", "id": 164769, "pid": 76337, "tid": -914061504, "ts": 1716454223885514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885588, "dur": 1,
    "args": {
      "External id": 164780,
      "cbid": 251, "correlation": 164780
    }
  },
  {
    "ph": "f", "id": 164780, "pid": 76337, "tid": -914061504, "ts": 1716454223885588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885592, "dur": 0,
    "args": {
      "External id": 164781,
      "cbid": 251, "correlation": 164781
    }
  },
  {
    "ph": "f", "id": 164781, "pid": 76337, "tid": -914061504, "ts": 1716454223885592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224033214, "dur": 11,
    "args": {
      "External id": 164782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164782,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 164782, "pid": 5, "tid": 7, "ts": 1716454224033214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885593, "dur": 12,
    "args": {
      "External id": 164782,
      "cbid": 211, "correlation": 164782
    }
  },
  {
    "ph": "s", "id": 164782, "pid": 76337, "tid": -914061504, "ts": 1716454223885593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224033226, "dur": 5,
    "args": {
      "External id": 164784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164784,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 164784, "pid": 5, "tid": 7, "ts": 1716454224033226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885607, "dur": 6,
    "args": {
      "External id": 164784,
      "cbid": 211, "correlation": 164784
    }
  },
  {
    "ph": "s", "id": 164784, "pid": 76337, "tid": -914061504, "ts": 1716454223885607,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885668, "dur": 1,
    "args": {
      "External id": 164795,
      "cbid": 251, "correlation": 164795
    }
  },
  {
    "ph": "f", "id": 164795, "pid": 76337, "tid": -914061504, "ts": 1716454223885668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885671, "dur": 0,
    "args": {
      "External id": 164796,
      "cbid": 251, "correlation": 164796
    }
  },
  {
    "ph": "f", "id": 164796, "pid": 76337, "tid": -914061504, "ts": 1716454223885671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224033233, "dur": 6,
    "args": {
      "External id": 164797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164797,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 164797, "pid": 5, "tid": 7, "ts": 1716454224033233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885673, "dur": 12,
    "args": {
      "External id": 164797,
      "cbid": 211, "correlation": 164797
    }
  },
  {
    "ph": "s", "id": 164797, "pid": 76337, "tid": -914061504, "ts": 1716454223885673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224033241, "dur": 3,
    "args": {
      "External id": 164799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164799,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 164799, "pid": 5, "tid": 7, "ts": 1716454224033241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885687, "dur": 7,
    "args": {
      "External id": 164799,
      "cbid": 211, "correlation": 164799
    }
  },
  {
    "ph": "s", "id": 164799, "pid": 76337, "tid": -914061504, "ts": 1716454223885687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224033245, "dur": 157,
    "args": {
      "External id": 164820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164820,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164820, "pid": 5, "tid": 7, "ts": 1716454224033245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885761, "dur": 12,
    "args": {
      "External id": 164820,
      "cbid": 211, "correlation": 164820
    }
  },
  {
    "ph": "s", "id": 164820, "pid": 76337, "tid": -914061504, "ts": 1716454223885761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885857, "dur": 1,
    "args": {
      "External id": 164838,
      "cbid": 251, "correlation": 164838
    }
  },
  {
    "ph": "f", "id": 164838, "pid": 76337, "tid": -914061504, "ts": 1716454223885857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224033403, "dur": 109,
    "args": {
      "External id": 164840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164840,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164840, "pid": 5, "tid": 7, "ts": 1716454224033403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885863, "dur": 13,
    "args": {
      "External id": 164840,
      "cbid": 211, "correlation": 164840
    }
  },
  {
    "ph": "s", "id": 164840, "pid": 76337, "tid": -914061504, "ts": 1716454223885863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224033513, "dur": 35,
    "args": {
      "External id": 164848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164848,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164848, "pid": 5, "tid": 7, "ts": 1716454224033513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885932, "dur": 13,
    "args": {
      "External id": 164848,
      "cbid": 211, "correlation": 164848
    }
  },
  {
    "ph": "s", "id": 164848, "pid": 76337, "tid": -914061504, "ts": 1716454223885932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224033549, "dur": 67,
    "args": {
      "External id": 164856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164856,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164856, "pid": 5, "tid": 7, "ts": 1716454224033549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223885981, "dur": 11,
    "args": {
      "External id": 164856,
      "cbid": 211, "correlation": 164856
    }
  },
  {
    "ph": "s", "id": 164856, "pid": 76337, "tid": -914061504, "ts": 1716454223885981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224033618, "dur": 93,
    "args": {
      "External id": 164878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164878,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164878, "pid": 5, "tid": 7, "ts": 1716454224033618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886034, "dur": 11,
    "args": {
      "External id": 164878,
      "cbid": 211, "correlation": 164878
    }
  },
  {
    "ph": "s", "id": 164878, "pid": 76337, "tid": -914061504, "ts": 1716454223886034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886121, "dur": 1,
    "args": {
      "External id": 164894,
      "cbid": 251, "correlation": 164894
    }
  },
  {
    "ph": "f", "id": 164894, "pid": 76337, "tid": -914061504, "ts": 1716454223886121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224033712, "dur": 578,
    "args": {
      "External id": 164896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164896,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 164896, "pid": 5, "tid": 7, "ts": 1716454224033712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886127, "dur": 13,
    "args": {
      "External id": 164896,
      "cbid": 211, "correlation": 164896
    }
  },
  {
    "ph": "s", "id": 164896, "pid": 76337, "tid": -914061504, "ts": 1716454223886127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224034291, "dur": 243,
    "args": {
      "External id": 164904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164904,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164904, "pid": 5, "tid": 7, "ts": 1716454224034291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886192, "dur": 12,
    "args": {
      "External id": 164904,
      "cbid": 211, "correlation": 164904
    }
  },
  {
    "ph": "s", "id": 164904, "pid": 76337, "tid": -914061504, "ts": 1716454223886192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224034536, "dur": 254,
    "args": {
      "External id": 164912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164912,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164912, "pid": 5, "tid": 7, "ts": 1716454224034536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886222, "dur": 8,
    "args": {
      "External id": 164912,
      "cbid": 211, "correlation": 164912
    }
  },
  {
    "ph": "s", "id": 164912, "pid": 76337, "tid": -914061504, "ts": 1716454223886222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886304, "dur": 1,
    "args": {
      "External id": 164928,
      "cbid": 251, "correlation": 164928
    }
  },
  {
    "ph": "f", "id": 164928, "pid": 76337, "tid": -914061504, "ts": 1716454223886304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886309, "dur": 0,
    "args": {
      "External id": 164930,
      "cbid": 251, "correlation": 164930
    }
  },
  {
    "ph": "f", "id": 164930, "pid": 76337, "tid": -914061504, "ts": 1716454223886309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224034791, "dur": 361,
    "args": {
      "External id": 164931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164931,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 164931, "pid": 5, "tid": 7, "ts": 1716454224034791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886311, "dur": 12,
    "args": {
      "External id": 164931,
      "cbid": 211, "correlation": 164931
    }
  },
  {
    "ph": "s", "id": 164931, "pid": 76337, "tid": -914061504, "ts": 1716454223886311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224035153, "dur": 50,
    "args": {
      "External id": 164939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164939,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164939, "pid": 5, "tid": 7, "ts": 1716454224035153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886354, "dur": 185,
    "args": {
      "External id": 164939,
      "cbid": 211, "correlation": 164939
    }
  },
  {
    "ph": "s", "id": 164939, "pid": 76337, "tid": -914061504, "ts": 1716454223886354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224035204, "dur": 158,
    "args": {
      "External id": 164950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164950, "pid": 5, "tid": 7, "ts": 1716454224035204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886597, "dur": 71,
    "args": {
      "External id": 164950,
      "cbid": 211, "correlation": 164950
    }
  },
  {
    "ph": "s", "id": 164950, "pid": 76337, "tid": -914061504, "ts": 1716454223886597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886722, "dur": 0,
    "args": {
      "External id": 164962,
      "cbid": 317, "correlation": 164962
    }
  },
  {
    "ph": "f", "id": 164962, "pid": 76337, "tid": -914061504, "ts": 1716454223886722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886723, "dur": 0,
    "args": {
      "External id": 164963,
      "cbid": 203, "correlation": 164963
    }
  },
  {
    "ph": "f", "id": 164963, "pid": 76337, "tid": -914061504, "ts": 1716454223886723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886724, "dur": 0,
    "args": {
      "External id": 164964,
      "cbid": 205, "correlation": 164964
    }
  },
  {
    "ph": "f", "id": 164964, "pid": 76337, "tid": -914061504, "ts": 1716454223886724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886747, "dur": 1,
    "args": {
      "External id": 164968,
      "cbid": 251, "correlation": 164968
    }
  },
  {
    "ph": "f", "id": 164968, "pid": 76337, "tid": -914061504, "ts": 1716454223886747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886749, "dur": 0,
    "args": {
      "External id": 164969,
      "cbid": 251, "correlation": 164969
    }
  },
  {
    "ph": "f", "id": 164969, "pid": 76337, "tid": -914061504, "ts": 1716454223886749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886750, "dur": 0,
    "args": {
      "External id": 164970,
      "cbid": 251, "correlation": 164970
    }
  },
  {
    "ph": "f", "id": 164970, "pid": 76337, "tid": -914061504, "ts": 1716454223886750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886750, "dur": 0,
    "args": {
      "External id": 164971,
      "cbid": 251, "correlation": 164971
    }
  },
  {
    "ph": "f", "id": 164971, "pid": 76337, "tid": -914061504, "ts": 1716454223886750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886751, "dur": 0,
    "args": {
      "External id": 164972,
      "cbid": 251, "correlation": 164972
    }
  },
  {
    "ph": "f", "id": 164972, "pid": 76337, "tid": -914061504, "ts": 1716454223886751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886752, "dur": 0,
    "args": {
      "External id": 164973,
      "cbid": 251, "correlation": 164973
    }
  },
  {
    "ph": "f", "id": 164973, "pid": 76337, "tid": -914061504, "ts": 1716454223886752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886753, "dur": 0,
    "args": {
      "External id": 164974,
      "cbid": 251, "correlation": 164974
    }
  },
  {
    "ph": "f", "id": 164974, "pid": 76337, "tid": -914061504, "ts": 1716454223886753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886753, "dur": 0,
    "args": {
      "External id": 164975,
      "cbid": 251, "correlation": 164975
    }
  },
  {
    "ph": "f", "id": 164975, "pid": 76337, "tid": -914061504, "ts": 1716454223886753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886754, "dur": 0,
    "args": {
      "External id": 164976,
      "cbid": 251, "correlation": 164976
    }
  },
  {
    "ph": "f", "id": 164976, "pid": 76337, "tid": -914061504, "ts": 1716454223886754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224035364, "dur": 116,
    "args": {
      "External id": 164977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164977,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 164977, "pid": 5, "tid": 7, "ts": 1716454224035364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886757, "dur": 38,
    "args": {
      "External id": 164977,
      "cbid": 211, "correlation": 164977
    }
  },
  {
    "ph": "s", "id": 164977, "pid": 76337, "tid": -914061504, "ts": 1716454223886757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224035481, "dur": 59,
    "args": {
      "External id": 164983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164983,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164983, "pid": 5, "tid": 7, "ts": 1716454224035481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223886817, "dur": 283,
    "args": {
      "External id": 164983,
      "cbid": 211, "correlation": 164983
    }
  },
  {
    "ph": "s", "id": 164983, "pid": 76337, "tid": -914061504, "ts": 1716454223886817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224035542, "dur": 49,
    "args": {
      "External id": 164991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 164991,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 164991, "pid": 5, "tid": 7, "ts": 1716454224035542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887124, "dur": 9,
    "args": {
      "External id": 164991,
      "cbid": 211, "correlation": 164991
    }
  },
  {
    "ph": "s", "id": 164991, "pid": 76337, "tid": -914061504, "ts": 1716454223887124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224035593, "dur": 51,
    "args": {
      "External id": 165011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165011,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 165011, "pid": 5, "tid": 7, "ts": 1716454224035593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887206, "dur": 11,
    "args": {
      "External id": 165011,
      "cbid": 211, "correlation": 165011
    }
  },
  {
    "ph": "s", "id": 165011, "pid": 76337, "tid": -914061504, "ts": 1716454223887206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224035645, "dur": 4,
    "args": {
      "External id": 165023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165023,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 165023, "pid": 5, "tid": 7, "ts": 1716454224035645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887227, "dur": 6,
    "args": {
      "External id": 165023,
      "cbid": 211, "correlation": 165023
    }
  },
  {
    "ph": "s", "id": 165023, "pid": 76337, "tid": -914061504, "ts": 1716454223887227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224035651, "dur": 56,
    "args": {
      "External id": 165026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165026,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165026, "pid": 5, "tid": 7, "ts": 1716454224035651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887245, "dur": 111,
    "args": {
      "External id": 165026,
      "cbid": 211, "correlation": 165026
    }
  },
  {
    "ph": "s", "id": 165026, "pid": 76337, "tid": -914061504, "ts": 1716454223887245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224035708, "dur": 37,
    "args": {
      "External id": 165035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165035,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165035, "pid": 5, "tid": 7, "ts": 1716454224035708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887395, "dur": 11,
    "args": {
      "External id": 165035,
      "cbid": 211, "correlation": 165035
    }
  },
  {
    "ph": "s", "id": 165035, "pid": 76337, "tid": -914061504, "ts": 1716454223887395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887452, "dur": 0,
    "args": {
      "External id": 165045,
      "cbid": 317, "correlation": 165045
    }
  },
  {
    "ph": "f", "id": 165045, "pid": 76337, "tid": -914061504, "ts": 1716454223887452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887453, "dur": 0,
    "args": {
      "External id": 165046,
      "cbid": 203, "correlation": 165046
    }
  },
  {
    "ph": "f", "id": 165046, "pid": 76337, "tid": -914061504, "ts": 1716454223887453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887454, "dur": 0,
    "args": {
      "External id": 165047,
      "cbid": 205, "correlation": 165047
    }
  },
  {
    "ph": "f", "id": 165047, "pid": 76337, "tid": -914061504, "ts": 1716454223887454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224035747, "dur": 41,
    "args": {
      "External id": 165051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165051,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165051, "pid": 5, "tid": 7, "ts": 1716454224035747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887470, "dur": 12,
    "args": {
      "External id": 165051,
      "cbid": 211, "correlation": 165051
    }
  },
  {
    "ph": "s", "id": 165051, "pid": 76337, "tid": -914061504, "ts": 1716454223887470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224035789, "dur": 3,
    "args": {
      "External id": 165053,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165053,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.80327868852459
    }
  },
  {
    "ph": "f", "id": 165053, "pid": 5, "tid": 7, "ts": 1716454224035789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887486, "dur": 20,
    "args": {
      "External id": 165053,
      "cbid": 51, "correlation": 165053
    }
  },
  {
    "ph": "s", "id": 165053, "pid": 76337, "tid": -914061504, "ts": 1716454223887486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887511, "dur": 2,
    "args": {
      "External id": 165055,
      "cbid": 200, "correlation": 165055
    }
  },
  {
    "ph": "f", "id": 165055, "pid": 76337, "tid": -914061504, "ts": 1716454223887511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887514, "dur": 0,
    "args": {
      "External id": 165056,
      "cbid": 200, "correlation": 165056
    }
  },
  {
    "ph": "f", "id": 165056, "pid": 76337, "tid": -914061504, "ts": 1716454223887514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887515, "dur": 0,
    "args": {
      "External id": 165057,
      "cbid": 200, "correlation": 165057
    }
  },
  {
    "ph": "f", "id": 165057, "pid": 76337, "tid": -914061504, "ts": 1716454223887515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887515, "dur": 0,
    "args": {
      "External id": 165058,
      "cbid": 200, "correlation": 165058
    }
  },
  {
    "ph": "f", "id": 165058, "pid": 76337, "tid": -914061504, "ts": 1716454223887515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887516, "dur": 4,
    "args": {
      "External id": 165059,
      "cbid": 15, "correlation": 165059
    }
  },
  {
    "ph": "f", "id": 165059, "pid": 76337, "tid": -914061504, "ts": 1716454223887516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887521, "dur": 1,
    "args": {
      "External id": 165060,
      "cbid": 251, "correlation": 165060
    }
  },
  {
    "ph": "f", "id": 165060, "pid": 76337, "tid": -914061504, "ts": 1716454223887521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224035794, "dur": 24,
    "args": {
      "External id": 165061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165061,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165061, "pid": 5, "tid": 7, "ts": 1716454224035794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887524, "dur": 8,
    "args": {
      "External id": 165061,
      "cbid": 211, "correlation": 165061
    }
  },
  {
    "ph": "s", "id": 165061, "pid": 76337, "tid": -914061504, "ts": 1716454223887524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224035819, "dur": 4,
    "args": {
      "External id": 165063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165063,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 165063, "pid": 5, "tid": 7, "ts": 1716454224035819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887537, "dur": 6,
    "args": {
      "External id": 165063,
      "cbid": 211, "correlation": 165063
    }
  },
  {
    "ph": "s", "id": 165063, "pid": 76337, "tid": -914061504, "ts": 1716454223887537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887546, "dur": 0,
    "args": {
      "External id": 165064,
      "cbid": 51, "correlation": 165064
    }
  },
  {
    "ph": "s", "id": 165064, "pid": 76337, "tid": -914061504, "ts": 1716454223887546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224035824, "dur": 190,
    "args": {
      "External id": 165065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165065,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165065, "pid": 5, "tid": 7, "ts": 1716454224035824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887547, "dur": 202,
    "args": {
      "External id": 165065,
      "cbid": 211, "correlation": 165065
    }
  },
  {
    "ph": "s", "id": 165065, "pid": 76337, "tid": -914061504, "ts": 1716454223887547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224036015, "dur": 6,
    "args": {
      "External id": 165066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165066,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165066, "pid": 5, "tid": 7, "ts": 1716454224036015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887753, "dur": 6,
    "args": {
      "External id": 165066,
      "cbid": 211, "correlation": 165066
    }
  },
  {
    "ph": "s", "id": 165066, "pid": 76337, "tid": -914061504, "ts": 1716454223887753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224036023, "dur": 5,
    "args": {
      "External id": 165072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 165072, "pid": 5, "tid": 7, "ts": 1716454224036023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223887784, "dur": 9,
    "args": {
      "External id": 165072,
      "cbid": 211, "correlation": 165072
    }
  },
  {
    "ph": "s", "id": 165072, "pid": 76337, "tid": -914061504, "ts": 1716454223887784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224036029, "dur": 3,
    "args": {
      "External id": 165080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165080,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165080, "pid": 5, "tid": 7, "ts": 1716454224036029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223889484, "dur": 15,
    "args": {
      "External id": 165080,
      "cbid": 211, "correlation": 165080
    }
  },
  {
    "ph": "s", "id": 165080, "pid": 76337, "tid": -914061504, "ts": 1716454223889484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224036033, "dur": 3,
    "args": {
      "External id": 165088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165088,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165088, "pid": 5, "tid": 7, "ts": 1716454224036033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223889524, "dur": 11,
    "args": {
      "External id": 165088,
      "cbid": 211, "correlation": 165088
    }
  },
  {
    "ph": "s", "id": 165088, "pid": 76337, "tid": -914061504, "ts": 1716454223889524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224036038, "dur": 3,
    "args": {
      "External id": 165096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165096,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165096, "pid": 5, "tid": 7, "ts": 1716454224036038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223889551, "dur": 8,
    "args": {
      "External id": 165096,
      "cbid": 211, "correlation": 165096
    }
  },
  {
    "ph": "s", "id": 165096, "pid": 76337, "tid": -914061504, "ts": 1716454223889551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224036042, "dur": 3,
    "args": {
      "External id": 165105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165105,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165105, "pid": 5, "tid": 7, "ts": 1716454224036042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223889730, "dur": 14,
    "args": {
      "External id": 165105,
      "cbid": 211, "correlation": 165105
    }
  },
  {
    "ph": "s", "id": 165105, "pid": 76337, "tid": -914061504, "ts": 1716454223889730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224036047, "dur": 3,
    "args": {
      "External id": 165114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165114,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165114, "pid": 5, "tid": 7, "ts": 1716454224036047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223889760, "dur": 7,
    "args": {
      "External id": 165114,
      "cbid": 211, "correlation": 165114
    }
  },
  {
    "ph": "s", "id": 165114, "pid": 76337, "tid": -914061504, "ts": 1716454223889760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224036051, "dur": 3,
    "args": {
      "External id": 165122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165122,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165122, "pid": 5, "tid": 7, "ts": 1716454224036051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223889784, "dur": 8,
    "args": {
      "External id": 165122,
      "cbid": 211, "correlation": 165122
    }
  },
  {
    "ph": "s", "id": 165122, "pid": 76337, "tid": -914061504, "ts": 1716454223889784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224036055, "dur": 3,
    "args": {
      "External id": 165130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165130,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165130, "pid": 5, "tid": 7, "ts": 1716454224036055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223890055, "dur": 16,
    "args": {
      "External id": 165130,
      "cbid": 211, "correlation": 165130
    }
  },
  {
    "ph": "s", "id": 165130, "pid": 76337, "tid": -914061504, "ts": 1716454223890055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224036060, "dur": 3,
    "args": {
      "External id": 165138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165138,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165138, "pid": 5, "tid": 7, "ts": 1716454224036060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454223890087, "dur": 7,
    "args": {
      "External id": 165138,
      "cbid": 211, "correlation": 165138
    }
  },
  {
    "ph": "s", "id": 165138, "pid": 76337, "tid": -914061504, "ts": 1716454223890087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036065, "dur": 1,
    "args": {
      "External id": 165148,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165148,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 165148, "pid": 5, "tid": 7, "ts": 1716454224036065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454223890154, "dur": 35,
    "args": {
      "External id": 165148,
      "cbid": 41, "correlation": 165148
    }
  },
  {
    "ph": "s", "id": 165148, "pid": 76337, "tid": -914061504, "ts": 1716454223890154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454223890191, "dur": 145891,
    "args": {
      "External id": 165149,
      "cbid": 131, "correlation": 165149
    }
  },
  {
    "ph": "f", "id": 165149, "pid": 76337, "tid": -914061504, "ts": 1716454223890191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224036228, "dur": 3,
    "args": {
      "External id": 165157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165157,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165157, "pid": 5, "tid": 7, "ts": 1716454224036228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036206, "dur": 23,
    "args": {
      "External id": 165157,
      "cbid": 211, "correlation": 165157
    }
  },
  {
    "ph": "s", "id": 165157, "pid": 76337, "tid": -914061504, "ts": 1716454224036206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036317, "dur": 3,
    "args": {
      "External id": 165166,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165166,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002380952380952381
    }
  },
  {
    "ph": "f", "id": 165166, "pid": 5, "tid": 7, "ts": 1716454224036317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036288, "dur": 28,
    "args": {
      "External id": 165166,
      "cbid": 41, "correlation": 165166
    }
  },
  {
    "ph": "s", "id": 165166, "pid": 76337, "tid": -914061504, "ts": 1716454224036288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224036406, "dur": 4,
    "args": {
      "External id": 165176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165176,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165176, "pid": 5, "tid": 7, "ts": 1716454224036406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036390, "dur": 16,
    "args": {
      "External id": 165176,
      "cbid": 211, "correlation": 165176
    }
  },
  {
    "ph": "s", "id": 165176, "pid": 76337, "tid": -914061504, "ts": 1716454224036390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036509, "dur": 1,
    "args": {
      "External id": 165186,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165186,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004545454545454545
    }
  },
  {
    "ph": "f", "id": 165186, "pid": 5, "tid": 7, "ts": 1716454224036509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036491, "dur": 16,
    "args": {
      "External id": 165186,
      "cbid": 41, "correlation": 165186
    }
  },
  {
    "ph": "s", "id": 165186, "pid": 76337, "tid": -914061504, "ts": 1716454224036491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036508, "dur": 9,
    "args": {
      "External id": 165187,
      "cbid": 131, "correlation": 165187
    }
  },
  {
    "ph": "f", "id": 165187, "pid": 76337, "tid": -914061504, "ts": 1716454224036508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036578, "dur": 3,
    "args": {
      "External id": 165194,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165194,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.415841584158414
    }
  },
  {
    "ph": "f", "id": 165194, "pid": 5, "tid": 7, "ts": 1716454224036578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036555, "dur": 23,
    "args": {
      "External id": 165194,
      "cbid": 41, "correlation": 165194
    }
  },
  {
    "ph": "s", "id": 165194, "pid": 76337, "tid": -914061504, "ts": 1716454224036555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036663, "dur": 3,
    "args": {
      "External id": 165213,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165213,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 165213, "pid": 5, "tid": 7, "ts": 1716454224036663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036644, "dur": 18,
    "args": {
      "External id": 165213,
      "cbid": 41, "correlation": 165213
    }
  },
  {
    "ph": "s", "id": 165213, "pid": 76337, "tid": -914061504, "ts": 1716454224036644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224036702, "dur": 3,
    "args": {
      "External id": 165219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165219, "pid": 5, "tid": 7, "ts": 1716454224036702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036690, "dur": 12,
    "args": {
      "External id": 165219,
      "cbid": 211, "correlation": 165219
    }
  },
  {
    "ph": "s", "id": 165219, "pid": 76337, "tid": -914061504, "ts": 1716454224036690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224036716, "dur": 6,
    "args": {
      "External id": 165221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165221,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 165221, "pid": 5, "tid": 7, "ts": 1716454224036716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036706, "dur": 9,
    "args": {
      "External id": 165221,
      "cbid": 211, "correlation": 165221
    }
  },
  {
    "ph": "s", "id": 165221, "pid": 76337, "tid": -914061504, "ts": 1716454224036706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454224036724, "dur": 3,
    "args": {
      "External id": 165223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165223,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165223, "pid": 5, "tid": 7, "ts": 1716454224036724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036716, "dur": 6,
    "args": {
      "External id": 165223,
      "cbid": 211, "correlation": 165223
    }
  },
  {
    "ph": "s", "id": 165223, "pid": 76337, "tid": -914061504, "ts": 1716454224036716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036757, "dur": 2,
    "args": {
      "External id": 165231,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165231,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 165231, "pid": 5, "tid": 7, "ts": 1716454224036757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036743, "dur": 13,
    "args": {
      "External id": 165231,
      "cbid": 41, "correlation": 165231
    }
  },
  {
    "ph": "s", "id": 165231, "pid": 76337, "tid": -914061504, "ts": 1716454224036743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224036802, "dur": 3,
    "args": {
      "External id": 165245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165245,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165245, "pid": 5, "tid": 7, "ts": 1716454224036802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036792, "dur": 12,
    "args": {
      "External id": 165245,
      "cbid": 211, "correlation": 165245
    }
  },
  {
    "ph": "s", "id": 165245, "pid": 76337, "tid": -914061504, "ts": 1716454224036792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224036825, "dur": 2,
    "args": {
      "External id": 165259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165259,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165259, "pid": 5, "tid": 7, "ts": 1716454224036825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036817, "dur": 6,
    "args": {
      "External id": 165259,
      "cbid": 211, "correlation": 165259
    }
  },
  {
    "ph": "s", "id": 165259, "pid": 76337, "tid": -914061504, "ts": 1716454224036817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224036874, "dur": 6,
    "args": {
      "External id": 165266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165266,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165266, "pid": 5, "tid": 7, "ts": 1716454224036874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036863, "dur": 12,
    "args": {
      "External id": 165266,
      "cbid": 211, "correlation": 165266
    }
  },
  {
    "ph": "s", "id": 165266, "pid": 76337, "tid": -914061504, "ts": 1716454224036863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224036885, "dur": 6,
    "args": {
      "External id": 165269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165269,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165269, "pid": 5, "tid": 7, "ts": 1716454224036885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036878, "dur": 7,
    "args": {
      "External id": 165269,
      "cbid": 211, "correlation": 165269
    }
  },
  {
    "ph": "s", "id": 165269, "pid": 76337, "tid": -914061504, "ts": 1716454224036878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224036894, "dur": 3,
    "args": {
      "External id": 165271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165271,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165271, "pid": 5, "tid": 7, "ts": 1716454224036894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036886, "dur": 7,
    "args": {
      "External id": 165271,
      "cbid": 211, "correlation": 165271
    }
  },
  {
    "ph": "s", "id": 165271, "pid": 76337, "tid": -914061504, "ts": 1716454224036886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036918, "dur": 2,
    "args": {
      "External id": 165274,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165274,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002777777777777778
    }
  },
  {
    "ph": "f", "id": 165274, "pid": 5, "tid": 7, "ts": 1716454224036918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036904, "dur": 13,
    "args": {
      "External id": 165274,
      "cbid": 41, "correlation": 165274
    }
  },
  {
    "ph": "s", "id": 165274, "pid": 76337, "tid": -914061504, "ts": 1716454224036904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224036969, "dur": 4,
    "args": {
      "External id": 165290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165290,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165290, "pid": 5, "tid": 7, "ts": 1716454224036969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036957, "dur": 12,
    "args": {
      "External id": 165290,
      "cbid": 211, "correlation": 165290
    }
  },
  {
    "ph": "s", "id": 165290, "pid": 76337, "tid": -914061504, "ts": 1716454224036957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224036997, "dur": 3,
    "args": {
      "External id": 165295,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165295,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003125
    }
  },
  {
    "ph": "f", "id": 165295, "pid": 5, "tid": 7, "ts": 1716454224036997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224036982, "dur": 14,
    "args": {
      "External id": 165295,
      "cbid": 41, "correlation": 165295
    }
  },
  {
    "ph": "s", "id": 165295, "pid": 76337, "tid": -914061504, "ts": 1716454224036982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224037023, "dur": 1,
    "args": {
      "External id": 165301,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165301,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 165301, "pid": 5, "tid": 7, "ts": 1716454224037023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037006, "dur": 26,
    "args": {
      "External id": 165301,
      "cbid": 41, "correlation": 165301
    }
  },
  {
    "ph": "s", "id": 165301, "pid": 76337, "tid": -914061504, "ts": 1716454224037006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037032, "dur": 3,
    "args": {
      "External id": 165302,
      "cbid": 131, "correlation": 165302
    }
  },
  {
    "ph": "f", "id": 165302, "pid": 76337, "tid": -914061504, "ts": 1716454224037032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224037082, "dur": 3,
    "args": {
      "External id": 165310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165310,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165310, "pid": 5, "tid": 7, "ts": 1716454224037082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037069, "dur": 12,
    "args": {
      "External id": 165310,
      "cbid": 211, "correlation": 165310
    }
  },
  {
    "ph": "s", "id": 165310, "pid": 76337, "tid": -914061504, "ts": 1716454224037069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224037113, "dur": 3,
    "args": {
      "External id": 165320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165320,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165320, "pid": 5, "tid": 7, "ts": 1716454224037113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037103, "dur": 9,
    "args": {
      "External id": 165320,
      "cbid": 211, "correlation": 165320
    }
  },
  {
    "ph": "s", "id": 165320, "pid": 76337, "tid": -914061504, "ts": 1716454224037103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224037136, "dur": 3,
    "args": {
      "External id": 165329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165329,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165329, "pid": 5, "tid": 7, "ts": 1716454224037136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037127, "dur": 7,
    "args": {
      "External id": 165329,
      "cbid": 211, "correlation": 165329
    }
  },
  {
    "ph": "s", "id": 165329, "pid": 76337, "tid": -914061504, "ts": 1716454224037127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224037244, "dur": 12,
    "args": {
      "External id": 165339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165339,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165339, "pid": 5, "tid": 7, "ts": 1716454224037244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037227, "dur": 19,
    "args": {
      "External id": 165339,
      "cbid": 211, "correlation": 165339
    }
  },
  {
    "ph": "s", "id": 165339, "pid": 76337, "tid": -914061504, "ts": 1716454224037227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224037283, "dur": 3,
    "args": {
      "External id": 165347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165347,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165347, "pid": 5, "tid": 7, "ts": 1716454224037283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037274, "dur": 8,
    "args": {
      "External id": 165347,
      "cbid": 211, "correlation": 165347
    }
  },
  {
    "ph": "s", "id": 165347, "pid": 76337, "tid": -914061504, "ts": 1716454224037274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224037327, "dur": 12,
    "args": {
      "External id": 165357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165357,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165357, "pid": 5, "tid": 7, "ts": 1716454224037327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037315, "dur": 11,
    "args": {
      "External id": 165357,
      "cbid": 211, "correlation": 165357
    }
  },
  {
    "ph": "s", "id": 165357, "pid": 76337, "tid": -914061504, "ts": 1716454224037315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224037356, "dur": 10,
    "args": {
      "External id": 165365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165365,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165365, "pid": 5, "tid": 7, "ts": 1716454224037356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037346, "dur": 8,
    "args": {
      "External id": 165365,
      "cbid": 211, "correlation": 165365
    }
  },
  {
    "ph": "s", "id": 165365, "pid": 76337, "tid": -914061504, "ts": 1716454224037346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224037382, "dur": 3,
    "args": {
      "External id": 165374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165374, "pid": 5, "tid": 7, "ts": 1716454224037382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037373, "dur": 9,
    "args": {
      "External id": 165374,
      "cbid": 211, "correlation": 165374
    }
  },
  {
    "ph": "s", "id": 165374, "pid": 76337, "tid": -914061504, "ts": 1716454224037373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224037406, "dur": 5,
    "args": {
      "External id": 165383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165383,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165383, "pid": 5, "tid": 7, "ts": 1716454224037406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037397, "dur": 7,
    "args": {
      "External id": 165383,
      "cbid": 211, "correlation": 165383
    }
  },
  {
    "ph": "s", "id": 165383, "pid": 76337, "tid": -914061504, "ts": 1716454224037397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224037442, "dur": 8,
    "args": {
      "External id": 165393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165393,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165393, "pid": 5, "tid": 7, "ts": 1716454224037442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037432, "dur": 11,
    "args": {
      "External id": 165393,
      "cbid": 211, "correlation": 165393
    }
  },
  {
    "ph": "s", "id": 165393, "pid": 76337, "tid": -914061504, "ts": 1716454224037432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224037754, "dur": 3,
    "args": {
      "External id": 165402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165402,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165402, "pid": 5, "tid": 7, "ts": 1716454224037754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037738, "dur": 16,
    "args": {
      "External id": 165402,
      "cbid": 211, "correlation": 165402
    }
  },
  {
    "ph": "s", "id": 165402, "pid": 76337, "tid": -914061504, "ts": 1716454224037738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224037780, "dur": 3,
    "args": {
      "External id": 165410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165410,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165410, "pid": 5, "tid": 7, "ts": 1716454224037780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037771, "dur": 8,
    "args": {
      "External id": 165410,
      "cbid": 211, "correlation": 165410
    }
  },
  {
    "ph": "s", "id": 165410, "pid": 76337, "tid": -914061504, "ts": 1716454224037771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224037832, "dur": 1,
    "args": {
      "External id": 165420,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165420,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0053226879574184965
    }
  },
  {
    "ph": "f", "id": 165420, "pid": 5, "tid": 7, "ts": 1716454224037832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037817, "dur": 13,
    "args": {
      "External id": 165420,
      "cbid": 41, "correlation": 165420
    }
  },
  {
    "ph": "s", "id": 165420, "pid": 76337, "tid": -914061504, "ts": 1716454224037817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037831, "dur": 8,
    "args": {
      "External id": 165421,
      "cbid": 131, "correlation": 165421
    }
  },
  {
    "ph": "f", "id": 165421, "pid": 76337, "tid": -914061504, "ts": 1716454224037831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224037921, "dur": 2,
    "args": {
      "External id": 165429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165429, "pid": 5, "tid": 7, "ts": 1716454224037921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037907, "dur": 14,
    "args": {
      "External id": 165429,
      "cbid": 211, "correlation": 165429
    }
  },
  {
    "ph": "s", "id": 165429, "pid": 76337, "tid": -914061504, "ts": 1716454224037907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038003, "dur": 3,
    "args": {
      "External id": 165438,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165438,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0026041666666666665
    }
  },
  {
    "ph": "f", "id": 165438, "pid": 5, "tid": 7, "ts": 1716454224038003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224037984, "dur": 19,
    "args": {
      "External id": 165438,
      "cbid": 41, "correlation": 165438
    }
  },
  {
    "ph": "s", "id": 165438, "pid": 76337, "tid": -914061504, "ts": 1716454224037984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224038075, "dur": 3,
    "args": {
      "External id": 165448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165448,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165448, "pid": 5, "tid": 7, "ts": 1716454224038075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038061, "dur": 13,
    "args": {
      "External id": 165448,
      "cbid": 211, "correlation": 165448
    }
  },
  {
    "ph": "s", "id": 165448, "pid": 76337, "tid": -914061504, "ts": 1716454224038061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038128, "dur": 1,
    "args": {
      "External id": 165458,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165458,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 165458, "pid": 5, "tid": 7, "ts": 1716454224038128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038113, "dur": 13,
    "args": {
      "External id": 165458,
      "cbid": 41, "correlation": 165458
    }
  },
  {
    "ph": "s", "id": 165458, "pid": 76337, "tid": -914061504, "ts": 1716454224038113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038127, "dur": 8,
    "args": {
      "External id": 165459,
      "cbid": 131, "correlation": 165459
    }
  },
  {
    "ph": "f", "id": 165459, "pid": 76337, "tid": -914061504, "ts": 1716454224038127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038188, "dur": 3,
    "args": {
      "External id": 165466,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165466,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 165466, "pid": 5, "tid": 7, "ts": 1716454224038188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038169, "dur": 18,
    "args": {
      "External id": 165466,
      "cbid": 41, "correlation": 165466
    }
  },
  {
    "ph": "s", "id": 165466, "pid": 76337, "tid": -914061504, "ts": 1716454224038169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038235, "dur": 1,
    "args": {
      "External id": 165477,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165477,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 165477, "pid": 5, "tid": 7, "ts": 1716454224038235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038223, "dur": 10,
    "args": {
      "External id": 165477,
      "cbid": 41, "correlation": 165477
    }
  },
  {
    "ph": "s", "id": 165477, "pid": 76337, "tid": -914061504, "ts": 1716454224038223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038234, "dur": 8,
    "args": {
      "External id": 165478,
      "cbid": 131, "correlation": 165478
    }
  },
  {
    "ph": "f", "id": 165478, "pid": 76337, "tid": -914061504, "ts": 1716454224038234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224038283, "dur": 3,
    "args": {
      "External id": 165486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165486,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165486, "pid": 5, "tid": 7, "ts": 1716454224038283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038270, "dur": 13,
    "args": {
      "External id": 165486,
      "cbid": 211, "correlation": 165486
    }
  },
  {
    "ph": "s", "id": 165486, "pid": 76337, "tid": -914061504, "ts": 1716454224038270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224038313, "dur": 3,
    "args": {
      "External id": 165496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165496,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165496, "pid": 5, "tid": 7, "ts": 1716454224038313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038303, "dur": 8,
    "args": {
      "External id": 165496,
      "cbid": 211, "correlation": 165496
    }
  },
  {
    "ph": "s", "id": 165496, "pid": 76337, "tid": -914061504, "ts": 1716454224038303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224038334, "dur": 3,
    "args": {
      "External id": 165505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165505,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165505, "pid": 5, "tid": 7, "ts": 1716454224038334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038326, "dur": 7,
    "args": {
      "External id": 165505,
      "cbid": 211, "correlation": 165505
    }
  },
  {
    "ph": "s", "id": 165505, "pid": 76337, "tid": -914061504, "ts": 1716454224038326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224038402, "dur": 5,
    "args": {
      "External id": 165513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165513,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165513, "pid": 5, "tid": 7, "ts": 1716454224038402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038388, "dur": 14,
    "args": {
      "External id": 165513,
      "cbid": 211, "correlation": 165513
    }
  },
  {
    "ph": "s", "id": 165513, "pid": 76337, "tid": -914061504, "ts": 1716454224038388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224038442, "dur": 3,
    "args": {
      "External id": 165522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165522,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165522, "pid": 5, "tid": 7, "ts": 1716454224038442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038432, "dur": 9,
    "args": {
      "External id": 165522,
      "cbid": 211, "correlation": 165522
    }
  },
  {
    "ph": "s", "id": 165522, "pid": 76337, "tid": -914061504, "ts": 1716454224038432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224038465, "dur": 3,
    "args": {
      "External id": 165531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165531,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165531, "pid": 5, "tid": 7, "ts": 1716454224038465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038457, "dur": 7,
    "args": {
      "External id": 165531,
      "cbid": 211, "correlation": 165531
    }
  },
  {
    "ph": "s", "id": 165531, "pid": 76337, "tid": -914061504, "ts": 1716454224038457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224038527, "dur": 3,
    "args": {
      "External id": 165539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165539,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165539, "pid": 5, "tid": 7, "ts": 1716454224038527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038516, "dur": 10,
    "args": {
      "External id": 165539,
      "cbid": 211, "correlation": 165539
    }
  },
  {
    "ph": "s", "id": 165539, "pid": 76337, "tid": -914061504, "ts": 1716454224038516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224038585, "dur": 1,
    "args": {
      "External id": 165547,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165547,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 165547, "pid": 5, "tid": 7, "ts": 1716454224038585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038569, "dur": 26,
    "args": {
      "External id": 165547,
      "cbid": 41, "correlation": 165547
    }
  },
  {
    "ph": "s", "id": 165547, "pid": 76337, "tid": -914061504, "ts": 1716454224038569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038596, "dur": 4,
    "args": {
      "External id": 165548,
      "cbid": 131, "correlation": 165548
    }
  },
  {
    "ph": "f", "id": 165548, "pid": 76337, "tid": -914061504, "ts": 1716454224038596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038658, "dur": 1,
    "args": {
      "External id": 165558,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165558,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 165558, "pid": 5, "tid": 7, "ts": 1716454224038658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038645, "dur": 10,
    "args": {
      "External id": 165558,
      "cbid": 41, "correlation": 165558
    }
  },
  {
    "ph": "s", "id": 165558, "pid": 76337, "tid": -914061504, "ts": 1716454224038645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038656, "dur": 7,
    "args": {
      "External id": 165559,
      "cbid": 131, "correlation": 165559
    }
  },
  {
    "ph": "f", "id": 165559, "pid": 76337, "tid": -914061504, "ts": 1716454224038656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038710, "dur": 1,
    "args": {
      "External id": 165568,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165568,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 165568, "pid": 5, "tid": 7, "ts": 1716454224038710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038700, "dur": 8,
    "args": {
      "External id": 165568,
      "cbid": 41, "correlation": 165568
    }
  },
  {
    "ph": "s", "id": 165568, "pid": 76337, "tid": -914061504, "ts": 1716454224038700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038709, "dur": 8,
    "args": {
      "External id": 165569,
      "cbid": 131, "correlation": 165569
    }
  },
  {
    "ph": "f", "id": 165569, "pid": 76337, "tid": -914061504, "ts": 1716454224038709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224038783, "dur": 4,
    "args": {
      "External id": 165576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165576, "pid": 5, "tid": 7, "ts": 1716454224038783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038767, "dur": 17,
    "args": {
      "External id": 165576,
      "cbid": 211, "correlation": 165576
    }
  },
  {
    "ph": "s", "id": 165576, "pid": 76337, "tid": -914061504, "ts": 1716454224038767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454224038821, "dur": 4,
    "args": {
      "External id": 165596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165596,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165596, "pid": 5, "tid": 7, "ts": 1716454224038821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038809, "dur": 12,
    "args": {
      "External id": 165596,
      "cbid": 211, "correlation": 165596
    }
  },
  {
    "ph": "s", "id": 165596, "pid": 76337, "tid": -914061504, "ts": 1716454224038809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038822, "dur": 0,
    "args": {
      "External id": 165597,
      "cbid": 11, "correlation": 165597
    }
  },
  {
    "ph": "f", "id": 165597, "pid": 76337, "tid": -914061504, "ts": 1716454224038822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038823, "dur": 0,
    "args": {
      "External id": 165598,
      "cbid": 11, "correlation": 165598
    }
  },
  {
    "ph": "f", "id": 165598, "pid": 76337, "tid": -914061504, "ts": 1716454224038823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224038836, "dur": 1,
    "args": {
      "External id": 165601,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165601,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 165601, "pid": 5, "tid": 7, "ts": 1716454224038836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038824, "dur": 21,
    "args": {
      "External id": 165601,
      "cbid": 41, "correlation": 165601
    }
  },
  {
    "ph": "s", "id": 165601, "pid": 76337, "tid": -914061504, "ts": 1716454224038824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038846, "dur": 3,
    "args": {
      "External id": 165602,
      "cbid": 131, "correlation": 165602
    }
  },
  {
    "ph": "f", "id": 165602, "pid": 76337, "tid": -914061504, "ts": 1716454224038846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224038874, "dur": 3,
    "args": {
      "External id": 165626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165626, "pid": 5, "tid": 7, "ts": 1716454224038874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038864, "dur": 9,
    "args": {
      "External id": 165626,
      "cbid": 211, "correlation": 165626
    }
  },
  {
    "ph": "s", "id": 165626, "pid": 76337, "tid": -914061504, "ts": 1716454224038864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038874, "dur": 0,
    "args": {
      "External id": 165627,
      "cbid": 11, "correlation": 165627
    }
  },
  {
    "ph": "f", "id": 165627, "pid": 76337, "tid": -914061504, "ts": 1716454224038874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038874, "dur": 0,
    "args": {
      "External id": 165628,
      "cbid": 11, "correlation": 165628
    }
  },
  {
    "ph": "f", "id": 165628, "pid": 76337, "tid": -914061504, "ts": 1716454224038874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038876, "dur": 1,
    "args": {
      "External id": 165630,
      "cbid": 200, "correlation": 165630
    }
  },
  {
    "ph": "f", "id": 165630, "pid": 76337, "tid": -914061504, "ts": 1716454224038876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224038886, "dur": 4,
    "args": {
      "External id": 165632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165632,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165632, "pid": 5, "tid": 7, "ts": 1716454224038886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038878, "dur": 7,
    "args": {
      "External id": 165632,
      "cbid": 211, "correlation": 165632
    }
  },
  {
    "ph": "s", "id": 165632, "pid": 76337, "tid": -914061504, "ts": 1716454224038878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038887, "dur": 0,
    "args": {
      "External id": 165633,
      "cbid": 11, "correlation": 165633
    }
  },
  {
    "ph": "f", "id": 165633, "pid": 76337, "tid": -914061504, "ts": 1716454224038887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038887, "dur": 0,
    "args": {
      "External id": 165634,
      "cbid": 11, "correlation": 165634
    }
  },
  {
    "ph": "f", "id": 165634, "pid": 76337, "tid": -914061504, "ts": 1716454224038887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224038925, "dur": 1,
    "args": {
      "External id": 165641,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165641,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 165641, "pid": 5, "tid": 7, "ts": 1716454224038925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038914, "dur": 19,
    "args": {
      "External id": 165641,
      "cbid": 41, "correlation": 165641
    }
  },
  {
    "ph": "s", "id": 165641, "pid": 76337, "tid": -914061504, "ts": 1716454224038914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038934, "dur": 3,
    "args": {
      "External id": 165642,
      "cbid": 131, "correlation": 165642
    }
  },
  {
    "ph": "f", "id": 165642, "pid": 76337, "tid": -914061504, "ts": 1716454224038934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224038992, "dur": 1,
    "args": {
      "External id": 165652,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 165652,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 165652, "pid": 5, "tid": 7, "ts": 1716454224038992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038972, "dur": 18,
    "args": {
      "External id": 165652,
      "cbid": 41, "correlation": 165652
    }
  },
  {
    "ph": "s", "id": 165652, "pid": 76337, "tid": -914061504, "ts": 1716454224038972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224038991, "dur": 7,
    "args": {
      "External id": 165653,
      "cbid": 131, "correlation": 165653
    }
  },
  {
    "ph": "f", "id": 165653, "pid": 76337, "tid": -914061504, "ts": 1716454224038991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224039062, "dur": 5,
    "args": {
      "External id": 165660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165660,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165660, "pid": 5, "tid": 7, "ts": 1716454224039062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039047, "dur": 16,
    "args": {
      "External id": 165660,
      "cbid": 211, "correlation": 165660
    }
  },
  {
    "ph": "s", "id": 165660, "pid": 76337, "tid": -914061504, "ts": 1716454224039047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224039131, "dur": 3,
    "args": {
      "External id": 165669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165669,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165669, "pid": 5, "tid": 7, "ts": 1716454224039131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039118, "dur": 13,
    "args": {
      "External id": 165669,
      "cbid": 211, "correlation": 165669
    }
  },
  {
    "ph": "s", "id": 165669, "pid": 76337, "tid": -914061504, "ts": 1716454224039118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224039167, "dur": 3,
    "args": {
      "External id": 165677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165677,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165677, "pid": 5, "tid": 7, "ts": 1716454224039167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039158, "dur": 9,
    "args": {
      "External id": 165677,
      "cbid": 211, "correlation": 165677
    }
  },
  {
    "ph": "s", "id": 165677, "pid": 76337, "tid": -914061504, "ts": 1716454224039158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224039200, "dur": 4,
    "args": {
      "External id": 165685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165685,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165685, "pid": 5, "tid": 7, "ts": 1716454224039200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039189, "dur": 11,
    "args": {
      "External id": 165685,
      "cbid": 211, "correlation": 165685
    }
  },
  {
    "ph": "s", "id": 165685, "pid": 76337, "tid": -914061504, "ts": 1716454224039189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224039228, "dur": 4,
    "args": {
      "External id": 165693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165693,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165693, "pid": 5, "tid": 7, "ts": 1716454224039228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039219, "dur": 9,
    "args": {
      "External id": 165693,
      "cbid": 211, "correlation": 165693
    }
  },
  {
    "ph": "s", "id": 165693, "pid": 76337, "tid": -914061504, "ts": 1716454224039219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224039255, "dur": 3,
    "args": {
      "External id": 165701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165701,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165701, "pid": 5, "tid": 7, "ts": 1716454224039255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039246, "dur": 8,
    "args": {
      "External id": 165701,
      "cbid": 211, "correlation": 165701
    }
  },
  {
    "ph": "s", "id": 165701, "pid": 76337, "tid": -914061504, "ts": 1716454224039246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224039280, "dur": 3,
    "args": {
      "External id": 165709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165709,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165709, "pid": 5, "tid": 7, "ts": 1716454224039280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039271, "dur": 8,
    "args": {
      "External id": 165709,
      "cbid": 211, "correlation": 165709
    }
  },
  {
    "ph": "s", "id": 165709, "pid": 76337, "tid": -914061504, "ts": 1716454224039271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224039302, "dur": 4,
    "args": {
      "External id": 165717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165717,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165717, "pid": 5, "tid": 7, "ts": 1716454224039302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039294, "dur": 7,
    "args": {
      "External id": 165717,
      "cbid": 211, "correlation": 165717
    }
  },
  {
    "ph": "s", "id": 165717, "pid": 76337, "tid": -914061504, "ts": 1716454224039294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224039321, "dur": 5,
    "args": {
      "External id": 165725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165725,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165725, "pid": 5, "tid": 7, "ts": 1716454224039321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039313, "dur": 7,
    "args": {
      "External id": 165725,
      "cbid": 211, "correlation": 165725
    }
  },
  {
    "ph": "s", "id": 165725, "pid": 76337, "tid": -914061504, "ts": 1716454224039313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224039342, "dur": 3,
    "args": {
      "External id": 165733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165733,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165733, "pid": 5, "tid": 7, "ts": 1716454224039342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039333, "dur": 7,
    "args": {
      "External id": 165733,
      "cbid": 211, "correlation": 165733
    }
  },
  {
    "ph": "s", "id": 165733, "pid": 76337, "tid": -914061504, "ts": 1716454224039333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224039398, "dur": 3,
    "args": {
      "External id": 165741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165741,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 165741, "pid": 5, "tid": 7, "ts": 1716454224039398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039387, "dur": 10,
    "args": {
      "External id": 165741,
      "cbid": 211, "correlation": 165741
    }
  },
  {
    "ph": "s", "id": 165741, "pid": 76337, "tid": -914061504, "ts": 1716454224039387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224039423, "dur": 4,
    "args": {
      "External id": 165749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165749,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165749, "pid": 5, "tid": 7, "ts": 1716454224039423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039414, "dur": 8,
    "args": {
      "External id": 165749,
      "cbid": 211, "correlation": 165749
    }
  },
  {
    "ph": "s", "id": 165749, "pid": 76337, "tid": -914061504, "ts": 1716454224039414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224039447, "dur": 4,
    "args": {
      "External id": 165757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165757,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165757, "pid": 5, "tid": 7, "ts": 1716454224039447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039439, "dur": 8,
    "args": {
      "External id": 165757,
      "cbid": 211, "correlation": 165757
    }
  },
  {
    "ph": "s", "id": 165757, "pid": 76337, "tid": -914061504, "ts": 1716454224039439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224039466, "dur": 3,
    "args": {
      "External id": 165765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165765,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 165765, "pid": 5, "tid": 7, "ts": 1716454224039466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039458, "dur": 7,
    "args": {
      "External id": 165765,
      "cbid": 211, "correlation": 165765
    }
  },
  {
    "ph": "s", "id": 165765, "pid": 76337, "tid": -914061504, "ts": 1716454224039458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224039867, "dur": 5,
    "args": {
      "External id": 165774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165774,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165774, "pid": 5, "tid": 7, "ts": 1716454224039867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039851, "dur": 17,
    "args": {
      "External id": 165774,
      "cbid": 211, "correlation": 165774
    }
  },
  {
    "ph": "s", "id": 165774, "pid": 76337, "tid": -914061504, "ts": 1716454224039851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224039904, "dur": 5,
    "args": {
      "External id": 165783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165783,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165783, "pid": 5, "tid": 7, "ts": 1716454224039904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224039894, "dur": 9,
    "args": {
      "External id": 165783,
      "cbid": 211, "correlation": 165783
    }
  },
  {
    "ph": "s", "id": 165783, "pid": 76337, "tid": -914061504, "ts": 1716454224039894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224040040, "dur": 3,
    "args": {
      "External id": 165799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165799, "pid": 5, "tid": 7, "ts": 1716454224040040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040025, "dur": 15,
    "args": {
      "External id": 165799,
      "cbid": 211, "correlation": 165799
    }
  },
  {
    "ph": "s", "id": 165799, "pid": 76337, "tid": -914061504, "ts": 1716454224040025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040073, "dur": 3,
    "args": {
      "External id": 165807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165807,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165807, "pid": 5, "tid": 7, "ts": 1716454224040073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040063, "dur": 10,
    "args": {
      "External id": 165807,
      "cbid": 211, "correlation": 165807
    }
  },
  {
    "ph": "s", "id": 165807, "pid": 76337, "tid": -914061504, "ts": 1716454224040063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040105, "dur": 3,
    "args": {
      "External id": 165815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165815,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165815, "pid": 5, "tid": 7, "ts": 1716454224040105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040095, "dur": 8,
    "args": {
      "External id": 165815,
      "cbid": 211, "correlation": 165815
    }
  },
  {
    "ph": "s", "id": 165815, "pid": 76337, "tid": -914061504, "ts": 1716454224040095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040137, "dur": 4,
    "args": {
      "External id": 165823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165823,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165823, "pid": 5, "tid": 7, "ts": 1716454224040137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040127, "dur": 8,
    "args": {
      "External id": 165823,
      "cbid": 211, "correlation": 165823
    }
  },
  {
    "ph": "s", "id": 165823, "pid": 76337, "tid": -914061504, "ts": 1716454224040127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224040192, "dur": 4,
    "args": {
      "External id": 165835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165835,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165835, "pid": 5, "tid": 7, "ts": 1716454224040192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040180, "dur": 12,
    "args": {
      "External id": 165835,
      "cbid": 211, "correlation": 165835
    }
  },
  {
    "ph": "s", "id": 165835, "pid": 76337, "tid": -914061504, "ts": 1716454224040180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224040238, "dur": 4,
    "args": {
      "External id": 165846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165846, "pid": 5, "tid": 7, "ts": 1716454224040238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040226, "dur": 11,
    "args": {
      "External id": 165846,
      "cbid": 211, "correlation": 165846
    }
  },
  {
    "ph": "s", "id": 165846, "pid": 76337, "tid": -914061504, "ts": 1716454224040226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040268, "dur": 3,
    "args": {
      "External id": 165854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165854, "pid": 5, "tid": 7, "ts": 1716454224040268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040258, "dur": 8,
    "args": {
      "External id": 165854,
      "cbid": 211, "correlation": 165854
    }
  },
  {
    "ph": "s", "id": 165854, "pid": 76337, "tid": -914061504, "ts": 1716454224040258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040301, "dur": 5,
    "args": {
      "External id": 165862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165862,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165862, "pid": 5, "tid": 7, "ts": 1716454224040301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040291, "dur": 10,
    "args": {
      "External id": 165862,
      "cbid": 211, "correlation": 165862
    }
  },
  {
    "ph": "s", "id": 165862, "pid": 76337, "tid": -914061504, "ts": 1716454224040291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040331, "dur": 5,
    "args": {
      "External id": 165870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165870,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165870, "pid": 5, "tid": 7, "ts": 1716454224040331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040321, "dur": 9,
    "args": {
      "External id": 165870,
      "cbid": 211, "correlation": 165870
    }
  },
  {
    "ph": "s", "id": 165870, "pid": 76337, "tid": -914061504, "ts": 1716454224040321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224040361, "dur": 4,
    "args": {
      "External id": 165879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165879,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165879, "pid": 5, "tid": 7, "ts": 1716454224040361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040351, "dur": 9,
    "args": {
      "External id": 165879,
      "cbid": 211, "correlation": 165879
    }
  },
  {
    "ph": "s", "id": 165879, "pid": 76337, "tid": -914061504, "ts": 1716454224040351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224040422, "dur": 4,
    "args": {
      "External id": 165892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165892,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165892, "pid": 5, "tid": 7, "ts": 1716454224040422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040408, "dur": 13,
    "args": {
      "External id": 165892,
      "cbid": 211, "correlation": 165892
    }
  },
  {
    "ph": "s", "id": 165892, "pid": 76337, "tid": -914061504, "ts": 1716454224040408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224040461, "dur": 5,
    "args": {
      "External id": 165902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165902,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 165902, "pid": 5, "tid": 7, "ts": 1716454224040461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040450, "dur": 10,
    "args": {
      "External id": 165902,
      "cbid": 211, "correlation": 165902
    }
  },
  {
    "ph": "s", "id": 165902, "pid": 76337, "tid": -914061504, "ts": 1716454224040450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040577, "dur": 4,
    "args": {
      "External id": 165919,
      "cbid": 251, "correlation": 165919
    }
  },
  {
    "ph": "f", "id": 165919, "pid": 76337, "tid": -914061504, "ts": 1716454224040577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454224040604, "dur": 11,
    "args": {
      "External id": 165921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165921,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 165921, "pid": 5, "tid": 7, "ts": 1716454224040604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040588, "dur": 16,
    "args": {
      "External id": 165921,
      "cbid": 211, "correlation": 165921
    }
  },
  {
    "ph": "s", "id": 165921, "pid": 76337, "tid": -914061504, "ts": 1716454224040588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224040660, "dur": 3,
    "args": {
      "External id": 165929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165929,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 165929, "pid": 5, "tid": 7, "ts": 1716454224040660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040648, "dur": 12,
    "args": {
      "External id": 165929,
      "cbid": 211, "correlation": 165929
    }
  },
  {
    "ph": "s", "id": 165929, "pid": 76337, "tid": -914061504, "ts": 1716454224040648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040717, "dur": 1,
    "args": {
      "External id": 165945,
      "cbid": 251, "correlation": 165945
    }
  },
  {
    "ph": "f", "id": 165945, "pid": 76337, "tid": -914061504, "ts": 1716454224040717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040723, "dur": 0,
    "args": {
      "External id": 165947,
      "cbid": 251, "correlation": 165947
    }
  },
  {
    "ph": "f", "id": 165947, "pid": 76337, "tid": -914061504, "ts": 1716454224040723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224040738, "dur": 13,
    "args": {
      "External id": 165948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165948,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 165948, "pid": 5, "tid": 7, "ts": 1716454224040738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040725, "dur": 13,
    "args": {
      "External id": 165948,
      "cbid": 211, "correlation": 165948
    }
  },
  {
    "ph": "s", "id": 165948, "pid": 76337, "tid": -914061504, "ts": 1716454224040725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224040754, "dur": 5,
    "args": {
      "External id": 165950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165950,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 165950, "pid": 5, "tid": 7, "ts": 1716454224040754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040743, "dur": 10,
    "args": {
      "External id": 165950,
      "cbid": 211, "correlation": 165950
    }
  },
  {
    "ph": "s", "id": 165950, "pid": 76337, "tid": -914061504, "ts": 1716454224040743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040848, "dur": 1,
    "args": {
      "External id": 165960,
      "cbid": 317, "correlation": 165960
    }
  },
  {
    "ph": "f", "id": 165960, "pid": 76337, "tid": -914061504, "ts": 1716454224040848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040850, "dur": 1,
    "args": {
      "External id": 165961,
      "cbid": 203, "correlation": 165961
    }
  },
  {
    "ph": "f", "id": 165961, "pid": 76337, "tid": -914061504, "ts": 1716454224040850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040852, "dur": 1,
    "args": {
      "External id": 165962,
      "cbid": 205, "correlation": 165962
    }
  },
  {
    "ph": "f", "id": 165962, "pid": 76337, "tid": -914061504, "ts": 1716454224040852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224040903, "dur": 7,
    "args": {
      "External id": 165966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165966,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165966, "pid": 5, "tid": 7, "ts": 1716454224040903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040889, "dur": 14,
    "args": {
      "External id": 165966,
      "cbid": 211, "correlation": 165966
    }
  },
  {
    "ph": "s", "id": 165966, "pid": 76337, "tid": -914061504, "ts": 1716454224040889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224040914, "dur": 4,
    "args": {
      "External id": 165968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165968,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 165968, "pid": 5, "tid": 7, "ts": 1716454224040914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040906, "dur": 6,
    "args": {
      "External id": 165968,
      "cbid": 211, "correlation": 165968
    }
  },
  {
    "ph": "s", "id": 165968, "pid": 76337, "tid": -914061504, "ts": 1716454224040906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224040932, "dur": 3,
    "args": {
      "External id": 165970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 165970, "pid": 5, "tid": 7, "ts": 1716454224040932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040922, "dur": 9,
    "args": {
      "External id": 165970,
      "cbid": 211, "correlation": 165970
    }
  },
  {
    "ph": "s", "id": 165970, "pid": 76337, "tid": -914061504, "ts": 1716454224040922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040937, "dur": 0,
    "args": {
      "External id": 165971,
      "cbid": 51, "correlation": 165971
    }
  },
  {
    "ph": "s", "id": 165971, "pid": 76337, "tid": -914061504, "ts": 1716454224040937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224040948, "dur": 85,
    "args": {
      "External id": 165972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165972,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 165972, "pid": 5, "tid": 7, "ts": 1716454224040948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040939, "dur": 8,
    "args": {
      "External id": 165972,
      "cbid": 211, "correlation": 165972
    }
  },
  {
    "ph": "s", "id": 165972, "pid": 76337, "tid": -914061504, "ts": 1716454224040939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224041034, "dur": 60,
    "args": {
      "External id": 165977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 165977, "pid": 5, "tid": 7, "ts": 1716454224041034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224040983, "dur": 10,
    "args": {
      "External id": 165977,
      "cbid": 211, "correlation": 165977
    }
  },
  {
    "ph": "s", "id": 165977, "pid": 76337, "tid": -914061504, "ts": 1716454224040983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224042806, "dur": 51,
    "args": {
      "External id": 165997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 165997,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 165997, "pid": 5, "tid": 7, "ts": 1716454224042806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042789, "dur": 17,
    "args": {
      "External id": 165997,
      "cbid": 211, "correlation": 165997
    }
  },
  {
    "ph": "s", "id": 165997, "pid": 76337, "tid": -914061504, "ts": 1716454224042789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224042858, "dur": 4,
    "args": {
      "External id": 166009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166009,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166009, "pid": 5, "tid": 7, "ts": 1716454224042858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042818, "dur": 9,
    "args": {
      "External id": 166009,
      "cbid": 211, "correlation": 166009
    }
  },
  {
    "ph": "s", "id": 166009, "pid": 76337, "tid": -914061504, "ts": 1716454224042818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224042864, "dur": 56,
    "args": {
      "External id": 166012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166012,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166012, "pid": 5, "tid": 7, "ts": 1716454224042864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042842, "dur": 7,
    "args": {
      "External id": 166012,
      "cbid": 211, "correlation": 166012
    }
  },
  {
    "ph": "s", "id": 166012, "pid": 76337, "tid": -914061504, "ts": 1716454224042842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224042921, "dur": 36,
    "args": {
      "External id": 166021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166021,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166021, "pid": 5, "tid": 7, "ts": 1716454224042921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042888, "dur": 10,
    "args": {
      "External id": 166021,
      "cbid": 211, "correlation": 166021
    }
  },
  {
    "ph": "s", "id": 166021, "pid": 76337, "tid": -914061504, "ts": 1716454224042888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042945, "dur": 0,
    "args": {
      "External id": 166031,
      "cbid": 317, "correlation": 166031
    }
  },
  {
    "ph": "f", "id": 166031, "pid": 76337, "tid": -914061504, "ts": 1716454224042945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042946, "dur": 0,
    "args": {
      "External id": 166032,
      "cbid": 203, "correlation": 166032
    }
  },
  {
    "ph": "f", "id": 166032, "pid": 76337, "tid": -914061504, "ts": 1716454224042946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042947, "dur": 0,
    "args": {
      "External id": 166033,
      "cbid": 205, "correlation": 166033
    }
  },
  {
    "ph": "f", "id": 166033, "pid": 76337, "tid": -914061504, "ts": 1716454224042947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224042987, "dur": 40,
    "args": {
      "External id": 166037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166037,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166037, "pid": 5, "tid": 7, "ts": 1716454224042987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042966, "dur": 20,
    "args": {
      "External id": 166037,
      "cbid": 211, "correlation": 166037
    }
  },
  {
    "ph": "s", "id": 166037, "pid": 76337, "tid": -914061504, "ts": 1716454224042966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224043028, "dur": 14,
    "args": {
      "External id": 166039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166039,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166039, "pid": 5, "tid": 7, "ts": 1716454224043028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224042989, "dur": 6,
    "args": {
      "External id": 166039,
      "cbid": 211, "correlation": 166039
    }
  },
  {
    "ph": "s", "id": 166039, "pid": 76337, "tid": -914061504, "ts": 1716454224042989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224043043, "dur": 3,
    "args": {
      "External id": 166041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166041,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166041, "pid": 5, "tid": 7, "ts": 1716454224043043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043001, "dur": 6,
    "args": {
      "External id": 166041,
      "cbid": 211, "correlation": 166041
    }
  },
  {
    "ph": "s", "id": 166041, "pid": 76337, "tid": -914061504, "ts": 1716454224043001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043011, "dur": 0,
    "args": {
      "External id": 166042,
      "cbid": 51, "correlation": 166042
    }
  },
  {
    "ph": "s", "id": 166042, "pid": 76337, "tid": -914061504, "ts": 1716454224043011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224043048, "dur": 705,
    "args": {
      "External id": 166043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166043,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166043, "pid": 5, "tid": 7, "ts": 1716454224043048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043012, "dur": 7,
    "args": {
      "External id": 166043,
      "cbid": 211, "correlation": 166043
    }
  },
  {
    "ph": "s", "id": 166043, "pid": 76337, "tid": -914061504, "ts": 1716454224043012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224043754, "dur": 60,
    "args": {
      "External id": 166048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166048,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166048, "pid": 5, "tid": 7, "ts": 1716454224043754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043043, "dur": 9,
    "args": {
      "External id": 166048,
      "cbid": 211, "correlation": 166048
    }
  },
  {
    "ph": "s", "id": 166048, "pid": 76337, "tid": -914061504, "ts": 1716454224043043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224043816, "dur": 5,
    "args": {
      "External id": 166056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166056,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166056, "pid": 5, "tid": 7, "ts": 1716454224043816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043088, "dur": 9,
    "args": {
      "External id": 166056,
      "cbid": 211, "correlation": 166056
    }
  },
  {
    "ph": "s", "id": 166056, "pid": 76337, "tid": -914061504, "ts": 1716454224043088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043153, "dur": 1,
    "args": {
      "External id": 166072,
      "cbid": 251, "correlation": 166072
    }
  },
  {
    "ph": "f", "id": 166072, "pid": 76337, "tid": -914061504, "ts": 1716454224043153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043158, "dur": 0,
    "args": {
      "External id": 166074,
      "cbid": 251, "correlation": 166074
    }
  },
  {
    "ph": "f", "id": 166074, "pid": 76337, "tid": -914061504, "ts": 1716454224043158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224043822, "dur": 9,
    "args": {
      "External id": 166075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166075,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 166075, "pid": 5, "tid": 7, "ts": 1716454224043822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043160, "dur": 12,
    "args": {
      "External id": 166075,
      "cbid": 211, "correlation": 166075
    }
  },
  {
    "ph": "s", "id": 166075, "pid": 76337, "tid": -914061504, "ts": 1716454224043160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224043832, "dur": 4,
    "args": {
      "External id": 166077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166077,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 166077, "pid": 5, "tid": 7, "ts": 1716454224043832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043174, "dur": 6,
    "args": {
      "External id": 166077,
      "cbid": 211, "correlation": 166077
    }
  },
  {
    "ph": "s", "id": 166077, "pid": 76337, "tid": -914061504, "ts": 1716454224043174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224043837, "dur": 55,
    "args": {
      "External id": 166087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166087, "pid": 5, "tid": 7, "ts": 1716454224043837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043234, "dur": 12,
    "args": {
      "External id": 166087,
      "cbid": 211, "correlation": 166087
    }
  },
  {
    "ph": "s", "id": 166087, "pid": 76337, "tid": -914061504, "ts": 1716454224043234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224043893, "dur": 51,
    "args": {
      "External id": 166107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166107,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 166107, "pid": 5, "tid": 7, "ts": 1716454224043893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043299, "dur": 10,
    "args": {
      "External id": 166107,
      "cbid": 211, "correlation": 166107
    }
  },
  {
    "ph": "s", "id": 166107, "pid": 76337, "tid": -914061504, "ts": 1716454224043299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224043946, "dur": 4,
    "args": {
      "External id": 166119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166119,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166119, "pid": 5, "tid": 7, "ts": 1716454224043946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043320, "dur": 7,
    "args": {
      "External id": 166119,
      "cbid": 211, "correlation": 166119
    }
  },
  {
    "ph": "s", "id": 166119, "pid": 76337, "tid": -914061504, "ts": 1716454224043320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224043951, "dur": 57,
    "args": {
      "External id": 166122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166122,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166122, "pid": 5, "tid": 7, "ts": 1716454224043951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043339, "dur": 7,
    "args": {
      "External id": 166122,
      "cbid": 211, "correlation": 166122
    }
  },
  {
    "ph": "s", "id": 166122, "pid": 76337, "tid": -914061504, "ts": 1716454224043339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224044010, "dur": 36,
    "args": {
      "External id": 166131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166131,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166131, "pid": 5, "tid": 7, "ts": 1716454224044010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043381, "dur": 10,
    "args": {
      "External id": 166131,
      "cbid": 211, "correlation": 166131
    }
  },
  {
    "ph": "s", "id": 166131, "pid": 76337, "tid": -914061504, "ts": 1716454224043381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043450, "dur": 0,
    "args": {
      "External id": 166141,
      "cbid": 317, "correlation": 166141
    }
  },
  {
    "ph": "f", "id": 166141, "pid": 76337, "tid": -914061504, "ts": 1716454224043450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043451, "dur": 0,
    "args": {
      "External id": 166142,
      "cbid": 203, "correlation": 166142
    }
  },
  {
    "ph": "f", "id": 166142, "pid": 76337, "tid": -914061504, "ts": 1716454224043451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043452, "dur": 0,
    "args": {
      "External id": 166143,
      "cbid": 205, "correlation": 166143
    }
  },
  {
    "ph": "f", "id": 166143, "pid": 76337, "tid": -914061504, "ts": 1716454224043452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224044047, "dur": 40,
    "args": {
      "External id": 166147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166147,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166147, "pid": 5, "tid": 7, "ts": 1716454224044047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043467, "dur": 13,
    "args": {
      "External id": 166147,
      "cbid": 211, "correlation": 166147
    }
  },
  {
    "ph": "s", "id": 166147, "pid": 76337, "tid": -914061504, "ts": 1716454224043467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224044088, "dur": 14,
    "args": {
      "External id": 166149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166149,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166149, "pid": 5, "tid": 7, "ts": 1716454224044088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043482, "dur": 5,
    "args": {
      "External id": 166149,
      "cbid": 211, "correlation": 166149
    }
  },
  {
    "ph": "s", "id": 166149, "pid": 76337, "tid": -914061504, "ts": 1716454224043482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224044104, "dur": 3,
    "args": {
      "External id": 166151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166151,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166151, "pid": 5, "tid": 7, "ts": 1716454224044104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043491, "dur": 5,
    "args": {
      "External id": 166151,
      "cbid": 211, "correlation": 166151
    }
  },
  {
    "ph": "s", "id": 166151, "pid": 76337, "tid": -914061504, "ts": 1716454224043491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043499, "dur": 0,
    "args": {
      "External id": 166152,
      "cbid": 51, "correlation": 166152
    }
  },
  {
    "ph": "s", "id": 166152, "pid": 76337, "tid": -914061504, "ts": 1716454224043499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224044109, "dur": 702,
    "args": {
      "External id": 166153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166153,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166153, "pid": 5, "tid": 7, "ts": 1716454224044109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043500, "dur": 5,
    "args": {
      "External id": 166153,
      "cbid": 211, "correlation": 166153
    }
  },
  {
    "ph": "s", "id": 166153, "pid": 76337, "tid": -914061504, "ts": 1716454224043500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224044813, "dur": 59,
    "args": {
      "External id": 166158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166158, "pid": 5, "tid": 7, "ts": 1716454224044813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043528, "dur": 8,
    "args": {
      "External id": 166158,
      "cbid": 211, "correlation": 166158
    }
  },
  {
    "ph": "s", "id": 166158, "pid": 76337, "tid": -914061504, "ts": 1716454224043528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224044873, "dur": 50,
    "args": {
      "External id": 166166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166166,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166166, "pid": 5, "tid": 7, "ts": 1716454224044873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043559, "dur": 9,
    "args": {
      "External id": 166166,
      "cbid": 211, "correlation": 166166
    }
  },
  {
    "ph": "s", "id": 166166, "pid": 76337, "tid": -914061504, "ts": 1716454224043559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224044924, "dur": 35,
    "args": {
      "External id": 166174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166174,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166174, "pid": 5, "tid": 7, "ts": 1716454224044924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043590, "dur": 9,
    "args": {
      "External id": 166174,
      "cbid": 211, "correlation": 166174
    }
  },
  {
    "ph": "s", "id": 166174, "pid": 76337, "tid": -914061504, "ts": 1716454224043590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224044961, "dur": 51,
    "args": {
      "External id": 166194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166194,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 166194, "pid": 5, "tid": 7, "ts": 1716454224044961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043674, "dur": 12,
    "args": {
      "External id": 166194,
      "cbid": 211, "correlation": 166194
    }
  },
  {
    "ph": "s", "id": 166194, "pid": 76337, "tid": -914061504, "ts": 1716454224043674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224045014, "dur": 4,
    "args": {
      "External id": 166206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166206,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166206, "pid": 5, "tid": 7, "ts": 1716454224045014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043695, "dur": 6,
    "args": {
      "External id": 166206,
      "cbid": 211, "correlation": 166206
    }
  },
  {
    "ph": "s", "id": 166206, "pid": 76337, "tid": -914061504, "ts": 1716454224043695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224045019, "dur": 56,
    "args": {
      "External id": 166209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166209, "pid": 5, "tid": 7, "ts": 1716454224045019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043713, "dur": 7,
    "args": {
      "External id": 166209,
      "cbid": 211, "correlation": 166209
    }
  },
  {
    "ph": "s", "id": 166209, "pid": 76337, "tid": -914061504, "ts": 1716454224043713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043772, "dur": 0,
    "args": {
      "External id": 166220,
      "cbid": 317, "correlation": 166220
    }
  },
  {
    "ph": "f", "id": 166220, "pid": 76337, "tid": -914061504, "ts": 1716454224043772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043773, "dur": 0,
    "args": {
      "External id": 166221,
      "cbid": 203, "correlation": 166221
    }
  },
  {
    "ph": "f", "id": 166221, "pid": 76337, "tid": -914061504, "ts": 1716454224043773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043774, "dur": 0,
    "args": {
      "External id": 166222,
      "cbid": 205, "correlation": 166222
    }
  },
  {
    "ph": "f", "id": 166222, "pid": 76337, "tid": -914061504, "ts": 1716454224043774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043810, "dur": 2,
    "args": {
      "External id": 166226,
      "cbid": 251, "correlation": 166226
    }
  },
  {
    "ph": "f", "id": 166226, "pid": 76337, "tid": -914061504, "ts": 1716454224043810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043813, "dur": 1,
    "args": {
      "External id": 166227,
      "cbid": 251, "correlation": 166227
    }
  },
  {
    "ph": "f", "id": 166227, "pid": 76337, "tid": -914061504, "ts": 1716454224043813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043815, "dur": 1,
    "args": {
      "External id": 166228,
      "cbid": 251, "correlation": 166228
    }
  },
  {
    "ph": "f", "id": 166228, "pid": 76337, "tid": -914061504, "ts": 1716454224043815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043818, "dur": 1,
    "args": {
      "External id": 166229,
      "cbid": 251, "correlation": 166229
    }
  },
  {
    "ph": "f", "id": 166229, "pid": 76337, "tid": -914061504, "ts": 1716454224043818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043820, "dur": 1,
    "args": {
      "External id": 166230,
      "cbid": 251, "correlation": 166230
    }
  },
  {
    "ph": "f", "id": 166230, "pid": 76337, "tid": -914061504, "ts": 1716454224043820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043821, "dur": 1,
    "args": {
      "External id": 166231,
      "cbid": 251, "correlation": 166231
    }
  },
  {
    "ph": "f", "id": 166231, "pid": 76337, "tid": -914061504, "ts": 1716454224043821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043823, "dur": 1,
    "args": {
      "External id": 166232,
      "cbid": 251, "correlation": 166232
    }
  },
  {
    "ph": "f", "id": 166232, "pid": 76337, "tid": -914061504, "ts": 1716454224043823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043825, "dur": 1,
    "args": {
      "External id": 166233,
      "cbid": 251, "correlation": 166233
    }
  },
  {
    "ph": "f", "id": 166233, "pid": 76337, "tid": -914061504, "ts": 1716454224043825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043828, "dur": 0,
    "args": {
      "External id": 166234,
      "cbid": 251, "correlation": 166234
    }
  },
  {
    "ph": "f", "id": 166234, "pid": 76337, "tid": -914061504, "ts": 1716454224043828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224045076, "dur": 114,
    "args": {
      "External id": 166235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166235,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166235, "pid": 5, "tid": 7, "ts": 1716454224045076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043831, "dur": 15,
    "args": {
      "External id": 166235,
      "cbid": 211, "correlation": 166235
    }
  },
  {
    "ph": "s", "id": 166235, "pid": 76337, "tid": -914061504, "ts": 1716454224043831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224045192, "dur": 60,
    "args": {
      "External id": 166241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166241, "pid": 5, "tid": 7, "ts": 1716454224045192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043870, "dur": 8,
    "args": {
      "External id": 166241,
      "cbid": 211, "correlation": 166241
    }
  },
  {
    "ph": "s", "id": 166241, "pid": 76337, "tid": -914061504, "ts": 1716454224043870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224045253, "dur": 562,
    "args": {
      "External id": 166250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166250,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166250, "pid": 5, "tid": 7, "ts": 1716454224045253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224043963, "dur": 24,
    "args": {
      "External id": 166250,
      "cbid": 211, "correlation": 166250
    }
  },
  {
    "ph": "s", "id": 166250, "pid": 76337, "tid": -914061504, "ts": 1716454224043963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224045816, "dur": 183,
    "args": {
      "External id": 166272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166272,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166272, "pid": 5, "tid": 7, "ts": 1716454224045816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044045, "dur": 13,
    "args": {
      "External id": 166272,
      "cbid": 211, "correlation": 166272
    }
  },
  {
    "ph": "s", "id": 166272, "pid": 76337, "tid": -914061504, "ts": 1716454224044045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044161, "dur": 2,
    "args": {
      "External id": 166283,
      "cbid": 251, "correlation": 166283
    }
  },
  {
    "ph": "f", "id": 166283, "pid": 76337, "tid": -914061504, "ts": 1716454224044161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224046000, "dur": 197,
    "args": {
      "External id": 166284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166284,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166284, "pid": 5, "tid": 7, "ts": 1716454224046000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044168, "dur": 14,
    "args": {
      "External id": 166284,
      "cbid": 211, "correlation": 166284
    }
  },
  {
    "ph": "s", "id": 166284, "pid": 76337, "tid": -914061504, "ts": 1716454224044168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044242, "dur": 1,
    "args": {
      "External id": 166295,
      "cbid": 251, "correlation": 166295
    }
  },
  {
    "ph": "f", "id": 166295, "pid": 76337, "tid": -914061504, "ts": 1716454224044242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224046199, "dur": 188,
    "args": {
      "External id": 166296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166296,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166296, "pid": 5, "tid": 7, "ts": 1716454224046199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044246, "dur": 12,
    "args": {
      "External id": 166296,
      "cbid": 211, "correlation": 166296
    }
  },
  {
    "ph": "s", "id": 166296, "pid": 76337, "tid": -914061504, "ts": 1716454224044246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044311, "dur": 1,
    "args": {
      "External id": 166307,
      "cbid": 251, "correlation": 166307
    }
  },
  {
    "ph": "f", "id": 166307, "pid": 76337, "tid": -914061504, "ts": 1716454224044311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224046389, "dur": 191,
    "args": {
      "External id": 166308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166308,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166308, "pid": 5, "tid": 7, "ts": 1716454224046389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044315, "dur": 11,
    "args": {
      "External id": 166308,
      "cbid": 211, "correlation": 166308
    }
  },
  {
    "ph": "s", "id": 166308, "pid": 76337, "tid": -914061504, "ts": 1716454224044315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224046581, "dur": 18828,
    "args": {
      "External id": 166329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166329,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166329, "pid": 5, "tid": 7, "ts": 1716454224046581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044423, "dur": 15,
    "args": {
      "External id": 166329,
      "cbid": 211, "correlation": 166329
    }
  },
  {
    "ph": "s", "id": 166329, "pid": 76337, "tid": -914061504, "ts": 1716454224044423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044534, "dur": 3,
    "args": {
      "External id": 166347,
      "cbid": 251, "correlation": 166347
    }
  },
  {
    "ph": "f", "id": 166347, "pid": 76337, "tid": -914061504, "ts": 1716454224044534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224065410, "dur": 203,
    "args": {
      "External id": 166349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166349,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166349, "pid": 5, "tid": 7, "ts": 1716454224065410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044542, "dur": 13,
    "args": {
      "External id": 166349,
      "cbid": 211, "correlation": 166349
    }
  },
  {
    "ph": "s", "id": 166349, "pid": 76337, "tid": -914061504, "ts": 1716454224044542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224065615, "dur": 67,
    "args": {
      "External id": 166357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166357,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166357, "pid": 5, "tid": 7, "ts": 1716454224065615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044614, "dur": 12,
    "args": {
      "External id": 166357,
      "cbid": 211, "correlation": 166357
    }
  },
  {
    "ph": "s", "id": 166357, "pid": 76337, "tid": -914061504, "ts": 1716454224044614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224065683, "dur": 97,
    "args": {
      "External id": 166365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166365,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166365, "pid": 5, "tid": 7, "ts": 1716454224065683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044653, "dur": 9,
    "args": {
      "External id": 166365,
      "cbid": 211, "correlation": 166365
    }
  },
  {
    "ph": "s", "id": 166365, "pid": 76337, "tid": -914061504, "ts": 1716454224044653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224065781, "dur": 55,
    "args": {
      "External id": 166376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166376,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166376, "pid": 5, "tid": 7, "ts": 1716454224065781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044735, "dur": 14,
    "args": {
      "External id": 166376,
      "cbid": 211, "correlation": 166376
    }
  },
  {
    "ph": "s", "id": 166376, "pid": 76337, "tid": -914061504, "ts": 1716454224044735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224065837, "dur": 92,
    "args": {
      "External id": 166398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166398,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166398, "pid": 5, "tid": 7, "ts": 1716454224065837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044769, "dur": 7,
    "args": {
      "External id": 166398,
      "cbid": 211, "correlation": 166398
    }
  },
  {
    "ph": "s", "id": 166398, "pid": 76337, "tid": -914061504, "ts": 1716454224044769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044850, "dur": 1,
    "args": {
      "External id": 166409,
      "cbid": 251, "correlation": 166409
    }
  },
  {
    "ph": "f", "id": 166409, "pid": 76337, "tid": -914061504, "ts": 1716454224044850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224065931, "dur": 104,
    "args": {
      "External id": 166410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166410,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166410, "pid": 5, "tid": 7, "ts": 1716454224065931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044856, "dur": 12,
    "args": {
      "External id": 166410,
      "cbid": 211, "correlation": 166410
    }
  },
  {
    "ph": "s", "id": 166410, "pid": 76337, "tid": -914061504, "ts": 1716454224044856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044941, "dur": 1,
    "args": {
      "External id": 166421,
      "cbid": 251, "correlation": 166421
    }
  },
  {
    "ph": "f", "id": 166421, "pid": 76337, "tid": -914061504, "ts": 1716454224044941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044945, "dur": 0,
    "args": {
      "External id": 166422,
      "cbid": 251, "correlation": 166422
    }
  },
  {
    "ph": "f", "id": 166422, "pid": 76337, "tid": -914061504, "ts": 1716454224044945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224066036, "dur": 10,
    "args": {
      "External id": 166423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166423,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 166423, "pid": 5, "tid": 7, "ts": 1716454224066036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044947, "dur": 14,
    "args": {
      "External id": 166423,
      "cbid": 211, "correlation": 166423
    }
  },
  {
    "ph": "s", "id": 166423, "pid": 76337, "tid": -914061504, "ts": 1716454224044947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224066047, "dur": 5,
    "args": {
      "External id": 166425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166425,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 166425, "pid": 5, "tid": 7, "ts": 1716454224066047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224044964, "dur": 7,
    "args": {
      "External id": 166425,
      "cbid": 211, "correlation": 166425
    }
  },
  {
    "ph": "s", "id": 166425, "pid": 76337, "tid": -914061504, "ts": 1716454224044964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045037, "dur": 1,
    "args": {
      "External id": 166436,
      "cbid": 251, "correlation": 166436
    }
  },
  {
    "ph": "f", "id": 166436, "pid": 76337, "tid": -914061504, "ts": 1716454224045037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045041, "dur": 0,
    "args": {
      "External id": 166437,
      "cbid": 251, "correlation": 166437
    }
  },
  {
    "ph": "f", "id": 166437, "pid": 76337, "tid": -914061504, "ts": 1716454224045041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224066054, "dur": 6,
    "args": {
      "External id": 166438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166438,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 166438, "pid": 5, "tid": 7, "ts": 1716454224066054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045042, "dur": 12,
    "args": {
      "External id": 166438,
      "cbid": 211, "correlation": 166438
    }
  },
  {
    "ph": "s", "id": 166438, "pid": 76337, "tid": -914061504, "ts": 1716454224045042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224066061, "dur": 3,
    "args": {
      "External id": 166440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166440,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 166440, "pid": 5, "tid": 7, "ts": 1716454224066061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045057, "dur": 5,
    "args": {
      "External id": 166440,
      "cbid": 211, "correlation": 166440
    }
  },
  {
    "ph": "s", "id": 166440, "pid": 76337, "tid": -914061504, "ts": 1716454224045057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224066066, "dur": 156,
    "args": {
      "External id": 166461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166461,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166461, "pid": 5, "tid": 7, "ts": 1716454224066066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045130, "dur": 13,
    "args": {
      "External id": 166461,
      "cbid": 211, "correlation": 166461
    }
  },
  {
    "ph": "s", "id": 166461, "pid": 76337, "tid": -914061504, "ts": 1716454224045130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045228, "dur": 2,
    "args": {
      "External id": 166479,
      "cbid": 251, "correlation": 166479
    }
  },
  {
    "ph": "f", "id": 166479, "pid": 76337, "tid": -914061504, "ts": 1716454224045228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224066224, "dur": 107,
    "args": {
      "External id": 166481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166481,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166481, "pid": 5, "tid": 7, "ts": 1716454224066224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045234, "dur": 14,
    "args": {
      "External id": 166481,
      "cbid": 211, "correlation": 166481
    }
  },
  {
    "ph": "s", "id": 166481, "pid": 76337, "tid": -914061504, "ts": 1716454224045234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224066332, "dur": 35,
    "args": {
      "External id": 166489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166489,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166489, "pid": 5, "tid": 7, "ts": 1716454224066332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045305, "dur": 12,
    "args": {
      "External id": 166489,
      "cbid": 211, "correlation": 166489
    }
  },
  {
    "ph": "s", "id": 166489, "pid": 76337, "tid": -914061504, "ts": 1716454224045305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224066368, "dur": 67,
    "args": {
      "External id": 166497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166497, "pid": 5, "tid": 7, "ts": 1716454224066368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045346, "dur": 9,
    "args": {
      "External id": 166497,
      "cbid": 211, "correlation": 166497
    }
  },
  {
    "ph": "s", "id": 166497, "pid": 76337, "tid": -914061504, "ts": 1716454224045346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224066436, "dur": 93,
    "args": {
      "External id": 166519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166519,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166519, "pid": 5, "tid": 7, "ts": 1716454224066436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045399, "dur": 10,
    "args": {
      "External id": 166519,
      "cbid": 211, "correlation": 166519
    }
  },
  {
    "ph": "s", "id": 166519, "pid": 76337, "tid": -914061504, "ts": 1716454224045399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045493, "dur": 1,
    "args": {
      "External id": 166535,
      "cbid": 251, "correlation": 166535
    }
  },
  {
    "ph": "f", "id": 166535, "pid": 76337, "tid": -914061504, "ts": 1716454224045493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224066530, "dur": 580,
    "args": {
      "External id": 166537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166537,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166537, "pid": 5, "tid": 7, "ts": 1716454224066530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045498, "dur": 13,
    "args": {
      "External id": 166537,
      "cbid": 211, "correlation": 166537
    }
  },
  {
    "ph": "s", "id": 166537, "pid": 76337, "tid": -914061504, "ts": 1716454224045498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224067111, "dur": 243,
    "args": {
      "External id": 166545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166545, "pid": 5, "tid": 7, "ts": 1716454224067111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045577, "dur": 15,
    "args": {
      "External id": 166545,
      "cbid": 211, "correlation": 166545
    }
  },
  {
    "ph": "s", "id": 166545, "pid": 76337, "tid": -914061504, "ts": 1716454224045577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224067356, "dur": 253,
    "args": {
      "External id": 166553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166553, "pid": 5, "tid": 7, "ts": 1716454224067356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045612, "dur": 8,
    "args": {
      "External id": 166553,
      "cbid": 211, "correlation": 166553
    }
  },
  {
    "ph": "s", "id": 166553, "pid": 76337, "tid": -914061504, "ts": 1716454224045612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045695, "dur": 2,
    "args": {
      "External id": 166569,
      "cbid": 251, "correlation": 166569
    }
  },
  {
    "ph": "f", "id": 166569, "pid": 76337, "tid": -914061504, "ts": 1716454224045695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045701, "dur": 0,
    "args": {
      "External id": 166571,
      "cbid": 251, "correlation": 166571
    }
  },
  {
    "ph": "f", "id": 166571, "pid": 76337, "tid": -914061504, "ts": 1716454224045701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224067611, "dur": 361,
    "args": {
      "External id": 166572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166572,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 166572, "pid": 5, "tid": 7, "ts": 1716454224067611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045707, "dur": 13,
    "args": {
      "External id": 166572,
      "cbid": 211, "correlation": 166572
    }
  },
  {
    "ph": "s", "id": 166572, "pid": 76337, "tid": -914061504, "ts": 1716454224045707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224067973, "dur": 50,
    "args": {
      "External id": 166580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166580,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166580, "pid": 5, "tid": 7, "ts": 1716454224067973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045750, "dur": 10,
    "args": {
      "External id": 166580,
      "cbid": 211, "correlation": 166580
    }
  },
  {
    "ph": "s", "id": 166580, "pid": 76337, "tid": -914061504, "ts": 1716454224045750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224068024, "dur": 159,
    "args": {
      "External id": 166591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166591,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166591, "pid": 5, "tid": 7, "ts": 1716454224068024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045820, "dur": 13,
    "args": {
      "External id": 166591,
      "cbid": 211, "correlation": 166591
    }
  },
  {
    "ph": "s", "id": 166591, "pid": 76337, "tid": -914061504, "ts": 1716454224045820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045885, "dur": 0,
    "args": {
      "External id": 166603,
      "cbid": 317, "correlation": 166603
    }
  },
  {
    "ph": "f", "id": 166603, "pid": 76337, "tid": -914061504, "ts": 1716454224045885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045886, "dur": 0,
    "args": {
      "External id": 166604,
      "cbid": 203, "correlation": 166604
    }
  },
  {
    "ph": "f", "id": 166604, "pid": 76337, "tid": -914061504, "ts": 1716454224045886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045887, "dur": 0,
    "args": {
      "External id": 166605,
      "cbid": 205, "correlation": 166605
    }
  },
  {
    "ph": "f", "id": 166605, "pid": 76337, "tid": -914061504, "ts": 1716454224045887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045910, "dur": 1,
    "args": {
      "External id": 166609,
      "cbid": 251, "correlation": 166609
    }
  },
  {
    "ph": "f", "id": 166609, "pid": 76337, "tid": -914061504, "ts": 1716454224045910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045912, "dur": 0,
    "args": {
      "External id": 166610,
      "cbid": 251, "correlation": 166610
    }
  },
  {
    "ph": "f", "id": 166610, "pid": 76337, "tid": -914061504, "ts": 1716454224045912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045913, "dur": 0,
    "args": {
      "External id": 166611,
      "cbid": 251, "correlation": 166611
    }
  },
  {
    "ph": "f", "id": 166611, "pid": 76337, "tid": -914061504, "ts": 1716454224045913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045914, "dur": 0,
    "args": {
      "External id": 166612,
      "cbid": 251, "correlation": 166612
    }
  },
  {
    "ph": "f", "id": 166612, "pid": 76337, "tid": -914061504, "ts": 1716454224045914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045915, "dur": 0,
    "args": {
      "External id": 166613,
      "cbid": 251, "correlation": 166613
    }
  },
  {
    "ph": "f", "id": 166613, "pid": 76337, "tid": -914061504, "ts": 1716454224045915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045915, "dur": 0,
    "args": {
      "External id": 166614,
      "cbid": 251, "correlation": 166614
    }
  },
  {
    "ph": "f", "id": 166614, "pid": 76337, "tid": -914061504, "ts": 1716454224045915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045916, "dur": 0,
    "args": {
      "External id": 166615,
      "cbid": 251, "correlation": 166615
    }
  },
  {
    "ph": "f", "id": 166615, "pid": 76337, "tid": -914061504, "ts": 1716454224045916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045917, "dur": 0,
    "args": {
      "External id": 166616,
      "cbid": 251, "correlation": 166616
    }
  },
  {
    "ph": "f", "id": 166616, "pid": 76337, "tid": -914061504, "ts": 1716454224045917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045918, "dur": 0,
    "args": {
      "External id": 166617,
      "cbid": 251, "correlation": 166617
    }
  },
  {
    "ph": "f", "id": 166617, "pid": 76337, "tid": -914061504, "ts": 1716454224045918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224068184, "dur": 116,
    "args": {
      "External id": 166618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166618,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166618, "pid": 5, "tid": 7, "ts": 1716454224068184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045920, "dur": 13,
    "args": {
      "External id": 166618,
      "cbid": 211, "correlation": 166618
    }
  },
  {
    "ph": "s", "id": 166618, "pid": 76337, "tid": -914061504, "ts": 1716454224045920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224068302, "dur": 60,
    "args": {
      "External id": 166624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166624,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166624, "pid": 5, "tid": 7, "ts": 1716454224068302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045955, "dur": 9,
    "args": {
      "External id": 166624,
      "cbid": 211, "correlation": 166624
    }
  },
  {
    "ph": "s", "id": 166624, "pid": 76337, "tid": -914061504, "ts": 1716454224045955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224068363, "dur": 50,
    "args": {
      "External id": 166632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166632,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166632, "pid": 5, "tid": 7, "ts": 1716454224068363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224045997, "dur": 10,
    "args": {
      "External id": 166632,
      "cbid": 211, "correlation": 166632
    }
  },
  {
    "ph": "s", "id": 166632, "pid": 76337, "tid": -914061504, "ts": 1716454224045997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224068414, "dur": 52,
    "args": {
      "External id": 166652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166652,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 166652, "pid": 5, "tid": 7, "ts": 1716454224068414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046071, "dur": 12,
    "args": {
      "External id": 166652,
      "cbid": 211, "correlation": 166652
    }
  },
  {
    "ph": "s", "id": 166652, "pid": 76337, "tid": -914061504, "ts": 1716454224046071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224068467, "dur": 5,
    "args": {
      "External id": 166664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166664,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166664, "pid": 5, "tid": 7, "ts": 1716454224068467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046093, "dur": 6,
    "args": {
      "External id": 166664,
      "cbid": 211, "correlation": 166664
    }
  },
  {
    "ph": "s", "id": 166664, "pid": 76337, "tid": -914061504, "ts": 1716454224046093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224068473, "dur": 57,
    "args": {
      "External id": 166667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166667, "pid": 5, "tid": 7, "ts": 1716454224068473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046111, "dur": 6,
    "args": {
      "External id": 166667,
      "cbid": 211, "correlation": 166667
    }
  },
  {
    "ph": "s", "id": 166667, "pid": 76337, "tid": -914061504, "ts": 1716454224046111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224068531, "dur": 37,
    "args": {
      "External id": 166676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166676,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166676, "pid": 5, "tid": 7, "ts": 1716454224068531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046150, "dur": 10,
    "args": {
      "External id": 166676,
      "cbid": 211, "correlation": 166676
    }
  },
  {
    "ph": "s", "id": 166676, "pid": 76337, "tid": -914061504, "ts": 1716454224046150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046203, "dur": 0,
    "args": {
      "External id": 166686,
      "cbid": 317, "correlation": 166686
    }
  },
  {
    "ph": "f", "id": 166686, "pid": 76337, "tid": -914061504, "ts": 1716454224046203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046203, "dur": 0,
    "args": {
      "External id": 166687,
      "cbid": 203, "correlation": 166687
    }
  },
  {
    "ph": "f", "id": 166687, "pid": 76337, "tid": -914061504, "ts": 1716454224046203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046204, "dur": 0,
    "args": {
      "External id": 166688,
      "cbid": 205, "correlation": 166688
    }
  },
  {
    "ph": "f", "id": 166688, "pid": 76337, "tid": -914061504, "ts": 1716454224046204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224068570, "dur": 41,
    "args": {
      "External id": 166692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166692, "pid": 5, "tid": 7, "ts": 1716454224068570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046218, "dur": 12,
    "args": {
      "External id": 166692,
      "cbid": 211, "correlation": 166692
    }
  },
  {
    "ph": "s", "id": 166692, "pid": 76337, "tid": -914061504, "ts": 1716454224046218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224068612, "dur": 14,
    "args": {
      "External id": 166694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166694,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166694, "pid": 5, "tid": 7, "ts": 1716454224068612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046233, "dur": 5,
    "args": {
      "External id": 166694,
      "cbid": 211, "correlation": 166694
    }
  },
  {
    "ph": "s", "id": 166694, "pid": 76337, "tid": -914061504, "ts": 1716454224046233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224068627, "dur": 4,
    "args": {
      "External id": 166696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166696, "pid": 5, "tid": 7, "ts": 1716454224068627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046242, "dur": 6,
    "args": {
      "External id": 166696,
      "cbid": 211, "correlation": 166696
    }
  },
  {
    "ph": "s", "id": 166696, "pid": 76337, "tid": -914061504, "ts": 1716454224046242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046251, "dur": 0,
    "args": {
      "External id": 166697,
      "cbid": 51, "correlation": 166697
    }
  },
  {
    "ph": "s", "id": 166697, "pid": 76337, "tid": -914061504, "ts": 1716454224046251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224068633, "dur": 707,
    "args": {
      "External id": 166698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166698,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166698, "pid": 5, "tid": 7, "ts": 1716454224068633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046252, "dur": 5,
    "args": {
      "External id": 166698,
      "cbid": 211, "correlation": 166698
    }
  },
  {
    "ph": "s", "id": 166698, "pid": 76337, "tid": -914061504, "ts": 1716454224046252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224069341, "dur": 59,
    "args": {
      "External id": 166703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166703, "pid": 5, "tid": 7, "ts": 1716454224069341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046279, "dur": 9,
    "args": {
      "External id": 166703,
      "cbid": 211, "correlation": 166703
    }
  },
  {
    "ph": "s", "id": 166703, "pid": 76337, "tid": -914061504, "ts": 1716454224046279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224069401, "dur": 4,
    "args": {
      "External id": 166711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166711,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166711, "pid": 5, "tid": 7, "ts": 1716454224069401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046321, "dur": 10,
    "args": {
      "External id": 166711,
      "cbid": 211, "correlation": 166711
    }
  },
  {
    "ph": "s", "id": 166711, "pid": 76337, "tid": -914061504, "ts": 1716454224046321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046387, "dur": 1,
    "args": {
      "External id": 166727,
      "cbid": 251, "correlation": 166727
    }
  },
  {
    "ph": "f", "id": 166727, "pid": 76337, "tid": -914061504, "ts": 1716454224046387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046392, "dur": 0,
    "args": {
      "External id": 166729,
      "cbid": 251, "correlation": 166729
    }
  },
  {
    "ph": "f", "id": 166729, "pid": 76337, "tid": -914061504, "ts": 1716454224046392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224069407, "dur": 11,
    "args": {
      "External id": 166730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166730,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 166730, "pid": 5, "tid": 7, "ts": 1716454224069407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046394, "dur": 11,
    "args": {
      "External id": 166730,
      "cbid": 211, "correlation": 166730
    }
  },
  {
    "ph": "s", "id": 166730, "pid": 76337, "tid": -914061504, "ts": 1716454224046394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224069420, "dur": 5,
    "args": {
      "External id": 166732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166732,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 166732, "pid": 5, "tid": 7, "ts": 1716454224069420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046407, "dur": 5,
    "args": {
      "External id": 166732,
      "cbid": 211, "correlation": 166732
    }
  },
  {
    "ph": "s", "id": 166732, "pid": 76337, "tid": -914061504, "ts": 1716454224046407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224069426, "dur": 53,
    "args": {
      "External id": 166742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166742, "pid": 5, "tid": 7, "ts": 1716454224069426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046464, "dur": 13,
    "args": {
      "External id": 166742,
      "cbid": 211, "correlation": 166742
    }
  },
  {
    "ph": "s", "id": 166742, "pid": 76337, "tid": -914061504, "ts": 1716454224046464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224069480, "dur": 52,
    "args": {
      "External id": 166762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166762,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 166762, "pid": 5, "tid": 7, "ts": 1716454224069480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046532, "dur": 11,
    "args": {
      "External id": 166762,
      "cbid": 211, "correlation": 166762
    }
  },
  {
    "ph": "s", "id": 166762, "pid": 76337, "tid": -914061504, "ts": 1716454224046532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224069534, "dur": 4,
    "args": {
      "External id": 166774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166774,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166774, "pid": 5, "tid": 7, "ts": 1716454224069534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046553, "dur": 6,
    "args": {
      "External id": 166774,
      "cbid": 211, "correlation": 166774
    }
  },
  {
    "ph": "s", "id": 166774, "pid": 76337, "tid": -914061504, "ts": 1716454224046553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224069539, "dur": 57,
    "args": {
      "External id": 166777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166777, "pid": 5, "tid": 7, "ts": 1716454224069539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046571, "dur": 7,
    "args": {
      "External id": 166777,
      "cbid": 211, "correlation": 166777
    }
  },
  {
    "ph": "s", "id": 166777, "pid": 76337, "tid": -914061504, "ts": 1716454224046571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224069597, "dur": 36,
    "args": {
      "External id": 166786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166786,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166786, "pid": 5, "tid": 7, "ts": 1716454224069597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046612, "dur": 10,
    "args": {
      "External id": 166786,
      "cbid": 211, "correlation": 166786
    }
  },
  {
    "ph": "s", "id": 166786, "pid": 76337, "tid": -914061504, "ts": 1716454224046612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046675, "dur": 0,
    "args": {
      "External id": 166796,
      "cbid": 317, "correlation": 166796
    }
  },
  {
    "ph": "f", "id": 166796, "pid": 76337, "tid": -914061504, "ts": 1716454224046675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046675, "dur": 0,
    "args": {
      "External id": 166797,
      "cbid": 203, "correlation": 166797
    }
  },
  {
    "ph": "f", "id": 166797, "pid": 76337, "tid": -914061504, "ts": 1716454224046675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046676, "dur": 0,
    "args": {
      "External id": 166798,
      "cbid": 205, "correlation": 166798
    }
  },
  {
    "ph": "f", "id": 166798, "pid": 76337, "tid": -914061504, "ts": 1716454224046676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224069634, "dur": 40,
    "args": {
      "External id": 166802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166802, "pid": 5, "tid": 7, "ts": 1716454224069634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046691, "dur": 12,
    "args": {
      "External id": 166802,
      "cbid": 211, "correlation": 166802
    }
  },
  {
    "ph": "s", "id": 166802, "pid": 76337, "tid": -914061504, "ts": 1716454224046691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224069676, "dur": 14,
    "args": {
      "External id": 166804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166804,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166804, "pid": 5, "tid": 7, "ts": 1716454224069676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046705, "dur": 5,
    "args": {
      "External id": 166804,
      "cbid": 211, "correlation": 166804
    }
  },
  {
    "ph": "s", "id": 166804, "pid": 76337, "tid": -914061504, "ts": 1716454224046705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224069691, "dur": 3,
    "args": {
      "External id": 166806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166806,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166806, "pid": 5, "tid": 7, "ts": 1716454224069691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046714, "dur": 5,
    "args": {
      "External id": 166806,
      "cbid": 211, "correlation": 166806
    }
  },
  {
    "ph": "s", "id": 166806, "pid": 76337, "tid": -914061504, "ts": 1716454224046714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046722, "dur": 0,
    "args": {
      "External id": 166807,
      "cbid": 51, "correlation": 166807
    }
  },
  {
    "ph": "s", "id": 166807, "pid": 76337, "tid": -914061504, "ts": 1716454224046722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224069696, "dur": 699,
    "args": {
      "External id": 166808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166808,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166808, "pid": 5, "tid": 7, "ts": 1716454224069696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046723, "dur": 5,
    "args": {
      "External id": 166808,
      "cbid": 211, "correlation": 166808
    }
  },
  {
    "ph": "s", "id": 166808, "pid": 76337, "tid": -914061504, "ts": 1716454224046723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224070396, "dur": 60,
    "args": {
      "External id": 166813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166813, "pid": 5, "tid": 7, "ts": 1716454224070396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046751, "dur": 8,
    "args": {
      "External id": 166813,
      "cbid": 211, "correlation": 166813
    }
  },
  {
    "ph": "s", "id": 166813, "pid": 76337, "tid": -914061504, "ts": 1716454224046751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224070457, "dur": 49,
    "args": {
      "External id": 166821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166821,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166821, "pid": 5, "tid": 7, "ts": 1716454224070457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046784, "dur": 8,
    "args": {
      "External id": 166821,
      "cbid": 211, "correlation": 166821
    }
  },
  {
    "ph": "s", "id": 166821, "pid": 76337, "tid": -914061504, "ts": 1716454224046784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224070508, "dur": 35,
    "args": {
      "External id": 166829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166829,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166829, "pid": 5, "tid": 7, "ts": 1716454224070508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046813, "dur": 8,
    "args": {
      "External id": 166829,
      "cbid": 211, "correlation": 166829
    }
  },
  {
    "ph": "s", "id": 166829, "pid": 76337, "tid": -914061504, "ts": 1716454224046813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224070544, "dur": 51,
    "args": {
      "External id": 166849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166849,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 166849, "pid": 5, "tid": 7, "ts": 1716454224070544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046893, "dur": 12,
    "args": {
      "External id": 166849,
      "cbid": 211, "correlation": 166849
    }
  },
  {
    "ph": "s", "id": 166849, "pid": 76337, "tid": -914061504, "ts": 1716454224046893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224070597, "dur": 4,
    "args": {
      "External id": 166861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166861,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 166861, "pid": 5, "tid": 7, "ts": 1716454224070597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046915, "dur": 6,
    "args": {
      "External id": 166861,
      "cbid": 211, "correlation": 166861
    }
  },
  {
    "ph": "s", "id": 166861, "pid": 76337, "tid": -914061504, "ts": 1716454224046915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224070602, "dur": 56,
    "args": {
      "External id": 166864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166864, "pid": 5, "tid": 7, "ts": 1716454224070602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046932, "dur": 8,
    "args": {
      "External id": 166864,
      "cbid": 211, "correlation": 166864
    }
  },
  {
    "ph": "s", "id": 166864, "pid": 76337, "tid": -914061504, "ts": 1716454224046932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046998, "dur": 0,
    "args": {
      "External id": 166875,
      "cbid": 317, "correlation": 166875
    }
  },
  {
    "ph": "f", "id": 166875, "pid": 76337, "tid": -914061504, "ts": 1716454224046998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224046999, "dur": 0,
    "args": {
      "External id": 166876,
      "cbid": 203, "correlation": 166876
    }
  },
  {
    "ph": "f", "id": 166876, "pid": 76337, "tid": -914061504, "ts": 1716454224046999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047000, "dur": 0,
    "args": {
      "External id": 166877,
      "cbid": 205, "correlation": 166877
    }
  },
  {
    "ph": "f", "id": 166877, "pid": 76337, "tid": -914061504, "ts": 1716454224047000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047021, "dur": 1,
    "args": {
      "External id": 166881,
      "cbid": 251, "correlation": 166881
    }
  },
  {
    "ph": "f", "id": 166881, "pid": 76337, "tid": -914061504, "ts": 1716454224047021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047023, "dur": 0,
    "args": {
      "External id": 166882,
      "cbid": 251, "correlation": 166882
    }
  },
  {
    "ph": "f", "id": 166882, "pid": 76337, "tid": -914061504, "ts": 1716454224047023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047024, "dur": 0,
    "args": {
      "External id": 166883,
      "cbid": 251, "correlation": 166883
    }
  },
  {
    "ph": "f", "id": 166883, "pid": 76337, "tid": -914061504, "ts": 1716454224047024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047025, "dur": 0,
    "args": {
      "External id": 166884,
      "cbid": 251, "correlation": 166884
    }
  },
  {
    "ph": "f", "id": 166884, "pid": 76337, "tid": -914061504, "ts": 1716454224047025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047026, "dur": 0,
    "args": {
      "External id": 166885,
      "cbid": 251, "correlation": 166885
    }
  },
  {
    "ph": "f", "id": 166885, "pid": 76337, "tid": -914061504, "ts": 1716454224047026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047026, "dur": 0,
    "args": {
      "External id": 166886,
      "cbid": 251, "correlation": 166886
    }
  },
  {
    "ph": "f", "id": 166886, "pid": 76337, "tid": -914061504, "ts": 1716454224047026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047028, "dur": 0,
    "args": {
      "External id": 166887,
      "cbid": 251, "correlation": 166887
    }
  },
  {
    "ph": "f", "id": 166887, "pid": 76337, "tid": -914061504, "ts": 1716454224047028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047028, "dur": 0,
    "args": {
      "External id": 166888,
      "cbid": 251, "correlation": 166888
    }
  },
  {
    "ph": "f", "id": 166888, "pid": 76337, "tid": -914061504, "ts": 1716454224047028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047030, "dur": 0,
    "args": {
      "External id": 166889,
      "cbid": 251, "correlation": 166889
    }
  },
  {
    "ph": "f", "id": 166889, "pid": 76337, "tid": -914061504, "ts": 1716454224047030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224070659, "dur": 110,
    "args": {
      "External id": 166890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166890,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166890, "pid": 5, "tid": 7, "ts": 1716454224070659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047032, "dur": 13,
    "args": {
      "External id": 166890,
      "cbid": 211, "correlation": 166890
    }
  },
  {
    "ph": "s", "id": 166890, "pid": 76337, "tid": -914061504, "ts": 1716454224047032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224070770, "dur": 60,
    "args": {
      "External id": 166896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166896,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166896, "pid": 5, "tid": 7, "ts": 1716454224070770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047068, "dur": 8,
    "args": {
      "External id": 166896,
      "cbid": 211, "correlation": 166896
    }
  },
  {
    "ph": "s", "id": 166896, "pid": 76337, "tid": -914061504, "ts": 1716454224047068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224070832, "dur": 600,
    "args": {
      "External id": 166905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166905,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166905, "pid": 5, "tid": 7, "ts": 1716454224070832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047151, "dur": 14,
    "args": {
      "External id": 166905,
      "cbid": 211, "correlation": 166905
    }
  },
  {
    "ph": "s", "id": 166905, "pid": 76337, "tid": -914061504, "ts": 1716454224047151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224071433, "dur": 182,
    "args": {
      "External id": 166927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166927,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 166927, "pid": 5, "tid": 7, "ts": 1716454224071433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047208, "dur": 10,
    "args": {
      "External id": 166927,
      "cbid": 211, "correlation": 166927
    }
  },
  {
    "ph": "s", "id": 166927, "pid": 76337, "tid": -914061504, "ts": 1716454224047208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047293, "dur": 1,
    "args": {
      "External id": 166938,
      "cbid": 251, "correlation": 166938
    }
  },
  {
    "ph": "f", "id": 166938, "pid": 76337, "tid": -914061504, "ts": 1716454224047293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224071617, "dur": 197,
    "args": {
      "External id": 166939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166939,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166939, "pid": 5, "tid": 7, "ts": 1716454224071617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047298, "dur": 13,
    "args": {
      "External id": 166939,
      "cbid": 211, "correlation": 166939
    }
  },
  {
    "ph": "s", "id": 166939, "pid": 76337, "tid": -914061504, "ts": 1716454224047298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047366, "dur": 1,
    "args": {
      "External id": 166950,
      "cbid": 251, "correlation": 166950
    }
  },
  {
    "ph": "f", "id": 166950, "pid": 76337, "tid": -914061504, "ts": 1716454224047366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224071815, "dur": 190,
    "args": {
      "External id": 166951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166951,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166951, "pid": 5, "tid": 7, "ts": 1716454224071815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047370, "dur": 11,
    "args": {
      "External id": 166951,
      "cbid": 211, "correlation": 166951
    }
  },
  {
    "ph": "s", "id": 166951, "pid": 76337, "tid": -914061504, "ts": 1716454224047370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047433, "dur": 1,
    "args": {
      "External id": 166962,
      "cbid": 251, "correlation": 166962
    }
  },
  {
    "ph": "f", "id": 166962, "pid": 76337, "tid": -914061504, "ts": 1716454224047433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224072006, "dur": 190,
    "args": {
      "External id": 166963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166963,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 166963, "pid": 5, "tid": 7, "ts": 1716454224072006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047437, "dur": 12,
    "args": {
      "External id": 166963,
      "cbid": 211, "correlation": 166963
    }
  },
  {
    "ph": "s", "id": 166963, "pid": 76337, "tid": -914061504, "ts": 1716454224047437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224072197, "dur": 18772,
    "args": {
      "External id": 166984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 166984,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 166984, "pid": 5, "tid": 7, "ts": 1716454224072197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047516, "dur": 12,
    "args": {
      "External id": 166984,
      "cbid": 211, "correlation": 166984
    }
  },
  {
    "ph": "s", "id": 166984, "pid": 76337, "tid": -914061504, "ts": 1716454224047516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047613, "dur": 1,
    "args": {
      "External id": 167002,
      "cbid": 251, "correlation": 167002
    }
  },
  {
    "ph": "f", "id": 167002, "pid": 76337, "tid": -914061504, "ts": 1716454224047613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224090970, "dur": 203,
    "args": {
      "External id": 167004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167004,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167004, "pid": 5, "tid": 7, "ts": 1716454224090970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047619, "dur": 13,
    "args": {
      "External id": 167004,
      "cbid": 211, "correlation": 167004
    }
  },
  {
    "ph": "s", "id": 167004, "pid": 76337, "tid": -914061504, "ts": 1716454224047619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224091175, "dur": 66,
    "args": {
      "External id": 167012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167012,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167012, "pid": 5, "tid": 7, "ts": 1716454224091175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047689, "dur": 12,
    "args": {
      "External id": 167012,
      "cbid": 211, "correlation": 167012
    }
  },
  {
    "ph": "s", "id": 167012, "pid": 76337, "tid": -914061504, "ts": 1716454224047689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224091243, "dur": 97,
    "args": {
      "External id": 167020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167020,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167020, "pid": 5, "tid": 7, "ts": 1716454224091243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047727, "dur": 9,
    "args": {
      "External id": 167020,
      "cbid": 211, "correlation": 167020
    }
  },
  {
    "ph": "s", "id": 167020, "pid": 76337, "tid": -914061504, "ts": 1716454224047727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224091341, "dur": 56,
    "args": {
      "External id": 167031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167031, "pid": 5, "tid": 7, "ts": 1716454224091341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047798, "dur": 12,
    "args": {
      "External id": 167031,
      "cbid": 211, "correlation": 167031
    }
  },
  {
    "ph": "s", "id": 167031, "pid": 76337, "tid": -914061504, "ts": 1716454224047798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224091398, "dur": 92,
    "args": {
      "External id": 167053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167053,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167053, "pid": 5, "tid": 7, "ts": 1716454224091398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047829, "dur": 8,
    "args": {
      "External id": 167053,
      "cbid": 211, "correlation": 167053
    }
  },
  {
    "ph": "s", "id": 167053, "pid": 76337, "tid": -914061504, "ts": 1716454224047829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047913, "dur": 1,
    "args": {
      "External id": 167064,
      "cbid": 251, "correlation": 167064
    }
  },
  {
    "ph": "f", "id": 167064, "pid": 76337, "tid": -914061504, "ts": 1716454224047913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224091492, "dur": 106,
    "args": {
      "External id": 167065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167065,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167065, "pid": 5, "tid": 7, "ts": 1716454224091492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047918, "dur": 12,
    "args": {
      "External id": 167065,
      "cbid": 211, "correlation": 167065
    }
  },
  {
    "ph": "s", "id": 167065, "pid": 76337, "tid": -914061504, "ts": 1716454224047918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224047997, "dur": 1,
    "args": {
      "External id": 167076,
      "cbid": 251, "correlation": 167076
    }
  },
  {
    "ph": "f", "id": 167076, "pid": 76337, "tid": -914061504, "ts": 1716454224047997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048001, "dur": 0,
    "args": {
      "External id": 167077,
      "cbid": 251, "correlation": 167077
    }
  },
  {
    "ph": "f", "id": 167077, "pid": 76337, "tid": -914061504, "ts": 1716454224048001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224091599, "dur": 10,
    "args": {
      "External id": 167078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167078,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 167078, "pid": 5, "tid": 7, "ts": 1716454224091599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048002, "dur": 13,
    "args": {
      "External id": 167078,
      "cbid": 211, "correlation": 167078
    }
  },
  {
    "ph": "s", "id": 167078, "pid": 76337, "tid": -914061504, "ts": 1716454224048002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224091610, "dur": 5,
    "args": {
      "External id": 167080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167080,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 167080, "pid": 5, "tid": 7, "ts": 1716454224091610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048017, "dur": 6,
    "args": {
      "External id": 167080,
      "cbid": 211, "correlation": 167080
    }
  },
  {
    "ph": "s", "id": 167080, "pid": 76337, "tid": -914061504, "ts": 1716454224048017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048079, "dur": 1,
    "args": {
      "External id": 167091,
      "cbid": 251, "correlation": 167091
    }
  },
  {
    "ph": "f", "id": 167091, "pid": 76337, "tid": -914061504, "ts": 1716454224048079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048082, "dur": 0,
    "args": {
      "External id": 167092,
      "cbid": 251, "correlation": 167092
    }
  },
  {
    "ph": "f", "id": 167092, "pid": 76337, "tid": -914061504, "ts": 1716454224048082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224091617, "dur": 6,
    "args": {
      "External id": 167093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167093,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 167093, "pid": 5, "tid": 7, "ts": 1716454224091617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048084, "dur": 12,
    "args": {
      "External id": 167093,
      "cbid": 211, "correlation": 167093
    }
  },
  {
    "ph": "s", "id": 167093, "pid": 76337, "tid": -914061504, "ts": 1716454224048084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224091624, "dur": 3,
    "args": {
      "External id": 167095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167095,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 167095, "pid": 5, "tid": 7, "ts": 1716454224091624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048098, "dur": 5,
    "args": {
      "External id": 167095,
      "cbid": 211, "correlation": 167095
    }
  },
  {
    "ph": "s", "id": 167095, "pid": 76337, "tid": -914061504, "ts": 1716454224048098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224091629, "dur": 156,
    "args": {
      "External id": 167116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167116,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 167116, "pid": 5, "tid": 7, "ts": 1716454224091629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048171, "dur": 12,
    "args": {
      "External id": 167116,
      "cbid": 211, "correlation": 167116
    }
  },
  {
    "ph": "s", "id": 167116, "pid": 76337, "tid": -914061504, "ts": 1716454224048171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048266, "dur": 1,
    "args": {
      "External id": 167134,
      "cbid": 251, "correlation": 167134
    }
  },
  {
    "ph": "f", "id": 167134, "pid": 76337, "tid": -914061504, "ts": 1716454224048266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224091787, "dur": 106,
    "args": {
      "External id": 167136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167136,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 167136, "pid": 5, "tid": 7, "ts": 1716454224091787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048272, "dur": 13,
    "args": {
      "External id": 167136,
      "cbid": 211, "correlation": 167136
    }
  },
  {
    "ph": "s", "id": 167136, "pid": 76337, "tid": -914061504, "ts": 1716454224048272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224091894, "dur": 35,
    "args": {
      "External id": 167144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167144,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167144, "pid": 5, "tid": 7, "ts": 1716454224091894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048342, "dur": 12,
    "args": {
      "External id": 167144,
      "cbid": 211, "correlation": 167144
    }
  },
  {
    "ph": "s", "id": 167144, "pid": 76337, "tid": -914061504, "ts": 1716454224048342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224091930, "dur": 67,
    "args": {
      "External id": 167152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167152,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167152, "pid": 5, "tid": 7, "ts": 1716454224091930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048382, "dur": 9,
    "args": {
      "External id": 167152,
      "cbid": 211, "correlation": 167152
    }
  },
  {
    "ph": "s", "id": 167152, "pid": 76337, "tid": -914061504, "ts": 1716454224048382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224091999, "dur": 92,
    "args": {
      "External id": 167174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167174,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167174, "pid": 5, "tid": 7, "ts": 1716454224091999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048433, "dur": 10,
    "args": {
      "External id": 167174,
      "cbid": 211, "correlation": 167174
    }
  },
  {
    "ph": "s", "id": 167174, "pid": 76337, "tid": -914061504, "ts": 1716454224048433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048520, "dur": 1,
    "args": {
      "External id": 167190,
      "cbid": 251, "correlation": 167190
    }
  },
  {
    "ph": "f", "id": 167190, "pid": 76337, "tid": -914061504, "ts": 1716454224048520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224092092, "dur": 577,
    "args": {
      "External id": 167192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167192,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167192, "pid": 5, "tid": 7, "ts": 1716454224092092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048525, "dur": 12,
    "args": {
      "External id": 167192,
      "cbid": 211, "correlation": 167192
    }
  },
  {
    "ph": "s", "id": 167192, "pid": 76337, "tid": -914061504, "ts": 1716454224048525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224092671, "dur": 243,
    "args": {
      "External id": 167200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167200, "pid": 5, "tid": 7, "ts": 1716454224092671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048590, "dur": 13,
    "args": {
      "External id": 167200,
      "cbid": 211, "correlation": 167200
    }
  },
  {
    "ph": "s", "id": 167200, "pid": 76337, "tid": -914061504, "ts": 1716454224048590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224092916, "dur": 249,
    "args": {
      "External id": 167208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167208,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167208, "pid": 5, "tid": 7, "ts": 1716454224092916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048620, "dur": 8,
    "args": {
      "External id": 167208,
      "cbid": 211, "correlation": 167208
    }
  },
  {
    "ph": "s", "id": 167208, "pid": 76337, "tid": -914061504, "ts": 1716454224048620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048702, "dur": 1,
    "args": {
      "External id": 167224,
      "cbid": 251, "correlation": 167224
    }
  },
  {
    "ph": "f", "id": 167224, "pid": 76337, "tid": -914061504, "ts": 1716454224048702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048707, "dur": 0,
    "args": {
      "External id": 167226,
      "cbid": 251, "correlation": 167226
    }
  },
  {
    "ph": "f", "id": 167226, "pid": 76337, "tid": -914061504, "ts": 1716454224048707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224093166, "dur": 362,
    "args": {
      "External id": 167227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167227,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 167227, "pid": 5, "tid": 7, "ts": 1716454224093166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048710, "dur": 13,
    "args": {
      "External id": 167227,
      "cbid": 211, "correlation": 167227
    }
  },
  {
    "ph": "s", "id": 167227, "pid": 76337, "tid": -914061504, "ts": 1716454224048710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224093529, "dur": 50,
    "args": {
      "External id": 167235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167235,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167235, "pid": 5, "tid": 7, "ts": 1716454224093529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048752, "dur": 10,
    "args": {
      "External id": 167235,
      "cbid": 211, "correlation": 167235
    }
  },
  {
    "ph": "s", "id": 167235, "pid": 76337, "tid": -914061504, "ts": 1716454224048752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224093580, "dur": 158,
    "args": {
      "External id": 167246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167246, "pid": 5, "tid": 7, "ts": 1716454224093580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048819, "dur": 12,
    "args": {
      "External id": 167246,
      "cbid": 211, "correlation": 167246
    }
  },
  {
    "ph": "s", "id": 167246, "pid": 76337, "tid": -914061504, "ts": 1716454224048819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048881, "dur": 0,
    "args": {
      "External id": 167258,
      "cbid": 317, "correlation": 167258
    }
  },
  {
    "ph": "f", "id": 167258, "pid": 76337, "tid": -914061504, "ts": 1716454224048881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048882, "dur": 0,
    "args": {
      "External id": 167259,
      "cbid": 203, "correlation": 167259
    }
  },
  {
    "ph": "f", "id": 167259, "pid": 76337, "tid": -914061504, "ts": 1716454224048882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048883, "dur": 0,
    "args": {
      "External id": 167260,
      "cbid": 205, "correlation": 167260
    }
  },
  {
    "ph": "f", "id": 167260, "pid": 76337, "tid": -914061504, "ts": 1716454224048883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048906, "dur": 1,
    "args": {
      "External id": 167264,
      "cbid": 251, "correlation": 167264
    }
  },
  {
    "ph": "f", "id": 167264, "pid": 76337, "tid": -914061504, "ts": 1716454224048906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048908, "dur": 0,
    "args": {
      "External id": 167265,
      "cbid": 251, "correlation": 167265
    }
  },
  {
    "ph": "f", "id": 167265, "pid": 76337, "tid": -914061504, "ts": 1716454224048908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048908, "dur": 0,
    "args": {
      "External id": 167266,
      "cbid": 251, "correlation": 167266
    }
  },
  {
    "ph": "f", "id": 167266, "pid": 76337, "tid": -914061504, "ts": 1716454224048908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048909, "dur": 0,
    "args": {
      "External id": 167267,
      "cbid": 251, "correlation": 167267
    }
  },
  {
    "ph": "f", "id": 167267, "pid": 76337, "tid": -914061504, "ts": 1716454224048909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048910, "dur": 0,
    "args": {
      "External id": 167268,
      "cbid": 251, "correlation": 167268
    }
  },
  {
    "ph": "f", "id": 167268, "pid": 76337, "tid": -914061504, "ts": 1716454224048910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048911, "dur": 0,
    "args": {
      "External id": 167269,
      "cbid": 251, "correlation": 167269
    }
  },
  {
    "ph": "f", "id": 167269, "pid": 76337, "tid": -914061504, "ts": 1716454224048911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048912, "dur": 0,
    "args": {
      "External id": 167270,
      "cbid": 251, "correlation": 167270
    }
  },
  {
    "ph": "f", "id": 167270, "pid": 76337, "tid": -914061504, "ts": 1716454224048912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048912, "dur": 0,
    "args": {
      "External id": 167271,
      "cbid": 251, "correlation": 167271
    }
  },
  {
    "ph": "f", "id": 167271, "pid": 76337, "tid": -914061504, "ts": 1716454224048912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048914, "dur": 0,
    "args": {
      "External id": 167272,
      "cbid": 251, "correlation": 167272
    }
  },
  {
    "ph": "f", "id": 167272, "pid": 76337, "tid": -914061504, "ts": 1716454224048914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224093740, "dur": 115,
    "args": {
      "External id": 167273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167273,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 167273, "pid": 5, "tid": 7, "ts": 1716454224093740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048916, "dur": 12,
    "args": {
      "External id": 167273,
      "cbid": 211, "correlation": 167273
    }
  },
  {
    "ph": "s", "id": 167273, "pid": 76337, "tid": -914061504, "ts": 1716454224048916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224093856, "dur": 60,
    "args": {
      "External id": 167279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167279, "pid": 5, "tid": 7, "ts": 1716454224093856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048950, "dur": 9,
    "args": {
      "External id": 167279,
      "cbid": 211, "correlation": 167279
    }
  },
  {
    "ph": "s", "id": 167279, "pid": 76337, "tid": -914061504, "ts": 1716454224048950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224093917, "dur": 50,
    "args": {
      "External id": 167287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167287,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167287, "pid": 5, "tid": 7, "ts": 1716454224093917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224048989, "dur": 9,
    "args": {
      "External id": 167287,
      "cbid": 211, "correlation": 167287
    }
  },
  {
    "ph": "s", "id": 167287, "pid": 76337, "tid": -914061504, "ts": 1716454224048989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049064, "dur": 0,
    "args": {
      "External id": 167297,
      "cbid": 317, "correlation": 167297
    }
  },
  {
    "ph": "f", "id": 167297, "pid": 76337, "tid": -914061504, "ts": 1716454224049064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049065, "dur": 0,
    "args": {
      "External id": 167298,
      "cbid": 203, "correlation": 167298
    }
  },
  {
    "ph": "f", "id": 167298, "pid": 76337, "tid": -914061504, "ts": 1716454224049065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049065, "dur": 0,
    "args": {
      "External id": 167299,
      "cbid": 205, "correlation": 167299
    }
  },
  {
    "ph": "f", "id": 167299, "pid": 76337, "tid": -914061504, "ts": 1716454224049065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224093968, "dur": 40,
    "args": {
      "External id": 167303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167303, "pid": 5, "tid": 7, "ts": 1716454224093968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049081, "dur": 13,
    "args": {
      "External id": 167303,
      "cbid": 211, "correlation": 167303
    }
  },
  {
    "ph": "s", "id": 167303, "pid": 76337, "tid": -914061504, "ts": 1716454224049081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224094010, "dur": 15,
    "args": {
      "External id": 167305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167305,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167305, "pid": 5, "tid": 7, "ts": 1716454224094010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049097, "dur": 5,
    "args": {
      "External id": 167305,
      "cbid": 211, "correlation": 167305
    }
  },
  {
    "ph": "s", "id": 167305, "pid": 76337, "tid": -914061504, "ts": 1716454224049097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224094027, "dur": 1,
    "args": {
      "External id": 167307,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 167307,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 167307, "pid": 5, "tid": 7, "ts": 1716454224094027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049115, "dur": 16,
    "args": {
      "External id": 167307,
      "cbid": 51, "correlation": 167307
    }
  },
  {
    "ph": "s", "id": 167307, "pid": 76337, "tid": -914061504, "ts": 1716454224049115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224094030, "dur": 366,
    "args": {
      "External id": 167308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167308,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167308, "pid": 5, "tid": 7, "ts": 1716454224094030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049132, "dur": 10,
    "args": {
      "External id": 167308,
      "cbid": 211, "correlation": 167308
    }
  },
  {
    "ph": "s", "id": 167308, "pid": 76337, "tid": -914061504, "ts": 1716454224049132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224094398, "dur": 13,
    "args": {
      "External id": 167310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167310, "pid": 5, "tid": 7, "ts": 1716454224094398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049150, "dur": 8,
    "args": {
      "External id": 167310,
      "cbid": 211, "correlation": 167310
    }
  },
  {
    "ph": "s", "id": 167310, "pid": 76337, "tid": -914061504, "ts": 1716454224049150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224094412, "dur": 15,
    "args": {
      "External id": 167316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167316, "pid": 5, "tid": 7, "ts": 1716454224094412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049180, "dur": 10,
    "args": {
      "External id": 167316,
      "cbid": 211, "correlation": 167316
    }
  },
  {
    "ph": "s", "id": 167316, "pid": 76337, "tid": -914061504, "ts": 1716454224049180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224094428, "dur": 19,
    "args": {
      "External id": 167336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167336,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 167336, "pid": 5, "tid": 7, "ts": 1716454224094428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049273, "dur": 12,
    "args": {
      "External id": 167336,
      "cbid": 211, "correlation": 167336
    }
  },
  {
    "ph": "s", "id": 167336, "pid": 76337, "tid": -914061504, "ts": 1716454224049273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224094448, "dur": 4,
    "args": {
      "External id": 167348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167348,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 167348, "pid": 5, "tid": 7, "ts": 1716454224094448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049296, "dur": 7,
    "args": {
      "External id": 167348,
      "cbid": 211, "correlation": 167348
    }
  },
  {
    "ph": "s", "id": 167348, "pid": 76337, "tid": -914061504, "ts": 1716454224049296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224094454, "dur": 17,
    "args": {
      "External id": 167351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167351,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167351, "pid": 5, "tid": 7, "ts": 1716454224094454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049314, "dur": 7,
    "args": {
      "External id": 167351,
      "cbid": 211, "correlation": 167351
    }
  },
  {
    "ph": "s", "id": 167351, "pid": 76337, "tid": -914061504, "ts": 1716454224049314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224094472, "dur": 12,
    "args": {
      "External id": 167360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167360,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167360, "pid": 5, "tid": 7, "ts": 1716454224094472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049354, "dur": 10,
    "args": {
      "External id": 167360,
      "cbid": 211, "correlation": 167360
    }
  },
  {
    "ph": "s", "id": 167360, "pid": 76337, "tid": -914061504, "ts": 1716454224049354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049410, "dur": 0,
    "args": {
      "External id": 167370,
      "cbid": 317, "correlation": 167370
    }
  },
  {
    "ph": "f", "id": 167370, "pid": 76337, "tid": -914061504, "ts": 1716454224049410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049411, "dur": 0,
    "args": {
      "External id": 167371,
      "cbid": 203, "correlation": 167371
    }
  },
  {
    "ph": "f", "id": 167371, "pid": 76337, "tid": -914061504, "ts": 1716454224049411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049412, "dur": 0,
    "args": {
      "External id": 167372,
      "cbid": 205, "correlation": 167372
    }
  },
  {
    "ph": "f", "id": 167372, "pid": 76337, "tid": -914061504, "ts": 1716454224049412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224094485, "dur": 12,
    "args": {
      "External id": 167376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167376,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167376, "pid": 5, "tid": 7, "ts": 1716454224094485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049426, "dur": 12,
    "args": {
      "External id": 167376,
      "cbid": 211, "correlation": 167376
    }
  },
  {
    "ph": "s", "id": 167376, "pid": 76337, "tid": -914061504, "ts": 1716454224049426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224094499, "dur": 25,
    "args": {
      "External id": 167378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167378, "pid": 5, "tid": 7, "ts": 1716454224094499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049440, "dur": 6,
    "args": {
      "External id": 167378,
      "cbid": 211, "correlation": 167378
    }
  },
  {
    "ph": "s", "id": 167378, "pid": 76337, "tid": -914061504, "ts": 1716454224049440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224094525, "dur": 4,
    "args": {
      "External id": 167380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 167380, "pid": 5, "tid": 7, "ts": 1716454224094525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049452, "dur": 5,
    "args": {
      "External id": 167380,
      "cbid": 211, "correlation": 167380
    }
  },
  {
    "ph": "s", "id": 167380, "pid": 76337, "tid": -914061504, "ts": 1716454224049452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049461, "dur": 0,
    "args": {
      "External id": 167381,
      "cbid": 51, "correlation": 167381
    }
  },
  {
    "ph": "s", "id": 167381, "pid": 76337, "tid": -914061504, "ts": 1716454224049461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224094530, "dur": 358,
    "args": {
      "External id": 167382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167382,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167382, "pid": 5, "tid": 7, "ts": 1716454224094530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049462, "dur": 7,
    "args": {
      "External id": 167382,
      "cbid": 211, "correlation": 167382
    }
  },
  {
    "ph": "s", "id": 167382, "pid": 76337, "tid": -914061504, "ts": 1716454224049462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224094889, "dur": 20,
    "args": {
      "External id": 167383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167383,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167383, "pid": 5, "tid": 7, "ts": 1716454224094889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049472, "dur": 5,
    "args": {
      "External id": 167383,
      "cbid": 211, "correlation": 167383
    }
  },
  {
    "ph": "s", "id": 167383, "pid": 76337, "tid": -914061504, "ts": 1716454224049472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224094911, "dur": 32,
    "args": {
      "External id": 167389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167389, "pid": 5, "tid": 7, "ts": 1716454224094911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049500, "dur": 8,
    "args": {
      "External id": 167389,
      "cbid": 211, "correlation": 167389
    }
  },
  {
    "ph": "s", "id": 167389, "pid": 76337, "tid": -914061504, "ts": 1716454224049500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224094944, "dur": 4,
    "args": {
      "External id": 167397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167397,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 167397, "pid": 5, "tid": 7, "ts": 1716454224094944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049543, "dur": 10,
    "args": {
      "External id": 167397,
      "cbid": 211, "correlation": 167397
    }
  },
  {
    "ph": "s", "id": 167397, "pid": 76337, "tid": -914061504, "ts": 1716454224049543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049609, "dur": 1,
    "args": {
      "External id": 167413,
      "cbid": 251, "correlation": 167413
    }
  },
  {
    "ph": "f", "id": 167413, "pid": 76337, "tid": -914061504, "ts": 1716454224049609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049614, "dur": 0,
    "args": {
      "External id": 167415,
      "cbid": 251, "correlation": 167415
    }
  },
  {
    "ph": "f", "id": 167415, "pid": 76337, "tid": -914061504, "ts": 1716454224049614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224094950, "dur": 13,
    "args": {
      "External id": 167416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167416,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 167416, "pid": 5, "tid": 7, "ts": 1716454224094950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049616, "dur": 12,
    "args": {
      "External id": 167416,
      "cbid": 211, "correlation": 167416
    }
  },
  {
    "ph": "s", "id": 167416, "pid": 76337, "tid": -914061504, "ts": 1716454224049616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224094964, "dur": 5,
    "args": {
      "External id": 167418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167418,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 167418, "pid": 5, "tid": 7, "ts": 1716454224094964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049630, "dur": 5,
    "args": {
      "External id": 167418,
      "cbid": 211, "correlation": 167418
    }
  },
  {
    "ph": "s", "id": 167418, "pid": 76337, "tid": -914061504, "ts": 1716454224049630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224094970, "dur": 29,
    "args": {
      "External id": 167428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167428, "pid": 5, "tid": 7, "ts": 1716454224094970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049689, "dur": 13,
    "args": {
      "External id": 167428,
      "cbid": 211, "correlation": 167428
    }
  },
  {
    "ph": "s", "id": 167428, "pid": 76337, "tid": -914061504, "ts": 1716454224049689,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224095000, "dur": 30,
    "args": {
      "External id": 167448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167448,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 167448, "pid": 5, "tid": 7, "ts": 1716454224095000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049758, "dur": 11,
    "args": {
      "External id": 167448,
      "cbid": 211, "correlation": 167448
    }
  },
  {
    "ph": "s", "id": 167448, "pid": 76337, "tid": -914061504, "ts": 1716454224049758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224095031, "dur": 4,
    "args": {
      "External id": 167460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167460,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 167460, "pid": 5, "tid": 7, "ts": 1716454224095031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049779, "dur": 6,
    "args": {
      "External id": 167460,
      "cbid": 211, "correlation": 167460
    }
  },
  {
    "ph": "s", "id": 167460, "pid": 76337, "tid": -914061504, "ts": 1716454224049779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224095037, "dur": 30,
    "args": {
      "External id": 167463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167463, "pid": 5, "tid": 7, "ts": 1716454224095037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049799, "dur": 7,
    "args": {
      "External id": 167463,
      "cbid": 211, "correlation": 167463
    }
  },
  {
    "ph": "s", "id": 167463, "pid": 76337, "tid": -914061504, "ts": 1716454224049799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224095068, "dur": 21,
    "args": {
      "External id": 167472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167472,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167472, "pid": 5, "tid": 7, "ts": 1716454224095068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049840, "dur": 11,
    "args": {
      "External id": 167472,
      "cbid": 211, "correlation": 167472
    }
  },
  {
    "ph": "s", "id": 167472, "pid": 76337, "tid": -914061504, "ts": 1716454224049840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049910, "dur": 0,
    "args": {
      "External id": 167482,
      "cbid": 317, "correlation": 167482
    }
  },
  {
    "ph": "f", "id": 167482, "pid": 76337, "tid": -914061504, "ts": 1716454224049910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049911, "dur": 0,
    "args": {
      "External id": 167483,
      "cbid": 203, "correlation": 167483
    }
  },
  {
    "ph": "f", "id": 167483, "pid": 76337, "tid": -914061504, "ts": 1716454224049911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049912, "dur": 0,
    "args": {
      "External id": 167484,
      "cbid": 205, "correlation": 167484
    }
  },
  {
    "ph": "f", "id": 167484, "pid": 76337, "tid": -914061504, "ts": 1716454224049912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224095090, "dur": 25,
    "args": {
      "External id": 167488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167488,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167488, "pid": 5, "tid": 7, "ts": 1716454224095090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049931, "dur": 12,
    "args": {
      "External id": 167488,
      "cbid": 211, "correlation": 167488
    }
  },
  {
    "ph": "s", "id": 167488, "pid": 76337, "tid": -914061504, "ts": 1716454224049931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224095117, "dur": 45,
    "args": {
      "External id": 167490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167490, "pid": 5, "tid": 7, "ts": 1716454224095117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049946, "dur": 6,
    "args": {
      "External id": 167490,
      "cbid": 211, "correlation": 167490
    }
  },
  {
    "ph": "s", "id": 167490, "pid": 76337, "tid": -914061504, "ts": 1716454224049946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224095163, "dur": 656,
    "args": {
      "External id": 167492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167492,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167492, "pid": 5, "tid": 7, "ts": 1716454224095163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049963, "dur": 10,
    "args": {
      "External id": 167492,
      "cbid": 211, "correlation": 167492
    }
  },
  {
    "ph": "s", "id": 167492, "pid": 76337, "tid": -914061504, "ts": 1716454224049963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224095820, "dur": 21,
    "args": {
      "External id": 167494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167494,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167494, "pid": 5, "tid": 7, "ts": 1716454224095820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224049985, "dur": 6,
    "args": {
      "External id": 167494,
      "cbid": 211, "correlation": 167494
    }
  },
  {
    "ph": "s", "id": 167494, "pid": 76337, "tid": -914061504, "ts": 1716454224049985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224095842, "dur": 32,
    "args": {
      "External id": 167500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167500, "pid": 5, "tid": 7, "ts": 1716454224095842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050015, "dur": 9,
    "args": {
      "External id": 167500,
      "cbid": 211, "correlation": 167500
    }
  },
  {
    "ph": "s", "id": 167500, "pid": 76337, "tid": -914061504, "ts": 1716454224050015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050075, "dur": 0,
    "args": {
      "External id": 167510,
      "cbid": 317, "correlation": 167510
    }
  },
  {
    "ph": "f", "id": 167510, "pid": 76337, "tid": -914061504, "ts": 1716454224050075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050076, "dur": 0,
    "args": {
      "External id": 167511,
      "cbid": 203, "correlation": 167511
    }
  },
  {
    "ph": "f", "id": 167511, "pid": 76337, "tid": -914061504, "ts": 1716454224050076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050077, "dur": 0,
    "args": {
      "External id": 167512,
      "cbid": 205, "correlation": 167512
    }
  },
  {
    "ph": "f", "id": 167512, "pid": 76337, "tid": -914061504, "ts": 1716454224050077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050103, "dur": 1,
    "args": {
      "External id": 167516,
      "cbid": 251, "correlation": 167516
    }
  },
  {
    "ph": "f", "id": 167516, "pid": 76337, "tid": -914061504, "ts": 1716454224050103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050105, "dur": 0,
    "args": {
      "External id": 167517,
      "cbid": 251, "correlation": 167517
    }
  },
  {
    "ph": "f", "id": 167517, "pid": 76337, "tid": -914061504, "ts": 1716454224050105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050106, "dur": 0,
    "args": {
      "External id": 167518,
      "cbid": 251, "correlation": 167518
    }
  },
  {
    "ph": "f", "id": 167518, "pid": 76337, "tid": -914061504, "ts": 1716454224050106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050107, "dur": 0,
    "args": {
      "External id": 167519,
      "cbid": 251, "correlation": 167519
    }
  },
  {
    "ph": "f", "id": 167519, "pid": 76337, "tid": -914061504, "ts": 1716454224050107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050108, "dur": 0,
    "args": {
      "External id": 167520,
      "cbid": 251, "correlation": 167520
    }
  },
  {
    "ph": "f", "id": 167520, "pid": 76337, "tid": -914061504, "ts": 1716454224050108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050109, "dur": 0,
    "args": {
      "External id": 167521,
      "cbid": 251, "correlation": 167521
    }
  },
  {
    "ph": "f", "id": 167521, "pid": 76337, "tid": -914061504, "ts": 1716454224050109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050110, "dur": 0,
    "args": {
      "External id": 167522,
      "cbid": 251, "correlation": 167522
    }
  },
  {
    "ph": "f", "id": 167522, "pid": 76337, "tid": -914061504, "ts": 1716454224050110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050111, "dur": 1,
    "args": {
      "External id": 167523,
      "cbid": 251, "correlation": 167523
    }
  },
  {
    "ph": "f", "id": 167523, "pid": 76337, "tid": -914061504, "ts": 1716454224050111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050113, "dur": 0,
    "args": {
      "External id": 167524,
      "cbid": 251, "correlation": 167524
    }
  },
  {
    "ph": "f", "id": 167524, "pid": 76337, "tid": -914061504, "ts": 1716454224050113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224095876, "dur": 51,
    "args": {
      "External id": 167525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167525,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 167525, "pid": 5, "tid": 7, "ts": 1716454224095876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050116, "dur": 13,
    "args": {
      "External id": 167525,
      "cbid": 211, "correlation": 167525
    }
  },
  {
    "ph": "s", "id": 167525, "pid": 76337, "tid": -914061504, "ts": 1716454224050116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224095928, "dur": 32,
    "args": {
      "External id": 167531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167531, "pid": 5, "tid": 7, "ts": 1716454224095928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050150, "dur": 8,
    "args": {
      "External id": 167531,
      "cbid": 211, "correlation": 167531
    }
  },
  {
    "ph": "s", "id": 167531, "pid": 76337, "tid": -914061504, "ts": 1716454224050150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224095962, "dur": 27,
    "args": {
      "External id": 167539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167539,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167539, "pid": 5, "tid": 7, "ts": 1716454224095962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050179, "dur": 8,
    "args": {
      "External id": 167539,
      "cbid": 211, "correlation": 167539
    }
  },
  {
    "ph": "s", "id": 167539, "pid": 76337, "tid": -914061504, "ts": 1716454224050179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224095990, "dur": 19,
    "args": {
      "External id": 167547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167547,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167547, "pid": 5, "tid": 7, "ts": 1716454224095990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050213, "dur": 9,
    "args": {
      "External id": 167547,
      "cbid": 211, "correlation": 167547
    }
  },
  {
    "ph": "s", "id": 167547, "pid": 76337, "tid": -914061504, "ts": 1716454224050213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224096010, "dur": 31,
    "args": {
      "External id": 167567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167567,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 167567, "pid": 5, "tid": 7, "ts": 1716454224096010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050303, "dur": 12,
    "args": {
      "External id": 167567,
      "cbid": 211, "correlation": 167567
    }
  },
  {
    "ph": "s", "id": 167567, "pid": 76337, "tid": -914061504, "ts": 1716454224050303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224096042, "dur": 4,
    "args": {
      "External id": 167579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167579,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 167579, "pid": 5, "tid": 7, "ts": 1716454224096042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050325, "dur": 6,
    "args": {
      "External id": 167579,
      "cbid": 211, "correlation": 167579
    }
  },
  {
    "ph": "s", "id": 167579, "pid": 76337, "tid": -914061504, "ts": 1716454224050325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224096048, "dur": 30,
    "args": {
      "External id": 167582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167582, "pid": 5, "tid": 7, "ts": 1716454224096048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050343, "dur": 7,
    "args": {
      "External id": 167582,
      "cbid": 211, "correlation": 167582
    }
  },
  {
    "ph": "s", "id": 167582, "pid": 76337, "tid": -914061504, "ts": 1716454224050343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050401, "dur": 0,
    "args": {
      "External id": 167593,
      "cbid": 317, "correlation": 167593
    }
  },
  {
    "ph": "f", "id": 167593, "pid": 76337, "tid": -914061504, "ts": 1716454224050401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050402, "dur": 0,
    "args": {
      "External id": 167594,
      "cbid": 203, "correlation": 167594
    }
  },
  {
    "ph": "f", "id": 167594, "pid": 76337, "tid": -914061504, "ts": 1716454224050402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050403, "dur": 0,
    "args": {
      "External id": 167595,
      "cbid": 205, "correlation": 167595
    }
  },
  {
    "ph": "f", "id": 167595, "pid": 76337, "tid": -914061504, "ts": 1716454224050403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224096079, "dur": 22,
    "args": {
      "External id": 167599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167599,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167599, "pid": 5, "tid": 7, "ts": 1716454224096079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050417, "dur": 11,
    "args": {
      "External id": 167599,
      "cbid": 211, "correlation": 167599
    }
  },
  {
    "ph": "s", "id": 167599, "pid": 76337, "tid": -914061504, "ts": 1716454224050417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224096103, "dur": 121,
    "args": {
      "External id": 167601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167601,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167601, "pid": 5, "tid": 7, "ts": 1716454224096103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050437, "dur": 10,
    "args": {
      "External id": 167601,
      "cbid": 211, "correlation": 167601
    }
  },
  {
    "ph": "s", "id": 167601, "pid": 76337, "tid": -914061504, "ts": 1716454224050437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224096225, "dur": 22,
    "args": {
      "External id": 167603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167603,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167603, "pid": 5, "tid": 7, "ts": 1716454224096225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050451, "dur": 5,
    "args": {
      "External id": 167603,
      "cbid": 211, "correlation": 167603
    }
  },
  {
    "ph": "s", "id": 167603, "pid": 76337, "tid": -914061504, "ts": 1716454224050451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224096248, "dur": 32,
    "args": {
      "External id": 167609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167609,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167609, "pid": 5, "tid": 7, "ts": 1716454224096248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050478, "dur": 8,
    "args": {
      "External id": 167609,
      "cbid": 211, "correlation": 167609
    }
  },
  {
    "ph": "s", "id": 167609, "pid": 76337, "tid": -914061504, "ts": 1716454224050478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224096282, "dur": 178,
    "args": {
      "External id": 167618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167618,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167618, "pid": 5, "tid": 7, "ts": 1716454224096282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050572, "dur": 16,
    "args": {
      "External id": 167618,
      "cbid": 211, "correlation": 167618
    }
  },
  {
    "ph": "s", "id": 167618, "pid": 76337, "tid": -914061504, "ts": 1716454224050572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224096462, "dur": 66,
    "args": {
      "External id": 167640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167640,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167640, "pid": 5, "tid": 7, "ts": 1716454224096462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050636, "dur": 12,
    "args": {
      "External id": 167640,
      "cbid": 211, "correlation": 167640
    }
  },
  {
    "ph": "s", "id": 167640, "pid": 76337, "tid": -914061504, "ts": 1716454224050636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050741, "dur": 2,
    "args": {
      "External id": 167651,
      "cbid": 251, "correlation": 167651
    }
  },
  {
    "ph": "f", "id": 167651, "pid": 76337, "tid": -914061504, "ts": 1716454224050741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224096529, "dur": 159,
    "args": {
      "External id": 167652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167652,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167652, "pid": 5, "tid": 7, "ts": 1716454224096529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050747, "dur": 13,
    "args": {
      "External id": 167652,
      "cbid": 211, "correlation": 167652
    }
  },
  {
    "ph": "s", "id": 167652, "pid": 76337, "tid": -914061504, "ts": 1716454224050747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050817, "dur": 1,
    "args": {
      "External id": 167663,
      "cbid": 251, "correlation": 167663
    }
  },
  {
    "ph": "f", "id": 167663, "pid": 76337, "tid": -914061504, "ts": 1716454224050817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224096689, "dur": 147,
    "args": {
      "External id": 167664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167664,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167664, "pid": 5, "tid": 7, "ts": 1716454224096689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050821, "dur": 11,
    "args": {
      "External id": 167664,
      "cbid": 211, "correlation": 167664
    }
  },
  {
    "ph": "s", "id": 167664, "pid": 76337, "tid": -914061504, "ts": 1716454224050821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050885, "dur": 1,
    "args": {
      "External id": 167675,
      "cbid": 251, "correlation": 167675
    }
  },
  {
    "ph": "f", "id": 167675, "pid": 76337, "tid": -914061504, "ts": 1716454224050885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224096837, "dur": 144,
    "args": {
      "External id": 167676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167676,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167676, "pid": 5, "tid": 7, "ts": 1716454224096837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050889, "dur": 12,
    "args": {
      "External id": 167676,
      "cbid": 211, "correlation": 167676
    }
  },
  {
    "ph": "s", "id": 167676, "pid": 76337, "tid": -914061504, "ts": 1716454224050889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224096983, "dur": 1963,
    "args": {
      "External id": 167697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167697,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 167697, "pid": 5, "tid": 7, "ts": 1716454224096983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224050997, "dur": 14,
    "args": {
      "External id": 167697,
      "cbid": 211, "correlation": 167697
    }
  },
  {
    "ph": "s", "id": 167697, "pid": 76337, "tid": -914061504, "ts": 1716454224050997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051111, "dur": 1,
    "args": {
      "External id": 167715,
      "cbid": 251, "correlation": 167715
    }
  },
  {
    "ph": "f", "id": 167715, "pid": 76337, "tid": -914061504, "ts": 1716454224051111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224098947, "dur": 149,
    "args": {
      "External id": 167717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167717,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 167717, "pid": 5, "tid": 7, "ts": 1716454224098947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051117, "dur": 13,
    "args": {
      "External id": 167717,
      "cbid": 211, "correlation": 167717
    }
  },
  {
    "ph": "s", "id": 167717, "pid": 76337, "tid": -914061504, "ts": 1716454224051117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224099097, "dur": 35,
    "args": {
      "External id": 167725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167725,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167725, "pid": 5, "tid": 7, "ts": 1716454224099097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051190, "dur": 12,
    "args": {
      "External id": 167725,
      "cbid": 211, "correlation": 167725
    }
  },
  {
    "ph": "s", "id": 167725, "pid": 76337, "tid": -914061504, "ts": 1716454224051190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224099134, "dur": 50,
    "args": {
      "External id": 167733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167733,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167733, "pid": 5, "tid": 7, "ts": 1716454224099134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051229, "dur": 9,
    "args": {
      "External id": 167733,
      "cbid": 211, "correlation": 167733
    }
  },
  {
    "ph": "s", "id": 167733, "pid": 76337, "tid": -914061504, "ts": 1716454224051229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224099185, "dur": 29,
    "args": {
      "External id": 167744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167744, "pid": 5, "tid": 7, "ts": 1716454224099185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051310, "dur": 13,
    "args": {
      "External id": 167744,
      "cbid": 211, "correlation": 167744
    }
  },
  {
    "ph": "s", "id": 167744, "pid": 76337, "tid": -914061504, "ts": 1716454224051310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224099216, "dur": 35,
    "args": {
      "External id": 167766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167766,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167766, "pid": 5, "tid": 7, "ts": 1716454224099216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051343, "dur": 7,
    "args": {
      "External id": 167766,
      "cbid": 211, "correlation": 167766
    }
  },
  {
    "ph": "s", "id": 167766, "pid": 76337, "tid": -914061504, "ts": 1716454224051343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051427, "dur": 1,
    "args": {
      "External id": 167777,
      "cbid": 251, "correlation": 167777
    }
  },
  {
    "ph": "f", "id": 167777, "pid": 76337, "tid": -914061504, "ts": 1716454224051427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224099252, "dur": 90,
    "args": {
      "External id": 167778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167778,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167778, "pid": 5, "tid": 7, "ts": 1716454224099252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051432, "dur": 14,
    "args": {
      "External id": 167778,
      "cbid": 211, "correlation": 167778
    }
  },
  {
    "ph": "s", "id": 167778, "pid": 76337, "tid": -914061504, "ts": 1716454224051432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051505, "dur": 1,
    "args": {
      "External id": 167789,
      "cbid": 251, "correlation": 167789
    }
  },
  {
    "ph": "f", "id": 167789, "pid": 76337, "tid": -914061504, "ts": 1716454224051505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051509, "dur": 0,
    "args": {
      "External id": 167790,
      "cbid": 251, "correlation": 167790
    }
  },
  {
    "ph": "f", "id": 167790, "pid": 76337, "tid": -914061504, "ts": 1716454224051509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224099343, "dur": 12,
    "args": {
      "External id": 167791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167791,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 167791, "pid": 5, "tid": 7, "ts": 1716454224099343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051511, "dur": 12,
    "args": {
      "External id": 167791,
      "cbid": 211, "correlation": 167791
    }
  },
  {
    "ph": "s", "id": 167791, "pid": 76337, "tid": -914061504, "ts": 1716454224051511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224099357, "dur": 5,
    "args": {
      "External id": 167793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167793,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 167793, "pid": 5, "tid": 7, "ts": 1716454224099357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051526, "dur": 8,
    "args": {
      "External id": 167793,
      "cbid": 211, "correlation": 167793
    }
  },
  {
    "ph": "s", "id": 167793, "pid": 76337, "tid": -914061504, "ts": 1716454224051526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051587, "dur": 1,
    "args": {
      "External id": 167804,
      "cbid": 251, "correlation": 167804
    }
  },
  {
    "ph": "f", "id": 167804, "pid": 76337, "tid": -914061504, "ts": 1716454224051587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051590, "dur": 0,
    "args": {
      "External id": 167805,
      "cbid": 251, "correlation": 167805
    }
  },
  {
    "ph": "f", "id": 167805, "pid": 76337, "tid": -914061504, "ts": 1716454224051590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224099363, "dur": 7,
    "args": {
      "External id": 167806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167806,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 167806, "pid": 5, "tid": 7, "ts": 1716454224099363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051591, "dur": 11,
    "args": {
      "External id": 167806,
      "cbid": 211, "correlation": 167806
    }
  },
  {
    "ph": "s", "id": 167806, "pid": 76337, "tid": -914061504, "ts": 1716454224051591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224099371, "dur": 3,
    "args": {
      "External id": 167808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167808,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 167808, "pid": 5, "tid": 7, "ts": 1716454224099371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051604, "dur": 5,
    "args": {
      "External id": 167808,
      "cbid": 211, "correlation": 167808
    }
  },
  {
    "ph": "s", "id": 167808, "pid": 76337, "tid": -914061504, "ts": 1716454224051604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224099376, "dur": 92,
    "args": {
      "External id": 167829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167829,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 167829, "pid": 5, "tid": 7, "ts": 1716454224099376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051677, "dur": 13,
    "args": {
      "External id": 167829,
      "cbid": 211, "correlation": 167829
    }
  },
  {
    "ph": "s", "id": 167829, "pid": 76337, "tid": -914061504, "ts": 1716454224051677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051774, "dur": 1,
    "args": {
      "External id": 167847,
      "cbid": 251, "correlation": 167847
    }
  },
  {
    "ph": "f", "id": 167847, "pid": 76337, "tid": -914061504, "ts": 1716454224051774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224099469, "dur": 102,
    "args": {
      "External id": 167849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167849,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167849, "pid": 5, "tid": 7, "ts": 1716454224099469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051780, "dur": 14,
    "args": {
      "External id": 167849,
      "cbid": 211, "correlation": 167849
    }
  },
  {
    "ph": "s", "id": 167849, "pid": 76337, "tid": -914061504, "ts": 1716454224051780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224099573, "dur": 19,
    "args": {
      "External id": 167857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167857,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167857, "pid": 5, "tid": 7, "ts": 1716454224099573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051850, "dur": 12,
    "args": {
      "External id": 167857,
      "cbid": 211, "correlation": 167857
    }
  },
  {
    "ph": "s", "id": 167857, "pid": 76337, "tid": -914061504, "ts": 1716454224051850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224099593, "dur": 38,
    "args": {
      "External id": 167865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167865,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167865, "pid": 5, "tid": 7, "ts": 1716454224099593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051892, "dur": 9,
    "args": {
      "External id": 167865,
      "cbid": 211, "correlation": 167865
    }
  },
  {
    "ph": "s", "id": 167865, "pid": 76337, "tid": -914061504, "ts": 1716454224051892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224099631, "dur": 35,
    "args": {
      "External id": 167887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167887,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167887, "pid": 5, "tid": 7, "ts": 1716454224099631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224051945, "dur": 10,
    "args": {
      "External id": 167887,
      "cbid": 211, "correlation": 167887
    }
  },
  {
    "ph": "s", "id": 167887, "pid": 76337, "tid": -914061504, "ts": 1716454224051945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052046, "dur": 1,
    "args": {
      "External id": 167903,
      "cbid": 251, "correlation": 167903
    }
  },
  {
    "ph": "f", "id": 167903, "pid": 76337, "tid": -914061504, "ts": 1716454224052046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052051, "dur": 0,
    "args": {
      "External id": 167905,
      "cbid": 251, "correlation": 167905
    }
  },
  {
    "ph": "f", "id": 167905, "pid": 76337, "tid": -914061504, "ts": 1716454224052051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224099667, "dur": 544,
    "args": {
      "External id": 167906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167906,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 167906, "pid": 5, "tid": 7, "ts": 1716454224099667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052056, "dur": 13,
    "args": {
      "External id": 167906,
      "cbid": 211, "correlation": 167906
    }
  },
  {
    "ph": "s", "id": 167906, "pid": 76337, "tid": -914061504, "ts": 1716454224052056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224100213, "dur": 126,
    "args": {
      "External id": 167914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167914, "pid": 5, "tid": 7, "ts": 1716454224100213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052129, "dur": 14,
    "args": {
      "External id": 167914,
      "cbid": 211, "correlation": 167914
    }
  },
  {
    "ph": "s", "id": 167914, "pid": 76337, "tid": -914061504, "ts": 1716454224052129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224100341, "dur": 128,
    "args": {
      "External id": 167922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167922, "pid": 5, "tid": 7, "ts": 1716454224100341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052165, "dur": 9,
    "args": {
      "External id": 167922,
      "cbid": 211, "correlation": 167922
    }
  },
  {
    "ph": "s", "id": 167922, "pid": 76337, "tid": -914061504, "ts": 1716454224052165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052244, "dur": 1,
    "args": {
      "External id": 167938,
      "cbid": 251, "correlation": 167938
    }
  },
  {
    "ph": "f", "id": 167938, "pid": 76337, "tid": -914061504, "ts": 1716454224052244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224100470, "dur": 260,
    "args": {
      "External id": 167940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167940,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167940, "pid": 5, "tid": 7, "ts": 1716454224100470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052250, "dur": 12,
    "args": {
      "External id": 167940,
      "cbid": 211, "correlation": 167940
    }
  },
  {
    "ph": "s", "id": 167940, "pid": 76337, "tid": -914061504, "ts": 1716454224052250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224100731, "dur": 27,
    "args": {
      "External id": 167948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167948,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167948, "pid": 5, "tid": 7, "ts": 1716454224100731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052292, "dur": 9,
    "args": {
      "External id": 167948,
      "cbid": 211, "correlation": 167948
    }
  },
  {
    "ph": "s", "id": 167948, "pid": 76337, "tid": -914061504, "ts": 1716454224052292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224100760, "dur": 82,
    "args": {
      "External id": 167959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167959, "pid": 5, "tid": 7, "ts": 1716454224100760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052362, "dur": 12,
    "args": {
      "External id": 167959,
      "cbid": 211, "correlation": 167959
    }
  },
  {
    "ph": "s", "id": 167959, "pid": 76337, "tid": -914061504, "ts": 1716454224052362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052429, "dur": 0,
    "args": {
      "External id": 167971,
      "cbid": 317, "correlation": 167971
    }
  },
  {
    "ph": "f", "id": 167971, "pid": 76337, "tid": -914061504, "ts": 1716454224052429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052430, "dur": 0,
    "args": {
      "External id": 167972,
      "cbid": 203, "correlation": 167972
    }
  },
  {
    "ph": "f", "id": 167972, "pid": 76337, "tid": -914061504, "ts": 1716454224052430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052431, "dur": 0,
    "args": {
      "External id": 167973,
      "cbid": 205, "correlation": 167973
    }
  },
  {
    "ph": "f", "id": 167973, "pid": 76337, "tid": -914061504, "ts": 1716454224052431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224100843, "dur": 23,
    "args": {
      "External id": 167977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167977,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167977, "pid": 5, "tid": 7, "ts": 1716454224100843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052449, "dur": 12,
    "args": {
      "External id": 167977,
      "cbid": 211, "correlation": 167977
    }
  },
  {
    "ph": "s", "id": 167977, "pid": 76337, "tid": -914061504, "ts": 1716454224052449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224100867, "dur": 121,
    "args": {
      "External id": 167979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167979,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 167979, "pid": 5, "tid": 7, "ts": 1716454224100867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052468, "dur": 7,
    "args": {
      "External id": 167979,
      "cbid": 211, "correlation": 167979
    }
  },
  {
    "ph": "s", "id": 167979, "pid": 76337, "tid": -914061504, "ts": 1716454224052468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224100990, "dur": 23,
    "args": {
      "External id": 167981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167981,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167981, "pid": 5, "tid": 7, "ts": 1716454224100990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052479, "dur": 6,
    "args": {
      "External id": 167981,
      "cbid": 211, "correlation": 167981
    }
  },
  {
    "ph": "s", "id": 167981, "pid": 76337, "tid": -914061504, "ts": 1716454224052479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224101015, "dur": 33,
    "args": {
      "External id": 167987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167987, "pid": 5, "tid": 7, "ts": 1716454224101015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052507, "dur": 8,
    "args": {
      "External id": 167987,
      "cbid": 211, "correlation": 167987
    }
  },
  {
    "ph": "s", "id": 167987, "pid": 76337, "tid": -914061504, "ts": 1716454224052507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224101049, "dur": 27,
    "args": {
      "External id": 167995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 167995,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 167995, "pid": 5, "tid": 7, "ts": 1716454224101049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052539, "dur": 8,
    "args": {
      "External id": 167995,
      "cbid": 211, "correlation": 167995
    }
  },
  {
    "ph": "s", "id": 167995, "pid": 76337, "tid": -914061504, "ts": 1716454224052539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224101077, "dur": 30,
    "args": {
      "External id": 168015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168015,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 168015, "pid": 5, "tid": 7, "ts": 1716454224101077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052615, "dur": 12,
    "args": {
      "External id": 168015,
      "cbid": 211, "correlation": 168015
    }
  },
  {
    "ph": "s", "id": 168015, "pid": 76337, "tid": -914061504, "ts": 1716454224052615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224101108, "dur": 4,
    "args": {
      "External id": 168027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168027,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 168027, "pid": 5, "tid": 7, "ts": 1716454224101108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052638, "dur": 6,
    "args": {
      "External id": 168027,
      "cbid": 211, "correlation": 168027
    }
  },
  {
    "ph": "s", "id": 168027, "pid": 76337, "tid": -914061504, "ts": 1716454224052638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224101114, "dur": 30,
    "args": {
      "External id": 168030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168030, "pid": 5, "tid": 7, "ts": 1716454224101114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052656, "dur": 7,
    "args": {
      "External id": 168030,
      "cbid": 211, "correlation": 168030
    }
  },
  {
    "ph": "s", "id": 168030, "pid": 76337, "tid": -914061504, "ts": 1716454224052656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224101146, "dur": 22,
    "args": {
      "External id": 168039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168039,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168039, "pid": 5, "tid": 7, "ts": 1716454224101146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052697, "dur": 10,
    "args": {
      "External id": 168039,
      "cbid": 211, "correlation": 168039
    }
  },
  {
    "ph": "s", "id": 168039, "pid": 76337, "tid": -914061504, "ts": 1716454224052697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052749, "dur": 0,
    "args": {
      "External id": 168049,
      "cbid": 317, "correlation": 168049
    }
  },
  {
    "ph": "f", "id": 168049, "pid": 76337, "tid": -914061504, "ts": 1716454224052749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052750, "dur": 0,
    "args": {
      "External id": 168050,
      "cbid": 203, "correlation": 168050
    }
  },
  {
    "ph": "f", "id": 168050, "pid": 76337, "tid": -914061504, "ts": 1716454224052750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052751, "dur": 0,
    "args": {
      "External id": 168051,
      "cbid": 205, "correlation": 168051
    }
  },
  {
    "ph": "f", "id": 168051, "pid": 76337, "tid": -914061504, "ts": 1716454224052751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224101169, "dur": 21,
    "args": {
      "External id": 168055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168055,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168055, "pid": 5, "tid": 7, "ts": 1716454224101169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052764, "dur": 11,
    "args": {
      "External id": 168055,
      "cbid": 211, "correlation": 168055
    }
  },
  {
    "ph": "s", "id": 168055, "pid": 76337, "tid": -914061504, "ts": 1716454224052764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224101192, "dur": 44,
    "args": {
      "External id": 168057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168057,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168057, "pid": 5, "tid": 7, "ts": 1716454224101192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052777, "dur": 5,
    "args": {
      "External id": 168057,
      "cbid": 211, "correlation": 168057
    }
  },
  {
    "ph": "s", "id": 168057, "pid": 76337, "tid": -914061504, "ts": 1716454224052777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224101237, "dur": 657,
    "args": {
      "External id": 168059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168059,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168059, "pid": 5, "tid": 7, "ts": 1716454224101237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052789, "dur": 7,
    "args": {
      "External id": 168059,
      "cbid": 211, "correlation": 168059
    }
  },
  {
    "ph": "s", "id": 168059, "pid": 76337, "tid": -914061504, "ts": 1716454224052789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224101895, "dur": 21,
    "args": {
      "External id": 168061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168061,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168061, "pid": 5, "tid": 7, "ts": 1716454224101895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052799, "dur": 5,
    "args": {
      "External id": 168061,
      "cbid": 211, "correlation": 168061
    }
  },
  {
    "ph": "s", "id": 168061, "pid": 76337, "tid": -914061504, "ts": 1716454224052799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224101918, "dur": 33,
    "args": {
      "External id": 168067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168067, "pid": 5, "tid": 7, "ts": 1716454224101918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052826, "dur": 9,
    "args": {
      "External id": 168067,
      "cbid": 211, "correlation": 168067
    }
  },
  {
    "ph": "s", "id": 168067, "pid": 76337, "tid": -914061504, "ts": 1716454224052826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224101952, "dur": 4,
    "args": {
      "External id": 168075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168075,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 168075, "pid": 5, "tid": 7, "ts": 1716454224101952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052870, "dur": 9,
    "args": {
      "External id": 168075,
      "cbid": 211, "correlation": 168075
    }
  },
  {
    "ph": "s", "id": 168075, "pid": 76337, "tid": -914061504, "ts": 1716454224052870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052939, "dur": 1,
    "args": {
      "External id": 168091,
      "cbid": 251, "correlation": 168091
    }
  },
  {
    "ph": "f", "id": 168091, "pid": 76337, "tid": -914061504, "ts": 1716454224052939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052944, "dur": 0,
    "args": {
      "External id": 168093,
      "cbid": 251, "correlation": 168093
    }
  },
  {
    "ph": "f", "id": 168093, "pid": 76337, "tid": -914061504, "ts": 1716454224052944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224101958, "dur": 13,
    "args": {
      "External id": 168094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168094,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 168094, "pid": 5, "tid": 7, "ts": 1716454224101958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052946, "dur": 11,
    "args": {
      "External id": 168094,
      "cbid": 211, "correlation": 168094
    }
  },
  {
    "ph": "s", "id": 168094, "pid": 76337, "tid": -914061504, "ts": 1716454224052946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224101972, "dur": 5,
    "args": {
      "External id": 168096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 168096, "pid": 5, "tid": 7, "ts": 1716454224101972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224052959, "dur": 5,
    "args": {
      "External id": 168096,
      "cbid": 211, "correlation": 168096
    }
  },
  {
    "ph": "s", "id": 168096, "pid": 76337, "tid": -914061504, "ts": 1716454224052959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224101978, "dur": 29,
    "args": {
      "External id": 168106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168106, "pid": 5, "tid": 7, "ts": 1716454224101978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053025, "dur": 14,
    "args": {
      "External id": 168106,
      "cbid": 211, "correlation": 168106
    }
  },
  {
    "ph": "s", "id": 168106, "pid": 76337, "tid": -914061504, "ts": 1716454224053025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224102009, "dur": 31,
    "args": {
      "External id": 168126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168126,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 168126, "pid": 5, "tid": 7, "ts": 1716454224102009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053094, "dur": 11,
    "args": {
      "External id": 168126,
      "cbid": 211, "correlation": 168126
    }
  },
  {
    "ph": "s", "id": 168126, "pid": 76337, "tid": -914061504, "ts": 1716454224053094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224102041, "dur": 4,
    "args": {
      "External id": 168138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168138,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 168138, "pid": 5, "tid": 7, "ts": 1716454224102041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053114, "dur": 6,
    "args": {
      "External id": 168138,
      "cbid": 211, "correlation": 168138
    }
  },
  {
    "ph": "s", "id": 168138, "pid": 76337, "tid": -914061504, "ts": 1716454224053114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224102047, "dur": 31,
    "args": {
      "External id": 168141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168141, "pid": 5, "tid": 7, "ts": 1716454224102047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053133, "dur": 7,
    "args": {
      "External id": 168141,
      "cbid": 211, "correlation": 168141
    }
  },
  {
    "ph": "s", "id": 168141, "pid": 76337, "tid": -914061504, "ts": 1716454224053133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224102079, "dur": 20,
    "args": {
      "External id": 168150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168150,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168150, "pid": 5, "tid": 7, "ts": 1716454224102079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053174, "dur": 10,
    "args": {
      "External id": 168150,
      "cbid": 211, "correlation": 168150
    }
  },
  {
    "ph": "s", "id": 168150, "pid": 76337, "tid": -914061504, "ts": 1716454224053174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053237, "dur": 0,
    "args": {
      "External id": 168160,
      "cbid": 317, "correlation": 168160
    }
  },
  {
    "ph": "f", "id": 168160, "pid": 76337, "tid": -914061504, "ts": 1716454224053237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053238, "dur": 0,
    "args": {
      "External id": 168161,
      "cbid": 203, "correlation": 168161
    }
  },
  {
    "ph": "f", "id": 168161, "pid": 76337, "tid": -914061504, "ts": 1716454224053238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053238, "dur": 0,
    "args": {
      "External id": 168162,
      "cbid": 205, "correlation": 168162
    }
  },
  {
    "ph": "f", "id": 168162, "pid": 76337, "tid": -914061504, "ts": 1716454224053238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224102100, "dur": 22,
    "args": {
      "External id": 168166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168166,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168166, "pid": 5, "tid": 7, "ts": 1716454224102100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053252, "dur": 12,
    "args": {
      "External id": 168166,
      "cbid": 211, "correlation": 168166
    }
  },
  {
    "ph": "s", "id": 168166, "pid": 76337, "tid": -914061504, "ts": 1716454224053252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224102123, "dur": 44,
    "args": {
      "External id": 168168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168168, "pid": 5, "tid": 7, "ts": 1716454224102123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053266, "dur": 5,
    "args": {
      "External id": 168168,
      "cbid": 211, "correlation": 168168
    }
  },
  {
    "ph": "s", "id": 168168, "pid": 76337, "tid": -914061504, "ts": 1716454224053266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224102169, "dur": 651,
    "args": {
      "External id": 168170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168170,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168170, "pid": 5, "tid": 7, "ts": 1716454224102169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053277, "dur": 6,
    "args": {
      "External id": 168170,
      "cbid": 211, "correlation": 168170
    }
  },
  {
    "ph": "s", "id": 168170, "pid": 76337, "tid": -914061504, "ts": 1716454224053277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224102821, "dur": 22,
    "args": {
      "External id": 168172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168172,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168172, "pid": 5, "tid": 7, "ts": 1716454224102821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053287, "dur": 5,
    "args": {
      "External id": 168172,
      "cbid": 211, "correlation": 168172
    }
  },
  {
    "ph": "s", "id": 168172, "pid": 76337, "tid": -914061504, "ts": 1716454224053287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224102844, "dur": 33,
    "args": {
      "External id": 168178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168178, "pid": 5, "tid": 7, "ts": 1716454224102844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053314, "dur": 8,
    "args": {
      "External id": 168178,
      "cbid": 211, "correlation": 168178
    }
  },
  {
    "ph": "s", "id": 168178, "pid": 76337, "tid": -914061504, "ts": 1716454224053314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224102878, "dur": 27,
    "args": {
      "External id": 168186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168186,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168186, "pid": 5, "tid": 7, "ts": 1716454224102878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053347, "dur": 9,
    "args": {
      "External id": 168186,
      "cbid": 211, "correlation": 168186
    }
  },
  {
    "ph": "s", "id": 168186, "pid": 76337, "tid": -914061504, "ts": 1716454224053347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224102906, "dur": 20,
    "args": {
      "External id": 168194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168194,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168194, "pid": 5, "tid": 7, "ts": 1716454224102906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053376, "dur": 10,
    "args": {
      "External id": 168194,
      "cbid": 211, "correlation": 168194
    }
  },
  {
    "ph": "s", "id": 168194, "pid": 76337, "tid": -914061504, "ts": 1716454224053376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224102928, "dur": 30,
    "args": {
      "External id": 168214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168214,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 168214, "pid": 5, "tid": 7, "ts": 1716454224102928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053455, "dur": 12,
    "args": {
      "External id": 168214,
      "cbid": 211, "correlation": 168214
    }
  },
  {
    "ph": "s", "id": 168214, "pid": 76337, "tid": -914061504, "ts": 1716454224053455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224102959, "dur": 4,
    "args": {
      "External id": 168226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168226,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 168226, "pid": 5, "tid": 7, "ts": 1716454224102959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053477, "dur": 6,
    "args": {
      "External id": 168226,
      "cbid": 211, "correlation": 168226
    }
  },
  {
    "ph": "s", "id": 168226, "pid": 76337, "tid": -914061504, "ts": 1716454224053477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224102964, "dur": 30,
    "args": {
      "External id": 168229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168229,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168229, "pid": 5, "tid": 7, "ts": 1716454224102964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053495, "dur": 6,
    "args": {
      "External id": 168229,
      "cbid": 211, "correlation": 168229
    }
  },
  {
    "ph": "s", "id": 168229, "pid": 76337, "tid": -914061504, "ts": 1716454224053495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053553, "dur": 0,
    "args": {
      "External id": 168240,
      "cbid": 317, "correlation": 168240
    }
  },
  {
    "ph": "f", "id": 168240, "pid": 76337, "tid": -914061504, "ts": 1716454224053553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053554, "dur": 0,
    "args": {
      "External id": 168241,
      "cbid": 203, "correlation": 168241
    }
  },
  {
    "ph": "f", "id": 168241, "pid": 76337, "tid": -914061504, "ts": 1716454224053554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053554, "dur": 0,
    "args": {
      "External id": 168242,
      "cbid": 205, "correlation": 168242
    }
  },
  {
    "ph": "f", "id": 168242, "pid": 76337, "tid": -914061504, "ts": 1716454224053554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224102995, "dur": 22,
    "args": {
      "External id": 168246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168246,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168246, "pid": 5, "tid": 7, "ts": 1716454224102995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053567, "dur": 12,
    "args": {
      "External id": 168246,
      "cbid": 211, "correlation": 168246
    }
  },
  {
    "ph": "s", "id": 168246, "pid": 76337, "tid": -914061504, "ts": 1716454224053567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224103019, "dur": 117,
    "args": {
      "External id": 168248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168248,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168248, "pid": 5, "tid": 7, "ts": 1716454224103019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053586, "dur": 6,
    "args": {
      "External id": 168248,
      "cbid": 211, "correlation": 168248
    }
  },
  {
    "ph": "s", "id": 168248, "pid": 76337, "tid": -914061504, "ts": 1716454224053586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224103137, "dur": 21,
    "args": {
      "External id": 168250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168250,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168250, "pid": 5, "tid": 7, "ts": 1716454224103137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053595, "dur": 5,
    "args": {
      "External id": 168250,
      "cbid": 211, "correlation": 168250
    }
  },
  {
    "ph": "s", "id": 168250, "pid": 76337, "tid": -914061504, "ts": 1716454224053595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224103159, "dur": 33,
    "args": {
      "External id": 168256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168256, "pid": 5, "tid": 7, "ts": 1716454224103159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053622, "dur": 8,
    "args": {
      "External id": 168256,
      "cbid": 211, "correlation": 168256
    }
  },
  {
    "ph": "s", "id": 168256, "pid": 76337, "tid": -914061504, "ts": 1716454224053622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224103194, "dur": 181,
    "args": {
      "External id": 168265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168265,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168265, "pid": 5, "tid": 7, "ts": 1716454224103194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053705, "dur": 14,
    "args": {
      "External id": 168265,
      "cbid": 211, "correlation": 168265
    }
  },
  {
    "ph": "s", "id": 168265, "pid": 76337, "tid": -914061504, "ts": 1716454224053705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224103376, "dur": 65,
    "args": {
      "External id": 168287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168287,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168287, "pid": 5, "tid": 7, "ts": 1716454224103376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053763, "dur": 10,
    "args": {
      "External id": 168287,
      "cbid": 211, "correlation": 168287
    }
  },
  {
    "ph": "s", "id": 168287, "pid": 76337, "tid": -914061504, "ts": 1716454224053763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053850, "dur": 2,
    "args": {
      "External id": 168298,
      "cbid": 251, "correlation": 168298
    }
  },
  {
    "ph": "f", "id": 168298, "pid": 76337, "tid": -914061504, "ts": 1716454224053850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224103442, "dur": 152,
    "args": {
      "External id": 168299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168299,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168299, "pid": 5, "tid": 7, "ts": 1716454224103442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053856, "dur": 13,
    "args": {
      "External id": 168299,
      "cbid": 211, "correlation": 168299
    }
  },
  {
    "ph": "s", "id": 168299, "pid": 76337, "tid": -914061504, "ts": 1716454224053856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053925, "dur": 1,
    "args": {
      "External id": 168310,
      "cbid": 251, "correlation": 168310
    }
  },
  {
    "ph": "f", "id": 168310, "pid": 76337, "tid": -914061504, "ts": 1716454224053925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224103596, "dur": 144,
    "args": {
      "External id": 168311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168311,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168311, "pid": 5, "tid": 7, "ts": 1716454224103596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224053929, "dur": 12,
    "args": {
      "External id": 168311,
      "cbid": 211, "correlation": 168311
    }
  },
  {
    "ph": "s", "id": 168311, "pid": 76337, "tid": -914061504, "ts": 1716454224053929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054009, "dur": 1,
    "args": {
      "External id": 168322,
      "cbid": 251, "correlation": 168322
    }
  },
  {
    "ph": "f", "id": 168322, "pid": 76337, "tid": -914061504, "ts": 1716454224054009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224103742, "dur": 143,
    "args": {
      "External id": 168323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168323,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168323, "pid": 5, "tid": 7, "ts": 1716454224103742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054013, "dur": 12,
    "args": {
      "External id": 168323,
      "cbid": 211, "correlation": 168323
    }
  },
  {
    "ph": "s", "id": 168323, "pid": 76337, "tid": -914061504, "ts": 1716454224054013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224103886, "dur": 1963,
    "args": {
      "External id": 168344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168344,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 168344, "pid": 5, "tid": 7, "ts": 1716454224103886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054096, "dur": 13,
    "args": {
      "External id": 168344,
      "cbid": 211, "correlation": 168344
    }
  },
  {
    "ph": "s", "id": 168344, "pid": 76337, "tid": -914061504, "ts": 1716454224054096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054196, "dur": 1,
    "args": {
      "External id": 168362,
      "cbid": 251, "correlation": 168362
    }
  },
  {
    "ph": "f", "id": 168362, "pid": 76337, "tid": -914061504, "ts": 1716454224054196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224105851, "dur": 148,
    "args": {
      "External id": 168364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168364,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 168364, "pid": 5, "tid": 7, "ts": 1716454224105851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054202, "dur": 13,
    "args": {
      "External id": 168364,
      "cbid": 211, "correlation": 168364
    }
  },
  {
    "ph": "s", "id": 168364, "pid": 76337, "tid": -914061504, "ts": 1716454224054202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224106000, "dur": 35,
    "args": {
      "External id": 168372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168372,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168372, "pid": 5, "tid": 7, "ts": 1716454224106000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054271, "dur": 12,
    "args": {
      "External id": 168372,
      "cbid": 211, "correlation": 168372
    }
  },
  {
    "ph": "s", "id": 168372, "pid": 76337, "tid": -914061504, "ts": 1716454224054271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224106036, "dur": 50,
    "args": {
      "External id": 168380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168380,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168380, "pid": 5, "tid": 7, "ts": 1716454224106036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054309, "dur": 8,
    "args": {
      "External id": 168380,
      "cbid": 211, "correlation": 168380
    }
  },
  {
    "ph": "s", "id": 168380, "pid": 76337, "tid": -914061504, "ts": 1716454224054309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224106088, "dur": 31,
    "args": {
      "External id": 168391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168391, "pid": 5, "tid": 7, "ts": 1716454224106088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054381, "dur": 13,
    "args": {
      "External id": 168391,
      "cbid": 211, "correlation": 168391
    }
  },
  {
    "ph": "s", "id": 168391, "pid": 76337, "tid": -914061504, "ts": 1716454224054381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224106120, "dur": 35,
    "args": {
      "External id": 168413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168413,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168413, "pid": 5, "tid": 7, "ts": 1716454224106120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054412, "dur": 8,
    "args": {
      "External id": 168413,
      "cbid": 211, "correlation": 168413
    }
  },
  {
    "ph": "s", "id": 168413, "pid": 76337, "tid": -914061504, "ts": 1716454224054412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054497, "dur": 1,
    "args": {
      "External id": 168424,
      "cbid": 251, "correlation": 168424
    }
  },
  {
    "ph": "f", "id": 168424, "pid": 76337, "tid": -914061504, "ts": 1716454224054497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224106156, "dur": 90,
    "args": {
      "External id": 168425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168425,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168425, "pid": 5, "tid": 7, "ts": 1716454224106156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054502, "dur": 12,
    "args": {
      "External id": 168425,
      "cbid": 211, "correlation": 168425
    }
  },
  {
    "ph": "s", "id": 168425, "pid": 76337, "tid": -914061504, "ts": 1716454224054502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054571, "dur": 1,
    "args": {
      "External id": 168436,
      "cbid": 251, "correlation": 168436
    }
  },
  {
    "ph": "f", "id": 168436, "pid": 76337, "tid": -914061504, "ts": 1716454224054571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054574, "dur": 0,
    "args": {
      "External id": 168437,
      "cbid": 251, "correlation": 168437
    }
  },
  {
    "ph": "f", "id": 168437, "pid": 76337, "tid": -914061504, "ts": 1716454224054574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224106248, "dur": 11,
    "args": {
      "External id": 168438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168438,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 168438, "pid": 5, "tid": 7, "ts": 1716454224106248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054576, "dur": 12,
    "args": {
      "External id": 168438,
      "cbid": 211, "correlation": 168438
    }
  },
  {
    "ph": "s", "id": 168438, "pid": 76337, "tid": -914061504, "ts": 1716454224054576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224106260, "dur": 5,
    "args": {
      "External id": 168440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168440,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 168440, "pid": 5, "tid": 7, "ts": 1716454224106260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054590, "dur": 6,
    "args": {
      "External id": 168440,
      "cbid": 211, "correlation": 168440
    }
  },
  {
    "ph": "s", "id": 168440, "pid": 76337, "tid": -914061504, "ts": 1716454224054590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054648, "dur": 1,
    "args": {
      "External id": 168451,
      "cbid": 251, "correlation": 168451
    }
  },
  {
    "ph": "f", "id": 168451, "pid": 76337, "tid": -914061504, "ts": 1716454224054648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054651, "dur": 0,
    "args": {
      "External id": 168452,
      "cbid": 251, "correlation": 168452
    }
  },
  {
    "ph": "f", "id": 168452, "pid": 76337, "tid": -914061504, "ts": 1716454224054651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224106266, "dur": 7,
    "args": {
      "External id": 168453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168453,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 168453, "pid": 5, "tid": 7, "ts": 1716454224106266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054653, "dur": 12,
    "args": {
      "External id": 168453,
      "cbid": 211, "correlation": 168453
    }
  },
  {
    "ph": "s", "id": 168453, "pid": 76337, "tid": -914061504, "ts": 1716454224054653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224106275, "dur": 3,
    "args": {
      "External id": 168455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168455,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 168455, "pid": 5, "tid": 7, "ts": 1716454224106275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054666, "dur": 6,
    "args": {
      "External id": 168455,
      "cbid": 211, "correlation": 168455
    }
  },
  {
    "ph": "s", "id": 168455, "pid": 76337, "tid": -914061504, "ts": 1716454224054666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224106279, "dur": 92,
    "args": {
      "External id": 168476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168476,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 168476, "pid": 5, "tid": 7, "ts": 1716454224106279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054742, "dur": 13,
    "args": {
      "External id": 168476,
      "cbid": 211, "correlation": 168476
    }
  },
  {
    "ph": "s", "id": 168476, "pid": 76337, "tid": -914061504, "ts": 1716454224054742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054841, "dur": 1,
    "args": {
      "External id": 168494,
      "cbid": 251, "correlation": 168494
    }
  },
  {
    "ph": "f", "id": 168494, "pid": 76337, "tid": -914061504, "ts": 1716454224054841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224106373, "dur": 87,
    "args": {
      "External id": 168496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168496,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168496, "pid": 5, "tid": 7, "ts": 1716454224106373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054847, "dur": 13,
    "args": {
      "External id": 168496,
      "cbid": 211, "correlation": 168496
    }
  },
  {
    "ph": "s", "id": 168496, "pid": 76337, "tid": -914061504, "ts": 1716454224054847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224106461, "dur": 19,
    "args": {
      "External id": 168504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168504,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168504, "pid": 5, "tid": 7, "ts": 1716454224106461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054916, "dur": 11,
    "args": {
      "External id": 168504,
      "cbid": 211, "correlation": 168504
    }
  },
  {
    "ph": "s", "id": 168504, "pid": 76337, "tid": -914061504, "ts": 1716454224054916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224106482, "dur": 37,
    "args": {
      "External id": 168512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168512,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168512, "pid": 5, "tid": 7, "ts": 1716454224106482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224054957, "dur": 9,
    "args": {
      "External id": 168512,
      "cbid": 211, "correlation": 168512
    }
  },
  {
    "ph": "s", "id": 168512, "pid": 76337, "tid": -914061504, "ts": 1716454224054957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224106519, "dur": 35,
    "args": {
      "External id": 168534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168534,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168534, "pid": 5, "tid": 7, "ts": 1716454224106519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055018, "dur": 10,
    "args": {
      "External id": 168534,
      "cbid": 211, "correlation": 168534
    }
  },
  {
    "ph": "s", "id": 168534, "pid": 76337, "tid": -914061504, "ts": 1716454224055018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055109, "dur": 1,
    "args": {
      "External id": 168550,
      "cbid": 251, "correlation": 168550
    }
  },
  {
    "ph": "f", "id": 168550, "pid": 76337, "tid": -914061504, "ts": 1716454224055109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055114, "dur": 0,
    "args": {
      "External id": 168552,
      "cbid": 251, "correlation": 168552
    }
  },
  {
    "ph": "f", "id": 168552, "pid": 76337, "tid": -914061504, "ts": 1716454224055114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224106556, "dur": 543,
    "args": {
      "External id": 168553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168553,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 168553, "pid": 5, "tid": 7, "ts": 1716454224106556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055117, "dur": 13,
    "args": {
      "External id": 168553,
      "cbid": 211, "correlation": 168553
    }
  },
  {
    "ph": "s", "id": 168553, "pid": 76337, "tid": -914061504, "ts": 1716454224055117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224107100, "dur": 127,
    "args": {
      "External id": 168561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168561,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168561, "pid": 5, "tid": 7, "ts": 1716454224107100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055182, "dur": 13,
    "args": {
      "External id": 168561,
      "cbid": 211, "correlation": 168561
    }
  },
  {
    "ph": "s", "id": 168561, "pid": 76337, "tid": -914061504, "ts": 1716454224055182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224107228, "dur": 127,
    "args": {
      "External id": 168569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168569, "pid": 5, "tid": 7, "ts": 1716454224107228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055213, "dur": 8,
    "args": {
      "External id": 168569,
      "cbid": 211, "correlation": 168569
    }
  },
  {
    "ph": "s", "id": 168569, "pid": 76337, "tid": -914061504, "ts": 1716454224055213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055290, "dur": 1,
    "args": {
      "External id": 168585,
      "cbid": 251, "correlation": 168585
    }
  },
  {
    "ph": "f", "id": 168585, "pid": 76337, "tid": -914061504, "ts": 1716454224055290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224107357, "dur": 305,
    "args": {
      "External id": 168587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168587,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168587, "pid": 5, "tid": 7, "ts": 1716454224107357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055296, "dur": 12,
    "args": {
      "External id": 168587,
      "cbid": 211, "correlation": 168587
    }
  },
  {
    "ph": "s", "id": 168587, "pid": 76337, "tid": -914061504, "ts": 1716454224055296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224107663, "dur": 27,
    "args": {
      "External id": 168595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168595,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168595, "pid": 5, "tid": 7, "ts": 1716454224107663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055337, "dur": 10,
    "args": {
      "External id": 168595,
      "cbid": 211, "correlation": 168595
    }
  },
  {
    "ph": "s", "id": 168595, "pid": 76337, "tid": -914061504, "ts": 1716454224055337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224107691, "dur": 81,
    "args": {
      "External id": 168606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168606,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168606, "pid": 5, "tid": 7, "ts": 1716454224107691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055406, "dur": 12,
    "args": {
      "External id": 168606,
      "cbid": 211, "correlation": 168606
    }
  },
  {
    "ph": "s", "id": 168606, "pid": 76337, "tid": -914061504, "ts": 1716454224055406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055468, "dur": 0,
    "args": {
      "External id": 168618,
      "cbid": 317, "correlation": 168618
    }
  },
  {
    "ph": "f", "id": 168618, "pid": 76337, "tid": -914061504, "ts": 1716454224055468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055469, "dur": 0,
    "args": {
      "External id": 168619,
      "cbid": 203, "correlation": 168619
    }
  },
  {
    "ph": "f", "id": 168619, "pid": 76337, "tid": -914061504, "ts": 1716454224055469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055470, "dur": 0,
    "args": {
      "External id": 168620,
      "cbid": 205, "correlation": 168620
    }
  },
  {
    "ph": "f", "id": 168620, "pid": 76337, "tid": -914061504, "ts": 1716454224055470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224107774, "dur": 22,
    "args": {
      "External id": 168624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168624, "pid": 5, "tid": 7, "ts": 1716454224107774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055486, "dur": 13,
    "args": {
      "External id": 168624,
      "cbid": 211, "correlation": 168624
    }
  },
  {
    "ph": "s", "id": 168624, "pid": 76337, "tid": -914061504, "ts": 1716454224055486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224107798, "dur": 123,
    "args": {
      "External id": 168626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168626,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168626, "pid": 5, "tid": 7, "ts": 1716454224107798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055505, "dur": 6,
    "args": {
      "External id": 168626,
      "cbid": 211, "correlation": 168626
    }
  },
  {
    "ph": "s", "id": 168626, "pid": 76337, "tid": -914061504, "ts": 1716454224055505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224107922, "dur": 23,
    "args": {
      "External id": 168628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168628,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168628, "pid": 5, "tid": 7, "ts": 1716454224107922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055516, "dur": 5,
    "args": {
      "External id": 168628,
      "cbid": 211, "correlation": 168628
    }
  },
  {
    "ph": "s", "id": 168628, "pid": 76337, "tid": -914061504, "ts": 1716454224055516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224107946, "dur": 33,
    "args": {
      "External id": 168634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168634,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168634, "pid": 5, "tid": 7, "ts": 1716454224107946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055543, "dur": 9,
    "args": {
      "External id": 168634,
      "cbid": 211, "correlation": 168634
    }
  },
  {
    "ph": "s", "id": 168634, "pid": 76337, "tid": -914061504, "ts": 1716454224055543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224107980, "dur": 27,
    "args": {
      "External id": 168642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168642,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168642, "pid": 5, "tid": 7, "ts": 1716454224107980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055575, "dur": 8,
    "args": {
      "External id": 168642,
      "cbid": 211, "correlation": 168642
    }
  },
  {
    "ph": "s", "id": 168642, "pid": 76337, "tid": -914061504, "ts": 1716454224055575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055649, "dur": 0,
    "args": {
      "External id": 168652,
      "cbid": 317, "correlation": 168652
    }
  },
  {
    "ph": "f", "id": 168652, "pid": 76337, "tid": -914061504, "ts": 1716454224055649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055650, "dur": 0,
    "args": {
      "External id": 168653,
      "cbid": 203, "correlation": 168653
    }
  },
  {
    "ph": "f", "id": 168653, "pid": 76337, "tid": -914061504, "ts": 1716454224055650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055650, "dur": 0,
    "args": {
      "External id": 168654,
      "cbid": 205, "correlation": 168654
    }
  },
  {
    "ph": "f", "id": 168654, "pid": 76337, "tid": -914061504, "ts": 1716454224055650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224108008, "dur": 23,
    "args": {
      "External id": 168658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168658, "pid": 5, "tid": 7, "ts": 1716454224108008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055665, "dur": 12,
    "args": {
      "External id": 168658,
      "cbid": 211, "correlation": 168658
    }
  },
  {
    "ph": "s", "id": 168658, "pid": 76337, "tid": -914061504, "ts": 1716454224055665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224108032, "dur": 44,
    "args": {
      "External id": 168660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168660, "pid": 5, "tid": 7, "ts": 1716454224108032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055679, "dur": 5,
    "args": {
      "External id": 168660,
      "cbid": 211, "correlation": 168660
    }
  },
  {
    "ph": "s", "id": 168660, "pid": 76337, "tid": -914061504, "ts": 1716454224055679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224108077, "dur": 236,
    "args": {
      "External id": 168662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168662,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 168662, "pid": 5, "tid": 7, "ts": 1716454224108077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055691, "dur": 8,
    "args": {
      "External id": 168662,
      "cbid": 211, "correlation": 168662
    }
  },
  {
    "ph": "s", "id": 168662, "pid": 76337, "tid": -914061504, "ts": 1716454224055691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224108314, "dur": 6,
    "args": {
      "External id": 168664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168664,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168664, "pid": 5, "tid": 7, "ts": 1716454224108314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055702, "dur": 5,
    "args": {
      "External id": 168664,
      "cbid": 211, "correlation": 168664
    }
  },
  {
    "ph": "s", "id": 168664, "pid": 76337, "tid": -914061504, "ts": 1716454224055702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224108322, "dur": 9,
    "args": {
      "External id": 168670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168670,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168670, "pid": 5, "tid": 7, "ts": 1716454224108322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055728, "dur": 8,
    "args": {
      "External id": 168670,
      "cbid": 211, "correlation": 168670
    }
  },
  {
    "ph": "s", "id": 168670, "pid": 76337, "tid": -914061504, "ts": 1716454224055728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224108333, "dur": 12,
    "args": {
      "External id": 168690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168690,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 168690, "pid": 5, "tid": 7, "ts": 1716454224108333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055825, "dur": 12,
    "args": {
      "External id": 168690,
      "cbid": 211, "correlation": 168690
    }
  },
  {
    "ph": "s", "id": 168690, "pid": 76337, "tid": -914061504, "ts": 1716454224055825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224108346, "dur": 4,
    "args": {
      "External id": 168702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168702,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 168702, "pid": 5, "tid": 7, "ts": 1716454224108346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055848, "dur": 7,
    "args": {
      "External id": 168702,
      "cbid": 211, "correlation": 168702
    }
  },
  {
    "ph": "s", "id": 168702, "pid": 76337, "tid": -914061504, "ts": 1716454224055848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224108351, "dur": 12,
    "args": {
      "External id": 168705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168705,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168705, "pid": 5, "tid": 7, "ts": 1716454224108351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055866, "dur": 6,
    "args": {
      "External id": 168705,
      "cbid": 211, "correlation": 168705
    }
  },
  {
    "ph": "s", "id": 168705, "pid": 76337, "tid": -914061504, "ts": 1716454224055866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224108365, "dur": 7,
    "args": {
      "External id": 168714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168714,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168714, "pid": 5, "tid": 7, "ts": 1716454224108365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055906, "dur": 10,
    "args": {
      "External id": 168714,
      "cbid": 211, "correlation": 168714
    }
  },
  {
    "ph": "s", "id": 168714, "pid": 76337, "tid": -914061504, "ts": 1716454224055906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055959, "dur": 0,
    "args": {
      "External id": 168724,
      "cbid": 317, "correlation": 168724
    }
  },
  {
    "ph": "f", "id": 168724, "pid": 76337, "tid": -914061504, "ts": 1716454224055959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055960, "dur": 0,
    "args": {
      "External id": 168725,
      "cbid": 203, "correlation": 168725
    }
  },
  {
    "ph": "f", "id": 168725, "pid": 76337, "tid": -914061504, "ts": 1716454224055960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055961, "dur": 0,
    "args": {
      "External id": 168726,
      "cbid": 205, "correlation": 168726
    }
  },
  {
    "ph": "f", "id": 168726, "pid": 76337, "tid": -914061504, "ts": 1716454224055961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224108373, "dur": 6,
    "args": {
      "External id": 168730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168730,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168730, "pid": 5, "tid": 7, "ts": 1716454224108373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224055985, "dur": 12,
    "args": {
      "External id": 168730,
      "cbid": 211, "correlation": 168730
    }
  },
  {
    "ph": "s", "id": 168730, "pid": 76337, "tid": -914061504, "ts": 1716454224055985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224108380, "dur": 84,
    "args": {
      "External id": 168732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168732,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168732, "pid": 5, "tid": 7, "ts": 1716454224108380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056000, "dur": 5,
    "args": {
      "External id": 168732,
      "cbid": 211, "correlation": 168732
    }
  },
  {
    "ph": "s", "id": 168732, "pid": 76337, "tid": -914061504, "ts": 1716454224056000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224108466, "dur": 1,
    "args": {
      "External id": 168734,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 168734,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 168734, "pid": 5, "tid": 7, "ts": 1716454224108466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056013, "dur": 14,
    "args": {
      "External id": 168734,
      "cbid": 51, "correlation": 168734
    }
  },
  {
    "ph": "s", "id": 168734, "pid": 76337, "tid": -914061504, "ts": 1716454224056013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224108470, "dur": 541,
    "args": {
      "External id": 168735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168735,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168735, "pid": 5, "tid": 7, "ts": 1716454224108470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056028, "dur": 9,
    "args": {
      "External id": 168735,
      "cbid": 211, "correlation": 168735
    }
  },
  {
    "ph": "s", "id": 168735, "pid": 76337, "tid": -914061504, "ts": 1716454224056028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224109012, "dur": 12,
    "args": {
      "External id": 168737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168737,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168737, "pid": 5, "tid": 7, "ts": 1716454224109012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056041, "dur": 6,
    "args": {
      "External id": 168737,
      "cbid": 211, "correlation": 168737
    }
  },
  {
    "ph": "s", "id": 168737, "pid": 76337, "tid": -914061504, "ts": 1716454224056041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224109025, "dur": 15,
    "args": {
      "External id": 168743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168743, "pid": 5, "tid": 7, "ts": 1716454224109025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056070, "dur": 9,
    "args": {
      "External id": 168743,
      "cbid": 211, "correlation": 168743
    }
  },
  {
    "ph": "s", "id": 168743, "pid": 76337, "tid": -914061504, "ts": 1716454224056070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224109041, "dur": 4,
    "args": {
      "External id": 168751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168751,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 168751, "pid": 5, "tid": 7, "ts": 1716454224109041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056115, "dur": 9,
    "args": {
      "External id": 168751,
      "cbid": 211, "correlation": 168751
    }
  },
  {
    "ph": "s", "id": 168751, "pid": 76337, "tid": -914061504, "ts": 1716454224056115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056179, "dur": 1,
    "args": {
      "External id": 168767,
      "cbid": 251, "correlation": 168767
    }
  },
  {
    "ph": "f", "id": 168767, "pid": 76337, "tid": -914061504, "ts": 1716454224056179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056184, "dur": 0,
    "args": {
      "External id": 168769,
      "cbid": 251, "correlation": 168769
    }
  },
  {
    "ph": "f", "id": 168769, "pid": 76337, "tid": -914061504, "ts": 1716454224056184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224109047, "dur": 13,
    "args": {
      "External id": 168770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168770,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168770, "pid": 5, "tid": 7, "ts": 1716454224109047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056186, "dur": 12,
    "args": {
      "External id": 168770,
      "cbid": 211, "correlation": 168770
    }
  },
  {
    "ph": "s", "id": 168770, "pid": 76337, "tid": -914061504, "ts": 1716454224056186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224109061, "dur": 5,
    "args": {
      "External id": 168772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168772,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168772, "pid": 5, "tid": 7, "ts": 1716454224109061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056200, "dur": 6,
    "args": {
      "External id": 168772,
      "cbid": 211, "correlation": 168772
    }
  },
  {
    "ph": "s", "id": 168772, "pid": 76337, "tid": -914061504, "ts": 1716454224056200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224109068, "dur": 18,
    "args": {
      "External id": 168782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168782,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168782, "pid": 5, "tid": 7, "ts": 1716454224109068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056258, "dur": 12,
    "args": {
      "External id": 168782,
      "cbid": 211, "correlation": 168782
    }
  },
  {
    "ph": "s", "id": 168782, "pid": 76337, "tid": -914061504, "ts": 1716454224056258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224109087, "dur": 18,
    "args": {
      "External id": 168802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168802,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 168802, "pid": 5, "tid": 7, "ts": 1716454224109087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056324, "dur": 10,
    "args": {
      "External id": 168802,
      "cbid": 211, "correlation": 168802
    }
  },
  {
    "ph": "s", "id": 168802, "pid": 76337, "tid": -914061504, "ts": 1716454224056324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224109106, "dur": 5,
    "args": {
      "External id": 168814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168814,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 168814, "pid": 5, "tid": 7, "ts": 1716454224109106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056345, "dur": 7,
    "args": {
      "External id": 168814,
      "cbid": 211, "correlation": 168814
    }
  },
  {
    "ph": "s", "id": 168814, "pid": 76337, "tid": -914061504, "ts": 1716454224056345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224109112, "dur": 17,
    "args": {
      "External id": 168817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168817,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168817, "pid": 5, "tid": 7, "ts": 1716454224109112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056364, "dur": 6,
    "args": {
      "External id": 168817,
      "cbid": 211, "correlation": 168817
    }
  },
  {
    "ph": "s", "id": 168817, "pid": 76337, "tid": -914061504, "ts": 1716454224056364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224109130, "dur": 11,
    "args": {
      "External id": 168826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168826,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168826, "pid": 5, "tid": 7, "ts": 1716454224109130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056405, "dur": 9,
    "args": {
      "External id": 168826,
      "cbid": 211, "correlation": 168826
    }
  },
  {
    "ph": "s", "id": 168826, "pid": 76337, "tid": -914061504, "ts": 1716454224056405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056467, "dur": 0,
    "args": {
      "External id": 168836,
      "cbid": 317, "correlation": 168836
    }
  },
  {
    "ph": "f", "id": 168836, "pid": 76337, "tid": -914061504, "ts": 1716454224056467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056468, "dur": 0,
    "args": {
      "External id": 168837,
      "cbid": 203, "correlation": 168837
    }
  },
  {
    "ph": "f", "id": 168837, "pid": 76337, "tid": -914061504, "ts": 1716454224056468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056469, "dur": 0,
    "args": {
      "External id": 168838,
      "cbid": 205, "correlation": 168838
    }
  },
  {
    "ph": "f", "id": 168838, "pid": 76337, "tid": -914061504, "ts": 1716454224056469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224109142, "dur": 12,
    "args": {
      "External id": 168842,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168842,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168842, "pid": 5, "tid": 7, "ts": 1716454224109142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056486, "dur": 12,
    "args": {
      "External id": 168842,
      "cbid": 211, "correlation": 168842
    }
  },
  {
    "ph": "s", "id": 168842, "pid": 76337, "tid": -914061504, "ts": 1716454224056486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224109155, "dur": 164,
    "args": {
      "External id": 168844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168844,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168844, "pid": 5, "tid": 7, "ts": 1716454224109155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056502, "dur": 5,
    "args": {
      "External id": 168844,
      "cbid": 211, "correlation": 168844
    }
  },
  {
    "ph": "s", "id": 168844, "pid": 76337, "tid": -914061504, "ts": 1716454224056502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224109321, "dur": 1,
    "args": {
      "External id": 168846,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 168846,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 168846, "pid": 5, "tid": 7, "ts": 1716454224109321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056514, "dur": 9,
    "args": {
      "External id": 168846,
      "cbid": 51, "correlation": 168846
    }
  },
  {
    "ph": "s", "id": 168846, "pid": 76337, "tid": -914061504, "ts": 1716454224056514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224109325, "dur": 666,
    "args": {
      "External id": 168847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168847,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 168847, "pid": 5, "tid": 7, "ts": 1716454224109325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056525, "dur": 8,
    "args": {
      "External id": 168847,
      "cbid": 211, "correlation": 168847
    }
  },
  {
    "ph": "s", "id": 168847, "pid": 76337, "tid": -914061504, "ts": 1716454224056525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224109992, "dur": 13,
    "args": {
      "External id": 168849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168849,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168849, "pid": 5, "tid": 7, "ts": 1716454224109992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056537, "dur": 5,
    "args": {
      "External id": 168849,
      "cbid": 211, "correlation": 168849
    }
  },
  {
    "ph": "s", "id": 168849, "pid": 76337, "tid": -914061504, "ts": 1716454224056537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224110007, "dur": 15,
    "args": {
      "External id": 168855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168855,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168855, "pid": 5, "tid": 7, "ts": 1716454224110007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056566, "dur": 8,
    "args": {
      "External id": 168855,
      "cbid": 211, "correlation": 168855
    }
  },
  {
    "ph": "s", "id": 168855, "pid": 76337, "tid": -914061504, "ts": 1716454224056566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056624, "dur": 0,
    "args": {
      "External id": 168865,
      "cbid": 317, "correlation": 168865
    }
  },
  {
    "ph": "f", "id": 168865, "pid": 76337, "tid": -914061504, "ts": 1716454224056624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056625, "dur": 0,
    "args": {
      "External id": 168866,
      "cbid": 203, "correlation": 168866
    }
  },
  {
    "ph": "f", "id": 168866, "pid": 76337, "tid": -914061504, "ts": 1716454224056625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056626, "dur": 0,
    "args": {
      "External id": 168867,
      "cbid": 205, "correlation": 168867
    }
  },
  {
    "ph": "f", "id": 168867, "pid": 76337, "tid": -914061504, "ts": 1716454224056626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224110023, "dur": 9,
    "args": {
      "External id": 168871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168871,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168871, "pid": 5, "tid": 7, "ts": 1716454224110023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056641, "dur": 12,
    "args": {
      "External id": 168871,
      "cbid": 211, "correlation": 168871
    }
  },
  {
    "ph": "s", "id": 168871, "pid": 76337, "tid": -914061504, "ts": 1716454224056641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224110033, "dur": 3,
    "args": {
      "External id": 168873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168873,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 168873, "pid": 5, "tid": 7, "ts": 1716454224110033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056661, "dur": 6,
    "args": {
      "External id": 168873,
      "cbid": 211, "correlation": 168873
    }
  },
  {
    "ph": "s", "id": 168873, "pid": 76337, "tid": -914061504, "ts": 1716454224056661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056672, "dur": 0,
    "args": {
      "External id": 168874,
      "cbid": 51, "correlation": 168874
    }
  },
  {
    "ph": "s", "id": 168874, "pid": 76337, "tid": -914061504, "ts": 1716454224056672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224110038, "dur": 58,
    "args": {
      "External id": 168875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168875,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 168875, "pid": 5, "tid": 7, "ts": 1716454224110038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056673, "dur": 7,
    "args": {
      "External id": 168875,
      "cbid": 211, "correlation": 168875
    }
  },
  {
    "ph": "s", "id": 168875, "pid": 76337, "tid": -914061504, "ts": 1716454224056673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224110097, "dur": 14,
    "args": {
      "External id": 168880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168880, "pid": 5, "tid": 7, "ts": 1716454224110097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056701, "dur": 9,
    "args": {
      "External id": 168880,
      "cbid": 211, "correlation": 168880
    }
  },
  {
    "ph": "s", "id": 168880, "pid": 76337, "tid": -914061504, "ts": 1716454224056701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224110113, "dur": 12,
    "args": {
      "External id": 168888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168888,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168888, "pid": 5, "tid": 7, "ts": 1716454224110113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056730, "dur": 8,
    "args": {
      "External id": 168888,
      "cbid": 211, "correlation": 168888
    }
  },
  {
    "ph": "s", "id": 168888, "pid": 76337, "tid": -914061504, "ts": 1716454224056730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224110126, "dur": 11,
    "args": {
      "External id": 168896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168896,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168896, "pid": 5, "tid": 7, "ts": 1716454224110126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056759, "dur": 8,
    "args": {
      "External id": 168896,
      "cbid": 211, "correlation": 168896
    }
  },
  {
    "ph": "s", "id": 168896, "pid": 76337, "tid": -914061504, "ts": 1716454224056759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224110138, "dur": 18,
    "args": {
      "External id": 168916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168916,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 168916, "pid": 5, "tid": 7, "ts": 1716454224110138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056840, "dur": 13,
    "args": {
      "External id": 168916,
      "cbid": 211, "correlation": 168916
    }
  },
  {
    "ph": "s", "id": 168916, "pid": 76337, "tid": -914061504, "ts": 1716454224056840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224110158, "dur": 5,
    "args": {
      "External id": 168928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168928,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 168928, "pid": 5, "tid": 7, "ts": 1716454224110158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056863, "dur": 6,
    "args": {
      "External id": 168928,
      "cbid": 211, "correlation": 168928
    }
  },
  {
    "ph": "s", "id": 168928, "pid": 76337, "tid": -914061504, "ts": 1716454224056863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224110163, "dur": 17,
    "args": {
      "External id": 168931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168931, "pid": 5, "tid": 7, "ts": 1716454224110163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056880, "dur": 6,
    "args": {
      "External id": 168931,
      "cbid": 211, "correlation": 168931
    }
  },
  {
    "ph": "s", "id": 168931, "pid": 76337, "tid": -914061504, "ts": 1716454224056880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056938, "dur": 0,
    "args": {
      "External id": 168942,
      "cbid": 317, "correlation": 168942
    }
  },
  {
    "ph": "f", "id": 168942, "pid": 76337, "tid": -914061504, "ts": 1716454224056938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056938, "dur": 0,
    "args": {
      "External id": 168943,
      "cbid": 203, "correlation": 168943
    }
  },
  {
    "ph": "f", "id": 168943, "pid": 76337, "tid": -914061504, "ts": 1716454224056938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056939, "dur": 0,
    "args": {
      "External id": 168944,
      "cbid": 205, "correlation": 168944
    }
  },
  {
    "ph": "f", "id": 168944, "pid": 76337, "tid": -914061504, "ts": 1716454224056939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224110182, "dur": 11,
    "args": {
      "External id": 168948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168948, "pid": 5, "tid": 7, "ts": 1716454224110182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056955, "dur": 11,
    "args": {
      "External id": 168948,
      "cbid": 211, "correlation": 168948
    }
  },
  {
    "ph": "s", "id": 168948, "pid": 76337, "tid": -914061504, "ts": 1716454224056955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224110194, "dur": 4,
    "args": {
      "External id": 168950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168950,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 168950, "pid": 5, "tid": 7, "ts": 1716454224110194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056972, "dur": 15,
    "args": {
      "External id": 168950,
      "cbid": 211, "correlation": 168950
    }
  },
  {
    "ph": "s", "id": 168950, "pid": 76337, "tid": -914061504, "ts": 1716454224056972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056991, "dur": 0,
    "args": {
      "External id": 168951,
      "cbid": 51, "correlation": 168951
    }
  },
  {
    "ph": "s", "id": 168951, "pid": 76337, "tid": -914061504, "ts": 1716454224056991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224110199, "dur": 99,
    "args": {
      "External id": 168952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168952,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 168952, "pid": 5, "tid": 7, "ts": 1716454224110199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224056992, "dur": 7,
    "args": {
      "External id": 168952,
      "cbid": 211, "correlation": 168952
    }
  },
  {
    "ph": "s", "id": 168952, "pid": 76337, "tid": -914061504, "ts": 1716454224056992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224110300, "dur": 16,
    "args": {
      "External id": 168957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168957, "pid": 5, "tid": 7, "ts": 1716454224110300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057021, "dur": 9,
    "args": {
      "External id": 168957,
      "cbid": 211, "correlation": 168957
    }
  },
  {
    "ph": "s", "id": 168957, "pid": 76337, "tid": -914061504, "ts": 1716454224057021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224110317, "dur": 84,
    "args": {
      "External id": 168966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168966,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168966, "pid": 5, "tid": 7, "ts": 1716454224110317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057106, "dur": 14,
    "args": {
      "External id": 168966,
      "cbid": 211, "correlation": 168966
    }
  },
  {
    "ph": "s", "id": 168966, "pid": 76337, "tid": -914061504, "ts": 1716454224057106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224110403, "dur": 31,
    "args": {
      "External id": 168988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 168988,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 168988, "pid": 5, "tid": 7, "ts": 1716454224110403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057162, "dur": 10,
    "args": {
      "External id": 168988,
      "cbid": 211, "correlation": 168988
    }
  },
  {
    "ph": "s", "id": 168988, "pid": 76337, "tid": -914061504, "ts": 1716454224057162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057253, "dur": 2,
    "args": {
      "External id": 168999,
      "cbid": 251, "correlation": 168999
    }
  },
  {
    "ph": "f", "id": 168999, "pid": 76337, "tid": -914061504, "ts": 1716454224057253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224110435, "dur": 165,
    "args": {
      "External id": 169000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169000,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169000, "pid": 5, "tid": 7, "ts": 1716454224110435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057258, "dur": 13,
    "args": {
      "External id": 169000,
      "cbid": 211, "correlation": 169000
    }
  },
  {
    "ph": "s", "id": 169000, "pid": 76337, "tid": -914061504, "ts": 1716454224057258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057328, "dur": 1,
    "args": {
      "External id": 169011,
      "cbid": 251, "correlation": 169011
    }
  },
  {
    "ph": "f", "id": 169011, "pid": 76337, "tid": -914061504, "ts": 1716454224057328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224110601, "dur": 160,
    "args": {
      "External id": 169012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169012,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169012, "pid": 5, "tid": 7, "ts": 1716454224110601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057332, "dur": 11,
    "args": {
      "External id": 169012,
      "cbid": 211, "correlation": 169012
    }
  },
  {
    "ph": "s", "id": 169012, "pid": 76337, "tid": -914061504, "ts": 1716454224057332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057396, "dur": 1,
    "args": {
      "External id": 169023,
      "cbid": 251, "correlation": 169023
    }
  },
  {
    "ph": "f", "id": 169023, "pid": 76337, "tid": -914061504, "ts": 1716454224057396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224110762, "dur": 158,
    "args": {
      "External id": 169024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169024,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169024, "pid": 5, "tid": 7, "ts": 1716454224110762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057400, "dur": 11,
    "args": {
      "External id": 169024,
      "cbid": 211, "correlation": 169024
    }
  },
  {
    "ph": "s", "id": 169024, "pid": 76337, "tid": -914061504, "ts": 1716454224057400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224110922, "dur": 341,
    "args": {
      "External id": 169049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169049,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169049, "pid": 5, "tid": 7, "ts": 1716454224110922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057488, "dur": 14,
    "args": {
      "External id": 169049,
      "cbid": 211, "correlation": 169049
    }
  },
  {
    "ph": "s", "id": 169049, "pid": 76337, "tid": -914061504, "ts": 1716454224057488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057588, "dur": 1,
    "args": {
      "External id": 169067,
      "cbid": 251, "correlation": 169067
    }
  },
  {
    "ph": "f", "id": 169067, "pid": 76337, "tid": -914061504, "ts": 1716454224057588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224111264, "dur": 167,
    "args": {
      "External id": 169069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169069,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169069, "pid": 5, "tid": 7, "ts": 1716454224111264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057594, "dur": 14,
    "args": {
      "External id": 169069,
      "cbid": 211, "correlation": 169069
    }
  },
  {
    "ph": "s", "id": 169069, "pid": 76337, "tid": -914061504, "ts": 1716454224057594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224111432, "dur": 19,
    "args": {
      "External id": 169077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169077,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169077, "pid": 5, "tid": 7, "ts": 1716454224111432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057664, "dur": 12,
    "args": {
      "External id": 169077,
      "cbid": 211, "correlation": 169077
    }
  },
  {
    "ph": "s", "id": 169077, "pid": 76337, "tid": -914061504, "ts": 1716454224057664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224111452, "dur": 27,
    "args": {
      "External id": 169085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169085,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169085, "pid": 5, "tid": 7, "ts": 1716454224111452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057703, "dur": 9,
    "args": {
      "External id": 169085,
      "cbid": 211, "correlation": 169085
    }
  },
  {
    "ph": "s", "id": 169085, "pid": 76337, "tid": -914061504, "ts": 1716454224057703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224111481, "dur": 19,
    "args": {
      "External id": 169096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169096, "pid": 5, "tid": 7, "ts": 1716454224111481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057775, "dur": 12,
    "args": {
      "External id": 169096,
      "cbid": 211, "correlation": 169096
    }
  },
  {
    "ph": "s", "id": 169096, "pid": 76337, "tid": -914061504, "ts": 1716454224057775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224111502, "dur": 17,
    "args": {
      "External id": 169118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169118,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169118, "pid": 5, "tid": 7, "ts": 1716454224111502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057805, "dur": 8,
    "args": {
      "External id": 169118,
      "cbid": 211, "correlation": 169118
    }
  },
  {
    "ph": "s", "id": 169118, "pid": 76337, "tid": -914061504, "ts": 1716454224057805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057890, "dur": 1,
    "args": {
      "External id": 169129,
      "cbid": 251, "correlation": 169129
    }
  },
  {
    "ph": "f", "id": 169129, "pid": 76337, "tid": -914061504, "ts": 1716454224057890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224111520, "dur": 90,
    "args": {
      "External id": 169130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169130,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169130, "pid": 5, "tid": 7, "ts": 1716454224111520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057897, "dur": 15,
    "args": {
      "External id": 169130,
      "cbid": 211, "correlation": 169130
    }
  },
  {
    "ph": "s", "id": 169130, "pid": 76337, "tid": -914061504, "ts": 1716454224057897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057968, "dur": 1,
    "args": {
      "External id": 169141,
      "cbid": 251, "correlation": 169141
    }
  },
  {
    "ph": "f", "id": 169141, "pid": 76337, "tid": -914061504, "ts": 1716454224057968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057971, "dur": 0,
    "args": {
      "External id": 169142,
      "cbid": 251, "correlation": 169142
    }
  },
  {
    "ph": "f", "id": 169142, "pid": 76337, "tid": -914061504, "ts": 1716454224057971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224111611, "dur": 13,
    "args": {
      "External id": 169143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169143,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169143, "pid": 5, "tid": 7, "ts": 1716454224111611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057981, "dur": 13,
    "args": {
      "External id": 169143,
      "cbid": 211, "correlation": 169143
    }
  },
  {
    "ph": "s", "id": 169143, "pid": 76337, "tid": -914061504, "ts": 1716454224057981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224111625, "dur": 6,
    "args": {
      "External id": 169145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169145,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169145, "pid": 5, "tid": 7, "ts": 1716454224111625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224057996, "dur": 6,
    "args": {
      "External id": 169145,
      "cbid": 211, "correlation": 169145
    }
  },
  {
    "ph": "s", "id": 169145, "pid": 76337, "tid": -914061504, "ts": 1716454224057996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058056, "dur": 1,
    "args": {
      "External id": 169156,
      "cbid": 251, "correlation": 169156
    }
  },
  {
    "ph": "f", "id": 169156, "pid": 76337, "tid": -914061504, "ts": 1716454224058056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058060, "dur": 0,
    "args": {
      "External id": 169157,
      "cbid": 251, "correlation": 169157
    }
  },
  {
    "ph": "f", "id": 169157, "pid": 76337, "tid": -914061504, "ts": 1716454224058060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224111632, "dur": 9,
    "args": {
      "External id": 169158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169158,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169158, "pid": 5, "tid": 7, "ts": 1716454224111632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058062, "dur": 11,
    "args": {
      "External id": 169158,
      "cbid": 211, "correlation": 169158
    }
  },
  {
    "ph": "s", "id": 169158, "pid": 76337, "tid": -914061504, "ts": 1716454224058062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224111642, "dur": 3,
    "args": {
      "External id": 169160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169160,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169160, "pid": 5, "tid": 7, "ts": 1716454224111642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058075, "dur": 5,
    "args": {
      "External id": 169160,
      "cbid": 211, "correlation": 169160
    }
  },
  {
    "ph": "s", "id": 169160, "pid": 76337, "tid": -914061504, "ts": 1716454224058075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224111647, "dur": 55,
    "args": {
      "External id": 169185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169185,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169185, "pid": 5, "tid": 7, "ts": 1716454224111647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058152, "dur": 13,
    "args": {
      "External id": 169185,
      "cbid": 211, "correlation": 169185
    }
  },
  {
    "ph": "s", "id": 169185, "pid": 76337, "tid": -914061504, "ts": 1716454224058152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058252, "dur": 2,
    "args": {
      "External id": 169203,
      "cbid": 251, "correlation": 169203
    }
  },
  {
    "ph": "f", "id": 169203, "pid": 76337, "tid": -914061504, "ts": 1716454224058252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224111703, "dur": 92,
    "args": {
      "External id": 169205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169205,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169205, "pid": 5, "tid": 7, "ts": 1716454224111703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058260, "dur": 15,
    "args": {
      "External id": 169205,
      "cbid": 211, "correlation": 169205
    }
  },
  {
    "ph": "s", "id": 169205, "pid": 76337, "tid": -914061504, "ts": 1716454224058260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224111796, "dur": 9,
    "args": {
      "External id": 169213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169213,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169213, "pid": 5, "tid": 7, "ts": 1716454224111796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058331, "dur": 12,
    "args": {
      "External id": 169213,
      "cbid": 211, "correlation": 169213
    }
  },
  {
    "ph": "s", "id": 169213, "pid": 76337, "tid": -914061504, "ts": 1716454224058331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224111807, "dur": 21,
    "args": {
      "External id": 169221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169221,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169221, "pid": 5, "tid": 7, "ts": 1716454224111807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058373, "dur": 9,
    "args": {
      "External id": 169221,
      "cbid": 211, "correlation": 169221
    }
  },
  {
    "ph": "s", "id": 169221, "pid": 76337, "tid": -914061504, "ts": 1716454224058373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224111830, "dur": 18,
    "args": {
      "External id": 169243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169243,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169243, "pid": 5, "tid": 7, "ts": 1716454224111830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058424, "dur": 10,
    "args": {
      "External id": 169243,
      "cbid": 211, "correlation": 169243
    }
  },
  {
    "ph": "s", "id": 169243, "pid": 76337, "tid": -914061504, "ts": 1716454224058424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058514, "dur": 2,
    "args": {
      "External id": 169259,
      "cbid": 251, "correlation": 169259
    }
  },
  {
    "ph": "f", "id": 169259, "pid": 76337, "tid": -914061504, "ts": 1716454224058514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058520, "dur": 0,
    "args": {
      "External id": 169261,
      "cbid": 251, "correlation": 169261
    }
  },
  {
    "ph": "f", "id": 169261, "pid": 76337, "tid": -914061504, "ts": 1716454224058520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224111849, "dur": 495,
    "args": {
      "External id": 169262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169262,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169262, "pid": 5, "tid": 7, "ts": 1716454224111849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058522, "dur": 15,
    "args": {
      "External id": 169262,
      "cbid": 211, "correlation": 169262
    }
  },
  {
    "ph": "s", "id": 169262, "pid": 76337, "tid": -914061504, "ts": 1716454224058522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224112345, "dur": 66,
    "args": {
      "External id": 169270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169270, "pid": 5, "tid": 7, "ts": 1716454224112345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058590, "dur": 12,
    "args": {
      "External id": 169270,
      "cbid": 211, "correlation": 169270
    }
  },
  {
    "ph": "s", "id": 169270, "pid": 76337, "tid": -914061504, "ts": 1716454224058590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224112413, "dur": 69,
    "args": {
      "External id": 169278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169278, "pid": 5, "tid": 7, "ts": 1716454224112413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058622, "dur": 8,
    "args": {
      "External id": 169278,
      "cbid": 211, "correlation": 169278
    }
  },
  {
    "ph": "s", "id": 169278, "pid": 76337, "tid": -914061504, "ts": 1716454224058622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058701, "dur": 1,
    "args": {
      "External id": 169294,
      "cbid": 251, "correlation": 169294
    }
  },
  {
    "ph": "f", "id": 169294, "pid": 76337, "tid": -914061504, "ts": 1716454224058701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224112484, "dur": 1,
    "args": {
      "External id": 169296,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 169296,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 169296, "pid": 5, "tid": 7, "ts": 1716454224112484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058707, "dur": 12,
    "args": {
      "External id": 169296,
      "cbid": 51, "correlation": 169296
    }
  },
  {
    "ph": "s", "id": 169296, "pid": 76337, "tid": -914061504, "ts": 1716454224058707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224112487, "dur": 269,
    "args": {
      "External id": 169297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169297,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169297, "pid": 5, "tid": 7, "ts": 1716454224112487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058720, "dur": 11,
    "args": {
      "External id": 169297,
      "cbid": 211, "correlation": 169297
    }
  },
  {
    "ph": "s", "id": 169297, "pid": 76337, "tid": -914061504, "ts": 1716454224058720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224112757, "dur": 15,
    "args": {
      "External id": 169305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169305,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169305, "pid": 5, "tid": 7, "ts": 1716454224112757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058762, "dur": 10,
    "args": {
      "External id": 169305,
      "cbid": 211, "correlation": 169305
    }
  },
  {
    "ph": "s", "id": 169305, "pid": 76337, "tid": -914061504, "ts": 1716454224058762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224112773, "dur": 38,
    "args": {
      "External id": 169316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169316,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169316, "pid": 5, "tid": 7, "ts": 1716454224112773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058829, "dur": 12,
    "args": {
      "External id": 169316,
      "cbid": 211, "correlation": 169316
    }
  },
  {
    "ph": "s", "id": 169316, "pid": 76337, "tid": -914061504, "ts": 1716454224058829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058893, "dur": 0,
    "args": {
      "External id": 169328,
      "cbid": 317, "correlation": 169328
    }
  },
  {
    "ph": "f", "id": 169328, "pid": 76337, "tid": -914061504, "ts": 1716454224058893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058894, "dur": 0,
    "args": {
      "External id": 169329,
      "cbid": 203, "correlation": 169329
    }
  },
  {
    "ph": "f", "id": 169329, "pid": 76337, "tid": -914061504, "ts": 1716454224058894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058895, "dur": 0,
    "args": {
      "External id": 169330,
      "cbid": 205, "correlation": 169330
    }
  },
  {
    "ph": "f", "id": 169330, "pid": 76337, "tid": -914061504, "ts": 1716454224058895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224112813, "dur": 13,
    "args": {
      "External id": 169334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169334,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169334, "pid": 5, "tid": 7, "ts": 1716454224112813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058910, "dur": 13,
    "args": {
      "External id": 169334,
      "cbid": 211, "correlation": 169334
    }
  },
  {
    "ph": "s", "id": 169334, "pid": 76337, "tid": -914061504, "ts": 1716454224058910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224112827, "dur": 4,
    "args": {
      "External id": 169336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169336,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169336, "pid": 5, "tid": 7, "ts": 1716454224112827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058927, "dur": 6,
    "args": {
      "External id": 169336,
      "cbid": 211, "correlation": 169336
    }
  },
  {
    "ph": "s", "id": 169336, "pid": 76337, "tid": -914061504, "ts": 1716454224058927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058936, "dur": 0,
    "args": {
      "External id": 169337,
      "cbid": 51, "correlation": 169337
    }
  },
  {
    "ph": "s", "id": 169337, "pid": 76337, "tid": -914061504, "ts": 1716454224058936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224112833, "dur": 97,
    "args": {
      "External id": 169338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169338,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 169338, "pid": 5, "tid": 7, "ts": 1716454224112833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058937, "dur": 5,
    "args": {
      "External id": 169338,
      "cbid": 211, "correlation": 169338
    }
  },
  {
    "ph": "s", "id": 169338, "pid": 76337, "tid": -914061504, "ts": 1716454224058937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224112931, "dur": 17,
    "args": {
      "External id": 169343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169343,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169343, "pid": 5, "tid": 7, "ts": 1716454224112931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224058964, "dur": 18,
    "args": {
      "External id": 169343,
      "cbid": 211, "correlation": 169343
    }
  },
  {
    "ph": "s", "id": 169343, "pid": 76337, "tid": -914061504, "ts": 1716454224058964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224112949, "dur": 11,
    "args": {
      "External id": 169351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169351,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169351, "pid": 5, "tid": 7, "ts": 1716454224112949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059005, "dur": 10,
    "args": {
      "External id": 169351,
      "cbid": 211, "correlation": 169351
    }
  },
  {
    "ph": "s", "id": 169351, "pid": 76337, "tid": -914061504, "ts": 1716454224059005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224112962, "dur": 18,
    "args": {
      "External id": 169371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169371,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 169371, "pid": 5, "tid": 7, "ts": 1716454224112962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059080, "dur": 12,
    "args": {
      "External id": 169371,
      "cbid": 211, "correlation": 169371
    }
  },
  {
    "ph": "s", "id": 169371, "pid": 76337, "tid": -914061504, "ts": 1716454224059080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224112982, "dur": 5,
    "args": {
      "External id": 169383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169383,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 169383, "pid": 5, "tid": 7, "ts": 1716454224112982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059102, "dur": 6,
    "args": {
      "External id": 169383,
      "cbid": 211, "correlation": 169383
    }
  },
  {
    "ph": "s", "id": 169383, "pid": 76337, "tid": -914061504, "ts": 1716454224059102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224112988, "dur": 18,
    "args": {
      "External id": 169386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169386, "pid": 5, "tid": 7, "ts": 1716454224112988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059121, "dur": 6,
    "args": {
      "External id": 169386,
      "cbid": 211, "correlation": 169386
    }
  },
  {
    "ph": "s", "id": 169386, "pid": 76337, "tid": -914061504, "ts": 1716454224059121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224113007, "dur": 12,
    "args": {
      "External id": 169395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169395,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169395, "pid": 5, "tid": 7, "ts": 1716454224113007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059158, "dur": 10,
    "args": {
      "External id": 169395,
      "cbid": 211, "correlation": 169395
    }
  },
  {
    "ph": "s", "id": 169395, "pid": 76337, "tid": -914061504, "ts": 1716454224059158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059209, "dur": 0,
    "args": {
      "External id": 169405,
      "cbid": 317, "correlation": 169405
    }
  },
  {
    "ph": "f", "id": 169405, "pid": 76337, "tid": -914061504, "ts": 1716454224059209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059210, "dur": 0,
    "args": {
      "External id": 169406,
      "cbid": 203, "correlation": 169406
    }
  },
  {
    "ph": "f", "id": 169406, "pid": 76337, "tid": -914061504, "ts": 1716454224059210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059211, "dur": 0,
    "args": {
      "External id": 169407,
      "cbid": 205, "correlation": 169407
    }
  },
  {
    "ph": "f", "id": 169407, "pid": 76337, "tid": -914061504, "ts": 1716454224059211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224113020, "dur": 11,
    "args": {
      "External id": 169411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169411,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169411, "pid": 5, "tid": 7, "ts": 1716454224113020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059226, "dur": 12,
    "args": {
      "External id": 169411,
      "cbid": 211, "correlation": 169411
    }
  },
  {
    "ph": "s", "id": 169411, "pid": 76337, "tid": -914061504, "ts": 1716454224059226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224113033, "dur": 165,
    "args": {
      "External id": 169413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169413,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169413, "pid": 5, "tid": 7, "ts": 1716454224113033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059240, "dur": 5,
    "args": {
      "External id": 169413,
      "cbid": 211, "correlation": 169413
    }
  },
  {
    "ph": "s", "id": 169413, "pid": 76337, "tid": -914061504, "ts": 1716454224059240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224113201, "dur": 1,
    "args": {
      "External id": 169415,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 169415,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 169415, "pid": 5, "tid": 7, "ts": 1716454224113201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059251, "dur": 8,
    "args": {
      "External id": 169415,
      "cbid": 51, "correlation": 169415
    }
  },
  {
    "ph": "s", "id": 169415, "pid": 76337, "tid": -914061504, "ts": 1716454224059251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224113204, "dur": 669,
    "args": {
      "External id": 169416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169416,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169416, "pid": 5, "tid": 7, "ts": 1716454224113204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059260, "dur": 7,
    "args": {
      "External id": 169416,
      "cbid": 211, "correlation": 169416
    }
  },
  {
    "ph": "s", "id": 169416, "pid": 76337, "tid": -914061504, "ts": 1716454224059260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224113875, "dur": 15,
    "args": {
      "External id": 169418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169418,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169418, "pid": 5, "tid": 7, "ts": 1716454224113875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059271, "dur": 5,
    "args": {
      "External id": 169418,
      "cbid": 211, "correlation": 169418
    }
  },
  {
    "ph": "s", "id": 169418, "pid": 76337, "tid": -914061504, "ts": 1716454224059271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224113891, "dur": 15,
    "args": {
      "External id": 169424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169424,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169424, "pid": 5, "tid": 7, "ts": 1716454224113891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059299, "dur": 9,
    "args": {
      "External id": 169424,
      "cbid": 211, "correlation": 169424
    }
  },
  {
    "ph": "s", "id": 169424, "pid": 76337, "tid": -914061504, "ts": 1716454224059299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224113907, "dur": 4,
    "args": {
      "External id": 169432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169432,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 169432, "pid": 5, "tid": 7, "ts": 1716454224113907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059342, "dur": 9,
    "args": {
      "External id": 169432,
      "cbid": 211, "correlation": 169432
    }
  },
  {
    "ph": "s", "id": 169432, "pid": 76337, "tid": -914061504, "ts": 1716454224059342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059407, "dur": 1,
    "args": {
      "External id": 169448,
      "cbid": 251, "correlation": 169448
    }
  },
  {
    "ph": "f", "id": 169448, "pid": 76337, "tid": -914061504, "ts": 1716454224059407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059412, "dur": 0,
    "args": {
      "External id": 169450,
      "cbid": 251, "correlation": 169450
    }
  },
  {
    "ph": "f", "id": 169450, "pid": 76337, "tid": -914061504, "ts": 1716454224059412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224113912, "dur": 13,
    "args": {
      "External id": 169451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169451,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169451, "pid": 5, "tid": 7, "ts": 1716454224113912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059414, "dur": 12,
    "args": {
      "External id": 169451,
      "cbid": 211, "correlation": 169451
    }
  },
  {
    "ph": "s", "id": 169451, "pid": 76337, "tid": -914061504, "ts": 1716454224059414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224113927, "dur": 5,
    "args": {
      "External id": 169453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169453,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169453, "pid": 5, "tid": 7, "ts": 1716454224113927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059428, "dur": 6,
    "args": {
      "External id": 169453,
      "cbid": 211, "correlation": 169453
    }
  },
  {
    "ph": "s", "id": 169453, "pid": 76337, "tid": -914061504, "ts": 1716454224059428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224113933, "dur": 17,
    "args": {
      "External id": 169463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169463, "pid": 5, "tid": 7, "ts": 1716454224113933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059486, "dur": 12,
    "args": {
      "External id": 169463,
      "cbid": 211, "correlation": 169463
    }
  },
  {
    "ph": "s", "id": 169463, "pid": 76337, "tid": -914061504, "ts": 1716454224059486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224113951, "dur": 18,
    "args": {
      "External id": 169483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169483,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 169483, "pid": 5, "tid": 7, "ts": 1716454224113951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059552, "dur": 11,
    "args": {
      "External id": 169483,
      "cbid": 211, "correlation": 169483
    }
  },
  {
    "ph": "s", "id": 169483, "pid": 76337, "tid": -914061504, "ts": 1716454224059552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224113970, "dur": 5,
    "args": {
      "External id": 169495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169495,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 169495, "pid": 5, "tid": 7, "ts": 1716454224113970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059574, "dur": 7,
    "args": {
      "External id": 169495,
      "cbid": 211, "correlation": 169495
    }
  },
  {
    "ph": "s", "id": 169495, "pid": 76337, "tid": -914061504, "ts": 1716454224059574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224113976, "dur": 17,
    "args": {
      "External id": 169498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169498, "pid": 5, "tid": 7, "ts": 1716454224113976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059593, "dur": 6,
    "args": {
      "External id": 169498,
      "cbid": 211, "correlation": 169498
    }
  },
  {
    "ph": "s", "id": 169498, "pid": 76337, "tid": -914061504, "ts": 1716454224059593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224113995, "dur": 11,
    "args": {
      "External id": 169507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169507,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169507, "pid": 5, "tid": 7, "ts": 1716454224113995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059633, "dur": 9,
    "args": {
      "External id": 169507,
      "cbid": 211, "correlation": 169507
    }
  },
  {
    "ph": "s", "id": 169507, "pid": 76337, "tid": -914061504, "ts": 1716454224059633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059695, "dur": 0,
    "args": {
      "External id": 169517,
      "cbid": 317, "correlation": 169517
    }
  },
  {
    "ph": "f", "id": 169517, "pid": 76337, "tid": -914061504, "ts": 1716454224059695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059695, "dur": 0,
    "args": {
      "External id": 169518,
      "cbid": 203, "correlation": 169518
    }
  },
  {
    "ph": "f", "id": 169518, "pid": 76337, "tid": -914061504, "ts": 1716454224059695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059696, "dur": 0,
    "args": {
      "External id": 169519,
      "cbid": 205, "correlation": 169519
    }
  },
  {
    "ph": "f", "id": 169519, "pid": 76337, "tid": -914061504, "ts": 1716454224059696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224114007, "dur": 11,
    "args": {
      "External id": 169523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169523,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169523, "pid": 5, "tid": 7, "ts": 1716454224114007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059710, "dur": 13,
    "args": {
      "External id": 169523,
      "cbid": 211, "correlation": 169523
    }
  },
  {
    "ph": "s", "id": 169523, "pid": 76337, "tid": -914061504, "ts": 1716454224059710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224114020, "dur": 164,
    "args": {
      "External id": 169525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169525,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169525, "pid": 5, "tid": 7, "ts": 1716454224114020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059726, "dur": 5,
    "args": {
      "External id": 169525,
      "cbid": 211, "correlation": 169525
    }
  },
  {
    "ph": "s", "id": 169525, "pid": 76337, "tid": -914061504, "ts": 1716454224059726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224114186, "dur": 1,
    "args": {
      "External id": 169527,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 169527,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 169527, "pid": 5, "tid": 7, "ts": 1716454224114186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059737, "dur": 6,
    "args": {
      "External id": 169527,
      "cbid": 51, "correlation": 169527
    }
  },
  {
    "ph": "s", "id": 169527, "pid": 76337, "tid": -914061504, "ts": 1716454224059737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224114190, "dur": 652,
    "args": {
      "External id": 169528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169528,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169528, "pid": 5, "tid": 7, "ts": 1716454224114190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059744, "dur": 6,
    "args": {
      "External id": 169528,
      "cbid": 211, "correlation": 169528
    }
  },
  {
    "ph": "s", "id": 169528, "pid": 76337, "tid": -914061504, "ts": 1716454224059744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224114844, "dur": 13,
    "args": {
      "External id": 169530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169530,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169530, "pid": 5, "tid": 7, "ts": 1716454224114844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059753, "dur": 5,
    "args": {
      "External id": 169530,
      "cbid": 211, "correlation": 169530
    }
  },
  {
    "ph": "s", "id": 169530, "pid": 76337, "tid": -914061504, "ts": 1716454224059753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224114858, "dur": 15,
    "args": {
      "External id": 169536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169536, "pid": 5, "tid": 7, "ts": 1716454224114858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059782, "dur": 8,
    "args": {
      "External id": 169536,
      "cbid": 211, "correlation": 169536
    }
  },
  {
    "ph": "s", "id": 169536, "pid": 76337, "tid": -914061504, "ts": 1716454224059782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224114874, "dur": 12,
    "args": {
      "External id": 169544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169544,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169544, "pid": 5, "tid": 7, "ts": 1716454224114874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059813, "dur": 8,
    "args": {
      "External id": 169544,
      "cbid": 211, "correlation": 169544
    }
  },
  {
    "ph": "s", "id": 169544, "pid": 76337, "tid": -914061504, "ts": 1716454224059813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224114888, "dur": 10,
    "args": {
      "External id": 169552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169552,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169552, "pid": 5, "tid": 7, "ts": 1716454224114888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059842, "dur": 8,
    "args": {
      "External id": 169552,
      "cbid": 211, "correlation": 169552
    }
  },
  {
    "ph": "s", "id": 169552, "pid": 76337, "tid": -914061504, "ts": 1716454224059842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224114899, "dur": 18,
    "args": {
      "External id": 169572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169572,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 169572, "pid": 5, "tid": 7, "ts": 1716454224114899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059920, "dur": 13,
    "args": {
      "External id": 169572,
      "cbid": 211, "correlation": 169572
    }
  },
  {
    "ph": "s", "id": 169572, "pid": 76337, "tid": -914061504, "ts": 1716454224059920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224114919, "dur": 4,
    "args": {
      "External id": 169584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169584,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 169584, "pid": 5, "tid": 7, "ts": 1716454224114919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059942, "dur": 6,
    "args": {
      "External id": 169584,
      "cbid": 211, "correlation": 169584
    }
  },
  {
    "ph": "s", "id": 169584, "pid": 76337, "tid": -914061504, "ts": 1716454224059942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224114924, "dur": 16,
    "args": {
      "External id": 169587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169587,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169587, "pid": 5, "tid": 7, "ts": 1716454224114924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224059960, "dur": 7,
    "args": {
      "External id": 169587,
      "cbid": 211, "correlation": 169587
    }
  },
  {
    "ph": "s", "id": 169587, "pid": 76337, "tid": -914061504, "ts": 1716454224059960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060031, "dur": 0,
    "args": {
      "External id": 169598,
      "cbid": 317, "correlation": 169598
    }
  },
  {
    "ph": "f", "id": 169598, "pid": 76337, "tid": -914061504, "ts": 1716454224060031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060032, "dur": 0,
    "args": {
      "External id": 169599,
      "cbid": 203, "correlation": 169599
    }
  },
  {
    "ph": "f", "id": 169599, "pid": 76337, "tid": -914061504, "ts": 1716454224060032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060032, "dur": 0,
    "args": {
      "External id": 169600,
      "cbid": 205, "correlation": 169600
    }
  },
  {
    "ph": "f", "id": 169600, "pid": 76337, "tid": -914061504, "ts": 1716454224060032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224114942, "dur": 11,
    "args": {
      "External id": 169604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169604,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169604, "pid": 5, "tid": 7, "ts": 1716454224114942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060045, "dur": 12,
    "args": {
      "External id": 169604,
      "cbid": 211, "correlation": 169604
    }
  },
  {
    "ph": "s", "id": 169604, "pid": 76337, "tid": -914061504, "ts": 1716454224060045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224114954, "dur": 4,
    "args": {
      "External id": 169606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169606,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169606, "pid": 5, "tid": 7, "ts": 1716454224114954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060062, "dur": 6,
    "args": {
      "External id": 169606,
      "cbid": 211, "correlation": 169606
    }
  },
  {
    "ph": "s", "id": 169606, "pid": 76337, "tid": -914061504, "ts": 1716454224060062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060070, "dur": 0,
    "args": {
      "External id": 169607,
      "cbid": 51, "correlation": 169607
    }
  },
  {
    "ph": "s", "id": 169607, "pid": 76337, "tid": -914061504, "ts": 1716454224060070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224114960, "dur": 96,
    "args": {
      "External id": 169608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169608,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 169608, "pid": 5, "tid": 7, "ts": 1716454224114960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060071, "dur": 5,
    "args": {
      "External id": 169608,
      "cbid": 211, "correlation": 169608
    }
  },
  {
    "ph": "s", "id": 169608, "pid": 76337, "tid": -914061504, "ts": 1716454224060071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224115057, "dur": 15,
    "args": {
      "External id": 169613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169613,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169613, "pid": 5, "tid": 7, "ts": 1716454224115057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060098, "dur": 8,
    "args": {
      "External id": 169613,
      "cbid": 211, "correlation": 169613
    }
  },
  {
    "ph": "s", "id": 169613, "pid": 76337, "tid": -914061504, "ts": 1716454224060098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224115074, "dur": 85,
    "args": {
      "External id": 169622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169622,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169622, "pid": 5, "tid": 7, "ts": 1716454224115074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060180, "dur": 14,
    "args": {
      "External id": 169622,
      "cbid": 211, "correlation": 169622
    }
  },
  {
    "ph": "s", "id": 169622, "pid": 76337, "tid": -914061504, "ts": 1716454224060180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224115160, "dur": 30,
    "args": {
      "External id": 169644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169644,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169644, "pid": 5, "tid": 7, "ts": 1716454224115160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060238, "dur": 10,
    "args": {
      "External id": 169644,
      "cbid": 211, "correlation": 169644
    }
  },
  {
    "ph": "s", "id": 169644, "pid": 76337, "tid": -914061504, "ts": 1716454224060238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060324, "dur": 2,
    "args": {
      "External id": 169655,
      "cbid": 251, "correlation": 169655
    }
  },
  {
    "ph": "f", "id": 169655, "pid": 76337, "tid": -914061504, "ts": 1716454224060324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224115191, "dur": 162,
    "args": {
      "External id": 169656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169656,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169656, "pid": 5, "tid": 7, "ts": 1716454224115191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060331, "dur": 13,
    "args": {
      "External id": 169656,
      "cbid": 211, "correlation": 169656
    }
  },
  {
    "ph": "s", "id": 169656, "pid": 76337, "tid": -914061504, "ts": 1716454224060331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060400, "dur": 1,
    "args": {
      "External id": 169667,
      "cbid": 251, "correlation": 169667
    }
  },
  {
    "ph": "f", "id": 169667, "pid": 76337, "tid": -914061504, "ts": 1716454224060400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224115354, "dur": 160,
    "args": {
      "External id": 169668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169668,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169668, "pid": 5, "tid": 7, "ts": 1716454224115354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060404, "dur": 12,
    "args": {
      "External id": 169668,
      "cbid": 211, "correlation": 169668
    }
  },
  {
    "ph": "s", "id": 169668, "pid": 76337, "tid": -914061504, "ts": 1716454224060404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060468, "dur": 1,
    "args": {
      "External id": 169679,
      "cbid": 251, "correlation": 169679
    }
  },
  {
    "ph": "f", "id": 169679, "pid": 76337, "tid": -914061504, "ts": 1716454224060468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224115516, "dur": 161,
    "args": {
      "External id": 169680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169680,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169680, "pid": 5, "tid": 7, "ts": 1716454224115516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060472, "dur": 12,
    "args": {
      "External id": 169680,
      "cbid": 211, "correlation": 169680
    }
  },
  {
    "ph": "s", "id": 169680, "pid": 76337, "tid": -914061504, "ts": 1716454224060472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224115678, "dur": 342,
    "args": {
      "External id": 169705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169705,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169705, "pid": 5, "tid": 7, "ts": 1716454224115678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060556, "dur": 12,
    "args": {
      "External id": 169705,
      "cbid": 211, "correlation": 169705
    }
  },
  {
    "ph": "s", "id": 169705, "pid": 76337, "tid": -914061504, "ts": 1716454224060556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060655, "dur": 1,
    "args": {
      "External id": 169723,
      "cbid": 251, "correlation": 169723
    }
  },
  {
    "ph": "f", "id": 169723, "pid": 76337, "tid": -914061504, "ts": 1716454224060655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224116022, "dur": 167,
    "args": {
      "External id": 169725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169725,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169725, "pid": 5, "tid": 7, "ts": 1716454224116022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060661, "dur": 14,
    "args": {
      "External id": 169725,
      "cbid": 211, "correlation": 169725
    }
  },
  {
    "ph": "s", "id": 169725, "pid": 76337, "tid": -914061504, "ts": 1716454224060661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224116190, "dur": 19,
    "args": {
      "External id": 169733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169733,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169733, "pid": 5, "tid": 7, "ts": 1716454224116190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060731, "dur": 13,
    "args": {
      "External id": 169733,
      "cbid": 211, "correlation": 169733
    }
  },
  {
    "ph": "s", "id": 169733, "pid": 76337, "tid": -914061504, "ts": 1716454224060731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224116211, "dur": 28,
    "args": {
      "External id": 169741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169741,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169741, "pid": 5, "tid": 7, "ts": 1716454224116211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060770, "dur": 8,
    "args": {
      "External id": 169741,
      "cbid": 211, "correlation": 169741
    }
  },
  {
    "ph": "s", "id": 169741, "pid": 76337, "tid": -914061504, "ts": 1716454224060770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224116240, "dur": 19,
    "args": {
      "External id": 169752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169752, "pid": 5, "tid": 7, "ts": 1716454224116240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060841, "dur": 13,
    "args": {
      "External id": 169752,
      "cbid": 211, "correlation": 169752
    }
  },
  {
    "ph": "s", "id": 169752, "pid": 76337, "tid": -914061504, "ts": 1716454224060841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224116260, "dur": 16,
    "args": {
      "External id": 169774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169774,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169774, "pid": 5, "tid": 7, "ts": 1716454224116260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060872, "dur": 7,
    "args": {
      "External id": 169774,
      "cbid": 211, "correlation": 169774
    }
  },
  {
    "ph": "s", "id": 169774, "pid": 76337, "tid": -914061504, "ts": 1716454224060872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060957, "dur": 1,
    "args": {
      "External id": 169785,
      "cbid": 251, "correlation": 169785
    }
  },
  {
    "ph": "f", "id": 169785, "pid": 76337, "tid": -914061504, "ts": 1716454224060957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224116278, "dur": 90,
    "args": {
      "External id": 169786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169786,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169786, "pid": 5, "tid": 7, "ts": 1716454224116278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224060963, "dur": 25,
    "args": {
      "External id": 169786,
      "cbid": 211, "correlation": 169786
    }
  },
  {
    "ph": "s", "id": 169786, "pid": 76337, "tid": -914061504, "ts": 1716454224060963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061043, "dur": 1,
    "args": {
      "External id": 169797,
      "cbid": 251, "correlation": 169797
    }
  },
  {
    "ph": "f", "id": 169797, "pid": 76337, "tid": -914061504, "ts": 1716454224061043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061047, "dur": 0,
    "args": {
      "External id": 169798,
      "cbid": 251, "correlation": 169798
    }
  },
  {
    "ph": "f", "id": 169798, "pid": 76337, "tid": -914061504, "ts": 1716454224061047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224116369, "dur": 12,
    "args": {
      "External id": 169799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169799,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169799, "pid": 5, "tid": 7, "ts": 1716454224116369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061049, "dur": 12,
    "args": {
      "External id": 169799,
      "cbid": 211, "correlation": 169799
    }
  },
  {
    "ph": "s", "id": 169799, "pid": 76337, "tid": -914061504, "ts": 1716454224061049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224116382, "dur": 6,
    "args": {
      "External id": 169801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169801,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169801, "pid": 5, "tid": 7, "ts": 1716454224116382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061063, "dur": 6,
    "args": {
      "External id": 169801,
      "cbid": 211, "correlation": 169801
    }
  },
  {
    "ph": "s", "id": 169801, "pid": 76337, "tid": -914061504, "ts": 1716454224061063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061119, "dur": 1,
    "args": {
      "External id": 169812,
      "cbid": 251, "correlation": 169812
    }
  },
  {
    "ph": "f", "id": 169812, "pid": 76337, "tid": -914061504, "ts": 1716454224061119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061123, "dur": 0,
    "args": {
      "External id": 169813,
      "cbid": 251, "correlation": 169813
    }
  },
  {
    "ph": "f", "id": 169813, "pid": 76337, "tid": -914061504, "ts": 1716454224061123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224116390, "dur": 8,
    "args": {
      "External id": 169814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169814,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169814, "pid": 5, "tid": 7, "ts": 1716454224116390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061124, "dur": 12,
    "args": {
      "External id": 169814,
      "cbid": 211, "correlation": 169814
    }
  },
  {
    "ph": "s", "id": 169814, "pid": 76337, "tid": -914061504, "ts": 1716454224061124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224116399, "dur": 4,
    "args": {
      "External id": 169816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169816,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169816, "pid": 5, "tid": 7, "ts": 1716454224116399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061138, "dur": 5,
    "args": {
      "External id": 169816,
      "cbid": 211, "correlation": 169816
    }
  },
  {
    "ph": "s", "id": 169816, "pid": 76337, "tid": -914061504, "ts": 1716454224061138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224116404, "dur": 55,
    "args": {
      "External id": 169841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169841,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169841, "pid": 5, "tid": 7, "ts": 1716454224116404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061214, "dur": 12,
    "args": {
      "External id": 169841,
      "cbid": 211, "correlation": 169841
    }
  },
  {
    "ph": "s", "id": 169841, "pid": 76337, "tid": -914061504, "ts": 1716454224061214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061312, "dur": 1,
    "args": {
      "External id": 169859,
      "cbid": 251, "correlation": 169859
    }
  },
  {
    "ph": "f", "id": 169859, "pid": 76337, "tid": -914061504, "ts": 1716454224061312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224116460, "dur": 91,
    "args": {
      "External id": 169861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169861,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169861, "pid": 5, "tid": 7, "ts": 1716454224116460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061317, "dur": 14,
    "args": {
      "External id": 169861,
      "cbid": 211, "correlation": 169861
    }
  },
  {
    "ph": "s", "id": 169861, "pid": 76337, "tid": -914061504, "ts": 1716454224061317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224116553, "dur": 10,
    "args": {
      "External id": 169869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169869,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169869, "pid": 5, "tid": 7, "ts": 1716454224116553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061388, "dur": 12,
    "args": {
      "External id": 169869,
      "cbid": 211, "correlation": 169869
    }
  },
  {
    "ph": "s", "id": 169869, "pid": 76337, "tid": -914061504, "ts": 1716454224061388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224116564, "dur": 21,
    "args": {
      "External id": 169877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169877,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169877, "pid": 5, "tid": 7, "ts": 1716454224116564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061428, "dur": 10,
    "args": {
      "External id": 169877,
      "cbid": 211, "correlation": 169877
    }
  },
  {
    "ph": "s", "id": 169877, "pid": 76337, "tid": -914061504, "ts": 1716454224061428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224116586, "dur": 18,
    "args": {
      "External id": 169899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169899,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169899, "pid": 5, "tid": 7, "ts": 1716454224116586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061480, "dur": 11,
    "args": {
      "External id": 169899,
      "cbid": 211, "correlation": 169899
    }
  },
  {
    "ph": "s", "id": 169899, "pid": 76337, "tid": -914061504, "ts": 1716454224061480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061567, "dur": 1,
    "args": {
      "External id": 169915,
      "cbid": 251, "correlation": 169915
    }
  },
  {
    "ph": "f", "id": 169915, "pid": 76337, "tid": -914061504, "ts": 1716454224061567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061572, "dur": 0,
    "args": {
      "External id": 169917,
      "cbid": 251, "correlation": 169917
    }
  },
  {
    "ph": "f", "id": 169917, "pid": 76337, "tid": -914061504, "ts": 1716454224061572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224116605, "dur": 495,
    "args": {
      "External id": 169918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169918,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 169918, "pid": 5, "tid": 7, "ts": 1716454224116605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061574, "dur": 13,
    "args": {
      "External id": 169918,
      "cbid": 211, "correlation": 169918
    }
  },
  {
    "ph": "s", "id": 169918, "pid": 76337, "tid": -914061504, "ts": 1716454224061574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224117102, "dur": 66,
    "args": {
      "External id": 169926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169926, "pid": 5, "tid": 7, "ts": 1716454224117102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061638, "dur": 13,
    "args": {
      "External id": 169926,
      "cbid": 211, "correlation": 169926
    }
  },
  {
    "ph": "s", "id": 169926, "pid": 76337, "tid": -914061504, "ts": 1716454224061638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224117169, "dur": 68,
    "args": {
      "External id": 169934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169934, "pid": 5, "tid": 7, "ts": 1716454224117169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061669, "dur": 8,
    "args": {
      "External id": 169934,
      "cbid": 211, "correlation": 169934
    }
  },
  {
    "ph": "s", "id": 169934, "pid": 76337, "tid": -914061504, "ts": 1716454224061669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061748, "dur": 1,
    "args": {
      "External id": 169950,
      "cbid": 251, "correlation": 169950
    }
  },
  {
    "ph": "f", "id": 169950, "pid": 76337, "tid": -914061504, "ts": 1716454224061748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224117239, "dur": 1,
    "args": {
      "External id": 169952,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 169952,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 169952, "pid": 5, "tid": 7, "ts": 1716454224117239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061753, "dur": 9,
    "args": {
      "External id": 169952,
      "cbid": 51, "correlation": 169952
    }
  },
  {
    "ph": "s", "id": 169952, "pid": 76337, "tid": -914061504, "ts": 1716454224061753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224117243, "dur": 272,
    "args": {
      "External id": 169953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169953,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169953, "pid": 5, "tid": 7, "ts": 1716454224117243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061763, "dur": 11,
    "args": {
      "External id": 169953,
      "cbid": 211, "correlation": 169953
    }
  },
  {
    "ph": "s", "id": 169953, "pid": 76337, "tid": -914061504, "ts": 1716454224061763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224117516, "dur": 14,
    "args": {
      "External id": 169961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169961,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169961, "pid": 5, "tid": 7, "ts": 1716454224117516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061805, "dur": 10,
    "args": {
      "External id": 169961,
      "cbid": 211, "correlation": 169961
    }
  },
  {
    "ph": "s", "id": 169961, "pid": 76337, "tid": -914061504, "ts": 1716454224061805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224117532, "dur": 38,
    "args": {
      "External id": 169972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169972, "pid": 5, "tid": 7, "ts": 1716454224117532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061873, "dur": 12,
    "args": {
      "External id": 169972,
      "cbid": 211, "correlation": 169972
    }
  },
  {
    "ph": "s", "id": 169972, "pid": 76337, "tid": -914061504, "ts": 1716454224061873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061937, "dur": 0,
    "args": {
      "External id": 169984,
      "cbid": 317, "correlation": 169984
    }
  },
  {
    "ph": "f", "id": 169984, "pid": 76337, "tid": -914061504, "ts": 1716454224061937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061937, "dur": 0,
    "args": {
      "External id": 169985,
      "cbid": 203, "correlation": 169985
    }
  },
  {
    "ph": "f", "id": 169985, "pid": 76337, "tid": -914061504, "ts": 1716454224061937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061938, "dur": 0,
    "args": {
      "External id": 169986,
      "cbid": 205, "correlation": 169986
    }
  },
  {
    "ph": "f", "id": 169986, "pid": 76337, "tid": -914061504, "ts": 1716454224061938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224117571, "dur": 13,
    "args": {
      "External id": 169990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169990,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169990, "pid": 5, "tid": 7, "ts": 1716454224117571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061954, "dur": 12,
    "args": {
      "External id": 169990,
      "cbid": 211, "correlation": 169990
    }
  },
  {
    "ph": "s", "id": 169990, "pid": 76337, "tid": -914061504, "ts": 1716454224061954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224117585, "dur": 4,
    "args": {
      "External id": 169992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169992,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 169992, "pid": 5, "tid": 7, "ts": 1716454224117585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061971, "dur": 15,
    "args": {
      "External id": 169992,
      "cbid": 211, "correlation": 169992
    }
  },
  {
    "ph": "s", "id": 169992, "pid": 76337, "tid": -914061504, "ts": 1716454224061971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061989, "dur": 0,
    "args": {
      "External id": 169993,
      "cbid": 51, "correlation": 169993
    }
  },
  {
    "ph": "s", "id": 169993, "pid": 76337, "tid": -914061504, "ts": 1716454224061989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224117591, "dur": 99,
    "args": {
      "External id": 169994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169994,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 169994, "pid": 5, "tid": 7, "ts": 1716454224117591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224061990, "dur": 6,
    "args": {
      "External id": 169994,
      "cbid": 211, "correlation": 169994
    }
  },
  {
    "ph": "s", "id": 169994, "pid": 76337, "tid": -914061504, "ts": 1716454224061990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224117691, "dur": 16,
    "args": {
      "External id": 169999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 169999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 169999, "pid": 5, "tid": 7, "ts": 1716454224117691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062018, "dur": 9,
    "args": {
      "External id": 169999,
      "cbid": 211, "correlation": 169999
    }
  },
  {
    "ph": "s", "id": 169999, "pid": 76337, "tid": -914061504, "ts": 1716454224062018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224117709, "dur": 14,
    "args": {
      "External id": 170007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170007,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170007, "pid": 5, "tid": 7, "ts": 1716454224117709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062051, "dur": 8,
    "args": {
      "External id": 170007,
      "cbid": 211, "correlation": 170007
    }
  },
  {
    "ph": "s", "id": 170007, "pid": 76337, "tid": -914061504, "ts": 1716454224062051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062122, "dur": 0,
    "args": {
      "External id": 170017,
      "cbid": 317, "correlation": 170017
    }
  },
  {
    "ph": "f", "id": 170017, "pid": 76337, "tid": -914061504, "ts": 1716454224062122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062123, "dur": 0,
    "args": {
      "External id": 170018,
      "cbid": 203, "correlation": 170018
    }
  },
  {
    "ph": "f", "id": 170018, "pid": 76337, "tid": -914061504, "ts": 1716454224062123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062124, "dur": 0,
    "args": {
      "External id": 170019,
      "cbid": 205, "correlation": 170019
    }
  },
  {
    "ph": "f", "id": 170019, "pid": 76337, "tid": -914061504, "ts": 1716454224062124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224117724, "dur": 13,
    "args": {
      "External id": 170023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170023, "pid": 5, "tid": 7, "ts": 1716454224117724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062138, "dur": 13,
    "args": {
      "External id": 170023,
      "cbid": 211, "correlation": 170023
    }
  },
  {
    "ph": "s", "id": 170023, "pid": 76337, "tid": -914061504, "ts": 1716454224062138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224117738, "dur": 164,
    "args": {
      "External id": 170025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170025,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170025, "pid": 5, "tid": 7, "ts": 1716454224117738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062153, "dur": 5,
    "args": {
      "External id": 170025,
      "cbid": 211, "correlation": 170025
    }
  },
  {
    "ph": "s", "id": 170025, "pid": 76337, "tid": -914061504, "ts": 1716454224062153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224117904, "dur": 1,
    "args": {
      "External id": 170027,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170027,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5765765765765766
    }
  },
  {
    "ph": "f", "id": 170027, "pid": 5, "tid": 7, "ts": 1716454224117904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062165, "dur": 6,
    "args": {
      "External id": 170027,
      "cbid": 51, "correlation": 170027
    }
  },
  {
    "ph": "s", "id": 170027, "pid": 76337, "tid": -914061504, "ts": 1716454224062165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224117908, "dur": 199,
    "args": {
      "External id": 170028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170028,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 170028, "pid": 5, "tid": 7, "ts": 1716454224117908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062172, "dur": 8,
    "args": {
      "External id": 170028,
      "cbid": 211, "correlation": 170028
    }
  },
  {
    "ph": "s", "id": 170028, "pid": 76337, "tid": -914061504, "ts": 1716454224062172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118109, "dur": 6,
    "args": {
      "External id": 170030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170030,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170030, "pid": 5, "tid": 7, "ts": 1716454224118109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062185, "dur": 5,
    "args": {
      "External id": 170030,
      "cbid": 211, "correlation": 170030
    }
  },
  {
    "ph": "s", "id": 170030, "pid": 76337, "tid": -914061504, "ts": 1716454224062185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224118117, "dur": 6,
    "args": {
      "External id": 170036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170036, "pid": 5, "tid": 7, "ts": 1716454224118117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062213, "dur": 9,
    "args": {
      "External id": 170036,
      "cbid": 211, "correlation": 170036
    }
  },
  {
    "ph": "s", "id": 170036, "pid": 76337, "tid": -914061504, "ts": 1716454224062213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118124, "dur": 11,
    "args": {
      "External id": 170056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170056,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170056, "pid": 5, "tid": 7, "ts": 1716454224118124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062309, "dur": 12,
    "args": {
      "External id": 170056,
      "cbid": 211, "correlation": 170056
    }
  },
  {
    "ph": "s", "id": 170056, "pid": 76337, "tid": -914061504, "ts": 1716454224062309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224118136, "dur": 4,
    "args": {
      "External id": 170068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170068,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170068, "pid": 5, "tid": 7, "ts": 1716454224118136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062331, "dur": 6,
    "args": {
      "External id": 170068,
      "cbid": 211, "correlation": 170068
    }
  },
  {
    "ph": "s", "id": 170068, "pid": 76337, "tid": -914061504, "ts": 1716454224062331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224118142, "dur": 9,
    "args": {
      "External id": 170071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170071,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170071, "pid": 5, "tid": 7, "ts": 1716454224118142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062351, "dur": 6,
    "args": {
      "External id": 170071,
      "cbid": 211, "correlation": 170071
    }
  },
  {
    "ph": "s", "id": 170071, "pid": 76337, "tid": -914061504, "ts": 1716454224062351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224118152, "dur": 5,
    "args": {
      "External id": 170080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170080,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170080, "pid": 5, "tid": 7, "ts": 1716454224118152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062390, "dur": 10,
    "args": {
      "External id": 170080,
      "cbid": 211, "correlation": 170080
    }
  },
  {
    "ph": "s", "id": 170080, "pid": 76337, "tid": -914061504, "ts": 1716454224062390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062441, "dur": 0,
    "args": {
      "External id": 170090,
      "cbid": 317, "correlation": 170090
    }
  },
  {
    "ph": "f", "id": 170090, "pid": 76337, "tid": -914061504, "ts": 1716454224062441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062442, "dur": 0,
    "args": {
      "External id": 170091,
      "cbid": 203, "correlation": 170091
    }
  },
  {
    "ph": "f", "id": 170091, "pid": 76337, "tid": -914061504, "ts": 1716454224062442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062442, "dur": 0,
    "args": {
      "External id": 170092,
      "cbid": 205, "correlation": 170092
    }
  },
  {
    "ph": "f", "id": 170092, "pid": 76337, "tid": -914061504, "ts": 1716454224062442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118159, "dur": 5,
    "args": {
      "External id": 170096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170096, "pid": 5, "tid": 7, "ts": 1716454224118159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062457, "dur": 11,
    "args": {
      "External id": 170096,
      "cbid": 211, "correlation": 170096
    }
  },
  {
    "ph": "s", "id": 170096, "pid": 76337, "tid": -914061504, "ts": 1716454224062457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118165, "dur": 163,
    "args": {
      "External id": 170098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170098,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170098, "pid": 5, "tid": 7, "ts": 1716454224118165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062471, "dur": 5,
    "args": {
      "External id": 170098,
      "cbid": 211, "correlation": 170098
    }
  },
  {
    "ph": "s", "id": 170098, "pid": 76337, "tid": -914061504, "ts": 1716454224062471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224118330, "dur": 1,
    "args": {
      "External id": 170100,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170100,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 170100, "pid": 5, "tid": 7, "ts": 1716454224118330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062483, "dur": 7,
    "args": {
      "External id": 170100,
      "cbid": 51, "correlation": 170100
    }
  },
  {
    "ph": "s", "id": 170100, "pid": 76337, "tid": -914061504, "ts": 1716454224062483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224118334, "dur": 272,
    "args": {
      "External id": 170101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170101,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170101, "pid": 5, "tid": 7, "ts": 1716454224118334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062491, "dur": 7,
    "args": {
      "External id": 170101,
      "cbid": 211, "correlation": 170101
    }
  },
  {
    "ph": "s", "id": 170101, "pid": 76337, "tid": -914061504, "ts": 1716454224062491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118607, "dur": 6,
    "args": {
      "External id": 170103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170103,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170103, "pid": 5, "tid": 7, "ts": 1716454224118607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062502, "dur": 5,
    "args": {
      "External id": 170103,
      "cbid": 211, "correlation": 170103
    }
  },
  {
    "ph": "s", "id": 170103, "pid": 76337, "tid": -914061504, "ts": 1716454224062502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224118614, "dur": 6,
    "args": {
      "External id": 170109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170109,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170109, "pid": 5, "tid": 7, "ts": 1716454224118614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062529, "dur": 8,
    "args": {
      "External id": 170109,
      "cbid": 211, "correlation": 170109
    }
  },
  {
    "ph": "s", "id": 170109, "pid": 76337, "tid": -914061504, "ts": 1716454224062529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224118622, "dur": 4,
    "args": {
      "External id": 170117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170117,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 170117, "pid": 5, "tid": 7, "ts": 1716454224118622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062573, "dur": 9,
    "args": {
      "External id": 170117,
      "cbid": 211, "correlation": 170117
    }
  },
  {
    "ph": "s", "id": 170117, "pid": 76337, "tid": -914061504, "ts": 1716454224062573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062636, "dur": 1,
    "args": {
      "External id": 170133,
      "cbid": 251, "correlation": 170133
    }
  },
  {
    "ph": "f", "id": 170133, "pid": 76337, "tid": -914061504, "ts": 1716454224062636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062641, "dur": 0,
    "args": {
      "External id": 170135,
      "cbid": 251, "correlation": 170135
    }
  },
  {
    "ph": "f", "id": 170135, "pid": 76337, "tid": -914061504, "ts": 1716454224062641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224118627, "dur": 13,
    "args": {
      "External id": 170136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170136,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170136, "pid": 5, "tid": 7, "ts": 1716454224118627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062644, "dur": 12,
    "args": {
      "External id": 170136,
      "cbid": 211, "correlation": 170136
    }
  },
  {
    "ph": "s", "id": 170136, "pid": 76337, "tid": -914061504, "ts": 1716454224062644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224118642, "dur": 5,
    "args": {
      "External id": 170138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170138,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170138, "pid": 5, "tid": 7, "ts": 1716454224118642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062658, "dur": 5,
    "args": {
      "External id": 170138,
      "cbid": 211, "correlation": 170138
    }
  },
  {
    "ph": "s", "id": 170138, "pid": 76337, "tid": -914061504, "ts": 1716454224062658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224118649, "dur": 6,
    "args": {
      "External id": 170148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170148,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170148, "pid": 5, "tid": 7, "ts": 1716454224118649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062715, "dur": 12,
    "args": {
      "External id": 170148,
      "cbid": 211, "correlation": 170148
    }
  },
  {
    "ph": "s", "id": 170148, "pid": 76337, "tid": -914061504, "ts": 1716454224062715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118656, "dur": 10,
    "args": {
      "External id": 170168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170168,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170168, "pid": 5, "tid": 7, "ts": 1716454224118656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062781, "dur": 11,
    "args": {
      "External id": 170168,
      "cbid": 211, "correlation": 170168
    }
  },
  {
    "ph": "s", "id": 170168, "pid": 76337, "tid": -914061504, "ts": 1716454224062781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224118667, "dur": 4,
    "args": {
      "External id": 170180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170180,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170180, "pid": 5, "tid": 7, "ts": 1716454224118667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062803, "dur": 6,
    "args": {
      "External id": 170180,
      "cbid": 211, "correlation": 170180
    }
  },
  {
    "ph": "s", "id": 170180, "pid": 76337, "tid": -914061504, "ts": 1716454224062803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224118672, "dur": 7,
    "args": {
      "External id": 170183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170183,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170183, "pid": 5, "tid": 7, "ts": 1716454224118672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062822, "dur": 6,
    "args": {
      "External id": 170183,
      "cbid": 211, "correlation": 170183
    }
  },
  {
    "ph": "s", "id": 170183, "pid": 76337, "tid": -914061504, "ts": 1716454224062822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224118680, "dur": 5,
    "args": {
      "External id": 170192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170192,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170192, "pid": 5, "tid": 7, "ts": 1716454224118680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062862, "dur": 9,
    "args": {
      "External id": 170192,
      "cbid": 211, "correlation": 170192
    }
  },
  {
    "ph": "s", "id": 170192, "pid": 76337, "tid": -914061504, "ts": 1716454224062862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062925, "dur": 0,
    "args": {
      "External id": 170202,
      "cbid": 317, "correlation": 170202
    }
  },
  {
    "ph": "f", "id": 170202, "pid": 76337, "tid": -914061504, "ts": 1716454224062925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062926, "dur": 0,
    "args": {
      "External id": 170203,
      "cbid": 203, "correlation": 170203
    }
  },
  {
    "ph": "f", "id": 170203, "pid": 76337, "tid": -914061504, "ts": 1716454224062926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062926, "dur": 0,
    "args": {
      "External id": 170204,
      "cbid": 205, "correlation": 170204
    }
  },
  {
    "ph": "f", "id": 170204, "pid": 76337, "tid": -914061504, "ts": 1716454224062926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118686, "dur": 5,
    "args": {
      "External id": 170208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170208,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170208, "pid": 5, "tid": 7, "ts": 1716454224118686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062941, "dur": 13,
    "args": {
      "External id": 170208,
      "cbid": 211, "correlation": 170208
    }
  },
  {
    "ph": "s", "id": 170208, "pid": 76337, "tid": -914061504, "ts": 1716454224062941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224118693, "dur": 164,
    "args": {
      "External id": 170210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170210,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170210, "pid": 5, "tid": 7, "ts": 1716454224118693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062956, "dur": 5,
    "args": {
      "External id": 170210,
      "cbid": 211, "correlation": 170210
    }
  },
  {
    "ph": "s", "id": 170210, "pid": 76337, "tid": -914061504, "ts": 1716454224062956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224118858, "dur": 1,
    "args": {
      "External id": 170212,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170212,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 170212, "pid": 5, "tid": 7, "ts": 1716454224118858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062967, "dur": 15,
    "args": {
      "External id": 170212,
      "cbid": 51, "correlation": 170212
    }
  },
  {
    "ph": "s", "id": 170212, "pid": 76337, "tid": -914061504, "ts": 1716454224062967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224118862, "dur": 261,
    "args": {
      "External id": 170213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170213,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170213, "pid": 5, "tid": 7, "ts": 1716454224118862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062983, "dur": 7,
    "args": {
      "External id": 170213,
      "cbid": 211, "correlation": 170213
    }
  },
  {
    "ph": "s", "id": 170213, "pid": 76337, "tid": -914061504, "ts": 1716454224062983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119125, "dur": 6,
    "args": {
      "External id": 170215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170215,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170215, "pid": 5, "tid": 7, "ts": 1716454224119125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224062993, "dur": 5,
    "args": {
      "External id": 170215,
      "cbid": 211, "correlation": 170215
    }
  },
  {
    "ph": "s", "id": 170215, "pid": 76337, "tid": -914061504, "ts": 1716454224062993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224119132, "dur": 6,
    "args": {
      "External id": 170221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170221,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170221, "pid": 5, "tid": 7, "ts": 1716454224119132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063023, "dur": 8,
    "args": {
      "External id": 170221,
      "cbid": 211, "correlation": 170221
    }
  },
  {
    "ph": "s", "id": 170221, "pid": 76337, "tid": -914061504, "ts": 1716454224063023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224119139, "dur": 5,
    "args": {
      "External id": 170229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170229,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170229, "pid": 5, "tid": 7, "ts": 1716454224119139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063057, "dur": 9,
    "args": {
      "External id": 170229,
      "cbid": 211, "correlation": 170229
    }
  },
  {
    "ph": "s", "id": 170229, "pid": 76337, "tid": -914061504, "ts": 1716454224063057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224119146, "dur": 4,
    "args": {
      "External id": 170237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170237,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170237, "pid": 5, "tid": 7, "ts": 1716454224119146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063086, "dur": 8,
    "args": {
      "External id": 170237,
      "cbid": 211, "correlation": 170237
    }
  },
  {
    "ph": "s", "id": 170237, "pid": 76337, "tid": -914061504, "ts": 1716454224063086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119152, "dur": 10,
    "args": {
      "External id": 170257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170257,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170257, "pid": 5, "tid": 7, "ts": 1716454224119152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063161, "dur": 12,
    "args": {
      "External id": 170257,
      "cbid": 211, "correlation": 170257
    }
  },
  {
    "ph": "s", "id": 170257, "pid": 76337, "tid": -914061504, "ts": 1716454224063161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224119163, "dur": 4,
    "args": {
      "External id": 170269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170269,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170269, "pid": 5, "tid": 7, "ts": 1716454224119163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063182, "dur": 6,
    "args": {
      "External id": 170269,
      "cbid": 211, "correlation": 170269
    }
  },
  {
    "ph": "s", "id": 170269, "pid": 76337, "tid": -914061504, "ts": 1716454224063182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224119168, "dur": 7,
    "args": {
      "External id": 170272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170272, "pid": 5, "tid": 7, "ts": 1716454224119168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063200, "dur": 8,
    "args": {
      "External id": 170272,
      "cbid": 211, "correlation": 170272
    }
  },
  {
    "ph": "s", "id": 170272, "pid": 76337, "tid": -914061504, "ts": 1716454224063200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224119176, "dur": 5,
    "args": {
      "External id": 170281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170281,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170281, "pid": 5, "tid": 7, "ts": 1716454224119176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063239, "dur": 9,
    "args": {
      "External id": 170281,
      "cbid": 211, "correlation": 170281
    }
  },
  {
    "ph": "s", "id": 170281, "pid": 76337, "tid": -914061504, "ts": 1716454224063239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063290, "dur": 0,
    "args": {
      "External id": 170291,
      "cbid": 317, "correlation": 170291
    }
  },
  {
    "ph": "f", "id": 170291, "pid": 76337, "tid": -914061504, "ts": 1716454224063290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063291, "dur": 0,
    "args": {
      "External id": 170292,
      "cbid": 203, "correlation": 170292
    }
  },
  {
    "ph": "f", "id": 170292, "pid": 76337, "tid": -914061504, "ts": 1716454224063291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063292, "dur": 0,
    "args": {
      "External id": 170293,
      "cbid": 205, "correlation": 170293
    }
  },
  {
    "ph": "f", "id": 170293, "pid": 76337, "tid": -914061504, "ts": 1716454224063292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119182, "dur": 5,
    "args": {
      "External id": 170297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170297,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170297, "pid": 5, "tid": 7, "ts": 1716454224119182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063305, "dur": 12,
    "args": {
      "External id": 170297,
      "cbid": 211, "correlation": 170297
    }
  },
  {
    "ph": "s", "id": 170297, "pid": 76337, "tid": -914061504, "ts": 1716454224063305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119188, "dur": 163,
    "args": {
      "External id": 170299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170299,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170299, "pid": 5, "tid": 7, "ts": 1716454224119188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063319, "dur": 5,
    "args": {
      "External id": 170299,
      "cbid": 211, "correlation": 170299
    }
  },
  {
    "ph": "s", "id": 170299, "pid": 76337, "tid": -914061504, "ts": 1716454224063319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224119353, "dur": 1,
    "args": {
      "External id": 170301,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170301,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 170301, "pid": 5, "tid": 7, "ts": 1716454224119353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063330, "dur": 6,
    "args": {
      "External id": 170301,
      "cbid": 51, "correlation": 170301
    }
  },
  {
    "ph": "s", "id": 170301, "pid": 76337, "tid": -914061504, "ts": 1716454224063330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224119356, "dur": 260,
    "args": {
      "External id": 170302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170302,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170302, "pid": 5, "tid": 7, "ts": 1716454224119356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063337, "dur": 6,
    "args": {
      "External id": 170302,
      "cbid": 211, "correlation": 170302
    }
  },
  {
    "ph": "s", "id": 170302, "pid": 76337, "tid": -914061504, "ts": 1716454224063337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119618, "dur": 6,
    "args": {
      "External id": 170304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170304, "pid": 5, "tid": 7, "ts": 1716454224119618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063346, "dur": 5,
    "args": {
      "External id": 170304,
      "cbid": 211, "correlation": 170304
    }
  },
  {
    "ph": "s", "id": 170304, "pid": 76337, "tid": -914061504, "ts": 1716454224063346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224119625, "dur": 6,
    "args": {
      "External id": 170310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170310,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170310, "pid": 5, "tid": 7, "ts": 1716454224119625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063375, "dur": 9,
    "args": {
      "External id": 170310,
      "cbid": 211, "correlation": 170310
    }
  },
  {
    "ph": "s", "id": 170310, "pid": 76337, "tid": -914061504, "ts": 1716454224063375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224119633, "dur": 4,
    "args": {
      "External id": 170318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170318,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 170318, "pid": 5, "tid": 7, "ts": 1716454224119633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063419, "dur": 9,
    "args": {
      "External id": 170318,
      "cbid": 211, "correlation": 170318
    }
  },
  {
    "ph": "s", "id": 170318, "pid": 76337, "tid": -914061504, "ts": 1716454224063419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063482, "dur": 1,
    "args": {
      "External id": 170334,
      "cbid": 251, "correlation": 170334
    }
  },
  {
    "ph": "f", "id": 170334, "pid": 76337, "tid": -914061504, "ts": 1716454224063482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063487, "dur": 0,
    "args": {
      "External id": 170336,
      "cbid": 251, "correlation": 170336
    }
  },
  {
    "ph": "f", "id": 170336, "pid": 76337, "tid": -914061504, "ts": 1716454224063487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224119638, "dur": 11,
    "args": {
      "External id": 170337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170337,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170337, "pid": 5, "tid": 7, "ts": 1716454224119638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063489, "dur": 11,
    "args": {
      "External id": 170337,
      "cbid": 211, "correlation": 170337
    }
  },
  {
    "ph": "s", "id": 170337, "pid": 76337, "tid": -914061504, "ts": 1716454224063489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224119650, "dur": 4,
    "args": {
      "External id": 170339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170339,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170339, "pid": 5, "tid": 7, "ts": 1716454224119650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063501, "dur": 5,
    "args": {
      "External id": 170339,
      "cbid": 211, "correlation": 170339
    }
  },
  {
    "ph": "s", "id": 170339, "pid": 76337, "tid": -914061504, "ts": 1716454224063501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224119654, "dur": 6,
    "args": {
      "External id": 170349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170349, "pid": 5, "tid": 7, "ts": 1716454224119654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063557, "dur": 13,
    "args": {
      "External id": 170349,
      "cbid": 211, "correlation": 170349
    }
  },
  {
    "ph": "s", "id": 170349, "pid": 76337, "tid": -914061504, "ts": 1716454224063557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119662, "dur": 10,
    "args": {
      "External id": 170369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170369,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170369, "pid": 5, "tid": 7, "ts": 1716454224119662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063624, "dur": 11,
    "args": {
      "External id": 170369,
      "cbid": 211, "correlation": 170369
    }
  },
  {
    "ph": "s", "id": 170369, "pid": 76337, "tid": -914061504, "ts": 1716454224063624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224119673, "dur": 4,
    "args": {
      "External id": 170381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170381,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170381, "pid": 5, "tid": 7, "ts": 1716454224119673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063645, "dur": 6,
    "args": {
      "External id": 170381,
      "cbid": 211, "correlation": 170381
    }
  },
  {
    "ph": "s", "id": 170381, "pid": 76337, "tid": -914061504, "ts": 1716454224063645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224119678, "dur": 7,
    "args": {
      "External id": 170384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170384, "pid": 5, "tid": 7, "ts": 1716454224119678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063665, "dur": 6,
    "args": {
      "External id": 170384,
      "cbid": 211, "correlation": 170384
    }
  },
  {
    "ph": "s", "id": 170384, "pid": 76337, "tid": -914061504, "ts": 1716454224063665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224119686, "dur": 5,
    "args": {
      "External id": 170393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170393,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170393, "pid": 5, "tid": 7, "ts": 1716454224119686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063705, "dur": 10,
    "args": {
      "External id": 170393,
      "cbid": 211, "correlation": 170393
    }
  },
  {
    "ph": "s", "id": 170393, "pid": 76337, "tid": -914061504, "ts": 1716454224063705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063767, "dur": 0,
    "args": {
      "External id": 170403,
      "cbid": 317, "correlation": 170403
    }
  },
  {
    "ph": "f", "id": 170403, "pid": 76337, "tid": -914061504, "ts": 1716454224063767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063768, "dur": 0,
    "args": {
      "External id": 170404,
      "cbid": 203, "correlation": 170404
    }
  },
  {
    "ph": "f", "id": 170404, "pid": 76337, "tid": -914061504, "ts": 1716454224063768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063769, "dur": 0,
    "args": {
      "External id": 170405,
      "cbid": 205, "correlation": 170405
    }
  },
  {
    "ph": "f", "id": 170405, "pid": 76337, "tid": -914061504, "ts": 1716454224063769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119692, "dur": 5,
    "args": {
      "External id": 170409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170409,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170409, "pid": 5, "tid": 7, "ts": 1716454224119692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063782, "dur": 12,
    "args": {
      "External id": 170409,
      "cbid": 211, "correlation": 170409
    }
  },
  {
    "ph": "s", "id": 170409, "pid": 76337, "tid": -914061504, "ts": 1716454224063782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224119699, "dur": 164,
    "args": {
      "External id": 170411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170411,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170411, "pid": 5, "tid": 7, "ts": 1716454224119699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063796, "dur": 5,
    "args": {
      "External id": 170411,
      "cbid": 211, "correlation": 170411
    }
  },
  {
    "ph": "s", "id": 170411, "pid": 76337, "tid": -914061504, "ts": 1716454224063796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224119865, "dur": 1,
    "args": {
      "External id": 170413,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170413,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 170413, "pid": 5, "tid": 7, "ts": 1716454224119865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063807, "dur": 6,
    "args": {
      "External id": 170413,
      "cbid": 51, "correlation": 170413
    }
  },
  {
    "ph": "s", "id": 170413, "pid": 76337, "tid": -914061504, "ts": 1716454224063807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224119869, "dur": 261,
    "args": {
      "External id": 170414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170414,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170414, "pid": 5, "tid": 7, "ts": 1716454224119869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063814, "dur": 7,
    "args": {
      "External id": 170414,
      "cbid": 211, "correlation": 170414
    }
  },
  {
    "ph": "s", "id": 170414, "pid": 76337, "tid": -914061504, "ts": 1716454224063814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120131, "dur": 6,
    "args": {
      "External id": 170416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170416,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170416, "pid": 5, "tid": 7, "ts": 1716454224120131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063827, "dur": 5,
    "args": {
      "External id": 170416,
      "cbid": 211, "correlation": 170416
    }
  },
  {
    "ph": "s", "id": 170416, "pid": 76337, "tid": -914061504, "ts": 1716454224063827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224120139, "dur": 6,
    "args": {
      "External id": 170422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170422,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170422, "pid": 5, "tid": 7, "ts": 1716454224120139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063854, "dur": 8,
    "args": {
      "External id": 170422,
      "cbid": 211, "correlation": 170422
    }
  },
  {
    "ph": "s", "id": 170422, "pid": 76337, "tid": -914061504, "ts": 1716454224063854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224120146, "dur": 5,
    "args": {
      "External id": 170430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170430,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170430, "pid": 5, "tid": 7, "ts": 1716454224120146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063886, "dur": 8,
    "args": {
      "External id": 170430,
      "cbid": 211, "correlation": 170430
    }
  },
  {
    "ph": "s", "id": 170430, "pid": 76337, "tid": -914061504, "ts": 1716454224063886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224120153, "dur": 4,
    "args": {
      "External id": 170438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170438,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170438, "pid": 5, "tid": 7, "ts": 1716454224120153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224063915, "dur": 9,
    "args": {
      "External id": 170438,
      "cbid": 211, "correlation": 170438
    }
  },
  {
    "ph": "s", "id": 170438, "pid": 76337, "tid": -914061504, "ts": 1716454224063915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120158, "dur": 10,
    "args": {
      "External id": 170458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170458,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170458, "pid": 5, "tid": 7, "ts": 1716454224120158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064029, "dur": 13,
    "args": {
      "External id": 170458,
      "cbid": 211, "correlation": 170458
    }
  },
  {
    "ph": "s", "id": 170458, "pid": 76337, "tid": -914061504, "ts": 1716454224064029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224120170, "dur": 4,
    "args": {
      "External id": 170470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170470,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170470, "pid": 5, "tid": 7, "ts": 1716454224120170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064052, "dur": 6,
    "args": {
      "External id": 170470,
      "cbid": 211, "correlation": 170470
    }
  },
  {
    "ph": "s", "id": 170470, "pid": 76337, "tid": -914061504, "ts": 1716454224064052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224120174, "dur": 7,
    "args": {
      "External id": 170473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170473, "pid": 5, "tid": 7, "ts": 1716454224120174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064071, "dur": 6,
    "args": {
      "External id": 170473,
      "cbid": 211, "correlation": 170473
    }
  },
  {
    "ph": "s", "id": 170473, "pid": 76337, "tid": -914061504, "ts": 1716454224064071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224120182, "dur": 5,
    "args": {
      "External id": 170482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170482,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170482, "pid": 5, "tid": 7, "ts": 1716454224120182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064108, "dur": 10,
    "args": {
      "External id": 170482,
      "cbid": 211, "correlation": 170482
    }
  },
  {
    "ph": "s", "id": 170482, "pid": 76337, "tid": -914061504, "ts": 1716454224064108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064160, "dur": 0,
    "args": {
      "External id": 170492,
      "cbid": 317, "correlation": 170492
    }
  },
  {
    "ph": "f", "id": 170492, "pid": 76337, "tid": -914061504, "ts": 1716454224064160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064161, "dur": 0,
    "args": {
      "External id": 170493,
      "cbid": 203, "correlation": 170493
    }
  },
  {
    "ph": "f", "id": 170493, "pid": 76337, "tid": -914061504, "ts": 1716454224064161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064162, "dur": 0,
    "args": {
      "External id": 170494,
      "cbid": 205, "correlation": 170494
    }
  },
  {
    "ph": "f", "id": 170494, "pid": 76337, "tid": -914061504, "ts": 1716454224064162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120188, "dur": 5,
    "args": {
      "External id": 170498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170498,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170498, "pid": 5, "tid": 7, "ts": 1716454224120188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064176, "dur": 11,
    "args": {
      "External id": 170498,
      "cbid": 211, "correlation": 170498
    }
  },
  {
    "ph": "s", "id": 170498, "pid": 76337, "tid": -914061504, "ts": 1716454224064176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120195, "dur": 164,
    "args": {
      "External id": 170500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170500,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170500, "pid": 5, "tid": 7, "ts": 1716454224120195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064190, "dur": 5,
    "args": {
      "External id": 170500,
      "cbid": 211, "correlation": 170500
    }
  },
  {
    "ph": "s", "id": 170500, "pid": 76337, "tid": -914061504, "ts": 1716454224064190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224120360, "dur": 1,
    "args": {
      "External id": 170502,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170502,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 170502, "pid": 5, "tid": 7, "ts": 1716454224120360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064200, "dur": 6,
    "args": {
      "External id": 170502,
      "cbid": 51, "correlation": 170502
    }
  },
  {
    "ph": "s", "id": 170502, "pid": 76337, "tid": -914061504, "ts": 1716454224064200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224120364, "dur": 260,
    "args": {
      "External id": 170503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170503,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170503, "pid": 5, "tid": 7, "ts": 1716454224120364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064207, "dur": 6,
    "args": {
      "External id": 170503,
      "cbid": 211, "correlation": 170503
    }
  },
  {
    "ph": "s", "id": 170503, "pid": 76337, "tid": -914061504, "ts": 1716454224064207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120626, "dur": 6,
    "args": {
      "External id": 170505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170505,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170505, "pid": 5, "tid": 7, "ts": 1716454224120626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064217, "dur": 6,
    "args": {
      "External id": 170505,
      "cbid": 211, "correlation": 170505
    }
  },
  {
    "ph": "s", "id": 170505, "pid": 76337, "tid": -914061504, "ts": 1716454224064217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224120633, "dur": 6,
    "args": {
      "External id": 170511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170511, "pid": 5, "tid": 7, "ts": 1716454224120633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064246, "dur": 8,
    "args": {
      "External id": 170511,
      "cbid": 211, "correlation": 170511
    }
  },
  {
    "ph": "s", "id": 170511, "pid": 76337, "tid": -914061504, "ts": 1716454224064246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224120640, "dur": 4,
    "args": {
      "External id": 170519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170519,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 170519, "pid": 5, "tid": 7, "ts": 1716454224120640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064291, "dur": 9,
    "args": {
      "External id": 170519,
      "cbid": 211, "correlation": 170519
    }
  },
  {
    "ph": "s", "id": 170519, "pid": 76337, "tid": -914061504, "ts": 1716454224064291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064354, "dur": 1,
    "args": {
      "External id": 170535,
      "cbid": 251, "correlation": 170535
    }
  },
  {
    "ph": "f", "id": 170535, "pid": 76337, "tid": -914061504, "ts": 1716454224064354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064359, "dur": 0,
    "args": {
      "External id": 170537,
      "cbid": 251, "correlation": 170537
    }
  },
  {
    "ph": "f", "id": 170537, "pid": 76337, "tid": -914061504, "ts": 1716454224064359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224120645, "dur": 10,
    "args": {
      "External id": 170538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170538,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170538, "pid": 5, "tid": 7, "ts": 1716454224120645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064361, "dur": 12,
    "args": {
      "External id": 170538,
      "cbid": 211, "correlation": 170538
    }
  },
  {
    "ph": "s", "id": 170538, "pid": 76337, "tid": -914061504, "ts": 1716454224064361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224120657, "dur": 4,
    "args": {
      "External id": 170540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170540,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170540, "pid": 5, "tid": 7, "ts": 1716454224120657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064375, "dur": 5,
    "args": {
      "External id": 170540,
      "cbid": 211, "correlation": 170540
    }
  },
  {
    "ph": "s", "id": 170540, "pid": 76337, "tid": -914061504, "ts": 1716454224064375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224120662, "dur": 6,
    "args": {
      "External id": 170550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170550,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170550, "pid": 5, "tid": 7, "ts": 1716454224120662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064431, "dur": 12,
    "args": {
      "External id": 170550,
      "cbid": 211, "correlation": 170550
    }
  },
  {
    "ph": "s", "id": 170550, "pid": 76337, "tid": -914061504, "ts": 1716454224064431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120669, "dur": 10,
    "args": {
      "External id": 170570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170570,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170570, "pid": 5, "tid": 7, "ts": 1716454224120669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064497, "dur": 10,
    "args": {
      "External id": 170570,
      "cbid": 211, "correlation": 170570
    }
  },
  {
    "ph": "s", "id": 170570, "pid": 76337, "tid": -914061504, "ts": 1716454224064497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224120680, "dur": 4,
    "args": {
      "External id": 170582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170582,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170582, "pid": 5, "tid": 7, "ts": 1716454224120680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064517, "dur": 6,
    "args": {
      "External id": 170582,
      "cbid": 211, "correlation": 170582
    }
  },
  {
    "ph": "s", "id": 170582, "pid": 76337, "tid": -914061504, "ts": 1716454224064517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224120685, "dur": 7,
    "args": {
      "External id": 170585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170585,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170585, "pid": 5, "tid": 7, "ts": 1716454224120685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064537, "dur": 6,
    "args": {
      "External id": 170585,
      "cbid": 211, "correlation": 170585
    }
  },
  {
    "ph": "s", "id": 170585, "pid": 76337, "tid": -914061504, "ts": 1716454224064537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224120693, "dur": 5,
    "args": {
      "External id": 170594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170594,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170594, "pid": 5, "tid": 7, "ts": 1716454224120693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064577, "dur": 10,
    "args": {
      "External id": 170594,
      "cbid": 211, "correlation": 170594
    }
  },
  {
    "ph": "s", "id": 170594, "pid": 76337, "tid": -914061504, "ts": 1716454224064577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064639, "dur": 0,
    "args": {
      "External id": 170604,
      "cbid": 317, "correlation": 170604
    }
  },
  {
    "ph": "f", "id": 170604, "pid": 76337, "tid": -914061504, "ts": 1716454224064639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064640, "dur": 0,
    "args": {
      "External id": 170605,
      "cbid": 203, "correlation": 170605
    }
  },
  {
    "ph": "f", "id": 170605, "pid": 76337, "tid": -914061504, "ts": 1716454224064640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064641, "dur": 0,
    "args": {
      "External id": 170606,
      "cbid": 205, "correlation": 170606
    }
  },
  {
    "ph": "f", "id": 170606, "pid": 76337, "tid": -914061504, "ts": 1716454224064641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120699, "dur": 5,
    "args": {
      "External id": 170610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170610,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170610, "pid": 5, "tid": 7, "ts": 1716454224120699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064654, "dur": 12,
    "args": {
      "External id": 170610,
      "cbid": 211, "correlation": 170610
    }
  },
  {
    "ph": "s", "id": 170610, "pid": 76337, "tid": -914061504, "ts": 1716454224064654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224120706, "dur": 163,
    "args": {
      "External id": 170612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170612,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170612, "pid": 5, "tid": 7, "ts": 1716454224120706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064669, "dur": 6,
    "args": {
      "External id": 170612,
      "cbid": 211, "correlation": 170612
    }
  },
  {
    "ph": "s", "id": 170612, "pid": 76337, "tid": -914061504, "ts": 1716454224064669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224120870, "dur": 1,
    "args": {
      "External id": 170614,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170614,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 170614, "pid": 5, "tid": 7, "ts": 1716454224120870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064680, "dur": 6,
    "args": {
      "External id": 170614,
      "cbid": 51, "correlation": 170614
    }
  },
  {
    "ph": "s", "id": 170614, "pid": 76337, "tid": -914061504, "ts": 1716454224064680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224120874, "dur": 261,
    "args": {
      "External id": 170615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170615,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170615, "pid": 5, "tid": 7, "ts": 1716454224120874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064687, "dur": 6,
    "args": {
      "External id": 170615,
      "cbid": 211, "correlation": 170615
    }
  },
  {
    "ph": "s", "id": 170615, "pid": 76337, "tid": -914061504, "ts": 1716454224064687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121136, "dur": 6,
    "args": {
      "External id": 170617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170617,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170617, "pid": 5, "tid": 7, "ts": 1716454224121136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064697, "dur": 5,
    "args": {
      "External id": 170617,
      "cbid": 211, "correlation": 170617
    }
  },
  {
    "ph": "s", "id": 170617, "pid": 76337, "tid": -914061504, "ts": 1716454224064697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121143, "dur": 6,
    "args": {
      "External id": 170623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170623,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170623, "pid": 5, "tid": 7, "ts": 1716454224121143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064725, "dur": 9,
    "args": {
      "External id": 170623,
      "cbid": 211, "correlation": 170623
    }
  },
  {
    "ph": "s", "id": 170623, "pid": 76337, "tid": -914061504, "ts": 1716454224064725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224121151, "dur": 5,
    "args": {
      "External id": 170631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170631,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170631, "pid": 5, "tid": 7, "ts": 1716454224121151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064757, "dur": 8,
    "args": {
      "External id": 170631,
      "cbid": 211, "correlation": 170631
    }
  },
  {
    "ph": "s", "id": 170631, "pid": 76337, "tid": -914061504, "ts": 1716454224064757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224121157, "dur": 4,
    "args": {
      "External id": 170639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170639,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170639, "pid": 5, "tid": 7, "ts": 1716454224121157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064786, "dur": 8,
    "args": {
      "External id": 170639,
      "cbid": 211, "correlation": 170639
    }
  },
  {
    "ph": "s", "id": 170639, "pid": 76337, "tid": -914061504, "ts": 1716454224064786,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121163, "dur": 10,
    "args": {
      "External id": 170659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170659,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 170659, "pid": 5, "tid": 7, "ts": 1716454224121163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064933, "dur": 13,
    "args": {
      "External id": 170659,
      "cbid": 211, "correlation": 170659
    }
  },
  {
    "ph": "s", "id": 170659, "pid": 76337, "tid": -914061504, "ts": 1716454224064933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224121174, "dur": 4,
    "args": {
      "External id": 170671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170671,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 170671, "pid": 5, "tid": 7, "ts": 1716454224121174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064956, "dur": 6,
    "args": {
      "External id": 170671,
      "cbid": 211, "correlation": 170671
    }
  },
  {
    "ph": "s", "id": 170671, "pid": 76337, "tid": -914061504, "ts": 1716454224064956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121179, "dur": 7,
    "args": {
      "External id": 170674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170674, "pid": 5, "tid": 7, "ts": 1716454224121179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224064982, "dur": 8,
    "args": {
      "External id": 170674,
      "cbid": 211, "correlation": 170674
    }
  },
  {
    "ph": "s", "id": 170674, "pid": 76337, "tid": -914061504, "ts": 1716454224064982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065043, "dur": 0,
    "args": {
      "External id": 170685,
      "cbid": 317, "correlation": 170685
    }
  },
  {
    "ph": "f", "id": 170685, "pid": 76337, "tid": -914061504, "ts": 1716454224065043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065043, "dur": 0,
    "args": {
      "External id": 170686,
      "cbid": 203, "correlation": 170686
    }
  },
  {
    "ph": "f", "id": 170686, "pid": 76337, "tid": -914061504, "ts": 1716454224065043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065044, "dur": 0,
    "args": {
      "External id": 170687,
      "cbid": 205, "correlation": 170687
    }
  },
  {
    "ph": "f", "id": 170687, "pid": 76337, "tid": -914061504, "ts": 1716454224065044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121187, "dur": 5,
    "args": {
      "External id": 170691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170691,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170691, "pid": 5, "tid": 7, "ts": 1716454224121187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065060, "dur": 13,
    "args": {
      "External id": 170691,
      "cbid": 211, "correlation": 170691
    }
  },
  {
    "ph": "s", "id": 170691, "pid": 76337, "tid": -914061504, "ts": 1716454224065060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224121193, "dur": 38,
    "args": {
      "External id": 170693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170693,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 170693, "pid": 5, "tid": 7, "ts": 1716454224121193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065080, "dur": 8,
    "args": {
      "External id": 170693,
      "cbid": 211, "correlation": 170693
    }
  },
  {
    "ph": "s", "id": 170693, "pid": 76337, "tid": -914061504, "ts": 1716454224065080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121232, "dur": 5,
    "args": {
      "External id": 170695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170695,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170695, "pid": 5, "tid": 7, "ts": 1716454224121232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065092, "dur": 5,
    "args": {
      "External id": 170695,
      "cbid": 211, "correlation": 170695
    }
  },
  {
    "ph": "s", "id": 170695, "pid": 76337, "tid": -914061504, "ts": 1716454224065092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121239, "dur": 6,
    "args": {
      "External id": 170701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170701, "pid": 5, "tid": 7, "ts": 1716454224121239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065119, "dur": 9,
    "args": {
      "External id": 170701,
      "cbid": 211, "correlation": 170701
    }
  },
  {
    "ph": "s", "id": 170701, "pid": 76337, "tid": -914061504, "ts": 1716454224065119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224121246, "dur": 20,
    "args": {
      "External id": 170710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170710,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170710, "pid": 5, "tid": 7, "ts": 1716454224121246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065202, "dur": 14,
    "args": {
      "External id": 170710,
      "cbid": 211, "correlation": 170710
    }
  },
  {
    "ph": "s", "id": 170710, "pid": 76337, "tid": -914061504, "ts": 1716454224065202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121267, "dur": 10,
    "args": {
      "External id": 170732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170732,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 170732, "pid": 5, "tid": 7, "ts": 1716454224121267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065258, "dur": 10,
    "args": {
      "External id": 170732,
      "cbid": 211, "correlation": 170732
    }
  },
  {
    "ph": "s", "id": 170732, "pid": 76337, "tid": -914061504, "ts": 1716454224065258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065349, "dur": 2,
    "args": {
      "External id": 170743,
      "cbid": 251, "correlation": 170743
    }
  },
  {
    "ph": "f", "id": 170743, "pid": 76337, "tid": -914061504, "ts": 1716454224065349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065353, "dur": 0,
    "args": {
      "External id": 170744,
      "cbid": 251, "correlation": 170744
    }
  },
  {
    "ph": "f", "id": 170744, "pid": 76337, "tid": -914061504, "ts": 1716454224065353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224121279, "dur": 55,
    "args": {
      "External id": 170745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170745,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 170745, "pid": 5, "tid": 7, "ts": 1716454224121279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065356, "dur": 15,
    "args": {
      "External id": 170745,
      "cbid": 211, "correlation": 170745
    }
  },
  {
    "ph": "s", "id": 170745, "pid": 76337, "tid": -914061504, "ts": 1716454224065356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065428, "dur": 1,
    "args": {
      "External id": 170756,
      "cbid": 251, "correlation": 170756
    }
  },
  {
    "ph": "f", "id": 170756, "pid": 76337, "tid": -914061504, "ts": 1716454224065428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065431, "dur": 0,
    "args": {
      "External id": 170757,
      "cbid": 251, "correlation": 170757
    }
  },
  {
    "ph": "f", "id": 170757, "pid": 76337, "tid": -914061504, "ts": 1716454224065431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224121335, "dur": 53,
    "args": {
      "External id": 170758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170758,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 170758, "pid": 5, "tid": 7, "ts": 1716454224121335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065433, "dur": 11,
    "args": {
      "External id": 170758,
      "cbid": 211, "correlation": 170758
    }
  },
  {
    "ph": "s", "id": 170758, "pid": 76337, "tid": -914061504, "ts": 1716454224065433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065498, "dur": 1,
    "args": {
      "External id": 170769,
      "cbid": 251, "correlation": 170769
    }
  },
  {
    "ph": "f", "id": 170769, "pid": 76337, "tid": -914061504, "ts": 1716454224065498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065501, "dur": 0,
    "args": {
      "External id": 170770,
      "cbid": 251, "correlation": 170770
    }
  },
  {
    "ph": "f", "id": 170770, "pid": 76337, "tid": -914061504, "ts": 1716454224065501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224121390, "dur": 54,
    "args": {
      "External id": 170771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170771,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 170771, "pid": 5, "tid": 7, "ts": 1716454224121390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065503, "dur": 12,
    "args": {
      "External id": 170771,
      "cbid": 211, "correlation": 170771
    }
  },
  {
    "ph": "s", "id": 170771, "pid": 76337, "tid": -914061504, "ts": 1716454224065503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224121445, "dur": 55,
    "args": {
      "External id": 170796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170796,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170796, "pid": 5, "tid": 7, "ts": 1716454224121445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065590, "dur": 13,
    "args": {
      "External id": 170796,
      "cbid": 211, "correlation": 170796
    }
  },
  {
    "ph": "s", "id": 170796, "pid": 76337, "tid": -914061504, "ts": 1716454224065590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065689, "dur": 1,
    "args": {
      "External id": 170814,
      "cbid": 251, "correlation": 170814
    }
  },
  {
    "ph": "f", "id": 170814, "pid": 76337, "tid": -914061504, "ts": 1716454224065689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224121501, "dur": 64,
    "args": {
      "External id": 170816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170816,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 170816, "pid": 5, "tid": 7, "ts": 1716454224121501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065695, "dur": 13,
    "args": {
      "External id": 170816,
      "cbid": 211, "correlation": 170816
    }
  },
  {
    "ph": "s", "id": 170816, "pid": 76337, "tid": -914061504, "ts": 1716454224065695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224121566, "dur": 6,
    "args": {
      "External id": 170824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170824,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170824, "pid": 5, "tid": 7, "ts": 1716454224121566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065766, "dur": 12,
    "args": {
      "External id": 170824,
      "cbid": 211, "correlation": 170824
    }
  },
  {
    "ph": "s", "id": 170824, "pid": 76337, "tid": -914061504, "ts": 1716454224065766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224121573, "dur": 7,
    "args": {
      "External id": 170832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170832,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 170832, "pid": 5, "tid": 7, "ts": 1716454224121573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065803, "dur": 9,
    "args": {
      "External id": 170832,
      "cbid": 211, "correlation": 170832
    }
  },
  {
    "ph": "s", "id": 170832, "pid": 76337, "tid": -914061504, "ts": 1716454224065803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121582, "dur": 8,
    "args": {
      "External id": 170843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170843,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170843, "pid": 5, "tid": 7, "ts": 1716454224121582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065876, "dur": 13,
    "args": {
      "External id": 170843,
      "cbid": 211, "correlation": 170843
    }
  },
  {
    "ph": "s", "id": 170843, "pid": 76337, "tid": -914061504, "ts": 1716454224065876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121591, "dur": 9,
    "args": {
      "External id": 170865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170865,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 170865, "pid": 5, "tid": 7, "ts": 1716454224121591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224065908, "dur": 7,
    "args": {
      "External id": 170865,
      "cbid": 211, "correlation": 170865
    }
  },
  {
    "ph": "s", "id": 170865, "pid": 76337, "tid": -914061504, "ts": 1716454224065908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066002, "dur": 2,
    "args": {
      "External id": 170876,
      "cbid": 251, "correlation": 170876
    }
  },
  {
    "ph": "f", "id": 170876, "pid": 76337, "tid": -914061504, "ts": 1716454224066002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224121602, "dur": 1,
    "args": {
      "External id": 170877,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170877,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3
    }
  },
  {
    "ph": "f", "id": 170877, "pid": 5, "tid": 7, "ts": 1716454224121602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066007, "dur": 12,
    "args": {
      "External id": 170877,
      "cbid": 51, "correlation": 170877
    }
  },
  {
    "ph": "s", "id": 170877, "pid": 76337, "tid": -914061504, "ts": 1716454224066007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224121605, "dur": 36,
    "args": {
      "External id": 170878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170878,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 170878, "pid": 5, "tid": 7, "ts": 1716454224121605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066021, "dur": 13,
    "args": {
      "External id": 170878,
      "cbid": 211, "correlation": 170878
    }
  },
  {
    "ph": "s", "id": 170878, "pid": 76337, "tid": -914061504, "ts": 1716454224066021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066094, "dur": 1,
    "args": {
      "External id": 170889,
      "cbid": 251, "correlation": 170889
    }
  },
  {
    "ph": "f", "id": 170889, "pid": 76337, "tid": -914061504, "ts": 1716454224066094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066098, "dur": 0,
    "args": {
      "External id": 170890,
      "cbid": 251, "correlation": 170890
    }
  },
  {
    "ph": "f", "id": 170890, "pid": 76337, "tid": -914061504, "ts": 1716454224066098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224121643, "dur": 12,
    "args": {
      "External id": 170891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170891,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170891, "pid": 5, "tid": 7, "ts": 1716454224121643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066099, "dur": 12,
    "args": {
      "External id": 170891,
      "cbid": 211, "correlation": 170891
    }
  },
  {
    "ph": "s", "id": 170891, "pid": 76337, "tid": -914061504, "ts": 1716454224066099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224121657, "dur": 6,
    "args": {
      "External id": 170893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170893,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170893, "pid": 5, "tid": 7, "ts": 1716454224121657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066113, "dur": 6,
    "args": {
      "External id": 170893,
      "cbid": 211, "correlation": 170893
    }
  },
  {
    "ph": "s", "id": 170893, "pid": 76337, "tid": -914061504, "ts": 1716454224066113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066171, "dur": 1,
    "args": {
      "External id": 170904,
      "cbid": 251, "correlation": 170904
    }
  },
  {
    "ph": "f", "id": 170904, "pid": 76337, "tid": -914061504, "ts": 1716454224066171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066174, "dur": 0,
    "args": {
      "External id": 170905,
      "cbid": 251, "correlation": 170905
    }
  },
  {
    "ph": "f", "id": 170905, "pid": 76337, "tid": -914061504, "ts": 1716454224066174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224121663, "dur": 8,
    "args": {
      "External id": 170906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170906,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170906, "pid": 5, "tid": 7, "ts": 1716454224121663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066176, "dur": 11,
    "args": {
      "External id": 170906,
      "cbid": 211, "correlation": 170906
    }
  },
  {
    "ph": "s", "id": 170906, "pid": 76337, "tid": -914061504, "ts": 1716454224066176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224121673, "dur": 4,
    "args": {
      "External id": 170908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170908,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 170908, "pid": 5, "tid": 7, "ts": 1716454224121673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066188, "dur": 5,
    "args": {
      "External id": 170908,
      "cbid": 211, "correlation": 170908
    }
  },
  {
    "ph": "s", "id": 170908, "pid": 76337, "tid": -914061504, "ts": 1716454224066188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224121677, "dur": 20,
    "args": {
      "External id": 170933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170933,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 170933, "pid": 5, "tid": 7, "ts": 1716454224121677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066268, "dur": 12,
    "args": {
      "External id": 170933,
      "cbid": 211, "correlation": 170933
    }
  },
  {
    "ph": "s", "id": 170933, "pid": 76337, "tid": -914061504, "ts": 1716454224066268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066368, "dur": 2,
    "args": {
      "External id": 170951,
      "cbid": 251, "correlation": 170951
    }
  },
  {
    "ph": "f", "id": 170951, "pid": 76337, "tid": -914061504, "ts": 1716454224066368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224121700, "dur": 1,
    "args": {
      "External id": 170953,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 170953,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3
    }
  },
  {
    "ph": "f", "id": 170953, "pid": 5, "tid": 7, "ts": 1716454224121700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066373, "dur": 10,
    "args": {
      "External id": 170953,
      "cbid": 51, "correlation": 170953
    }
  },
  {
    "ph": "s", "id": 170953, "pid": 76337, "tid": -914061504, "ts": 1716454224066373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224121703, "dur": 37,
    "args": {
      "External id": 170954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170954,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 170954, "pid": 5, "tid": 7, "ts": 1716454224121703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066385, "dur": 13,
    "args": {
      "External id": 170954,
      "cbid": 211, "correlation": 170954
    }
  },
  {
    "ph": "s", "id": 170954, "pid": 76337, "tid": -914061504, "ts": 1716454224066385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224121742, "dur": 4,
    "args": {
      "External id": 170962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170962,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170962, "pid": 5, "tid": 7, "ts": 1716454224121742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066457, "dur": 13,
    "args": {
      "External id": 170962,
      "cbid": 211, "correlation": 170962
    }
  },
  {
    "ph": "s", "id": 170962, "pid": 76337, "tid": -914061504, "ts": 1716454224066457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121747, "dur": 8,
    "args": {
      "External id": 170970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 170970, "pid": 5, "tid": 7, "ts": 1716454224121747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066498, "dur": 9,
    "args": {
      "External id": 170970,
      "cbid": 211, "correlation": 170970
    }
  },
  {
    "ph": "s", "id": 170970, "pid": 76337, "tid": -914061504, "ts": 1716454224066498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224121757, "dur": 8,
    "args": {
      "External id": 170992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 170992,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 170992, "pid": 5, "tid": 7, "ts": 1716454224121757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066550, "dur": 10,
    "args": {
      "External id": 170992,
      "cbid": 211, "correlation": 170992
    }
  },
  {
    "ph": "s", "id": 170992, "pid": 76337, "tid": -914061504, "ts": 1716454224066550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066642, "dur": 1,
    "args": {
      "External id": 171008,
      "cbid": 251, "correlation": 171008
    }
  },
  {
    "ph": "f", "id": 171008, "pid": 76337, "tid": -914061504, "ts": 1716454224066642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066647, "dur": 0,
    "args": {
      "External id": 171010,
      "cbid": 251, "correlation": 171010
    }
  },
  {
    "ph": "f", "id": 171010, "pid": 76337, "tid": -914061504, "ts": 1716454224066647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224121767, "dur": 189,
    "args": {
      "External id": 171011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171011,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171011, "pid": 5, "tid": 7, "ts": 1716454224121767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066649, "dur": 13,
    "args": {
      "External id": 171011,
      "cbid": 211, "correlation": 171011
    }
  },
  {
    "ph": "s", "id": 171011, "pid": 76337, "tid": -914061504, "ts": 1716454224066649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121957, "dur": 21,
    "args": {
      "External id": 171019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171019, "pid": 5, "tid": 7, "ts": 1716454224121957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066714, "dur": 13,
    "args": {
      "External id": 171019,
      "cbid": 211, "correlation": 171019
    }
  },
  {
    "ph": "s", "id": 171019, "pid": 76337, "tid": -914061504, "ts": 1716454224066714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224121979, "dur": 21,
    "args": {
      "External id": 171027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171027, "pid": 5, "tid": 7, "ts": 1716454224121979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066746, "dur": 8,
    "args": {
      "External id": 171027,
      "cbid": 211, "correlation": 171027
    }
  },
  {
    "ph": "s", "id": 171027, "pid": 76337, "tid": -914061504, "ts": 1716454224066746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066828, "dur": 1,
    "args": {
      "External id": 171043,
      "cbid": 251, "correlation": 171043
    }
  },
  {
    "ph": "f", "id": 171043, "pid": 76337, "tid": -914061504, "ts": 1716454224066828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224122003, "dur": 1,
    "args": {
      "External id": 171045,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171045,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 171045, "pid": 5, "tid": 7, "ts": 1716454224122003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066833, "dur": 8,
    "args": {
      "External id": 171045,
      "cbid": 51, "correlation": 171045
    }
  },
  {
    "ph": "s", "id": 171045, "pid": 76337, "tid": -914061504, "ts": 1716454224066833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224122006, "dur": 109,
    "args": {
      "External id": 171046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171046,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 171046, "pid": 5, "tid": 7, "ts": 1716454224122006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066843, "dur": 11,
    "args": {
      "External id": 171046,
      "cbid": 211, "correlation": 171046
    }
  },
  {
    "ph": "s", "id": 171046, "pid": 76337, "tid": -914061504, "ts": 1716454224066843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224122117, "dur": 6,
    "args": {
      "External id": 171054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171054,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171054, "pid": 5, "tid": 7, "ts": 1716454224122117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066885, "dur": 10,
    "args": {
      "External id": 171054,
      "cbid": 211, "correlation": 171054
    }
  },
  {
    "ph": "s", "id": 171054, "pid": 76337, "tid": -914061504, "ts": 1716454224066885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224122124, "dur": 10,
    "args": {
      "External id": 171065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171065, "pid": 5, "tid": 7, "ts": 1716454224122124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224066953, "dur": 12,
    "args": {
      "External id": 171065,
      "cbid": 211, "correlation": 171065
    }
  },
  {
    "ph": "s", "id": 171065, "pid": 76337, "tid": -914061504, "ts": 1716454224066953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067029, "dur": 0,
    "args": {
      "External id": 171077,
      "cbid": 317, "correlation": 171077
    }
  },
  {
    "ph": "f", "id": 171077, "pid": 76337, "tid": -914061504, "ts": 1716454224067029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067030, "dur": 0,
    "args": {
      "External id": 171078,
      "cbid": 203, "correlation": 171078
    }
  },
  {
    "ph": "f", "id": 171078, "pid": 76337, "tid": -914061504, "ts": 1716454224067030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067031, "dur": 0,
    "args": {
      "External id": 171079,
      "cbid": 205, "correlation": 171079
    }
  },
  {
    "ph": "f", "id": 171079, "pid": 76337, "tid": -914061504, "ts": 1716454224067031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122136, "dur": 5,
    "args": {
      "External id": 171083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171083,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171083, "pid": 5, "tid": 7, "ts": 1716454224122136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067046, "dur": 12,
    "args": {
      "External id": 171083,
      "cbid": 211, "correlation": 171083
    }
  },
  {
    "ph": "s", "id": 171083, "pid": 76337, "tid": -914061504, "ts": 1716454224067046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224122142, "dur": 38,
    "args": {
      "External id": 171085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171085,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 171085, "pid": 5, "tid": 7, "ts": 1716454224122142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067065, "dur": 8,
    "args": {
      "External id": 171085,
      "cbid": 211, "correlation": 171085
    }
  },
  {
    "ph": "s", "id": 171085, "pid": 76337, "tid": -914061504, "ts": 1716454224067065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122182, "dur": 6,
    "args": {
      "External id": 171087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171087,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171087, "pid": 5, "tid": 7, "ts": 1716454224122182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067077, "dur": 6,
    "args": {
      "External id": 171087,
      "cbid": 211, "correlation": 171087
    }
  },
  {
    "ph": "s", "id": 171087, "pid": 76337, "tid": -914061504, "ts": 1716454224067077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224122189, "dur": 7,
    "args": {
      "External id": 171093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171093, "pid": 5, "tid": 7, "ts": 1716454224122189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067105, "dur": 9,
    "args": {
      "External id": 171093,
      "cbid": 211, "correlation": 171093
    }
  },
  {
    "ph": "s", "id": 171093, "pid": 76337, "tid": -914061504, "ts": 1716454224067105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224122197, "dur": 5,
    "args": {
      "External id": 171101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171101,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171101, "pid": 5, "tid": 7, "ts": 1716454224122197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067137, "dur": 8,
    "args": {
      "External id": 171101,
      "cbid": 211, "correlation": 171101
    }
  },
  {
    "ph": "s", "id": 171101, "pid": 76337, "tid": -914061504, "ts": 1716454224067137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122203, "dur": 11,
    "args": {
      "External id": 171121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171121,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171121, "pid": 5, "tid": 7, "ts": 1716454224122203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067210, "dur": 12,
    "args": {
      "External id": 171121,
      "cbid": 211, "correlation": 171121
    }
  },
  {
    "ph": "s", "id": 171121, "pid": 76337, "tid": -914061504, "ts": 1716454224067210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224122216, "dur": 4,
    "args": {
      "External id": 171133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171133,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 171133, "pid": 5, "tid": 7, "ts": 1716454224122216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067232, "dur": 6,
    "args": {
      "External id": 171133,
      "cbid": 211, "correlation": 171133
    }
  },
  {
    "ph": "s", "id": 171133, "pid": 76337, "tid": -914061504, "ts": 1716454224067232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224122222, "dur": 8,
    "args": {
      "External id": 171136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171136,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171136, "pid": 5, "tid": 7, "ts": 1716454224122222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067249, "dur": 7,
    "args": {
      "External id": 171136,
      "cbid": 211, "correlation": 171136
    }
  },
  {
    "ph": "s", "id": 171136, "pid": 76337, "tid": -914061504, "ts": 1716454224067249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224122232, "dur": 5,
    "args": {
      "External id": 171145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171145,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171145, "pid": 5, "tid": 7, "ts": 1716454224122232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067288, "dur": 9,
    "args": {
      "External id": 171145,
      "cbid": 211, "correlation": 171145
    }
  },
  {
    "ph": "s", "id": 171145, "pid": 76337, "tid": -914061504, "ts": 1716454224067288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067339, "dur": 0,
    "args": {
      "External id": 171155,
      "cbid": 317, "correlation": 171155
    }
  },
  {
    "ph": "f", "id": 171155, "pid": 76337, "tid": -914061504, "ts": 1716454224067339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067340, "dur": 0,
    "args": {
      "External id": 171156,
      "cbid": 203, "correlation": 171156
    }
  },
  {
    "ph": "f", "id": 171156, "pid": 76337, "tid": -914061504, "ts": 1716454224067340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067340, "dur": 0,
    "args": {
      "External id": 171157,
      "cbid": 205, "correlation": 171157
    }
  },
  {
    "ph": "f", "id": 171157, "pid": 76337, "tid": -914061504, "ts": 1716454224067340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122238, "dur": 5,
    "args": {
      "External id": 171161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171161,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171161, "pid": 5, "tid": 7, "ts": 1716454224122238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067354, "dur": 11,
    "args": {
      "External id": 171161,
      "cbid": 211, "correlation": 171161
    }
  },
  {
    "ph": "s", "id": 171161, "pid": 76337, "tid": -914061504, "ts": 1716454224067354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122245, "dur": 163,
    "args": {
      "External id": 171163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171163,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171163, "pid": 5, "tid": 7, "ts": 1716454224122245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067368, "dur": 6,
    "args": {
      "External id": 171163,
      "cbid": 211, "correlation": 171163
    }
  },
  {
    "ph": "s", "id": 171163, "pid": 76337, "tid": -914061504, "ts": 1716454224067368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224122410, "dur": 1,
    "args": {
      "External id": 171165,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171165,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 171165, "pid": 5, "tid": 7, "ts": 1716454224122410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067379, "dur": 6,
    "args": {
      "External id": 171165,
      "cbid": 51, "correlation": 171165
    }
  },
  {
    "ph": "s", "id": 171165, "pid": 76337, "tid": -914061504, "ts": 1716454224067379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224122414, "dur": 272,
    "args": {
      "External id": 171166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171166,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171166, "pid": 5, "tid": 7, "ts": 1716454224122414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067386, "dur": 6,
    "args": {
      "External id": 171166,
      "cbid": 211, "correlation": 171166
    }
  },
  {
    "ph": "s", "id": 171166, "pid": 76337, "tid": -914061504, "ts": 1716454224067386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122687, "dur": 6,
    "args": {
      "External id": 171168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171168,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171168, "pid": 5, "tid": 7, "ts": 1716454224122687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067396, "dur": 5,
    "args": {
      "External id": 171168,
      "cbid": 211, "correlation": 171168
    }
  },
  {
    "ph": "s", "id": 171168, "pid": 76337, "tid": -914061504, "ts": 1716454224067396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224122694, "dur": 7,
    "args": {
      "External id": 171174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171174, "pid": 5, "tid": 7, "ts": 1716454224122694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067424, "dur": 8,
    "args": {
      "External id": 171174,
      "cbid": 211, "correlation": 171174
    }
  },
  {
    "ph": "s", "id": 171174, "pid": 76337, "tid": -914061504, "ts": 1716454224067424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224122702, "dur": 4,
    "args": {
      "External id": 171182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171182,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 171182, "pid": 5, "tid": 7, "ts": 1716454224122702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067468, "dur": 10,
    "args": {
      "External id": 171182,
      "cbid": 211, "correlation": 171182
    }
  },
  {
    "ph": "s", "id": 171182, "pid": 76337, "tid": -914061504, "ts": 1716454224067468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067534, "dur": 1,
    "args": {
      "External id": 171198,
      "cbid": 251, "correlation": 171198
    }
  },
  {
    "ph": "f", "id": 171198, "pid": 76337, "tid": -914061504, "ts": 1716454224067534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067539, "dur": 0,
    "args": {
      "External id": 171200,
      "cbid": 251, "correlation": 171200
    }
  },
  {
    "ph": "f", "id": 171200, "pid": 76337, "tid": -914061504, "ts": 1716454224067539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224122707, "dur": 12,
    "args": {
      "External id": 171201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171201,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171201, "pid": 5, "tid": 7, "ts": 1716454224122707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067541, "dur": 11,
    "args": {
      "External id": 171201,
      "cbid": 211, "correlation": 171201
    }
  },
  {
    "ph": "s", "id": 171201, "pid": 76337, "tid": -914061504, "ts": 1716454224067541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224122721, "dur": 5,
    "args": {
      "External id": 171203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171203,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171203, "pid": 5, "tid": 7, "ts": 1716454224122721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067554, "dur": 5,
    "args": {
      "External id": 171203,
      "cbid": 211, "correlation": 171203
    }
  },
  {
    "ph": "s", "id": 171203, "pid": 76337, "tid": -914061504, "ts": 1716454224067554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224122728, "dur": 6,
    "args": {
      "External id": 171213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171213, "pid": 5, "tid": 7, "ts": 1716454224122728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067610, "dur": 13,
    "args": {
      "External id": 171213,
      "cbid": 211, "correlation": 171213
    }
  },
  {
    "ph": "s", "id": 171213, "pid": 76337, "tid": -914061504, "ts": 1716454224067610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122735, "dur": 10,
    "args": {
      "External id": 171233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171233,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171233, "pid": 5, "tid": 7, "ts": 1716454224122735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067678, "dur": 10,
    "args": {
      "External id": 171233,
      "cbid": 211, "correlation": 171233
    }
  },
  {
    "ph": "s", "id": 171233, "pid": 76337, "tid": -914061504, "ts": 1716454224067678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224122746, "dur": 4,
    "args": {
      "External id": 171245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171245,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 171245, "pid": 5, "tid": 7, "ts": 1716454224122746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067698, "dur": 6,
    "args": {
      "External id": 171245,
      "cbid": 211, "correlation": 171245
    }
  },
  {
    "ph": "s", "id": 171245, "pid": 76337, "tid": -914061504, "ts": 1716454224067698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224122751, "dur": 7,
    "args": {
      "External id": 171248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171248, "pid": 5, "tid": 7, "ts": 1716454224122751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067716, "dur": 7,
    "args": {
      "External id": 171248,
      "cbid": 211, "correlation": 171248
    }
  },
  {
    "ph": "s", "id": 171248, "pid": 76337, "tid": -914061504, "ts": 1716454224067716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224122760, "dur": 5,
    "args": {
      "External id": 171257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171257,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171257, "pid": 5, "tid": 7, "ts": 1716454224122760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067757, "dur": 9,
    "args": {
      "External id": 171257,
      "cbid": 211, "correlation": 171257
    }
  },
  {
    "ph": "s", "id": 171257, "pid": 76337, "tid": -914061504, "ts": 1716454224067757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067819, "dur": 0,
    "args": {
      "External id": 171267,
      "cbid": 317, "correlation": 171267
    }
  },
  {
    "ph": "f", "id": 171267, "pid": 76337, "tid": -914061504, "ts": 1716454224067819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067819, "dur": 0,
    "args": {
      "External id": 171268,
      "cbid": 203, "correlation": 171268
    }
  },
  {
    "ph": "f", "id": 171268, "pid": 76337, "tid": -914061504, "ts": 1716454224067819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067820, "dur": 0,
    "args": {
      "External id": 171269,
      "cbid": 205, "correlation": 171269
    }
  },
  {
    "ph": "f", "id": 171269, "pid": 76337, "tid": -914061504, "ts": 1716454224067820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122766, "dur": 5,
    "args": {
      "External id": 171273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171273,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171273, "pid": 5, "tid": 7, "ts": 1716454224122766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067835, "dur": 12,
    "args": {
      "External id": 171273,
      "cbid": 211, "correlation": 171273
    }
  },
  {
    "ph": "s", "id": 171273, "pid": 76337, "tid": -914061504, "ts": 1716454224067835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224122772, "dur": 163,
    "args": {
      "External id": 171275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171275,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171275, "pid": 5, "tid": 7, "ts": 1716454224122772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067849, "dur": 5,
    "args": {
      "External id": 171275,
      "cbid": 211, "correlation": 171275
    }
  },
  {
    "ph": "s", "id": 171275, "pid": 76337, "tid": -914061504, "ts": 1716454224067849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224122938, "dur": 1,
    "args": {
      "External id": 171277,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171277,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 171277, "pid": 5, "tid": 7, "ts": 1716454224122938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067859, "dur": 6,
    "args": {
      "External id": 171277,
      "cbid": 51, "correlation": 171277
    }
  },
  {
    "ph": "s", "id": 171277, "pid": 76337, "tid": -914061504, "ts": 1716454224067859,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224122942, "dur": 261,
    "args": {
      "External id": 171278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171278,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171278, "pid": 5, "tid": 7, "ts": 1716454224122942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067867, "dur": 6,
    "args": {
      "External id": 171278,
      "cbid": 211, "correlation": 171278
    }
  },
  {
    "ph": "s", "id": 171278, "pid": 76337, "tid": -914061504, "ts": 1716454224067867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224123204, "dur": 6,
    "args": {
      "External id": 171280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171280,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171280, "pid": 5, "tid": 7, "ts": 1716454224123204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067876, "dur": 6,
    "args": {
      "External id": 171280,
      "cbid": 211, "correlation": 171280
    }
  },
  {
    "ph": "s", "id": 171280, "pid": 76337, "tid": -914061504, "ts": 1716454224067876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224123211, "dur": 6,
    "args": {
      "External id": 171286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171286, "pid": 5, "tid": 7, "ts": 1716454224123211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067905, "dur": 8,
    "args": {
      "External id": 171286,
      "cbid": 211, "correlation": 171286
    }
  },
  {
    "ph": "s", "id": 171286, "pid": 76337, "tid": -914061504, "ts": 1716454224067905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224123219, "dur": 5,
    "args": {
      "External id": 171294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171294,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171294, "pid": 5, "tid": 7, "ts": 1716454224123219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067938, "dur": 8,
    "args": {
      "External id": 171294,
      "cbid": 211, "correlation": 171294
    }
  },
  {
    "ph": "s", "id": 171294, "pid": 76337, "tid": -914061504, "ts": 1716454224067938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224123225, "dur": 4,
    "args": {
      "External id": 171302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171302,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171302, "pid": 5, "tid": 7, "ts": 1716454224123225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224067967, "dur": 21,
    "args": {
      "External id": 171302,
      "cbid": 211, "correlation": 171302
    }
  },
  {
    "ph": "s", "id": 171302, "pid": 76337, "tid": -914061504, "ts": 1716454224067967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224123231, "dur": 12,
    "args": {
      "External id": 171311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171311,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171311, "pid": 5, "tid": 7, "ts": 1716454224123231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068069, "dur": 16,
    "args": {
      "External id": 171311,
      "cbid": 211, "correlation": 171311
    }
  },
  {
    "ph": "s", "id": 171311, "pid": 76337, "tid": -914061504, "ts": 1716454224068069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224123244, "dur": 12,
    "args": {
      "External id": 171331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171331,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171331, "pid": 5, "tid": 7, "ts": 1716454224123244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068139, "dur": 10,
    "args": {
      "External id": 171331,
      "cbid": 211, "correlation": 171331
    }
  },
  {
    "ph": "s", "id": 171331, "pid": 76337, "tid": -914061504, "ts": 1716454224068139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224123258, "dur": 4,
    "args": {
      "External id": 171343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171343,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171343, "pid": 5, "tid": 7, "ts": 1716454224123258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068160, "dur": 6,
    "args": {
      "External id": 171343,
      "cbid": 211, "correlation": 171343
    }
  },
  {
    "ph": "s", "id": 171343, "pid": 76337, "tid": -914061504, "ts": 1716454224068160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224123263, "dur": 10,
    "args": {
      "External id": 171346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171346, "pid": 5, "tid": 7, "ts": 1716454224123263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068178, "dur": 7,
    "args": {
      "External id": 171346,
      "cbid": 211, "correlation": 171346
    }
  },
  {
    "ph": "s", "id": 171346, "pid": 76337, "tid": -914061504, "ts": 1716454224068178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224123275, "dur": 6,
    "args": {
      "External id": 171355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171355,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171355, "pid": 5, "tid": 7, "ts": 1716454224123275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068217, "dur": 10,
    "args": {
      "External id": 171355,
      "cbid": 211, "correlation": 171355
    }
  },
  {
    "ph": "s", "id": 171355, "pid": 76337, "tid": -914061504, "ts": 1716454224068217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068271, "dur": 0,
    "args": {
      "External id": 171365,
      "cbid": 317, "correlation": 171365
    }
  },
  {
    "ph": "f", "id": 171365, "pid": 76337, "tid": -914061504, "ts": 1716454224068271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068271, "dur": 0,
    "args": {
      "External id": 171366,
      "cbid": 203, "correlation": 171366
    }
  },
  {
    "ph": "f", "id": 171366, "pid": 76337, "tid": -914061504, "ts": 1716454224068271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068272, "dur": 0,
    "args": {
      "External id": 171367,
      "cbid": 205, "correlation": 171367
    }
  },
  {
    "ph": "f", "id": 171367, "pid": 76337, "tid": -914061504, "ts": 1716454224068272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224123282, "dur": 7,
    "args": {
      "External id": 171371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171371, "pid": 5, "tid": 7, "ts": 1716454224123282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068288, "dur": 12,
    "args": {
      "External id": 171371,
      "cbid": 211, "correlation": 171371
    }
  },
  {
    "ph": "s", "id": 171371, "pid": 76337, "tid": -914061504, "ts": 1716454224068288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224123290, "dur": 322,
    "args": {
      "External id": 171373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171373,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171373, "pid": 5, "tid": 7, "ts": 1716454224123290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068303, "dur": 5,
    "args": {
      "External id": 171373,
      "cbid": 211, "correlation": 171373
    }
  },
  {
    "ph": "s", "id": 171373, "pid": 76337, "tid": -914061504, "ts": 1716454224068303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224123614, "dur": 1,
    "args": {
      "External id": 171375,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171375,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 171375, "pid": 5, "tid": 7, "ts": 1716454224123614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068314, "dur": 7,
    "args": {
      "External id": 171375,
      "cbid": 51, "correlation": 171375
    }
  },
  {
    "ph": "s", "id": 171375, "pid": 76337, "tid": -914061504, "ts": 1716454224068314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224123617, "dur": 498,
    "args": {
      "External id": 171376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171376,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171376, "pid": 5, "tid": 7, "ts": 1716454224123617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068322, "dur": 6,
    "args": {
      "External id": 171376,
      "cbid": 211, "correlation": 171376
    }
  },
  {
    "ph": "s", "id": 171376, "pid": 76337, "tid": -914061504, "ts": 1716454224068322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124117, "dur": 5,
    "args": {
      "External id": 171378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171378, "pid": 5, "tid": 7, "ts": 1716454224124117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068333, "dur": 6,
    "args": {
      "External id": 171378,
      "cbid": 211, "correlation": 171378
    }
  },
  {
    "ph": "s", "id": 171378, "pid": 76337, "tid": -914061504, "ts": 1716454224068333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224124124, "dur": 6,
    "args": {
      "External id": 171384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171384, "pid": 5, "tid": 7, "ts": 1716454224124124,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068363, "dur": 9,
    "args": {
      "External id": 171384,
      "cbid": 211, "correlation": 171384
    }
  },
  {
    "ph": "s", "id": 171384, "pid": 76337, "tid": -914061504, "ts": 1716454224068363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224124131, "dur": 4,
    "args": {
      "External id": 171392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171392,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 171392, "pid": 5, "tid": 7, "ts": 1716454224124131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068407, "dur": 9,
    "args": {
      "External id": 171392,
      "cbid": 211, "correlation": 171392
    }
  },
  {
    "ph": "s", "id": 171392, "pid": 76337, "tid": -914061504, "ts": 1716454224068407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068470, "dur": 1,
    "args": {
      "External id": 171408,
      "cbid": 251, "correlation": 171408
    }
  },
  {
    "ph": "f", "id": 171408, "pid": 76337, "tid": -914061504, "ts": 1716454224068470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068475, "dur": 0,
    "args": {
      "External id": 171410,
      "cbid": 251, "correlation": 171410
    }
  },
  {
    "ph": "f", "id": 171410, "pid": 76337, "tid": -914061504, "ts": 1716454224068475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224124136, "dur": 11,
    "args": {
      "External id": 171411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171411,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171411, "pid": 5, "tid": 7, "ts": 1716454224124136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068477, "dur": 12,
    "args": {
      "External id": 171411,
      "cbid": 211, "correlation": 171411
    }
  },
  {
    "ph": "s", "id": 171411, "pid": 76337, "tid": -914061504, "ts": 1716454224068477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224124149, "dur": 5,
    "args": {
      "External id": 171413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171413,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171413, "pid": 5, "tid": 7, "ts": 1716454224124149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068491, "dur": 6,
    "args": {
      "External id": 171413,
      "cbid": 211, "correlation": 171413
    }
  },
  {
    "ph": "s", "id": 171413, "pid": 76337, "tid": -914061504, "ts": 1716454224068491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224124155, "dur": 6,
    "args": {
      "External id": 171423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171423,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171423, "pid": 5, "tid": 7, "ts": 1716454224124155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068548, "dur": 12,
    "args": {
      "External id": 171423,
      "cbid": 211, "correlation": 171423
    }
  },
  {
    "ph": "s", "id": 171423, "pid": 76337, "tid": -914061504, "ts": 1716454224068548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124162, "dur": 9,
    "args": {
      "External id": 171443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171443,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171443, "pid": 5, "tid": 7, "ts": 1716454224124162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068613, "dur": 11,
    "args": {
      "External id": 171443,
      "cbid": 211, "correlation": 171443
    }
  },
  {
    "ph": "s", "id": 171443, "pid": 76337, "tid": -914061504, "ts": 1716454224068613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224124173, "dur": 4,
    "args": {
      "External id": 171455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171455,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 171455, "pid": 5, "tid": 7, "ts": 1716454224124173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068634, "dur": 6,
    "args": {
      "External id": 171455,
      "cbid": 211, "correlation": 171455
    }
  },
  {
    "ph": "s", "id": 171455, "pid": 76337, "tid": -914061504, "ts": 1716454224068634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224124178, "dur": 7,
    "args": {
      "External id": 171458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171458, "pid": 5, "tid": 7, "ts": 1716454224124178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068653, "dur": 6,
    "args": {
      "External id": 171458,
      "cbid": 211, "correlation": 171458
    }
  },
  {
    "ph": "s", "id": 171458, "pid": 76337, "tid": -914061504, "ts": 1716454224068653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224124187, "dur": 5,
    "args": {
      "External id": 171467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171467,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171467, "pid": 5, "tid": 7, "ts": 1716454224124187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068693, "dur": 10,
    "args": {
      "External id": 171467,
      "cbid": 211, "correlation": 171467
    }
  },
  {
    "ph": "s", "id": 171467, "pid": 76337, "tid": -914061504, "ts": 1716454224068693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068755, "dur": 0,
    "args": {
      "External id": 171477,
      "cbid": 317, "correlation": 171477
    }
  },
  {
    "ph": "f", "id": 171477, "pid": 76337, "tid": -914061504, "ts": 1716454224068755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068755, "dur": 0,
    "args": {
      "External id": 171478,
      "cbid": 203, "correlation": 171478
    }
  },
  {
    "ph": "f", "id": 171478, "pid": 76337, "tid": -914061504, "ts": 1716454224068755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068756, "dur": 0,
    "args": {
      "External id": 171479,
      "cbid": 205, "correlation": 171479
    }
  },
  {
    "ph": "f", "id": 171479, "pid": 76337, "tid": -914061504, "ts": 1716454224068756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124192, "dur": 5,
    "args": {
      "External id": 171483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171483,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171483, "pid": 5, "tid": 7, "ts": 1716454224124192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068770, "dur": 12,
    "args": {
      "External id": 171483,
      "cbid": 211, "correlation": 171483
    }
  },
  {
    "ph": "s", "id": 171483, "pid": 76337, "tid": -914061504, "ts": 1716454224068770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124199, "dur": 163,
    "args": {
      "External id": 171485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171485,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171485, "pid": 5, "tid": 7, "ts": 1716454224124199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068784, "dur": 5,
    "args": {
      "External id": 171485,
      "cbid": 211, "correlation": 171485
    }
  },
  {
    "ph": "s", "id": 171485, "pid": 76337, "tid": -914061504, "ts": 1716454224068784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224124364, "dur": 1,
    "args": {
      "External id": 171487,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171487,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 171487, "pid": 5, "tid": 7, "ts": 1716454224124364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068796, "dur": 6,
    "args": {
      "External id": 171487,
      "cbid": 51, "correlation": 171487
    }
  },
  {
    "ph": "s", "id": 171487, "pid": 76337, "tid": -914061504, "ts": 1716454224068796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224124368, "dur": 262,
    "args": {
      "External id": 171488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171488,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171488, "pid": 5, "tid": 7, "ts": 1716454224124368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068803, "dur": 6,
    "args": {
      "External id": 171488,
      "cbid": 211, "correlation": 171488
    }
  },
  {
    "ph": "s", "id": 171488, "pid": 76337, "tid": -914061504, "ts": 1716454224068803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124631, "dur": 6,
    "args": {
      "External id": 171490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171490, "pid": 5, "tid": 7, "ts": 1716454224124631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068813, "dur": 5,
    "args": {
      "External id": 171490,
      "cbid": 211, "correlation": 171490
    }
  },
  {
    "ph": "s", "id": 171490, "pid": 76337, "tid": -914061504, "ts": 1716454224068813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224124638, "dur": 6,
    "args": {
      "External id": 171496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171496, "pid": 5, "tid": 7, "ts": 1716454224124638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068841, "dur": 9,
    "args": {
      "External id": 171496,
      "cbid": 211, "correlation": 171496
    }
  },
  {
    "ph": "s", "id": 171496, "pid": 76337, "tid": -914061504, "ts": 1716454224068841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068901, "dur": 0,
    "args": {
      "External id": 171506,
      "cbid": 317, "correlation": 171506
    }
  },
  {
    "ph": "f", "id": 171506, "pid": 76337, "tid": -914061504, "ts": 1716454224068901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068902, "dur": 0,
    "args": {
      "External id": 171507,
      "cbid": 203, "correlation": 171507
    }
  },
  {
    "ph": "f", "id": 171507, "pid": 76337, "tid": -914061504, "ts": 1716454224068902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068902, "dur": 0,
    "args": {
      "External id": 171508,
      "cbid": 205, "correlation": 171508
    }
  },
  {
    "ph": "f", "id": 171508, "pid": 76337, "tid": -914061504, "ts": 1716454224068902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124645, "dur": 8,
    "args": {
      "External id": 171512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171512,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171512, "pid": 5, "tid": 7, "ts": 1716454224124645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068917, "dur": 11,
    "args": {
      "External id": 171512,
      "cbid": 211, "correlation": 171512
    }
  },
  {
    "ph": "s", "id": 171512, "pid": 76337, "tid": -914061504, "ts": 1716454224068917,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224124654, "dur": 3,
    "args": {
      "External id": 171514,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171514,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4563106796116505
    }
  },
  {
    "ph": "f", "id": 171514, "pid": 5, "tid": 7, "ts": 1716454224124654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068934, "dur": 15,
    "args": {
      "External id": 171514,
      "cbid": 51, "correlation": 171514
    }
  },
  {
    "ph": "s", "id": 171514, "pid": 76337, "tid": -914061504, "ts": 1716454224068934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224124659, "dur": 95,
    "args": {
      "External id": 171515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171515,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 171515, "pid": 5, "tid": 7, "ts": 1716454224124659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068950, "dur": 6,
    "args": {
      "External id": 171515,
      "cbid": 211, "correlation": 171515
    }
  },
  {
    "ph": "s", "id": 171515, "pid": 76337, "tid": -914061504, "ts": 1716454224068950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124755, "dur": 5,
    "args": {
      "External id": 171517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171517,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171517, "pid": 5, "tid": 7, "ts": 1716454224124755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068960, "dur": 5,
    "args": {
      "External id": 171517,
      "cbid": 211, "correlation": 171517
    }
  },
  {
    "ph": "s", "id": 171517, "pid": 76337, "tid": -914061504, "ts": 1716454224068960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224124762, "dur": 6,
    "args": {
      "External id": 171523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171523,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171523, "pid": 5, "tid": 7, "ts": 1716454224124762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224068996, "dur": 10,
    "args": {
      "External id": 171523,
      "cbid": 211, "correlation": 171523
    }
  },
  {
    "ph": "s", "id": 171523, "pid": 76337, "tid": -914061504, "ts": 1716454224068996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224124769, "dur": 5,
    "args": {
      "External id": 171531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171531,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171531, "pid": 5, "tid": 7, "ts": 1716454224124769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069027, "dur": 8,
    "args": {
      "External id": 171531,
      "cbid": 211, "correlation": 171531
    }
  },
  {
    "ph": "s", "id": 171531, "pid": 76337, "tid": -914061504, "ts": 1716454224069027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224124776, "dur": 5,
    "args": {
      "External id": 171539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171539,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171539, "pid": 5, "tid": 7, "ts": 1716454224124776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069056, "dur": 8,
    "args": {
      "External id": 171539,
      "cbid": 211, "correlation": 171539
    }
  },
  {
    "ph": "s", "id": 171539, "pid": 76337, "tid": -914061504, "ts": 1716454224069056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224124782, "dur": 11,
    "args": {
      "External id": 171548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171548,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171548, "pid": 5, "tid": 7, "ts": 1716454224124782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069099, "dur": 10,
    "args": {
      "External id": 171548,
      "cbid": 211, "correlation": 171548
    }
  },
  {
    "ph": "s", "id": 171548, "pid": 76337, "tid": -914061504, "ts": 1716454224069099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124794, "dur": 12,
    "args": {
      "External id": 171568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171568,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171568, "pid": 5, "tid": 7, "ts": 1716454224124794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069169, "dur": 11,
    "args": {
      "External id": 171568,
      "cbid": 211, "correlation": 171568
    }
  },
  {
    "ph": "s", "id": 171568, "pid": 76337, "tid": -914061504, "ts": 1716454224069169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224124808, "dur": 4,
    "args": {
      "External id": 171580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171580,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171580, "pid": 5, "tid": 7, "ts": 1716454224124808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069190, "dur": 7,
    "args": {
      "External id": 171580,
      "cbid": 211, "correlation": 171580
    }
  },
  {
    "ph": "s", "id": 171580, "pid": 76337, "tid": -914061504, "ts": 1716454224069190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224124813, "dur": 11,
    "args": {
      "External id": 171583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171583,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171583, "pid": 5, "tid": 7, "ts": 1716454224124813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069208, "dur": 6,
    "args": {
      "External id": 171583,
      "cbid": 211, "correlation": 171583
    }
  },
  {
    "ph": "s", "id": 171583, "pid": 76337, "tid": -914061504, "ts": 1716454224069208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224124826, "dur": 6,
    "args": {
      "External id": 171592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171592,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171592, "pid": 5, "tid": 7, "ts": 1716454224124826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069247, "dur": 10,
    "args": {
      "External id": 171592,
      "cbid": 211, "correlation": 171592
    }
  },
  {
    "ph": "s", "id": 171592, "pid": 76337, "tid": -914061504, "ts": 1716454224069247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069298, "dur": 0,
    "args": {
      "External id": 171602,
      "cbid": 317, "correlation": 171602
    }
  },
  {
    "ph": "f", "id": 171602, "pid": 76337, "tid": -914061504, "ts": 1716454224069298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069299, "dur": 0,
    "args": {
      "External id": 171603,
      "cbid": 203, "correlation": 171603
    }
  },
  {
    "ph": "f", "id": 171603, "pid": 76337, "tid": -914061504, "ts": 1716454224069299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069300, "dur": 0,
    "args": {
      "External id": 171604,
      "cbid": 205, "correlation": 171604
    }
  },
  {
    "ph": "f", "id": 171604, "pid": 76337, "tid": -914061504, "ts": 1716454224069300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124834, "dur": 7,
    "args": {
      "External id": 171608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171608,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171608, "pid": 5, "tid": 7, "ts": 1716454224124834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069313, "dur": 11,
    "args": {
      "External id": 171608,
      "cbid": 211, "correlation": 171608
    }
  },
  {
    "ph": "s", "id": 171608, "pid": 76337, "tid": -914061504, "ts": 1716454224069313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224124842, "dur": 321,
    "args": {
      "External id": 171610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171610,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171610, "pid": 5, "tid": 7, "ts": 1716454224124842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069327, "dur": 5,
    "args": {
      "External id": 171610,
      "cbid": 211, "correlation": 171610
    }
  },
  {
    "ph": "s", "id": 171610, "pid": 76337, "tid": -914061504, "ts": 1716454224069327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224125165, "dur": 1,
    "args": {
      "External id": 171612,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171612,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 171612, "pid": 5, "tid": 7, "ts": 1716454224125165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069337, "dur": 6,
    "args": {
      "External id": 171612,
      "cbid": 51, "correlation": 171612
    }
  },
  {
    "ph": "s", "id": 171612, "pid": 76337, "tid": -914061504, "ts": 1716454224069337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224125169, "dur": 501,
    "args": {
      "External id": 171613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171613,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171613, "pid": 5, "tid": 7, "ts": 1716454224125169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069345, "dur": 6,
    "args": {
      "External id": 171613,
      "cbid": 211, "correlation": 171613
    }
  },
  {
    "ph": "s", "id": 171613, "pid": 76337, "tid": -914061504, "ts": 1716454224069345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224125672, "dur": 6,
    "args": {
      "External id": 171615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171615,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171615, "pid": 5, "tid": 7, "ts": 1716454224125672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069356, "dur": 5,
    "args": {
      "External id": 171615,
      "cbid": 211, "correlation": 171615
    }
  },
  {
    "ph": "s", "id": 171615, "pid": 76337, "tid": -914061504, "ts": 1716454224069356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224125678, "dur": 6,
    "args": {
      "External id": 171621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171621, "pid": 5, "tid": 7, "ts": 1716454224125678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069384, "dur": 9,
    "args": {
      "External id": 171621,
      "cbid": 211, "correlation": 171621
    }
  },
  {
    "ph": "s", "id": 171621, "pid": 76337, "tid": -914061504, "ts": 1716454224069384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224125686, "dur": 4,
    "args": {
      "External id": 171629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171629,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 171629, "pid": 5, "tid": 7, "ts": 1716454224125686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069427, "dur": 9,
    "args": {
      "External id": 171629,
      "cbid": 211, "correlation": 171629
    }
  },
  {
    "ph": "s", "id": 171629, "pid": 76337, "tid": -914061504, "ts": 1716454224069427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069489, "dur": 1,
    "args": {
      "External id": 171645,
      "cbid": 251, "correlation": 171645
    }
  },
  {
    "ph": "f", "id": 171645, "pid": 76337, "tid": -914061504, "ts": 1716454224069489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069495, "dur": 0,
    "args": {
      "External id": 171647,
      "cbid": 251, "correlation": 171647
    }
  },
  {
    "ph": "f", "id": 171647, "pid": 76337, "tid": -914061504, "ts": 1716454224069495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224125691, "dur": 12,
    "args": {
      "External id": 171648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171648,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171648, "pid": 5, "tid": 7, "ts": 1716454224125691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069497, "dur": 11,
    "args": {
      "External id": 171648,
      "cbid": 211, "correlation": 171648
    }
  },
  {
    "ph": "s", "id": 171648, "pid": 76337, "tid": -914061504, "ts": 1716454224069497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224125704, "dur": 5,
    "args": {
      "External id": 171650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171650,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171650, "pid": 5, "tid": 7, "ts": 1716454224125704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069510, "dur": 5,
    "args": {
      "External id": 171650,
      "cbid": 211, "correlation": 171650
    }
  },
  {
    "ph": "s", "id": 171650, "pid": 76337, "tid": -914061504, "ts": 1716454224069510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224125711, "dur": 6,
    "args": {
      "External id": 171660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171660,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171660, "pid": 5, "tid": 7, "ts": 1716454224125711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069567, "dur": 13,
    "args": {
      "External id": 171660,
      "cbid": 211, "correlation": 171660
    }
  },
  {
    "ph": "s", "id": 171660, "pid": 76337, "tid": -914061504, "ts": 1716454224069567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224125718, "dur": 10,
    "args": {
      "External id": 171680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171680,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171680, "pid": 5, "tid": 7, "ts": 1716454224125718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069633, "dur": 11,
    "args": {
      "External id": 171680,
      "cbid": 211, "correlation": 171680
    }
  },
  {
    "ph": "s", "id": 171680, "pid": 76337, "tid": -914061504, "ts": 1716454224069633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224125729, "dur": 4,
    "args": {
      "External id": 171692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171692,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 171692, "pid": 5, "tid": 7, "ts": 1716454224125729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069655, "dur": 6,
    "args": {
      "External id": 171692,
      "cbid": 211, "correlation": 171692
    }
  },
  {
    "ph": "s", "id": 171692, "pid": 76337, "tid": -914061504, "ts": 1716454224069655,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224125734, "dur": 7,
    "args": {
      "External id": 171695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171695,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171695, "pid": 5, "tid": 7, "ts": 1716454224125734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069674, "dur": 6,
    "args": {
      "External id": 171695,
      "cbid": 211, "correlation": 171695
    }
  },
  {
    "ph": "s", "id": 171695, "pid": 76337, "tid": -914061504, "ts": 1716454224069674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224125742, "dur": 5,
    "args": {
      "External id": 171704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171704,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171704, "pid": 5, "tid": 7, "ts": 1716454224125742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069715, "dur": 10,
    "args": {
      "External id": 171704,
      "cbid": 211, "correlation": 171704
    }
  },
  {
    "ph": "s", "id": 171704, "pid": 76337, "tid": -914061504, "ts": 1716454224069715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069776, "dur": 0,
    "args": {
      "External id": 171714,
      "cbid": 317, "correlation": 171714
    }
  },
  {
    "ph": "f", "id": 171714, "pid": 76337, "tid": -914061504, "ts": 1716454224069776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069777, "dur": 0,
    "args": {
      "External id": 171715,
      "cbid": 203, "correlation": 171715
    }
  },
  {
    "ph": "f", "id": 171715, "pid": 76337, "tid": -914061504, "ts": 1716454224069777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069778, "dur": 0,
    "args": {
      "External id": 171716,
      "cbid": 205, "correlation": 171716
    }
  },
  {
    "ph": "f", "id": 171716, "pid": 76337, "tid": -914061504, "ts": 1716454224069778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224125749, "dur": 5,
    "args": {
      "External id": 171720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171720,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171720, "pid": 5, "tid": 7, "ts": 1716454224125749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069791, "dur": 12,
    "args": {
      "External id": 171720,
      "cbid": 211, "correlation": 171720
    }
  },
  {
    "ph": "s", "id": 171720, "pid": 76337, "tid": -914061504, "ts": 1716454224069791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224125755, "dur": 164,
    "args": {
      "External id": 171722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171722,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171722, "pid": 5, "tid": 7, "ts": 1716454224125755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069806, "dur": 6,
    "args": {
      "External id": 171722,
      "cbid": 211, "correlation": 171722
    }
  },
  {
    "ph": "s", "id": 171722, "pid": 76337, "tid": -914061504, "ts": 1716454224069806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224125921, "dur": 1,
    "args": {
      "External id": 171724,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171724,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 171724, "pid": 5, "tid": 7, "ts": 1716454224125921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069817, "dur": 6,
    "args": {
      "External id": 171724,
      "cbid": 51, "correlation": 171724
    }
  },
  {
    "ph": "s", "id": 171724, "pid": 76337, "tid": -914061504, "ts": 1716454224069817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224125925, "dur": 261,
    "args": {
      "External id": 171725,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171725,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171725, "pid": 5, "tid": 7, "ts": 1716454224125925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069824, "dur": 6,
    "args": {
      "External id": 171725,
      "cbid": 211, "correlation": 171725
    }
  },
  {
    "ph": "s", "id": 171725, "pid": 76337, "tid": -914061504, "ts": 1716454224069824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224126187, "dur": 6,
    "args": {
      "External id": 171727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171727,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171727, "pid": 5, "tid": 7, "ts": 1716454224126187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069834, "dur": 5,
    "args": {
      "External id": 171727,
      "cbid": 211, "correlation": 171727
    }
  },
  {
    "ph": "s", "id": 171727, "pid": 76337, "tid": -914061504, "ts": 1716454224069834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224126194, "dur": 6,
    "args": {
      "External id": 171733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171733,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171733, "pid": 5, "tid": 7, "ts": 1716454224126194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069862, "dur": 8,
    "args": {
      "External id": 171733,
      "cbid": 211, "correlation": 171733
    }
  },
  {
    "ph": "s", "id": 171733, "pid": 76337, "tid": -914061504, "ts": 1716454224069862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069921, "dur": 0,
    "args": {
      "External id": 171743,
      "cbid": 317, "correlation": 171743
    }
  },
  {
    "ph": "f", "id": 171743, "pid": 76337, "tid": -914061504, "ts": 1716454224069921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069922, "dur": 0,
    "args": {
      "External id": 171744,
      "cbid": 203, "correlation": 171744
    }
  },
  {
    "ph": "f", "id": 171744, "pid": 76337, "tid": -914061504, "ts": 1716454224069922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069923, "dur": 0,
    "args": {
      "External id": 171745,
      "cbid": 205, "correlation": 171745
    }
  },
  {
    "ph": "f", "id": 171745, "pid": 76337, "tid": -914061504, "ts": 1716454224069923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224126202, "dur": 8,
    "args": {
      "External id": 171749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171749,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171749, "pid": 5, "tid": 7, "ts": 1716454224126202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069934, "dur": 12,
    "args": {
      "External id": 171749,
      "cbid": 211, "correlation": 171749
    }
  },
  {
    "ph": "s", "id": 171749, "pid": 76337, "tid": -914061504, "ts": 1716454224069934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224126211, "dur": 3,
    "args": {
      "External id": 171751,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171751,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 171751, "pid": 5, "tid": 7, "ts": 1716454224126211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069951, "dur": 10,
    "args": {
      "External id": 171751,
      "cbid": 51, "correlation": 171751
    }
  },
  {
    "ph": "s", "id": 171751, "pid": 76337, "tid": -914061504, "ts": 1716454224069951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224126215, "dur": 96,
    "args": {
      "External id": 171752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171752,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 171752, "pid": 5, "tid": 7, "ts": 1716454224126215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069962, "dur": 6,
    "args": {
      "External id": 171752,
      "cbid": 211, "correlation": 171752
    }
  },
  {
    "ph": "s", "id": 171752, "pid": 76337, "tid": -914061504, "ts": 1716454224069962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224126313, "dur": 5,
    "args": {
      "External id": 171754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171754,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171754, "pid": 5, "tid": 7, "ts": 1716454224126313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224069971, "dur": 12,
    "args": {
      "External id": 171754,
      "cbid": 211, "correlation": 171754
    }
  },
  {
    "ph": "s", "id": 171754, "pid": 76337, "tid": -914061504, "ts": 1716454224069971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224126320, "dur": 6,
    "args": {
      "External id": 171760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171760,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171760, "pid": 5, "tid": 7, "ts": 1716454224126320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070006, "dur": 10,
    "args": {
      "External id": 171760,
      "cbid": 211, "correlation": 171760
    }
  },
  {
    "ph": "s", "id": 171760, "pid": 76337, "tid": -914061504, "ts": 1716454224070006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224126327, "dur": 5,
    "args": {
      "External id": 171768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171768,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171768, "pid": 5, "tid": 7, "ts": 1716454224126327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070036, "dur": 8,
    "args": {
      "External id": 171768,
      "cbid": 211, "correlation": 171768
    }
  },
  {
    "ph": "s", "id": 171768, "pid": 76337, "tid": -914061504, "ts": 1716454224070036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224126334, "dur": 4,
    "args": {
      "External id": 171776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171776,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171776, "pid": 5, "tid": 7, "ts": 1716454224126334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070066, "dur": 8,
    "args": {
      "External id": 171776,
      "cbid": 211, "correlation": 171776
    }
  },
  {
    "ph": "s", "id": 171776, "pid": 76337, "tid": -914061504, "ts": 1716454224070066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224126340, "dur": 11,
    "args": {
      "External id": 171785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171785,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171785, "pid": 5, "tid": 7, "ts": 1716454224126340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070108, "dur": 10,
    "args": {
      "External id": 171785,
      "cbid": 211, "correlation": 171785
    }
  },
  {
    "ph": "s", "id": 171785, "pid": 76337, "tid": -914061504, "ts": 1716454224070108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224126352, "dur": 12,
    "args": {
      "External id": 171805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171805,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171805, "pid": 5, "tid": 7, "ts": 1716454224126352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070177, "dur": 11,
    "args": {
      "External id": 171805,
      "cbid": 211, "correlation": 171805
    }
  },
  {
    "ph": "s", "id": 171805, "pid": 76337, "tid": -914061504, "ts": 1716454224070177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224126366, "dur": 4,
    "args": {
      "External id": 171817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171817,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171817, "pid": 5, "tid": 7, "ts": 1716454224126366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070198, "dur": 6,
    "args": {
      "External id": 171817,
      "cbid": 211, "correlation": 171817
    }
  },
  {
    "ph": "s", "id": 171817, "pid": 76337, "tid": -914061504, "ts": 1716454224070198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224126371, "dur": 11,
    "args": {
      "External id": 171820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171820, "pid": 5, "tid": 7, "ts": 1716454224126371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070217, "dur": 6,
    "args": {
      "External id": 171820,
      "cbid": 211, "correlation": 171820
    }
  },
  {
    "ph": "s", "id": 171820, "pid": 76337, "tid": -914061504, "ts": 1716454224070217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224126384, "dur": 6,
    "args": {
      "External id": 171829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171829,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171829, "pid": 5, "tid": 7, "ts": 1716454224126384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070254, "dur": 10,
    "args": {
      "External id": 171829,
      "cbid": 211, "correlation": 171829
    }
  },
  {
    "ph": "s", "id": 171829, "pid": 76337, "tid": -914061504, "ts": 1716454224070254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070305, "dur": 0,
    "args": {
      "External id": 171839,
      "cbid": 317, "correlation": 171839
    }
  },
  {
    "ph": "f", "id": 171839, "pid": 76337, "tid": -914061504, "ts": 1716454224070305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070306, "dur": 0,
    "args": {
      "External id": 171840,
      "cbid": 203, "correlation": 171840
    }
  },
  {
    "ph": "f", "id": 171840, "pid": 76337, "tid": -914061504, "ts": 1716454224070306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070307, "dur": 0,
    "args": {
      "External id": 171841,
      "cbid": 205, "correlation": 171841
    }
  },
  {
    "ph": "f", "id": 171841, "pid": 76337, "tid": -914061504, "ts": 1716454224070307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224126391, "dur": 7,
    "args": {
      "External id": 171845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171845,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171845, "pid": 5, "tid": 7, "ts": 1716454224126391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070321, "dur": 11,
    "args": {
      "External id": 171845,
      "cbid": 211, "correlation": 171845
    }
  },
  {
    "ph": "s", "id": 171845, "pid": 76337, "tid": -914061504, "ts": 1716454224070321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224126399, "dur": 322,
    "args": {
      "External id": 171847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171847,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171847, "pid": 5, "tid": 7, "ts": 1716454224126399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070334, "dur": 5,
    "args": {
      "External id": 171847,
      "cbid": 211, "correlation": 171847
    }
  },
  {
    "ph": "s", "id": 171847, "pid": 76337, "tid": -914061504, "ts": 1716454224070334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224126723, "dur": 1,
    "args": {
      "External id": 171849,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171849,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 171849, "pid": 5, "tid": 7, "ts": 1716454224126723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070345, "dur": 6,
    "args": {
      "External id": 171849,
      "cbid": 51, "correlation": 171849
    }
  },
  {
    "ph": "s", "id": 171849, "pid": 76337, "tid": -914061504, "ts": 1716454224070345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224126727, "dur": 500,
    "args": {
      "External id": 171850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171850,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171850, "pid": 5, "tid": 7, "ts": 1716454224126727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070353, "dur": 6,
    "args": {
      "External id": 171850,
      "cbid": 211, "correlation": 171850
    }
  },
  {
    "ph": "s", "id": 171850, "pid": 76337, "tid": -914061504, "ts": 1716454224070353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127228, "dur": 5,
    "args": {
      "External id": 171852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171852,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171852, "pid": 5, "tid": 7, "ts": 1716454224127228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070363, "dur": 5,
    "args": {
      "External id": 171852,
      "cbid": 211, "correlation": 171852
    }
  },
  {
    "ph": "s", "id": 171852, "pid": 76337, "tid": -914061504, "ts": 1716454224070363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224127235, "dur": 6,
    "args": {
      "External id": 171858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171858, "pid": 5, "tid": 7, "ts": 1716454224127235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070390, "dur": 9,
    "args": {
      "External id": 171858,
      "cbid": 211, "correlation": 171858
    }
  },
  {
    "ph": "s", "id": 171858, "pid": 76337, "tid": -914061504, "ts": 1716454224070390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224127242, "dur": 4,
    "args": {
      "External id": 171866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171866,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 171866, "pid": 5, "tid": 7, "ts": 1716454224127242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070434, "dur": 9,
    "args": {
      "External id": 171866,
      "cbid": 211, "correlation": 171866
    }
  },
  {
    "ph": "s", "id": 171866, "pid": 76337, "tid": -914061504, "ts": 1716454224070434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070496, "dur": 1,
    "args": {
      "External id": 171882,
      "cbid": 251, "correlation": 171882
    }
  },
  {
    "ph": "f", "id": 171882, "pid": 76337, "tid": -914061504, "ts": 1716454224070496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070501, "dur": 0,
    "args": {
      "External id": 171884,
      "cbid": 251, "correlation": 171884
    }
  },
  {
    "ph": "f", "id": 171884, "pid": 76337, "tid": -914061504, "ts": 1716454224070501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224127248, "dur": 12,
    "args": {
      "External id": 171885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171885,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171885, "pid": 5, "tid": 7, "ts": 1716454224127248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070503, "dur": 12,
    "args": {
      "External id": 171885,
      "cbid": 211, "correlation": 171885
    }
  },
  {
    "ph": "s", "id": 171885, "pid": 76337, "tid": -914061504, "ts": 1716454224070503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224127261, "dur": 5,
    "args": {
      "External id": 171887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171887,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171887, "pid": 5, "tid": 7, "ts": 1716454224127261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070517, "dur": 6,
    "args": {
      "External id": 171887,
      "cbid": 211, "correlation": 171887
    }
  },
  {
    "ph": "s", "id": 171887, "pid": 76337, "tid": -914061504, "ts": 1716454224070517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224127268, "dur": 6,
    "args": {
      "External id": 171897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171897,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171897, "pid": 5, "tid": 7, "ts": 1716454224127268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070574, "dur": 12,
    "args": {
      "External id": 171897,
      "cbid": 211, "correlation": 171897
    }
  },
  {
    "ph": "s", "id": 171897, "pid": 76337, "tid": -914061504, "ts": 1716454224070574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127275, "dur": 9,
    "args": {
      "External id": 171917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171917,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 171917, "pid": 5, "tid": 7, "ts": 1716454224127275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070639, "dur": 11,
    "args": {
      "External id": 171917,
      "cbid": 211, "correlation": 171917
    }
  },
  {
    "ph": "s", "id": 171917, "pid": 76337, "tid": -914061504, "ts": 1716454224070639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224127286, "dur": 4,
    "args": {
      "External id": 171929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171929,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 171929, "pid": 5, "tid": 7, "ts": 1716454224127286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070659, "dur": 7,
    "args": {
      "External id": 171929,
      "cbid": 211, "correlation": 171929
    }
  },
  {
    "ph": "s", "id": 171929, "pid": 76337, "tid": -914061504, "ts": 1716454224070659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224127291, "dur": 7,
    "args": {
      "External id": 171932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171932,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171932, "pid": 5, "tid": 7, "ts": 1716454224127291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070680, "dur": 6,
    "args": {
      "External id": 171932,
      "cbid": 211, "correlation": 171932
    }
  },
  {
    "ph": "s", "id": 171932, "pid": 76337, "tid": -914061504, "ts": 1716454224070680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224127299, "dur": 5,
    "args": {
      "External id": 171941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171941,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171941, "pid": 5, "tid": 7, "ts": 1716454224127299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070720, "dur": 10,
    "args": {
      "External id": 171941,
      "cbid": 211, "correlation": 171941
    }
  },
  {
    "ph": "s", "id": 171941, "pid": 76337, "tid": -914061504, "ts": 1716454224070720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070782, "dur": 0,
    "args": {
      "External id": 171951,
      "cbid": 317, "correlation": 171951
    }
  },
  {
    "ph": "f", "id": 171951, "pid": 76337, "tid": -914061504, "ts": 1716454224070782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070783, "dur": 0,
    "args": {
      "External id": 171952,
      "cbid": 203, "correlation": 171952
    }
  },
  {
    "ph": "f", "id": 171952, "pid": 76337, "tid": -914061504, "ts": 1716454224070783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070783, "dur": 0,
    "args": {
      "External id": 171953,
      "cbid": 205, "correlation": 171953
    }
  },
  {
    "ph": "f", "id": 171953, "pid": 76337, "tid": -914061504, "ts": 1716454224070783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127304, "dur": 5,
    "args": {
      "External id": 171957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171957,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171957, "pid": 5, "tid": 7, "ts": 1716454224127304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070797, "dur": 12,
    "args": {
      "External id": 171957,
      "cbid": 211, "correlation": 171957
    }
  },
  {
    "ph": "s", "id": 171957, "pid": 76337, "tid": -914061504, "ts": 1716454224070797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127311, "dur": 164,
    "args": {
      "External id": 171959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171959,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171959, "pid": 5, "tid": 7, "ts": 1716454224127311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070811, "dur": 5,
    "args": {
      "External id": 171959,
      "cbid": 211, "correlation": 171959
    }
  },
  {
    "ph": "s", "id": 171959, "pid": 76337, "tid": -914061504, "ts": 1716454224070811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224127477, "dur": 1,
    "args": {
      "External id": 171961,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171961,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 171961, "pid": 5, "tid": 7, "ts": 1716454224127477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070822, "dur": 6,
    "args": {
      "External id": 171961,
      "cbid": 51, "correlation": 171961
    }
  },
  {
    "ph": "s", "id": 171961, "pid": 76337, "tid": -914061504, "ts": 1716454224070822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224127480, "dur": 262,
    "args": {
      "External id": 171962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171962,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 171962, "pid": 5, "tid": 7, "ts": 1716454224127480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070830, "dur": 6,
    "args": {
      "External id": 171962,
      "cbid": 211, "correlation": 171962
    }
  },
  {
    "ph": "s", "id": 171962, "pid": 76337, "tid": -914061504, "ts": 1716454224070830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127744, "dur": 6,
    "args": {
      "External id": 171964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171964,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171964, "pid": 5, "tid": 7, "ts": 1716454224127744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070839, "dur": 5,
    "args": {
      "External id": 171964,
      "cbid": 211, "correlation": 171964
    }
  },
  {
    "ph": "s", "id": 171964, "pid": 76337, "tid": -914061504, "ts": 1716454224070839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224127751, "dur": 6,
    "args": {
      "External id": 171970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171970, "pid": 5, "tid": 7, "ts": 1716454224127751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070867, "dur": 9,
    "args": {
      "External id": 171970,
      "cbid": 211, "correlation": 171970
    }
  },
  {
    "ph": "s", "id": 171970, "pid": 76337, "tid": -914061504, "ts": 1716454224070867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070926, "dur": 0,
    "args": {
      "External id": 171980,
      "cbid": 317, "correlation": 171980
    }
  },
  {
    "ph": "f", "id": 171980, "pid": 76337, "tid": -914061504, "ts": 1716454224070926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070927, "dur": 0,
    "args": {
      "External id": 171981,
      "cbid": 203, "correlation": 171981
    }
  },
  {
    "ph": "f", "id": 171981, "pid": 76337, "tid": -914061504, "ts": 1716454224070927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070928, "dur": 0,
    "args": {
      "External id": 171982,
      "cbid": 205, "correlation": 171982
    }
  },
  {
    "ph": "f", "id": 171982, "pid": 76337, "tid": -914061504, "ts": 1716454224070928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127758, "dur": 8,
    "args": {
      "External id": 171986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171986, "pid": 5, "tid": 7, "ts": 1716454224127758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070939, "dur": 11,
    "args": {
      "External id": 171986,
      "cbid": 211, "correlation": 171986
    }
  },
  {
    "ph": "s", "id": 171986, "pid": 76337, "tid": -914061504, "ts": 1716454224070939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224127768, "dur": 3,
    "args": {
      "External id": 171988,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 171988,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 171988, "pid": 5, "tid": 7, "ts": 1716454224127768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070956, "dur": 9,
    "args": {
      "External id": 171988,
      "cbid": 51, "correlation": 171988
    }
  },
  {
    "ph": "s", "id": 171988, "pid": 76337, "tid": -914061504, "ts": 1716454224070956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224127772, "dur": 96,
    "args": {
      "External id": 171989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171989,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 171989, "pid": 5, "tid": 7, "ts": 1716454224127772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070966, "dur": 14,
    "args": {
      "External id": 171989,
      "cbid": 211, "correlation": 171989
    }
  },
  {
    "ph": "s", "id": 171989, "pid": 76337, "tid": -914061504, "ts": 1716454224070966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127869, "dur": 6,
    "args": {
      "External id": 171991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171991,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 171991, "pid": 5, "tid": 7, "ts": 1716454224127869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224070984, "dur": 6,
    "args": {
      "External id": 171991,
      "cbid": 211, "correlation": 171991
    }
  },
  {
    "ph": "s", "id": 171991, "pid": 76337, "tid": -914061504, "ts": 1716454224070984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224127877, "dur": 6,
    "args": {
      "External id": 171997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 171997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 171997, "pid": 5, "tid": 7, "ts": 1716454224127877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071011, "dur": 9,
    "args": {
      "External id": 171997,
      "cbid": 211, "correlation": 171997
    }
  },
  {
    "ph": "s", "id": 171997, "pid": 76337, "tid": -914061504, "ts": 1716454224071011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224127884, "dur": 5,
    "args": {
      "External id": 172005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172005,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 172005, "pid": 5, "tid": 7, "ts": 1716454224127884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071041, "dur": 8,
    "args": {
      "External id": 172005,
      "cbid": 211, "correlation": 172005
    }
  },
  {
    "ph": "s", "id": 172005, "pid": 76337, "tid": -914061504, "ts": 1716454224071041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224127891, "dur": 5,
    "args": {
      "External id": 172013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172013,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 172013, "pid": 5, "tid": 7, "ts": 1716454224127891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071070, "dur": 9,
    "args": {
      "External id": 172013,
      "cbid": 211, "correlation": 172013
    }
  },
  {
    "ph": "s", "id": 172013, "pid": 76337, "tid": -914061504, "ts": 1716454224071070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224127897, "dur": 14,
    "args": {
      "External id": 172024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172024,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172024, "pid": 5, "tid": 7, "ts": 1716454224127897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071150, "dur": 14,
    "args": {
      "External id": 172024,
      "cbid": 211, "correlation": 172024
    }
  },
  {
    "ph": "s", "id": 172024, "pid": 76337, "tid": -914061504, "ts": 1716454224071150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071207, "dur": 0,
    "args": {
      "External id": 172034,
      "cbid": 317, "correlation": 172034
    }
  },
  {
    "ph": "f", "id": 172034, "pid": 76337, "tid": -914061504, "ts": 1716454224071207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071208, "dur": 0,
    "args": {
      "External id": 172035,
      "cbid": 203, "correlation": 172035
    }
  },
  {
    "ph": "f", "id": 172035, "pid": 76337, "tid": -914061504, "ts": 1716454224071208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071208, "dur": 0,
    "args": {
      "External id": 172036,
      "cbid": 205, "correlation": 172036
    }
  },
  {
    "ph": "f", "id": 172036, "pid": 76337, "tid": -914061504, "ts": 1716454224071208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127912, "dur": 9,
    "args": {
      "External id": 172040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172040,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172040, "pid": 5, "tid": 7, "ts": 1716454224127912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071223, "dur": 11,
    "args": {
      "External id": 172040,
      "cbid": 211, "correlation": 172040
    }
  },
  {
    "ph": "s", "id": 172040, "pid": 76337, "tid": -914061504, "ts": 1716454224071223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224127922, "dur": 164,
    "args": {
      "External id": 172042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172042,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172042, "pid": 5, "tid": 7, "ts": 1716454224127922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071236, "dur": 5,
    "args": {
      "External id": 172042,
      "cbid": 211, "correlation": 172042
    }
  },
  {
    "ph": "s", "id": 172042, "pid": 76337, "tid": -914061504, "ts": 1716454224071236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224128089, "dur": 1,
    "args": {
      "External id": 172044,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 172044,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 172044, "pid": 5, "tid": 7, "ts": 1716454224128089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071247, "dur": 6,
    "args": {
      "External id": 172044,
      "cbid": 51, "correlation": 172044
    }
  },
  {
    "ph": "s", "id": 172044, "pid": 76337, "tid": -914061504, "ts": 1716454224071247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224128092, "dur": 654,
    "args": {
      "External id": 172045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172045,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172045, "pid": 5, "tid": 7, "ts": 1716454224128092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071254, "dur": 6,
    "args": {
      "External id": 172045,
      "cbid": 211, "correlation": 172045
    }
  },
  {
    "ph": "s", "id": 172045, "pid": 76337, "tid": -914061504, "ts": 1716454224071254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224128747, "dur": 12,
    "args": {
      "External id": 172047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172047,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172047, "pid": 5, "tid": 7, "ts": 1716454224128747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071265, "dur": 6,
    "args": {
      "External id": 172047,
      "cbid": 211, "correlation": 172047
    }
  },
  {
    "ph": "s", "id": 172047, "pid": 76337, "tid": -914061504, "ts": 1716454224071265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224128760, "dur": 15,
    "args": {
      "External id": 172053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172053,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172053, "pid": 5, "tid": 7, "ts": 1716454224128760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071295, "dur": 9,
    "args": {
      "External id": 172053,
      "cbid": 211, "correlation": 172053
    }
  },
  {
    "ph": "s", "id": 172053, "pid": 76337, "tid": -914061504, "ts": 1716454224071295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224128776, "dur": 30,
    "args": {
      "External id": 172062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172062,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172062, "pid": 5, "tid": 7, "ts": 1716454224128776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071388, "dur": 13,
    "args": {
      "External id": 172062,
      "cbid": 211, "correlation": 172062
    }
  },
  {
    "ph": "s", "id": 172062, "pid": 76337, "tid": -914061504, "ts": 1716454224071388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224128808, "dur": 30,
    "args": {
      "External id": 172082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172082,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 172082, "pid": 5, "tid": 7, "ts": 1716454224128808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071458, "dur": 11,
    "args": {
      "External id": 172082,
      "cbid": 211, "correlation": 172082
    }
  },
  {
    "ph": "s", "id": 172082, "pid": 76337, "tid": -914061504, "ts": 1716454224071458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224128839, "dur": 4,
    "args": {
      "External id": 172094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172094,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172094, "pid": 5, "tid": 7, "ts": 1716454224128839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071479, "dur": 6,
    "args": {
      "External id": 172094,
      "cbid": 211, "correlation": 172094
    }
  },
  {
    "ph": "s", "id": 172094, "pid": 76337, "tid": -914061504, "ts": 1716454224071479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224128845, "dur": 31,
    "args": {
      "External id": 172097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172097, "pid": 5, "tid": 7, "ts": 1716454224128845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071497, "dur": 6,
    "args": {
      "External id": 172097,
      "cbid": 211, "correlation": 172097
    }
  },
  {
    "ph": "s", "id": 172097, "pid": 76337, "tid": -914061504, "ts": 1716454224071497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224128877, "dur": 22,
    "args": {
      "External id": 172106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172106,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172106, "pid": 5, "tid": 7, "ts": 1716454224128877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071536, "dur": 9,
    "args": {
      "External id": 172106,
      "cbid": 211, "correlation": 172106
    }
  },
  {
    "ph": "s", "id": 172106, "pid": 76337, "tid": -914061504, "ts": 1716454224071536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071588, "dur": 0,
    "args": {
      "External id": 172116,
      "cbid": 317, "correlation": 172116
    }
  },
  {
    "ph": "f", "id": 172116, "pid": 76337, "tid": -914061504, "ts": 1716454224071588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071589, "dur": 0,
    "args": {
      "External id": 172117,
      "cbid": 203, "correlation": 172117
    }
  },
  {
    "ph": "f", "id": 172117, "pid": 76337, "tid": -914061504, "ts": 1716454224071589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071590, "dur": 0,
    "args": {
      "External id": 172118,
      "cbid": 205, "correlation": 172118
    }
  },
  {
    "ph": "f", "id": 172118, "pid": 76337, "tid": -914061504, "ts": 1716454224071590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224128900, "dur": 22,
    "args": {
      "External id": 172122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172122, "pid": 5, "tid": 7, "ts": 1716454224128900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071605, "dur": 11,
    "args": {
      "External id": 172122,
      "cbid": 211, "correlation": 172122
    }
  },
  {
    "ph": "s", "id": 172122, "pid": 76337, "tid": -914061504, "ts": 1716454224071605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224128923, "dur": 324,
    "args": {
      "External id": 172124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172124, "pid": 5, "tid": 7, "ts": 1716454224128923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071619, "dur": 6,
    "args": {
      "External id": 172124,
      "cbid": 211, "correlation": 172124
    }
  },
  {
    "ph": "s", "id": 172124, "pid": 76337, "tid": -914061504, "ts": 1716454224071619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224129249, "dur": 1,
    "args": {
      "External id": 172126,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 172126,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5878750765462339
    }
  },
  {
    "ph": "f", "id": 172126, "pid": 5, "tid": 7, "ts": 1716454224129249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071631, "dur": 6,
    "args": {
      "External id": 172126,
      "cbid": 51, "correlation": 172126
    }
  },
  {
    "ph": "s", "id": 172126, "pid": 76337, "tid": -914061504, "ts": 1716454224071631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224129253, "dur": 1246,
    "args": {
      "External id": 172127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172127,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172127, "pid": 5, "tid": 7, "ts": 1716454224129253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071638, "dur": 6,
    "args": {
      "External id": 172127,
      "cbid": 211, "correlation": 172127
    }
  },
  {
    "ph": "s", "id": 172127, "pid": 76337, "tid": -914061504, "ts": 1716454224071638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224130500, "dur": 12,
    "args": {
      "External id": 172129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172129, "pid": 5, "tid": 7, "ts": 1716454224130500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071648, "dur": 5,
    "args": {
      "External id": 172129,
      "cbid": 211, "correlation": 172129
    }
  },
  {
    "ph": "s", "id": 172129, "pid": 76337, "tid": -914061504, "ts": 1716454224071648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224130514, "dur": 15,
    "args": {
      "External id": 172135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172135, "pid": 5, "tid": 7, "ts": 1716454224130514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071676, "dur": 8,
    "args": {
      "External id": 172135,
      "cbid": 211, "correlation": 172135
    }
  },
  {
    "ph": "s", "id": 172135, "pid": 76337, "tid": -914061504, "ts": 1716454224071676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224130530, "dur": 4,
    "args": {
      "External id": 172143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172143,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 172143, "pid": 5, "tid": 7, "ts": 1716454224130530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071719, "dur": 10,
    "args": {
      "External id": 172143,
      "cbid": 211, "correlation": 172143
    }
  },
  {
    "ph": "s", "id": 172143, "pid": 76337, "tid": -914061504, "ts": 1716454224071719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071784, "dur": 1,
    "args": {
      "External id": 172159,
      "cbid": 251, "correlation": 172159
    }
  },
  {
    "ph": "f", "id": 172159, "pid": 76337, "tid": -914061504, "ts": 1716454224071784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071789, "dur": 0,
    "args": {
      "External id": 172161,
      "cbid": 251, "correlation": 172161
    }
  },
  {
    "ph": "f", "id": 172161, "pid": 76337, "tid": -914061504, "ts": 1716454224071789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224130535, "dur": 13,
    "args": {
      "External id": 172162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172162,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172162, "pid": 5, "tid": 7, "ts": 1716454224130535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071791, "dur": 12,
    "args": {
      "External id": 172162,
      "cbid": 211, "correlation": 172162
    }
  },
  {
    "ph": "s", "id": 172162, "pid": 76337, "tid": -914061504, "ts": 1716454224071791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224130549, "dur": 5,
    "args": {
      "External id": 172164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172164,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172164, "pid": 5, "tid": 7, "ts": 1716454224130549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071804, "dur": 5,
    "args": {
      "External id": 172164,
      "cbid": 211, "correlation": 172164
    }
  },
  {
    "ph": "s", "id": 172164, "pid": 76337, "tid": -914061504, "ts": 1716454224071804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224130555, "dur": 17,
    "args": {
      "External id": 172174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172174, "pid": 5, "tid": 7, "ts": 1716454224130555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071862, "dur": 12,
    "args": {
      "External id": 172174,
      "cbid": 211, "correlation": 172174
    }
  },
  {
    "ph": "s", "id": 172174, "pid": 76337, "tid": -914061504, "ts": 1716454224071862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224130574, "dur": 17,
    "args": {
      "External id": 172194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172194,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 172194, "pid": 5, "tid": 7, "ts": 1716454224130574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071928, "dur": 11,
    "args": {
      "External id": 172194,
      "cbid": 211, "correlation": 172194
    }
  },
  {
    "ph": "s", "id": 172194, "pid": 76337, "tid": -914061504, "ts": 1716454224071928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224130592, "dur": 5,
    "args": {
      "External id": 172206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172206,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 172206, "pid": 5, "tid": 7, "ts": 1716454224130592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071949, "dur": 6,
    "args": {
      "External id": 172206,
      "cbid": 211, "correlation": 172206
    }
  },
  {
    "ph": "s", "id": 172206, "pid": 76337, "tid": -914061504, "ts": 1716454224071949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224130598, "dur": 17,
    "args": {
      "External id": 172209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172209, "pid": 5, "tid": 7, "ts": 1716454224130598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224071967, "dur": 15,
    "args": {
      "External id": 172209,
      "cbid": 211, "correlation": 172209
    }
  },
  {
    "ph": "s", "id": 172209, "pid": 76337, "tid": -914061504, "ts": 1716454224071967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224130616, "dur": 11,
    "args": {
      "External id": 172218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172218,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172218, "pid": 5, "tid": 7, "ts": 1716454224130616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072016, "dur": 11,
    "args": {
      "External id": 172218,
      "cbid": 211, "correlation": 172218
    }
  },
  {
    "ph": "s", "id": 172218, "pid": 76337, "tid": -914061504, "ts": 1716454224072016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072080, "dur": 0,
    "args": {
      "External id": 172228,
      "cbid": 317, "correlation": 172228
    }
  },
  {
    "ph": "f", "id": 172228, "pid": 76337, "tid": -914061504, "ts": 1716454224072080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072081, "dur": 0,
    "args": {
      "External id": 172229,
      "cbid": 203, "correlation": 172229
    }
  },
  {
    "ph": "f", "id": 172229, "pid": 76337, "tid": -914061504, "ts": 1716454224072081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072082, "dur": 0,
    "args": {
      "External id": 172230,
      "cbid": 205, "correlation": 172230
    }
  },
  {
    "ph": "f", "id": 172230, "pid": 76337, "tid": -914061504, "ts": 1716454224072082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224130629, "dur": 11,
    "args": {
      "External id": 172234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172234,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172234, "pid": 5, "tid": 7, "ts": 1716454224130629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072095, "dur": 12,
    "args": {
      "External id": 172234,
      "cbid": 211, "correlation": 172234
    }
  },
  {
    "ph": "s", "id": 172234, "pid": 76337, "tid": -914061504, "ts": 1716454224072095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224130641, "dur": 163,
    "args": {
      "External id": 172236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172236,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172236, "pid": 5, "tid": 7, "ts": 1716454224130641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072109, "dur": 5,
    "args": {
      "External id": 172236,
      "cbid": 211, "correlation": 172236
    }
  },
  {
    "ph": "s", "id": 172236, "pid": 76337, "tid": -914061504, "ts": 1716454224072109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224130807, "dur": 1,
    "args": {
      "External id": 172238,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 172238,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 172238, "pid": 5, "tid": 7, "ts": 1716454224130807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072120, "dur": 7,
    "args": {
      "External id": 172238,
      "cbid": 51, "correlation": 172238
    }
  },
  {
    "ph": "s", "id": 172238, "pid": 76337, "tid": -914061504, "ts": 1716454224072120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224130810, "dur": 652,
    "args": {
      "External id": 172239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172239,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172239, "pid": 5, "tid": 7, "ts": 1716454224130810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072128, "dur": 7,
    "args": {
      "External id": 172239,
      "cbid": 211, "correlation": 172239
    }
  },
  {
    "ph": "s", "id": 172239, "pid": 76337, "tid": -914061504, "ts": 1716454224072128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224131463, "dur": 13,
    "args": {
      "External id": 172241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172241,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172241, "pid": 5, "tid": 7, "ts": 1716454224131463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072138, "dur": 5,
    "args": {
      "External id": 172241,
      "cbid": 211, "correlation": 172241
    }
  },
  {
    "ph": "s", "id": 172241, "pid": 76337, "tid": -914061504, "ts": 1716454224072138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224131478, "dur": 15,
    "args": {
      "External id": 172247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172247, "pid": 5, "tid": 7, "ts": 1716454224131478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072166, "dur": 9,
    "args": {
      "External id": 172247,
      "cbid": 211, "correlation": 172247
    }
  },
  {
    "ph": "s", "id": 172247, "pid": 76337, "tid": -914061504, "ts": 1716454224072166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072224, "dur": 0,
    "args": {
      "External id": 172257,
      "cbid": 317, "correlation": 172257
    }
  },
  {
    "ph": "f", "id": 172257, "pid": 76337, "tid": -914061504, "ts": 1716454224072224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072224, "dur": 0,
    "args": {
      "External id": 172258,
      "cbid": 203, "correlation": 172258
    }
  },
  {
    "ph": "f", "id": 172258, "pid": 76337, "tid": -914061504, "ts": 1716454224072224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072225, "dur": 0,
    "args": {
      "External id": 172259,
      "cbid": 205, "correlation": 172259
    }
  },
  {
    "ph": "f", "id": 172259, "pid": 76337, "tid": -914061504, "ts": 1716454224072225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224131493, "dur": 21,
    "args": {
      "External id": 172263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172263,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172263, "pid": 5, "tid": 7, "ts": 1716454224131493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072239, "dur": 11,
    "args": {
      "External id": 172263,
      "cbid": 211, "correlation": 172263
    }
  },
  {
    "ph": "s", "id": 172263, "pid": 76337, "tid": -914061504, "ts": 1716454224072239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224131516, "dur": 4,
    "args": {
      "External id": 172265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172265, "pid": 5, "tid": 7, "ts": 1716454224131516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072256, "dur": 6,
    "args": {
      "External id": 172265,
      "cbid": 211, "correlation": 172265
    }
  },
  {
    "ph": "s", "id": 172265, "pid": 76337, "tid": -914061504, "ts": 1716454224072256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072265, "dur": 0,
    "args": {
      "External id": 172266,
      "cbid": 51, "correlation": 172266
    }
  },
  {
    "ph": "s", "id": 172266, "pid": 76337, "tid": -914061504, "ts": 1716454224072265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224131521, "dur": 176,
    "args": {
      "External id": 172267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172267,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 172267, "pid": 5, "tid": 7, "ts": 1716454224131521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072266, "dur": 5,
    "args": {
      "External id": 172267,
      "cbid": 211, "correlation": 172267
    }
  },
  {
    "ph": "s", "id": 172267, "pid": 76337, "tid": -914061504, "ts": 1716454224072266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224131699, "dur": 16,
    "args": {
      "External id": 172272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172272, "pid": 5, "tid": 7, "ts": 1716454224131699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072293, "dur": 8,
    "args": {
      "External id": 172272,
      "cbid": 211, "correlation": 172272
    }
  },
  {
    "ph": "s", "id": 172272, "pid": 76337, "tid": -914061504, "ts": 1716454224072293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224131716, "dur": 12,
    "args": {
      "External id": 172280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172280,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172280, "pid": 5, "tid": 7, "ts": 1716454224131716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072321, "dur": 8,
    "args": {
      "External id": 172280,
      "cbid": 211, "correlation": 172280
    }
  },
  {
    "ph": "s", "id": 172280, "pid": 76337, "tid": -914061504, "ts": 1716454224072321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224131729, "dur": 11,
    "args": {
      "External id": 172288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172288,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172288, "pid": 5, "tid": 7, "ts": 1716454224131729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072350, "dur": 8,
    "args": {
      "External id": 172288,
      "cbid": 211, "correlation": 172288
    }
  },
  {
    "ph": "s", "id": 172288, "pid": 76337, "tid": -914061504, "ts": 1716454224072350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224131741, "dur": 19,
    "args": {
      "External id": 172308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172308,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 172308, "pid": 5, "tid": 7, "ts": 1716454224131741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072432, "dur": 12,
    "args": {
      "External id": 172308,
      "cbid": 211, "correlation": 172308
    }
  },
  {
    "ph": "s", "id": 172308, "pid": 76337, "tid": -914061504, "ts": 1716454224072432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224131761, "dur": 4,
    "args": {
      "External id": 172320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172320,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 172320, "pid": 5, "tid": 7, "ts": 1716454224131761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072454, "dur": 6,
    "args": {
      "External id": 172320,
      "cbid": 211, "correlation": 172320
    }
  },
  {
    "ph": "s", "id": 172320, "pid": 76337, "tid": -914061504, "ts": 1716454224072454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224131767, "dur": 17,
    "args": {
      "External id": 172323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172323,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172323, "pid": 5, "tid": 7, "ts": 1716454224131767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072471, "dur": 7,
    "args": {
      "External id": 172323,
      "cbid": 211, "correlation": 172323
    }
  },
  {
    "ph": "s", "id": 172323, "pid": 76337, "tid": -914061504, "ts": 1716454224072471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072528, "dur": 0,
    "args": {
      "External id": 172334,
      "cbid": 317, "correlation": 172334
    }
  },
  {
    "ph": "f", "id": 172334, "pid": 76337, "tid": -914061504, "ts": 1716454224072528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072529, "dur": 0,
    "args": {
      "External id": 172335,
      "cbid": 203, "correlation": 172335
    }
  },
  {
    "ph": "f", "id": 172335, "pid": 76337, "tid": -914061504, "ts": 1716454224072529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072530, "dur": 0,
    "args": {
      "External id": 172336,
      "cbid": 205, "correlation": 172336
    }
  },
  {
    "ph": "f", "id": 172336, "pid": 76337, "tid": -914061504, "ts": 1716454224072530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224131785, "dur": 12,
    "args": {
      "External id": 172340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172340,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172340, "pid": 5, "tid": 7, "ts": 1716454224131785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072543, "dur": 11,
    "args": {
      "External id": 172340,
      "cbid": 211, "correlation": 172340
    }
  },
  {
    "ph": "s", "id": 172340, "pid": 76337, "tid": -914061504, "ts": 1716454224072543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224131798, "dur": 3,
    "args": {
      "External id": 172342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172342,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172342, "pid": 5, "tid": 7, "ts": 1716454224131798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072559, "dur": 6,
    "args": {
      "External id": 172342,
      "cbid": 211, "correlation": 172342
    }
  },
  {
    "ph": "s", "id": 172342, "pid": 76337, "tid": -914061504, "ts": 1716454224072559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072567, "dur": 0,
    "args": {
      "External id": 172343,
      "cbid": 51, "correlation": 172343
    }
  },
  {
    "ph": "s", "id": 172343, "pid": 76337, "tid": -914061504, "ts": 1716454224072567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224131803, "dur": 90,
    "args": {
      "External id": 172344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172344,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 172344, "pid": 5, "tid": 7, "ts": 1716454224131803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072568, "dur": 5,
    "args": {
      "External id": 172344,
      "cbid": 211, "correlation": 172344
    }
  },
  {
    "ph": "s", "id": 172344, "pid": 76337, "tid": -914061504, "ts": 1716454224072568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224131895, "dur": 16,
    "args": {
      "External id": 172349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172349, "pid": 5, "tid": 7, "ts": 1716454224131895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072595, "dur": 9,
    "args": {
      "External id": 172349,
      "cbid": 211, "correlation": 172349
    }
  },
  {
    "ph": "s", "id": 172349, "pid": 76337, "tid": -914061504, "ts": 1716454224072595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224131912, "dur": 85,
    "args": {
      "External id": 172358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172358,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172358, "pid": 5, "tid": 7, "ts": 1716454224131912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072678, "dur": 15,
    "args": {
      "External id": 172358,
      "cbid": 211, "correlation": 172358
    }
  },
  {
    "ph": "s", "id": 172358, "pid": 76337, "tid": -914061504, "ts": 1716454224072678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224131999, "dur": 30,
    "args": {
      "External id": 172380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172380,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172380, "pid": 5, "tid": 7, "ts": 1716454224131999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072737, "dur": 10,
    "args": {
      "External id": 172380,
      "cbid": 211, "correlation": 172380
    }
  },
  {
    "ph": "s", "id": 172380, "pid": 76337, "tid": -914061504, "ts": 1716454224072737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072830, "dur": 1,
    "args": {
      "External id": 172391,
      "cbid": 251, "correlation": 172391
    }
  },
  {
    "ph": "f", "id": 172391, "pid": 76337, "tid": -914061504, "ts": 1716454224072830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224132031, "dur": 167,
    "args": {
      "External id": 172392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172392,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172392, "pid": 5, "tid": 7, "ts": 1716454224132031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072836, "dur": 13,
    "args": {
      "External id": 172392,
      "cbid": 211, "correlation": 172392
    }
  },
  {
    "ph": "s", "id": 172392, "pid": 76337, "tid": -914061504, "ts": 1716454224072836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072907, "dur": 1,
    "args": {
      "External id": 172403,
      "cbid": 251, "correlation": 172403
    }
  },
  {
    "ph": "f", "id": 172403, "pid": 76337, "tid": -914061504, "ts": 1716454224072907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224132199, "dur": 160,
    "args": {
      "External id": 172404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172404,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172404, "pid": 5, "tid": 7, "ts": 1716454224132199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072911, "dur": 11,
    "args": {
      "External id": 172404,
      "cbid": 211, "correlation": 172404
    }
  },
  {
    "ph": "s", "id": 172404, "pid": 76337, "tid": -914061504, "ts": 1716454224072911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072984, "dur": 1,
    "args": {
      "External id": 172415,
      "cbid": 251, "correlation": 172415
    }
  },
  {
    "ph": "f", "id": 172415, "pid": 76337, "tid": -914061504, "ts": 1716454224072984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224132360, "dur": 159,
    "args": {
      "External id": 172416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172416,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172416, "pid": 5, "tid": 7, "ts": 1716454224132360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224072989, "dur": 12,
    "args": {
      "External id": 172416,
      "cbid": 211, "correlation": 172416
    }
  },
  {
    "ph": "s", "id": 172416, "pid": 76337, "tid": -914061504, "ts": 1716454224072989,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224132520, "dur": 338,
    "args": {
      "External id": 172441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172441,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172441, "pid": 5, "tid": 7, "ts": 1716454224132520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073077, "dur": 14,
    "args": {
      "External id": 172441,
      "cbid": 211, "correlation": 172441
    }
  },
  {
    "ph": "s", "id": 172441, "pid": 76337, "tid": -914061504, "ts": 1716454224073077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073179, "dur": 1,
    "args": {
      "External id": 172459,
      "cbid": 251, "correlation": 172459
    }
  },
  {
    "ph": "f", "id": 172459, "pid": 76337, "tid": -914061504, "ts": 1716454224073179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224132860, "dur": 167,
    "args": {
      "External id": 172461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172461,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172461, "pid": 5, "tid": 7, "ts": 1716454224132860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073185, "dur": 14,
    "args": {
      "External id": 172461,
      "cbid": 211, "correlation": 172461
    }
  },
  {
    "ph": "s", "id": 172461, "pid": 76337, "tid": -914061504, "ts": 1716454224073185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224133028, "dur": 20,
    "args": {
      "External id": 172469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172469,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172469, "pid": 5, "tid": 7, "ts": 1716454224133028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073256, "dur": 12,
    "args": {
      "External id": 172469,
      "cbid": 211, "correlation": 172469
    }
  },
  {
    "ph": "s", "id": 172469, "pid": 76337, "tid": -914061504, "ts": 1716454224073256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224133049, "dur": 28,
    "args": {
      "External id": 172477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172477,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172477, "pid": 5, "tid": 7, "ts": 1716454224133049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073295, "dur": 8,
    "args": {
      "External id": 172477,
      "cbid": 211, "correlation": 172477
    }
  },
  {
    "ph": "s", "id": 172477, "pid": 76337, "tid": -914061504, "ts": 1716454224073295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224133078, "dur": 18,
    "args": {
      "External id": 172488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172488, "pid": 5, "tid": 7, "ts": 1716454224133078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073368, "dur": 12,
    "args": {
      "External id": 172488,
      "cbid": 211, "correlation": 172488
    }
  },
  {
    "ph": "s", "id": 172488, "pid": 76337, "tid": -914061504, "ts": 1716454224073368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224133098, "dur": 16,
    "args": {
      "External id": 172510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172510,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172510, "pid": 5, "tid": 7, "ts": 1716454224133098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073399, "dur": 8,
    "args": {
      "External id": 172510,
      "cbid": 211, "correlation": 172510
    }
  },
  {
    "ph": "s", "id": 172510, "pid": 76337, "tid": -914061504, "ts": 1716454224073399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073486, "dur": 1,
    "args": {
      "External id": 172521,
      "cbid": 251, "correlation": 172521
    }
  },
  {
    "ph": "f", "id": 172521, "pid": 76337, "tid": -914061504, "ts": 1716454224073486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224133115, "dur": 89,
    "args": {
      "External id": 172522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172522,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172522, "pid": 5, "tid": 7, "ts": 1716454224133115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073491, "dur": 14,
    "args": {
      "External id": 172522,
      "cbid": 211, "correlation": 172522
    }
  },
  {
    "ph": "s", "id": 172522, "pid": 76337, "tid": -914061504, "ts": 1716454224073491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073563, "dur": 1,
    "args": {
      "External id": 172533,
      "cbid": 251, "correlation": 172533
    }
  },
  {
    "ph": "f", "id": 172533, "pid": 76337, "tid": -914061504, "ts": 1716454224073563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073567, "dur": 0,
    "args": {
      "External id": 172534,
      "cbid": 251, "correlation": 172534
    }
  },
  {
    "ph": "f", "id": 172534, "pid": 76337, "tid": -914061504, "ts": 1716454224073567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224133206, "dur": 12,
    "args": {
      "External id": 172535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172535,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172535, "pid": 5, "tid": 7, "ts": 1716454224133206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073569, "dur": 12,
    "args": {
      "External id": 172535,
      "cbid": 211, "correlation": 172535
    }
  },
  {
    "ph": "s", "id": 172535, "pid": 76337, "tid": -914061504, "ts": 1716454224073569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224133220, "dur": 5,
    "args": {
      "External id": 172537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172537,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172537, "pid": 5, "tid": 7, "ts": 1716454224133220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073583, "dur": 6,
    "args": {
      "External id": 172537,
      "cbid": 211, "correlation": 172537
    }
  },
  {
    "ph": "s", "id": 172537, "pid": 76337, "tid": -914061504, "ts": 1716454224073583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073640, "dur": 1,
    "args": {
      "External id": 172548,
      "cbid": 251, "correlation": 172548
    }
  },
  {
    "ph": "f", "id": 172548, "pid": 76337, "tid": -914061504, "ts": 1716454224073640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073643, "dur": 0,
    "args": {
      "External id": 172549,
      "cbid": 251, "correlation": 172549
    }
  },
  {
    "ph": "f", "id": 172549, "pid": 76337, "tid": -914061504, "ts": 1716454224073643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224133226, "dur": 9,
    "args": {
      "External id": 172550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172550,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172550, "pid": 5, "tid": 7, "ts": 1716454224133226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073645, "dur": 12,
    "args": {
      "External id": 172550,
      "cbid": 211, "correlation": 172550
    }
  },
  {
    "ph": "s", "id": 172550, "pid": 76337, "tid": -914061504, "ts": 1716454224073645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224133236, "dur": 4,
    "args": {
      "External id": 172552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172552,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172552, "pid": 5, "tid": 7, "ts": 1716454224133236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073658, "dur": 6,
    "args": {
      "External id": 172552,
      "cbid": 211, "correlation": 172552
    }
  },
  {
    "ph": "s", "id": 172552, "pid": 76337, "tid": -914061504, "ts": 1716454224073658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224133241, "dur": 56,
    "args": {
      "External id": 172577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172577,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172577, "pid": 5, "tid": 7, "ts": 1716454224133241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073735, "dur": 13,
    "args": {
      "External id": 172577,
      "cbid": 211, "correlation": 172577
    }
  },
  {
    "ph": "s", "id": 172577, "pid": 76337, "tid": -914061504, "ts": 1716454224073735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073833, "dur": 2,
    "args": {
      "External id": 172595,
      "cbid": 251, "correlation": 172595
    }
  },
  {
    "ph": "f", "id": 172595, "pid": 76337, "tid": -914061504, "ts": 1716454224073833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224133298, "dur": 93,
    "args": {
      "External id": 172597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172597,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172597, "pid": 5, "tid": 7, "ts": 1716454224133298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073840, "dur": 14,
    "args": {
      "External id": 172597,
      "cbid": 211, "correlation": 172597
    }
  },
  {
    "ph": "s", "id": 172597, "pid": 76337, "tid": -914061504, "ts": 1716454224073840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224133392, "dur": 9,
    "args": {
      "External id": 172605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172605,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172605, "pid": 5, "tid": 7, "ts": 1716454224133392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073910, "dur": 12,
    "args": {
      "External id": 172605,
      "cbid": 211, "correlation": 172605
    }
  },
  {
    "ph": "s", "id": 172605, "pid": 76337, "tid": -914061504, "ts": 1716454224073910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224133402, "dur": 21,
    "args": {
      "External id": 172613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172613,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172613, "pid": 5, "tid": 7, "ts": 1716454224133402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224073951, "dur": 9,
    "args": {
      "External id": 172613,
      "cbid": 211, "correlation": 172613
    }
  },
  {
    "ph": "s", "id": 172613, "pid": 76337, "tid": -914061504, "ts": 1716454224073951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224133425, "dur": 18,
    "args": {
      "External id": 172635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172635,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172635, "pid": 5, "tid": 7, "ts": 1716454224133425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074011, "dur": 11,
    "args": {
      "External id": 172635,
      "cbid": 211, "correlation": 172635
    }
  },
  {
    "ph": "s", "id": 172635, "pid": 76337, "tid": -914061504, "ts": 1716454224074011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074100, "dur": 1,
    "args": {
      "External id": 172651,
      "cbid": 251, "correlation": 172651
    }
  },
  {
    "ph": "f", "id": 172651, "pid": 76337, "tid": -914061504, "ts": 1716454224074100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074105, "dur": 0,
    "args": {
      "External id": 172653,
      "cbid": 251, "correlation": 172653
    }
  },
  {
    "ph": "f", "id": 172653, "pid": 76337, "tid": -914061504, "ts": 1716454224074105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224133444, "dur": 496,
    "args": {
      "External id": 172654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172654,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172654, "pid": 5, "tid": 7, "ts": 1716454224133444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074107, "dur": 13,
    "args": {
      "External id": 172654,
      "cbid": 211, "correlation": 172654
    }
  },
  {
    "ph": "s", "id": 172654, "pid": 76337, "tid": -914061504, "ts": 1716454224074107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224133941, "dur": 67,
    "args": {
      "External id": 172662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172662, "pid": 5, "tid": 7, "ts": 1716454224133941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074174, "dur": 13,
    "args": {
      "External id": 172662,
      "cbid": 211, "correlation": 172662
    }
  },
  {
    "ph": "s", "id": 172662, "pid": 76337, "tid": -914061504, "ts": 1716454224074174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224134010, "dur": 66,
    "args": {
      "External id": 172670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172670,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172670, "pid": 5, "tid": 7, "ts": 1716454224134010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074207, "dur": 8,
    "args": {
      "External id": 172670,
      "cbid": 211, "correlation": 172670
    }
  },
  {
    "ph": "s", "id": 172670, "pid": 76337, "tid": -914061504, "ts": 1716454224074207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074286, "dur": 1,
    "args": {
      "External id": 172686,
      "cbid": 251, "correlation": 172686
    }
  },
  {
    "ph": "f", "id": 172686, "pid": 76337, "tid": -914061504, "ts": 1716454224074286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224134078, "dur": 1,
    "args": {
      "External id": 172688,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 172688,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 172688, "pid": 5, "tid": 7, "ts": 1716454224134078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074291, "dur": 11,
    "args": {
      "External id": 172688,
      "cbid": 51, "correlation": 172688
    }
  },
  {
    "ph": "s", "id": 172688, "pid": 76337, "tid": -914061504, "ts": 1716454224074291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224134082, "dur": 272,
    "args": {
      "External id": 172689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172689,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172689, "pid": 5, "tid": 7, "ts": 1716454224134082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074304, "dur": 11,
    "args": {
      "External id": 172689,
      "cbid": 211, "correlation": 172689
    }
  },
  {
    "ph": "s", "id": 172689, "pid": 76337, "tid": -914061504, "ts": 1716454224074304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224134355, "dur": 14,
    "args": {
      "External id": 172697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172697,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172697, "pid": 5, "tid": 7, "ts": 1716454224134355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074346, "dur": 11,
    "args": {
      "External id": 172697,
      "cbid": 211, "correlation": 172697
    }
  },
  {
    "ph": "s", "id": 172697, "pid": 76337, "tid": -914061504, "ts": 1716454224074346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224134370, "dur": 38,
    "args": {
      "External id": 172708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172708, "pid": 5, "tid": 7, "ts": 1716454224134370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074415, "dur": 12,
    "args": {
      "External id": 172708,
      "cbid": 211, "correlation": 172708
    }
  },
  {
    "ph": "s", "id": 172708, "pid": 76337, "tid": -914061504, "ts": 1716454224074415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074480, "dur": 0,
    "args": {
      "External id": 172720,
      "cbid": 317, "correlation": 172720
    }
  },
  {
    "ph": "f", "id": 172720, "pid": 76337, "tid": -914061504, "ts": 1716454224074480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074481, "dur": 0,
    "args": {
      "External id": 172721,
      "cbid": 203, "correlation": 172721
    }
  },
  {
    "ph": "f", "id": 172721, "pid": 76337, "tid": -914061504, "ts": 1716454224074481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074482, "dur": 0,
    "args": {
      "External id": 172722,
      "cbid": 205, "correlation": 172722
    }
  },
  {
    "ph": "f", "id": 172722, "pid": 76337, "tid": -914061504, "ts": 1716454224074482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224134410, "dur": 15,
    "args": {
      "External id": 172726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172726,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172726, "pid": 5, "tid": 7, "ts": 1716454224134410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074497, "dur": 12,
    "args": {
      "External id": 172726,
      "cbid": 211, "correlation": 172726
    }
  },
  {
    "ph": "s", "id": 172726, "pid": 76337, "tid": -914061504, "ts": 1716454224074497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224134426, "dur": 4,
    "args": {
      "External id": 172728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172728,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172728, "pid": 5, "tid": 7, "ts": 1716454224134426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074514, "dur": 6,
    "args": {
      "External id": 172728,
      "cbid": 211, "correlation": 172728
    }
  },
  {
    "ph": "s", "id": 172728, "pid": 76337, "tid": -914061504, "ts": 1716454224074514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074522, "dur": 0,
    "args": {
      "External id": 172729,
      "cbid": 51, "correlation": 172729
    }
  },
  {
    "ph": "s", "id": 172729, "pid": 76337, "tid": -914061504, "ts": 1716454224074522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224134431, "dur": 98,
    "args": {
      "External id": 172730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172730,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 172730, "pid": 5, "tid": 7, "ts": 1716454224134431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074523, "dur": 5,
    "args": {
      "External id": 172730,
      "cbid": 211, "correlation": 172730
    }
  },
  {
    "ph": "s", "id": 172730, "pid": 76337, "tid": -914061504, "ts": 1716454224074523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224134531, "dur": 16,
    "args": {
      "External id": 172735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172735,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172735, "pid": 5, "tid": 7, "ts": 1716454224134531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074551, "dur": 8,
    "args": {
      "External id": 172735,
      "cbid": 211, "correlation": 172735
    }
  },
  {
    "ph": "s", "id": 172735, "pid": 76337, "tid": -914061504, "ts": 1716454224074551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224134548, "dur": 12,
    "args": {
      "External id": 172743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172743,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172743, "pid": 5, "tid": 7, "ts": 1716454224134548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074582, "dur": 8,
    "args": {
      "External id": 172743,
      "cbid": 211, "correlation": 172743
    }
  },
  {
    "ph": "s", "id": 172743, "pid": 76337, "tid": -914061504, "ts": 1716454224074582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224134561, "dur": 30,
    "args": {
      "External id": 172752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172752,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172752, "pid": 5, "tid": 7, "ts": 1716454224134561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074622, "dur": 10,
    "args": {
      "External id": 172752,
      "cbid": 211, "correlation": 172752
    }
  },
  {
    "ph": "s", "id": 172752, "pid": 76337, "tid": -914061504, "ts": 1716454224074622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224134592, "dur": 30,
    "args": {
      "External id": 172772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172772,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 172772, "pid": 5, "tid": 7, "ts": 1716454224134592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074693, "dur": 12,
    "args": {
      "External id": 172772,
      "cbid": 211, "correlation": 172772
    }
  },
  {
    "ph": "s", "id": 172772, "pid": 76337, "tid": -914061504, "ts": 1716454224074693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224134624, "dur": 5,
    "args": {
      "External id": 172784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172784,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172784, "pid": 5, "tid": 7, "ts": 1716454224134624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074715, "dur": 6,
    "args": {
      "External id": 172784,
      "cbid": 211, "correlation": 172784
    }
  },
  {
    "ph": "s", "id": 172784, "pid": 76337, "tid": -914061504, "ts": 1716454224074715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224134631, "dur": 31,
    "args": {
      "External id": 172787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172787,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172787, "pid": 5, "tid": 7, "ts": 1716454224134631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074733, "dur": 7,
    "args": {
      "External id": 172787,
      "cbid": 211, "correlation": 172787
    }
  },
  {
    "ph": "s", "id": 172787, "pid": 76337, "tid": -914061504, "ts": 1716454224074733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224134663, "dur": 22,
    "args": {
      "External id": 172796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172796,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172796, "pid": 5, "tid": 7, "ts": 1716454224134663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074772, "dur": 10,
    "args": {
      "External id": 172796,
      "cbid": 211, "correlation": 172796
    }
  },
  {
    "ph": "s", "id": 172796, "pid": 76337, "tid": -914061504, "ts": 1716454224074772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074825, "dur": 0,
    "args": {
      "External id": 172806,
      "cbid": 317, "correlation": 172806
    }
  },
  {
    "ph": "f", "id": 172806, "pid": 76337, "tid": -914061504, "ts": 1716454224074825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074826, "dur": 0,
    "args": {
      "External id": 172807,
      "cbid": 203, "correlation": 172807
    }
  },
  {
    "ph": "f", "id": 172807, "pid": 76337, "tid": -914061504, "ts": 1716454224074826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074827, "dur": 0,
    "args": {
      "External id": 172808,
      "cbid": 205, "correlation": 172808
    }
  },
  {
    "ph": "f", "id": 172808, "pid": 76337, "tid": -914061504, "ts": 1716454224074827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224134686, "dur": 22,
    "args": {
      "External id": 172812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172812, "pid": 5, "tid": 7, "ts": 1716454224134686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074840, "dur": 12,
    "args": {
      "External id": 172812,
      "cbid": 211, "correlation": 172812
    }
  },
  {
    "ph": "s", "id": 172812, "pid": 76337, "tid": -914061504, "ts": 1716454224074840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224134710, "dur": 323,
    "args": {
      "External id": 172814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172814,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172814, "pid": 5, "tid": 7, "ts": 1716454224134710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074855, "dur": 5,
    "args": {
      "External id": 172814,
      "cbid": 211, "correlation": 172814
    }
  },
  {
    "ph": "s", "id": 172814, "pid": 76337, "tid": -914061504, "ts": 1716454224074855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224135035, "dur": 1,
    "args": {
      "External id": 172816,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 172816,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 172816, "pid": 5, "tid": 7, "ts": 1716454224135035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074866, "dur": 7,
    "args": {
      "External id": 172816,
      "cbid": 51, "correlation": 172816
    }
  },
  {
    "ph": "s", "id": 172816, "pid": 76337, "tid": -914061504, "ts": 1716454224074866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224135039, "dur": 1265,
    "args": {
      "External id": 172817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172817,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172817, "pid": 5, "tid": 7, "ts": 1716454224135039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074874, "dur": 6,
    "args": {
      "External id": 172817,
      "cbid": 211, "correlation": 172817
    }
  },
  {
    "ph": "s", "id": 172817, "pid": 76337, "tid": -914061504, "ts": 1716454224074874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224136305, "dur": 13,
    "args": {
      "External id": 172819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172819,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172819, "pid": 5, "tid": 7, "ts": 1716454224136305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074884, "dur": 5,
    "args": {
      "External id": 172819,
      "cbid": 211, "correlation": 172819
    }
  },
  {
    "ph": "s", "id": 172819, "pid": 76337, "tid": -914061504, "ts": 1716454224074884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224136319, "dur": 15,
    "args": {
      "External id": 172825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172825,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172825, "pid": 5, "tid": 7, "ts": 1716454224136319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074913, "dur": 8,
    "args": {
      "External id": 172825,
      "cbid": 211, "correlation": 172825
    }
  },
  {
    "ph": "s", "id": 172825, "pid": 76337, "tid": -914061504, "ts": 1716454224074913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224136336, "dur": 4,
    "args": {
      "External id": 172833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172833,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 172833, "pid": 5, "tid": 7, "ts": 1716454224136336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224074958, "dur": 9,
    "args": {
      "External id": 172833,
      "cbid": 211, "correlation": 172833
    }
  },
  {
    "ph": "s", "id": 172833, "pid": 76337, "tid": -914061504, "ts": 1716454224074958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075031, "dur": 1,
    "args": {
      "External id": 172849,
      "cbid": 251, "correlation": 172849
    }
  },
  {
    "ph": "f", "id": 172849, "pid": 76337, "tid": -914061504, "ts": 1716454224075031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075036, "dur": 0,
    "args": {
      "External id": 172851,
      "cbid": 251, "correlation": 172851
    }
  },
  {
    "ph": "f", "id": 172851, "pid": 76337, "tid": -914061504, "ts": 1716454224075036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224136341, "dur": 13,
    "args": {
      "External id": 172852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172852,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172852, "pid": 5, "tid": 7, "ts": 1716454224136341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075038, "dur": 12,
    "args": {
      "External id": 172852,
      "cbid": 211, "correlation": 172852
    }
  },
  {
    "ph": "s", "id": 172852, "pid": 76337, "tid": -914061504, "ts": 1716454224075038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224136356, "dur": 6,
    "args": {
      "External id": 172854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172854,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172854, "pid": 5, "tid": 7, "ts": 1716454224136356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075051, "dur": 6,
    "args": {
      "External id": 172854,
      "cbid": 211, "correlation": 172854
    }
  },
  {
    "ph": "s", "id": 172854, "pid": 76337, "tid": -914061504, "ts": 1716454224075051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224136362, "dur": 17,
    "args": {
      "External id": 172864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172864, "pid": 5, "tid": 7, "ts": 1716454224136362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075111, "dur": 11,
    "args": {
      "External id": 172864,
      "cbid": 211, "correlation": 172864
    }
  },
  {
    "ph": "s", "id": 172864, "pid": 76337, "tid": -914061504, "ts": 1716454224075111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224136381, "dur": 17,
    "args": {
      "External id": 172884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172884,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 172884, "pid": 5, "tid": 7, "ts": 1716454224136381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075177, "dur": 10,
    "args": {
      "External id": 172884,
      "cbid": 211, "correlation": 172884
    }
  },
  {
    "ph": "s", "id": 172884, "pid": 76337, "tid": -914061504, "ts": 1716454224075177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224136399, "dur": 5,
    "args": {
      "External id": 172896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172896,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 172896, "pid": 5, "tid": 7, "ts": 1716454224136399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075197, "dur": 6,
    "args": {
      "External id": 172896,
      "cbid": 211, "correlation": 172896
    }
  },
  {
    "ph": "s", "id": 172896, "pid": 76337, "tid": -914061504, "ts": 1716454224075197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224136405, "dur": 17,
    "args": {
      "External id": 172899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172899,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172899, "pid": 5, "tid": 7, "ts": 1716454224136405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075216, "dur": 6,
    "args": {
      "External id": 172899,
      "cbid": 211, "correlation": 172899
    }
  },
  {
    "ph": "s", "id": 172899, "pid": 76337, "tid": -914061504, "ts": 1716454224075216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224136424, "dur": 11,
    "args": {
      "External id": 172908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172908,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172908, "pid": 5, "tid": 7, "ts": 1716454224136424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075256, "dur": 10,
    "args": {
      "External id": 172908,
      "cbid": 211, "correlation": 172908
    }
  },
  {
    "ph": "s", "id": 172908, "pid": 76337, "tid": -914061504, "ts": 1716454224075256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075318, "dur": 0,
    "args": {
      "External id": 172918,
      "cbid": 317, "correlation": 172918
    }
  },
  {
    "ph": "f", "id": 172918, "pid": 76337, "tid": -914061504, "ts": 1716454224075318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075319, "dur": 0,
    "args": {
      "External id": 172919,
      "cbid": 203, "correlation": 172919
    }
  },
  {
    "ph": "f", "id": 172919, "pid": 76337, "tid": -914061504, "ts": 1716454224075319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075319, "dur": 0,
    "args": {
      "External id": 172920,
      "cbid": 205, "correlation": 172920
    }
  },
  {
    "ph": "f", "id": 172920, "pid": 76337, "tid": -914061504, "ts": 1716454224075319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224136436, "dur": 12,
    "args": {
      "External id": 172924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172924,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172924, "pid": 5, "tid": 7, "ts": 1716454224136436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075333, "dur": 12,
    "args": {
      "External id": 172924,
      "cbid": 211, "correlation": 172924
    }
  },
  {
    "ph": "s", "id": 172924, "pid": 76337, "tid": -914061504, "ts": 1716454224075333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224136449, "dur": 164,
    "args": {
      "External id": 172926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172926,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172926, "pid": 5, "tid": 7, "ts": 1716454224136449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075347, "dur": 5,
    "args": {
      "External id": 172926,
      "cbid": 211, "correlation": 172926
    }
  },
  {
    "ph": "s", "id": 172926, "pid": 76337, "tid": -914061504, "ts": 1716454224075347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224136615, "dur": 1,
    "args": {
      "External id": 172928,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 172928,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 172928, "pid": 5, "tid": 7, "ts": 1716454224136615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075358, "dur": 7,
    "args": {
      "External id": 172928,
      "cbid": 51, "correlation": 172928
    }
  },
  {
    "ph": "s", "id": 172928, "pid": 76337, "tid": -914061504, "ts": 1716454224075358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224136619, "dur": 653,
    "args": {
      "External id": 172929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172929,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 172929, "pid": 5, "tid": 7, "ts": 1716454224136619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075366, "dur": 6,
    "args": {
      "External id": 172929,
      "cbid": 211, "correlation": 172929
    }
  },
  {
    "ph": "s", "id": 172929, "pid": 76337, "tid": -914061504, "ts": 1716454224075366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224137273, "dur": 13,
    "args": {
      "External id": 172931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172931,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172931, "pid": 5, "tid": 7, "ts": 1716454224137273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075375, "dur": 5,
    "args": {
      "External id": 172931,
      "cbid": 211, "correlation": 172931
    }
  },
  {
    "ph": "s", "id": 172931, "pid": 76337, "tid": -914061504, "ts": 1716454224075375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224137287, "dur": 15,
    "args": {
      "External id": 172937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172937,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172937, "pid": 5, "tid": 7, "ts": 1716454224137287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075404, "dur": 10,
    "args": {
      "External id": 172937,
      "cbid": 211, "correlation": 172937
    }
  },
  {
    "ph": "s", "id": 172937, "pid": 76337, "tid": -914061504, "ts": 1716454224075404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075464, "dur": 0,
    "args": {
      "External id": 172947,
      "cbid": 317, "correlation": 172947
    }
  },
  {
    "ph": "f", "id": 172947, "pid": 76337, "tid": -914061504, "ts": 1716454224075464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075465, "dur": 0,
    "args": {
      "External id": 172948,
      "cbid": 203, "correlation": 172948
    }
  },
  {
    "ph": "f", "id": 172948, "pid": 76337, "tid": -914061504, "ts": 1716454224075465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075465, "dur": 0,
    "args": {
      "External id": 172949,
      "cbid": 205, "correlation": 172949
    }
  },
  {
    "ph": "f", "id": 172949, "pid": 76337, "tid": -914061504, "ts": 1716454224075465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224137303, "dur": 21,
    "args": {
      "External id": 172953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172953,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172953, "pid": 5, "tid": 7, "ts": 1716454224137303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075477, "dur": 12,
    "args": {
      "External id": 172953,
      "cbid": 211, "correlation": 172953
    }
  },
  {
    "ph": "s", "id": 172953, "pid": 76337, "tid": -914061504, "ts": 1716454224075477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224137326, "dur": 4,
    "args": {
      "External id": 172955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172955,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 172955, "pid": 5, "tid": 7, "ts": 1716454224137326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075492, "dur": 6,
    "args": {
      "External id": 172955,
      "cbid": 211, "correlation": 172955
    }
  },
  {
    "ph": "s", "id": 172955, "pid": 76337, "tid": -914061504, "ts": 1716454224075492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075501, "dur": 0,
    "args": {
      "External id": 172956,
      "cbid": 51, "correlation": 172956
    }
  },
  {
    "ph": "s", "id": 172956, "pid": 76337, "tid": -914061504, "ts": 1716454224075501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224137331, "dur": 173,
    "args": {
      "External id": 172957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172957,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 172957, "pid": 5, "tid": 7, "ts": 1716454224137331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075502, "dur": 5,
    "args": {
      "External id": 172957,
      "cbid": 211, "correlation": 172957
    }
  },
  {
    "ph": "s", "id": 172957, "pid": 76337, "tid": -914061504, "ts": 1716454224075502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224137505, "dur": 16,
    "args": {
      "External id": 172962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172962,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172962, "pid": 5, "tid": 7, "ts": 1716454224137505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075527, "dur": 8,
    "args": {
      "External id": 172962,
      "cbid": 211, "correlation": 172962
    }
  },
  {
    "ph": "s", "id": 172962, "pid": 76337, "tid": -914061504, "ts": 1716454224075527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224137522, "dur": 12,
    "args": {
      "External id": 172970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172970,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172970, "pid": 5, "tid": 7, "ts": 1716454224137522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075556, "dur": 8,
    "args": {
      "External id": 172970,
      "cbid": 211, "correlation": 172970
    }
  },
  {
    "ph": "s", "id": 172970, "pid": 76337, "tid": -914061504, "ts": 1716454224075556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224137535, "dur": 10,
    "args": {
      "External id": 172978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172978,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 172978, "pid": 5, "tid": 7, "ts": 1716454224137535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075584, "dur": 8,
    "args": {
      "External id": 172978,
      "cbid": 211, "correlation": 172978
    }
  },
  {
    "ph": "s", "id": 172978, "pid": 76337, "tid": -914061504, "ts": 1716454224075584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224137547, "dur": 18,
    "args": {
      "External id": 172998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 172998,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 172998, "pid": 5, "tid": 7, "ts": 1716454224137547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075669, "dur": 12,
    "args": {
      "External id": 172998,
      "cbid": 211, "correlation": 172998
    }
  },
  {
    "ph": "s", "id": 172998, "pid": 76337, "tid": -914061504, "ts": 1716454224075669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224137566, "dur": 5,
    "args": {
      "External id": 173010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173010,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 173010, "pid": 5, "tid": 7, "ts": 1716454224137566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075691, "dur": 6,
    "args": {
      "External id": 173010,
      "cbid": 211, "correlation": 173010
    }
  },
  {
    "ph": "s", "id": 173010, "pid": 76337, "tid": -914061504, "ts": 1716454224075691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224137572, "dur": 17,
    "args": {
      "External id": 173013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173013, "pid": 5, "tid": 7, "ts": 1716454224137572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075709, "dur": 7,
    "args": {
      "External id": 173013,
      "cbid": 211, "correlation": 173013
    }
  },
  {
    "ph": "s", "id": 173013, "pid": 76337, "tid": -914061504, "ts": 1716454224075709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075767, "dur": 0,
    "args": {
      "External id": 173024,
      "cbid": 317, "correlation": 173024
    }
  },
  {
    "ph": "f", "id": 173024, "pid": 76337, "tid": -914061504, "ts": 1716454224075767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075768, "dur": 0,
    "args": {
      "External id": 173025,
      "cbid": 203, "correlation": 173025
    }
  },
  {
    "ph": "f", "id": 173025, "pid": 76337, "tid": -914061504, "ts": 1716454224075768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075769, "dur": 0,
    "args": {
      "External id": 173026,
      "cbid": 205, "correlation": 173026
    }
  },
  {
    "ph": "f", "id": 173026, "pid": 76337, "tid": -914061504, "ts": 1716454224075769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224137591, "dur": 12,
    "args": {
      "External id": 173030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173030,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173030, "pid": 5, "tid": 7, "ts": 1716454224137591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075782, "dur": 12,
    "args": {
      "External id": 173030,
      "cbid": 211, "correlation": 173030
    }
  },
  {
    "ph": "s", "id": 173030, "pid": 76337, "tid": -914061504, "ts": 1716454224075782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224137604, "dur": 3,
    "args": {
      "External id": 173032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173032,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173032, "pid": 5, "tid": 7, "ts": 1716454224137604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075797, "dur": 6,
    "args": {
      "External id": 173032,
      "cbid": 211, "correlation": 173032
    }
  },
  {
    "ph": "s", "id": 173032, "pid": 76337, "tid": -914061504, "ts": 1716454224075797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075805, "dur": 0,
    "args": {
      "External id": 173033,
      "cbid": 51, "correlation": 173033
    }
  },
  {
    "ph": "s", "id": 173033, "pid": 76337, "tid": -914061504, "ts": 1716454224075805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224137609, "dur": 91,
    "args": {
      "External id": 173034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173034,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 173034, "pid": 5, "tid": 7, "ts": 1716454224137609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075806, "dur": 5,
    "args": {
      "External id": 173034,
      "cbid": 211, "correlation": 173034
    }
  },
  {
    "ph": "s", "id": 173034, "pid": 76337, "tid": -914061504, "ts": 1716454224075806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224137701, "dur": 16,
    "args": {
      "External id": 173039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173039, "pid": 5, "tid": 7, "ts": 1716454224137701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075833, "dur": 8,
    "args": {
      "External id": 173039,
      "cbid": 211, "correlation": 173039
    }
  },
  {
    "ph": "s", "id": 173039, "pid": 76337, "tid": -914061504, "ts": 1716454224075833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224137718, "dur": 84,
    "args": {
      "External id": 173048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173048,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173048, "pid": 5, "tid": 7, "ts": 1716454224137718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075914, "dur": 15,
    "args": {
      "External id": 173048,
      "cbid": 211, "correlation": 173048
    }
  },
  {
    "ph": "s", "id": 173048, "pid": 76337, "tid": -914061504, "ts": 1716454224075914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224137803, "dur": 31,
    "args": {
      "External id": 173070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173070,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173070, "pid": 5, "tid": 7, "ts": 1716454224137803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224075971, "dur": 20,
    "args": {
      "External id": 173070,
      "cbid": 211, "correlation": 173070
    }
  },
  {
    "ph": "s", "id": 173070, "pid": 76337, "tid": -914061504, "ts": 1716454224075971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076070, "dur": 1,
    "args": {
      "External id": 173081,
      "cbid": 251, "correlation": 173081
    }
  },
  {
    "ph": "f", "id": 173081, "pid": 76337, "tid": -914061504, "ts": 1716454224076070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224137835, "dur": 166,
    "args": {
      "External id": 173082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173082,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173082, "pid": 5, "tid": 7, "ts": 1716454224137835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076076, "dur": 13,
    "args": {
      "External id": 173082,
      "cbid": 211, "correlation": 173082
    }
  },
  {
    "ph": "s", "id": 173082, "pid": 76337, "tid": -914061504, "ts": 1716454224076076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076145, "dur": 1,
    "args": {
      "External id": 173093,
      "cbid": 251, "correlation": 173093
    }
  },
  {
    "ph": "f", "id": 173093, "pid": 76337, "tid": -914061504, "ts": 1716454224076145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224138003, "dur": 160,
    "args": {
      "External id": 173094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173094,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173094, "pid": 5, "tid": 7, "ts": 1716454224138003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076150, "dur": 11,
    "args": {
      "External id": 173094,
      "cbid": 211, "correlation": 173094
    }
  },
  {
    "ph": "s", "id": 173094, "pid": 76337, "tid": -914061504, "ts": 1716454224076150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076214, "dur": 1,
    "args": {
      "External id": 173105,
      "cbid": 251, "correlation": 173105
    }
  },
  {
    "ph": "f", "id": 173105, "pid": 76337, "tid": -914061504, "ts": 1716454224076214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224138164, "dur": 160,
    "args": {
      "External id": 173106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173106,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173106, "pid": 5, "tid": 7, "ts": 1716454224138164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076219, "dur": 12,
    "args": {
      "External id": 173106,
      "cbid": 211, "correlation": 173106
    }
  },
  {
    "ph": "s", "id": 173106, "pid": 76337, "tid": -914061504, "ts": 1716454224076219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224138325, "dur": 339,
    "args": {
      "External id": 173131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173131,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173131, "pid": 5, "tid": 7, "ts": 1716454224138325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076304, "dur": 13,
    "args": {
      "External id": 173131,
      "cbid": 211, "correlation": 173131
    }
  },
  {
    "ph": "s", "id": 173131, "pid": 76337, "tid": -914061504, "ts": 1716454224076304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076404, "dur": 1,
    "args": {
      "External id": 173149,
      "cbid": 251, "correlation": 173149
    }
  },
  {
    "ph": "f", "id": 173149, "pid": 76337, "tid": -914061504, "ts": 1716454224076404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224138665, "dur": 166,
    "args": {
      "External id": 173151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173151,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173151, "pid": 5, "tid": 7, "ts": 1716454224138665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076410, "dur": 13,
    "args": {
      "External id": 173151,
      "cbid": 211, "correlation": 173151
    }
  },
  {
    "ph": "s", "id": 173151, "pid": 76337, "tid": -914061504, "ts": 1716454224076410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224138833, "dur": 19,
    "args": {
      "External id": 173159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173159,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173159, "pid": 5, "tid": 7, "ts": 1716454224138833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076482, "dur": 12,
    "args": {
      "External id": 173159,
      "cbid": 211, "correlation": 173159
    }
  },
  {
    "ph": "s", "id": 173159, "pid": 76337, "tid": -914061504, "ts": 1716454224076482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224138853, "dur": 27,
    "args": {
      "External id": 173167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173167,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173167, "pid": 5, "tid": 7, "ts": 1716454224138853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076520, "dur": 9,
    "args": {
      "External id": 173167,
      "cbid": 211, "correlation": 173167
    }
  },
  {
    "ph": "s", "id": 173167, "pid": 76337, "tid": -914061504, "ts": 1716454224076520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224138882, "dur": 18,
    "args": {
      "External id": 173178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173178, "pid": 5, "tid": 7, "ts": 1716454224138882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076592, "dur": 12,
    "args": {
      "External id": 173178,
      "cbid": 211, "correlation": 173178
    }
  },
  {
    "ph": "s", "id": 173178, "pid": 76337, "tid": -914061504, "ts": 1716454224076592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224138901, "dur": 16,
    "args": {
      "External id": 173200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173200,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173200, "pid": 5, "tid": 7, "ts": 1716454224138901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076622, "dur": 9,
    "args": {
      "External id": 173200,
      "cbid": 211, "correlation": 173200
    }
  },
  {
    "ph": "s", "id": 173200, "pid": 76337, "tid": -914061504, "ts": 1716454224076622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076709, "dur": 1,
    "args": {
      "External id": 173211,
      "cbid": 251, "correlation": 173211
    }
  },
  {
    "ph": "f", "id": 173211, "pid": 76337, "tid": -914061504, "ts": 1716454224076709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224138919, "dur": 90,
    "args": {
      "External id": 173212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173212,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173212, "pid": 5, "tid": 7, "ts": 1716454224138919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076714, "dur": 14,
    "args": {
      "External id": 173212,
      "cbid": 211, "correlation": 173212
    }
  },
  {
    "ph": "s", "id": 173212, "pid": 76337, "tid": -914061504, "ts": 1716454224076714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076784, "dur": 1,
    "args": {
      "External id": 173223,
      "cbid": 251, "correlation": 173223
    }
  },
  {
    "ph": "f", "id": 173223, "pid": 76337, "tid": -914061504, "ts": 1716454224076784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076788, "dur": 0,
    "args": {
      "External id": 173224,
      "cbid": 251, "correlation": 173224
    }
  },
  {
    "ph": "f", "id": 173224, "pid": 76337, "tid": -914061504, "ts": 1716454224076788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224139010, "dur": 12,
    "args": {
      "External id": 173225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173225,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173225, "pid": 5, "tid": 7, "ts": 1716454224139010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076789, "dur": 12,
    "args": {
      "External id": 173225,
      "cbid": 211, "correlation": 173225
    }
  },
  {
    "ph": "s", "id": 173225, "pid": 76337, "tid": -914061504, "ts": 1716454224076789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224139023, "dur": 6,
    "args": {
      "External id": 173227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173227,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173227, "pid": 5, "tid": 7, "ts": 1716454224139023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076803, "dur": 6,
    "args": {
      "External id": 173227,
      "cbid": 211, "correlation": 173227
    }
  },
  {
    "ph": "s", "id": 173227, "pid": 76337, "tid": -914061504, "ts": 1716454224076803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076860, "dur": 1,
    "args": {
      "External id": 173238,
      "cbid": 251, "correlation": 173238
    }
  },
  {
    "ph": "f", "id": 173238, "pid": 76337, "tid": -914061504, "ts": 1716454224076860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076863, "dur": 0,
    "args": {
      "External id": 173239,
      "cbid": 251, "correlation": 173239
    }
  },
  {
    "ph": "f", "id": 173239, "pid": 76337, "tid": -914061504, "ts": 1716454224076863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224139030, "dur": 9,
    "args": {
      "External id": 173240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173240,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173240, "pid": 5, "tid": 7, "ts": 1716454224139030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076865, "dur": 12,
    "args": {
      "External id": 173240,
      "cbid": 211, "correlation": 173240
    }
  },
  {
    "ph": "s", "id": 173240, "pid": 76337, "tid": -914061504, "ts": 1716454224076865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224139040, "dur": 4,
    "args": {
      "External id": 173242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173242,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173242, "pid": 5, "tid": 7, "ts": 1716454224139040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076878, "dur": 6,
    "args": {
      "External id": 173242,
      "cbid": 211, "correlation": 173242
    }
  },
  {
    "ph": "s", "id": 173242, "pid": 76337, "tid": -914061504, "ts": 1716454224076878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224139045, "dur": 56,
    "args": {
      "External id": 173267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173267,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173267, "pid": 5, "tid": 7, "ts": 1716454224139045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224076956, "dur": 12,
    "args": {
      "External id": 173267,
      "cbid": 211, "correlation": 173267
    }
  },
  {
    "ph": "s", "id": 173267, "pid": 76337, "tid": -914061504, "ts": 1716454224076956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077065, "dur": 1,
    "args": {
      "External id": 173285,
      "cbid": 251, "correlation": 173285
    }
  },
  {
    "ph": "f", "id": 173285, "pid": 76337, "tid": -914061504, "ts": 1716454224077065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224139102, "dur": 91,
    "args": {
      "External id": 173287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173287,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173287, "pid": 5, "tid": 7, "ts": 1716454224139102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077071, "dur": 15,
    "args": {
      "External id": 173287,
      "cbid": 211, "correlation": 173287
    }
  },
  {
    "ph": "s", "id": 173287, "pid": 76337, "tid": -914061504, "ts": 1716454224077071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224139194, "dur": 10,
    "args": {
      "External id": 173295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173295,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173295, "pid": 5, "tid": 7, "ts": 1716454224139194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077142, "dur": 12,
    "args": {
      "External id": 173295,
      "cbid": 211, "correlation": 173295
    }
  },
  {
    "ph": "s", "id": 173295, "pid": 76337, "tid": -914061504, "ts": 1716454224077142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224139205, "dur": 22,
    "args": {
      "External id": 173303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173303,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173303, "pid": 5, "tid": 7, "ts": 1716454224139205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077184, "dur": 9,
    "args": {
      "External id": 173303,
      "cbid": 211, "correlation": 173303
    }
  },
  {
    "ph": "s", "id": 173303, "pid": 76337, "tid": -914061504, "ts": 1716454224077184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224139228, "dur": 17,
    "args": {
      "External id": 173325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173325,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173325, "pid": 5, "tid": 7, "ts": 1716454224139228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077235, "dur": 10,
    "args": {
      "External id": 173325,
      "cbid": 211, "correlation": 173325
    }
  },
  {
    "ph": "s", "id": 173325, "pid": 76337, "tid": -914061504, "ts": 1716454224077235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077320, "dur": 1,
    "args": {
      "External id": 173341,
      "cbid": 251, "correlation": 173341
    }
  },
  {
    "ph": "f", "id": 173341, "pid": 76337, "tid": -914061504, "ts": 1716454224077320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077325, "dur": 0,
    "args": {
      "External id": 173343,
      "cbid": 251, "correlation": 173343
    }
  },
  {
    "ph": "f", "id": 173343, "pid": 76337, "tid": -914061504, "ts": 1716454224077325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224139247, "dur": 497,
    "args": {
      "External id": 173344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173344,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173344, "pid": 5, "tid": 7, "ts": 1716454224139247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077327, "dur": 14,
    "args": {
      "External id": 173344,
      "cbid": 211, "correlation": 173344
    }
  },
  {
    "ph": "s", "id": 173344, "pid": 76337, "tid": -914061504, "ts": 1716454224077327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224139745, "dur": 66,
    "args": {
      "External id": 173352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173352, "pid": 5, "tid": 7, "ts": 1716454224139745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077395, "dur": 12,
    "args": {
      "External id": 173352,
      "cbid": 211, "correlation": 173352
    }
  },
  {
    "ph": "s", "id": 173352, "pid": 76337, "tid": -914061504, "ts": 1716454224077395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224139813, "dur": 67,
    "args": {
      "External id": 173360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173360,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173360, "pid": 5, "tid": 7, "ts": 1716454224139813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077425, "dur": 9,
    "args": {
      "External id": 173360,
      "cbid": 211, "correlation": 173360
    }
  },
  {
    "ph": "s", "id": 173360, "pid": 76337, "tid": -914061504, "ts": 1716454224077425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077506, "dur": 1,
    "args": {
      "External id": 173376,
      "cbid": 251, "correlation": 173376
    }
  },
  {
    "ph": "f", "id": 173376, "pid": 76337, "tid": -914061504, "ts": 1716454224077506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224139882, "dur": 1,
    "args": {
      "External id": 173378,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 173378,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 173378, "pid": 5, "tid": 7, "ts": 1716454224139882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077511, "dur": 9,
    "args": {
      "External id": 173378,
      "cbid": 51, "correlation": 173378
    }
  },
  {
    "ph": "s", "id": 173378, "pid": 76337, "tid": -914061504, "ts": 1716454224077511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224139886, "dur": 271,
    "args": {
      "External id": 173379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173379,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173379, "pid": 5, "tid": 7, "ts": 1716454224139886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077522, "dur": 12,
    "args": {
      "External id": 173379,
      "cbid": 211, "correlation": 173379
    }
  },
  {
    "ph": "s", "id": 173379, "pid": 76337, "tid": -914061504, "ts": 1716454224077522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224140158, "dur": 14,
    "args": {
      "External id": 173387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173387,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173387, "pid": 5, "tid": 7, "ts": 1716454224140158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077564, "dur": 10,
    "args": {
      "External id": 173387,
      "cbid": 211, "correlation": 173387
    }
  },
  {
    "ph": "s", "id": 173387, "pid": 76337, "tid": -914061504, "ts": 1716454224077564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224140173, "dur": 38,
    "args": {
      "External id": 173398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173398,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173398, "pid": 5, "tid": 7, "ts": 1716454224140173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077633, "dur": 12,
    "args": {
      "External id": 173398,
      "cbid": 211, "correlation": 173398
    }
  },
  {
    "ph": "s", "id": 173398, "pid": 76337, "tid": -914061504, "ts": 1716454224077633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077695, "dur": 0,
    "args": {
      "External id": 173410,
      "cbid": 317, "correlation": 173410
    }
  },
  {
    "ph": "f", "id": 173410, "pid": 76337, "tid": -914061504, "ts": 1716454224077695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077696, "dur": 0,
    "args": {
      "External id": 173411,
      "cbid": 203, "correlation": 173411
    }
  },
  {
    "ph": "f", "id": 173411, "pid": 76337, "tid": -914061504, "ts": 1716454224077696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077697, "dur": 0,
    "args": {
      "External id": 173412,
      "cbid": 205, "correlation": 173412
    }
  },
  {
    "ph": "f", "id": 173412, "pid": 76337, "tid": -914061504, "ts": 1716454224077697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224140212, "dur": 15,
    "args": {
      "External id": 173416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173416,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173416, "pid": 5, "tid": 7, "ts": 1716454224140212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077712, "dur": 12,
    "args": {
      "External id": 173416,
      "cbid": 211, "correlation": 173416
    }
  },
  {
    "ph": "s", "id": 173416, "pid": 76337, "tid": -914061504, "ts": 1716454224077712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224140228, "dur": 4,
    "args": {
      "External id": 173418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173418, "pid": 5, "tid": 7, "ts": 1716454224140228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077728, "dur": 6,
    "args": {
      "External id": 173418,
      "cbid": 211, "correlation": 173418
    }
  },
  {
    "ph": "s", "id": 173418, "pid": 76337, "tid": -914061504, "ts": 1716454224077728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077737, "dur": 0,
    "args": {
      "External id": 173419,
      "cbid": 51, "correlation": 173419
    }
  },
  {
    "ph": "s", "id": 173419, "pid": 76337, "tid": -914061504, "ts": 1716454224077737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224140233, "dur": 96,
    "args": {
      "External id": 173420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173420,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 173420, "pid": 5, "tid": 7, "ts": 1716454224140233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077738, "dur": 5,
    "args": {
      "External id": 173420,
      "cbid": 211, "correlation": 173420
    }
  },
  {
    "ph": "s", "id": 173420, "pid": 76337, "tid": -914061504, "ts": 1716454224077738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224140331, "dur": 17,
    "args": {
      "External id": 173425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173425,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173425, "pid": 5, "tid": 7, "ts": 1716454224140331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077765, "dur": 9,
    "args": {
      "External id": 173425,
      "cbid": 211, "correlation": 173425
    }
  },
  {
    "ph": "s", "id": 173425, "pid": 76337, "tid": -914061504, "ts": 1716454224077765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224140348, "dur": 12,
    "args": {
      "External id": 173433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173433,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173433, "pid": 5, "tid": 7, "ts": 1716454224140348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077797, "dur": 8,
    "args": {
      "External id": 173433,
      "cbid": 211, "correlation": 173433
    }
  },
  {
    "ph": "s", "id": 173433, "pid": 76337, "tid": -914061504, "ts": 1716454224077797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224140361, "dur": 25,
    "args": {
      "External id": 173442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173442,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173442, "pid": 5, "tid": 7, "ts": 1716454224140361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077837, "dur": 10,
    "args": {
      "External id": 173442,
      "cbid": 211, "correlation": 173442
    }
  },
  {
    "ph": "s", "id": 173442, "pid": 76337, "tid": -914061504, "ts": 1716454224077837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224140388, "dur": 25,
    "args": {
      "External id": 173462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173462,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 173462, "pid": 5, "tid": 7, "ts": 1716454224140388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077907, "dur": 12,
    "args": {
      "External id": 173462,
      "cbid": 211, "correlation": 173462
    }
  },
  {
    "ph": "s", "id": 173462, "pid": 76337, "tid": -914061504, "ts": 1716454224077907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224140414, "dur": 5,
    "args": {
      "External id": 173474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173474,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 173474, "pid": 5, "tid": 7, "ts": 1716454224140414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077929, "dur": 7,
    "args": {
      "External id": 173474,
      "cbid": 211, "correlation": 173474
    }
  },
  {
    "ph": "s", "id": 173474, "pid": 76337, "tid": -914061504, "ts": 1716454224077929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224140420, "dur": 24,
    "args": {
      "External id": 173477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173477, "pid": 5, "tid": 7, "ts": 1716454224140420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077948, "dur": 6,
    "args": {
      "External id": 173477,
      "cbid": 211, "correlation": 173477
    }
  },
  {
    "ph": "s", "id": 173477, "pid": 76337, "tid": -914061504, "ts": 1716454224077948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224140446, "dur": 17,
    "args": {
      "External id": 173486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173486,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173486, "pid": 5, "tid": 7, "ts": 1716454224140446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224077995, "dur": 11,
    "args": {
      "External id": 173486,
      "cbid": 211, "correlation": 173486
    }
  },
  {
    "ph": "s", "id": 173486, "pid": 76337, "tid": -914061504, "ts": 1716454224077995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078048, "dur": 0,
    "args": {
      "External id": 173496,
      "cbid": 317, "correlation": 173496
    }
  },
  {
    "ph": "f", "id": 173496, "pid": 76337, "tid": -914061504, "ts": 1716454224078048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078049, "dur": 0,
    "args": {
      "External id": 173497,
      "cbid": 203, "correlation": 173497
    }
  },
  {
    "ph": "f", "id": 173497, "pid": 76337, "tid": -914061504, "ts": 1716454224078049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078050, "dur": 0,
    "args": {
      "External id": 173498,
      "cbid": 205, "correlation": 173498
    }
  },
  {
    "ph": "f", "id": 173498, "pid": 76337, "tid": -914061504, "ts": 1716454224078050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224140464, "dur": 18,
    "args": {
      "External id": 173502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173502,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173502, "pid": 5, "tid": 7, "ts": 1716454224140464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078065, "dur": 11,
    "args": {
      "External id": 173502,
      "cbid": 211, "correlation": 173502
    }
  },
  {
    "ph": "s", "id": 173502, "pid": 76337, "tid": -914061504, "ts": 1716454224078065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224140483, "dur": 241,
    "args": {
      "External id": 173504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173504,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173504, "pid": 5, "tid": 7, "ts": 1716454224140483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078079, "dur": 6,
    "args": {
      "External id": 173504,
      "cbid": 211, "correlation": 173504
    }
  },
  {
    "ph": "s", "id": 173504, "pid": 76337, "tid": -914061504, "ts": 1716454224078079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224140726, "dur": 1,
    "args": {
      "External id": 173506,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 173506,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5263157894736842
    }
  },
  {
    "ph": "f", "id": 173506, "pid": 5, "tid": 7, "ts": 1716454224140726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078091, "dur": 8,
    "args": {
      "External id": 173506,
      "cbid": 51, "correlation": 173506
    }
  },
  {
    "ph": "s", "id": 173506, "pid": 76337, "tid": -914061504, "ts": 1716454224078091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224140730, "dur": 810,
    "args": {
      "External id": 173507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173507,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173507, "pid": 5, "tid": 7, "ts": 1716454224140730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078100, "dur": 6,
    "args": {
      "External id": 173507,
      "cbid": 211, "correlation": 173507
    }
  },
  {
    "ph": "s", "id": 173507, "pid": 76337, "tid": -914061504, "ts": 1716454224078100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224141541, "dur": 13,
    "args": {
      "External id": 173509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173509,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173509, "pid": 5, "tid": 7, "ts": 1716454224141541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078111, "dur": 5,
    "args": {
      "External id": 173509,
      "cbid": 211, "correlation": 173509
    }
  },
  {
    "ph": "s", "id": 173509, "pid": 76337, "tid": -914061504, "ts": 1716454224078111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224141556, "dur": 15,
    "args": {
      "External id": 173515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173515,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173515, "pid": 5, "tid": 7, "ts": 1716454224141556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078139, "dur": 9,
    "args": {
      "External id": 173515,
      "cbid": 211, "correlation": 173515
    }
  },
  {
    "ph": "s", "id": 173515, "pid": 76337, "tid": -914061504, "ts": 1716454224078139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224141572, "dur": 4,
    "args": {
      "External id": 173523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173523,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 173523, "pid": 5, "tid": 7, "ts": 1716454224141572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078184, "dur": 9,
    "args": {
      "External id": 173523,
      "cbid": 211, "correlation": 173523
    }
  },
  {
    "ph": "s", "id": 173523, "pid": 76337, "tid": -914061504, "ts": 1716454224078184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078250, "dur": 1,
    "args": {
      "External id": 173539,
      "cbid": 251, "correlation": 173539
    }
  },
  {
    "ph": "f", "id": 173539, "pid": 76337, "tid": -914061504, "ts": 1716454224078250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078255, "dur": 0,
    "args": {
      "External id": 173541,
      "cbid": 251, "correlation": 173541
    }
  },
  {
    "ph": "f", "id": 173541, "pid": 76337, "tid": -914061504, "ts": 1716454224078255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224141577, "dur": 14,
    "args": {
      "External id": 173542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173542,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173542, "pid": 5, "tid": 7, "ts": 1716454224141577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078256, "dur": 11,
    "args": {
      "External id": 173542,
      "cbid": 211, "correlation": 173542
    }
  },
  {
    "ph": "s", "id": 173542, "pid": 76337, "tid": -914061504, "ts": 1716454224078256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224141592, "dur": 5,
    "args": {
      "External id": 173544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173544,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173544, "pid": 5, "tid": 7, "ts": 1716454224141592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078269, "dur": 5,
    "args": {
      "External id": 173544,
      "cbid": 211, "correlation": 173544
    }
  },
  {
    "ph": "s", "id": 173544, "pid": 76337, "tid": -914061504, "ts": 1716454224078269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224141599, "dur": 17,
    "args": {
      "External id": 173554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173554,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173554, "pid": 5, "tid": 7, "ts": 1716454224141599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078327, "dur": 12,
    "args": {
      "External id": 173554,
      "cbid": 211, "correlation": 173554
    }
  },
  {
    "ph": "s", "id": 173554, "pid": 76337, "tid": -914061504, "ts": 1716454224078327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224141616, "dur": 19,
    "args": {
      "External id": 173574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173574,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 173574, "pid": 5, "tid": 7, "ts": 1716454224141616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078392, "dur": 11,
    "args": {
      "External id": 173574,
      "cbid": 211, "correlation": 173574
    }
  },
  {
    "ph": "s", "id": 173574, "pid": 76337, "tid": -914061504, "ts": 1716454224078392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224141636, "dur": 4,
    "args": {
      "External id": 173586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173586,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 173586, "pid": 5, "tid": 7, "ts": 1716454224141636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078413, "dur": 6,
    "args": {
      "External id": 173586,
      "cbid": 211, "correlation": 173586
    }
  },
  {
    "ph": "s", "id": 173586, "pid": 76337, "tid": -914061504, "ts": 1716454224078413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224141642, "dur": 17,
    "args": {
      "External id": 173589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173589, "pid": 5, "tid": 7, "ts": 1716454224141642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078431, "dur": 6,
    "args": {
      "External id": 173589,
      "cbid": 211, "correlation": 173589
    }
  },
  {
    "ph": "s", "id": 173589, "pid": 76337, "tid": -914061504, "ts": 1716454224078431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224141661, "dur": 10,
    "args": {
      "External id": 173598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173598,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173598, "pid": 5, "tid": 7, "ts": 1716454224141661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078471, "dur": 10,
    "args": {
      "External id": 173598,
      "cbid": 211, "correlation": 173598
    }
  },
  {
    "ph": "s", "id": 173598, "pid": 76337, "tid": -914061504, "ts": 1716454224078471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078534, "dur": 0,
    "args": {
      "External id": 173608,
      "cbid": 317, "correlation": 173608
    }
  },
  {
    "ph": "f", "id": 173608, "pid": 76337, "tid": -914061504, "ts": 1716454224078534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078535, "dur": 0,
    "args": {
      "External id": 173609,
      "cbid": 203, "correlation": 173609
    }
  },
  {
    "ph": "f", "id": 173609, "pid": 76337, "tid": -914061504, "ts": 1716454224078535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078536, "dur": 0,
    "args": {
      "External id": 173610,
      "cbid": 205, "correlation": 173610
    }
  },
  {
    "ph": "f", "id": 173610, "pid": 76337, "tid": -914061504, "ts": 1716454224078536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224141672, "dur": 11,
    "args": {
      "External id": 173614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173614,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173614, "pid": 5, "tid": 7, "ts": 1716454224141672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078549, "dur": 12,
    "args": {
      "External id": 173614,
      "cbid": 211, "correlation": 173614
    }
  },
  {
    "ph": "s", "id": 173614, "pid": 76337, "tid": -914061504, "ts": 1716454224078549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224141684, "dur": 163,
    "args": {
      "External id": 173616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173616,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173616, "pid": 5, "tid": 7, "ts": 1716454224141684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078563, "dur": 5,
    "args": {
      "External id": 173616,
      "cbid": 211, "correlation": 173616
    }
  },
  {
    "ph": "s", "id": 173616, "pid": 76337, "tid": -914061504, "ts": 1716454224078563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224141850, "dur": 1,
    "args": {
      "External id": 173618,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 173618,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 173618, "pid": 5, "tid": 7, "ts": 1716454224141850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078575, "dur": 7,
    "args": {
      "External id": 173618,
      "cbid": 51, "correlation": 173618
    }
  },
  {
    "ph": "s", "id": 173618, "pid": 76337, "tid": -914061504, "ts": 1716454224078575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224141854, "dur": 649,
    "args": {
      "External id": 173619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173619,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173619, "pid": 5, "tid": 7, "ts": 1716454224141854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078583, "dur": 6,
    "args": {
      "External id": 173619,
      "cbid": 211, "correlation": 173619
    }
  },
  {
    "ph": "s", "id": 173619, "pid": 76337, "tid": -914061504, "ts": 1716454224078583,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224142504, "dur": 13,
    "args": {
      "External id": 173621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173621,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173621, "pid": 5, "tid": 7, "ts": 1716454224142504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078594, "dur": 5,
    "args": {
      "External id": 173621,
      "cbid": 211, "correlation": 173621
    }
  },
  {
    "ph": "s", "id": 173621, "pid": 76337, "tid": -914061504, "ts": 1716454224078594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224142519, "dur": 15,
    "args": {
      "External id": 173627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173627,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173627, "pid": 5, "tid": 7, "ts": 1716454224142519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078622, "dur": 10,
    "args": {
      "External id": 173627,
      "cbid": 211, "correlation": 173627
    }
  },
  {
    "ph": "s", "id": 173627, "pid": 76337, "tid": -914061504, "ts": 1716454224078622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078681, "dur": 0,
    "args": {
      "External id": 173637,
      "cbid": 317, "correlation": 173637
    }
  },
  {
    "ph": "f", "id": 173637, "pid": 76337, "tid": -914061504, "ts": 1716454224078681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078682, "dur": 0,
    "args": {
      "External id": 173638,
      "cbid": 203, "correlation": 173638
    }
  },
  {
    "ph": "f", "id": 173638, "pid": 76337, "tid": -914061504, "ts": 1716454224078682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078683, "dur": 0,
    "args": {
      "External id": 173639,
      "cbid": 205, "correlation": 173639
    }
  },
  {
    "ph": "f", "id": 173639, "pid": 76337, "tid": -914061504, "ts": 1716454224078683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224142535, "dur": 17,
    "args": {
      "External id": 173643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173643,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173643, "pid": 5, "tid": 7, "ts": 1716454224142535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078697, "dur": 11,
    "args": {
      "External id": 173643,
      "cbid": 211, "correlation": 173643
    }
  },
  {
    "ph": "s", "id": 173643, "pid": 76337, "tid": -914061504, "ts": 1716454224078697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224142553, "dur": 4,
    "args": {
      "External id": 173645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173645,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173645, "pid": 5, "tid": 7, "ts": 1716454224142553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078713, "dur": 6,
    "args": {
      "External id": 173645,
      "cbid": 211, "correlation": 173645
    }
  },
  {
    "ph": "s", "id": 173645, "pid": 76337, "tid": -914061504, "ts": 1716454224078713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078721, "dur": 0,
    "args": {
      "External id": 173646,
      "cbid": 51, "correlation": 173646
    }
  },
  {
    "ph": "s", "id": 173646, "pid": 76337, "tid": -914061504, "ts": 1716454224078721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224142559, "dur": 132,
    "args": {
      "External id": 173647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173647,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 173647, "pid": 5, "tid": 7, "ts": 1716454224142559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078722, "dur": 5,
    "args": {
      "External id": 173647,
      "cbid": 211, "correlation": 173647
    }
  },
  {
    "ph": "s", "id": 173647, "pid": 76337, "tid": -914061504, "ts": 1716454224078722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224142692, "dur": 16,
    "args": {
      "External id": 173652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173652,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173652, "pid": 5, "tid": 7, "ts": 1716454224142692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078748, "dur": 8,
    "args": {
      "External id": 173652,
      "cbid": 211, "correlation": 173652
    }
  },
  {
    "ph": "s", "id": 173652, "pid": 76337, "tid": -914061504, "ts": 1716454224078748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224142709, "dur": 11,
    "args": {
      "External id": 173660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173660,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173660, "pid": 5, "tid": 7, "ts": 1716454224142709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078777, "dur": 8,
    "args": {
      "External id": 173660,
      "cbid": 211, "correlation": 173660
    }
  },
  {
    "ph": "s", "id": 173660, "pid": 76337, "tid": -914061504, "ts": 1716454224078777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224142721, "dur": 11,
    "args": {
      "External id": 173668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173668,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173668, "pid": 5, "tid": 7, "ts": 1716454224142721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078805, "dur": 9,
    "args": {
      "External id": 173668,
      "cbid": 211, "correlation": 173668
    }
  },
  {
    "ph": "s", "id": 173668, "pid": 76337, "tid": -914061504, "ts": 1716454224078805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224142733, "dur": 19,
    "args": {
      "External id": 173688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173688,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 173688, "pid": 5, "tid": 7, "ts": 1716454224142733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078890, "dur": 12,
    "args": {
      "External id": 173688,
      "cbid": 211, "correlation": 173688
    }
  },
  {
    "ph": "s", "id": 173688, "pid": 76337, "tid": -914061504, "ts": 1716454224078890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224142753, "dur": 4,
    "args": {
      "External id": 173700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173700,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 173700, "pid": 5, "tid": 7, "ts": 1716454224142753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078912, "dur": 6,
    "args": {
      "External id": 173700,
      "cbid": 211, "correlation": 173700
    }
  },
  {
    "ph": "s", "id": 173700, "pid": 76337, "tid": -914061504, "ts": 1716454224078912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224142759, "dur": 17,
    "args": {
      "External id": 173703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173703, "pid": 5, "tid": 7, "ts": 1716454224142759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078929, "dur": 7,
    "args": {
      "External id": 173703,
      "cbid": 211, "correlation": 173703
    }
  },
  {
    "ph": "s", "id": 173703, "pid": 76337, "tid": -914061504, "ts": 1716454224078929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078994, "dur": 0,
    "args": {
      "External id": 173714,
      "cbid": 317, "correlation": 173714
    }
  },
  {
    "ph": "f", "id": 173714, "pid": 76337, "tid": -914061504, "ts": 1716454224078994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078995, "dur": 0,
    "args": {
      "External id": 173715,
      "cbid": 203, "correlation": 173715
    }
  },
  {
    "ph": "f", "id": 173715, "pid": 76337, "tid": -914061504, "ts": 1716454224078995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224078995, "dur": 0,
    "args": {
      "External id": 173716,
      "cbid": 205, "correlation": 173716
    }
  },
  {
    "ph": "f", "id": 173716, "pid": 76337, "tid": -914061504, "ts": 1716454224078995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224142777, "dur": 11,
    "args": {
      "External id": 173720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173720,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173720, "pid": 5, "tid": 7, "ts": 1716454224142777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079009, "dur": 12,
    "args": {
      "External id": 173720,
      "cbid": 211, "correlation": 173720
    }
  },
  {
    "ph": "s", "id": 173720, "pid": 76337, "tid": -914061504, "ts": 1716454224079009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224142789, "dur": 3,
    "args": {
      "External id": 173722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173722,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173722, "pid": 5, "tid": 7, "ts": 1716454224142789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079025, "dur": 6,
    "args": {
      "External id": 173722,
      "cbid": 211, "correlation": 173722
    }
  },
  {
    "ph": "s", "id": 173722, "pid": 76337, "tid": -914061504, "ts": 1716454224079025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079034, "dur": 0,
    "args": {
      "External id": 173723,
      "cbid": 51, "correlation": 173723
    }
  },
  {
    "ph": "s", "id": 173723, "pid": 76337, "tid": -914061504, "ts": 1716454224079034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224142794, "dur": 92,
    "args": {
      "External id": 173724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173724,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 173724, "pid": 5, "tid": 7, "ts": 1716454224142794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079035, "dur": 5,
    "args": {
      "External id": 173724,
      "cbid": 211, "correlation": 173724
    }
  },
  {
    "ph": "s", "id": 173724, "pid": 76337, "tid": -914061504, "ts": 1716454224079035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224142887, "dur": 16,
    "args": {
      "External id": 173729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173729,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173729, "pid": 5, "tid": 7, "ts": 1716454224142887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079061, "dur": 9,
    "args": {
      "External id": 173729,
      "cbid": 211, "correlation": 173729
    }
  },
  {
    "ph": "s", "id": 173729, "pid": 76337, "tid": -914061504, "ts": 1716454224079061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224142904, "dur": 85,
    "args": {
      "External id": 173738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173738,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173738, "pid": 5, "tid": 7, "ts": 1716454224142904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079142, "dur": 13,
    "args": {
      "External id": 173738,
      "cbid": 211, "correlation": 173738
    }
  },
  {
    "ph": "s", "id": 173738, "pid": 76337, "tid": -914061504, "ts": 1716454224079142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224142990, "dur": 30,
    "args": {
      "External id": 173760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173760,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173760, "pid": 5, "tid": 7, "ts": 1716454224142990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079200, "dur": 10,
    "args": {
      "External id": 173760,
      "cbid": 211, "correlation": 173760
    }
  },
  {
    "ph": "s", "id": 173760, "pid": 76337, "tid": -914061504, "ts": 1716454224079200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079288, "dur": 1,
    "args": {
      "External id": 173771,
      "cbid": 251, "correlation": 173771
    }
  },
  {
    "ph": "f", "id": 173771, "pid": 76337, "tid": -914061504, "ts": 1716454224079288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224143021, "dur": 165,
    "args": {
      "External id": 173772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173772,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173772, "pid": 5, "tid": 7, "ts": 1716454224143021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079294, "dur": 14,
    "args": {
      "External id": 173772,
      "cbid": 211, "correlation": 173772
    }
  },
  {
    "ph": "s", "id": 173772, "pid": 76337, "tid": -914061504, "ts": 1716454224079294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079365, "dur": 1,
    "args": {
      "External id": 173783,
      "cbid": 251, "correlation": 173783
    }
  },
  {
    "ph": "f", "id": 173783, "pid": 76337, "tid": -914061504, "ts": 1716454224079365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224143187, "dur": 159,
    "args": {
      "External id": 173784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173784,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173784, "pid": 5, "tid": 7, "ts": 1716454224143187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079369, "dur": 13,
    "args": {
      "External id": 173784,
      "cbid": 211, "correlation": 173784
    }
  },
  {
    "ph": "s", "id": 173784, "pid": 76337, "tid": -914061504, "ts": 1716454224079369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079435, "dur": 1,
    "args": {
      "External id": 173795,
      "cbid": 251, "correlation": 173795
    }
  },
  {
    "ph": "f", "id": 173795, "pid": 76337, "tid": -914061504, "ts": 1716454224079435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224143348, "dur": 158,
    "args": {
      "External id": 173796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173796,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173796, "pid": 5, "tid": 7, "ts": 1716454224143348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079439, "dur": 11,
    "args": {
      "External id": 173796,
      "cbid": 211, "correlation": 173796
    }
  },
  {
    "ph": "s", "id": 173796, "pid": 76337, "tid": -914061504, "ts": 1716454224079439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224143508, "dur": 336,
    "args": {
      "External id": 173821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173821,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173821, "pid": 5, "tid": 7, "ts": 1716454224143508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079524, "dur": 13,
    "args": {
      "External id": 173821,
      "cbid": 211, "correlation": 173821
    }
  },
  {
    "ph": "s", "id": 173821, "pid": 76337, "tid": -914061504, "ts": 1716454224079524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079625, "dur": 1,
    "args": {
      "External id": 173839,
      "cbid": 251, "correlation": 173839
    }
  },
  {
    "ph": "f", "id": 173839, "pid": 76337, "tid": -914061504, "ts": 1716454224079625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224143845, "dur": 167,
    "args": {
      "External id": 173841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173841,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173841, "pid": 5, "tid": 7, "ts": 1716454224143845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079631, "dur": 13,
    "args": {
      "External id": 173841,
      "cbid": 211, "correlation": 173841
    }
  },
  {
    "ph": "s", "id": 173841, "pid": 76337, "tid": -914061504, "ts": 1716454224079631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224144014, "dur": 19,
    "args": {
      "External id": 173849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173849,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173849, "pid": 5, "tid": 7, "ts": 1716454224144014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079701, "dur": 12,
    "args": {
      "External id": 173849,
      "cbid": 211, "correlation": 173849
    }
  },
  {
    "ph": "s", "id": 173849, "pid": 76337, "tid": -914061504, "ts": 1716454224079701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224144034, "dur": 28,
    "args": {
      "External id": 173857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173857,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173857, "pid": 5, "tid": 7, "ts": 1716454224144034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079739, "dur": 8,
    "args": {
      "External id": 173857,
      "cbid": 211, "correlation": 173857
    }
  },
  {
    "ph": "s", "id": 173857, "pid": 76337, "tid": -914061504, "ts": 1716454224079739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224144062, "dur": 19,
    "args": {
      "External id": 173868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173868,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173868, "pid": 5, "tid": 7, "ts": 1716454224144062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079810, "dur": 12,
    "args": {
      "External id": 173868,
      "cbid": 211, "correlation": 173868
    }
  },
  {
    "ph": "s", "id": 173868, "pid": 76337, "tid": -914061504, "ts": 1716454224079810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224144082, "dur": 16,
    "args": {
      "External id": 173890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173890,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173890, "pid": 5, "tid": 7, "ts": 1716454224144082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079842, "dur": 7,
    "args": {
      "External id": 173890,
      "cbid": 211, "correlation": 173890
    }
  },
  {
    "ph": "s", "id": 173890, "pid": 76337, "tid": -914061504, "ts": 1716454224079842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079927, "dur": 1,
    "args": {
      "External id": 173901,
      "cbid": 251, "correlation": 173901
    }
  },
  {
    "ph": "f", "id": 173901, "pid": 76337, "tid": -914061504, "ts": 1716454224079927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224144100, "dur": 90,
    "args": {
      "External id": 173902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173902,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173902, "pid": 5, "tid": 7, "ts": 1716454224144100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224079932, "dur": 13,
    "args": {
      "External id": 173902,
      "cbid": 211, "correlation": 173902
    }
  },
  {
    "ph": "s", "id": 173902, "pid": 76337, "tid": -914061504, "ts": 1716454224079932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080009, "dur": 1,
    "args": {
      "External id": 173913,
      "cbid": 251, "correlation": 173913
    }
  },
  {
    "ph": "f", "id": 173913, "pid": 76337, "tid": -914061504, "ts": 1716454224080009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080012, "dur": 0,
    "args": {
      "External id": 173914,
      "cbid": 251, "correlation": 173914
    }
  },
  {
    "ph": "f", "id": 173914, "pid": 76337, "tid": -914061504, "ts": 1716454224080012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224144191, "dur": 13,
    "args": {
      "External id": 173915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173915,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173915, "pid": 5, "tid": 7, "ts": 1716454224144191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080014, "dur": 12,
    "args": {
      "External id": 173915,
      "cbid": 211, "correlation": 173915
    }
  },
  {
    "ph": "s", "id": 173915, "pid": 76337, "tid": -914061504, "ts": 1716454224080014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224144205, "dur": 6,
    "args": {
      "External id": 173917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173917,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173917, "pid": 5, "tid": 7, "ts": 1716454224144205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080028, "dur": 6,
    "args": {
      "External id": 173917,
      "cbid": 211, "correlation": 173917
    }
  },
  {
    "ph": "s", "id": 173917, "pid": 76337, "tid": -914061504, "ts": 1716454224080028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080087, "dur": 1,
    "args": {
      "External id": 173928,
      "cbid": 251, "correlation": 173928
    }
  },
  {
    "ph": "f", "id": 173928, "pid": 76337, "tid": -914061504, "ts": 1716454224080087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080091, "dur": 0,
    "args": {
      "External id": 173929,
      "cbid": 251, "correlation": 173929
    }
  },
  {
    "ph": "f", "id": 173929, "pid": 76337, "tid": -914061504, "ts": 1716454224080091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224144212, "dur": 9,
    "args": {
      "External id": 173930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173930,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173930, "pid": 5, "tid": 7, "ts": 1716454224144212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080092, "dur": 12,
    "args": {
      "External id": 173930,
      "cbid": 211, "correlation": 173930
    }
  },
  {
    "ph": "s", "id": 173930, "pid": 76337, "tid": -914061504, "ts": 1716454224080092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224144222, "dur": 4,
    "args": {
      "External id": 173932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173932,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173932, "pid": 5, "tid": 7, "ts": 1716454224144222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080105, "dur": 5,
    "args": {
      "External id": 173932,
      "cbid": 211, "correlation": 173932
    }
  },
  {
    "ph": "s", "id": 173932, "pid": 76337, "tid": -914061504, "ts": 1716454224080105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224144227, "dur": 55,
    "args": {
      "External id": 173957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173957,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 173957, "pid": 5, "tid": 7, "ts": 1716454224144227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080183, "dur": 13,
    "args": {
      "External id": 173957,
      "cbid": 211, "correlation": 173957
    }
  },
  {
    "ph": "s", "id": 173957, "pid": 76337, "tid": -914061504, "ts": 1716454224080183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080282, "dur": 1,
    "args": {
      "External id": 173975,
      "cbid": 251, "correlation": 173975
    }
  },
  {
    "ph": "f", "id": 173975, "pid": 76337, "tid": -914061504, "ts": 1716454224080282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224144284, "dur": 91,
    "args": {
      "External id": 173977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173977,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 173977, "pid": 5, "tid": 7, "ts": 1716454224144284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080288, "dur": 14,
    "args": {
      "External id": 173977,
      "cbid": 211, "correlation": 173977
    }
  },
  {
    "ph": "s", "id": 173977, "pid": 76337, "tid": -914061504, "ts": 1716454224080288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224144376, "dur": 10,
    "args": {
      "External id": 173985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173985,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173985, "pid": 5, "tid": 7, "ts": 1716454224144376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080360, "dur": 12,
    "args": {
      "External id": 173985,
      "cbid": 211, "correlation": 173985
    }
  },
  {
    "ph": "s", "id": 173985, "pid": 76337, "tid": -914061504, "ts": 1716454224080360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224144387, "dur": 21,
    "args": {
      "External id": 173993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 173993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 173993, "pid": 5, "tid": 7, "ts": 1716454224144387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080401, "dur": 9,
    "args": {
      "External id": 173993,
      "cbid": 211, "correlation": 173993
    }
  },
  {
    "ph": "s", "id": 173993, "pid": 76337, "tid": -914061504, "ts": 1716454224080401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224144410, "dur": 18,
    "args": {
      "External id": 174015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174015,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174015, "pid": 5, "tid": 7, "ts": 1716454224144410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080452, "dur": 10,
    "args": {
      "External id": 174015,
      "cbid": 211, "correlation": 174015
    }
  },
  {
    "ph": "s", "id": 174015, "pid": 76337, "tid": -914061504, "ts": 1716454224080452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080541, "dur": 1,
    "args": {
      "External id": 174031,
      "cbid": 251, "correlation": 174031
    }
  },
  {
    "ph": "f", "id": 174031, "pid": 76337, "tid": -914061504, "ts": 1716454224080541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080545, "dur": 0,
    "args": {
      "External id": 174033,
      "cbid": 251, "correlation": 174033
    }
  },
  {
    "ph": "f", "id": 174033, "pid": 76337, "tid": -914061504, "ts": 1716454224080545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224144429, "dur": 499,
    "args": {
      "External id": 174034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174034,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174034, "pid": 5, "tid": 7, "ts": 1716454224144429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080547, "dur": 12,
    "args": {
      "External id": 174034,
      "cbid": 211, "correlation": 174034
    }
  },
  {
    "ph": "s", "id": 174034, "pid": 76337, "tid": -914061504, "ts": 1716454224080547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224144929, "dur": 67,
    "args": {
      "External id": 174042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174042,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174042, "pid": 5, "tid": 7, "ts": 1716454224144929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080612, "dur": 12,
    "args": {
      "External id": 174042,
      "cbid": 211, "correlation": 174042
    }
  },
  {
    "ph": "s", "id": 174042, "pid": 76337, "tid": -914061504, "ts": 1716454224080612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224144997, "dur": 67,
    "args": {
      "External id": 174050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174050, "pid": 5, "tid": 7, "ts": 1716454224144997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080644, "dur": 8,
    "args": {
      "External id": 174050,
      "cbid": 211, "correlation": 174050
    }
  },
  {
    "ph": "s", "id": 174050, "pid": 76337, "tid": -914061504, "ts": 1716454224080644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080722, "dur": 1,
    "args": {
      "External id": 174066,
      "cbid": 251, "correlation": 174066
    }
  },
  {
    "ph": "f", "id": 174066, "pid": 76337, "tid": -914061504, "ts": 1716454224080722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224145066, "dur": 1,
    "args": {
      "External id": 174068,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 174068,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 174068, "pid": 5, "tid": 7, "ts": 1716454224145066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080727, "dur": 10,
    "args": {
      "External id": 174068,
      "cbid": 51, "correlation": 174068
    }
  },
  {
    "ph": "s", "id": 174068, "pid": 76337, "tid": -914061504, "ts": 1716454224080727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224145069, "dur": 271,
    "args": {
      "External id": 174069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174069,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 174069, "pid": 5, "tid": 7, "ts": 1716454224145069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080739, "dur": 11,
    "args": {
      "External id": 174069,
      "cbid": 211, "correlation": 174069
    }
  },
  {
    "ph": "s", "id": 174069, "pid": 76337, "tid": -914061504, "ts": 1716454224080739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224145342, "dur": 15,
    "args": {
      "External id": 174077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174077,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174077, "pid": 5, "tid": 7, "ts": 1716454224145342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080781, "dur": 11,
    "args": {
      "External id": 174077,
      "cbid": 211, "correlation": 174077
    }
  },
  {
    "ph": "s", "id": 174077, "pid": 76337, "tid": -914061504, "ts": 1716454224080781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224145358, "dur": 38,
    "args": {
      "External id": 174088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174088,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174088, "pid": 5, "tid": 7, "ts": 1716454224145358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080849, "dur": 12,
    "args": {
      "External id": 174088,
      "cbid": 211, "correlation": 174088
    }
  },
  {
    "ph": "s", "id": 174088, "pid": 76337, "tid": -914061504, "ts": 1716454224080849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080912, "dur": 0,
    "args": {
      "External id": 174100,
      "cbid": 317, "correlation": 174100
    }
  },
  {
    "ph": "f", "id": 174100, "pid": 76337, "tid": -914061504, "ts": 1716454224080912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080913, "dur": 0,
    "args": {
      "External id": 174101,
      "cbid": 203, "correlation": 174101
    }
  },
  {
    "ph": "f", "id": 174101, "pid": 76337, "tid": -914061504, "ts": 1716454224080913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080914, "dur": 0,
    "args": {
      "External id": 174102,
      "cbid": 205, "correlation": 174102
    }
  },
  {
    "ph": "f", "id": 174102, "pid": 76337, "tid": -914061504, "ts": 1716454224080914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224145397, "dur": 13,
    "args": {
      "External id": 174106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174106,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174106, "pid": 5, "tid": 7, "ts": 1716454224145397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080928, "dur": 12,
    "args": {
      "External id": 174106,
      "cbid": 211, "correlation": 174106
    }
  },
  {
    "ph": "s", "id": 174106, "pid": 76337, "tid": -914061504, "ts": 1716454224080928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224145411, "dur": 4,
    "args": {
      "External id": 174108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174108,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 174108, "pid": 5, "tid": 7, "ts": 1716454224145411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080946, "dur": 6,
    "args": {
      "External id": 174108,
      "cbid": 211, "correlation": 174108
    }
  },
  {
    "ph": "s", "id": 174108, "pid": 76337, "tid": -914061504, "ts": 1716454224080946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080954, "dur": 0,
    "args": {
      "External id": 174109,
      "cbid": 51, "correlation": 174109
    }
  },
  {
    "ph": "s", "id": 174109, "pid": 76337, "tid": -914061504, "ts": 1716454224080954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224145416, "dur": 98,
    "args": {
      "External id": 174110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174110,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 174110, "pid": 5, "tid": 7, "ts": 1716454224145416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224080955, "dur": 5,
    "args": {
      "External id": 174110,
      "cbid": 211, "correlation": 174110
    }
  },
  {
    "ph": "s", "id": 174110, "pid": 76337, "tid": -914061504, "ts": 1716454224080955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224145516, "dur": 16,
    "args": {
      "External id": 174115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174115,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174115, "pid": 5, "tid": 7, "ts": 1716454224145516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081613, "dur": 12,
    "args": {
      "External id": 174115,
      "cbid": 211, "correlation": 174115
    }
  },
  {
    "ph": "s", "id": 174115, "pid": 76337, "tid": -914061504, "ts": 1716454224081613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224145533, "dur": 12,
    "args": {
      "External id": 174123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174123,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174123, "pid": 5, "tid": 7, "ts": 1716454224145533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081712, "dur": 11,
    "args": {
      "External id": 174123,
      "cbid": 211, "correlation": 174123
    }
  },
  {
    "ph": "s", "id": 174123, "pid": 76337, "tid": -914061504, "ts": 1716454224081712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224145546, "dur": 57,
    "args": {
      "External id": 174134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174134,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174134, "pid": 5, "tid": 7, "ts": 1716454224145546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081863, "dur": 15,
    "args": {
      "External id": 174134,
      "cbid": 211, "correlation": 174134
    }
  },
  {
    "ph": "s", "id": 174134, "pid": 76337, "tid": -914061504, "ts": 1716454224081863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081931, "dur": 0,
    "args": {
      "External id": 174144,
      "cbid": 317, "correlation": 174144
    }
  },
  {
    "ph": "f", "id": 174144, "pid": 76337, "tid": -914061504, "ts": 1716454224081931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081932, "dur": 0,
    "args": {
      "External id": 174145,
      "cbid": 203, "correlation": 174145
    }
  },
  {
    "ph": "f", "id": 174145, "pid": 76337, "tid": -914061504, "ts": 1716454224081932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081933, "dur": 0,
    "args": {
      "External id": 174146,
      "cbid": 205, "correlation": 174146
    }
  },
  {
    "ph": "f", "id": 174146, "pid": 76337, "tid": -914061504, "ts": 1716454224081933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224145605, "dur": 41,
    "args": {
      "External id": 174150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174150,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174150, "pid": 5, "tid": 7, "ts": 1716454224145605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081950, "dur": 12,
    "args": {
      "External id": 174150,
      "cbid": 211, "correlation": 174150
    }
  },
  {
    "ph": "s", "id": 174150, "pid": 76337, "tid": -914061504, "ts": 1716454224081950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224145648, "dur": 163,
    "args": {
      "External id": 174152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174152,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174152, "pid": 5, "tid": 7, "ts": 1716454224145648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081964, "dur": 6,
    "args": {
      "External id": 174152,
      "cbid": 211, "correlation": 174152
    }
  },
  {
    "ph": "s", "id": 174152, "pid": 76337, "tid": -914061504, "ts": 1716454224081964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224145811, "dur": 1968,
    "args": {
      "External id": 174154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174154,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174154, "pid": 5, "tid": 7, "ts": 1716454224145811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224081987, "dur": 9,
    "args": {
      "External id": 174154,
      "cbid": 211, "correlation": 174154
    }
  },
  {
    "ph": "s", "id": 174154, "pid": 76337, "tid": -914061504, "ts": 1716454224081987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224147781, "dur": 39,
    "args": {
      "External id": 174156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174156, "pid": 5, "tid": 7, "ts": 1716454224147781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082001, "dur": 5,
    "args": {
      "External id": 174156,
      "cbid": 211, "correlation": 174156
    }
  },
  {
    "ph": "s", "id": 174156, "pid": 76337, "tid": -914061504, "ts": 1716454224082001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224147821, "dur": 59,
    "args": {
      "External id": 174162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174162, "pid": 5, "tid": 7, "ts": 1716454224147821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082032, "dur": 9,
    "args": {
      "External id": 174162,
      "cbid": 211, "correlation": 174162
    }
  },
  {
    "ph": "s", "id": 174162, "pid": 76337, "tid": -914061504, "ts": 1716454224082032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224147881, "dur": 85,
    "args": {
      "External id": 174171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174171,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174171, "pid": 5, "tid": 7, "ts": 1716454224147881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082191, "dur": 15,
    "args": {
      "External id": 174171,
      "cbid": 211, "correlation": 174171
    }
  },
  {
    "ph": "s", "id": 174171, "pid": 76337, "tid": -914061504, "ts": 1716454224082191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224147968, "dur": 74,
    "args": {
      "External id": 174191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174191,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 174191, "pid": 5, "tid": 7, "ts": 1716454224147968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082289, "dur": 13,
    "args": {
      "External id": 174191,
      "cbid": 211, "correlation": 174191
    }
  },
  {
    "ph": "s", "id": 174191, "pid": 76337, "tid": -914061504, "ts": 1716454224082289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224148043, "dur": 5,
    "args": {
      "External id": 174203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174203,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 174203, "pid": 5, "tid": 7, "ts": 1716454224148043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082312, "dur": 6,
    "args": {
      "External id": 174203,
      "cbid": 211, "correlation": 174203
    }
  },
  {
    "ph": "s", "id": 174203, "pid": 76337, "tid": -914061504, "ts": 1716454224082312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224148049, "dur": 82,
    "args": {
      "External id": 174206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174206,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174206, "pid": 5, "tid": 7, "ts": 1716454224148049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082332, "dur": 7,
    "args": {
      "External id": 174206,
      "cbid": 211, "correlation": 174206
    }
  },
  {
    "ph": "s", "id": 174206, "pid": 76337, "tid": -914061504, "ts": 1716454224082332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224148133, "dur": 54,
    "args": {
      "External id": 174215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174215,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174215, "pid": 5, "tid": 7, "ts": 1716454224148133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082378, "dur": 10,
    "args": {
      "External id": 174215,
      "cbid": 211, "correlation": 174215
    }
  },
  {
    "ph": "s", "id": 174215, "pid": 76337, "tid": -914061504, "ts": 1716454224082378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082434, "dur": 0,
    "args": {
      "External id": 174225,
      "cbid": 317, "correlation": 174225
    }
  },
  {
    "ph": "f", "id": 174225, "pid": 76337, "tid": -914061504, "ts": 1716454224082434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082435, "dur": 0,
    "args": {
      "External id": 174226,
      "cbid": 203, "correlation": 174226
    }
  },
  {
    "ph": "f", "id": 174226, "pid": 76337, "tid": -914061504, "ts": 1716454224082435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082436, "dur": 0,
    "args": {
      "External id": 174227,
      "cbid": 205, "correlation": 174227
    }
  },
  {
    "ph": "f", "id": 174227, "pid": 76337, "tid": -914061504, "ts": 1716454224082436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224148188, "dur": 58,
    "args": {
      "External id": 174231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174231,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174231, "pid": 5, "tid": 7, "ts": 1716454224148188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082452, "dur": 12,
    "args": {
      "External id": 174231,
      "cbid": 211, "correlation": 174231
    }
  },
  {
    "ph": "s", "id": 174231, "pid": 76337, "tid": -914061504, "ts": 1716454224082452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224148247, "dur": 122,
    "args": {
      "External id": 174233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174233,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174233, "pid": 5, "tid": 7, "ts": 1716454224148247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082466, "dur": 5,
    "args": {
      "External id": 174233,
      "cbid": 211, "correlation": 174233
    }
  },
  {
    "ph": "s", "id": 174233, "pid": 76337, "tid": -914061504, "ts": 1716454224082466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224148370, "dur": 1889,
    "args": {
      "External id": 174235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174235,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174235, "pid": 5, "tid": 7, "ts": 1716454224148370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082478, "dur": 6,
    "args": {
      "External id": 174235,
      "cbid": 211, "correlation": 174235
    }
  },
  {
    "ph": "s", "id": 174235, "pid": 76337, "tid": -914061504, "ts": 1716454224082478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224150261, "dur": 21,
    "args": {
      "External id": 174237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174237, "pid": 5, "tid": 7, "ts": 1716454224150261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082488, "dur": 5,
    "args": {
      "External id": 174237,
      "cbid": 211, "correlation": 174237
    }
  },
  {
    "ph": "s", "id": 174237, "pid": 76337, "tid": -914061504, "ts": 1716454224082488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224150283, "dur": 33,
    "args": {
      "External id": 174243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174243,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174243, "pid": 5, "tid": 7, "ts": 1716454224150283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082516, "dur": 9,
    "args": {
      "External id": 174243,
      "cbid": 211, "correlation": 174243
    }
  },
  {
    "ph": "s", "id": 174243, "pid": 76337, "tid": -914061504, "ts": 1716454224082516,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224150318, "dur": 4,
    "args": {
      "External id": 174251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174251,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 174251, "pid": 5, "tid": 7, "ts": 1716454224150318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082560, "dur": 10,
    "args": {
      "External id": 174251,
      "cbid": 211, "correlation": 174251
    }
  },
  {
    "ph": "s", "id": 174251, "pid": 76337, "tid": -914061504, "ts": 1716454224082560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082647, "dur": 2,
    "args": {
      "External id": 174267,
      "cbid": 251, "correlation": 174267
    }
  },
  {
    "ph": "f", "id": 174267, "pid": 76337, "tid": -914061504, "ts": 1716454224082647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082654, "dur": 0,
    "args": {
      "External id": 174269,
      "cbid": 251, "correlation": 174269
    }
  },
  {
    "ph": "f", "id": 174269, "pid": 76337, "tid": -914061504, "ts": 1716454224082654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224150323, "dur": 12,
    "args": {
      "External id": 174270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174270,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 174270, "pid": 5, "tid": 7, "ts": 1716454224150323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082656, "dur": 13,
    "args": {
      "External id": 174270,
      "cbid": 211, "correlation": 174270
    }
  },
  {
    "ph": "s", "id": 174270, "pid": 76337, "tid": -914061504, "ts": 1716454224082656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224150337, "dur": 5,
    "args": {
      "External id": 174272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174272,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 174272, "pid": 5, "tid": 7, "ts": 1716454224150337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082671, "dur": 6,
    "args": {
      "External id": 174272,
      "cbid": 211, "correlation": 174272
    }
  },
  {
    "ph": "s", "id": 174272, "pid": 76337, "tid": -914061504, "ts": 1716454224082671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224150343, "dur": 29,
    "args": {
      "External id": 174282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174282, "pid": 5, "tid": 7, "ts": 1716454224150343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082743, "dur": 14,
    "args": {
      "External id": 174282,
      "cbid": 211, "correlation": 174282
    }
  },
  {
    "ph": "s", "id": 174282, "pid": 76337, "tid": -914061504, "ts": 1716454224082743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224150373, "dur": 30,
    "args": {
      "External id": 174302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174302,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 174302, "pid": 5, "tid": 7, "ts": 1716454224150373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082814, "dur": 12,
    "args": {
      "External id": 174302,
      "cbid": 211, "correlation": 174302
    }
  },
  {
    "ph": "s", "id": 174302, "pid": 76337, "tid": -914061504, "ts": 1716454224082814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224150404, "dur": 5,
    "args": {
      "External id": 174314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174314,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 174314, "pid": 5, "tid": 7, "ts": 1716454224150404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082836, "dur": 6,
    "args": {
      "External id": 174314,
      "cbid": 211, "correlation": 174314
    }
  },
  {
    "ph": "s", "id": 174314, "pid": 76337, "tid": -914061504, "ts": 1716454224082836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224150410, "dur": 30,
    "args": {
      "External id": 174317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174317,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174317, "pid": 5, "tid": 7, "ts": 1716454224150410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082853, "dur": 6,
    "args": {
      "External id": 174317,
      "cbid": 211, "correlation": 174317
    }
  },
  {
    "ph": "s", "id": 174317, "pid": 76337, "tid": -914061504, "ts": 1716454224082853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224150442, "dur": 20,
    "args": {
      "External id": 174326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174326,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174326, "pid": 5, "tid": 7, "ts": 1716454224150442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082894, "dur": 9,
    "args": {
      "External id": 174326,
      "cbid": 211, "correlation": 174326
    }
  },
  {
    "ph": "s", "id": 174326, "pid": 76337, "tid": -914061504, "ts": 1716454224082894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082964, "dur": 0,
    "args": {
      "External id": 174336,
      "cbid": 317, "correlation": 174336
    }
  },
  {
    "ph": "f", "id": 174336, "pid": 76337, "tid": -914061504, "ts": 1716454224082964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082965, "dur": 0,
    "args": {
      "External id": 174337,
      "cbid": 203, "correlation": 174337
    }
  },
  {
    "ph": "f", "id": 174337, "pid": 76337, "tid": -914061504, "ts": 1716454224082965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082965, "dur": 0,
    "args": {
      "External id": 174338,
      "cbid": 205, "correlation": 174338
    }
  },
  {
    "ph": "f", "id": 174338, "pid": 76337, "tid": -914061504, "ts": 1716454224082965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224150464, "dur": 24,
    "args": {
      "External id": 174342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174342,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174342, "pid": 5, "tid": 7, "ts": 1716454224150464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224082990, "dur": 13,
    "args": {
      "External id": 174342,
      "cbid": 211, "correlation": 174342
    }
  },
  {
    "ph": "s", "id": 174342, "pid": 76337, "tid": -914061504, "ts": 1716454224082990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224150488, "dur": 44,
    "args": {
      "External id": 174344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174344, "pid": 5, "tid": 7, "ts": 1716454224150488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083005, "dur": 5,
    "args": {
      "External id": 174344,
      "cbid": 211, "correlation": 174344
    }
  },
  {
    "ph": "s", "id": 174344, "pid": 76337, "tid": -914061504, "ts": 1716454224083005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224150534, "dur": 648,
    "args": {
      "External id": 174346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174346,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174346, "pid": 5, "tid": 7, "ts": 1716454224150534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083017, "dur": 7,
    "args": {
      "External id": 174346,
      "cbid": 211, "correlation": 174346
    }
  },
  {
    "ph": "s", "id": 174346, "pid": 76337, "tid": -914061504, "ts": 1716454224083017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224151183, "dur": 23,
    "args": {
      "External id": 174348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174348, "pid": 5, "tid": 7, "ts": 1716454224151183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083028, "dur": 5,
    "args": {
      "External id": 174348,
      "cbid": 211, "correlation": 174348
    }
  },
  {
    "ph": "s", "id": 174348, "pid": 76337, "tid": -914061504, "ts": 1716454224083028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224151207, "dur": 32,
    "args": {
      "External id": 174354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174354, "pid": 5, "tid": 7, "ts": 1716454224151207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083056, "dur": 9,
    "args": {
      "External id": 174354,
      "cbid": 211, "correlation": 174354
    }
  },
  {
    "ph": "s", "id": 174354, "pid": 76337, "tid": -914061504, "ts": 1716454224083056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083115, "dur": 0,
    "args": {
      "External id": 174364,
      "cbid": 317, "correlation": 174364
    }
  },
  {
    "ph": "f", "id": 174364, "pid": 76337, "tid": -914061504, "ts": 1716454224083115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083115, "dur": 0,
    "args": {
      "External id": 174365,
      "cbid": 203, "correlation": 174365
    }
  },
  {
    "ph": "f", "id": 174365, "pid": 76337, "tid": -914061504, "ts": 1716454224083115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083116, "dur": 0,
    "args": {
      "External id": 174366,
      "cbid": 205, "correlation": 174366
    }
  },
  {
    "ph": "f", "id": 174366, "pid": 76337, "tid": -914061504, "ts": 1716454224083116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224151240, "dur": 57,
    "args": {
      "External id": 174370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174370,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174370, "pid": 5, "tid": 7, "ts": 1716454224151240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083129, "dur": 11,
    "args": {
      "External id": 174370,
      "cbid": 211, "correlation": 174370
    }
  },
  {
    "ph": "s", "id": 174370, "pid": 76337, "tid": -914061504, "ts": 1716454224083129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224151298, "dur": 270,
    "args": {
      "External id": 174372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174372,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174372, "pid": 5, "tid": 7, "ts": 1716454224151298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083147, "dur": 8,
    "args": {
      "External id": 174372,
      "cbid": 211, "correlation": 174372
    }
  },
  {
    "ph": "s", "id": 174372, "pid": 76337, "tid": -914061504, "ts": 1716454224083147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224151569, "dur": 23,
    "args": {
      "External id": 174374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174374,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174374, "pid": 5, "tid": 7, "ts": 1716454224151569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083159, "dur": 5,
    "args": {
      "External id": 174374,
      "cbid": 211, "correlation": 174374
    }
  },
  {
    "ph": "s", "id": 174374, "pid": 76337, "tid": -914061504, "ts": 1716454224083159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224151593, "dur": 32,
    "args": {
      "External id": 174380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174380,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174380, "pid": 5, "tid": 7, "ts": 1716454224151593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083185, "dur": 8,
    "args": {
      "External id": 174380,
      "cbid": 211, "correlation": 174380
    }
  },
  {
    "ph": "s", "id": 174380, "pid": 76337, "tid": -914061504, "ts": 1716454224083185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224151627, "dur": 27,
    "args": {
      "External id": 174388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174388,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174388, "pid": 5, "tid": 7, "ts": 1716454224151627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083213, "dur": 8,
    "args": {
      "External id": 174388,
      "cbid": 211, "correlation": 174388
    }
  },
  {
    "ph": "s", "id": 174388, "pid": 76337, "tid": -914061504, "ts": 1716454224083213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224151655, "dur": 20,
    "args": {
      "External id": 174396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174396,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174396, "pid": 5, "tid": 7, "ts": 1716454224151655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083243, "dur": 9,
    "args": {
      "External id": 174396,
      "cbid": 211, "correlation": 174396
    }
  },
  {
    "ph": "s", "id": 174396, "pid": 76337, "tid": -914061504, "ts": 1716454224083243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224151675, "dur": 31,
    "args": {
      "External id": 174416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174416,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 174416, "pid": 5, "tid": 7, "ts": 1716454224151675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083332, "dur": 13,
    "args": {
      "External id": 174416,
      "cbid": 211, "correlation": 174416
    }
  },
  {
    "ph": "s", "id": 174416, "pid": 76337, "tid": -914061504, "ts": 1716454224083332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224151708, "dur": 4,
    "args": {
      "External id": 174428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174428,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 174428, "pid": 5, "tid": 7, "ts": 1716454224151708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083355, "dur": 6,
    "args": {
      "External id": 174428,
      "cbid": 211, "correlation": 174428
    }
  },
  {
    "ph": "s", "id": 174428, "pid": 76337, "tid": -914061504, "ts": 1716454224083355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224151714, "dur": 31,
    "args": {
      "External id": 174431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174431,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174431, "pid": 5, "tid": 7, "ts": 1716454224151714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083373, "dur": 7,
    "args": {
      "External id": 174431,
      "cbid": 211, "correlation": 174431
    }
  },
  {
    "ph": "s", "id": 174431, "pid": 76337, "tid": -914061504, "ts": 1716454224083373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083434, "dur": 0,
    "args": {
      "External id": 174442,
      "cbid": 317, "correlation": 174442
    }
  },
  {
    "ph": "f", "id": 174442, "pid": 76337, "tid": -914061504, "ts": 1716454224083434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083435, "dur": 0,
    "args": {
      "External id": 174443,
      "cbid": 203, "correlation": 174443
    }
  },
  {
    "ph": "f", "id": 174443, "pid": 76337, "tid": -914061504, "ts": 1716454224083435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083436, "dur": 0,
    "args": {
      "External id": 174444,
      "cbid": 205, "correlation": 174444
    }
  },
  {
    "ph": "f", "id": 174444, "pid": 76337, "tid": -914061504, "ts": 1716454224083436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224151745, "dur": 22,
    "args": {
      "External id": 174448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174448,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174448, "pid": 5, "tid": 7, "ts": 1716454224151745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083450, "dur": 12,
    "args": {
      "External id": 174448,
      "cbid": 211, "correlation": 174448
    }
  },
  {
    "ph": "s", "id": 174448, "pid": 76337, "tid": -914061504, "ts": 1716454224083450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224151769, "dur": 105,
    "args": {
      "External id": 174450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174450,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174450, "pid": 5, "tid": 7, "ts": 1716454224151769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083468, "dur": 7,
    "args": {
      "External id": 174450,
      "cbid": 211, "correlation": 174450
    }
  },
  {
    "ph": "s", "id": 174450, "pid": 76337, "tid": -914061504, "ts": 1716454224083468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224151875, "dur": 22,
    "args": {
      "External id": 174452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174452, "pid": 5, "tid": 7, "ts": 1716454224151875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083479, "dur": 5,
    "args": {
      "External id": 174452,
      "cbid": 211, "correlation": 174452
    }
  },
  {
    "ph": "s", "id": 174452, "pid": 76337, "tid": -914061504, "ts": 1716454224083479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224151899, "dur": 33,
    "args": {
      "External id": 174458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174458, "pid": 5, "tid": 7, "ts": 1716454224151899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083506, "dur": 8,
    "args": {
      "External id": 174458,
      "cbid": 211, "correlation": 174458
    }
  },
  {
    "ph": "s", "id": 174458, "pid": 76337, "tid": -914061504, "ts": 1716454224083506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224151933, "dur": 163,
    "args": {
      "External id": 174467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174467,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174467, "pid": 5, "tid": 7, "ts": 1716454224151933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083597, "dur": 15,
    "args": {
      "External id": 174467,
      "cbid": 211, "correlation": 174467
    }
  },
  {
    "ph": "s", "id": 174467, "pid": 76337, "tid": -914061504, "ts": 1716454224083597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224152098, "dur": 65,
    "args": {
      "External id": 174489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174489,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174489, "pid": 5, "tid": 7, "ts": 1716454224152098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083672, "dur": 13,
    "args": {
      "External id": 174489,
      "cbid": 211, "correlation": 174489
    }
  },
  {
    "ph": "s", "id": 174489, "pid": 76337, "tid": -914061504, "ts": 1716454224083672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083775, "dur": 2,
    "args": {
      "External id": 174500,
      "cbid": 251, "correlation": 174500
    }
  },
  {
    "ph": "f", "id": 174500, "pid": 76337, "tid": -914061504, "ts": 1716454224083775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224152164, "dur": 155,
    "args": {
      "External id": 174501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174501,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174501, "pid": 5, "tid": 7, "ts": 1716454224152164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083781, "dur": 13,
    "args": {
      "External id": 174501,
      "cbid": 211, "correlation": 174501
    }
  },
  {
    "ph": "s", "id": 174501, "pid": 76337, "tid": -914061504, "ts": 1716454224083781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083854, "dur": 1,
    "args": {
      "External id": 174512,
      "cbid": 251, "correlation": 174512
    }
  },
  {
    "ph": "f", "id": 174512, "pid": 76337, "tid": -914061504, "ts": 1716454224083854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224152321, "dur": 147,
    "args": {
      "External id": 174513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174513,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174513, "pid": 5, "tid": 7, "ts": 1716454224152321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083858, "dur": 11,
    "args": {
      "External id": 174513,
      "cbid": 211, "correlation": 174513
    }
  },
  {
    "ph": "s", "id": 174513, "pid": 76337, "tid": -914061504, "ts": 1716454224083858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083927, "dur": 1,
    "args": {
      "External id": 174524,
      "cbid": 251, "correlation": 174524
    }
  },
  {
    "ph": "f", "id": 174524, "pid": 76337, "tid": -914061504, "ts": 1716454224083927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224152470, "dur": 145,
    "args": {
      "External id": 174525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174525,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174525, "pid": 5, "tid": 7, "ts": 1716454224152470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224083931, "dur": 11,
    "args": {
      "External id": 174525,
      "cbid": 211, "correlation": 174525
    }
  },
  {
    "ph": "s", "id": 174525, "pid": 76337, "tid": -914061504, "ts": 1716454224083931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224152616, "dur": 1949,
    "args": {
      "External id": 174546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174546,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 174546, "pid": 5, "tid": 7, "ts": 1716454224152616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084031, "dur": 14,
    "args": {
      "External id": 174546,
      "cbid": 211, "correlation": 174546
    }
  },
  {
    "ph": "s", "id": 174546, "pid": 76337, "tid": -914061504, "ts": 1716454224084031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084141, "dur": 2,
    "args": {
      "External id": 174564,
      "cbid": 251, "correlation": 174564
    }
  },
  {
    "ph": "f", "id": 174564, "pid": 76337, "tid": -914061504, "ts": 1716454224084141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224154566, "dur": 148,
    "args": {
      "External id": 174566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174566,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 174566, "pid": 5, "tid": 7, "ts": 1716454224154566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084147, "dur": 14,
    "args": {
      "External id": 174566,
      "cbid": 211, "correlation": 174566
    }
  },
  {
    "ph": "s", "id": 174566, "pid": 76337, "tid": -914061504, "ts": 1716454224084147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224154716, "dur": 35,
    "args": {
      "External id": 174574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174574,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174574, "pid": 5, "tid": 7, "ts": 1716454224154716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084221, "dur": 13,
    "args": {
      "External id": 174574,
      "cbid": 211, "correlation": 174574
    }
  },
  {
    "ph": "s", "id": 174574, "pid": 76337, "tid": -914061504, "ts": 1716454224084221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224154753, "dur": 51,
    "args": {
      "External id": 174582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174582,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174582, "pid": 5, "tid": 7, "ts": 1716454224154753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084261, "dur": 8,
    "args": {
      "External id": 174582,
      "cbid": 211, "correlation": 174582
    }
  },
  {
    "ph": "s", "id": 174582, "pid": 76337, "tid": -914061504, "ts": 1716454224084261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224154805, "dur": 30,
    "args": {
      "External id": 174593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174593,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174593, "pid": 5, "tid": 7, "ts": 1716454224154805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084341, "dur": 14,
    "args": {
      "External id": 174593,
      "cbid": 211, "correlation": 174593
    }
  },
  {
    "ph": "s", "id": 174593, "pid": 76337, "tid": -914061504, "ts": 1716454224084341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224154836, "dur": 34,
    "args": {
      "External id": 174615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174615,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174615, "pid": 5, "tid": 7, "ts": 1716454224154836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084373, "dur": 8,
    "args": {
      "External id": 174615,
      "cbid": 211, "correlation": 174615
    }
  },
  {
    "ph": "s", "id": 174615, "pid": 76337, "tid": -914061504, "ts": 1716454224084373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084462, "dur": 1,
    "args": {
      "External id": 174626,
      "cbid": 251, "correlation": 174626
    }
  },
  {
    "ph": "f", "id": 174626, "pid": 76337, "tid": -914061504, "ts": 1716454224084462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224154872, "dur": 92,
    "args": {
      "External id": 174627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174627,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174627, "pid": 5, "tid": 7, "ts": 1716454224154872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084467, "dur": 14,
    "args": {
      "External id": 174627,
      "cbid": 211, "correlation": 174627
    }
  },
  {
    "ph": "s", "id": 174627, "pid": 76337, "tid": -914061504, "ts": 1716454224084467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084542, "dur": 1,
    "args": {
      "External id": 174638,
      "cbid": 251, "correlation": 174638
    }
  },
  {
    "ph": "f", "id": 174638, "pid": 76337, "tid": -914061504, "ts": 1716454224084542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084546, "dur": 0,
    "args": {
      "External id": 174639,
      "cbid": 251, "correlation": 174639
    }
  },
  {
    "ph": "f", "id": 174639, "pid": 76337, "tid": -914061504, "ts": 1716454224084546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224154966, "dur": 10,
    "args": {
      "External id": 174640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174640,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 174640, "pid": 5, "tid": 7, "ts": 1716454224154966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084548, "dur": 12,
    "args": {
      "External id": 174640,
      "cbid": 211, "correlation": 174640
    }
  },
  {
    "ph": "s", "id": 174640, "pid": 76337, "tid": -914061504, "ts": 1716454224084548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224154977, "dur": 5,
    "args": {
      "External id": 174642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174642,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 174642, "pid": 5, "tid": 7, "ts": 1716454224154977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084561, "dur": 6,
    "args": {
      "External id": 174642,
      "cbid": 211, "correlation": 174642
    }
  },
  {
    "ph": "s", "id": 174642, "pid": 76337, "tid": -914061504, "ts": 1716454224084561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084621, "dur": 1,
    "args": {
      "External id": 174653,
      "cbid": 251, "correlation": 174653
    }
  },
  {
    "ph": "f", "id": 174653, "pid": 76337, "tid": -914061504, "ts": 1716454224084621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084625, "dur": 0,
    "args": {
      "External id": 174654,
      "cbid": 251, "correlation": 174654
    }
  },
  {
    "ph": "f", "id": 174654, "pid": 76337, "tid": -914061504, "ts": 1716454224084625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224154983, "dur": 7,
    "args": {
      "External id": 174655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174655,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 174655, "pid": 5, "tid": 7, "ts": 1716454224154983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084627, "dur": 12,
    "args": {
      "External id": 174655,
      "cbid": 211, "correlation": 174655
    }
  },
  {
    "ph": "s", "id": 174655, "pid": 76337, "tid": -914061504, "ts": 1716454224084627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224154992, "dur": 3,
    "args": {
      "External id": 174657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174657,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 174657, "pid": 5, "tid": 7, "ts": 1716454224154992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084640, "dur": 5,
    "args": {
      "External id": 174657,
      "cbid": 211, "correlation": 174657
    }
  },
  {
    "ph": "s", "id": 174657, "pid": 76337, "tid": -914061504, "ts": 1716454224084640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224154996, "dur": 91,
    "args": {
      "External id": 174678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174678,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 174678, "pid": 5, "tid": 7, "ts": 1716454224154996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084714, "dur": 13,
    "args": {
      "External id": 174678,
      "cbid": 211, "correlation": 174678
    }
  },
  {
    "ph": "s", "id": 174678, "pid": 76337, "tid": -914061504, "ts": 1716454224084714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084816, "dur": 2,
    "args": {
      "External id": 174696,
      "cbid": 251, "correlation": 174696
    }
  },
  {
    "ph": "f", "id": 174696, "pid": 76337, "tid": -914061504, "ts": 1716454224084816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224155089, "dur": 100,
    "args": {
      "External id": 174698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174698,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174698, "pid": 5, "tid": 7, "ts": 1716454224155089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084822, "dur": 14,
    "args": {
      "External id": 174698,
      "cbid": 211, "correlation": 174698
    }
  },
  {
    "ph": "s", "id": 174698, "pid": 76337, "tid": -914061504, "ts": 1716454224084822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224155190, "dur": 20,
    "args": {
      "External id": 174706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174706,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174706, "pid": 5, "tid": 7, "ts": 1716454224155190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084895, "dur": 12,
    "args": {
      "External id": 174706,
      "cbid": 211, "correlation": 174706
    }
  },
  {
    "ph": "s", "id": 174706, "pid": 76337, "tid": -914061504, "ts": 1716454224084895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224155211, "dur": 38,
    "args": {
      "External id": 174714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174714,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174714, "pid": 5, "tid": 7, "ts": 1716454224155211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084938, "dur": 9,
    "args": {
      "External id": 174714,
      "cbid": 211, "correlation": 174714
    }
  },
  {
    "ph": "s", "id": 174714, "pid": 76337, "tid": -914061504, "ts": 1716454224084938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224155250, "dur": 35,
    "args": {
      "External id": 174736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174736,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174736, "pid": 5, "tid": 7, "ts": 1716454224155250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224084998, "dur": 11,
    "args": {
      "External id": 174736,
      "cbid": 211, "correlation": 174736
    }
  },
  {
    "ph": "s", "id": 174736, "pid": 76337, "tid": -914061504, "ts": 1716454224084998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085096, "dur": 1,
    "args": {
      "External id": 174752,
      "cbid": 251, "correlation": 174752
    }
  },
  {
    "ph": "f", "id": 174752, "pid": 76337, "tid": -914061504, "ts": 1716454224085096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085101, "dur": 0,
    "args": {
      "External id": 174754,
      "cbid": 251, "correlation": 174754
    }
  },
  {
    "ph": "f", "id": 174754, "pid": 76337, "tid": -914061504, "ts": 1716454224085101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224155286, "dur": 541,
    "args": {
      "External id": 174755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174755,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 174755, "pid": 5, "tid": 7, "ts": 1716454224155286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085106, "dur": 12,
    "args": {
      "External id": 174755,
      "cbid": 211, "correlation": 174755
    }
  },
  {
    "ph": "s", "id": 174755, "pid": 76337, "tid": -914061504, "ts": 1716454224085106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224155829, "dur": 125,
    "args": {
      "External id": 174763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174763,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174763, "pid": 5, "tid": 7, "ts": 1716454224155829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085175, "dur": 14,
    "args": {
      "External id": 174763,
      "cbid": 211, "correlation": 174763
    }
  },
  {
    "ph": "s", "id": 174763, "pid": 76337, "tid": -914061504, "ts": 1716454224085175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224155956, "dur": 129,
    "args": {
      "External id": 174771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174771, "pid": 5, "tid": 7, "ts": 1716454224155956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085210, "dur": 8,
    "args": {
      "External id": 174771,
      "cbid": 211, "correlation": 174771
    }
  },
  {
    "ph": "s", "id": 174771, "pid": 76337, "tid": -914061504, "ts": 1716454224085210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085289, "dur": 1,
    "args": {
      "External id": 174787,
      "cbid": 251, "correlation": 174787
    }
  },
  {
    "ph": "f", "id": 174787, "pid": 76337, "tid": -914061504, "ts": 1716454224085289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224156085, "dur": 307,
    "args": {
      "External id": 174789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174789,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174789, "pid": 5, "tid": 7, "ts": 1716454224156085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085294, "dur": 12,
    "args": {
      "External id": 174789,
      "cbid": 211, "correlation": 174789
    }
  },
  {
    "ph": "s", "id": 174789, "pid": 76337, "tid": -914061504, "ts": 1716454224085294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224156394, "dur": 27,
    "args": {
      "External id": 174797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174797,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174797, "pid": 5, "tid": 7, "ts": 1716454224156394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085337, "dur": 9,
    "args": {
      "External id": 174797,
      "cbid": 211, "correlation": 174797
    }
  },
  {
    "ph": "s", "id": 174797, "pid": 76337, "tid": -914061504, "ts": 1716454224085337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224156422, "dur": 82,
    "args": {
      "External id": 174808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174808,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174808, "pid": 5, "tid": 7, "ts": 1716454224156422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085406, "dur": 12,
    "args": {
      "External id": 174808,
      "cbid": 211, "correlation": 174808
    }
  },
  {
    "ph": "s", "id": 174808, "pid": 76337, "tid": -914061504, "ts": 1716454224085406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085473, "dur": 0,
    "args": {
      "External id": 174820,
      "cbid": 317, "correlation": 174820
    }
  },
  {
    "ph": "f", "id": 174820, "pid": 76337, "tid": -914061504, "ts": 1716454224085473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085473, "dur": 0,
    "args": {
      "External id": 174821,
      "cbid": 203, "correlation": 174821
    }
  },
  {
    "ph": "f", "id": 174821, "pid": 76337, "tid": -914061504, "ts": 1716454224085473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085474, "dur": 0,
    "args": {
      "External id": 174822,
      "cbid": 205, "correlation": 174822
    }
  },
  {
    "ph": "f", "id": 174822, "pid": 76337, "tid": -914061504, "ts": 1716454224085474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224156505, "dur": 23,
    "args": {
      "External id": 174826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174826,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174826, "pid": 5, "tid": 7, "ts": 1716454224156505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085490, "dur": 12,
    "args": {
      "External id": 174826,
      "cbid": 211, "correlation": 174826
    }
  },
  {
    "ph": "s", "id": 174826, "pid": 76337, "tid": -914061504, "ts": 1716454224085490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224156529, "dur": 121,
    "args": {
      "External id": 174828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174828,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174828, "pid": 5, "tid": 7, "ts": 1716454224156529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085508, "dur": 6,
    "args": {
      "External id": 174828,
      "cbid": 211, "correlation": 174828
    }
  },
  {
    "ph": "s", "id": 174828, "pid": 76337, "tid": -914061504, "ts": 1716454224085508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224156652, "dur": 23,
    "args": {
      "External id": 174830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174830,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174830, "pid": 5, "tid": 7, "ts": 1716454224156652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085518, "dur": 5,
    "args": {
      "External id": 174830,
      "cbid": 211, "correlation": 174830
    }
  },
  {
    "ph": "s", "id": 174830, "pid": 76337, "tid": -914061504, "ts": 1716454224085518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224156676, "dur": 33,
    "args": {
      "External id": 174836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174836,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174836, "pid": 5, "tid": 7, "ts": 1716454224156676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085547, "dur": 9,
    "args": {
      "External id": 174836,
      "cbid": 211, "correlation": 174836
    }
  },
  {
    "ph": "s", "id": 174836, "pid": 76337, "tid": -914061504, "ts": 1716454224085547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224156711, "dur": 26,
    "args": {
      "External id": 174844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174844,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174844, "pid": 5, "tid": 7, "ts": 1716454224156711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085579, "dur": 8,
    "args": {
      "External id": 174844,
      "cbid": 211, "correlation": 174844
    }
  },
  {
    "ph": "s", "id": 174844, "pid": 76337, "tid": -914061504, "ts": 1716454224085579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224156738, "dur": 54,
    "args": {
      "External id": 174853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174853,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174853, "pid": 5, "tid": 7, "ts": 1716454224156738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085616, "dur": 10,
    "args": {
      "External id": 174853,
      "cbid": 211, "correlation": 174853
    }
  },
  {
    "ph": "s", "id": 174853, "pid": 76337, "tid": -914061504, "ts": 1716454224085616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224156793, "dur": 53,
    "args": {
      "External id": 174873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174873,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 174873, "pid": 5, "tid": 7, "ts": 1716454224156793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085692, "dur": 11,
    "args": {
      "External id": 174873,
      "cbid": 211, "correlation": 174873
    }
  },
  {
    "ph": "s", "id": 174873, "pid": 76337, "tid": -914061504, "ts": 1716454224085692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224156848, "dur": 5,
    "args": {
      "External id": 174885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174885,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 174885, "pid": 5, "tid": 7, "ts": 1716454224156848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085713, "dur": 6,
    "args": {
      "External id": 174885,
      "cbid": 211, "correlation": 174885
    }
  },
  {
    "ph": "s", "id": 174885, "pid": 76337, "tid": -914061504, "ts": 1716454224085713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224156854, "dur": 57,
    "args": {
      "External id": 174888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174888, "pid": 5, "tid": 7, "ts": 1716454224156854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085732, "dur": 6,
    "args": {
      "External id": 174888,
      "cbid": 211, "correlation": 174888
    }
  },
  {
    "ph": "s", "id": 174888, "pid": 76337, "tid": -914061504, "ts": 1716454224085732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224156912, "dur": 37,
    "args": {
      "External id": 174897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174897,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174897, "pid": 5, "tid": 7, "ts": 1716454224156912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085770, "dur": 11,
    "args": {
      "External id": 174897,
      "cbid": 211, "correlation": 174897
    }
  },
  {
    "ph": "s", "id": 174897, "pid": 76337, "tid": -914061504, "ts": 1716454224085770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085825, "dur": 0,
    "args": {
      "External id": 174907,
      "cbid": 317, "correlation": 174907
    }
  },
  {
    "ph": "f", "id": 174907, "pid": 76337, "tid": -914061504, "ts": 1716454224085825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085826, "dur": 0,
    "args": {
      "External id": 174908,
      "cbid": 203, "correlation": 174908
    }
  },
  {
    "ph": "f", "id": 174908, "pid": 76337, "tid": -914061504, "ts": 1716454224085826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085827, "dur": 0,
    "args": {
      "External id": 174909,
      "cbid": 205, "correlation": 174909
    }
  },
  {
    "ph": "f", "id": 174909, "pid": 76337, "tid": -914061504, "ts": 1716454224085827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224156950, "dur": 39,
    "args": {
      "External id": 174913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174913,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174913, "pid": 5, "tid": 7, "ts": 1716454224156950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085842, "dur": 11,
    "args": {
      "External id": 174913,
      "cbid": 211, "correlation": 174913
    }
  },
  {
    "ph": "s", "id": 174913, "pid": 76337, "tid": -914061504, "ts": 1716454224085842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224156990, "dur": 83,
    "args": {
      "External id": 174915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174915,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174915, "pid": 5, "tid": 7, "ts": 1716454224156990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085856, "dur": 5,
    "args": {
      "External id": 174915,
      "cbid": 211, "correlation": 174915
    }
  },
  {
    "ph": "s", "id": 174915, "pid": 76337, "tid": -914061504, "ts": 1716454224085856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224157074, "dur": 1279,
    "args": {
      "External id": 174917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174917,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 174917, "pid": 5, "tid": 7, "ts": 1716454224157074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085867, "dur": 6,
    "args": {
      "External id": 174917,
      "cbid": 211, "correlation": 174917
    }
  },
  {
    "ph": "s", "id": 174917, "pid": 76337, "tid": -914061504, "ts": 1716454224085867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224158354, "dur": 21,
    "args": {
      "External id": 174919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174919,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174919, "pid": 5, "tid": 7, "ts": 1716454224158354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085878, "dur": 6,
    "args": {
      "External id": 174919,
      "cbid": 211, "correlation": 174919
    }
  },
  {
    "ph": "s", "id": 174919, "pid": 76337, "tid": -914061504, "ts": 1716454224085878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224158377, "dur": 33,
    "args": {
      "External id": 174925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174925, "pid": 5, "tid": 7, "ts": 1716454224158377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085906, "dur": 8,
    "args": {
      "External id": 174925,
      "cbid": 211, "correlation": 174925
    }
  },
  {
    "ph": "s", "id": 174925, "pid": 76337, "tid": -914061504, "ts": 1716454224085906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224158411, "dur": 4,
    "args": {
      "External id": 174933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174933,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 174933, "pid": 5, "tid": 7, "ts": 1716454224158411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224085950, "dur": 9,
    "args": {
      "External id": 174933,
      "cbid": 211, "correlation": 174933
    }
  },
  {
    "ph": "s", "id": 174933, "pid": 76337, "tid": -914061504, "ts": 1716454224085950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086024, "dur": 1,
    "args": {
      "External id": 174949,
      "cbid": 251, "correlation": 174949
    }
  },
  {
    "ph": "f", "id": 174949, "pid": 76337, "tid": -914061504, "ts": 1716454224086024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086029, "dur": 0,
    "args": {
      "External id": 174951,
      "cbid": 251, "correlation": 174951
    }
  },
  {
    "ph": "f", "id": 174951, "pid": 76337, "tid": -914061504, "ts": 1716454224086029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224158416, "dur": 12,
    "args": {
      "External id": 174952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174952,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 174952, "pid": 5, "tid": 7, "ts": 1716454224158416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086031, "dur": 12,
    "args": {
      "External id": 174952,
      "cbid": 211, "correlation": 174952
    }
  },
  {
    "ph": "s", "id": 174952, "pid": 76337, "tid": -914061504, "ts": 1716454224086031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224158430, "dur": 5,
    "args": {
      "External id": 174954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174954,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 174954, "pid": 5, "tid": 7, "ts": 1716454224158430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086045, "dur": 5,
    "args": {
      "External id": 174954,
      "cbid": 211, "correlation": 174954
    }
  },
  {
    "ph": "s", "id": 174954, "pid": 76337, "tid": -914061504, "ts": 1716454224086045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224158436, "dur": 29,
    "args": {
      "External id": 174964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174964,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174964, "pid": 5, "tid": 7, "ts": 1716454224158436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086103, "dur": 12,
    "args": {
      "External id": 174964,
      "cbid": 211, "correlation": 174964
    }
  },
  {
    "ph": "s", "id": 174964, "pid": 76337, "tid": -914061504, "ts": 1716454224086103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224158467, "dur": 31,
    "args": {
      "External id": 174984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174984,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 174984, "pid": 5, "tid": 7, "ts": 1716454224158467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086171, "dur": 11,
    "args": {
      "External id": 174984,
      "cbid": 211, "correlation": 174984
    }
  },
  {
    "ph": "s", "id": 174984, "pid": 76337, "tid": -914061504, "ts": 1716454224086171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224158499, "dur": 4,
    "args": {
      "External id": 174996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174996,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 174996, "pid": 5, "tid": 7, "ts": 1716454224158499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086192, "dur": 6,
    "args": {
      "External id": 174996,
      "cbid": 211, "correlation": 174996
    }
  },
  {
    "ph": "s", "id": 174996, "pid": 76337, "tid": -914061504, "ts": 1716454224086192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224158504, "dur": 30,
    "args": {
      "External id": 174999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 174999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 174999, "pid": 5, "tid": 7, "ts": 1716454224158504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086211, "dur": 7,
    "args": {
      "External id": 174999,
      "cbid": 211, "correlation": 174999
    }
  },
  {
    "ph": "s", "id": 174999, "pid": 76337, "tid": -914061504, "ts": 1716454224086211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224158535, "dur": 21,
    "args": {
      "External id": 175008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175008,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175008, "pid": 5, "tid": 7, "ts": 1716454224158535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086252, "dur": 10,
    "args": {
      "External id": 175008,
      "cbid": 211, "correlation": 175008
    }
  },
  {
    "ph": "s", "id": 175008, "pid": 76337, "tid": -914061504, "ts": 1716454224086252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086318, "dur": 0,
    "args": {
      "External id": 175018,
      "cbid": 317, "correlation": 175018
    }
  },
  {
    "ph": "f", "id": 175018, "pid": 76337, "tid": -914061504, "ts": 1716454224086318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086319, "dur": 0,
    "args": {
      "External id": 175019,
      "cbid": 203, "correlation": 175019
    }
  },
  {
    "ph": "f", "id": 175019, "pid": 76337, "tid": -914061504, "ts": 1716454224086319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086320, "dur": 0,
    "args": {
      "External id": 175020,
      "cbid": 205, "correlation": 175020
    }
  },
  {
    "ph": "f", "id": 175020, "pid": 76337, "tid": -914061504, "ts": 1716454224086320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224158558, "dur": 22,
    "args": {
      "External id": 175024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175024,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175024, "pid": 5, "tid": 7, "ts": 1716454224158558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086334, "dur": 15,
    "args": {
      "External id": 175024,
      "cbid": 211, "correlation": 175024
    }
  },
  {
    "ph": "s", "id": 175024, "pid": 76337, "tid": -914061504, "ts": 1716454224086334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224158581, "dur": 44,
    "args": {
      "External id": 175026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175026,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175026, "pid": 5, "tid": 7, "ts": 1716454224158581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086351, "dur": 5,
    "args": {
      "External id": 175026,
      "cbid": 211, "correlation": 175026
    }
  },
  {
    "ph": "s", "id": 175026, "pid": 76337, "tid": -914061504, "ts": 1716454224086351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224158625, "dur": 646,
    "args": {
      "External id": 175028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175028,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175028, "pid": 5, "tid": 7, "ts": 1716454224158625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086362, "dur": 6,
    "args": {
      "External id": 175028,
      "cbid": 211, "correlation": 175028
    }
  },
  {
    "ph": "s", "id": 175028, "pid": 76337, "tid": -914061504, "ts": 1716454224086362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224159272, "dur": 21,
    "args": {
      "External id": 175030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175030,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175030, "pid": 5, "tid": 7, "ts": 1716454224159272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086372, "dur": 5,
    "args": {
      "External id": 175030,
      "cbid": 211, "correlation": 175030
    }
  },
  {
    "ph": "s", "id": 175030, "pid": 76337, "tid": -914061504, "ts": 1716454224086372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224159295, "dur": 33,
    "args": {
      "External id": 175036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175036, "pid": 5, "tid": 7, "ts": 1716454224159295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086400, "dur": 8,
    "args": {
      "External id": 175036,
      "cbid": 211, "correlation": 175036
    }
  },
  {
    "ph": "s", "id": 175036, "pid": 76337, "tid": -914061504, "ts": 1716454224086400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086461, "dur": 0,
    "args": {
      "External id": 175046,
      "cbid": 317, "correlation": 175046
    }
  },
  {
    "ph": "f", "id": 175046, "pid": 76337, "tid": -914061504, "ts": 1716454224086461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086462, "dur": 0,
    "args": {
      "External id": 175047,
      "cbid": 203, "correlation": 175047
    }
  },
  {
    "ph": "f", "id": 175047, "pid": 76337, "tid": -914061504, "ts": 1716454224086462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086462, "dur": 0,
    "args": {
      "External id": 175048,
      "cbid": 205, "correlation": 175048
    }
  },
  {
    "ph": "f", "id": 175048, "pid": 76337, "tid": -914061504, "ts": 1716454224086462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224159329, "dur": 39,
    "args": {
      "External id": 175052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175052,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175052, "pid": 5, "tid": 7, "ts": 1716454224159329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086475, "dur": 13,
    "args": {
      "External id": 175052,
      "cbid": 211, "correlation": 175052
    }
  },
  {
    "ph": "s", "id": 175052, "pid": 76337, "tid": -914061504, "ts": 1716454224086475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224159370, "dur": 191,
    "args": {
      "External id": 175054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175054,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175054, "pid": 5, "tid": 7, "ts": 1716454224159370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086493, "dur": 6,
    "args": {
      "External id": 175054,
      "cbid": 211, "correlation": 175054
    }
  },
  {
    "ph": "s", "id": 175054, "pid": 76337, "tid": -914061504, "ts": 1716454224086493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224159562, "dur": 22,
    "args": {
      "External id": 175056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175056, "pid": 5, "tid": 7, "ts": 1716454224159562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086502, "dur": 5,
    "args": {
      "External id": 175056,
      "cbid": 211, "correlation": 175056
    }
  },
  {
    "ph": "s", "id": 175056, "pid": 76337, "tid": -914061504, "ts": 1716454224086502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224159585, "dur": 32,
    "args": {
      "External id": 175062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175062, "pid": 5, "tid": 7, "ts": 1716454224159585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086528, "dur": 9,
    "args": {
      "External id": 175062,
      "cbid": 211, "correlation": 175062
    }
  },
  {
    "ph": "s", "id": 175062, "pid": 76337, "tid": -914061504, "ts": 1716454224086528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224159619, "dur": 27,
    "args": {
      "External id": 175070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175070,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175070, "pid": 5, "tid": 7, "ts": 1716454224159619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086558, "dur": 8,
    "args": {
      "External id": 175070,
      "cbid": 211, "correlation": 175070
    }
  },
  {
    "ph": "s", "id": 175070, "pid": 76337, "tid": -914061504, "ts": 1716454224086558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224159647, "dur": 20,
    "args": {
      "External id": 175078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175078,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175078, "pid": 5, "tid": 7, "ts": 1716454224159647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086587, "dur": 8,
    "args": {
      "External id": 175078,
      "cbid": 211, "correlation": 175078
    }
  },
  {
    "ph": "s", "id": 175078, "pid": 76337, "tid": -914061504, "ts": 1716454224086587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224159668, "dur": 30,
    "args": {
      "External id": 175098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175098,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 175098, "pid": 5, "tid": 7, "ts": 1716454224159668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086673, "dur": 13,
    "args": {
      "External id": 175098,
      "cbid": 211, "correlation": 175098
    }
  },
  {
    "ph": "s", "id": 175098, "pid": 76337, "tid": -914061504, "ts": 1716454224086673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224159700, "dur": 4,
    "args": {
      "External id": 175110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175110,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 175110, "pid": 5, "tid": 7, "ts": 1716454224159700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086696, "dur": 6,
    "args": {
      "External id": 175110,
      "cbid": 211, "correlation": 175110
    }
  },
  {
    "ph": "s", "id": 175110, "pid": 76337, "tid": -914061504, "ts": 1716454224086696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224159705, "dur": 30,
    "args": {
      "External id": 175113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175113,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175113, "pid": 5, "tid": 7, "ts": 1716454224159705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086713, "dur": 6,
    "args": {
      "External id": 175113,
      "cbid": 211, "correlation": 175113
    }
  },
  {
    "ph": "s", "id": 175113, "pid": 76337, "tid": -914061504, "ts": 1716454224086713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086773, "dur": 0,
    "args": {
      "External id": 175124,
      "cbid": 317, "correlation": 175124
    }
  },
  {
    "ph": "f", "id": 175124, "pid": 76337, "tid": -914061504, "ts": 1716454224086773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086773, "dur": 0,
    "args": {
      "External id": 175125,
      "cbid": 203, "correlation": 175125
    }
  },
  {
    "ph": "f", "id": 175125, "pid": 76337, "tid": -914061504, "ts": 1716454224086773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086774, "dur": 0,
    "args": {
      "External id": 175126,
      "cbid": 205, "correlation": 175126
    }
  },
  {
    "ph": "f", "id": 175126, "pid": 76337, "tid": -914061504, "ts": 1716454224086774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224159737, "dur": 23,
    "args": {
      "External id": 175130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175130,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175130, "pid": 5, "tid": 7, "ts": 1716454224159737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086789, "dur": 12,
    "args": {
      "External id": 175130,
      "cbid": 211, "correlation": 175130
    }
  },
  {
    "ph": "s", "id": 175130, "pid": 76337, "tid": -914061504, "ts": 1716454224086789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224159761, "dur": 106,
    "args": {
      "External id": 175132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175132,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175132, "pid": 5, "tid": 7, "ts": 1716454224159761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086807, "dur": 6,
    "args": {
      "External id": 175132,
      "cbid": 211, "correlation": 175132
    }
  },
  {
    "ph": "s", "id": 175132, "pid": 76337, "tid": -914061504, "ts": 1716454224086807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224159868, "dur": 23,
    "args": {
      "External id": 175134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175134, "pid": 5, "tid": 7, "ts": 1716454224159868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086817, "dur": 5,
    "args": {
      "External id": 175134,
      "cbid": 211, "correlation": 175134
    }
  },
  {
    "ph": "s", "id": 175134, "pid": 76337, "tid": -914061504, "ts": 1716454224086817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224159892, "dur": 33,
    "args": {
      "External id": 175140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175140, "pid": 5, "tid": 7, "ts": 1716454224159892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086844, "dur": 8,
    "args": {
      "External id": 175140,
      "cbid": 211, "correlation": 175140
    }
  },
  {
    "ph": "s", "id": 175140, "pid": 76337, "tid": -914061504, "ts": 1716454224086844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224159926, "dur": 164,
    "args": {
      "External id": 175149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175149,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175149, "pid": 5, "tid": 7, "ts": 1716454224159926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086931, "dur": 15,
    "args": {
      "External id": 175149,
      "cbid": 211, "correlation": 175149
    }
  },
  {
    "ph": "s", "id": 175149, "pid": 76337, "tid": -914061504, "ts": 1716454224086931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224160091, "dur": 64,
    "args": {
      "External id": 175171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175171,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175171, "pid": 5, "tid": 7, "ts": 1716454224160091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224086999, "dur": 11,
    "args": {
      "External id": 175171,
      "cbid": 211, "correlation": 175171
    }
  },
  {
    "ph": "s", "id": 175171, "pid": 76337, "tid": -914061504, "ts": 1716454224086999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087093, "dur": 1,
    "args": {
      "External id": 175182,
      "cbid": 251, "correlation": 175182
    }
  },
  {
    "ph": "f", "id": 175182, "pid": 76337, "tid": -914061504, "ts": 1716454224087093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224160157, "dur": 152,
    "args": {
      "External id": 175183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175183,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175183, "pid": 5, "tid": 7, "ts": 1716454224160157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087098, "dur": 13,
    "args": {
      "External id": 175183,
      "cbid": 211, "correlation": 175183
    }
  },
  {
    "ph": "s", "id": 175183, "pid": 76337, "tid": -914061504, "ts": 1716454224087098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087170, "dur": 1,
    "args": {
      "External id": 175194,
      "cbid": 251, "correlation": 175194
    }
  },
  {
    "ph": "f", "id": 175194, "pid": 76337, "tid": -914061504, "ts": 1716454224087170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224160310, "dur": 144,
    "args": {
      "External id": 175195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175195,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175195, "pid": 5, "tid": 7, "ts": 1716454224160310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087174, "dur": 12,
    "args": {
      "External id": 175195,
      "cbid": 211, "correlation": 175195
    }
  },
  {
    "ph": "s", "id": 175195, "pid": 76337, "tid": -914061504, "ts": 1716454224087174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087244, "dur": 1,
    "args": {
      "External id": 175206,
      "cbid": 251, "correlation": 175206
    }
  },
  {
    "ph": "f", "id": 175206, "pid": 76337, "tid": -914061504, "ts": 1716454224087244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224160456, "dur": 146,
    "args": {
      "External id": 175207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175207,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175207, "pid": 5, "tid": 7, "ts": 1716454224160456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087248, "dur": 12,
    "args": {
      "External id": 175207,
      "cbid": 211, "correlation": 175207
    }
  },
  {
    "ph": "s", "id": 175207, "pid": 76337, "tid": -914061504, "ts": 1716454224087248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224160603, "dur": 1946,
    "args": {
      "External id": 175228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175228,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 175228, "pid": 5, "tid": 7, "ts": 1716454224160603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087330, "dur": 13,
    "args": {
      "External id": 175228,
      "cbid": 211, "correlation": 175228
    }
  },
  {
    "ph": "s", "id": 175228, "pid": 76337, "tid": -914061504, "ts": 1716454224087330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087432, "dur": 1,
    "args": {
      "External id": 175246,
      "cbid": 251, "correlation": 175246
    }
  },
  {
    "ph": "f", "id": 175246, "pid": 76337, "tid": -914061504, "ts": 1716454224087432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224162551, "dur": 148,
    "args": {
      "External id": 175248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175248,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 175248, "pid": 5, "tid": 7, "ts": 1716454224162551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087438, "dur": 14,
    "args": {
      "External id": 175248,
      "cbid": 211, "correlation": 175248
    }
  },
  {
    "ph": "s", "id": 175248, "pid": 76337, "tid": -914061504, "ts": 1716454224087438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224162700, "dur": 36,
    "args": {
      "External id": 175256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175256,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175256, "pid": 5, "tid": 7, "ts": 1716454224162700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087512, "dur": 12,
    "args": {
      "External id": 175256,
      "cbid": 211, "correlation": 175256
    }
  },
  {
    "ph": "s", "id": 175256, "pid": 76337, "tid": -914061504, "ts": 1716454224087512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224162737, "dur": 50,
    "args": {
      "External id": 175264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175264,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175264, "pid": 5, "tid": 7, "ts": 1716454224162737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087551, "dur": 8,
    "args": {
      "External id": 175264,
      "cbid": 211, "correlation": 175264
    }
  },
  {
    "ph": "s", "id": 175264, "pid": 76337, "tid": -914061504, "ts": 1716454224087551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224162789, "dur": 29,
    "args": {
      "External id": 175275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175275,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175275, "pid": 5, "tid": 7, "ts": 1716454224162789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087622, "dur": 12,
    "args": {
      "External id": 175275,
      "cbid": 211, "correlation": 175275
    }
  },
  {
    "ph": "s", "id": 175275, "pid": 76337, "tid": -914061504, "ts": 1716454224087622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224162819, "dur": 34,
    "args": {
      "External id": 175297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175297,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175297, "pid": 5, "tid": 7, "ts": 1716454224162819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087654, "dur": 8,
    "args": {
      "External id": 175297,
      "cbid": 211, "correlation": 175297
    }
  },
  {
    "ph": "s", "id": 175297, "pid": 76337, "tid": -914061504, "ts": 1716454224087654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087744, "dur": 1,
    "args": {
      "External id": 175308,
      "cbid": 251, "correlation": 175308
    }
  },
  {
    "ph": "f", "id": 175308, "pid": 76337, "tid": -914061504, "ts": 1716454224087744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224162855, "dur": 91,
    "args": {
      "External id": 175309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175309,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175309, "pid": 5, "tid": 7, "ts": 1716454224162855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087750, "dur": 13,
    "args": {
      "External id": 175309,
      "cbid": 211, "correlation": 175309
    }
  },
  {
    "ph": "s", "id": 175309, "pid": 76337, "tid": -914061504, "ts": 1716454224087750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087822, "dur": 1,
    "args": {
      "External id": 175320,
      "cbid": 251, "correlation": 175320
    }
  },
  {
    "ph": "f", "id": 175320, "pid": 76337, "tid": -914061504, "ts": 1716454224087822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087825, "dur": 0,
    "args": {
      "External id": 175321,
      "cbid": 251, "correlation": 175321
    }
  },
  {
    "ph": "f", "id": 175321, "pid": 76337, "tid": -914061504, "ts": 1716454224087825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224162947, "dur": 11,
    "args": {
      "External id": 175322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175322,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 175322, "pid": 5, "tid": 7, "ts": 1716454224162947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087827, "dur": 12,
    "args": {
      "External id": 175322,
      "cbid": 211, "correlation": 175322
    }
  },
  {
    "ph": "s", "id": 175322, "pid": 76337, "tid": -914061504, "ts": 1716454224087827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224162959, "dur": 5,
    "args": {
      "External id": 175324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175324,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 175324, "pid": 5, "tid": 7, "ts": 1716454224162959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087841, "dur": 6,
    "args": {
      "External id": 175324,
      "cbid": 211, "correlation": 175324
    }
  },
  {
    "ph": "s", "id": 175324, "pid": 76337, "tid": -914061504, "ts": 1716454224087841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087902, "dur": 1,
    "args": {
      "External id": 175335,
      "cbid": 251, "correlation": 175335
    }
  },
  {
    "ph": "f", "id": 175335, "pid": 76337, "tid": -914061504, "ts": 1716454224087902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087906, "dur": 0,
    "args": {
      "External id": 175336,
      "cbid": 251, "correlation": 175336
    }
  },
  {
    "ph": "f", "id": 175336, "pid": 76337, "tid": -914061504, "ts": 1716454224087906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224162966, "dur": 7,
    "args": {
      "External id": 175337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175337,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 175337, "pid": 5, "tid": 7, "ts": 1716454224162966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087907, "dur": 11,
    "args": {
      "External id": 175337,
      "cbid": 211, "correlation": 175337
    }
  },
  {
    "ph": "s", "id": 175337, "pid": 76337, "tid": -914061504, "ts": 1716454224087907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224162974, "dur": 3,
    "args": {
      "External id": 175339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175339,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 175339, "pid": 5, "tid": 7, "ts": 1716454224162974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224087920, "dur": 5,
    "args": {
      "External id": 175339,
      "cbid": 211, "correlation": 175339
    }
  },
  {
    "ph": "s", "id": 175339, "pid": 76337, "tid": -914061504, "ts": 1716454224087920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224162979, "dur": 92,
    "args": {
      "External id": 175360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175360,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 175360, "pid": 5, "tid": 7, "ts": 1716454224162979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088002, "dur": 13,
    "args": {
      "External id": 175360,
      "cbid": 211, "correlation": 175360
    }
  },
  {
    "ph": "s", "id": 175360, "pid": 76337, "tid": -914061504, "ts": 1716454224088002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088104, "dur": 1,
    "args": {
      "External id": 175378,
      "cbid": 251, "correlation": 175378
    }
  },
  {
    "ph": "f", "id": 175378, "pid": 76337, "tid": -914061504, "ts": 1716454224088104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224163072, "dur": 85,
    "args": {
      "External id": 175380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175380,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175380, "pid": 5, "tid": 7, "ts": 1716454224163072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088110, "dur": 13,
    "args": {
      "External id": 175380,
      "cbid": 211, "correlation": 175380
    }
  },
  {
    "ph": "s", "id": 175380, "pid": 76337, "tid": -914061504, "ts": 1716454224088110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224163158, "dur": 19,
    "args": {
      "External id": 175388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175388,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175388, "pid": 5, "tid": 7, "ts": 1716454224163158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088182, "dur": 12,
    "args": {
      "External id": 175388,
      "cbid": 211, "correlation": 175388
    }
  },
  {
    "ph": "s", "id": 175388, "pid": 76337, "tid": -914061504, "ts": 1716454224088182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224163178, "dur": 38,
    "args": {
      "External id": 175396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175396, "pid": 5, "tid": 7, "ts": 1716454224163178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088224, "dur": 9,
    "args": {
      "External id": 175396,
      "cbid": 211, "correlation": 175396
    }
  },
  {
    "ph": "s", "id": 175396, "pid": 76337, "tid": -914061504, "ts": 1716454224088224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224163218, "dur": 35,
    "args": {
      "External id": 175418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175418,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175418, "pid": 5, "tid": 7, "ts": 1716454224163218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088278, "dur": 10,
    "args": {
      "External id": 175418,
      "cbid": 211, "correlation": 175418
    }
  },
  {
    "ph": "s", "id": 175418, "pid": 76337, "tid": -914061504, "ts": 1716454224088278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088372, "dur": 1,
    "args": {
      "External id": 175434,
      "cbid": 251, "correlation": 175434
    }
  },
  {
    "ph": "f", "id": 175434, "pid": 76337, "tid": -914061504, "ts": 1716454224088372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088377, "dur": 0,
    "args": {
      "External id": 175436,
      "cbid": 251, "correlation": 175436
    }
  },
  {
    "ph": "f", "id": 175436, "pid": 76337, "tid": -914061504, "ts": 1716454224088377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224163254, "dur": 540,
    "args": {
      "External id": 175437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175437,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 175437, "pid": 5, "tid": 7, "ts": 1716454224163254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088381, "dur": 12,
    "args": {
      "External id": 175437,
      "cbid": 211, "correlation": 175437
    }
  },
  {
    "ph": "s", "id": 175437, "pid": 76337, "tid": -914061504, "ts": 1716454224088381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224163795, "dur": 124,
    "args": {
      "External id": 175445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175445, "pid": 5, "tid": 7, "ts": 1716454224163795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088445, "dur": 13,
    "args": {
      "External id": 175445,
      "cbid": 211, "correlation": 175445
    }
  },
  {
    "ph": "s", "id": 175445, "pid": 76337, "tid": -914061504, "ts": 1716454224088445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224163920, "dur": 128,
    "args": {
      "External id": 175453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175453,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175453, "pid": 5, "tid": 7, "ts": 1716454224163920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088476, "dur": 8,
    "args": {
      "External id": 175453,
      "cbid": 211, "correlation": 175453
    }
  },
  {
    "ph": "s", "id": 175453, "pid": 76337, "tid": -914061504, "ts": 1716454224088476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088553, "dur": 1,
    "args": {
      "External id": 175469,
      "cbid": 251, "correlation": 175469
    }
  },
  {
    "ph": "f", "id": 175469, "pid": 76337, "tid": -914061504, "ts": 1716454224088553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224164049, "dur": 305,
    "args": {
      "External id": 175471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175471,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175471, "pid": 5, "tid": 7, "ts": 1716454224164049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088559, "dur": 12,
    "args": {
      "External id": 175471,
      "cbid": 211, "correlation": 175471
    }
  },
  {
    "ph": "s", "id": 175471, "pid": 76337, "tid": -914061504, "ts": 1716454224088559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224164356, "dur": 27,
    "args": {
      "External id": 175479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175479,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175479, "pid": 5, "tid": 7, "ts": 1716454224164356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088600, "dur": 10,
    "args": {
      "External id": 175479,
      "cbid": 211, "correlation": 175479
    }
  },
  {
    "ph": "s", "id": 175479, "pid": 76337, "tid": -914061504, "ts": 1716454224088600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224164383, "dur": 81,
    "args": {
      "External id": 175490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175490, "pid": 5, "tid": 7, "ts": 1716454224164383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088669, "dur": 12,
    "args": {
      "External id": 175490,
      "cbid": 211, "correlation": 175490
    }
  },
  {
    "ph": "s", "id": 175490, "pid": 76337, "tid": -914061504, "ts": 1716454224088669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088732, "dur": 0,
    "args": {
      "External id": 175502,
      "cbid": 317, "correlation": 175502
    }
  },
  {
    "ph": "f", "id": 175502, "pid": 76337, "tid": -914061504, "ts": 1716454224088732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088733, "dur": 0,
    "args": {
      "External id": 175503,
      "cbid": 203, "correlation": 175503
    }
  },
  {
    "ph": "f", "id": 175503, "pid": 76337, "tid": -914061504, "ts": 1716454224088733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088733, "dur": 0,
    "args": {
      "External id": 175504,
      "cbid": 205, "correlation": 175504
    }
  },
  {
    "ph": "f", "id": 175504, "pid": 76337, "tid": -914061504, "ts": 1716454224088733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224164466, "dur": 23,
    "args": {
      "External id": 175508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175508,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175508, "pid": 5, "tid": 7, "ts": 1716454224164466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088748, "dur": 13,
    "args": {
      "External id": 175508,
      "cbid": 211, "correlation": 175508
    }
  },
  {
    "ph": "s", "id": 175508, "pid": 76337, "tid": -914061504, "ts": 1716454224088748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224164490, "dur": 121,
    "args": {
      "External id": 175510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175510,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175510, "pid": 5, "tid": 7, "ts": 1716454224164490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088767, "dur": 7,
    "args": {
      "External id": 175510,
      "cbid": 211, "correlation": 175510
    }
  },
  {
    "ph": "s", "id": 175510, "pid": 76337, "tid": -914061504, "ts": 1716454224088767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224164612, "dur": 23,
    "args": {
      "External id": 175512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175512,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175512, "pid": 5, "tid": 7, "ts": 1716454224164612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088777, "dur": 5,
    "args": {
      "External id": 175512,
      "cbid": 211, "correlation": 175512
    }
  },
  {
    "ph": "s", "id": 175512, "pid": 76337, "tid": -914061504, "ts": 1716454224088777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224164637, "dur": 33,
    "args": {
      "External id": 175518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175518,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175518, "pid": 5, "tid": 7, "ts": 1716454224164637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088804, "dur": 9,
    "args": {
      "External id": 175518,
      "cbid": 211, "correlation": 175518
    }
  },
  {
    "ph": "s", "id": 175518, "pid": 76337, "tid": -914061504, "ts": 1716454224088804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224164671, "dur": 26,
    "args": {
      "External id": 175526,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175526,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175526, "pid": 5, "tid": 7, "ts": 1716454224164671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088836, "dur": 8,
    "args": {
      "External id": 175526,
      "cbid": 211, "correlation": 175526
    }
  },
  {
    "ph": "s", "id": 175526, "pid": 76337, "tid": -914061504, "ts": 1716454224088836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224164699, "dur": 47,
    "args": {
      "External id": 175535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175535,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175535, "pid": 5, "tid": 7, "ts": 1716454224164699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088875, "dur": 10,
    "args": {
      "External id": 175535,
      "cbid": 211, "correlation": 175535
    }
  },
  {
    "ph": "s", "id": 175535, "pid": 76337, "tid": -914061504, "ts": 1716454224088875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224164747, "dur": 41,
    "args": {
      "External id": 175555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175555,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 175555, "pid": 5, "tid": 7, "ts": 1716454224164747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088950, "dur": 13,
    "args": {
      "External id": 175555,
      "cbid": 211, "correlation": 175555
    }
  },
  {
    "ph": "s", "id": 175555, "pid": 76337, "tid": -914061504, "ts": 1716454224088950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224164790, "dur": 5,
    "args": {
      "External id": 175567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175567,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 175567, "pid": 5, "tid": 7, "ts": 1716454224164790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224088972, "dur": 15,
    "args": {
      "External id": 175567,
      "cbid": 211, "correlation": 175567
    }
  },
  {
    "ph": "s", "id": 175567, "pid": 76337, "tid": -914061504, "ts": 1716454224088972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224164797, "dur": 44,
    "args": {
      "External id": 175570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175570, "pid": 5, "tid": 7, "ts": 1716454224164797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089000, "dur": 7,
    "args": {
      "External id": 175570,
      "cbid": 211, "correlation": 175570
    }
  },
  {
    "ph": "s", "id": 175570, "pid": 76337, "tid": -914061504, "ts": 1716454224089000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224164843, "dur": 29,
    "args": {
      "External id": 175579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175579,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175579, "pid": 5, "tid": 7, "ts": 1716454224164843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089041, "dur": 10,
    "args": {
      "External id": 175579,
      "cbid": 211, "correlation": 175579
    }
  },
  {
    "ph": "s", "id": 175579, "pid": 76337, "tid": -914061504, "ts": 1716454224089041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089093, "dur": 0,
    "args": {
      "External id": 175589,
      "cbid": 317, "correlation": 175589
    }
  },
  {
    "ph": "f", "id": 175589, "pid": 76337, "tid": -914061504, "ts": 1716454224089093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089094, "dur": 0,
    "args": {
      "External id": 175590,
      "cbid": 203, "correlation": 175590
    }
  },
  {
    "ph": "f", "id": 175590, "pid": 76337, "tid": -914061504, "ts": 1716454224089094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089094, "dur": 0,
    "args": {
      "External id": 175591,
      "cbid": 205, "correlation": 175591
    }
  },
  {
    "ph": "f", "id": 175591, "pid": 76337, "tid": -914061504, "ts": 1716454224089094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224164873, "dur": 31,
    "args": {
      "External id": 175595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175595,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175595, "pid": 5, "tid": 7, "ts": 1716454224164873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089110, "dur": 11,
    "args": {
      "External id": 175595,
      "cbid": 211, "correlation": 175595
    }
  },
  {
    "ph": "s", "id": 175595, "pid": 76337, "tid": -914061504, "ts": 1716454224089110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224164905, "dur": 63,
    "args": {
      "External id": 175597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175597,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175597, "pid": 5, "tid": 7, "ts": 1716454224164905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089123, "dur": 5,
    "args": {
      "External id": 175597,
      "cbid": 211, "correlation": 175597
    }
  },
  {
    "ph": "s", "id": 175597, "pid": 76337, "tid": -914061504, "ts": 1716454224089123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224164970, "dur": 964,
    "args": {
      "External id": 175599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175599,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175599, "pid": 5, "tid": 7, "ts": 1716454224164970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089134, "dur": 6,
    "args": {
      "External id": 175599,
      "cbid": 211, "correlation": 175599
    }
  },
  {
    "ph": "s", "id": 175599, "pid": 76337, "tid": -914061504, "ts": 1716454224089134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224165936, "dur": 21,
    "args": {
      "External id": 175601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175601,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175601, "pid": 5, "tid": 7, "ts": 1716454224165936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089144, "dur": 5,
    "args": {
      "External id": 175601,
      "cbid": 211, "correlation": 175601
    }
  },
  {
    "ph": "s", "id": 175601, "pid": 76337, "tid": -914061504, "ts": 1716454224089144,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224165958, "dur": 33,
    "args": {
      "External id": 175607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175607,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175607, "pid": 5, "tid": 7, "ts": 1716454224165958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089172, "dur": 8,
    "args": {
      "External id": 175607,
      "cbid": 211, "correlation": 175607
    }
  },
  {
    "ph": "s", "id": 175607, "pid": 76337, "tid": -914061504, "ts": 1716454224089172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224165992, "dur": 4,
    "args": {
      "External id": 175615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175615,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 175615, "pid": 5, "tid": 7, "ts": 1716454224165992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089215, "dur": 9,
    "args": {
      "External id": 175615,
      "cbid": 211, "correlation": 175615
    }
  },
  {
    "ph": "s", "id": 175615, "pid": 76337, "tid": -914061504, "ts": 1716454224089215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089284, "dur": 1,
    "args": {
      "External id": 175631,
      "cbid": 251, "correlation": 175631
    }
  },
  {
    "ph": "f", "id": 175631, "pid": 76337, "tid": -914061504, "ts": 1716454224089284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089289, "dur": 0,
    "args": {
      "External id": 175633,
      "cbid": 251, "correlation": 175633
    }
  },
  {
    "ph": "f", "id": 175633, "pid": 76337, "tid": -914061504, "ts": 1716454224089289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224165998, "dur": 12,
    "args": {
      "External id": 175634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175634,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 175634, "pid": 5, "tid": 7, "ts": 1716454224165998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089291, "dur": 11,
    "args": {
      "External id": 175634,
      "cbid": 211, "correlation": 175634
    }
  },
  {
    "ph": "s", "id": 175634, "pid": 76337, "tid": -914061504, "ts": 1716454224089291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224166012, "dur": 5,
    "args": {
      "External id": 175636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175636,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 175636, "pid": 5, "tid": 7, "ts": 1716454224166012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089304, "dur": 6,
    "args": {
      "External id": 175636,
      "cbid": 211, "correlation": 175636
    }
  },
  {
    "ph": "s", "id": 175636, "pid": 76337, "tid": -914061504, "ts": 1716454224089304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224166018, "dur": 30,
    "args": {
      "External id": 175646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175646,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175646, "pid": 5, "tid": 7, "ts": 1716454224166018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089362, "dur": 12,
    "args": {
      "External id": 175646,
      "cbid": 211, "correlation": 175646
    }
  },
  {
    "ph": "s", "id": 175646, "pid": 76337, "tid": -914061504, "ts": 1716454224089362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224166049, "dur": 31,
    "args": {
      "External id": 175666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175666,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 175666, "pid": 5, "tid": 7, "ts": 1716454224166049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089430, "dur": 11,
    "args": {
      "External id": 175666,
      "cbid": 211, "correlation": 175666
    }
  },
  {
    "ph": "s", "id": 175666, "pid": 76337, "tid": -914061504, "ts": 1716454224089430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224166082, "dur": 4,
    "args": {
      "External id": 175678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175678,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 175678, "pid": 5, "tid": 7, "ts": 1716454224166082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089450, "dur": 6,
    "args": {
      "External id": 175678,
      "cbid": 211, "correlation": 175678
    }
  },
  {
    "ph": "s", "id": 175678, "pid": 76337, "tid": -914061504, "ts": 1716454224089450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224166087, "dur": 30,
    "args": {
      "External id": 175681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175681,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175681, "pid": 5, "tid": 7, "ts": 1716454224166087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089470, "dur": 7,
    "args": {
      "External id": 175681,
      "cbid": 211, "correlation": 175681
    }
  },
  {
    "ph": "s", "id": 175681, "pid": 76337, "tid": -914061504, "ts": 1716454224089470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224166119, "dur": 21,
    "args": {
      "External id": 175690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175690,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175690, "pid": 5, "tid": 7, "ts": 1716454224166119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089510, "dur": 10,
    "args": {
      "External id": 175690,
      "cbid": 211, "correlation": 175690
    }
  },
  {
    "ph": "s", "id": 175690, "pid": 76337, "tid": -914061504, "ts": 1716454224089510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089576, "dur": 0,
    "args": {
      "External id": 175700,
      "cbid": 317, "correlation": 175700
    }
  },
  {
    "ph": "f", "id": 175700, "pid": 76337, "tid": -914061504, "ts": 1716454224089576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089577, "dur": 0,
    "args": {
      "External id": 175701,
      "cbid": 203, "correlation": 175701
    }
  },
  {
    "ph": "f", "id": 175701, "pid": 76337, "tid": -914061504, "ts": 1716454224089577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089578, "dur": 0,
    "args": {
      "External id": 175702,
      "cbid": 205, "correlation": 175702
    }
  },
  {
    "ph": "f", "id": 175702, "pid": 76337, "tid": -914061504, "ts": 1716454224089578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224166140, "dur": 22,
    "args": {
      "External id": 175706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175706,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175706, "pid": 5, "tid": 7, "ts": 1716454224166140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089592, "dur": 12,
    "args": {
      "External id": 175706,
      "cbid": 211, "correlation": 175706
    }
  },
  {
    "ph": "s", "id": 175706, "pid": 76337, "tid": -914061504, "ts": 1716454224089592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224166164, "dur": 45,
    "args": {
      "External id": 175708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175708,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175708, "pid": 5, "tid": 7, "ts": 1716454224166164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089606, "dur": 6,
    "args": {
      "External id": 175708,
      "cbid": 211, "correlation": 175708
    }
  },
  {
    "ph": "s", "id": 175708, "pid": 76337, "tid": -914061504, "ts": 1716454224089606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224166210, "dur": 644,
    "args": {
      "External id": 175710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175710,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175710, "pid": 5, "tid": 7, "ts": 1716454224166210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089619, "dur": 6,
    "args": {
      "External id": 175710,
      "cbid": 211, "correlation": 175710
    }
  },
  {
    "ph": "s", "id": 175710, "pid": 76337, "tid": -914061504, "ts": 1716454224089619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224166856, "dur": 23,
    "args": {
      "External id": 175712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175712,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175712, "pid": 5, "tid": 7, "ts": 1716454224166856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089628, "dur": 5,
    "args": {
      "External id": 175712,
      "cbid": 211, "correlation": 175712
    }
  },
  {
    "ph": "s", "id": 175712, "pid": 76337, "tid": -914061504, "ts": 1716454224089628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224166880, "dur": 34,
    "args": {
      "External id": 175718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175718, "pid": 5, "tid": 7, "ts": 1716454224166880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089656, "dur": 9,
    "args": {
      "External id": 175718,
      "cbid": 211, "correlation": 175718
    }
  },
  {
    "ph": "s", "id": 175718, "pid": 76337, "tid": -914061504, "ts": 1716454224089656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089718, "dur": 0,
    "args": {
      "External id": 175728,
      "cbid": 317, "correlation": 175728
    }
  },
  {
    "ph": "f", "id": 175728, "pid": 76337, "tid": -914061504, "ts": 1716454224089718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089719, "dur": 0,
    "args": {
      "External id": 175729,
      "cbid": 203, "correlation": 175729
    }
  },
  {
    "ph": "f", "id": 175729, "pid": 76337, "tid": -914061504, "ts": 1716454224089719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089720, "dur": 0,
    "args": {
      "External id": 175730,
      "cbid": 205, "correlation": 175730
    }
  },
  {
    "ph": "f", "id": 175730, "pid": 76337, "tid": -914061504, "ts": 1716454224089720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224166915, "dur": 31,
    "args": {
      "External id": 175734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175734,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175734, "pid": 5, "tid": 7, "ts": 1716454224166915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089733, "dur": 12,
    "args": {
      "External id": 175734,
      "cbid": 211, "correlation": 175734
    }
  },
  {
    "ph": "s", "id": 175734, "pid": 76337, "tid": -914061504, "ts": 1716454224089733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224166948, "dur": 154,
    "args": {
      "External id": 175736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175736,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175736, "pid": 5, "tid": 7, "ts": 1716454224166948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089751, "dur": 6,
    "args": {
      "External id": 175736,
      "cbid": 211, "correlation": 175736
    }
  },
  {
    "ph": "s", "id": 175736, "pid": 76337, "tid": -914061504, "ts": 1716454224089751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224167103, "dur": 22,
    "args": {
      "External id": 175738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175738, "pid": 5, "tid": 7, "ts": 1716454224167103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089761, "dur": 6,
    "args": {
      "External id": 175738,
      "cbid": 211, "correlation": 175738
    }
  },
  {
    "ph": "s", "id": 175738, "pid": 76337, "tid": -914061504, "ts": 1716454224089761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224167126, "dur": 33,
    "args": {
      "External id": 175744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175744, "pid": 5, "tid": 7, "ts": 1716454224167126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089787, "dur": 8,
    "args": {
      "External id": 175744,
      "cbid": 211, "correlation": 175744
    }
  },
  {
    "ph": "s", "id": 175744, "pid": 76337, "tid": -914061504, "ts": 1716454224089787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224167160, "dur": 27,
    "args": {
      "External id": 175752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175752,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175752, "pid": 5, "tid": 7, "ts": 1716454224167160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089818, "dur": 8,
    "args": {
      "External id": 175752,
      "cbid": 211, "correlation": 175752
    }
  },
  {
    "ph": "s", "id": 175752, "pid": 76337, "tid": -914061504, "ts": 1716454224089818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224167188, "dur": 20,
    "args": {
      "External id": 175760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175760,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175760, "pid": 5, "tid": 7, "ts": 1716454224167188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089846, "dur": 8,
    "args": {
      "External id": 175760,
      "cbid": 211, "correlation": 175760
    }
  },
  {
    "ph": "s", "id": 175760, "pid": 76337, "tid": -914061504, "ts": 1716454224089846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224167210, "dur": 30,
    "args": {
      "External id": 175780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175780,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 175780, "pid": 5, "tid": 7, "ts": 1716454224167210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089931, "dur": 13,
    "args": {
      "External id": 175780,
      "cbid": 211, "correlation": 175780
    }
  },
  {
    "ph": "s", "id": 175780, "pid": 76337, "tid": -914061504, "ts": 1716454224089931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224167241, "dur": 4,
    "args": {
      "External id": 175792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175792,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 175792, "pid": 5, "tid": 7, "ts": 1716454224167241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089953, "dur": 6,
    "args": {
      "External id": 175792,
      "cbid": 211, "correlation": 175792
    }
  },
  {
    "ph": "s", "id": 175792, "pid": 76337, "tid": -914061504, "ts": 1716454224089953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224167247, "dur": 31,
    "args": {
      "External id": 175795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175795,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175795, "pid": 5, "tid": 7, "ts": 1716454224167247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224089971, "dur": 16,
    "args": {
      "External id": 175795,
      "cbid": 211, "correlation": 175795
    }
  },
  {
    "ph": "s", "id": 175795, "pid": 76337, "tid": -914061504, "ts": 1716454224089971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090043, "dur": 0,
    "args": {
      "External id": 175806,
      "cbid": 317, "correlation": 175806
    }
  },
  {
    "ph": "f", "id": 175806, "pid": 76337, "tid": -914061504, "ts": 1716454224090043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090043, "dur": 0,
    "args": {
      "External id": 175807,
      "cbid": 203, "correlation": 175807
    }
  },
  {
    "ph": "f", "id": 175807, "pid": 76337, "tid": -914061504, "ts": 1716454224090043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090044, "dur": 0,
    "args": {
      "External id": 175808,
      "cbid": 205, "correlation": 175808
    }
  },
  {
    "ph": "f", "id": 175808, "pid": 76337, "tid": -914061504, "ts": 1716454224090044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224167280, "dur": 22,
    "args": {
      "External id": 175812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175812, "pid": 5, "tid": 7, "ts": 1716454224167280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090057, "dur": 13,
    "args": {
      "External id": 175812,
      "cbid": 211, "correlation": 175812
    }
  },
  {
    "ph": "s", "id": 175812, "pid": 76337, "tid": -914061504, "ts": 1716454224090057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224167303, "dur": 105,
    "args": {
      "External id": 175814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175814,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175814, "pid": 5, "tid": 7, "ts": 1716454224167303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090076, "dur": 6,
    "args": {
      "External id": 175814,
      "cbid": 211, "correlation": 175814
    }
  },
  {
    "ph": "s", "id": 175814, "pid": 76337, "tid": -914061504, "ts": 1716454224090076,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224167410, "dur": 21,
    "args": {
      "External id": 175816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175816,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175816, "pid": 5, "tid": 7, "ts": 1716454224167410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090085, "dur": 5,
    "args": {
      "External id": 175816,
      "cbid": 211, "correlation": 175816
    }
  },
  {
    "ph": "s", "id": 175816, "pid": 76337, "tid": -914061504, "ts": 1716454224090085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224167432, "dur": 32,
    "args": {
      "External id": 175822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175822, "pid": 5, "tid": 7, "ts": 1716454224167432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090112, "dur": 9,
    "args": {
      "External id": 175822,
      "cbid": 211, "correlation": 175822
    }
  },
  {
    "ph": "s", "id": 175822, "pid": 76337, "tid": -914061504, "ts": 1716454224090112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224167466, "dur": 195,
    "args": {
      "External id": 175831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175831,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175831, "pid": 5, "tid": 7, "ts": 1716454224167466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090208, "dur": 16,
    "args": {
      "External id": 175831,
      "cbid": 211, "correlation": 175831
    }
  },
  {
    "ph": "s", "id": 175831, "pid": 76337, "tid": -914061504, "ts": 1716454224090208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224167662, "dur": 65,
    "args": {
      "External id": 175853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175853,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175853, "pid": 5, "tid": 7, "ts": 1716454224167662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090269, "dur": 10,
    "args": {
      "External id": 175853,
      "cbid": 211, "correlation": 175853
    }
  },
  {
    "ph": "s", "id": 175853, "pid": 76337, "tid": -914061504, "ts": 1716454224090269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090359, "dur": 1,
    "args": {
      "External id": 175864,
      "cbid": 251, "correlation": 175864
    }
  },
  {
    "ph": "f", "id": 175864, "pid": 76337, "tid": -914061504, "ts": 1716454224090359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224167728, "dur": 151,
    "args": {
      "External id": 175865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175865,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175865, "pid": 5, "tid": 7, "ts": 1716454224167728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090364, "dur": 14,
    "args": {
      "External id": 175865,
      "cbid": 211, "correlation": 175865
    }
  },
  {
    "ph": "s", "id": 175865, "pid": 76337, "tid": -914061504, "ts": 1716454224090364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090438, "dur": 1,
    "args": {
      "External id": 175876,
      "cbid": 251, "correlation": 175876
    }
  },
  {
    "ph": "f", "id": 175876, "pid": 76337, "tid": -914061504, "ts": 1716454224090438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224167881, "dur": 144,
    "args": {
      "External id": 175877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175877,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175877, "pid": 5, "tid": 7, "ts": 1716454224167881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090442, "dur": 12,
    "args": {
      "External id": 175877,
      "cbid": 211, "correlation": 175877
    }
  },
  {
    "ph": "s", "id": 175877, "pid": 76337, "tid": -914061504, "ts": 1716454224090442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090511, "dur": 1,
    "args": {
      "External id": 175888,
      "cbid": 251, "correlation": 175888
    }
  },
  {
    "ph": "f", "id": 175888, "pid": 76337, "tid": -914061504, "ts": 1716454224090511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224168027, "dur": 143,
    "args": {
      "External id": 175889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175889,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175889, "pid": 5, "tid": 7, "ts": 1716454224168027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090515, "dur": 12,
    "args": {
      "External id": 175889,
      "cbid": 211, "correlation": 175889
    }
  },
  {
    "ph": "s", "id": 175889, "pid": 76337, "tid": -914061504, "ts": 1716454224090515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224168171, "dur": 1950,
    "args": {
      "External id": 175910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175910,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 175910, "pid": 5, "tid": 7, "ts": 1716454224168171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090598, "dur": 12,
    "args": {
      "External id": 175910,
      "cbid": 211, "correlation": 175910
    }
  },
  {
    "ph": "s", "id": 175910, "pid": 76337, "tid": -914061504, "ts": 1716454224090598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090700, "dur": 1,
    "args": {
      "External id": 175928,
      "cbid": 251, "correlation": 175928
    }
  },
  {
    "ph": "f", "id": 175928, "pid": 76337, "tid": -914061504, "ts": 1716454224090700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224170123, "dur": 148,
    "args": {
      "External id": 175930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175930,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 175930, "pid": 5, "tid": 7, "ts": 1716454224170123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090705, "dur": 14,
    "args": {
      "External id": 175930,
      "cbid": 211, "correlation": 175930
    }
  },
  {
    "ph": "s", "id": 175930, "pid": 76337, "tid": -914061504, "ts": 1716454224090705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224170272, "dur": 36,
    "args": {
      "External id": 175938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175938,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175938, "pid": 5, "tid": 7, "ts": 1716454224170272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090780, "dur": 12,
    "args": {
      "External id": 175938,
      "cbid": 211, "correlation": 175938
    }
  },
  {
    "ph": "s", "id": 175938, "pid": 76337, "tid": -914061504, "ts": 1716454224090780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224170309, "dur": 50,
    "args": {
      "External id": 175946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175946,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175946, "pid": 5, "tid": 7, "ts": 1716454224170309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090819, "dur": 10,
    "args": {
      "External id": 175946,
      "cbid": 211, "correlation": 175946
    }
  },
  {
    "ph": "s", "id": 175946, "pid": 76337, "tid": -914061504, "ts": 1716454224090819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224170360, "dur": 30,
    "args": {
      "External id": 175957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175957,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175957, "pid": 5, "tid": 7, "ts": 1716454224170360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090891, "dur": 13,
    "args": {
      "External id": 175957,
      "cbid": 211, "correlation": 175957
    }
  },
  {
    "ph": "s", "id": 175957, "pid": 76337, "tid": -914061504, "ts": 1716454224090891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224170392, "dur": 35,
    "args": {
      "External id": 175979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175979,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 175979, "pid": 5, "tid": 7, "ts": 1716454224170392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224090922, "dur": 8,
    "args": {
      "External id": 175979,
      "cbid": 211, "correlation": 175979
    }
  },
  {
    "ph": "s", "id": 175979, "pid": 76337, "tid": -914061504, "ts": 1716454224090922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091018, "dur": 1,
    "args": {
      "External id": 175990,
      "cbid": 251, "correlation": 175990
    }
  },
  {
    "ph": "f", "id": 175990, "pid": 76337, "tid": -914061504, "ts": 1716454224091018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224170428, "dur": 91,
    "args": {
      "External id": 175991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 175991,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 175991, "pid": 5, "tid": 7, "ts": 1716454224170428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091023, "dur": 14,
    "args": {
      "External id": 175991,
      "cbid": 211, "correlation": 175991
    }
  },
  {
    "ph": "s", "id": 175991, "pid": 76337, "tid": -914061504, "ts": 1716454224091023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091097, "dur": 1,
    "args": {
      "External id": 176002,
      "cbid": 251, "correlation": 176002
    }
  },
  {
    "ph": "f", "id": 176002, "pid": 76337, "tid": -914061504, "ts": 1716454224091097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091101, "dur": 0,
    "args": {
      "External id": 176003,
      "cbid": 251, "correlation": 176003
    }
  },
  {
    "ph": "f", "id": 176003, "pid": 76337, "tid": -914061504, "ts": 1716454224091101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224170520, "dur": 11,
    "args": {
      "External id": 176004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176004,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 176004, "pid": 5, "tid": 7, "ts": 1716454224170520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091102, "dur": 12,
    "args": {
      "External id": 176004,
      "cbid": 211, "correlation": 176004
    }
  },
  {
    "ph": "s", "id": 176004, "pid": 76337, "tid": -914061504, "ts": 1716454224091102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224170532, "dur": 5,
    "args": {
      "External id": 176006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176006,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 176006, "pid": 5, "tid": 7, "ts": 1716454224170532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091116, "dur": 6,
    "args": {
      "External id": 176006,
      "cbid": 211, "correlation": 176006
    }
  },
  {
    "ph": "s", "id": 176006, "pid": 76337, "tid": -914061504, "ts": 1716454224091116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091177, "dur": 1,
    "args": {
      "External id": 176017,
      "cbid": 251, "correlation": 176017
    }
  },
  {
    "ph": "f", "id": 176017, "pid": 76337, "tid": -914061504, "ts": 1716454224091177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091181, "dur": 0,
    "args": {
      "External id": 176018,
      "cbid": 251, "correlation": 176018
    }
  },
  {
    "ph": "f", "id": 176018, "pid": 76337, "tid": -914061504, "ts": 1716454224091181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224170538, "dur": 7,
    "args": {
      "External id": 176019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176019,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 176019, "pid": 5, "tid": 7, "ts": 1716454224170538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091183, "dur": 12,
    "args": {
      "External id": 176019,
      "cbid": 211, "correlation": 176019
    }
  },
  {
    "ph": "s", "id": 176019, "pid": 76337, "tid": -914061504, "ts": 1716454224091183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224170547, "dur": 3,
    "args": {
      "External id": 176021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176021,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 176021, "pid": 5, "tid": 7, "ts": 1716454224170547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091196, "dur": 5,
    "args": {
      "External id": 176021,
      "cbid": 211, "correlation": 176021
    }
  },
  {
    "ph": "s", "id": 176021, "pid": 76337, "tid": -914061504, "ts": 1716454224091196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224170551, "dur": 91,
    "args": {
      "External id": 176042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176042,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 176042, "pid": 5, "tid": 7, "ts": 1716454224170551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091269, "dur": 14,
    "args": {
      "External id": 176042,
      "cbid": 211, "correlation": 176042
    }
  },
  {
    "ph": "s", "id": 176042, "pid": 76337, "tid": -914061504, "ts": 1716454224091269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091369, "dur": 1,
    "args": {
      "External id": 176060,
      "cbid": 251, "correlation": 176060
    }
  },
  {
    "ph": "f", "id": 176060, "pid": 76337, "tid": -914061504, "ts": 1716454224091369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224170644, "dur": 84,
    "args": {
      "External id": 176062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176062,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176062, "pid": 5, "tid": 7, "ts": 1716454224170644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091375, "dur": 14,
    "args": {
      "External id": 176062,
      "cbid": 211, "correlation": 176062
    }
  },
  {
    "ph": "s", "id": 176062, "pid": 76337, "tid": -914061504, "ts": 1716454224091375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224170729, "dur": 20,
    "args": {
      "External id": 176070,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176070,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176070, "pid": 5, "tid": 7, "ts": 1716454224170729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091448, "dur": 12,
    "args": {
      "External id": 176070,
      "cbid": 211, "correlation": 176070
    }
  },
  {
    "ph": "s", "id": 176070, "pid": 76337, "tid": -914061504, "ts": 1716454224091448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224170750, "dur": 38,
    "args": {
      "External id": 176078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176078,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176078, "pid": 5, "tid": 7, "ts": 1716454224170750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091490, "dur": 9,
    "args": {
      "External id": 176078,
      "cbid": 211, "correlation": 176078
    }
  },
  {
    "ph": "s", "id": 176078, "pid": 76337, "tid": -914061504, "ts": 1716454224091490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224170789, "dur": 35,
    "args": {
      "External id": 176100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176100,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176100, "pid": 5, "tid": 7, "ts": 1716454224170789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091542, "dur": 10,
    "args": {
      "External id": 176100,
      "cbid": 211, "correlation": 176100
    }
  },
  {
    "ph": "s", "id": 176100, "pid": 76337, "tid": -914061504, "ts": 1716454224091542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091638, "dur": 1,
    "args": {
      "External id": 176116,
      "cbid": 251, "correlation": 176116
    }
  },
  {
    "ph": "f", "id": 176116, "pid": 76337, "tid": -914061504, "ts": 1716454224091638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091643, "dur": 0,
    "args": {
      "External id": 176118,
      "cbid": 251, "correlation": 176118
    }
  },
  {
    "ph": "f", "id": 176118, "pid": 76337, "tid": -914061504, "ts": 1716454224091643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224170825, "dur": 539,
    "args": {
      "External id": 176119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176119,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 176119, "pid": 5, "tid": 7, "ts": 1716454224170825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091647, "dur": 13,
    "args": {
      "External id": 176119,
      "cbid": 211, "correlation": 176119
    }
  },
  {
    "ph": "s", "id": 176119, "pid": 76337, "tid": -914061504, "ts": 1716454224091647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224171366, "dur": 126,
    "args": {
      "External id": 176127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176127,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176127, "pid": 5, "tid": 7, "ts": 1716454224171366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091711, "dur": 13,
    "args": {
      "External id": 176127,
      "cbid": 211, "correlation": 176127
    }
  },
  {
    "ph": "s", "id": 176127, "pid": 76337, "tid": -914061504, "ts": 1716454224091711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224171493, "dur": 128,
    "args": {
      "External id": 176135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176135, "pid": 5, "tid": 7, "ts": 1716454224171493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091742, "dur": 8,
    "args": {
      "External id": 176135,
      "cbid": 211, "correlation": 176135
    }
  },
  {
    "ph": "s", "id": 176135, "pid": 76337, "tid": -914061504, "ts": 1716454224091742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091820, "dur": 1,
    "args": {
      "External id": 176151,
      "cbid": 251, "correlation": 176151
    }
  },
  {
    "ph": "f", "id": 176151, "pid": 76337, "tid": -914061504, "ts": 1716454224091820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224171623, "dur": 307,
    "args": {
      "External id": 176153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176153,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176153, "pid": 5, "tid": 7, "ts": 1716454224171623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091826, "dur": 13,
    "args": {
      "External id": 176153,
      "cbid": 211, "correlation": 176153
    }
  },
  {
    "ph": "s", "id": 176153, "pid": 76337, "tid": -914061504, "ts": 1716454224091826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224171931, "dur": 27,
    "args": {
      "External id": 176161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176161,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176161, "pid": 5, "tid": 7, "ts": 1716454224171931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091869, "dur": 10,
    "args": {
      "External id": 176161,
      "cbid": 211, "correlation": 176161
    }
  },
  {
    "ph": "s", "id": 176161, "pid": 76337, "tid": -914061504, "ts": 1716454224091869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224171959, "dur": 81,
    "args": {
      "External id": 176172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176172,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176172, "pid": 5, "tid": 7, "ts": 1716454224171959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224091934, "dur": 13,
    "args": {
      "External id": 176172,
      "cbid": 211, "correlation": 176172
    }
  },
  {
    "ph": "s", "id": 176172, "pid": 76337, "tid": -914061504, "ts": 1716454224091934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092008, "dur": 0,
    "args": {
      "External id": 176184,
      "cbid": 317, "correlation": 176184
    }
  },
  {
    "ph": "f", "id": 176184, "pid": 76337, "tid": -914061504, "ts": 1716454224092008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092009, "dur": 0,
    "args": {
      "External id": 176185,
      "cbid": 203, "correlation": 176185
    }
  },
  {
    "ph": "f", "id": 176185, "pid": 76337, "tid": -914061504, "ts": 1716454224092009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092009, "dur": 0,
    "args": {
      "External id": 176186,
      "cbid": 205, "correlation": 176186
    }
  },
  {
    "ph": "f", "id": 176186, "pid": 76337, "tid": -914061504, "ts": 1716454224092009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224172042, "dur": 22,
    "args": {
      "External id": 176190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176190,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176190, "pid": 5, "tid": 7, "ts": 1716454224172042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092024, "dur": 12,
    "args": {
      "External id": 176190,
      "cbid": 211, "correlation": 176190
    }
  },
  {
    "ph": "s", "id": 176190, "pid": 76337, "tid": -914061504, "ts": 1716454224092024,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224172065, "dur": 119,
    "args": {
      "External id": 176192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176192,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176192, "pid": 5, "tid": 7, "ts": 1716454224172065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092044, "dur": 6,
    "args": {
      "External id": 176192,
      "cbid": 211, "correlation": 176192
    }
  },
  {
    "ph": "s", "id": 176192, "pid": 76337, "tid": -914061504, "ts": 1716454224092044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224172186, "dur": 22,
    "args": {
      "External id": 176194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176194,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176194, "pid": 5, "tid": 7, "ts": 1716454224172186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092054, "dur": 5,
    "args": {
      "External id": 176194,
      "cbid": 211, "correlation": 176194
    }
  },
  {
    "ph": "s", "id": 176194, "pid": 76337, "tid": -914061504, "ts": 1716454224092054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224172210, "dur": 33,
    "args": {
      "External id": 176200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176200, "pid": 5, "tid": 7, "ts": 1716454224172210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092081, "dur": 10,
    "args": {
      "External id": 176200,
      "cbid": 211, "correlation": 176200
    }
  },
  {
    "ph": "s", "id": 176200, "pid": 76337, "tid": -914061504, "ts": 1716454224092081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224172244, "dur": 27,
    "args": {
      "External id": 176208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176208,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176208, "pid": 5, "tid": 7, "ts": 1716454224172244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092113, "dur": 8,
    "args": {
      "External id": 176208,
      "cbid": 211, "correlation": 176208
    }
  },
  {
    "ph": "s", "id": 176208, "pid": 76337, "tid": -914061504, "ts": 1716454224092113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224172272, "dur": 102,
    "args": {
      "External id": 176219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176219,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176219, "pid": 5, "tid": 7, "ts": 1716454224172272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092181, "dur": 12,
    "args": {
      "External id": 176219,
      "cbid": 211, "correlation": 176219
    }
  },
  {
    "ph": "s", "id": 176219, "pid": 76337, "tid": -914061504, "ts": 1716454224092181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092241, "dur": 0,
    "args": {
      "External id": 176229,
      "cbid": 317, "correlation": 176229
    }
  },
  {
    "ph": "f", "id": 176229, "pid": 76337, "tid": -914061504, "ts": 1716454224092241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092242, "dur": 0,
    "args": {
      "External id": 176230,
      "cbid": 203, "correlation": 176230
    }
  },
  {
    "ph": "f", "id": 176230, "pid": 76337, "tid": -914061504, "ts": 1716454224092242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092243, "dur": 0,
    "args": {
      "External id": 176231,
      "cbid": 205, "correlation": 176231
    }
  },
  {
    "ph": "f", "id": 176231, "pid": 76337, "tid": -914061504, "ts": 1716454224092243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224172375, "dur": 74,
    "args": {
      "External id": 176235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176235, "pid": 5, "tid": 7, "ts": 1716454224172375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092258, "dur": 12,
    "args": {
      "External id": 176235,
      "cbid": 211, "correlation": 176235
    }
  },
  {
    "ph": "s", "id": 176235, "pid": 76337, "tid": -914061504, "ts": 1716454224092258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224172450, "dur": 44,
    "args": {
      "External id": 176237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176237, "pid": 5, "tid": 7, "ts": 1716454224172450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092272, "dur": 5,
    "args": {
      "External id": 176237,
      "cbid": 211, "correlation": 176237
    }
  },
  {
    "ph": "s", "id": 176237, "pid": 76337, "tid": -914061504, "ts": 1716454224092272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224172496, "dur": 4,
    "args": {
      "External id": 176239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176239,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176239, "pid": 5, "tid": 7, "ts": 1716454224172496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092283, "dur": 7,
    "args": {
      "External id": 176239,
      "cbid": 211, "correlation": 176239
    }
  },
  {
    "ph": "s", "id": 176239, "pid": 76337, "tid": -914061504, "ts": 1716454224092283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092294, "dur": 0,
    "args": {
      "External id": 176240,
      "cbid": 51, "correlation": 176240
    }
  },
  {
    "ph": "s", "id": 176240, "pid": 76337, "tid": -914061504, "ts": 1716454224092294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224172501, "dur": 2230,
    "args": {
      "External id": 176241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176241,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176241, "pid": 5, "tid": 7, "ts": 1716454224172501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092296, "dur": 5,
    "args": {
      "External id": 176241,
      "cbid": 211, "correlation": 176241
    }
  },
  {
    "ph": "s", "id": 176241, "pid": 76337, "tid": -914061504, "ts": 1716454224092296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224174732, "dur": 112,
    "args": {
      "External id": 176246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176246,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176246, "pid": 5, "tid": 7, "ts": 1716454224174732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092324, "dur": 8,
    "args": {
      "External id": 176246,
      "cbid": 211, "correlation": 176246
    }
  },
  {
    "ph": "s", "id": 176246, "pid": 76337, "tid": -914061504, "ts": 1716454224092324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224174846, "dur": 169,
    "args": {
      "External id": 176255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176255,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176255, "pid": 5, "tid": 7, "ts": 1716454224174846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092442, "dur": 14,
    "args": {
      "External id": 176255,
      "cbid": 211, "correlation": 176255
    }
  },
  {
    "ph": "s", "id": 176255, "pid": 76337, "tid": -914061504, "ts": 1716454224092442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224175016, "dur": 129,
    "args": {
      "External id": 176275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176275,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 176275, "pid": 5, "tid": 7, "ts": 1716454224175016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092518, "dur": 12,
    "args": {
      "External id": 176275,
      "cbid": 211, "correlation": 176275
    }
  },
  {
    "ph": "s", "id": 176275, "pid": 76337, "tid": -914061504, "ts": 1716454224092518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224175147, "dur": 4,
    "args": {
      "External id": 176287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176287,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 176287, "pid": 5, "tid": 7, "ts": 1716454224175147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092539, "dur": 7,
    "args": {
      "External id": 176287,
      "cbid": 211, "correlation": 176287
    }
  },
  {
    "ph": "s", "id": 176287, "pid": 76337, "tid": -914061504, "ts": 1716454224092539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224175152, "dur": 159,
    "args": {
      "External id": 176290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176290, "pid": 5, "tid": 7, "ts": 1716454224175152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092559, "dur": 7,
    "args": {
      "External id": 176290,
      "cbid": 211, "correlation": 176290
    }
  },
  {
    "ph": "s", "id": 176290, "pid": 76337, "tid": -914061504, "ts": 1716454224092559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224175312, "dur": 102,
    "args": {
      "External id": 176299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176299,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176299, "pid": 5, "tid": 7, "ts": 1716454224175312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092600, "dur": 9,
    "args": {
      "External id": 176299,
      "cbid": 211, "correlation": 176299
    }
  },
  {
    "ph": "s", "id": 176299, "pid": 76337, "tid": -914061504, "ts": 1716454224092600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092657, "dur": 0,
    "args": {
      "External id": 176309,
      "cbid": 317, "correlation": 176309
    }
  },
  {
    "ph": "f", "id": 176309, "pid": 76337, "tid": -914061504, "ts": 1716454224092657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092657, "dur": 0,
    "args": {
      "External id": 176310,
      "cbid": 203, "correlation": 176310
    }
  },
  {
    "ph": "f", "id": 176310, "pid": 76337, "tid": -914061504, "ts": 1716454224092657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092658, "dur": 0,
    "args": {
      "External id": 176311,
      "cbid": 205, "correlation": 176311
    }
  },
  {
    "ph": "f", "id": 176311, "pid": 76337, "tid": -914061504, "ts": 1716454224092658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224175416, "dur": 110,
    "args": {
      "External id": 176315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176315,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176315, "pid": 5, "tid": 7, "ts": 1716454224175416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092674, "dur": 11,
    "args": {
      "External id": 176315,
      "cbid": 211, "correlation": 176315
    }
  },
  {
    "ph": "s", "id": 176315, "pid": 76337, "tid": -914061504, "ts": 1716454224092674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224175527, "dur": 34,
    "args": {
      "External id": 176317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176317,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176317, "pid": 5, "tid": 7, "ts": 1716454224175527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092688, "dur": 6,
    "args": {
      "External id": 176317,
      "cbid": 211, "correlation": 176317
    }
  },
  {
    "ph": "s", "id": 176317, "pid": 76337, "tid": -914061504, "ts": 1716454224092688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224175562, "dur": 4,
    "args": {
      "External id": 176319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176319,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176319, "pid": 5, "tid": 7, "ts": 1716454224175562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092699, "dur": 5,
    "args": {
      "External id": 176319,
      "cbid": 211, "correlation": 176319
    }
  },
  {
    "ph": "s", "id": 176319, "pid": 76337, "tid": -914061504, "ts": 1716454224092699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092708, "dur": 0,
    "args": {
      "External id": 176320,
      "cbid": 51, "correlation": 176320
    }
  },
  {
    "ph": "s", "id": 176320, "pid": 76337, "tid": -914061504, "ts": 1716454224092708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224175567, "dur": 2009,
    "args": {
      "External id": 176321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176321,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176321, "pid": 5, "tid": 7, "ts": 1716454224175567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092709, "dur": 6,
    "args": {
      "External id": 176321,
      "cbid": 211, "correlation": 176321
    }
  },
  {
    "ph": "s", "id": 176321, "pid": 76337, "tid": -914061504, "ts": 1716454224092709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224177577, "dur": 58,
    "args": {
      "External id": 176326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176326, "pid": 5, "tid": 7, "ts": 1716454224177577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092739, "dur": 9,
    "args": {
      "External id": 176326,
      "cbid": 211, "correlation": 176326
    }
  },
  {
    "ph": "s", "id": 176326, "pid": 76337, "tid": -914061504, "ts": 1716454224092739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224177637, "dur": 4,
    "args": {
      "External id": 176334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176334,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176334, "pid": 5, "tid": 7, "ts": 1716454224177637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092782, "dur": 10,
    "args": {
      "External id": 176334,
      "cbid": 211, "correlation": 176334
    }
  },
  {
    "ph": "s", "id": 176334, "pid": 76337, "tid": -914061504, "ts": 1716454224092782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092850, "dur": 1,
    "args": {
      "External id": 176350,
      "cbid": 251, "correlation": 176350
    }
  },
  {
    "ph": "f", "id": 176350, "pid": 76337, "tid": -914061504, "ts": 1716454224092850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092855, "dur": 0,
    "args": {
      "External id": 176352,
      "cbid": 251, "correlation": 176352
    }
  },
  {
    "ph": "f", "id": 176352, "pid": 76337, "tid": -914061504, "ts": 1716454224092855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224177643, "dur": 11,
    "args": {
      "External id": 176353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176353,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 176353, "pid": 5, "tid": 7, "ts": 1716454224177643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092857, "dur": 11,
    "args": {
      "External id": 176353,
      "cbid": 211, "correlation": 176353
    }
  },
  {
    "ph": "s", "id": 176353, "pid": 76337, "tid": -914061504, "ts": 1716454224092857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224177655, "dur": 5,
    "args": {
      "External id": 176355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176355,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 176355, "pid": 5, "tid": 7, "ts": 1716454224177655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092870, "dur": 5,
    "args": {
      "External id": 176355,
      "cbid": 211, "correlation": 176355
    }
  },
  {
    "ph": "s", "id": 176355, "pid": 76337, "tid": -914061504, "ts": 1716454224092870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224177662, "dur": 55,
    "args": {
      "External id": 176365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176365,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176365, "pid": 5, "tid": 7, "ts": 1716454224177662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224092927, "dur": 13,
    "args": {
      "External id": 176365,
      "cbid": 211, "correlation": 176365
    }
  },
  {
    "ph": "s", "id": 176365, "pid": 76337, "tid": -914061504, "ts": 1716454224092927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224177718, "dur": 53,
    "args": {
      "External id": 176385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176385,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 176385, "pid": 5, "tid": 7, "ts": 1716454224177718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093009, "dur": 12,
    "args": {
      "External id": 176385,
      "cbid": 211, "correlation": 176385
    }
  },
  {
    "ph": "s", "id": 176385, "pid": 76337, "tid": -914061504, "ts": 1716454224093009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224177772, "dur": 4,
    "args": {
      "External id": 176397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176397,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176397, "pid": 5, "tid": 7, "ts": 1716454224177772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093030, "dur": 6,
    "args": {
      "External id": 176397,
      "cbid": 211, "correlation": 176397
    }
  },
  {
    "ph": "s", "id": 176397, "pid": 76337, "tid": -914061504, "ts": 1716454224093030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224177777, "dur": 55,
    "args": {
      "External id": 176400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176400,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176400, "pid": 5, "tid": 7, "ts": 1716454224177777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093049, "dur": 6,
    "args": {
      "External id": 176400,
      "cbid": 211, "correlation": 176400
    }
  },
  {
    "ph": "s", "id": 176400, "pid": 76337, "tid": -914061504, "ts": 1716454224093049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224177834, "dur": 36,
    "args": {
      "External id": 176409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176409,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176409, "pid": 5, "tid": 7, "ts": 1716454224177834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093091, "dur": 10,
    "args": {
      "External id": 176409,
      "cbid": 211, "correlation": 176409
    }
  },
  {
    "ph": "s", "id": 176409, "pid": 76337, "tid": -914061504, "ts": 1716454224093091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093157, "dur": 0,
    "args": {
      "External id": 176419,
      "cbid": 317, "correlation": 176419
    }
  },
  {
    "ph": "f", "id": 176419, "pid": 76337, "tid": -914061504, "ts": 1716454224093157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093158, "dur": 0,
    "args": {
      "External id": 176420,
      "cbid": 203, "correlation": 176420
    }
  },
  {
    "ph": "f", "id": 176420, "pid": 76337, "tid": -914061504, "ts": 1716454224093158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093159, "dur": 0,
    "args": {
      "External id": 176421,
      "cbid": 205, "correlation": 176421
    }
  },
  {
    "ph": "f", "id": 176421, "pid": 76337, "tid": -914061504, "ts": 1716454224093159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224177871, "dur": 40,
    "args": {
      "External id": 176425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176425,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176425, "pid": 5, "tid": 7, "ts": 1716454224177871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093174, "dur": 12,
    "args": {
      "External id": 176425,
      "cbid": 211, "correlation": 176425
    }
  },
  {
    "ph": "s", "id": 176425, "pid": 76337, "tid": -914061504, "ts": 1716454224093174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224177912, "dur": 14,
    "args": {
      "External id": 176427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176427,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176427, "pid": 5, "tid": 7, "ts": 1716454224177912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093189, "dur": 5,
    "args": {
      "External id": 176427,
      "cbid": 211, "correlation": 176427
    }
  },
  {
    "ph": "s", "id": 176427, "pid": 76337, "tid": -914061504, "ts": 1716454224093189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224177928, "dur": 4,
    "args": {
      "External id": 176429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176429, "pid": 5, "tid": 7, "ts": 1716454224177928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093199, "dur": 5,
    "args": {
      "External id": 176429,
      "cbid": 211, "correlation": 176429
    }
  },
  {
    "ph": "s", "id": 176429, "pid": 76337, "tid": -914061504, "ts": 1716454224093199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093208, "dur": 0,
    "args": {
      "External id": 176430,
      "cbid": 51, "correlation": 176430
    }
  },
  {
    "ph": "s", "id": 176430, "pid": 76337, "tid": -914061504, "ts": 1716454224093208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224177932, "dur": 698,
    "args": {
      "External id": 176431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176431,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176431, "pid": 5, "tid": 7, "ts": 1716454224177932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093208, "dur": 5,
    "args": {
      "External id": 176431,
      "cbid": 211, "correlation": 176431
    }
  },
  {
    "ph": "s", "id": 176431, "pid": 76337, "tid": -914061504, "ts": 1716454224093208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224178632, "dur": 60,
    "args": {
      "External id": 176436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176436,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176436, "pid": 5, "tid": 7, "ts": 1716454224178632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093236, "dur": 9,
    "args": {
      "External id": 176436,
      "cbid": 211, "correlation": 176436
    }
  },
  {
    "ph": "s", "id": 176436, "pid": 76337, "tid": -914061504, "ts": 1716454224093236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093297, "dur": 0,
    "args": {
      "External id": 176446,
      "cbid": 317, "correlation": 176446
    }
  },
  {
    "ph": "f", "id": 176446, "pid": 76337, "tid": -914061504, "ts": 1716454224093297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093298, "dur": 0,
    "args": {
      "External id": 176447,
      "cbid": 203, "correlation": 176447
    }
  },
  {
    "ph": "f", "id": 176447, "pid": 76337, "tid": -914061504, "ts": 1716454224093298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093299, "dur": 0,
    "args": {
      "External id": 176448,
      "cbid": 205, "correlation": 176448
    }
  },
  {
    "ph": "f", "id": 176448, "pid": 76337, "tid": -914061504, "ts": 1716454224093299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224178693, "dur": 4,
    "args": {
      "External id": 176452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176452, "pid": 5, "tid": 7, "ts": 1716454224178693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093314, "dur": 12,
    "args": {
      "External id": 176452,
      "cbid": 211, "correlation": 176452
    }
  },
  {
    "ph": "s", "id": 176452, "pid": 76337, "tid": -914061504, "ts": 1716454224093314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093330, "dur": 0,
    "args": {
      "External id": 176453,
      "cbid": 51, "correlation": 176453
    }
  },
  {
    "ph": "s", "id": 176453, "pid": 76337, "tid": -914061504, "ts": 1716454224093330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224178698, "dur": 266,
    "args": {
      "External id": 176454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176454,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176454, "pid": 5, "tid": 7, "ts": 1716454224178698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093331, "dur": 7,
    "args": {
      "External id": 176454,
      "cbid": 211, "correlation": 176454
    }
  },
  {
    "ph": "s", "id": 176454, "pid": 76337, "tid": -914061504, "ts": 1716454224093331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224178965, "dur": 59,
    "args": {
      "External id": 176459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176459,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176459, "pid": 5, "tid": 7, "ts": 1716454224178965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093359, "dur": 8,
    "args": {
      "External id": 176459,
      "cbid": 211, "correlation": 176459
    }
  },
  {
    "ph": "s", "id": 176459, "pid": 76337, "tid": -914061504, "ts": 1716454224093359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224179025, "dur": 50,
    "args": {
      "External id": 176467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176467,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176467, "pid": 5, "tid": 7, "ts": 1716454224179025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093387, "dur": 8,
    "args": {
      "External id": 176467,
      "cbid": 211, "correlation": 176467
    }
  },
  {
    "ph": "s", "id": 176467, "pid": 76337, "tid": -914061504, "ts": 1716454224093387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224179076, "dur": 36,
    "args": {
      "External id": 176475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176475,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176475, "pid": 5, "tid": 7, "ts": 1716454224179076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093416, "dur": 9,
    "args": {
      "External id": 176475,
      "cbid": 211, "correlation": 176475
    }
  },
  {
    "ph": "s", "id": 176475, "pid": 76337, "tid": -914061504, "ts": 1716454224093416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224179113, "dur": 52,
    "args": {
      "External id": 176495,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176495,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 176495, "pid": 5, "tid": 7, "ts": 1716454224179113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093504, "dur": 13,
    "args": {
      "External id": 176495,
      "cbid": 211, "correlation": 176495
    }
  },
  {
    "ph": "s", "id": 176495, "pid": 76337, "tid": -914061504, "ts": 1716454224093504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224179166, "dur": 4,
    "args": {
      "External id": 176507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176507,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 176507, "pid": 5, "tid": 7, "ts": 1716454224179166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093526, "dur": 6,
    "args": {
      "External id": 176507,
      "cbid": 211, "correlation": 176507
    }
  },
  {
    "ph": "s", "id": 176507, "pid": 76337, "tid": -914061504, "ts": 1716454224093526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224179171, "dur": 55,
    "args": {
      "External id": 176510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176510,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176510, "pid": 5, "tid": 7, "ts": 1716454224179171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093544, "dur": 7,
    "args": {
      "External id": 176510,
      "cbid": 211, "correlation": 176510
    }
  },
  {
    "ph": "s", "id": 176510, "pid": 76337, "tid": -914061504, "ts": 1716454224093544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093604, "dur": 0,
    "args": {
      "External id": 176521,
      "cbid": 317, "correlation": 176521
    }
  },
  {
    "ph": "f", "id": 176521, "pid": 76337, "tid": -914061504, "ts": 1716454224093604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093605, "dur": 0,
    "args": {
      "External id": 176522,
      "cbid": 203, "correlation": 176522
    }
  },
  {
    "ph": "f", "id": 176522, "pid": 76337, "tid": -914061504, "ts": 1716454224093605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093606, "dur": 0,
    "args": {
      "External id": 176523,
      "cbid": 205, "correlation": 176523
    }
  },
  {
    "ph": "f", "id": 176523, "pid": 76337, "tid": -914061504, "ts": 1716454224093606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093636, "dur": 2,
    "args": {
      "External id": 176527,
      "cbid": 251, "correlation": 176527
    }
  },
  {
    "ph": "f", "id": 176527, "pid": 76337, "tid": -914061504, "ts": 1716454224093636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093640, "dur": 1,
    "args": {
      "External id": 176528,
      "cbid": 251, "correlation": 176528
    }
  },
  {
    "ph": "f", "id": 176528, "pid": 76337, "tid": -914061504, "ts": 1716454224093640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093642, "dur": 1,
    "args": {
      "External id": 176529,
      "cbid": 251, "correlation": 176529
    }
  },
  {
    "ph": "f", "id": 176529, "pid": 76337, "tid": -914061504, "ts": 1716454224093642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093644, "dur": 1,
    "args": {
      "External id": 176530,
      "cbid": 251, "correlation": 176530
    }
  },
  {
    "ph": "f", "id": 176530, "pid": 76337, "tid": -914061504, "ts": 1716454224093644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093646, "dur": 1,
    "args": {
      "External id": 176531,
      "cbid": 251, "correlation": 176531
    }
  },
  {
    "ph": "f", "id": 176531, "pid": 76337, "tid": -914061504, "ts": 1716454224093646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093648, "dur": 1,
    "args": {
      "External id": 176532,
      "cbid": 251, "correlation": 176532
    }
  },
  {
    "ph": "f", "id": 176532, "pid": 76337, "tid": -914061504, "ts": 1716454224093648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093650, "dur": 1,
    "args": {
      "External id": 176533,
      "cbid": 251, "correlation": 176533
    }
  },
  {
    "ph": "f", "id": 176533, "pid": 76337, "tid": -914061504, "ts": 1716454224093650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093652, "dur": 1,
    "args": {
      "External id": 176534,
      "cbid": 251, "correlation": 176534
    }
  },
  {
    "ph": "f", "id": 176534, "pid": 76337, "tid": -914061504, "ts": 1716454224093652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093654, "dur": 0,
    "args": {
      "External id": 176535,
      "cbid": 251, "correlation": 176535
    }
  },
  {
    "ph": "f", "id": 176535, "pid": 76337, "tid": -914061504, "ts": 1716454224093654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224179227, "dur": 114,
    "args": {
      "External id": 176536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176536,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 176536, "pid": 5, "tid": 7, "ts": 1716454224179227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093657, "dur": 13,
    "args": {
      "External id": 176536,
      "cbid": 211, "correlation": 176536
    }
  },
  {
    "ph": "s", "id": 176536, "pid": 76337, "tid": -914061504, "ts": 1716454224093657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224179342, "dur": 60,
    "args": {
      "External id": 176542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176542, "pid": 5, "tid": 7, "ts": 1716454224179342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093693, "dur": 10,
    "args": {
      "External id": 176542,
      "cbid": 211, "correlation": 176542
    }
  },
  {
    "ph": "s", "id": 176542, "pid": 76337, "tid": -914061504, "ts": 1716454224093693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224179403, "dur": 587,
    "args": {
      "External id": 176551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176551,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176551, "pid": 5, "tid": 7, "ts": 1716454224179403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093780, "dur": 14,
    "args": {
      "External id": 176551,
      "cbid": 211, "correlation": 176551
    }
  },
  {
    "ph": "s", "id": 176551, "pid": 76337, "tid": -914061504, "ts": 1716454224093780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224179991, "dur": 181,
    "args": {
      "External id": 176573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176573,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176573, "pid": 5, "tid": 7, "ts": 1716454224179991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093842, "dur": 12,
    "args": {
      "External id": 176573,
      "cbid": 211, "correlation": 176573
    }
  },
  {
    "ph": "s", "id": 176573, "pid": 76337, "tid": -914061504, "ts": 1716454224093842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093935, "dur": 1,
    "args": {
      "External id": 176584,
      "cbid": 251, "correlation": 176584
    }
  },
  {
    "ph": "f", "id": 176584, "pid": 76337, "tid": -914061504, "ts": 1716454224093935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224180174, "dur": 197,
    "args": {
      "External id": 176585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176585,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176585, "pid": 5, "tid": 7, "ts": 1716454224180174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224093940, "dur": 14,
    "args": {
      "External id": 176585,
      "cbid": 211, "correlation": 176585
    }
  },
  {
    "ph": "s", "id": 176585, "pid": 76337, "tid": -914061504, "ts": 1716454224093940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094022, "dur": 1,
    "args": {
      "External id": 176596,
      "cbid": 251, "correlation": 176596
    }
  },
  {
    "ph": "f", "id": 176596, "pid": 76337, "tid": -914061504, "ts": 1716454224094022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224180372, "dur": 187,
    "args": {
      "External id": 176597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176597,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176597, "pid": 5, "tid": 7, "ts": 1716454224180372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094027, "dur": 12,
    "args": {
      "External id": 176597,
      "cbid": 211, "correlation": 176597
    }
  },
  {
    "ph": "s", "id": 176597, "pid": 76337, "tid": -914061504, "ts": 1716454224094027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094095, "dur": 1,
    "args": {
      "External id": 176608,
      "cbid": 251, "correlation": 176608
    }
  },
  {
    "ph": "f", "id": 176608, "pid": 76337, "tid": -914061504, "ts": 1716454224094095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224180561, "dur": 185,
    "args": {
      "External id": 176609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176609,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176609, "pid": 5, "tid": 7, "ts": 1716454224180561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094100, "dur": 12,
    "args": {
      "External id": 176609,
      "cbid": 211, "correlation": 176609
    }
  },
  {
    "ph": "s", "id": 176609, "pid": 76337, "tid": -914061504, "ts": 1716454224094100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224180747, "dur": 18572,
    "args": {
      "External id": 176630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176630,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 176630, "pid": 5, "tid": 7, "ts": 1716454224180747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094183, "dur": 14,
    "args": {
      "External id": 176630,
      "cbid": 211, "correlation": 176630
    }
  },
  {
    "ph": "s", "id": 176630, "pid": 76337, "tid": -914061504, "ts": 1716454224094183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094282, "dur": 1,
    "args": {
      "External id": 176648,
      "cbid": 251, "correlation": 176648
    }
  },
  {
    "ph": "f", "id": 176648, "pid": 76337, "tid": -914061504, "ts": 1716454224094282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224199321, "dur": 207,
    "args": {
      "External id": 176650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176650,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176650, "pid": 5, "tid": 7, "ts": 1716454224199321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094288, "dur": 14,
    "args": {
      "External id": 176650,
      "cbid": 211, "correlation": 176650
    }
  },
  {
    "ph": "s", "id": 176650, "pid": 76337, "tid": -914061504, "ts": 1716454224094288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224199529, "dur": 66,
    "args": {
      "External id": 176658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176658,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176658, "pid": 5, "tid": 7, "ts": 1716454224199529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094363, "dur": 12,
    "args": {
      "External id": 176658,
      "cbid": 211, "correlation": 176658
    }
  },
  {
    "ph": "s", "id": 176658, "pid": 76337, "tid": -914061504, "ts": 1716454224094363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224199597, "dur": 96,
    "args": {
      "External id": 176666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176666,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176666, "pid": 5, "tid": 7, "ts": 1716454224199597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094404, "dur": 8,
    "args": {
      "External id": 176666,
      "cbid": 211, "correlation": 176666
    }
  },
  {
    "ph": "s", "id": 176666, "pid": 76337, "tid": -914061504, "ts": 1716454224094404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224199695, "dur": 54,
    "args": {
      "External id": 176677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176677,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176677, "pid": 5, "tid": 7, "ts": 1716454224199695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094479, "dur": 13,
    "args": {
      "External id": 176677,
      "cbid": 211, "correlation": 176677
    }
  },
  {
    "ph": "s", "id": 176677, "pid": 76337, "tid": -914061504, "ts": 1716454224094479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224199750, "dur": 93,
    "args": {
      "External id": 176699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176699,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176699, "pid": 5, "tid": 7, "ts": 1716454224199750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094512, "dur": 8,
    "args": {
      "External id": 176699,
      "cbid": 211, "correlation": 176699
    }
  },
  {
    "ph": "s", "id": 176699, "pid": 76337, "tid": -914061504, "ts": 1716454224094512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094598, "dur": 1,
    "args": {
      "External id": 176710,
      "cbid": 251, "correlation": 176710
    }
  },
  {
    "ph": "f", "id": 176710, "pid": 76337, "tid": -914061504, "ts": 1716454224094598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224199845, "dur": 103,
    "args": {
      "External id": 176711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176711,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176711, "pid": 5, "tid": 7, "ts": 1716454224199845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094604, "dur": 13,
    "args": {
      "External id": 176711,
      "cbid": 211, "correlation": 176711
    }
  },
  {
    "ph": "s", "id": 176711, "pid": 76337, "tid": -914061504, "ts": 1716454224094604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094681, "dur": 1,
    "args": {
      "External id": 176722,
      "cbid": 251, "correlation": 176722
    }
  },
  {
    "ph": "f", "id": 176722, "pid": 76337, "tid": -914061504, "ts": 1716454224094681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094686, "dur": 0,
    "args": {
      "External id": 176723,
      "cbid": 251, "correlation": 176723
    }
  },
  {
    "ph": "f", "id": 176723, "pid": 76337, "tid": -914061504, "ts": 1716454224094686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224199950, "dur": 10,
    "args": {
      "External id": 176724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176724,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 176724, "pid": 5, "tid": 7, "ts": 1716454224199950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094688, "dur": 14,
    "args": {
      "External id": 176724,
      "cbid": 211, "correlation": 176724
    }
  },
  {
    "ph": "s", "id": 176724, "pid": 76337, "tid": -914061504, "ts": 1716454224094688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224199961, "dur": 5,
    "args": {
      "External id": 176726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176726,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 176726, "pid": 5, "tid": 7, "ts": 1716454224199961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094706, "dur": 8,
    "args": {
      "External id": 176726,
      "cbid": 211, "correlation": 176726
    }
  },
  {
    "ph": "s", "id": 176726, "pid": 76337, "tid": -914061504, "ts": 1716454224094706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094771, "dur": 1,
    "args": {
      "External id": 176737,
      "cbid": 251, "correlation": 176737
    }
  },
  {
    "ph": "f", "id": 176737, "pid": 76337, "tid": -914061504, "ts": 1716454224094771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094774, "dur": 0,
    "args": {
      "External id": 176738,
      "cbid": 251, "correlation": 176738
    }
  },
  {
    "ph": "f", "id": 176738, "pid": 76337, "tid": -914061504, "ts": 1716454224094774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224199968, "dur": 6,
    "args": {
      "External id": 176739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176739,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 176739, "pid": 5, "tid": 7, "ts": 1716454224199968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094776, "dur": 12,
    "args": {
      "External id": 176739,
      "cbid": 211, "correlation": 176739
    }
  },
  {
    "ph": "s", "id": 176739, "pid": 76337, "tid": -914061504, "ts": 1716454224094776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224199975, "dur": 3,
    "args": {
      "External id": 176741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176741,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 176741, "pid": 5, "tid": 7, "ts": 1716454224199975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094789, "dur": 5,
    "args": {
      "External id": 176741,
      "cbid": 211, "correlation": 176741
    }
  },
  {
    "ph": "s", "id": 176741, "pid": 76337, "tid": -914061504, "ts": 1716454224094789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224199980, "dur": 156,
    "args": {
      "External id": 176762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176762,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 176762, "pid": 5, "tid": 7, "ts": 1716454224199980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094863, "dur": 12,
    "args": {
      "External id": 176762,
      "cbid": 211, "correlation": 176762
    }
  },
  {
    "ph": "s", "id": 176762, "pid": 76337, "tid": -914061504, "ts": 1716454224094863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094964, "dur": 2,
    "args": {
      "External id": 176780,
      "cbid": 251, "correlation": 176780
    }
  },
  {
    "ph": "f", "id": 176780, "pid": 76337, "tid": -914061504, "ts": 1716454224094964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224200138, "dur": 109,
    "args": {
      "External id": 176782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176782,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 176782, "pid": 5, "tid": 7, "ts": 1716454224200138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224094970, "dur": 21,
    "args": {
      "External id": 176782,
      "cbid": 211, "correlation": 176782
    }
  },
  {
    "ph": "s", "id": 176782, "pid": 76337, "tid": -914061504, "ts": 1716454224094970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224200248, "dur": 34,
    "args": {
      "External id": 176790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176790,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176790, "pid": 5, "tid": 7, "ts": 1716454224200248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095052, "dur": 12,
    "args": {
      "External id": 176790,
      "cbid": 211, "correlation": 176790
    }
  },
  {
    "ph": "s", "id": 176790, "pid": 76337, "tid": -914061504, "ts": 1716454224095052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224200283, "dur": 68,
    "args": {
      "External id": 176798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176798,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176798, "pid": 5, "tid": 7, "ts": 1716454224200283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095093, "dur": 10,
    "args": {
      "External id": 176798,
      "cbid": 211, "correlation": 176798
    }
  },
  {
    "ph": "s", "id": 176798, "pid": 76337, "tid": -914061504, "ts": 1716454224095093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224200353, "dur": 93,
    "args": {
      "External id": 176820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176820,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176820, "pid": 5, "tid": 7, "ts": 1716454224200353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095147, "dur": 11,
    "args": {
      "External id": 176820,
      "cbid": 211, "correlation": 176820
    }
  },
  {
    "ph": "s", "id": 176820, "pid": 76337, "tid": -914061504, "ts": 1716454224095147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095239, "dur": 1,
    "args": {
      "External id": 176836,
      "cbid": 251, "correlation": 176836
    }
  },
  {
    "ph": "f", "id": 176836, "pid": 76337, "tid": -914061504, "ts": 1716454224095239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224200446, "dur": 575,
    "args": {
      "External id": 176838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176838,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 176838, "pid": 5, "tid": 7, "ts": 1716454224200446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095245, "dur": 13,
    "args": {
      "External id": 176838,
      "cbid": 211, "correlation": 176838
    }
  },
  {
    "ph": "s", "id": 176838, "pid": 76337, "tid": -914061504, "ts": 1716454224095245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224201023, "dur": 244,
    "args": {
      "External id": 176846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176846, "pid": 5, "tid": 7, "ts": 1716454224201023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095311, "dur": 13,
    "args": {
      "External id": 176846,
      "cbid": 211, "correlation": 176846
    }
  },
  {
    "ph": "s", "id": 176846, "pid": 76337, "tid": -914061504, "ts": 1716454224095311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224201269, "dur": 253,
    "args": {
      "External id": 176854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176854, "pid": 5, "tid": 7, "ts": 1716454224201269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095342, "dur": 9,
    "args": {
      "External id": 176854,
      "cbid": 211, "correlation": 176854
    }
  },
  {
    "ph": "s", "id": 176854, "pid": 76337, "tid": -914061504, "ts": 1716454224095342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095429, "dur": 1,
    "args": {
      "External id": 176870,
      "cbid": 251, "correlation": 176870
    }
  },
  {
    "ph": "f", "id": 176870, "pid": 76337, "tid": -914061504, "ts": 1716454224095429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095434, "dur": 0,
    "args": {
      "External id": 176872,
      "cbid": 251, "correlation": 176872
    }
  },
  {
    "ph": "f", "id": 176872, "pid": 76337, "tid": -914061504, "ts": 1716454224095434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224201523, "dur": 357,
    "args": {
      "External id": 176873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176873,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 176873, "pid": 5, "tid": 7, "ts": 1716454224201523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095437, "dur": 13,
    "args": {
      "External id": 176873,
      "cbid": 211, "correlation": 176873
    }
  },
  {
    "ph": "s", "id": 176873, "pid": 76337, "tid": -914061504, "ts": 1716454224095437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224201881, "dur": 50,
    "args": {
      "External id": 176881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176881,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176881, "pid": 5, "tid": 7, "ts": 1716454224201881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095479, "dur": 10,
    "args": {
      "External id": 176881,
      "cbid": 211, "correlation": 176881
    }
  },
  {
    "ph": "s", "id": 176881, "pid": 76337, "tid": -914061504, "ts": 1716454224095479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224201933, "dur": 160,
    "args": {
      "External id": 176892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176892, "pid": 5, "tid": 7, "ts": 1716454224201933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095546, "dur": 13,
    "args": {
      "External id": 176892,
      "cbid": 211, "correlation": 176892
    }
  },
  {
    "ph": "s", "id": 176892, "pid": 76337, "tid": -914061504, "ts": 1716454224095546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095613, "dur": 0,
    "args": {
      "External id": 176904,
      "cbid": 317, "correlation": 176904
    }
  },
  {
    "ph": "f", "id": 176904, "pid": 76337, "tid": -914061504, "ts": 1716454224095613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095614, "dur": 0,
    "args": {
      "External id": 176905,
      "cbid": 203, "correlation": 176905
    }
  },
  {
    "ph": "f", "id": 176905, "pid": 76337, "tid": -914061504, "ts": 1716454224095614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095615, "dur": 0,
    "args": {
      "External id": 176906,
      "cbid": 205, "correlation": 176906
    }
  },
  {
    "ph": "f", "id": 176906, "pid": 76337, "tid": -914061504, "ts": 1716454224095615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095639, "dur": 1,
    "args": {
      "External id": 176910,
      "cbid": 251, "correlation": 176910
    }
  },
  {
    "ph": "f", "id": 176910, "pid": 76337, "tid": -914061504, "ts": 1716454224095639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095641, "dur": 0,
    "args": {
      "External id": 176911,
      "cbid": 251, "correlation": 176911
    }
  },
  {
    "ph": "f", "id": 176911, "pid": 76337, "tid": -914061504, "ts": 1716454224095641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095641, "dur": 0,
    "args": {
      "External id": 176912,
      "cbid": 251, "correlation": 176912
    }
  },
  {
    "ph": "f", "id": 176912, "pid": 76337, "tid": -914061504, "ts": 1716454224095641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095643, "dur": 0,
    "args": {
      "External id": 176913,
      "cbid": 251, "correlation": 176913
    }
  },
  {
    "ph": "f", "id": 176913, "pid": 76337, "tid": -914061504, "ts": 1716454224095643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095644, "dur": 0,
    "args": {
      "External id": 176914,
      "cbid": 251, "correlation": 176914
    }
  },
  {
    "ph": "f", "id": 176914, "pid": 76337, "tid": -914061504, "ts": 1716454224095644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095645, "dur": 0,
    "args": {
      "External id": 176915,
      "cbid": 251, "correlation": 176915
    }
  },
  {
    "ph": "f", "id": 176915, "pid": 76337, "tid": -914061504, "ts": 1716454224095645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095646, "dur": 0,
    "args": {
      "External id": 176916,
      "cbid": 251, "correlation": 176916
    }
  },
  {
    "ph": "f", "id": 176916, "pid": 76337, "tid": -914061504, "ts": 1716454224095646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095647, "dur": 0,
    "args": {
      "External id": 176917,
      "cbid": 251, "correlation": 176917
    }
  },
  {
    "ph": "f", "id": 176917, "pid": 76337, "tid": -914061504, "ts": 1716454224095647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095649, "dur": 0,
    "args": {
      "External id": 176918,
      "cbid": 251, "correlation": 176918
    }
  },
  {
    "ph": "f", "id": 176918, "pid": 76337, "tid": -914061504, "ts": 1716454224095649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224202094, "dur": 115,
    "args": {
      "External id": 176919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176919,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 176919, "pid": 5, "tid": 7, "ts": 1716454224202094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095651, "dur": 12,
    "args": {
      "External id": 176919,
      "cbid": 211, "correlation": 176919
    }
  },
  {
    "ph": "s", "id": 176919, "pid": 76337, "tid": -914061504, "ts": 1716454224095651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224202210, "dur": 60,
    "args": {
      "External id": 176925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176925, "pid": 5, "tid": 7, "ts": 1716454224202210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095685, "dur": 9,
    "args": {
      "External id": 176925,
      "cbid": 211, "correlation": 176925
    }
  },
  {
    "ph": "s", "id": 176925, "pid": 76337, "tid": -914061504, "ts": 1716454224095685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224202271, "dur": 50,
    "args": {
      "External id": 176933,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176933,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176933, "pid": 5, "tid": 7, "ts": 1716454224202271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095716, "dur": 8,
    "args": {
      "External id": 176933,
      "cbid": 211, "correlation": 176933
    }
  },
  {
    "ph": "s", "id": 176933, "pid": 76337, "tid": -914061504, "ts": 1716454224095716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224202322, "dur": 98,
    "args": {
      "External id": 176942,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176942,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176942, "pid": 5, "tid": 7, "ts": 1716454224202322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095755, "dur": 11,
    "args": {
      "External id": 176942,
      "cbid": 211, "correlation": 176942
    }
  },
  {
    "ph": "s", "id": 176942, "pid": 76337, "tid": -914061504, "ts": 1716454224095755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224202421, "dur": 92,
    "args": {
      "External id": 176962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176962,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 176962, "pid": 5, "tid": 7, "ts": 1716454224202421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095832, "dur": 12,
    "args": {
      "External id": 176962,
      "cbid": 211, "correlation": 176962
    }
  },
  {
    "ph": "s", "id": 176962, "pid": 76337, "tid": -914061504, "ts": 1716454224095832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224202514, "dur": 4,
    "args": {
      "External id": 176974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176974,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 176974, "pid": 5, "tid": 7, "ts": 1716454224202514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095855, "dur": 7,
    "args": {
      "External id": 176974,
      "cbid": 211, "correlation": 176974
    }
  },
  {
    "ph": "s", "id": 176974, "pid": 76337, "tid": -914061504, "ts": 1716454224095855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224202520, "dur": 108,
    "args": {
      "External id": 176977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176977,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176977, "pid": 5, "tid": 7, "ts": 1716454224202520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095874, "dur": 7,
    "args": {
      "External id": 176977,
      "cbid": 211, "correlation": 176977
    }
  },
  {
    "ph": "s", "id": 176977, "pid": 76337, "tid": -914061504, "ts": 1716454224095874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224202630, "dur": 69,
    "args": {
      "External id": 176986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 176986,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 176986, "pid": 5, "tid": 7, "ts": 1716454224202630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095914, "dur": 9,
    "args": {
      "External id": 176986,
      "cbid": 211, "correlation": 176986
    }
  },
  {
    "ph": "s", "id": 176986, "pid": 76337, "tid": -914061504, "ts": 1716454224095914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095967, "dur": 0,
    "args": {
      "External id": 176996,
      "cbid": 317, "correlation": 176996
    }
  },
  {
    "ph": "f", "id": 176996, "pid": 76337, "tid": -914061504, "ts": 1716454224095967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095968, "dur": 0,
    "args": {
      "External id": 176997,
      "cbid": 203, "correlation": 176997
    }
  },
  {
    "ph": "f", "id": 176997, "pid": 76337, "tid": -914061504, "ts": 1716454224095968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095969, "dur": 0,
    "args": {
      "External id": 176998,
      "cbid": 205, "correlation": 176998
    }
  },
  {
    "ph": "f", "id": 176998, "pid": 76337, "tid": -914061504, "ts": 1716454224095969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224202700, "dur": 76,
    "args": {
      "External id": 177002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177002,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177002, "pid": 5, "tid": 7, "ts": 1716454224202700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224095996, "dur": 13,
    "args": {
      "External id": 177002,
      "cbid": 211, "correlation": 177002
    }
  },
  {
    "ph": "s", "id": 177002, "pid": 76337, "tid": -914061504, "ts": 1716454224095996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224202777, "dur": 24,
    "args": {
      "External id": 177004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177004,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177004, "pid": 5, "tid": 7, "ts": 1716454224202777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096012, "dur": 5,
    "args": {
      "External id": 177004,
      "cbid": 211, "correlation": 177004
    }
  },
  {
    "ph": "s", "id": 177004, "pid": 76337, "tid": -914061504, "ts": 1716454224096012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224202803, "dur": 4,
    "args": {
      "External id": 177006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177006,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177006, "pid": 5, "tid": 7, "ts": 1716454224202803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096021, "dur": 6,
    "args": {
      "External id": 177006,
      "cbid": 211, "correlation": 177006
    }
  },
  {
    "ph": "s", "id": 177006, "pid": 76337, "tid": -914061504, "ts": 1716454224096021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096030, "dur": 0,
    "args": {
      "External id": 177007,
      "cbid": 51, "correlation": 177007
    }
  },
  {
    "ph": "s", "id": 177007, "pid": 76337, "tid": -914061504, "ts": 1716454224096030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224202809, "dur": 1371,
    "args": {
      "External id": 177008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177008,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177008, "pid": 5, "tid": 7, "ts": 1716454224202809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096031, "dur": 5,
    "args": {
      "External id": 177008,
      "cbid": 211, "correlation": 177008
    }
  },
  {
    "ph": "s", "id": 177008, "pid": 76337, "tid": -914061504, "ts": 1716454224096031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224204181, "dur": 59,
    "args": {
      "External id": 177013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177013,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177013, "pid": 5, "tid": 7, "ts": 1716454224204181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096059, "dur": 9,
    "args": {
      "External id": 177013,
      "cbid": 211, "correlation": 177013
    }
  },
  {
    "ph": "s", "id": 177013, "pid": 76337, "tid": -914061504, "ts": 1716454224096059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224204241, "dur": 4,
    "args": {
      "External id": 177021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177021,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177021, "pid": 5, "tid": 7, "ts": 1716454224204241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096105, "dur": 9,
    "args": {
      "External id": 177021,
      "cbid": 211, "correlation": 177021
    }
  },
  {
    "ph": "s", "id": 177021, "pid": 76337, "tid": -914061504, "ts": 1716454224096105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096173, "dur": 1,
    "args": {
      "External id": 177037,
      "cbid": 251, "correlation": 177037
    }
  },
  {
    "ph": "f", "id": 177037, "pid": 76337, "tid": -914061504, "ts": 1716454224096173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096178, "dur": 0,
    "args": {
      "External id": 177039,
      "cbid": 251, "correlation": 177039
    }
  },
  {
    "ph": "f", "id": 177039, "pid": 76337, "tid": -914061504, "ts": 1716454224096178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224204247, "dur": 11,
    "args": {
      "External id": 177040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177040,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 177040, "pid": 5, "tid": 7, "ts": 1716454224204247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096180, "dur": 12,
    "args": {
      "External id": 177040,
      "cbid": 211, "correlation": 177040
    }
  },
  {
    "ph": "s", "id": 177040, "pid": 76337, "tid": -914061504, "ts": 1716454224096180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224204259, "dur": 5,
    "args": {
      "External id": 177042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177042,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 177042, "pid": 5, "tid": 7, "ts": 1716454224204259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096193, "dur": 5,
    "args": {
      "External id": 177042,
      "cbid": 211, "correlation": 177042
    }
  },
  {
    "ph": "s", "id": 177042, "pid": 76337, "tid": -914061504, "ts": 1716454224096193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224204266, "dur": 55,
    "args": {
      "External id": 177052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177052,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177052, "pid": 5, "tid": 7, "ts": 1716454224204266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096250, "dur": 12,
    "args": {
      "External id": 177052,
      "cbid": 211, "correlation": 177052
    }
  },
  {
    "ph": "s", "id": 177052, "pid": 76337, "tid": -914061504, "ts": 1716454224096250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224204322, "dur": 51,
    "args": {
      "External id": 177072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177072,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 177072, "pid": 5, "tid": 7, "ts": 1716454224204322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096319, "dur": 11,
    "args": {
      "External id": 177072,
      "cbid": 211, "correlation": 177072
    }
  },
  {
    "ph": "s", "id": 177072, "pid": 76337, "tid": -914061504, "ts": 1716454224096319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224204375, "dur": 4,
    "args": {
      "External id": 177084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177084,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177084, "pid": 5, "tid": 7, "ts": 1716454224204375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096340, "dur": 6,
    "args": {
      "External id": 177084,
      "cbid": 211, "correlation": 177084
    }
  },
  {
    "ph": "s", "id": 177084, "pid": 76337, "tid": -914061504, "ts": 1716454224096340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224204380, "dur": 55,
    "args": {
      "External id": 177087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177087,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177087, "pid": 5, "tid": 7, "ts": 1716454224204380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096359, "dur": 6,
    "args": {
      "External id": 177087,
      "cbid": 211, "correlation": 177087
    }
  },
  {
    "ph": "s", "id": 177087, "pid": 76337, "tid": -914061504, "ts": 1716454224096359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224204436, "dur": 36,
    "args": {
      "External id": 177096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177096,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177096, "pid": 5, "tid": 7, "ts": 1716454224204436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096398, "dur": 11,
    "args": {
      "External id": 177096,
      "cbid": 211, "correlation": 177096
    }
  },
  {
    "ph": "s", "id": 177096, "pid": 76337, "tid": -914061504, "ts": 1716454224096398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096465, "dur": 0,
    "args": {
      "External id": 177106,
      "cbid": 317, "correlation": 177106
    }
  },
  {
    "ph": "f", "id": 177106, "pid": 76337, "tid": -914061504, "ts": 1716454224096465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096466, "dur": 0,
    "args": {
      "External id": 177107,
      "cbid": 203, "correlation": 177107
    }
  },
  {
    "ph": "f", "id": 177107, "pid": 76337, "tid": -914061504, "ts": 1716454224096466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096467, "dur": 0,
    "args": {
      "External id": 177108,
      "cbid": 205, "correlation": 177108
    }
  },
  {
    "ph": "f", "id": 177108, "pid": 76337, "tid": -914061504, "ts": 1716454224096467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224204474, "dur": 40,
    "args": {
      "External id": 177112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177112,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177112, "pid": 5, "tid": 7, "ts": 1716454224204474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096481, "dur": 12,
    "args": {
      "External id": 177112,
      "cbid": 211, "correlation": 177112
    }
  },
  {
    "ph": "s", "id": 177112, "pid": 76337, "tid": -914061504, "ts": 1716454224096481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224204515, "dur": 14,
    "args": {
      "External id": 177114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177114,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177114, "pid": 5, "tid": 7, "ts": 1716454224204515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096496, "dur": 5,
    "args": {
      "External id": 177114,
      "cbid": 211, "correlation": 177114
    }
  },
  {
    "ph": "s", "id": 177114, "pid": 76337, "tid": -914061504, "ts": 1716454224096496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224204531, "dur": 3,
    "args": {
      "External id": 177116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177116,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177116, "pid": 5, "tid": 7, "ts": 1716454224204531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096505, "dur": 5,
    "args": {
      "External id": 177116,
      "cbid": 211, "correlation": 177116
    }
  },
  {
    "ph": "s", "id": 177116, "pid": 76337, "tid": -914061504, "ts": 1716454224096505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096513, "dur": 0,
    "args": {
      "External id": 177117,
      "cbid": 51, "correlation": 177117
    }
  },
  {
    "ph": "s", "id": 177117, "pid": 76337, "tid": -914061504, "ts": 1716454224096513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224204535, "dur": 702,
    "args": {
      "External id": 177118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177118,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177118, "pid": 5, "tid": 7, "ts": 1716454224204535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096514, "dur": 5,
    "args": {
      "External id": 177118,
      "cbid": 211, "correlation": 177118
    }
  },
  {
    "ph": "s", "id": 177118, "pid": 76337, "tid": -914061504, "ts": 1716454224096514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224205238, "dur": 60,
    "args": {
      "External id": 177123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177123, "pid": 5, "tid": 7, "ts": 1716454224205238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096541, "dur": 8,
    "args": {
      "External id": 177123,
      "cbid": 211, "correlation": 177123
    }
  },
  {
    "ph": "s", "id": 177123, "pid": 76337, "tid": -914061504, "ts": 1716454224096541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096598, "dur": 0,
    "args": {
      "External id": 177133,
      "cbid": 317, "correlation": 177133
    }
  },
  {
    "ph": "f", "id": 177133, "pid": 76337, "tid": -914061504, "ts": 1716454224096598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096599, "dur": 0,
    "args": {
      "External id": 177134,
      "cbid": 203, "correlation": 177134
    }
  },
  {
    "ph": "f", "id": 177134, "pid": 76337, "tid": -914061504, "ts": 1716454224096599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096599, "dur": 0,
    "args": {
      "External id": 177135,
      "cbid": 205, "correlation": 177135
    }
  },
  {
    "ph": "f", "id": 177135, "pid": 76337, "tid": -914061504, "ts": 1716454224096599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224205299, "dur": 75,
    "args": {
      "External id": 177139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177139,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177139, "pid": 5, "tid": 7, "ts": 1716454224205299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096613, "dur": 11,
    "args": {
      "External id": 177139,
      "cbid": 211, "correlation": 177139
    }
  },
  {
    "ph": "s", "id": 177139, "pid": 76337, "tid": -914061504, "ts": 1716454224096613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224205375, "dur": 209,
    "args": {
      "External id": 177141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177141,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177141, "pid": 5, "tid": 7, "ts": 1716454224205375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096632, "dur": 8,
    "args": {
      "External id": 177141,
      "cbid": 211, "correlation": 177141
    }
  },
  {
    "ph": "s", "id": 177141, "pid": 76337, "tid": -914061504, "ts": 1716454224096632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224205586, "dur": 39,
    "args": {
      "External id": 177143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177143,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177143, "pid": 5, "tid": 7, "ts": 1716454224205586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096644, "dur": 6,
    "args": {
      "External id": 177143,
      "cbid": 211, "correlation": 177143
    }
  },
  {
    "ph": "s", "id": 177143, "pid": 76337, "tid": -914061504, "ts": 1716454224096644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224205626, "dur": 59,
    "args": {
      "External id": 177149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177149,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177149, "pid": 5, "tid": 7, "ts": 1716454224205626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096671, "dur": 8,
    "args": {
      "External id": 177149,
      "cbid": 211, "correlation": 177149
    }
  },
  {
    "ph": "s", "id": 177149, "pid": 76337, "tid": -914061504, "ts": 1716454224096671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224205687, "dur": 50,
    "args": {
      "External id": 177157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177157,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177157, "pid": 5, "tid": 7, "ts": 1716454224205687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096700, "dur": 9,
    "args": {
      "External id": 177157,
      "cbid": 211, "correlation": 177157
    }
  },
  {
    "ph": "s", "id": 177157, "pid": 76337, "tid": -914061504, "ts": 1716454224096700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224205738, "dur": 35,
    "args": {
      "External id": 177165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177165,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177165, "pid": 5, "tid": 7, "ts": 1716454224205738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096729, "dur": 8,
    "args": {
      "External id": 177165,
      "cbid": 211, "correlation": 177165
    }
  },
  {
    "ph": "s", "id": 177165, "pid": 76337, "tid": -914061504, "ts": 1716454224096729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224205775, "dur": 53,
    "args": {
      "External id": 177185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177185,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 177185, "pid": 5, "tid": 7, "ts": 1716454224205775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096817, "dur": 13,
    "args": {
      "External id": 177185,
      "cbid": 211, "correlation": 177185
    }
  },
  {
    "ph": "s", "id": 177185, "pid": 76337, "tid": -914061504, "ts": 1716454224096817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224205829, "dur": 5,
    "args": {
      "External id": 177197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177197,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177197, "pid": 5, "tid": 7, "ts": 1716454224205829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096839, "dur": 7,
    "args": {
      "External id": 177197,
      "cbid": 211, "correlation": 177197
    }
  },
  {
    "ph": "s", "id": 177197, "pid": 76337, "tid": -914061504, "ts": 1716454224096839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224205835, "dur": 56,
    "args": {
      "External id": 177200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177200,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177200, "pid": 5, "tid": 7, "ts": 1716454224205835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096857, "dur": 7,
    "args": {
      "External id": 177200,
      "cbid": 211, "correlation": 177200
    }
  },
  {
    "ph": "s", "id": 177200, "pid": 76337, "tid": -914061504, "ts": 1716454224096857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096919, "dur": 0,
    "args": {
      "External id": 177211,
      "cbid": 317, "correlation": 177211
    }
  },
  {
    "ph": "f", "id": 177211, "pid": 76337, "tid": -914061504, "ts": 1716454224096919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096920, "dur": 0,
    "args": {
      "External id": 177212,
      "cbid": 203, "correlation": 177212
    }
  },
  {
    "ph": "f", "id": 177212, "pid": 76337, "tid": -914061504, "ts": 1716454224096920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096920, "dur": 0,
    "args": {
      "External id": 177213,
      "cbid": 205, "correlation": 177213
    }
  },
  {
    "ph": "f", "id": 177213, "pid": 76337, "tid": -914061504, "ts": 1716454224096920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096942, "dur": 1,
    "args": {
      "External id": 177217,
      "cbid": 251, "correlation": 177217
    }
  },
  {
    "ph": "f", "id": 177217, "pid": 76337, "tid": -914061504, "ts": 1716454224096942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096944, "dur": 0,
    "args": {
      "External id": 177218,
      "cbid": 251, "correlation": 177218
    }
  },
  {
    "ph": "f", "id": 177218, "pid": 76337, "tid": -914061504, "ts": 1716454224096944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096945, "dur": 0,
    "args": {
      "External id": 177219,
      "cbid": 251, "correlation": 177219
    }
  },
  {
    "ph": "f", "id": 177219, "pid": 76337, "tid": -914061504, "ts": 1716454224096945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096946, "dur": 0,
    "args": {
      "External id": 177220,
      "cbid": 251, "correlation": 177220
    }
  },
  {
    "ph": "f", "id": 177220, "pid": 76337, "tid": -914061504, "ts": 1716454224096946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096946, "dur": 0,
    "args": {
      "External id": 177221,
      "cbid": 251, "correlation": 177221
    }
  },
  {
    "ph": "f", "id": 177221, "pid": 76337, "tid": -914061504, "ts": 1716454224096946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096947, "dur": 0,
    "args": {
      "External id": 177222,
      "cbid": 251, "correlation": 177222
    }
  },
  {
    "ph": "f", "id": 177222, "pid": 76337, "tid": -914061504, "ts": 1716454224096947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096948, "dur": 0,
    "args": {
      "External id": 177223,
      "cbid": 251, "correlation": 177223
    }
  },
  {
    "ph": "f", "id": 177223, "pid": 76337, "tid": -914061504, "ts": 1716454224096948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096949, "dur": 0,
    "args": {
      "External id": 177224,
      "cbid": 251, "correlation": 177224
    }
  },
  {
    "ph": "f", "id": 177224, "pid": 76337, "tid": -914061504, "ts": 1716454224096949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096950, "dur": 0,
    "args": {
      "External id": 177225,
      "cbid": 251, "correlation": 177225
    }
  },
  {
    "ph": "f", "id": 177225, "pid": 76337, "tid": -914061504, "ts": 1716454224096950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224205892, "dur": 113,
    "args": {
      "External id": 177226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177226,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 177226, "pid": 5, "tid": 7, "ts": 1716454224205892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096952, "dur": 13,
    "args": {
      "External id": 177226,
      "cbid": 211, "correlation": 177226
    }
  },
  {
    "ph": "s", "id": 177226, "pid": 76337, "tid": -914061504, "ts": 1716454224096952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224206006, "dur": 60,
    "args": {
      "External id": 177232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177232,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177232, "pid": 5, "tid": 7, "ts": 1716454224206006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224096996, "dur": 9,
    "args": {
      "External id": 177232,
      "cbid": 211, "correlation": 177232
    }
  },
  {
    "ph": "s", "id": 177232, "pid": 76337, "tid": -914061504, "ts": 1716454224096996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224206067, "dur": 582,
    "args": {
      "External id": 177241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177241,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177241, "pid": 5, "tid": 7, "ts": 1716454224206067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097082, "dur": 14,
    "args": {
      "External id": 177241,
      "cbid": 211, "correlation": 177241
    }
  },
  {
    "ph": "s", "id": 177241, "pid": 76337, "tid": -914061504, "ts": 1716454224097082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224206651, "dur": 183,
    "args": {
      "External id": 177263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177263,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177263, "pid": 5, "tid": 7, "ts": 1716454224206651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097142, "dur": 11,
    "args": {
      "External id": 177263,
      "cbid": 211, "correlation": 177263
    }
  },
  {
    "ph": "s", "id": 177263, "pid": 76337, "tid": -914061504, "ts": 1716454224097142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097232, "dur": 1,
    "args": {
      "External id": 177274,
      "cbid": 251, "correlation": 177274
    }
  },
  {
    "ph": "f", "id": 177274, "pid": 76337, "tid": -914061504, "ts": 1716454224097232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224206835, "dur": 199,
    "args": {
      "External id": 177275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177275,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177275, "pid": 5, "tid": 7, "ts": 1716454224206835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097237, "dur": 13,
    "args": {
      "External id": 177275,
      "cbid": 211, "correlation": 177275
    }
  },
  {
    "ph": "s", "id": 177275, "pid": 76337, "tid": -914061504, "ts": 1716454224097237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097309, "dur": 1,
    "args": {
      "External id": 177286,
      "cbid": 251, "correlation": 177286
    }
  },
  {
    "ph": "f", "id": 177286, "pid": 76337, "tid": -914061504, "ts": 1716454224097309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224207036, "dur": 191,
    "args": {
      "External id": 177287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177287,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177287, "pid": 5, "tid": 7, "ts": 1716454224207036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097313, "dur": 12,
    "args": {
      "External id": 177287,
      "cbid": 211, "correlation": 177287
    }
  },
  {
    "ph": "s", "id": 177287, "pid": 76337, "tid": -914061504, "ts": 1716454224097313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097379, "dur": 1,
    "args": {
      "External id": 177298,
      "cbid": 251, "correlation": 177298
    }
  },
  {
    "ph": "f", "id": 177298, "pid": 76337, "tid": -914061504, "ts": 1716454224097379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224207228, "dur": 192,
    "args": {
      "External id": 177299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177299,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177299, "pid": 5, "tid": 7, "ts": 1716454224207228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097383, "dur": 11,
    "args": {
      "External id": 177299,
      "cbid": 211, "correlation": 177299
    }
  },
  {
    "ph": "s", "id": 177299, "pid": 76337, "tid": -914061504, "ts": 1716454224097383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224207421, "dur": 18792,
    "args": {
      "External id": 177320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177320,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 177320, "pid": 5, "tid": 7, "ts": 1716454224207421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097465, "dur": 12,
    "args": {
      "External id": 177320,
      "cbid": 211, "correlation": 177320
    }
  },
  {
    "ph": "s", "id": 177320, "pid": 76337, "tid": -914061504, "ts": 1716454224097465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097565, "dur": 1,
    "args": {
      "External id": 177338,
      "cbid": 251, "correlation": 177338
    }
  },
  {
    "ph": "f", "id": 177338, "pid": 76337, "tid": -914061504, "ts": 1716454224097565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224226214, "dur": 203,
    "args": {
      "External id": 177340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177340,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177340, "pid": 5, "tid": 7, "ts": 1716454224226214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097571, "dur": 13,
    "args": {
      "External id": 177340,
      "cbid": 211, "correlation": 177340
    }
  },
  {
    "ph": "s", "id": 177340, "pid": 76337, "tid": -914061504, "ts": 1716454224097571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224226418, "dur": 66,
    "args": {
      "External id": 177348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177348,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177348, "pid": 5, "tid": 7, "ts": 1716454224226418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097643, "dur": 12,
    "args": {
      "External id": 177348,
      "cbid": 211, "correlation": 177348
    }
  },
  {
    "ph": "s", "id": 177348, "pid": 76337, "tid": -914061504, "ts": 1716454224097643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224226485, "dur": 97,
    "args": {
      "External id": 177356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177356,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177356, "pid": 5, "tid": 7, "ts": 1716454224226485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097683, "dur": 9,
    "args": {
      "External id": 177356,
      "cbid": 211, "correlation": 177356
    }
  },
  {
    "ph": "s", "id": 177356, "pid": 76337, "tid": -914061504, "ts": 1716454224097683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224226584, "dur": 55,
    "args": {
      "External id": 177367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177367,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177367, "pid": 5, "tid": 7, "ts": 1716454224226584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097757, "dur": 13,
    "args": {
      "External id": 177367,
      "cbid": 211, "correlation": 177367
    }
  },
  {
    "ph": "s", "id": 177367, "pid": 76337, "tid": -914061504, "ts": 1716454224097757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224226640, "dur": 92,
    "args": {
      "External id": 177389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177389,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177389, "pid": 5, "tid": 7, "ts": 1716454224226640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224097789, "dur": 1166,
    "args": {
      "External id": 177389,
      "cbid": 211, "correlation": 177389
    }
  },
  {
    "ph": "s", "id": 177389, "pid": 76337, "tid": -914061504, "ts": 1716454224097789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099044, "dur": 1,
    "args": {
      "External id": 177400,
      "cbid": 251, "correlation": 177400
    }
  },
  {
    "ph": "f", "id": 177400, "pid": 76337, "tid": -914061504, "ts": 1716454224099044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224226733, "dur": 106,
    "args": {
      "External id": 177401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177401,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177401, "pid": 5, "tid": 7, "ts": 1716454224226733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099049, "dur": 58,
    "args": {
      "External id": 177401,
      "cbid": 211, "correlation": 177401
    }
  },
  {
    "ph": "s", "id": 177401, "pid": 76337, "tid": -914061504, "ts": 1716454224099049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099168, "dur": 1,
    "args": {
      "External id": 177412,
      "cbid": 251, "correlation": 177412
    }
  },
  {
    "ph": "f", "id": 177412, "pid": 76337, "tid": -914061504, "ts": 1716454224099168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099172, "dur": 0,
    "args": {
      "External id": 177413,
      "cbid": 251, "correlation": 177413
    }
  },
  {
    "ph": "f", "id": 177413, "pid": 76337, "tid": -914061504, "ts": 1716454224099172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224226841, "dur": 10,
    "args": {
      "External id": 177414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177414,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 177414, "pid": 5, "tid": 7, "ts": 1716454224226841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099174, "dur": 13,
    "args": {
      "External id": 177414,
      "cbid": 211, "correlation": 177414
    }
  },
  {
    "ph": "s", "id": 177414, "pid": 76337, "tid": -914061504, "ts": 1716454224099174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224226852, "dur": 5,
    "args": {
      "External id": 177416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177416,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 177416, "pid": 5, "tid": 7, "ts": 1716454224226852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099188, "dur": 6,
    "args": {
      "External id": 177416,
      "cbid": 211, "correlation": 177416
    }
  },
  {
    "ph": "s", "id": 177416, "pid": 76337, "tid": -914061504, "ts": 1716454224099188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099251, "dur": 1,
    "args": {
      "External id": 177427,
      "cbid": 251, "correlation": 177427
    }
  },
  {
    "ph": "f", "id": 177427, "pid": 76337, "tid": -914061504, "ts": 1716454224099251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099254, "dur": 0,
    "args": {
      "External id": 177428,
      "cbid": 251, "correlation": 177428
    }
  },
  {
    "ph": "f", "id": 177428, "pid": 76337, "tid": -914061504, "ts": 1716454224099254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224226859, "dur": 6,
    "args": {
      "External id": 177429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177429,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 177429, "pid": 5, "tid": 7, "ts": 1716454224226859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099256, "dur": 12,
    "args": {
      "External id": 177429,
      "cbid": 211, "correlation": 177429
    }
  },
  {
    "ph": "s", "id": 177429, "pid": 76337, "tid": -914061504, "ts": 1716454224099256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224226866, "dur": 3,
    "args": {
      "External id": 177431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177431,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 177431, "pid": 5, "tid": 7, "ts": 1716454224226866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099269, "dur": 5,
    "args": {
      "External id": 177431,
      "cbid": 211, "correlation": 177431
    }
  },
  {
    "ph": "s", "id": 177431, "pid": 76337, "tid": -914061504, "ts": 1716454224099269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224226871, "dur": 156,
    "args": {
      "External id": 177452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177452,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 177452, "pid": 5, "tid": 7, "ts": 1716454224226871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099344, "dur": 12,
    "args": {
      "External id": 177452,
      "cbid": 211, "correlation": 177452
    }
  },
  {
    "ph": "s", "id": 177452, "pid": 76337, "tid": -914061504, "ts": 1716454224099344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099445, "dur": 1,
    "args": {
      "External id": 177470,
      "cbid": 251, "correlation": 177470
    }
  },
  {
    "ph": "f", "id": 177470, "pid": 76337, "tid": -914061504, "ts": 1716454224099445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224227029, "dur": 106,
    "args": {
      "External id": 177472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177472,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 177472, "pid": 5, "tid": 7, "ts": 1716454224227029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099451, "dur": 14,
    "args": {
      "External id": 177472,
      "cbid": 211, "correlation": 177472
    }
  },
  {
    "ph": "s", "id": 177472, "pid": 76337, "tid": -914061504, "ts": 1716454224099451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224227136, "dur": 35,
    "args": {
      "External id": 177480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177480,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177480, "pid": 5, "tid": 7, "ts": 1716454224227136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099526, "dur": 12,
    "args": {
      "External id": 177480,
      "cbid": 211, "correlation": 177480
    }
  },
  {
    "ph": "s", "id": 177480, "pid": 76337, "tid": -914061504, "ts": 1716454224099526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224227172, "dur": 67,
    "args": {
      "External id": 177488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177488,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177488, "pid": 5, "tid": 7, "ts": 1716454224227172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099567, "dur": 10,
    "args": {
      "External id": 177488,
      "cbid": 211, "correlation": 177488
    }
  },
  {
    "ph": "s", "id": 177488, "pid": 76337, "tid": -914061504, "ts": 1716454224099567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224227240, "dur": 92,
    "args": {
      "External id": 177510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177510,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177510, "pid": 5, "tid": 7, "ts": 1716454224227240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099619, "dur": 11,
    "args": {
      "External id": 177510,
      "cbid": 211, "correlation": 177510
    }
  },
  {
    "ph": "s", "id": 177510, "pid": 76337, "tid": -914061504, "ts": 1716454224099619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099712, "dur": 1,
    "args": {
      "External id": 177526,
      "cbid": 251, "correlation": 177526
    }
  },
  {
    "ph": "f", "id": 177526, "pid": 76337, "tid": -914061504, "ts": 1716454224099712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224227334, "dur": 580,
    "args": {
      "External id": 177528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177528,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177528, "pid": 5, "tid": 7, "ts": 1716454224227334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099718, "dur": 14,
    "args": {
      "External id": 177528,
      "cbid": 211, "correlation": 177528
    }
  },
  {
    "ph": "s", "id": 177528, "pid": 76337, "tid": -914061504, "ts": 1716454224099718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224227915, "dur": 246,
    "args": {
      "External id": 177536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177536, "pid": 5, "tid": 7, "ts": 1716454224227915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099785, "dur": 12,
    "args": {
      "External id": 177536,
      "cbid": 211, "correlation": 177536
    }
  },
  {
    "ph": "s", "id": 177536, "pid": 76337, "tid": -914061504, "ts": 1716454224099785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224228162, "dur": 254,
    "args": {
      "External id": 177544,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177544,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177544, "pid": 5, "tid": 7, "ts": 1716454224228162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099817, "dur": 8,
    "args": {
      "External id": 177544,
      "cbid": 211, "correlation": 177544
    }
  },
  {
    "ph": "s", "id": 177544, "pid": 76337, "tid": -914061504, "ts": 1716454224099817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099902, "dur": 1,
    "args": {
      "External id": 177560,
      "cbid": 251, "correlation": 177560
    }
  },
  {
    "ph": "f", "id": 177560, "pid": 76337, "tid": -914061504, "ts": 1716454224099902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099907, "dur": 0,
    "args": {
      "External id": 177562,
      "cbid": 251, "correlation": 177562
    }
  },
  {
    "ph": "f", "id": 177562, "pid": 76337, "tid": -914061504, "ts": 1716454224099907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224228418, "dur": 361,
    "args": {
      "External id": 177563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177563,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 177563, "pid": 5, "tid": 7, "ts": 1716454224228418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099910, "dur": 13,
    "args": {
      "External id": 177563,
      "cbid": 211, "correlation": 177563
    }
  },
  {
    "ph": "s", "id": 177563, "pid": 76337, "tid": -914061504, "ts": 1716454224099910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224228780, "dur": 50,
    "args": {
      "External id": 177571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177571,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177571, "pid": 5, "tid": 7, "ts": 1716454224228780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224099952, "dur": 10,
    "args": {
      "External id": 177571,
      "cbid": 211, "correlation": 177571
    }
  },
  {
    "ph": "s", "id": 177571, "pid": 76337, "tid": -914061504, "ts": 1716454224099952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224228831, "dur": 159,
    "args": {
      "External id": 177582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177582,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177582, "pid": 5, "tid": 7, "ts": 1716454224228831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100032, "dur": 190,
    "args": {
      "External id": 177582,
      "cbid": 211, "correlation": 177582
    }
  },
  {
    "ph": "s", "id": 177582, "pid": 76337, "tid": -914061504, "ts": 1716454224100032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100281, "dur": 0,
    "args": {
      "External id": 177594,
      "cbid": 317, "correlation": 177594
    }
  },
  {
    "ph": "f", "id": 177594, "pid": 76337, "tid": -914061504, "ts": 1716454224100281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100282, "dur": 0,
    "args": {
      "External id": 177595,
      "cbid": 203, "correlation": 177595
    }
  },
  {
    "ph": "f", "id": 177595, "pid": 76337, "tid": -914061504, "ts": 1716454224100282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100282, "dur": 1,
    "args": {
      "External id": 177596,
      "cbid": 205, "correlation": 177596
    }
  },
  {
    "ph": "f", "id": 177596, "pid": 76337, "tid": -914061504, "ts": 1716454224100282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100309, "dur": 1,
    "args": {
      "External id": 177600,
      "cbid": 251, "correlation": 177600
    }
  },
  {
    "ph": "f", "id": 177600, "pid": 76337, "tid": -914061504, "ts": 1716454224100309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100311, "dur": 0,
    "args": {
      "External id": 177601,
      "cbid": 251, "correlation": 177601
    }
  },
  {
    "ph": "f", "id": 177601, "pid": 76337, "tid": -914061504, "ts": 1716454224100311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100312, "dur": 0,
    "args": {
      "External id": 177602,
      "cbid": 251, "correlation": 177602
    }
  },
  {
    "ph": "f", "id": 177602, "pid": 76337, "tid": -914061504, "ts": 1716454224100312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100313, "dur": 0,
    "args": {
      "External id": 177603,
      "cbid": 251, "correlation": 177603
    }
  },
  {
    "ph": "f", "id": 177603, "pid": 76337, "tid": -914061504, "ts": 1716454224100313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100314, "dur": 0,
    "args": {
      "External id": 177604,
      "cbid": 251, "correlation": 177604
    }
  },
  {
    "ph": "f", "id": 177604, "pid": 76337, "tid": -914061504, "ts": 1716454224100314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100314, "dur": 0,
    "args": {
      "External id": 177605,
      "cbid": 251, "correlation": 177605
    }
  },
  {
    "ph": "f", "id": 177605, "pid": 76337, "tid": -914061504, "ts": 1716454224100314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100315, "dur": 0,
    "args": {
      "External id": 177606,
      "cbid": 251, "correlation": 177606
    }
  },
  {
    "ph": "f", "id": 177606, "pid": 76337, "tid": -914061504, "ts": 1716454224100315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100316, "dur": 0,
    "args": {
      "External id": 177607,
      "cbid": 251, "correlation": 177607
    }
  },
  {
    "ph": "f", "id": 177607, "pid": 76337, "tid": -914061504, "ts": 1716454224100316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100317, "dur": 0,
    "args": {
      "External id": 177608,
      "cbid": 251, "correlation": 177608
    }
  },
  {
    "ph": "f", "id": 177608, "pid": 76337, "tid": -914061504, "ts": 1716454224100317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224228992, "dur": 116,
    "args": {
      "External id": 177609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177609,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 177609, "pid": 5, "tid": 7, "ts": 1716454224228992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100319, "dur": 31,
    "args": {
      "External id": 177609,
      "cbid": 211, "correlation": 177609
    }
  },
  {
    "ph": "s", "id": 177609, "pid": 76337, "tid": -914061504, "ts": 1716454224100319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224229109, "dur": 60,
    "args": {
      "External id": 177615,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177615,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177615, "pid": 5, "tid": 7, "ts": 1716454224229109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100373, "dur": 104,
    "args": {
      "External id": 177615,
      "cbid": 211, "correlation": 177615
    }
  },
  {
    "ph": "s", "id": 177615, "pid": 76337, "tid": -914061504, "ts": 1716454224100373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224229170, "dur": 50,
    "args": {
      "External id": 177623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177623,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177623, "pid": 5, "tid": 7, "ts": 1716454224229170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100501, "dur": 238,
    "args": {
      "External id": 177623,
      "cbid": 211, "correlation": 177623
    }
  },
  {
    "ph": "s", "id": 177623, "pid": 76337, "tid": -914061504, "ts": 1716454224100501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224229221, "dur": 99,
    "args": {
      "External id": 177632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177632,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177632, "pid": 5, "tid": 7, "ts": 1716454224229221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100773, "dur": 11,
    "args": {
      "External id": 177632,
      "cbid": 211, "correlation": 177632
    }
  },
  {
    "ph": "s", "id": 177632, "pid": 76337, "tid": -914061504, "ts": 1716454224100773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224229322, "dur": 92,
    "args": {
      "External id": 177652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177652,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 177652, "pid": 5, "tid": 7, "ts": 1716454224229322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100853, "dur": 12,
    "args": {
      "External id": 177652,
      "cbid": 211, "correlation": 177652
    }
  },
  {
    "ph": "s", "id": 177652, "pid": 76337, "tid": -914061504, "ts": 1716454224100853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224229415, "dur": 5,
    "args": {
      "External id": 177664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177664,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 177664, "pid": 5, "tid": 7, "ts": 1716454224229415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100876, "dur": 7,
    "args": {
      "External id": 177664,
      "cbid": 211, "correlation": 177664
    }
  },
  {
    "ph": "s", "id": 177664, "pid": 76337, "tid": -914061504, "ts": 1716454224100876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224229421, "dur": 109,
    "args": {
      "External id": 177667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177667, "pid": 5, "tid": 7, "ts": 1716454224229421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224100895, "dur": 103,
    "args": {
      "External id": 177667,
      "cbid": 211, "correlation": 177667
    }
  },
  {
    "ph": "s", "id": 177667, "pid": 76337, "tid": -914061504, "ts": 1716454224100895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224229531, "dur": 69,
    "args": {
      "External id": 177676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177676,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177676, "pid": 5, "tid": 7, "ts": 1716454224229531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101034, "dur": 10,
    "args": {
      "External id": 177676,
      "cbid": 211, "correlation": 177676
    }
  },
  {
    "ph": "s", "id": 177676, "pid": 76337, "tid": -914061504, "ts": 1716454224101034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101091, "dur": 0,
    "args": {
      "External id": 177686,
      "cbid": 317, "correlation": 177686
    }
  },
  {
    "ph": "f", "id": 177686, "pid": 76337, "tid": -914061504, "ts": 1716454224101091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101092, "dur": 0,
    "args": {
      "External id": 177687,
      "cbid": 203, "correlation": 177687
    }
  },
  {
    "ph": "f", "id": 177687, "pid": 76337, "tid": -914061504, "ts": 1716454224101092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101093, "dur": 0,
    "args": {
      "External id": 177688,
      "cbid": 205, "correlation": 177688
    }
  },
  {
    "ph": "f", "id": 177688, "pid": 76337, "tid": -914061504, "ts": 1716454224101093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224229601, "dur": 77,
    "args": {
      "External id": 177692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177692, "pid": 5, "tid": 7, "ts": 1716454224229601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101110, "dur": 12,
    "args": {
      "External id": 177692,
      "cbid": 211, "correlation": 177692
    }
  },
  {
    "ph": "s", "id": 177692, "pid": 76337, "tid": -914061504, "ts": 1716454224101110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224229679, "dur": 24,
    "args": {
      "External id": 177694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177694,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177694, "pid": 5, "tid": 7, "ts": 1716454224229679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101125, "dur": 5,
    "args": {
      "External id": 177694,
      "cbid": 211, "correlation": 177694
    }
  },
  {
    "ph": "s", "id": 177694, "pid": 76337, "tid": -914061504, "ts": 1716454224101125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224229704, "dur": 4,
    "args": {
      "External id": 177696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177696, "pid": 5, "tid": 7, "ts": 1716454224229704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101136, "dur": 6,
    "args": {
      "External id": 177696,
      "cbid": 211, "correlation": 177696
    }
  },
  {
    "ph": "s", "id": 177696, "pid": 76337, "tid": -914061504, "ts": 1716454224101136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101146, "dur": 0,
    "args": {
      "External id": 177697,
      "cbid": 51, "correlation": 177697
    }
  },
  {
    "ph": "s", "id": 177697, "pid": 76337, "tid": -914061504, "ts": 1716454224101146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224229709, "dur": 1375,
    "args": {
      "External id": 177698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177698,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177698, "pid": 5, "tid": 7, "ts": 1716454224229709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101146, "dur": 6,
    "args": {
      "External id": 177698,
      "cbid": 211, "correlation": 177698
    }
  },
  {
    "ph": "s", "id": 177698, "pid": 76337, "tid": -914061504, "ts": 1716454224101146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224231085, "dur": 59,
    "args": {
      "External id": 177703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177703,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177703, "pid": 5, "tid": 7, "ts": 1716454224231085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101175, "dur": 9,
    "args": {
      "External id": 177703,
      "cbid": 211, "correlation": 177703
    }
  },
  {
    "ph": "s", "id": 177703, "pid": 76337, "tid": -914061504, "ts": 1716454224101175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224231145, "dur": 5,
    "args": {
      "External id": 177711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177711,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177711, "pid": 5, "tid": 7, "ts": 1716454224231145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101220, "dur": 10,
    "args": {
      "External id": 177711,
      "cbid": 211, "correlation": 177711
    }
  },
  {
    "ph": "s", "id": 177711, "pid": 76337, "tid": -914061504, "ts": 1716454224101220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101292, "dur": 1,
    "args": {
      "External id": 177727,
      "cbid": 251, "correlation": 177727
    }
  },
  {
    "ph": "f", "id": 177727, "pid": 76337, "tid": -914061504, "ts": 1716454224101292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101297, "dur": 0,
    "args": {
      "External id": 177729,
      "cbid": 251, "correlation": 177729
    }
  },
  {
    "ph": "f", "id": 177729, "pid": 76337, "tid": -914061504, "ts": 1716454224101297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224231152, "dur": 12,
    "args": {
      "External id": 177730,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177730,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 177730, "pid": 5, "tid": 7, "ts": 1716454224231152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101299, "dur": 12,
    "args": {
      "External id": 177730,
      "cbid": 211, "correlation": 177730
    }
  },
  {
    "ph": "s", "id": 177730, "pid": 76337, "tid": -914061504, "ts": 1716454224101299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224231165, "dur": 5,
    "args": {
      "External id": 177732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177732,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 177732, "pid": 5, "tid": 7, "ts": 1716454224231165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101314, "dur": 5,
    "args": {
      "External id": 177732,
      "cbid": 211, "correlation": 177732
    }
  },
  {
    "ph": "s", "id": 177732, "pid": 76337, "tid": -914061504, "ts": 1716454224101314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224231171, "dur": 54,
    "args": {
      "External id": 177742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177742,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177742, "pid": 5, "tid": 7, "ts": 1716454224231171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101372, "dur": 532,
    "args": {
      "External id": 177742,
      "cbid": 211, "correlation": 177742
    }
  },
  {
    "ph": "s", "id": 177742, "pid": 76337, "tid": -914061504, "ts": 1716454224101372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224231227, "dur": 52,
    "args": {
      "External id": 177762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177762,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 177762, "pid": 5, "tid": 7, "ts": 1716454224231227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101961, "dur": 19,
    "args": {
      "External id": 177762,
      "cbid": 211, "correlation": 177762
    }
  },
  {
    "ph": "s", "id": 177762, "pid": 76337, "tid": -914061504, "ts": 1716454224101961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224231280, "dur": 4,
    "args": {
      "External id": 177774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177774,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177774, "pid": 5, "tid": 7, "ts": 1716454224231280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224101991, "dur": 7,
    "args": {
      "External id": 177774,
      "cbid": 211, "correlation": 177774
    }
  },
  {
    "ph": "s", "id": 177774, "pid": 76337, "tid": -914061504, "ts": 1716454224101991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224231285, "dur": 55,
    "args": {
      "External id": 177777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177777,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177777, "pid": 5, "tid": 7, "ts": 1716454224231285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102013, "dur": 7,
    "args": {
      "External id": 177777,
      "cbid": 211, "correlation": 177777
    }
  },
  {
    "ph": "s", "id": 177777, "pid": 76337, "tid": -914061504, "ts": 1716454224102013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224231342, "dur": 37,
    "args": {
      "External id": 177786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177786,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177786, "pid": 5, "tid": 7, "ts": 1716454224231342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102055, "dur": 10,
    "args": {
      "External id": 177786,
      "cbid": 211, "correlation": 177786
    }
  },
  {
    "ph": "s", "id": 177786, "pid": 76337, "tid": -914061504, "ts": 1716454224102055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102121, "dur": 0,
    "args": {
      "External id": 177796,
      "cbid": 317, "correlation": 177796
    }
  },
  {
    "ph": "f", "id": 177796, "pid": 76337, "tid": -914061504, "ts": 1716454224102121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102122, "dur": 0,
    "args": {
      "External id": 177797,
      "cbid": 203, "correlation": 177797
    }
  },
  {
    "ph": "f", "id": 177797, "pid": 76337, "tid": -914061504, "ts": 1716454224102122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102123, "dur": 0,
    "args": {
      "External id": 177798,
      "cbid": 205, "correlation": 177798
    }
  },
  {
    "ph": "f", "id": 177798, "pid": 76337, "tid": -914061504, "ts": 1716454224102123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224231380, "dur": 40,
    "args": {
      "External id": 177802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177802,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177802, "pid": 5, "tid": 7, "ts": 1716454224231380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102137, "dur": 12,
    "args": {
      "External id": 177802,
      "cbid": 211, "correlation": 177802
    }
  },
  {
    "ph": "s", "id": 177802, "pid": 76337, "tid": -914061504, "ts": 1716454224102137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224231422, "dur": 14,
    "args": {
      "External id": 177804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177804,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177804, "pid": 5, "tid": 7, "ts": 1716454224231422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102152, "dur": 5,
    "args": {
      "External id": 177804,
      "cbid": 211, "correlation": 177804
    }
  },
  {
    "ph": "s", "id": 177804, "pid": 76337, "tid": -914061504, "ts": 1716454224102152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224231437, "dur": 3,
    "args": {
      "External id": 177806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177806,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177806, "pid": 5, "tid": 7, "ts": 1716454224231437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102161, "dur": 5,
    "args": {
      "External id": 177806,
      "cbid": 211, "correlation": 177806
    }
  },
  {
    "ph": "s", "id": 177806, "pid": 76337, "tid": -914061504, "ts": 1716454224102161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102170, "dur": 0,
    "args": {
      "External id": 177807,
      "cbid": 51, "correlation": 177807
    }
  },
  {
    "ph": "s", "id": 177807, "pid": 76337, "tid": -914061504, "ts": 1716454224102170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224231442, "dur": 701,
    "args": {
      "External id": 177808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177808,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177808, "pid": 5, "tid": 7, "ts": 1716454224231442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102171, "dur": 5,
    "args": {
      "External id": 177808,
      "cbid": 211, "correlation": 177808
    }
  },
  {
    "ph": "s", "id": 177808, "pid": 76337, "tid": -914061504, "ts": 1716454224102171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224232145, "dur": 60,
    "args": {
      "External id": 177813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177813,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177813, "pid": 5, "tid": 7, "ts": 1716454224232145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102199, "dur": 9,
    "args": {
      "External id": 177813,
      "cbid": 211, "correlation": 177813
    }
  },
  {
    "ph": "s", "id": 177813, "pid": 76337, "tid": -914061504, "ts": 1716454224102199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102258, "dur": 0,
    "args": {
      "External id": 177823,
      "cbid": 317, "correlation": 177823
    }
  },
  {
    "ph": "f", "id": 177823, "pid": 76337, "tid": -914061504, "ts": 1716454224102258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102259, "dur": 0,
    "args": {
      "External id": 177824,
      "cbid": 203, "correlation": 177824
    }
  },
  {
    "ph": "f", "id": 177824, "pid": 76337, "tid": -914061504, "ts": 1716454224102259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102259, "dur": 0,
    "args": {
      "External id": 177825,
      "cbid": 205, "correlation": 177825
    }
  },
  {
    "ph": "f", "id": 177825, "pid": 76337, "tid": -914061504, "ts": 1716454224102259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224232206, "dur": 75,
    "args": {
      "External id": 177829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177829,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177829, "pid": 5, "tid": 7, "ts": 1716454224232206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102271, "dur": 11,
    "args": {
      "External id": 177829,
      "cbid": 211, "correlation": 177829
    }
  },
  {
    "ph": "s", "id": 177829, "pid": 76337, "tid": -914061504, "ts": 1716454224102271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224232282, "dur": 209,
    "args": {
      "External id": 177831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177831,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177831, "pid": 5, "tid": 7, "ts": 1716454224232282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102291, "dur": 7,
    "args": {
      "External id": 177831,
      "cbid": 211, "correlation": 177831
    }
  },
  {
    "ph": "s", "id": 177831, "pid": 76337, "tid": -914061504, "ts": 1716454224102291,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224232493, "dur": 38,
    "args": {
      "External id": 177833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177833,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177833, "pid": 5, "tid": 7, "ts": 1716454224232493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102303, "dur": 5,
    "args": {
      "External id": 177833,
      "cbid": 211, "correlation": 177833
    }
  },
  {
    "ph": "s", "id": 177833, "pid": 76337, "tid": -914061504, "ts": 1716454224102303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224232532, "dur": 60,
    "args": {
      "External id": 177839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177839, "pid": 5, "tid": 7, "ts": 1716454224232532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102329, "dur": 500,
    "args": {
      "External id": 177839,
      "cbid": 211, "correlation": 177839
    }
  },
  {
    "ph": "s", "id": 177839, "pid": 76337, "tid": -914061504, "ts": 1716454224102329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224232593, "dur": 50,
    "args": {
      "External id": 177847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177847,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177847, "pid": 5, "tid": 7, "ts": 1716454224232593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102849, "dur": 9,
    "args": {
      "External id": 177847,
      "cbid": 211, "correlation": 177847
    }
  },
  {
    "ph": "s", "id": 177847, "pid": 76337, "tid": -914061504, "ts": 1716454224102849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224232644, "dur": 35,
    "args": {
      "External id": 177855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177855,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177855, "pid": 5, "tid": 7, "ts": 1716454224232644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102879, "dur": 9,
    "args": {
      "External id": 177855,
      "cbid": 211, "correlation": 177855
    }
  },
  {
    "ph": "s", "id": 177855, "pid": 76337, "tid": -914061504, "ts": 1716454224102879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224232680, "dur": 52,
    "args": {
      "External id": 177875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177875,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 177875, "pid": 5, "tid": 7, "ts": 1716454224232680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102967, "dur": 20,
    "args": {
      "External id": 177875,
      "cbid": 211, "correlation": 177875
    }
  },
  {
    "ph": "s", "id": 177875, "pid": 76337, "tid": -914061504, "ts": 1716454224102967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224232734, "dur": 4,
    "args": {
      "External id": 177887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177887,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 177887, "pid": 5, "tid": 7, "ts": 1716454224232734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224102997, "dur": 7,
    "args": {
      "External id": 177887,
      "cbid": 211, "correlation": 177887
    }
  },
  {
    "ph": "s", "id": 177887, "pid": 76337, "tid": -914061504, "ts": 1716454224102997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224232740, "dur": 55,
    "args": {
      "External id": 177890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177890, "pid": 5, "tid": 7, "ts": 1716454224232740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103016, "dur": 7,
    "args": {
      "External id": 177890,
      "cbid": 211, "correlation": 177890
    }
  },
  {
    "ph": "s", "id": 177890, "pid": 76337, "tid": -914061504, "ts": 1716454224103016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103078, "dur": 0,
    "args": {
      "External id": 177901,
      "cbid": 317, "correlation": 177901
    }
  },
  {
    "ph": "f", "id": 177901, "pid": 76337, "tid": -914061504, "ts": 1716454224103078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103078, "dur": 0,
    "args": {
      "External id": 177902,
      "cbid": 203, "correlation": 177902
    }
  },
  {
    "ph": "f", "id": 177902, "pid": 76337, "tid": -914061504, "ts": 1716454224103078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103079, "dur": 0,
    "args": {
      "External id": 177903,
      "cbid": 205, "correlation": 177903
    }
  },
  {
    "ph": "f", "id": 177903, "pid": 76337, "tid": -914061504, "ts": 1716454224103079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103101, "dur": 1,
    "args": {
      "External id": 177907,
      "cbid": 251, "correlation": 177907
    }
  },
  {
    "ph": "f", "id": 177907, "pid": 76337, "tid": -914061504, "ts": 1716454224103101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103104, "dur": 0,
    "args": {
      "External id": 177908,
      "cbid": 251, "correlation": 177908
    }
  },
  {
    "ph": "f", "id": 177908, "pid": 76337, "tid": -914061504, "ts": 1716454224103104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103105, "dur": 0,
    "args": {
      "External id": 177909,
      "cbid": 251, "correlation": 177909
    }
  },
  {
    "ph": "f", "id": 177909, "pid": 76337, "tid": -914061504, "ts": 1716454224103105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103106, "dur": 0,
    "args": {
      "External id": 177910,
      "cbid": 251, "correlation": 177910
    }
  },
  {
    "ph": "f", "id": 177910, "pid": 76337, "tid": -914061504, "ts": 1716454224103106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103106, "dur": 0,
    "args": {
      "External id": 177911,
      "cbid": 251, "correlation": 177911
    }
  },
  {
    "ph": "f", "id": 177911, "pid": 76337, "tid": -914061504, "ts": 1716454224103106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103107, "dur": 0,
    "args": {
      "External id": 177912,
      "cbid": 251, "correlation": 177912
    }
  },
  {
    "ph": "f", "id": 177912, "pid": 76337, "tid": -914061504, "ts": 1716454224103107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103108, "dur": 0,
    "args": {
      "External id": 177913,
      "cbid": 251, "correlation": 177913
    }
  },
  {
    "ph": "f", "id": 177913, "pid": 76337, "tid": -914061504, "ts": 1716454224103108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103109, "dur": 0,
    "args": {
      "External id": 177914,
      "cbid": 251, "correlation": 177914
    }
  },
  {
    "ph": "f", "id": 177914, "pid": 76337, "tid": -914061504, "ts": 1716454224103109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103111, "dur": 0,
    "args": {
      "External id": 177915,
      "cbid": 251, "correlation": 177915
    }
  },
  {
    "ph": "f", "id": 177915, "pid": 76337, "tid": -914061504, "ts": 1716454224103111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224232796, "dur": 115,
    "args": {
      "External id": 177916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177916,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 177916, "pid": 5, "tid": 7, "ts": 1716454224232796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103113, "dur": 12,
    "args": {
      "External id": 177916,
      "cbid": 211, "correlation": 177916
    }
  },
  {
    "ph": "s", "id": 177916, "pid": 76337, "tid": -914061504, "ts": 1716454224103113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224232912, "dur": 60,
    "args": {
      "External id": 177922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177922, "pid": 5, "tid": 7, "ts": 1716454224232912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103148, "dur": 9,
    "args": {
      "External id": 177922,
      "cbid": 211, "correlation": 177922
    }
  },
  {
    "ph": "s", "id": 177922, "pid": 76337, "tid": -914061504, "ts": 1716454224103148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224232973, "dur": 645,
    "args": {
      "External id": 177931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177931,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177931, "pid": 5, "tid": 7, "ts": 1716454224232973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103254, "dur": 15,
    "args": {
      "External id": 177931,
      "cbid": 211, "correlation": 177931
    }
  },
  {
    "ph": "s", "id": 177931, "pid": 76337, "tid": -914061504, "ts": 1716454224103254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224233620, "dur": 182,
    "args": {
      "External id": 177953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177953,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 177953, "pid": 5, "tid": 7, "ts": 1716454224233620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103317, "dur": 10,
    "args": {
      "External id": 177953,
      "cbid": 211, "correlation": 177953
    }
  },
  {
    "ph": "s", "id": 177953, "pid": 76337, "tid": -914061504, "ts": 1716454224103317,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103413, "dur": 2,
    "args": {
      "External id": 177964,
      "cbid": 251, "correlation": 177964
    }
  },
  {
    "ph": "f", "id": 177964, "pid": 76337, "tid": -914061504, "ts": 1716454224103413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224233804, "dur": 195,
    "args": {
      "External id": 177965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177965,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177965, "pid": 5, "tid": 7, "ts": 1716454224233804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103419, "dur": 14,
    "args": {
      "External id": 177965,
      "cbid": 211, "correlation": 177965
    }
  },
  {
    "ph": "s", "id": 177965, "pid": 76337, "tid": -914061504, "ts": 1716454224103419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103491, "dur": 1,
    "args": {
      "External id": 177976,
      "cbid": 251, "correlation": 177976
    }
  },
  {
    "ph": "f", "id": 177976, "pid": 76337, "tid": -914061504, "ts": 1716454224103491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224234000, "dur": 188,
    "args": {
      "External id": 177977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177977,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177977, "pid": 5, "tid": 7, "ts": 1716454224234000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103495, "dur": 12,
    "args": {
      "External id": 177977,
      "cbid": 211, "correlation": 177977
    }
  },
  {
    "ph": "s", "id": 177977, "pid": 76337, "tid": -914061504, "ts": 1716454224103495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103563, "dur": 1,
    "args": {
      "External id": 177988,
      "cbid": 251, "correlation": 177988
    }
  },
  {
    "ph": "f", "id": 177988, "pid": 76337, "tid": -914061504, "ts": 1716454224103563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224234190, "dur": 189,
    "args": {
      "External id": 177989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 177989,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 177989, "pid": 5, "tid": 7, "ts": 1716454224234190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103568, "dur": 11,
    "args": {
      "External id": 177989,
      "cbid": 211, "correlation": 177989
    }
  },
  {
    "ph": "s", "id": 177989, "pid": 76337, "tid": -914061504, "ts": 1716454224103568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224234380, "dur": 18769,
    "args": {
      "External id": 178010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178010,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 178010, "pid": 5, "tid": 7, "ts": 1716454224234380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103649, "dur": 12,
    "args": {
      "External id": 178010,
      "cbid": 211, "correlation": 178010
    }
  },
  {
    "ph": "s", "id": 178010, "pid": 76337, "tid": -914061504, "ts": 1716454224103649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103753, "dur": 1,
    "args": {
      "External id": 178028,
      "cbid": 251, "correlation": 178028
    }
  },
  {
    "ph": "f", "id": 178028, "pid": 76337, "tid": -914061504, "ts": 1716454224103753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224253150, "dur": 201,
    "args": {
      "External id": 178030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178030,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 178030, "pid": 5, "tid": 7, "ts": 1716454224253150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103758, "dur": 14,
    "args": {
      "External id": 178030,
      "cbid": 211, "correlation": 178030
    }
  },
  {
    "ph": "s", "id": 178030, "pid": 76337, "tid": -914061504, "ts": 1716454224103758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224253353, "dur": 67,
    "args": {
      "External id": 178038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178038,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178038, "pid": 5, "tid": 7, "ts": 1716454224253353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103833, "dur": 12,
    "args": {
      "External id": 178038,
      "cbid": 211, "correlation": 178038
    }
  },
  {
    "ph": "s", "id": 178038, "pid": 76337, "tid": -914061504, "ts": 1716454224103833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224253421, "dur": 97,
    "args": {
      "External id": 178046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178046,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178046, "pid": 5, "tid": 7, "ts": 1716454224253421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103872, "dur": 21,
    "args": {
      "External id": 178046,
      "cbid": 211, "correlation": 178046
    }
  },
  {
    "ph": "s", "id": 178046, "pid": 76337, "tid": -914061504, "ts": 1716454224103872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224253519, "dur": 55,
    "args": {
      "External id": 178057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178057, "pid": 5, "tid": 7, "ts": 1716454224253519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224103962, "dur": 1898,
    "args": {
      "External id": 178057,
      "cbid": 211, "correlation": 178057
    }
  },
  {
    "ph": "s", "id": 178057, "pid": 76337, "tid": -914061504, "ts": 1716454224103962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224253575, "dur": 92,
    "args": {
      "External id": 178079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178079,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178079, "pid": 5, "tid": 7, "ts": 1716454224253575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224105880, "dur": 127,
    "args": {
      "External id": 178079,
      "cbid": 211, "correlation": 178079
    }
  },
  {
    "ph": "s", "id": 178079, "pid": 76337, "tid": -914061504, "ts": 1716454224105880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106087, "dur": 1,
    "args": {
      "External id": 178090,
      "cbid": 251, "correlation": 178090
    }
  },
  {
    "ph": "f", "id": 178090, "pid": 76337, "tid": -914061504, "ts": 1716454224106087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224253669, "dur": 103,
    "args": {
      "External id": 178091,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178091,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 178091, "pid": 5, "tid": 7, "ts": 1716454224253669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106092, "dur": 14,
    "args": {
      "External id": 178091,
      "cbid": 211, "correlation": 178091
    }
  },
  {
    "ph": "s", "id": 178091, "pid": 76337, "tid": -914061504, "ts": 1716454224106092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106170, "dur": 1,
    "args": {
      "External id": 178102,
      "cbid": 251, "correlation": 178102
    }
  },
  {
    "ph": "f", "id": 178102, "pid": 76337, "tid": -914061504, "ts": 1716454224106170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106173, "dur": 0,
    "args": {
      "External id": 178103,
      "cbid": 251, "correlation": 178103
    }
  },
  {
    "ph": "f", "id": 178103, "pid": 76337, "tid": -914061504, "ts": 1716454224106173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224253774, "dur": 10,
    "args": {
      "External id": 178104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178104,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178104, "pid": 5, "tid": 7, "ts": 1716454224253774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106175, "dur": 13,
    "args": {
      "External id": 178104,
      "cbid": 211, "correlation": 178104
    }
  },
  {
    "ph": "s", "id": 178104, "pid": 76337, "tid": -914061504, "ts": 1716454224106175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224253785, "dur": 5,
    "args": {
      "External id": 178106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178106,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 178106, "pid": 5, "tid": 7, "ts": 1716454224253785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106190, "dur": 6,
    "args": {
      "External id": 178106,
      "cbid": 211, "correlation": 178106
    }
  },
  {
    "ph": "s", "id": 178106, "pid": 76337, "tid": -914061504, "ts": 1716454224106190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106251, "dur": 1,
    "args": {
      "External id": 178117,
      "cbid": 251, "correlation": 178117
    }
  },
  {
    "ph": "f", "id": 178117, "pid": 76337, "tid": -914061504, "ts": 1716454224106251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106255, "dur": 0,
    "args": {
      "External id": 178118,
      "cbid": 251, "correlation": 178118
    }
  },
  {
    "ph": "f", "id": 178118, "pid": 76337, "tid": -914061504, "ts": 1716454224106255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224253792, "dur": 6,
    "args": {
      "External id": 178119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178119,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178119, "pid": 5, "tid": 7, "ts": 1716454224253792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106257, "dur": 12,
    "args": {
      "External id": 178119,
      "cbid": 211, "correlation": 178119
    }
  },
  {
    "ph": "s", "id": 178119, "pid": 76337, "tid": -914061504, "ts": 1716454224106257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224253799, "dur": 3,
    "args": {
      "External id": 178121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178121,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 178121, "pid": 5, "tid": 7, "ts": 1716454224253799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106270, "dur": 6,
    "args": {
      "External id": 178121,
      "cbid": 211, "correlation": 178121
    }
  },
  {
    "ph": "s", "id": 178121, "pid": 76337, "tid": -914061504, "ts": 1716454224106270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224253804, "dur": 158,
    "args": {
      "External id": 178142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178142,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 178142, "pid": 5, "tid": 7, "ts": 1716454224253804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106344, "dur": 13,
    "args": {
      "External id": 178142,
      "cbid": 211, "correlation": 178142
    }
  },
  {
    "ph": "s", "id": 178142, "pid": 76337, "tid": -914061504, "ts": 1716454224106344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106444, "dur": 1,
    "args": {
      "External id": 178160,
      "cbid": 251, "correlation": 178160
    }
  },
  {
    "ph": "f", "id": 178160, "pid": 76337, "tid": -914061504, "ts": 1716454224106444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224253964, "dur": 109,
    "args": {
      "External id": 178162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178162,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 178162, "pid": 5, "tid": 7, "ts": 1716454224253964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106450, "dur": 14,
    "args": {
      "External id": 178162,
      "cbid": 211, "correlation": 178162
    }
  },
  {
    "ph": "s", "id": 178162, "pid": 76337, "tid": -914061504, "ts": 1716454224106450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224254074, "dur": 34,
    "args": {
      "External id": 178170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178170,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178170, "pid": 5, "tid": 7, "ts": 1716454224254074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106522, "dur": 12,
    "args": {
      "External id": 178170,
      "cbid": 211, "correlation": 178170
    }
  },
  {
    "ph": "s", "id": 178170, "pid": 76337, "tid": -914061504, "ts": 1716454224106522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224254109, "dur": 66,
    "args": {
      "External id": 178178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178178,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178178, "pid": 5, "tid": 7, "ts": 1716454224254109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106563, "dur": 9,
    "args": {
      "External id": 178178,
      "cbid": 211, "correlation": 178178
    }
  },
  {
    "ph": "s", "id": 178178, "pid": 76337, "tid": -914061504, "ts": 1716454224106563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224254177, "dur": 92,
    "args": {
      "External id": 178200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178200,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178200, "pid": 5, "tid": 7, "ts": 1716454224254177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106616, "dur": 10,
    "args": {
      "External id": 178200,
      "cbid": 211, "correlation": 178200
    }
  },
  {
    "ph": "s", "id": 178200, "pid": 76337, "tid": -914061504, "ts": 1716454224106616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106707, "dur": 1,
    "args": {
      "External id": 178216,
      "cbid": 251, "correlation": 178216
    }
  },
  {
    "ph": "f", "id": 178216, "pid": 76337, "tid": -914061504, "ts": 1716454224106707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224254271, "dur": 578,
    "args": {
      "External id": 178218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178218,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 178218, "pid": 5, "tid": 7, "ts": 1716454224254271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106713, "dur": 13,
    "args": {
      "External id": 178218,
      "cbid": 211, "correlation": 178218
    }
  },
  {
    "ph": "s", "id": 178218, "pid": 76337, "tid": -914061504, "ts": 1716454224106713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224254850, "dur": 247,
    "args": {
      "External id": 178226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178226, "pid": 5, "tid": 7, "ts": 1716454224254850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106780, "dur": 13,
    "args": {
      "External id": 178226,
      "cbid": 211, "correlation": 178226
    }
  },
  {
    "ph": "s", "id": 178226, "pid": 76337, "tid": -914061504, "ts": 1716454224106780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224255098, "dur": 253,
    "args": {
      "External id": 178234,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178234,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178234, "pid": 5, "tid": 7, "ts": 1716454224255098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106812, "dur": 9,
    "args": {
      "External id": 178234,
      "cbid": 211, "correlation": 178234
    }
  },
  {
    "ph": "s", "id": 178234, "pid": 76337, "tid": -914061504, "ts": 1716454224106812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106897, "dur": 1,
    "args": {
      "External id": 178250,
      "cbid": 251, "correlation": 178250
    }
  },
  {
    "ph": "f", "id": 178250, "pid": 76337, "tid": -914061504, "ts": 1716454224106897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106902, "dur": 0,
    "args": {
      "External id": 178252,
      "cbid": 251, "correlation": 178252
    }
  },
  {
    "ph": "f", "id": 178252, "pid": 76337, "tid": -914061504, "ts": 1716454224106902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224255352, "dur": 358,
    "args": {
      "External id": 178253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178253,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178253, "pid": 5, "tid": 7, "ts": 1716454224255352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106905, "dur": 14,
    "args": {
      "External id": 178253,
      "cbid": 211, "correlation": 178253
    }
  },
  {
    "ph": "s", "id": 178253, "pid": 76337, "tid": -914061504, "ts": 1716454224106905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224255712, "dur": 50,
    "args": {
      "External id": 178261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178261,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178261, "pid": 5, "tid": 7, "ts": 1716454224255712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224106948, "dur": 161,
    "args": {
      "External id": 178261,
      "cbid": 211, "correlation": 178261
    }
  },
  {
    "ph": "s", "id": 178261, "pid": 76337, "tid": -914061504, "ts": 1716454224106948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224255763, "dur": 158,
    "args": {
      "External id": 178272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178272, "pid": 5, "tid": 7, "ts": 1716454224255763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107168, "dur": 69,
    "args": {
      "External id": 178272,
      "cbid": 211, "correlation": 178272
    }
  },
  {
    "ph": "s", "id": 178272, "pid": 76337, "tid": -914061504, "ts": 1716454224107168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107294, "dur": 0,
    "args": {
      "External id": 178284,
      "cbid": 317, "correlation": 178284
    }
  },
  {
    "ph": "f", "id": 178284, "pid": 76337, "tid": -914061504, "ts": 1716454224107294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107295, "dur": 0,
    "args": {
      "External id": 178285,
      "cbid": 203, "correlation": 178285
    }
  },
  {
    "ph": "f", "id": 178285, "pid": 76337, "tid": -914061504, "ts": 1716454224107295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107295, "dur": 0,
    "args": {
      "External id": 178286,
      "cbid": 205, "correlation": 178286
    }
  },
  {
    "ph": "f", "id": 178286, "pid": 76337, "tid": -914061504, "ts": 1716454224107295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107318, "dur": 1,
    "args": {
      "External id": 178290,
      "cbid": 251, "correlation": 178290
    }
  },
  {
    "ph": "f", "id": 178290, "pid": 76337, "tid": -914061504, "ts": 1716454224107318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107320, "dur": 0,
    "args": {
      "External id": 178291,
      "cbid": 251, "correlation": 178291
    }
  },
  {
    "ph": "f", "id": 178291, "pid": 76337, "tid": -914061504, "ts": 1716454224107320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107321, "dur": 0,
    "args": {
      "External id": 178292,
      "cbid": 251, "correlation": 178292
    }
  },
  {
    "ph": "f", "id": 178292, "pid": 76337, "tid": -914061504, "ts": 1716454224107321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107322, "dur": 0,
    "args": {
      "External id": 178293,
      "cbid": 251, "correlation": 178293
    }
  },
  {
    "ph": "f", "id": 178293, "pid": 76337, "tid": -914061504, "ts": 1716454224107322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107323, "dur": 0,
    "args": {
      "External id": 178294,
      "cbid": 251, "correlation": 178294
    }
  },
  {
    "ph": "f", "id": 178294, "pid": 76337, "tid": -914061504, "ts": 1716454224107323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107323, "dur": 0,
    "args": {
      "External id": 178295,
      "cbid": 251, "correlation": 178295
    }
  },
  {
    "ph": "f", "id": 178295, "pid": 76337, "tid": -914061504, "ts": 1716454224107323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107324, "dur": 0,
    "args": {
      "External id": 178296,
      "cbid": 251, "correlation": 178296
    }
  },
  {
    "ph": "f", "id": 178296, "pid": 76337, "tid": -914061504, "ts": 1716454224107324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107325, "dur": 0,
    "args": {
      "External id": 178297,
      "cbid": 251, "correlation": 178297
    }
  },
  {
    "ph": "f", "id": 178297, "pid": 76337, "tid": -914061504, "ts": 1716454224107325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107326, "dur": 0,
    "args": {
      "External id": 178298,
      "cbid": 251, "correlation": 178298
    }
  },
  {
    "ph": "f", "id": 178298, "pid": 76337, "tid": -914061504, "ts": 1716454224107326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224255922, "dur": 116,
    "args": {
      "External id": 178299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178299,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 178299, "pid": 5, "tid": 7, "ts": 1716454224255922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107328, "dur": 37,
    "args": {
      "External id": 178299,
      "cbid": 211, "correlation": 178299
    }
  },
  {
    "ph": "s", "id": 178299, "pid": 76337, "tid": -914061504, "ts": 1716454224107328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224256040, "dur": 60,
    "args": {
      "External id": 178305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178305,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178305, "pid": 5, "tid": 7, "ts": 1716454224256040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107388, "dur": 282,
    "args": {
      "External id": 178305,
      "cbid": 211, "correlation": 178305
    }
  },
  {
    "ph": "s", "id": 178305, "pid": 76337, "tid": -914061504, "ts": 1716454224107388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224256101, "dur": 50,
    "args": {
      "External id": 178313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178313,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178313, "pid": 5, "tid": 7, "ts": 1716454224256101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107694, "dur": 9,
    "args": {
      "External id": 178313,
      "cbid": 211, "correlation": 178313
    }
  },
  {
    "ph": "s", "id": 178313, "pid": 76337, "tid": -914061504, "ts": 1716454224107694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224256152, "dur": 53,
    "args": {
      "External id": 178333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178333,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 178333, "pid": 5, "tid": 7, "ts": 1716454224256152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107779, "dur": 12,
    "args": {
      "External id": 178333,
      "cbid": 211, "correlation": 178333
    }
  },
  {
    "ph": "s", "id": 178333, "pid": 76337, "tid": -914061504, "ts": 1716454224107779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224256207, "dur": 5,
    "args": {
      "External id": 178345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178345,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 178345, "pid": 5, "tid": 7, "ts": 1716454224256207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107802, "dur": 7,
    "args": {
      "External id": 178345,
      "cbid": 211, "correlation": 178345
    }
  },
  {
    "ph": "s", "id": 178345, "pid": 76337, "tid": -914061504, "ts": 1716454224107802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224256213, "dur": 56,
    "args": {
      "External id": 178348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178348,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178348, "pid": 5, "tid": 7, "ts": 1716454224256213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107820, "dur": 108,
    "args": {
      "External id": 178348,
      "cbid": 211, "correlation": 178348
    }
  },
  {
    "ph": "s", "id": 178348, "pid": 76337, "tid": -914061504, "ts": 1716454224107820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224256270, "dur": 36,
    "args": {
      "External id": 178357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178357,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178357, "pid": 5, "tid": 7, "ts": 1716454224256270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224107967, "dur": 19,
    "args": {
      "External id": 178357,
      "cbid": 211, "correlation": 178357
    }
  },
  {
    "ph": "s", "id": 178357, "pid": 76337, "tid": -914061504, "ts": 1716454224107967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108035, "dur": 0,
    "args": {
      "External id": 178367,
      "cbid": 317, "correlation": 178367
    }
  },
  {
    "ph": "f", "id": 178367, "pid": 76337, "tid": -914061504, "ts": 1716454224108035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108036, "dur": 0,
    "args": {
      "External id": 178368,
      "cbid": 203, "correlation": 178368
    }
  },
  {
    "ph": "f", "id": 178368, "pid": 76337, "tid": -914061504, "ts": 1716454224108036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108037, "dur": 0,
    "args": {
      "External id": 178369,
      "cbid": 205, "correlation": 178369
    }
  },
  {
    "ph": "f", "id": 178369, "pid": 76337, "tid": -914061504, "ts": 1716454224108037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224256307, "dur": 40,
    "args": {
      "External id": 178373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178373,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178373, "pid": 5, "tid": 7, "ts": 1716454224256307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108052, "dur": 12,
    "args": {
      "External id": 178373,
      "cbid": 211, "correlation": 178373
    }
  },
  {
    "ph": "s", "id": 178373, "pid": 76337, "tid": -914061504, "ts": 1716454224108052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224256349, "dur": 4,
    "args": {
      "External id": 178375,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178375,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.428571428571429
    }
  },
  {
    "ph": "f", "id": 178375, "pid": 5, "tid": 7, "ts": 1716454224256349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108068, "dur": 21,
    "args": {
      "External id": 178375,
      "cbid": 51, "correlation": 178375
    }
  },
  {
    "ph": "s", "id": 178375, "pid": 76337, "tid": -914061504, "ts": 1716454224108068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108095, "dur": 1,
    "args": {
      "External id": 178377,
      "cbid": 200, "correlation": 178377
    }
  },
  {
    "ph": "f", "id": 178377, "pid": 76337, "tid": -914061504, "ts": 1716454224108095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108096, "dur": 0,
    "args": {
      "External id": 178378,
      "cbid": 200, "correlation": 178378
    }
  },
  {
    "ph": "f", "id": 178378, "pid": 76337, "tid": -914061504, "ts": 1716454224108096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108097, "dur": 0,
    "args": {
      "External id": 178379,
      "cbid": 200, "correlation": 178379
    }
  },
  {
    "ph": "f", "id": 178379, "pid": 76337, "tid": -914061504, "ts": 1716454224108097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108097, "dur": 0,
    "args": {
      "External id": 178380,
      "cbid": 200, "correlation": 178380
    }
  },
  {
    "ph": "f", "id": 178380, "pid": 76337, "tid": -914061504, "ts": 1716454224108097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108099, "dur": 3,
    "args": {
      "External id": 178381,
      "cbid": 15, "correlation": 178381
    }
  },
  {
    "ph": "f", "id": 178381, "pid": 76337, "tid": -914061504, "ts": 1716454224108099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108103, "dur": 1,
    "args": {
      "External id": 178382,
      "cbid": 251, "correlation": 178382
    }
  },
  {
    "ph": "f", "id": 178382, "pid": 76337, "tid": -914061504, "ts": 1716454224108103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224256354, "dur": 23,
    "args": {
      "External id": 178383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178383,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178383, "pid": 5, "tid": 7, "ts": 1716454224256354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108106, "dur": 8,
    "args": {
      "External id": 178383,
      "cbid": 211, "correlation": 178383
    }
  },
  {
    "ph": "s", "id": 178383, "pid": 76337, "tid": -914061504, "ts": 1716454224108106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224256378, "dur": 4,
    "args": {
      "External id": 178385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178385,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 178385, "pid": 5, "tid": 7, "ts": 1716454224256378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108121, "dur": 5,
    "args": {
      "External id": 178385,
      "cbid": 211, "correlation": 178385
    }
  },
  {
    "ph": "s", "id": 178385, "pid": 76337, "tid": -914061504, "ts": 1716454224108121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108130, "dur": 0,
    "args": {
      "External id": 178386,
      "cbid": 51, "correlation": 178386
    }
  },
  {
    "ph": "s", "id": 178386, "pid": 76337, "tid": -914061504, "ts": 1716454224108130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224256384, "dur": 190,
    "args": {
      "External id": 178387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178387,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178387, "pid": 5, "tid": 7, "ts": 1716454224256384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108131, "dur": 191,
    "args": {
      "External id": 178387,
      "cbid": 211, "correlation": 178387
    }
  },
  {
    "ph": "s", "id": 178387, "pid": 76337, "tid": -914061504, "ts": 1716454224108131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224256575, "dur": 6,
    "args": {
      "External id": 178388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178388,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178388, "pid": 5, "tid": 7, "ts": 1716454224256575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108325, "dur": 6,
    "args": {
      "External id": 178388,
      "cbid": 211, "correlation": 178388
    }
  },
  {
    "ph": "s", "id": 178388, "pid": 76337, "tid": -914061504, "ts": 1716454224108325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224256582, "dur": 5,
    "args": {
      "External id": 178394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 178394, "pid": 5, "tid": 7, "ts": 1716454224256582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224108355, "dur": 9,
    "args": {
      "External id": 178394,
      "cbid": 211, "correlation": 178394
    }
  },
  {
    "ph": "s", "id": 178394, "pid": 76337, "tid": -914061504, "ts": 1716454224108355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224256588, "dur": 3,
    "args": {
      "External id": 178402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178402,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178402, "pid": 5, "tid": 7, "ts": 1716454224256588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224110508, "dur": 17,
    "args": {
      "External id": 178402,
      "cbid": 211, "correlation": 178402
    }
  },
  {
    "ph": "s", "id": 178402, "pid": 76337, "tid": -914061504, "ts": 1716454224110508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224256593, "dur": 3,
    "args": {
      "External id": 178410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178410,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178410, "pid": 5, "tid": 7, "ts": 1716454224256593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224110552, "dur": 11,
    "args": {
      "External id": 178410,
      "cbid": 211, "correlation": 178410
    }
  },
  {
    "ph": "s", "id": 178410, "pid": 76337, "tid": -914061504, "ts": 1716454224110552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224256597, "dur": 3,
    "args": {
      "External id": 178418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178418,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178418, "pid": 5, "tid": 7, "ts": 1716454224256597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224110579, "dur": 9,
    "args": {
      "External id": 178418,
      "cbid": 211, "correlation": 178418
    }
  },
  {
    "ph": "s", "id": 178418, "pid": 76337, "tid": -914061504, "ts": 1716454224110579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224256601, "dur": 3,
    "args": {
      "External id": 178427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178427,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178427, "pid": 5, "tid": 7, "ts": 1716454224256601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224110788, "dur": 15,
    "args": {
      "External id": 178427,
      "cbid": 211, "correlation": 178427
    }
  },
  {
    "ph": "s", "id": 178427, "pid": 76337, "tid": -914061504, "ts": 1716454224110788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224256605, "dur": 3,
    "args": {
      "External id": 178436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178436,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178436, "pid": 5, "tid": 7, "ts": 1716454224256605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224110818, "dur": 8,
    "args": {
      "External id": 178436,
      "cbid": 211, "correlation": 178436
    }
  },
  {
    "ph": "s", "id": 178436, "pid": 76337, "tid": -914061504, "ts": 1716454224110818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224256609, "dur": 3,
    "args": {
      "External id": 178444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178444,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178444, "pid": 5, "tid": 7, "ts": 1716454224256609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224110845, "dur": 8,
    "args": {
      "External id": 178444,
      "cbid": 211, "correlation": 178444
    }
  },
  {
    "ph": "s", "id": 178444, "pid": 76337, "tid": -914061504, "ts": 1716454224110845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224256614, "dur": 3,
    "args": {
      "External id": 178452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178452,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178452, "pid": 5, "tid": 7, "ts": 1716454224256614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224111115, "dur": 16,
    "args": {
      "External id": 178452,
      "cbid": 211, "correlation": 178452
    }
  },
  {
    "ph": "s", "id": 178452, "pid": 76337, "tid": -914061504, "ts": 1716454224111115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224256618, "dur": 3,
    "args": {
      "External id": 178460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178460,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178460, "pid": 5, "tid": 7, "ts": 1716454224256618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224111147, "dur": 8,
    "args": {
      "External id": 178460,
      "cbid": 211, "correlation": 178460
    }
  },
  {
    "ph": "s", "id": 178460, "pid": 76337, "tid": -914061504, "ts": 1716454224111147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224256623, "dur": 1,
    "args": {
      "External id": 178470,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178470,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 178470, "pid": 5, "tid": 7, "ts": 1716454224256623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224111230, "dur": 36,
    "args": {
      "External id": 178470,
      "cbid": 41, "correlation": 178470
    }
  },
  {
    "ph": "s", "id": 178470, "pid": 76337, "tid": -914061504, "ts": 1716454224111230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224111267, "dur": 145383,
    "args": {
      "External id": 178471,
      "cbid": 131, "correlation": 178471
    }
  },
  {
    "ph": "f", "id": 178471, "pid": 76337, "tid": -914061504, "ts": 1716454224111267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224256988, "dur": 3,
    "args": {
      "External id": 178479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178479, "pid": 5, "tid": 7, "ts": 1716454224256988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224256951, "dur": 39,
    "args": {
      "External id": 178479,
      "cbid": 211, "correlation": 178479
    }
  },
  {
    "ph": "s", "id": 178479, "pid": 76337, "tid": -914061504, "ts": 1716454224256951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257100, "dur": 3,
    "args": {
      "External id": 178488,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178488,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002380952380952381
    }
  },
  {
    "ph": "f", "id": 178488, "pid": 5, "tid": 7, "ts": 1716454224257100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257067, "dur": 32,
    "args": {
      "External id": 178488,
      "cbid": 41, "correlation": 178488
    }
  },
  {
    "ph": "s", "id": 178488, "pid": 76337, "tid": -914061504, "ts": 1716454224257067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224257212, "dur": 4,
    "args": {
      "External id": 178498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178498,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178498, "pid": 5, "tid": 7, "ts": 1716454224257212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257194, "dur": 19,
    "args": {
      "External id": 178498,
      "cbid": 211, "correlation": 178498
    }
  },
  {
    "ph": "s", "id": 178498, "pid": 76337, "tid": -914061504, "ts": 1716454224257194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257309, "dur": 1,
    "args": {
      "External id": 178508,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178508,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 178508, "pid": 5, "tid": 7, "ts": 1716454224257309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257288, "dur": 20,
    "args": {
      "External id": 178508,
      "cbid": 41, "correlation": 178508
    }
  },
  {
    "ph": "s", "id": 178508, "pid": 76337, "tid": -914061504, "ts": 1716454224257288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257309, "dur": 8,
    "args": {
      "External id": 178509,
      "cbid": 131, "correlation": 178509
    }
  },
  {
    "ph": "f", "id": 178509, "pid": 76337, "tid": -914061504, "ts": 1716454224257309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257388, "dur": 3,
    "args": {
      "External id": 178516,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178516,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.425255338904364
    }
  },
  {
    "ph": "f", "id": 178516, "pid": 5, "tid": 7, "ts": 1716454224257388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257364, "dur": 23,
    "args": {
      "External id": 178516,
      "cbid": 41, "correlation": 178516
    }
  },
  {
    "ph": "s", "id": 178516, "pid": 76337, "tid": -914061504, "ts": 1716454224257364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257501, "dur": 3,
    "args": {
      "External id": 178535,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178535,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005263157894736842
    }
  },
  {
    "ph": "f", "id": 178535, "pid": 5, "tid": 7, "ts": 1716454224257501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257479, "dur": 21,
    "args": {
      "External id": 178535,
      "cbid": 41, "correlation": 178535
    }
  },
  {
    "ph": "s", "id": 178535, "pid": 76337, "tid": -914061504, "ts": 1716454224257479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224257546, "dur": 3,
    "args": {
      "External id": 178541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178541, "pid": 5, "tid": 7, "ts": 1716454224257546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257534, "dur": 12,
    "args": {
      "External id": 178541,
      "cbid": 211, "correlation": 178541
    }
  },
  {
    "ph": "s", "id": 178541, "pid": 76337, "tid": -914061504, "ts": 1716454224257534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224257559, "dur": 6,
    "args": {
      "External id": 178543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178543,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 178543, "pid": 5, "tid": 7, "ts": 1716454224257559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257549, "dur": 9,
    "args": {
      "External id": 178543,
      "cbid": 211, "correlation": 178543
    }
  },
  {
    "ph": "s", "id": 178543, "pid": 76337, "tid": -914061504, "ts": 1716454224257549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454224257568, "dur": 3,
    "args": {
      "External id": 178545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178545,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178545, "pid": 5, "tid": 7, "ts": 1716454224257568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257560, "dur": 6,
    "args": {
      "External id": 178545,
      "cbid": 211, "correlation": 178545
    }
  },
  {
    "ph": "s", "id": 178545, "pid": 76337, "tid": -914061504, "ts": 1716454224257560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257606, "dur": 2,
    "args": {
      "External id": 178553,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178553,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 178553, "pid": 5, "tid": 7, "ts": 1716454224257606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257592, "dur": 14,
    "args": {
      "External id": 178553,
      "cbid": 41, "correlation": 178553
    }
  },
  {
    "ph": "s", "id": 178553, "pid": 76337, "tid": -914061504, "ts": 1716454224257592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224257657, "dur": 3,
    "args": {
      "External id": 178567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178567,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178567, "pid": 5, "tid": 7, "ts": 1716454224257657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257646, "dur": 13,
    "args": {
      "External id": 178567,
      "cbid": 211, "correlation": 178567
    }
  },
  {
    "ph": "s", "id": 178567, "pid": 76337, "tid": -914061504, "ts": 1716454224257646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224257677, "dur": 2,
    "args": {
      "External id": 178581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178581,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178581, "pid": 5, "tid": 7, "ts": 1716454224257677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257670, "dur": 6,
    "args": {
      "External id": 178581,
      "cbid": 211, "correlation": 178581
    }
  },
  {
    "ph": "s", "id": 178581, "pid": 76337, "tid": -914061504, "ts": 1716454224257670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224257720, "dur": 6,
    "args": {
      "External id": 178588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178588,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178588, "pid": 5, "tid": 7, "ts": 1716454224257720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257709, "dur": 11,
    "args": {
      "External id": 178588,
      "cbid": 211, "correlation": 178588
    }
  },
  {
    "ph": "s", "id": 178588, "pid": 76337, "tid": -914061504, "ts": 1716454224257709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224257731, "dur": 6,
    "args": {
      "External id": 178591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178591,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178591, "pid": 5, "tid": 7, "ts": 1716454224257731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257723, "dur": 7,
    "args": {
      "External id": 178591,
      "cbid": 211, "correlation": 178591
    }
  },
  {
    "ph": "s", "id": 178591, "pid": 76337, "tid": -914061504, "ts": 1716454224257723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224257740, "dur": 3,
    "args": {
      "External id": 178593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178593,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178593, "pid": 5, "tid": 7, "ts": 1716454224257740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257732, "dur": 7,
    "args": {
      "External id": 178593,
      "cbid": 211, "correlation": 178593
    }
  },
  {
    "ph": "s", "id": 178593, "pid": 76337, "tid": -914061504, "ts": 1716454224257732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257760, "dur": 2,
    "args": {
      "External id": 178596,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178596,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 178596, "pid": 5, "tid": 7, "ts": 1716454224257760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257748, "dur": 12,
    "args": {
      "External id": 178596,
      "cbid": 41, "correlation": 178596
    }
  },
  {
    "ph": "s", "id": 178596, "pid": 76337, "tid": -914061504, "ts": 1716454224257748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224257819, "dur": 4,
    "args": {
      "External id": 178612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178612,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178612, "pid": 5, "tid": 7, "ts": 1716454224257819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257806, "dur": 13,
    "args": {
      "External id": 178612,
      "cbid": 211, "correlation": 178612
    }
  },
  {
    "ph": "s", "id": 178612, "pid": 76337, "tid": -914061504, "ts": 1716454224257806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224257841, "dur": 3,
    "args": {
      "External id": 178617,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178617,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003125
    }
  },
  {
    "ph": "f", "id": 178617, "pid": 5, "tid": 7, "ts": 1716454224257841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257825, "dur": 14,
    "args": {
      "External id": 178617,
      "cbid": 41, "correlation": 178617
    }
  },
  {
    "ph": "s", "id": 178617, "pid": 76337, "tid": -914061504, "ts": 1716454224257825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224257877, "dur": 1,
    "args": {
      "External id": 178623,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178623,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005787037037037037
    }
  },
  {
    "ph": "f", "id": 178623, "pid": 5, "tid": 7, "ts": 1716454224257877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257849, "dur": 37,
    "args": {
      "External id": 178623,
      "cbid": 41, "correlation": 178623
    }
  },
  {
    "ph": "s", "id": 178623, "pid": 76337, "tid": -914061504, "ts": 1716454224257849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257887, "dur": 4,
    "args": {
      "External id": 178624,
      "cbid": 131, "correlation": 178624
    }
  },
  {
    "ph": "f", "id": 178624, "pid": 76337, "tid": -914061504, "ts": 1716454224257887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224257942, "dur": 3,
    "args": {
      "External id": 178632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178632,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178632, "pid": 5, "tid": 7, "ts": 1716454224257942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257928, "dur": 14,
    "args": {
      "External id": 178632,
      "cbid": 211, "correlation": 178632
    }
  },
  {
    "ph": "s", "id": 178632, "pid": 76337, "tid": -914061504, "ts": 1716454224257928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224257972, "dur": 3,
    "args": {
      "External id": 178642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178642,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178642, "pid": 5, "tid": 7, "ts": 1716454224257972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257963, "dur": 9,
    "args": {
      "External id": 178642,
      "cbid": 211, "correlation": 178642
    }
  },
  {
    "ph": "s", "id": 178642, "pid": 76337, "tid": -914061504, "ts": 1716454224257963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224258007, "dur": 3,
    "args": {
      "External id": 178651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178651,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178651, "pid": 5, "tid": 7, "ts": 1716454224258007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224257997, "dur": 9,
    "args": {
      "External id": 178651,
      "cbid": 211, "correlation": 178651
    }
  },
  {
    "ph": "s", "id": 178651, "pid": 76337, "tid": -914061504, "ts": 1716454224257997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224258155, "dur": 12,
    "args": {
      "External id": 178661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178661,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178661, "pid": 5, "tid": 7, "ts": 1716454224258155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258141, "dur": 15,
    "args": {
      "External id": 178661,
      "cbid": 211, "correlation": 178661
    }
  },
  {
    "ph": "s", "id": 178661, "pid": 76337, "tid": -914061504, "ts": 1716454224258141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224258195, "dur": 3,
    "args": {
      "External id": 178669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178669,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178669, "pid": 5, "tid": 7, "ts": 1716454224258195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258186, "dur": 8,
    "args": {
      "External id": 178669,
      "cbid": 211, "correlation": 178669
    }
  },
  {
    "ph": "s", "id": 178669, "pid": 76337, "tid": -914061504, "ts": 1716454224258186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224258242, "dur": 11,
    "args": {
      "External id": 178679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178679,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178679, "pid": 5, "tid": 7, "ts": 1716454224258242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258230, "dur": 12,
    "args": {
      "External id": 178679,
      "cbid": 211, "correlation": 178679
    }
  },
  {
    "ph": "s", "id": 178679, "pid": 76337, "tid": -914061504, "ts": 1716454224258230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224258279, "dur": 10,
    "args": {
      "External id": 178687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178687,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178687, "pid": 5, "tid": 7, "ts": 1716454224258279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258269, "dur": 10,
    "args": {
      "External id": 178687,
      "cbid": 211, "correlation": 178687
    }
  },
  {
    "ph": "s", "id": 178687, "pid": 76337, "tid": -914061504, "ts": 1716454224258269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224258306, "dur": 3,
    "args": {
      "External id": 178696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178696, "pid": 5, "tid": 7, "ts": 1716454224258306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258297, "dur": 9,
    "args": {
      "External id": 178696,
      "cbid": 211, "correlation": 178696
    }
  },
  {
    "ph": "s", "id": 178696, "pid": 76337, "tid": -914061504, "ts": 1716454224258297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224258332, "dur": 5,
    "args": {
      "External id": 178705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178705,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178705, "pid": 5, "tid": 7, "ts": 1716454224258332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258323, "dur": 8,
    "args": {
      "External id": 178705,
      "cbid": 211, "correlation": 178705
    }
  },
  {
    "ph": "s", "id": 178705, "pid": 76337, "tid": -914061504, "ts": 1716454224258323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224258377, "dur": 8,
    "args": {
      "External id": 178715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178715,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178715, "pid": 5, "tid": 7, "ts": 1716454224258377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258366, "dur": 11,
    "args": {
      "External id": 178715,
      "cbid": 211, "correlation": 178715
    }
  },
  {
    "ph": "s", "id": 178715, "pid": 76337, "tid": -914061504, "ts": 1716454224258366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224258765, "dur": 3,
    "args": {
      "External id": 178724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178724,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178724, "pid": 5, "tid": 7, "ts": 1716454224258765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258748, "dur": 17,
    "args": {
      "External id": 178724,
      "cbid": 211, "correlation": 178724
    }
  },
  {
    "ph": "s", "id": 178724, "pid": 76337, "tid": -914061504, "ts": 1716454224258748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224258793, "dur": 3,
    "args": {
      "External id": 178732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178732,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178732, "pid": 5, "tid": 7, "ts": 1716454224258793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258783, "dur": 10,
    "args": {
      "External id": 178732,
      "cbid": 211, "correlation": 178732
    }
  },
  {
    "ph": "s", "id": 178732, "pid": 76337, "tid": -914061504, "ts": 1716454224258783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224258845, "dur": 1,
    "args": {
      "External id": 178742,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178742,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 178742, "pid": 5, "tid": 7, "ts": 1716454224258845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258830, "dur": 13,
    "args": {
      "External id": 178742,
      "cbid": 41, "correlation": 178742
    }
  },
  {
    "ph": "s", "id": 178742, "pid": 76337, "tid": -914061504, "ts": 1716454224258830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258844, "dur": 8,
    "args": {
      "External id": 178743,
      "cbid": 131, "correlation": 178743
    }
  },
  {
    "ph": "f", "id": 178743, "pid": 76337, "tid": -914061504, "ts": 1716454224258844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224258935, "dur": 2,
    "args": {
      "External id": 178751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178751, "pid": 5, "tid": 7, "ts": 1716454224258935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258921, "dur": 14,
    "args": {
      "External id": 178751,
      "cbid": 211, "correlation": 178751
    }
  },
  {
    "ph": "s", "id": 178751, "pid": 76337, "tid": -914061504, "ts": 1716454224258921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224259017, "dur": 3,
    "args": {
      "External id": 178760,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178760,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0025252525252525255
    }
  },
  {
    "ph": "f", "id": 178760, "pid": 5, "tid": 7, "ts": 1716454224259017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224258998, "dur": 19,
    "args": {
      "External id": 178760,
      "cbid": 41, "correlation": 178760
    }
  },
  {
    "ph": "s", "id": 178760, "pid": 76337, "tid": -914061504, "ts": 1716454224258998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224259089, "dur": 3,
    "args": {
      "External id": 178770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178770,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 178770, "pid": 5, "tid": 7, "ts": 1716454224259089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259075, "dur": 14,
    "args": {
      "External id": 178770,
      "cbid": 211, "correlation": 178770
    }
  },
  {
    "ph": "s", "id": 178770, "pid": 76337, "tid": -914061504, "ts": 1716454224259075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224259140, "dur": 1,
    "args": {
      "External id": 178780,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178780,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 178780, "pid": 5, "tid": 7, "ts": 1716454224259140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259127, "dur": 12,
    "args": {
      "External id": 178780,
      "cbid": 41, "correlation": 178780
    }
  },
  {
    "ph": "s", "id": 178780, "pid": 76337, "tid": -914061504, "ts": 1716454224259127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259139, "dur": 8,
    "args": {
      "External id": 178781,
      "cbid": 131, "correlation": 178781
    }
  },
  {
    "ph": "f", "id": 178781, "pid": 76337, "tid": -914061504, "ts": 1716454224259139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224259200, "dur": 3,
    "args": {
      "External id": 178788,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178788,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.346938775510203
    }
  },
  {
    "ph": "f", "id": 178788, "pid": 5, "tid": 7, "ts": 1716454224259200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259181, "dur": 18,
    "args": {
      "External id": 178788,
      "cbid": 41, "correlation": 178788
    }
  },
  {
    "ph": "s", "id": 178788, "pid": 76337, "tid": -914061504, "ts": 1716454224259181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224259254, "dur": 1,
    "args": {
      "External id": 178799,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178799,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 178799, "pid": 5, "tid": 7, "ts": 1716454224259254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259240, "dur": 11,
    "args": {
      "External id": 178799,
      "cbid": 41, "correlation": 178799
    }
  },
  {
    "ph": "s", "id": 178799, "pid": 76337, "tid": -914061504, "ts": 1716454224259240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259252, "dur": 8,
    "args": {
      "External id": 178800,
      "cbid": 131, "correlation": 178800
    }
  },
  {
    "ph": "f", "id": 178800, "pid": 76337, "tid": -914061504, "ts": 1716454224259252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224259302, "dur": 3,
    "args": {
      "External id": 178808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178808,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178808, "pid": 5, "tid": 7, "ts": 1716454224259302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259288, "dur": 13,
    "args": {
      "External id": 178808,
      "cbid": 211, "correlation": 178808
    }
  },
  {
    "ph": "s", "id": 178808, "pid": 76337, "tid": -914061504, "ts": 1716454224259288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224259331, "dur": 3,
    "args": {
      "External id": 178818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178818,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178818, "pid": 5, "tid": 7, "ts": 1716454224259331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259321, "dur": 8,
    "args": {
      "External id": 178818,
      "cbid": 211, "correlation": 178818
    }
  },
  {
    "ph": "s", "id": 178818, "pid": 76337, "tid": -914061504, "ts": 1716454224259321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224259353, "dur": 3,
    "args": {
      "External id": 178827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178827,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178827, "pid": 5, "tid": 7, "ts": 1716454224259353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259344, "dur": 7,
    "args": {
      "External id": 178827,
      "cbid": 211, "correlation": 178827
    }
  },
  {
    "ph": "s", "id": 178827, "pid": 76337, "tid": -914061504, "ts": 1716454224259344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224259422, "dur": 5,
    "args": {
      "External id": 178835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178835,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178835, "pid": 5, "tid": 7, "ts": 1716454224259422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259409, "dur": 14,
    "args": {
      "External id": 178835,
      "cbid": 211, "correlation": 178835
    }
  },
  {
    "ph": "s", "id": 178835, "pid": 76337, "tid": -914061504, "ts": 1716454224259409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224259463, "dur": 3,
    "args": {
      "External id": 178844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178844,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178844, "pid": 5, "tid": 7, "ts": 1716454224259463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259452, "dur": 10,
    "args": {
      "External id": 178844,
      "cbid": 211, "correlation": 178844
    }
  },
  {
    "ph": "s", "id": 178844, "pid": 76337, "tid": -914061504, "ts": 1716454224259452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224259486, "dur": 3,
    "args": {
      "External id": 178853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178853,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178853, "pid": 5, "tid": 7, "ts": 1716454224259486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259477, "dur": 7,
    "args": {
      "External id": 178853,
      "cbid": 211, "correlation": 178853
    }
  },
  {
    "ph": "s", "id": 178853, "pid": 76337, "tid": -914061504, "ts": 1716454224259477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224259568, "dur": 3,
    "args": {
      "External id": 178861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178861,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 178861, "pid": 5, "tid": 7, "ts": 1716454224259568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259557, "dur": 10,
    "args": {
      "External id": 178861,
      "cbid": 211, "correlation": 178861
    }
  },
  {
    "ph": "s", "id": 178861, "pid": 76337, "tid": -914061504, "ts": 1716454224259557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224259630, "dur": 1,
    "args": {
      "External id": 178869,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178869,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 178869, "pid": 5, "tid": 7, "ts": 1716454224259630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259614, "dur": 26,
    "args": {
      "External id": 178869,
      "cbid": 41, "correlation": 178869
    }
  },
  {
    "ph": "s", "id": 178869, "pid": 76337, "tid": -914061504, "ts": 1716454224259614,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259640, "dur": 3,
    "args": {
      "External id": 178870,
      "cbid": 131, "correlation": 178870
    }
  },
  {
    "ph": "f", "id": 178870, "pid": 76337, "tid": -914061504, "ts": 1716454224259640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224259713, "dur": 1,
    "args": {
      "External id": 178880,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178880,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 178880, "pid": 5, "tid": 7, "ts": 1716454224259713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259698, "dur": 12,
    "args": {
      "External id": 178880,
      "cbid": 41, "correlation": 178880
    }
  },
  {
    "ph": "s", "id": 178880, "pid": 76337, "tid": -914061504, "ts": 1716454224259698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259711, "dur": 8,
    "args": {
      "External id": 178881,
      "cbid": 131, "correlation": 178881
    }
  },
  {
    "ph": "f", "id": 178881, "pid": 76337, "tid": -914061504, "ts": 1716454224259711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224259775, "dur": 1,
    "args": {
      "External id": 178890,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178890,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 178890, "pid": 5, "tid": 7, "ts": 1716454224259775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259763, "dur": 9,
    "args": {
      "External id": 178890,
      "cbid": 41, "correlation": 178890
    }
  },
  {
    "ph": "s", "id": 178890, "pid": 76337, "tid": -914061504, "ts": 1716454224259763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259773, "dur": 8,
    "args": {
      "External id": 178891,
      "cbid": 131, "correlation": 178891
    }
  },
  {
    "ph": "f", "id": 178891, "pid": 76337, "tid": -914061504, "ts": 1716454224259773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224259854, "dur": 4,
    "args": {
      "External id": 178898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178898, "pid": 5, "tid": 7, "ts": 1716454224259854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259837, "dur": 18,
    "args": {
      "External id": 178898,
      "cbid": 211, "correlation": 178898
    }
  },
  {
    "ph": "s", "id": 178898, "pid": 76337, "tid": -914061504, "ts": 1716454224259837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454224259897, "dur": 4,
    "args": {
      "External id": 178918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178918,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178918, "pid": 5, "tid": 7, "ts": 1716454224259897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259884, "dur": 13,
    "args": {
      "External id": 178918,
      "cbid": 211, "correlation": 178918
    }
  },
  {
    "ph": "s", "id": 178918, "pid": 76337, "tid": -914061504, "ts": 1716454224259884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259898, "dur": 0,
    "args": {
      "External id": 178919,
      "cbid": 11, "correlation": 178919
    }
  },
  {
    "ph": "f", "id": 178919, "pid": 76337, "tid": -914061504, "ts": 1716454224259898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259899, "dur": 0,
    "args": {
      "External id": 178920,
      "cbid": 11, "correlation": 178920
    }
  },
  {
    "ph": "f", "id": 178920, "pid": 76337, "tid": -914061504, "ts": 1716454224259899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224259912, "dur": 1,
    "args": {
      "External id": 178923,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178923,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 178923, "pid": 5, "tid": 7, "ts": 1716454224259912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259900, "dur": 21,
    "args": {
      "External id": 178923,
      "cbid": 41, "correlation": 178923
    }
  },
  {
    "ph": "s", "id": 178923, "pid": 76337, "tid": -914061504, "ts": 1716454224259900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259922, "dur": 3,
    "args": {
      "External id": 178924,
      "cbid": 131, "correlation": 178924
    }
  },
  {
    "ph": "f", "id": 178924, "pid": 76337, "tid": -914061504, "ts": 1716454224259922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224259951, "dur": 3,
    "args": {
      "External id": 178948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178948,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178948, "pid": 5, "tid": 7, "ts": 1716454224259951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259940, "dur": 10,
    "args": {
      "External id": 178948,
      "cbid": 211, "correlation": 178948
    }
  },
  {
    "ph": "s", "id": 178948, "pid": 76337, "tid": -914061504, "ts": 1716454224259940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259951, "dur": 0,
    "args": {
      "External id": 178949,
      "cbid": 11, "correlation": 178949
    }
  },
  {
    "ph": "f", "id": 178949, "pid": 76337, "tid": -914061504, "ts": 1716454224259951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259952, "dur": 0,
    "args": {
      "External id": 178950,
      "cbid": 11, "correlation": 178950
    }
  },
  {
    "ph": "f", "id": 178950, "pid": 76337, "tid": -914061504, "ts": 1716454224259952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259953, "dur": 1,
    "args": {
      "External id": 178952,
      "cbid": 200, "correlation": 178952
    }
  },
  {
    "ph": "f", "id": 178952, "pid": 76337, "tid": -914061504, "ts": 1716454224259953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224259964, "dur": 4,
    "args": {
      "External id": 178954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178954,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178954, "pid": 5, "tid": 7, "ts": 1716454224259964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259956, "dur": 8,
    "args": {
      "External id": 178954,
      "cbid": 211, "correlation": 178954
    }
  },
  {
    "ph": "s", "id": 178954, "pid": 76337, "tid": -914061504, "ts": 1716454224259956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259965, "dur": 0,
    "args": {
      "External id": 178955,
      "cbid": 11, "correlation": 178955
    }
  },
  {
    "ph": "f", "id": 178955, "pid": 76337, "tid": -914061504, "ts": 1716454224259965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224259965, "dur": 0,
    "args": {
      "External id": 178956,
      "cbid": 11, "correlation": 178956
    }
  },
  {
    "ph": "f", "id": 178956, "pid": 76337, "tid": -914061504, "ts": 1716454224259965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224260014, "dur": 1,
    "args": {
      "External id": 178963,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178963,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 178963, "pid": 5, "tid": 7, "ts": 1716454224260014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260002, "dur": 21,
    "args": {
      "External id": 178963,
      "cbid": 41, "correlation": 178963
    }
  },
  {
    "ph": "s", "id": 178963, "pid": 76337, "tid": -914061504, "ts": 1716454224260002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260024, "dur": 3,
    "args": {
      "External id": 178964,
      "cbid": 131, "correlation": 178964
    }
  },
  {
    "ph": "f", "id": 178964, "pid": 76337, "tid": -914061504, "ts": 1716454224260024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224260075, "dur": 1,
    "args": {
      "External id": 178974,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 178974,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 178974, "pid": 5, "tid": 7, "ts": 1716454224260075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260063, "dur": 10,
    "args": {
      "External id": 178974,
      "cbid": 41, "correlation": 178974
    }
  },
  {
    "ph": "s", "id": 178974, "pid": 76337, "tid": -914061504, "ts": 1716454224260063,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260073, "dur": 8,
    "args": {
      "External id": 178975,
      "cbid": 131, "correlation": 178975
    }
  },
  {
    "ph": "f", "id": 178975, "pid": 76337, "tid": -914061504, "ts": 1716454224260073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224260147, "dur": 5,
    "args": {
      "External id": 178982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178982,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178982, "pid": 5, "tid": 7, "ts": 1716454224260147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260131, "dur": 16,
    "args": {
      "External id": 178982,
      "cbid": 211, "correlation": 178982
    }
  },
  {
    "ph": "s", "id": 178982, "pid": 76337, "tid": -914061504, "ts": 1716454224260131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224260225, "dur": 3,
    "args": {
      "External id": 178991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178991,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178991, "pid": 5, "tid": 7, "ts": 1716454224260225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260212, "dur": 13,
    "args": {
      "External id": 178991,
      "cbid": 211, "correlation": 178991
    }
  },
  {
    "ph": "s", "id": 178991, "pid": 76337, "tid": -914061504, "ts": 1716454224260212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224260261, "dur": 3,
    "args": {
      "External id": 178999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 178999,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 178999, "pid": 5, "tid": 7, "ts": 1716454224260261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260251, "dur": 9,
    "args": {
      "External id": 178999,
      "cbid": 211, "correlation": 178999
    }
  },
  {
    "ph": "s", "id": 178999, "pid": 76337, "tid": -914061504, "ts": 1716454224260251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224260295, "dur": 4,
    "args": {
      "External id": 179007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179007,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179007, "pid": 5, "tid": 7, "ts": 1716454224260295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260284, "dur": 12,
    "args": {
      "External id": 179007,
      "cbid": 211, "correlation": 179007
    }
  },
  {
    "ph": "s", "id": 179007, "pid": 76337, "tid": -914061504, "ts": 1716454224260284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224260327, "dur": 4,
    "args": {
      "External id": 179015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179015,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179015, "pid": 5, "tid": 7, "ts": 1716454224260327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260316, "dur": 10,
    "args": {
      "External id": 179015,
      "cbid": 211, "correlation": 179015
    }
  },
  {
    "ph": "s", "id": 179015, "pid": 76337, "tid": -914061504, "ts": 1716454224260316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224260354, "dur": 3,
    "args": {
      "External id": 179023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179023,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179023, "pid": 5, "tid": 7, "ts": 1716454224260354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260345, "dur": 8,
    "args": {
      "External id": 179023,
      "cbid": 211, "correlation": 179023
    }
  },
  {
    "ph": "s", "id": 179023, "pid": 76337, "tid": -914061504, "ts": 1716454224260345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224260381, "dur": 3,
    "args": {
      "External id": 179031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179031,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179031, "pid": 5, "tid": 7, "ts": 1716454224260381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260371, "dur": 10,
    "args": {
      "External id": 179031,
      "cbid": 211, "correlation": 179031
    }
  },
  {
    "ph": "s", "id": 179031, "pid": 76337, "tid": -914061504, "ts": 1716454224260371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224260402, "dur": 4,
    "args": {
      "External id": 179039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179039,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179039, "pid": 5, "tid": 7, "ts": 1716454224260402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260393, "dur": 7,
    "args": {
      "External id": 179039,
      "cbid": 211, "correlation": 179039
    }
  },
  {
    "ph": "s", "id": 179039, "pid": 76337, "tid": -914061504, "ts": 1716454224260393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224260420, "dur": 4,
    "args": {
      "External id": 179047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179047,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179047, "pid": 5, "tid": 7, "ts": 1716454224260420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260412, "dur": 7,
    "args": {
      "External id": 179047,
      "cbid": 211, "correlation": 179047
    }
  },
  {
    "ph": "s", "id": 179047, "pid": 76337, "tid": -914061504, "ts": 1716454224260412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224260440, "dur": 3,
    "args": {
      "External id": 179055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179055,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179055, "pid": 5, "tid": 7, "ts": 1716454224260440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260433, "dur": 7,
    "args": {
      "External id": 179055,
      "cbid": 211, "correlation": 179055
    }
  },
  {
    "ph": "s", "id": 179055, "pid": 76337, "tid": -914061504, "ts": 1716454224260433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224260507, "dur": 3,
    "args": {
      "External id": 179063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179063,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179063, "pid": 5, "tid": 7, "ts": 1716454224260507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260495, "dur": 11,
    "args": {
      "External id": 179063,
      "cbid": 211, "correlation": 179063
    }
  },
  {
    "ph": "s", "id": 179063, "pid": 76337, "tid": -914061504, "ts": 1716454224260495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224260533, "dur": 4,
    "args": {
      "External id": 179071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179071,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179071, "pid": 5, "tid": 7, "ts": 1716454224260533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260524, "dur": 8,
    "args": {
      "External id": 179071,
      "cbid": 211, "correlation": 179071
    }
  },
  {
    "ph": "s", "id": 179071, "pid": 76337, "tid": -914061504, "ts": 1716454224260524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224260555, "dur": 4,
    "args": {
      "External id": 179079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179079,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179079, "pid": 5, "tid": 7, "ts": 1716454224260555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260546, "dur": 8,
    "args": {
      "External id": 179079,
      "cbid": 211, "correlation": 179079
    }
  },
  {
    "ph": "s", "id": 179079, "pid": 76337, "tid": -914061504, "ts": 1716454224260546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224260574, "dur": 3,
    "args": {
      "External id": 179087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179087,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179087, "pid": 5, "tid": 7, "ts": 1716454224260574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224260566, "dur": 7,
    "args": {
      "External id": 179087,
      "cbid": 211, "correlation": 179087
    }
  },
  {
    "ph": "s", "id": 179087, "pid": 76337, "tid": -914061504, "ts": 1716454224260566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224261141, "dur": 5,
    "args": {
      "External id": 179096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179096,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179096, "pid": 5, "tid": 7, "ts": 1716454224261141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261125, "dur": 16,
    "args": {
      "External id": 179096,
      "cbid": 211, "correlation": 179096
    }
  },
  {
    "ph": "s", "id": 179096, "pid": 76337, "tid": -914061504, "ts": 1716454224261125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224261179, "dur": 5,
    "args": {
      "External id": 179105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179105,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179105, "pid": 5, "tid": 7, "ts": 1716454224261179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261169, "dur": 9,
    "args": {
      "External id": 179105,
      "cbid": 211, "correlation": 179105
    }
  },
  {
    "ph": "s", "id": 179105, "pid": 76337, "tid": -914061504, "ts": 1716454224261169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224261374, "dur": 3,
    "args": {
      "External id": 179121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179121, "pid": 5, "tid": 7, "ts": 1716454224261374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261360, "dur": 15,
    "args": {
      "External id": 179121,
      "cbid": 211, "correlation": 179121
    }
  },
  {
    "ph": "s", "id": 179121, "pid": 76337, "tid": -914061504, "ts": 1716454224261360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224261409, "dur": 3,
    "args": {
      "External id": 179129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179129,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179129, "pid": 5, "tid": 7, "ts": 1716454224261409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261399, "dur": 9,
    "args": {
      "External id": 179129,
      "cbid": 211, "correlation": 179129
    }
  },
  {
    "ph": "s", "id": 179129, "pid": 76337, "tid": -914061504, "ts": 1716454224261399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224261439, "dur": 3,
    "args": {
      "External id": 179137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179137,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179137, "pid": 5, "tid": 7, "ts": 1716454224261439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261429, "dur": 8,
    "args": {
      "External id": 179137,
      "cbid": 211, "correlation": 179137
    }
  },
  {
    "ph": "s", "id": 179137, "pid": 76337, "tid": -914061504, "ts": 1716454224261429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224261470, "dur": 4,
    "args": {
      "External id": 179145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179145,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179145, "pid": 5, "tid": 7, "ts": 1716454224261470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261461, "dur": 8,
    "args": {
      "External id": 179145,
      "cbid": 211, "correlation": 179145
    }
  },
  {
    "ph": "s", "id": 179145, "pid": 76337, "tid": -914061504, "ts": 1716454224261461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224261524, "dur": 4,
    "args": {
      "External id": 179157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179157,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179157, "pid": 5, "tid": 7, "ts": 1716454224261524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261512, "dur": 12,
    "args": {
      "External id": 179157,
      "cbid": 211, "correlation": 179157
    }
  },
  {
    "ph": "s", "id": 179157, "pid": 76337, "tid": -914061504, "ts": 1716454224261512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224261569, "dur": 4,
    "args": {
      "External id": 179168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179168, "pid": 5, "tid": 7, "ts": 1716454224261569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261557, "dur": 11,
    "args": {
      "External id": 179168,
      "cbid": 211, "correlation": 179168
    }
  },
  {
    "ph": "s", "id": 179168, "pid": 76337, "tid": -914061504, "ts": 1716454224261557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224261600, "dur": 3,
    "args": {
      "External id": 179176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179176, "pid": 5, "tid": 7, "ts": 1716454224261600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261590, "dur": 9,
    "args": {
      "External id": 179176,
      "cbid": 211, "correlation": 179176
    }
  },
  {
    "ph": "s", "id": 179176, "pid": 76337, "tid": -914061504, "ts": 1716454224261590,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224261637, "dur": 5,
    "args": {
      "External id": 179184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179184,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179184, "pid": 5, "tid": 7, "ts": 1716454224261637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261626, "dur": 11,
    "args": {
      "External id": 179184,
      "cbid": 211, "correlation": 179184
    }
  },
  {
    "ph": "s", "id": 179184, "pid": 76337, "tid": -914061504, "ts": 1716454224261626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224261668, "dur": 5,
    "args": {
      "External id": 179192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179192,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179192, "pid": 5, "tid": 7, "ts": 1716454224261668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261658, "dur": 10,
    "args": {
      "External id": 179192,
      "cbid": 211, "correlation": 179192
    }
  },
  {
    "ph": "s", "id": 179192, "pid": 76337, "tid": -914061504, "ts": 1716454224261658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224261705, "dur": 4,
    "args": {
      "External id": 179201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179201,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179201, "pid": 5, "tid": 7, "ts": 1716454224261705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261694, "dur": 10,
    "args": {
      "External id": 179201,
      "cbid": 211, "correlation": 179201
    }
  },
  {
    "ph": "s", "id": 179201, "pid": 76337, "tid": -914061504, "ts": 1716454224261694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224261765, "dur": 4,
    "args": {
      "External id": 179214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179214,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179214, "pid": 5, "tid": 7, "ts": 1716454224261765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261753, "dur": 13,
    "args": {
      "External id": 179214,
      "cbid": 211, "correlation": 179214
    }
  },
  {
    "ph": "s", "id": 179214, "pid": 76337, "tid": -914061504, "ts": 1716454224261753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224261808, "dur": 5,
    "args": {
      "External id": 179224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179224,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179224, "pid": 5, "tid": 7, "ts": 1716454224261808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261797, "dur": 11,
    "args": {
      "External id": 179224,
      "cbid": 211, "correlation": 179224
    }
  },
  {
    "ph": "s", "id": 179224, "pid": 76337, "tid": -914061504, "ts": 1716454224261797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224261989, "dur": 5,
    "args": {
      "External id": 179241,
      "cbid": 251, "correlation": 179241
    }
  },
  {
    "ph": "f", "id": 179241, "pid": 76337, "tid": -914061504, "ts": 1716454224261989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454224262020, "dur": 12,
    "args": {
      "External id": 179243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179243,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179243, "pid": 5, "tid": 7, "ts": 1716454224262020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262004, "dur": 17,
    "args": {
      "External id": 179243,
      "cbid": 211, "correlation": 179243
    }
  },
  {
    "ph": "s", "id": 179243, "pid": 76337, "tid": -914061504, "ts": 1716454224262004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224262105, "dur": 3,
    "args": {
      "External id": 179251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179251,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179251, "pid": 5, "tid": 7, "ts": 1716454224262105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262092, "dur": 13,
    "args": {
      "External id": 179251,
      "cbid": 211, "correlation": 179251
    }
  },
  {
    "ph": "s", "id": 179251, "pid": 76337, "tid": -914061504, "ts": 1716454224262092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262170, "dur": 2,
    "args": {
      "External id": 179267,
      "cbid": 251, "correlation": 179267
    }
  },
  {
    "ph": "f", "id": 179267, "pid": 76337, "tid": -914061504, "ts": 1716454224262170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262177, "dur": 0,
    "args": {
      "External id": 179269,
      "cbid": 251, "correlation": 179269
    }
  },
  {
    "ph": "f", "id": 179269, "pid": 76337, "tid": -914061504, "ts": 1716454224262177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224262194, "dur": 13,
    "args": {
      "External id": 179270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179270,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179270, "pid": 5, "tid": 7, "ts": 1716454224262194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262180, "dur": 15,
    "args": {
      "External id": 179270,
      "cbid": 211, "correlation": 179270
    }
  },
  {
    "ph": "s", "id": 179270, "pid": 76337, "tid": -914061504, "ts": 1716454224262180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224262210, "dur": 5,
    "args": {
      "External id": 179272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179272,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179272, "pid": 5, "tid": 7, "ts": 1716454224262210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262200, "dur": 9,
    "args": {
      "External id": 179272,
      "cbid": 211, "correlation": 179272
    }
  },
  {
    "ph": "s", "id": 179272, "pid": 76337, "tid": -914061504, "ts": 1716454224262200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262339, "dur": 1,
    "args": {
      "External id": 179282,
      "cbid": 317, "correlation": 179282
    }
  },
  {
    "ph": "f", "id": 179282, "pid": 76337, "tid": -914061504, "ts": 1716454224262339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262341, "dur": 1,
    "args": {
      "External id": 179283,
      "cbid": 203, "correlation": 179283
    }
  },
  {
    "ph": "f", "id": 179283, "pid": 76337, "tid": -914061504, "ts": 1716454224262341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262343, "dur": 1,
    "args": {
      "External id": 179284,
      "cbid": 205, "correlation": 179284
    }
  },
  {
    "ph": "f", "id": 179284, "pid": 76337, "tid": -914061504, "ts": 1716454224262343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224262403, "dur": 6,
    "args": {
      "External id": 179288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179288,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179288, "pid": 5, "tid": 7, "ts": 1716454224262403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262386, "dur": 17,
    "args": {
      "External id": 179288,
      "cbid": 211, "correlation": 179288
    }
  },
  {
    "ph": "s", "id": 179288, "pid": 76337, "tid": -914061504, "ts": 1716454224262386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224262415, "dur": 4,
    "args": {
      "External id": 179290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179290,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 179290, "pid": 5, "tid": 7, "ts": 1716454224262415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262407, "dur": 7,
    "args": {
      "External id": 179290,
      "cbid": 211, "correlation": 179290
    }
  },
  {
    "ph": "s", "id": 179290, "pid": 76337, "tid": -914061504, "ts": 1716454224262407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224262436, "dur": 3,
    "args": {
      "External id": 179292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179292, "pid": 5, "tid": 7, "ts": 1716454224262436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262427, "dur": 8,
    "args": {
      "External id": 179292,
      "cbid": 211, "correlation": 179292
    }
  },
  {
    "ph": "s", "id": 179292, "pid": 76337, "tid": -914061504, "ts": 1716454224262427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262440, "dur": 0,
    "args": {
      "External id": 179293,
      "cbid": 51, "correlation": 179293
    }
  },
  {
    "ph": "s", "id": 179293, "pid": 76337, "tid": -914061504, "ts": 1716454224262440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224262450, "dur": 87,
    "args": {
      "External id": 179294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179294,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179294, "pid": 5, "tid": 7, "ts": 1716454224262450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262442, "dur": 7,
    "args": {
      "External id": 179294,
      "cbid": 211, "correlation": 179294
    }
  },
  {
    "ph": "s", "id": 179294, "pid": 76337, "tid": -914061504, "ts": 1716454224262442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224262539, "dur": 60,
    "args": {
      "External id": 179299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179299, "pid": 5, "tid": 7, "ts": 1716454224262539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224262481, "dur": 11,
    "args": {
      "External id": 179299,
      "cbid": 211, "correlation": 179299
    }
  },
  {
    "ph": "s", "id": 179299, "pid": 76337, "tid": -914061504, "ts": 1716454224262481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224264426, "dur": 52,
    "args": {
      "External id": 179319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179319,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 179319, "pid": 5, "tid": 7, "ts": 1716454224264426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264409, "dur": 17,
    "args": {
      "External id": 179319,
      "cbid": 211, "correlation": 179319
    }
  },
  {
    "ph": "s", "id": 179319, "pid": 76337, "tid": -914061504, "ts": 1716454224264409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224264479, "dur": 4,
    "args": {
      "External id": 179331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179331,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179331, "pid": 5, "tid": 7, "ts": 1716454224264479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264438, "dur": 8,
    "args": {
      "External id": 179331,
      "cbid": 211, "correlation": 179331
    }
  },
  {
    "ph": "s", "id": 179331, "pid": 76337, "tid": -914061504, "ts": 1716454224264438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224264485, "dur": 58,
    "args": {
      "External id": 179334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179334,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179334, "pid": 5, "tid": 7, "ts": 1716454224264485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264464, "dur": 8,
    "args": {
      "External id": 179334,
      "cbid": 211, "correlation": 179334
    }
  },
  {
    "ph": "s", "id": 179334, "pid": 76337, "tid": -914061504, "ts": 1716454224264464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224264544, "dur": 36,
    "args": {
      "External id": 179343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179343,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179343, "pid": 5, "tid": 7, "ts": 1716454224264544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264513, "dur": 11,
    "args": {
      "External id": 179343,
      "cbid": 211, "correlation": 179343
    }
  },
  {
    "ph": "s", "id": 179343, "pid": 76337, "tid": -914061504, "ts": 1716454224264513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264576, "dur": 0,
    "args": {
      "External id": 179353,
      "cbid": 317, "correlation": 179353
    }
  },
  {
    "ph": "f", "id": 179353, "pid": 76337, "tid": -914061504, "ts": 1716454224264576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264577, "dur": 0,
    "args": {
      "External id": 179354,
      "cbid": 203, "correlation": 179354
    }
  },
  {
    "ph": "f", "id": 179354, "pid": 76337, "tid": -914061504, "ts": 1716454224264577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264578, "dur": 0,
    "args": {
      "External id": 179355,
      "cbid": 205, "correlation": 179355
    }
  },
  {
    "ph": "f", "id": 179355, "pid": 76337, "tid": -914061504, "ts": 1716454224264578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224264613, "dur": 40,
    "args": {
      "External id": 179359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179359,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179359, "pid": 5, "tid": 7, "ts": 1716454224264613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264595, "dur": 17,
    "args": {
      "External id": 179359,
      "cbid": 211, "correlation": 179359
    }
  },
  {
    "ph": "s", "id": 179359, "pid": 76337, "tid": -914061504, "ts": 1716454224264595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224264654, "dur": 14,
    "args": {
      "External id": 179361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179361,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179361, "pid": 5, "tid": 7, "ts": 1716454224264654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264615, "dur": 6,
    "args": {
      "External id": 179361,
      "cbid": 211, "correlation": 179361
    }
  },
  {
    "ph": "s", "id": 179361, "pid": 76337, "tid": -914061504, "ts": 1716454224264615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224264669, "dur": 3,
    "args": {
      "External id": 179363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179363,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179363, "pid": 5, "tid": 7, "ts": 1716454224264669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264626, "dur": 6,
    "args": {
      "External id": 179363,
      "cbid": 211, "correlation": 179363
    }
  },
  {
    "ph": "s", "id": 179363, "pid": 76337, "tid": -914061504, "ts": 1716454224264626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264636, "dur": 0,
    "args": {
      "External id": 179364,
      "cbid": 51, "correlation": 179364
    }
  },
  {
    "ph": "s", "id": 179364, "pid": 76337, "tid": -914061504, "ts": 1716454224264636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224264674, "dur": 704,
    "args": {
      "External id": 179365,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179365,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179365, "pid": 5, "tid": 7, "ts": 1716454224264674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264638, "dur": 8,
    "args": {
      "External id": 179365,
      "cbid": 211, "correlation": 179365
    }
  },
  {
    "ph": "s", "id": 179365, "pid": 76337, "tid": -914061504, "ts": 1716454224264638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224265379, "dur": 60,
    "args": {
      "External id": 179370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179370,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179370, "pid": 5, "tid": 7, "ts": 1716454224265379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264669, "dur": 9,
    "args": {
      "External id": 179370,
      "cbid": 211, "correlation": 179370
    }
  },
  {
    "ph": "s", "id": 179370, "pid": 76337, "tid": -914061504, "ts": 1716454224264669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224265440, "dur": 4,
    "args": {
      "External id": 179378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179378,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179378, "pid": 5, "tid": 7, "ts": 1716454224265440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264715, "dur": 9,
    "args": {
      "External id": 179378,
      "cbid": 211, "correlation": 179378
    }
  },
  {
    "ph": "s", "id": 179378, "pid": 76337, "tid": -914061504, "ts": 1716454224264715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264784, "dur": 2,
    "args": {
      "External id": 179394,
      "cbid": 251, "correlation": 179394
    }
  },
  {
    "ph": "f", "id": 179394, "pid": 76337, "tid": -914061504, "ts": 1716454224264784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264790, "dur": 0,
    "args": {
      "External id": 179396,
      "cbid": 251, "correlation": 179396
    }
  },
  {
    "ph": "f", "id": 179396, "pid": 76337, "tid": -914061504, "ts": 1716454224264790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224265445, "dur": 9,
    "args": {
      "External id": 179397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179397,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 179397, "pid": 5, "tid": 7, "ts": 1716454224265445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264792, "dur": 13,
    "args": {
      "External id": 179397,
      "cbid": 211, "correlation": 179397
    }
  },
  {
    "ph": "s", "id": 179397, "pid": 76337, "tid": -914061504, "ts": 1716454224264792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224265455, "dur": 4,
    "args": {
      "External id": 179399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179399,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 179399, "pid": 5, "tid": 7, "ts": 1716454224265455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264807, "dur": 6,
    "args": {
      "External id": 179399,
      "cbid": 211, "correlation": 179399
    }
  },
  {
    "ph": "s", "id": 179399, "pid": 76337, "tid": -914061504, "ts": 1716454224264807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224265461, "dur": 55,
    "args": {
      "External id": 179409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179409, "pid": 5, "tid": 7, "ts": 1716454224265461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264868, "dur": 12,
    "args": {
      "External id": 179409,
      "cbid": 211, "correlation": 179409
    }
  },
  {
    "ph": "s", "id": 179409, "pid": 76337, "tid": -914061504, "ts": 1716454224264868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224265517, "dur": 52,
    "args": {
      "External id": 179429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179429,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 179429, "pid": 5, "tid": 7, "ts": 1716454224265517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264936, "dur": 11,
    "args": {
      "External id": 179429,
      "cbid": 211, "correlation": 179429
    }
  },
  {
    "ph": "s", "id": 179429, "pid": 76337, "tid": -914061504, "ts": 1716454224264936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224265571, "dur": 4,
    "args": {
      "External id": 179441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179441,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179441, "pid": 5, "tid": 7, "ts": 1716454224265571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264958, "dur": 6,
    "args": {
      "External id": 179441,
      "cbid": 211, "correlation": 179441
    }
  },
  {
    "ph": "s", "id": 179441, "pid": 76337, "tid": -914061504, "ts": 1716454224264958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224265576, "dur": 56,
    "args": {
      "External id": 179444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179444,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179444, "pid": 5, "tid": 7, "ts": 1716454224265576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224264986, "dur": 7,
    "args": {
      "External id": 179444,
      "cbid": 211, "correlation": 179444
    }
  },
  {
    "ph": "s", "id": 179444, "pid": 76337, "tid": -914061504, "ts": 1716454224264986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224265633, "dur": 37,
    "args": {
      "External id": 179453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179453,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179453, "pid": 5, "tid": 7, "ts": 1716454224265633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265029, "dur": 10,
    "args": {
      "External id": 179453,
      "cbid": 211, "correlation": 179453
    }
  },
  {
    "ph": "s", "id": 179453, "pid": 76337, "tid": -914061504, "ts": 1716454224265029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265109, "dur": 0,
    "args": {
      "External id": 179463,
      "cbid": 317, "correlation": 179463
    }
  },
  {
    "ph": "f", "id": 179463, "pid": 76337, "tid": -914061504, "ts": 1716454224265109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265110, "dur": 0,
    "args": {
      "External id": 179464,
      "cbid": 203, "correlation": 179464
    }
  },
  {
    "ph": "f", "id": 179464, "pid": 76337, "tid": -914061504, "ts": 1716454224265110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265110, "dur": 0,
    "args": {
      "External id": 179465,
      "cbid": 205, "correlation": 179465
    }
  },
  {
    "ph": "f", "id": 179465, "pid": 76337, "tid": -914061504, "ts": 1716454224265110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224265672, "dur": 39,
    "args": {
      "External id": 179469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179469,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179469, "pid": 5, "tid": 7, "ts": 1716454224265672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265125, "dur": 12,
    "args": {
      "External id": 179469,
      "cbid": 211, "correlation": 179469
    }
  },
  {
    "ph": "s", "id": 179469, "pid": 76337, "tid": -914061504, "ts": 1716454224265125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224265712, "dur": 14,
    "args": {
      "External id": 179471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179471,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179471, "pid": 5, "tid": 7, "ts": 1716454224265712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265139, "dur": 6,
    "args": {
      "External id": 179471,
      "cbid": 211, "correlation": 179471
    }
  },
  {
    "ph": "s", "id": 179471, "pid": 76337, "tid": -914061504, "ts": 1716454224265139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224265728, "dur": 3,
    "args": {
      "External id": 179473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179473,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179473, "pid": 5, "tid": 7, "ts": 1716454224265728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265149, "dur": 6,
    "args": {
      "External id": 179473,
      "cbid": 211, "correlation": 179473
    }
  },
  {
    "ph": "s", "id": 179473, "pid": 76337, "tid": -914061504, "ts": 1716454224265149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265159, "dur": 0,
    "args": {
      "External id": 179474,
      "cbid": 51, "correlation": 179474
    }
  },
  {
    "ph": "s", "id": 179474, "pid": 76337, "tid": -914061504, "ts": 1716454224265159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224265733, "dur": 700,
    "args": {
      "External id": 179475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179475,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179475, "pid": 5, "tid": 7, "ts": 1716454224265733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265160, "dur": 5,
    "args": {
      "External id": 179475,
      "cbid": 211, "correlation": 179475
    }
  },
  {
    "ph": "s", "id": 179475, "pid": 76337, "tid": -914061504, "ts": 1716454224265160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224266435, "dur": 60,
    "args": {
      "External id": 179480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179480, "pid": 5, "tid": 7, "ts": 1716454224266435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265188, "dur": 8,
    "args": {
      "External id": 179480,
      "cbid": 211, "correlation": 179480
    }
  },
  {
    "ph": "s", "id": 179480, "pid": 76337, "tid": -914061504, "ts": 1716454224265188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224266496, "dur": 50,
    "args": {
      "External id": 179488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179488,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179488, "pid": 5, "tid": 7, "ts": 1716454224266496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265221, "dur": 8,
    "args": {
      "External id": 179488,
      "cbid": 211, "correlation": 179488
    }
  },
  {
    "ph": "s", "id": 179488, "pid": 76337, "tid": -914061504, "ts": 1716454224265221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224266547, "dur": 35,
    "args": {
      "External id": 179496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179496,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179496, "pid": 5, "tid": 7, "ts": 1716454224266547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265252, "dur": 10,
    "args": {
      "External id": 179496,
      "cbid": 211, "correlation": 179496
    }
  },
  {
    "ph": "s", "id": 179496, "pid": 76337, "tid": -914061504, "ts": 1716454224265252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224266583, "dur": 53,
    "args": {
      "External id": 179516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179516,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 179516, "pid": 5, "tid": 7, "ts": 1716454224266583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265346, "dur": 13,
    "args": {
      "External id": 179516,
      "cbid": 211, "correlation": 179516
    }
  },
  {
    "ph": "s", "id": 179516, "pid": 76337, "tid": -914061504, "ts": 1716454224265346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224266638, "dur": 4,
    "args": {
      "External id": 179528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179528,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179528, "pid": 5, "tid": 7, "ts": 1716454224266638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265370, "dur": 6,
    "args": {
      "External id": 179528,
      "cbid": 211, "correlation": 179528
    }
  },
  {
    "ph": "s", "id": 179528, "pid": 76337, "tid": -914061504, "ts": 1716454224265370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224266643, "dur": 56,
    "args": {
      "External id": 179531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179531, "pid": 5, "tid": 7, "ts": 1716454224266643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265387, "dur": 6,
    "args": {
      "External id": 179531,
      "cbid": 211, "correlation": 179531
    }
  },
  {
    "ph": "s", "id": 179531, "pid": 76337, "tid": -914061504, "ts": 1716454224265387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265449, "dur": 0,
    "args": {
      "External id": 179542,
      "cbid": 317, "correlation": 179542
    }
  },
  {
    "ph": "f", "id": 179542, "pid": 76337, "tid": -914061504, "ts": 1716454224265449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265450, "dur": 0,
    "args": {
      "External id": 179543,
      "cbid": 203, "correlation": 179543
    }
  },
  {
    "ph": "f", "id": 179543, "pid": 76337, "tid": -914061504, "ts": 1716454224265450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265451, "dur": 0,
    "args": {
      "External id": 179544,
      "cbid": 205, "correlation": 179544
    }
  },
  {
    "ph": "f", "id": 179544, "pid": 76337, "tid": -914061504, "ts": 1716454224265451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265491, "dur": 3,
    "args": {
      "External id": 179548,
      "cbid": 251, "correlation": 179548
    }
  },
  {
    "ph": "f", "id": 179548, "pid": 76337, "tid": -914061504, "ts": 1716454224265491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265495, "dur": 1,
    "args": {
      "External id": 179549,
      "cbid": 251, "correlation": 179549
    }
  },
  {
    "ph": "f", "id": 179549, "pid": 76337, "tid": -914061504, "ts": 1716454224265495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265497, "dur": 1,
    "args": {
      "External id": 179550,
      "cbid": 251, "correlation": 179550
    }
  },
  {
    "ph": "f", "id": 179550, "pid": 76337, "tid": -914061504, "ts": 1716454224265497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265499, "dur": 1,
    "args": {
      "External id": 179551,
      "cbid": 251, "correlation": 179551
    }
  },
  {
    "ph": "f", "id": 179551, "pid": 76337, "tid": -914061504, "ts": 1716454224265499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265501, "dur": 1,
    "args": {
      "External id": 179552,
      "cbid": 251, "correlation": 179552
    }
  },
  {
    "ph": "f", "id": 179552, "pid": 76337, "tid": -914061504, "ts": 1716454224265501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265503, "dur": 1,
    "args": {
      "External id": 179553,
      "cbid": 251, "correlation": 179553
    }
  },
  {
    "ph": "f", "id": 179553, "pid": 76337, "tid": -914061504, "ts": 1716454224265503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265505, "dur": 1,
    "args": {
      "External id": 179554,
      "cbid": 251, "correlation": 179554
    }
  },
  {
    "ph": "f", "id": 179554, "pid": 76337, "tid": -914061504, "ts": 1716454224265505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265507, "dur": 1,
    "args": {
      "External id": 179555,
      "cbid": 251, "correlation": 179555
    }
  },
  {
    "ph": "f", "id": 179555, "pid": 76337, "tid": -914061504, "ts": 1716454224265507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265510, "dur": 0,
    "args": {
      "External id": 179556,
      "cbid": 251, "correlation": 179556
    }
  },
  {
    "ph": "f", "id": 179556, "pid": 76337, "tid": -914061504, "ts": 1716454224265510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224266701, "dur": 114,
    "args": {
      "External id": 179557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179557,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 179557, "pid": 5, "tid": 7, "ts": 1716454224266701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265514, "dur": 15,
    "args": {
      "External id": 179557,
      "cbid": 211, "correlation": 179557
    }
  },
  {
    "ph": "s", "id": 179557, "pid": 76337, "tid": -914061504, "ts": 1716454224265514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224266816, "dur": 60,
    "args": {
      "External id": 179563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179563,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179563, "pid": 5, "tid": 7, "ts": 1716454224266816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265552, "dur": 9,
    "args": {
      "External id": 179563,
      "cbid": 211, "correlation": 179563
    }
  },
  {
    "ph": "s", "id": 179563, "pid": 76337, "tid": -914061504, "ts": 1716454224265552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224266877, "dur": 575,
    "args": {
      "External id": 179572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179572,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179572, "pid": 5, "tid": 7, "ts": 1716454224266877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265658, "dur": 16,
    "args": {
      "External id": 179572,
      "cbid": 211, "correlation": 179572
    }
  },
  {
    "ph": "s", "id": 179572, "pid": 76337, "tid": -914061504, "ts": 1716454224265658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224267453, "dur": 183,
    "args": {
      "External id": 179594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179594,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179594, "pid": 5, "tid": 7, "ts": 1716454224267453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265735, "dur": 12,
    "args": {
      "External id": 179594,
      "cbid": 211, "correlation": 179594
    }
  },
  {
    "ph": "s", "id": 179594, "pid": 76337, "tid": -914061504, "ts": 1716454224265735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265878, "dur": 2,
    "args": {
      "External id": 179605,
      "cbid": 251, "correlation": 179605
    }
  },
  {
    "ph": "f", "id": 179605, "pid": 76337, "tid": -914061504, "ts": 1716454224265878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224267637, "dur": 196,
    "args": {
      "External id": 179606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179606,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179606, "pid": 5, "tid": 7, "ts": 1716454224267637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265885, "dur": 15,
    "args": {
      "External id": 179606,
      "cbid": 211, "correlation": 179606
    }
  },
  {
    "ph": "s", "id": 179606, "pid": 76337, "tid": -914061504, "ts": 1716454224265885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265960, "dur": 1,
    "args": {
      "External id": 179617,
      "cbid": 251, "correlation": 179617
    }
  },
  {
    "ph": "f", "id": 179617, "pid": 76337, "tid": -914061504, "ts": 1716454224265960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224267835, "dur": 190,
    "args": {
      "External id": 179618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179618,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179618, "pid": 5, "tid": 7, "ts": 1716454224267835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224265964, "dur": 19,
    "args": {
      "External id": 179618,
      "cbid": 211, "correlation": 179618
    }
  },
  {
    "ph": "s", "id": 179618, "pid": 76337, "tid": -914061504, "ts": 1716454224265964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266039, "dur": 1,
    "args": {
      "External id": 179629,
      "cbid": 251, "correlation": 179629
    }
  },
  {
    "ph": "f", "id": 179629, "pid": 76337, "tid": -914061504, "ts": 1716454224266039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224268026, "dur": 186,
    "args": {
      "External id": 179630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179630,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179630, "pid": 5, "tid": 7, "ts": 1716454224268026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266043, "dur": 13,
    "args": {
      "External id": 179630,
      "cbid": 211, "correlation": 179630
    }
  },
  {
    "ph": "s", "id": 179630, "pid": 76337, "tid": -914061504, "ts": 1716454224266043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224268213, "dur": 18802,
    "args": {
      "External id": 179651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179651,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 179651, "pid": 5, "tid": 7, "ts": 1716454224268213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266166, "dur": 15,
    "args": {
      "External id": 179651,
      "cbid": 211, "correlation": 179651
    }
  },
  {
    "ph": "s", "id": 179651, "pid": 76337, "tid": -914061504, "ts": 1716454224266166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266306, "dur": 2,
    "args": {
      "External id": 179669,
      "cbid": 251, "correlation": 179669
    }
  },
  {
    "ph": "f", "id": 179669, "pid": 76337, "tid": -914061504, "ts": 1716454224266306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224287017, "dur": 204,
    "args": {
      "External id": 179671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179671,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179671, "pid": 5, "tid": 7, "ts": 1716454224287017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266313, "dur": 13,
    "args": {
      "External id": 179671,
      "cbid": 211, "correlation": 179671
    }
  },
  {
    "ph": "s", "id": 179671, "pid": 76337, "tid": -914061504, "ts": 1716454224266313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224287222, "dur": 66,
    "args": {
      "External id": 179679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179679,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179679, "pid": 5, "tid": 7, "ts": 1716454224287222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266387, "dur": 13,
    "args": {
      "External id": 179679,
      "cbid": 211, "correlation": 179679
    }
  },
  {
    "ph": "s", "id": 179679, "pid": 76337, "tid": -914061504, "ts": 1716454224266387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224287289, "dur": 96,
    "args": {
      "External id": 179687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179687,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179687, "pid": 5, "tid": 7, "ts": 1716454224287289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266428, "dur": 9,
    "args": {
      "External id": 179687,
      "cbid": 211, "correlation": 179687
    }
  },
  {
    "ph": "s", "id": 179687, "pid": 76337, "tid": -914061504, "ts": 1716454224266428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224287387, "dur": 55,
    "args": {
      "External id": 179698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179698,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179698, "pid": 5, "tid": 7, "ts": 1716454224287387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266520, "dur": 16,
    "args": {
      "External id": 179698,
      "cbid": 211, "correlation": 179698
    }
  },
  {
    "ph": "s", "id": 179698, "pid": 76337, "tid": -914061504, "ts": 1716454224266520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224287443, "dur": 93,
    "args": {
      "External id": 179720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179720,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179720, "pid": 5, "tid": 7, "ts": 1716454224287443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266556, "dur": 7,
    "args": {
      "External id": 179720,
      "cbid": 211, "correlation": 179720
    }
  },
  {
    "ph": "s", "id": 179720, "pid": 76337, "tid": -914061504, "ts": 1716454224266556,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266640, "dur": 1,
    "args": {
      "External id": 179731,
      "cbid": 251, "correlation": 179731
    }
  },
  {
    "ph": "f", "id": 179731, "pid": 76337, "tid": -914061504, "ts": 1716454224266640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224287537, "dur": 107,
    "args": {
      "External id": 179732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179732,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179732, "pid": 5, "tid": 7, "ts": 1716454224287537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266646, "dur": 13,
    "args": {
      "External id": 179732,
      "cbid": 211, "correlation": 179732
    }
  },
  {
    "ph": "s", "id": 179732, "pid": 76337, "tid": -914061504, "ts": 1716454224266646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266741, "dur": 2,
    "args": {
      "External id": 179743,
      "cbid": 251, "correlation": 179743
    }
  },
  {
    "ph": "f", "id": 179743, "pid": 76337, "tid": -914061504, "ts": 1716454224266741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266747, "dur": 0,
    "args": {
      "External id": 179744,
      "cbid": 251, "correlation": 179744
    }
  },
  {
    "ph": "f", "id": 179744, "pid": 76337, "tid": -914061504, "ts": 1716454224266747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224287646, "dur": 11,
    "args": {
      "External id": 179745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179745,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179745, "pid": 5, "tid": 7, "ts": 1716454224287646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266749, "dur": 15,
    "args": {
      "External id": 179745,
      "cbid": 211, "correlation": 179745
    }
  },
  {
    "ph": "s", "id": 179745, "pid": 76337, "tid": -914061504, "ts": 1716454224266749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224287658, "dur": 5,
    "args": {
      "External id": 179747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179747,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 179747, "pid": 5, "tid": 7, "ts": 1716454224287658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266768, "dur": 7,
    "args": {
      "External id": 179747,
      "cbid": 211, "correlation": 179747
    }
  },
  {
    "ph": "s", "id": 179747, "pid": 76337, "tid": -914061504, "ts": 1716454224266768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266831, "dur": 1,
    "args": {
      "External id": 179758,
      "cbid": 251, "correlation": 179758
    }
  },
  {
    "ph": "f", "id": 179758, "pid": 76337, "tid": -914061504, "ts": 1716454224266831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266835, "dur": 0,
    "args": {
      "External id": 179759,
      "cbid": 251, "correlation": 179759
    }
  },
  {
    "ph": "f", "id": 179759, "pid": 76337, "tid": -914061504, "ts": 1716454224266835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224287664, "dur": 6,
    "args": {
      "External id": 179760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179760,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 179760, "pid": 5, "tid": 7, "ts": 1716454224287664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266836, "dur": 11,
    "args": {
      "External id": 179760,
      "cbid": 211, "correlation": 179760
    }
  },
  {
    "ph": "s", "id": 179760, "pid": 76337, "tid": -914061504, "ts": 1716454224266836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224287672, "dur": 3,
    "args": {
      "External id": 179762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179762,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 179762, "pid": 5, "tid": 7, "ts": 1716454224287672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266849, "dur": 6,
    "args": {
      "External id": 179762,
      "cbid": 211, "correlation": 179762
    }
  },
  {
    "ph": "s", "id": 179762, "pid": 76337, "tid": -914061504, "ts": 1716454224266849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224287677, "dur": 157,
    "args": {
      "External id": 179783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179783,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 179783, "pid": 5, "tid": 7, "ts": 1716454224287677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224266923, "dur": 12,
    "args": {
      "External id": 179783,
      "cbid": 211, "correlation": 179783
    }
  },
  {
    "ph": "s", "id": 179783, "pid": 76337, "tid": -914061504, "ts": 1716454224266923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267029, "dur": 2,
    "args": {
      "External id": 179801,
      "cbid": 251, "correlation": 179801
    }
  },
  {
    "ph": "f", "id": 179801, "pid": 76337, "tid": -914061504, "ts": 1716454224267029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224287835, "dur": 107,
    "args": {
      "External id": 179803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179803,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 179803, "pid": 5, "tid": 7, "ts": 1716454224287835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267035, "dur": 14,
    "args": {
      "External id": 179803,
      "cbid": 211, "correlation": 179803
    }
  },
  {
    "ph": "s", "id": 179803, "pid": 76337, "tid": -914061504, "ts": 1716454224267035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224287943, "dur": 34,
    "args": {
      "External id": 179811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179811,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179811, "pid": 5, "tid": 7, "ts": 1716454224287943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267108, "dur": 12,
    "args": {
      "External id": 179811,
      "cbid": 211, "correlation": 179811
    }
  },
  {
    "ph": "s", "id": 179811, "pid": 76337, "tid": -914061504, "ts": 1716454224267108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224287978, "dur": 68,
    "args": {
      "External id": 179819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179819,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179819, "pid": 5, "tid": 7, "ts": 1716454224287978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267148, "dur": 10,
    "args": {
      "External id": 179819,
      "cbid": 211, "correlation": 179819
    }
  },
  {
    "ph": "s", "id": 179819, "pid": 76337, "tid": -914061504, "ts": 1716454224267148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224288048, "dur": 92,
    "args": {
      "External id": 179841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179841,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179841, "pid": 5, "tid": 7, "ts": 1716454224288048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267200, "dur": 11,
    "args": {
      "External id": 179841,
      "cbid": 211, "correlation": 179841
    }
  },
  {
    "ph": "s", "id": 179841, "pid": 76337, "tid": -914061504, "ts": 1716454224267200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267301, "dur": 1,
    "args": {
      "External id": 179857,
      "cbid": 251, "correlation": 179857
    }
  },
  {
    "ph": "f", "id": 179857, "pid": 76337, "tid": -914061504, "ts": 1716454224267301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224288142, "dur": 578,
    "args": {
      "External id": 179859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179859,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 179859, "pid": 5, "tid": 7, "ts": 1716454224288142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267307, "dur": 14,
    "args": {
      "External id": 179859,
      "cbid": 211, "correlation": 179859
    }
  },
  {
    "ph": "s", "id": 179859, "pid": 76337, "tid": -914061504, "ts": 1716454224267307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224288721, "dur": 246,
    "args": {
      "External id": 179867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179867, "pid": 5, "tid": 7, "ts": 1716454224288721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267392, "dur": 15,
    "args": {
      "External id": 179867,
      "cbid": 211, "correlation": 179867
    }
  },
  {
    "ph": "s", "id": 179867, "pid": 76337, "tid": -914061504, "ts": 1716454224267392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224288968, "dur": 252,
    "args": {
      "External id": 179875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179875,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179875, "pid": 5, "tid": 7, "ts": 1716454224288968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267428, "dur": 9,
    "args": {
      "External id": 179875,
      "cbid": 211, "correlation": 179875
    }
  },
  {
    "ph": "s", "id": 179875, "pid": 76337, "tid": -914061504, "ts": 1716454224267428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267516, "dur": 2,
    "args": {
      "External id": 179891,
      "cbid": 251, "correlation": 179891
    }
  },
  {
    "ph": "f", "id": 179891, "pid": 76337, "tid": -914061504, "ts": 1716454224267516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267522, "dur": 0,
    "args": {
      "External id": 179893,
      "cbid": 251, "correlation": 179893
    }
  },
  {
    "ph": "f", "id": 179893, "pid": 76337, "tid": -914061504, "ts": 1716454224267522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224289222, "dur": 360,
    "args": {
      "External id": 179894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179894,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 179894, "pid": 5, "tid": 7, "ts": 1716454224289222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267527, "dur": 14,
    "args": {
      "External id": 179894,
      "cbid": 211, "correlation": 179894
    }
  },
  {
    "ph": "s", "id": 179894, "pid": 76337, "tid": -914061504, "ts": 1716454224267527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224289583, "dur": 50,
    "args": {
      "External id": 179902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179902,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179902, "pid": 5, "tid": 7, "ts": 1716454224289583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267571, "dur": 10,
    "args": {
      "External id": 179902,
      "cbid": 211, "correlation": 179902
    }
  },
  {
    "ph": "s", "id": 179902, "pid": 76337, "tid": -914061504, "ts": 1716454224267571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224289635, "dur": 159,
    "args": {
      "External id": 179913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179913,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179913, "pid": 5, "tid": 7, "ts": 1716454224289635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267642, "dur": 12,
    "args": {
      "External id": 179913,
      "cbid": 211, "correlation": 179913
    }
  },
  {
    "ph": "s", "id": 179913, "pid": 76337, "tid": -914061504, "ts": 1716454224267642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267712, "dur": 0,
    "args": {
      "External id": 179925,
      "cbid": 317, "correlation": 179925
    }
  },
  {
    "ph": "f", "id": 179925, "pid": 76337, "tid": -914061504, "ts": 1716454224267712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267713, "dur": 0,
    "args": {
      "External id": 179926,
      "cbid": 203, "correlation": 179926
    }
  },
  {
    "ph": "f", "id": 179926, "pid": 76337, "tid": -914061504, "ts": 1716454224267713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267714, "dur": 0,
    "args": {
      "External id": 179927,
      "cbid": 205, "correlation": 179927
    }
  },
  {
    "ph": "f", "id": 179927, "pid": 76337, "tid": -914061504, "ts": 1716454224267714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267737, "dur": 1,
    "args": {
      "External id": 179931,
      "cbid": 251, "correlation": 179931
    }
  },
  {
    "ph": "f", "id": 179931, "pid": 76337, "tid": -914061504, "ts": 1716454224267737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267739, "dur": 0,
    "args": {
      "External id": 179932,
      "cbid": 251, "correlation": 179932
    }
  },
  {
    "ph": "f", "id": 179932, "pid": 76337, "tid": -914061504, "ts": 1716454224267739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267740, "dur": 0,
    "args": {
      "External id": 179933,
      "cbid": 251, "correlation": 179933
    }
  },
  {
    "ph": "f", "id": 179933, "pid": 76337, "tid": -914061504, "ts": 1716454224267740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267740, "dur": 0,
    "args": {
      "External id": 179934,
      "cbid": 251, "correlation": 179934
    }
  },
  {
    "ph": "f", "id": 179934, "pid": 76337, "tid": -914061504, "ts": 1716454224267740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267741, "dur": 0,
    "args": {
      "External id": 179935,
      "cbid": 251, "correlation": 179935
    }
  },
  {
    "ph": "f", "id": 179935, "pid": 76337, "tid": -914061504, "ts": 1716454224267741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267742, "dur": 0,
    "args": {
      "External id": 179936,
      "cbid": 251, "correlation": 179936
    }
  },
  {
    "ph": "f", "id": 179936, "pid": 76337, "tid": -914061504, "ts": 1716454224267742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267743, "dur": 0,
    "args": {
      "External id": 179937,
      "cbid": 251, "correlation": 179937
    }
  },
  {
    "ph": "f", "id": 179937, "pid": 76337, "tid": -914061504, "ts": 1716454224267743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267744, "dur": 0,
    "args": {
      "External id": 179938,
      "cbid": 251, "correlation": 179938
    }
  },
  {
    "ph": "f", "id": 179938, "pid": 76337, "tid": -914061504, "ts": 1716454224267744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267745, "dur": 0,
    "args": {
      "External id": 179939,
      "cbid": 251, "correlation": 179939
    }
  },
  {
    "ph": "f", "id": 179939, "pid": 76337, "tid": -914061504, "ts": 1716454224267745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224289794, "dur": 114,
    "args": {
      "External id": 179940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179940,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 179940, "pid": 5, "tid": 7, "ts": 1716454224289794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267747, "dur": 13,
    "args": {
      "External id": 179940,
      "cbid": 211, "correlation": 179940
    }
  },
  {
    "ph": "s", "id": 179940, "pid": 76337, "tid": -914061504, "ts": 1716454224267747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224289910, "dur": 60,
    "args": {
      "External id": 179946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179946,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179946, "pid": 5, "tid": 7, "ts": 1716454224289910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267783, "dur": 9,
    "args": {
      "External id": 179946,
      "cbid": 211, "correlation": 179946
    }
  },
  {
    "ph": "s", "id": 179946, "pid": 76337, "tid": -914061504, "ts": 1716454224267783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224289971, "dur": 50,
    "args": {
      "External id": 179954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179954,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179954, "pid": 5, "tid": 7, "ts": 1716454224289971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267817, "dur": 8,
    "args": {
      "External id": 179954,
      "cbid": 211, "correlation": 179954
    }
  },
  {
    "ph": "s", "id": 179954, "pid": 76337, "tid": -914061504, "ts": 1716454224267817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224290023, "dur": 52,
    "args": {
      "External id": 179974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179974,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 179974, "pid": 5, "tid": 7, "ts": 1716454224290023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267890, "dur": 12,
    "args": {
      "External id": 179974,
      "cbid": 211, "correlation": 179974
    }
  },
  {
    "ph": "s", "id": 179974, "pid": 76337, "tid": -914061504, "ts": 1716454224267890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224290076, "dur": 4,
    "args": {
      "External id": 179986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179986,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 179986, "pid": 5, "tid": 7, "ts": 1716454224290076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267911, "dur": 7,
    "args": {
      "External id": 179986,
      "cbid": 211, "correlation": 179986
    }
  },
  {
    "ph": "s", "id": 179986, "pid": 76337, "tid": -914061504, "ts": 1716454224267911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224290082, "dur": 55,
    "args": {
      "External id": 179989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179989, "pid": 5, "tid": 7, "ts": 1716454224290082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267930, "dur": 7,
    "args": {
      "External id": 179989,
      "cbid": 211, "correlation": 179989
    }
  },
  {
    "ph": "s", "id": 179989, "pid": 76337, "tid": -914061504, "ts": 1716454224267930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224290139, "dur": 37,
    "args": {
      "External id": 179998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 179998,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 179998, "pid": 5, "tid": 7, "ts": 1716454224290139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224267969, "dur": 18,
    "args": {
      "External id": 179998,
      "cbid": 211, "correlation": 179998
    }
  },
  {
    "ph": "s", "id": 179998, "pid": 76337, "tid": -914061504, "ts": 1716454224267969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268031, "dur": 0,
    "args": {
      "External id": 180008,
      "cbid": 317, "correlation": 180008
    }
  },
  {
    "ph": "f", "id": 180008, "pid": 76337, "tid": -914061504, "ts": 1716454224268031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268031, "dur": 0,
    "args": {
      "External id": 180009,
      "cbid": 203, "correlation": 180009
    }
  },
  {
    "ph": "f", "id": 180009, "pid": 76337, "tid": -914061504, "ts": 1716454224268031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268032, "dur": 0,
    "args": {
      "External id": 180010,
      "cbid": 205, "correlation": 180010
    }
  },
  {
    "ph": "f", "id": 180010, "pid": 76337, "tid": -914061504, "ts": 1716454224268032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224290177, "dur": 40,
    "args": {
      "External id": 180014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180014, "pid": 5, "tid": 7, "ts": 1716454224290177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268046, "dur": 11,
    "args": {
      "External id": 180014,
      "cbid": 211, "correlation": 180014
    }
  },
  {
    "ph": "s", "id": 180014, "pid": 76337, "tid": -914061504, "ts": 1716454224268046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224290218, "dur": 14,
    "args": {
      "External id": 180016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180016,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180016, "pid": 5, "tid": 7, "ts": 1716454224290218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268060, "dur": 5,
    "args": {
      "External id": 180016,
      "cbid": 211, "correlation": 180016
    }
  },
  {
    "ph": "s", "id": 180016, "pid": 76337, "tid": -914061504, "ts": 1716454224268060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224290234, "dur": 4,
    "args": {
      "External id": 180018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180018, "pid": 5, "tid": 7, "ts": 1716454224290234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268070, "dur": 6,
    "args": {
      "External id": 180018,
      "cbid": 211, "correlation": 180018
    }
  },
  {
    "ph": "s", "id": 180018, "pid": 76337, "tid": -914061504, "ts": 1716454224268070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268079, "dur": 0,
    "args": {
      "External id": 180019,
      "cbid": 51, "correlation": 180019
    }
  },
  {
    "ph": "s", "id": 180019, "pid": 76337, "tid": -914061504, "ts": 1716454224268079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224290239, "dur": 706,
    "args": {
      "External id": 180020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180020,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180020, "pid": 5, "tid": 7, "ts": 1716454224290239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268080, "dur": 5,
    "args": {
      "External id": 180020,
      "cbid": 211, "correlation": 180020
    }
  },
  {
    "ph": "s", "id": 180020, "pid": 76337, "tid": -914061504, "ts": 1716454224268080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224290947, "dur": 60,
    "args": {
      "External id": 180025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180025, "pid": 5, "tid": 7, "ts": 1716454224290947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268107, "dur": 9,
    "args": {
      "External id": 180025,
      "cbid": 211, "correlation": 180025
    }
  },
  {
    "ph": "s", "id": 180025, "pid": 76337, "tid": -914061504, "ts": 1716454224268107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224291008, "dur": 3,
    "args": {
      "External id": 180033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180033,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180033, "pid": 5, "tid": 7, "ts": 1716454224291008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268150, "dur": 9,
    "args": {
      "External id": 180033,
      "cbid": 211, "correlation": 180033
    }
  },
  {
    "ph": "s", "id": 180033, "pid": 76337, "tid": -914061504, "ts": 1716454224268150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268215, "dur": 2,
    "args": {
      "External id": 180049,
      "cbid": 251, "correlation": 180049
    }
  },
  {
    "ph": "f", "id": 180049, "pid": 76337, "tid": -914061504, "ts": 1716454224268215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268221, "dur": 0,
    "args": {
      "External id": 180051,
      "cbid": 251, "correlation": 180051
    }
  },
  {
    "ph": "f", "id": 180051, "pid": 76337, "tid": -914061504, "ts": 1716454224268221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224291012, "dur": 11,
    "args": {
      "External id": 180052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180052,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 180052, "pid": 5, "tid": 7, "ts": 1716454224291012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268223, "dur": 11,
    "args": {
      "External id": 180052,
      "cbid": 211, "correlation": 180052
    }
  },
  {
    "ph": "s", "id": 180052, "pid": 76337, "tid": -914061504, "ts": 1716454224268223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224291025, "dur": 5,
    "args": {
      "External id": 180054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180054,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 180054, "pid": 5, "tid": 7, "ts": 1716454224291025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268236, "dur": 6,
    "args": {
      "External id": 180054,
      "cbid": 211, "correlation": 180054
    }
  },
  {
    "ph": "s", "id": 180054, "pid": 76337, "tid": -914061504, "ts": 1716454224268236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224291031, "dur": 53,
    "args": {
      "External id": 180064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180064, "pid": 5, "tid": 7, "ts": 1716454224291031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268293, "dur": 12,
    "args": {
      "External id": 180064,
      "cbid": 211, "correlation": 180064
    }
  },
  {
    "ph": "s", "id": 180064, "pid": 76337, "tid": -914061504, "ts": 1716454224268293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224291086, "dur": 52,
    "args": {
      "External id": 180084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180084,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 180084, "pid": 5, "tid": 7, "ts": 1716454224291086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268360, "dur": 11,
    "args": {
      "External id": 180084,
      "cbid": 211, "correlation": 180084
    }
  },
  {
    "ph": "s", "id": 180084, "pid": 76337, "tid": -914061504, "ts": 1716454224268360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224291139, "dur": 4,
    "args": {
      "External id": 180096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180096,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180096, "pid": 5, "tid": 7, "ts": 1716454224291139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268380, "dur": 6,
    "args": {
      "External id": 180096,
      "cbid": 211, "correlation": 180096
    }
  },
  {
    "ph": "s", "id": 180096, "pid": 76337, "tid": -914061504, "ts": 1716454224268380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224291144, "dur": 54,
    "args": {
      "External id": 180099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180099, "pid": 5, "tid": 7, "ts": 1716454224291144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268399, "dur": 6,
    "args": {
      "External id": 180099,
      "cbid": 211, "correlation": 180099
    }
  },
  {
    "ph": "s", "id": 180099, "pid": 76337, "tid": -914061504, "ts": 1716454224268399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224291200, "dur": 37,
    "args": {
      "External id": 180108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180108,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180108, "pid": 5, "tid": 7, "ts": 1716454224291200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268439, "dur": 10,
    "args": {
      "External id": 180108,
      "cbid": 211, "correlation": 180108
    }
  },
  {
    "ph": "s", "id": 180108, "pid": 76337, "tid": -914061504, "ts": 1716454224268439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268503, "dur": 0,
    "args": {
      "External id": 180118,
      "cbid": 317, "correlation": 180118
    }
  },
  {
    "ph": "f", "id": 180118, "pid": 76337, "tid": -914061504, "ts": 1716454224268503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268503, "dur": 0,
    "args": {
      "External id": 180119,
      "cbid": 203, "correlation": 180119
    }
  },
  {
    "ph": "f", "id": 180119, "pid": 76337, "tid": -914061504, "ts": 1716454224268503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268504, "dur": 0,
    "args": {
      "External id": 180120,
      "cbid": 205, "correlation": 180120
    }
  },
  {
    "ph": "f", "id": 180120, "pid": 76337, "tid": -914061504, "ts": 1716454224268504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224291238, "dur": 39,
    "args": {
      "External id": 180124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180124, "pid": 5, "tid": 7, "ts": 1716454224291238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268518, "dur": 13,
    "args": {
      "External id": 180124,
      "cbid": 211, "correlation": 180124
    }
  },
  {
    "ph": "s", "id": 180124, "pid": 76337, "tid": -914061504, "ts": 1716454224268518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224291278, "dur": 14,
    "args": {
      "External id": 180126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180126,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180126, "pid": 5, "tid": 7, "ts": 1716454224291278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268533, "dur": 5,
    "args": {
      "External id": 180126,
      "cbid": 211, "correlation": 180126
    }
  },
  {
    "ph": "s", "id": 180126, "pid": 76337, "tid": -914061504, "ts": 1716454224268533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224291294, "dur": 3,
    "args": {
      "External id": 180128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180128,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180128, "pid": 5, "tid": 7, "ts": 1716454224291294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268542, "dur": 5,
    "args": {
      "External id": 180128,
      "cbid": 211, "correlation": 180128
    }
  },
  {
    "ph": "s", "id": 180128, "pid": 76337, "tid": -914061504, "ts": 1716454224268542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268550, "dur": 0,
    "args": {
      "External id": 180129,
      "cbid": 51, "correlation": 180129
    }
  },
  {
    "ph": "s", "id": 180129, "pid": 76337, "tid": -914061504, "ts": 1716454224268550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224291298, "dur": 697,
    "args": {
      "External id": 180130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180130,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180130, "pid": 5, "tid": 7, "ts": 1716454224291298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268551, "dur": 5,
    "args": {
      "External id": 180130,
      "cbid": 211, "correlation": 180130
    }
  },
  {
    "ph": "s", "id": 180130, "pid": 76337, "tid": -914061504, "ts": 1716454224268551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224291997, "dur": 59,
    "args": {
      "External id": 180135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180135, "pid": 5, "tid": 7, "ts": 1716454224291997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268578, "dur": 10,
    "args": {
      "External id": 180135,
      "cbid": 211, "correlation": 180135
    }
  },
  {
    "ph": "s", "id": 180135, "pid": 76337, "tid": -914061504, "ts": 1716454224268578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224292057, "dur": 50,
    "args": {
      "External id": 180143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180143,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180143, "pid": 5, "tid": 7, "ts": 1716454224292057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268611, "dur": 9,
    "args": {
      "External id": 180143,
      "cbid": 211, "correlation": 180143
    }
  },
  {
    "ph": "s", "id": 180143, "pid": 76337, "tid": -914061504, "ts": 1716454224268611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224292108, "dur": 35,
    "args": {
      "External id": 180151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180151,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180151, "pid": 5, "tid": 7, "ts": 1716454224292108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268641, "dur": 9,
    "args": {
      "External id": 180151,
      "cbid": 211, "correlation": 180151
    }
  },
  {
    "ph": "s", "id": 180151, "pid": 76337, "tid": -914061504, "ts": 1716454224268641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224292144, "dur": 51,
    "args": {
      "External id": 180171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180171,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 180171, "pid": 5, "tid": 7, "ts": 1716454224292144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268725, "dur": 13,
    "args": {
      "External id": 180171,
      "cbid": 211, "correlation": 180171
    }
  },
  {
    "ph": "s", "id": 180171, "pid": 76337, "tid": -914061504, "ts": 1716454224268725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224292197, "dur": 4,
    "args": {
      "External id": 180183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180183,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180183, "pid": 5, "tid": 7, "ts": 1716454224292197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268747, "dur": 6,
    "args": {
      "External id": 180183,
      "cbid": 211, "correlation": 180183
    }
  },
  {
    "ph": "s", "id": 180183, "pid": 76337, "tid": -914061504, "ts": 1716454224268747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224292202, "dur": 55,
    "args": {
      "External id": 180186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180186, "pid": 5, "tid": 7, "ts": 1716454224292202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268764, "dur": 6,
    "args": {
      "External id": 180186,
      "cbid": 211, "correlation": 180186
    }
  },
  {
    "ph": "s", "id": 180186, "pid": 76337, "tid": -914061504, "ts": 1716454224268764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268821, "dur": 0,
    "args": {
      "External id": 180197,
      "cbid": 317, "correlation": 180197
    }
  },
  {
    "ph": "f", "id": 180197, "pid": 76337, "tid": -914061504, "ts": 1716454224268821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268821, "dur": 0,
    "args": {
      "External id": 180198,
      "cbid": 203, "correlation": 180198
    }
  },
  {
    "ph": "f", "id": 180198, "pid": 76337, "tid": -914061504, "ts": 1716454224268821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268822, "dur": 0,
    "args": {
      "External id": 180199,
      "cbid": 205, "correlation": 180199
    }
  },
  {
    "ph": "f", "id": 180199, "pid": 76337, "tid": -914061504, "ts": 1716454224268822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268844, "dur": 1,
    "args": {
      "External id": 180203,
      "cbid": 251, "correlation": 180203
    }
  },
  {
    "ph": "f", "id": 180203, "pid": 76337, "tid": -914061504, "ts": 1716454224268844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268846, "dur": 0,
    "args": {
      "External id": 180204,
      "cbid": 251, "correlation": 180204
    }
  },
  {
    "ph": "f", "id": 180204, "pid": 76337, "tid": -914061504, "ts": 1716454224268846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268847, "dur": 0,
    "args": {
      "External id": 180205,
      "cbid": 251, "correlation": 180205
    }
  },
  {
    "ph": "f", "id": 180205, "pid": 76337, "tid": -914061504, "ts": 1716454224268847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268847, "dur": 0,
    "args": {
      "External id": 180206,
      "cbid": 251, "correlation": 180206
    }
  },
  {
    "ph": "f", "id": 180206, "pid": 76337, "tid": -914061504, "ts": 1716454224268847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268848, "dur": 0,
    "args": {
      "External id": 180207,
      "cbid": 251, "correlation": 180207
    }
  },
  {
    "ph": "f", "id": 180207, "pid": 76337, "tid": -914061504, "ts": 1716454224268848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268849, "dur": 0,
    "args": {
      "External id": 180208,
      "cbid": 251, "correlation": 180208
    }
  },
  {
    "ph": "f", "id": 180208, "pid": 76337, "tid": -914061504, "ts": 1716454224268849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268850, "dur": 0,
    "args": {
      "External id": 180209,
      "cbid": 251, "correlation": 180209
    }
  },
  {
    "ph": "f", "id": 180209, "pid": 76337, "tid": -914061504, "ts": 1716454224268850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268850, "dur": 0,
    "args": {
      "External id": 180210,
      "cbid": 251, "correlation": 180210
    }
  },
  {
    "ph": "f", "id": 180210, "pid": 76337, "tid": -914061504, "ts": 1716454224268850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268852, "dur": 0,
    "args": {
      "External id": 180211,
      "cbid": 251, "correlation": 180211
    }
  },
  {
    "ph": "f", "id": 180211, "pid": 76337, "tid": -914061504, "ts": 1716454224268852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224292258, "dur": 111,
    "args": {
      "External id": 180212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180212,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 180212, "pid": 5, "tid": 7, "ts": 1716454224292258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268854, "dur": 12,
    "args": {
      "External id": 180212,
      "cbid": 211, "correlation": 180212
    }
  },
  {
    "ph": "s", "id": 180212, "pid": 76337, "tid": -914061504, "ts": 1716454224268854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224292371, "dur": 60,
    "args": {
      "External id": 180218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180218,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180218, "pid": 5, "tid": 7, "ts": 1716454224292371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268889, "dur": 9,
    "args": {
      "External id": 180218,
      "cbid": 211, "correlation": 180218
    }
  },
  {
    "ph": "s", "id": 180218, "pid": 76337, "tid": -914061504, "ts": 1716454224268889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224292432, "dur": 600,
    "args": {
      "External id": 180227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180227,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180227, "pid": 5, "tid": 7, "ts": 1716454224292432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224268972, "dur": 25,
    "args": {
      "External id": 180227,
      "cbid": 211, "correlation": 180227
    }
  },
  {
    "ph": "s", "id": 180227, "pid": 76337, "tid": -914061504, "ts": 1716454224268972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224293033, "dur": 181,
    "args": {
      "External id": 180249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180249,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180249, "pid": 5, "tid": 7, "ts": 1716454224293033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269040, "dur": 10,
    "args": {
      "External id": 180249,
      "cbid": 211, "correlation": 180249
    }
  },
  {
    "ph": "s", "id": 180249, "pid": 76337, "tid": -914061504, "ts": 1716454224269040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269125, "dur": 1,
    "args": {
      "External id": 180260,
      "cbid": 251, "correlation": 180260
    }
  },
  {
    "ph": "f", "id": 180260, "pid": 76337, "tid": -914061504, "ts": 1716454224269125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224293216, "dur": 196,
    "args": {
      "External id": 180261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180261,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180261, "pid": 5, "tid": 7, "ts": 1716454224293216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269131, "dur": 13,
    "args": {
      "External id": 180261,
      "cbid": 211, "correlation": 180261
    }
  },
  {
    "ph": "s", "id": 180261, "pid": 76337, "tid": -914061504, "ts": 1716454224269131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269199, "dur": 1,
    "args": {
      "External id": 180272,
      "cbid": 251, "correlation": 180272
    }
  },
  {
    "ph": "f", "id": 180272, "pid": 76337, "tid": -914061504, "ts": 1716454224269199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224293414, "dur": 190,
    "args": {
      "External id": 180273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180273,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180273, "pid": 5, "tid": 7, "ts": 1716454224293414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269203, "dur": 11,
    "args": {
      "External id": 180273,
      "cbid": 211, "correlation": 180273
    }
  },
  {
    "ph": "s", "id": 180273, "pid": 76337, "tid": -914061504, "ts": 1716454224269203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269265, "dur": 1,
    "args": {
      "External id": 180284,
      "cbid": 251, "correlation": 180284
    }
  },
  {
    "ph": "f", "id": 180284, "pid": 76337, "tid": -914061504, "ts": 1716454224269265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224293605, "dur": 189,
    "args": {
      "External id": 180285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180285,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180285, "pid": 5, "tid": 7, "ts": 1716454224293605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269269, "dur": 12,
    "args": {
      "External id": 180285,
      "cbid": 211, "correlation": 180285
    }
  },
  {
    "ph": "s", "id": 180285, "pid": 76337, "tid": -914061504, "ts": 1716454224269269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224293796, "dur": 18777,
    "args": {
      "External id": 180306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180306,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 180306, "pid": 5, "tid": 7, "ts": 1716454224293796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269349, "dur": 12,
    "args": {
      "External id": 180306,
      "cbid": 211, "correlation": 180306
    }
  },
  {
    "ph": "s", "id": 180306, "pid": 76337, "tid": -914061504, "ts": 1716454224269349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269447, "dur": 1,
    "args": {
      "External id": 180324,
      "cbid": 251, "correlation": 180324
    }
  },
  {
    "ph": "f", "id": 180324, "pid": 76337, "tid": -914061504, "ts": 1716454224269447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224312574, "dur": 203,
    "args": {
      "External id": 180326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180326,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180326, "pid": 5, "tid": 7, "ts": 1716454224312574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269452, "dur": 13,
    "args": {
      "External id": 180326,
      "cbid": 211, "correlation": 180326
    }
  },
  {
    "ph": "s", "id": 180326, "pid": 76337, "tid": -914061504, "ts": 1716454224269452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224312779, "dur": 67,
    "args": {
      "External id": 180334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180334,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180334, "pid": 5, "tid": 7, "ts": 1716454224312779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269522, "dur": 13,
    "args": {
      "External id": 180334,
      "cbid": 211, "correlation": 180334
    }
  },
  {
    "ph": "s", "id": 180334, "pid": 76337, "tid": -914061504, "ts": 1716454224269522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224312847, "dur": 96,
    "args": {
      "External id": 180342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180342,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180342, "pid": 5, "tid": 7, "ts": 1716454224312847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269561, "dur": 8,
    "args": {
      "External id": 180342,
      "cbid": 211, "correlation": 180342
    }
  },
  {
    "ph": "s", "id": 180342, "pid": 76337, "tid": -914061504, "ts": 1716454224269561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224312944, "dur": 54,
    "args": {
      "External id": 180353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180353,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180353, "pid": 5, "tid": 7, "ts": 1716454224312944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269631, "dur": 12,
    "args": {
      "External id": 180353,
      "cbid": 211, "correlation": 180353
    }
  },
  {
    "ph": "s", "id": 180353, "pid": 76337, "tid": -914061504, "ts": 1716454224269631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224313000, "dur": 92,
    "args": {
      "External id": 180375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180375,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180375, "pid": 5, "tid": 7, "ts": 1716454224313000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269662, "dur": 8,
    "args": {
      "External id": 180375,
      "cbid": 211, "correlation": 180375
    }
  },
  {
    "ph": "s", "id": 180375, "pid": 76337, "tid": -914061504, "ts": 1716454224269662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269744, "dur": 1,
    "args": {
      "External id": 180386,
      "cbid": 251, "correlation": 180386
    }
  },
  {
    "ph": "f", "id": 180386, "pid": 76337, "tid": -914061504, "ts": 1716454224269744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224313094, "dur": 104,
    "args": {
      "External id": 180387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180387,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180387, "pid": 5, "tid": 7, "ts": 1716454224313094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269749, "dur": 13,
    "args": {
      "External id": 180387,
      "cbid": 211, "correlation": 180387
    }
  },
  {
    "ph": "s", "id": 180387, "pid": 76337, "tid": -914061504, "ts": 1716454224269749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269820, "dur": 1,
    "args": {
      "External id": 180398,
      "cbid": 251, "correlation": 180398
    }
  },
  {
    "ph": "f", "id": 180398, "pid": 76337, "tid": -914061504, "ts": 1716454224269820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269824, "dur": 0,
    "args": {
      "External id": 180399,
      "cbid": 251, "correlation": 180399
    }
  },
  {
    "ph": "f", "id": 180399, "pid": 76337, "tid": -914061504, "ts": 1716454224269824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224313199, "dur": 10,
    "args": {
      "External id": 180400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180400,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 180400, "pid": 5, "tid": 7, "ts": 1716454224313199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269825, "dur": 13,
    "args": {
      "External id": 180400,
      "cbid": 211, "correlation": 180400
    }
  },
  {
    "ph": "s", "id": 180400, "pid": 76337, "tid": -914061504, "ts": 1716454224269825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224313210, "dur": 5,
    "args": {
      "External id": 180402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180402,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 180402, "pid": 5, "tid": 7, "ts": 1716454224313210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269840, "dur": 6,
    "args": {
      "External id": 180402,
      "cbid": 211, "correlation": 180402
    }
  },
  {
    "ph": "s", "id": 180402, "pid": 76337, "tid": -914061504, "ts": 1716454224269840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269899, "dur": 1,
    "args": {
      "External id": 180413,
      "cbid": 251, "correlation": 180413
    }
  },
  {
    "ph": "f", "id": 180413, "pid": 76337, "tid": -914061504, "ts": 1716454224269899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269903, "dur": 0,
    "args": {
      "External id": 180414,
      "cbid": 251, "correlation": 180414
    }
  },
  {
    "ph": "f", "id": 180414, "pid": 76337, "tid": -914061504, "ts": 1716454224269903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224313217, "dur": 6,
    "args": {
      "External id": 180415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180415,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 180415, "pid": 5, "tid": 7, "ts": 1716454224313217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269904, "dur": 12,
    "args": {
      "External id": 180415,
      "cbid": 211, "correlation": 180415
    }
  },
  {
    "ph": "s", "id": 180415, "pid": 76337, "tid": -914061504, "ts": 1716454224269904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224313224, "dur": 3,
    "args": {
      "External id": 180417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180417,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 180417, "pid": 5, "tid": 7, "ts": 1716454224313224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269918, "dur": 5,
    "args": {
      "External id": 180417,
      "cbid": 211, "correlation": 180417
    }
  },
  {
    "ph": "s", "id": 180417, "pid": 76337, "tid": -914061504, "ts": 1716454224269918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224313229, "dur": 156,
    "args": {
      "External id": 180438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180438,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 180438, "pid": 5, "tid": 7, "ts": 1716454224313229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224269999, "dur": 12,
    "args": {
      "External id": 180438,
      "cbid": 211, "correlation": 180438
    }
  },
  {
    "ph": "s", "id": 180438, "pid": 76337, "tid": -914061504, "ts": 1716454224269999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270097, "dur": 1,
    "args": {
      "External id": 180456,
      "cbid": 251, "correlation": 180456
    }
  },
  {
    "ph": "f", "id": 180456, "pid": 76337, "tid": -914061504, "ts": 1716454224270097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224313387, "dur": 107,
    "args": {
      "External id": 180458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180458,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 180458, "pid": 5, "tid": 7, "ts": 1716454224313387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270103, "dur": 13,
    "args": {
      "External id": 180458,
      "cbid": 211, "correlation": 180458
    }
  },
  {
    "ph": "s", "id": 180458, "pid": 76337, "tid": -914061504, "ts": 1716454224270103,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224313495, "dur": 34,
    "args": {
      "External id": 180466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180466,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180466, "pid": 5, "tid": 7, "ts": 1716454224313495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270172, "dur": 12,
    "args": {
      "External id": 180466,
      "cbid": 211, "correlation": 180466
    }
  },
  {
    "ph": "s", "id": 180466, "pid": 76337, "tid": -914061504, "ts": 1716454224270172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224313531, "dur": 67,
    "args": {
      "External id": 180474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180474,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180474, "pid": 5, "tid": 7, "ts": 1716454224313531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270213, "dur": 9,
    "args": {
      "External id": 180474,
      "cbid": 211, "correlation": 180474
    }
  },
  {
    "ph": "s", "id": 180474, "pid": 76337, "tid": -914061504, "ts": 1716454224270213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224313600, "dur": 92,
    "args": {
      "External id": 180496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180496,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180496, "pid": 5, "tid": 7, "ts": 1716454224313600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270264, "dur": 10,
    "args": {
      "External id": 180496,
      "cbid": 211, "correlation": 180496
    }
  },
  {
    "ph": "s", "id": 180496, "pid": 76337, "tid": -914061504, "ts": 1716454224270264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270350, "dur": 1,
    "args": {
      "External id": 180512,
      "cbid": 251, "correlation": 180512
    }
  },
  {
    "ph": "f", "id": 180512, "pid": 76337, "tid": -914061504, "ts": 1716454224270350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224313694, "dur": 581,
    "args": {
      "External id": 180514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180514,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180514, "pid": 5, "tid": 7, "ts": 1716454224313694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270356, "dur": 12,
    "args": {
      "External id": 180514,
      "cbid": 211, "correlation": 180514
    }
  },
  {
    "ph": "s", "id": 180514, "pid": 76337, "tid": -914061504, "ts": 1716454224270356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224314276, "dur": 246,
    "args": {
      "External id": 180522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180522, "pid": 5, "tid": 7, "ts": 1716454224314276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270421, "dur": 12,
    "args": {
      "External id": 180522,
      "cbid": 211, "correlation": 180522
    }
  },
  {
    "ph": "s", "id": 180522, "pid": 76337, "tid": -914061504, "ts": 1716454224270421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224314523, "dur": 254,
    "args": {
      "External id": 180530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180530,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180530, "pid": 5, "tid": 7, "ts": 1716454224314523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270453, "dur": 8,
    "args": {
      "External id": 180530,
      "cbid": 211, "correlation": 180530
    }
  },
  {
    "ph": "s", "id": 180530, "pid": 76337, "tid": -914061504, "ts": 1716454224270453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270543, "dur": 1,
    "args": {
      "External id": 180546,
      "cbid": 251, "correlation": 180546
    }
  },
  {
    "ph": "f", "id": 180546, "pid": 76337, "tid": -914061504, "ts": 1716454224270543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270549, "dur": 0,
    "args": {
      "External id": 180548,
      "cbid": 251, "correlation": 180548
    }
  },
  {
    "ph": "f", "id": 180548, "pid": 76337, "tid": -914061504, "ts": 1716454224270549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224314779, "dur": 360,
    "args": {
      "External id": 180549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180549,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 180549, "pid": 5, "tid": 7, "ts": 1716454224314779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270551, "dur": 13,
    "args": {
      "External id": 180549,
      "cbid": 211, "correlation": 180549
    }
  },
  {
    "ph": "s", "id": 180549, "pid": 76337, "tid": -914061504, "ts": 1716454224270551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224315140, "dur": 50,
    "args": {
      "External id": 180557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180557,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180557, "pid": 5, "tid": 7, "ts": 1716454224315140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270594, "dur": 10,
    "args": {
      "External id": 180557,
      "cbid": 211, "correlation": 180557
    }
  },
  {
    "ph": "s", "id": 180557, "pid": 76337, "tid": -914061504, "ts": 1716454224270594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224315191, "dur": 159,
    "args": {
      "External id": 180568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180568, "pid": 5, "tid": 7, "ts": 1716454224315191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270660, "dur": 12,
    "args": {
      "External id": 180568,
      "cbid": 211, "correlation": 180568
    }
  },
  {
    "ph": "s", "id": 180568, "pid": 76337, "tid": -914061504, "ts": 1716454224270660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270723, "dur": 0,
    "args": {
      "External id": 180580,
      "cbid": 317, "correlation": 180580
    }
  },
  {
    "ph": "f", "id": 180580, "pid": 76337, "tid": -914061504, "ts": 1716454224270723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270724, "dur": 0,
    "args": {
      "External id": 180581,
      "cbid": 203, "correlation": 180581
    }
  },
  {
    "ph": "f", "id": 180581, "pid": 76337, "tid": -914061504, "ts": 1716454224270724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270725, "dur": 0,
    "args": {
      "External id": 180582,
      "cbid": 205, "correlation": 180582
    }
  },
  {
    "ph": "f", "id": 180582, "pid": 76337, "tid": -914061504, "ts": 1716454224270725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270747, "dur": 1,
    "args": {
      "External id": 180586,
      "cbid": 251, "correlation": 180586
    }
  },
  {
    "ph": "f", "id": 180586, "pid": 76337, "tid": -914061504, "ts": 1716454224270747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270749, "dur": 0,
    "args": {
      "External id": 180587,
      "cbid": 251, "correlation": 180587
    }
  },
  {
    "ph": "f", "id": 180587, "pid": 76337, "tid": -914061504, "ts": 1716454224270749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270750, "dur": 0,
    "args": {
      "External id": 180588,
      "cbid": 251, "correlation": 180588
    }
  },
  {
    "ph": "f", "id": 180588, "pid": 76337, "tid": -914061504, "ts": 1716454224270750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270750, "dur": 0,
    "args": {
      "External id": 180589,
      "cbid": 251, "correlation": 180589
    }
  },
  {
    "ph": "f", "id": 180589, "pid": 76337, "tid": -914061504, "ts": 1716454224270750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270751, "dur": 0,
    "args": {
      "External id": 180590,
      "cbid": 251, "correlation": 180590
    }
  },
  {
    "ph": "f", "id": 180590, "pid": 76337, "tid": -914061504, "ts": 1716454224270751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270752, "dur": 0,
    "args": {
      "External id": 180591,
      "cbid": 251, "correlation": 180591
    }
  },
  {
    "ph": "f", "id": 180591, "pid": 76337, "tid": -914061504, "ts": 1716454224270752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270753, "dur": 0,
    "args": {
      "External id": 180592,
      "cbid": 251, "correlation": 180592
    }
  },
  {
    "ph": "f", "id": 180592, "pid": 76337, "tid": -914061504, "ts": 1716454224270753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270753, "dur": 0,
    "args": {
      "External id": 180593,
      "cbid": 251, "correlation": 180593
    }
  },
  {
    "ph": "f", "id": 180593, "pid": 76337, "tid": -914061504, "ts": 1716454224270753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270755, "dur": 0,
    "args": {
      "External id": 180594,
      "cbid": 251, "correlation": 180594
    }
  },
  {
    "ph": "f", "id": 180594, "pid": 76337, "tid": -914061504, "ts": 1716454224270755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224315352, "dur": 116,
    "args": {
      "External id": 180595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180595,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 180595, "pid": 5, "tid": 7, "ts": 1716454224315352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270757, "dur": 12,
    "args": {
      "External id": 180595,
      "cbid": 211, "correlation": 180595
    }
  },
  {
    "ph": "s", "id": 180595, "pid": 76337, "tid": -914061504, "ts": 1716454224270757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224315469, "dur": 60,
    "args": {
      "External id": 180601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180601,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180601, "pid": 5, "tid": 7, "ts": 1716454224315469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270791, "dur": 10,
    "args": {
      "External id": 180601,
      "cbid": 211, "correlation": 180601
    }
  },
  {
    "ph": "s", "id": 180601, "pid": 76337, "tid": -914061504, "ts": 1716454224270791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224315531, "dur": 49,
    "args": {
      "External id": 180609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180609,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180609, "pid": 5, "tid": 7, "ts": 1716454224315531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270824, "dur": 8,
    "args": {
      "External id": 180609,
      "cbid": 211, "correlation": 180609
    }
  },
  {
    "ph": "s", "id": 180609, "pid": 76337, "tid": -914061504, "ts": 1716454224270824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270907, "dur": 0,
    "args": {
      "External id": 180619,
      "cbid": 317, "correlation": 180619
    }
  },
  {
    "ph": "f", "id": 180619, "pid": 76337, "tid": -914061504, "ts": 1716454224270907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270907, "dur": 0,
    "args": {
      "External id": 180620,
      "cbid": 203, "correlation": 180620
    }
  },
  {
    "ph": "f", "id": 180620, "pid": 76337, "tid": -914061504, "ts": 1716454224270907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270908, "dur": 0,
    "args": {
      "External id": 180621,
      "cbid": 205, "correlation": 180621
    }
  },
  {
    "ph": "f", "id": 180621, "pid": 76337, "tid": -914061504, "ts": 1716454224270908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224315581, "dur": 41,
    "args": {
      "External id": 180625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180625, "pid": 5, "tid": 7, "ts": 1716454224315581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270924, "dur": 12,
    "args": {
      "External id": 180625,
      "cbid": 211, "correlation": 180625
    }
  },
  {
    "ph": "s", "id": 180625, "pid": 76337, "tid": -914061504, "ts": 1716454224270924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224315624, "dur": 15,
    "args": {
      "External id": 180627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180627,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180627, "pid": 5, "tid": 7, "ts": 1716454224315624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270939, "dur": 6,
    "args": {
      "External id": 180627,
      "cbid": 211, "correlation": 180627
    }
  },
  {
    "ph": "s", "id": 180627, "pid": 76337, "tid": -914061504, "ts": 1716454224270939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224315641, "dur": 1,
    "args": {
      "External id": 180629,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 180629,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 180629, "pid": 5, "tid": 7, "ts": 1716454224315641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270959, "dur": 25,
    "args": {
      "External id": 180629,
      "cbid": 51, "correlation": 180629
    }
  },
  {
    "ph": "s", "id": 180629, "pid": 76337, "tid": -914061504, "ts": 1716454224270959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224315645, "dur": 364,
    "args": {
      "External id": 180630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180630,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180630, "pid": 5, "tid": 7, "ts": 1716454224315645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224270986, "dur": 11,
    "args": {
      "External id": 180630,
      "cbid": 211, "correlation": 180630
    }
  },
  {
    "ph": "s", "id": 180630, "pid": 76337, "tid": -914061504, "ts": 1716454224270986,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316010, "dur": 14,
    "args": {
      "External id": 180632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180632,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180632, "pid": 5, "tid": 7, "ts": 1716454224316010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271005, "dur": 7,
    "args": {
      "External id": 180632,
      "cbid": 211, "correlation": 180632
    }
  },
  {
    "ph": "s", "id": 180632, "pid": 76337, "tid": -914061504, "ts": 1716454224271005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224316025, "dur": 15,
    "args": {
      "External id": 180638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180638, "pid": 5, "tid": 7, "ts": 1716454224316025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271035, "dur": 9,
    "args": {
      "External id": 180638,
      "cbid": 211, "correlation": 180638
    }
  },
  {
    "ph": "s", "id": 180638, "pid": 76337, "tid": -914061504, "ts": 1716454224271035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316042, "dur": 19,
    "args": {
      "External id": 180658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180658,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 180658, "pid": 5, "tid": 7, "ts": 1716454224316042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271134, "dur": 14,
    "args": {
      "External id": 180658,
      "cbid": 211, "correlation": 180658
    }
  },
  {
    "ph": "s", "id": 180658, "pid": 76337, "tid": -914061504, "ts": 1716454224271134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224316062, "dur": 4,
    "args": {
      "External id": 180670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180670,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180670, "pid": 5, "tid": 7, "ts": 1716454224316062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271158, "dur": 6,
    "args": {
      "External id": 180670,
      "cbid": 211, "correlation": 180670
    }
  },
  {
    "ph": "s", "id": 180670, "pid": 76337, "tid": -914061504, "ts": 1716454224271158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224316068, "dur": 18,
    "args": {
      "External id": 180673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180673,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180673, "pid": 5, "tid": 7, "ts": 1716454224316068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271175, "dur": 7,
    "args": {
      "External id": 180673,
      "cbid": 211, "correlation": 180673
    }
  },
  {
    "ph": "s", "id": 180673, "pid": 76337, "tid": -914061504, "ts": 1716454224271175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224316087, "dur": 11,
    "args": {
      "External id": 180682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180682,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180682, "pid": 5, "tid": 7, "ts": 1716454224316087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271216, "dur": 9,
    "args": {
      "External id": 180682,
      "cbid": 211, "correlation": 180682
    }
  },
  {
    "ph": "s", "id": 180682, "pid": 76337, "tid": -914061504, "ts": 1716454224271216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271274, "dur": 0,
    "args": {
      "External id": 180692,
      "cbid": 317, "correlation": 180692
    }
  },
  {
    "ph": "f", "id": 180692, "pid": 76337, "tid": -914061504, "ts": 1716454224271274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271274, "dur": 0,
    "args": {
      "External id": 180693,
      "cbid": 203, "correlation": 180693
    }
  },
  {
    "ph": "f", "id": 180693, "pid": 76337, "tid": -914061504, "ts": 1716454224271274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271275, "dur": 0,
    "args": {
      "External id": 180694,
      "cbid": 205, "correlation": 180694
    }
  },
  {
    "ph": "f", "id": 180694, "pid": 76337, "tid": -914061504, "ts": 1716454224271275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316100, "dur": 11,
    "args": {
      "External id": 180698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180698,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180698, "pid": 5, "tid": 7, "ts": 1716454224316100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271290, "dur": 13,
    "args": {
      "External id": 180698,
      "cbid": 211, "correlation": 180698
    }
  },
  {
    "ph": "s", "id": 180698, "pid": 76337, "tid": -914061504, "ts": 1716454224271290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316112, "dur": 25,
    "args": {
      "External id": 180700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180700, "pid": 5, "tid": 7, "ts": 1716454224316112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271305, "dur": 5,
    "args": {
      "External id": 180700,
      "cbid": 211, "correlation": 180700
    }
  },
  {
    "ph": "s", "id": 180700, "pid": 76337, "tid": -914061504, "ts": 1716454224271305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224316138, "dur": 4,
    "args": {
      "External id": 180702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 180702, "pid": 5, "tid": 7, "ts": 1716454224316138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271316, "dur": 5,
    "args": {
      "External id": 180702,
      "cbid": 211, "correlation": 180702
    }
  },
  {
    "ph": "s", "id": 180702, "pid": 76337, "tid": -914061504, "ts": 1716454224271316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271325, "dur": 0,
    "args": {
      "External id": 180703,
      "cbid": 51, "correlation": 180703
    }
  },
  {
    "ph": "s", "id": 180703, "pid": 76337, "tid": -914061504, "ts": 1716454224271325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224316143, "dur": 360,
    "args": {
      "External id": 180704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180704,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180704, "pid": 5, "tid": 7, "ts": 1716454224316143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271326, "dur": 7,
    "args": {
      "External id": 180704,
      "cbid": 211, "correlation": 180704
    }
  },
  {
    "ph": "s", "id": 180704, "pid": 76337, "tid": -914061504, "ts": 1716454224271326,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316504, "dur": 19,
    "args": {
      "External id": 180705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180705,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180705, "pid": 5, "tid": 7, "ts": 1716454224316504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271336, "dur": 5,
    "args": {
      "External id": 180705,
      "cbid": 211, "correlation": 180705
    }
  },
  {
    "ph": "s", "id": 180705, "pid": 76337, "tid": -914061504, "ts": 1716454224271336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224316525, "dur": 34,
    "args": {
      "External id": 180711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180711, "pid": 5, "tid": 7, "ts": 1716454224316525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271364, "dur": 9,
    "args": {
      "External id": 180711,
      "cbid": 211, "correlation": 180711
    }
  },
  {
    "ph": "s", "id": 180711, "pid": 76337, "tid": -914061504, "ts": 1716454224271364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224316560, "dur": 4,
    "args": {
      "External id": 180719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180719,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 180719, "pid": 5, "tid": 7, "ts": 1716454224316560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271407, "dur": 9,
    "args": {
      "External id": 180719,
      "cbid": 211, "correlation": 180719
    }
  },
  {
    "ph": "s", "id": 180719, "pid": 76337, "tid": -914061504, "ts": 1716454224271407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271473, "dur": 1,
    "args": {
      "External id": 180735,
      "cbid": 251, "correlation": 180735
    }
  },
  {
    "ph": "f", "id": 180735, "pid": 76337, "tid": -914061504, "ts": 1716454224271473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271478, "dur": 0,
    "args": {
      "External id": 180737,
      "cbid": 251, "correlation": 180737
    }
  },
  {
    "ph": "f", "id": 180737, "pid": 76337, "tid": -914061504, "ts": 1716454224271478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224316565, "dur": 13,
    "args": {
      "External id": 180738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180738,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 180738, "pid": 5, "tid": 7, "ts": 1716454224316565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271480, "dur": 12,
    "args": {
      "External id": 180738,
      "cbid": 211, "correlation": 180738
    }
  },
  {
    "ph": "s", "id": 180738, "pid": 76337, "tid": -914061504, "ts": 1716454224271480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224316579, "dur": 5,
    "args": {
      "External id": 180740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180740,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 180740, "pid": 5, "tid": 7, "ts": 1716454224316579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271493, "dur": 5,
    "args": {
      "External id": 180740,
      "cbid": 211, "correlation": 180740
    }
  },
  {
    "ph": "s", "id": 180740, "pid": 76337, "tid": -914061504, "ts": 1716454224271493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224316585, "dur": 30,
    "args": {
      "External id": 180750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180750, "pid": 5, "tid": 7, "ts": 1716454224316585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271552, "dur": 13,
    "args": {
      "External id": 180750,
      "cbid": 211, "correlation": 180750
    }
  },
  {
    "ph": "s", "id": 180750, "pid": 76337, "tid": -914061504, "ts": 1716454224271552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316616, "dur": 30,
    "args": {
      "External id": 180770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180770,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 180770, "pid": 5, "tid": 7, "ts": 1716454224316616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271619, "dur": 11,
    "args": {
      "External id": 180770,
      "cbid": 211, "correlation": 180770
    }
  },
  {
    "ph": "s", "id": 180770, "pid": 76337, "tid": -914061504, "ts": 1716454224271619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224316648, "dur": 4,
    "args": {
      "External id": 180782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180782,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 180782, "pid": 5, "tid": 7, "ts": 1716454224316648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271640, "dur": 6,
    "args": {
      "External id": 180782,
      "cbid": 211, "correlation": 180782
    }
  },
  {
    "ph": "s", "id": 180782, "pid": 76337, "tid": -914061504, "ts": 1716454224271640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224316653, "dur": 30,
    "args": {
      "External id": 180785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180785, "pid": 5, "tid": 7, "ts": 1716454224316653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271659, "dur": 6,
    "args": {
      "External id": 180785,
      "cbid": 211, "correlation": 180785
    }
  },
  {
    "ph": "s", "id": 180785, "pid": 76337, "tid": -914061504, "ts": 1716454224271659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224316684, "dur": 20,
    "args": {
      "External id": 180794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180794,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180794, "pid": 5, "tid": 7, "ts": 1716454224316684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271699, "dur": 10,
    "args": {
      "External id": 180794,
      "cbid": 211, "correlation": 180794
    }
  },
  {
    "ph": "s", "id": 180794, "pid": 76337, "tid": -914061504, "ts": 1716454224271699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271766, "dur": 0,
    "args": {
      "External id": 180804,
      "cbid": 317, "correlation": 180804
    }
  },
  {
    "ph": "f", "id": 180804, "pid": 76337, "tid": -914061504, "ts": 1716454224271766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271767, "dur": 0,
    "args": {
      "External id": 180805,
      "cbid": 203, "correlation": 180805
    }
  },
  {
    "ph": "f", "id": 180805, "pid": 76337, "tid": -914061504, "ts": 1716454224271767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271768, "dur": 0,
    "args": {
      "External id": 180806,
      "cbid": 205, "correlation": 180806
    }
  },
  {
    "ph": "f", "id": 180806, "pid": 76337, "tid": -914061504, "ts": 1716454224271768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316706, "dur": 21,
    "args": {
      "External id": 180810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180810, "pid": 5, "tid": 7, "ts": 1716454224316706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271787, "dur": 12,
    "args": {
      "External id": 180810,
      "cbid": 211, "correlation": 180810
    }
  },
  {
    "ph": "s", "id": 180810, "pid": 76337, "tid": -914061504, "ts": 1716454224271787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224316729, "dur": 44,
    "args": {
      "External id": 180812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180812, "pid": 5, "tid": 7, "ts": 1716454224316729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271802, "dur": 6,
    "args": {
      "External id": 180812,
      "cbid": 211, "correlation": 180812
    }
  },
  {
    "ph": "s", "id": 180812, "pid": 76337, "tid": -914061504, "ts": 1716454224271802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224316774, "dur": 657,
    "args": {
      "External id": 180814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180814,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180814, "pid": 5, "tid": 7, "ts": 1716454224316774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271816, "dur": 9,
    "args": {
      "External id": 180814,
      "cbid": 211, "correlation": 180814
    }
  },
  {
    "ph": "s", "id": 180814, "pid": 76337, "tid": -914061504, "ts": 1716454224271816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224317433, "dur": 20,
    "args": {
      "External id": 180816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180816,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180816, "pid": 5, "tid": 7, "ts": 1716454224317433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271829, "dur": 5,
    "args": {
      "External id": 180816,
      "cbid": 211, "correlation": 180816
    }
  },
  {
    "ph": "s", "id": 180816, "pid": 76337, "tid": -914061504, "ts": 1716454224271829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224317454, "dur": 34,
    "args": {
      "External id": 180822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180822, "pid": 5, "tid": 7, "ts": 1716454224317454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271858, "dur": 9,
    "args": {
      "External id": 180822,
      "cbid": 211, "correlation": 180822
    }
  },
  {
    "ph": "s", "id": 180822, "pid": 76337, "tid": -914061504, "ts": 1716454224271858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271916, "dur": 0,
    "args": {
      "External id": 180832,
      "cbid": 317, "correlation": 180832
    }
  },
  {
    "ph": "f", "id": 180832, "pid": 76337, "tid": -914061504, "ts": 1716454224271916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271917, "dur": 0,
    "args": {
      "External id": 180833,
      "cbid": 203, "correlation": 180833
    }
  },
  {
    "ph": "f", "id": 180833, "pid": 76337, "tid": -914061504, "ts": 1716454224271917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271918, "dur": 0,
    "args": {
      "External id": 180834,
      "cbid": 205, "correlation": 180834
    }
  },
  {
    "ph": "f", "id": 180834, "pid": 76337, "tid": -914061504, "ts": 1716454224271918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271940, "dur": 1,
    "args": {
      "External id": 180838,
      "cbid": 251, "correlation": 180838
    }
  },
  {
    "ph": "f", "id": 180838, "pid": 76337, "tid": -914061504, "ts": 1716454224271940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271941, "dur": 0,
    "args": {
      "External id": 180839,
      "cbid": 251, "correlation": 180839
    }
  },
  {
    "ph": "f", "id": 180839, "pid": 76337, "tid": -914061504, "ts": 1716454224271941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271942, "dur": 0,
    "args": {
      "External id": 180840,
      "cbid": 251, "correlation": 180840
    }
  },
  {
    "ph": "f", "id": 180840, "pid": 76337, "tid": -914061504, "ts": 1716454224271942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271943, "dur": 0,
    "args": {
      "External id": 180841,
      "cbid": 251, "correlation": 180841
    }
  },
  {
    "ph": "f", "id": 180841, "pid": 76337, "tid": -914061504, "ts": 1716454224271943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271944, "dur": 0,
    "args": {
      "External id": 180842,
      "cbid": 251, "correlation": 180842
    }
  },
  {
    "ph": "f", "id": 180842, "pid": 76337, "tid": -914061504, "ts": 1716454224271944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271944, "dur": 0,
    "args": {
      "External id": 180843,
      "cbid": 251, "correlation": 180843
    }
  },
  {
    "ph": "f", "id": 180843, "pid": 76337, "tid": -914061504, "ts": 1716454224271944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271945, "dur": 0,
    "args": {
      "External id": 180844,
      "cbid": 251, "correlation": 180844
    }
  },
  {
    "ph": "f", "id": 180844, "pid": 76337, "tid": -914061504, "ts": 1716454224271945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271946, "dur": 0,
    "args": {
      "External id": 180845,
      "cbid": 251, "correlation": 180845
    }
  },
  {
    "ph": "f", "id": 180845, "pid": 76337, "tid": -914061504, "ts": 1716454224271946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271947, "dur": 0,
    "args": {
      "External id": 180846,
      "cbid": 251, "correlation": 180846
    }
  },
  {
    "ph": "f", "id": 180846, "pid": 76337, "tid": -914061504, "ts": 1716454224271947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224317489, "dur": 53,
    "args": {
      "External id": 180847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180847,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 180847, "pid": 5, "tid": 7, "ts": 1716454224317489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271949, "dur": 13,
    "args": {
      "External id": 180847,
      "cbid": 211, "correlation": 180847
    }
  },
  {
    "ph": "s", "id": 180847, "pid": 76337, "tid": -914061504, "ts": 1716454224271949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224317543, "dur": 33,
    "args": {
      "External id": 180853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180853, "pid": 5, "tid": 7, "ts": 1716454224317543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224271990, "dur": 9,
    "args": {
      "External id": 180853,
      "cbid": 211, "correlation": 180853
    }
  },
  {
    "ph": "s", "id": 180853, "pid": 76337, "tid": -914061504, "ts": 1716454224271990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224317577, "dur": 27,
    "args": {
      "External id": 180861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180861,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180861, "pid": 5, "tid": 7, "ts": 1716454224317577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272021, "dur": 8,
    "args": {
      "External id": 180861,
      "cbid": 211, "correlation": 180861
    }
  },
  {
    "ph": "s", "id": 180861, "pid": 76337, "tid": -914061504, "ts": 1716454224272021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224317605, "dur": 20,
    "args": {
      "External id": 180869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180869,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180869, "pid": 5, "tid": 7, "ts": 1716454224317605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272050, "dur": 9,
    "args": {
      "External id": 180869,
      "cbid": 211, "correlation": 180869
    }
  },
  {
    "ph": "s", "id": 180869, "pid": 76337, "tid": -914061504, "ts": 1716454224272050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224317626, "dur": 30,
    "args": {
      "External id": 180889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180889,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 180889, "pid": 5, "tid": 7, "ts": 1716454224317626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272133, "dur": 13,
    "args": {
      "External id": 180889,
      "cbid": 211, "correlation": 180889
    }
  },
  {
    "ph": "s", "id": 180889, "pid": 76337, "tid": -914061504, "ts": 1716454224272133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224317657, "dur": 4,
    "args": {
      "External id": 180901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180901,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 180901, "pid": 5, "tid": 7, "ts": 1716454224317657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272155, "dur": 7,
    "args": {
      "External id": 180901,
      "cbid": 211, "correlation": 180901
    }
  },
  {
    "ph": "s", "id": 180901, "pid": 76337, "tid": -914061504, "ts": 1716454224272155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224317662, "dur": 30,
    "args": {
      "External id": 180904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180904,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180904, "pid": 5, "tid": 7, "ts": 1716454224317662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272174, "dur": 7,
    "args": {
      "External id": 180904,
      "cbid": 211, "correlation": 180904
    }
  },
  {
    "ph": "s", "id": 180904, "pid": 76337, "tid": -914061504, "ts": 1716454224272174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272234, "dur": 0,
    "args": {
      "External id": 180915,
      "cbid": 317, "correlation": 180915
    }
  },
  {
    "ph": "f", "id": 180915, "pid": 76337, "tid": -914061504, "ts": 1716454224272234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272235, "dur": 0,
    "args": {
      "External id": 180916,
      "cbid": 203, "correlation": 180916
    }
  },
  {
    "ph": "f", "id": 180916, "pid": 76337, "tid": -914061504, "ts": 1716454224272235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272236, "dur": 0,
    "args": {
      "External id": 180917,
      "cbid": 205, "correlation": 180917
    }
  },
  {
    "ph": "f", "id": 180917, "pid": 76337, "tid": -914061504, "ts": 1716454224272236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224317693, "dur": 21,
    "args": {
      "External id": 180921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180921,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180921, "pid": 5, "tid": 7, "ts": 1716454224317693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272251, "dur": 12,
    "args": {
      "External id": 180921,
      "cbid": 211, "correlation": 180921
    }
  },
  {
    "ph": "s", "id": 180921, "pid": 76337, "tid": -914061504, "ts": 1716454224272251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224317716, "dur": 121,
    "args": {
      "External id": 180923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180923,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180923, "pid": 5, "tid": 7, "ts": 1716454224317716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272272, "dur": 8,
    "args": {
      "External id": 180923,
      "cbid": 211, "correlation": 180923
    }
  },
  {
    "ph": "s", "id": 180923, "pid": 76337, "tid": -914061504, "ts": 1716454224272272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224317839, "dur": 21,
    "args": {
      "External id": 180925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180925,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180925, "pid": 5, "tid": 7, "ts": 1716454224317839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272284, "dur": 5,
    "args": {
      "External id": 180925,
      "cbid": 211, "correlation": 180925
    }
  },
  {
    "ph": "s", "id": 180925, "pid": 76337, "tid": -914061504, "ts": 1716454224272284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224317861, "dur": 33,
    "args": {
      "External id": 180931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180931,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180931, "pid": 5, "tid": 7, "ts": 1716454224317861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272312, "dur": 9,
    "args": {
      "External id": 180931,
      "cbid": 211, "correlation": 180931
    }
  },
  {
    "ph": "s", "id": 180931, "pid": 76337, "tid": -914061504, "ts": 1716454224272312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224317895, "dur": 201,
    "args": {
      "External id": 180940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180940,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180940, "pid": 5, "tid": 7, "ts": 1716454224317895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272394, "dur": 14,
    "args": {
      "External id": 180940,
      "cbid": 211, "correlation": 180940
    }
  },
  {
    "ph": "s", "id": 180940, "pid": 76337, "tid": -914061504, "ts": 1716454224272394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224318098, "dur": 66,
    "args": {
      "External id": 180962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180962,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 180962, "pid": 5, "tid": 7, "ts": 1716454224318098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272452, "dur": 10,
    "args": {
      "External id": 180962,
      "cbid": 211, "correlation": 180962
    }
  },
  {
    "ph": "s", "id": 180962, "pid": 76337, "tid": -914061504, "ts": 1716454224272452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272547, "dur": 1,
    "args": {
      "External id": 180973,
      "cbid": 251, "correlation": 180973
    }
  },
  {
    "ph": "f", "id": 180973, "pid": 76337, "tid": -914061504, "ts": 1716454224272547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224318165, "dur": 156,
    "args": {
      "External id": 180974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180974,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180974, "pid": 5, "tid": 7, "ts": 1716454224318165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272552, "dur": 14,
    "args": {
      "External id": 180974,
      "cbid": 211, "correlation": 180974
    }
  },
  {
    "ph": "s", "id": 180974, "pid": 76337, "tid": -914061504, "ts": 1716454224272552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272623, "dur": 1,
    "args": {
      "External id": 180985,
      "cbid": 251, "correlation": 180985
    }
  },
  {
    "ph": "f", "id": 180985, "pid": 76337, "tid": -914061504, "ts": 1716454224272623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224318323, "dur": 147,
    "args": {
      "External id": 180986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180986,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180986, "pid": 5, "tid": 7, "ts": 1716454224318323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272627, "dur": 11,
    "args": {
      "External id": 180986,
      "cbid": 211, "correlation": 180986
    }
  },
  {
    "ph": "s", "id": 180986, "pid": 76337, "tid": -914061504, "ts": 1716454224272627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272693, "dur": 1,
    "args": {
      "External id": 180997,
      "cbid": 251, "correlation": 180997
    }
  },
  {
    "ph": "f", "id": 180997, "pid": 76337, "tid": -914061504, "ts": 1716454224272693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224318471, "dur": 145,
    "args": {
      "External id": 180998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 180998,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 180998, "pid": 5, "tid": 7, "ts": 1716454224318471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272697, "dur": 11,
    "args": {
      "External id": 180998,
      "cbid": 211, "correlation": 180998
    }
  },
  {
    "ph": "s", "id": 180998, "pid": 76337, "tid": -914061504, "ts": 1716454224272697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224318618, "dur": 1965,
    "args": {
      "External id": 181019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181019,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 181019, "pid": 5, "tid": 7, "ts": 1716454224318618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272781, "dur": 14,
    "args": {
      "External id": 181019,
      "cbid": 211, "correlation": 181019
    }
  },
  {
    "ph": "s", "id": 181019, "pid": 76337, "tid": -914061504, "ts": 1716454224272781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272885, "dur": 1,
    "args": {
      "External id": 181037,
      "cbid": 251, "correlation": 181037
    }
  },
  {
    "ph": "f", "id": 181037, "pid": 76337, "tid": -914061504, "ts": 1716454224272885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224320583, "dur": 148,
    "args": {
      "External id": 181039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181039,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 181039, "pid": 5, "tid": 7, "ts": 1716454224320583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272891, "dur": 13,
    "args": {
      "External id": 181039,
      "cbid": 211, "correlation": 181039
    }
  },
  {
    "ph": "s", "id": 181039, "pid": 76337, "tid": -914061504, "ts": 1716454224272891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224320733, "dur": 36,
    "args": {
      "External id": 181047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181047,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181047, "pid": 5, "tid": 7, "ts": 1716454224320733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224272963, "dur": 21,
    "args": {
      "External id": 181047,
      "cbid": 211, "correlation": 181047
    }
  },
  {
    "ph": "s", "id": 181047, "pid": 76337, "tid": -914061504, "ts": 1716454224272963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224320770, "dur": 50,
    "args": {
      "External id": 181055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181055,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181055, "pid": 5, "tid": 7, "ts": 1716454224320770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273012, "dur": 9,
    "args": {
      "External id": 181055,
      "cbid": 211, "correlation": 181055
    }
  },
  {
    "ph": "s", "id": 181055, "pid": 76337, "tid": -914061504, "ts": 1716454224273012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224320821, "dur": 31,
    "args": {
      "External id": 181066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181066, "pid": 5, "tid": 7, "ts": 1716454224320821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273086, "dur": 13,
    "args": {
      "External id": 181066,
      "cbid": 211, "correlation": 181066
    }
  },
  {
    "ph": "s", "id": 181066, "pid": 76337, "tid": -914061504, "ts": 1716454224273086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224320853, "dur": 35,
    "args": {
      "External id": 181088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181088,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181088, "pid": 5, "tid": 7, "ts": 1716454224320853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273118, "dur": 8,
    "args": {
      "External id": 181088,
      "cbid": 211, "correlation": 181088
    }
  },
  {
    "ph": "s", "id": 181088, "pid": 76337, "tid": -914061504, "ts": 1716454224273118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273201, "dur": 1,
    "args": {
      "External id": 181099,
      "cbid": 251, "correlation": 181099
    }
  },
  {
    "ph": "f", "id": 181099, "pid": 76337, "tid": -914061504, "ts": 1716454224273201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224320889, "dur": 90,
    "args": {
      "External id": 181100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181100,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181100, "pid": 5, "tid": 7, "ts": 1716454224320889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273207, "dur": 14,
    "args": {
      "External id": 181100,
      "cbid": 211, "correlation": 181100
    }
  },
  {
    "ph": "s", "id": 181100, "pid": 76337, "tid": -914061504, "ts": 1716454224273207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273277, "dur": 1,
    "args": {
      "External id": 181111,
      "cbid": 251, "correlation": 181111
    }
  },
  {
    "ph": "f", "id": 181111, "pid": 76337, "tid": -914061504, "ts": 1716454224273277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273281, "dur": 0,
    "args": {
      "External id": 181112,
      "cbid": 251, "correlation": 181112
    }
  },
  {
    "ph": "f", "id": 181112, "pid": 76337, "tid": -914061504, "ts": 1716454224273281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224320981, "dur": 12,
    "args": {
      "External id": 181113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181113,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 181113, "pid": 5, "tid": 7, "ts": 1716454224320981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273282, "dur": 12,
    "args": {
      "External id": 181113,
      "cbid": 211, "correlation": 181113
    }
  },
  {
    "ph": "s", "id": 181113, "pid": 76337, "tid": -914061504, "ts": 1716454224273282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224320994, "dur": 5,
    "args": {
      "External id": 181115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181115,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 181115, "pid": 5, "tid": 7, "ts": 1716454224320994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273297, "dur": 7,
    "args": {
      "External id": 181115,
      "cbid": 211, "correlation": 181115
    }
  },
  {
    "ph": "s", "id": 181115, "pid": 76337, "tid": -914061504, "ts": 1716454224273297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273355, "dur": 1,
    "args": {
      "External id": 181126,
      "cbid": 251, "correlation": 181126
    }
  },
  {
    "ph": "f", "id": 181126, "pid": 76337, "tid": -914061504, "ts": 1716454224273355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273359, "dur": 0,
    "args": {
      "External id": 181127,
      "cbid": 251, "correlation": 181127
    }
  },
  {
    "ph": "f", "id": 181127, "pid": 76337, "tid": -914061504, "ts": 1716454224273359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224321000, "dur": 7,
    "args": {
      "External id": 181128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181128,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 181128, "pid": 5, "tid": 7, "ts": 1716454224321000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273360, "dur": 12,
    "args": {
      "External id": 181128,
      "cbid": 211, "correlation": 181128
    }
  },
  {
    "ph": "s", "id": 181128, "pid": 76337, "tid": -914061504, "ts": 1716454224273360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224321009, "dur": 4,
    "args": {
      "External id": 181130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181130,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 181130, "pid": 5, "tid": 7, "ts": 1716454224321009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273374, "dur": 5,
    "args": {
      "External id": 181130,
      "cbid": 211, "correlation": 181130
    }
  },
  {
    "ph": "s", "id": 181130, "pid": 76337, "tid": -914061504, "ts": 1716454224273374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224321014, "dur": 92,
    "args": {
      "External id": 181151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181151,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 181151, "pid": 5, "tid": 7, "ts": 1716454224321014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273447, "dur": 12,
    "args": {
      "External id": 181151,
      "cbid": 211, "correlation": 181151
    }
  },
  {
    "ph": "s", "id": 181151, "pid": 76337, "tid": -914061504, "ts": 1716454224273447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273544, "dur": 1,
    "args": {
      "External id": 181169,
      "cbid": 251, "correlation": 181169
    }
  },
  {
    "ph": "f", "id": 181169, "pid": 76337, "tid": -914061504, "ts": 1716454224273544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224321107, "dur": 98,
    "args": {
      "External id": 181171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181171,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181171, "pid": 5, "tid": 7, "ts": 1716454224321107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273550, "dur": 13,
    "args": {
      "External id": 181171,
      "cbid": 211, "correlation": 181171
    }
  },
  {
    "ph": "s", "id": 181171, "pid": 76337, "tid": -914061504, "ts": 1716454224273550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224321207, "dur": 19,
    "args": {
      "External id": 181179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181179,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181179, "pid": 5, "tid": 7, "ts": 1716454224321207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273620, "dur": 12,
    "args": {
      "External id": 181179,
      "cbid": 211, "correlation": 181179
    }
  },
  {
    "ph": "s", "id": 181179, "pid": 76337, "tid": -914061504, "ts": 1716454224273620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224321227, "dur": 37,
    "args": {
      "External id": 181187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181187,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181187, "pid": 5, "tid": 7, "ts": 1716454224321227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273661, "dur": 10,
    "args": {
      "External id": 181187,
      "cbid": 211, "correlation": 181187
    }
  },
  {
    "ph": "s", "id": 181187, "pid": 76337, "tid": -914061504, "ts": 1716454224273661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224321265, "dur": 35,
    "args": {
      "External id": 181209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181209,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181209, "pid": 5, "tid": 7, "ts": 1716454224321265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273712, "dur": 11,
    "args": {
      "External id": 181209,
      "cbid": 211, "correlation": 181209
    }
  },
  {
    "ph": "s", "id": 181209, "pid": 76337, "tid": -914061504, "ts": 1716454224273712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273807, "dur": 1,
    "args": {
      "External id": 181225,
      "cbid": 251, "correlation": 181225
    }
  },
  {
    "ph": "f", "id": 181225, "pid": 76337, "tid": -914061504, "ts": 1716454224273807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273812, "dur": 0,
    "args": {
      "External id": 181227,
      "cbid": 251, "correlation": 181227
    }
  },
  {
    "ph": "f", "id": 181227, "pid": 76337, "tid": -914061504, "ts": 1716454224273812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224321302, "dur": 547,
    "args": {
      "External id": 181228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181228,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 181228, "pid": 5, "tid": 7, "ts": 1716454224321302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273815, "dur": 13,
    "args": {
      "External id": 181228,
      "cbid": 211, "correlation": 181228
    }
  },
  {
    "ph": "s", "id": 181228, "pid": 76337, "tid": -914061504, "ts": 1716454224273815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224321850, "dur": 126,
    "args": {
      "External id": 181236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181236, "pid": 5, "tid": 7, "ts": 1716454224321850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273881, "dur": 12,
    "args": {
      "External id": 181236,
      "cbid": 211, "correlation": 181236
    }
  },
  {
    "ph": "s", "id": 181236, "pid": 76337, "tid": -914061504, "ts": 1716454224273881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224321978, "dur": 129,
    "args": {
      "External id": 181244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181244, "pid": 5, "tid": 7, "ts": 1716454224321978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224273912, "dur": 9,
    "args": {
      "External id": 181244,
      "cbid": 211, "correlation": 181244
    }
  },
  {
    "ph": "s", "id": 181244, "pid": 76337, "tid": -914061504, "ts": 1716454224273912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274002, "dur": 1,
    "args": {
      "External id": 181260,
      "cbid": 251, "correlation": 181260
    }
  },
  {
    "ph": "f", "id": 181260, "pid": 76337, "tid": -914061504, "ts": 1716454224274002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224322108, "dur": 308,
    "args": {
      "External id": 181262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181262,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181262, "pid": 5, "tid": 7, "ts": 1716454224322108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274008, "dur": 14,
    "args": {
      "External id": 181262,
      "cbid": 211, "correlation": 181262
    }
  },
  {
    "ph": "s", "id": 181262, "pid": 76337, "tid": -914061504, "ts": 1716454224274008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224322417, "dur": 27,
    "args": {
      "External id": 181270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181270,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181270, "pid": 5, "tid": 7, "ts": 1716454224322417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274052, "dur": 10,
    "args": {
      "External id": 181270,
      "cbid": 211, "correlation": 181270
    }
  },
  {
    "ph": "s", "id": 181270, "pid": 76337, "tid": -914061504, "ts": 1716454224274052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224322445, "dur": 82,
    "args": {
      "External id": 181281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181281,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181281, "pid": 5, "tid": 7, "ts": 1716454224322445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274121, "dur": 12,
    "args": {
      "External id": 181281,
      "cbid": 211, "correlation": 181281
    }
  },
  {
    "ph": "s", "id": 181281, "pid": 76337, "tid": -914061504, "ts": 1716454224274121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274189, "dur": 0,
    "args": {
      "External id": 181293,
      "cbid": 317, "correlation": 181293
    }
  },
  {
    "ph": "f", "id": 181293, "pid": 76337, "tid": -914061504, "ts": 1716454224274189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274190, "dur": 0,
    "args": {
      "External id": 181294,
      "cbid": 203, "correlation": 181294
    }
  },
  {
    "ph": "f", "id": 181294, "pid": 76337, "tid": -914061504, "ts": 1716454224274190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274191, "dur": 0,
    "args": {
      "External id": 181295,
      "cbid": 205, "correlation": 181295
    }
  },
  {
    "ph": "f", "id": 181295, "pid": 76337, "tid": -914061504, "ts": 1716454224274191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224322529, "dur": 22,
    "args": {
      "External id": 181299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181299,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181299, "pid": 5, "tid": 7, "ts": 1716454224322529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274206, "dur": 13,
    "args": {
      "External id": 181299,
      "cbid": 211, "correlation": 181299
    }
  },
  {
    "ph": "s", "id": 181299, "pid": 76337, "tid": -914061504, "ts": 1716454224274206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224322552, "dur": 122,
    "args": {
      "External id": 181301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181301,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181301, "pid": 5, "tid": 7, "ts": 1716454224322552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274225, "dur": 7,
    "args": {
      "External id": 181301,
      "cbid": 211, "correlation": 181301
    }
  },
  {
    "ph": "s", "id": 181301, "pid": 76337, "tid": -914061504, "ts": 1716454224274225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224322676, "dur": 22,
    "args": {
      "External id": 181303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181303,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181303, "pid": 5, "tid": 7, "ts": 1716454224322676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274236, "dur": 5,
    "args": {
      "External id": 181303,
      "cbid": 211, "correlation": 181303
    }
  },
  {
    "ph": "s", "id": 181303, "pid": 76337, "tid": -914061504, "ts": 1716454224274236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224322699, "dur": 33,
    "args": {
      "External id": 181309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181309, "pid": 5, "tid": 7, "ts": 1716454224322699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274263, "dur": 9,
    "args": {
      "External id": 181309,
      "cbid": 211, "correlation": 181309
    }
  },
  {
    "ph": "s", "id": 181309, "pid": 76337, "tid": -914061504, "ts": 1716454224274263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224322733, "dur": 27,
    "args": {
      "External id": 181317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181317,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181317, "pid": 5, "tid": 7, "ts": 1716454224322733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274295, "dur": 8,
    "args": {
      "External id": 181317,
      "cbid": 211, "correlation": 181317
    }
  },
  {
    "ph": "s", "id": 181317, "pid": 76337, "tid": -914061504, "ts": 1716454224274295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224322761, "dur": 31,
    "args": {
      "External id": 181337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181337,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 181337, "pid": 5, "tid": 7, "ts": 1716454224322761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274368, "dur": 12,
    "args": {
      "External id": 181337,
      "cbid": 211, "correlation": 181337
    }
  },
  {
    "ph": "s", "id": 181337, "pid": 76337, "tid": -914061504, "ts": 1716454224274368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224322793, "dur": 4,
    "args": {
      "External id": 181349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181349,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 181349, "pid": 5, "tid": 7, "ts": 1716454224322793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274390, "dur": 6,
    "args": {
      "External id": 181349,
      "cbid": 211, "correlation": 181349
    }
  },
  {
    "ph": "s", "id": 181349, "pid": 76337, "tid": -914061504, "ts": 1716454224274390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224322799, "dur": 33,
    "args": {
      "External id": 181352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181352, "pid": 5, "tid": 7, "ts": 1716454224322799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274407, "dur": 7,
    "args": {
      "External id": 181352,
      "cbid": 211, "correlation": 181352
    }
  },
  {
    "ph": "s", "id": 181352, "pid": 76337, "tid": -914061504, "ts": 1716454224274407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224322833, "dur": 21,
    "args": {
      "External id": 181361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181361,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181361, "pid": 5, "tid": 7, "ts": 1716454224322833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274446, "dur": 9,
    "args": {
      "External id": 181361,
      "cbid": 211, "correlation": 181361
    }
  },
  {
    "ph": "s", "id": 181361, "pid": 76337, "tid": -914061504, "ts": 1716454224274446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274498, "dur": 0,
    "args": {
      "External id": 181371,
      "cbid": 317, "correlation": 181371
    }
  },
  {
    "ph": "f", "id": 181371, "pid": 76337, "tid": -914061504, "ts": 1716454224274498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274499, "dur": 0,
    "args": {
      "External id": 181372,
      "cbid": 203, "correlation": 181372
    }
  },
  {
    "ph": "f", "id": 181372, "pid": 76337, "tid": -914061504, "ts": 1716454224274499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274499, "dur": 0,
    "args": {
      "External id": 181373,
      "cbid": 205, "correlation": 181373
    }
  },
  {
    "ph": "f", "id": 181373, "pid": 76337, "tid": -914061504, "ts": 1716454224274499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224322856, "dur": 23,
    "args": {
      "External id": 181377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181377,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181377, "pid": 5, "tid": 7, "ts": 1716454224322856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274512, "dur": 12,
    "args": {
      "External id": 181377,
      "cbid": 211, "correlation": 181377
    }
  },
  {
    "ph": "s", "id": 181377, "pid": 76337, "tid": -914061504, "ts": 1716454224274512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224322880, "dur": 45,
    "args": {
      "External id": 181379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181379,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181379, "pid": 5, "tid": 7, "ts": 1716454224322880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274527, "dur": 5,
    "args": {
      "External id": 181379,
      "cbid": 211, "correlation": 181379
    }
  },
  {
    "ph": "s", "id": 181379, "pid": 76337, "tid": -914061504, "ts": 1716454224274527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224322926, "dur": 659,
    "args": {
      "External id": 181381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181381,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181381, "pid": 5, "tid": 7, "ts": 1716454224322926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274538, "dur": 6,
    "args": {
      "External id": 181381,
      "cbid": 211, "correlation": 181381
    }
  },
  {
    "ph": "s", "id": 181381, "pid": 76337, "tid": -914061504, "ts": 1716454224274538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224323587, "dur": 24,
    "args": {
      "External id": 181383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181383,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181383, "pid": 5, "tid": 7, "ts": 1716454224323587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274548, "dur": 5,
    "args": {
      "External id": 181383,
      "cbid": 211, "correlation": 181383
    }
  },
  {
    "ph": "s", "id": 181383, "pid": 76337, "tid": -914061504, "ts": 1716454224274548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224323612, "dur": 34,
    "args": {
      "External id": 181389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181389, "pid": 5, "tid": 7, "ts": 1716454224323612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274576, "dur": 8,
    "args": {
      "External id": 181389,
      "cbid": 211, "correlation": 181389
    }
  },
  {
    "ph": "s", "id": 181389, "pid": 76337, "tid": -914061504, "ts": 1716454224274576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224323647, "dur": 3,
    "args": {
      "External id": 181397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181397,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 181397, "pid": 5, "tid": 7, "ts": 1716454224323647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274619, "dur": 10,
    "args": {
      "External id": 181397,
      "cbid": 211, "correlation": 181397
    }
  },
  {
    "ph": "s", "id": 181397, "pid": 76337, "tid": -914061504, "ts": 1716454224274619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274684, "dur": 1,
    "args": {
      "External id": 181413,
      "cbid": 251, "correlation": 181413
    }
  },
  {
    "ph": "f", "id": 181413, "pid": 76337, "tid": -914061504, "ts": 1716454224274684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274689, "dur": 0,
    "args": {
      "External id": 181415,
      "cbid": 251, "correlation": 181415
    }
  },
  {
    "ph": "f", "id": 181415, "pid": 76337, "tid": -914061504, "ts": 1716454224274689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224323652, "dur": 13,
    "args": {
      "External id": 181416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181416,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 181416, "pid": 5, "tid": 7, "ts": 1716454224323652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274691, "dur": 11,
    "args": {
      "External id": 181416,
      "cbid": 211, "correlation": 181416
    }
  },
  {
    "ph": "s", "id": 181416, "pid": 76337, "tid": -914061504, "ts": 1716454224274691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224323666, "dur": 5,
    "args": {
      "External id": 181418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181418,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 181418, "pid": 5, "tid": 7, "ts": 1716454224323666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274704, "dur": 5,
    "args": {
      "External id": 181418,
      "cbid": 211, "correlation": 181418
    }
  },
  {
    "ph": "s", "id": 181418, "pid": 76337, "tid": -914061504, "ts": 1716454224274704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224323672, "dur": 29,
    "args": {
      "External id": 181428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181428,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181428, "pid": 5, "tid": 7, "ts": 1716454224323672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274763, "dur": 14,
    "args": {
      "External id": 181428,
      "cbid": 211, "correlation": 181428
    }
  },
  {
    "ph": "s", "id": 181428, "pid": 76337, "tid": -914061504, "ts": 1716454224274763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224323703, "dur": 31,
    "args": {
      "External id": 181448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181448,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 181448, "pid": 5, "tid": 7, "ts": 1716454224323703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274831, "dur": 11,
    "args": {
      "External id": 181448,
      "cbid": 211, "correlation": 181448
    }
  },
  {
    "ph": "s", "id": 181448, "pid": 76337, "tid": -914061504, "ts": 1716454224274831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224323735, "dur": 4,
    "args": {
      "External id": 181460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181460,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 181460, "pid": 5, "tid": 7, "ts": 1716454224323735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274851, "dur": 7,
    "args": {
      "External id": 181460,
      "cbid": 211, "correlation": 181460
    }
  },
  {
    "ph": "s", "id": 181460, "pid": 76337, "tid": -914061504, "ts": 1716454224274851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224323740, "dur": 30,
    "args": {
      "External id": 181463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181463, "pid": 5, "tid": 7, "ts": 1716454224323740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274871, "dur": 7,
    "args": {
      "External id": 181463,
      "cbid": 211, "correlation": 181463
    }
  },
  {
    "ph": "s", "id": 181463, "pid": 76337, "tid": -914061504, "ts": 1716454224274871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224323771, "dur": 20,
    "args": {
      "External id": 181472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181472,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181472, "pid": 5, "tid": 7, "ts": 1716454224323771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274912, "dur": 10,
    "args": {
      "External id": 181472,
      "cbid": 211, "correlation": 181472
    }
  },
  {
    "ph": "s", "id": 181472, "pid": 76337, "tid": -914061504, "ts": 1716454224274912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274984, "dur": 0,
    "args": {
      "External id": 181482,
      "cbid": 317, "correlation": 181482
    }
  },
  {
    "ph": "f", "id": 181482, "pid": 76337, "tid": -914061504, "ts": 1716454224274984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274985, "dur": 0,
    "args": {
      "External id": 181483,
      "cbid": 203, "correlation": 181483
    }
  },
  {
    "ph": "f", "id": 181483, "pid": 76337, "tid": -914061504, "ts": 1716454224274985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224274986, "dur": 0,
    "args": {
      "External id": 181484,
      "cbid": 205, "correlation": 181484
    }
  },
  {
    "ph": "f", "id": 181484, "pid": 76337, "tid": -914061504, "ts": 1716454224274986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224323792, "dur": 23,
    "args": {
      "External id": 181488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181488,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181488, "pid": 5, "tid": 7, "ts": 1716454224323792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275000, "dur": 12,
    "args": {
      "External id": 181488,
      "cbid": 211, "correlation": 181488
    }
  },
  {
    "ph": "s", "id": 181488, "pid": 76337, "tid": -914061504, "ts": 1716454224275000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224323816, "dur": 45,
    "args": {
      "External id": 181490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181490, "pid": 5, "tid": 7, "ts": 1716454224323816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275014, "dur": 5,
    "args": {
      "External id": 181490,
      "cbid": 211, "correlation": 181490
    }
  },
  {
    "ph": "s", "id": 181490, "pid": 76337, "tid": -914061504, "ts": 1716454224275014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224323862, "dur": 651,
    "args": {
      "External id": 181492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181492,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181492, "pid": 5, "tid": 7, "ts": 1716454224323862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275025, "dur": 7,
    "args": {
      "External id": 181492,
      "cbid": 211, "correlation": 181492
    }
  },
  {
    "ph": "s", "id": 181492, "pid": 76337, "tid": -914061504, "ts": 1716454224275025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224324515, "dur": 21,
    "args": {
      "External id": 181494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181494,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181494, "pid": 5, "tid": 7, "ts": 1716454224324515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275036, "dur": 5,
    "args": {
      "External id": 181494,
      "cbid": 211, "correlation": 181494
    }
  },
  {
    "ph": "s", "id": 181494, "pid": 76337, "tid": -914061504, "ts": 1716454224275036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224324537, "dur": 34,
    "args": {
      "External id": 181500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181500, "pid": 5, "tid": 7, "ts": 1716454224324537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275064, "dur": 8,
    "args": {
      "External id": 181500,
      "cbid": 211, "correlation": 181500
    }
  },
  {
    "ph": "s", "id": 181500, "pid": 76337, "tid": -914061504, "ts": 1716454224275064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224324572, "dur": 27,
    "args": {
      "External id": 181508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181508,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181508, "pid": 5, "tid": 7, "ts": 1716454224324572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275097, "dur": 8,
    "args": {
      "External id": 181508,
      "cbid": 211, "correlation": 181508
    }
  },
  {
    "ph": "s", "id": 181508, "pid": 76337, "tid": -914061504, "ts": 1716454224275097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224324600, "dur": 19,
    "args": {
      "External id": 181516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181516,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181516, "pid": 5, "tid": 7, "ts": 1716454224324600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275126, "dur": 9,
    "args": {
      "External id": 181516,
      "cbid": 211, "correlation": 181516
    }
  },
  {
    "ph": "s", "id": 181516, "pid": 76337, "tid": -914061504, "ts": 1716454224275126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224324621, "dur": 31,
    "args": {
      "External id": 181536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181536,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 181536, "pid": 5, "tid": 7, "ts": 1716454224324621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275206, "dur": 13,
    "args": {
      "External id": 181536,
      "cbid": 211, "correlation": 181536
    }
  },
  {
    "ph": "s", "id": 181536, "pid": 76337, "tid": -914061504, "ts": 1716454224275206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224324652, "dur": 4,
    "args": {
      "External id": 181548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181548,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 181548, "pid": 5, "tid": 7, "ts": 1716454224324652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275229, "dur": 7,
    "args": {
      "External id": 181548,
      "cbid": 211, "correlation": 181548
    }
  },
  {
    "ph": "s", "id": 181548, "pid": 76337, "tid": -914061504, "ts": 1716454224275229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224324658, "dur": 30,
    "args": {
      "External id": 181551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181551,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181551, "pid": 5, "tid": 7, "ts": 1716454224324658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275247, "dur": 6,
    "args": {
      "External id": 181551,
      "cbid": 211, "correlation": 181551
    }
  },
  {
    "ph": "s", "id": 181551, "pid": 76337, "tid": -914061504, "ts": 1716454224275247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275306, "dur": 0,
    "args": {
      "External id": 181562,
      "cbid": 317, "correlation": 181562
    }
  },
  {
    "ph": "f", "id": 181562, "pid": 76337, "tid": -914061504, "ts": 1716454224275306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275306, "dur": 0,
    "args": {
      "External id": 181563,
      "cbid": 203, "correlation": 181563
    }
  },
  {
    "ph": "f", "id": 181563, "pid": 76337, "tid": -914061504, "ts": 1716454224275306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275307, "dur": 0,
    "args": {
      "External id": 181564,
      "cbid": 205, "correlation": 181564
    }
  },
  {
    "ph": "f", "id": 181564, "pid": 76337, "tid": -914061504, "ts": 1716454224275307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224324689, "dur": 23,
    "args": {
      "External id": 181568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181568,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181568, "pid": 5, "tid": 7, "ts": 1716454224324689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275320, "dur": 12,
    "args": {
      "External id": 181568,
      "cbid": 211, "correlation": 181568
    }
  },
  {
    "ph": "s", "id": 181568, "pid": 76337, "tid": -914061504, "ts": 1716454224275320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224324713, "dur": 117,
    "args": {
      "External id": 181570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181570,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181570, "pid": 5, "tid": 7, "ts": 1716454224324713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275338, "dur": 7,
    "args": {
      "External id": 181570,
      "cbid": 211, "correlation": 181570
    }
  },
  {
    "ph": "s", "id": 181570, "pid": 76337, "tid": -914061504, "ts": 1716454224275338,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224324832, "dur": 22,
    "args": {
      "External id": 181572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181572,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181572, "pid": 5, "tid": 7, "ts": 1716454224324832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275348, "dur": 5,
    "args": {
      "External id": 181572,
      "cbid": 211, "correlation": 181572
    }
  },
  {
    "ph": "s", "id": 181572, "pid": 76337, "tid": -914061504, "ts": 1716454224275348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224324855, "dur": 32,
    "args": {
      "External id": 181578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181578, "pid": 5, "tid": 7, "ts": 1716454224324855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275376, "dur": 9,
    "args": {
      "External id": 181578,
      "cbid": 211, "correlation": 181578
    }
  },
  {
    "ph": "s", "id": 181578, "pid": 76337, "tid": -914061504, "ts": 1716454224275376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224324889, "dur": 199,
    "args": {
      "External id": 181587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181587,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181587, "pid": 5, "tid": 7, "ts": 1716454224324889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275459, "dur": 14,
    "args": {
      "External id": 181587,
      "cbid": 211, "correlation": 181587
    }
  },
  {
    "ph": "s", "id": 181587, "pid": 76337, "tid": -914061504, "ts": 1716454224275459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224325089, "dur": 65,
    "args": {
      "External id": 181609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181609,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181609, "pid": 5, "tid": 7, "ts": 1716454224325089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275517, "dur": 10,
    "args": {
      "External id": 181609,
      "cbid": 211, "correlation": 181609
    }
  },
  {
    "ph": "s", "id": 181609, "pid": 76337, "tid": -914061504, "ts": 1716454224275517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275606, "dur": 2,
    "args": {
      "External id": 181620,
      "cbid": 251, "correlation": 181620
    }
  },
  {
    "ph": "f", "id": 181620, "pid": 76337, "tid": -914061504, "ts": 1716454224275606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224325156, "dur": 154,
    "args": {
      "External id": 181621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181621,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181621, "pid": 5, "tid": 7, "ts": 1716454224325156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275612, "dur": 14,
    "args": {
      "External id": 181621,
      "cbid": 211, "correlation": 181621
    }
  },
  {
    "ph": "s", "id": 181621, "pid": 76337, "tid": -914061504, "ts": 1716454224275612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275683, "dur": 1,
    "args": {
      "External id": 181632,
      "cbid": 251, "correlation": 181632
    }
  },
  {
    "ph": "f", "id": 181632, "pid": 76337, "tid": -914061504, "ts": 1716454224275683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224325311, "dur": 145,
    "args": {
      "External id": 181633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181633,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181633, "pid": 5, "tid": 7, "ts": 1716454224325311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275687, "dur": 12,
    "args": {
      "External id": 181633,
      "cbid": 211, "correlation": 181633
    }
  },
  {
    "ph": "s", "id": 181633, "pid": 76337, "tid": -914061504, "ts": 1716454224275687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275752, "dur": 1,
    "args": {
      "External id": 181644,
      "cbid": 251, "correlation": 181644
    }
  },
  {
    "ph": "f", "id": 181644, "pid": 76337, "tid": -914061504, "ts": 1716454224275752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224325457, "dur": 141,
    "args": {
      "External id": 181645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181645,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181645, "pid": 5, "tid": 7, "ts": 1716454224325457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275756, "dur": 11,
    "args": {
      "External id": 181645,
      "cbid": 211, "correlation": 181645
    }
  },
  {
    "ph": "s", "id": 181645, "pid": 76337, "tid": -914061504, "ts": 1716454224275756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224325600, "dur": 1870,
    "args": {
      "External id": 181666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181666,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 181666, "pid": 5, "tid": 7, "ts": 1716454224325600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275837, "dur": 13,
    "args": {
      "External id": 181666,
      "cbid": 211, "correlation": 181666
    }
  },
  {
    "ph": "s", "id": 181666, "pid": 76337, "tid": -914061504, "ts": 1716454224275837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275938, "dur": 1,
    "args": {
      "External id": 181684,
      "cbid": 251, "correlation": 181684
    }
  },
  {
    "ph": "f", "id": 181684, "pid": 76337, "tid": -914061504, "ts": 1716454224275938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224327471, "dur": 143,
    "args": {
      "External id": 181686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181686,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 181686, "pid": 5, "tid": 7, "ts": 1716454224327471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224275944, "dur": 13,
    "args": {
      "External id": 181686,
      "cbid": 211, "correlation": 181686
    }
  },
  {
    "ph": "s", "id": 181686, "pid": 76337, "tid": -914061504, "ts": 1716454224275944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224327615, "dur": 35,
    "args": {
      "External id": 181694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181694,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181694, "pid": 5, "tid": 7, "ts": 1716454224327615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276026, "dur": 13,
    "args": {
      "External id": 181694,
      "cbid": 211, "correlation": 181694
    }
  },
  {
    "ph": "s", "id": 181694, "pid": 76337, "tid": -914061504, "ts": 1716454224276026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224327652, "dur": 51,
    "args": {
      "External id": 181702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181702,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181702, "pid": 5, "tid": 7, "ts": 1716454224327652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276066, "dur": 8,
    "args": {
      "External id": 181702,
      "cbid": 211, "correlation": 181702
    }
  },
  {
    "ph": "s", "id": 181702, "pid": 76337, "tid": -914061504, "ts": 1716454224276066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224327704, "dur": 31,
    "args": {
      "External id": 181713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181713,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181713, "pid": 5, "tid": 7, "ts": 1716454224327704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276137, "dur": 14,
    "args": {
      "External id": 181713,
      "cbid": 211, "correlation": 181713
    }
  },
  {
    "ph": "s", "id": 181713, "pid": 76337, "tid": -914061504, "ts": 1716454224276137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224327735, "dur": 33,
    "args": {
      "External id": 181735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181735,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181735, "pid": 5, "tid": 7, "ts": 1716454224327735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276168, "dur": 7,
    "args": {
      "External id": 181735,
      "cbid": 211, "correlation": 181735
    }
  },
  {
    "ph": "s", "id": 181735, "pid": 76337, "tid": -914061504, "ts": 1716454224276168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276253, "dur": 1,
    "args": {
      "External id": 181746,
      "cbid": 251, "correlation": 181746
    }
  },
  {
    "ph": "f", "id": 181746, "pid": 76337, "tid": -914061504, "ts": 1716454224276253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224327770, "dur": 74,
    "args": {
      "External id": 181747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181747,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181747, "pid": 5, "tid": 7, "ts": 1716454224327770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276259, "dur": 13,
    "args": {
      "External id": 181747,
      "cbid": 211, "correlation": 181747
    }
  },
  {
    "ph": "s", "id": 181747, "pid": 76337, "tid": -914061504, "ts": 1716454224276259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276327, "dur": 1,
    "args": {
      "External id": 181758,
      "cbid": 251, "correlation": 181758
    }
  },
  {
    "ph": "f", "id": 181758, "pid": 76337, "tid": -914061504, "ts": 1716454224276327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276331, "dur": 0,
    "args": {
      "External id": 181759,
      "cbid": 251, "correlation": 181759
    }
  },
  {
    "ph": "f", "id": 181759, "pid": 76337, "tid": -914061504, "ts": 1716454224276331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224327845, "dur": 10,
    "args": {
      "External id": 181760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181760,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 181760, "pid": 5, "tid": 7, "ts": 1716454224327845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276332, "dur": 12,
    "args": {
      "External id": 181760,
      "cbid": 211, "correlation": 181760
    }
  },
  {
    "ph": "s", "id": 181760, "pid": 76337, "tid": -914061504, "ts": 1716454224276332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224327856, "dur": 5,
    "args": {
      "External id": 181762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181762,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 181762, "pid": 5, "tid": 7, "ts": 1716454224327856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276346, "dur": 7,
    "args": {
      "External id": 181762,
      "cbid": 211, "correlation": 181762
    }
  },
  {
    "ph": "s", "id": 181762, "pid": 76337, "tid": -914061504, "ts": 1716454224276346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276405, "dur": 1,
    "args": {
      "External id": 181773,
      "cbid": 251, "correlation": 181773
    }
  },
  {
    "ph": "f", "id": 181773, "pid": 76337, "tid": -914061504, "ts": 1716454224276405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276409, "dur": 0,
    "args": {
      "External id": 181774,
      "cbid": 251, "correlation": 181774
    }
  },
  {
    "ph": "f", "id": 181774, "pid": 76337, "tid": -914061504, "ts": 1716454224276409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224327862, "dur": 7,
    "args": {
      "External id": 181775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181775,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 181775, "pid": 5, "tid": 7, "ts": 1716454224327862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276410, "dur": 11,
    "args": {
      "External id": 181775,
      "cbid": 211, "correlation": 181775
    }
  },
  {
    "ph": "s", "id": 181775, "pid": 76337, "tid": -914061504, "ts": 1716454224276410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224327870, "dur": 3,
    "args": {
      "External id": 181777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181777,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 181777, "pid": 5, "tid": 7, "ts": 1716454224327870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276423, "dur": 6,
    "args": {
      "External id": 181777,
      "cbid": 211, "correlation": 181777
    }
  },
  {
    "ph": "s", "id": 181777, "pid": 76337, "tid": -914061504, "ts": 1716454224276423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224327874, "dur": 88,
    "args": {
      "External id": 181798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181798,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 181798, "pid": 5, "tid": 7, "ts": 1716454224327874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276496, "dur": 13,
    "args": {
      "External id": 181798,
      "cbid": 211, "correlation": 181798
    }
  },
  {
    "ph": "s", "id": 181798, "pid": 76337, "tid": -914061504, "ts": 1716454224276496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276595, "dur": 1,
    "args": {
      "External id": 181816,
      "cbid": 251, "correlation": 181816
    }
  },
  {
    "ph": "f", "id": 181816, "pid": 76337, "tid": -914061504, "ts": 1716454224276595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224327963, "dur": 95,
    "args": {
      "External id": 181818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181818,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181818, "pid": 5, "tid": 7, "ts": 1716454224327963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276601, "dur": 14,
    "args": {
      "External id": 181818,
      "cbid": 211, "correlation": 181818
    }
  },
  {
    "ph": "s", "id": 181818, "pid": 76337, "tid": -914061504, "ts": 1716454224276601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224328060, "dur": 19,
    "args": {
      "External id": 181826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181826,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181826, "pid": 5, "tid": 7, "ts": 1716454224328060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276671, "dur": 12,
    "args": {
      "External id": 181826,
      "cbid": 211, "correlation": 181826
    }
  },
  {
    "ph": "s", "id": 181826, "pid": 76337, "tid": -914061504, "ts": 1716454224276671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224328081, "dur": 37,
    "args": {
      "External id": 181834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181834,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181834, "pid": 5, "tid": 7, "ts": 1716454224328081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276713, "dur": 9,
    "args": {
      "External id": 181834,
      "cbid": 211, "correlation": 181834
    }
  },
  {
    "ph": "s", "id": 181834, "pid": 76337, "tid": -914061504, "ts": 1716454224276713,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224328119, "dur": 33,
    "args": {
      "External id": 181856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181856,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181856, "pid": 5, "tid": 7, "ts": 1716454224328119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276764, "dur": 10,
    "args": {
      "External id": 181856,
      "cbid": 211, "correlation": 181856
    }
  },
  {
    "ph": "s", "id": 181856, "pid": 76337, "tid": -914061504, "ts": 1716454224276764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276853, "dur": 1,
    "args": {
      "External id": 181872,
      "cbid": 251, "correlation": 181872
    }
  },
  {
    "ph": "f", "id": 181872, "pid": 76337, "tid": -914061504, "ts": 1716454224276853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276858, "dur": 0,
    "args": {
      "External id": 181874,
      "cbid": 251, "correlation": 181874
    }
  },
  {
    "ph": "f", "id": 181874, "pid": 76337, "tid": -914061504, "ts": 1716454224276858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224328153, "dur": 525,
    "args": {
      "External id": 181875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181875,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 181875, "pid": 5, "tid": 7, "ts": 1716454224328153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276861, "dur": 13,
    "args": {
      "External id": 181875,
      "cbid": 211, "correlation": 181875
    }
  },
  {
    "ph": "s", "id": 181875, "pid": 76337, "tid": -914061504, "ts": 1716454224276861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224328679, "dur": 121,
    "args": {
      "External id": 181883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181883, "pid": 5, "tid": 7, "ts": 1716454224328679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276930, "dur": 12,
    "args": {
      "External id": 181883,
      "cbid": 211, "correlation": 181883
    }
  },
  {
    "ph": "s", "id": 181883, "pid": 76337, "tid": -914061504, "ts": 1716454224276930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224328802, "dur": 128,
    "args": {
      "External id": 181891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181891, "pid": 5, "tid": 7, "ts": 1716454224328802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224276960, "dur": 9,
    "args": {
      "External id": 181891,
      "cbid": 211, "correlation": 181891
    }
  },
  {
    "ph": "s", "id": 181891, "pid": 76337, "tid": -914061504, "ts": 1716454224276960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277046, "dur": 1,
    "args": {
      "External id": 181907,
      "cbid": 251, "correlation": 181907
    }
  },
  {
    "ph": "f", "id": 181907, "pid": 76337, "tid": -914061504, "ts": 1716454224277046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224328932, "dur": 298,
    "args": {
      "External id": 181909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181909,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181909, "pid": 5, "tid": 7, "ts": 1716454224328932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277051, "dur": 13,
    "args": {
      "External id": 181909,
      "cbid": 211, "correlation": 181909
    }
  },
  {
    "ph": "s", "id": 181909, "pid": 76337, "tid": -914061504, "ts": 1716454224277051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224329231, "dur": 27,
    "args": {
      "External id": 181917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181917,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181917, "pid": 5, "tid": 7, "ts": 1716454224329231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277094, "dur": 10,
    "args": {
      "External id": 181917,
      "cbid": 211, "correlation": 181917
    }
  },
  {
    "ph": "s", "id": 181917, "pid": 76337, "tid": -914061504, "ts": 1716454224277094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224329259, "dur": 78,
    "args": {
      "External id": 181928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181928, "pid": 5, "tid": 7, "ts": 1716454224329259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277161, "dur": 13,
    "args": {
      "External id": 181928,
      "cbid": 211, "correlation": 181928
    }
  },
  {
    "ph": "s", "id": 181928, "pid": 76337, "tid": -914061504, "ts": 1716454224277161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277228, "dur": 0,
    "args": {
      "External id": 181940,
      "cbid": 317, "correlation": 181940
    }
  },
  {
    "ph": "f", "id": 181940, "pid": 76337, "tid": -914061504, "ts": 1716454224277228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277228, "dur": 0,
    "args": {
      "External id": 181941,
      "cbid": 203, "correlation": 181941
    }
  },
  {
    "ph": "f", "id": 181941, "pid": 76337, "tid": -914061504, "ts": 1716454224277228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277229, "dur": 0,
    "args": {
      "External id": 181942,
      "cbid": 205, "correlation": 181942
    }
  },
  {
    "ph": "f", "id": 181942, "pid": 76337, "tid": -914061504, "ts": 1716454224277229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329339, "dur": 23,
    "args": {
      "External id": 181946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181946,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181946, "pid": 5, "tid": 7, "ts": 1716454224329339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277244, "dur": 12,
    "args": {
      "External id": 181946,
      "cbid": 211, "correlation": 181946
    }
  },
  {
    "ph": "s", "id": 181946, "pid": 76337, "tid": -914061504, "ts": 1716454224277244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224329363, "dur": 116,
    "args": {
      "External id": 181948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181948,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 181948, "pid": 5, "tid": 7, "ts": 1716454224329363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277262, "dur": 6,
    "args": {
      "External id": 181948,
      "cbid": 211, "correlation": 181948
    }
  },
  {
    "ph": "s", "id": 181948, "pid": 76337, "tid": -914061504, "ts": 1716454224277262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329481, "dur": 23,
    "args": {
      "External id": 181950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181950,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181950, "pid": 5, "tid": 7, "ts": 1716454224329481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277273, "dur": 5,
    "args": {
      "External id": 181950,
      "cbid": 211, "correlation": 181950
    }
  },
  {
    "ph": "s", "id": 181950, "pid": 76337, "tid": -914061504, "ts": 1716454224277273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224329505, "dur": 32,
    "args": {
      "External id": 181956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181956,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181956, "pid": 5, "tid": 7, "ts": 1716454224329505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277301, "dur": 8,
    "args": {
      "External id": 181956,
      "cbid": 211, "correlation": 181956
    }
  },
  {
    "ph": "s", "id": 181956, "pid": 76337, "tid": -914061504, "ts": 1716454224277301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224329538, "dur": 27,
    "args": {
      "External id": 181964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181964,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181964, "pid": 5, "tid": 7, "ts": 1716454224329538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277333, "dur": 8,
    "args": {
      "External id": 181964,
      "cbid": 211, "correlation": 181964
    }
  },
  {
    "ph": "s", "id": 181964, "pid": 76337, "tid": -914061504, "ts": 1716454224277333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277405, "dur": 0,
    "args": {
      "External id": 181974,
      "cbid": 317, "correlation": 181974
    }
  },
  {
    "ph": "f", "id": 181974, "pid": 76337, "tid": -914061504, "ts": 1716454224277405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277406, "dur": 0,
    "args": {
      "External id": 181975,
      "cbid": 203, "correlation": 181975
    }
  },
  {
    "ph": "f", "id": 181975, "pid": 76337, "tid": -914061504, "ts": 1716454224277406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277406, "dur": 0,
    "args": {
      "External id": 181976,
      "cbid": 205, "correlation": 181976
    }
  },
  {
    "ph": "f", "id": 181976, "pid": 76337, "tid": -914061504, "ts": 1716454224277406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329567, "dur": 21,
    "args": {
      "External id": 181980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181980,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181980, "pid": 5, "tid": 7, "ts": 1716454224329567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277421, "dur": 13,
    "args": {
      "External id": 181980,
      "cbid": 211, "correlation": 181980
    }
  },
  {
    "ph": "s", "id": 181980, "pid": 76337, "tid": -914061504, "ts": 1716454224277421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329589, "dur": 43,
    "args": {
      "External id": 181982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181982,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181982, "pid": 5, "tid": 7, "ts": 1716454224329589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277436, "dur": 5,
    "args": {
      "External id": 181982,
      "cbid": 211, "correlation": 181982
    }
  },
  {
    "ph": "s", "id": 181982, "pid": 76337, "tid": -914061504, "ts": 1716454224277436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224329633, "dur": 229,
    "args": {
      "External id": 181984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181984,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 181984, "pid": 5, "tid": 7, "ts": 1716454224329633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277448, "dur": 7,
    "args": {
      "External id": 181984,
      "cbid": 211, "correlation": 181984
    }
  },
  {
    "ph": "s", "id": 181984, "pid": 76337, "tid": -914061504, "ts": 1716454224277448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329864, "dur": 6,
    "args": {
      "External id": 181986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181986, "pid": 5, "tid": 7, "ts": 1716454224329864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277458, "dur": 5,
    "args": {
      "External id": 181986,
      "cbid": 211, "correlation": 181986
    }
  },
  {
    "ph": "s", "id": 181986, "pid": 76337, "tid": -914061504, "ts": 1716454224277458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224329871, "dur": 9,
    "args": {
      "External id": 181992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 181992,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 181992, "pid": 5, "tid": 7, "ts": 1716454224329871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277485, "dur": 8,
    "args": {
      "External id": 181992,
      "cbid": 211, "correlation": 181992
    }
  },
  {
    "ph": "s", "id": 181992, "pid": 76337, "tid": -914061504, "ts": 1716454224277485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329882, "dur": 12,
    "args": {
      "External id": 182012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182012,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182012, "pid": 5, "tid": 7, "ts": 1716454224329882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277577, "dur": 13,
    "args": {
      "External id": 182012,
      "cbid": 211, "correlation": 182012
    }
  },
  {
    "ph": "s", "id": 182012, "pid": 76337, "tid": -914061504, "ts": 1716454224277577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224329894, "dur": 4,
    "args": {
      "External id": 182024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182024,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 182024, "pid": 5, "tid": 7, "ts": 1716454224329894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277600, "dur": 6,
    "args": {
      "External id": 182024,
      "cbid": 211, "correlation": 182024
    }
  },
  {
    "ph": "s", "id": 182024, "pid": 76337, "tid": -914061504, "ts": 1716454224277600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224329900, "dur": 12,
    "args": {
      "External id": 182027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182027,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182027, "pid": 5, "tid": 7, "ts": 1716454224329900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277618, "dur": 6,
    "args": {
      "External id": 182027,
      "cbid": 211, "correlation": 182027
    }
  },
  {
    "ph": "s", "id": 182027, "pid": 76337, "tid": -914061504, "ts": 1716454224277618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224329912, "dur": 6,
    "args": {
      "External id": 182036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182036,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182036, "pid": 5, "tid": 7, "ts": 1716454224329912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277656, "dur": 9,
    "args": {
      "External id": 182036,
      "cbid": 211, "correlation": 182036
    }
  },
  {
    "ph": "s", "id": 182036, "pid": 76337, "tid": -914061504, "ts": 1716454224277656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277713, "dur": 0,
    "args": {
      "External id": 182046,
      "cbid": 317, "correlation": 182046
    }
  },
  {
    "ph": "f", "id": 182046, "pid": 76337, "tid": -914061504, "ts": 1716454224277713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277714, "dur": 0,
    "args": {
      "External id": 182047,
      "cbid": 203, "correlation": 182047
    }
  },
  {
    "ph": "f", "id": 182047, "pid": 76337, "tid": -914061504, "ts": 1716454224277714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277714, "dur": 0,
    "args": {
      "External id": 182048,
      "cbid": 205, "correlation": 182048
    }
  },
  {
    "ph": "f", "id": 182048, "pid": 76337, "tid": -914061504, "ts": 1716454224277714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329920, "dur": 5,
    "args": {
      "External id": 182052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182052,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182052, "pid": 5, "tid": 7, "ts": 1716454224329920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277730, "dur": 12,
    "args": {
      "External id": 182052,
      "cbid": 211, "correlation": 182052
    }
  },
  {
    "ph": "s", "id": 182052, "pid": 76337, "tid": -914061504, "ts": 1716454224277730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224329927, "dur": 81,
    "args": {
      "External id": 182054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182054,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182054, "pid": 5, "tid": 7, "ts": 1716454224329927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277745, "dur": 5,
    "args": {
      "External id": 182054,
      "cbid": 211, "correlation": 182054
    }
  },
  {
    "ph": "s", "id": 182054, "pid": 76337, "tid": -914061504, "ts": 1716454224277745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224330010, "dur": 1,
    "args": {
      "External id": 182056,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 182056,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6122448979591837
    }
  },
  {
    "ph": "f", "id": 182056, "pid": 5, "tid": 7, "ts": 1716454224330010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277758, "dur": 9,
    "args": {
      "External id": 182056,
      "cbid": 51, "correlation": 182056
    }
  },
  {
    "ph": "s", "id": 182056, "pid": 76337, "tid": -914061504, "ts": 1716454224277758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224330014, "dur": 530,
    "args": {
      "External id": 182057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182057,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182057, "pid": 5, "tid": 7, "ts": 1716454224330014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277769, "dur": 9,
    "args": {
      "External id": 182057,
      "cbid": 211, "correlation": 182057
    }
  },
  {
    "ph": "s", "id": 182057, "pid": 76337, "tid": -914061504, "ts": 1716454224277769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224330545, "dur": 11,
    "args": {
      "External id": 182059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182059,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182059, "pid": 5, "tid": 7, "ts": 1716454224330545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277781, "dur": 5,
    "args": {
      "External id": 182059,
      "cbid": 211, "correlation": 182059
    }
  },
  {
    "ph": "s", "id": 182059, "pid": 76337, "tid": -914061504, "ts": 1716454224277781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224330558, "dur": 14,
    "args": {
      "External id": 182065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182065,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182065, "pid": 5, "tid": 7, "ts": 1716454224330558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277809, "dur": 8,
    "args": {
      "External id": 182065,
      "cbid": 211, "correlation": 182065
    }
  },
  {
    "ph": "s", "id": 182065, "pid": 76337, "tid": -914061504, "ts": 1716454224277809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224330573, "dur": 3,
    "args": {
      "External id": 182073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182073,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 182073, "pid": 5, "tid": 7, "ts": 1716454224330573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277853, "dur": 9,
    "args": {
      "External id": 182073,
      "cbid": 211, "correlation": 182073
    }
  },
  {
    "ph": "s", "id": 182073, "pid": 76337, "tid": -914061504, "ts": 1716454224277853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277917, "dur": 1,
    "args": {
      "External id": 182089,
      "cbid": 251, "correlation": 182089
    }
  },
  {
    "ph": "f", "id": 182089, "pid": 76337, "tid": -914061504, "ts": 1716454224277917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277922, "dur": 0,
    "args": {
      "External id": 182091,
      "cbid": 251, "correlation": 182091
    }
  },
  {
    "ph": "f", "id": 182091, "pid": 76337, "tid": -914061504, "ts": 1716454224277922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224330578, "dur": 13,
    "args": {
      "External id": 182092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182092,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182092, "pid": 5, "tid": 7, "ts": 1716454224330578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277924, "dur": 12,
    "args": {
      "External id": 182092,
      "cbid": 211, "correlation": 182092
    }
  },
  {
    "ph": "s", "id": 182092, "pid": 76337, "tid": -914061504, "ts": 1716454224277924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224330592, "dur": 5,
    "args": {
      "External id": 182094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182094,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182094, "pid": 5, "tid": 7, "ts": 1716454224330592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224277937, "dur": 6,
    "args": {
      "External id": 182094,
      "cbid": 211, "correlation": 182094
    }
  },
  {
    "ph": "s", "id": 182094, "pid": 76337, "tid": -914061504, "ts": 1716454224277937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224330599, "dur": 16,
    "args": {
      "External id": 182104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182104,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182104, "pid": 5, "tid": 7, "ts": 1716454224330599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278005, "dur": 13,
    "args": {
      "External id": 182104,
      "cbid": 211, "correlation": 182104
    }
  },
  {
    "ph": "s", "id": 182104, "pid": 76337, "tid": -914061504, "ts": 1716454224278005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224330616, "dur": 19,
    "args": {
      "External id": 182124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182124,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182124, "pid": 5, "tid": 7, "ts": 1716454224330616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278071, "dur": 12,
    "args": {
      "External id": 182124,
      "cbid": 211, "correlation": 182124
    }
  },
  {
    "ph": "s", "id": 182124, "pid": 76337, "tid": -914061504, "ts": 1716454224278071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224330636, "dur": 5,
    "args": {
      "External id": 182136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182136,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 182136, "pid": 5, "tid": 7, "ts": 1716454224330636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278093, "dur": 6,
    "args": {
      "External id": 182136,
      "cbid": 211, "correlation": 182136
    }
  },
  {
    "ph": "s", "id": 182136, "pid": 76337, "tid": -914061504, "ts": 1716454224278093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224330642, "dur": 17,
    "args": {
      "External id": 182139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182139,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182139, "pid": 5, "tid": 7, "ts": 1716454224330642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278112, "dur": 7,
    "args": {
      "External id": 182139,
      "cbid": 211, "correlation": 182139
    }
  },
  {
    "ph": "s", "id": 182139, "pid": 76337, "tid": -914061504, "ts": 1716454224278112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224330660, "dur": 11,
    "args": {
      "External id": 182148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182148,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182148, "pid": 5, "tid": 7, "ts": 1716454224330660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278153, "dur": 10,
    "args": {
      "External id": 182148,
      "cbid": 211, "correlation": 182148
    }
  },
  {
    "ph": "s", "id": 182148, "pid": 76337, "tid": -914061504, "ts": 1716454224278153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278219, "dur": 0,
    "args": {
      "External id": 182158,
      "cbid": 317, "correlation": 182158
    }
  },
  {
    "ph": "f", "id": 182158, "pid": 76337, "tid": -914061504, "ts": 1716454224278219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278220, "dur": 0,
    "args": {
      "External id": 182159,
      "cbid": 203, "correlation": 182159
    }
  },
  {
    "ph": "f", "id": 182159, "pid": 76337, "tid": -914061504, "ts": 1716454224278220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278221, "dur": 0,
    "args": {
      "External id": 182160,
      "cbid": 205, "correlation": 182160
    }
  },
  {
    "ph": "f", "id": 182160, "pid": 76337, "tid": -914061504, "ts": 1716454224278221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224330672, "dur": 11,
    "args": {
      "External id": 182164,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182164,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182164, "pid": 5, "tid": 7, "ts": 1716454224330672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278239, "dur": 12,
    "args": {
      "External id": 182164,
      "cbid": 211, "correlation": 182164
    }
  },
  {
    "ph": "s", "id": 182164, "pid": 76337, "tid": -914061504, "ts": 1716454224278239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224330685, "dur": 160,
    "args": {
      "External id": 182166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182166,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182166, "pid": 5, "tid": 7, "ts": 1716454224330685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278253, "dur": 5,
    "args": {
      "External id": 182166,
      "cbid": 211, "correlation": 182166
    }
  },
  {
    "ph": "s", "id": 182166, "pid": 76337, "tid": -914061504, "ts": 1716454224278253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224330847, "dur": 1,
    "args": {
      "External id": 182168,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 182168,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 182168, "pid": 5, "tid": 7, "ts": 1716454224330847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278265, "dur": 7,
    "args": {
      "External id": 182168,
      "cbid": 51, "correlation": 182168
    }
  },
  {
    "ph": "s", "id": 182168, "pid": 76337, "tid": -914061504, "ts": 1716454224278265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224330851, "dur": 658,
    "args": {
      "External id": 182169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182169,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182169, "pid": 5, "tid": 7, "ts": 1716454224330851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278273, "dur": 6,
    "args": {
      "External id": 182169,
      "cbid": 211, "correlation": 182169
    }
  },
  {
    "ph": "s", "id": 182169, "pid": 76337, "tid": -914061504, "ts": 1716454224278273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224331510, "dur": 12,
    "args": {
      "External id": 182171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182171,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182171, "pid": 5, "tid": 7, "ts": 1716454224331510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278284, "dur": 6,
    "args": {
      "External id": 182171,
      "cbid": 211, "correlation": 182171
    }
  },
  {
    "ph": "s", "id": 182171, "pid": 76337, "tid": -914061504, "ts": 1716454224278284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224331524, "dur": 14,
    "args": {
      "External id": 182177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182177,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182177, "pid": 5, "tid": 7, "ts": 1716454224331524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278313, "dur": 8,
    "args": {
      "External id": 182177,
      "cbid": 211, "correlation": 182177
    }
  },
  {
    "ph": "s", "id": 182177, "pid": 76337, "tid": -914061504, "ts": 1716454224278313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278371, "dur": 0,
    "args": {
      "External id": 182187,
      "cbid": 317, "correlation": 182187
    }
  },
  {
    "ph": "f", "id": 182187, "pid": 76337, "tid": -914061504, "ts": 1716454224278371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278372, "dur": 0,
    "args": {
      "External id": 182188,
      "cbid": 203, "correlation": 182188
    }
  },
  {
    "ph": "f", "id": 182188, "pid": 76337, "tid": -914061504, "ts": 1716454224278372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278372, "dur": 0,
    "args": {
      "External id": 182189,
      "cbid": 205, "correlation": 182189
    }
  },
  {
    "ph": "f", "id": 182189, "pid": 76337, "tid": -914061504, "ts": 1716454224278372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224331540, "dur": 8,
    "args": {
      "External id": 182193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182193,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182193, "pid": 5, "tid": 7, "ts": 1716454224331540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278388, "dur": 12,
    "args": {
      "External id": 182193,
      "cbid": 211, "correlation": 182193
    }
  },
  {
    "ph": "s", "id": 182193, "pid": 76337, "tid": -914061504, "ts": 1716454224278388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224331549, "dur": 3,
    "args": {
      "External id": 182195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182195,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182195, "pid": 5, "tid": 7, "ts": 1716454224331549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278406, "dur": 6,
    "args": {
      "External id": 182195,
      "cbid": 211, "correlation": 182195
    }
  },
  {
    "ph": "s", "id": 182195, "pid": 76337, "tid": -914061504, "ts": 1716454224278406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278415, "dur": 0,
    "args": {
      "External id": 182196,
      "cbid": 51, "correlation": 182196
    }
  },
  {
    "ph": "s", "id": 182196, "pid": 76337, "tid": -914061504, "ts": 1716454224278415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224331554, "dur": 56,
    "args": {
      "External id": 182197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182197,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 182197, "pid": 5, "tid": 7, "ts": 1716454224331554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278416, "dur": 5,
    "args": {
      "External id": 182197,
      "cbid": 211, "correlation": 182197
    }
  },
  {
    "ph": "s", "id": 182197, "pid": 76337, "tid": -914061504, "ts": 1716454224278416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224331611, "dur": 14,
    "args": {
      "External id": 182202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182202,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182202, "pid": 5, "tid": 7, "ts": 1716454224331611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278442, "dur": 8,
    "args": {
      "External id": 182202,
      "cbid": 211, "correlation": 182202
    }
  },
  {
    "ph": "s", "id": 182202, "pid": 76337, "tid": -914061504, "ts": 1716454224278442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224331626, "dur": 12,
    "args": {
      "External id": 182210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182210,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182210, "pid": 5, "tid": 7, "ts": 1716454224331626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278470, "dur": 8,
    "args": {
      "External id": 182210,
      "cbid": 211, "correlation": 182210
    }
  },
  {
    "ph": "s", "id": 182210, "pid": 76337, "tid": -914061504, "ts": 1716454224278470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224331639, "dur": 11,
    "args": {
      "External id": 182218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182218,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182218, "pid": 5, "tid": 7, "ts": 1716454224331639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278500, "dur": 8,
    "args": {
      "External id": 182218,
      "cbid": 211, "correlation": 182218
    }
  },
  {
    "ph": "s", "id": 182218, "pid": 76337, "tid": -914061504, "ts": 1716454224278500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224331651, "dur": 19,
    "args": {
      "External id": 182238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182238,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182238, "pid": 5, "tid": 7, "ts": 1716454224331651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278580, "dur": 13,
    "args": {
      "External id": 182238,
      "cbid": 211, "correlation": 182238
    }
  },
  {
    "ph": "s", "id": 182238, "pid": 76337, "tid": -914061504, "ts": 1716454224278580,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224331671, "dur": 4,
    "args": {
      "External id": 182250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182250,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 182250, "pid": 5, "tid": 7, "ts": 1716454224331671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278603, "dur": 6,
    "args": {
      "External id": 182250,
      "cbid": 211, "correlation": 182250
    }
  },
  {
    "ph": "s", "id": 182250, "pid": 76337, "tid": -914061504, "ts": 1716454224278603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224331677, "dur": 16,
    "args": {
      "External id": 182253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182253, "pid": 5, "tid": 7, "ts": 1716454224331677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278620, "dur": 7,
    "args": {
      "External id": 182253,
      "cbid": 211, "correlation": 182253
    }
  },
  {
    "ph": "s", "id": 182253, "pid": 76337, "tid": -914061504, "ts": 1716454224278620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278680, "dur": 0,
    "args": {
      "External id": 182264,
      "cbid": 317, "correlation": 182264
    }
  },
  {
    "ph": "f", "id": 182264, "pid": 76337, "tid": -914061504, "ts": 1716454224278680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278681, "dur": 0,
    "args": {
      "External id": 182265,
      "cbid": 203, "correlation": 182265
    }
  },
  {
    "ph": "f", "id": 182265, "pid": 76337, "tid": -914061504, "ts": 1716454224278681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278682, "dur": 0,
    "args": {
      "External id": 182266,
      "cbid": 205, "correlation": 182266
    }
  },
  {
    "ph": "f", "id": 182266, "pid": 76337, "tid": -914061504, "ts": 1716454224278682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224331694, "dur": 11,
    "args": {
      "External id": 182270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182270, "pid": 5, "tid": 7, "ts": 1716454224331694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278698, "dur": 12,
    "args": {
      "External id": 182270,
      "cbid": 211, "correlation": 182270
    }
  },
  {
    "ph": "s", "id": 182270, "pid": 76337, "tid": -914061504, "ts": 1716454224278698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224331707, "dur": 3,
    "args": {
      "External id": 182272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182272,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182272, "pid": 5, "tid": 7, "ts": 1716454224331707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278714, "dur": 6,
    "args": {
      "External id": 182272,
      "cbid": 211, "correlation": 182272
    }
  },
  {
    "ph": "s", "id": 182272, "pid": 76337, "tid": -914061504, "ts": 1716454224278714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278723, "dur": 0,
    "args": {
      "External id": 182273,
      "cbid": 51, "correlation": 182273
    }
  },
  {
    "ph": "s", "id": 182273, "pid": 76337, "tid": -914061504, "ts": 1716454224278723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224331711, "dur": 95,
    "args": {
      "External id": 182274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182274,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 182274, "pid": 5, "tid": 7, "ts": 1716454224331711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278724, "dur": 6,
    "args": {
      "External id": 182274,
      "cbid": 211, "correlation": 182274
    }
  },
  {
    "ph": "s", "id": 182274, "pid": 76337, "tid": -914061504, "ts": 1716454224278724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224331808, "dur": 15,
    "args": {
      "External id": 182279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182279, "pid": 5, "tid": 7, "ts": 1716454224331808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278753, "dur": 8,
    "args": {
      "External id": 182279,
      "cbid": 211, "correlation": 182279
    }
  },
  {
    "ph": "s", "id": 182279, "pid": 76337, "tid": -914061504, "ts": 1716454224278753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224331824, "dur": 81,
    "args": {
      "External id": 182288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182288,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182288, "pid": 5, "tid": 7, "ts": 1716454224331824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278835, "dur": 14,
    "args": {
      "External id": 182288,
      "cbid": 211, "correlation": 182288
    }
  },
  {
    "ph": "s", "id": 182288, "pid": 76337, "tid": -914061504, "ts": 1716454224278835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224331906, "dur": 30,
    "args": {
      "External id": 182310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182310,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182310, "pid": 5, "tid": 7, "ts": 1716454224331906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224278896, "dur": 11,
    "args": {
      "External id": 182310,
      "cbid": 211, "correlation": 182310
    }
  },
  {
    "ph": "s", "id": 182310, "pid": 76337, "tid": -914061504, "ts": 1716454224278896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279002, "dur": 2,
    "args": {
      "External id": 182321,
      "cbid": 251, "correlation": 182321
    }
  },
  {
    "ph": "f", "id": 182321, "pid": 76337, "tid": -914061504, "ts": 1716454224279002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224331937, "dur": 163,
    "args": {
      "External id": 182322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182322,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182322, "pid": 5, "tid": 7, "ts": 1716454224331937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279008, "dur": 13,
    "args": {
      "External id": 182322,
      "cbid": 211, "correlation": 182322
    }
  },
  {
    "ph": "s", "id": 182322, "pid": 76337, "tid": -914061504, "ts": 1716454224279008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279081, "dur": 1,
    "args": {
      "External id": 182333,
      "cbid": 251, "correlation": 182333
    }
  },
  {
    "ph": "f", "id": 182333, "pid": 76337, "tid": -914061504, "ts": 1716454224279081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224332101, "dur": 135,
    "args": {
      "External id": 182334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182334,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182334, "pid": 5, "tid": 7, "ts": 1716454224332101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279085, "dur": 11,
    "args": {
      "External id": 182334,
      "cbid": 211, "correlation": 182334
    }
  },
  {
    "ph": "s", "id": 182334, "pid": 76337, "tid": -914061504, "ts": 1716454224279085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279151, "dur": 1,
    "args": {
      "External id": 182345,
      "cbid": 251, "correlation": 182345
    }
  },
  {
    "ph": "f", "id": 182345, "pid": 76337, "tid": -914061504, "ts": 1716454224279151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224332238, "dur": 154,
    "args": {
      "External id": 182346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182346,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182346, "pid": 5, "tid": 7, "ts": 1716454224332238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279156, "dur": 11,
    "args": {
      "External id": 182346,
      "cbid": 211, "correlation": 182346
    }
  },
  {
    "ph": "s", "id": 182346, "pid": 76337, "tid": -914061504, "ts": 1716454224279156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224332393, "dur": 331,
    "args": {
      "External id": 182371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182371,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182371, "pid": 5, "tid": 7, "ts": 1716454224332393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279242, "dur": 14,
    "args": {
      "External id": 182371,
      "cbid": 211, "correlation": 182371
    }
  },
  {
    "ph": "s", "id": 182371, "pid": 76337, "tid": -914061504, "ts": 1716454224279242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279347, "dur": 1,
    "args": {
      "External id": 182389,
      "cbid": 251, "correlation": 182389
    }
  },
  {
    "ph": "f", "id": 182389, "pid": 76337, "tid": -914061504, "ts": 1716454224279347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224332725, "dur": 164,
    "args": {
      "External id": 182391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182391,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182391, "pid": 5, "tid": 7, "ts": 1716454224332725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279353, "dur": 14,
    "args": {
      "External id": 182391,
      "cbid": 211, "correlation": 182391
    }
  },
  {
    "ph": "s", "id": 182391, "pid": 76337, "tid": -914061504, "ts": 1716454224279353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224332891, "dur": 19,
    "args": {
      "External id": 182399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182399,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182399, "pid": 5, "tid": 7, "ts": 1716454224332891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279425, "dur": 12,
    "args": {
      "External id": 182399,
      "cbid": 211, "correlation": 182399
    }
  },
  {
    "ph": "s", "id": 182399, "pid": 76337, "tid": -914061504, "ts": 1716454224279425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224332912, "dur": 27,
    "args": {
      "External id": 182407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182407,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182407, "pid": 5, "tid": 7, "ts": 1716454224332912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279465, "dur": 9,
    "args": {
      "External id": 182407,
      "cbid": 211, "correlation": 182407
    }
  },
  {
    "ph": "s", "id": 182407, "pid": 76337, "tid": -914061504, "ts": 1716454224279465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224332940, "dur": 18,
    "args": {
      "External id": 182418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182418, "pid": 5, "tid": 7, "ts": 1716454224332940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279537, "dur": 12,
    "args": {
      "External id": 182418,
      "cbid": 211, "correlation": 182418
    }
  },
  {
    "ph": "s", "id": 182418, "pid": 76337, "tid": -914061504, "ts": 1716454224279537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224332960, "dur": 16,
    "args": {
      "External id": 182440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182440,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182440, "pid": 5, "tid": 7, "ts": 1716454224332960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279569, "dur": 8,
    "args": {
      "External id": 182440,
      "cbid": 211, "correlation": 182440
    }
  },
  {
    "ph": "s", "id": 182440, "pid": 76337, "tid": -914061504, "ts": 1716454224279569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279653, "dur": 1,
    "args": {
      "External id": 182451,
      "cbid": 251, "correlation": 182451
    }
  },
  {
    "ph": "f", "id": 182451, "pid": 76337, "tid": -914061504, "ts": 1716454224279653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224332977, "dur": 89,
    "args": {
      "External id": 182452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182452,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182452, "pid": 5, "tid": 7, "ts": 1716454224332977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279660, "dur": 16,
    "args": {
      "External id": 182452,
      "cbid": 211, "correlation": 182452
    }
  },
  {
    "ph": "s", "id": 182452, "pid": 76337, "tid": -914061504, "ts": 1716454224279660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279732, "dur": 1,
    "args": {
      "External id": 182463,
      "cbid": 251, "correlation": 182463
    }
  },
  {
    "ph": "f", "id": 182463, "pid": 76337, "tid": -914061504, "ts": 1716454224279732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279736, "dur": 0,
    "args": {
      "External id": 182464,
      "cbid": 251, "correlation": 182464
    }
  },
  {
    "ph": "f", "id": 182464, "pid": 76337, "tid": -914061504, "ts": 1716454224279736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224333068, "dur": 12,
    "args": {
      "External id": 182465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182465,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182465, "pid": 5, "tid": 7, "ts": 1716454224333068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279738, "dur": 12,
    "args": {
      "External id": 182465,
      "cbid": 211, "correlation": 182465
    }
  },
  {
    "ph": "s", "id": 182465, "pid": 76337, "tid": -914061504, "ts": 1716454224279738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224333081, "dur": 6,
    "args": {
      "External id": 182467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182467,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182467, "pid": 5, "tid": 7, "ts": 1716454224333081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279751, "dur": 6,
    "args": {
      "External id": 182467,
      "cbid": 211, "correlation": 182467
    }
  },
  {
    "ph": "s", "id": 182467, "pid": 76337, "tid": -914061504, "ts": 1716454224279751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279809, "dur": 1,
    "args": {
      "External id": 182478,
      "cbid": 251, "correlation": 182478
    }
  },
  {
    "ph": "f", "id": 182478, "pid": 76337, "tid": -914061504, "ts": 1716454224279809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279812, "dur": 0,
    "args": {
      "External id": 182479,
      "cbid": 251, "correlation": 182479
    }
  },
  {
    "ph": "f", "id": 182479, "pid": 76337, "tid": -914061504, "ts": 1716454224279812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224333088, "dur": 8,
    "args": {
      "External id": 182480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182480,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182480, "pid": 5, "tid": 7, "ts": 1716454224333088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279814, "dur": 12,
    "args": {
      "External id": 182480,
      "cbid": 211, "correlation": 182480
    }
  },
  {
    "ph": "s", "id": 182480, "pid": 76337, "tid": -914061504, "ts": 1716454224279814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224333097, "dur": 3,
    "args": {
      "External id": 182482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182482,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182482, "pid": 5, "tid": 7, "ts": 1716454224333097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279827, "dur": 5,
    "args": {
      "External id": 182482,
      "cbid": 211, "correlation": 182482
    }
  },
  {
    "ph": "s", "id": 182482, "pid": 76337, "tid": -914061504, "ts": 1716454224279827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224333101, "dur": 54,
    "args": {
      "External id": 182507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182507,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182507, "pid": 5, "tid": 7, "ts": 1716454224333101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224279903, "dur": 12,
    "args": {
      "External id": 182507,
      "cbid": 211, "correlation": 182507
    }
  },
  {
    "ph": "s", "id": 182507, "pid": 76337, "tid": -914061504, "ts": 1716454224279903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280015, "dur": 2,
    "args": {
      "External id": 182525,
      "cbid": 251, "correlation": 182525
    }
  },
  {
    "ph": "f", "id": 182525, "pid": 76337, "tid": -914061504, "ts": 1716454224280015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224333157, "dur": 90,
    "args": {
      "External id": 182527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182527,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182527, "pid": 5, "tid": 7, "ts": 1716454224333157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280021, "dur": 15,
    "args": {
      "External id": 182527,
      "cbid": 211, "correlation": 182527
    }
  },
  {
    "ph": "s", "id": 182527, "pid": 76337, "tid": -914061504, "ts": 1716454224280021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224333248, "dur": 9,
    "args": {
      "External id": 182535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182535,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182535, "pid": 5, "tid": 7, "ts": 1716454224333248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280094, "dur": 12,
    "args": {
      "External id": 182535,
      "cbid": 211, "correlation": 182535
    }
  },
  {
    "ph": "s", "id": 182535, "pid": 76337, "tid": -914061504, "ts": 1716454224280094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224333259, "dur": 21,
    "args": {
      "External id": 182543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182543,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182543, "pid": 5, "tid": 7, "ts": 1716454224333259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280135, "dur": 9,
    "args": {
      "External id": 182543,
      "cbid": 211, "correlation": 182543
    }
  },
  {
    "ph": "s", "id": 182543, "pid": 76337, "tid": -914061504, "ts": 1716454224280135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224333281, "dur": 18,
    "args": {
      "External id": 182565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182565,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182565, "pid": 5, "tid": 7, "ts": 1716454224333281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280187, "dur": 9,
    "args": {
      "External id": 182565,
      "cbid": 211, "correlation": 182565
    }
  },
  {
    "ph": "s", "id": 182565, "pid": 76337, "tid": -914061504, "ts": 1716454224280187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280276, "dur": 2,
    "args": {
      "External id": 182581,
      "cbid": 251, "correlation": 182581
    }
  },
  {
    "ph": "f", "id": 182581, "pid": 76337, "tid": -914061504, "ts": 1716454224280276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280283, "dur": 0,
    "args": {
      "External id": 182583,
      "cbid": 251, "correlation": 182583
    }
  },
  {
    "ph": "f", "id": 182583, "pid": 76337, "tid": -914061504, "ts": 1716454224280283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224333300, "dur": 494,
    "args": {
      "External id": 182584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182584,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182584, "pid": 5, "tid": 7, "ts": 1716454224333300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280285, "dur": 15,
    "args": {
      "External id": 182584,
      "cbid": 211, "correlation": 182584
    }
  },
  {
    "ph": "s", "id": 182584, "pid": 76337, "tid": -914061504, "ts": 1716454224280285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224333795, "dur": 65,
    "args": {
      "External id": 182592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182592, "pid": 5, "tid": 7, "ts": 1716454224333795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280352, "dur": 12,
    "args": {
      "External id": 182592,
      "cbid": 211, "correlation": 182592
    }
  },
  {
    "ph": "s", "id": 182592, "pid": 76337, "tid": -914061504, "ts": 1716454224280352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224333862, "dur": 68,
    "args": {
      "External id": 182600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182600, "pid": 5, "tid": 7, "ts": 1716454224333862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280384, "dur": 8,
    "args": {
      "External id": 182600,
      "cbid": 211, "correlation": 182600
    }
  },
  {
    "ph": "s", "id": 182600, "pid": 76337, "tid": -914061504, "ts": 1716454224280384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280464, "dur": 1,
    "args": {
      "External id": 182616,
      "cbid": 251, "correlation": 182616
    }
  },
  {
    "ph": "f", "id": 182616, "pid": 76337, "tid": -914061504, "ts": 1716454224280464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224333932, "dur": 1,
    "args": {
      "External id": 182618,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 182618,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 182618, "pid": 5, "tid": 7, "ts": 1716454224333932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280469, "dur": 12,
    "args": {
      "External id": 182618,
      "cbid": 51, "correlation": 182618
    }
  },
  {
    "ph": "s", "id": 182618, "pid": 76337, "tid": -914061504, "ts": 1716454224280469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224333936, "dur": 266,
    "args": {
      "External id": 182619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182619,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182619, "pid": 5, "tid": 7, "ts": 1716454224333936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280484, "dur": 11,
    "args": {
      "External id": 182619,
      "cbid": 211, "correlation": 182619
    }
  },
  {
    "ph": "s", "id": 182619, "pid": 76337, "tid": -914061504, "ts": 1716454224280484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224334204, "dur": 14,
    "args": {
      "External id": 182627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182627,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182627, "pid": 5, "tid": 7, "ts": 1716454224334204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280525, "dur": 11,
    "args": {
      "External id": 182627,
      "cbid": 211, "correlation": 182627
    }
  },
  {
    "ph": "s", "id": 182627, "pid": 76337, "tid": -914061504, "ts": 1716454224280525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224334219, "dur": 37,
    "args": {
      "External id": 182638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182638,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182638, "pid": 5, "tid": 7, "ts": 1716454224334219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280593, "dur": 12,
    "args": {
      "External id": 182638,
      "cbid": 211, "correlation": 182638
    }
  },
  {
    "ph": "s", "id": 182638, "pid": 76337, "tid": -914061504, "ts": 1716454224280593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280662, "dur": 0,
    "args": {
      "External id": 182650,
      "cbid": 317, "correlation": 182650
    }
  },
  {
    "ph": "f", "id": 182650, "pid": 76337, "tid": -914061504, "ts": 1716454224280662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280662, "dur": 0,
    "args": {
      "External id": 182651,
      "cbid": 203, "correlation": 182651
    }
  },
  {
    "ph": "f", "id": 182651, "pid": 76337, "tid": -914061504, "ts": 1716454224280662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280663, "dur": 0,
    "args": {
      "External id": 182652,
      "cbid": 205, "correlation": 182652
    }
  },
  {
    "ph": "f", "id": 182652, "pid": 76337, "tid": -914061504, "ts": 1716454224280663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224334258, "dur": 13,
    "args": {
      "External id": 182656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182656,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182656, "pid": 5, "tid": 7, "ts": 1716454224334258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280679, "dur": 13,
    "args": {
      "External id": 182656,
      "cbid": 211, "correlation": 182656
    }
  },
  {
    "ph": "s", "id": 182656, "pid": 76337, "tid": -914061504, "ts": 1716454224280679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224334272, "dur": 4,
    "args": {
      "External id": 182658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182658,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182658, "pid": 5, "tid": 7, "ts": 1716454224334272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280696, "dur": 6,
    "args": {
      "External id": 182658,
      "cbid": 211, "correlation": 182658
    }
  },
  {
    "ph": "s", "id": 182658, "pid": 76337, "tid": -914061504, "ts": 1716454224280696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280705, "dur": 0,
    "args": {
      "External id": 182659,
      "cbid": 51, "correlation": 182659
    }
  },
  {
    "ph": "s", "id": 182659, "pid": 76337, "tid": -914061504, "ts": 1716454224280705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224334277, "dur": 94,
    "args": {
      "External id": 182660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182660,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 182660, "pid": 5, "tid": 7, "ts": 1716454224334277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280706, "dur": 5,
    "args": {
      "External id": 182660,
      "cbid": 211, "correlation": 182660
    }
  },
  {
    "ph": "s", "id": 182660, "pid": 76337, "tid": -914061504, "ts": 1716454224280706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224334373, "dur": 16,
    "args": {
      "External id": 182665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182665,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182665, "pid": 5, "tid": 7, "ts": 1716454224334373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280733, "dur": 10,
    "args": {
      "External id": 182665,
      "cbid": 211, "correlation": 182665
    }
  },
  {
    "ph": "s", "id": 182665, "pid": 76337, "tid": -914061504, "ts": 1716454224280733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224334390, "dur": 11,
    "args": {
      "External id": 182673,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182673,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182673, "pid": 5, "tid": 7, "ts": 1716454224334390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280766, "dur": 8,
    "args": {
      "External id": 182673,
      "cbid": 211, "correlation": 182673
    }
  },
  {
    "ph": "s", "id": 182673, "pid": 76337, "tid": -914061504, "ts": 1716454224280766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224334403, "dur": 17,
    "args": {
      "External id": 182693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182693,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182693, "pid": 5, "tid": 7, "ts": 1716454224334403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280838, "dur": 12,
    "args": {
      "External id": 182693,
      "cbid": 211, "correlation": 182693
    }
  },
  {
    "ph": "s", "id": 182693, "pid": 76337, "tid": -914061504, "ts": 1716454224280838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224334421, "dur": 5,
    "args": {
      "External id": 182705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182705,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 182705, "pid": 5, "tid": 7, "ts": 1716454224334421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280861, "dur": 6,
    "args": {
      "External id": 182705,
      "cbid": 211, "correlation": 182705
    }
  },
  {
    "ph": "s", "id": 182705, "pid": 76337, "tid": -914061504, "ts": 1716454224280861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224334427, "dur": 18,
    "args": {
      "External id": 182708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182708, "pid": 5, "tid": 7, "ts": 1716454224334427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280879, "dur": 7,
    "args": {
      "External id": 182708,
      "cbid": 211, "correlation": 182708
    }
  },
  {
    "ph": "s", "id": 182708, "pid": 76337, "tid": -914061504, "ts": 1716454224280879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224334447, "dur": 12,
    "args": {
      "External id": 182717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182717,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182717, "pid": 5, "tid": 7, "ts": 1716454224334447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280918, "dur": 9,
    "args": {
      "External id": 182717,
      "cbid": 211, "correlation": 182717
    }
  },
  {
    "ph": "s", "id": 182717, "pid": 76337, "tid": -914061504, "ts": 1716454224280918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280969, "dur": 0,
    "args": {
      "External id": 182727,
      "cbid": 317, "correlation": 182727
    }
  },
  {
    "ph": "f", "id": 182727, "pid": 76337, "tid": -914061504, "ts": 1716454224280969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280970, "dur": 0,
    "args": {
      "External id": 182728,
      "cbid": 203, "correlation": 182728
    }
  },
  {
    "ph": "f", "id": 182728, "pid": 76337, "tid": -914061504, "ts": 1716454224280970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280970, "dur": 0,
    "args": {
      "External id": 182729,
      "cbid": 205, "correlation": 182729
    }
  },
  {
    "ph": "f", "id": 182729, "pid": 76337, "tid": -914061504, "ts": 1716454224280970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224334460, "dur": 11,
    "args": {
      "External id": 182733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182733,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182733, "pid": 5, "tid": 7, "ts": 1716454224334460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224280993, "dur": 12,
    "args": {
      "External id": 182733,
      "cbid": 211, "correlation": 182733
    }
  },
  {
    "ph": "s", "id": 182733, "pid": 76337, "tid": -914061504, "ts": 1716454224280993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224334472, "dur": 160,
    "args": {
      "External id": 182735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182735,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182735, "pid": 5, "tid": 7, "ts": 1716454224334472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281008, "dur": 5,
    "args": {
      "External id": 182735,
      "cbid": 211, "correlation": 182735
    }
  },
  {
    "ph": "s", "id": 182735, "pid": 76337, "tid": -914061504, "ts": 1716454224281008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224334634, "dur": 1,
    "args": {
      "External id": 182737,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 182737,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 182737, "pid": 5, "tid": 7, "ts": 1716454224334634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281019, "dur": 6,
    "args": {
      "External id": 182737,
      "cbid": 51, "correlation": 182737
    }
  },
  {
    "ph": "s", "id": 182737, "pid": 76337, "tid": -914061504, "ts": 1716454224281019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224334638, "dur": 657,
    "args": {
      "External id": 182738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182738,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182738, "pid": 5, "tid": 7, "ts": 1716454224334638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281027, "dur": 6,
    "args": {
      "External id": 182738,
      "cbid": 211, "correlation": 182738
    }
  },
  {
    "ph": "s", "id": 182738, "pid": 76337, "tid": -914061504, "ts": 1716454224281027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224335296, "dur": 13,
    "args": {
      "External id": 182740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182740,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182740, "pid": 5, "tid": 7, "ts": 1716454224335296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281037, "dur": 5,
    "args": {
      "External id": 182740,
      "cbid": 211, "correlation": 182740
    }
  },
  {
    "ph": "s", "id": 182740, "pid": 76337, "tid": -914061504, "ts": 1716454224281037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224335311, "dur": 14,
    "args": {
      "External id": 182746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182746,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182746, "pid": 5, "tid": 7, "ts": 1716454224335311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281065, "dur": 8,
    "args": {
      "External id": 182746,
      "cbid": 211, "correlation": 182746
    }
  },
  {
    "ph": "s", "id": 182746, "pid": 76337, "tid": -914061504, "ts": 1716454224281065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224335326, "dur": 3,
    "args": {
      "External id": 182754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182754,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 182754, "pid": 5, "tid": 7, "ts": 1716454224335326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281108, "dur": 9,
    "args": {
      "External id": 182754,
      "cbid": 211, "correlation": 182754
    }
  },
  {
    "ph": "s", "id": 182754, "pid": 76337, "tid": -914061504, "ts": 1716454224281108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281172, "dur": 1,
    "args": {
      "External id": 182770,
      "cbid": 251, "correlation": 182770
    }
  },
  {
    "ph": "f", "id": 182770, "pid": 76337, "tid": -914061504, "ts": 1716454224281172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281177, "dur": 0,
    "args": {
      "External id": 182772,
      "cbid": 251, "correlation": 182772
    }
  },
  {
    "ph": "f", "id": 182772, "pid": 76337, "tid": -914061504, "ts": 1716454224281177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224335331, "dur": 13,
    "args": {
      "External id": 182773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182773,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182773, "pid": 5, "tid": 7, "ts": 1716454224335331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281179, "dur": 11,
    "args": {
      "External id": 182773,
      "cbid": 211, "correlation": 182773
    }
  },
  {
    "ph": "s", "id": 182773, "pid": 76337, "tid": -914061504, "ts": 1716454224281179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224335345, "dur": 5,
    "args": {
      "External id": 182775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182775,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182775, "pid": 5, "tid": 7, "ts": 1716454224335345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281192, "dur": 5,
    "args": {
      "External id": 182775,
      "cbid": 211, "correlation": 182775
    }
  },
  {
    "ph": "s", "id": 182775, "pid": 76337, "tid": -914061504, "ts": 1716454224281192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224335351, "dur": 16,
    "args": {
      "External id": 182785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182785, "pid": 5, "tid": 7, "ts": 1716454224335351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281250, "dur": 12,
    "args": {
      "External id": 182785,
      "cbid": 211, "correlation": 182785
    }
  },
  {
    "ph": "s", "id": 182785, "pid": 76337, "tid": -914061504, "ts": 1716454224281250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224335369, "dur": 18,
    "args": {
      "External id": 182805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182805,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182805, "pid": 5, "tid": 7, "ts": 1716454224335369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281316, "dur": 10,
    "args": {
      "External id": 182805,
      "cbid": 211, "correlation": 182805
    }
  },
  {
    "ph": "s", "id": 182805, "pid": 76337, "tid": -914061504, "ts": 1716454224281316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224335389, "dur": 4,
    "args": {
      "External id": 182817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182817,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 182817, "pid": 5, "tid": 7, "ts": 1716454224335389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281336, "dur": 6,
    "args": {
      "External id": 182817,
      "cbid": 211, "correlation": 182817
    }
  },
  {
    "ph": "s", "id": 182817, "pid": 76337, "tid": -914061504, "ts": 1716454224281336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224335394, "dur": 17,
    "args": {
      "External id": 182820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182820, "pid": 5, "tid": 7, "ts": 1716454224335394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281357, "dur": 6,
    "args": {
      "External id": 182820,
      "cbid": 211, "correlation": 182820
    }
  },
  {
    "ph": "s", "id": 182820, "pid": 76337, "tid": -914061504, "ts": 1716454224281357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224335412, "dur": 11,
    "args": {
      "External id": 182829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182829,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182829, "pid": 5, "tid": 7, "ts": 1716454224335412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281396, "dur": 10,
    "args": {
      "External id": 182829,
      "cbid": 211, "correlation": 182829
    }
  },
  {
    "ph": "s", "id": 182829, "pid": 76337, "tid": -914061504, "ts": 1716454224281396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281458, "dur": 0,
    "args": {
      "External id": 182839,
      "cbid": 317, "correlation": 182839
    }
  },
  {
    "ph": "f", "id": 182839, "pid": 76337, "tid": -914061504, "ts": 1716454224281458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281459, "dur": 0,
    "args": {
      "External id": 182840,
      "cbid": 203, "correlation": 182840
    }
  },
  {
    "ph": "f", "id": 182840, "pid": 76337, "tid": -914061504, "ts": 1716454224281459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281459, "dur": 0,
    "args": {
      "External id": 182841,
      "cbid": 205, "correlation": 182841
    }
  },
  {
    "ph": "f", "id": 182841, "pid": 76337, "tid": -914061504, "ts": 1716454224281459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224335425, "dur": 11,
    "args": {
      "External id": 182845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182845,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182845, "pid": 5, "tid": 7, "ts": 1716454224335425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281473, "dur": 12,
    "args": {
      "External id": 182845,
      "cbid": 211, "correlation": 182845
    }
  },
  {
    "ph": "s", "id": 182845, "pid": 76337, "tid": -914061504, "ts": 1716454224281473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224335437, "dur": 160,
    "args": {
      "External id": 182847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182847,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182847, "pid": 5, "tid": 7, "ts": 1716454224335437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281487, "dur": 5,
    "args": {
      "External id": 182847,
      "cbid": 211, "correlation": 182847
    }
  },
  {
    "ph": "s", "id": 182847, "pid": 76337, "tid": -914061504, "ts": 1716454224281487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224335599, "dur": 1,
    "args": {
      "External id": 182849,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 182849,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 182849, "pid": 5, "tid": 7, "ts": 1716454224335599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281498, "dur": 8,
    "args": {
      "External id": 182849,
      "cbid": 51, "correlation": 182849
    }
  },
  {
    "ph": "s", "id": 182849, "pid": 76337, "tid": -914061504, "ts": 1716454224281498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224335603, "dur": 640,
    "args": {
      "External id": 182850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182850,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182850, "pid": 5, "tid": 7, "ts": 1716454224335603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281507, "dur": 6,
    "args": {
      "External id": 182850,
      "cbid": 211, "correlation": 182850
    }
  },
  {
    "ph": "s", "id": 182850, "pid": 76337, "tid": -914061504, "ts": 1716454224281507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224336245, "dur": 12,
    "args": {
      "External id": 182852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182852,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182852, "pid": 5, "tid": 7, "ts": 1716454224336245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281517, "dur": 5,
    "args": {
      "External id": 182852,
      "cbid": 211, "correlation": 182852
    }
  },
  {
    "ph": "s", "id": 182852, "pid": 76337, "tid": -914061504, "ts": 1716454224281517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224336258, "dur": 14,
    "args": {
      "External id": 182858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182858, "pid": 5, "tid": 7, "ts": 1716454224336258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281544, "dur": 9,
    "args": {
      "External id": 182858,
      "cbid": 211, "correlation": 182858
    }
  },
  {
    "ph": "s", "id": 182858, "pid": 76337, "tid": -914061504, "ts": 1716454224281544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224336273, "dur": 12,
    "args": {
      "External id": 182866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182866,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182866, "pid": 5, "tid": 7, "ts": 1716454224336273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281577, "dur": 8,
    "args": {
      "External id": 182866,
      "cbid": 211, "correlation": 182866
    }
  },
  {
    "ph": "s", "id": 182866, "pid": 76337, "tid": -914061504, "ts": 1716454224281577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224336286, "dur": 11,
    "args": {
      "External id": 182874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182874,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182874, "pid": 5, "tid": 7, "ts": 1716454224336286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281605, "dur": 8,
    "args": {
      "External id": 182874,
      "cbid": 211, "correlation": 182874
    }
  },
  {
    "ph": "s", "id": 182874, "pid": 76337, "tid": -914061504, "ts": 1716454224281605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224336299, "dur": 18,
    "args": {
      "External id": 182894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182894,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 182894, "pid": 5, "tid": 7, "ts": 1716454224336299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281682, "dur": 12,
    "args": {
      "External id": 182894,
      "cbid": 211, "correlation": 182894
    }
  },
  {
    "ph": "s", "id": 182894, "pid": 76337, "tid": -914061504, "ts": 1716454224281682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224336318, "dur": 4,
    "args": {
      "External id": 182906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182906,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 182906, "pid": 5, "tid": 7, "ts": 1716454224336318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281704, "dur": 7,
    "args": {
      "External id": 182906,
      "cbid": 211, "correlation": 182906
    }
  },
  {
    "ph": "s", "id": 182906, "pid": 76337, "tid": -914061504, "ts": 1716454224281704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224336323, "dur": 16,
    "args": {
      "External id": 182909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182909, "pid": 5, "tid": 7, "ts": 1716454224336323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281723, "dur": 6,
    "args": {
      "External id": 182909,
      "cbid": 211, "correlation": 182909
    }
  },
  {
    "ph": "s", "id": 182909, "pid": 76337, "tid": -914061504, "ts": 1716454224281723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281779, "dur": 0,
    "args": {
      "External id": 182920,
      "cbid": 317, "correlation": 182920
    }
  },
  {
    "ph": "f", "id": 182920, "pid": 76337, "tid": -914061504, "ts": 1716454224281779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281780, "dur": 0,
    "args": {
      "External id": 182921,
      "cbid": 203, "correlation": 182921
    }
  },
  {
    "ph": "f", "id": 182921, "pid": 76337, "tid": -914061504, "ts": 1716454224281780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281781, "dur": 0,
    "args": {
      "External id": 182922,
      "cbid": 205, "correlation": 182922
    }
  },
  {
    "ph": "f", "id": 182922, "pid": 76337, "tid": -914061504, "ts": 1716454224281781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224336340, "dur": 11,
    "args": {
      "External id": 182926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182926,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182926, "pid": 5, "tid": 7, "ts": 1716454224336340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281795, "dur": 12,
    "args": {
      "External id": 182926,
      "cbid": 211, "correlation": 182926
    }
  },
  {
    "ph": "s", "id": 182926, "pid": 76337, "tid": -914061504, "ts": 1716454224281795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224336353, "dur": 4,
    "args": {
      "External id": 182928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 182928, "pid": 5, "tid": 7, "ts": 1716454224336353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281812, "dur": 6,
    "args": {
      "External id": 182928,
      "cbid": 211, "correlation": 182928
    }
  },
  {
    "ph": "s", "id": 182928, "pid": 76337, "tid": -914061504, "ts": 1716454224281812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281820, "dur": 0,
    "args": {
      "External id": 182929,
      "cbid": 51, "correlation": 182929
    }
  },
  {
    "ph": "s", "id": 182929, "pid": 76337, "tid": -914061504, "ts": 1716454224281820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224336358, "dur": 93,
    "args": {
      "External id": 182930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182930,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 182930, "pid": 5, "tid": 7, "ts": 1716454224336358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281821, "dur": 5,
    "args": {
      "External id": 182930,
      "cbid": 211, "correlation": 182930
    }
  },
  {
    "ph": "s", "id": 182930, "pid": 76337, "tid": -914061504, "ts": 1716454224281821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224336452, "dur": 16,
    "args": {
      "External id": 182935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182935, "pid": 5, "tid": 7, "ts": 1716454224336452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281848, "dur": 8,
    "args": {
      "External id": 182935,
      "cbid": 211, "correlation": 182935
    }
  },
  {
    "ph": "s", "id": 182935, "pid": 76337, "tid": -914061504, "ts": 1716454224281848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224336469, "dur": 83,
    "args": {
      "External id": 182944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182944,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182944, "pid": 5, "tid": 7, "ts": 1716454224336469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224281930, "dur": 14,
    "args": {
      "External id": 182944,
      "cbid": 211, "correlation": 182944
    }
  },
  {
    "ph": "s", "id": 182944, "pid": 76337, "tid": -914061504, "ts": 1716454224281930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224336553, "dur": 29,
    "args": {
      "External id": 182966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182966,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 182966, "pid": 5, "tid": 7, "ts": 1716454224336553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282000, "dur": 11,
    "args": {
      "External id": 182966,
      "cbid": 211, "correlation": 182966
    }
  },
  {
    "ph": "s", "id": 182966, "pid": 76337, "tid": -914061504, "ts": 1716454224282000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282089, "dur": 1,
    "args": {
      "External id": 182977,
      "cbid": 251, "correlation": 182977
    }
  },
  {
    "ph": "f", "id": 182977, "pid": 76337, "tid": -914061504, "ts": 1716454224282089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224336584, "dur": 161,
    "args": {
      "External id": 182978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182978,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182978, "pid": 5, "tid": 7, "ts": 1716454224336584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282094, "dur": 12,
    "args": {
      "External id": 182978,
      "cbid": 211, "correlation": 182978
    }
  },
  {
    "ph": "s", "id": 182978, "pid": 76337, "tid": -914061504, "ts": 1716454224282094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282162, "dur": 1,
    "args": {
      "External id": 182989,
      "cbid": 251, "correlation": 182989
    }
  },
  {
    "ph": "f", "id": 182989, "pid": 76337, "tid": -914061504, "ts": 1716454224282162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224336747, "dur": 157,
    "args": {
      "External id": 182990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 182990,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 182990, "pid": 5, "tid": 7, "ts": 1716454224336747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282167, "dur": 12,
    "args": {
      "External id": 182990,
      "cbid": 211, "correlation": 182990
    }
  },
  {
    "ph": "s", "id": 182990, "pid": 76337, "tid": -914061504, "ts": 1716454224282167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282231, "dur": 1,
    "args": {
      "External id": 183001,
      "cbid": 251, "correlation": 183001
    }
  },
  {
    "ph": "f", "id": 183001, "pid": 76337, "tid": -914061504, "ts": 1716454224282231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224336906, "dur": 156,
    "args": {
      "External id": 183002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183002,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183002, "pid": 5, "tid": 7, "ts": 1716454224336906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282235, "dur": 11,
    "args": {
      "External id": 183002,
      "cbid": 211, "correlation": 183002
    }
  },
  {
    "ph": "s", "id": 183002, "pid": 76337, "tid": -914061504, "ts": 1716454224282235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224337063, "dur": 336,
    "args": {
      "External id": 183027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183027,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183027, "pid": 5, "tid": 7, "ts": 1716454224337063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282318, "dur": 13,
    "args": {
      "External id": 183027,
      "cbid": 211, "correlation": 183027
    }
  },
  {
    "ph": "s", "id": 183027, "pid": 76337, "tid": -914061504, "ts": 1716454224282318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282415, "dur": 1,
    "args": {
      "External id": 183045,
      "cbid": 251, "correlation": 183045
    }
  },
  {
    "ph": "f", "id": 183045, "pid": 76337, "tid": -914061504, "ts": 1716454224282415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224337400, "dur": 164,
    "args": {
      "External id": 183047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183047,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183047, "pid": 5, "tid": 7, "ts": 1716454224337400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282421, "dur": 14,
    "args": {
      "External id": 183047,
      "cbid": 211, "correlation": 183047
    }
  },
  {
    "ph": "s", "id": 183047, "pid": 76337, "tid": -914061504, "ts": 1716454224282421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224337566, "dur": 19,
    "args": {
      "External id": 183055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183055,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183055, "pid": 5, "tid": 7, "ts": 1716454224337566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282491, "dur": 12,
    "args": {
      "External id": 183055,
      "cbid": 211, "correlation": 183055
    }
  },
  {
    "ph": "s", "id": 183055, "pid": 76337, "tid": -914061504, "ts": 1716454224282491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224337586, "dur": 27,
    "args": {
      "External id": 183063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183063,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183063, "pid": 5, "tid": 7, "ts": 1716454224337586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282530, "dur": 8,
    "args": {
      "External id": 183063,
      "cbid": 211, "correlation": 183063
    }
  },
  {
    "ph": "s", "id": 183063, "pid": 76337, "tid": -914061504, "ts": 1716454224282530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224337615, "dur": 18,
    "args": {
      "External id": 183074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183074, "pid": 5, "tid": 7, "ts": 1716454224337615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282600, "dur": 13,
    "args": {
      "External id": 183074,
      "cbid": 211, "correlation": 183074
    }
  },
  {
    "ph": "s", "id": 183074, "pid": 76337, "tid": -914061504, "ts": 1716454224282600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224337634, "dur": 16,
    "args": {
      "External id": 183096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183096,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183096, "pid": 5, "tid": 7, "ts": 1716454224337634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282631, "dur": 8,
    "args": {
      "External id": 183096,
      "cbid": 211, "correlation": 183096
    }
  },
  {
    "ph": "s", "id": 183096, "pid": 76337, "tid": -914061504, "ts": 1716454224282631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282714, "dur": 1,
    "args": {
      "External id": 183107,
      "cbid": 251, "correlation": 183107
    }
  },
  {
    "ph": "f", "id": 183107, "pid": 76337, "tid": -914061504, "ts": 1716454224282714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224337651, "dur": 87,
    "args": {
      "External id": 183108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183108,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 183108, "pid": 5, "tid": 7, "ts": 1716454224337651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282719, "dur": 14,
    "args": {
      "External id": 183108,
      "cbid": 211, "correlation": 183108
    }
  },
  {
    "ph": "s", "id": 183108, "pid": 76337, "tid": -914061504, "ts": 1716454224282719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282788, "dur": 1,
    "args": {
      "External id": 183119,
      "cbid": 251, "correlation": 183119
    }
  },
  {
    "ph": "f", "id": 183119, "pid": 76337, "tid": -914061504, "ts": 1716454224282788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282792, "dur": 0,
    "args": {
      "External id": 183120,
      "cbid": 251, "correlation": 183120
    }
  },
  {
    "ph": "f", "id": 183120, "pid": 76337, "tid": -914061504, "ts": 1716454224282792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224337739, "dur": 12,
    "args": {
      "External id": 183121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183121,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183121, "pid": 5, "tid": 7, "ts": 1716454224337739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282794, "dur": 12,
    "args": {
      "External id": 183121,
      "cbid": 211, "correlation": 183121
    }
  },
  {
    "ph": "s", "id": 183121, "pid": 76337, "tid": -914061504, "ts": 1716454224282794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224337753, "dur": 5,
    "args": {
      "External id": 183123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183123,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183123, "pid": 5, "tid": 7, "ts": 1716454224337753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282807, "dur": 6,
    "args": {
      "External id": 183123,
      "cbid": 211, "correlation": 183123
    }
  },
  {
    "ph": "s", "id": 183123, "pid": 76337, "tid": -914061504, "ts": 1716454224282807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282865, "dur": 1,
    "args": {
      "External id": 183134,
      "cbid": 251, "correlation": 183134
    }
  },
  {
    "ph": "f", "id": 183134, "pid": 76337, "tid": -914061504, "ts": 1716454224282865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282868, "dur": 0,
    "args": {
      "External id": 183135,
      "cbid": 251, "correlation": 183135
    }
  },
  {
    "ph": "f", "id": 183135, "pid": 76337, "tid": -914061504, "ts": 1716454224282868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224337760, "dur": 8,
    "args": {
      "External id": 183136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183136,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183136, "pid": 5, "tid": 7, "ts": 1716454224337760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282870, "dur": 12,
    "args": {
      "External id": 183136,
      "cbid": 211, "correlation": 183136
    }
  },
  {
    "ph": "s", "id": 183136, "pid": 76337, "tid": -914061504, "ts": 1716454224282870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224337769, "dur": 3,
    "args": {
      "External id": 183138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183138,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183138, "pid": 5, "tid": 7, "ts": 1716454224337769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282883, "dur": 6,
    "args": {
      "External id": 183138,
      "cbid": 211, "correlation": 183138
    }
  },
  {
    "ph": "s", "id": 183138, "pid": 76337, "tid": -914061504, "ts": 1716454224282883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224337773, "dur": 53,
    "args": {
      "External id": 183163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183163,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183163, "pid": 5, "tid": 7, "ts": 1716454224337773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224282960, "dur": 12,
    "args": {
      "External id": 183163,
      "cbid": 211, "correlation": 183163
    }
  },
  {
    "ph": "s", "id": 183163, "pid": 76337, "tid": -914061504, "ts": 1716454224282960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283068, "dur": 1,
    "args": {
      "External id": 183181,
      "cbid": 251, "correlation": 183181
    }
  },
  {
    "ph": "f", "id": 183181, "pid": 76337, "tid": -914061504, "ts": 1716454224283068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224337828, "dur": 90,
    "args": {
      "External id": 183183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183183,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 183183, "pid": 5, "tid": 7, "ts": 1716454224337828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283074, "dur": 14,
    "args": {
      "External id": 183183,
      "cbid": 211, "correlation": 183183
    }
  },
  {
    "ph": "s", "id": 183183, "pid": 76337, "tid": -914061504, "ts": 1716454224283074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224337919, "dur": 9,
    "args": {
      "External id": 183191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183191,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183191, "pid": 5, "tid": 7, "ts": 1716454224337919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283143, "dur": 12,
    "args": {
      "External id": 183191,
      "cbid": 211, "correlation": 183191
    }
  },
  {
    "ph": "s", "id": 183191, "pid": 76337, "tid": -914061504, "ts": 1716454224283143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224337929, "dur": 20,
    "args": {
      "External id": 183199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183199,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183199, "pid": 5, "tid": 7, "ts": 1716454224337929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283184, "dur": 10,
    "args": {
      "External id": 183199,
      "cbid": 211, "correlation": 183199
    }
  },
  {
    "ph": "s", "id": 183199, "pid": 76337, "tid": -914061504, "ts": 1716454224283184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224337951, "dur": 17,
    "args": {
      "External id": 183221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183221,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183221, "pid": 5, "tid": 7, "ts": 1716454224337951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283236, "dur": 11,
    "args": {
      "External id": 183221,
      "cbid": 211, "correlation": 183221
    }
  },
  {
    "ph": "s", "id": 183221, "pid": 76337, "tid": -914061504, "ts": 1716454224283236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283323, "dur": 1,
    "args": {
      "External id": 183237,
      "cbid": 251, "correlation": 183237
    }
  },
  {
    "ph": "f", "id": 183237, "pid": 76337, "tid": -914061504, "ts": 1716454224283323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283328, "dur": 0,
    "args": {
      "External id": 183239,
      "cbid": 251, "correlation": 183239
    }
  },
  {
    "ph": "f", "id": 183239, "pid": 76337, "tid": -914061504, "ts": 1716454224283328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224337969, "dur": 493,
    "args": {
      "External id": 183240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183240,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183240, "pid": 5, "tid": 7, "ts": 1716454224337969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283330, "dur": 13,
    "args": {
      "External id": 183240,
      "cbid": 211, "correlation": 183240
    }
  },
  {
    "ph": "s", "id": 183240, "pid": 76337, "tid": -914061504, "ts": 1716454224283330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224338463, "dur": 65,
    "args": {
      "External id": 183248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183248, "pid": 5, "tid": 7, "ts": 1716454224338463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283394, "dur": 13,
    "args": {
      "External id": 183248,
      "cbid": 211, "correlation": 183248
    }
  },
  {
    "ph": "s", "id": 183248, "pid": 76337, "tid": -914061504, "ts": 1716454224283394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224338529, "dur": 67,
    "args": {
      "External id": 183256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183256, "pid": 5, "tid": 7, "ts": 1716454224338529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283425, "dur": 8,
    "args": {
      "External id": 183256,
      "cbid": 211, "correlation": 183256
    }
  },
  {
    "ph": "s", "id": 183256, "pid": 76337, "tid": -914061504, "ts": 1716454224283425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283505, "dur": 1,
    "args": {
      "External id": 183272,
      "cbid": 251, "correlation": 183272
    }
  },
  {
    "ph": "f", "id": 183272, "pid": 76337, "tid": -914061504, "ts": 1716454224283505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224338598, "dur": 1,
    "args": {
      "External id": 183274,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183274,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 183274, "pid": 5, "tid": 7, "ts": 1716454224338598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283510, "dur": 9,
    "args": {
      "External id": 183274,
      "cbid": 51, "correlation": 183274
    }
  },
  {
    "ph": "s", "id": 183274, "pid": 76337, "tid": -914061504, "ts": 1716454224283510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224338602, "dur": 268,
    "args": {
      "External id": 183275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183275,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 183275, "pid": 5, "tid": 7, "ts": 1716454224338602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283521, "dur": 11,
    "args": {
      "External id": 183275,
      "cbid": 211, "correlation": 183275
    }
  },
  {
    "ph": "s", "id": 183275, "pid": 76337, "tid": -914061504, "ts": 1716454224283521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224338872, "dur": 14,
    "args": {
      "External id": 183283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183283,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183283, "pid": 5, "tid": 7, "ts": 1716454224338872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283563, "dur": 10,
    "args": {
      "External id": 183283,
      "cbid": 211, "correlation": 183283
    }
  },
  {
    "ph": "s", "id": 183283, "pid": 76337, "tid": -914061504, "ts": 1716454224283563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224338886, "dur": 37,
    "args": {
      "External id": 183294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183294,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183294, "pid": 5, "tid": 7, "ts": 1716454224338886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283631, "dur": 12,
    "args": {
      "External id": 183294,
      "cbid": 211, "correlation": 183294
    }
  },
  {
    "ph": "s", "id": 183294, "pid": 76337, "tid": -914061504, "ts": 1716454224283631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283694, "dur": 0,
    "args": {
      "External id": 183306,
      "cbid": 317, "correlation": 183306
    }
  },
  {
    "ph": "f", "id": 183306, "pid": 76337, "tid": -914061504, "ts": 1716454224283694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283695, "dur": 0,
    "args": {
      "External id": 183307,
      "cbid": 203, "correlation": 183307
    }
  },
  {
    "ph": "f", "id": 183307, "pid": 76337, "tid": -914061504, "ts": 1716454224283695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283695, "dur": 0,
    "args": {
      "External id": 183308,
      "cbid": 205, "correlation": 183308
    }
  },
  {
    "ph": "f", "id": 183308, "pid": 76337, "tid": -914061504, "ts": 1716454224283695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224338925, "dur": 13,
    "args": {
      "External id": 183312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183312,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183312, "pid": 5, "tid": 7, "ts": 1716454224338925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283711, "dur": 12,
    "args": {
      "External id": 183312,
      "cbid": 211, "correlation": 183312
    }
  },
  {
    "ph": "s", "id": 183312, "pid": 76337, "tid": -914061504, "ts": 1716454224283711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224338939, "dur": 4,
    "args": {
      "External id": 183314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 183314, "pid": 5, "tid": 7, "ts": 1716454224338939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283728, "dur": 6,
    "args": {
      "External id": 183314,
      "cbid": 211, "correlation": 183314
    }
  },
  {
    "ph": "s", "id": 183314, "pid": 76337, "tid": -914061504, "ts": 1716454224283728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283736, "dur": 0,
    "args": {
      "External id": 183315,
      "cbid": 51, "correlation": 183315
    }
  },
  {
    "ph": "s", "id": 183315, "pid": 76337, "tid": -914061504, "ts": 1716454224283736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224338945, "dur": 95,
    "args": {
      "External id": 183316,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183316,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 183316, "pid": 5, "tid": 7, "ts": 1716454224338945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283737, "dur": 5,
    "args": {
      "External id": 183316,
      "cbid": 211, "correlation": 183316
    }
  },
  {
    "ph": "s", "id": 183316, "pid": 76337, "tid": -914061504, "ts": 1716454224283737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224339041, "dur": 16,
    "args": {
      "External id": 183321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183321, "pid": 5, "tid": 7, "ts": 1716454224339041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283765, "dur": 9,
    "args": {
      "External id": 183321,
      "cbid": 211, "correlation": 183321
    }
  },
  {
    "ph": "s", "id": 183321, "pid": 76337, "tid": -914061504, "ts": 1716454224283765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224339058, "dur": 11,
    "args": {
      "External id": 183329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183329,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183329, "pid": 5, "tid": 7, "ts": 1716454224339058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283796, "dur": 8,
    "args": {
      "External id": 183329,
      "cbid": 211, "correlation": 183329
    }
  },
  {
    "ph": "s", "id": 183329, "pid": 76337, "tid": -914061504, "ts": 1716454224283796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283867, "dur": 0,
    "args": {
      "External id": 183339,
      "cbid": 317, "correlation": 183339
    }
  },
  {
    "ph": "f", "id": 183339, "pid": 76337, "tid": -914061504, "ts": 1716454224283867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283868, "dur": 0,
    "args": {
      "External id": 183340,
      "cbid": 203, "correlation": 183340
    }
  },
  {
    "ph": "f", "id": 183340, "pid": 76337, "tid": -914061504, "ts": 1716454224283868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283869, "dur": 0,
    "args": {
      "External id": 183341,
      "cbid": 205, "correlation": 183341
    }
  },
  {
    "ph": "f", "id": 183341, "pid": 76337, "tid": -914061504, "ts": 1716454224283869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339071, "dur": 11,
    "args": {
      "External id": 183345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183345,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183345, "pid": 5, "tid": 7, "ts": 1716454224339071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283884, "dur": 12,
    "args": {
      "External id": 183345,
      "cbid": 211, "correlation": 183345
    }
  },
  {
    "ph": "s", "id": 183345, "pid": 76337, "tid": -914061504, "ts": 1716454224283884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339083, "dur": 160,
    "args": {
      "External id": 183347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183347,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183347, "pid": 5, "tid": 7, "ts": 1716454224339083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283898, "dur": 5,
    "args": {
      "External id": 183347,
      "cbid": 211, "correlation": 183347
    }
  },
  {
    "ph": "s", "id": 183347, "pid": 76337, "tid": -914061504, "ts": 1716454224283898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224339245, "dur": 1,
    "args": {
      "External id": 183349,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183349,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 183349, "pid": 5, "tid": 7, "ts": 1716454224339245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283910, "dur": 7,
    "args": {
      "External id": 183349,
      "cbid": 51, "correlation": 183349
    }
  },
  {
    "ph": "s", "id": 183349, "pid": 76337, "tid": -914061504, "ts": 1716454224283910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224339249, "dur": 198,
    "args": {
      "External id": 183350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183350,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 183350, "pid": 5, "tid": 7, "ts": 1716454224339249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283918, "dur": 8,
    "args": {
      "External id": 183350,
      "cbid": 211, "correlation": 183350
    }
  },
  {
    "ph": "s", "id": 183350, "pid": 76337, "tid": -914061504, "ts": 1716454224283918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339448, "dur": 6,
    "args": {
      "External id": 183352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183352,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183352, "pid": 5, "tid": 7, "ts": 1716454224339448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283931, "dur": 5,
    "args": {
      "External id": 183352,
      "cbid": 211, "correlation": 183352
    }
  },
  {
    "ph": "s", "id": 183352, "pid": 76337, "tid": -914061504, "ts": 1716454224283931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224339455, "dur": 6,
    "args": {
      "External id": 183358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183358, "pid": 5, "tid": 7, "ts": 1716454224339455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224283958, "dur": 9,
    "args": {
      "External id": 183358,
      "cbid": 211, "correlation": 183358
    }
  },
  {
    "ph": "s", "id": 183358, "pid": 76337, "tid": -914061504, "ts": 1716454224283958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339462, "dur": 11,
    "args": {
      "External id": 183378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183378,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183378, "pid": 5, "tid": 7, "ts": 1716454224339462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284073, "dur": 13,
    "args": {
      "External id": 183378,
      "cbid": 211, "correlation": 183378
    }
  },
  {
    "ph": "s", "id": 183378, "pid": 76337, "tid": -914061504, "ts": 1716454224284073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224339475, "dur": 4,
    "args": {
      "External id": 183390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183390,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183390, "pid": 5, "tid": 7, "ts": 1716454224339475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284096, "dur": 6,
    "args": {
      "External id": 183390,
      "cbid": 211, "correlation": 183390
    }
  },
  {
    "ph": "s", "id": 183390, "pid": 76337, "tid": -914061504, "ts": 1716454224284096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224339480, "dur": 8,
    "args": {
      "External id": 183393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183393,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183393, "pid": 5, "tid": 7, "ts": 1716454224339480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284116, "dur": 7,
    "args": {
      "External id": 183393,
      "cbid": 211, "correlation": 183393
    }
  },
  {
    "ph": "s", "id": 183393, "pid": 76337, "tid": -914061504, "ts": 1716454224284116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224339489, "dur": 5,
    "args": {
      "External id": 183402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183402,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183402, "pid": 5, "tid": 7, "ts": 1716454224339489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284155, "dur": 11,
    "args": {
      "External id": 183402,
      "cbid": 211, "correlation": 183402
    }
  },
  {
    "ph": "s", "id": 183402, "pid": 76337, "tid": -914061504, "ts": 1716454224284155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284208, "dur": 0,
    "args": {
      "External id": 183412,
      "cbid": 317, "correlation": 183412
    }
  },
  {
    "ph": "f", "id": 183412, "pid": 76337, "tid": -914061504, "ts": 1716454224284208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284208, "dur": 0,
    "args": {
      "External id": 183413,
      "cbid": 203, "correlation": 183413
    }
  },
  {
    "ph": "f", "id": 183413, "pid": 76337, "tid": -914061504, "ts": 1716454224284208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284209, "dur": 0,
    "args": {
      "External id": 183414,
      "cbid": 205, "correlation": 183414
    }
  },
  {
    "ph": "f", "id": 183414, "pid": 76337, "tid": -914061504, "ts": 1716454224284209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339496, "dur": 5,
    "args": {
      "External id": 183418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183418,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183418, "pid": 5, "tid": 7, "ts": 1716454224339496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284224, "dur": 11,
    "args": {
      "External id": 183418,
      "cbid": 211, "correlation": 183418
    }
  },
  {
    "ph": "s", "id": 183418, "pid": 76337, "tid": -914061504, "ts": 1716454224284224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339502, "dur": 160,
    "args": {
      "External id": 183420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183420,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183420, "pid": 5, "tid": 7, "ts": 1716454224339502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284238, "dur": 5,
    "args": {
      "External id": 183420,
      "cbid": 211, "correlation": 183420
    }
  },
  {
    "ph": "s", "id": 183420, "pid": 76337, "tid": -914061504, "ts": 1716454224284238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224339664, "dur": 1,
    "args": {
      "External id": 183422,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183422,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 183422, "pid": 5, "tid": 7, "ts": 1716454224339664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284249, "dur": 6,
    "args": {
      "External id": 183422,
      "cbid": 51, "correlation": 183422
    }
  },
  {
    "ph": "s", "id": 183422, "pid": 76337, "tid": -914061504, "ts": 1716454224284249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224339668, "dur": 266,
    "args": {
      "External id": 183423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183423,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183423, "pid": 5, "tid": 7, "ts": 1716454224339668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284257, "dur": 6,
    "args": {
      "External id": 183423,
      "cbid": 211, "correlation": 183423
    }
  },
  {
    "ph": "s", "id": 183423, "pid": 76337, "tid": -914061504, "ts": 1716454224284257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339934, "dur": 5,
    "args": {
      "External id": 183425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183425,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183425, "pid": 5, "tid": 7, "ts": 1716454224339934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284268, "dur": 5,
    "args": {
      "External id": 183425,
      "cbid": 211, "correlation": 183425
    }
  },
  {
    "ph": "s", "id": 183425, "pid": 76337, "tid": -914061504, "ts": 1716454224284268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224339941, "dur": 6,
    "args": {
      "External id": 183431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183431,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183431, "pid": 5, "tid": 7, "ts": 1716454224339941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284296, "dur": 8,
    "args": {
      "External id": 183431,
      "cbid": 211, "correlation": 183431
    }
  },
  {
    "ph": "s", "id": 183431, "pid": 76337, "tid": -914061504, "ts": 1716454224284296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224339948, "dur": 3,
    "args": {
      "External id": 183439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183439,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 183439, "pid": 5, "tid": 7, "ts": 1716454224339948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284340, "dur": 9,
    "args": {
      "External id": 183439,
      "cbid": 211, "correlation": 183439
    }
  },
  {
    "ph": "s", "id": 183439, "pid": 76337, "tid": -914061504, "ts": 1716454224284340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284406, "dur": 1,
    "args": {
      "External id": 183455,
      "cbid": 251, "correlation": 183455
    }
  },
  {
    "ph": "f", "id": 183455, "pid": 76337, "tid": -914061504, "ts": 1716454224284406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284411, "dur": 0,
    "args": {
      "External id": 183457,
      "cbid": 251, "correlation": 183457
    }
  },
  {
    "ph": "f", "id": 183457, "pid": 76337, "tid": -914061504, "ts": 1716454224284411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224339953, "dur": 13,
    "args": {
      "External id": 183458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183458,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183458, "pid": 5, "tid": 7, "ts": 1716454224339953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284414, "dur": 12,
    "args": {
      "External id": 183458,
      "cbid": 211, "correlation": 183458
    }
  },
  {
    "ph": "s", "id": 183458, "pid": 76337, "tid": -914061504, "ts": 1716454224284414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224339967, "dur": 5,
    "args": {
      "External id": 183460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183460,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183460, "pid": 5, "tid": 7, "ts": 1716454224339967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284427, "dur": 6,
    "args": {
      "External id": 183460,
      "cbid": 211, "correlation": 183460
    }
  },
  {
    "ph": "s", "id": 183460, "pid": 76337, "tid": -914061504, "ts": 1716454224284427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224339973, "dur": 6,
    "args": {
      "External id": 183470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183470,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183470, "pid": 5, "tid": 7, "ts": 1716454224339973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284485, "dur": 12,
    "args": {
      "External id": 183470,
      "cbid": 211, "correlation": 183470
    }
  },
  {
    "ph": "s", "id": 183470, "pid": 76337, "tid": -914061504, "ts": 1716454224284485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224339980, "dur": 9,
    "args": {
      "External id": 183490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183490,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183490, "pid": 5, "tid": 7, "ts": 1716454224339980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284551, "dur": 10,
    "args": {
      "External id": 183490,
      "cbid": 211, "correlation": 183490
    }
  },
  {
    "ph": "s", "id": 183490, "pid": 76337, "tid": -914061504, "ts": 1716454224284551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224339991, "dur": 3,
    "args": {
      "External id": 183502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183502,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183502, "pid": 5, "tid": 7, "ts": 1716454224339991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284572, "dur": 7,
    "args": {
      "External id": 183502,
      "cbid": 211, "correlation": 183502
    }
  },
  {
    "ph": "s", "id": 183502, "pid": 76337, "tid": -914061504, "ts": 1716454224284572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224339995, "dur": 7,
    "args": {
      "External id": 183505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183505,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183505, "pid": 5, "tid": 7, "ts": 1716454224339995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284592, "dur": 6,
    "args": {
      "External id": 183505,
      "cbid": 211, "correlation": 183505
    }
  },
  {
    "ph": "s", "id": 183505, "pid": 76337, "tid": -914061504, "ts": 1716454224284592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224340004, "dur": 4,
    "args": {
      "External id": 183514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183514,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183514, "pid": 5, "tid": 7, "ts": 1716454224340004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284632, "dur": 10,
    "args": {
      "External id": 183514,
      "cbid": 211, "correlation": 183514
    }
  },
  {
    "ph": "s", "id": 183514, "pid": 76337, "tid": -914061504, "ts": 1716454224284632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284699, "dur": 0,
    "args": {
      "External id": 183524,
      "cbid": 317, "correlation": 183524
    }
  },
  {
    "ph": "f", "id": 183524, "pid": 76337, "tid": -914061504, "ts": 1716454224284699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284700, "dur": 0,
    "args": {
      "External id": 183525,
      "cbid": 203, "correlation": 183525
    }
  },
  {
    "ph": "f", "id": 183525, "pid": 76337, "tid": -914061504, "ts": 1716454224284700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284701, "dur": 0,
    "args": {
      "External id": 183526,
      "cbid": 205, "correlation": 183526
    }
  },
  {
    "ph": "f", "id": 183526, "pid": 76337, "tid": -914061504, "ts": 1716454224284701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340010, "dur": 5,
    "args": {
      "External id": 183530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183530,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183530, "pid": 5, "tid": 7, "ts": 1716454224340010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284715, "dur": 12,
    "args": {
      "External id": 183530,
      "cbid": 211, "correlation": 183530
    }
  },
  {
    "ph": "s", "id": 183530, "pid": 76337, "tid": -914061504, "ts": 1716454224284715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340016, "dur": 160,
    "args": {
      "External id": 183532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183532,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183532, "pid": 5, "tid": 7, "ts": 1716454224340016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284730, "dur": 5,
    "args": {
      "External id": 183532,
      "cbid": 211, "correlation": 183532
    }
  },
  {
    "ph": "s", "id": 183532, "pid": 76337, "tid": -914061504, "ts": 1716454224284730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224340178, "dur": 1,
    "args": {
      "External id": 183534,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183534,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 183534, "pid": 5, "tid": 7, "ts": 1716454224340178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284741, "dur": 6,
    "args": {
      "External id": 183534,
      "cbid": 51, "correlation": 183534
    }
  },
  {
    "ph": "s", "id": 183534, "pid": 76337, "tid": -914061504, "ts": 1716454224284741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224340182, "dur": 255,
    "args": {
      "External id": 183535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183535,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183535, "pid": 5, "tid": 7, "ts": 1716454224340182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284748, "dur": 6,
    "args": {
      "External id": 183535,
      "cbid": 211, "correlation": 183535
    }
  },
  {
    "ph": "s", "id": 183535, "pid": 76337, "tid": -914061504, "ts": 1716454224284748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340438, "dur": 6,
    "args": {
      "External id": 183537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183537,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183537, "pid": 5, "tid": 7, "ts": 1716454224340438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284757, "dur": 5,
    "args": {
      "External id": 183537,
      "cbid": 211, "correlation": 183537
    }
  },
  {
    "ph": "s", "id": 183537, "pid": 76337, "tid": -914061504, "ts": 1716454224284757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224340446, "dur": 6,
    "args": {
      "External id": 183543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183543,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183543, "pid": 5, "tid": 7, "ts": 1716454224340446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284787, "dur": 8,
    "args": {
      "External id": 183543,
      "cbid": 211, "correlation": 183543
    }
  },
  {
    "ph": "s", "id": 183543, "pid": 76337, "tid": -914061504, "ts": 1716454224284787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224340453, "dur": 5,
    "args": {
      "External id": 183551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183551,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183551, "pid": 5, "tid": 7, "ts": 1716454224340453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284819, "dur": 8,
    "args": {
      "External id": 183551,
      "cbid": 211, "correlation": 183551
    }
  },
  {
    "ph": "s", "id": 183551, "pid": 76337, "tid": -914061504, "ts": 1716454224284819,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224340460, "dur": 4,
    "args": {
      "External id": 183559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183559,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183559, "pid": 5, "tid": 7, "ts": 1716454224340460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284849, "dur": 8,
    "args": {
      "External id": 183559,
      "cbid": 211, "correlation": 183559
    }
  },
  {
    "ph": "s", "id": 183559, "pid": 76337, "tid": -914061504, "ts": 1716454224284849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340465, "dur": 9,
    "args": {
      "External id": 183579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183579,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183579, "pid": 5, "tid": 7, "ts": 1716454224340465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284922, "dur": 13,
    "args": {
      "External id": 183579,
      "cbid": 211, "correlation": 183579
    }
  },
  {
    "ph": "s", "id": 183579, "pid": 76337, "tid": -914061504, "ts": 1716454224284922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224340476, "dur": 4,
    "args": {
      "External id": 183591,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183591,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183591, "pid": 5, "tid": 7, "ts": 1716454224340476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284945, "dur": 6,
    "args": {
      "External id": 183591,
      "cbid": 211, "correlation": 183591
    }
  },
  {
    "ph": "s", "id": 183591, "pid": 76337, "tid": -914061504, "ts": 1716454224284945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224340481, "dur": 6,
    "args": {
      "External id": 183594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183594,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183594, "pid": 5, "tid": 7, "ts": 1716454224340481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224284963, "dur": 6,
    "args": {
      "External id": 183594,
      "cbid": 211, "correlation": 183594
    }
  },
  {
    "ph": "s", "id": 183594, "pid": 76337, "tid": -914061504, "ts": 1716454224284963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224340489, "dur": 4,
    "args": {
      "External id": 183603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183603,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183603, "pid": 5, "tid": 7, "ts": 1716454224340489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285010, "dur": 10,
    "args": {
      "External id": 183603,
      "cbid": 211, "correlation": 183603
    }
  },
  {
    "ph": "s", "id": 183603, "pid": 76337, "tid": -914061504, "ts": 1716454224285010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285062, "dur": 0,
    "args": {
      "External id": 183613,
      "cbid": 317, "correlation": 183613
    }
  },
  {
    "ph": "f", "id": 183613, "pid": 76337, "tid": -914061504, "ts": 1716454224285062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285063, "dur": 0,
    "args": {
      "External id": 183614,
      "cbid": 203, "correlation": 183614
    }
  },
  {
    "ph": "f", "id": 183614, "pid": 76337, "tid": -914061504, "ts": 1716454224285063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285064, "dur": 0,
    "args": {
      "External id": 183615,
      "cbid": 205, "correlation": 183615
    }
  },
  {
    "ph": "f", "id": 183615, "pid": 76337, "tid": -914061504, "ts": 1716454224285064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340495, "dur": 5,
    "args": {
      "External id": 183619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183619,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183619, "pid": 5, "tid": 7, "ts": 1716454224340495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285077, "dur": 12,
    "args": {
      "External id": 183619,
      "cbid": 211, "correlation": 183619
    }
  },
  {
    "ph": "s", "id": 183619, "pid": 76337, "tid": -914061504, "ts": 1716454224285077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340501, "dur": 160,
    "args": {
      "External id": 183621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183621,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183621, "pid": 5, "tid": 7, "ts": 1716454224340501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285091, "dur": 5,
    "args": {
      "External id": 183621,
      "cbid": 211, "correlation": 183621
    }
  },
  {
    "ph": "s", "id": 183621, "pid": 76337, "tid": -914061504, "ts": 1716454224285091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224340663, "dur": 1,
    "args": {
      "External id": 183623,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183623,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 183623, "pid": 5, "tid": 7, "ts": 1716454224340663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285102, "dur": 6,
    "args": {
      "External id": 183623,
      "cbid": 51, "correlation": 183623
    }
  },
  {
    "ph": "s", "id": 183623, "pid": 76337, "tid": -914061504, "ts": 1716454224285102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224340666, "dur": 254,
    "args": {
      "External id": 183624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183624,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183624, "pid": 5, "tid": 7, "ts": 1716454224340666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285109, "dur": 6,
    "args": {
      "External id": 183624,
      "cbid": 211, "correlation": 183624
    }
  },
  {
    "ph": "s", "id": 183624, "pid": 76337, "tid": -914061504, "ts": 1716454224285109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340922, "dur": 6,
    "args": {
      "External id": 183626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183626, "pid": 5, "tid": 7, "ts": 1716454224340922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285118, "dur": 5,
    "args": {
      "External id": 183626,
      "cbid": 211, "correlation": 183626
    }
  },
  {
    "ph": "s", "id": 183626, "pid": 76337, "tid": -914061504, "ts": 1716454224285118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224340929, "dur": 6,
    "args": {
      "External id": 183632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183632,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183632, "pid": 5, "tid": 7, "ts": 1716454224340929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285147, "dur": 8,
    "args": {
      "External id": 183632,
      "cbid": 211, "correlation": 183632
    }
  },
  {
    "ph": "s", "id": 183632, "pid": 76337, "tid": -914061504, "ts": 1716454224285147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224340936, "dur": 3,
    "args": {
      "External id": 183640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183640,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 183640, "pid": 5, "tid": 7, "ts": 1716454224340936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285190, "dur": 9,
    "args": {
      "External id": 183640,
      "cbid": 211, "correlation": 183640
    }
  },
  {
    "ph": "s", "id": 183640, "pid": 76337, "tid": -914061504, "ts": 1716454224285190,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285253, "dur": 1,
    "args": {
      "External id": 183656,
      "cbid": 251, "correlation": 183656
    }
  },
  {
    "ph": "f", "id": 183656, "pid": 76337, "tid": -914061504, "ts": 1716454224285253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285258, "dur": 0,
    "args": {
      "External id": 183658,
      "cbid": 251, "correlation": 183658
    }
  },
  {
    "ph": "f", "id": 183658, "pid": 76337, "tid": -914061504, "ts": 1716454224285258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224340940, "dur": 10,
    "args": {
      "External id": 183659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183659,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183659, "pid": 5, "tid": 7, "ts": 1716454224340940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285260, "dur": 11,
    "args": {
      "External id": 183659,
      "cbid": 211, "correlation": 183659
    }
  },
  {
    "ph": "s", "id": 183659, "pid": 76337, "tid": -914061504, "ts": 1716454224285260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224340952, "dur": 4,
    "args": {
      "External id": 183661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183661,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183661, "pid": 5, "tid": 7, "ts": 1716454224340952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285273, "dur": 5,
    "args": {
      "External id": 183661,
      "cbid": 211, "correlation": 183661
    }
  },
  {
    "ph": "s", "id": 183661, "pid": 76337, "tid": -914061504, "ts": 1716454224285273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224340957, "dur": 6,
    "args": {
      "External id": 183671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183671, "pid": 5, "tid": 7, "ts": 1716454224340957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285329, "dur": 13,
    "args": {
      "External id": 183671,
      "cbid": 211, "correlation": 183671
    }
  },
  {
    "ph": "s", "id": 183671, "pid": 76337, "tid": -914061504, "ts": 1716454224285329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340964, "dur": 9,
    "args": {
      "External id": 183691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183691,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183691, "pid": 5, "tid": 7, "ts": 1716454224340964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285394, "dur": 11,
    "args": {
      "External id": 183691,
      "cbid": 211, "correlation": 183691
    }
  },
  {
    "ph": "s", "id": 183691, "pid": 76337, "tid": -914061504, "ts": 1716454224285394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224340974, "dur": 4,
    "args": {
      "External id": 183703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183703,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183703, "pid": 5, "tid": 7, "ts": 1716454224340974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285415, "dur": 6,
    "args": {
      "External id": 183703,
      "cbid": 211, "correlation": 183703
    }
  },
  {
    "ph": "s", "id": 183703, "pid": 76337, "tid": -914061504, "ts": 1716454224285415,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224340980, "dur": 6,
    "args": {
      "External id": 183706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183706, "pid": 5, "tid": 7, "ts": 1716454224340980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285434, "dur": 6,
    "args": {
      "External id": 183706,
      "cbid": 211, "correlation": 183706
    }
  },
  {
    "ph": "s", "id": 183706, "pid": 76337, "tid": -914061504, "ts": 1716454224285434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224340987, "dur": 5,
    "args": {
      "External id": 183715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183715,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183715, "pid": 5, "tid": 7, "ts": 1716454224340987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285474, "dur": 10,
    "args": {
      "External id": 183715,
      "cbid": 211, "correlation": 183715
    }
  },
  {
    "ph": "s", "id": 183715, "pid": 76337, "tid": -914061504, "ts": 1716454224285474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285536, "dur": 0,
    "args": {
      "External id": 183725,
      "cbid": 317, "correlation": 183725
    }
  },
  {
    "ph": "f", "id": 183725, "pid": 76337, "tid": -914061504, "ts": 1716454224285536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285536, "dur": 0,
    "args": {
      "External id": 183726,
      "cbid": 203, "correlation": 183726
    }
  },
  {
    "ph": "f", "id": 183726, "pid": 76337, "tid": -914061504, "ts": 1716454224285536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285537, "dur": 0,
    "args": {
      "External id": 183727,
      "cbid": 205, "correlation": 183727
    }
  },
  {
    "ph": "f", "id": 183727, "pid": 76337, "tid": -914061504, "ts": 1716454224285537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224340993, "dur": 5,
    "args": {
      "External id": 183731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183731,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183731, "pid": 5, "tid": 7, "ts": 1716454224340993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285551, "dur": 12,
    "args": {
      "External id": 183731,
      "cbid": 211, "correlation": 183731
    }
  },
  {
    "ph": "s", "id": 183731, "pid": 76337, "tid": -914061504, "ts": 1716454224285551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341000, "dur": 160,
    "args": {
      "External id": 183733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183733,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183733, "pid": 5, "tid": 7, "ts": 1716454224341000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285566, "dur": 5,
    "args": {
      "External id": 183733,
      "cbid": 211, "correlation": 183733
    }
  },
  {
    "ph": "s", "id": 183733, "pid": 76337, "tid": -914061504, "ts": 1716454224285566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224341162, "dur": 1,
    "args": {
      "External id": 183735,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183735,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 183735, "pid": 5, "tid": 7, "ts": 1716454224341162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285576, "dur": 6,
    "args": {
      "External id": 183735,
      "cbid": 51, "correlation": 183735
    }
  },
  {
    "ph": "s", "id": 183735, "pid": 76337, "tid": -914061504, "ts": 1716454224285576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224341166, "dur": 254,
    "args": {
      "External id": 183736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183736,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183736, "pid": 5, "tid": 7, "ts": 1716454224341166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285584, "dur": 6,
    "args": {
      "External id": 183736,
      "cbid": 211, "correlation": 183736
    }
  },
  {
    "ph": "s", "id": 183736, "pid": 76337, "tid": -914061504, "ts": 1716454224285584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341421, "dur": 6,
    "args": {
      "External id": 183738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183738, "pid": 5, "tid": 7, "ts": 1716454224341421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285595, "dur": 6,
    "args": {
      "External id": 183738,
      "cbid": 211, "correlation": 183738
    }
  },
  {
    "ph": "s", "id": 183738, "pid": 76337, "tid": -914061504, "ts": 1716454224285595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224341428, "dur": 6,
    "args": {
      "External id": 183744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183744,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183744, "pid": 5, "tid": 7, "ts": 1716454224341428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285624, "dur": 9,
    "args": {
      "External id": 183744,
      "cbid": 211, "correlation": 183744
    }
  },
  {
    "ph": "s", "id": 183744, "pid": 76337, "tid": -914061504, "ts": 1716454224285624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224341435, "dur": 5,
    "args": {
      "External id": 183752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183752,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183752, "pid": 5, "tid": 7, "ts": 1716454224341435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285657, "dur": 8,
    "args": {
      "External id": 183752,
      "cbid": 211, "correlation": 183752
    }
  },
  {
    "ph": "s", "id": 183752, "pid": 76337, "tid": -914061504, "ts": 1716454224285657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224341441, "dur": 4,
    "args": {
      "External id": 183760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183760,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183760, "pid": 5, "tid": 7, "ts": 1716454224341441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285686, "dur": 8,
    "args": {
      "External id": 183760,
      "cbid": 211, "correlation": 183760
    }
  },
  {
    "ph": "s", "id": 183760, "pid": 76337, "tid": -914061504, "ts": 1716454224285686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341447, "dur": 10,
    "args": {
      "External id": 183780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183780,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183780, "pid": 5, "tid": 7, "ts": 1716454224341447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285797, "dur": 14,
    "args": {
      "External id": 183780,
      "cbid": 211, "correlation": 183780
    }
  },
  {
    "ph": "s", "id": 183780, "pid": 76337, "tid": -914061504, "ts": 1716454224285797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224341458, "dur": 4,
    "args": {
      "External id": 183792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183792,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183792, "pid": 5, "tid": 7, "ts": 1716454224341458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285820, "dur": 6,
    "args": {
      "External id": 183792,
      "cbid": 211, "correlation": 183792
    }
  },
  {
    "ph": "s", "id": 183792, "pid": 76337, "tid": -914061504, "ts": 1716454224285820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224341463, "dur": 7,
    "args": {
      "External id": 183795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183795,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183795, "pid": 5, "tid": 7, "ts": 1716454224341463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285838, "dur": 6,
    "args": {
      "External id": 183795,
      "cbid": 211, "correlation": 183795
    }
  },
  {
    "ph": "s", "id": 183795, "pid": 76337, "tid": -914061504, "ts": 1716454224285838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224341471, "dur": 4,
    "args": {
      "External id": 183804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183804,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183804, "pid": 5, "tid": 7, "ts": 1716454224341471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285876, "dur": 9,
    "args": {
      "External id": 183804,
      "cbid": 211, "correlation": 183804
    }
  },
  {
    "ph": "s", "id": 183804, "pid": 76337, "tid": -914061504, "ts": 1716454224285876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285932, "dur": 0,
    "args": {
      "External id": 183814,
      "cbid": 317, "correlation": 183814
    }
  },
  {
    "ph": "f", "id": 183814, "pid": 76337, "tid": -914061504, "ts": 1716454224285932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285933, "dur": 0,
    "args": {
      "External id": 183815,
      "cbid": 203, "correlation": 183815
    }
  },
  {
    "ph": "f", "id": 183815, "pid": 76337, "tid": -914061504, "ts": 1716454224285933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285934, "dur": 0,
    "args": {
      "External id": 183816,
      "cbid": 205, "correlation": 183816
    }
  },
  {
    "ph": "f", "id": 183816, "pid": 76337, "tid": -914061504, "ts": 1716454224285934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341477, "dur": 5,
    "args": {
      "External id": 183820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183820,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183820, "pid": 5, "tid": 7, "ts": 1716454224341477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285947, "dur": 12,
    "args": {
      "External id": 183820,
      "cbid": 211, "correlation": 183820
    }
  },
  {
    "ph": "s", "id": 183820, "pid": 76337, "tid": -914061504, "ts": 1716454224285947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341483, "dur": 159,
    "args": {
      "External id": 183822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183822,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183822, "pid": 5, "tid": 7, "ts": 1716454224341483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285961, "dur": 6,
    "args": {
      "External id": 183822,
      "cbid": 211, "correlation": 183822
    }
  },
  {
    "ph": "s", "id": 183822, "pid": 76337, "tid": -914061504, "ts": 1716454224285961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224341644, "dur": 1,
    "args": {
      "External id": 183824,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183824,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 183824, "pid": 5, "tid": 7, "ts": 1716454224341644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285972, "dur": 15,
    "args": {
      "External id": 183824,
      "cbid": 51, "correlation": 183824
    }
  },
  {
    "ph": "s", "id": 183824, "pid": 76337, "tid": -914061504, "ts": 1716454224285972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224341648, "dur": 254,
    "args": {
      "External id": 183825,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183825,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183825, "pid": 5, "tid": 7, "ts": 1716454224341648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285988, "dur": 7,
    "args": {
      "External id": 183825,
      "cbid": 211, "correlation": 183825
    }
  },
  {
    "ph": "s", "id": 183825, "pid": 76337, "tid": -914061504, "ts": 1716454224285988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341903, "dur": 6,
    "args": {
      "External id": 183827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183827,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183827, "pid": 5, "tid": 7, "ts": 1716454224341903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224285998, "dur": 5,
    "args": {
      "External id": 183827,
      "cbid": 211, "correlation": 183827
    }
  },
  {
    "ph": "s", "id": 183827, "pid": 76337, "tid": -914061504, "ts": 1716454224285998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224341910, "dur": 6,
    "args": {
      "External id": 183833,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183833,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183833, "pid": 5, "tid": 7, "ts": 1716454224341910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286027, "dur": 8,
    "args": {
      "External id": 183833,
      "cbid": 211, "correlation": 183833
    }
  },
  {
    "ph": "s", "id": 183833, "pid": 76337, "tid": -914061504, "ts": 1716454224286027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224341918, "dur": 3,
    "args": {
      "External id": 183841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183841,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 183841, "pid": 5, "tid": 7, "ts": 1716454224341918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286071, "dur": 10,
    "args": {
      "External id": 183841,
      "cbid": 211, "correlation": 183841
    }
  },
  {
    "ph": "s", "id": 183841, "pid": 76337, "tid": -914061504, "ts": 1716454224286071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286133, "dur": 1,
    "args": {
      "External id": 183857,
      "cbid": 251, "correlation": 183857
    }
  },
  {
    "ph": "f", "id": 183857, "pid": 76337, "tid": -914061504, "ts": 1716454224286133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286138, "dur": 0,
    "args": {
      "External id": 183859,
      "cbid": 251, "correlation": 183859
    }
  },
  {
    "ph": "f", "id": 183859, "pid": 76337, "tid": -914061504, "ts": 1716454224286138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224341922, "dur": 10,
    "args": {
      "External id": 183860,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183860,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183860, "pid": 5, "tid": 7, "ts": 1716454224341922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286140, "dur": 11,
    "args": {
      "External id": 183860,
      "cbid": 211, "correlation": 183860
    }
  },
  {
    "ph": "s", "id": 183860, "pid": 76337, "tid": -914061504, "ts": 1716454224286140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224341933, "dur": 4,
    "args": {
      "External id": 183862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183862,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183862, "pid": 5, "tid": 7, "ts": 1716454224341933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286153, "dur": 5,
    "args": {
      "External id": 183862,
      "cbid": 211, "correlation": 183862
    }
  },
  {
    "ph": "s", "id": 183862, "pid": 76337, "tid": -914061504, "ts": 1716454224286153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224341938, "dur": 6,
    "args": {
      "External id": 183872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183872,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183872, "pid": 5, "tid": 7, "ts": 1716454224341938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286209, "dur": 13,
    "args": {
      "External id": 183872,
      "cbid": 211, "correlation": 183872
    }
  },
  {
    "ph": "s", "id": 183872, "pid": 76337, "tid": -914061504, "ts": 1716454224286209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341945, "dur": 9,
    "args": {
      "External id": 183892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183892,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183892, "pid": 5, "tid": 7, "ts": 1716454224341945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286276, "dur": 10,
    "args": {
      "External id": 183892,
      "cbid": 211, "correlation": 183892
    }
  },
  {
    "ph": "s", "id": 183892, "pid": 76337, "tid": -914061504, "ts": 1716454224286276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224341956, "dur": 4,
    "args": {
      "External id": 183904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183904,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183904, "pid": 5, "tid": 7, "ts": 1716454224341956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286295, "dur": 6,
    "args": {
      "External id": 183904,
      "cbid": 211, "correlation": 183904
    }
  },
  {
    "ph": "s", "id": 183904, "pid": 76337, "tid": -914061504, "ts": 1716454224286295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224341961, "dur": 6,
    "args": {
      "External id": 183907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183907,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183907, "pid": 5, "tid": 7, "ts": 1716454224341961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286314, "dur": 7,
    "args": {
      "External id": 183907,
      "cbid": 211, "correlation": 183907
    }
  },
  {
    "ph": "s", "id": 183907, "pid": 76337, "tid": -914061504, "ts": 1716454224286314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224341969, "dur": 4,
    "args": {
      "External id": 183916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183916,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183916, "pid": 5, "tid": 7, "ts": 1716454224341969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286354, "dur": 9,
    "args": {
      "External id": 183916,
      "cbid": 211, "correlation": 183916
    }
  },
  {
    "ph": "s", "id": 183916, "pid": 76337, "tid": -914061504, "ts": 1716454224286354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286420, "dur": 0,
    "args": {
      "External id": 183926,
      "cbid": 317, "correlation": 183926
    }
  },
  {
    "ph": "f", "id": 183926, "pid": 76337, "tid": -914061504, "ts": 1716454224286420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286421, "dur": 0,
    "args": {
      "External id": 183927,
      "cbid": 203, "correlation": 183927
    }
  },
  {
    "ph": "f", "id": 183927, "pid": 76337, "tid": -914061504, "ts": 1716454224286421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286421, "dur": 0,
    "args": {
      "External id": 183928,
      "cbid": 205, "correlation": 183928
    }
  },
  {
    "ph": "f", "id": 183928, "pid": 76337, "tid": -914061504, "ts": 1716454224286421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341974, "dur": 5,
    "args": {
      "External id": 183932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183932,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183932, "pid": 5, "tid": 7, "ts": 1716454224341974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286436, "dur": 12,
    "args": {
      "External id": 183932,
      "cbid": 211, "correlation": 183932
    }
  },
  {
    "ph": "s", "id": 183932, "pid": 76337, "tid": -914061504, "ts": 1716454224286436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224341981, "dur": 159,
    "args": {
      "External id": 183934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183934,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183934, "pid": 5, "tid": 7, "ts": 1716454224341981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286450, "dur": 5,
    "args": {
      "External id": 183934,
      "cbid": 211, "correlation": 183934
    }
  },
  {
    "ph": "s", "id": 183934, "pid": 76337, "tid": -914061504, "ts": 1716454224286450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224342141, "dur": 1,
    "args": {
      "External id": 183936,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 183936,
      "bytes": 240, "memory bandwidth (GB/s)": 0.1563517915309446
    }
  },
  {
    "ph": "f", "id": 183936, "pid": 5, "tid": 7, "ts": 1716454224342141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286461, "dur": 6,
    "args": {
      "External id": 183936,
      "cbid": 51, "correlation": 183936
    }
  },
  {
    "ph": "s", "id": 183936, "pid": 76337, "tid": -914061504, "ts": 1716454224286461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224342145, "dur": 254,
    "args": {
      "External id": 183937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183937,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 183937, "pid": 5, "tid": 7, "ts": 1716454224342145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286468, "dur": 6,
    "args": {
      "External id": 183937,
      "cbid": 211, "correlation": 183937
    }
  },
  {
    "ph": "s", "id": 183937, "pid": 76337, "tid": -914061504, "ts": 1716454224286468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224342401, "dur": 6,
    "args": {
      "External id": 183939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183939,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 183939, "pid": 5, "tid": 7, "ts": 1716454224342401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286479, "dur": 5,
    "args": {
      "External id": 183939,
      "cbid": 211, "correlation": 183939
    }
  },
  {
    "ph": "s", "id": 183939, "pid": 76337, "tid": -914061504, "ts": 1716454224286479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224342407, "dur": 6,
    "args": {
      "External id": 183945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183945,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183945, "pid": 5, "tid": 7, "ts": 1716454224342407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286506, "dur": 8,
    "args": {
      "External id": 183945,
      "cbid": 211, "correlation": 183945
    }
  },
  {
    "ph": "s", "id": 183945, "pid": 76337, "tid": -914061504, "ts": 1716454224286506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224342415, "dur": 5,
    "args": {
      "External id": 183953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183953,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183953, "pid": 5, "tid": 7, "ts": 1716454224342415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286540, "dur": 8,
    "args": {
      "External id": 183953,
      "cbid": 211, "correlation": 183953
    }
  },
  {
    "ph": "s", "id": 183953, "pid": 76337, "tid": -914061504, "ts": 1716454224286540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224342421, "dur": 4,
    "args": {
      "External id": 183961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183961,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183961, "pid": 5, "tid": 7, "ts": 1716454224342421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286568, "dur": 9,
    "args": {
      "External id": 183961,
      "cbid": 211, "correlation": 183961
    }
  },
  {
    "ph": "s", "id": 183961, "pid": 76337, "tid": -914061504, "ts": 1716454224286568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224342427, "dur": 9,
    "args": {
      "External id": 183981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183981,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 183981, "pid": 5, "tid": 7, "ts": 1716454224342427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286765, "dur": 15,
    "args": {
      "External id": 183981,
      "cbid": 211, "correlation": 183981
    }
  },
  {
    "ph": "s", "id": 183981, "pid": 76337, "tid": -914061504, "ts": 1716454224286765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224342437, "dur": 4,
    "args": {
      "External id": 183993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183993,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 183993, "pid": 5, "tid": 7, "ts": 1716454224342437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286790, "dur": 6,
    "args": {
      "External id": 183993,
      "cbid": 211, "correlation": 183993
    }
  },
  {
    "ph": "s", "id": 183993, "pid": 76337, "tid": -914061504, "ts": 1716454224286790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224342442, "dur": 6,
    "args": {
      "External id": 183996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 183996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 183996, "pid": 5, "tid": 7, "ts": 1716454224342442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286808, "dur": 7,
    "args": {
      "External id": 183996,
      "cbid": 211, "correlation": 183996
    }
  },
  {
    "ph": "s", "id": 183996, "pid": 76337, "tid": -914061504, "ts": 1716454224286808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286870, "dur": 0,
    "args": {
      "External id": 184007,
      "cbid": 317, "correlation": 184007
    }
  },
  {
    "ph": "f", "id": 184007, "pid": 76337, "tid": -914061504, "ts": 1716454224286870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286871, "dur": 0,
    "args": {
      "External id": 184008,
      "cbid": 203, "correlation": 184008
    }
  },
  {
    "ph": "f", "id": 184008, "pid": 76337, "tid": -914061504, "ts": 1716454224286871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286872, "dur": 0,
    "args": {
      "External id": 184009,
      "cbid": 205, "correlation": 184009
    }
  },
  {
    "ph": "f", "id": 184009, "pid": 76337, "tid": -914061504, "ts": 1716454224286872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224342450, "dur": 5,
    "args": {
      "External id": 184013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184013,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184013, "pid": 5, "tid": 7, "ts": 1716454224342450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286888, "dur": 12,
    "args": {
      "External id": 184013,
      "cbid": 211, "correlation": 184013
    }
  },
  {
    "ph": "s", "id": 184013, "pid": 76337, "tid": -914061504, "ts": 1716454224286888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224342456, "dur": 37,
    "args": {
      "External id": 184015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184015,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 184015, "pid": 5, "tid": 7, "ts": 1716454224342456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286907, "dur": 9,
    "args": {
      "External id": 184015,
      "cbid": 211, "correlation": 184015
    }
  },
  {
    "ph": "s", "id": 184015, "pid": 76337, "tid": -914061504, "ts": 1716454224286907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224342494, "dur": 5,
    "args": {
      "External id": 184017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184017,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184017, "pid": 5, "tid": 7, "ts": 1716454224342494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286920, "dur": 5,
    "args": {
      "External id": 184017,
      "cbid": 211, "correlation": 184017
    }
  },
  {
    "ph": "s", "id": 184017, "pid": 76337, "tid": -914061504, "ts": 1716454224286920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224342500, "dur": 6,
    "args": {
      "External id": 184023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184023,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184023, "pid": 5, "tid": 7, "ts": 1716454224342500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224286948, "dur": 8,
    "args": {
      "External id": 184023,
      "cbid": 211, "correlation": 184023
    }
  },
  {
    "ph": "s", "id": 184023, "pid": 76337, "tid": -914061504, "ts": 1716454224286948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224342507, "dur": 20,
    "args": {
      "External id": 184032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184032,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184032, "pid": 5, "tid": 7, "ts": 1716454224342507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287040, "dur": 14,
    "args": {
      "External id": 184032,
      "cbid": 211, "correlation": 184032
    }
  },
  {
    "ph": "s", "id": 184032, "pid": 76337, "tid": -914061504, "ts": 1716454224287040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224342529, "dur": 10,
    "args": {
      "External id": 184054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184054,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 184054, "pid": 5, "tid": 7, "ts": 1716454224342529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287098, "dur": 10,
    "args": {
      "External id": 184054,
      "cbid": 211, "correlation": 184054
    }
  },
  {
    "ph": "s", "id": 184054, "pid": 76337, "tid": -914061504, "ts": 1716454224287098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287188, "dur": 2,
    "args": {
      "External id": 184065,
      "cbid": 251, "correlation": 184065
    }
  },
  {
    "ph": "f", "id": 184065, "pid": 76337, "tid": -914061504, "ts": 1716454224287188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287192, "dur": 0,
    "args": {
      "External id": 184066,
      "cbid": 251, "correlation": 184066
    }
  },
  {
    "ph": "f", "id": 184066, "pid": 76337, "tid": -914061504, "ts": 1716454224287192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224342540, "dur": 53,
    "args": {
      "External id": 184067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184067,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 184067, "pid": 5, "tid": 7, "ts": 1716454224342540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287196, "dur": 14,
    "args": {
      "External id": 184067,
      "cbid": 211, "correlation": 184067
    }
  },
  {
    "ph": "s", "id": 184067, "pid": 76337, "tid": -914061504, "ts": 1716454224287196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287266, "dur": 1,
    "args": {
      "External id": 184078,
      "cbid": 251, "correlation": 184078
    }
  },
  {
    "ph": "f", "id": 184078, "pid": 76337, "tid": -914061504, "ts": 1716454224287266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287270, "dur": 0,
    "args": {
      "External id": 184079,
      "cbid": 251, "correlation": 184079
    }
  },
  {
    "ph": "f", "id": 184079, "pid": 76337, "tid": -914061504, "ts": 1716454224287270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224342594, "dur": 52,
    "args": {
      "External id": 184080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184080,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 184080, "pid": 5, "tid": 7, "ts": 1716454224342594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287271, "dur": 12,
    "args": {
      "External id": 184080,
      "cbid": 211, "correlation": 184080
    }
  },
  {
    "ph": "s", "id": 184080, "pid": 76337, "tid": -914061504, "ts": 1716454224287271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287338, "dur": 1,
    "args": {
      "External id": 184091,
      "cbid": 251, "correlation": 184091
    }
  },
  {
    "ph": "f", "id": 184091, "pid": 76337, "tid": -914061504, "ts": 1716454224287338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287342, "dur": 0,
    "args": {
      "External id": 184092,
      "cbid": 251, "correlation": 184092
    }
  },
  {
    "ph": "f", "id": 184092, "pid": 76337, "tid": -914061504, "ts": 1716454224287342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224342648, "dur": 53,
    "args": {
      "External id": 184093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184093,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 184093, "pid": 5, "tid": 7, "ts": 1716454224342648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287344, "dur": 11,
    "args": {
      "External id": 184093,
      "cbid": 211, "correlation": 184093
    }
  },
  {
    "ph": "s", "id": 184093, "pid": 76337, "tid": -914061504, "ts": 1716454224287344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224342701, "dur": 54,
    "args": {
      "External id": 184118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184118,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184118, "pid": 5, "tid": 7, "ts": 1716454224342701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287426, "dur": 13,
    "args": {
      "External id": 184118,
      "cbid": 211, "correlation": 184118
    }
  },
  {
    "ph": "s", "id": 184118, "pid": 76337, "tid": -914061504, "ts": 1716454224287426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287526, "dur": 1,
    "args": {
      "External id": 184136,
      "cbid": 251, "correlation": 184136
    }
  },
  {
    "ph": "f", "id": 184136, "pid": 76337, "tid": -914061504, "ts": 1716454224287526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224342757, "dur": 62,
    "args": {
      "External id": 184138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184138,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 184138, "pid": 5, "tid": 7, "ts": 1716454224342757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287532, "dur": 14,
    "args": {
      "External id": 184138,
      "cbid": 211, "correlation": 184138
    }
  },
  {
    "ph": "s", "id": 184138, "pid": 76337, "tid": -914061504, "ts": 1716454224287532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224342821, "dur": 6,
    "args": {
      "External id": 184146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184146,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184146, "pid": 5, "tid": 7, "ts": 1716454224342821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287602, "dur": 12,
    "args": {
      "External id": 184146,
      "cbid": 211, "correlation": 184146
    }
  },
  {
    "ph": "s", "id": 184146, "pid": 76337, "tid": -914061504, "ts": 1716454224287602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224342828, "dur": 7,
    "args": {
      "External id": 184154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184154,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184154, "pid": 5, "tid": 7, "ts": 1716454224342828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287641, "dur": 8,
    "args": {
      "External id": 184154,
      "cbid": 211, "correlation": 184154
    }
  },
  {
    "ph": "s", "id": 184154, "pid": 76337, "tid": -914061504, "ts": 1716454224287641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224342837, "dur": 8,
    "args": {
      "External id": 184165,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184165,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184165, "pid": 5, "tid": 7, "ts": 1716454224342837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287712, "dur": 12,
    "args": {
      "External id": 184165,
      "cbid": 211, "correlation": 184165
    }
  },
  {
    "ph": "s", "id": 184165, "pid": 76337, "tid": -914061504, "ts": 1716454224287712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224342846, "dur": 8,
    "args": {
      "External id": 184187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184187,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 184187, "pid": 5, "tid": 7, "ts": 1716454224342846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287745, "dur": 8,
    "args": {
      "External id": 184187,
      "cbid": 211, "correlation": 184187
    }
  },
  {
    "ph": "s", "id": 184187, "pid": 76337, "tid": -914061504, "ts": 1716454224287745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287830, "dur": 2,
    "args": {
      "External id": 184198,
      "cbid": 251, "correlation": 184198
    }
  },
  {
    "ph": "f", "id": 184198, "pid": 76337, "tid": -914061504, "ts": 1716454224287830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224342856, "dur": 1,
    "args": {
      "External id": 184199,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184199,
      "bytes": 480, "memory bandwidth (GB/s)": 0.3
    }
  },
  {
    "ph": "f", "id": 184199, "pid": 5, "tid": 7, "ts": 1716454224342856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287836, "dur": 12,
    "args": {
      "External id": 184199,
      "cbid": 51, "correlation": 184199
    }
  },
  {
    "ph": "s", "id": 184199, "pid": 76337, "tid": -914061504, "ts": 1716454224287836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224342860, "dur": 36,
    "args": {
      "External id": 184200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184200,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 184200, "pid": 5, "tid": 7, "ts": 1716454224342860,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287849, "dur": 12,
    "args": {
      "External id": 184200,
      "cbid": 211, "correlation": 184200
    }
  },
  {
    "ph": "s", "id": 184200, "pid": 76337, "tid": -914061504, "ts": 1716454224287849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287921, "dur": 1,
    "args": {
      "External id": 184211,
      "cbid": 251, "correlation": 184211
    }
  },
  {
    "ph": "f", "id": 184211, "pid": 76337, "tid": -914061504, "ts": 1716454224287921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287926, "dur": 0,
    "args": {
      "External id": 184212,
      "cbid": 251, "correlation": 184212
    }
  },
  {
    "ph": "f", "id": 184212, "pid": 76337, "tid": -914061504, "ts": 1716454224287926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224342897, "dur": 12,
    "args": {
      "External id": 184213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184213,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184213, "pid": 5, "tid": 7, "ts": 1716454224342897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287927, "dur": 12,
    "args": {
      "External id": 184213,
      "cbid": 211, "correlation": 184213
    }
  },
  {
    "ph": "s", "id": 184213, "pid": 76337, "tid": -914061504, "ts": 1716454224287927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224342910, "dur": 5,
    "args": {
      "External id": 184215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184215,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184215, "pid": 5, "tid": 7, "ts": 1716454224342910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224287941, "dur": 7,
    "args": {
      "External id": 184215,
      "cbid": 211, "correlation": 184215
    }
  },
  {
    "ph": "s", "id": 184215, "pid": 76337, "tid": -914061504, "ts": 1716454224287941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288008, "dur": 1,
    "args": {
      "External id": 184226,
      "cbid": 251, "correlation": 184226
    }
  },
  {
    "ph": "f", "id": 184226, "pid": 76337, "tid": -914061504, "ts": 1716454224288008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288011, "dur": 0,
    "args": {
      "External id": 184227,
      "cbid": 251, "correlation": 184227
    }
  },
  {
    "ph": "f", "id": 184227, "pid": 76337, "tid": -914061504, "ts": 1716454224288011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224342917, "dur": 8,
    "args": {
      "External id": 184228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184228,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184228, "pid": 5, "tid": 7, "ts": 1716454224342917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288013, "dur": 13,
    "args": {
      "External id": 184228,
      "cbid": 211, "correlation": 184228
    }
  },
  {
    "ph": "s", "id": 184228, "pid": 76337, "tid": -914061504, "ts": 1716454224288013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224342926, "dur": 3,
    "args": {
      "External id": 184230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184230,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184230, "pid": 5, "tid": 7, "ts": 1716454224342926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288027, "dur": 6,
    "args": {
      "External id": 184230,
      "cbid": 211, "correlation": 184230
    }
  },
  {
    "ph": "s", "id": 184230, "pid": 76337, "tid": -914061504, "ts": 1716454224288027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224342931, "dur": 20,
    "args": {
      "External id": 184255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184255,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 184255, "pid": 5, "tid": 7, "ts": 1716454224342931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288107, "dur": 12,
    "args": {
      "External id": 184255,
      "cbid": 211, "correlation": 184255
    }
  },
  {
    "ph": "s", "id": 184255, "pid": 76337, "tid": -914061504, "ts": 1716454224288107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288213, "dur": 2,
    "args": {
      "External id": 184273,
      "cbid": 251, "correlation": 184273
    }
  },
  {
    "ph": "f", "id": 184273, "pid": 76337, "tid": -914061504, "ts": 1716454224288213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224342952, "dur": 1,
    "args": {
      "External id": 184275,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184275,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 184275, "pid": 5, "tid": 7, "ts": 1716454224342952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288219, "dur": 10,
    "args": {
      "External id": 184275,
      "cbid": 51, "correlation": 184275
    }
  },
  {
    "ph": "s", "id": 184275, "pid": 76337, "tid": -914061504, "ts": 1716454224288219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224342956, "dur": 36,
    "args": {
      "External id": 184276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184276,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 184276, "pid": 5, "tid": 7, "ts": 1716454224342956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288231, "dur": 12,
    "args": {
      "External id": 184276,
      "cbid": 211, "correlation": 184276
    }
  },
  {
    "ph": "s", "id": 184276, "pid": 76337, "tid": -914061504, "ts": 1716454224288231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224342993, "dur": 4,
    "args": {
      "External id": 184284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184284,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184284, "pid": 5, "tid": 7, "ts": 1716454224342993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288302, "dur": 12,
    "args": {
      "External id": 184284,
      "cbid": 211, "correlation": 184284
    }
  },
  {
    "ph": "s", "id": 184284, "pid": 76337, "tid": -914061504, "ts": 1716454224288302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224342998, "dur": 8,
    "args": {
      "External id": 184292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184292, "pid": 5, "tid": 7, "ts": 1716454224342998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288344, "dur": 10,
    "args": {
      "External id": 184292,
      "cbid": 211, "correlation": 184292
    }
  },
  {
    "ph": "s", "id": 184292, "pid": 76337, "tid": -914061504, "ts": 1716454224288344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343008, "dur": 8,
    "args": {
      "External id": 184314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184314,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 184314, "pid": 5, "tid": 7, "ts": 1716454224343008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288396, "dur": 11,
    "args": {
      "External id": 184314,
      "cbid": 211, "correlation": 184314
    }
  },
  {
    "ph": "s", "id": 184314, "pid": 76337, "tid": -914061504, "ts": 1716454224288396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288488, "dur": 1,
    "args": {
      "External id": 184330,
      "cbid": 251, "correlation": 184330
    }
  },
  {
    "ph": "f", "id": 184330, "pid": 76337, "tid": -914061504, "ts": 1716454224288488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288493, "dur": 0,
    "args": {
      "External id": 184332,
      "cbid": 251, "correlation": 184332
    }
  },
  {
    "ph": "f", "id": 184332, "pid": 76337, "tid": -914061504, "ts": 1716454224288493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224343017, "dur": 188,
    "args": {
      "External id": 184333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184333,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184333, "pid": 5, "tid": 7, "ts": 1716454224343017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288495, "dur": 13,
    "args": {
      "External id": 184333,
      "cbid": 211, "correlation": 184333
    }
  },
  {
    "ph": "s", "id": 184333, "pid": 76337, "tid": -914061504, "ts": 1716454224288495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343207, "dur": 21,
    "args": {
      "External id": 184341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184341,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184341, "pid": 5, "tid": 7, "ts": 1716454224343207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288562, "dur": 12,
    "args": {
      "External id": 184341,
      "cbid": 211, "correlation": 184341
    }
  },
  {
    "ph": "s", "id": 184341, "pid": 76337, "tid": -914061504, "ts": 1716454224288562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343229, "dur": 21,
    "args": {
      "External id": 184349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184349, "pid": 5, "tid": 7, "ts": 1716454224343229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288592, "dur": 8,
    "args": {
      "External id": 184349,
      "cbid": 211, "correlation": 184349
    }
  },
  {
    "ph": "s", "id": 184349, "pid": 76337, "tid": -914061504, "ts": 1716454224288592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288674, "dur": 1,
    "args": {
      "External id": 184365,
      "cbid": 251, "correlation": 184365
    }
  },
  {
    "ph": "f", "id": 184365, "pid": 76337, "tid": -914061504, "ts": 1716454224288674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224343252, "dur": 1,
    "args": {
      "External id": 184367,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184367,
      "bytes": 120, "memory bandwidth (GB/s)": 0.0797872340425532
    }
  },
  {
    "ph": "f", "id": 184367, "pid": 5, "tid": 7, "ts": 1716454224343252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288679, "dur": 9,
    "args": {
      "External id": 184367,
      "cbid": 51, "correlation": 184367
    }
  },
  {
    "ph": "s", "id": 184367, "pid": 76337, "tid": -914061504, "ts": 1716454224288679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224343256, "dur": 109,
    "args": {
      "External id": 184368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184368,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 184368, "pid": 5, "tid": 7, "ts": 1716454224343256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288690, "dur": 12,
    "args": {
      "External id": 184368,
      "cbid": 211, "correlation": 184368
    }
  },
  {
    "ph": "s", "id": 184368, "pid": 76337, "tid": -914061504, "ts": 1716454224288690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224343366, "dur": 5,
    "args": {
      "External id": 184376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184376,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184376, "pid": 5, "tid": 7, "ts": 1716454224343366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288732, "dur": 10,
    "args": {
      "External id": 184376,
      "cbid": 211, "correlation": 184376
    }
  },
  {
    "ph": "s", "id": 184376, "pid": 76337, "tid": -914061504, "ts": 1716454224288732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343373, "dur": 10,
    "args": {
      "External id": 184387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184387, "pid": 5, "tid": 7, "ts": 1716454224343373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288799, "dur": 12,
    "args": {
      "External id": 184387,
      "cbid": 211, "correlation": 184387
    }
  },
  {
    "ph": "s", "id": 184387, "pid": 76337, "tid": -914061504, "ts": 1716454224288799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288866, "dur": 0,
    "args": {
      "External id": 184399,
      "cbid": 317, "correlation": 184399
    }
  },
  {
    "ph": "f", "id": 184399, "pid": 76337, "tid": -914061504, "ts": 1716454224288866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288867, "dur": 0,
    "args": {
      "External id": 184400,
      "cbid": 203, "correlation": 184400
    }
  },
  {
    "ph": "f", "id": 184400, "pid": 76337, "tid": -914061504, "ts": 1716454224288867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288868, "dur": 0,
    "args": {
      "External id": 184401,
      "cbid": 205, "correlation": 184401
    }
  },
  {
    "ph": "f", "id": 184401, "pid": 76337, "tid": -914061504, "ts": 1716454224288868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343384, "dur": 6,
    "args": {
      "External id": 184405,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184405,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184405, "pid": 5, "tid": 7, "ts": 1716454224343384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288884, "dur": 12,
    "args": {
      "External id": 184405,
      "cbid": 211, "correlation": 184405
    }
  },
  {
    "ph": "s", "id": 184405, "pid": 76337, "tid": -914061504, "ts": 1716454224288884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224343391, "dur": 37,
    "args": {
      "External id": 184407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184407,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 184407, "pid": 5, "tid": 7, "ts": 1716454224343391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288903, "dur": 7,
    "args": {
      "External id": 184407,
      "cbid": 211, "correlation": 184407
    }
  },
  {
    "ph": "s", "id": 184407, "pid": 76337, "tid": -914061504, "ts": 1716454224288903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343429, "dur": 6,
    "args": {
      "External id": 184409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184409,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184409, "pid": 5, "tid": 7, "ts": 1716454224343429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288913, "dur": 6,
    "args": {
      "External id": 184409,
      "cbid": 211, "correlation": 184409
    }
  },
  {
    "ph": "s", "id": 184409, "pid": 76337, "tid": -914061504, "ts": 1716454224288913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343436, "dur": 7,
    "args": {
      "External id": 184415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184415, "pid": 5, "tid": 7, "ts": 1716454224343436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288941, "dur": 8,
    "args": {
      "External id": 184415,
      "cbid": 211, "correlation": 184415
    }
  },
  {
    "ph": "s", "id": 184415, "pid": 76337, "tid": -914061504, "ts": 1716454224288941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224343444, "dur": 5,
    "args": {
      "External id": 184423,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184423,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184423, "pid": 5, "tid": 7, "ts": 1716454224343444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224288981, "dur": 9,
    "args": {
      "External id": 184423,
      "cbid": 211, "correlation": 184423
    }
  },
  {
    "ph": "s", "id": 184423, "pid": 76337, "tid": -914061504, "ts": 1716454224288981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343450, "dur": 11,
    "args": {
      "External id": 184443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184443,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 184443, "pid": 5, "tid": 7, "ts": 1716454224343450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289056, "dur": 12,
    "args": {
      "External id": 184443,
      "cbid": 211, "correlation": 184443
    }
  },
  {
    "ph": "s", "id": 184443, "pid": 76337, "tid": -914061504, "ts": 1716454224289056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224343462, "dur": 4,
    "args": {
      "External id": 184455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184455,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 184455, "pid": 5, "tid": 7, "ts": 1716454224343462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289078, "dur": 6,
    "args": {
      "External id": 184455,
      "cbid": 211, "correlation": 184455
    }
  },
  {
    "ph": "s", "id": 184455, "pid": 76337, "tid": -914061504, "ts": 1716454224289078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343468, "dur": 8,
    "args": {
      "External id": 184458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184458,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184458, "pid": 5, "tid": 7, "ts": 1716454224343468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289096, "dur": 6,
    "args": {
      "External id": 184458,
      "cbid": 211, "correlation": 184458
    }
  },
  {
    "ph": "s", "id": 184458, "pid": 76337, "tid": -914061504, "ts": 1716454224289096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224343477, "dur": 5,
    "args": {
      "External id": 184467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184467,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184467, "pid": 5, "tid": 7, "ts": 1716454224343477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289135, "dur": 10,
    "args": {
      "External id": 184467,
      "cbid": 211, "correlation": 184467
    }
  },
  {
    "ph": "s", "id": 184467, "pid": 76337, "tid": -914061504, "ts": 1716454224289135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289191, "dur": 0,
    "args": {
      "External id": 184477,
      "cbid": 317, "correlation": 184477
    }
  },
  {
    "ph": "f", "id": 184477, "pid": 76337, "tid": -914061504, "ts": 1716454224289191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289191, "dur": 0,
    "args": {
      "External id": 184478,
      "cbid": 203, "correlation": 184478
    }
  },
  {
    "ph": "f", "id": 184478, "pid": 76337, "tid": -914061504, "ts": 1716454224289191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289192, "dur": 0,
    "args": {
      "External id": 184479,
      "cbid": 205, "correlation": 184479
    }
  },
  {
    "ph": "f", "id": 184479, "pid": 76337, "tid": -914061504, "ts": 1716454224289192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343484, "dur": 5,
    "args": {
      "External id": 184483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184483,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184483, "pid": 5, "tid": 7, "ts": 1716454224343484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289206, "dur": 12,
    "args": {
      "External id": 184483,
      "cbid": 211, "correlation": 184483
    }
  },
  {
    "ph": "s", "id": 184483, "pid": 76337, "tid": -914061504, "ts": 1716454224289206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343490, "dur": 160,
    "args": {
      "External id": 184485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184485,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184485, "pid": 5, "tid": 7, "ts": 1716454224343490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289220, "dur": 6,
    "args": {
      "External id": 184485,
      "cbid": 211, "correlation": 184485
    }
  },
  {
    "ph": "s", "id": 184485, "pid": 76337, "tid": -914061504, "ts": 1716454224289220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224343652, "dur": 1,
    "args": {
      "External id": 184487,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184487,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 184487, "pid": 5, "tid": 7, "ts": 1716454224343652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289232, "dur": 6,
    "args": {
      "External id": 184487,
      "cbid": 51, "correlation": 184487
    }
  },
  {
    "ph": "s", "id": 184487, "pid": 76337, "tid": -914061504, "ts": 1716454224289232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224343655, "dur": 265,
    "args": {
      "External id": 184488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184488,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184488, "pid": 5, "tid": 7, "ts": 1716454224343655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289239, "dur": 6,
    "args": {
      "External id": 184488,
      "cbid": 211, "correlation": 184488
    }
  },
  {
    "ph": "s", "id": 184488, "pid": 76337, "tid": -914061504, "ts": 1716454224289239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343921, "dur": 5,
    "args": {
      "External id": 184490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184490,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184490, "pid": 5, "tid": 7, "ts": 1716454224343921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289249, "dur": 5,
    "args": {
      "External id": 184490,
      "cbid": 211, "correlation": 184490
    }
  },
  {
    "ph": "s", "id": 184490, "pid": 76337, "tid": -914061504, "ts": 1716454224289249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343928, "dur": 6,
    "args": {
      "External id": 184496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184496, "pid": 5, "tid": 7, "ts": 1716454224343928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289277, "dur": 9,
    "args": {
      "External id": 184496,
      "cbid": 211, "correlation": 184496
    }
  },
  {
    "ph": "s", "id": 184496, "pid": 76337, "tid": -914061504, "ts": 1716454224289277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224343936, "dur": 3,
    "args": {
      "External id": 184504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184504,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 184504, "pid": 5, "tid": 7, "ts": 1716454224343936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289322, "dur": 10,
    "args": {
      "External id": 184504,
      "cbid": 211, "correlation": 184504
    }
  },
  {
    "ph": "s", "id": 184504, "pid": 76337, "tid": -914061504, "ts": 1716454224289322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289389, "dur": 1,
    "args": {
      "External id": 184520,
      "cbid": 251, "correlation": 184520
    }
  },
  {
    "ph": "f", "id": 184520, "pid": 76337, "tid": -914061504, "ts": 1716454224289389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289394, "dur": 0,
    "args": {
      "External id": 184522,
      "cbid": 251, "correlation": 184522
    }
  },
  {
    "ph": "f", "id": 184522, "pid": 76337, "tid": -914061504, "ts": 1716454224289394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224343940, "dur": 12,
    "args": {
      "External id": 184523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184523,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184523, "pid": 5, "tid": 7, "ts": 1716454224343940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289395, "dur": 11,
    "args": {
      "External id": 184523,
      "cbid": 211, "correlation": 184523
    }
  },
  {
    "ph": "s", "id": 184523, "pid": 76337, "tid": -914061504, "ts": 1716454224289395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224343954, "dur": 5,
    "args": {
      "External id": 184525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184525,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184525, "pid": 5, "tid": 7, "ts": 1716454224343954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289408, "dur": 5,
    "args": {
      "External id": 184525,
      "cbid": 211, "correlation": 184525
    }
  },
  {
    "ph": "s", "id": 184525, "pid": 76337, "tid": -914061504, "ts": 1716454224289408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343960, "dur": 6,
    "args": {
      "External id": 184535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184535, "pid": 5, "tid": 7, "ts": 1716454224343960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289467, "dur": 13,
    "args": {
      "External id": 184535,
      "cbid": 211, "correlation": 184535
    }
  },
  {
    "ph": "s", "id": 184535, "pid": 76337, "tid": -914061504, "ts": 1716454224289467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343967, "dur": 10,
    "args": {
      "External id": 184555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184555,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 184555, "pid": 5, "tid": 7, "ts": 1716454224343967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289535, "dur": 11,
    "args": {
      "External id": 184555,
      "cbid": 211, "correlation": 184555
    }
  },
  {
    "ph": "s", "id": 184555, "pid": 76337, "tid": -914061504, "ts": 1716454224289535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224343978, "dur": 4,
    "args": {
      "External id": 184567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184567,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 184567, "pid": 5, "tid": 7, "ts": 1716454224343978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289555, "dur": 6,
    "args": {
      "External id": 184567,
      "cbid": 211, "correlation": 184567
    }
  },
  {
    "ph": "s", "id": 184567, "pid": 76337, "tid": -914061504, "ts": 1716454224289555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224343983, "dur": 7,
    "args": {
      "External id": 184570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184570, "pid": 5, "tid": 7, "ts": 1716454224343983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289574, "dur": 7,
    "args": {
      "External id": 184570,
      "cbid": 211, "correlation": 184570
    }
  },
  {
    "ph": "s", "id": 184570, "pid": 76337, "tid": -914061504, "ts": 1716454224289574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224343991, "dur": 4,
    "args": {
      "External id": 184579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184579,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184579, "pid": 5, "tid": 7, "ts": 1716454224343991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289615, "dur": 9,
    "args": {
      "External id": 184579,
      "cbid": 211, "correlation": 184579
    }
  },
  {
    "ph": "s", "id": 184579, "pid": 76337, "tid": -914061504, "ts": 1716454224289615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289677, "dur": 0,
    "args": {
      "External id": 184589,
      "cbid": 317, "correlation": 184589
    }
  },
  {
    "ph": "f", "id": 184589, "pid": 76337, "tid": -914061504, "ts": 1716454224289677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289678, "dur": 0,
    "args": {
      "External id": 184590,
      "cbid": 203, "correlation": 184590
    }
  },
  {
    "ph": "f", "id": 184590, "pid": 76337, "tid": -914061504, "ts": 1716454224289678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289678, "dur": 0,
    "args": {
      "External id": 184591,
      "cbid": 205, "correlation": 184591
    }
  },
  {
    "ph": "f", "id": 184591, "pid": 76337, "tid": -914061504, "ts": 1716454224289678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224343997, "dur": 5,
    "args": {
      "External id": 184595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184595,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184595, "pid": 5, "tid": 7, "ts": 1716454224343997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289693, "dur": 12,
    "args": {
      "External id": 184595,
      "cbid": 211, "correlation": 184595
    }
  },
  {
    "ph": "s", "id": 184595, "pid": 76337, "tid": -914061504, "ts": 1716454224289693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224344003, "dur": 159,
    "args": {
      "External id": 184597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184597,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184597, "pid": 5, "tid": 7, "ts": 1716454224344003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289707, "dur": 5,
    "args": {
      "External id": 184597,
      "cbid": 211, "correlation": 184597
    }
  },
  {
    "ph": "s", "id": 184597, "pid": 76337, "tid": -914061504, "ts": 1716454224289707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224344164, "dur": 1,
    "args": {
      "External id": 184599,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184599,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 184599, "pid": 5, "tid": 7, "ts": 1716454224344164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289718, "dur": 6,
    "args": {
      "External id": 184599,
      "cbid": 51, "correlation": 184599
    }
  },
  {
    "ph": "s", "id": 184599, "pid": 76337, "tid": -914061504, "ts": 1716454224289718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224344168, "dur": 255,
    "args": {
      "External id": 184600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184600,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184600, "pid": 5, "tid": 7, "ts": 1716454224344168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289725, "dur": 6,
    "args": {
      "External id": 184600,
      "cbid": 211, "correlation": 184600
    }
  },
  {
    "ph": "s", "id": 184600, "pid": 76337, "tid": -914061504, "ts": 1716454224289725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224344424, "dur": 6,
    "args": {
      "External id": 184602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184602,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184602, "pid": 5, "tid": 7, "ts": 1716454224344424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289735, "dur": 6,
    "args": {
      "External id": 184602,
      "cbid": 211, "correlation": 184602
    }
  },
  {
    "ph": "s", "id": 184602, "pid": 76337, "tid": -914061504, "ts": 1716454224289735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224344431, "dur": 6,
    "args": {
      "External id": 184608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184608, "pid": 5, "tid": 7, "ts": 1716454224344431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289764, "dur": 8,
    "args": {
      "External id": 184608,
      "cbid": 211, "correlation": 184608
    }
  },
  {
    "ph": "s", "id": 184608, "pid": 76337, "tid": -914061504, "ts": 1716454224289764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224344439, "dur": 5,
    "args": {
      "External id": 184616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184616,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184616, "pid": 5, "tid": 7, "ts": 1716454224344439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289797, "dur": 8,
    "args": {
      "External id": 184616,
      "cbid": 211, "correlation": 184616
    }
  },
  {
    "ph": "s", "id": 184616, "pid": 76337, "tid": -914061504, "ts": 1716454224289797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224344445, "dur": 4,
    "args": {
      "External id": 184624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184624,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184624, "pid": 5, "tid": 7, "ts": 1716454224344445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289826, "dur": 8,
    "args": {
      "External id": 184624,
      "cbid": 211, "correlation": 184624
    }
  },
  {
    "ph": "s", "id": 184624, "pid": 76337, "tid": -914061504, "ts": 1716454224289826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224344451, "dur": 12,
    "args": {
      "External id": 184633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184633,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184633, "pid": 5, "tid": 7, "ts": 1716454224344451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224289926, "dur": 14,
    "args": {
      "External id": 184633,
      "cbid": 211, "correlation": 184633
    }
  },
  {
    "ph": "s", "id": 184633, "pid": 76337, "tid": -914061504, "ts": 1716454224289926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224344464, "dur": 12,
    "args": {
      "External id": 184653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184653,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 184653, "pid": 5, "tid": 7, "ts": 1716454224344464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290014, "dur": 12,
    "args": {
      "External id": 184653,
      "cbid": 211, "correlation": 184653
    }
  },
  {
    "ph": "s", "id": 184653, "pid": 76337, "tid": -914061504, "ts": 1716454224290014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224344477, "dur": 4,
    "args": {
      "External id": 184665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184665,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184665, "pid": 5, "tid": 7, "ts": 1716454224344477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290036, "dur": 7,
    "args": {
      "External id": 184665,
      "cbid": 211, "correlation": 184665
    }
  },
  {
    "ph": "s", "id": 184665, "pid": 76337, "tid": -914061504, "ts": 1716454224290036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224344482, "dur": 10,
    "args": {
      "External id": 184668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184668,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184668, "pid": 5, "tid": 7, "ts": 1716454224344482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290056, "dur": 6,
    "args": {
      "External id": 184668,
      "cbid": 211, "correlation": 184668
    }
  },
  {
    "ph": "s", "id": 184668, "pid": 76337, "tid": -914061504, "ts": 1716454224290056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224344493, "dur": 6,
    "args": {
      "External id": 184677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184677,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184677, "pid": 5, "tid": 7, "ts": 1716454224344493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290095, "dur": 10,
    "args": {
      "External id": 184677,
      "cbid": 211, "correlation": 184677
    }
  },
  {
    "ph": "s", "id": 184677, "pid": 76337, "tid": -914061504, "ts": 1716454224290095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290152, "dur": 0,
    "args": {
      "External id": 184687,
      "cbid": 317, "correlation": 184687
    }
  },
  {
    "ph": "f", "id": 184687, "pid": 76337, "tid": -914061504, "ts": 1716454224290152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290153, "dur": 0,
    "args": {
      "External id": 184688,
      "cbid": 203, "correlation": 184688
    }
  },
  {
    "ph": "f", "id": 184688, "pid": 76337, "tid": -914061504, "ts": 1716454224290153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290153, "dur": 0,
    "args": {
      "External id": 184689,
      "cbid": 205, "correlation": 184689
    }
  },
  {
    "ph": "f", "id": 184689, "pid": 76337, "tid": -914061504, "ts": 1716454224290153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224344501, "dur": 6,
    "args": {
      "External id": 184693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184693,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184693, "pid": 5, "tid": 7, "ts": 1716454224344501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290169, "dur": 11,
    "args": {
      "External id": 184693,
      "cbid": 211, "correlation": 184693
    }
  },
  {
    "ph": "s", "id": 184693, "pid": 76337, "tid": -914061504, "ts": 1716454224290169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224344509, "dur": 314,
    "args": {
      "External id": 184695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184695,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184695, "pid": 5, "tid": 7, "ts": 1716454224344509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290183, "dur": 5,
    "args": {
      "External id": 184695,
      "cbid": 211, "correlation": 184695
    }
  },
  {
    "ph": "s", "id": 184695, "pid": 76337, "tid": -914061504, "ts": 1716454224290183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224344825, "dur": 1,
    "args": {
      "External id": 184697,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184697,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 184697, "pid": 5, "tid": 7, "ts": 1716454224344825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290194, "dur": 7,
    "args": {
      "External id": 184697,
      "cbid": 51, "correlation": 184697
    }
  },
  {
    "ph": "s", "id": 184697, "pid": 76337, "tid": -914061504, "ts": 1716454224290194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224344829, "dur": 488,
    "args": {
      "External id": 184698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184698,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184698, "pid": 5, "tid": 7, "ts": 1716454224344829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290202, "dur": 6,
    "args": {
      "External id": 184698,
      "cbid": 211, "correlation": 184698
    }
  },
  {
    "ph": "s", "id": 184698, "pid": 76337, "tid": -914061504, "ts": 1716454224290202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345318, "dur": 6,
    "args": {
      "External id": 184700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184700, "pid": 5, "tid": 7, "ts": 1716454224345318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290212, "dur": 5,
    "args": {
      "External id": 184700,
      "cbid": 211, "correlation": 184700
    }
  },
  {
    "ph": "s", "id": 184700, "pid": 76337, "tid": -914061504, "ts": 1716454224290212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224345325, "dur": 6,
    "args": {
      "External id": 184706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184706, "pid": 5, "tid": 7, "ts": 1716454224345325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290241, "dur": 10,
    "args": {
      "External id": 184706,
      "cbid": 211, "correlation": 184706
    }
  },
  {
    "ph": "s", "id": 184706, "pid": 76337, "tid": -914061504, "ts": 1716454224290241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224345332, "dur": 3,
    "args": {
      "External id": 184714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184714,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 184714, "pid": 5, "tid": 7, "ts": 1716454224345332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290285, "dur": 9,
    "args": {
      "External id": 184714,
      "cbid": 211, "correlation": 184714
    }
  },
  {
    "ph": "s", "id": 184714, "pid": 76337, "tid": -914061504, "ts": 1716454224290285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290350, "dur": 1,
    "args": {
      "External id": 184730,
      "cbid": 251, "correlation": 184730
    }
  },
  {
    "ph": "f", "id": 184730, "pid": 76337, "tid": -914061504, "ts": 1716454224290350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290355, "dur": 0,
    "args": {
      "External id": 184732,
      "cbid": 251, "correlation": 184732
    }
  },
  {
    "ph": "f", "id": 184732, "pid": 76337, "tid": -914061504, "ts": 1716454224290355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224345336, "dur": 11,
    "args": {
      "External id": 184733,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184733,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184733, "pid": 5, "tid": 7, "ts": 1716454224345336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290357, "dur": 11,
    "args": {
      "External id": 184733,
      "cbid": 211, "correlation": 184733
    }
  },
  {
    "ph": "s", "id": 184733, "pid": 76337, "tid": -914061504, "ts": 1716454224290357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224345349, "dur": 4,
    "args": {
      "External id": 184735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184735,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184735, "pid": 5, "tid": 7, "ts": 1716454224345349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290370, "dur": 5,
    "args": {
      "External id": 184735,
      "cbid": 211, "correlation": 184735
    }
  },
  {
    "ph": "s", "id": 184735, "pid": 76337, "tid": -914061504, "ts": 1716454224290370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224345355, "dur": 6,
    "args": {
      "External id": 184745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184745,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184745, "pid": 5, "tid": 7, "ts": 1716454224345355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290426, "dur": 12,
    "args": {
      "External id": 184745,
      "cbid": 211, "correlation": 184745
    }
  },
  {
    "ph": "s", "id": 184745, "pid": 76337, "tid": -914061504, "ts": 1716454224290426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345361, "dur": 9,
    "args": {
      "External id": 184765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184765,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 184765, "pid": 5, "tid": 7, "ts": 1716454224345361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290492, "dur": 12,
    "args": {
      "External id": 184765,
      "cbid": 211, "correlation": 184765
    }
  },
  {
    "ph": "s", "id": 184765, "pid": 76337, "tid": -914061504, "ts": 1716454224290492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224345372, "dur": 4,
    "args": {
      "External id": 184777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184777,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 184777, "pid": 5, "tid": 7, "ts": 1716454224345372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290514, "dur": 6,
    "args": {
      "External id": 184777,
      "cbid": 211, "correlation": 184777
    }
  },
  {
    "ph": "s", "id": 184777, "pid": 76337, "tid": -914061504, "ts": 1716454224290514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224345377, "dur": 7,
    "args": {
      "External id": 184780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184780,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184780, "pid": 5, "tid": 7, "ts": 1716454224345377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290533, "dur": 6,
    "args": {
      "External id": 184780,
      "cbid": 211, "correlation": 184780
    }
  },
  {
    "ph": "s", "id": 184780, "pid": 76337, "tid": -914061504, "ts": 1716454224290533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224345385, "dur": 4,
    "args": {
      "External id": 184789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184789,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184789, "pid": 5, "tid": 7, "ts": 1716454224345385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290574, "dur": 9,
    "args": {
      "External id": 184789,
      "cbid": 211, "correlation": 184789
    }
  },
  {
    "ph": "s", "id": 184789, "pid": 76337, "tid": -914061504, "ts": 1716454224290574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290636, "dur": 0,
    "args": {
      "External id": 184799,
      "cbid": 317, "correlation": 184799
    }
  },
  {
    "ph": "f", "id": 184799, "pid": 76337, "tid": -914061504, "ts": 1716454224290636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290637, "dur": 0,
    "args": {
      "External id": 184800,
      "cbid": 203, "correlation": 184800
    }
  },
  {
    "ph": "f", "id": 184800, "pid": 76337, "tid": -914061504, "ts": 1716454224290637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290637, "dur": 0,
    "args": {
      "External id": 184801,
      "cbid": 205, "correlation": 184801
    }
  },
  {
    "ph": "f", "id": 184801, "pid": 76337, "tid": -914061504, "ts": 1716454224290637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345391, "dur": 5,
    "args": {
      "External id": 184805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184805,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184805, "pid": 5, "tid": 7, "ts": 1716454224345391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290651, "dur": 13,
    "args": {
      "External id": 184805,
      "cbid": 211, "correlation": 184805
    }
  },
  {
    "ph": "s", "id": 184805, "pid": 76337, "tid": -914061504, "ts": 1716454224290651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345397, "dur": 159,
    "args": {
      "External id": 184807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184807,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184807, "pid": 5, "tid": 7, "ts": 1716454224345397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290666, "dur": 5,
    "args": {
      "External id": 184807,
      "cbid": 211, "correlation": 184807
    }
  },
  {
    "ph": "s", "id": 184807, "pid": 76337, "tid": -914061504, "ts": 1716454224290666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224345558, "dur": 1,
    "args": {
      "External id": 184809,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184809,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 184809, "pid": 5, "tid": 7, "ts": 1716454224345558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290677, "dur": 7,
    "args": {
      "External id": 184809,
      "cbid": 51, "correlation": 184809
    }
  },
  {
    "ph": "s", "id": 184809, "pid": 76337, "tid": -914061504, "ts": 1716454224290677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224345562, "dur": 254,
    "args": {
      "External id": 184810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184810,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184810, "pid": 5, "tid": 7, "ts": 1716454224345562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290686, "dur": 6,
    "args": {
      "External id": 184810,
      "cbid": 211, "correlation": 184810
    }
  },
  {
    "ph": "s", "id": 184810, "pid": 76337, "tid": -914061504, "ts": 1716454224290686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345817, "dur": 6,
    "args": {
      "External id": 184812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184812, "pid": 5, "tid": 7, "ts": 1716454224345817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290695, "dur": 5,
    "args": {
      "External id": 184812,
      "cbid": 211, "correlation": 184812
    }
  },
  {
    "ph": "s", "id": 184812, "pid": 76337, "tid": -914061504, "ts": 1716454224290695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224345824, "dur": 6,
    "args": {
      "External id": 184818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184818, "pid": 5, "tid": 7, "ts": 1716454224345824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290724, "dur": 8,
    "args": {
      "External id": 184818,
      "cbid": 211, "correlation": 184818
    }
  },
  {
    "ph": "s", "id": 184818, "pid": 76337, "tid": -914061504, "ts": 1716454224290724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290782, "dur": 0,
    "args": {
      "External id": 184828,
      "cbid": 317, "correlation": 184828
    }
  },
  {
    "ph": "f", "id": 184828, "pid": 76337, "tid": -914061504, "ts": 1716454224290782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290782, "dur": 0,
    "args": {
      "External id": 184829,
      "cbid": 203, "correlation": 184829
    }
  },
  {
    "ph": "f", "id": 184829, "pid": 76337, "tid": -914061504, "ts": 1716454224290782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290783, "dur": 0,
    "args": {
      "External id": 184830,
      "cbid": 205, "correlation": 184830
    }
  },
  {
    "ph": "f", "id": 184830, "pid": 76337, "tid": -914061504, "ts": 1716454224290783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345832, "dur": 7,
    "args": {
      "External id": 184834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184834,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184834, "pid": 5, "tid": 7, "ts": 1716454224345832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290797, "dur": 12,
    "args": {
      "External id": 184834,
      "cbid": 211, "correlation": 184834
    }
  },
  {
    "ph": "s", "id": 184834, "pid": 76337, "tid": -914061504, "ts": 1716454224290797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224345840, "dur": 3,
    "args": {
      "External id": 184836,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184836,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4150943396226414
    }
  },
  {
    "ph": "f", "id": 184836, "pid": 5, "tid": 7, "ts": 1716454224345840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290815, "dur": 14,
    "args": {
      "External id": 184836,
      "cbid": 51, "correlation": 184836
    }
  },
  {
    "ph": "s", "id": 184836, "pid": 76337, "tid": -914061504, "ts": 1716454224290815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224345844, "dur": 96,
    "args": {
      "External id": 184837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184837,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 184837, "pid": 5, "tid": 7, "ts": 1716454224345844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290830, "dur": 7,
    "args": {
      "External id": 184837,
      "cbid": 211, "correlation": 184837
    }
  },
  {
    "ph": "s", "id": 184837, "pid": 76337, "tid": -914061504, "ts": 1716454224290830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345942, "dur": 5,
    "args": {
      "External id": 184839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184839,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184839, "pid": 5, "tid": 7, "ts": 1716454224345942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290841, "dur": 5,
    "args": {
      "External id": 184839,
      "cbid": 211, "correlation": 184839
    }
  },
  {
    "ph": "s", "id": 184839, "pid": 76337, "tid": -914061504, "ts": 1716454224290841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224345948, "dur": 6,
    "args": {
      "External id": 184845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184845,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184845, "pid": 5, "tid": 7, "ts": 1716454224345948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290869, "dur": 8,
    "args": {
      "External id": 184845,
      "cbid": 211, "correlation": 184845
    }
  },
  {
    "ph": "s", "id": 184845, "pid": 76337, "tid": -914061504, "ts": 1716454224290869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224345956, "dur": 5,
    "args": {
      "External id": 184853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184853,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184853, "pid": 5, "tid": 7, "ts": 1716454224345956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290897, "dur": 8,
    "args": {
      "External id": 184853,
      "cbid": 211, "correlation": 184853
    }
  },
  {
    "ph": "s", "id": 184853, "pid": 76337, "tid": -914061504, "ts": 1716454224290897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224345962, "dur": 4,
    "args": {
      "External id": 184861,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184861,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184861, "pid": 5, "tid": 7, "ts": 1716454224345962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290926, "dur": 8,
    "args": {
      "External id": 184861,
      "cbid": 211, "correlation": 184861
    }
  },
  {
    "ph": "s", "id": 184861, "pid": 76337, "tid": -914061504, "ts": 1716454224290926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224345967, "dur": 11,
    "args": {
      "External id": 184870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184870,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184870, "pid": 5, "tid": 7, "ts": 1716454224345967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224290970, "dur": 45,
    "args": {
      "External id": 184870,
      "cbid": 211, "correlation": 184870
    }
  },
  {
    "ph": "s", "id": 184870, "pid": 76337, "tid": -914061504, "ts": 1716454224290970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224345980, "dur": 12,
    "args": {
      "External id": 184890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184890,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 184890, "pid": 5, "tid": 7, "ts": 1716454224345980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291078, "dur": 11,
    "args": {
      "External id": 184890,
      "cbid": 211, "correlation": 184890
    }
  },
  {
    "ph": "s", "id": 184890, "pid": 76337, "tid": -914061504, "ts": 1716454224291078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224345993, "dur": 4,
    "args": {
      "External id": 184902,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184902,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184902, "pid": 5, "tid": 7, "ts": 1716454224345993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291100, "dur": 6,
    "args": {
      "External id": 184902,
      "cbid": 211, "correlation": 184902
    }
  },
  {
    "ph": "s", "id": 184902, "pid": 76337, "tid": -914061504, "ts": 1716454224291100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224345999, "dur": 10,
    "args": {
      "External id": 184905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184905,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184905, "pid": 5, "tid": 7, "ts": 1716454224345999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291118, "dur": 6,
    "args": {
      "External id": 184905,
      "cbid": 211, "correlation": 184905
    }
  },
  {
    "ph": "s", "id": 184905, "pid": 76337, "tid": -914061504, "ts": 1716454224291118,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224346010, "dur": 6,
    "args": {
      "External id": 184914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184914,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184914, "pid": 5, "tid": 7, "ts": 1716454224346010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291155, "dur": 10,
    "args": {
      "External id": 184914,
      "cbid": 211, "correlation": 184914
    }
  },
  {
    "ph": "s", "id": 184914, "pid": 76337, "tid": -914061504, "ts": 1716454224291155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291211, "dur": 0,
    "args": {
      "External id": 184924,
      "cbid": 317, "correlation": 184924
    }
  },
  {
    "ph": "f", "id": 184924, "pid": 76337, "tid": -914061504, "ts": 1716454224291211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291212, "dur": 0,
    "args": {
      "External id": 184925,
      "cbid": 203, "correlation": 184925
    }
  },
  {
    "ph": "f", "id": 184925, "pid": 76337, "tid": -914061504, "ts": 1716454224291212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291212, "dur": 0,
    "args": {
      "External id": 184926,
      "cbid": 205, "correlation": 184926
    }
  },
  {
    "ph": "f", "id": 184926, "pid": 76337, "tid": -914061504, "ts": 1716454224291212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224346018, "dur": 6,
    "args": {
      "External id": 184930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184930,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184930, "pid": 5, "tid": 7, "ts": 1716454224346018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291227, "dur": 11,
    "args": {
      "External id": 184930,
      "cbid": 211, "correlation": 184930
    }
  },
  {
    "ph": "s", "id": 184930, "pid": 76337, "tid": -914061504, "ts": 1716454224291227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224346026, "dur": 314,
    "args": {
      "External id": 184932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184932,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184932, "pid": 5, "tid": 7, "ts": 1716454224346026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291241, "dur": 5,
    "args": {
      "External id": 184932,
      "cbid": 211, "correlation": 184932
    }
  },
  {
    "ph": "s", "id": 184932, "pid": 76337, "tid": -914061504, "ts": 1716454224291241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224346342, "dur": 1,
    "args": {
      "External id": 184934,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 184934,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 184934, "pid": 5, "tid": 7, "ts": 1716454224346342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291251, "dur": 6,
    "args": {
      "External id": 184934,
      "cbid": 51, "correlation": 184934
    }
  },
  {
    "ph": "s", "id": 184934, "pid": 76337, "tid": -914061504, "ts": 1716454224291251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224346345, "dur": 489,
    "args": {
      "External id": 184935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184935,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184935, "pid": 5, "tid": 7, "ts": 1716454224346345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291259, "dur": 6,
    "args": {
      "External id": 184935,
      "cbid": 211, "correlation": 184935
    }
  },
  {
    "ph": "s", "id": 184935, "pid": 76337, "tid": -914061504, "ts": 1716454224291259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224346836, "dur": 5,
    "args": {
      "External id": 184937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184937,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 184937, "pid": 5, "tid": 7, "ts": 1716454224346836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291269, "dur": 6,
    "args": {
      "External id": 184937,
      "cbid": 211, "correlation": 184937
    }
  },
  {
    "ph": "s", "id": 184937, "pid": 76337, "tid": -914061504, "ts": 1716454224291269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224346842, "dur": 7,
    "args": {
      "External id": 184943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184943, "pid": 5, "tid": 7, "ts": 1716454224346842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291297, "dur": 8,
    "args": {
      "External id": 184943,
      "cbid": 211, "correlation": 184943
    }
  },
  {
    "ph": "s", "id": 184943, "pid": 76337, "tid": -914061504, "ts": 1716454224291297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224346850, "dur": 3,
    "args": {
      "External id": 184951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184951,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 184951, "pid": 5, "tid": 7, "ts": 1716454224346850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291342, "dur": 10,
    "args": {
      "External id": 184951,
      "cbid": 211, "correlation": 184951
    }
  },
  {
    "ph": "s", "id": 184951, "pid": 76337, "tid": -914061504, "ts": 1716454224291342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291404, "dur": 1,
    "args": {
      "External id": 184967,
      "cbid": 251, "correlation": 184967
    }
  },
  {
    "ph": "f", "id": 184967, "pid": 76337, "tid": -914061504, "ts": 1716454224291404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291410, "dur": 0,
    "args": {
      "External id": 184969,
      "cbid": 251, "correlation": 184969
    }
  },
  {
    "ph": "f", "id": 184969, "pid": 76337, "tid": -914061504, "ts": 1716454224291410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224346855, "dur": 12,
    "args": {
      "External id": 184970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184970,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184970, "pid": 5, "tid": 7, "ts": 1716454224346855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291411, "dur": 15,
    "args": {
      "External id": 184970,
      "cbid": 211, "correlation": 184970
    }
  },
  {
    "ph": "s", "id": 184970, "pid": 76337, "tid": -914061504, "ts": 1716454224291411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224346868, "dur": 5,
    "args": {
      "External id": 184972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184972,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 184972, "pid": 5, "tid": 7, "ts": 1716454224346868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291428, "dur": 5,
    "args": {
      "External id": 184972,
      "cbid": 211, "correlation": 184972
    }
  },
  {
    "ph": "s", "id": 184972, "pid": 76337, "tid": -914061504, "ts": 1716454224291428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224346875, "dur": 6,
    "args": {
      "External id": 184982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 184982,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 184982, "pid": 5, "tid": 7, "ts": 1716454224346875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291485, "dur": 11,
    "args": {
      "External id": 184982,
      "cbid": 211, "correlation": 184982
    }
  },
  {
    "ph": "s", "id": 184982, "pid": 76337, "tid": -914061504, "ts": 1716454224291485,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224346882, "dur": 9,
    "args": {
      "External id": 185002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185002,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 185002, "pid": 5, "tid": 7, "ts": 1716454224346882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291551, "dur": 10,
    "args": {
      "External id": 185002,
      "cbid": 211, "correlation": 185002
    }
  },
  {
    "ph": "s", "id": 185002, "pid": 76337, "tid": -914061504, "ts": 1716454224291551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224346892, "dur": 4,
    "args": {
      "External id": 185014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185014,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 185014, "pid": 5, "tid": 7, "ts": 1716454224346892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291571, "dur": 6,
    "args": {
      "External id": 185014,
      "cbid": 211, "correlation": 185014
    }
  },
  {
    "ph": "s", "id": 185014, "pid": 76337, "tid": -914061504, "ts": 1716454224291571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224346897, "dur": 6,
    "args": {
      "External id": 185017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185017,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185017, "pid": 5, "tid": 7, "ts": 1716454224346897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291591, "dur": 6,
    "args": {
      "External id": 185017,
      "cbid": 211, "correlation": 185017
    }
  },
  {
    "ph": "s", "id": 185017, "pid": 76337, "tid": -914061504, "ts": 1716454224291591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224346905, "dur": 4,
    "args": {
      "External id": 185026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185026,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185026, "pid": 5, "tid": 7, "ts": 1716454224346905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291630, "dur": 10,
    "args": {
      "External id": 185026,
      "cbid": 211, "correlation": 185026
    }
  },
  {
    "ph": "s", "id": 185026, "pid": 76337, "tid": -914061504, "ts": 1716454224291630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291696, "dur": 0,
    "args": {
      "External id": 185036,
      "cbid": 317, "correlation": 185036
    }
  },
  {
    "ph": "f", "id": 185036, "pid": 76337, "tid": -914061504, "ts": 1716454224291696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291697, "dur": 0,
    "args": {
      "External id": 185037,
      "cbid": 203, "correlation": 185037
    }
  },
  {
    "ph": "f", "id": 185037, "pid": 76337, "tid": -914061504, "ts": 1716454224291697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291698, "dur": 0,
    "args": {
      "External id": 185038,
      "cbid": 205, "correlation": 185038
    }
  },
  {
    "ph": "f", "id": 185038, "pid": 76337, "tid": -914061504, "ts": 1716454224291698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224346911, "dur": 5,
    "args": {
      "External id": 185042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185042,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185042, "pid": 5, "tid": 7, "ts": 1716454224346911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291711, "dur": 12,
    "args": {
      "External id": 185042,
      "cbid": 211, "correlation": 185042
    }
  },
  {
    "ph": "s", "id": 185042, "pid": 76337, "tid": -914061504, "ts": 1716454224291711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224346917, "dur": 159,
    "args": {
      "External id": 185044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185044,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185044, "pid": 5, "tid": 7, "ts": 1716454224346917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291726, "dur": 5,
    "args": {
      "External id": 185044,
      "cbid": 211, "correlation": 185044
    }
  },
  {
    "ph": "s", "id": 185044, "pid": 76337, "tid": -914061504, "ts": 1716454224291726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224347078, "dur": 1,
    "args": {
      "External id": 185046,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185046,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 185046, "pid": 5, "tid": 7, "ts": 1716454224347078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291737, "dur": 7,
    "args": {
      "External id": 185046,
      "cbid": 51, "correlation": 185046
    }
  },
  {
    "ph": "s", "id": 185046, "pid": 76337, "tid": -914061504, "ts": 1716454224291737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224347082, "dur": 256,
    "args": {
      "External id": 185047,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185047,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185047, "pid": 5, "tid": 7, "ts": 1716454224347082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291746, "dur": 6,
    "args": {
      "External id": 185047,
      "cbid": 211, "correlation": 185047
    }
  },
  {
    "ph": "s", "id": 185047, "pid": 76337, "tid": -914061504, "ts": 1716454224291746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224347338, "dur": 6,
    "args": {
      "External id": 185049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185049,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185049, "pid": 5, "tid": 7, "ts": 1716454224347338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291755, "dur": 5,
    "args": {
      "External id": 185049,
      "cbid": 211, "correlation": 185049
    }
  },
  {
    "ph": "s", "id": 185049, "pid": 76337, "tid": -914061504, "ts": 1716454224291755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224347345, "dur": 6,
    "args": {
      "External id": 185055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185055,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185055, "pid": 5, "tid": 7, "ts": 1716454224347345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291783, "dur": 9,
    "args": {
      "External id": 185055,
      "cbid": 211, "correlation": 185055
    }
  },
  {
    "ph": "s", "id": 185055, "pid": 76337, "tid": -914061504, "ts": 1716454224291783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291843, "dur": 0,
    "args": {
      "External id": 185065,
      "cbid": 317, "correlation": 185065
    }
  },
  {
    "ph": "f", "id": 185065, "pid": 76337, "tid": -914061504, "ts": 1716454224291843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291844, "dur": 0,
    "args": {
      "External id": 185066,
      "cbid": 203, "correlation": 185066
    }
  },
  {
    "ph": "f", "id": 185066, "pid": 76337, "tid": -914061504, "ts": 1716454224291844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291845, "dur": 0,
    "args": {
      "External id": 185067,
      "cbid": 205, "correlation": 185067
    }
  },
  {
    "ph": "f", "id": 185067, "pid": 76337, "tid": -914061504, "ts": 1716454224291845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224347353, "dur": 7,
    "args": {
      "External id": 185071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185071,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185071, "pid": 5, "tid": 7, "ts": 1716454224347353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291856, "dur": 12,
    "args": {
      "External id": 185071,
      "cbid": 211, "correlation": 185071
    }
  },
  {
    "ph": "s", "id": 185071, "pid": 76337, "tid": -914061504, "ts": 1716454224291856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224347361, "dur": 3,
    "args": {
      "External id": 185073,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185073,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3888888888888888
    }
  },
  {
    "ph": "f", "id": 185073, "pid": 5, "tid": 7, "ts": 1716454224347361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291873, "dur": 10,
    "args": {
      "External id": 185073,
      "cbid": 51, "correlation": 185073
    }
  },
  {
    "ph": "s", "id": 185073, "pid": 76337, "tid": -914061504, "ts": 1716454224291873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224347366, "dur": 95,
    "args": {
      "External id": 185074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185074,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 185074, "pid": 5, "tid": 7, "ts": 1716454224347366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291884, "dur": 7,
    "args": {
      "External id": 185074,
      "cbid": 211, "correlation": 185074
    }
  },
  {
    "ph": "s", "id": 185074, "pid": 76337, "tid": -914061504, "ts": 1716454224291884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224347462, "dur": 5,
    "args": {
      "External id": 185076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185076,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185076, "pid": 5, "tid": 7, "ts": 1716454224347462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291894, "dur": 6,
    "args": {
      "External id": 185076,
      "cbid": 211, "correlation": 185076
    }
  },
  {
    "ph": "s", "id": 185076, "pid": 76337, "tid": -914061504, "ts": 1716454224291894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224347469, "dur": 6,
    "args": {
      "External id": 185082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185082,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185082, "pid": 5, "tid": 7, "ts": 1716454224347469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291920, "dur": 8,
    "args": {
      "External id": 185082,
      "cbid": 211, "correlation": 185082
    }
  },
  {
    "ph": "s", "id": 185082, "pid": 76337, "tid": -914061504, "ts": 1716454224291920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224347476, "dur": 5,
    "args": {
      "External id": 185090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185090,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185090, "pid": 5, "tid": 7, "ts": 1716454224347476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291950, "dur": 8,
    "args": {
      "External id": 185090,
      "cbid": 211, "correlation": 185090
    }
  },
  {
    "ph": "s", "id": 185090, "pid": 76337, "tid": -914061504, "ts": 1716454224291950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224347482, "dur": 4,
    "args": {
      "External id": 185098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185098,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185098, "pid": 5, "tid": 7, "ts": 1716454224347482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224291987, "dur": 9,
    "args": {
      "External id": 185098,
      "cbid": 211, "correlation": 185098
    }
  },
  {
    "ph": "s", "id": 185098, "pid": 76337, "tid": -914061504, "ts": 1716454224291987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224347488, "dur": 11,
    "args": {
      "External id": 185107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185107,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185107, "pid": 5, "tid": 7, "ts": 1716454224347488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292032, "dur": 11,
    "args": {
      "External id": 185107,
      "cbid": 211, "correlation": 185107
    }
  },
  {
    "ph": "s", "id": 185107, "pid": 76337, "tid": -914061504, "ts": 1716454224292032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224347500, "dur": 12,
    "args": {
      "External id": 185127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185127,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 185127, "pid": 5, "tid": 7, "ts": 1716454224347500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292106, "dur": 12,
    "args": {
      "External id": 185127,
      "cbid": 211, "correlation": 185127
    }
  },
  {
    "ph": "s", "id": 185127, "pid": 76337, "tid": -914061504, "ts": 1716454224292106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224347514, "dur": 4,
    "args": {
      "External id": 185139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185139,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185139, "pid": 5, "tid": 7, "ts": 1716454224347514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292127, "dur": 6,
    "args": {
      "External id": 185139,
      "cbid": 211, "correlation": 185139
    }
  },
  {
    "ph": "s", "id": 185139, "pid": 76337, "tid": -914061504, "ts": 1716454224292127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224347520, "dur": 10,
    "args": {
      "External id": 185142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185142, "pid": 5, "tid": 7, "ts": 1716454224347520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292146, "dur": 7,
    "args": {
      "External id": 185142,
      "cbid": 211, "correlation": 185142
    }
  },
  {
    "ph": "s", "id": 185142, "pid": 76337, "tid": -914061504, "ts": 1716454224292146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224347531, "dur": 6,
    "args": {
      "External id": 185151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185151,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185151, "pid": 5, "tid": 7, "ts": 1716454224347531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292183, "dur": 10,
    "args": {
      "External id": 185151,
      "cbid": 211, "correlation": 185151
    }
  },
  {
    "ph": "s", "id": 185151, "pid": 76337, "tid": -914061504, "ts": 1716454224292183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292235, "dur": 0,
    "args": {
      "External id": 185161,
      "cbid": 317, "correlation": 185161
    }
  },
  {
    "ph": "f", "id": 185161, "pid": 76337, "tid": -914061504, "ts": 1716454224292235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292236, "dur": 0,
    "args": {
      "External id": 185162,
      "cbid": 203, "correlation": 185162
    }
  },
  {
    "ph": "f", "id": 185162, "pid": 76337, "tid": -914061504, "ts": 1716454224292236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292237, "dur": 0,
    "args": {
      "External id": 185163,
      "cbid": 205, "correlation": 185163
    }
  },
  {
    "ph": "f", "id": 185163, "pid": 76337, "tid": -914061504, "ts": 1716454224292237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224347539, "dur": 6,
    "args": {
      "External id": 185167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185167,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185167, "pid": 5, "tid": 7, "ts": 1716454224347539,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292251, "dur": 11,
    "args": {
      "External id": 185167,
      "cbid": 211, "correlation": 185167
    }
  },
  {
    "ph": "s", "id": 185167, "pid": 76337, "tid": -914061504, "ts": 1716454224292251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224347547, "dur": 314,
    "args": {
      "External id": 185169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185169,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185169, "pid": 5, "tid": 7, "ts": 1716454224347547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292265, "dur": 5,
    "args": {
      "External id": 185169,
      "cbid": 211, "correlation": 185169
    }
  },
  {
    "ph": "s", "id": 185169, "pid": 76337, "tid": -914061504, "ts": 1716454224292265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224347863, "dur": 1,
    "args": {
      "External id": 185171,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185171,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 185171, "pid": 5, "tid": 7, "ts": 1716454224347863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292275, "dur": 6,
    "args": {
      "External id": 185171,
      "cbid": 51, "correlation": 185171
    }
  },
  {
    "ph": "s", "id": 185171, "pid": 76337, "tid": -914061504, "ts": 1716454224292275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224347866, "dur": 490,
    "args": {
      "External id": 185172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185172,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185172, "pid": 5, "tid": 7, "ts": 1716454224347866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292282, "dur": 6,
    "args": {
      "External id": 185172,
      "cbid": 211, "correlation": 185172
    }
  },
  {
    "ph": "s", "id": 185172, "pid": 76337, "tid": -914061504, "ts": 1716454224292282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348357, "dur": 6,
    "args": {
      "External id": 185174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185174,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185174, "pid": 5, "tid": 7, "ts": 1716454224348357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292292, "dur": 6,
    "args": {
      "External id": 185174,
      "cbid": 211, "correlation": 185174
    }
  },
  {
    "ph": "s", "id": 185174, "pid": 76337, "tid": -914061504, "ts": 1716454224292292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224348364, "dur": 6,
    "args": {
      "External id": 185180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185180, "pid": 5, "tid": 7, "ts": 1716454224348364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292320, "dur": 8,
    "args": {
      "External id": 185180,
      "cbid": 211, "correlation": 185180
    }
  },
  {
    "ph": "s", "id": 185180, "pid": 76337, "tid": -914061504, "ts": 1716454224292320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224348371, "dur": 3,
    "args": {
      "External id": 185188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185188,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 185188, "pid": 5, "tid": 7, "ts": 1716454224348371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292363, "dur": 9,
    "args": {
      "External id": 185188,
      "cbid": 211, "correlation": 185188
    }
  },
  {
    "ph": "s", "id": 185188, "pid": 76337, "tid": -914061504, "ts": 1716454224292363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292426, "dur": 1,
    "args": {
      "External id": 185204,
      "cbid": 251, "correlation": 185204
    }
  },
  {
    "ph": "f", "id": 185204, "pid": 76337, "tid": -914061504, "ts": 1716454224292426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292431, "dur": 0,
    "args": {
      "External id": 185206,
      "cbid": 251, "correlation": 185206
    }
  },
  {
    "ph": "f", "id": 185206, "pid": 76337, "tid": -914061504, "ts": 1716454224292431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224348376, "dur": 12,
    "args": {
      "External id": 185207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185207,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185207, "pid": 5, "tid": 7, "ts": 1716454224348376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292433, "dur": 11,
    "args": {
      "External id": 185207,
      "cbid": 211, "correlation": 185207
    }
  },
  {
    "ph": "s", "id": 185207, "pid": 76337, "tid": -914061504, "ts": 1716454224292433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224348389, "dur": 5,
    "args": {
      "External id": 185209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185209,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185209, "pid": 5, "tid": 7, "ts": 1716454224348389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292446, "dur": 6,
    "args": {
      "External id": 185209,
      "cbid": 211, "correlation": 185209
    }
  },
  {
    "ph": "s", "id": 185209, "pid": 76337, "tid": -914061504, "ts": 1716454224292446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224348395, "dur": 6,
    "args": {
      "External id": 185219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185219,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185219, "pid": 5, "tid": 7, "ts": 1716454224348395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292502, "dur": 12,
    "args": {
      "External id": 185219,
      "cbid": 211, "correlation": 185219
    }
  },
  {
    "ph": "s", "id": 185219, "pid": 76337, "tid": -914061504, "ts": 1716454224292502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348402, "dur": 9,
    "args": {
      "External id": 185239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185239,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 185239, "pid": 5, "tid": 7, "ts": 1716454224348402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292567, "dur": 11,
    "args": {
      "External id": 185239,
      "cbid": 211, "correlation": 185239
    }
  },
  {
    "ph": "s", "id": 185239, "pid": 76337, "tid": -914061504, "ts": 1716454224292567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224348413, "dur": 4,
    "args": {
      "External id": 185251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185251,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 185251, "pid": 5, "tid": 7, "ts": 1716454224348413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292588, "dur": 6,
    "args": {
      "External id": 185251,
      "cbid": 211, "correlation": 185251
    }
  },
  {
    "ph": "s", "id": 185251, "pid": 76337, "tid": -914061504, "ts": 1716454224292588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224348418, "dur": 7,
    "args": {
      "External id": 185254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185254,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185254, "pid": 5, "tid": 7, "ts": 1716454224348418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292608, "dur": 6,
    "args": {
      "External id": 185254,
      "cbid": 211, "correlation": 185254
    }
  },
  {
    "ph": "s", "id": 185254, "pid": 76337, "tid": -914061504, "ts": 1716454224292608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224348426, "dur": 4,
    "args": {
      "External id": 185263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185263,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185263, "pid": 5, "tid": 7, "ts": 1716454224348426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292646, "dur": 10,
    "args": {
      "External id": 185263,
      "cbid": 211, "correlation": 185263
    }
  },
  {
    "ph": "s", "id": 185263, "pid": 76337, "tid": -914061504, "ts": 1716454224292646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292713, "dur": 0,
    "args": {
      "External id": 185273,
      "cbid": 317, "correlation": 185273
    }
  },
  {
    "ph": "f", "id": 185273, "pid": 76337, "tid": -914061504, "ts": 1716454224292713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292714, "dur": 0,
    "args": {
      "External id": 185274,
      "cbid": 203, "correlation": 185274
    }
  },
  {
    "ph": "f", "id": 185274, "pid": 76337, "tid": -914061504, "ts": 1716454224292714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292715, "dur": 0,
    "args": {
      "External id": 185275,
      "cbid": 205, "correlation": 185275
    }
  },
  {
    "ph": "f", "id": 185275, "pid": 76337, "tid": -914061504, "ts": 1716454224292715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348432, "dur": 5,
    "args": {
      "External id": 185279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185279,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185279, "pid": 5, "tid": 7, "ts": 1716454224348432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292728, "dur": 12,
    "args": {
      "External id": 185279,
      "cbid": 211, "correlation": 185279
    }
  },
  {
    "ph": "s", "id": 185279, "pid": 76337, "tid": -914061504, "ts": 1716454224292728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348438, "dur": 159,
    "args": {
      "External id": 185281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185281,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185281, "pid": 5, "tid": 7, "ts": 1716454224348438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292742, "dur": 5,
    "args": {
      "External id": 185281,
      "cbid": 211, "correlation": 185281
    }
  },
  {
    "ph": "s", "id": 185281, "pid": 76337, "tid": -914061504, "ts": 1716454224292742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224348599, "dur": 1,
    "args": {
      "External id": 185283,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185283,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 185283, "pid": 5, "tid": 7, "ts": 1716454224348599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292753, "dur": 7,
    "args": {
      "External id": 185283,
      "cbid": 51, "correlation": 185283
    }
  },
  {
    "ph": "s", "id": 185283, "pid": 76337, "tid": -914061504, "ts": 1716454224292753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224348603, "dur": 254,
    "args": {
      "External id": 185284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185284,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185284, "pid": 5, "tid": 7, "ts": 1716454224348603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292761, "dur": 6,
    "args": {
      "External id": 185284,
      "cbid": 211, "correlation": 185284
    }
  },
  {
    "ph": "s", "id": 185284, "pid": 76337, "tid": -914061504, "ts": 1716454224292761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348859, "dur": 6,
    "args": {
      "External id": 185286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185286,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185286, "pid": 5, "tid": 7, "ts": 1716454224348859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292771, "dur": 5,
    "args": {
      "External id": 185286,
      "cbid": 211, "correlation": 185286
    }
  },
  {
    "ph": "s", "id": 185286, "pid": 76337, "tid": -914061504, "ts": 1716454224292771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224348866, "dur": 6,
    "args": {
      "External id": 185292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185292, "pid": 5, "tid": 7, "ts": 1716454224348866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292799, "dur": 9,
    "args": {
      "External id": 185292,
      "cbid": 211, "correlation": 185292
    }
  },
  {
    "ph": "s", "id": 185292, "pid": 76337, "tid": -914061504, "ts": 1716454224292799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292857, "dur": 0,
    "args": {
      "External id": 185302,
      "cbid": 317, "correlation": 185302
    }
  },
  {
    "ph": "f", "id": 185302, "pid": 76337, "tid": -914061504, "ts": 1716454224292857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292858, "dur": 0,
    "args": {
      "External id": 185303,
      "cbid": 203, "correlation": 185303
    }
  },
  {
    "ph": "f", "id": 185303, "pid": 76337, "tid": -914061504, "ts": 1716454224292858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292859, "dur": 0,
    "args": {
      "External id": 185304,
      "cbid": 205, "correlation": 185304
    }
  },
  {
    "ph": "f", "id": 185304, "pid": 76337, "tid": -914061504, "ts": 1716454224292859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348874, "dur": 7,
    "args": {
      "External id": 185308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185308,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185308, "pid": 5, "tid": 7, "ts": 1716454224348874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292871, "dur": 11,
    "args": {
      "External id": 185308,
      "cbid": 211, "correlation": 185308
    }
  },
  {
    "ph": "s", "id": 185308, "pid": 76337, "tid": -914061504, "ts": 1716454224292871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224348882, "dur": 3,
    "args": {
      "External id": 185310,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185310,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.530612244897959
    }
  },
  {
    "ph": "f", "id": 185310, "pid": 5, "tid": 7, "ts": 1716454224348882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292887, "dur": 10,
    "args": {
      "External id": 185310,
      "cbid": 51, "correlation": 185310
    }
  },
  {
    "ph": "s", "id": 185310, "pid": 76337, "tid": -914061504, "ts": 1716454224292887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224348886, "dur": 93,
    "args": {
      "External id": 185311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185311,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 185311, "pid": 5, "tid": 7, "ts": 1716454224348886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292899, "dur": 6,
    "args": {
      "External id": 185311,
      "cbid": 211, "correlation": 185311
    }
  },
  {
    "ph": "s", "id": 185311, "pid": 76337, "tid": -914061504, "ts": 1716454224292899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224348981, "dur": 5,
    "args": {
      "External id": 185313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185313,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185313, "pid": 5, "tid": 7, "ts": 1716454224348981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292908, "dur": 5,
    "args": {
      "External id": 185313,
      "cbid": 211, "correlation": 185313
    }
  },
  {
    "ph": "s", "id": 185313, "pid": 76337, "tid": -914061504, "ts": 1716454224292908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224348987, "dur": 6,
    "args": {
      "External id": 185319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185319,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185319, "pid": 5, "tid": 7, "ts": 1716454224348987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292935, "dur": 8,
    "args": {
      "External id": 185319,
      "cbid": 211, "correlation": 185319
    }
  },
  {
    "ph": "s", "id": 185319, "pid": 76337, "tid": -914061504, "ts": 1716454224292935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224348994, "dur": 5,
    "args": {
      "External id": 185327,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185327,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185327, "pid": 5, "tid": 7, "ts": 1716454224348994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224292963, "dur": 8,
    "args": {
      "External id": 185327,
      "cbid": 211, "correlation": 185327
    }
  },
  {
    "ph": "s", "id": 185327, "pid": 76337, "tid": -914061504, "ts": 1716454224292963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224349001, "dur": 4,
    "args": {
      "External id": 185335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185335,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 185335, "pid": 5, "tid": 7, "ts": 1716454224349001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293000, "dur": 9,
    "args": {
      "External id": 185335,
      "cbid": 211, "correlation": 185335
    }
  },
  {
    "ph": "s", "id": 185335, "pid": 76337, "tid": -914061504, "ts": 1716454224293000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224349006, "dur": 14,
    "args": {
      "External id": 185346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185346,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185346, "pid": 5, "tid": 7, "ts": 1716454224349006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293102, "dur": 14,
    "args": {
      "External id": 185346,
      "cbid": 211, "correlation": 185346
    }
  },
  {
    "ph": "s", "id": 185346, "pid": 76337, "tid": -914061504, "ts": 1716454224293102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293159, "dur": 0,
    "args": {
      "External id": 185356,
      "cbid": 317, "correlation": 185356
    }
  },
  {
    "ph": "f", "id": 185356, "pid": 76337, "tid": -914061504, "ts": 1716454224293159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293160, "dur": 0,
    "args": {
      "External id": 185357,
      "cbid": 203, "correlation": 185357
    }
  },
  {
    "ph": "f", "id": 185357, "pid": 76337, "tid": -914061504, "ts": 1716454224293160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293160, "dur": 0,
    "args": {
      "External id": 185358,
      "cbid": 205, "correlation": 185358
    }
  },
  {
    "ph": "f", "id": 185358, "pid": 76337, "tid": -914061504, "ts": 1716454224293160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224349022, "dur": 8,
    "args": {
      "External id": 185362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185362,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185362, "pid": 5, "tid": 7, "ts": 1716454224349022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293176, "dur": 11,
    "args": {
      "External id": 185362,
      "cbid": 211, "correlation": 185362
    }
  },
  {
    "ph": "s", "id": 185362, "pid": 76337, "tid": -914061504, "ts": 1716454224293176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224349031, "dur": 160,
    "args": {
      "External id": 185364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185364,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185364, "pid": 5, "tid": 7, "ts": 1716454224349031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293189, "dur": 5,
    "args": {
      "External id": 185364,
      "cbid": 211, "correlation": 185364
    }
  },
  {
    "ph": "s", "id": 185364, "pid": 76337, "tid": -914061504, "ts": 1716454224293189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224349193, "dur": 1,
    "args": {
      "External id": 185366,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185366,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 185366, "pid": 5, "tid": 7, "ts": 1716454224349193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293200, "dur": 6,
    "args": {
      "External id": 185366,
      "cbid": 51, "correlation": 185366
    }
  },
  {
    "ph": "s", "id": 185366, "pid": 76337, "tid": -914061504, "ts": 1716454224293200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224349197, "dur": 643,
    "args": {
      "External id": 185367,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185367,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185367, "pid": 5, "tid": 7, "ts": 1716454224349197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293207, "dur": 6,
    "args": {
      "External id": 185367,
      "cbid": 211, "correlation": 185367
    }
  },
  {
    "ph": "s", "id": 185367, "pid": 76337, "tid": -914061504, "ts": 1716454224293207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224349842, "dur": 12,
    "args": {
      "External id": 185369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185369,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185369, "pid": 5, "tid": 7, "ts": 1716454224349842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293217, "dur": 6,
    "args": {
      "External id": 185369,
      "cbid": 211, "correlation": 185369
    }
  },
  {
    "ph": "s", "id": 185369, "pid": 76337, "tid": -914061504, "ts": 1716454224293217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224349855, "dur": 15,
    "args": {
      "External id": 185375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185375,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185375, "pid": 5, "tid": 7, "ts": 1716454224349855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293247, "dur": 9,
    "args": {
      "External id": 185375,
      "cbid": 211, "correlation": 185375
    }
  },
  {
    "ph": "s", "id": 185375, "pid": 76337, "tid": -914061504, "ts": 1716454224293247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224349871, "dur": 30,
    "args": {
      "External id": 185384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185384,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185384, "pid": 5, "tid": 7, "ts": 1716454224349871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293345, "dur": 13,
    "args": {
      "External id": 185384,
      "cbid": 211, "correlation": 185384
    }
  },
  {
    "ph": "s", "id": 185384, "pid": 76337, "tid": -914061504, "ts": 1716454224293345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224349902, "dur": 31,
    "args": {
      "External id": 185404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185404,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 185404, "pid": 5, "tid": 7, "ts": 1716454224349902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293417, "dur": 12,
    "args": {
      "External id": 185404,
      "cbid": 211, "correlation": 185404
    }
  },
  {
    "ph": "s", "id": 185404, "pid": 76337, "tid": -914061504, "ts": 1716454224293417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224349934, "dur": 5,
    "args": {
      "External id": 185416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185416,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185416, "pid": 5, "tid": 7, "ts": 1716454224349934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293439, "dur": 6,
    "args": {
      "External id": 185416,
      "cbid": 211, "correlation": 185416
    }
  },
  {
    "ph": "s", "id": 185416, "pid": 76337, "tid": -914061504, "ts": 1716454224293439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224349940, "dur": 30,
    "args": {
      "External id": 185419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185419, "pid": 5, "tid": 7, "ts": 1716454224349940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293457, "dur": 6,
    "args": {
      "External id": 185419,
      "cbid": 211, "correlation": 185419
    }
  },
  {
    "ph": "s", "id": 185419, "pid": 76337, "tid": -914061504, "ts": 1716454224293457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224349972, "dur": 22,
    "args": {
      "External id": 185428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185428,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185428, "pid": 5, "tid": 7, "ts": 1716454224349972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293496, "dur": 10,
    "args": {
      "External id": 185428,
      "cbid": 211, "correlation": 185428
    }
  },
  {
    "ph": "s", "id": 185428, "pid": 76337, "tid": -914061504, "ts": 1716454224293496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293550, "dur": 0,
    "args": {
      "External id": 185438,
      "cbid": 317, "correlation": 185438
    }
  },
  {
    "ph": "f", "id": 185438, "pid": 76337, "tid": -914061504, "ts": 1716454224293550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293551, "dur": 0,
    "args": {
      "External id": 185439,
      "cbid": 203, "correlation": 185439
    }
  },
  {
    "ph": "f", "id": 185439, "pid": 76337, "tid": -914061504, "ts": 1716454224293551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293552, "dur": 0,
    "args": {
      "External id": 185440,
      "cbid": 205, "correlation": 185440
    }
  },
  {
    "ph": "f", "id": 185440, "pid": 76337, "tid": -914061504, "ts": 1716454224293552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224349994, "dur": 22,
    "args": {
      "External id": 185444,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185444,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185444, "pid": 5, "tid": 7, "ts": 1716454224349994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293567, "dur": 12,
    "args": {
      "External id": 185444,
      "cbid": 211, "correlation": 185444
    }
  },
  {
    "ph": "s", "id": 185444, "pid": 76337, "tid": -914061504, "ts": 1716454224293567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224350017, "dur": 313,
    "args": {
      "External id": 185446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185446, "pid": 5, "tid": 7, "ts": 1716454224350017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293582, "dur": 6,
    "args": {
      "External id": 185446,
      "cbid": 211, "correlation": 185446
    }
  },
  {
    "ph": "s", "id": 185446, "pid": 76337, "tid": -914061504, "ts": 1716454224293582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224350333, "dur": 1,
    "args": {
      "External id": 185448,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185448,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 185448, "pid": 5, "tid": 7, "ts": 1716454224350333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293593, "dur": 7,
    "args": {
      "External id": 185448,
      "cbid": 51, "correlation": 185448
    }
  },
  {
    "ph": "s", "id": 185448, "pid": 76337, "tid": -914061504, "ts": 1716454224293593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224350336, "dur": 1228,
    "args": {
      "External id": 185449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185449,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185449, "pid": 5, "tid": 7, "ts": 1716454224350336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293601, "dur": 6,
    "args": {
      "External id": 185449,
      "cbid": 211, "correlation": 185449
    }
  },
  {
    "ph": "s", "id": 185449, "pid": 76337, "tid": -914061504, "ts": 1716454224293601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224351566, "dur": 13,
    "args": {
      "External id": 185451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185451, "pid": 5, "tid": 7, "ts": 1716454224351566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293611, "dur": 5,
    "args": {
      "External id": 185451,
      "cbid": 211, "correlation": 185451
    }
  },
  {
    "ph": "s", "id": 185451, "pid": 76337, "tid": -914061504, "ts": 1716454224293611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224351580, "dur": 15,
    "args": {
      "External id": 185457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185457, "pid": 5, "tid": 7, "ts": 1716454224351580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293640, "dur": 9,
    "args": {
      "External id": 185457,
      "cbid": 211, "correlation": 185457
    }
  },
  {
    "ph": "s", "id": 185457, "pid": 76337, "tid": -914061504, "ts": 1716454224293640,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224351596, "dur": 3,
    "args": {
      "External id": 185465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185465,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 185465, "pid": 5, "tid": 7, "ts": 1716454224351596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293684, "dur": 10,
    "args": {
      "External id": 185465,
      "cbid": 211, "correlation": 185465
    }
  },
  {
    "ph": "s", "id": 185465, "pid": 76337, "tid": -914061504, "ts": 1716454224293684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293748, "dur": 1,
    "args": {
      "External id": 185481,
      "cbid": 251, "correlation": 185481
    }
  },
  {
    "ph": "f", "id": 185481, "pid": 76337, "tid": -914061504, "ts": 1716454224293748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293753, "dur": 0,
    "args": {
      "External id": 185483,
      "cbid": 251, "correlation": 185483
    }
  },
  {
    "ph": "f", "id": 185483, "pid": 76337, "tid": -914061504, "ts": 1716454224293753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224351601, "dur": 12,
    "args": {
      "External id": 185484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185484,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185484, "pid": 5, "tid": 7, "ts": 1716454224351601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293755, "dur": 11,
    "args": {
      "External id": 185484,
      "cbid": 211, "correlation": 185484
    }
  },
  {
    "ph": "s", "id": 185484, "pid": 76337, "tid": -914061504, "ts": 1716454224293755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224351614, "dur": 5,
    "args": {
      "External id": 185486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185486,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185486, "pid": 5, "tid": 7, "ts": 1716454224351614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293768, "dur": 5,
    "args": {
      "External id": 185486,
      "cbid": 211, "correlation": 185486
    }
  },
  {
    "ph": "s", "id": 185486, "pid": 76337, "tid": -914061504, "ts": 1716454224293768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224351620, "dur": 17,
    "args": {
      "External id": 185496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185496,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185496, "pid": 5, "tid": 7, "ts": 1716454224351620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293824, "dur": 12,
    "args": {
      "External id": 185496,
      "cbid": 211, "correlation": 185496
    }
  },
  {
    "ph": "s", "id": 185496, "pid": 76337, "tid": -914061504, "ts": 1716454224293824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224351638, "dur": 19,
    "args": {
      "External id": 185516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185516,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 185516, "pid": 5, "tid": 7, "ts": 1716454224351638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293890, "dur": 11,
    "args": {
      "External id": 185516,
      "cbid": 211, "correlation": 185516
    }
  },
  {
    "ph": "s", "id": 185516, "pid": 76337, "tid": -914061504, "ts": 1716454224293890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224351658, "dur": 4,
    "args": {
      "External id": 185528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185528,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 185528, "pid": 5, "tid": 7, "ts": 1716454224351658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293911, "dur": 6,
    "args": {
      "External id": 185528,
      "cbid": 211, "correlation": 185528
    }
  },
  {
    "ph": "s", "id": 185528, "pid": 76337, "tid": -914061504, "ts": 1716454224293911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224351664, "dur": 16,
    "args": {
      "External id": 185531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185531,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185531, "pid": 5, "tid": 7, "ts": 1716454224351664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293929, "dur": 6,
    "args": {
      "External id": 185531,
      "cbid": 211, "correlation": 185531
    }
  },
  {
    "ph": "s", "id": 185531, "pid": 76337, "tid": -914061504, "ts": 1716454224293929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224351681, "dur": 12,
    "args": {
      "External id": 185540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185540,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185540, "pid": 5, "tid": 7, "ts": 1716454224351681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224293969, "dur": 19,
    "args": {
      "External id": 185540,
      "cbid": 211, "correlation": 185540
    }
  },
  {
    "ph": "s", "id": 185540, "pid": 76337, "tid": -914061504, "ts": 1716454224293969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294042, "dur": 0,
    "args": {
      "External id": 185550,
      "cbid": 317, "correlation": 185550
    }
  },
  {
    "ph": "f", "id": 185550, "pid": 76337, "tid": -914061504, "ts": 1716454224294042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294043, "dur": 0,
    "args": {
      "External id": 185551,
      "cbid": 203, "correlation": 185551
    }
  },
  {
    "ph": "f", "id": 185551, "pid": 76337, "tid": -914061504, "ts": 1716454224294043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294044, "dur": 0,
    "args": {
      "External id": 185552,
      "cbid": 205, "correlation": 185552
    }
  },
  {
    "ph": "f", "id": 185552, "pid": 76337, "tid": -914061504, "ts": 1716454224294044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224351694, "dur": 11,
    "args": {
      "External id": 185556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185556,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185556, "pid": 5, "tid": 7, "ts": 1716454224351694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294058, "dur": 12,
    "args": {
      "External id": 185556,
      "cbid": 211, "correlation": 185556
    }
  },
  {
    "ph": "s", "id": 185556, "pid": 76337, "tid": -914061504, "ts": 1716454224294058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224351706, "dur": 160,
    "args": {
      "External id": 185558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185558,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185558, "pid": 5, "tid": 7, "ts": 1716454224351706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294072, "dur": 5,
    "args": {
      "External id": 185558,
      "cbid": 211, "correlation": 185558
    }
  },
  {
    "ph": "s", "id": 185558, "pid": 76337, "tid": -914061504, "ts": 1716454224294072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224351868, "dur": 1,
    "args": {
      "External id": 185560,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 185560,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 185560, "pid": 5, "tid": 7, "ts": 1716454224351868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294083, "dur": 7,
    "args": {
      "External id": 185560,
      "cbid": 51, "correlation": 185560
    }
  },
  {
    "ph": "s", "id": 185560, "pid": 76337, "tid": -914061504, "ts": 1716454224294083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224351872, "dur": 640,
    "args": {
      "External id": 185561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185561,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185561, "pid": 5, "tid": 7, "ts": 1716454224351872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294091, "dur": 7,
    "args": {
      "External id": 185561,
      "cbid": 211, "correlation": 185561
    }
  },
  {
    "ph": "s", "id": 185561, "pid": 76337, "tid": -914061504, "ts": 1716454224294091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224352513, "dur": 13,
    "args": {
      "External id": 185563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185563,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185563, "pid": 5, "tid": 7, "ts": 1716454224352513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294102, "dur": 5,
    "args": {
      "External id": 185563,
      "cbid": 211, "correlation": 185563
    }
  },
  {
    "ph": "s", "id": 185563, "pid": 76337, "tid": -914061504, "ts": 1716454224294102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224352528, "dur": 14,
    "args": {
      "External id": 185569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185569, "pid": 5, "tid": 7, "ts": 1716454224352528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294130, "dur": 8,
    "args": {
      "External id": 185569,
      "cbid": 211, "correlation": 185569
    }
  },
  {
    "ph": "s", "id": 185569, "pid": 76337, "tid": -914061504, "ts": 1716454224294130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294188, "dur": 0,
    "args": {
      "External id": 185579,
      "cbid": 317, "correlation": 185579
    }
  },
  {
    "ph": "f", "id": 185579, "pid": 76337, "tid": -914061504, "ts": 1716454224294188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294189, "dur": 0,
    "args": {
      "External id": 185580,
      "cbid": 203, "correlation": 185580
    }
  },
  {
    "ph": "f", "id": 185580, "pid": 76337, "tid": -914061504, "ts": 1716454224294189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294190, "dur": 0,
    "args": {
      "External id": 185581,
      "cbid": 205, "correlation": 185581
    }
  },
  {
    "ph": "f", "id": 185581, "pid": 76337, "tid": -914061504, "ts": 1716454224294190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224352543, "dur": 21,
    "args": {
      "External id": 185585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185585,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185585, "pid": 5, "tid": 7, "ts": 1716454224352543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294205, "dur": 11,
    "args": {
      "External id": 185585,
      "cbid": 211, "correlation": 185585
    }
  },
  {
    "ph": "s", "id": 185585, "pid": 76337, "tid": -914061504, "ts": 1716454224294205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224352566, "dur": 4,
    "args": {
      "External id": 185587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185587,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 185587, "pid": 5, "tid": 7, "ts": 1716454224352566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294221, "dur": 6,
    "args": {
      "External id": 185587,
      "cbid": 211, "correlation": 185587
    }
  },
  {
    "ph": "s", "id": 185587, "pid": 76337, "tid": -914061504, "ts": 1716454224294221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294231, "dur": 0,
    "args": {
      "External id": 185588,
      "cbid": 51, "correlation": 185588
    }
  },
  {
    "ph": "s", "id": 185588, "pid": 76337, "tid": -914061504, "ts": 1716454224294231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224352571, "dur": 173,
    "args": {
      "External id": 185589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185589,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 185589, "pid": 5, "tid": 7, "ts": 1716454224352571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294231, "dur": 6,
    "args": {
      "External id": 185589,
      "cbid": 211, "correlation": 185589
    }
  },
  {
    "ph": "s", "id": 185589, "pid": 76337, "tid": -914061504, "ts": 1716454224294231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224352745, "dur": 15,
    "args": {
      "External id": 185594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185594,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185594, "pid": 5, "tid": 7, "ts": 1716454224352745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294258, "dur": 9,
    "args": {
      "External id": 185594,
      "cbid": 211, "correlation": 185594
    }
  },
  {
    "ph": "s", "id": 185594, "pid": 76337, "tid": -914061504, "ts": 1716454224294258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224352762, "dur": 12,
    "args": {
      "External id": 185602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185602,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185602, "pid": 5, "tid": 7, "ts": 1716454224352762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294287, "dur": 8,
    "args": {
      "External id": 185602,
      "cbid": 211, "correlation": 185602
    }
  },
  {
    "ph": "s", "id": 185602, "pid": 76337, "tid": -914061504, "ts": 1716454224294287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224352774, "dur": 10,
    "args": {
      "External id": 185610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185610,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185610, "pid": 5, "tid": 7, "ts": 1716454224352774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294316, "dur": 8,
    "args": {
      "External id": 185610,
      "cbid": 211, "correlation": 185610
    }
  },
  {
    "ph": "s", "id": 185610, "pid": 76337, "tid": -914061504, "ts": 1716454224294316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224352786, "dur": 18,
    "args": {
      "External id": 185630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185630,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 185630, "pid": 5, "tid": 7, "ts": 1716454224352786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294397, "dur": 13,
    "args": {
      "External id": 185630,
      "cbid": 211, "correlation": 185630
    }
  },
  {
    "ph": "s", "id": 185630, "pid": 76337, "tid": -914061504, "ts": 1716454224294397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224352805, "dur": 4,
    "args": {
      "External id": 185642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185642,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 185642, "pid": 5, "tid": 7, "ts": 1716454224352805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294420, "dur": 6,
    "args": {
      "External id": 185642,
      "cbid": 211, "correlation": 185642
    }
  },
  {
    "ph": "s", "id": 185642, "pid": 76337, "tid": -914061504, "ts": 1716454224294420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224352810, "dur": 16,
    "args": {
      "External id": 185645,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185645,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185645, "pid": 5, "tid": 7, "ts": 1716454224352810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294438, "dur": 6,
    "args": {
      "External id": 185645,
      "cbid": 211, "correlation": 185645
    }
  },
  {
    "ph": "s", "id": 185645, "pid": 76337, "tid": -914061504, "ts": 1716454224294438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294495, "dur": 0,
    "args": {
      "External id": 185656,
      "cbid": 317, "correlation": 185656
    }
  },
  {
    "ph": "f", "id": 185656, "pid": 76337, "tid": -914061504, "ts": 1716454224294495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294496, "dur": 0,
    "args": {
      "External id": 185657,
      "cbid": 203, "correlation": 185657
    }
  },
  {
    "ph": "f", "id": 185657, "pid": 76337, "tid": -914061504, "ts": 1716454224294496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294497, "dur": 0,
    "args": {
      "External id": 185658,
      "cbid": 205, "correlation": 185658
    }
  },
  {
    "ph": "f", "id": 185658, "pid": 76337, "tid": -914061504, "ts": 1716454224294497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224352828, "dur": 11,
    "args": {
      "External id": 185662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185662,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185662, "pid": 5, "tid": 7, "ts": 1716454224352828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294511, "dur": 12,
    "args": {
      "External id": 185662,
      "cbid": 211, "correlation": 185662
    }
  },
  {
    "ph": "s", "id": 185662, "pid": 76337, "tid": -914061504, "ts": 1716454224294511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224352841, "dur": 3,
    "args": {
      "External id": 185664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185664,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 185664, "pid": 5, "tid": 7, "ts": 1716454224352841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294528, "dur": 6,
    "args": {
      "External id": 185664,
      "cbid": 211, "correlation": 185664
    }
  },
  {
    "ph": "s", "id": 185664, "pid": 76337, "tid": -914061504, "ts": 1716454224294528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294536, "dur": 0,
    "args": {
      "External id": 185665,
      "cbid": 51, "correlation": 185665
    }
  },
  {
    "ph": "s", "id": 185665, "pid": 76337, "tid": -914061504, "ts": 1716454224294536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224352845, "dur": 89,
    "args": {
      "External id": 185666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185666,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 185666, "pid": 5, "tid": 7, "ts": 1716454224352845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294537, "dur": 5,
    "args": {
      "External id": 185666,
      "cbid": 211, "correlation": 185666
    }
  },
  {
    "ph": "s", "id": 185666, "pid": 76337, "tid": -914061504, "ts": 1716454224294537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224352935, "dur": 16,
    "args": {
      "External id": 185671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185671, "pid": 5, "tid": 7, "ts": 1716454224352935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294564, "dur": 8,
    "args": {
      "External id": 185671,
      "cbid": 211, "correlation": 185671
    }
  },
  {
    "ph": "s", "id": 185671, "pid": 76337, "tid": -914061504, "ts": 1716454224294564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224352952, "dur": 83,
    "args": {
      "External id": 185680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185680,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185680, "pid": 5, "tid": 7, "ts": 1716454224352952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294645, "dur": 16,
    "args": {
      "External id": 185680,
      "cbid": 211, "correlation": 185680
    }
  },
  {
    "ph": "s", "id": 185680, "pid": 76337, "tid": -914061504, "ts": 1716454224294645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224353036, "dur": 29,
    "args": {
      "External id": 185702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185702,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185702, "pid": 5, "tid": 7, "ts": 1716454224353036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294705, "dur": 11,
    "args": {
      "External id": 185702,
      "cbid": 211, "correlation": 185702
    }
  },
  {
    "ph": "s", "id": 185702, "pid": 76337, "tid": -914061504, "ts": 1716454224294705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294802, "dur": 2,
    "args": {
      "External id": 185713,
      "cbid": 251, "correlation": 185713
    }
  },
  {
    "ph": "f", "id": 185713, "pid": 76337, "tid": -914061504, "ts": 1716454224294802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224353067, "dur": 140,
    "args": {
      "External id": 185714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185714,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185714, "pid": 5, "tid": 7, "ts": 1716454224353067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294809, "dur": 14,
    "args": {
      "External id": 185714,
      "cbid": 211, "correlation": 185714
    }
  },
  {
    "ph": "s", "id": 185714, "pid": 76337, "tid": -914061504, "ts": 1716454224294809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294880, "dur": 1,
    "args": {
      "External id": 185725,
      "cbid": 251, "correlation": 185725
    }
  },
  {
    "ph": "f", "id": 185725, "pid": 76337, "tid": -914061504, "ts": 1716454224294880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224353208, "dur": 157,
    "args": {
      "External id": 185726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185726,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185726, "pid": 5, "tid": 7, "ts": 1716454224353208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294884, "dur": 11,
    "args": {
      "External id": 185726,
      "cbid": 211, "correlation": 185726
    }
  },
  {
    "ph": "s", "id": 185726, "pid": 76337, "tid": -914061504, "ts": 1716454224294884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294949, "dur": 1,
    "args": {
      "External id": 185737,
      "cbid": 251, "correlation": 185737
    }
  },
  {
    "ph": "f", "id": 185737, "pid": 76337, "tid": -914061504, "ts": 1716454224294949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224353366, "dur": 135,
    "args": {
      "External id": 185738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185738,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185738, "pid": 5, "tid": 7, "ts": 1716454224353366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224294953, "dur": 12,
    "args": {
      "External id": 185738,
      "cbid": 211, "correlation": 185738
    }
  },
  {
    "ph": "s", "id": 185738, "pid": 76337, "tid": -914061504, "ts": 1716454224294953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224353503, "dur": 332,
    "args": {
      "External id": 185763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185763,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185763, "pid": 5, "tid": 7, "ts": 1716454224353503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295051, "dur": 14,
    "args": {
      "External id": 185763,
      "cbid": 211, "correlation": 185763
    }
  },
  {
    "ph": "s", "id": 185763, "pid": 76337, "tid": -914061504, "ts": 1716454224295051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295154, "dur": 1,
    "args": {
      "External id": 185781,
      "cbid": 251, "correlation": 185781
    }
  },
  {
    "ph": "f", "id": 185781, "pid": 76337, "tid": -914061504, "ts": 1716454224295154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224353836, "dur": 164,
    "args": {
      "External id": 185783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185783,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185783, "pid": 5, "tid": 7, "ts": 1716454224353836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295160, "dur": 14,
    "args": {
      "External id": 185783,
      "cbid": 211, "correlation": 185783
    }
  },
  {
    "ph": "s", "id": 185783, "pid": 76337, "tid": -914061504, "ts": 1716454224295160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224354002, "dur": 19,
    "args": {
      "External id": 185791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185791,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185791, "pid": 5, "tid": 7, "ts": 1716454224354002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295231, "dur": 12,
    "args": {
      "External id": 185791,
      "cbid": 211, "correlation": 185791
    }
  },
  {
    "ph": "s", "id": 185791, "pid": 76337, "tid": -914061504, "ts": 1716454224295231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224354023, "dur": 27,
    "args": {
      "External id": 185799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185799,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185799, "pid": 5, "tid": 7, "ts": 1716454224354023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295270, "dur": 8,
    "args": {
      "External id": 185799,
      "cbid": 211, "correlation": 185799
    }
  },
  {
    "ph": "s", "id": 185799, "pid": 76337, "tid": -914061504, "ts": 1716454224295270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224354051, "dur": 19,
    "args": {
      "External id": 185810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185810, "pid": 5, "tid": 7, "ts": 1716454224354051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295342, "dur": 13,
    "args": {
      "External id": 185810,
      "cbid": 211, "correlation": 185810
    }
  },
  {
    "ph": "s", "id": 185810, "pid": 76337, "tid": -914061504, "ts": 1716454224295342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224354071, "dur": 16,
    "args": {
      "External id": 185832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185832,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185832, "pid": 5, "tid": 7, "ts": 1716454224354071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295374, "dur": 7,
    "args": {
      "External id": 185832,
      "cbid": 211, "correlation": 185832
    }
  },
  {
    "ph": "s", "id": 185832, "pid": 76337, "tid": -914061504, "ts": 1716454224295374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295458, "dur": 1,
    "args": {
      "External id": 185843,
      "cbid": 251, "correlation": 185843
    }
  },
  {
    "ph": "f", "id": 185843, "pid": 76337, "tid": -914061504, "ts": 1716454224295458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224354089, "dur": 88,
    "args": {
      "External id": 185844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185844,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 185844, "pid": 5, "tid": 7, "ts": 1716454224354089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295464, "dur": 14,
    "args": {
      "External id": 185844,
      "cbid": 211, "correlation": 185844
    }
  },
  {
    "ph": "s", "id": 185844, "pid": 76337, "tid": -914061504, "ts": 1716454224295464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295535, "dur": 1,
    "args": {
      "External id": 185855,
      "cbid": 251, "correlation": 185855
    }
  },
  {
    "ph": "f", "id": 185855, "pid": 76337, "tid": -914061504, "ts": 1716454224295535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295538, "dur": 0,
    "args": {
      "External id": 185856,
      "cbid": 251, "correlation": 185856
    }
  },
  {
    "ph": "f", "id": 185856, "pid": 76337, "tid": -914061504, "ts": 1716454224295538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224354178, "dur": 12,
    "args": {
      "External id": 185857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185857,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185857, "pid": 5, "tid": 7, "ts": 1716454224354178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295540, "dur": 12,
    "args": {
      "External id": 185857,
      "cbid": 211, "correlation": 185857
    }
  },
  {
    "ph": "s", "id": 185857, "pid": 76337, "tid": -914061504, "ts": 1716454224295540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224354192, "dur": 5,
    "args": {
      "External id": 185859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185859,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185859, "pid": 5, "tid": 7, "ts": 1716454224354192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295553, "dur": 6,
    "args": {
      "External id": 185859,
      "cbid": 211, "correlation": 185859
    }
  },
  {
    "ph": "s", "id": 185859, "pid": 76337, "tid": -914061504, "ts": 1716454224295553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295611, "dur": 1,
    "args": {
      "External id": 185870,
      "cbid": 251, "correlation": 185870
    }
  },
  {
    "ph": "f", "id": 185870, "pid": 76337, "tid": -914061504, "ts": 1716454224295611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295615, "dur": 0,
    "args": {
      "External id": 185871,
      "cbid": 251, "correlation": 185871
    }
  },
  {
    "ph": "f", "id": 185871, "pid": 76337, "tid": -914061504, "ts": 1716454224295615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224354199, "dur": 8,
    "args": {
      "External id": 185872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185872,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185872, "pid": 5, "tid": 7, "ts": 1716454224354199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295616, "dur": 12,
    "args": {
      "External id": 185872,
      "cbid": 211, "correlation": 185872
    }
  },
  {
    "ph": "s", "id": 185872, "pid": 76337, "tid": -914061504, "ts": 1716454224295616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224354208, "dur": 3,
    "args": {
      "External id": 185874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185874,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185874, "pid": 5, "tid": 7, "ts": 1716454224354208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295630, "dur": 6,
    "args": {
      "External id": 185874,
      "cbid": 211, "correlation": 185874
    }
  },
  {
    "ph": "s", "id": 185874, "pid": 76337, "tid": -914061504, "ts": 1716454224295630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224354213, "dur": 54,
    "args": {
      "External id": 185899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185899,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185899, "pid": 5, "tid": 7, "ts": 1716454224354213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295707, "dur": 12,
    "args": {
      "External id": 185899,
      "cbid": 211, "correlation": 185899
    }
  },
  {
    "ph": "s", "id": 185899, "pid": 76337, "tid": -914061504, "ts": 1716454224295707,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295806, "dur": 1,
    "args": {
      "External id": 185917,
      "cbid": 251, "correlation": 185917
    }
  },
  {
    "ph": "f", "id": 185917, "pid": 76337, "tid": -914061504, "ts": 1716454224295806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224354268, "dur": 88,
    "args": {
      "External id": 185919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185919,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 185919, "pid": 5, "tid": 7, "ts": 1716454224354268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295813, "dur": 14,
    "args": {
      "External id": 185919,
      "cbid": 211, "correlation": 185919
    }
  },
  {
    "ph": "s", "id": 185919, "pid": 76337, "tid": -914061504, "ts": 1716454224295813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224354357, "dur": 9,
    "args": {
      "External id": 185927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185927,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185927, "pid": 5, "tid": 7, "ts": 1716454224354357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295882, "dur": 12,
    "args": {
      "External id": 185927,
      "cbid": 211, "correlation": 185927
    }
  },
  {
    "ph": "s", "id": 185927, "pid": 76337, "tid": -914061504, "ts": 1716454224295882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224354367, "dur": 20,
    "args": {
      "External id": 185935,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185935,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185935, "pid": 5, "tid": 7, "ts": 1716454224354367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295923, "dur": 9,
    "args": {
      "External id": 185935,
      "cbid": 211, "correlation": 185935
    }
  },
  {
    "ph": "s", "id": 185935, "pid": 76337, "tid": -914061504, "ts": 1716454224295923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224354389, "dur": 17,
    "args": {
      "External id": 185957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185957,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185957, "pid": 5, "tid": 7, "ts": 1716454224354389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224295982, "dur": 11,
    "args": {
      "External id": 185957,
      "cbid": 211, "correlation": 185957
    }
  },
  {
    "ph": "s", "id": 185957, "pid": 76337, "tid": -914061504, "ts": 1716454224295982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296073, "dur": 1,
    "args": {
      "External id": 185973,
      "cbid": 251, "correlation": 185973
    }
  },
  {
    "ph": "f", "id": 185973, "pid": 76337, "tid": -914061504, "ts": 1716454224296073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296078, "dur": 0,
    "args": {
      "External id": 185975,
      "cbid": 251, "correlation": 185975
    }
  },
  {
    "ph": "f", "id": 185975, "pid": 76337, "tid": -914061504, "ts": 1716454224296078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224354407, "dur": 495,
    "args": {
      "External id": 185976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185976,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 185976, "pid": 5, "tid": 7, "ts": 1716454224354407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296079, "dur": 13,
    "args": {
      "External id": 185976,
      "cbid": 211, "correlation": 185976
    }
  },
  {
    "ph": "s", "id": 185976, "pid": 76337, "tid": -914061504, "ts": 1716454224296079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224354904, "dur": 65,
    "args": {
      "External id": 185984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185984,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185984, "pid": 5, "tid": 7, "ts": 1716454224354904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296148, "dur": 12,
    "args": {
      "External id": 185984,
      "cbid": 211, "correlation": 185984
    }
  },
  {
    "ph": "s", "id": 185984, "pid": 76337, "tid": -914061504, "ts": 1716454224296148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224354970, "dur": 66,
    "args": {
      "External id": 185992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 185992,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 185992, "pid": 5, "tid": 7, "ts": 1716454224354970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296181, "dur": 9,
    "args": {
      "External id": 185992,
      "cbid": 211, "correlation": 185992
    }
  },
  {
    "ph": "s", "id": 185992, "pid": 76337, "tid": -914061504, "ts": 1716454224296181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296262, "dur": 1,
    "args": {
      "External id": 186008,
      "cbid": 251, "correlation": 186008
    }
  },
  {
    "ph": "f", "id": 186008, "pid": 76337, "tid": -914061504, "ts": 1716454224296262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224355039, "dur": 1,
    "args": {
      "External id": 186010,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 186010,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 186010, "pid": 5, "tid": 7, "ts": 1716454224355039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296267, "dur": 10,
    "args": {
      "External id": 186010,
      "cbid": 51, "correlation": 186010
    }
  },
  {
    "ph": "s", "id": 186010, "pid": 76337, "tid": -914061504, "ts": 1716454224296267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224355042, "dur": 267,
    "args": {
      "External id": 186011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186011,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186011, "pid": 5, "tid": 7, "ts": 1716454224355042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296279, "dur": 12,
    "args": {
      "External id": 186011,
      "cbid": 211, "correlation": 186011
    }
  },
  {
    "ph": "s", "id": 186011, "pid": 76337, "tid": -914061504, "ts": 1716454224296279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224355311, "dur": 14,
    "args": {
      "External id": 186019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186019,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186019, "pid": 5, "tid": 7, "ts": 1716454224355311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296322, "dur": 10,
    "args": {
      "External id": 186019,
      "cbid": 211, "correlation": 186019
    }
  },
  {
    "ph": "s", "id": 186019, "pid": 76337, "tid": -914061504, "ts": 1716454224296322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224355326, "dur": 37,
    "args": {
      "External id": 186030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186030, "pid": 5, "tid": 7, "ts": 1716454224355326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296391, "dur": 12,
    "args": {
      "External id": 186030,
      "cbid": 211, "correlation": 186030
    }
  },
  {
    "ph": "s", "id": 186030, "pid": 76337, "tid": -914061504, "ts": 1716454224296391,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296456, "dur": 0,
    "args": {
      "External id": 186042,
      "cbid": 317, "correlation": 186042
    }
  },
  {
    "ph": "f", "id": 186042, "pid": 76337, "tid": -914061504, "ts": 1716454224296456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296457, "dur": 0,
    "args": {
      "External id": 186043,
      "cbid": 203, "correlation": 186043
    }
  },
  {
    "ph": "f", "id": 186043, "pid": 76337, "tid": -914061504, "ts": 1716454224296457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296458, "dur": 0,
    "args": {
      "External id": 186044,
      "cbid": 205, "correlation": 186044
    }
  },
  {
    "ph": "f", "id": 186044, "pid": 76337, "tid": -914061504, "ts": 1716454224296458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224355365, "dur": 13,
    "args": {
      "External id": 186048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186048,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186048, "pid": 5, "tid": 7, "ts": 1716454224355365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296473, "dur": 12,
    "args": {
      "External id": 186048,
      "cbid": 211, "correlation": 186048
    }
  },
  {
    "ph": "s", "id": 186048, "pid": 76337, "tid": -914061504, "ts": 1716454224296473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224355379, "dur": 4,
    "args": {
      "External id": 186050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186050,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186050, "pid": 5, "tid": 7, "ts": 1716454224355379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296490, "dur": 6,
    "args": {
      "External id": 186050,
      "cbid": 211, "correlation": 186050
    }
  },
  {
    "ph": "s", "id": 186050, "pid": 76337, "tid": -914061504, "ts": 1716454224296490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296499, "dur": 0,
    "args": {
      "External id": 186051,
      "cbid": 51, "correlation": 186051
    }
  },
  {
    "ph": "s", "id": 186051, "pid": 76337, "tid": -914061504, "ts": 1716454224296499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224355384, "dur": 96,
    "args": {
      "External id": 186052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186052,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 186052, "pid": 5, "tid": 7, "ts": 1716454224355384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296500, "dur": 5,
    "args": {
      "External id": 186052,
      "cbid": 211, "correlation": 186052
    }
  },
  {
    "ph": "s", "id": 186052, "pid": 76337, "tid": -914061504, "ts": 1716454224296500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224355481, "dur": 16,
    "args": {
      "External id": 186057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186057,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186057, "pid": 5, "tid": 7, "ts": 1716454224355481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296527, "dur": 9,
    "args": {
      "External id": 186057,
      "cbid": 211, "correlation": 186057
    }
  },
  {
    "ph": "s", "id": 186057, "pid": 76337, "tid": -914061504, "ts": 1716454224296527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224355497, "dur": 12,
    "args": {
      "External id": 186065,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186065,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186065, "pid": 5, "tid": 7, "ts": 1716454224355497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296559, "dur": 8,
    "args": {
      "External id": 186065,
      "cbid": 211, "correlation": 186065
    }
  },
  {
    "ph": "s", "id": 186065, "pid": 76337, "tid": -914061504, "ts": 1716454224296559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224355510, "dur": 30,
    "args": {
      "External id": 186074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186074,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186074, "pid": 5, "tid": 7, "ts": 1716454224355510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296600, "dur": 10,
    "args": {
      "External id": 186074,
      "cbid": 211, "correlation": 186074
    }
  },
  {
    "ph": "s", "id": 186074, "pid": 76337, "tid": -914061504, "ts": 1716454224296600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224355542, "dur": 31,
    "args": {
      "External id": 186094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186094,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 186094, "pid": 5, "tid": 7, "ts": 1716454224355542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296671, "dur": 12,
    "args": {
      "External id": 186094,
      "cbid": 211, "correlation": 186094
    }
  },
  {
    "ph": "s", "id": 186094, "pid": 76337, "tid": -914061504, "ts": 1716454224296671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224355574, "dur": 5,
    "args": {
      "External id": 186106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186106,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186106, "pid": 5, "tid": 7, "ts": 1716454224355574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296692, "dur": 7,
    "args": {
      "External id": 186106,
      "cbid": 211, "correlation": 186106
    }
  },
  {
    "ph": "s", "id": 186106, "pid": 76337, "tid": -914061504, "ts": 1716454224296692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224355581, "dur": 30,
    "args": {
      "External id": 186109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186109,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186109, "pid": 5, "tid": 7, "ts": 1716454224355581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296711, "dur": 6,
    "args": {
      "External id": 186109,
      "cbid": 211, "correlation": 186109
    }
  },
  {
    "ph": "s", "id": 186109, "pid": 76337, "tid": -914061504, "ts": 1716454224296711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224355613, "dur": 21,
    "args": {
      "External id": 186118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186118,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186118, "pid": 5, "tid": 7, "ts": 1716454224355613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296750, "dur": 9,
    "args": {
      "External id": 186118,
      "cbid": 211, "correlation": 186118
    }
  },
  {
    "ph": "s", "id": 186118, "pid": 76337, "tid": -914061504, "ts": 1716454224296750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296806, "dur": 0,
    "args": {
      "External id": 186128,
      "cbid": 317, "correlation": 186128
    }
  },
  {
    "ph": "f", "id": 186128, "pid": 76337, "tid": -914061504, "ts": 1716454224296806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296807, "dur": 0,
    "args": {
      "External id": 186129,
      "cbid": 203, "correlation": 186129
    }
  },
  {
    "ph": "f", "id": 186129, "pid": 76337, "tid": -914061504, "ts": 1716454224296807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296808, "dur": 0,
    "args": {
      "External id": 186130,
      "cbid": 205, "correlation": 186130
    }
  },
  {
    "ph": "f", "id": 186130, "pid": 76337, "tid": -914061504, "ts": 1716454224296808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224355635, "dur": 21,
    "args": {
      "External id": 186134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186134, "pid": 5, "tid": 7, "ts": 1716454224355635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296821, "dur": 12,
    "args": {
      "External id": 186134,
      "cbid": 211, "correlation": 186134
    }
  },
  {
    "ph": "s", "id": 186134, "pid": 76337, "tid": -914061504, "ts": 1716454224296821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224355658, "dur": 314,
    "args": {
      "External id": 186136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186136,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186136, "pid": 5, "tid": 7, "ts": 1716454224355658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296835, "dur": 5,
    "args": {
      "External id": 186136,
      "cbid": 211, "correlation": 186136
    }
  },
  {
    "ph": "s", "id": 186136, "pid": 76337, "tid": -914061504, "ts": 1716454224296835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224355975, "dur": 1,
    "args": {
      "External id": 186138,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 186138,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 186138, "pid": 5, "tid": 7, "ts": 1716454224355975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296847, "dur": 8,
    "args": {
      "External id": 186138,
      "cbid": 51, "correlation": 186138
    }
  },
  {
    "ph": "s", "id": 186138, "pid": 76337, "tid": -914061504, "ts": 1716454224296847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224355978, "dur": 1243,
    "args": {
      "External id": 186139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186139,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186139, "pid": 5, "tid": 7, "ts": 1716454224355978,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296855, "dur": 6,
    "args": {
      "External id": 186139,
      "cbid": 211, "correlation": 186139
    }
  },
  {
    "ph": "s", "id": 186139, "pid": 76337, "tid": -914061504, "ts": 1716454224296855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224357223, "dur": 13,
    "args": {
      "External id": 186141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186141,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186141, "pid": 5, "tid": 7, "ts": 1716454224357223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296865, "dur": 5,
    "args": {
      "External id": 186141,
      "cbid": 211, "correlation": 186141
    }
  },
  {
    "ph": "s", "id": 186141, "pid": 76337, "tid": -914061504, "ts": 1716454224296865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224357237, "dur": 14,
    "args": {
      "External id": 186147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186147,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186147, "pid": 5, "tid": 7, "ts": 1716454224357237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296893, "dur": 9,
    "args": {
      "External id": 186147,
      "cbid": 211, "correlation": 186147
    }
  },
  {
    "ph": "s", "id": 186147, "pid": 76337, "tid": -914061504, "ts": 1716454224296893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224357253, "dur": 3,
    "args": {
      "External id": 186155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186155,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 186155, "pid": 5, "tid": 7, "ts": 1716454224357253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224296937, "dur": 9,
    "args": {
      "External id": 186155,
      "cbid": 211, "correlation": 186155
    }
  },
  {
    "ph": "s", "id": 186155, "pid": 76337, "tid": -914061504, "ts": 1716454224296937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297010, "dur": 1,
    "args": {
      "External id": 186171,
      "cbid": 251, "correlation": 186171
    }
  },
  {
    "ph": "f", "id": 186171, "pid": 76337, "tid": -914061504, "ts": 1716454224297010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297015, "dur": 0,
    "args": {
      "External id": 186173,
      "cbid": 251, "correlation": 186173
    }
  },
  {
    "ph": "f", "id": 186173, "pid": 76337, "tid": -914061504, "ts": 1716454224297015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224357257, "dur": 13,
    "args": {
      "External id": 186174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186174,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186174, "pid": 5, "tid": 7, "ts": 1716454224357257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297017, "dur": 11,
    "args": {
      "External id": 186174,
      "cbid": 211, "correlation": 186174
    }
  },
  {
    "ph": "s", "id": 186174, "pid": 76337, "tid": -914061504, "ts": 1716454224297017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224357272, "dur": 5,
    "args": {
      "External id": 186176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186176,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186176, "pid": 5, "tid": 7, "ts": 1716454224357272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297030, "dur": 5,
    "args": {
      "External id": 186176,
      "cbid": 211, "correlation": 186176
    }
  },
  {
    "ph": "s", "id": 186176, "pid": 76337, "tid": -914061504, "ts": 1716454224297030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224357278, "dur": 16,
    "args": {
      "External id": 186186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186186, "pid": 5, "tid": 7, "ts": 1716454224357278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297088, "dur": 12,
    "args": {
      "External id": 186186,
      "cbid": 211, "correlation": 186186
    }
  },
  {
    "ph": "s", "id": 186186, "pid": 76337, "tid": -914061504, "ts": 1716454224297088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224357295, "dur": 18,
    "args": {
      "External id": 186206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186206,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 186206, "pid": 5, "tid": 7, "ts": 1716454224357295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297154, "dur": 11,
    "args": {
      "External id": 186206,
      "cbid": 211, "correlation": 186206
    }
  },
  {
    "ph": "s", "id": 186206, "pid": 76337, "tid": -914061504, "ts": 1716454224297154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224357314, "dur": 4,
    "args": {
      "External id": 186218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186218,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 186218, "pid": 5, "tid": 7, "ts": 1716454224357314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297175, "dur": 6,
    "args": {
      "External id": 186218,
      "cbid": 211, "correlation": 186218
    }
  },
  {
    "ph": "s", "id": 186218, "pid": 76337, "tid": -914061504, "ts": 1716454224297175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224357320, "dur": 16,
    "args": {
      "External id": 186221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186221,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186221, "pid": 5, "tid": 7, "ts": 1716454224357320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297193, "dur": 7,
    "args": {
      "External id": 186221,
      "cbid": 211, "correlation": 186221
    }
  },
  {
    "ph": "s", "id": 186221, "pid": 76337, "tid": -914061504, "ts": 1716454224297193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224357338, "dur": 11,
    "args": {
      "External id": 186230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186230,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186230, "pid": 5, "tid": 7, "ts": 1716454224357338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297233, "dur": 10,
    "args": {
      "External id": 186230,
      "cbid": 211, "correlation": 186230
    }
  },
  {
    "ph": "s", "id": 186230, "pid": 76337, "tid": -914061504, "ts": 1716454224297233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297295, "dur": 0,
    "args": {
      "External id": 186240,
      "cbid": 317, "correlation": 186240
    }
  },
  {
    "ph": "f", "id": 186240, "pid": 76337, "tid": -914061504, "ts": 1716454224297295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297295, "dur": 0,
    "args": {
      "External id": 186241,
      "cbid": 203, "correlation": 186241
    }
  },
  {
    "ph": "f", "id": 186241, "pid": 76337, "tid": -914061504, "ts": 1716454224297295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297296, "dur": 0,
    "args": {
      "External id": 186242,
      "cbid": 205, "correlation": 186242
    }
  },
  {
    "ph": "f", "id": 186242, "pid": 76337, "tid": -914061504, "ts": 1716454224297296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224357349, "dur": 11,
    "args": {
      "External id": 186246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186246,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186246, "pid": 5, "tid": 7, "ts": 1716454224357349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297310, "dur": 12,
    "args": {
      "External id": 186246,
      "cbid": 211, "correlation": 186246
    }
  },
  {
    "ph": "s", "id": 186246, "pid": 76337, "tid": -914061504, "ts": 1716454224297310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224357362, "dur": 161,
    "args": {
      "External id": 186248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186248,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186248, "pid": 5, "tid": 7, "ts": 1716454224357362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297325, "dur": 5,
    "args": {
      "External id": 186248,
      "cbid": 211, "correlation": 186248
    }
  },
  {
    "ph": "s", "id": 186248, "pid": 76337, "tid": -914061504, "ts": 1716454224297325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224357525, "dur": 1,
    "args": {
      "External id": 186250,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 186250,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 186250, "pid": 5, "tid": 7, "ts": 1716454224357525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297335, "dur": 6,
    "args": {
      "External id": 186250,
      "cbid": 51, "correlation": 186250
    }
  },
  {
    "ph": "s", "id": 186250, "pid": 76337, "tid": -914061504, "ts": 1716454224297335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224357529, "dur": 642,
    "args": {
      "External id": 186251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186251,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186251, "pid": 5, "tid": 7, "ts": 1716454224357529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297343, "dur": 6,
    "args": {
      "External id": 186251,
      "cbid": 211, "correlation": 186251
    }
  },
  {
    "ph": "s", "id": 186251, "pid": 76337, "tid": -914061504, "ts": 1716454224297343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224358172, "dur": 13,
    "args": {
      "External id": 186253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186253,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186253, "pid": 5, "tid": 7, "ts": 1716454224358172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297353, "dur": 5,
    "args": {
      "External id": 186253,
      "cbid": 211, "correlation": 186253
    }
  },
  {
    "ph": "s", "id": 186253, "pid": 76337, "tid": -914061504, "ts": 1716454224297353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224358186, "dur": 14,
    "args": {
      "External id": 186259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186259,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186259, "pid": 5, "tid": 7, "ts": 1716454224358186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297382, "dur": 8,
    "args": {
      "External id": 186259,
      "cbid": 211, "correlation": 186259
    }
  },
  {
    "ph": "s", "id": 186259, "pid": 76337, "tid": -914061504, "ts": 1716454224297382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297441, "dur": 0,
    "args": {
      "External id": 186269,
      "cbid": 317, "correlation": 186269
    }
  },
  {
    "ph": "f", "id": 186269, "pid": 76337, "tid": -914061504, "ts": 1716454224297441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297442, "dur": 0,
    "args": {
      "External id": 186270,
      "cbid": 203, "correlation": 186270
    }
  },
  {
    "ph": "f", "id": 186270, "pid": 76337, "tid": -914061504, "ts": 1716454224297442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297442, "dur": 0,
    "args": {
      "External id": 186271,
      "cbid": 205, "correlation": 186271
    }
  },
  {
    "ph": "f", "id": 186271, "pid": 76337, "tid": -914061504, "ts": 1716454224297442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224358202, "dur": 21,
    "args": {
      "External id": 186275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186275,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186275, "pid": 5, "tid": 7, "ts": 1716454224358202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297454, "dur": 12,
    "args": {
      "External id": 186275,
      "cbid": 211, "correlation": 186275
    }
  },
  {
    "ph": "s", "id": 186275, "pid": 76337, "tid": -914061504, "ts": 1716454224297454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224358224, "dur": 4,
    "args": {
      "External id": 186277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186277,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186277, "pid": 5, "tid": 7, "ts": 1716454224358224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297470, "dur": 6,
    "args": {
      "External id": 186277,
      "cbid": 211, "correlation": 186277
    }
  },
  {
    "ph": "s", "id": 186277, "pid": 76337, "tid": -914061504, "ts": 1716454224297470,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297479, "dur": 0,
    "args": {
      "External id": 186278,
      "cbid": 51, "correlation": 186278
    }
  },
  {
    "ph": "s", "id": 186278, "pid": 76337, "tid": -914061504, "ts": 1716454224297479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224358229, "dur": 169,
    "args": {
      "External id": 186279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186279,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 186279, "pid": 5, "tid": 7, "ts": 1716454224358229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297479, "dur": 5,
    "args": {
      "External id": 186279,
      "cbid": 211, "correlation": 186279
    }
  },
  {
    "ph": "s", "id": 186279, "pid": 76337, "tid": -914061504, "ts": 1716454224297479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224358399, "dur": 16,
    "args": {
      "External id": 186284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186284,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186284, "pid": 5, "tid": 7, "ts": 1716454224358399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297504, "dur": 9,
    "args": {
      "External id": 186284,
      "cbid": 211, "correlation": 186284
    }
  },
  {
    "ph": "s", "id": 186284, "pid": 76337, "tid": -914061504, "ts": 1716454224297504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224358416, "dur": 12,
    "args": {
      "External id": 186292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186292,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186292, "pid": 5, "tid": 7, "ts": 1716454224358416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297533, "dur": 8,
    "args": {
      "External id": 186292,
      "cbid": 211, "correlation": 186292
    }
  },
  {
    "ph": "s", "id": 186292, "pid": 76337, "tid": -914061504, "ts": 1716454224297533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224358429, "dur": 10,
    "args": {
      "External id": 186300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186300,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186300, "pid": 5, "tid": 7, "ts": 1716454224358429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297561, "dur": 9,
    "args": {
      "External id": 186300,
      "cbid": 211, "correlation": 186300
    }
  },
  {
    "ph": "s", "id": 186300, "pid": 76337, "tid": -914061504, "ts": 1716454224297561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224358440, "dur": 18,
    "args": {
      "External id": 186320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186320,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 186320, "pid": 5, "tid": 7, "ts": 1716454224358440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297643, "dur": 12,
    "args": {
      "External id": 186320,
      "cbid": 211, "correlation": 186320
    }
  },
  {
    "ph": "s", "id": 186320, "pid": 76337, "tid": -914061504, "ts": 1716454224297643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224358460, "dur": 5,
    "args": {
      "External id": 186332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186332,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 186332, "pid": 5, "tid": 7, "ts": 1716454224358460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297665, "dur": 7,
    "args": {
      "External id": 186332,
      "cbid": 211, "correlation": 186332
    }
  },
  {
    "ph": "s", "id": 186332, "pid": 76337, "tid": -914061504, "ts": 1716454224297665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224358466, "dur": 17,
    "args": {
      "External id": 186335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186335,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186335, "pid": 5, "tid": 7, "ts": 1716454224358466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297683, "dur": 6,
    "args": {
      "External id": 186335,
      "cbid": 211, "correlation": 186335
    }
  },
  {
    "ph": "s", "id": 186335, "pid": 76337, "tid": -914061504, "ts": 1716454224297683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297739, "dur": 0,
    "args": {
      "External id": 186346,
      "cbid": 317, "correlation": 186346
    }
  },
  {
    "ph": "f", "id": 186346, "pid": 76337, "tid": -914061504, "ts": 1716454224297739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297740, "dur": 0,
    "args": {
      "External id": 186347,
      "cbid": 203, "correlation": 186347
    }
  },
  {
    "ph": "f", "id": 186347, "pid": 76337, "tid": -914061504, "ts": 1716454224297740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297741, "dur": 0,
    "args": {
      "External id": 186348,
      "cbid": 205, "correlation": 186348
    }
  },
  {
    "ph": "f", "id": 186348, "pid": 76337, "tid": -914061504, "ts": 1716454224297741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224358484, "dur": 11,
    "args": {
      "External id": 186352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186352,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186352, "pid": 5, "tid": 7, "ts": 1716454224358484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297754, "dur": 12,
    "args": {
      "External id": 186352,
      "cbid": 211, "correlation": 186352
    }
  },
  {
    "ph": "s", "id": 186352, "pid": 76337, "tid": -914061504, "ts": 1716454224297754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224358496, "dur": 4,
    "args": {
      "External id": 186354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186354,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186354, "pid": 5, "tid": 7, "ts": 1716454224358496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297770, "dur": 6,
    "args": {
      "External id": 186354,
      "cbid": 211, "correlation": 186354
    }
  },
  {
    "ph": "s", "id": 186354, "pid": 76337, "tid": -914061504, "ts": 1716454224297770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297779, "dur": 0,
    "args": {
      "External id": 186355,
      "cbid": 51, "correlation": 186355
    }
  },
  {
    "ph": "s", "id": 186355, "pid": 76337, "tid": -914061504, "ts": 1716454224297779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224358501, "dur": 89,
    "args": {
      "External id": 186356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186356,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 186356, "pid": 5, "tid": 7, "ts": 1716454224358501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297779, "dur": 5,
    "args": {
      "External id": 186356,
      "cbid": 211, "correlation": 186356
    }
  },
  {
    "ph": "s", "id": 186356, "pid": 76337, "tid": -914061504, "ts": 1716454224297779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224358591, "dur": 15,
    "args": {
      "External id": 186361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186361, "pid": 5, "tid": 7, "ts": 1716454224358591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297805, "dur": 9,
    "args": {
      "External id": 186361,
      "cbid": 211, "correlation": 186361
    }
  },
  {
    "ph": "s", "id": 186361, "pid": 76337, "tid": -914061504, "ts": 1716454224297805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224358608, "dur": 83,
    "args": {
      "External id": 186370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186370,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186370, "pid": 5, "tid": 7, "ts": 1716454224358608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297887, "dur": 14,
    "args": {
      "External id": 186370,
      "cbid": 211, "correlation": 186370
    }
  },
  {
    "ph": "s", "id": 186370, "pid": 76337, "tid": -914061504, "ts": 1716454224297887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224358692, "dur": 30,
    "args": {
      "External id": 186392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186392,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186392, "pid": 5, "tid": 7, "ts": 1716454224358692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224297944, "dur": 10,
    "args": {
      "External id": 186392,
      "cbid": 211, "correlation": 186392
    }
  },
  {
    "ph": "s", "id": 186392, "pid": 76337, "tid": -914061504, "ts": 1716454224297944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298043, "dur": 1,
    "args": {
      "External id": 186403,
      "cbid": 251, "correlation": 186403
    }
  },
  {
    "ph": "f", "id": 186403, "pid": 76337, "tid": -914061504, "ts": 1716454224298043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224358723, "dur": 162,
    "args": {
      "External id": 186404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186404,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186404, "pid": 5, "tid": 7, "ts": 1716454224358723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298049, "dur": 14,
    "args": {
      "External id": 186404,
      "cbid": 211, "correlation": 186404
    }
  },
  {
    "ph": "s", "id": 186404, "pid": 76337, "tid": -914061504, "ts": 1716454224298049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298120, "dur": 1,
    "args": {
      "External id": 186415,
      "cbid": 251, "correlation": 186415
    }
  },
  {
    "ph": "f", "id": 186415, "pid": 76337, "tid": -914061504, "ts": 1716454224298120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224358886, "dur": 158,
    "args": {
      "External id": 186416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186416,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186416, "pid": 5, "tid": 7, "ts": 1716454224358886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298125, "dur": 11,
    "args": {
      "External id": 186416,
      "cbid": 211, "correlation": 186416
    }
  },
  {
    "ph": "s", "id": 186416, "pid": 76337, "tid": -914061504, "ts": 1716454224298125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298190, "dur": 1,
    "args": {
      "External id": 186427,
      "cbid": 251, "correlation": 186427
    }
  },
  {
    "ph": "f", "id": 186427, "pid": 76337, "tid": -914061504, "ts": 1716454224298190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224359046, "dur": 157,
    "args": {
      "External id": 186428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186428,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186428, "pid": 5, "tid": 7, "ts": 1716454224359046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298194, "dur": 11,
    "args": {
      "External id": 186428,
      "cbid": 211, "correlation": 186428
    }
  },
  {
    "ph": "s", "id": 186428, "pid": 76337, "tid": -914061504, "ts": 1716454224298194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224359204, "dur": 332,
    "args": {
      "External id": 186453,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186453,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186453, "pid": 5, "tid": 7, "ts": 1716454224359204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298278, "dur": 13,
    "args": {
      "External id": 186453,
      "cbid": 211, "correlation": 186453
    }
  },
  {
    "ph": "s", "id": 186453, "pid": 76337, "tid": -914061504, "ts": 1716454224298278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298377, "dur": 1,
    "args": {
      "External id": 186471,
      "cbid": 251, "correlation": 186471
    }
  },
  {
    "ph": "f", "id": 186471, "pid": 76337, "tid": -914061504, "ts": 1716454224298377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224359538, "dur": 163,
    "args": {
      "External id": 186473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186473,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186473, "pid": 5, "tid": 7, "ts": 1716454224359538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298383, "dur": 13,
    "args": {
      "External id": 186473,
      "cbid": 211, "correlation": 186473
    }
  },
  {
    "ph": "s", "id": 186473, "pid": 76337, "tid": -914061504, "ts": 1716454224298383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224359703, "dur": 19,
    "args": {
      "External id": 186481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186481,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186481, "pid": 5, "tid": 7, "ts": 1716454224359703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298454, "dur": 12,
    "args": {
      "External id": 186481,
      "cbid": 211, "correlation": 186481
    }
  },
  {
    "ph": "s", "id": 186481, "pid": 76337, "tid": -914061504, "ts": 1716454224298454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224359723, "dur": 27,
    "args": {
      "External id": 186489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186489,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186489, "pid": 5, "tid": 7, "ts": 1716454224359723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298493, "dur": 8,
    "args": {
      "External id": 186489,
      "cbid": 211, "correlation": 186489
    }
  },
  {
    "ph": "s", "id": 186489, "pid": 76337, "tid": -914061504, "ts": 1716454224298493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224359752, "dur": 19,
    "args": {
      "External id": 186500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186500, "pid": 5, "tid": 7, "ts": 1716454224359752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298565, "dur": 12,
    "args": {
      "External id": 186500,
      "cbid": 211, "correlation": 186500
    }
  },
  {
    "ph": "s", "id": 186500, "pid": 76337, "tid": -914061504, "ts": 1716454224298565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224359772, "dur": 16,
    "args": {
      "External id": 186522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186522,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186522, "pid": 5, "tid": 7, "ts": 1716454224359772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298596, "dur": 8,
    "args": {
      "External id": 186522,
      "cbid": 211, "correlation": 186522
    }
  },
  {
    "ph": "s", "id": 186522, "pid": 76337, "tid": -914061504, "ts": 1716454224298596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298681, "dur": 1,
    "args": {
      "External id": 186533,
      "cbid": 251, "correlation": 186533
    }
  },
  {
    "ph": "f", "id": 186533, "pid": 76337, "tid": -914061504, "ts": 1716454224298681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224359790, "dur": 87,
    "args": {
      "External id": 186534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186534,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186534, "pid": 5, "tid": 7, "ts": 1716454224359790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298687, "dur": 13,
    "args": {
      "External id": 186534,
      "cbid": 211, "correlation": 186534
    }
  },
  {
    "ph": "s", "id": 186534, "pid": 76337, "tid": -914061504, "ts": 1716454224298687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298756, "dur": 1,
    "args": {
      "External id": 186545,
      "cbid": 251, "correlation": 186545
    }
  },
  {
    "ph": "f", "id": 186545, "pid": 76337, "tid": -914061504, "ts": 1716454224298756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298760, "dur": 0,
    "args": {
      "External id": 186546,
      "cbid": 251, "correlation": 186546
    }
  },
  {
    "ph": "f", "id": 186546, "pid": 76337, "tid": -914061504, "ts": 1716454224298760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224359878, "dur": 12,
    "args": {
      "External id": 186547,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186547,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186547, "pid": 5, "tid": 7, "ts": 1716454224359878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298761, "dur": 12,
    "args": {
      "External id": 186547,
      "cbid": 211, "correlation": 186547
    }
  },
  {
    "ph": "s", "id": 186547, "pid": 76337, "tid": -914061504, "ts": 1716454224298761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224359892, "dur": 5,
    "args": {
      "External id": 186549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186549,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186549, "pid": 5, "tid": 7, "ts": 1716454224359892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298775, "dur": 6,
    "args": {
      "External id": 186549,
      "cbid": 211, "correlation": 186549
    }
  },
  {
    "ph": "s", "id": 186549, "pid": 76337, "tid": -914061504, "ts": 1716454224298775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298833, "dur": 1,
    "args": {
      "External id": 186560,
      "cbid": 251, "correlation": 186560
    }
  },
  {
    "ph": "f", "id": 186560, "pid": 76337, "tid": -914061504, "ts": 1716454224298833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298836, "dur": 0,
    "args": {
      "External id": 186561,
      "cbid": 251, "correlation": 186561
    }
  },
  {
    "ph": "f", "id": 186561, "pid": 76337, "tid": -914061504, "ts": 1716454224298836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224359899, "dur": 8,
    "args": {
      "External id": 186562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186562,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186562, "pid": 5, "tid": 7, "ts": 1716454224359899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298838, "dur": 12,
    "args": {
      "External id": 186562,
      "cbid": 211, "correlation": 186562
    }
  },
  {
    "ph": "s", "id": 186562, "pid": 76337, "tid": -914061504, "ts": 1716454224298838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224359908, "dur": 3,
    "args": {
      "External id": 186564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186564,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186564, "pid": 5, "tid": 7, "ts": 1716454224359908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298851, "dur": 5,
    "args": {
      "External id": 186564,
      "cbid": 211, "correlation": 186564
    }
  },
  {
    "ph": "s", "id": 186564, "pid": 76337, "tid": -914061504, "ts": 1716454224298851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224359913, "dur": 54,
    "args": {
      "External id": 186589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186589,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186589, "pid": 5, "tid": 7, "ts": 1716454224359913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224298927, "dur": 13,
    "args": {
      "External id": 186589,
      "cbid": 211, "correlation": 186589
    }
  },
  {
    "ph": "s", "id": 186589, "pid": 76337, "tid": -914061504, "ts": 1716454224298927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299036, "dur": 1,
    "args": {
      "External id": 186607,
      "cbid": 251, "correlation": 186607
    }
  },
  {
    "ph": "f", "id": 186607, "pid": 76337, "tid": -914061504, "ts": 1716454224299036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224359968, "dur": 89,
    "args": {
      "External id": 186609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186609,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186609, "pid": 5, "tid": 7, "ts": 1716454224359968,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299042, "dur": 14,
    "args": {
      "External id": 186609,
      "cbid": 211, "correlation": 186609
    }
  },
  {
    "ph": "s", "id": 186609, "pid": 76337, "tid": -914061504, "ts": 1716454224299042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224360059, "dur": 10,
    "args": {
      "External id": 186617,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186617,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186617, "pid": 5, "tid": 7, "ts": 1716454224360059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299112, "dur": 12,
    "args": {
      "External id": 186617,
      "cbid": 211, "correlation": 186617
    }
  },
  {
    "ph": "s", "id": 186617, "pid": 76337, "tid": -914061504, "ts": 1716454224299112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224360070, "dur": 21,
    "args": {
      "External id": 186625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186625,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186625, "pid": 5, "tid": 7, "ts": 1716454224360070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299153, "dur": 9,
    "args": {
      "External id": 186625,
      "cbid": 211, "correlation": 186625
    }
  },
  {
    "ph": "s", "id": 186625, "pid": 76337, "tid": -914061504, "ts": 1716454224299153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224360092, "dur": 17,
    "args": {
      "External id": 186647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186647,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186647, "pid": 5, "tid": 7, "ts": 1716454224360092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299204, "dur": 10,
    "args": {
      "External id": 186647,
      "cbid": 211, "correlation": 186647
    }
  },
  {
    "ph": "s", "id": 186647, "pid": 76337, "tid": -914061504, "ts": 1716454224299204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299290, "dur": 1,
    "args": {
      "External id": 186663,
      "cbid": 251, "correlation": 186663
    }
  },
  {
    "ph": "f", "id": 186663, "pid": 76337, "tid": -914061504, "ts": 1716454224299290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299296, "dur": 0,
    "args": {
      "External id": 186665,
      "cbid": 251, "correlation": 186665
    }
  },
  {
    "ph": "f", "id": 186665, "pid": 76337, "tid": -914061504, "ts": 1716454224299296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224360110, "dur": 494,
    "args": {
      "External id": 186666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186666,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186666, "pid": 5, "tid": 7, "ts": 1716454224360110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299297, "dur": 13,
    "args": {
      "External id": 186666,
      "cbid": 211, "correlation": 186666
    }
  },
  {
    "ph": "s", "id": 186666, "pid": 76337, "tid": -914061504, "ts": 1716454224299297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224360605, "dur": 66,
    "args": {
      "External id": 186674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186674, "pid": 5, "tid": 7, "ts": 1716454224360605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299363, "dur": 12,
    "args": {
      "External id": 186674,
      "cbid": 211, "correlation": 186674
    }
  },
  {
    "ph": "s", "id": 186674, "pid": 76337, "tid": -914061504, "ts": 1716454224299363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224360673, "dur": 66,
    "args": {
      "External id": 186682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186682,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186682, "pid": 5, "tid": 7, "ts": 1716454224360673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299394, "dur": 8,
    "args": {
      "External id": 186682,
      "cbid": 211, "correlation": 186682
    }
  },
  {
    "ph": "s", "id": 186682, "pid": 76337, "tid": -914061504, "ts": 1716454224299394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299472, "dur": 1,
    "args": {
      "External id": 186698,
      "cbid": 251, "correlation": 186698
    }
  },
  {
    "ph": "f", "id": 186698, "pid": 76337, "tid": -914061504, "ts": 1716454224299472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224360740, "dur": 1,
    "args": {
      "External id": 186700,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 186700,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 186700, "pid": 5, "tid": 7, "ts": 1716454224360740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299477, "dur": 9,
    "args": {
      "External id": 186700,
      "cbid": 51, "correlation": 186700
    }
  },
  {
    "ph": "s", "id": 186700, "pid": 76337, "tid": -914061504, "ts": 1716454224299477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224360744, "dur": 265,
    "args": {
      "External id": 186701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186701,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186701, "pid": 5, "tid": 7, "ts": 1716454224360744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299487, "dur": 11,
    "args": {
      "External id": 186701,
      "cbid": 211, "correlation": 186701
    }
  },
  {
    "ph": "s", "id": 186701, "pid": 76337, "tid": -914061504, "ts": 1716454224299487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224361010, "dur": 14,
    "args": {
      "External id": 186709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186709,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186709, "pid": 5, "tid": 7, "ts": 1716454224361010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299529, "dur": 11,
    "args": {
      "External id": 186709,
      "cbid": 211, "correlation": 186709
    }
  },
  {
    "ph": "s", "id": 186709, "pid": 76337, "tid": -914061504, "ts": 1716454224299529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224361025, "dur": 38,
    "args": {
      "External id": 186720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186720,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186720, "pid": 5, "tid": 7, "ts": 1716454224361025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299598, "dur": 12,
    "args": {
      "External id": 186720,
      "cbid": 211, "correlation": 186720
    }
  },
  {
    "ph": "s", "id": 186720, "pid": 76337, "tid": -914061504, "ts": 1716454224299598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299662, "dur": 0,
    "args": {
      "External id": 186732,
      "cbid": 317, "correlation": 186732
    }
  },
  {
    "ph": "f", "id": 186732, "pid": 76337, "tid": -914061504, "ts": 1716454224299662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299663, "dur": 0,
    "args": {
      "External id": 186733,
      "cbid": 203, "correlation": 186733
    }
  },
  {
    "ph": "f", "id": 186733, "pid": 76337, "tid": -914061504, "ts": 1716454224299663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299664, "dur": 0,
    "args": {
      "External id": 186734,
      "cbid": 205, "correlation": 186734
    }
  },
  {
    "ph": "f", "id": 186734, "pid": 76337, "tid": -914061504, "ts": 1716454224299664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224361065, "dur": 14,
    "args": {
      "External id": 186738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186738,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186738, "pid": 5, "tid": 7, "ts": 1716454224361065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299679, "dur": 12,
    "args": {
      "External id": 186738,
      "cbid": 211, "correlation": 186738
    }
  },
  {
    "ph": "s", "id": 186738, "pid": 76337, "tid": -914061504, "ts": 1716454224299679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224361080, "dur": 4,
    "args": {
      "External id": 186740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186740, "pid": 5, "tid": 7, "ts": 1716454224361080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299696, "dur": 6,
    "args": {
      "External id": 186740,
      "cbid": 211, "correlation": 186740
    }
  },
  {
    "ph": "s", "id": 186740, "pid": 76337, "tid": -914061504, "ts": 1716454224299696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299704, "dur": 0,
    "args": {
      "External id": 186741,
      "cbid": 51, "correlation": 186741
    }
  },
  {
    "ph": "s", "id": 186741, "pid": 76337, "tid": -914061504, "ts": 1716454224299704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224361085, "dur": 94,
    "args": {
      "External id": 186742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186742,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 186742, "pid": 5, "tid": 7, "ts": 1716454224361085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299705, "dur": 5,
    "args": {
      "External id": 186742,
      "cbid": 211, "correlation": 186742
    }
  },
  {
    "ph": "s", "id": 186742, "pid": 76337, "tid": -914061504, "ts": 1716454224299705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224361181, "dur": 16,
    "args": {
      "External id": 186747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186747,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186747, "pid": 5, "tid": 7, "ts": 1716454224361181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299732, "dur": 9,
    "args": {
      "External id": 186747,
      "cbid": 211, "correlation": 186747
    }
  },
  {
    "ph": "s", "id": 186747, "pid": 76337, "tid": -914061504, "ts": 1716454224299732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224361198, "dur": 13,
    "args": {
      "External id": 186755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186755,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186755, "pid": 5, "tid": 7, "ts": 1716454224361198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299764, "dur": 8,
    "args": {
      "External id": 186755,
      "cbid": 211, "correlation": 186755
    }
  },
  {
    "ph": "s", "id": 186755, "pid": 76337, "tid": -914061504, "ts": 1716454224299764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224361213, "dur": 25,
    "args": {
      "External id": 186764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186764,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186764, "pid": 5, "tid": 7, "ts": 1716454224361213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299802, "dur": 11,
    "args": {
      "External id": 186764,
      "cbid": 211, "correlation": 186764
    }
  },
  {
    "ph": "s", "id": 186764, "pid": 76337, "tid": -914061504, "ts": 1716454224299802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224361239, "dur": 24,
    "args": {
      "External id": 186784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186784,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 186784, "pid": 5, "tid": 7, "ts": 1716454224361239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299873, "dur": 11,
    "args": {
      "External id": 186784,
      "cbid": 211, "correlation": 186784
    }
  },
  {
    "ph": "s", "id": 186784, "pid": 76337, "tid": -914061504, "ts": 1716454224299873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224361264, "dur": 5,
    "args": {
      "External id": 186796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186796,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 186796, "pid": 5, "tid": 7, "ts": 1716454224361264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299895, "dur": 7,
    "args": {
      "External id": 186796,
      "cbid": 211, "correlation": 186796
    }
  },
  {
    "ph": "s", "id": 186796, "pid": 76337, "tid": -914061504, "ts": 1716454224299895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224361270, "dur": 25,
    "args": {
      "External id": 186799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186799,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186799, "pid": 5, "tid": 7, "ts": 1716454224361270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299913, "dur": 6,
    "args": {
      "External id": 186799,
      "cbid": 211, "correlation": 186799
    }
  },
  {
    "ph": "s", "id": 186799, "pid": 76337, "tid": -914061504, "ts": 1716454224299913,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224361296, "dur": 17,
    "args": {
      "External id": 186808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186808,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186808, "pid": 5, "tid": 7, "ts": 1716454224361296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224299953, "dur": 9,
    "args": {
      "External id": 186808,
      "cbid": 211, "correlation": 186808
    }
  },
  {
    "ph": "s", "id": 186808, "pid": 76337, "tid": -914061504, "ts": 1716454224299953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300017, "dur": 0,
    "args": {
      "External id": 186818,
      "cbid": 317, "correlation": 186818
    }
  },
  {
    "ph": "f", "id": 186818, "pid": 76337, "tid": -914061504, "ts": 1716454224300017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300018, "dur": 0,
    "args": {
      "External id": 186819,
      "cbid": 203, "correlation": 186819
    }
  },
  {
    "ph": "f", "id": 186819, "pid": 76337, "tid": -914061504, "ts": 1716454224300018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300018, "dur": 0,
    "args": {
      "External id": 186820,
      "cbid": 205, "correlation": 186820
    }
  },
  {
    "ph": "f", "id": 186820, "pid": 76337, "tid": -914061504, "ts": 1716454224300018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224361315, "dur": 17,
    "args": {
      "External id": 186824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186824,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186824, "pid": 5, "tid": 7, "ts": 1716454224361315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300034, "dur": 12,
    "args": {
      "External id": 186824,
      "cbid": 211, "correlation": 186824
    }
  },
  {
    "ph": "s", "id": 186824, "pid": 76337, "tid": -914061504, "ts": 1716454224300034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224361333, "dur": 238,
    "args": {
      "External id": 186826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186826,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186826, "pid": 5, "tid": 7, "ts": 1716454224361333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300049, "dur": 5,
    "args": {
      "External id": 186826,
      "cbid": 211, "correlation": 186826
    }
  },
  {
    "ph": "s", "id": 186826, "pid": 76337, "tid": -914061504, "ts": 1716454224300049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224361573, "dur": 1,
    "args": {
      "External id": 186828,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 186828,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 186828, "pid": 5, "tid": 7, "ts": 1716454224361573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300060, "dur": 8,
    "args": {
      "External id": 186828,
      "cbid": 51, "correlation": 186828
    }
  },
  {
    "ph": "s", "id": 186828, "pid": 76337, "tid": -914061504, "ts": 1716454224300060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224361577, "dur": 806,
    "args": {
      "External id": 186829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186829,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186829, "pid": 5, "tid": 7, "ts": 1716454224361577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300069, "dur": 6,
    "args": {
      "External id": 186829,
      "cbid": 211, "correlation": 186829
    }
  },
  {
    "ph": "s", "id": 186829, "pid": 76337, "tid": -914061504, "ts": 1716454224300069,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224362384, "dur": 13,
    "args": {
      "External id": 186831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186831,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186831, "pid": 5, "tid": 7, "ts": 1716454224362384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300080, "dur": 5,
    "args": {
      "External id": 186831,
      "cbid": 211, "correlation": 186831
    }
  },
  {
    "ph": "s", "id": 186831, "pid": 76337, "tid": -914061504, "ts": 1716454224300080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224362399, "dur": 14,
    "args": {
      "External id": 186837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186837,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186837, "pid": 5, "tid": 7, "ts": 1716454224362399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300108, "dur": 8,
    "args": {
      "External id": 186837,
      "cbid": 211, "correlation": 186837
    }
  },
  {
    "ph": "s", "id": 186837, "pid": 76337, "tid": -914061504, "ts": 1716454224300108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224362414, "dur": 3,
    "args": {
      "External id": 186845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186845,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 186845, "pid": 5, "tid": 7, "ts": 1716454224362414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300153, "dur": 9,
    "args": {
      "External id": 186845,
      "cbid": 211, "correlation": 186845
    }
  },
  {
    "ph": "s", "id": 186845, "pid": 76337, "tid": -914061504, "ts": 1716454224300153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300218, "dur": 1,
    "args": {
      "External id": 186861,
      "cbid": 251, "correlation": 186861
    }
  },
  {
    "ph": "f", "id": 186861, "pid": 76337, "tid": -914061504, "ts": 1716454224300218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300224, "dur": 0,
    "args": {
      "External id": 186863,
      "cbid": 251, "correlation": 186863
    }
  },
  {
    "ph": "f", "id": 186863, "pid": 76337, "tid": -914061504, "ts": 1716454224300224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224362419, "dur": 14,
    "args": {
      "External id": 186864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186864,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186864, "pid": 5, "tid": 7, "ts": 1716454224362419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300225, "dur": 11,
    "args": {
      "External id": 186864,
      "cbid": 211, "correlation": 186864
    }
  },
  {
    "ph": "s", "id": 186864, "pid": 76337, "tid": -914061504, "ts": 1716454224300225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224362434, "dur": 5,
    "args": {
      "External id": 186866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186866,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186866, "pid": 5, "tid": 7, "ts": 1716454224362434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300238, "dur": 6,
    "args": {
      "External id": 186866,
      "cbid": 211, "correlation": 186866
    }
  },
  {
    "ph": "s", "id": 186866, "pid": 76337, "tid": -914061504, "ts": 1716454224300238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224362441, "dur": 16,
    "args": {
      "External id": 186876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186876, "pid": 5, "tid": 7, "ts": 1716454224362441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300297, "dur": 12,
    "args": {
      "External id": 186876,
      "cbid": 211, "correlation": 186876
    }
  },
  {
    "ph": "s", "id": 186876, "pid": 76337, "tid": -914061504, "ts": 1716454224300297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224362458, "dur": 19,
    "args": {
      "External id": 186896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186896,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 186896, "pid": 5, "tid": 7, "ts": 1716454224362458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300364, "dur": 10,
    "args": {
      "External id": 186896,
      "cbid": 211, "correlation": 186896
    }
  },
  {
    "ph": "s", "id": 186896, "pid": 76337, "tid": -914061504, "ts": 1716454224300364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224362478, "dur": 4,
    "args": {
      "External id": 186908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186908,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 186908, "pid": 5, "tid": 7, "ts": 1716454224362478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300384, "dur": 7,
    "args": {
      "External id": 186908,
      "cbid": 211, "correlation": 186908
    }
  },
  {
    "ph": "s", "id": 186908, "pid": 76337, "tid": -914061504, "ts": 1716454224300384,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224362484, "dur": 16,
    "args": {
      "External id": 186911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186911,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186911, "pid": 5, "tid": 7, "ts": 1716454224362484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300403, "dur": 7,
    "args": {
      "External id": 186911,
      "cbid": 211, "correlation": 186911
    }
  },
  {
    "ph": "s", "id": 186911, "pid": 76337, "tid": -914061504, "ts": 1716454224300403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224362501, "dur": 10,
    "args": {
      "External id": 186920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186920,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186920, "pid": 5, "tid": 7, "ts": 1716454224362501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300444, "dur": 10,
    "args": {
      "External id": 186920,
      "cbid": 211, "correlation": 186920
    }
  },
  {
    "ph": "s", "id": 186920, "pid": 76337, "tid": -914061504, "ts": 1716454224300444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300506, "dur": 0,
    "args": {
      "External id": 186930,
      "cbid": 317, "correlation": 186930
    }
  },
  {
    "ph": "f", "id": 186930, "pid": 76337, "tid": -914061504, "ts": 1716454224300506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300507, "dur": 0,
    "args": {
      "External id": 186931,
      "cbid": 203, "correlation": 186931
    }
  },
  {
    "ph": "f", "id": 186931, "pid": 76337, "tid": -914061504, "ts": 1716454224300507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300508, "dur": 0,
    "args": {
      "External id": 186932,
      "cbid": 205, "correlation": 186932
    }
  },
  {
    "ph": "f", "id": 186932, "pid": 76337, "tid": -914061504, "ts": 1716454224300508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224362513, "dur": 12,
    "args": {
      "External id": 186936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186936,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186936, "pid": 5, "tid": 7, "ts": 1716454224362513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300521, "dur": 11,
    "args": {
      "External id": 186936,
      "cbid": 211, "correlation": 186936
    }
  },
  {
    "ph": "s", "id": 186936, "pid": 76337, "tid": -914061504, "ts": 1716454224300521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224362525, "dur": 159,
    "args": {
      "External id": 186938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186938,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186938, "pid": 5, "tid": 7, "ts": 1716454224362525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300535, "dur": 5,
    "args": {
      "External id": 186938,
      "cbid": 211, "correlation": 186938
    }
  },
  {
    "ph": "s", "id": 186938, "pid": 76337, "tid": -914061504, "ts": 1716454224300535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224362687, "dur": 1,
    "args": {
      "External id": 186940,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 186940,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 186940, "pid": 5, "tid": 7, "ts": 1716454224362687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300546, "dur": 7,
    "args": {
      "External id": 186940,
      "cbid": 51, "correlation": 186940
    }
  },
  {
    "ph": "s", "id": 186940, "pid": 76337, "tid": -914061504, "ts": 1716454224300546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224362691, "dur": 641,
    "args": {
      "External id": 186941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186941,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 186941, "pid": 5, "tid": 7, "ts": 1716454224362691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300554, "dur": 6,
    "args": {
      "External id": 186941,
      "cbid": 211, "correlation": 186941
    }
  },
  {
    "ph": "s", "id": 186941, "pid": 76337, "tid": -914061504, "ts": 1716454224300554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224363333, "dur": 12,
    "args": {
      "External id": 186943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186943,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186943, "pid": 5, "tid": 7, "ts": 1716454224363333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300565, "dur": 5,
    "args": {
      "External id": 186943,
      "cbid": 211, "correlation": 186943
    }
  },
  {
    "ph": "s", "id": 186943, "pid": 76337, "tid": -914061504, "ts": 1716454224300565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224363347, "dur": 14,
    "args": {
      "External id": 186949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186949,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186949, "pid": 5, "tid": 7, "ts": 1716454224363347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300593, "dur": 10,
    "args": {
      "External id": 186949,
      "cbid": 211, "correlation": 186949
    }
  },
  {
    "ph": "s", "id": 186949, "pid": 76337, "tid": -914061504, "ts": 1716454224300593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300652, "dur": 0,
    "args": {
      "External id": 186959,
      "cbid": 317, "correlation": 186959
    }
  },
  {
    "ph": "f", "id": 186959, "pid": 76337, "tid": -914061504, "ts": 1716454224300652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300653, "dur": 0,
    "args": {
      "External id": 186960,
      "cbid": 203, "correlation": 186960
    }
  },
  {
    "ph": "f", "id": 186960, "pid": 76337, "tid": -914061504, "ts": 1716454224300653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300654, "dur": 0,
    "args": {
      "External id": 186961,
      "cbid": 205, "correlation": 186961
    }
  },
  {
    "ph": "f", "id": 186961, "pid": 76337, "tid": -914061504, "ts": 1716454224300654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224363362, "dur": 17,
    "args": {
      "External id": 186965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186965,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186965, "pid": 5, "tid": 7, "ts": 1716454224363362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300668, "dur": 12,
    "args": {
      "External id": 186965,
      "cbid": 211, "correlation": 186965
    }
  },
  {
    "ph": "s", "id": 186965, "pid": 76337, "tid": -914061504, "ts": 1716454224300668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224363380, "dur": 4,
    "args": {
      "External id": 186967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186967,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 186967, "pid": 5, "tid": 7, "ts": 1716454224363380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300685, "dur": 6,
    "args": {
      "External id": 186967,
      "cbid": 211, "correlation": 186967
    }
  },
  {
    "ph": "s", "id": 186967, "pid": 76337, "tid": -914061504, "ts": 1716454224300685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300693, "dur": 0,
    "args": {
      "External id": 186968,
      "cbid": 51, "correlation": 186968
    }
  },
  {
    "ph": "s", "id": 186968, "pid": 76337, "tid": -914061504, "ts": 1716454224300693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224363385, "dur": 128,
    "args": {
      "External id": 186969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186969,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 186969, "pid": 5, "tid": 7, "ts": 1716454224363385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300694, "dur": 5,
    "args": {
      "External id": 186969,
      "cbid": 211, "correlation": 186969
    }
  },
  {
    "ph": "s", "id": 186969, "pid": 76337, "tid": -914061504, "ts": 1716454224300694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224363515, "dur": 15,
    "args": {
      "External id": 186974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186974,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186974, "pid": 5, "tid": 7, "ts": 1716454224363515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300720, "dur": 8,
    "args": {
      "External id": 186974,
      "cbid": 211, "correlation": 186974
    }
  },
  {
    "ph": "s", "id": 186974, "pid": 76337, "tid": -914061504, "ts": 1716454224300720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224363532, "dur": 13,
    "args": {
      "External id": 186982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186982,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186982, "pid": 5, "tid": 7, "ts": 1716454224363532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300749, "dur": 8,
    "args": {
      "External id": 186982,
      "cbid": 211, "correlation": 186982
    }
  },
  {
    "ph": "s", "id": 186982, "pid": 76337, "tid": -914061504, "ts": 1716454224300749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224363546, "dur": 10,
    "args": {
      "External id": 186990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 186990,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 186990, "pid": 5, "tid": 7, "ts": 1716454224363546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300778, "dur": 8,
    "args": {
      "External id": 186990,
      "cbid": 211, "correlation": 186990
    }
  },
  {
    "ph": "s", "id": 186990, "pid": 76337, "tid": -914061504, "ts": 1716454224300778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224363557, "dur": 18,
    "args": {
      "External id": 187010,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187010,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 187010, "pid": 5, "tid": 7, "ts": 1716454224363557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300865, "dur": 12,
    "args": {
      "External id": 187010,
      "cbid": 211, "correlation": 187010
    }
  },
  {
    "ph": "s", "id": 187010, "pid": 76337, "tid": -914061504, "ts": 1716454224300865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224363577, "dur": 4,
    "args": {
      "External id": 187022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187022,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 187022, "pid": 5, "tid": 7, "ts": 1716454224363577,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300887, "dur": 6,
    "args": {
      "External id": 187022,
      "cbid": 211, "correlation": 187022
    }
  },
  {
    "ph": "s", "id": 187022, "pid": 76337, "tid": -914061504, "ts": 1716454224300887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224363583, "dur": 17,
    "args": {
      "External id": 187025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187025, "pid": 5, "tid": 7, "ts": 1716454224363583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300905, "dur": 7,
    "args": {
      "External id": 187025,
      "cbid": 211, "correlation": 187025
    }
  },
  {
    "ph": "s", "id": 187025, "pid": 76337, "tid": -914061504, "ts": 1716454224300905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300962, "dur": 0,
    "args": {
      "External id": 187036,
      "cbid": 317, "correlation": 187036
    }
  },
  {
    "ph": "f", "id": 187036, "pid": 76337, "tid": -914061504, "ts": 1716454224300962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300963, "dur": 0,
    "args": {
      "External id": 187037,
      "cbid": 203, "correlation": 187037
    }
  },
  {
    "ph": "f", "id": 187037, "pid": 76337, "tid": -914061504, "ts": 1716454224300963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300964, "dur": 0,
    "args": {
      "External id": 187038,
      "cbid": 205, "correlation": 187038
    }
  },
  {
    "ph": "f", "id": 187038, "pid": 76337, "tid": -914061504, "ts": 1716454224300964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224363601, "dur": 11,
    "args": {
      "External id": 187042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187042,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187042, "pid": 5, "tid": 7, "ts": 1716454224363601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224300985, "dur": 12,
    "args": {
      "External id": 187042,
      "cbid": 211, "correlation": 187042
    }
  },
  {
    "ph": "s", "id": 187042, "pid": 76337, "tid": -914061504, "ts": 1716454224300985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224363613, "dur": 3,
    "args": {
      "External id": 187044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187044,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 187044, "pid": 5, "tid": 7, "ts": 1716454224363613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301002, "dur": 6,
    "args": {
      "External id": 187044,
      "cbid": 211, "correlation": 187044
    }
  },
  {
    "ph": "s", "id": 187044, "pid": 76337, "tid": -914061504, "ts": 1716454224301002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301010, "dur": 0,
    "args": {
      "External id": 187045,
      "cbid": 51, "correlation": 187045
    }
  },
  {
    "ph": "s", "id": 187045, "pid": 76337, "tid": -914061504, "ts": 1716454224301010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224363618, "dur": 89,
    "args": {
      "External id": 187046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187046,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 187046, "pid": 5, "tid": 7, "ts": 1716454224363618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301011, "dur": 6,
    "args": {
      "External id": 187046,
      "cbid": 211, "correlation": 187046
    }
  },
  {
    "ph": "s", "id": 187046, "pid": 76337, "tid": -914061504, "ts": 1716454224301011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224363708, "dur": 15,
    "args": {
      "External id": 187051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187051,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187051, "pid": 5, "tid": 7, "ts": 1716454224363708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301039, "dur": 8,
    "args": {
      "External id": 187051,
      "cbid": 211, "correlation": 187051
    }
  },
  {
    "ph": "s", "id": 187051, "pid": 76337, "tid": -914061504, "ts": 1716454224301039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224363724, "dur": 83,
    "args": {
      "External id": 187060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187060,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187060, "pid": 5, "tid": 7, "ts": 1716454224363724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301124, "dur": 14,
    "args": {
      "External id": 187060,
      "cbid": 211, "correlation": 187060
    }
  },
  {
    "ph": "s", "id": 187060, "pid": 76337, "tid": -914061504, "ts": 1716454224301124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224363808, "dur": 29,
    "args": {
      "External id": 187082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187082,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187082, "pid": 5, "tid": 7, "ts": 1716454224363808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301183, "dur": 10,
    "args": {
      "External id": 187082,
      "cbid": 211, "correlation": 187082
    }
  },
  {
    "ph": "s", "id": 187082, "pid": 76337, "tid": -914061504, "ts": 1716454224301183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301281, "dur": 1,
    "args": {
      "External id": 187093,
      "cbid": 251, "correlation": 187093
    }
  },
  {
    "ph": "f", "id": 187093, "pid": 76337, "tid": -914061504, "ts": 1716454224301281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224363839, "dur": 162,
    "args": {
      "External id": 187094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187094,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187094, "pid": 5, "tid": 7, "ts": 1716454224363839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301287, "dur": 13,
    "args": {
      "External id": 187094,
      "cbid": 211, "correlation": 187094
    }
  },
  {
    "ph": "s", "id": 187094, "pid": 76337, "tid": -914061504, "ts": 1716454224301287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301358, "dur": 1,
    "args": {
      "External id": 187105,
      "cbid": 251, "correlation": 187105
    }
  },
  {
    "ph": "f", "id": 187105, "pid": 76337, "tid": -914061504, "ts": 1716454224301358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224364003, "dur": 156,
    "args": {
      "External id": 187106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187106,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187106, "pid": 5, "tid": 7, "ts": 1716454224364003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301362, "dur": 11,
    "args": {
      "External id": 187106,
      "cbid": 211, "correlation": 187106
    }
  },
  {
    "ph": "s", "id": 187106, "pid": 76337, "tid": -914061504, "ts": 1716454224301362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301426, "dur": 1,
    "args": {
      "External id": 187117,
      "cbid": 251, "correlation": 187117
    }
  },
  {
    "ph": "f", "id": 187117, "pid": 76337, "tid": -914061504, "ts": 1716454224301426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224364159, "dur": 158,
    "args": {
      "External id": 187118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187118,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187118, "pid": 5, "tid": 7, "ts": 1716454224364159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301430, "dur": 11,
    "args": {
      "External id": 187118,
      "cbid": 211, "correlation": 187118
    }
  },
  {
    "ph": "s", "id": 187118, "pid": 76337, "tid": -914061504, "ts": 1716454224301430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224364318, "dur": 331,
    "args": {
      "External id": 187143,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187143,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187143, "pid": 5, "tid": 7, "ts": 1716454224364318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301518, "dur": 13,
    "args": {
      "External id": 187143,
      "cbid": 211, "correlation": 187143
    }
  },
  {
    "ph": "s", "id": 187143, "pid": 76337, "tid": -914061504, "ts": 1716454224301518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301623, "dur": 1,
    "args": {
      "External id": 187161,
      "cbid": 251, "correlation": 187161
    }
  },
  {
    "ph": "f", "id": 187161, "pid": 76337, "tid": -914061504, "ts": 1716454224301623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224364650, "dur": 165,
    "args": {
      "External id": 187163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187163,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187163, "pid": 5, "tid": 7, "ts": 1716454224364650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301629, "dur": 13,
    "args": {
      "External id": 187163,
      "cbid": 211, "correlation": 187163
    }
  },
  {
    "ph": "s", "id": 187163, "pid": 76337, "tid": -914061504, "ts": 1716454224301629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224364817, "dur": 19,
    "args": {
      "External id": 187171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187171,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187171, "pid": 5, "tid": 7, "ts": 1716454224364817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301699, "dur": 13,
    "args": {
      "External id": 187171,
      "cbid": 211, "correlation": 187171
    }
  },
  {
    "ph": "s", "id": 187171, "pid": 76337, "tid": -914061504, "ts": 1716454224301699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224364837, "dur": 27,
    "args": {
      "External id": 187179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187179,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187179, "pid": 5, "tid": 7, "ts": 1716454224364837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301739, "dur": 8,
    "args": {
      "External id": 187179,
      "cbid": 211, "correlation": 187179
    }
  },
  {
    "ph": "s", "id": 187179, "pid": 76337, "tid": -914061504, "ts": 1716454224301739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224364866, "dur": 18,
    "args": {
      "External id": 187190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187190,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187190, "pid": 5, "tid": 7, "ts": 1716454224364866,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301814, "dur": 13,
    "args": {
      "External id": 187190,
      "cbid": 211, "correlation": 187190
    }
  },
  {
    "ph": "s", "id": 187190, "pid": 76337, "tid": -914061504, "ts": 1716454224301814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224364885, "dur": 16,
    "args": {
      "External id": 187212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187212,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187212, "pid": 5, "tid": 7, "ts": 1716454224364885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301845, "dur": 7,
    "args": {
      "External id": 187212,
      "cbid": 211, "correlation": 187212
    }
  },
  {
    "ph": "s", "id": 187212, "pid": 76337, "tid": -914061504, "ts": 1716454224301845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301930, "dur": 1,
    "args": {
      "External id": 187223,
      "cbid": 251, "correlation": 187223
    }
  },
  {
    "ph": "f", "id": 187223, "pid": 76337, "tid": -914061504, "ts": 1716454224301930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224364902, "dur": 88,
    "args": {
      "External id": 187224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187224,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 187224, "pid": 5, "tid": 7, "ts": 1716454224364902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224301935, "dur": 13,
    "args": {
      "External id": 187224,
      "cbid": 211, "correlation": 187224
    }
  },
  {
    "ph": "s", "id": 187224, "pid": 76337, "tid": -914061504, "ts": 1716454224301935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302017, "dur": 1,
    "args": {
      "External id": 187235,
      "cbid": 251, "correlation": 187235
    }
  },
  {
    "ph": "f", "id": 187235, "pid": 76337, "tid": -914061504, "ts": 1716454224302017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302021, "dur": 0,
    "args": {
      "External id": 187236,
      "cbid": 251, "correlation": 187236
    }
  },
  {
    "ph": "f", "id": 187236, "pid": 76337, "tid": -914061504, "ts": 1716454224302021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224364991, "dur": 12,
    "args": {
      "External id": 187237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187237,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187237, "pid": 5, "tid": 7, "ts": 1716454224364991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302023, "dur": 13,
    "args": {
      "External id": 187237,
      "cbid": 211, "correlation": 187237
    }
  },
  {
    "ph": "s", "id": 187237, "pid": 76337, "tid": -914061504, "ts": 1716454224302023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224365005, "dur": 6,
    "args": {
      "External id": 187239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187239,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187239, "pid": 5, "tid": 7, "ts": 1716454224365005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302038, "dur": 6,
    "args": {
      "External id": 187239,
      "cbid": 211, "correlation": 187239
    }
  },
  {
    "ph": "s", "id": 187239, "pid": 76337, "tid": -914061504, "ts": 1716454224302038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302097, "dur": 1,
    "args": {
      "External id": 187250,
      "cbid": 251, "correlation": 187250
    }
  },
  {
    "ph": "f", "id": 187250, "pid": 76337, "tid": -914061504, "ts": 1716454224302097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302100, "dur": 0,
    "args": {
      "External id": 187251,
      "cbid": 251, "correlation": 187251
    }
  },
  {
    "ph": "f", "id": 187251, "pid": 76337, "tid": -914061504, "ts": 1716454224302100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224365012, "dur": 8,
    "args": {
      "External id": 187252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187252,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187252, "pid": 5, "tid": 7, "ts": 1716454224365012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302102, "dur": 12,
    "args": {
      "External id": 187252,
      "cbid": 211, "correlation": 187252
    }
  },
  {
    "ph": "s", "id": 187252, "pid": 76337, "tid": -914061504, "ts": 1716454224302102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224365022, "dur": 3,
    "args": {
      "External id": 187254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187254,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187254, "pid": 5, "tid": 7, "ts": 1716454224365022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302116, "dur": 5,
    "args": {
      "External id": 187254,
      "cbid": 211, "correlation": 187254
    }
  },
  {
    "ph": "s", "id": 187254, "pid": 76337, "tid": -914061504, "ts": 1716454224302116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224365026, "dur": 54,
    "args": {
      "External id": 187279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187279,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187279, "pid": 5, "tid": 7, "ts": 1716454224365026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302191, "dur": 12,
    "args": {
      "External id": 187279,
      "cbid": 211, "correlation": 187279
    }
  },
  {
    "ph": "s", "id": 187279, "pid": 76337, "tid": -914061504, "ts": 1716454224302191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302290, "dur": 1,
    "args": {
      "External id": 187297,
      "cbid": 251, "correlation": 187297
    }
  },
  {
    "ph": "f", "id": 187297, "pid": 76337, "tid": -914061504, "ts": 1716454224302290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224365081, "dur": 90,
    "args": {
      "External id": 187299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187299,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 187299, "pid": 5, "tid": 7, "ts": 1716454224365081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302296, "dur": 15,
    "args": {
      "External id": 187299,
      "cbid": 211, "correlation": 187299
    }
  },
  {
    "ph": "s", "id": 187299, "pid": 76337, "tid": -914061504, "ts": 1716454224302296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224365173, "dur": 9,
    "args": {
      "External id": 187307,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187307,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187307, "pid": 5, "tid": 7, "ts": 1716454224365173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302368, "dur": 12,
    "args": {
      "External id": 187307,
      "cbid": 211, "correlation": 187307
    }
  },
  {
    "ph": "s", "id": 187307, "pid": 76337, "tid": -914061504, "ts": 1716454224302368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224365183, "dur": 21,
    "args": {
      "External id": 187315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187315, "pid": 5, "tid": 7, "ts": 1716454224365183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302411, "dur": 9,
    "args": {
      "External id": 187315,
      "cbid": 211, "correlation": 187315
    }
  },
  {
    "ph": "s", "id": 187315, "pid": 76337, "tid": -914061504, "ts": 1716454224302411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224365206, "dur": 18,
    "args": {
      "External id": 187337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187337,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187337, "pid": 5, "tid": 7, "ts": 1716454224365206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302463, "dur": 10,
    "args": {
      "External id": 187337,
      "cbid": 211, "correlation": 187337
    }
  },
  {
    "ph": "s", "id": 187337, "pid": 76337, "tid": -914061504, "ts": 1716454224302463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302550, "dur": 1,
    "args": {
      "External id": 187353,
      "cbid": 251, "correlation": 187353
    }
  },
  {
    "ph": "f", "id": 187353, "pid": 76337, "tid": -914061504, "ts": 1716454224302550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302555, "dur": 0,
    "args": {
      "External id": 187355,
      "cbid": 251, "correlation": 187355
    }
  },
  {
    "ph": "f", "id": 187355, "pid": 76337, "tid": -914061504, "ts": 1716454224302555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224365225, "dur": 493,
    "args": {
      "External id": 187356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187356,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187356, "pid": 5, "tid": 7, "ts": 1716454224365225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302557, "dur": 14,
    "args": {
      "External id": 187356,
      "cbid": 211, "correlation": 187356
    }
  },
  {
    "ph": "s", "id": 187356, "pid": 76337, "tid": -914061504, "ts": 1716454224302557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224365720, "dur": 65,
    "args": {
      "External id": 187364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187364, "pid": 5, "tid": 7, "ts": 1716454224365720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302628, "dur": 12,
    "args": {
      "External id": 187364,
      "cbid": 211, "correlation": 187364
    }
  },
  {
    "ph": "s", "id": 187364, "pid": 76337, "tid": -914061504, "ts": 1716454224302628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224365787, "dur": 67,
    "args": {
      "External id": 187372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187372,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187372, "pid": 5, "tid": 7, "ts": 1716454224365787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302662, "dur": 9,
    "args": {
      "External id": 187372,
      "cbid": 211, "correlation": 187372
    }
  },
  {
    "ph": "s", "id": 187372, "pid": 76337, "tid": -914061504, "ts": 1716454224302662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302742, "dur": 1,
    "args": {
      "External id": 187388,
      "cbid": 251, "correlation": 187388
    }
  },
  {
    "ph": "f", "id": 187388, "pid": 76337, "tid": -914061504, "ts": 1716454224302742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224365855, "dur": 1,
    "args": {
      "External id": 187390,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 187390,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 187390, "pid": 5, "tid": 7, "ts": 1716454224365855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302746, "dur": 11,
    "args": {
      "External id": 187390,
      "cbid": 51, "correlation": 187390
    }
  },
  {
    "ph": "s", "id": 187390, "pid": 76337, "tid": -914061504, "ts": 1716454224302746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224365859, "dur": 267,
    "args": {
      "External id": 187391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187391,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 187391, "pid": 5, "tid": 7, "ts": 1716454224365859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302759, "dur": 11,
    "args": {
      "External id": 187391,
      "cbid": 211, "correlation": 187391
    }
  },
  {
    "ph": "s", "id": 187391, "pid": 76337, "tid": -914061504, "ts": 1716454224302759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224366127, "dur": 14,
    "args": {
      "External id": 187399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187399,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187399, "pid": 5, "tid": 7, "ts": 1716454224366127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302800, "dur": 11,
    "args": {
      "External id": 187399,
      "cbid": 211, "correlation": 187399
    }
  },
  {
    "ph": "s", "id": 187399, "pid": 76337, "tid": -914061504, "ts": 1716454224302800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224366143, "dur": 37,
    "args": {
      "External id": 187410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187410,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187410, "pid": 5, "tid": 7, "ts": 1716454224366143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302870, "dur": 12,
    "args": {
      "External id": 187410,
      "cbid": 211, "correlation": 187410
    }
  },
  {
    "ph": "s", "id": 187410, "pid": 76337, "tid": -914061504, "ts": 1716454224302870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302934, "dur": 0,
    "args": {
      "External id": 187422,
      "cbid": 317, "correlation": 187422
    }
  },
  {
    "ph": "f", "id": 187422, "pid": 76337, "tid": -914061504, "ts": 1716454224302934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302935, "dur": 0,
    "args": {
      "External id": 187423,
      "cbid": 203, "correlation": 187423
    }
  },
  {
    "ph": "f", "id": 187423, "pid": 76337, "tid": -914061504, "ts": 1716454224302935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302935, "dur": 0,
    "args": {
      "External id": 187424,
      "cbid": 205, "correlation": 187424
    }
  },
  {
    "ph": "f", "id": 187424, "pid": 76337, "tid": -914061504, "ts": 1716454224302935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224366182, "dur": 13,
    "args": {
      "External id": 187428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187428,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187428, "pid": 5, "tid": 7, "ts": 1716454224366182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302951, "dur": 13,
    "args": {
      "External id": 187428,
      "cbid": 211, "correlation": 187428
    }
  },
  {
    "ph": "s", "id": 187428, "pid": 76337, "tid": -914061504, "ts": 1716454224302951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224366196, "dur": 4,
    "args": {
      "External id": 187430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187430,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 187430, "pid": 5, "tid": 7, "ts": 1716454224366196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302968, "dur": 13,
    "args": {
      "External id": 187430,
      "cbid": 211, "correlation": 187430
    }
  },
  {
    "ph": "s", "id": 187430, "pid": 76337, "tid": -914061504, "ts": 1716454224302968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302984, "dur": 0,
    "args": {
      "External id": 187431,
      "cbid": 51, "correlation": 187431
    }
  },
  {
    "ph": "s", "id": 187431, "pid": 76337, "tid": -914061504, "ts": 1716454224302984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224366201, "dur": 96,
    "args": {
      "External id": 187432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187432,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 187432, "pid": 5, "tid": 7, "ts": 1716454224366201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224302985, "dur": 6,
    "args": {
      "External id": 187432,
      "cbid": 211, "correlation": 187432
    }
  },
  {
    "ph": "s", "id": 187432, "pid": 76337, "tid": -914061504, "ts": 1716454224302985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224366299, "dur": 16,
    "args": {
      "External id": 187437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187437,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187437, "pid": 5, "tid": 7, "ts": 1716454224366299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303014, "dur": 9,
    "args": {
      "External id": 187437,
      "cbid": 211, "correlation": 187437
    }
  },
  {
    "ph": "s", "id": 187437, "pid": 76337, "tid": -914061504, "ts": 1716454224303014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224366316, "dur": 12,
    "args": {
      "External id": 187445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187445,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187445, "pid": 5, "tid": 7, "ts": 1716454224366316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303045, "dur": 9,
    "args": {
      "External id": 187445,
      "cbid": 211, "correlation": 187445
    }
  },
  {
    "ph": "s", "id": 187445, "pid": 76337, "tid": -914061504, "ts": 1716454224303045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224366329, "dur": 55,
    "args": {
      "External id": 187456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187456,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187456, "pid": 5, "tid": 7, "ts": 1716454224366329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303113, "dur": 12,
    "args": {
      "External id": 187456,
      "cbid": 211, "correlation": 187456
    }
  },
  {
    "ph": "s", "id": 187456, "pid": 76337, "tid": -914061504, "ts": 1716454224303113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303169, "dur": 0,
    "args": {
      "External id": 187466,
      "cbid": 317, "correlation": 187466
    }
  },
  {
    "ph": "f", "id": 187466, "pid": 76337, "tid": -914061504, "ts": 1716454224303169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303170, "dur": 0,
    "args": {
      "External id": 187467,
      "cbid": 203, "correlation": 187467
    }
  },
  {
    "ph": "f", "id": 187467, "pid": 76337, "tid": -914061504, "ts": 1716454224303170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303170, "dur": 0,
    "args": {
      "External id": 187468,
      "cbid": 205, "correlation": 187468
    }
  },
  {
    "ph": "f", "id": 187468, "pid": 76337, "tid": -914061504, "ts": 1716454224303170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224366385, "dur": 38,
    "args": {
      "External id": 187472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187472,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187472, "pid": 5, "tid": 7, "ts": 1716454224366385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303187, "dur": 12,
    "args": {
      "External id": 187472,
      "cbid": 211, "correlation": 187472
    }
  },
  {
    "ph": "s", "id": 187472, "pid": 76337, "tid": -914061504, "ts": 1716454224303187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224366424, "dur": 159,
    "args": {
      "External id": 187474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187474,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187474, "pid": 5, "tid": 7, "ts": 1716454224366424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303201, "dur": 6,
    "args": {
      "External id": 187474,
      "cbid": 211, "correlation": 187474
    }
  },
  {
    "ph": "s", "id": 187474, "pid": 76337, "tid": -914061504, "ts": 1716454224303201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224366585, "dur": 1975,
    "args": {
      "External id": 187476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187476,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187476, "pid": 5, "tid": 7, "ts": 1716454224366585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303214, "dur": 9,
    "args": {
      "External id": 187476,
      "cbid": 211, "correlation": 187476
    }
  },
  {
    "ph": "s", "id": 187476, "pid": 76337, "tid": -914061504, "ts": 1716454224303214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224368561, "dur": 38,
    "args": {
      "External id": 187478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187478, "pid": 5, "tid": 7, "ts": 1716454224368561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303227, "dur": 5,
    "args": {
      "External id": 187478,
      "cbid": 211, "correlation": 187478
    }
  },
  {
    "ph": "s", "id": 187478, "pid": 76337, "tid": -914061504, "ts": 1716454224303227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224368601, "dur": 59,
    "args": {
      "External id": 187484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187484, "pid": 5, "tid": 7, "ts": 1716454224368601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303256, "dur": 8,
    "args": {
      "External id": 187484,
      "cbid": 211, "correlation": 187484
    }
  },
  {
    "ph": "s", "id": 187484, "pid": 76337, "tid": -914061504, "ts": 1716454224303256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224368661, "dur": 82,
    "args": {
      "External id": 187493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187493,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187493, "pid": 5, "tid": 7, "ts": 1716454224368661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303349, "dur": 14,
    "args": {
      "External id": 187493,
      "cbid": 211, "correlation": 187493
    }
  },
  {
    "ph": "s", "id": 187493, "pid": 76337, "tid": -914061504, "ts": 1716454224303349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224368745, "dur": 73,
    "args": {
      "External id": 187513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187513,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 187513, "pid": 5, "tid": 7, "ts": 1716454224368745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303422, "dur": 11,
    "args": {
      "External id": 187513,
      "cbid": 211, "correlation": 187513
    }
  },
  {
    "ph": "s", "id": 187513, "pid": 76337, "tid": -914061504, "ts": 1716454224303422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224368819, "dur": 5,
    "args": {
      "External id": 187525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187525,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 187525, "pid": 5, "tid": 7, "ts": 1716454224368819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303442, "dur": 6,
    "args": {
      "External id": 187525,
      "cbid": 211, "correlation": 187525
    }
  },
  {
    "ph": "s", "id": 187525, "pid": 76337, "tid": -914061504, "ts": 1716454224303442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224368825, "dur": 81,
    "args": {
      "External id": 187528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187528,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187528, "pid": 5, "tid": 7, "ts": 1716454224368825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303462, "dur": 7,
    "args": {
      "External id": 187528,
      "cbid": 211, "correlation": 187528
    }
  },
  {
    "ph": "s", "id": 187528, "pid": 76337, "tid": -914061504, "ts": 1716454224303462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224368907, "dur": 53,
    "args": {
      "External id": 187537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187537,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187537, "pid": 5, "tid": 7, "ts": 1716454224368907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303503, "dur": 10,
    "args": {
      "External id": 187537,
      "cbid": 211, "correlation": 187537
    }
  },
  {
    "ph": "s", "id": 187537, "pid": 76337, "tid": -914061504, "ts": 1716454224303503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303555, "dur": 0,
    "args": {
      "External id": 187547,
      "cbid": 317, "correlation": 187547
    }
  },
  {
    "ph": "f", "id": 187547, "pid": 76337, "tid": -914061504, "ts": 1716454224303555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303555, "dur": 0,
    "args": {
      "External id": 187548,
      "cbid": 203, "correlation": 187548
    }
  },
  {
    "ph": "f", "id": 187548, "pid": 76337, "tid": -914061504, "ts": 1716454224303555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303556, "dur": 0,
    "args": {
      "External id": 187549,
      "cbid": 205, "correlation": 187549
    }
  },
  {
    "ph": "f", "id": 187549, "pid": 76337, "tid": -914061504, "ts": 1716454224303556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224368962, "dur": 57,
    "args": {
      "External id": 187553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187553,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187553, "pid": 5, "tid": 7, "ts": 1716454224368962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303573, "dur": 11,
    "args": {
      "External id": 187553,
      "cbid": 211, "correlation": 187553
    }
  },
  {
    "ph": "s", "id": 187553, "pid": 76337, "tid": -914061504, "ts": 1716454224303573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224369021, "dur": 121,
    "args": {
      "External id": 187555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187555,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187555, "pid": 5, "tid": 7, "ts": 1716454224369021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303587, "dur": 5,
    "args": {
      "External id": 187555,
      "cbid": 211, "correlation": 187555
    }
  },
  {
    "ph": "s", "id": 187555, "pid": 76337, "tid": -914061504, "ts": 1716454224303587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224369142, "dur": 1880,
    "args": {
      "External id": 187557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187557,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187557, "pid": 5, "tid": 7, "ts": 1716454224369142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303598, "dur": 7,
    "args": {
      "External id": 187557,
      "cbid": 211, "correlation": 187557
    }
  },
  {
    "ph": "s", "id": 187557, "pid": 76337, "tid": -914061504, "ts": 1716454224303598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224371024, "dur": 19,
    "args": {
      "External id": 187559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187559, "pid": 5, "tid": 7, "ts": 1716454224371024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303609, "dur": 5,
    "args": {
      "External id": 187559,
      "cbid": 211, "correlation": 187559
    }
  },
  {
    "ph": "s", "id": 187559, "pid": 76337, "tid": -914061504, "ts": 1716454224303609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224371044, "dur": 32,
    "args": {
      "External id": 187565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187565,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187565, "pid": 5, "tid": 7, "ts": 1716454224371044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303636, "dur": 8,
    "args": {
      "External id": 187565,
      "cbid": 211, "correlation": 187565
    }
  },
  {
    "ph": "s", "id": 187565, "pid": 76337, "tid": -914061504, "ts": 1716454224303636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224371078, "dur": 4,
    "args": {
      "External id": 187573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187573,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 187573, "pid": 5, "tid": 7, "ts": 1716454224371078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303680, "dur": 9,
    "args": {
      "External id": 187573,
      "cbid": 211, "correlation": 187573
    }
  },
  {
    "ph": "s", "id": 187573, "pid": 76337, "tid": -914061504, "ts": 1716454224303680,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303749, "dur": 1,
    "args": {
      "External id": 187589,
      "cbid": 251, "correlation": 187589
    }
  },
  {
    "ph": "f", "id": 187589, "pid": 76337, "tid": -914061504, "ts": 1716454224303749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303755, "dur": 0,
    "args": {
      "External id": 187591,
      "cbid": 251, "correlation": 187591
    }
  },
  {
    "ph": "f", "id": 187591, "pid": 76337, "tid": -914061504, "ts": 1716454224303755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224371083, "dur": 12,
    "args": {
      "External id": 187592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187592,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 187592, "pid": 5, "tid": 7, "ts": 1716454224371083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303757, "dur": 11,
    "args": {
      "External id": 187592,
      "cbid": 211, "correlation": 187592
    }
  },
  {
    "ph": "s", "id": 187592, "pid": 76337, "tid": -914061504, "ts": 1716454224303757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224371096, "dur": 5,
    "args": {
      "External id": 187594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187594,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 187594, "pid": 5, "tid": 7, "ts": 1716454224371096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303770, "dur": 5,
    "args": {
      "External id": 187594,
      "cbid": 211, "correlation": 187594
    }
  },
  {
    "ph": "s", "id": 187594, "pid": 76337, "tid": -914061504, "ts": 1716454224303770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224371102, "dur": 29,
    "args": {
      "External id": 187604,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187604,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187604, "pid": 5, "tid": 7, "ts": 1716454224371102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303831, "dur": 12,
    "args": {
      "External id": 187604,
      "cbid": 211, "correlation": 187604
    }
  },
  {
    "ph": "s", "id": 187604, "pid": 76337, "tid": -914061504, "ts": 1716454224303831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224371132, "dur": 30,
    "args": {
      "External id": 187624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187624,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 187624, "pid": 5, "tid": 7, "ts": 1716454224371132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303897, "dur": 11,
    "args": {
      "External id": 187624,
      "cbid": 211, "correlation": 187624
    }
  },
  {
    "ph": "s", "id": 187624, "pid": 76337, "tid": -914061504, "ts": 1716454224303897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224371164, "dur": 4,
    "args": {
      "External id": 187636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187636,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 187636, "pid": 5, "tid": 7, "ts": 1716454224371164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303918, "dur": 6,
    "args": {
      "External id": 187636,
      "cbid": 211, "correlation": 187636
    }
  },
  {
    "ph": "s", "id": 187636, "pid": 76337, "tid": -914061504, "ts": 1716454224303918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224371169, "dur": 29,
    "args": {
      "External id": 187639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187639,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187639, "pid": 5, "tid": 7, "ts": 1716454224371169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303936, "dur": 6,
    "args": {
      "External id": 187639,
      "cbid": 211, "correlation": 187639
    }
  },
  {
    "ph": "s", "id": 187639, "pid": 76337, "tid": -914061504, "ts": 1716454224303936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224371199, "dur": 20,
    "args": {
      "External id": 187648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187648,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187648, "pid": 5, "tid": 7, "ts": 1716454224371199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224303983, "dur": 11,
    "args": {
      "External id": 187648,
      "cbid": 211, "correlation": 187648
    }
  },
  {
    "ph": "s", "id": 187648, "pid": 76337, "tid": -914061504, "ts": 1716454224303983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304049, "dur": 0,
    "args": {
      "External id": 187658,
      "cbid": 317, "correlation": 187658
    }
  },
  {
    "ph": "f", "id": 187658, "pid": 76337, "tid": -914061504, "ts": 1716454224304049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304050, "dur": 0,
    "args": {
      "External id": 187659,
      "cbid": 203, "correlation": 187659
    }
  },
  {
    "ph": "f", "id": 187659, "pid": 76337, "tid": -914061504, "ts": 1716454224304050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304050, "dur": 0,
    "args": {
      "External id": 187660,
      "cbid": 205, "correlation": 187660
    }
  },
  {
    "ph": "f", "id": 187660, "pid": 76337, "tid": -914061504, "ts": 1716454224304050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224371221, "dur": 21,
    "args": {
      "External id": 187664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187664,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187664, "pid": 5, "tid": 7, "ts": 1716454224371221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304067, "dur": 12,
    "args": {
      "External id": 187664,
      "cbid": 211, "correlation": 187664
    }
  },
  {
    "ph": "s", "id": 187664, "pid": 76337, "tid": -914061504, "ts": 1716454224304067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224371243, "dur": 43,
    "args": {
      "External id": 187666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187666,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187666, "pid": 5, "tid": 7, "ts": 1716454224371243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304082, "dur": 5,
    "args": {
      "External id": 187666,
      "cbid": 211, "correlation": 187666
    }
  },
  {
    "ph": "s", "id": 187666, "pid": 76337, "tid": -914061504, "ts": 1716454224304082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224371288, "dur": 640,
    "args": {
      "External id": 187668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187668,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187668, "pid": 5, "tid": 7, "ts": 1716454224371288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304093, "dur": 6,
    "args": {
      "External id": 187668,
      "cbid": 211, "correlation": 187668
    }
  },
  {
    "ph": "s", "id": 187668, "pid": 76337, "tid": -914061504, "ts": 1716454224304093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224371929, "dur": 23,
    "args": {
      "External id": 187670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187670, "pid": 5, "tid": 7, "ts": 1716454224371929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304104, "dur": 5,
    "args": {
      "External id": 187670,
      "cbid": 211, "correlation": 187670
    }
  },
  {
    "ph": "s", "id": 187670, "pid": 76337, "tid": -914061504, "ts": 1716454224304104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224371953, "dur": 32,
    "args": {
      "External id": 187676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187676, "pid": 5, "tid": 7, "ts": 1716454224371953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304132, "dur": 8,
    "args": {
      "External id": 187676,
      "cbid": 211, "correlation": 187676
    }
  },
  {
    "ph": "s", "id": 187676, "pid": 76337, "tid": -914061504, "ts": 1716454224304132,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304190, "dur": 0,
    "args": {
      "External id": 187686,
      "cbid": 317, "correlation": 187686
    }
  },
  {
    "ph": "f", "id": 187686, "pid": 76337, "tid": -914061504, "ts": 1716454224304190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304191, "dur": 0,
    "args": {
      "External id": 187687,
      "cbid": 203, "correlation": 187687
    }
  },
  {
    "ph": "f", "id": 187687, "pid": 76337, "tid": -914061504, "ts": 1716454224304191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304192, "dur": 0,
    "args": {
      "External id": 187688,
      "cbid": 205, "correlation": 187688
    }
  },
  {
    "ph": "f", "id": 187688, "pid": 76337, "tid": -914061504, "ts": 1716454224304192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224371986, "dur": 56,
    "args": {
      "External id": 187692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187692, "pid": 5, "tid": 7, "ts": 1716454224371986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304206, "dur": 12,
    "args": {
      "External id": 187692,
      "cbid": 211, "correlation": 187692
    }
  },
  {
    "ph": "s", "id": 187692, "pid": 76337, "tid": -914061504, "ts": 1716454224304206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224372044, "dur": 263,
    "args": {
      "External id": 187694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187694,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187694, "pid": 5, "tid": 7, "ts": 1716454224372044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304224, "dur": 8,
    "args": {
      "External id": 187694,
      "cbid": 211, "correlation": 187694
    }
  },
  {
    "ph": "s", "id": 187694, "pid": 76337, "tid": -914061504, "ts": 1716454224304224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224372308, "dur": 23,
    "args": {
      "External id": 187696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187696,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187696, "pid": 5, "tid": 7, "ts": 1716454224372308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304235, "dur": 5,
    "args": {
      "External id": 187696,
      "cbid": 211, "correlation": 187696
    }
  },
  {
    "ph": "s", "id": 187696, "pid": 76337, "tid": -914061504, "ts": 1716454224304235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224372332, "dur": 32,
    "args": {
      "External id": 187702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187702,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187702, "pid": 5, "tid": 7, "ts": 1716454224372332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304261, "dur": 8,
    "args": {
      "External id": 187702,
      "cbid": 211, "correlation": 187702
    }
  },
  {
    "ph": "s", "id": 187702, "pid": 76337, "tid": -914061504, "ts": 1716454224304261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224372365, "dur": 27,
    "args": {
      "External id": 187710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187710,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187710, "pid": 5, "tid": 7, "ts": 1716454224372365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304289, "dur": 8,
    "args": {
      "External id": 187710,
      "cbid": 211, "correlation": 187710
    }
  },
  {
    "ph": "s", "id": 187710, "pid": 76337, "tid": -914061504, "ts": 1716454224304289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224372394, "dur": 19,
    "args": {
      "External id": 187718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187718,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187718, "pid": 5, "tid": 7, "ts": 1716454224372394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304318, "dur": 9,
    "args": {
      "External id": 187718,
      "cbid": 211, "correlation": 187718
    }
  },
  {
    "ph": "s", "id": 187718, "pid": 76337, "tid": -914061504, "ts": 1716454224304318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224372414, "dur": 31,
    "args": {
      "External id": 187738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187738,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 187738, "pid": 5, "tid": 7, "ts": 1716454224372414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304401, "dur": 13,
    "args": {
      "External id": 187738,
      "cbid": 211, "correlation": 187738
    }
  },
  {
    "ph": "s", "id": 187738, "pid": 76337, "tid": -914061504, "ts": 1716454224304401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224372447, "dur": 4,
    "args": {
      "External id": 187750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187750,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 187750, "pid": 5, "tid": 7, "ts": 1716454224372447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304423, "dur": 6,
    "args": {
      "External id": 187750,
      "cbid": 211, "correlation": 187750
    }
  },
  {
    "ph": "s", "id": 187750, "pid": 76337, "tid": -914061504, "ts": 1716454224304423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224372453, "dur": 30,
    "args": {
      "External id": 187753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187753,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187753, "pid": 5, "tid": 7, "ts": 1716454224372453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304440, "dur": 7,
    "args": {
      "External id": 187753,
      "cbid": 211, "correlation": 187753
    }
  },
  {
    "ph": "s", "id": 187753, "pid": 76337, "tid": -914061504, "ts": 1716454224304440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304498, "dur": 0,
    "args": {
      "External id": 187764,
      "cbid": 317, "correlation": 187764
    }
  },
  {
    "ph": "f", "id": 187764, "pid": 76337, "tid": -914061504, "ts": 1716454224304498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304499, "dur": 0,
    "args": {
      "External id": 187765,
      "cbid": 203, "correlation": 187765
    }
  },
  {
    "ph": "f", "id": 187765, "pid": 76337, "tid": -914061504, "ts": 1716454224304499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304500, "dur": 0,
    "args": {
      "External id": 187766,
      "cbid": 205, "correlation": 187766
    }
  },
  {
    "ph": "f", "id": 187766, "pid": 76337, "tid": -914061504, "ts": 1716454224304500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224372484, "dur": 21,
    "args": {
      "External id": 187770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187770, "pid": 5, "tid": 7, "ts": 1716454224372484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304514, "dur": 12,
    "args": {
      "External id": 187770,
      "cbid": 211, "correlation": 187770
    }
  },
  {
    "ph": "s", "id": 187770, "pid": 76337, "tid": -914061504, "ts": 1716454224304514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224372507, "dur": 104,
    "args": {
      "External id": 187772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187772,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187772, "pid": 5, "tid": 7, "ts": 1716454224372507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304532, "dur": 7,
    "args": {
      "External id": 187772,
      "cbid": 211, "correlation": 187772
    }
  },
  {
    "ph": "s", "id": 187772, "pid": 76337, "tid": -914061504, "ts": 1716454224304532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224372612, "dur": 23,
    "args": {
      "External id": 187774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187774,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187774, "pid": 5, "tid": 7, "ts": 1716454224372612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304543, "dur": 5,
    "args": {
      "External id": 187774,
      "cbid": 211, "correlation": 187774
    }
  },
  {
    "ph": "s", "id": 187774, "pid": 76337, "tid": -914061504, "ts": 1716454224304543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224372636, "dur": 32,
    "args": {
      "External id": 187780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187780,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187780, "pid": 5, "tid": 7, "ts": 1716454224372636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304571, "dur": 8,
    "args": {
      "External id": 187780,
      "cbid": 211, "correlation": 187780
    }
  },
  {
    "ph": "s", "id": 187780, "pid": 76337, "tid": -914061504, "ts": 1716454224304571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224372669, "dur": 192,
    "args": {
      "External id": 187789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187789,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187789, "pid": 5, "tid": 7, "ts": 1716454224372669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304652, "dur": 15,
    "args": {
      "External id": 187789,
      "cbid": 211, "correlation": 187789
    }
  },
  {
    "ph": "s", "id": 187789, "pid": 76337, "tid": -914061504, "ts": 1716454224304652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224372863, "dur": 64,
    "args": {
      "External id": 187811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187811,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187811, "pid": 5, "tid": 7, "ts": 1716454224372863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304711, "dur": 10,
    "args": {
      "External id": 187811,
      "cbid": 211, "correlation": 187811
    }
  },
  {
    "ph": "s", "id": 187811, "pid": 76337, "tid": -914061504, "ts": 1716454224304711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304801, "dur": 1,
    "args": {
      "External id": 187822,
      "cbid": 251, "correlation": 187822
    }
  },
  {
    "ph": "f", "id": 187822, "pid": 76337, "tid": -914061504, "ts": 1716454224304801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224372928, "dur": 152,
    "args": {
      "External id": 187823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187823,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187823, "pid": 5, "tid": 7, "ts": 1716454224372928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304807, "dur": 13,
    "args": {
      "External id": 187823,
      "cbid": 211, "correlation": 187823
    }
  },
  {
    "ph": "s", "id": 187823, "pid": 76337, "tid": -914061504, "ts": 1716454224304807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304876, "dur": 1,
    "args": {
      "External id": 187834,
      "cbid": 251, "correlation": 187834
    }
  },
  {
    "ph": "f", "id": 187834, "pid": 76337, "tid": -914061504, "ts": 1716454224304876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224373081, "dur": 146,
    "args": {
      "External id": 187835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187835,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187835, "pid": 5, "tid": 7, "ts": 1716454224373081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304880, "dur": 11,
    "args": {
      "External id": 187835,
      "cbid": 211, "correlation": 187835
    }
  },
  {
    "ph": "s", "id": 187835, "pid": 76337, "tid": -914061504, "ts": 1716454224304880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304945, "dur": 1,
    "args": {
      "External id": 187846,
      "cbid": 251, "correlation": 187846
    }
  },
  {
    "ph": "f", "id": 187846, "pid": 76337, "tid": -914061504, "ts": 1716454224304945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224373229, "dur": 143,
    "args": {
      "External id": 187847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187847,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187847, "pid": 5, "tid": 7, "ts": 1716454224373229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224304949, "dur": 12,
    "args": {
      "External id": 187847,
      "cbid": 211, "correlation": 187847
    }
  },
  {
    "ph": "s", "id": 187847, "pid": 76337, "tid": -914061504, "ts": 1716454224304949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224373374, "dur": 1910,
    "args": {
      "External id": 187868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187868,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 187868, "pid": 5, "tid": 7, "ts": 1716454224373374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305040, "dur": 14,
    "args": {
      "External id": 187868,
      "cbid": 211, "correlation": 187868
    }
  },
  {
    "ph": "s", "id": 187868, "pid": 76337, "tid": -914061504, "ts": 1716454224305040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305140, "dur": 1,
    "args": {
      "External id": 187886,
      "cbid": 251, "correlation": 187886
    }
  },
  {
    "ph": "f", "id": 187886, "pid": 76337, "tid": -914061504, "ts": 1716454224305140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224375285, "dur": 149,
    "args": {
      "External id": 187888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187888,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 187888, "pid": 5, "tid": 7, "ts": 1716454224375285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305146, "dur": 14,
    "args": {
      "External id": 187888,
      "cbid": 211, "correlation": 187888
    }
  },
  {
    "ph": "s", "id": 187888, "pid": 76337, "tid": -914061504, "ts": 1716454224305146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224375435, "dur": 35,
    "args": {
      "External id": 187896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187896,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187896, "pid": 5, "tid": 7, "ts": 1716454224375435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305217, "dur": 12,
    "args": {
      "External id": 187896,
      "cbid": 211, "correlation": 187896
    }
  },
  {
    "ph": "s", "id": 187896, "pid": 76337, "tid": -914061504, "ts": 1716454224305217,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224375471, "dur": 51,
    "args": {
      "External id": 187904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187904,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187904, "pid": 5, "tid": 7, "ts": 1716454224375471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305256, "dur": 9,
    "args": {
      "External id": 187904,
      "cbid": 211, "correlation": 187904
    }
  },
  {
    "ph": "s", "id": 187904, "pid": 76337, "tid": -914061504, "ts": 1716454224305256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224375523, "dur": 29,
    "args": {
      "External id": 187915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187915,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187915, "pid": 5, "tid": 7, "ts": 1716454224375523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305327, "dur": 12,
    "args": {
      "External id": 187915,
      "cbid": 211, "correlation": 187915
    }
  },
  {
    "ph": "s", "id": 187915, "pid": 76337, "tid": -914061504, "ts": 1716454224305327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224375553, "dur": 33,
    "args": {
      "External id": 187937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187937,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 187937, "pid": 5, "tid": 7, "ts": 1716454224375553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305358, "dur": 8,
    "args": {
      "External id": 187937,
      "cbid": 211, "correlation": 187937
    }
  },
  {
    "ph": "s", "id": 187937, "pid": 76337, "tid": -914061504, "ts": 1716454224305358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305442, "dur": 1,
    "args": {
      "External id": 187948,
      "cbid": 251, "correlation": 187948
    }
  },
  {
    "ph": "f", "id": 187948, "pid": 76337, "tid": -914061504, "ts": 1716454224305442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224375588, "dur": 76,
    "args": {
      "External id": 187949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187949,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 187949, "pid": 5, "tid": 7, "ts": 1716454224375588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305448, "dur": 14,
    "args": {
      "External id": 187949,
      "cbid": 211, "correlation": 187949
    }
  },
  {
    "ph": "s", "id": 187949, "pid": 76337, "tid": -914061504, "ts": 1716454224305448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305517, "dur": 1,
    "args": {
      "External id": 187960,
      "cbid": 251, "correlation": 187960
    }
  },
  {
    "ph": "f", "id": 187960, "pid": 76337, "tid": -914061504, "ts": 1716454224305517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305520, "dur": 0,
    "args": {
      "External id": 187961,
      "cbid": 251, "correlation": 187961
    }
  },
  {
    "ph": "f", "id": 187961, "pid": 76337, "tid": -914061504, "ts": 1716454224305520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224375666, "dur": 11,
    "args": {
      "External id": 187962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187962,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 187962, "pid": 5, "tid": 7, "ts": 1716454224375666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305522, "dur": 12,
    "args": {
      "External id": 187962,
      "cbid": 211, "correlation": 187962
    }
  },
  {
    "ph": "s", "id": 187962, "pid": 76337, "tid": -914061504, "ts": 1716454224305522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224375678, "dur": 5,
    "args": {
      "External id": 187964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187964,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 187964, "pid": 5, "tid": 7, "ts": 1716454224375678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305536, "dur": 6,
    "args": {
      "External id": 187964,
      "cbid": 211, "correlation": 187964
    }
  },
  {
    "ph": "s", "id": 187964, "pid": 76337, "tid": -914061504, "ts": 1716454224305536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305593, "dur": 1,
    "args": {
      "External id": 187975,
      "cbid": 251, "correlation": 187975
    }
  },
  {
    "ph": "f", "id": 187975, "pid": 76337, "tid": -914061504, "ts": 1716454224305593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305596, "dur": 0,
    "args": {
      "External id": 187976,
      "cbid": 251, "correlation": 187976
    }
  },
  {
    "ph": "f", "id": 187976, "pid": 76337, "tid": -914061504, "ts": 1716454224305596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224375684, "dur": 7,
    "args": {
      "External id": 187977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187977,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 187977, "pid": 5, "tid": 7, "ts": 1716454224375684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305598, "dur": 12,
    "args": {
      "External id": 187977,
      "cbid": 211, "correlation": 187977
    }
  },
  {
    "ph": "s", "id": 187977, "pid": 76337, "tid": -914061504, "ts": 1716454224305598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224375692, "dur": 3,
    "args": {
      "External id": 187979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 187979,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 187979, "pid": 5, "tid": 7, "ts": 1716454224375692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305612, "dur": 5,
    "args": {
      "External id": 187979,
      "cbid": 211, "correlation": 187979
    }
  },
  {
    "ph": "s", "id": 187979, "pid": 76337, "tid": -914061504, "ts": 1716454224305612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224375697, "dur": 91,
    "args": {
      "External id": 188000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188000,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 188000, "pid": 5, "tid": 7, "ts": 1716454224375697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305684, "dur": 13,
    "args": {
      "External id": 188000,
      "cbid": 211, "correlation": 188000
    }
  },
  {
    "ph": "s", "id": 188000, "pid": 76337, "tid": -914061504, "ts": 1716454224305684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305782, "dur": 1,
    "args": {
      "External id": 188018,
      "cbid": 251, "correlation": 188018
    }
  },
  {
    "ph": "f", "id": 188018, "pid": 76337, "tid": -914061504, "ts": 1716454224305782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224375789, "dur": 96,
    "args": {
      "External id": 188020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188020,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188020, "pid": 5, "tid": 7, "ts": 1716454224375789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305788, "dur": 13,
    "args": {
      "External id": 188020,
      "cbid": 211, "correlation": 188020
    }
  },
  {
    "ph": "s", "id": 188020, "pid": 76337, "tid": -914061504, "ts": 1716454224305788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224375887, "dur": 19,
    "args": {
      "External id": 188028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188028,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188028, "pid": 5, "tid": 7, "ts": 1716454224375887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305857, "dur": 13,
    "args": {
      "External id": 188028,
      "cbid": 211, "correlation": 188028
    }
  },
  {
    "ph": "s", "id": 188028, "pid": 76337, "tid": -914061504, "ts": 1716454224305857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224375907, "dur": 37,
    "args": {
      "External id": 188036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188036,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188036, "pid": 5, "tid": 7, "ts": 1716454224375907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305898, "dur": 9,
    "args": {
      "External id": 188036,
      "cbid": 211, "correlation": 188036
    }
  },
  {
    "ph": "s", "id": 188036, "pid": 76337, "tid": -914061504, "ts": 1716454224305898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224375945, "dur": 34,
    "args": {
      "External id": 188058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188058,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188058, "pid": 5, "tid": 7, "ts": 1716454224375945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224305949, "dur": 10,
    "args": {
      "External id": 188058,
      "cbid": 211, "correlation": 188058
    }
  },
  {
    "ph": "s", "id": 188058, "pid": 76337, "tid": -914061504, "ts": 1716454224305949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306052, "dur": 1,
    "args": {
      "External id": 188074,
      "cbid": 251, "correlation": 188074
    }
  },
  {
    "ph": "f", "id": 188074, "pid": 76337, "tid": -914061504, "ts": 1716454224306052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306057, "dur": 0,
    "args": {
      "External id": 188076,
      "cbid": 251, "correlation": 188076
    }
  },
  {
    "ph": "f", "id": 188076, "pid": 76337, "tid": -914061504, "ts": 1716454224306057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224375980, "dur": 532,
    "args": {
      "External id": 188077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188077,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 188077, "pid": 5, "tid": 7, "ts": 1716454224375980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306062, "dur": 13,
    "args": {
      "External id": 188077,
      "cbid": 211, "correlation": 188077
    }
  },
  {
    "ph": "s", "id": 188077, "pid": 76337, "tid": -914061504, "ts": 1716454224306062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224376514, "dur": 124,
    "args": {
      "External id": 188085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188085,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188085, "pid": 5, "tid": 7, "ts": 1716454224376514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306129, "dur": 12,
    "args": {
      "External id": 188085,
      "cbid": 211, "correlation": 188085
    }
  },
  {
    "ph": "s", "id": 188085, "pid": 76337, "tid": -914061504, "ts": 1716454224306129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224376639, "dur": 129,
    "args": {
      "External id": 188093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188093, "pid": 5, "tid": 7, "ts": 1716454224376639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306160, "dur": 8,
    "args": {
      "External id": 188093,
      "cbid": 211, "correlation": 188093
    }
  },
  {
    "ph": "s", "id": 188093, "pid": 76337, "tid": -914061504, "ts": 1716454224306160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306236, "dur": 1,
    "args": {
      "External id": 188109,
      "cbid": 251, "correlation": 188109
    }
  },
  {
    "ph": "f", "id": 188109, "pid": 76337, "tid": -914061504, "ts": 1716454224306236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224376770, "dur": 299,
    "args": {
      "External id": 188111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188111,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188111, "pid": 5, "tid": 7, "ts": 1716454224376770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306242, "dur": 12,
    "args": {
      "External id": 188111,
      "cbid": 211, "correlation": 188111
    }
  },
  {
    "ph": "s", "id": 188111, "pid": 76337, "tid": -914061504, "ts": 1716454224306242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224377070, "dur": 27,
    "args": {
      "External id": 188119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188119,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188119, "pid": 5, "tid": 7, "ts": 1716454224377070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306284, "dur": 10,
    "args": {
      "External id": 188119,
      "cbid": 211, "correlation": 188119
    }
  },
  {
    "ph": "s", "id": 188119, "pid": 76337, "tid": -914061504, "ts": 1716454224306284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224377099, "dur": 80,
    "args": {
      "External id": 188130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188130,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188130, "pid": 5, "tid": 7, "ts": 1716454224377099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306350, "dur": 13,
    "args": {
      "External id": 188130,
      "cbid": 211, "correlation": 188130
    }
  },
  {
    "ph": "s", "id": 188130, "pid": 76337, "tid": -914061504, "ts": 1716454224306350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306414, "dur": 0,
    "args": {
      "External id": 188142,
      "cbid": 317, "correlation": 188142
    }
  },
  {
    "ph": "f", "id": 188142, "pid": 76337, "tid": -914061504, "ts": 1716454224306414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306414, "dur": 0,
    "args": {
      "External id": 188143,
      "cbid": 203, "correlation": 188143
    }
  },
  {
    "ph": "f", "id": 188143, "pid": 76337, "tid": -914061504, "ts": 1716454224306414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306415, "dur": 0,
    "args": {
      "External id": 188144,
      "cbid": 205, "correlation": 188144
    }
  },
  {
    "ph": "f", "id": 188144, "pid": 76337, "tid": -914061504, "ts": 1716454224306415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224377180, "dur": 24,
    "args": {
      "External id": 188148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188148,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188148, "pid": 5, "tid": 7, "ts": 1716454224377180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306430, "dur": 12,
    "args": {
      "External id": 188148,
      "cbid": 211, "correlation": 188148
    }
  },
  {
    "ph": "s", "id": 188148, "pid": 76337, "tid": -914061504, "ts": 1716454224306430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224377205, "dur": 118,
    "args": {
      "External id": 188150,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188150,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188150, "pid": 5, "tid": 7, "ts": 1716454224377205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306448, "dur": 6,
    "args": {
      "External id": 188150,
      "cbid": 211, "correlation": 188150
    }
  },
  {
    "ph": "s", "id": 188150, "pid": 76337, "tid": -914061504, "ts": 1716454224306448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224377325, "dur": 25,
    "args": {
      "External id": 188152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188152,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188152, "pid": 5, "tid": 7, "ts": 1716454224377325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306458, "dur": 6,
    "args": {
      "External id": 188152,
      "cbid": 211, "correlation": 188152
    }
  },
  {
    "ph": "s", "id": 188152, "pid": 76337, "tid": -914061504, "ts": 1716454224306458,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224377351, "dur": 32,
    "args": {
      "External id": 188158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188158,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188158, "pid": 5, "tid": 7, "ts": 1716454224377351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306486, "dur": 8,
    "args": {
      "External id": 188158,
      "cbid": 211, "correlation": 188158
    }
  },
  {
    "ph": "s", "id": 188158, "pid": 76337, "tid": -914061504, "ts": 1716454224306486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224377385, "dur": 27,
    "args": {
      "External id": 188166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188166,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188166, "pid": 5, "tid": 7, "ts": 1716454224377385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306518, "dur": 8,
    "args": {
      "External id": 188166,
      "cbid": 211, "correlation": 188166
    }
  },
  {
    "ph": "s", "id": 188166, "pid": 76337, "tid": -914061504, "ts": 1716454224306518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224377413, "dur": 54,
    "args": {
      "External id": 188175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188175,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188175, "pid": 5, "tid": 7, "ts": 1716454224377413,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306555, "dur": 11,
    "args": {
      "External id": 188175,
      "cbid": 211, "correlation": 188175
    }
  },
  {
    "ph": "s", "id": 188175, "pid": 76337, "tid": -914061504, "ts": 1716454224306555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224377468, "dur": 53,
    "args": {
      "External id": 188195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188195,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 188195, "pid": 5, "tid": 7, "ts": 1716454224377468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306626, "dur": 11,
    "args": {
      "External id": 188195,
      "cbid": 211, "correlation": 188195
    }
  },
  {
    "ph": "s", "id": 188195, "pid": 76337, "tid": -914061504, "ts": 1716454224306626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224377522, "dur": 5,
    "args": {
      "External id": 188207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188207,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 188207, "pid": 5, "tid": 7, "ts": 1716454224377522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306648, "dur": 6,
    "args": {
      "External id": 188207,
      "cbid": 211, "correlation": 188207
    }
  },
  {
    "ph": "s", "id": 188207, "pid": 76337, "tid": -914061504, "ts": 1716454224306648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224377528, "dur": 57,
    "args": {
      "External id": 188210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188210,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188210, "pid": 5, "tid": 7, "ts": 1716454224377528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306667, "dur": 6,
    "args": {
      "External id": 188210,
      "cbid": 211, "correlation": 188210
    }
  },
  {
    "ph": "s", "id": 188210, "pid": 76337, "tid": -914061504, "ts": 1716454224306667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224377586, "dur": 37,
    "args": {
      "External id": 188219,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188219,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188219, "pid": 5, "tid": 7, "ts": 1716454224377586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306705, "dur": 10,
    "args": {
      "External id": 188219,
      "cbid": 211, "correlation": 188219
    }
  },
  {
    "ph": "s", "id": 188219, "pid": 76337, "tid": -914061504, "ts": 1716454224306705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306756, "dur": 0,
    "args": {
      "External id": 188229,
      "cbid": 317, "correlation": 188229
    }
  },
  {
    "ph": "f", "id": 188229, "pid": 76337, "tid": -914061504, "ts": 1716454224306756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306756, "dur": 0,
    "args": {
      "External id": 188230,
      "cbid": 203, "correlation": 188230
    }
  },
  {
    "ph": "f", "id": 188230, "pid": 76337, "tid": -914061504, "ts": 1716454224306756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306757, "dur": 0,
    "args": {
      "External id": 188231,
      "cbid": 205, "correlation": 188231
    }
  },
  {
    "ph": "f", "id": 188231, "pid": 76337, "tid": -914061504, "ts": 1716454224306757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224377624, "dur": 39,
    "args": {
      "External id": 188235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188235,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188235, "pid": 5, "tid": 7, "ts": 1716454224377624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306773, "dur": 11,
    "args": {
      "External id": 188235,
      "cbid": 211, "correlation": 188235
    }
  },
  {
    "ph": "s", "id": 188235, "pid": 76337, "tid": -914061504, "ts": 1716454224306773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224377665, "dur": 82,
    "args": {
      "External id": 188237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188237,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188237, "pid": 5, "tid": 7, "ts": 1716454224377665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306787, "dur": 5,
    "args": {
      "External id": 188237,
      "cbid": 211, "correlation": 188237
    }
  },
  {
    "ph": "s", "id": 188237, "pid": 76337, "tid": -914061504, "ts": 1716454224306787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224377748, "dur": 1266,
    "args": {
      "External id": 188239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188239,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188239, "pid": 5, "tid": 7, "ts": 1716454224377748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306797, "dur": 6,
    "args": {
      "External id": 188239,
      "cbid": 211, "correlation": 188239
    }
  },
  {
    "ph": "s", "id": 188239, "pid": 76337, "tid": -914061504, "ts": 1716454224306797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224379016, "dur": 22,
    "args": {
      "External id": 188241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188241,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188241, "pid": 5, "tid": 7, "ts": 1716454224379016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306807, "dur": 5,
    "args": {
      "External id": 188241,
      "cbid": 211, "correlation": 188241
    }
  },
  {
    "ph": "s", "id": 188241, "pid": 76337, "tid": -914061504, "ts": 1716454224306807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224379040, "dur": 33,
    "args": {
      "External id": 188247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188247, "pid": 5, "tid": 7, "ts": 1716454224379040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306835, "dur": 8,
    "args": {
      "External id": 188247,
      "cbid": 211, "correlation": 188247
    }
  },
  {
    "ph": "s", "id": 188247, "pid": 76337, "tid": -914061504, "ts": 1716454224306835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224379074, "dur": 3,
    "args": {
      "External id": 188255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188255,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 188255, "pid": 5, "tid": 7, "ts": 1716454224379074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306878, "dur": 9,
    "args": {
      "External id": 188255,
      "cbid": 211, "correlation": 188255
    }
  },
  {
    "ph": "s", "id": 188255, "pid": 76337, "tid": -914061504, "ts": 1716454224306878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306943, "dur": 1,
    "args": {
      "External id": 188271,
      "cbid": 251, "correlation": 188271
    }
  },
  {
    "ph": "f", "id": 188271, "pid": 76337, "tid": -914061504, "ts": 1716454224306943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306948, "dur": 0,
    "args": {
      "External id": 188273,
      "cbid": 251, "correlation": 188273
    }
  },
  {
    "ph": "f", "id": 188273, "pid": 76337, "tid": -914061504, "ts": 1716454224306948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224379079, "dur": 12,
    "args": {
      "External id": 188274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188274,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 188274, "pid": 5, "tid": 7, "ts": 1716454224379079,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306950, "dur": 11,
    "args": {
      "External id": 188274,
      "cbid": 211, "correlation": 188274
    }
  },
  {
    "ph": "s", "id": 188274, "pid": 76337, "tid": -914061504, "ts": 1716454224306950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224379091, "dur": 5,
    "args": {
      "External id": 188276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188276,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 188276, "pid": 5, "tid": 7, "ts": 1716454224379091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224306964, "dur": 6,
    "args": {
      "External id": 188276,
      "cbid": 211, "correlation": 188276
    }
  },
  {
    "ph": "s", "id": 188276, "pid": 76337, "tid": -914061504, "ts": 1716454224306964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224379098, "dur": 28,
    "args": {
      "External id": 188286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188286,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188286, "pid": 5, "tid": 7, "ts": 1716454224379098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307031, "dur": 13,
    "args": {
      "External id": 188286,
      "cbid": 211, "correlation": 188286
    }
  },
  {
    "ph": "s", "id": 188286, "pid": 76337, "tid": -914061504, "ts": 1716454224307031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224379128, "dur": 31,
    "args": {
      "External id": 188306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188306,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 188306, "pid": 5, "tid": 7, "ts": 1716454224379128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307098, "dur": 11,
    "args": {
      "External id": 188306,
      "cbid": 211, "correlation": 188306
    }
  },
  {
    "ph": "s", "id": 188306, "pid": 76337, "tid": -914061504, "ts": 1716454224307098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224379160, "dur": 4,
    "args": {
      "External id": 188318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188318,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 188318, "pid": 5, "tid": 7, "ts": 1716454224379160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307120, "dur": 6,
    "args": {
      "External id": 188318,
      "cbid": 211, "correlation": 188318
    }
  },
  {
    "ph": "s", "id": 188318, "pid": 76337, "tid": -914061504, "ts": 1716454224307120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224379165, "dur": 29,
    "args": {
      "External id": 188321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188321, "pid": 5, "tid": 7, "ts": 1716454224379165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307139, "dur": 6,
    "args": {
      "External id": 188321,
      "cbid": 211, "correlation": 188321
    }
  },
  {
    "ph": "s", "id": 188321, "pid": 76337, "tid": -914061504, "ts": 1716454224307139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224379196, "dur": 20,
    "args": {
      "External id": 188330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188330,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188330, "pid": 5, "tid": 7, "ts": 1716454224379196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307179, "dur": 10,
    "args": {
      "External id": 188330,
      "cbid": 211, "correlation": 188330
    }
  },
  {
    "ph": "s", "id": 188330, "pid": 76337, "tid": -914061504, "ts": 1716454224307179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307242, "dur": 0,
    "args": {
      "External id": 188340,
      "cbid": 317, "correlation": 188340
    }
  },
  {
    "ph": "f", "id": 188340, "pid": 76337, "tid": -914061504, "ts": 1716454224307242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307243, "dur": 0,
    "args": {
      "External id": 188341,
      "cbid": 203, "correlation": 188341
    }
  },
  {
    "ph": "f", "id": 188341, "pid": 76337, "tid": -914061504, "ts": 1716454224307243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307243, "dur": 0,
    "args": {
      "External id": 188342,
      "cbid": 205, "correlation": 188342
    }
  },
  {
    "ph": "f", "id": 188342, "pid": 76337, "tid": -914061504, "ts": 1716454224307243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224379217, "dur": 23,
    "args": {
      "External id": 188346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188346,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188346, "pid": 5, "tid": 7, "ts": 1716454224379217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307257, "dur": 13,
    "args": {
      "External id": 188346,
      "cbid": 211, "correlation": 188346
    }
  },
  {
    "ph": "s", "id": 188346, "pid": 76337, "tid": -914061504, "ts": 1716454224307257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224379242, "dur": 44,
    "args": {
      "External id": 188348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188348,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188348, "pid": 5, "tid": 7, "ts": 1716454224379242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307273, "dur": 5,
    "args": {
      "External id": 188348,
      "cbid": 211, "correlation": 188348
    }
  },
  {
    "ph": "s", "id": 188348, "pid": 76337, "tid": -914061504, "ts": 1716454224307273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224379287, "dur": 637,
    "args": {
      "External id": 188350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188350,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188350, "pid": 5, "tid": 7, "ts": 1716454224379287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307284, "dur": 6,
    "args": {
      "External id": 188350,
      "cbid": 211, "correlation": 188350
    }
  },
  {
    "ph": "s", "id": 188350, "pid": 76337, "tid": -914061504, "ts": 1716454224307284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224379925, "dur": 21,
    "args": {
      "External id": 188352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188352,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188352, "pid": 5, "tid": 7, "ts": 1716454224379925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307293, "dur": 5,
    "args": {
      "External id": 188352,
      "cbid": 211, "correlation": 188352
    }
  },
  {
    "ph": "s", "id": 188352, "pid": 76337, "tid": -914061504, "ts": 1716454224307293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224379948, "dur": 32,
    "args": {
      "External id": 188358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188358, "pid": 5, "tid": 7, "ts": 1716454224379948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307322, "dur": 8,
    "args": {
      "External id": 188358,
      "cbid": 211, "correlation": 188358
    }
  },
  {
    "ph": "s", "id": 188358, "pid": 76337, "tid": -914061504, "ts": 1716454224307322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307380, "dur": 0,
    "args": {
      "External id": 188368,
      "cbid": 317, "correlation": 188368
    }
  },
  {
    "ph": "f", "id": 188368, "pid": 76337, "tid": -914061504, "ts": 1716454224307380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307380, "dur": 0,
    "args": {
      "External id": 188369,
      "cbid": 203, "correlation": 188369
    }
  },
  {
    "ph": "f", "id": 188369, "pid": 76337, "tid": -914061504, "ts": 1716454224307380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307381, "dur": 0,
    "args": {
      "External id": 188370,
      "cbid": 205, "correlation": 188370
    }
  },
  {
    "ph": "f", "id": 188370, "pid": 76337, "tid": -914061504, "ts": 1716454224307381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224379981, "dur": 38,
    "args": {
      "External id": 188374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188374,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188374, "pid": 5, "tid": 7, "ts": 1716454224379981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307393, "dur": 12,
    "args": {
      "External id": 188374,
      "cbid": 211, "correlation": 188374
    }
  },
  {
    "ph": "s", "id": 188374, "pid": 76337, "tid": -914061504, "ts": 1716454224307393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224380020, "dur": 186,
    "args": {
      "External id": 188376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188376,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188376, "pid": 5, "tid": 7, "ts": 1716454224380020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307411, "dur": 7,
    "args": {
      "External id": 188376,
      "cbid": 211, "correlation": 188376
    }
  },
  {
    "ph": "s", "id": 188376, "pid": 76337, "tid": -914061504, "ts": 1716454224307411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224380207, "dur": 22,
    "args": {
      "External id": 188378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188378, "pid": 5, "tid": 7, "ts": 1716454224380207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307421, "dur": 5,
    "args": {
      "External id": 188378,
      "cbid": 211, "correlation": 188378
    }
  },
  {
    "ph": "s", "id": 188378, "pid": 76337, "tid": -914061504, "ts": 1716454224307421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224380231, "dur": 32,
    "args": {
      "External id": 188384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188384, "pid": 5, "tid": 7, "ts": 1716454224380231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307447, "dur": 8,
    "args": {
      "External id": 188384,
      "cbid": 211, "correlation": 188384
    }
  },
  {
    "ph": "s", "id": 188384, "pid": 76337, "tid": -914061504, "ts": 1716454224307447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224380265, "dur": 27,
    "args": {
      "External id": 188392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188392,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188392, "pid": 5, "tid": 7, "ts": 1716454224380265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307476, "dur": 9,
    "args": {
      "External id": 188392,
      "cbid": 211, "correlation": 188392
    }
  },
  {
    "ph": "s", "id": 188392, "pid": 76337, "tid": -914061504, "ts": 1716454224307476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224380293, "dur": 20,
    "args": {
      "External id": 188400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188400,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188400, "pid": 5, "tid": 7, "ts": 1716454224380293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307506, "dur": 8,
    "args": {
      "External id": 188400,
      "cbid": 211, "correlation": 188400
    }
  },
  {
    "ph": "s", "id": 188400, "pid": 76337, "tid": -914061504, "ts": 1716454224307506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224380314, "dur": 30,
    "args": {
      "External id": 188420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188420,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 188420, "pid": 5, "tid": 7, "ts": 1716454224380314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307588, "dur": 12,
    "args": {
      "External id": 188420,
      "cbid": 211, "correlation": 188420
    }
  },
  {
    "ph": "s", "id": 188420, "pid": 76337, "tid": -914061504, "ts": 1716454224307588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224380346, "dur": 4,
    "args": {
      "External id": 188432,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188432,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 188432, "pid": 5, "tid": 7, "ts": 1716454224380346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307610, "dur": 6,
    "args": {
      "External id": 188432,
      "cbid": 211, "correlation": 188432
    }
  },
  {
    "ph": "s", "id": 188432, "pid": 76337, "tid": -914061504, "ts": 1716454224307610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224380351, "dur": 30,
    "args": {
      "External id": 188435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188435,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188435, "pid": 5, "tid": 7, "ts": 1716454224380351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307628, "dur": 6,
    "args": {
      "External id": 188435,
      "cbid": 211, "correlation": 188435
    }
  },
  {
    "ph": "s", "id": 188435, "pid": 76337, "tid": -914061504, "ts": 1716454224307628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307685, "dur": 0,
    "args": {
      "External id": 188446,
      "cbid": 317, "correlation": 188446
    }
  },
  {
    "ph": "f", "id": 188446, "pid": 76337, "tid": -914061504, "ts": 1716454224307685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307686, "dur": 0,
    "args": {
      "External id": 188447,
      "cbid": 203, "correlation": 188447
    }
  },
  {
    "ph": "f", "id": 188447, "pid": 76337, "tid": -914061504, "ts": 1716454224307686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307686, "dur": 0,
    "args": {
      "External id": 188448,
      "cbid": 205, "correlation": 188448
    }
  },
  {
    "ph": "f", "id": 188448, "pid": 76337, "tid": -914061504, "ts": 1716454224307686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224380382, "dur": 22,
    "args": {
      "External id": 188452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188452,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188452, "pid": 5, "tid": 7, "ts": 1716454224380382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307699, "dur": 12,
    "args": {
      "External id": 188452,
      "cbid": 211, "correlation": 188452
    }
  },
  {
    "ph": "s", "id": 188452, "pid": 76337, "tid": -914061504, "ts": 1716454224307699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224380406, "dur": 103,
    "args": {
      "External id": 188454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188454,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188454, "pid": 5, "tid": 7, "ts": 1716454224380406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307717, "dur": 7,
    "args": {
      "External id": 188454,
      "cbid": 211, "correlation": 188454
    }
  },
  {
    "ph": "s", "id": 188454, "pid": 76337, "tid": -914061504, "ts": 1716454224307717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224380510, "dur": 25,
    "args": {
      "External id": 188456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188456, "pid": 5, "tid": 7, "ts": 1716454224380510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307727, "dur": 5,
    "args": {
      "External id": 188456,
      "cbid": 211, "correlation": 188456
    }
  },
  {
    "ph": "s", "id": 188456, "pid": 76337, "tid": -914061504, "ts": 1716454224307727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224380536, "dur": 32,
    "args": {
      "External id": 188462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188462, "pid": 5, "tid": 7, "ts": 1716454224380536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307754, "dur": 8,
    "args": {
      "External id": 188462,
      "cbid": 211, "correlation": 188462
    }
  },
  {
    "ph": "s", "id": 188462, "pid": 76337, "tid": -914061504, "ts": 1716454224307754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224380569, "dur": 196,
    "args": {
      "External id": 188471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188471,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188471, "pid": 5, "tid": 7, "ts": 1716454224380569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307835, "dur": 13,
    "args": {
      "External id": 188471,
      "cbid": 211, "correlation": 188471
    }
  },
  {
    "ph": "s", "id": 188471, "pid": 76337, "tid": -914061504, "ts": 1716454224307835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224380766, "dur": 64,
    "args": {
      "External id": 188493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188493,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188493, "pid": 5, "tid": 7, "ts": 1716454224380766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307890, "dur": 10,
    "args": {
      "External id": 188493,
      "cbid": 211, "correlation": 188493
    }
  },
  {
    "ph": "s", "id": 188493, "pid": 76337, "tid": -914061504, "ts": 1716454224307890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307988, "dur": 1,
    "args": {
      "External id": 188504,
      "cbid": 251, "correlation": 188504
    }
  },
  {
    "ph": "f", "id": 188504, "pid": 76337, "tid": -914061504, "ts": 1716454224307988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224380831, "dur": 152,
    "args": {
      "External id": 188505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188505,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188505, "pid": 5, "tid": 7, "ts": 1716454224380831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224307994, "dur": 14,
    "args": {
      "External id": 188505,
      "cbid": 211, "correlation": 188505
    }
  },
  {
    "ph": "s", "id": 188505, "pid": 76337, "tid": -914061504, "ts": 1716454224307994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308065, "dur": 1,
    "args": {
      "External id": 188516,
      "cbid": 251, "correlation": 188516
    }
  },
  {
    "ph": "f", "id": 188516, "pid": 76337, "tid": -914061504, "ts": 1716454224308065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224380984, "dur": 144,
    "args": {
      "External id": 188517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188517,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188517, "pid": 5, "tid": 7, "ts": 1716454224380984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308070, "dur": 12,
    "args": {
      "External id": 188517,
      "cbid": 211, "correlation": 188517
    }
  },
  {
    "ph": "s", "id": 188517, "pid": 76337, "tid": -914061504, "ts": 1716454224308070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308135, "dur": 1,
    "args": {
      "External id": 188528,
      "cbid": 251, "correlation": 188528
    }
  },
  {
    "ph": "f", "id": 188528, "pid": 76337, "tid": -914061504, "ts": 1716454224308135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224381129, "dur": 143,
    "args": {
      "External id": 188529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188529,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188529, "pid": 5, "tid": 7, "ts": 1716454224381129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308139, "dur": 11,
    "args": {
      "External id": 188529,
      "cbid": 211, "correlation": 188529
    }
  },
  {
    "ph": "s", "id": 188529, "pid": 76337, "tid": -914061504, "ts": 1716454224308139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224381273, "dur": 1909,
    "args": {
      "External id": 188550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188550,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 188550, "pid": 5, "tid": 7, "ts": 1716454224381273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308221, "dur": 13,
    "args": {
      "External id": 188550,
      "cbid": 211, "correlation": 188550
    }
  },
  {
    "ph": "s", "id": 188550, "pid": 76337, "tid": -914061504, "ts": 1716454224308221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308318, "dur": 1,
    "args": {
      "External id": 188568,
      "cbid": 251, "correlation": 188568
    }
  },
  {
    "ph": "f", "id": 188568, "pid": 76337, "tid": -914061504, "ts": 1716454224308318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224383183, "dur": 147,
    "args": {
      "External id": 188570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188570,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 188570, "pid": 5, "tid": 7, "ts": 1716454224383183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308324, "dur": 14,
    "args": {
      "External id": 188570,
      "cbid": 211, "correlation": 188570
    }
  },
  {
    "ph": "s", "id": 188570, "pid": 76337, "tid": -914061504, "ts": 1716454224308324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224383332, "dur": 35,
    "args": {
      "External id": 188578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188578,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188578, "pid": 5, "tid": 7, "ts": 1716454224383332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308394, "dur": 11,
    "args": {
      "External id": 188578,
      "cbid": 211, "correlation": 188578
    }
  },
  {
    "ph": "s", "id": 188578, "pid": 76337, "tid": -914061504, "ts": 1716454224308394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224383369, "dur": 51,
    "args": {
      "External id": 188586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188586,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188586, "pid": 5, "tid": 7, "ts": 1716454224383369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308433, "dur": 9,
    "args": {
      "External id": 188586,
      "cbid": 211, "correlation": 188586
    }
  },
  {
    "ph": "s", "id": 188586, "pid": 76337, "tid": -914061504, "ts": 1716454224308433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224383422, "dur": 29,
    "args": {
      "External id": 188597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188597,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188597, "pid": 5, "tid": 7, "ts": 1716454224383422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308504, "dur": 12,
    "args": {
      "External id": 188597,
      "cbid": 211, "correlation": 188597
    }
  },
  {
    "ph": "s", "id": 188597, "pid": 76337, "tid": -914061504, "ts": 1716454224308504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224383452, "dur": 34,
    "args": {
      "External id": 188619,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188619,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188619, "pid": 5, "tid": 7, "ts": 1716454224383452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308536, "dur": 7,
    "args": {
      "External id": 188619,
      "cbid": 211, "correlation": 188619
    }
  },
  {
    "ph": "s", "id": 188619, "pid": 76337, "tid": -914061504, "ts": 1716454224308536,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308620, "dur": 1,
    "args": {
      "External id": 188630,
      "cbid": 251, "correlation": 188630
    }
  },
  {
    "ph": "f", "id": 188630, "pid": 76337, "tid": -914061504, "ts": 1716454224308620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224383487, "dur": 90,
    "args": {
      "External id": 188631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188631,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188631, "pid": 5, "tid": 7, "ts": 1716454224383487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308625, "dur": 14,
    "args": {
      "External id": 188631,
      "cbid": 211, "correlation": 188631
    }
  },
  {
    "ph": "s", "id": 188631, "pid": 76337, "tid": -914061504, "ts": 1716454224308625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308695, "dur": 1,
    "args": {
      "External id": 188642,
      "cbid": 251, "correlation": 188642
    }
  },
  {
    "ph": "f", "id": 188642, "pid": 76337, "tid": -914061504, "ts": 1716454224308695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308699, "dur": 0,
    "args": {
      "External id": 188643,
      "cbid": 251, "correlation": 188643
    }
  },
  {
    "ph": "f", "id": 188643, "pid": 76337, "tid": -914061504, "ts": 1716454224308699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224383578, "dur": 11,
    "args": {
      "External id": 188644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188644,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 188644, "pid": 5, "tid": 7, "ts": 1716454224383578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308700, "dur": 12,
    "args": {
      "External id": 188644,
      "cbid": 211, "correlation": 188644
    }
  },
  {
    "ph": "s", "id": 188644, "pid": 76337, "tid": -914061504, "ts": 1716454224308700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224383590, "dur": 5,
    "args": {
      "External id": 188646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188646,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 188646, "pid": 5, "tid": 7, "ts": 1716454224383590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308714, "dur": 6,
    "args": {
      "External id": 188646,
      "cbid": 211, "correlation": 188646
    }
  },
  {
    "ph": "s", "id": 188646, "pid": 76337, "tid": -914061504, "ts": 1716454224308714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308772, "dur": 1,
    "args": {
      "External id": 188657,
      "cbid": 251, "correlation": 188657
    }
  },
  {
    "ph": "f", "id": 188657, "pid": 76337, "tid": -914061504, "ts": 1716454224308772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308775, "dur": 0,
    "args": {
      "External id": 188658,
      "cbid": 251, "correlation": 188658
    }
  },
  {
    "ph": "f", "id": 188658, "pid": 76337, "tid": -914061504, "ts": 1716454224308775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224383597, "dur": 7,
    "args": {
      "External id": 188659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188659,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 188659, "pid": 5, "tid": 7, "ts": 1716454224383597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308776, "dur": 13,
    "args": {
      "External id": 188659,
      "cbid": 211, "correlation": 188659
    }
  },
  {
    "ph": "s", "id": 188659, "pid": 76337, "tid": -914061504, "ts": 1716454224308776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224383605, "dur": 3,
    "args": {
      "External id": 188661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188661,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 188661, "pid": 5, "tid": 7, "ts": 1716454224383605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308791, "dur": 5,
    "args": {
      "External id": 188661,
      "cbid": 211, "correlation": 188661
    }
  },
  {
    "ph": "s", "id": 188661, "pid": 76337, "tid": -914061504, "ts": 1716454224308791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224383609, "dur": 90,
    "args": {
      "External id": 188682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188682,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 188682, "pid": 5, "tid": 7, "ts": 1716454224383609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308864, "dur": 12,
    "args": {
      "External id": 188682,
      "cbid": 211, "correlation": 188682
    }
  },
  {
    "ph": "s", "id": 188682, "pid": 76337, "tid": -914061504, "ts": 1716454224308864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308961, "dur": 1,
    "args": {
      "External id": 188700,
      "cbid": 251, "correlation": 188700
    }
  },
  {
    "ph": "f", "id": 188700, "pid": 76337, "tid": -914061504, "ts": 1716454224308961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224383701, "dur": 97,
    "args": {
      "External id": 188702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188702,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188702, "pid": 5, "tid": 7, "ts": 1716454224383701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224308967, "dur": 22,
    "args": {
      "External id": 188702,
      "cbid": 211, "correlation": 188702
    }
  },
  {
    "ph": "s", "id": 188702, "pid": 76337, "tid": -914061504, "ts": 1716454224308967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224383799, "dur": 19,
    "args": {
      "External id": 188710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188710,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188710, "pid": 5, "tid": 7, "ts": 1716454224383799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309046, "dur": 12,
    "args": {
      "External id": 188710,
      "cbid": 211, "correlation": 188710
    }
  },
  {
    "ph": "s", "id": 188710, "pid": 76337, "tid": -914061504, "ts": 1716454224309046,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224383819, "dur": 38,
    "args": {
      "External id": 188718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188718, "pid": 5, "tid": 7, "ts": 1716454224383819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309088, "dur": 9,
    "args": {
      "External id": 188718,
      "cbid": 211, "correlation": 188718
    }
  },
  {
    "ph": "s", "id": 188718, "pid": 76337, "tid": -914061504, "ts": 1716454224309088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224383858, "dur": 35,
    "args": {
      "External id": 188740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188740,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188740, "pid": 5, "tid": 7, "ts": 1716454224383858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309138, "dur": 10,
    "args": {
      "External id": 188740,
      "cbid": 211, "correlation": 188740
    }
  },
  {
    "ph": "s", "id": 188740, "pid": 76337, "tid": -914061504, "ts": 1716454224309138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309225, "dur": 1,
    "args": {
      "External id": 188756,
      "cbid": 251, "correlation": 188756
    }
  },
  {
    "ph": "f", "id": 188756, "pid": 76337, "tid": -914061504, "ts": 1716454224309225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309230, "dur": 0,
    "args": {
      "External id": 188758,
      "cbid": 251, "correlation": 188758
    }
  },
  {
    "ph": "f", "id": 188758, "pid": 76337, "tid": -914061504, "ts": 1716454224309230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224383894, "dur": 531,
    "args": {
      "External id": 188759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188759,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 188759, "pid": 5, "tid": 7, "ts": 1716454224383894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309234, "dur": 13,
    "args": {
      "External id": 188759,
      "cbid": 211, "correlation": 188759
    }
  },
  {
    "ph": "s", "id": 188759, "pid": 76337, "tid": -914061504, "ts": 1716454224309234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224384427, "dur": 126,
    "args": {
      "External id": 188767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188767, "pid": 5, "tid": 7, "ts": 1716454224384427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309299, "dur": 13,
    "args": {
      "External id": 188767,
      "cbid": 211, "correlation": 188767
    }
  },
  {
    "ph": "s", "id": 188767, "pid": 76337, "tid": -914061504, "ts": 1716454224309299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224384554, "dur": 130,
    "args": {
      "External id": 188775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188775,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188775, "pid": 5, "tid": 7, "ts": 1716454224384554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309329, "dur": 9,
    "args": {
      "External id": 188775,
      "cbid": 211, "correlation": 188775
    }
  },
  {
    "ph": "s", "id": 188775, "pid": 76337, "tid": -914061504, "ts": 1716454224309329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309406, "dur": 1,
    "args": {
      "External id": 188791,
      "cbid": 251, "correlation": 188791
    }
  },
  {
    "ph": "f", "id": 188791, "pid": 76337, "tid": -914061504, "ts": 1716454224309406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224384685, "dur": 303,
    "args": {
      "External id": 188793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188793,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188793, "pid": 5, "tid": 7, "ts": 1716454224384685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309412, "dur": 12,
    "args": {
      "External id": 188793,
      "cbid": 211, "correlation": 188793
    }
  },
  {
    "ph": "s", "id": 188793, "pid": 76337, "tid": -914061504, "ts": 1716454224309412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224384990, "dur": 26,
    "args": {
      "External id": 188801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188801,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188801, "pid": 5, "tid": 7, "ts": 1716454224384990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309455, "dur": 9,
    "args": {
      "External id": 188801,
      "cbid": 211, "correlation": 188801
    }
  },
  {
    "ph": "s", "id": 188801, "pid": 76337, "tid": -914061504, "ts": 1716454224309455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224385018, "dur": 81,
    "args": {
      "External id": 188812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188812, "pid": 5, "tid": 7, "ts": 1716454224385018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309522, "dur": 12,
    "args": {
      "External id": 188812,
      "cbid": 211, "correlation": 188812
    }
  },
  {
    "ph": "s", "id": 188812, "pid": 76337, "tid": -914061504, "ts": 1716454224309522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309586, "dur": 0,
    "args": {
      "External id": 188824,
      "cbid": 317, "correlation": 188824
    }
  },
  {
    "ph": "f", "id": 188824, "pid": 76337, "tid": -914061504, "ts": 1716454224309586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309587, "dur": 0,
    "args": {
      "External id": 188825,
      "cbid": 203, "correlation": 188825
    }
  },
  {
    "ph": "f", "id": 188825, "pid": 76337, "tid": -914061504, "ts": 1716454224309587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309588, "dur": 0,
    "args": {
      "External id": 188826,
      "cbid": 205, "correlation": 188826
    }
  },
  {
    "ph": "f", "id": 188826, "pid": 76337, "tid": -914061504, "ts": 1716454224309588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224385100, "dur": 23,
    "args": {
      "External id": 188830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188830,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188830, "pid": 5, "tid": 7, "ts": 1716454224385100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309603, "dur": 12,
    "args": {
      "External id": 188830,
      "cbid": 211, "correlation": 188830
    }
  },
  {
    "ph": "s", "id": 188830, "pid": 76337, "tid": -914061504, "ts": 1716454224309603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224385125, "dur": 120,
    "args": {
      "External id": 188832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188832,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188832, "pid": 5, "tid": 7, "ts": 1716454224385125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309621, "dur": 6,
    "args": {
      "External id": 188832,
      "cbid": 211, "correlation": 188832
    }
  },
  {
    "ph": "s", "id": 188832, "pid": 76337, "tid": -914061504, "ts": 1716454224309621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224385246, "dur": 24,
    "args": {
      "External id": 188834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188834,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188834, "pid": 5, "tid": 7, "ts": 1716454224385246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309631, "dur": 5,
    "args": {
      "External id": 188834,
      "cbid": 211, "correlation": 188834
    }
  },
  {
    "ph": "s", "id": 188834, "pid": 76337, "tid": -914061504, "ts": 1716454224309631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224385271, "dur": 33,
    "args": {
      "External id": 188840,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188840,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188840, "pid": 5, "tid": 7, "ts": 1716454224385271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309660, "dur": 8,
    "args": {
      "External id": 188840,
      "cbid": 211, "correlation": 188840
    }
  },
  {
    "ph": "s", "id": 188840, "pid": 76337, "tid": -914061504, "ts": 1716454224309660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224385305, "dur": 27,
    "args": {
      "External id": 188848,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188848,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188848, "pid": 5, "tid": 7, "ts": 1716454224385305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309692, "dur": 8,
    "args": {
      "External id": 188848,
      "cbid": 211, "correlation": 188848
    }
  },
  {
    "ph": "s", "id": 188848, "pid": 76337, "tid": -914061504, "ts": 1716454224309692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224385333, "dur": 45,
    "args": {
      "External id": 188857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188857,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188857, "pid": 5, "tid": 7, "ts": 1716454224385333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309728, "dur": 10,
    "args": {
      "External id": 188857,
      "cbid": 211, "correlation": 188857
    }
  },
  {
    "ph": "s", "id": 188857, "pid": 76337, "tid": -914061504, "ts": 1716454224309728,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224385379, "dur": 42,
    "args": {
      "External id": 188877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188877,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 188877, "pid": 5, "tid": 7, "ts": 1716454224385379,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309800, "dur": 11,
    "args": {
      "External id": 188877,
      "cbid": 211, "correlation": 188877
    }
  },
  {
    "ph": "s", "id": 188877, "pid": 76337, "tid": -914061504, "ts": 1716454224309800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224385423, "dur": 5,
    "args": {
      "External id": 188889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188889,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 188889, "pid": 5, "tid": 7, "ts": 1716454224385423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309821, "dur": 6,
    "args": {
      "External id": 188889,
      "cbid": 211, "correlation": 188889
    }
  },
  {
    "ph": "s", "id": 188889, "pid": 76337, "tid": -914061504, "ts": 1716454224309821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224385429, "dur": 43,
    "args": {
      "External id": 188892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188892, "pid": 5, "tid": 7, "ts": 1716454224385429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309840, "dur": 7,
    "args": {
      "External id": 188892,
      "cbid": 211, "correlation": 188892
    }
  },
  {
    "ph": "s", "id": 188892, "pid": 76337, "tid": -914061504, "ts": 1716454224309840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224385473, "dur": 31,
    "args": {
      "External id": 188901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188901,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188901, "pid": 5, "tid": 7, "ts": 1716454224385473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309879, "dur": 10,
    "args": {
      "External id": 188901,
      "cbid": 211, "correlation": 188901
    }
  },
  {
    "ph": "s", "id": 188901, "pid": 76337, "tid": -914061504, "ts": 1716454224309879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309931, "dur": 0,
    "args": {
      "External id": 188911,
      "cbid": 317, "correlation": 188911
    }
  },
  {
    "ph": "f", "id": 188911, "pid": 76337, "tid": -914061504, "ts": 1716454224309931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309932, "dur": 0,
    "args": {
      "External id": 188912,
      "cbid": 203, "correlation": 188912
    }
  },
  {
    "ph": "f", "id": 188912, "pid": 76337, "tid": -914061504, "ts": 1716454224309932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309932, "dur": 0,
    "args": {
      "External id": 188913,
      "cbid": 205, "correlation": 188913
    }
  },
  {
    "ph": "f", "id": 188913, "pid": 76337, "tid": -914061504, "ts": 1716454224309932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224385505, "dur": 31,
    "args": {
      "External id": 188917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188917,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188917, "pid": 5, "tid": 7, "ts": 1716454224385505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309948, "dur": 11,
    "args": {
      "External id": 188917,
      "cbid": 211, "correlation": 188917
    }
  },
  {
    "ph": "s", "id": 188917, "pid": 76337, "tid": -914061504, "ts": 1716454224309948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224385538, "dur": 63,
    "args": {
      "External id": 188919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188919,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188919, "pid": 5, "tid": 7, "ts": 1716454224385538,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309961, "dur": 5,
    "args": {
      "External id": 188919,
      "cbid": 211, "correlation": 188919
    }
  },
  {
    "ph": "s", "id": 188919, "pid": 76337, "tid": -914061504, "ts": 1716454224309961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224385603, "dur": 966,
    "args": {
      "External id": 188921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188921,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 188921, "pid": 5, "tid": 7, "ts": 1716454224385603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309972, "dur": 15,
    "args": {
      "External id": 188921,
      "cbid": 211, "correlation": 188921
    }
  },
  {
    "ph": "s", "id": 188921, "pid": 76337, "tid": -914061504, "ts": 1716454224309972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224386570, "dur": 21,
    "args": {
      "External id": 188923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188923,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188923, "pid": 5, "tid": 7, "ts": 1716454224386570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224309991, "dur": 5,
    "args": {
      "External id": 188923,
      "cbid": 211, "correlation": 188923
    }
  },
  {
    "ph": "s", "id": 188923, "pid": 76337, "tid": -914061504, "ts": 1716454224309991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224386592, "dur": 33,
    "args": {
      "External id": 188929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188929,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188929, "pid": 5, "tid": 7, "ts": 1716454224386592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310019, "dur": 8,
    "args": {
      "External id": 188929,
      "cbid": 211, "correlation": 188929
    }
  },
  {
    "ph": "s", "id": 188929, "pid": 76337, "tid": -914061504, "ts": 1716454224310019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224386626, "dur": 3,
    "args": {
      "External id": 188937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188937,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 188937, "pid": 5, "tid": 7, "ts": 1716454224386626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310064, "dur": 9,
    "args": {
      "External id": 188937,
      "cbid": 211, "correlation": 188937
    }
  },
  {
    "ph": "s", "id": 188937, "pid": 76337, "tid": -914061504, "ts": 1716454224310064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310129, "dur": 1,
    "args": {
      "External id": 188953,
      "cbid": 251, "correlation": 188953
    }
  },
  {
    "ph": "f", "id": 188953, "pid": 76337, "tid": -914061504, "ts": 1716454224310129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310134, "dur": 0,
    "args": {
      "External id": 188955,
      "cbid": 251, "correlation": 188955
    }
  },
  {
    "ph": "f", "id": 188955, "pid": 76337, "tid": -914061504, "ts": 1716454224310134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224386631, "dur": 12,
    "args": {
      "External id": 188956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188956,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 188956, "pid": 5, "tid": 7, "ts": 1716454224386631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310135, "dur": 12,
    "args": {
      "External id": 188956,
      "cbid": 211, "correlation": 188956
    }
  },
  {
    "ph": "s", "id": 188956, "pid": 76337, "tid": -914061504, "ts": 1716454224310135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224386644, "dur": 5,
    "args": {
      "External id": 188958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188958,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 188958, "pid": 5, "tid": 7, "ts": 1716454224386644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310149, "dur": 5,
    "args": {
      "External id": 188958,
      "cbid": 211, "correlation": 188958
    }
  },
  {
    "ph": "s", "id": 188958, "pid": 76337, "tid": -914061504, "ts": 1716454224310149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224386651, "dur": 29,
    "args": {
      "External id": 188968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188968,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 188968, "pid": 5, "tid": 7, "ts": 1716454224386651,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310206, "dur": 12,
    "args": {
      "External id": 188968,
      "cbid": 211, "correlation": 188968
    }
  },
  {
    "ph": "s", "id": 188968, "pid": 76337, "tid": -914061504, "ts": 1716454224310206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224386680, "dur": 31,
    "args": {
      "External id": 188988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 188988,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 188988, "pid": 5, "tid": 7, "ts": 1716454224386680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310272, "dur": 10,
    "args": {
      "External id": 188988,
      "cbid": 211, "correlation": 188988
    }
  },
  {
    "ph": "s", "id": 188988, "pid": 76337, "tid": -914061504, "ts": 1716454224310272,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224386713, "dur": 4,
    "args": {
      "External id": 189000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189000,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 189000, "pid": 5, "tid": 7, "ts": 1716454224386713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310292, "dur": 7,
    "args": {
      "External id": 189000,
      "cbid": 211, "correlation": 189000
    }
  },
  {
    "ph": "s", "id": 189000, "pid": 76337, "tid": -914061504, "ts": 1716454224310292,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224386718, "dur": 30,
    "args": {
      "External id": 189003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189003,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189003, "pid": 5, "tid": 7, "ts": 1716454224386718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310312, "dur": 6,
    "args": {
      "External id": 189003,
      "cbid": 211, "correlation": 189003
    }
  },
  {
    "ph": "s", "id": 189003, "pid": 76337, "tid": -914061504, "ts": 1716454224310312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224386750, "dur": 20,
    "args": {
      "External id": 189012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189012,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189012, "pid": 5, "tid": 7, "ts": 1716454224386750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310352, "dur": 9,
    "args": {
      "External id": 189012,
      "cbid": 211, "correlation": 189012
    }
  },
  {
    "ph": "s", "id": 189012, "pid": 76337, "tid": -914061504, "ts": 1716454224310352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310414, "dur": 0,
    "args": {
      "External id": 189022,
      "cbid": 317, "correlation": 189022
    }
  },
  {
    "ph": "f", "id": 189022, "pid": 76337, "tid": -914061504, "ts": 1716454224310414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310414, "dur": 0,
    "args": {
      "External id": 189023,
      "cbid": 203, "correlation": 189023
    }
  },
  {
    "ph": "f", "id": 189023, "pid": 76337, "tid": -914061504, "ts": 1716454224310414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310415, "dur": 0,
    "args": {
      "External id": 189024,
      "cbid": 205, "correlation": 189024
    }
  },
  {
    "ph": "f", "id": 189024, "pid": 76337, "tid": -914061504, "ts": 1716454224310415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224386771, "dur": 22,
    "args": {
      "External id": 189028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189028,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189028, "pid": 5, "tid": 7, "ts": 1716454224386771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310429, "dur": 12,
    "args": {
      "External id": 189028,
      "cbid": 211, "correlation": 189028
    }
  },
  {
    "ph": "s", "id": 189028, "pid": 76337, "tid": -914061504, "ts": 1716454224310429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224386794, "dur": 44,
    "args": {
      "External id": 189030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189030,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189030, "pid": 5, "tid": 7, "ts": 1716454224386794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310444, "dur": 6,
    "args": {
      "External id": 189030,
      "cbid": 211, "correlation": 189030
    }
  },
  {
    "ph": "s", "id": 189030, "pid": 76337, "tid": -914061504, "ts": 1716454224310444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224386840, "dur": 646,
    "args": {
      "External id": 189032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189032,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189032, "pid": 5, "tid": 7, "ts": 1716454224386840,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310456, "dur": 6,
    "args": {
      "External id": 189032,
      "cbid": 211, "correlation": 189032
    }
  },
  {
    "ph": "s", "id": 189032, "pid": 76337, "tid": -914061504, "ts": 1716454224310456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224387487, "dur": 22,
    "args": {
      "External id": 189034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189034,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189034, "pid": 5, "tid": 7, "ts": 1716454224387487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310465, "dur": 5,
    "args": {
      "External id": 189034,
      "cbid": 211, "correlation": 189034
    }
  },
  {
    "ph": "s", "id": 189034, "pid": 76337, "tid": -914061504, "ts": 1716454224310465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224387510, "dur": 33,
    "args": {
      "External id": 189040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189040, "pid": 5, "tid": 7, "ts": 1716454224387510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310493, "dur": 9,
    "args": {
      "External id": 189040,
      "cbid": 211, "correlation": 189040
    }
  },
  {
    "ph": "s", "id": 189040, "pid": 76337, "tid": -914061504, "ts": 1716454224310493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310551, "dur": 0,
    "args": {
      "External id": 189050,
      "cbid": 317, "correlation": 189050
    }
  },
  {
    "ph": "f", "id": 189050, "pid": 76337, "tid": -914061504, "ts": 1716454224310551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310552, "dur": 0,
    "args": {
      "External id": 189051,
      "cbid": 203, "correlation": 189051
    }
  },
  {
    "ph": "f", "id": 189051, "pid": 76337, "tid": -914061504, "ts": 1716454224310552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310553, "dur": 0,
    "args": {
      "External id": 189052,
      "cbid": 205, "correlation": 189052
    }
  },
  {
    "ph": "f", "id": 189052, "pid": 76337, "tid": -914061504, "ts": 1716454224310553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224387544, "dur": 30,
    "args": {
      "External id": 189056,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189056,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189056, "pid": 5, "tid": 7, "ts": 1716454224387544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310566, "dur": 12,
    "args": {
      "External id": 189056,
      "cbid": 211, "correlation": 189056
    }
  },
  {
    "ph": "s", "id": 189056, "pid": 76337, "tid": -914061504, "ts": 1716454224310566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224387575, "dur": 153,
    "args": {
      "External id": 189058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189058,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189058, "pid": 5, "tid": 7, "ts": 1716454224387575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310584, "dur": 7,
    "args": {
      "External id": 189058,
      "cbid": 211, "correlation": 189058
    }
  },
  {
    "ph": "s", "id": 189058, "pid": 76337, "tid": -914061504, "ts": 1716454224310584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224387730, "dur": 25,
    "args": {
      "External id": 189060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189060,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189060, "pid": 5, "tid": 7, "ts": 1716454224387730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310596, "dur": 5,
    "args": {
      "External id": 189060,
      "cbid": 211, "correlation": 189060
    }
  },
  {
    "ph": "s", "id": 189060, "pid": 76337, "tid": -914061504, "ts": 1716454224310596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224387756, "dur": 32,
    "args": {
      "External id": 189066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189066, "pid": 5, "tid": 7, "ts": 1716454224387756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310621, "dur": 8,
    "args": {
      "External id": 189066,
      "cbid": 211, "correlation": 189066
    }
  },
  {
    "ph": "s", "id": 189066, "pid": 76337, "tid": -914061504, "ts": 1716454224310621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224387789, "dur": 27,
    "args": {
      "External id": 189074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189074,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189074, "pid": 5, "tid": 7, "ts": 1716454224387789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310650, "dur": 8,
    "args": {
      "External id": 189074,
      "cbid": 211, "correlation": 189074
    }
  },
  {
    "ph": "s", "id": 189074, "pid": 76337, "tid": -914061504, "ts": 1716454224310650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224387817, "dur": 20,
    "args": {
      "External id": 189082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189082,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189082, "pid": 5, "tid": 7, "ts": 1716454224387817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310678, "dur": 8,
    "args": {
      "External id": 189082,
      "cbid": 211, "correlation": 189082
    }
  },
  {
    "ph": "s", "id": 189082, "pid": 76337, "tid": -914061504, "ts": 1716454224310678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224387839, "dur": 31,
    "args": {
      "External id": 189102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189102,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 189102, "pid": 5, "tid": 7, "ts": 1716454224387839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310761, "dur": 12,
    "args": {
      "External id": 189102,
      "cbid": 211, "correlation": 189102
    }
  },
  {
    "ph": "s", "id": 189102, "pid": 76337, "tid": -914061504, "ts": 1716454224310761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224387871, "dur": 5,
    "args": {
      "External id": 189114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189114,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 189114, "pid": 5, "tid": 7, "ts": 1716454224387871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310782, "dur": 6,
    "args": {
      "External id": 189114,
      "cbid": 211, "correlation": 189114
    }
  },
  {
    "ph": "s", "id": 189114, "pid": 76337, "tid": -914061504, "ts": 1716454224310782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224387877, "dur": 31,
    "args": {
      "External id": 189117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189117,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189117, "pid": 5, "tid": 7, "ts": 1716454224387877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310801, "dur": 6,
    "args": {
      "External id": 189117,
      "cbid": 211, "correlation": 189117
    }
  },
  {
    "ph": "s", "id": 189117, "pid": 76337, "tid": -914061504, "ts": 1716454224310801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310857, "dur": 0,
    "args": {
      "External id": 189128,
      "cbid": 317, "correlation": 189128
    }
  },
  {
    "ph": "f", "id": 189128, "pid": 76337, "tid": -914061504, "ts": 1716454224310857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310858, "dur": 0,
    "args": {
      "External id": 189129,
      "cbid": 203, "correlation": 189129
    }
  },
  {
    "ph": "f", "id": 189129, "pid": 76337, "tid": -914061504, "ts": 1716454224310858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310859, "dur": 0,
    "args": {
      "External id": 189130,
      "cbid": 205, "correlation": 189130
    }
  },
  {
    "ph": "f", "id": 189130, "pid": 76337, "tid": -914061504, "ts": 1716454224310859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224387909, "dur": 22,
    "args": {
      "External id": 189134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189134, "pid": 5, "tid": 7, "ts": 1716454224387909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310872, "dur": 12,
    "args": {
      "External id": 189134,
      "cbid": 211, "correlation": 189134
    }
  },
  {
    "ph": "s", "id": 189134, "pid": 76337, "tid": -914061504, "ts": 1716454224310872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224387933, "dur": 106,
    "args": {
      "External id": 189136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189136,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189136, "pid": 5, "tid": 7, "ts": 1716454224387933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310889, "dur": 6,
    "args": {
      "External id": 189136,
      "cbid": 211, "correlation": 189136
    }
  },
  {
    "ph": "s", "id": 189136, "pid": 76337, "tid": -914061504, "ts": 1716454224310889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224388040, "dur": 24,
    "args": {
      "External id": 189138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189138,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189138, "pid": 5, "tid": 7, "ts": 1716454224388040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310900, "dur": 5,
    "args": {
      "External id": 189138,
      "cbid": 211, "correlation": 189138
    }
  },
  {
    "ph": "s", "id": 189138, "pid": 76337, "tid": -914061504, "ts": 1716454224310900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224388065, "dur": 33,
    "args": {
      "External id": 189144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189144, "pid": 5, "tid": 7, "ts": 1716454224388065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224310926, "dur": 8,
    "args": {
      "External id": 189144,
      "cbid": 211, "correlation": 189144
    }
  },
  {
    "ph": "s", "id": 189144, "pid": 76337, "tid": -914061504, "ts": 1716454224310926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224388099, "dur": 199,
    "args": {
      "External id": 189153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189153,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189153, "pid": 5, "tid": 7, "ts": 1716454224388099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311015, "dur": 14,
    "args": {
      "External id": 189153,
      "cbid": 211, "correlation": 189153
    }
  },
  {
    "ph": "s", "id": 189153, "pid": 76337, "tid": -914061504, "ts": 1716454224311015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224388299, "dur": 66,
    "args": {
      "External id": 189175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189175,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189175, "pid": 5, "tid": 7, "ts": 1716454224388299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311073, "dur": 10,
    "args": {
      "External id": 189175,
      "cbid": 211, "correlation": 189175
    }
  },
  {
    "ph": "s", "id": 189175, "pid": 76337, "tid": -914061504, "ts": 1716454224311073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311160, "dur": 1,
    "args": {
      "External id": 189186,
      "cbid": 251, "correlation": 189186
    }
  },
  {
    "ph": "f", "id": 189186, "pid": 76337, "tid": -914061504, "ts": 1716454224311160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224388366, "dur": 151,
    "args": {
      "External id": 189187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189187,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189187, "pid": 5, "tid": 7, "ts": 1716454224388366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311165, "dur": 13,
    "args": {
      "External id": 189187,
      "cbid": 211, "correlation": 189187
    }
  },
  {
    "ph": "s", "id": 189187, "pid": 76337, "tid": -914061504, "ts": 1716454224311165,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311235, "dur": 1,
    "args": {
      "External id": 189198,
      "cbid": 251, "correlation": 189198
    }
  },
  {
    "ph": "f", "id": 189198, "pid": 76337, "tid": -914061504, "ts": 1716454224311235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224388519, "dur": 145,
    "args": {
      "External id": 189199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189199,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189199, "pid": 5, "tid": 7, "ts": 1716454224388519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311239, "dur": 12,
    "args": {
      "External id": 189199,
      "cbid": 211, "correlation": 189199
    }
  },
  {
    "ph": "s", "id": 189199, "pid": 76337, "tid": -914061504, "ts": 1716454224311239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311306, "dur": 1,
    "args": {
      "External id": 189210,
      "cbid": 251, "correlation": 189210
    }
  },
  {
    "ph": "f", "id": 189210, "pid": 76337, "tid": -914061504, "ts": 1716454224311306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224388665, "dur": 145,
    "args": {
      "External id": 189211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189211,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189211, "pid": 5, "tid": 7, "ts": 1716454224388665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311310, "dur": 11,
    "args": {
      "External id": 189211,
      "cbid": 211, "correlation": 189211
    }
  },
  {
    "ph": "s", "id": 189211, "pid": 76337, "tid": -914061504, "ts": 1716454224311310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224388812, "dur": 1944,
    "args": {
      "External id": 189232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189232,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 189232, "pid": 5, "tid": 7, "ts": 1716454224388812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311389, "dur": 13,
    "args": {
      "External id": 189232,
      "cbid": 211, "correlation": 189232
    }
  },
  {
    "ph": "s", "id": 189232, "pid": 76337, "tid": -914061504, "ts": 1716454224311389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311486, "dur": 1,
    "args": {
      "External id": 189250,
      "cbid": 251, "correlation": 189250
    }
  },
  {
    "ph": "f", "id": 189250, "pid": 76337, "tid": -914061504, "ts": 1716454224311486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224390757, "dur": 149,
    "args": {
      "External id": 189252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189252,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 189252, "pid": 5, "tid": 7, "ts": 1716454224390757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311492, "dur": 14,
    "args": {
      "External id": 189252,
      "cbid": 211, "correlation": 189252
    }
  },
  {
    "ph": "s", "id": 189252, "pid": 76337, "tid": -914061504, "ts": 1716454224311492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224390908, "dur": 35,
    "args": {
      "External id": 189260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189260,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189260, "pid": 5, "tid": 7, "ts": 1716454224390908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311561, "dur": 12,
    "args": {
      "External id": 189260,
      "cbid": 211, "correlation": 189260
    }
  },
  {
    "ph": "s", "id": 189260, "pid": 76337, "tid": -914061504, "ts": 1716454224311561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224390945, "dur": 52,
    "args": {
      "External id": 189268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189268,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189268, "pid": 5, "tid": 7, "ts": 1716454224390945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311600, "dur": 9,
    "args": {
      "External id": 189268,
      "cbid": 211, "correlation": 189268
    }
  },
  {
    "ph": "s", "id": 189268, "pid": 76337, "tid": -914061504, "ts": 1716454224311600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224390997, "dur": 31,
    "args": {
      "External id": 189279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189279,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189279, "pid": 5, "tid": 7, "ts": 1716454224390997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311670, "dur": 12,
    "args": {
      "External id": 189279,
      "cbid": 211, "correlation": 189279
    }
  },
  {
    "ph": "s", "id": 189279, "pid": 76337, "tid": -914061504, "ts": 1716454224311670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224391029, "dur": 34,
    "args": {
      "External id": 189301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189301,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189301, "pid": 5, "tid": 7, "ts": 1716454224391029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311701, "dur": 8,
    "args": {
      "External id": 189301,
      "cbid": 211, "correlation": 189301
    }
  },
  {
    "ph": "s", "id": 189301, "pid": 76337, "tid": -914061504, "ts": 1716454224311701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311784, "dur": 1,
    "args": {
      "External id": 189312,
      "cbid": 251, "correlation": 189312
    }
  },
  {
    "ph": "f", "id": 189312, "pid": 76337, "tid": -914061504, "ts": 1716454224311784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224391065, "dur": 91,
    "args": {
      "External id": 189313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189313,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189313, "pid": 5, "tid": 7, "ts": 1716454224391065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311790, "dur": 14,
    "args": {
      "External id": 189313,
      "cbid": 211, "correlation": 189313
    }
  },
  {
    "ph": "s", "id": 189313, "pid": 76337, "tid": -914061504, "ts": 1716454224311790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311858, "dur": 1,
    "args": {
      "External id": 189324,
      "cbid": 251, "correlation": 189324
    }
  },
  {
    "ph": "f", "id": 189324, "pid": 76337, "tid": -914061504, "ts": 1716454224311858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311862, "dur": 0,
    "args": {
      "External id": 189325,
      "cbid": 251, "correlation": 189325
    }
  },
  {
    "ph": "f", "id": 189325, "pid": 76337, "tid": -914061504, "ts": 1716454224311862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224391158, "dur": 11,
    "args": {
      "External id": 189326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189326,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 189326, "pid": 5, "tid": 7, "ts": 1716454224391158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311863, "dur": 12,
    "args": {
      "External id": 189326,
      "cbid": 211, "correlation": 189326
    }
  },
  {
    "ph": "s", "id": 189326, "pid": 76337, "tid": -914061504, "ts": 1716454224311863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224391170, "dur": 5,
    "args": {
      "External id": 189328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189328,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 189328, "pid": 5, "tid": 7, "ts": 1716454224391170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311877, "dur": 6,
    "args": {
      "External id": 189328,
      "cbid": 211, "correlation": 189328
    }
  },
  {
    "ph": "s", "id": 189328, "pid": 76337, "tid": -914061504, "ts": 1716454224311877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311934, "dur": 1,
    "args": {
      "External id": 189339,
      "cbid": 251, "correlation": 189339
    }
  },
  {
    "ph": "f", "id": 189339, "pid": 76337, "tid": -914061504, "ts": 1716454224311934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311937, "dur": 0,
    "args": {
      "External id": 189340,
      "cbid": 251, "correlation": 189340
    }
  },
  {
    "ph": "f", "id": 189340, "pid": 76337, "tid": -914061504, "ts": 1716454224311937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224391177, "dur": 7,
    "args": {
      "External id": 189341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189341,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 189341, "pid": 5, "tid": 7, "ts": 1716454224391177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311939, "dur": 12,
    "args": {
      "External id": 189341,
      "cbid": 211, "correlation": 189341
    }
  },
  {
    "ph": "s", "id": 189341, "pid": 76337, "tid": -914061504, "ts": 1716454224311939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224391186, "dur": 4,
    "args": {
      "External id": 189343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189343,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 189343, "pid": 5, "tid": 7, "ts": 1716454224391186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224311952, "dur": 7,
    "args": {
      "External id": 189343,
      "cbid": 211, "correlation": 189343
    }
  },
  {
    "ph": "s", "id": 189343, "pid": 76337, "tid": -914061504, "ts": 1716454224311952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224391190, "dur": 92,
    "args": {
      "External id": 189364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189364,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 189364, "pid": 5, "tid": 7, "ts": 1716454224391190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312035, "dur": 13,
    "args": {
      "External id": 189364,
      "cbid": 211, "correlation": 189364
    }
  },
  {
    "ph": "s", "id": 189364, "pid": 76337, "tid": -914061504, "ts": 1716454224312035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312133, "dur": 1,
    "args": {
      "External id": 189382,
      "cbid": 251, "correlation": 189382
    }
  },
  {
    "ph": "f", "id": 189382, "pid": 76337, "tid": -914061504, "ts": 1716454224312133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224391283, "dur": 97,
    "args": {
      "External id": 189384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189384,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189384, "pid": 5, "tid": 7, "ts": 1716454224391283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312139, "dur": 13,
    "args": {
      "External id": 189384,
      "cbid": 211, "correlation": 189384
    }
  },
  {
    "ph": "s", "id": 189384, "pid": 76337, "tid": -914061504, "ts": 1716454224312139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224391381, "dur": 19,
    "args": {
      "External id": 189392,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189392,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189392, "pid": 5, "tid": 7, "ts": 1716454224391381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312210, "dur": 12,
    "args": {
      "External id": 189392,
      "cbid": 211, "correlation": 189392
    }
  },
  {
    "ph": "s", "id": 189392, "pid": 76337, "tid": -914061504, "ts": 1716454224312210,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224391402, "dur": 38,
    "args": {
      "External id": 189400,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189400,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189400, "pid": 5, "tid": 7, "ts": 1716454224391402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312251, "dur": 10,
    "args": {
      "External id": 189400,
      "cbid": 211, "correlation": 189400
    }
  },
  {
    "ph": "s", "id": 189400, "pid": 76337, "tid": -914061504, "ts": 1716454224312251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224391441, "dur": 34,
    "args": {
      "External id": 189422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189422,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189422, "pid": 5, "tid": 7, "ts": 1716454224391441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312302, "dur": 10,
    "args": {
      "External id": 189422,
      "cbid": 211, "correlation": 189422
    }
  },
  {
    "ph": "s", "id": 189422, "pid": 76337, "tid": -914061504, "ts": 1716454224312302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312392, "dur": 1,
    "args": {
      "External id": 189438,
      "cbid": 251, "correlation": 189438
    }
  },
  {
    "ph": "f", "id": 189438, "pid": 76337, "tid": -914061504, "ts": 1716454224312392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312397, "dur": 0,
    "args": {
      "External id": 189440,
      "cbid": 251, "correlation": 189440
    }
  },
  {
    "ph": "f", "id": 189440, "pid": 76337, "tid": -914061504, "ts": 1716454224312397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224391477, "dur": 538,
    "args": {
      "External id": 189441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189441,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 189441, "pid": 5, "tid": 7, "ts": 1716454224391477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312400, "dur": 14,
    "args": {
      "External id": 189441,
      "cbid": 211, "correlation": 189441
    }
  },
  {
    "ph": "s", "id": 189441, "pid": 76337, "tid": -914061504, "ts": 1716454224312400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224392016, "dur": 124,
    "args": {
      "External id": 189449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189449,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189449, "pid": 5, "tid": 7, "ts": 1716454224392016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312465, "dur": 12,
    "args": {
      "External id": 189449,
      "cbid": 211, "correlation": 189449
    }
  },
  {
    "ph": "s", "id": 189449, "pid": 76337, "tid": -914061504, "ts": 1716454224312465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224392141, "dur": 129,
    "args": {
      "External id": 189457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189457, "pid": 5, "tid": 7, "ts": 1716454224392141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312495, "dur": 8,
    "args": {
      "External id": 189457,
      "cbid": 211, "correlation": 189457
    }
  },
  {
    "ph": "s", "id": 189457, "pid": 76337, "tid": -914061504, "ts": 1716454224312495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312573, "dur": 1,
    "args": {
      "External id": 189473,
      "cbid": 251, "correlation": 189473
    }
  },
  {
    "ph": "f", "id": 189473, "pid": 76337, "tid": -914061504, "ts": 1716454224312573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224392272, "dur": 309,
    "args": {
      "External id": 189475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189475,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189475, "pid": 5, "tid": 7, "ts": 1716454224392272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312578, "dur": 12,
    "args": {
      "External id": 189475,
      "cbid": 211, "correlation": 189475
    }
  },
  {
    "ph": "s", "id": 189475, "pid": 76337, "tid": -914061504, "ts": 1716454224312578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224392583, "dur": 28,
    "args": {
      "External id": 189483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189483,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189483, "pid": 5, "tid": 7, "ts": 1716454224392583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312621, "dur": 10,
    "args": {
      "External id": 189483,
      "cbid": 211, "correlation": 189483
    }
  },
  {
    "ph": "s", "id": 189483, "pid": 76337, "tid": -914061504, "ts": 1716454224312621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224392612, "dur": 82,
    "args": {
      "External id": 189494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189494,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189494, "pid": 5, "tid": 7, "ts": 1716454224392612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312688, "dur": 12,
    "args": {
      "External id": 189494,
      "cbid": 211, "correlation": 189494
    }
  },
  {
    "ph": "s", "id": 189494, "pid": 76337, "tid": -914061504, "ts": 1716454224312688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312750, "dur": 0,
    "args": {
      "External id": 189506,
      "cbid": 317, "correlation": 189506
    }
  },
  {
    "ph": "f", "id": 189506, "pid": 76337, "tid": -914061504, "ts": 1716454224312750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312751, "dur": 0,
    "args": {
      "External id": 189507,
      "cbid": 203, "correlation": 189507
    }
  },
  {
    "ph": "f", "id": 189507, "pid": 76337, "tid": -914061504, "ts": 1716454224312751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312751, "dur": 0,
    "args": {
      "External id": 189508,
      "cbid": 205, "correlation": 189508
    }
  },
  {
    "ph": "f", "id": 189508, "pid": 76337, "tid": -914061504, "ts": 1716454224312751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224392695, "dur": 23,
    "args": {
      "External id": 189512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189512,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189512, "pid": 5, "tid": 7, "ts": 1716454224392695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312766, "dur": 12,
    "args": {
      "External id": 189512,
      "cbid": 211, "correlation": 189512
    }
  },
  {
    "ph": "s", "id": 189512, "pid": 76337, "tid": -914061504, "ts": 1716454224312766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224392719, "dur": 121,
    "args": {
      "External id": 189514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189514,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189514, "pid": 5, "tid": 7, "ts": 1716454224392719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312785, "dur": 6,
    "args": {
      "External id": 189514,
      "cbid": 211, "correlation": 189514
    }
  },
  {
    "ph": "s", "id": 189514, "pid": 76337, "tid": -914061504, "ts": 1716454224312785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224392841, "dur": 25,
    "args": {
      "External id": 189516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189516,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189516, "pid": 5, "tid": 7, "ts": 1716454224392841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312794, "dur": 5,
    "args": {
      "External id": 189516,
      "cbid": 211, "correlation": 189516
    }
  },
  {
    "ph": "s", "id": 189516, "pid": 76337, "tid": -914061504, "ts": 1716454224312794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224392867, "dur": 33,
    "args": {
      "External id": 189522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189522, "pid": 5, "tid": 7, "ts": 1716454224392867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312822, "dur": 8,
    "args": {
      "External id": 189522,
      "cbid": 211, "correlation": 189522
    }
  },
  {
    "ph": "s", "id": 189522, "pid": 76337, "tid": -914061504, "ts": 1716454224312822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224392901, "dur": 27,
    "args": {
      "External id": 189530,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189530,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189530, "pid": 5, "tid": 7, "ts": 1716454224392901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312853, "dur": 9,
    "args": {
      "External id": 189530,
      "cbid": 211, "correlation": 189530
    }
  },
  {
    "ph": "s", "id": 189530, "pid": 76337, "tid": -914061504, "ts": 1716454224312853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224392929, "dur": 102,
    "args": {
      "External id": 189541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189541,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189541, "pid": 5, "tid": 7, "ts": 1716454224392929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312918, "dur": 11,
    "args": {
      "External id": 189541,
      "cbid": 211, "correlation": 189541
    }
  },
  {
    "ph": "s", "id": 189541, "pid": 76337, "tid": -914061504, "ts": 1716454224312918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312972, "dur": 7,
    "args": {
      "External id": 189551,
      "cbid": 317, "correlation": 189551
    }
  },
  {
    "ph": "f", "id": 189551, "pid": 76337, "tid": -914061504, "ts": 1716454224312972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312980, "dur": 0,
    "args": {
      "External id": 189552,
      "cbid": 203, "correlation": 189552
    }
  },
  {
    "ph": "f", "id": 189552, "pid": 76337, "tid": -914061504, "ts": 1716454224312980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312981, "dur": 0,
    "args": {
      "External id": 189553,
      "cbid": 205, "correlation": 189553
    }
  },
  {
    "ph": "f", "id": 189553, "pid": 76337, "tid": -914061504, "ts": 1716454224312981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224393033, "dur": 73,
    "args": {
      "External id": 189557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189557,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189557, "pid": 5, "tid": 7, "ts": 1716454224393033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224312995, "dur": 12,
    "args": {
      "External id": 189557,
      "cbid": 211, "correlation": 189557
    }
  },
  {
    "ph": "s", "id": 189557, "pid": 76337, "tid": -914061504, "ts": 1716454224312995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224393107, "dur": 45,
    "args": {
      "External id": 189559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189559, "pid": 5, "tid": 7, "ts": 1716454224393107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313010, "dur": 6,
    "args": {
      "External id": 189559,
      "cbid": 211, "correlation": 189559
    }
  },
  {
    "ph": "s", "id": 189559, "pid": 76337, "tid": -914061504, "ts": 1716454224313010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224393153, "dur": 4,
    "args": {
      "External id": 189561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189561,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189561, "pid": 5, "tid": 7, "ts": 1716454224393153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313021, "dur": 6,
    "args": {
      "External id": 189561,
      "cbid": 211, "correlation": 189561
    }
  },
  {
    "ph": "s", "id": 189561, "pid": 76337, "tid": -914061504, "ts": 1716454224313021,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313030, "dur": 0,
    "args": {
      "External id": 189562,
      "cbid": 51, "correlation": 189562
    }
  },
  {
    "ph": "s", "id": 189562, "pid": 76337, "tid": -914061504, "ts": 1716454224313030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224393158, "dur": 2245,
    "args": {
      "External id": 189563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189563,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189563, "pid": 5, "tid": 7, "ts": 1716454224393158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313031, "dur": 5,
    "args": {
      "External id": 189563,
      "cbid": 211, "correlation": 189563
    }
  },
  {
    "ph": "s", "id": 189563, "pid": 76337, "tid": -914061504, "ts": 1716454224313031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224395404, "dur": 114,
    "args": {
      "External id": 189568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189568,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189568, "pid": 5, "tid": 7, "ts": 1716454224395404,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313059, "dur": 9,
    "args": {
      "External id": 189568,
      "cbid": 211, "correlation": 189568
    }
  },
  {
    "ph": "s", "id": 189568, "pid": 76337, "tid": -914061504, "ts": 1716454224313059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224395520, "dur": 167,
    "args": {
      "External id": 189577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189577,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189577, "pid": 5, "tid": 7, "ts": 1716454224395520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313152, "dur": 14,
    "args": {
      "External id": 189577,
      "cbid": 211, "correlation": 189577
    }
  },
  {
    "ph": "s", "id": 189577, "pid": 76337, "tid": -914061504, "ts": 1716454224313152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224395688, "dur": 128,
    "args": {
      "External id": 189597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189597,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 189597, "pid": 5, "tid": 7, "ts": 1716454224395688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313224, "dur": 11,
    "args": {
      "External id": 189597,
      "cbid": 211, "correlation": 189597
    }
  },
  {
    "ph": "s", "id": 189597, "pid": 76337, "tid": -914061504, "ts": 1716454224313224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224395817, "dur": 4,
    "args": {
      "External id": 189609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189609,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 189609, "pid": 5, "tid": 7, "ts": 1716454224395817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313245, "dur": 6,
    "args": {
      "External id": 189609,
      "cbid": 211, "correlation": 189609
    }
  },
  {
    "ph": "s", "id": 189609, "pid": 76337, "tid": -914061504, "ts": 1716454224313245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224395823, "dur": 159,
    "args": {
      "External id": 189612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189612, "pid": 5, "tid": 7, "ts": 1716454224395823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313264, "dur": 7,
    "args": {
      "External id": 189612,
      "cbid": 211, "correlation": 189612
    }
  },
  {
    "ph": "s", "id": 189612, "pid": 76337, "tid": -914061504, "ts": 1716454224313264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224395983, "dur": 101,
    "args": {
      "External id": 189621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189621,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189621, "pid": 5, "tid": 7, "ts": 1716454224395983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313304, "dur": 11,
    "args": {
      "External id": 189621,
      "cbid": 211, "correlation": 189621
    }
  },
  {
    "ph": "s", "id": 189621, "pid": 76337, "tid": -914061504, "ts": 1716454224313304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313357, "dur": 0,
    "args": {
      "External id": 189631,
      "cbid": 317, "correlation": 189631
    }
  },
  {
    "ph": "f", "id": 189631, "pid": 76337, "tid": -914061504, "ts": 1716454224313357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313357, "dur": 0,
    "args": {
      "External id": 189632,
      "cbid": 203, "correlation": 189632
    }
  },
  {
    "ph": "f", "id": 189632, "pid": 76337, "tid": -914061504, "ts": 1716454224313357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313358, "dur": 0,
    "args": {
      "External id": 189633,
      "cbid": 205, "correlation": 189633
    }
  },
  {
    "ph": "f", "id": 189633, "pid": 76337, "tid": -914061504, "ts": 1716454224313358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224396086, "dur": 112,
    "args": {
      "External id": 189637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189637,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189637, "pid": 5, "tid": 7, "ts": 1716454224396086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313374, "dur": 11,
    "args": {
      "External id": 189637,
      "cbid": 211, "correlation": 189637
    }
  },
  {
    "ph": "s", "id": 189637, "pid": 76337, "tid": -914061504, "ts": 1716454224313374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224396199, "dur": 34,
    "args": {
      "External id": 189639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189639,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189639, "pid": 5, "tid": 7, "ts": 1716454224396199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313388, "dur": 5,
    "args": {
      "External id": 189639,
      "cbid": 211, "correlation": 189639
    }
  },
  {
    "ph": "s", "id": 189639, "pid": 76337, "tid": -914061504, "ts": 1716454224313388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224396235, "dur": 4,
    "args": {
      "External id": 189641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189641,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189641, "pid": 5, "tid": 7, "ts": 1716454224396235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313398, "dur": 5,
    "args": {
      "External id": 189641,
      "cbid": 211, "correlation": 189641
    }
  },
  {
    "ph": "s", "id": 189641, "pid": 76337, "tid": -914061504, "ts": 1716454224313398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313407, "dur": 0,
    "args": {
      "External id": 189642,
      "cbid": 51, "correlation": 189642
    }
  },
  {
    "ph": "s", "id": 189642, "pid": 76337, "tid": -914061504, "ts": 1716454224313407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224396240, "dur": 2008,
    "args": {
      "External id": 189643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189643,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189643, "pid": 5, "tid": 7, "ts": 1716454224396240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313408, "dur": 7,
    "args": {
      "External id": 189643,
      "cbid": 211, "correlation": 189643
    }
  },
  {
    "ph": "s", "id": 189643, "pid": 76337, "tid": -914061504, "ts": 1716454224313408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224398249, "dur": 59,
    "args": {
      "External id": 189648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189648,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189648, "pid": 5, "tid": 7, "ts": 1716454224398249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313438, "dur": 8,
    "args": {
      "External id": 189648,
      "cbid": 211, "correlation": 189648
    }
  },
  {
    "ph": "s", "id": 189648, "pid": 76337, "tid": -914061504, "ts": 1716454224313438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224398310, "dur": 3,
    "args": {
      "External id": 189656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189656,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189656, "pid": 5, "tid": 7, "ts": 1716454224398310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313481, "dur": 9,
    "args": {
      "External id": 189656,
      "cbid": 211, "correlation": 189656
    }
  },
  {
    "ph": "s", "id": 189656, "pid": 76337, "tid": -914061504, "ts": 1716454224313481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313547, "dur": 1,
    "args": {
      "External id": 189672,
      "cbid": 251, "correlation": 189672
    }
  },
  {
    "ph": "f", "id": 189672, "pid": 76337, "tid": -914061504, "ts": 1716454224313547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313552, "dur": 0,
    "args": {
      "External id": 189674,
      "cbid": 251, "correlation": 189674
    }
  },
  {
    "ph": "f", "id": 189674, "pid": 76337, "tid": -914061504, "ts": 1716454224313552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224398315, "dur": 11,
    "args": {
      "External id": 189675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189675,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 189675, "pid": 5, "tid": 7, "ts": 1716454224398315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313554, "dur": 12,
    "args": {
      "External id": 189675,
      "cbid": 211, "correlation": 189675
    }
  },
  {
    "ph": "s", "id": 189675, "pid": 76337, "tid": -914061504, "ts": 1716454224313554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224398327, "dur": 5,
    "args": {
      "External id": 189677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189677,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 189677, "pid": 5, "tid": 7, "ts": 1716454224398327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313568, "dur": 5,
    "args": {
      "External id": 189677,
      "cbid": 211, "correlation": 189677
    }
  },
  {
    "ph": "s", "id": 189677, "pid": 76337, "tid": -914061504, "ts": 1716454224313568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224398334, "dur": 55,
    "args": {
      "External id": 189687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189687,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189687, "pid": 5, "tid": 7, "ts": 1716454224398334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313625, "dur": 12,
    "args": {
      "External id": 189687,
      "cbid": 211, "correlation": 189687
    }
  },
  {
    "ph": "s", "id": 189687, "pid": 76337, "tid": -914061504, "ts": 1716454224313625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224398390, "dur": 51,
    "args": {
      "External id": 189707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189707,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 189707, "pid": 5, "tid": 7, "ts": 1716454224398390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313692, "dur": 11,
    "args": {
      "External id": 189707,
      "cbid": 211, "correlation": 189707
    }
  },
  {
    "ph": "s", "id": 189707, "pid": 76337, "tid": -914061504, "ts": 1716454224313692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224398442, "dur": 4,
    "args": {
      "External id": 189719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189719,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189719, "pid": 5, "tid": 7, "ts": 1716454224398442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313712, "dur": 7,
    "args": {
      "External id": 189719,
      "cbid": 211, "correlation": 189719
    }
  },
  {
    "ph": "s", "id": 189719, "pid": 76337, "tid": -914061504, "ts": 1716454224313712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224398447, "dur": 56,
    "args": {
      "External id": 189722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189722,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189722, "pid": 5, "tid": 7, "ts": 1716454224398447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313732, "dur": 6,
    "args": {
      "External id": 189722,
      "cbid": 211, "correlation": 189722
    }
  },
  {
    "ph": "s", "id": 189722, "pid": 76337, "tid": -914061504, "ts": 1716454224313732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224398504, "dur": 37,
    "args": {
      "External id": 189731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189731,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189731, "pid": 5, "tid": 7, "ts": 1716454224398504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313773, "dur": 9,
    "args": {
      "External id": 189731,
      "cbid": 211, "correlation": 189731
    }
  },
  {
    "ph": "s", "id": 189731, "pid": 76337, "tid": -914061504, "ts": 1716454224313773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313835, "dur": 0,
    "args": {
      "External id": 189741,
      "cbid": 317, "correlation": 189741
    }
  },
  {
    "ph": "f", "id": 189741, "pid": 76337, "tid": -914061504, "ts": 1716454224313835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313836, "dur": 0,
    "args": {
      "External id": 189742,
      "cbid": 203, "correlation": 189742
    }
  },
  {
    "ph": "f", "id": 189742, "pid": 76337, "tid": -914061504, "ts": 1716454224313836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313837, "dur": 0,
    "args": {
      "External id": 189743,
      "cbid": 205, "correlation": 189743
    }
  },
  {
    "ph": "f", "id": 189743, "pid": 76337, "tid": -914061504, "ts": 1716454224313837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224398543, "dur": 40,
    "args": {
      "External id": 189747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189747,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189747, "pid": 5, "tid": 7, "ts": 1716454224398543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313855, "dur": 12,
    "args": {
      "External id": 189747,
      "cbid": 211, "correlation": 189747
    }
  },
  {
    "ph": "s", "id": 189747, "pid": 76337, "tid": -914061504, "ts": 1716454224313855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224398584, "dur": 14,
    "args": {
      "External id": 189749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189749,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189749, "pid": 5, "tid": 7, "ts": 1716454224398584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313870, "dur": 6,
    "args": {
      "External id": 189749,
      "cbid": 211, "correlation": 189749
    }
  },
  {
    "ph": "s", "id": 189749, "pid": 76337, "tid": -914061504, "ts": 1716454224313870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224398600, "dur": 4,
    "args": {
      "External id": 189751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189751,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189751, "pid": 5, "tid": 7, "ts": 1716454224398600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313880, "dur": 5,
    "args": {
      "External id": 189751,
      "cbid": 211, "correlation": 189751
    }
  },
  {
    "ph": "s", "id": 189751, "pid": 76337, "tid": -914061504, "ts": 1716454224313880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313888, "dur": 0,
    "args": {
      "External id": 189752,
      "cbid": 51, "correlation": 189752
    }
  },
  {
    "ph": "s", "id": 189752, "pid": 76337, "tid": -914061504, "ts": 1716454224313888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224398604, "dur": 697,
    "args": {
      "External id": 189753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189753,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189753, "pid": 5, "tid": 7, "ts": 1716454224398604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313889, "dur": 5,
    "args": {
      "External id": 189753,
      "cbid": 211, "correlation": 189753
    }
  },
  {
    "ph": "s", "id": 189753, "pid": 76337, "tid": -914061504, "ts": 1716454224313889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224399303, "dur": 60,
    "args": {
      "External id": 189758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189758,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189758, "pid": 5, "tid": 7, "ts": 1716454224399303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313916, "dur": 9,
    "args": {
      "External id": 189758,
      "cbid": 211, "correlation": 189758
    }
  },
  {
    "ph": "s", "id": 189758, "pid": 76337, "tid": -914061504, "ts": 1716454224313916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313981, "dur": 0,
    "args": {
      "External id": 189768,
      "cbid": 317, "correlation": 189768
    }
  },
  {
    "ph": "f", "id": 189768, "pid": 76337, "tid": -914061504, "ts": 1716454224313981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313982, "dur": 0,
    "args": {
      "External id": 189769,
      "cbid": 203, "correlation": 189769
    }
  },
  {
    "ph": "f", "id": 189769, "pid": 76337, "tid": -914061504, "ts": 1716454224313982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313983, "dur": 0,
    "args": {
      "External id": 189770,
      "cbid": 205, "correlation": 189770
    }
  },
  {
    "ph": "f", "id": 189770, "pid": 76337, "tid": -914061504, "ts": 1716454224313983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224399364, "dur": 3,
    "args": {
      "External id": 189774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189774, "pid": 5, "tid": 7, "ts": 1716454224399364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224313998, "dur": 12,
    "args": {
      "External id": 189774,
      "cbid": 211, "correlation": 189774
    }
  },
  {
    "ph": "s", "id": 189774, "pid": 76337, "tid": -914061504, "ts": 1716454224313998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314014, "dur": 0,
    "args": {
      "External id": 189775,
      "cbid": 51, "correlation": 189775
    }
  },
  {
    "ph": "s", "id": 189775, "pid": 76337, "tid": -914061504, "ts": 1716454224314014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224399369, "dur": 265,
    "args": {
      "External id": 189776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189776,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189776, "pid": 5, "tid": 7, "ts": 1716454224399369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314015, "dur": 8,
    "args": {
      "External id": 189776,
      "cbid": 211, "correlation": 189776
    }
  },
  {
    "ph": "s", "id": 189776, "pid": 76337, "tid": -914061504, "ts": 1716454224314015,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224399635, "dur": 59,
    "args": {
      "External id": 189781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189781,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189781, "pid": 5, "tid": 7, "ts": 1716454224399635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314044, "dur": 8,
    "args": {
      "External id": 189781,
      "cbid": 211, "correlation": 189781
    }
  },
  {
    "ph": "s", "id": 189781, "pid": 76337, "tid": -914061504, "ts": 1716454224314044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224399695, "dur": 51,
    "args": {
      "External id": 189789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189789, "pid": 5, "tid": 7, "ts": 1716454224399695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314074, "dur": 8,
    "args": {
      "External id": 189789,
      "cbid": 211, "correlation": 189789
    }
  },
  {
    "ph": "s", "id": 189789, "pid": 76337, "tid": -914061504, "ts": 1716454224314074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224399747, "dur": 35,
    "args": {
      "External id": 189797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189797,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189797, "pid": 5, "tid": 7, "ts": 1716454224399747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314102, "dur": 8,
    "args": {
      "External id": 189797,
      "cbid": 211, "correlation": 189797
    }
  },
  {
    "ph": "s", "id": 189797, "pid": 76337, "tid": -914061504, "ts": 1716454224314102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224399783, "dur": 51,
    "args": {
      "External id": 189817,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189817,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 189817, "pid": 5, "tid": 7, "ts": 1716454224399783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314184, "dur": 12,
    "args": {
      "External id": 189817,
      "cbid": 211, "correlation": 189817
    }
  },
  {
    "ph": "s", "id": 189817, "pid": 76337, "tid": -914061504, "ts": 1716454224314184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224399836, "dur": 4,
    "args": {
      "External id": 189829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189829,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 189829, "pid": 5, "tid": 7, "ts": 1716454224399836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314206, "dur": 6,
    "args": {
      "External id": 189829,
      "cbid": 211, "correlation": 189829
    }
  },
  {
    "ph": "s", "id": 189829, "pid": 76337, "tid": -914061504, "ts": 1716454224314206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224399841, "dur": 55,
    "args": {
      "External id": 189832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189832,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189832, "pid": 5, "tid": 7, "ts": 1716454224399841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314224, "dur": 6,
    "args": {
      "External id": 189832,
      "cbid": 211, "correlation": 189832
    }
  },
  {
    "ph": "s", "id": 189832, "pid": 76337, "tid": -914061504, "ts": 1716454224314224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314280, "dur": 0,
    "args": {
      "External id": 189843,
      "cbid": 317, "correlation": 189843
    }
  },
  {
    "ph": "f", "id": 189843, "pid": 76337, "tid": -914061504, "ts": 1716454224314280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314281, "dur": 0,
    "args": {
      "External id": 189844,
      "cbid": 203, "correlation": 189844
    }
  },
  {
    "ph": "f", "id": 189844, "pid": 76337, "tid": -914061504, "ts": 1716454224314281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314282, "dur": 0,
    "args": {
      "External id": 189845,
      "cbid": 205, "correlation": 189845
    }
  },
  {
    "ph": "f", "id": 189845, "pid": 76337, "tid": -914061504, "ts": 1716454224314282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314313, "dur": 2,
    "args": {
      "External id": 189849,
      "cbid": 251, "correlation": 189849
    }
  },
  {
    "ph": "f", "id": 189849, "pid": 76337, "tid": -914061504, "ts": 1716454224314313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314317, "dur": 1,
    "args": {
      "External id": 189850,
      "cbid": 251, "correlation": 189850
    }
  },
  {
    "ph": "f", "id": 189850, "pid": 76337, "tid": -914061504, "ts": 1716454224314317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314318, "dur": 1,
    "args": {
      "External id": 189851,
      "cbid": 251, "correlation": 189851
    }
  },
  {
    "ph": "f", "id": 189851, "pid": 76337, "tid": -914061504, "ts": 1716454224314318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314320, "dur": 1,
    "args": {
      "External id": 189852,
      "cbid": 251, "correlation": 189852
    }
  },
  {
    "ph": "f", "id": 189852, "pid": 76337, "tid": -914061504, "ts": 1716454224314320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314322, "dur": 1,
    "args": {
      "External id": 189853,
      "cbid": 251, "correlation": 189853
    }
  },
  {
    "ph": "f", "id": 189853, "pid": 76337, "tid": -914061504, "ts": 1716454224314322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314324, "dur": 2,
    "args": {
      "External id": 189854,
      "cbid": 251, "correlation": 189854
    }
  },
  {
    "ph": "f", "id": 189854, "pid": 76337, "tid": -914061504, "ts": 1716454224314324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314326, "dur": 1,
    "args": {
      "External id": 189855,
      "cbid": 251, "correlation": 189855
    }
  },
  {
    "ph": "f", "id": 189855, "pid": 76337, "tid": -914061504, "ts": 1716454224314326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314328, "dur": 1,
    "args": {
      "External id": 189856,
      "cbid": 251, "correlation": 189856
    }
  },
  {
    "ph": "f", "id": 189856, "pid": 76337, "tid": -914061504, "ts": 1716454224314328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314331, "dur": 0,
    "args": {
      "External id": 189857,
      "cbid": 251, "correlation": 189857
    }
  },
  {
    "ph": "f", "id": 189857, "pid": 76337, "tid": -914061504, "ts": 1716454224314331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224399897, "dur": 116,
    "args": {
      "External id": 189858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189858,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 189858, "pid": 5, "tid": 7, "ts": 1716454224399897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314334, "dur": 13,
    "args": {
      "External id": 189858,
      "cbid": 211, "correlation": 189858
    }
  },
  {
    "ph": "s", "id": 189858, "pid": 76337, "tid": -914061504, "ts": 1716454224314334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224400015, "dur": 60,
    "args": {
      "External id": 189864,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189864,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189864, "pid": 5, "tid": 7, "ts": 1716454224400015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314370, "dur": 9,
    "args": {
      "External id": 189864,
      "cbid": 211, "correlation": 189864
    }
  },
  {
    "ph": "s", "id": 189864, "pid": 76337, "tid": -914061504, "ts": 1716454224314370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224400076, "dur": 609,
    "args": {
      "External id": 189873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189873,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189873, "pid": 5, "tid": 7, "ts": 1716454224400076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314451, "dur": 14,
    "args": {
      "External id": 189873,
      "cbid": 211, "correlation": 189873
    }
  },
  {
    "ph": "s", "id": 189873, "pid": 76337, "tid": -914061504, "ts": 1716454224314451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224400687, "dur": 181,
    "args": {
      "External id": 189895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189895,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189895, "pid": 5, "tid": 7, "ts": 1716454224400687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314509, "dur": 11,
    "args": {
      "External id": 189895,
      "cbid": 211, "correlation": 189895
    }
  },
  {
    "ph": "s", "id": 189895, "pid": 76337, "tid": -914061504, "ts": 1716454224314509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314597, "dur": 1,
    "args": {
      "External id": 189906,
      "cbid": 251, "correlation": 189906
    }
  },
  {
    "ph": "f", "id": 189906, "pid": 76337, "tid": -914061504, "ts": 1716454224314597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224400869, "dur": 197,
    "args": {
      "External id": 189907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189907,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189907, "pid": 5, "tid": 7, "ts": 1716454224400869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314602, "dur": 14,
    "args": {
      "External id": 189907,
      "cbid": 211, "correlation": 189907
    }
  },
  {
    "ph": "s", "id": 189907, "pid": 76337, "tid": -914061504, "ts": 1716454224314602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314672, "dur": 1,
    "args": {
      "External id": 189918,
      "cbid": 251, "correlation": 189918
    }
  },
  {
    "ph": "f", "id": 189918, "pid": 76337, "tid": -914061504, "ts": 1716454224314672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224401068, "dur": 187,
    "args": {
      "External id": 189919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189919,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189919, "pid": 5, "tid": 7, "ts": 1716454224401068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314676, "dur": 12,
    "args": {
      "External id": 189919,
      "cbid": 211, "correlation": 189919
    }
  },
  {
    "ph": "s", "id": 189919, "pid": 76337, "tid": -914061504, "ts": 1716454224314676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314740, "dur": 1,
    "args": {
      "External id": 189930,
      "cbid": 251, "correlation": 189930
    }
  },
  {
    "ph": "f", "id": 189930, "pid": 76337, "tid": -914061504, "ts": 1716454224314740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224401256, "dur": 185,
    "args": {
      "External id": 189931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189931,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189931, "pid": 5, "tid": 7, "ts": 1716454224401256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314744, "dur": 12,
    "args": {
      "External id": 189931,
      "cbid": 211, "correlation": 189931
    }
  },
  {
    "ph": "s", "id": 189931, "pid": 76337, "tid": -914061504, "ts": 1716454224314744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224401443, "dur": 18686,
    "args": {
      "External id": 189952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189952,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 189952, "pid": 5, "tid": 7, "ts": 1716454224401443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314828, "dur": 13,
    "args": {
      "External id": 189952,
      "cbid": 211, "correlation": 189952
    }
  },
  {
    "ph": "s", "id": 189952, "pid": 76337, "tid": -914061504, "ts": 1716454224314828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314925, "dur": 1,
    "args": {
      "External id": 189970,
      "cbid": 251, "correlation": 189970
    }
  },
  {
    "ph": "f", "id": 189970, "pid": 76337, "tid": -914061504, "ts": 1716454224314925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224420131, "dur": 203,
    "args": {
      "External id": 189972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189972,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 189972, "pid": 5, "tid": 7, "ts": 1716454224420131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224314930, "dur": 13,
    "args": {
      "External id": 189972,
      "cbid": 211, "correlation": 189972
    }
  },
  {
    "ph": "s", "id": 189972, "pid": 76337, "tid": -914061504, "ts": 1716454224314930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224420335, "dur": 67,
    "args": {
      "External id": 189980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189980,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189980, "pid": 5, "tid": 7, "ts": 1716454224420335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315008, "dur": 14,
    "args": {
      "External id": 189980,
      "cbid": 211, "correlation": 189980
    }
  },
  {
    "ph": "s", "id": 189980, "pid": 76337, "tid": -914061504, "ts": 1716454224315008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224420403, "dur": 96,
    "args": {
      "External id": 189988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189988,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189988, "pid": 5, "tid": 7, "ts": 1716454224420403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315048, "dur": 9,
    "args": {
      "External id": 189988,
      "cbid": 211, "correlation": 189988
    }
  },
  {
    "ph": "s", "id": 189988, "pid": 76337, "tid": -914061504, "ts": 1716454224315048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224420500, "dur": 53,
    "args": {
      "External id": 189999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 189999,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 189999, "pid": 5, "tid": 7, "ts": 1716454224420500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315122, "dur": 12,
    "args": {
      "External id": 189999,
      "cbid": 211, "correlation": 189999
    }
  },
  {
    "ph": "s", "id": 189999, "pid": 76337, "tid": -914061504, "ts": 1716454224315122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224420555, "dur": 93,
    "args": {
      "External id": 190021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190021,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190021, "pid": 5, "tid": 7, "ts": 1716454224420555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315153, "dur": 8,
    "args": {
      "External id": 190021,
      "cbid": 211, "correlation": 190021
    }
  },
  {
    "ph": "s", "id": 190021, "pid": 76337, "tid": -914061504, "ts": 1716454224315153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315237, "dur": 1,
    "args": {
      "External id": 190032,
      "cbid": 251, "correlation": 190032
    }
  },
  {
    "ph": "f", "id": 190032, "pid": 76337, "tid": -914061504, "ts": 1716454224315237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224420649, "dur": 106,
    "args": {
      "External id": 190033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190033,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190033, "pid": 5, "tid": 7, "ts": 1716454224420649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315242, "dur": 13,
    "args": {
      "External id": 190033,
      "cbid": 211, "correlation": 190033
    }
  },
  {
    "ph": "s", "id": 190033, "pid": 76337, "tid": -914061504, "ts": 1716454224315242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315316, "dur": 1,
    "args": {
      "External id": 190044,
      "cbid": 251, "correlation": 190044
    }
  },
  {
    "ph": "f", "id": 190044, "pid": 76337, "tid": -914061504, "ts": 1716454224315316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315320, "dur": 0,
    "args": {
      "External id": 190045,
      "cbid": 251, "correlation": 190045
    }
  },
  {
    "ph": "f", "id": 190045, "pid": 76337, "tid": -914061504, "ts": 1716454224315320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224420756, "dur": 10,
    "args": {
      "External id": 190046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190046,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 190046, "pid": 5, "tid": 7, "ts": 1716454224420756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315322, "dur": 14,
    "args": {
      "External id": 190046,
      "cbid": 211, "correlation": 190046
    }
  },
  {
    "ph": "s", "id": 190046, "pid": 76337, "tid": -914061504, "ts": 1716454224315322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224420767, "dur": 5,
    "args": {
      "External id": 190048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190048,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190048, "pid": 5, "tid": 7, "ts": 1716454224420767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315340, "dur": 7,
    "args": {
      "External id": 190048,
      "cbid": 211, "correlation": 190048
    }
  },
  {
    "ph": "s", "id": 190048, "pid": 76337, "tid": -914061504, "ts": 1716454224315340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315402, "dur": 1,
    "args": {
      "External id": 190059,
      "cbid": 251, "correlation": 190059
    }
  },
  {
    "ph": "f", "id": 190059, "pid": 76337, "tid": -914061504, "ts": 1716454224315402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315405, "dur": 0,
    "args": {
      "External id": 190060,
      "cbid": 251, "correlation": 190060
    }
  },
  {
    "ph": "f", "id": 190060, "pid": 76337, "tid": -914061504, "ts": 1716454224315405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224420774, "dur": 6,
    "args": {
      "External id": 190061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190061,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 190061, "pid": 5, "tid": 7, "ts": 1716454224420774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315407, "dur": 12,
    "args": {
      "External id": 190061,
      "cbid": 211, "correlation": 190061
    }
  },
  {
    "ph": "s", "id": 190061, "pid": 76337, "tid": -914061504, "ts": 1716454224315407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224420781, "dur": 3,
    "args": {
      "External id": 190063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190063,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190063, "pid": 5, "tid": 7, "ts": 1716454224420781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315421, "dur": 5,
    "args": {
      "External id": 190063,
      "cbid": 211, "correlation": 190063
    }
  },
  {
    "ph": "s", "id": 190063, "pid": 76337, "tid": -914061504, "ts": 1716454224315421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224420786, "dur": 157,
    "args": {
      "External id": 190084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190084,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190084, "pid": 5, "tid": 7, "ts": 1716454224420786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315495, "dur": 13,
    "args": {
      "External id": 190084,
      "cbid": 211, "correlation": 190084
    }
  },
  {
    "ph": "s", "id": 190084, "pid": 76337, "tid": -914061504, "ts": 1716454224315495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315591, "dur": 1,
    "args": {
      "External id": 190102,
      "cbid": 251, "correlation": 190102
    }
  },
  {
    "ph": "f", "id": 190102, "pid": 76337, "tid": -914061504, "ts": 1716454224315591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224420944, "dur": 108,
    "args": {
      "External id": 190104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190104,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190104, "pid": 5, "tid": 7, "ts": 1716454224420944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315597, "dur": 14,
    "args": {
      "External id": 190104,
      "cbid": 211, "correlation": 190104
    }
  },
  {
    "ph": "s", "id": 190104, "pid": 76337, "tid": -914061504, "ts": 1716454224315597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224421054, "dur": 35,
    "args": {
      "External id": 190112,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190112,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190112, "pid": 5, "tid": 7, "ts": 1716454224421054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315667, "dur": 12,
    "args": {
      "External id": 190112,
      "cbid": 211, "correlation": 190112
    }
  },
  {
    "ph": "s", "id": 190112, "pid": 76337, "tid": -914061504, "ts": 1716454224315667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224421090, "dur": 67,
    "args": {
      "External id": 190120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190120,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190120, "pid": 5, "tid": 7, "ts": 1716454224421090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315708, "dur": 10,
    "args": {
      "External id": 190120,
      "cbid": 211, "correlation": 190120
    }
  },
  {
    "ph": "s", "id": 190120, "pid": 76337, "tid": -914061504, "ts": 1716454224315708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224421158, "dur": 93,
    "args": {
      "External id": 190142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190142,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190142, "pid": 5, "tid": 7, "ts": 1716454224421158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315759, "dur": 10,
    "args": {
      "External id": 190142,
      "cbid": 211, "correlation": 190142
    }
  },
  {
    "ph": "s", "id": 190142, "pid": 76337, "tid": -914061504, "ts": 1716454224315759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315845, "dur": 1,
    "args": {
      "External id": 190158,
      "cbid": 251, "correlation": 190158
    }
  },
  {
    "ph": "f", "id": 190158, "pid": 76337, "tid": -914061504, "ts": 1716454224315845,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224421253, "dur": 583,
    "args": {
      "External id": 190160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190160,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190160, "pid": 5, "tid": 7, "ts": 1716454224421253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315851, "dur": 13,
    "args": {
      "External id": 190160,
      "cbid": 211, "correlation": 190160
    }
  },
  {
    "ph": "s", "id": 190160, "pid": 76337, "tid": -914061504, "ts": 1716454224315851,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224421837, "dur": 245,
    "args": {
      "External id": 190168,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190168,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190168, "pid": 5, "tid": 7, "ts": 1716454224421837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315916, "dur": 13,
    "args": {
      "External id": 190168,
      "cbid": 211, "correlation": 190168
    }
  },
  {
    "ph": "s", "id": 190168, "pid": 76337, "tid": -914061504, "ts": 1716454224315916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224422084, "dur": 252,
    "args": {
      "External id": 190176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190176,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190176, "pid": 5, "tid": 7, "ts": 1716454224422084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224315948, "dur": 8,
    "args": {
      "External id": 190176,
      "cbid": 211, "correlation": 190176
    }
  },
  {
    "ph": "s", "id": 190176, "pid": 76337, "tid": -914061504, "ts": 1716454224315948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316041, "dur": 1,
    "args": {
      "External id": 190192,
      "cbid": 251, "correlation": 190192
    }
  },
  {
    "ph": "f", "id": 190192, "pid": 76337, "tid": -914061504, "ts": 1716454224316041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316046, "dur": 0,
    "args": {
      "External id": 190194,
      "cbid": 251, "correlation": 190194
    }
  },
  {
    "ph": "f", "id": 190194, "pid": 76337, "tid": -914061504, "ts": 1716454224316046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224422337, "dur": 359,
    "args": {
      "External id": 190195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190195,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 190195, "pid": 5, "tid": 7, "ts": 1716454224422337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316049, "dur": 14,
    "args": {
      "External id": 190195,
      "cbid": 211, "correlation": 190195
    }
  },
  {
    "ph": "s", "id": 190195, "pid": 76337, "tid": -914061504, "ts": 1716454224316049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224422697, "dur": 50,
    "args": {
      "External id": 190203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190203,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190203, "pid": 5, "tid": 7, "ts": 1716454224422697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316092, "dur": 10,
    "args": {
      "External id": 190203,
      "cbid": 211, "correlation": 190203
    }
  },
  {
    "ph": "s", "id": 190203, "pid": 76337, "tid": -914061504, "ts": 1716454224316092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224422748, "dur": 159,
    "args": {
      "External id": 190214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190214, "pid": 5, "tid": 7, "ts": 1716454224422748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316160, "dur": 13,
    "args": {
      "External id": 190214,
      "cbid": 211, "correlation": 190214
    }
  },
  {
    "ph": "s", "id": 190214, "pid": 76337, "tid": -914061504, "ts": 1716454224316160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316224, "dur": 0,
    "args": {
      "External id": 190226,
      "cbid": 317, "correlation": 190226
    }
  },
  {
    "ph": "f", "id": 190226, "pid": 76337, "tid": -914061504, "ts": 1716454224316224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316225, "dur": 0,
    "args": {
      "External id": 190227,
      "cbid": 203, "correlation": 190227
    }
  },
  {
    "ph": "f", "id": 190227, "pid": 76337, "tid": -914061504, "ts": 1716454224316225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316226, "dur": 0,
    "args": {
      "External id": 190228,
      "cbid": 205, "correlation": 190228
    }
  },
  {
    "ph": "f", "id": 190228, "pid": 76337, "tid": -914061504, "ts": 1716454224316226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316250, "dur": 1,
    "args": {
      "External id": 190232,
      "cbid": 251, "correlation": 190232
    }
  },
  {
    "ph": "f", "id": 190232, "pid": 76337, "tid": -914061504, "ts": 1716454224316250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316252, "dur": 0,
    "args": {
      "External id": 190233,
      "cbid": 251, "correlation": 190233
    }
  },
  {
    "ph": "f", "id": 190233, "pid": 76337, "tid": -914061504, "ts": 1716454224316252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316252, "dur": 0,
    "args": {
      "External id": 190234,
      "cbid": 251, "correlation": 190234
    }
  },
  {
    "ph": "f", "id": 190234, "pid": 76337, "tid": -914061504, "ts": 1716454224316252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316253, "dur": 0,
    "args": {
      "External id": 190235,
      "cbid": 251, "correlation": 190235
    }
  },
  {
    "ph": "f", "id": 190235, "pid": 76337, "tid": -914061504, "ts": 1716454224316253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316254, "dur": 0,
    "args": {
      "External id": 190236,
      "cbid": 251, "correlation": 190236
    }
  },
  {
    "ph": "f", "id": 190236, "pid": 76337, "tid": -914061504, "ts": 1716454224316254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316255, "dur": 0,
    "args": {
      "External id": 190237,
      "cbid": 251, "correlation": 190237
    }
  },
  {
    "ph": "f", "id": 190237, "pid": 76337, "tid": -914061504, "ts": 1716454224316255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316256, "dur": 0,
    "args": {
      "External id": 190238,
      "cbid": 251, "correlation": 190238
    }
  },
  {
    "ph": "f", "id": 190238, "pid": 76337, "tid": -914061504, "ts": 1716454224316256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316257, "dur": 0,
    "args": {
      "External id": 190239,
      "cbid": 251, "correlation": 190239
    }
  },
  {
    "ph": "f", "id": 190239, "pid": 76337, "tid": -914061504, "ts": 1716454224316257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316258, "dur": 0,
    "args": {
      "External id": 190240,
      "cbid": 251, "correlation": 190240
    }
  },
  {
    "ph": "f", "id": 190240, "pid": 76337, "tid": -914061504, "ts": 1716454224316258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224422909, "dur": 115,
    "args": {
      "External id": 190241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190241,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190241, "pid": 5, "tid": 7, "ts": 1716454224422909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316261, "dur": 12,
    "args": {
      "External id": 190241,
      "cbid": 211, "correlation": 190241
    }
  },
  {
    "ph": "s", "id": 190241, "pid": 76337, "tid": -914061504, "ts": 1716454224316261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224423025, "dur": 59,
    "args": {
      "External id": 190247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190247,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190247, "pid": 5, "tid": 7, "ts": 1716454224423025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316296, "dur": 9,
    "args": {
      "External id": 190247,
      "cbid": 211, "correlation": 190247
    }
  },
  {
    "ph": "s", "id": 190247, "pid": 76337, "tid": -914061504, "ts": 1716454224316296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224423086, "dur": 50,
    "args": {
      "External id": 190255,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190255,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190255, "pid": 5, "tid": 7, "ts": 1716454224423086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316327, "dur": 8,
    "args": {
      "External id": 190255,
      "cbid": 211, "correlation": 190255
    }
  },
  {
    "ph": "s", "id": 190255, "pid": 76337, "tid": -914061504, "ts": 1716454224316327,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224423137, "dur": 99,
    "args": {
      "External id": 190264,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190264,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190264, "pid": 5, "tid": 7, "ts": 1716454224423137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316367, "dur": 10,
    "args": {
      "External id": 190264,
      "cbid": 211, "correlation": 190264
    }
  },
  {
    "ph": "s", "id": 190264, "pid": 76337, "tid": -914061504, "ts": 1716454224316367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224423237, "dur": 92,
    "args": {
      "External id": 190284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190284,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 190284, "pid": 5, "tid": 7, "ts": 1716454224423237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316439, "dur": 11,
    "args": {
      "External id": 190284,
      "cbid": 211, "correlation": 190284
    }
  },
  {
    "ph": "s", "id": 190284, "pid": 76337, "tid": -914061504, "ts": 1716454224316439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224423331, "dur": 5,
    "args": {
      "External id": 190296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190296,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190296, "pid": 5, "tid": 7, "ts": 1716454224423331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316460, "dur": 8,
    "args": {
      "External id": 190296,
      "cbid": 211, "correlation": 190296
    }
  },
  {
    "ph": "s", "id": 190296, "pid": 76337, "tid": -914061504, "ts": 1716454224316460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224423336, "dur": 108,
    "args": {
      "External id": 190299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190299,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190299, "pid": 5, "tid": 7, "ts": 1716454224423336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316480, "dur": 6,
    "args": {
      "External id": 190299,
      "cbid": 211, "correlation": 190299
    }
  },
  {
    "ph": "s", "id": 190299, "pid": 76337, "tid": -914061504, "ts": 1716454224316480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224423446, "dur": 70,
    "args": {
      "External id": 190308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190308,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190308, "pid": 5, "tid": 7, "ts": 1716454224423446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316520, "dur": 10,
    "args": {
      "External id": 190308,
      "cbid": 211, "correlation": 190308
    }
  },
  {
    "ph": "s", "id": 190308, "pid": 76337, "tid": -914061504, "ts": 1716454224316520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316572, "dur": 0,
    "args": {
      "External id": 190318,
      "cbid": 317, "correlation": 190318
    }
  },
  {
    "ph": "f", "id": 190318, "pid": 76337, "tid": -914061504, "ts": 1716454224316572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316573, "dur": 0,
    "args": {
      "External id": 190319,
      "cbid": 203, "correlation": 190319
    }
  },
  {
    "ph": "f", "id": 190319, "pid": 76337, "tid": -914061504, "ts": 1716454224316573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316574, "dur": 0,
    "args": {
      "External id": 190320,
      "cbid": 205, "correlation": 190320
    }
  },
  {
    "ph": "f", "id": 190320, "pid": 76337, "tid": -914061504, "ts": 1716454224316574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224423517, "dur": 75,
    "args": {
      "External id": 190324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190324,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190324, "pid": 5, "tid": 7, "ts": 1716454224423517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316589, "dur": 11,
    "args": {
      "External id": 190324,
      "cbid": 211, "correlation": 190324
    }
  },
  {
    "ph": "s", "id": 190324, "pid": 76337, "tid": -914061504, "ts": 1716454224316589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224423594, "dur": 24,
    "args": {
      "External id": 190326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190326,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190326, "pid": 5, "tid": 7, "ts": 1716454224423594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316603, "dur": 5,
    "args": {
      "External id": 190326,
      "cbid": 211, "correlation": 190326
    }
  },
  {
    "ph": "s", "id": 190326, "pid": 76337, "tid": -914061504, "ts": 1716454224316603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224423620, "dur": 4,
    "args": {
      "External id": 190328,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190328,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 190328, "pid": 5, "tid": 7, "ts": 1716454224423620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316613, "dur": 6,
    "args": {
      "External id": 190328,
      "cbid": 211, "correlation": 190328
    }
  },
  {
    "ph": "s", "id": 190328, "pid": 76337, "tid": -914061504, "ts": 1716454224316613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316622, "dur": 0,
    "args": {
      "External id": 190329,
      "cbid": 51, "correlation": 190329
    }
  },
  {
    "ph": "s", "id": 190329, "pid": 76337, "tid": -914061504, "ts": 1716454224316622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224423625, "dur": 1373,
    "args": {
      "External id": 190330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190330,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190330, "pid": 5, "tid": 7, "ts": 1716454224423625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316623, "dur": 5,
    "args": {
      "External id": 190330,
      "cbid": 211, "correlation": 190330
    }
  },
  {
    "ph": "s", "id": 190330, "pid": 76337, "tid": -914061504, "ts": 1716454224316623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224425000, "dur": 59,
    "args": {
      "External id": 190335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190335,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190335, "pid": 5, "tid": 7, "ts": 1716454224425000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316650, "dur": 8,
    "args": {
      "External id": 190335,
      "cbid": 211, "correlation": 190335
    }
  },
  {
    "ph": "s", "id": 190335, "pid": 76337, "tid": -914061504, "ts": 1716454224316650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224425061, "dur": 3,
    "args": {
      "External id": 190343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190343,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 190343, "pid": 5, "tid": 7, "ts": 1716454224425061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316694, "dur": 10,
    "args": {
      "External id": 190343,
      "cbid": 211, "correlation": 190343
    }
  },
  {
    "ph": "s", "id": 190343, "pid": 76337, "tid": -914061504, "ts": 1716454224316694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316760, "dur": 2,
    "args": {
      "External id": 190359,
      "cbid": 251, "correlation": 190359
    }
  },
  {
    "ph": "f", "id": 190359, "pid": 76337, "tid": -914061504, "ts": 1716454224316760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316766, "dur": 0,
    "args": {
      "External id": 190361,
      "cbid": 251, "correlation": 190361
    }
  },
  {
    "ph": "f", "id": 190361, "pid": 76337, "tid": -914061504, "ts": 1716454224316766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224425065, "dur": 12,
    "args": {
      "External id": 190362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190362,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 190362, "pid": 5, "tid": 7, "ts": 1716454224425065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316768, "dur": 11,
    "args": {
      "External id": 190362,
      "cbid": 211, "correlation": 190362
    }
  },
  {
    "ph": "s", "id": 190362, "pid": 76337, "tid": -914061504, "ts": 1716454224316768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224425078, "dur": 5,
    "args": {
      "External id": 190364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190364,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190364, "pid": 5, "tid": 7, "ts": 1716454224425078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316781, "dur": 6,
    "args": {
      "External id": 190364,
      "cbid": 211, "correlation": 190364
    }
  },
  {
    "ph": "s", "id": 190364, "pid": 76337, "tid": -914061504, "ts": 1716454224316781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224425085, "dur": 54,
    "args": {
      "External id": 190374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190374,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190374, "pid": 5, "tid": 7, "ts": 1716454224425085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316840, "dur": 12,
    "args": {
      "External id": 190374,
      "cbid": 211, "correlation": 190374
    }
  },
  {
    "ph": "s", "id": 190374, "pid": 76337, "tid": -914061504, "ts": 1716454224316840,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224425140, "dur": 51,
    "args": {
      "External id": 190394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190394,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 190394, "pid": 5, "tid": 7, "ts": 1716454224425140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316906, "dur": 12,
    "args": {
      "External id": 190394,
      "cbid": 211, "correlation": 190394
    }
  },
  {
    "ph": "s", "id": 190394, "pid": 76337, "tid": -914061504, "ts": 1716454224316906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224425193, "dur": 4,
    "args": {
      "External id": 190406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190406,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 190406, "pid": 5, "tid": 7, "ts": 1716454224425193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316927, "dur": 6,
    "args": {
      "External id": 190406,
      "cbid": 211, "correlation": 190406
    }
  },
  {
    "ph": "s", "id": 190406, "pid": 76337, "tid": -914061504, "ts": 1716454224316927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224425198, "dur": 56,
    "args": {
      "External id": 190409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190409,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190409, "pid": 5, "tid": 7, "ts": 1716454224425198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316946, "dur": 6,
    "args": {
      "External id": 190409,
      "cbid": 211, "correlation": 190409
    }
  },
  {
    "ph": "s", "id": 190409, "pid": 76337, "tid": -914061504, "ts": 1716454224316946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224425255, "dur": 38,
    "args": {
      "External id": 190418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190418,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190418, "pid": 5, "tid": 7, "ts": 1716454224425255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224316995, "dur": 10,
    "args": {
      "External id": 190418,
      "cbid": 211, "correlation": 190418
    }
  },
  {
    "ph": "s", "id": 190418, "pid": 76337, "tid": -914061504, "ts": 1716454224316995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317059, "dur": 0,
    "args": {
      "External id": 190428,
      "cbid": 317, "correlation": 190428
    }
  },
  {
    "ph": "f", "id": 190428, "pid": 76337, "tid": -914061504, "ts": 1716454224317059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317060, "dur": 0,
    "args": {
      "External id": 190429,
      "cbid": 203, "correlation": 190429
    }
  },
  {
    "ph": "f", "id": 190429, "pid": 76337, "tid": -914061504, "ts": 1716454224317060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317061, "dur": 0,
    "args": {
      "External id": 190430,
      "cbid": 205, "correlation": 190430
    }
  },
  {
    "ph": "f", "id": 190430, "pid": 76337, "tid": -914061504, "ts": 1716454224317061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224425294, "dur": 40,
    "args": {
      "External id": 190434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190434,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190434, "pid": 5, "tid": 7, "ts": 1716454224425294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317075, "dur": 12,
    "args": {
      "External id": 190434,
      "cbid": 211, "correlation": 190434
    }
  },
  {
    "ph": "s", "id": 190434, "pid": 76337, "tid": -914061504, "ts": 1716454224317075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224425335, "dur": 14,
    "args": {
      "External id": 190436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190436,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190436, "pid": 5, "tid": 7, "ts": 1716454224425335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317089, "dur": 5,
    "args": {
      "External id": 190436,
      "cbid": 211, "correlation": 190436
    }
  },
  {
    "ph": "s", "id": 190436, "pid": 76337, "tid": -914061504, "ts": 1716454224317089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224425351, "dur": 3,
    "args": {
      "External id": 190438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190438,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 190438, "pid": 5, "tid": 7, "ts": 1716454224425351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317099, "dur": 5,
    "args": {
      "External id": 190438,
      "cbid": 211, "correlation": 190438
    }
  },
  {
    "ph": "s", "id": 190438, "pid": 76337, "tid": -914061504, "ts": 1716454224317099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317107, "dur": 0,
    "args": {
      "External id": 190439,
      "cbid": 51, "correlation": 190439
    }
  },
  {
    "ph": "s", "id": 190439, "pid": 76337, "tid": -914061504, "ts": 1716454224317107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224425356, "dur": 702,
    "args": {
      "External id": 190440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190440,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190440, "pid": 5, "tid": 7, "ts": 1716454224425356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317108, "dur": 5,
    "args": {
      "External id": 190440,
      "cbid": 211, "correlation": 190440
    }
  },
  {
    "ph": "s", "id": 190440, "pid": 76337, "tid": -914061504, "ts": 1716454224317108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224426059, "dur": 61,
    "args": {
      "External id": 190445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190445,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190445, "pid": 5, "tid": 7, "ts": 1716454224426059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317135, "dur": 8,
    "args": {
      "External id": 190445,
      "cbid": 211, "correlation": 190445
    }
  },
  {
    "ph": "s", "id": 190445, "pid": 76337, "tid": -914061504, "ts": 1716454224317135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317192, "dur": 0,
    "args": {
      "External id": 190455,
      "cbid": 317, "correlation": 190455
    }
  },
  {
    "ph": "f", "id": 190455, "pid": 76337, "tid": -914061504, "ts": 1716454224317192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317193, "dur": 0,
    "args": {
      "External id": 190456,
      "cbid": 203, "correlation": 190456
    }
  },
  {
    "ph": "f", "id": 190456, "pid": 76337, "tid": -914061504, "ts": 1716454224317193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317193, "dur": 0,
    "args": {
      "External id": 190457,
      "cbid": 205, "correlation": 190457
    }
  },
  {
    "ph": "f", "id": 190457, "pid": 76337, "tid": -914061504, "ts": 1716454224317193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224426121, "dur": 74,
    "args": {
      "External id": 190461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190461,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190461, "pid": 5, "tid": 7, "ts": 1716454224426121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317206, "dur": 12,
    "args": {
      "External id": 190461,
      "cbid": 211, "correlation": 190461
    }
  },
  {
    "ph": "s", "id": 190461, "pid": 76337, "tid": -914061504, "ts": 1716454224317206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224426196, "dur": 208,
    "args": {
      "External id": 190463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190463,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190463, "pid": 5, "tid": 7, "ts": 1716454224426196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317225, "dur": 8,
    "args": {
      "External id": 190463,
      "cbid": 211, "correlation": 190463
    }
  },
  {
    "ph": "s", "id": 190463, "pid": 76337, "tid": -914061504, "ts": 1716454224317225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224426405, "dur": 40,
    "args": {
      "External id": 190465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190465,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190465, "pid": 5, "tid": 7, "ts": 1716454224426405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317238, "dur": 202,
    "args": {
      "External id": 190465,
      "cbid": 211, "correlation": 190465
    }
  },
  {
    "ph": "s", "id": 190465, "pid": 76337, "tid": -914061504, "ts": 1716454224317238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224426447, "dur": 59,
    "args": {
      "External id": 190471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190471,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190471, "pid": 5, "tid": 7, "ts": 1716454224426447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317462, "dur": 9,
    "args": {
      "External id": 190471,
      "cbid": 211, "correlation": 190471
    }
  },
  {
    "ph": "s", "id": 190471, "pid": 76337, "tid": -914061504, "ts": 1716454224317462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224426507, "dur": 49,
    "args": {
      "External id": 190479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190479,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190479, "pid": 5, "tid": 7, "ts": 1716454224426507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317493, "dur": 8,
    "args": {
      "External id": 190479,
      "cbid": 211, "correlation": 190479
    }
  },
  {
    "ph": "s", "id": 190479, "pid": 76337, "tid": -914061504, "ts": 1716454224317493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224426558, "dur": 35,
    "args": {
      "External id": 190487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190487,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190487, "pid": 5, "tid": 7, "ts": 1716454224426558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317520, "dur": 30,
    "args": {
      "External id": 190487,
      "cbid": 211, "correlation": 190487
    }
  },
  {
    "ph": "s", "id": 190487, "pid": 76337, "tid": -914061504, "ts": 1716454224317520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224426594, "dur": 54,
    "args": {
      "External id": 190507,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190507,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 190507, "pid": 5, "tid": 7, "ts": 1716454224426594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317626, "dur": 13,
    "args": {
      "External id": 190507,
      "cbid": 211, "correlation": 190507
    }
  },
  {
    "ph": "s", "id": 190507, "pid": 76337, "tid": -914061504, "ts": 1716454224317626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224426649, "dur": 4,
    "args": {
      "External id": 190519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190519,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 190519, "pid": 5, "tid": 7, "ts": 1716454224426649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317650, "dur": 6,
    "args": {
      "External id": 190519,
      "cbid": 211, "correlation": 190519
    }
  },
  {
    "ph": "s", "id": 190519, "pid": 76337, "tid": -914061504, "ts": 1716454224317650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224426655, "dur": 56,
    "args": {
      "External id": 190522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190522,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190522, "pid": 5, "tid": 7, "ts": 1716454224426655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317667, "dur": 6,
    "args": {
      "External id": 190522,
      "cbid": 211, "correlation": 190522
    }
  },
  {
    "ph": "s", "id": 190522, "pid": 76337, "tid": -914061504, "ts": 1716454224317667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317725, "dur": 0,
    "args": {
      "External id": 190533,
      "cbid": 317, "correlation": 190533
    }
  },
  {
    "ph": "f", "id": 190533, "pid": 76337, "tid": -914061504, "ts": 1716454224317725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317726, "dur": 0,
    "args": {
      "External id": 190534,
      "cbid": 203, "correlation": 190534
    }
  },
  {
    "ph": "f", "id": 190534, "pid": 76337, "tid": -914061504, "ts": 1716454224317726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317727, "dur": 0,
    "args": {
      "External id": 190535,
      "cbid": 205, "correlation": 190535
    }
  },
  {
    "ph": "f", "id": 190535, "pid": 76337, "tid": -914061504, "ts": 1716454224317727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317750, "dur": 1,
    "args": {
      "External id": 190539,
      "cbid": 251, "correlation": 190539
    }
  },
  {
    "ph": "f", "id": 190539, "pid": 76337, "tid": -914061504, "ts": 1716454224317750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317752, "dur": 0,
    "args": {
      "External id": 190540,
      "cbid": 251, "correlation": 190540
    }
  },
  {
    "ph": "f", "id": 190540, "pid": 76337, "tid": -914061504, "ts": 1716454224317752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317752, "dur": 0,
    "args": {
      "External id": 190541,
      "cbid": 251, "correlation": 190541
    }
  },
  {
    "ph": "f", "id": 190541, "pid": 76337, "tid": -914061504, "ts": 1716454224317752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317753, "dur": 0,
    "args": {
      "External id": 190542,
      "cbid": 251, "correlation": 190542
    }
  },
  {
    "ph": "f", "id": 190542, "pid": 76337, "tid": -914061504, "ts": 1716454224317753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317754, "dur": 0,
    "args": {
      "External id": 190543,
      "cbid": 251, "correlation": 190543
    }
  },
  {
    "ph": "f", "id": 190543, "pid": 76337, "tid": -914061504, "ts": 1716454224317754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317755, "dur": 0,
    "args": {
      "External id": 190544,
      "cbid": 251, "correlation": 190544
    }
  },
  {
    "ph": "f", "id": 190544, "pid": 76337, "tid": -914061504, "ts": 1716454224317755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317756, "dur": 0,
    "args": {
      "External id": 190545,
      "cbid": 251, "correlation": 190545
    }
  },
  {
    "ph": "f", "id": 190545, "pid": 76337, "tid": -914061504, "ts": 1716454224317756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317756, "dur": 0,
    "args": {
      "External id": 190546,
      "cbid": 251, "correlation": 190546
    }
  },
  {
    "ph": "f", "id": 190546, "pid": 76337, "tid": -914061504, "ts": 1716454224317756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317758, "dur": 0,
    "args": {
      "External id": 190547,
      "cbid": 251, "correlation": 190547
    }
  },
  {
    "ph": "f", "id": 190547, "pid": 76337, "tid": -914061504, "ts": 1716454224317758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224426712, "dur": 113,
    "args": {
      "External id": 190548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190548,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190548, "pid": 5, "tid": 7, "ts": 1716454224426712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317760, "dur": 13,
    "args": {
      "External id": 190548,
      "cbid": 211, "correlation": 190548
    }
  },
  {
    "ph": "s", "id": 190548, "pid": 76337, "tid": -914061504, "ts": 1716454224317760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224426826, "dur": 60,
    "args": {
      "External id": 190554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190554,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190554, "pid": 5, "tid": 7, "ts": 1716454224426826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317795, "dur": 9,
    "args": {
      "External id": 190554,
      "cbid": 211, "correlation": 190554
    }
  },
  {
    "ph": "s", "id": 190554, "pid": 76337, "tid": -914061504, "ts": 1716454224317795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224426887, "dur": 557,
    "args": {
      "External id": 190563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190563,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190563, "pid": 5, "tid": 7, "ts": 1716454224426887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317877, "dur": 15,
    "args": {
      "External id": 190563,
      "cbid": 211, "correlation": 190563
    }
  },
  {
    "ph": "s", "id": 190563, "pid": 76337, "tid": -914061504, "ts": 1716454224317877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224427445, "dur": 182,
    "args": {
      "External id": 190585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190585,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190585, "pid": 5, "tid": 7, "ts": 1716454224427445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224317936, "dur": 10,
    "args": {
      "External id": 190585,
      "cbid": 211, "correlation": 190585
    }
  },
  {
    "ph": "s", "id": 190585, "pid": 76337, "tid": -914061504, "ts": 1716454224317936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318030, "dur": 1,
    "args": {
      "External id": 190596,
      "cbid": 251, "correlation": 190596
    }
  },
  {
    "ph": "f", "id": 190596, "pid": 76337, "tid": -914061504, "ts": 1716454224318030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224427628, "dur": 196,
    "args": {
      "External id": 190597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190597,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190597, "pid": 5, "tid": 7, "ts": 1716454224427628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318035, "dur": 15,
    "args": {
      "External id": 190597,
      "cbid": 211, "correlation": 190597
    }
  },
  {
    "ph": "s", "id": 190597, "pid": 76337, "tid": -914061504, "ts": 1716454224318035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318105, "dur": 1,
    "args": {
      "External id": 190608,
      "cbid": 251, "correlation": 190608
    }
  },
  {
    "ph": "f", "id": 190608, "pid": 76337, "tid": -914061504, "ts": 1716454224318105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224427826, "dur": 189,
    "args": {
      "External id": 190609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190609,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190609, "pid": 5, "tid": 7, "ts": 1716454224427826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318109, "dur": 11,
    "args": {
      "External id": 190609,
      "cbid": 211, "correlation": 190609
    }
  },
  {
    "ph": "s", "id": 190609, "pid": 76337, "tid": -914061504, "ts": 1716454224318109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318171, "dur": 1,
    "args": {
      "External id": 190620,
      "cbid": 251, "correlation": 190620
    }
  },
  {
    "ph": "f", "id": 190620, "pid": 76337, "tid": -914061504, "ts": 1716454224318171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224428016, "dur": 188,
    "args": {
      "External id": 190621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190621,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190621, "pid": 5, "tid": 7, "ts": 1716454224428016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318176, "dur": 11,
    "args": {
      "External id": 190621,
      "cbid": 211, "correlation": 190621
    }
  },
  {
    "ph": "s", "id": 190621, "pid": 76337, "tid": -914061504, "ts": 1716454224318176,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224428206, "dur": 18742,
    "args": {
      "External id": 190642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190642,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190642, "pid": 5, "tid": 7, "ts": 1716454224428206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318257, "dur": 12,
    "args": {
      "External id": 190642,
      "cbid": 211, "correlation": 190642
    }
  },
  {
    "ph": "s", "id": 190642, "pid": 76337, "tid": -914061504, "ts": 1716454224318257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318354, "dur": 1,
    "args": {
      "External id": 190660,
      "cbid": 251, "correlation": 190660
    }
  },
  {
    "ph": "f", "id": 190660, "pid": 76337, "tid": -914061504, "ts": 1716454224318354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224446949, "dur": 204,
    "args": {
      "External id": 190662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190662,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190662, "pid": 5, "tid": 7, "ts": 1716454224446949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318359, "dur": 13,
    "args": {
      "External id": 190662,
      "cbid": 211, "correlation": 190662
    }
  },
  {
    "ph": "s", "id": 190662, "pid": 76337, "tid": -914061504, "ts": 1716454224318359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224447155, "dur": 66,
    "args": {
      "External id": 190670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190670,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190670, "pid": 5, "tid": 7, "ts": 1716454224447155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318428, "dur": 13,
    "args": {
      "External id": 190670,
      "cbid": 211, "correlation": 190670
    }
  },
  {
    "ph": "s", "id": 190670, "pid": 76337, "tid": -914061504, "ts": 1716454224318428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224447222, "dur": 96,
    "args": {
      "External id": 190678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190678,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190678, "pid": 5, "tid": 7, "ts": 1716454224447222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318468, "dur": 13,
    "args": {
      "External id": 190678,
      "cbid": 211, "correlation": 190678
    }
  },
  {
    "ph": "s", "id": 190678, "pid": 76337, "tid": -914061504, "ts": 1716454224318468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224447320, "dur": 54,
    "args": {
      "External id": 190689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190689,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190689, "pid": 5, "tid": 7, "ts": 1716454224447320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318544, "dur": 83,
    "args": {
      "External id": 190689,
      "cbid": 211, "correlation": 190689
    }
  },
  {
    "ph": "s", "id": 190689, "pid": 76337, "tid": -914061504, "ts": 1716454224318544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224447375, "dur": 92,
    "args": {
      "External id": 190711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190711,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190711, "pid": 5, "tid": 7, "ts": 1716454224447375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224318646, "dur": 1945,
    "args": {
      "External id": 190711,
      "cbid": 211, "correlation": 190711
    }
  },
  {
    "ph": "s", "id": 190711, "pid": 76337, "tid": -914061504, "ts": 1716454224318646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320671, "dur": 1,
    "args": {
      "External id": 190722,
      "cbid": 251, "correlation": 190722
    }
  },
  {
    "ph": "f", "id": 190722, "pid": 76337, "tid": -914061504, "ts": 1716454224320671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224447469, "dur": 105,
    "args": {
      "External id": 190723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190723,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190723, "pid": 5, "tid": 7, "ts": 1716454224447469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320676, "dur": 67,
    "args": {
      "External id": 190723,
      "cbid": 211, "correlation": 190723
    }
  },
  {
    "ph": "s", "id": 190723, "pid": 76337, "tid": -914061504, "ts": 1716454224320676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320802, "dur": 1,
    "args": {
      "External id": 190734,
      "cbid": 251, "correlation": 190734
    }
  },
  {
    "ph": "f", "id": 190734, "pid": 76337, "tid": -914061504, "ts": 1716454224320802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320806, "dur": 0,
    "args": {
      "External id": 190735,
      "cbid": 251, "correlation": 190735
    }
  },
  {
    "ph": "f", "id": 190735, "pid": 76337, "tid": -914061504, "ts": 1716454224320806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224447575, "dur": 10,
    "args": {
      "External id": 190736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190736,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 190736, "pid": 5, "tid": 7, "ts": 1716454224447575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320807, "dur": 14,
    "args": {
      "External id": 190736,
      "cbid": 211, "correlation": 190736
    }
  },
  {
    "ph": "s", "id": 190736, "pid": 76337, "tid": -914061504, "ts": 1716454224320807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224447587, "dur": 5,
    "args": {
      "External id": 190738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190738,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190738, "pid": 5, "tid": 7, "ts": 1716454224447587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320823, "dur": 6,
    "args": {
      "External id": 190738,
      "cbid": 211, "correlation": 190738
    }
  },
  {
    "ph": "s", "id": 190738, "pid": 76337, "tid": -914061504, "ts": 1716454224320823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320883, "dur": 1,
    "args": {
      "External id": 190749,
      "cbid": 251, "correlation": 190749
    }
  },
  {
    "ph": "f", "id": 190749, "pid": 76337, "tid": -914061504, "ts": 1716454224320883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320887, "dur": 0,
    "args": {
      "External id": 190750,
      "cbid": 251, "correlation": 190750
    }
  },
  {
    "ph": "f", "id": 190750, "pid": 76337, "tid": -914061504, "ts": 1716454224320887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224447593, "dur": 6,
    "args": {
      "External id": 190751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190751,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 190751, "pid": 5, "tid": 7, "ts": 1716454224447593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320888, "dur": 12,
    "args": {
      "External id": 190751,
      "cbid": 211, "correlation": 190751
    }
  },
  {
    "ph": "s", "id": 190751, "pid": 76337, "tid": -914061504, "ts": 1716454224320888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224447600, "dur": 3,
    "args": {
      "External id": 190753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190753,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190753, "pid": 5, "tid": 7, "ts": 1716454224447600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320901, "dur": 5,
    "args": {
      "External id": 190753,
      "cbid": 211, "correlation": 190753
    }
  },
  {
    "ph": "s", "id": 190753, "pid": 76337, "tid": -914061504, "ts": 1716454224320901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224447605, "dur": 157,
    "args": {
      "External id": 190774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190774,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190774, "pid": 5, "tid": 7, "ts": 1716454224447605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224320983, "dur": 13,
    "args": {
      "External id": 190774,
      "cbid": 211, "correlation": 190774
    }
  },
  {
    "ph": "s", "id": 190774, "pid": 76337, "tid": -914061504, "ts": 1716454224320983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321081, "dur": 1,
    "args": {
      "External id": 190792,
      "cbid": 251, "correlation": 190792
    }
  },
  {
    "ph": "f", "id": 190792, "pid": 76337, "tid": -914061504, "ts": 1716454224321081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224447764, "dur": 106,
    "args": {
      "External id": 190794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190794,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190794, "pid": 5, "tid": 7, "ts": 1716454224447764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321087, "dur": 13,
    "args": {
      "External id": 190794,
      "cbid": 211, "correlation": 190794
    }
  },
  {
    "ph": "s", "id": 190794, "pid": 76337, "tid": -914061504, "ts": 1716454224321087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224447871, "dur": 35,
    "args": {
      "External id": 190802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190802,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190802, "pid": 5, "tid": 7, "ts": 1716454224447871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321156, "dur": 13,
    "args": {
      "External id": 190802,
      "cbid": 211, "correlation": 190802
    }
  },
  {
    "ph": "s", "id": 190802, "pid": 76337, "tid": -914061504, "ts": 1716454224321156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224447908, "dur": 67,
    "args": {
      "External id": 190810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190810,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190810, "pid": 5, "tid": 7, "ts": 1716454224447908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321197, "dur": 9,
    "args": {
      "External id": 190810,
      "cbid": 211, "correlation": 190810
    }
  },
  {
    "ph": "s", "id": 190810, "pid": 76337, "tid": -914061504, "ts": 1716454224321197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224447976, "dur": 93,
    "args": {
      "External id": 190832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190832,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190832, "pid": 5, "tid": 7, "ts": 1716454224447976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321248, "dur": 10,
    "args": {
      "External id": 190832,
      "cbid": 211, "correlation": 190832
    }
  },
  {
    "ph": "s", "id": 190832, "pid": 76337, "tid": -914061504, "ts": 1716454224321248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321334, "dur": 1,
    "args": {
      "External id": 190848,
      "cbid": 251, "correlation": 190848
    }
  },
  {
    "ph": "f", "id": 190848, "pid": 76337, "tid": -914061504, "ts": 1716454224321334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224448070, "dur": 577,
    "args": {
      "External id": 190850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190850,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 190850, "pid": 5, "tid": 7, "ts": 1716454224448070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321340, "dur": 12,
    "args": {
      "External id": 190850,
      "cbid": 211, "correlation": 190850
    }
  },
  {
    "ph": "s", "id": 190850, "pid": 76337, "tid": -914061504, "ts": 1716454224321340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224448648, "dur": 245,
    "args": {
      "External id": 190858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190858, "pid": 5, "tid": 7, "ts": 1716454224448648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321405, "dur": 13,
    "args": {
      "External id": 190858,
      "cbid": 211, "correlation": 190858
    }
  },
  {
    "ph": "s", "id": 190858, "pid": 76337, "tid": -914061504, "ts": 1716454224321405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224448894, "dur": 253,
    "args": {
      "External id": 190866,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190866,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190866, "pid": 5, "tid": 7, "ts": 1716454224448894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321436, "dur": 8,
    "args": {
      "External id": 190866,
      "cbid": 211, "correlation": 190866
    }
  },
  {
    "ph": "s", "id": 190866, "pid": 76337, "tid": -914061504, "ts": 1716454224321436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321518, "dur": 1,
    "args": {
      "External id": 190882,
      "cbid": 251, "correlation": 190882
    }
  },
  {
    "ph": "f", "id": 190882, "pid": 76337, "tid": -914061504, "ts": 1716454224321518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321523, "dur": 0,
    "args": {
      "External id": 190884,
      "cbid": 251, "correlation": 190884
    }
  },
  {
    "ph": "f", "id": 190884, "pid": 76337, "tid": -914061504, "ts": 1716454224321523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224449148, "dur": 359,
    "args": {
      "External id": 190885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190885,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 190885, "pid": 5, "tid": 7, "ts": 1716454224449148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321525, "dur": 13,
    "args": {
      "External id": 190885,
      "cbid": 211, "correlation": 190885
    }
  },
  {
    "ph": "s", "id": 190885, "pid": 76337, "tid": -914061504, "ts": 1716454224321525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224449508, "dur": 50,
    "args": {
      "External id": 190893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190893,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190893, "pid": 5, "tid": 7, "ts": 1716454224449508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321568, "dur": 10,
    "args": {
      "External id": 190893,
      "cbid": 211, "correlation": 190893
    }
  },
  {
    "ph": "s", "id": 190893, "pid": 76337, "tid": -914061504, "ts": 1716454224321568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224449560, "dur": 160,
    "args": {
      "External id": 190904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190904,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190904, "pid": 5, "tid": 7, "ts": 1716454224449560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321635, "dur": 225,
    "args": {
      "External id": 190904,
      "cbid": 211, "correlation": 190904
    }
  },
  {
    "ph": "s", "id": 190904, "pid": 76337, "tid": -914061504, "ts": 1716454224321635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321912, "dur": 0,
    "args": {
      "External id": 190916,
      "cbid": 317, "correlation": 190916
    }
  },
  {
    "ph": "f", "id": 190916, "pid": 76337, "tid": -914061504, "ts": 1716454224321912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321913, "dur": 0,
    "args": {
      "External id": 190917,
      "cbid": 203, "correlation": 190917
    }
  },
  {
    "ph": "f", "id": 190917, "pid": 76337, "tid": -914061504, "ts": 1716454224321913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321914, "dur": 0,
    "args": {
      "External id": 190918,
      "cbid": 205, "correlation": 190918
    }
  },
  {
    "ph": "f", "id": 190918, "pid": 76337, "tid": -914061504, "ts": 1716454224321914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321936, "dur": 1,
    "args": {
      "External id": 190922,
      "cbid": 251, "correlation": 190922
    }
  },
  {
    "ph": "f", "id": 190922, "pid": 76337, "tid": -914061504, "ts": 1716454224321936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321938, "dur": 0,
    "args": {
      "External id": 190923,
      "cbid": 251, "correlation": 190923
    }
  },
  {
    "ph": "f", "id": 190923, "pid": 76337, "tid": -914061504, "ts": 1716454224321938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321938, "dur": 0,
    "args": {
      "External id": 190924,
      "cbid": 251, "correlation": 190924
    }
  },
  {
    "ph": "f", "id": 190924, "pid": 76337, "tid": -914061504, "ts": 1716454224321938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321939, "dur": 0,
    "args": {
      "External id": 190925,
      "cbid": 251, "correlation": 190925
    }
  },
  {
    "ph": "f", "id": 190925, "pid": 76337, "tid": -914061504, "ts": 1716454224321939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321940, "dur": 0,
    "args": {
      "External id": 190926,
      "cbid": 251, "correlation": 190926
    }
  },
  {
    "ph": "f", "id": 190926, "pid": 76337, "tid": -914061504, "ts": 1716454224321940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321940, "dur": 0,
    "args": {
      "External id": 190927,
      "cbid": 251, "correlation": 190927
    }
  },
  {
    "ph": "f", "id": 190927, "pid": 76337, "tid": -914061504, "ts": 1716454224321940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321941, "dur": 0,
    "args": {
      "External id": 190928,
      "cbid": 251, "correlation": 190928
    }
  },
  {
    "ph": "f", "id": 190928, "pid": 76337, "tid": -914061504, "ts": 1716454224321941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321942, "dur": 0,
    "args": {
      "External id": 190929,
      "cbid": 251, "correlation": 190929
    }
  },
  {
    "ph": "f", "id": 190929, "pid": 76337, "tid": -914061504, "ts": 1716454224321942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321943, "dur": 0,
    "args": {
      "External id": 190930,
      "cbid": 251, "correlation": 190930
    }
  },
  {
    "ph": "f", "id": 190930, "pid": 76337, "tid": -914061504, "ts": 1716454224321943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224449720, "dur": 115,
    "args": {
      "External id": 190931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190931,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 190931, "pid": 5, "tid": 7, "ts": 1716454224449720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224321945, "dur": 42,
    "args": {
      "External id": 190931,
      "cbid": 211, "correlation": 190931
    }
  },
  {
    "ph": "s", "id": 190931, "pid": 76337, "tid": -914061504, "ts": 1716454224321945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224449836, "dur": 61,
    "args": {
      "External id": 190937,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190937,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190937, "pid": 5, "tid": 7, "ts": 1716454224449836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322011, "dur": 104,
    "args": {
      "External id": 190937,
      "cbid": 211, "correlation": 190937
    }
  },
  {
    "ph": "s", "id": 190937, "pid": 76337, "tid": -914061504, "ts": 1716454224322011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224449898, "dur": 50,
    "args": {
      "External id": 190945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190945,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190945, "pid": 5, "tid": 7, "ts": 1716454224449898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322139, "dur": 285,
    "args": {
      "External id": 190945,
      "cbid": 211, "correlation": 190945
    }
  },
  {
    "ph": "s", "id": 190945, "pid": 76337, "tid": -914061504, "ts": 1716454224322139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224449949, "dur": 98,
    "args": {
      "External id": 190954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190954,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190954, "pid": 5, "tid": 7, "ts": 1716454224449949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322455, "dur": 10,
    "args": {
      "External id": 190954,
      "cbid": 211, "correlation": 190954
    }
  },
  {
    "ph": "s", "id": 190954, "pid": 76337, "tid": -914061504, "ts": 1716454224322455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224450048, "dur": 95,
    "args": {
      "External id": 190974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190974,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 190974, "pid": 5, "tid": 7, "ts": 1716454224450048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322527, "dur": 12,
    "args": {
      "External id": 190974,
      "cbid": 211, "correlation": 190974
    }
  },
  {
    "ph": "s", "id": 190974, "pid": 76337, "tid": -914061504, "ts": 1716454224322527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224450144, "dur": 4,
    "args": {
      "External id": 190986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190986,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 190986, "pid": 5, "tid": 7, "ts": 1716454224450144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322550, "dur": 10,
    "args": {
      "External id": 190986,
      "cbid": 211, "correlation": 190986
    }
  },
  {
    "ph": "s", "id": 190986, "pid": 76337, "tid": -914061504, "ts": 1716454224322550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224450150, "dur": 109,
    "args": {
      "External id": 190989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190989,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190989, "pid": 5, "tid": 7, "ts": 1716454224450150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322572, "dur": 110,
    "args": {
      "External id": 190989,
      "cbid": 211, "correlation": 190989
    }
  },
  {
    "ph": "s", "id": 190989, "pid": 76337, "tid": -914061504, "ts": 1716454224322572,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224450261, "dur": 69,
    "args": {
      "External id": 190998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 190998,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 190998, "pid": 5, "tid": 7, "ts": 1716454224450261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322716, "dur": 11,
    "args": {
      "External id": 190998,
      "cbid": 211, "correlation": 190998
    }
  },
  {
    "ph": "s", "id": 190998, "pid": 76337, "tid": -914061504, "ts": 1716454224322716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322769, "dur": 0,
    "args": {
      "External id": 191008,
      "cbid": 317, "correlation": 191008
    }
  },
  {
    "ph": "f", "id": 191008, "pid": 76337, "tid": -914061504, "ts": 1716454224322769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322770, "dur": 0,
    "args": {
      "External id": 191009,
      "cbid": 203, "correlation": 191009
    }
  },
  {
    "ph": "f", "id": 191009, "pid": 76337, "tid": -914061504, "ts": 1716454224322770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322771, "dur": 0,
    "args": {
      "External id": 191010,
      "cbid": 205, "correlation": 191010
    }
  },
  {
    "ph": "f", "id": 191010, "pid": 76337, "tid": -914061504, "ts": 1716454224322771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224450331, "dur": 76,
    "args": {
      "External id": 191014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191014, "pid": 5, "tid": 7, "ts": 1716454224450331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322784, "dur": 12,
    "args": {
      "External id": 191014,
      "cbid": 211, "correlation": 191014
    }
  },
  {
    "ph": "s", "id": 191014, "pid": 76337, "tid": -914061504, "ts": 1716454224322784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224450409, "dur": 24,
    "args": {
      "External id": 191016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191016,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191016, "pid": 5, "tid": 7, "ts": 1716454224450409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322798, "dur": 5,
    "args": {
      "External id": 191016,
      "cbid": 211, "correlation": 191016
    }
  },
  {
    "ph": "s", "id": 191016, "pid": 76337, "tid": -914061504, "ts": 1716454224322798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224450434, "dur": 4,
    "args": {
      "External id": 191018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191018, "pid": 5, "tid": 7, "ts": 1716454224450434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322809, "dur": 6,
    "args": {
      "External id": 191018,
      "cbid": 211, "correlation": 191018
    }
  },
  {
    "ph": "s", "id": 191018, "pid": 76337, "tid": -914061504, "ts": 1716454224322809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322817, "dur": 0,
    "args": {
      "External id": 191019,
      "cbid": 51, "correlation": 191019
    }
  },
  {
    "ph": "s", "id": 191019, "pid": 76337, "tid": -914061504, "ts": 1716454224322817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224450439, "dur": 1375,
    "args": {
      "External id": 191020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191020,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191020, "pid": 5, "tid": 7, "ts": 1716454224450439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322818, "dur": 5,
    "args": {
      "External id": 191020,
      "cbid": 211, "correlation": 191020
    }
  },
  {
    "ph": "s", "id": 191020, "pid": 76337, "tid": -914061504, "ts": 1716454224322818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224451816, "dur": 59,
    "args": {
      "External id": 191025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191025,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191025, "pid": 5, "tid": 7, "ts": 1716454224451816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322845, "dur": 9,
    "args": {
      "External id": 191025,
      "cbid": 211, "correlation": 191025
    }
  },
  {
    "ph": "s", "id": 191025, "pid": 76337, "tid": -914061504, "ts": 1716454224322845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224451876, "dur": 4,
    "args": {
      "External id": 191033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191033,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191033, "pid": 5, "tid": 7, "ts": 1716454224451876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322889, "dur": 9,
    "args": {
      "External id": 191033,
      "cbid": 211, "correlation": 191033
    }
  },
  {
    "ph": "s", "id": 191033, "pid": 76337, "tid": -914061504, "ts": 1716454224322889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322955, "dur": 1,
    "args": {
      "External id": 191049,
      "cbid": 251, "correlation": 191049
    }
  },
  {
    "ph": "f", "id": 191049, "pid": 76337, "tid": -914061504, "ts": 1716454224322955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322960, "dur": 0,
    "args": {
      "External id": 191051,
      "cbid": 251, "correlation": 191051
    }
  },
  {
    "ph": "f", "id": 191051, "pid": 76337, "tid": -914061504, "ts": 1716454224322960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224451881, "dur": 11,
    "args": {
      "External id": 191052,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191052,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 191052, "pid": 5, "tid": 7, "ts": 1716454224451881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322962, "dur": 20,
    "args": {
      "External id": 191052,
      "cbid": 211, "correlation": 191052
    }
  },
  {
    "ph": "s", "id": 191052, "pid": 76337, "tid": -914061504, "ts": 1716454224322962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224451893, "dur": 5,
    "args": {
      "External id": 191054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191054,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 191054, "pid": 5, "tid": 7, "ts": 1716454224451893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224322984, "dur": 6,
    "args": {
      "External id": 191054,
      "cbid": 211, "correlation": 191054
    }
  },
  {
    "ph": "s", "id": 191054, "pid": 76337, "tid": -914061504, "ts": 1716454224322984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224451900, "dur": 55,
    "args": {
      "External id": 191064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191064,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191064, "pid": 5, "tid": 7, "ts": 1716454224451900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323043, "dur": 553,
    "args": {
      "External id": 191064,
      "cbid": 211, "correlation": 191064
    }
  },
  {
    "ph": "s", "id": 191064, "pid": 76337, "tid": -914061504, "ts": 1716454224323043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224451956, "dur": 53,
    "args": {
      "External id": 191084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191084,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 191084, "pid": 5, "tid": 7, "ts": 1716454224451956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323651, "dur": 11,
    "args": {
      "External id": 191084,
      "cbid": 211, "correlation": 191084
    }
  },
  {
    "ph": "s", "id": 191084, "pid": 76337, "tid": -914061504, "ts": 1716454224323651,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224452011, "dur": 4,
    "args": {
      "External id": 191096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191096,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191096, "pid": 5, "tid": 7, "ts": 1716454224452011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323671, "dur": 6,
    "args": {
      "External id": 191096,
      "cbid": 211, "correlation": 191096
    }
  },
  {
    "ph": "s", "id": 191096, "pid": 76337, "tid": -914061504, "ts": 1716454224323671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224452016, "dur": 56,
    "args": {
      "External id": 191099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191099,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191099, "pid": 5, "tid": 7, "ts": 1716454224452016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323691, "dur": 6,
    "args": {
      "External id": 191099,
      "cbid": 211, "correlation": 191099
    }
  },
  {
    "ph": "s", "id": 191099, "pid": 76337, "tid": -914061504, "ts": 1716454224323691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224452074, "dur": 37,
    "args": {
      "External id": 191108,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191108,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191108, "pid": 5, "tid": 7, "ts": 1716454224452074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323732, "dur": 10,
    "args": {
      "External id": 191108,
      "cbid": 211, "correlation": 191108
    }
  },
  {
    "ph": "s", "id": 191108, "pid": 76337, "tid": -914061504, "ts": 1716454224323732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323794, "dur": 0,
    "args": {
      "External id": 191118,
      "cbid": 317, "correlation": 191118
    }
  },
  {
    "ph": "f", "id": 191118, "pid": 76337, "tid": -914061504, "ts": 1716454224323794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323795, "dur": 0,
    "args": {
      "External id": 191119,
      "cbid": 203, "correlation": 191119
    }
  },
  {
    "ph": "f", "id": 191119, "pid": 76337, "tid": -914061504, "ts": 1716454224323795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323796, "dur": 0,
    "args": {
      "External id": 191120,
      "cbid": 205, "correlation": 191120
    }
  },
  {
    "ph": "f", "id": 191120, "pid": 76337, "tid": -914061504, "ts": 1716454224323796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224452112, "dur": 40,
    "args": {
      "External id": 191124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191124,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191124, "pid": 5, "tid": 7, "ts": 1716454224452112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323810, "dur": 12,
    "args": {
      "External id": 191124,
      "cbid": 211, "correlation": 191124
    }
  },
  {
    "ph": "s", "id": 191124, "pid": 76337, "tid": -914061504, "ts": 1716454224323810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224452154, "dur": 14,
    "args": {
      "External id": 191126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191126,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191126, "pid": 5, "tid": 7, "ts": 1716454224452154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323824, "dur": 5,
    "args": {
      "External id": 191126,
      "cbid": 211, "correlation": 191126
    }
  },
  {
    "ph": "s", "id": 191126, "pid": 76337, "tid": -914061504, "ts": 1716454224323824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224452169, "dur": 3,
    "args": {
      "External id": 191128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191128,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191128, "pid": 5, "tid": 7, "ts": 1716454224452169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323833, "dur": 6,
    "args": {
      "External id": 191128,
      "cbid": 211, "correlation": 191128
    }
  },
  {
    "ph": "s", "id": 191128, "pid": 76337, "tid": -914061504, "ts": 1716454224323833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323843, "dur": 0,
    "args": {
      "External id": 191129,
      "cbid": 51, "correlation": 191129
    }
  },
  {
    "ph": "s", "id": 191129, "pid": 76337, "tid": -914061504, "ts": 1716454224323843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224452174, "dur": 703,
    "args": {
      "External id": 191130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191130,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191130, "pid": 5, "tid": 7, "ts": 1716454224452174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323844, "dur": 5,
    "args": {
      "External id": 191130,
      "cbid": 211, "correlation": 191130
    }
  },
  {
    "ph": "s", "id": 191130, "pid": 76337, "tid": -914061504, "ts": 1716454224323844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224452878, "dur": 60,
    "args": {
      "External id": 191135,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191135,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191135, "pid": 5, "tid": 7, "ts": 1716454224452878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323871, "dur": 9,
    "args": {
      "External id": 191135,
      "cbid": 211, "correlation": 191135
    }
  },
  {
    "ph": "s", "id": 191135, "pid": 76337, "tid": -914061504, "ts": 1716454224323871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323929, "dur": 0,
    "args": {
      "External id": 191145,
      "cbid": 317, "correlation": 191145
    }
  },
  {
    "ph": "f", "id": 191145, "pid": 76337, "tid": -914061504, "ts": 1716454224323929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323930, "dur": 0,
    "args": {
      "External id": 191146,
      "cbid": 203, "correlation": 191146
    }
  },
  {
    "ph": "f", "id": 191146, "pid": 76337, "tid": -914061504, "ts": 1716454224323930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323930, "dur": 0,
    "args": {
      "External id": 191147,
      "cbid": 205, "correlation": 191147
    }
  },
  {
    "ph": "f", "id": 191147, "pid": 76337, "tid": -914061504, "ts": 1716454224323930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224452939, "dur": 75,
    "args": {
      "External id": 191151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191151,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191151, "pid": 5, "tid": 7, "ts": 1716454224452939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323943, "dur": 11,
    "args": {
      "External id": 191151,
      "cbid": 211, "correlation": 191151
    }
  },
  {
    "ph": "s", "id": 191151, "pid": 76337, "tid": -914061504, "ts": 1716454224323943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224453016, "dur": 209,
    "args": {
      "External id": 191153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191153,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191153, "pid": 5, "tid": 7, "ts": 1716454224453016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323960, "dur": 6,
    "args": {
      "External id": 191153,
      "cbid": 211, "correlation": 191153
    }
  },
  {
    "ph": "s", "id": 191153, "pid": 76337, "tid": -914061504, "ts": 1716454224323960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224453226, "dur": 39,
    "args": {
      "External id": 191155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191155,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191155, "pid": 5, "tid": 7, "ts": 1716454224453226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224323971, "dur": 14,
    "args": {
      "External id": 191155,
      "cbid": 211, "correlation": 191155
    }
  },
  {
    "ph": "s", "id": 191155, "pid": 76337, "tid": -914061504, "ts": 1716454224323971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224453266, "dur": 59,
    "args": {
      "External id": 191161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191161,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191161, "pid": 5, "tid": 7, "ts": 1716454224453266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324006, "dur": 516,
    "args": {
      "External id": 191161,
      "cbid": 211, "correlation": 191161
    }
  },
  {
    "ph": "s", "id": 191161, "pid": 76337, "tid": -914061504, "ts": 1716454224324006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224453326, "dur": 50,
    "args": {
      "External id": 191169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191169,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191169, "pid": 5, "tid": 7, "ts": 1716454224453326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324544, "dur": 8,
    "args": {
      "External id": 191169,
      "cbid": 211, "correlation": 191169
    }
  },
  {
    "ph": "s", "id": 191169, "pid": 76337, "tid": -914061504, "ts": 1716454224324544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224453377, "dur": 35,
    "args": {
      "External id": 191177,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191177,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191177, "pid": 5, "tid": 7, "ts": 1716454224453377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324573, "dur": 9,
    "args": {
      "External id": 191177,
      "cbid": 211, "correlation": 191177
    }
  },
  {
    "ph": "s", "id": 191177, "pid": 76337, "tid": -914061504, "ts": 1716454224324573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224453414, "dur": 52,
    "args": {
      "External id": 191197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191197,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 191197, "pid": 5, "tid": 7, "ts": 1716454224453414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324656, "dur": 13,
    "args": {
      "External id": 191197,
      "cbid": 211, "correlation": 191197
    }
  },
  {
    "ph": "s", "id": 191197, "pid": 76337, "tid": -914061504, "ts": 1716454224324656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224453468, "dur": 5,
    "args": {
      "External id": 191209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191209,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191209, "pid": 5, "tid": 7, "ts": 1716454224453468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324679, "dur": 6,
    "args": {
      "External id": 191209,
      "cbid": 211, "correlation": 191209
    }
  },
  {
    "ph": "s", "id": 191209, "pid": 76337, "tid": -914061504, "ts": 1716454224324679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224453474, "dur": 56,
    "args": {
      "External id": 191212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191212, "pid": 5, "tid": 7, "ts": 1716454224453474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324696, "dur": 7,
    "args": {
      "External id": 191212,
      "cbid": 211, "correlation": 191212
    }
  },
  {
    "ph": "s", "id": 191212, "pid": 76337, "tid": -914061504, "ts": 1716454224324696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324753, "dur": 0,
    "args": {
      "External id": 191223,
      "cbid": 317, "correlation": 191223
    }
  },
  {
    "ph": "f", "id": 191223, "pid": 76337, "tid": -914061504, "ts": 1716454224324753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324754, "dur": 0,
    "args": {
      "External id": 191224,
      "cbid": 203, "correlation": 191224
    }
  },
  {
    "ph": "f", "id": 191224, "pid": 76337, "tid": -914061504, "ts": 1716454224324754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324755, "dur": 0,
    "args": {
      "External id": 191225,
      "cbid": 205, "correlation": 191225
    }
  },
  {
    "ph": "f", "id": 191225, "pid": 76337, "tid": -914061504, "ts": 1716454224324755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324777, "dur": 1,
    "args": {
      "External id": 191229,
      "cbid": 251, "correlation": 191229
    }
  },
  {
    "ph": "f", "id": 191229, "pid": 76337, "tid": -914061504, "ts": 1716454224324777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324779, "dur": 0,
    "args": {
      "External id": 191230,
      "cbid": 251, "correlation": 191230
    }
  },
  {
    "ph": "f", "id": 191230, "pid": 76337, "tid": -914061504, "ts": 1716454224324779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324780, "dur": 0,
    "args": {
      "External id": 191231,
      "cbid": 251, "correlation": 191231
    }
  },
  {
    "ph": "f", "id": 191231, "pid": 76337, "tid": -914061504, "ts": 1716454224324780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324781, "dur": 0,
    "args": {
      "External id": 191232,
      "cbid": 251, "correlation": 191232
    }
  },
  {
    "ph": "f", "id": 191232, "pid": 76337, "tid": -914061504, "ts": 1716454224324781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324781, "dur": 0,
    "args": {
      "External id": 191233,
      "cbid": 251, "correlation": 191233
    }
  },
  {
    "ph": "f", "id": 191233, "pid": 76337, "tid": -914061504, "ts": 1716454224324781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324782, "dur": 0,
    "args": {
      "External id": 191234,
      "cbid": 251, "correlation": 191234
    }
  },
  {
    "ph": "f", "id": 191234, "pid": 76337, "tid": -914061504, "ts": 1716454224324782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324783, "dur": 0,
    "args": {
      "External id": 191235,
      "cbid": 251, "correlation": 191235
    }
  },
  {
    "ph": "f", "id": 191235, "pid": 76337, "tid": -914061504, "ts": 1716454224324783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324783, "dur": 0,
    "args": {
      "External id": 191236,
      "cbid": 251, "correlation": 191236
    }
  },
  {
    "ph": "f", "id": 191236, "pid": 76337, "tid": -914061504, "ts": 1716454224324783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324785, "dur": 0,
    "args": {
      "External id": 191237,
      "cbid": 251, "correlation": 191237
    }
  },
  {
    "ph": "f", "id": 191237, "pid": 76337, "tid": -914061504, "ts": 1716454224324785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224453532, "dur": 115,
    "args": {
      "External id": 191238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191238,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 191238, "pid": 5, "tid": 7, "ts": 1716454224453532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324787, "dur": 12,
    "args": {
      "External id": 191238,
      "cbid": 211, "correlation": 191238
    }
  },
  {
    "ph": "s", "id": 191238, "pid": 76337, "tid": -914061504, "ts": 1716454224324787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224453648, "dur": 60,
    "args": {
      "External id": 191244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191244, "pid": 5, "tid": 7, "ts": 1716454224453648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324822, "dur": 9,
    "args": {
      "External id": 191244,
      "cbid": 211, "correlation": 191244
    }
  },
  {
    "ph": "s", "id": 191244, "pid": 76337, "tid": -914061504, "ts": 1716454224324822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224453709, "dur": 606,
    "args": {
      "External id": 191253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191253,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191253, "pid": 5, "tid": 7, "ts": 1716454224453709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324903, "dur": 14,
    "args": {
      "External id": 191253,
      "cbid": 211, "correlation": 191253
    }
  },
  {
    "ph": "s", "id": 191253, "pid": 76337, "tid": -914061504, "ts": 1716454224324903,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224454316, "dur": 182,
    "args": {
      "External id": 191275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191275,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191275, "pid": 5, "tid": 7, "ts": 1716454224454316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224324960, "dur": 11,
    "args": {
      "External id": 191275,
      "cbid": 211, "correlation": 191275
    }
  },
  {
    "ph": "s", "id": 191275, "pid": 76337, "tid": -914061504, "ts": 1716454224324960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325053, "dur": 1,
    "args": {
      "External id": 191286,
      "cbid": 251, "correlation": 191286
    }
  },
  {
    "ph": "f", "id": 191286, "pid": 76337, "tid": -914061504, "ts": 1716454224325053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224454500, "dur": 194,
    "args": {
      "External id": 191287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191287,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191287, "pid": 5, "tid": 7, "ts": 1716454224454500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325059, "dur": 14,
    "args": {
      "External id": 191287,
      "cbid": 211, "correlation": 191287
    }
  },
  {
    "ph": "s", "id": 191287, "pid": 76337, "tid": -914061504, "ts": 1716454224325059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325127, "dur": 1,
    "args": {
      "External id": 191298,
      "cbid": 251, "correlation": 191298
    }
  },
  {
    "ph": "f", "id": 191298, "pid": 76337, "tid": -914061504, "ts": 1716454224325127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224454696, "dur": 190,
    "args": {
      "External id": 191299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191299,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191299, "pid": 5, "tid": 7, "ts": 1716454224454696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325130, "dur": 11,
    "args": {
      "External id": 191299,
      "cbid": 211, "correlation": 191299
    }
  },
  {
    "ph": "s", "id": 191299, "pid": 76337, "tid": -914061504, "ts": 1716454224325130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325192, "dur": 1,
    "args": {
      "External id": 191310,
      "cbid": 251, "correlation": 191310
    }
  },
  {
    "ph": "f", "id": 191310, "pid": 76337, "tid": -914061504, "ts": 1716454224325192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224454887, "dur": 187,
    "args": {
      "External id": 191311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191311,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191311, "pid": 5, "tid": 7, "ts": 1716454224454887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325196, "dur": 11,
    "args": {
      "External id": 191311,
      "cbid": 211, "correlation": 191311
    }
  },
  {
    "ph": "s", "id": 191311, "pid": 76337, "tid": -914061504, "ts": 1716454224325196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224455076, "dur": 18741,
    "args": {
      "External id": 191332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191332,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 191332, "pid": 5, "tid": 7, "ts": 1716454224455076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325278, "dur": 12,
    "args": {
      "External id": 191332,
      "cbid": 211, "correlation": 191332
    }
  },
  {
    "ph": "s", "id": 191332, "pid": 76337, "tid": -914061504, "ts": 1716454224325278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325374, "dur": 1,
    "args": {
      "External id": 191350,
      "cbid": 251, "correlation": 191350
    }
  },
  {
    "ph": "f", "id": 191350, "pid": 76337, "tid": -914061504, "ts": 1716454224325374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224473818, "dur": 203,
    "args": {
      "External id": 191352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191352,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191352, "pid": 5, "tid": 7, "ts": 1716454224473818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325380, "dur": 13,
    "args": {
      "External id": 191352,
      "cbid": 211, "correlation": 191352
    }
  },
  {
    "ph": "s", "id": 191352, "pid": 76337, "tid": -914061504, "ts": 1716454224325380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224474023, "dur": 66,
    "args": {
      "External id": 191360,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191360,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191360, "pid": 5, "tid": 7, "ts": 1716454224474023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325449, "dur": 18,
    "args": {
      "External id": 191360,
      "cbid": 211, "correlation": 191360
    }
  },
  {
    "ph": "s", "id": 191360, "pid": 76337, "tid": -914061504, "ts": 1716454224325449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224474090, "dur": 96,
    "args": {
      "External id": 191368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191368,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191368, "pid": 5, "tid": 7, "ts": 1716454224474090,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325494, "dur": 113,
    "args": {
      "External id": 191368,
      "cbid": 211, "correlation": 191368
    }
  },
  {
    "ph": "s", "id": 191368, "pid": 76337, "tid": -914061504, "ts": 1716454224325494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224474188, "dur": 55,
    "args": {
      "External id": 191379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191379, "pid": 5, "tid": 7, "ts": 1716454224474188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224325670, "dur": 1810,
    "args": {
      "External id": 191379,
      "cbid": 211, "correlation": 191379
    }
  },
  {
    "ph": "s", "id": 191379, "pid": 76337, "tid": -914061504, "ts": 1716454224325670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224474244, "dur": 93,
    "args": {
      "External id": 191401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191401,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191401, "pid": 5, "tid": 7, "ts": 1716454224474244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327501, "dur": 122,
    "args": {
      "External id": 191401,
      "cbid": 211, "correlation": 191401
    }
  },
  {
    "ph": "s", "id": 191401, "pid": 76337, "tid": -914061504, "ts": 1716454224327501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327701, "dur": 1,
    "args": {
      "External id": 191412,
      "cbid": 251, "correlation": 191412
    }
  },
  {
    "ph": "f", "id": 191412, "pid": 76337, "tid": -914061504, "ts": 1716454224327701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224474338, "dur": 104,
    "args": {
      "External id": 191413,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191413,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191413, "pid": 5, "tid": 7, "ts": 1716454224474338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327706, "dur": 13,
    "args": {
      "External id": 191413,
      "cbid": 211, "correlation": 191413
    }
  },
  {
    "ph": "s", "id": 191413, "pid": 76337, "tid": -914061504, "ts": 1716454224327706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327777, "dur": 1,
    "args": {
      "External id": 191424,
      "cbid": 251, "correlation": 191424
    }
  },
  {
    "ph": "f", "id": 191424, "pid": 76337, "tid": -914061504, "ts": 1716454224327777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327781, "dur": 0,
    "args": {
      "External id": 191425,
      "cbid": 251, "correlation": 191425
    }
  },
  {
    "ph": "f", "id": 191425, "pid": 76337, "tid": -914061504, "ts": 1716454224327781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224474443, "dur": 11,
    "args": {
      "External id": 191426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191426,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191426, "pid": 5, "tid": 7, "ts": 1716454224474443,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327783, "dur": 12,
    "args": {
      "External id": 191426,
      "cbid": 211, "correlation": 191426
    }
  },
  {
    "ph": "s", "id": 191426, "pid": 76337, "tid": -914061504, "ts": 1716454224327783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224474455, "dur": 5,
    "args": {
      "External id": 191428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191428,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 191428, "pid": 5, "tid": 7, "ts": 1716454224474455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327797, "dur": 6,
    "args": {
      "External id": 191428,
      "cbid": 211, "correlation": 191428
    }
  },
  {
    "ph": "s", "id": 191428, "pid": 76337, "tid": -914061504, "ts": 1716454224327797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327858, "dur": 1,
    "args": {
      "External id": 191439,
      "cbid": 251, "correlation": 191439
    }
  },
  {
    "ph": "f", "id": 191439, "pid": 76337, "tid": -914061504, "ts": 1716454224327858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327861, "dur": 0,
    "args": {
      "External id": 191440,
      "cbid": 251, "correlation": 191440
    }
  },
  {
    "ph": "f", "id": 191440, "pid": 76337, "tid": -914061504, "ts": 1716454224327861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224474462, "dur": 6,
    "args": {
      "External id": 191441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191441,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191441, "pid": 5, "tid": 7, "ts": 1716454224474462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327863, "dur": 12,
    "args": {
      "External id": 191441,
      "cbid": 211, "correlation": 191441
    }
  },
  {
    "ph": "s", "id": 191441, "pid": 76337, "tid": -914061504, "ts": 1716454224327863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224474469, "dur": 3,
    "args": {
      "External id": 191443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191443,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 191443, "pid": 5, "tid": 7, "ts": 1716454224474469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327876, "dur": 6,
    "args": {
      "External id": 191443,
      "cbid": 211, "correlation": 191443
    }
  },
  {
    "ph": "s", "id": 191443, "pid": 76337, "tid": -914061504, "ts": 1716454224327876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224474474, "dur": 157,
    "args": {
      "External id": 191464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191464,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 191464, "pid": 5, "tid": 7, "ts": 1716454224474474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224327950, "dur": 12,
    "args": {
      "External id": 191464,
      "cbid": 211, "correlation": 191464
    }
  },
  {
    "ph": "s", "id": 191464, "pid": 76337, "tid": -914061504, "ts": 1716454224327950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328053, "dur": 1,
    "args": {
      "External id": 191482,
      "cbid": 251, "correlation": 191482
    }
  },
  {
    "ph": "f", "id": 191482, "pid": 76337, "tid": -914061504, "ts": 1716454224328053,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224474633, "dur": 106,
    "args": {
      "External id": 191484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191484,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 191484, "pid": 5, "tid": 7, "ts": 1716454224474633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328058, "dur": 14,
    "args": {
      "External id": 191484,
      "cbid": 211, "correlation": 191484
    }
  },
  {
    "ph": "s", "id": 191484, "pid": 76337, "tid": -914061504, "ts": 1716454224328058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224474740, "dur": 35,
    "args": {
      "External id": 191492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191492,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191492, "pid": 5, "tid": 7, "ts": 1716454224474740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328128, "dur": 13,
    "args": {
      "External id": 191492,
      "cbid": 211, "correlation": 191492
    }
  },
  {
    "ph": "s", "id": 191492, "pid": 76337, "tid": -914061504, "ts": 1716454224328128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224474776, "dur": 67,
    "args": {
      "External id": 191500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191500,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191500, "pid": 5, "tid": 7, "ts": 1716454224474776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328170, "dur": 9,
    "args": {
      "External id": 191500,
      "cbid": 211, "correlation": 191500
    }
  },
  {
    "ph": "s", "id": 191500, "pid": 76337, "tid": -914061504, "ts": 1716454224328170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224474844, "dur": 93,
    "args": {
      "External id": 191522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191522,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191522, "pid": 5, "tid": 7, "ts": 1716454224474844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328220, "dur": 10,
    "args": {
      "External id": 191522,
      "cbid": 211, "correlation": 191522
    }
  },
  {
    "ph": "s", "id": 191522, "pid": 76337, "tid": -914061504, "ts": 1716454224328220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328308, "dur": 1,
    "args": {
      "External id": 191538,
      "cbid": 251, "correlation": 191538
    }
  },
  {
    "ph": "f", "id": 191538, "pid": 76337, "tid": -914061504, "ts": 1716454224328308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224474938, "dur": 577,
    "args": {
      "External id": 191540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191540,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 191540, "pid": 5, "tid": 7, "ts": 1716454224474938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328314, "dur": 12,
    "args": {
      "External id": 191540,
      "cbid": 211, "correlation": 191540
    }
  },
  {
    "ph": "s", "id": 191540, "pid": 76337, "tid": -914061504, "ts": 1716454224328314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224475517, "dur": 244,
    "args": {
      "External id": 191548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191548, "pid": 5, "tid": 7, "ts": 1716454224475517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328378, "dur": 13,
    "args": {
      "External id": 191548,
      "cbid": 211, "correlation": 191548
    }
  },
  {
    "ph": "s", "id": 191548, "pid": 76337, "tid": -914061504, "ts": 1716454224328378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224475762, "dur": 253,
    "args": {
      "External id": 191556,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191556,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191556, "pid": 5, "tid": 7, "ts": 1716454224475762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328409, "dur": 8,
    "args": {
      "External id": 191556,
      "cbid": 211, "correlation": 191556
    }
  },
  {
    "ph": "s", "id": 191556, "pid": 76337, "tid": -914061504, "ts": 1716454224328409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328489, "dur": 1,
    "args": {
      "External id": 191572,
      "cbid": 251, "correlation": 191572
    }
  },
  {
    "ph": "f", "id": 191572, "pid": 76337, "tid": -914061504, "ts": 1716454224328489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328495, "dur": 0,
    "args": {
      "External id": 191574,
      "cbid": 251, "correlation": 191574
    }
  },
  {
    "ph": "f", "id": 191574, "pid": 76337, "tid": -914061504, "ts": 1716454224328495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224476016, "dur": 360,
    "args": {
      "External id": 191575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191575,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191575, "pid": 5, "tid": 7, "ts": 1716454224476016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328498, "dur": 13,
    "args": {
      "External id": 191575,
      "cbid": 211, "correlation": 191575
    }
  },
  {
    "ph": "s", "id": 191575, "pid": 76337, "tid": -914061504, "ts": 1716454224328498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224476378, "dur": 50,
    "args": {
      "External id": 191583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191583,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191583, "pid": 5, "tid": 7, "ts": 1716454224476378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328539, "dur": 148,
    "args": {
      "External id": 191583,
      "cbid": 211, "correlation": 191583
    }
  },
  {
    "ph": "s", "id": 191583, "pid": 76337, "tid": -914061504, "ts": 1716454224328539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224476429, "dur": 159,
    "args": {
      "External id": 191594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191594,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191594, "pid": 5, "tid": 7, "ts": 1716454224476429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328745, "dur": 66,
    "args": {
      "External id": 191594,
      "cbid": 211, "correlation": 191594
    }
  },
  {
    "ph": "s", "id": 191594, "pid": 76337, "tid": -914061504, "ts": 1716454224328745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328864, "dur": 0,
    "args": {
      "External id": 191606,
      "cbid": 317, "correlation": 191606
    }
  },
  {
    "ph": "f", "id": 191606, "pid": 76337, "tid": -914061504, "ts": 1716454224328864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328865, "dur": 0,
    "args": {
      "External id": 191607,
      "cbid": 203, "correlation": 191607
    }
  },
  {
    "ph": "f", "id": 191607, "pid": 76337, "tid": -914061504, "ts": 1716454224328865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328865, "dur": 0,
    "args": {
      "External id": 191608,
      "cbid": 205, "correlation": 191608
    }
  },
  {
    "ph": "f", "id": 191608, "pid": 76337, "tid": -914061504, "ts": 1716454224328865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328889, "dur": 1,
    "args": {
      "External id": 191612,
      "cbid": 251, "correlation": 191612
    }
  },
  {
    "ph": "f", "id": 191612, "pid": 76337, "tid": -914061504, "ts": 1716454224328889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328891, "dur": 0,
    "args": {
      "External id": 191613,
      "cbid": 251, "correlation": 191613
    }
  },
  {
    "ph": "f", "id": 191613, "pid": 76337, "tid": -914061504, "ts": 1716454224328891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328892, "dur": 0,
    "args": {
      "External id": 191614,
      "cbid": 251, "correlation": 191614
    }
  },
  {
    "ph": "f", "id": 191614, "pid": 76337, "tid": -914061504, "ts": 1716454224328892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328893, "dur": 0,
    "args": {
      "External id": 191615,
      "cbid": 251, "correlation": 191615
    }
  },
  {
    "ph": "f", "id": 191615, "pid": 76337, "tid": -914061504, "ts": 1716454224328893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328893, "dur": 0,
    "args": {
      "External id": 191616,
      "cbid": 251, "correlation": 191616
    }
  },
  {
    "ph": "f", "id": 191616, "pid": 76337, "tid": -914061504, "ts": 1716454224328893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328894, "dur": 0,
    "args": {
      "External id": 191617,
      "cbid": 251, "correlation": 191617
    }
  },
  {
    "ph": "f", "id": 191617, "pid": 76337, "tid": -914061504, "ts": 1716454224328894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328895, "dur": 0,
    "args": {
      "External id": 191618,
      "cbid": 251, "correlation": 191618
    }
  },
  {
    "ph": "f", "id": 191618, "pid": 76337, "tid": -914061504, "ts": 1716454224328895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328896, "dur": 0,
    "args": {
      "External id": 191619,
      "cbid": 251, "correlation": 191619
    }
  },
  {
    "ph": "f", "id": 191619, "pid": 76337, "tid": -914061504, "ts": 1716454224328896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328897, "dur": 0,
    "args": {
      "External id": 191620,
      "cbid": 251, "correlation": 191620
    }
  },
  {
    "ph": "f", "id": 191620, "pid": 76337, "tid": -914061504, "ts": 1716454224328897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224476589, "dur": 117,
    "args": {
      "External id": 191621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191621,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 191621, "pid": 5, "tid": 7, "ts": 1716454224476589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328899, "dur": 42,
    "args": {
      "External id": 191621,
      "cbid": 211, "correlation": 191621
    }
  },
  {
    "ph": "s", "id": 191621, "pid": 76337, "tid": -914061504, "ts": 1716454224328899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224476707, "dur": 60,
    "args": {
      "External id": 191627,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191627,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191627, "pid": 5, "tid": 7, "ts": 1716454224476707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224328963, "dur": 276,
    "args": {
      "External id": 191627,
      "cbid": 211, "correlation": 191627
    }
  },
  {
    "ph": "s", "id": 191627, "pid": 76337, "tid": -914061504, "ts": 1716454224328963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224476768, "dur": 50,
    "args": {
      "External id": 191635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191635,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191635, "pid": 5, "tid": 7, "ts": 1716454224476768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329264, "dur": 11,
    "args": {
      "External id": 191635,
      "cbid": 211, "correlation": 191635
    }
  },
  {
    "ph": "s", "id": 191635, "pid": 76337, "tid": -914061504, "ts": 1716454224329264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224476819, "dur": 51,
    "args": {
      "External id": 191655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191655,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 191655, "pid": 5, "tid": 7, "ts": 1716454224476819,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329345, "dur": 12,
    "args": {
      "External id": 191655,
      "cbid": 211, "correlation": 191655
    }
  },
  {
    "ph": "s", "id": 191655, "pid": 76337, "tid": -914061504, "ts": 1716454224329345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224476872, "dur": 5,
    "args": {
      "External id": 191667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191667,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191667, "pid": 5, "tid": 7, "ts": 1716454224476872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329368, "dur": 7,
    "args": {
      "External id": 191667,
      "cbid": 211, "correlation": 191667
    }
  },
  {
    "ph": "s", "id": 191667, "pid": 76337, "tid": -914061504, "ts": 1716454224329368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224476878, "dur": 56,
    "args": {
      "External id": 191670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191670,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191670, "pid": 5, "tid": 7, "ts": 1716454224476878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329386, "dur": 101,
    "args": {
      "External id": 191670,
      "cbid": 211, "correlation": 191670
    }
  },
  {
    "ph": "s", "id": 191670, "pid": 76337, "tid": -914061504, "ts": 1716454224329386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224476936, "dur": 37,
    "args": {
      "External id": 191679,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191679,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191679, "pid": 5, "tid": 7, "ts": 1716454224476936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329526, "dur": 11,
    "args": {
      "External id": 191679,
      "cbid": 211, "correlation": 191679
    }
  },
  {
    "ph": "s", "id": 191679, "pid": 76337, "tid": -914061504, "ts": 1716454224329526,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329581, "dur": 0,
    "args": {
      "External id": 191689,
      "cbid": 317, "correlation": 191689
    }
  },
  {
    "ph": "f", "id": 191689, "pid": 76337, "tid": -914061504, "ts": 1716454224329581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329582, "dur": 0,
    "args": {
      "External id": 191690,
      "cbid": 203, "correlation": 191690
    }
  },
  {
    "ph": "f", "id": 191690, "pid": 76337, "tid": -914061504, "ts": 1716454224329582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329583, "dur": 0,
    "args": {
      "External id": 191691,
      "cbid": 205, "correlation": 191691
    }
  },
  {
    "ph": "f", "id": 191691, "pid": 76337, "tid": -914061504, "ts": 1716454224329583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224476974, "dur": 40,
    "args": {
      "External id": 191695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191695,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191695, "pid": 5, "tid": 7, "ts": 1716454224476974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329599, "dur": 12,
    "args": {
      "External id": 191695,
      "cbid": 211, "correlation": 191695
    }
  },
  {
    "ph": "s", "id": 191695, "pid": 76337, "tid": -914061504, "ts": 1716454224329599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224477016, "dur": 3,
    "args": {
      "External id": 191697,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191697,
      "bytes": 46080, "memory bandwidth (GB/s)": 12.307692307692308
    }
  },
  {
    "ph": "f", "id": 191697, "pid": 5, "tid": 7, "ts": 1716454224477016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329615, "dur": 20,
    "args": {
      "External id": 191697,
      "cbid": 51, "correlation": 191697
    }
  },
  {
    "ph": "s", "id": 191697, "pid": 76337, "tid": -914061504, "ts": 1716454224329615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329640, "dur": 2,
    "args": {
      "External id": 191699,
      "cbid": 200, "correlation": 191699
    }
  },
  {
    "ph": "f", "id": 191699, "pid": 76337, "tid": -914061504, "ts": 1716454224329640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329642, "dur": 0,
    "args": {
      "External id": 191700,
      "cbid": 200, "correlation": 191700
    }
  },
  {
    "ph": "f", "id": 191700, "pid": 76337, "tid": -914061504, "ts": 1716454224329642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329643, "dur": 0,
    "args": {
      "External id": 191701,
      "cbid": 200, "correlation": 191701
    }
  },
  {
    "ph": "f", "id": 191701, "pid": 76337, "tid": -914061504, "ts": 1716454224329643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329643, "dur": 0,
    "args": {
      "External id": 191702,
      "cbid": 200, "correlation": 191702
    }
  },
  {
    "ph": "f", "id": 191702, "pid": 76337, "tid": -914061504, "ts": 1716454224329643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329644, "dur": 4,
    "args": {
      "External id": 191703,
      "cbid": 15, "correlation": 191703
    }
  },
  {
    "ph": "f", "id": 191703, "pid": 76337, "tid": -914061504, "ts": 1716454224329644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329649, "dur": 1,
    "args": {
      "External id": 191704,
      "cbid": 251, "correlation": 191704
    }
  },
  {
    "ph": "f", "id": 191704, "pid": 76337, "tid": -914061504, "ts": 1716454224329649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224477021, "dur": 23,
    "args": {
      "External id": 191705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191705,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191705, "pid": 5, "tid": 7, "ts": 1716454224477021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329652, "dur": 8,
    "args": {
      "External id": 191705,
      "cbid": 211, "correlation": 191705
    }
  },
  {
    "ph": "s", "id": 191705, "pid": 76337, "tid": -914061504, "ts": 1716454224329652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224477045, "dur": 4,
    "args": {
      "External id": 191707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191707,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 191707, "pid": 5, "tid": 7, "ts": 1716454224477045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329666, "dur": 6,
    "args": {
      "External id": 191707,
      "cbid": 211, "correlation": 191707
    }
  },
  {
    "ph": "s", "id": 191707, "pid": 76337, "tid": -914061504, "ts": 1716454224329666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329676, "dur": 0,
    "args": {
      "External id": 191708,
      "cbid": 51, "correlation": 191708
    }
  },
  {
    "ph": "s", "id": 191708, "pid": 76337, "tid": -914061504, "ts": 1716454224329676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224477051, "dur": 191,
    "args": {
      "External id": 191709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191709,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191709, "pid": 5, "tid": 7, "ts": 1716454224477051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329677, "dur": 194,
    "args": {
      "External id": 191709,
      "cbid": 211, "correlation": 191709
    }
  },
  {
    "ph": "s", "id": 191709, "pid": 76337, "tid": -914061504, "ts": 1716454224329677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224477243, "dur": 7,
    "args": {
      "External id": 191710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191710,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191710, "pid": 5, "tid": 7, "ts": 1716454224477243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329875, "dur": 5,
    "args": {
      "External id": 191710,
      "cbid": 211, "correlation": 191710
    }
  },
  {
    "ph": "s", "id": 191710, "pid": 76337, "tid": -914061504, "ts": 1716454224329875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224477251, "dur": 5,
    "args": {
      "External id": 191716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191716,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 191716, "pid": 5, "tid": 7, "ts": 1716454224477251,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224329905, "dur": 9,
    "args": {
      "External id": 191716,
      "cbid": 211, "correlation": 191716
    }
  },
  {
    "ph": "s", "id": 191716, "pid": 76337, "tid": -914061504, "ts": 1716454224329905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224477258, "dur": 3,
    "args": {
      "External id": 191724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191724,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191724, "pid": 5, "tid": 7, "ts": 1716454224477258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224331609, "dur": 17,
    "args": {
      "External id": 191724,
      "cbid": 211, "correlation": 191724
    }
  },
  {
    "ph": "s", "id": 191724, "pid": 76337, "tid": -914061504, "ts": 1716454224331609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224477262, "dur": 3,
    "args": {
      "External id": 191732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191732,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191732, "pid": 5, "tid": 7, "ts": 1716454224477262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224331650, "dur": 11,
    "args": {
      "External id": 191732,
      "cbid": 211, "correlation": 191732
    }
  },
  {
    "ph": "s", "id": 191732, "pid": 76337, "tid": -914061504, "ts": 1716454224331650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224477267, "dur": 3,
    "args": {
      "External id": 191740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191740,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191740, "pid": 5, "tid": 7, "ts": 1716454224477267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224331678, "dur": 8,
    "args": {
      "External id": 191740,
      "cbid": 211, "correlation": 191740
    }
  },
  {
    "ph": "s", "id": 191740, "pid": 76337, "tid": -914061504, "ts": 1716454224331678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224477270, "dur": 3,
    "args": {
      "External id": 191749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191749,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191749, "pid": 5, "tid": 7, "ts": 1716454224477270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224331864, "dur": 14,
    "args": {
      "External id": 191749,
      "cbid": 211, "correlation": 191749
    }
  },
  {
    "ph": "s", "id": 191749, "pid": 76337, "tid": -914061504, "ts": 1716454224331864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224477275, "dur": 3,
    "args": {
      "External id": 191758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191758,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191758, "pid": 5, "tid": 7, "ts": 1716454224477275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224331893, "dur": 7,
    "args": {
      "External id": 191758,
      "cbid": 211, "correlation": 191758
    }
  },
  {
    "ph": "s", "id": 191758, "pid": 76337, "tid": -914061504, "ts": 1716454224331893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224477279, "dur": 3,
    "args": {
      "External id": 191766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191766,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191766, "pid": 5, "tid": 7, "ts": 1716454224477279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224331919, "dur": 8,
    "args": {
      "External id": 191766,
      "cbid": 211, "correlation": 191766
    }
  },
  {
    "ph": "s", "id": 191766, "pid": 76337, "tid": -914061504, "ts": 1716454224331919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224477283, "dur": 3,
    "args": {
      "External id": 191774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191774,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191774, "pid": 5, "tid": 7, "ts": 1716454224477283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224332191, "dur": 16,
    "args": {
      "External id": 191774,
      "cbid": 211, "correlation": 191774
    }
  },
  {
    "ph": "s", "id": 191774, "pid": 76337, "tid": -914061504, "ts": 1716454224332191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224477288, "dur": 3,
    "args": {
      "External id": 191782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191782,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191782, "pid": 5, "tid": 7, "ts": 1716454224477288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224332222, "dur": 7,
    "args": {
      "External id": 191782,
      "cbid": 211, "correlation": 191782
    }
  },
  {
    "ph": "s", "id": 191782, "pid": 76337, "tid": -914061504, "ts": 1716454224332222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224477293, "dur": 1,
    "args": {
      "External id": 191792,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191792,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 191792, "pid": 5, "tid": 7, "ts": 1716454224477293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224332288, "dur": 34,
    "args": {
      "External id": 191792,
      "cbid": 41, "correlation": 191792
    }
  },
  {
    "ph": "s", "id": 191792, "pid": 76337, "tid": -914061504, "ts": 1716454224332288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224332323, "dur": 145001,
    "args": {
      "External id": 191793,
      "cbid": 131, "correlation": 191793
    }
  },
  {
    "ph": "f", "id": 191793, "pid": 76337, "tid": -914061504, "ts": 1716454224332323,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224477592, "dur": 3,
    "args": {
      "External id": 191801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191801,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191801, "pid": 5, "tid": 7, "ts": 1716454224477592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224477540, "dur": 54,
    "args": {
      "External id": 191801,
      "cbid": 211, "correlation": 191801
    }
  },
  {
    "ph": "s", "id": 191801, "pid": 76337, "tid": -914061504, "ts": 1716454224477540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224477702, "dur": 3,
    "args": {
      "External id": 191810,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191810,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002380952380952381
    }
  },
  {
    "ph": "f", "id": 191810, "pid": 5, "tid": 7, "ts": 1716454224477702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224477662, "dur": 41,
    "args": {
      "External id": 191810,
      "cbid": 41, "correlation": 191810
    }
  },
  {
    "ph": "s", "id": 191810, "pid": 76337, "tid": -914061504, "ts": 1716454224477662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224477809, "dur": 4,
    "args": {
      "External id": 191820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191820,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 191820, "pid": 5, "tid": 7, "ts": 1716454224477809,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224477792, "dur": 19,
    "args": {
      "External id": 191820,
      "cbid": 211, "correlation": 191820
    }
  },
  {
    "ph": "s", "id": 191820, "pid": 76337, "tid": -914061504, "ts": 1716454224477792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224477903, "dur": 1,
    "args": {
      "External id": 191830,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191830,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 191830, "pid": 5, "tid": 7, "ts": 1716454224477903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224477878, "dur": 24,
    "args": {
      "External id": 191830,
      "cbid": 41, "correlation": 191830
    }
  },
  {
    "ph": "s", "id": 191830, "pid": 76337, "tid": -914061504, "ts": 1716454224477878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224477902, "dur": 8,
    "args": {
      "External id": 191831,
      "cbid": 131, "correlation": 191831
    }
  },
  {
    "ph": "f", "id": 191831, "pid": 76337, "tid": -914061504, "ts": 1716454224477902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224477992, "dur": 3,
    "args": {
      "External id": 191838,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191838,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.425255338904364
    }
  },
  {
    "ph": "f", "id": 191838, "pid": 5, "tid": 7, "ts": 1716454224477992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224477959, "dur": 33,
    "args": {
      "External id": 191838,
      "cbid": 41, "correlation": 191838
    }
  },
  {
    "ph": "s", "id": 191838, "pid": 76337, "tid": -914061504, "ts": 1716454224477959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224478108, "dur": 3,
    "args": {
      "External id": 191857,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191857,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 191857, "pid": 5, "tid": 7, "ts": 1716454224478108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478083, "dur": 25,
    "args": {
      "External id": 191857,
      "cbid": 41, "correlation": 191857
    }
  },
  {
    "ph": "s", "id": 191857, "pid": 76337, "tid": -914061504, "ts": 1716454224478083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224478150, "dur": 3,
    "args": {
      "External id": 191863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191863, "pid": 5, "tid": 7, "ts": 1716454224478150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478139, "dur": 12,
    "args": {
      "External id": 191863,
      "cbid": 211, "correlation": 191863
    }
  },
  {
    "ph": "s", "id": 191863, "pid": 76337, "tid": -914061504, "ts": 1716454224478139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224478164, "dur": 6,
    "args": {
      "External id": 191865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191865,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 191865, "pid": 5, "tid": 7, "ts": 1716454224478164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478154, "dur": 9,
    "args": {
      "External id": 191865,
      "cbid": 211, "correlation": 191865
    }
  },
  {
    "ph": "s", "id": 191865, "pid": 76337, "tid": -914061504, "ts": 1716454224478154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454224478172, "dur": 3,
    "args": {
      "External id": 191867,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191867,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191867, "pid": 5, "tid": 7, "ts": 1716454224478172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478164, "dur": 7,
    "args": {
      "External id": 191867,
      "cbid": 211, "correlation": 191867
    }
  },
  {
    "ph": "s", "id": 191867, "pid": 76337, "tid": -914061504, "ts": 1716454224478164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224478211, "dur": 2,
    "args": {
      "External id": 191875,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191875,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 191875, "pid": 5, "tid": 7, "ts": 1716454224478211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478196, "dur": 14,
    "args": {
      "External id": 191875,
      "cbid": 41, "correlation": 191875
    }
  },
  {
    "ph": "s", "id": 191875, "pid": 76337, "tid": -914061504, "ts": 1716454224478196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224478264, "dur": 3,
    "args": {
      "External id": 191889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191889,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191889, "pid": 5, "tid": 7, "ts": 1716454224478264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478253, "dur": 12,
    "args": {
      "External id": 191889,
      "cbid": 211, "correlation": 191889
    }
  },
  {
    "ph": "s", "id": 191889, "pid": 76337, "tid": -914061504, "ts": 1716454224478253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224478287, "dur": 2,
    "args": {
      "External id": 191903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191903,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191903, "pid": 5, "tid": 7, "ts": 1716454224478287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478279, "dur": 7,
    "args": {
      "External id": 191903,
      "cbid": 211, "correlation": 191903
    }
  },
  {
    "ph": "s", "id": 191903, "pid": 76337, "tid": -914061504, "ts": 1716454224478279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224478329, "dur": 6,
    "args": {
      "External id": 191910,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191910,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191910, "pid": 5, "tid": 7, "ts": 1716454224478329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478318, "dur": 11,
    "args": {
      "External id": 191910,
      "cbid": 211, "correlation": 191910
    }
  },
  {
    "ph": "s", "id": 191910, "pid": 76337, "tid": -914061504, "ts": 1716454224478318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224478339, "dur": 6,
    "args": {
      "External id": 191913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191913,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191913, "pid": 5, "tid": 7, "ts": 1716454224478339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478332, "dur": 7,
    "args": {
      "External id": 191913,
      "cbid": 211, "correlation": 191913
    }
  },
  {
    "ph": "s", "id": 191913, "pid": 76337, "tid": -914061504, "ts": 1716454224478332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224478350, "dur": 3,
    "args": {
      "External id": 191915,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191915,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191915, "pid": 5, "tid": 7, "ts": 1716454224478350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478341, "dur": 8,
    "args": {
      "External id": 191915,
      "cbid": 211, "correlation": 191915
    }
  },
  {
    "ph": "s", "id": 191915, "pid": 76337, "tid": -914061504, "ts": 1716454224478341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224478371, "dur": 2,
    "args": {
      "External id": 191918,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191918,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 191918, "pid": 5, "tid": 7, "ts": 1716454224478371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478357, "dur": 13,
    "args": {
      "External id": 191918,
      "cbid": 41, "correlation": 191918
    }
  },
  {
    "ph": "s", "id": 191918, "pid": 76337, "tid": -914061504, "ts": 1716454224478357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224478429, "dur": 4,
    "args": {
      "External id": 191934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191934,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 191934, "pid": 5, "tid": 7, "ts": 1716454224478429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478416, "dur": 14,
    "args": {
      "External id": 191934,
      "cbid": 211, "correlation": 191934
    }
  },
  {
    "ph": "s", "id": 191934, "pid": 76337, "tid": -914061504, "ts": 1716454224478416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224478453, "dur": 3,
    "args": {
      "External id": 191939,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191939,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003125
    }
  },
  {
    "ph": "f", "id": 191939, "pid": 5, "tid": 7, "ts": 1716454224478453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478434, "dur": 18,
    "args": {
      "External id": 191939,
      "cbid": 41, "correlation": 191939
    }
  },
  {
    "ph": "s", "id": 191939, "pid": 76337, "tid": -914061504, "ts": 1716454224478434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224478518, "dur": 1,
    "args": {
      "External id": 191945,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 191945,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005896226415094339
    }
  },
  {
    "ph": "f", "id": 191945, "pid": 5, "tid": 7, "ts": 1716454224478518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478463, "dur": 64,
    "args": {
      "External id": 191945,
      "cbid": 41, "correlation": 191945
    }
  },
  {
    "ph": "s", "id": 191945, "pid": 76337, "tid": -914061504, "ts": 1716454224478463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478528, "dur": 4,
    "args": {
      "External id": 191946,
      "cbid": 131, "correlation": 191946
    }
  },
  {
    "ph": "f", "id": 191946, "pid": 76337, "tid": -914061504, "ts": 1716454224478528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224478581, "dur": 3,
    "args": {
      "External id": 191954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191954,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191954, "pid": 5, "tid": 7, "ts": 1716454224478581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478568, "dur": 16,
    "args": {
      "External id": 191954,
      "cbid": 211, "correlation": 191954
    }
  },
  {
    "ph": "s", "id": 191954, "pid": 76337, "tid": -914061504, "ts": 1716454224478568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224478618, "dur": 3,
    "args": {
      "External id": 191964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191964,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191964, "pid": 5, "tid": 7, "ts": 1716454224478618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478608, "dur": 8,
    "args": {
      "External id": 191964,
      "cbid": 211, "correlation": 191964
    }
  },
  {
    "ph": "s", "id": 191964, "pid": 76337, "tid": -914061504, "ts": 1716454224478608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224478645, "dur": 3,
    "args": {
      "External id": 191973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191973,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191973, "pid": 5, "tid": 7, "ts": 1716454224478645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478635, "dur": 9,
    "args": {
      "External id": 191973,
      "cbid": 211, "correlation": 191973
    }
  },
  {
    "ph": "s", "id": 191973, "pid": 76337, "tid": -914061504, "ts": 1716454224478635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224478776, "dur": 12,
    "args": {
      "External id": 191983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191983,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191983, "pid": 5, "tid": 7, "ts": 1716454224478776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478762, "dur": 14,
    "args": {
      "External id": 191983,
      "cbid": 211, "correlation": 191983
    }
  },
  {
    "ph": "s", "id": 191983, "pid": 76337, "tid": -914061504, "ts": 1716454224478762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224478817, "dur": 3,
    "args": {
      "External id": 191991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 191991,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 191991, "pid": 5, "tid": 7, "ts": 1716454224478817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478808, "dur": 9,
    "args": {
      "External id": 191991,
      "cbid": 211, "correlation": 191991
    }
  },
  {
    "ph": "s", "id": 191991, "pid": 76337, "tid": -914061504, "ts": 1716454224478808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224478871, "dur": 12,
    "args": {
      "External id": 192001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192001,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192001, "pid": 5, "tid": 7, "ts": 1716454224478871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478854, "dur": 17,
    "args": {
      "External id": 192001,
      "cbid": 211, "correlation": 192001
    }
  },
  {
    "ph": "s", "id": 192001, "pid": 76337, "tid": -914061504, "ts": 1716454224478854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224478907, "dur": 10,
    "args": {
      "External id": 192009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192009,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192009, "pid": 5, "tid": 7, "ts": 1716454224478907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478896, "dur": 10,
    "args": {
      "External id": 192009,
      "cbid": 211, "correlation": 192009
    }
  },
  {
    "ph": "s", "id": 192009, "pid": 76337, "tid": -914061504, "ts": 1716454224478896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224478937, "dur": 3,
    "args": {
      "External id": 192018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192018, "pid": 5, "tid": 7, "ts": 1716454224478937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478927, "dur": 9,
    "args": {
      "External id": 192018,
      "cbid": 211, "correlation": 192018
    }
  },
  {
    "ph": "s", "id": 192018, "pid": 76337, "tid": -914061504, "ts": 1716454224478927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224478962, "dur": 5,
    "args": {
      "External id": 192027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192027,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192027, "pid": 5, "tid": 7, "ts": 1716454224478962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224478953, "dur": 8,
    "args": {
      "External id": 192027,
      "cbid": 211, "correlation": 192027
    }
  },
  {
    "ph": "s", "id": 192027, "pid": 76337, "tid": -914061504, "ts": 1716454224478953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224479019, "dur": 8,
    "args": {
      "External id": 192037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192037,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192037, "pid": 5, "tid": 7, "ts": 1716454224479019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479007, "dur": 11,
    "args": {
      "External id": 192037,
      "cbid": 211, "correlation": 192037
    }
  },
  {
    "ph": "s", "id": 192037, "pid": 76337, "tid": -914061504, "ts": 1716454224479007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224479376, "dur": 3,
    "args": {
      "External id": 192046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192046,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192046, "pid": 5, "tid": 7, "ts": 1716454224479376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479361, "dur": 15,
    "args": {
      "External id": 192046,
      "cbid": 211, "correlation": 192046
    }
  },
  {
    "ph": "s", "id": 192046, "pid": 76337, "tid": -914061504, "ts": 1716454224479361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224479407, "dur": 3,
    "args": {
      "External id": 192054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192054,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192054, "pid": 5, "tid": 7, "ts": 1716454224479407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479397, "dur": 10,
    "args": {
      "External id": 192054,
      "cbid": 211, "correlation": 192054
    }
  },
  {
    "ph": "s", "id": 192054, "pid": 76337, "tid": -914061504, "ts": 1716454224479397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224479459, "dur": 1,
    "args": {
      "External id": 192064,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192064,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 192064, "pid": 5, "tid": 7, "ts": 1716454224479459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479443, "dur": 14,
    "args": {
      "External id": 192064,
      "cbid": 41, "correlation": 192064
    }
  },
  {
    "ph": "s", "id": 192064, "pid": 76337, "tid": -914061504, "ts": 1716454224479443,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479458, "dur": 8,
    "args": {
      "External id": 192065,
      "cbid": 131, "correlation": 192065
    }
  },
  {
    "ph": "f", "id": 192065, "pid": 76337, "tid": -914061504, "ts": 1716454224479458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224479549, "dur": 3,
    "args": {
      "External id": 192073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192073, "pid": 5, "tid": 7, "ts": 1716454224479549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479533, "dur": 16,
    "args": {
      "External id": 192073,
      "cbid": 211, "correlation": 192073
    }
  },
  {
    "ph": "s", "id": 192073, "pid": 76337, "tid": -914061504, "ts": 1716454224479533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224479621, "dur": 3,
    "args": {
      "External id": 192082,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192082,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002577319587628866
    }
  },
  {
    "ph": "f", "id": 192082, "pid": 5, "tid": 7, "ts": 1716454224479621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479603, "dur": 17,
    "args": {
      "External id": 192082,
      "cbid": 41, "correlation": 192082
    }
  },
  {
    "ph": "s", "id": 192082, "pid": 76337, "tid": -914061504, "ts": 1716454224479603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224479692, "dur": 3,
    "args": {
      "External id": 192092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192092,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192092, "pid": 5, "tid": 7, "ts": 1716454224479692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479678, "dur": 14,
    "args": {
      "External id": 192092,
      "cbid": 211, "correlation": 192092
    }
  },
  {
    "ph": "s", "id": 192092, "pid": 76337, "tid": -914061504, "ts": 1716454224479678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224479744, "dur": 1,
    "args": {
      "External id": 192102,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192102,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 192102, "pid": 5, "tid": 7, "ts": 1716454224479744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479730, "dur": 13,
    "args": {
      "External id": 192102,
      "cbid": 41, "correlation": 192102
    }
  },
  {
    "ph": "s", "id": 192102, "pid": 76337, "tid": -914061504, "ts": 1716454224479730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479743, "dur": 7,
    "args": {
      "External id": 192103,
      "cbid": 131, "correlation": 192103
    }
  },
  {
    "ph": "f", "id": 192103, "pid": 76337, "tid": -914061504, "ts": 1716454224479743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224479803, "dur": 3,
    "args": {
      "External id": 192110,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192110,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.670103092783506
    }
  },
  {
    "ph": "f", "id": 192110, "pid": 5, "tid": 7, "ts": 1716454224479803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479783, "dur": 20,
    "args": {
      "External id": 192110,
      "cbid": 41, "correlation": 192110
    }
  },
  {
    "ph": "s", "id": 192110, "pid": 76337, "tid": -914061504, "ts": 1716454224479783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224479851, "dur": 1,
    "args": {
      "External id": 192121,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192121,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 192121, "pid": 5, "tid": 7, "ts": 1716454224479851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479838, "dur": 11,
    "args": {
      "External id": 192121,
      "cbid": 41, "correlation": 192121
    }
  },
  {
    "ph": "s", "id": 192121, "pid": 76337, "tid": -914061504, "ts": 1716454224479838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479850, "dur": 8,
    "args": {
      "External id": 192122,
      "cbid": 131, "correlation": 192122
    }
  },
  {
    "ph": "f", "id": 192122, "pid": 76337, "tid": -914061504, "ts": 1716454224479850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224479901, "dur": 3,
    "args": {
      "External id": 192130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192130,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192130, "pid": 5, "tid": 7, "ts": 1716454224479901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479887, "dur": 14,
    "args": {
      "External id": 192130,
      "cbid": 211, "correlation": 192130
    }
  },
  {
    "ph": "s", "id": 192130, "pid": 76337, "tid": -914061504, "ts": 1716454224479887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224479929, "dur": 3,
    "args": {
      "External id": 192140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192140,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192140, "pid": 5, "tid": 7, "ts": 1716454224479929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479920, "dur": 8,
    "args": {
      "External id": 192140,
      "cbid": 211, "correlation": 192140
    }
  },
  {
    "ph": "s", "id": 192140, "pid": 76337, "tid": -914061504, "ts": 1716454224479920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224479952, "dur": 3,
    "args": {
      "External id": 192149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192149,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192149, "pid": 5, "tid": 7, "ts": 1716454224479952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224479942, "dur": 8,
    "args": {
      "External id": 192149,
      "cbid": 211, "correlation": 192149
    }
  },
  {
    "ph": "s", "id": 192149, "pid": 76337, "tid": -914061504, "ts": 1716454224479942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224480029, "dur": 6,
    "args": {
      "External id": 192157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192157,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192157, "pid": 5, "tid": 7, "ts": 1716454224480029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480016, "dur": 14,
    "args": {
      "External id": 192157,
      "cbid": 211, "correlation": 192157
    }
  },
  {
    "ph": "s", "id": 192157, "pid": 76337, "tid": -914061504, "ts": 1716454224480016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480070, "dur": 3,
    "args": {
      "External id": 192166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192166,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192166, "pid": 5, "tid": 7, "ts": 1716454224480070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480060, "dur": 9,
    "args": {
      "External id": 192166,
      "cbid": 211, "correlation": 192166
    }
  },
  {
    "ph": "s", "id": 192166, "pid": 76337, "tid": -914061504, "ts": 1716454224480060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224480093, "dur": 3,
    "args": {
      "External id": 192175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192175,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192175, "pid": 5, "tid": 7, "ts": 1716454224480093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480084, "dur": 7,
    "args": {
      "External id": 192175,
      "cbid": 211, "correlation": 192175
    }
  },
  {
    "ph": "s", "id": 192175, "pid": 76337, "tid": -914061504, "ts": 1716454224480084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480171, "dur": 3,
    "args": {
      "External id": 192183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192183,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192183, "pid": 5, "tid": 7, "ts": 1716454224480171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480160, "dur": 10,
    "args": {
      "External id": 192183,
      "cbid": 211, "correlation": 192183
    }
  },
  {
    "ph": "s", "id": 192183, "pid": 76337, "tid": -914061504, "ts": 1716454224480160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224480234, "dur": 1,
    "args": {
      "External id": 192191,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192191,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 192191, "pid": 5, "tid": 7, "ts": 1716454224480234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480216, "dur": 28,
    "args": {
      "External id": 192191,
      "cbid": 41, "correlation": 192191
    }
  },
  {
    "ph": "s", "id": 192191, "pid": 76337, "tid": -914061504, "ts": 1716454224480216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480245, "dur": 3,
    "args": {
      "External id": 192192,
      "cbid": 131, "correlation": 192192
    }
  },
  {
    "ph": "f", "id": 192192, "pid": 76337, "tid": -914061504, "ts": 1716454224480245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224480310, "dur": 1,
    "args": {
      "External id": 192202,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192202,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 192202, "pid": 5, "tid": 7, "ts": 1716454224480310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480297, "dur": 11,
    "args": {
      "External id": 192202,
      "cbid": 41, "correlation": 192202
    }
  },
  {
    "ph": "s", "id": 192202, "pid": 76337, "tid": -914061504, "ts": 1716454224480297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480309, "dur": 7,
    "args": {
      "External id": 192203,
      "cbid": 131, "correlation": 192203
    }
  },
  {
    "ph": "f", "id": 192203, "pid": 76337, "tid": -914061504, "ts": 1716454224480309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224480367, "dur": 1,
    "args": {
      "External id": 192212,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192212,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 192212, "pid": 5, "tid": 7, "ts": 1716454224480367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480356, "dur": 9,
    "args": {
      "External id": 192212,
      "cbid": 41, "correlation": 192212
    }
  },
  {
    "ph": "s", "id": 192212, "pid": 76337, "tid": -914061504, "ts": 1716454224480356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480365, "dur": 8,
    "args": {
      "External id": 192213,
      "cbid": 131, "correlation": 192213
    }
  },
  {
    "ph": "f", "id": 192213, "pid": 76337, "tid": -914061504, "ts": 1716454224480365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224480445, "dur": 4,
    "args": {
      "External id": 192220,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192220,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192220, "pid": 5, "tid": 7, "ts": 1716454224480445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480428, "dur": 18,
    "args": {
      "External id": 192220,
      "cbid": 211, "correlation": 192220
    }
  },
  {
    "ph": "s", "id": 192220, "pid": 76337, "tid": -914061504, "ts": 1716454224480428,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454224480486, "dur": 4,
    "args": {
      "External id": 192240,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192240,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192240, "pid": 5, "tid": 7, "ts": 1716454224480486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480473, "dur": 13,
    "args": {
      "External id": 192240,
      "cbid": 211, "correlation": 192240
    }
  },
  {
    "ph": "s", "id": 192240, "pid": 76337, "tid": -914061504, "ts": 1716454224480473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480487, "dur": 0,
    "args": {
      "External id": 192241,
      "cbid": 11, "correlation": 192241
    }
  },
  {
    "ph": "f", "id": 192241, "pid": 76337, "tid": -914061504, "ts": 1716454224480487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480487, "dur": 0,
    "args": {
      "External id": 192242,
      "cbid": 11, "correlation": 192242
    }
  },
  {
    "ph": "f", "id": 192242, "pid": 76337, "tid": -914061504, "ts": 1716454224480487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224480504, "dur": 1,
    "args": {
      "External id": 192245,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192245,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 192245, "pid": 5, "tid": 7, "ts": 1716454224480504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480489, "dur": 24,
    "args": {
      "External id": 192245,
      "cbid": 41, "correlation": 192245
    }
  },
  {
    "ph": "s", "id": 192245, "pid": 76337, "tid": -914061504, "ts": 1716454224480489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480513, "dur": 3,
    "args": {
      "External id": 192246,
      "cbid": 131, "correlation": 192246
    }
  },
  {
    "ph": "f", "id": 192246, "pid": 76337, "tid": -914061504, "ts": 1716454224480513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224480543, "dur": 3,
    "args": {
      "External id": 192270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192270,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192270, "pid": 5, "tid": 7, "ts": 1716454224480543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480533, "dur": 9,
    "args": {
      "External id": 192270,
      "cbid": 211, "correlation": 192270
    }
  },
  {
    "ph": "s", "id": 192270, "pid": 76337, "tid": -914061504, "ts": 1716454224480533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480543, "dur": 0,
    "args": {
      "External id": 192271,
      "cbid": 11, "correlation": 192271
    }
  },
  {
    "ph": "f", "id": 192271, "pid": 76337, "tid": -914061504, "ts": 1716454224480543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480543, "dur": 0,
    "args": {
      "External id": 192272,
      "cbid": 11, "correlation": 192272
    }
  },
  {
    "ph": "f", "id": 192272, "pid": 76337, "tid": -914061504, "ts": 1716454224480543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480545, "dur": 1,
    "args": {
      "External id": 192274,
      "cbid": 200, "correlation": 192274
    }
  },
  {
    "ph": "f", "id": 192274, "pid": 76337, "tid": -914061504, "ts": 1716454224480545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224480555, "dur": 4,
    "args": {
      "External id": 192276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192276,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192276, "pid": 5, "tid": 7, "ts": 1716454224480555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480547, "dur": 8,
    "args": {
      "External id": 192276,
      "cbid": 211, "correlation": 192276
    }
  },
  {
    "ph": "s", "id": 192276, "pid": 76337, "tid": -914061504, "ts": 1716454224480547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480556, "dur": 0,
    "args": {
      "External id": 192277,
      "cbid": 11, "correlation": 192277
    }
  },
  {
    "ph": "f", "id": 192277, "pid": 76337, "tid": -914061504, "ts": 1716454224480556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480556, "dur": 0,
    "args": {
      "External id": 192278,
      "cbid": 11, "correlation": 192278
    }
  },
  {
    "ph": "f", "id": 192278, "pid": 76337, "tid": -914061504, "ts": 1716454224480556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224480602, "dur": 1,
    "args": {
      "External id": 192285,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192285,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 192285, "pid": 5, "tid": 7, "ts": 1716454224480602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480588, "dur": 21,
    "args": {
      "External id": 192285,
      "cbid": 41, "correlation": 192285
    }
  },
  {
    "ph": "s", "id": 192285, "pid": 76337, "tid": -914061504, "ts": 1716454224480588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480611, "dur": 3,
    "args": {
      "External id": 192286,
      "cbid": 131, "correlation": 192286
    }
  },
  {
    "ph": "f", "id": 192286, "pid": 76337, "tid": -914061504, "ts": 1716454224480611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224480663, "dur": 1,
    "args": {
      "External id": 192296,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 192296,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 192296, "pid": 5, "tid": 7, "ts": 1716454224480663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480649, "dur": 12,
    "args": {
      "External id": 192296,
      "cbid": 41, "correlation": 192296
    }
  },
  {
    "ph": "s", "id": 192296, "pid": 76337, "tid": -914061504, "ts": 1716454224480649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480662, "dur": 7,
    "args": {
      "External id": 192297,
      "cbid": 131, "correlation": 192297
    }
  },
  {
    "ph": "f", "id": 192297, "pid": 76337, "tid": -914061504, "ts": 1716454224480662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224480738, "dur": 5,
    "args": {
      "External id": 192304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192304,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192304, "pid": 5, "tid": 7, "ts": 1716454224480738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480723, "dur": 15,
    "args": {
      "External id": 192304,
      "cbid": 211, "correlation": 192304
    }
  },
  {
    "ph": "s", "id": 192304, "pid": 76337, "tid": -914061504, "ts": 1716454224480723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480817, "dur": 3,
    "args": {
      "External id": 192313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192313,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192313, "pid": 5, "tid": 7, "ts": 1716454224480817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480799, "dur": 19,
    "args": {
      "External id": 192313,
      "cbid": 211, "correlation": 192313
    }
  },
  {
    "ph": "s", "id": 192313, "pid": 76337, "tid": -914061504, "ts": 1716454224480799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480855, "dur": 3,
    "args": {
      "External id": 192321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192321,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192321, "pid": 5, "tid": 7, "ts": 1716454224480855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480845, "dur": 10,
    "args": {
      "External id": 192321,
      "cbid": 211, "correlation": 192321
    }
  },
  {
    "ph": "s", "id": 192321, "pid": 76337, "tid": -914061504, "ts": 1716454224480845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480891, "dur": 4,
    "args": {
      "External id": 192329,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192329,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192329, "pid": 5, "tid": 7, "ts": 1716454224480891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480879, "dur": 12,
    "args": {
      "External id": 192329,
      "cbid": 211, "correlation": 192329
    }
  },
  {
    "ph": "s", "id": 192329, "pid": 76337, "tid": -914061504, "ts": 1716454224480879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480923, "dur": 4,
    "args": {
      "External id": 192337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192337,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192337, "pid": 5, "tid": 7, "ts": 1716454224480923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480912, "dur": 11,
    "args": {
      "External id": 192337,
      "cbid": 211, "correlation": 192337
    }
  },
  {
    "ph": "s", "id": 192337, "pid": 76337, "tid": -914061504, "ts": 1716454224480912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224480949, "dur": 3,
    "args": {
      "External id": 192345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192345,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192345, "pid": 5, "tid": 7, "ts": 1716454224480949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480941, "dur": 8,
    "args": {
      "External id": 192345,
      "cbid": 211, "correlation": 192345
    }
  },
  {
    "ph": "s", "id": 192345, "pid": 76337, "tid": -914061504, "ts": 1716454224480941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224480982, "dur": 3,
    "args": {
      "External id": 192353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192353,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192353, "pid": 5, "tid": 7, "ts": 1716454224480982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480966, "dur": 17,
    "args": {
      "External id": 192353,
      "cbid": 211, "correlation": 192353
    }
  },
  {
    "ph": "s", "id": 192353, "pid": 76337, "tid": -914061504, "ts": 1716454224480966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224481007, "dur": 4,
    "args": {
      "External id": 192361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192361,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192361, "pid": 5, "tid": 7, "ts": 1716454224481007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224480998, "dur": 8,
    "args": {
      "External id": 192361,
      "cbid": 211, "correlation": 192361
    }
  },
  {
    "ph": "s", "id": 192361, "pid": 76337, "tid": -914061504, "ts": 1716454224480998,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224481027, "dur": 5,
    "args": {
      "External id": 192369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192369,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192369, "pid": 5, "tid": 7, "ts": 1716454224481027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481018, "dur": 7,
    "args": {
      "External id": 192369,
      "cbid": 211, "correlation": 192369
    }
  },
  {
    "ph": "s", "id": 192369, "pid": 76337, "tid": -914061504, "ts": 1716454224481018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224481046, "dur": 3,
    "args": {
      "External id": 192377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192377,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192377, "pid": 5, "tid": 7, "ts": 1716454224481046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481038, "dur": 7,
    "args": {
      "External id": 192377,
      "cbid": 211, "correlation": 192377
    }
  },
  {
    "ph": "s", "id": 192377, "pid": 76337, "tid": -914061504, "ts": 1716454224481038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224481110, "dur": 3,
    "args": {
      "External id": 192385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192385,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192385, "pid": 5, "tid": 7, "ts": 1716454224481110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481099, "dur": 11,
    "args": {
      "External id": 192385,
      "cbid": 211, "correlation": 192385
    }
  },
  {
    "ph": "s", "id": 192385, "pid": 76337, "tid": -914061504, "ts": 1716454224481099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224481136, "dur": 4,
    "args": {
      "External id": 192393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192393,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192393, "pid": 5, "tid": 7, "ts": 1716454224481136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481128, "dur": 8,
    "args": {
      "External id": 192393,
      "cbid": 211, "correlation": 192393
    }
  },
  {
    "ph": "s", "id": 192393, "pid": 76337, "tid": -914061504, "ts": 1716454224481128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224481159, "dur": 4,
    "args": {
      "External id": 192401,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192401,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192401, "pid": 5, "tid": 7, "ts": 1716454224481159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481150, "dur": 7,
    "args": {
      "External id": 192401,
      "cbid": 211, "correlation": 192401
    }
  },
  {
    "ph": "s", "id": 192401, "pid": 76337, "tid": -914061504, "ts": 1716454224481150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224481178, "dur": 3,
    "args": {
      "External id": 192409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192409,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 192409, "pid": 5, "tid": 7, "ts": 1716454224481178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481170, "dur": 6,
    "args": {
      "External id": 192409,
      "cbid": 211, "correlation": 192409
    }
  },
  {
    "ph": "s", "id": 192409, "pid": 76337, "tid": -914061504, "ts": 1716454224481170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224481618, "dur": 5,
    "args": {
      "External id": 192418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192418,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192418, "pid": 5, "tid": 7, "ts": 1716454224481618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481601, "dur": 17,
    "args": {
      "External id": 192418,
      "cbid": 211, "correlation": 192418
    }
  },
  {
    "ph": "s", "id": 192418, "pid": 76337, "tid": -914061504, "ts": 1716454224481601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224481657, "dur": 5,
    "args": {
      "External id": 192427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192427,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192427, "pid": 5, "tid": 7, "ts": 1716454224481657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481646, "dur": 10,
    "args": {
      "External id": 192427,
      "cbid": 211, "correlation": 192427
    }
  },
  {
    "ph": "s", "id": 192427, "pid": 76337, "tid": -914061504, "ts": 1716454224481646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224481805, "dur": 3,
    "args": {
      "External id": 192443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192443, "pid": 5, "tid": 7, "ts": 1716454224481805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481790, "dur": 15,
    "args": {
      "External id": 192443,
      "cbid": 211, "correlation": 192443
    }
  },
  {
    "ph": "s", "id": 192443, "pid": 76337, "tid": -914061504, "ts": 1716454224481790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224481838, "dur": 3,
    "args": {
      "External id": 192451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192451,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192451, "pid": 5, "tid": 7, "ts": 1716454224481838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481829, "dur": 9,
    "args": {
      "External id": 192451,
      "cbid": 211, "correlation": 192451
    }
  },
  {
    "ph": "s", "id": 192451, "pid": 76337, "tid": -914061504, "ts": 1716454224481829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224481871, "dur": 3,
    "args": {
      "External id": 192459,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192459,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192459, "pid": 5, "tid": 7, "ts": 1716454224481871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481861, "dur": 8,
    "args": {
      "External id": 192459,
      "cbid": 211, "correlation": 192459
    }
  },
  {
    "ph": "s", "id": 192459, "pid": 76337, "tid": -914061504, "ts": 1716454224481861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224481906, "dur": 4,
    "args": {
      "External id": 192467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192467,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192467, "pid": 5, "tid": 7, "ts": 1716454224481906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481895, "dur": 9,
    "args": {
      "External id": 192467,
      "cbid": 211, "correlation": 192467
    }
  },
  {
    "ph": "s", "id": 192467, "pid": 76337, "tid": -914061504, "ts": 1716454224481895,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224481962, "dur": 4,
    "args": {
      "External id": 192479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192479,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192479, "pid": 5, "tid": 7, "ts": 1716454224481962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224481948, "dur": 13,
    "args": {
      "External id": 192479,
      "cbid": 211, "correlation": 192479
    }
  },
  {
    "ph": "s", "id": 192479, "pid": 76337, "tid": -914061504, "ts": 1716454224481948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224482019, "dur": 4,
    "args": {
      "External id": 192490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192490, "pid": 5, "tid": 7, "ts": 1716454224482019,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482006, "dur": 13,
    "args": {
      "External id": 192490,
      "cbid": 211, "correlation": 192490
    }
  },
  {
    "ph": "s", "id": 192490, "pid": 76337, "tid": -914061504, "ts": 1716454224482006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224482052, "dur": 3,
    "args": {
      "External id": 192498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192498, "pid": 5, "tid": 7, "ts": 1716454224482052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482041, "dur": 10,
    "args": {
      "External id": 192498,
      "cbid": 211, "correlation": 192498
    }
  },
  {
    "ph": "s", "id": 192498, "pid": 76337, "tid": -914061504, "ts": 1716454224482041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224482088, "dur": 5,
    "args": {
      "External id": 192506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192506,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192506, "pid": 5, "tid": 7, "ts": 1716454224482088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482077, "dur": 11,
    "args": {
      "External id": 192506,
      "cbid": 211, "correlation": 192506
    }
  },
  {
    "ph": "s", "id": 192506, "pid": 76337, "tid": -914061504, "ts": 1716454224482077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224482121, "dur": 5,
    "args": {
      "External id": 192514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192514,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192514, "pid": 5, "tid": 7, "ts": 1716454224482121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482111, "dur": 10,
    "args": {
      "External id": 192514,
      "cbid": 211, "correlation": 192514
    }
  },
  {
    "ph": "s", "id": 192514, "pid": 76337, "tid": -914061504, "ts": 1716454224482111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224482153, "dur": 4,
    "args": {
      "External id": 192523,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192523,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192523, "pid": 5, "tid": 7, "ts": 1716454224482153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482142, "dur": 11,
    "args": {
      "External id": 192523,
      "cbid": 211, "correlation": 192523
    }
  },
  {
    "ph": "s", "id": 192523, "pid": 76337, "tid": -914061504, "ts": 1716454224482142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224482216, "dur": 4,
    "args": {
      "External id": 192536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192536,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192536, "pid": 5, "tid": 7, "ts": 1716454224482216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482203, "dur": 13,
    "args": {
      "External id": 192536,
      "cbid": 211, "correlation": 192536
    }
  },
  {
    "ph": "s", "id": 192536, "pid": 76337, "tid": -914061504, "ts": 1716454224482203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224482256, "dur": 5,
    "args": {
      "External id": 192546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192546,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 192546, "pid": 5, "tid": 7, "ts": 1716454224482256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482246, "dur": 10,
    "args": {
      "External id": 192546,
      "cbid": 211, "correlation": 192546
    }
  },
  {
    "ph": "s", "id": 192546, "pid": 76337, "tid": -914061504, "ts": 1716454224482246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482396, "dur": 4,
    "args": {
      "External id": 192563,
      "cbid": 251, "correlation": 192563
    }
  },
  {
    "ph": "f", "id": 192563, "pid": 76337, "tid": -914061504, "ts": 1716454224482396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454224482426, "dur": 11,
    "args": {
      "External id": 192565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192565,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192565, "pid": 5, "tid": 7, "ts": 1716454224482426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482410, "dur": 18,
    "args": {
      "External id": 192565,
      "cbid": 211, "correlation": 192565
    }
  },
  {
    "ph": "s", "id": 192565, "pid": 76337, "tid": -914061504, "ts": 1716454224482410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224482499, "dur": 4,
    "args": {
      "External id": 192573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192573,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192573, "pid": 5, "tid": 7, "ts": 1716454224482499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482486, "dur": 12,
    "args": {
      "External id": 192573,
      "cbid": 211, "correlation": 192573
    }
  },
  {
    "ph": "s", "id": 192573, "pid": 76337, "tid": -914061504, "ts": 1716454224482486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482559, "dur": 2,
    "args": {
      "External id": 192589,
      "cbid": 251, "correlation": 192589
    }
  },
  {
    "ph": "f", "id": 192589, "pid": 76337, "tid": -914061504, "ts": 1716454224482559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482565, "dur": 0,
    "args": {
      "External id": 192591,
      "cbid": 251, "correlation": 192591
    }
  },
  {
    "ph": "f", "id": 192591, "pid": 76337, "tid": -914061504, "ts": 1716454224482565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224482582, "dur": 13,
    "args": {
      "External id": 192592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192592,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192592, "pid": 5, "tid": 7, "ts": 1716454224482582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482567, "dur": 14,
    "args": {
      "External id": 192592,
      "cbid": 211, "correlation": 192592
    }
  },
  {
    "ph": "s", "id": 192592, "pid": 76337, "tid": -914061504, "ts": 1716454224482567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224482597, "dur": 5,
    "args": {
      "External id": 192594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192594,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192594, "pid": 5, "tid": 7, "ts": 1716454224482597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482586, "dur": 9,
    "args": {
      "External id": 192594,
      "cbid": 211, "correlation": 192594
    }
  },
  {
    "ph": "s", "id": 192594, "pid": 76337, "tid": -914061504, "ts": 1716454224482586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482711, "dur": 1,
    "args": {
      "External id": 192604,
      "cbid": 317, "correlation": 192604
    }
  },
  {
    "ph": "f", "id": 192604, "pid": 76337, "tid": -914061504, "ts": 1716454224482711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482714, "dur": 1,
    "args": {
      "External id": 192605,
      "cbid": 203, "correlation": 192605
    }
  },
  {
    "ph": "f", "id": 192605, "pid": 76337, "tid": -914061504, "ts": 1716454224482714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482716, "dur": 1,
    "args": {
      "External id": 192606,
      "cbid": 205, "correlation": 192606
    }
  },
  {
    "ph": "f", "id": 192606, "pid": 76337, "tid": -914061504, "ts": 1716454224482716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224482771, "dur": 7,
    "args": {
      "External id": 192610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192610,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192610, "pid": 5, "tid": 7, "ts": 1716454224482771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482756, "dur": 15,
    "args": {
      "External id": 192610,
      "cbid": 211, "correlation": 192610
    }
  },
  {
    "ph": "s", "id": 192610, "pid": 76337, "tid": -914061504, "ts": 1716454224482756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224482782, "dur": 4,
    "args": {
      "External id": 192612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192612,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 192612, "pid": 5, "tid": 7, "ts": 1716454224482782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482775, "dur": 6,
    "args": {
      "External id": 192612,
      "cbid": 211, "correlation": 192612
    }
  },
  {
    "ph": "s", "id": 192612, "pid": 76337, "tid": -914061504, "ts": 1716454224482775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224482803, "dur": 3,
    "args": {
      "External id": 192614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192614, "pid": 5, "tid": 7, "ts": 1716454224482803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482793, "dur": 8,
    "args": {
      "External id": 192614,
      "cbid": 211, "correlation": 192614
    }
  },
  {
    "ph": "s", "id": 192614, "pid": 76337, "tid": -914061504, "ts": 1716454224482793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482808, "dur": 0,
    "args": {
      "External id": 192615,
      "cbid": 51, "correlation": 192615
    }
  },
  {
    "ph": "s", "id": 192615, "pid": 76337, "tid": -914061504, "ts": 1716454224482808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224482818, "dur": 86,
    "args": {
      "External id": 192616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192616,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192616, "pid": 5, "tid": 7, "ts": 1716454224482818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482810, "dur": 7,
    "args": {
      "External id": 192616,
      "cbid": 211, "correlation": 192616
    }
  },
  {
    "ph": "s", "id": 192616, "pid": 76337, "tid": -914061504, "ts": 1716454224482810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224482905, "dur": 60,
    "args": {
      "External id": 192621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192621, "pid": 5, "tid": 7, "ts": 1716454224482905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224482850, "dur": 11,
    "args": {
      "External id": 192621,
      "cbid": 211, "correlation": 192621
    }
  },
  {
    "ph": "s", "id": 192621, "pid": 76337, "tid": -914061504, "ts": 1716454224482850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224484757, "dur": 51,
    "args": {
      "External id": 192641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192641,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 192641, "pid": 5, "tid": 7, "ts": 1716454224484757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484739, "dur": 18,
    "args": {
      "External id": 192641,
      "cbid": 211, "correlation": 192641
    }
  },
  {
    "ph": "s", "id": 192641, "pid": 76337, "tid": -914061504, "ts": 1716454224484739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224484810, "dur": 4,
    "args": {
      "External id": 192653,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192653,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192653, "pid": 5, "tid": 7, "ts": 1716454224484810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484769, "dur": 8,
    "args": {
      "External id": 192653,
      "cbid": 211, "correlation": 192653
    }
  },
  {
    "ph": "s", "id": 192653, "pid": 76337, "tid": -914061504, "ts": 1716454224484769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224484815, "dur": 57,
    "args": {
      "External id": 192656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192656,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192656, "pid": 5, "tid": 7, "ts": 1716454224484815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484794, "dur": 9,
    "args": {
      "External id": 192656,
      "cbid": 211, "correlation": 192656
    }
  },
  {
    "ph": "s", "id": 192656, "pid": 76337, "tid": -914061504, "ts": 1716454224484794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224484873, "dur": 36,
    "args": {
      "External id": 192665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192665,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192665, "pid": 5, "tid": 7, "ts": 1716454224484873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484841, "dur": 11,
    "args": {
      "External id": 192665,
      "cbid": 211, "correlation": 192665
    }
  },
  {
    "ph": "s", "id": 192665, "pid": 76337, "tid": -914061504, "ts": 1716454224484841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484897, "dur": 0,
    "args": {
      "External id": 192675,
      "cbid": 317, "correlation": 192675
    }
  },
  {
    "ph": "f", "id": 192675, "pid": 76337, "tid": -914061504, "ts": 1716454224484897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484898, "dur": 0,
    "args": {
      "External id": 192676,
      "cbid": 203, "correlation": 192676
    }
  },
  {
    "ph": "f", "id": 192676, "pid": 76337, "tid": -914061504, "ts": 1716454224484898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484899, "dur": 0,
    "args": {
      "External id": 192677,
      "cbid": 205, "correlation": 192677
    }
  },
  {
    "ph": "f", "id": 192677, "pid": 76337, "tid": -914061504, "ts": 1716454224484899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224484930, "dur": 40,
    "args": {
      "External id": 192681,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192681,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192681, "pid": 5, "tid": 7, "ts": 1716454224484930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484918, "dur": 12,
    "args": {
      "External id": 192681,
      "cbid": 211, "correlation": 192681
    }
  },
  {
    "ph": "s", "id": 192681, "pid": 76337, "tid": -914061504, "ts": 1716454224484918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224484972, "dur": 15,
    "args": {
      "External id": 192683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192683,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192683, "pid": 5, "tid": 7, "ts": 1716454224484972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484932, "dur": 6,
    "args": {
      "External id": 192683,
      "cbid": 211, "correlation": 192683
    }
  },
  {
    "ph": "s", "id": 192683, "pid": 76337, "tid": -914061504, "ts": 1716454224484932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224484988, "dur": 3,
    "args": {
      "External id": 192685,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192685,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192685, "pid": 5, "tid": 7, "ts": 1716454224484988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484945, "dur": 7,
    "args": {
      "External id": 192685,
      "cbid": 211, "correlation": 192685
    }
  },
  {
    "ph": "s", "id": 192685, "pid": 76337, "tid": -914061504, "ts": 1716454224484945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484956, "dur": 0,
    "args": {
      "External id": 192686,
      "cbid": 51, "correlation": 192686
    }
  },
  {
    "ph": "s", "id": 192686, "pid": 76337, "tid": -914061504, "ts": 1716454224484956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224484993, "dur": 705,
    "args": {
      "External id": 192687,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192687,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192687, "pid": 5, "tid": 7, "ts": 1716454224484993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484957, "dur": 6,
    "args": {
      "External id": 192687,
      "cbid": 211, "correlation": 192687
    }
  },
  {
    "ph": "s", "id": 192687, "pid": 76337, "tid": -914061504, "ts": 1716454224484957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224485699, "dur": 60,
    "args": {
      "External id": 192692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192692,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192692, "pid": 5, "tid": 7, "ts": 1716454224485699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224484994, "dur": 11,
    "args": {
      "External id": 192692,
      "cbid": 211, "correlation": 192692
    }
  },
  {
    "ph": "s", "id": 192692, "pid": 76337, "tid": -914061504, "ts": 1716454224484994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224485760, "dur": 4,
    "args": {
      "External id": 192700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192700,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192700, "pid": 5, "tid": 7, "ts": 1716454224485760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485040, "dur": 9,
    "args": {
      "External id": 192700,
      "cbid": 211, "correlation": 192700
    }
  },
  {
    "ph": "s", "id": 192700, "pid": 76337, "tid": -914061504, "ts": 1716454224485040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485107, "dur": 2,
    "args": {
      "External id": 192716,
      "cbid": 251, "correlation": 192716
    }
  },
  {
    "ph": "f", "id": 192716, "pid": 76337, "tid": -914061504, "ts": 1716454224485107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485113, "dur": 0,
    "args": {
      "External id": 192718,
      "cbid": 251, "correlation": 192718
    }
  },
  {
    "ph": "f", "id": 192718, "pid": 76337, "tid": -914061504, "ts": 1716454224485113,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224485765, "dur": 9,
    "args": {
      "External id": 192719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192719,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 192719, "pid": 5, "tid": 7, "ts": 1716454224485765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485115, "dur": 11,
    "args": {
      "External id": 192719,
      "cbid": 211, "correlation": 192719
    }
  },
  {
    "ph": "s", "id": 192719, "pid": 76337, "tid": -914061504, "ts": 1716454224485115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224485775, "dur": 4,
    "args": {
      "External id": 192721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192721,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 192721, "pid": 5, "tid": 7, "ts": 1716454224485775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485128, "dur": 6,
    "args": {
      "External id": 192721,
      "cbid": 211, "correlation": 192721
    }
  },
  {
    "ph": "s", "id": 192721, "pid": 76337, "tid": -914061504, "ts": 1716454224485128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224485780, "dur": 54,
    "args": {
      "External id": 192731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192731, "pid": 5, "tid": 7, "ts": 1716454224485780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485186, "dur": 12,
    "args": {
      "External id": 192731,
      "cbid": 211, "correlation": 192731
    }
  },
  {
    "ph": "s", "id": 192731, "pid": 76337, "tid": -914061504, "ts": 1716454224485186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224485836, "dur": 51,
    "args": {
      "External id": 192751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192751,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 192751, "pid": 5, "tid": 7, "ts": 1716454224485836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485252, "dur": 11,
    "args": {
      "External id": 192751,
      "cbid": 211, "correlation": 192751
    }
  },
  {
    "ph": "s", "id": 192751, "pid": 76337, "tid": -914061504, "ts": 1716454224485252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224485888, "dur": 4,
    "args": {
      "External id": 192763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192763,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192763, "pid": 5, "tid": 7, "ts": 1716454224485888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485273, "dur": 6,
    "args": {
      "External id": 192763,
      "cbid": 211, "correlation": 192763
    }
  },
  {
    "ph": "s", "id": 192763, "pid": 76337, "tid": -914061504, "ts": 1716454224485273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224485894, "dur": 56,
    "args": {
      "External id": 192766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192766,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192766, "pid": 5, "tid": 7, "ts": 1716454224485894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485290, "dur": 7,
    "args": {
      "External id": 192766,
      "cbid": 211, "correlation": 192766
    }
  },
  {
    "ph": "s", "id": 192766, "pid": 76337, "tid": -914061504, "ts": 1716454224485290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224485951, "dur": 37,
    "args": {
      "External id": 192775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192775,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192775, "pid": 5, "tid": 7, "ts": 1716454224485951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485331, "dur": 10,
    "args": {
      "External id": 192775,
      "cbid": 211, "correlation": 192775
    }
  },
  {
    "ph": "s", "id": 192775, "pid": 76337, "tid": -914061504, "ts": 1716454224485331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485407, "dur": 0,
    "args": {
      "External id": 192785,
      "cbid": 317, "correlation": 192785
    }
  },
  {
    "ph": "f", "id": 192785, "pid": 76337, "tid": -914061504, "ts": 1716454224485407,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485408, "dur": 0,
    "args": {
      "External id": 192786,
      "cbid": 203, "correlation": 192786
    }
  },
  {
    "ph": "f", "id": 192786, "pid": 76337, "tid": -914061504, "ts": 1716454224485408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485408, "dur": 0,
    "args": {
      "External id": 192787,
      "cbid": 205, "correlation": 192787
    }
  },
  {
    "ph": "f", "id": 192787, "pid": 76337, "tid": -914061504, "ts": 1716454224485408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224485989, "dur": 39,
    "args": {
      "External id": 192791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192791,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192791, "pid": 5, "tid": 7, "ts": 1716454224485989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485422, "dur": 12,
    "args": {
      "External id": 192791,
      "cbid": 211, "correlation": 192791
    }
  },
  {
    "ph": "s", "id": 192791, "pid": 76337, "tid": -914061504, "ts": 1716454224485422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224486029, "dur": 14,
    "args": {
      "External id": 192793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192793,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192793, "pid": 5, "tid": 7, "ts": 1716454224486029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485437, "dur": 5,
    "args": {
      "External id": 192793,
      "cbid": 211, "correlation": 192793
    }
  },
  {
    "ph": "s", "id": 192793, "pid": 76337, "tid": -914061504, "ts": 1716454224485437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224486045, "dur": 3,
    "args": {
      "External id": 192795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192795,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192795, "pid": 5, "tid": 7, "ts": 1716454224486045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485446, "dur": 6,
    "args": {
      "External id": 192795,
      "cbid": 211, "correlation": 192795
    }
  },
  {
    "ph": "s", "id": 192795, "pid": 76337, "tid": -914061504, "ts": 1716454224485446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485455, "dur": 0,
    "args": {
      "External id": 192796,
      "cbid": 51, "correlation": 192796
    }
  },
  {
    "ph": "s", "id": 192796, "pid": 76337, "tid": -914061504, "ts": 1716454224485455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224486050, "dur": 697,
    "args": {
      "External id": 192797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192797,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192797, "pid": 5, "tid": 7, "ts": 1716454224486050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485456, "dur": 5,
    "args": {
      "External id": 192797,
      "cbid": 211, "correlation": 192797
    }
  },
  {
    "ph": "s", "id": 192797, "pid": 76337, "tid": -914061504, "ts": 1716454224485456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224486749, "dur": 60,
    "args": {
      "External id": 192802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192802,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192802, "pid": 5, "tid": 7, "ts": 1716454224486749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485484, "dur": 9,
    "args": {
      "External id": 192802,
      "cbid": 211, "correlation": 192802
    }
  },
  {
    "ph": "s", "id": 192802, "pid": 76337, "tid": -914061504, "ts": 1716454224485484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224486810, "dur": 50,
    "args": {
      "External id": 192810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192810,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192810, "pid": 5, "tid": 7, "ts": 1716454224486810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485517, "dur": 8,
    "args": {
      "External id": 192810,
      "cbid": 211, "correlation": 192810
    }
  },
  {
    "ph": "s", "id": 192810, "pid": 76337, "tid": -914061504, "ts": 1716454224485517,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224486861, "dur": 36,
    "args": {
      "External id": 192818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192818,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192818, "pid": 5, "tid": 7, "ts": 1716454224486861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485547, "dur": 10,
    "args": {
      "External id": 192818,
      "cbid": 211, "correlation": 192818
    }
  },
  {
    "ph": "s", "id": 192818, "pid": 76337, "tid": -914061504, "ts": 1716454224485547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224486898, "dur": 52,
    "args": {
      "External id": 192838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192838,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 192838, "pid": 5, "tid": 7, "ts": 1716454224486898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485635, "dur": 13,
    "args": {
      "External id": 192838,
      "cbid": 211, "correlation": 192838
    }
  },
  {
    "ph": "s", "id": 192838, "pid": 76337, "tid": -914061504, "ts": 1716454224485635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224486952, "dur": 4,
    "args": {
      "External id": 192850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192850,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 192850, "pid": 5, "tid": 7, "ts": 1716454224486952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485659, "dur": 7,
    "args": {
      "External id": 192850,
      "cbid": 211, "correlation": 192850
    }
  },
  {
    "ph": "s", "id": 192850, "pid": 76337, "tid": -914061504, "ts": 1716454224485659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224486957, "dur": 55,
    "args": {
      "External id": 192853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192853, "pid": 5, "tid": 7, "ts": 1716454224486957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485676, "dur": 7,
    "args": {
      "External id": 192853,
      "cbid": 211, "correlation": 192853
    }
  },
  {
    "ph": "s", "id": 192853, "pid": 76337, "tid": -914061504, "ts": 1716454224485676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485735, "dur": 0,
    "args": {
      "External id": 192864,
      "cbid": 317, "correlation": 192864
    }
  },
  {
    "ph": "f", "id": 192864, "pid": 76337, "tid": -914061504, "ts": 1716454224485735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485736, "dur": 0,
    "args": {
      "External id": 192865,
      "cbid": 203, "correlation": 192865
    }
  },
  {
    "ph": "f", "id": 192865, "pid": 76337, "tid": -914061504, "ts": 1716454224485736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485736, "dur": 0,
    "args": {
      "External id": 192866,
      "cbid": 205, "correlation": 192866
    }
  },
  {
    "ph": "f", "id": 192866, "pid": 76337, "tid": -914061504, "ts": 1716454224485736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485772, "dur": 2,
    "args": {
      "External id": 192870,
      "cbid": 251, "correlation": 192870
    }
  },
  {
    "ph": "f", "id": 192870, "pid": 76337, "tid": -914061504, "ts": 1716454224485772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485775, "dur": 1,
    "args": {
      "External id": 192871,
      "cbid": 251, "correlation": 192871
    }
  },
  {
    "ph": "f", "id": 192871, "pid": 76337, "tid": -914061504, "ts": 1716454224485775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485777, "dur": 2,
    "args": {
      "External id": 192872,
      "cbid": 251, "correlation": 192872
    }
  },
  {
    "ph": "f", "id": 192872, "pid": 76337, "tid": -914061504, "ts": 1716454224485777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485780, "dur": 1,
    "args": {
      "External id": 192873,
      "cbid": 251, "correlation": 192873
    }
  },
  {
    "ph": "f", "id": 192873, "pid": 76337, "tid": -914061504, "ts": 1716454224485780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485782, "dur": 1,
    "args": {
      "External id": 192874,
      "cbid": 251, "correlation": 192874
    }
  },
  {
    "ph": "f", "id": 192874, "pid": 76337, "tid": -914061504, "ts": 1716454224485782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485784, "dur": 1,
    "args": {
      "External id": 192875,
      "cbid": 251, "correlation": 192875
    }
  },
  {
    "ph": "f", "id": 192875, "pid": 76337, "tid": -914061504, "ts": 1716454224485784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485786, "dur": 1,
    "args": {
      "External id": 192876,
      "cbid": 251, "correlation": 192876
    }
  },
  {
    "ph": "f", "id": 192876, "pid": 76337, "tid": -914061504, "ts": 1716454224485786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485788, "dur": 1,
    "args": {
      "External id": 192877,
      "cbid": 251, "correlation": 192877
    }
  },
  {
    "ph": "f", "id": 192877, "pid": 76337, "tid": -914061504, "ts": 1716454224485788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485791, "dur": 0,
    "args": {
      "External id": 192878,
      "cbid": 251, "correlation": 192878
    }
  },
  {
    "ph": "f", "id": 192878, "pid": 76337, "tid": -914061504, "ts": 1716454224485791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224487013, "dur": 116,
    "args": {
      "External id": 192879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192879,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 192879, "pid": 5, "tid": 7, "ts": 1716454224487013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485795, "dur": 16,
    "args": {
      "External id": 192879,
      "cbid": 211, "correlation": 192879
    }
  },
  {
    "ph": "s", "id": 192879, "pid": 76337, "tid": -914061504, "ts": 1716454224485795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224487131, "dur": 60,
    "args": {
      "External id": 192885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192885,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192885, "pid": 5, "tid": 7, "ts": 1716454224487131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485834, "dur": 9,
    "args": {
      "External id": 192885,
      "cbid": 211, "correlation": 192885
    }
  },
  {
    "ph": "s", "id": 192885, "pid": 76337, "tid": -914061504, "ts": 1716454224485834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224487192, "dur": 498,
    "args": {
      "External id": 192894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192894,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192894, "pid": 5, "tid": 7, "ts": 1716454224487192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224485935, "dur": 15,
    "args": {
      "External id": 192894,
      "cbid": 211, "correlation": 192894
    }
  },
  {
    "ph": "s", "id": 192894, "pid": 76337, "tid": -914061504, "ts": 1716454224485935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224487691, "dur": 182,
    "args": {
      "External id": 192916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192916,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 192916, "pid": 5, "tid": 7, "ts": 1716454224487691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486020, "dur": 13,
    "args": {
      "External id": 192916,
      "cbid": 211, "correlation": 192916
    }
  },
  {
    "ph": "s", "id": 192916, "pid": 76337, "tid": -914061504, "ts": 1716454224486020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486146, "dur": 2,
    "args": {
      "External id": 192927,
      "cbid": 251, "correlation": 192927
    }
  },
  {
    "ph": "f", "id": 192927, "pid": 76337, "tid": -914061504, "ts": 1716454224486146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224487875, "dur": 198,
    "args": {
      "External id": 192928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192928,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192928, "pid": 5, "tid": 7, "ts": 1716454224487875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486153, "dur": 14,
    "args": {
      "External id": 192928,
      "cbid": 211, "correlation": 192928
    }
  },
  {
    "ph": "s", "id": 192928, "pid": 76337, "tid": -914061504, "ts": 1716454224486153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486230, "dur": 1,
    "args": {
      "External id": 192939,
      "cbid": 251, "correlation": 192939
    }
  },
  {
    "ph": "f", "id": 192939, "pid": 76337, "tid": -914061504, "ts": 1716454224486230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224488074, "dur": 186,
    "args": {
      "External id": 192940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192940,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192940, "pid": 5, "tid": 7, "ts": 1716454224488074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486234, "dur": 13,
    "args": {
      "External id": 192940,
      "cbid": 211, "correlation": 192940
    }
  },
  {
    "ph": "s", "id": 192940, "pid": 76337, "tid": -914061504, "ts": 1716454224486234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486300, "dur": 1,
    "args": {
      "External id": 192951,
      "cbid": 251, "correlation": 192951
    }
  },
  {
    "ph": "f", "id": 192951, "pid": 76337, "tid": -914061504, "ts": 1716454224486300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224488261, "dur": 187,
    "args": {
      "External id": 192952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192952,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192952, "pid": 5, "tid": 7, "ts": 1716454224488261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486304, "dur": 11,
    "args": {
      "External id": 192952,
      "cbid": 211, "correlation": 192952
    }
  },
  {
    "ph": "s", "id": 192952, "pid": 76337, "tid": -914061504, "ts": 1716454224486304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224488450, "dur": 18689,
    "args": {
      "External id": 192973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192973,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 192973, "pid": 5, "tid": 7, "ts": 1716454224488450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486423, "dur": 16,
    "args": {
      "External id": 192973,
      "cbid": 211, "correlation": 192973
    }
  },
  {
    "ph": "s", "id": 192973, "pid": 76337, "tid": -914061504, "ts": 1716454224486423,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486544, "dur": 2,
    "args": {
      "External id": 192991,
      "cbid": 251, "correlation": 192991
    }
  },
  {
    "ph": "f", "id": 192991, "pid": 76337, "tid": -914061504, "ts": 1716454224486544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224507140, "dur": 200,
    "args": {
      "External id": 192993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 192993,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 192993, "pid": 5, "tid": 7, "ts": 1716454224507140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486551, "dur": 13,
    "args": {
      "External id": 192993,
      "cbid": 211, "correlation": 192993
    }
  },
  {
    "ph": "s", "id": 192993, "pid": 76337, "tid": -914061504, "ts": 1716454224486551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224507342, "dur": 67,
    "args": {
      "External id": 193001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193001,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193001, "pid": 5, "tid": 7, "ts": 1716454224507342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486624, "dur": 13,
    "args": {
      "External id": 193001,
      "cbid": 211, "correlation": 193001
    }
  },
  {
    "ph": "s", "id": 193001, "pid": 76337, "tid": -914061504, "ts": 1716454224486624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224507410, "dur": 96,
    "args": {
      "External id": 193009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193009,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193009, "pid": 5, "tid": 7, "ts": 1716454224507410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486665, "dur": 9,
    "args": {
      "External id": 193009,
      "cbid": 211, "correlation": 193009
    }
  },
  {
    "ph": "s", "id": 193009, "pid": 76337, "tid": -914061504, "ts": 1716454224486665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224507507, "dur": 55,
    "args": {
      "External id": 193020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193020,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193020, "pid": 5, "tid": 7, "ts": 1716454224507507,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486756, "dur": 14,
    "args": {
      "External id": 193020,
      "cbid": 211, "correlation": 193020
    }
  },
  {
    "ph": "s", "id": 193020, "pid": 76337, "tid": -914061504, "ts": 1716454224486756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224507564, "dur": 92,
    "args": {
      "External id": 193042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193042,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193042, "pid": 5, "tid": 7, "ts": 1716454224507564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486792, "dur": 8,
    "args": {
      "External id": 193042,
      "cbid": 211, "correlation": 193042
    }
  },
  {
    "ph": "s", "id": 193042, "pid": 76337, "tid": -914061504, "ts": 1716454224486792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486877, "dur": 1,
    "args": {
      "External id": 193053,
      "cbid": 251, "correlation": 193053
    }
  },
  {
    "ph": "f", "id": 193053, "pid": 76337, "tid": -914061504, "ts": 1716454224486877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224507657, "dur": 105,
    "args": {
      "External id": 193054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193054,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193054, "pid": 5, "tid": 7, "ts": 1716454224507657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486882, "dur": 13,
    "args": {
      "External id": 193054,
      "cbid": 211, "correlation": 193054
    }
  },
  {
    "ph": "s", "id": 193054, "pid": 76337, "tid": -914061504, "ts": 1716454224486882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486964, "dur": 1,
    "args": {
      "External id": 193065,
      "cbid": 251, "correlation": 193065
    }
  },
  {
    "ph": "f", "id": 193065, "pid": 76337, "tid": -914061504, "ts": 1716454224486964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486969, "dur": 0,
    "args": {
      "External id": 193066,
      "cbid": 251, "correlation": 193066
    }
  },
  {
    "ph": "f", "id": 193066, "pid": 76337, "tid": -914061504, "ts": 1716454224486969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224507764, "dur": 10,
    "args": {
      "External id": 193067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193067,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 193067, "pid": 5, "tid": 7, "ts": 1716454224507764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224486971, "dur": 26,
    "args": {
      "External id": 193067,
      "cbid": 211, "correlation": 193067
    }
  },
  {
    "ph": "s", "id": 193067, "pid": 76337, "tid": -914061504, "ts": 1716454224486971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224507775, "dur": 5,
    "args": {
      "External id": 193069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193069,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 193069, "pid": 5, "tid": 7, "ts": 1716454224507775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487001, "dur": 8,
    "args": {
      "External id": 193069,
      "cbid": 211, "correlation": 193069
    }
  },
  {
    "ph": "s", "id": 193069, "pid": 76337, "tid": -914061504, "ts": 1716454224487001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487067, "dur": 1,
    "args": {
      "External id": 193080,
      "cbid": 251, "correlation": 193080
    }
  },
  {
    "ph": "f", "id": 193080, "pid": 76337, "tid": -914061504, "ts": 1716454224487067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487070, "dur": 0,
    "args": {
      "External id": 193081,
      "cbid": 251, "correlation": 193081
    }
  },
  {
    "ph": "f", "id": 193081, "pid": 76337, "tid": -914061504, "ts": 1716454224487070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224507782, "dur": 6,
    "args": {
      "External id": 193082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193082,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 193082, "pid": 5, "tid": 7, "ts": 1716454224507782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487072, "dur": 13,
    "args": {
      "External id": 193082,
      "cbid": 211, "correlation": 193082
    }
  },
  {
    "ph": "s", "id": 193082, "pid": 76337, "tid": -914061504, "ts": 1716454224487072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224507789, "dur": 3,
    "args": {
      "External id": 193084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193084,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 193084, "pid": 5, "tid": 7, "ts": 1716454224507789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487086, "dur": 5,
    "args": {
      "External id": 193084,
      "cbid": 211, "correlation": 193084
    }
  },
  {
    "ph": "s", "id": 193084, "pid": 76337, "tid": -914061504, "ts": 1716454224487086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224507794, "dur": 156,
    "args": {
      "External id": 193105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193105,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193105, "pid": 5, "tid": 7, "ts": 1716454224507794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487159, "dur": 12,
    "args": {
      "External id": 193105,
      "cbid": 211, "correlation": 193105
    }
  },
  {
    "ph": "s", "id": 193105, "pid": 76337, "tid": -914061504, "ts": 1716454224487159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487257, "dur": 2,
    "args": {
      "External id": 193123,
      "cbid": 251, "correlation": 193123
    }
  },
  {
    "ph": "f", "id": 193123, "pid": 76337, "tid": -914061504, "ts": 1716454224487257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224507952, "dur": 106,
    "args": {
      "External id": 193125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193125,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193125, "pid": 5, "tid": 7, "ts": 1716454224507952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487263, "dur": 15,
    "args": {
      "External id": 193125,
      "cbid": 211, "correlation": 193125
    }
  },
  {
    "ph": "s", "id": 193125, "pid": 76337, "tid": -914061504, "ts": 1716454224487263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224508060, "dur": 34,
    "args": {
      "External id": 193133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193133,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193133, "pid": 5, "tid": 7, "ts": 1716454224508060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487339, "dur": 12,
    "args": {
      "External id": 193133,
      "cbid": 211, "correlation": 193133
    }
  },
  {
    "ph": "s", "id": 193133, "pid": 76337, "tid": -914061504, "ts": 1716454224487339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224508095, "dur": 68,
    "args": {
      "External id": 193141,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193141,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193141, "pid": 5, "tid": 7, "ts": 1716454224508095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487381, "dur": 9,
    "args": {
      "External id": 193141,
      "cbid": 211, "correlation": 193141
    }
  },
  {
    "ph": "s", "id": 193141, "pid": 76337, "tid": -914061504, "ts": 1716454224487381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224508165, "dur": 92,
    "args": {
      "External id": 193163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193163,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193163, "pid": 5, "tid": 7, "ts": 1716454224508165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487432, "dur": 10,
    "args": {
      "External id": 193163,
      "cbid": 211, "correlation": 193163
    }
  },
  {
    "ph": "s", "id": 193163, "pid": 76337, "tid": -914061504, "ts": 1716454224487432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487523, "dur": 1,
    "args": {
      "External id": 193179,
      "cbid": 251, "correlation": 193179
    }
  },
  {
    "ph": "f", "id": 193179, "pid": 76337, "tid": -914061504, "ts": 1716454224487523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224508259, "dur": 576,
    "args": {
      "External id": 193181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193181,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193181, "pid": 5, "tid": 7, "ts": 1716454224508259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487529, "dur": 13,
    "args": {
      "External id": 193181,
      "cbid": 211, "correlation": 193181
    }
  },
  {
    "ph": "s", "id": 193181, "pid": 76337, "tid": -914061504, "ts": 1716454224487529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224508836, "dur": 245,
    "args": {
      "External id": 193189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193189, "pid": 5, "tid": 7, "ts": 1716454224508836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487611, "dur": 15,
    "args": {
      "External id": 193189,
      "cbid": 211, "correlation": 193189
    }
  },
  {
    "ph": "s", "id": 193189, "pid": 76337, "tid": -914061504, "ts": 1716454224487611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224509083, "dur": 250,
    "args": {
      "External id": 193197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193197,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193197, "pid": 5, "tid": 7, "ts": 1716454224509083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487648, "dur": 8,
    "args": {
      "External id": 193197,
      "cbid": 211, "correlation": 193197
    }
  },
  {
    "ph": "s", "id": 193197, "pid": 76337, "tid": -914061504, "ts": 1716454224487648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487733, "dur": 2,
    "args": {
      "External id": 193213,
      "cbid": 251, "correlation": 193213
    }
  },
  {
    "ph": "f", "id": 193213, "pid": 76337, "tid": -914061504, "ts": 1716454224487733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487739, "dur": 0,
    "args": {
      "External id": 193215,
      "cbid": 251, "correlation": 193215
    }
  },
  {
    "ph": "f", "id": 193215, "pid": 76337, "tid": -914061504, "ts": 1716454224487739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224509334, "dur": 355,
    "args": {
      "External id": 193216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193216,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 193216, "pid": 5, "tid": 7, "ts": 1716454224509334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487744, "dur": 13,
    "args": {
      "External id": 193216,
      "cbid": 211, "correlation": 193216
    }
  },
  {
    "ph": "s", "id": 193216, "pid": 76337, "tid": -914061504, "ts": 1716454224487744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224509691, "dur": 49,
    "args": {
      "External id": 193224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193224,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193224, "pid": 5, "tid": 7, "ts": 1716454224509691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487787, "dur": 10,
    "args": {
      "External id": 193224,
      "cbid": 211, "correlation": 193224
    }
  },
  {
    "ph": "s", "id": 193224, "pid": 76337, "tid": -914061504, "ts": 1716454224487787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224509742, "dur": 157,
    "args": {
      "External id": 193235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193235,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193235, "pid": 5, "tid": 7, "ts": 1716454224509742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487861, "dur": 13,
    "args": {
      "External id": 193235,
      "cbid": 211, "correlation": 193235
    }
  },
  {
    "ph": "s", "id": 193235, "pid": 76337, "tid": -914061504, "ts": 1716454224487861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487928, "dur": 0,
    "args": {
      "External id": 193247,
      "cbid": 317, "correlation": 193247
    }
  },
  {
    "ph": "f", "id": 193247, "pid": 76337, "tid": -914061504, "ts": 1716454224487928,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487929, "dur": 0,
    "args": {
      "External id": 193248,
      "cbid": 203, "correlation": 193248
    }
  },
  {
    "ph": "f", "id": 193248, "pid": 76337, "tid": -914061504, "ts": 1716454224487929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487930, "dur": 0,
    "args": {
      "External id": 193249,
      "cbid": 205, "correlation": 193249
    }
  },
  {
    "ph": "f", "id": 193249, "pid": 76337, "tid": -914061504, "ts": 1716454224487930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487952, "dur": 1,
    "args": {
      "External id": 193253,
      "cbid": 251, "correlation": 193253
    }
  },
  {
    "ph": "f", "id": 193253, "pid": 76337, "tid": -914061504, "ts": 1716454224487952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487954, "dur": 0,
    "args": {
      "External id": 193254,
      "cbid": 251, "correlation": 193254
    }
  },
  {
    "ph": "f", "id": 193254, "pid": 76337, "tid": -914061504, "ts": 1716454224487954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487955, "dur": 0,
    "args": {
      "External id": 193255,
      "cbid": 251, "correlation": 193255
    }
  },
  {
    "ph": "f", "id": 193255, "pid": 76337, "tid": -914061504, "ts": 1716454224487955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487956, "dur": 0,
    "args": {
      "External id": 193256,
      "cbid": 251, "correlation": 193256
    }
  },
  {
    "ph": "f", "id": 193256, "pid": 76337, "tid": -914061504, "ts": 1716454224487956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487956, "dur": 1,
    "args": {
      "External id": 193257,
      "cbid": 251, "correlation": 193257
    }
  },
  {
    "ph": "f", "id": 193257, "pid": 76337, "tid": -914061504, "ts": 1716454224487956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487958, "dur": 0,
    "args": {
      "External id": 193258,
      "cbid": 251, "correlation": 193258
    }
  },
  {
    "ph": "f", "id": 193258, "pid": 76337, "tid": -914061504, "ts": 1716454224487958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487958, "dur": 0,
    "args": {
      "External id": 193259,
      "cbid": 251, "correlation": 193259
    }
  },
  {
    "ph": "f", "id": 193259, "pid": 76337, "tid": -914061504, "ts": 1716454224487958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487959, "dur": 0,
    "args": {
      "External id": 193260,
      "cbid": 251, "correlation": 193260
    }
  },
  {
    "ph": "f", "id": 193260, "pid": 76337, "tid": -914061504, "ts": 1716454224487959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487960, "dur": 0,
    "args": {
      "External id": 193261,
      "cbid": 251, "correlation": 193261
    }
  },
  {
    "ph": "f", "id": 193261, "pid": 76337, "tid": -914061504, "ts": 1716454224487960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224509900, "dur": 113,
    "args": {
      "External id": 193262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193262,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193262, "pid": 5, "tid": 7, "ts": 1716454224509900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224487963, "dur": 22,
    "args": {
      "External id": 193262,
      "cbid": 211, "correlation": 193262
    }
  },
  {
    "ph": "s", "id": 193262, "pid": 76337, "tid": -914061504, "ts": 1716454224487963,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224510015, "dur": 60,
    "args": {
      "External id": 193268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193268,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193268, "pid": 5, "tid": 7, "ts": 1716454224510015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488007, "dur": 9,
    "args": {
      "External id": 193268,
      "cbid": 211, "correlation": 193268
    }
  },
  {
    "ph": "s", "id": 193268, "pid": 76337, "tid": -914061504, "ts": 1716454224488007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224510076, "dur": 50,
    "args": {
      "External id": 193276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193276,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193276, "pid": 5, "tid": 7, "ts": 1716454224510076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488042, "dur": 8,
    "args": {
      "External id": 193276,
      "cbid": 211, "correlation": 193276
    }
  },
  {
    "ph": "s", "id": 193276, "pid": 76337, "tid": -914061504, "ts": 1716454224488042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224510127, "dur": 51,
    "args": {
      "External id": 193296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193296,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 193296, "pid": 5, "tid": 7, "ts": 1716454224510127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488115, "dur": 13,
    "args": {
      "External id": 193296,
      "cbid": 211, "correlation": 193296
    }
  },
  {
    "ph": "s", "id": 193296, "pid": 76337, "tid": -914061504, "ts": 1716454224488115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224510179, "dur": 4,
    "args": {
      "External id": 193308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193308,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193308, "pid": 5, "tid": 7, "ts": 1716454224510179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488138, "dur": 7,
    "args": {
      "External id": 193308,
      "cbid": 211, "correlation": 193308
    }
  },
  {
    "ph": "s", "id": 193308, "pid": 76337, "tid": -914061504, "ts": 1716454224488138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224510185, "dur": 57,
    "args": {
      "External id": 193311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193311, "pid": 5, "tid": 7, "ts": 1716454224510185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488156, "dur": 6,
    "args": {
      "External id": 193311,
      "cbid": 211, "correlation": 193311
    }
  },
  {
    "ph": "s", "id": 193311, "pid": 76337, "tid": -914061504, "ts": 1716454224488156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224510244, "dur": 37,
    "args": {
      "External id": 193320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193320,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193320, "pid": 5, "tid": 7, "ts": 1716454224510244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488196, "dur": 10,
    "args": {
      "External id": 193320,
      "cbid": 211, "correlation": 193320
    }
  },
  {
    "ph": "s", "id": 193320, "pid": 76337, "tid": -914061504, "ts": 1716454224488196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488247, "dur": 0,
    "args": {
      "External id": 193330,
      "cbid": 317, "correlation": 193330
    }
  },
  {
    "ph": "f", "id": 193330, "pid": 76337, "tid": -914061504, "ts": 1716454224488247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488247, "dur": 0,
    "args": {
      "External id": 193331,
      "cbid": 203, "correlation": 193331
    }
  },
  {
    "ph": "f", "id": 193331, "pid": 76337, "tid": -914061504, "ts": 1716454224488247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488248, "dur": 0,
    "args": {
      "External id": 193332,
      "cbid": 205, "correlation": 193332
    }
  },
  {
    "ph": "f", "id": 193332, "pid": 76337, "tid": -914061504, "ts": 1716454224488248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224510282, "dur": 41,
    "args": {
      "External id": 193336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193336, "pid": 5, "tid": 7, "ts": 1716454224510282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488262, "dur": 12,
    "args": {
      "External id": 193336,
      "cbid": 211, "correlation": 193336
    }
  },
  {
    "ph": "s", "id": 193336, "pid": 76337, "tid": -914061504, "ts": 1716454224488262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224510324, "dur": 15,
    "args": {
      "External id": 193338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193338,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193338, "pid": 5, "tid": 7, "ts": 1716454224510324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488277, "dur": 6,
    "args": {
      "External id": 193338,
      "cbid": 211, "correlation": 193338
    }
  },
  {
    "ph": "s", "id": 193338, "pid": 76337, "tid": -914061504, "ts": 1716454224488277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224510340, "dur": 3,
    "args": {
      "External id": 193340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193340, "pid": 5, "tid": 7, "ts": 1716454224510340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488287, "dur": 5,
    "args": {
      "External id": 193340,
      "cbid": 211, "correlation": 193340
    }
  },
  {
    "ph": "s", "id": 193340, "pid": 76337, "tid": -914061504, "ts": 1716454224488287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488296, "dur": 0,
    "args": {
      "External id": 193341,
      "cbid": 51, "correlation": 193341
    }
  },
  {
    "ph": "s", "id": 193341, "pid": 76337, "tid": -914061504, "ts": 1716454224488296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224510345, "dur": 703,
    "args": {
      "External id": 193342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193342,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193342, "pid": 5, "tid": 7, "ts": 1716454224510345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488296, "dur": 5,
    "args": {
      "External id": 193342,
      "cbid": 211, "correlation": 193342
    }
  },
  {
    "ph": "s", "id": 193342, "pid": 76337, "tid": -914061504, "ts": 1716454224488296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224511049, "dur": 59,
    "args": {
      "External id": 193347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193347, "pid": 5, "tid": 7, "ts": 1716454224511049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488323, "dur": 10,
    "args": {
      "External id": 193347,
      "cbid": 211, "correlation": 193347
    }
  },
  {
    "ph": "s", "id": 193347, "pid": 76337, "tid": -914061504, "ts": 1716454224488323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224511110, "dur": 3,
    "args": {
      "External id": 193355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193355,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193355, "pid": 5, "tid": 7, "ts": 1716454224511110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488366, "dur": 10,
    "args": {
      "External id": 193355,
      "cbid": 211, "correlation": 193355
    }
  },
  {
    "ph": "s", "id": 193355, "pid": 76337, "tid": -914061504, "ts": 1716454224488366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488433, "dur": 1,
    "args": {
      "External id": 193371,
      "cbid": 251, "correlation": 193371
    }
  },
  {
    "ph": "f", "id": 193371, "pid": 76337, "tid": -914061504, "ts": 1716454224488433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488438, "dur": 0,
    "args": {
      "External id": 193373,
      "cbid": 251, "correlation": 193373
    }
  },
  {
    "ph": "f", "id": 193373, "pid": 76337, "tid": -914061504, "ts": 1716454224488438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224511115, "dur": 11,
    "args": {
      "External id": 193374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193374,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 193374, "pid": 5, "tid": 7, "ts": 1716454224511115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488439, "dur": 11,
    "args": {
      "External id": 193374,
      "cbid": 211, "correlation": 193374
    }
  },
  {
    "ph": "s", "id": 193374, "pid": 76337, "tid": -914061504, "ts": 1716454224488439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224511128, "dur": 5,
    "args": {
      "External id": 193376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193376,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 193376, "pid": 5, "tid": 7, "ts": 1716454224511128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488453, "dur": 5,
    "args": {
      "External id": 193376,
      "cbid": 211, "correlation": 193376
    }
  },
  {
    "ph": "s", "id": 193376, "pid": 76337, "tid": -914061504, "ts": 1716454224488453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224511134, "dur": 52,
    "args": {
      "External id": 193386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193386, "pid": 5, "tid": 7, "ts": 1716454224511134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488511, "dur": 12,
    "args": {
      "External id": 193386,
      "cbid": 211, "correlation": 193386
    }
  },
  {
    "ph": "s", "id": 193386, "pid": 76337, "tid": -914061504, "ts": 1716454224488511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224511188, "dur": 50,
    "args": {
      "External id": 193406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193406,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 193406, "pid": 5, "tid": 7, "ts": 1716454224511188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488578, "dur": 11,
    "args": {
      "External id": 193406,
      "cbid": 211, "correlation": 193406
    }
  },
  {
    "ph": "s", "id": 193406, "pid": 76337, "tid": -914061504, "ts": 1716454224488578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224511239, "dur": 4,
    "args": {
      "External id": 193418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193418,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193418, "pid": 5, "tid": 7, "ts": 1716454224511239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488598, "dur": 10,
    "args": {
      "External id": 193418,
      "cbid": 211, "correlation": 193418
    }
  },
  {
    "ph": "s", "id": 193418, "pid": 76337, "tid": -914061504, "ts": 1716454224488598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224511244, "dur": 55,
    "args": {
      "External id": 193421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193421, "pid": 5, "tid": 7, "ts": 1716454224511244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488620, "dur": 7,
    "args": {
      "External id": 193421,
      "cbid": 211, "correlation": 193421
    }
  },
  {
    "ph": "s", "id": 193421, "pid": 76337, "tid": -914061504, "ts": 1716454224488620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224511301, "dur": 36,
    "args": {
      "External id": 193430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193430,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193430, "pid": 5, "tid": 7, "ts": 1716454224511301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488661, "dur": 10,
    "args": {
      "External id": 193430,
      "cbid": 211, "correlation": 193430
    }
  },
  {
    "ph": "s", "id": 193430, "pid": 76337, "tid": -914061504, "ts": 1716454224488661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488726, "dur": 0,
    "args": {
      "External id": 193440,
      "cbid": 317, "correlation": 193440
    }
  },
  {
    "ph": "f", "id": 193440, "pid": 76337, "tid": -914061504, "ts": 1716454224488726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488727, "dur": 0,
    "args": {
      "External id": 193441,
      "cbid": 203, "correlation": 193441
    }
  },
  {
    "ph": "f", "id": 193441, "pid": 76337, "tid": -914061504, "ts": 1716454224488727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488728, "dur": 0,
    "args": {
      "External id": 193442,
      "cbid": 205, "correlation": 193442
    }
  },
  {
    "ph": "f", "id": 193442, "pid": 76337, "tid": -914061504, "ts": 1716454224488728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224511338, "dur": 40,
    "args": {
      "External id": 193446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193446, "pid": 5, "tid": 7, "ts": 1716454224511338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488741, "dur": 12,
    "args": {
      "External id": 193446,
      "cbid": 211, "correlation": 193446
    }
  },
  {
    "ph": "s", "id": 193446, "pid": 76337, "tid": -914061504, "ts": 1716454224488741,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224511380, "dur": 14,
    "args": {
      "External id": 193448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193448,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193448, "pid": 5, "tid": 7, "ts": 1716454224511380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488756, "dur": 5,
    "args": {
      "External id": 193448,
      "cbid": 211, "correlation": 193448
    }
  },
  {
    "ph": "s", "id": 193448, "pid": 76337, "tid": -914061504, "ts": 1716454224488756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224511395, "dur": 3,
    "args": {
      "External id": 193450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193450,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193450, "pid": 5, "tid": 7, "ts": 1716454224511395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488765, "dur": 5,
    "args": {
      "External id": 193450,
      "cbid": 211, "correlation": 193450
    }
  },
  {
    "ph": "s", "id": 193450, "pid": 76337, "tid": -914061504, "ts": 1716454224488765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488774, "dur": 0,
    "args": {
      "External id": 193451,
      "cbid": 51, "correlation": 193451
    }
  },
  {
    "ph": "s", "id": 193451, "pid": 76337, "tid": -914061504, "ts": 1716454224488774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224511400, "dur": 697,
    "args": {
      "External id": 193452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193452,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193452, "pid": 5, "tid": 7, "ts": 1716454224511400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488774, "dur": 6,
    "args": {
      "External id": 193452,
      "cbid": 211, "correlation": 193452
    }
  },
  {
    "ph": "s", "id": 193452, "pid": 76337, "tid": -914061504, "ts": 1716454224488774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224512098, "dur": 59,
    "args": {
      "External id": 193457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193457, "pid": 5, "tid": 7, "ts": 1716454224512098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488802, "dur": 8,
    "args": {
      "External id": 193457,
      "cbid": 211, "correlation": 193457
    }
  },
  {
    "ph": "s", "id": 193457, "pid": 76337, "tid": -914061504, "ts": 1716454224488802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224512159, "dur": 50,
    "args": {
      "External id": 193465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193465,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193465, "pid": 5, "tid": 7, "ts": 1716454224512159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488835, "dur": 8,
    "args": {
      "External id": 193465,
      "cbid": 211, "correlation": 193465
    }
  },
  {
    "ph": "s", "id": 193465, "pid": 76337, "tid": -914061504, "ts": 1716454224488835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224512209, "dur": 35,
    "args": {
      "External id": 193473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193473,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193473, "pid": 5, "tid": 7, "ts": 1716454224512209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488863, "dur": 8,
    "args": {
      "External id": 193473,
      "cbid": 211, "correlation": 193473
    }
  },
  {
    "ph": "s", "id": 193473, "pid": 76337, "tid": -914061504, "ts": 1716454224488863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224512246, "dur": 52,
    "args": {
      "External id": 193493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193493,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 193493, "pid": 5, "tid": 7, "ts": 1716454224512246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488942, "dur": 12,
    "args": {
      "External id": 193493,
      "cbid": 211, "correlation": 193493
    }
  },
  {
    "ph": "s", "id": 193493, "pid": 76337, "tid": -914061504, "ts": 1716454224488942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224512299, "dur": 4,
    "args": {
      "External id": 193505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193505,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193505, "pid": 5, "tid": 7, "ts": 1716454224512299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488964, "dur": 6,
    "args": {
      "External id": 193505,
      "cbid": 211, "correlation": 193505
    }
  },
  {
    "ph": "s", "id": 193505, "pid": 76337, "tid": -914061504, "ts": 1716454224488964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224512304, "dur": 55,
    "args": {
      "External id": 193508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193508, "pid": 5, "tid": 7, "ts": 1716454224512304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224488990, "dur": 7,
    "args": {
      "External id": 193508,
      "cbid": 211, "correlation": 193508
    }
  },
  {
    "ph": "s", "id": 193508, "pid": 76337, "tid": -914061504, "ts": 1716454224488990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489048, "dur": 0,
    "args": {
      "External id": 193519,
      "cbid": 317, "correlation": 193519
    }
  },
  {
    "ph": "f", "id": 193519, "pid": 76337, "tid": -914061504, "ts": 1716454224489048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489049, "dur": 0,
    "args": {
      "External id": 193520,
      "cbid": 203, "correlation": 193520
    }
  },
  {
    "ph": "f", "id": 193520, "pid": 76337, "tid": -914061504, "ts": 1716454224489049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489050, "dur": 0,
    "args": {
      "External id": 193521,
      "cbid": 205, "correlation": 193521
    }
  },
  {
    "ph": "f", "id": 193521, "pid": 76337, "tid": -914061504, "ts": 1716454224489050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489070, "dur": 1,
    "args": {
      "External id": 193525,
      "cbid": 251, "correlation": 193525
    }
  },
  {
    "ph": "f", "id": 193525, "pid": 76337, "tid": -914061504, "ts": 1716454224489070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489072, "dur": 0,
    "args": {
      "External id": 193526,
      "cbid": 251, "correlation": 193526
    }
  },
  {
    "ph": "f", "id": 193526, "pid": 76337, "tid": -914061504, "ts": 1716454224489072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489073, "dur": 0,
    "args": {
      "External id": 193527,
      "cbid": 251, "correlation": 193527
    }
  },
  {
    "ph": "f", "id": 193527, "pid": 76337, "tid": -914061504, "ts": 1716454224489073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489074, "dur": 0,
    "args": {
      "External id": 193528,
      "cbid": 251, "correlation": 193528
    }
  },
  {
    "ph": "f", "id": 193528, "pid": 76337, "tid": -914061504, "ts": 1716454224489074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489074, "dur": 0,
    "args": {
      "External id": 193529,
      "cbid": 251, "correlation": 193529
    }
  },
  {
    "ph": "f", "id": 193529, "pid": 76337, "tid": -914061504, "ts": 1716454224489074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489075, "dur": 0,
    "args": {
      "External id": 193530,
      "cbid": 251, "correlation": 193530
    }
  },
  {
    "ph": "f", "id": 193530, "pid": 76337, "tid": -914061504, "ts": 1716454224489075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489076, "dur": 0,
    "args": {
      "External id": 193531,
      "cbid": 251, "correlation": 193531
    }
  },
  {
    "ph": "f", "id": 193531, "pid": 76337, "tid": -914061504, "ts": 1716454224489076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489076, "dur": 0,
    "args": {
      "External id": 193532,
      "cbid": 251, "correlation": 193532
    }
  },
  {
    "ph": "f", "id": 193532, "pid": 76337, "tid": -914061504, "ts": 1716454224489076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489078, "dur": 0,
    "args": {
      "External id": 193533,
      "cbid": 251, "correlation": 193533
    }
  },
  {
    "ph": "f", "id": 193533, "pid": 76337, "tid": -914061504, "ts": 1716454224489078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224512360, "dur": 112,
    "args": {
      "External id": 193534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193534,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193534, "pid": 5, "tid": 7, "ts": 1716454224512360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489080, "dur": 13,
    "args": {
      "External id": 193534,
      "cbid": 211, "correlation": 193534
    }
  },
  {
    "ph": "s", "id": 193534, "pid": 76337, "tid": -914061504, "ts": 1716454224489080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224512473, "dur": 60,
    "args": {
      "External id": 193540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193540,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193540, "pid": 5, "tid": 7, "ts": 1716454224512473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489115, "dur": 9,
    "args": {
      "External id": 193540,
      "cbid": 211, "correlation": 193540
    }
  },
  {
    "ph": "s", "id": 193540, "pid": 76337, "tid": -914061504, "ts": 1716454224489115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224512534, "dur": 653,
    "args": {
      "External id": 193549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193549,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193549, "pid": 5, "tid": 7, "ts": 1716454224512534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489198, "dur": 13,
    "args": {
      "External id": 193549,
      "cbid": 211, "correlation": 193549
    }
  },
  {
    "ph": "s", "id": 193549, "pid": 76337, "tid": -914061504, "ts": 1716454224489198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224513188, "dur": 182,
    "args": {
      "External id": 193571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193571,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193571, "pid": 5, "tid": 7, "ts": 1716454224513188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489258, "dur": 11,
    "args": {
      "External id": 193571,
      "cbid": 211, "correlation": 193571
    }
  },
  {
    "ph": "s", "id": 193571, "pid": 76337, "tid": -914061504, "ts": 1716454224489258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489346, "dur": 1,
    "args": {
      "External id": 193582,
      "cbid": 251, "correlation": 193582
    }
  },
  {
    "ph": "f", "id": 193582, "pid": 76337, "tid": -914061504, "ts": 1716454224489346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224513371, "dur": 196,
    "args": {
      "External id": 193583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193583,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193583, "pid": 5, "tid": 7, "ts": 1716454224513371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489351, "dur": 13,
    "args": {
      "External id": 193583,
      "cbid": 211, "correlation": 193583
    }
  },
  {
    "ph": "s", "id": 193583, "pid": 76337, "tid": -914061504, "ts": 1716454224489351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489422, "dur": 1,
    "args": {
      "External id": 193594,
      "cbid": 251, "correlation": 193594
    }
  },
  {
    "ph": "f", "id": 193594, "pid": 76337, "tid": -914061504, "ts": 1716454224489422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224513568, "dur": 185,
    "args": {
      "External id": 193595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193595,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193595, "pid": 5, "tid": 7, "ts": 1716454224513568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489427, "dur": 13,
    "args": {
      "External id": 193595,
      "cbid": 211, "correlation": 193595
    }
  },
  {
    "ph": "s", "id": 193595, "pid": 76337, "tid": -914061504, "ts": 1716454224489427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489490, "dur": 1,
    "args": {
      "External id": 193606,
      "cbid": 251, "correlation": 193606
    }
  },
  {
    "ph": "f", "id": 193606, "pid": 76337, "tid": -914061504, "ts": 1716454224489490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224513755, "dur": 186,
    "args": {
      "External id": 193607,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193607,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193607, "pid": 5, "tid": 7, "ts": 1716454224513755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489495, "dur": 11,
    "args": {
      "External id": 193607,
      "cbid": 211, "correlation": 193607
    }
  },
  {
    "ph": "s", "id": 193607, "pid": 76337, "tid": -914061504, "ts": 1716454224489495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224513942, "dur": 18696,
    "args": {
      "External id": 193628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193628,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193628, "pid": 5, "tid": 7, "ts": 1716454224513942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489574, "dur": 14,
    "args": {
      "External id": 193628,
      "cbid": 211, "correlation": 193628
    }
  },
  {
    "ph": "s", "id": 193628, "pid": 76337, "tid": -914061504, "ts": 1716454224489574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489673, "dur": 1,
    "args": {
      "External id": 193646,
      "cbid": 251, "correlation": 193646
    }
  },
  {
    "ph": "f", "id": 193646, "pid": 76337, "tid": -914061504, "ts": 1716454224489673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224532638, "dur": 203,
    "args": {
      "External id": 193648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193648,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193648, "pid": 5, "tid": 7, "ts": 1716454224532638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489679, "dur": 14,
    "args": {
      "External id": 193648,
      "cbid": 211, "correlation": 193648
    }
  },
  {
    "ph": "s", "id": 193648, "pid": 76337, "tid": -914061504, "ts": 1716454224489679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224532842, "dur": 66,
    "args": {
      "External id": 193656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193656,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193656, "pid": 5, "tid": 7, "ts": 1716454224532842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489750, "dur": 12,
    "args": {
      "External id": 193656,
      "cbid": 211, "correlation": 193656
    }
  },
  {
    "ph": "s", "id": 193656, "pid": 76337, "tid": -914061504, "ts": 1716454224489750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224532910, "dur": 97,
    "args": {
      "External id": 193664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193664,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193664, "pid": 5, "tid": 7, "ts": 1716454224532910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489790, "dur": 8,
    "args": {
      "External id": 193664,
      "cbid": 211, "correlation": 193664
    }
  },
  {
    "ph": "s", "id": 193664, "pid": 76337, "tid": -914061504, "ts": 1716454224489790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224533008, "dur": 54,
    "args": {
      "External id": 193675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193675,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193675, "pid": 5, "tid": 7, "ts": 1716454224533008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489860, "dur": 12,
    "args": {
      "External id": 193675,
      "cbid": 211, "correlation": 193675
    }
  },
  {
    "ph": "s", "id": 193675, "pid": 76337, "tid": -914061504, "ts": 1716454224489860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224533064, "dur": 92,
    "args": {
      "External id": 193697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193697,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193697, "pid": 5, "tid": 7, "ts": 1716454224533064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489894, "dur": 7,
    "args": {
      "External id": 193697,
      "cbid": 211, "correlation": 193697
    }
  },
  {
    "ph": "s", "id": 193697, "pid": 76337, "tid": -914061504, "ts": 1716454224489894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489986, "dur": 1,
    "args": {
      "External id": 193708,
      "cbid": 251, "correlation": 193708
    }
  },
  {
    "ph": "f", "id": 193708, "pid": 76337, "tid": -914061504, "ts": 1716454224489986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224533157, "dur": 105,
    "args": {
      "External id": 193709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193709,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193709, "pid": 5, "tid": 7, "ts": 1716454224533157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224489992, "dur": 13,
    "args": {
      "External id": 193709,
      "cbid": 211, "correlation": 193709
    }
  },
  {
    "ph": "s", "id": 193709, "pid": 76337, "tid": -914061504, "ts": 1716454224489992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490065, "dur": 1,
    "args": {
      "External id": 193720,
      "cbid": 251, "correlation": 193720
    }
  },
  {
    "ph": "f", "id": 193720, "pid": 76337, "tid": -914061504, "ts": 1716454224490065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490069, "dur": 0,
    "args": {
      "External id": 193721,
      "cbid": 251, "correlation": 193721
    }
  },
  {
    "ph": "f", "id": 193721, "pid": 76337, "tid": -914061504, "ts": 1716454224490069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224533264, "dur": 10,
    "args": {
      "External id": 193722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193722,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 193722, "pid": 5, "tid": 7, "ts": 1716454224533264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490071, "dur": 12,
    "args": {
      "External id": 193722,
      "cbid": 211, "correlation": 193722
    }
  },
  {
    "ph": "s", "id": 193722, "pid": 76337, "tid": -914061504, "ts": 1716454224490071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224533275, "dur": 5,
    "args": {
      "External id": 193724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193724,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 193724, "pid": 5, "tid": 7, "ts": 1716454224533275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490084, "dur": 7,
    "args": {
      "External id": 193724,
      "cbid": 211, "correlation": 193724
    }
  },
  {
    "ph": "s", "id": 193724, "pid": 76337, "tid": -914061504, "ts": 1716454224490084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490146, "dur": 1,
    "args": {
      "External id": 193735,
      "cbid": 251, "correlation": 193735
    }
  },
  {
    "ph": "f", "id": 193735, "pid": 76337, "tid": -914061504, "ts": 1716454224490146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490149, "dur": 0,
    "args": {
      "External id": 193736,
      "cbid": 251, "correlation": 193736
    }
  },
  {
    "ph": "f", "id": 193736, "pid": 76337, "tid": -914061504, "ts": 1716454224490149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224533282, "dur": 6,
    "args": {
      "External id": 193737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193737,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 193737, "pid": 5, "tid": 7, "ts": 1716454224533282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490150, "dur": 11,
    "args": {
      "External id": 193737,
      "cbid": 211, "correlation": 193737
    }
  },
  {
    "ph": "s", "id": 193737, "pid": 76337, "tid": -914061504, "ts": 1716454224490150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224533289, "dur": 3,
    "args": {
      "External id": 193739,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193739,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 193739, "pid": 5, "tid": 7, "ts": 1716454224533289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490163, "dur": 5,
    "args": {
      "External id": 193739,
      "cbid": 211, "correlation": 193739
    }
  },
  {
    "ph": "s", "id": 193739, "pid": 76337, "tid": -914061504, "ts": 1716454224490163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224533294, "dur": 154,
    "args": {
      "External id": 193760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193760,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193760, "pid": 5, "tid": 7, "ts": 1716454224533294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490238, "dur": 12,
    "args": {
      "External id": 193760,
      "cbid": 211, "correlation": 193760
    }
  },
  {
    "ph": "s", "id": 193760, "pid": 76337, "tid": -914061504, "ts": 1716454224490238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490333, "dur": 1,
    "args": {
      "External id": 193778,
      "cbid": 251, "correlation": 193778
    }
  },
  {
    "ph": "f", "id": 193778, "pid": 76337, "tid": -914061504, "ts": 1716454224490333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224533449, "dur": 106,
    "args": {
      "External id": 193780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193780,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193780, "pid": 5, "tid": 7, "ts": 1716454224533449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490340, "dur": 13,
    "args": {
      "External id": 193780,
      "cbid": 211, "correlation": 193780
    }
  },
  {
    "ph": "s", "id": 193780, "pid": 76337, "tid": -914061504, "ts": 1716454224490340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224533556, "dur": 35,
    "args": {
      "External id": 193788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193788,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193788, "pid": 5, "tid": 7, "ts": 1716454224533556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490408, "dur": 12,
    "args": {
      "External id": 193788,
      "cbid": 211, "correlation": 193788
    }
  },
  {
    "ph": "s", "id": 193788, "pid": 76337, "tid": -914061504, "ts": 1716454224490408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224533592, "dur": 66,
    "args": {
      "External id": 193796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193796,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193796, "pid": 5, "tid": 7, "ts": 1716454224533592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490449, "dur": 9,
    "args": {
      "External id": 193796,
      "cbid": 211, "correlation": 193796
    }
  },
  {
    "ph": "s", "id": 193796, "pid": 76337, "tid": -914061504, "ts": 1716454224490449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224533659, "dur": 92,
    "args": {
      "External id": 193818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193818,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193818, "pid": 5, "tid": 7, "ts": 1716454224533659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490501, "dur": 10,
    "args": {
      "External id": 193818,
      "cbid": 211, "correlation": 193818
    }
  },
  {
    "ph": "s", "id": 193818, "pid": 76337, "tid": -914061504, "ts": 1716454224490501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490586, "dur": 1,
    "args": {
      "External id": 193834,
      "cbid": 251, "correlation": 193834
    }
  },
  {
    "ph": "f", "id": 193834, "pid": 76337, "tid": -914061504, "ts": 1716454224490586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224533753, "dur": 576,
    "args": {
      "External id": 193836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193836,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193836, "pid": 5, "tid": 7, "ts": 1716454224533753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490592, "dur": 12,
    "args": {
      "External id": 193836,
      "cbid": 211, "correlation": 193836
    }
  },
  {
    "ph": "s", "id": 193836, "pid": 76337, "tid": -914061504, "ts": 1716454224490592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224534331, "dur": 242,
    "args": {
      "External id": 193844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193844, "pid": 5, "tid": 7, "ts": 1716454224534331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490656, "dur": 12,
    "args": {
      "External id": 193844,
      "cbid": 211, "correlation": 193844
    }
  },
  {
    "ph": "s", "id": 193844, "pid": 76337, "tid": -914061504, "ts": 1716454224490656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224534574, "dur": 251,
    "args": {
      "External id": 193852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193852,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193852, "pid": 5, "tid": 7, "ts": 1716454224534574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490686, "dur": 9,
    "args": {
      "External id": 193852,
      "cbid": 211, "correlation": 193852
    }
  },
  {
    "ph": "s", "id": 193852, "pid": 76337, "tid": -914061504, "ts": 1716454224490686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490768, "dur": 1,
    "args": {
      "External id": 193868,
      "cbid": 251, "correlation": 193868
    }
  },
  {
    "ph": "f", "id": 193868, "pid": 76337, "tid": -914061504, "ts": 1716454224490768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490773, "dur": 0,
    "args": {
      "External id": 193870,
      "cbid": 251, "correlation": 193870
    }
  },
  {
    "ph": "f", "id": 193870, "pid": 76337, "tid": -914061504, "ts": 1716454224490773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224534827, "dur": 358,
    "args": {
      "External id": 193871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193871,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 193871, "pid": 5, "tid": 7, "ts": 1716454224534827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490776, "dur": 13,
    "args": {
      "External id": 193871,
      "cbid": 211, "correlation": 193871
    }
  },
  {
    "ph": "s", "id": 193871, "pid": 76337, "tid": -914061504, "ts": 1716454224490776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224535186, "dur": 50,
    "args": {
      "External id": 193879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193879,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193879, "pid": 5, "tid": 7, "ts": 1716454224535186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490818, "dur": 10,
    "args": {
      "External id": 193879,
      "cbid": 211, "correlation": 193879
    }
  },
  {
    "ph": "s", "id": 193879, "pid": 76337, "tid": -914061504, "ts": 1716454224490818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224535238, "dur": 157,
    "args": {
      "External id": 193890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193890, "pid": 5, "tid": 7, "ts": 1716454224535238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490886, "dur": 13,
    "args": {
      "External id": 193890,
      "cbid": 211, "correlation": 193890
    }
  },
  {
    "ph": "s", "id": 193890, "pid": 76337, "tid": -914061504, "ts": 1716454224490886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490951, "dur": 0,
    "args": {
      "External id": 193902,
      "cbid": 317, "correlation": 193902
    }
  },
  {
    "ph": "f", "id": 193902, "pid": 76337, "tid": -914061504, "ts": 1716454224490951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490951, "dur": 0,
    "args": {
      "External id": 193903,
      "cbid": 203, "correlation": 193903
    }
  },
  {
    "ph": "f", "id": 193903, "pid": 76337, "tid": -914061504, "ts": 1716454224490951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490952, "dur": 0,
    "args": {
      "External id": 193904,
      "cbid": 205, "correlation": 193904
    }
  },
  {
    "ph": "f", "id": 193904, "pid": 76337, "tid": -914061504, "ts": 1716454224490952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490980, "dur": 1,
    "args": {
      "External id": 193908,
      "cbid": 251, "correlation": 193908
    }
  },
  {
    "ph": "f", "id": 193908, "pid": 76337, "tid": -914061504, "ts": 1716454224490980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490982, "dur": 0,
    "args": {
      "External id": 193909,
      "cbid": 251, "correlation": 193909
    }
  },
  {
    "ph": "f", "id": 193909, "pid": 76337, "tid": -914061504, "ts": 1716454224490982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490983, "dur": 0,
    "args": {
      "External id": 193910,
      "cbid": 251, "correlation": 193910
    }
  },
  {
    "ph": "f", "id": 193910, "pid": 76337, "tid": -914061504, "ts": 1716454224490983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490984, "dur": 0,
    "args": {
      "External id": 193911,
      "cbid": 251, "correlation": 193911
    }
  },
  {
    "ph": "f", "id": 193911, "pid": 76337, "tid": -914061504, "ts": 1716454224490984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490985, "dur": 0,
    "args": {
      "External id": 193912,
      "cbid": 251, "correlation": 193912
    }
  },
  {
    "ph": "f", "id": 193912, "pid": 76337, "tid": -914061504, "ts": 1716454224490985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490985, "dur": 0,
    "args": {
      "External id": 193913,
      "cbid": 251, "correlation": 193913
    }
  },
  {
    "ph": "f", "id": 193913, "pid": 76337, "tid": -914061504, "ts": 1716454224490985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490986, "dur": 0,
    "args": {
      "External id": 193914,
      "cbid": 251, "correlation": 193914
    }
  },
  {
    "ph": "f", "id": 193914, "pid": 76337, "tid": -914061504, "ts": 1716454224490986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490987, "dur": 0,
    "args": {
      "External id": 193915,
      "cbid": 251, "correlation": 193915
    }
  },
  {
    "ph": "f", "id": 193915, "pid": 76337, "tid": -914061504, "ts": 1716454224490987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490988, "dur": 0,
    "args": {
      "External id": 193916,
      "cbid": 251, "correlation": 193916
    }
  },
  {
    "ph": "f", "id": 193916, "pid": 76337, "tid": -914061504, "ts": 1716454224490988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224535396, "dur": 116,
    "args": {
      "External id": 193917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193917,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 193917, "pid": 5, "tid": 7, "ts": 1716454224535396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224490990, "dur": 13,
    "args": {
      "External id": 193917,
      "cbid": 211, "correlation": 193917
    }
  },
  {
    "ph": "s", "id": 193917, "pid": 76337, "tid": -914061504, "ts": 1716454224490990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224535514, "dur": 59,
    "args": {
      "External id": 193923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193923, "pid": 5, "tid": 7, "ts": 1716454224535514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491026, "dur": 9,
    "args": {
      "External id": 193923,
      "cbid": 211, "correlation": 193923
    }
  },
  {
    "ph": "s", "id": 193923, "pid": 76337, "tid": -914061504, "ts": 1716454224491026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224535574, "dur": 50,
    "args": {
      "External id": 193931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193931,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193931, "pid": 5, "tid": 7, "ts": 1716454224535574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491059, "dur": 8,
    "args": {
      "External id": 193931,
      "cbid": 211, "correlation": 193931
    }
  },
  {
    "ph": "s", "id": 193931, "pid": 76337, "tid": -914061504, "ts": 1716454224491059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491130, "dur": 0,
    "args": {
      "External id": 193941,
      "cbid": 317, "correlation": 193941
    }
  },
  {
    "ph": "f", "id": 193941, "pid": 76337, "tid": -914061504, "ts": 1716454224491130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491131, "dur": 0,
    "args": {
      "External id": 193942,
      "cbid": 203, "correlation": 193942
    }
  },
  {
    "ph": "f", "id": 193942, "pid": 76337, "tid": -914061504, "ts": 1716454224491131,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491132, "dur": 0,
    "args": {
      "External id": 193943,
      "cbid": 205, "correlation": 193943
    }
  },
  {
    "ph": "f", "id": 193943, "pid": 76337, "tid": -914061504, "ts": 1716454224491132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224535625, "dur": 42,
    "args": {
      "External id": 193947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193947, "pid": 5, "tid": 7, "ts": 1716454224535625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491148, "dur": 12,
    "args": {
      "External id": 193947,
      "cbid": 211, "correlation": 193947
    }
  },
  {
    "ph": "s", "id": 193947, "pid": 76337, "tid": -914061504, "ts": 1716454224491148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224535668, "dur": 15,
    "args": {
      "External id": 193949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193949,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193949, "pid": 5, "tid": 7, "ts": 1716454224535668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491163, "dur": 5,
    "args": {
      "External id": 193949,
      "cbid": 211, "correlation": 193949
    }
  },
  {
    "ph": "s", "id": 193949, "pid": 76337, "tid": -914061504, "ts": 1716454224491163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224535685, "dur": 1,
    "args": {
      "External id": 193951,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 193951,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 193951, "pid": 5, "tid": 7, "ts": 1716454224535685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491181, "dur": 21,
    "args": {
      "External id": 193951,
      "cbid": 51, "correlation": 193951
    }
  },
  {
    "ph": "s", "id": 193951, "pid": 76337, "tid": -914061504, "ts": 1716454224491181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224535688, "dur": 364,
    "args": {
      "External id": 193952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193952,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 193952, "pid": 5, "tid": 7, "ts": 1716454224535688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491204, "dur": 10,
    "args": {
      "External id": 193952,
      "cbid": 211, "correlation": 193952
    }
  },
  {
    "ph": "s", "id": 193952, "pid": 76337, "tid": -914061504, "ts": 1716454224491204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536054, "dur": 13,
    "args": {
      "External id": 193954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193954,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193954, "pid": 5, "tid": 7, "ts": 1716454224536054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491222, "dur": 7,
    "args": {
      "External id": 193954,
      "cbid": 211, "correlation": 193954
    }
  },
  {
    "ph": "s", "id": 193954, "pid": 76337, "tid": -914061504, "ts": 1716454224491222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224536068, "dur": 15,
    "args": {
      "External id": 193960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193960, "pid": 5, "tid": 7, "ts": 1716454224536068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491253, "dur": 8,
    "args": {
      "External id": 193960,
      "cbid": 211, "correlation": 193960
    }
  },
  {
    "ph": "s", "id": 193960, "pid": 76337, "tid": -914061504, "ts": 1716454224491253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536084, "dur": 19,
    "args": {
      "External id": 193980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193980,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 193980, "pid": 5, "tid": 7, "ts": 1716454224536084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491349, "dur": 12,
    "args": {
      "External id": 193980,
      "cbid": 211, "correlation": 193980
    }
  },
  {
    "ph": "s", "id": 193980, "pid": 76337, "tid": -914061504, "ts": 1716454224491349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224536105, "dur": 4,
    "args": {
      "External id": 193992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193992,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 193992, "pid": 5, "tid": 7, "ts": 1716454224536105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491371, "dur": 6,
    "args": {
      "External id": 193992,
      "cbid": 211, "correlation": 193992
    }
  },
  {
    "ph": "s", "id": 193992, "pid": 76337, "tid": -914061504, "ts": 1716454224491371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224536110, "dur": 17,
    "args": {
      "External id": 193995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 193995,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 193995, "pid": 5, "tid": 7, "ts": 1716454224536110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491390, "dur": 6,
    "args": {
      "External id": 193995,
      "cbid": 211, "correlation": 193995
    }
  },
  {
    "ph": "s", "id": 193995, "pid": 76337, "tid": -914061504, "ts": 1716454224491390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224536129, "dur": 12,
    "args": {
      "External id": 194004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194004,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194004, "pid": 5, "tid": 7, "ts": 1716454224536129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491429, "dur": 11,
    "args": {
      "External id": 194004,
      "cbid": 211, "correlation": 194004
    }
  },
  {
    "ph": "s", "id": 194004, "pid": 76337, "tid": -914061504, "ts": 1716454224491429,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491490, "dur": 0,
    "args": {
      "External id": 194014,
      "cbid": 317, "correlation": 194014
    }
  },
  {
    "ph": "f", "id": 194014, "pid": 76337, "tid": -914061504, "ts": 1716454224491490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491491, "dur": 0,
    "args": {
      "External id": 194015,
      "cbid": 203, "correlation": 194015
    }
  },
  {
    "ph": "f", "id": 194015, "pid": 76337, "tid": -914061504, "ts": 1716454224491491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491492, "dur": 0,
    "args": {
      "External id": 194016,
      "cbid": 205, "correlation": 194016
    }
  },
  {
    "ph": "f", "id": 194016, "pid": 76337, "tid": -914061504, "ts": 1716454224491492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536142, "dur": 11,
    "args": {
      "External id": 194020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194020,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194020, "pid": 5, "tid": 7, "ts": 1716454224536142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491508, "dur": 12,
    "args": {
      "External id": 194020,
      "cbid": 211, "correlation": 194020
    }
  },
  {
    "ph": "s", "id": 194020, "pid": 76337, "tid": -914061504, "ts": 1716454224491508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536155, "dur": 24,
    "args": {
      "External id": 194022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194022,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194022, "pid": 5, "tid": 7, "ts": 1716454224536155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491522, "dur": 5,
    "args": {
      "External id": 194022,
      "cbid": 211, "correlation": 194022
    }
  },
  {
    "ph": "s", "id": 194022, "pid": 76337, "tid": -914061504, "ts": 1716454224491522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224536180, "dur": 4,
    "args": {
      "External id": 194024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194024,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 194024, "pid": 5, "tid": 7, "ts": 1716454224536180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491533, "dur": 6,
    "args": {
      "External id": 194024,
      "cbid": 211, "correlation": 194024
    }
  },
  {
    "ph": "s", "id": 194024, "pid": 76337, "tid": -914061504, "ts": 1716454224491533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491543, "dur": 0,
    "args": {
      "External id": 194025,
      "cbid": 51, "correlation": 194025
    }
  },
  {
    "ph": "s", "id": 194025, "pid": 76337, "tid": -914061504, "ts": 1716454224491543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224536185, "dur": 358,
    "args": {
      "External id": 194026,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194026,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194026, "pid": 5, "tid": 7, "ts": 1716454224536185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491544, "dur": 7,
    "args": {
      "External id": 194026,
      "cbid": 211, "correlation": 194026
    }
  },
  {
    "ph": "s", "id": 194026, "pid": 76337, "tid": -914061504, "ts": 1716454224491544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536545, "dur": 20,
    "args": {
      "External id": 194027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194027,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194027, "pid": 5, "tid": 7, "ts": 1716454224536545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491554, "dur": 5,
    "args": {
      "External id": 194027,
      "cbid": 211, "correlation": 194027
    }
  },
  {
    "ph": "s", "id": 194027, "pid": 76337, "tid": -914061504, "ts": 1716454224491554,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224536566, "dur": 32,
    "args": {
      "External id": 194033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194033, "pid": 5, "tid": 7, "ts": 1716454224536566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491581, "dur": 9,
    "args": {
      "External id": 194033,
      "cbid": 211, "correlation": 194033
    }
  },
  {
    "ph": "s", "id": 194033, "pid": 76337, "tid": -914061504, "ts": 1716454224491581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224536599, "dur": 3,
    "args": {
      "External id": 194041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194041,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 194041, "pid": 5, "tid": 7, "ts": 1716454224536599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491627, "dur": 10,
    "args": {
      "External id": 194041,
      "cbid": 211, "correlation": 194041
    }
  },
  {
    "ph": "s", "id": 194041, "pid": 76337, "tid": -914061504, "ts": 1716454224491627,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491696, "dur": 1,
    "args": {
      "External id": 194057,
      "cbid": 251, "correlation": 194057
    }
  },
  {
    "ph": "f", "id": 194057, "pid": 76337, "tid": -914061504, "ts": 1716454224491696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491701, "dur": 0,
    "args": {
      "External id": 194059,
      "cbid": 251, "correlation": 194059
    }
  },
  {
    "ph": "f", "id": 194059, "pid": 76337, "tid": -914061504, "ts": 1716454224491701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224536604, "dur": 13,
    "args": {
      "External id": 194060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194060,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 194060, "pid": 5, "tid": 7, "ts": 1716454224536604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491703, "dur": 11,
    "args": {
      "External id": 194060,
      "cbid": 211, "correlation": 194060
    }
  },
  {
    "ph": "s", "id": 194060, "pid": 76337, "tid": -914061504, "ts": 1716454224491703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224536618, "dur": 5,
    "args": {
      "External id": 194062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194062,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 194062, "pid": 5, "tid": 7, "ts": 1716454224536618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491717, "dur": 5,
    "args": {
      "External id": 194062,
      "cbid": 211, "correlation": 194062
    }
  },
  {
    "ph": "s", "id": 194062, "pid": 76337, "tid": -914061504, "ts": 1716454224491717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224536624, "dur": 30,
    "args": {
      "External id": 194072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194072, "pid": 5, "tid": 7, "ts": 1716454224536624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491775, "dur": 13,
    "args": {
      "External id": 194072,
      "cbid": 211, "correlation": 194072
    }
  },
  {
    "ph": "s", "id": 194072, "pid": 76337, "tid": -914061504, "ts": 1716454224491775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536655, "dur": 31,
    "args": {
      "External id": 194092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194092,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 194092, "pid": 5, "tid": 7, "ts": 1716454224536655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491848, "dur": 11,
    "args": {
      "External id": 194092,
      "cbid": 211, "correlation": 194092
    }
  },
  {
    "ph": "s", "id": 194092, "pid": 76337, "tid": -914061504, "ts": 1716454224491848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224536687, "dur": 4,
    "args": {
      "External id": 194104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194104,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 194104, "pid": 5, "tid": 7, "ts": 1716454224536687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491871, "dur": 6,
    "args": {
      "External id": 194104,
      "cbid": 211, "correlation": 194104
    }
  },
  {
    "ph": "s", "id": 194104, "pid": 76337, "tid": -914061504, "ts": 1716454224491871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224536693, "dur": 29,
    "args": {
      "External id": 194107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194107, "pid": 5, "tid": 7, "ts": 1716454224536693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491891, "dur": 6,
    "args": {
      "External id": 194107,
      "cbid": 211, "correlation": 194107
    }
  },
  {
    "ph": "s", "id": 194107, "pid": 76337, "tid": -914061504, "ts": 1716454224491891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224536724, "dur": 20,
    "args": {
      "External id": 194116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194116,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194116, "pid": 5, "tid": 7, "ts": 1716454224536724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224491932, "dur": 10,
    "args": {
      "External id": 194116,
      "cbid": 211, "correlation": 194116
    }
  },
  {
    "ph": "s", "id": 194116, "pid": 76337, "tid": -914061504, "ts": 1716454224491932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492004, "dur": 0,
    "args": {
      "External id": 194126,
      "cbid": 317, "correlation": 194126
    }
  },
  {
    "ph": "f", "id": 194126, "pid": 76337, "tid": -914061504, "ts": 1716454224492004,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492005, "dur": 0,
    "args": {
      "External id": 194127,
      "cbid": 203, "correlation": 194127
    }
  },
  {
    "ph": "f", "id": 194127, "pid": 76337, "tid": -914061504, "ts": 1716454224492005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492005, "dur": 0,
    "args": {
      "External id": 194128,
      "cbid": 205, "correlation": 194128
    }
  },
  {
    "ph": "f", "id": 194128, "pid": 76337, "tid": -914061504, "ts": 1716454224492005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536746, "dur": 23,
    "args": {
      "External id": 194132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194132, "pid": 5, "tid": 7, "ts": 1716454224536746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492023, "dur": 13,
    "args": {
      "External id": 194132,
      "cbid": 211, "correlation": 194132
    }
  },
  {
    "ph": "s", "id": 194132, "pid": 76337, "tid": -914061504, "ts": 1716454224492023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224536770, "dur": 44,
    "args": {
      "External id": 194134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194134, "pid": 5, "tid": 7, "ts": 1716454224536770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492039, "dur": 5,
    "args": {
      "External id": 194134,
      "cbid": 211, "correlation": 194134
    }
  },
  {
    "ph": "s", "id": 194134, "pid": 76337, "tid": -914061504, "ts": 1716454224492039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224536816, "dur": 655,
    "args": {
      "External id": 194136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194136,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194136, "pid": 5, "tid": 7, "ts": 1716454224536816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492053, "dur": 9,
    "args": {
      "External id": 194136,
      "cbid": 211, "correlation": 194136
    }
  },
  {
    "ph": "s", "id": 194136, "pid": 76337, "tid": -914061504, "ts": 1716454224492053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224537472, "dur": 21,
    "args": {
      "External id": 194138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194138,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194138, "pid": 5, "tid": 7, "ts": 1716454224537472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492066, "dur": 5,
    "args": {
      "External id": 194138,
      "cbid": 211, "correlation": 194138
    }
  },
  {
    "ph": "s", "id": 194138, "pid": 76337, "tid": -914061504, "ts": 1716454224492066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224537494, "dur": 33,
    "args": {
      "External id": 194144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194144, "pid": 5, "tid": 7, "ts": 1716454224537494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492095, "dur": 12,
    "args": {
      "External id": 194144,
      "cbid": 211, "correlation": 194144
    }
  },
  {
    "ph": "s", "id": 194144, "pid": 76337, "tid": -914061504, "ts": 1716454224492095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492158, "dur": 0,
    "args": {
      "External id": 194154,
      "cbid": 317, "correlation": 194154
    }
  },
  {
    "ph": "f", "id": 194154, "pid": 76337, "tid": -914061504, "ts": 1716454224492158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492159, "dur": 0,
    "args": {
      "External id": 194155,
      "cbid": 203, "correlation": 194155
    }
  },
  {
    "ph": "f", "id": 194155, "pid": 76337, "tid": -914061504, "ts": 1716454224492159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492160, "dur": 0,
    "args": {
      "External id": 194156,
      "cbid": 205, "correlation": 194156
    }
  },
  {
    "ph": "f", "id": 194156, "pid": 76337, "tid": -914061504, "ts": 1716454224492160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492181, "dur": 1,
    "args": {
      "External id": 194160,
      "cbid": 251, "correlation": 194160
    }
  },
  {
    "ph": "f", "id": 194160, "pid": 76337, "tid": -914061504, "ts": 1716454224492181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492183, "dur": 0,
    "args": {
      "External id": 194161,
      "cbid": 251, "correlation": 194161
    }
  },
  {
    "ph": "f", "id": 194161, "pid": 76337, "tid": -914061504, "ts": 1716454224492183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492184, "dur": 0,
    "args": {
      "External id": 194162,
      "cbid": 251, "correlation": 194162
    }
  },
  {
    "ph": "f", "id": 194162, "pid": 76337, "tid": -914061504, "ts": 1716454224492184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492185, "dur": 0,
    "args": {
      "External id": 194163,
      "cbid": 251, "correlation": 194163
    }
  },
  {
    "ph": "f", "id": 194163, "pid": 76337, "tid": -914061504, "ts": 1716454224492185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492186, "dur": 0,
    "args": {
      "External id": 194164,
      "cbid": 251, "correlation": 194164
    }
  },
  {
    "ph": "f", "id": 194164, "pid": 76337, "tid": -914061504, "ts": 1716454224492186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492187, "dur": 0,
    "args": {
      "External id": 194165,
      "cbid": 251, "correlation": 194165
    }
  },
  {
    "ph": "f", "id": 194165, "pid": 76337, "tid": -914061504, "ts": 1716454224492187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492188, "dur": 0,
    "args": {
      "External id": 194166,
      "cbid": 251, "correlation": 194166
    }
  },
  {
    "ph": "f", "id": 194166, "pid": 76337, "tid": -914061504, "ts": 1716454224492188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492188, "dur": 0,
    "args": {
      "External id": 194167,
      "cbid": 251, "correlation": 194167
    }
  },
  {
    "ph": "f", "id": 194167, "pid": 76337, "tid": -914061504, "ts": 1716454224492188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492190, "dur": 0,
    "args": {
      "External id": 194168,
      "cbid": 251, "correlation": 194168
    }
  },
  {
    "ph": "f", "id": 194168, "pid": 76337, "tid": -914061504, "ts": 1716454224492190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224537529, "dur": 52,
    "args": {
      "External id": 194169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194169,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 194169, "pid": 5, "tid": 7, "ts": 1716454224537529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492192, "dur": 12,
    "args": {
      "External id": 194169,
      "cbid": 211, "correlation": 194169
    }
  },
  {
    "ph": "s", "id": 194169, "pid": 76337, "tid": -914061504, "ts": 1716454224492192,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224537582, "dur": 32,
    "args": {
      "External id": 194175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194175, "pid": 5, "tid": 7, "ts": 1716454224537582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492224, "dur": 11,
    "args": {
      "External id": 194175,
      "cbid": 211, "correlation": 194175
    }
  },
  {
    "ph": "s", "id": 194175, "pid": 76337, "tid": -914061504, "ts": 1716454224492224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224537615, "dur": 27,
    "args": {
      "External id": 194183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194183,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194183, "pid": 5, "tid": 7, "ts": 1716454224537615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492256, "dur": 8,
    "args": {
      "External id": 194183,
      "cbid": 211, "correlation": 194183
    }
  },
  {
    "ph": "s", "id": 194183, "pid": 76337, "tid": -914061504, "ts": 1716454224492256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224537643, "dur": 20,
    "args": {
      "External id": 194191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194191,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194191, "pid": 5, "tid": 7, "ts": 1716454224537643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492288, "dur": 8,
    "args": {
      "External id": 194191,
      "cbid": 211, "correlation": 194191
    }
  },
  {
    "ph": "s", "id": 194191, "pid": 76337, "tid": -914061504, "ts": 1716454224492288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224537664, "dur": 30,
    "args": {
      "External id": 194211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194211,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 194211, "pid": 5, "tid": 7, "ts": 1716454224537664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492371, "dur": 13,
    "args": {
      "External id": 194211,
      "cbid": 211, "correlation": 194211
    }
  },
  {
    "ph": "s", "id": 194211, "pid": 76337, "tid": -914061504, "ts": 1716454224492371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224537695, "dur": 4,
    "args": {
      "External id": 194223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194223,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 194223, "pid": 5, "tid": 7, "ts": 1716454224537695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492394, "dur": 6,
    "args": {
      "External id": 194223,
      "cbid": 211, "correlation": 194223
    }
  },
  {
    "ph": "s", "id": 194223, "pid": 76337, "tid": -914061504, "ts": 1716454224492394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224537701, "dur": 29,
    "args": {
      "External id": 194226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194226, "pid": 5, "tid": 7, "ts": 1716454224537701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492411, "dur": 6,
    "args": {
      "External id": 194226,
      "cbid": 211, "correlation": 194226
    }
  },
  {
    "ph": "s", "id": 194226, "pid": 76337, "tid": -914061504, "ts": 1716454224492411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492472, "dur": 0,
    "args": {
      "External id": 194237,
      "cbid": 317, "correlation": 194237
    }
  },
  {
    "ph": "f", "id": 194237, "pid": 76337, "tid": -914061504, "ts": 1716454224492472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492473, "dur": 0,
    "args": {
      "External id": 194238,
      "cbid": 203, "correlation": 194238
    }
  },
  {
    "ph": "f", "id": 194238, "pid": 76337, "tid": -914061504, "ts": 1716454224492473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492473, "dur": 0,
    "args": {
      "External id": 194239,
      "cbid": 205, "correlation": 194239
    }
  },
  {
    "ph": "f", "id": 194239, "pid": 76337, "tid": -914061504, "ts": 1716454224492473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224537732, "dur": 22,
    "args": {
      "External id": 194243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194243,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194243, "pid": 5, "tid": 7, "ts": 1716454224537732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492489, "dur": 12,
    "args": {
      "External id": 194243,
      "cbid": 211, "correlation": 194243
    }
  },
  {
    "ph": "s", "id": 194243, "pid": 76337, "tid": -914061504, "ts": 1716454224492489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224537755, "dur": 121,
    "args": {
      "External id": 194245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194245,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194245, "pid": 5, "tid": 7, "ts": 1716454224537755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492509, "dur": 8,
    "args": {
      "External id": 194245,
      "cbid": 211, "correlation": 194245
    }
  },
  {
    "ph": "s", "id": 194245, "pid": 76337, "tid": -914061504, "ts": 1716454224492509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224537877, "dur": 23,
    "args": {
      "External id": 194247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194247,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194247, "pid": 5, "tid": 7, "ts": 1716454224537877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492521, "dur": 5,
    "args": {
      "External id": 194247,
      "cbid": 211, "correlation": 194247
    }
  },
  {
    "ph": "s", "id": 194247, "pid": 76337, "tid": -914061504, "ts": 1716454224492521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224537902, "dur": 32,
    "args": {
      "External id": 194253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194253,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194253, "pid": 5, "tid": 7, "ts": 1716454224537902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492550, "dur": 8,
    "args": {
      "External id": 194253,
      "cbid": 211, "correlation": 194253
    }
  },
  {
    "ph": "s", "id": 194253, "pid": 76337, "tid": -914061504, "ts": 1716454224492550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224537935, "dur": 164,
    "args": {
      "External id": 194262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194262,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194262, "pid": 5, "tid": 7, "ts": 1716454224537935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492633, "dur": 14,
    "args": {
      "External id": 194262,
      "cbid": 211, "correlation": 194262
    }
  },
  {
    "ph": "s", "id": 194262, "pid": 76337, "tid": -914061504, "ts": 1716454224492633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224538100, "dur": 66,
    "args": {
      "External id": 194284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194284,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194284, "pid": 5, "tid": 7, "ts": 1716454224538100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492695, "dur": 10,
    "args": {
      "External id": 194284,
      "cbid": 211, "correlation": 194284
    }
  },
  {
    "ph": "s", "id": 194284, "pid": 76337, "tid": -914061504, "ts": 1716454224492695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492787, "dur": 1,
    "args": {
      "External id": 194295,
      "cbid": 251, "correlation": 194295
    }
  },
  {
    "ph": "f", "id": 194295, "pid": 76337, "tid": -914061504, "ts": 1716454224492787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224538167, "dur": 158,
    "args": {
      "External id": 194296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194296,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194296, "pid": 5, "tid": 7, "ts": 1716454224538167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492792, "dur": 13,
    "args": {
      "External id": 194296,
      "cbid": 211, "correlation": 194296
    }
  },
  {
    "ph": "s", "id": 194296, "pid": 76337, "tid": -914061504, "ts": 1716454224492792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492862, "dur": 1,
    "args": {
      "External id": 194307,
      "cbid": 251, "correlation": 194307
    }
  },
  {
    "ph": "f", "id": 194307, "pid": 76337, "tid": -914061504, "ts": 1716454224492862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224538326, "dur": 148,
    "args": {
      "External id": 194308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194308,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194308, "pid": 5, "tid": 7, "ts": 1716454224538326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492866, "dur": 11,
    "args": {
      "External id": 194308,
      "cbid": 211, "correlation": 194308
    }
  },
  {
    "ph": "s", "id": 194308, "pid": 76337, "tid": -914061504, "ts": 1716454224492866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492930, "dur": 1,
    "args": {
      "External id": 194319,
      "cbid": 251, "correlation": 194319
    }
  },
  {
    "ph": "f", "id": 194319, "pid": 76337, "tid": -914061504, "ts": 1716454224492930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224538475, "dur": 143,
    "args": {
      "External id": 194320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194320,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194320, "pid": 5, "tid": 7, "ts": 1716454224538475,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224492934, "dur": 12,
    "args": {
      "External id": 194320,
      "cbid": 211, "correlation": 194320
    }
  },
  {
    "ph": "s", "id": 194320, "pid": 76337, "tid": -914061504, "ts": 1716454224492934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224538620, "dur": 1946,
    "args": {
      "External id": 194341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194341,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 194341, "pid": 5, "tid": 7, "ts": 1716454224538620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493027, "dur": 15,
    "args": {
      "External id": 194341,
      "cbid": 211, "correlation": 194341
    }
  },
  {
    "ph": "s", "id": 194341, "pid": 76337, "tid": -914061504, "ts": 1716454224493027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493128, "dur": 1,
    "args": {
      "External id": 194359,
      "cbid": 251, "correlation": 194359
    }
  },
  {
    "ph": "f", "id": 194359, "pid": 76337, "tid": -914061504, "ts": 1716454224493128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224540568, "dur": 146,
    "args": {
      "External id": 194361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194361,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 194361, "pid": 5, "tid": 7, "ts": 1716454224540568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493134, "dur": 14,
    "args": {
      "External id": 194361,
      "cbid": 211, "correlation": 194361
    }
  },
  {
    "ph": "s", "id": 194361, "pid": 76337, "tid": -914061504, "ts": 1716454224493134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224540715, "dur": 35,
    "args": {
      "External id": 194369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194369,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194369, "pid": 5, "tid": 7, "ts": 1716454224540715,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493206, "dur": 12,
    "args": {
      "External id": 194369,
      "cbid": 211, "correlation": 194369
    }
  },
  {
    "ph": "s", "id": 194369, "pid": 76337, "tid": -914061504, "ts": 1716454224493206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224540751, "dur": 51,
    "args": {
      "External id": 194377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194377,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194377, "pid": 5, "tid": 7, "ts": 1716454224540751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493246, "dur": 9,
    "args": {
      "External id": 194377,
      "cbid": 211, "correlation": 194377
    }
  },
  {
    "ph": "s", "id": 194377, "pid": 76337, "tid": -914061504, "ts": 1716454224493246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224540803, "dur": 31,
    "args": {
      "External id": 194388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194388,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194388, "pid": 5, "tid": 7, "ts": 1716454224540803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493318, "dur": 13,
    "args": {
      "External id": 194388,
      "cbid": 211, "correlation": 194388
    }
  },
  {
    "ph": "s", "id": 194388, "pid": 76337, "tid": -914061504, "ts": 1716454224493318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224540835, "dur": 34,
    "args": {
      "External id": 194410,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194410,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194410, "pid": 5, "tid": 7, "ts": 1716454224540835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493350, "dur": 7,
    "args": {
      "External id": 194410,
      "cbid": 211, "correlation": 194410
    }
  },
  {
    "ph": "s", "id": 194410, "pid": 76337, "tid": -914061504, "ts": 1716454224493350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493433, "dur": 1,
    "args": {
      "External id": 194421,
      "cbid": 251, "correlation": 194421
    }
  },
  {
    "ph": "f", "id": 194421, "pid": 76337, "tid": -914061504, "ts": 1716454224493433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224540871, "dur": 90,
    "args": {
      "External id": 194422,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194422,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194422, "pid": 5, "tid": 7, "ts": 1716454224540871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493438, "dur": 14,
    "args": {
      "External id": 194422,
      "cbid": 211, "correlation": 194422
    }
  },
  {
    "ph": "s", "id": 194422, "pid": 76337, "tid": -914061504, "ts": 1716454224493438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493508, "dur": 1,
    "args": {
      "External id": 194433,
      "cbid": 251, "correlation": 194433
    }
  },
  {
    "ph": "f", "id": 194433, "pid": 76337, "tid": -914061504, "ts": 1716454224493508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493511, "dur": 0,
    "args": {
      "External id": 194434,
      "cbid": 251, "correlation": 194434
    }
  },
  {
    "ph": "f", "id": 194434, "pid": 76337, "tid": -914061504, "ts": 1716454224493511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224540962, "dur": 12,
    "args": {
      "External id": 194435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194435,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 194435, "pid": 5, "tid": 7, "ts": 1716454224540962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493513, "dur": 12,
    "args": {
      "External id": 194435,
      "cbid": 211, "correlation": 194435
    }
  },
  {
    "ph": "s", "id": 194435, "pid": 76337, "tid": -914061504, "ts": 1716454224493513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224540975, "dur": 5,
    "args": {
      "External id": 194437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194437,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 194437, "pid": 5, "tid": 7, "ts": 1716454224540975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493527, "dur": 7,
    "args": {
      "External id": 194437,
      "cbid": 211, "correlation": 194437
    }
  },
  {
    "ph": "s", "id": 194437, "pid": 76337, "tid": -914061504, "ts": 1716454224493527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493587, "dur": 1,
    "args": {
      "External id": 194448,
      "cbid": 251, "correlation": 194448
    }
  },
  {
    "ph": "f", "id": 194448, "pid": 76337, "tid": -914061504, "ts": 1716454224493587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493590, "dur": 0,
    "args": {
      "External id": 194449,
      "cbid": 251, "correlation": 194449
    }
  },
  {
    "ph": "f", "id": 194449, "pid": 76337, "tid": -914061504, "ts": 1716454224493590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224540981, "dur": 7,
    "args": {
      "External id": 194450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194450,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 194450, "pid": 5, "tid": 7, "ts": 1716454224540981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493592, "dur": 12,
    "args": {
      "External id": 194450,
      "cbid": 211, "correlation": 194450
    }
  },
  {
    "ph": "s", "id": 194450, "pid": 76337, "tid": -914061504, "ts": 1716454224493592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224540990, "dur": 4,
    "args": {
      "External id": 194452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194452,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 194452, "pid": 5, "tid": 7, "ts": 1716454224540990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493605, "dur": 5,
    "args": {
      "External id": 194452,
      "cbid": 211, "correlation": 194452
    }
  },
  {
    "ph": "s", "id": 194452, "pid": 76337, "tid": -914061504, "ts": 1716454224493605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224540995, "dur": 90,
    "args": {
      "External id": 194473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194473,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 194473, "pid": 5, "tid": 7, "ts": 1716454224540995,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493678, "dur": 13,
    "args": {
      "External id": 194473,
      "cbid": 211, "correlation": 194473
    }
  },
  {
    "ph": "s", "id": 194473, "pid": 76337, "tid": -914061504, "ts": 1716454224493678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493779, "dur": 1,
    "args": {
      "External id": 194491,
      "cbid": 251, "correlation": 194491
    }
  },
  {
    "ph": "f", "id": 194491, "pid": 76337, "tid": -914061504, "ts": 1716454224493779,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224541086, "dur": 98,
    "args": {
      "External id": 194493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194493,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194493, "pid": 5, "tid": 7, "ts": 1716454224541086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493785, "dur": 13,
    "args": {
      "External id": 194493,
      "cbid": 211, "correlation": 194493
    }
  },
  {
    "ph": "s", "id": 194493, "pid": 76337, "tid": -914061504, "ts": 1716454224493785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224541186, "dur": 19,
    "args": {
      "External id": 194501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194501,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194501, "pid": 5, "tid": 7, "ts": 1716454224541186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493855, "dur": 12,
    "args": {
      "External id": 194501,
      "cbid": 211, "correlation": 194501
    }
  },
  {
    "ph": "s", "id": 194501, "pid": 76337, "tid": -914061504, "ts": 1716454224493855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224541206, "dur": 37,
    "args": {
      "External id": 194509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194509,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194509, "pid": 5, "tid": 7, "ts": 1716454224541206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493896, "dur": 9,
    "args": {
      "External id": 194509,
      "cbid": 211, "correlation": 194509
    }
  },
  {
    "ph": "s", "id": 194509, "pid": 76337, "tid": -914061504, "ts": 1716454224493896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224541245, "dur": 34,
    "args": {
      "External id": 194531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194531,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194531, "pid": 5, "tid": 7, "ts": 1716454224541245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224493950, "dur": 12,
    "args": {
      "External id": 194531,
      "cbid": 211, "correlation": 194531
    }
  },
  {
    "ph": "s", "id": 194531, "pid": 76337, "tid": -914061504, "ts": 1716454224493950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494050, "dur": 1,
    "args": {
      "External id": 194547,
      "cbid": 251, "correlation": 194547
    }
  },
  {
    "ph": "f", "id": 194547, "pid": 76337, "tid": -914061504, "ts": 1716454224494050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494055, "dur": 0,
    "args": {
      "External id": 194549,
      "cbid": 251, "correlation": 194549
    }
  },
  {
    "ph": "f", "id": 194549, "pid": 76337, "tid": -914061504, "ts": 1716454224494055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224541281, "dur": 540,
    "args": {
      "External id": 194550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194550,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 194550, "pid": 5, "tid": 7, "ts": 1716454224541281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494059, "dur": 13,
    "args": {
      "External id": 194550,
      "cbid": 211, "correlation": 194550
    }
  },
  {
    "ph": "s", "id": 194550, "pid": 76337, "tid": -914061504, "ts": 1716454224494059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224541822, "dur": 125,
    "args": {
      "External id": 194558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194558,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194558, "pid": 5, "tid": 7, "ts": 1716454224541822,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494126, "dur": 13,
    "args": {
      "External id": 194558,
      "cbid": 211, "correlation": 194558
    }
  },
  {
    "ph": "s", "id": 194558, "pid": 76337, "tid": -914061504, "ts": 1716454224494126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224541948, "dur": 127,
    "args": {
      "External id": 194566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194566, "pid": 5, "tid": 7, "ts": 1716454224541948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494158, "dur": 8,
    "args": {
      "External id": 194566,
      "cbid": 211, "correlation": 194566
    }
  },
  {
    "ph": "s", "id": 194566, "pid": 76337, "tid": -914061504, "ts": 1716454224494158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494236, "dur": 1,
    "args": {
      "External id": 194582,
      "cbid": 251, "correlation": 194582
    }
  },
  {
    "ph": "f", "id": 194582, "pid": 76337, "tid": -914061504, "ts": 1716454224494236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224542077, "dur": 257,
    "args": {
      "External id": 194584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194584,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194584, "pid": 5, "tid": 7, "ts": 1716454224542077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494241, "dur": 12,
    "args": {
      "External id": 194584,
      "cbid": 211, "correlation": 194584
    }
  },
  {
    "ph": "s", "id": 194584, "pid": 76337, "tid": -914061504, "ts": 1716454224494241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224542335, "dur": 27,
    "args": {
      "External id": 194592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194592,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194592, "pid": 5, "tid": 7, "ts": 1716454224542335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494283, "dur": 10,
    "args": {
      "External id": 194592,
      "cbid": 211, "correlation": 194592
    }
  },
  {
    "ph": "s", "id": 194592, "pid": 76337, "tid": -914061504, "ts": 1716454224494283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224542364, "dur": 81,
    "args": {
      "External id": 194603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194603,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194603, "pid": 5, "tid": 7, "ts": 1716454224542364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494351, "dur": 13,
    "args": {
      "External id": 194603,
      "cbid": 211, "correlation": 194603
    }
  },
  {
    "ph": "s", "id": 194603, "pid": 76337, "tid": -914061504, "ts": 1716454224494351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494416, "dur": 0,
    "args": {
      "External id": 194615,
      "cbid": 317, "correlation": 194615
    }
  },
  {
    "ph": "f", "id": 194615, "pid": 76337, "tid": -914061504, "ts": 1716454224494416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494417, "dur": 0,
    "args": {
      "External id": 194616,
      "cbid": 203, "correlation": 194616
    }
  },
  {
    "ph": "f", "id": 194616, "pid": 76337, "tid": -914061504, "ts": 1716454224494417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494418, "dur": 0,
    "args": {
      "External id": 194617,
      "cbid": 205, "correlation": 194617
    }
  },
  {
    "ph": "f", "id": 194617, "pid": 76337, "tid": -914061504, "ts": 1716454224494418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224542446, "dur": 25,
    "args": {
      "External id": 194621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194621,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194621, "pid": 5, "tid": 7, "ts": 1716454224542446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494433, "dur": 12,
    "args": {
      "External id": 194621,
      "cbid": 211, "correlation": 194621
    }
  },
  {
    "ph": "s", "id": 194621, "pid": 76337, "tid": -914061504, "ts": 1716454224494433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224542473, "dur": 120,
    "args": {
      "External id": 194623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194623,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194623, "pid": 5, "tid": 7, "ts": 1716454224542473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494451, "dur": 7,
    "args": {
      "External id": 194623,
      "cbid": 211, "correlation": 194623
    }
  },
  {
    "ph": "s", "id": 194623, "pid": 76337, "tid": -914061504, "ts": 1716454224494451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224542594, "dur": 23,
    "args": {
      "External id": 194625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194625,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194625, "pid": 5, "tid": 7, "ts": 1716454224542594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494463, "dur": 5,
    "args": {
      "External id": 194625,
      "cbid": 211, "correlation": 194625
    }
  },
  {
    "ph": "s", "id": 194625, "pid": 76337, "tid": -914061504, "ts": 1716454224494463,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224542618, "dur": 32,
    "args": {
      "External id": 194631,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194631,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194631, "pid": 5, "tid": 7, "ts": 1716454224542618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494490, "dur": 8,
    "args": {
      "External id": 194631,
      "cbid": 211, "correlation": 194631
    }
  },
  {
    "ph": "s", "id": 194631, "pid": 76337, "tid": -914061504, "ts": 1716454224494490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224542652, "dur": 27,
    "args": {
      "External id": 194639,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194639,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194639, "pid": 5, "tid": 7, "ts": 1716454224542652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494522, "dur": 8,
    "args": {
      "External id": 194639,
      "cbid": 211, "correlation": 194639
    }
  },
  {
    "ph": "s", "id": 194639, "pid": 76337, "tid": -914061504, "ts": 1716454224494522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224542680, "dur": 31,
    "args": {
      "External id": 194659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194659,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 194659, "pid": 5, "tid": 7, "ts": 1716454224542680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494599, "dur": 13,
    "args": {
      "External id": 194659,
      "cbid": 211, "correlation": 194659
    }
  },
  {
    "ph": "s", "id": 194659, "pid": 76337, "tid": -914061504, "ts": 1716454224494599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224542713, "dur": 5,
    "args": {
      "External id": 194671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194671,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 194671, "pid": 5, "tid": 7, "ts": 1716454224542713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494621, "dur": 6,
    "args": {
      "External id": 194671,
      "cbid": 211, "correlation": 194671
    }
  },
  {
    "ph": "s", "id": 194671, "pid": 76337, "tid": -914061504, "ts": 1716454224494621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224542719, "dur": 30,
    "args": {
      "External id": 194674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194674, "pid": 5, "tid": 7, "ts": 1716454224542719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494639, "dur": 6,
    "args": {
      "External id": 194674,
      "cbid": 211, "correlation": 194674
    }
  },
  {
    "ph": "s", "id": 194674, "pid": 76337, "tid": -914061504, "ts": 1716454224494639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224542751, "dur": 21,
    "args": {
      "External id": 194683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194683,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194683, "pid": 5, "tid": 7, "ts": 1716454224542751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494678, "dur": 9,
    "args": {
      "External id": 194683,
      "cbid": 211, "correlation": 194683
    }
  },
  {
    "ph": "s", "id": 194683, "pid": 76337, "tid": -914061504, "ts": 1716454224494678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494729, "dur": 0,
    "args": {
      "External id": 194693,
      "cbid": 317, "correlation": 194693
    }
  },
  {
    "ph": "f", "id": 194693, "pid": 76337, "tid": -914061504, "ts": 1716454224494729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494730, "dur": 0,
    "args": {
      "External id": 194694,
      "cbid": 203, "correlation": 194694
    }
  },
  {
    "ph": "f", "id": 194694, "pid": 76337, "tid": -914061504, "ts": 1716454224494730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494731, "dur": 0,
    "args": {
      "External id": 194695,
      "cbid": 205, "correlation": 194695
    }
  },
  {
    "ph": "f", "id": 194695, "pid": 76337, "tid": -914061504, "ts": 1716454224494731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224542773, "dur": 22,
    "args": {
      "External id": 194699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194699,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194699, "pid": 5, "tid": 7, "ts": 1716454224542773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494743, "dur": 11,
    "args": {
      "External id": 194699,
      "cbid": 211, "correlation": 194699
    }
  },
  {
    "ph": "s", "id": 194699, "pid": 76337, "tid": -914061504, "ts": 1716454224494743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224542796, "dur": 44,
    "args": {
      "External id": 194701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194701,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194701, "pid": 5, "tid": 7, "ts": 1716454224542796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494757, "dur": 6,
    "args": {
      "External id": 194701,
      "cbid": 211, "correlation": 194701
    }
  },
  {
    "ph": "s", "id": 194701, "pid": 76337, "tid": -914061504, "ts": 1716454224494757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224542841, "dur": 656,
    "args": {
      "External id": 194703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194703,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194703, "pid": 5, "tid": 7, "ts": 1716454224542841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494769, "dur": 6,
    "args": {
      "External id": 194703,
      "cbid": 211, "correlation": 194703
    }
  },
  {
    "ph": "s", "id": 194703, "pid": 76337, "tid": -914061504, "ts": 1716454224494769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224543498, "dur": 21,
    "args": {
      "External id": 194705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194705,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194705, "pid": 5, "tid": 7, "ts": 1716454224543498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494779, "dur": 5,
    "args": {
      "External id": 194705,
      "cbid": 211, "correlation": 194705
    }
  },
  {
    "ph": "s", "id": 194705, "pid": 76337, "tid": -914061504, "ts": 1716454224494779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224543521, "dur": 33,
    "args": {
      "External id": 194711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194711,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194711, "pid": 5, "tid": 7, "ts": 1716454224543521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494805, "dur": 9,
    "args": {
      "External id": 194711,
      "cbid": 211, "correlation": 194711
    }
  },
  {
    "ph": "s", "id": 194711, "pid": 76337, "tid": -914061504, "ts": 1716454224494805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224543555, "dur": 4,
    "args": {
      "External id": 194719,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194719,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 194719, "pid": 5, "tid": 7, "ts": 1716454224543555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494848, "dur": 9,
    "args": {
      "External id": 194719,
      "cbid": 211, "correlation": 194719
    }
  },
  {
    "ph": "s", "id": 194719, "pid": 76337, "tid": -914061504, "ts": 1716454224494848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494914, "dur": 1,
    "args": {
      "External id": 194735,
      "cbid": 251, "correlation": 194735
    }
  },
  {
    "ph": "f", "id": 194735, "pid": 76337, "tid": -914061504, "ts": 1716454224494914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494919, "dur": 0,
    "args": {
      "External id": 194737,
      "cbid": 251, "correlation": 194737
    }
  },
  {
    "ph": "f", "id": 194737, "pid": 76337, "tid": -914061504, "ts": 1716454224494919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224543560, "dur": 13,
    "args": {
      "External id": 194738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194738,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 194738, "pid": 5, "tid": 7, "ts": 1716454224543560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494921, "dur": 11,
    "args": {
      "External id": 194738,
      "cbid": 211, "correlation": 194738
    }
  },
  {
    "ph": "s", "id": 194738, "pid": 76337, "tid": -914061504, "ts": 1716454224494921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224543574, "dur": 5,
    "args": {
      "External id": 194740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194740,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 194740, "pid": 5, "tid": 7, "ts": 1716454224543574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224494934, "dur": 5,
    "args": {
      "External id": 194740,
      "cbid": 211, "correlation": 194740
    }
  },
  {
    "ph": "s", "id": 194740, "pid": 76337, "tid": -914061504, "ts": 1716454224494934,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224543581, "dur": 29,
    "args": {
      "External id": 194750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194750,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194750, "pid": 5, "tid": 7, "ts": 1716454224543581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495006, "dur": 14,
    "args": {
      "External id": 194750,
      "cbid": 211, "correlation": 194750
    }
  },
  {
    "ph": "s", "id": 194750, "pid": 76337, "tid": -914061504, "ts": 1716454224495006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224543612, "dur": 31,
    "args": {
      "External id": 194770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194770,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 194770, "pid": 5, "tid": 7, "ts": 1716454224543612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495075, "dur": 11,
    "args": {
      "External id": 194770,
      "cbid": 211, "correlation": 194770
    }
  },
  {
    "ph": "s", "id": 194770, "pid": 76337, "tid": -914061504, "ts": 1716454224495075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224543644, "dur": 5,
    "args": {
      "External id": 194782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194782,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 194782, "pid": 5, "tid": 7, "ts": 1716454224543644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495096, "dur": 7,
    "args": {
      "External id": 194782,
      "cbid": 211, "correlation": 194782
    }
  },
  {
    "ph": "s", "id": 194782, "pid": 76337, "tid": -914061504, "ts": 1716454224495096,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224543649, "dur": 30,
    "args": {
      "External id": 194785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194785,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194785, "pid": 5, "tid": 7, "ts": 1716454224543649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495115, "dur": 6,
    "args": {
      "External id": 194785,
      "cbid": 211, "correlation": 194785
    }
  },
  {
    "ph": "s", "id": 194785, "pid": 76337, "tid": -914061504, "ts": 1716454224495115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224543681, "dur": 20,
    "args": {
      "External id": 194794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194794,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194794, "pid": 5, "tid": 7, "ts": 1716454224543681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495156, "dur": 10,
    "args": {
      "External id": 194794,
      "cbid": 211, "correlation": 194794
    }
  },
  {
    "ph": "s", "id": 194794, "pid": 76337, "tid": -914061504, "ts": 1716454224495156,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495219, "dur": 0,
    "args": {
      "External id": 194804,
      "cbid": 317, "correlation": 194804
    }
  },
  {
    "ph": "f", "id": 194804, "pid": 76337, "tid": -914061504, "ts": 1716454224495219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495219, "dur": 0,
    "args": {
      "External id": 194805,
      "cbid": 203, "correlation": 194805
    }
  },
  {
    "ph": "f", "id": 194805, "pid": 76337, "tid": -914061504, "ts": 1716454224495219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495220, "dur": 0,
    "args": {
      "External id": 194806,
      "cbid": 205, "correlation": 194806
    }
  },
  {
    "ph": "f", "id": 194806, "pid": 76337, "tid": -914061504, "ts": 1716454224495220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224543702, "dur": 23,
    "args": {
      "External id": 194810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194810, "pid": 5, "tid": 7, "ts": 1716454224543702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495234, "dur": 12,
    "args": {
      "External id": 194810,
      "cbid": 211, "correlation": 194810
    }
  },
  {
    "ph": "s", "id": 194810, "pid": 76337, "tid": -914061504, "ts": 1716454224495234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224543726, "dur": 44,
    "args": {
      "External id": 194812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194812, "pid": 5, "tid": 7, "ts": 1716454224543726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495248, "dur": 5,
    "args": {
      "External id": 194812,
      "cbid": 211, "correlation": 194812
    }
  },
  {
    "ph": "s", "id": 194812, "pid": 76337, "tid": -914061504, "ts": 1716454224495248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224543772, "dur": 646,
    "args": {
      "External id": 194814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194814,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194814, "pid": 5, "tid": 7, "ts": 1716454224543772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495259, "dur": 6,
    "args": {
      "External id": 194814,
      "cbid": 211, "correlation": 194814
    }
  },
  {
    "ph": "s", "id": 194814, "pid": 76337, "tid": -914061504, "ts": 1716454224495259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224544419, "dur": 21,
    "args": {
      "External id": 194816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194816,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194816, "pid": 5, "tid": 7, "ts": 1716454224544419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495269, "dur": 5,
    "args": {
      "External id": 194816,
      "cbid": 211, "correlation": 194816
    }
  },
  {
    "ph": "s", "id": 194816, "pid": 76337, "tid": -914061504, "ts": 1716454224495269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224544442, "dur": 33,
    "args": {
      "External id": 194822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194822, "pid": 5, "tid": 7, "ts": 1716454224544442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495297, "dur": 8,
    "args": {
      "External id": 194822,
      "cbid": 211, "correlation": 194822
    }
  },
  {
    "ph": "s", "id": 194822, "pid": 76337, "tid": -914061504, "ts": 1716454224495297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224544476, "dur": 26,
    "args": {
      "External id": 194830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194830,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194830, "pid": 5, "tid": 7, "ts": 1716454224544476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495329, "dur": 8,
    "args": {
      "External id": 194830,
      "cbid": 211, "correlation": 194830
    }
  },
  {
    "ph": "s", "id": 194830, "pid": 76337, "tid": -914061504, "ts": 1716454224495329,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224544503, "dur": 20,
    "args": {
      "External id": 194838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194838,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194838, "pid": 5, "tid": 7, "ts": 1716454224544503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495359, "dur": 8,
    "args": {
      "External id": 194838,
      "cbid": 211, "correlation": 194838
    }
  },
  {
    "ph": "s", "id": 194838, "pid": 76337, "tid": -914061504, "ts": 1716454224495359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224544524, "dur": 30,
    "args": {
      "External id": 194858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194858,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 194858, "pid": 5, "tid": 7, "ts": 1716454224544524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495437, "dur": 12,
    "args": {
      "External id": 194858,
      "cbid": 211, "correlation": 194858
    }
  },
  {
    "ph": "s", "id": 194858, "pid": 76337, "tid": -914061504, "ts": 1716454224495437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224544555, "dur": 5,
    "args": {
      "External id": 194870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194870,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 194870, "pid": 5, "tid": 7, "ts": 1716454224544555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495459, "dur": 6,
    "args": {
      "External id": 194870,
      "cbid": 211, "correlation": 194870
    }
  },
  {
    "ph": "s", "id": 194870, "pid": 76337, "tid": -914061504, "ts": 1716454224495459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224544561, "dur": 30,
    "args": {
      "External id": 194873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194873,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194873, "pid": 5, "tid": 7, "ts": 1716454224544561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495477, "dur": 6,
    "args": {
      "External id": 194873,
      "cbid": 211, "correlation": 194873
    }
  },
  {
    "ph": "s", "id": 194873, "pid": 76337, "tid": -914061504, "ts": 1716454224495477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495535, "dur": 0,
    "args": {
      "External id": 194884,
      "cbid": 317, "correlation": 194884
    }
  },
  {
    "ph": "f", "id": 194884, "pid": 76337, "tid": -914061504, "ts": 1716454224495535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495535, "dur": 0,
    "args": {
      "External id": 194885,
      "cbid": 203, "correlation": 194885
    }
  },
  {
    "ph": "f", "id": 194885, "pid": 76337, "tid": -914061504, "ts": 1716454224495535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495536, "dur": 0,
    "args": {
      "External id": 194886,
      "cbid": 205, "correlation": 194886
    }
  },
  {
    "ph": "f", "id": 194886, "pid": 76337, "tid": -914061504, "ts": 1716454224495536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224544593, "dur": 22,
    "args": {
      "External id": 194890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194890,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194890, "pid": 5, "tid": 7, "ts": 1716454224544593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495549, "dur": 12,
    "args": {
      "External id": 194890,
      "cbid": 211, "correlation": 194890
    }
  },
  {
    "ph": "s", "id": 194890, "pid": 76337, "tid": -914061504, "ts": 1716454224495549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224544616, "dur": 116,
    "args": {
      "External id": 194892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194892,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194892, "pid": 5, "tid": 7, "ts": 1716454224544616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495567, "dur": 6,
    "args": {
      "External id": 194892,
      "cbid": 211, "correlation": 194892
    }
  },
  {
    "ph": "s", "id": 194892, "pid": 76337, "tid": -914061504, "ts": 1716454224495567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224544734, "dur": 22,
    "args": {
      "External id": 194894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194894,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194894, "pid": 5, "tid": 7, "ts": 1716454224544734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495577, "dur": 9,
    "args": {
      "External id": 194894,
      "cbid": 211, "correlation": 194894
    }
  },
  {
    "ph": "s", "id": 194894, "pid": 76337, "tid": -914061504, "ts": 1716454224495577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224544757, "dur": 32,
    "args": {
      "External id": 194900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194900, "pid": 5, "tid": 7, "ts": 1716454224544757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495608, "dur": 10,
    "args": {
      "External id": 194900,
      "cbid": 211, "correlation": 194900
    }
  },
  {
    "ph": "s", "id": 194900, "pid": 76337, "tid": -914061504, "ts": 1716454224495608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224544791, "dur": 171,
    "args": {
      "External id": 194909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194909,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194909, "pid": 5, "tid": 7, "ts": 1716454224544791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495690, "dur": 14,
    "args": {
      "External id": 194909,
      "cbid": 211, "correlation": 194909
    }
  },
  {
    "ph": "s", "id": 194909, "pid": 76337, "tid": -914061504, "ts": 1716454224495690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224544963, "dur": 65,
    "args": {
      "External id": 194931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194931,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 194931, "pid": 5, "tid": 7, "ts": 1716454224544963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495752, "dur": 11,
    "args": {
      "External id": 194931,
      "cbid": 211, "correlation": 194931
    }
  },
  {
    "ph": "s", "id": 194931, "pid": 76337, "tid": -914061504, "ts": 1716454224495752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495841, "dur": 2,
    "args": {
      "External id": 194942,
      "cbid": 251, "correlation": 194942
    }
  },
  {
    "ph": "f", "id": 194942, "pid": 76337, "tid": -914061504, "ts": 1716454224495841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224545029, "dur": 155,
    "args": {
      "External id": 194943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194943,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194943, "pid": 5, "tid": 7, "ts": 1716454224545029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495847, "dur": 13,
    "args": {
      "External id": 194943,
      "cbid": 211, "correlation": 194943
    }
  },
  {
    "ph": "s", "id": 194943, "pid": 76337, "tid": -914061504, "ts": 1716454224495847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495918, "dur": 1,
    "args": {
      "External id": 194954,
      "cbid": 251, "correlation": 194954
    }
  },
  {
    "ph": "f", "id": 194954, "pid": 76337, "tid": -914061504, "ts": 1716454224495918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224545185, "dur": 145,
    "args": {
      "External id": 194955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194955,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194955, "pid": 5, "tid": 7, "ts": 1716454224545185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495922, "dur": 11,
    "args": {
      "External id": 194955,
      "cbid": 211, "correlation": 194955
    }
  },
  {
    "ph": "s", "id": 194955, "pid": 76337, "tid": -914061504, "ts": 1716454224495922,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224495996, "dur": 1,
    "args": {
      "External id": 194966,
      "cbid": 251, "correlation": 194966
    }
  },
  {
    "ph": "f", "id": 194966, "pid": 76337, "tid": -914061504, "ts": 1716454224495996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224545332, "dur": 142,
    "args": {
      "External id": 194967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194967,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 194967, "pid": 5, "tid": 7, "ts": 1716454224545332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496000, "dur": 12,
    "args": {
      "External id": 194967,
      "cbid": 211, "correlation": 194967
    }
  },
  {
    "ph": "s", "id": 194967, "pid": 76337, "tid": -914061504, "ts": 1716454224496000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224545476, "dur": 1946,
    "args": {
      "External id": 194988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 194988,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 194988, "pid": 5, "tid": 7, "ts": 1716454224545476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496084, "dur": 13,
    "args": {
      "External id": 194988,
      "cbid": 211, "correlation": 194988
    }
  },
  {
    "ph": "s", "id": 194988, "pid": 76337, "tid": -914061504, "ts": 1716454224496084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496182, "dur": 1,
    "args": {
      "External id": 195006,
      "cbid": 251, "correlation": 195006
    }
  },
  {
    "ph": "f", "id": 195006, "pid": 76337, "tid": -914061504, "ts": 1716454224496182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224547424, "dur": 147,
    "args": {
      "External id": 195008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195008,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195008, "pid": 5, "tid": 7, "ts": 1716454224547424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496188, "dur": 13,
    "args": {
      "External id": 195008,
      "cbid": 211, "correlation": 195008
    }
  },
  {
    "ph": "s", "id": 195008, "pid": 76337, "tid": -914061504, "ts": 1716454224496188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224547572, "dur": 36,
    "args": {
      "External id": 195016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195016,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195016, "pid": 5, "tid": 7, "ts": 1716454224547572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496259, "dur": 13,
    "args": {
      "External id": 195016,
      "cbid": 211, "correlation": 195016
    }
  },
  {
    "ph": "s", "id": 195016, "pid": 76337, "tid": -914061504, "ts": 1716454224496259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224547609, "dur": 51,
    "args": {
      "External id": 195024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195024,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195024, "pid": 5, "tid": 7, "ts": 1716454224547609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496298, "dur": 9,
    "args": {
      "External id": 195024,
      "cbid": 211, "correlation": 195024
    }
  },
  {
    "ph": "s", "id": 195024, "pid": 76337, "tid": -914061504, "ts": 1716454224496298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224547661, "dur": 30,
    "args": {
      "External id": 195035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195035,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195035, "pid": 5, "tid": 7, "ts": 1716454224547661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496370, "dur": 12,
    "args": {
      "External id": 195035,
      "cbid": 211, "correlation": 195035
    }
  },
  {
    "ph": "s", "id": 195035, "pid": 76337, "tid": -914061504, "ts": 1716454224496370,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224547692, "dur": 34,
    "args": {
      "External id": 195057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195057,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195057, "pid": 5, "tid": 7, "ts": 1716454224547692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496400, "dur": 7,
    "args": {
      "External id": 195057,
      "cbid": 211, "correlation": 195057
    }
  },
  {
    "ph": "s", "id": 195057, "pid": 76337, "tid": -914061504, "ts": 1716454224496400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496489, "dur": 1,
    "args": {
      "External id": 195068,
      "cbid": 251, "correlation": 195068
    }
  },
  {
    "ph": "f", "id": 195068, "pid": 76337, "tid": -914061504, "ts": 1716454224496489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224547728, "dur": 89,
    "args": {
      "External id": 195069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195069,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195069, "pid": 5, "tid": 7, "ts": 1716454224547728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496494, "dur": 13,
    "args": {
      "External id": 195069,
      "cbid": 211, "correlation": 195069
    }
  },
  {
    "ph": "s", "id": 195069, "pid": 76337, "tid": -914061504, "ts": 1716454224496494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496562, "dur": 1,
    "args": {
      "External id": 195080,
      "cbid": 251, "correlation": 195080
    }
  },
  {
    "ph": "f", "id": 195080, "pid": 76337, "tid": -914061504, "ts": 1716454224496562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496567, "dur": 0,
    "args": {
      "External id": 195081,
      "cbid": 251, "correlation": 195081
    }
  },
  {
    "ph": "f", "id": 195081, "pid": 76337, "tid": -914061504, "ts": 1716454224496567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224547818, "dur": 12,
    "args": {
      "External id": 195082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195082,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 195082, "pid": 5, "tid": 7, "ts": 1716454224547818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496569, "dur": 12,
    "args": {
      "External id": 195082,
      "cbid": 211, "correlation": 195082
    }
  },
  {
    "ph": "s", "id": 195082, "pid": 76337, "tid": -914061504, "ts": 1716454224496569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224547832, "dur": 5,
    "args": {
      "External id": 195084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195084,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 195084, "pid": 5, "tid": 7, "ts": 1716454224547832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496582, "dur": 6,
    "args": {
      "External id": 195084,
      "cbid": 211, "correlation": 195084
    }
  },
  {
    "ph": "s", "id": 195084, "pid": 76337, "tid": -914061504, "ts": 1716454224496582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496640, "dur": 1,
    "args": {
      "External id": 195095,
      "cbid": 251, "correlation": 195095
    }
  },
  {
    "ph": "f", "id": 195095, "pid": 76337, "tid": -914061504, "ts": 1716454224496640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496644, "dur": 0,
    "args": {
      "External id": 195096,
      "cbid": 251, "correlation": 195096
    }
  },
  {
    "ph": "f", "id": 195096, "pid": 76337, "tid": -914061504, "ts": 1716454224496644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224547838, "dur": 7,
    "args": {
      "External id": 195097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195097,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 195097, "pid": 5, "tid": 7, "ts": 1716454224547838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496645, "dur": 12,
    "args": {
      "External id": 195097,
      "cbid": 211, "correlation": 195097
    }
  },
  {
    "ph": "s", "id": 195097, "pid": 76337, "tid": -914061504, "ts": 1716454224496645,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224547846, "dur": 4,
    "args": {
      "External id": 195099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195099,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 195099, "pid": 5, "tid": 7, "ts": 1716454224547846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496658, "dur": 6,
    "args": {
      "External id": 195099,
      "cbid": 211, "correlation": 195099
    }
  },
  {
    "ph": "s", "id": 195099, "pid": 76337, "tid": -914061504, "ts": 1716454224496658,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224547851, "dur": 94,
    "args": {
      "External id": 195120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195120,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 195120, "pid": 5, "tid": 7, "ts": 1716454224547851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496732, "dur": 12,
    "args": {
      "External id": 195120,
      "cbid": 211, "correlation": 195120
    }
  },
  {
    "ph": "s", "id": 195120, "pid": 76337, "tid": -914061504, "ts": 1716454224496732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496832, "dur": 1,
    "args": {
      "External id": 195138,
      "cbid": 251, "correlation": 195138
    }
  },
  {
    "ph": "f", "id": 195138, "pid": 76337, "tid": -914061504, "ts": 1716454224496832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224547946, "dur": 99,
    "args": {
      "External id": 195140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195140,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195140, "pid": 5, "tid": 7, "ts": 1716454224547946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496838, "dur": 13,
    "args": {
      "External id": 195140,
      "cbid": 211, "correlation": 195140
    }
  },
  {
    "ph": "s", "id": 195140, "pid": 76337, "tid": -914061504, "ts": 1716454224496838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224548047, "dur": 19,
    "args": {
      "External id": 195148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195148,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195148, "pid": 5, "tid": 7, "ts": 1716454224548047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496906, "dur": 12,
    "args": {
      "External id": 195148,
      "cbid": 211, "correlation": 195148
    }
  },
  {
    "ph": "s", "id": 195148, "pid": 76337, "tid": -914061504, "ts": 1716454224496906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224548068, "dur": 37,
    "args": {
      "External id": 195156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195156,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195156, "pid": 5, "tid": 7, "ts": 1716454224548068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224496948, "dur": 9,
    "args": {
      "External id": 195156,
      "cbid": 211, "correlation": 195156
    }
  },
  {
    "ph": "s", "id": 195156, "pid": 76337, "tid": -914061504, "ts": 1716454224496948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224548106, "dur": 35,
    "args": {
      "External id": 195178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195178,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195178, "pid": 5, "tid": 7, "ts": 1716454224548106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497008, "dur": 11,
    "args": {
      "External id": 195178,
      "cbid": 211, "correlation": 195178
    }
  },
  {
    "ph": "s", "id": 195178, "pid": 76337, "tid": -914061504, "ts": 1716454224497008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497098, "dur": 1,
    "args": {
      "External id": 195194,
      "cbid": 251, "correlation": 195194
    }
  },
  {
    "ph": "f", "id": 195194, "pid": 76337, "tid": -914061504, "ts": 1716454224497098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497103, "dur": 0,
    "args": {
      "External id": 195196,
      "cbid": 251, "correlation": 195196
    }
  },
  {
    "ph": "f", "id": 195196, "pid": 76337, "tid": -914061504, "ts": 1716454224497103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224548143, "dur": 540,
    "args": {
      "External id": 195197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195197,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195197, "pid": 5, "tid": 7, "ts": 1716454224548143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497106, "dur": 13,
    "args": {
      "External id": 195197,
      "cbid": 211, "correlation": 195197
    }
  },
  {
    "ph": "s", "id": 195197, "pid": 76337, "tid": -914061504, "ts": 1716454224497106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224548684, "dur": 126,
    "args": {
      "External id": 195205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195205, "pid": 5, "tid": 7, "ts": 1716454224548684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497171, "dur": 13,
    "args": {
      "External id": 195205,
      "cbid": 211, "correlation": 195205
    }
  },
  {
    "ph": "s", "id": 195205, "pid": 76337, "tid": -914061504, "ts": 1716454224497171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224548811, "dur": 128,
    "args": {
      "External id": 195213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195213, "pid": 5, "tid": 7, "ts": 1716454224548811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497201, "dur": 8,
    "args": {
      "External id": 195213,
      "cbid": 211, "correlation": 195213
    }
  },
  {
    "ph": "s", "id": 195213, "pid": 76337, "tid": -914061504, "ts": 1716454224497201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497278, "dur": 2,
    "args": {
      "External id": 195229,
      "cbid": 251, "correlation": 195229
    }
  },
  {
    "ph": "f", "id": 195229, "pid": 76337, "tid": -914061504, "ts": 1716454224497278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224548940, "dur": 302,
    "args": {
      "External id": 195231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195231,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195231, "pid": 5, "tid": 7, "ts": 1716454224548940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497284, "dur": 12,
    "args": {
      "External id": 195231,
      "cbid": 211, "correlation": 195231
    }
  },
  {
    "ph": "s", "id": 195231, "pid": 76337, "tid": -914061504, "ts": 1716454224497284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224549243, "dur": 27,
    "args": {
      "External id": 195239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195239,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195239, "pid": 5, "tid": 7, "ts": 1716454224549243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497325, "dur": 10,
    "args": {
      "External id": 195239,
      "cbid": 211, "correlation": 195239
    }
  },
  {
    "ph": "s", "id": 195239, "pid": 76337, "tid": -914061504, "ts": 1716454224497325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224549271, "dur": 81,
    "args": {
      "External id": 195250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195250,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195250, "pid": 5, "tid": 7, "ts": 1716454224549271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497393, "dur": 12,
    "args": {
      "External id": 195250,
      "cbid": 211, "correlation": 195250
    }
  },
  {
    "ph": "s", "id": 195250, "pid": 76337, "tid": -914061504, "ts": 1716454224497393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497456, "dur": 0,
    "args": {
      "External id": 195262,
      "cbid": 317, "correlation": 195262
    }
  },
  {
    "ph": "f", "id": 195262, "pid": 76337, "tid": -914061504, "ts": 1716454224497456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497457, "dur": 0,
    "args": {
      "External id": 195263,
      "cbid": 203, "correlation": 195263
    }
  },
  {
    "ph": "f", "id": 195263, "pid": 76337, "tid": -914061504, "ts": 1716454224497457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497457, "dur": 0,
    "args": {
      "External id": 195264,
      "cbid": 205, "correlation": 195264
    }
  },
  {
    "ph": "f", "id": 195264, "pid": 76337, "tid": -914061504, "ts": 1716454224497457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549354, "dur": 23,
    "args": {
      "External id": 195268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195268,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195268, "pid": 5, "tid": 7, "ts": 1716454224549354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497472, "dur": 13,
    "args": {
      "External id": 195268,
      "cbid": 211, "correlation": 195268
    }
  },
  {
    "ph": "s", "id": 195268, "pid": 76337, "tid": -914061504, "ts": 1716454224497472,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224549378, "dur": 121,
    "args": {
      "External id": 195270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195270,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195270, "pid": 5, "tid": 7, "ts": 1716454224549378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497491, "dur": 7,
    "args": {
      "External id": 195270,
      "cbid": 211, "correlation": 195270
    }
  },
  {
    "ph": "s", "id": 195270, "pid": 76337, "tid": -914061504, "ts": 1716454224497491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549500, "dur": 22,
    "args": {
      "External id": 195272,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195272,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195272, "pid": 5, "tid": 7, "ts": 1716454224549500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497502, "dur": 5,
    "args": {
      "External id": 195272,
      "cbid": 211, "correlation": 195272
    }
  },
  {
    "ph": "s", "id": 195272, "pid": 76337, "tid": -914061504, "ts": 1716454224497502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224549523, "dur": 32,
    "args": {
      "External id": 195278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195278,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195278, "pid": 5, "tid": 7, "ts": 1716454224549523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497530, "dur": 9,
    "args": {
      "External id": 195278,
      "cbid": 211, "correlation": 195278
    }
  },
  {
    "ph": "s", "id": 195278, "pid": 76337, "tid": -914061504, "ts": 1716454224497530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224549556, "dur": 26,
    "args": {
      "External id": 195286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195286,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195286, "pid": 5, "tid": 7, "ts": 1716454224549556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497561, "dur": 8,
    "args": {
      "External id": 195286,
      "cbid": 211, "correlation": 195286
    }
  },
  {
    "ph": "s", "id": 195286, "pid": 76337, "tid": -914061504, "ts": 1716454224497561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497637, "dur": 0,
    "args": {
      "External id": 195296,
      "cbid": 317, "correlation": 195296
    }
  },
  {
    "ph": "f", "id": 195296, "pid": 76337, "tid": -914061504, "ts": 1716454224497637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497638, "dur": 0,
    "args": {
      "External id": 195297,
      "cbid": 203, "correlation": 195297
    }
  },
  {
    "ph": "f", "id": 195297, "pid": 76337, "tid": -914061504, "ts": 1716454224497638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497638, "dur": 0,
    "args": {
      "External id": 195298,
      "cbid": 205, "correlation": 195298
    }
  },
  {
    "ph": "f", "id": 195298, "pid": 76337, "tid": -914061504, "ts": 1716454224497638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549584, "dur": 24,
    "args": {
      "External id": 195302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195302,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195302, "pid": 5, "tid": 7, "ts": 1716454224549584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497652, "dur": 12,
    "args": {
      "External id": 195302,
      "cbid": 211, "correlation": 195302
    }
  },
  {
    "ph": "s", "id": 195302, "pid": 76337, "tid": -914061504, "ts": 1716454224497652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549610, "dur": 44,
    "args": {
      "External id": 195304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195304,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195304, "pid": 5, "tid": 7, "ts": 1716454224549610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497667, "dur": 5,
    "args": {
      "External id": 195304,
      "cbid": 211, "correlation": 195304
    }
  },
  {
    "ph": "s", "id": 195304, "pid": 76337, "tid": -914061504, "ts": 1716454224497667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224549656, "dur": 234,
    "args": {
      "External id": 195306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195306,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 195306, "pid": 5, "tid": 7, "ts": 1716454224549656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497679, "dur": 7,
    "args": {
      "External id": 195306,
      "cbid": 211, "correlation": 195306
    }
  },
  {
    "ph": "s", "id": 195306, "pid": 76337, "tid": -914061504, "ts": 1716454224497679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549891, "dur": 7,
    "args": {
      "External id": 195308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195308,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195308, "pid": 5, "tid": 7, "ts": 1716454224549891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497690, "dur": 5,
    "args": {
      "External id": 195308,
      "cbid": 211, "correlation": 195308
    }
  },
  {
    "ph": "s", "id": 195308, "pid": 76337, "tid": -914061504, "ts": 1716454224497690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224549899, "dur": 9,
    "args": {
      "External id": 195314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195314, "pid": 5, "tid": 7, "ts": 1716454224549899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497716, "dur": 8,
    "args": {
      "External id": 195314,
      "cbid": 211, "correlation": 195314
    }
  },
  {
    "ph": "s", "id": 195314, "pid": 76337, "tid": -914061504, "ts": 1716454224497716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549909, "dur": 12,
    "args": {
      "External id": 195334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195334,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 195334, "pid": 5, "tid": 7, "ts": 1716454224549909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497809, "dur": 12,
    "args": {
      "External id": 195334,
      "cbid": 211, "correlation": 195334
    }
  },
  {
    "ph": "s", "id": 195334, "pid": 76337, "tid": -914061504, "ts": 1716454224497809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224549923, "dur": 5,
    "args": {
      "External id": 195346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195346,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 195346, "pid": 5, "tid": 7, "ts": 1716454224549923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497831, "dur": 7,
    "args": {
      "External id": 195346,
      "cbid": 211, "correlation": 195346
    }
  },
  {
    "ph": "s", "id": 195346, "pid": 76337, "tid": -914061504, "ts": 1716454224497831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224549929, "dur": 12,
    "args": {
      "External id": 195349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195349,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195349, "pid": 5, "tid": 7, "ts": 1716454224549929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497849, "dur": 7,
    "args": {
      "External id": 195349,
      "cbid": 211, "correlation": 195349
    }
  },
  {
    "ph": "s", "id": 195349, "pid": 76337, "tid": -914061504, "ts": 1716454224497849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224549942, "dur": 7,
    "args": {
      "External id": 195358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195358,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195358, "pid": 5, "tid": 7, "ts": 1716454224549942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497889, "dur": 10,
    "args": {
      "External id": 195358,
      "cbid": 211, "correlation": 195358
    }
  },
  {
    "ph": "s", "id": 195358, "pid": 76337, "tid": -914061504, "ts": 1716454224497889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497945, "dur": 0,
    "args": {
      "External id": 195368,
      "cbid": 317, "correlation": 195368
    }
  },
  {
    "ph": "f", "id": 195368, "pid": 76337, "tid": -914061504, "ts": 1716454224497945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497945, "dur": 0,
    "args": {
      "External id": 195369,
      "cbid": 203, "correlation": 195369
    }
  },
  {
    "ph": "f", "id": 195369, "pid": 76337, "tid": -914061504, "ts": 1716454224497945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497946, "dur": 0,
    "args": {
      "External id": 195370,
      "cbid": 205, "correlation": 195370
    }
  },
  {
    "ph": "f", "id": 195370, "pid": 76337, "tid": -914061504, "ts": 1716454224497946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549950, "dur": 6,
    "args": {
      "External id": 195374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195374,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195374, "pid": 5, "tid": 7, "ts": 1716454224549950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497962, "dur": 19,
    "args": {
      "External id": 195374,
      "cbid": 211, "correlation": 195374
    }
  },
  {
    "ph": "s", "id": 195374, "pid": 76337, "tid": -914061504, "ts": 1716454224497962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224549957, "dur": 84,
    "args": {
      "External id": 195376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195376,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195376, "pid": 5, "tid": 7, "ts": 1716454224549957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497984, "dur": 6,
    "args": {
      "External id": 195376,
      "cbid": 211, "correlation": 195376
    }
  },
  {
    "ph": "s", "id": 195376, "pid": 76337, "tid": -914061504, "ts": 1716454224497984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224550043, "dur": 1,
    "args": {
      "External id": 195378,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 195378,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 195378, "pid": 5, "tid": 7, "ts": 1716454224550043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224497997, "dur": 9,
    "args": {
      "External id": 195378,
      "cbid": 51, "correlation": 195378
    }
  },
  {
    "ph": "s", "id": 195378, "pid": 76337, "tid": -914061504, "ts": 1716454224497997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224550047, "dur": 543,
    "args": {
      "External id": 195379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195379,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195379, "pid": 5, "tid": 7, "ts": 1716454224550047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498007, "dur": 8,
    "args": {
      "External id": 195379,
      "cbid": 211, "correlation": 195379
    }
  },
  {
    "ph": "s", "id": 195379, "pid": 76337, "tid": -914061504, "ts": 1716454224498007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224550591, "dur": 12,
    "args": {
      "External id": 195381,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195381,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195381, "pid": 5, "tid": 7, "ts": 1716454224550591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498019, "dur": 5,
    "args": {
      "External id": 195381,
      "cbid": 211, "correlation": 195381
    }
  },
  {
    "ph": "s", "id": 195381, "pid": 76337, "tid": -914061504, "ts": 1716454224498019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224550605, "dur": 14,
    "args": {
      "External id": 195387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195387,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195387, "pid": 5, "tid": 7, "ts": 1716454224550605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498047, "dur": 8,
    "args": {
      "External id": 195387,
      "cbid": 211, "correlation": 195387
    }
  },
  {
    "ph": "s", "id": 195387, "pid": 76337, "tid": -914061504, "ts": 1716454224498047,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224550620, "dur": 3,
    "args": {
      "External id": 195395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195395,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 195395, "pid": 5, "tid": 7, "ts": 1716454224550620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498095, "dur": 10,
    "args": {
      "External id": 195395,
      "cbid": 211, "correlation": 195395
    }
  },
  {
    "ph": "s", "id": 195395, "pid": 76337, "tid": -914061504, "ts": 1716454224498095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498159, "dur": 1,
    "args": {
      "External id": 195411,
      "cbid": 251, "correlation": 195411
    }
  },
  {
    "ph": "f", "id": 195411, "pid": 76337, "tid": -914061504, "ts": 1716454224498159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498164, "dur": 0,
    "args": {
      "External id": 195413,
      "cbid": 251, "correlation": 195413
    }
  },
  {
    "ph": "f", "id": 195413, "pid": 76337, "tid": -914061504, "ts": 1716454224498164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224550625, "dur": 13,
    "args": {
      "External id": 195414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195414,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195414, "pid": 5, "tid": 7, "ts": 1716454224550625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498166, "dur": 11,
    "args": {
      "External id": 195414,
      "cbid": 211, "correlation": 195414
    }
  },
  {
    "ph": "s", "id": 195414, "pid": 76337, "tid": -914061504, "ts": 1716454224498166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224550639, "dur": 5,
    "args": {
      "External id": 195416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195416,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195416, "pid": 5, "tid": 7, "ts": 1716454224550639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498179, "dur": 6,
    "args": {
      "External id": 195416,
      "cbid": 211, "correlation": 195416
    }
  },
  {
    "ph": "s", "id": 195416, "pid": 76337, "tid": -914061504, "ts": 1716454224498179,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224550646, "dur": 17,
    "args": {
      "External id": 195426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195426,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195426, "pid": 5, "tid": 7, "ts": 1716454224550646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498237, "dur": 12,
    "args": {
      "External id": 195426,
      "cbid": 211, "correlation": 195426
    }
  },
  {
    "ph": "s", "id": 195426, "pid": 76337, "tid": -914061504, "ts": 1716454224498237,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224550664, "dur": 18,
    "args": {
      "External id": 195446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195446,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 195446, "pid": 5, "tid": 7, "ts": 1716454224550664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498303, "dur": 10,
    "args": {
      "External id": 195446,
      "cbid": 211, "correlation": 195446
    }
  },
  {
    "ph": "s", "id": 195446, "pid": 76337, "tid": -914061504, "ts": 1716454224498303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224550684, "dur": 4,
    "args": {
      "External id": 195458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195458,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 195458, "pid": 5, "tid": 7, "ts": 1716454224550684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498324, "dur": 6,
    "args": {
      "External id": 195458,
      "cbid": 211, "correlation": 195458
    }
  },
  {
    "ph": "s", "id": 195458, "pid": 76337, "tid": -914061504, "ts": 1716454224498324,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224550690, "dur": 17,
    "args": {
      "External id": 195461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195461,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195461, "pid": 5, "tid": 7, "ts": 1716454224550690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498343, "dur": 6,
    "args": {
      "External id": 195461,
      "cbid": 211, "correlation": 195461
    }
  },
  {
    "ph": "s", "id": 195461, "pid": 76337, "tid": -914061504, "ts": 1716454224498343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224550708, "dur": 11,
    "args": {
      "External id": 195470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195470,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195470, "pid": 5, "tid": 7, "ts": 1716454224550708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498383, "dur": 11,
    "args": {
      "External id": 195470,
      "cbid": 211, "correlation": 195470
    }
  },
  {
    "ph": "s", "id": 195470, "pid": 76337, "tid": -914061504, "ts": 1716454224498383,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498446, "dur": 0,
    "args": {
      "External id": 195480,
      "cbid": 317, "correlation": 195480
    }
  },
  {
    "ph": "f", "id": 195480, "pid": 76337, "tid": -914061504, "ts": 1716454224498446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498447, "dur": 0,
    "args": {
      "External id": 195481,
      "cbid": 203, "correlation": 195481
    }
  },
  {
    "ph": "f", "id": 195481, "pid": 76337, "tid": -914061504, "ts": 1716454224498447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498448, "dur": 0,
    "args": {
      "External id": 195482,
      "cbid": 205, "correlation": 195482
    }
  },
  {
    "ph": "f", "id": 195482, "pid": 76337, "tid": -914061504, "ts": 1716454224498448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224550720, "dur": 11,
    "args": {
      "External id": 195486,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195486,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195486, "pid": 5, "tid": 7, "ts": 1716454224550720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498465, "dur": 12,
    "args": {
      "External id": 195486,
      "cbid": 211, "correlation": 195486
    }
  },
  {
    "ph": "s", "id": 195486, "pid": 76337, "tid": -914061504, "ts": 1716454224498465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224550733, "dur": 163,
    "args": {
      "External id": 195488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195488,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195488, "pid": 5, "tid": 7, "ts": 1716454224550733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498479, "dur": 5,
    "args": {
      "External id": 195488,
      "cbid": 211, "correlation": 195488
    }
  },
  {
    "ph": "s", "id": 195488, "pid": 76337, "tid": -914061504, "ts": 1716454224498479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224550898, "dur": 1,
    "args": {
      "External id": 195490,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 195490,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 195490, "pid": 5, "tid": 7, "ts": 1716454224550898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498491, "dur": 6,
    "args": {
      "External id": 195490,
      "cbid": 51, "correlation": 195490
    }
  },
  {
    "ph": "s", "id": 195490, "pid": 76337, "tid": -914061504, "ts": 1716454224498491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224550902, "dur": 662,
    "args": {
      "External id": 195491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195491,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195491, "pid": 5, "tid": 7, "ts": 1716454224550902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498499, "dur": 6,
    "args": {
      "External id": 195491,
      "cbid": 211, "correlation": 195491
    }
  },
  {
    "ph": "s", "id": 195491, "pid": 76337, "tid": -914061504, "ts": 1716454224498499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224551565, "dur": 12,
    "args": {
      "External id": 195493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195493,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195493, "pid": 5, "tid": 7, "ts": 1716454224551565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498509, "dur": 5,
    "args": {
      "External id": 195493,
      "cbid": 211, "correlation": 195493
    }
  },
  {
    "ph": "s", "id": 195493, "pid": 76337, "tid": -914061504, "ts": 1716454224498509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224551578, "dur": 14,
    "args": {
      "External id": 195499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195499,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195499, "pid": 5, "tid": 7, "ts": 1716454224551578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498538, "dur": 9,
    "args": {
      "External id": 195499,
      "cbid": 211, "correlation": 195499
    }
  },
  {
    "ph": "s", "id": 195499, "pid": 76337, "tid": -914061504, "ts": 1716454224498538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498596, "dur": 0,
    "args": {
      "External id": 195509,
      "cbid": 317, "correlation": 195509
    }
  },
  {
    "ph": "f", "id": 195509, "pid": 76337, "tid": -914061504, "ts": 1716454224498596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498597, "dur": 0,
    "args": {
      "External id": 195510,
      "cbid": 203, "correlation": 195510
    }
  },
  {
    "ph": "f", "id": 195510, "pid": 76337, "tid": -914061504, "ts": 1716454224498597,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498598, "dur": 0,
    "args": {
      "External id": 195511,
      "cbid": 205, "correlation": 195511
    }
  },
  {
    "ph": "f", "id": 195511, "pid": 76337, "tid": -914061504, "ts": 1716454224498598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224551594, "dur": 8,
    "args": {
      "External id": 195515,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195515,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195515, "pid": 5, "tid": 7, "ts": 1716454224551594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498612, "dur": 11,
    "args": {
      "External id": 195515,
      "cbid": 211, "correlation": 195515
    }
  },
  {
    "ph": "s", "id": 195515, "pid": 76337, "tid": -914061504, "ts": 1716454224498612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224551604, "dur": 4,
    "args": {
      "External id": 195517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195517,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195517, "pid": 5, "tid": 7, "ts": 1716454224551604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498628, "dur": 6,
    "args": {
      "External id": 195517,
      "cbid": 211, "correlation": 195517
    }
  },
  {
    "ph": "s", "id": 195517, "pid": 76337, "tid": -914061504, "ts": 1716454224498628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498638, "dur": 0,
    "args": {
      "External id": 195518,
      "cbid": 51, "correlation": 195518
    }
  },
  {
    "ph": "s", "id": 195518, "pid": 76337, "tid": -914061504, "ts": 1716454224498638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224551609, "dur": 57,
    "args": {
      "External id": 195519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195519,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 195519, "pid": 5, "tid": 7, "ts": 1716454224551609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498639, "dur": 5,
    "args": {
      "External id": 195519,
      "cbid": 211, "correlation": 195519
    }
  },
  {
    "ph": "s", "id": 195519, "pid": 76337, "tid": -914061504, "ts": 1716454224498639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224551667, "dur": 14,
    "args": {
      "External id": 195524,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195524,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195524, "pid": 5, "tid": 7, "ts": 1716454224551667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498664, "dur": 8,
    "args": {
      "External id": 195524,
      "cbid": 211, "correlation": 195524
    }
  },
  {
    "ph": "s", "id": 195524, "pid": 76337, "tid": -914061504, "ts": 1716454224498664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224551682, "dur": 12,
    "args": {
      "External id": 195532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195532,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195532, "pid": 5, "tid": 7, "ts": 1716454224551682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498694, "dur": 8,
    "args": {
      "External id": 195532,
      "cbid": 211, "correlation": 195532
    }
  },
  {
    "ph": "s", "id": 195532, "pid": 76337, "tid": -914061504, "ts": 1716454224498694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224551696, "dur": 10,
    "args": {
      "External id": 195540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195540,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195540, "pid": 5, "tid": 7, "ts": 1716454224551696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498722, "dur": 8,
    "args": {
      "External id": 195540,
      "cbid": 211, "correlation": 195540
    }
  },
  {
    "ph": "s", "id": 195540, "pid": 76337, "tid": -914061504, "ts": 1716454224498722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224551707, "dur": 19,
    "args": {
      "External id": 195560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195560,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 195560, "pid": 5, "tid": 7, "ts": 1716454224551707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498804, "dur": 12,
    "args": {
      "External id": 195560,
      "cbid": 211, "correlation": 195560
    }
  },
  {
    "ph": "s", "id": 195560, "pid": 76337, "tid": -914061504, "ts": 1716454224498804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224551727, "dur": 5,
    "args": {
      "External id": 195572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195572,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 195572, "pid": 5, "tid": 7, "ts": 1716454224551727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498826, "dur": 6,
    "args": {
      "External id": 195572,
      "cbid": 211, "correlation": 195572
    }
  },
  {
    "ph": "s", "id": 195572, "pid": 76337, "tid": -914061504, "ts": 1716454224498826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224551734, "dur": 17,
    "args": {
      "External id": 195575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195575, "pid": 5, "tid": 7, "ts": 1716454224551734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498844, "dur": 6,
    "args": {
      "External id": 195575,
      "cbid": 211, "correlation": 195575
    }
  },
  {
    "ph": "s", "id": 195575, "pid": 76337, "tid": -914061504, "ts": 1716454224498844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498904, "dur": 0,
    "args": {
      "External id": 195586,
      "cbid": 317, "correlation": 195586
    }
  },
  {
    "ph": "f", "id": 195586, "pid": 76337, "tid": -914061504, "ts": 1716454224498904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498904, "dur": 0,
    "args": {
      "External id": 195587,
      "cbid": 203, "correlation": 195587
    }
  },
  {
    "ph": "f", "id": 195587, "pid": 76337, "tid": -914061504, "ts": 1716454224498904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498905, "dur": 0,
    "args": {
      "External id": 195588,
      "cbid": 205, "correlation": 195588
    }
  },
  {
    "ph": "f", "id": 195588, "pid": 76337, "tid": -914061504, "ts": 1716454224498905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224551752, "dur": 12,
    "args": {
      "External id": 195592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195592,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195592, "pid": 5, "tid": 7, "ts": 1716454224551752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498921, "dur": 12,
    "args": {
      "External id": 195592,
      "cbid": 211, "correlation": 195592
    }
  },
  {
    "ph": "s", "id": 195592, "pid": 76337, "tid": -914061504, "ts": 1716454224498921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224551765, "dur": 3,
    "args": {
      "External id": 195594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195594,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195594, "pid": 5, "tid": 7, "ts": 1716454224551765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498937, "dur": 9,
    "args": {
      "External id": 195594,
      "cbid": 211, "correlation": 195594
    }
  },
  {
    "ph": "s", "id": 195594, "pid": 76337, "tid": -914061504, "ts": 1716454224498937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498950, "dur": 0,
    "args": {
      "External id": 195595,
      "cbid": 51, "correlation": 195595
    }
  },
  {
    "ph": "s", "id": 195595, "pid": 76337, "tid": -914061504, "ts": 1716454224498950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224551770, "dur": 97,
    "args": {
      "External id": 195596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195596,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 195596, "pid": 5, "tid": 7, "ts": 1716454224551770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498951, "dur": 6,
    "args": {
      "External id": 195596,
      "cbid": 211, "correlation": 195596
    }
  },
  {
    "ph": "s", "id": 195596, "pid": 76337, "tid": -914061504, "ts": 1716454224498951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224551868, "dur": 15,
    "args": {
      "External id": 195601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195601,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195601, "pid": 5, "tid": 7, "ts": 1716454224551868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224498987, "dur": 10,
    "args": {
      "External id": 195601,
      "cbid": 211, "correlation": 195601
    }
  },
  {
    "ph": "s", "id": 195601, "pid": 76337, "tid": -914061504, "ts": 1716454224498987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224551885, "dur": 84,
    "args": {
      "External id": 195610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195610,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195610, "pid": 5, "tid": 7, "ts": 1716454224551885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499070, "dur": 14,
    "args": {
      "External id": 195610,
      "cbid": 211, "correlation": 195610
    }
  },
  {
    "ph": "s", "id": 195610, "pid": 76337, "tid": -914061504, "ts": 1716454224499070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224551970, "dur": 31,
    "args": {
      "External id": 195632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195632,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195632, "pid": 5, "tid": 7, "ts": 1716454224551970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499126, "dur": 11,
    "args": {
      "External id": 195632,
      "cbid": 211, "correlation": 195632
    }
  },
  {
    "ph": "s", "id": 195632, "pid": 76337, "tid": -914061504, "ts": 1716454224499126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499219, "dur": 1,
    "args": {
      "External id": 195643,
      "cbid": 251, "correlation": 195643
    }
  },
  {
    "ph": "f", "id": 195643, "pid": 76337, "tid": -914061504, "ts": 1716454224499219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224552003, "dur": 162,
    "args": {
      "External id": 195644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195644,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195644, "pid": 5, "tid": 7, "ts": 1716454224552003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499224, "dur": 14,
    "args": {
      "External id": 195644,
      "cbid": 211, "correlation": 195644
    }
  },
  {
    "ph": "s", "id": 195644, "pid": 76337, "tid": -914061504, "ts": 1716454224499224,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499298, "dur": 1,
    "args": {
      "External id": 195655,
      "cbid": 251, "correlation": 195655
    }
  },
  {
    "ph": "f", "id": 195655, "pid": 76337, "tid": -914061504, "ts": 1716454224499298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224552166, "dur": 158,
    "args": {
      "External id": 195656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195656,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195656, "pid": 5, "tid": 7, "ts": 1716454224552166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499302, "dur": 11,
    "args": {
      "External id": 195656,
      "cbid": 211, "correlation": 195656
    }
  },
  {
    "ph": "s", "id": 195656, "pid": 76337, "tid": -914061504, "ts": 1716454224499302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499367, "dur": 1,
    "args": {
      "External id": 195667,
      "cbid": 251, "correlation": 195667
    }
  },
  {
    "ph": "f", "id": 195667, "pid": 76337, "tid": -914061504, "ts": 1716454224499367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224552325, "dur": 159,
    "args": {
      "External id": 195668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195668,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195668, "pid": 5, "tid": 7, "ts": 1716454224552325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499371, "dur": 11,
    "args": {
      "External id": 195668,
      "cbid": 211, "correlation": 195668
    }
  },
  {
    "ph": "s", "id": 195668, "pid": 76337, "tid": -914061504, "ts": 1716454224499371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224552485, "dur": 339,
    "args": {
      "External id": 195693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195693,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195693, "pid": 5, "tid": 7, "ts": 1716454224552485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499457, "dur": 13,
    "args": {
      "External id": 195693,
      "cbid": 211, "correlation": 195693
    }
  },
  {
    "ph": "s", "id": 195693, "pid": 76337, "tid": -914061504, "ts": 1716454224499457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499561, "dur": 1,
    "args": {
      "External id": 195711,
      "cbid": 251, "correlation": 195711
    }
  },
  {
    "ph": "f", "id": 195711, "pid": 76337, "tid": -914061504, "ts": 1716454224499561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224552825, "dur": 164,
    "args": {
      "External id": 195713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195713,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195713, "pid": 5, "tid": 7, "ts": 1716454224552825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499567, "dur": 13,
    "args": {
      "External id": 195713,
      "cbid": 211, "correlation": 195713
    }
  },
  {
    "ph": "s", "id": 195713, "pid": 76337, "tid": -914061504, "ts": 1716454224499567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224552991, "dur": 20,
    "args": {
      "External id": 195721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195721,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195721, "pid": 5, "tid": 7, "ts": 1716454224552991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499636, "dur": 13,
    "args": {
      "External id": 195721,
      "cbid": 211, "correlation": 195721
    }
  },
  {
    "ph": "s", "id": 195721, "pid": 76337, "tid": -914061504, "ts": 1716454224499636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224553012, "dur": 27,
    "args": {
      "External id": 195729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195729,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195729, "pid": 5, "tid": 7, "ts": 1716454224553012,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499675, "dur": 8,
    "args": {
      "External id": 195729,
      "cbid": 211, "correlation": 195729
    }
  },
  {
    "ph": "s", "id": 195729, "pid": 76337, "tid": -914061504, "ts": 1716454224499675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224553040, "dur": 19,
    "args": {
      "External id": 195740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195740,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195740, "pid": 5, "tid": 7, "ts": 1716454224553040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499748, "dur": 13,
    "args": {
      "External id": 195740,
      "cbid": 211, "correlation": 195740
    }
  },
  {
    "ph": "s", "id": 195740, "pid": 76337, "tid": -914061504, "ts": 1716454224499748,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224553061, "dur": 16,
    "args": {
      "External id": 195762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195762,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195762, "pid": 5, "tid": 7, "ts": 1716454224553061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499783, "dur": 8,
    "args": {
      "External id": 195762,
      "cbid": 211, "correlation": 195762
    }
  },
  {
    "ph": "s", "id": 195762, "pid": 76337, "tid": -914061504, "ts": 1716454224499783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499871, "dur": 2,
    "args": {
      "External id": 195773,
      "cbid": 251, "correlation": 195773
    }
  },
  {
    "ph": "f", "id": 195773, "pid": 76337, "tid": -914061504, "ts": 1716454224499871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224553078, "dur": 89,
    "args": {
      "External id": 195774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195774,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195774, "pid": 5, "tid": 7, "ts": 1716454224553078,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499878, "dur": 15,
    "args": {
      "External id": 195774,
      "cbid": 211, "correlation": 195774
    }
  },
  {
    "ph": "s", "id": 195774, "pid": 76337, "tid": -914061504, "ts": 1716454224499878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499953, "dur": 1,
    "args": {
      "External id": 195785,
      "cbid": 251, "correlation": 195785
    }
  },
  {
    "ph": "f", "id": 195785, "pid": 76337, "tid": -914061504, "ts": 1716454224499953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499956, "dur": 0,
    "args": {
      "External id": 195786,
      "cbid": 251, "correlation": 195786
    }
  },
  {
    "ph": "f", "id": 195786, "pid": 76337, "tid": -914061504, "ts": 1716454224499956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224553168, "dur": 12,
    "args": {
      "External id": 195787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195787,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195787, "pid": 5, "tid": 7, "ts": 1716454224553168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499958, "dur": 12,
    "args": {
      "External id": 195787,
      "cbid": 211, "correlation": 195787
    }
  },
  {
    "ph": "s", "id": 195787, "pid": 76337, "tid": -914061504, "ts": 1716454224499958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224553182, "dur": 5,
    "args": {
      "External id": 195789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195789,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195789, "pid": 5, "tid": 7, "ts": 1716454224553182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224499971, "dur": 13,
    "args": {
      "External id": 195789,
      "cbid": 211, "correlation": 195789
    }
  },
  {
    "ph": "s", "id": 195789, "pid": 76337, "tid": -914061504, "ts": 1716454224499971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500039, "dur": 1,
    "args": {
      "External id": 195800,
      "cbid": 251, "correlation": 195800
    }
  },
  {
    "ph": "f", "id": 195800, "pid": 76337, "tid": -914061504, "ts": 1716454224500039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500043, "dur": 0,
    "args": {
      "External id": 195801,
      "cbid": 251, "correlation": 195801
    }
  },
  {
    "ph": "f", "id": 195801, "pid": 76337, "tid": -914061504, "ts": 1716454224500043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224553188, "dur": 8,
    "args": {
      "External id": 195802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195802,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195802, "pid": 5, "tid": 7, "ts": 1716454224553188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500045, "dur": 12,
    "args": {
      "External id": 195802,
      "cbid": 211, "correlation": 195802
    }
  },
  {
    "ph": "s", "id": 195802, "pid": 76337, "tid": -914061504, "ts": 1716454224500045,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224553198, "dur": 3,
    "args": {
      "External id": 195804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195804,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195804, "pid": 5, "tid": 7, "ts": 1716454224553198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500058, "dur": 5,
    "args": {
      "External id": 195804,
      "cbid": 211, "correlation": 195804
    }
  },
  {
    "ph": "s", "id": 195804, "pid": 76337, "tid": -914061504, "ts": 1716454224500058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224553203, "dur": 55,
    "args": {
      "External id": 195829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195829,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195829, "pid": 5, "tid": 7, "ts": 1716454224553203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500134, "dur": 13,
    "args": {
      "External id": 195829,
      "cbid": 211, "correlation": 195829
    }
  },
  {
    "ph": "s", "id": 195829, "pid": 76337, "tid": -914061504, "ts": 1716454224500134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500236, "dur": 2,
    "args": {
      "External id": 195847,
      "cbid": 251, "correlation": 195847
    }
  },
  {
    "ph": "f", "id": 195847, "pid": 76337, "tid": -914061504, "ts": 1716454224500236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224553259, "dur": 91,
    "args": {
      "External id": 195849,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195849,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195849, "pid": 5, "tid": 7, "ts": 1716454224553259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500244, "dur": 15,
    "args": {
      "External id": 195849,
      "cbid": 211, "correlation": 195849
    }
  },
  {
    "ph": "s", "id": 195849, "pid": 76337, "tid": -914061504, "ts": 1716454224500244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224553352, "dur": 10,
    "args": {
      "External id": 195857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195857,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195857, "pid": 5, "tid": 7, "ts": 1716454224553352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500315, "dur": 12,
    "args": {
      "External id": 195857,
      "cbid": 211, "correlation": 195857
    }
  },
  {
    "ph": "s", "id": 195857, "pid": 76337, "tid": -914061504, "ts": 1716454224500315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224553363, "dur": 20,
    "args": {
      "External id": 195865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195865,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195865, "pid": 5, "tid": 7, "ts": 1716454224553363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500357, "dur": 9,
    "args": {
      "External id": 195865,
      "cbid": 211, "correlation": 195865
    }
  },
  {
    "ph": "s", "id": 195865, "pid": 76337, "tid": -914061504, "ts": 1716454224500357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224553385, "dur": 17,
    "args": {
      "External id": 195887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195887,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195887, "pid": 5, "tid": 7, "ts": 1716454224553385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500411, "dur": 10,
    "args": {
      "External id": 195887,
      "cbid": 211, "correlation": 195887
    }
  },
  {
    "ph": "s", "id": 195887, "pid": 76337, "tid": -914061504, "ts": 1716454224500411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500505, "dur": 2,
    "args": {
      "External id": 195903,
      "cbid": 251, "correlation": 195903
    }
  },
  {
    "ph": "f", "id": 195903, "pid": 76337, "tid": -914061504, "ts": 1716454224500505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500510, "dur": 0,
    "args": {
      "External id": 195905,
      "cbid": 251, "correlation": 195905
    }
  },
  {
    "ph": "f", "id": 195905, "pid": 76337, "tid": -914061504, "ts": 1716454224500510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224553403, "dur": 497,
    "args": {
      "External id": 195906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195906,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 195906, "pid": 5, "tid": 7, "ts": 1716454224553403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500513, "dur": 15,
    "args": {
      "External id": 195906,
      "cbid": 211, "correlation": 195906
    }
  },
  {
    "ph": "s", "id": 195906, "pid": 76337, "tid": -914061504, "ts": 1716454224500513,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224553901, "dur": 66,
    "args": {
      "External id": 195914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195914,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195914, "pid": 5, "tid": 7, "ts": 1716454224553901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500581, "dur": 13,
    "args": {
      "External id": 195914,
      "cbid": 211, "correlation": 195914
    }
  },
  {
    "ph": "s", "id": 195914, "pid": 76337, "tid": -914061504, "ts": 1716454224500581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224553969, "dur": 69,
    "args": {
      "External id": 195922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195922,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195922, "pid": 5, "tid": 7, "ts": 1716454224553969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500612, "dur": 8,
    "args": {
      "External id": 195922,
      "cbid": 211, "correlation": 195922
    }
  },
  {
    "ph": "s", "id": 195922, "pid": 76337, "tid": -914061504, "ts": 1716454224500612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500693, "dur": 1,
    "args": {
      "External id": 195938,
      "cbid": 251, "correlation": 195938
    }
  },
  {
    "ph": "f", "id": 195938, "pid": 76337, "tid": -914061504, "ts": 1716454224500693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224554040, "dur": 1,
    "args": {
      "External id": 195940,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 195940,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 195940, "pid": 5, "tid": 7, "ts": 1716454224554040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500698, "dur": 14,
    "args": {
      "External id": 195940,
      "cbid": 51, "correlation": 195940
    }
  },
  {
    "ph": "s", "id": 195940, "pid": 76337, "tid": -914061504, "ts": 1716454224500698,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224554044, "dur": 271,
    "args": {
      "External id": 195941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195941,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195941, "pid": 5, "tid": 7, "ts": 1716454224554044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500714, "dur": 11,
    "args": {
      "External id": 195941,
      "cbid": 211, "correlation": 195941
    }
  },
  {
    "ph": "s", "id": 195941, "pid": 76337, "tid": -914061504, "ts": 1716454224500714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224554316, "dur": 16,
    "args": {
      "External id": 195949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195949,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195949, "pid": 5, "tid": 7, "ts": 1716454224554316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500757, "dur": 10,
    "args": {
      "External id": 195949,
      "cbid": 211, "correlation": 195949
    }
  },
  {
    "ph": "s", "id": 195949, "pid": 76337, "tid": -914061504, "ts": 1716454224500757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224554333, "dur": 38,
    "args": {
      "External id": 195960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195960,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195960, "pid": 5, "tid": 7, "ts": 1716454224554333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500826, "dur": 12,
    "args": {
      "External id": 195960,
      "cbid": 211, "correlation": 195960
    }
  },
  {
    "ph": "s", "id": 195960, "pid": 76337, "tid": -914061504, "ts": 1716454224500826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500893, "dur": 0,
    "args": {
      "External id": 195972,
      "cbid": 317, "correlation": 195972
    }
  },
  {
    "ph": "f", "id": 195972, "pid": 76337, "tid": -914061504, "ts": 1716454224500893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500893, "dur": 0,
    "args": {
      "External id": 195973,
      "cbid": 203, "correlation": 195973
    }
  },
  {
    "ph": "f", "id": 195973, "pid": 76337, "tid": -914061504, "ts": 1716454224500893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500894, "dur": 0,
    "args": {
      "External id": 195974,
      "cbid": 205, "correlation": 195974
    }
  },
  {
    "ph": "f", "id": 195974, "pid": 76337, "tid": -914061504, "ts": 1716454224500894,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224554373, "dur": 13,
    "args": {
      "External id": 195978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195978,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195978, "pid": 5, "tid": 7, "ts": 1716454224554373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500914, "dur": 12,
    "args": {
      "External id": 195978,
      "cbid": 211, "correlation": 195978
    }
  },
  {
    "ph": "s", "id": 195978, "pid": 76337, "tid": -914061504, "ts": 1716454224500914,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224554387, "dur": 4,
    "args": {
      "External id": 195980,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195980,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 195980, "pid": 5, "tid": 7, "ts": 1716454224554387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500931, "dur": 6,
    "args": {
      "External id": 195980,
      "cbid": 211, "correlation": 195980
    }
  },
  {
    "ph": "s", "id": 195980, "pid": 76337, "tid": -914061504, "ts": 1716454224500931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500940, "dur": 0,
    "args": {
      "External id": 195981,
      "cbid": 51, "correlation": 195981
    }
  },
  {
    "ph": "s", "id": 195981, "pid": 76337, "tid": -914061504, "ts": 1716454224500940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224554392, "dur": 97,
    "args": {
      "External id": 195982,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195982,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 195982, "pid": 5, "tid": 7, "ts": 1716454224554392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500941, "dur": 5,
    "args": {
      "External id": 195982,
      "cbid": 211, "correlation": 195982
    }
  },
  {
    "ph": "s", "id": 195982, "pid": 76337, "tid": -914061504, "ts": 1716454224500941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224554491, "dur": 16,
    "args": {
      "External id": 195987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195987,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195987, "pid": 5, "tid": 7, "ts": 1716454224554491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224500969, "dur": 17,
    "args": {
      "External id": 195987,
      "cbid": 211, "correlation": 195987
    }
  },
  {
    "ph": "s", "id": 195987, "pid": 76337, "tid": -914061504, "ts": 1716454224500969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224554508, "dur": 11,
    "args": {
      "External id": 195995,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 195995,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 195995, "pid": 5, "tid": 7, "ts": 1716454224554508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501011, "dur": 9,
    "args": {
      "External id": 195995,
      "cbid": 211, "correlation": 195995
    }
  },
  {
    "ph": "s", "id": 195995, "pid": 76337, "tid": -914061504, "ts": 1716454224501011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224554521, "dur": 19,
    "args": {
      "External id": 196015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196015,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 196015, "pid": 5, "tid": 7, "ts": 1716454224554521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501091, "dur": 13,
    "args": {
      "External id": 196015,
      "cbid": 211, "correlation": 196015
    }
  },
  {
    "ph": "s", "id": 196015, "pid": 76337, "tid": -914061504, "ts": 1716454224501091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224554542, "dur": 4,
    "args": {
      "External id": 196027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196027,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 196027, "pid": 5, "tid": 7, "ts": 1716454224554542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501116, "dur": 7,
    "args": {
      "External id": 196027,
      "cbid": 211, "correlation": 196027
    }
  },
  {
    "ph": "s", "id": 196027, "pid": 76337, "tid": -914061504, "ts": 1716454224501116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224554547, "dur": 19,
    "args": {
      "External id": 196030,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196030,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196030, "pid": 5, "tid": 7, "ts": 1716454224554547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501135, "dur": 7,
    "args": {
      "External id": 196030,
      "cbid": 211, "correlation": 196030
    }
  },
  {
    "ph": "s", "id": 196030, "pid": 76337, "tid": -914061504, "ts": 1716454224501135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224554567, "dur": 11,
    "args": {
      "External id": 196039,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196039,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196039, "pid": 5, "tid": 7, "ts": 1716454224554567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501175, "dur": 10,
    "args": {
      "External id": 196039,
      "cbid": 211, "correlation": 196039
    }
  },
  {
    "ph": "s", "id": 196039, "pid": 76337, "tid": -914061504, "ts": 1716454224501175,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501228, "dur": 0,
    "args": {
      "External id": 196049,
      "cbid": 317, "correlation": 196049
    }
  },
  {
    "ph": "f", "id": 196049, "pid": 76337, "tid": -914061504, "ts": 1716454224501228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501229, "dur": 0,
    "args": {
      "External id": 196050,
      "cbid": 203, "correlation": 196050
    }
  },
  {
    "ph": "f", "id": 196050, "pid": 76337, "tid": -914061504, "ts": 1716454224501229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501229, "dur": 0,
    "args": {
      "External id": 196051,
      "cbid": 205, "correlation": 196051
    }
  },
  {
    "ph": "f", "id": 196051, "pid": 76337, "tid": -914061504, "ts": 1716454224501229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224554580, "dur": 11,
    "args": {
      "External id": 196055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196055,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196055, "pid": 5, "tid": 7, "ts": 1716454224554580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501243, "dur": 12,
    "args": {
      "External id": 196055,
      "cbid": 211, "correlation": 196055
    }
  },
  {
    "ph": "s", "id": 196055, "pid": 76337, "tid": -914061504, "ts": 1716454224501243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224554592, "dur": 162,
    "args": {
      "External id": 196057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196057,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196057, "pid": 5, "tid": 7, "ts": 1716454224554592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501258, "dur": 5,
    "args": {
      "External id": 196057,
      "cbid": 211, "correlation": 196057
    }
  },
  {
    "ph": "s", "id": 196057, "pid": 76337, "tid": -914061504, "ts": 1716454224501258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224554757, "dur": 1,
    "args": {
      "External id": 196059,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196059,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 196059, "pid": 5, "tid": 7, "ts": 1716454224554757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501270, "dur": 7,
    "args": {
      "External id": 196059,
      "cbid": 51, "correlation": 196059
    }
  },
  {
    "ph": "s", "id": 196059, "pid": 76337, "tid": -914061504, "ts": 1716454224501270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224554761, "dur": 664,
    "args": {
      "External id": 196060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196060,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196060, "pid": 5, "tid": 7, "ts": 1716454224554761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501278, "dur": 6,
    "args": {
      "External id": 196060,
      "cbid": 211, "correlation": 196060
    }
  },
  {
    "ph": "s", "id": 196060, "pid": 76337, "tid": -914061504, "ts": 1716454224501278,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224555426, "dur": 13,
    "args": {
      "External id": 196062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196062,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196062, "pid": 5, "tid": 7, "ts": 1716454224555426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501289, "dur": 5,
    "args": {
      "External id": 196062,
      "cbid": 211, "correlation": 196062
    }
  },
  {
    "ph": "s", "id": 196062, "pid": 76337, "tid": -914061504, "ts": 1716454224501289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224555441, "dur": 15,
    "args": {
      "External id": 196068,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196068,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196068, "pid": 5, "tid": 7, "ts": 1716454224555441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501318, "dur": 9,
    "args": {
      "External id": 196068,
      "cbid": 211, "correlation": 196068
    }
  },
  {
    "ph": "s", "id": 196068, "pid": 76337, "tid": -914061504, "ts": 1716454224501318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224555457, "dur": 3,
    "args": {
      "External id": 196076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196076,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 196076, "pid": 5, "tid": 7, "ts": 1716454224555457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501362, "dur": 9,
    "args": {
      "External id": 196076,
      "cbid": 211, "correlation": 196076
    }
  },
  {
    "ph": "s", "id": 196076, "pid": 76337, "tid": -914061504, "ts": 1716454224501362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501430, "dur": 1,
    "args": {
      "External id": 196092,
      "cbid": 251, "correlation": 196092
    }
  },
  {
    "ph": "f", "id": 196092, "pid": 76337, "tid": -914061504, "ts": 1716454224501430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501435, "dur": 0,
    "args": {
      "External id": 196094,
      "cbid": 251, "correlation": 196094
    }
  },
  {
    "ph": "f", "id": 196094, "pid": 76337, "tid": -914061504, "ts": 1716454224501435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224555462, "dur": 13,
    "args": {
      "External id": 196095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196095,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196095, "pid": 5, "tid": 7, "ts": 1716454224555462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501437, "dur": 12,
    "args": {
      "External id": 196095,
      "cbid": 211, "correlation": 196095
    }
  },
  {
    "ph": "s", "id": 196095, "pid": 76337, "tid": -914061504, "ts": 1716454224501437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224555477, "dur": 5,
    "args": {
      "External id": 196097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196097,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196097, "pid": 5, "tid": 7, "ts": 1716454224555477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501452, "dur": 6,
    "args": {
      "External id": 196097,
      "cbid": 211, "correlation": 196097
    }
  },
  {
    "ph": "s", "id": 196097, "pid": 76337, "tid": -914061504, "ts": 1716454224501452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224555483, "dur": 17,
    "args": {
      "External id": 196107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196107, "pid": 5, "tid": 7, "ts": 1716454224555483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501510, "dur": 12,
    "args": {
      "External id": 196107,
      "cbid": 211, "correlation": 196107
    }
  },
  {
    "ph": "s", "id": 196107, "pid": 76337, "tid": -914061504, "ts": 1716454224501510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224555502, "dur": 18,
    "args": {
      "External id": 196127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196127,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 196127, "pid": 5, "tid": 7, "ts": 1716454224555502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501577, "dur": 10,
    "args": {
      "External id": 196127,
      "cbid": 211, "correlation": 196127
    }
  },
  {
    "ph": "s", "id": 196127, "pid": 76337, "tid": -914061504, "ts": 1716454224501577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224555521, "dur": 4,
    "args": {
      "External id": 196139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196139,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 196139, "pid": 5, "tid": 7, "ts": 1716454224555521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501597, "dur": 7,
    "args": {
      "External id": 196139,
      "cbid": 211, "correlation": 196139
    }
  },
  {
    "ph": "s", "id": 196139, "pid": 76337, "tid": -914061504, "ts": 1716454224501597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224555526, "dur": 16,
    "args": {
      "External id": 196142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196142,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196142, "pid": 5, "tid": 7, "ts": 1716454224555526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501616, "dur": 6,
    "args": {
      "External id": 196142,
      "cbid": 211, "correlation": 196142
    }
  },
  {
    "ph": "s", "id": 196142, "pid": 76337, "tid": -914061504, "ts": 1716454224501616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224555544, "dur": 11,
    "args": {
      "External id": 196151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196151,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196151, "pid": 5, "tid": 7, "ts": 1716454224555544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501656, "dur": 9,
    "args": {
      "External id": 196151,
      "cbid": 211, "correlation": 196151
    }
  },
  {
    "ph": "s", "id": 196151, "pid": 76337, "tid": -914061504, "ts": 1716454224501656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501718, "dur": 0,
    "args": {
      "External id": 196161,
      "cbid": 317, "correlation": 196161
    }
  },
  {
    "ph": "f", "id": 196161, "pid": 76337, "tid": -914061504, "ts": 1716454224501718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501719, "dur": 0,
    "args": {
      "External id": 196162,
      "cbid": 203, "correlation": 196162
    }
  },
  {
    "ph": "f", "id": 196162, "pid": 76337, "tid": -914061504, "ts": 1716454224501719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501719, "dur": 0,
    "args": {
      "External id": 196163,
      "cbid": 205, "correlation": 196163
    }
  },
  {
    "ph": "f", "id": 196163, "pid": 76337, "tid": -914061504, "ts": 1716454224501719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224555556, "dur": 12,
    "args": {
      "External id": 196167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196167,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196167, "pid": 5, "tid": 7, "ts": 1716454224555556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501733, "dur": 12,
    "args": {
      "External id": 196167,
      "cbid": 211, "correlation": 196167
    }
  },
  {
    "ph": "s", "id": 196167, "pid": 76337, "tid": -914061504, "ts": 1716454224501733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224555569, "dur": 162,
    "args": {
      "External id": 196169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196169,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196169, "pid": 5, "tid": 7, "ts": 1716454224555569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501747, "dur": 6,
    "args": {
      "External id": 196169,
      "cbid": 211, "correlation": 196169
    }
  },
  {
    "ph": "s", "id": 196169, "pid": 76337, "tid": -914061504, "ts": 1716454224501747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224555733, "dur": 1,
    "args": {
      "External id": 196171,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196171,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 196171, "pid": 5, "tid": 7, "ts": 1716454224555733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501759, "dur": 6,
    "args": {
      "External id": 196171,
      "cbid": 51, "correlation": 196171
    }
  },
  {
    "ph": "s", "id": 196171, "pid": 76337, "tid": -914061504, "ts": 1716454224501759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224555737, "dur": 648,
    "args": {
      "External id": 196172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196172,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196172, "pid": 5, "tid": 7, "ts": 1716454224555737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501766, "dur": 6,
    "args": {
      "External id": 196172,
      "cbid": 211, "correlation": 196172
    }
  },
  {
    "ph": "s", "id": 196172, "pid": 76337, "tid": -914061504, "ts": 1716454224501766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224556386, "dur": 13,
    "args": {
      "External id": 196174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196174,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196174, "pid": 5, "tid": 7, "ts": 1716454224556386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501778, "dur": 5,
    "args": {
      "External id": 196174,
      "cbid": 211, "correlation": 196174
    }
  },
  {
    "ph": "s", "id": 196174, "pid": 76337, "tid": -914061504, "ts": 1716454224501778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224556400, "dur": 15,
    "args": {
      "External id": 196180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196180, "pid": 5, "tid": 7, "ts": 1716454224556400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501806, "dur": 9,
    "args": {
      "External id": 196180,
      "cbid": 211, "correlation": 196180
    }
  },
  {
    "ph": "s", "id": 196180, "pid": 76337, "tid": -914061504, "ts": 1716454224501806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224556416, "dur": 12,
    "args": {
      "External id": 196188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196188,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196188, "pid": 5, "tid": 7, "ts": 1716454224556416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501838, "dur": 8,
    "args": {
      "External id": 196188,
      "cbid": 211, "correlation": 196188
    }
  },
  {
    "ph": "s", "id": 196188, "pid": 76337, "tid": -914061504, "ts": 1716454224501838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224556429, "dur": 10,
    "args": {
      "External id": 196196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196196,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196196, "pid": 5, "tid": 7, "ts": 1716454224556429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501868, "dur": 8,
    "args": {
      "External id": 196196,
      "cbid": 211, "correlation": 196196
    }
  },
  {
    "ph": "s", "id": 196196, "pid": 76337, "tid": -914061504, "ts": 1716454224501868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224556441, "dur": 18,
    "args": {
      "External id": 196216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196216,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 196216, "pid": 5, "tid": 7, "ts": 1716454224556441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501950, "dur": 13,
    "args": {
      "External id": 196216,
      "cbid": 211, "correlation": 196216
    }
  },
  {
    "ph": "s", "id": 196216, "pid": 76337, "tid": -914061504, "ts": 1716454224501950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224556460, "dur": 4,
    "args": {
      "External id": 196228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196228,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 196228, "pid": 5, "tid": 7, "ts": 1716454224556460,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224501982, "dur": 7,
    "args": {
      "External id": 196228,
      "cbid": 211, "correlation": 196228
    }
  },
  {
    "ph": "s", "id": 196228, "pid": 76337, "tid": -914061504, "ts": 1716454224501982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224556466, "dur": 16,
    "args": {
      "External id": 196231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196231, "pid": 5, "tid": 7, "ts": 1716454224556466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502000, "dur": 6,
    "args": {
      "External id": 196231,
      "cbid": 211, "correlation": 196231
    }
  },
  {
    "ph": "s", "id": 196231, "pid": 76337, "tid": -914061504, "ts": 1716454224502000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502058, "dur": 0,
    "args": {
      "External id": 196242,
      "cbid": 317, "correlation": 196242
    }
  },
  {
    "ph": "f", "id": 196242, "pid": 76337, "tid": -914061504, "ts": 1716454224502058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502059, "dur": 0,
    "args": {
      "External id": 196243,
      "cbid": 203, "correlation": 196243
    }
  },
  {
    "ph": "f", "id": 196243, "pid": 76337, "tid": -914061504, "ts": 1716454224502059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502060, "dur": 0,
    "args": {
      "External id": 196244,
      "cbid": 205, "correlation": 196244
    }
  },
  {
    "ph": "f", "id": 196244, "pid": 76337, "tid": -914061504, "ts": 1716454224502060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224556483, "dur": 12,
    "args": {
      "External id": 196248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196248,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196248, "pid": 5, "tid": 7, "ts": 1716454224556483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502074, "dur": 12,
    "args": {
      "External id": 196248,
      "cbid": 211, "correlation": 196248
    }
  },
  {
    "ph": "s", "id": 196248, "pid": 76337, "tid": -914061504, "ts": 1716454224502074,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224556496, "dur": 4,
    "args": {
      "External id": 196250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196250,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 196250, "pid": 5, "tid": 7, "ts": 1716454224556496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502090, "dur": 6,
    "args": {
      "External id": 196250,
      "cbid": 211, "correlation": 196250
    }
  },
  {
    "ph": "s", "id": 196250, "pid": 76337, "tid": -914061504, "ts": 1716454224502090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502099, "dur": 0,
    "args": {
      "External id": 196251,
      "cbid": 51, "correlation": 196251
    }
  },
  {
    "ph": "s", "id": 196251, "pid": 76337, "tid": -914061504, "ts": 1716454224502099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224556501, "dur": 95,
    "args": {
      "External id": 196252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196252,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 196252, "pid": 5, "tid": 7, "ts": 1716454224556501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502099, "dur": 5,
    "args": {
      "External id": 196252,
      "cbid": 211, "correlation": 196252
    }
  },
  {
    "ph": "s", "id": 196252, "pid": 76337, "tid": -914061504, "ts": 1716454224502099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224556598, "dur": 16,
    "args": {
      "External id": 196257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196257, "pid": 5, "tid": 7, "ts": 1716454224556598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502127, "dur": 8,
    "args": {
      "External id": 196257,
      "cbid": 211, "correlation": 196257
    }
  },
  {
    "ph": "s", "id": 196257, "pid": 76337, "tid": -914061504, "ts": 1716454224502127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224556615, "dur": 83,
    "args": {
      "External id": 196266,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196266,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196266, "pid": 5, "tid": 7, "ts": 1716454224556615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502213, "dur": 14,
    "args": {
      "External id": 196266,
      "cbid": 211, "correlation": 196266
    }
  },
  {
    "ph": "s", "id": 196266, "pid": 76337, "tid": -914061504, "ts": 1716454224502213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224556698, "dur": 31,
    "args": {
      "External id": 196288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196288,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196288, "pid": 5, "tid": 7, "ts": 1716454224556698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502273, "dur": 10,
    "args": {
      "External id": 196288,
      "cbid": 211, "correlation": 196288
    }
  },
  {
    "ph": "s", "id": 196288, "pid": 76337, "tid": -914061504, "ts": 1716454224502273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502367, "dur": 1,
    "args": {
      "External id": 196299,
      "cbid": 251, "correlation": 196299
    }
  },
  {
    "ph": "f", "id": 196299, "pid": 76337, "tid": -914061504, "ts": 1716454224502367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224556731, "dur": 165,
    "args": {
      "External id": 196300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196300,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196300, "pid": 5, "tid": 7, "ts": 1716454224556731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502373, "dur": 13,
    "args": {
      "External id": 196300,
      "cbid": 211, "correlation": 196300
    }
  },
  {
    "ph": "s", "id": 196300, "pid": 76337, "tid": -914061504, "ts": 1716454224502373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502446, "dur": 1,
    "args": {
      "External id": 196311,
      "cbid": 251, "correlation": 196311
    }
  },
  {
    "ph": "f", "id": 196311, "pid": 76337, "tid": -914061504, "ts": 1716454224502446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224556897, "dur": 158,
    "args": {
      "External id": 196312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196312,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196312, "pid": 5, "tid": 7, "ts": 1716454224556897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502450, "dur": 12,
    "args": {
      "External id": 196312,
      "cbid": 211, "correlation": 196312
    }
  },
  {
    "ph": "s", "id": 196312, "pid": 76337, "tid": -914061504, "ts": 1716454224502450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502517, "dur": 1,
    "args": {
      "External id": 196323,
      "cbid": 251, "correlation": 196323
    }
  },
  {
    "ph": "f", "id": 196323, "pid": 76337, "tid": -914061504, "ts": 1716454224502517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224557057, "dur": 157,
    "args": {
      "External id": 196324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196324,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196324, "pid": 5, "tid": 7, "ts": 1716454224557057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502521, "dur": 14,
    "args": {
      "External id": 196324,
      "cbid": 211, "correlation": 196324
    }
  },
  {
    "ph": "s", "id": 196324, "pid": 76337, "tid": -914061504, "ts": 1716454224502521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224557215, "dur": 336,
    "args": {
      "External id": 196349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196349,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196349, "pid": 5, "tid": 7, "ts": 1716454224557215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502612, "dur": 13,
    "args": {
      "External id": 196349,
      "cbid": 211, "correlation": 196349
    }
  },
  {
    "ph": "s", "id": 196349, "pid": 76337, "tid": -914061504, "ts": 1716454224502612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502717, "dur": 1,
    "args": {
      "External id": 196367,
      "cbid": 251, "correlation": 196367
    }
  },
  {
    "ph": "f", "id": 196367, "pid": 76337, "tid": -914061504, "ts": 1716454224502717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224557553, "dur": 167,
    "args": {
      "External id": 196369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196369,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196369, "pid": 5, "tid": 7, "ts": 1716454224557553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502723, "dur": 13,
    "args": {
      "External id": 196369,
      "cbid": 211, "correlation": 196369
    }
  },
  {
    "ph": "s", "id": 196369, "pid": 76337, "tid": -914061504, "ts": 1716454224502723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224557721, "dur": 20,
    "args": {
      "External id": 196377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196377,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196377, "pid": 5, "tid": 7, "ts": 1716454224557721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502793, "dur": 12,
    "args": {
      "External id": 196377,
      "cbid": 211, "correlation": 196377
    }
  },
  {
    "ph": "s", "id": 196377, "pid": 76337, "tid": -914061504, "ts": 1716454224502793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224557742, "dur": 27,
    "args": {
      "External id": 196385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196385,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196385, "pid": 5, "tid": 7, "ts": 1716454224557742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502832, "dur": 9,
    "args": {
      "External id": 196385,
      "cbid": 211, "correlation": 196385
    }
  },
  {
    "ph": "s", "id": 196385, "pid": 76337, "tid": -914061504, "ts": 1716454224502832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224557771, "dur": 19,
    "args": {
      "External id": 196396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196396,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196396, "pid": 5, "tid": 7, "ts": 1716454224557771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502905, "dur": 14,
    "args": {
      "External id": 196396,
      "cbid": 211, "correlation": 196396
    }
  },
  {
    "ph": "s", "id": 196396, "pid": 76337, "tid": -914061504, "ts": 1716454224502905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224557791, "dur": 16,
    "args": {
      "External id": 196418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196418,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196418, "pid": 5, "tid": 7, "ts": 1716454224557791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224502936, "dur": 7,
    "args": {
      "External id": 196418,
      "cbid": 211, "correlation": 196418
    }
  },
  {
    "ph": "s", "id": 196418, "pid": 76337, "tid": -914061504, "ts": 1716454224502936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503031, "dur": 1,
    "args": {
      "External id": 196429,
      "cbid": 251, "correlation": 196429
    }
  },
  {
    "ph": "f", "id": 196429, "pid": 76337, "tid": -914061504, "ts": 1716454224503031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224557808, "dur": 89,
    "args": {
      "External id": 196430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196430,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 196430, "pid": 5, "tid": 7, "ts": 1716454224557808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503037, "dur": 13,
    "args": {
      "External id": 196430,
      "cbid": 211, "correlation": 196430
    }
  },
  {
    "ph": "s", "id": 196430, "pid": 76337, "tid": -914061504, "ts": 1716454224503037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503111, "dur": 1,
    "args": {
      "External id": 196441,
      "cbid": 251, "correlation": 196441
    }
  },
  {
    "ph": "f", "id": 196441, "pid": 76337, "tid": -914061504, "ts": 1716454224503111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503115, "dur": 0,
    "args": {
      "External id": 196442,
      "cbid": 251, "correlation": 196442
    }
  },
  {
    "ph": "f", "id": 196442, "pid": 76337, "tid": -914061504, "ts": 1716454224503115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224557898, "dur": 12,
    "args": {
      "External id": 196443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196443,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196443, "pid": 5, "tid": 7, "ts": 1716454224557898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503117, "dur": 12,
    "args": {
      "External id": 196443,
      "cbid": 211, "correlation": 196443
    }
  },
  {
    "ph": "s", "id": 196443, "pid": 76337, "tid": -914061504, "ts": 1716454224503117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224557912, "dur": 6,
    "args": {
      "External id": 196445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196445,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196445, "pid": 5, "tid": 7, "ts": 1716454224557912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503130, "dur": 6,
    "args": {
      "External id": 196445,
      "cbid": 211, "correlation": 196445
    }
  },
  {
    "ph": "s", "id": 196445, "pid": 76337, "tid": -914061504, "ts": 1716454224503130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503189, "dur": 1,
    "args": {
      "External id": 196456,
      "cbid": 251, "correlation": 196456
    }
  },
  {
    "ph": "f", "id": 196456, "pid": 76337, "tid": -914061504, "ts": 1716454224503189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503192, "dur": 0,
    "args": {
      "External id": 196457,
      "cbid": 251, "correlation": 196457
    }
  },
  {
    "ph": "f", "id": 196457, "pid": 76337, "tid": -914061504, "ts": 1716454224503192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224557919, "dur": 8,
    "args": {
      "External id": 196458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196458,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196458, "pid": 5, "tid": 7, "ts": 1716454224557919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503194, "dur": 11,
    "args": {
      "External id": 196458,
      "cbid": 211, "correlation": 196458
    }
  },
  {
    "ph": "s", "id": 196458, "pid": 76337, "tid": -914061504, "ts": 1716454224503194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224557929, "dur": 4,
    "args": {
      "External id": 196460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196460,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196460, "pid": 5, "tid": 7, "ts": 1716454224557929,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503207, "dur": 6,
    "args": {
      "External id": 196460,
      "cbid": 211, "correlation": 196460
    }
  },
  {
    "ph": "s", "id": 196460, "pid": 76337, "tid": -914061504, "ts": 1716454224503207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224557933, "dur": 54,
    "args": {
      "External id": 196485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196485,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196485, "pid": 5, "tid": 7, "ts": 1716454224557933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503283, "dur": 12,
    "args": {
      "External id": 196485,
      "cbid": 211, "correlation": 196485
    }
  },
  {
    "ph": "s", "id": 196485, "pid": 76337, "tid": -914061504, "ts": 1716454224503283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503381, "dur": 1,
    "args": {
      "External id": 196503,
      "cbid": 251, "correlation": 196503
    }
  },
  {
    "ph": "f", "id": 196503, "pid": 76337, "tid": -914061504, "ts": 1716454224503381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224557989, "dur": 91,
    "args": {
      "External id": 196505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196505,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 196505, "pid": 5, "tid": 7, "ts": 1716454224557989,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503387, "dur": 14,
    "args": {
      "External id": 196505,
      "cbid": 211, "correlation": 196505
    }
  },
  {
    "ph": "s", "id": 196505, "pid": 76337, "tid": -914061504, "ts": 1716454224503387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224558082, "dur": 10,
    "args": {
      "External id": 196513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196513,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196513, "pid": 5, "tid": 7, "ts": 1716454224558082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503456, "dur": 13,
    "args": {
      "External id": 196513,
      "cbid": 211, "correlation": 196513
    }
  },
  {
    "ph": "s", "id": 196513, "pid": 76337, "tid": -914061504, "ts": 1716454224503456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224558093, "dur": 20,
    "args": {
      "External id": 196521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196521,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196521, "pid": 5, "tid": 7, "ts": 1716454224558093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503497, "dur": 9,
    "args": {
      "External id": 196521,
      "cbid": 211, "correlation": 196521
    }
  },
  {
    "ph": "s", "id": 196521, "pid": 76337, "tid": -914061504, "ts": 1716454224503497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224558114, "dur": 17,
    "args": {
      "External id": 196543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196543,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196543, "pid": 5, "tid": 7, "ts": 1716454224558114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503551, "dur": 11,
    "args": {
      "External id": 196543,
      "cbid": 211, "correlation": 196543
    }
  },
  {
    "ph": "s", "id": 196543, "pid": 76337, "tid": -914061504, "ts": 1716454224503551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503643, "dur": 1,
    "args": {
      "External id": 196559,
      "cbid": 251, "correlation": 196559
    }
  },
  {
    "ph": "f", "id": 196559, "pid": 76337, "tid": -914061504, "ts": 1716454224503643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503648, "dur": 0,
    "args": {
      "External id": 196561,
      "cbid": 251, "correlation": 196561
    }
  },
  {
    "ph": "f", "id": 196561, "pid": 76337, "tid": -914061504, "ts": 1716454224503648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224558133, "dur": 494,
    "args": {
      "External id": 196562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196562,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196562, "pid": 5, "tid": 7, "ts": 1716454224558133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503650, "dur": 15,
    "args": {
      "External id": 196562,
      "cbid": 211, "correlation": 196562
    }
  },
  {
    "ph": "s", "id": 196562, "pid": 76337, "tid": -914061504, "ts": 1716454224503650,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224558628, "dur": 66,
    "args": {
      "External id": 196570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196570,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196570, "pid": 5, "tid": 7, "ts": 1716454224558628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503722, "dur": 13,
    "args": {
      "External id": 196570,
      "cbid": 211, "correlation": 196570
    }
  },
  {
    "ph": "s", "id": 196570, "pid": 76337, "tid": -914061504, "ts": 1716454224503722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224558696, "dur": 69,
    "args": {
      "External id": 196578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196578,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196578, "pid": 5, "tid": 7, "ts": 1716454224558696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503754, "dur": 9,
    "args": {
      "External id": 196578,
      "cbid": 211, "correlation": 196578
    }
  },
  {
    "ph": "s", "id": 196578, "pid": 76337, "tid": -914061504, "ts": 1716454224503754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503834, "dur": 1,
    "args": {
      "External id": 196594,
      "cbid": 251, "correlation": 196594
    }
  },
  {
    "ph": "f", "id": 196594, "pid": 76337, "tid": -914061504, "ts": 1716454224503834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224558767, "dur": 1,
    "args": {
      "External id": 196596,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196596,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 196596, "pid": 5, "tid": 7, "ts": 1716454224558767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503839, "dur": 10,
    "args": {
      "External id": 196596,
      "cbid": 51, "correlation": 196596
    }
  },
  {
    "ph": "s", "id": 196596, "pid": 76337, "tid": -914061504, "ts": 1716454224503839,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224558770, "dur": 272,
    "args": {
      "External id": 196597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196597,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 196597, "pid": 5, "tid": 7, "ts": 1716454224558770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503850, "dur": 12,
    "args": {
      "External id": 196597,
      "cbid": 211, "correlation": 196597
    }
  },
  {
    "ph": "s", "id": 196597, "pid": 76337, "tid": -914061504, "ts": 1716454224503850,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224559044, "dur": 14,
    "args": {
      "External id": 196605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196605,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196605, "pid": 5, "tid": 7, "ts": 1716454224559044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503892, "dur": 10,
    "args": {
      "External id": 196605,
      "cbid": 211, "correlation": 196605
    }
  },
  {
    "ph": "s", "id": 196605, "pid": 76337, "tid": -914061504, "ts": 1716454224503892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224559059, "dur": 38,
    "args": {
      "External id": 196616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196616,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196616, "pid": 5, "tid": 7, "ts": 1716454224559059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224503961, "dur": 20,
    "args": {
      "External id": 196616,
      "cbid": 211, "correlation": 196616
    }
  },
  {
    "ph": "s", "id": 196616, "pid": 76337, "tid": -914061504, "ts": 1716454224503961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504034, "dur": 0,
    "args": {
      "External id": 196628,
      "cbid": 317, "correlation": 196628
    }
  },
  {
    "ph": "f", "id": 196628, "pid": 76337, "tid": -914061504, "ts": 1716454224504034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504035, "dur": 0,
    "args": {
      "External id": 196629,
      "cbid": 203, "correlation": 196629
    }
  },
  {
    "ph": "f", "id": 196629, "pid": 76337, "tid": -914061504, "ts": 1716454224504035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504036, "dur": 0,
    "args": {
      "External id": 196630,
      "cbid": 205, "correlation": 196630
    }
  },
  {
    "ph": "f", "id": 196630, "pid": 76337, "tid": -914061504, "ts": 1716454224504036,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559098, "dur": 14,
    "args": {
      "External id": 196634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196634,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196634, "pid": 5, "tid": 7, "ts": 1716454224559098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504051, "dur": 13,
    "args": {
      "External id": 196634,
      "cbid": 211, "correlation": 196634
    }
  },
  {
    "ph": "s", "id": 196634, "pid": 76337, "tid": -914061504, "ts": 1716454224504051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224559114, "dur": 4,
    "args": {
      "External id": 196636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 196636, "pid": 5, "tid": 7, "ts": 1716454224559114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504070, "dur": 6,
    "args": {
      "External id": 196636,
      "cbid": 211, "correlation": 196636
    }
  },
  {
    "ph": "s", "id": 196636, "pid": 76337, "tid": -914061504, "ts": 1716454224504070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504078, "dur": 0,
    "args": {
      "External id": 196637,
      "cbid": 51, "correlation": 196637
    }
  },
  {
    "ph": "s", "id": 196637, "pid": 76337, "tid": -914061504, "ts": 1716454224504078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224559119, "dur": 98,
    "args": {
      "External id": 196638,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196638,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 196638, "pid": 5, "tid": 7, "ts": 1716454224559119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504079, "dur": 5,
    "args": {
      "External id": 196638,
      "cbid": 211, "correlation": 196638
    }
  },
  {
    "ph": "s", "id": 196638, "pid": 76337, "tid": -914061504, "ts": 1716454224504079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224559219, "dur": 16,
    "args": {
      "External id": 196643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196643, "pid": 5, "tid": 7, "ts": 1716454224559219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504107, "dur": 9,
    "args": {
      "External id": 196643,
      "cbid": 211, "correlation": 196643
    }
  },
  {
    "ph": "s", "id": 196643, "pid": 76337, "tid": -914061504, "ts": 1716454224504107,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224559236, "dur": 11,
    "args": {
      "External id": 196651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196651,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196651, "pid": 5, "tid": 7, "ts": 1716454224559236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504139, "dur": 8,
    "args": {
      "External id": 196651,
      "cbid": 211, "correlation": 196651
    }
  },
  {
    "ph": "s", "id": 196651, "pid": 76337, "tid": -914061504, "ts": 1716454224504139,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504211, "dur": 0,
    "args": {
      "External id": 196661,
      "cbid": 317, "correlation": 196661
    }
  },
  {
    "ph": "f", "id": 196661, "pid": 76337, "tid": -914061504, "ts": 1716454224504211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504212, "dur": 0,
    "args": {
      "External id": 196662,
      "cbid": 203, "correlation": 196662
    }
  },
  {
    "ph": "f", "id": 196662, "pid": 76337, "tid": -914061504, "ts": 1716454224504212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504212, "dur": 0,
    "args": {
      "External id": 196663,
      "cbid": 205, "correlation": 196663
    }
  },
  {
    "ph": "f", "id": 196663, "pid": 76337, "tid": -914061504, "ts": 1716454224504212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559249, "dur": 12,
    "args": {
      "External id": 196667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196667,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196667, "pid": 5, "tid": 7, "ts": 1716454224559249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504227, "dur": 12,
    "args": {
      "External id": 196667,
      "cbid": 211, "correlation": 196667
    }
  },
  {
    "ph": "s", "id": 196667, "pid": 76337, "tid": -914061504, "ts": 1716454224504227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559262, "dur": 162,
    "args": {
      "External id": 196669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196669,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196669, "pid": 5, "tid": 7, "ts": 1716454224559262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504242, "dur": 5,
    "args": {
      "External id": 196669,
      "cbid": 211, "correlation": 196669
    }
  },
  {
    "ph": "s", "id": 196669, "pid": 76337, "tid": -914061504, "ts": 1716454224504242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224559427, "dur": 1,
    "args": {
      "External id": 196671,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196671,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5769230769230769
    }
  },
  {
    "ph": "f", "id": 196671, "pid": 5, "tid": 7, "ts": 1716454224559427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504253, "dur": 7,
    "args": {
      "External id": 196671,
      "cbid": 51, "correlation": 196671
    }
  },
  {
    "ph": "s", "id": 196671, "pid": 76337, "tid": -914061504, "ts": 1716454224504253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224559430, "dur": 198,
    "args": {
      "External id": 196672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196672,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 196672, "pid": 5, "tid": 7, "ts": 1716454224559430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504262, "dur": 7,
    "args": {
      "External id": 196672,
      "cbid": 211, "correlation": 196672
    }
  },
  {
    "ph": "s", "id": 196672, "pid": 76337, "tid": -914061504, "ts": 1716454224504262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559630, "dur": 6,
    "args": {
      "External id": 196674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196674, "pid": 5, "tid": 7, "ts": 1716454224559630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504274, "dur": 5,
    "args": {
      "External id": 196674,
      "cbid": 211, "correlation": 196674
    }
  },
  {
    "ph": "s", "id": 196674, "pid": 76337, "tid": -914061504, "ts": 1716454224504274,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224559638, "dur": 6,
    "args": {
      "External id": 196680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196680, "pid": 5, "tid": 7, "ts": 1716454224559638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504301, "dur": 9,
    "args": {
      "External id": 196680,
      "cbid": 211, "correlation": 196680
    }
  },
  {
    "ph": "s", "id": 196680, "pid": 76337, "tid": -914061504, "ts": 1716454224504301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559645, "dur": 11,
    "args": {
      "External id": 196700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196700,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 196700, "pid": 5, "tid": 7, "ts": 1716454224559645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504402, "dur": 13,
    "args": {
      "External id": 196700,
      "cbid": 211, "correlation": 196700
    }
  },
  {
    "ph": "s", "id": 196700, "pid": 76337, "tid": -914061504, "ts": 1716454224504402,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224559657, "dur": 4,
    "args": {
      "External id": 196712,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196712,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 196712, "pid": 5, "tid": 7, "ts": 1716454224559657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504426, "dur": 6,
    "args": {
      "External id": 196712,
      "cbid": 211, "correlation": 196712
    }
  },
  {
    "ph": "s", "id": 196712, "pid": 76337, "tid": -914061504, "ts": 1716454224504426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224559663, "dur": 8,
    "args": {
      "External id": 196715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196715,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196715, "pid": 5, "tid": 7, "ts": 1716454224559663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504444, "dur": 7,
    "args": {
      "External id": 196715,
      "cbid": 211, "correlation": 196715
    }
  },
  {
    "ph": "s", "id": 196715, "pid": 76337, "tid": -914061504, "ts": 1716454224504444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224559673, "dur": 5,
    "args": {
      "External id": 196724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196724,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196724, "pid": 5, "tid": 7, "ts": 1716454224559673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504488, "dur": 10,
    "args": {
      "External id": 196724,
      "cbid": 211, "correlation": 196724
    }
  },
  {
    "ph": "s", "id": 196724, "pid": 76337, "tid": -914061504, "ts": 1716454224504488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504543, "dur": 0,
    "args": {
      "External id": 196734,
      "cbid": 317, "correlation": 196734
    }
  },
  {
    "ph": "f", "id": 196734, "pid": 76337, "tid": -914061504, "ts": 1716454224504543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504543, "dur": 0,
    "args": {
      "External id": 196735,
      "cbid": 203, "correlation": 196735
    }
  },
  {
    "ph": "f", "id": 196735, "pid": 76337, "tid": -914061504, "ts": 1716454224504543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504544, "dur": 0,
    "args": {
      "External id": 196736,
      "cbid": 205, "correlation": 196736
    }
  },
  {
    "ph": "f", "id": 196736, "pid": 76337, "tid": -914061504, "ts": 1716454224504544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559679, "dur": 5,
    "args": {
      "External id": 196740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196740,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196740, "pid": 5, "tid": 7, "ts": 1716454224559679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504559, "dur": 11,
    "args": {
      "External id": 196740,
      "cbid": 211, "correlation": 196740
    }
  },
  {
    "ph": "s", "id": 196740, "pid": 76337, "tid": -914061504, "ts": 1716454224504559,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224559686, "dur": 163,
    "args": {
      "External id": 196742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196742,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196742, "pid": 5, "tid": 7, "ts": 1716454224559686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504573, "dur": 5,
    "args": {
      "External id": 196742,
      "cbid": 211, "correlation": 196742
    }
  },
  {
    "ph": "s", "id": 196742, "pid": 76337, "tid": -914061504, "ts": 1716454224504573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224559850, "dur": 1,
    "args": {
      "External id": 196744,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196744,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 196744, "pid": 5, "tid": 7, "ts": 1716454224559850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504584, "dur": 6,
    "args": {
      "External id": 196744,
      "cbid": 51, "correlation": 196744
    }
  },
  {
    "ph": "s", "id": 196744, "pid": 76337, "tid": -914061504, "ts": 1716454224504584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224559854, "dur": 270,
    "args": {
      "External id": 196745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196745,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196745, "pid": 5, "tid": 7, "ts": 1716454224559854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504591, "dur": 6,
    "args": {
      "External id": 196745,
      "cbid": 211, "correlation": 196745
    }
  },
  {
    "ph": "s", "id": 196745, "pid": 76337, "tid": -914061504, "ts": 1716454224504591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560126, "dur": 6,
    "args": {
      "External id": 196747,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196747,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196747, "pid": 5, "tid": 7, "ts": 1716454224560126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504603, "dur": 6,
    "args": {
      "External id": 196747,
      "cbid": 211, "correlation": 196747
    }
  },
  {
    "ph": "s", "id": 196747, "pid": 76337, "tid": -914061504, "ts": 1716454224504603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224560133, "dur": 6,
    "args": {
      "External id": 196753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196753,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196753, "pid": 5, "tid": 7, "ts": 1716454224560133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504632, "dur": 8,
    "args": {
      "External id": 196753,
      "cbid": 211, "correlation": 196753
    }
  },
  {
    "ph": "s", "id": 196753, "pid": 76337, "tid": -914061504, "ts": 1716454224504632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224560140, "dur": 3,
    "args": {
      "External id": 196761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196761,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 196761, "pid": 5, "tid": 7, "ts": 1716454224560140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504675, "dur": 9,
    "args": {
      "External id": 196761,
      "cbid": 211, "correlation": 196761
    }
  },
  {
    "ph": "s", "id": 196761, "pid": 76337, "tid": -914061504, "ts": 1716454224504675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504743, "dur": 1,
    "args": {
      "External id": 196777,
      "cbid": 251, "correlation": 196777
    }
  },
  {
    "ph": "f", "id": 196777, "pid": 76337, "tid": -914061504, "ts": 1716454224504743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504748, "dur": 0,
    "args": {
      "External id": 196779,
      "cbid": 251, "correlation": 196779
    }
  },
  {
    "ph": "f", "id": 196779, "pid": 76337, "tid": -914061504, "ts": 1716454224504748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224560144, "dur": 13,
    "args": {
      "External id": 196780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196780,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196780, "pid": 5, "tid": 7, "ts": 1716454224560144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504750, "dur": 12,
    "args": {
      "External id": 196780,
      "cbid": 211, "correlation": 196780
    }
  },
  {
    "ph": "s", "id": 196780, "pid": 76337, "tid": -914061504, "ts": 1716454224504750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224560159, "dur": 5,
    "args": {
      "External id": 196782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196782,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196782, "pid": 5, "tid": 7, "ts": 1716454224560159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504764, "dur": 5,
    "args": {
      "External id": 196782,
      "cbid": 211, "correlation": 196782
    }
  },
  {
    "ph": "s", "id": 196782, "pid": 76337, "tid": -914061504, "ts": 1716454224504764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224560165, "dur": 6,
    "args": {
      "External id": 196792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196792,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196792, "pid": 5, "tid": 7, "ts": 1716454224560165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504821, "dur": 12,
    "args": {
      "External id": 196792,
      "cbid": 211, "correlation": 196792
    }
  },
  {
    "ph": "s", "id": 196792, "pid": 76337, "tid": -914061504, "ts": 1716454224504821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560172, "dur": 10,
    "args": {
      "External id": 196812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196812,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 196812, "pid": 5, "tid": 7, "ts": 1716454224560172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504887, "dur": 11,
    "args": {
      "External id": 196812,
      "cbid": 211, "correlation": 196812
    }
  },
  {
    "ph": "s", "id": 196812, "pid": 76337, "tid": -914061504, "ts": 1716454224504887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224560183, "dur": 4,
    "args": {
      "External id": 196824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196824,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 196824, "pid": 5, "tid": 7, "ts": 1716454224560183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504909, "dur": 6,
    "args": {
      "External id": 196824,
      "cbid": 211, "correlation": 196824
    }
  },
  {
    "ph": "s", "id": 196824, "pid": 76337, "tid": -914061504, "ts": 1716454224504909,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224560188, "dur": 7,
    "args": {
      "External id": 196827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196827,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196827, "pid": 5, "tid": 7, "ts": 1716454224560188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504928, "dur": 6,
    "args": {
      "External id": 196827,
      "cbid": 211, "correlation": 196827
    }
  },
  {
    "ph": "s", "id": 196827, "pid": 76337, "tid": -914061504, "ts": 1716454224504928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224560196, "dur": 5,
    "args": {
      "External id": 196836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196836,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196836, "pid": 5, "tid": 7, "ts": 1716454224560196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224504968, "dur": 17,
    "args": {
      "External id": 196836,
      "cbid": 211, "correlation": 196836
    }
  },
  {
    "ph": "s", "id": 196836, "pid": 76337, "tid": -914061504, "ts": 1716454224504968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505040, "dur": 0,
    "args": {
      "External id": 196846,
      "cbid": 317, "correlation": 196846
    }
  },
  {
    "ph": "f", "id": 196846, "pid": 76337, "tid": -914061504, "ts": 1716454224505040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505041, "dur": 0,
    "args": {
      "External id": 196847,
      "cbid": 203, "correlation": 196847
    }
  },
  {
    "ph": "f", "id": 196847, "pid": 76337, "tid": -914061504, "ts": 1716454224505041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505042, "dur": 0,
    "args": {
      "External id": 196848,
      "cbid": 205, "correlation": 196848
    }
  },
  {
    "ph": "f", "id": 196848, "pid": 76337, "tid": -914061504, "ts": 1716454224505042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560202, "dur": 5,
    "args": {
      "External id": 196852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196852,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196852, "pid": 5, "tid": 7, "ts": 1716454224560202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505056, "dur": 12,
    "args": {
      "External id": 196852,
      "cbid": 211, "correlation": 196852
    }
  },
  {
    "ph": "s", "id": 196852, "pid": 76337, "tid": -914061504, "ts": 1716454224505056,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560208, "dur": 161,
    "args": {
      "External id": 196854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196854,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196854, "pid": 5, "tid": 7, "ts": 1716454224560208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505071, "dur": 5,
    "args": {
      "External id": 196854,
      "cbid": 211, "correlation": 196854
    }
  },
  {
    "ph": "s", "id": 196854, "pid": 76337, "tid": -914061504, "ts": 1716454224505071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224560372, "dur": 1,
    "args": {
      "External id": 196856,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196856,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 196856, "pid": 5, "tid": 7, "ts": 1716454224560372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505082, "dur": 6,
    "args": {
      "External id": 196856,
      "cbid": 51, "correlation": 196856
    }
  },
  {
    "ph": "s", "id": 196856, "pid": 76337, "tid": -914061504, "ts": 1716454224505082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224560375, "dur": 260,
    "args": {
      "External id": 196857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196857,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196857, "pid": 5, "tid": 7, "ts": 1716454224560375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505089, "dur": 6,
    "args": {
      "External id": 196857,
      "cbid": 211, "correlation": 196857
    }
  },
  {
    "ph": "s", "id": 196857, "pid": 76337, "tid": -914061504, "ts": 1716454224505089,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560637, "dur": 6,
    "args": {
      "External id": 196859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196859,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196859, "pid": 5, "tid": 7, "ts": 1716454224560637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505099, "dur": 5,
    "args": {
      "External id": 196859,
      "cbid": 211, "correlation": 196859
    }
  },
  {
    "ph": "s", "id": 196859, "pid": 76337, "tid": -914061504, "ts": 1716454224505099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224560644, "dur": 6,
    "args": {
      "External id": 196865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196865,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196865, "pid": 5, "tid": 7, "ts": 1716454224560644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505127, "dur": 9,
    "args": {
      "External id": 196865,
      "cbid": 211, "correlation": 196865
    }
  },
  {
    "ph": "s", "id": 196865, "pid": 76337, "tid": -914061504, "ts": 1716454224505127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224560652, "dur": 5,
    "args": {
      "External id": 196873,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196873,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196873, "pid": 5, "tid": 7, "ts": 1716454224560652,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505161, "dur": 8,
    "args": {
      "External id": 196873,
      "cbid": 211, "correlation": 196873
    }
  },
  {
    "ph": "s", "id": 196873, "pid": 76337, "tid": -914061504, "ts": 1716454224505161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224560658, "dur": 4,
    "args": {
      "External id": 196881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196881,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196881, "pid": 5, "tid": 7, "ts": 1716454224560658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505189, "dur": 9,
    "args": {
      "External id": 196881,
      "cbid": 211, "correlation": 196881
    }
  },
  {
    "ph": "s", "id": 196881, "pid": 76337, "tid": -914061504, "ts": 1716454224505189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560664, "dur": 9,
    "args": {
      "External id": 196901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196901,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 196901, "pid": 5, "tid": 7, "ts": 1716454224560664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505265, "dur": 12,
    "args": {
      "External id": 196901,
      "cbid": 211, "correlation": 196901
    }
  },
  {
    "ph": "s", "id": 196901, "pid": 76337, "tid": -914061504, "ts": 1716454224505265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224560675, "dur": 4,
    "args": {
      "External id": 196913,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196913,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 196913, "pid": 5, "tid": 7, "ts": 1716454224560675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505286, "dur": 6,
    "args": {
      "External id": 196913,
      "cbid": 211, "correlation": 196913
    }
  },
  {
    "ph": "s", "id": 196913, "pid": 76337, "tid": -914061504, "ts": 1716454224505286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224560680, "dur": 7,
    "args": {
      "External id": 196916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196916,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196916, "pid": 5, "tid": 7, "ts": 1716454224560680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505304, "dur": 7,
    "args": {
      "External id": 196916,
      "cbid": 211, "correlation": 196916
    }
  },
  {
    "ph": "s", "id": 196916, "pid": 76337, "tid": -914061504, "ts": 1716454224505304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224560688, "dur": 5,
    "args": {
      "External id": 196925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196925,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196925, "pid": 5, "tid": 7, "ts": 1716454224560688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505342, "dur": 9,
    "args": {
      "External id": 196925,
      "cbid": 211, "correlation": 196925
    }
  },
  {
    "ph": "s", "id": 196925, "pid": 76337, "tid": -914061504, "ts": 1716454224505342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505394, "dur": 0,
    "args": {
      "External id": 196935,
      "cbid": 317, "correlation": 196935
    }
  },
  {
    "ph": "f", "id": 196935, "pid": 76337, "tid": -914061504, "ts": 1716454224505394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505394, "dur": 0,
    "args": {
      "External id": 196936,
      "cbid": 203, "correlation": 196936
    }
  },
  {
    "ph": "f", "id": 196936, "pid": 76337, "tid": -914061504, "ts": 1716454224505394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505395, "dur": 0,
    "args": {
      "External id": 196937,
      "cbid": 205, "correlation": 196937
    }
  },
  {
    "ph": "f", "id": 196937, "pid": 76337, "tid": -914061504, "ts": 1716454224505395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560694, "dur": 5,
    "args": {
      "External id": 196941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196941,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196941, "pid": 5, "tid": 7, "ts": 1716454224560694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505409, "dur": 11,
    "args": {
      "External id": 196941,
      "cbid": 211, "correlation": 196941
    }
  },
  {
    "ph": "s", "id": 196941, "pid": 76337, "tid": -914061504, "ts": 1716454224505409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224560700, "dur": 163,
    "args": {
      "External id": 196943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196943,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196943, "pid": 5, "tid": 7, "ts": 1716454224560700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505422, "dur": 5,
    "args": {
      "External id": 196943,
      "cbid": 211, "correlation": 196943
    }
  },
  {
    "ph": "s", "id": 196943, "pid": 76337, "tid": -914061504, "ts": 1716454224505422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224560865, "dur": 1,
    "args": {
      "External id": 196945,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 196945,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 196945, "pid": 5, "tid": 7, "ts": 1716454224560865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505433, "dur": 6,
    "args": {
      "External id": 196945,
      "cbid": 51, "correlation": 196945
    }
  },
  {
    "ph": "s", "id": 196945, "pid": 76337, "tid": -914061504, "ts": 1716454224505433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224560869, "dur": 258,
    "args": {
      "External id": 196946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196946,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196946, "pid": 5, "tid": 7, "ts": 1716454224560869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505440, "dur": 6,
    "args": {
      "External id": 196946,
      "cbid": 211, "correlation": 196946
    }
  },
  {
    "ph": "s", "id": 196946, "pid": 76337, "tid": -914061504, "ts": 1716454224505440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561128, "dur": 6,
    "args": {
      "External id": 196948,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196948,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 196948, "pid": 5, "tid": 7, "ts": 1716454224561128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505449, "dur": 5,
    "args": {
      "External id": 196948,
      "cbid": 211, "correlation": 196948
    }
  },
  {
    "ph": "s", "id": 196948, "pid": 76337, "tid": -914061504, "ts": 1716454224505449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224561135, "dur": 6,
    "args": {
      "External id": 196954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196954,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196954, "pid": 5, "tid": 7, "ts": 1716454224561135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505477, "dur": 8,
    "args": {
      "External id": 196954,
      "cbid": 211, "correlation": 196954
    }
  },
  {
    "ph": "s", "id": 196954, "pid": 76337, "tid": -914061504, "ts": 1716454224505477,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224561143, "dur": 3,
    "args": {
      "External id": 196962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196962,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 196962, "pid": 5, "tid": 7, "ts": 1716454224561143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505520, "dur": 10,
    "args": {
      "External id": 196962,
      "cbid": 211, "correlation": 196962
    }
  },
  {
    "ph": "s", "id": 196962, "pid": 76337, "tid": -914061504, "ts": 1716454224505520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505582, "dur": 1,
    "args": {
      "External id": 196978,
      "cbid": 251, "correlation": 196978
    }
  },
  {
    "ph": "f", "id": 196978, "pid": 76337, "tid": -914061504, "ts": 1716454224505582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505587, "dur": 0,
    "args": {
      "External id": 196980,
      "cbid": 251, "correlation": 196980
    }
  },
  {
    "ph": "f", "id": 196980, "pid": 76337, "tid": -914061504, "ts": 1716454224505587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224561147, "dur": 10,
    "args": {
      "External id": 196981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196981,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196981, "pid": 5, "tid": 7, "ts": 1716454224561147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505588, "dur": 11,
    "args": {
      "External id": 196981,
      "cbid": 211, "correlation": 196981
    }
  },
  {
    "ph": "s", "id": 196981, "pid": 76337, "tid": -914061504, "ts": 1716454224505588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224561159, "dur": 4,
    "args": {
      "External id": 196983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196983,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 196983, "pid": 5, "tid": 7, "ts": 1716454224561159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505601, "dur": 9,
    "args": {
      "External id": 196983,
      "cbid": 211, "correlation": 196983
    }
  },
  {
    "ph": "s", "id": 196983, "pid": 76337, "tid": -914061504, "ts": 1716454224505601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224561164, "dur": 6,
    "args": {
      "External id": 196993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 196993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 196993, "pid": 5, "tid": 7, "ts": 1716454224561164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505661, "dur": 12,
    "args": {
      "External id": 196993,
      "cbid": 211, "correlation": 196993
    }
  },
  {
    "ph": "s", "id": 196993, "pid": 76337, "tid": -914061504, "ts": 1716454224505661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561171, "dur": 10,
    "args": {
      "External id": 197013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197013,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197013, "pid": 5, "tid": 7, "ts": 1716454224561171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505727, "dur": 10,
    "args": {
      "External id": 197013,
      "cbid": 211, "correlation": 197013
    }
  },
  {
    "ph": "s", "id": 197013, "pid": 76337, "tid": -914061504, "ts": 1716454224505727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224561182, "dur": 4,
    "args": {
      "External id": 197025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197025,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 197025, "pid": 5, "tid": 7, "ts": 1716454224561182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505747, "dur": 6,
    "args": {
      "External id": 197025,
      "cbid": 211, "correlation": 197025
    }
  },
  {
    "ph": "s", "id": 197025, "pid": 76337, "tid": -914061504, "ts": 1716454224505747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224561187, "dur": 7,
    "args": {
      "External id": 197028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197028, "pid": 5, "tid": 7, "ts": 1716454224561187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505767, "dur": 6,
    "args": {
      "External id": 197028,
      "cbid": 211, "correlation": 197028
    }
  },
  {
    "ph": "s", "id": 197028, "pid": 76337, "tid": -914061504, "ts": 1716454224505767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224561194, "dur": 5,
    "args": {
      "External id": 197037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197037,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197037, "pid": 5, "tid": 7, "ts": 1716454224561194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505806, "dur": 11,
    "args": {
      "External id": 197037,
      "cbid": 211, "correlation": 197037
    }
  },
  {
    "ph": "s", "id": 197037, "pid": 76337, "tid": -914061504, "ts": 1716454224505806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505870, "dur": 0,
    "args": {
      "External id": 197047,
      "cbid": 317, "correlation": 197047
    }
  },
  {
    "ph": "f", "id": 197047, "pid": 76337, "tid": -914061504, "ts": 1716454224505870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505870, "dur": 0,
    "args": {
      "External id": 197048,
      "cbid": 203, "correlation": 197048
    }
  },
  {
    "ph": "f", "id": 197048, "pid": 76337, "tid": -914061504, "ts": 1716454224505870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505871, "dur": 0,
    "args": {
      "External id": 197049,
      "cbid": 205, "correlation": 197049
    }
  },
  {
    "ph": "f", "id": 197049, "pid": 76337, "tid": -914061504, "ts": 1716454224505871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561200, "dur": 5,
    "args": {
      "External id": 197053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197053,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197053, "pid": 5, "tid": 7, "ts": 1716454224561200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505885, "dur": 12,
    "args": {
      "External id": 197053,
      "cbid": 211, "correlation": 197053
    }
  },
  {
    "ph": "s", "id": 197053, "pid": 76337, "tid": -914061504, "ts": 1716454224505885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561207, "dur": 162,
    "args": {
      "External id": 197055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197055,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197055, "pid": 5, "tid": 7, "ts": 1716454224561207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505899, "dur": 5,
    "args": {
      "External id": 197055,
      "cbid": 211, "correlation": 197055
    }
  },
  {
    "ph": "s", "id": 197055, "pid": 76337, "tid": -914061504, "ts": 1716454224505899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224561371, "dur": 1,
    "args": {
      "External id": 197057,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197057,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 197057, "pid": 5, "tid": 7, "ts": 1716454224561371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505910, "dur": 7,
    "args": {
      "External id": 197057,
      "cbid": 51, "correlation": 197057
    }
  },
  {
    "ph": "s", "id": 197057, "pid": 76337, "tid": -914061504, "ts": 1716454224505910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224561374, "dur": 259,
    "args": {
      "External id": 197058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197058,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197058, "pid": 5, "tid": 7, "ts": 1716454224561374,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505918, "dur": 6,
    "args": {
      "External id": 197058,
      "cbid": 211, "correlation": 197058
    }
  },
  {
    "ph": "s", "id": 197058, "pid": 76337, "tid": -914061504, "ts": 1716454224505918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561635, "dur": 6,
    "args": {
      "External id": 197060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197060,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197060, "pid": 5, "tid": 7, "ts": 1716454224561635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505929, "dur": 5,
    "args": {
      "External id": 197060,
      "cbid": 211, "correlation": 197060
    }
  },
  {
    "ph": "s", "id": 197060, "pid": 76337, "tid": -914061504, "ts": 1716454224505929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224561641, "dur": 6,
    "args": {
      "External id": 197066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197066,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197066, "pid": 5, "tid": 7, "ts": 1716454224561641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505957, "dur": 9,
    "args": {
      "External id": 197066,
      "cbid": 211, "correlation": 197066
    }
  },
  {
    "ph": "s", "id": 197066, "pid": 76337, "tid": -914061504, "ts": 1716454224505957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224561649, "dur": 5,
    "args": {
      "External id": 197074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197074,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197074, "pid": 5, "tid": 7, "ts": 1716454224561649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224505999, "dur": 10,
    "args": {
      "External id": 197074,
      "cbid": 211, "correlation": 197074
    }
  },
  {
    "ph": "s", "id": 197074, "pid": 76337, "tid": -914061504, "ts": 1716454224505999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224561655, "dur": 4,
    "args": {
      "External id": 197082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197082,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197082, "pid": 5, "tid": 7, "ts": 1716454224561655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506032, "dur": 8,
    "args": {
      "External id": 197082,
      "cbid": 211, "correlation": 197082
    }
  },
  {
    "ph": "s", "id": 197082, "pid": 76337, "tid": -914061504, "ts": 1716454224506032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561661, "dur": 10,
    "args": {
      "External id": 197102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197102,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197102, "pid": 5, "tid": 7, "ts": 1716454224561661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506136, "dur": 12,
    "args": {
      "External id": 197102,
      "cbid": 211, "correlation": 197102
    }
  },
  {
    "ph": "s", "id": 197102, "pid": 76337, "tid": -914061504, "ts": 1716454224506136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224561672, "dur": 4,
    "args": {
      "External id": 197114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197114,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 197114, "pid": 5, "tid": 7, "ts": 1716454224561672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506158, "dur": 6,
    "args": {
      "External id": 197114,
      "cbid": 211, "correlation": 197114
    }
  },
  {
    "ph": "s", "id": 197114, "pid": 76337, "tid": -914061504, "ts": 1716454224506158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224561677, "dur": 6,
    "args": {
      "External id": 197117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197117,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197117, "pid": 5, "tid": 7, "ts": 1716454224561677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506177, "dur": 6,
    "args": {
      "External id": 197117,
      "cbid": 211, "correlation": 197117
    }
  },
  {
    "ph": "s", "id": 197117, "pid": 76337, "tid": -914061504, "ts": 1716454224506177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224561685, "dur": 5,
    "args": {
      "External id": 197126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197126,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197126, "pid": 5, "tid": 7, "ts": 1716454224561685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506214, "dur": 10,
    "args": {
      "External id": 197126,
      "cbid": 211, "correlation": 197126
    }
  },
  {
    "ph": "s", "id": 197126, "pid": 76337, "tid": -914061504, "ts": 1716454224506214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506267, "dur": 0,
    "args": {
      "External id": 197136,
      "cbid": 317, "correlation": 197136
    }
  },
  {
    "ph": "f", "id": 197136, "pid": 76337, "tid": -914061504, "ts": 1716454224506267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506268, "dur": 0,
    "args": {
      "External id": 197137,
      "cbid": 203, "correlation": 197137
    }
  },
  {
    "ph": "f", "id": 197137, "pid": 76337, "tid": -914061504, "ts": 1716454224506268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506269, "dur": 0,
    "args": {
      "External id": 197138,
      "cbid": 205, "correlation": 197138
    }
  },
  {
    "ph": "f", "id": 197138, "pid": 76337, "tid": -914061504, "ts": 1716454224506269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561690, "dur": 5,
    "args": {
      "External id": 197142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197142,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197142, "pid": 5, "tid": 7, "ts": 1716454224561690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506283, "dur": 11,
    "args": {
      "External id": 197142,
      "cbid": 211, "correlation": 197142
    }
  },
  {
    "ph": "s", "id": 197142, "pid": 76337, "tid": -914061504, "ts": 1716454224506283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224561697, "dur": 162,
    "args": {
      "External id": 197144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197144,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197144, "pid": 5, "tid": 7, "ts": 1716454224561697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506297, "dur": 5,
    "args": {
      "External id": 197144,
      "cbid": 211, "correlation": 197144
    }
  },
  {
    "ph": "s", "id": 197144, "pid": 76337, "tid": -914061504, "ts": 1716454224506297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224561861, "dur": 1,
    "args": {
      "External id": 197146,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197146,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 197146, "pid": 5, "tid": 7, "ts": 1716454224561861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506308, "dur": 6,
    "args": {
      "External id": 197146,
      "cbid": 51, "correlation": 197146
    }
  },
  {
    "ph": "s", "id": 197146, "pid": 76337, "tid": -914061504, "ts": 1716454224506308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224561865, "dur": 259,
    "args": {
      "External id": 197147,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197147,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197147, "pid": 5, "tid": 7, "ts": 1716454224561865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506315, "dur": 6,
    "args": {
      "External id": 197147,
      "cbid": 211, "correlation": 197147
    }
  },
  {
    "ph": "s", "id": 197147, "pid": 76337, "tid": -914061504, "ts": 1716454224506315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562125, "dur": 5,
    "args": {
      "External id": 197149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197149,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197149, "pid": 5, "tid": 7, "ts": 1716454224562125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506325, "dur": 5,
    "args": {
      "External id": 197149,
      "cbid": 211, "correlation": 197149
    }
  },
  {
    "ph": "s", "id": 197149, "pid": 76337, "tid": -914061504, "ts": 1716454224506325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224562132, "dur": 6,
    "args": {
      "External id": 197155,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197155,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197155, "pid": 5, "tid": 7, "ts": 1716454224562132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506354, "dur": 8,
    "args": {
      "External id": 197155,
      "cbid": 211, "correlation": 197155
    }
  },
  {
    "ph": "s", "id": 197155, "pid": 76337, "tid": -914061504, "ts": 1716454224506354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224562139, "dur": 3,
    "args": {
      "External id": 197163,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197163,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 197163, "pid": 5, "tid": 7, "ts": 1716454224562139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506397, "dur": 9,
    "args": {
      "External id": 197163,
      "cbid": 211, "correlation": 197163
    }
  },
  {
    "ph": "s", "id": 197163, "pid": 76337, "tid": -914061504, "ts": 1716454224506397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506458, "dur": 1,
    "args": {
      "External id": 197179,
      "cbid": 251, "correlation": 197179
    }
  },
  {
    "ph": "f", "id": 197179, "pid": 76337, "tid": -914061504, "ts": 1716454224506458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506463, "dur": 0,
    "args": {
      "External id": 197181,
      "cbid": 251, "correlation": 197181
    }
  },
  {
    "ph": "f", "id": 197181, "pid": 76337, "tid": -914061504, "ts": 1716454224506463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224562143, "dur": 10,
    "args": {
      "External id": 197182,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197182,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197182, "pid": 5, "tid": 7, "ts": 1716454224562143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506465, "dur": 12,
    "args": {
      "External id": 197182,
      "cbid": 211, "correlation": 197182
    }
  },
  {
    "ph": "s", "id": 197182, "pid": 76337, "tid": -914061504, "ts": 1716454224506465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224562155, "dur": 4,
    "args": {
      "External id": 197184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197184,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197184, "pid": 5, "tid": 7, "ts": 1716454224562155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506479, "dur": 5,
    "args": {
      "External id": 197184,
      "cbid": 211, "correlation": 197184
    }
  },
  {
    "ph": "s", "id": 197184, "pid": 76337, "tid": -914061504, "ts": 1716454224506479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224562160, "dur": 6,
    "args": {
      "External id": 197194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197194,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197194, "pid": 5, "tid": 7, "ts": 1716454224562160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506535, "dur": 12,
    "args": {
      "External id": 197194,
      "cbid": 211, "correlation": 197194
    }
  },
  {
    "ph": "s", "id": 197194, "pid": 76337, "tid": -914061504, "ts": 1716454224506535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562167, "dur": 9,
    "args": {
      "External id": 197214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197214,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197214, "pid": 5, "tid": 7, "ts": 1716454224562167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506600, "dur": 11,
    "args": {
      "External id": 197214,
      "cbid": 211, "correlation": 197214
    }
  },
  {
    "ph": "s", "id": 197214, "pid": 76337, "tid": -914061504, "ts": 1716454224506600,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224562178, "dur": 4,
    "args": {
      "External id": 197226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197226,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 197226, "pid": 5, "tid": 7, "ts": 1716454224562178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506621, "dur": 6,
    "args": {
      "External id": 197226,
      "cbid": 211, "correlation": 197226
    }
  },
  {
    "ph": "s", "id": 197226, "pid": 76337, "tid": -914061504, "ts": 1716454224506621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224562183, "dur": 7,
    "args": {
      "External id": 197229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197229,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197229, "pid": 5, "tid": 7, "ts": 1716454224562183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506639, "dur": 6,
    "args": {
      "External id": 197229,
      "cbid": 211, "correlation": 197229
    }
  },
  {
    "ph": "s", "id": 197229, "pid": 76337, "tid": -914061504, "ts": 1716454224506639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224562191, "dur": 5,
    "args": {
      "External id": 197238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197238,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197238, "pid": 5, "tid": 7, "ts": 1716454224562191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506678, "dur": 11,
    "args": {
      "External id": 197238,
      "cbid": 211, "correlation": 197238
    }
  },
  {
    "ph": "s", "id": 197238, "pid": 76337, "tid": -914061504, "ts": 1716454224506678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506743, "dur": 0,
    "args": {
      "External id": 197248,
      "cbid": 317, "correlation": 197248
    }
  },
  {
    "ph": "f", "id": 197248, "pid": 76337, "tid": -914061504, "ts": 1716454224506743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506744, "dur": 0,
    "args": {
      "External id": 197249,
      "cbid": 203, "correlation": 197249
    }
  },
  {
    "ph": "f", "id": 197249, "pid": 76337, "tid": -914061504, "ts": 1716454224506744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506745, "dur": 0,
    "args": {
      "External id": 197250,
      "cbid": 205, "correlation": 197250
    }
  },
  {
    "ph": "f", "id": 197250, "pid": 76337, "tid": -914061504, "ts": 1716454224506745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562197, "dur": 5,
    "args": {
      "External id": 197254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197254,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197254, "pid": 5, "tid": 7, "ts": 1716454224562197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506758, "dur": 12,
    "args": {
      "External id": 197254,
      "cbid": 211, "correlation": 197254
    }
  },
  {
    "ph": "s", "id": 197254, "pid": 76337, "tid": -914061504, "ts": 1716454224506758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562203, "dur": 162,
    "args": {
      "External id": 197256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197256,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197256, "pid": 5, "tid": 7, "ts": 1716454224562203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506772, "dur": 5,
    "args": {
      "External id": 197256,
      "cbid": 211, "correlation": 197256
    }
  },
  {
    "ph": "s", "id": 197256, "pid": 76337, "tid": -914061504, "ts": 1716454224506772,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224562367, "dur": 1,
    "args": {
      "External id": 197258,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197258,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15296367112810708
    }
  },
  {
    "ph": "f", "id": 197258, "pid": 5, "tid": 7, "ts": 1716454224562367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506783, "dur": 7,
    "args": {
      "External id": 197258,
      "cbid": 51, "correlation": 197258
    }
  },
  {
    "ph": "s", "id": 197258, "pid": 76337, "tid": -914061504, "ts": 1716454224506783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224562371, "dur": 258,
    "args": {
      "External id": 197259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197259,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197259, "pid": 5, "tid": 7, "ts": 1716454224562371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506791, "dur": 6,
    "args": {
      "External id": 197259,
      "cbid": 211, "correlation": 197259
    }
  },
  {
    "ph": "s", "id": 197259, "pid": 76337, "tid": -914061504, "ts": 1716454224506791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562630, "dur": 5,
    "args": {
      "External id": 197261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197261,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197261, "pid": 5, "tid": 7, "ts": 1716454224562630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506801, "dur": 5,
    "args": {
      "External id": 197261,
      "cbid": 211, "correlation": 197261
    }
  },
  {
    "ph": "s", "id": 197261, "pid": 76337, "tid": -914061504, "ts": 1716454224506801,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224562637, "dur": 6,
    "args": {
      "External id": 197267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197267,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197267, "pid": 5, "tid": 7, "ts": 1716454224562637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506829, "dur": 9,
    "args": {
      "External id": 197267,
      "cbid": 211, "correlation": 197267
    }
  },
  {
    "ph": "s", "id": 197267, "pid": 76337, "tid": -914061504, "ts": 1716454224506829,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224562644, "dur": 5,
    "args": {
      "External id": 197275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197275,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197275, "pid": 5, "tid": 7, "ts": 1716454224562644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506862, "dur": 8,
    "args": {
      "External id": 197275,
      "cbid": 211, "correlation": 197275
    }
  },
  {
    "ph": "s", "id": 197275, "pid": 76337, "tid": -914061504, "ts": 1716454224506862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224562650, "dur": 4,
    "args": {
      "External id": 197283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197283,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197283, "pid": 5, "tid": 7, "ts": 1716454224562650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224506891, "dur": 9,
    "args": {
      "External id": 197283,
      "cbid": 211, "correlation": 197283
    }
  },
  {
    "ph": "s", "id": 197283, "pid": 76337, "tid": -914061504, "ts": 1716454224506891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562656, "dur": 10,
    "args": {
      "External id": 197303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197303,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197303, "pid": 5, "tid": 7, "ts": 1716454224562656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507057, "dur": 14,
    "args": {
      "External id": 197303,
      "cbid": 211, "correlation": 197303
    }
  },
  {
    "ph": "s", "id": 197303, "pid": 76337, "tid": -914061504, "ts": 1716454224507057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224562667, "dur": 4,
    "args": {
      "External id": 197315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197315,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 197315, "pid": 5, "tid": 7, "ts": 1716454224562667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507081, "dur": 6,
    "args": {
      "External id": 197315,
      "cbid": 211, "correlation": 197315
    }
  },
  {
    "ph": "s", "id": 197315, "pid": 76337, "tid": -914061504, "ts": 1716454224507081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224562672, "dur": 7,
    "args": {
      "External id": 197318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197318, "pid": 5, "tid": 7, "ts": 1716454224562672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507100, "dur": 7,
    "args": {
      "External id": 197318,
      "cbid": 211, "correlation": 197318
    }
  },
  {
    "ph": "s", "id": 197318, "pid": 76337, "tid": -914061504, "ts": 1716454224507100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507159, "dur": 0,
    "args": {
      "External id": 197329,
      "cbid": 317, "correlation": 197329
    }
  },
  {
    "ph": "f", "id": 197329, "pid": 76337, "tid": -914061504, "ts": 1716454224507159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507160, "dur": 0,
    "args": {
      "External id": 197330,
      "cbid": 203, "correlation": 197330
    }
  },
  {
    "ph": "f", "id": 197330, "pid": 76337, "tid": -914061504, "ts": 1716454224507160,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507161, "dur": 0,
    "args": {
      "External id": 197331,
      "cbid": 205, "correlation": 197331
    }
  },
  {
    "ph": "f", "id": 197331, "pid": 76337, "tid": -914061504, "ts": 1716454224507161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562680, "dur": 5,
    "args": {
      "External id": 197335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197335,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197335, "pid": 5, "tid": 7, "ts": 1716454224562680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507178, "dur": 13,
    "args": {
      "External id": 197335,
      "cbid": 211, "correlation": 197335
    }
  },
  {
    "ph": "s", "id": 197335, "pid": 76337, "tid": -914061504, "ts": 1716454224507178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224562687, "dur": 37,
    "args": {
      "External id": 197337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197337,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 197337, "pid": 5, "tid": 7, "ts": 1716454224562687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507198, "dur": 9,
    "args": {
      "External id": 197337,
      "cbid": 211, "correlation": 197337
    }
  },
  {
    "ph": "s", "id": 197337, "pid": 76337, "tid": -914061504, "ts": 1716454224507198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562725, "dur": 5,
    "args": {
      "External id": 197339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197339,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197339, "pid": 5, "tid": 7, "ts": 1716454224562725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507211, "dur": 5,
    "args": {
      "External id": 197339,
      "cbid": 211, "correlation": 197339
    }
  },
  {
    "ph": "s", "id": 197339, "pid": 76337, "tid": -914061504, "ts": 1716454224507211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224562731, "dur": 6,
    "args": {
      "External id": 197345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197345,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197345, "pid": 5, "tid": 7, "ts": 1716454224562731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507239, "dur": 8,
    "args": {
      "External id": 197345,
      "cbid": 211, "correlation": 197345
    }
  },
  {
    "ph": "s", "id": 197345, "pid": 76337, "tid": -914061504, "ts": 1716454224507239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224562738, "dur": 20,
    "args": {
      "External id": 197354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197354,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197354, "pid": 5, "tid": 7, "ts": 1716454224562738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507321, "dur": 14,
    "args": {
      "External id": 197354,
      "cbid": 211, "correlation": 197354
    }
  },
  {
    "ph": "s", "id": 197354, "pid": 76337, "tid": -914061504, "ts": 1716454224507321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224562759, "dur": 11,
    "args": {
      "External id": 197376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197376,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 197376, "pid": 5, "tid": 7, "ts": 1716454224562759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507378, "dur": 11,
    "args": {
      "External id": 197376,
      "cbid": 211, "correlation": 197376
    }
  },
  {
    "ph": "s", "id": 197376, "pid": 76337, "tid": -914061504, "ts": 1716454224507378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507468, "dur": 2,
    "args": {
      "External id": 197387,
      "cbid": 251, "correlation": 197387
    }
  },
  {
    "ph": "f", "id": 197387, "pid": 76337, "tid": -914061504, "ts": 1716454224507468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507473, "dur": 0,
    "args": {
      "External id": 197388,
      "cbid": 251, "correlation": 197388
    }
  },
  {
    "ph": "f", "id": 197388, "pid": 76337, "tid": -914061504, "ts": 1716454224507473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224562771, "dur": 54,
    "args": {
      "External id": 197389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197389,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 197389, "pid": 5, "tid": 7, "ts": 1716454224562771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507476, "dur": 15,
    "args": {
      "External id": 197389,
      "cbid": 211, "correlation": 197389
    }
  },
  {
    "ph": "s", "id": 197389, "pid": 76337, "tid": -914061504, "ts": 1716454224507476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507548, "dur": 1,
    "args": {
      "External id": 197400,
      "cbid": 251, "correlation": 197400
    }
  },
  {
    "ph": "f", "id": 197400, "pid": 76337, "tid": -914061504, "ts": 1716454224507548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507551, "dur": 0,
    "args": {
      "External id": 197401,
      "cbid": 251, "correlation": 197401
    }
  },
  {
    "ph": "f", "id": 197401, "pid": 76337, "tid": -914061504, "ts": 1716454224507551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224562826, "dur": 53,
    "args": {
      "External id": 197402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197402,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 197402, "pid": 5, "tid": 7, "ts": 1716454224562826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507553, "dur": 11,
    "args": {
      "External id": 197402,
      "cbid": 211, "correlation": 197402
    }
  },
  {
    "ph": "s", "id": 197402, "pid": 76337, "tid": -914061504, "ts": 1716454224507553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507618, "dur": 1,
    "args": {
      "External id": 197413,
      "cbid": 251, "correlation": 197413
    }
  },
  {
    "ph": "f", "id": 197413, "pid": 76337, "tid": -914061504, "ts": 1716454224507618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507622, "dur": 0,
    "args": {
      "External id": 197414,
      "cbid": 251, "correlation": 197414
    }
  },
  {
    "ph": "f", "id": 197414, "pid": 76337, "tid": -914061504, "ts": 1716454224507622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224562881, "dur": 54,
    "args": {
      "External id": 197415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197415,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 197415, "pid": 5, "tid": 7, "ts": 1716454224562881,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507624, "dur": 12,
    "args": {
      "External id": 197415,
      "cbid": 211, "correlation": 197415
    }
  },
  {
    "ph": "s", "id": 197415, "pid": 76337, "tid": -914061504, "ts": 1716454224507624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224562936, "dur": 55,
    "args": {
      "External id": 197440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197440,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197440, "pid": 5, "tid": 7, "ts": 1716454224562936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507708, "dur": 12,
    "args": {
      "External id": 197440,
      "cbid": 211, "correlation": 197440
    }
  },
  {
    "ph": "s", "id": 197440, "pid": 76337, "tid": -914061504, "ts": 1716454224507708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507807, "dur": 1,
    "args": {
      "External id": 197458,
      "cbid": 251, "correlation": 197458
    }
  },
  {
    "ph": "f", "id": 197458, "pid": 76337, "tid": -914061504, "ts": 1716454224507807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224562992, "dur": 63,
    "args": {
      "External id": 197460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197460,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 197460, "pid": 5, "tid": 7, "ts": 1716454224562992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507813, "dur": 13,
    "args": {
      "External id": 197460,
      "cbid": 211, "correlation": 197460
    }
  },
  {
    "ph": "s", "id": 197460, "pid": 76337, "tid": -914061504, "ts": 1716454224507813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224563056, "dur": 6,
    "args": {
      "External id": 197468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197468,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197468, "pid": 5, "tid": 7, "ts": 1716454224563056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507882, "dur": 13,
    "args": {
      "External id": 197468,
      "cbid": 211, "correlation": 197468
    }
  },
  {
    "ph": "s", "id": 197468, "pid": 76337, "tid": -914061504, "ts": 1716454224507882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224563064, "dur": 7,
    "args": {
      "External id": 197476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197476,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197476, "pid": 5, "tid": 7, "ts": 1716454224563064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507920, "dur": 8,
    "args": {
      "External id": 197476,
      "cbid": 211, "correlation": 197476
    }
  },
  {
    "ph": "s", "id": 197476, "pid": 76337, "tid": -914061504, "ts": 1716454224507920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563072, "dur": 7,
    "args": {
      "External id": 197487,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197487,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197487, "pid": 5, "tid": 7, "ts": 1716454224563072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224507999, "dur": 13,
    "args": {
      "External id": 197487,
      "cbid": 211, "correlation": 197487
    }
  },
  {
    "ph": "s", "id": 197487, "pid": 76337, "tid": -914061504, "ts": 1716454224507999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563081, "dur": 9,
    "args": {
      "External id": 197509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197509,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 197509, "pid": 5, "tid": 7, "ts": 1716454224563081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508031, "dur": 9,
    "args": {
      "External id": 197509,
      "cbid": 211, "correlation": 197509
    }
  },
  {
    "ph": "s", "id": 197509, "pid": 76337, "tid": -914061504, "ts": 1716454224508031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508118, "dur": 2,
    "args": {
      "External id": 197520,
      "cbid": 251, "correlation": 197520
    }
  },
  {
    "ph": "f", "id": 197520, "pid": 76337, "tid": -914061504, "ts": 1716454224508118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224563092, "dur": 1,
    "args": {
      "External id": 197521,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197521,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 197521, "pid": 5, "tid": 7, "ts": 1716454224563092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508124, "dur": 10,
    "args": {
      "External id": 197521,
      "cbid": 51, "correlation": 197521
    }
  },
  {
    "ph": "s", "id": 197521, "pid": 76337, "tid": -914061504, "ts": 1716454224508124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224563096, "dur": 37,
    "args": {
      "External id": 197522,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197522,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 197522, "pid": 5, "tid": 7, "ts": 1716454224563096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508136, "dur": 12,
    "args": {
      "External id": 197522,
      "cbid": 211, "correlation": 197522
    }
  },
  {
    "ph": "s", "id": 197522, "pid": 76337, "tid": -914061504, "ts": 1716454224508136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508207, "dur": 1,
    "args": {
      "External id": 197533,
      "cbid": 251, "correlation": 197533
    }
  },
  {
    "ph": "f", "id": 197533, "pid": 76337, "tid": -914061504, "ts": 1716454224508207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508210, "dur": 0,
    "args": {
      "External id": 197534,
      "cbid": 251, "correlation": 197534
    }
  },
  {
    "ph": "f", "id": 197534, "pid": 76337, "tid": -914061504, "ts": 1716454224508210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224563134, "dur": 12,
    "args": {
      "External id": 197535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197535,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197535, "pid": 5, "tid": 7, "ts": 1716454224563134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508212, "dur": 12,
    "args": {
      "External id": 197535,
      "cbid": 211, "correlation": 197535
    }
  },
  {
    "ph": "s", "id": 197535, "pid": 76337, "tid": -914061504, "ts": 1716454224508212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224563147, "dur": 5,
    "args": {
      "External id": 197537,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197537,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197537, "pid": 5, "tid": 7, "ts": 1716454224563147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508225, "dur": 6,
    "args": {
      "External id": 197537,
      "cbid": 211, "correlation": 197537
    }
  },
  {
    "ph": "s", "id": 197537, "pid": 76337, "tid": -914061504, "ts": 1716454224508225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508283, "dur": 1,
    "args": {
      "External id": 197548,
      "cbid": 251, "correlation": 197548
    }
  },
  {
    "ph": "f", "id": 197548, "pid": 76337, "tid": -914061504, "ts": 1716454224508283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508287, "dur": 0,
    "args": {
      "External id": 197549,
      "cbid": 251, "correlation": 197549
    }
  },
  {
    "ph": "f", "id": 197549, "pid": 76337, "tid": -914061504, "ts": 1716454224508287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224563154, "dur": 8,
    "args": {
      "External id": 197550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197550,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197550, "pid": 5, "tid": 7, "ts": 1716454224563154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508289, "dur": 11,
    "args": {
      "External id": 197550,
      "cbid": 211, "correlation": 197550
    }
  },
  {
    "ph": "s", "id": 197550, "pid": 76337, "tid": -914061504, "ts": 1716454224508289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224563163, "dur": 4,
    "args": {
      "External id": 197552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197552,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197552, "pid": 5, "tid": 7, "ts": 1716454224563163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508302, "dur": 5,
    "args": {
      "External id": 197552,
      "cbid": 211, "correlation": 197552
    }
  },
  {
    "ph": "s", "id": 197552, "pid": 76337, "tid": -914061504, "ts": 1716454224508302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224563168, "dur": 20,
    "args": {
      "External id": 197577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197577,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 197577, "pid": 5, "tid": 7, "ts": 1716454224563168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508380, "dur": 13,
    "args": {
      "External id": 197577,
      "cbid": 211, "correlation": 197577
    }
  },
  {
    "ph": "s", "id": 197577, "pid": 76337, "tid": -914061504, "ts": 1716454224508380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508483, "dur": 2,
    "args": {
      "External id": 197595,
      "cbid": 251, "correlation": 197595
    }
  },
  {
    "ph": "f", "id": 197595, "pid": 76337, "tid": -914061504, "ts": 1716454224508483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224563190, "dur": 1,
    "args": {
      "External id": 197597,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197597,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 197597, "pid": 5, "tid": 7, "ts": 1716454224563190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508489, "dur": 11,
    "args": {
      "External id": 197597,
      "cbid": 51, "correlation": 197597
    }
  },
  {
    "ph": "s", "id": 197597, "pid": 76337, "tid": -914061504, "ts": 1716454224508489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224563194, "dur": 37,
    "args": {
      "External id": 197598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197598,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 197598, "pid": 5, "tid": 7, "ts": 1716454224563194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508501, "dur": 13,
    "args": {
      "External id": 197598,
      "cbid": 211, "correlation": 197598
    }
  },
  {
    "ph": "s", "id": 197598, "pid": 76337, "tid": -914061504, "ts": 1716454224508501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224563232, "dur": 4,
    "args": {
      "External id": 197606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197606,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197606, "pid": 5, "tid": 7, "ts": 1716454224563232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508573, "dur": 12,
    "args": {
      "External id": 197606,
      "cbid": 211, "correlation": 197606
    }
  },
  {
    "ph": "s", "id": 197606, "pid": 76337, "tid": -914061504, "ts": 1716454224508573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563237, "dur": 8,
    "args": {
      "External id": 197614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197614, "pid": 5, "tid": 7, "ts": 1716454224563237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508615, "dur": 9,
    "args": {
      "External id": 197614,
      "cbid": 211, "correlation": 197614
    }
  },
  {
    "ph": "s", "id": 197614, "pid": 76337, "tid": -914061504, "ts": 1716454224508615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563247, "dur": 8,
    "args": {
      "External id": 197636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197636,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 197636, "pid": 5, "tid": 7, "ts": 1716454224563247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508666, "dur": 10,
    "args": {
      "External id": 197636,
      "cbid": 211, "correlation": 197636
    }
  },
  {
    "ph": "s", "id": 197636, "pid": 76337, "tid": -914061504, "ts": 1716454224508666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508758, "dur": 1,
    "args": {
      "External id": 197652,
      "cbid": 251, "correlation": 197652
    }
  },
  {
    "ph": "f", "id": 197652, "pid": 76337, "tid": -914061504, "ts": 1716454224508758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508763, "dur": 0,
    "args": {
      "External id": 197654,
      "cbid": 251, "correlation": 197654
    }
  },
  {
    "ph": "f", "id": 197654, "pid": 76337, "tid": -914061504, "ts": 1716454224508763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224563256, "dur": 190,
    "args": {
      "External id": 197655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197655,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197655, "pid": 5, "tid": 7, "ts": 1716454224563256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508765, "dur": 13,
    "args": {
      "External id": 197655,
      "cbid": 211, "correlation": 197655
    }
  },
  {
    "ph": "s", "id": 197655, "pid": 76337, "tid": -914061504, "ts": 1716454224508765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563448, "dur": 21,
    "args": {
      "External id": 197663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197663,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197663, "pid": 5, "tid": 7, "ts": 1716454224563448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508832, "dur": 12,
    "args": {
      "External id": 197663,
      "cbid": 211, "correlation": 197663
    }
  },
  {
    "ph": "s", "id": 197663, "pid": 76337, "tid": -914061504, "ts": 1716454224508832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563471, "dur": 21,
    "args": {
      "External id": 197671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197671, "pid": 5, "tid": 7, "ts": 1716454224563471,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508864, "dur": 8,
    "args": {
      "External id": 197671,
      "cbid": 211, "correlation": 197671
    }
  },
  {
    "ph": "s", "id": 197671, "pid": 76337, "tid": -914061504, "ts": 1716454224508864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508945, "dur": 1,
    "args": {
      "External id": 197687,
      "cbid": 251, "correlation": 197687
    }
  },
  {
    "ph": "f", "id": 197687, "pid": 76337, "tid": -914061504, "ts": 1716454224508945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224563494, "dur": 1,
    "args": {
      "External id": 197689,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197689,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 197689, "pid": 5, "tid": 7, "ts": 1716454224563494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508951, "dur": 9,
    "args": {
      "External id": 197689,
      "cbid": 51, "correlation": 197689
    }
  },
  {
    "ph": "s", "id": 197689, "pid": 76337, "tid": -914061504, "ts": 1716454224508951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224563497, "dur": 110,
    "args": {
      "External id": 197690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197690,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 197690, "pid": 5, "tid": 7, "ts": 1716454224563497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224508961, "dur": 22,
    "args": {
      "External id": 197690,
      "cbid": 211, "correlation": 197690
    }
  },
  {
    "ph": "s", "id": 197690, "pid": 76337, "tid": -914061504, "ts": 1716454224508961,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224563609, "dur": 6,
    "args": {
      "External id": 197698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197698,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197698, "pid": 5, "tid": 7, "ts": 1716454224563609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509016, "dur": 11,
    "args": {
      "External id": 197698,
      "cbid": 211, "correlation": 197698
    }
  },
  {
    "ph": "s", "id": 197698, "pid": 76337, "tid": -914061504, "ts": 1716454224509016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563616, "dur": 10,
    "args": {
      "External id": 197709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197709, "pid": 5, "tid": 7, "ts": 1716454224563616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509086, "dur": 13,
    "args": {
      "External id": 197709,
      "cbid": 211, "correlation": 197709
    }
  },
  {
    "ph": "s", "id": 197709, "pid": 76337, "tid": -914061504, "ts": 1716454224509086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509153, "dur": 0,
    "args": {
      "External id": 197721,
      "cbid": 317, "correlation": 197721
    }
  },
  {
    "ph": "f", "id": 197721, "pid": 76337, "tid": -914061504, "ts": 1716454224509153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509154, "dur": 0,
    "args": {
      "External id": 197722,
      "cbid": 203, "correlation": 197722
    }
  },
  {
    "ph": "f", "id": 197722, "pid": 76337, "tid": -914061504, "ts": 1716454224509154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509155, "dur": 0,
    "args": {
      "External id": 197723,
      "cbid": 205, "correlation": 197723
    }
  },
  {
    "ph": "f", "id": 197723, "pid": 76337, "tid": -914061504, "ts": 1716454224509155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563627, "dur": 5,
    "args": {
      "External id": 197727,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197727,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197727, "pid": 5, "tid": 7, "ts": 1716454224563627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509170, "dur": 12,
    "args": {
      "External id": 197727,
      "cbid": 211, "correlation": 197727
    }
  },
  {
    "ph": "s", "id": 197727, "pid": 76337, "tid": -914061504, "ts": 1716454224509170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224563634, "dur": 37,
    "args": {
      "External id": 197729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197729,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 197729, "pid": 5, "tid": 7, "ts": 1716454224563634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509189, "dur": 7,
    "args": {
      "External id": 197729,
      "cbid": 211, "correlation": 197729
    }
  },
  {
    "ph": "s", "id": 197729, "pid": 76337, "tid": -914061504, "ts": 1716454224509189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563673, "dur": 6,
    "args": {
      "External id": 197731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197731,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197731, "pid": 5, "tid": 7, "ts": 1716454224563673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509200, "dur": 6,
    "args": {
      "External id": 197731,
      "cbid": 211, "correlation": 197731
    }
  },
  {
    "ph": "s", "id": 197731, "pid": 76337, "tid": -914061504, "ts": 1716454224509200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563680, "dur": 7,
    "args": {
      "External id": 197737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197737, "pid": 5, "tid": 7, "ts": 1716454224563680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509227, "dur": 8,
    "args": {
      "External id": 197737,
      "cbid": 211, "correlation": 197737
    }
  },
  {
    "ph": "s", "id": 197737, "pid": 76337, "tid": -914061504, "ts": 1716454224509227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224563689, "dur": 6,
    "args": {
      "External id": 197745,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197745,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197745, "pid": 5, "tid": 7, "ts": 1716454224563689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509259, "dur": 8,
    "args": {
      "External id": 197745,
      "cbid": 211, "correlation": 197745
    }
  },
  {
    "ph": "s", "id": 197745, "pid": 76337, "tid": -914061504, "ts": 1716454224509259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563696, "dur": 11,
    "args": {
      "External id": 197765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197765,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197765, "pid": 5, "tid": 7, "ts": 1716454224563696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509332, "dur": 12,
    "args": {
      "External id": 197765,
      "cbid": 211, "correlation": 197765
    }
  },
  {
    "ph": "s", "id": 197765, "pid": 76337, "tid": -914061504, "ts": 1716454224509332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224563709, "dur": 4,
    "args": {
      "External id": 197777,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197777,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 197777, "pid": 5, "tid": 7, "ts": 1716454224563709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509354, "dur": 7,
    "args": {
      "External id": 197777,
      "cbid": 211, "correlation": 197777
    }
  },
  {
    "ph": "s", "id": 197777, "pid": 76337, "tid": -914061504, "ts": 1716454224509354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224563714, "dur": 9,
    "args": {
      "External id": 197780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197780,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197780, "pid": 5, "tid": 7, "ts": 1716454224563714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509373, "dur": 7,
    "args": {
      "External id": 197780,
      "cbid": 211, "correlation": 197780
    }
  },
  {
    "ph": "s", "id": 197780, "pid": 76337, "tid": -914061504, "ts": 1716454224509373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224563724, "dur": 5,
    "args": {
      "External id": 197789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197789,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197789, "pid": 5, "tid": 7, "ts": 1716454224563724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509412, "dur": 10,
    "args": {
      "External id": 197789,
      "cbid": 211, "correlation": 197789
    }
  },
  {
    "ph": "s", "id": 197789, "pid": 76337, "tid": -914061504, "ts": 1716454224509412,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509464, "dur": 0,
    "args": {
      "External id": 197799,
      "cbid": 317, "correlation": 197799
    }
  },
  {
    "ph": "f", "id": 197799, "pid": 76337, "tid": -914061504, "ts": 1716454224509464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509465, "dur": 0,
    "args": {
      "External id": 197800,
      "cbid": 203, "correlation": 197800
    }
  },
  {
    "ph": "f", "id": 197800, "pid": 76337, "tid": -914061504, "ts": 1716454224509465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509466, "dur": 0,
    "args": {
      "External id": 197801,
      "cbid": 205, "correlation": 197801
    }
  },
  {
    "ph": "f", "id": 197801, "pid": 76337, "tid": -914061504, "ts": 1716454224509466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563731, "dur": 5,
    "args": {
      "External id": 197805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197805,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197805, "pid": 5, "tid": 7, "ts": 1716454224563731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509479, "dur": 11,
    "args": {
      "External id": 197805,
      "cbid": 211, "correlation": 197805
    }
  },
  {
    "ph": "s", "id": 197805, "pid": 76337, "tid": -914061504, "ts": 1716454224509479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224563737, "dur": 162,
    "args": {
      "External id": 197807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197807,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197807, "pid": 5, "tid": 7, "ts": 1716454224563737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509493, "dur": 5,
    "args": {
      "External id": 197807,
      "cbid": 211, "correlation": 197807
    }
  },
  {
    "ph": "s", "id": 197807, "pid": 76337, "tid": -914061504, "ts": 1716454224509493,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224563902, "dur": 1,
    "args": {
      "External id": 197809,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197809,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 197809, "pid": 5, "tid": 7, "ts": 1716454224563902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509504, "dur": 7,
    "args": {
      "External id": 197809,
      "cbid": 51, "correlation": 197809
    }
  },
  {
    "ph": "s", "id": 197809, "pid": 76337, "tid": -914061504, "ts": 1716454224509504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224563905, "dur": 269,
    "args": {
      "External id": 197810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197810,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197810, "pid": 5, "tid": 7, "ts": 1716454224563905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509512, "dur": 6,
    "args": {
      "External id": 197810,
      "cbid": 211, "correlation": 197810
    }
  },
  {
    "ph": "s", "id": 197810, "pid": 76337, "tid": -914061504, "ts": 1716454224509512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564176, "dur": 6,
    "args": {
      "External id": 197812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197812,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197812, "pid": 5, "tid": 7, "ts": 1716454224564176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509522, "dur": 5,
    "args": {
      "External id": 197812,
      "cbid": 211, "correlation": 197812
    }
  },
  {
    "ph": "s", "id": 197812, "pid": 76337, "tid": -914061504, "ts": 1716454224509522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224564183, "dur": 6,
    "args": {
      "External id": 197818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197818, "pid": 5, "tid": 7, "ts": 1716454224564183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509549, "dur": 9,
    "args": {
      "External id": 197818,
      "cbid": 211, "correlation": 197818
    }
  },
  {
    "ph": "s", "id": 197818, "pid": 76337, "tid": -914061504, "ts": 1716454224509549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224564191, "dur": 3,
    "args": {
      "External id": 197826,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197826,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 197826, "pid": 5, "tid": 7, "ts": 1716454224564191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509593, "dur": 10,
    "args": {
      "External id": 197826,
      "cbid": 211, "correlation": 197826
    }
  },
  {
    "ph": "s", "id": 197826, "pid": 76337, "tid": -914061504, "ts": 1716454224509593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509658, "dur": 1,
    "args": {
      "External id": 197842,
      "cbid": 251, "correlation": 197842
    }
  },
  {
    "ph": "f", "id": 197842, "pid": 76337, "tid": -914061504, "ts": 1716454224509658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509664, "dur": 0,
    "args": {
      "External id": 197844,
      "cbid": 251, "correlation": 197844
    }
  },
  {
    "ph": "f", "id": 197844, "pid": 76337, "tid": -914061504, "ts": 1716454224509664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224564195, "dur": 13,
    "args": {
      "External id": 197845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197845,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197845, "pid": 5, "tid": 7, "ts": 1716454224564195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509666, "dur": 11,
    "args": {
      "External id": 197845,
      "cbid": 211, "correlation": 197845
    }
  },
  {
    "ph": "s", "id": 197845, "pid": 76337, "tid": -914061504, "ts": 1716454224509666,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224564209, "dur": 5,
    "args": {
      "External id": 197847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197847,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197847, "pid": 5, "tid": 7, "ts": 1716454224564209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509679, "dur": 5,
    "args": {
      "External id": 197847,
      "cbid": 211, "correlation": 197847
    }
  },
  {
    "ph": "s", "id": 197847, "pid": 76337, "tid": -914061504, "ts": 1716454224509679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224564215, "dur": 6,
    "args": {
      "External id": 197857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197857,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197857, "pid": 5, "tid": 7, "ts": 1716454224564215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509736, "dur": 12,
    "args": {
      "External id": 197857,
      "cbid": 211, "correlation": 197857
    }
  },
  {
    "ph": "s", "id": 197857, "pid": 76337, "tid": -914061504, "ts": 1716454224509736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564222, "dur": 10,
    "args": {
      "External id": 197877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197877,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197877, "pid": 5, "tid": 7, "ts": 1716454224564222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509802, "dur": 11,
    "args": {
      "External id": 197877,
      "cbid": 211, "correlation": 197877
    }
  },
  {
    "ph": "s", "id": 197877, "pid": 76337, "tid": -914061504, "ts": 1716454224509802,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224564233, "dur": 3,
    "args": {
      "External id": 197889,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197889,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 197889, "pid": 5, "tid": 7, "ts": 1716454224564233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509823, "dur": 6,
    "args": {
      "External id": 197889,
      "cbid": 211, "correlation": 197889
    }
  },
  {
    "ph": "s", "id": 197889, "pid": 76337, "tid": -914061504, "ts": 1716454224509823,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224564238, "dur": 7,
    "args": {
      "External id": 197892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197892, "pid": 5, "tid": 7, "ts": 1716454224564238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509842, "dur": 6,
    "args": {
      "External id": 197892,
      "cbid": 211, "correlation": 197892
    }
  },
  {
    "ph": "s", "id": 197892, "pid": 76337, "tid": -914061504, "ts": 1716454224509842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224564246, "dur": 5,
    "args": {
      "External id": 197901,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197901,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197901, "pid": 5, "tid": 7, "ts": 1716454224564246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509882, "dur": 9,
    "args": {
      "External id": 197901,
      "cbid": 211, "correlation": 197901
    }
  },
  {
    "ph": "s", "id": 197901, "pid": 76337, "tid": -914061504, "ts": 1716454224509882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509944, "dur": 0,
    "args": {
      "External id": 197911,
      "cbid": 317, "correlation": 197911
    }
  },
  {
    "ph": "f", "id": 197911, "pid": 76337, "tid": -914061504, "ts": 1716454224509944,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509945, "dur": 0,
    "args": {
      "External id": 197912,
      "cbid": 203, "correlation": 197912
    }
  },
  {
    "ph": "f", "id": 197912, "pid": 76337, "tid": -914061504, "ts": 1716454224509945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509945, "dur": 0,
    "args": {
      "External id": 197913,
      "cbid": 205, "correlation": 197913
    }
  },
  {
    "ph": "f", "id": 197913, "pid": 76337, "tid": -914061504, "ts": 1716454224509945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564252, "dur": 5,
    "args": {
      "External id": 197917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197917,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197917, "pid": 5, "tid": 7, "ts": 1716454224564252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509959, "dur": 13,
    "args": {
      "External id": 197917,
      "cbid": 211, "correlation": 197917
    }
  },
  {
    "ph": "s", "id": 197917, "pid": 76337, "tid": -914061504, "ts": 1716454224509959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564259, "dur": 163,
    "args": {
      "External id": 197919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197919,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197919, "pid": 5, "tid": 7, "ts": 1716454224564259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509982, "dur": 6,
    "args": {
      "External id": 197919,
      "cbid": 211, "correlation": 197919
    }
  },
  {
    "ph": "s", "id": 197919, "pid": 76337, "tid": -914061504, "ts": 1716454224509982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224564423, "dur": 1,
    "args": {
      "External id": 197921,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 197921,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 197921, "pid": 5, "tid": 7, "ts": 1716454224564423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224509993, "dur": 6,
    "args": {
      "External id": 197921,
      "cbid": 51, "correlation": 197921
    }
  },
  {
    "ph": "s", "id": 197921, "pid": 76337, "tid": -914061504, "ts": 1716454224509993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224564427, "dur": 259,
    "args": {
      "External id": 197922,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197922,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197922, "pid": 5, "tid": 7, "ts": 1716454224564427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510001, "dur": 6,
    "args": {
      "External id": 197922,
      "cbid": 211, "correlation": 197922
    }
  },
  {
    "ph": "s", "id": 197922, "pid": 76337, "tid": -914061504, "ts": 1716454224510001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564688, "dur": 6,
    "args": {
      "External id": 197924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197924,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197924, "pid": 5, "tid": 7, "ts": 1716454224564688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510010, "dur": 6,
    "args": {
      "External id": 197924,
      "cbid": 211, "correlation": 197924
    }
  },
  {
    "ph": "s", "id": 197924, "pid": 76337, "tid": -914061504, "ts": 1716454224510010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224564695, "dur": 6,
    "args": {
      "External id": 197930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197930,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197930, "pid": 5, "tid": 7, "ts": 1716454224564695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510040, "dur": 9,
    "args": {
      "External id": 197930,
      "cbid": 211, "correlation": 197930
    }
  },
  {
    "ph": "s", "id": 197930, "pid": 76337, "tid": -914061504, "ts": 1716454224510040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224564703, "dur": 4,
    "args": {
      "External id": 197938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197938,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197938, "pid": 5, "tid": 7, "ts": 1716454224564703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510073, "dur": 8,
    "args": {
      "External id": 197938,
      "cbid": 211, "correlation": 197938
    }
  },
  {
    "ph": "s", "id": 197938, "pid": 76337, "tid": -914061504, "ts": 1716454224510073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224564709, "dur": 4,
    "args": {
      "External id": 197946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197946,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 197946, "pid": 5, "tid": 7, "ts": 1716454224564709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510102, "dur": 8,
    "args": {
      "External id": 197946,
      "cbid": 211, "correlation": 197946
    }
  },
  {
    "ph": "s", "id": 197946, "pid": 76337, "tid": -914061504, "ts": 1716454224510102,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224564714, "dur": 11,
    "args": {
      "External id": 197955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197955,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197955, "pid": 5, "tid": 7, "ts": 1716454224564714,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510191, "dur": 14,
    "args": {
      "External id": 197955,
      "cbid": 211, "correlation": 197955
    }
  },
  {
    "ph": "s", "id": 197955, "pid": 76337, "tid": -914061504, "ts": 1716454224510191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564727, "dur": 12,
    "args": {
      "External id": 197975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197975,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 197975, "pid": 5, "tid": 7, "ts": 1716454224564727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510260, "dur": 11,
    "args": {
      "External id": 197975,
      "cbid": 211, "correlation": 197975
    }
  },
  {
    "ph": "s", "id": 197975, "pid": 76337, "tid": -914061504, "ts": 1716454224510260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224564740, "dur": 4,
    "args": {
      "External id": 197987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197987,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 197987, "pid": 5, "tid": 7, "ts": 1716454224564740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510281, "dur": 6,
    "args": {
      "External id": 197987,
      "cbid": 211, "correlation": 197987
    }
  },
  {
    "ph": "s", "id": 197987, "pid": 76337, "tid": -914061504, "ts": 1716454224510281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224564746, "dur": 10,
    "args": {
      "External id": 197990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197990,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197990, "pid": 5, "tid": 7, "ts": 1716454224564746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510299, "dur": 7,
    "args": {
      "External id": 197990,
      "cbid": 211, "correlation": 197990
    }
  },
  {
    "ph": "s", "id": 197990, "pid": 76337, "tid": -914061504, "ts": 1716454224510299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224564757, "dur": 6,
    "args": {
      "External id": 197999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 197999,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 197999, "pid": 5, "tid": 7, "ts": 1716454224564757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510339, "dur": 9,
    "args": {
      "External id": 197999,
      "cbid": 211, "correlation": 197999
    }
  },
  {
    "ph": "s", "id": 197999, "pid": 76337, "tid": -914061504, "ts": 1716454224510339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510391, "dur": 0,
    "args": {
      "External id": 198009,
      "cbid": 317, "correlation": 198009
    }
  },
  {
    "ph": "f", "id": 198009, "pid": 76337, "tid": -914061504, "ts": 1716454224510391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510392, "dur": 0,
    "args": {
      "External id": 198010,
      "cbid": 203, "correlation": 198010
    }
  },
  {
    "ph": "f", "id": 198010, "pid": 76337, "tid": -914061504, "ts": 1716454224510392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510393, "dur": 0,
    "args": {
      "External id": 198011,
      "cbid": 205, "correlation": 198011
    }
  },
  {
    "ph": "f", "id": 198011, "pid": 76337, "tid": -914061504, "ts": 1716454224510393,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564765, "dur": 6,
    "args": {
      "External id": 198015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198015,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198015, "pid": 5, "tid": 7, "ts": 1716454224564765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510407, "dur": 12,
    "args": {
      "External id": 198015,
      "cbid": 211, "correlation": 198015
    }
  },
  {
    "ph": "s", "id": 198015, "pid": 76337, "tid": -914061504, "ts": 1716454224510407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224564773, "dur": 320,
    "args": {
      "External id": 198017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198017,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198017, "pid": 5, "tid": 7, "ts": 1716454224564773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510422, "dur": 5,
    "args": {
      "External id": 198017,
      "cbid": 211, "correlation": 198017
    }
  },
  {
    "ph": "s", "id": 198017, "pid": 76337, "tid": -914061504, "ts": 1716454224510422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224565095, "dur": 1,
    "args": {
      "External id": 198019,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198019,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 198019, "pid": 5, "tid": 7, "ts": 1716454224565095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510433, "dur": 8,
    "args": {
      "External id": 198019,
      "cbid": 51, "correlation": 198019
    }
  },
  {
    "ph": "s", "id": 198019, "pid": 76337, "tid": -914061504, "ts": 1716454224510433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224565099, "dur": 494,
    "args": {
      "External id": 198020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198020,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198020, "pid": 5, "tid": 7, "ts": 1716454224565099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510442, "dur": 6,
    "args": {
      "External id": 198020,
      "cbid": 211, "correlation": 198020
    }
  },
  {
    "ph": "s", "id": 198020, "pid": 76337, "tid": -914061504, "ts": 1716454224510442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224565594, "dur": 6,
    "args": {
      "External id": 198022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198022,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198022, "pid": 5, "tid": 7, "ts": 1716454224565594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510451, "dur": 5,
    "args": {
      "External id": 198022,
      "cbid": 211, "correlation": 198022
    }
  },
  {
    "ph": "s", "id": 198022, "pid": 76337, "tid": -914061504, "ts": 1716454224510451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224565601, "dur": 6,
    "args": {
      "External id": 198028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198028, "pid": 5, "tid": 7, "ts": 1716454224565601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510481, "dur": 9,
    "args": {
      "External id": 198028,
      "cbid": 211, "correlation": 198028
    }
  },
  {
    "ph": "s", "id": 198028, "pid": 76337, "tid": -914061504, "ts": 1716454224510481,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224565609, "dur": 3,
    "args": {
      "External id": 198036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198036,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 198036, "pid": 5, "tid": 7, "ts": 1716454224565609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510524, "dur": 9,
    "args": {
      "External id": 198036,
      "cbid": 211, "correlation": 198036
    }
  },
  {
    "ph": "s", "id": 198036, "pid": 76337, "tid": -914061504, "ts": 1716454224510524,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510587, "dur": 1,
    "args": {
      "External id": 198052,
      "cbid": 251, "correlation": 198052
    }
  },
  {
    "ph": "f", "id": 198052, "pid": 76337, "tid": -914061504, "ts": 1716454224510587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510592, "dur": 0,
    "args": {
      "External id": 198054,
      "cbid": 251, "correlation": 198054
    }
  },
  {
    "ph": "f", "id": 198054, "pid": 76337, "tid": -914061504, "ts": 1716454224510592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224565614, "dur": 11,
    "args": {
      "External id": 198055,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198055,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198055, "pid": 5, "tid": 7, "ts": 1716454224565614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510594, "dur": 11,
    "args": {
      "External id": 198055,
      "cbid": 211, "correlation": 198055
    }
  },
  {
    "ph": "s", "id": 198055, "pid": 76337, "tid": -914061504, "ts": 1716454224510594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224565626, "dur": 5,
    "args": {
      "External id": 198057,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198057,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198057, "pid": 5, "tid": 7, "ts": 1716454224565626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510606, "dur": 6,
    "args": {
      "External id": 198057,
      "cbid": 211, "correlation": 198057
    }
  },
  {
    "ph": "s", "id": 198057, "pid": 76337, "tid": -914061504, "ts": 1716454224510606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224565632, "dur": 6,
    "args": {
      "External id": 198067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198067,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198067, "pid": 5, "tid": 7, "ts": 1716454224565632,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510663, "dur": 12,
    "args": {
      "External id": 198067,
      "cbid": 211, "correlation": 198067
    }
  },
  {
    "ph": "s", "id": 198067, "pid": 76337, "tid": -914061504, "ts": 1716454224510663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224565639, "dur": 9,
    "args": {
      "External id": 198087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198087,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198087, "pid": 5, "tid": 7, "ts": 1716454224565639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510729, "dur": 11,
    "args": {
      "External id": 198087,
      "cbid": 211, "correlation": 198087
    }
  },
  {
    "ph": "s", "id": 198087, "pid": 76337, "tid": -914061504, "ts": 1716454224510729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224565650, "dur": 4,
    "args": {
      "External id": 198099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198099,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 198099, "pid": 5, "tid": 7, "ts": 1716454224565650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510750, "dur": 6,
    "args": {
      "External id": 198099,
      "cbid": 211, "correlation": 198099
    }
  },
  {
    "ph": "s", "id": 198099, "pid": 76337, "tid": -914061504, "ts": 1716454224510750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224565655, "dur": 7,
    "args": {
      "External id": 198102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198102, "pid": 5, "tid": 7, "ts": 1716454224565655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510769, "dur": 6,
    "args": {
      "External id": 198102,
      "cbid": 211, "correlation": 198102
    }
  },
  {
    "ph": "s", "id": 198102, "pid": 76337, "tid": -914061504, "ts": 1716454224510769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224565663, "dur": 5,
    "args": {
      "External id": 198111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198111,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198111, "pid": 5, "tid": 7, "ts": 1716454224565663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510808, "dur": 10,
    "args": {
      "External id": 198111,
      "cbid": 211, "correlation": 198111
    }
  },
  {
    "ph": "s", "id": 198111, "pid": 76337, "tid": -914061504, "ts": 1716454224510808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510871, "dur": 0,
    "args": {
      "External id": 198121,
      "cbid": 317, "correlation": 198121
    }
  },
  {
    "ph": "f", "id": 198121, "pid": 76337, "tid": -914061504, "ts": 1716454224510871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510872, "dur": 0,
    "args": {
      "External id": 198122,
      "cbid": 203, "correlation": 198122
    }
  },
  {
    "ph": "f", "id": 198122, "pid": 76337, "tid": -914061504, "ts": 1716454224510872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510873, "dur": 0,
    "args": {
      "External id": 198123,
      "cbid": 205, "correlation": 198123
    }
  },
  {
    "ph": "f", "id": 198123, "pid": 76337, "tid": -914061504, "ts": 1716454224510873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224565669, "dur": 5,
    "args": {
      "External id": 198127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198127,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198127, "pid": 5, "tid": 7, "ts": 1716454224565669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510887, "dur": 12,
    "args": {
      "External id": 198127,
      "cbid": 211, "correlation": 198127
    }
  },
  {
    "ph": "s", "id": 198127, "pid": 76337, "tid": -914061504, "ts": 1716454224510887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224565675, "dur": 161,
    "args": {
      "External id": 198129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198129, "pid": 5, "tid": 7, "ts": 1716454224565675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510901, "dur": 5,
    "args": {
      "External id": 198129,
      "cbid": 211, "correlation": 198129
    }
  },
  {
    "ph": "s", "id": 198129, "pid": 76337, "tid": -914061504, "ts": 1716454224510901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224565838, "dur": 1,
    "args": {
      "External id": 198131,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198131,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 198131, "pid": 5, "tid": 7, "ts": 1716454224565838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510911, "dur": 8,
    "args": {
      "External id": 198131,
      "cbid": 51, "correlation": 198131
    }
  },
  {
    "ph": "s", "id": 198131, "pid": 76337, "tid": -914061504, "ts": 1716454224510911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224565842, "dur": 258,
    "args": {
      "External id": 198132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198132,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198132, "pid": 5, "tid": 7, "ts": 1716454224565842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510920, "dur": 6,
    "args": {
      "External id": 198132,
      "cbid": 211, "correlation": 198132
    }
  },
  {
    "ph": "s", "id": 198132, "pid": 76337, "tid": -914061504, "ts": 1716454224510920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224566102, "dur": 6,
    "args": {
      "External id": 198134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198134, "pid": 5, "tid": 7, "ts": 1716454224566102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510930, "dur": 5,
    "args": {
      "External id": 198134,
      "cbid": 211, "correlation": 198134
    }
  },
  {
    "ph": "s", "id": 198134, "pid": 76337, "tid": -914061504, "ts": 1716454224510930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224566109, "dur": 6,
    "args": {
      "External id": 198140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198140, "pid": 5, "tid": 7, "ts": 1716454224566109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224510958, "dur": 10,
    "args": {
      "External id": 198140,
      "cbid": 211, "correlation": 198140
    }
  },
  {
    "ph": "s", "id": 198140, "pid": 76337, "tid": -914061504, "ts": 1716454224510958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511027, "dur": 0,
    "args": {
      "External id": 198150,
      "cbid": 317, "correlation": 198150
    }
  },
  {
    "ph": "f", "id": 198150, "pid": 76337, "tid": -914061504, "ts": 1716454224511027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511028, "dur": 0,
    "args": {
      "External id": 198151,
      "cbid": 203, "correlation": 198151
    }
  },
  {
    "ph": "f", "id": 198151, "pid": 76337, "tid": -914061504, "ts": 1716454224511028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511028, "dur": 0,
    "args": {
      "External id": 198152,
      "cbid": 205, "correlation": 198152
    }
  },
  {
    "ph": "f", "id": 198152, "pid": 76337, "tid": -914061504, "ts": 1716454224511028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224566117, "dur": 7,
    "args": {
      "External id": 198156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198156, "pid": 5, "tid": 7, "ts": 1716454224566117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511043, "dur": 12,
    "args": {
      "External id": 198156,
      "cbid": 211, "correlation": 198156
    }
  },
  {
    "ph": "s", "id": 198156, "pid": 76337, "tid": -914061504, "ts": 1716454224511043,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224566125, "dur": 3,
    "args": {
      "External id": 198158,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198158,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 198158, "pid": 5, "tid": 7, "ts": 1716454224566125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511062, "dur": 21,
    "args": {
      "External id": 198158,
      "cbid": 51, "correlation": 198158
    }
  },
  {
    "ph": "s", "id": 198158, "pid": 76337, "tid": -914061504, "ts": 1716454224511062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224566129, "dur": 97,
    "args": {
      "External id": 198159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198159,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 198159, "pid": 5, "tid": 7, "ts": 1716454224566129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511084, "dur": 6,
    "args": {
      "External id": 198159,
      "cbid": 211, "correlation": 198159
    }
  },
  {
    "ph": "s", "id": 198159, "pid": 76337, "tid": -914061504, "ts": 1716454224511084,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224566228, "dur": 6,
    "args": {
      "External id": 198161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198161,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198161, "pid": 5, "tid": 7, "ts": 1716454224566228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511095, "dur": 5,
    "args": {
      "External id": 198161,
      "cbid": 211, "correlation": 198161
    }
  },
  {
    "ph": "s", "id": 198161, "pid": 76337, "tid": -914061504, "ts": 1716454224511095,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224566235, "dur": 6,
    "args": {
      "External id": 198167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198167,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198167, "pid": 5, "tid": 7, "ts": 1716454224566235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511123, "dur": 9,
    "args": {
      "External id": 198167,
      "cbid": 211, "correlation": 198167
    }
  },
  {
    "ph": "s", "id": 198167, "pid": 76337, "tid": -914061504, "ts": 1716454224511123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224566242, "dur": 5,
    "args": {
      "External id": 198175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198175,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198175, "pid": 5, "tid": 7, "ts": 1716454224566242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511153, "dur": 9,
    "args": {
      "External id": 198175,
      "cbid": 211, "correlation": 198175
    }
  },
  {
    "ph": "s", "id": 198175, "pid": 76337, "tid": -914061504, "ts": 1716454224511153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224566248, "dur": 4,
    "args": {
      "External id": 198183,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198183,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198183, "pid": 5, "tid": 7, "ts": 1716454224566248,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511183, "dur": 8,
    "args": {
      "External id": 198183,
      "cbid": 211, "correlation": 198183
    }
  },
  {
    "ph": "s", "id": 198183, "pid": 76337, "tid": -914061504, "ts": 1716454224511183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224566254, "dur": 11,
    "args": {
      "External id": 198192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198192,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198192, "pid": 5, "tid": 7, "ts": 1716454224566254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511227, "dur": 10,
    "args": {
      "External id": 198192,
      "cbid": 211, "correlation": 198192
    }
  },
  {
    "ph": "s", "id": 198192, "pid": 76337, "tid": -914061504, "ts": 1716454224511227,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224566266, "dur": 12,
    "args": {
      "External id": 198212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198212,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198212, "pid": 5, "tid": 7, "ts": 1716454224566266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511296, "dur": 11,
    "args": {
      "External id": 198212,
      "cbid": 211, "correlation": 198212
    }
  },
  {
    "ph": "s", "id": 198212, "pid": 76337, "tid": -914061504, "ts": 1716454224511296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224566280, "dur": 4,
    "args": {
      "External id": 198224,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198224,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198224, "pid": 5, "tid": 7, "ts": 1716454224566280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511319, "dur": 6,
    "args": {
      "External id": 198224,
      "cbid": 211, "correlation": 198224
    }
  },
  {
    "ph": "s", "id": 198224, "pid": 76337, "tid": -914061504, "ts": 1716454224511319,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224566285, "dur": 10,
    "args": {
      "External id": 198227,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198227,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198227, "pid": 5, "tid": 7, "ts": 1716454224566285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511336, "dur": 7,
    "args": {
      "External id": 198227,
      "cbid": 211, "correlation": 198227
    }
  },
  {
    "ph": "s", "id": 198227, "pid": 76337, "tid": -914061504, "ts": 1716454224511336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224566297, "dur": 6,
    "args": {
      "External id": 198236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198236,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198236, "pid": 5, "tid": 7, "ts": 1716454224566297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511375, "dur": 10,
    "args": {
      "External id": 198236,
      "cbid": 211, "correlation": 198236
    }
  },
  {
    "ph": "s", "id": 198236, "pid": 76337, "tid": -914061504, "ts": 1716454224511375,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511428, "dur": 0,
    "args": {
      "External id": 198246,
      "cbid": 317, "correlation": 198246
    }
  },
  {
    "ph": "f", "id": 198246, "pid": 76337, "tid": -914061504, "ts": 1716454224511428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511428, "dur": 0,
    "args": {
      "External id": 198247,
      "cbid": 203, "correlation": 198247
    }
  },
  {
    "ph": "f", "id": 198247, "pid": 76337, "tid": -914061504, "ts": 1716454224511428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511429, "dur": 0,
    "args": {
      "External id": 198248,
      "cbid": 205, "correlation": 198248
    }
  },
  {
    "ph": "f", "id": 198248, "pid": 76337, "tid": -914061504, "ts": 1716454224511429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224566305, "dur": 6,
    "args": {
      "External id": 198252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198252,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198252, "pid": 5, "tid": 7, "ts": 1716454224566305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511442, "dur": 11,
    "args": {
      "External id": 198252,
      "cbid": 211, "correlation": 198252
    }
  },
  {
    "ph": "s", "id": 198252, "pid": 76337, "tid": -914061504, "ts": 1716454224511442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224566313, "dur": 321,
    "args": {
      "External id": 198254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198254,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198254, "pid": 5, "tid": 7, "ts": 1716454224566313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511456, "dur": 5,
    "args": {
      "External id": 198254,
      "cbid": 211, "correlation": 198254
    }
  },
  {
    "ph": "s", "id": 198254, "pid": 76337, "tid": -914061504, "ts": 1716454224511456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224566635, "dur": 1,
    "args": {
      "External id": 198256,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198256,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 198256, "pid": 5, "tid": 7, "ts": 1716454224566635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511467, "dur": 6,
    "args": {
      "External id": 198256,
      "cbid": 51, "correlation": 198256
    }
  },
  {
    "ph": "s", "id": 198256, "pid": 76337, "tid": -914061504, "ts": 1716454224511467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224566639, "dur": 497,
    "args": {
      "External id": 198257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198257,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198257, "pid": 5, "tid": 7, "ts": 1716454224566639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511474, "dur": 6,
    "args": {
      "External id": 198257,
      "cbid": 211, "correlation": 198257
    }
  },
  {
    "ph": "s", "id": 198257, "pid": 76337, "tid": -914061504, "ts": 1716454224511474,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567137, "dur": 6,
    "args": {
      "External id": 198259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198259,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198259, "pid": 5, "tid": 7, "ts": 1716454224567137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511484, "dur": 5,
    "args": {
      "External id": 198259,
      "cbid": 211, "correlation": 198259
    }
  },
  {
    "ph": "s", "id": 198259, "pid": 76337, "tid": -914061504, "ts": 1716454224511484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224567144, "dur": 6,
    "args": {
      "External id": 198265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198265,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198265, "pid": 5, "tid": 7, "ts": 1716454224567144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511512, "dur": 9,
    "args": {
      "External id": 198265,
      "cbid": 211, "correlation": 198265
    }
  },
  {
    "ph": "s", "id": 198265, "pid": 76337, "tid": -914061504, "ts": 1716454224511512,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224567152, "dur": 3,
    "args": {
      "External id": 198273,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198273,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 198273, "pid": 5, "tid": 7, "ts": 1716454224567152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511555, "dur": 10,
    "args": {
      "External id": 198273,
      "cbid": 211, "correlation": 198273
    }
  },
  {
    "ph": "s", "id": 198273, "pid": 76337, "tid": -914061504, "ts": 1716454224511555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511617, "dur": 1,
    "args": {
      "External id": 198289,
      "cbid": 251, "correlation": 198289
    }
  },
  {
    "ph": "f", "id": 198289, "pid": 76337, "tid": -914061504, "ts": 1716454224511617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511623, "dur": 0,
    "args": {
      "External id": 198291,
      "cbid": 251, "correlation": 198291
    }
  },
  {
    "ph": "f", "id": 198291, "pid": 76337, "tid": -914061504, "ts": 1716454224511623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224567156, "dur": 13,
    "args": {
      "External id": 198292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198292,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198292, "pid": 5, "tid": 7, "ts": 1716454224567156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511624, "dur": 11,
    "args": {
      "External id": 198292,
      "cbid": 211, "correlation": 198292
    }
  },
  {
    "ph": "s", "id": 198292, "pid": 76337, "tid": -914061504, "ts": 1716454224511624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224567171, "dur": 5,
    "args": {
      "External id": 198294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198294,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198294, "pid": 5, "tid": 7, "ts": 1716454224567171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511637, "dur": 5,
    "args": {
      "External id": 198294,
      "cbid": 211, "correlation": 198294
    }
  },
  {
    "ph": "s", "id": 198294, "pid": 76337, "tid": -914061504, "ts": 1716454224511637,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224567178, "dur": 6,
    "args": {
      "External id": 198304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198304,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198304, "pid": 5, "tid": 7, "ts": 1716454224567178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511693, "dur": 12,
    "args": {
      "External id": 198304,
      "cbid": 211, "correlation": 198304
    }
  },
  {
    "ph": "s", "id": 198304, "pid": 76337, "tid": -914061504, "ts": 1716454224511693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567185, "dur": 9,
    "args": {
      "External id": 198324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198324,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198324, "pid": 5, "tid": 7, "ts": 1716454224567185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511759, "dur": 11,
    "args": {
      "External id": 198324,
      "cbid": 211, "correlation": 198324
    }
  },
  {
    "ph": "s", "id": 198324, "pid": 76337, "tid": -914061504, "ts": 1716454224511759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224567195, "dur": 4,
    "args": {
      "External id": 198336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198336,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 198336, "pid": 5, "tid": 7, "ts": 1716454224567195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511780, "dur": 6,
    "args": {
      "External id": 198336,
      "cbid": 211, "correlation": 198336
    }
  },
  {
    "ph": "s", "id": 198336, "pid": 76337, "tid": -914061504, "ts": 1716454224511780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224567200, "dur": 7,
    "args": {
      "External id": 198339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198339,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198339, "pid": 5, "tid": 7, "ts": 1716454224567200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511798, "dur": 6,
    "args": {
      "External id": 198339,
      "cbid": 211, "correlation": 198339
    }
  },
  {
    "ph": "s", "id": 198339, "pid": 76337, "tid": -914061504, "ts": 1716454224511798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224567208, "dur": 5,
    "args": {
      "External id": 198348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198348,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198348, "pid": 5, "tid": 7, "ts": 1716454224567208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511838, "dur": 10,
    "args": {
      "External id": 198348,
      "cbid": 211, "correlation": 198348
    }
  },
  {
    "ph": "s", "id": 198348, "pid": 76337, "tid": -914061504, "ts": 1716454224511838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511900, "dur": 0,
    "args": {
      "External id": 198358,
      "cbid": 317, "correlation": 198358
    }
  },
  {
    "ph": "f", "id": 198358, "pid": 76337, "tid": -914061504, "ts": 1716454224511900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511901, "dur": 0,
    "args": {
      "External id": 198359,
      "cbid": 203, "correlation": 198359
    }
  },
  {
    "ph": "f", "id": 198359, "pid": 76337, "tid": -914061504, "ts": 1716454224511901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511901, "dur": 0,
    "args": {
      "External id": 198360,
      "cbid": 205, "correlation": 198360
    }
  },
  {
    "ph": "f", "id": 198360, "pid": 76337, "tid": -914061504, "ts": 1716454224511901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567214, "dur": 5,
    "args": {
      "External id": 198364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198364,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198364, "pid": 5, "tid": 7, "ts": 1716454224567214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511916, "dur": 12,
    "args": {
      "External id": 198364,
      "cbid": 211, "correlation": 198364
    }
  },
  {
    "ph": "s", "id": 198364, "pid": 76337, "tid": -914061504, "ts": 1716454224511916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567221, "dur": 162,
    "args": {
      "External id": 198366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198366,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198366, "pid": 5, "tid": 7, "ts": 1716454224567221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511931, "dur": 5,
    "args": {
      "External id": 198366,
      "cbid": 211, "correlation": 198366
    }
  },
  {
    "ph": "s", "id": 198366, "pid": 76337, "tid": -914061504, "ts": 1716454224511931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224567385, "dur": 1,
    "args": {
      "External id": 198368,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198368,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 198368, "pid": 5, "tid": 7, "ts": 1716454224567385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511941, "dur": 6,
    "args": {
      "External id": 198368,
      "cbid": 51, "correlation": 198368
    }
  },
  {
    "ph": "s", "id": 198368, "pid": 76337, "tid": -914061504, "ts": 1716454224511941,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224567388, "dur": 260,
    "args": {
      "External id": 198369,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198369,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198369, "pid": 5, "tid": 7, "ts": 1716454224567388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511949, "dur": 6,
    "args": {
      "External id": 198369,
      "cbid": 211, "correlation": 198369
    }
  },
  {
    "ph": "s", "id": 198369, "pid": 76337, "tid": -914061504, "ts": 1716454224511949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567649, "dur": 6,
    "args": {
      "External id": 198371,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198371,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198371, "pid": 5, "tid": 7, "ts": 1716454224567649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511960, "dur": 6,
    "args": {
      "External id": 198371,
      "cbid": 211, "correlation": 198371
    }
  },
  {
    "ph": "s", "id": 198371, "pid": 76337, "tid": -914061504, "ts": 1716454224511960,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224567657, "dur": 6,
    "args": {
      "External id": 198377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198377,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198377, "pid": 5, "tid": 7, "ts": 1716454224567657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224511997, "dur": 9,
    "args": {
      "External id": 198377,
      "cbid": 211, "correlation": 198377
    }
  },
  {
    "ph": "s", "id": 198377, "pid": 76337, "tid": -914061504, "ts": 1716454224511997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512057, "dur": 0,
    "args": {
      "External id": 198387,
      "cbid": 317, "correlation": 198387
    }
  },
  {
    "ph": "f", "id": 198387, "pid": 76337, "tid": -914061504, "ts": 1716454224512057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512058, "dur": 0,
    "args": {
      "External id": 198388,
      "cbid": 203, "correlation": 198388
    }
  },
  {
    "ph": "f", "id": 198388, "pid": 76337, "tid": -914061504, "ts": 1716454224512058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512059, "dur": 0,
    "args": {
      "External id": 198389,
      "cbid": 205, "correlation": 198389
    }
  },
  {
    "ph": "f", "id": 198389, "pid": 76337, "tid": -914061504, "ts": 1716454224512059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567664, "dur": 7,
    "args": {
      "External id": 198393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198393,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198393, "pid": 5, "tid": 7, "ts": 1716454224567664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512071, "dur": 11,
    "args": {
      "External id": 198393,
      "cbid": 211, "correlation": 198393
    }
  },
  {
    "ph": "s", "id": 198393, "pid": 76337, "tid": -914061504, "ts": 1716454224512071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224567673, "dur": 3,
    "args": {
      "External id": 198395,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198395,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.3513513513513513
    }
  },
  {
    "ph": "f", "id": 198395, "pid": 5, "tid": 7, "ts": 1716454224567673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512088, "dur": 10,
    "args": {
      "External id": 198395,
      "cbid": 51, "correlation": 198395
    }
  },
  {
    "ph": "s", "id": 198395, "pid": 76337, "tid": -914061504, "ts": 1716454224512088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224567677, "dur": 94,
    "args": {
      "External id": 198396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198396,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 198396, "pid": 5, "tid": 7, "ts": 1716454224567677,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512098, "dur": 6,
    "args": {
      "External id": 198396,
      "cbid": 211, "correlation": 198396
    }
  },
  {
    "ph": "s", "id": 198396, "pid": 76337, "tid": -914061504, "ts": 1716454224512098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567772, "dur": 6,
    "args": {
      "External id": 198398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198398,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198398, "pid": 5, "tid": 7, "ts": 1716454224567772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512108, "dur": 6,
    "args": {
      "External id": 198398,
      "cbid": 211, "correlation": 198398
    }
  },
  {
    "ph": "s", "id": 198398, "pid": 76337, "tid": -914061504, "ts": 1716454224512108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224567780, "dur": 6,
    "args": {
      "External id": 198404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198404,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198404, "pid": 5, "tid": 7, "ts": 1716454224567780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512135, "dur": 8,
    "args": {
      "External id": 198404,
      "cbid": 211, "correlation": 198404
    }
  },
  {
    "ph": "s", "id": 198404, "pid": 76337, "tid": -914061504, "ts": 1716454224512135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224567787, "dur": 5,
    "args": {
      "External id": 198412,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198412,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198412, "pid": 5, "tid": 7, "ts": 1716454224567787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512166, "dur": 8,
    "args": {
      "External id": 198412,
      "cbid": 211, "correlation": 198412
    }
  },
  {
    "ph": "s", "id": 198412, "pid": 76337, "tid": -914061504, "ts": 1716454224512166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224567793, "dur": 4,
    "args": {
      "External id": 198420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198420,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198420, "pid": 5, "tid": 7, "ts": 1716454224567793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512194, "dur": 8,
    "args": {
      "External id": 198420,
      "cbid": 211, "correlation": 198420
    }
  },
  {
    "ph": "s", "id": 198420, "pid": 76337, "tid": -914061504, "ts": 1716454224512194,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224567799, "dur": 11,
    "args": {
      "External id": 198429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198429,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198429, "pid": 5, "tid": 7, "ts": 1716454224567799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512238, "dur": 10,
    "args": {
      "External id": 198429,
      "cbid": 211, "correlation": 198429
    }
  },
  {
    "ph": "s", "id": 198429, "pid": 76337, "tid": -914061504, "ts": 1716454224512238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567812, "dur": 12,
    "args": {
      "External id": 198449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198449,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198449, "pid": 5, "tid": 7, "ts": 1716454224567812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512308, "dur": 12,
    "args": {
      "External id": 198449,
      "cbid": 211, "correlation": 198449
    }
  },
  {
    "ph": "s", "id": 198449, "pid": 76337, "tid": -914061504, "ts": 1716454224512308,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224567825, "dur": 4,
    "args": {
      "External id": 198461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198461,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198461, "pid": 5, "tid": 7, "ts": 1716454224567825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512330, "dur": 6,
    "args": {
      "External id": 198461,
      "cbid": 211, "correlation": 198461
    }
  },
  {
    "ph": "s", "id": 198461, "pid": 76337, "tid": -914061504, "ts": 1716454224512330,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224567831, "dur": 11,
    "args": {
      "External id": 198464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198464, "pid": 5, "tid": 7, "ts": 1716454224567831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512347, "dur": 6,
    "args": {
      "External id": 198464,
      "cbid": 211, "correlation": 198464
    }
  },
  {
    "ph": "s", "id": 198464, "pid": 76337, "tid": -914061504, "ts": 1716454224512347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224567843, "dur": 6,
    "args": {
      "External id": 198473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198473,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198473, "pid": 5, "tid": 7, "ts": 1716454224567843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512385, "dur": 9,
    "args": {
      "External id": 198473,
      "cbid": 211, "correlation": 198473
    }
  },
  {
    "ph": "s", "id": 198473, "pid": 76337, "tid": -914061504, "ts": 1716454224512385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512437, "dur": 0,
    "args": {
      "External id": 198483,
      "cbid": 317, "correlation": 198483
    }
  },
  {
    "ph": "f", "id": 198483, "pid": 76337, "tid": -914061504, "ts": 1716454224512437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512438, "dur": 0,
    "args": {
      "External id": 198484,
      "cbid": 203, "correlation": 198484
    }
  },
  {
    "ph": "f", "id": 198484, "pid": 76337, "tid": -914061504, "ts": 1716454224512438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512438, "dur": 0,
    "args": {
      "External id": 198485,
      "cbid": 205, "correlation": 198485
    }
  },
  {
    "ph": "f", "id": 198485, "pid": 76337, "tid": -914061504, "ts": 1716454224512438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567850, "dur": 7,
    "args": {
      "External id": 198489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198489,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198489, "pid": 5, "tid": 7, "ts": 1716454224567850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512451, "dur": 12,
    "args": {
      "External id": 198489,
      "cbid": 211, "correlation": 198489
    }
  },
  {
    "ph": "s", "id": 198489, "pid": 76337, "tid": -914061504, "ts": 1716454224512451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224567858, "dur": 319,
    "args": {
      "External id": 198491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198491,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198491, "pid": 5, "tid": 7, "ts": 1716454224567858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512466, "dur": 5,
    "args": {
      "External id": 198491,
      "cbid": 211, "correlation": 198491
    }
  },
  {
    "ph": "s", "id": 198491, "pid": 76337, "tid": -914061504, "ts": 1716454224512466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224568180, "dur": 1,
    "args": {
      "External id": 198493,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198493,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 198493, "pid": 5, "tid": 7, "ts": 1716454224568180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512476, "dur": 6,
    "args": {
      "External id": 198493,
      "cbid": 51, "correlation": 198493
    }
  },
  {
    "ph": "s", "id": 198493, "pid": 76337, "tid": -914061504, "ts": 1716454224512476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224568183, "dur": 495,
    "args": {
      "External id": 198494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198494,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198494, "pid": 5, "tid": 7, "ts": 1716454224568183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512484, "dur": 6,
    "args": {
      "External id": 198494,
      "cbid": 211, "correlation": 198494
    }
  },
  {
    "ph": "s", "id": 198494, "pid": 76337, "tid": -914061504, "ts": 1716454224512484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224568680, "dur": 5,
    "args": {
      "External id": 198496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198496,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198496, "pid": 5, "tid": 7, "ts": 1716454224568680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512494, "dur": 5,
    "args": {
      "External id": 198496,
      "cbid": 211, "correlation": 198496
    }
  },
  {
    "ph": "s", "id": 198496, "pid": 76337, "tid": -914061504, "ts": 1716454224512494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224568687, "dur": 6,
    "args": {
      "External id": 198502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198502, "pid": 5, "tid": 7, "ts": 1716454224568687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512522, "dur": 9,
    "args": {
      "External id": 198502,
      "cbid": 211, "correlation": 198502
    }
  },
  {
    "ph": "s", "id": 198502, "pid": 76337, "tid": -914061504, "ts": 1716454224512522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224568695, "dur": 3,
    "args": {
      "External id": 198510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198510,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 198510, "pid": 5, "tid": 7, "ts": 1716454224568695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512567, "dur": 9,
    "args": {
      "External id": 198510,
      "cbid": 211, "correlation": 198510
    }
  },
  {
    "ph": "s", "id": 198510, "pid": 76337, "tid": -914061504, "ts": 1716454224512567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512628, "dur": 1,
    "args": {
      "External id": 198526,
      "cbid": 251, "correlation": 198526
    }
  },
  {
    "ph": "f", "id": 198526, "pid": 76337, "tid": -914061504, "ts": 1716454224512628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512633, "dur": 0,
    "args": {
      "External id": 198528,
      "cbid": 251, "correlation": 198528
    }
  },
  {
    "ph": "f", "id": 198528, "pid": 76337, "tid": -914061504, "ts": 1716454224512633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224568699, "dur": 13,
    "args": {
      "External id": 198529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198529,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198529, "pid": 5, "tid": 7, "ts": 1716454224568699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512635, "dur": 11,
    "args": {
      "External id": 198529,
      "cbid": 211, "correlation": 198529
    }
  },
  {
    "ph": "s", "id": 198529, "pid": 76337, "tid": -914061504, "ts": 1716454224512635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224568713, "dur": 5,
    "args": {
      "External id": 198531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198531,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198531, "pid": 5, "tid": 7, "ts": 1716454224568713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512647, "dur": 5,
    "args": {
      "External id": 198531,
      "cbid": 211, "correlation": 198531
    }
  },
  {
    "ph": "s", "id": 198531, "pid": 76337, "tid": -914061504, "ts": 1716454224512647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224568720, "dur": 6,
    "args": {
      "External id": 198541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198541,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198541, "pid": 5, "tid": 7, "ts": 1716454224568720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512702, "dur": 12,
    "args": {
      "External id": 198541,
      "cbid": 211, "correlation": 198541
    }
  },
  {
    "ph": "s", "id": 198541, "pid": 76337, "tid": -914061504, "ts": 1716454224512702,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224568727, "dur": 9,
    "args": {
      "External id": 198561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198561,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198561, "pid": 5, "tid": 7, "ts": 1716454224568727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512769, "dur": 11,
    "args": {
      "External id": 198561,
      "cbid": 211, "correlation": 198561
    }
  },
  {
    "ph": "s", "id": 198561, "pid": 76337, "tid": -914061504, "ts": 1716454224512769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224568737, "dur": 4,
    "args": {
      "External id": 198573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198573,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 198573, "pid": 5, "tid": 7, "ts": 1716454224568737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512789, "dur": 6,
    "args": {
      "External id": 198573,
      "cbid": 211, "correlation": 198573
    }
  },
  {
    "ph": "s", "id": 198573, "pid": 76337, "tid": -914061504, "ts": 1716454224512789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224568742, "dur": 6,
    "args": {
      "External id": 198576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198576,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198576, "pid": 5, "tid": 7, "ts": 1716454224568742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512808, "dur": 7,
    "args": {
      "External id": 198576,
      "cbid": 211, "correlation": 198576
    }
  },
  {
    "ph": "s", "id": 198576, "pid": 76337, "tid": -914061504, "ts": 1716454224512808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224568750, "dur": 5,
    "args": {
      "External id": 198585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198585,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198585, "pid": 5, "tid": 7, "ts": 1716454224568750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512848, "dur": 10,
    "args": {
      "External id": 198585,
      "cbid": 211, "correlation": 198585
    }
  },
  {
    "ph": "s", "id": 198585, "pid": 76337, "tid": -914061504, "ts": 1716454224512848,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512910, "dur": 0,
    "args": {
      "External id": 198595,
      "cbid": 317, "correlation": 198595
    }
  },
  {
    "ph": "f", "id": 198595, "pid": 76337, "tid": -914061504, "ts": 1716454224512910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512911, "dur": 0,
    "args": {
      "External id": 198596,
      "cbid": 203, "correlation": 198596
    }
  },
  {
    "ph": "f", "id": 198596, "pid": 76337, "tid": -914061504, "ts": 1716454224512911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512911, "dur": 0,
    "args": {
      "External id": 198597,
      "cbid": 205, "correlation": 198597
    }
  },
  {
    "ph": "f", "id": 198597, "pid": 76337, "tid": -914061504, "ts": 1716454224512911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224568756, "dur": 5,
    "args": {
      "External id": 198601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198601,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198601, "pid": 5, "tid": 7, "ts": 1716454224568756,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512926, "dur": 12,
    "args": {
      "External id": 198601,
      "cbid": 211, "correlation": 198601
    }
  },
  {
    "ph": "s", "id": 198601, "pid": 76337, "tid": -914061504, "ts": 1716454224512926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224568762, "dur": 163,
    "args": {
      "External id": 198603,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198603,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198603, "pid": 5, "tid": 7, "ts": 1716454224568762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512940, "dur": 5,
    "args": {
      "External id": 198603,
      "cbid": 211, "correlation": 198603
    }
  },
  {
    "ph": "s", "id": 198603, "pid": 76337, "tid": -914061504, "ts": 1716454224512940,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224568927, "dur": 1,
    "args": {
      "External id": 198605,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198605,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 198605, "pid": 5, "tid": 7, "ts": 1716454224568927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512951, "dur": 6,
    "args": {
      "External id": 198605,
      "cbid": 51, "correlation": 198605
    }
  },
  {
    "ph": "s", "id": 198605, "pid": 76337, "tid": -914061504, "ts": 1716454224512951,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224568931, "dur": 259,
    "args": {
      "External id": 198606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198606,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198606, "pid": 5, "tid": 7, "ts": 1716454224568931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512958, "dur": 6,
    "args": {
      "External id": 198606,
      "cbid": 211, "correlation": 198606
    }
  },
  {
    "ph": "s", "id": 198606, "pid": 76337, "tid": -914061504, "ts": 1716454224512958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224569191, "dur": 6,
    "args": {
      "External id": 198608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198608,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198608, "pid": 5, "tid": 7, "ts": 1716454224569191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224512968, "dur": 13,
    "args": {
      "External id": 198608,
      "cbid": 211, "correlation": 198608
    }
  },
  {
    "ph": "s", "id": 198608, "pid": 76337, "tid": -914061504, "ts": 1716454224512968,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224569198, "dur": 6,
    "args": {
      "External id": 198614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198614,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198614, "pid": 5, "tid": 7, "ts": 1716454224569198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513004, "dur": 9,
    "args": {
      "External id": 198614,
      "cbid": 211, "correlation": 198614
    }
  },
  {
    "ph": "s", "id": 198614, "pid": 76337, "tid": -914061504, "ts": 1716454224513004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513063, "dur": 0,
    "args": {
      "External id": 198624,
      "cbid": 317, "correlation": 198624
    }
  },
  {
    "ph": "f", "id": 198624, "pid": 76337, "tid": -914061504, "ts": 1716454224513063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513064, "dur": 0,
    "args": {
      "External id": 198625,
      "cbid": 203, "correlation": 198625
    }
  },
  {
    "ph": "f", "id": 198625, "pid": 76337, "tid": -914061504, "ts": 1716454224513064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513065, "dur": 0,
    "args": {
      "External id": 198626,
      "cbid": 205, "correlation": 198626
    }
  },
  {
    "ph": "f", "id": 198626, "pid": 76337, "tid": -914061504, "ts": 1716454224513065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224569206, "dur": 8,
    "args": {
      "External id": 198630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198630,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198630, "pid": 5, "tid": 7, "ts": 1716454224569206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513077, "dur": 12,
    "args": {
      "External id": 198630,
      "cbid": 211, "correlation": 198630
    }
  },
  {
    "ph": "s", "id": 198630, "pid": 76337, "tid": -914061504, "ts": 1716454224513077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224569215, "dur": 3,
    "args": {
      "External id": 198632,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198632,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 198632, "pid": 5, "tid": 7, "ts": 1716454224569215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513094, "dur": 10,
    "args": {
      "External id": 198632,
      "cbid": 51, "correlation": 198632
    }
  },
  {
    "ph": "s", "id": 198632, "pid": 76337, "tid": -914061504, "ts": 1716454224513094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224569219, "dur": 93,
    "args": {
      "External id": 198633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198633,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 198633, "pid": 5, "tid": 7, "ts": 1716454224569219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513105, "dur": 6,
    "args": {
      "External id": 198633,
      "cbid": 211, "correlation": 198633
    }
  },
  {
    "ph": "s", "id": 198633, "pid": 76337, "tid": -914061504, "ts": 1716454224513105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224569314, "dur": 5,
    "args": {
      "External id": 198635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198635,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198635, "pid": 5, "tid": 7, "ts": 1716454224569314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513115, "dur": 6,
    "args": {
      "External id": 198635,
      "cbid": 211, "correlation": 198635
    }
  },
  {
    "ph": "s", "id": 198635, "pid": 76337, "tid": -914061504, "ts": 1716454224513115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224569320, "dur": 6,
    "args": {
      "External id": 198641,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198641,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198641, "pid": 5, "tid": 7, "ts": 1716454224569320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513142, "dur": 8,
    "args": {
      "External id": 198641,
      "cbid": 211, "correlation": 198641
    }
  },
  {
    "ph": "s", "id": 198641, "pid": 76337, "tid": -914061504, "ts": 1716454224513142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224569328, "dur": 5,
    "args": {
      "External id": 198649,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198649,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198649, "pid": 5, "tid": 7, "ts": 1716454224569328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513171, "dur": 8,
    "args": {
      "External id": 198649,
      "cbid": 211, "correlation": 198649
    }
  },
  {
    "ph": "s", "id": 198649, "pid": 76337, "tid": -914061504, "ts": 1716454224513171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224569334, "dur": 4,
    "args": {
      "External id": 198657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198657,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 198657, "pid": 5, "tid": 7, "ts": 1716454224569334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513199, "dur": 8,
    "args": {
      "External id": 198657,
      "cbid": 211, "correlation": 198657
    }
  },
  {
    "ph": "s", "id": 198657, "pid": 76337, "tid": -914061504, "ts": 1716454224513199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224569339, "dur": 14,
    "args": {
      "External id": 198668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198668,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198668, "pid": 5, "tid": 7, "ts": 1716454224569339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513290, "dur": 14,
    "args": {
      "External id": 198668,
      "cbid": 211, "correlation": 198668
    }
  },
  {
    "ph": "s", "id": 198668, "pid": 76337, "tid": -914061504, "ts": 1716454224513290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513348, "dur": 0,
    "args": {
      "External id": 198678,
      "cbid": 317, "correlation": 198678
    }
  },
  {
    "ph": "f", "id": 198678, "pid": 76337, "tid": -914061504, "ts": 1716454224513348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513349, "dur": 0,
    "args": {
      "External id": 198679,
      "cbid": 203, "correlation": 198679
    }
  },
  {
    "ph": "f", "id": 198679, "pid": 76337, "tid": -914061504, "ts": 1716454224513349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513349, "dur": 0,
    "args": {
      "External id": 198680,
      "cbid": 205, "correlation": 198680
    }
  },
  {
    "ph": "f", "id": 198680, "pid": 76337, "tid": -914061504, "ts": 1716454224513349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224569355, "dur": 9,
    "args": {
      "External id": 198684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198684,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198684, "pid": 5, "tid": 7, "ts": 1716454224569355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513365, "dur": 12,
    "args": {
      "External id": 198684,
      "cbid": 211, "correlation": 198684
    }
  },
  {
    "ph": "s", "id": 198684, "pid": 76337, "tid": -914061504, "ts": 1716454224513365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224569365, "dur": 162,
    "args": {
      "External id": 198686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198686,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198686, "pid": 5, "tid": 7, "ts": 1716454224569365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513380, "dur": 5,
    "args": {
      "External id": 198686,
      "cbid": 211, "correlation": 198686
    }
  },
  {
    "ph": "s", "id": 198686, "pid": 76337, "tid": -914061504, "ts": 1716454224513380,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224569529, "dur": 1,
    "args": {
      "External id": 198688,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198688,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 198688, "pid": 5, "tid": 7, "ts": 1716454224569529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513390, "dur": 6,
    "args": {
      "External id": 198688,
      "cbid": 51, "correlation": 198688
    }
  },
  {
    "ph": "s", "id": 198688, "pid": 76337, "tid": -914061504, "ts": 1716454224513390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224569533, "dur": 648,
    "args": {
      "External id": 198689,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198689,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198689, "pid": 5, "tid": 7, "ts": 1716454224569533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513397, "dur": 6,
    "args": {
      "External id": 198689,
      "cbid": 211, "correlation": 198689
    }
  },
  {
    "ph": "s", "id": 198689, "pid": 76337, "tid": -914061504, "ts": 1716454224513397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224570183, "dur": 13,
    "args": {
      "External id": 198691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198691,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198691, "pid": 5, "tid": 7, "ts": 1716454224570183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513407, "dur": 5,
    "args": {
      "External id": 198691,
      "cbid": 211, "correlation": 198691
    }
  },
  {
    "ph": "s", "id": 198691, "pid": 76337, "tid": -914061504, "ts": 1716454224513407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224570197, "dur": 14,
    "args": {
      "External id": 198697,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198697,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198697, "pid": 5, "tid": 7, "ts": 1716454224570197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513436, "dur": 9,
    "args": {
      "External id": 198697,
      "cbid": 211, "correlation": 198697
    }
  },
  {
    "ph": "s", "id": 198697, "pid": 76337, "tid": -914061504, "ts": 1716454224513436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224570213, "dur": 30,
    "args": {
      "External id": 198706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198706,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198706, "pid": 5, "tid": 7, "ts": 1716454224570213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513527, "dur": 13,
    "args": {
      "External id": 198706,
      "cbid": 211, "correlation": 198706
    }
  },
  {
    "ph": "s", "id": 198706, "pid": 76337, "tid": -914061504, "ts": 1716454224513527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224570244, "dur": 31,
    "args": {
      "External id": 198726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198726,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198726, "pid": 5, "tid": 7, "ts": 1716454224570244,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513595, "dur": 11,
    "args": {
      "External id": 198726,
      "cbid": 211, "correlation": 198726
    }
  },
  {
    "ph": "s", "id": 198726, "pid": 76337, "tid": -914061504, "ts": 1716454224513595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224570276, "dur": 4,
    "args": {
      "External id": 198738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198738,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198738, "pid": 5, "tid": 7, "ts": 1716454224570276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513619, "dur": 7,
    "args": {
      "External id": 198738,
      "cbid": 211, "correlation": 198738
    }
  },
  {
    "ph": "s", "id": 198738, "pid": 76337, "tid": -914061504, "ts": 1716454224513619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224570282, "dur": 30,
    "args": {
      "External id": 198741,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198741,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198741, "pid": 5, "tid": 7, "ts": 1716454224570282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513638, "dur": 6,
    "args": {
      "External id": 198741,
      "cbid": 211, "correlation": 198741
    }
  },
  {
    "ph": "s", "id": 198741, "pid": 76337, "tid": -914061504, "ts": 1716454224513638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224570313, "dur": 21,
    "args": {
      "External id": 198750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198750,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198750, "pid": 5, "tid": 7, "ts": 1716454224570313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513678, "dur": 10,
    "args": {
      "External id": 198750,
      "cbid": 211, "correlation": 198750
    }
  },
  {
    "ph": "s", "id": 198750, "pid": 76337, "tid": -914061504, "ts": 1716454224513678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513733, "dur": 0,
    "args": {
      "External id": 198760,
      "cbid": 317, "correlation": 198760
    }
  },
  {
    "ph": "f", "id": 198760, "pid": 76337, "tid": -914061504, "ts": 1716454224513733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513734, "dur": 0,
    "args": {
      "External id": 198761,
      "cbid": 203, "correlation": 198761
    }
  },
  {
    "ph": "f", "id": 198761, "pid": 76337, "tid": -914061504, "ts": 1716454224513734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513735, "dur": 0,
    "args": {
      "External id": 198762,
      "cbid": 205, "correlation": 198762
    }
  },
  {
    "ph": "f", "id": 198762, "pid": 76337, "tid": -914061504, "ts": 1716454224513735,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224570335, "dur": 22,
    "args": {
      "External id": 198766,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198766,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198766, "pid": 5, "tid": 7, "ts": 1716454224570335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513751, "dur": 11,
    "args": {
      "External id": 198766,
      "cbid": 211, "correlation": 198766
    }
  },
  {
    "ph": "s", "id": 198766, "pid": 76337, "tid": -914061504, "ts": 1716454224513751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224570359, "dur": 319,
    "args": {
      "External id": 198768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198768, "pid": 5, "tid": 7, "ts": 1716454224570359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513765, "dur": 5,
    "args": {
      "External id": 198768,
      "cbid": 211, "correlation": 198768
    }
  },
  {
    "ph": "s", "id": 198768, "pid": 76337, "tid": -914061504, "ts": 1716454224513765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224570680, "dur": 1,
    "args": {
      "External id": 198770,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198770,
      "bytes": 960, "memory bandwidth (GB/s)": 0.600375234521576
    }
  },
  {
    "ph": "f", "id": 198770, "pid": 5, "tid": 7, "ts": 1716454224570680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513776, "dur": 7,
    "args": {
      "External id": 198770,
      "cbid": 51, "correlation": 198770
    }
  },
  {
    "ph": "s", "id": 198770, "pid": 76337, "tid": -914061504, "ts": 1716454224513776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224570684, "dur": 1240,
    "args": {
      "External id": 198771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198771,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198771, "pid": 5, "tid": 7, "ts": 1716454224570684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513784, "dur": 6,
    "args": {
      "External id": 198771,
      "cbid": 211, "correlation": 198771
    }
  },
  {
    "ph": "s", "id": 198771, "pid": 76337, "tid": -914061504, "ts": 1716454224513784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224571925, "dur": 12,
    "args": {
      "External id": 198773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198773,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198773, "pid": 5, "tid": 7, "ts": 1716454224571925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513794, "dur": 5,
    "args": {
      "External id": 198773,
      "cbid": 211, "correlation": 198773
    }
  },
  {
    "ph": "s", "id": 198773, "pid": 76337, "tid": -914061504, "ts": 1716454224513794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224571939, "dur": 15,
    "args": {
      "External id": 198779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198779, "pid": 5, "tid": 7, "ts": 1716454224571939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513824, "dur": 10,
    "args": {
      "External id": 198779,
      "cbid": 211, "correlation": 198779
    }
  },
  {
    "ph": "s", "id": 198779, "pid": 76337, "tid": -914061504, "ts": 1716454224513824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224571955, "dur": 3,
    "args": {
      "External id": 198787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198787,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 198787, "pid": 5, "tid": 7, "ts": 1716454224571955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513869, "dur": 10,
    "args": {
      "External id": 198787,
      "cbid": 211, "correlation": 198787
    }
  },
  {
    "ph": "s", "id": 198787, "pid": 76337, "tid": -914061504, "ts": 1716454224513869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513937, "dur": 1,
    "args": {
      "External id": 198803,
      "cbid": 251, "correlation": 198803
    }
  },
  {
    "ph": "f", "id": 198803, "pid": 76337, "tid": -914061504, "ts": 1716454224513937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513942, "dur": 0,
    "args": {
      "External id": 198805,
      "cbid": 251, "correlation": 198805
    }
  },
  {
    "ph": "f", "id": 198805, "pid": 76337, "tid": -914061504, "ts": 1716454224513942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224571960, "dur": 12,
    "args": {
      "External id": 198806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198806,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198806, "pid": 5, "tid": 7, "ts": 1716454224571960,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513944, "dur": 11,
    "args": {
      "External id": 198806,
      "cbid": 211, "correlation": 198806
    }
  },
  {
    "ph": "s", "id": 198806, "pid": 76337, "tid": -914061504, "ts": 1716454224513944,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224571973, "dur": 5,
    "args": {
      "External id": 198808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198808,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198808, "pid": 5, "tid": 7, "ts": 1716454224571973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224513957, "dur": 5,
    "args": {
      "External id": 198808,
      "cbid": 211, "correlation": 198808
    }
  },
  {
    "ph": "s", "id": 198808, "pid": 76337, "tid": -914061504, "ts": 1716454224513957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224571980, "dur": 16,
    "args": {
      "External id": 198818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198818,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198818, "pid": 5, "tid": 7, "ts": 1716454224571980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514025, "dur": 13,
    "args": {
      "External id": 198818,
      "cbid": 211, "correlation": 198818
    }
  },
  {
    "ph": "s", "id": 198818, "pid": 76337, "tid": -914061504, "ts": 1716454224514025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224571997, "dur": 17,
    "args": {
      "External id": 198838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198838,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198838, "pid": 5, "tid": 7, "ts": 1716454224571997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514094, "dur": 11,
    "args": {
      "External id": 198838,
      "cbid": 211, "correlation": 198838
    }
  },
  {
    "ph": "s", "id": 198838, "pid": 76337, "tid": -914061504, "ts": 1716454224514094,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224572016, "dur": 4,
    "args": {
      "External id": 198850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198850,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 198850, "pid": 5, "tid": 7, "ts": 1716454224572016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514114, "dur": 6,
    "args": {
      "External id": 198850,
      "cbid": 211, "correlation": 198850
    }
  },
  {
    "ph": "s", "id": 198850, "pid": 76337, "tid": -914061504, "ts": 1716454224514114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224572022, "dur": 17,
    "args": {
      "External id": 198853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198853,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198853, "pid": 5, "tid": 7, "ts": 1716454224572022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514134, "dur": 7,
    "args": {
      "External id": 198853,
      "cbid": 211, "correlation": 198853
    }
  },
  {
    "ph": "s", "id": 198853, "pid": 76337, "tid": -914061504, "ts": 1716454224514134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224572040, "dur": 11,
    "args": {
      "External id": 198862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198862,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198862, "pid": 5, "tid": 7, "ts": 1716454224572040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514174, "dur": 10,
    "args": {
      "External id": 198862,
      "cbid": 211, "correlation": 198862
    }
  },
  {
    "ph": "s", "id": 198862, "pid": 76337, "tid": -914061504, "ts": 1716454224514174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514236, "dur": 0,
    "args": {
      "External id": 198872,
      "cbid": 317, "correlation": 198872
    }
  },
  {
    "ph": "f", "id": 198872, "pid": 76337, "tid": -914061504, "ts": 1716454224514236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514237, "dur": 0,
    "args": {
      "External id": 198873,
      "cbid": 203, "correlation": 198873
    }
  },
  {
    "ph": "f", "id": 198873, "pid": 76337, "tid": -914061504, "ts": 1716454224514237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514238, "dur": 0,
    "args": {
      "External id": 198874,
      "cbid": 205, "correlation": 198874
    }
  },
  {
    "ph": "f", "id": 198874, "pid": 76337, "tid": -914061504, "ts": 1716454224514238,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224572052, "dur": 11,
    "args": {
      "External id": 198878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198878,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198878, "pid": 5, "tid": 7, "ts": 1716454224572052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514255, "dur": 12,
    "args": {
      "External id": 198878,
      "cbid": 211, "correlation": 198878
    }
  },
  {
    "ph": "s", "id": 198878, "pid": 76337, "tid": -914061504, "ts": 1716454224514255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224572064, "dur": 163,
    "args": {
      "External id": 198880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198880,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198880, "pid": 5, "tid": 7, "ts": 1716454224572064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514270, "dur": 5,
    "args": {
      "External id": 198880,
      "cbid": 211, "correlation": 198880
    }
  },
  {
    "ph": "s", "id": 198880, "pid": 76337, "tid": -914061504, "ts": 1716454224514270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224572230, "dur": 1,
    "args": {
      "External id": 198882,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 198882,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 198882, "pid": 5, "tid": 7, "ts": 1716454224572230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514281, "dur": 8,
    "args": {
      "External id": 198882,
      "cbid": 51, "correlation": 198882
    }
  },
  {
    "ph": "s", "id": 198882, "pid": 76337, "tid": -914061504, "ts": 1716454224514281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224572233, "dur": 649,
    "args": {
      "External id": 198883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198883,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 198883, "pid": 5, "tid": 7, "ts": 1716454224572233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514290, "dur": 7,
    "args": {
      "External id": 198883,
      "cbid": 211, "correlation": 198883
    }
  },
  {
    "ph": "s", "id": 198883, "pid": 76337, "tid": -914061504, "ts": 1716454224514290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224572883, "dur": 12,
    "args": {
      "External id": 198885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198885, "pid": 5, "tid": 7, "ts": 1716454224572883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514300, "dur": 5,
    "args": {
      "External id": 198885,
      "cbid": 211, "correlation": 198885
    }
  },
  {
    "ph": "s", "id": 198885, "pid": 76337, "tid": -914061504, "ts": 1716454224514300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224572897, "dur": 15,
    "args": {
      "External id": 198891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198891, "pid": 5, "tid": 7, "ts": 1716454224572897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514328, "dur": 9,
    "args": {
      "External id": 198891,
      "cbid": 211, "correlation": 198891
    }
  },
  {
    "ph": "s", "id": 198891, "pid": 76337, "tid": -914061504, "ts": 1716454224514328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514387, "dur": 0,
    "args": {
      "External id": 198901,
      "cbid": 317, "correlation": 198901
    }
  },
  {
    "ph": "f", "id": 198901, "pid": 76337, "tid": -914061504, "ts": 1716454224514387,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514388, "dur": 0,
    "args": {
      "External id": 198902,
      "cbid": 203, "correlation": 198902
    }
  },
  {
    "ph": "f", "id": 198902, "pid": 76337, "tid": -914061504, "ts": 1716454224514388,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514389, "dur": 0,
    "args": {
      "External id": 198903,
      "cbid": 205, "correlation": 198903
    }
  },
  {
    "ph": "f", "id": 198903, "pid": 76337, "tid": -914061504, "ts": 1716454224514389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224572913, "dur": 21,
    "args": {
      "External id": 198907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198907,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198907, "pid": 5, "tid": 7, "ts": 1716454224572913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514403, "dur": 11,
    "args": {
      "External id": 198907,
      "cbid": 211, "correlation": 198907
    }
  },
  {
    "ph": "s", "id": 198907, "pid": 76337, "tid": -914061504, "ts": 1716454224514403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224572935, "dur": 4,
    "args": {
      "External id": 198909,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198909,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 198909, "pid": 5, "tid": 7, "ts": 1716454224572935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514420, "dur": 6,
    "args": {
      "External id": 198909,
      "cbid": 211, "correlation": 198909
    }
  },
  {
    "ph": "s", "id": 198909, "pid": 76337, "tid": -914061504, "ts": 1716454224514420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514430, "dur": 0,
    "args": {
      "External id": 198910,
      "cbid": 51, "correlation": 198910
    }
  },
  {
    "ph": "s", "id": 198910, "pid": 76337, "tid": -914061504, "ts": 1716454224514430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224572941, "dur": 175,
    "args": {
      "External id": 198911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198911,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 198911, "pid": 5, "tid": 7, "ts": 1716454224572941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514430, "dur": 6,
    "args": {
      "External id": 198911,
      "cbid": 211, "correlation": 198911
    }
  },
  {
    "ph": "s", "id": 198911, "pid": 76337, "tid": -914061504, "ts": 1716454224514430,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224573117, "dur": 16,
    "args": {
      "External id": 198916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198916,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198916, "pid": 5, "tid": 7, "ts": 1716454224573117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514457, "dur": 8,
    "args": {
      "External id": 198916,
      "cbid": 211, "correlation": 198916
    }
  },
  {
    "ph": "s", "id": 198916, "pid": 76337, "tid": -914061504, "ts": 1716454224514457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224573135, "dur": 11,
    "args": {
      "External id": 198924,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198924,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198924, "pid": 5, "tid": 7, "ts": 1716454224573135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514486, "dur": 8,
    "args": {
      "External id": 198924,
      "cbid": 211, "correlation": 198924
    }
  },
  {
    "ph": "s", "id": 198924, "pid": 76337, "tid": -914061504, "ts": 1716454224514486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224573147, "dur": 10,
    "args": {
      "External id": 198932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198932,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198932, "pid": 5, "tid": 7, "ts": 1716454224573147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514514, "dur": 8,
    "args": {
      "External id": 198932,
      "cbid": 211, "correlation": 198932
    }
  },
  {
    "ph": "s", "id": 198932, "pid": 76337, "tid": -914061504, "ts": 1716454224514514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224573158, "dur": 20,
    "args": {
      "External id": 198952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198952,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 198952, "pid": 5, "tid": 7, "ts": 1716454224573158,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514597, "dur": 12,
    "args": {
      "External id": 198952,
      "cbid": 211, "correlation": 198952
    }
  },
  {
    "ph": "s", "id": 198952, "pid": 76337, "tid": -914061504, "ts": 1716454224514597,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224573179, "dur": 5,
    "args": {
      "External id": 198964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198964,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 198964, "pid": 5, "tid": 7, "ts": 1716454224573179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514619, "dur": 6,
    "args": {
      "External id": 198964,
      "cbid": 211, "correlation": 198964
    }
  },
  {
    "ph": "s", "id": 198964, "pid": 76337, "tid": -914061504, "ts": 1716454224514619,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224573186, "dur": 17,
    "args": {
      "External id": 198967,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198967,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198967, "pid": 5, "tid": 7, "ts": 1716454224573186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514636, "dur": 7,
    "args": {
      "External id": 198967,
      "cbid": 211, "correlation": 198967
    }
  },
  {
    "ph": "s", "id": 198967, "pid": 76337, "tid": -914061504, "ts": 1716454224514636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514694, "dur": 0,
    "args": {
      "External id": 198978,
      "cbid": 317, "correlation": 198978
    }
  },
  {
    "ph": "f", "id": 198978, "pid": 76337, "tid": -914061504, "ts": 1716454224514694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514695, "dur": 0,
    "args": {
      "External id": 198979,
      "cbid": 203, "correlation": 198979
    }
  },
  {
    "ph": "f", "id": 198979, "pid": 76337, "tid": -914061504, "ts": 1716454224514695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514696, "dur": 0,
    "args": {
      "External id": 198980,
      "cbid": 205, "correlation": 198980
    }
  },
  {
    "ph": "f", "id": 198980, "pid": 76337, "tid": -914061504, "ts": 1716454224514696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224573204, "dur": 11,
    "args": {
      "External id": 198984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198984,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198984, "pid": 5, "tid": 7, "ts": 1716454224573204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514709, "dur": 12,
    "args": {
      "External id": 198984,
      "cbid": 211, "correlation": 198984
    }
  },
  {
    "ph": "s", "id": 198984, "pid": 76337, "tid": -914061504, "ts": 1716454224514709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224573216, "dur": 3,
    "args": {
      "External id": 198986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198986,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 198986, "pid": 5, "tid": 7, "ts": 1716454224573216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514725, "dur": 5,
    "args": {
      "External id": 198986,
      "cbid": 211, "correlation": 198986
    }
  },
  {
    "ph": "s", "id": 198986, "pid": 76337, "tid": -914061504, "ts": 1716454224514725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514733, "dur": 0,
    "args": {
      "External id": 198987,
      "cbid": 51, "correlation": 198987
    }
  },
  {
    "ph": "s", "id": 198987, "pid": 76337, "tid": -914061504, "ts": 1716454224514733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224573221, "dur": 93,
    "args": {
      "External id": 198988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198988,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 198988, "pid": 5, "tid": 7, "ts": 1716454224573221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514734, "dur": 5,
    "args": {
      "External id": 198988,
      "cbid": 211, "correlation": 198988
    }
  },
  {
    "ph": "s", "id": 198988, "pid": 76337, "tid": -914061504, "ts": 1716454224514734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224573315, "dur": 15,
    "args": {
      "External id": 198993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 198993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 198993, "pid": 5, "tid": 7, "ts": 1716454224573315,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514761, "dur": 9,
    "args": {
      "External id": 198993,
      "cbid": 211, "correlation": 198993
    }
  },
  {
    "ph": "s", "id": 198993, "pid": 76337, "tid": -914061504, "ts": 1716454224514761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224573332, "dur": 83,
    "args": {
      "External id": 199002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199002,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199002, "pid": 5, "tid": 7, "ts": 1716454224573332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514843, "dur": 15,
    "args": {
      "External id": 199002,
      "cbid": 211, "correlation": 199002
    }
  },
  {
    "ph": "s", "id": 199002, "pid": 76337, "tid": -914061504, "ts": 1716454224514843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224573417, "dur": 30,
    "args": {
      "External id": 199024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199024,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199024, "pid": 5, "tid": 7, "ts": 1716454224573417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224514905, "dur": 11,
    "args": {
      "External id": 199024,
      "cbid": 211, "correlation": 199024
    }
  },
  {
    "ph": "s", "id": 199024, "pid": 76337, "tid": -914061504, "ts": 1716454224514905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515005, "dur": 2,
    "args": {
      "External id": 199035,
      "cbid": 251, "correlation": 199035
    }
  },
  {
    "ph": "f", "id": 199035, "pid": 76337, "tid": -914061504, "ts": 1716454224515005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224573448, "dur": 163,
    "args": {
      "External id": 199036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199036,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199036, "pid": 5, "tid": 7, "ts": 1716454224573448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515011, "dur": 13,
    "args": {
      "External id": 199036,
      "cbid": 211, "correlation": 199036
    }
  },
  {
    "ph": "s", "id": 199036, "pid": 76337, "tid": -914061504, "ts": 1716454224515011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515083, "dur": 1,
    "args": {
      "External id": 199047,
      "cbid": 251, "correlation": 199047
    }
  },
  {
    "ph": "f", "id": 199047, "pid": 76337, "tid": -914061504, "ts": 1716454224515083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224573612, "dur": 157,
    "args": {
      "External id": 199048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199048,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199048, "pid": 5, "tid": 7, "ts": 1716454224573612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515087, "dur": 12,
    "args": {
      "External id": 199048,
      "cbid": 211, "correlation": 199048
    }
  },
  {
    "ph": "s", "id": 199048, "pid": 76337, "tid": -914061504, "ts": 1716454224515087,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515153, "dur": 1,
    "args": {
      "External id": 199059,
      "cbid": 251, "correlation": 199059
    }
  },
  {
    "ph": "f", "id": 199059, "pid": 76337, "tid": -914061504, "ts": 1716454224515153,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224573770, "dur": 136,
    "args": {
      "External id": 199060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199060,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199060, "pid": 5, "tid": 7, "ts": 1716454224573770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515157, "dur": 11,
    "args": {
      "External id": 199060,
      "cbid": 211, "correlation": 199060
    }
  },
  {
    "ph": "s", "id": 199060, "pid": 76337, "tid": -914061504, "ts": 1716454224515157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224573908, "dur": 334,
    "args": {
      "External id": 199085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199085,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199085, "pid": 5, "tid": 7, "ts": 1716454224573908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515244, "dur": 14,
    "args": {
      "External id": 199085,
      "cbid": 211, "correlation": 199085
    }
  },
  {
    "ph": "s", "id": 199085, "pid": 76337, "tid": -914061504, "ts": 1716454224515244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515346, "dur": 1,
    "args": {
      "External id": 199103,
      "cbid": 251, "correlation": 199103
    }
  },
  {
    "ph": "f", "id": 199103, "pid": 76337, "tid": -914061504, "ts": 1716454224515346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224574243, "dur": 167,
    "args": {
      "External id": 199105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199105,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199105, "pid": 5, "tid": 7, "ts": 1716454224574243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515352, "dur": 13,
    "args": {
      "External id": 199105,
      "cbid": 211, "correlation": 199105
    }
  },
  {
    "ph": "s", "id": 199105, "pid": 76337, "tid": -914061504, "ts": 1716454224515352,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224574412, "dur": 19,
    "args": {
      "External id": 199113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199113,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199113, "pid": 5, "tid": 7, "ts": 1716454224574412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515422, "dur": 12,
    "args": {
      "External id": 199113,
      "cbid": 211, "correlation": 199113
    }
  },
  {
    "ph": "s", "id": 199113, "pid": 76337, "tid": -914061504, "ts": 1716454224515422,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224574432, "dur": 27,
    "args": {
      "External id": 199121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199121,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199121, "pid": 5, "tid": 7, "ts": 1716454224574432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515461, "dur": 8,
    "args": {
      "External id": 199121,
      "cbid": 211, "correlation": 199121
    }
  },
  {
    "ph": "s", "id": 199121, "pid": 76337, "tid": -914061504, "ts": 1716454224515461,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224574461, "dur": 19,
    "args": {
      "External id": 199132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199132, "pid": 5, "tid": 7, "ts": 1716454224574461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515534, "dur": 12,
    "args": {
      "External id": 199132,
      "cbid": 211, "correlation": 199132
    }
  },
  {
    "ph": "s", "id": 199132, "pid": 76337, "tid": -914061504, "ts": 1716454224515534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224574481, "dur": 16,
    "args": {
      "External id": 199154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199154,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199154, "pid": 5, "tid": 7, "ts": 1716454224574481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515566, "dur": 8,
    "args": {
      "External id": 199154,
      "cbid": 211, "correlation": 199154
    }
  },
  {
    "ph": "s", "id": 199154, "pid": 76337, "tid": -914061504, "ts": 1716454224515566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515650, "dur": 2,
    "args": {
      "External id": 199165,
      "cbid": 251, "correlation": 199165
    }
  },
  {
    "ph": "f", "id": 199165, "pid": 76337, "tid": -914061504, "ts": 1716454224515650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224574499, "dur": 89,
    "args": {
      "External id": 199166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199166,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199166, "pid": 5, "tid": 7, "ts": 1716454224574499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515656, "dur": 13,
    "args": {
      "External id": 199166,
      "cbid": 211, "correlation": 199166
    }
  },
  {
    "ph": "s", "id": 199166, "pid": 76337, "tid": -914061504, "ts": 1716454224515656,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515726, "dur": 1,
    "args": {
      "External id": 199177,
      "cbid": 251, "correlation": 199177
    }
  },
  {
    "ph": "f", "id": 199177, "pid": 76337, "tid": -914061504, "ts": 1716454224515726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515730, "dur": 0,
    "args": {
      "External id": 199178,
      "cbid": 251, "correlation": 199178
    }
  },
  {
    "ph": "f", "id": 199178, "pid": 76337, "tid": -914061504, "ts": 1716454224515730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224574589, "dur": 13,
    "args": {
      "External id": 199179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199179,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199179, "pid": 5, "tid": 7, "ts": 1716454224574589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515732, "dur": 12,
    "args": {
      "External id": 199179,
      "cbid": 211, "correlation": 199179
    }
  },
  {
    "ph": "s", "id": 199179, "pid": 76337, "tid": -914061504, "ts": 1716454224515732,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224574603, "dur": 6,
    "args": {
      "External id": 199181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199181,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199181, "pid": 5, "tid": 7, "ts": 1716454224574603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515745, "dur": 6,
    "args": {
      "External id": 199181,
      "cbid": 211, "correlation": 199181
    }
  },
  {
    "ph": "s", "id": 199181, "pid": 76337, "tid": -914061504, "ts": 1716454224515745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515802, "dur": 1,
    "args": {
      "External id": 199192,
      "cbid": 251, "correlation": 199192
    }
  },
  {
    "ph": "f", "id": 199192, "pid": 76337, "tid": -914061504, "ts": 1716454224515802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515805, "dur": 0,
    "args": {
      "External id": 199193,
      "cbid": 251, "correlation": 199193
    }
  },
  {
    "ph": "f", "id": 199193, "pid": 76337, "tid": -914061504, "ts": 1716454224515805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224574610, "dur": 9,
    "args": {
      "External id": 199194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199194,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199194, "pid": 5, "tid": 7, "ts": 1716454224574610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515807, "dur": 12,
    "args": {
      "External id": 199194,
      "cbid": 211, "correlation": 199194
    }
  },
  {
    "ph": "s", "id": 199194, "pid": 76337, "tid": -914061504, "ts": 1716454224515807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224574620, "dur": 4,
    "args": {
      "External id": 199196,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199196,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199196, "pid": 5, "tid": 7, "ts": 1716454224574620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515820, "dur": 5,
    "args": {
      "External id": 199196,
      "cbid": 211, "correlation": 199196
    }
  },
  {
    "ph": "s", "id": 199196, "pid": 76337, "tid": -914061504, "ts": 1716454224515820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224574625, "dur": 54,
    "args": {
      "External id": 199221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199221,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199221, "pid": 5, "tid": 7, "ts": 1716454224574625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224515897, "dur": 13,
    "args": {
      "External id": 199221,
      "cbid": 211, "correlation": 199221
    }
  },
  {
    "ph": "s", "id": 199221, "pid": 76337, "tid": -914061504, "ts": 1716454224515897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516003, "dur": 1,
    "args": {
      "External id": 199239,
      "cbid": 251, "correlation": 199239
    }
  },
  {
    "ph": "f", "id": 199239, "pid": 76337, "tid": -914061504, "ts": 1716454224516003,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224574681, "dur": 92,
    "args": {
      "External id": 199241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199241,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199241, "pid": 5, "tid": 7, "ts": 1716454224574681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516010, "dur": 15,
    "args": {
      "External id": 199241,
      "cbid": 211, "correlation": 199241
    }
  },
  {
    "ph": "s", "id": 199241, "pid": 76337, "tid": -914061504, "ts": 1716454224516010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224574774, "dur": 9,
    "args": {
      "External id": 199249,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199249,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199249, "pid": 5, "tid": 7, "ts": 1716454224574774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516080, "dur": 12,
    "args": {
      "External id": 199249,
      "cbid": 211, "correlation": 199249
    }
  },
  {
    "ph": "s", "id": 199249, "pid": 76337, "tid": -914061504, "ts": 1716454224516080,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224574784, "dur": 21,
    "args": {
      "External id": 199257,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199257,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199257, "pid": 5, "tid": 7, "ts": 1716454224574784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516122, "dur": 9,
    "args": {
      "External id": 199257,
      "cbid": 211, "correlation": 199257
    }
  },
  {
    "ph": "s", "id": 199257, "pid": 76337, "tid": -914061504, "ts": 1716454224516122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224574807, "dur": 18,
    "args": {
      "External id": 199279,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199279,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199279, "pid": 5, "tid": 7, "ts": 1716454224574807,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516174, "dur": 10,
    "args": {
      "External id": 199279,
      "cbid": 211, "correlation": 199279
    }
  },
  {
    "ph": "s", "id": 199279, "pid": 76337, "tid": -914061504, "ts": 1716454224516174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516261, "dur": 1,
    "args": {
      "External id": 199295,
      "cbid": 251, "correlation": 199295
    }
  },
  {
    "ph": "f", "id": 199295, "pid": 76337, "tid": -914061504, "ts": 1716454224516261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516266, "dur": 0,
    "args": {
      "External id": 199297,
      "cbid": 251, "correlation": 199297
    }
  },
  {
    "ph": "f", "id": 199297, "pid": 76337, "tid": -914061504, "ts": 1716454224516266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224574826, "dur": 499,
    "args": {
      "External id": 199298,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199298,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199298, "pid": 5, "tid": 7, "ts": 1716454224574826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516268, "dur": 13,
    "args": {
      "External id": 199298,
      "cbid": 211, "correlation": 199298
    }
  },
  {
    "ph": "s", "id": 199298, "pid": 76337, "tid": -914061504, "ts": 1716454224516268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224575326, "dur": 67,
    "args": {
      "External id": 199306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199306,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199306, "pid": 5, "tid": 7, "ts": 1716454224575326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516334, "dur": 12,
    "args": {
      "External id": 199306,
      "cbid": 211, "correlation": 199306
    }
  },
  {
    "ph": "s", "id": 199306, "pid": 76337, "tid": -914061504, "ts": 1716454224516334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224575394, "dur": 67,
    "args": {
      "External id": 199314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199314,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199314, "pid": 5, "tid": 7, "ts": 1716454224575394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516366, "dur": 9,
    "args": {
      "External id": 199314,
      "cbid": 211, "correlation": 199314
    }
  },
  {
    "ph": "s", "id": 199314, "pid": 76337, "tid": -914061504, "ts": 1716454224516366,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516446, "dur": 1,
    "args": {
      "External id": 199330,
      "cbid": 251, "correlation": 199330
    }
  },
  {
    "ph": "f", "id": 199330, "pid": 76337, "tid": -914061504, "ts": 1716454224516446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224575463, "dur": 1,
    "args": {
      "External id": 199332,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 199332,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 199332, "pid": 5, "tid": 7, "ts": 1716454224575463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516451, "dur": 9,
    "args": {
      "External id": 199332,
      "cbid": 51, "correlation": 199332
    }
  },
  {
    "ph": "s", "id": 199332, "pid": 76337, "tid": -914061504, "ts": 1716454224516451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224575467, "dur": 273,
    "args": {
      "External id": 199333,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199333,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199333, "pid": 5, "tid": 7, "ts": 1716454224575467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516462, "dur": 12,
    "args": {
      "External id": 199333,
      "cbid": 211, "correlation": 199333
    }
  },
  {
    "ph": "s", "id": 199333, "pid": 76337, "tid": -914061504, "ts": 1716454224516462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224575741, "dur": 14,
    "args": {
      "External id": 199341,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199341,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199341, "pid": 5, "tid": 7, "ts": 1716454224575741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516504, "dur": 10,
    "args": {
      "External id": 199341,
      "cbid": 211, "correlation": 199341
    }
  },
  {
    "ph": "s", "id": 199341, "pid": 76337, "tid": -914061504, "ts": 1716454224516504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224575757, "dur": 38,
    "args": {
      "External id": 199352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199352, "pid": 5, "tid": 7, "ts": 1716454224575757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516574, "dur": 12,
    "args": {
      "External id": 199352,
      "cbid": 211, "correlation": 199352
    }
  },
  {
    "ph": "s", "id": 199352, "pid": 76337, "tid": -914061504, "ts": 1716454224516574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516639, "dur": 0,
    "args": {
      "External id": 199364,
      "cbid": 317, "correlation": 199364
    }
  },
  {
    "ph": "f", "id": 199364, "pid": 76337, "tid": -914061504, "ts": 1716454224516639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516640, "dur": 0,
    "args": {
      "External id": 199365,
      "cbid": 203, "correlation": 199365
    }
  },
  {
    "ph": "f", "id": 199365, "pid": 76337, "tid": -914061504, "ts": 1716454224516640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516641, "dur": 0,
    "args": {
      "External id": 199366,
      "cbid": 205, "correlation": 199366
    }
  },
  {
    "ph": "f", "id": 199366, "pid": 76337, "tid": -914061504, "ts": 1716454224516641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224575796, "dur": 14,
    "args": {
      "External id": 199370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199370,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199370, "pid": 5, "tid": 7, "ts": 1716454224575796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516657, "dur": 13,
    "args": {
      "External id": 199370,
      "cbid": 211, "correlation": 199370
    }
  },
  {
    "ph": "s", "id": 199370, "pid": 76337, "tid": -914061504, "ts": 1716454224516657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224575811, "dur": 4,
    "args": {
      "External id": 199372,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199372,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199372, "pid": 5, "tid": 7, "ts": 1716454224575811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516674, "dur": 6,
    "args": {
      "External id": 199372,
      "cbid": 211, "correlation": 199372
    }
  },
  {
    "ph": "s", "id": 199372, "pid": 76337, "tid": -914061504, "ts": 1716454224516674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516683, "dur": 0,
    "args": {
      "External id": 199373,
      "cbid": 51, "correlation": 199373
    }
  },
  {
    "ph": "s", "id": 199373, "pid": 76337, "tid": -914061504, "ts": 1716454224516683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224575816, "dur": 97,
    "args": {
      "External id": 199374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199374,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 199374, "pid": 5, "tid": 7, "ts": 1716454224575816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516684, "dur": 5,
    "args": {
      "External id": 199374,
      "cbid": 211, "correlation": 199374
    }
  },
  {
    "ph": "s", "id": 199374, "pid": 76337, "tid": -914061504, "ts": 1716454224516684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224575914, "dur": 16,
    "args": {
      "External id": 199379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199379,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199379, "pid": 5, "tid": 7, "ts": 1716454224575914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516710, "dur": 9,
    "args": {
      "External id": 199379,
      "cbid": 211, "correlation": 199379
    }
  },
  {
    "ph": "s", "id": 199379, "pid": 76337, "tid": -914061504, "ts": 1716454224516710,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224575932, "dur": 11,
    "args": {
      "External id": 199387,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199387,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199387, "pid": 5, "tid": 7, "ts": 1716454224575932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516742, "dur": 8,
    "args": {
      "External id": 199387,
      "cbid": 211, "correlation": 199387
    }
  },
  {
    "ph": "s", "id": 199387, "pid": 76337, "tid": -914061504, "ts": 1716454224516742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224575945, "dur": 30,
    "args": {
      "External id": 199396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199396,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199396, "pid": 5, "tid": 7, "ts": 1716454224575945,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516781, "dur": 10,
    "args": {
      "External id": 199396,
      "cbid": 211, "correlation": 199396
    }
  },
  {
    "ph": "s", "id": 199396, "pid": 76337, "tid": -914061504, "ts": 1716454224516781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224575976, "dur": 31,
    "args": {
      "External id": 199416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199416,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 199416, "pid": 5, "tid": 7, "ts": 1716454224575976,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516853, "dur": 11,
    "args": {
      "External id": 199416,
      "cbid": 211, "correlation": 199416
    }
  },
  {
    "ph": "s", "id": 199416, "pid": 76337, "tid": -914061504, "ts": 1716454224516853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224576008, "dur": 5,
    "args": {
      "External id": 199428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199428,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199428, "pid": 5, "tid": 7, "ts": 1716454224576008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516880, "dur": 7,
    "args": {
      "External id": 199428,
      "cbid": 211, "correlation": 199428
    }
  },
  {
    "ph": "s", "id": 199428, "pid": 76337, "tid": -914061504, "ts": 1716454224516880,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224576015, "dur": 31,
    "args": {
      "External id": 199431,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199431,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199431, "pid": 5, "tid": 7, "ts": 1716454224576015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516899, "dur": 7,
    "args": {
      "External id": 199431,
      "cbid": 211, "correlation": 199431
    }
  },
  {
    "ph": "s", "id": 199431, "pid": 76337, "tid": -914061504, "ts": 1716454224516899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224576048, "dur": 21,
    "args": {
      "External id": 199440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199440,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199440, "pid": 5, "tid": 7, "ts": 1716454224576048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516939, "dur": 10,
    "args": {
      "External id": 199440,
      "cbid": 211, "correlation": 199440
    }
  },
  {
    "ph": "s", "id": 199440, "pid": 76337, "tid": -914061504, "ts": 1716454224516939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516999, "dur": 0,
    "args": {
      "External id": 199450,
      "cbid": 317, "correlation": 199450
    }
  },
  {
    "ph": "f", "id": 199450, "pid": 76337, "tid": -914061504, "ts": 1716454224516999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224516999, "dur": 0,
    "args": {
      "External id": 199451,
      "cbid": 203, "correlation": 199451
    }
  },
  {
    "ph": "f", "id": 199451, "pid": 76337, "tid": -914061504, "ts": 1716454224516999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517000, "dur": 0,
    "args": {
      "External id": 199452,
      "cbid": 205, "correlation": 199452
    }
  },
  {
    "ph": "f", "id": 199452, "pid": 76337, "tid": -914061504, "ts": 1716454224517000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224576070, "dur": 21,
    "args": {
      "External id": 199456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199456, "pid": 5, "tid": 7, "ts": 1716454224576070,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517016, "dur": 12,
    "args": {
      "External id": 199456,
      "cbid": 211, "correlation": 199456
    }
  },
  {
    "ph": "s", "id": 199456, "pid": 76337, "tid": -914061504, "ts": 1716454224517016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224576093, "dur": 319,
    "args": {
      "External id": 199458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199458,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199458, "pid": 5, "tid": 7, "ts": 1716454224576093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517031, "dur": 5,
    "args": {
      "External id": 199458,
      "cbid": 211, "correlation": 199458
    }
  },
  {
    "ph": "s", "id": 199458, "pid": 76337, "tid": -914061504, "ts": 1716454224517031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224576414, "dur": 1,
    "args": {
      "External id": 199460,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 199460,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 199460, "pid": 5, "tid": 7, "ts": 1716454224576414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517042, "dur": 6,
    "args": {
      "External id": 199460,
      "cbid": 51, "correlation": 199460
    }
  },
  {
    "ph": "s", "id": 199460, "pid": 76337, "tid": -914061504, "ts": 1716454224517042,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224576418, "dur": 1257,
    "args": {
      "External id": 199461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199461,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199461, "pid": 5, "tid": 7, "ts": 1716454224576418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517050, "dur": 6,
    "args": {
      "External id": 199461,
      "cbid": 211, "correlation": 199461
    }
  },
  {
    "ph": "s", "id": 199461, "pid": 76337, "tid": -914061504, "ts": 1716454224517050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224577676, "dur": 12,
    "args": {
      "External id": 199463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199463,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199463, "pid": 5, "tid": 7, "ts": 1716454224577676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517060, "dur": 6,
    "args": {
      "External id": 199463,
      "cbid": 211, "correlation": 199463
    }
  },
  {
    "ph": "s", "id": 199463, "pid": 76337, "tid": -914061504, "ts": 1716454224517060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224577690, "dur": 15,
    "args": {
      "External id": 199469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199469,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199469, "pid": 5, "tid": 7, "ts": 1716454224577690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517092, "dur": 9,
    "args": {
      "External id": 199469,
      "cbid": 211, "correlation": 199469
    }
  },
  {
    "ph": "s", "id": 199469, "pid": 76337, "tid": -914061504, "ts": 1716454224517092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224577706, "dur": 3,
    "args": {
      "External id": 199477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199477,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 199477, "pid": 5, "tid": 7, "ts": 1716454224577706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517137, "dur": 10,
    "args": {
      "External id": 199477,
      "cbid": 211, "correlation": 199477
    }
  },
  {
    "ph": "s", "id": 199477, "pid": 76337, "tid": -914061504, "ts": 1716454224517137,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517202, "dur": 1,
    "args": {
      "External id": 199493,
      "cbid": 251, "correlation": 199493
    }
  },
  {
    "ph": "f", "id": 199493, "pid": 76337, "tid": -914061504, "ts": 1716454224517202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517207, "dur": 0,
    "args": {
      "External id": 199495,
      "cbid": 251, "correlation": 199495
    }
  },
  {
    "ph": "f", "id": 199495, "pid": 76337, "tid": -914061504, "ts": 1716454224517207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224577711, "dur": 13,
    "args": {
      "External id": 199496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199496,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199496, "pid": 5, "tid": 7, "ts": 1716454224577711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517209, "dur": 12,
    "args": {
      "External id": 199496,
      "cbid": 211, "correlation": 199496
    }
  },
  {
    "ph": "s", "id": 199496, "pid": 76337, "tid": -914061504, "ts": 1716454224517209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224577725, "dur": 5,
    "args": {
      "External id": 199498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199498,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199498, "pid": 5, "tid": 7, "ts": 1716454224577725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517222, "dur": 6,
    "args": {
      "External id": 199498,
      "cbid": 211, "correlation": 199498
    }
  },
  {
    "ph": "s", "id": 199498, "pid": 76337, "tid": -914061504, "ts": 1716454224517222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224577732, "dur": 17,
    "args": {
      "External id": 199508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199508, "pid": 5, "tid": 7, "ts": 1716454224577732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517281, "dur": 12,
    "args": {
      "External id": 199508,
      "cbid": 211, "correlation": 199508
    }
  },
  {
    "ph": "s", "id": 199508, "pid": 76337, "tid": -914061504, "ts": 1716454224517281,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224577750, "dur": 17,
    "args": {
      "External id": 199528,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199528,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 199528, "pid": 5, "tid": 7, "ts": 1716454224577750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517347, "dur": 10,
    "args": {
      "External id": 199528,
      "cbid": 211, "correlation": 199528
    }
  },
  {
    "ph": "s", "id": 199528, "pid": 76337, "tid": -914061504, "ts": 1716454224517347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224577769, "dur": 4,
    "args": {
      "External id": 199540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199540,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 199540, "pid": 5, "tid": 7, "ts": 1716454224577769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517367, "dur": 6,
    "args": {
      "External id": 199540,
      "cbid": 211, "correlation": 199540
    }
  },
  {
    "ph": "s", "id": 199540, "pid": 76337, "tid": -914061504, "ts": 1716454224517367,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224577774, "dur": 17,
    "args": {
      "External id": 199543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199543,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199543, "pid": 5, "tid": 7, "ts": 1716454224577774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517387, "dur": 6,
    "args": {
      "External id": 199543,
      "cbid": 211, "correlation": 199543
    }
  },
  {
    "ph": "s", "id": 199543, "pid": 76337, "tid": -914061504, "ts": 1716454224517387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224577792, "dur": 11,
    "args": {
      "External id": 199552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199552,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199552, "pid": 5, "tid": 7, "ts": 1716454224577792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517426, "dur": 10,
    "args": {
      "External id": 199552,
      "cbid": 211, "correlation": 199552
    }
  },
  {
    "ph": "s", "id": 199552, "pid": 76337, "tid": -914061504, "ts": 1716454224517426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517489, "dur": 0,
    "args": {
      "External id": 199562,
      "cbid": 317, "correlation": 199562
    }
  },
  {
    "ph": "f", "id": 199562, "pid": 76337, "tid": -914061504, "ts": 1716454224517489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517490, "dur": 0,
    "args": {
      "External id": 199563,
      "cbid": 203, "correlation": 199563
    }
  },
  {
    "ph": "f", "id": 199563, "pid": 76337, "tid": -914061504, "ts": 1716454224517490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517490, "dur": 0,
    "args": {
      "External id": 199564,
      "cbid": 205, "correlation": 199564
    }
  },
  {
    "ph": "f", "id": 199564, "pid": 76337, "tid": -914061504, "ts": 1716454224517490,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224577804, "dur": 12,
    "args": {
      "External id": 199568,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199568,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199568, "pid": 5, "tid": 7, "ts": 1716454224577804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517505, "dur": 12,
    "args": {
      "External id": 199568,
      "cbid": 211, "correlation": 199568
    }
  },
  {
    "ph": "s", "id": 199568, "pid": 76337, "tid": -914061504, "ts": 1716454224517505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224577817, "dur": 162,
    "args": {
      "External id": 199570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199570,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199570, "pid": 5, "tid": 7, "ts": 1716454224577817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517518, "dur": 5,
    "args": {
      "External id": 199570,
      "cbid": 211, "correlation": 199570
    }
  },
  {
    "ph": "s", "id": 199570, "pid": 76337, "tid": -914061504, "ts": 1716454224517518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224577981, "dur": 1,
    "args": {
      "External id": 199572,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 199572,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 199572, "pid": 5, "tid": 7, "ts": 1716454224577981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517530, "dur": 6,
    "args": {
      "External id": 199572,
      "cbid": 51, "correlation": 199572
    }
  },
  {
    "ph": "s", "id": 199572, "pid": 76337, "tid": -914061504, "ts": 1716454224517530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224577985, "dur": 650,
    "args": {
      "External id": 199573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199573,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199573, "pid": 5, "tid": 7, "ts": 1716454224577985,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517537, "dur": 6,
    "args": {
      "External id": 199573,
      "cbid": 211, "correlation": 199573
    }
  },
  {
    "ph": "s", "id": 199573, "pid": 76337, "tid": -914061504, "ts": 1716454224517537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224578636, "dur": 12,
    "args": {
      "External id": 199575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199575,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199575, "pid": 5, "tid": 7, "ts": 1716454224578636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517548, "dur": 5,
    "args": {
      "External id": 199575,
      "cbid": 211, "correlation": 199575
    }
  },
  {
    "ph": "s", "id": 199575, "pid": 76337, "tid": -914061504, "ts": 1716454224517548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224578649, "dur": 15,
    "args": {
      "External id": 199581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199581,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199581, "pid": 5, "tid": 7, "ts": 1716454224578649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517576, "dur": 9,
    "args": {
      "External id": 199581,
      "cbid": 211, "correlation": 199581
    }
  },
  {
    "ph": "s", "id": 199581, "pid": 76337, "tid": -914061504, "ts": 1716454224517576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517634, "dur": 0,
    "args": {
      "External id": 199591,
      "cbid": 317, "correlation": 199591
    }
  },
  {
    "ph": "f", "id": 199591, "pid": 76337, "tid": -914061504, "ts": 1716454224517634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517635, "dur": 0,
    "args": {
      "External id": 199592,
      "cbid": 203, "correlation": 199592
    }
  },
  {
    "ph": "f", "id": 199592, "pid": 76337, "tid": -914061504, "ts": 1716454224517635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517635, "dur": 0,
    "args": {
      "External id": 199593,
      "cbid": 205, "correlation": 199593
    }
  },
  {
    "ph": "f", "id": 199593, "pid": 76337, "tid": -914061504, "ts": 1716454224517635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224578665, "dur": 22,
    "args": {
      "External id": 199597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199597,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199597, "pid": 5, "tid": 7, "ts": 1716454224578665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517647, "dur": 11,
    "args": {
      "External id": 199597,
      "cbid": 211, "correlation": 199597
    }
  },
  {
    "ph": "s", "id": 199597, "pid": 76337, "tid": -914061504, "ts": 1716454224517647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224578689, "dur": 4,
    "args": {
      "External id": 199599,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199599,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199599, "pid": 5, "tid": 7, "ts": 1716454224578689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517662, "dur": 6,
    "args": {
      "External id": 199599,
      "cbid": 211, "correlation": 199599
    }
  },
  {
    "ph": "s", "id": 199599, "pid": 76337, "tid": -914061504, "ts": 1716454224517662,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517670, "dur": 0,
    "args": {
      "External id": 199600,
      "cbid": 51, "correlation": 199600
    }
  },
  {
    "ph": "s", "id": 199600, "pid": 76337, "tid": -914061504, "ts": 1716454224517670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224578694, "dur": 173,
    "args": {
      "External id": 199601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199601,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 199601, "pid": 5, "tid": 7, "ts": 1716454224578694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517671, "dur": 5,
    "args": {
      "External id": 199601,
      "cbid": 211, "correlation": 199601
    }
  },
  {
    "ph": "s", "id": 199601, "pid": 76337, "tid": -914061504, "ts": 1716454224517671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224578869, "dur": 16,
    "args": {
      "External id": 199606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199606,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199606, "pid": 5, "tid": 7, "ts": 1716454224578869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517697, "dur": 8,
    "args": {
      "External id": 199606,
      "cbid": 211, "correlation": 199606
    }
  },
  {
    "ph": "s", "id": 199606, "pid": 76337, "tid": -914061504, "ts": 1716454224517697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224578886, "dur": 12,
    "args": {
      "External id": 199614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199614,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199614, "pid": 5, "tid": 7, "ts": 1716454224578886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517725, "dur": 8,
    "args": {
      "External id": 199614,
      "cbid": 211, "correlation": 199614
    }
  },
  {
    "ph": "s", "id": 199614, "pid": 76337, "tid": -914061504, "ts": 1716454224517725,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224578900, "dur": 10,
    "args": {
      "External id": 199622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199622,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199622, "pid": 5, "tid": 7, "ts": 1716454224578900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517754, "dur": 8,
    "args": {
      "External id": 199622,
      "cbid": 211, "correlation": 199622
    }
  },
  {
    "ph": "s", "id": 199622, "pid": 76337, "tid": -914061504, "ts": 1716454224517754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224578911, "dur": 18,
    "args": {
      "External id": 199642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199642,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 199642, "pid": 5, "tid": 7, "ts": 1716454224578911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517835, "dur": 13,
    "args": {
      "External id": 199642,
      "cbid": 211, "correlation": 199642
    }
  },
  {
    "ph": "s", "id": 199642, "pid": 76337, "tid": -914061504, "ts": 1716454224517835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224578930, "dur": 5,
    "args": {
      "External id": 199654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199654,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 199654, "pid": 5, "tid": 7, "ts": 1716454224578930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517858, "dur": 6,
    "args": {
      "External id": 199654,
      "cbid": 211, "correlation": 199654
    }
  },
  {
    "ph": "s", "id": 199654, "pid": 76337, "tid": -914061504, "ts": 1716454224517858,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224578936, "dur": 16,
    "args": {
      "External id": 199657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199657,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199657, "pid": 5, "tid": 7, "ts": 1716454224578936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517876, "dur": 6,
    "args": {
      "External id": 199657,
      "cbid": 211, "correlation": 199657
    }
  },
  {
    "ph": "s", "id": 199657, "pid": 76337, "tid": -914061504, "ts": 1716454224517876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517933, "dur": 0,
    "args": {
      "External id": 199668,
      "cbid": 317, "correlation": 199668
    }
  },
  {
    "ph": "f", "id": 199668, "pid": 76337, "tid": -914061504, "ts": 1716454224517933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517934, "dur": 0,
    "args": {
      "External id": 199669,
      "cbid": 203, "correlation": 199669
    }
  },
  {
    "ph": "f", "id": 199669, "pid": 76337, "tid": -914061504, "ts": 1716454224517934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517935, "dur": 0,
    "args": {
      "External id": 199670,
      "cbid": 205, "correlation": 199670
    }
  },
  {
    "ph": "f", "id": 199670, "pid": 76337, "tid": -914061504, "ts": 1716454224517935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224578954, "dur": 11,
    "args": {
      "External id": 199674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199674, "pid": 5, "tid": 7, "ts": 1716454224578954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517949, "dur": 11,
    "args": {
      "External id": 199674,
      "cbid": 211, "correlation": 199674
    }
  },
  {
    "ph": "s", "id": 199674, "pid": 76337, "tid": -914061504, "ts": 1716454224517949,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224578966, "dur": 3,
    "args": {
      "External id": 199676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199676,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199676, "pid": 5, "tid": 7, "ts": 1716454224578966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517964, "dur": 6,
    "args": {
      "External id": 199676,
      "cbid": 211, "correlation": 199676
    }
  },
  {
    "ph": "s", "id": 199676, "pid": 76337, "tid": -914061504, "ts": 1716454224517964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517972, "dur": 7,
    "args": {
      "External id": 199677,
      "cbid": 51, "correlation": 199677
    }
  },
  {
    "ph": "s", "id": 199677, "pid": 76337, "tid": -914061504, "ts": 1716454224517972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224578971, "dur": 91,
    "args": {
      "External id": 199678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199678,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 199678, "pid": 5, "tid": 7, "ts": 1716454224578971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224517980, "dur": 6,
    "args": {
      "External id": 199678,
      "cbid": 211, "correlation": 199678
    }
  },
  {
    "ph": "s", "id": 199678, "pid": 76337, "tid": -914061504, "ts": 1716454224517980,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224579063, "dur": 16,
    "args": {
      "External id": 199683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199683, "pid": 5, "tid": 7, "ts": 1716454224579063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518009, "dur": 8,
    "args": {
      "External id": 199683,
      "cbid": 211, "correlation": 199683
    }
  },
  {
    "ph": "s", "id": 199683, "pid": 76337, "tid": -914061504, "ts": 1716454224518009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224579081, "dur": 82,
    "args": {
      "External id": 199692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199692,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199692, "pid": 5, "tid": 7, "ts": 1716454224579081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518090, "dur": 15,
    "args": {
      "External id": 199692,
      "cbid": 211, "correlation": 199692
    }
  },
  {
    "ph": "s", "id": 199692, "pid": 76337, "tid": -914061504, "ts": 1716454224518090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224579164, "dur": 30,
    "args": {
      "External id": 199714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199714,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199714, "pid": 5, "tid": 7, "ts": 1716454224579164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518151, "dur": 10,
    "args": {
      "External id": 199714,
      "cbid": 211, "correlation": 199714
    }
  },
  {
    "ph": "s", "id": 199714, "pid": 76337, "tid": -914061504, "ts": 1716454224518151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518240, "dur": 1,
    "args": {
      "External id": 199725,
      "cbid": 251, "correlation": 199725
    }
  },
  {
    "ph": "f", "id": 199725, "pid": 76337, "tid": -914061504, "ts": 1716454224518240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224579196, "dur": 164,
    "args": {
      "External id": 199726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199726,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199726, "pid": 5, "tid": 7, "ts": 1716454224579196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518246, "dur": 13,
    "args": {
      "External id": 199726,
      "cbid": 211, "correlation": 199726
    }
  },
  {
    "ph": "s", "id": 199726, "pid": 76337, "tid": -914061504, "ts": 1716454224518246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518316, "dur": 1,
    "args": {
      "External id": 199737,
      "cbid": 251, "correlation": 199737
    }
  },
  {
    "ph": "f", "id": 199737, "pid": 76337, "tid": -914061504, "ts": 1716454224518316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224579361, "dur": 157,
    "args": {
      "External id": 199738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199738,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199738, "pid": 5, "tid": 7, "ts": 1716454224579361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518320, "dur": 11,
    "args": {
      "External id": 199738,
      "cbid": 211, "correlation": 199738
    }
  },
  {
    "ph": "s", "id": 199738, "pid": 76337, "tid": -914061504, "ts": 1716454224518320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518385, "dur": 1,
    "args": {
      "External id": 199749,
      "cbid": 251, "correlation": 199749
    }
  },
  {
    "ph": "f", "id": 199749, "pid": 76337, "tid": -914061504, "ts": 1716454224518385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224579520, "dur": 159,
    "args": {
      "External id": 199750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199750,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199750, "pid": 5, "tid": 7, "ts": 1716454224579520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518389, "dur": 12,
    "args": {
      "External id": 199750,
      "cbid": 211, "correlation": 199750
    }
  },
  {
    "ph": "s", "id": 199750, "pid": 76337, "tid": -914061504, "ts": 1716454224518389,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224579680, "dur": 334,
    "args": {
      "External id": 199775,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199775,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199775, "pid": 5, "tid": 7, "ts": 1716454224579680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518473, "dur": 13,
    "args": {
      "External id": 199775,
      "cbid": 211, "correlation": 199775
    }
  },
  {
    "ph": "s", "id": 199775, "pid": 76337, "tid": -914061504, "ts": 1716454224518473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518573, "dur": 1,
    "args": {
      "External id": 199793,
      "cbid": 251, "correlation": 199793
    }
  },
  {
    "ph": "f", "id": 199793, "pid": 76337, "tid": -914061504, "ts": 1716454224518573,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224580016, "dur": 167,
    "args": {
      "External id": 199795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199795,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199795, "pid": 5, "tid": 7, "ts": 1716454224580016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518578, "dur": 13,
    "args": {
      "External id": 199795,
      "cbid": 211, "correlation": 199795
    }
  },
  {
    "ph": "s", "id": 199795, "pid": 76337, "tid": -914061504, "ts": 1716454224518578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224580184, "dur": 19,
    "args": {
      "External id": 199803,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199803,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199803, "pid": 5, "tid": 7, "ts": 1716454224580184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518647, "dur": 13,
    "args": {
      "External id": 199803,
      "cbid": 211, "correlation": 199803
    }
  },
  {
    "ph": "s", "id": 199803, "pid": 76337, "tid": -914061504, "ts": 1716454224518647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224580205, "dur": 27,
    "args": {
      "External id": 199811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199811,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199811, "pid": 5, "tid": 7, "ts": 1716454224580205,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518686, "dur": 8,
    "args": {
      "External id": 199811,
      "cbid": 211, "correlation": 199811
    }
  },
  {
    "ph": "s", "id": 199811, "pid": 76337, "tid": -914061504, "ts": 1716454224518686,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224580234, "dur": 18,
    "args": {
      "External id": 199822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199822, "pid": 5, "tid": 7, "ts": 1716454224580234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518758, "dur": 12,
    "args": {
      "External id": 199822,
      "cbid": 211, "correlation": 199822
    }
  },
  {
    "ph": "s", "id": 199822, "pid": 76337, "tid": -914061504, "ts": 1716454224518758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224580253, "dur": 16,
    "args": {
      "External id": 199844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199844,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199844, "pid": 5, "tid": 7, "ts": 1716454224580253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518788, "dur": 8,
    "args": {
      "External id": 199844,
      "cbid": 211, "correlation": 199844
    }
  },
  {
    "ph": "s", "id": 199844, "pid": 76337, "tid": -914061504, "ts": 1716454224518788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518873, "dur": 1,
    "args": {
      "External id": 199855,
      "cbid": 251, "correlation": 199855
    }
  },
  {
    "ph": "f", "id": 199855, "pid": 76337, "tid": -914061504, "ts": 1716454224518873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224580271, "dur": 89,
    "args": {
      "External id": 199856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199856,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199856, "pid": 5, "tid": 7, "ts": 1716454224580271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518879, "dur": 13,
    "args": {
      "External id": 199856,
      "cbid": 211, "correlation": 199856
    }
  },
  {
    "ph": "s", "id": 199856, "pid": 76337, "tid": -914061504, "ts": 1716454224518879,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518947, "dur": 1,
    "args": {
      "External id": 199867,
      "cbid": 251, "correlation": 199867
    }
  },
  {
    "ph": "f", "id": 199867, "pid": 76337, "tid": -914061504, "ts": 1716454224518947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518950, "dur": 0,
    "args": {
      "External id": 199868,
      "cbid": 251, "correlation": 199868
    }
  },
  {
    "ph": "f", "id": 199868, "pid": 76337, "tid": -914061504, "ts": 1716454224518950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224580362, "dur": 13,
    "args": {
      "External id": 199869,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199869,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199869, "pid": 5, "tid": 7, "ts": 1716454224580362,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518952, "dur": 12,
    "args": {
      "External id": 199869,
      "cbid": 211, "correlation": 199869
    }
  },
  {
    "ph": "s", "id": 199869, "pid": 76337, "tid": -914061504, "ts": 1716454224518952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224580376, "dur": 6,
    "args": {
      "External id": 199871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199871,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199871, "pid": 5, "tid": 7, "ts": 1716454224580376,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224518966, "dur": 6,
    "args": {
      "External id": 199871,
      "cbid": 211, "correlation": 199871
    }
  },
  {
    "ph": "s", "id": 199871, "pid": 76337, "tid": -914061504, "ts": 1716454224518966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519034, "dur": 1,
    "args": {
      "External id": 199882,
      "cbid": 251, "correlation": 199882
    }
  },
  {
    "ph": "f", "id": 199882, "pid": 76337, "tid": -914061504, "ts": 1716454224519034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519038, "dur": 0,
    "args": {
      "External id": 199883,
      "cbid": 251, "correlation": 199883
    }
  },
  {
    "ph": "f", "id": 199883, "pid": 76337, "tid": -914061504, "ts": 1716454224519038,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224580382, "dur": 8,
    "args": {
      "External id": 199884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199884,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199884, "pid": 5, "tid": 7, "ts": 1716454224580382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519039, "dur": 12,
    "args": {
      "External id": 199884,
      "cbid": 211, "correlation": 199884
    }
  },
  {
    "ph": "s", "id": 199884, "pid": 76337, "tid": -914061504, "ts": 1716454224519039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224580392, "dur": 4,
    "args": {
      "External id": 199886,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199886,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199886, "pid": 5, "tid": 7, "ts": 1716454224580392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519053, "dur": 6,
    "args": {
      "External id": 199886,
      "cbid": 211, "correlation": 199886
    }
  },
  {
    "ph": "s", "id": 199886, "pid": 76337, "tid": -914061504, "ts": 1716454224519053,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224580397, "dur": 55,
    "args": {
      "External id": 199911,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199911,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199911, "pid": 5, "tid": 7, "ts": 1716454224580397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519130, "dur": 12,
    "args": {
      "External id": 199911,
      "cbid": 211, "correlation": 199911
    }
  },
  {
    "ph": "s", "id": 199911, "pid": 76337, "tid": -914061504, "ts": 1716454224519130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519228, "dur": 1,
    "args": {
      "External id": 199929,
      "cbid": 251, "correlation": 199929
    }
  },
  {
    "ph": "f", "id": 199929, "pid": 76337, "tid": -914061504, "ts": 1716454224519228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224580453, "dur": 91,
    "args": {
      "External id": 199931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199931,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 199931, "pid": 5, "tid": 7, "ts": 1716454224580453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519234, "dur": 14,
    "args": {
      "External id": 199931,
      "cbid": 211, "correlation": 199931
    }
  },
  {
    "ph": "s", "id": 199931, "pid": 76337, "tid": -914061504, "ts": 1716454224519234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224580546, "dur": 10,
    "args": {
      "External id": 199939,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199939,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199939, "pid": 5, "tid": 7, "ts": 1716454224580546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519303, "dur": 13,
    "args": {
      "External id": 199939,
      "cbid": 211, "correlation": 199939
    }
  },
  {
    "ph": "s", "id": 199939, "pid": 76337, "tid": -914061504, "ts": 1716454224519303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224580557, "dur": 21,
    "args": {
      "External id": 199947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199947,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199947, "pid": 5, "tid": 7, "ts": 1716454224580557,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519344, "dur": 9,
    "args": {
      "External id": 199947,
      "cbid": 211, "correlation": 199947
    }
  },
  {
    "ph": "s", "id": 199947, "pid": 76337, "tid": -914061504, "ts": 1716454224519344,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224580580, "dur": 18,
    "args": {
      "External id": 199969,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199969,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199969, "pid": 5, "tid": 7, "ts": 1716454224580580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519397, "dur": 10,
    "args": {
      "External id": 199969,
      "cbid": 211, "correlation": 199969
    }
  },
  {
    "ph": "s", "id": 199969, "pid": 76337, "tid": -914061504, "ts": 1716454224519397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519484, "dur": 1,
    "args": {
      "External id": 199985,
      "cbid": 251, "correlation": 199985
    }
  },
  {
    "ph": "f", "id": 199985, "pid": 76337, "tid": -914061504, "ts": 1716454224519484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519489, "dur": 0,
    "args": {
      "External id": 199987,
      "cbid": 251, "correlation": 199987
    }
  },
  {
    "ph": "f", "id": 199987, "pid": 76337, "tid": -914061504, "ts": 1716454224519489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224580599, "dur": 498,
    "args": {
      "External id": 199988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199988,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 199988, "pid": 5, "tid": 7, "ts": 1716454224580599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519491, "dur": 12,
    "args": {
      "External id": 199988,
      "cbid": 211, "correlation": 199988
    }
  },
  {
    "ph": "s", "id": 199988, "pid": 76337, "tid": -914061504, "ts": 1716454224519491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224581098, "dur": 67,
    "args": {
      "External id": 199996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 199996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 199996, "pid": 5, "tid": 7, "ts": 1716454224581098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519555, "dur": 13,
    "args": {
      "External id": 199996,
      "cbid": 211, "correlation": 199996
    }
  },
  {
    "ph": "s", "id": 199996, "pid": 76337, "tid": -914061504, "ts": 1716454224519555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224581166, "dur": 68,
    "args": {
      "External id": 200004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200004,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200004, "pid": 5, "tid": 7, "ts": 1716454224581166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519587, "dur": 8,
    "args": {
      "External id": 200004,
      "cbid": 211, "correlation": 200004
    }
  },
  {
    "ph": "s", "id": 200004, "pid": 76337, "tid": -914061504, "ts": 1716454224519587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519668, "dur": 1,
    "args": {
      "External id": 200020,
      "cbid": 251, "correlation": 200020
    }
  },
  {
    "ph": "f", "id": 200020, "pid": 76337, "tid": -914061504, "ts": 1716454224519668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224581236, "dur": 1,
    "args": {
      "External id": 200022,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 200022,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 200022, "pid": 5, "tid": 7, "ts": 1716454224581236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519673, "dur": 9,
    "args": {
      "External id": 200022,
      "cbid": 51, "correlation": 200022
    }
  },
  {
    "ph": "s", "id": 200022, "pid": 76337, "tid": -914061504, "ts": 1716454224519673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224581240, "dur": 270,
    "args": {
      "External id": 200023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200023,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200023, "pid": 5, "tid": 7, "ts": 1716454224581240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519684, "dur": 11,
    "args": {
      "External id": 200023,
      "cbid": 211, "correlation": 200023
    }
  },
  {
    "ph": "s", "id": 200023, "pid": 76337, "tid": -914061504, "ts": 1716454224519684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224581511, "dur": 14,
    "args": {
      "External id": 200031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200031,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200031, "pid": 5, "tid": 7, "ts": 1716454224581511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519727, "dur": 10,
    "args": {
      "External id": 200031,
      "cbid": 211, "correlation": 200031
    }
  },
  {
    "ph": "s", "id": 200031, "pid": 76337, "tid": -914061504, "ts": 1716454224519727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224581526, "dur": 37,
    "args": {
      "External id": 200042,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200042,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200042, "pid": 5, "tid": 7, "ts": 1716454224581526,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519794, "dur": 12,
    "args": {
      "External id": 200042,
      "cbid": 211, "correlation": 200042
    }
  },
  {
    "ph": "s", "id": 200042, "pid": 76337, "tid": -914061504, "ts": 1716454224519794,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519858, "dur": 0,
    "args": {
      "External id": 200054,
      "cbid": 317, "correlation": 200054
    }
  },
  {
    "ph": "f", "id": 200054, "pid": 76337, "tid": -914061504, "ts": 1716454224519858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519859, "dur": 0,
    "args": {
      "External id": 200055,
      "cbid": 203, "correlation": 200055
    }
  },
  {
    "ph": "f", "id": 200055, "pid": 76337, "tid": -914061504, "ts": 1716454224519859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519859, "dur": 0,
    "args": {
      "External id": 200056,
      "cbid": 205, "correlation": 200056
    }
  },
  {
    "ph": "f", "id": 200056, "pid": 76337, "tid": -914061504, "ts": 1716454224519859,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224581565, "dur": 14,
    "args": {
      "External id": 200060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200060,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200060, "pid": 5, "tid": 7, "ts": 1716454224581565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519875, "dur": 12,
    "args": {
      "External id": 200060,
      "cbid": 211, "correlation": 200060
    }
  },
  {
    "ph": "s", "id": 200060, "pid": 76337, "tid": -914061504, "ts": 1716454224519875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224581580, "dur": 4,
    "args": {
      "External id": 200062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200062, "pid": 5, "tid": 7, "ts": 1716454224581580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519892, "dur": 6,
    "args": {
      "External id": 200062,
      "cbid": 211, "correlation": 200062
    }
  },
  {
    "ph": "s", "id": 200062, "pid": 76337, "tid": -914061504, "ts": 1716454224519892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519900, "dur": 0,
    "args": {
      "External id": 200063,
      "cbid": 51, "correlation": 200063
    }
  },
  {
    "ph": "s", "id": 200063, "pid": 76337, "tid": -914061504, "ts": 1716454224519900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224581586, "dur": 97,
    "args": {
      "External id": 200064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200064,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 200064, "pid": 5, "tid": 7, "ts": 1716454224581586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519901, "dur": 5,
    "args": {
      "External id": 200064,
      "cbid": 211, "correlation": 200064
    }
  },
  {
    "ph": "s", "id": 200064, "pid": 76337, "tid": -914061504, "ts": 1716454224519901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224581684, "dur": 17,
    "args": {
      "External id": 200069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200069,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200069, "pid": 5, "tid": 7, "ts": 1716454224581684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519928, "dur": 8,
    "args": {
      "External id": 200069,
      "cbid": 211, "correlation": 200069
    }
  },
  {
    "ph": "s", "id": 200069, "pid": 76337, "tid": -914061504, "ts": 1716454224519928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224581702, "dur": 13,
    "args": {
      "External id": 200077,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200077,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200077, "pid": 5, "tid": 7, "ts": 1716454224581702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224519959, "dur": 9,
    "args": {
      "External id": 200077,
      "cbid": 211, "correlation": 200077
    }
  },
  {
    "ph": "s", "id": 200077, "pid": 76337, "tid": -914061504, "ts": 1716454224519959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224581716, "dur": 24,
    "args": {
      "External id": 200086,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200086,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200086, "pid": 5, "tid": 7, "ts": 1716454224581716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520007, "dur": 12,
    "args": {
      "External id": 200086,
      "cbid": 211, "correlation": 200086
    }
  },
  {
    "ph": "s", "id": 200086, "pid": 76337, "tid": -914061504, "ts": 1716454224520007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224581742, "dur": 24,
    "args": {
      "External id": 200106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200106,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 200106, "pid": 5, "tid": 7, "ts": 1716454224581742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520082, "dur": 12,
    "args": {
      "External id": 200106,
      "cbid": 211, "correlation": 200106
    }
  },
  {
    "ph": "s", "id": 200106, "pid": 76337, "tid": -914061504, "ts": 1716454224520082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224581767, "dur": 5,
    "args": {
      "External id": 200118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200118,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 200118, "pid": 5, "tid": 7, "ts": 1716454224581767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520104, "dur": 6,
    "args": {
      "External id": 200118,
      "cbid": 211, "correlation": 200118
    }
  },
  {
    "ph": "s", "id": 200118, "pid": 76337, "tid": -914061504, "ts": 1716454224520104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224581773, "dur": 26,
    "args": {
      "External id": 200121,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200121,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200121, "pid": 5, "tid": 7, "ts": 1716454224581773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520123, "dur": 7,
    "args": {
      "External id": 200121,
      "cbid": 211, "correlation": 200121
    }
  },
  {
    "ph": "s", "id": 200121, "pid": 76337, "tid": -914061504, "ts": 1716454224520123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224581800, "dur": 17,
    "args": {
      "External id": 200130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200130,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200130, "pid": 5, "tid": 7, "ts": 1716454224581800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520162, "dur": 10,
    "args": {
      "External id": 200130,
      "cbid": 211, "correlation": 200130
    }
  },
  {
    "ph": "s", "id": 200130, "pid": 76337, "tid": -914061504, "ts": 1716454224520162,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520214, "dur": 0,
    "args": {
      "External id": 200140,
      "cbid": 317, "correlation": 200140
    }
  },
  {
    "ph": "f", "id": 200140, "pid": 76337, "tid": -914061504, "ts": 1716454224520214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520215, "dur": 0,
    "args": {
      "External id": 200141,
      "cbid": 203, "correlation": 200141
    }
  },
  {
    "ph": "f", "id": 200141, "pid": 76337, "tid": -914061504, "ts": 1716454224520215,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520216, "dur": 0,
    "args": {
      "External id": 200142,
      "cbid": 205, "correlation": 200142
    }
  },
  {
    "ph": "f", "id": 200142, "pid": 76337, "tid": -914061504, "ts": 1716454224520216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224581818, "dur": 18,
    "args": {
      "External id": 200146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200146,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200146, "pid": 5, "tid": 7, "ts": 1716454224581818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520231, "dur": 11,
    "args": {
      "External id": 200146,
      "cbid": 211, "correlation": 200146
    }
  },
  {
    "ph": "s", "id": 200146, "pid": 76337, "tid": -914061504, "ts": 1716454224520231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224581837, "dur": 240,
    "args": {
      "External id": 200148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200148,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200148, "pid": 5, "tid": 7, "ts": 1716454224581837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520245, "dur": 5,
    "args": {
      "External id": 200148,
      "cbid": 211, "correlation": 200148
    }
  },
  {
    "ph": "s", "id": 200148, "pid": 76337, "tid": -914061504, "ts": 1716454224520245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224582080, "dur": 1,
    "args": {
      "External id": 200150,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 200150,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 200150, "pid": 5, "tid": 7, "ts": 1716454224582080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520256, "dur": 9,
    "args": {
      "External id": 200150,
      "cbid": 51, "correlation": 200150
    }
  },
  {
    "ph": "s", "id": 200150, "pid": 76337, "tid": -914061504, "ts": 1716454224520256,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224582084, "dur": 815,
    "args": {
      "External id": 200151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200151,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200151, "pid": 5, "tid": 7, "ts": 1716454224582084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520266, "dur": 6,
    "args": {
      "External id": 200151,
      "cbid": 211, "correlation": 200151
    }
  },
  {
    "ph": "s", "id": 200151, "pid": 76337, "tid": -914061504, "ts": 1716454224520266,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224582900, "dur": 14,
    "args": {
      "External id": 200153,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200153,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200153, "pid": 5, "tid": 7, "ts": 1716454224582900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520277, "dur": 5,
    "args": {
      "External id": 200153,
      "cbid": 211, "correlation": 200153
    }
  },
  {
    "ph": "s", "id": 200153, "pid": 76337, "tid": -914061504, "ts": 1716454224520277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224582915, "dur": 15,
    "args": {
      "External id": 200159,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200159,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200159, "pid": 5, "tid": 7, "ts": 1716454224582915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520304, "dur": 8,
    "args": {
      "External id": 200159,
      "cbid": 211, "correlation": 200159
    }
  },
  {
    "ph": "s", "id": 200159, "pid": 76337, "tid": -914061504, "ts": 1716454224520304,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224582931, "dur": 4,
    "args": {
      "External id": 200167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200167,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 200167, "pid": 5, "tid": 7, "ts": 1716454224582931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520348, "dur": 9,
    "args": {
      "External id": 200167,
      "cbid": 211, "correlation": 200167
    }
  },
  {
    "ph": "s", "id": 200167, "pid": 76337, "tid": -914061504, "ts": 1716454224520348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520417, "dur": 1,
    "args": {
      "External id": 200183,
      "cbid": 251, "correlation": 200183
    }
  },
  {
    "ph": "f", "id": 200183, "pid": 76337, "tid": -914061504, "ts": 1716454224520417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520422, "dur": 0,
    "args": {
      "External id": 200185,
      "cbid": 251, "correlation": 200185
    }
  },
  {
    "ph": "f", "id": 200185, "pid": 76337, "tid": -914061504, "ts": 1716454224520422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224582936, "dur": 13,
    "args": {
      "External id": 200186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200186,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200186, "pid": 5, "tid": 7, "ts": 1716454224582936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520424, "dur": 11,
    "args": {
      "External id": 200186,
      "cbid": 211, "correlation": 200186
    }
  },
  {
    "ph": "s", "id": 200186, "pid": 76337, "tid": -914061504, "ts": 1716454224520424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224582951, "dur": 5,
    "args": {
      "External id": 200188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200188,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200188, "pid": 5, "tid": 7, "ts": 1716454224582951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520437, "dur": 5,
    "args": {
      "External id": 200188,
      "cbid": 211, "correlation": 200188
    }
  },
  {
    "ph": "s", "id": 200188, "pid": 76337, "tid": -914061504, "ts": 1716454224520437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224582957, "dur": 17,
    "args": {
      "External id": 200198,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200198,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200198, "pid": 5, "tid": 7, "ts": 1716454224582957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520495, "dur": 12,
    "args": {
      "External id": 200198,
      "cbid": 211, "correlation": 200198
    }
  },
  {
    "ph": "s", "id": 200198, "pid": 76337, "tid": -914061504, "ts": 1716454224520495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224582975, "dur": 17,
    "args": {
      "External id": 200218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200218,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 200218, "pid": 5, "tid": 7, "ts": 1716454224582975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520561, "dur": 11,
    "args": {
      "External id": 200218,
      "cbid": 211, "correlation": 200218
    }
  },
  {
    "ph": "s", "id": 200218, "pid": 76337, "tid": -914061504, "ts": 1716454224520561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224582994, "dur": 4,
    "args": {
      "External id": 200230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200230,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 200230, "pid": 5, "tid": 7, "ts": 1716454224582994,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520582, "dur": 6,
    "args": {
      "External id": 200230,
      "cbid": 211, "correlation": 200230
    }
  },
  {
    "ph": "s", "id": 200230, "pid": 76337, "tid": -914061504, "ts": 1716454224520582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224582999, "dur": 17,
    "args": {
      "External id": 200233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200233,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200233, "pid": 5, "tid": 7, "ts": 1716454224582999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520601, "dur": 7,
    "args": {
      "External id": 200233,
      "cbid": 211, "correlation": 200233
    }
  },
  {
    "ph": "s", "id": 200233, "pid": 76337, "tid": -914061504, "ts": 1716454224520601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224583017, "dur": 11,
    "args": {
      "External id": 200242,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200242,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200242, "pid": 5, "tid": 7, "ts": 1716454224583017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520642, "dur": 9,
    "args": {
      "External id": 200242,
      "cbid": 211, "correlation": 200242
    }
  },
  {
    "ph": "s", "id": 200242, "pid": 76337, "tid": -914061504, "ts": 1716454224520642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520704, "dur": 0,
    "args": {
      "External id": 200252,
      "cbid": 317, "correlation": 200252
    }
  },
  {
    "ph": "f", "id": 200252, "pid": 76337, "tid": -914061504, "ts": 1716454224520704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520705, "dur": 0,
    "args": {
      "External id": 200253,
      "cbid": 203, "correlation": 200253
    }
  },
  {
    "ph": "f", "id": 200253, "pid": 76337, "tid": -914061504, "ts": 1716454224520705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520705, "dur": 0,
    "args": {
      "External id": 200254,
      "cbid": 205, "correlation": 200254
    }
  },
  {
    "ph": "f", "id": 200254, "pid": 76337, "tid": -914061504, "ts": 1716454224520705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224583028, "dur": 11,
    "args": {
      "External id": 200258,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200258,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200258, "pid": 5, "tid": 7, "ts": 1716454224583028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520719, "dur": 12,
    "args": {
      "External id": 200258,
      "cbid": 211, "correlation": 200258
    }
  },
  {
    "ph": "s", "id": 200258, "pid": 76337, "tid": -914061504, "ts": 1716454224520719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224583041, "dur": 163,
    "args": {
      "External id": 200260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200260,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200260, "pid": 5, "tid": 7, "ts": 1716454224583041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520734, "dur": 5,
    "args": {
      "External id": 200260,
      "cbid": 211, "correlation": 200260
    }
  },
  {
    "ph": "s", "id": 200260, "pid": 76337, "tid": -914061504, "ts": 1716454224520734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224583206, "dur": 1,
    "args": {
      "External id": 200262,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 200262,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 200262, "pid": 5, "tid": 7, "ts": 1716454224583206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520745, "dur": 6,
    "args": {
      "External id": 200262,
      "cbid": 51, "correlation": 200262
    }
  },
  {
    "ph": "s", "id": 200262, "pid": 76337, "tid": -914061504, "ts": 1716454224520745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224583210, "dur": 650,
    "args": {
      "External id": 200263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200263,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200263, "pid": 5, "tid": 7, "ts": 1716454224583210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520752, "dur": 6,
    "args": {
      "External id": 200263,
      "cbid": 211, "correlation": 200263
    }
  },
  {
    "ph": "s", "id": 200263, "pid": 76337, "tid": -914061504, "ts": 1716454224520752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224583861, "dur": 14,
    "args": {
      "External id": 200265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200265,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200265, "pid": 5, "tid": 7, "ts": 1716454224583861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520763, "dur": 5,
    "args": {
      "External id": 200265,
      "cbid": 211, "correlation": 200265
    }
  },
  {
    "ph": "s", "id": 200265, "pid": 76337, "tid": -914061504, "ts": 1716454224520763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224583876, "dur": 15,
    "args": {
      "External id": 200271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200271,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200271, "pid": 5, "tid": 7, "ts": 1716454224583876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520792, "dur": 8,
    "args": {
      "External id": 200271,
      "cbid": 211, "correlation": 200271
    }
  },
  {
    "ph": "s", "id": 200271, "pid": 76337, "tid": -914061504, "ts": 1716454224520792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520850, "dur": 0,
    "args": {
      "External id": 200281,
      "cbid": 317, "correlation": 200281
    }
  },
  {
    "ph": "f", "id": 200281, "pid": 76337, "tid": -914061504, "ts": 1716454224520850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520851, "dur": 0,
    "args": {
      "External id": 200282,
      "cbid": 203, "correlation": 200282
    }
  },
  {
    "ph": "f", "id": 200282, "pid": 76337, "tid": -914061504, "ts": 1716454224520851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520852, "dur": 0,
    "args": {
      "External id": 200283,
      "cbid": 205, "correlation": 200283
    }
  },
  {
    "ph": "f", "id": 200283, "pid": 76337, "tid": -914061504, "ts": 1716454224520852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224583892, "dur": 17,
    "args": {
      "External id": 200287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200287,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200287, "pid": 5, "tid": 7, "ts": 1716454224583892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520865, "dur": 13,
    "args": {
      "External id": 200287,
      "cbid": 211, "correlation": 200287
    }
  },
  {
    "ph": "s", "id": 200287, "pid": 76337, "tid": -914061504, "ts": 1716454224520865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224583911, "dur": 4,
    "args": {
      "External id": 200289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200289, "pid": 5, "tid": 7, "ts": 1716454224583911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520882, "dur": 5,
    "args": {
      "External id": 200289,
      "cbid": 211, "correlation": 200289
    }
  },
  {
    "ph": "s", "id": 200289, "pid": 76337, "tid": -914061504, "ts": 1716454224520882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520890, "dur": 0,
    "args": {
      "External id": 200290,
      "cbid": 51, "correlation": 200290
    }
  },
  {
    "ph": "s", "id": 200290, "pid": 76337, "tid": -914061504, "ts": 1716454224520890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224583916, "dur": 132,
    "args": {
      "External id": 200291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200291,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 200291, "pid": 5, "tid": 7, "ts": 1716454224583916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520891, "dur": 5,
    "args": {
      "External id": 200291,
      "cbid": 211, "correlation": 200291
    }
  },
  {
    "ph": "s", "id": 200291, "pid": 76337, "tid": -914061504, "ts": 1716454224520891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224584049, "dur": 16,
    "args": {
      "External id": 200296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200296,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200296, "pid": 5, "tid": 7, "ts": 1716454224584049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520916, "dur": 9,
    "args": {
      "External id": 200296,
      "cbid": 211, "correlation": 200296
    }
  },
  {
    "ph": "s", "id": 200296, "pid": 76337, "tid": -914061504, "ts": 1716454224520916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224584066, "dur": 12,
    "args": {
      "External id": 200304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200304,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200304, "pid": 5, "tid": 7, "ts": 1716454224584066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520945, "dur": 7,
    "args": {
      "External id": 200304,
      "cbid": 211, "correlation": 200304
    }
  },
  {
    "ph": "s", "id": 200304, "pid": 76337, "tid": -914061504, "ts": 1716454224520945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224584080, "dur": 10,
    "args": {
      "External id": 200312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200312,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200312, "pid": 5, "tid": 7, "ts": 1716454224584080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224520972, "dur": 16,
    "args": {
      "External id": 200312,
      "cbid": 211, "correlation": 200312
    }
  },
  {
    "ph": "s", "id": 200312, "pid": 76337, "tid": -914061504, "ts": 1716454224520972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224584092, "dur": 18,
    "args": {
      "External id": 200332,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200332,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 200332, "pid": 5, "tid": 7, "ts": 1716454224584092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521068, "dur": 14,
    "args": {
      "External id": 200332,
      "cbid": 211, "correlation": 200332
    }
  },
  {
    "ph": "s", "id": 200332, "pid": 76337, "tid": -914061504, "ts": 1716454224521068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224584111, "dur": 4,
    "args": {
      "External id": 200344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200344,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 200344, "pid": 5, "tid": 7, "ts": 1716454224584111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521092, "dur": 6,
    "args": {
      "External id": 200344,
      "cbid": 211, "correlation": 200344
    }
  },
  {
    "ph": "s", "id": 200344, "pid": 76337, "tid": -914061504, "ts": 1716454224521092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224584117, "dur": 17,
    "args": {
      "External id": 200347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200347, "pid": 5, "tid": 7, "ts": 1716454224584117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521109, "dur": 6,
    "args": {
      "External id": 200347,
      "cbid": 211, "correlation": 200347
    }
  },
  {
    "ph": "s", "id": 200347, "pid": 76337, "tid": -914061504, "ts": 1716454224521109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521167, "dur": 0,
    "args": {
      "External id": 200358,
      "cbid": 317, "correlation": 200358
    }
  },
  {
    "ph": "f", "id": 200358, "pid": 76337, "tid": -914061504, "ts": 1716454224521167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521167, "dur": 0,
    "args": {
      "External id": 200359,
      "cbid": 203, "correlation": 200359
    }
  },
  {
    "ph": "f", "id": 200359, "pid": 76337, "tid": -914061504, "ts": 1716454224521167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521168, "dur": 0,
    "args": {
      "External id": 200360,
      "cbid": 205, "correlation": 200360
    }
  },
  {
    "ph": "f", "id": 200360, "pid": 76337, "tid": -914061504, "ts": 1716454224521168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224584135, "dur": 11,
    "args": {
      "External id": 200364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200364,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200364, "pid": 5, "tid": 7, "ts": 1716454224584135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521182, "dur": 11,
    "args": {
      "External id": 200364,
      "cbid": 211, "correlation": 200364
    }
  },
  {
    "ph": "s", "id": 200364, "pid": 76337, "tid": -914061504, "ts": 1716454224521182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224584147, "dur": 3,
    "args": {
      "External id": 200366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200366,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200366, "pid": 5, "tid": 7, "ts": 1716454224584147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521197, "dur": 5,
    "args": {
      "External id": 200366,
      "cbid": 211, "correlation": 200366
    }
  },
  {
    "ph": "s", "id": 200366, "pid": 76337, "tid": -914061504, "ts": 1716454224521197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521205, "dur": 0,
    "args": {
      "External id": 200367,
      "cbid": 51, "correlation": 200367
    }
  },
  {
    "ph": "s", "id": 200367, "pid": 76337, "tid": -914061504, "ts": 1716454224521205,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224584152, "dur": 91,
    "args": {
      "External id": 200368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200368,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 200368, "pid": 5, "tid": 7, "ts": 1716454224584152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521206, "dur": 5,
    "args": {
      "External id": 200368,
      "cbid": 211, "correlation": 200368
    }
  },
  {
    "ph": "s", "id": 200368, "pid": 76337, "tid": -914061504, "ts": 1716454224521206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224584245, "dur": 15,
    "args": {
      "External id": 200373,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200373,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200373, "pid": 5, "tid": 7, "ts": 1716454224584245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521233, "dur": 8,
    "args": {
      "External id": 200373,
      "cbid": 211, "correlation": 200373
    }
  },
  {
    "ph": "s", "id": 200373, "pid": 76337, "tid": -914061504, "ts": 1716454224521233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224584262, "dur": 84,
    "args": {
      "External id": 200382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200382,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200382, "pid": 5, "tid": 7, "ts": 1716454224584262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521313, "dur": 14,
    "args": {
      "External id": 200382,
      "cbid": 211, "correlation": 200382
    }
  },
  {
    "ph": "s", "id": 200382, "pid": 76337, "tid": -914061504, "ts": 1716454224521313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224584347, "dur": 30,
    "args": {
      "External id": 200404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200404,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200404, "pid": 5, "tid": 7, "ts": 1716454224584347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521369, "dur": 11,
    "args": {
      "External id": 200404,
      "cbid": 211, "correlation": 200404
    }
  },
  {
    "ph": "s", "id": 200404, "pid": 76337, "tid": -914061504, "ts": 1716454224521369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521459, "dur": 1,
    "args": {
      "External id": 200415,
      "cbid": 251, "correlation": 200415
    }
  },
  {
    "ph": "f", "id": 200415, "pid": 76337, "tid": -914061504, "ts": 1716454224521459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224584378, "dur": 142,
    "args": {
      "External id": 200416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200416,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200416, "pid": 5, "tid": 7, "ts": 1716454224584378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521464, "dur": 13,
    "args": {
      "External id": 200416,
      "cbid": 211, "correlation": 200416
    }
  },
  {
    "ph": "s", "id": 200416, "pid": 76337, "tid": -914061504, "ts": 1716454224521464,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521535, "dur": 1,
    "args": {
      "External id": 200427,
      "cbid": 251, "correlation": 200427
    }
  },
  {
    "ph": "f", "id": 200427, "pid": 76337, "tid": -914061504, "ts": 1716454224521535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224584522, "dur": 159,
    "args": {
      "External id": 200428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200428,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200428, "pid": 5, "tid": 7, "ts": 1716454224584522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521539, "dur": 11,
    "args": {
      "External id": 200428,
      "cbid": 211, "correlation": 200428
    }
  },
  {
    "ph": "s", "id": 200428, "pid": 76337, "tid": -914061504, "ts": 1716454224521539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521602, "dur": 1,
    "args": {
      "External id": 200439,
      "cbid": 251, "correlation": 200439
    }
  },
  {
    "ph": "f", "id": 200439, "pid": 76337, "tid": -914061504, "ts": 1716454224521602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224584682, "dur": 158,
    "args": {
      "External id": 200440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200440,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200440, "pid": 5, "tid": 7, "ts": 1716454224584682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521606, "dur": 11,
    "args": {
      "External id": 200440,
      "cbid": 211, "correlation": 200440
    }
  },
  {
    "ph": "s", "id": 200440, "pid": 76337, "tid": -914061504, "ts": 1716454224521606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224584842, "dur": 336,
    "args": {
      "External id": 200465,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200465,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200465, "pid": 5, "tid": 7, "ts": 1716454224584842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521691, "dur": 12,
    "args": {
      "External id": 200465,
      "cbid": 211, "correlation": 200465
    }
  },
  {
    "ph": "s", "id": 200465, "pid": 76337, "tid": -914061504, "ts": 1716454224521691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521791, "dur": 1,
    "args": {
      "External id": 200483,
      "cbid": 251, "correlation": 200483
    }
  },
  {
    "ph": "f", "id": 200483, "pid": 76337, "tid": -914061504, "ts": 1716454224521791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224585179, "dur": 166,
    "args": {
      "External id": 200485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200485,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200485, "pid": 5, "tid": 7, "ts": 1716454224585179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521797, "dur": 14,
    "args": {
      "External id": 200485,
      "cbid": 211, "correlation": 200485
    }
  },
  {
    "ph": "s", "id": 200485, "pid": 76337, "tid": -914061504, "ts": 1716454224521797,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224585347, "dur": 19,
    "args": {
      "External id": 200493,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200493,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200493, "pid": 5, "tid": 7, "ts": 1716454224585347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521866, "dur": 12,
    "args": {
      "External id": 200493,
      "cbid": 211, "correlation": 200493
    }
  },
  {
    "ph": "s", "id": 200493, "pid": 76337, "tid": -914061504, "ts": 1716454224521866,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224585367, "dur": 27,
    "args": {
      "External id": 200501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200501,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200501, "pid": 5, "tid": 7, "ts": 1716454224585367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521905, "dur": 8,
    "args": {
      "External id": 200501,
      "cbid": 211, "correlation": 200501
    }
  },
  {
    "ph": "s", "id": 200501, "pid": 76337, "tid": -914061504, "ts": 1716454224521905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224585396, "dur": 18,
    "args": {
      "External id": 200512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200512,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200512, "pid": 5, "tid": 7, "ts": 1716454224585396,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224521984, "dur": 14,
    "args": {
      "External id": 200512,
      "cbid": 211, "correlation": 200512
    }
  },
  {
    "ph": "s", "id": 200512, "pid": 76337, "tid": -914061504, "ts": 1716454224521984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224585415, "dur": 16,
    "args": {
      "External id": 200534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200534,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200534, "pid": 5, "tid": 7, "ts": 1716454224585415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522020, "dur": 8,
    "args": {
      "External id": 200534,
      "cbid": 211, "correlation": 200534
    }
  },
  {
    "ph": "s", "id": 200534, "pid": 76337, "tid": -914061504, "ts": 1716454224522020,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522107, "dur": 1,
    "args": {
      "External id": 200545,
      "cbid": 251, "correlation": 200545
    }
  },
  {
    "ph": "f", "id": 200545, "pid": 76337, "tid": -914061504, "ts": 1716454224522107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224585433, "dur": 90,
    "args": {
      "External id": 200546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200546,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200546, "pid": 5, "tid": 7, "ts": 1716454224585433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522113, "dur": 13,
    "args": {
      "External id": 200546,
      "cbid": 211, "correlation": 200546
    }
  },
  {
    "ph": "s", "id": 200546, "pid": 76337, "tid": -914061504, "ts": 1716454224522113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522180, "dur": 1,
    "args": {
      "External id": 200557,
      "cbid": 251, "correlation": 200557
    }
  },
  {
    "ph": "f", "id": 200557, "pid": 76337, "tid": -914061504, "ts": 1716454224522180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522184, "dur": 0,
    "args": {
      "External id": 200558,
      "cbid": 251, "correlation": 200558
    }
  },
  {
    "ph": "f", "id": 200558, "pid": 76337, "tid": -914061504, "ts": 1716454224522184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224585524, "dur": 12,
    "args": {
      "External id": 200559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200559,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200559, "pid": 5, "tid": 7, "ts": 1716454224585524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522186, "dur": 12,
    "args": {
      "External id": 200559,
      "cbid": 211, "correlation": 200559
    }
  },
  {
    "ph": "s", "id": 200559, "pid": 76337, "tid": -914061504, "ts": 1716454224522186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224585537, "dur": 6,
    "args": {
      "External id": 200561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200561,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200561, "pid": 5, "tid": 7, "ts": 1716454224585537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522200, "dur": 6,
    "args": {
      "External id": 200561,
      "cbid": 211, "correlation": 200561
    }
  },
  {
    "ph": "s", "id": 200561, "pid": 76337, "tid": -914061504, "ts": 1716454224522200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522257, "dur": 1,
    "args": {
      "External id": 200572,
      "cbid": 251, "correlation": 200572
    }
  },
  {
    "ph": "f", "id": 200572, "pid": 76337, "tid": -914061504, "ts": 1716454224522257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522260, "dur": 0,
    "args": {
      "External id": 200573,
      "cbid": 251, "correlation": 200573
    }
  },
  {
    "ph": "f", "id": 200573, "pid": 76337, "tid": -914061504, "ts": 1716454224522260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224585545, "dur": 8,
    "args": {
      "External id": 200574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200574,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200574, "pid": 5, "tid": 7, "ts": 1716454224585545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522262, "dur": 11,
    "args": {
      "External id": 200574,
      "cbid": 211, "correlation": 200574
    }
  },
  {
    "ph": "s", "id": 200574, "pid": 76337, "tid": -914061504, "ts": 1716454224522262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224585554, "dur": 4,
    "args": {
      "External id": 200576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200576,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200576, "pid": 5, "tid": 7, "ts": 1716454224585554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522275, "dur": 5,
    "args": {
      "External id": 200576,
      "cbid": 211, "correlation": 200576
    }
  },
  {
    "ph": "s", "id": 200576, "pid": 76337, "tid": -914061504, "ts": 1716454224522275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224585559, "dur": 56,
    "args": {
      "External id": 200601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200601,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200601, "pid": 5, "tid": 7, "ts": 1716454224585559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522351, "dur": 12,
    "args": {
      "External id": 200601,
      "cbid": 211, "correlation": 200601
    }
  },
  {
    "ph": "s", "id": 200601, "pid": 76337, "tid": -914061504, "ts": 1716454224522351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522449, "dur": 1,
    "args": {
      "External id": 200619,
      "cbid": 251, "correlation": 200619
    }
  },
  {
    "ph": "f", "id": 200619, "pid": 76337, "tid": -914061504, "ts": 1716454224522449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224585616, "dur": 90,
    "args": {
      "External id": 200621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200621,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200621, "pid": 5, "tid": 7, "ts": 1716454224585616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522455, "dur": 13,
    "args": {
      "External id": 200621,
      "cbid": 211, "correlation": 200621
    }
  },
  {
    "ph": "s", "id": 200621, "pid": 76337, "tid": -914061504, "ts": 1716454224522455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224585708, "dur": 10,
    "args": {
      "External id": 200629,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200629,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200629, "pid": 5, "tid": 7, "ts": 1716454224585708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522525, "dur": 12,
    "args": {
      "External id": 200629,
      "cbid": 211, "correlation": 200629
    }
  },
  {
    "ph": "s", "id": 200629, "pid": 76337, "tid": -914061504, "ts": 1716454224522525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224585719, "dur": 21,
    "args": {
      "External id": 200637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200637,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200637, "pid": 5, "tid": 7, "ts": 1716454224585719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522566, "dur": 9,
    "args": {
      "External id": 200637,
      "cbid": 211, "correlation": 200637
    }
  },
  {
    "ph": "s", "id": 200637, "pid": 76337, "tid": -914061504, "ts": 1716454224522566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224585741, "dur": 18,
    "args": {
      "External id": 200659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200659,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200659, "pid": 5, "tid": 7, "ts": 1716454224585741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522618, "dur": 10,
    "args": {
      "External id": 200659,
      "cbid": 211, "correlation": 200659
    }
  },
  {
    "ph": "s", "id": 200659, "pid": 76337, "tid": -914061504, "ts": 1716454224522618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522704, "dur": 1,
    "args": {
      "External id": 200675,
      "cbid": 251, "correlation": 200675
    }
  },
  {
    "ph": "f", "id": 200675, "pid": 76337, "tid": -914061504, "ts": 1716454224522704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522709, "dur": 0,
    "args": {
      "External id": 200677,
      "cbid": 251, "correlation": 200677
    }
  },
  {
    "ph": "f", "id": 200677, "pid": 76337, "tid": -914061504, "ts": 1716454224522709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224585761, "dur": 497,
    "args": {
      "External id": 200678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200678,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200678, "pid": 5, "tid": 7, "ts": 1716454224585761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522711, "dur": 13,
    "args": {
      "External id": 200678,
      "cbid": 211, "correlation": 200678
    }
  },
  {
    "ph": "s", "id": 200678, "pid": 76337, "tid": -914061504, "ts": 1716454224522711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224586259, "dur": 66,
    "args": {
      "External id": 200686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200686,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200686, "pid": 5, "tid": 7, "ts": 1716454224586259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522776, "dur": 12,
    "args": {
      "External id": 200686,
      "cbid": 211, "correlation": 200686
    }
  },
  {
    "ph": "s", "id": 200686, "pid": 76337, "tid": -914061504, "ts": 1716454224522776,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224586326, "dur": 67,
    "args": {
      "External id": 200694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200694, "pid": 5, "tid": 7, "ts": 1716454224586326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522806, "dur": 8,
    "args": {
      "External id": 200694,
      "cbid": 211, "correlation": 200694
    }
  },
  {
    "ph": "s", "id": 200694, "pid": 76337, "tid": -914061504, "ts": 1716454224522806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522884, "dur": 1,
    "args": {
      "External id": 200710,
      "cbid": 251, "correlation": 200710
    }
  },
  {
    "ph": "f", "id": 200710, "pid": 76337, "tid": -914061504, "ts": 1716454224522884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224586395, "dur": 1,
    "args": {
      "External id": 200712,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 200712,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 200712, "pid": 5, "tid": 7, "ts": 1716454224586395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522889, "dur": 10,
    "args": {
      "External id": 200712,
      "cbid": 51, "correlation": 200712
    }
  },
  {
    "ph": "s", "id": 200712, "pid": 76337, "tid": -914061504, "ts": 1716454224522889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224586399, "dur": 271,
    "args": {
      "External id": 200713,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200713,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200713, "pid": 5, "tid": 7, "ts": 1716454224586399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522900, "dur": 11,
    "args": {
      "External id": 200713,
      "cbid": 211, "correlation": 200713
    }
  },
  {
    "ph": "s", "id": 200713, "pid": 76337, "tid": -914061504, "ts": 1716454224522900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224586672, "dur": 14,
    "args": {
      "External id": 200721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200721,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200721, "pid": 5, "tid": 7, "ts": 1716454224586672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224522942, "dur": 11,
    "args": {
      "External id": 200721,
      "cbid": 211, "correlation": 200721
    }
  },
  {
    "ph": "s", "id": 200721, "pid": 76337, "tid": -914061504, "ts": 1716454224522942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224586687, "dur": 38,
    "args": {
      "External id": 200732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200732,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200732, "pid": 5, "tid": 7, "ts": 1716454224586687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523018, "dur": 13,
    "args": {
      "External id": 200732,
      "cbid": 211, "correlation": 200732
    }
  },
  {
    "ph": "s", "id": 200732, "pid": 76337, "tid": -914061504, "ts": 1716454224523018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523082, "dur": 0,
    "args": {
      "External id": 200744,
      "cbid": 317, "correlation": 200744
    }
  },
  {
    "ph": "f", "id": 200744, "pid": 76337, "tid": -914061504, "ts": 1716454224523082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523083, "dur": 0,
    "args": {
      "External id": 200745,
      "cbid": 203, "correlation": 200745
    }
  },
  {
    "ph": "f", "id": 200745, "pid": 76337, "tid": -914061504, "ts": 1716454224523083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523084, "dur": 0,
    "args": {
      "External id": 200746,
      "cbid": 205, "correlation": 200746
    }
  },
  {
    "ph": "f", "id": 200746, "pid": 76337, "tid": -914061504, "ts": 1716454224523084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224586726, "dur": 13,
    "args": {
      "External id": 200750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200750,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200750, "pid": 5, "tid": 7, "ts": 1716454224586726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523100, "dur": 12,
    "args": {
      "External id": 200750,
      "cbid": 211, "correlation": 200750
    }
  },
  {
    "ph": "s", "id": 200750, "pid": 76337, "tid": -914061504, "ts": 1716454224523100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224586740, "dur": 4,
    "args": {
      "External id": 200752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200752,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 200752, "pid": 5, "tid": 7, "ts": 1716454224586740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523117, "dur": 6,
    "args": {
      "External id": 200752,
      "cbid": 211, "correlation": 200752
    }
  },
  {
    "ph": "s", "id": 200752, "pid": 76337, "tid": -914061504, "ts": 1716454224523117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523126, "dur": 0,
    "args": {
      "External id": 200753,
      "cbid": 51, "correlation": 200753
    }
  },
  {
    "ph": "s", "id": 200753, "pid": 76337, "tid": -914061504, "ts": 1716454224523126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224586745, "dur": 98,
    "args": {
      "External id": 200754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200754,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 200754, "pid": 5, "tid": 7, "ts": 1716454224586745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523126, "dur": 5,
    "args": {
      "External id": 200754,
      "cbid": 211, "correlation": 200754
    }
  },
  {
    "ph": "s", "id": 200754, "pid": 76337, "tid": -914061504, "ts": 1716454224523126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224586844, "dur": 16,
    "args": {
      "External id": 200759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200759,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200759, "pid": 5, "tid": 7, "ts": 1716454224586844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523154, "dur": 8,
    "args": {
      "External id": 200759,
      "cbid": 211, "correlation": 200759
    }
  },
  {
    "ph": "s", "id": 200759, "pid": 76337, "tid": -914061504, "ts": 1716454224523154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224586862, "dur": 11,
    "args": {
      "External id": 200767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200767,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200767, "pid": 5, "tid": 7, "ts": 1716454224586862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523185, "dur": 8,
    "args": {
      "External id": 200767,
      "cbid": 211, "correlation": 200767
    }
  },
  {
    "ph": "s", "id": 200767, "pid": 76337, "tid": -914061504, "ts": 1716454224523185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224586874, "dur": 58,
    "args": {
      "External id": 200778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200778,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200778, "pid": 5, "tid": 7, "ts": 1716454224586874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523247, "dur": 12,
    "args": {
      "External id": 200778,
      "cbid": 211, "correlation": 200778
    }
  },
  {
    "ph": "s", "id": 200778, "pid": 76337, "tid": -914061504, "ts": 1716454224523247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523303, "dur": 0,
    "args": {
      "External id": 200788,
      "cbid": 317, "correlation": 200788
    }
  },
  {
    "ph": "f", "id": 200788, "pid": 76337, "tid": -914061504, "ts": 1716454224523303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523304, "dur": 0,
    "args": {
      "External id": 200789,
      "cbid": 203, "correlation": 200789
    }
  },
  {
    "ph": "f", "id": 200789, "pid": 76337, "tid": -914061504, "ts": 1716454224523304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523304, "dur": 0,
    "args": {
      "External id": 200790,
      "cbid": 205, "correlation": 200790
    }
  },
  {
    "ph": "f", "id": 200790, "pid": 76337, "tid": -914061504, "ts": 1716454224523304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224586933, "dur": 38,
    "args": {
      "External id": 200794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200794,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200794, "pid": 5, "tid": 7, "ts": 1716454224586933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523320, "dur": 11,
    "args": {
      "External id": 200794,
      "cbid": 211, "correlation": 200794
    }
  },
  {
    "ph": "s", "id": 200794, "pid": 76337, "tid": -914061504, "ts": 1716454224523320,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224586972, "dur": 163,
    "args": {
      "External id": 200796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200796, "pid": 5, "tid": 7, "ts": 1716454224586972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523334, "dur": 5,
    "args": {
      "External id": 200796,
      "cbid": 211, "correlation": 200796
    }
  },
  {
    "ph": "s", "id": 200796, "pid": 76337, "tid": -914061504, "ts": 1716454224523334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224587137, "dur": 1970,
    "args": {
      "External id": 200798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200798,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200798, "pid": 5, "tid": 7, "ts": 1716454224587137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523346, "dur": 8,
    "args": {
      "External id": 200798,
      "cbid": 211, "correlation": 200798
    }
  },
  {
    "ph": "s", "id": 200798, "pid": 76337, "tid": -914061504, "ts": 1716454224523346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224589107, "dur": 39,
    "args": {
      "External id": 200800,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200800,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200800, "pid": 5, "tid": 7, "ts": 1716454224589107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523358, "dur": 5,
    "args": {
      "External id": 200800,
      "cbid": 211, "correlation": 200800
    }
  },
  {
    "ph": "s", "id": 200800, "pid": 76337, "tid": -914061504, "ts": 1716454224523358,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224589148, "dur": 58,
    "args": {
      "External id": 200806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200806,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200806, "pid": 5, "tid": 7, "ts": 1716454224589148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523385, "dur": 9,
    "args": {
      "External id": 200806,
      "cbid": 211, "correlation": 200806
    }
  },
  {
    "ph": "s", "id": 200806, "pid": 76337, "tid": -914061504, "ts": 1716454224523385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224589207, "dur": 86,
    "args": {
      "External id": 200815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200815,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200815, "pid": 5, "tid": 7, "ts": 1716454224589207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523475, "dur": 13,
    "args": {
      "External id": 200815,
      "cbid": 211, "correlation": 200815
    }
  },
  {
    "ph": "s", "id": 200815, "pid": 76337, "tid": -914061504, "ts": 1716454224523475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224589295, "dur": 73,
    "args": {
      "External id": 200835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200835,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 200835, "pid": 5, "tid": 7, "ts": 1716454224589295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523545, "dur": 12,
    "args": {
      "External id": 200835,
      "cbid": 211, "correlation": 200835
    }
  },
  {
    "ph": "s", "id": 200835, "pid": 76337, "tid": -914061504, "ts": 1716454224523545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224589369, "dur": 5,
    "args": {
      "External id": 200847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200847,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 200847, "pid": 5, "tid": 7, "ts": 1716454224589369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523567, "dur": 6,
    "args": {
      "External id": 200847,
      "cbid": 211, "correlation": 200847
    }
  },
  {
    "ph": "s", "id": 200847, "pid": 76337, "tid": -914061504, "ts": 1716454224523567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224589375, "dur": 81,
    "args": {
      "External id": 200850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200850,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200850, "pid": 5, "tid": 7, "ts": 1716454224589375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523585, "dur": 7,
    "args": {
      "External id": 200850,
      "cbid": 211, "correlation": 200850
    }
  },
  {
    "ph": "s", "id": 200850, "pid": 76337, "tid": -914061504, "ts": 1716454224523585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224589457, "dur": 54,
    "args": {
      "External id": 200859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200859,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200859, "pid": 5, "tid": 7, "ts": 1716454224589457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523629, "dur": 10,
    "args": {
      "External id": 200859,
      "cbid": 211, "correlation": 200859
    }
  },
  {
    "ph": "s", "id": 200859, "pid": 76337, "tid": -914061504, "ts": 1716454224523629,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523682, "dur": 0,
    "args": {
      "External id": 200869,
      "cbid": 317, "correlation": 200869
    }
  },
  {
    "ph": "f", "id": 200869, "pid": 76337, "tid": -914061504, "ts": 1716454224523682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523683, "dur": 0,
    "args": {
      "External id": 200870,
      "cbid": 203, "correlation": 200870
    }
  },
  {
    "ph": "f", "id": 200870, "pid": 76337, "tid": -914061504, "ts": 1716454224523683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523683, "dur": 0,
    "args": {
      "External id": 200871,
      "cbid": 205, "correlation": 200871
    }
  },
  {
    "ph": "f", "id": 200871, "pid": 76337, "tid": -914061504, "ts": 1716454224523683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224589512, "dur": 55,
    "args": {
      "External id": 200875,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200875,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200875, "pid": 5, "tid": 7, "ts": 1716454224589512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523700, "dur": 12,
    "args": {
      "External id": 200875,
      "cbid": 211, "correlation": 200875
    }
  },
  {
    "ph": "s", "id": 200875, "pid": 76337, "tid": -914061504, "ts": 1716454224523700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224589569, "dur": 123,
    "args": {
      "External id": 200877,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200877,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200877, "pid": 5, "tid": 7, "ts": 1716454224589569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523715, "dur": 5,
    "args": {
      "External id": 200877,
      "cbid": 211, "correlation": 200877
    }
  },
  {
    "ph": "s", "id": 200877, "pid": 76337, "tid": -914061504, "ts": 1716454224523715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224589693, "dur": 1895,
    "args": {
      "External id": 200879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200879,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200879, "pid": 5, "tid": 7, "ts": 1716454224589693,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523726, "dur": 6,
    "args": {
      "External id": 200879,
      "cbid": 211, "correlation": 200879
    }
  },
  {
    "ph": "s", "id": 200879, "pid": 76337, "tid": -914061504, "ts": 1716454224523726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224591589, "dur": 20,
    "args": {
      "External id": 200881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200881, "pid": 5, "tid": 7, "ts": 1716454224591589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523736, "dur": 5,
    "args": {
      "External id": 200881,
      "cbid": 211, "correlation": 200881
    }
  },
  {
    "ph": "s", "id": 200881, "pid": 76337, "tid": -914061504, "ts": 1716454224523736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224591610, "dur": 33,
    "args": {
      "External id": 200887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200887,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200887, "pid": 5, "tid": 7, "ts": 1716454224591610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523764, "dur": 8,
    "args": {
      "External id": 200887,
      "cbid": 211, "correlation": 200887
    }
  },
  {
    "ph": "s", "id": 200887, "pid": 76337, "tid": -914061504, "ts": 1716454224523764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224591645, "dur": 4,
    "args": {
      "External id": 200895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200895,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 200895, "pid": 5, "tid": 7, "ts": 1716454224591645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523808, "dur": 10,
    "args": {
      "External id": 200895,
      "cbid": 211, "correlation": 200895
    }
  },
  {
    "ph": "s", "id": 200895, "pid": 76337, "tid": -914061504, "ts": 1716454224523808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523874, "dur": 1,
    "args": {
      "External id": 200911,
      "cbid": 251, "correlation": 200911
    }
  },
  {
    "ph": "f", "id": 200911, "pid": 76337, "tid": -914061504, "ts": 1716454224523874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523879, "dur": 0,
    "args": {
      "External id": 200913,
      "cbid": 251, "correlation": 200913
    }
  },
  {
    "ph": "f", "id": 200913, "pid": 76337, "tid": -914061504, "ts": 1716454224523879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224591650, "dur": 12,
    "args": {
      "External id": 200914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200914,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 200914, "pid": 5, "tid": 7, "ts": 1716454224591650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523881, "dur": 11,
    "args": {
      "External id": 200914,
      "cbid": 211, "correlation": 200914
    }
  },
  {
    "ph": "s", "id": 200914, "pid": 76337, "tid": -914061504, "ts": 1716454224523881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224591663, "dur": 5,
    "args": {
      "External id": 200916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200916,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 200916, "pid": 5, "tid": 7, "ts": 1716454224591663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523894, "dur": 5,
    "args": {
      "External id": 200916,
      "cbid": 211, "correlation": 200916
    }
  },
  {
    "ph": "s", "id": 200916, "pid": 76337, "tid": -914061504, "ts": 1716454224523894,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224591670, "dur": 30,
    "args": {
      "External id": 200926,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200926,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200926, "pid": 5, "tid": 7, "ts": 1716454224591670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224523952, "dur": 13,
    "args": {
      "External id": 200926,
      "cbid": 211, "correlation": 200926
    }
  },
  {
    "ph": "s", "id": 200926, "pid": 76337, "tid": -914061504, "ts": 1716454224523952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224591701, "dur": 31,
    "args": {
      "External id": 200946,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200946,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 200946, "pid": 5, "tid": 7, "ts": 1716454224591701,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524027, "dur": 11,
    "args": {
      "External id": 200946,
      "cbid": 211, "correlation": 200946
    }
  },
  {
    "ph": "s", "id": 200946, "pid": 76337, "tid": -914061504, "ts": 1716454224524027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224591734, "dur": 5,
    "args": {
      "External id": 200958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200958,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 200958, "pid": 5, "tid": 7, "ts": 1716454224591734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524048, "dur": 6,
    "args": {
      "External id": 200958,
      "cbid": 211, "correlation": 200958
    }
  },
  {
    "ph": "s", "id": 200958, "pid": 76337, "tid": -914061504, "ts": 1716454224524048,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224591740, "dur": 30,
    "args": {
      "External id": 200961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200961,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200961, "pid": 5, "tid": 7, "ts": 1716454224591740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524066, "dur": 6,
    "args": {
      "External id": 200961,
      "cbid": 211, "correlation": 200961
    }
  },
  {
    "ph": "s", "id": 200961, "pid": 76337, "tid": -914061504, "ts": 1716454224524066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224591772, "dur": 21,
    "args": {
      "External id": 200970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200970,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200970, "pid": 5, "tid": 7, "ts": 1716454224591772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524106, "dur": 10,
    "args": {
      "External id": 200970,
      "cbid": 211, "correlation": 200970
    }
  },
  {
    "ph": "s", "id": 200970, "pid": 76337, "tid": -914061504, "ts": 1716454224524106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524169, "dur": 0,
    "args": {
      "External id": 200980,
      "cbid": 317, "correlation": 200980
    }
  },
  {
    "ph": "f", "id": 200980, "pid": 76337, "tid": -914061504, "ts": 1716454224524169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524169, "dur": 0,
    "args": {
      "External id": 200981,
      "cbid": 203, "correlation": 200981
    }
  },
  {
    "ph": "f", "id": 200981, "pid": 76337, "tid": -914061504, "ts": 1716454224524169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524170, "dur": 0,
    "args": {
      "External id": 200982,
      "cbid": 205, "correlation": 200982
    }
  },
  {
    "ph": "f", "id": 200982, "pid": 76337, "tid": -914061504, "ts": 1716454224524170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224591793, "dur": 23,
    "args": {
      "External id": 200986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200986,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200986, "pid": 5, "tid": 7, "ts": 1716454224591793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524186, "dur": 11,
    "args": {
      "External id": 200986,
      "cbid": 211, "correlation": 200986
    }
  },
  {
    "ph": "s", "id": 200986, "pid": 76337, "tid": -914061504, "ts": 1716454224524186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224591818, "dur": 44,
    "args": {
      "External id": 200988,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200988,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200988, "pid": 5, "tid": 7, "ts": 1716454224591818,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524200, "dur": 5,
    "args": {
      "External id": 200988,
      "cbid": 211, "correlation": 200988
    }
  },
  {
    "ph": "s", "id": 200988, "pid": 76337, "tid": -914061504, "ts": 1716454224524200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224591864, "dur": 646,
    "args": {
      "External id": 200990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200990,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 200990, "pid": 5, "tid": 7, "ts": 1716454224591864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524212, "dur": 6,
    "args": {
      "External id": 200990,
      "cbid": 211, "correlation": 200990
    }
  },
  {
    "ph": "s", "id": 200990, "pid": 76337, "tid": -914061504, "ts": 1716454224524212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224592511, "dur": 22,
    "args": {
      "External id": 200992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200992, "pid": 5, "tid": 7, "ts": 1716454224592511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524222, "dur": 5,
    "args": {
      "External id": 200992,
      "cbid": 211, "correlation": 200992
    }
  },
  {
    "ph": "s", "id": 200992, "pid": 76337, "tid": -914061504, "ts": 1716454224524222,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224592534, "dur": 33,
    "args": {
      "External id": 200998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 200998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 200998, "pid": 5, "tid": 7, "ts": 1716454224592534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524250, "dur": 10,
    "args": {
      "External id": 200998,
      "cbid": 211, "correlation": 200998
    }
  },
  {
    "ph": "s", "id": 200998, "pid": 76337, "tid": -914061504, "ts": 1716454224524250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524309, "dur": 0,
    "args": {
      "External id": 201008,
      "cbid": 317, "correlation": 201008
    }
  },
  {
    "ph": "f", "id": 201008, "pid": 76337, "tid": -914061504, "ts": 1716454224524309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524309, "dur": 0,
    "args": {
      "External id": 201009,
      "cbid": 203, "correlation": 201009
    }
  },
  {
    "ph": "f", "id": 201009, "pid": 76337, "tid": -914061504, "ts": 1716454224524309,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524310, "dur": 0,
    "args": {
      "External id": 201010,
      "cbid": 205, "correlation": 201010
    }
  },
  {
    "ph": "f", "id": 201010, "pid": 76337, "tid": -914061504, "ts": 1716454224524310,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224592568, "dur": 56,
    "args": {
      "External id": 201014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201014,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201014, "pid": 5, "tid": 7, "ts": 1716454224592568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524323, "dur": 11,
    "args": {
      "External id": 201014,
      "cbid": 211, "correlation": 201014
    }
  },
  {
    "ph": "s", "id": 201014, "pid": 76337, "tid": -914061504, "ts": 1716454224524323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224592625, "dur": 270,
    "args": {
      "External id": 201016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201016,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201016, "pid": 5, "tid": 7, "ts": 1716454224592625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524341, "dur": 8,
    "args": {
      "External id": 201016,
      "cbid": 211, "correlation": 201016
    }
  },
  {
    "ph": "s", "id": 201016, "pid": 76337, "tid": -914061504, "ts": 1716454224524341,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224592896, "dur": 22,
    "args": {
      "External id": 201018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201018,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201018, "pid": 5, "tid": 7, "ts": 1716454224592896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524353, "dur": 6,
    "args": {
      "External id": 201018,
      "cbid": 211, "correlation": 201018
    }
  },
  {
    "ph": "s", "id": 201018, "pid": 76337, "tid": -914061504, "ts": 1716454224524353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224592919, "dur": 32,
    "args": {
      "External id": 201024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201024,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201024, "pid": 5, "tid": 7, "ts": 1716454224592919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524379, "dur": 8,
    "args": {
      "External id": 201024,
      "cbid": 211, "correlation": 201024
    }
  },
  {
    "ph": "s", "id": 201024, "pid": 76337, "tid": -914061504, "ts": 1716454224524379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224592953, "dur": 27,
    "args": {
      "External id": 201032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201032,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201032, "pid": 5, "tid": 7, "ts": 1716454224592953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524408, "dur": 8,
    "args": {
      "External id": 201032,
      "cbid": 211, "correlation": 201032
    }
  },
  {
    "ph": "s", "id": 201032, "pid": 76337, "tid": -914061504, "ts": 1716454224524408,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224592981, "dur": 20,
    "args": {
      "External id": 201040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201040,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201040, "pid": 5, "tid": 7, "ts": 1716454224592981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524437, "dur": 8,
    "args": {
      "External id": 201040,
      "cbid": 211, "correlation": 201040
    }
  },
  {
    "ph": "s", "id": 201040, "pid": 76337, "tid": -914061504, "ts": 1716454224524437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224593002, "dur": 30,
    "args": {
      "External id": 201060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201060,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 201060, "pid": 5, "tid": 7, "ts": 1716454224593002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524520, "dur": 12,
    "args": {
      "External id": 201060,
      "cbid": 211, "correlation": 201060
    }
  },
  {
    "ph": "s", "id": 201060, "pid": 76337, "tid": -914061504, "ts": 1716454224524520,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224593033, "dur": 5,
    "args": {
      "External id": 201072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201072,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 201072, "pid": 5, "tid": 7, "ts": 1716454224593033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524542, "dur": 6,
    "args": {
      "External id": 201072,
      "cbid": 211, "correlation": 201072
    }
  },
  {
    "ph": "s", "id": 201072, "pid": 76337, "tid": -914061504, "ts": 1716454224524542,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224593039, "dur": 30,
    "args": {
      "External id": 201075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201075,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201075, "pid": 5, "tid": 7, "ts": 1716454224593039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524560, "dur": 7,
    "args": {
      "External id": 201075,
      "cbid": 211, "correlation": 201075
    }
  },
  {
    "ph": "s", "id": 201075, "pid": 76337, "tid": -914061504, "ts": 1716454224524560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524617, "dur": 0,
    "args": {
      "External id": 201086,
      "cbid": 317, "correlation": 201086
    }
  },
  {
    "ph": "f", "id": 201086, "pid": 76337, "tid": -914061504, "ts": 1716454224524617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524617, "dur": 0,
    "args": {
      "External id": 201087,
      "cbid": 203, "correlation": 201087
    }
  },
  {
    "ph": "f", "id": 201087, "pid": 76337, "tid": -914061504, "ts": 1716454224524617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524618, "dur": 0,
    "args": {
      "External id": 201088,
      "cbid": 205, "correlation": 201088
    }
  },
  {
    "ph": "f", "id": 201088, "pid": 76337, "tid": -914061504, "ts": 1716454224524618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224593071, "dur": 21,
    "args": {
      "External id": 201092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201092,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201092, "pid": 5, "tid": 7, "ts": 1716454224593071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524632, "dur": 11,
    "args": {
      "External id": 201092,
      "cbid": 211, "correlation": 201092
    }
  },
  {
    "ph": "s", "id": 201092, "pid": 76337, "tid": -914061504, "ts": 1716454224524632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224593094, "dur": 105,
    "args": {
      "External id": 201094,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201094,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201094, "pid": 5, "tid": 7, "ts": 1716454224593094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524649, "dur": 6,
    "args": {
      "External id": 201094,
      "cbid": 211, "correlation": 201094
    }
  },
  {
    "ph": "s", "id": 201094, "pid": 76337, "tid": -914061504, "ts": 1716454224524649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224593200, "dur": 23,
    "args": {
      "External id": 201096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201096, "pid": 5, "tid": 7, "ts": 1716454224593200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524660, "dur": 5,
    "args": {
      "External id": 201096,
      "cbid": 211, "correlation": 201096
    }
  },
  {
    "ph": "s", "id": 201096, "pid": 76337, "tid": -914061504, "ts": 1716454224524660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224593225, "dur": 33,
    "args": {
      "External id": 201102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201102, "pid": 5, "tid": 7, "ts": 1716454224593225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524687, "dur": 8,
    "args": {
      "External id": 201102,
      "cbid": 211, "correlation": 201102
    }
  },
  {
    "ph": "s", "id": 201102, "pid": 76337, "tid": -914061504, "ts": 1716454224524687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224593259, "dur": 199,
    "args": {
      "External id": 201111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201111,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201111, "pid": 5, "tid": 7, "ts": 1716454224593259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524769, "dur": 14,
    "args": {
      "External id": 201111,
      "cbid": 211, "correlation": 201111
    }
  },
  {
    "ph": "s", "id": 201111, "pid": 76337, "tid": -914061504, "ts": 1716454224524769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224593459, "dur": 65,
    "args": {
      "External id": 201133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201133,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201133, "pid": 5, "tid": 7, "ts": 1716454224593459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524826, "dur": 10,
    "args": {
      "External id": 201133,
      "cbid": 211, "correlation": 201133
    }
  },
  {
    "ph": "s", "id": 201133, "pid": 76337, "tid": -914061504, "ts": 1716454224524826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524916, "dur": 1,
    "args": {
      "External id": 201144,
      "cbid": 251, "correlation": 201144
    }
  },
  {
    "ph": "f", "id": 201144, "pid": 76337, "tid": -914061504, "ts": 1716454224524916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224593525, "dur": 152,
    "args": {
      "External id": 201145,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201145,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201145, "pid": 5, "tid": 7, "ts": 1716454224593525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524921, "dur": 13,
    "args": {
      "External id": 201145,
      "cbid": 211, "correlation": 201145
    }
  },
  {
    "ph": "s", "id": 201145, "pid": 76337, "tid": -914061504, "ts": 1716454224524921,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224524998, "dur": 1,
    "args": {
      "External id": 201156,
      "cbid": 251, "correlation": 201156
    }
  },
  {
    "ph": "f", "id": 201156, "pid": 76337, "tid": -914061504, "ts": 1716454224524998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224593679, "dur": 147,
    "args": {
      "External id": 201157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201157,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201157, "pid": 5, "tid": 7, "ts": 1716454224593679,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525002, "dur": 13,
    "args": {
      "External id": 201157,
      "cbid": 211, "correlation": 201157
    }
  },
  {
    "ph": "s", "id": 201157, "pid": 76337, "tid": -914061504, "ts": 1716454224525002,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525071, "dur": 1,
    "args": {
      "External id": 201168,
      "cbid": 251, "correlation": 201168
    }
  },
  {
    "ph": "f", "id": 201168, "pid": 76337, "tid": -914061504, "ts": 1716454224525071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224593827, "dur": 145,
    "args": {
      "External id": 201169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201169,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201169, "pid": 5, "tid": 7, "ts": 1716454224593827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525075, "dur": 11,
    "args": {
      "External id": 201169,
      "cbid": 211, "correlation": 201169
    }
  },
  {
    "ph": "s", "id": 201169, "pid": 76337, "tid": -914061504, "ts": 1716454224525075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224593973, "dur": 1948,
    "args": {
      "External id": 201190,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201190,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 201190, "pid": 5, "tid": 7, "ts": 1716454224593973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525160, "dur": 14,
    "args": {
      "External id": 201190,
      "cbid": 211, "correlation": 201190
    }
  },
  {
    "ph": "s", "id": 201190, "pid": 76337, "tid": -914061504, "ts": 1716454224525160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525260, "dur": 2,
    "args": {
      "External id": 201208,
      "cbid": 251, "correlation": 201208
    }
  },
  {
    "ph": "f", "id": 201208, "pid": 76337, "tid": -914061504, "ts": 1716454224525260,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224595923, "dur": 150,
    "args": {
      "External id": 201210,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201210,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 201210, "pid": 5, "tid": 7, "ts": 1716454224595923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525267, "dur": 13,
    "args": {
      "External id": 201210,
      "cbid": 211, "correlation": 201210
    }
  },
  {
    "ph": "s", "id": 201210, "pid": 76337, "tid": -914061504, "ts": 1716454224525267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224596074, "dur": 35,
    "args": {
      "External id": 201218,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201218,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201218, "pid": 5, "tid": 7, "ts": 1716454224596074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525337, "dur": 12,
    "args": {
      "External id": 201218,
      "cbid": 211, "correlation": 201218
    }
  },
  {
    "ph": "s", "id": 201218, "pid": 76337, "tid": -914061504, "ts": 1716454224525337,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224596111, "dur": 50,
    "args": {
      "External id": 201226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201226,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201226, "pid": 5, "tid": 7, "ts": 1716454224596111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525376, "dur": 8,
    "args": {
      "External id": 201226,
      "cbid": 211, "correlation": 201226
    }
  },
  {
    "ph": "s", "id": 201226, "pid": 76337, "tid": -914061504, "ts": 1716454224525376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224596162, "dur": 30,
    "args": {
      "External id": 201237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201237,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201237, "pid": 5, "tid": 7, "ts": 1716454224596162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525447, "dur": 12,
    "args": {
      "External id": 201237,
      "cbid": 211, "correlation": 201237
    }
  },
  {
    "ph": "s", "id": 201237, "pid": 76337, "tid": -914061504, "ts": 1716454224525447,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224596193, "dur": 34,
    "args": {
      "External id": 201259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201259,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201259, "pid": 5, "tid": 7, "ts": 1716454224596193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525478, "dur": 8,
    "args": {
      "External id": 201259,
      "cbid": 211, "correlation": 201259
    }
  },
  {
    "ph": "s", "id": 201259, "pid": 76337, "tid": -914061504, "ts": 1716454224525478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525562, "dur": 1,
    "args": {
      "External id": 201270,
      "cbid": 251, "correlation": 201270
    }
  },
  {
    "ph": "f", "id": 201270, "pid": 76337, "tid": -914061504, "ts": 1716454224525562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224596229, "dur": 91,
    "args": {
      "External id": 201271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201271,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201271, "pid": 5, "tid": 7, "ts": 1716454224596229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525568, "dur": 12,
    "args": {
      "External id": 201271,
      "cbid": 211, "correlation": 201271
    }
  },
  {
    "ph": "s", "id": 201271, "pid": 76337, "tid": -914061504, "ts": 1716454224525568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525636, "dur": 1,
    "args": {
      "External id": 201282,
      "cbid": 251, "correlation": 201282
    }
  },
  {
    "ph": "f", "id": 201282, "pid": 76337, "tid": -914061504, "ts": 1716454224525636,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525639, "dur": 0,
    "args": {
      "External id": 201283,
      "cbid": 251, "correlation": 201283
    }
  },
  {
    "ph": "f", "id": 201283, "pid": 76337, "tid": -914061504, "ts": 1716454224525639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224596321, "dur": 11,
    "args": {
      "External id": 201284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201284,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 201284, "pid": 5, "tid": 7, "ts": 1716454224596321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525641, "dur": 12,
    "args": {
      "External id": 201284,
      "cbid": 211, "correlation": 201284
    }
  },
  {
    "ph": "s", "id": 201284, "pid": 76337, "tid": -914061504, "ts": 1716454224525641,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224596333, "dur": 5,
    "args": {
      "External id": 201286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201286,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 201286, "pid": 5, "tid": 7, "ts": 1716454224596333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525654, "dur": 6,
    "args": {
      "External id": 201286,
      "cbid": 211, "correlation": 201286
    }
  },
  {
    "ph": "s", "id": 201286, "pid": 76337, "tid": -914061504, "ts": 1716454224525654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525712, "dur": 1,
    "args": {
      "External id": 201297,
      "cbid": 251, "correlation": 201297
    }
  },
  {
    "ph": "f", "id": 201297, "pid": 76337, "tid": -914061504, "ts": 1716454224525712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525716, "dur": 0,
    "args": {
      "External id": 201298,
      "cbid": 251, "correlation": 201298
    }
  },
  {
    "ph": "f", "id": 201298, "pid": 76337, "tid": -914061504, "ts": 1716454224525716,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224596339, "dur": 8,
    "args": {
      "External id": 201299,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201299,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 201299, "pid": 5, "tid": 7, "ts": 1716454224596339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525718, "dur": 11,
    "args": {
      "External id": 201299,
      "cbid": 211, "correlation": 201299
    }
  },
  {
    "ph": "s", "id": 201299, "pid": 76337, "tid": -914061504, "ts": 1716454224525718,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224596348, "dur": 4,
    "args": {
      "External id": 201301,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201301,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 201301, "pid": 5, "tid": 7, "ts": 1716454224596348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525731, "dur": 5,
    "args": {
      "External id": 201301,
      "cbid": 211, "correlation": 201301
    }
  },
  {
    "ph": "s", "id": 201301, "pid": 76337, "tid": -914061504, "ts": 1716454224525731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224596354, "dur": 93,
    "args": {
      "External id": 201322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201322,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 201322, "pid": 5, "tid": 7, "ts": 1716454224596354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525803, "dur": 13,
    "args": {
      "External id": 201322,
      "cbid": 211, "correlation": 201322
    }
  },
  {
    "ph": "s", "id": 201322, "pid": 76337, "tid": -914061504, "ts": 1716454224525803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525899, "dur": 1,
    "args": {
      "External id": 201340,
      "cbid": 251, "correlation": 201340
    }
  },
  {
    "ph": "f", "id": 201340, "pid": 76337, "tid": -914061504, "ts": 1716454224525899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224596448, "dur": 98,
    "args": {
      "External id": 201342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201342,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201342, "pid": 5, "tid": 7, "ts": 1716454224596448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525904, "dur": 14,
    "args": {
      "External id": 201342,
      "cbid": 211, "correlation": 201342
    }
  },
  {
    "ph": "s", "id": 201342, "pid": 76337, "tid": -914061504, "ts": 1716454224525904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224596548, "dur": 19,
    "args": {
      "External id": 201350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201350,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201350, "pid": 5, "tid": 7, "ts": 1716454224596548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224525983, "dur": 13,
    "args": {
      "External id": 201350,
      "cbid": 211, "correlation": 201350
    }
  },
  {
    "ph": "s", "id": 201350, "pid": 76337, "tid": -914061504, "ts": 1716454224525983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224596568, "dur": 37,
    "args": {
      "External id": 201358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201358,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201358, "pid": 5, "tid": 7, "ts": 1716454224596568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526027, "dur": 9,
    "args": {
      "External id": 201358,
      "cbid": 211, "correlation": 201358
    }
  },
  {
    "ph": "s", "id": 201358, "pid": 76337, "tid": -914061504, "ts": 1716454224526027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224596607, "dur": 35,
    "args": {
      "External id": 201380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201380,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201380, "pid": 5, "tid": 7, "ts": 1716454224596607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526078, "dur": 10,
    "args": {
      "External id": 201380,
      "cbid": 211, "correlation": 201380
    }
  },
  {
    "ph": "s", "id": 201380, "pid": 76337, "tid": -914061504, "ts": 1716454224526078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526167, "dur": 2,
    "args": {
      "External id": 201396,
      "cbid": 251, "correlation": 201396
    }
  },
  {
    "ph": "f", "id": 201396, "pid": 76337, "tid": -914061504, "ts": 1716454224526167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526172, "dur": 0,
    "args": {
      "External id": 201398,
      "cbid": 251, "correlation": 201398
    }
  },
  {
    "ph": "f", "id": 201398, "pid": 76337, "tid": -914061504, "ts": 1716454224526172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224596643, "dur": 540,
    "args": {
      "External id": 201399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201399,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 201399, "pid": 5, "tid": 7, "ts": 1716454224596643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526177, "dur": 13,
    "args": {
      "External id": 201399,
      "cbid": 211, "correlation": 201399
    }
  },
  {
    "ph": "s", "id": 201399, "pid": 76337, "tid": -914061504, "ts": 1716454224526177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224597184, "dur": 126,
    "args": {
      "External id": 201407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201407,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201407, "pid": 5, "tid": 7, "ts": 1716454224597184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526242, "dur": 12,
    "args": {
      "External id": 201407,
      "cbid": 211, "correlation": 201407
    }
  },
  {
    "ph": "s", "id": 201407, "pid": 76337, "tid": -914061504, "ts": 1716454224526242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224597312, "dur": 128,
    "args": {
      "External id": 201415,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201415,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201415, "pid": 5, "tid": 7, "ts": 1716454224597312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526275, "dur": 8,
    "args": {
      "External id": 201415,
      "cbid": 211, "correlation": 201415
    }
  },
  {
    "ph": "s", "id": 201415, "pid": 76337, "tid": -914061504, "ts": 1716454224526275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526352, "dur": 1,
    "args": {
      "External id": 201431,
      "cbid": 251, "correlation": 201431
    }
  },
  {
    "ph": "f", "id": 201431, "pid": 76337, "tid": -914061504, "ts": 1716454224526352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224597441, "dur": 304,
    "args": {
      "External id": 201433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201433,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201433, "pid": 5, "tid": 7, "ts": 1716454224597441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526357, "dur": 13,
    "args": {
      "External id": 201433,
      "cbid": 211, "correlation": 201433
    }
  },
  {
    "ph": "s", "id": 201433, "pid": 76337, "tid": -914061504, "ts": 1716454224526357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224597746, "dur": 27,
    "args": {
      "External id": 201441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201441,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201441, "pid": 5, "tid": 7, "ts": 1716454224597746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526399, "dur": 10,
    "args": {
      "External id": 201441,
      "cbid": 211, "correlation": 201441
    }
  },
  {
    "ph": "s", "id": 201441, "pid": 76337, "tid": -914061504, "ts": 1716454224526399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224597774, "dur": 81,
    "args": {
      "External id": 201452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201452,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201452, "pid": 5, "tid": 7, "ts": 1716454224597774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526467, "dur": 12,
    "args": {
      "External id": 201452,
      "cbid": 211, "correlation": 201452
    }
  },
  {
    "ph": "s", "id": 201452, "pid": 76337, "tid": -914061504, "ts": 1716454224526467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526529, "dur": 0,
    "args": {
      "External id": 201464,
      "cbid": 317, "correlation": 201464
    }
  },
  {
    "ph": "f", "id": 201464, "pid": 76337, "tid": -914061504, "ts": 1716454224526529,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526530, "dur": 0,
    "args": {
      "External id": 201465,
      "cbid": 203, "correlation": 201465
    }
  },
  {
    "ph": "f", "id": 201465, "pid": 76337, "tid": -914061504, "ts": 1716454224526530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526530, "dur": 0,
    "args": {
      "External id": 201466,
      "cbid": 205, "correlation": 201466
    }
  },
  {
    "ph": "f", "id": 201466, "pid": 76337, "tid": -914061504, "ts": 1716454224526530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224597857, "dur": 22,
    "args": {
      "External id": 201470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201470,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201470, "pid": 5, "tid": 7, "ts": 1716454224597857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526546, "dur": 12,
    "args": {
      "External id": 201470,
      "cbid": 211, "correlation": 201470
    }
  },
  {
    "ph": "s", "id": 201470, "pid": 76337, "tid": -914061504, "ts": 1716454224526546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224597880, "dur": 121,
    "args": {
      "External id": 201472,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201472,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201472, "pid": 5, "tid": 7, "ts": 1716454224597880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526565, "dur": 7,
    "args": {
      "External id": 201472,
      "cbid": 211, "correlation": 201472
    }
  },
  {
    "ph": "s", "id": 201472, "pid": 76337, "tid": -914061504, "ts": 1716454224526565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224598002, "dur": 24,
    "args": {
      "External id": 201474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201474,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201474, "pid": 5, "tid": 7, "ts": 1716454224598002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526576, "dur": 5,
    "args": {
      "External id": 201474,
      "cbid": 211, "correlation": 201474
    }
  },
  {
    "ph": "s", "id": 201474, "pid": 76337, "tid": -914061504, "ts": 1716454224526576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224598027, "dur": 32,
    "args": {
      "External id": 201480,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201480,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201480, "pid": 5, "tid": 7, "ts": 1716454224598027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526603, "dur": 9,
    "args": {
      "External id": 201480,
      "cbid": 211, "correlation": 201480
    }
  },
  {
    "ph": "s", "id": 201480, "pid": 76337, "tid": -914061504, "ts": 1716454224526603,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224598061, "dur": 27,
    "args": {
      "External id": 201488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201488,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201488, "pid": 5, "tid": 7, "ts": 1716454224598061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526635, "dur": 8,
    "args": {
      "External id": 201488,
      "cbid": 211, "correlation": 201488
    }
  },
  {
    "ph": "s", "id": 201488, "pid": 76337, "tid": -914061504, "ts": 1716454224526635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224598089, "dur": 54,
    "args": {
      "External id": 201497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201497,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201497, "pid": 5, "tid": 7, "ts": 1716454224598089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526673, "dur": 10,
    "args": {
      "External id": 201497,
      "cbid": 211, "correlation": 201497
    }
  },
  {
    "ph": "s", "id": 201497, "pid": 76337, "tid": -914061504, "ts": 1716454224526673,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224598144, "dur": 52,
    "args": {
      "External id": 201517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201517,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 201517, "pid": 5, "tid": 7, "ts": 1716454224598144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526742, "dur": 11,
    "args": {
      "External id": 201517,
      "cbid": 211, "correlation": 201517
    }
  },
  {
    "ph": "s", "id": 201517, "pid": 76337, "tid": -914061504, "ts": 1716454224526742,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224598198, "dur": 4,
    "args": {
      "External id": 201529,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201529,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 201529, "pid": 5, "tid": 7, "ts": 1716454224598198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526763, "dur": 7,
    "args": {
      "External id": 201529,
      "cbid": 211, "correlation": 201529
    }
  },
  {
    "ph": "s", "id": 201529, "pid": 76337, "tid": -914061504, "ts": 1716454224526763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224598204, "dur": 56,
    "args": {
      "External id": 201532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201532,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201532, "pid": 5, "tid": 7, "ts": 1716454224598204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526782, "dur": 6,
    "args": {
      "External id": 201532,
      "cbid": 211, "correlation": 201532
    }
  },
  {
    "ph": "s", "id": 201532, "pid": 76337, "tid": -914061504, "ts": 1716454224526782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224598261, "dur": 37,
    "args": {
      "External id": 201541,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201541,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201541, "pid": 5, "tid": 7, "ts": 1716454224598261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526822, "dur": 9,
    "args": {
      "External id": 201541,
      "cbid": 211, "correlation": 201541
    }
  },
  {
    "ph": "s", "id": 201541, "pid": 76337, "tid": -914061504, "ts": 1716454224526822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526874, "dur": 0,
    "args": {
      "External id": 201551,
      "cbid": 317, "correlation": 201551
    }
  },
  {
    "ph": "f", "id": 201551, "pid": 76337, "tid": -914061504, "ts": 1716454224526874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526875, "dur": 0,
    "args": {
      "External id": 201552,
      "cbid": 203, "correlation": 201552
    }
  },
  {
    "ph": "f", "id": 201552, "pid": 76337, "tid": -914061504, "ts": 1716454224526875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526875, "dur": 0,
    "args": {
      "External id": 201553,
      "cbid": 205, "correlation": 201553
    }
  },
  {
    "ph": "f", "id": 201553, "pid": 76337, "tid": -914061504, "ts": 1716454224526875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224598300, "dur": 40,
    "args": {
      "External id": 201557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201557,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201557, "pid": 5, "tid": 7, "ts": 1716454224598300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526890, "dur": 11,
    "args": {
      "External id": 201557,
      "cbid": 211, "correlation": 201557
    }
  },
  {
    "ph": "s", "id": 201557, "pid": 76337, "tid": -914061504, "ts": 1716454224526890,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224598341, "dur": 83,
    "args": {
      "External id": 201559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201559,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201559, "pid": 5, "tid": 7, "ts": 1716454224598341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526904, "dur": 5,
    "args": {
      "External id": 201559,
      "cbid": 211, "correlation": 201559
    }
  },
  {
    "ph": "s", "id": 201559, "pid": 76337, "tid": -914061504, "ts": 1716454224526904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224598426, "dur": 1282,
    "args": {
      "External id": 201561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201561,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201561, "pid": 5, "tid": 7, "ts": 1716454224598426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526915, "dur": 7,
    "args": {
      "External id": 201561,
      "cbid": 211, "correlation": 201561
    }
  },
  {
    "ph": "s", "id": 201561, "pid": 76337, "tid": -914061504, "ts": 1716454224526915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224599710, "dur": 20,
    "args": {
      "External id": 201563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201563,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201563, "pid": 5, "tid": 7, "ts": 1716454224599710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526926, "dur": 5,
    "args": {
      "External id": 201563,
      "cbid": 211, "correlation": 201563
    }
  },
  {
    "ph": "s", "id": 201563, "pid": 76337, "tid": -914061504, "ts": 1716454224526926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224599731, "dur": 33,
    "args": {
      "External id": 201569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201569, "pid": 5, "tid": 7, "ts": 1716454224599731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224526953, "dur": 9,
    "args": {
      "External id": 201569,
      "cbid": 211, "correlation": 201569
    }
  },
  {
    "ph": "s", "id": 201569, "pid": 76337, "tid": -914061504, "ts": 1716454224526953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224599766, "dur": 3,
    "args": {
      "External id": 201577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201577,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 201577, "pid": 5, "tid": 7, "ts": 1716454224599766,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527005, "dur": 11,
    "args": {
      "External id": 201577,
      "cbid": 211, "correlation": 201577
    }
  },
  {
    "ph": "s", "id": 201577, "pid": 76337, "tid": -914061504, "ts": 1716454224527005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527071, "dur": 1,
    "args": {
      "External id": 201593,
      "cbid": 251, "correlation": 201593
    }
  },
  {
    "ph": "f", "id": 201593, "pid": 76337, "tid": -914061504, "ts": 1716454224527071,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527076, "dur": 0,
    "args": {
      "External id": 201595,
      "cbid": 251, "correlation": 201595
    }
  },
  {
    "ph": "f", "id": 201595, "pid": 76337, "tid": -914061504, "ts": 1716454224527076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224599771, "dur": 13,
    "args": {
      "External id": 201596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201596,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 201596, "pid": 5, "tid": 7, "ts": 1716454224599771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527078, "dur": 11,
    "args": {
      "External id": 201596,
      "cbid": 211, "correlation": 201596
    }
  },
  {
    "ph": "s", "id": 201596, "pid": 76337, "tid": -914061504, "ts": 1716454224527078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224599785, "dur": 5,
    "args": {
      "External id": 201598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201598,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 201598, "pid": 5, "tid": 7, "ts": 1716454224599785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527091, "dur": 5,
    "args": {
      "External id": 201598,
      "cbid": 211, "correlation": 201598
    }
  },
  {
    "ph": "s", "id": 201598, "pid": 76337, "tid": -914061504, "ts": 1716454224527091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224599791, "dur": 28,
    "args": {
      "External id": 201608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201608,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201608, "pid": 5, "tid": 7, "ts": 1716454224599791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527149, "dur": 13,
    "args": {
      "External id": 201608,
      "cbid": 211, "correlation": 201608
    }
  },
  {
    "ph": "s", "id": 201608, "pid": 76337, "tid": -914061504, "ts": 1716454224527149,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224599821, "dur": 31,
    "args": {
      "External id": 201628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201628,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 201628, "pid": 5, "tid": 7, "ts": 1716454224599821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527214, "dur": 11,
    "args": {
      "External id": 201628,
      "cbid": 211, "correlation": 201628
    }
  },
  {
    "ph": "s", "id": 201628, "pid": 76337, "tid": -914061504, "ts": 1716454224527214,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224599853, "dur": 4,
    "args": {
      "External id": 201640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201640,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 201640, "pid": 5, "tid": 7, "ts": 1716454224599853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527235, "dur": 6,
    "args": {
      "External id": 201640,
      "cbid": 211, "correlation": 201640
    }
  },
  {
    "ph": "s", "id": 201640, "pid": 76337, "tid": -914061504, "ts": 1716454224527235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224599858, "dur": 31,
    "args": {
      "External id": 201643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201643, "pid": 5, "tid": 7, "ts": 1716454224599858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527253, "dur": 7,
    "args": {
      "External id": 201643,
      "cbid": 211, "correlation": 201643
    }
  },
  {
    "ph": "s", "id": 201643, "pid": 76337, "tid": -914061504, "ts": 1716454224527253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224599891, "dur": 20,
    "args": {
      "External id": 201652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201652,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201652, "pid": 5, "tid": 7, "ts": 1716454224599891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527295, "dur": 10,
    "args": {
      "External id": 201652,
      "cbid": 211, "correlation": 201652
    }
  },
  {
    "ph": "s", "id": 201652, "pid": 76337, "tid": -914061504, "ts": 1716454224527295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527357, "dur": 0,
    "args": {
      "External id": 201662,
      "cbid": 317, "correlation": 201662
    }
  },
  {
    "ph": "f", "id": 201662, "pid": 76337, "tid": -914061504, "ts": 1716454224527357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527358, "dur": 0,
    "args": {
      "External id": 201663,
      "cbid": 203, "correlation": 201663
    }
  },
  {
    "ph": "f", "id": 201663, "pid": 76337, "tid": -914061504, "ts": 1716454224527358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527358, "dur": 0,
    "args": {
      "External id": 201664,
      "cbid": 205, "correlation": 201664
    }
  },
  {
    "ph": "f", "id": 201664, "pid": 76337, "tid": -914061504, "ts": 1716454224527358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224599912, "dur": 22,
    "args": {
      "External id": 201668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201668,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201668, "pid": 5, "tid": 7, "ts": 1716454224599912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527373, "dur": 12,
    "args": {
      "External id": 201668,
      "cbid": 211, "correlation": 201668
    }
  },
  {
    "ph": "s", "id": 201668, "pid": 76337, "tid": -914061504, "ts": 1716454224527373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224599935, "dur": 44,
    "args": {
      "External id": 201670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201670,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201670, "pid": 5, "tid": 7, "ts": 1716454224599935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527387, "dur": 5,
    "args": {
      "External id": 201670,
      "cbid": 211, "correlation": 201670
    }
  },
  {
    "ph": "s", "id": 201670, "pid": 76337, "tid": -914061504, "ts": 1716454224527387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224599980, "dur": 645,
    "args": {
      "External id": 201672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201672,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201672, "pid": 5, "tid": 7, "ts": 1716454224599980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527398, "dur": 8,
    "args": {
      "External id": 201672,
      "cbid": 211, "correlation": 201672
    }
  },
  {
    "ph": "s", "id": 201672, "pid": 76337, "tid": -914061504, "ts": 1716454224527398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224600627, "dur": 22,
    "args": {
      "External id": 201674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201674, "pid": 5, "tid": 7, "ts": 1716454224600627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527410, "dur": 6,
    "args": {
      "External id": 201674,
      "cbid": 211, "correlation": 201674
    }
  },
  {
    "ph": "s", "id": 201674, "pid": 76337, "tid": -914061504, "ts": 1716454224527410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224600649, "dur": 32,
    "args": {
      "External id": 201680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201680, "pid": 5, "tid": 7, "ts": 1716454224600649,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527439, "dur": 8,
    "args": {
      "External id": 201680,
      "cbid": 211, "correlation": 201680
    }
  },
  {
    "ph": "s", "id": 201680, "pid": 76337, "tid": -914061504, "ts": 1716454224527439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527497, "dur": 0,
    "args": {
      "External id": 201690,
      "cbid": 317, "correlation": 201690
    }
  },
  {
    "ph": "f", "id": 201690, "pid": 76337, "tid": -914061504, "ts": 1716454224527497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527498, "dur": 0,
    "args": {
      "External id": 201691,
      "cbid": 203, "correlation": 201691
    }
  },
  {
    "ph": "f", "id": 201691, "pid": 76337, "tid": -914061504, "ts": 1716454224527498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527498, "dur": 0,
    "args": {
      "External id": 201692,
      "cbid": 205, "correlation": 201692
    }
  },
  {
    "ph": "f", "id": 201692, "pid": 76337, "tid": -914061504, "ts": 1716454224527498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224600683, "dur": 38,
    "args": {
      "External id": 201696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201696,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201696, "pid": 5, "tid": 7, "ts": 1716454224600683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527511, "dur": 12,
    "args": {
      "External id": 201696,
      "cbid": 211, "correlation": 201696
    }
  },
  {
    "ph": "s", "id": 201696, "pid": 76337, "tid": -914061504, "ts": 1716454224527511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224600722, "dur": 190,
    "args": {
      "External id": 201698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201698,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201698, "pid": 5, "tid": 7, "ts": 1716454224600722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527529, "dur": 6,
    "args": {
      "External id": 201698,
      "cbid": 211, "correlation": 201698
    }
  },
  {
    "ph": "s", "id": 201698, "pid": 76337, "tid": -914061504, "ts": 1716454224527529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224600914, "dur": 22,
    "args": {
      "External id": 201700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201700, "pid": 5, "tid": 7, "ts": 1716454224600914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527538, "dur": 5,
    "args": {
      "External id": 201700,
      "cbid": 211, "correlation": 201700
    }
  },
  {
    "ph": "s", "id": 201700, "pid": 76337, "tid": -914061504, "ts": 1716454224527538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224600937, "dur": 32,
    "args": {
      "External id": 201706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201706,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201706, "pid": 5, "tid": 7, "ts": 1716454224600937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527564, "dur": 8,
    "args": {
      "External id": 201706,
      "cbid": 211, "correlation": 201706
    }
  },
  {
    "ph": "s", "id": 201706, "pid": 76337, "tid": -914061504, "ts": 1716454224527564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224600971, "dur": 27,
    "args": {
      "External id": 201714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201714,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201714, "pid": 5, "tid": 7, "ts": 1716454224600971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527593, "dur": 8,
    "args": {
      "External id": 201714,
      "cbid": 211, "correlation": 201714
    }
  },
  {
    "ph": "s", "id": 201714, "pid": 76337, "tid": -914061504, "ts": 1716454224527593,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224600999, "dur": 19,
    "args": {
      "External id": 201722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201722,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201722, "pid": 5, "tid": 7, "ts": 1716454224600999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527622, "dur": 8,
    "args": {
      "External id": 201722,
      "cbid": 211, "correlation": 201722
    }
  },
  {
    "ph": "s", "id": 201722, "pid": 76337, "tid": -914061504, "ts": 1716454224527622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224601020, "dur": 30,
    "args": {
      "External id": 201742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201742,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 201742, "pid": 5, "tid": 7, "ts": 1716454224601020,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527703, "dur": 13,
    "args": {
      "External id": 201742,
      "cbid": 211, "correlation": 201742
    }
  },
  {
    "ph": "s", "id": 201742, "pid": 76337, "tid": -914061504, "ts": 1716454224527703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224601051, "dur": 5,
    "args": {
      "External id": 201754,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201754,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 201754, "pid": 5, "tid": 7, "ts": 1716454224601051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527726, "dur": 6,
    "args": {
      "External id": 201754,
      "cbid": 211, "correlation": 201754
    }
  },
  {
    "ph": "s", "id": 201754, "pid": 76337, "tid": -914061504, "ts": 1716454224527726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224601057, "dur": 30,
    "args": {
      "External id": 201757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201757,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201757, "pid": 5, "tid": 7, "ts": 1716454224601057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527743, "dur": 6,
    "args": {
      "External id": 201757,
      "cbid": 211, "correlation": 201757
    }
  },
  {
    "ph": "s", "id": 201757, "pid": 76337, "tid": -914061504, "ts": 1716454224527743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527800, "dur": 0,
    "args": {
      "External id": 201768,
      "cbid": 317, "correlation": 201768
    }
  },
  {
    "ph": "f", "id": 201768, "pid": 76337, "tid": -914061504, "ts": 1716454224527800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527801, "dur": 0,
    "args": {
      "External id": 201769,
      "cbid": 203, "correlation": 201769
    }
  },
  {
    "ph": "f", "id": 201769, "pid": 76337, "tid": -914061504, "ts": 1716454224527801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527801, "dur": 0,
    "args": {
      "External id": 201770,
      "cbid": 205, "correlation": 201770
    }
  },
  {
    "ph": "f", "id": 201770, "pid": 76337, "tid": -914061504, "ts": 1716454224527801,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224601089, "dur": 21,
    "args": {
      "External id": 201774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201774,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201774, "pid": 5, "tid": 7, "ts": 1716454224601089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527815, "dur": 11,
    "args": {
      "External id": 201774,
      "cbid": 211, "correlation": 201774
    }
  },
  {
    "ph": "s", "id": 201774, "pid": 76337, "tid": -914061504, "ts": 1716454224527815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224601111, "dur": 105,
    "args": {
      "External id": 201776,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201776,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201776, "pid": 5, "tid": 7, "ts": 1716454224601111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527832, "dur": 6,
    "args": {
      "External id": 201776,
      "cbid": 211, "correlation": 201776
    }
  },
  {
    "ph": "s", "id": 201776, "pid": 76337, "tid": -914061504, "ts": 1716454224527832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224601218, "dur": 23,
    "args": {
      "External id": 201778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201778, "pid": 5, "tid": 7, "ts": 1716454224601218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527842, "dur": 5,
    "args": {
      "External id": 201778,
      "cbid": 211, "correlation": 201778
    }
  },
  {
    "ph": "s", "id": 201778, "pid": 76337, "tid": -914061504, "ts": 1716454224527842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224601242, "dur": 32,
    "args": {
      "External id": 201784,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201784,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201784, "pid": 5, "tid": 7, "ts": 1716454224601242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527870, "dur": 8,
    "args": {
      "External id": 201784,
      "cbid": 211, "correlation": 201784
    }
  },
  {
    "ph": "s", "id": 201784, "pid": 76337, "tid": -914061504, "ts": 1716454224527870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224601276, "dur": 190,
    "args": {
      "External id": 201793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201793,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201793, "pid": 5, "tid": 7, "ts": 1716454224601276,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224527950, "dur": 15,
    "args": {
      "External id": 201793,
      "cbid": 211, "correlation": 201793
    }
  },
  {
    "ph": "s", "id": 201793, "pid": 76337, "tid": -914061504, "ts": 1716454224527950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224601467, "dur": 65,
    "args": {
      "External id": 201815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201815,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201815, "pid": 5, "tid": 7, "ts": 1716454224601467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528016, "dur": 11,
    "args": {
      "External id": 201815,
      "cbid": 211, "correlation": 201815
    }
  },
  {
    "ph": "s", "id": 201815, "pid": 76337, "tid": -914061504, "ts": 1716454224528016,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528103, "dur": 1,
    "args": {
      "External id": 201826,
      "cbid": 251, "correlation": 201826
    }
  },
  {
    "ph": "f", "id": 201826, "pid": 76337, "tid": -914061504, "ts": 1716454224528103,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224601534, "dur": 153,
    "args": {
      "External id": 201827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201827,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201827, "pid": 5, "tid": 7, "ts": 1716454224601534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528109, "dur": 14,
    "args": {
      "External id": 201827,
      "cbid": 211, "correlation": 201827
    }
  },
  {
    "ph": "s", "id": 201827, "pid": 76337, "tid": -914061504, "ts": 1716454224528109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528179, "dur": 1,
    "args": {
      "External id": 201838,
      "cbid": 251, "correlation": 201838
    }
  },
  {
    "ph": "f", "id": 201838, "pid": 76337, "tid": -914061504, "ts": 1716454224528179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224601688, "dur": 146,
    "args": {
      "External id": 201839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201839,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201839, "pid": 5, "tid": 7, "ts": 1716454224601688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528182, "dur": 11,
    "args": {
      "External id": 201839,
      "cbid": 211, "correlation": 201839
    }
  },
  {
    "ph": "s", "id": 201839, "pid": 76337, "tid": -914061504, "ts": 1716454224528182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528247, "dur": 1,
    "args": {
      "External id": 201850,
      "cbid": 251, "correlation": 201850
    }
  },
  {
    "ph": "f", "id": 201850, "pid": 76337, "tid": -914061504, "ts": 1716454224528247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224601836, "dur": 145,
    "args": {
      "External id": 201851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201851,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201851, "pid": 5, "tid": 7, "ts": 1716454224601836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528251, "dur": 12,
    "args": {
      "External id": 201851,
      "cbid": 211, "correlation": 201851
    }
  },
  {
    "ph": "s", "id": 201851, "pid": 76337, "tid": -914061504, "ts": 1716454224528251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224601982, "dur": 1951,
    "args": {
      "External id": 201872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201872,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 201872, "pid": 5, "tid": 7, "ts": 1716454224601982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528332, "dur": 12,
    "args": {
      "External id": 201872,
      "cbid": 211, "correlation": 201872
    }
  },
  {
    "ph": "s", "id": 201872, "pid": 76337, "tid": -914061504, "ts": 1716454224528332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528430, "dur": 1,
    "args": {
      "External id": 201890,
      "cbid": 251, "correlation": 201890
    }
  },
  {
    "ph": "f", "id": 201890, "pid": 76337, "tid": -914061504, "ts": 1716454224528430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224603935, "dur": 150,
    "args": {
      "External id": 201892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201892,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 201892, "pid": 5, "tid": 7, "ts": 1716454224603935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528436, "dur": 13,
    "args": {
      "External id": 201892,
      "cbid": 211, "correlation": 201892
    }
  },
  {
    "ph": "s", "id": 201892, "pid": 76337, "tid": -914061504, "ts": 1716454224528436,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224604086, "dur": 35,
    "args": {
      "External id": 201900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201900,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201900, "pid": 5, "tid": 7, "ts": 1716454224604086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528505, "dur": 13,
    "args": {
      "External id": 201900,
      "cbid": 211, "correlation": 201900
    }
  },
  {
    "ph": "s", "id": 201900, "pid": 76337, "tid": -914061504, "ts": 1716454224528505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224604123, "dur": 51,
    "args": {
      "External id": 201908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201908,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201908, "pid": 5, "tid": 7, "ts": 1716454224604123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528545, "dur": 8,
    "args": {
      "External id": 201908,
      "cbid": 211, "correlation": 201908
    }
  },
  {
    "ph": "s", "id": 201908, "pid": 76337, "tid": -914061504, "ts": 1716454224528545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224604175, "dur": 30,
    "args": {
      "External id": 201919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201919,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201919, "pid": 5, "tid": 7, "ts": 1716454224604175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528615, "dur": 13,
    "args": {
      "External id": 201919,
      "cbid": 211, "correlation": 201919
    }
  },
  {
    "ph": "s", "id": 201919, "pid": 76337, "tid": -914061504, "ts": 1716454224528615,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224604206, "dur": 34,
    "args": {
      "External id": 201941,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201941,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 201941, "pid": 5, "tid": 7, "ts": 1716454224604206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528646, "dur": 7,
    "args": {
      "External id": 201941,
      "cbid": 211, "correlation": 201941
    }
  },
  {
    "ph": "s", "id": 201941, "pid": 76337, "tid": -914061504, "ts": 1716454224528646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528731, "dur": 1,
    "args": {
      "External id": 201952,
      "cbid": 251, "correlation": 201952
    }
  },
  {
    "ph": "f", "id": 201952, "pid": 76337, "tid": -914061504, "ts": 1716454224528731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224604242, "dur": 90,
    "args": {
      "External id": 201953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201953,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 201953, "pid": 5, "tid": 7, "ts": 1716454224604242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528736, "dur": 13,
    "args": {
      "External id": 201953,
      "cbid": 211, "correlation": 201953
    }
  },
  {
    "ph": "s", "id": 201953, "pid": 76337, "tid": -914061504, "ts": 1716454224528736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528805, "dur": 1,
    "args": {
      "External id": 201964,
      "cbid": 251, "correlation": 201964
    }
  },
  {
    "ph": "f", "id": 201964, "pid": 76337, "tid": -914061504, "ts": 1716454224528805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528810, "dur": 0,
    "args": {
      "External id": 201965,
      "cbid": 251, "correlation": 201965
    }
  },
  {
    "ph": "f", "id": 201965, "pid": 76337, "tid": -914061504, "ts": 1716454224528810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224604333, "dur": 12,
    "args": {
      "External id": 201966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201966,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 201966, "pid": 5, "tid": 7, "ts": 1716454224604333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528812, "dur": 11,
    "args": {
      "External id": 201966,
      "cbid": 211, "correlation": 201966
    }
  },
  {
    "ph": "s", "id": 201966, "pid": 76337, "tid": -914061504, "ts": 1716454224528812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224604346, "dur": 5,
    "args": {
      "External id": 201968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201968,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 201968, "pid": 5, "tid": 7, "ts": 1716454224604346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528825, "dur": 5,
    "args": {
      "External id": 201968,
      "cbid": 211, "correlation": 201968
    }
  },
  {
    "ph": "s", "id": 201968, "pid": 76337, "tid": -914061504, "ts": 1716454224528825,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528882, "dur": 1,
    "args": {
      "External id": 201979,
      "cbid": 251, "correlation": 201979
    }
  },
  {
    "ph": "f", "id": 201979, "pid": 76337, "tid": -914061504, "ts": 1716454224528882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528885, "dur": 0,
    "args": {
      "External id": 201980,
      "cbid": 251, "correlation": 201980
    }
  },
  {
    "ph": "f", "id": 201980, "pid": 76337, "tid": -914061504, "ts": 1716454224528885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224604352, "dur": 7,
    "args": {
      "External id": 201981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201981,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 201981, "pid": 5, "tid": 7, "ts": 1716454224604352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528886, "dur": 11,
    "args": {
      "External id": 201981,
      "cbid": 211, "correlation": 201981
    }
  },
  {
    "ph": "s", "id": 201981, "pid": 76337, "tid": -914061504, "ts": 1716454224528886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224604361, "dur": 4,
    "args": {
      "External id": 201983,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 201983,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 201983, "pid": 5, "tid": 7, "ts": 1716454224604361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528899, "dur": 6,
    "args": {
      "External id": 201983,
      "cbid": 211, "correlation": 201983
    }
  },
  {
    "ph": "s", "id": 201983, "pid": 76337, "tid": -914061504, "ts": 1716454224528899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224604366, "dur": 91,
    "args": {
      "External id": 202004,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202004,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 202004, "pid": 5, "tid": 7, "ts": 1716454224604366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224528981, "dur": 13,
    "args": {
      "External id": 202004,
      "cbid": 211, "correlation": 202004
    }
  },
  {
    "ph": "s", "id": 202004, "pid": 76337, "tid": -914061504, "ts": 1716454224528981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529080, "dur": 1,
    "args": {
      "External id": 202022,
      "cbid": 251, "correlation": 202022
    }
  },
  {
    "ph": "f", "id": 202022, "pid": 76337, "tid": -914061504, "ts": 1716454224529080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224604459, "dur": 98,
    "args": {
      "External id": 202024,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202024,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202024, "pid": 5, "tid": 7, "ts": 1716454224604459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529086, "dur": 13,
    "args": {
      "External id": 202024,
      "cbid": 211, "correlation": 202024
    }
  },
  {
    "ph": "s", "id": 202024, "pid": 76337, "tid": -914061504, "ts": 1716454224529086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224604558, "dur": 19,
    "args": {
      "External id": 202032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202032,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202032, "pid": 5, "tid": 7, "ts": 1716454224604558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529154, "dur": 13,
    "args": {
      "External id": 202032,
      "cbid": 211, "correlation": 202032
    }
  },
  {
    "ph": "s", "id": 202032, "pid": 76337, "tid": -914061504, "ts": 1716454224529154,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224604578, "dur": 38,
    "args": {
      "External id": 202040,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202040,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202040, "pid": 5, "tid": 7, "ts": 1716454224604578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529197, "dur": 9,
    "args": {
      "External id": 202040,
      "cbid": 211, "correlation": 202040
    }
  },
  {
    "ph": "s", "id": 202040, "pid": 76337, "tid": -914061504, "ts": 1716454224529197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224604618, "dur": 35,
    "args": {
      "External id": 202062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202062,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202062, "pid": 5, "tid": 7, "ts": 1716454224604618,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529248, "dur": 10,
    "args": {
      "External id": 202062,
      "cbid": 211, "correlation": 202062
    }
  },
  {
    "ph": "s", "id": 202062, "pid": 76337, "tid": -914061504, "ts": 1716454224529248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529337, "dur": 1,
    "args": {
      "External id": 202078,
      "cbid": 251, "correlation": 202078
    }
  },
  {
    "ph": "f", "id": 202078, "pid": 76337, "tid": -914061504, "ts": 1716454224529337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529342, "dur": 0,
    "args": {
      "External id": 202080,
      "cbid": 251, "correlation": 202080
    }
  },
  {
    "ph": "f", "id": 202080, "pid": 76337, "tid": -914061504, "ts": 1716454224529342,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224604654, "dur": 538,
    "args": {
      "External id": 202081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202081,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 202081, "pid": 5, "tid": 7, "ts": 1716454224604654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529345, "dur": 13,
    "args": {
      "External id": 202081,
      "cbid": 211, "correlation": 202081
    }
  },
  {
    "ph": "s", "id": 202081, "pid": 76337, "tid": -914061504, "ts": 1716454224529345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224605193, "dur": 127,
    "args": {
      "External id": 202089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202089, "pid": 5, "tid": 7, "ts": 1716454224605193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529411, "dur": 13,
    "args": {
      "External id": 202089,
      "cbid": 211, "correlation": 202089
    }
  },
  {
    "ph": "s", "id": 202089, "pid": 76337, "tid": -914061504, "ts": 1716454224529411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224605321, "dur": 130,
    "args": {
      "External id": 202097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202097,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202097, "pid": 5, "tid": 7, "ts": 1716454224605321,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529442, "dur": 8,
    "args": {
      "External id": 202097,
      "cbid": 211, "correlation": 202097
    }
  },
  {
    "ph": "s", "id": 202097, "pid": 76337, "tid": -914061504, "ts": 1716454224529442,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529517, "dur": 1,
    "args": {
      "External id": 202113,
      "cbid": 251, "correlation": 202113
    }
  },
  {
    "ph": "f", "id": 202113, "pid": 76337, "tid": -914061504, "ts": 1716454224529517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224605452, "dur": 309,
    "args": {
      "External id": 202115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202115,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202115, "pid": 5, "tid": 7, "ts": 1716454224605452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529523, "dur": 12,
    "args": {
      "External id": 202115,
      "cbid": 211, "correlation": 202115
    }
  },
  {
    "ph": "s", "id": 202115, "pid": 76337, "tid": -914061504, "ts": 1716454224529523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224605763, "dur": 27,
    "args": {
      "External id": 202123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202123,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202123, "pid": 5, "tid": 7, "ts": 1716454224605763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529565, "dur": 9,
    "args": {
      "External id": 202123,
      "cbid": 211, "correlation": 202123
    }
  },
  {
    "ph": "s", "id": 202123, "pid": 76337, "tid": -914061504, "ts": 1716454224529565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224605791, "dur": 82,
    "args": {
      "External id": 202134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202134,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202134, "pid": 5, "tid": 7, "ts": 1716454224605791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529631, "dur": 12,
    "args": {
      "External id": 202134,
      "cbid": 211, "correlation": 202134
    }
  },
  {
    "ph": "s", "id": 202134, "pid": 76337, "tid": -914061504, "ts": 1716454224529631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529694, "dur": 0,
    "args": {
      "External id": 202146,
      "cbid": 317, "correlation": 202146
    }
  },
  {
    "ph": "f", "id": 202146, "pid": 76337, "tid": -914061504, "ts": 1716454224529694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529695, "dur": 0,
    "args": {
      "External id": 202147,
      "cbid": 203, "correlation": 202147
    }
  },
  {
    "ph": "f", "id": 202147, "pid": 76337, "tid": -914061504, "ts": 1716454224529695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529696, "dur": 0,
    "args": {
      "External id": 202148,
      "cbid": 205, "correlation": 202148
    }
  },
  {
    "ph": "f", "id": 202148, "pid": 76337, "tid": -914061504, "ts": 1716454224529696,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224605875, "dur": 24,
    "args": {
      "External id": 202152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202152,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202152, "pid": 5, "tid": 7, "ts": 1716454224605875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529711, "dur": 12,
    "args": {
      "External id": 202152,
      "cbid": 211, "correlation": 202152
    }
  },
  {
    "ph": "s", "id": 202152, "pid": 76337, "tid": -914061504, "ts": 1716454224529711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224605900, "dur": 121,
    "args": {
      "External id": 202154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202154,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202154, "pid": 5, "tid": 7, "ts": 1716454224605900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529729, "dur": 6,
    "args": {
      "External id": 202154,
      "cbid": 211, "correlation": 202154
    }
  },
  {
    "ph": "s", "id": 202154, "pid": 76337, "tid": -914061504, "ts": 1716454224529729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224606022, "dur": 22,
    "args": {
      "External id": 202156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202156, "pid": 5, "tid": 7, "ts": 1716454224606022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529739, "dur": 5,
    "args": {
      "External id": 202156,
      "cbid": 211, "correlation": 202156
    }
  },
  {
    "ph": "s", "id": 202156, "pid": 76337, "tid": -914061504, "ts": 1716454224529739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224606045, "dur": 33,
    "args": {
      "External id": 202162,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202162,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202162, "pid": 5, "tid": 7, "ts": 1716454224606045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529768, "dur": 8,
    "args": {
      "External id": 202162,
      "cbid": 211, "correlation": 202162
    }
  },
  {
    "ph": "s", "id": 202162, "pid": 76337, "tid": -914061504, "ts": 1716454224529768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224606080, "dur": 26,
    "args": {
      "External id": 202170,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202170,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202170, "pid": 5, "tid": 7, "ts": 1716454224606080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529799, "dur": 8,
    "args": {
      "External id": 202170,
      "cbid": 211, "correlation": 202170
    }
  },
  {
    "ph": "s", "id": 202170, "pid": 76337, "tid": -914061504, "ts": 1716454224529799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224606107, "dur": 45,
    "args": {
      "External id": 202179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202179,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202179, "pid": 5, "tid": 7, "ts": 1716454224606107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529836, "dur": 10,
    "args": {
      "External id": 202179,
      "cbid": 211, "correlation": 202179
    }
  },
  {
    "ph": "s", "id": 202179, "pid": 76337, "tid": -914061504, "ts": 1716454224529836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224606154, "dur": 42,
    "args": {
      "External id": 202199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202199,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 202199, "pid": 5, "tid": 7, "ts": 1716454224606154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529906, "dur": 12,
    "args": {
      "External id": 202199,
      "cbid": 211, "correlation": 202199
    }
  },
  {
    "ph": "s", "id": 202199, "pid": 76337, "tid": -914061504, "ts": 1716454224529906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224606197, "dur": 5,
    "args": {
      "External id": 202211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202211,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 202211, "pid": 5, "tid": 7, "ts": 1716454224606197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529928, "dur": 6,
    "args": {
      "External id": 202211,
      "cbid": 211, "correlation": 202211
    }
  },
  {
    "ph": "s", "id": 202211, "pid": 76337, "tid": -914061504, "ts": 1716454224529928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224606203, "dur": 44,
    "args": {
      "External id": 202214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202214, "pid": 5, "tid": 7, "ts": 1716454224606203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529946, "dur": 7,
    "args": {
      "External id": 202214,
      "cbid": 211, "correlation": 202214
    }
  },
  {
    "ph": "s", "id": 202214, "pid": 76337, "tid": -914061504, "ts": 1716454224529946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224606249, "dur": 29,
    "args": {
      "External id": 202223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202223,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202223, "pid": 5, "tid": 7, "ts": 1716454224606249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224529994, "dur": 10,
    "args": {
      "External id": 202223,
      "cbid": 211, "correlation": 202223
    }
  },
  {
    "ph": "s", "id": 202223, "pid": 76337, "tid": -914061504, "ts": 1716454224529994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530048, "dur": 0,
    "args": {
      "External id": 202233,
      "cbid": 317, "correlation": 202233
    }
  },
  {
    "ph": "f", "id": 202233, "pid": 76337, "tid": -914061504, "ts": 1716454224530048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530049, "dur": 0,
    "args": {
      "External id": 202234,
      "cbid": 203, "correlation": 202234
    }
  },
  {
    "ph": "f", "id": 202234, "pid": 76337, "tid": -914061504, "ts": 1716454224530049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530049, "dur": 0,
    "args": {
      "External id": 202235,
      "cbid": 205, "correlation": 202235
    }
  },
  {
    "ph": "f", "id": 202235, "pid": 76337, "tid": -914061504, "ts": 1716454224530049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224606280, "dur": 31,
    "args": {
      "External id": 202239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202239,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202239, "pid": 5, "tid": 7, "ts": 1716454224606280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530065, "dur": 11,
    "args": {
      "External id": 202239,
      "cbid": 211, "correlation": 202239
    }
  },
  {
    "ph": "s", "id": 202239, "pid": 76337, "tid": -914061504, "ts": 1716454224530065,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224606312, "dur": 64,
    "args": {
      "External id": 202241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202241,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202241, "pid": 5, "tid": 7, "ts": 1716454224606312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530079, "dur": 5,
    "args": {
      "External id": 202241,
      "cbid": 211, "correlation": 202241
    }
  },
  {
    "ph": "s", "id": 202241, "pid": 76337, "tid": -914061504, "ts": 1716454224530079,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224606377, "dur": 975,
    "args": {
      "External id": 202243,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202243,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202243, "pid": 5, "tid": 7, "ts": 1716454224606377,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530090, "dur": 6,
    "args": {
      "External id": 202243,
      "cbid": 211, "correlation": 202243
    }
  },
  {
    "ph": "s", "id": 202243, "pid": 76337, "tid": -914061504, "ts": 1716454224530090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224607354, "dur": 20,
    "args": {
      "External id": 202245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202245,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202245, "pid": 5, "tid": 7, "ts": 1716454224607354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530100, "dur": 5,
    "args": {
      "External id": 202245,
      "cbid": 211, "correlation": 202245
    }
  },
  {
    "ph": "s", "id": 202245, "pid": 76337, "tid": -914061504, "ts": 1716454224530100,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224607375, "dur": 33,
    "args": {
      "External id": 202251,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202251,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202251, "pid": 5, "tid": 7, "ts": 1716454224607375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530128, "dur": 9,
    "args": {
      "External id": 202251,
      "cbid": 211, "correlation": 202251
    }
  },
  {
    "ph": "s", "id": 202251, "pid": 76337, "tid": -914061504, "ts": 1716454224530128,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224607410, "dur": 3,
    "args": {
      "External id": 202259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202259,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 202259, "pid": 5, "tid": 7, "ts": 1716454224607410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530171, "dur": 9,
    "args": {
      "External id": 202259,
      "cbid": 211, "correlation": 202259
    }
  },
  {
    "ph": "s", "id": 202259, "pid": 76337, "tid": -914061504, "ts": 1716454224530171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530234, "dur": 1,
    "args": {
      "External id": 202275,
      "cbid": 251, "correlation": 202275
    }
  },
  {
    "ph": "f", "id": 202275, "pid": 76337, "tid": -914061504, "ts": 1716454224530234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530239, "dur": 0,
    "args": {
      "External id": 202277,
      "cbid": 251, "correlation": 202277
    }
  },
  {
    "ph": "f", "id": 202277, "pid": 76337, "tid": -914061504, "ts": 1716454224530239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224607414, "dur": 12,
    "args": {
      "External id": 202278,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202278,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 202278, "pid": 5, "tid": 7, "ts": 1716454224607414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530241, "dur": 11,
    "args": {
      "External id": 202278,
      "cbid": 211, "correlation": 202278
    }
  },
  {
    "ph": "s", "id": 202278, "pid": 76337, "tid": -914061504, "ts": 1716454224530241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224607428, "dur": 5,
    "args": {
      "External id": 202280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202280,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 202280, "pid": 5, "tid": 7, "ts": 1716454224607428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530254, "dur": 6,
    "args": {
      "External id": 202280,
      "cbid": 211, "correlation": 202280
    }
  },
  {
    "ph": "s", "id": 202280, "pid": 76337, "tid": -914061504, "ts": 1716454224530254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224607435, "dur": 30,
    "args": {
      "External id": 202290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202290,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202290, "pid": 5, "tid": 7, "ts": 1716454224607435,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530312, "dur": 12,
    "args": {
      "External id": 202290,
      "cbid": 211, "correlation": 202290
    }
  },
  {
    "ph": "s", "id": 202290, "pid": 76337, "tid": -914061504, "ts": 1716454224530312,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224607466, "dur": 32,
    "args": {
      "External id": 202310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202310,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 202310, "pid": 5, "tid": 7, "ts": 1716454224607466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530377, "dur": 10,
    "args": {
      "External id": 202310,
      "cbid": 211, "correlation": 202310
    }
  },
  {
    "ph": "s", "id": 202310, "pid": 76337, "tid": -914061504, "ts": 1716454224530377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224607499, "dur": 4,
    "args": {
      "External id": 202322,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202322,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 202322, "pid": 5, "tid": 7, "ts": 1716454224607499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530397, "dur": 6,
    "args": {
      "External id": 202322,
      "cbid": 211, "correlation": 202322
    }
  },
  {
    "ph": "s", "id": 202322, "pid": 76337, "tid": -914061504, "ts": 1716454224530397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224607505, "dur": 30,
    "args": {
      "External id": 202325,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202325,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202325, "pid": 5, "tid": 7, "ts": 1716454224607505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530417, "dur": 6,
    "args": {
      "External id": 202325,
      "cbid": 211, "correlation": 202325
    }
  },
  {
    "ph": "s", "id": 202325, "pid": 76337, "tid": -914061504, "ts": 1716454224530417,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224607536, "dur": 20,
    "args": {
      "External id": 202334,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202334,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202334, "pid": 5, "tid": 7, "ts": 1716454224607536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530456, "dur": 10,
    "args": {
      "External id": 202334,
      "cbid": 211, "correlation": 202334
    }
  },
  {
    "ph": "s", "id": 202334, "pid": 76337, "tid": -914061504, "ts": 1716454224530456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530520, "dur": 0,
    "args": {
      "External id": 202344,
      "cbid": 317, "correlation": 202344
    }
  },
  {
    "ph": "f", "id": 202344, "pid": 76337, "tid": -914061504, "ts": 1716454224530520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530521, "dur": 0,
    "args": {
      "External id": 202345,
      "cbid": 203, "correlation": 202345
    }
  },
  {
    "ph": "f", "id": 202345, "pid": 76337, "tid": -914061504, "ts": 1716454224530521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530522, "dur": 0,
    "args": {
      "External id": 202346,
      "cbid": 205, "correlation": 202346
    }
  },
  {
    "ph": "f", "id": 202346, "pid": 76337, "tid": -914061504, "ts": 1716454224530522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224607558, "dur": 22,
    "args": {
      "External id": 202350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202350,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202350, "pid": 5, "tid": 7, "ts": 1716454224607558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530535, "dur": 12,
    "args": {
      "External id": 202350,
      "cbid": 211, "correlation": 202350
    }
  },
  {
    "ph": "s", "id": 202350, "pid": 76337, "tid": -914061504, "ts": 1716454224530535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224607581, "dur": 44,
    "args": {
      "External id": 202352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202352,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202352, "pid": 5, "tid": 7, "ts": 1716454224607581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530549, "dur": 5,
    "args": {
      "External id": 202352,
      "cbid": 211, "correlation": 202352
    }
  },
  {
    "ph": "s", "id": 202352, "pid": 76337, "tid": -914061504, "ts": 1716454224530549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224607626, "dur": 649,
    "args": {
      "External id": 202354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202354,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202354, "pid": 5, "tid": 7, "ts": 1716454224607626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530560, "dur": 7,
    "args": {
      "External id": 202354,
      "cbid": 211, "correlation": 202354
    }
  },
  {
    "ph": "s", "id": 202354, "pid": 76337, "tid": -914061504, "ts": 1716454224530560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224608277, "dur": 24,
    "args": {
      "External id": 202356,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202356,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202356, "pid": 5, "tid": 7, "ts": 1716454224608277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530571, "dur": 5,
    "args": {
      "External id": 202356,
      "cbid": 211, "correlation": 202356
    }
  },
  {
    "ph": "s", "id": 202356, "pid": 76337, "tid": -914061504, "ts": 1716454224530571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224608302, "dur": 34,
    "args": {
      "External id": 202362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202362, "pid": 5, "tid": 7, "ts": 1716454224608302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530598, "dur": 8,
    "args": {
      "External id": 202362,
      "cbid": 211, "correlation": 202362
    }
  },
  {
    "ph": "s", "id": 202362, "pid": 76337, "tid": -914061504, "ts": 1716454224530598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530655, "dur": 0,
    "args": {
      "External id": 202372,
      "cbid": 317, "correlation": 202372
    }
  },
  {
    "ph": "f", "id": 202372, "pid": 76337, "tid": -914061504, "ts": 1716454224530655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530656, "dur": 0,
    "args": {
      "External id": 202373,
      "cbid": 203, "correlation": 202373
    }
  },
  {
    "ph": "f", "id": 202373, "pid": 76337, "tid": -914061504, "ts": 1716454224530656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530657, "dur": 0,
    "args": {
      "External id": 202374,
      "cbid": 205, "correlation": 202374
    }
  },
  {
    "ph": "f", "id": 202374, "pid": 76337, "tid": -914061504, "ts": 1716454224530657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224608337, "dur": 30,
    "args": {
      "External id": 202378,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202378,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202378, "pid": 5, "tid": 7, "ts": 1716454224608337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530670, "dur": 11,
    "args": {
      "External id": 202378,
      "cbid": 211, "correlation": 202378
    }
  },
  {
    "ph": "s", "id": 202378, "pid": 76337, "tid": -914061504, "ts": 1716454224530670,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224608369, "dur": 153,
    "args": {
      "External id": 202380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202380,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202380, "pid": 5, "tid": 7, "ts": 1716454224608369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530687, "dur": 6,
    "args": {
      "External id": 202380,
      "cbid": 211, "correlation": 202380
    }
  },
  {
    "ph": "s", "id": 202380, "pid": 76337, "tid": -914061504, "ts": 1716454224530687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224608523, "dur": 21,
    "args": {
      "External id": 202382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202382,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202382, "pid": 5, "tid": 7, "ts": 1716454224608523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530697, "dur": 5,
    "args": {
      "External id": 202382,
      "cbid": 211, "correlation": 202382
    }
  },
  {
    "ph": "s", "id": 202382, "pid": 76337, "tid": -914061504, "ts": 1716454224530697,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224608546, "dur": 33,
    "args": {
      "External id": 202388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202388,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202388, "pid": 5, "tid": 7, "ts": 1716454224608546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530723, "dur": 8,
    "args": {
      "External id": 202388,
      "cbid": 211, "correlation": 202388
    }
  },
  {
    "ph": "s", "id": 202388, "pid": 76337, "tid": -914061504, "ts": 1716454224530723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224608580, "dur": 27,
    "args": {
      "External id": 202396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202396,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202396, "pid": 5, "tid": 7, "ts": 1716454224608580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530751, "dur": 8,
    "args": {
      "External id": 202396,
      "cbid": 211, "correlation": 202396
    }
  },
  {
    "ph": "s", "id": 202396, "pid": 76337, "tid": -914061504, "ts": 1716454224530751,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224608608, "dur": 19,
    "args": {
      "External id": 202404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202404,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202404, "pid": 5, "tid": 7, "ts": 1716454224608608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530780, "dur": 8,
    "args": {
      "External id": 202404,
      "cbid": 211, "correlation": 202404
    }
  },
  {
    "ph": "s", "id": 202404, "pid": 76337, "tid": -914061504, "ts": 1716454224530780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224608629, "dur": 29,
    "args": {
      "External id": 202424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202424,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 202424, "pid": 5, "tid": 7, "ts": 1716454224608629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530861, "dur": 13,
    "args": {
      "External id": 202424,
      "cbid": 211, "correlation": 202424
    }
  },
  {
    "ph": "s", "id": 202424, "pid": 76337, "tid": -914061504, "ts": 1716454224530861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224608659, "dur": 5,
    "args": {
      "External id": 202436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202436,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 202436, "pid": 5, "tid": 7, "ts": 1716454224608659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530883, "dur": 6,
    "args": {
      "External id": 202436,
      "cbid": 211, "correlation": 202436
    }
  },
  {
    "ph": "s", "id": 202436, "pid": 76337, "tid": -914061504, "ts": 1716454224530883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224608666, "dur": 32,
    "args": {
      "External id": 202439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202439,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202439, "pid": 5, "tid": 7, "ts": 1716454224608666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530901, "dur": 7,
    "args": {
      "External id": 202439,
      "cbid": 211, "correlation": 202439
    }
  },
  {
    "ph": "s", "id": 202439, "pid": 76337, "tid": -914061504, "ts": 1716454224530901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530957, "dur": 0,
    "args": {
      "External id": 202450,
      "cbid": 317, "correlation": 202450
    }
  },
  {
    "ph": "f", "id": 202450, "pid": 76337, "tid": -914061504, "ts": 1716454224530957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530958, "dur": 0,
    "args": {
      "External id": 202451,
      "cbid": 203, "correlation": 202451
    }
  },
  {
    "ph": "f", "id": 202451, "pid": 76337, "tid": -914061504, "ts": 1716454224530958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530959, "dur": 0,
    "args": {
      "External id": 202452,
      "cbid": 205, "correlation": 202452
    }
  },
  {
    "ph": "f", "id": 202452, "pid": 76337, "tid": -914061504, "ts": 1716454224530959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224608698, "dur": 22,
    "args": {
      "External id": 202456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202456, "pid": 5, "tid": 7, "ts": 1716454224608698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530973, "dur": 21,
    "args": {
      "External id": 202456,
      "cbid": 211, "correlation": 202456
    }
  },
  {
    "ph": "s", "id": 202456, "pid": 76337, "tid": -914061504, "ts": 1716454224530973,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224608722, "dur": 106,
    "args": {
      "External id": 202458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202458,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202458, "pid": 5, "tid": 7, "ts": 1716454224608722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224530999, "dur": 6,
    "args": {
      "External id": 202458,
      "cbid": 211, "correlation": 202458
    }
  },
  {
    "ph": "s", "id": 202458, "pid": 76337, "tid": -914061504, "ts": 1716454224530999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224608829, "dur": 24,
    "args": {
      "External id": 202460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202460,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202460, "pid": 5, "tid": 7, "ts": 1716454224608829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531008, "dur": 5,
    "args": {
      "External id": 202460,
      "cbid": 211, "correlation": 202460
    }
  },
  {
    "ph": "s", "id": 202460, "pid": 76337, "tid": -914061504, "ts": 1716454224531008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224608854, "dur": 32,
    "args": {
      "External id": 202466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202466, "pid": 5, "tid": 7, "ts": 1716454224608854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531037, "dur": 9,
    "args": {
      "External id": 202466,
      "cbid": 211, "correlation": 202466
    }
  },
  {
    "ph": "s", "id": 202466, "pid": 76337, "tid": -914061504, "ts": 1716454224531037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224608887, "dur": 203,
    "args": {
      "External id": 202475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202475,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202475, "pid": 5, "tid": 7, "ts": 1716454224608887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531119, "dur": 14,
    "args": {
      "External id": 202475,
      "cbid": 211, "correlation": 202475
    }
  },
  {
    "ph": "s", "id": 202475, "pid": 76337, "tid": -914061504, "ts": 1716454224531119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224609092, "dur": 66,
    "args": {
      "External id": 202497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202497,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202497, "pid": 5, "tid": 7, "ts": 1716454224609092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531178, "dur": 10,
    "args": {
      "External id": 202497,
      "cbid": 211, "correlation": 202497
    }
  },
  {
    "ph": "s", "id": 202497, "pid": 76337, "tid": -914061504, "ts": 1716454224531178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531264, "dur": 2,
    "args": {
      "External id": 202508,
      "cbid": 251, "correlation": 202508
    }
  },
  {
    "ph": "f", "id": 202508, "pid": 76337, "tid": -914061504, "ts": 1716454224531264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224609159, "dur": 153,
    "args": {
      "External id": 202509,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202509,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202509, "pid": 5, "tid": 7, "ts": 1716454224609159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531270, "dur": 13,
    "args": {
      "External id": 202509,
      "cbid": 211, "correlation": 202509
    }
  },
  {
    "ph": "s", "id": 202509, "pid": 76337, "tid": -914061504, "ts": 1716454224531270,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531341, "dur": 1,
    "args": {
      "External id": 202520,
      "cbid": 251, "correlation": 202520
    }
  },
  {
    "ph": "f", "id": 202520, "pid": 76337, "tid": -914061504, "ts": 1716454224531341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224609314, "dur": 146,
    "args": {
      "External id": 202521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202521,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202521, "pid": 5, "tid": 7, "ts": 1716454224609314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531345, "dur": 11,
    "args": {
      "External id": 202521,
      "cbid": 211, "correlation": 202521
    }
  },
  {
    "ph": "s", "id": 202521, "pid": 76337, "tid": -914061504, "ts": 1716454224531345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531411, "dur": 1,
    "args": {
      "External id": 202532,
      "cbid": 251, "correlation": 202532
    }
  },
  {
    "ph": "f", "id": 202532, "pid": 76337, "tid": -914061504, "ts": 1716454224531411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224609461, "dur": 146,
    "args": {
      "External id": 202533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202533,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202533, "pid": 5, "tid": 7, "ts": 1716454224609461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531416, "dur": 12,
    "args": {
      "External id": 202533,
      "cbid": 211, "correlation": 202533
    }
  },
  {
    "ph": "s", "id": 202533, "pid": 76337, "tid": -914061504, "ts": 1716454224531416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224609609, "dur": 1969,
    "args": {
      "External id": 202554,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202554,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 202554, "pid": 5, "tid": 7, "ts": 1716454224609609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531495, "dur": 12,
    "args": {
      "External id": 202554,
      "cbid": 211, "correlation": 202554
    }
  },
  {
    "ph": "s", "id": 202554, "pid": 76337, "tid": -914061504, "ts": 1716454224531495,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531593, "dur": 1,
    "args": {
      "External id": 202572,
      "cbid": 251, "correlation": 202572
    }
  },
  {
    "ph": "f", "id": 202572, "pid": 76337, "tid": -914061504, "ts": 1716454224531593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224611579, "dur": 150,
    "args": {
      "External id": 202574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202574,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 202574, "pid": 5, "tid": 7, "ts": 1716454224611579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531599, "dur": 13,
    "args": {
      "External id": 202574,
      "cbid": 211, "correlation": 202574
    }
  },
  {
    "ph": "s", "id": 202574, "pid": 76337, "tid": -914061504, "ts": 1716454224531599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224611730, "dur": 36,
    "args": {
      "External id": 202582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202582,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202582, "pid": 5, "tid": 7, "ts": 1716454224611730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531667, "dur": 12,
    "args": {
      "External id": 202582,
      "cbid": 211, "correlation": 202582
    }
  },
  {
    "ph": "s", "id": 202582, "pid": 76337, "tid": -914061504, "ts": 1716454224531667,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224611767, "dur": 51,
    "args": {
      "External id": 202590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202590,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202590, "pid": 5, "tid": 7, "ts": 1716454224611767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531706, "dur": 9,
    "args": {
      "External id": 202590,
      "cbid": 211, "correlation": 202590
    }
  },
  {
    "ph": "s", "id": 202590, "pid": 76337, "tid": -914061504, "ts": 1716454224531706,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224611820, "dur": 30,
    "args": {
      "External id": 202601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202601,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202601, "pid": 5, "tid": 7, "ts": 1716454224611820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531777, "dur": 12,
    "args": {
      "External id": 202601,
      "cbid": 211, "correlation": 202601
    }
  },
  {
    "ph": "s", "id": 202601, "pid": 76337, "tid": -914061504, "ts": 1716454224531777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224611851, "dur": 35,
    "args": {
      "External id": 202623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202623,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202623, "pid": 5, "tid": 7, "ts": 1716454224611851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531808, "dur": 7,
    "args": {
      "External id": 202623,
      "cbid": 211, "correlation": 202623
    }
  },
  {
    "ph": "s", "id": 202623, "pid": 76337, "tid": -914061504, "ts": 1716454224531808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531891, "dur": 1,
    "args": {
      "External id": 202634,
      "cbid": 251, "correlation": 202634
    }
  },
  {
    "ph": "f", "id": 202634, "pid": 76337, "tid": -914061504, "ts": 1716454224531891,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224611888, "dur": 91,
    "args": {
      "External id": 202635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202635,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202635, "pid": 5, "tid": 7, "ts": 1716454224611888,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531896, "dur": 13,
    "args": {
      "External id": 202635,
      "cbid": 211, "correlation": 202635
    }
  },
  {
    "ph": "s", "id": 202635, "pid": 76337, "tid": -914061504, "ts": 1716454224531896,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531965, "dur": 1,
    "args": {
      "External id": 202646,
      "cbid": 251, "correlation": 202646
    }
  },
  {
    "ph": "f", "id": 202646, "pid": 76337, "tid": -914061504, "ts": 1716454224531965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531969, "dur": 0,
    "args": {
      "External id": 202647,
      "cbid": 251, "correlation": 202647
    }
  },
  {
    "ph": "f", "id": 202647, "pid": 76337, "tid": -914061504, "ts": 1716454224531969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224611980, "dur": 12,
    "args": {
      "External id": 202648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202648,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 202648, "pid": 5, "tid": 7, "ts": 1716454224611980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531971, "dur": 20,
    "args": {
      "External id": 202648,
      "cbid": 211, "correlation": 202648
    }
  },
  {
    "ph": "s", "id": 202648, "pid": 76337, "tid": -914061504, "ts": 1716454224531971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224611993, "dur": 5,
    "args": {
      "External id": 202650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202650,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 202650, "pid": 5, "tid": 7, "ts": 1716454224611993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224531992, "dur": 6,
    "args": {
      "External id": 202650,
      "cbid": 211, "correlation": 202650
    }
  },
  {
    "ph": "s", "id": 202650, "pid": 76337, "tid": -914061504, "ts": 1716454224531992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532052, "dur": 1,
    "args": {
      "External id": 202661,
      "cbid": 251, "correlation": 202661
    }
  },
  {
    "ph": "f", "id": 202661, "pid": 76337, "tid": -914061504, "ts": 1716454224532052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532055, "dur": 0,
    "args": {
      "External id": 202662,
      "cbid": 251, "correlation": 202662
    }
  },
  {
    "ph": "f", "id": 202662, "pid": 76337, "tid": -914061504, "ts": 1716454224532055,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224611999, "dur": 7,
    "args": {
      "External id": 202663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202663,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 202663, "pid": 5, "tid": 7, "ts": 1716454224611999,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532057, "dur": 12,
    "args": {
      "External id": 202663,
      "cbid": 211, "correlation": 202663
    }
  },
  {
    "ph": "s", "id": 202663, "pid": 76337, "tid": -914061504, "ts": 1716454224532057,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224612008, "dur": 4,
    "args": {
      "External id": 202665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202665,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 202665, "pid": 5, "tid": 7, "ts": 1716454224612008,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532071, "dur": 5,
    "args": {
      "External id": 202665,
      "cbid": 211, "correlation": 202665
    }
  },
  {
    "ph": "s", "id": 202665, "pid": 76337, "tid": -914061504, "ts": 1716454224532071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224612013, "dur": 92,
    "args": {
      "External id": 202686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202686,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 202686, "pid": 5, "tid": 7, "ts": 1716454224612013,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532143, "dur": 12,
    "args": {
      "External id": 202686,
      "cbid": 211, "correlation": 202686
    }
  },
  {
    "ph": "s", "id": 202686, "pid": 76337, "tid": -914061504, "ts": 1716454224532143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532240, "dur": 1,
    "args": {
      "External id": 202704,
      "cbid": 251, "correlation": 202704
    }
  },
  {
    "ph": "f", "id": 202704, "pid": 76337, "tid": -914061504, "ts": 1716454224532240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224612106, "dur": 97,
    "args": {
      "External id": 202706,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202706,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202706, "pid": 5, "tid": 7, "ts": 1716454224612106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532245, "dur": 13,
    "args": {
      "External id": 202706,
      "cbid": 211, "correlation": 202706
    }
  },
  {
    "ph": "s", "id": 202706, "pid": 76337, "tid": -914061504, "ts": 1716454224532245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224612204, "dur": 19,
    "args": {
      "External id": 202714,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202714,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202714, "pid": 5, "tid": 7, "ts": 1716454224612204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532314, "dur": 13,
    "args": {
      "External id": 202714,
      "cbid": 211, "correlation": 202714
    }
  },
  {
    "ph": "s", "id": 202714, "pid": 76337, "tid": -914061504, "ts": 1716454224532314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224612225, "dur": 37,
    "args": {
      "External id": 202722,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202722,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202722, "pid": 5, "tid": 7, "ts": 1716454224612225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532355, "dur": 9,
    "args": {
      "External id": 202722,
      "cbid": 211, "correlation": 202722
    }
  },
  {
    "ph": "s", "id": 202722, "pid": 76337, "tid": -914061504, "ts": 1716454224532355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224612263, "dur": 35,
    "args": {
      "External id": 202744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202744,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202744, "pid": 5, "tid": 7, "ts": 1716454224612263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532407, "dur": 10,
    "args": {
      "External id": 202744,
      "cbid": 211, "correlation": 202744
    }
  },
  {
    "ph": "s", "id": 202744, "pid": 76337, "tid": -914061504, "ts": 1716454224532407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532495, "dur": 1,
    "args": {
      "External id": 202760,
      "cbid": 251, "correlation": 202760
    }
  },
  {
    "ph": "f", "id": 202760, "pid": 76337, "tid": -914061504, "ts": 1716454224532495,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532499, "dur": 0,
    "args": {
      "External id": 202762,
      "cbid": 251, "correlation": 202762
    }
  },
  {
    "ph": "f", "id": 202762, "pid": 76337, "tid": -914061504, "ts": 1716454224532499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224612300, "dur": 543,
    "args": {
      "External id": 202763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202763,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 202763, "pid": 5, "tid": 7, "ts": 1716454224612300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532503, "dur": 12,
    "args": {
      "External id": 202763,
      "cbid": 211, "correlation": 202763
    }
  },
  {
    "ph": "s", "id": 202763, "pid": 76337, "tid": -914061504, "ts": 1716454224532503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224612844, "dur": 126,
    "args": {
      "External id": 202771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202771,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202771, "pid": 5, "tid": 7, "ts": 1716454224612844,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532567, "dur": 13,
    "args": {
      "External id": 202771,
      "cbid": 211, "correlation": 202771
    }
  },
  {
    "ph": "s", "id": 202771, "pid": 76337, "tid": -914061504, "ts": 1716454224532567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224612972, "dur": 127,
    "args": {
      "External id": 202779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202779, "pid": 5, "tid": 7, "ts": 1716454224612972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532598, "dur": 8,
    "args": {
      "External id": 202779,
      "cbid": 211, "correlation": 202779
    }
  },
  {
    "ph": "s", "id": 202779, "pid": 76337, "tid": -914061504, "ts": 1716454224532598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532675, "dur": 1,
    "args": {
      "External id": 202795,
      "cbid": 251, "correlation": 202795
    }
  },
  {
    "ph": "f", "id": 202795, "pid": 76337, "tid": -914061504, "ts": 1716454224532675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224613101, "dur": 306,
    "args": {
      "External id": 202797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202797,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202797, "pid": 5, "tid": 7, "ts": 1716454224613101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532681, "dur": 12,
    "args": {
      "External id": 202797,
      "cbid": 211, "correlation": 202797
    }
  },
  {
    "ph": "s", "id": 202797, "pid": 76337, "tid": -914061504, "ts": 1716454224532681,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224613408, "dur": 27,
    "args": {
      "External id": 202805,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202805,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202805, "pid": 5, "tid": 7, "ts": 1716454224613408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532724, "dur": 10,
    "args": {
      "External id": 202805,
      "cbid": 211, "correlation": 202805
    }
  },
  {
    "ph": "s", "id": 202805, "pid": 76337, "tid": -914061504, "ts": 1716454224532724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224613437, "dur": 82,
    "args": {
      "External id": 202816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202816,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202816, "pid": 5, "tid": 7, "ts": 1716454224613437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532791, "dur": 12,
    "args": {
      "External id": 202816,
      "cbid": 211, "correlation": 202816
    }
  },
  {
    "ph": "s", "id": 202816, "pid": 76337, "tid": -914061504, "ts": 1716454224532791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532853, "dur": 0,
    "args": {
      "External id": 202828,
      "cbid": 317, "correlation": 202828
    }
  },
  {
    "ph": "f", "id": 202828, "pid": 76337, "tid": -914061504, "ts": 1716454224532853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532854, "dur": 0,
    "args": {
      "External id": 202829,
      "cbid": 203, "correlation": 202829
    }
  },
  {
    "ph": "f", "id": 202829, "pid": 76337, "tid": -914061504, "ts": 1716454224532854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532855, "dur": 0,
    "args": {
      "External id": 202830,
      "cbid": 205, "correlation": 202830
    }
  },
  {
    "ph": "f", "id": 202830, "pid": 76337, "tid": -914061504, "ts": 1716454224532855,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224613520, "dur": 23,
    "args": {
      "External id": 202834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202834,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202834, "pid": 5, "tid": 7, "ts": 1716454224613520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532869, "dur": 13,
    "args": {
      "External id": 202834,
      "cbid": 211, "correlation": 202834
    }
  },
  {
    "ph": "s", "id": 202834, "pid": 76337, "tid": -914061504, "ts": 1716454224532869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224613545, "dur": 121,
    "args": {
      "External id": 202836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202836,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202836, "pid": 5, "tid": 7, "ts": 1716454224613545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532889, "dur": 6,
    "args": {
      "External id": 202836,
      "cbid": 211, "correlation": 202836
    }
  },
  {
    "ph": "s", "id": 202836, "pid": 76337, "tid": -914061504, "ts": 1716454224532889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224613667, "dur": 23,
    "args": {
      "External id": 202838,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202838,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202838, "pid": 5, "tid": 7, "ts": 1716454224613667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532899, "dur": 5,
    "args": {
      "External id": 202838,
      "cbid": 211, "correlation": 202838
    }
  },
  {
    "ph": "s", "id": 202838, "pid": 76337, "tid": -914061504, "ts": 1716454224532899,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224613692, "dur": 33,
    "args": {
      "External id": 202844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202844, "pid": 5, "tid": 7, "ts": 1716454224613692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532927, "dur": 8,
    "args": {
      "External id": 202844,
      "cbid": 211, "correlation": 202844
    }
  },
  {
    "ph": "s", "id": 202844, "pid": 76337, "tid": -914061504, "ts": 1716454224532927,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224613726, "dur": 27,
    "args": {
      "External id": 202852,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202852,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202852, "pid": 5, "tid": 7, "ts": 1716454224613726,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224532958, "dur": 8,
    "args": {
      "External id": 202852,
      "cbid": 211, "correlation": 202852
    }
  },
  {
    "ph": "s", "id": 202852, "pid": 76337, "tid": -914061504, "ts": 1716454224532958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224613754, "dur": 102,
    "args": {
      "External id": 202863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202863,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202863, "pid": 5, "tid": 7, "ts": 1716454224613754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533029, "dur": 12,
    "args": {
      "External id": 202863,
      "cbid": 211, "correlation": 202863
    }
  },
  {
    "ph": "s", "id": 202863, "pid": 76337, "tid": -914061504, "ts": 1716454224533029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533086, "dur": 0,
    "args": {
      "External id": 202873,
      "cbid": 317, "correlation": 202873
    }
  },
  {
    "ph": "f", "id": 202873, "pid": 76337, "tid": -914061504, "ts": 1716454224533086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533086, "dur": 0,
    "args": {
      "External id": 202874,
      "cbid": 203, "correlation": 202874
    }
  },
  {
    "ph": "f", "id": 202874, "pid": 76337, "tid": -914061504, "ts": 1716454224533086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533087, "dur": 0,
    "args": {
      "External id": 202875,
      "cbid": 205, "correlation": 202875
    }
  },
  {
    "ph": "f", "id": 202875, "pid": 76337, "tid": -914061504, "ts": 1716454224533087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224613858, "dur": 75,
    "args": {
      "External id": 202879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202879,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202879, "pid": 5, "tid": 7, "ts": 1716454224613858,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533101, "dur": 11,
    "args": {
      "External id": 202879,
      "cbid": 211, "correlation": 202879
    }
  },
  {
    "ph": "s", "id": 202879, "pid": 76337, "tid": -914061504, "ts": 1716454224533101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224613934, "dur": 45,
    "args": {
      "External id": 202881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202881, "pid": 5, "tid": 7, "ts": 1716454224613934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533115, "dur": 5,
    "args": {
      "External id": 202881,
      "cbid": 211, "correlation": 202881
    }
  },
  {
    "ph": "s", "id": 202881, "pid": 76337, "tid": -914061504, "ts": 1716454224533115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224613980, "dur": 4,
    "args": {
      "External id": 202883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202883,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 202883, "pid": 5, "tid": 7, "ts": 1716454224613980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533126, "dur": 6,
    "args": {
      "External id": 202883,
      "cbid": 211, "correlation": 202883
    }
  },
  {
    "ph": "s", "id": 202883, "pid": 76337, "tid": -914061504, "ts": 1716454224533126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533134, "dur": 0,
    "args": {
      "External id": 202884,
      "cbid": 51, "correlation": 202884
    }
  },
  {
    "ph": "s", "id": 202884, "pid": 76337, "tid": -914061504, "ts": 1716454224533134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224613986, "dur": 2239,
    "args": {
      "External id": 202885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202885,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202885, "pid": 5, "tid": 7, "ts": 1716454224613986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533135, "dur": 5,
    "args": {
      "External id": 202885,
      "cbid": 211, "correlation": 202885
    }
  },
  {
    "ph": "s", "id": 202885, "pid": 76337, "tid": -914061504, "ts": 1716454224533135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224616226, "dur": 113,
    "args": {
      "External id": 202890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202890,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202890, "pid": 5, "tid": 7, "ts": 1716454224616226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533163, "dur": 9,
    "args": {
      "External id": 202890,
      "cbid": 211, "correlation": 202890
    }
  },
  {
    "ph": "s", "id": 202890, "pid": 76337, "tid": -914061504, "ts": 1716454224533163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224616340, "dur": 168,
    "args": {
      "External id": 202899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202899,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202899, "pid": 5, "tid": 7, "ts": 1716454224616340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533254, "dur": 13,
    "args": {
      "External id": 202899,
      "cbid": 211, "correlation": 202899
    }
  },
  {
    "ph": "s", "id": 202899, "pid": 76337, "tid": -914061504, "ts": 1716454224533254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224616509, "dur": 131,
    "args": {
      "External id": 202919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202919,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 202919, "pid": 5, "tid": 7, "ts": 1716454224616509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533325, "dur": 11,
    "args": {
      "External id": 202919,
      "cbid": 211, "correlation": 202919
    }
  },
  {
    "ph": "s", "id": 202919, "pid": 76337, "tid": -914061504, "ts": 1716454224533325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224616642, "dur": 4,
    "args": {
      "External id": 202931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202931,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 202931, "pid": 5, "tid": 7, "ts": 1716454224616642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533346, "dur": 6,
    "args": {
      "External id": 202931,
      "cbid": 211, "correlation": 202931
    }
  },
  {
    "ph": "s", "id": 202931, "pid": 76337, "tid": -914061504, "ts": 1716454224533346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224616647, "dur": 159,
    "args": {
      "External id": 202934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202934,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202934, "pid": 5, "tid": 7, "ts": 1716454224616647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533364, "dur": 7,
    "args": {
      "External id": 202934,
      "cbid": 211, "correlation": 202934
    }
  },
  {
    "ph": "s", "id": 202934, "pid": 76337, "tid": -914061504, "ts": 1716454224533364,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224616808, "dur": 101,
    "args": {
      "External id": 202943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202943,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202943, "pid": 5, "tid": 7, "ts": 1716454224616808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533404, "dur": 10,
    "args": {
      "External id": 202943,
      "cbid": 211, "correlation": 202943
    }
  },
  {
    "ph": "s", "id": 202943, "pid": 76337, "tid": -914061504, "ts": 1716454224533404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533456, "dur": 0,
    "args": {
      "External id": 202953,
      "cbid": 317, "correlation": 202953
    }
  },
  {
    "ph": "f", "id": 202953, "pid": 76337, "tid": -914061504, "ts": 1716454224533456,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533457, "dur": 0,
    "args": {
      "External id": 202954,
      "cbid": 203, "correlation": 202954
    }
  },
  {
    "ph": "f", "id": 202954, "pid": 76337, "tid": -914061504, "ts": 1716454224533457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533458, "dur": 0,
    "args": {
      "External id": 202955,
      "cbid": 205, "correlation": 202955
    }
  },
  {
    "ph": "f", "id": 202955, "pid": 76337, "tid": -914061504, "ts": 1716454224533458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224616910, "dur": 111,
    "args": {
      "External id": 202959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202959,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202959, "pid": 5, "tid": 7, "ts": 1716454224616910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533475, "dur": 12,
    "args": {
      "External id": 202959,
      "cbid": 211, "correlation": 202959
    }
  },
  {
    "ph": "s", "id": 202959, "pid": 76337, "tid": -914061504, "ts": 1716454224533475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224617022, "dur": 34,
    "args": {
      "External id": 202961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202961,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202961, "pid": 5, "tid": 7, "ts": 1716454224617022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533489, "dur": 5,
    "args": {
      "External id": 202961,
      "cbid": 211, "correlation": 202961
    }
  },
  {
    "ph": "s", "id": 202961, "pid": 76337, "tid": -914061504, "ts": 1716454224533489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224617057, "dur": 4,
    "args": {
      "External id": 202963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202963,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 202963, "pid": 5, "tid": 7, "ts": 1716454224617057,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533499, "dur": 5,
    "args": {
      "External id": 202963,
      "cbid": 211, "correlation": 202963
    }
  },
  {
    "ph": "s", "id": 202963, "pid": 76337, "tid": -914061504, "ts": 1716454224533499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533508, "dur": 0,
    "args": {
      "External id": 202964,
      "cbid": 51, "correlation": 202964
    }
  },
  {
    "ph": "s", "id": 202964, "pid": 76337, "tid": -914061504, "ts": 1716454224533508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224617062, "dur": 2023,
    "args": {
      "External id": 202965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202965,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 202965, "pid": 5, "tid": 7, "ts": 1716454224617062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533509, "dur": 5,
    "args": {
      "External id": 202965,
      "cbid": 211, "correlation": 202965
    }
  },
  {
    "ph": "s", "id": 202965, "pid": 76337, "tid": -914061504, "ts": 1716454224533509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224619086, "dur": 58,
    "args": {
      "External id": 202970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202970,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 202970, "pid": 5, "tid": 7, "ts": 1716454224619086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533538, "dur": 8,
    "args": {
      "External id": 202970,
      "cbid": 211, "correlation": 202970
    }
  },
  {
    "ph": "s", "id": 202970, "pid": 76337, "tid": -914061504, "ts": 1716454224533538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224619146, "dur": 3,
    "args": {
      "External id": 202978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202978,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 202978, "pid": 5, "tid": 7, "ts": 1716454224619146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533581, "dur": 9,
    "args": {
      "External id": 202978,
      "cbid": 211, "correlation": 202978
    }
  },
  {
    "ph": "s", "id": 202978, "pid": 76337, "tid": -914061504, "ts": 1716454224533581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533645, "dur": 1,
    "args": {
      "External id": 202994,
      "cbid": 251, "correlation": 202994
    }
  },
  {
    "ph": "f", "id": 202994, "pid": 76337, "tid": -914061504, "ts": 1716454224533645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533650, "dur": 0,
    "args": {
      "External id": 202996,
      "cbid": 251, "correlation": 202996
    }
  },
  {
    "ph": "f", "id": 202996, "pid": 76337, "tid": -914061504, "ts": 1716454224533650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224619150, "dur": 12,
    "args": {
      "External id": 202997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202997,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 202997, "pid": 5, "tid": 7, "ts": 1716454224619150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533652, "dur": 11,
    "args": {
      "External id": 202997,
      "cbid": 211, "correlation": 202997
    }
  },
  {
    "ph": "s", "id": 202997, "pid": 76337, "tid": -914061504, "ts": 1716454224533652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224619163, "dur": 5,
    "args": {
      "External id": 202999,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 202999,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 202999, "pid": 5, "tid": 7, "ts": 1716454224619163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533665, "dur": 6,
    "args": {
      "External id": 202999,
      "cbid": 211, "correlation": 202999
    }
  },
  {
    "ph": "s", "id": 202999, "pid": 76337, "tid": -914061504, "ts": 1716454224533665,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224619170, "dur": 54,
    "args": {
      "External id": 203009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203009,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203009, "pid": 5, "tid": 7, "ts": 1716454224619170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533724, "dur": 12,
    "args": {
      "External id": 203009,
      "cbid": 211, "correlation": 203009
    }
  },
  {
    "ph": "s", "id": 203009, "pid": 76337, "tid": -914061504, "ts": 1716454224533724,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224619225, "dur": 51,
    "args": {
      "External id": 203029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203029,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 203029, "pid": 5, "tid": 7, "ts": 1716454224619225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533791, "dur": 11,
    "args": {
      "External id": 203029,
      "cbid": 211, "correlation": 203029
    }
  },
  {
    "ph": "s", "id": 203029, "pid": 76337, "tid": -914061504, "ts": 1716454224533791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224619278, "dur": 4,
    "args": {
      "External id": 203041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203041,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203041, "pid": 5, "tid": 7, "ts": 1716454224619278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533812, "dur": 6,
    "args": {
      "External id": 203041,
      "cbid": 211, "correlation": 203041
    }
  },
  {
    "ph": "s", "id": 203041, "pid": 76337, "tid": -914061504, "ts": 1716454224533812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224619283, "dur": 57,
    "args": {
      "External id": 203044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203044,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203044, "pid": 5, "tid": 7, "ts": 1716454224619283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533830, "dur": 6,
    "args": {
      "External id": 203044,
      "cbid": 211, "correlation": 203044
    }
  },
  {
    "ph": "s", "id": 203044, "pid": 76337, "tid": -914061504, "ts": 1716454224533830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224619341, "dur": 37,
    "args": {
      "External id": 203053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203053,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203053, "pid": 5, "tid": 7, "ts": 1716454224619341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533870, "dur": 11,
    "args": {
      "External id": 203053,
      "cbid": 211, "correlation": 203053
    }
  },
  {
    "ph": "s", "id": 203053, "pid": 76337, "tid": -914061504, "ts": 1716454224533870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533934, "dur": 0,
    "args": {
      "External id": 203063,
      "cbid": 317, "correlation": 203063
    }
  },
  {
    "ph": "f", "id": 203063, "pid": 76337, "tid": -914061504, "ts": 1716454224533934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533935, "dur": 0,
    "args": {
      "External id": 203064,
      "cbid": 203, "correlation": 203064
    }
  },
  {
    "ph": "f", "id": 203064, "pid": 76337, "tid": -914061504, "ts": 1716454224533935,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533936, "dur": 0,
    "args": {
      "External id": 203065,
      "cbid": 205, "correlation": 203065
    }
  },
  {
    "ph": "f", "id": 203065, "pid": 76337, "tid": -914061504, "ts": 1716454224533936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224619380, "dur": 40,
    "args": {
      "External id": 203069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203069,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203069, "pid": 5, "tid": 7, "ts": 1716454224619380,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533952, "dur": 12,
    "args": {
      "External id": 203069,
      "cbid": 211, "correlation": 203069
    }
  },
  {
    "ph": "s", "id": 203069, "pid": 76337, "tid": -914061504, "ts": 1716454224533952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224619421, "dur": 14,
    "args": {
      "External id": 203071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203071,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203071, "pid": 5, "tid": 7, "ts": 1716454224619421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533967, "dur": 13,
    "args": {
      "External id": 203071,
      "cbid": 211, "correlation": 203071
    }
  },
  {
    "ph": "s", "id": 203071, "pid": 76337, "tid": -914061504, "ts": 1716454224533967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224619437, "dur": 3,
    "args": {
      "External id": 203073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203073,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203073, "pid": 5, "tid": 7, "ts": 1716454224619437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533985, "dur": 6,
    "args": {
      "External id": 203073,
      "cbid": 211, "correlation": 203073
    }
  },
  {
    "ph": "s", "id": 203073, "pid": 76337, "tid": -914061504, "ts": 1716454224533985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533994, "dur": 0,
    "args": {
      "External id": 203074,
      "cbid": 51, "correlation": 203074
    }
  },
  {
    "ph": "s", "id": 203074, "pid": 76337, "tid": -914061504, "ts": 1716454224533994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224619441, "dur": 703,
    "args": {
      "External id": 203075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203075,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203075, "pid": 5, "tid": 7, "ts": 1716454224619441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224533995, "dur": 5,
    "args": {
      "External id": 203075,
      "cbid": 211, "correlation": 203075
    }
  },
  {
    "ph": "s", "id": 203075, "pid": 76337, "tid": -914061504, "ts": 1716454224533995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224620145, "dur": 60,
    "args": {
      "External id": 203080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203080,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203080, "pid": 5, "tid": 7, "ts": 1716454224620145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534023, "dur": 9,
    "args": {
      "External id": 203080,
      "cbid": 211, "correlation": 203080
    }
  },
  {
    "ph": "s", "id": 203080, "pid": 76337, "tid": -914061504, "ts": 1716454224534023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534083, "dur": 0,
    "args": {
      "External id": 203090,
      "cbid": 317, "correlation": 203090
    }
  },
  {
    "ph": "f", "id": 203090, "pid": 76337, "tid": -914061504, "ts": 1716454224534083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534083, "dur": 0,
    "args": {
      "External id": 203091,
      "cbid": 203, "correlation": 203091
    }
  },
  {
    "ph": "f", "id": 203091, "pid": 76337, "tid": -914061504, "ts": 1716454224534083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534084, "dur": 0,
    "args": {
      "External id": 203092,
      "cbid": 205, "correlation": 203092
    }
  },
  {
    "ph": "f", "id": 203092, "pid": 76337, "tid": -914061504, "ts": 1716454224534084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224620206, "dur": 4,
    "args": {
      "External id": 203096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203096,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203096, "pid": 5, "tid": 7, "ts": 1716454224620206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534099, "dur": 11,
    "args": {
      "External id": 203096,
      "cbid": 211, "correlation": 203096
    }
  },
  {
    "ph": "s", "id": 203096, "pid": 76337, "tid": -914061504, "ts": 1716454224534099,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534115, "dur": 0,
    "args": {
      "External id": 203097,
      "cbid": 51, "correlation": 203097
    }
  },
  {
    "ph": "s", "id": 203097, "pid": 76337, "tid": -914061504, "ts": 1716454224534115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224620211, "dur": 268,
    "args": {
      "External id": 203098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203098,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203098, "pid": 5, "tid": 7, "ts": 1716454224620211,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534116, "dur": 8,
    "args": {
      "External id": 203098,
      "cbid": 211, "correlation": 203098
    }
  },
  {
    "ph": "s", "id": 203098, "pid": 76337, "tid": -914061504, "ts": 1716454224534116,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224620481, "dur": 59,
    "args": {
      "External id": 203103,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203103,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203103, "pid": 5, "tid": 7, "ts": 1716454224620481,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534145, "dur": 8,
    "args": {
      "External id": 203103,
      "cbid": 211, "correlation": 203103
    }
  },
  {
    "ph": "s", "id": 203103, "pid": 76337, "tid": -914061504, "ts": 1716454224534145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224620542, "dur": 51,
    "args": {
      "External id": 203111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203111,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203111, "pid": 5, "tid": 7, "ts": 1716454224620542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534173, "dur": 8,
    "args": {
      "External id": 203111,
      "cbid": 211, "correlation": 203111
    }
  },
  {
    "ph": "s", "id": 203111, "pid": 76337, "tid": -914061504, "ts": 1716454224534173,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224620594, "dur": 35,
    "args": {
      "External id": 203119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203119,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203119, "pid": 5, "tid": 7, "ts": 1716454224620594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534202, "dur": 8,
    "args": {
      "External id": 203119,
      "cbid": 211, "correlation": 203119
    }
  },
  {
    "ph": "s", "id": 203119, "pid": 76337, "tid": -914061504, "ts": 1716454224534202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224620631, "dur": 51,
    "args": {
      "External id": 203139,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203139,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 203139, "pid": 5, "tid": 7, "ts": 1716454224620631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534285, "dur": 12,
    "args": {
      "External id": 203139,
      "cbid": 211, "correlation": 203139
    }
  },
  {
    "ph": "s", "id": 203139, "pid": 76337, "tid": -914061504, "ts": 1716454224534285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224620683, "dur": 4,
    "args": {
      "External id": 203151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203151,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203151, "pid": 5, "tid": 7, "ts": 1716454224620683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534307, "dur": 6,
    "args": {
      "External id": 203151,
      "cbid": 211, "correlation": 203151
    }
  },
  {
    "ph": "s", "id": 203151, "pid": 76337, "tid": -914061504, "ts": 1716454224534307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224620688, "dur": 56,
    "args": {
      "External id": 203154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203154,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203154, "pid": 5, "tid": 7, "ts": 1716454224620688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534325, "dur": 6,
    "args": {
      "External id": 203154,
      "cbid": 211, "correlation": 203154
    }
  },
  {
    "ph": "s", "id": 203154, "pid": 76337, "tid": -914061504, "ts": 1716454224534325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534382, "dur": 0,
    "args": {
      "External id": 203165,
      "cbid": 317, "correlation": 203165
    }
  },
  {
    "ph": "f", "id": 203165, "pid": 76337, "tid": -914061504, "ts": 1716454224534382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534382, "dur": 0,
    "args": {
      "External id": 203166,
      "cbid": 203, "correlation": 203166
    }
  },
  {
    "ph": "f", "id": 203166, "pid": 76337, "tid": -914061504, "ts": 1716454224534382,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534383, "dur": 0,
    "args": {
      "External id": 203167,
      "cbid": 205, "correlation": 203167
    }
  },
  {
    "ph": "f", "id": 203167, "pid": 76337, "tid": -914061504, "ts": 1716454224534383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534415, "dur": 2,
    "args": {
      "External id": 203171,
      "cbid": 251, "correlation": 203171
    }
  },
  {
    "ph": "f", "id": 203171, "pid": 76337, "tid": -914061504, "ts": 1716454224534415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534419, "dur": 1,
    "args": {
      "External id": 203172,
      "cbid": 251, "correlation": 203172
    }
  },
  {
    "ph": "f", "id": 203172, "pid": 76337, "tid": -914061504, "ts": 1716454224534419,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534420, "dur": 0,
    "args": {
      "External id": 203173,
      "cbid": 251, "correlation": 203173
    }
  },
  {
    "ph": "f", "id": 203173, "pid": 76337, "tid": -914061504, "ts": 1716454224534420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534422, "dur": 1,
    "args": {
      "External id": 203174,
      "cbid": 251, "correlation": 203174
    }
  },
  {
    "ph": "f", "id": 203174, "pid": 76337, "tid": -914061504, "ts": 1716454224534422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534423, "dur": 1,
    "args": {
      "External id": 203175,
      "cbid": 251, "correlation": 203175
    }
  },
  {
    "ph": "f", "id": 203175, "pid": 76337, "tid": -914061504, "ts": 1716454224534423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534425, "dur": 1,
    "args": {
      "External id": 203176,
      "cbid": 251, "correlation": 203176
    }
  },
  {
    "ph": "f", "id": 203176, "pid": 76337, "tid": -914061504, "ts": 1716454224534425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534427, "dur": 1,
    "args": {
      "External id": 203177,
      "cbid": 251, "correlation": 203177
    }
  },
  {
    "ph": "f", "id": 203177, "pid": 76337, "tid": -914061504, "ts": 1716454224534427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534429, "dur": 1,
    "args": {
      "External id": 203178,
      "cbid": 251, "correlation": 203178
    }
  },
  {
    "ph": "f", "id": 203178, "pid": 76337, "tid": -914061504, "ts": 1716454224534429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534432, "dur": 0,
    "args": {
      "External id": 203179,
      "cbid": 251, "correlation": 203179
    }
  },
  {
    "ph": "f", "id": 203179, "pid": 76337, "tid": -914061504, "ts": 1716454224534432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224620745, "dur": 114,
    "args": {
      "External id": 203180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203180,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203180, "pid": 5, "tid": 7, "ts": 1716454224620745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534434, "dur": 13,
    "args": {
      "External id": 203180,
      "cbid": 211, "correlation": 203180
    }
  },
  {
    "ph": "s", "id": 203180, "pid": 76337, "tid": -914061504, "ts": 1716454224534434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224620861, "dur": 59,
    "args": {
      "External id": 203186,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203186,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203186, "pid": 5, "tid": 7, "ts": 1716454224620861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534471, "dur": 9,
    "args": {
      "External id": 203186,
      "cbid": 211, "correlation": 203186
    }
  },
  {
    "ph": "s", "id": 203186, "pid": 76337, "tid": -914061504, "ts": 1716454224534471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224620922, "dur": 568,
    "args": {
      "External id": 203195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203195,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203195, "pid": 5, "tid": 7, "ts": 1716454224620922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534553, "dur": 14,
    "args": {
      "External id": 203195,
      "cbid": 211, "correlation": 203195
    }
  },
  {
    "ph": "s", "id": 203195, "pid": 76337, "tid": -914061504, "ts": 1716454224534553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224621491, "dur": 183,
    "args": {
      "External id": 203217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203217,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203217, "pid": 5, "tid": 7, "ts": 1716454224621491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534616, "dur": 12,
    "args": {
      "External id": 203217,
      "cbid": 211, "correlation": 203217
    }
  },
  {
    "ph": "s", "id": 203217, "pid": 76337, "tid": -914061504, "ts": 1716454224534616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534706, "dur": 1,
    "args": {
      "External id": 203228,
      "cbid": 251, "correlation": 203228
    }
  },
  {
    "ph": "f", "id": 203228, "pid": 76337, "tid": -914061504, "ts": 1716454224534706,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224621675, "dur": 199,
    "args": {
      "External id": 203229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203229,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203229, "pid": 5, "tid": 7, "ts": 1716454224621675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534711, "dur": 14,
    "args": {
      "External id": 203229,
      "cbid": 211, "correlation": 203229
    }
  },
  {
    "ph": "s", "id": 203229, "pid": 76337, "tid": -914061504, "ts": 1716454224534711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534783, "dur": 1,
    "args": {
      "External id": 203240,
      "cbid": 251, "correlation": 203240
    }
  },
  {
    "ph": "f", "id": 203240, "pid": 76337, "tid": -914061504, "ts": 1716454224534783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224621875, "dur": 188,
    "args": {
      "External id": 203241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203241,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203241, "pid": 5, "tid": 7, "ts": 1716454224621875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534787, "dur": 11,
    "args": {
      "External id": 203241,
      "cbid": 211, "correlation": 203241
    }
  },
  {
    "ph": "s", "id": 203241, "pid": 76337, "tid": -914061504, "ts": 1716454224534787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534850, "dur": 1,
    "args": {
      "External id": 203252,
      "cbid": 251, "correlation": 203252
    }
  },
  {
    "ph": "f", "id": 203252, "pid": 76337, "tid": -914061504, "ts": 1716454224534850,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224622065, "dur": 187,
    "args": {
      "External id": 203253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203253,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203253, "pid": 5, "tid": 7, "ts": 1716454224622065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534854, "dur": 11,
    "args": {
      "External id": 203253,
      "cbid": 211, "correlation": 203253
    }
  },
  {
    "ph": "s", "id": 203253, "pid": 76337, "tid": -914061504, "ts": 1716454224534854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224622254, "dur": 18715,
    "args": {
      "External id": 203274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203274,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203274, "pid": 5, "tid": 7, "ts": 1716454224622254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224534938, "dur": 14,
    "args": {
      "External id": 203274,
      "cbid": 211, "correlation": 203274
    }
  },
  {
    "ph": "s", "id": 203274, "pid": 76337, "tid": -914061504, "ts": 1716454224534938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535044, "dur": 1,
    "args": {
      "External id": 203292,
      "cbid": 251, "correlation": 203292
    }
  },
  {
    "ph": "f", "id": 203292, "pid": 76337, "tid": -914061504, "ts": 1716454224535044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224640970, "dur": 203,
    "args": {
      "External id": 203294,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203294,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203294, "pid": 5, "tid": 7, "ts": 1716454224640970,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535050, "dur": 13,
    "args": {
      "External id": 203294,
      "cbid": 211, "correlation": 203294
    }
  },
  {
    "ph": "s", "id": 203294, "pid": 76337, "tid": -914061504, "ts": 1716454224535050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224641175, "dur": 67,
    "args": {
      "External id": 203302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203302,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203302, "pid": 5, "tid": 7, "ts": 1716454224641175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535119, "dur": 13,
    "args": {
      "External id": 203302,
      "cbid": 211, "correlation": 203302
    }
  },
  {
    "ph": "s", "id": 203302, "pid": 76337, "tid": -914061504, "ts": 1716454224535119,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224641243, "dur": 97,
    "args": {
      "External id": 203310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203310,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203310, "pid": 5, "tid": 7, "ts": 1716454224641243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535158, "dur": 9,
    "args": {
      "External id": 203310,
      "cbid": 211, "correlation": 203310
    }
  },
  {
    "ph": "s", "id": 203310, "pid": 76337, "tid": -914061504, "ts": 1716454224535158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224641341, "dur": 54,
    "args": {
      "External id": 203321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203321,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203321, "pid": 5, "tid": 7, "ts": 1716454224641341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535232, "dur": 12,
    "args": {
      "External id": 203321,
      "cbid": 211, "correlation": 203321
    }
  },
  {
    "ph": "s", "id": 203321, "pid": 76337, "tid": -914061504, "ts": 1716454224535232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224641397, "dur": 93,
    "args": {
      "External id": 203343,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203343,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203343, "pid": 5, "tid": 7, "ts": 1716454224641397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535263, "dur": 8,
    "args": {
      "External id": 203343,
      "cbid": 211, "correlation": 203343
    }
  },
  {
    "ph": "s", "id": 203343, "pid": 76337, "tid": -914061504, "ts": 1716454224535263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535346, "dur": 1,
    "args": {
      "External id": 203354,
      "cbid": 251, "correlation": 203354
    }
  },
  {
    "ph": "f", "id": 203354, "pid": 76337, "tid": -914061504, "ts": 1716454224535346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224641491, "dur": 104,
    "args": {
      "External id": 203355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203355,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203355, "pid": 5, "tid": 7, "ts": 1716454224641491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535351, "dur": 13,
    "args": {
      "External id": 203355,
      "cbid": 211, "correlation": 203355
    }
  },
  {
    "ph": "s", "id": 203355, "pid": 76337, "tid": -914061504, "ts": 1716454224535351,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535425, "dur": 1,
    "args": {
      "External id": 203366,
      "cbid": 251, "correlation": 203366
    }
  },
  {
    "ph": "f", "id": 203366, "pid": 76337, "tid": -914061504, "ts": 1716454224535425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535429, "dur": 0,
    "args": {
      "External id": 203367,
      "cbid": 251, "correlation": 203367
    }
  },
  {
    "ph": "f", "id": 203367, "pid": 76337, "tid": -914061504, "ts": 1716454224535429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224641596, "dur": 10,
    "args": {
      "External id": 203368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203368,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 203368, "pid": 5, "tid": 7, "ts": 1716454224641596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535431, "dur": 13,
    "args": {
      "External id": 203368,
      "cbid": 211, "correlation": 203368
    }
  },
  {
    "ph": "s", "id": 203368, "pid": 76337, "tid": -914061504, "ts": 1716454224535431,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224641608, "dur": 5,
    "args": {
      "External id": 203370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203370,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 203370, "pid": 5, "tid": 7, "ts": 1716454224641608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535448, "dur": 7,
    "args": {
      "External id": 203370,
      "cbid": 211, "correlation": 203370
    }
  },
  {
    "ph": "s", "id": 203370, "pid": 76337, "tid": -914061504, "ts": 1716454224535448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535511, "dur": 1,
    "args": {
      "External id": 203381,
      "cbid": 251, "correlation": 203381
    }
  },
  {
    "ph": "f", "id": 203381, "pid": 76337, "tid": -914061504, "ts": 1716454224535511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535514, "dur": 0,
    "args": {
      "External id": 203382,
      "cbid": 251, "correlation": 203382
    }
  },
  {
    "ph": "f", "id": 203382, "pid": 76337, "tid": -914061504, "ts": 1716454224535514,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224641614, "dur": 6,
    "args": {
      "External id": 203383,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203383,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 203383, "pid": 5, "tid": 7, "ts": 1716454224641614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535515, "dur": 13,
    "args": {
      "External id": 203383,
      "cbid": 211, "correlation": 203383
    }
  },
  {
    "ph": "s", "id": 203383, "pid": 76337, "tid": -914061504, "ts": 1716454224535515,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224641622, "dur": 3,
    "args": {
      "External id": 203385,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203385,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 203385, "pid": 5, "tid": 7, "ts": 1716454224641622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535530, "dur": 5,
    "args": {
      "External id": 203385,
      "cbid": 211, "correlation": 203385
    }
  },
  {
    "ph": "s", "id": 203385, "pid": 76337, "tid": -914061504, "ts": 1716454224535530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224641627, "dur": 158,
    "args": {
      "External id": 203406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203406,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203406, "pid": 5, "tid": 7, "ts": 1716454224641627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535602, "dur": 12,
    "args": {
      "External id": 203406,
      "cbid": 211, "correlation": 203406
    }
  },
  {
    "ph": "s", "id": 203406, "pid": 76337, "tid": -914061504, "ts": 1716454224535602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535698, "dur": 1,
    "args": {
      "External id": 203424,
      "cbid": 251, "correlation": 203424
    }
  },
  {
    "ph": "f", "id": 203424, "pid": 76337, "tid": -914061504, "ts": 1716454224535698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224641786, "dur": 108,
    "args": {
      "External id": 203426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203426,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203426, "pid": 5, "tid": 7, "ts": 1716454224641786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535704, "dur": 14,
    "args": {
      "External id": 203426,
      "cbid": 211, "correlation": 203426
    }
  },
  {
    "ph": "s", "id": 203426, "pid": 76337, "tid": -914061504, "ts": 1716454224535704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224641895, "dur": 35,
    "args": {
      "External id": 203434,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203434,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203434, "pid": 5, "tid": 7, "ts": 1716454224641895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535773, "dur": 13,
    "args": {
      "External id": 203434,
      "cbid": 211, "correlation": 203434
    }
  },
  {
    "ph": "s", "id": 203434, "pid": 76337, "tid": -914061504, "ts": 1716454224535773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224641931, "dur": 66,
    "args": {
      "External id": 203442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203442,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203442, "pid": 5, "tid": 7, "ts": 1716454224641931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535814, "dur": 9,
    "args": {
      "External id": 203442,
      "cbid": 211, "correlation": 203442
    }
  },
  {
    "ph": "s", "id": 203442, "pid": 76337, "tid": -914061504, "ts": 1716454224535814,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224641998, "dur": 93,
    "args": {
      "External id": 203464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203464,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203464, "pid": 5, "tid": 7, "ts": 1716454224641998,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535865, "dur": 10,
    "args": {
      "External id": 203464,
      "cbid": 211, "correlation": 203464
    }
  },
  {
    "ph": "s", "id": 203464, "pid": 76337, "tid": -914061504, "ts": 1716454224535865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535951, "dur": 1,
    "args": {
      "External id": 203480,
      "cbid": 251, "correlation": 203480
    }
  },
  {
    "ph": "f", "id": 203480, "pid": 76337, "tid": -914061504, "ts": 1716454224535951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224642092, "dur": 577,
    "args": {
      "External id": 203482,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203482,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203482, "pid": 5, "tid": 7, "ts": 1716454224642092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224535956, "dur": 12,
    "args": {
      "External id": 203482,
      "cbid": 211, "correlation": 203482
    }
  },
  {
    "ph": "s", "id": 203482, "pid": 76337, "tid": -914061504, "ts": 1716454224535956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224642671, "dur": 245,
    "args": {
      "External id": 203490,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203490,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203490, "pid": 5, "tid": 7, "ts": 1716454224642671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536029, "dur": 13,
    "args": {
      "External id": 203490,
      "cbid": 211, "correlation": 203490
    }
  },
  {
    "ph": "s", "id": 203490, "pid": 76337, "tid": -914061504, "ts": 1716454224536029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224642918, "dur": 253,
    "args": {
      "External id": 203498,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203498,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203498, "pid": 5, "tid": 7, "ts": 1716454224642918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536060, "dur": 8,
    "args": {
      "External id": 203498,
      "cbid": 211, "correlation": 203498
    }
  },
  {
    "ph": "s", "id": 203498, "pid": 76337, "tid": -914061504, "ts": 1716454224536060,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536143, "dur": 1,
    "args": {
      "External id": 203514,
      "cbid": 251, "correlation": 203514
    }
  },
  {
    "ph": "f", "id": 203514, "pid": 76337, "tid": -914061504, "ts": 1716454224536143,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536148, "dur": 0,
    "args": {
      "External id": 203516,
      "cbid": 251, "correlation": 203516
    }
  },
  {
    "ph": "f", "id": 203516, "pid": 76337, "tid": -914061504, "ts": 1716454224536148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224643172, "dur": 360,
    "args": {
      "External id": 203517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203517,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 203517, "pid": 5, "tid": 7, "ts": 1716454224643172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536151, "dur": 12,
    "args": {
      "External id": 203517,
      "cbid": 211, "correlation": 203517
    }
  },
  {
    "ph": "s", "id": 203517, "pid": 76337, "tid": -914061504, "ts": 1716454224536151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224643533, "dur": 50,
    "args": {
      "External id": 203525,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203525,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203525, "pid": 5, "tid": 7, "ts": 1716454224643533,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536193, "dur": 10,
    "args": {
      "External id": 203525,
      "cbid": 211, "correlation": 203525
    }
  },
  {
    "ph": "s", "id": 203525, "pid": 76337, "tid": -914061504, "ts": 1716454224536193,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224643584, "dur": 159,
    "args": {
      "External id": 203536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203536, "pid": 5, "tid": 7, "ts": 1716454224643584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536260, "dur": 12,
    "args": {
      "External id": 203536,
      "cbid": 211, "correlation": 203536
    }
  },
  {
    "ph": "s", "id": 203536, "pid": 76337, "tid": -914061504, "ts": 1716454224536260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536324, "dur": 0,
    "args": {
      "External id": 203548,
      "cbid": 317, "correlation": 203548
    }
  },
  {
    "ph": "f", "id": 203548, "pid": 76337, "tid": -914061504, "ts": 1716454224536324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536325, "dur": 0,
    "args": {
      "External id": 203549,
      "cbid": 203, "correlation": 203549
    }
  },
  {
    "ph": "f", "id": 203549, "pid": 76337, "tid": -914061504, "ts": 1716454224536325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536326, "dur": 0,
    "args": {
      "External id": 203550,
      "cbid": 205, "correlation": 203550
    }
  },
  {
    "ph": "f", "id": 203550, "pid": 76337, "tid": -914061504, "ts": 1716454224536326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536350, "dur": 1,
    "args": {
      "External id": 203554,
      "cbid": 251, "correlation": 203554
    }
  },
  {
    "ph": "f", "id": 203554, "pid": 76337, "tid": -914061504, "ts": 1716454224536350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536352, "dur": 0,
    "args": {
      "External id": 203555,
      "cbid": 251, "correlation": 203555
    }
  },
  {
    "ph": "f", "id": 203555, "pid": 76337, "tid": -914061504, "ts": 1716454224536352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536353, "dur": 0,
    "args": {
      "External id": 203556,
      "cbid": 251, "correlation": 203556
    }
  },
  {
    "ph": "f", "id": 203556, "pid": 76337, "tid": -914061504, "ts": 1716454224536353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536353, "dur": 0,
    "args": {
      "External id": 203557,
      "cbid": 251, "correlation": 203557
    }
  },
  {
    "ph": "f", "id": 203557, "pid": 76337, "tid": -914061504, "ts": 1716454224536353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536354, "dur": 0,
    "args": {
      "External id": 203558,
      "cbid": 251, "correlation": 203558
    }
  },
  {
    "ph": "f", "id": 203558, "pid": 76337, "tid": -914061504, "ts": 1716454224536354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536355, "dur": 0,
    "args": {
      "External id": 203559,
      "cbid": 251, "correlation": 203559
    }
  },
  {
    "ph": "f", "id": 203559, "pid": 76337, "tid": -914061504, "ts": 1716454224536355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536356, "dur": 0,
    "args": {
      "External id": 203560,
      "cbid": 251, "correlation": 203560
    }
  },
  {
    "ph": "f", "id": 203560, "pid": 76337, "tid": -914061504, "ts": 1716454224536356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536357, "dur": 0,
    "args": {
      "External id": 203561,
      "cbid": 251, "correlation": 203561
    }
  },
  {
    "ph": "f", "id": 203561, "pid": 76337, "tid": -914061504, "ts": 1716454224536357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536358, "dur": 0,
    "args": {
      "External id": 203562,
      "cbid": 251, "correlation": 203562
    }
  },
  {
    "ph": "f", "id": 203562, "pid": 76337, "tid": -914061504, "ts": 1716454224536358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224643745, "dur": 115,
    "args": {
      "External id": 203563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203563,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203563, "pid": 5, "tid": 7, "ts": 1716454224643745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536360, "dur": 12,
    "args": {
      "External id": 203563,
      "cbid": 211, "correlation": 203563
    }
  },
  {
    "ph": "s", "id": 203563, "pid": 76337, "tid": -914061504, "ts": 1716454224536360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224643861, "dur": 60,
    "args": {
      "External id": 203569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203569,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203569, "pid": 5, "tid": 7, "ts": 1716454224643861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536396, "dur": 8,
    "args": {
      "External id": 203569,
      "cbid": 211, "correlation": 203569
    }
  },
  {
    "ph": "s", "id": 203569, "pid": 76337, "tid": -914061504, "ts": 1716454224536396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224643922, "dur": 50,
    "args": {
      "External id": 203577,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203577,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203577, "pid": 5, "tid": 7, "ts": 1716454224643922,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536427, "dur": 9,
    "args": {
      "External id": 203577,
      "cbid": 211, "correlation": 203577
    }
  },
  {
    "ph": "s", "id": 203577, "pid": 76337, "tid": -914061504, "ts": 1716454224536427,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224643973, "dur": 98,
    "args": {
      "External id": 203586,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203586,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203586, "pid": 5, "tid": 7, "ts": 1716454224643973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536466, "dur": 10,
    "args": {
      "External id": 203586,
      "cbid": 211, "correlation": 203586
    }
  },
  {
    "ph": "s", "id": 203586, "pid": 76337, "tid": -914061504, "ts": 1716454224536466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224644072, "dur": 92,
    "args": {
      "External id": 203606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203606,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 203606, "pid": 5, "tid": 7, "ts": 1716454224644072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536539, "dur": 11,
    "args": {
      "External id": 203606,
      "cbid": 211, "correlation": 203606
    }
  },
  {
    "ph": "s", "id": 203606, "pid": 76337, "tid": -914061504, "ts": 1716454224536539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224644165, "dur": 4,
    "args": {
      "External id": 203618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203618,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 203618, "pid": 5, "tid": 7, "ts": 1716454224644165,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536560, "dur": 6,
    "args": {
      "External id": 203618,
      "cbid": 211, "correlation": 203618
    }
  },
  {
    "ph": "s", "id": 203618, "pid": 76337, "tid": -914061504, "ts": 1716454224536560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224644171, "dur": 108,
    "args": {
      "External id": 203621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203621,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203621, "pid": 5, "tid": 7, "ts": 1716454224644171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536579, "dur": 7,
    "args": {
      "External id": 203621,
      "cbid": 211, "correlation": 203621
    }
  },
  {
    "ph": "s", "id": 203621, "pid": 76337, "tid": -914061504, "ts": 1716454224536579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224644281, "dur": 70,
    "args": {
      "External id": 203630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203630,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203630, "pid": 5, "tid": 7, "ts": 1716454224644281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536618, "dur": 10,
    "args": {
      "External id": 203630,
      "cbid": 211, "correlation": 203630
    }
  },
  {
    "ph": "s", "id": 203630, "pid": 76337, "tid": -914061504, "ts": 1716454224536618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536669, "dur": 0,
    "args": {
      "External id": 203640,
      "cbid": 317, "correlation": 203640
    }
  },
  {
    "ph": "f", "id": 203640, "pid": 76337, "tid": -914061504, "ts": 1716454224536669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536670, "dur": 0,
    "args": {
      "External id": 203641,
      "cbid": 203, "correlation": 203641
    }
  },
  {
    "ph": "f", "id": 203641, "pid": 76337, "tid": -914061504, "ts": 1716454224536670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536671, "dur": 0,
    "args": {
      "External id": 203642,
      "cbid": 205, "correlation": 203642
    }
  },
  {
    "ph": "f", "id": 203642, "pid": 76337, "tid": -914061504, "ts": 1716454224536671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224644352, "dur": 75,
    "args": {
      "External id": 203646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203646,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203646, "pid": 5, "tid": 7, "ts": 1716454224644352,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536687, "dur": 11,
    "args": {
      "External id": 203646,
      "cbid": 211, "correlation": 203646
    }
  },
  {
    "ph": "s", "id": 203646, "pid": 76337, "tid": -914061504, "ts": 1716454224536687,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224644429, "dur": 24,
    "args": {
      "External id": 203648,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203648,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203648, "pid": 5, "tid": 7, "ts": 1716454224644429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536701, "dur": 5,
    "args": {
      "External id": 203648,
      "cbid": 211, "correlation": 203648
    }
  },
  {
    "ph": "s", "id": 203648, "pid": 76337, "tid": -914061504, "ts": 1716454224536701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224644454, "dur": 4,
    "args": {
      "External id": 203650,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203650,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203650, "pid": 5, "tid": 7, "ts": 1716454224644454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536712, "dur": 5,
    "args": {
      "External id": 203650,
      "cbid": 211, "correlation": 203650
    }
  },
  {
    "ph": "s", "id": 203650, "pid": 76337, "tid": -914061504, "ts": 1716454224536712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536720, "dur": 0,
    "args": {
      "External id": 203651,
      "cbid": 51, "correlation": 203651
    }
  },
  {
    "ph": "s", "id": 203651, "pid": 76337, "tid": -914061504, "ts": 1716454224536720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224644459, "dur": 1371,
    "args": {
      "External id": 203652,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203652,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203652, "pid": 5, "tid": 7, "ts": 1716454224644459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536721, "dur": 5,
    "args": {
      "External id": 203652,
      "cbid": 211, "correlation": 203652
    }
  },
  {
    "ph": "s", "id": 203652, "pid": 76337, "tid": -914061504, "ts": 1716454224536721,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224645832, "dur": 59,
    "args": {
      "External id": 203657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203657,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203657, "pid": 5, "tid": 7, "ts": 1716454224645832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536749, "dur": 8,
    "args": {
      "External id": 203657,
      "cbid": 211, "correlation": 203657
    }
  },
  {
    "ph": "s", "id": 203657, "pid": 76337, "tid": -914061504, "ts": 1716454224536749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224645893, "dur": 3,
    "args": {
      "External id": 203665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203665,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203665, "pid": 5, "tid": 7, "ts": 1716454224645893,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536792, "dur": 9,
    "args": {
      "External id": 203665,
      "cbid": 211, "correlation": 203665
    }
  },
  {
    "ph": "s", "id": 203665, "pid": 76337, "tid": -914061504, "ts": 1716454224536792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536857, "dur": 1,
    "args": {
      "External id": 203681,
      "cbid": 251, "correlation": 203681
    }
  },
  {
    "ph": "f", "id": 203681, "pid": 76337, "tid": -914061504, "ts": 1716454224536857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536862, "dur": 0,
    "args": {
      "External id": 203683,
      "cbid": 251, "correlation": 203683
    }
  },
  {
    "ph": "f", "id": 203683, "pid": 76337, "tid": -914061504, "ts": 1716454224536862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224645898, "dur": 11,
    "args": {
      "External id": 203684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203684,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 203684, "pid": 5, "tid": 7, "ts": 1716454224645898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536864, "dur": 11,
    "args": {
      "External id": 203684,
      "cbid": 211, "correlation": 203684
    }
  },
  {
    "ph": "s", "id": 203684, "pid": 76337, "tid": -914061504, "ts": 1716454224536864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224645910, "dur": 5,
    "args": {
      "External id": 203686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203686,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 203686, "pid": 5, "tid": 7, "ts": 1716454224645910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536878, "dur": 6,
    "args": {
      "External id": 203686,
      "cbid": 211, "correlation": 203686
    }
  },
  {
    "ph": "s", "id": 203686, "pid": 76337, "tid": -914061504, "ts": 1716454224536878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224645917, "dur": 55,
    "args": {
      "External id": 203696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203696,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203696, "pid": 5, "tid": 7, "ts": 1716454224645917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224536936, "dur": 12,
    "args": {
      "External id": 203696,
      "cbid": 211, "correlation": 203696
    }
  },
  {
    "ph": "s", "id": 203696, "pid": 76337, "tid": -914061504, "ts": 1716454224536936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224645973, "dur": 52,
    "args": {
      "External id": 203716,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203716,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 203716, "pid": 5, "tid": 7, "ts": 1716454224645973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537010, "dur": 12,
    "args": {
      "External id": 203716,
      "cbid": 211, "correlation": 203716
    }
  },
  {
    "ph": "s", "id": 203716, "pid": 76337, "tid": -914061504, "ts": 1716454224537010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224646027, "dur": 4,
    "args": {
      "External id": 203728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203728,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203728, "pid": 5, "tid": 7, "ts": 1716454224646027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537032, "dur": 7,
    "args": {
      "External id": 203728,
      "cbid": 211, "correlation": 203728
    }
  },
  {
    "ph": "s", "id": 203728, "pid": 76337, "tid": -914061504, "ts": 1716454224537032,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224646032, "dur": 54,
    "args": {
      "External id": 203731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203731,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203731, "pid": 5, "tid": 7, "ts": 1716454224646032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537051, "dur": 6,
    "args": {
      "External id": 203731,
      "cbid": 211, "correlation": 203731
    }
  },
  {
    "ph": "s", "id": 203731, "pid": 76337, "tid": -914061504, "ts": 1716454224537051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224646087, "dur": 36,
    "args": {
      "External id": 203740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203740,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203740, "pid": 5, "tid": 7, "ts": 1716454224646087,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537092, "dur": 10,
    "args": {
      "External id": 203740,
      "cbid": 211, "correlation": 203740
    }
  },
  {
    "ph": "s", "id": 203740, "pid": 76337, "tid": -914061504, "ts": 1716454224537092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537155, "dur": 0,
    "args": {
      "External id": 203750,
      "cbid": 317, "correlation": 203750
    }
  },
  {
    "ph": "f", "id": 203750, "pid": 76337, "tid": -914061504, "ts": 1716454224537155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537155, "dur": 0,
    "args": {
      "External id": 203751,
      "cbid": 203, "correlation": 203751
    }
  },
  {
    "ph": "f", "id": 203751, "pid": 76337, "tid": -914061504, "ts": 1716454224537155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537156, "dur": 0,
    "args": {
      "External id": 203752,
      "cbid": 205, "correlation": 203752
    }
  },
  {
    "ph": "f", "id": 203752, "pid": 76337, "tid": -914061504, "ts": 1716454224537156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224646125, "dur": 41,
    "args": {
      "External id": 203756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203756,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203756, "pid": 5, "tid": 7, "ts": 1716454224646125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537170, "dur": 12,
    "args": {
      "External id": 203756,
      "cbid": 211, "correlation": 203756
    }
  },
  {
    "ph": "s", "id": 203756, "pid": 76337, "tid": -914061504, "ts": 1716454224537170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224646167, "dur": 14,
    "args": {
      "External id": 203758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203758,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203758, "pid": 5, "tid": 7, "ts": 1716454224646167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537185, "dur": 6,
    "args": {
      "External id": 203758,
      "cbid": 211, "correlation": 203758
    }
  },
  {
    "ph": "s", "id": 203758, "pid": 76337, "tid": -914061504, "ts": 1716454224537185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224646183, "dur": 3,
    "args": {
      "External id": 203760,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203760,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203760, "pid": 5, "tid": 7, "ts": 1716454224646183,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537195, "dur": 5,
    "args": {
      "External id": 203760,
      "cbid": 211, "correlation": 203760
    }
  },
  {
    "ph": "s", "id": 203760, "pid": 76337, "tid": -914061504, "ts": 1716454224537195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537203, "dur": 0,
    "args": {
      "External id": 203761,
      "cbid": 51, "correlation": 203761
    }
  },
  {
    "ph": "s", "id": 203761, "pid": 76337, "tid": -914061504, "ts": 1716454224537203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224646187, "dur": 701,
    "args": {
      "External id": 203762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203762,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203762, "pid": 5, "tid": 7, "ts": 1716454224646187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537204, "dur": 5,
    "args": {
      "External id": 203762,
      "cbid": 211, "correlation": 203762
    }
  },
  {
    "ph": "s", "id": 203762, "pid": 76337, "tid": -914061504, "ts": 1716454224537204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224646890, "dur": 60,
    "args": {
      "External id": 203767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203767,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203767, "pid": 5, "tid": 7, "ts": 1716454224646890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537231, "dur": 9,
    "args": {
      "External id": 203767,
      "cbid": 211, "correlation": 203767
    }
  },
  {
    "ph": "s", "id": 203767, "pid": 76337, "tid": -914061504, "ts": 1716454224537231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537289, "dur": 0,
    "args": {
      "External id": 203777,
      "cbid": 317, "correlation": 203777
    }
  },
  {
    "ph": "f", "id": 203777, "pid": 76337, "tid": -914061504, "ts": 1716454224537289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537290, "dur": 0,
    "args": {
      "External id": 203778,
      "cbid": 203, "correlation": 203778
    }
  },
  {
    "ph": "f", "id": 203778, "pid": 76337, "tid": -914061504, "ts": 1716454224537290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537291, "dur": 0,
    "args": {
      "External id": 203779,
      "cbid": 205, "correlation": 203779
    }
  },
  {
    "ph": "f", "id": 203779, "pid": 76337, "tid": -914061504, "ts": 1716454224537291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224646951, "dur": 77,
    "args": {
      "External id": 203783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203783,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203783, "pid": 5, "tid": 7, "ts": 1716454224646951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537303, "dur": 11,
    "args": {
      "External id": 203783,
      "cbid": 211, "correlation": 203783
    }
  },
  {
    "ph": "s", "id": 203783, "pid": 76337, "tid": -914061504, "ts": 1716454224537303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224647029, "dur": 211,
    "args": {
      "External id": 203785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203785,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203785, "pid": 5, "tid": 7, "ts": 1716454224647029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537322, "dur": 7,
    "args": {
      "External id": 203785,
      "cbid": 211, "correlation": 203785
    }
  },
  {
    "ph": "s", "id": 203785, "pid": 76337, "tid": -914061504, "ts": 1716454224537322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224647241, "dur": 39,
    "args": {
      "External id": 203787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203787,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203787, "pid": 5, "tid": 7, "ts": 1716454224647241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537334, "dur": 145,
    "args": {
      "External id": 203787,
      "cbid": 211, "correlation": 203787
    }
  },
  {
    "ph": "s", "id": 203787, "pid": 76337, "tid": -914061504, "ts": 1716454224537334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224647282, "dur": 60,
    "args": {
      "External id": 203793,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203793,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203793, "pid": 5, "tid": 7, "ts": 1716454224647282,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537500, "dur": 9,
    "args": {
      "External id": 203793,
      "cbid": 211, "correlation": 203793
    }
  },
  {
    "ph": "s", "id": 203793, "pid": 76337, "tid": -914061504, "ts": 1716454224537500,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224647344, "dur": 50,
    "args": {
      "External id": 203801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203801,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203801, "pid": 5, "tid": 7, "ts": 1716454224647344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537530, "dur": 8,
    "args": {
      "External id": 203801,
      "cbid": 211, "correlation": 203801
    }
  },
  {
    "ph": "s", "id": 203801, "pid": 76337, "tid": -914061504, "ts": 1716454224537530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224647395, "dur": 36,
    "args": {
      "External id": 203809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203809,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203809, "pid": 5, "tid": 7, "ts": 1716454224647395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537560, "dur": 29,
    "args": {
      "External id": 203809,
      "cbid": 211, "correlation": 203809
    }
  },
  {
    "ph": "s", "id": 203809, "pid": 76337, "tid": -914061504, "ts": 1716454224537560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224647432, "dur": 51,
    "args": {
      "External id": 203829,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203829,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 203829, "pid": 5, "tid": 7, "ts": 1716454224647432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537668, "dur": 13,
    "args": {
      "External id": 203829,
      "cbid": 211, "correlation": 203829
    }
  },
  {
    "ph": "s", "id": 203829, "pid": 76337, "tid": -914061504, "ts": 1716454224537668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224647485, "dur": 4,
    "args": {
      "External id": 203841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203841,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 203841, "pid": 5, "tid": 7, "ts": 1716454224647485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537690, "dur": 6,
    "args": {
      "External id": 203841,
      "cbid": 211, "correlation": 203841
    }
  },
  {
    "ph": "s", "id": 203841, "pid": 76337, "tid": -914061504, "ts": 1716454224537690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224647491, "dur": 55,
    "args": {
      "External id": 203844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203844,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203844, "pid": 5, "tid": 7, "ts": 1716454224647491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537709, "dur": 7,
    "args": {
      "External id": 203844,
      "cbid": 211, "correlation": 203844
    }
  },
  {
    "ph": "s", "id": 203844, "pid": 76337, "tid": -914061504, "ts": 1716454224537709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537767, "dur": 0,
    "args": {
      "External id": 203855,
      "cbid": 317, "correlation": 203855
    }
  },
  {
    "ph": "f", "id": 203855, "pid": 76337, "tid": -914061504, "ts": 1716454224537767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537768, "dur": 0,
    "args": {
      "External id": 203856,
      "cbid": 203, "correlation": 203856
    }
  },
  {
    "ph": "f", "id": 203856, "pid": 76337, "tid": -914061504, "ts": 1716454224537768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537769, "dur": 0,
    "args": {
      "External id": 203857,
      "cbid": 205, "correlation": 203857
    }
  },
  {
    "ph": "f", "id": 203857, "pid": 76337, "tid": -914061504, "ts": 1716454224537769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537790, "dur": 1,
    "args": {
      "External id": 203861,
      "cbid": 251, "correlation": 203861
    }
  },
  {
    "ph": "f", "id": 203861, "pid": 76337, "tid": -914061504, "ts": 1716454224537790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537792, "dur": 0,
    "args": {
      "External id": 203862,
      "cbid": 251, "correlation": 203862
    }
  },
  {
    "ph": "f", "id": 203862, "pid": 76337, "tid": -914061504, "ts": 1716454224537792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537792, "dur": 0,
    "args": {
      "External id": 203863,
      "cbid": 251, "correlation": 203863
    }
  },
  {
    "ph": "f", "id": 203863, "pid": 76337, "tid": -914061504, "ts": 1716454224537792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537793, "dur": 0,
    "args": {
      "External id": 203864,
      "cbid": 251, "correlation": 203864
    }
  },
  {
    "ph": "f", "id": 203864, "pid": 76337, "tid": -914061504, "ts": 1716454224537793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537794, "dur": 0,
    "args": {
      "External id": 203865,
      "cbid": 251, "correlation": 203865
    }
  },
  {
    "ph": "f", "id": 203865, "pid": 76337, "tid": -914061504, "ts": 1716454224537794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537794, "dur": 0,
    "args": {
      "External id": 203866,
      "cbid": 251, "correlation": 203866
    }
  },
  {
    "ph": "f", "id": 203866, "pid": 76337, "tid": -914061504, "ts": 1716454224537794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537795, "dur": 0,
    "args": {
      "External id": 203867,
      "cbid": 251, "correlation": 203867
    }
  },
  {
    "ph": "f", "id": 203867, "pid": 76337, "tid": -914061504, "ts": 1716454224537795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537796, "dur": 0,
    "args": {
      "External id": 203868,
      "cbid": 251, "correlation": 203868
    }
  },
  {
    "ph": "f", "id": 203868, "pid": 76337, "tid": -914061504, "ts": 1716454224537796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537797, "dur": 0,
    "args": {
      "External id": 203869,
      "cbid": 251, "correlation": 203869
    }
  },
  {
    "ph": "f", "id": 203869, "pid": 76337, "tid": -914061504, "ts": 1716454224537797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224647547, "dur": 112,
    "args": {
      "External id": 203870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203870,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203870, "pid": 5, "tid": 7, "ts": 1716454224647547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537799, "dur": 13,
    "args": {
      "External id": 203870,
      "cbid": 211, "correlation": 203870
    }
  },
  {
    "ph": "s", "id": 203870, "pid": 76337, "tid": -914061504, "ts": 1716454224537799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224647661, "dur": 59,
    "args": {
      "External id": 203876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203876,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203876, "pid": 5, "tid": 7, "ts": 1716454224647661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537835, "dur": 9,
    "args": {
      "External id": 203876,
      "cbid": 211, "correlation": 203876
    }
  },
  {
    "ph": "s", "id": 203876, "pid": 76337, "tid": -914061504, "ts": 1716454224537835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224647721, "dur": 577,
    "args": {
      "External id": 203885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203885,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203885, "pid": 5, "tid": 7, "ts": 1716454224647721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537918, "dur": 14,
    "args": {
      "External id": 203885,
      "cbid": 211, "correlation": 203885
    }
  },
  {
    "ph": "s", "id": 203885, "pid": 76337, "tid": -914061504, "ts": 1716454224537918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224648299, "dur": 182,
    "args": {
      "External id": 203907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203907,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203907, "pid": 5, "tid": 7, "ts": 1716454224648299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224537983, "dur": 11,
    "args": {
      "External id": 203907,
      "cbid": 211, "correlation": 203907
    }
  },
  {
    "ph": "s", "id": 203907, "pid": 76337, "tid": -914061504, "ts": 1716454224537983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538072, "dur": 1,
    "args": {
      "External id": 203918,
      "cbid": 251, "correlation": 203918
    }
  },
  {
    "ph": "f", "id": 203918, "pid": 76337, "tid": -914061504, "ts": 1716454224538072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224648483, "dur": 196,
    "args": {
      "External id": 203919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203919,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203919, "pid": 5, "tid": 7, "ts": 1716454224648483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538077, "dur": 13,
    "args": {
      "External id": 203919,
      "cbid": 211, "correlation": 203919
    }
  },
  {
    "ph": "s", "id": 203919, "pid": 76337, "tid": -914061504, "ts": 1716454224538077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538144, "dur": 1,
    "args": {
      "External id": 203930,
      "cbid": 251, "correlation": 203930
    }
  },
  {
    "ph": "f", "id": 203930, "pid": 76337, "tid": -914061504, "ts": 1716454224538144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224648681, "dur": 190,
    "args": {
      "External id": 203931,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203931,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203931, "pid": 5, "tid": 7, "ts": 1716454224648681,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538148, "dur": 11,
    "args": {
      "External id": 203931,
      "cbid": 211, "correlation": 203931
    }
  },
  {
    "ph": "s", "id": 203931, "pid": 76337, "tid": -914061504, "ts": 1716454224538148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538212, "dur": 1,
    "args": {
      "External id": 203942,
      "cbid": 251, "correlation": 203942
    }
  },
  {
    "ph": "f", "id": 203942, "pid": 76337, "tid": -914061504, "ts": 1716454224538212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224648872, "dur": 189,
    "args": {
      "External id": 203943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203943,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203943, "pid": 5, "tid": 7, "ts": 1716454224648872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538216, "dur": 11,
    "args": {
      "External id": 203943,
      "cbid": 211, "correlation": 203943
    }
  },
  {
    "ph": "s", "id": 203943, "pid": 76337, "tid": -914061504, "ts": 1716454224538216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224649062, "dur": 18775,
    "args": {
      "External id": 203964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203964,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 203964, "pid": 5, "tid": 7, "ts": 1716454224649062,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538296, "dur": 12,
    "args": {
      "External id": 203964,
      "cbid": 211, "correlation": 203964
    }
  },
  {
    "ph": "s", "id": 203964, "pid": 76337, "tid": -914061504, "ts": 1716454224538296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538392, "dur": 1,
    "args": {
      "External id": 203982,
      "cbid": 251, "correlation": 203982
    }
  },
  {
    "ph": "f", "id": 203982, "pid": 76337, "tid": -914061504, "ts": 1716454224538392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224667838, "dur": 200,
    "args": {
      "External id": 203984,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203984,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 203984, "pid": 5, "tid": 7, "ts": 1716454224667838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538397, "dur": 13,
    "args": {
      "External id": 203984,
      "cbid": 211, "correlation": 203984
    }
  },
  {
    "ph": "s", "id": 203984, "pid": 76337, "tid": -914061504, "ts": 1716454224538397,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224668040, "dur": 66,
    "args": {
      "External id": 203992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 203992,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 203992, "pid": 5, "tid": 7, "ts": 1716454224668040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538467, "dur": 12,
    "args": {
      "External id": 203992,
      "cbid": 211, "correlation": 203992
    }
  },
  {
    "ph": "s", "id": 203992, "pid": 76337, "tid": -914061504, "ts": 1716454224538467,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224668108, "dur": 97,
    "args": {
      "External id": 204000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204000,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204000, "pid": 5, "tid": 7, "ts": 1716454224668108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538506, "dur": 9,
    "args": {
      "External id": 204000,
      "cbid": 211, "correlation": 204000
    }
  },
  {
    "ph": "s", "id": 204000, "pid": 76337, "tid": -914061504, "ts": 1716454224538506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224668206, "dur": 54,
    "args": {
      "External id": 204011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204011,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204011, "pid": 5, "tid": 7, "ts": 1716454224668206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538579, "dur": 50,
    "args": {
      "External id": 204011,
      "cbid": 211, "correlation": 204011
    }
  },
  {
    "ph": "s", "id": 204011, "pid": 76337, "tid": -914061504, "ts": 1716454224538579,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224668261, "dur": 93,
    "args": {
      "External id": 204033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204033,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204033, "pid": 5, "tid": 7, "ts": 1716454224668261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224538648, "dur": 1928,
    "args": {
      "External id": 204033,
      "cbid": 211, "correlation": 204033
    }
  },
  {
    "ph": "s", "id": 204033, "pid": 76337, "tid": -914061504, "ts": 1716454224538648,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540654, "dur": 1,
    "args": {
      "External id": 204044,
      "cbid": 251, "correlation": 204044
    }
  },
  {
    "ph": "f", "id": 204044, "pid": 76337, "tid": -914061504, "ts": 1716454224540654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224668355, "dur": 105,
    "args": {
      "External id": 204045,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204045,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204045, "pid": 5, "tid": 7, "ts": 1716454224668355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540659, "dur": 65,
    "args": {
      "External id": 204045,
      "cbid": 211, "correlation": 204045
    }
  },
  {
    "ph": "s", "id": 204045, "pid": 76337, "tid": -914061504, "ts": 1716454224540659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540784, "dur": 1,
    "args": {
      "External id": 204056,
      "cbid": 251, "correlation": 204056
    }
  },
  {
    "ph": "f", "id": 204056, "pid": 76337, "tid": -914061504, "ts": 1716454224540784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540788, "dur": 0,
    "args": {
      "External id": 204057,
      "cbid": 251, "correlation": 204057
    }
  },
  {
    "ph": "f", "id": 204057, "pid": 76337, "tid": -914061504, "ts": 1716454224540788,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224668461, "dur": 10,
    "args": {
      "External id": 204058,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204058,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 204058, "pid": 5, "tid": 7, "ts": 1716454224668461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540789, "dur": 12,
    "args": {
      "External id": 204058,
      "cbid": 211, "correlation": 204058
    }
  },
  {
    "ph": "s", "id": 204058, "pid": 76337, "tid": -914061504, "ts": 1716454224540789,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224668473, "dur": 5,
    "args": {
      "External id": 204060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204060,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 204060, "pid": 5, "tid": 7, "ts": 1716454224668473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540803, "dur": 6,
    "args": {
      "External id": 204060,
      "cbid": 211, "correlation": 204060
    }
  },
  {
    "ph": "s", "id": 204060, "pid": 76337, "tid": -914061504, "ts": 1716454224540803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540863, "dur": 1,
    "args": {
      "External id": 204071,
      "cbid": 251, "correlation": 204071
    }
  },
  {
    "ph": "f", "id": 204071, "pid": 76337, "tid": -914061504, "ts": 1716454224540863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540867, "dur": 0,
    "args": {
      "External id": 204072,
      "cbid": 251, "correlation": 204072
    }
  },
  {
    "ph": "f", "id": 204072, "pid": 76337, "tid": -914061504, "ts": 1716454224540867,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224668480, "dur": 6,
    "args": {
      "External id": 204073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204073,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 204073, "pid": 5, "tid": 7, "ts": 1716454224668480,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540868, "dur": 12,
    "args": {
      "External id": 204073,
      "cbid": 211, "correlation": 204073
    }
  },
  {
    "ph": "s", "id": 204073, "pid": 76337, "tid": -914061504, "ts": 1716454224540868,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224668487, "dur": 3,
    "args": {
      "External id": 204075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204075,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 204075, "pid": 5, "tid": 7, "ts": 1716454224668487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540882, "dur": 6,
    "args": {
      "External id": 204075,
      "cbid": 211, "correlation": 204075
    }
  },
  {
    "ph": "s", "id": 204075, "pid": 76337, "tid": -914061504, "ts": 1716454224540882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224668492, "dur": 157,
    "args": {
      "External id": 204096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204096,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204096, "pid": 5, "tid": 7, "ts": 1716454224668492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224540955, "dur": 16,
    "args": {
      "External id": 204096,
      "cbid": 211, "correlation": 204096
    }
  },
  {
    "ph": "s", "id": 204096, "pid": 76337, "tid": -914061504, "ts": 1716454224540955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541064, "dur": 1,
    "args": {
      "External id": 204114,
      "cbid": 251, "correlation": 204114
    }
  },
  {
    "ph": "f", "id": 204114, "pid": 76337, "tid": -914061504, "ts": 1716454224541064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224668650, "dur": 107,
    "args": {
      "External id": 204116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204116,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204116, "pid": 5, "tid": 7, "ts": 1716454224668650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541070, "dur": 14,
    "args": {
      "External id": 204116,
      "cbid": 211, "correlation": 204116
    }
  },
  {
    "ph": "s", "id": 204116, "pid": 76337, "tid": -914061504, "ts": 1716454224541070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224668758, "dur": 35,
    "args": {
      "External id": 204124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204124,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204124, "pid": 5, "tid": 7, "ts": 1716454224668758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541141, "dur": 12,
    "args": {
      "External id": 204124,
      "cbid": 211, "correlation": 204124
    }
  },
  {
    "ph": "s", "id": 204124, "pid": 76337, "tid": -914061504, "ts": 1716454224541141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224668795, "dur": 66,
    "args": {
      "External id": 204132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204132,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204132, "pid": 5, "tid": 7, "ts": 1716454224668795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541183, "dur": 9,
    "args": {
      "External id": 204132,
      "cbid": 211, "correlation": 204132
    }
  },
  {
    "ph": "s", "id": 204132, "pid": 76337, "tid": -914061504, "ts": 1716454224541183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224668862, "dur": 93,
    "args": {
      "External id": 204154,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204154,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204154, "pid": 5, "tid": 7, "ts": 1716454224668862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541233, "dur": 10,
    "args": {
      "External id": 204154,
      "cbid": 211, "correlation": 204154
    }
  },
  {
    "ph": "s", "id": 204154, "pid": 76337, "tid": -914061504, "ts": 1716454224541233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541319, "dur": 1,
    "args": {
      "External id": 204170,
      "cbid": 251, "correlation": 204170
    }
  },
  {
    "ph": "f", "id": 204170, "pid": 76337, "tid": -914061504, "ts": 1716454224541319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224668956, "dur": 579,
    "args": {
      "External id": 204172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204172,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204172, "pid": 5, "tid": 7, "ts": 1716454224668956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541325, "dur": 13,
    "args": {
      "External id": 204172,
      "cbid": 211, "correlation": 204172
    }
  },
  {
    "ph": "s", "id": 204172, "pid": 76337, "tid": -914061504, "ts": 1716454224541325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224669536, "dur": 243,
    "args": {
      "External id": 204180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204180,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204180, "pid": 5, "tid": 7, "ts": 1716454224669536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541390, "dur": 12,
    "args": {
      "External id": 204180,
      "cbid": 211, "correlation": 204180
    }
  },
  {
    "ph": "s", "id": 204180, "pid": 76337, "tid": -914061504, "ts": 1716454224541390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224669781, "dur": 251,
    "args": {
      "External id": 204188,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204188,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204188, "pid": 5, "tid": 7, "ts": 1716454224669781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541420, "dur": 9,
    "args": {
      "External id": 204188,
      "cbid": 211, "correlation": 204188
    }
  },
  {
    "ph": "s", "id": 204188, "pid": 76337, "tid": -914061504, "ts": 1716454224541420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541501, "dur": 1,
    "args": {
      "External id": 204204,
      "cbid": 251, "correlation": 204204
    }
  },
  {
    "ph": "f", "id": 204204, "pid": 76337, "tid": -914061504, "ts": 1716454224541501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541506, "dur": 0,
    "args": {
      "External id": 204206,
      "cbid": 251, "correlation": 204206
    }
  },
  {
    "ph": "f", "id": 204206, "pid": 76337, "tid": -914061504, "ts": 1716454224541506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224670033, "dur": 360,
    "args": {
      "External id": 204207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204207,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 204207, "pid": 5, "tid": 7, "ts": 1716454224670033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541509, "dur": 12,
    "args": {
      "External id": 204207,
      "cbid": 211, "correlation": 204207
    }
  },
  {
    "ph": "s", "id": 204207, "pid": 76337, "tid": -914061504, "ts": 1716454224541509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224670395, "dur": 50,
    "args": {
      "External id": 204215,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204215,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204215, "pid": 5, "tid": 7, "ts": 1716454224670395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541551, "dur": 10,
    "args": {
      "External id": 204215,
      "cbid": 211, "correlation": 204215
    }
  },
  {
    "ph": "s", "id": 204215, "pid": 76337, "tid": -914061504, "ts": 1716454224541551,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224670446, "dur": 159,
    "args": {
      "External id": 204226,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204226,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204226, "pid": 5, "tid": 7, "ts": 1716454224670446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541617, "dur": 215,
    "args": {
      "External id": 204226,
      "cbid": 211, "correlation": 204226
    }
  },
  {
    "ph": "s", "id": 204226, "pid": 76337, "tid": -914061504, "ts": 1716454224541617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541884, "dur": 0,
    "args": {
      "External id": 204238,
      "cbid": 317, "correlation": 204238
    }
  },
  {
    "ph": "f", "id": 204238, "pid": 76337, "tid": -914061504, "ts": 1716454224541884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541885, "dur": 0,
    "args": {
      "External id": 204239,
      "cbid": 203, "correlation": 204239
    }
  },
  {
    "ph": "f", "id": 204239, "pid": 76337, "tid": -914061504, "ts": 1716454224541885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541885, "dur": 0,
    "args": {
      "External id": 204240,
      "cbid": 205, "correlation": 204240
    }
  },
  {
    "ph": "f", "id": 204240, "pid": 76337, "tid": -914061504, "ts": 1716454224541885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541908, "dur": 1,
    "args": {
      "External id": 204244,
      "cbid": 251, "correlation": 204244
    }
  },
  {
    "ph": "f", "id": 204244, "pid": 76337, "tid": -914061504, "ts": 1716454224541908,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541910, "dur": 0,
    "args": {
      "External id": 204245,
      "cbid": 251, "correlation": 204245
    }
  },
  {
    "ph": "f", "id": 204245, "pid": 76337, "tid": -914061504, "ts": 1716454224541910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541911, "dur": 0,
    "args": {
      "External id": 204246,
      "cbid": 251, "correlation": 204246
    }
  },
  {
    "ph": "f", "id": 204246, "pid": 76337, "tid": -914061504, "ts": 1716454224541911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541912, "dur": 0,
    "args": {
      "External id": 204247,
      "cbid": 251, "correlation": 204247
    }
  },
  {
    "ph": "f", "id": 204247, "pid": 76337, "tid": -914061504, "ts": 1716454224541912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541912, "dur": 0,
    "args": {
      "External id": 204248,
      "cbid": 251, "correlation": 204248
    }
  },
  {
    "ph": "f", "id": 204248, "pid": 76337, "tid": -914061504, "ts": 1716454224541912,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541913, "dur": 0,
    "args": {
      "External id": 204249,
      "cbid": 251, "correlation": 204249
    }
  },
  {
    "ph": "f", "id": 204249, "pid": 76337, "tid": -914061504, "ts": 1716454224541913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541914, "dur": 0,
    "args": {
      "External id": 204250,
      "cbid": 251, "correlation": 204250
    }
  },
  {
    "ph": "f", "id": 204250, "pid": 76337, "tid": -914061504, "ts": 1716454224541914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541915, "dur": 0,
    "args": {
      "External id": 204251,
      "cbid": 251, "correlation": 204251
    }
  },
  {
    "ph": "f", "id": 204251, "pid": 76337, "tid": -914061504, "ts": 1716454224541915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541916, "dur": 0,
    "args": {
      "External id": 204252,
      "cbid": 251, "correlation": 204252
    }
  },
  {
    "ph": "f", "id": 204252, "pid": 76337, "tid": -914061504, "ts": 1716454224541916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224670606, "dur": 115,
    "args": {
      "External id": 204253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204253,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204253, "pid": 5, "tid": 7, "ts": 1716454224670606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541918, "dur": 39,
    "args": {
      "External id": 204253,
      "cbid": 211, "correlation": 204253
    }
  },
  {
    "ph": "s", "id": 204253, "pid": 76337, "tid": -914061504, "ts": 1716454224541918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224670722, "dur": 61,
    "args": {
      "External id": 204259,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204259,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204259, "pid": 5, "tid": 7, "ts": 1716454224670722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224541988, "dur": 97,
    "args": {
      "External id": 204259,
      "cbid": 211, "correlation": 204259
    }
  },
  {
    "ph": "s", "id": 204259, "pid": 76337, "tid": -914061504, "ts": 1716454224541988,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224670784, "dur": 50,
    "args": {
      "External id": 204267,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204267,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204267, "pid": 5, "tid": 7, "ts": 1716454224670784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542109, "dur": 233,
    "args": {
      "External id": 204267,
      "cbid": 211, "correlation": 204267
    }
  },
  {
    "ph": "s", "id": 204267, "pid": 76337, "tid": -914061504, "ts": 1716454224542109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224670836, "dur": 99,
    "args": {
      "External id": 204276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204276,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204276, "pid": 5, "tid": 7, "ts": 1716454224670836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542374, "dur": 10,
    "args": {
      "External id": 204276,
      "cbid": 211, "correlation": 204276
    }
  },
  {
    "ph": "s", "id": 204276, "pid": 76337, "tid": -914061504, "ts": 1716454224542374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224670936, "dur": 93,
    "args": {
      "External id": 204296,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204296,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 204296, "pid": 5, "tid": 7, "ts": 1716454224670936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542446, "dur": 12,
    "args": {
      "External id": 204296,
      "cbid": 211, "correlation": 204296
    }
  },
  {
    "ph": "s", "id": 204296, "pid": 76337, "tid": -914061504, "ts": 1716454224542446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224671031, "dur": 5,
    "args": {
      "External id": 204308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204308,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 204308, "pid": 5, "tid": 7, "ts": 1716454224671031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542468, "dur": 12,
    "args": {
      "External id": 204308,
      "cbid": 211, "correlation": 204308
    }
  },
  {
    "ph": "s", "id": 204308, "pid": 76337, "tid": -914061504, "ts": 1716454224542468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224671037, "dur": 108,
    "args": {
      "External id": 204311,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204311,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204311, "pid": 5, "tid": 7, "ts": 1716454224671037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542492, "dur": 109,
    "args": {
      "External id": 204311,
      "cbid": 211, "correlation": 204311
    }
  },
  {
    "ph": "s", "id": 204311, "pid": 76337, "tid": -914061504, "ts": 1716454224542492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224671147, "dur": 68,
    "args": {
      "External id": 204320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204320,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204320, "pid": 5, "tid": 7, "ts": 1716454224671147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542635, "dur": 10,
    "args": {
      "External id": 204320,
      "cbid": 211, "correlation": 204320
    }
  },
  {
    "ph": "s", "id": 204320, "pid": 76337, "tid": -914061504, "ts": 1716454224542635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542686, "dur": 0,
    "args": {
      "External id": 204330,
      "cbid": 317, "correlation": 204330
    }
  },
  {
    "ph": "f", "id": 204330, "pid": 76337, "tid": -914061504, "ts": 1716454224542686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542687, "dur": 0,
    "args": {
      "External id": 204331,
      "cbid": 203, "correlation": 204331
    }
  },
  {
    "ph": "f", "id": 204331, "pid": 76337, "tid": -914061504, "ts": 1716454224542687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542688, "dur": 0,
    "args": {
      "External id": 204332,
      "cbid": 205, "correlation": 204332
    }
  },
  {
    "ph": "f", "id": 204332, "pid": 76337, "tid": -914061504, "ts": 1716454224542688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224671217, "dur": 76,
    "args": {
      "External id": 204336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204336, "pid": 5, "tid": 7, "ts": 1716454224671217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542703, "dur": 12,
    "args": {
      "External id": 204336,
      "cbid": 211, "correlation": 204336
    }
  },
  {
    "ph": "s", "id": 204336, "pid": 76337, "tid": -914061504, "ts": 1716454224542703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224671294, "dur": 24,
    "args": {
      "External id": 204338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204338,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204338, "pid": 5, "tid": 7, "ts": 1716454224671294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542717, "dur": 6,
    "args": {
      "External id": 204338,
      "cbid": 211, "correlation": 204338
    }
  },
  {
    "ph": "s", "id": 204338, "pid": 76337, "tid": -914061504, "ts": 1716454224542717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224671320, "dur": 4,
    "args": {
      "External id": 204340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 204340, "pid": 5, "tid": 7, "ts": 1716454224671320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542729, "dur": 6,
    "args": {
      "External id": 204340,
      "cbid": 211, "correlation": 204340
    }
  },
  {
    "ph": "s", "id": 204340, "pid": 76337, "tid": -914061504, "ts": 1716454224542729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542737, "dur": 0,
    "args": {
      "External id": 204341,
      "cbid": 51, "correlation": 204341
    }
  },
  {
    "ph": "s", "id": 204341, "pid": 76337, "tid": -914061504, "ts": 1716454224542737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224671325, "dur": 1372,
    "args": {
      "External id": 204342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204342,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204342, "pid": 5, "tid": 7, "ts": 1716454224671325,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542738, "dur": 5,
    "args": {
      "External id": 204342,
      "cbid": 211, "correlation": 204342
    }
  },
  {
    "ph": "s", "id": 204342, "pid": 76337, "tid": -914061504, "ts": 1716454224542738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224672698, "dur": 59,
    "args": {
      "External id": 204347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204347,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204347, "pid": 5, "tid": 7, "ts": 1716454224672698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542765, "dur": 9,
    "args": {
      "External id": 204347,
      "cbid": 211, "correlation": 204347
    }
  },
  {
    "ph": "s", "id": 204347, "pid": 76337, "tid": -914061504, "ts": 1716454224542765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224672758, "dur": 3,
    "args": {
      "External id": 204355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204355,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 204355, "pid": 5, "tid": 7, "ts": 1716454224672758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542809, "dur": 9,
    "args": {
      "External id": 204355,
      "cbid": 211, "correlation": 204355
    }
  },
  {
    "ph": "s", "id": 204355, "pid": 76337, "tid": -914061504, "ts": 1716454224542809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542876, "dur": 1,
    "args": {
      "External id": 204371,
      "cbid": 251, "correlation": 204371
    }
  },
  {
    "ph": "f", "id": 204371, "pid": 76337, "tid": -914061504, "ts": 1716454224542876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542882, "dur": 0,
    "args": {
      "External id": 204373,
      "cbid": 251, "correlation": 204373
    }
  },
  {
    "ph": "f", "id": 204373, "pid": 76337, "tid": -914061504, "ts": 1716454224542882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224672763, "dur": 11,
    "args": {
      "External id": 204374,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204374,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 204374, "pid": 5, "tid": 7, "ts": 1716454224672763,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542884, "dur": 12,
    "args": {
      "External id": 204374,
      "cbid": 211, "correlation": 204374
    }
  },
  {
    "ph": "s", "id": 204374, "pid": 76337, "tid": -914061504, "ts": 1716454224542884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224672775, "dur": 5,
    "args": {
      "External id": 204376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204376,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 204376, "pid": 5, "tid": 7, "ts": 1716454224672775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542898, "dur": 6,
    "args": {
      "External id": 204376,
      "cbid": 211, "correlation": 204376
    }
  },
  {
    "ph": "s", "id": 204376, "pid": 76337, "tid": -914061504, "ts": 1716454224542898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224672782, "dur": 54,
    "args": {
      "External id": 204386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204386,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204386, "pid": 5, "tid": 7, "ts": 1716454224672782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224542955, "dur": 553,
    "args": {
      "External id": 204386,
      "cbid": 211, "correlation": 204386
    }
  },
  {
    "ph": "s", "id": 204386, "pid": 76337, "tid": -914061504, "ts": 1716454224542955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224672837, "dur": 52,
    "args": {
      "External id": 204406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204406,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 204406, "pid": 5, "tid": 7, "ts": 1716454224672837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543565, "dur": 11,
    "args": {
      "External id": 204406,
      "cbid": 211, "correlation": 204406
    }
  },
  {
    "ph": "s", "id": 204406, "pid": 76337, "tid": -914061504, "ts": 1716454224543565,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224672890, "dur": 4,
    "args": {
      "External id": 204418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204418,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 204418, "pid": 5, "tid": 7, "ts": 1716454224672890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543585, "dur": 7,
    "args": {
      "External id": 204418,
      "cbid": 211, "correlation": 204418
    }
  },
  {
    "ph": "s", "id": 204418, "pid": 76337, "tid": -914061504, "ts": 1716454224543585,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224672895, "dur": 54,
    "args": {
      "External id": 204421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204421,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204421, "pid": 5, "tid": 7, "ts": 1716454224672895,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543604, "dur": 7,
    "args": {
      "External id": 204421,
      "cbid": 211, "correlation": 204421
    }
  },
  {
    "ph": "s", "id": 204421, "pid": 76337, "tid": -914061504, "ts": 1716454224543604,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224672951, "dur": 36,
    "args": {
      "External id": 204430,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204430,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204430, "pid": 5, "tid": 7, "ts": 1716454224672951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543644, "dur": 10,
    "args": {
      "External id": 204430,
      "cbid": 211, "correlation": 204430
    }
  },
  {
    "ph": "s", "id": 204430, "pid": 76337, "tid": -914061504, "ts": 1716454224543644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543707, "dur": 0,
    "args": {
      "External id": 204440,
      "cbid": 317, "correlation": 204440
    }
  },
  {
    "ph": "f", "id": 204440, "pid": 76337, "tid": -914061504, "ts": 1716454224543707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543708, "dur": 0,
    "args": {
      "External id": 204441,
      "cbid": 203, "correlation": 204441
    }
  },
  {
    "ph": "f", "id": 204441, "pid": 76337, "tid": -914061504, "ts": 1716454224543708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543709, "dur": 0,
    "args": {
      "External id": 204442,
      "cbid": 205, "correlation": 204442
    }
  },
  {
    "ph": "f", "id": 204442, "pid": 76337, "tid": -914061504, "ts": 1716454224543709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224672988, "dur": 40,
    "args": {
      "External id": 204446,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204446,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204446, "pid": 5, "tid": 7, "ts": 1716454224672988,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543723, "dur": 12,
    "args": {
      "External id": 204446,
      "cbid": 211, "correlation": 204446
    }
  },
  {
    "ph": "s", "id": 204446, "pid": 76337, "tid": -914061504, "ts": 1716454224543723,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224673029, "dur": 14,
    "args": {
      "External id": 204448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204448,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204448, "pid": 5, "tid": 7, "ts": 1716454224673029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543737, "dur": 5,
    "args": {
      "External id": 204448,
      "cbid": 211, "correlation": 204448
    }
  },
  {
    "ph": "s", "id": 204448, "pid": 76337, "tid": -914061504, "ts": 1716454224543737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224673045, "dur": 3,
    "args": {
      "External id": 204450,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204450,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 204450, "pid": 5, "tid": 7, "ts": 1716454224673045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543747, "dur": 6,
    "args": {
      "External id": 204450,
      "cbid": 211, "correlation": 204450
    }
  },
  {
    "ph": "s", "id": 204450, "pid": 76337, "tid": -914061504, "ts": 1716454224543747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543756, "dur": 0,
    "args": {
      "External id": 204451,
      "cbid": 51, "correlation": 204451
    }
  },
  {
    "ph": "s", "id": 204451, "pid": 76337, "tid": -914061504, "ts": 1716454224543756,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224673050, "dur": 702,
    "args": {
      "External id": 204452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204452,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204452, "pid": 5, "tid": 7, "ts": 1716454224673050,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543757, "dur": 5,
    "args": {
      "External id": 204452,
      "cbid": 211, "correlation": 204452
    }
  },
  {
    "ph": "s", "id": 204452, "pid": 76337, "tid": -914061504, "ts": 1716454224543757,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224673753, "dur": 60,
    "args": {
      "External id": 204457,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204457,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204457, "pid": 5, "tid": 7, "ts": 1716454224673753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543785, "dur": 9,
    "args": {
      "External id": 204457,
      "cbid": 211, "correlation": 204457
    }
  },
  {
    "ph": "s", "id": 204457, "pid": 76337, "tid": -914061504, "ts": 1716454224543785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543842, "dur": 0,
    "args": {
      "External id": 204467,
      "cbid": 317, "correlation": 204467
    }
  },
  {
    "ph": "f", "id": 204467, "pid": 76337, "tid": -914061504, "ts": 1716454224543842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543842, "dur": 0,
    "args": {
      "External id": 204468,
      "cbid": 203, "correlation": 204468
    }
  },
  {
    "ph": "f", "id": 204468, "pid": 76337, "tid": -914061504, "ts": 1716454224543842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543843, "dur": 0,
    "args": {
      "External id": 204469,
      "cbid": 205, "correlation": 204469
    }
  },
  {
    "ph": "f", "id": 204469, "pid": 76337, "tid": -914061504, "ts": 1716454224543843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224673814, "dur": 75,
    "args": {
      "External id": 204473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204473,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204473, "pid": 5, "tid": 7, "ts": 1716454224673814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543855, "dur": 11,
    "args": {
      "External id": 204473,
      "cbid": 211, "correlation": 204473
    }
  },
  {
    "ph": "s", "id": 204473, "pid": 76337, "tid": -914061504, "ts": 1716454224543855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224673890, "dur": 206,
    "args": {
      "External id": 204475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204475,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204475, "pid": 5, "tid": 7, "ts": 1716454224673890,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543873, "dur": 6,
    "args": {
      "External id": 204475,
      "cbid": 211, "correlation": 204475
    }
  },
  {
    "ph": "s", "id": 204475, "pid": 76337, "tid": -914061504, "ts": 1716454224543873,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224674097, "dur": 39,
    "args": {
      "External id": 204477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204477,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204477, "pid": 5, "tid": 7, "ts": 1716454224674097,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543884, "dur": 5,
    "args": {
      "External id": 204477,
      "cbid": 211, "correlation": 204477
    }
  },
  {
    "ph": "s", "id": 204477, "pid": 76337, "tid": -914061504, "ts": 1716454224543884,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224674137, "dur": 60,
    "args": {
      "External id": 204483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204483,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204483, "pid": 5, "tid": 7, "ts": 1716454224674137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224543910, "dur": 516,
    "args": {
      "External id": 204483,
      "cbid": 211, "correlation": 204483
    }
  },
  {
    "ph": "s", "id": 204483, "pid": 76337, "tid": -914061504, "ts": 1716454224543910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224674198, "dur": 50,
    "args": {
      "External id": 204491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204491,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204491, "pid": 5, "tid": 7, "ts": 1716454224674198,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544449, "dur": 9,
    "args": {
      "External id": 204491,
      "cbid": 211, "correlation": 204491
    }
  },
  {
    "ph": "s", "id": 204491, "pid": 76337, "tid": -914061504, "ts": 1716454224544449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224674249, "dur": 35,
    "args": {
      "External id": 204499,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204499,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204499, "pid": 5, "tid": 7, "ts": 1716454224674249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544479, "dur": 8,
    "args": {
      "External id": 204499,
      "cbid": 211, "correlation": 204499
    }
  },
  {
    "ph": "s", "id": 204499, "pid": 76337, "tid": -914061504, "ts": 1716454224544479,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224674286, "dur": 50,
    "args": {
      "External id": 204519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204519,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 204519, "pid": 5, "tid": 7, "ts": 1716454224674286,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544561, "dur": 13,
    "args": {
      "External id": 204519,
      "cbid": 211, "correlation": 204519
    }
  },
  {
    "ph": "s", "id": 204519, "pid": 76337, "tid": -914061504, "ts": 1716454224544561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224674337, "dur": 4,
    "args": {
      "External id": 204531,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204531,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 204531, "pid": 5, "tid": 7, "ts": 1716454224674337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544584, "dur": 6,
    "args": {
      "External id": 204531,
      "cbid": 211, "correlation": 204531
    }
  },
  {
    "ph": "s", "id": 204531, "pid": 76337, "tid": -914061504, "ts": 1716454224544584,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224674343, "dur": 55,
    "args": {
      "External id": 204534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204534,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204534, "pid": 5, "tid": 7, "ts": 1716454224674343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544601, "dur": 7,
    "args": {
      "External id": 204534,
      "cbid": 211, "correlation": 204534
    }
  },
  {
    "ph": "s", "id": 204534, "pid": 76337, "tid": -914061504, "ts": 1716454224544601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544659, "dur": 0,
    "args": {
      "External id": 204545,
      "cbid": 317, "correlation": 204545
    }
  },
  {
    "ph": "f", "id": 204545, "pid": 76337, "tid": -914061504, "ts": 1716454224544659,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544660, "dur": 0,
    "args": {
      "External id": 204546,
      "cbid": 203, "correlation": 204546
    }
  },
  {
    "ph": "f", "id": 204546, "pid": 76337, "tid": -914061504, "ts": 1716454224544660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544660, "dur": 0,
    "args": {
      "External id": 204547,
      "cbid": 205, "correlation": 204547
    }
  },
  {
    "ph": "f", "id": 204547, "pid": 76337, "tid": -914061504, "ts": 1716454224544660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544682, "dur": 1,
    "args": {
      "External id": 204551,
      "cbid": 251, "correlation": 204551
    }
  },
  {
    "ph": "f", "id": 204551, "pid": 76337, "tid": -914061504, "ts": 1716454224544682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544684, "dur": 0,
    "args": {
      "External id": 204552,
      "cbid": 251, "correlation": 204552
    }
  },
  {
    "ph": "f", "id": 204552, "pid": 76337, "tid": -914061504, "ts": 1716454224544684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544685, "dur": 0,
    "args": {
      "External id": 204553,
      "cbid": 251, "correlation": 204553
    }
  },
  {
    "ph": "f", "id": 204553, "pid": 76337, "tid": -914061504, "ts": 1716454224544685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544686, "dur": 0,
    "args": {
      "External id": 204554,
      "cbid": 251, "correlation": 204554
    }
  },
  {
    "ph": "f", "id": 204554, "pid": 76337, "tid": -914061504, "ts": 1716454224544686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544686, "dur": 0,
    "args": {
      "External id": 204555,
      "cbid": 251, "correlation": 204555
    }
  },
  {
    "ph": "f", "id": 204555, "pid": 76337, "tid": -914061504, "ts": 1716454224544686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544687, "dur": 0,
    "args": {
      "External id": 204556,
      "cbid": 251, "correlation": 204556
    }
  },
  {
    "ph": "f", "id": 204556, "pid": 76337, "tid": -914061504, "ts": 1716454224544687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544688, "dur": 0,
    "args": {
      "External id": 204557,
      "cbid": 251, "correlation": 204557
    }
  },
  {
    "ph": "f", "id": 204557, "pid": 76337, "tid": -914061504, "ts": 1716454224544688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544688, "dur": 0,
    "args": {
      "External id": 204558,
      "cbid": 251, "correlation": 204558
    }
  },
  {
    "ph": "f", "id": 204558, "pid": 76337, "tid": -914061504, "ts": 1716454224544688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544690, "dur": 0,
    "args": {
      "External id": 204559,
      "cbid": 251, "correlation": 204559
    }
  },
  {
    "ph": "f", "id": 204559, "pid": 76337, "tid": -914061504, "ts": 1716454224544690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224674399, "dur": 113,
    "args": {
      "External id": 204560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204560,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204560, "pid": 5, "tid": 7, "ts": 1716454224674399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544692, "dur": 12,
    "args": {
      "External id": 204560,
      "cbid": 211, "correlation": 204560
    }
  },
  {
    "ph": "s", "id": 204560, "pid": 76337, "tid": -914061504, "ts": 1716454224544692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224674513, "dur": 59,
    "args": {
      "External id": 204566,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204566,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204566, "pid": 5, "tid": 7, "ts": 1716454224674513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544727, "dur": 9,
    "args": {
      "External id": 204566,
      "cbid": 211, "correlation": 204566
    }
  },
  {
    "ph": "s", "id": 204566, "pid": 76337, "tid": -914061504, "ts": 1716454224544727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224674574, "dur": 600,
    "args": {
      "External id": 204575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204575,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204575, "pid": 5, "tid": 7, "ts": 1716454224674574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544810, "dur": 15,
    "args": {
      "External id": 204575,
      "cbid": 211, "correlation": 204575
    }
  },
  {
    "ph": "s", "id": 204575, "pid": 76337, "tid": -914061504, "ts": 1716454224544810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224675175, "dur": 183,
    "args": {
      "External id": 204597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204597,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204597, "pid": 5, "tid": 7, "ts": 1716454224675175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544867, "dur": 11,
    "args": {
      "External id": 204597,
      "cbid": 211, "correlation": 204597
    }
  },
  {
    "ph": "s", "id": 204597, "pid": 76337, "tid": -914061504, "ts": 1716454224544867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544954, "dur": 1,
    "args": {
      "External id": 204608,
      "cbid": 251, "correlation": 204608
    }
  },
  {
    "ph": "f", "id": 204608, "pid": 76337, "tid": -914061504, "ts": 1716454224544954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224675359, "dur": 199,
    "args": {
      "External id": 204609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204609,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204609, "pid": 5, "tid": 7, "ts": 1716454224675359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224544959, "dur": 21,
    "args": {
      "External id": 204609,
      "cbid": 211, "correlation": 204609
    }
  },
  {
    "ph": "s", "id": 204609, "pid": 76337, "tid": -914061504, "ts": 1716454224544959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545037, "dur": 1,
    "args": {
      "External id": 204620,
      "cbid": 251, "correlation": 204620
    }
  },
  {
    "ph": "f", "id": 204620, "pid": 76337, "tid": -914061504, "ts": 1716454224545037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224675560, "dur": 190,
    "args": {
      "External id": 204621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204621,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204621, "pid": 5, "tid": 7, "ts": 1716454224675560,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545041, "dur": 12,
    "args": {
      "External id": 204621,
      "cbid": 211, "correlation": 204621
    }
  },
  {
    "ph": "s", "id": 204621, "pid": 76337, "tid": -914061504, "ts": 1716454224545041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545105, "dur": 1,
    "args": {
      "External id": 204632,
      "cbid": 251, "correlation": 204632
    }
  },
  {
    "ph": "f", "id": 204632, "pid": 76337, "tid": -914061504, "ts": 1716454224545105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224675751, "dur": 187,
    "args": {
      "External id": 204633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204633,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204633, "pid": 5, "tid": 7, "ts": 1716454224675751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545109, "dur": 12,
    "args": {
      "External id": 204633,
      "cbid": 211, "correlation": 204633
    }
  },
  {
    "ph": "s", "id": 204633, "pid": 76337, "tid": -914061504, "ts": 1716454224545109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224675940, "dur": 18734,
    "args": {
      "External id": 204654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204654,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204654, "pid": 5, "tid": 7, "ts": 1716454224675940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545188, "dur": 12,
    "args": {
      "External id": 204654,
      "cbid": 211, "correlation": 204654
    }
  },
  {
    "ph": "s", "id": 204654, "pid": 76337, "tid": -914061504, "ts": 1716454224545188,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545284, "dur": 1,
    "args": {
      "External id": 204672,
      "cbid": 251, "correlation": 204672
    }
  },
  {
    "ph": "f", "id": 204672, "pid": 76337, "tid": -914061504, "ts": 1716454224545284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224694675, "dur": 203,
    "args": {
      "External id": 204674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204674,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204674, "pid": 5, "tid": 7, "ts": 1716454224694675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545290, "dur": 13,
    "args": {
      "External id": 204674,
      "cbid": 211, "correlation": 204674
    }
  },
  {
    "ph": "s", "id": 204674, "pid": 76337, "tid": -914061504, "ts": 1716454224545290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224694880, "dur": 66,
    "args": {
      "External id": 204682,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204682,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204682, "pid": 5, "tid": 7, "ts": 1716454224694880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545359, "dur": 13,
    "args": {
      "External id": 204682,
      "cbid": 211, "correlation": 204682
    }
  },
  {
    "ph": "s", "id": 204682, "pid": 76337, "tid": -914061504, "ts": 1716454224545359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224694947, "dur": 97,
    "args": {
      "External id": 204690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204690,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204690, "pid": 5, "tid": 7, "ts": 1716454224694947,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545399, "dur": 85,
    "args": {
      "External id": 204690,
      "cbid": 211, "correlation": 204690
    }
  },
  {
    "ph": "s", "id": 204690, "pid": 76337, "tid": -914061504, "ts": 1716454224545399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224695046, "dur": 54,
    "args": {
      "External id": 204701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204701, "pid": 5, "tid": 7, "ts": 1716454224695046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224545546, "dur": 1887,
    "args": {
      "External id": 204701,
      "cbid": 211, "correlation": 204701
    }
  },
  {
    "ph": "s", "id": 204701, "pid": 76337, "tid": -914061504, "ts": 1716454224545546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224695101, "dur": 93,
    "args": {
      "External id": 204723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204723,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204723, "pid": 5, "tid": 7, "ts": 1716454224695101,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547453, "dur": 126,
    "args": {
      "External id": 204723,
      "cbid": 211, "correlation": 204723
    }
  },
  {
    "ph": "s", "id": 204723, "pid": 76337, "tid": -914061504, "ts": 1716454224547453,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547655, "dur": 1,
    "args": {
      "External id": 204734,
      "cbid": 251, "correlation": 204734
    }
  },
  {
    "ph": "f", "id": 204734, "pid": 76337, "tid": -914061504, "ts": 1716454224547655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224695195, "dur": 104,
    "args": {
      "External id": 204735,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204735,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204735, "pid": 5, "tid": 7, "ts": 1716454224695195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547661, "dur": 13,
    "args": {
      "External id": 204735,
      "cbid": 211, "correlation": 204735
    }
  },
  {
    "ph": "s", "id": 204735, "pid": 76337, "tid": -914061504, "ts": 1716454224547661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547733, "dur": 1,
    "args": {
      "External id": 204746,
      "cbid": 251, "correlation": 204746
    }
  },
  {
    "ph": "f", "id": 204746, "pid": 76337, "tid": -914061504, "ts": 1716454224547733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547737, "dur": 0,
    "args": {
      "External id": 204747,
      "cbid": 251, "correlation": 204747
    }
  },
  {
    "ph": "f", "id": 204747, "pid": 76337, "tid": -914061504, "ts": 1716454224547737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224695300, "dur": 11,
    "args": {
      "External id": 204748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204748,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 204748, "pid": 5, "tid": 7, "ts": 1716454224695300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547739, "dur": 12,
    "args": {
      "External id": 204748,
      "cbid": 211, "correlation": 204748
    }
  },
  {
    "ph": "s", "id": 204748, "pid": 76337, "tid": -914061504, "ts": 1716454224547739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224695312, "dur": 5,
    "args": {
      "External id": 204750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204750,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 204750, "pid": 5, "tid": 7, "ts": 1716454224695312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547753, "dur": 6,
    "args": {
      "External id": 204750,
      "cbid": 211, "correlation": 204750
    }
  },
  {
    "ph": "s", "id": 204750, "pid": 76337, "tid": -914061504, "ts": 1716454224547753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547813, "dur": 1,
    "args": {
      "External id": 204761,
      "cbid": 251, "correlation": 204761
    }
  },
  {
    "ph": "f", "id": 204761, "pid": 76337, "tid": -914061504, "ts": 1716454224547813,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547816, "dur": 0,
    "args": {
      "External id": 204762,
      "cbid": 251, "correlation": 204762
    }
  },
  {
    "ph": "f", "id": 204762, "pid": 76337, "tid": -914061504, "ts": 1716454224547816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224695318, "dur": 6,
    "args": {
      "External id": 204763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204763,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 204763, "pid": 5, "tid": 7, "ts": 1716454224695318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547818, "dur": 12,
    "args": {
      "External id": 204763,
      "cbid": 211, "correlation": 204763
    }
  },
  {
    "ph": "s", "id": 204763, "pid": 76337, "tid": -914061504, "ts": 1716454224547818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224695326, "dur": 3,
    "args": {
      "External id": 204765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204765,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 204765, "pid": 5, "tid": 7, "ts": 1716454224695326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547831, "dur": 6,
    "args": {
      "External id": 204765,
      "cbid": 211, "correlation": 204765
    }
  },
  {
    "ph": "s", "id": 204765, "pid": 76337, "tid": -914061504, "ts": 1716454224547831,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224695331, "dur": 155,
    "args": {
      "External id": 204786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204786,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204786, "pid": 5, "tid": 7, "ts": 1716454224695331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224547905, "dur": 12,
    "args": {
      "External id": 204786,
      "cbid": 211, "correlation": 204786
    }
  },
  {
    "ph": "s", "id": 204786, "pid": 76337, "tid": -914061504, "ts": 1716454224547905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548009, "dur": 1,
    "args": {
      "External id": 204804,
      "cbid": 251, "correlation": 204804
    }
  },
  {
    "ph": "f", "id": 204804, "pid": 76337, "tid": -914061504, "ts": 1716454224548009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224695487, "dur": 107,
    "args": {
      "External id": 204806,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204806,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204806, "pid": 5, "tid": 7, "ts": 1716454224695487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548014, "dur": 14,
    "args": {
      "External id": 204806,
      "cbid": 211, "correlation": 204806
    }
  },
  {
    "ph": "s", "id": 204806, "pid": 76337, "tid": -914061504, "ts": 1716454224548014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224695596, "dur": 34,
    "args": {
      "External id": 204814,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204814,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204814, "pid": 5, "tid": 7, "ts": 1716454224695596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548085, "dur": 12,
    "args": {
      "External id": 204814,
      "cbid": 211, "correlation": 204814
    }
  },
  {
    "ph": "s", "id": 204814, "pid": 76337, "tid": -914061504, "ts": 1716454224548085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224695631, "dur": 67,
    "args": {
      "External id": 204822,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204822,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204822, "pid": 5, "tid": 7, "ts": 1716454224695631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548126, "dur": 10,
    "args": {
      "External id": 204822,
      "cbid": 211, "correlation": 204822
    }
  },
  {
    "ph": "s", "id": 204822, "pid": 76337, "tid": -914061504, "ts": 1716454224548126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224695699, "dur": 93,
    "args": {
      "External id": 204844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204844,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204844, "pid": 5, "tid": 7, "ts": 1716454224695699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548177, "dur": 10,
    "args": {
      "External id": 204844,
      "cbid": 211, "correlation": 204844
    }
  },
  {
    "ph": "s", "id": 204844, "pid": 76337, "tid": -914061504, "ts": 1716454224548177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548263, "dur": 1,
    "args": {
      "External id": 204860,
      "cbid": 251, "correlation": 204860
    }
  },
  {
    "ph": "f", "id": 204860, "pid": 76337, "tid": -914061504, "ts": 1716454224548263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224695794, "dur": 580,
    "args": {
      "External id": 204862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204862,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 204862, "pid": 5, "tid": 7, "ts": 1716454224695794,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548269, "dur": 12,
    "args": {
      "External id": 204862,
      "cbid": 211, "correlation": 204862
    }
  },
  {
    "ph": "s", "id": 204862, "pid": 76337, "tid": -914061504, "ts": 1716454224548269,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224696375, "dur": 248,
    "args": {
      "External id": 204870,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204870,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204870, "pid": 5, "tid": 7, "ts": 1716454224696375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548334, "dur": 13,
    "args": {
      "External id": 204870,
      "cbid": 211, "correlation": 204870
    }
  },
  {
    "ph": "s", "id": 204870, "pid": 76337, "tid": -914061504, "ts": 1716454224548334,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224696624, "dur": 250,
    "args": {
      "External id": 204878,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204878,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204878, "pid": 5, "tid": 7, "ts": 1716454224696624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548365, "dur": 8,
    "args": {
      "External id": 204878,
      "cbid": 211, "correlation": 204878
    }
  },
  {
    "ph": "s", "id": 204878, "pid": 76337, "tid": -914061504, "ts": 1716454224548365,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548447, "dur": 1,
    "args": {
      "External id": 204894,
      "cbid": 251, "correlation": 204894
    }
  },
  {
    "ph": "f", "id": 204894, "pid": 76337, "tid": -914061504, "ts": 1716454224548447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548452, "dur": 0,
    "args": {
      "External id": 204896,
      "cbid": 251, "correlation": 204896
    }
  },
  {
    "ph": "f", "id": 204896, "pid": 76337, "tid": -914061504, "ts": 1716454224548452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224696875, "dur": 360,
    "args": {
      "External id": 204897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204897,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 204897, "pid": 5, "tid": 7, "ts": 1716454224696875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548454, "dur": 12,
    "args": {
      "External id": 204897,
      "cbid": 211, "correlation": 204897
    }
  },
  {
    "ph": "s", "id": 204897, "pid": 76337, "tid": -914061504, "ts": 1716454224548454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224697236, "dur": 50,
    "args": {
      "External id": 204905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204905,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204905, "pid": 5, "tid": 7, "ts": 1716454224697236,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548497, "dur": 195,
    "args": {
      "External id": 204905,
      "cbid": 211, "correlation": 204905
    }
  },
  {
    "ph": "s", "id": 204905, "pid": 76337, "tid": -914061504, "ts": 1716454224548497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224697288, "dur": 159,
    "args": {
      "External id": 204916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204916,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204916, "pid": 5, "tid": 7, "ts": 1716454224697288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548750, "dur": 70,
    "args": {
      "External id": 204916,
      "cbid": 211, "correlation": 204916
    }
  },
  {
    "ph": "s", "id": 204916, "pid": 76337, "tid": -914061504, "ts": 1716454224548750,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548873, "dur": 0,
    "args": {
      "External id": 204928,
      "cbid": 317, "correlation": 204928
    }
  },
  {
    "ph": "f", "id": 204928, "pid": 76337, "tid": -914061504, "ts": 1716454224548873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548874, "dur": 0,
    "args": {
      "External id": 204929,
      "cbid": 203, "correlation": 204929
    }
  },
  {
    "ph": "f", "id": 204929, "pid": 76337, "tid": -914061504, "ts": 1716454224548874,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548875, "dur": 0,
    "args": {
      "External id": 204930,
      "cbid": 205, "correlation": 204930
    }
  },
  {
    "ph": "f", "id": 204930, "pid": 76337, "tid": -914061504, "ts": 1716454224548875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548898, "dur": 1,
    "args": {
      "External id": 204934,
      "cbid": 251, "correlation": 204934
    }
  },
  {
    "ph": "f", "id": 204934, "pid": 76337, "tid": -914061504, "ts": 1716454224548898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548900, "dur": 0,
    "args": {
      "External id": 204935,
      "cbid": 251, "correlation": 204935
    }
  },
  {
    "ph": "f", "id": 204935, "pid": 76337, "tid": -914061504, "ts": 1716454224548900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548900, "dur": 0,
    "args": {
      "External id": 204936,
      "cbid": 251, "correlation": 204936
    }
  },
  {
    "ph": "f", "id": 204936, "pid": 76337, "tid": -914061504, "ts": 1716454224548900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548901, "dur": 0,
    "args": {
      "External id": 204937,
      "cbid": 251, "correlation": 204937
    }
  },
  {
    "ph": "f", "id": 204937, "pid": 76337, "tid": -914061504, "ts": 1716454224548901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548902, "dur": 1,
    "args": {
      "External id": 204938,
      "cbid": 251, "correlation": 204938
    }
  },
  {
    "ph": "f", "id": 204938, "pid": 76337, "tid": -914061504, "ts": 1716454224548902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548903, "dur": 0,
    "args": {
      "External id": 204939,
      "cbid": 251, "correlation": 204939
    }
  },
  {
    "ph": "f", "id": 204939, "pid": 76337, "tid": -914061504, "ts": 1716454224548903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548904, "dur": 0,
    "args": {
      "External id": 204940,
      "cbid": 251, "correlation": 204940
    }
  },
  {
    "ph": "f", "id": 204940, "pid": 76337, "tid": -914061504, "ts": 1716454224548904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548905, "dur": 0,
    "args": {
      "External id": 204941,
      "cbid": 251, "correlation": 204941
    }
  },
  {
    "ph": "f", "id": 204941, "pid": 76337, "tid": -914061504, "ts": 1716454224548905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548906, "dur": 0,
    "args": {
      "External id": 204942,
      "cbid": 251, "correlation": 204942
    }
  },
  {
    "ph": "f", "id": 204942, "pid": 76337, "tid": -914061504, "ts": 1716454224548906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224697448, "dur": 115,
    "args": {
      "External id": 204943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204943,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 204943, "pid": 5, "tid": 7, "ts": 1716454224697448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548908, "dur": 41,
    "args": {
      "External id": 204943,
      "cbid": 211, "correlation": 204943
    }
  },
  {
    "ph": "s", "id": 204943, "pid": 76337, "tid": -914061504, "ts": 1716454224548908,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224697565, "dur": 60,
    "args": {
      "External id": 204949,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204949,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204949, "pid": 5, "tid": 7, "ts": 1716454224697565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224548971, "dur": 280,
    "args": {
      "External id": 204949,
      "cbid": 211, "correlation": 204949
    }
  },
  {
    "ph": "s", "id": 204949, "pid": 76337, "tid": -914061504, "ts": 1716454224548971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224697626, "dur": 51,
    "args": {
      "External id": 204957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204957,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204957, "pid": 5, "tid": 7, "ts": 1716454224697626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549276, "dur": 9,
    "args": {
      "External id": 204957,
      "cbid": 211, "correlation": 204957
    }
  },
  {
    "ph": "s", "id": 204957, "pid": 76337, "tid": -914061504, "ts": 1716454224549276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224697678, "dur": 53,
    "args": {
      "External id": 204977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204977,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 204977, "pid": 5, "tid": 7, "ts": 1716454224697678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549356, "dur": 12,
    "args": {
      "External id": 204977,
      "cbid": 211, "correlation": 204977
    }
  },
  {
    "ph": "s", "id": 204977, "pid": 76337, "tid": -914061504, "ts": 1716454224549356,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224697732, "dur": 4,
    "args": {
      "External id": 204989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204989,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 204989, "pid": 5, "tid": 7, "ts": 1716454224697732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549378, "dur": 6,
    "args": {
      "External id": 204989,
      "cbid": 211, "correlation": 204989
    }
  },
  {
    "ph": "s", "id": 204989, "pid": 76337, "tid": -914061504, "ts": 1716454224549378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224697738, "dur": 56,
    "args": {
      "External id": 204992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 204992,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 204992, "pid": 5, "tid": 7, "ts": 1716454224697738,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549396, "dur": 110,
    "args": {
      "External id": 204992,
      "cbid": 211, "correlation": 204992
    }
  },
  {
    "ph": "s", "id": 204992, "pid": 76337, "tid": -914061504, "ts": 1716454224549396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224697795, "dur": 38,
    "args": {
      "External id": 205001,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205001,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205001, "pid": 5, "tid": 7, "ts": 1716454224697795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549547, "dur": 10,
    "args": {
      "External id": 205001,
      "cbid": 211, "correlation": 205001
    }
  },
  {
    "ph": "s", "id": 205001, "pid": 76337, "tid": -914061504, "ts": 1716454224549547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549602, "dur": 0,
    "args": {
      "External id": 205011,
      "cbid": 317, "correlation": 205011
    }
  },
  {
    "ph": "f", "id": 205011, "pid": 76337, "tid": -914061504, "ts": 1716454224549602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549604, "dur": 0,
    "args": {
      "External id": 205012,
      "cbid": 203, "correlation": 205012
    }
  },
  {
    "ph": "f", "id": 205012, "pid": 76337, "tid": -914061504, "ts": 1716454224549604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549604, "dur": 0,
    "args": {
      "External id": 205013,
      "cbid": 205, "correlation": 205013
    }
  },
  {
    "ph": "f", "id": 205013, "pid": 76337, "tid": -914061504, "ts": 1716454224549604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224697835, "dur": 42,
    "args": {
      "External id": 205017,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205017,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205017, "pid": 5, "tid": 7, "ts": 1716454224697835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549620, "dur": 13,
    "args": {
      "External id": 205017,
      "cbid": 211, "correlation": 205017
    }
  },
  {
    "ph": "s", "id": 205017, "pid": 76337, "tid": -914061504, "ts": 1716454224549620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224697878, "dur": 3,
    "args": {
      "External id": 205019,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205019,
      "bytes": 46080, "memory bandwidth (GB/s)": 11.900826446280991
    }
  },
  {
    "ph": "f", "id": 205019, "pid": 5, "tid": 7, "ts": 1716454224697878,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549636, "dur": 16,
    "args": {
      "External id": 205019,
      "cbid": 51, "correlation": 205019
    }
  },
  {
    "ph": "s", "id": 205019, "pid": 76337, "tid": -914061504, "ts": 1716454224549636,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549662, "dur": 2,
    "args": {
      "External id": 205021,
      "cbid": 200, "correlation": 205021
    }
  },
  {
    "ph": "f", "id": 205021, "pid": 76337, "tid": -914061504, "ts": 1716454224549662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549664, "dur": 0,
    "args": {
      "External id": 205022,
      "cbid": 200, "correlation": 205022
    }
  },
  {
    "ph": "f", "id": 205022, "pid": 76337, "tid": -914061504, "ts": 1716454224549664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549665, "dur": 0,
    "args": {
      "External id": 205023,
      "cbid": 200, "correlation": 205023
    }
  },
  {
    "ph": "f", "id": 205023, "pid": 76337, "tid": -914061504, "ts": 1716454224549665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549665, "dur": 0,
    "args": {
      "External id": 205024,
      "cbid": 200, "correlation": 205024
    }
  },
  {
    "ph": "f", "id": 205024, "pid": 76337, "tid": -914061504, "ts": 1716454224549665,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaFuncGetAttributes", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549667, "dur": 4,
    "args": {
      "External id": 205025,
      "cbid": 15, "correlation": 205025
    }
  },
  {
    "ph": "f", "id": 205025, "pid": 76337, "tid": -914061504, "ts": 1716454224549667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549672, "dur": 1,
    "args": {
      "External id": 205026,
      "cbid": 251, "correlation": 205026
    }
  },
  {
    "ph": "f", "id": 205026, "pid": 76337, "tid": -914061504, "ts": 1716454224549672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void tensorTransformGeneric<__half, __half, float, true, false, false, (cudnnKernelDataType_t)0>(cudnnTensorTransformStruct, tensorTransformParams, int, unsigned long, __half const*, __half*, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224697883, "dur": 25,
    "args": {
      "External id": 205027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205027,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 2,
      "warps per SM": 64,
      "grid": [160, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205027, "pid": 5, "tid": 7, "ts": 1716454224697883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549675, "dur": 9,
    "args": {
      "External id": 205027,
      "cbid": 211, "correlation": 205027
    }
  },
  {
    "ph": "s", "id": 205027, "pid": 76337, "tid": -914061504, "ts": 1716454224549675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224697909, "dur": 4,
    "args": {
      "External id": 205029,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205029,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 205029, "pid": 5, "tid": 7, "ts": 1716454224697909,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549690, "dur": 6,
    "args": {
      "External id": 205029,
      "cbid": 211, "correlation": 205029
    }
  },
  {
    "ph": "s", "id": 205029, "pid": 76337, "tid": -914061504, "ts": 1716454224549690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549699, "dur": 0,
    "args": {
      "External id": 205030,
      "cbid": 51, "correlation": 205030
    }
  },
  {
    "ph": "s", "id": 205030, "pid": 76337, "tid": -914061504, "ts": 1716454224549699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224697914, "dur": 191,
    "args": {
      "External id": 205031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205031,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205031, "pid": 5, "tid": 7, "ts": 1716454224697914,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549700, "dur": 199,
    "args": {
      "External id": 205031,
      "cbid": 211, "correlation": 205031
    }
  },
  {
    "ph": "s", "id": 205031, "pid": 76337, "tid": -914061504, "ts": 1716454224549700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224698107, "dur": 7,
    "args": {
      "External id": 205032,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205032,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205032, "pid": 5, "tid": 7, "ts": 1716454224698107,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549902, "dur": 5,
    "args": {
      "External id": 205032,
      "cbid": 211, "correlation": 205032
    }
  },
  {
    "ph": "s", "id": 205032, "pid": 76337, "tid": -914061504, "ts": 1716454224549902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224698115, "dur": 5,
    "args": {
      "External id": 205038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205038,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [192, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 15
    }
  },
  {
    "ph": "f", "id": 205038, "pid": 5, "tid": 7, "ts": 1716454224698115,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224549931, "dur": 9,
    "args": {
      "External id": 205038,
      "cbid": 211, "correlation": 205038
    }
  },
  {
    "ph": "s", "id": 205038, "pid": 76337, "tid": -914061504, "ts": 1716454224549931,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224698121, "dur": 3,
    "args": {
      "External id": 205046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205046,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205046, "pid": 5, "tid": 7, "ts": 1716454224698121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224551623, "dur": 15,
    "args": {
      "External id": 205046,
      "cbid": 211, "correlation": 205046
    }
  },
  {
    "ph": "s", "id": 205046, "pid": 76337, "tid": -914061504, "ts": 1716454224551623,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224698125, "dur": 3,
    "args": {
      "External id": 205054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205054,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205054, "pid": 5, "tid": 7, "ts": 1716454224698125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224551664, "dur": 10,
    "args": {
      "External id": 205054,
      "cbid": 211, "correlation": 205054
    }
  },
  {
    "ph": "s", "id": 205054, "pid": 76337, "tid": -914061504, "ts": 1716454224551664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224698130, "dur": 3,
    "args": {
      "External id": 205062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205062,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205062, "pid": 5, "tid": 7, "ts": 1716454224698130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224551691, "dur": 8,
    "args": {
      "External id": 205062,
      "cbid": 211, "correlation": 205062
    }
  },
  {
    "ph": "s", "id": 205062, "pid": 76337, "tid": -914061504, "ts": 1716454224551691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224698134, "dur": 3,
    "args": {
      "External id": 205071,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205071,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205071, "pid": 5, "tid": 7, "ts": 1716454224698134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224551863, "dur": 14,
    "args": {
      "External id": 205071,
      "cbid": 211, "correlation": 205071
    }
  },
  {
    "ph": "s", "id": 205071, "pid": 76337, "tid": -914061504, "ts": 1716454224551863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224698138, "dur": 3,
    "args": {
      "External id": 205080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205080,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205080, "pid": 5, "tid": 7, "ts": 1716454224698138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224551892, "dur": 7,
    "args": {
      "External id": 205080,
      "cbid": 211, "correlation": 205080
    }
  },
  {
    "ph": "s", "id": 205080, "pid": 76337, "tid": -914061504, "ts": 1716454224551892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224698142, "dur": 3,
    "args": {
      "External id": 205088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205088,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205088, "pid": 5, "tid": 7, "ts": 1716454224698142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224551918, "dur": 8,
    "args": {
      "External id": 205088,
      "cbid": 211, "correlation": 205088
    }
  },
  {
    "ph": "s", "id": 205088, "pid": 76337, "tid": -914061504, "ts": 1716454224551918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224698146, "dur": 3,
    "args": {
      "External id": 205096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205096,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205096, "pid": 5, "tid": 7, "ts": 1716454224698146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224552183, "dur": 15,
    "args": {
      "External id": 205096,
      "cbid": 211, "correlation": 205096
    }
  },
  {
    "ph": "s", "id": 205096, "pid": 76337, "tid": -914061504, "ts": 1716454224552183,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224698150, "dur": 3,
    "args": {
      "External id": 205104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205104,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205104, "pid": 5, "tid": 7, "ts": 1716454224698150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224552215, "dur": 8,
    "args": {
      "External id": 205104,
      "cbid": 211, "correlation": 205104
    }
  },
  {
    "ph": "s", "id": 205104, "pid": 76337, "tid": -914061504, "ts": 1716454224552215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224698155, "dur": 1,
    "args": {
      "External id": 205114,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205114,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 205114, "pid": 5, "tid": 7, "ts": 1716454224698155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224552280, "dur": 35,
    "args": {
      "External id": 205114,
      "cbid": 41, "correlation": 205114
    }
  },
  {
    "ph": "s", "id": 205114, "pid": 76337, "tid": -914061504, "ts": 1716454224552280,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224552316, "dur": 145856,
    "args": {
      "External id": 205115,
      "cbid": 131, "correlation": 205115
    }
  },
  {
    "ph": "f", "id": 205115, "pid": 76337, "tid": -914061504, "ts": 1716454224552316,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224698360, "dur": 3,
    "args": {
      "External id": 205123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205123,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205123, "pid": 5, "tid": 7, "ts": 1716454224698360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698332, "dur": 31,
    "args": {
      "External id": 205123,
      "cbid": 211, "correlation": 205123
    }
  },
  {
    "ph": "s", "id": 205123, "pid": 76337, "tid": -914061504, "ts": 1716454224698332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224698458, "dur": 3,
    "args": {
      "External id": 205132,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205132,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002380952380952381
    }
  },
  {
    "ph": "f", "id": 205132, "pid": 5, "tid": 7, "ts": 1716454224698458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698426, "dur": 33,
    "args": {
      "External id": 205132,
      "cbid": 41, "correlation": 205132
    }
  },
  {
    "ph": "s", "id": 205132, "pid": 76337, "tid": -914061504, "ts": 1716454224698426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224698565, "dur": 4,
    "args": {
      "External id": 205142,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205142,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205142, "pid": 5, "tid": 7, "ts": 1716454224698565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698545, "dur": 22,
    "args": {
      "External id": 205142,
      "cbid": 211, "correlation": 205142
    }
  },
  {
    "ph": "s", "id": 205142, "pid": 76337, "tid": -914061504, "ts": 1716454224698545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224698645, "dur": 1,
    "args": {
      "External id": 205152,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205152,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 205152, "pid": 5, "tid": 7, "ts": 1716454224698645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698626, "dur": 17,
    "args": {
      "External id": 205152,
      "cbid": 41, "correlation": 205152
    }
  },
  {
    "ph": "s", "id": 205152, "pid": 76337, "tid": -914061504, "ts": 1716454224698626,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698644, "dur": 8,
    "args": {
      "External id": 205153,
      "cbid": 131, "correlation": 205153
    }
  },
  {
    "ph": "f", "id": 205153, "pid": 76337, "tid": -914061504, "ts": 1716454224698644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224698722, "dur": 3,
    "args": {
      "External id": 205160,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205160,
      "bytes": 98304, "memory bandwidth (GB/s)": 30.415841584158414
    }
  },
  {
    "ph": "f", "id": 205160, "pid": 5, "tid": 7, "ts": 1716454224698722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698694, "dur": 28,
    "args": {
      "External id": 205160,
      "cbid": 41, "correlation": 205160
    }
  },
  {
    "ph": "s", "id": 205160, "pid": 76337, "tid": -914061504, "ts": 1716454224698694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224698815, "dur": 3,
    "args": {
      "External id": 205179,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205179,
      "bytes": 16, "memory bandwidth (GB/s)": 0.005263157894736842
    }
  },
  {
    "ph": "f", "id": 205179, "pid": 5, "tid": 7, "ts": 1716454224698815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698796, "dur": 18,
    "args": {
      "External id": 205179,
      "cbid": 41, "correlation": 205179
    }
  },
  {
    "ph": "s", "id": 205179, "pid": 76337, "tid": -914061504, "ts": 1716454224698796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_set_info(int, int, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224698853, "dur": 3,
    "args": {
      "External id": 205185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205185,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205185, "pid": 5, "tid": 7, "ts": 1716454224698853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698842, "dur": 11,
    "args": {
      "External id": 205185,
      "cbid": 211, "correlation": 205185
    }
  },
  {
    "ph": "s", "id": 205185, "pid": 76337, "tid": -914061504, "ts": 1716454224698842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void getf2_cta_32x32<float, float>(int, int, int, int, float*, int, int*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224698868, "dur": 6,
    "args": {
      "External id": 205187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205187,
      "registers per thread": 26,
      "shared memory": 4480,
      "blocks per SM": 0.0125,
      "warps per SM": 0.4,
      "grid": [1, 1, 1],
      "block": [1024, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 205187, "pid": 5, "tid": 7, "ts": 1716454224698868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698857, "dur": 10,
    "args": {
      "External id": 205187,
      "cbid": 211, "correlation": 205187
    }
  },
  {
    "ph": "s", "id": 205187, "pid": 76337, "tid": -914061504, "ts": 1716454224698857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "getrf2_reset_info(int*)", "pid": 5, "tid": 7,
    "ts": 1716454224698877, "dur": 3,
    "args": {
      "External id": 205189,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205189,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.000390625,
      "grid": [1, 1, 1],
      "block": [1, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205189, "pid": 5, "tid": 7, "ts": 1716454224698877,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698869, "dur": 7,
    "args": {
      "External id": 205189,
      "cbid": 211, "correlation": 205189
    }
  },
  {
    "ph": "s", "id": 205189, "pid": 76337, "tid": -914061504, "ts": 1716454224698869,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224698911, "dur": 2,
    "args": {
      "External id": 205197,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205197,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002717391304347826
    }
  },
  {
    "ph": "f", "id": 205197, "pid": 5, "tid": 7, "ts": 1716454224698911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698897, "dur": 13,
    "args": {
      "External id": 205197,
      "cbid": 41, "correlation": 205197
    }
  },
  {
    "ph": "s", "id": 205197, "pid": 76337, "tid": -914061504, "ts": 1716454224698897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224698956, "dur": 3,
    "args": {
      "External id": 205211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205211,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205211, "pid": 5, "tid": 7, "ts": 1716454224698956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698946, "dur": 12,
    "args": {
      "External id": 205211,
      "cbid": 211, "correlation": 205211
    }
  },
  {
    "ph": "s", "id": 205211, "pid": 76337, "tid": -914061504, "ts": 1716454224698946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#4}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224698986, "dur": 2,
    "args": {
      "External id": 205225,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205225,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.025,
      "grid": [1, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205225, "pid": 5, "tid": 7, "ts": 1716454224698986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224698971, "dur": 15,
    "args": {
      "External id": 205225,
      "cbid": 211, "correlation": 205225
    }
  },
  {
    "ph": "s", "id": 205225, "pid": 76337, "tid": -914061504, "ts": 1716454224698971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_lower_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224699022, "dur": 6,
    "args": {
      "External id": 205232,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205232,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205232, "pid": 5, "tid": 7, "ts": 1716454224699022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699012, "dur": 11,
    "args": {
      "External id": 205232,
      "cbid": 211, "correlation": 205232
    }
  },
  {
    "ph": "s", "id": 205232, "pid": 76337, "tid": -914061504, "ts": 1716454224699012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void trsm_batch_left_upper_kernel<float>(cublasTrsmBatchParams<float>, float const* const*, float* const*, float const*, float)", "pid": 5, "tid": 7,
    "ts": 1716454224699033, "dur": 6,
    "args": {
      "External id": 205235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205235,
      "registers per thread": 56,
      "shared memory": 192,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [32, 8, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205235, "pid": 5, "tid": 7, "ts": 1716454224699033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699025, "dur": 7,
    "args": {
      "External id": 205235,
      "cbid": 211, "correlation": 205235
    }
  },
  {
    "ph": "s", "id": 205235, "pid": 76337, "tid": -914061504, "ts": 1716454224699025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void laswp_kernel_reverse<float, false>(int, float* const*, int, int, int, int const*, int, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224699042, "dur": 3,
    "args": {
      "External id": 205237,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205237,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [1, 256, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205237, "pid": 5, "tid": 7, "ts": 1716454224699042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699034, "dur": 7,
    "args": {
      "External id": 205237,
      "cbid": 211, "correlation": 205237
    }
  },
  {
    "ph": "s", "id": 205237, "pid": 76337, "tid": -914061504, "ts": 1716454224699034,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224699064, "dur": 2,
    "args": {
      "External id": 205240,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205240,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0027472527472527475
    }
  },
  {
    "ph": "f", "id": 205240, "pid": 5, "tid": 7, "ts": 1716454224699064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699050, "dur": 13,
    "args": {
      "External id": 205240,
      "cbid": 41, "correlation": 205240
    }
  },
  {
    "ph": "s", "id": 205240, "pid": 76337, "tid": -914061504, "ts": 1716454224699050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#11}::operator()() const::{lambda(bool)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224699121, "dur": 4,
    "args": {
      "External id": 205256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205256,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205256, "pid": 5, "tid": 7, "ts": 1716454224699121,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699108, "dur": 13,
    "args": {
      "External id": 205256,
      "cbid": 211, "correlation": 205256
    }
  },
  {
    "ph": "s", "id": 205256, "pid": 76337, "tid": -914061504, "ts": 1716454224699108,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224699141, "dur": 3,
    "args": {
      "External id": 205261,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205261,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0003125
    }
  },
  {
    "ph": "f", "id": 205261, "pid": 5, "tid": 7, "ts": 1716454224699141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699126, "dur": 13,
    "args": {
      "External id": 205261,
      "cbid": 41, "correlation": 205261
    }
  },
  {
    "ph": "s", "id": 205261, "pid": 76337, "tid": -914061504, "ts": 1716454224699126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224699172, "dur": 1,
    "args": {
      "External id": 205267,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205267,
      "bytes": 1, "memory bandwidth (GB/s)": 0.0005896226415094339
    }
  },
  {
    "ph": "f", "id": 205267, "pid": 5, "tid": 7, "ts": 1716454224699172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699150, "dur": 31,
    "args": {
      "External id": 205267,
      "cbid": 41, "correlation": 205267
    }
  },
  {
    "ph": "s", "id": 205267, "pid": 76337, "tid": -914061504, "ts": 1716454224699150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699181, "dur": 3,
    "args": {
      "External id": 205268,
      "cbid": 131, "correlation": 205268
    }
  },
  {
    "ph": "f", "id": 205268, "pid": 76337, "tid": -914061504, "ts": 1716454224699181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224699243, "dur": 3,
    "args": {
      "External id": 205276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205276,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205276, "pid": 5, "tid": 7, "ts": 1716454224699243,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699225, "dur": 18,
    "args": {
      "External id": 205276,
      "cbid": 211, "correlation": 205276
    }
  },
  {
    "ph": "s", "id": 205276, "pid": 76337, "tid": -914061504, "ts": 1716454224699225,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224699280, "dur": 3,
    "args": {
      "External id": 205286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205286,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205286, "pid": 5, "tid": 7, "ts": 1716454224699280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699267, "dur": 12,
    "args": {
      "External id": 205286,
      "cbid": 211, "correlation": 205286
    }
  },
  {
    "ph": "s", "id": 205286, "pid": 76337, "tid": -914061504, "ts": 1716454224699267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224699307, "dur": 3,
    "args": {
      "External id": 205295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205295,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205295, "pid": 5, "tid": 7, "ts": 1716454224699307,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699297, "dur": 9,
    "args": {
      "External id": 205295,
      "cbid": 211, "correlation": 205295
    }
  },
  {
    "ph": "s", "id": 205295, "pid": 76337, "tid": -914061504, "ts": 1716454224699297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224699428, "dur": 12,
    "args": {
      "External id": 205305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205305,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205305, "pid": 5, "tid": 7, "ts": 1716454224699428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699410, "dur": 19,
    "args": {
      "External id": 205305,
      "cbid": 211, "correlation": 205305
    }
  },
  {
    "ph": "s", "id": 205305, "pid": 76337, "tid": -914061504, "ts": 1716454224699410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224699470, "dur": 3,
    "args": {
      "External id": 205313,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205313,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205313, "pid": 5, "tid": 7, "ts": 1716454224699470,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699460, "dur": 9,
    "args": {
      "External id": 205313,
      "cbid": 211, "correlation": 205313
    }
  },
  {
    "ph": "s", "id": 205313, "pid": 76337, "tid": -914061504, "ts": 1716454224699460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224699517, "dur": 12,
    "args": {
      "External id": 205323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205323,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205323, "pid": 5, "tid": 7, "ts": 1716454224699517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699503, "dur": 14,
    "args": {
      "External id": 205323,
      "cbid": 211, "correlation": 205323
    }
  },
  {
    "ph": "s", "id": 205323, "pid": 76337, "tid": -914061504, "ts": 1716454224699503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224699550, "dur": 10,
    "args": {
      "External id": 205331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205331,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205331, "pid": 5, "tid": 7, "ts": 1716454224699550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699540, "dur": 9,
    "args": {
      "External id": 205331,
      "cbid": 211, "correlation": 205331
    }
  },
  {
    "ph": "s", "id": 205331, "pid": 76337, "tid": -914061504, "ts": 1716454224699540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224699581, "dur": 3,
    "args": {
      "External id": 205340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205340,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205340, "pid": 5, "tid": 7, "ts": 1716454224699581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699570, "dur": 12,
    "args": {
      "External id": 205340,
      "cbid": 211, "correlation": 205340
    }
  },
  {
    "ph": "s", "id": 205340, "pid": 76337, "tid": -914061504, "ts": 1716454224699570,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1> >(int, at::native::CUDAFunctor_add<float>, at::detail::Array<char*, 3>, TrivialOffsetCalculator<2, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<2>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224699610, "dur": 5,
    "args": {
      "External id": 205349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205349,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205349, "pid": 5, "tid": 7, "ts": 1716454224699610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699601, "dur": 8,
    "args": {
      "External id": 205349,
      "cbid": 211, "correlation": 205349
    }
  },
  {
    "ph": "s", "id": 205349, "pid": 76337, "tid": -914061504, "ts": 1716454224699601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224699650, "dur": 8,
    "args": {
      "External id": 205359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205359,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205359, "pid": 5, "tid": 7, "ts": 1716454224699650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699639, "dur": 10,
    "args": {
      "External id": 205359,
      "cbid": 211, "correlation": 205359
    }
  },
  {
    "ph": "s", "id": 205359, "pid": 76337, "tid": -914061504, "ts": 1716454224699639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224699973, "dur": 3,
    "args": {
      "External id": 205368,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205368,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205368, "pid": 5, "tid": 7, "ts": 1716454224699973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224699957, "dur": 24,
    "args": {
      "External id": 205368,
      "cbid": 211, "correlation": 205368
    }
  },
  {
    "ph": "s", "id": 205368, "pid": 76337, "tid": -914061504, "ts": 1716454224699957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224700015, "dur": 3,
    "args": {
      "External id": 205376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205376,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205376, "pid": 5, "tid": 7, "ts": 1716454224700015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700004, "dur": 11,
    "args": {
      "External id": 205376,
      "cbid": 211, "correlation": 205376
    }
  },
  {
    "ph": "s", "id": 205376, "pid": 76337, "tid": -914061504, "ts": 1716454224700004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700067, "dur": 1,
    "args": {
      "External id": 205386,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205386,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005211726384364821
    }
  },
  {
    "ph": "f", "id": 205386, "pid": 5, "tid": 7, "ts": 1716454224700067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700052, "dur": 12,
    "args": {
      "External id": 205386,
      "cbid": 41, "correlation": 205386
    }
  },
  {
    "ph": "s", "id": 205386, "pid": 76337, "tid": -914061504, "ts": 1716454224700052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700066, "dur": 8,
    "args": {
      "External id": 205387,
      "cbid": 131, "correlation": 205387
    }
  },
  {
    "ph": "f", "id": 205387, "pid": 76337, "tid": -914061504, "ts": 1716454224700066,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::FillFunctor<float>, at::detail::Array<char*, 1> >(int, at::native::FillFunctor<float>, at::detail::Array<char*, 1>)", "pid": 5, "tid": 7,
    "ts": 1716454224700155, "dur": 2,
    "args": {
      "External id": 205395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205395, "pid": 5, "tid": 7, "ts": 1716454224700155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700141, "dur": 14,
    "args": {
      "External id": 205395,
      "cbid": 211, "correlation": 205395
    }
  },
  {
    "ph": "s", "id": 205395, "pid": 76337, "tid": -914061504, "ts": 1716454224700141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700227, "dur": 3,
    "args": {
      "External id": 205404,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205404,
      "bytes": 8, "memory bandwidth (GB/s)": 0.002551020408163265
    }
  },
  {
    "ph": "f", "id": 205404, "pid": 5, "tid": 7, "ts": 1716454224700227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700209, "dur": 18,
    "args": {
      "External id": 205404,
      "cbid": 41, "correlation": 205404
    }
  },
  {
    "ph": "s", "id": 205404, "pid": 76337, "tid": -914061504, "ts": 1716454224700209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 1, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224700297, "dur": 3,
    "args": {
      "External id": 205414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205414,
      "registers per thread": 18,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205414, "pid": 5, "tid": 7, "ts": 1716454224700297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700283, "dur": 13,
    "args": {
      "External id": 205414,
      "cbid": 211, "correlation": 205414
    }
  },
  {
    "ph": "s", "id": 205414, "pid": 76337, "tid": -914061504, "ts": 1716454224700283,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700349, "dur": 1,
    "args": {
      "External id": 205424,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205424,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005208333333333333
    }
  },
  {
    "ph": "f", "id": 205424, "pid": 5, "tid": 7, "ts": 1716454224700349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700335, "dur": 12,
    "args": {
      "External id": 205424,
      "cbid": 41, "correlation": 205424
    }
  },
  {
    "ph": "s", "id": 205424, "pid": 76337, "tid": -914061504, "ts": 1716454224700335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700348, "dur": 8,
    "args": {
      "External id": 205425,
      "cbid": 131, "correlation": 205425
    }
  },
  {
    "ph": "f", "id": 205425, "pid": 76337, "tid": -914061504, "ts": 1716454224700348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoD (Device -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700410, "dur": 3,
    "args": {
      "External id": 205432,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205432,
      "bytes": 98304, "memory bandwidth (GB/s)": 31.670103092783506
    }
  },
  {
    "ph": "f", "id": 205432, "pid": 5, "tid": 7, "ts": 1716454224700410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700390, "dur": 19,
    "args": {
      "External id": 205432,
      "cbid": 41, "correlation": 205432
    }
  },
  {
    "ph": "s", "id": 205432, "pid": 76337, "tid": -914061504, "ts": 1716454224700390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700457, "dur": 1,
    "args": {
      "External id": 205443,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205443,
      "bytes": 2, "memory bandwidth (GB/s)": 0.0013020833333333333
    }
  },
  {
    "ph": "f", "id": 205443, "pid": 5, "tid": 7, "ts": 1716454224700457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700445, "dur": 10,
    "args": {
      "External id": 205443,
      "cbid": 41, "correlation": 205443
    }
  },
  {
    "ph": "s", "id": 205443, "pid": 76337, "tid": -914061504, "ts": 1716454224700445,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700455, "dur": 8,
    "args": {
      "External id": 205444,
      "cbid": 131, "correlation": 205444
    }
  },
  {
    "ph": "f", "id": 205444, "pid": 76337, "tid": -914061504, "ts": 1716454224700455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224700505, "dur": 3,
    "args": {
      "External id": 205452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205452,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205452, "pid": 5, "tid": 7, "ts": 1716454224700505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700492, "dur": 12,
    "args": {
      "External id": 205452,
      "cbid": 211, "correlation": 205452
    }
  },
  {
    "ph": "s", "id": 205452, "pid": 76337, "tid": -914061504, "ts": 1716454224700492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224700534, "dur": 3,
    "args": {
      "External id": 205462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205462,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205462, "pid": 5, "tid": 7, "ts": 1716454224700534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700525, "dur": 8,
    "args": {
      "External id": 205462,
      "cbid": 211, "correlation": 205462
    }
  },
  {
    "ph": "s", "id": 205462, "pid": 76337, "tid": -914061504, "ts": 1716454224700525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224700556, "dur": 3,
    "args": {
      "External id": 205471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205471,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205471, "pid": 5, "tid": 7, "ts": 1716454224700556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700547, "dur": 7,
    "args": {
      "External id": 205471,
      "cbid": 211, "correlation": 205471
    }
  },
  {
    "ph": "s", "id": 205471, "pid": 76337, "tid": -914061504, "ts": 1716454224700547,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224700628, "dur": 5,
    "args": {
      "External id": 205479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205479,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205479, "pid": 5, "tid": 7, "ts": 1716454224700628,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700611, "dur": 18,
    "args": {
      "External id": 205479,
      "cbid": 211, "correlation": 205479
    }
  },
  {
    "ph": "s", "id": 205479, "pid": 76337, "tid": -914061504, "ts": 1716454224700611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224700670, "dur": 3,
    "args": {
      "External id": 205488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205488,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205488, "pid": 5, "tid": 7, "ts": 1716454224700670,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700659, "dur": 10,
    "args": {
      "External id": 205488,
      "cbid": 211, "correlation": 205488
    }
  },
  {
    "ph": "s", "id": 205488, "pid": 76337, "tid": -914061504, "ts": 1716454224700659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224700694, "dur": 3,
    "args": {
      "External id": 205497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205497,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205497, "pid": 5, "tid": 7, "ts": 1716454224700694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700685, "dur": 7,
    "args": {
      "External id": 205497,
      "cbid": 211, "correlation": 205497
    }
  },
  {
    "ph": "s", "id": 205497, "pid": 76337, "tid": -914061504, "ts": 1716454224700685,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224700755, "dur": 3,
    "args": {
      "External id": 205505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205505,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205505, "pid": 5, "tid": 7, "ts": 1716454224700755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700744, "dur": 9,
    "args": {
      "External id": 205505,
      "cbid": 211, "correlation": 205505
    }
  },
  {
    "ph": "s", "id": 205505, "pid": 76337, "tid": -914061504, "ts": 1716454224700744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224700816, "dur": 1,
    "args": {
      "External id": 205513,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205513,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004464285714285714
    }
  },
  {
    "ph": "f", "id": 205513, "pid": 5, "tid": 7, "ts": 1716454224700816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700798, "dur": 28,
    "args": {
      "External id": 205513,
      "cbid": 41, "correlation": 205513
    }
  },
  {
    "ph": "s", "id": 205513, "pid": 76337, "tid": -914061504, "ts": 1716454224700798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700827, "dur": 3,
    "args": {
      "External id": 205514,
      "cbid": 131, "correlation": 205514
    }
  },
  {
    "ph": "f", "id": 205514, "pid": 76337, "tid": -914061504, "ts": 1716454224700827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700889, "dur": 1,
    "args": {
      "External id": 205524,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205524,
      "bytes": 42, "memory bandwidth (GB/s)": 0.02734375
    }
  },
  {
    "ph": "f", "id": 205524, "pid": 5, "tid": 7, "ts": 1716454224700889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700876, "dur": 10,
    "args": {
      "External id": 205524,
      "cbid": 41, "correlation": 205524
    }
  },
  {
    "ph": "s", "id": 205524, "pid": 76337, "tid": -914061504, "ts": 1716454224700876,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700887, "dur": 7,
    "args": {
      "External id": 205525,
      "cbid": 131, "correlation": 205525
    }
  },
  {
    "ph": "f", "id": 205525, "pid": 76337, "tid": -914061504, "ts": 1716454224700887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224700943, "dur": 1,
    "args": {
      "External id": 205534,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205534,
      "bytes": 8, "memory bandwidth (GB/s)": 0.005319148936170213
    }
  },
  {
    "ph": "f", "id": 205534, "pid": 5, "tid": 7, "ts": 1716454224700943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700933, "dur": 8,
    "args": {
      "External id": 205534,
      "cbid": 41, "correlation": 205534
    }
  },
  {
    "ph": "s", "id": 205534, "pid": 76337, "tid": -914061504, "ts": 1716454224700933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224700942, "dur": 8,
    "args": {
      "External id": 205535,
      "cbid": 131, "correlation": 205535
    }
  },
  {
    "ph": "f", "id": 205535, "pid": 76337, "tid": -914061504, "ts": 1716454224700942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<long, long, bool, at::native::(anonymous namespace)::CompareEqFunctor<long> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224701029, "dur": 4,
    "args": {
      "External id": 205542,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205542,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205542, "pid": 5, "tid": 7, "ts": 1716454224701029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701009, "dur": 22,
    "args": {
      "External id": 205542,
      "cbid": 211, "correlation": 205542
    }
  },
  {
    "ph": "s", "id": 205542, "pid": 76337, "tid": -914061504, "ts": 1716454224701009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceReduceSingleTileKernel<at_cuda_detail::cub::DeviceReducePolicy<bool, int, int, at_cuda_detail::cub::Sum>::Policy600, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int>(at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, int*, int, at_cuda_detail::cub::Sum, int)", "pid": 5, "tid": 7,
    "ts": 1716454224701067, "dur": 4,
    "args": {
      "External id": 205562,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205562,
      "registers per thread": 40,
      "shared memory": 44,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205562, "pid": 5, "tid": 7, "ts": 1716454224701067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701055, "dur": 12,
    "args": {
      "External id": 205562,
      "cbid": 211, "correlation": 205562
    }
  },
  {
    "ph": "s", "id": 205562, "pid": 76337, "tid": -914061504, "ts": 1716454224701055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701068, "dur": 0,
    "args": {
      "External id": 205563,
      "cbid": 11, "correlation": 205563
    }
  },
  {
    "ph": "f", "id": 205563, "pid": 76337, "tid": -914061504, "ts": 1716454224701068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701069, "dur": 0,
    "args": {
      "External id": 205564,
      "cbid": 11, "correlation": 205564
    }
  },
  {
    "ph": "f", "id": 205564, "pid": 76337, "tid": -914061504, "ts": 1716454224701069,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224701082, "dur": 1,
    "args": {
      "External id": 205567,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205567,
      "bytes": 4, "memory bandwidth (GB/s)": 0.0023584905660377358
    }
  },
  {
    "ph": "f", "id": 205567, "pid": 5, "tid": 7, "ts": 1716454224701082,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701070, "dur": 21,
    "args": {
      "External id": 205567,
      "cbid": 41, "correlation": 205567
    }
  },
  {
    "ph": "s", "id": 205567, "pid": 76337, "tid": -914061504, "ts": 1716454224701070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701092, "dur": 3,
    "args": {
      "External id": 205568,
      "cbid": 131, "correlation": 205568
    }
  },
  {
    "ph": "f", "id": 205568, "pid": 76337, "tid": -914061504, "ts": 1716454224701092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceCompactInitKernel<at_cuda_detail::cub::ScanTileState<int, true>, int*>(at_cuda_detail::cub::ScanTileState<int, true>, int, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224701120, "dur": 3,
    "args": {
      "External id": 205592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205592,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205592, "pid": 5, "tid": 7, "ts": 1716454224701120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701110, "dur": 9,
    "args": {
      "External id": 205592,
      "cbid": 211, "correlation": 205592
    }
  },
  {
    "ph": "s", "id": 205592, "pid": 76337, "tid": -914061504, "ts": 1716454224701110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701120, "dur": 0,
    "args": {
      "External id": 205593,
      "cbid": 11, "correlation": 205593
    }
  },
  {
    "ph": "f", "id": 205593, "pid": 76337, "tid": -914061504, "ts": 1716454224701120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701120, "dur": 0,
    "args": {
      "External id": 205594,
      "cbid": 11, "correlation": 205594
    }
  },
  {
    "ph": "f", "id": 205594, "pid": 76337, "tid": -914061504, "ts": 1716454224701120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetAttribute", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701122, "dur": 1,
    "args": {
      "External id": 205596,
      "cbid": 200, "correlation": 205596
    }
  },
  {
    "ph": "f", "id": 205596, "pid": 76337, "tid": -914061504, "ts": 1716454224701122,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at_cuda_detail::cub::DeviceSelectSweepKernel<at_cuda_detail::cub::DispatchSelectIf<at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>::PtxSelectIfPolicyT, at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, false>(at_cuda_detail::cub::CountingInputIterator<long, long>, at_cuda_detail::cub::TransformInputIterator<bool, at::native::(anonymous namespace)::NonZeroOp<bool>, bool*, long>, long*, int*, at_cuda_detail::cub::ScanTileState<int, true>, at_cuda_detail::cub::NullType, at_cuda_detail::cub::NullType, int, int)", "pid": 5, "tid": 7,
    "ts": 1716454224701132, "dur": 4,
    "args": {
      "External id": 205598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205598,
      "registers per thread": 40,
      "shared memory": 5120,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205598, "pid": 5, "tid": 7, "ts": 1716454224701132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701125, "dur": 8,
    "args": {
      "External id": 205598,
      "cbid": 211, "correlation": 205598
    }
  },
  {
    "ph": "s", "id": 205598, "pid": 76337, "tid": -914061504, "ts": 1716454224701125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701133, "dur": 0,
    "args": {
      "External id": 205599,
      "cbid": 11, "correlation": 205599
    }
  },
  {
    "ph": "f", "id": 205599, "pid": 76337, "tid": -914061504, "ts": 1716454224701133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaPeekAtLastError", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701134, "dur": 0,
    "args": {
      "External id": 205600,
      "cbid": 11, "correlation": 205600
    }
  },
  {
    "ph": "f", "id": 205600, "pid": 76337, "tid": -914061504, "ts": 1716454224701134,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy DtoH (Device -> Pageable)", "pid": 5, "tid": 7,
    "ts": 1716454224701171, "dur": 1,
    "args": {
      "External id": 205607,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205607,
      "bytes": 8, "memory bandwidth (GB/s)": 0.004807692307692308
    }
  },
  {
    "ph": "f", "id": 205607, "pid": 5, "tid": 7, "ts": 1716454224701171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701159, "dur": 20,
    "args": {
      "External id": 205607,
      "cbid": 41, "correlation": 205607
    }
  },
  {
    "ph": "s", "id": 205607, "pid": 76337, "tid": -914061504, "ts": 1716454224701159,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701180, "dur": 3,
    "args": {
      "External id": 205608,
      "cbid": 131, "correlation": 205608
    }
  },
  {
    "ph": "f", "id": 205608, "pid": 76337, "tid": -914061504, "ts": 1716454224701180,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memcpy", "name": "Memcpy HtoD (Pageable -> Device)", "pid": 5, "tid": 7,
    "ts": 1716454224701230, "dur": 1,
    "args": {
      "External id": 205618,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 205618,
      "bytes": 8, "memory bandwidth (GB/s)": 0.0047169811320754715
    }
  },
  {
    "ph": "f", "id": 205618, "pid": 5, "tid": 7, "ts": 1716454224701230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemcpyAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701218, "dur": 9,
    "args": {
      "External id": 205618,
      "cbid": 41, "correlation": 205618
    }
  },
  {
    "ph": "s", "id": 205618, "pid": 76337, "tid": -914061504, "ts": 1716454224701218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamSynchronize", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701229, "dur": 8,
    "args": {
      "External id": 205619,
      "cbid": 131, "correlation": 205619
    }
  },
  {
    "ph": "f", "id": 205619, "pid": 76337, "tid": -914061504, "ts": 1716454224701229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::index_elementwise_kernel<128, 4, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_index_kernel<at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1}>(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>, at::native::index_kernel_impl<at::native::OpaqueType<2> >(at::TensorIteratorBase&, c10::ArrayRef<long>, c10::ArrayRef<long>)::{lambda(char*, char*, long)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224701299, "dur": 5,
    "args": {
      "External id": 205626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205626,
      "registers per thread": 40,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205626, "pid": 5, "tid": 7, "ts": 1716454224701299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701284, "dur": 16,
    "args": {
      "External id": 205626,
      "cbid": 211, "correlation": 205626
    }
  },
  {
    "ph": "s", "id": 205626, "pid": 76337, "tid": -914061504, "ts": 1716454224701284,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::pow_tensor_scalar_kernel_impl<c10::Half, c10::Half>(at::TensorIteratorBase&, c10::Half)::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224701370, "dur": 3,
    "args": {
      "External id": 205635,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205635,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205635, "pid": 5, "tid": 7, "ts": 1716454224701370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701357, "dur": 12,
    "args": {
      "External id": 205635,
      "cbid": 211, "correlation": 205635
    }
  },
  {
    "ph": "s", "id": 205635, "pid": 76337, "tid": -914061504, "ts": 1716454224701357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnSelf_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224701405, "dur": 3,
    "args": {
      "External id": 205643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205643,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205643, "pid": 5, "tid": 7, "ts": 1716454224701405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701394, "dur": 10,
    "args": {
      "External id": 205643,
      "cbid": 211, "correlation": 205643
    }
  },
  {
    "ph": "s", "id": 205643, "pid": 76337, "tid": -914061504, "ts": 1716454224701394,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::sqrt_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224701436, "dur": 4,
    "args": {
      "External id": 205651,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205651,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205651, "pid": 5, "tid": 7, "ts": 1716454224701436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701425, "dur": 11,
    "args": {
      "External id": 205651,
      "cbid": 211, "correlation": 205651
    }
  },
  {
    "ph": "s", "id": 205651, "pid": 76337, "tid": -914061504, "ts": 1716454224701425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::reciprocal_kernel_cuda(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#5}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224701467, "dur": 4,
    "args": {
      "External id": 205659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205659,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205659, "pid": 5, "tid": 7, "ts": 1716454224701467,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701456, "dur": 9,
    "args": {
      "External id": 205659,
      "cbid": 211, "correlation": 205659
    }
  },
  {
    "ph": "s", "id": 205659, "pid": 76337, "tid": -914061504, "ts": 1716454224701456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224701492, "dur": 3,
    "args": {
      "External id": 205667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205667,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205667, "pid": 5, "tid": 7, "ts": 1716454224701492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701483, "dur": 8,
    "args": {
      "External id": 205667,
      "cbid": 211, "correlation": 205667
    }
  },
  {
    "ph": "s", "id": 205667, "pid": 76337, "tid": -914061504, "ts": 1716454224701483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3> >(int, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224701522, "dur": 3,
    "args": {
      "External id": 205675,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205675,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205675, "pid": 5, "tid": 7, "ts": 1716454224701522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701509, "dur": 14,
    "args": {
      "External id": 205675,
      "cbid": 211, "correlation": 205675
    }
  },
  {
    "ph": "s", "id": 205675, "pid": 76337, "tid": -914061504, "ts": 1716454224701509,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224701546, "dur": 4,
    "args": {
      "External id": 205683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205683,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205683, "pid": 5, "tid": 7, "ts": 1716454224701546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701537, "dur": 8,
    "args": {
      "External id": 205683,
      "cbid": 211, "correlation": 205683
    }
  },
  {
    "ph": "s", "id": 205683, "pid": 76337, "tid": -914061504, "ts": 1716454224701537,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224701565, "dur": 5,
    "args": {
      "External id": 205691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205691,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205691, "pid": 5, "tid": 7, "ts": 1716454224701565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701557, "dur": 6,
    "args": {
      "External id": 205691,
      "cbid": 211, "correlation": 205691
    }
  },
  {
    "ph": "s", "id": 205691, "pid": 76337, "tid": -914061504, "ts": 1716454224701557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224701584, "dur": 3,
    "args": {
      "External id": 205699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205699,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205699, "pid": 5, "tid": 7, "ts": 1716454224701584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701576, "dur": 7,
    "args": {
      "External id": 205699,
      "cbid": 211, "correlation": 205699
    }
  },
  {
    "ph": "s", "id": 205699, "pid": 76337, "tid": -914061504, "ts": 1716454224701576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2> >(int, at::native::CUDAFunctorOnOther_add<c10::Half>, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224701644, "dur": 3,
    "args": {
      "External id": 205707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205707,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 0.3,
      "warps per SM": 1.2,
      "grid": [24, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 205707, "pid": 5, "tid": 7, "ts": 1716454224701644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701631, "dur": 12,
    "args": {
      "External id": 205707,
      "cbid": 211, "correlation": 205707
    }
  },
  {
    "ph": "s", "id": 205707, "pid": 76337, "tid": -914061504, "ts": 1716454224701631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224701669, "dur": 4,
    "args": {
      "External id": 205715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205715,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205715, "pid": 5, "tid": 7, "ts": 1716454224701669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701661, "dur": 7,
    "args": {
      "External id": 205715,
      "cbid": 211, "correlation": 205715
    }
  },
  {
    "ph": "s", "id": 205715, "pid": 76337, "tid": -914061504, "ts": 1716454224701661,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224701692, "dur": 4,
    "args": {
      "External id": 205723,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205723,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205723, "pid": 5, "tid": 7, "ts": 1716454224701692,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701683, "dur": 8,
    "args": {
      "External id": 205723,
      "cbid": 211, "correlation": 205723
    }
  },
  {
    "ph": "s", "id": 205723, "pid": 76337, "tid": -914061504, "ts": 1716454224701683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224701710, "dur": 3,
    "args": {
      "External id": 205731,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205731,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [96, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 205731, "pid": 5, "tid": 7, "ts": 1716454224701710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224701703, "dur": 6,
    "args": {
      "External id": 205731,
      "cbid": 211, "correlation": 205731
    }
  },
  {
    "ph": "s", "id": 205731, "pid": 76337, "tid": -914061504, "ts": 1716454224701703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224702142, "dur": 5,
    "args": {
      "External id": 205740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205740,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205740, "pid": 5, "tid": 7, "ts": 1716454224702142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702124, "dur": 19,
    "args": {
      "External id": 205740,
      "cbid": 211, "correlation": 205740
    }
  },
  {
    "ph": "s", "id": 205740, "pid": 76337, "tid": -914061504, "ts": 1716454224702124,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224702182, "dur": 5,
    "args": {
      "External id": 205749,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205749,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205749, "pid": 5, "tid": 7, "ts": 1716454224702182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702171, "dur": 10,
    "args": {
      "External id": 205749,
      "cbid": 211, "correlation": 205749
    }
  },
  {
    "ph": "s", "id": 205749, "pid": 76337, "tid": -914061504, "ts": 1716454224702171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void (anonymous namespace)::elementwise_kernel_with_index<int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>(int, at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}, function_traits<at::native::arange_cuda_out(c10::Scalar const&, c10::Scalar const&, c10::Scalar const&, at::Tensor&)::{lambda()#1}::operator()() const::{lambda()#7}::operator()() const::{lambda(long)#1}>::result_type*)", "pid": 5, "tid": 7,
    "ts": 1716454224702312, "dur": 3,
    "args": {
      "External id": 205765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205765,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.075,
      "grid": [3, 1, 1],
      "block": [64, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205765, "pid": 5, "tid": 7, "ts": 1716454224702312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702298, "dur": 14,
    "args": {
      "External id": 205765,
      "cbid": 211, "correlation": 205765
    }
  },
  {
    "ph": "s", "id": 205765, "pid": 76337, "tid": -914061504, "ts": 1716454224702298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224702346, "dur": 3,
    "args": {
      "External id": 205773,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205773,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205773, "pid": 5, "tid": 7, "ts": 1716454224702346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702336, "dur": 8,
    "args": {
      "External id": 205773,
      "cbid": 211, "correlation": 205773
    }
  },
  {
    "ph": "s", "id": 205773, "pid": 76337, "tid": -914061504, "ts": 1716454224702336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224702381, "dur": 3,
    "args": {
      "External id": 205781,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205781,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205781, "pid": 5, "tid": 7, "ts": 1716454224702381,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702371, "dur": 9,
    "args": {
      "External id": 205781,
      "cbid": 211, "correlation": 205781
    }
  },
  {
    "ph": "s", "id": 205781, "pid": 76337, "tid": -914061504, "ts": 1716454224702371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::exp_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224702414, "dur": 4,
    "args": {
      "External id": 205789,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205789,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205789, "pid": 5, "tid": 7, "ts": 1716454224702414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702403, "dur": 10,
    "args": {
      "External id": 205789,
      "cbid": 211, "correlation": 205789
    }
  },
  {
    "ph": "s", "id": 205789, "pid": 76337, "tid": -914061504, "ts": 1716454224702403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#7}::operator()() const::{lambda(float)#1} const&)::{lambda(int)#2})", "pid": 5, "tid": 7,
    "ts": 1716454224702469, "dur": 4,
    "args": {
      "External id": 205801,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205801,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.05,
      "grid": [1, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205801, "pid": 5, "tid": 7, "ts": 1716454224702469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702456, "dur": 13,
    "args": {
      "External id": 205801,
      "cbid": 211, "correlation": 205801
    }
  },
  {
    "ph": "s", "id": 205801, "pid": 76337, "tid": -914061504, "ts": 1716454224702456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 2, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224702519, "dur": 4,
    "args": {
      "External id": 205812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205812, "pid": 5, "tid": 7, "ts": 1716454224702519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702502, "dur": 17,
    "args": {
      "External id": 205812,
      "cbid": 211, "correlation": 205812
    }
  },
  {
    "ph": "s", "id": 205812, "pid": 76337, "tid": -914061504, "ts": 1716454224702502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224702550, "dur": 3,
    "args": {
      "External id": 205820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205820, "pid": 5, "tid": 7, "ts": 1716454224702550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702541, "dur": 8,
    "args": {
      "External id": 205820,
      "cbid": 211, "correlation": 205820
    }
  },
  {
    "ph": "s", "id": 205820, "pid": 76337, "tid": -914061504, "ts": 1716454224702541,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::sin_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224702588, "dur": 5,
    "args": {
      "External id": 205828,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205828,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205828, "pid": 5, "tid": 7, "ts": 1716454224702588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702575, "dur": 14,
    "args": {
      "External id": 205828,
      "cbid": 211, "correlation": 205828
    }
  },
  {
    "ph": "s", "id": 205828, "pid": 76337, "tid": -914061504, "ts": 1716454224702575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2> >(int, at::native::cos_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#2}::operator()() const::{lambda(float)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224702622, "dur": 5,
    "args": {
      "External id": 205836,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205836,
      "registers per thread": 30,
      "shared memory": 0,
      "blocks per SM": 0.0375,
      "warps per SM": 0.15,
      "grid": [3, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205836, "pid": 5, "tid": 7, "ts": 1716454224702622,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702611, "dur": 10,
    "args": {
      "External id": 205836,
      "cbid": 211, "correlation": 205836
    }
  },
  {
    "ph": "s", "id": 205836, "pid": 76337, "tid": -914061504, "ts": 1716454224702611,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 128, 1>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224702655, "dur": 4,
    "args": {
      "External id": 205845,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205845,
      "registers per thread": 22,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205845, "pid": 5, "tid": 7, "ts": 1716454224702655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702642, "dur": 13,
    "args": {
      "External id": 205845,
      "cbid": 211, "correlation": 205845
    }
  },
  {
    "ph": "s", "id": 205845, "pid": 76337, "tid": -914061504, "ts": 1716454224702642,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<float, unsigned int, 2, 64, 64>(float*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<float, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224702720, "dur": 4,
    "args": {
      "External id": 205858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205858,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205858, "pid": 5, "tid": 7, "ts": 1716454224702720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702705, "dur": 16,
    "args": {
      "External id": 205858,
      "cbid": 211, "correlation": 205858
    }
  },
  {
    "ph": "s", "id": 205858, "pid": 76337, "tid": -914061504, "ts": 1716454224702705,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1> >(int, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>, TrivialOffsetCalculator<1, unsigned int>, TrivialOffsetCalculator<1, unsigned int>, at::native::memory::LoadWithCast<1>, at::native::memory::StoreWithCast<1>)", "pid": 5, "tid": 7,
    "ts": 1716454224702764, "dur": 5,
    "args": {
      "External id": 205868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205868,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.0625,
      "warps per SM": 0.25,
      "grid": [5, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 205868, "pid": 5, "tid": 7, "ts": 1716454224702764,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702753, "dur": 11,
    "args": {
      "External id": 205868,
      "cbid": 211, "correlation": 205868
    }
  },
  {
    "ph": "s", "id": 205868, "pid": 76337, "tid": -914061504, "ts": 1716454224702753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702896, "dur": 6,
    "args": {
      "External id": 205885,
      "cbid": 251, "correlation": 205885
    }
  },
  {
    "ph": "f", "id": 205885, "pid": 76337, "tid": -914061504, "ts": 1716454224702896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void gemmSN_TN_kernel<float, 128, 16, 2, 4, 8, 9, false, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half> >(cublasGemmSmallNParams<cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half const>, cublasGemvTensorStridedBatched<__half>, float>)", "pid": 5, "tid": 7,
    "ts": 1716454224702926, "dur": 12,
    "args": {
      "External id": 205887,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205887,
      "registers per thread": 72,
      "shared memory": 13824,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [160, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 205887, "pid": 5, "tid": 7, "ts": 1716454224702926,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702910, "dur": 17,
    "args": {
      "External id": 205887,
      "cbid": 211, "correlation": 205887
    }
  },
  {
    "ph": "s", "id": 205887, "pid": 76337, "tid": -914061504, "ts": 1716454224702910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224703001, "dur": 4,
    "args": {
      "External id": 205895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205895,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 205895, "pid": 5, "tid": 7, "ts": 1716454224703001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224702984, "dur": 16,
    "args": {
      "External id": 205895,
      "cbid": 211, "correlation": 205895
    }
  },
  {
    "ph": "s", "id": 205895, "pid": 76337, "tid": -914061504, "ts": 1716454224702984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703061, "dur": 2,
    "args": {
      "External id": 205911,
      "cbid": 251, "correlation": 205911
    }
  },
  {
    "ph": "f", "id": 205911, "pid": 76337, "tid": -914061504, "ts": 1716454224703061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703067, "dur": 0,
    "args": {
      "External id": 205913,
      "cbid": 251, "correlation": 205913
    }
  },
  {
    "ph": "f", "id": 205913, "pid": 76337, "tid": -914061504, "ts": 1716454224703067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224703084, "dur": 13,
    "args": {
      "External id": 205914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205914,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 205914, "pid": 5, "tid": 7, "ts": 1716454224703084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703070, "dur": 14,
    "args": {
      "External id": 205914,
      "cbid": 211, "correlation": 205914
    }
  },
  {
    "ph": "s", "id": 205914, "pid": 76337, "tid": -914061504, "ts": 1716454224703070,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224703100, "dur": 5,
    "args": {
      "External id": 205916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205916,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 205916, "pid": 5, "tid": 7, "ts": 1716454224703100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703090, "dur": 8,
    "args": {
      "External id": 205916,
      "cbid": 211, "correlation": 205916
    }
  },
  {
    "ph": "s", "id": 205916, "pid": 76337, "tid": -914061504, "ts": 1716454224703090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703204, "dur": 1,
    "args": {
      "External id": 205926,
      "cbid": 317, "correlation": 205926
    }
  },
  {
    "ph": "f", "id": 205926, "pid": 76337, "tid": -914061504, "ts": 1716454224703204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703206, "dur": 1,
    "args": {
      "External id": 205927,
      "cbid": 203, "correlation": 205927
    }
  },
  {
    "ph": "f", "id": 205927, "pid": 76337, "tid": -914061504, "ts": 1716454224703206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703208, "dur": 1,
    "args": {
      "External id": 205928,
      "cbid": 205, "correlation": 205928
    }
  },
  {
    "ph": "f", "id": 205928, "pid": 76337, "tid": -914061504, "ts": 1716454224703208,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224703265, "dur": 7,
    "args": {
      "External id": 205932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205932,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 9.6,
      "warps per SM": 76.8,
      "grid": [96, 1, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205932, "pid": 5, "tid": 7, "ts": 1716454224703265,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703251, "dur": 14,
    "args": {
      "External id": 205932,
      "cbid": 211, "correlation": 205932
    }
  },
  {
    "ph": "s", "id": 205932, "pid": 76337, "tid": -914061504, "ts": 1716454224703251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224703277, "dur": 4,
    "args": {
      "External id": 205934,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205934,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 4,
      "warps per SM": 32,
      "grid": [1, 1, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 50
    }
  },
  {
    "ph": "f", "id": 205934, "pid": 5, "tid": 7, "ts": 1716454224703277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703268, "dur": 7,
    "args": {
      "External id": 205934,
      "cbid": 211, "correlation": 205934
    }
  },
  {
    "ph": "s", "id": 205934, "pid": 76337, "tid": -914061504, "ts": 1716454224703268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224703297, "dur": 3,
    "args": {
      "External id": 205936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 205936, "pid": 5, "tid": 7, "ts": 1716454224703297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703288, "dur": 8,
    "args": {
      "External id": 205936,
      "cbid": 211, "correlation": 205936
    }
  },
  {
    "ph": "s", "id": 205936, "pid": 76337, "tid": -914061504, "ts": 1716454224703288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703303, "dur": 0,
    "args": {
      "External id": 205937,
      "cbid": 51, "correlation": 205937
    }
  },
  {
    "ph": "s", "id": 205937, "pid": 76337, "tid": -914061504, "ts": 1716454224703303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224703314, "dur": 84,
    "args": {
      "External id": 205938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205938,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [96, 5, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 205938, "pid": 5, "tid": 7, "ts": 1716454224703314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703305, "dur": 8,
    "args": {
      "External id": 205938,
      "cbid": 211, "correlation": 205938
    }
  },
  {
    "ph": "s", "id": 205938, "pid": 76337, "tid": -914061504, "ts": 1716454224703305,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224703400, "dur": 60,
    "args": {
      "External id": 205943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205943, "pid": 5, "tid": 7, "ts": 1716454224703400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224703343, "dur": 10,
    "args": {
      "External id": 205943,
      "cbid": 211, "correlation": 205943
    }
  },
  {
    "ph": "s", "id": 205943, "pid": 76337, "tid": -914061504, "ts": 1716454224703343,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224705168, "dur": 52,
    "args": {
      "External id": 205963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205963,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 205963, "pid": 5, "tid": 7, "ts": 1716454224705168,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705147, "dur": 21,
    "args": {
      "External id": 205963,
      "cbid": 211, "correlation": 205963
    }
  },
  {
    "ph": "s", "id": 205963, "pid": 76337, "tid": -914061504, "ts": 1716454224705147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224705222, "dur": 5,
    "args": {
      "External id": 205975,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205975,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 205975, "pid": 5, "tid": 7, "ts": 1716454224705222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705180, "dur": 8,
    "args": {
      "External id": 205975,
      "cbid": 211, "correlation": 205975
    }
  },
  {
    "ph": "s", "id": 205975, "pid": 76337, "tid": -914061504, "ts": 1716454224705180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224705227, "dur": 56,
    "args": {
      "External id": 205978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205978,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205978, "pid": 5, "tid": 7, "ts": 1716454224705227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705206, "dur": 7,
    "args": {
      "External id": 205978,
      "cbid": 211, "correlation": 205978
    }
  },
  {
    "ph": "s", "id": 205978, "pid": 76337, "tid": -914061504, "ts": 1716454224705206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224705285, "dur": 36,
    "args": {
      "External id": 205987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 205987,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 205987, "pid": 5, "tid": 7, "ts": 1716454224705285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705252, "dur": 10,
    "args": {
      "External id": 205987,
      "cbid": 211, "correlation": 205987
    }
  },
  {
    "ph": "s", "id": 205987, "pid": 76337, "tid": -914061504, "ts": 1716454224705252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705311, "dur": 0,
    "args": {
      "External id": 205997,
      "cbid": 317, "correlation": 205997
    }
  },
  {
    "ph": "f", "id": 205997, "pid": 76337, "tid": -914061504, "ts": 1716454224705311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705312, "dur": 0,
    "args": {
      "External id": 205998,
      "cbid": 203, "correlation": 205998
    }
  },
  {
    "ph": "f", "id": 205998, "pid": 76337, "tid": -914061504, "ts": 1716454224705312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705313, "dur": 0,
    "args": {
      "External id": 205999,
      "cbid": 205, "correlation": 205999
    }
  },
  {
    "ph": "f", "id": 205999, "pid": 76337, "tid": -914061504, "ts": 1716454224705313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224705344, "dur": 40,
    "args": {
      "External id": 206003,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206003,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206003, "pid": 5, "tid": 7, "ts": 1716454224705344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705331, "dur": 13,
    "args": {
      "External id": 206003,
      "cbid": 211, "correlation": 206003
    }
  },
  {
    "ph": "s", "id": 206003, "pid": 76337, "tid": -914061504, "ts": 1716454224705331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224705386, "dur": 14,
    "args": {
      "External id": 206005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206005,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206005, "pid": 5, "tid": 7, "ts": 1716454224705386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705347, "dur": 6,
    "args": {
      "External id": 206005,
      "cbid": 211, "correlation": 206005
    }
  },
  {
    "ph": "s", "id": 206005, "pid": 76337, "tid": -914061504, "ts": 1716454224705347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224705401, "dur": 3,
    "args": {
      "External id": 206007,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206007,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206007, "pid": 5, "tid": 7, "ts": 1716454224705401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705359, "dur": 6,
    "args": {
      "External id": 206007,
      "cbid": 211, "correlation": 206007
    }
  },
  {
    "ph": "s", "id": 206007, "pid": 76337, "tid": -914061504, "ts": 1716454224705359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705368, "dur": 0,
    "args": {
      "External id": 206008,
      "cbid": 51, "correlation": 206008
    }
  },
  {
    "ph": "s", "id": 206008, "pid": 76337, "tid": -914061504, "ts": 1716454224705368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224705405, "dur": 706,
    "args": {
      "External id": 206009,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206009,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206009, "pid": 5, "tid": 7, "ts": 1716454224705405,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705369, "dur": 7,
    "args": {
      "External id": 206009,
      "cbid": 211, "correlation": 206009
    }
  },
  {
    "ph": "s", "id": 206009, "pid": 76337, "tid": -914061504, "ts": 1716454224705369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224706112, "dur": 59,
    "args": {
      "External id": 206014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206014,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206014, "pid": 5, "tid": 7, "ts": 1716454224706112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705401, "dur": 9,
    "args": {
      "External id": 206014,
      "cbid": 211, "correlation": 206014
    }
  },
  {
    "ph": "s", "id": 206014, "pid": 76337, "tid": -914061504, "ts": 1716454224705401,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224706173, "dur": 4,
    "args": {
      "External id": 206022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206022,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206022, "pid": 5, "tid": 7, "ts": 1716454224706173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705444, "dur": 9,
    "args": {
      "External id": 206022,
      "cbid": 211, "correlation": 206022
    }
  },
  {
    "ph": "s", "id": 206022, "pid": 76337, "tid": -914061504, "ts": 1716454224705444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705513, "dur": 1,
    "args": {
      "External id": 206038,
      "cbid": 251, "correlation": 206038
    }
  },
  {
    "ph": "f", "id": 206038, "pid": 76337, "tid": -914061504, "ts": 1716454224705513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705519, "dur": 0,
    "args": {
      "External id": 206040,
      "cbid": 251, "correlation": 206040
    }
  },
  {
    "ph": "f", "id": 206040, "pid": 76337, "tid": -914061504, "ts": 1716454224705519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224706178, "dur": 9,
    "args": {
      "External id": 206041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206041,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 206041, "pid": 5, "tid": 7, "ts": 1716454224706178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705521, "dur": 11,
    "args": {
      "External id": 206041,
      "cbid": 211, "correlation": 206041
    }
  },
  {
    "ph": "s", "id": 206041, "pid": 76337, "tid": -914061504, "ts": 1716454224705521,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224706188, "dur": 4,
    "args": {
      "External id": 206043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206043,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 206043, "pid": 5, "tid": 7, "ts": 1716454224706188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705534, "dur": 7,
    "args": {
      "External id": 206043,
      "cbid": 211, "correlation": 206043
    }
  },
  {
    "ph": "s", "id": 206043, "pid": 76337, "tid": -914061504, "ts": 1716454224705534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224706193, "dur": 55,
    "args": {
      "External id": 206053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206053,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206053, "pid": 5, "tid": 7, "ts": 1716454224706193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705596, "dur": 12,
    "args": {
      "External id": 206053,
      "cbid": 211, "correlation": 206053
    }
  },
  {
    "ph": "s", "id": 206053, "pid": 76337, "tid": -914061504, "ts": 1716454224705596,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224706250, "dur": 51,
    "args": {
      "External id": 206073,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206073,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 206073, "pid": 5, "tid": 7, "ts": 1716454224706250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705663, "dur": 10,
    "args": {
      "External id": 206073,
      "cbid": 211, "correlation": 206073
    }
  },
  {
    "ph": "s", "id": 206073, "pid": 76337, "tid": -914061504, "ts": 1716454224705663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224706302, "dur": 4,
    "args": {
      "External id": 206085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206085,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206085, "pid": 5, "tid": 7, "ts": 1716454224706302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705683, "dur": 7,
    "args": {
      "External id": 206085,
      "cbid": 211, "correlation": 206085
    }
  },
  {
    "ph": "s", "id": 206085, "pid": 76337, "tid": -914061504, "ts": 1716454224705683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224706308, "dur": 56,
    "args": {
      "External id": 206088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206088,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206088, "pid": 5, "tid": 7, "ts": 1716454224706308,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705704, "dur": 6,
    "args": {
      "External id": 206088,
      "cbid": 211, "correlation": 206088
    }
  },
  {
    "ph": "s", "id": 206088, "pid": 76337, "tid": -914061504, "ts": 1716454224705704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224706364, "dur": 37,
    "args": {
      "External id": 206097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206097,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206097, "pid": 5, "tid": 7, "ts": 1716454224706364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705744, "dur": 10,
    "args": {
      "External id": 206097,
      "cbid": 211, "correlation": 206097
    }
  },
  {
    "ph": "s", "id": 206097, "pid": 76337, "tid": -914061504, "ts": 1716454224705744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705814, "dur": 0,
    "args": {
      "External id": 206107,
      "cbid": 317, "correlation": 206107
    }
  },
  {
    "ph": "f", "id": 206107, "pid": 76337, "tid": -914061504, "ts": 1716454224705814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705815, "dur": 0,
    "args": {
      "External id": 206108,
      "cbid": 203, "correlation": 206108
    }
  },
  {
    "ph": "f", "id": 206108, "pid": 76337, "tid": -914061504, "ts": 1716454224705815,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705816, "dur": 0,
    "args": {
      "External id": 206109,
      "cbid": 205, "correlation": 206109
    }
  },
  {
    "ph": "f", "id": 206109, "pid": 76337, "tid": -914061504, "ts": 1716454224705816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224706402, "dur": 41,
    "args": {
      "External id": 206113,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206113,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206113, "pid": 5, "tid": 7, "ts": 1716454224706402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705830, "dur": 13,
    "args": {
      "External id": 206113,
      "cbid": 211, "correlation": 206113
    }
  },
  {
    "ph": "s", "id": 206113, "pid": 76337, "tid": -914061504, "ts": 1716454224705830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224706445, "dur": 14,
    "args": {
      "External id": 206115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206115,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206115, "pid": 5, "tid": 7, "ts": 1716454224706445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705845, "dur": 5,
    "args": {
      "External id": 206115,
      "cbid": 211, "correlation": 206115
    }
  },
  {
    "ph": "s", "id": 206115, "pid": 76337, "tid": -914061504, "ts": 1716454224705845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224706461, "dur": 4,
    "args": {
      "External id": 206117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206117,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206117, "pid": 5, "tid": 7, "ts": 1716454224706461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705854, "dur": 5,
    "args": {
      "External id": 206117,
      "cbid": 211, "correlation": 206117
    }
  },
  {
    "ph": "s", "id": 206117, "pid": 76337, "tid": -914061504, "ts": 1716454224705854,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705862, "dur": 0,
    "args": {
      "External id": 206118,
      "cbid": 51, "correlation": 206118
    }
  },
  {
    "ph": "s", "id": 206118, "pid": 76337, "tid": -914061504, "ts": 1716454224705862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224706466, "dur": 699,
    "args": {
      "External id": 206119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206119,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206119, "pid": 5, "tid": 7, "ts": 1716454224706466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705863, "dur": 5,
    "args": {
      "External id": 206119,
      "cbid": 211, "correlation": 206119
    }
  },
  {
    "ph": "s", "id": 206119, "pid": 76337, "tid": -914061504, "ts": 1716454224705863,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224707166, "dur": 59,
    "args": {
      "External id": 206124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206124,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206124, "pid": 5, "tid": 7, "ts": 1716454224707166,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705892, "dur": 8,
    "args": {
      "External id": 206124,
      "cbid": 211, "correlation": 206124
    }
  },
  {
    "ph": "s", "id": 206124, "pid": 76337, "tid": -914061504, "ts": 1716454224705892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224707226, "dur": 50,
    "args": {
      "External id": 206132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206132,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206132, "pid": 5, "tid": 7, "ts": 1716454224707226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705924, "dur": 8,
    "args": {
      "External id": 206132,
      "cbid": 211, "correlation": 206132
    }
  },
  {
    "ph": "s", "id": 206132, "pid": 76337, "tid": -914061504, "ts": 1716454224705924,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224707278, "dur": 35,
    "args": {
      "External id": 206140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206140,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206140, "pid": 5, "tid": 7, "ts": 1716454224707278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224705956, "dur": 9,
    "args": {
      "External id": 206140,
      "cbid": 211, "correlation": 206140
    }
  },
  {
    "ph": "s", "id": 206140, "pid": 76337, "tid": -914061504, "ts": 1716454224705956,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224707314, "dur": 53,
    "args": {
      "External id": 206160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206160,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 206160, "pid": 5, "tid": 7, "ts": 1716454224707314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706050, "dur": 13,
    "args": {
      "External id": 206160,
      "cbid": 211, "correlation": 206160
    }
  },
  {
    "ph": "s", "id": 206160, "pid": 76337, "tid": -914061504, "ts": 1716454224706050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224707368, "dur": 4,
    "args": {
      "External id": 206172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206172,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206172, "pid": 5, "tid": 7, "ts": 1716454224707368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706072, "dur": 6,
    "args": {
      "External id": 206172,
      "cbid": 211, "correlation": 206172
    }
  },
  {
    "ph": "s", "id": 206172, "pid": 76337, "tid": -914061504, "ts": 1716454224706072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224707373, "dur": 56,
    "args": {
      "External id": 206175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206175, "pid": 5, "tid": 7, "ts": 1716454224707373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706090, "dur": 7,
    "args": {
      "External id": 206175,
      "cbid": 211, "correlation": 206175
    }
  },
  {
    "ph": "s", "id": 206175, "pid": 76337, "tid": -914061504, "ts": 1716454224706090,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706149, "dur": 0,
    "args": {
      "External id": 206186,
      "cbid": 317, "correlation": 206186
    }
  },
  {
    "ph": "f", "id": 206186, "pid": 76337, "tid": -914061504, "ts": 1716454224706149,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706150, "dur": 0,
    "args": {
      "External id": 206187,
      "cbid": 203, "correlation": 206187
    }
  },
  {
    "ph": "f", "id": 206187, "pid": 76337, "tid": -914061504, "ts": 1716454224706150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706150, "dur": 0,
    "args": {
      "External id": 206188,
      "cbid": 205, "correlation": 206188
    }
  },
  {
    "ph": "f", "id": 206188, "pid": 76337, "tid": -914061504, "ts": 1716454224706150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706185, "dur": 2,
    "args": {
      "External id": 206192,
      "cbid": 251, "correlation": 206192
    }
  },
  {
    "ph": "f", "id": 206192, "pid": 76337, "tid": -914061504, "ts": 1716454224706185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706189, "dur": 1,
    "args": {
      "External id": 206193,
      "cbid": 251, "correlation": 206193
    }
  },
  {
    "ph": "f", "id": 206193, "pid": 76337, "tid": -914061504, "ts": 1716454224706189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706191, "dur": 1,
    "args": {
      "External id": 206194,
      "cbid": 251, "correlation": 206194
    }
  },
  {
    "ph": "f", "id": 206194, "pid": 76337, "tid": -914061504, "ts": 1716454224706191,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706193, "dur": 1,
    "args": {
      "External id": 206195,
      "cbid": 251, "correlation": 206195
    }
  },
  {
    "ph": "f", "id": 206195, "pid": 76337, "tid": -914061504, "ts": 1716454224706193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706195, "dur": 1,
    "args": {
      "External id": 206196,
      "cbid": 251, "correlation": 206196
    }
  },
  {
    "ph": "f", "id": 206196, "pid": 76337, "tid": -914061504, "ts": 1716454224706195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706197, "dur": 1,
    "args": {
      "External id": 206197,
      "cbid": 251, "correlation": 206197
    }
  },
  {
    "ph": "f", "id": 206197, "pid": 76337, "tid": -914061504, "ts": 1716454224706197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706199, "dur": 1,
    "args": {
      "External id": 206198,
      "cbid": 251, "correlation": 206198
    }
  },
  {
    "ph": "f", "id": 206198, "pid": 76337, "tid": -914061504, "ts": 1716454224706199,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706201, "dur": 1,
    "args": {
      "External id": 206199,
      "cbid": 251, "correlation": 206199
    }
  },
  {
    "ph": "f", "id": 206199, "pid": 76337, "tid": -914061504, "ts": 1716454224706201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706204, "dur": 0,
    "args": {
      "External id": 206200,
      "cbid": 251, "correlation": 206200
    }
  },
  {
    "ph": "f", "id": 206200, "pid": 76337, "tid": -914061504, "ts": 1716454224706204,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224707430, "dur": 115,
    "args": {
      "External id": 206201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206201,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206201, "pid": 5, "tid": 7, "ts": 1716454224707430,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706208, "dur": 14,
    "args": {
      "External id": 206201,
      "cbid": 211, "correlation": 206201
    }
  },
  {
    "ph": "s", "id": 206201, "pid": 76337, "tid": -914061504, "ts": 1716454224706208,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224707547, "dur": 60,
    "args": {
      "External id": 206207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206207,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206207, "pid": 5, "tid": 7, "ts": 1716454224707547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706246, "dur": 9,
    "args": {
      "External id": 206207,
      "cbid": 211, "correlation": 206207
    }
  },
  {
    "ph": "s", "id": 206207, "pid": 76337, "tid": -914061504, "ts": 1716454224706246,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224707609, "dur": 560,
    "args": {
      "External id": 206216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206216,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206216, "pid": 5, "tid": 7, "ts": 1716454224707609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706347, "dur": 27,
    "args": {
      "External id": 206216,
      "cbid": 211, "correlation": 206216
    }
  },
  {
    "ph": "s", "id": 206216, "pid": 76337, "tid": -914061504, "ts": 1716454224706347,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224708170, "dur": 183,
    "args": {
      "External id": 206238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206238,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206238, "pid": 5, "tid": 7, "ts": 1716454224708170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706433, "dur": 12,
    "args": {
      "External id": 206238,
      "cbid": 211, "correlation": 206238
    }
  },
  {
    "ph": "s", "id": 206238, "pid": 76337, "tid": -914061504, "ts": 1716454224706433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706550, "dur": 2,
    "args": {
      "External id": 206249,
      "cbid": 251, "correlation": 206249
    }
  },
  {
    "ph": "f", "id": 206249, "pid": 76337, "tid": -914061504, "ts": 1716454224706550,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224708354, "dur": 196,
    "args": {
      "External id": 206250,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206250,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206250, "pid": 5, "tid": 7, "ts": 1716454224708354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706558, "dur": 15,
    "args": {
      "External id": 206250,
      "cbid": 211, "correlation": 206250
    }
  },
  {
    "ph": "s", "id": 206250, "pid": 76337, "tid": -914061504, "ts": 1716454224706558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706631, "dur": 1,
    "args": {
      "External id": 206261,
      "cbid": 251, "correlation": 206261
    }
  },
  {
    "ph": "f", "id": 206261, "pid": 76337, "tid": -914061504, "ts": 1716454224706631,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224708551, "dur": 190,
    "args": {
      "External id": 206262,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206262,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206262, "pid": 5, "tid": 7, "ts": 1716454224708551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706635, "dur": 11,
    "args": {
      "External id": 206262,
      "cbid": 211, "correlation": 206262
    }
  },
  {
    "ph": "s", "id": 206262, "pid": 76337, "tid": -914061504, "ts": 1716454224706635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706700, "dur": 1,
    "args": {
      "External id": 206273,
      "cbid": 251, "correlation": 206273
    }
  },
  {
    "ph": "f", "id": 206273, "pid": 76337, "tid": -914061504, "ts": 1716454224706700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224708742, "dur": 188,
    "args": {
      "External id": 206274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206274,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206274, "pid": 5, "tid": 7, "ts": 1716454224708742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706704, "dur": 12,
    "args": {
      "External id": 206274,
      "cbid": 211, "correlation": 206274
    }
  },
  {
    "ph": "s", "id": 206274, "pid": 76337, "tid": -914061504, "ts": 1716454224706704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224708932, "dur": 18774,
    "args": {
      "External id": 206295,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206295,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206295, "pid": 5, "tid": 7, "ts": 1716454224708932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706816, "dur": 14,
    "args": {
      "External id": 206295,
      "cbid": 211, "correlation": 206295
    }
  },
  {
    "ph": "s", "id": 206295, "pid": 76337, "tid": -914061504, "ts": 1716454224706816,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706932, "dur": 2,
    "args": {
      "External id": 206313,
      "cbid": 251, "correlation": 206313
    }
  },
  {
    "ph": "f", "id": 206313, "pid": 76337, "tid": -914061504, "ts": 1716454224706932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224727707, "dur": 210,
    "args": {
      "External id": 206315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206315,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206315, "pid": 5, "tid": 7, "ts": 1716454224727707,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224706939, "dur": 14,
    "args": {
      "External id": 206315,
      "cbid": 211, "correlation": 206315
    }
  },
  {
    "ph": "s", "id": 206315, "pid": 76337, "tid": -914061504, "ts": 1716454224706939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224727919, "dur": 67,
    "args": {
      "External id": 206323,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206323,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206323, "pid": 5, "tid": 7, "ts": 1716454224727919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707022, "dur": 13,
    "args": {
      "External id": 206323,
      "cbid": 211, "correlation": 206323
    }
  },
  {
    "ph": "s", "id": 206323, "pid": 76337, "tid": -914061504, "ts": 1716454224707022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224727987, "dur": 96,
    "args": {
      "External id": 206331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206331,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206331, "pid": 5, "tid": 7, "ts": 1716454224727987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707062, "dur": 9,
    "args": {
      "External id": 206331,
      "cbid": 211, "correlation": 206331
    }
  },
  {
    "ph": "s", "id": 206331, "pid": 76337, "tid": -914061504, "ts": 1716454224707062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224728084, "dur": 54,
    "args": {
      "External id": 206342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206342,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206342, "pid": 5, "tid": 7, "ts": 1716454224728084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707148, "dur": 14,
    "args": {
      "External id": 206342,
      "cbid": 211, "correlation": 206342
    }
  },
  {
    "ph": "s", "id": 206342, "pid": 76337, "tid": -914061504, "ts": 1716454224707148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224728140, "dur": 93,
    "args": {
      "External id": 206364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206364,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206364, "pid": 5, "tid": 7, "ts": 1716454224728140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707182, "dur": 7,
    "args": {
      "External id": 206364,
      "cbid": 211, "correlation": 206364
    }
  },
  {
    "ph": "s", "id": 206364, "pid": 76337, "tid": -914061504, "ts": 1716454224707182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707266, "dur": 1,
    "args": {
      "External id": 206375,
      "cbid": 251, "correlation": 206375
    }
  },
  {
    "ph": "f", "id": 206375, "pid": 76337, "tid": -914061504, "ts": 1716454224707266,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224728234, "dur": 103,
    "args": {
      "External id": 206376,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206376,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206376, "pid": 5, "tid": 7, "ts": 1716454224728234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707271, "dur": 12,
    "args": {
      "External id": 206376,
      "cbid": 211, "correlation": 206376
    }
  },
  {
    "ph": "s", "id": 206376, "pid": 76337, "tid": -914061504, "ts": 1716454224707271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707354, "dur": 2,
    "args": {
      "External id": 206387,
      "cbid": 251, "correlation": 206387
    }
  },
  {
    "ph": "f", "id": 206387, "pid": 76337, "tid": -914061504, "ts": 1716454224707354,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707359, "dur": 0,
    "args": {
      "External id": 206388,
      "cbid": 251, "correlation": 206388
    }
  },
  {
    "ph": "f", "id": 206388, "pid": 76337, "tid": -914061504, "ts": 1716454224707359,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224728338, "dur": 10,
    "args": {
      "External id": 206389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206389,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 206389, "pid": 5, "tid": 7, "ts": 1716454224728338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707361, "dur": 14,
    "args": {
      "External id": 206389,
      "cbid": 211, "correlation": 206389
    }
  },
  {
    "ph": "s", "id": 206389, "pid": 76337, "tid": -914061504, "ts": 1716454224707361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224728350, "dur": 5,
    "args": {
      "External id": 206391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206391,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 206391, "pid": 5, "tid": 7, "ts": 1716454224728350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707379, "dur": 7,
    "args": {
      "External id": 206391,
      "cbid": 211, "correlation": 206391
    }
  },
  {
    "ph": "s", "id": 206391, "pid": 76337, "tid": -914061504, "ts": 1716454224707379,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707441, "dur": 1,
    "args": {
      "External id": 206402,
      "cbid": 251, "correlation": 206402
    }
  },
  {
    "ph": "f", "id": 206402, "pid": 76337, "tid": -914061504, "ts": 1716454224707441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707444, "dur": 0,
    "args": {
      "External id": 206403,
      "cbid": 251, "correlation": 206403
    }
  },
  {
    "ph": "f", "id": 206403, "pid": 76337, "tid": -914061504, "ts": 1716454224707444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224728356, "dur": 6,
    "args": {
      "External id": 206404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206404,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 206404, "pid": 5, "tid": 7, "ts": 1716454224728356,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707446, "dur": 12,
    "args": {
      "External id": 206404,
      "cbid": 211, "correlation": 206404
    }
  },
  {
    "ph": "s", "id": 206404, "pid": 76337, "tid": -914061504, "ts": 1716454224707446,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224728363, "dur": 3,
    "args": {
      "External id": 206406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206406,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 206406, "pid": 5, "tid": 7, "ts": 1716454224728363,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707460, "dur": 5,
    "args": {
      "External id": 206406,
      "cbid": 211, "correlation": 206406
    }
  },
  {
    "ph": "s", "id": 206406, "pid": 76337, "tid": -914061504, "ts": 1716454224707460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224728368, "dur": 156,
    "args": {
      "External id": 206427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206427,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206427, "pid": 5, "tid": 7, "ts": 1716454224728368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707535, "dur": 13,
    "args": {
      "External id": 206427,
      "cbid": 211, "correlation": 206427
    }
  },
  {
    "ph": "s", "id": 206427, "pid": 76337, "tid": -914061504, "ts": 1716454224707535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707633, "dur": 2,
    "args": {
      "External id": 206445,
      "cbid": 251, "correlation": 206445
    }
  },
  {
    "ph": "f", "id": 206445, "pid": 76337, "tid": -914061504, "ts": 1716454224707633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224728525, "dur": 107,
    "args": {
      "External id": 206447,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206447,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206447, "pid": 5, "tid": 7, "ts": 1716454224728525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707639, "dur": 15,
    "args": {
      "External id": 206447,
      "cbid": 211, "correlation": 206447
    }
  },
  {
    "ph": "s", "id": 206447, "pid": 76337, "tid": -914061504, "ts": 1716454224707639,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224728634, "dur": 34,
    "args": {
      "External id": 206455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206455,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206455, "pid": 5, "tid": 7, "ts": 1716454224728634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707712, "dur": 12,
    "args": {
      "External id": 206455,
      "cbid": 211, "correlation": 206455
    }
  },
  {
    "ph": "s", "id": 206455, "pid": 76337, "tid": -914061504, "ts": 1716454224707712,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224728669, "dur": 68,
    "args": {
      "External id": 206463,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206463,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206463, "pid": 5, "tid": 7, "ts": 1716454224728669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707753, "dur": 9,
    "args": {
      "External id": 206463,
      "cbid": 211, "correlation": 206463
    }
  },
  {
    "ph": "s", "id": 206463, "pid": 76337, "tid": -914061504, "ts": 1716454224707753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224728739, "dur": 93,
    "args": {
      "External id": 206485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206485,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206485, "pid": 5, "tid": 7, "ts": 1716454224728739,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707805, "dur": 10,
    "args": {
      "External id": 206485,
      "cbid": 211, "correlation": 206485
    }
  },
  {
    "ph": "s", "id": 206485, "pid": 76337, "tid": -914061504, "ts": 1716454224707805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707897, "dur": 1,
    "args": {
      "External id": 206501,
      "cbid": 251, "correlation": 206501
    }
  },
  {
    "ph": "f", "id": 206501, "pid": 76337, "tid": -914061504, "ts": 1716454224707897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224728833, "dur": 580,
    "args": {
      "External id": 206503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206503,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206503, "pid": 5, "tid": 7, "ts": 1716454224728833,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707902, "dur": 12,
    "args": {
      "External id": 206503,
      "cbid": 211, "correlation": 206503
    }
  },
  {
    "ph": "s", "id": 206503, "pid": 76337, "tid": -914061504, "ts": 1716454224707902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224729414, "dur": 247,
    "args": {
      "External id": 206511,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206511,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206511, "pid": 5, "tid": 7, "ts": 1716454224729414,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224707990, "dur": 15,
    "args": {
      "External id": 206511,
      "cbid": 211, "correlation": 206511
    }
  },
  {
    "ph": "s", "id": 206511, "pid": 76337, "tid": -914061504, "ts": 1716454224707990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224729662, "dur": 252,
    "args": {
      "External id": 206519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206519,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206519, "pid": 5, "tid": 7, "ts": 1716454224729662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708026, "dur": 8,
    "args": {
      "External id": 206519,
      "cbid": 211, "correlation": 206519
    }
  },
  {
    "ph": "s", "id": 206519, "pid": 76337, "tid": -914061504, "ts": 1716454224708026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708111, "dur": 2,
    "args": {
      "External id": 206535,
      "cbid": 251, "correlation": 206535
    }
  },
  {
    "ph": "f", "id": 206535, "pid": 76337, "tid": -914061504, "ts": 1716454224708111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708116, "dur": 0,
    "args": {
      "External id": 206537,
      "cbid": 251, "correlation": 206537
    }
  },
  {
    "ph": "f", "id": 206537, "pid": 76337, "tid": -914061504, "ts": 1716454224708116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224729916, "dur": 360,
    "args": {
      "External id": 206538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206538,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 206538, "pid": 5, "tid": 7, "ts": 1716454224729916,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708120, "dur": 14,
    "args": {
      "External id": 206538,
      "cbid": 211, "correlation": 206538
    }
  },
  {
    "ph": "s", "id": 206538, "pid": 76337, "tid": -914061504, "ts": 1716454224708120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224730277, "dur": 49,
    "args": {
      "External id": 206546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206546,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206546, "pid": 5, "tid": 7, "ts": 1716454224730277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708164, "dur": 10,
    "args": {
      "External id": 206546,
      "cbid": 211, "correlation": 206546
    }
  },
  {
    "ph": "s", "id": 206546, "pid": 76337, "tid": -914061504, "ts": 1716454224708164,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224730328, "dur": 158,
    "args": {
      "External id": 206557,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206557,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206557, "pid": 5, "tid": 7, "ts": 1716454224730328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708235, "dur": 13,
    "args": {
      "External id": 206557,
      "cbid": 211, "correlation": 206557
    }
  },
  {
    "ph": "s", "id": 206557, "pid": 76337, "tid": -914061504, "ts": 1716454224708235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708301, "dur": 0,
    "args": {
      "External id": 206569,
      "cbid": 317, "correlation": 206569
    }
  },
  {
    "ph": "f", "id": 206569, "pid": 76337, "tid": -914061504, "ts": 1716454224708301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708302, "dur": 0,
    "args": {
      "External id": 206570,
      "cbid": 203, "correlation": 206570
    }
  },
  {
    "ph": "f", "id": 206570, "pid": 76337, "tid": -914061504, "ts": 1716454224708302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708303, "dur": 0,
    "args": {
      "External id": 206571,
      "cbid": 205, "correlation": 206571
    }
  },
  {
    "ph": "f", "id": 206571, "pid": 76337, "tid": -914061504, "ts": 1716454224708303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708327, "dur": 1,
    "args": {
      "External id": 206575,
      "cbid": 251, "correlation": 206575
    }
  },
  {
    "ph": "f", "id": 206575, "pid": 76337, "tid": -914061504, "ts": 1716454224708327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708329, "dur": 0,
    "args": {
      "External id": 206576,
      "cbid": 251, "correlation": 206576
    }
  },
  {
    "ph": "f", "id": 206576, "pid": 76337, "tid": -914061504, "ts": 1716454224708329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708330, "dur": 0,
    "args": {
      "External id": 206577,
      "cbid": 251, "correlation": 206577
    }
  },
  {
    "ph": "f", "id": 206577, "pid": 76337, "tid": -914061504, "ts": 1716454224708330,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708331, "dur": 0,
    "args": {
      "External id": 206578,
      "cbid": 251, "correlation": 206578
    }
  },
  {
    "ph": "f", "id": 206578, "pid": 76337, "tid": -914061504, "ts": 1716454224708331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708332, "dur": 0,
    "args": {
      "External id": 206579,
      "cbid": 251, "correlation": 206579
    }
  },
  {
    "ph": "f", "id": 206579, "pid": 76337, "tid": -914061504, "ts": 1716454224708332,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708333, "dur": 0,
    "args": {
      "External id": 206580,
      "cbid": 251, "correlation": 206580
    }
  },
  {
    "ph": "f", "id": 206580, "pid": 76337, "tid": -914061504, "ts": 1716454224708333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708334, "dur": 0,
    "args": {
      "External id": 206581,
      "cbid": 251, "correlation": 206581
    }
  },
  {
    "ph": "f", "id": 206581, "pid": 76337, "tid": -914061504, "ts": 1716454224708334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708335, "dur": 0,
    "args": {
      "External id": 206582,
      "cbid": 251, "correlation": 206582
    }
  },
  {
    "ph": "f", "id": 206582, "pid": 76337, "tid": -914061504, "ts": 1716454224708335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708336, "dur": 0,
    "args": {
      "External id": 206583,
      "cbid": 251, "correlation": 206583
    }
  },
  {
    "ph": "f", "id": 206583, "pid": 76337, "tid": -914061504, "ts": 1716454224708336,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224730487, "dur": 117,
    "args": {
      "External id": 206584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206584,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206584, "pid": 5, "tid": 7, "ts": 1716454224730487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708339, "dur": 13,
    "args": {
      "External id": 206584,
      "cbid": 211, "correlation": 206584
    }
  },
  {
    "ph": "s", "id": 206584, "pid": 76337, "tid": -914061504, "ts": 1716454224708339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224730605, "dur": 59,
    "args": {
      "External id": 206590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206590,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206590, "pid": 5, "tid": 7, "ts": 1716454224730605,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708374, "dur": 9,
    "args": {
      "External id": 206590,
      "cbid": 211, "correlation": 206590
    }
  },
  {
    "ph": "s", "id": 206590, "pid": 76337, "tid": -914061504, "ts": 1716454224708374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224730666, "dur": 50,
    "args": {
      "External id": 206598,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206598,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206598, "pid": 5, "tid": 7, "ts": 1716454224730666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708407, "dur": 9,
    "args": {
      "External id": 206598,
      "cbid": 211, "correlation": 206598
    }
  },
  {
    "ph": "s", "id": 206598, "pid": 76337, "tid": -914061504, "ts": 1716454224708407,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224730717, "dur": 53,
    "args": {
      "External id": 206618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206618,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 206618, "pid": 5, "tid": 7, "ts": 1716454224730717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708482, "dur": 12,
    "args": {
      "External id": 206618,
      "cbid": 211, "correlation": 206618
    }
  },
  {
    "ph": "s", "id": 206618, "pid": 76337, "tid": -914061504, "ts": 1716454224708482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224730771, "dur": 5,
    "args": {
      "External id": 206630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206630,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206630, "pid": 5, "tid": 7, "ts": 1716454224730771,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708504, "dur": 6,
    "args": {
      "External id": 206630,
      "cbid": 211, "correlation": 206630
    }
  },
  {
    "ph": "s", "id": 206630, "pid": 76337, "tid": -914061504, "ts": 1716454224708504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224730777, "dur": 56,
    "args": {
      "External id": 206633,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206633,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206633, "pid": 5, "tid": 7, "ts": 1716454224730777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708522, "dur": 7,
    "args": {
      "External id": 206633,
      "cbid": 211, "correlation": 206633
    }
  },
  {
    "ph": "s", "id": 206633, "pid": 76337, "tid": -914061504, "ts": 1716454224708522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224730835, "dur": 37,
    "args": {
      "External id": 206642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206642,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206642, "pid": 5, "tid": 7, "ts": 1716454224730835,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708562, "dur": 10,
    "args": {
      "External id": 206642,
      "cbid": 211, "correlation": 206642
    }
  },
  {
    "ph": "s", "id": 206642, "pid": 76337, "tid": -914061504, "ts": 1716454224708562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708614, "dur": 0,
    "args": {
      "External id": 206652,
      "cbid": 317, "correlation": 206652
    }
  },
  {
    "ph": "f", "id": 206652, "pid": 76337, "tid": -914061504, "ts": 1716454224708614,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708615, "dur": 0,
    "args": {
      "External id": 206653,
      "cbid": 203, "correlation": 206653
    }
  },
  {
    "ph": "f", "id": 206653, "pid": 76337, "tid": -914061504, "ts": 1716454224708615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708616, "dur": 0,
    "args": {
      "External id": 206654,
      "cbid": 205, "correlation": 206654
    }
  },
  {
    "ph": "f", "id": 206654, "pid": 76337, "tid": -914061504, "ts": 1716454224708616,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224730873, "dur": 40,
    "args": {
      "External id": 206658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206658,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206658, "pid": 5, "tid": 7, "ts": 1716454224730873,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708630, "dur": 12,
    "args": {
      "External id": 206658,
      "cbid": 211, "correlation": 206658
    }
  },
  {
    "ph": "s", "id": 206658, "pid": 76337, "tid": -914061504, "ts": 1716454224708630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224730915, "dur": 14,
    "args": {
      "External id": 206660,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206660,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206660, "pid": 5, "tid": 7, "ts": 1716454224730915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708644, "dur": 5,
    "args": {
      "External id": 206660,
      "cbid": 211, "correlation": 206660
    }
  },
  {
    "ph": "s", "id": 206660, "pid": 76337, "tid": -914061504, "ts": 1716454224708644,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224730931, "dur": 4,
    "args": {
      "External id": 206662,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206662,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206662, "pid": 5, "tid": 7, "ts": 1716454224730931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708654, "dur": 6,
    "args": {
      "External id": 206662,
      "cbid": 211, "correlation": 206662
    }
  },
  {
    "ph": "s", "id": 206662, "pid": 76337, "tid": -914061504, "ts": 1716454224708654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708663, "dur": 0,
    "args": {
      "External id": 206663,
      "cbid": 51, "correlation": 206663
    }
  },
  {
    "ph": "s", "id": 206663, "pid": 76337, "tid": -914061504, "ts": 1716454224708663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224730936, "dur": 706,
    "args": {
      "External id": 206664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206664,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206664, "pid": 5, "tid": 7, "ts": 1716454224730936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708664, "dur": 5,
    "args": {
      "External id": 206664,
      "cbid": 211, "correlation": 206664
    }
  },
  {
    "ph": "s", "id": 206664, "pid": 76337, "tid": -914061504, "ts": 1716454224708664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224731643, "dur": 59,
    "args": {
      "External id": 206669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206669, "pid": 5, "tid": 7, "ts": 1716454224731643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708691, "dur": 9,
    "args": {
      "External id": 206669,
      "cbid": 211, "correlation": 206669
    }
  },
  {
    "ph": "s", "id": 206669, "pid": 76337, "tid": -914061504, "ts": 1716454224708691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224731704, "dur": 4,
    "args": {
      "External id": 206677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206677,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206677, "pid": 5, "tid": 7, "ts": 1716454224731704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708734, "dur": 10,
    "args": {
      "External id": 206677,
      "cbid": 211, "correlation": 206677
    }
  },
  {
    "ph": "s", "id": 206677, "pid": 76337, "tid": -914061504, "ts": 1716454224708734,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708800, "dur": 1,
    "args": {
      "External id": 206693,
      "cbid": 251, "correlation": 206693
    }
  },
  {
    "ph": "f", "id": 206693, "pid": 76337, "tid": -914061504, "ts": 1716454224708800,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708805, "dur": 0,
    "args": {
      "External id": 206695,
      "cbid": 251, "correlation": 206695
    }
  },
  {
    "ph": "f", "id": 206695, "pid": 76337, "tid": -914061504, "ts": 1716454224708805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224731709, "dur": 11,
    "args": {
      "External id": 206696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206696,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 206696, "pid": 5, "tid": 7, "ts": 1716454224731709,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708807, "dur": 11,
    "args": {
      "External id": 206696,
      "cbid": 211, "correlation": 206696
    }
  },
  {
    "ph": "s", "id": 206696, "pid": 76337, "tid": -914061504, "ts": 1716454224708807,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224731721, "dur": 5,
    "args": {
      "External id": 206698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206698,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 206698, "pid": 5, "tid": 7, "ts": 1716454224731721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708821, "dur": 5,
    "args": {
      "External id": 206698,
      "cbid": 211, "correlation": 206698
    }
  },
  {
    "ph": "s", "id": 206698, "pid": 76337, "tid": -914061504, "ts": 1716454224708821,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224731728, "dur": 54,
    "args": {
      "External id": 206708,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206708,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206708, "pid": 5, "tid": 7, "ts": 1716454224731728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708878, "dur": 13,
    "args": {
      "External id": 206708,
      "cbid": 211, "correlation": 206708
    }
  },
  {
    "ph": "s", "id": 206708, "pid": 76337, "tid": -914061504, "ts": 1716454224708878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224731783, "dur": 51,
    "args": {
      "External id": 206728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206728,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 206728, "pid": 5, "tid": 7, "ts": 1716454224731783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708946, "dur": 11,
    "args": {
      "External id": 206728,
      "cbid": 211, "correlation": 206728
    }
  },
  {
    "ph": "s", "id": 206728, "pid": 76337, "tid": -914061504, "ts": 1716454224708946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224731836, "dur": 4,
    "args": {
      "External id": 206740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206740,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206740, "pid": 5, "tid": 7, "ts": 1716454224731836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708966, "dur": 6,
    "args": {
      "External id": 206740,
      "cbid": 211, "correlation": 206740
    }
  },
  {
    "ph": "s", "id": 206740, "pid": 76337, "tid": -914061504, "ts": 1716454224708966,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224731841, "dur": 55,
    "args": {
      "External id": 206743,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206743,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206743, "pid": 5, "tid": 7, "ts": 1716454224731841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224708994, "dur": 7,
    "args": {
      "External id": 206743,
      "cbid": 211, "correlation": 206743
    }
  },
  {
    "ph": "s", "id": 206743, "pid": 76337, "tid": -914061504, "ts": 1716454224708994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224731898, "dur": 37,
    "args": {
      "External id": 206752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206752,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206752, "pid": 5, "tid": 7, "ts": 1716454224731898,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709036, "dur": 10,
    "args": {
      "External id": 206752,
      "cbid": 211, "correlation": 206752
    }
  },
  {
    "ph": "s", "id": 206752, "pid": 76337, "tid": -914061504, "ts": 1716454224709036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709099, "dur": 0,
    "args": {
      "External id": 206762,
      "cbid": 317, "correlation": 206762
    }
  },
  {
    "ph": "f", "id": 206762, "pid": 76337, "tid": -914061504, "ts": 1716454224709099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709099, "dur": 0,
    "args": {
      "External id": 206763,
      "cbid": 203, "correlation": 206763
    }
  },
  {
    "ph": "f", "id": 206763, "pid": 76337, "tid": -914061504, "ts": 1716454224709099,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709100, "dur": 0,
    "args": {
      "External id": 206764,
      "cbid": 205, "correlation": 206764
    }
  },
  {
    "ph": "f", "id": 206764, "pid": 76337, "tid": -914061504, "ts": 1716454224709100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224731936, "dur": 39,
    "args": {
      "External id": 206768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206768,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206768, "pid": 5, "tid": 7, "ts": 1716454224731936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709113, "dur": 12,
    "args": {
      "External id": 206768,
      "cbid": 211, "correlation": 206768
    }
  },
  {
    "ph": "s", "id": 206768, "pid": 76337, "tid": -914061504, "ts": 1716454224709113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224731977, "dur": 14,
    "args": {
      "External id": 206770,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206770,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206770, "pid": 5, "tid": 7, "ts": 1716454224731977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709129, "dur": 6,
    "args": {
      "External id": 206770,
      "cbid": 211, "correlation": 206770
    }
  },
  {
    "ph": "s", "id": 206770, "pid": 76337, "tid": -914061504, "ts": 1716454224709129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224731992, "dur": 3,
    "args": {
      "External id": 206772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206772,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206772, "pid": 5, "tid": 7, "ts": 1716454224731992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709138, "dur": 5,
    "args": {
      "External id": 206772,
      "cbid": 211, "correlation": 206772
    }
  },
  {
    "ph": "s", "id": 206772, "pid": 76337, "tid": -914061504, "ts": 1716454224709138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709146, "dur": 0,
    "args": {
      "External id": 206773,
      "cbid": 51, "correlation": 206773
    }
  },
  {
    "ph": "s", "id": 206773, "pid": 76337, "tid": -914061504, "ts": 1716454224709146,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224731997, "dur": 700,
    "args": {
      "External id": 206774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206774,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206774, "pid": 5, "tid": 7, "ts": 1716454224731997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709147, "dur": 5,
    "args": {
      "External id": 206774,
      "cbid": 211, "correlation": 206774
    }
  },
  {
    "ph": "s", "id": 206774, "pid": 76337, "tid": -914061504, "ts": 1716454224709147,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224732698, "dur": 59,
    "args": {
      "External id": 206779,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206779,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206779, "pid": 5, "tid": 7, "ts": 1716454224732698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709174, "dur": 9,
    "args": {
      "External id": 206779,
      "cbid": 211, "correlation": 206779
    }
  },
  {
    "ph": "s", "id": 206779, "pid": 76337, "tid": -914061504, "ts": 1716454224709174,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224732759, "dur": 50,
    "args": {
      "External id": 206787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206787,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206787, "pid": 5, "tid": 7, "ts": 1716454224732759,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709207, "dur": 8,
    "args": {
      "External id": 206787,
      "cbid": 211, "correlation": 206787
    }
  },
  {
    "ph": "s", "id": 206787, "pid": 76337, "tid": -914061504, "ts": 1716454224709207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224732810, "dur": 36,
    "args": {
      "External id": 206795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206795,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206795, "pid": 5, "tid": 7, "ts": 1716454224732810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709236, "dur": 8,
    "args": {
      "External id": 206795,
      "cbid": 211, "correlation": 206795
    }
  },
  {
    "ph": "s", "id": 206795, "pid": 76337, "tid": -914061504, "ts": 1716454224709236,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224732847, "dur": 52,
    "args": {
      "External id": 206815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206815,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 206815, "pid": 5, "tid": 7, "ts": 1716454224732847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709314, "dur": 13,
    "args": {
      "External id": 206815,
      "cbid": 211, "correlation": 206815
    }
  },
  {
    "ph": "s", "id": 206815, "pid": 76337, "tid": -914061504, "ts": 1716454224709314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224732900, "dur": 4,
    "args": {
      "External id": 206827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206827,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 206827, "pid": 5, "tid": 7, "ts": 1716454224732900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709336, "dur": 6,
    "args": {
      "External id": 206827,
      "cbid": 211, "correlation": 206827
    }
  },
  {
    "ph": "s", "id": 206827, "pid": 76337, "tid": -914061504, "ts": 1716454224709336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224732905, "dur": 55,
    "args": {
      "External id": 206830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206830, "pid": 5, "tid": 7, "ts": 1716454224732905,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709354, "dur": 6,
    "args": {
      "External id": 206830,
      "cbid": 211, "correlation": 206830
    }
  },
  {
    "ph": "s", "id": 206830, "pid": 76337, "tid": -914061504, "ts": 1716454224709354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709410, "dur": 0,
    "args": {
      "External id": 206841,
      "cbid": 317, "correlation": 206841
    }
  },
  {
    "ph": "f", "id": 206841, "pid": 76337, "tid": -914061504, "ts": 1716454224709410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709411, "dur": 0,
    "args": {
      "External id": 206842,
      "cbid": 203, "correlation": 206842
    }
  },
  {
    "ph": "f", "id": 206842, "pid": 76337, "tid": -914061504, "ts": 1716454224709411,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709412, "dur": 0,
    "args": {
      "External id": 206843,
      "cbid": 205, "correlation": 206843
    }
  },
  {
    "ph": "f", "id": 206843, "pid": 76337, "tid": -914061504, "ts": 1716454224709412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709434, "dur": 1,
    "args": {
      "External id": 206847,
      "cbid": 251, "correlation": 206847
    }
  },
  {
    "ph": "f", "id": 206847, "pid": 76337, "tid": -914061504, "ts": 1716454224709434,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709436, "dur": 0,
    "args": {
      "External id": 206848,
      "cbid": 251, "correlation": 206848
    }
  },
  {
    "ph": "f", "id": 206848, "pid": 76337, "tid": -914061504, "ts": 1716454224709436,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709437, "dur": 0,
    "args": {
      "External id": 206849,
      "cbid": 251, "correlation": 206849
    }
  },
  {
    "ph": "f", "id": 206849, "pid": 76337, "tid": -914061504, "ts": 1716454224709437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709438, "dur": 0,
    "args": {
      "External id": 206850,
      "cbid": 251, "correlation": 206850
    }
  },
  {
    "ph": "f", "id": 206850, "pid": 76337, "tid": -914061504, "ts": 1716454224709438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709438, "dur": 0,
    "args": {
      "External id": 206851,
      "cbid": 251, "correlation": 206851
    }
  },
  {
    "ph": "f", "id": 206851, "pid": 76337, "tid": -914061504, "ts": 1716454224709438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709439, "dur": 0,
    "args": {
      "External id": 206852,
      "cbid": 251, "correlation": 206852
    }
  },
  {
    "ph": "f", "id": 206852, "pid": 76337, "tid": -914061504, "ts": 1716454224709439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709440, "dur": 0,
    "args": {
      "External id": 206853,
      "cbid": 251, "correlation": 206853
    }
  },
  {
    "ph": "f", "id": 206853, "pid": 76337, "tid": -914061504, "ts": 1716454224709440,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709441, "dur": 0,
    "args": {
      "External id": 206854,
      "cbid": 251, "correlation": 206854
    }
  },
  {
    "ph": "f", "id": 206854, "pid": 76337, "tid": -914061504, "ts": 1716454224709441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709442, "dur": 0,
    "args": {
      "External id": 206855,
      "cbid": 251, "correlation": 206855
    }
  },
  {
    "ph": "f", "id": 206855, "pid": 76337, "tid": -914061504, "ts": 1716454224709442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224732962, "dur": 112,
    "args": {
      "External id": 206856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206856,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206856, "pid": 5, "tid": 7, "ts": 1716454224732962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709444, "dur": 12,
    "args": {
      "External id": 206856,
      "cbid": 211, "correlation": 206856
    }
  },
  {
    "ph": "s", "id": 206856, "pid": 76337, "tid": -914061504, "ts": 1716454224709444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224733075, "dur": 59,
    "args": {
      "External id": 206862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206862,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206862, "pid": 5, "tid": 7, "ts": 1716454224733075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709480, "dur": 8,
    "args": {
      "External id": 206862,
      "cbid": 211, "correlation": 206862
    }
  },
  {
    "ph": "s", "id": 206862, "pid": 76337, "tid": -914061504, "ts": 1716454224709480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224733136, "dur": 567,
    "args": {
      "External id": 206871,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206871,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206871, "pid": 5, "tid": 7, "ts": 1716454224733136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709561, "dur": 15,
    "args": {
      "External id": 206871,
      "cbid": 211, "correlation": 206871
    }
  },
  {
    "ph": "s", "id": 206871, "pid": 76337, "tid": -914061504, "ts": 1716454224709561,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224733704, "dur": 182,
    "args": {
      "External id": 206893,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206893,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206893, "pid": 5, "tid": 7, "ts": 1716454224733704,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709617, "dur": 11,
    "args": {
      "External id": 206893,
      "cbid": 211, "correlation": 206893
    }
  },
  {
    "ph": "s", "id": 206893, "pid": 76337, "tid": -914061504, "ts": 1716454224709617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709703, "dur": 1,
    "args": {
      "External id": 206904,
      "cbid": 251, "correlation": 206904
    }
  },
  {
    "ph": "f", "id": 206904, "pid": 76337, "tid": -914061504, "ts": 1716454224709703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224733887, "dur": 197,
    "args": {
      "External id": 206905,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206905,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206905, "pid": 5, "tid": 7, "ts": 1716454224733887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709708, "dur": 14,
    "args": {
      "External id": 206905,
      "cbid": 211, "correlation": 206905
    }
  },
  {
    "ph": "s", "id": 206905, "pid": 76337, "tid": -914061504, "ts": 1716454224709708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709777, "dur": 1,
    "args": {
      "External id": 206916,
      "cbid": 251, "correlation": 206916
    }
  },
  {
    "ph": "f", "id": 206916, "pid": 76337, "tid": -914061504, "ts": 1716454224709777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224734085, "dur": 188,
    "args": {
      "External id": 206917,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206917,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206917, "pid": 5, "tid": 7, "ts": 1716454224734085,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709781, "dur": 11,
    "args": {
      "External id": 206917,
      "cbid": 211, "correlation": 206917
    }
  },
  {
    "ph": "s", "id": 206917, "pid": 76337, "tid": -914061504, "ts": 1716454224709781,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709843, "dur": 1,
    "args": {
      "External id": 206928,
      "cbid": 251, "correlation": 206928
    }
  },
  {
    "ph": "f", "id": 206928, "pid": 76337, "tid": -914061504, "ts": 1716454224709843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224734274, "dur": 186,
    "args": {
      "External id": 206929,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206929,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206929, "pid": 5, "tid": 7, "ts": 1716454224734274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709847, "dur": 11,
    "args": {
      "External id": 206929,
      "cbid": 211, "correlation": 206929
    }
  },
  {
    "ph": "s", "id": 206929, "pid": 76337, "tid": -914061504, "ts": 1716454224709847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224734462, "dur": 18799,
    "args": {
      "External id": 206950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206950,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 206950, "pid": 5, "tid": 7, "ts": 1716454224734462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224709928, "dur": 12,
    "args": {
      "External id": 206950,
      "cbid": 211, "correlation": 206950
    }
  },
  {
    "ph": "s", "id": 206950, "pid": 76337, "tid": -914061504, "ts": 1716454224709928,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710034, "dur": 1,
    "args": {
      "External id": 206968,
      "cbid": 251, "correlation": 206968
    }
  },
  {
    "ph": "f", "id": 206968, "pid": 76337, "tid": -914061504, "ts": 1716454224710034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224753262, "dur": 205,
    "args": {
      "External id": 206970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206970,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 206970, "pid": 5, "tid": 7, "ts": 1716454224753262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710040, "dur": 14,
    "args": {
      "External id": 206970,
      "cbid": 211, "correlation": 206970
    }
  },
  {
    "ph": "s", "id": 206970, "pid": 76337, "tid": -914061504, "ts": 1716454224710040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224753469, "dur": 66,
    "args": {
      "External id": 206978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206978,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206978, "pid": 5, "tid": 7, "ts": 1716454224753469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710112, "dur": 13,
    "args": {
      "External id": 206978,
      "cbid": 211, "correlation": 206978
    }
  },
  {
    "ph": "s", "id": 206978, "pid": 76337, "tid": -914061504, "ts": 1716454224710112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224753537, "dur": 96,
    "args": {
      "External id": 206986,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206986,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206986, "pid": 5, "tid": 7, "ts": 1716454224753537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710152, "dur": 9,
    "args": {
      "External id": 206986,
      "cbid": 211, "correlation": 206986
    }
  },
  {
    "ph": "s", "id": 206986, "pid": 76337, "tid": -914061504, "ts": 1716454224710152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224753634, "dur": 55,
    "args": {
      "External id": 206997,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 206997,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 206997, "pid": 5, "tid": 7, "ts": 1716454224753634,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710226, "dur": 12,
    "args": {
      "External id": 206997,
      "cbid": 211, "correlation": 206997
    }
  },
  {
    "ph": "s", "id": 206997, "pid": 76337, "tid": -914061504, "ts": 1716454224710226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224753690, "dur": 93,
    "args": {
      "External id": 207019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207019,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207019, "pid": 5, "tid": 7, "ts": 1716454224753690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710257, "dur": 8,
    "args": {
      "External id": 207019,
      "cbid": 211, "correlation": 207019
    }
  },
  {
    "ph": "s", "id": 207019, "pid": 76337, "tid": -914061504, "ts": 1716454224710257,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710343, "dur": 1,
    "args": {
      "External id": 207030,
      "cbid": 251, "correlation": 207030
    }
  },
  {
    "ph": "f", "id": 207030, "pid": 76337, "tid": -914061504, "ts": 1716454224710343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224753785, "dur": 103,
    "args": {
      "External id": 207031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207031,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207031, "pid": 5, "tid": 7, "ts": 1716454224753785,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710348, "dur": 13,
    "args": {
      "External id": 207031,
      "cbid": 211, "correlation": 207031
    }
  },
  {
    "ph": "s", "id": 207031, "pid": 76337, "tid": -914061504, "ts": 1716454224710348,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710420, "dur": 1,
    "args": {
      "External id": 207042,
      "cbid": 251, "correlation": 207042
    }
  },
  {
    "ph": "f", "id": 207042, "pid": 76337, "tid": -914061504, "ts": 1716454224710420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710423, "dur": 0,
    "args": {
      "External id": 207043,
      "cbid": 251, "correlation": 207043
    }
  },
  {
    "ph": "f", "id": 207043, "pid": 76337, "tid": -914061504, "ts": 1716454224710423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224753889, "dur": 10,
    "args": {
      "External id": 207044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207044,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 207044, "pid": 5, "tid": 7, "ts": 1716454224753889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710425, "dur": 12,
    "args": {
      "External id": 207044,
      "cbid": 211, "correlation": 207044
    }
  },
  {
    "ph": "s", "id": 207044, "pid": 76337, "tid": -914061504, "ts": 1716454224710425,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224753901, "dur": 5,
    "args": {
      "External id": 207046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207046,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 207046, "pid": 5, "tid": 7, "ts": 1716454224753901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710439, "dur": 6,
    "args": {
      "External id": 207046,
      "cbid": 211, "correlation": 207046
    }
  },
  {
    "ph": "s", "id": 207046, "pid": 76337, "tid": -914061504, "ts": 1716454224710439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710500, "dur": 1,
    "args": {
      "External id": 207057,
      "cbid": 251, "correlation": 207057
    }
  },
  {
    "ph": "f", "id": 207057, "pid": 76337, "tid": -914061504, "ts": 1716454224710500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710503, "dur": 0,
    "args": {
      "External id": 207058,
      "cbid": 251, "correlation": 207058
    }
  },
  {
    "ph": "f", "id": 207058, "pid": 76337, "tid": -914061504, "ts": 1716454224710503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224753907, "dur": 6,
    "args": {
      "External id": 207059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207059,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 207059, "pid": 5, "tid": 7, "ts": 1716454224753907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710504, "dur": 13,
    "args": {
      "External id": 207059,
      "cbid": 211, "correlation": 207059
    }
  },
  {
    "ph": "s", "id": 207059, "pid": 76337, "tid": -914061504, "ts": 1716454224710504,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224753915, "dur": 3,
    "args": {
      "External id": 207061,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207061,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 207061, "pid": 5, "tid": 7, "ts": 1716454224753915,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710519, "dur": 5,
    "args": {
      "External id": 207061,
      "cbid": 211, "correlation": 207061
    }
  },
  {
    "ph": "s", "id": 207061, "pid": 76337, "tid": -914061504, "ts": 1716454224710519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224753920, "dur": 156,
    "args": {
      "External id": 207082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207082,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 207082, "pid": 5, "tid": 7, "ts": 1716454224753920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710592, "dur": 12,
    "args": {
      "External id": 207082,
      "cbid": 211, "correlation": 207082
    }
  },
  {
    "ph": "s", "id": 207082, "pid": 76337, "tid": -914061504, "ts": 1716454224710592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710689, "dur": 1,
    "args": {
      "External id": 207100,
      "cbid": 251, "correlation": 207100
    }
  },
  {
    "ph": "f", "id": 207100, "pid": 76337, "tid": -914061504, "ts": 1716454224710689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224754077, "dur": 107,
    "args": {
      "External id": 207102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207102,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 207102, "pid": 5, "tid": 7, "ts": 1716454224754077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710695, "dur": 14,
    "args": {
      "External id": 207102,
      "cbid": 211, "correlation": 207102
    }
  },
  {
    "ph": "s", "id": 207102, "pid": 76337, "tid": -914061504, "ts": 1716454224710695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224754186, "dur": 35,
    "args": {
      "External id": 207110,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207110,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207110, "pid": 5, "tid": 7, "ts": 1716454224754186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710764, "dur": 11,
    "args": {
      "External id": 207110,
      "cbid": 211, "correlation": 207110
    }
  },
  {
    "ph": "s", "id": 207110, "pid": 76337, "tid": -914061504, "ts": 1716454224710764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224754222, "dur": 66,
    "args": {
      "External id": 207118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207118,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207118, "pid": 5, "tid": 7, "ts": 1716454224754222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710805, "dur": 10,
    "args": {
      "External id": 207118,
      "cbid": 211, "correlation": 207118
    }
  },
  {
    "ph": "s", "id": 207118, "pid": 76337, "tid": -914061504, "ts": 1716454224710805,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224754290, "dur": 93,
    "args": {
      "External id": 207140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207140,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207140, "pid": 5, "tid": 7, "ts": 1716454224754290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710857, "dur": 11,
    "args": {
      "External id": 207140,
      "cbid": 211, "correlation": 207140
    }
  },
  {
    "ph": "s", "id": 207140, "pid": 76337, "tid": -914061504, "ts": 1716454224710857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710942, "dur": 1,
    "args": {
      "External id": 207156,
      "cbid": 251, "correlation": 207156
    }
  },
  {
    "ph": "f", "id": 207156, "pid": 76337, "tid": -914061504, "ts": 1716454224710942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224754384, "dur": 579,
    "args": {
      "External id": 207158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207158,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207158, "pid": 5, "tid": 7, "ts": 1716454224754384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224710948, "dur": 13,
    "args": {
      "External id": 207158,
      "cbid": 211, "correlation": 207158
    }
  },
  {
    "ph": "s", "id": 207158, "pid": 76337, "tid": -914061504, "ts": 1716454224710948,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224754964, "dur": 244,
    "args": {
      "External id": 207166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207166, "pid": 5, "tid": 7, "ts": 1716454224754964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711023, "dur": 13,
    "args": {
      "External id": 207166,
      "cbid": 211, "correlation": 207166
    }
  },
  {
    "ph": "s", "id": 207166, "pid": 76337, "tid": -914061504, "ts": 1716454224711023,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224755209, "dur": 253,
    "args": {
      "External id": 207174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207174,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207174, "pid": 5, "tid": 7, "ts": 1716454224755209,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711054, "dur": 8,
    "args": {
      "External id": 207174,
      "cbid": 211, "correlation": 207174
    }
  },
  {
    "ph": "s", "id": 207174, "pid": 76337, "tid": -914061504, "ts": 1716454224711054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711136, "dur": 1,
    "args": {
      "External id": 207190,
      "cbid": 251, "correlation": 207190
    }
  },
  {
    "ph": "f", "id": 207190, "pid": 76337, "tid": -914061504, "ts": 1716454224711136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711140, "dur": 0,
    "args": {
      "External id": 207192,
      "cbid": 251, "correlation": 207192
    }
  },
  {
    "ph": "f", "id": 207192, "pid": 76337, "tid": -914061504, "ts": 1716454224711140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224755464, "dur": 363,
    "args": {
      "External id": 207193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207193,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 207193, "pid": 5, "tid": 7, "ts": 1716454224755464,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711143, "dur": 14,
    "args": {
      "External id": 207193,
      "cbid": 211, "correlation": 207193
    }
  },
  {
    "ph": "s", "id": 207193, "pid": 76337, "tid": -914061504, "ts": 1716454224711143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224755828, "dur": 50,
    "args": {
      "External id": 207201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207201,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207201, "pid": 5, "tid": 7, "ts": 1716454224755828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711185, "dur": 10,
    "args": {
      "External id": 207201,
      "cbid": 211, "correlation": 207201
    }
  },
  {
    "ph": "s", "id": 207201, "pid": 76337, "tid": -914061504, "ts": 1716454224711185,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224755879, "dur": 160,
    "args": {
      "External id": 207212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207212, "pid": 5, "tid": 7, "ts": 1716454224755879,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711253, "dur": 12,
    "args": {
      "External id": 207212,
      "cbid": 211, "correlation": 207212
    }
  },
  {
    "ph": "s", "id": 207212, "pid": 76337, "tid": -914061504, "ts": 1716454224711253,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711318, "dur": 0,
    "args": {
      "External id": 207224,
      "cbid": 317, "correlation": 207224
    }
  },
  {
    "ph": "f", "id": 207224, "pid": 76337, "tid": -914061504, "ts": 1716454224711318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711318, "dur": 0,
    "args": {
      "External id": 207225,
      "cbid": 203, "correlation": 207225
    }
  },
  {
    "ph": "f", "id": 207225, "pid": 76337, "tid": -914061504, "ts": 1716454224711318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711319, "dur": 0,
    "args": {
      "External id": 207226,
      "cbid": 205, "correlation": 207226
    }
  },
  {
    "ph": "f", "id": 207226, "pid": 76337, "tid": -914061504, "ts": 1716454224711319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711341, "dur": 1,
    "args": {
      "External id": 207230,
      "cbid": 251, "correlation": 207230
    }
  },
  {
    "ph": "f", "id": 207230, "pid": 76337, "tid": -914061504, "ts": 1716454224711341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711343, "dur": 0,
    "args": {
      "External id": 207231,
      "cbid": 251, "correlation": 207231
    }
  },
  {
    "ph": "f", "id": 207231, "pid": 76337, "tid": -914061504, "ts": 1716454224711343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711344, "dur": 0,
    "args": {
      "External id": 207232,
      "cbid": 251, "correlation": 207232
    }
  },
  {
    "ph": "f", "id": 207232, "pid": 76337, "tid": -914061504, "ts": 1716454224711344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711345, "dur": 0,
    "args": {
      "External id": 207233,
      "cbid": 251, "correlation": 207233
    }
  },
  {
    "ph": "f", "id": 207233, "pid": 76337, "tid": -914061504, "ts": 1716454224711345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711346, "dur": 0,
    "args": {
      "External id": 207234,
      "cbid": 251, "correlation": 207234
    }
  },
  {
    "ph": "f", "id": 207234, "pid": 76337, "tid": -914061504, "ts": 1716454224711346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711347, "dur": 0,
    "args": {
      "External id": 207235,
      "cbid": 251, "correlation": 207235
    }
  },
  {
    "ph": "f", "id": 207235, "pid": 76337, "tid": -914061504, "ts": 1716454224711347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711348, "dur": 0,
    "args": {
      "External id": 207236,
      "cbid": 251, "correlation": 207236
    }
  },
  {
    "ph": "f", "id": 207236, "pid": 76337, "tid": -914061504, "ts": 1716454224711348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711349, "dur": 0,
    "args": {
      "External id": 207237,
      "cbid": 251, "correlation": 207237
    }
  },
  {
    "ph": "f", "id": 207237, "pid": 76337, "tid": -914061504, "ts": 1716454224711349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711350, "dur": 0,
    "args": {
      "External id": 207238,
      "cbid": 251, "correlation": 207238
    }
  },
  {
    "ph": "f", "id": 207238, "pid": 76337, "tid": -914061504, "ts": 1716454224711350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224756040, "dur": 116,
    "args": {
      "External id": 207239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207239,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 207239, "pid": 5, "tid": 7, "ts": 1716454224756040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711353, "dur": 12,
    "args": {
      "External id": 207239,
      "cbid": 211, "correlation": 207239
    }
  },
  {
    "ph": "s", "id": 207239, "pid": 76337, "tid": -914061504, "ts": 1716454224711353,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224756157, "dur": 60,
    "args": {
      "External id": 207245,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207245,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207245, "pid": 5, "tid": 7, "ts": 1716454224756157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711387, "dur": 9,
    "args": {
      "External id": 207245,
      "cbid": 211, "correlation": 207245
    }
  },
  {
    "ph": "s", "id": 207245, "pid": 76337, "tid": -914061504, "ts": 1716454224711387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224756218, "dur": 50,
    "args": {
      "External id": 207253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207253,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207253, "pid": 5, "tid": 7, "ts": 1716454224756218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711418, "dur": 8,
    "args": {
      "External id": 207253,
      "cbid": 211, "correlation": 207253
    }
  },
  {
    "ph": "s", "id": 207253, "pid": 76337, "tid": -914061504, "ts": 1716454224711418,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711491, "dur": 0,
    "args": {
      "External id": 207263,
      "cbid": 317, "correlation": 207263
    }
  },
  {
    "ph": "f", "id": 207263, "pid": 76337, "tid": -914061504, "ts": 1716454224711491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711492, "dur": 0,
    "args": {
      "External id": 207264,
      "cbid": 203, "correlation": 207264
    }
  },
  {
    "ph": "f", "id": 207264, "pid": 76337, "tid": -914061504, "ts": 1716454224711492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711493, "dur": 0,
    "args": {
      "External id": 207265,
      "cbid": 205, "correlation": 207265
    }
  },
  {
    "ph": "f", "id": 207265, "pid": 76337, "tid": -914061504, "ts": 1716454224711493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224756269, "dur": 42,
    "args": {
      "External id": 207269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207269,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207269, "pid": 5, "tid": 7, "ts": 1716454224756269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711510, "dur": 12,
    "args": {
      "External id": 207269,
      "cbid": 211, "correlation": 207269
    }
  },
  {
    "ph": "s", "id": 207269, "pid": 76337, "tid": -914061504, "ts": 1716454224711510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224756313, "dur": 14,
    "args": {
      "External id": 207271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207271,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207271, "pid": 5, "tid": 7, "ts": 1716454224756313,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711525, "dur": 5,
    "args": {
      "External id": 207271,
      "cbid": 211, "correlation": 207271
    }
  },
  {
    "ph": "s", "id": 207271, "pid": 76337, "tid": -914061504, "ts": 1716454224711525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224756329, "dur": 1,
    "args": {
      "External id": 207273,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 207273,
      "bytes": 1536, "memory bandwidth (GB/s)": 0.9056603773584906
    }
  },
  {
    "ph": "f", "id": 207273, "pid": 5, "tid": 7, "ts": 1716454224756329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711544, "dur": 17,
    "args": {
      "External id": 207273,
      "cbid": 51, "correlation": 207273
    }
  },
  {
    "ph": "s", "id": 207273, "pid": 76337, "tid": -914061504, "ts": 1716454224711544,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224756333, "dur": 363,
    "args": {
      "External id": 207274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207274,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 7.2,
      "warps per SM": 57.6,
      "grid": [2, 96, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207274, "pid": 5, "tid": 7, "ts": 1716454224756333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711563, "dur": 9,
    "args": {
      "External id": 207274,
      "cbid": 211, "correlation": 207274
    }
  },
  {
    "ph": "s", "id": 207274, "pid": 76337, "tid": -914061504, "ts": 1716454224711563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224756698, "dur": 12,
    "args": {
      "External id": 207276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207276,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207276, "pid": 5, "tid": 7, "ts": 1716454224756698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711581, "dur": 8,
    "args": {
      "External id": 207276,
      "cbid": 211, "correlation": 207276
    }
  },
  {
    "ph": "s", "id": 207276, "pid": 76337, "tid": -914061504, "ts": 1716454224711581,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224756712, "dur": 15,
    "args": {
      "External id": 207282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207282, "pid": 5, "tid": 7, "ts": 1716454224756712,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711612, "dur": 8,
    "args": {
      "External id": 207282,
      "cbid": 211, "correlation": 207282
    }
  },
  {
    "ph": "s", "id": 207282, "pid": 76337, "tid": -914061504, "ts": 1716454224711612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224756728, "dur": 19,
    "args": {
      "External id": 207302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207302,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 207302, "pid": 5, "tid": 7, "ts": 1716454224756728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711704, "dur": 12,
    "args": {
      "External id": 207302,
      "cbid": 211, "correlation": 207302
    }
  },
  {
    "ph": "s", "id": 207302, "pid": 76337, "tid": -914061504, "ts": 1716454224711704,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224756749, "dur": 5,
    "args": {
      "External id": 207314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207314,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 207314, "pid": 5, "tid": 7, "ts": 1716454224756749,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711726, "dur": 6,
    "args": {
      "External id": 207314,
      "cbid": 211, "correlation": 207314
    }
  },
  {
    "ph": "s", "id": 207314, "pid": 76337, "tid": -914061504, "ts": 1716454224711726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224756755, "dur": 18,
    "args": {
      "External id": 207317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207317,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207317, "pid": 5, "tid": 7, "ts": 1716454224756755,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711745, "dur": 7,
    "args": {
      "External id": 207317,
      "cbid": 211, "correlation": 207317
    }
  },
  {
    "ph": "s", "id": 207317, "pid": 76337, "tid": -914061504, "ts": 1716454224711745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224756773, "dur": 11,
    "args": {
      "External id": 207326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207326,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207326, "pid": 5, "tid": 7, "ts": 1716454224756773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711784, "dur": 10,
    "args": {
      "External id": 207326,
      "cbid": 211, "correlation": 207326
    }
  },
  {
    "ph": "s", "id": 207326, "pid": 76337, "tid": -914061504, "ts": 1716454224711784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711841, "dur": 0,
    "args": {
      "External id": 207336,
      "cbid": 317, "correlation": 207336
    }
  },
  {
    "ph": "f", "id": 207336, "pid": 76337, "tid": -914061504, "ts": 1716454224711841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711841, "dur": 0,
    "args": {
      "External id": 207337,
      "cbid": 203, "correlation": 207337
    }
  },
  {
    "ph": "f", "id": 207337, "pid": 76337, "tid": -914061504, "ts": 1716454224711841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711842, "dur": 0,
    "args": {
      "External id": 207338,
      "cbid": 205, "correlation": 207338
    }
  },
  {
    "ph": "f", "id": 207338, "pid": 76337, "tid": -914061504, "ts": 1716454224711842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224756786, "dur": 11,
    "args": {
      "External id": 207342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207342,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [24, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207342, "pid": 5, "tid": 7, "ts": 1716454224756786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711857, "dur": 12,
    "args": {
      "External id": 207342,
      "cbid": 211, "correlation": 207342
    }
  },
  {
    "ph": "s", "id": 207342, "pid": 76337, "tid": -914061504, "ts": 1716454224711857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224756798, "dur": 24,
    "args": {
      "External id": 207344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 10, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207344, "pid": 5, "tid": 7, "ts": 1716454224756798,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711871, "dur": 5,
    "args": {
      "External id": 207344,
      "cbid": 211, "correlation": 207344
    }
  },
  {
    "ph": "s", "id": 207344, "pid": 76337, "tid": -914061504, "ts": 1716454224711871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224756824, "dur": 3,
    "args": {
      "External id": 207346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.05,
      "warps per SM": 0.4,
      "grid": [4, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 1
    }
  },
  {
    "ph": "f", "id": 207346, "pid": 5, "tid": 7, "ts": 1716454224756824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711882, "dur": 6,
    "args": {
      "External id": 207346,
      "cbid": 211, "correlation": 207346
    }
  },
  {
    "ph": "s", "id": 207346, "pid": 76337, "tid": -914061504, "ts": 1716454224711882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711892, "dur": 0,
    "args": {
      "External id": 207347,
      "cbid": 51, "correlation": 207347
    }
  },
  {
    "ph": "s", "id": 207347, "pid": 76337, "tid": -914061504, "ts": 1716454224711892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224756829, "dur": 359,
    "args": {
      "External id": 207348,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207348,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207348, "pid": 5, "tid": 7, "ts": 1716454224756829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711893, "dur": 7,
    "args": {
      "External id": 207348,
      "cbid": 211, "correlation": 207348
    }
  },
  {
    "ph": "s", "id": 207348, "pid": 76337, "tid": -914061504, "ts": 1716454224711893,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224757190, "dur": 21,
    "args": {
      "External id": 207349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207349,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207349, "pid": 5, "tid": 7, "ts": 1716454224757190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711902, "dur": 5,
    "args": {
      "External id": 207349,
      "cbid": 211, "correlation": 207349
    }
  },
  {
    "ph": "s", "id": 207349, "pid": 76337, "tid": -914061504, "ts": 1716454224711902,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224757212, "dur": 33,
    "args": {
      "External id": 207355,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207355,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207355, "pid": 5, "tid": 7, "ts": 1716454224757212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711930, "dur": 9,
    "args": {
      "External id": 207355,
      "cbid": 211, "correlation": 207355
    }
  },
  {
    "ph": "s", "id": 207355, "pid": 76337, "tid": -914061504, "ts": 1716454224711930,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224757246, "dur": 3,
    "args": {
      "External id": 207363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207363,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 207363, "pid": 5, "tid": 7, "ts": 1716454224757246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224711983, "dur": 11,
    "args": {
      "External id": 207363,
      "cbid": 211, "correlation": 207363
    }
  },
  {
    "ph": "s", "id": 207363, "pid": 76337, "tid": -914061504, "ts": 1716454224711983,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712051, "dur": 1,
    "args": {
      "External id": 207379,
      "cbid": 251, "correlation": 207379
    }
  },
  {
    "ph": "f", "id": 207379, "pid": 76337, "tid": -914061504, "ts": 1716454224712051,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712056, "dur": 0,
    "args": {
      "External id": 207381,
      "cbid": 251, "correlation": 207381
    }
  },
  {
    "ph": "f", "id": 207381, "pid": 76337, "tid": -914061504, "ts": 1716454224712056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224757250, "dur": 12,
    "args": {
      "External id": 207382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207382,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 207382, "pid": 5, "tid": 7, "ts": 1716454224757250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712059, "dur": 12,
    "args": {
      "External id": 207382,
      "cbid": 211, "correlation": 207382
    }
  },
  {
    "ph": "s", "id": 207382, "pid": 76337, "tid": -914061504, "ts": 1716454224712059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224757264, "dur": 5,
    "args": {
      "External id": 207384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207384,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 207384, "pid": 5, "tid": 7, "ts": 1716454224757264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712072, "dur": 6,
    "args": {
      "External id": 207384,
      "cbid": 211, "correlation": 207384
    }
  },
  {
    "ph": "s", "id": 207384, "pid": 76337, "tid": -914061504, "ts": 1716454224712072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224757271, "dur": 30,
    "args": {
      "External id": 207394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207394,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207394, "pid": 5, "tid": 7, "ts": 1716454224757271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712131, "dur": 12,
    "args": {
      "External id": 207394,
      "cbid": 211, "correlation": 207394
    }
  },
  {
    "ph": "s", "id": 207394, "pid": 76337, "tid": -914061504, "ts": 1716454224712131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224757302, "dur": 30,
    "args": {
      "External id": 207414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207414,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 207414, "pid": 5, "tid": 7, "ts": 1716454224757302,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712198, "dur": 11,
    "args": {
      "External id": 207414,
      "cbid": 211, "correlation": 207414
    }
  },
  {
    "ph": "s", "id": 207414, "pid": 76337, "tid": -914061504, "ts": 1716454224712198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224757333, "dur": 4,
    "args": {
      "External id": 207426,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207426,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 207426, "pid": 5, "tid": 7, "ts": 1716454224757333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712219, "dur": 6,
    "args": {
      "External id": 207426,
      "cbid": 211, "correlation": 207426
    }
  },
  {
    "ph": "s", "id": 207426, "pid": 76337, "tid": -914061504, "ts": 1716454224712219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224757338, "dur": 30,
    "args": {
      "External id": 207429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207429, "pid": 5, "tid": 7, "ts": 1716454224757338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712238, "dur": 6,
    "args": {
      "External id": 207429,
      "cbid": 211, "correlation": 207429
    }
  },
  {
    "ph": "s", "id": 207429, "pid": 76337, "tid": -914061504, "ts": 1716454224712238,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224757369, "dur": 22,
    "args": {
      "External id": 207438,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207438,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207438, "pid": 5, "tid": 7, "ts": 1716454224757369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712279, "dur": 9,
    "args": {
      "External id": 207438,
      "cbid": 211, "correlation": 207438
    }
  },
  {
    "ph": "s", "id": 207438, "pid": 76337, "tid": -914061504, "ts": 1716454224712279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712343, "dur": 0,
    "args": {
      "External id": 207448,
      "cbid": 317, "correlation": 207448
    }
  },
  {
    "ph": "f", "id": 207448, "pid": 76337, "tid": -914061504, "ts": 1716454224712343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712343, "dur": 0,
    "args": {
      "External id": 207449,
      "cbid": 203, "correlation": 207449
    }
  },
  {
    "ph": "f", "id": 207449, "pid": 76337, "tid": -914061504, "ts": 1716454224712343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712344, "dur": 0,
    "args": {
      "External id": 207450,
      "cbid": 205, "correlation": 207450
    }
  },
  {
    "ph": "f", "id": 207450, "pid": 76337, "tid": -914061504, "ts": 1716454224712344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224757392, "dur": 23,
    "args": {
      "External id": 207454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207454,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207454, "pid": 5, "tid": 7, "ts": 1716454224757392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712362, "dur": 12,
    "args": {
      "External id": 207454,
      "cbid": 211, "correlation": 207454
    }
  },
  {
    "ph": "s", "id": 207454, "pid": 76337, "tid": -914061504, "ts": 1716454224712362,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224757417, "dur": 44,
    "args": {
      "External id": 207456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207456, "pid": 5, "tid": 7, "ts": 1716454224757417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712377, "dur": 5,
    "args": {
      "External id": 207456,
      "cbid": 211, "correlation": 207456
    }
  },
  {
    "ph": "s", "id": 207456, "pid": 76337, "tid": -914061504, "ts": 1716454224712377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224757463, "dur": 659,
    "args": {
      "External id": 207458,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207458,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207458, "pid": 5, "tid": 7, "ts": 1716454224757463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712392, "dur": 9,
    "args": {
      "External id": 207458,
      "cbid": 211, "correlation": 207458
    }
  },
  {
    "ph": "s", "id": 207458, "pid": 76337, "tid": -914061504, "ts": 1716454224712392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224758123, "dur": 23,
    "args": {
      "External id": 207460,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207460,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207460, "pid": 5, "tid": 7, "ts": 1716454224758123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712405, "dur": 5,
    "args": {
      "External id": 207460,
      "cbid": 211, "correlation": 207460
    }
  },
  {
    "ph": "s", "id": 207460, "pid": 76337, "tid": -914061504, "ts": 1716454224712405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224758147, "dur": 33,
    "args": {
      "External id": 207466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207466,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207466, "pid": 5, "tid": 7, "ts": 1716454224758147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712434, "dur": 8,
    "args": {
      "External id": 207466,
      "cbid": 211, "correlation": 207466
    }
  },
  {
    "ph": "s", "id": 207466, "pid": 76337, "tid": -914061504, "ts": 1716454224712434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712492, "dur": 0,
    "args": {
      "External id": 207476,
      "cbid": 317, "correlation": 207476
    }
  },
  {
    "ph": "f", "id": 207476, "pid": 76337, "tid": -914061504, "ts": 1716454224712492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712493, "dur": 0,
    "args": {
      "External id": 207477,
      "cbid": 203, "correlation": 207477
    }
  },
  {
    "ph": "f", "id": 207477, "pid": 76337, "tid": -914061504, "ts": 1716454224712493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712493, "dur": 0,
    "args": {
      "External id": 207478,
      "cbid": 205, "correlation": 207478
    }
  },
  {
    "ph": "f", "id": 207478, "pid": 76337, "tid": -914061504, "ts": 1716454224712493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712515, "dur": 1,
    "args": {
      "External id": 207482,
      "cbid": 251, "correlation": 207482
    }
  },
  {
    "ph": "f", "id": 207482, "pid": 76337, "tid": -914061504, "ts": 1716454224712515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712517, "dur": 0,
    "args": {
      "External id": 207483,
      "cbid": 251, "correlation": 207483
    }
  },
  {
    "ph": "f", "id": 207483, "pid": 76337, "tid": -914061504, "ts": 1716454224712517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712518, "dur": 0,
    "args": {
      "External id": 207484,
      "cbid": 251, "correlation": 207484
    }
  },
  {
    "ph": "f", "id": 207484, "pid": 76337, "tid": -914061504, "ts": 1716454224712518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712519, "dur": 0,
    "args": {
      "External id": 207485,
      "cbid": 251, "correlation": 207485
    }
  },
  {
    "ph": "f", "id": 207485, "pid": 76337, "tid": -914061504, "ts": 1716454224712519,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712520, "dur": 0,
    "args": {
      "External id": 207486,
      "cbid": 251, "correlation": 207486
    }
  },
  {
    "ph": "f", "id": 207486, "pid": 76337, "tid": -914061504, "ts": 1716454224712520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712521, "dur": 0,
    "args": {
      "External id": 207487,
      "cbid": 251, "correlation": 207487
    }
  },
  {
    "ph": "f", "id": 207487, "pid": 76337, "tid": -914061504, "ts": 1716454224712521,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712522, "dur": 0,
    "args": {
      "External id": 207488,
      "cbid": 251, "correlation": 207488
    }
  },
  {
    "ph": "f", "id": 207488, "pid": 76337, "tid": -914061504, "ts": 1716454224712522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712523, "dur": 0,
    "args": {
      "External id": 207489,
      "cbid": 251, "correlation": 207489
    }
  },
  {
    "ph": "f", "id": 207489, "pid": 76337, "tid": -914061504, "ts": 1716454224712523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712524, "dur": 0,
    "args": {
      "External id": 207490,
      "cbid": 251, "correlation": 207490
    }
  },
  {
    "ph": "f", "id": 207490, "pid": 76337, "tid": -914061504, "ts": 1716454224712524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224758182, "dur": 52,
    "args": {
      "External id": 207491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207491,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [12, 5, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 207491, "pid": 5, "tid": 7, "ts": 1716454224758182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712527, "dur": 12,
    "args": {
      "External id": 207491,
      "cbid": 211, "correlation": 207491
    }
  },
  {
    "ph": "s", "id": 207491, "pid": 76337, "tid": -914061504, "ts": 1716454224712527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224758235, "dur": 32,
    "args": {
      "External id": 207497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207497,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207497, "pid": 5, "tid": 7, "ts": 1716454224758235,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712558, "dur": 8,
    "args": {
      "External id": 207497,
      "cbid": 211, "correlation": 207497
    }
  },
  {
    "ph": "s", "id": 207497, "pid": 76337, "tid": -914061504, "ts": 1716454224712558,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224758268, "dur": 27,
    "args": {
      "External id": 207505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207505,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207505, "pid": 5, "tid": 7, "ts": 1716454224758268,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712588, "dur": 8,
    "args": {
      "External id": 207505,
      "cbid": 211, "correlation": 207505
    }
  },
  {
    "ph": "s", "id": 207505, "pid": 76337, "tid": -914061504, "ts": 1716454224712588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224758296, "dur": 19,
    "args": {
      "External id": 207513,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207513,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207513, "pid": 5, "tid": 7, "ts": 1716454224758296,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712616, "dur": 9,
    "args": {
      "External id": 207513,
      "cbid": 211, "correlation": 207513
    }
  },
  {
    "ph": "s", "id": 207513, "pid": 76337, "tid": -914061504, "ts": 1716454224712616,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224758317, "dur": 29,
    "args": {
      "External id": 207533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207533,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 207533, "pid": 5, "tid": 7, "ts": 1716454224758317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712699, "dur": 12,
    "args": {
      "External id": 207533,
      "cbid": 211, "correlation": 207533
    }
  },
  {
    "ph": "s", "id": 207533, "pid": 76337, "tid": -914061504, "ts": 1716454224712699,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224758347, "dur": 4,
    "args": {
      "External id": 207545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207545,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 207545, "pid": 5, "tid": 7, "ts": 1716454224758347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712720, "dur": 7,
    "args": {
      "External id": 207545,
      "cbid": 211, "correlation": 207545
    }
  },
  {
    "ph": "s", "id": 207545, "pid": 76337, "tid": -914061504, "ts": 1716454224712720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224758353, "dur": 29,
    "args": {
      "External id": 207548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207548,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207548, "pid": 5, "tid": 7, "ts": 1716454224758353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712738, "dur": 6,
    "args": {
      "External id": 207548,
      "cbid": 211, "correlation": 207548
    }
  },
  {
    "ph": "s", "id": 207548, "pid": 76337, "tid": -914061504, "ts": 1716454224712738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712796, "dur": 0,
    "args": {
      "External id": 207559,
      "cbid": 317, "correlation": 207559
    }
  },
  {
    "ph": "f", "id": 207559, "pid": 76337, "tid": -914061504, "ts": 1716454224712796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712797, "dur": 0,
    "args": {
      "External id": 207560,
      "cbid": 203, "correlation": 207560
    }
  },
  {
    "ph": "f", "id": 207560, "pid": 76337, "tid": -914061504, "ts": 1716454224712797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712797, "dur": 0,
    "args": {
      "External id": 207561,
      "cbid": 205, "correlation": 207561
    }
  },
  {
    "ph": "f", "id": 207561, "pid": 76337, "tid": -914061504, "ts": 1716454224712797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224758383, "dur": 23,
    "args": {
      "External id": 207565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207565,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207565, "pid": 5, "tid": 7, "ts": 1716454224758383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712812, "dur": 13,
    "args": {
      "External id": 207565,
      "cbid": 211, "correlation": 207565
    }
  },
  {
    "ph": "s", "id": 207565, "pid": 76337, "tid": -914061504, "ts": 1716454224712812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224758408, "dur": 121,
    "args": {
      "External id": 207567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207567,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207567, "pid": 5, "tid": 7, "ts": 1716454224758408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712834, "dur": 9,
    "args": {
      "External id": 207567,
      "cbid": 211, "correlation": 207567
    }
  },
  {
    "ph": "s", "id": 207567, "pid": 76337, "tid": -914061504, "ts": 1716454224712834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224758531, "dur": 23,
    "args": {
      "External id": 207569,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207569,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207569, "pid": 5, "tid": 7, "ts": 1716454224758531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712846, "dur": 5,
    "args": {
      "External id": 207569,
      "cbid": 211, "correlation": 207569
    }
  },
  {
    "ph": "s", "id": 207569, "pid": 76337, "tid": -914061504, "ts": 1716454224712846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224758555, "dur": 33,
    "args": {
      "External id": 207575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207575,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207575, "pid": 5, "tid": 7, "ts": 1716454224758555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712874, "dur": 8,
    "args": {
      "External id": 207575,
      "cbid": 211, "correlation": 207575
    }
  },
  {
    "ph": "s", "id": 207575, "pid": 76337, "tid": -914061504, "ts": 1716454224712874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224758589, "dur": 199,
    "args": {
      "External id": 207584,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207584,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207584, "pid": 5, "tid": 7, "ts": 1716454224758589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224712957, "dur": 15,
    "args": {
      "External id": 207584,
      "cbid": 211, "correlation": 207584
    }
  },
  {
    "ph": "s", "id": 207584, "pid": 76337, "tid": -914061504, "ts": 1716454224712957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224758789, "dur": 66,
    "args": {
      "External id": 207606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207606,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207606, "pid": 5, "tid": 7, "ts": 1716454224758789,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713025, "dur": 11,
    "args": {
      "External id": 207606,
      "cbid": 211, "correlation": 207606
    }
  },
  {
    "ph": "s", "id": 207606, "pid": 76337, "tid": -914061504, "ts": 1716454224713025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713116, "dur": 1,
    "args": {
      "External id": 207617,
      "cbid": 251, "correlation": 207617
    }
  },
  {
    "ph": "f", "id": 207617, "pid": 76337, "tid": -914061504, "ts": 1716454224713116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224758856, "dur": 157,
    "args": {
      "External id": 207618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207618,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207618, "pid": 5, "tid": 7, "ts": 1716454224758856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713121, "dur": 13,
    "args": {
      "External id": 207618,
      "cbid": 211, "correlation": 207618
    }
  },
  {
    "ph": "s", "id": 207618, "pid": 76337, "tid": -914061504, "ts": 1716454224713121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713192, "dur": 1,
    "args": {
      "External id": 207629,
      "cbid": 251, "correlation": 207629
    }
  },
  {
    "ph": "f", "id": 207629, "pid": 76337, "tid": -914061504, "ts": 1716454224713192,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224759014, "dur": 149,
    "args": {
      "External id": 207630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207630,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207630, "pid": 5, "tid": 7, "ts": 1716454224759014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713196, "dur": 11,
    "args": {
      "External id": 207630,
      "cbid": 211, "correlation": 207630
    }
  },
  {
    "ph": "s", "id": 207630, "pid": 76337, "tid": -914061504, "ts": 1716454224713196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713261, "dur": 1,
    "args": {
      "External id": 207641,
      "cbid": 251, "correlation": 207641
    }
  },
  {
    "ph": "f", "id": 207641, "pid": 76337, "tid": -914061504, "ts": 1716454224713261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224759164, "dur": 147,
    "args": {
      "External id": 207642,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207642,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207642, "pid": 5, "tid": 7, "ts": 1716454224759164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713265, "dur": 12,
    "args": {
      "External id": 207642,
      "cbid": 211, "correlation": 207642
    }
  },
  {
    "ph": "s", "id": 207642, "pid": 76337, "tid": -914061504, "ts": 1716454224713265,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224759312, "dur": 1965,
    "args": {
      "External id": 207663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207663,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 207663, "pid": 5, "tid": 7, "ts": 1716454224759312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713350, "dur": 13,
    "args": {
      "External id": 207663,
      "cbid": 211, "correlation": 207663
    }
  },
  {
    "ph": "s", "id": 207663, "pid": 76337, "tid": -914061504, "ts": 1716454224713350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713450, "dur": 1,
    "args": {
      "External id": 207681,
      "cbid": 251, "correlation": 207681
    }
  },
  {
    "ph": "f", "id": 207681, "pid": 76337, "tid": -914061504, "ts": 1716454224713450,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224761278, "dur": 147,
    "args": {
      "External id": 207683,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207683,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 207683, "pid": 5, "tid": 7, "ts": 1716454224761278,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713456, "dur": 13,
    "args": {
      "External id": 207683,
      "cbid": 211, "correlation": 207683
    }
  },
  {
    "ph": "s", "id": 207683, "pid": 76337, "tid": -914061504, "ts": 1716454224713456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224761427, "dur": 35,
    "args": {
      "External id": 207691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207691,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207691, "pid": 5, "tid": 7, "ts": 1716454224761427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713527, "dur": 13,
    "args": {
      "External id": 207691,
      "cbid": 211, "correlation": 207691
    }
  },
  {
    "ph": "s", "id": 207691, "pid": 76337, "tid": -914061504, "ts": 1716454224713527,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224761463, "dur": 50,
    "args": {
      "External id": 207699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207699,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207699, "pid": 5, "tid": 7, "ts": 1716454224761463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713566, "dur": 10,
    "args": {
      "External id": 207699,
      "cbid": 211, "correlation": 207699
    }
  },
  {
    "ph": "s", "id": 207699, "pid": 76337, "tid": -914061504, "ts": 1716454224713566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224761515, "dur": 31,
    "args": {
      "External id": 207710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207710, "pid": 5, "tid": 7, "ts": 1716454224761515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713638, "dur": 13,
    "args": {
      "External id": 207710,
      "cbid": 211, "correlation": 207710
    }
  },
  {
    "ph": "s", "id": 207710, "pid": 76337, "tid": -914061504, "ts": 1716454224713638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224761547, "dur": 35,
    "args": {
      "External id": 207732,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207732,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207732, "pid": 5, "tid": 7, "ts": 1716454224761547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713669, "dur": 9,
    "args": {
      "External id": 207732,
      "cbid": 211, "correlation": 207732
    }
  },
  {
    "ph": "s", "id": 207732, "pid": 76337, "tid": -914061504, "ts": 1716454224713669,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713754, "dur": 1,
    "args": {
      "External id": 207743,
      "cbid": 251, "correlation": 207743
    }
  },
  {
    "ph": "f", "id": 207743, "pid": 76337, "tid": -914061504, "ts": 1716454224713754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224761583, "dur": 91,
    "args": {
      "External id": 207744,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207744,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207744, "pid": 5, "tid": 7, "ts": 1716454224761583,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713759, "dur": 13,
    "args": {
      "External id": 207744,
      "cbid": 211, "correlation": 207744
    }
  },
  {
    "ph": "s", "id": 207744, "pid": 76337, "tid": -914061504, "ts": 1716454224713759,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713828, "dur": 1,
    "args": {
      "External id": 207755,
      "cbid": 251, "correlation": 207755
    }
  },
  {
    "ph": "f", "id": 207755, "pid": 76337, "tid": -914061504, "ts": 1716454224713828,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713831, "dur": 0,
    "args": {
      "External id": 207756,
      "cbid": 251, "correlation": 207756
    }
  },
  {
    "ph": "f", "id": 207756, "pid": 76337, "tid": -914061504, "ts": 1716454224713831,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224761676, "dur": 11,
    "args": {
      "External id": 207757,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207757,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 207757, "pid": 5, "tid": 7, "ts": 1716454224761676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713833, "dur": 12,
    "args": {
      "External id": 207757,
      "cbid": 211, "correlation": 207757
    }
  },
  {
    "ph": "s", "id": 207757, "pid": 76337, "tid": -914061504, "ts": 1716454224713833,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224761688, "dur": 5,
    "args": {
      "External id": 207759,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207759,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 207759, "pid": 5, "tid": 7, "ts": 1716454224761688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713847, "dur": 7,
    "args": {
      "External id": 207759,
      "cbid": 211, "correlation": 207759
    }
  },
  {
    "ph": "s", "id": 207759, "pid": 76337, "tid": -914061504, "ts": 1716454224713847,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713906, "dur": 1,
    "args": {
      "External id": 207770,
      "cbid": 251, "correlation": 207770
    }
  },
  {
    "ph": "f", "id": 207770, "pid": 76337, "tid": -914061504, "ts": 1716454224713906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713910, "dur": 0,
    "args": {
      "External id": 207771,
      "cbid": 251, "correlation": 207771
    }
  },
  {
    "ph": "f", "id": 207771, "pid": 76337, "tid": -914061504, "ts": 1716454224713910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224761694, "dur": 7,
    "args": {
      "External id": 207772,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207772,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 207772, "pid": 5, "tid": 7, "ts": 1716454224761694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713911, "dur": 12,
    "args": {
      "External id": 207772,
      "cbid": 211, "correlation": 207772
    }
  },
  {
    "ph": "s", "id": 207772, "pid": 76337, "tid": -914061504, "ts": 1716454224713911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224761703, "dur": 4,
    "args": {
      "External id": 207774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207774,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 207774, "pid": 5, "tid": 7, "ts": 1716454224761703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224713925, "dur": 6,
    "args": {
      "External id": 207774,
      "cbid": 211, "correlation": 207774
    }
  },
  {
    "ph": "s", "id": 207774, "pid": 76337, "tid": -914061504, "ts": 1716454224713925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224761708, "dur": 93,
    "args": {
      "External id": 207795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207795,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 207795, "pid": 5, "tid": 7, "ts": 1716454224761708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714007, "dur": 13,
    "args": {
      "External id": 207795,
      "cbid": 211, "correlation": 207795
    }
  },
  {
    "ph": "s", "id": 207795, "pid": 76337, "tid": -914061504, "ts": 1716454224714007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714106, "dur": 1,
    "args": {
      "External id": 207813,
      "cbid": 251, "correlation": 207813
    }
  },
  {
    "ph": "f", "id": 207813, "pid": 76337, "tid": -914061504, "ts": 1716454224714106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224761802, "dur": 100,
    "args": {
      "External id": 207815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207815,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207815, "pid": 5, "tid": 7, "ts": 1716454224761802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714112, "dur": 14,
    "args": {
      "External id": 207815,
      "cbid": 211, "correlation": 207815
    }
  },
  {
    "ph": "s", "id": 207815, "pid": 76337, "tid": -914061504, "ts": 1716454224714112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224761904, "dur": 20,
    "args": {
      "External id": 207823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207823,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207823, "pid": 5, "tid": 7, "ts": 1716454224761904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714182, "dur": 12,
    "args": {
      "External id": 207823,
      "cbid": 211, "correlation": 207823
    }
  },
  {
    "ph": "s", "id": 207823, "pid": 76337, "tid": -914061504, "ts": 1716454224714182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224761925, "dur": 37,
    "args": {
      "External id": 207831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207831,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207831, "pid": 5, "tid": 7, "ts": 1716454224761925,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714223, "dur": 9,
    "args": {
      "External id": 207831,
      "cbid": 211, "correlation": 207831
    }
  },
  {
    "ph": "s", "id": 207831, "pid": 76337, "tid": -914061504, "ts": 1716454224714223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224761963, "dur": 35,
    "args": {
      "External id": 207853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207853,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207853, "pid": 5, "tid": 7, "ts": 1716454224761963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714275, "dur": 11,
    "args": {
      "External id": 207853,
      "cbid": 211, "correlation": 207853
    }
  },
  {
    "ph": "s", "id": 207853, "pid": 76337, "tid": -914061504, "ts": 1716454224714275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714364, "dur": 1,
    "args": {
      "External id": 207869,
      "cbid": 251, "correlation": 207869
    }
  },
  {
    "ph": "f", "id": 207869, "pid": 76337, "tid": -914061504, "ts": 1716454224714364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714369, "dur": 0,
    "args": {
      "External id": 207871,
      "cbid": 251, "correlation": 207871
    }
  },
  {
    "ph": "f", "id": 207871, "pid": 76337, "tid": -914061504, "ts": 1716454224714369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224762000, "dur": 545,
    "args": {
      "External id": 207872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207872,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 207872, "pid": 5, "tid": 7, "ts": 1716454224762000,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714373, "dur": 13,
    "args": {
      "External id": 207872,
      "cbid": 211, "correlation": 207872
    }
  },
  {
    "ph": "s", "id": 207872, "pid": 76337, "tid": -914061504, "ts": 1716454224714373,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224762546, "dur": 127,
    "args": {
      "External id": 207880,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207880,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207880, "pid": 5, "tid": 7, "ts": 1716454224762546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714438, "dur": 12,
    "args": {
      "External id": 207880,
      "cbid": 211, "correlation": 207880
    }
  },
  {
    "ph": "s", "id": 207880, "pid": 76337, "tid": -914061504, "ts": 1716454224714438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224762674, "dur": 127,
    "args": {
      "External id": 207888,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207888,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207888, "pid": 5, "tid": 7, "ts": 1716454224762674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714469, "dur": 9,
    "args": {
      "External id": 207888,
      "cbid": 211, "correlation": 207888
    }
  },
  {
    "ph": "s", "id": 207888, "pid": 76337, "tid": -914061504, "ts": 1716454224714469,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714547, "dur": 1,
    "args": {
      "External id": 207904,
      "cbid": 251, "correlation": 207904
    }
  },
  {
    "ph": "f", "id": 207904, "pid": 76337, "tid": -914061504, "ts": 1716454224714547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224762802, "dur": 306,
    "args": {
      "External id": 207906,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207906,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207906, "pid": 5, "tid": 7, "ts": 1716454224762802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714553, "dur": 12,
    "args": {
      "External id": 207906,
      "cbid": 211, "correlation": 207906
    }
  },
  {
    "ph": "s", "id": 207906, "pid": 76337, "tid": -914061504, "ts": 1716454224714553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224763110, "dur": 27,
    "args": {
      "External id": 207914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207914,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207914, "pid": 5, "tid": 7, "ts": 1716454224763110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714595, "dur": 9,
    "args": {
      "External id": 207914,
      "cbid": 211, "correlation": 207914
    }
  },
  {
    "ph": "s", "id": 207914, "pid": 76337, "tid": -914061504, "ts": 1716454224714595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224763138, "dur": 82,
    "args": {
      "External id": 207925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207925,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207925, "pid": 5, "tid": 7, "ts": 1716454224763138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714663, "dur": 13,
    "args": {
      "External id": 207925,
      "cbid": 211, "correlation": 207925
    }
  },
  {
    "ph": "s", "id": 207925, "pid": 76337, "tid": -914061504, "ts": 1716454224714663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714728, "dur": 0,
    "args": {
      "External id": 207937,
      "cbid": 317, "correlation": 207937
    }
  },
  {
    "ph": "f", "id": 207937, "pid": 76337, "tid": -914061504, "ts": 1716454224714728,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714729, "dur": 0,
    "args": {
      "External id": 207938,
      "cbid": 203, "correlation": 207938
    }
  },
  {
    "ph": "f", "id": 207938, "pid": 76337, "tid": -914061504, "ts": 1716454224714729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714730, "dur": 0,
    "args": {
      "External id": 207939,
      "cbid": 205, "correlation": 207939
    }
  },
  {
    "ph": "f", "id": 207939, "pid": 76337, "tid": -914061504, "ts": 1716454224714730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224763222, "dur": 22,
    "args": {
      "External id": 207943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207943,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207943, "pid": 5, "tid": 7, "ts": 1716454224763222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714744, "dur": 12,
    "args": {
      "External id": 207943,
      "cbid": 211, "correlation": 207943
    }
  },
  {
    "ph": "s", "id": 207943, "pid": 76337, "tid": -914061504, "ts": 1716454224714744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224763245, "dur": 121,
    "args": {
      "External id": 207945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207945,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 207945, "pid": 5, "tid": 7, "ts": 1716454224763245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714763, "dur": 8,
    "args": {
      "External id": 207945,
      "cbid": 211, "correlation": 207945
    }
  },
  {
    "ph": "s", "id": 207945, "pid": 76337, "tid": -914061504, "ts": 1716454224714763,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224763368, "dur": 22,
    "args": {
      "External id": 207947,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207947,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207947, "pid": 5, "tid": 7, "ts": 1716454224763368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714775, "dur": 5,
    "args": {
      "External id": 207947,
      "cbid": 211, "correlation": 207947
    }
  },
  {
    "ph": "s", "id": 207947, "pid": 76337, "tid": -914061504, "ts": 1716454224714775,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224763391, "dur": 33,
    "args": {
      "External id": 207953,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207953,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207953, "pid": 5, "tid": 7, "ts": 1716454224763391,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714803, "dur": 8,
    "args": {
      "External id": 207953,
      "cbid": 211, "correlation": 207953
    }
  },
  {
    "ph": "s", "id": 207953, "pid": 76337, "tid": -914061504, "ts": 1716454224714803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224763425, "dur": 26,
    "args": {
      "External id": 207961,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207961,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207961, "pid": 5, "tid": 7, "ts": 1716454224763425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714834, "dur": 8,
    "args": {
      "External id": 207961,
      "cbid": 211, "correlation": 207961
    }
  },
  {
    "ph": "s", "id": 207961, "pid": 76337, "tid": -914061504, "ts": 1716454224714834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224763453, "dur": 30,
    "args": {
      "External id": 207981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207981,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 207981, "pid": 5, "tid": 7, "ts": 1716454224763453,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714906, "dur": 12,
    "args": {
      "External id": 207981,
      "cbid": 211, "correlation": 207981
    }
  },
  {
    "ph": "s", "id": 207981, "pid": 76337, "tid": -914061504, "ts": 1716454224714906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224763484, "dur": 5,
    "args": {
      "External id": 207993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207993,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 207993, "pid": 5, "tid": 7, "ts": 1716454224763484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714929, "dur": 6,
    "args": {
      "External id": 207993,
      "cbid": 211, "correlation": 207993
    }
  },
  {
    "ph": "s", "id": 207993, "pid": 76337, "tid": -914061504, "ts": 1716454224714929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224763491, "dur": 31,
    "args": {
      "External id": 207996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 207996,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 207996, "pid": 5, "tid": 7, "ts": 1716454224763491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714947, "dur": 7,
    "args": {
      "External id": 207996,
      "cbid": 211, "correlation": 207996
    }
  },
  {
    "ph": "s", "id": 207996, "pid": 76337, "tid": -914061504, "ts": 1716454224714947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224763523, "dur": 22,
    "args": {
      "External id": 208005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208005,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208005, "pid": 5, "tid": 7, "ts": 1716454224763523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224714994, "dur": 10,
    "args": {
      "External id": 208005,
      "cbid": 211, "correlation": 208005
    }
  },
  {
    "ph": "s", "id": 208005, "pid": 76337, "tid": -914061504, "ts": 1716454224714994,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715047, "dur": 0,
    "args": {
      "External id": 208015,
      "cbid": 317, "correlation": 208015
    }
  },
  {
    "ph": "f", "id": 208015, "pid": 76337, "tid": -914061504, "ts": 1716454224715047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715048, "dur": 0,
    "args": {
      "External id": 208016,
      "cbid": 203, "correlation": 208016
    }
  },
  {
    "ph": "f", "id": 208016, "pid": 76337, "tid": -914061504, "ts": 1716454224715048,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715049, "dur": 0,
    "args": {
      "External id": 208017,
      "cbid": 205, "correlation": 208017
    }
  },
  {
    "ph": "f", "id": 208017, "pid": 76337, "tid": -914061504, "ts": 1716454224715049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224763546, "dur": 22,
    "args": {
      "External id": 208021,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208021,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208021, "pid": 5, "tid": 7, "ts": 1716454224763546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715061, "dur": 13,
    "args": {
      "External id": 208021,
      "cbid": 211, "correlation": 208021
    }
  },
  {
    "ph": "s", "id": 208021, "pid": 76337, "tid": -914061504, "ts": 1716454224715061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224763569, "dur": 45,
    "args": {
      "External id": 208023,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208023,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208023, "pid": 5, "tid": 7, "ts": 1716454224763569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715077, "dur": 5,
    "args": {
      "External id": 208023,
      "cbid": 211, "correlation": 208023
    }
  },
  {
    "ph": "s", "id": 208023, "pid": 76337, "tid": -914061504, "ts": 1716454224715077,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224763615, "dur": 658,
    "args": {
      "External id": 208025,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208025,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208025, "pid": 5, "tid": 7, "ts": 1716454224763615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715088, "dur": 6,
    "args": {
      "External id": 208025,
      "cbid": 211, "correlation": 208025
    }
  },
  {
    "ph": "s", "id": 208025, "pid": 76337, "tid": -914061504, "ts": 1716454224715088,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224764274, "dur": 21,
    "args": {
      "External id": 208027,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208027,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208027, "pid": 5, "tid": 7, "ts": 1716454224764274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715097, "dur": 5,
    "args": {
      "External id": 208027,
      "cbid": 211, "correlation": 208027
    }
  },
  {
    "ph": "s", "id": 208027, "pid": 76337, "tid": -914061504, "ts": 1716454224715097,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224764297, "dur": 32,
    "args": {
      "External id": 208033,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208033,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208033, "pid": 5, "tid": 7, "ts": 1716454224764297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715125, "dur": 9,
    "args": {
      "External id": 208033,
      "cbid": 211, "correlation": 208033
    }
  },
  {
    "ph": "s", "id": 208033, "pid": 76337, "tid": -914061504, "ts": 1716454224715125,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224764331, "dur": 3,
    "args": {
      "External id": 208041,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208041,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 208041, "pid": 5, "tid": 7, "ts": 1716454224764331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715168, "dur": 10,
    "args": {
      "External id": 208041,
      "cbid": 211, "correlation": 208041
    }
  },
  {
    "ph": "s", "id": 208041, "pid": 76337, "tid": -914061504, "ts": 1716454224715168,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715234, "dur": 1,
    "args": {
      "External id": 208057,
      "cbid": 251, "correlation": 208057
    }
  },
  {
    "ph": "f", "id": 208057, "pid": 76337, "tid": -914061504, "ts": 1716454224715234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715240, "dur": 0,
    "args": {
      "External id": 208059,
      "cbid": 251, "correlation": 208059
    }
  },
  {
    "ph": "f", "id": 208059, "pid": 76337, "tid": -914061504, "ts": 1716454224715240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224764335, "dur": 12,
    "args": {
      "External id": 208060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208060,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 208060, "pid": 5, "tid": 7, "ts": 1716454224764335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715242, "dur": 11,
    "args": {
      "External id": 208060,
      "cbid": 211, "correlation": 208060
    }
  },
  {
    "ph": "s", "id": 208060, "pid": 76337, "tid": -914061504, "ts": 1716454224715242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224764349, "dur": 5,
    "args": {
      "External id": 208062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208062,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 208062, "pid": 5, "tid": 7, "ts": 1716454224764349,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715255, "dur": 5,
    "args": {
      "External id": 208062,
      "cbid": 211, "correlation": 208062
    }
  },
  {
    "ph": "s", "id": 208062, "pid": 76337, "tid": -914061504, "ts": 1716454224715255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224764355, "dur": 30,
    "args": {
      "External id": 208072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208072,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208072, "pid": 5, "tid": 7, "ts": 1716454224764355,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715313, "dur": 13,
    "args": {
      "External id": 208072,
      "cbid": 211, "correlation": 208072
    }
  },
  {
    "ph": "s", "id": 208072, "pid": 76337, "tid": -914061504, "ts": 1716454224715313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224764386, "dur": 30,
    "args": {
      "External id": 208092,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208092,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 208092, "pid": 5, "tid": 7, "ts": 1716454224764386,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715381, "dur": 10,
    "args": {
      "External id": 208092,
      "cbid": 211, "correlation": 208092
    }
  },
  {
    "ph": "s", "id": 208092, "pid": 76337, "tid": -914061504, "ts": 1716454224715381,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224764418, "dur": 4,
    "args": {
      "External id": 208104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208104,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 208104, "pid": 5, "tid": 7, "ts": 1716454224764418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715400, "dur": 6,
    "args": {
      "External id": 208104,
      "cbid": 211, "correlation": 208104
    }
  },
  {
    "ph": "s", "id": 208104, "pid": 76337, "tid": -914061504, "ts": 1716454224715400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224764423, "dur": 30,
    "args": {
      "External id": 208107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208107,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208107, "pid": 5, "tid": 7, "ts": 1716454224764423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715421, "dur": 6,
    "args": {
      "External id": 208107,
      "cbid": 211, "correlation": 208107
    }
  },
  {
    "ph": "s", "id": 208107, "pid": 76337, "tid": -914061504, "ts": 1716454224715421,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224764455, "dur": 20,
    "args": {
      "External id": 208116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208116,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208116, "pid": 5, "tid": 7, "ts": 1716454224764455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715460, "dur": 10,
    "args": {
      "External id": 208116,
      "cbid": 211, "correlation": 208116
    }
  },
  {
    "ph": "s", "id": 208116, "pid": 76337, "tid": -914061504, "ts": 1716454224715460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715524, "dur": 0,
    "args": {
      "External id": 208126,
      "cbid": 317, "correlation": 208126
    }
  },
  {
    "ph": "f", "id": 208126, "pid": 76337, "tid": -914061504, "ts": 1716454224715524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715525, "dur": 0,
    "args": {
      "External id": 208127,
      "cbid": 203, "correlation": 208127
    }
  },
  {
    "ph": "f", "id": 208127, "pid": 76337, "tid": -914061504, "ts": 1716454224715525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715525, "dur": 0,
    "args": {
      "External id": 208128,
      "cbid": 205, "correlation": 208128
    }
  },
  {
    "ph": "f", "id": 208128, "pid": 76337, "tid": -914061504, "ts": 1716454224715525,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224764476, "dur": 22,
    "args": {
      "External id": 208132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208132,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208132, "pid": 5, "tid": 7, "ts": 1716454224764476,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715539, "dur": 12,
    "args": {
      "External id": 208132,
      "cbid": 211, "correlation": 208132
    }
  },
  {
    "ph": "s", "id": 208132, "pid": 76337, "tid": -914061504, "ts": 1716454224715539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224764500, "dur": 44,
    "args": {
      "External id": 208134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208134, "pid": 5, "tid": 7, "ts": 1716454224764500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715553, "dur": 5,
    "args": {
      "External id": 208134,
      "cbid": 211, "correlation": 208134
    }
  },
  {
    "ph": "s", "id": 208134, "pid": 76337, "tid": -914061504, "ts": 1716454224715553,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224764546, "dur": 650,
    "args": {
      "External id": 208136,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208136,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208136, "pid": 5, "tid": 7, "ts": 1716454224764546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715564, "dur": 7,
    "args": {
      "External id": 208136,
      "cbid": 211, "correlation": 208136
    }
  },
  {
    "ph": "s", "id": 208136, "pid": 76337, "tid": -914061504, "ts": 1716454224715564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224765197, "dur": 22,
    "args": {
      "External id": 208138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208138,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208138, "pid": 5, "tid": 7, "ts": 1716454224765197,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715574, "dur": 5,
    "args": {
      "External id": 208138,
      "cbid": 211, "correlation": 208138
    }
  },
  {
    "ph": "s", "id": 208138, "pid": 76337, "tid": -914061504, "ts": 1716454224715574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224765220, "dur": 33,
    "args": {
      "External id": 208144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208144, "pid": 5, "tid": 7, "ts": 1716454224765220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715602, "dur": 8,
    "args": {
      "External id": 208144,
      "cbid": 211, "correlation": 208144
    }
  },
  {
    "ph": "s", "id": 208144, "pid": 76337, "tid": -914061504, "ts": 1716454224715602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224765255, "dur": 27,
    "args": {
      "External id": 208152,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208152,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208152, "pid": 5, "tid": 7, "ts": 1716454224765255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715635, "dur": 8,
    "args": {
      "External id": 208152,
      "cbid": 211, "correlation": 208152
    }
  },
  {
    "ph": "s", "id": 208152, "pid": 76337, "tid": -914061504, "ts": 1716454224715635,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224765283, "dur": 20,
    "args": {
      "External id": 208160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208160,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208160, "pid": 5, "tid": 7, "ts": 1716454224765283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715664, "dur": 8,
    "args": {
      "External id": 208160,
      "cbid": 211, "correlation": 208160
    }
  },
  {
    "ph": "s", "id": 208160, "pid": 76337, "tid": -914061504, "ts": 1716454224715664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224765304, "dur": 30,
    "args": {
      "External id": 208180,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208180,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 208180, "pid": 5, "tid": 7, "ts": 1716454224765304,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715743, "dur": 12,
    "args": {
      "External id": 208180,
      "cbid": 211, "correlation": 208180
    }
  },
  {
    "ph": "s", "id": 208180, "pid": 76337, "tid": -914061504, "ts": 1716454224715743,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224765335, "dur": 4,
    "args": {
      "External id": 208192,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208192,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 208192, "pid": 5, "tid": 7, "ts": 1716454224765335,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715765, "dur": 7,
    "args": {
      "External id": 208192,
      "cbid": 211, "correlation": 208192
    }
  },
  {
    "ph": "s", "id": 208192, "pid": 76337, "tid": -914061504, "ts": 1716454224715765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224765340, "dur": 30,
    "args": {
      "External id": 208195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208195,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208195, "pid": 5, "tid": 7, "ts": 1716454224765340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715783, "dur": 6,
    "args": {
      "External id": 208195,
      "cbid": 211, "correlation": 208195
    }
  },
  {
    "ph": "s", "id": 208195, "pid": 76337, "tid": -914061504, "ts": 1716454224715783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715841, "dur": 0,
    "args": {
      "External id": 208206,
      "cbid": 317, "correlation": 208206
    }
  },
  {
    "ph": "f", "id": 208206, "pid": 76337, "tid": -914061504, "ts": 1716454224715841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715842, "dur": 0,
    "args": {
      "External id": 208207,
      "cbid": 203, "correlation": 208207
    }
  },
  {
    "ph": "f", "id": 208207, "pid": 76337, "tid": -914061504, "ts": 1716454224715842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715843, "dur": 0,
    "args": {
      "External id": 208208,
      "cbid": 205, "correlation": 208208
    }
  },
  {
    "ph": "f", "id": 208208, "pid": 76337, "tid": -914061504, "ts": 1716454224715843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224765372, "dur": 22,
    "args": {
      "External id": 208212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208212,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208212, "pid": 5, "tid": 7, "ts": 1716454224765372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715855, "dur": 12,
    "args": {
      "External id": 208212,
      "cbid": 211, "correlation": 208212
    }
  },
  {
    "ph": "s", "id": 208212, "pid": 76337, "tid": -914061504, "ts": 1716454224715855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224765395, "dur": 117,
    "args": {
      "External id": 208214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208214,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208214, "pid": 5, "tid": 7, "ts": 1716454224765395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715874, "dur": 6,
    "args": {
      "External id": 208214,
      "cbid": 211, "correlation": 208214
    }
  },
  {
    "ph": "s", "id": 208214, "pid": 76337, "tid": -914061504, "ts": 1716454224715874,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224765513, "dur": 21,
    "args": {
      "External id": 208216,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208216,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208216, "pid": 5, "tid": 7, "ts": 1716454224765513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715883, "dur": 5,
    "args": {
      "External id": 208216,
      "cbid": 211, "correlation": 208216
    }
  },
  {
    "ph": "s", "id": 208216, "pid": 76337, "tid": -914061504, "ts": 1716454224715883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224765535, "dur": 32,
    "args": {
      "External id": 208222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208222,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208222, "pid": 5, "tid": 7, "ts": 1716454224765535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224715910, "dur": 9,
    "args": {
      "External id": 208222,
      "cbid": 211, "correlation": 208222
    }
  },
  {
    "ph": "s", "id": 208222, "pid": 76337, "tid": -914061504, "ts": 1716454224715910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224765569, "dur": 199,
    "args": {
      "External id": 208231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208231,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208231, "pid": 5, "tid": 7, "ts": 1716454224765569,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716000, "dur": 14,
    "args": {
      "External id": 208231,
      "cbid": 211, "correlation": 208231
    }
  },
  {
    "ph": "s", "id": 208231, "pid": 76337, "tid": -914061504, "ts": 1716454224716000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224765769, "dur": 66,
    "args": {
      "External id": 208253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208253,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208253, "pid": 5, "tid": 7, "ts": 1716454224765769,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716058, "dur": 10,
    "args": {
      "External id": 208253,
      "cbid": 211, "correlation": 208253
    }
  },
  {
    "ph": "s", "id": 208253, "pid": 76337, "tid": -914061504, "ts": 1716454224716058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716146, "dur": 1,
    "args": {
      "External id": 208264,
      "cbid": 251, "correlation": 208264
    }
  },
  {
    "ph": "f", "id": 208264, "pid": 76337, "tid": -914061504, "ts": 1716454224716146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224765836, "dur": 154,
    "args": {
      "External id": 208265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208265,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208265, "pid": 5, "tid": 7, "ts": 1716454224765836,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716152, "dur": 13,
    "args": {
      "External id": 208265,
      "cbid": 211, "correlation": 208265
    }
  },
  {
    "ph": "s", "id": 208265, "pid": 76337, "tid": -914061504, "ts": 1716454224716152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716222, "dur": 1,
    "args": {
      "External id": 208276,
      "cbid": 251, "correlation": 208276
    }
  },
  {
    "ph": "f", "id": 208276, "pid": 76337, "tid": -914061504, "ts": 1716454224716222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224765992, "dur": 149,
    "args": {
      "External id": 208277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208277,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208277, "pid": 5, "tid": 7, "ts": 1716454224765992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716226, "dur": 11,
    "args": {
      "External id": 208277,
      "cbid": 211, "correlation": 208277
    }
  },
  {
    "ph": "s", "id": 208277, "pid": 76337, "tid": -914061504, "ts": 1716454224716226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716292, "dur": 1,
    "args": {
      "External id": 208288,
      "cbid": 251, "correlation": 208288
    }
  },
  {
    "ph": "f", "id": 208288, "pid": 76337, "tid": -914061504, "ts": 1716454224716292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224766142, "dur": 145,
    "args": {
      "External id": 208289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208289,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208289, "pid": 5, "tid": 7, "ts": 1716454224766142,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716295, "dur": 10,
    "args": {
      "External id": 208289,
      "cbid": 211, "correlation": 208289
    }
  },
  {
    "ph": "s", "id": 208289, "pid": 76337, "tid": -914061504, "ts": 1716454224716295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224766289, "dur": 1972,
    "args": {
      "External id": 208310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208310,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 208310, "pid": 5, "tid": 7, "ts": 1716454224766289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716377, "dur": 13,
    "args": {
      "External id": 208310,
      "cbid": 211, "correlation": 208310
    }
  },
  {
    "ph": "s", "id": 208310, "pid": 76337, "tid": -914061504, "ts": 1716454224716377,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716474, "dur": 1,
    "args": {
      "External id": 208328,
      "cbid": 251, "correlation": 208328
    }
  },
  {
    "ph": "f", "id": 208328, "pid": 76337, "tid": -914061504, "ts": 1716454224716474,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224768262, "dur": 148,
    "args": {
      "External id": 208330,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208330,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 208330, "pid": 5, "tid": 7, "ts": 1716454224768262,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716480, "dur": 13,
    "args": {
      "External id": 208330,
      "cbid": 211, "correlation": 208330
    }
  },
  {
    "ph": "s", "id": 208330, "pid": 76337, "tid": -914061504, "ts": 1716454224716480,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224768412, "dur": 35,
    "args": {
      "External id": 208338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208338,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208338, "pid": 5, "tid": 7, "ts": 1716454224768412,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716550, "dur": 13,
    "args": {
      "External id": 208338,
      "cbid": 211, "correlation": 208338
    }
  },
  {
    "ph": "s", "id": 208338, "pid": 76337, "tid": -914061504, "ts": 1716454224716550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224768448, "dur": 50,
    "args": {
      "External id": 208346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208346,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208346, "pid": 5, "tid": 7, "ts": 1716454224768448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716591, "dur": 9,
    "args": {
      "External id": 208346,
      "cbid": 211, "correlation": 208346
    }
  },
  {
    "ph": "s", "id": 208346, "pid": 76337, "tid": -914061504, "ts": 1716454224716591,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224768499, "dur": 30,
    "args": {
      "External id": 208357,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208357,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208357, "pid": 5, "tid": 7, "ts": 1716454224768499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716663, "dur": 13,
    "args": {
      "External id": 208357,
      "cbid": 211, "correlation": 208357
    }
  },
  {
    "ph": "s", "id": 208357, "pid": 76337, "tid": -914061504, "ts": 1716454224716663,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224768530, "dur": 34,
    "args": {
      "External id": 208379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208379,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208379, "pid": 5, "tid": 7, "ts": 1716454224768530,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716693, "dur": 7,
    "args": {
      "External id": 208379,
      "cbid": 211, "correlation": 208379
    }
  },
  {
    "ph": "s", "id": 208379, "pid": 76337, "tid": -914061504, "ts": 1716454224716693,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716777, "dur": 1,
    "args": {
      "External id": 208390,
      "cbid": 251, "correlation": 208390
    }
  },
  {
    "ph": "f", "id": 208390, "pid": 76337, "tid": -914061504, "ts": 1716454224716777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224768566, "dur": 89,
    "args": {
      "External id": 208391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208391,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208391, "pid": 5, "tid": 7, "ts": 1716454224768566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716782, "dur": 13,
    "args": {
      "External id": 208391,
      "cbid": 211, "correlation": 208391
    }
  },
  {
    "ph": "s", "id": 208391, "pid": 76337, "tid": -914061504, "ts": 1716454224716782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716851, "dur": 1,
    "args": {
      "External id": 208402,
      "cbid": 251, "correlation": 208402
    }
  },
  {
    "ph": "f", "id": 208402, "pid": 76337, "tid": -914061504, "ts": 1716454224716851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716854, "dur": 0,
    "args": {
      "External id": 208403,
      "cbid": 251, "correlation": 208403
    }
  },
  {
    "ph": "f", "id": 208403, "pid": 76337, "tid": -914061504, "ts": 1716454224716854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224768656, "dur": 12,
    "args": {
      "External id": 208404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208404,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 208404, "pid": 5, "tid": 7, "ts": 1716454224768656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716856, "dur": 13,
    "args": {
      "External id": 208404,
      "cbid": 211, "correlation": 208404
    }
  },
  {
    "ph": "s", "id": 208404, "pid": 76337, "tid": -914061504, "ts": 1716454224716856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224768669, "dur": 5,
    "args": {
      "External id": 208406,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208406,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 208406, "pid": 5, "tid": 7, "ts": 1716454224768669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716870, "dur": 6,
    "args": {
      "External id": 208406,
      "cbid": 211, "correlation": 208406
    }
  },
  {
    "ph": "s", "id": 208406, "pid": 76337, "tid": -914061504, "ts": 1716454224716870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716927, "dur": 1,
    "args": {
      "External id": 208417,
      "cbid": 251, "correlation": 208417
    }
  },
  {
    "ph": "f", "id": 208417, "pid": 76337, "tid": -914061504, "ts": 1716454224716927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716931, "dur": 0,
    "args": {
      "External id": 208418,
      "cbid": 251, "correlation": 208418
    }
  },
  {
    "ph": "f", "id": 208418, "pid": 76337, "tid": -914061504, "ts": 1716454224716931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224768675, "dur": 7,
    "args": {
      "External id": 208419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208419,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 208419, "pid": 5, "tid": 7, "ts": 1716454224768675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716932, "dur": 12,
    "args": {
      "External id": 208419,
      "cbid": 211, "correlation": 208419
    }
  },
  {
    "ph": "s", "id": 208419, "pid": 76337, "tid": -914061504, "ts": 1716454224716932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224768684, "dur": 3,
    "args": {
      "External id": 208421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208421,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 208421, "pid": 5, "tid": 7, "ts": 1716454224768684,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224716945, "dur": 5,
    "args": {
      "External id": 208421,
      "cbid": 211, "correlation": 208421
    }
  },
  {
    "ph": "s", "id": 208421, "pid": 76337, "tid": -914061504, "ts": 1716454224716945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224768688, "dur": 94,
    "args": {
      "External id": 208442,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208442,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 208442, "pid": 5, "tid": 7, "ts": 1716454224768688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717030, "dur": 13,
    "args": {
      "External id": 208442,
      "cbid": 211, "correlation": 208442
    }
  },
  {
    "ph": "s", "id": 208442, "pid": 76337, "tid": -914061504, "ts": 1716454224717030,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717128, "dur": 1,
    "args": {
      "External id": 208460,
      "cbid": 251, "correlation": 208460
    }
  },
  {
    "ph": "f", "id": 208460, "pid": 76337, "tid": -914061504, "ts": 1716454224717128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224768783, "dur": 99,
    "args": {
      "External id": 208462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208462,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208462, "pid": 5, "tid": 7, "ts": 1716454224768783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717133, "dur": 13,
    "args": {
      "External id": 208462,
      "cbid": 211, "correlation": 208462
    }
  },
  {
    "ph": "s", "id": 208462, "pid": 76337, "tid": -914061504, "ts": 1716454224717133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224768884, "dur": 19,
    "args": {
      "External id": 208470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208470,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208470, "pid": 5, "tid": 7, "ts": 1716454224768884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717202, "dur": 13,
    "args": {
      "External id": 208470,
      "cbid": 211, "correlation": 208470
    }
  },
  {
    "ph": "s", "id": 208470, "pid": 76337, "tid": -914061504, "ts": 1716454224717202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224768904, "dur": 37,
    "args": {
      "External id": 208478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208478,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208478, "pid": 5, "tid": 7, "ts": 1716454224768904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717244, "dur": 9,
    "args": {
      "External id": 208478,
      "cbid": 211, "correlation": 208478
    }
  },
  {
    "ph": "s", "id": 208478, "pid": 76337, "tid": -914061504, "ts": 1716454224717244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224768942, "dur": 35,
    "args": {
      "External id": 208500,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208500,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208500, "pid": 5, "tid": 7, "ts": 1716454224768942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717295, "dur": 10,
    "args": {
      "External id": 208500,
      "cbid": 211, "correlation": 208500
    }
  },
  {
    "ph": "s", "id": 208500, "pid": 76337, "tid": -914061504, "ts": 1716454224717295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717384, "dur": 1,
    "args": {
      "External id": 208516,
      "cbid": 251, "correlation": 208516
    }
  },
  {
    "ph": "f", "id": 208516, "pid": 76337, "tid": -914061504, "ts": 1716454224717384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717389, "dur": 0,
    "args": {
      "External id": 208518,
      "cbid": 251, "correlation": 208518
    }
  },
  {
    "ph": "f", "id": 208518, "pid": 76337, "tid": -914061504, "ts": 1716454224717389,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224768979, "dur": 543,
    "args": {
      "External id": 208519,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208519,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 208519, "pid": 5, "tid": 7, "ts": 1716454224768979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717392, "dur": 12,
    "args": {
      "External id": 208519,
      "cbid": 211, "correlation": 208519
    }
  },
  {
    "ph": "s", "id": 208519, "pid": 76337, "tid": -914061504, "ts": 1716454224717392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224769523, "dur": 124,
    "args": {
      "External id": 208527,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208527,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208527, "pid": 5, "tid": 7, "ts": 1716454224769523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717456, "dur": 13,
    "args": {
      "External id": 208527,
      "cbid": 211, "correlation": 208527
    }
  },
  {
    "ph": "s", "id": 208527, "pid": 76337, "tid": -914061504, "ts": 1716454224717456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224769648, "dur": 126,
    "args": {
      "External id": 208535,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208535,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208535, "pid": 5, "tid": 7, "ts": 1716454224769648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717487, "dur": 8,
    "args": {
      "External id": 208535,
      "cbid": 211, "correlation": 208535
    }
  },
  {
    "ph": "s", "id": 208535, "pid": 76337, "tid": -914061504, "ts": 1716454224717487,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717563, "dur": 1,
    "args": {
      "External id": 208551,
      "cbid": 251, "correlation": 208551
    }
  },
  {
    "ph": "f", "id": 208551, "pid": 76337, "tid": -914061504, "ts": 1716454224717563,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224769775, "dur": 307,
    "args": {
      "External id": 208553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208553,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208553, "pid": 5, "tid": 7, "ts": 1716454224769775,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717568, "dur": 12,
    "args": {
      "External id": 208553,
      "cbid": 211, "correlation": 208553
    }
  },
  {
    "ph": "s", "id": 208553, "pid": 76337, "tid": -914061504, "ts": 1716454224717568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224770084, "dur": 27,
    "args": {
      "External id": 208561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208561,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208561, "pid": 5, "tid": 7, "ts": 1716454224770084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717610, "dur": 10,
    "args": {
      "External id": 208561,
      "cbid": 211, "correlation": 208561
    }
  },
  {
    "ph": "s", "id": 208561, "pid": 76337, "tid": -914061504, "ts": 1716454224717610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224770112, "dur": 82,
    "args": {
      "External id": 208572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208572,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208572, "pid": 5, "tid": 7, "ts": 1716454224770112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717679, "dur": 12,
    "args": {
      "External id": 208572,
      "cbid": 211, "correlation": 208572
    }
  },
  {
    "ph": "s", "id": 208572, "pid": 76337, "tid": -914061504, "ts": 1716454224717679,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717741, "dur": 0,
    "args": {
      "External id": 208584,
      "cbid": 317, "correlation": 208584
    }
  },
  {
    "ph": "f", "id": 208584, "pid": 76337, "tid": -914061504, "ts": 1716454224717741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717742, "dur": 0,
    "args": {
      "External id": 208585,
      "cbid": 203, "correlation": 208585
    }
  },
  {
    "ph": "f", "id": 208585, "pid": 76337, "tid": -914061504, "ts": 1716454224717742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717742, "dur": 0,
    "args": {
      "External id": 208586,
      "cbid": 205, "correlation": 208586
    }
  },
  {
    "ph": "f", "id": 208586, "pid": 76337, "tid": -914061504, "ts": 1716454224717742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770196, "dur": 23,
    "args": {
      "External id": 208590,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208590,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208590, "pid": 5, "tid": 7, "ts": 1716454224770196,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717758, "dur": 12,
    "args": {
      "External id": 208590,
      "cbid": 211, "correlation": 208590
    }
  },
  {
    "ph": "s", "id": 208590, "pid": 76337, "tid": -914061504, "ts": 1716454224717758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224770219, "dur": 122,
    "args": {
      "External id": 208592,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208592,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208592, "pid": 5, "tid": 7, "ts": 1716454224770219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717777, "dur": 6,
    "args": {
      "External id": 208592,
      "cbid": 211, "correlation": 208592
    }
  },
  {
    "ph": "s", "id": 208592, "pid": 76337, "tid": -914061504, "ts": 1716454224717777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770343, "dur": 21,
    "args": {
      "External id": 208594,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208594,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208594, "pid": 5, "tid": 7, "ts": 1716454224770343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717787, "dur": 5,
    "args": {
      "External id": 208594,
      "cbid": 211, "correlation": 208594
    }
  },
  {
    "ph": "s", "id": 208594, "pid": 76337, "tid": -914061504, "ts": 1716454224717787,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224770365, "dur": 33,
    "args": {
      "External id": 208600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208600,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208600, "pid": 5, "tid": 7, "ts": 1716454224770365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717815, "dur": 8,
    "args": {
      "External id": 208600,
      "cbid": 211, "correlation": 208600
    }
  },
  {
    "ph": "s", "id": 208600, "pid": 76337, "tid": -914061504, "ts": 1716454224717815,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224770399, "dur": 27,
    "args": {
      "External id": 208608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208608,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208608, "pid": 5, "tid": 7, "ts": 1716454224770399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717846, "dur": 8,
    "args": {
      "External id": 208608,
      "cbid": 211, "correlation": 208608
    }
  },
  {
    "ph": "s", "id": 208608, "pid": 76337, "tid": -914061504, "ts": 1716454224717846,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717917, "dur": 0,
    "args": {
      "External id": 208618,
      "cbid": 317, "correlation": 208618
    }
  },
  {
    "ph": "f", "id": 208618, "pid": 76337, "tid": -914061504, "ts": 1716454224717917,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717918, "dur": 0,
    "args": {
      "External id": 208619,
      "cbid": 203, "correlation": 208619
    }
  },
  {
    "ph": "f", "id": 208619, "pid": 76337, "tid": -914061504, "ts": 1716454224717918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717918, "dur": 0,
    "args": {
      "External id": 208620,
      "cbid": 205, "correlation": 208620
    }
  },
  {
    "ph": "f", "id": 208620, "pid": 76337, "tid": -914061504, "ts": 1716454224717918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770427, "dur": 23,
    "args": {
      "External id": 208624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208624,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208624, "pid": 5, "tid": 7, "ts": 1716454224770427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717933, "dur": 12,
    "args": {
      "External id": 208624,
      "cbid": 211, "correlation": 208624
    }
  },
  {
    "ph": "s", "id": 208624, "pid": 76337, "tid": -914061504, "ts": 1716454224717933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770452, "dur": 46,
    "args": {
      "External id": 208626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208626,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208626, "pid": 5, "tid": 7, "ts": 1716454224770452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717947, "dur": 5,
    "args": {
      "External id": 208626,
      "cbid": 211, "correlation": 208626
    }
  },
  {
    "ph": "s", "id": 208626, "pid": 76337, "tid": -914061504, "ts": 1716454224717947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224770498, "dur": 236,
    "args": {
      "External id": 208628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208628,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [5, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 208628, "pid": 5, "tid": 7, "ts": 1716454224770498,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717959, "dur": 6,
    "args": {
      "External id": 208628,
      "cbid": 211, "correlation": 208628
    }
  },
  {
    "ph": "s", "id": 208628, "pid": 76337, "tid": -914061504, "ts": 1716454224717959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770736, "dur": 7,
    "args": {
      "External id": 208630,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208630,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208630, "pid": 5, "tid": 7, "ts": 1716454224770736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224717969, "dur": 14,
    "args": {
      "External id": 208630,
      "cbid": 211, "correlation": 208630
    }
  },
  {
    "ph": "s", "id": 208630, "pid": 76337, "tid": -914061504, "ts": 1716454224717969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224770744, "dur": 9,
    "args": {
      "External id": 208636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208636, "pid": 5, "tid": 7, "ts": 1716454224770744,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718006, "dur": 9,
    "args": {
      "External id": 208636,
      "cbid": 211, "correlation": 208636
    }
  },
  {
    "ph": "s", "id": 208636, "pid": 76337, "tid": -914061504, "ts": 1716454224718006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770754, "dur": 12,
    "args": {
      "External id": 208656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208656,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 208656, "pid": 5, "tid": 7, "ts": 1716454224770754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718098, "dur": 13,
    "args": {
      "External id": 208656,
      "cbid": 211, "correlation": 208656
    }
  },
  {
    "ph": "s", "id": 208656, "pid": 76337, "tid": -914061504, "ts": 1716454224718098,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224770768, "dur": 4,
    "args": {
      "External id": 208668,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208668,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 208668, "pid": 5, "tid": 7, "ts": 1716454224770768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718121, "dur": 6,
    "args": {
      "External id": 208668,
      "cbid": 211, "correlation": 208668
    }
  },
  {
    "ph": "s", "id": 208668, "pid": 76337, "tid": -914061504, "ts": 1716454224718121,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224770773, "dur": 13,
    "args": {
      "External id": 208671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208671,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208671, "pid": 5, "tid": 7, "ts": 1716454224770773,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718138, "dur": 6,
    "args": {
      "External id": 208671,
      "cbid": 211, "correlation": 208671
    }
  },
  {
    "ph": "s", "id": 208671, "pid": 76337, "tid": -914061504, "ts": 1716454224718138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224770787, "dur": 7,
    "args": {
      "External id": 208680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208680,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208680, "pid": 5, "tid": 7, "ts": 1716454224770787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718178, "dur": 10,
    "args": {
      "External id": 208680,
      "cbid": 211, "correlation": 208680
    }
  },
  {
    "ph": "s", "id": 208680, "pid": 76337, "tid": -914061504, "ts": 1716454224718178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718230, "dur": 0,
    "args": {
      "External id": 208690,
      "cbid": 317, "correlation": 208690
    }
  },
  {
    "ph": "f", "id": 208690, "pid": 76337, "tid": -914061504, "ts": 1716454224718230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718231, "dur": 0,
    "args": {
      "External id": 208691,
      "cbid": 203, "correlation": 208691
    }
  },
  {
    "ph": "f", "id": 208691, "pid": 76337, "tid": -914061504, "ts": 1716454224718231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718232, "dur": 0,
    "args": {
      "External id": 208692,
      "cbid": 205, "correlation": 208692
    }
  },
  {
    "ph": "f", "id": 208692, "pid": 76337, "tid": -914061504, "ts": 1716454224718232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770795, "dur": 6,
    "args": {
      "External id": 208696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208696,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208696, "pid": 5, "tid": 7, "ts": 1716454224770795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718248, "dur": 12,
    "args": {
      "External id": 208696,
      "cbid": 211, "correlation": 208696
    }
  },
  {
    "ph": "s", "id": 208696, "pid": 76337, "tid": -914061504, "ts": 1716454224718248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224770802, "dur": 84,
    "args": {
      "External id": 208698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208698,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 20, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208698, "pid": 5, "tid": 7, "ts": 1716454224770802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718263, "dur": 5,
    "args": {
      "External id": 208698,
      "cbid": 211, "correlation": 208698
    }
  },
  {
    "ph": "s", "id": 208698, "pid": 76337, "tid": -914061504, "ts": 1716454224718263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224770889, "dur": 1,
    "args": {
      "External id": 208700,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 208700,
      "bytes": 960, "memory bandwidth (GB/s)": 0.6
    }
  },
  {
    "ph": "f", "id": 208700, "pid": 5, "tid": 7, "ts": 1716454224770889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718275, "dur": 8,
    "args": {
      "External id": 208700,
      "cbid": 51, "correlation": 208700
    }
  },
  {
    "ph": "s", "id": 208700, "pid": 76337, "tid": -914061504, "ts": 1716454224718275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224770892, "dur": 545,
    "args": {
      "External id": 208701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208701,
      "registers per thread": 236,
      "shared memory": 18432,
      "blocks per SM": 30,
      "warps per SM": 120,
      "grid": [10, 12, 20],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208701, "pid": 5, "tid": 7, "ts": 1716454224770892,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718285, "dur": 8,
    "args": {
      "External id": 208701,
      "cbid": 211, "correlation": 208701
    }
  },
  {
    "ph": "s", "id": 208701, "pid": 76337, "tid": -914061504, "ts": 1716454224718285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224771439, "dur": 12,
    "args": {
      "External id": 208703,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208703,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208703, "pid": 5, "tid": 7, "ts": 1716454224771439,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718298, "dur": 5,
    "args": {
      "External id": 208703,
      "cbid": 211, "correlation": 208703
    }
  },
  {
    "ph": "s", "id": 208703, "pid": 76337, "tid": -914061504, "ts": 1716454224718298,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224771452, "dur": 15,
    "args": {
      "External id": 208709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208709,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208709, "pid": 5, "tid": 7, "ts": 1716454224771452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718325, "dur": 8,
    "args": {
      "External id": 208709,
      "cbid": 211, "correlation": 208709
    }
  },
  {
    "ph": "s", "id": 208709, "pid": 76337, "tid": -914061504, "ts": 1716454224718325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224771468, "dur": 3,
    "args": {
      "External id": 208717,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208717,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 208717, "pid": 5, "tid": 7, "ts": 1716454224771468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718369, "dur": 9,
    "args": {
      "External id": 208717,
      "cbid": 211, "correlation": 208717
    }
  },
  {
    "ph": "s", "id": 208717, "pid": 76337, "tid": -914061504, "ts": 1716454224718369,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718432, "dur": 1,
    "args": {
      "External id": 208733,
      "cbid": 251, "correlation": 208733
    }
  },
  {
    "ph": "f", "id": 208733, "pid": 76337, "tid": -914061504, "ts": 1716454224718432,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718437, "dur": 0,
    "args": {
      "External id": 208735,
      "cbid": 251, "correlation": 208735
    }
  },
  {
    "ph": "f", "id": 208735, "pid": 76337, "tid": -914061504, "ts": 1716454224718437,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224771473, "dur": 13,
    "args": {
      "External id": 208736,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208736,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208736, "pid": 5, "tid": 7, "ts": 1716454224771473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718439, "dur": 12,
    "args": {
      "External id": 208736,
      "cbid": 211, "correlation": 208736
    }
  },
  {
    "ph": "s", "id": 208736, "pid": 76337, "tid": -914061504, "ts": 1716454224718439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224771487, "dur": 5,
    "args": {
      "External id": 208738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208738,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208738, "pid": 5, "tid": 7, "ts": 1716454224771487,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718452, "dur": 5,
    "args": {
      "External id": 208738,
      "cbid": 211, "correlation": 208738
    }
  },
  {
    "ph": "s", "id": 208738, "pid": 76337, "tid": -914061504, "ts": 1716454224718452,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224771494, "dur": 17,
    "args": {
      "External id": 208748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208748,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208748, "pid": 5, "tid": 7, "ts": 1716454224771494,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718510, "dur": 11,
    "args": {
      "External id": 208748,
      "cbid": 211, "correlation": 208748
    }
  },
  {
    "ph": "s", "id": 208748, "pid": 76337, "tid": -914061504, "ts": 1716454224718510,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224771512, "dur": 18,
    "args": {
      "External id": 208768,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208768,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 208768, "pid": 5, "tid": 7, "ts": 1716454224771512,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718576, "dur": 11,
    "args": {
      "External id": 208768,
      "cbid": 211, "correlation": 208768
    }
  },
  {
    "ph": "s", "id": 208768, "pid": 76337, "tid": -914061504, "ts": 1716454224718576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224771531, "dur": 5,
    "args": {
      "External id": 208780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208780,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 208780, "pid": 5, "tid": 7, "ts": 1716454224771531,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718598, "dur": 6,
    "args": {
      "External id": 208780,
      "cbid": 211, "correlation": 208780
    }
  },
  {
    "ph": "s", "id": 208780, "pid": 76337, "tid": -914061504, "ts": 1716454224718598,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224771537, "dur": 17,
    "args": {
      "External id": 208783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208783,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208783, "pid": 5, "tid": 7, "ts": 1716454224771537,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718617, "dur": 6,
    "args": {
      "External id": 208783,
      "cbid": 211, "correlation": 208783
    }
  },
  {
    "ph": "s", "id": 208783, "pid": 76337, "tid": -914061504, "ts": 1716454224718617,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224771555, "dur": 11,
    "args": {
      "External id": 208792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208792,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208792, "pid": 5, "tid": 7, "ts": 1716454224771555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718657, "dur": 9,
    "args": {
      "External id": 208792,
      "cbid": 211, "correlation": 208792
    }
  },
  {
    "ph": "s", "id": 208792, "pid": 76337, "tid": -914061504, "ts": 1716454224718657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718719, "dur": 0,
    "args": {
      "External id": 208802,
      "cbid": 317, "correlation": 208802
    }
  },
  {
    "ph": "f", "id": 208802, "pid": 76337, "tid": -914061504, "ts": 1716454224718719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718720, "dur": 0,
    "args": {
      "External id": 208803,
      "cbid": 203, "correlation": 208803
    }
  },
  {
    "ph": "f", "id": 208803, "pid": 76337, "tid": -914061504, "ts": 1716454224718720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718720, "dur": 0,
    "args": {
      "External id": 208804,
      "cbid": 205, "correlation": 208804
    }
  },
  {
    "ph": "f", "id": 208804, "pid": 76337, "tid": -914061504, "ts": 1716454224718720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224771567, "dur": 11,
    "args": {
      "External id": 208808,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208808,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208808, "pid": 5, "tid": 7, "ts": 1716454224771567,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718738, "dur": 13,
    "args": {
      "External id": 208808,
      "cbid": 211, "correlation": 208808
    }
  },
  {
    "ph": "s", "id": 208808, "pid": 76337, "tid": -914061504, "ts": 1716454224718738,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224771579, "dur": 165,
    "args": {
      "External id": 208810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208810,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208810, "pid": 5, "tid": 7, "ts": 1716454224771579,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718753, "dur": 5,
    "args": {
      "External id": 208810,
      "cbid": 211, "correlation": 208810
    }
  },
  {
    "ph": "s", "id": 208810, "pid": 76337, "tid": -914061504, "ts": 1716454224718753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224771746, "dur": 1,
    "args": {
      "External id": 208812,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 208812,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 208812, "pid": 5, "tid": 7, "ts": 1716454224771746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718765, "dur": 7,
    "args": {
      "External id": 208812,
      "cbid": 51, "correlation": 208812
    }
  },
  {
    "ph": "s", "id": 208812, "pid": 76337, "tid": -914061504, "ts": 1716454224718765,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224771750, "dur": 666,
    "args": {
      "External id": 208813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208813,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208813, "pid": 5, "tid": 7, "ts": 1716454224771750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718773, "dur": 6,
    "args": {
      "External id": 208813,
      "cbid": 211, "correlation": 208813
    }
  },
  {
    "ph": "s", "id": 208813, "pid": 76337, "tid": -914061504, "ts": 1716454224718773,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224772417, "dur": 12,
    "args": {
      "External id": 208815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208815,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208815, "pid": 5, "tid": 7, "ts": 1716454224772417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718783, "dur": 5,
    "args": {
      "External id": 208815,
      "cbid": 211, "correlation": 208815
    }
  },
  {
    "ph": "s", "id": 208815, "pid": 76337, "tid": -914061504, "ts": 1716454224718783,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224772431, "dur": 15,
    "args": {
      "External id": 208821,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208821,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208821, "pid": 5, "tid": 7, "ts": 1716454224772431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718812, "dur": 9,
    "args": {
      "External id": 208821,
      "cbid": 211, "correlation": 208821
    }
  },
  {
    "ph": "s", "id": 208821, "pid": 76337, "tid": -914061504, "ts": 1716454224718812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718871, "dur": 0,
    "args": {
      "External id": 208831,
      "cbid": 317, "correlation": 208831
    }
  },
  {
    "ph": "f", "id": 208831, "pid": 76337, "tid": -914061504, "ts": 1716454224718871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718872, "dur": 0,
    "args": {
      "External id": 208832,
      "cbid": 203, "correlation": 208832
    }
  },
  {
    "ph": "f", "id": 208832, "pid": 76337, "tid": -914061504, "ts": 1716454224718872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718872, "dur": 0,
    "args": {
      "External id": 208833,
      "cbid": 205, "correlation": 208833
    }
  },
  {
    "ph": "f", "id": 208833, "pid": 76337, "tid": -914061504, "ts": 1716454224718872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224772447, "dur": 8,
    "args": {
      "External id": 208837,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208837,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [6, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208837, "pid": 5, "tid": 7, "ts": 1716454224772447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718888, "dur": 13,
    "args": {
      "External id": 208837,
      "cbid": 211, "correlation": 208837
    }
  },
  {
    "ph": "s", "id": 208837, "pid": 76337, "tid": -914061504, "ts": 1716454224718888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224772457, "dur": 3,
    "args": {
      "External id": 208839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208839,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 208839, "pid": 5, "tid": 7, "ts": 1716454224772457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718907, "dur": 6,
    "args": {
      "External id": 208839,
      "cbid": 211, "correlation": 208839
    }
  },
  {
    "ph": "s", "id": 208839, "pid": 76337, "tid": -914061504, "ts": 1716454224718907,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718916, "dur": 0,
    "args": {
      "External id": 208840,
      "cbid": 51, "correlation": 208840
    }
  },
  {
    "ph": "s", "id": 208840, "pid": 76337, "tid": -914061504, "ts": 1716454224718916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224772461, "dur": 57,
    "args": {
      "External id": 208841,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208841,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [12, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 208841, "pid": 5, "tid": 7, "ts": 1716454224772461,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718916, "dur": 5,
    "args": {
      "External id": 208841,
      "cbid": 211, "correlation": 208841
    }
  },
  {
    "ph": "s", "id": 208841, "pid": 76337, "tid": -914061504, "ts": 1716454224718916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224772520, "dur": 14,
    "args": {
      "External id": 208846,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208846,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208846, "pid": 5, "tid": 7, "ts": 1716454224772520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718943, "dur": 8,
    "args": {
      "External id": 208846,
      "cbid": 211, "correlation": 208846
    }
  },
  {
    "ph": "s", "id": 208846, "pid": 76337, "tid": -914061504, "ts": 1716454224718943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224772536, "dur": 12,
    "args": {
      "External id": 208854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208854,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208854, "pid": 5, "tid": 7, "ts": 1716454224772536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224718971, "dur": 17,
    "args": {
      "External id": 208854,
      "cbid": 211, "correlation": 208854
    }
  },
  {
    "ph": "s", "id": 208854, "pid": 76337, "tid": -914061504, "ts": 1716454224718971,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224772549, "dur": 10,
    "args": {
      "External id": 208862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208862,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208862, "pid": 5, "tid": 7, "ts": 1716454224772549,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719011, "dur": 8,
    "args": {
      "External id": 208862,
      "cbid": 211, "correlation": 208862
    }
  },
  {
    "ph": "s", "id": 208862, "pid": 76337, "tid": -914061504, "ts": 1716454224719011,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224772561, "dur": 18,
    "args": {
      "External id": 208882,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208882,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 208882, "pid": 5, "tid": 7, "ts": 1716454224772561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719091, "dur": 12,
    "args": {
      "External id": 208882,
      "cbid": 211, "correlation": 208882
    }
  },
  {
    "ph": "s", "id": 208882, "pid": 76337, "tid": -914061504, "ts": 1716454224719091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224772580, "dur": 4,
    "args": {
      "External id": 208894,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208894,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 208894, "pid": 5, "tid": 7, "ts": 1716454224772580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719114, "dur": 6,
    "args": {
      "External id": 208894,
      "cbid": 211, "correlation": 208894
    }
  },
  {
    "ph": "s", "id": 208894, "pid": 76337, "tid": -914061504, "ts": 1716454224719114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224772586, "dur": 17,
    "args": {
      "External id": 208897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208897,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208897, "pid": 5, "tid": 7, "ts": 1716454224772586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719131, "dur": 6,
    "args": {
      "External id": 208897,
      "cbid": 211, "correlation": 208897
    }
  },
  {
    "ph": "s", "id": 208897, "pid": 76337, "tid": -914061504, "ts": 1716454224719131,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719189, "dur": 0,
    "args": {
      "External id": 208908,
      "cbid": 317, "correlation": 208908
    }
  },
  {
    "ph": "f", "id": 208908, "pid": 76337, "tid": -914061504, "ts": 1716454224719189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719190, "dur": 0,
    "args": {
      "External id": 208909,
      "cbid": 203, "correlation": 208909
    }
  },
  {
    "ph": "f", "id": 208909, "pid": 76337, "tid": -914061504, "ts": 1716454224719190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719190, "dur": 0,
    "args": {
      "External id": 208910,
      "cbid": 205, "correlation": 208910
    }
  },
  {
    "ph": "f", "id": 208910, "pid": 76337, "tid": -914061504, "ts": 1716454224719190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224772604, "dur": 11,
    "args": {
      "External id": 208914,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208914,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208914, "pid": 5, "tid": 7, "ts": 1716454224772604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719207, "dur": 12,
    "args": {
      "External id": 208914,
      "cbid": 211, "correlation": 208914
    }
  },
  {
    "ph": "s", "id": 208914, "pid": 76337, "tid": -914061504, "ts": 1716454224719207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224772617, "dur": 3,
    "args": {
      "External id": 208916,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208916,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 208916, "pid": 5, "tid": 7, "ts": 1716454224772617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719223, "dur": 5,
    "args": {
      "External id": 208916,
      "cbid": 211, "correlation": 208916
    }
  },
  {
    "ph": "s", "id": 208916, "pid": 76337, "tid": -914061504, "ts": 1716454224719223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719232, "dur": 0,
    "args": {
      "External id": 208917,
      "cbid": 51, "correlation": 208917
    }
  },
  {
    "ph": "s", "id": 208917, "pid": 76337, "tid": -914061504, "ts": 1716454224719232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224772621, "dur": 97,
    "args": {
      "External id": 208918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208918,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 208918, "pid": 5, "tid": 7, "ts": 1716454224772621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719233, "dur": 6,
    "args": {
      "External id": 208918,
      "cbid": 211, "correlation": 208918
    }
  },
  {
    "ph": "s", "id": 208918, "pid": 76337, "tid": -914061504, "ts": 1716454224719233,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224772720, "dur": 16,
    "args": {
      "External id": 208923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208923, "pid": 5, "tid": 7, "ts": 1716454224772720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719261, "dur": 8,
    "args": {
      "External id": 208923,
      "cbid": 211, "correlation": 208923
    }
  },
  {
    "ph": "s", "id": 208923, "pid": 76337, "tid": -914061504, "ts": 1716454224719261,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224772737, "dur": 85,
    "args": {
      "External id": 208932,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208932,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208932, "pid": 5, "tid": 7, "ts": 1716454224772737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719342, "dur": 14,
    "args": {
      "External id": 208932,
      "cbid": 211, "correlation": 208932
    }
  },
  {
    "ph": "s", "id": 208932, "pid": 76337, "tid": -914061504, "ts": 1716454224719342,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224772824, "dur": 31,
    "args": {
      "External id": 208954,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208954,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 208954, "pid": 5, "tid": 7, "ts": 1716454224772824,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719398, "dur": 10,
    "args": {
      "External id": 208954,
      "cbid": 211, "correlation": 208954
    }
  },
  {
    "ph": "s", "id": 208954, "pid": 76337, "tid": -914061504, "ts": 1716454224719398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719485, "dur": 1,
    "args": {
      "External id": 208965,
      "cbid": 251, "correlation": 208965
    }
  },
  {
    "ph": "f", "id": 208965, "pid": 76337, "tid": -914061504, "ts": 1716454224719485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224772856, "dur": 166,
    "args": {
      "External id": 208966,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208966,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208966, "pid": 5, "tid": 7, "ts": 1716454224772856,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719491, "dur": 13,
    "args": {
      "External id": 208966,
      "cbid": 211, "correlation": 208966
    }
  },
  {
    "ph": "s", "id": 208966, "pid": 76337, "tid": -914061504, "ts": 1716454224719491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719562, "dur": 1,
    "args": {
      "External id": 208977,
      "cbid": 251, "correlation": 208977
    }
  },
  {
    "ph": "f", "id": 208977, "pid": 76337, "tid": -914061504, "ts": 1716454224719562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224773023, "dur": 159,
    "args": {
      "External id": 208978,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208978,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208978, "pid": 5, "tid": 7, "ts": 1716454224773023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719566, "dur": 11,
    "args": {
      "External id": 208978,
      "cbid": 211, "correlation": 208978
    }
  },
  {
    "ph": "s", "id": 208978, "pid": 76337, "tid": -914061504, "ts": 1716454224719566,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719630, "dur": 1,
    "args": {
      "External id": 208989,
      "cbid": 251, "correlation": 208989
    }
  },
  {
    "ph": "f", "id": 208989, "pid": 76337, "tid": -914061504, "ts": 1716454224719630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224773184, "dur": 158,
    "args": {
      "External id": 208990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 208990,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 208990, "pid": 5, "tid": 7, "ts": 1716454224773184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719634, "dur": 12,
    "args": {
      "External id": 208990,
      "cbid": 211, "correlation": 208990
    }
  },
  {
    "ph": "s", "id": 208990, "pid": 76337, "tid": -914061504, "ts": 1716454224719634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224773343, "dur": 338,
    "args": {
      "External id": 209015,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209015,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209015, "pid": 5, "tid": 7, "ts": 1716454224773343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719720, "dur": 13,
    "args": {
      "External id": 209015,
      "cbid": 211, "correlation": 209015
    }
  },
  {
    "ph": "s", "id": 209015, "pid": 76337, "tid": -914061504, "ts": 1716454224719720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719821, "dur": 1,
    "args": {
      "External id": 209033,
      "cbid": 251, "correlation": 209033
    }
  },
  {
    "ph": "f", "id": 209033, "pid": 76337, "tid": -914061504, "ts": 1716454224719821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224773682, "dur": 167,
    "args": {
      "External id": 209035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209035,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209035, "pid": 5, "tid": 7, "ts": 1716454224773682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719827, "dur": 13,
    "args": {
      "External id": 209035,
      "cbid": 211, "correlation": 209035
    }
  },
  {
    "ph": "s", "id": 209035, "pid": 76337, "tid": -914061504, "ts": 1716454224719827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224773851, "dur": 19,
    "args": {
      "External id": 209043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209043,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209043, "pid": 5, "tid": 7, "ts": 1716454224773851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719897, "dur": 12,
    "args": {
      "External id": 209043,
      "cbid": 211, "correlation": 209043
    }
  },
  {
    "ph": "s", "id": 209043, "pid": 76337, "tid": -914061504, "ts": 1716454224719897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224773871, "dur": 27,
    "args": {
      "External id": 209051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209051,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209051, "pid": 5, "tid": 7, "ts": 1716454224773871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224719936, "dur": 9,
    "args": {
      "External id": 209051,
      "cbid": 211, "correlation": 209051
    }
  },
  {
    "ph": "s", "id": 209051, "pid": 76337, "tid": -914061504, "ts": 1716454224719936,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224773900, "dur": 18,
    "args": {
      "External id": 209062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209062,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209062, "pid": 5, "tid": 7, "ts": 1716454224773900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720017, "dur": 13,
    "args": {
      "External id": 209062,
      "cbid": 211, "correlation": 209062
    }
  },
  {
    "ph": "s", "id": 209062, "pid": 76337, "tid": -914061504, "ts": 1716454224720017,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224773919, "dur": 16,
    "args": {
      "External id": 209084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209084,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209084, "pid": 5, "tid": 7, "ts": 1716454224773919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720049, "dur": 7,
    "args": {
      "External id": 209084,
      "cbid": 211, "correlation": 209084
    }
  },
  {
    "ph": "s", "id": 209084, "pid": 76337, "tid": -914061504, "ts": 1716454224720049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720135, "dur": 1,
    "args": {
      "External id": 209095,
      "cbid": 251, "correlation": 209095
    }
  },
  {
    "ph": "f", "id": 209095, "pid": 76337, "tid": -914061504, "ts": 1716454224720135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224773938, "dur": 89,
    "args": {
      "External id": 209096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209096,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209096, "pid": 5, "tid": 7, "ts": 1716454224773938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720141, "dur": 16,
    "args": {
      "External id": 209096,
      "cbid": 211, "correlation": 209096
    }
  },
  {
    "ph": "s", "id": 209096, "pid": 76337, "tid": -914061504, "ts": 1716454224720141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720213, "dur": 1,
    "args": {
      "External id": 209107,
      "cbid": 251, "correlation": 209107
    }
  },
  {
    "ph": "f", "id": 209107, "pid": 76337, "tid": -914061504, "ts": 1716454224720213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720217, "dur": 0,
    "args": {
      "External id": 209108,
      "cbid": 251, "correlation": 209108
    }
  },
  {
    "ph": "f", "id": 209108, "pid": 76337, "tid": -914061504, "ts": 1716454224720217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224774028, "dur": 12,
    "args": {
      "External id": 209109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209109,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209109, "pid": 5, "tid": 7, "ts": 1716454224774028,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720219, "dur": 12,
    "args": {
      "External id": 209109,
      "cbid": 211, "correlation": 209109
    }
  },
  {
    "ph": "s", "id": 209109, "pid": 76337, "tid": -914061504, "ts": 1716454224720219,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224774042, "dur": 6,
    "args": {
      "External id": 209111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209111,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209111, "pid": 5, "tid": 7, "ts": 1716454224774042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720232, "dur": 6,
    "args": {
      "External id": 209111,
      "cbid": 211, "correlation": 209111
    }
  },
  {
    "ph": "s", "id": 209111, "pid": 76337, "tid": -914061504, "ts": 1716454224720232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720291, "dur": 1,
    "args": {
      "External id": 209122,
      "cbid": 251, "correlation": 209122
    }
  },
  {
    "ph": "f", "id": 209122, "pid": 76337, "tid": -914061504, "ts": 1716454224720291,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720295, "dur": 0,
    "args": {
      "External id": 209123,
      "cbid": 251, "correlation": 209123
    }
  },
  {
    "ph": "f", "id": 209123, "pid": 76337, "tid": -914061504, "ts": 1716454224720295,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224774049, "dur": 8,
    "args": {
      "External id": 209124,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209124,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209124, "pid": 5, "tid": 7, "ts": 1716454224774049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720296, "dur": 12,
    "args": {
      "External id": 209124,
      "cbid": 211, "correlation": 209124
    }
  },
  {
    "ph": "s", "id": 209124, "pid": 76337, "tid": -914061504, "ts": 1716454224720296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224774059, "dur": 3,
    "args": {
      "External id": 209126,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209126,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209126, "pid": 5, "tid": 7, "ts": 1716454224774059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720310, "dur": 5,
    "args": {
      "External id": 209126,
      "cbid": 211, "correlation": 209126
    }
  },
  {
    "ph": "s", "id": 209126, "pid": 76337, "tid": -914061504, "ts": 1716454224720310,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224774063, "dur": 55,
    "args": {
      "External id": 209151,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209151,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209151, "pid": 5, "tid": 7, "ts": 1716454224774063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720387, "dur": 13,
    "args": {
      "External id": 209151,
      "cbid": 211, "correlation": 209151
    }
  },
  {
    "ph": "s", "id": 209151, "pid": 76337, "tid": -914061504, "ts": 1716454224720387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720486, "dur": 2,
    "args": {
      "External id": 209169,
      "cbid": 251, "correlation": 209169
    }
  },
  {
    "ph": "f", "id": 209169, "pid": 76337, "tid": -914061504, "ts": 1716454224720486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224774120, "dur": 92,
    "args": {
      "External id": 209171,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209171,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209171, "pid": 5, "tid": 7, "ts": 1716454224774120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720494, "dur": 14,
    "args": {
      "External id": 209171,
      "cbid": 211, "correlation": 209171
    }
  },
  {
    "ph": "s", "id": 209171, "pid": 76337, "tid": -914061504, "ts": 1716454224720494,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224774213, "dur": 9,
    "args": {
      "External id": 209179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209179,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209179, "pid": 5, "tid": 7, "ts": 1716454224774213,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720564, "dur": 12,
    "args": {
      "External id": 209179,
      "cbid": 211, "correlation": 209179
    }
  },
  {
    "ph": "s", "id": 209179, "pid": 76337, "tid": -914061504, "ts": 1716454224720564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224774224, "dur": 22,
    "args": {
      "External id": 209187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209187,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209187, "pid": 5, "tid": 7, "ts": 1716454224774224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720606, "dur": 9,
    "args": {
      "External id": 209187,
      "cbid": 211, "correlation": 209187
    }
  },
  {
    "ph": "s", "id": 209187, "pid": 76337, "tid": -914061504, "ts": 1716454224720606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224774247, "dur": 18,
    "args": {
      "External id": 209209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209209,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209209, "pid": 5, "tid": 7, "ts": 1716454224774247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720657, "dur": 10,
    "args": {
      "External id": 209209,
      "cbid": 211, "correlation": 209209
    }
  },
  {
    "ph": "s", "id": 209209, "pid": 76337, "tid": -914061504, "ts": 1716454224720657,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720745, "dur": 2,
    "args": {
      "External id": 209225,
      "cbid": 251, "correlation": 209225
    }
  },
  {
    "ph": "f", "id": 209225, "pid": 76337, "tid": -914061504, "ts": 1716454224720745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720752, "dur": 0,
    "args": {
      "External id": 209227,
      "cbid": 251, "correlation": 209227
    }
  },
  {
    "ph": "f", "id": 209227, "pid": 76337, "tid": -914061504, "ts": 1716454224720752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224774267, "dur": 497,
    "args": {
      "External id": 209228,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209228,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209228, "pid": 5, "tid": 7, "ts": 1716454224774267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720754, "dur": 15,
    "args": {
      "External id": 209228,
      "cbid": 211, "correlation": 209228
    }
  },
  {
    "ph": "s", "id": 209228, "pid": 76337, "tid": -914061504, "ts": 1716454224720754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224774765, "dur": 66,
    "args": {
      "External id": 209236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209236,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209236, "pid": 5, "tid": 7, "ts": 1716454224774765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720822, "dur": 12,
    "args": {
      "External id": 209236,
      "cbid": 211, "correlation": 209236
    }
  },
  {
    "ph": "s", "id": 209236, "pid": 76337, "tid": -914061504, "ts": 1716454224720822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224774832, "dur": 68,
    "args": {
      "External id": 209244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209244,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209244, "pid": 5, "tid": 7, "ts": 1716454224774832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720853, "dur": 8,
    "args": {
      "External id": 209244,
      "cbid": 211, "correlation": 209244
    }
  },
  {
    "ph": "s", "id": 209244, "pid": 76337, "tid": -914061504, "ts": 1716454224720853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720933, "dur": 1,
    "args": {
      "External id": 209260,
      "cbid": 251, "correlation": 209260
    }
  },
  {
    "ph": "f", "id": 209260, "pid": 76337, "tid": -914061504, "ts": 1716454224720933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224774902, "dur": 1,
    "args": {
      "External id": 209262,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 209262,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 209262, "pid": 5, "tid": 7, "ts": 1716454224774902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720939, "dur": 12,
    "args": {
      "External id": 209262,
      "cbid": 51, "correlation": 209262
    }
  },
  {
    "ph": "s", "id": 209262, "pid": 76337, "tid": -914061504, "ts": 1716454224720939,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224774906, "dur": 271,
    "args": {
      "External id": 209263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209263,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209263, "pid": 5, "tid": 7, "ts": 1716454224774906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224720953, "dur": 11,
    "args": {
      "External id": 209263,
      "cbid": 211, "correlation": 209263
    }
  },
  {
    "ph": "s", "id": 209263, "pid": 76337, "tid": -914061504, "ts": 1716454224720953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224775178, "dur": 14,
    "args": {
      "External id": 209271,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209271,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209271, "pid": 5, "tid": 7, "ts": 1716454224775178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721005, "dur": 11,
    "args": {
      "External id": 209271,
      "cbid": 211, "correlation": 209271
    }
  },
  {
    "ph": "s", "id": 209271, "pid": 76337, "tid": -914061504, "ts": 1716454224721005,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224775193, "dur": 38,
    "args": {
      "External id": 209282,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209282,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209282, "pid": 5, "tid": 7, "ts": 1716454224775193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721073, "dur": 12,
    "args": {
      "External id": 209282,
      "cbid": 211, "correlation": 209282
    }
  },
  {
    "ph": "s", "id": 209282, "pid": 76337, "tid": -914061504, "ts": 1716454224721073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721138, "dur": 0,
    "args": {
      "External id": 209294,
      "cbid": 317, "correlation": 209294
    }
  },
  {
    "ph": "f", "id": 209294, "pid": 76337, "tid": -914061504, "ts": 1716454224721138,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721139, "dur": 0,
    "args": {
      "External id": 209295,
      "cbid": 203, "correlation": 209295
    }
  },
  {
    "ph": "f", "id": 209295, "pid": 76337, "tid": -914061504, "ts": 1716454224721139,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721140, "dur": 0,
    "args": {
      "External id": 209296,
      "cbid": 205, "correlation": 209296
    }
  },
  {
    "ph": "f", "id": 209296, "pid": 76337, "tid": -914061504, "ts": 1716454224721140,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224775233, "dur": 13,
    "args": {
      "External id": 209300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209300,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209300, "pid": 5, "tid": 7, "ts": 1716454224775233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721155, "dur": 12,
    "args": {
      "External id": 209300,
      "cbid": 211, "correlation": 209300
    }
  },
  {
    "ph": "s", "id": 209300, "pid": 76337, "tid": -914061504, "ts": 1716454224721155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224775247, "dur": 4,
    "args": {
      "External id": 209302,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209302,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209302, "pid": 5, "tid": 7, "ts": 1716454224775247,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721172, "dur": 6,
    "args": {
      "External id": 209302,
      "cbid": 211, "correlation": 209302
    }
  },
  {
    "ph": "s", "id": 209302, "pid": 76337, "tid": -914061504, "ts": 1716454224721172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721181, "dur": 0,
    "args": {
      "External id": 209303,
      "cbid": 51, "correlation": 209303
    }
  },
  {
    "ph": "s", "id": 209303, "pid": 76337, "tid": -914061504, "ts": 1716454224721181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224775252, "dur": 99,
    "args": {
      "External id": 209304,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209304,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 209304, "pid": 5, "tid": 7, "ts": 1716454224775252,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721181, "dur": 5,
    "args": {
      "External id": 209304,
      "cbid": 211, "correlation": 209304
    }
  },
  {
    "ph": "s", "id": 209304, "pid": 76337, "tid": -914061504, "ts": 1716454224721181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224775353, "dur": 16,
    "args": {
      "External id": 209309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209309,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209309, "pid": 5, "tid": 7, "ts": 1716454224775353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721209, "dur": 8,
    "args": {
      "External id": 209309,
      "cbid": 211, "correlation": 209309
    }
  },
  {
    "ph": "s", "id": 209309, "pid": 76337, "tid": -914061504, "ts": 1716454224721209,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224775370, "dur": 11,
    "args": {
      "External id": 209317,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209317,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209317, "pid": 5, "tid": 7, "ts": 1716454224775370,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721240, "dur": 10,
    "args": {
      "External id": 209317,
      "cbid": 211, "correlation": 209317
    }
  },
  {
    "ph": "s", "id": 209317, "pid": 76337, "tid": -914061504, "ts": 1716454224721240,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224775383, "dur": 19,
    "args": {
      "External id": 209337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209337,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 209337, "pid": 5, "tid": 7, "ts": 1716454224775383,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721313, "dur": 12,
    "args": {
      "External id": 209337,
      "cbid": 211, "correlation": 209337
    }
  },
  {
    "ph": "s", "id": 209337, "pid": 76337, "tid": -914061504, "ts": 1716454224721313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224775403, "dur": 4,
    "args": {
      "External id": 209349,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209349,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 209349, "pid": 5, "tid": 7, "ts": 1716454224775403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721335, "dur": 7,
    "args": {
      "External id": 209349,
      "cbid": 211, "correlation": 209349
    }
  },
  {
    "ph": "s", "id": 209349, "pid": 76337, "tid": -914061504, "ts": 1716454224721335,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224775408, "dur": 18,
    "args": {
      "External id": 209352,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209352,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209352, "pid": 5, "tid": 7, "ts": 1716454224775408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721354, "dur": 6,
    "args": {
      "External id": 209352,
      "cbid": 211, "correlation": 209352
    }
  },
  {
    "ph": "s", "id": 209352, "pid": 76337, "tid": -914061504, "ts": 1716454224721354,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224775428, "dur": 12,
    "args": {
      "External id": 209361,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209361,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209361, "pid": 5, "tid": 7, "ts": 1716454224775428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721392, "dur": 10,
    "args": {
      "External id": 209361,
      "cbid": 211, "correlation": 209361
    }
  },
  {
    "ph": "s", "id": 209361, "pid": 76337, "tid": -914061504, "ts": 1716454224721392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721444, "dur": 0,
    "args": {
      "External id": 209371,
      "cbid": 317, "correlation": 209371
    }
  },
  {
    "ph": "f", "id": 209371, "pid": 76337, "tid": -914061504, "ts": 1716454224721444,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721445, "dur": 0,
    "args": {
      "External id": 209372,
      "cbid": 203, "correlation": 209372
    }
  },
  {
    "ph": "f", "id": 209372, "pid": 76337, "tid": -914061504, "ts": 1716454224721445,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721446, "dur": 0,
    "args": {
      "External id": 209373,
      "cbid": 205, "correlation": 209373
    }
  },
  {
    "ph": "f", "id": 209373, "pid": 76337, "tid": -914061504, "ts": 1716454224721446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224775442, "dur": 11,
    "args": {
      "External id": 209377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209377,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209377, "pid": 5, "tid": 7, "ts": 1716454224775442,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721459, "dur": 11,
    "args": {
      "External id": 209377,
      "cbid": 211, "correlation": 209377
    }
  },
  {
    "ph": "s", "id": 209377, "pid": 76337, "tid": -914061504, "ts": 1716454224721459,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224775454, "dur": 164,
    "args": {
      "External id": 209379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209379,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209379, "pid": 5, "tid": 7, "ts": 1716454224775454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721473, "dur": 5,
    "args": {
      "External id": 209379,
      "cbid": 211, "correlation": 209379
    }
  },
  {
    "ph": "s", "id": 209379, "pid": 76337, "tid": -914061504, "ts": 1716454224721473,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224775620, "dur": 1,
    "args": {
      "External id": 209381,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 209381,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 209381, "pid": 5, "tid": 7, "ts": 1716454224775620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721484, "dur": 8,
    "args": {
      "External id": 209381,
      "cbid": 51, "correlation": 209381
    }
  },
  {
    "ph": "s", "id": 209381, "pid": 76337, "tid": -914061504, "ts": 1716454224721484,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224775624, "dur": 665,
    "args": {
      "External id": 209382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209382,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209382, "pid": 5, "tid": 7, "ts": 1716454224775624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721492, "dur": 6,
    "args": {
      "External id": 209382,
      "cbid": 211, "correlation": 209382
    }
  },
  {
    "ph": "s", "id": 209382, "pid": 76337, "tid": -914061504, "ts": 1716454224721492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224776290, "dur": 14,
    "args": {
      "External id": 209384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209384,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209384, "pid": 5, "tid": 7, "ts": 1716454224776290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721503, "dur": 5,
    "args": {
      "External id": 209384,
      "cbid": 211, "correlation": 209384
    }
  },
  {
    "ph": "s", "id": 209384, "pid": 76337, "tid": -914061504, "ts": 1716454224721503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224776305, "dur": 15,
    "args": {
      "External id": 209390,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209390,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209390, "pid": 5, "tid": 7, "ts": 1716454224776305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721530, "dur": 9,
    "args": {
      "External id": 209390,
      "cbid": 211, "correlation": 209390
    }
  },
  {
    "ph": "s", "id": 209390, "pid": 76337, "tid": -914061504, "ts": 1716454224721530,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224776322, "dur": 3,
    "args": {
      "External id": 209398,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209398,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 209398, "pid": 5, "tid": 7, "ts": 1716454224776322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721574, "dur": 9,
    "args": {
      "External id": 209398,
      "cbid": 211, "correlation": 209398
    }
  },
  {
    "ph": "s", "id": 209398, "pid": 76337, "tid": -914061504, "ts": 1716454224721574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721638, "dur": 1,
    "args": {
      "External id": 209414,
      "cbid": 251, "correlation": 209414
    }
  },
  {
    "ph": "f", "id": 209414, "pid": 76337, "tid": -914061504, "ts": 1716454224721638,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721644, "dur": 0,
    "args": {
      "External id": 209416,
      "cbid": 251, "correlation": 209416
    }
  },
  {
    "ph": "f", "id": 209416, "pid": 76337, "tid": -914061504, "ts": 1716454224721644,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224776326, "dur": 13,
    "args": {
      "External id": 209417,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209417,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209417, "pid": 5, "tid": 7, "ts": 1716454224776326,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721646, "dur": 11,
    "args": {
      "External id": 209417,
      "cbid": 211, "correlation": 209417
    }
  },
  {
    "ph": "s", "id": 209417, "pid": 76337, "tid": -914061504, "ts": 1716454224721646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224776341, "dur": 5,
    "args": {
      "External id": 209419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209419,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209419, "pid": 5, "tid": 7, "ts": 1716454224776341,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721659, "dur": 5,
    "args": {
      "External id": 209419,
      "cbid": 211, "correlation": 209419
    }
  },
  {
    "ph": "s", "id": 209419, "pid": 76337, "tid": -914061504, "ts": 1716454224721659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224776347, "dur": 17,
    "args": {
      "External id": 209429,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209429,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209429, "pid": 5, "tid": 7, "ts": 1716454224776347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721716, "dur": 12,
    "args": {
      "External id": 209429,
      "cbid": 211, "correlation": 209429
    }
  },
  {
    "ph": "s", "id": 209429, "pid": 76337, "tid": -914061504, "ts": 1716454224721716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224776366, "dur": 18,
    "args": {
      "External id": 209449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209449,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 209449, "pid": 5, "tid": 7, "ts": 1716454224776366,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721782, "dur": 12,
    "args": {
      "External id": 209449,
      "cbid": 211, "correlation": 209449
    }
  },
  {
    "ph": "s", "id": 209449, "pid": 76337, "tid": -914061504, "ts": 1716454224721782,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224776385, "dur": 4,
    "args": {
      "External id": 209461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209461,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 209461, "pid": 5, "tid": 7, "ts": 1716454224776385,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721803, "dur": 6,
    "args": {
      "External id": 209461,
      "cbid": 211, "correlation": 209461
    }
  },
  {
    "ph": "s", "id": 209461, "pid": 76337, "tid": -914061504, "ts": 1716454224721803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224776390, "dur": 17,
    "args": {
      "External id": 209464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209464,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209464, "pid": 5, "tid": 7, "ts": 1716454224776390,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721822, "dur": 6,
    "args": {
      "External id": 209464,
      "cbid": 211, "correlation": 209464
    }
  },
  {
    "ph": "s", "id": 209464, "pid": 76337, "tid": -914061504, "ts": 1716454224721822,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224776408, "dur": 11,
    "args": {
      "External id": 209473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209473,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209473, "pid": 5, "tid": 7, "ts": 1716454224776408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721862, "dur": 10,
    "args": {
      "External id": 209473,
      "cbid": 211, "correlation": 209473
    }
  },
  {
    "ph": "s", "id": 209473, "pid": 76337, "tid": -914061504, "ts": 1716454224721862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721923, "dur": 0,
    "args": {
      "External id": 209483,
      "cbid": 317, "correlation": 209483
    }
  },
  {
    "ph": "f", "id": 209483, "pid": 76337, "tid": -914061504, "ts": 1716454224721923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721924, "dur": 0,
    "args": {
      "External id": 209484,
      "cbid": 203, "correlation": 209484
    }
  },
  {
    "ph": "f", "id": 209484, "pid": 76337, "tid": -914061504, "ts": 1716454224721924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721924, "dur": 0,
    "args": {
      "External id": 209485,
      "cbid": 205, "correlation": 209485
    }
  },
  {
    "ph": "f", "id": 209485, "pid": 76337, "tid": -914061504, "ts": 1716454224721924,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224776421, "dur": 11,
    "args": {
      "External id": 209489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209489,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209489, "pid": 5, "tid": 7, "ts": 1716454224776421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721938, "dur": 13,
    "args": {
      "External id": 209489,
      "cbid": 211, "correlation": 209489
    }
  },
  {
    "ph": "s", "id": 209489, "pid": 76337, "tid": -914061504, "ts": 1716454224721938,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224776433, "dur": 164,
    "args": {
      "External id": 209491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209491,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209491, "pid": 5, "tid": 7, "ts": 1716454224776433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721954, "dur": 5,
    "args": {
      "External id": 209491,
      "cbid": 211, "correlation": 209491
    }
  },
  {
    "ph": "s", "id": 209491, "pid": 76337, "tid": -914061504, "ts": 1716454224721954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224776600, "dur": 1,
    "args": {
      "External id": 209493,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 209493,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 209493, "pid": 5, "tid": 7, "ts": 1716454224776600,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721964, "dur": 6,
    "args": {
      "External id": 209493,
      "cbid": 51, "correlation": 209493
    }
  },
  {
    "ph": "s", "id": 209493, "pid": 76337, "tid": -914061504, "ts": 1716454224721964,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224776603, "dur": 654,
    "args": {
      "External id": 209494,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209494,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209494, "pid": 5, "tid": 7, "ts": 1716454224776603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721972, "dur": 14,
    "args": {
      "External id": 209494,
      "cbid": 211, "correlation": 209494
    }
  },
  {
    "ph": "s", "id": 209494, "pid": 76337, "tid": -914061504, "ts": 1716454224721972,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224777259, "dur": 13,
    "args": {
      "External id": 209496,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209496,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209496, "pid": 5, "tid": 7, "ts": 1716454224777259,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224721990, "dur": 6,
    "args": {
      "External id": 209496,
      "cbid": 211, "correlation": 209496
    }
  },
  {
    "ph": "s", "id": 209496, "pid": 76337, "tid": -914061504, "ts": 1716454224721990,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224777273, "dur": 15,
    "args": {
      "External id": 209502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209502,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209502, "pid": 5, "tid": 7, "ts": 1716454224777273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722019, "dur": 9,
    "args": {
      "External id": 209502,
      "cbid": 211, "correlation": 209502
    }
  },
  {
    "ph": "s", "id": 209502, "pid": 76337, "tid": -914061504, "ts": 1716454224722019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224777289, "dur": 12,
    "args": {
      "External id": 209510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209510,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209510, "pid": 5, "tid": 7, "ts": 1716454224777289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722052, "dur": 8,
    "args": {
      "External id": 209510,
      "cbid": 211, "correlation": 209510
    }
  },
  {
    "ph": "s", "id": 209510, "pid": 76337, "tid": -914061504, "ts": 1716454224722052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224777303, "dur": 10,
    "args": {
      "External id": 209518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209518,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209518, "pid": 5, "tid": 7, "ts": 1716454224777303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722081, "dur": 9,
    "args": {
      "External id": 209518,
      "cbid": 211, "correlation": 209518
    }
  },
  {
    "ph": "s", "id": 209518, "pid": 76337, "tid": -914061504, "ts": 1716454224722081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224777314, "dur": 18,
    "args": {
      "External id": 209538,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209538,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 209538, "pid": 5, "tid": 7, "ts": 1716454224777314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722160, "dur": 12,
    "args": {
      "External id": 209538,
      "cbid": 211, "correlation": 209538
    }
  },
  {
    "ph": "s", "id": 209538, "pid": 76337, "tid": -914061504, "ts": 1716454224722160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224777333, "dur": 4,
    "args": {
      "External id": 209550,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209550,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 209550, "pid": 5, "tid": 7, "ts": 1716454224777333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722182, "dur": 6,
    "args": {
      "External id": 209550,
      "cbid": 211, "correlation": 209550
    }
  },
  {
    "ph": "s", "id": 209550, "pid": 76337, "tid": -914061504, "ts": 1716454224722182,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224777339, "dur": 16,
    "args": {
      "External id": 209553,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209553,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209553, "pid": 5, "tid": 7, "ts": 1716454224777339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722200, "dur": 7,
    "args": {
      "External id": 209553,
      "cbid": 211, "correlation": 209553
    }
  },
  {
    "ph": "s", "id": 209553, "pid": 76337, "tid": -914061504, "ts": 1716454224722200,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722256, "dur": 0,
    "args": {
      "External id": 209564,
      "cbid": 317, "correlation": 209564
    }
  },
  {
    "ph": "f", "id": 209564, "pid": 76337, "tid": -914061504, "ts": 1716454224722256,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722257, "dur": 0,
    "args": {
      "External id": 209565,
      "cbid": 203, "correlation": 209565
    }
  },
  {
    "ph": "f", "id": 209565, "pid": 76337, "tid": -914061504, "ts": 1716454224722257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722258, "dur": 0,
    "args": {
      "External id": 209566,
      "cbid": 205, "correlation": 209566
    }
  },
  {
    "ph": "f", "id": 209566, "pid": 76337, "tid": -914061504, "ts": 1716454224722258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224777357, "dur": 11,
    "args": {
      "External id": 209570,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209570,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209570, "pid": 5, "tid": 7, "ts": 1716454224777357,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722271, "dur": 11,
    "args": {
      "External id": 209570,
      "cbid": 211, "correlation": 209570
    }
  },
  {
    "ph": "s", "id": 209570, "pid": 76337, "tid": -914061504, "ts": 1716454224722271,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224777369, "dur": 4,
    "args": {
      "External id": 209572,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209572,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209572, "pid": 5, "tid": 7, "ts": 1716454224777369,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722286, "dur": 7,
    "args": {
      "External id": 209572,
      "cbid": 211, "correlation": 209572
    }
  },
  {
    "ph": "s", "id": 209572, "pid": 76337, "tid": -914061504, "ts": 1716454224722286,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722296, "dur": 0,
    "args": {
      "External id": 209573,
      "cbid": 51, "correlation": 209573
    }
  },
  {
    "ph": "s", "id": 209573, "pid": 76337, "tid": -914061504, "ts": 1716454224722296,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224777375, "dur": 95,
    "args": {
      "External id": 209574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209574,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 209574, "pid": 5, "tid": 7, "ts": 1716454224777375,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722297, "dur": 5,
    "args": {
      "External id": 209574,
      "cbid": 211, "correlation": 209574
    }
  },
  {
    "ph": "s", "id": 209574, "pid": 76337, "tid": -914061504, "ts": 1716454224722297,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224777472, "dur": 16,
    "args": {
      "External id": 209579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209579,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209579, "pid": 5, "tid": 7, "ts": 1716454224777472,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722323, "dur": 8,
    "args": {
      "External id": 209579,
      "cbid": 211, "correlation": 209579
    }
  },
  {
    "ph": "s", "id": 209579, "pid": 76337, "tid": -914061504, "ts": 1716454224722323,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224777489, "dur": 85,
    "args": {
      "External id": 209588,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209588,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209588, "pid": 5, "tid": 7, "ts": 1716454224777489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722404, "dur": 14,
    "args": {
      "External id": 209588,
      "cbid": 211, "correlation": 209588
    }
  },
  {
    "ph": "s", "id": 209588, "pid": 76337, "tid": -914061504, "ts": 1716454224722404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224777576, "dur": 30,
    "args": {
      "External id": 209610,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209610,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209610, "pid": 5, "tid": 7, "ts": 1716454224777576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722460, "dur": 10,
    "args": {
      "External id": 209610,
      "cbid": 211, "correlation": 209610
    }
  },
  {
    "ph": "s", "id": 209610, "pid": 76337, "tid": -914061504, "ts": 1716454224722460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722547, "dur": 1,
    "args": {
      "External id": 209621,
      "cbid": 251, "correlation": 209621
    }
  },
  {
    "ph": "f", "id": 209621, "pid": 76337, "tid": -914061504, "ts": 1716454224722547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224777607, "dur": 166,
    "args": {
      "External id": 209622,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209622,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209622, "pid": 5, "tid": 7, "ts": 1716454224777607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722552, "dur": 13,
    "args": {
      "External id": 209622,
      "cbid": 211, "correlation": 209622
    }
  },
  {
    "ph": "s", "id": 209622, "pid": 76337, "tid": -914061504, "ts": 1716454224722552,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722621, "dur": 1,
    "args": {
      "External id": 209633,
      "cbid": 251, "correlation": 209633
    }
  },
  {
    "ph": "f", "id": 209633, "pid": 76337, "tid": -914061504, "ts": 1716454224722621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224777774, "dur": 164,
    "args": {
      "External id": 209634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209634,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209634, "pid": 5, "tid": 7, "ts": 1716454224777774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722625, "dur": 11,
    "args": {
      "External id": 209634,
      "cbid": 211, "correlation": 209634
    }
  },
  {
    "ph": "s", "id": 209634, "pid": 76337, "tid": -914061504, "ts": 1716454224722625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722689, "dur": 1,
    "args": {
      "External id": 209645,
      "cbid": 251, "correlation": 209645
    }
  },
  {
    "ph": "f", "id": 209645, "pid": 76337, "tid": -914061504, "ts": 1716454224722689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224777939, "dur": 161,
    "args": {
      "External id": 209646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209646,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209646, "pid": 5, "tid": 7, "ts": 1716454224777939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722694, "dur": 12,
    "args": {
      "External id": 209646,
      "cbid": 211, "correlation": 209646
    }
  },
  {
    "ph": "s", "id": 209646, "pid": 76337, "tid": -914061504, "ts": 1716454224722694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224778102, "dur": 343,
    "args": {
      "External id": 209671,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209671,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209671, "pid": 5, "tid": 7, "ts": 1716454224778102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722777, "dur": 12,
    "args": {
      "External id": 209671,
      "cbid": 211, "correlation": 209671
    }
  },
  {
    "ph": "s", "id": 209671, "pid": 76337, "tid": -914061504, "ts": 1716454224722777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722875, "dur": 1,
    "args": {
      "External id": 209689,
      "cbid": 251, "correlation": 209689
    }
  },
  {
    "ph": "f", "id": 209689, "pid": 76337, "tid": -914061504, "ts": 1716454224722875,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224778446, "dur": 166,
    "args": {
      "External id": 209691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209691,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209691, "pid": 5, "tid": 7, "ts": 1716454224778446,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722881, "dur": 13,
    "args": {
      "External id": 209691,
      "cbid": 211, "correlation": 209691
    }
  },
  {
    "ph": "s", "id": 209691, "pid": 76337, "tid": -914061504, "ts": 1716454224722881,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224778613, "dur": 19,
    "args": {
      "External id": 209699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209699,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209699, "pid": 5, "tid": 7, "ts": 1716454224778613,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224722952, "dur": 12,
    "args": {
      "External id": 209699,
      "cbid": 211, "correlation": 209699
    }
  },
  {
    "ph": "s", "id": 209699, "pid": 76337, "tid": -914061504, "ts": 1716454224722952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224778633, "dur": 27,
    "args": {
      "External id": 209707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209707,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209707, "pid": 5, "tid": 7, "ts": 1716454224778633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723000, "dur": 9,
    "args": {
      "External id": 209707,
      "cbid": 211, "correlation": 209707
    }
  },
  {
    "ph": "s", "id": 209707, "pid": 76337, "tid": -914061504, "ts": 1716454224723000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224778662, "dur": 19,
    "args": {
      "External id": 209718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209718,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209718, "pid": 5, "tid": 7, "ts": 1716454224778662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723072, "dur": 12,
    "args": {
      "External id": 209718,
      "cbid": 211, "correlation": 209718
    }
  },
  {
    "ph": "s", "id": 209718, "pid": 76337, "tid": -914061504, "ts": 1716454224723072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224778682, "dur": 16,
    "args": {
      "External id": 209740,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209740,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209740, "pid": 5, "tid": 7, "ts": 1716454224778682,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723104, "dur": 8,
    "args": {
      "External id": 209740,
      "cbid": 211, "correlation": 209740
    }
  },
  {
    "ph": "s", "id": 209740, "pid": 76337, "tid": -914061504, "ts": 1716454224723104,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723189, "dur": 1,
    "args": {
      "External id": 209751,
      "cbid": 251, "correlation": 209751
    }
  },
  {
    "ph": "f", "id": 209751, "pid": 76337, "tid": -914061504, "ts": 1716454224723189,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224778700, "dur": 90,
    "args": {
      "External id": 209752,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209752,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209752, "pid": 5, "tid": 7, "ts": 1716454224778700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723195, "dur": 14,
    "args": {
      "External id": 209752,
      "cbid": 211, "correlation": 209752
    }
  },
  {
    "ph": "s", "id": 209752, "pid": 76337, "tid": -914061504, "ts": 1716454224723195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723263, "dur": 1,
    "args": {
      "External id": 209763,
      "cbid": 251, "correlation": 209763
    }
  },
  {
    "ph": "f", "id": 209763, "pid": 76337, "tid": -914061504, "ts": 1716454224723263,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723267, "dur": 0,
    "args": {
      "External id": 209764,
      "cbid": 251, "correlation": 209764
    }
  },
  {
    "ph": "f", "id": 209764, "pid": 76337, "tid": -914061504, "ts": 1716454224723267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224778790, "dur": 12,
    "args": {
      "External id": 209765,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209765,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209765, "pid": 5, "tid": 7, "ts": 1716454224778790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723268, "dur": 12,
    "args": {
      "External id": 209765,
      "cbid": 211, "correlation": 209765
    }
  },
  {
    "ph": "s", "id": 209765, "pid": 76337, "tid": -914061504, "ts": 1716454224723268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224778804, "dur": 5,
    "args": {
      "External id": 209767,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209767,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209767, "pid": 5, "tid": 7, "ts": 1716454224778804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723282, "dur": 6,
    "args": {
      "External id": 209767,
      "cbid": 211, "correlation": 209767
    }
  },
  {
    "ph": "s", "id": 209767, "pid": 76337, "tid": -914061504, "ts": 1716454224723282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723340, "dur": 1,
    "args": {
      "External id": 209778,
      "cbid": 251, "correlation": 209778
    }
  },
  {
    "ph": "f", "id": 209778, "pid": 76337, "tid": -914061504, "ts": 1716454224723340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723343, "dur": 0,
    "args": {
      "External id": 209779,
      "cbid": 251, "correlation": 209779
    }
  },
  {
    "ph": "f", "id": 209779, "pid": 76337, "tid": -914061504, "ts": 1716454224723343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224778811, "dur": 8,
    "args": {
      "External id": 209780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209780,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209780, "pid": 5, "tid": 7, "ts": 1716454224778811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723345, "dur": 12,
    "args": {
      "External id": 209780,
      "cbid": 211, "correlation": 209780
    }
  },
  {
    "ph": "s", "id": 209780, "pid": 76337, "tid": -914061504, "ts": 1716454224723345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224778821, "dur": 3,
    "args": {
      "External id": 209782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209782,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209782, "pid": 5, "tid": 7, "ts": 1716454224778821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723359, "dur": 6,
    "args": {
      "External id": 209782,
      "cbid": 211, "correlation": 209782
    }
  },
  {
    "ph": "s", "id": 209782, "pid": 76337, "tid": -914061504, "ts": 1716454224723359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224778825, "dur": 56,
    "args": {
      "External id": 209807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209807,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209807, "pid": 5, "tid": 7, "ts": 1716454224778825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723435, "dur": 13,
    "args": {
      "External id": 209807,
      "cbid": 211, "correlation": 209807
    }
  },
  {
    "ph": "s", "id": 209807, "pid": 76337, "tid": -914061504, "ts": 1716454224723435,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723532, "dur": 1,
    "args": {
      "External id": 209825,
      "cbid": 251, "correlation": 209825
    }
  },
  {
    "ph": "f", "id": 209825, "pid": 76337, "tid": -914061504, "ts": 1716454224723532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224778882, "dur": 92,
    "args": {
      "External id": 209827,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209827,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209827, "pid": 5, "tid": 7, "ts": 1716454224778882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723538, "dur": 14,
    "args": {
      "External id": 209827,
      "cbid": 211, "correlation": 209827
    }
  },
  {
    "ph": "s", "id": 209827, "pid": 76337, "tid": -914061504, "ts": 1716454224723538,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224778975, "dur": 10,
    "args": {
      "External id": 209835,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209835,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209835, "pid": 5, "tid": 7, "ts": 1716454224778975,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723608, "dur": 12,
    "args": {
      "External id": 209835,
      "cbid": 211, "correlation": 209835
    }
  },
  {
    "ph": "s", "id": 209835, "pid": 76337, "tid": -914061504, "ts": 1716454224723608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224778986, "dur": 21,
    "args": {
      "External id": 209843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209843,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209843, "pid": 5, "tid": 7, "ts": 1716454224778986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723649, "dur": 9,
    "args": {
      "External id": 209843,
      "cbid": 211, "correlation": 209843
    }
  },
  {
    "ph": "s", "id": 209843, "pid": 76337, "tid": -914061504, "ts": 1716454224723649,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224779009, "dur": 17,
    "args": {
      "External id": 209865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209865,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209865, "pid": 5, "tid": 7, "ts": 1716454224779009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723701, "dur": 10,
    "args": {
      "External id": 209865,
      "cbid": 211, "correlation": 209865
    }
  },
  {
    "ph": "s", "id": 209865, "pid": 76337, "tid": -914061504, "ts": 1716454224723701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723787, "dur": 1,
    "args": {
      "External id": 209881,
      "cbid": 251, "correlation": 209881
    }
  },
  {
    "ph": "f", "id": 209881, "pid": 76337, "tid": -914061504, "ts": 1716454224723787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723792, "dur": 0,
    "args": {
      "External id": 209883,
      "cbid": 251, "correlation": 209883
    }
  },
  {
    "ph": "f", "id": 209883, "pid": 76337, "tid": -914061504, "ts": 1716454224723792,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224779027, "dur": 493,
    "args": {
      "External id": 209884,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209884,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 209884, "pid": 5, "tid": 7, "ts": 1716454224779027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723795, "dur": 12,
    "args": {
      "External id": 209884,
      "cbid": 211, "correlation": 209884
    }
  },
  {
    "ph": "s", "id": 209884, "pid": 76337, "tid": -914061504, "ts": 1716454224723795,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224779522, "dur": 65,
    "args": {
      "External id": 209892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209892,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209892, "pid": 5, "tid": 7, "ts": 1716454224779522,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723860, "dur": 12,
    "args": {
      "External id": 209892,
      "cbid": 211, "correlation": 209892
    }
  },
  {
    "ph": "s", "id": 209892, "pid": 76337, "tid": -914061504, "ts": 1716454224723860,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224779588, "dur": 68,
    "args": {
      "External id": 209900,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209900,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209900, "pid": 5, "tid": 7, "ts": 1716454224779588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723889, "dur": 9,
    "args": {
      "External id": 209900,
      "cbid": 211, "correlation": 209900
    }
  },
  {
    "ph": "s", "id": 209900, "pid": 76337, "tid": -914061504, "ts": 1716454224723889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723971, "dur": 10,
    "args": {
      "External id": 209916,
      "cbid": 251, "correlation": 209916
    }
  },
  {
    "ph": "f", "id": 209916, "pid": 76337, "tid": -914061504, "ts": 1716454224723971,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224779658, "dur": 1,
    "args": {
      "External id": 209918,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 209918,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 209918, "pid": 5, "tid": 7, "ts": 1716454224779658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723985, "dur": 11,
    "args": {
      "External id": 209918,
      "cbid": 51, "correlation": 209918
    }
  },
  {
    "ph": "s", "id": 209918, "pid": 76337, "tid": -914061504, "ts": 1716454224723985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224779662, "dur": 273,
    "args": {
      "External id": 209919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209919,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209919, "pid": 5, "tid": 7, "ts": 1716454224779662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224723997, "dur": 12,
    "args": {
      "External id": 209919,
      "cbid": 211, "correlation": 209919
    }
  },
  {
    "ph": "s", "id": 209919, "pid": 76337, "tid": -914061504, "ts": 1716454224723997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224779936, "dur": 14,
    "args": {
      "External id": 209927,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209927,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209927, "pid": 5, "tid": 7, "ts": 1716454224779936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724041, "dur": 10,
    "args": {
      "External id": 209927,
      "cbid": 211, "correlation": 209927
    }
  },
  {
    "ph": "s", "id": 209927, "pid": 76337, "tid": -914061504, "ts": 1716454224724041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224779951, "dur": 38,
    "args": {
      "External id": 209938,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209938,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209938, "pid": 5, "tid": 7, "ts": 1716454224779951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724109, "dur": 12,
    "args": {
      "External id": 209938,
      "cbid": 211, "correlation": 209938
    }
  },
  {
    "ph": "s", "id": 209938, "pid": 76337, "tid": -914061504, "ts": 1716454224724109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724172, "dur": 0,
    "args": {
      "External id": 209950,
      "cbid": 317, "correlation": 209950
    }
  },
  {
    "ph": "f", "id": 209950, "pid": 76337, "tid": -914061504, "ts": 1716454224724172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724173, "dur": 0,
    "args": {
      "External id": 209951,
      "cbid": 203, "correlation": 209951
    }
  },
  {
    "ph": "f", "id": 209951, "pid": 76337, "tid": -914061504, "ts": 1716454224724173,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724174, "dur": 0,
    "args": {
      "External id": 209952,
      "cbid": 205, "correlation": 209952
    }
  },
  {
    "ph": "f", "id": 209952, "pid": 76337, "tid": -914061504, "ts": 1716454224724174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224779991, "dur": 13,
    "args": {
      "External id": 209956,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209956,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209956, "pid": 5, "tid": 7, "ts": 1716454224779991,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724189, "dur": 13,
    "args": {
      "External id": 209956,
      "cbid": 211, "correlation": 209956
    }
  },
  {
    "ph": "s", "id": 209956, "pid": 76337, "tid": -914061504, "ts": 1716454224724189,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224780006, "dur": 4,
    "args": {
      "External id": 209958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209958,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 209958, "pid": 5, "tid": 7, "ts": 1716454224780006,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724207, "dur": 6,
    "args": {
      "External id": 209958,
      "cbid": 211, "correlation": 209958
    }
  },
  {
    "ph": "s", "id": 209958, "pid": 76337, "tid": -914061504, "ts": 1716454224724207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724215, "dur": 0,
    "args": {
      "External id": 209959,
      "cbid": 51, "correlation": 209959
    }
  },
  {
    "ph": "s", "id": 209959, "pid": 76337, "tid": -914061504, "ts": 1716454224724215,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224780011, "dur": 99,
    "args": {
      "External id": 209960,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209960,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 209960, "pid": 5, "tid": 7, "ts": 1716454224780011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724216, "dur": 5,
    "args": {
      "External id": 209960,
      "cbid": 211, "correlation": 209960
    }
  },
  {
    "ph": "s", "id": 209960, "pid": 76337, "tid": -914061504, "ts": 1716454224724216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224780111, "dur": 17,
    "args": {
      "External id": 209965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209965, "pid": 5, "tid": 7, "ts": 1716454224780111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724244, "dur": 8,
    "args": {
      "External id": 209965,
      "cbid": 211, "correlation": 209965
    }
  },
  {
    "ph": "s", "id": 209965, "pid": 76337, "tid": -914061504, "ts": 1716454224724244,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224780129, "dur": 13,
    "args": {
      "External id": 209973,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209973,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209973, "pid": 5, "tid": 7, "ts": 1716454224780129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724275, "dur": 8,
    "args": {
      "External id": 209973,
      "cbid": 211, "correlation": 209973
    }
  },
  {
    "ph": "s", "id": 209973, "pid": 76337, "tid": -914061504, "ts": 1716454224724275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724344, "dur": 0,
    "args": {
      "External id": 209983,
      "cbid": 317, "correlation": 209983
    }
  },
  {
    "ph": "f", "id": 209983, "pid": 76337, "tid": -914061504, "ts": 1716454224724344,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724345, "dur": 0,
    "args": {
      "External id": 209984,
      "cbid": 203, "correlation": 209984
    }
  },
  {
    "ph": "f", "id": 209984, "pid": 76337, "tid": -914061504, "ts": 1716454224724345,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724346, "dur": 0,
    "args": {
      "External id": 209985,
      "cbid": 205, "correlation": 209985
    }
  },
  {
    "ph": "f", "id": 209985, "pid": 76337, "tid": -914061504, "ts": 1716454224724346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224780144, "dur": 12,
    "args": {
      "External id": 209989,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209989,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209989, "pid": 5, "tid": 7, "ts": 1716454224780144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724360, "dur": 12,
    "args": {
      "External id": 209989,
      "cbid": 211, "correlation": 209989
    }
  },
  {
    "ph": "s", "id": 209989, "pid": 76337, "tid": -914061504, "ts": 1716454224724360,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224780157, "dur": 164,
    "args": {
      "External id": 209991,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209991,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209991, "pid": 5, "tid": 7, "ts": 1716454224780157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724374, "dur": 5,
    "args": {
      "External id": 209991,
      "cbid": 211, "correlation": 209991
    }
  },
  {
    "ph": "s", "id": 209991, "pid": 76337, "tid": -914061504, "ts": 1716454224724374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224780324, "dur": 1,
    "args": {
      "External id": 209993,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 209993,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 209993, "pid": 5, "tid": 7, "ts": 1716454224780324,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724386, "dur": 7,
    "args": {
      "External id": 209993,
      "cbid": 51, "correlation": 209993
    }
  },
  {
    "ph": "s", "id": 209993, "pid": 76337, "tid": -914061504, "ts": 1716454224724386,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224780327, "dur": 199,
    "args": {
      "External id": 209994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209994,
      "registers per thread": 144,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 209994, "pid": 5, "tid": 7, "ts": 1716454224780327,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724395, "dur": 8,
    "args": {
      "External id": 209994,
      "cbid": 211, "correlation": 209994
    }
  },
  {
    "ph": "s", "id": 209994, "pid": 76337, "tid": -914061504, "ts": 1716454224724395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224780528, "dur": 7,
    "args": {
      "External id": 209996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 209996,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 209996, "pid": 5, "tid": 7, "ts": 1716454224780528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724409, "dur": 5,
    "args": {
      "External id": 209996,
      "cbid": 211, "correlation": 209996
    }
  },
  {
    "ph": "s", "id": 209996, "pid": 76337, "tid": -914061504, "ts": 1716454224724409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224780536, "dur": 6,
    "args": {
      "External id": 210002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210002, "pid": 5, "tid": 7, "ts": 1716454224780536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724437, "dur": 8,
    "args": {
      "External id": 210002,
      "cbid": 211, "correlation": 210002
    }
  },
  {
    "ph": "s", "id": 210002, "pid": 76337, "tid": -914061504, "ts": 1716454224724437,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224780543, "dur": 11,
    "args": {
      "External id": 210022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210022,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210022, "pid": 5, "tid": 7, "ts": 1716454224780543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724528, "dur": 12,
    "args": {
      "External id": 210022,
      "cbid": 211, "correlation": 210022
    }
  },
  {
    "ph": "s", "id": 210022, "pid": 76337, "tid": -914061504, "ts": 1716454224724528,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224780556, "dur": 4,
    "args": {
      "External id": 210034,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210034,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210034, "pid": 5, "tid": 7, "ts": 1716454224780556,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724550, "dur": 6,
    "args": {
      "External id": 210034,
      "cbid": 211, "correlation": 210034
    }
  },
  {
    "ph": "s", "id": 210034, "pid": 76337, "tid": -914061504, "ts": 1716454224724550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224780561, "dur": 9,
    "args": {
      "External id": 210037,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210037,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210037, "pid": 5, "tid": 7, "ts": 1716454224780561,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724569, "dur": 6,
    "args": {
      "External id": 210037,
      "cbid": 211, "correlation": 210037
    }
  },
  {
    "ph": "s", "id": 210037, "pid": 76337, "tid": -914061504, "ts": 1716454224724569,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224780571, "dur": 5,
    "args": {
      "External id": 210046,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210046,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210046, "pid": 5, "tid": 7, "ts": 1716454224780571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724609, "dur": 10,
    "args": {
      "External id": 210046,
      "cbid": 211, "correlation": 210046
    }
  },
  {
    "ph": "s", "id": 210046, "pid": 76337, "tid": -914061504, "ts": 1716454224724609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724661, "dur": 0,
    "args": {
      "External id": 210056,
      "cbid": 317, "correlation": 210056
    }
  },
  {
    "ph": "f", "id": 210056, "pid": 76337, "tid": -914061504, "ts": 1716454224724661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724661, "dur": 0,
    "args": {
      "External id": 210057,
      "cbid": 203, "correlation": 210057
    }
  },
  {
    "ph": "f", "id": 210057, "pid": 76337, "tid": -914061504, "ts": 1716454224724661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724662, "dur": 0,
    "args": {
      "External id": 210058,
      "cbid": 205, "correlation": 210058
    }
  },
  {
    "ph": "f", "id": 210058, "pid": 76337, "tid": -914061504, "ts": 1716454224724662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224780578, "dur": 5,
    "args": {
      "External id": 210062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210062,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210062, "pid": 5, "tid": 7, "ts": 1716454224780578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724677, "dur": 12,
    "args": {
      "External id": 210062,
      "cbid": 211, "correlation": 210062
    }
  },
  {
    "ph": "s", "id": 210062, "pid": 76337, "tid": -914061504, "ts": 1716454224724677,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224780584, "dur": 164,
    "args": {
      "External id": 210064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210064,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210064, "pid": 5, "tid": 7, "ts": 1716454224780584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724692, "dur": 5,
    "args": {
      "External id": 210064,
      "cbid": 211, "correlation": 210064
    }
  },
  {
    "ph": "s", "id": 210064, "pid": 76337, "tid": -914061504, "ts": 1716454224724692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224780750, "dur": 1,
    "args": {
      "External id": 210066,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210066,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 210066, "pid": 5, "tid": 7, "ts": 1716454224780750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724703, "dur": 6,
    "args": {
      "External id": 210066,
      "cbid": 51, "correlation": 210066
    }
  },
  {
    "ph": "s", "id": 210066, "pid": 76337, "tid": -914061504, "ts": 1716454224724703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224780753, "dur": 272,
    "args": {
      "External id": 210067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210067,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210067, "pid": 5, "tid": 7, "ts": 1716454224780753,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724711, "dur": 6,
    "args": {
      "External id": 210067,
      "cbid": 211, "correlation": 210067
    }
  },
  {
    "ph": "s", "id": 210067, "pid": 76337, "tid": -914061504, "ts": 1716454224724711,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781027, "dur": 6,
    "args": {
      "External id": 210069,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210069,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210069, "pid": 5, "tid": 7, "ts": 1716454224781027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724720, "dur": 5,
    "args": {
      "External id": 210069,
      "cbid": 211, "correlation": 210069
    }
  },
  {
    "ph": "s", "id": 210069, "pid": 76337, "tid": -914061504, "ts": 1716454224724720,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224781034, "dur": 6,
    "args": {
      "External id": 210075,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210075,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210075, "pid": 5, "tid": 7, "ts": 1716454224781034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724749, "dur": 8,
    "args": {
      "External id": 210075,
      "cbid": 211, "correlation": 210075
    }
  },
  {
    "ph": "s", "id": 210075, "pid": 76337, "tid": -914061504, "ts": 1716454224724749,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224781042, "dur": 3,
    "args": {
      "External id": 210083,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210083,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 210083, "pid": 5, "tid": 7, "ts": 1716454224781042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724792, "dur": 9,
    "args": {
      "External id": 210083,
      "cbid": 211, "correlation": 210083
    }
  },
  {
    "ph": "s", "id": 210083, "pid": 76337, "tid": -914061504, "ts": 1716454224724792,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724857, "dur": 1,
    "args": {
      "External id": 210099,
      "cbid": 251, "correlation": 210099
    }
  },
  {
    "ph": "f", "id": 210099, "pid": 76337, "tid": -914061504, "ts": 1716454224724857,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724862, "dur": 0,
    "args": {
      "External id": 210101,
      "cbid": 251, "correlation": 210101
    }
  },
  {
    "ph": "f", "id": 210101, "pid": 76337, "tid": -914061504, "ts": 1716454224724862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224781046, "dur": 14,
    "args": {
      "External id": 210102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210102,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210102, "pid": 5, "tid": 7, "ts": 1716454224781046,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724864, "dur": 11,
    "args": {
      "External id": 210102,
      "cbid": 211, "correlation": 210102
    }
  },
  {
    "ph": "s", "id": 210102, "pid": 76337, "tid": -914061504, "ts": 1716454224724864,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224781061, "dur": 5,
    "args": {
      "External id": 210104,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210104,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210104, "pid": 5, "tid": 7, "ts": 1716454224781061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724877, "dur": 6,
    "args": {
      "External id": 210104,
      "cbid": 211, "correlation": 210104
    }
  },
  {
    "ph": "s", "id": 210104, "pid": 76337, "tid": -914061504, "ts": 1716454224724877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224781068, "dur": 6,
    "args": {
      "External id": 210114,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210114,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210114, "pid": 5, "tid": 7, "ts": 1716454224781068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224724935, "dur": 12,
    "args": {
      "External id": 210114,
      "cbid": 211, "correlation": 210114
    }
  },
  {
    "ph": "s", "id": 210114, "pid": 76337, "tid": -914061504, "ts": 1716454224724935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781075, "dur": 10,
    "args": {
      "External id": 210134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210134,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210134, "pid": 5, "tid": 7, "ts": 1716454224781075,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725009, "dur": 11,
    "args": {
      "External id": 210134,
      "cbid": 211, "correlation": 210134
    }
  },
  {
    "ph": "s", "id": 210134, "pid": 76337, "tid": -914061504, "ts": 1716454224725009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224781086, "dur": 4,
    "args": {
      "External id": 210146,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210146,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210146, "pid": 5, "tid": 7, "ts": 1716454224781086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725031, "dur": 7,
    "args": {
      "External id": 210146,
      "cbid": 211, "correlation": 210146
    }
  },
  {
    "ph": "s", "id": 210146, "pid": 76337, "tid": -914061504, "ts": 1716454224725031,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224781091, "dur": 7,
    "args": {
      "External id": 210149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210149,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210149, "pid": 5, "tid": 7, "ts": 1716454224781091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725050, "dur": 6,
    "args": {
      "External id": 210149,
      "cbid": 211, "correlation": 210149
    }
  },
  {
    "ph": "s", "id": 210149, "pid": 76337, "tid": -914061504, "ts": 1716454224725050,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224781100, "dur": 5,
    "args": {
      "External id": 210158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210158,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210158, "pid": 5, "tid": 7, "ts": 1716454224781100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725091, "dur": 10,
    "args": {
      "External id": 210158,
      "cbid": 211, "correlation": 210158
    }
  },
  {
    "ph": "s", "id": 210158, "pid": 76337, "tid": -914061504, "ts": 1716454224725091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725154, "dur": 0,
    "args": {
      "External id": 210168,
      "cbid": 317, "correlation": 210168
    }
  },
  {
    "ph": "f", "id": 210168, "pid": 76337, "tid": -914061504, "ts": 1716454224725154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725154, "dur": 0,
    "args": {
      "External id": 210169,
      "cbid": 203, "correlation": 210169
    }
  },
  {
    "ph": "f", "id": 210169, "pid": 76337, "tid": -914061504, "ts": 1716454224725154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725155, "dur": 0,
    "args": {
      "External id": 210170,
      "cbid": 205, "correlation": 210170
    }
  },
  {
    "ph": "f", "id": 210170, "pid": 76337, "tid": -914061504, "ts": 1716454224725155,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781106, "dur": 5,
    "args": {
      "External id": 210174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210174,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210174, "pid": 5, "tid": 7, "ts": 1716454224781106,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725169, "dur": 12,
    "args": {
      "External id": 210174,
      "cbid": 211, "correlation": 210174
    }
  },
  {
    "ph": "s", "id": 210174, "pid": 76337, "tid": -914061504, "ts": 1716454224725169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781112, "dur": 163,
    "args": {
      "External id": 210176,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210176,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210176, "pid": 5, "tid": 7, "ts": 1716454224781112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725184, "dur": 6,
    "args": {
      "External id": 210176,
      "cbid": 211, "correlation": 210176
    }
  },
  {
    "ph": "s", "id": 210176, "pid": 76337, "tid": -914061504, "ts": 1716454224725184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224781277, "dur": 1,
    "args": {
      "External id": 210178,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210178,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 210178, "pid": 5, "tid": 7, "ts": 1716454224781277,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725195, "dur": 6,
    "args": {
      "External id": 210178,
      "cbid": 51, "correlation": 210178
    }
  },
  {
    "ph": "s", "id": 210178, "pid": 76337, "tid": -914061504, "ts": 1716454224725195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224781281, "dur": 261,
    "args": {
      "External id": 210179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210179,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210179, "pid": 5, "tid": 7, "ts": 1716454224781281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725202, "dur": 6,
    "args": {
      "External id": 210179,
      "cbid": 211, "correlation": 210179
    }
  },
  {
    "ph": "s", "id": 210179, "pid": 76337, "tid": -914061504, "ts": 1716454224725202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781543, "dur": 6,
    "args": {
      "External id": 210181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210181,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210181, "pid": 5, "tid": 7, "ts": 1716454224781543,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725212, "dur": 5,
    "args": {
      "External id": 210181,
      "cbid": 211, "correlation": 210181
    }
  },
  {
    "ph": "s", "id": 210181, "pid": 76337, "tid": -914061504, "ts": 1716454224725212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224781551, "dur": 6,
    "args": {
      "External id": 210187,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210187,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210187, "pid": 5, "tid": 7, "ts": 1716454224781551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725241, "dur": 8,
    "args": {
      "External id": 210187,
      "cbid": 211, "correlation": 210187
    }
  },
  {
    "ph": "s", "id": 210187, "pid": 76337, "tid": -914061504, "ts": 1716454224725241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224781558, "dur": 5,
    "args": {
      "External id": 210195,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210195,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210195, "pid": 5, "tid": 7, "ts": 1716454224781558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725273, "dur": 8,
    "args": {
      "External id": 210195,
      "cbid": 211, "correlation": 210195
    }
  },
  {
    "ph": "s", "id": 210195, "pid": 76337, "tid": -914061504, "ts": 1716454224725273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224781565, "dur": 4,
    "args": {
      "External id": 210203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210203,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210203, "pid": 5, "tid": 7, "ts": 1716454224781565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725303, "dur": 8,
    "args": {
      "External id": 210203,
      "cbid": 211, "correlation": 210203
    }
  },
  {
    "ph": "s", "id": 210203, "pid": 76337, "tid": -914061504, "ts": 1716454224725303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781570, "dur": 10,
    "args": {
      "External id": 210223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210223,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210223, "pid": 5, "tid": 7, "ts": 1716454224781570,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725376, "dur": 12,
    "args": {
      "External id": 210223,
      "cbid": 211, "correlation": 210223
    }
  },
  {
    "ph": "s", "id": 210223, "pid": 76337, "tid": -914061504, "ts": 1716454224725376,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224781582, "dur": 4,
    "args": {
      "External id": 210235,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210235,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210235, "pid": 5, "tid": 7, "ts": 1716454224781582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725398, "dur": 6,
    "args": {
      "External id": 210235,
      "cbid": 211, "correlation": 210235
    }
  },
  {
    "ph": "s", "id": 210235, "pid": 76337, "tid": -914061504, "ts": 1716454224725398,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224781587, "dur": 7,
    "args": {
      "External id": 210238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210238, "pid": 5, "tid": 7, "ts": 1716454224781587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725416, "dur": 6,
    "args": {
      "External id": 210238,
      "cbid": 211, "correlation": 210238
    }
  },
  {
    "ph": "s", "id": 210238, "pid": 76337, "tid": -914061504, "ts": 1716454224725416,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224781595, "dur": 5,
    "args": {
      "External id": 210247,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210247,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210247, "pid": 5, "tid": 7, "ts": 1716454224781595,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725454, "dur": 9,
    "args": {
      "External id": 210247,
      "cbid": 211, "correlation": 210247
    }
  },
  {
    "ph": "s", "id": 210247, "pid": 76337, "tid": -914061504, "ts": 1716454224725454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725504, "dur": 0,
    "args": {
      "External id": 210257,
      "cbid": 317, "correlation": 210257
    }
  },
  {
    "ph": "f", "id": 210257, "pid": 76337, "tid": -914061504, "ts": 1716454224725504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725505, "dur": 0,
    "args": {
      "External id": 210258,
      "cbid": 203, "correlation": 210258
    }
  },
  {
    "ph": "f", "id": 210258, "pid": 76337, "tid": -914061504, "ts": 1716454224725505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725505, "dur": 0,
    "args": {
      "External id": 210259,
      "cbid": 205, "correlation": 210259
    }
  },
  {
    "ph": "f", "id": 210259, "pid": 76337, "tid": -914061504, "ts": 1716454224725505,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781601, "dur": 5,
    "args": {
      "External id": 210263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210263,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210263, "pid": 5, "tid": 7, "ts": 1716454224781601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725518, "dur": 11,
    "args": {
      "External id": 210263,
      "cbid": 211, "correlation": 210263
    }
  },
  {
    "ph": "s", "id": 210263, "pid": 76337, "tid": -914061504, "ts": 1716454224725518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224781607, "dur": 163,
    "args": {
      "External id": 210265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210265,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210265, "pid": 5, "tid": 7, "ts": 1716454224781607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725532, "dur": 6,
    "args": {
      "External id": 210265,
      "cbid": 211, "correlation": 210265
    }
  },
  {
    "ph": "s", "id": 210265, "pid": 76337, "tid": -914061504, "ts": 1716454224725532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224781772, "dur": 1,
    "args": {
      "External id": 210267,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210267,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 210267, "pid": 5, "tid": 7, "ts": 1716454224781772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725543, "dur": 6,
    "args": {
      "External id": 210267,
      "cbid": 51, "correlation": 210267
    }
  },
  {
    "ph": "s", "id": 210267, "pid": 76337, "tid": -914061504, "ts": 1716454224725543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224781776, "dur": 260,
    "args": {
      "External id": 210268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210268,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210268, "pid": 5, "tid": 7, "ts": 1716454224781776,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725550, "dur": 6,
    "args": {
      "External id": 210268,
      "cbid": 211, "correlation": 210268
    }
  },
  {
    "ph": "s", "id": 210268, "pid": 76337, "tid": -914061504, "ts": 1716454224725550,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782037, "dur": 6,
    "args": {
      "External id": 210270,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210270,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210270, "pid": 5, "tid": 7, "ts": 1716454224782037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725560, "dur": 5,
    "args": {
      "External id": 210270,
      "cbid": 211, "correlation": 210270
    }
  },
  {
    "ph": "s", "id": 210270, "pid": 76337, "tid": -914061504, "ts": 1716454224725560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224782044, "dur": 6,
    "args": {
      "External id": 210276,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210276,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210276, "pid": 5, "tid": 7, "ts": 1716454224782044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725589, "dur": 9,
    "args": {
      "External id": 210276,
      "cbid": 211, "correlation": 210276
    }
  },
  {
    "ph": "s", "id": 210276, "pid": 76337, "tid": -914061504, "ts": 1716454224725589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224782052, "dur": 3,
    "args": {
      "External id": 210284,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210284,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 210284, "pid": 5, "tid": 7, "ts": 1716454224782052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725633, "dur": 10,
    "args": {
      "External id": 210284,
      "cbid": 211, "correlation": 210284
    }
  },
  {
    "ph": "s", "id": 210284, "pid": 76337, "tid": -914061504, "ts": 1716454224725633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725695, "dur": 1,
    "args": {
      "External id": 210300,
      "cbid": 251, "correlation": 210300
    }
  },
  {
    "ph": "f", "id": 210300, "pid": 76337, "tid": -914061504, "ts": 1716454224725695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725699, "dur": 0,
    "args": {
      "External id": 210302,
      "cbid": 251, "correlation": 210302
    }
  },
  {
    "ph": "f", "id": 210302, "pid": 76337, "tid": -914061504, "ts": 1716454224725699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224782056, "dur": 10,
    "args": {
      "External id": 210303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210303,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210303, "pid": 5, "tid": 7, "ts": 1716454224782056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725701, "dur": 10,
    "args": {
      "External id": 210303,
      "cbid": 211, "correlation": 210303
    }
  },
  {
    "ph": "s", "id": 210303, "pid": 76337, "tid": -914061504, "ts": 1716454224725701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224782068, "dur": 4,
    "args": {
      "External id": 210305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210305,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210305, "pid": 5, "tid": 7, "ts": 1716454224782068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725714, "dur": 5,
    "args": {
      "External id": 210305,
      "cbid": 211, "correlation": 210305
    }
  },
  {
    "ph": "s", "id": 210305, "pid": 76337, "tid": -914061504, "ts": 1716454224725714,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224782073, "dur": 6,
    "args": {
      "External id": 210315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210315, "pid": 5, "tid": 7, "ts": 1716454224782073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725770, "dur": 12,
    "args": {
      "External id": 210315,
      "cbid": 211, "correlation": 210315
    }
  },
  {
    "ph": "s", "id": 210315, "pid": 76337, "tid": -914061504, "ts": 1716454224725770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782080, "dur": 10,
    "args": {
      "External id": 210335,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210335,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210335, "pid": 5, "tid": 7, "ts": 1716454224782080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725835, "dur": 10,
    "args": {
      "External id": 210335,
      "cbid": 211, "correlation": 210335
    }
  },
  {
    "ph": "s", "id": 210335, "pid": 76337, "tid": -914061504, "ts": 1716454224725835,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224782091, "dur": 4,
    "args": {
      "External id": 210347,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210347,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210347, "pid": 5, "tid": 7, "ts": 1716454224782091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725855, "dur": 6,
    "args": {
      "External id": 210347,
      "cbid": 211, "correlation": 210347
    }
  },
  {
    "ph": "s", "id": 210347, "pid": 76337, "tid": -914061504, "ts": 1716454224725855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224782096, "dur": 7,
    "args": {
      "External id": 210350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210350, "pid": 5, "tid": 7, "ts": 1716454224782096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725872, "dur": 6,
    "args": {
      "External id": 210350,
      "cbid": 211, "correlation": 210350
    }
  },
  {
    "ph": "s", "id": 210350, "pid": 76337, "tid": -914061504, "ts": 1716454224725872,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224782104, "dur": 5,
    "args": {
      "External id": 210359,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210359,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210359, "pid": 5, "tid": 7, "ts": 1716454224782104,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725912, "dur": 10,
    "args": {
      "External id": 210359,
      "cbid": 211, "correlation": 210359
    }
  },
  {
    "ph": "s", "id": 210359, "pid": 76337, "tid": -914061504, "ts": 1716454224725912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725981, "dur": 0,
    "args": {
      "External id": 210369,
      "cbid": 317, "correlation": 210369
    }
  },
  {
    "ph": "f", "id": 210369, "pid": 76337, "tid": -914061504, "ts": 1716454224725981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725982, "dur": 0,
    "args": {
      "External id": 210370,
      "cbid": 203, "correlation": 210370
    }
  },
  {
    "ph": "f", "id": 210370, "pid": 76337, "tid": -914061504, "ts": 1716454224725982,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725983, "dur": 0,
    "args": {
      "External id": 210371,
      "cbid": 205, "correlation": 210371
    }
  },
  {
    "ph": "f", "id": 210371, "pid": 76337, "tid": -914061504, "ts": 1716454224725983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782110, "dur": 5,
    "args": {
      "External id": 210375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210375,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210375, "pid": 5, "tid": 7, "ts": 1716454224782110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224725997, "dur": 12,
    "args": {
      "External id": 210375,
      "cbid": 211, "correlation": 210375
    }
  },
  {
    "ph": "s", "id": 210375, "pid": 76337, "tid": -914061504, "ts": 1716454224725997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782116, "dur": 162,
    "args": {
      "External id": 210377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210377,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210377, "pid": 5, "tid": 7, "ts": 1716454224782116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726012, "dur": 5,
    "args": {
      "External id": 210377,
      "cbid": 211, "correlation": 210377
    }
  },
  {
    "ph": "s", "id": 210377, "pid": 76337, "tid": -914061504, "ts": 1716454224726012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224782281, "dur": 1,
    "args": {
      "External id": 210379,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210379,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 210379, "pid": 5, "tid": 7, "ts": 1716454224782281,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726022, "dur": 6,
    "args": {
      "External id": 210379,
      "cbid": 51, "correlation": 210379
    }
  },
  {
    "ph": "s", "id": 210379, "pid": 76337, "tid": -914061504, "ts": 1716454224726022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224782285, "dur": 261,
    "args": {
      "External id": 210380,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210380,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210380, "pid": 5, "tid": 7, "ts": 1716454224782285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726029, "dur": 7,
    "args": {
      "External id": 210380,
      "cbid": 211, "correlation": 210380
    }
  },
  {
    "ph": "s", "id": 210380, "pid": 76337, "tid": -914061504, "ts": 1716454224726029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782546, "dur": 6,
    "args": {
      "External id": 210382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210382,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210382, "pid": 5, "tid": 7, "ts": 1716454224782546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726040, "dur": 5,
    "args": {
      "External id": 210382,
      "cbid": 211, "correlation": 210382
    }
  },
  {
    "ph": "s", "id": 210382, "pid": 76337, "tid": -914061504, "ts": 1716454224726040,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224782554, "dur": 6,
    "args": {
      "External id": 210388,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210388,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210388, "pid": 5, "tid": 7, "ts": 1716454224782554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726068, "dur": 8,
    "args": {
      "External id": 210388,
      "cbid": 211, "correlation": 210388
    }
  },
  {
    "ph": "s", "id": 210388, "pid": 76337, "tid": -914061504, "ts": 1716454224726068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224782562, "dur": 5,
    "args": {
      "External id": 210396,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210396,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210396, "pid": 5, "tid": 7, "ts": 1716454224782562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726101, "dur": 8,
    "args": {
      "External id": 210396,
      "cbid": 211, "correlation": 210396
    }
  },
  {
    "ph": "s", "id": 210396, "pid": 76337, "tid": -914061504, "ts": 1716454224726101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224782568, "dur": 4,
    "args": {
      "External id": 210404,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210404,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210404, "pid": 5, "tid": 7, "ts": 1716454224782568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726129, "dur": 9,
    "args": {
      "External id": 210404,
      "cbid": 211, "correlation": 210404
    }
  },
  {
    "ph": "s", "id": 210404, "pid": 76337, "tid": -914061504, "ts": 1716454224726129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782574, "dur": 10,
    "args": {
      "External id": 210424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210424,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210424, "pid": 5, "tid": 7, "ts": 1716454224782574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726232, "dur": 13,
    "args": {
      "External id": 210424,
      "cbid": 211, "correlation": 210424
    }
  },
  {
    "ph": "s", "id": 210424, "pid": 76337, "tid": -914061504, "ts": 1716454224726232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224782585, "dur": 4,
    "args": {
      "External id": 210436,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210436,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210436, "pid": 5, "tid": 7, "ts": 1716454224782585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726255, "dur": 6,
    "args": {
      "External id": 210436,
      "cbid": 211, "correlation": 210436
    }
  },
  {
    "ph": "s", "id": 210436, "pid": 76337, "tid": -914061504, "ts": 1716454224726255,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224782590, "dur": 6,
    "args": {
      "External id": 210439,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210439,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210439, "pid": 5, "tid": 7, "ts": 1716454224782590,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726273, "dur": 6,
    "args": {
      "External id": 210439,
      "cbid": 211, "correlation": 210439
    }
  },
  {
    "ph": "s", "id": 210439, "pid": 76337, "tid": -914061504, "ts": 1716454224726273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224782598, "dur": 5,
    "args": {
      "External id": 210448,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210448,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210448, "pid": 5, "tid": 7, "ts": 1716454224782598,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726311, "dur": 10,
    "args": {
      "External id": 210448,
      "cbid": 211, "correlation": 210448
    }
  },
  {
    "ph": "s", "id": 210448, "pid": 76337, "tid": -914061504, "ts": 1716454224726311,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726364, "dur": 0,
    "args": {
      "External id": 210458,
      "cbid": 317, "correlation": 210458
    }
  },
  {
    "ph": "f", "id": 210458, "pid": 76337, "tid": -914061504, "ts": 1716454224726364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726365, "dur": 0,
    "args": {
      "External id": 210459,
      "cbid": 203, "correlation": 210459
    }
  },
  {
    "ph": "f", "id": 210459, "pid": 76337, "tid": -914061504, "ts": 1716454224726365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726365, "dur": 0,
    "args": {
      "External id": 210460,
      "cbid": 205, "correlation": 210460
    }
  },
  {
    "ph": "f", "id": 210460, "pid": 76337, "tid": -914061504, "ts": 1716454224726365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782604, "dur": 5,
    "args": {
      "External id": 210464,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210464,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210464, "pid": 5, "tid": 7, "ts": 1716454224782604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726378, "dur": 12,
    "args": {
      "External id": 210464,
      "cbid": 211, "correlation": 210464
    }
  },
  {
    "ph": "s", "id": 210464, "pid": 76337, "tid": -914061504, "ts": 1716454224726378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224782610, "dur": 164,
    "args": {
      "External id": 210466,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210466,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210466, "pid": 5, "tid": 7, "ts": 1716454224782610,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726393, "dur": 5,
    "args": {
      "External id": 210466,
      "cbid": 211, "correlation": 210466
    }
  },
  {
    "ph": "s", "id": 210466, "pid": 76337, "tid": -914061504, "ts": 1716454224726393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224782777, "dur": 1,
    "args": {
      "External id": 210468,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210468,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 210468, "pid": 5, "tid": 7, "ts": 1716454224782777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726403, "dur": 6,
    "args": {
      "External id": 210468,
      "cbid": 51, "correlation": 210468
    }
  },
  {
    "ph": "s", "id": 210468, "pid": 76337, "tid": -914061504, "ts": 1716454224726403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224782780, "dur": 260,
    "args": {
      "External id": 210469,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210469,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210469, "pid": 5, "tid": 7, "ts": 1716454224782780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726410, "dur": 6,
    "args": {
      "External id": 210469,
      "cbid": 211, "correlation": 210469
    }
  },
  {
    "ph": "s", "id": 210469, "pid": 76337, "tid": -914061504, "ts": 1716454224726410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783042, "dur": 6,
    "args": {
      "External id": 210471,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210471,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210471, "pid": 5, "tid": 7, "ts": 1716454224783042,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726420, "dur": 5,
    "args": {
      "External id": 210471,
      "cbid": 211, "correlation": 210471
    }
  },
  {
    "ph": "s", "id": 210471, "pid": 76337, "tid": -914061504, "ts": 1716454224726420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224783049, "dur": 6,
    "args": {
      "External id": 210477,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210477,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210477, "pid": 5, "tid": 7, "ts": 1716454224783049,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726448, "dur": 8,
    "args": {
      "External id": 210477,
      "cbid": 211, "correlation": 210477
    }
  },
  {
    "ph": "s", "id": 210477, "pid": 76337, "tid": -914061504, "ts": 1716454224726448,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224783056, "dur": 3,
    "args": {
      "External id": 210485,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210485,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 210485, "pid": 5, "tid": 7, "ts": 1716454224783056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726492, "dur": 10,
    "args": {
      "External id": 210485,
      "cbid": 211, "correlation": 210485
    }
  },
  {
    "ph": "s", "id": 210485, "pid": 76337, "tid": -914061504, "ts": 1716454224726492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726553, "dur": 1,
    "args": {
      "External id": 210501,
      "cbid": 251, "correlation": 210501
    }
  },
  {
    "ph": "f", "id": 210501, "pid": 76337, "tid": -914061504, "ts": 1716454224726553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726559, "dur": 0,
    "args": {
      "External id": 210503,
      "cbid": 251, "correlation": 210503
    }
  },
  {
    "ph": "f", "id": 210503, "pid": 76337, "tid": -914061504, "ts": 1716454224726559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224783061, "dur": 10,
    "args": {
      "External id": 210504,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210504,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210504, "pid": 5, "tid": 7, "ts": 1716454224783061,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726560, "dur": 11,
    "args": {
      "External id": 210504,
      "cbid": 211, "correlation": 210504
    }
  },
  {
    "ph": "s", "id": 210504, "pid": 76337, "tid": -914061504, "ts": 1716454224726560,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224783072, "dur": 4,
    "args": {
      "External id": 210506,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210506,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210506, "pid": 5, "tid": 7, "ts": 1716454224783072,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726573, "dur": 6,
    "args": {
      "External id": 210506,
      "cbid": 211, "correlation": 210506
    }
  },
  {
    "ph": "s", "id": 210506, "pid": 76337, "tid": -914061504, "ts": 1716454224726573,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224783077, "dur": 6,
    "args": {
      "External id": 210516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210516,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210516, "pid": 5, "tid": 7, "ts": 1716454224783077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726630, "dur": 12,
    "args": {
      "External id": 210516,
      "cbid": 211, "correlation": 210516
    }
  },
  {
    "ph": "s", "id": 210516, "pid": 76337, "tid": -914061504, "ts": 1716454224726630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783084, "dur": 9,
    "args": {
      "External id": 210536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210536,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210536, "pid": 5, "tid": 7, "ts": 1716454224783084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726695, "dur": 11,
    "args": {
      "External id": 210536,
      "cbid": 211, "correlation": 210536
    }
  },
  {
    "ph": "s", "id": 210536, "pid": 76337, "tid": -914061504, "ts": 1716454224726695,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224783095, "dur": 4,
    "args": {
      "External id": 210548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210548,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210548, "pid": 5, "tid": 7, "ts": 1716454224783095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726716, "dur": 6,
    "args": {
      "External id": 210548,
      "cbid": 211, "correlation": 210548
    }
  },
  {
    "ph": "s", "id": 210548, "pid": 76337, "tid": -914061504, "ts": 1716454224726716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224783100, "dur": 7,
    "args": {
      "External id": 210551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210551,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210551, "pid": 5, "tid": 7, "ts": 1716454224783100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726735, "dur": 6,
    "args": {
      "External id": 210551,
      "cbid": 211, "correlation": 210551
    }
  },
  {
    "ph": "s", "id": 210551, "pid": 76337, "tid": -914061504, "ts": 1716454224726735,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224783108, "dur": 5,
    "args": {
      "External id": 210560,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210560,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210560, "pid": 5, "tid": 7, "ts": 1716454224783108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726774, "dur": 11,
    "args": {
      "External id": 210560,
      "cbid": 211, "correlation": 210560
    }
  },
  {
    "ph": "s", "id": 210560, "pid": 76337, "tid": -914061504, "ts": 1716454224726774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726838, "dur": 0,
    "args": {
      "External id": 210570,
      "cbid": 317, "correlation": 210570
    }
  },
  {
    "ph": "f", "id": 210570, "pid": 76337, "tid": -914061504, "ts": 1716454224726838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726838, "dur": 0,
    "args": {
      "External id": 210571,
      "cbid": 203, "correlation": 210571
    }
  },
  {
    "ph": "f", "id": 210571, "pid": 76337, "tid": -914061504, "ts": 1716454224726838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726839, "dur": 0,
    "args": {
      "External id": 210572,
      "cbid": 205, "correlation": 210572
    }
  },
  {
    "ph": "f", "id": 210572, "pid": 76337, "tid": -914061504, "ts": 1716454224726839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783114, "dur": 5,
    "args": {
      "External id": 210576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210576,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210576, "pid": 5, "tid": 7, "ts": 1716454224783114,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726853, "dur": 12,
    "args": {
      "External id": 210576,
      "cbid": 211, "correlation": 210576
    }
  },
  {
    "ph": "s", "id": 210576, "pid": 76337, "tid": -914061504, "ts": 1716454224726853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783120, "dur": 163,
    "args": {
      "External id": 210578,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210578,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210578, "pid": 5, "tid": 7, "ts": 1716454224783120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726867, "dur": 5,
    "args": {
      "External id": 210578,
      "cbid": 211, "correlation": 210578
    }
  },
  {
    "ph": "s", "id": 210578, "pid": 76337, "tid": -914061504, "ts": 1716454224726867,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224783285, "dur": 1,
    "args": {
      "External id": 210580,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210580,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 210580, "pid": 5, "tid": 7, "ts": 1716454224783285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726878, "dur": 7,
    "args": {
      "External id": 210580,
      "cbid": 51, "correlation": 210580
    }
  },
  {
    "ph": "s", "id": 210580, "pid": 76337, "tid": -914061504, "ts": 1716454224726878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224783289, "dur": 261,
    "args": {
      "External id": 210581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210581,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210581, "pid": 5, "tid": 7, "ts": 1716454224783289,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726886, "dur": 6,
    "args": {
      "External id": 210581,
      "cbid": 211, "correlation": 210581
    }
  },
  {
    "ph": "s", "id": 210581, "pid": 76337, "tid": -914061504, "ts": 1716454224726886,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783551, "dur": 6,
    "args": {
      "External id": 210583,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210583,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210583, "pid": 5, "tid": 7, "ts": 1716454224783551,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726897, "dur": 5,
    "args": {
      "External id": 210583,
      "cbid": 211, "correlation": 210583
    }
  },
  {
    "ph": "s", "id": 210583, "pid": 76337, "tid": -914061504, "ts": 1716454224726897,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224783558, "dur": 6,
    "args": {
      "External id": 210589,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210589,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210589, "pid": 5, "tid": 7, "ts": 1716454224783558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726925, "dur": 9,
    "args": {
      "External id": 210589,
      "cbid": 211, "correlation": 210589
    }
  },
  {
    "ph": "s", "id": 210589, "pid": 76337, "tid": -914061504, "ts": 1716454224726925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224783566, "dur": 5,
    "args": {
      "External id": 210597,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210597,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210597, "pid": 5, "tid": 7, "ts": 1716454224783566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726958, "dur": 8,
    "args": {
      "External id": 210597,
      "cbid": 211, "correlation": 210597
    }
  },
  {
    "ph": "s", "id": 210597, "pid": 76337, "tid": -914061504, "ts": 1716454224726958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224783572, "dur": 4,
    "args": {
      "External id": 210605,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210605,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210605, "pid": 5, "tid": 7, "ts": 1716454224783572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224726996, "dur": 9,
    "args": {
      "External id": 210605,
      "cbid": 211, "correlation": 210605
    }
  },
  {
    "ph": "s", "id": 210605, "pid": 76337, "tid": -914061504, "ts": 1716454224726996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783578, "dur": 10,
    "args": {
      "External id": 210625,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210625,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 210625, "pid": 5, "tid": 7, "ts": 1716454224783578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727138, "dur": 13,
    "args": {
      "External id": 210625,
      "cbid": 211, "correlation": 210625
    }
  },
  {
    "ph": "s", "id": 210625, "pid": 76337, "tid": -914061504, "ts": 1716454224727138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224783589, "dur": 4,
    "args": {
      "External id": 210637,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210637,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 210637, "pid": 5, "tid": 7, "ts": 1716454224783589,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727161, "dur": 7,
    "args": {
      "External id": 210637,
      "cbid": 211, "correlation": 210637
    }
  },
  {
    "ph": "s", "id": 210637, "pid": 76337, "tid": -914061504, "ts": 1716454224727161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224783594, "dur": 7,
    "args": {
      "External id": 210640,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210640,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210640, "pid": 5, "tid": 7, "ts": 1716454224783594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727180, "dur": 7,
    "args": {
      "External id": 210640,
      "cbid": 211, "correlation": 210640
    }
  },
  {
    "ph": "s", "id": 210640, "pid": 76337, "tid": -914061504, "ts": 1716454224727180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727239, "dur": 0,
    "args": {
      "External id": 210651,
      "cbid": 317, "correlation": 210651
    }
  },
  {
    "ph": "f", "id": 210651, "pid": 76337, "tid": -914061504, "ts": 1716454224727239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727240, "dur": 0,
    "args": {
      "External id": 210652,
      "cbid": 203, "correlation": 210652
    }
  },
  {
    "ph": "f", "id": 210652, "pid": 76337, "tid": -914061504, "ts": 1716454224727240,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727241, "dur": 0,
    "args": {
      "External id": 210653,
      "cbid": 205, "correlation": 210653
    }
  },
  {
    "ph": "f", "id": 210653, "pid": 76337, "tid": -914061504, "ts": 1716454224727241,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783602, "dur": 5,
    "args": {
      "External id": 210657,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210657,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210657, "pid": 5, "tid": 7, "ts": 1716454224783602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727258, "dur": 12,
    "args": {
      "External id": 210657,
      "cbid": 211, "correlation": 210657
    }
  },
  {
    "ph": "s", "id": 210657, "pid": 76337, "tid": -914061504, "ts": 1716454224727258,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224783608, "dur": 37,
    "args": {
      "External id": 210659,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210659,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 210659, "pid": 5, "tid": 7, "ts": 1716454224783608,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727277, "dur": 9,
    "args": {
      "External id": 210659,
      "cbid": 211, "correlation": 210659
    }
  },
  {
    "ph": "s", "id": 210659, "pid": 76337, "tid": -914061504, "ts": 1716454224727277,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783647, "dur": 5,
    "args": {
      "External id": 210661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210661,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210661, "pid": 5, "tid": 7, "ts": 1716454224783647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727290, "dur": 5,
    "args": {
      "External id": 210661,
      "cbid": 211, "correlation": 210661
    }
  },
  {
    "ph": "s", "id": 210661, "pid": 76337, "tid": -914061504, "ts": 1716454224727290,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224783653, "dur": 6,
    "args": {
      "External id": 210667,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210667,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210667, "pid": 5, "tid": 7, "ts": 1716454224783653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727318, "dur": 9,
    "args": {
      "External id": 210667,
      "cbid": 211, "correlation": 210667
    }
  },
  {
    "ph": "s", "id": 210667, "pid": 76337, "tid": -914061504, "ts": 1716454224727318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224783661, "dur": 20,
    "args": {
      "External id": 210676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210676,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210676, "pid": 5, "tid": 7, "ts": 1716454224783661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727400, "dur": 14,
    "args": {
      "External id": 210676,
      "cbid": 211, "correlation": 210676
    }
  },
  {
    "ph": "s", "id": 210676, "pid": 76337, "tid": -914061504, "ts": 1716454224727400,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224783683, "dur": 11,
    "args": {
      "External id": 210698,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210698,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [768, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 60
    }
  },
  {
    "ph": "f", "id": 210698, "pid": 5, "tid": 7, "ts": 1716454224783683,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727457, "dur": 11,
    "args": {
      "External id": 210698,
      "cbid": 211, "correlation": 210698
    }
  },
  {
    "ph": "s", "id": 210698, "pid": 76337, "tid": -914061504, "ts": 1716454224727457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727547, "dur": 2,
    "args": {
      "External id": 210709,
      "cbid": 251, "correlation": 210709
    }
  },
  {
    "ph": "f", "id": 210709, "pid": 76337, "tid": -914061504, "ts": 1716454224727547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727552, "dur": 0,
    "args": {
      "External id": 210710,
      "cbid": 251, "correlation": 210710
    }
  },
  {
    "ph": "f", "id": 210710, "pid": 76337, "tid": -914061504, "ts": 1716454224727552,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224783695, "dur": 55,
    "args": {
      "External id": 210711,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210711,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 210711, "pid": 5, "tid": 7, "ts": 1716454224783695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727555, "dur": 14,
    "args": {
      "External id": 210711,
      "cbid": 211, "correlation": 210711
    }
  },
  {
    "ph": "s", "id": 210711, "pid": 76337, "tid": -914061504, "ts": 1716454224727555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727626, "dur": 1,
    "args": {
      "External id": 210722,
      "cbid": 251, "correlation": 210722
    }
  },
  {
    "ph": "f", "id": 210722, "pid": 76337, "tid": -914061504, "ts": 1716454224727626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727629, "dur": 0,
    "args": {
      "External id": 210723,
      "cbid": 251, "correlation": 210723
    }
  },
  {
    "ph": "f", "id": 210723, "pid": 76337, "tid": -914061504, "ts": 1716454224727629,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224783751, "dur": 53,
    "args": {
      "External id": 210724,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210724,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 210724, "pid": 5, "tid": 7, "ts": 1716454224783751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727631, "dur": 11,
    "args": {
      "External id": 210724,
      "cbid": 211, "correlation": 210724
    }
  },
  {
    "ph": "s", "id": 210724, "pid": 76337, "tid": -914061504, "ts": 1716454224727631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727695, "dur": 1,
    "args": {
      "External id": 210735,
      "cbid": 251, "correlation": 210735
    }
  },
  {
    "ph": "f", "id": 210735, "pid": 76337, "tid": -914061504, "ts": 1716454224727695,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727699, "dur": 0,
    "args": {
      "External id": 210736,
      "cbid": 251, "correlation": 210736
    }
  },
  {
    "ph": "f", "id": 210736, "pid": 76337, "tid": -914061504, "ts": 1716454224727699,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224783806, "dur": 54,
    "args": {
      "External id": 210737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210737,
      "registers per thread": 186,
      "shared memory": 32768,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [10, 12, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 210737, "pid": 5, "tid": 7, "ts": 1716454224783806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727701, "dur": 12,
    "args": {
      "External id": 210737,
      "cbid": 211, "correlation": 210737
    }
  },
  {
    "ph": "s", "id": 210737, "pid": 76337, "tid": -914061504, "ts": 1716454224727701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224783861, "dur": 58,
    "args": {
      "External id": 210762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210762,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 2.4,
      "warps per SM": 9.6,
      "grid": [3, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210762, "pid": 5, "tid": 7, "ts": 1716454224783861,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727785, "dur": 12,
    "args": {
      "External id": 210762,
      "cbid": 211, "correlation": 210762
    }
  },
  {
    "ph": "s", "id": 210762, "pid": 76337, "tid": -914061504, "ts": 1716454224727785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727883, "dur": 1,
    "args": {
      "External id": 210780,
      "cbid": 251, "correlation": 210780
    }
  },
  {
    "ph": "f", "id": 210780, "pid": 76337, "tid": -914061504, "ts": 1716454224727883,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224783921, "dur": 64,
    "args": {
      "External id": 210782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210782,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 0.75,
      "warps per SM": 3,
      "grid": [10, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 210782, "pid": 5, "tid": 7, "ts": 1716454224783921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727888, "dur": 13,
    "args": {
      "External id": 210782,
      "cbid": 211, "correlation": 210782
    }
  },
  {
    "ph": "s", "id": 210782, "pid": 76337, "tid": -914061504, "ts": 1716454224727888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224783986, "dur": 6,
    "args": {
      "External id": 210790,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210790,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210790, "pid": 5, "tid": 7, "ts": 1716454224783986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224727958, "dur": 13,
    "args": {
      "External id": 210790,
      "cbid": 211, "correlation": 210790
    }
  },
  {
    "ph": "s", "id": 210790, "pid": 76337, "tid": -914061504, "ts": 1716454224727958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224783993, "dur": 7,
    "args": {
      "External id": 210798,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210798,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210798, "pid": 5, "tid": 7, "ts": 1716454224783993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728007, "dur": 10,
    "args": {
      "External id": 210798,
      "cbid": 211, "correlation": 210798
    }
  },
  {
    "ph": "s", "id": 210798, "pid": 76337, "tid": -914061504, "ts": 1716454224728007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784002, "dur": 7,
    "args": {
      "External id": 210809,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210809,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210809, "pid": 5, "tid": 7, "ts": 1716454224784002,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728082, "dur": 12,
    "args": {
      "External id": 210809,
      "cbid": 211, "correlation": 210809
    }
  },
  {
    "ph": "s", "id": 210809, "pid": 76337, "tid": -914061504, "ts": 1716454224728082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784010, "dur": 9,
    "args": {
      "External id": 210831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210831,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 210831, "pid": 5, "tid": 7, "ts": 1716454224784010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728115, "dur": 7,
    "args": {
      "External id": 210831,
      "cbid": 211, "correlation": 210831
    }
  },
  {
    "ph": "s", "id": 210831, "pid": 76337, "tid": -914061504, "ts": 1716454224728115,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728200, "dur": 2,
    "args": {
      "External id": 210842,
      "cbid": 251, "correlation": 210842
    }
  },
  {
    "ph": "f", "id": 210842, "pid": 76337, "tid": -914061504, "ts": 1716454224728200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224784021, "dur": 1,
    "args": {
      "External id": 210843,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210843,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 210843, "pid": 5, "tid": 7, "ts": 1716454224784021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728206, "dur": 11,
    "args": {
      "External id": 210843,
      "cbid": 51, "correlation": 210843
    }
  },
  {
    "ph": "s", "id": 210843, "pid": 76337, "tid": -914061504, "ts": 1716454224728206,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224784025, "dur": 37,
    "args": {
      "External id": 210844,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210844,
      "registers per thread": 106,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 210844, "pid": 5, "tid": 7, "ts": 1716454224784025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728218, "dur": 13,
    "args": {
      "External id": 210844,
      "cbid": 211, "correlation": 210844
    }
  },
  {
    "ph": "s", "id": 210844, "pid": 76337, "tid": -914061504, "ts": 1716454224728218,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728290, "dur": 1,
    "args": {
      "External id": 210855,
      "cbid": 251, "correlation": 210855
    }
  },
  {
    "ph": "f", "id": 210855, "pid": 76337, "tid": -914061504, "ts": 1716454224728290,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728294, "dur": 0,
    "args": {
      "External id": 210856,
      "cbid": 251, "correlation": 210856
    }
  },
  {
    "ph": "f", "id": 210856, "pid": 76337, "tid": -914061504, "ts": 1716454224728294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224784063, "dur": 11,
    "args": {
      "External id": 210857,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210857,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210857, "pid": 5, "tid": 7, "ts": 1716454224784063,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728295, "dur": 12,
    "args": {
      "External id": 210857,
      "cbid": 211, "correlation": 210857
    }
  },
  {
    "ph": "s", "id": 210857, "pid": 76337, "tid": -914061504, "ts": 1716454224728295,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224784076, "dur": 6,
    "args": {
      "External id": 210859,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210859,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210859, "pid": 5, "tid": 7, "ts": 1716454224784076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728309, "dur": 7,
    "args": {
      "External id": 210859,
      "cbid": 211, "correlation": 210859
    }
  },
  {
    "ph": "s", "id": 210859, "pid": 76337, "tid": -914061504, "ts": 1716454224728309,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728367, "dur": 1,
    "args": {
      "External id": 210870,
      "cbid": 251, "correlation": 210870
    }
  },
  {
    "ph": "f", "id": 210870, "pid": 76337, "tid": -914061504, "ts": 1716454224728367,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728371, "dur": 0,
    "args": {
      "External id": 210871,
      "cbid": 251, "correlation": 210871
    }
  },
  {
    "ph": "f", "id": 210871, "pid": 76337, "tid": -914061504, "ts": 1716454224728371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224784083, "dur": 8,
    "args": {
      "External id": 210872,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210872,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210872, "pid": 5, "tid": 7, "ts": 1716454224784083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728372, "dur": 11,
    "args": {
      "External id": 210872,
      "cbid": 211, "correlation": 210872
    }
  },
  {
    "ph": "s", "id": 210872, "pid": 76337, "tid": -914061504, "ts": 1716454224728372,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224784093, "dur": 4,
    "args": {
      "External id": 210874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210874,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210874, "pid": 5, "tid": 7, "ts": 1716454224784093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728385, "dur": 5,
    "args": {
      "External id": 210874,
      "cbid": 211, "correlation": 210874
    }
  },
  {
    "ph": "s", "id": 210874, "pid": 76337, "tid": -914061504, "ts": 1716454224728385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224784098, "dur": 19,
    "args": {
      "External id": 210899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210899,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 1.6,
      "warps per SM": 6.4,
      "grid": [2, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 10
    }
  },
  {
    "ph": "f", "id": 210899, "pid": 5, "tid": 7, "ts": 1716454224784098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728465, "dur": 12,
    "args": {
      "External id": 210899,
      "cbid": 211, "correlation": 210899
    }
  },
  {
    "ph": "s", "id": 210899, "pid": 76337, "tid": -914061504, "ts": 1716454224728465,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728565, "dur": 2,
    "args": {
      "External id": 210917,
      "cbid": 251, "correlation": 210917
    }
  },
  {
    "ph": "f", "id": 210917, "pid": 76337, "tid": -914061504, "ts": 1716454224728565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224784119, "dur": 1,
    "args": {
      "External id": 210919,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 210919,
      "bytes": 480, "memory bandwidth (GB/s)": 0.30612244897959184
    }
  },
  {
    "ph": "f", "id": 210919, "pid": 5, "tid": 7, "ts": 1716454224784119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728571, "dur": 9,
    "args": {
      "External id": 210919,
      "cbid": 51, "correlation": 210919
    }
  },
  {
    "ph": "s", "id": 210919, "pid": 76337, "tid": -914061504, "ts": 1716454224728571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x64_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224784123, "dur": 37,
    "args": {
      "External id": 210920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210920,
      "registers per thread": 110,
      "shared memory": 16896,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [20, 6, 2],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 210920, "pid": 5, "tid": 7, "ts": 1716454224784123,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728582, "dur": 12,
    "args": {
      "External id": 210920,
      "cbid": 211, "correlation": 210920
    }
  },
  {
    "ph": "s", "id": 210920, "pid": 76337, "tid": -914061504, "ts": 1716454224728582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224784161, "dur": 4,
    "args": {
      "External id": 210928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210928,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210928, "pid": 5, "tid": 7, "ts": 1716454224784161,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728652, "dur": 13,
    "args": {
      "External id": 210928,
      "cbid": 211, "correlation": 210928
    }
  },
  {
    "ph": "s", "id": 210928, "pid": 76337, "tid": -914061504, "ts": 1716454224728652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784167, "dur": 8,
    "args": {
      "External id": 210936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 210936, "pid": 5, "tid": 7, "ts": 1716454224784167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728694, "dur": 9,
    "args": {
      "External id": 210936,
      "cbid": 211, "correlation": 210936
    }
  },
  {
    "ph": "s", "id": 210936, "pid": 76337, "tid": -914061504, "ts": 1716454224728694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784176, "dur": 8,
    "args": {
      "External id": 210958,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210958,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [384, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 30
    }
  },
  {
    "ph": "f", "id": 210958, "pid": 5, "tid": 7, "ts": 1716454224784176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728747, "dur": 10,
    "args": {
      "External id": 210958,
      "cbid": 211, "correlation": 210958
    }
  },
  {
    "ph": "s", "id": 210958, "pid": 76337, "tid": -914061504, "ts": 1716454224728747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728838, "dur": 1,
    "args": {
      "External id": 210974,
      "cbid": 251, "correlation": 210974
    }
  },
  {
    "ph": "f", "id": 210974, "pid": 76337, "tid": -914061504, "ts": 1716454224728838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728843, "dur": 0,
    "args": {
      "External id": 210976,
      "cbid": 251, "correlation": 210976
    }
  },
  {
    "ph": "f", "id": 210976, "pid": 76337, "tid": -914061504, "ts": 1716454224728843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224784186, "dur": 191,
    "args": {
      "External id": 210977,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210977,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [24, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 210977, "pid": 5, "tid": 7, "ts": 1716454224784186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728845, "dur": 13,
    "args": {
      "External id": 210977,
      "cbid": 211, "correlation": 210977
    }
  },
  {
    "ph": "s", "id": 210977, "pid": 76337, "tid": -914061504, "ts": 1716454224728845,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784378, "dur": 20,
    "args": {
      "External id": 210985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210985,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210985, "pid": 5, "tid": 7, "ts": 1716454224784378,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728912, "dur": 13,
    "args": {
      "External id": 210985,
      "cbid": 211, "correlation": 210985
    }
  },
  {
    "ph": "s", "id": 210985, "pid": 76337, "tid": -914061504, "ts": 1716454224728912,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784400, "dur": 23,
    "args": {
      "External id": 210993,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 210993,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 210993, "pid": 5, "tid": 7, "ts": 1716454224784400,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224728942, "dur": 8,
    "args": {
      "External id": 210993,
      "cbid": 211, "correlation": 210993
    }
  },
  {
    "ph": "s", "id": 210993, "pid": 76337, "tid": -914061504, "ts": 1716454224728942,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729034, "dur": 1,
    "args": {
      "External id": 211009,
      "cbid": 251, "correlation": 211009
    }
  },
  {
    "ph": "f", "id": 211009, "pid": 76337, "tid": -914061504, "ts": 1716454224729034,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224784425, "dur": 1,
    "args": {
      "External id": 211011,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211011,
      "bytes": 120, "memory bandwidth (GB/s)": 0.078125
    }
  },
  {
    "ph": "f", "id": 211011, "pid": 5, "tid": 7, "ts": 1716454224784425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729039, "dur": 8,
    "args": {
      "External id": 211011,
      "cbid": 51, "correlation": 211011
    }
  },
  {
    "ph": "s", "id": 211011, "pid": 76337, "tid": -914061504, "ts": 1716454224729039,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224784429, "dur": 111,
    "args": {
      "External id": 211012,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211012,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 1.875,
      "warps per SM": 7.5,
      "grid": [10, 3, 5],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 12
    }
  },
  {
    "ph": "f", "id": 211012, "pid": 5, "tid": 7, "ts": 1716454224784429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729049, "dur": 13,
    "args": {
      "External id": 211012,
      "cbid": 211, "correlation": 211012
    }
  },
  {
    "ph": "s", "id": 211012, "pid": 76337, "tid": -914061504, "ts": 1716454224729049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224784541, "dur": 6,
    "args": {
      "External id": 211020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211020,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211020, "pid": 5, "tid": 7, "ts": 1716454224784541,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729093, "dur": 10,
    "args": {
      "External id": 211020,
      "cbid": 211, "correlation": 211020
    }
  },
  {
    "ph": "s", "id": 211020, "pid": 76337, "tid": -914061504, "ts": 1716454224729093,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784548, "dur": 10,
    "args": {
      "External id": 211031,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211031,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211031, "pid": 5, "tid": 7, "ts": 1716454224784548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729163, "dur": 12,
    "args": {
      "External id": 211031,
      "cbid": 211, "correlation": 211031
    }
  },
  {
    "ph": "s", "id": 211031, "pid": 76337, "tid": -914061504, "ts": 1716454224729163,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729226, "dur": 0,
    "args": {
      "External id": 211043,
      "cbid": 317, "correlation": 211043
    }
  },
  {
    "ph": "f", "id": 211043, "pid": 76337, "tid": -914061504, "ts": 1716454224729226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729227, "dur": 0,
    "args": {
      "External id": 211044,
      "cbid": 203, "correlation": 211044
    }
  },
  {
    "ph": "f", "id": 211044, "pid": 76337, "tid": -914061504, "ts": 1716454224729227,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729228, "dur": 0,
    "args": {
      "External id": 211045,
      "cbid": 205, "correlation": 211045
    }
  },
  {
    "ph": "f", "id": 211045, "pid": 76337, "tid": -914061504, "ts": 1716454224729228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784559, "dur": 6,
    "args": {
      "External id": 211049,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211049,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211049, "pid": 5, "tid": 7, "ts": 1716454224784559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729243, "dur": 12,
    "args": {
      "External id": 211049,
      "cbid": 211, "correlation": 211049
    }
  },
  {
    "ph": "s", "id": 211049, "pid": 76337, "tid": -914061504, "ts": 1716454224729243,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224784566, "dur": 37,
    "args": {
      "External id": 211051,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211051,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [20, 6, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 211051, "pid": 5, "tid": 7, "ts": 1716454224784566,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729263, "dur": 6,
    "args": {
      "External id": 211051,
      "cbid": 211, "correlation": 211051
    }
  },
  {
    "ph": "s", "id": 211051, "pid": 76337, "tid": -914061504, "ts": 1716454224729263,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784604, "dur": 6,
    "args": {
      "External id": 211053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211053,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211053, "pid": 5, "tid": 7, "ts": 1716454224784604,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729273, "dur": 5,
    "args": {
      "External id": 211053,
      "cbid": 211, "correlation": 211053
    }
  },
  {
    "ph": "s", "id": 211053, "pid": 76337, "tid": -914061504, "ts": 1716454224729273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784611, "dur": 7,
    "args": {
      "External id": 211059,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211059,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211059, "pid": 5, "tid": 7, "ts": 1716454224784611,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729300, "dur": 9,
    "args": {
      "External id": 211059,
      "cbid": 211, "correlation": 211059
    }
  },
  {
    "ph": "s", "id": 211059, "pid": 76337, "tid": -914061504, "ts": 1716454224729300,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224784620, "dur": 5,
    "args": {
      "External id": 211067,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211067,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211067, "pid": 5, "tid": 7, "ts": 1716454224784620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729331, "dur": 8,
    "args": {
      "External id": 211067,
      "cbid": 211, "correlation": 211067
    }
  },
  {
    "ph": "s", "id": 211067, "pid": 76337, "tid": -914061504, "ts": 1716454224729331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784626, "dur": 11,
    "args": {
      "External id": 211087,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211087,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211087, "pid": 5, "tid": 7, "ts": 1716454224784626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729404, "dur": 12,
    "args": {
      "External id": 211087,
      "cbid": 211, "correlation": 211087
    }
  },
  {
    "ph": "s", "id": 211087, "pid": 76337, "tid": -914061504, "ts": 1716454224729404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224784639, "dur": 5,
    "args": {
      "External id": 211099,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211099,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 211099, "pid": 5, "tid": 7, "ts": 1716454224784639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729426, "dur": 6,
    "args": {
      "External id": 211099,
      "cbid": 211, "correlation": 211099
    }
  },
  {
    "ph": "s", "id": 211099, "pid": 76337, "tid": -914061504, "ts": 1716454224729426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224784645, "dur": 9,
    "args": {
      "External id": 211102,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211102,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211102, "pid": 5, "tid": 7, "ts": 1716454224784645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729444, "dur": 7,
    "args": {
      "External id": 211102,
      "cbid": 211, "correlation": 211102
    }
  },
  {
    "ph": "s", "id": 211102, "pid": 76337, "tid": -914061504, "ts": 1716454224729444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224784655, "dur": 5,
    "args": {
      "External id": 211111,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211111,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211111, "pid": 5, "tid": 7, "ts": 1716454224784655,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729483, "dur": 9,
    "args": {
      "External id": 211111,
      "cbid": 211, "correlation": 211111
    }
  },
  {
    "ph": "s", "id": 211111, "pid": 76337, "tid": -914061504, "ts": 1716454224729483,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729534, "dur": 0,
    "args": {
      "External id": 211121,
      "cbid": 317, "correlation": 211121
    }
  },
  {
    "ph": "f", "id": 211121, "pid": 76337, "tid": -914061504, "ts": 1716454224729534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729535, "dur": 0,
    "args": {
      "External id": 211122,
      "cbid": 203, "correlation": 211122
    }
  },
  {
    "ph": "f", "id": 211122, "pid": 76337, "tid": -914061504, "ts": 1716454224729535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729535, "dur": 0,
    "args": {
      "External id": 211123,
      "cbid": 205, "correlation": 211123
    }
  },
  {
    "ph": "f", "id": 211123, "pid": 76337, "tid": -914061504, "ts": 1716454224729535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784662, "dur": 5,
    "args": {
      "External id": 211127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211127,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211127, "pid": 5, "tid": 7, "ts": 1716454224784662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729548, "dur": 12,
    "args": {
      "External id": 211127,
      "cbid": 211, "correlation": 211127
    }
  },
  {
    "ph": "s", "id": 211127, "pid": 76337, "tid": -914061504, "ts": 1716454224729548,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224784668, "dur": 163,
    "args": {
      "External id": 211129,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211129,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211129, "pid": 5, "tid": 7, "ts": 1716454224784668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729563, "dur": 5,
    "args": {
      "External id": 211129,
      "cbid": 211, "correlation": 211129
    }
  },
  {
    "ph": "s", "id": 211129, "pid": 76337, "tid": -914061504, "ts": 1716454224729563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224784834, "dur": 1,
    "args": {
      "External id": 211131,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211131,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 211131, "pid": 5, "tid": 7, "ts": 1716454224784834,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729574, "dur": 7,
    "args": {
      "External id": 211131,
      "cbid": 51, "correlation": 211131
    }
  },
  {
    "ph": "s", "id": 211131, "pid": 76337, "tid": -914061504, "ts": 1716454224729574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224784837, "dur": 272,
    "args": {
      "External id": 211132,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211132,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211132, "pid": 5, "tid": 7, "ts": 1716454224784837,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729582, "dur": 7,
    "args": {
      "External id": 211132,
      "cbid": 211, "correlation": 211132
    }
  },
  {
    "ph": "s", "id": 211132, "pid": 76337, "tid": -914061504, "ts": 1716454224729582,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785111, "dur": 5,
    "args": {
      "External id": 211134,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211134,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211134, "pid": 5, "tid": 7, "ts": 1716454224785111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729592, "dur": 5,
    "args": {
      "External id": 211134,
      "cbid": 211, "correlation": 211134
    }
  },
  {
    "ph": "s", "id": 211134, "pid": 76337, "tid": -914061504, "ts": 1716454224729592,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224785118, "dur": 6,
    "args": {
      "External id": 211140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211140, "pid": 5, "tid": 7, "ts": 1716454224785118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729620, "dur": 9,
    "args": {
      "External id": 211140,
      "cbid": 211, "correlation": 211140
    }
  },
  {
    "ph": "s", "id": 211140, "pid": 76337, "tid": -914061504, "ts": 1716454224729620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224785125, "dur": 3,
    "args": {
      "External id": 211148,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211148,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 211148, "pid": 5, "tid": 7, "ts": 1716454224785125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729664, "dur": 9,
    "args": {
      "External id": 211148,
      "cbid": 211, "correlation": 211148
    }
  },
  {
    "ph": "s", "id": 211148, "pid": 76337, "tid": -914061504, "ts": 1716454224729664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729727, "dur": 1,
    "args": {
      "External id": 211164,
      "cbid": 251, "correlation": 211164
    }
  },
  {
    "ph": "f", "id": 211164, "pid": 76337, "tid": -914061504, "ts": 1716454224729727,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729732, "dur": 0,
    "args": {
      "External id": 211166,
      "cbid": 251, "correlation": 211166
    }
  },
  {
    "ph": "f", "id": 211166, "pid": 76337, "tid": -914061504, "ts": 1716454224729732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224785130, "dur": 13,
    "args": {
      "External id": 211167,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211167,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211167, "pid": 5, "tid": 7, "ts": 1716454224785130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729733, "dur": 11,
    "args": {
      "External id": 211167,
      "cbid": 211, "correlation": 211167
    }
  },
  {
    "ph": "s", "id": 211167, "pid": 76337, "tid": -914061504, "ts": 1716454224729733,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224785144, "dur": 5,
    "args": {
      "External id": 211169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211169,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211169, "pid": 5, "tid": 7, "ts": 1716454224785144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729746, "dur": 5,
    "args": {
      "External id": 211169,
      "cbid": 211, "correlation": 211169
    }
  },
  {
    "ph": "s", "id": 211169, "pid": 76337, "tid": -914061504, "ts": 1716454224729746,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224785150, "dur": 6,
    "args": {
      "External id": 211179,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211179,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211179, "pid": 5, "tid": 7, "ts": 1716454224785150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729803, "dur": 13,
    "args": {
      "External id": 211179,
      "cbid": 211, "correlation": 211179
    }
  },
  {
    "ph": "s", "id": 211179, "pid": 76337, "tid": -914061504, "ts": 1716454224729803,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785157, "dur": 10,
    "args": {
      "External id": 211199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211199,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211199, "pid": 5, "tid": 7, "ts": 1716454224785157,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729871, "dur": 10,
    "args": {
      "External id": 211199,
      "cbid": 211, "correlation": 211199
    }
  },
  {
    "ph": "s", "id": 211199, "pid": 76337, "tid": -914061504, "ts": 1716454224729871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224785169, "dur": 4,
    "args": {
      "External id": 211211,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211211,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 211211, "pid": 5, "tid": 7, "ts": 1716454224785169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729891, "dur": 6,
    "args": {
      "External id": 211211,
      "cbid": 211, "correlation": 211211
    }
  },
  {
    "ph": "s", "id": 211211, "pid": 76337, "tid": -914061504, "ts": 1716454224729891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224785174, "dur": 7,
    "args": {
      "External id": 211214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211214,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211214, "pid": 5, "tid": 7, "ts": 1716454224785174,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729911, "dur": 7,
    "args": {
      "External id": 211214,
      "cbid": 211, "correlation": 211214
    }
  },
  {
    "ph": "s", "id": 211214, "pid": 76337, "tid": -914061504, "ts": 1716454224729911,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224785182, "dur": 5,
    "args": {
      "External id": 211223,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211223,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211223, "pid": 5, "tid": 7, "ts": 1716454224785182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224729950, "dur": 10,
    "args": {
      "External id": 211223,
      "cbid": 211, "correlation": 211223
    }
  },
  {
    "ph": "s", "id": 211223, "pid": 76337, "tid": -914061504, "ts": 1716454224729950,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730022, "dur": 0,
    "args": {
      "External id": 211233,
      "cbid": 317, "correlation": 211233
    }
  },
  {
    "ph": "f", "id": 211233, "pid": 76337, "tid": -914061504, "ts": 1716454224730022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730023, "dur": 0,
    "args": {
      "External id": 211234,
      "cbid": 203, "correlation": 211234
    }
  },
  {
    "ph": "f", "id": 211234, "pid": 76337, "tid": -914061504, "ts": 1716454224730023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730024, "dur": 0,
    "args": {
      "External id": 211235,
      "cbid": 205, "correlation": 211235
    }
  },
  {
    "ph": "f", "id": 211235, "pid": 76337, "tid": -914061504, "ts": 1716454224730024,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785188, "dur": 5,
    "args": {
      "External id": 211239,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211239,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211239, "pid": 5, "tid": 7, "ts": 1716454224785188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730037, "dur": 12,
    "args": {
      "External id": 211239,
      "cbid": 211, "correlation": 211239
    }
  },
  {
    "ph": "s", "id": 211239, "pid": 76337, "tid": -914061504, "ts": 1716454224730037,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785195, "dur": 164,
    "args": {
      "External id": 211241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211241,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211241, "pid": 5, "tid": 7, "ts": 1716454224785195,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730052, "dur": 6,
    "args": {
      "External id": 211241,
      "cbid": 211, "correlation": 211241
    }
  },
  {
    "ph": "s", "id": 211241, "pid": 76337, "tid": -914061504, "ts": 1716454224730052,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224785361, "dur": 1,
    "args": {
      "External id": 211243,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211243,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 211243, "pid": 5, "tid": 7, "ts": 1716454224785361,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730064, "dur": 6,
    "args": {
      "External id": 211243,
      "cbid": 51, "correlation": 211243
    }
  },
  {
    "ph": "s", "id": 211243, "pid": 76337, "tid": -914061504, "ts": 1716454224730064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224785365, "dur": 261,
    "args": {
      "External id": 211244,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211244,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211244, "pid": 5, "tid": 7, "ts": 1716454224785365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730071, "dur": 6,
    "args": {
      "External id": 211244,
      "cbid": 211, "correlation": 211244
    }
  },
  {
    "ph": "s", "id": 211244, "pid": 76337, "tid": -914061504, "ts": 1716454224730071,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785627, "dur": 6,
    "args": {
      "External id": 211246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211246,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211246, "pid": 5, "tid": 7, "ts": 1716454224785627,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730081, "dur": 5,
    "args": {
      "External id": 211246,
      "cbid": 211, "correlation": 211246
    }
  },
  {
    "ph": "s", "id": 211246, "pid": 76337, "tid": -914061504, "ts": 1716454224730081,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224785635, "dur": 6,
    "args": {
      "External id": 211252,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211252,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211252, "pid": 5, "tid": 7, "ts": 1716454224785635,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730110, "dur": 8,
    "args": {
      "External id": 211252,
      "cbid": 211, "correlation": 211252
    }
  },
  {
    "ph": "s", "id": 211252, "pid": 76337, "tid": -914061504, "ts": 1716454224730110,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224785642, "dur": 5,
    "args": {
      "External id": 211260,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211260,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211260, "pid": 5, "tid": 7, "ts": 1716454224785642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730142, "dur": 8,
    "args": {
      "External id": 211260,
      "cbid": 211, "correlation": 211260
    }
  },
  {
    "ph": "s", "id": 211260, "pid": 76337, "tid": -914061504, "ts": 1716454224730142,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224785648, "dur": 4,
    "args": {
      "External id": 211268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211268,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211268, "pid": 5, "tid": 7, "ts": 1716454224785648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730171, "dur": 8,
    "args": {
      "External id": 211268,
      "cbid": 211, "correlation": 211268
    }
  },
  {
    "ph": "s", "id": 211268, "pid": 76337, "tid": -914061504, "ts": 1716454224730171,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224785654, "dur": 11,
    "args": {
      "External id": 211277,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211277,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211277, "pid": 5, "tid": 7, "ts": 1716454224785654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730260, "dur": 14,
    "args": {
      "External id": 211277,
      "cbid": 211, "correlation": 211277
    }
  },
  {
    "ph": "s", "id": 211277, "pid": 76337, "tid": -914061504, "ts": 1716454224730260,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785667, "dur": 12,
    "args": {
      "External id": 211297,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211297,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211297, "pid": 5, "tid": 7, "ts": 1716454224785667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730328, "dur": 11,
    "args": {
      "External id": 211297,
      "cbid": 211, "correlation": 211297
    }
  },
  {
    "ph": "s", "id": 211297, "pid": 76337, "tid": -914061504, "ts": 1716454224730328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224785680, "dur": 4,
    "args": {
      "External id": 211309,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211309,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211309, "pid": 5, "tid": 7, "ts": 1716454224785680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730349, "dur": 6,
    "args": {
      "External id": 211309,
      "cbid": 211, "correlation": 211309
    }
  },
  {
    "ph": "s", "id": 211309, "pid": 76337, "tid": -914061504, "ts": 1716454224730349,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224785686, "dur": 10,
    "args": {
      "External id": 211312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211312,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211312, "pid": 5, "tid": 7, "ts": 1716454224785686,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730368, "dur": 6,
    "args": {
      "External id": 211312,
      "cbid": 211, "correlation": 211312
    }
  },
  {
    "ph": "s", "id": 211312, "pid": 76337, "tid": -914061504, "ts": 1716454224730368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224785697, "dur": 6,
    "args": {
      "External id": 211321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211321,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211321, "pid": 5, "tid": 7, "ts": 1716454224785697,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730406, "dur": 9,
    "args": {
      "External id": 211321,
      "cbid": 211, "correlation": 211321
    }
  },
  {
    "ph": "s", "id": 211321, "pid": 76337, "tid": -914061504, "ts": 1716454224730406,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730458, "dur": 0,
    "args": {
      "External id": 211331,
      "cbid": 317, "correlation": 211331
    }
  },
  {
    "ph": "f", "id": 211331, "pid": 76337, "tid": -914061504, "ts": 1716454224730458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730458, "dur": 0,
    "args": {
      "External id": 211332,
      "cbid": 203, "correlation": 211332
    }
  },
  {
    "ph": "f", "id": 211332, "pid": 76337, "tid": -914061504, "ts": 1716454224730458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730459, "dur": 0,
    "args": {
      "External id": 211333,
      "cbid": 205, "correlation": 211333
    }
  },
  {
    "ph": "f", "id": 211333, "pid": 76337, "tid": -914061504, "ts": 1716454224730459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785705, "dur": 7,
    "args": {
      "External id": 211337,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211337,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211337, "pid": 5, "tid": 7, "ts": 1716454224785705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730475, "dur": 11,
    "args": {
      "External id": 211337,
      "cbid": 211, "correlation": 211337
    }
  },
  {
    "ph": "s", "id": 211337, "pid": 76337, "tid": -914061504, "ts": 1716454224730475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224785713, "dur": 322,
    "args": {
      "External id": 211339,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211339,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211339, "pid": 5, "tid": 7, "ts": 1716454224785713,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730488, "dur": 5,
    "args": {
      "External id": 211339,
      "cbid": 211, "correlation": 211339
    }
  },
  {
    "ph": "s", "id": 211339, "pid": 76337, "tid": -914061504, "ts": 1716454224730488,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224786037, "dur": 1,
    "args": {
      "External id": 211341,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211341,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 211341, "pid": 5, "tid": 7, "ts": 1716454224786037,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730499, "dur": 7,
    "args": {
      "External id": 211341,
      "cbid": 51, "correlation": 211341
    }
  },
  {
    "ph": "s", "id": 211341, "pid": 76337, "tid": -914061504, "ts": 1716454224730499,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224786040, "dur": 498,
    "args": {
      "External id": 211342,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211342,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211342, "pid": 5, "tid": 7, "ts": 1716454224786040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730508, "dur": 6,
    "args": {
      "External id": 211342,
      "cbid": 211, "correlation": 211342
    }
  },
  {
    "ph": "s", "id": 211342, "pid": 76337, "tid": -914061504, "ts": 1716454224730508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224786540, "dur": 6,
    "args": {
      "External id": 211344,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211344,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211344, "pid": 5, "tid": 7, "ts": 1716454224786540,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730518, "dur": 5,
    "args": {
      "External id": 211344,
      "cbid": 211, "correlation": 211344
    }
  },
  {
    "ph": "s", "id": 211344, "pid": 76337, "tid": -914061504, "ts": 1716454224730518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224786547, "dur": 6,
    "args": {
      "External id": 211350,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211350,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211350, "pid": 5, "tid": 7, "ts": 1716454224786547,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730545, "dur": 9,
    "args": {
      "External id": 211350,
      "cbid": 211, "correlation": 211350
    }
  },
  {
    "ph": "s", "id": 211350, "pid": 76337, "tid": -914061504, "ts": 1716454224730545,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224786554, "dur": 3,
    "args": {
      "External id": 211358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211358,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 211358, "pid": 5, "tid": 7, "ts": 1716454224786554,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730589, "dur": 10,
    "args": {
      "External id": 211358,
      "cbid": 211, "correlation": 211358
    }
  },
  {
    "ph": "s", "id": 211358, "pid": 76337, "tid": -914061504, "ts": 1716454224730589,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730653, "dur": 1,
    "args": {
      "External id": 211374,
      "cbid": 251, "correlation": 211374
    }
  },
  {
    "ph": "f", "id": 211374, "pid": 76337, "tid": -914061504, "ts": 1716454224730653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730658, "dur": 0,
    "args": {
      "External id": 211376,
      "cbid": 251, "correlation": 211376
    }
  },
  {
    "ph": "f", "id": 211376, "pid": 76337, "tid": -914061504, "ts": 1716454224730658,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224786559, "dur": 11,
    "args": {
      "External id": 211377,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211377,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211377, "pid": 5, "tid": 7, "ts": 1716454224786559,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730660, "dur": 11,
    "args": {
      "External id": 211377,
      "cbid": 211, "correlation": 211377
    }
  },
  {
    "ph": "s", "id": 211377, "pid": 76337, "tid": -914061504, "ts": 1716454224730660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224786572, "dur": 5,
    "args": {
      "External id": 211379,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211379,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211379, "pid": 5, "tid": 7, "ts": 1716454224786572,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730674, "dur": 6,
    "args": {
      "External id": 211379,
      "cbid": 211, "correlation": 211379
    }
  },
  {
    "ph": "s", "id": 211379, "pid": 76337, "tid": -914061504, "ts": 1716454224730674,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224786578, "dur": 6,
    "args": {
      "External id": 211389,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211389,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211389, "pid": 5, "tid": 7, "ts": 1716454224786578,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730731, "dur": 12,
    "args": {
      "External id": 211389,
      "cbid": 211, "correlation": 211389
    }
  },
  {
    "ph": "s", "id": 211389, "pid": 76337, "tid": -914061504, "ts": 1716454224730731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224786585, "dur": 9,
    "args": {
      "External id": 211409,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211409,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211409, "pid": 5, "tid": 7, "ts": 1716454224786585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730796, "dur": 11,
    "args": {
      "External id": 211409,
      "cbid": 211, "correlation": 211409
    }
  },
  {
    "ph": "s", "id": 211409, "pid": 76337, "tid": -914061504, "ts": 1716454224730796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224786596, "dur": 4,
    "args": {
      "External id": 211421,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211421,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 211421, "pid": 5, "tid": 7, "ts": 1716454224786596,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730818, "dur": 6,
    "args": {
      "External id": 211421,
      "cbid": 211, "correlation": 211421
    }
  },
  {
    "ph": "s", "id": 211421, "pid": 76337, "tid": -914061504, "ts": 1716454224730818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224786601, "dur": 7,
    "args": {
      "External id": 211424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211424,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211424, "pid": 5, "tid": 7, "ts": 1716454224786601,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730837, "dur": 6,
    "args": {
      "External id": 211424,
      "cbid": 211, "correlation": 211424
    }
  },
  {
    "ph": "s", "id": 211424, "pid": 76337, "tid": -914061504, "ts": 1716454224730837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224786609, "dur": 5,
    "args": {
      "External id": 211433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211433,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211433, "pid": 5, "tid": 7, "ts": 1716454224786609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730877, "dur": 10,
    "args": {
      "External id": 211433,
      "cbid": 211, "correlation": 211433
    }
  },
  {
    "ph": "s", "id": 211433, "pid": 76337, "tid": -914061504, "ts": 1716454224730877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730939, "dur": 0,
    "args": {
      "External id": 211443,
      "cbid": 317, "correlation": 211443
    }
  },
  {
    "ph": "f", "id": 211443, "pid": 76337, "tid": -914061504, "ts": 1716454224730939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730940, "dur": 0,
    "args": {
      "External id": 211444,
      "cbid": 203, "correlation": 211444
    }
  },
  {
    "ph": "f", "id": 211444, "pid": 76337, "tid": -914061504, "ts": 1716454224730940,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730941, "dur": 0,
    "args": {
      "External id": 211445,
      "cbid": 205, "correlation": 211445
    }
  },
  {
    "ph": "f", "id": 211445, "pid": 76337, "tid": -914061504, "ts": 1716454224730941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224786615, "dur": 5,
    "args": {
      "External id": 211449,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211449,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211449, "pid": 5, "tid": 7, "ts": 1716454224786615,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730954, "dur": 12,
    "args": {
      "External id": 211449,
      "cbid": 211, "correlation": 211449
    }
  },
  {
    "ph": "s", "id": 211449, "pid": 76337, "tid": -914061504, "ts": 1716454224730954,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224786621, "dur": 163,
    "args": {
      "External id": 211451,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211451,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211451, "pid": 5, "tid": 7, "ts": 1716454224786621,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730969, "dur": 13,
    "args": {
      "External id": 211451,
      "cbid": 211, "correlation": 211451
    }
  },
  {
    "ph": "s", "id": 211451, "pid": 76337, "tid": -914061504, "ts": 1716454224730969,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224786787, "dur": 1,
    "args": {
      "External id": 211453,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211453,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 211453, "pid": 5, "tid": 7, "ts": 1716454224786787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730987, "dur": 8,
    "args": {
      "External id": 211453,
      "cbid": 51, "correlation": 211453
    }
  },
  {
    "ph": "s", "id": 211453, "pid": 76337, "tid": -914061504, "ts": 1716454224730987,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224786790, "dur": 261,
    "args": {
      "External id": 211454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211454,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211454, "pid": 5, "tid": 7, "ts": 1716454224786790,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224730996, "dur": 7,
    "args": {
      "External id": 211454,
      "cbid": 211, "correlation": 211454
    }
  },
  {
    "ph": "s", "id": 211454, "pid": 76337, "tid": -914061504, "ts": 1716454224730996,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224787052, "dur": 6,
    "args": {
      "External id": 211456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211456,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211456, "pid": 5, "tid": 7, "ts": 1716454224787052,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731007, "dur": 5,
    "args": {
      "External id": 211456,
      "cbid": 211, "correlation": 211456
    }
  },
  {
    "ph": "s", "id": 211456, "pid": 76337, "tid": -914061504, "ts": 1716454224731007,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224787059, "dur": 6,
    "args": {
      "External id": 211462,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211462,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211462, "pid": 5, "tid": 7, "ts": 1716454224787059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731035, "dur": 9,
    "args": {
      "External id": 211462,
      "cbid": 211, "correlation": 211462
    }
  },
  {
    "ph": "s", "id": 211462, "pid": 76337, "tid": -914061504, "ts": 1716454224731035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731094, "dur": 0,
    "args": {
      "External id": 211472,
      "cbid": 317, "correlation": 211472
    }
  },
  {
    "ph": "f", "id": 211472, "pid": 76337, "tid": -914061504, "ts": 1716454224731094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731094, "dur": 0,
    "args": {
      "External id": 211473,
      "cbid": 203, "correlation": 211473
    }
  },
  {
    "ph": "f", "id": 211473, "pid": 76337, "tid": -914061504, "ts": 1716454224731094,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731095, "dur": 0,
    "args": {
      "External id": 211474,
      "cbid": 205, "correlation": 211474
    }
  },
  {
    "ph": "f", "id": 211474, "pid": 76337, "tid": -914061504, "ts": 1716454224731095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224787067, "dur": 8,
    "args": {
      "External id": 211478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211478, "pid": 5, "tid": 7, "ts": 1716454224787067,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731109, "dur": 12,
    "args": {
      "External id": 211478,
      "cbid": 211, "correlation": 211478
    }
  },
  {
    "ph": "s", "id": 211478, "pid": 76337, "tid": -914061504, "ts": 1716454224731109,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224787076, "dur": 3,
    "args": {
      "External id": 211480,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211480,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.5
    }
  },
  {
    "ph": "f", "id": 211480, "pid": 5, "tid": 7, "ts": 1716454224787076,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731127, "dur": 13,
    "args": {
      "External id": 211480,
      "cbid": 51, "correlation": 211480
    }
  },
  {
    "ph": "s", "id": 211480, "pid": 76337, "tid": -914061504, "ts": 1716454224731127,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224787080, "dur": 98,
    "args": {
      "External id": 211481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211481,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 211481, "pid": 5, "tid": 7, "ts": 1716454224787080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731141, "dur": 7,
    "args": {
      "External id": 211481,
      "cbid": 211, "correlation": 211481
    }
  },
  {
    "ph": "s", "id": 211481, "pid": 76337, "tid": -914061504, "ts": 1716454224731141,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224787179, "dur": 6,
    "args": {
      "External id": 211483,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211483,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211483, "pid": 5, "tid": 7, "ts": 1716454224787179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731152, "dur": 5,
    "args": {
      "External id": 211483,
      "cbid": 211, "correlation": 211483
    }
  },
  {
    "ph": "s", "id": 211483, "pid": 76337, "tid": -914061504, "ts": 1716454224731152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224787186, "dur": 6,
    "args": {
      "External id": 211489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211489,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211489, "pid": 5, "tid": 7, "ts": 1716454224787186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731178, "dur": 8,
    "args": {
      "External id": 211489,
      "cbid": 211, "correlation": 211489
    }
  },
  {
    "ph": "s", "id": 211489, "pid": 76337, "tid": -914061504, "ts": 1716454224731178,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224787194, "dur": 5,
    "args": {
      "External id": 211497,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211497,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211497, "pid": 5, "tid": 7, "ts": 1716454224787194,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731207, "dur": 7,
    "args": {
      "External id": 211497,
      "cbid": 211, "correlation": 211497
    }
  },
  {
    "ph": "s", "id": 211497, "pid": 76337, "tid": -914061504, "ts": 1716454224731207,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224787200, "dur": 4,
    "args": {
      "External id": 211505,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211505,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211505, "pid": 5, "tid": 7, "ts": 1716454224787200,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731235, "dur": 9,
    "args": {
      "External id": 211505,
      "cbid": 211, "correlation": 211505
    }
  },
  {
    "ph": "s", "id": 211505, "pid": 76337, "tid": -914061504, "ts": 1716454224731235,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224787206, "dur": 11,
    "args": {
      "External id": 211514,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211514,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211514, "pid": 5, "tid": 7, "ts": 1716454224787206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731279, "dur": 10,
    "args": {
      "External id": 211514,
      "cbid": 211, "correlation": 211514
    }
  },
  {
    "ph": "s", "id": 211514, "pid": 76337, "tid": -914061504, "ts": 1716454224731279,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224787218, "dur": 12,
    "args": {
      "External id": 211534,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211534,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211534, "pid": 5, "tid": 7, "ts": 1716454224787218,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731350, "dur": 12,
    "args": {
      "External id": 211534,
      "cbid": 211, "correlation": 211534
    }
  },
  {
    "ph": "s", "id": 211534, "pid": 76337, "tid": -914061504, "ts": 1716454224731350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224787231, "dur": 4,
    "args": {
      "External id": 211546,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211546,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211546, "pid": 5, "tid": 7, "ts": 1716454224787231,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731371, "dur": 6,
    "args": {
      "External id": 211546,
      "cbid": 211, "correlation": 211546
    }
  },
  {
    "ph": "s", "id": 211546, "pid": 76337, "tid": -914061504, "ts": 1716454224731371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224787237, "dur": 11,
    "args": {
      "External id": 211549,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211549,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211549, "pid": 5, "tid": 7, "ts": 1716454224787237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731388, "dur": 7,
    "args": {
      "External id": 211549,
      "cbid": 211, "correlation": 211549
    }
  },
  {
    "ph": "s", "id": 211549, "pid": 76337, "tid": -914061504, "ts": 1716454224731388,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224787249, "dur": 6,
    "args": {
      "External id": 211558,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211558,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211558, "pid": 5, "tid": 7, "ts": 1716454224787249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731426, "dur": 9,
    "args": {
      "External id": 211558,
      "cbid": 211, "correlation": 211558
    }
  },
  {
    "ph": "s", "id": 211558, "pid": 76337, "tid": -914061504, "ts": 1716454224731426,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731477, "dur": 0,
    "args": {
      "External id": 211568,
      "cbid": 317, "correlation": 211568
    }
  },
  {
    "ph": "f", "id": 211568, "pid": 76337, "tid": -914061504, "ts": 1716454224731477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731478, "dur": 0,
    "args": {
      "External id": 211569,
      "cbid": 203, "correlation": 211569
    }
  },
  {
    "ph": "f", "id": 211569, "pid": 76337, "tid": -914061504, "ts": 1716454224731478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731479, "dur": 0,
    "args": {
      "External id": 211570,
      "cbid": 205, "correlation": 211570
    }
  },
  {
    "ph": "f", "id": 211570, "pid": 76337, "tid": -914061504, "ts": 1716454224731479,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224787257, "dur": 6,
    "args": {
      "External id": 211574,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211574,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211574, "pid": 5, "tid": 7, "ts": 1716454224787257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731492, "dur": 12,
    "args": {
      "External id": 211574,
      "cbid": 211, "correlation": 211574
    }
  },
  {
    "ph": "s", "id": 211574, "pid": 76337, "tid": -914061504, "ts": 1716454224731492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224787264, "dur": 322,
    "args": {
      "External id": 211576,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211576,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211576, "pid": 5, "tid": 7, "ts": 1716454224787264,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731507, "dur": 5,
    "args": {
      "External id": 211576,
      "cbid": 211, "correlation": 211576
    }
  },
  {
    "ph": "s", "id": 211576, "pid": 76337, "tid": -914061504, "ts": 1716454224731507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224787588, "dur": 1,
    "args": {
      "External id": 211578,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211578,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 211578, "pid": 5, "tid": 7, "ts": 1716454224787588,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731518, "dur": 6,
    "args": {
      "External id": 211578,
      "cbid": 51, "correlation": 211578
    }
  },
  {
    "ph": "s", "id": 211578, "pid": 76337, "tid": -914061504, "ts": 1716454224731518,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224787592, "dur": 500,
    "args": {
      "External id": 211579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211579,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211579, "pid": 5, "tid": 7, "ts": 1716454224787592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731525, "dur": 6,
    "args": {
      "External id": 211579,
      "cbid": 211, "correlation": 211579
    }
  },
  {
    "ph": "s", "id": 211579, "pid": 76337, "tid": -914061504, "ts": 1716454224731525,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788093, "dur": 5,
    "args": {
      "External id": 211581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211581,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211581, "pid": 5, "tid": 7, "ts": 1716454224788093,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731534, "dur": 5,
    "args": {
      "External id": 211581,
      "cbid": 211, "correlation": 211581
    }
  },
  {
    "ph": "s", "id": 211581, "pid": 76337, "tid": -914061504, "ts": 1716454224731534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224788100, "dur": 6,
    "args": {
      "External id": 211587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211587,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211587, "pid": 5, "tid": 7, "ts": 1716454224788100,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731563, "dur": 8,
    "args": {
      "External id": 211587,
      "cbid": 211, "correlation": 211587
    }
  },
  {
    "ph": "s", "id": 211587, "pid": 76337, "tid": -914061504, "ts": 1716454224731563,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224788108, "dur": 3,
    "args": {
      "External id": 211595,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211595,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 211595, "pid": 5, "tid": 7, "ts": 1716454224788108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731606, "dur": 9,
    "args": {
      "External id": 211595,
      "cbid": 211, "correlation": 211595
    }
  },
  {
    "ph": "s", "id": 211595, "pid": 76337, "tid": -914061504, "ts": 1716454224731606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731668, "dur": 1,
    "args": {
      "External id": 211611,
      "cbid": 251, "correlation": 211611
    }
  },
  {
    "ph": "f", "id": 211611, "pid": 76337, "tid": -914061504, "ts": 1716454224731668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731673, "dur": 0,
    "args": {
      "External id": 211613,
      "cbid": 251, "correlation": 211613
    }
  },
  {
    "ph": "f", "id": 211613, "pid": 76337, "tid": -914061504, "ts": 1716454224731673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224788112, "dur": 13,
    "args": {
      "External id": 211614,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211614,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211614, "pid": 5, "tid": 7, "ts": 1716454224788112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731675, "dur": 11,
    "args": {
      "External id": 211614,
      "cbid": 211, "correlation": 211614
    }
  },
  {
    "ph": "s", "id": 211614, "pid": 76337, "tid": -914061504, "ts": 1716454224731675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224788126, "dur": 5,
    "args": {
      "External id": 211616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211616,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211616, "pid": 5, "tid": 7, "ts": 1716454224788126,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731688, "dur": 5,
    "args": {
      "External id": 211616,
      "cbid": 211, "correlation": 211616
    }
  },
  {
    "ph": "s", "id": 211616, "pid": 76337, "tid": -914061504, "ts": 1716454224731688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224788133, "dur": 6,
    "args": {
      "External id": 211626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211626,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211626, "pid": 5, "tid": 7, "ts": 1716454224788133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731744, "dur": 13,
    "args": {
      "External id": 211626,
      "cbid": 211, "correlation": 211626
    }
  },
  {
    "ph": "s", "id": 211626, "pid": 76337, "tid": -914061504, "ts": 1716454224731744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788141, "dur": 9,
    "args": {
      "External id": 211646,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211646,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211646, "pid": 5, "tid": 7, "ts": 1716454224788141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731812, "dur": 11,
    "args": {
      "External id": 211646,
      "cbid": 211, "correlation": 211646
    }
  },
  {
    "ph": "s", "id": 211646, "pid": 76337, "tid": -914061504, "ts": 1716454224731812,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224788151, "dur": 4,
    "args": {
      "External id": 211658,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211658,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 211658, "pid": 5, "tid": 7, "ts": 1716454224788151,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731832, "dur": 6,
    "args": {
      "External id": 211658,
      "cbid": 211, "correlation": 211658
    }
  },
  {
    "ph": "s", "id": 211658, "pid": 76337, "tid": -914061504, "ts": 1716454224731832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224788156, "dur": 7,
    "args": {
      "External id": 211661,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211661,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211661, "pid": 5, "tid": 7, "ts": 1716454224788156,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731852, "dur": 6,
    "args": {
      "External id": 211661,
      "cbid": 211, "correlation": 211661
    }
  },
  {
    "ph": "s", "id": 211661, "pid": 76337, "tid": -914061504, "ts": 1716454224731852,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224788164, "dur": 5,
    "args": {
      "External id": 211670,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211670,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211670, "pid": 5, "tid": 7, "ts": 1716454224788164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731892, "dur": 10,
    "args": {
      "External id": 211670,
      "cbid": 211, "correlation": 211670
    }
  },
  {
    "ph": "s", "id": 211670, "pid": 76337, "tid": -914061504, "ts": 1716454224731892,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731955, "dur": 0,
    "args": {
      "External id": 211680,
      "cbid": 317, "correlation": 211680
    }
  },
  {
    "ph": "f", "id": 211680, "pid": 76337, "tid": -914061504, "ts": 1716454224731955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731956, "dur": 0,
    "args": {
      "External id": 211681,
      "cbid": 203, "correlation": 211681
    }
  },
  {
    "ph": "f", "id": 211681, "pid": 76337, "tid": -914061504, "ts": 1716454224731956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731956, "dur": 0,
    "args": {
      "External id": 211682,
      "cbid": 205, "correlation": 211682
    }
  },
  {
    "ph": "f", "id": 211682, "pid": 76337, "tid": -914061504, "ts": 1716454224731956,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788170, "dur": 5,
    "args": {
      "External id": 211686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211686,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211686, "pid": 5, "tid": 7, "ts": 1716454224788170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731970, "dur": 21,
    "args": {
      "External id": 211686,
      "cbid": 211, "correlation": 211686
    }
  },
  {
    "ph": "s", "id": 211686, "pid": 76337, "tid": -914061504, "ts": 1716454224731970,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788177, "dur": 164,
    "args": {
      "External id": 211688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211688,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211688, "pid": 5, "tid": 7, "ts": 1716454224788177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224731993, "dur": 5,
    "args": {
      "External id": 211688,
      "cbid": 211, "correlation": 211688
    }
  },
  {
    "ph": "s", "id": 211688, "pid": 76337, "tid": -914061504, "ts": 1716454224731993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224788343, "dur": 1,
    "args": {
      "External id": 211690,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211690,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 211690, "pid": 5, "tid": 7, "ts": 1716454224788343,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732004, "dur": 7,
    "args": {
      "External id": 211690,
      "cbid": 51, "correlation": 211690
    }
  },
  {
    "ph": "s", "id": 211690, "pid": 76337, "tid": -914061504, "ts": 1716454224732004,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224788346, "dur": 261,
    "args": {
      "External id": 211691,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211691,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211691, "pid": 5, "tid": 7, "ts": 1716454224788346,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732012, "dur": 6,
    "args": {
      "External id": 211691,
      "cbid": 211, "correlation": 211691
    }
  },
  {
    "ph": "s", "id": 211691, "pid": 76337, "tid": -914061504, "ts": 1716454224732012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788609, "dur": 6,
    "args": {
      "External id": 211693,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211693,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211693, "pid": 5, "tid": 7, "ts": 1716454224788609,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732022, "dur": 5,
    "args": {
      "External id": 211693,
      "cbid": 211, "correlation": 211693
    }
  },
  {
    "ph": "s", "id": 211693, "pid": 76337, "tid": -914061504, "ts": 1716454224732022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224788617, "dur": 6,
    "args": {
      "External id": 211699,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211699,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211699, "pid": 5, "tid": 7, "ts": 1716454224788617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732051, "dur": 8,
    "args": {
      "External id": 211699,
      "cbid": 211, "correlation": 211699
    }
  },
  {
    "ph": "s", "id": 211699, "pid": 76337, "tid": -914061504, "ts": 1716454224732051,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732110, "dur": 0,
    "args": {
      "External id": 211709,
      "cbid": 317, "correlation": 211709
    }
  },
  {
    "ph": "f", "id": 211709, "pid": 76337, "tid": -914061504, "ts": 1716454224732110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732111, "dur": 0,
    "args": {
      "External id": 211710,
      "cbid": 203, "correlation": 211710
    }
  },
  {
    "ph": "f", "id": 211710, "pid": 76337, "tid": -914061504, "ts": 1716454224732111,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732112, "dur": 0,
    "args": {
      "External id": 211711,
      "cbid": 205, "correlation": 211711
    }
  },
  {
    "ph": "f", "id": 211711, "pid": 76337, "tid": -914061504, "ts": 1716454224732112,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788624, "dur": 8,
    "args": {
      "External id": 211715,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211715,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211715, "pid": 5, "tid": 7, "ts": 1716454224788624,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732123, "dur": 11,
    "args": {
      "External id": 211715,
      "cbid": 211, "correlation": 211715
    }
  },
  {
    "ph": "s", "id": 211715, "pid": 76337, "tid": -914061504, "ts": 1716454224732123,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224788633, "dur": 3,
    "args": {
      "External id": 211717,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211717,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4563106796116505
    }
  },
  {
    "ph": "f", "id": 211717, "pid": 5, "tid": 7, "ts": 1716454224788633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732140, "dur": 10,
    "args": {
      "External id": 211717,
      "cbid": 51, "correlation": 211717
    }
  },
  {
    "ph": "s", "id": 211717, "pid": 76337, "tid": -914061504, "ts": 1716454224732140,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224788637, "dur": 94,
    "args": {
      "External id": 211718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211718,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 211718, "pid": 5, "tid": 7, "ts": 1716454224788637,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732151, "dur": 6,
    "args": {
      "External id": 211718,
      "cbid": 211, "correlation": 211718
    }
  },
  {
    "ph": "s", "id": 211718, "pid": 76337, "tid": -914061504, "ts": 1716454224732151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788733, "dur": 6,
    "args": {
      "External id": 211720,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211720,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211720, "pid": 5, "tid": 7, "ts": 1716454224788733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732160, "dur": 5,
    "args": {
      "External id": 211720,
      "cbid": 211, "correlation": 211720
    }
  },
  {
    "ph": "s", "id": 211720, "pid": 76337, "tid": -914061504, "ts": 1716454224732160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224788740, "dur": 6,
    "args": {
      "External id": 211726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211726,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211726, "pid": 5, "tid": 7, "ts": 1716454224788740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732187, "dur": 9,
    "args": {
      "External id": 211726,
      "cbid": 211, "correlation": 211726
    }
  },
  {
    "ph": "s", "id": 211726, "pid": 76337, "tid": -914061504, "ts": 1716454224732187,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224788748, "dur": 5,
    "args": {
      "External id": 211734,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211734,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211734, "pid": 5, "tid": 7, "ts": 1716454224788748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732216, "dur": 8,
    "args": {
      "External id": 211734,
      "cbid": 211, "correlation": 211734
    }
  },
  {
    "ph": "s", "id": 211734, "pid": 76337, "tid": -914061504, "ts": 1716454224732216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224788754, "dur": 4,
    "args": {
      "External id": 211742,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211742,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211742, "pid": 5, "tid": 7, "ts": 1716454224788754,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732245, "dur": 8,
    "args": {
      "External id": 211742,
      "cbid": 211, "correlation": 211742
    }
  },
  {
    "ph": "s", "id": 211742, "pid": 76337, "tid": -914061504, "ts": 1716454224732245,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224788760, "dur": 11,
    "args": {
      "External id": 211751,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211751,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211751, "pid": 5, "tid": 7, "ts": 1716454224788760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732287, "dur": 10,
    "args": {
      "External id": 211751,
      "cbid": 211, "correlation": 211751
    }
  },
  {
    "ph": "s", "id": 211751, "pid": 76337, "tid": -914061504, "ts": 1716454224732287,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788772, "dur": 12,
    "args": {
      "External id": 211771,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211771,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211771, "pid": 5, "tid": 7, "ts": 1716454224788772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732357, "dur": 11,
    "args": {
      "External id": 211771,
      "cbid": 211, "correlation": 211771
    }
  },
  {
    "ph": "s", "id": 211771, "pid": 76337, "tid": -914061504, "ts": 1716454224732357,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224788786, "dur": 4,
    "args": {
      "External id": 211783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211783,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211783, "pid": 5, "tid": 7, "ts": 1716454224788786,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732378, "dur": 6,
    "args": {
      "External id": 211783,
      "cbid": 211, "correlation": 211783
    }
  },
  {
    "ph": "s", "id": 211783, "pid": 76337, "tid": -914061504, "ts": 1716454224732378,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224788791, "dur": 11,
    "args": {
      "External id": 211786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211786,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211786, "pid": 5, "tid": 7, "ts": 1716454224788791,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732395, "dur": 7,
    "args": {
      "External id": 211786,
      "cbid": 211, "correlation": 211786
    }
  },
  {
    "ph": "s", "id": 211786, "pid": 76337, "tid": -914061504, "ts": 1716454224732395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224788804, "dur": 6,
    "args": {
      "External id": 211795,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211795,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 24,
      "warps per SM": 96,
      "grid": [1920, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211795, "pid": 5, "tid": 7, "ts": 1716454224788804,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732432, "dur": 9,
    "args": {
      "External id": 211795,
      "cbid": 211, "correlation": 211795
    }
  },
  {
    "ph": "s", "id": 211795, "pid": 76337, "tid": -914061504, "ts": 1716454224732432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732482, "dur": 0,
    "args": {
      "External id": 211805,
      "cbid": 317, "correlation": 211805
    }
  },
  {
    "ph": "f", "id": 211805, "pid": 76337, "tid": -914061504, "ts": 1716454224732482,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732483, "dur": 0,
    "args": {
      "External id": 211806,
      "cbid": 203, "correlation": 211806
    }
  },
  {
    "ph": "f", "id": 211806, "pid": 76337, "tid": -914061504, "ts": 1716454224732483,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732484, "dur": 0,
    "args": {
      "External id": 211807,
      "cbid": 205, "correlation": 211807
    }
  },
  {
    "ph": "f", "id": 211807, "pid": 76337, "tid": -914061504, "ts": 1716454224732484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788812, "dur": 7,
    "args": {
      "External id": 211811,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211811,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211811, "pid": 5, "tid": 7, "ts": 1716454224788812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732498, "dur": 11,
    "args": {
      "External id": 211811,
      "cbid": 211, "correlation": 211811
    }
  },
  {
    "ph": "s", "id": 211811, "pid": 76337, "tid": -914061504, "ts": 1716454224732498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224788820, "dur": 322,
    "args": {
      "External id": 211813,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211813,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211813, "pid": 5, "tid": 7, "ts": 1716454224788820,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732511, "dur": 5,
    "args": {
      "External id": 211813,
      "cbid": 211, "correlation": 211813
    }
  },
  {
    "ph": "s", "id": 211813, "pid": 76337, "tid": -914061504, "ts": 1716454224732511,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224789144, "dur": 1,
    "args": {
      "External id": 211815,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211815,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 211815, "pid": 5, "tid": 7, "ts": 1716454224789144,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732522, "dur": 6,
    "args": {
      "External id": 211815,
      "cbid": 51, "correlation": 211815
    }
  },
  {
    "ph": "s", "id": 211815, "pid": 76337, "tid": -914061504, "ts": 1716454224732522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224789148, "dur": 499,
    "args": {
      "External id": 211816,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211816,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211816, "pid": 5, "tid": 7, "ts": 1716454224789148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732529, "dur": 6,
    "args": {
      "External id": 211816,
      "cbid": 211, "correlation": 211816
    }
  },
  {
    "ph": "s", "id": 211816, "pid": 76337, "tid": -914061504, "ts": 1716454224732529,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224789648, "dur": 5,
    "args": {
      "External id": 211818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211818,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211818, "pid": 5, "tid": 7, "ts": 1716454224789648,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732539, "dur": 6,
    "args": {
      "External id": 211818,
      "cbid": 211, "correlation": 211818
    }
  },
  {
    "ph": "s", "id": 211818, "pid": 76337, "tid": -914061504, "ts": 1716454224732539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224789654, "dur": 7,
    "args": {
      "External id": 211824,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211824,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211824, "pid": 5, "tid": 7, "ts": 1716454224789654,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732568, "dur": 8,
    "args": {
      "External id": 211824,
      "cbid": 211, "correlation": 211824
    }
  },
  {
    "ph": "s", "id": 211824, "pid": 76337, "tid": -914061504, "ts": 1716454224732568,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224789662, "dur": 3,
    "args": {
      "External id": 211832,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211832,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 211832, "pid": 5, "tid": 7, "ts": 1716454224789662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732612, "dur": 9,
    "args": {
      "External id": 211832,
      "cbid": 211, "correlation": 211832
    }
  },
  {
    "ph": "s", "id": 211832, "pid": 76337, "tid": -914061504, "ts": 1716454224732612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732674, "dur": 2,
    "args": {
      "External id": 211848,
      "cbid": 251, "correlation": 211848
    }
  },
  {
    "ph": "f", "id": 211848, "pid": 76337, "tid": -914061504, "ts": 1716454224732674,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732680, "dur": 0,
    "args": {
      "External id": 211850,
      "cbid": 251, "correlation": 211850
    }
  },
  {
    "ph": "f", "id": 211850, "pid": 76337, "tid": -914061504, "ts": 1716454224732680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224789667, "dur": 12,
    "args": {
      "External id": 211851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211851,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211851, "pid": 5, "tid": 7, "ts": 1716454224789667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732682, "dur": 12,
    "args": {
      "External id": 211851,
      "cbid": 211, "correlation": 211851
    }
  },
  {
    "ph": "s", "id": 211851, "pid": 76337, "tid": -914061504, "ts": 1716454224732682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224789680, "dur": 5,
    "args": {
      "External id": 211853,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211853,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211853, "pid": 5, "tid": 7, "ts": 1716454224789680,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732696, "dur": 5,
    "args": {
      "External id": 211853,
      "cbid": 211, "correlation": 211853
    }
  },
  {
    "ph": "s", "id": 211853, "pid": 76337, "tid": -914061504, "ts": 1716454224732696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224789687, "dur": 6,
    "args": {
      "External id": 211863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211863,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211863, "pid": 5, "tid": 7, "ts": 1716454224789687,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732752, "dur": 12,
    "args": {
      "External id": 211863,
      "cbid": 211, "correlation": 211863
    }
  },
  {
    "ph": "s", "id": 211863, "pid": 76337, "tid": -914061504, "ts": 1716454224732752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224789694, "dur": 9,
    "args": {
      "External id": 211883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211883,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 211883, "pid": 5, "tid": 7, "ts": 1716454224789694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732817, "dur": 11,
    "args": {
      "External id": 211883,
      "cbid": 211, "correlation": 211883
    }
  },
  {
    "ph": "s", "id": 211883, "pid": 76337, "tid": -914061504, "ts": 1716454224732817,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224789705, "dur": 4,
    "args": {
      "External id": 211895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211895,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 211895, "pid": 5, "tid": 7, "ts": 1716454224789705,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732838, "dur": 6,
    "args": {
      "External id": 211895,
      "cbid": 211, "correlation": 211895
    }
  },
  {
    "ph": "s", "id": 211895, "pid": 76337, "tid": -914061504, "ts": 1716454224732838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224789710, "dur": 6,
    "args": {
      "External id": 211898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211898,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211898, "pid": 5, "tid": 7, "ts": 1716454224789710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732857, "dur": 7,
    "args": {
      "External id": 211898,
      "cbid": 211, "correlation": 211898
    }
  },
  {
    "ph": "s", "id": 211898, "pid": 76337, "tid": -914061504, "ts": 1716454224732857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224789718, "dur": 5,
    "args": {
      "External id": 211907,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211907,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211907, "pid": 5, "tid": 7, "ts": 1716454224789718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732898, "dur": 10,
    "args": {
      "External id": 211907,
      "cbid": 211, "correlation": 211907
    }
  },
  {
    "ph": "s", "id": 211907, "pid": 76337, "tid": -914061504, "ts": 1716454224732898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732961, "dur": 0,
    "args": {
      "External id": 211917,
      "cbid": 317, "correlation": 211917
    }
  },
  {
    "ph": "f", "id": 211917, "pid": 76337, "tid": -914061504, "ts": 1716454224732961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732962, "dur": 0,
    "args": {
      "External id": 211918,
      "cbid": 203, "correlation": 211918
    }
  },
  {
    "ph": "f", "id": 211918, "pid": 76337, "tid": -914061504, "ts": 1716454224732962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732962, "dur": 0,
    "args": {
      "External id": 211919,
      "cbid": 205, "correlation": 211919
    }
  },
  {
    "ph": "f", "id": 211919, "pid": 76337, "tid": -914061504, "ts": 1716454224732962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224789724, "dur": 5,
    "args": {
      "External id": 211923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211923,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211923, "pid": 5, "tid": 7, "ts": 1716454224789724,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224732984, "dur": 13,
    "args": {
      "External id": 211923,
      "cbid": 211, "correlation": 211923
    }
  },
  {
    "ph": "s", "id": 211923, "pid": 76337, "tid": -914061504, "ts": 1716454224732984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224789730, "dur": 163,
    "args": {
      "External id": 211925,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211925,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211925, "pid": 5, "tid": 7, "ts": 1716454224789730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733000, "dur": 5,
    "args": {
      "External id": 211925,
      "cbid": 211, "correlation": 211925
    }
  },
  {
    "ph": "s", "id": 211925, "pid": 76337, "tid": -914061504, "ts": 1716454224733000,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224789896, "dur": 1,
    "args": {
      "External id": 211927,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211927,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 211927, "pid": 5, "tid": 7, "ts": 1716454224789896,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733010, "dur": 8,
    "args": {
      "External id": 211927,
      "cbid": 51, "correlation": 211927
    }
  },
  {
    "ph": "s", "id": 211927, "pid": 76337, "tid": -914061504, "ts": 1716454224733010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224789900, "dur": 261,
    "args": {
      "External id": 211928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211928,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 1.125,
      "warps per SM": 9,
      "grid": [5, 6, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 211928, "pid": 5, "tid": 7, "ts": 1716454224789900,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733019, "dur": 6,
    "args": {
      "External id": 211928,
      "cbid": 211, "correlation": 211928
    }
  },
  {
    "ph": "s", "id": 211928, "pid": 76337, "tid": -914061504, "ts": 1716454224733019,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224790162, "dur": 6,
    "args": {
      "External id": 211930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211930,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211930, "pid": 5, "tid": 7, "ts": 1716454224790162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733028, "dur": 5,
    "args": {
      "External id": 211930,
      "cbid": 211, "correlation": 211930
    }
  },
  {
    "ph": "s", "id": 211930, "pid": 76337, "tid": -914061504, "ts": 1716454224733028,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224790169, "dur": 6,
    "args": {
      "External id": 211936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211936,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211936, "pid": 5, "tid": 7, "ts": 1716454224790169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733058, "dur": 9,
    "args": {
      "External id": 211936,
      "cbid": 211, "correlation": 211936
    }
  },
  {
    "ph": "s", "id": 211936, "pid": 76337, "tid": -914061504, "ts": 1716454224733058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733117, "dur": 0,
    "args": {
      "External id": 211946,
      "cbid": 317, "correlation": 211946
    }
  },
  {
    "ph": "f", "id": 211946, "pid": 76337, "tid": -914061504, "ts": 1716454224733117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733118, "dur": 0,
    "args": {
      "External id": 211947,
      "cbid": 203, "correlation": 211947
    }
  },
  {
    "ph": "f", "id": 211947, "pid": 76337, "tid": -914061504, "ts": 1716454224733118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733119, "dur": 0,
    "args": {
      "External id": 211948,
      "cbid": 205, "correlation": 211948
    }
  },
  {
    "ph": "f", "id": 211948, "pid": 76337, "tid": -914061504, "ts": 1716454224733119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224790177, "dur": 8,
    "args": {
      "External id": 211952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211952,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 16,
      "warps per SM": 128,
      "grid": [2, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211952, "pid": 5, "tid": 7, "ts": 1716454224790177,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733130, "dur": 12,
    "args": {
      "External id": 211952,
      "cbid": 211, "correlation": 211952
    }
  },
  {
    "ph": "s", "id": 211952, "pid": 76337, "tid": -914061504, "ts": 1716454224733130,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224790186, "dur": 3,
    "args": {
      "External id": 211954,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 211954,
      "bytes": 4800, "memory bandwidth (GB/s)": 1.4851485148514851
    }
  },
  {
    "ph": "f", "id": 211954, "pid": 5, "tid": 7, "ts": 1716454224790186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733148, "dur": 9,
    "args": {
      "External id": 211954,
      "cbid": 51, "correlation": 211954
    }
  },
  {
    "ph": "s", "id": 211954, "pid": 76337, "tid": -914061504, "ts": 1716454224733148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x64x64_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224790190, "dur": 94,
    "args": {
      "External id": 211955,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211955,
      "registers per thread": 154,
      "shared memory": 16384,
      "blocks per SM": 15,
      "warps per SM": 60,
      "grid": [20, 6, 10],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 211955, "pid": 5, "tid": 7, "ts": 1716454224790190,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733158, "dur": 6,
    "args": {
      "External id": 211955,
      "cbid": 211, "correlation": 211955
    }
  },
  {
    "ph": "s", "id": 211955, "pid": 76337, "tid": -914061504, "ts": 1716454224733158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224790285, "dur": 6,
    "args": {
      "External id": 211957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211957,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 8,
      "warps per SM": 64,
      "grid": [2, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211957, "pid": 5, "tid": 7, "ts": 1716454224790285,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733167, "dur": 5,
    "args": {
      "External id": 211957,
      "cbid": 211, "correlation": 211957
    }
  },
  {
    "ph": "s", "id": 211957, "pid": 76337, "tid": -914061504, "ts": 1716454224733167,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224790292, "dur": 6,
    "args": {
      "External id": 211963,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211963,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211963, "pid": 5, "tid": 7, "ts": 1716454224790292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733195, "dur": 8,
    "args": {
      "External id": 211963,
      "cbid": 211, "correlation": 211963
    }
  },
  {
    "ph": "s", "id": 211963, "pid": 76337, "tid": -914061504, "ts": 1716454224733195,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224790300, "dur": 5,
    "args": {
      "External id": 211971,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211971,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211971, "pid": 5, "tid": 7, "ts": 1716454224790300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733223, "dur": 8,
    "args": {
      "External id": 211971,
      "cbid": 211, "correlation": 211971
    }
  },
  {
    "ph": "s", "id": 211971, "pid": 76337, "tid": -914061504, "ts": 1716454224733223,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224790306, "dur": 4,
    "args": {
      "External id": 211979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211979,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [960, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 75
    }
  },
  {
    "ph": "f", "id": 211979, "pid": 5, "tid": 7, "ts": 1716454224790306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733252, "dur": 8,
    "args": {
      "External id": 211979,
      "cbid": 211, "correlation": 211979
    }
  },
  {
    "ph": "s", "id": 211979, "pid": 76337, "tid": -914061504, "ts": 1716454224733252,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224790312, "dur": 14,
    "args": {
      "External id": 211990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 211990,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 16,
      "warps per SM": 256,
      "grid": [2, 1, 640],
      "block": [8, 16, 4],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 211990, "pid": 5, "tid": 7, "ts": 1716454224790312,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733340, "dur": 24,
    "args": {
      "External id": 211990,
      "cbid": 211, "correlation": 211990
    }
  },
  {
    "ph": "s", "id": 211990, "pid": 76337, "tid": -914061504, "ts": 1716454224733340,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733408, "dur": 0,
    "args": {
      "External id": 212000,
      "cbid": 317, "correlation": 212000
    }
  },
  {
    "ph": "f", "id": 212000, "pid": 76337, "tid": -914061504, "ts": 1716454224733408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733409, "dur": 0,
    "args": {
      "External id": 212001,
      "cbid": 203, "correlation": 212001
    }
  },
  {
    "ph": "f", "id": 212001, "pid": 76337, "tid": -914061504, "ts": 1716454224733409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733410, "dur": 0,
    "args": {
      "External id": 212002,
      "cbid": 205, "correlation": 212002
    }
  },
  {
    "ph": "f", "id": 212002, "pid": 76337, "tid": -914061504, "ts": 1716454224733410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224790328, "dur": 9,
    "args": {
      "External id": 212006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212006,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212006, "pid": 5, "tid": 7, "ts": 1716454224790328,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733424, "dur": 12,
    "args": {
      "External id": 212006,
      "cbid": 211, "correlation": 212006
    }
  },
  {
    "ph": "s", "id": 212006, "pid": 76337, "tid": -914061504, "ts": 1716454224733424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224790338, "dur": 164,
    "args": {
      "External id": 212008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212008,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212008, "pid": 5, "tid": 7, "ts": 1716454224790338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733438, "dur": 6,
    "args": {
      "External id": 212008,
      "cbid": 211, "correlation": 212008
    }
  },
  {
    "ph": "s", "id": 212008, "pid": 76337, "tid": -914061504, "ts": 1716454224733438,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224790504, "dur": 1,
    "args": {
      "External id": 212010,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 212010,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 212010, "pid": 5, "tid": 7, "ts": 1716454224790504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733449, "dur": 6,
    "args": {
      "External id": 212010,
      "cbid": 51, "correlation": 212010
    }
  },
  {
    "ph": "s", "id": 212010, "pid": 76337, "tid": -914061504, "ts": 1716454224733449,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224790508, "dur": 654,
    "args": {
      "External id": 212011,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212011,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212011, "pid": 5, "tid": 7, "ts": 1716454224790508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733457, "dur": 6,
    "args": {
      "External id": 212011,
      "cbid": 211, "correlation": 212011
    }
  },
  {
    "ph": "s", "id": 212011, "pid": 76337, "tid": -914061504, "ts": 1716454224733457,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224791164, "dur": 12,
    "args": {
      "External id": 212013,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212013,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212013, "pid": 5, "tid": 7, "ts": 1716454224791164,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733466, "dur": 5,
    "args": {
      "External id": 212013,
      "cbid": 211, "correlation": 212013
    }
  },
  {
    "ph": "s", "id": 212013, "pid": 76337, "tid": -914061504, "ts": 1716454224733466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224791178, "dur": 14,
    "args": {
      "External id": 212019,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212019,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212019, "pid": 5, "tid": 7, "ts": 1716454224791178,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733496, "dur": 9,
    "args": {
      "External id": 212019,
      "cbid": 211, "correlation": 212019
    }
  },
  {
    "ph": "s", "id": 212019, "pid": 76337, "tid": -914061504, "ts": 1716454224733496,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224791193, "dur": 29,
    "args": {
      "External id": 212028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212028,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212028, "pid": 5, "tid": 7, "ts": 1716454224791193,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733586, "dur": 13,
    "args": {
      "External id": 212028,
      "cbid": 211, "correlation": 212028
    }
  },
  {
    "ph": "s", "id": 212028, "pid": 76337, "tid": -914061504, "ts": 1716454224733586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224791224, "dur": 30,
    "args": {
      "External id": 212048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212048,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 212048, "pid": 5, "tid": 7, "ts": 1716454224791224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733654, "dur": 11,
    "args": {
      "External id": 212048,
      "cbid": 211, "correlation": 212048
    }
  },
  {
    "ph": "s", "id": 212048, "pid": 76337, "tid": -914061504, "ts": 1716454224733654,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224791255, "dur": 4,
    "args": {
      "External id": 212060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212060,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212060, "pid": 5, "tid": 7, "ts": 1716454224791255,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733675, "dur": 6,
    "args": {
      "External id": 212060,
      "cbid": 211, "correlation": 212060
    }
  },
  {
    "ph": "s", "id": 212060, "pid": 76337, "tid": -914061504, "ts": 1716454224733675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224791261, "dur": 30,
    "args": {
      "External id": 212063,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212063,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212063, "pid": 5, "tid": 7, "ts": 1716454224791261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733694, "dur": 6,
    "args": {
      "External id": 212063,
      "cbid": 211, "correlation": 212063
    }
  },
  {
    "ph": "s", "id": 212063, "pid": 76337, "tid": -914061504, "ts": 1716454224733694,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224791292, "dur": 21,
    "args": {
      "External id": 212072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212072,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212072, "pid": 5, "tid": 7, "ts": 1716454224791292,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733731, "dur": 10,
    "args": {
      "External id": 212072,
      "cbid": 211, "correlation": 212072
    }
  },
  {
    "ph": "s", "id": 212072, "pid": 76337, "tid": -914061504, "ts": 1716454224733731,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733782, "dur": 0,
    "args": {
      "External id": 212082,
      "cbid": 317, "correlation": 212082
    }
  },
  {
    "ph": "f", "id": 212082, "pid": 76337, "tid": -914061504, "ts": 1716454224733782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733783, "dur": 0,
    "args": {
      "External id": 212083,
      "cbid": 203, "correlation": 212083
    }
  },
  {
    "ph": "f", "id": 212083, "pid": 76337, "tid": -914061504, "ts": 1716454224733783,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733784, "dur": 0,
    "args": {
      "External id": 212084,
      "cbid": 205, "correlation": 212084
    }
  },
  {
    "ph": "f", "id": 212084, "pid": 76337, "tid": -914061504, "ts": 1716454224733784,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224791314, "dur": 23,
    "args": {
      "External id": 212088,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212088,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212088, "pid": 5, "tid": 7, "ts": 1716454224791314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733799, "dur": 12,
    "args": {
      "External id": 212088,
      "cbid": 211, "correlation": 212088
    }
  },
  {
    "ph": "s", "id": 212088, "pid": 76337, "tid": -914061504, "ts": 1716454224733799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224791338, "dur": 324,
    "args": {
      "External id": 212090,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212090,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212090, "pid": 5, "tid": 7, "ts": 1716454224791338,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733813, "dur": 5,
    "args": {
      "External id": 212090,
      "cbid": 211, "correlation": 212090
    }
  },
  {
    "ph": "s", "id": 212090, "pid": 76337, "tid": -914061504, "ts": 1716454224733813,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224791664, "dur": 1,
    "args": {
      "External id": 212092,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 212092,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 212092, "pid": 5, "tid": 7, "ts": 1716454224791664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733824, "dur": 7,
    "args": {
      "External id": 212092,
      "cbid": 51, "correlation": 212092
    }
  },
  {
    "ph": "s", "id": 212092, "pid": 76337, "tid": -914061504, "ts": 1716454224733824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224791668, "dur": 1248,
    "args": {
      "External id": 212093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212093,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212093, "pid": 5, "tid": 7, "ts": 1716454224791668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733832, "dur": 7,
    "args": {
      "External id": 212093,
      "cbid": 211, "correlation": 212093
    }
  },
  {
    "ph": "s", "id": 212093, "pid": 76337, "tid": -914061504, "ts": 1716454224733832,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224792918, "dur": 13,
    "args": {
      "External id": 212095,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212095,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212095, "pid": 5, "tid": 7, "ts": 1716454224792918,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733844, "dur": 5,
    "args": {
      "External id": 212095,
      "cbid": 211, "correlation": 212095
    }
  },
  {
    "ph": "s", "id": 212095, "pid": 76337, "tid": -914061504, "ts": 1716454224733844,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224792932, "dur": 15,
    "args": {
      "External id": 212101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212101, "pid": 5, "tid": 7, "ts": 1716454224792932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733871, "dur": 8,
    "args": {
      "External id": 212101,
      "cbid": 211, "correlation": 212101
    }
  },
  {
    "ph": "s", "id": 212101, "pid": 76337, "tid": -914061504, "ts": 1716454224733871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224792948, "dur": 3,
    "args": {
      "External id": 212109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212109,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 212109, "pid": 5, "tid": 7, "ts": 1716454224792948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733915, "dur": 9,
    "args": {
      "External id": 212109,
      "cbid": 211, "correlation": 212109
    }
  },
  {
    "ph": "s", "id": 212109, "pid": 76337, "tid": -914061504, "ts": 1716454224733915,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733987, "dur": 2,
    "args": {
      "External id": 212125,
      "cbid": 251, "correlation": 212125
    }
  },
  {
    "ph": "f", "id": 212125, "pid": 76337, "tid": -914061504, "ts": 1716454224733987,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733993, "dur": 0,
    "args": {
      "External id": 212127,
      "cbid": 251, "correlation": 212127
    }
  },
  {
    "ph": "f", "id": 212127, "pid": 76337, "tid": -914061504, "ts": 1716454224733993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224792953, "dur": 12,
    "args": {
      "External id": 212128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212128,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212128, "pid": 5, "tid": 7, "ts": 1716454224792953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224733995, "dur": 12,
    "args": {
      "External id": 212128,
      "cbid": 211, "correlation": 212128
    }
  },
  {
    "ph": "s", "id": 212128, "pid": 76337, "tid": -914061504, "ts": 1716454224733995,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224792966, "dur": 5,
    "args": {
      "External id": 212130,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212130,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212130, "pid": 5, "tid": 7, "ts": 1716454224792966,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734008, "dur": 5,
    "args": {
      "External id": 212130,
      "cbid": 211, "correlation": 212130
    }
  },
  {
    "ph": "s", "id": 212130, "pid": 76337, "tid": -914061504, "ts": 1716454224734008,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224792972, "dur": 17,
    "args": {
      "External id": 212140,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212140,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212140, "pid": 5, "tid": 7, "ts": 1716454224792972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734066, "dur": 12,
    "args": {
      "External id": 212140,
      "cbid": 211, "correlation": 212140
    }
  },
  {
    "ph": "s", "id": 212140, "pid": 76337, "tid": -914061504, "ts": 1716454224734066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224792990, "dur": 17,
    "args": {
      "External id": 212160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212160,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 212160, "pid": 5, "tid": 7, "ts": 1716454224792990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734134, "dur": 11,
    "args": {
      "External id": 212160,
      "cbid": 211, "correlation": 212160
    }
  },
  {
    "ph": "s", "id": 212160, "pid": 76337, "tid": -914061504, "ts": 1716454224734134,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224793009, "dur": 4,
    "args": {
      "External id": 212172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212172,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 212172, "pid": 5, "tid": 7, "ts": 1716454224793009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734155, "dur": 6,
    "args": {
      "External id": 212172,
      "cbid": 211, "correlation": 212172
    }
  },
  {
    "ph": "s", "id": 212172, "pid": 76337, "tid": -914061504, "ts": 1716454224734155,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224793014, "dur": 17,
    "args": {
      "External id": 212175,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212175,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212175, "pid": 5, "tid": 7, "ts": 1716454224793014,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734172, "dur": 6,
    "args": {
      "External id": 212175,
      "cbid": 211, "correlation": 212175
    }
  },
  {
    "ph": "s", "id": 212175, "pid": 76337, "tid": -914061504, "ts": 1716454224734172,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224793032, "dur": 11,
    "args": {
      "External id": 212184,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212184,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212184, "pid": 5, "tid": 7, "ts": 1716454224793032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734212, "dur": 10,
    "args": {
      "External id": 212184,
      "cbid": 211, "correlation": 212184
    }
  },
  {
    "ph": "s", "id": 212184, "pid": 76337, "tid": -914061504, "ts": 1716454224734212,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734273, "dur": 0,
    "args": {
      "External id": 212194,
      "cbid": 317, "correlation": 212194
    }
  },
  {
    "ph": "f", "id": 212194, "pid": 76337, "tid": -914061504, "ts": 1716454224734273,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734274, "dur": 0,
    "args": {
      "External id": 212195,
      "cbid": 203, "correlation": 212195
    }
  },
  {
    "ph": "f", "id": 212195, "pid": 76337, "tid": -914061504, "ts": 1716454224734274,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734275, "dur": 0,
    "args": {
      "External id": 212196,
      "cbid": 205, "correlation": 212196
    }
  },
  {
    "ph": "f", "id": 212196, "pid": 76337, "tid": -914061504, "ts": 1716454224734275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224793045, "dur": 12,
    "args": {
      "External id": 212200,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212200,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212200, "pid": 5, "tid": 7, "ts": 1716454224793045,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734289, "dur": 12,
    "args": {
      "External id": 212200,
      "cbid": 211, "correlation": 212200
    }
  },
  {
    "ph": "s", "id": 212200, "pid": 76337, "tid": -914061504, "ts": 1716454224734289,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224793058, "dur": 165,
    "args": {
      "External id": 212202,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212202,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212202, "pid": 5, "tid": 7, "ts": 1716454224793058,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734303, "dur": 5,
    "args": {
      "External id": 212202,
      "cbid": 211, "correlation": 212202
    }
  },
  {
    "ph": "s", "id": 212202, "pid": 76337, "tid": -914061504, "ts": 1716454224734303,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224793225, "dur": 1,
    "args": {
      "External id": 212204,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 212204,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5878750765462339
    }
  },
  {
    "ph": "f", "id": 212204, "pid": 5, "tid": 7, "ts": 1716454224793225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734314, "dur": 6,
    "args": {
      "External id": 212204,
      "cbid": 51, "correlation": 212204
    }
  },
  {
    "ph": "s", "id": 212204, "pid": 76337, "tid": -914061504, "ts": 1716454224734314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224793229, "dur": 652,
    "args": {
      "External id": 212205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212205,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212205, "pid": 5, "tid": 7, "ts": 1716454224793229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734321, "dur": 6,
    "args": {
      "External id": 212205,
      "cbid": 211, "correlation": 212205
    }
  },
  {
    "ph": "s", "id": 212205, "pid": 76337, "tid": -914061504, "ts": 1716454224734321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224793882, "dur": 13,
    "args": {
      "External id": 212207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212207,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212207, "pid": 5, "tid": 7, "ts": 1716454224793882,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734331, "dur": 5,
    "args": {
      "External id": 212207,
      "cbid": 211, "correlation": 212207
    }
  },
  {
    "ph": "s", "id": 212207, "pid": 76337, "tid": -914061504, "ts": 1716454224734331,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224793897, "dur": 15,
    "args": {
      "External id": 212213,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212213,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212213, "pid": 5, "tid": 7, "ts": 1716454224793897,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734359, "dur": 8,
    "args": {
      "External id": 212213,
      "cbid": 211, "correlation": 212213
    }
  },
  {
    "ph": "s", "id": 212213, "pid": 76337, "tid": -914061504, "ts": 1716454224734359,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734417, "dur": 0,
    "args": {
      "External id": 212223,
      "cbid": 317, "correlation": 212223
    }
  },
  {
    "ph": "f", "id": 212223, "pid": 76337, "tid": -914061504, "ts": 1716454224734417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734418, "dur": 0,
    "args": {
      "External id": 212224,
      "cbid": 203, "correlation": 212224
    }
  },
  {
    "ph": "f", "id": 212224, "pid": 76337, "tid": -914061504, "ts": 1716454224734418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734418, "dur": 0,
    "args": {
      "External id": 212225,
      "cbid": 205, "correlation": 212225
    }
  },
  {
    "ph": "f", "id": 212225, "pid": 76337, "tid": -914061504, "ts": 1716454224734418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224793913, "dur": 21,
    "args": {
      "External id": 212229,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212229,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212229, "pid": 5, "tid": 7, "ts": 1716454224793913,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734432, "dur": 12,
    "args": {
      "External id": 212229,
      "cbid": 211, "correlation": 212229
    }
  },
  {
    "ph": "s", "id": 212229, "pid": 76337, "tid": -914061504, "ts": 1716454224734432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224793936, "dur": 4,
    "args": {
      "External id": 212231,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212231,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212231, "pid": 5, "tid": 7, "ts": 1716454224793936,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734450, "dur": 6,
    "args": {
      "External id": 212231,
      "cbid": 211, "correlation": 212231
    }
  },
  {
    "ph": "s", "id": 212231, "pid": 76337, "tid": -914061504, "ts": 1716454224734450,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734460, "dur": 0,
    "args": {
      "External id": 212232,
      "cbid": 51, "correlation": 212232
    }
  },
  {
    "ph": "s", "id": 212232, "pid": 76337, "tid": -914061504, "ts": 1716454224734460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224793941, "dur": 175,
    "args": {
      "External id": 212233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212233,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 212233, "pid": 5, "tid": 7, "ts": 1716454224793941,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734460, "dur": 5,
    "args": {
      "External id": 212233,
      "cbid": 211, "correlation": 212233
    }
  },
  {
    "ph": "s", "id": 212233, "pid": 76337, "tid": -914061504, "ts": 1716454224734460,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224794117, "dur": 16,
    "args": {
      "External id": 212238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212238,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212238, "pid": 5, "tid": 7, "ts": 1716454224794117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734486, "dur": 8,
    "args": {
      "External id": 212238,
      "cbid": 211, "correlation": 212238
    }
  },
  {
    "ph": "s", "id": 212238, "pid": 76337, "tid": -914061504, "ts": 1716454224734486,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224794135, "dur": 12,
    "args": {
      "External id": 212246,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212246,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212246, "pid": 5, "tid": 7, "ts": 1716454224794135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734514, "dur": 8,
    "args": {
      "External id": 212246,
      "cbid": 211, "correlation": 212246
    }
  },
  {
    "ph": "s", "id": 212246, "pid": 76337, "tid": -914061504, "ts": 1716454224734514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224794148, "dur": 10,
    "args": {
      "External id": 212254,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212254,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212254, "pid": 5, "tid": 7, "ts": 1716454224794148,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734543, "dur": 8,
    "args": {
      "External id": 212254,
      "cbid": 211, "correlation": 212254
    }
  },
  {
    "ph": "s", "id": 212254, "pid": 76337, "tid": -914061504, "ts": 1716454224734543,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224794159, "dur": 19,
    "args": {
      "External id": 212274,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212274,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 212274, "pid": 5, "tid": 7, "ts": 1716454224794159,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734624, "dur": 13,
    "args": {
      "External id": 212274,
      "cbid": 211, "correlation": 212274
    }
  },
  {
    "ph": "s", "id": 212274, "pid": 76337, "tid": -914061504, "ts": 1716454224734624,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224794179, "dur": 5,
    "args": {
      "External id": 212286,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212286,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 212286, "pid": 5, "tid": 7, "ts": 1716454224794179,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734647, "dur": 6,
    "args": {
      "External id": 212286,
      "cbid": 211, "correlation": 212286
    }
  },
  {
    "ph": "s", "id": 212286, "pid": 76337, "tid": -914061504, "ts": 1716454224734647,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224794185, "dur": 16,
    "args": {
      "External id": 212289,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212289,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212289, "pid": 5, "tid": 7, "ts": 1716454224794185,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734664, "dur": 7,
    "args": {
      "External id": 212289,
      "cbid": 211, "correlation": 212289
    }
  },
  {
    "ph": "s", "id": 212289, "pid": 76337, "tid": -914061504, "ts": 1716454224734664,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734721, "dur": 0,
    "args": {
      "External id": 212300,
      "cbid": 317, "correlation": 212300
    }
  },
  {
    "ph": "f", "id": 212300, "pid": 76337, "tid": -914061504, "ts": 1716454224734721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734722, "dur": 0,
    "args": {
      "External id": 212301,
      "cbid": 203, "correlation": 212301
    }
  },
  {
    "ph": "f", "id": 212301, "pid": 76337, "tid": -914061504, "ts": 1716454224734722,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734723, "dur": 0,
    "args": {
      "External id": 212302,
      "cbid": 205, "correlation": 212302
    }
  },
  {
    "ph": "f", "id": 212302, "pid": 76337, "tid": -914061504, "ts": 1716454224734723,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224794203, "dur": 12,
    "args": {
      "External id": 212306,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212306,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212306, "pid": 5, "tid": 7, "ts": 1716454224794203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734736, "dur": 12,
    "args": {
      "External id": 212306,
      "cbid": 211, "correlation": 212306
    }
  },
  {
    "ph": "s", "id": 212306, "pid": 76337, "tid": -914061504, "ts": 1716454224734736,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224794216, "dur": 4,
    "args": {
      "External id": 212308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212308,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212308, "pid": 5, "tid": 7, "ts": 1716454224794216,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734752, "dur": 6,
    "args": {
      "External id": 212308,
      "cbid": 211, "correlation": 212308
    }
  },
  {
    "ph": "s", "id": 212308, "pid": 76337, "tid": -914061504, "ts": 1716454224734752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734760, "dur": 0,
    "args": {
      "External id": 212309,
      "cbid": 51, "correlation": 212309
    }
  },
  {
    "ph": "s", "id": 212309, "pid": 76337, "tid": -914061504, "ts": 1716454224734760,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224794221, "dur": 91,
    "args": {
      "External id": 212310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212310,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 212310, "pid": 5, "tid": 7, "ts": 1716454224794221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734761, "dur": 5,
    "args": {
      "External id": 212310,
      "cbid": 211, "correlation": 212310
    }
  },
  {
    "ph": "s", "id": 212310, "pid": 76337, "tid": -914061504, "ts": 1716454224734761,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224794314, "dur": 16,
    "args": {
      "External id": 212315,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212315,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212315, "pid": 5, "tid": 7, "ts": 1716454224794314,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734788, "dur": 8,
    "args": {
      "External id": 212315,
      "cbid": 211, "correlation": 212315
    }
  },
  {
    "ph": "s", "id": 212315, "pid": 76337, "tid": -914061504, "ts": 1716454224734788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224794331, "dur": 84,
    "args": {
      "External id": 212324,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212324,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212324, "pid": 5, "tid": 7, "ts": 1716454224794331,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734870, "dur": 16,
    "args": {
      "External id": 212324,
      "cbid": 211, "correlation": 212324
    }
  },
  {
    "ph": "s", "id": 212324, "pid": 76337, "tid": -914061504, "ts": 1716454224734870,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224794417, "dur": 30,
    "args": {
      "External id": 212346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212346,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212346, "pid": 5, "tid": 7, "ts": 1716454224794417,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224734929, "dur": 11,
    "args": {
      "External id": 212346,
      "cbid": 211, "correlation": 212346
    }
  },
  {
    "ph": "s", "id": 212346, "pid": 76337, "tid": -914061504, "ts": 1716454224734929,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735029, "dur": 2,
    "args": {
      "External id": 212357,
      "cbid": 251, "correlation": 212357
    }
  },
  {
    "ph": "f", "id": 212357, "pid": 76337, "tid": -914061504, "ts": 1716454224735029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224794449, "dur": 167,
    "args": {
      "External id": 212358,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212358,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212358, "pid": 5, "tid": 7, "ts": 1716454224794449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735036, "dur": 14,
    "args": {
      "External id": 212358,
      "cbid": 211, "correlation": 212358
    }
  },
  {
    "ph": "s", "id": 212358, "pid": 76337, "tid": -914061504, "ts": 1716454224735036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735108, "dur": 1,
    "args": {
      "External id": 212369,
      "cbid": 251, "correlation": 212369
    }
  },
  {
    "ph": "f", "id": 212369, "pid": 76337, "tid": -914061504, "ts": 1716454224735108,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224794617, "dur": 159,
    "args": {
      "External id": 212370,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212370,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212370, "pid": 5, "tid": 7, "ts": 1716454224794617,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735111, "dur": 11,
    "args": {
      "External id": 212370,
      "cbid": 211, "correlation": 212370
    }
  },
  {
    "ph": "s", "id": 212370, "pid": 76337, "tid": -914061504, "ts": 1716454224735111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735176, "dur": 1,
    "args": {
      "External id": 212381,
      "cbid": 251, "correlation": 212381
    }
  },
  {
    "ph": "f", "id": 212381, "pid": 76337, "tid": -914061504, "ts": 1716454224735176,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224794777, "dur": 160,
    "args": {
      "External id": 212382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212382,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212382, "pid": 5, "tid": 7, "ts": 1716454224794777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735180, "dur": 11,
    "args": {
      "External id": 212382,
      "cbid": 211, "correlation": 212382
    }
  },
  {
    "ph": "s", "id": 212382, "pid": 76337, "tid": -914061504, "ts": 1716454224735180,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224794938, "dur": 339,
    "args": {
      "External id": 212407,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212407,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212407, "pid": 5, "tid": 7, "ts": 1716454224794938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735267, "dur": 13,
    "args": {
      "External id": 212407,
      "cbid": 211, "correlation": 212407
    }
  },
  {
    "ph": "s", "id": 212407, "pid": 76337, "tid": -914061504, "ts": 1716454224735267,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735368, "dur": 1,
    "args": {
      "External id": 212425,
      "cbid": 251, "correlation": 212425
    }
  },
  {
    "ph": "f", "id": 212425, "pid": 76337, "tid": -914061504, "ts": 1716454224735368,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224795279, "dur": 168,
    "args": {
      "External id": 212427,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212427,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212427, "pid": 5, "tid": 7, "ts": 1716454224795279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735374, "dur": 14,
    "args": {
      "External id": 212427,
      "cbid": 211, "correlation": 212427
    }
  },
  {
    "ph": "s", "id": 212427, "pid": 76337, "tid": -914061504, "ts": 1716454224735374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224795448, "dur": 19,
    "args": {
      "External id": 212435,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212435,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212435, "pid": 5, "tid": 7, "ts": 1716454224795448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735444, "dur": 11,
    "args": {
      "External id": 212435,
      "cbid": 211, "correlation": 212435
    }
  },
  {
    "ph": "s", "id": 212435, "pid": 76337, "tid": -914061504, "ts": 1716454224735444,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224795469, "dur": 27,
    "args": {
      "External id": 212443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212443,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212443, "pid": 5, "tid": 7, "ts": 1716454224795469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735482, "dur": 8,
    "args": {
      "External id": 212443,
      "cbid": 211, "correlation": 212443
    }
  },
  {
    "ph": "s", "id": 212443, "pid": 76337, "tid": -914061504, "ts": 1716454224735482,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224795497, "dur": 18,
    "args": {
      "External id": 212454,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212454,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212454, "pid": 5, "tid": 7, "ts": 1716454224795497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735555, "dur": 12,
    "args": {
      "External id": 212454,
      "cbid": 211, "correlation": 212454
    }
  },
  {
    "ph": "s", "id": 212454, "pid": 76337, "tid": -914061504, "ts": 1716454224735555,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224795517, "dur": 16,
    "args": {
      "External id": 212476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212476,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212476, "pid": 5, "tid": 7, "ts": 1716454224795517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735587, "dur": 7,
    "args": {
      "External id": 212476,
      "cbid": 211, "correlation": 212476
    }
  },
  {
    "ph": "s", "id": 212476, "pid": 76337, "tid": -914061504, "ts": 1716454224735587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735673, "dur": 1,
    "args": {
      "External id": 212487,
      "cbid": 251, "correlation": 212487
    }
  },
  {
    "ph": "f", "id": 212487, "pid": 76337, "tid": -914061504, "ts": 1716454224735673,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224795534, "dur": 90,
    "args": {
      "External id": 212488,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212488,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212488, "pid": 5, "tid": 7, "ts": 1716454224795534,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735678, "dur": 14,
    "args": {
      "External id": 212488,
      "cbid": 211, "correlation": 212488
    }
  },
  {
    "ph": "s", "id": 212488, "pid": 76337, "tid": -914061504, "ts": 1716454224735678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735748, "dur": 1,
    "args": {
      "External id": 212499,
      "cbid": 251, "correlation": 212499
    }
  },
  {
    "ph": "f", "id": 212499, "pid": 76337, "tid": -914061504, "ts": 1716454224735748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735752, "dur": 0,
    "args": {
      "External id": 212500,
      "cbid": 251, "correlation": 212500
    }
  },
  {
    "ph": "f", "id": 212500, "pid": 76337, "tid": -914061504, "ts": 1716454224735752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224795625, "dur": 12,
    "args": {
      "External id": 212501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212501,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212501, "pid": 5, "tid": 7, "ts": 1716454224795625,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735754, "dur": 12,
    "args": {
      "External id": 212501,
      "cbid": 211, "correlation": 212501
    }
  },
  {
    "ph": "s", "id": 212501, "pid": 76337, "tid": -914061504, "ts": 1716454224735754,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224795639, "dur": 6,
    "args": {
      "External id": 212503,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212503,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212503, "pid": 5, "tid": 7, "ts": 1716454224795639,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735767, "dur": 6,
    "args": {
      "External id": 212503,
      "cbid": 211, "correlation": 212503
    }
  },
  {
    "ph": "s", "id": 212503, "pid": 76337, "tid": -914061504, "ts": 1716454224735767,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735825, "dur": 1,
    "args": {
      "External id": 212514,
      "cbid": 251, "correlation": 212514
    }
  },
  {
    "ph": "f", "id": 212514, "pid": 76337, "tid": -914061504, "ts": 1716454224735825,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735829, "dur": 0,
    "args": {
      "External id": 212515,
      "cbid": 251, "correlation": 212515
    }
  },
  {
    "ph": "f", "id": 212515, "pid": 76337, "tid": -914061504, "ts": 1716454224735829,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224795646, "dur": 8,
    "args": {
      "External id": 212516,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212516,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212516, "pid": 5, "tid": 7, "ts": 1716454224795646,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735830, "dur": 12,
    "args": {
      "External id": 212516,
      "cbid": 211, "correlation": 212516
    }
  },
  {
    "ph": "s", "id": 212516, "pid": 76337, "tid": -914061504, "ts": 1716454224735830,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224795656, "dur": 3,
    "args": {
      "External id": 212518,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212518,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212518, "pid": 5, "tid": 7, "ts": 1716454224795656,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735843, "dur": 5,
    "args": {
      "External id": 212518,
      "cbid": 211, "correlation": 212518
    }
  },
  {
    "ph": "s", "id": 212518, "pid": 76337, "tid": -914061504, "ts": 1716454224735843,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224795660, "dur": 56,
    "args": {
      "External id": 212543,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212543,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212543, "pid": 5, "tid": 7, "ts": 1716454224795660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224735920, "dur": 13,
    "args": {
      "External id": 212543,
      "cbid": 211, "correlation": 212543
    }
  },
  {
    "ph": "s", "id": 212543, "pid": 76337, "tid": -914061504, "ts": 1716454224735920,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736029, "dur": 2,
    "args": {
      "External id": 212561,
      "cbid": 251, "correlation": 212561
    }
  },
  {
    "ph": "f", "id": 212561, "pid": 76337, "tid": -914061504, "ts": 1716454224736029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224795717, "dur": 91,
    "args": {
      "External id": 212563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212563,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212563, "pid": 5, "tid": 7, "ts": 1716454224795717,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736035, "dur": 14,
    "args": {
      "External id": 212563,
      "cbid": 211, "correlation": 212563
    }
  },
  {
    "ph": "s", "id": 212563, "pid": 76337, "tid": -914061504, "ts": 1716454224736035,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224795810, "dur": 10,
    "args": {
      "External id": 212571,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212571,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212571, "pid": 5, "tid": 7, "ts": 1716454224795810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736106, "dur": 12,
    "args": {
      "External id": 212571,
      "cbid": 211, "correlation": 212571
    }
  },
  {
    "ph": "s", "id": 212571, "pid": 76337, "tid": -914061504, "ts": 1716454224736106,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224795821, "dur": 21,
    "args": {
      "External id": 212579,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212579,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212579, "pid": 5, "tid": 7, "ts": 1716454224795821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736148, "dur": 9,
    "args": {
      "External id": 212579,
      "cbid": 211, "correlation": 212579
    }
  },
  {
    "ph": "s", "id": 212579, "pid": 76337, "tid": -914061504, "ts": 1716454224736148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224795843, "dur": 17,
    "args": {
      "External id": 212601,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212601,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212601, "pid": 5, "tid": 7, "ts": 1716454224795843,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736199, "dur": 10,
    "args": {
      "External id": 212601,
      "cbid": 211, "correlation": 212601
    }
  },
  {
    "ph": "s", "id": 212601, "pid": 76337, "tid": -914061504, "ts": 1716454224736199,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736288, "dur": 1,
    "args": {
      "External id": 212617,
      "cbid": 251, "correlation": 212617
    }
  },
  {
    "ph": "f", "id": 212617, "pid": 76337, "tid": -914061504, "ts": 1716454224736288,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736293, "dur": 0,
    "args": {
      "External id": 212619,
      "cbid": 251, "correlation": 212619
    }
  },
  {
    "ph": "f", "id": 212619, "pid": 76337, "tid": -914061504, "ts": 1716454224736293,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224795862, "dur": 496,
    "args": {
      "External id": 212620,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212620,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212620, "pid": 5, "tid": 7, "ts": 1716454224795862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736294, "dur": 13,
    "args": {
      "External id": 212620,
      "cbid": 211, "correlation": 212620
    }
  },
  {
    "ph": "s", "id": 212620, "pid": 76337, "tid": -914061504, "ts": 1716454224736294,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224796360, "dur": 66,
    "args": {
      "External id": 212628,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212628,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212628, "pid": 5, "tid": 7, "ts": 1716454224796360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736361, "dur": 12,
    "args": {
      "External id": 212628,
      "cbid": 211, "correlation": 212628
    }
  },
  {
    "ph": "s", "id": 212628, "pid": 76337, "tid": -914061504, "ts": 1716454224736361,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224796427, "dur": 65,
    "args": {
      "External id": 212636,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212636,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212636, "pid": 5, "tid": 7, "ts": 1716454224796427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736393, "dur": 8,
    "args": {
      "External id": 212636,
      "cbid": 211, "correlation": 212636
    }
  },
  {
    "ph": "s", "id": 212636, "pid": 76337, "tid": -914061504, "ts": 1716454224736393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736473, "dur": 1,
    "args": {
      "External id": 212652,
      "cbid": 251, "correlation": 212652
    }
  },
  {
    "ph": "f", "id": 212652, "pid": 76337, "tid": -914061504, "ts": 1716454224736473,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224796493, "dur": 1,
    "args": {
      "External id": 212654,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 212654,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 212654, "pid": 5, "tid": 7, "ts": 1716454224796493,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736478, "dur": 10,
    "args": {
      "External id": 212654,
      "cbid": 51, "correlation": 212654
    }
  },
  {
    "ph": "s", "id": 212654, "pid": 76337, "tid": -914061504, "ts": 1716454224736478,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224796497, "dur": 274,
    "args": {
      "External id": 212655,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212655,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212655, "pid": 5, "tid": 7, "ts": 1716454224796497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736489, "dur": 11,
    "args": {
      "External id": 212655,
      "cbid": 211, "correlation": 212655
    }
  },
  {
    "ph": "s", "id": 212655, "pid": 76337, "tid": -914061504, "ts": 1716454224736489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224796772, "dur": 13,
    "args": {
      "External id": 212663,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212663,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212663, "pid": 5, "tid": 7, "ts": 1716454224796772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736533, "dur": 10,
    "args": {
      "External id": 212663,
      "cbid": 211, "correlation": 212663
    }
  },
  {
    "ph": "s", "id": 212663, "pid": 76337, "tid": -914061504, "ts": 1716454224736533,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224796787, "dur": 38,
    "args": {
      "External id": 212674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212674,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212674, "pid": 5, "tid": 7, "ts": 1716454224796787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736602, "dur": 12,
    "args": {
      "External id": 212674,
      "cbid": 211, "correlation": 212674
    }
  },
  {
    "ph": "s", "id": 212674, "pid": 76337, "tid": -914061504, "ts": 1716454224736602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736666, "dur": 0,
    "args": {
      "External id": 212686,
      "cbid": 317, "correlation": 212686
    }
  },
  {
    "ph": "f", "id": 212686, "pid": 76337, "tid": -914061504, "ts": 1716454224736666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736667, "dur": 0,
    "args": {
      "External id": 212687,
      "cbid": 203, "correlation": 212687
    }
  },
  {
    "ph": "f", "id": 212687, "pid": 76337, "tid": -914061504, "ts": 1716454224736667,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736668, "dur": 0,
    "args": {
      "External id": 212688,
      "cbid": 205, "correlation": 212688
    }
  },
  {
    "ph": "f", "id": 212688, "pid": 76337, "tid": -914061504, "ts": 1716454224736668,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224796826, "dur": 14,
    "args": {
      "External id": 212692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212692,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212692, "pid": 5, "tid": 7, "ts": 1716454224796826,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736684, "dur": 12,
    "args": {
      "External id": 212692,
      "cbid": 211, "correlation": 212692
    }
  },
  {
    "ph": "s", "id": 212692, "pid": 76337, "tid": -914061504, "ts": 1716454224736684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224796841, "dur": 4,
    "args": {
      "External id": 212694,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212694,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212694, "pid": 5, "tid": 7, "ts": 1716454224796841,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736700, "dur": 6,
    "args": {
      "External id": 212694,
      "cbid": 211, "correlation": 212694
    }
  },
  {
    "ph": "s", "id": 212694, "pid": 76337, "tid": -914061504, "ts": 1716454224736700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736708, "dur": 0,
    "args": {
      "External id": 212695,
      "cbid": 51, "correlation": 212695
    }
  },
  {
    "ph": "s", "id": 212695, "pid": 76337, "tid": -914061504, "ts": 1716454224736708,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224796847, "dur": 98,
    "args": {
      "External id": 212696,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212696,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 212696, "pid": 5, "tid": 7, "ts": 1716454224796847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736709, "dur": 5,
    "args": {
      "External id": 212696,
      "cbid": 211, "correlation": 212696
    }
  },
  {
    "ph": "s", "id": 212696, "pid": 76337, "tid": -914061504, "ts": 1716454224736709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224796946, "dur": 17,
    "args": {
      "External id": 212701,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212701,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212701, "pid": 5, "tid": 7, "ts": 1716454224796946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736737, "dur": 8,
    "args": {
      "External id": 212701,
      "cbid": 211, "correlation": 212701
    }
  },
  {
    "ph": "s", "id": 212701, "pid": 76337, "tid": -914061504, "ts": 1716454224736737,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224796964, "dur": 11,
    "args": {
      "External id": 212709,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212709,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212709, "pid": 5, "tid": 7, "ts": 1716454224796964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736769, "dur": 9,
    "args": {
      "External id": 212709,
      "cbid": 211, "correlation": 212709
    }
  },
  {
    "ph": "s", "id": 212709, "pid": 76337, "tid": -914061504, "ts": 1716454224736769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224796977, "dur": 30,
    "args": {
      "External id": 212718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212718,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212718, "pid": 5, "tid": 7, "ts": 1716454224796977,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736809, "dur": 10,
    "args": {
      "External id": 212718,
      "cbid": 211, "correlation": 212718
    }
  },
  {
    "ph": "s", "id": 212718, "pid": 76337, "tid": -914061504, "ts": 1716454224736809,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224797009, "dur": 31,
    "args": {
      "External id": 212738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212738,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 212738, "pid": 5, "tid": 7, "ts": 1716454224797009,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736883, "dur": 11,
    "args": {
      "External id": 212738,
      "cbid": 211, "correlation": 212738
    }
  },
  {
    "ph": "s", "id": 212738, "pid": 76337, "tid": -914061504, "ts": 1716454224736883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224797041, "dur": 5,
    "args": {
      "External id": 212750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212750,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 1,
      "warps per SM": 8,
      "grid": [80, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212750, "pid": 5, "tid": 7, "ts": 1716454224797041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736904, "dur": 6,
    "args": {
      "External id": 212750,
      "cbid": 211, "correlation": 212750
    }
  },
  {
    "ph": "s", "id": 212750, "pid": 76337, "tid": -914061504, "ts": 1716454224736904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224797047, "dur": 31,
    "args": {
      "External id": 212753,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212753,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212753, "pid": 5, "tid": 7, "ts": 1716454224797047,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736923, "dur": 8,
    "args": {
      "External id": 212753,
      "cbid": 211, "correlation": 212753
    }
  },
  {
    "ph": "s", "id": 212753, "pid": 76337, "tid": -914061504, "ts": 1716454224736923,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224797080, "dur": 21,
    "args": {
      "External id": 212762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212762,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212762, "pid": 5, "tid": 7, "ts": 1716454224797080,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224736962, "dur": 9,
    "args": {
      "External id": 212762,
      "cbid": 211, "correlation": 212762
    }
  },
  {
    "ph": "s", "id": 212762, "pid": 76337, "tid": -914061504, "ts": 1716454224736962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737025, "dur": 0,
    "args": {
      "External id": 212772,
      "cbid": 317, "correlation": 212772
    }
  },
  {
    "ph": "f", "id": 212772, "pid": 76337, "tid": -914061504, "ts": 1716454224737025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737026, "dur": 0,
    "args": {
      "External id": 212773,
      "cbid": 203, "correlation": 212773
    }
  },
  {
    "ph": "f", "id": 212773, "pid": 76337, "tid": -914061504, "ts": 1716454224737026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737027, "dur": 0,
    "args": {
      "External id": 212774,
      "cbid": 205, "correlation": 212774
    }
  },
  {
    "ph": "f", "id": 212774, "pid": 76337, "tid": -914061504, "ts": 1716454224737027,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224797102, "dur": 22,
    "args": {
      "External id": 212778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212778, "pid": 5, "tid": 7, "ts": 1716454224797102,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737041, "dur": 12,
    "args": {
      "External id": 212778,
      "cbid": 211, "correlation": 212778
    }
  },
  {
    "ph": "s", "id": 212778, "pid": 76337, "tid": -914061504, "ts": 1716454224737041,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224797125, "dur": 323,
    "args": {
      "External id": 212780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212780,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 1280,
      "warps per SM": 10240,
      "grid": [1, 80, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212780, "pid": 5, "tid": 7, "ts": 1716454224797125,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737055, "dur": 5,
    "args": {
      "External id": 212780,
      "cbid": 211, "correlation": 212780
    }
  },
  {
    "ph": "s", "id": 212780, "pid": 76337, "tid": -914061504, "ts": 1716454224737055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224797451, "dur": 1,
    "args": {
      "External id": 212782,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 212782,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 212782, "pid": 5, "tid": 7, "ts": 1716454224797451,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737067, "dur": 7,
    "args": {
      "External id": 212782,
      "cbid": 51, "correlation": 212782
    }
  },
  {
    "ph": "s", "id": 212782, "pid": 76337, "tid": -914061504, "ts": 1716454224737067,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224797455, "dur": 1265,
    "args": {
      "External id": 212783,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212783,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212783, "pid": 5, "tid": 7, "ts": 1716454224797455,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737075, "dur": 6,
    "args": {
      "External id": 212783,
      "cbid": 211, "correlation": 212783
    }
  },
  {
    "ph": "s", "id": 212783, "pid": 76337, "tid": -914061504, "ts": 1716454224737075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224798721, "dur": 13,
    "args": {
      "External id": 212785,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212785,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212785, "pid": 5, "tid": 7, "ts": 1716454224798721,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737086, "dur": 5,
    "args": {
      "External id": 212785,
      "cbid": 211, "correlation": 212785
    }
  },
  {
    "ph": "s", "id": 212785, "pid": 76337, "tid": -914061504, "ts": 1716454224737086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224798736, "dur": 15,
    "args": {
      "External id": 212791,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212791,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212791, "pid": 5, "tid": 7, "ts": 1716454224798736,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737114, "dur": 9,
    "args": {
      "External id": 212791,
      "cbid": 211, "correlation": 212791
    }
  },
  {
    "ph": "s", "id": 212791, "pid": 76337, "tid": -914061504, "ts": 1716454224737114,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224798752, "dur": 3,
    "args": {
      "External id": 212799,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212799,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 212799, "pid": 5, "tid": 7, "ts": 1716454224798752,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737158, "dur": 9,
    "args": {
      "External id": 212799,
      "cbid": 211, "correlation": 212799
    }
  },
  {
    "ph": "s", "id": 212799, "pid": 76337, "tid": -914061504, "ts": 1716454224737158,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737226, "dur": 1,
    "args": {
      "External id": 212815,
      "cbid": 251, "correlation": 212815
    }
  },
  {
    "ph": "f", "id": 212815, "pid": 76337, "tid": -914061504, "ts": 1716454224737226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737232, "dur": 0,
    "args": {
      "External id": 212817,
      "cbid": 251, "correlation": 212817
    }
  },
  {
    "ph": "f", "id": 212817, "pid": 76337, "tid": -914061504, "ts": 1716454224737232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224798757, "dur": 13,
    "args": {
      "External id": 212818,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212818,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212818, "pid": 5, "tid": 7, "ts": 1716454224798757,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737234, "dur": 11,
    "args": {
      "External id": 212818,
      "cbid": 211, "correlation": 212818
    }
  },
  {
    "ph": "s", "id": 212818, "pid": 76337, "tid": -914061504, "ts": 1716454224737234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224798772, "dur": 5,
    "args": {
      "External id": 212820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212820,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212820, "pid": 5, "tid": 7, "ts": 1716454224798772,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737247, "dur": 5,
    "args": {
      "External id": 212820,
      "cbid": 211, "correlation": 212820
    }
  },
  {
    "ph": "s", "id": 212820, "pid": 76337, "tid": -914061504, "ts": 1716454224737247,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224798778, "dur": 17,
    "args": {
      "External id": 212830,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212830,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212830, "pid": 5, "tid": 7, "ts": 1716454224798778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737306, "dur": 12,
    "args": {
      "External id": 212830,
      "cbid": 211, "correlation": 212830
    }
  },
  {
    "ph": "s", "id": 212830, "pid": 76337, "tid": -914061504, "ts": 1716454224737306,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224798796, "dur": 18,
    "args": {
      "External id": 212850,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212850,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 212850, "pid": 5, "tid": 7, "ts": 1716454224798796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737374, "dur": 11,
    "args": {
      "External id": 212850,
      "cbid": 211, "correlation": 212850
    }
  },
  {
    "ph": "s", "id": 212850, "pid": 76337, "tid": -914061504, "ts": 1716454224737374,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224798816, "dur": 4,
    "args": {
      "External id": 212862,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212862,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 212862, "pid": 5, "tid": 7, "ts": 1716454224798816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737395, "dur": 6,
    "args": {
      "External id": 212862,
      "cbid": 211, "correlation": 212862
    }
  },
  {
    "ph": "s", "id": 212862, "pid": 76337, "tid": -914061504, "ts": 1716454224737395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224798821, "dur": 16,
    "args": {
      "External id": 212865,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212865,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212865, "pid": 5, "tid": 7, "ts": 1716454224798821,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737413, "dur": 7,
    "args": {
      "External id": 212865,
      "cbid": 211, "correlation": 212865
    }
  },
  {
    "ph": "s", "id": 212865, "pid": 76337, "tid": -914061504, "ts": 1716454224737413,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224798839, "dur": 11,
    "args": {
      "External id": 212874,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212874,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212874, "pid": 5, "tid": 7, "ts": 1716454224798839,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737454, "dur": 10,
    "args": {
      "External id": 212874,
      "cbid": 211, "correlation": 212874
    }
  },
  {
    "ph": "s", "id": 212874, "pid": 76337, "tid": -914061504, "ts": 1716454224737454,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737516, "dur": 0,
    "args": {
      "External id": 212884,
      "cbid": 317, "correlation": 212884
    }
  },
  {
    "ph": "f", "id": 212884, "pid": 76337, "tid": -914061504, "ts": 1716454224737516,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737517, "dur": 0,
    "args": {
      "External id": 212885,
      "cbid": 203, "correlation": 212885
    }
  },
  {
    "ph": "f", "id": 212885, "pid": 76337, "tid": -914061504, "ts": 1716454224737517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737518, "dur": 0,
    "args": {
      "External id": 212886,
      "cbid": 205, "correlation": 212886
    }
  },
  {
    "ph": "f", "id": 212886, "pid": 76337, "tid": -914061504, "ts": 1716454224737518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224798851, "dur": 11,
    "args": {
      "External id": 212890,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212890,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212890, "pid": 5, "tid": 7, "ts": 1716454224798851,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737532, "dur": 11,
    "args": {
      "External id": 212890,
      "cbid": 211, "correlation": 212890
    }
  },
  {
    "ph": "s", "id": 212890, "pid": 76337, "tid": -914061504, "ts": 1716454224737532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224798864, "dur": 165,
    "args": {
      "External id": 212892,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212892,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212892, "pid": 5, "tid": 7, "ts": 1716454224798864,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737546, "dur": 5,
    "args": {
      "External id": 212892,
      "cbid": 211, "correlation": 212892
    }
  },
  {
    "ph": "s", "id": 212892, "pid": 76337, "tid": -914061504, "ts": 1716454224737546,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224799030, "dur": 1,
    "args": {
      "External id": 212894,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 212894,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5
    }
  },
  {
    "ph": "f", "id": 212894, "pid": 5, "tid": 7, "ts": 1716454224799030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737557, "dur": 7,
    "args": {
      "External id": 212894,
      "cbid": 51, "correlation": 212894
    }
  },
  {
    "ph": "s", "id": 212894, "pid": 76337, "tid": -914061504, "ts": 1716454224737557,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224799035, "dur": 653,
    "args": {
      "External id": 212895,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212895,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 212895, "pid": 5, "tid": 7, "ts": 1716454224799035,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737564, "dur": 7,
    "args": {
      "External id": 212895,
      "cbid": 211, "correlation": 212895
    }
  },
  {
    "ph": "s", "id": 212895, "pid": 76337, "tid": -914061504, "ts": 1716454224737564,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224799688, "dur": 14,
    "args": {
      "External id": 212897,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212897,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212897, "pid": 5, "tid": 7, "ts": 1716454224799688,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737576, "dur": 5,
    "args": {
      "External id": 212897,
      "cbid": 211, "correlation": 212897
    }
  },
  {
    "ph": "s", "id": 212897, "pid": 76337, "tid": -914061504, "ts": 1716454224737576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224799703, "dur": 16,
    "args": {
      "External id": 212903,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212903,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212903, "pid": 5, "tid": 7, "ts": 1716454224799703,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737605, "dur": 9,
    "args": {
      "External id": 212903,
      "cbid": 211, "correlation": 212903
    }
  },
  {
    "ph": "s", "id": 212903, "pid": 76337, "tid": -914061504, "ts": 1716454224737605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737662, "dur": 0,
    "args": {
      "External id": 212913,
      "cbid": 317, "correlation": 212913
    }
  },
  {
    "ph": "f", "id": 212913, "pid": 76337, "tid": -914061504, "ts": 1716454224737662,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737663, "dur": 0,
    "args": {
      "External id": 212914,
      "cbid": 203, "correlation": 212914
    }
  },
  {
    "ph": "f", "id": 212914, "pid": 76337, "tid": -914061504, "ts": 1716454224737663,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737664, "dur": 0,
    "args": {
      "External id": 212915,
      "cbid": 205, "correlation": 212915
    }
  },
  {
    "ph": "f", "id": 212915, "pid": 76337, "tid": -914061504, "ts": 1716454224737664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224799720, "dur": 21,
    "args": {
      "External id": 212919,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212919,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [6, 80, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212919, "pid": 5, "tid": 7, "ts": 1716454224799720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737676, "dur": 11,
    "args": {
      "External id": 212919,
      "cbid": 211, "correlation": 212919
    }
  },
  {
    "ph": "s", "id": 212919, "pid": 76337, "tid": -914061504, "ts": 1716454224737676,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224799743, "dur": 4,
    "args": {
      "External id": 212921,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212921,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212921, "pid": 5, "tid": 7, "ts": 1716454224799743,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737692, "dur": 6,
    "args": {
      "External id": 212921,
      "cbid": 211, "correlation": 212921
    }
  },
  {
    "ph": "s", "id": 212921, "pid": 76337, "tid": -914061504, "ts": 1716454224737692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737700, "dur": 0,
    "args": {
      "External id": 212922,
      "cbid": 51, "correlation": 212922
    }
  },
  {
    "ph": "s", "id": 212922, "pid": 76337, "tid": -914061504, "ts": 1716454224737700,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224799748, "dur": 172,
    "args": {
      "External id": 212923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212923,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 212923, "pid": 5, "tid": 7, "ts": 1716454224799748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737701, "dur": 5,
    "args": {
      "External id": 212923,
      "cbid": 211, "correlation": 212923
    }
  },
  {
    "ph": "s", "id": 212923, "pid": 76337, "tid": -914061504, "ts": 1716454224737701,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224799921, "dur": 16,
    "args": {
      "External id": 212928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212928,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212928, "pid": 5, "tid": 7, "ts": 1716454224799921,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737727, "dur": 8,
    "args": {
      "External id": 212928,
      "cbid": 211, "correlation": 212928
    }
  },
  {
    "ph": "s", "id": 212928, "pid": 76337, "tid": -914061504, "ts": 1716454224737727,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224799939, "dur": 13,
    "args": {
      "External id": 212936,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212936,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212936, "pid": 5, "tid": 7, "ts": 1716454224799939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737755, "dur": 8,
    "args": {
      "External id": 212936,
      "cbid": 211, "correlation": 212936
    }
  },
  {
    "ph": "s", "id": 212936, "pid": 76337, "tid": -914061504, "ts": 1716454224737755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224799953, "dur": 10,
    "args": {
      "External id": 212944,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212944,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212944, "pid": 5, "tid": 7, "ts": 1716454224799953,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737784, "dur": 8,
    "args": {
      "External id": 212944,
      "cbid": 211, "correlation": 212944
    }
  },
  {
    "ph": "s", "id": 212944, "pid": 76337, "tid": -914061504, "ts": 1716454224737784,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224799964, "dur": 18,
    "args": {
      "External id": 212964,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212964,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 212964, "pid": 5, "tid": 7, "ts": 1716454224799964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737865, "dur": 13,
    "args": {
      "External id": 212964,
      "cbid": 211, "correlation": 212964
    }
  },
  {
    "ph": "s", "id": 212964, "pid": 76337, "tid": -914061504, "ts": 1716454224737865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224799984, "dur": 4,
    "args": {
      "External id": 212976,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212976,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 212976, "pid": 5, "tid": 7, "ts": 1716454224799984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737888, "dur": 6,
    "args": {
      "External id": 212976,
      "cbid": 211, "correlation": 212976
    }
  },
  {
    "ph": "s", "id": 212976, "pid": 76337, "tid": -914061504, "ts": 1716454224737888,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224799990, "dur": 16,
    "args": {
      "External id": 212979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212979, "pid": 5, "tid": 7, "ts": 1716454224799990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737905, "dur": 6,
    "args": {
      "External id": 212979,
      "cbid": 211, "correlation": 212979
    }
  },
  {
    "ph": "s", "id": 212979, "pid": 76337, "tid": -914061504, "ts": 1716454224737905,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737962, "dur": 0,
    "args": {
      "External id": 212990,
      "cbid": 317, "correlation": 212990
    }
  },
  {
    "ph": "f", "id": 212990, "pid": 76337, "tid": -914061504, "ts": 1716454224737962,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737963, "dur": 0,
    "args": {
      "External id": 212991,
      "cbid": 203, "correlation": 212991
    }
  },
  {
    "ph": "f", "id": 212991, "pid": 76337, "tid": -914061504, "ts": 1716454224737963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737964, "dur": 0,
    "args": {
      "External id": 212992,
      "cbid": 205, "correlation": 212992
    }
  },
  {
    "ph": "f", "id": 212992, "pid": 76337, "tid": -914061504, "ts": 1716454224737964,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224800007, "dur": 12,
    "args": {
      "External id": 212996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212996,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 212996, "pid": 5, "tid": 7, "ts": 1716454224800007,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224737985, "dur": 12,
    "args": {
      "External id": 212996,
      "cbid": 211, "correlation": 212996
    }
  },
  {
    "ph": "s", "id": 212996, "pid": 76337, "tid": -914061504, "ts": 1716454224737985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224800021, "dur": 3,
    "args": {
      "External id": 212998,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 212998,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 212998, "pid": 5, "tid": 7, "ts": 1716454224800021,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738001, "dur": 6,
    "args": {
      "External id": 212998,
      "cbid": 211, "correlation": 212998
    }
  },
  {
    "ph": "s", "id": 212998, "pid": 76337, "tid": -914061504, "ts": 1716454224738001,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738009, "dur": 0,
    "args": {
      "External id": 212999,
      "cbid": 51, "correlation": 212999
    }
  },
  {
    "ph": "s", "id": 212999, "pid": 76337, "tid": -914061504, "ts": 1716454224738009,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224800025, "dur": 91,
    "args": {
      "External id": 213000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213000,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 213000, "pid": 5, "tid": 7, "ts": 1716454224800025,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738010, "dur": 5,
    "args": {
      "External id": 213000,
      "cbid": 211, "correlation": 213000
    }
  },
  {
    "ph": "s", "id": 213000, "pid": 76337, "tid": -914061504, "ts": 1716454224738010,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224800118, "dur": 16,
    "args": {
      "External id": 213005,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213005,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213005, "pid": 5, "tid": 7, "ts": 1716454224800118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738038, "dur": 8,
    "args": {
      "External id": 213005,
      "cbid": 211, "correlation": 213005
    }
  },
  {
    "ph": "s", "id": 213005, "pid": 76337, "tid": -914061504, "ts": 1716454224738038,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224800135, "dur": 85,
    "args": {
      "External id": 213014,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213014,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213014, "pid": 5, "tid": 7, "ts": 1716454224800135,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738120, "dur": 14,
    "args": {
      "External id": 213014,
      "cbid": 211, "correlation": 213014
    }
  },
  {
    "ph": "s", "id": 213014, "pid": 76337, "tid": -914061504, "ts": 1716454224738120,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224800221, "dur": 30,
    "args": {
      "External id": 213036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213036,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213036, "pid": 5, "tid": 7, "ts": 1716454224800221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738177, "dur": 10,
    "args": {
      "External id": 213036,
      "cbid": 211, "correlation": 213036
    }
  },
  {
    "ph": "s", "id": 213036, "pid": 76337, "tid": -914061504, "ts": 1716454224738177,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738270, "dur": 1,
    "args": {
      "External id": 213047,
      "cbid": 251, "correlation": 213047
    }
  },
  {
    "ph": "f", "id": 213047, "pid": 76337, "tid": -914061504, "ts": 1716454224738270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224800253, "dur": 165,
    "args": {
      "External id": 213048,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213048,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213048, "pid": 5, "tid": 7, "ts": 1716454224800253,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738276, "dur": 13,
    "args": {
      "External id": 213048,
      "cbid": 211, "correlation": 213048
    }
  },
  {
    "ph": "s", "id": 213048, "pid": 76337, "tid": -914061504, "ts": 1716454224738276,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738347, "dur": 1,
    "args": {
      "External id": 213059,
      "cbid": 251, "correlation": 213059
    }
  },
  {
    "ph": "f", "id": 213059, "pid": 76337, "tid": -914061504, "ts": 1716454224738347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224800420, "dur": 161,
    "args": {
      "External id": 213060,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213060,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213060, "pid": 5, "tid": 7, "ts": 1716454224800420,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738350, "dur": 11,
    "args": {
      "External id": 213060,
      "cbid": 211, "correlation": 213060
    }
  },
  {
    "ph": "s", "id": 213060, "pid": 76337, "tid": -914061504, "ts": 1716454224738350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738416, "dur": 1,
    "args": {
      "External id": 213071,
      "cbid": 251, "correlation": 213071
    }
  },
  {
    "ph": "f", "id": 213071, "pid": 76337, "tid": -914061504, "ts": 1716454224738416,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224800582, "dur": 159,
    "args": {
      "External id": 213072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213072,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213072, "pid": 5, "tid": 7, "ts": 1716454224800582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738420, "dur": 11,
    "args": {
      "External id": 213072,
      "cbid": 211, "correlation": 213072
    }
  },
  {
    "ph": "s", "id": 213072, "pid": 76337, "tid": -914061504, "ts": 1716454224738420,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224800742, "dur": 339,
    "args": {
      "External id": 213097,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213097,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213097, "pid": 5, "tid": 7, "ts": 1716454224800742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738506, "dur": 14,
    "args": {
      "External id": 213097,
      "cbid": 211, "correlation": 213097
    }
  },
  {
    "ph": "s", "id": 213097, "pid": 76337, "tid": -914061504, "ts": 1716454224738506,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738606, "dur": 1,
    "args": {
      "External id": 213115,
      "cbid": 251, "correlation": 213115
    }
  },
  {
    "ph": "f", "id": 213115, "pid": 76337, "tid": -914061504, "ts": 1716454224738606,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224801083, "dur": 166,
    "args": {
      "External id": 213117,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213117,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213117, "pid": 5, "tid": 7, "ts": 1716454224801083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738612, "dur": 14,
    "args": {
      "External id": 213117,
      "cbid": 211, "correlation": 213117
    }
  },
  {
    "ph": "s", "id": 213117, "pid": 76337, "tid": -914061504, "ts": 1716454224738612,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224801250, "dur": 19,
    "args": {
      "External id": 213125,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213125,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213125, "pid": 5, "tid": 7, "ts": 1716454224801250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738682, "dur": 12,
    "args": {
      "External id": 213125,
      "cbid": 211, "correlation": 213125
    }
  },
  {
    "ph": "s", "id": 213125, "pid": 76337, "tid": -914061504, "ts": 1716454224738682,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224801271, "dur": 28,
    "args": {
      "External id": 213133,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213133,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213133, "pid": 5, "tid": 7, "ts": 1716454224801271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738722, "dur": 8,
    "args": {
      "External id": 213133,
      "cbid": 211, "correlation": 213133
    }
  },
  {
    "ph": "s", "id": 213133, "pid": 76337, "tid": -914061504, "ts": 1716454224738722,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224801300, "dur": 18,
    "args": {
      "External id": 213144,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213144,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213144, "pid": 5, "tid": 7, "ts": 1716454224801300,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738793, "dur": 12,
    "args": {
      "External id": 213144,
      "cbid": 211, "correlation": 213144
    }
  },
  {
    "ph": "s", "id": 213144, "pid": 76337, "tid": -914061504, "ts": 1716454224738793,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224801319, "dur": 16,
    "args": {
      "External id": 213166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213166,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213166, "pid": 5, "tid": 7, "ts": 1716454224801319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738824, "dur": 8,
    "args": {
      "External id": 213166,
      "cbid": 211, "correlation": 213166
    }
  },
  {
    "ph": "s", "id": 213166, "pid": 76337, "tid": -914061504, "ts": 1716454224738824,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738911, "dur": 1,
    "args": {
      "External id": 213177,
      "cbid": 251, "correlation": 213177
    }
  },
  {
    "ph": "f", "id": 213177, "pid": 76337, "tid": -914061504, "ts": 1716454224738911,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224801337, "dur": 89,
    "args": {
      "External id": 213178,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213178,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213178, "pid": 5, "tid": 7, "ts": 1716454224801337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738916, "dur": 14,
    "args": {
      "External id": 213178,
      "cbid": 211, "correlation": 213178
    }
  },
  {
    "ph": "s", "id": 213178, "pid": 76337, "tid": -914061504, "ts": 1716454224738916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738993, "dur": 1,
    "args": {
      "External id": 213189,
      "cbid": 251, "correlation": 213189
    }
  },
  {
    "ph": "f", "id": 213189, "pid": 76337, "tid": -914061504, "ts": 1716454224738993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738997, "dur": 0,
    "args": {
      "External id": 213190,
      "cbid": 251, "correlation": 213190
    }
  },
  {
    "ph": "f", "id": 213190, "pid": 76337, "tid": -914061504, "ts": 1716454224738997,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224801428, "dur": 12,
    "args": {
      "External id": 213191,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213191,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213191, "pid": 5, "tid": 7, "ts": 1716454224801428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224738999, "dur": 12,
    "args": {
      "External id": 213191,
      "cbid": 211, "correlation": 213191
    }
  },
  {
    "ph": "s", "id": 213191, "pid": 76337, "tid": -914061504, "ts": 1716454224738999,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224801441, "dur": 5,
    "args": {
      "External id": 213193,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213193,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213193, "pid": 5, "tid": 7, "ts": 1716454224801441,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739013, "dur": 6,
    "args": {
      "External id": 213193,
      "cbid": 211, "correlation": 213193
    }
  },
  {
    "ph": "s", "id": 213193, "pid": 76337, "tid": -914061504, "ts": 1716454224739013,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739073, "dur": 1,
    "args": {
      "External id": 213204,
      "cbid": 251, "correlation": 213204
    }
  },
  {
    "ph": "f", "id": 213204, "pid": 76337, "tid": -914061504, "ts": 1716454224739073,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739077, "dur": 0,
    "args": {
      "External id": 213205,
      "cbid": 251, "correlation": 213205
    }
  },
  {
    "ph": "f", "id": 213205, "pid": 76337, "tid": -914061504, "ts": 1716454224739077,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224801448, "dur": 8,
    "args": {
      "External id": 213206,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213206,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213206, "pid": 5, "tid": 7, "ts": 1716454224801448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739078, "dur": 12,
    "args": {
      "External id": 213206,
      "cbid": 211, "correlation": 213206
    }
  },
  {
    "ph": "s", "id": 213206, "pid": 76337, "tid": -914061504, "ts": 1716454224739078,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224801458, "dur": 3,
    "args": {
      "External id": 213208,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213208,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213208, "pid": 5, "tid": 7, "ts": 1716454224801458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739092, "dur": 5,
    "args": {
      "External id": 213208,
      "cbid": 211, "correlation": 213208
    }
  },
  {
    "ph": "s", "id": 213208, "pid": 76337, "tid": -914061504, "ts": 1716454224739092,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224801462, "dur": 56,
    "args": {
      "External id": 213233,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213233,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213233, "pid": 5, "tid": 7, "ts": 1716454224801462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739169, "dur": 13,
    "args": {
      "External id": 213233,
      "cbid": 211, "correlation": 213233
    }
  },
  {
    "ph": "s", "id": 213233, "pid": 76337, "tid": -914061504, "ts": 1716454224739169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739269, "dur": 1,
    "args": {
      "External id": 213251,
      "cbid": 251, "correlation": 213251
    }
  },
  {
    "ph": "f", "id": 213251, "pid": 76337, "tid": -914061504, "ts": 1716454224739269,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224801520, "dur": 91,
    "args": {
      "External id": 213253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213253,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213253, "pid": 5, "tid": 7, "ts": 1716454224801520,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739275, "dur": 13,
    "args": {
      "External id": 213253,
      "cbid": 211, "correlation": 213253
    }
  },
  {
    "ph": "s", "id": 213253, "pid": 76337, "tid": -914061504, "ts": 1716454224739275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224801612, "dur": 9,
    "args": {
      "External id": 213261,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213261,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213261, "pid": 5, "tid": 7, "ts": 1716454224801612,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739345, "dur": 12,
    "args": {
      "External id": 213261,
      "cbid": 211, "correlation": 213261
    }
  },
  {
    "ph": "s", "id": 213261, "pid": 76337, "tid": -914061504, "ts": 1716454224739345,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224801623, "dur": 21,
    "args": {
      "External id": 213269,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213269,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213269, "pid": 5, "tid": 7, "ts": 1716454224801623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739387, "dur": 9,
    "args": {
      "External id": 213269,
      "cbid": 211, "correlation": 213269
    }
  },
  {
    "ph": "s", "id": 213269, "pid": 76337, "tid": -914061504, "ts": 1716454224739387,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224801645, "dur": 17,
    "args": {
      "External id": 213291,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213291,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213291, "pid": 5, "tid": 7, "ts": 1716454224801645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739440, "dur": 10,
    "args": {
      "External id": 213291,
      "cbid": 211, "correlation": 213291
    }
  },
  {
    "ph": "s", "id": 213291, "pid": 76337, "tid": -914061504, "ts": 1716454224739440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739527, "dur": 1,
    "args": {
      "External id": 213307,
      "cbid": 251, "correlation": 213307
    }
  },
  {
    "ph": "f", "id": 213307, "pid": 76337, "tid": -914061504, "ts": 1716454224739527,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739532, "dur": 0,
    "args": {
      "External id": 213309,
      "cbid": 251, "correlation": 213309
    }
  },
  {
    "ph": "f", "id": 213309, "pid": 76337, "tid": -914061504, "ts": 1716454224739532,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224801664, "dur": 498,
    "args": {
      "External id": 213310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213310,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213310, "pid": 5, "tid": 7, "ts": 1716454224801664,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739534, "dur": 12,
    "args": {
      "External id": 213310,
      "cbid": 211, "correlation": 213310
    }
  },
  {
    "ph": "s", "id": 213310, "pid": 76337, "tid": -914061504, "ts": 1716454224739534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224802163, "dur": 65,
    "args": {
      "External id": 213318,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213318,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213318, "pid": 5, "tid": 7, "ts": 1716454224802163,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739599, "dur": 12,
    "args": {
      "External id": 213318,
      "cbid": 211, "correlation": 213318
    }
  },
  {
    "ph": "s", "id": 213318, "pid": 76337, "tid": -914061504, "ts": 1716454224739599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224802229, "dur": 68,
    "args": {
      "External id": 213326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213326,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213326, "pid": 5, "tid": 7, "ts": 1716454224802229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739631, "dur": 8,
    "args": {
      "External id": 213326,
      "cbid": 211, "correlation": 213326
    }
  },
  {
    "ph": "s", "id": 213326, "pid": 76337, "tid": -914061504, "ts": 1716454224739631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739710, "dur": 1,
    "args": {
      "External id": 213342,
      "cbid": 251, "correlation": 213342
    }
  },
  {
    "ph": "f", "id": 213342, "pid": 76337, "tid": -914061504, "ts": 1716454224739710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224802299, "dur": 1,
    "args": {
      "External id": 213344,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 213344,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15306122448979592
    }
  },
  {
    "ph": "f", "id": 213344, "pid": 5, "tid": 7, "ts": 1716454224802299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739715, "dur": 9,
    "args": {
      "External id": 213344,
      "cbid": 51, "correlation": 213344
    }
  },
  {
    "ph": "s", "id": 213344, "pid": 76337, "tid": -914061504, "ts": 1716454224739715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224802303, "dur": 271,
    "args": {
      "External id": 213345,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213345,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213345, "pid": 5, "tid": 7, "ts": 1716454224802303,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739726, "dur": 11,
    "args": {
      "External id": 213345,
      "cbid": 211, "correlation": 213345
    }
  },
  {
    "ph": "s", "id": 213345, "pid": 76337, "tid": -914061504, "ts": 1716454224739726,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224802575, "dur": 14,
    "args": {
      "External id": 213353,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213353,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213353, "pid": 5, "tid": 7, "ts": 1716454224802575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739768, "dur": 11,
    "args": {
      "External id": 213353,
      "cbid": 211, "correlation": 213353
    }
  },
  {
    "ph": "s", "id": 213353, "pid": 76337, "tid": -914061504, "ts": 1716454224739768,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224802591, "dur": 38,
    "args": {
      "External id": 213364,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213364,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213364, "pid": 5, "tid": 7, "ts": 1716454224802591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739837, "dur": 12,
    "args": {
      "External id": 213364,
      "cbid": 211, "correlation": 213364
    }
  },
  {
    "ph": "s", "id": 213364, "pid": 76337, "tid": -914061504, "ts": 1716454224739837,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739903, "dur": 0,
    "args": {
      "External id": 213376,
      "cbid": 317, "correlation": 213376
    }
  },
  {
    "ph": "f", "id": 213376, "pid": 76337, "tid": -914061504, "ts": 1716454224739903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739903, "dur": 0,
    "args": {
      "External id": 213377,
      "cbid": 203, "correlation": 213377
    }
  },
  {
    "ph": "f", "id": 213377, "pid": 76337, "tid": -914061504, "ts": 1716454224739903,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739904, "dur": 0,
    "args": {
      "External id": 213378,
      "cbid": 205, "correlation": 213378
    }
  },
  {
    "ph": "f", "id": 213378, "pid": 76337, "tid": -914061504, "ts": 1716454224739904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224802630, "dur": 13,
    "args": {
      "External id": 213382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213382,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213382, "pid": 5, "tid": 7, "ts": 1716454224802630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739919, "dur": 13,
    "args": {
      "External id": 213382,
      "cbid": 211, "correlation": 213382
    }
  },
  {
    "ph": "s", "id": 213382, "pid": 76337, "tid": -914061504, "ts": 1716454224739919,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224802645, "dur": 4,
    "args": {
      "External id": 213384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213384,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213384, "pid": 5, "tid": 7, "ts": 1716454224802645,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739937, "dur": 6,
    "args": {
      "External id": 213384,
      "cbid": 211, "correlation": 213384
    }
  },
  {
    "ph": "s", "id": 213384, "pid": 76337, "tid": -914061504, "ts": 1716454224739937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739946, "dur": 0,
    "args": {
      "External id": 213385,
      "cbid": 51, "correlation": 213385
    }
  },
  {
    "ph": "s", "id": 213385, "pid": 76337, "tid": -914061504, "ts": 1716454224739946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224802650, "dur": 98,
    "args": {
      "External id": 213386,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213386,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 213386, "pid": 5, "tid": 7, "ts": 1716454224802650,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739946, "dur": 5,
    "args": {
      "External id": 213386,
      "cbid": 211, "correlation": 213386
    }
  },
  {
    "ph": "s", "id": 213386, "pid": 76337, "tid": -914061504, "ts": 1716454224739946,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224802750, "dur": 16,
    "args": {
      "External id": 213391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213391,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213391, "pid": 5, "tid": 7, "ts": 1716454224802750,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224739982, "dur": 9,
    "args": {
      "External id": 213391,
      "cbid": 211, "correlation": 213391
    }
  },
  {
    "ph": "s", "id": 213391, "pid": 76337, "tid": -914061504, "ts": 1716454224739982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224802768, "dur": 12,
    "args": {
      "External id": 213399,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213399,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213399, "pid": 5, "tid": 7, "ts": 1716454224802768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740014, "dur": 8,
    "args": {
      "External id": 213399,
      "cbid": 211, "correlation": 213399
    }
  },
  {
    "ph": "s", "id": 213399, "pid": 76337, "tid": -914061504, "ts": 1716454224740014,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224802781, "dur": 24,
    "args": {
      "External id": 213408,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213408,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213408, "pid": 5, "tid": 7, "ts": 1716454224802781,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740054, "dur": 10,
    "args": {
      "External id": 213408,
      "cbid": 211, "correlation": 213408
    }
  },
  {
    "ph": "s", "id": 213408, "pid": 76337, "tid": -914061504, "ts": 1716454224740054,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224802806, "dur": 24,
    "args": {
      "External id": 213428,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213428,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 213428, "pid": 5, "tid": 7, "ts": 1716454224802806,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740126, "dur": 12,
    "args": {
      "External id": 213428,
      "cbid": 211, "correlation": 213428
    }
  },
  {
    "ph": "s", "id": 213428, "pid": 76337, "tid": -914061504, "ts": 1716454224740126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224802832, "dur": 5,
    "args": {
      "External id": 213440,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213440,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 213440, "pid": 5, "tid": 7, "ts": 1716454224802832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740148, "dur": 6,
    "args": {
      "External id": 213440,
      "cbid": 211, "correlation": 213440
    }
  },
  {
    "ph": "s", "id": 213440, "pid": 76337, "tid": -914061504, "ts": 1716454224740148,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224802838, "dur": 25,
    "args": {
      "External id": 213443,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213443,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213443, "pid": 5, "tid": 7, "ts": 1716454224802838,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740166, "dur": 6,
    "args": {
      "External id": 213443,
      "cbid": 211, "correlation": 213443
    }
  },
  {
    "ph": "s", "id": 213443, "pid": 76337, "tid": -914061504, "ts": 1716454224740166,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224802865, "dur": 18,
    "args": {
      "External id": 213452,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213452,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 72,
      "warps per SM": 288,
      "grid": [5760, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213452, "pid": 5, "tid": 7, "ts": 1716454224802865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740204, "dur": 9,
    "args": {
      "External id": 213452,
      "cbid": 211, "correlation": 213452
    }
  },
  {
    "ph": "s", "id": 213452, "pid": 76337, "tid": -914061504, "ts": 1716454224740204,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740257, "dur": 0,
    "args": {
      "External id": 213462,
      "cbid": 317, "correlation": 213462
    }
  },
  {
    "ph": "f", "id": 213462, "pid": 76337, "tid": -914061504, "ts": 1716454224740257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740257, "dur": 0,
    "args": {
      "External id": 213463,
      "cbid": 203, "correlation": 213463
    }
  },
  {
    "ph": "f", "id": 213463, "pid": 76337, "tid": -914061504, "ts": 1716454224740257,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740258, "dur": 0,
    "args": {
      "External id": 213464,
      "cbid": 205, "correlation": 213464
    }
  },
  {
    "ph": "f", "id": 213464, "pid": 76337, "tid": -914061504, "ts": 1716454224740258,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224802884, "dur": 17,
    "args": {
      "External id": 213468,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213468,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213468, "pid": 5, "tid": 7, "ts": 1716454224802884,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740273, "dur": 12,
    "args": {
      "External id": 213468,
      "cbid": 211, "correlation": 213468
    }
  },
  {
    "ph": "s", "id": 213468, "pid": 76337, "tid": -914061504, "ts": 1716454224740273,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224802902, "dur": 242,
    "args": {
      "External id": 213470,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213470,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 960,
      "warps per SM": 7680,
      "grid": [1, 60, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213470, "pid": 5, "tid": 7, "ts": 1716454224802902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740288, "dur": 5,
    "args": {
      "External id": 213470,
      "cbid": 211, "correlation": 213470
    }
  },
  {
    "ph": "s", "id": 213470, "pid": 76337, "tid": -914061504, "ts": 1716454224740288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224803146, "dur": 1,
    "args": {
      "External id": 213472,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 213472,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 213472, "pid": 5, "tid": 7, "ts": 1716454224803146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740299, "dur": 8,
    "args": {
      "External id": 213472,
      "cbid": 51, "correlation": 213472
    }
  },
  {
    "ph": "s", "id": 213472, "pid": 76337, "tid": -914061504, "ts": 1716454224740299,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224803150, "dur": 813,
    "args": {
      "External id": 213473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213473,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 24, 2],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213473, "pid": 5, "tid": 7, "ts": 1716454224803150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740307, "dur": 6,
    "args": {
      "External id": 213473,
      "cbid": 211, "correlation": 213473
    }
  },
  {
    "ph": "s", "id": 213473, "pid": 76337, "tid": -914061504, "ts": 1716454224740307,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224803965, "dur": 14,
    "args": {
      "External id": 213475,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213475,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213475, "pid": 5, "tid": 7, "ts": 1716454224803965,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740318, "dur": 5,
    "args": {
      "External id": 213475,
      "cbid": 211, "correlation": 213475
    }
  },
  {
    "ph": "s", "id": 213475, "pid": 76337, "tid": -914061504, "ts": 1716454224740318,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224803980, "dur": 15,
    "args": {
      "External id": 213481,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213481,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213481, "pid": 5, "tid": 7, "ts": 1716454224803980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740346, "dur": 8,
    "args": {
      "External id": 213481,
      "cbid": 211, "correlation": 213481
    }
  },
  {
    "ph": "s", "id": 213481, "pid": 76337, "tid": -914061504, "ts": 1716454224740346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224803996, "dur": 4,
    "args": {
      "External id": 213489,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213489,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 213489, "pid": 5, "tid": 7, "ts": 1716454224803996,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740390, "dur": 9,
    "args": {
      "External id": 213489,
      "cbid": 211, "correlation": 213489
    }
  },
  {
    "ph": "s", "id": 213489, "pid": 76337, "tid": -914061504, "ts": 1716454224740390,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740454, "dur": 1,
    "args": {
      "External id": 213505,
      "cbid": 251, "correlation": 213505
    }
  },
  {
    "ph": "f", "id": 213505, "pid": 76337, "tid": -914061504, "ts": 1716454224740454,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740459, "dur": 0,
    "args": {
      "External id": 213507,
      "cbid": 251, "correlation": 213507
    }
  },
  {
    "ph": "f", "id": 213507, "pid": 76337, "tid": -914061504, "ts": 1716454224740459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224804001, "dur": 14,
    "args": {
      "External id": 213508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213508,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213508, "pid": 5, "tid": 7, "ts": 1716454224804001,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740462, "dur": 11,
    "args": {
      "External id": 213508,
      "cbid": 211, "correlation": 213508
    }
  },
  {
    "ph": "s", "id": 213508, "pid": 76337, "tid": -914061504, "ts": 1716454224740462,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224804016, "dur": 5,
    "args": {
      "External id": 213510,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213510,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213510, "pid": 5, "tid": 7, "ts": 1716454224804016,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740475, "dur": 5,
    "args": {
      "External id": 213510,
      "cbid": 211, "correlation": 213510
    }
  },
  {
    "ph": "s", "id": 213510, "pid": 76337, "tid": -914061504, "ts": 1716454224740475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224804022, "dur": 17,
    "args": {
      "External id": 213520,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213520,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213520, "pid": 5, "tid": 7, "ts": 1716454224804022,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740534, "dur": 12,
    "args": {
      "External id": 213520,
      "cbid": 211, "correlation": 213520
    }
  },
  {
    "ph": "s", "id": 213520, "pid": 76337, "tid": -914061504, "ts": 1716454224740534,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224804041, "dur": 17,
    "args": {
      "External id": 213540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213540,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 213540, "pid": 5, "tid": 7, "ts": 1716454224804041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740599, "dur": 11,
    "args": {
      "External id": 213540,
      "cbid": 211, "correlation": 213540
    }
  },
  {
    "ph": "s", "id": 213540, "pid": 76337, "tid": -914061504, "ts": 1716454224740599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224804060, "dur": 4,
    "args": {
      "External id": 213552,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213552,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 213552, "pid": 5, "tid": 7, "ts": 1716454224804060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740620, "dur": 6,
    "args": {
      "External id": 213552,
      "cbid": 211, "correlation": 213552
    }
  },
  {
    "ph": "s", "id": 213552, "pid": 76337, "tid": -914061504, "ts": 1716454224740620,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224804065, "dur": 16,
    "args": {
      "External id": 213555,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213555,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213555, "pid": 5, "tid": 7, "ts": 1716454224804065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740638, "dur": 6,
    "args": {
      "External id": 213555,
      "cbid": 211, "correlation": 213555
    }
  },
  {
    "ph": "s", "id": 213555, "pid": 76337, "tid": -914061504, "ts": 1716454224740638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224804083, "dur": 11,
    "args": {
      "External id": 213564,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213564,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213564, "pid": 5, "tid": 7, "ts": 1716454224804083,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740678, "dur": 10,
    "args": {
      "External id": 213564,
      "cbid": 211, "correlation": 213564
    }
  },
  {
    "ph": "s", "id": 213564, "pid": 76337, "tid": -914061504, "ts": 1716454224740678,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740740, "dur": 0,
    "args": {
      "External id": 213574,
      "cbid": 317, "correlation": 213574
    }
  },
  {
    "ph": "f", "id": 213574, "pid": 76337, "tid": -914061504, "ts": 1716454224740740,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740741, "dur": 0,
    "args": {
      "External id": 213575,
      "cbid": 203, "correlation": 213575
    }
  },
  {
    "ph": "f", "id": 213575, "pid": 76337, "tid": -914061504, "ts": 1716454224740741,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740742, "dur": 0,
    "args": {
      "External id": 213576,
      "cbid": 205, "correlation": 213576
    }
  },
  {
    "ph": "f", "id": 213576, "pid": 76337, "tid": -914061504, "ts": 1716454224740742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224804095, "dur": 12,
    "args": {
      "External id": 213580,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213580,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213580, "pid": 5, "tid": 7, "ts": 1716454224804095,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740755, "dur": 12,
    "args": {
      "External id": 213580,
      "cbid": 211, "correlation": 213580
    }
  },
  {
    "ph": "s", "id": 213580, "pid": 76337, "tid": -914061504, "ts": 1716454224740755,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224804109, "dur": 164,
    "args": {
      "External id": 213582,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213582,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213582, "pid": 5, "tid": 7, "ts": 1716454224804109,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740769, "dur": 5,
    "args": {
      "External id": 213582,
      "cbid": 211, "correlation": 213582
    }
  },
  {
    "ph": "s", "id": 213582, "pid": 76337, "tid": -914061504, "ts": 1716454224740769,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224804275, "dur": 1,
    "args": {
      "External id": 213584,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 213584,
      "bytes": 960, "memory bandwidth (GB/s)": 0.5882352941176471
    }
  },
  {
    "ph": "f", "id": 213584, "pid": 5, "tid": 7, "ts": 1716454224804275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740780, "dur": 7,
    "args": {
      "External id": 213584,
      "cbid": 51, "correlation": 213584
    }
  },
  {
    "ph": "s", "id": 213584, "pid": 76337, "tid": -914061504, "ts": 1716454224740780,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_indexed_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224804279, "dur": 653,
    "args": {
      "External id": 213585,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213585,
      "registers per thread": 254,
      "shared memory": 41984,
      "blocks per SM": 4.5,
      "warps per SM": 36,
      "grid": [5, 24, 3],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213585, "pid": 5, "tid": 7, "ts": 1716454224804279,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740788, "dur": 6,
    "args": {
      "External id": 213585,
      "cbid": 211, "correlation": 213585
    }
  },
  {
    "ph": "s", "id": 213585, "pid": 76337, "tid": -914061504, "ts": 1716454224740788,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224804933, "dur": 12,
    "args": {
      "External id": 213587,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213587,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213587, "pid": 5, "tid": 7, "ts": 1716454224804933,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740798, "dur": 5,
    "args": {
      "External id": 213587,
      "cbid": 211, "correlation": 213587
    }
  },
  {
    "ph": "s", "id": 213587, "pid": 76337, "tid": -914061504, "ts": 1716454224740798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224804946, "dur": 15,
    "args": {
      "External id": 213593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213593,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213593, "pid": 5, "tid": 7, "ts": 1716454224804946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740827, "dur": 9,
    "args": {
      "External id": 213593,
      "cbid": 211, "correlation": 213593
    }
  },
  {
    "ph": "s", "id": 213593, "pid": 76337, "tid": -914061504, "ts": 1716454224740827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740885, "dur": 0,
    "args": {
      "External id": 213603,
      "cbid": 317, "correlation": 213603
    }
  },
  {
    "ph": "f", "id": 213603, "pid": 76337, "tid": -914061504, "ts": 1716454224740885,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740886, "dur": 0,
    "args": {
      "External id": 213604,
      "cbid": 203, "correlation": 213604
    }
  },
  {
    "ph": "f", "id": 213604, "pid": 76337, "tid": -914061504, "ts": 1716454224740886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740887, "dur": 0,
    "args": {
      "External id": 213605,
      "cbid": 205, "correlation": 213605
    }
  },
  {
    "ph": "f", "id": 213605, "pid": 76337, "tid": -914061504, "ts": 1716454224740887,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224804963, "dur": 17,
    "args": {
      "External id": 213609,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213609,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 36,
      "warps per SM": 288,
      "grid": [6, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213609, "pid": 5, "tid": 7, "ts": 1716454224804963,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740901, "dur": 11,
    "args": {
      "External id": 213609,
      "cbid": 211, "correlation": 213609
    }
  },
  {
    "ph": "s", "id": 213609, "pid": 76337, "tid": -914061504, "ts": 1716454224740901,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224804981, "dur": 4,
    "args": {
      "External id": 213611,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213611,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213611, "pid": 5, "tid": 7, "ts": 1716454224804981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740916, "dur": 6,
    "args": {
      "External id": 213611,
      "cbid": 211, "correlation": 213611
    }
  },
  {
    "ph": "s", "id": 213611, "pid": 76337, "tid": -914061504, "ts": 1716454224740916,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740925, "dur": 0,
    "args": {
      "External id": 213612,
      "cbid": 51, "correlation": 213612
    }
  },
  {
    "ph": "s", "id": 213612, "pid": 76337, "tid": -914061504, "ts": 1716454224740925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224804986, "dur": 132,
    "args": {
      "External id": 213613,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213613,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 213613, "pid": 5, "tid": 7, "ts": 1716454224804986,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740926, "dur": 5,
    "args": {
      "External id": 213613,
      "cbid": 211, "correlation": 213613
    }
  },
  {
    "ph": "s", "id": 213613, "pid": 76337, "tid": -914061504, "ts": 1716454224740926,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224805120, "dur": 16,
    "args": {
      "External id": 213618,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213618,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213618, "pid": 5, "tid": 7, "ts": 1716454224805120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740952, "dur": 8,
    "args": {
      "External id": 213618,
      "cbid": 211, "correlation": 213618
    }
  },
  {
    "ph": "s", "id": 213618, "pid": 76337, "tid": -914061504, "ts": 1716454224740952,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224805137, "dur": 12,
    "args": {
      "External id": 213626,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213626,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213626, "pid": 5, "tid": 7, "ts": 1716454224805137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224740991, "dur": 9,
    "args": {
      "External id": 213626,
      "cbid": 211, "correlation": 213626
    }
  },
  {
    "ph": "s", "id": 213626, "pid": 76337, "tid": -914061504, "ts": 1716454224740991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224805150, "dur": 10,
    "args": {
      "External id": 213634,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213634,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213634, "pid": 5, "tid": 7, "ts": 1716454224805150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741022, "dur": 8,
    "args": {
      "External id": 213634,
      "cbid": 211, "correlation": 213634
    }
  },
  {
    "ph": "s", "id": 213634, "pid": 76337, "tid": -914061504, "ts": 1716454224741022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224805162, "dur": 19,
    "args": {
      "External id": 213654,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213654,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 213654, "pid": 5, "tid": 7, "ts": 1716454224805162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741105, "dur": 12,
    "args": {
      "External id": 213654,
      "cbid": 211, "correlation": 213654
    }
  },
  {
    "ph": "s", "id": 213654, "pid": 76337, "tid": -914061504, "ts": 1716454224741105,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224805182, "dur": 4,
    "args": {
      "External id": 213666,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213666,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 213666, "pid": 5, "tid": 7, "ts": 1716454224805182,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741126, "dur": 6,
    "args": {
      "External id": 213666,
      "cbid": 211, "correlation": 213666
    }
  },
  {
    "ph": "s", "id": 213666, "pid": 76337, "tid": -914061504, "ts": 1716454224741126,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224805188, "dur": 17,
    "args": {
      "External id": 213669,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213669,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213669, "pid": 5, "tid": 7, "ts": 1716454224805188,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741145, "dur": 6,
    "args": {
      "External id": 213669,
      "cbid": 211, "correlation": 213669
    }
  },
  {
    "ph": "s", "id": 213669, "pid": 76337, "tid": -914061504, "ts": 1716454224741145,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741201, "dur": 0,
    "args": {
      "External id": 213680,
      "cbid": 317, "correlation": 213680
    }
  },
  {
    "ph": "f", "id": 213680, "pid": 76337, "tid": -914061504, "ts": 1716454224741201,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741202, "dur": 0,
    "args": {
      "External id": 213681,
      "cbid": 203, "correlation": 213681
    }
  },
  {
    "ph": "f", "id": 213681, "pid": 76337, "tid": -914061504, "ts": 1716454224741202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741203, "dur": 0,
    "args": {
      "External id": 213682,
      "cbid": 205, "correlation": 213682
    }
  },
  {
    "ph": "f", "id": 213682, "pid": 76337, "tid": -914061504, "ts": 1716454224741203,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224805206, "dur": 13,
    "args": {
      "External id": 213686,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213686,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213686, "pid": 5, "tid": 7, "ts": 1716454224805206,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741216, "dur": 12,
    "args": {
      "External id": 213686,
      "cbid": 211, "correlation": 213686
    }
  },
  {
    "ph": "s", "id": 213686, "pid": 76337, "tid": -914061504, "ts": 1716454224741216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224805220, "dur": 3,
    "args": {
      "External id": 213688,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213688,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213688, "pid": 5, "tid": 7, "ts": 1716454224805220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741232, "dur": 6,
    "args": {
      "External id": 213688,
      "cbid": 211, "correlation": 213688
    }
  },
  {
    "ph": "s", "id": 213688, "pid": 76337, "tid": -914061504, "ts": 1716454224741232,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741241, "dur": 0,
    "args": {
      "External id": 213689,
      "cbid": 51, "correlation": 213689
    }
  },
  {
    "ph": "s", "id": 213689, "pid": 76337, "tid": -914061504, "ts": 1716454224741241,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224805224, "dur": 91,
    "args": {
      "External id": 213690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213690,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 213690, "pid": 5, "tid": 7, "ts": 1716454224805224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741242, "dur": 5,
    "args": {
      "External id": 213690,
      "cbid": 211, "correlation": 213690
    }
  },
  {
    "ph": "s", "id": 213690, "pid": 76337, "tid": -914061504, "ts": 1716454224741242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224805317, "dur": 16,
    "args": {
      "External id": 213695,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213695,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213695, "pid": 5, "tid": 7, "ts": 1716454224805317,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741268, "dur": 8,
    "args": {
      "External id": 213695,
      "cbid": 211, "correlation": 213695
    }
  },
  {
    "ph": "s", "id": 213695, "pid": 76337, "tid": -914061504, "ts": 1716454224741268,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224805334, "dur": 85,
    "args": {
      "External id": 213704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213704,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213704, "pid": 5, "tid": 7, "ts": 1716454224805334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741350, "dur": 13,
    "args": {
      "External id": 213704,
      "cbid": 211, "correlation": 213704
    }
  },
  {
    "ph": "s", "id": 213704, "pid": 76337, "tid": -914061504, "ts": 1716454224741350,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224805421, "dur": 30,
    "args": {
      "External id": 213726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213726,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [3072, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213726, "pid": 5, "tid": 7, "ts": 1716454224805421,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741405, "dur": 10,
    "args": {
      "External id": 213726,
      "cbid": 211, "correlation": 213726
    }
  },
  {
    "ph": "s", "id": 213726, "pid": 76337, "tid": -914061504, "ts": 1716454224741405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741496, "dur": 1,
    "args": {
      "External id": 213737,
      "cbid": 251, "correlation": 213737
    }
  },
  {
    "ph": "f", "id": 213737, "pid": 76337, "tid": -914061504, "ts": 1716454224741496,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224805452, "dur": 166,
    "args": {
      "External id": 213738,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213738,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213738, "pid": 5, "tid": 7, "ts": 1716454224805452,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741502, "dur": 13,
    "args": {
      "External id": 213738,
      "cbid": 211, "correlation": 213738
    }
  },
  {
    "ph": "s", "id": 213738, "pid": 76337, "tid": -914061504, "ts": 1716454224741502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741571, "dur": 1,
    "args": {
      "External id": 213749,
      "cbid": 251, "correlation": 213749
    }
  },
  {
    "ph": "f", "id": 213749, "pid": 76337, "tid": -914061504, "ts": 1716454224741571,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224805619, "dur": 159,
    "args": {
      "External id": 213750,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213750,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213750, "pid": 5, "tid": 7, "ts": 1716454224805619,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741575, "dur": 12,
    "args": {
      "External id": 213750,
      "cbid": 211, "correlation": 213750
    }
  },
  {
    "ph": "s", "id": 213750, "pid": 76337, "tid": -914061504, "ts": 1716454224741575,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741642, "dur": 1,
    "args": {
      "External id": 213761,
      "cbid": 251, "correlation": 213761
    }
  },
  {
    "ph": "f", "id": 213761, "pid": 76337, "tid": -914061504, "ts": 1716454224741642,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224805780, "dur": 161,
    "args": {
      "External id": 213762,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213762,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213762, "pid": 5, "tid": 7, "ts": 1716454224805780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741646, "dur": 11,
    "args": {
      "External id": 213762,
      "cbid": 211, "correlation": 213762
    }
  },
  {
    "ph": "s", "id": 213762, "pid": 76337, "tid": -914061504, "ts": 1716454224741646,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224805942, "dur": 337,
    "args": {
      "External id": 213787,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213787,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 9.6,
      "warps per SM": 38.4,
      "grid": [12, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213787, "pid": 5, "tid": 7, "ts": 1716454224805942,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741730, "dur": 13,
    "args": {
      "External id": 213787,
      "cbid": 211, "correlation": 213787
    }
  },
  {
    "ph": "s", "id": 213787, "pid": 76337, "tid": -914061504, "ts": 1716454224741730,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741830, "dur": 1,
    "args": {
      "External id": 213805,
      "cbid": 251, "correlation": 213805
    }
  },
  {
    "ph": "f", "id": 213805, "pid": 76337, "tid": -914061504, "ts": 1716454224741830,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224806280, "dur": 166,
    "args": {
      "External id": 213807,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213807,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [10, 24, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213807, "pid": 5, "tid": 7, "ts": 1716454224806280,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741836, "dur": 14,
    "args": {
      "External id": 213807,
      "cbid": 211, "correlation": 213807
    }
  },
  {
    "ph": "s", "id": 213807, "pid": 76337, "tid": -914061504, "ts": 1716454224741836,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224806448, "dur": 19,
    "args": {
      "External id": 213815,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213815,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213815, "pid": 5, "tid": 7, "ts": 1716454224806448,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741906, "dur": 12,
    "args": {
      "External id": 213815,
      "cbid": 211, "correlation": 213815
    }
  },
  {
    "ph": "s", "id": 213815, "pid": 76337, "tid": -914061504, "ts": 1716454224741906,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224806468, "dur": 28,
    "args": {
      "External id": 213823,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213823,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213823, "pid": 5, "tid": 7, "ts": 1716454224806468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224741945, "dur": 8,
    "args": {
      "External id": 213823,
      "cbid": 211, "correlation": 213823
    }
  },
  {
    "ph": "s", "id": 213823, "pid": 76337, "tid": -914061504, "ts": 1716454224741945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224806497, "dur": 18,
    "args": {
      "External id": 213834,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213834,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213834, "pid": 5, "tid": 7, "ts": 1716454224806497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742027, "dur": 14,
    "args": {
      "External id": 213834,
      "cbid": 211, "correlation": 213834
    }
  },
  {
    "ph": "s", "id": 213834, "pid": 76337, "tid": -914061504, "ts": 1716454224742027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224806517, "dur": 16,
    "args": {
      "External id": 213856,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213856,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213856, "pid": 5, "tid": 7, "ts": 1716454224806517,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742059, "dur": 7,
    "args": {
      "External id": 213856,
      "cbid": 211, "correlation": 213856
    }
  },
  {
    "ph": "s", "id": 213856, "pid": 76337, "tid": -914061504, "ts": 1716454224742059,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742146, "dur": 1,
    "args": {
      "External id": 213867,
      "cbid": 251, "correlation": 213867
    }
  },
  {
    "ph": "f", "id": 213867, "pid": 76337, "tid": -914061504, "ts": 1716454224742146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224806535, "dur": 90,
    "args": {
      "External id": 213868,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213868,
      "registers per thread": 250,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213868, "pid": 5, "tid": 7, "ts": 1716454224806535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742151, "dur": 13,
    "args": {
      "External id": 213868,
      "cbid": 211, "correlation": 213868
    }
  },
  {
    "ph": "s", "id": 213868, "pid": 76337, "tid": -914061504, "ts": 1716454224742151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742220, "dur": 1,
    "args": {
      "External id": 213879,
      "cbid": 251, "correlation": 213879
    }
  },
  {
    "ph": "f", "id": 213879, "pid": 76337, "tid": -914061504, "ts": 1716454224742220,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742224, "dur": 0,
    "args": {
      "External id": 213880,
      "cbid": 251, "correlation": 213880
    }
  },
  {
    "ph": "f", "id": 213880, "pid": 76337, "tid": -914061504, "ts": 1716454224742224,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224806626, "dur": 12,
    "args": {
      "External id": 213881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213881,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213881, "pid": 5, "tid": 7, "ts": 1716454224806626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742226, "dur": 12,
    "args": {
      "External id": 213881,
      "cbid": 211, "correlation": 213881
    }
  },
  {
    "ph": "s", "id": 213881, "pid": 76337, "tid": -914061504, "ts": 1716454224742226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224806640, "dur": 6,
    "args": {
      "External id": 213883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213883,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213883, "pid": 5, "tid": 7, "ts": 1716454224806640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742239, "dur": 7,
    "args": {
      "External id": 213883,
      "cbid": 211, "correlation": 213883
    }
  },
  {
    "ph": "s", "id": 213883, "pid": 76337, "tid": -914061504, "ts": 1716454224742239,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742298, "dur": 1,
    "args": {
      "External id": 213894,
      "cbid": 251, "correlation": 213894
    }
  },
  {
    "ph": "f", "id": 213894, "pid": 76337, "tid": -914061504, "ts": 1716454224742298,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742301, "dur": 0,
    "args": {
      "External id": 213895,
      "cbid": 251, "correlation": 213895
    }
  },
  {
    "ph": "f", "id": 213895, "pid": 76337, "tid": -914061504, "ts": 1716454224742301,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224806647, "dur": 9,
    "args": {
      "External id": 213896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213896,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 2,
      "warps per SM": 8,
      "grid": [8, 5, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213896, "pid": 5, "tid": 7, "ts": 1716454224806647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742302, "dur": 11,
    "args": {
      "External id": 213896,
      "cbid": 211, "correlation": 213896
    }
  },
  {
    "ph": "s", "id": 213896, "pid": 76337, "tid": -914061504, "ts": 1716454224742302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224806657, "dur": 3,
    "args": {
      "External id": 213898,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213898,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 8,
      "grid": [40, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213898, "pid": 5, "tid": 7, "ts": 1716454224806657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742315, "dur": 5,
    "args": {
      "External id": 213898,
      "cbid": 211, "correlation": 213898
    }
  },
  {
    "ph": "s", "id": 213898, "pid": 76337, "tid": -914061504, "ts": 1716454224742315,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, false>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224806661, "dur": 55,
    "args": {
      "External id": 213923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213923,
      "registers per thread": 168,
      "shared memory": 33280,
      "blocks per SM": 4.8,
      "warps per SM": 19.2,
      "grid": [6, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 213923, "pid": 5, "tid": 7, "ts": 1716454224806661,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742392, "dur": 13,
    "args": {
      "External id": 213923,
      "cbid": 211, "correlation": 213923
    }
  },
  {
    "ph": "s", "id": 213923, "pid": 76337, "tid": -914061504, "ts": 1716454224742392,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742491, "dur": 1,
    "args": {
      "External id": 213941,
      "cbid": 251, "correlation": 213941
    }
  },
  {
    "ph": "f", "id": 213941, "pid": 76337, "tid": -914061504, "ts": 1716454224742491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224806718, "dur": 92,
    "args": {
      "External id": 213943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213943,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [5, 12, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 213943, "pid": 5, "tid": 7, "ts": 1716454224806718,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742498, "dur": 14,
    "args": {
      "External id": 213943,
      "cbid": 211, "correlation": 213943
    }
  },
  {
    "ph": "s", "id": 213943, "pid": 76337, "tid": -914061504, "ts": 1716454224742498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224806812, "dur": 10,
    "args": {
      "External id": 213951,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213951,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213951, "pid": 5, "tid": 7, "ts": 1716454224806812,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742567, "dur": 12,
    "args": {
      "External id": 213951,
      "cbid": 211, "correlation": 213951
    }
  },
  {
    "ph": "s", "id": 213951, "pid": 76337, "tid": -914061504, "ts": 1716454224742567,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224806823, "dur": 22,
    "args": {
      "External id": 213959,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213959,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213959, "pid": 5, "tid": 7, "ts": 1716454224806823,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742608, "dur": 9,
    "args": {
      "External id": 213959,
      "cbid": 211, "correlation": 213959
    }
  },
  {
    "ph": "s", "id": 213959, "pid": 76337, "tid": -914061504, "ts": 1716454224742608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224806846, "dur": 18,
    "args": {
      "External id": 213981,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 213981,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [1536, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 213981, "pid": 5, "tid": 7, "ts": 1716454224806846,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742659, "dur": 10,
    "args": {
      "External id": 213981,
      "cbid": 211, "correlation": 213981
    }
  },
  {
    "ph": "s", "id": 213981, "pid": 76337, "tid": -914061504, "ts": 1716454224742659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742746, "dur": 1,
    "args": {
      "External id": 213997,
      "cbid": 251, "correlation": 213997
    }
  },
  {
    "ph": "f", "id": 213997, "pid": 76337, "tid": -914061504, "ts": 1716454224742746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742751, "dur": 0,
    "args": {
      "External id": 213999,
      "cbid": 251, "correlation": 213999
    }
  },
  {
    "ph": "f", "id": 213999, "pid": 76337, "tid": -914061504, "ts": 1716454224742751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8>(cutlass_70_tensorop_f16_s884gemm_relu_f16_128x128_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224806865, "dur": 499,
    "args": {
      "External id": 214000,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214000,
      "registers per thread": 254,
      "shared memory": 32768,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214000, "pid": 5, "tid": 7, "ts": 1716454224806865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742753, "dur": 12,
    "args": {
      "External id": 214000,
      "cbid": 211, "correlation": 214000
    }
  },
  {
    "ph": "s", "id": 214000, "pid": 76337, "tid": -914061504, "ts": 1716454224742753,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224807365, "dur": 66,
    "args": {
      "External id": 214008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214008,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214008, "pid": 5, "tid": 7, "ts": 1716454224807365,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742818, "dur": 12,
    "args": {
      "External id": 214008,
      "cbid": 211, "correlation": 214008
    }
  },
  {
    "ph": "s", "id": 214008, "pid": 76337, "tid": -914061504, "ts": 1716454224742818,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224807433, "dur": 65,
    "args": {
      "External id": 214016,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214016,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214016, "pid": 5, "tid": 7, "ts": 1716454224807433,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742849, "dur": 8,
    "args": {
      "External id": 214016,
      "cbid": 211, "correlation": 214016
    }
  },
  {
    "ph": "s", "id": 214016, "pid": 76337, "tid": -914061504, "ts": 1716454224742849,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742927, "dur": 1,
    "args": {
      "External id": 214032,
      "cbid": 251, "correlation": 214032
    }
  },
  {
    "ph": "f", "id": 214032, "pid": 76337, "tid": -914061504, "ts": 1716454224742927,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "gpu_memset", "name": "Memset (Device)", "pid": 5, "tid": 7,
    "ts": 1716454224807500, "dur": 1,
    "args": {
      "External id": 214034,
      "device": 5, "context": 1,
      "stream": 7, "correlation": 214034,
      "bytes": 240, "memory bandwidth (GB/s)": 0.15625
    }
  },
  {
    "ph": "f", "id": 214034, "pid": 5, "tid": 7, "ts": 1716454224807500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742932, "dur": 9,
    "args": {
      "External id": 214034,
      "cbid": 51, "correlation": 214034
    }
  },
  {
    "ph": "s", "id": 214034, "pid": 76337, "tid": -914061504, "ts": 1716454224742932,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_256x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224807504, "dur": 272,
    "args": {
      "External id": 214035,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214035,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 12, 4],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 214035, "pid": 5, "tid": 7, "ts": 1716454224807504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742943, "dur": 12,
    "args": {
      "External id": 214035,
      "cbid": 211, "correlation": 214035
    }
  },
  {
    "ph": "s", "id": 214035, "pid": 76337, "tid": -914061504, "ts": 1716454224742943,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224807778, "dur": 14,
    "args": {
      "External id": 214043,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214043,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214043, "pid": 5, "tid": 7, "ts": 1716454224807778,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224742993, "dur": 12,
    "args": {
      "External id": 214043,
      "cbid": 211, "correlation": 214043
    }
  },
  {
    "ph": "s", "id": 214043, "pid": 76337, "tid": -914061504, "ts": 1716454224742993,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224807793, "dur": 38,
    "args": {
      "External id": 214054,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214054,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214054, "pid": 5, "tid": 7, "ts": 1716454224807793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743062, "dur": 12,
    "args": {
      "External id": 214054,
      "cbid": 211, "correlation": 214054
    }
  },
  {
    "ph": "s", "id": 214054, "pid": 76337, "tid": -914061504, "ts": 1716454224743062,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743127, "dur": 0,
    "args": {
      "External id": 214066,
      "cbid": 317, "correlation": 214066
    }
  },
  {
    "ph": "f", "id": 214066, "pid": 76337, "tid": -914061504, "ts": 1716454224743127,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743128, "dur": 0,
    "args": {
      "External id": 214067,
      "cbid": 203, "correlation": 214067
    }
  },
  {
    "ph": "f", "id": 214067, "pid": 76337, "tid": -914061504, "ts": 1716454224743128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743128, "dur": 0,
    "args": {
      "External id": 214068,
      "cbid": 205, "correlation": 214068
    }
  },
  {
    "ph": "f", "id": 214068, "pid": 76337, "tid": -914061504, "ts": 1716454224743128,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224807832, "dur": 13,
    "args": {
      "External id": 214072,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214072,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 24,
      "warps per SM": 192,
      "grid": [6, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214072, "pid": 5, "tid": 7, "ts": 1716454224807832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743143, "dur": 13,
    "args": {
      "External id": 214072,
      "cbid": 211, "correlation": 214072
    }
  },
  {
    "ph": "s", "id": 214072, "pid": 76337, "tid": -914061504, "ts": 1716454224743143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224807847, "dur": 4,
    "args": {
      "External id": 214074,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214074,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.0125,
      "warps per SM": 0.1,
      "grid": [1, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 214074, "pid": 5, "tid": 7, "ts": 1716454224807847,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743161, "dur": 6,
    "args": {
      "External id": 214074,
      "cbid": 211, "correlation": 214074
    }
  },
  {
    "ph": "s", "id": 214074, "pid": 76337, "tid": -914061504, "ts": 1716454224743161,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743169, "dur": 0,
    "args": {
      "External id": 214075,
      "cbid": 51, "correlation": 214075
    }
  },
  {
    "ph": "s", "id": 214075, "pid": 76337, "tid": -914061504, "ts": 1716454224743169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224807852, "dur": 98,
    "args": {
      "External id": 214076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214076,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 1.5,
      "warps per SM": 6,
      "grid": [6, 20, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 214076, "pid": 5, "tid": 7, "ts": 1716454224807852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743170, "dur": 5,
    "args": {
      "External id": 214076,
      "cbid": 211, "correlation": 214076
    }
  },
  {
    "ph": "s", "id": 214076, "pid": 76337, "tid": -914061504, "ts": 1716454224743170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224807951, "dur": 17,
    "args": {
      "External id": 214081,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214081,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214081, "pid": 5, "tid": 7, "ts": 1716454224807951,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743197, "dur": 9,
    "args": {
      "External id": 214081,
      "cbid": 211, "correlation": 214081
    }
  },
  {
    "ph": "s", "id": 214081, "pid": 76337, "tid": -914061504, "ts": 1716454224743197,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224807969, "dur": 12,
    "args": {
      "External id": 214089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214089,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [3840, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214089, "pid": 5, "tid": 7, "ts": 1716454224807969,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743229, "dur": 8,
    "args": {
      "External id": 214089,
      "cbid": 211, "correlation": 214089
    }
  },
  {
    "ph": "s", "id": 214089, "pid": 76337, "tid": -914061504, "ts": 1716454224743229,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224807983, "dur": 57,
    "args": {
      "External id": 214100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214100,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 64,
      "warps per SM": 1024,
      "grid": [2, 1, 2560],
      "block": [16, 32, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214100, "pid": 5, "tid": 7, "ts": 1716454224807983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743293, "dur": 12,
    "args": {
      "External id": 214100,
      "cbid": 211, "correlation": 214100
    }
  },
  {
    "ph": "s", "id": 214100, "pid": 76337, "tid": -914061504, "ts": 1716454224743293,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743350, "dur": 0,
    "args": {
      "External id": 214110,
      "cbid": 317, "correlation": 214110
    }
  },
  {
    "ph": "f", "id": 214110, "pid": 76337, "tid": -914061504, "ts": 1716454224743350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743351, "dur": 0,
    "args": {
      "External id": 214111,
      "cbid": 203, "correlation": 214111
    }
  },
  {
    "ph": "f", "id": 214111, "pid": 76337, "tid": -914061504, "ts": 1716454224743351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743351, "dur": 0,
    "args": {
      "External id": 214112,
      "cbid": 205, "correlation": 214112
    }
  },
  {
    "ph": "f", "id": 214112, "pid": 76337, "tid": -914061504, "ts": 1716454224743351,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224808041, "dur": 39,
    "args": {
      "External id": 214116,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214116,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214116, "pid": 5, "tid": 7, "ts": 1716454224808041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743368, "dur": 11,
    "args": {
      "External id": 214116,
      "cbid": 211, "correlation": 214116
    }
  },
  {
    "ph": "s", "id": 214116, "pid": 76337, "tid": -914061504, "ts": 1716454224743368,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224808081, "dur": 164,
    "args": {
      "External id": 214118,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214118,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 640,
      "warps per SM": 5120,
      "grid": [1, 40, 1280],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214118, "pid": 5, "tid": 7, "ts": 1716454224808081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743382, "dur": 5,
    "args": {
      "External id": 214118,
      "cbid": 211, "correlation": 214118
    }
  },
  {
    "ph": "s", "id": 214118, "pid": 76337, "tid": -914061504, "ts": 1716454224743382,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224808246, "dur": 1974,
    "args": {
      "External id": 214120,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214120,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [10, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214120, "pid": 5, "tid": 7, "ts": 1716454224808246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743393, "dur": 8,
    "args": {
      "External id": 214120,
      "cbid": 211, "correlation": 214120
    }
  },
  {
    "ph": "s", "id": 214120, "pid": 76337, "tid": -914061504, "ts": 1716454224743393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224810222, "dur": 38,
    "args": {
      "External id": 214122,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214122,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214122, "pid": 5, "tid": 7, "ts": 1716454224810222,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743405, "dur": 5,
    "args": {
      "External id": 214122,
      "cbid": 211, "correlation": 214122
    }
  },
  {
    "ph": "s", "id": 214122, "pid": 76337, "tid": -914061504, "ts": 1716454224743405,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224810261, "dur": 59,
    "args": {
      "External id": 214128,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214128,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214128, "pid": 5, "tid": 7, "ts": 1716454224810261,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743432, "dur": 8,
    "args": {
      "External id": 214128,
      "cbid": 211, "correlation": 214128
    }
  },
  {
    "ph": "s", "id": 214128, "pid": 76337, "tid": -914061504, "ts": 1716454224743432,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224810322, "dur": 86,
    "args": {
      "External id": 214137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214137,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214137, "pid": 5, "tid": 7, "ts": 1716454224810322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743522, "dur": 13,
    "args": {
      "External id": 214137,
      "cbid": 211, "correlation": 214137
    }
  },
  {
    "ph": "s", "id": 214137, "pid": 76337, "tid": -914061504, "ts": 1716454224743522,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224810409, "dur": 74,
    "args": {
      "External id": 214157,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214157,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 214157, "pid": 5, "tid": 7, "ts": 1716454224810409,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743610, "dur": 12,
    "args": {
      "External id": 214157,
      "cbid": 211, "correlation": 214157
    }
  },
  {
    "ph": "s", "id": 214157, "pid": 76337, "tid": -914061504, "ts": 1716454224743610,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224810485, "dur": 5,
    "args": {
      "External id": 214169,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214169,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.75,
      "warps per SM": 6,
      "grid": [60, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 9
    }
  },
  {
    "ph": "f", "id": 214169, "pid": 5, "tid": 7, "ts": 1716454224810485,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743632, "dur": 6,
    "args": {
      "External id": 214169,
      "cbid": 211, "correlation": 214169
    }
  },
  {
    "ph": "s", "id": 214169, "pid": 76337, "tid": -914061504, "ts": 1716454224743632,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224810491, "dur": 83,
    "args": {
      "External id": 214172,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214172,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214172, "pid": 5, "tid": 7, "ts": 1716454224810491,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743652, "dur": 7,
    "args": {
      "External id": 214172,
      "cbid": 211, "correlation": 214172
    }
  },
  {
    "ph": "s", "id": 214172, "pid": 76337, "tid": -914061504, "ts": 1716454224743652,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224810576, "dur": 53,
    "args": {
      "External id": 214181,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214181,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 288,
      "warps per SM": 1152,
      "grid": [23040, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214181, "pid": 5, "tid": 7, "ts": 1716454224810576,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743692, "dur": 10,
    "args": {
      "External id": 214181,
      "cbid": 211, "correlation": 214181
    }
  },
  {
    "ph": "s", "id": 214181, "pid": 76337, "tid": -914061504, "ts": 1716454224743692,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743746, "dur": 0,
    "args": {
      "External id": 214191,
      "cbid": 317, "correlation": 214191
    }
  },
  {
    "ph": "f", "id": 214191, "pid": 76337, "tid": -914061504, "ts": 1716454224743746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743746, "dur": 0,
    "args": {
      "External id": 214192,
      "cbid": 203, "correlation": 214192
    }
  },
  {
    "ph": "f", "id": 214192, "pid": 76337, "tid": -914061504, "ts": 1716454224743746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743747, "dur": 0,
    "args": {
      "External id": 214193,
      "cbid": 205, "correlation": 214193
    }
  },
  {
    "ph": "f", "id": 214193, "pid": 76337, "tid": -914061504, "ts": 1716454224743747,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224810630, "dur": 58,
    "args": {
      "External id": 214197,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214197,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214197, "pid": 5, "tid": 7, "ts": 1716454224810630,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743764, "dur": 12,
    "args": {
      "External id": 214197,
      "cbid": 211, "correlation": 214197
    }
  },
  {
    "ph": "s", "id": 214197, "pid": 76337, "tid": -914061504, "ts": 1716454224743764,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224810689, "dur": 123,
    "args": {
      "External id": 214199,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214199,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 480,
      "warps per SM": 3840,
      "grid": [1, 60, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214199, "pid": 5, "tid": 7, "ts": 1716454224810689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743779, "dur": 5,
    "args": {
      "External id": 214199,
      "cbid": 211, "correlation": 214199
    }
  },
  {
    "ph": "s", "id": 214199, "pid": 76337, "tid": -914061504, "ts": 1716454224743779,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224810814, "dur": 1905,
    "args": {
      "External id": 214201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214201,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214201, "pid": 5, "tid": 7, "ts": 1716454224810814,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743790, "dur": 6,
    "args": {
      "External id": 214201,
      "cbid": 211, "correlation": 214201
    }
  },
  {
    "ph": "s", "id": 214201, "pid": 76337, "tid": -914061504, "ts": 1716454224743790,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224812720, "dur": 21,
    "args": {
      "External id": 214203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214203,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214203, "pid": 5, "tid": 7, "ts": 1716454224812720,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743800, "dur": 5,
    "args": {
      "External id": 214203,
      "cbid": 211, "correlation": 214203
    }
  },
  {
    "ph": "s", "id": 214203, "pid": 76337, "tid": -914061504, "ts": 1716454224743800,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224812742, "dur": 33,
    "args": {
      "External id": 214209,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214209,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214209, "pid": 5, "tid": 7, "ts": 1716454224812742,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743827, "dur": 8,
    "args": {
      "External id": 214209,
      "cbid": 211, "correlation": 214209
    }
  },
  {
    "ph": "s", "id": 214209, "pid": 76337, "tid": -914061504, "ts": 1716454224743827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224812777, "dur": 3,
    "args": {
      "External id": 214217,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214217,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 214217, "pid": 5, "tid": 7, "ts": 1716454224812777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743871, "dur": 9,
    "args": {
      "External id": 214217,
      "cbid": 211, "correlation": 214217
    }
  },
  {
    "ph": "s", "id": 214217, "pid": 76337, "tid": -914061504, "ts": 1716454224743871,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743937, "dur": 1,
    "args": {
      "External id": 214233,
      "cbid": 251, "correlation": 214233
    }
  },
  {
    "ph": "f", "id": 214233, "pid": 76337, "tid": -914061504, "ts": 1716454224743937,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743943, "dur": 0,
    "args": {
      "External id": 214235,
      "cbid": 251, "correlation": 214235
    }
  },
  {
    "ph": "f", "id": 214235, "pid": 76337, "tid": -914061504, "ts": 1716454224743943,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224812782, "dur": 12,
    "args": {
      "External id": 214236,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214236,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 214236, "pid": 5, "tid": 7, "ts": 1716454224812782,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743945, "dur": 11,
    "args": {
      "External id": 214236,
      "cbid": 211, "correlation": 214236
    }
  },
  {
    "ph": "s", "id": 214236, "pid": 76337, "tid": -914061504, "ts": 1716454224743945,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224812795, "dur": 5,
    "args": {
      "External id": 214238,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214238,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 214238, "pid": 5, "tid": 7, "ts": 1716454224812795,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224743959, "dur": 5,
    "args": {
      "External id": 214238,
      "cbid": 211, "correlation": 214238
    }
  },
  {
    "ph": "s", "id": 214238, "pid": 76337, "tid": -914061504, "ts": 1716454224743959,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224812802, "dur": 28,
    "args": {
      "External id": 214248,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214248,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214248, "pid": 5, "tid": 7, "ts": 1716454224812802,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744025, "dur": 13,
    "args": {
      "External id": 214248,
      "cbid": 211, "correlation": 214248
    }
  },
  {
    "ph": "s", "id": 214248, "pid": 76337, "tid": -914061504, "ts": 1716454224744025,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224812832, "dur": 32,
    "args": {
      "External id": 214268,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214268,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 214268, "pid": 5, "tid": 7, "ts": 1716454224812832,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744091, "dur": 11,
    "args": {
      "External id": 214268,
      "cbid": 211, "correlation": 214268
    }
  },
  {
    "ph": "s", "id": 214268, "pid": 76337, "tid": -914061504, "ts": 1716454224744091,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224812865, "dur": 4,
    "args": {
      "External id": 214280,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214280,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 214280, "pid": 5, "tid": 7, "ts": 1716454224812865,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744112, "dur": 6,
    "args": {
      "External id": 214280,
      "cbid": 211, "correlation": 214280
    }
  },
  {
    "ph": "s", "id": 214280, "pid": 76337, "tid": -914061504, "ts": 1716454224744112,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224812871, "dur": 29,
    "args": {
      "External id": 214283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214283,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214283, "pid": 5, "tid": 7, "ts": 1716454224812871,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744129, "dur": 6,
    "args": {
      "External id": 214283,
      "cbid": 211, "correlation": 214283
    }
  },
  {
    "ph": "s", "id": 214283, "pid": 76337, "tid": -914061504, "ts": 1716454224744129,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224812901, "dur": 20,
    "args": {
      "External id": 214292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214292,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214292, "pid": 5, "tid": 7, "ts": 1716454224812901,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744169, "dur": 9,
    "args": {
      "External id": 214292,
      "cbid": 211, "correlation": 214292
    }
  },
  {
    "ph": "s", "id": 214292, "pid": 76337, "tid": -914061504, "ts": 1716454224744169,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744232, "dur": 0,
    "args": {
      "External id": 214302,
      "cbid": 317, "correlation": 214302
    }
  },
  {
    "ph": "f", "id": 214302, "pid": 76337, "tid": -914061504, "ts": 1716454224744232,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744233, "dur": 0,
    "args": {
      "External id": 214303,
      "cbid": 203, "correlation": 214303
    }
  },
  {
    "ph": "f", "id": 214303, "pid": 76337, "tid": -914061504, "ts": 1716454224744233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744234, "dur": 0,
    "args": {
      "External id": 214304,
      "cbid": 205, "correlation": 214304
    }
  },
  {
    "ph": "f", "id": 214304, "pid": 76337, "tid": -914061504, "ts": 1716454224744234,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224812923, "dur": 22,
    "args": {
      "External id": 214308,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214308,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214308, "pid": 5, "tid": 7, "ts": 1716454224812923,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744249, "dur": 12,
    "args": {
      "External id": 214308,
      "cbid": 211, "correlation": 214308
    }
  },
  {
    "ph": "s", "id": 214308, "pid": 76337, "tid": -914061504, "ts": 1716454224744249,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224812946, "dur": 45,
    "args": {
      "External id": 214310,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214310,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214310, "pid": 5, "tid": 7, "ts": 1716454224812946,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744264, "dur": 5,
    "args": {
      "External id": 214310,
      "cbid": 211, "correlation": 214310
    }
  },
  {
    "ph": "s", "id": 214310, "pid": 76337, "tid": -914061504, "ts": 1716454224744264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224812992, "dur": 650,
    "args": {
      "External id": 214312,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214312,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214312, "pid": 5, "tid": 7, "ts": 1716454224812992,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744275, "dur": 6,
    "args": {
      "External id": 214312,
      "cbid": 211, "correlation": 214312
    }
  },
  {
    "ph": "s", "id": 214312, "pid": 76337, "tid": -914061504, "ts": 1716454224744275,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224813643, "dur": 22,
    "args": {
      "External id": 214314,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214314,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214314, "pid": 5, "tid": 7, "ts": 1716454224813643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744285, "dur": 5,
    "args": {
      "External id": 214314,
      "cbid": 211, "correlation": 214314
    }
  },
  {
    "ph": "s", "id": 214314, "pid": 76337, "tid": -914061504, "ts": 1716454224744285,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224813666, "dur": 33,
    "args": {
      "External id": 214320,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214320,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214320, "pid": 5, "tid": 7, "ts": 1716454224813666,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744314, "dur": 8,
    "args": {
      "External id": 214320,
      "cbid": 211, "correlation": 214320
    }
  },
  {
    "ph": "s", "id": 214320, "pid": 76337, "tid": -914061504, "ts": 1716454224744314,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744372, "dur": 0,
    "args": {
      "External id": 214330,
      "cbid": 317, "correlation": 214330
    }
  },
  {
    "ph": "f", "id": 214330, "pid": 76337, "tid": -914061504, "ts": 1716454224744372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744372, "dur": 0,
    "args": {
      "External id": 214331,
      "cbid": 203, "correlation": 214331
    }
  },
  {
    "ph": "f", "id": 214331, "pid": 76337, "tid": -914061504, "ts": 1716454224744372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744373, "dur": 0,
    "args": {
      "External id": 214332,
      "cbid": 205, "correlation": 214332
    }
  },
  {
    "ph": "f", "id": 214332, "pid": 76337, "tid": -914061504, "ts": 1716454224744373,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224813700, "dur": 57,
    "args": {
      "External id": 214336,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214336,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 144,
      "warps per SM": 1152,
      "grid": [24, 60, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214336, "pid": 5, "tid": 7, "ts": 1716454224813700,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744385, "dur": 12,
    "args": {
      "External id": 214336,
      "cbid": 211, "correlation": 214336
    }
  },
  {
    "ph": "s", "id": 214336, "pid": 76337, "tid": -914061504, "ts": 1716454224744385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224813758, "dur": 272,
    "args": {
      "External id": 214338,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214338,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214338, "pid": 5, "tid": 7, "ts": 1716454224813758,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744403, "dur": 7,
    "args": {
      "External id": 214338,
      "cbid": 211, "correlation": 214338
    }
  },
  {
    "ph": "s", "id": 214338, "pid": 76337, "tid": -914061504, "ts": 1716454224744403,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224814031, "dur": 21,
    "args": {
      "External id": 214340,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214340,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214340, "pid": 5, "tid": 7, "ts": 1716454224814031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744414, "dur": 5,
    "args": {
      "External id": 214340,
      "cbid": 211, "correlation": 214340
    }
  },
  {
    "ph": "s", "id": 214340, "pid": 76337, "tid": -914061504, "ts": 1716454224744414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224814054, "dur": 32,
    "args": {
      "External id": 214346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214346,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214346, "pid": 5, "tid": 7, "ts": 1716454224814054,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744439, "dur": 9,
    "args": {
      "External id": 214346,
      "cbid": 211, "correlation": 214346
    }
  },
  {
    "ph": "s", "id": 214346, "pid": 76337, "tid": -914061504, "ts": 1716454224744439,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224814088, "dur": 27,
    "args": {
      "External id": 214354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214354,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214354, "pid": 5, "tid": 7, "ts": 1716454224814088,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744468, "dur": 8,
    "args": {
      "External id": 214354,
      "cbid": 211, "correlation": 214354
    }
  },
  {
    "ph": "s", "id": 214354, "pid": 76337, "tid": -914061504, "ts": 1716454224744468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224814116, "dur": 20,
    "args": {
      "External id": 214362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214362,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214362, "pid": 5, "tid": 7, "ts": 1716454224814116,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744497, "dur": 8,
    "args": {
      "External id": 214362,
      "cbid": 211, "correlation": 214362
    }
  },
  {
    "ph": "s", "id": 214362, "pid": 76337, "tid": -914061504, "ts": 1716454224744497,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224814137, "dur": 31,
    "args": {
      "External id": 214382,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214382,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 214382, "pid": 5, "tid": 7, "ts": 1716454224814137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744578, "dur": 12,
    "args": {
      "External id": 214382,
      "cbid": 211, "correlation": 214382
    }
  },
  {
    "ph": "s", "id": 214382, "pid": 76337, "tid": -914061504, "ts": 1716454224744578,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224814169, "dur": 4,
    "args": {
      "External id": 214394,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214394,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 214394, "pid": 5, "tid": 7, "ts": 1716454224814169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744601, "dur": 6,
    "args": {
      "External id": 214394,
      "cbid": 211, "correlation": 214394
    }
  },
  {
    "ph": "s", "id": 214394, "pid": 76337, "tid": -914061504, "ts": 1716454224744601,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224814175, "dur": 31,
    "args": {
      "External id": 214397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214397,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214397, "pid": 5, "tid": 7, "ts": 1716454224814175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744618, "dur": 6,
    "args": {
      "External id": 214397,
      "cbid": 211, "correlation": 214397
    }
  },
  {
    "ph": "s", "id": 214397, "pid": 76337, "tid": -914061504, "ts": 1716454224744618,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744675, "dur": 0,
    "args": {
      "External id": 214408,
      "cbid": 317, "correlation": 214408
    }
  },
  {
    "ph": "f", "id": 214408, "pid": 76337, "tid": -914061504, "ts": 1716454224744675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744675, "dur": 0,
    "args": {
      "External id": 214409,
      "cbid": 203, "correlation": 214409
    }
  },
  {
    "ph": "f", "id": 214409, "pid": 76337, "tid": -914061504, "ts": 1716454224744675,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744676, "dur": 0,
    "args": {
      "External id": 214410,
      "cbid": 205, "correlation": 214410
    }
  },
  {
    "ph": "f", "id": 214410, "pid": 76337, "tid": -914061504, "ts": 1716454224744676,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224814207, "dur": 21,
    "args": {
      "External id": 214414,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214414,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214414, "pid": 5, "tid": 7, "ts": 1716454224814207,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744690, "dur": 13,
    "args": {
      "External id": 214414,
      "cbid": 211, "correlation": 214414
    }
  },
  {
    "ph": "s", "id": 214414, "pid": 76337, "tid": -914061504, "ts": 1716454224744690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224814229, "dur": 106,
    "args": {
      "External id": 214416,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214416,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214416, "pid": 5, "tid": 7, "ts": 1716454224814229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744709, "dur": 6,
    "args": {
      "External id": 214416,
      "cbid": 211, "correlation": 214416
    }
  },
  {
    "ph": "s", "id": 214416, "pid": 76337, "tid": -914061504, "ts": 1716454224744709,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224814337, "dur": 23,
    "args": {
      "External id": 214418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214418,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214418, "pid": 5, "tid": 7, "ts": 1716454224814337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744719, "dur": 5,
    "args": {
      "External id": 214418,
      "cbid": 211, "correlation": 214418
    }
  },
  {
    "ph": "s", "id": 214418, "pid": 76337, "tid": -914061504, "ts": 1716454224744719,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224814360, "dur": 32,
    "args": {
      "External id": 214424,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214424,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214424, "pid": 5, "tid": 7, "ts": 1716454224814360,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744747, "dur": 9,
    "args": {
      "External id": 214424,
      "cbid": 211, "correlation": 214424
    }
  },
  {
    "ph": "s", "id": 214424, "pid": 76337, "tid": -914061504, "ts": 1716454224744747,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224814394, "dur": 178,
    "args": {
      "External id": 214433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214433,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214433, "pid": 5, "tid": 7, "ts": 1716454224814394,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744828, "dur": 14,
    "args": {
      "External id": 214433,
      "cbid": 211, "correlation": 214433
    }
  },
  {
    "ph": "s", "id": 214433, "pid": 76337, "tid": -914061504, "ts": 1716454224744828,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224814574, "dur": 66,
    "args": {
      "External id": 214455,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214455,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214455, "pid": 5, "tid": 7, "ts": 1716454224814574,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744885, "dur": 10,
    "args": {
      "External id": 214455,
      "cbid": 211, "correlation": 214455
    }
  },
  {
    "ph": "s", "id": 214455, "pid": 76337, "tid": -914061504, "ts": 1716454224744885,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744984, "dur": 2,
    "args": {
      "External id": 214466,
      "cbid": 251, "correlation": 214466
    }
  },
  {
    "ph": "f", "id": 214466, "pid": 76337, "tid": -914061504, "ts": 1716454224744984,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224814641, "dur": 155,
    "args": {
      "External id": 214467,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214467,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214467, "pid": 5, "tid": 7, "ts": 1716454224814641,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224744991, "dur": 15,
    "args": {
      "External id": 214467,
      "cbid": 211, "correlation": 214467
    }
  },
  {
    "ph": "s", "id": 214467, "pid": 76337, "tid": -914061504, "ts": 1716454224744991,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745064, "dur": 1,
    "args": {
      "External id": 214478,
      "cbid": 251, "correlation": 214478
    }
  },
  {
    "ph": "f", "id": 214478, "pid": 76337, "tid": -914061504, "ts": 1716454224745064,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224814797, "dur": 149,
    "args": {
      "External id": 214479,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214479,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214479, "pid": 5, "tid": 7, "ts": 1716454224814797,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745068, "dur": 11,
    "args": {
      "External id": 214479,
      "cbid": 211, "correlation": 214479
    }
  },
  {
    "ph": "s", "id": 214479, "pid": 76337, "tid": -914061504, "ts": 1716454224745068,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745132, "dur": 1,
    "args": {
      "External id": 214490,
      "cbid": 251, "correlation": 214490
    }
  },
  {
    "ph": "f", "id": 214490, "pid": 76337, "tid": -914061504, "ts": 1716454224745132,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224814948, "dur": 143,
    "args": {
      "External id": 214491,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214491,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214491, "pid": 5, "tid": 7, "ts": 1716454224814948,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745136, "dur": 11,
    "args": {
      "External id": 214491,
      "cbid": 211, "correlation": 214491
    }
  },
  {
    "ph": "s", "id": 214491, "pid": 76337, "tid": -914061504, "ts": 1716454224745136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224815092, "dur": 1974,
    "args": {
      "External id": 214512,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214512,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 214512, "pid": 5, "tid": 7, "ts": 1716454224815092,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745220, "dur": 13,
    "args": {
      "External id": 214512,
      "cbid": 211, "correlation": 214512
    }
  },
  {
    "ph": "s", "id": 214512, "pid": 76337, "tid": -914061504, "ts": 1716454224745220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745319, "dur": 1,
    "args": {
      "External id": 214530,
      "cbid": 251, "correlation": 214530
    }
  },
  {
    "ph": "f", "id": 214530, "pid": 76337, "tid": -914061504, "ts": 1716454224745319,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224817068, "dur": 148,
    "args": {
      "External id": 214532,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214532,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 214532, "pid": 5, "tid": 7, "ts": 1716454224817068,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745325, "dur": 13,
    "args": {
      "External id": 214532,
      "cbid": 211, "correlation": 214532
    }
  },
  {
    "ph": "s", "id": 214532, "pid": 76337, "tid": -914061504, "ts": 1716454224745325,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224817217, "dur": 35,
    "args": {
      "External id": 214540,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214540,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214540, "pid": 5, "tid": 7, "ts": 1716454224817217,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745395, "dur": 13,
    "args": {
      "External id": 214540,
      "cbid": 211, "correlation": 214540
    }
  },
  {
    "ph": "s", "id": 214540, "pid": 76337, "tid": -914061504, "ts": 1716454224745395,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224817254, "dur": 50,
    "args": {
      "External id": 214548,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214548,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214548, "pid": 5, "tid": 7, "ts": 1716454224817254,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745434, "dur": 8,
    "args": {
      "External id": 214548,
      "cbid": 211, "correlation": 214548
    }
  },
  {
    "ph": "s", "id": 214548, "pid": 76337, "tid": -914061504, "ts": 1716454224745434,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224817305, "dur": 30,
    "args": {
      "External id": 214559,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214559,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214559, "pid": 5, "tid": 7, "ts": 1716454224817305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745508, "dur": 13,
    "args": {
      "External id": 214559,
      "cbid": 211, "correlation": 214559
    }
  },
  {
    "ph": "s", "id": 214559, "pid": 76337, "tid": -914061504, "ts": 1716454224745508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224817337, "dur": 34,
    "args": {
      "External id": 214581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214581,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214581, "pid": 5, "tid": 7, "ts": 1716454224817337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745539, "dur": 7,
    "args": {
      "External id": 214581,
      "cbid": 211, "correlation": 214581
    }
  },
  {
    "ph": "s", "id": 214581, "pid": 76337, "tid": -914061504, "ts": 1716454224745539,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745623, "dur": 1,
    "args": {
      "External id": 214592,
      "cbid": 251, "correlation": 214592
    }
  },
  {
    "ph": "f", "id": 214592, "pid": 76337, "tid": -914061504, "ts": 1716454224745623,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224817372, "dur": 91,
    "args": {
      "External id": 214593,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214593,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214593, "pid": 5, "tid": 7, "ts": 1716454224817372,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745628, "dur": 13,
    "args": {
      "External id": 214593,
      "cbid": 211, "correlation": 214593
    }
  },
  {
    "ph": "s", "id": 214593, "pid": 76337, "tid": -914061504, "ts": 1716454224745628,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745698, "dur": 1,
    "args": {
      "External id": 214604,
      "cbid": 251, "correlation": 214604
    }
  },
  {
    "ph": "f", "id": 214604, "pid": 76337, "tid": -914061504, "ts": 1716454224745698,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745702, "dur": 0,
    "args": {
      "External id": 214605,
      "cbid": 251, "correlation": 214605
    }
  },
  {
    "ph": "f", "id": 214605, "pid": 76337, "tid": -914061504, "ts": 1716454224745702,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224817465, "dur": 11,
    "args": {
      "External id": 214606,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214606,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 214606, "pid": 5, "tid": 7, "ts": 1716454224817465,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745703, "dur": 11,
    "args": {
      "External id": 214606,
      "cbid": 211, "correlation": 214606
    }
  },
  {
    "ph": "s", "id": 214606, "pid": 76337, "tid": -914061504, "ts": 1716454224745703,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224817477, "dur": 5,
    "args": {
      "External id": 214608,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214608,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 214608, "pid": 5, "tid": 7, "ts": 1716454224817477,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745717, "dur": 6,
    "args": {
      "External id": 214608,
      "cbid": 211, "correlation": 214608
    }
  },
  {
    "ph": "s", "id": 214608, "pid": 76337, "tid": -914061504, "ts": 1716454224745717,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745774, "dur": 1,
    "args": {
      "External id": 214619,
      "cbid": 251, "correlation": 214619
    }
  },
  {
    "ph": "f", "id": 214619, "pid": 76337, "tid": -914061504, "ts": 1716454224745774,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745777, "dur": 0,
    "args": {
      "External id": 214620,
      "cbid": 251, "correlation": 214620
    }
  },
  {
    "ph": "f", "id": 214620, "pid": 76337, "tid": -914061504, "ts": 1716454224745777,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224817484, "dur": 7,
    "args": {
      "External id": 214621,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214621,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 214621, "pid": 5, "tid": 7, "ts": 1716454224817484,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745778, "dur": 12,
    "args": {
      "External id": 214621,
      "cbid": 211, "correlation": 214621
    }
  },
  {
    "ph": "s", "id": 214621, "pid": 76337, "tid": -914061504, "ts": 1716454224745778,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224817492, "dur": 4,
    "args": {
      "External id": 214623,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214623,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 214623, "pid": 5, "tid": 7, "ts": 1716454224817492,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745791, "dur": 6,
    "args": {
      "External id": 214623,
      "cbid": 211, "correlation": 214623
    }
  },
  {
    "ph": "s", "id": 214623, "pid": 76337, "tid": -914061504, "ts": 1716454224745791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224817497, "dur": 93,
    "args": {
      "External id": 214644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214644,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 214644, "pid": 5, "tid": 7, "ts": 1716454224817497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745865, "dur": 12,
    "args": {
      "External id": 214644,
      "cbid": 211, "correlation": 214644
    }
  },
  {
    "ph": "s", "id": 214644, "pid": 76337, "tid": -914061504, "ts": 1716454224745865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745961, "dur": 1,
    "args": {
      "External id": 214662,
      "cbid": 251, "correlation": 214662
    }
  },
  {
    "ph": "f", "id": 214662, "pid": 76337, "tid": -914061504, "ts": 1716454224745961,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224817591, "dur": 98,
    "args": {
      "External id": 214664,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214664,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214664, "pid": 5, "tid": 7, "ts": 1716454224817591,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224745967, "dur": 21,
    "args": {
      "External id": 214664,
      "cbid": 211, "correlation": 214664
    }
  },
  {
    "ph": "s", "id": 214664, "pid": 76337, "tid": -914061504, "ts": 1716454224745967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224817691, "dur": 19,
    "args": {
      "External id": 214672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214672,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214672, "pid": 5, "tid": 7, "ts": 1716454224817691,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746044, "dur": 13,
    "args": {
      "External id": 214672,
      "cbid": 211, "correlation": 214672
    }
  },
  {
    "ph": "s", "id": 214672, "pid": 76337, "tid": -914061504, "ts": 1716454224746044,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224817711, "dur": 39,
    "args": {
      "External id": 214680,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214680,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214680, "pid": 5, "tid": 7, "ts": 1716454224817711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746086, "dur": 9,
    "args": {
      "External id": 214680,
      "cbid": 211, "correlation": 214680
    }
  },
  {
    "ph": "s", "id": 214680, "pid": 76337, "tid": -914061504, "ts": 1716454224746086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224817751, "dur": 35,
    "args": {
      "External id": 214702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214702,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214702, "pid": 5, "tid": 7, "ts": 1716454224817751,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746135, "dur": 10,
    "args": {
      "External id": 214702,
      "cbid": 211, "correlation": 214702
    }
  },
  {
    "ph": "s", "id": 214702, "pid": 76337, "tid": -914061504, "ts": 1716454224746135,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746225, "dur": 1,
    "args": {
      "External id": 214718,
      "cbid": 251, "correlation": 214718
    }
  },
  {
    "ph": "f", "id": 214718, "pid": 76337, "tid": -914061504, "ts": 1716454224746225,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746230, "dur": 0,
    "args": {
      "External id": 214720,
      "cbid": 251, "correlation": 214720
    }
  },
  {
    "ph": "f", "id": 214720, "pid": 76337, "tid": -914061504, "ts": 1716454224746230,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224817787, "dur": 545,
    "args": {
      "External id": 214721,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214721,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 214721, "pid": 5, "tid": 7, "ts": 1716454224817787,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746234, "dur": 13,
    "args": {
      "External id": 214721,
      "cbid": 211, "correlation": 214721
    }
  },
  {
    "ph": "s", "id": 214721, "pid": 76337, "tid": -914061504, "ts": 1716454224746234,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224818333, "dur": 127,
    "args": {
      "External id": 214729,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214729,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214729, "pid": 5, "tid": 7, "ts": 1716454224818333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746301, "dur": 12,
    "args": {
      "External id": 214729,
      "cbid": 211, "correlation": 214729
    }
  },
  {
    "ph": "s", "id": 214729, "pid": 76337, "tid": -914061504, "ts": 1716454224746301,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224818462, "dur": 130,
    "args": {
      "External id": 214737,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214737,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214737, "pid": 5, "tid": 7, "ts": 1716454224818462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746332, "dur": 8,
    "args": {
      "External id": 214737,
      "cbid": 211, "correlation": 214737
    }
  },
  {
    "ph": "s", "id": 214737, "pid": 76337, "tid": -914061504, "ts": 1716454224746332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746408, "dur": 1,
    "args": {
      "External id": 214753,
      "cbid": 251, "correlation": 214753
    }
  },
  {
    "ph": "f", "id": 214753, "pid": 76337, "tid": -914061504, "ts": 1716454224746408,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224818593, "dur": 309,
    "args": {
      "External id": 214755,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214755,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214755, "pid": 5, "tid": 7, "ts": 1716454224818593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746414, "dur": 12,
    "args": {
      "External id": 214755,
      "cbid": 211, "correlation": 214755
    }
  },
  {
    "ph": "s", "id": 214755, "pid": 76337, "tid": -914061504, "ts": 1716454224746414,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224818904, "dur": 27,
    "args": {
      "External id": 214763,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214763,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214763, "pid": 5, "tid": 7, "ts": 1716454224818904,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746456, "dur": 10,
    "args": {
      "External id": 214763,
      "cbid": 211, "correlation": 214763
    }
  },
  {
    "ph": "s", "id": 214763, "pid": 76337, "tid": -914061504, "ts": 1716454224746456,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224818932, "dur": 82,
    "args": {
      "External id": 214774,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214774,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214774, "pid": 5, "tid": 7, "ts": 1716454224818932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746523, "dur": 12,
    "args": {
      "External id": 214774,
      "cbid": 211, "correlation": 214774
    }
  },
  {
    "ph": "s", "id": 214774, "pid": 76337, "tid": -914061504, "ts": 1716454224746523,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746585, "dur": 0,
    "args": {
      "External id": 214786,
      "cbid": 317, "correlation": 214786
    }
  },
  {
    "ph": "f", "id": 214786, "pid": 76337, "tid": -914061504, "ts": 1716454224746585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746585, "dur": 0,
    "args": {
      "External id": 214787,
      "cbid": 203, "correlation": 214787
    }
  },
  {
    "ph": "f", "id": 214787, "pid": 76337, "tid": -914061504, "ts": 1716454224746585,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746586, "dur": 0,
    "args": {
      "External id": 214788,
      "cbid": 205, "correlation": 214788
    }
  },
  {
    "ph": "f", "id": 214788, "pid": 76337, "tid": -914061504, "ts": 1716454224746586,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224819015, "dur": 23,
    "args": {
      "External id": 214792,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214792,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214792, "pid": 5, "tid": 7, "ts": 1716454224819015,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746602, "dur": 12,
    "args": {
      "External id": 214792,
      "cbid": 211, "correlation": 214792
    }
  },
  {
    "ph": "s", "id": 214792, "pid": 76337, "tid": -914061504, "ts": 1716454224746602,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224819039, "dur": 122,
    "args": {
      "External id": 214794,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214794,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214794, "pid": 5, "tid": 7, "ts": 1716454224819039,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746621, "dur": 6,
    "args": {
      "External id": 214794,
      "cbid": 211, "correlation": 214794
    }
  },
  {
    "ph": "s", "id": 214794, "pid": 76337, "tid": -914061504, "ts": 1716454224746621,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224819162, "dur": 24,
    "args": {
      "External id": 214796,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214796,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214796, "pid": 5, "tid": 7, "ts": 1716454224819162,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746631, "dur": 5,
    "args": {
      "External id": 214796,
      "cbid": 211, "correlation": 214796
    }
  },
  {
    "ph": "s", "id": 214796, "pid": 76337, "tid": -914061504, "ts": 1716454224746631,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224819187, "dur": 32,
    "args": {
      "External id": 214802,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214802,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214802, "pid": 5, "tid": 7, "ts": 1716454224819187,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746659, "dur": 8,
    "args": {
      "External id": 214802,
      "cbid": 211, "correlation": 214802
    }
  },
  {
    "ph": "s", "id": 214802, "pid": 76337, "tid": -914061504, "ts": 1716454224746659,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224819221, "dur": 27,
    "args": {
      "External id": 214810,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214810,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214810, "pid": 5, "tid": 7, "ts": 1716454224819221,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746690, "dur": 9,
    "args": {
      "External id": 214810,
      "cbid": 211, "correlation": 214810
    }
  },
  {
    "ph": "s", "id": 214810, "pid": 76337, "tid": -914061504, "ts": 1716454224746690,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224819249, "dur": 55,
    "args": {
      "External id": 214819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214819,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214819, "pid": 5, "tid": 7, "ts": 1716454224819249,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746729, "dur": 10,
    "args": {
      "External id": 214819,
      "cbid": 211, "correlation": 214819
    }
  },
  {
    "ph": "s", "id": 214819, "pid": 76337, "tid": -914061504, "ts": 1716454224746729,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224819305, "dur": 52,
    "args": {
      "External id": 214839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214839,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 214839, "pid": 5, "tid": 7, "ts": 1716454224819305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746799, "dur": 12,
    "args": {
      "External id": 214839,
      "cbid": 211, "correlation": 214839
    }
  },
  {
    "ph": "s", "id": 214839, "pid": 76337, "tid": -914061504, "ts": 1716454224746799,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224819358, "dur": 4,
    "args": {
      "External id": 214851,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214851,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.5,
      "warps per SM": 4,
      "grid": [40, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 214851, "pid": 5, "tid": 7, "ts": 1716454224819358,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746820, "dur": 6,
    "args": {
      "External id": 214851,
      "cbid": 211, "correlation": 214851
    }
  },
  {
    "ph": "s", "id": 214851, "pid": 76337, "tid": -914061504, "ts": 1716454224746820,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224819364, "dur": 58,
    "args": {
      "External id": 214854,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214854,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214854, "pid": 5, "tid": 7, "ts": 1716454224819364,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746838, "dur": 6,
    "args": {
      "External id": 214854,
      "cbid": 211, "correlation": 214854
    }
  },
  {
    "ph": "s", "id": 214854, "pid": 76337, "tid": -914061504, "ts": 1716454224746838,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224819423, "dur": 37,
    "args": {
      "External id": 214863,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214863,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214863, "pid": 5, "tid": 7, "ts": 1716454224819423,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746878, "dur": 10,
    "args": {
      "External id": 214863,
      "cbid": 211, "correlation": 214863
    }
  },
  {
    "ph": "s", "id": 214863, "pid": 76337, "tid": -914061504, "ts": 1716454224746878,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746931, "dur": 0,
    "args": {
      "External id": 214873,
      "cbid": 317, "correlation": 214873
    }
  },
  {
    "ph": "f", "id": 214873, "pid": 76337, "tid": -914061504, "ts": 1716454224746931,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746932, "dur": 0,
    "args": {
      "External id": 214874,
      "cbid": 203, "correlation": 214874
    }
  },
  {
    "ph": "f", "id": 214874, "pid": 76337, "tid": -914061504, "ts": 1716454224746932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746932, "dur": 0,
    "args": {
      "External id": 214875,
      "cbid": 205, "correlation": 214875
    }
  },
  {
    "ph": "f", "id": 214875, "pid": 76337, "tid": -914061504, "ts": 1716454224746932,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224819462, "dur": 39,
    "args": {
      "External id": 214879,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214879,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214879, "pid": 5, "tid": 7, "ts": 1716454224819462,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746947, "dur": 12,
    "args": {
      "External id": 214879,
      "cbid": 211, "correlation": 214879
    }
  },
  {
    "ph": "s", "id": 214879, "pid": 76337, "tid": -914061504, "ts": 1716454224746947,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224819502, "dur": 84,
    "args": {
      "External id": 214881,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214881,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 320,
      "warps per SM": 2560,
      "grid": [1, 40, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214881, "pid": 5, "tid": 7, "ts": 1716454224819502,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746962, "dur": 5,
    "args": {
      "External id": 214881,
      "cbid": 211, "correlation": 214881
    }
  },
  {
    "ph": "s", "id": 214881, "pid": 76337, "tid": -914061504, "ts": 1716454224746962,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224819587, "dur": 1288,
    "args": {
      "External id": 214883,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214883,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214883, "pid": 5, "tid": 7, "ts": 1716454224819587,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746973, "dur": 21,
    "args": {
      "External id": 214883,
      "cbid": 211, "correlation": 214883
    }
  },
  {
    "ph": "s", "id": 214883, "pid": 76337, "tid": -914061504, "ts": 1716454224746973,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224820876, "dur": 21,
    "args": {
      "External id": 214885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214885,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214885, "pid": 5, "tid": 7, "ts": 1716454224820876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224746997, "dur": 6,
    "args": {
      "External id": 214885,
      "cbid": 211, "correlation": 214885
    }
  },
  {
    "ph": "s", "id": 214885, "pid": 76337, "tid": -914061504, "ts": 1716454224746997,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224820899, "dur": 34,
    "args": {
      "External id": 214891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214891, "pid": 5, "tid": 7, "ts": 1716454224820899,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747027, "dur": 8,
    "args": {
      "External id": 214891,
      "cbid": 211, "correlation": 214891
    }
  },
  {
    "ph": "s", "id": 214891, "pid": 76337, "tid": -914061504, "ts": 1716454224747027,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224820934, "dur": 3,
    "args": {
      "External id": 214899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214899,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 214899, "pid": 5, "tid": 7, "ts": 1716454224820934,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747072, "dur": 9,
    "args": {
      "External id": 214899,
      "cbid": 211, "correlation": 214899
    }
  },
  {
    "ph": "s", "id": 214899, "pid": 76337, "tid": -914061504, "ts": 1716454224747072,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747137, "dur": 1,
    "args": {
      "External id": 214915,
      "cbid": 251, "correlation": 214915
    }
  },
  {
    "ph": "f", "id": 214915, "pid": 76337, "tid": -914061504, "ts": 1716454224747137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747141, "dur": 0,
    "args": {
      "External id": 214917,
      "cbid": 251, "correlation": 214917
    }
  },
  {
    "ph": "f", "id": 214917, "pid": 76337, "tid": -914061504, "ts": 1716454224747141,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224820939, "dur": 12,
    "args": {
      "External id": 214918,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214918,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 214918, "pid": 5, "tid": 7, "ts": 1716454224820939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747143, "dur": 12,
    "args": {
      "External id": 214918,
      "cbid": 211, "correlation": 214918
    }
  },
  {
    "ph": "s", "id": 214918, "pid": 76337, "tid": -914061504, "ts": 1716454224747143,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224820952, "dur": 5,
    "args": {
      "External id": 214920,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214920,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 214920, "pid": 5, "tid": 7, "ts": 1716454224820952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747157, "dur": 5,
    "args": {
      "External id": 214920,
      "cbid": 211, "correlation": 214920
    }
  },
  {
    "ph": "s", "id": 214920, "pid": 76337, "tid": -914061504, "ts": 1716454224747157,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224820959, "dur": 29,
    "args": {
      "External id": 214930,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214930,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214930, "pid": 5, "tid": 7, "ts": 1716454224820959,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747216, "dur": 12,
    "args": {
      "External id": 214930,
      "cbid": 211, "correlation": 214930
    }
  },
  {
    "ph": "s", "id": 214930, "pid": 76337, "tid": -914061504, "ts": 1716454224747216,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224820990, "dur": 32,
    "args": {
      "External id": 214950,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214950,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 214950, "pid": 5, "tid": 7, "ts": 1716454224820990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747282, "dur": 11,
    "args": {
      "External id": 214950,
      "cbid": 211, "correlation": 214950
    }
  },
  {
    "ph": "s", "id": 214950, "pid": 76337, "tid": -914061504, "ts": 1716454224747282,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224821023, "dur": 4,
    "args": {
      "External id": 214962,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214962,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 214962, "pid": 5, "tid": 7, "ts": 1716454224821023,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747302, "dur": 7,
    "args": {
      "External id": 214962,
      "cbid": 211, "correlation": 214962
    }
  },
  {
    "ph": "s", "id": 214962, "pid": 76337, "tid": -914061504, "ts": 1716454224747302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224821029, "dur": 30,
    "args": {
      "External id": 214965,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214965,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214965, "pid": 5, "tid": 7, "ts": 1716454224821029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747322, "dur": 8,
    "args": {
      "External id": 214965,
      "cbid": 211, "correlation": 214965
    }
  },
  {
    "ph": "s", "id": 214965, "pid": 76337, "tid": -914061504, "ts": 1716454224747322,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224821060, "dur": 20,
    "args": {
      "External id": 214974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214974,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214974, "pid": 5, "tid": 7, "ts": 1716454224821060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747363, "dur": 9,
    "args": {
      "External id": 214974,
      "cbid": 211, "correlation": 214974
    }
  },
  {
    "ph": "s", "id": 214974, "pid": 76337, "tid": -914061504, "ts": 1716454224747363,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747425, "dur": 0,
    "args": {
      "External id": 214984,
      "cbid": 317, "correlation": 214984
    }
  },
  {
    "ph": "f", "id": 214984, "pid": 76337, "tid": -914061504, "ts": 1716454224747425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747426, "dur": 0,
    "args": {
      "External id": 214985,
      "cbid": 203, "correlation": 214985
    }
  },
  {
    "ph": "f", "id": 214985, "pid": 76337, "tid": -914061504, "ts": 1716454224747426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747426, "dur": 0,
    "args": {
      "External id": 214986,
      "cbid": 205, "correlation": 214986
    }
  },
  {
    "ph": "f", "id": 214986, "pid": 76337, "tid": -914061504, "ts": 1716454224747426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224821081, "dur": 23,
    "args": {
      "External id": 214990,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214990,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214990, "pid": 5, "tid": 7, "ts": 1716454224821081,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747440, "dur": 12,
    "args": {
      "External id": 214990,
      "cbid": 211, "correlation": 214990
    }
  },
  {
    "ph": "s", "id": 214990, "pid": 76337, "tid": -914061504, "ts": 1716454224747440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224821105, "dur": 45,
    "args": {
      "External id": 214992,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214992,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214992, "pid": 5, "tid": 7, "ts": 1716454224821105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747455, "dur": 5,
    "args": {
      "External id": 214992,
      "cbid": 211, "correlation": 214992
    }
  },
  {
    "ph": "s", "id": 214992, "pid": 76337, "tid": -914061504, "ts": 1716454224747455,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224821152, "dur": 650,
    "args": {
      "External id": 214994,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214994,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 214994, "pid": 5, "tid": 7, "ts": 1716454224821152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747466, "dur": 6,
    "args": {
      "External id": 214994,
      "cbid": 211, "correlation": 214994
    }
  },
  {
    "ph": "s", "id": 214994, "pid": 76337, "tid": -914061504, "ts": 1716454224747466,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224821803, "dur": 23,
    "args": {
      "External id": 214996,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 214996,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 214996, "pid": 5, "tid": 7, "ts": 1716454224821803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747476, "dur": 5,
    "args": {
      "External id": 214996,
      "cbid": 211, "correlation": 214996
    }
  },
  {
    "ph": "s", "id": 214996, "pid": 76337, "tid": -914061504, "ts": 1716454224747476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224821827, "dur": 34,
    "args": {
      "External id": 215002,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215002,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215002, "pid": 5, "tid": 7, "ts": 1716454224821827,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747503, "dur": 9,
    "args": {
      "External id": 215002,
      "cbid": 211, "correlation": 215002
    }
  },
  {
    "ph": "s", "id": 215002, "pid": 76337, "tid": -914061504, "ts": 1716454224747503,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747564, "dur": 0,
    "args": {
      "External id": 215012,
      "cbid": 317, "correlation": 215012
    }
  },
  {
    "ph": "f", "id": 215012, "pid": 76337, "tid": -914061504, "ts": 1716454224747564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747564, "dur": 0,
    "args": {
      "External id": 215013,
      "cbid": 203, "correlation": 215013
    }
  },
  {
    "ph": "f", "id": 215013, "pid": 76337, "tid": -914061504, "ts": 1716454224747564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747565, "dur": 0,
    "args": {
      "External id": 215014,
      "cbid": 205, "correlation": 215014
    }
  },
  {
    "ph": "f", "id": 215014, "pid": 76337, "tid": -914061504, "ts": 1716454224747565,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224821862, "dur": 39,
    "args": {
      "External id": 215018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215018,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [24, 40, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215018, "pid": 5, "tid": 7, "ts": 1716454224821862,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747577, "dur": 11,
    "args": {
      "External id": 215018,
      "cbid": 211, "correlation": 215018
    }
  },
  {
    "ph": "s", "id": 215018, "pid": 76337, "tid": -914061504, "ts": 1716454224747577,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224821902, "dur": 192,
    "args": {
      "External id": 215020,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215020,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215020, "pid": 5, "tid": 7, "ts": 1716454224821902,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747594, "dur": 7,
    "args": {
      "External id": 215020,
      "cbid": 211, "correlation": 215020
    }
  },
  {
    "ph": "s", "id": 215020, "pid": 76337, "tid": -914061504, "ts": 1716454224747594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224822096, "dur": 22,
    "args": {
      "External id": 215022,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215022,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215022, "pid": 5, "tid": 7, "ts": 1716454224822096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747605, "dur": 5,
    "args": {
      "External id": 215022,
      "cbid": 211, "correlation": 215022
    }
  },
  {
    "ph": "s", "id": 215022, "pid": 76337, "tid": -914061504, "ts": 1716454224747605,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224822119, "dur": 32,
    "args": {
      "External id": 215028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215028,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215028, "pid": 5, "tid": 7, "ts": 1716454224822119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747630, "dur": 8,
    "args": {
      "External id": 215028,
      "cbid": 211, "correlation": 215028
    }
  },
  {
    "ph": "s", "id": 215028, "pid": 76337, "tid": -914061504, "ts": 1716454224747630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224822152, "dur": 27,
    "args": {
      "External id": 215036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215036,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215036, "pid": 5, "tid": 7, "ts": 1716454224822152,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747660, "dur": 7,
    "args": {
      "External id": 215036,
      "cbid": 211, "correlation": 215036
    }
  },
  {
    "ph": "s", "id": 215036, "pid": 76337, "tid": -914061504, "ts": 1716454224747660,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224822181, "dur": 20,
    "args": {
      "External id": 215044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215044,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215044, "pid": 5, "tid": 7, "ts": 1716454224822181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747688, "dur": 8,
    "args": {
      "External id": 215044,
      "cbid": 211, "correlation": 215044
    }
  },
  {
    "ph": "s", "id": 215044, "pid": 76337, "tid": -914061504, "ts": 1716454224747688,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224822202, "dur": 30,
    "args": {
      "External id": 215064,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215064,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 215064, "pid": 5, "tid": 7, "ts": 1716454224822202,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747770, "dur": 12,
    "args": {
      "External id": 215064,
      "cbid": 211, "correlation": 215064
    }
  },
  {
    "ph": "s", "id": 215064, "pid": 76337, "tid": -914061504, "ts": 1716454224747770,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224822233, "dur": 4,
    "args": {
      "External id": 215076,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215076,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 215076, "pid": 5, "tid": 7, "ts": 1716454224822233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747791, "dur": 6,
    "args": {
      "External id": 215076,
      "cbid": 211, "correlation": 215076
    }
  },
  {
    "ph": "s", "id": 215076, "pid": 76337, "tid": -914061504, "ts": 1716454224747791,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224822239, "dur": 30,
    "args": {
      "External id": 215079,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215079,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215079, "pid": 5, "tid": 7, "ts": 1716454224822239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747810, "dur": 6,
    "args": {
      "External id": 215079,
      "cbid": 211, "correlation": 215079
    }
  },
  {
    "ph": "s", "id": 215079, "pid": 76337, "tid": -914061504, "ts": 1716454224747810,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747868, "dur": 0,
    "args": {
      "External id": 215090,
      "cbid": 317, "correlation": 215090
    }
  },
  {
    "ph": "f", "id": 215090, "pid": 76337, "tid": -914061504, "ts": 1716454224747868,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747869, "dur": 0,
    "args": {
      "External id": 215091,
      "cbid": 203, "correlation": 215091
    }
  },
  {
    "ph": "f", "id": 215091, "pid": 76337, "tid": -914061504, "ts": 1716454224747869,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747870, "dur": 0,
    "args": {
      "External id": 215092,
      "cbid": 205, "correlation": 215092
    }
  },
  {
    "ph": "f", "id": 215092, "pid": 76337, "tid": -914061504, "ts": 1716454224747870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224822271, "dur": 22,
    "args": {
      "External id": 215096,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215096,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215096, "pid": 5, "tid": 7, "ts": 1716454224822271,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747882, "dur": 12,
    "args": {
      "External id": 215096,
      "cbid": 211, "correlation": 215096
    }
  },
  {
    "ph": "s", "id": 215096, "pid": 76337, "tid": -914061504, "ts": 1716454224747882,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224822294, "dur": 106,
    "args": {
      "External id": 215098,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215098,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215098, "pid": 5, "tid": 7, "ts": 1716454224822294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747900, "dur": 7,
    "args": {
      "External id": 215098,
      "cbid": 211, "correlation": 215098
    }
  },
  {
    "ph": "s", "id": 215098, "pid": 76337, "tid": -914061504, "ts": 1716454224747900,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224822401, "dur": 22,
    "args": {
      "External id": 215100,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215100,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215100, "pid": 5, "tid": 7, "ts": 1716454224822401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747910, "dur": 5,
    "args": {
      "External id": 215100,
      "cbid": 211, "correlation": 215100
    }
  },
  {
    "ph": "s", "id": 215100, "pid": 76337, "tid": -914061504, "ts": 1716454224747910,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224822425, "dur": 31,
    "args": {
      "External id": 215106,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215106,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215106, "pid": 5, "tid": 7, "ts": 1716454224822425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224747937, "dur": 8,
    "args": {
      "External id": 215106,
      "cbid": 211, "correlation": 215106
    }
  },
  {
    "ph": "s", "id": 215106, "pid": 76337, "tid": -914061504, "ts": 1716454224747937,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224822457, "dur": 195,
    "args": {
      "External id": 215115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215115,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215115, "pid": 5, "tid": 7, "ts": 1716454224822457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748029, "dur": 14,
    "args": {
      "External id": 215115,
      "cbid": 211, "correlation": 215115
    }
  },
  {
    "ph": "s", "id": 215115, "pid": 76337, "tid": -914061504, "ts": 1716454224748029,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224822653, "dur": 65,
    "args": {
      "External id": 215137,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215137,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215137, "pid": 5, "tid": 7, "ts": 1716454224822653,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748085, "dur": 10,
    "args": {
      "External id": 215137,
      "cbid": 211, "correlation": 215137
    }
  },
  {
    "ph": "s", "id": 215137, "pid": 76337, "tid": -914061504, "ts": 1716454224748085,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748175, "dur": 1,
    "args": {
      "External id": 215148,
      "cbid": 251, "correlation": 215148
    }
  },
  {
    "ph": "f", "id": 215148, "pid": 76337, "tid": -914061504, "ts": 1716454224748175,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224822719, "dur": 152,
    "args": {
      "External id": 215149,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215149,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215149, "pid": 5, "tid": 7, "ts": 1716454224822719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748181, "dur": 13,
    "args": {
      "External id": 215149,
      "cbid": 211, "correlation": 215149
    }
  },
  {
    "ph": "s", "id": 215149, "pid": 76337, "tid": -914061504, "ts": 1716454224748181,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748250, "dur": 1,
    "args": {
      "External id": 215160,
      "cbid": 251, "correlation": 215160
    }
  },
  {
    "ph": "f", "id": 215160, "pid": 76337, "tid": -914061504, "ts": 1716454224748250,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224822872, "dur": 145,
    "args": {
      "External id": 215161,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215161,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215161, "pid": 5, "tid": 7, "ts": 1716454224822872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748254, "dur": 11,
    "args": {
      "External id": 215161,
      "cbid": 211, "correlation": 215161
    }
  },
  {
    "ph": "s", "id": 215161, "pid": 76337, "tid": -914061504, "ts": 1716454224748254,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748318, "dur": 1,
    "args": {
      "External id": 215172,
      "cbid": 251, "correlation": 215172
    }
  },
  {
    "ph": "f", "id": 215172, "pid": 76337, "tid": -914061504, "ts": 1716454224748318,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224823018, "dur": 147,
    "args": {
      "External id": 215173,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215173,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215173, "pid": 5, "tid": 7, "ts": 1716454224823018,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748321, "dur": 11,
    "args": {
      "External id": 215173,
      "cbid": 211, "correlation": 215173
    }
  },
  {
    "ph": "s", "id": 215173, "pid": 76337, "tid": -914061504, "ts": 1716454224748321,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224823167, "dur": 1965,
    "args": {
      "External id": 215194,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215194,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 215194, "pid": 5, "tid": 7, "ts": 1716454224823167,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748404, "dur": 13,
    "args": {
      "External id": 215194,
      "cbid": 211, "correlation": 215194
    }
  },
  {
    "ph": "s", "id": 215194, "pid": 76337, "tid": -914061504, "ts": 1716454224748404,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748499, "dur": 1,
    "args": {
      "External id": 215212,
      "cbid": 251, "correlation": 215212
    }
  },
  {
    "ph": "f", "id": 215212, "pid": 76337, "tid": -914061504, "ts": 1716454224748499,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224825133, "dur": 149,
    "args": {
      "External id": 215214,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215214,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 215214, "pid": 5, "tid": 7, "ts": 1716454224825133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748505, "dur": 13,
    "args": {
      "External id": 215214,
      "cbid": 211, "correlation": 215214
    }
  },
  {
    "ph": "s", "id": 215214, "pid": 76337, "tid": -914061504, "ts": 1716454224748505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224825283, "dur": 36,
    "args": {
      "External id": 215222,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215222,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215222, "pid": 5, "tid": 7, "ts": 1716454224825283,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748574, "dur": 12,
    "args": {
      "External id": 215222,
      "cbid": 211, "correlation": 215222
    }
  },
  {
    "ph": "s", "id": 215222, "pid": 76337, "tid": -914061504, "ts": 1716454224748574,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224825320, "dur": 50,
    "args": {
      "External id": 215230,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215230,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215230, "pid": 5, "tid": 7, "ts": 1716454224825320,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748613, "dur": 8,
    "args": {
      "External id": 215230,
      "cbid": 211, "correlation": 215230
    }
  },
  {
    "ph": "s", "id": 215230, "pid": 76337, "tid": -914061504, "ts": 1716454224748613,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224825371, "dur": 29,
    "args": {
      "External id": 215241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215241,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215241, "pid": 5, "tid": 7, "ts": 1716454224825371,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748684, "dur": 12,
    "args": {
      "External id": 215241,
      "cbid": 211, "correlation": 215241
    }
  },
  {
    "ph": "s", "id": 215241, "pid": 76337, "tid": -914061504, "ts": 1716454224748684,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224825402, "dur": 35,
    "args": {
      "External id": 215263,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215263,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215263, "pid": 5, "tid": 7, "ts": 1716454224825402,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748715, "dur": 7,
    "args": {
      "External id": 215263,
      "cbid": 211, "correlation": 215263
    }
  },
  {
    "ph": "s", "id": 215263, "pid": 76337, "tid": -914061504, "ts": 1716454224748715,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748799, "dur": 1,
    "args": {
      "External id": 215274,
      "cbid": 251, "correlation": 215274
    }
  },
  {
    "ph": "f", "id": 215274, "pid": 76337, "tid": -914061504, "ts": 1716454224748799,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224825438, "dur": 76,
    "args": {
      "External id": 215275,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215275,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215275, "pid": 5, "tid": 7, "ts": 1716454224825438,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748804, "dur": 14,
    "args": {
      "External id": 215275,
      "cbid": 211, "correlation": 215275
    }
  },
  {
    "ph": "s", "id": 215275, "pid": 76337, "tid": -914061504, "ts": 1716454224748804,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748872, "dur": 1,
    "args": {
      "External id": 215286,
      "cbid": 251, "correlation": 215286
    }
  },
  {
    "ph": "f", "id": 215286, "pid": 76337, "tid": -914061504, "ts": 1716454224748872,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748876, "dur": 0,
    "args": {
      "External id": 215287,
      "cbid": 251, "correlation": 215287
    }
  },
  {
    "ph": "f", "id": 215287, "pid": 76337, "tid": -914061504, "ts": 1716454224748876,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224825515, "dur": 11,
    "args": {
      "External id": 215288,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215288,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 215288, "pid": 5, "tid": 7, "ts": 1716454224825515,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748877, "dur": 12,
    "args": {
      "External id": 215288,
      "cbid": 211, "correlation": 215288
    }
  },
  {
    "ph": "s", "id": 215288, "pid": 76337, "tid": -914061504, "ts": 1716454224748877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224825528, "dur": 5,
    "args": {
      "External id": 215290,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215290,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 215290, "pid": 5, "tid": 7, "ts": 1716454224825528,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748891, "dur": 6,
    "args": {
      "External id": 215290,
      "cbid": 211, "correlation": 215290
    }
  },
  {
    "ph": "s", "id": 215290, "pid": 76337, "tid": -914061504, "ts": 1716454224748891,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748949, "dur": 1,
    "args": {
      "External id": 215301,
      "cbid": 251, "correlation": 215301
    }
  },
  {
    "ph": "f", "id": 215301, "pid": 76337, "tid": -914061504, "ts": 1716454224748949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748952, "dur": 0,
    "args": {
      "External id": 215302,
      "cbid": 251, "correlation": 215302
    }
  },
  {
    "ph": "f", "id": 215302, "pid": 76337, "tid": -914061504, "ts": 1716454224748952,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224825535, "dur": 7,
    "args": {
      "External id": 215303,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215303,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 215303, "pid": 5, "tid": 7, "ts": 1716454224825535,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748953, "dur": 12,
    "args": {
      "External id": 215303,
      "cbid": 211, "correlation": 215303
    }
  },
  {
    "ph": "s", "id": 215303, "pid": 76337, "tid": -914061504, "ts": 1716454224748953,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224825544, "dur": 4,
    "args": {
      "External id": 215305,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215305,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 215305, "pid": 5, "tid": 7, "ts": 1716454224825544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224748967, "dur": 13,
    "args": {
      "External id": 215305,
      "cbid": 211, "correlation": 215305
    }
  },
  {
    "ph": "s", "id": 215305, "pid": 76337, "tid": -914061504, "ts": 1716454224748967,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224825548, "dur": 93,
    "args": {
      "External id": 215326,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215326,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 215326, "pid": 5, "tid": 7, "ts": 1716454224825548,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749049, "dur": 13,
    "args": {
      "External id": 215326,
      "cbid": 211, "correlation": 215326
    }
  },
  {
    "ph": "s", "id": 215326, "pid": 76337, "tid": -914061504, "ts": 1716454224749049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749146, "dur": 1,
    "args": {
      "External id": 215344,
      "cbid": 251, "correlation": 215344
    }
  },
  {
    "ph": "f", "id": 215344, "pid": 76337, "tid": -914061504, "ts": 1716454224749146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224825643, "dur": 101,
    "args": {
      "External id": 215346,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215346,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215346, "pid": 5, "tid": 7, "ts": 1716454224825643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749152, "dur": 13,
    "args": {
      "External id": 215346,
      "cbid": 211, "correlation": 215346
    }
  },
  {
    "ph": "s", "id": 215346, "pid": 76337, "tid": -914061504, "ts": 1716454224749152,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224825745, "dur": 19,
    "args": {
      "External id": 215354,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215354,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215354, "pid": 5, "tid": 7, "ts": 1716454224825745,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749220, "dur": 12,
    "args": {
      "External id": 215354,
      "cbid": 211, "correlation": 215354
    }
  },
  {
    "ph": "s", "id": 215354, "pid": 76337, "tid": -914061504, "ts": 1716454224749220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224825765, "dur": 39,
    "args": {
      "External id": 215362,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215362,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215362, "pid": 5, "tid": 7, "ts": 1716454224825765,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749262, "dur": 10,
    "args": {
      "External id": 215362,
      "cbid": 211, "correlation": 215362
    }
  },
  {
    "ph": "s", "id": 215362, "pid": 76337, "tid": -914061504, "ts": 1716454224749262,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224825805, "dur": 35,
    "args": {
      "External id": 215384,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215384,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215384, "pid": 5, "tid": 7, "ts": 1716454224825805,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749313, "dur": 10,
    "args": {
      "External id": 215384,
      "cbid": 211, "correlation": 215384
    }
  },
  {
    "ph": "s", "id": 215384, "pid": 76337, "tid": -914061504, "ts": 1716454224749313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749401, "dur": 1,
    "args": {
      "External id": 215400,
      "cbid": 251, "correlation": 215400
    }
  },
  {
    "ph": "f", "id": 215400, "pid": 76337, "tid": -914061504, "ts": 1716454224749401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749406, "dur": 0,
    "args": {
      "External id": 215402,
      "cbid": 251, "correlation": 215402
    }
  },
  {
    "ph": "f", "id": 215402, "pid": 76337, "tid": -914061504, "ts": 1716454224749406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224825842, "dur": 541,
    "args": {
      "External id": 215403,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215403,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 215403, "pid": 5, "tid": 7, "ts": 1716454224825842,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749409, "dur": 13,
    "args": {
      "External id": 215403,
      "cbid": 211, "correlation": 215403
    }
  },
  {
    "ph": "s", "id": 215403, "pid": 76337, "tid": -914061504, "ts": 1716454224749409,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224826384, "dur": 125,
    "args": {
      "External id": 215411,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215411,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215411, "pid": 5, "tid": 7, "ts": 1716454224826384,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749476, "dur": 12,
    "args": {
      "External id": 215411,
      "cbid": 211, "correlation": 215411
    }
  },
  {
    "ph": "s", "id": 215411, "pid": 76337, "tid": -914061504, "ts": 1716454224749476,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224826511, "dur": 128,
    "args": {
      "External id": 215419,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215419,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215419, "pid": 5, "tid": 7, "ts": 1716454224826511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749505, "dur": 8,
    "args": {
      "External id": 215419,
      "cbid": 211, "correlation": 215419
    }
  },
  {
    "ph": "s", "id": 215419, "pid": 76337, "tid": -914061504, "ts": 1716454224749505,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749582, "dur": 1,
    "args": {
      "External id": 215435,
      "cbid": 251, "correlation": 215435
    }
  },
  {
    "ph": "f", "id": 215435, "pid": 76337, "tid": -914061504, "ts": 1716454224749582,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224826640, "dur": 312,
    "args": {
      "External id": 215437,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215437,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215437, "pid": 5, "tid": 7, "ts": 1716454224826640,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749587, "dur": 12,
    "args": {
      "External id": 215437,
      "cbid": 211, "correlation": 215437
    }
  },
  {
    "ph": "s", "id": 215437, "pid": 76337, "tid": -914061504, "ts": 1716454224749587,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224826954, "dur": 27,
    "args": {
      "External id": 215445,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215445,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215445, "pid": 5, "tid": 7, "ts": 1716454224826954,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749630, "dur": 10,
    "args": {
      "External id": 215445,
      "cbid": 211, "correlation": 215445
    }
  },
  {
    "ph": "s", "id": 215445, "pid": 76337, "tid": -914061504, "ts": 1716454224749630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224826981, "dur": 82,
    "args": {
      "External id": 215456,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215456,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215456, "pid": 5, "tid": 7, "ts": 1716454224826981,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749696, "dur": 12,
    "args": {
      "External id": 215456,
      "cbid": 211, "correlation": 215456
    }
  },
  {
    "ph": "s", "id": 215456, "pid": 76337, "tid": -914061504, "ts": 1716454224749696,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749760, "dur": 0,
    "args": {
      "External id": 215468,
      "cbid": 317, "correlation": 215468
    }
  },
  {
    "ph": "f", "id": 215468, "pid": 76337, "tid": -914061504, "ts": 1716454224749760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749761, "dur": 0,
    "args": {
      "External id": 215469,
      "cbid": 203, "correlation": 215469
    }
  },
  {
    "ph": "f", "id": 215469, "pid": 76337, "tid": -914061504, "ts": 1716454224749761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749761, "dur": 0,
    "args": {
      "External id": 215470,
      "cbid": 205, "correlation": 215470
    }
  },
  {
    "ph": "f", "id": 215470, "pid": 76337, "tid": -914061504, "ts": 1716454224749761,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224827065, "dur": 24,
    "args": {
      "External id": 215474,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215474,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215474, "pid": 5, "tid": 7, "ts": 1716454224827065,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749777, "dur": 12,
    "args": {
      "External id": 215474,
      "cbid": 211, "correlation": 215474
    }
  },
  {
    "ph": "s", "id": 215474, "pid": 76337, "tid": -914061504, "ts": 1716454224749777,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224827091, "dur": 120,
    "args": {
      "External id": 215476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215476,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215476, "pid": 5, "tid": 7, "ts": 1716454224827091,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749796, "dur": 7,
    "args": {
      "External id": 215476,
      "cbid": 211, "correlation": 215476
    }
  },
  {
    "ph": "s", "id": 215476, "pid": 76337, "tid": -914061504, "ts": 1716454224749796,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224827212, "dur": 23,
    "args": {
      "External id": 215478,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215478,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215478, "pid": 5, "tid": 7, "ts": 1716454224827212,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749806, "dur": 5,
    "args": {
      "External id": 215478,
      "cbid": 211, "correlation": 215478
    }
  },
  {
    "ph": "s", "id": 215478, "pid": 76337, "tid": -914061504, "ts": 1716454224749806,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224827237, "dur": 33,
    "args": {
      "External id": 215484,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215484,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215484, "pid": 5, "tid": 7, "ts": 1716454224827237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749834, "dur": 8,
    "args": {
      "External id": 215484,
      "cbid": 211, "correlation": 215484
    }
  },
  {
    "ph": "s", "id": 215484, "pid": 76337, "tid": -914061504, "ts": 1716454224749834,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224827272, "dur": 26,
    "args": {
      "External id": 215492,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215492,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215492, "pid": 5, "tid": 7, "ts": 1716454224827272,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749865, "dur": 8,
    "args": {
      "External id": 215492,
      "cbid": 211, "correlation": 215492
    }
  },
  {
    "ph": "s", "id": 215492, "pid": 76337, "tid": -914061504, "ts": 1716454224749865,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224827299, "dur": 50,
    "args": {
      "External id": 215501,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215501,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215501, "pid": 5, "tid": 7, "ts": 1716454224827299,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749904, "dur": 10,
    "args": {
      "External id": 215501,
      "cbid": 211, "correlation": 215501
    }
  },
  {
    "ph": "s", "id": 215501, "pid": 76337, "tid": -914061504, "ts": 1716454224749904,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224827350, "dur": 43,
    "args": {
      "External id": 215521,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215521,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 215521, "pid": 5, "tid": 7, "ts": 1716454224827350,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224749984, "dur": 12,
    "args": {
      "External id": 215521,
      "cbid": 211, "correlation": 215521
    }
  },
  {
    "ph": "s", "id": 215521, "pid": 76337, "tid": -914061504, "ts": 1716454224749984,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224827395, "dur": 5,
    "args": {
      "External id": 215533,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215533,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 215533, "pid": 5, "tid": 7, "ts": 1716454224827395,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750006, "dur": 6,
    "args": {
      "External id": 215533,
      "cbid": 211, "correlation": 215533
    }
  },
  {
    "ph": "s", "id": 215533, "pid": 76337, "tid": -914061504, "ts": 1716454224750006,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224827401, "dur": 44,
    "args": {
      "External id": 215536,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215536,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215536, "pid": 5, "tid": 7, "ts": 1716454224827401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750026, "dur": 7,
    "args": {
      "External id": 215536,
      "cbid": 211, "correlation": 215536
    }
  },
  {
    "ph": "s", "id": 215536, "pid": 76337, "tid": -914061504, "ts": 1716454224750026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224827447, "dur": 29,
    "args": {
      "External id": 215545,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215545,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 144,
      "warps per SM": 576,
      "grid": [11520, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215545, "pid": 5, "tid": 7, "ts": 1716454224827447,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750066, "dur": 11,
    "args": {
      "External id": 215545,
      "cbid": 211, "correlation": 215545
    }
  },
  {
    "ph": "s", "id": 215545, "pid": 76337, "tid": -914061504, "ts": 1716454224750066,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750118, "dur": 0,
    "args": {
      "External id": 215555,
      "cbid": 317, "correlation": 215555
    }
  },
  {
    "ph": "f", "id": 215555, "pid": 76337, "tid": -914061504, "ts": 1716454224750118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750119, "dur": 0,
    "args": {
      "External id": 215556,
      "cbid": 203, "correlation": 215556
    }
  },
  {
    "ph": "f", "id": 215556, "pid": 76337, "tid": -914061504, "ts": 1716454224750119,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750120, "dur": 0,
    "args": {
      "External id": 215557,
      "cbid": 205, "correlation": 215557
    }
  },
  {
    "ph": "f", "id": 215557, "pid": 76337, "tid": -914061504, "ts": 1716454224750120,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224827478, "dur": 31,
    "args": {
      "External id": 215561,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215561,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215561, "pid": 5, "tid": 7, "ts": 1716454224827478,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750136, "dur": 11,
    "args": {
      "External id": 215561,
      "cbid": 211, "correlation": 215561
    }
  },
  {
    "ph": "s", "id": 215561, "pid": 76337, "tid": -914061504, "ts": 1716454224750136,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224827510, "dur": 64,
    "args": {
      "External id": 215563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215563,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 240,
      "warps per SM": 1920,
      "grid": [1, 30, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215563, "pid": 5, "tid": 7, "ts": 1716454224827510,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750150, "dur": 5,
    "args": {
      "External id": 215563,
      "cbid": 211, "correlation": 215563
    }
  },
  {
    "ph": "s", "id": 215563, "pid": 76337, "tid": -914061504, "ts": 1716454224750150,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224827575, "dur": 968,
    "args": {
      "External id": 215565,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215565,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215565, "pid": 5, "tid": 7, "ts": 1716454224827575,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750160, "dur": 6,
    "args": {
      "External id": 215565,
      "cbid": 211, "correlation": 215565
    }
  },
  {
    "ph": "s", "id": 215565, "pid": 76337, "tid": -914061504, "ts": 1716454224750160,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224828545, "dur": 21,
    "args": {
      "External id": 215567,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215567,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215567, "pid": 5, "tid": 7, "ts": 1716454224828545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750170, "dur": 6,
    "args": {
      "External id": 215567,
      "cbid": 211, "correlation": 215567
    }
  },
  {
    "ph": "s", "id": 215567, "pid": 76337, "tid": -914061504, "ts": 1716454224750170,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224828568, "dur": 33,
    "args": {
      "External id": 215573,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215573,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215573, "pid": 5, "tid": 7, "ts": 1716454224828568,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750198, "dur": 8,
    "args": {
      "External id": 215573,
      "cbid": 211, "correlation": 215573
    }
  },
  {
    "ph": "s", "id": 215573, "pid": 76337, "tid": -914061504, "ts": 1716454224750198,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224828602, "dur": 3,
    "args": {
      "External id": 215581,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215581,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 215581, "pid": 5, "tid": 7, "ts": 1716454224828602,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750242, "dur": 9,
    "args": {
      "External id": 215581,
      "cbid": 211, "correlation": 215581
    }
  },
  {
    "ph": "s", "id": 215581, "pid": 76337, "tid": -914061504, "ts": 1716454224750242,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750306, "dur": 1,
    "args": {
      "External id": 215597,
      "cbid": 251, "correlation": 215597
    }
  },
  {
    "ph": "f", "id": 215597, "pid": 76337, "tid": -914061504, "ts": 1716454224750306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750311, "dur": 0,
    "args": {
      "External id": 215599,
      "cbid": 251, "correlation": 215599
    }
  },
  {
    "ph": "f", "id": 215599, "pid": 76337, "tid": -914061504, "ts": 1716454224750311,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224828607, "dur": 12,
    "args": {
      "External id": 215600,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215600,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 215600, "pid": 5, "tid": 7, "ts": 1716454224828607,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750313, "dur": 11,
    "args": {
      "External id": 215600,
      "cbid": 211, "correlation": 215600
    }
  },
  {
    "ph": "s", "id": 215600, "pid": 76337, "tid": -914061504, "ts": 1716454224750313,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224828620, "dur": 5,
    "args": {
      "External id": 215602,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215602,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 215602, "pid": 5, "tid": 7, "ts": 1716454224828620,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750328, "dur": 5,
    "args": {
      "External id": 215602,
      "cbid": 211, "correlation": 215602
    }
  },
  {
    "ph": "s", "id": 215602, "pid": 76337, "tid": -914061504, "ts": 1716454224750328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224828626, "dur": 29,
    "args": {
      "External id": 215612,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215612,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215612, "pid": 5, "tid": 7, "ts": 1716454224828626,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750385, "dur": 12,
    "args": {
      "External id": 215612,
      "cbid": 211, "correlation": 215612
    }
  },
  {
    "ph": "s", "id": 215612, "pid": 76337, "tid": -914061504, "ts": 1716454224750385,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224828657, "dur": 31,
    "args": {
      "External id": 215632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215632,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 215632, "pid": 5, "tid": 7, "ts": 1716454224828657,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750451, "dur": 11,
    "args": {
      "External id": 215632,
      "cbid": 211, "correlation": 215632
    }
  },
  {
    "ph": "s", "id": 215632, "pid": 76337, "tid": -914061504, "ts": 1716454224750451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224828689, "dur": 4,
    "args": {
      "External id": 215644,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215644,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 215644, "pid": 5, "tid": 7, "ts": 1716454224828689,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750471, "dur": 7,
    "args": {
      "External id": 215644,
      "cbid": 211, "correlation": 215644
    }
  },
  {
    "ph": "s", "id": 215644, "pid": 76337, "tid": -914061504, "ts": 1716454224750471,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224828694, "dur": 29,
    "args": {
      "External id": 215647,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215647,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215647, "pid": 5, "tid": 7, "ts": 1716454224828694,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750491, "dur": 7,
    "args": {
      "External id": 215647,
      "cbid": 211, "correlation": 215647
    }
  },
  {
    "ph": "s", "id": 215647, "pid": 76337, "tid": -914061504, "ts": 1716454224750491,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224828725, "dur": 20,
    "args": {
      "External id": 215656,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215656,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215656, "pid": 5, "tid": 7, "ts": 1716454224828725,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750532, "dur": 9,
    "args": {
      "External id": 215656,
      "cbid": 211, "correlation": 215656
    }
  },
  {
    "ph": "s", "id": 215656, "pid": 76337, "tid": -914061504, "ts": 1716454224750532,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750593, "dur": 0,
    "args": {
      "External id": 215666,
      "cbid": 317, "correlation": 215666
    }
  },
  {
    "ph": "f", "id": 215666, "pid": 76337, "tid": -914061504, "ts": 1716454224750593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750593, "dur": 0,
    "args": {
      "External id": 215667,
      "cbid": 203, "correlation": 215667
    }
  },
  {
    "ph": "f", "id": 215667, "pid": 76337, "tid": -914061504, "ts": 1716454224750593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750594, "dur": 0,
    "args": {
      "External id": 215668,
      "cbid": 205, "correlation": 215668
    }
  },
  {
    "ph": "f", "id": 215668, "pid": 76337, "tid": -914061504, "ts": 1716454224750594,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224828746, "dur": 23,
    "args": {
      "External id": 215672,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215672,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215672, "pid": 5, "tid": 7, "ts": 1716454224828746,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750608, "dur": 12,
    "args": {
      "External id": 215672,
      "cbid": 211, "correlation": 215672
    }
  },
  {
    "ph": "s", "id": 215672, "pid": 76337, "tid": -914061504, "ts": 1716454224750608,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224828770, "dur": 44,
    "args": {
      "External id": 215674,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215674,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215674, "pid": 5, "tid": 7, "ts": 1716454224828770,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750622, "dur": 6,
    "args": {
      "External id": 215674,
      "cbid": 211, "correlation": 215674
    }
  },
  {
    "ph": "s", "id": 215674, "pid": 76337, "tid": -914061504, "ts": 1716454224750622,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224828816, "dur": 648,
    "args": {
      "External id": 215676,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215676,
      "registers per thread": 254,
      "shared memory": 17408,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215676, "pid": 5, "tid": 7, "ts": 1716454224828816,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750634, "dur": 6,
    "args": {
      "External id": 215676,
      "cbid": 211, "correlation": 215676
    }
  },
  {
    "ph": "s", "id": 215676, "pid": 76337, "tid": -914061504, "ts": 1716454224750634,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224829466, "dur": 22,
    "args": {
      "External id": 215678,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215678,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215678, "pid": 5, "tid": 7, "ts": 1716454224829466,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750643, "dur": 5,
    "args": {
      "External id": 215678,
      "cbid": 211, "correlation": 215678
    }
  },
  {
    "ph": "s", "id": 215678, "pid": 76337, "tid": -914061504, "ts": 1716454224750643,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224829489, "dur": 33,
    "args": {
      "External id": 215684,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215684,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215684, "pid": 5, "tid": 7, "ts": 1716454224829489,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750671, "dur": 9,
    "args": {
      "External id": 215684,
      "cbid": 211, "correlation": 215684
    }
  },
  {
    "ph": "s", "id": 215684, "pid": 76337, "tid": -914061504, "ts": 1716454224750671,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750729, "dur": 0,
    "args": {
      "External id": 215694,
      "cbid": 317, "correlation": 215694
    }
  },
  {
    "ph": "f", "id": 215694, "pid": 76337, "tid": -914061504, "ts": 1716454224750729,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750730, "dur": 0,
    "args": {
      "External id": 215695,
      "cbid": 203, "correlation": 215695
    }
  },
  {
    "ph": "f", "id": 215695, "pid": 76337, "tid": -914061504, "ts": 1716454224750730,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750731, "dur": 0,
    "args": {
      "External id": 215696,
      "cbid": 205, "correlation": 215696
    }
  },
  {
    "ph": "f", "id": 215696, "pid": 76337, "tid": -914061504, "ts": 1716454224750731,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224829523, "dur": 30,
    "args": {
      "External id": 215700,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215700,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 72,
      "warps per SM": 576,
      "grid": [24, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215700, "pid": 5, "tid": 7, "ts": 1716454224829523,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750744, "dur": 11,
    "args": {
      "External id": 215700,
      "cbid": 211, "correlation": 215700
    }
  },
  {
    "ph": "s", "id": 215700, "pid": 76337, "tid": -914061504, "ts": 1716454224750744,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224829555, "dur": 153,
    "args": {
      "External id": 215702,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215702,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215702, "pid": 5, "tid": 7, "ts": 1716454224829555,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750762, "dur": 6,
    "args": {
      "External id": 215702,
      "cbid": 211, "correlation": 215702
    }
  },
  {
    "ph": "s", "id": 215702, "pid": 76337, "tid": -914061504, "ts": 1716454224750762,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224829710, "dur": 22,
    "args": {
      "External id": 215704,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215704,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215704, "pid": 5, "tid": 7, "ts": 1716454224829710,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750771, "dur": 5,
    "args": {
      "External id": 215704,
      "cbid": 211, "correlation": 215704
    }
  },
  {
    "ph": "s", "id": 215704, "pid": 76337, "tid": -914061504, "ts": 1716454224750771,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224829733, "dur": 33,
    "args": {
      "External id": 215710,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215710,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215710, "pid": 5, "tid": 7, "ts": 1716454224829733,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750798, "dur": 8,
    "args": {
      "External id": 215710,
      "cbid": 211, "correlation": 215710
    }
  },
  {
    "ph": "s", "id": 215710, "pid": 76337, "tid": -914061504, "ts": 1716454224750798,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224829767, "dur": 27,
    "args": {
      "External id": 215718,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215718,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215718, "pid": 5, "tid": 7, "ts": 1716454224829767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750826, "dur": 9,
    "args": {
      "External id": 215718,
      "cbid": 211, "correlation": 215718
    }
  },
  {
    "ph": "s", "id": 215718, "pid": 76337, "tid": -914061504, "ts": 1716454224750826,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224829796, "dur": 20,
    "args": {
      "External id": 215726,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215726,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215726, "pid": 5, "tid": 7, "ts": 1716454224829796,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750855, "dur": 8,
    "args": {
      "External id": 215726,
      "cbid": 211, "correlation": 215726
    }
  },
  {
    "ph": "s", "id": 215726, "pid": 76337, "tid": -914061504, "ts": 1716454224750855,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224829817, "dur": 30,
    "args": {
      "External id": 215746,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215746,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 215746, "pid": 5, "tid": 7, "ts": 1716454224829817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750935, "dur": 12,
    "args": {
      "External id": 215746,
      "cbid": 211, "correlation": 215746
    }
  },
  {
    "ph": "s", "id": 215746, "pid": 76337, "tid": -914061504, "ts": 1716454224750935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224829848, "dur": 4,
    "args": {
      "External id": 215758,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215758,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 215758, "pid": 5, "tid": 7, "ts": 1716454224829848,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750957, "dur": 6,
    "args": {
      "External id": 215758,
      "cbid": 211, "correlation": 215758
    }
  },
  {
    "ph": "s", "id": 215758, "pid": 76337, "tid": -914061504, "ts": 1716454224750957,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224829854, "dur": 31,
    "args": {
      "External id": 215761,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215761,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215761, "pid": 5, "tid": 7, "ts": 1716454224829854,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224750982, "dur": 9,
    "args": {
      "External id": 215761,
      "cbid": 211, "correlation": 215761
    }
  },
  {
    "ph": "s", "id": 215761, "pid": 76337, "tid": -914061504, "ts": 1716454224750982,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751043, "dur": 0,
    "args": {
      "External id": 215772,
      "cbid": 317, "correlation": 215772
    }
  },
  {
    "ph": "f", "id": 215772, "pid": 76337, "tid": -914061504, "ts": 1716454224751043,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751044, "dur": 0,
    "args": {
      "External id": 215773,
      "cbid": 203, "correlation": 215773
    }
  },
  {
    "ph": "f", "id": 215773, "pid": 76337, "tid": -914061504, "ts": 1716454224751044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751044, "dur": 0,
    "args": {
      "External id": 215774,
      "cbid": 205, "correlation": 215774
    }
  },
  {
    "ph": "f", "id": 215774, "pid": 76337, "tid": -914061504, "ts": 1716454224751044,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224829886, "dur": 23,
    "args": {
      "External id": 215778,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215778,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215778, "pid": 5, "tid": 7, "ts": 1716454224829886,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751058, "dur": 12,
    "args": {
      "External id": 215778,
      "cbid": 211, "correlation": 215778
    }
  },
  {
    "ph": "s", "id": 215778, "pid": 76337, "tid": -914061504, "ts": 1716454224751058,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224829910, "dur": 105,
    "args": {
      "External id": 215780,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215780,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215780, "pid": 5, "tid": 7, "ts": 1716454224829910,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751075, "dur": 6,
    "args": {
      "External id": 215780,
      "cbid": 211, "correlation": 215780
    }
  },
  {
    "ph": "s", "id": 215780, "pid": 76337, "tid": -914061504, "ts": 1716454224751075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224830017, "dur": 22,
    "args": {
      "External id": 215782,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215782,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215782, "pid": 5, "tid": 7, "ts": 1716454224830017,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751086, "dur": 5,
    "args": {
      "External id": 215782,
      "cbid": 211, "correlation": 215782
    }
  },
  {
    "ph": "s", "id": 215782, "pid": 76337, "tid": -914061504, "ts": 1716454224751086,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224830041, "dur": 32,
    "args": {
      "External id": 215788,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215788,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215788, "pid": 5, "tid": 7, "ts": 1716454224830041,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751113, "dur": 8,
    "args": {
      "External id": 215788,
      "cbid": 211, "correlation": 215788
    }
  },
  {
    "ph": "s", "id": 215788, "pid": 76337, "tid": -914061504, "ts": 1716454224751113,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224830074, "dur": 163,
    "args": {
      "External id": 215797,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215797,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215797, "pid": 5, "tid": 7, "ts": 1716454224830074,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751196, "dur": 13,
    "args": {
      "External id": 215797,
      "cbid": 211, "correlation": 215797
    }
  },
  {
    "ph": "s", "id": 215797, "pid": 76337, "tid": -914061504, "ts": 1716454224751196,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224830239, "dur": 65,
    "args": {
      "External id": 215819,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215819,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 153.6,
      "warps per SM": 614.4,
      "grid": [12288, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215819, "pid": 5, "tid": 7, "ts": 1716454224830239,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751251, "dur": 10,
    "args": {
      "External id": 215819,
      "cbid": 211, "correlation": 215819
    }
  },
  {
    "ph": "s", "id": 215819, "pid": 76337, "tid": -914061504, "ts": 1716454224751251,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751340, "dur": 1,
    "args": {
      "External id": 215830,
      "cbid": 251, "correlation": 215830
    }
  },
  {
    "ph": "f", "id": 215830, "pid": 76337, "tid": -914061504, "ts": 1716454224751340,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224830306, "dur": 150,
    "args": {
      "External id": 215831,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215831,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215831, "pid": 5, "tid": 7, "ts": 1716454224830306,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751346, "dur": 13,
    "args": {
      "External id": 215831,
      "cbid": 211, "correlation": 215831
    }
  },
  {
    "ph": "s", "id": 215831, "pid": 76337, "tid": -914061504, "ts": 1716454224751346,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751415, "dur": 1,
    "args": {
      "External id": 215842,
      "cbid": 251, "correlation": 215842
    }
  },
  {
    "ph": "f", "id": 215842, "pid": 76337, "tid": -914061504, "ts": 1716454224751415,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224830457, "dur": 144,
    "args": {
      "External id": 215843,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215843,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215843, "pid": 5, "tid": 7, "ts": 1716454224830457,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751419, "dur": 12,
    "args": {
      "External id": 215843,
      "cbid": 211, "correlation": 215843
    }
  },
  {
    "ph": "s", "id": 215843, "pid": 76337, "tid": -914061504, "ts": 1716454224751419,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751486, "dur": 1,
    "args": {
      "External id": 215854,
      "cbid": 251, "correlation": 215854
    }
  },
  {
    "ph": "f", "id": 215854, "pid": 76337, "tid": -914061504, "ts": 1716454224751486,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224830603, "dur": 144,
    "args": {
      "External id": 215855,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215855,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 6,
      "warps per SM": 24,
      "grid": [5, 96, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215855, "pid": 5, "tid": 7, "ts": 1716454224830603,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751490, "dur": 11,
    "args": {
      "External id": 215855,
      "cbid": 211, "correlation": 215855
    }
  },
  {
    "ph": "s", "id": 215855, "pid": 76337, "tid": -914061504, "ts": 1716454224751490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224830748, "dur": 1969,
    "args": {
      "External id": 215876,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215876,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 215876, "pid": 5, "tid": 7, "ts": 1716454224830748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751571, "dur": 13,
    "args": {
      "External id": 215876,
      "cbid": 211, "correlation": 215876
    }
  },
  {
    "ph": "s", "id": 215876, "pid": 76337, "tid": -914061504, "ts": 1716454224751571,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751669, "dur": 1,
    "args": {
      "External id": 215894,
      "cbid": 251, "correlation": 215894
    }
  },
  {
    "ph": "f", "id": 215894, "pid": 76337, "tid": -914061504, "ts": 1716454224751669,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224832719, "dur": 150,
    "args": {
      "External id": 215896,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215896,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3,
      "warps per SM": 24,
      "grid": [5, 48, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 215896, "pid": 5, "tid": 7, "ts": 1716454224832719,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751675, "dur": 14,
    "args": {
      "External id": 215896,
      "cbid": 211, "correlation": 215896
    }
  },
  {
    "ph": "s", "id": 215896, "pid": 76337, "tid": -914061504, "ts": 1716454224751675,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224832870, "dur": 35,
    "args": {
      "External id": 215904,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215904,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215904, "pid": 5, "tid": 7, "ts": 1716454224832870,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751745, "dur": 12,
    "args": {
      "External id": 215904,
      "cbid": 211, "correlation": 215904
    }
  },
  {
    "ph": "s", "id": 215904, "pid": 76337, "tid": -914061504, "ts": 1716454224751745,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224832907, "dur": 50,
    "args": {
      "External id": 215912,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215912,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215912, "pid": 5, "tid": 7, "ts": 1716454224832907,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751785, "dur": 9,
    "args": {
      "External id": 215912,
      "cbid": 211, "correlation": 215912
    }
  },
  {
    "ph": "s", "id": 215912, "pid": 76337, "tid": -914061504, "ts": 1716454224751785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224832958, "dur": 30,
    "args": {
      "External id": 215923,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215923,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215923, "pid": 5, "tid": 7, "ts": 1716454224832958,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751856, "dur": 12,
    "args": {
      "External id": 215923,
      "cbid": 211, "correlation": 215923
    }
  },
  {
    "ph": "s", "id": 215923, "pid": 76337, "tid": -914061504, "ts": 1716454224751856,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224832990, "dur": 35,
    "args": {
      "External id": 215945,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215945,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 215945, "pid": 5, "tid": 7, "ts": 1716454224832990,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751887, "dur": 8,
    "args": {
      "External id": 215945,
      "cbid": 211, "correlation": 215945
    }
  },
  {
    "ph": "s", "id": 215945, "pid": 76337, "tid": -914061504, "ts": 1716454224751887,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751973, "dur": 8,
    "args": {
      "External id": 215956,
      "cbid": 251, "correlation": 215956
    }
  },
  {
    "ph": "f", "id": 215956, "pid": 76337, "tid": -914061504, "ts": 1716454224751973,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224833026, "dur": 91,
    "args": {
      "External id": 215957,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215957,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 215957, "pid": 5, "tid": 7, "ts": 1716454224833026,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224751985, "dur": 14,
    "args": {
      "External id": 215957,
      "cbid": 211, "correlation": 215957
    }
  },
  {
    "ph": "s", "id": 215957, "pid": 76337, "tid": -914061504, "ts": 1716454224751985,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752056, "dur": 1,
    "args": {
      "External id": 215968,
      "cbid": 251, "correlation": 215968
    }
  },
  {
    "ph": "f", "id": 215968, "pid": 76337, "tid": -914061504, "ts": 1716454224752056,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752060, "dur": 0,
    "args": {
      "External id": 215969,
      "cbid": 251, "correlation": 215969
    }
  },
  {
    "ph": "f", "id": 215969, "pid": 76337, "tid": -914061504, "ts": 1716454224752060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224833118, "dur": 11,
    "args": {
      "External id": 215970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215970,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 215970, "pid": 5, "tid": 7, "ts": 1716454224833118,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752061, "dur": 12,
    "args": {
      "External id": 215970,
      "cbid": 211, "correlation": 215970
    }
  },
  {
    "ph": "s", "id": 215970, "pid": 76337, "tid": -914061504, "ts": 1716454224752061,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224833130, "dur": 5,
    "args": {
      "External id": 215972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215972,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 215972, "pid": 5, "tid": 7, "ts": 1716454224833130,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752075, "dur": 6,
    "args": {
      "External id": 215972,
      "cbid": 211, "correlation": 215972
    }
  },
  {
    "ph": "s", "id": 215972, "pid": 76337, "tid": -914061504, "ts": 1716454224752075,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752133, "dur": 1,
    "args": {
      "External id": 215983,
      "cbid": 251, "correlation": 215983
    }
  },
  {
    "ph": "f", "id": 215983, "pid": 76337, "tid": -914061504, "ts": 1716454224752133,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752136, "dur": 0,
    "args": {
      "External id": 215984,
      "cbid": 251, "correlation": 215984
    }
  },
  {
    "ph": "f", "id": 215984, "pid": 76337, "tid": -914061504, "ts": 1716454224752136,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224833137, "dur": 7,
    "args": {
      "External id": 215985,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215985,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 3, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 215985, "pid": 5, "tid": 7, "ts": 1716454224833137,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752138, "dur": 12,
    "args": {
      "External id": 215985,
      "cbid": 211, "correlation": 215985
    }
  },
  {
    "ph": "s", "id": 215985, "pid": 76337, "tid": -914061504, "ts": 1716454224752138,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, false, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224833145, "dur": 4,
    "args": {
      "External id": 215987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 215987,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 4,
      "grid": [20, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 6
    }
  },
  {
    "ph": "f", "id": 215987, "pid": 5, "tid": 7, "ts": 1716454224833145,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752151, "dur": 5,
    "args": {
      "External id": 215987,
      "cbid": 211, "correlation": 215987
    }
  },
  {
    "ph": "s", "id": 215987, "pid": 76337, "tid": -914061504, "ts": 1716454224752151,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 32l, 128l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224833150, "dur": 94,
    "args": {
      "External id": 216008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216008,
      "registers per thread": 168,
      "shared memory": 24960,
      "blocks per SM": 19.2,
      "warps per SM": 76.8,
      "grid": [24, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 216008, "pid": 5, "tid": 7, "ts": 1716454224833150,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752226, "dur": 13,
    "args": {
      "External id": 216008,
      "cbid": 211, "correlation": 216008
    }
  },
  {
    "ph": "s", "id": 216008, "pid": 76337, "tid": -914061504, "ts": 1716454224752226,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752322, "dur": 1,
    "args": {
      "External id": 216026,
      "cbid": 251, "correlation": 216026
    }
  },
  {
    "ph": "f", "id": 216026, "pid": 76337, "tid": -914061504, "ts": 1716454224752322,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224833245, "dur": 86,
    "args": {
      "External id": 216028,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216028,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216028, "pid": 5, "tid": 7, "ts": 1716454224833245,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752328, "dur": 14,
    "args": {
      "External id": 216028,
      "cbid": 211, "correlation": 216028
    }
  },
  {
    "ph": "s", "id": 216028, "pid": 76337, "tid": -914061504, "ts": 1716454224752328,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224833333, "dur": 19,
    "args": {
      "External id": 216036,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216036,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216036, "pid": 5, "tid": 7, "ts": 1716454224833333,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752399, "dur": 12,
    "args": {
      "External id": 216036,
      "cbid": 211, "correlation": 216036
    }
  },
  {
    "ph": "s", "id": 216036, "pid": 76337, "tid": -914061504, "ts": 1716454224752399,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224833353, "dur": 37,
    "args": {
      "External id": 216044,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216044,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216044, "pid": 5, "tid": 7, "ts": 1716454224833353,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752440, "dur": 10,
    "args": {
      "External id": 216044,
      "cbid": 211, "correlation": 216044
    }
  },
  {
    "ph": "s", "id": 216044, "pid": 76337, "tid": -914061504, "ts": 1716454224752440,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224833392, "dur": 35,
    "args": {
      "External id": 216066,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216066,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [6144, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216066, "pid": 5, "tid": 7, "ts": 1716454224833392,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752490, "dur": 10,
    "args": {
      "External id": 216066,
      "cbid": 211, "correlation": 216066
    }
  },
  {
    "ph": "s", "id": 216066, "pid": 76337, "tid": -914061504, "ts": 1716454224752490,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752580, "dur": 1,
    "args": {
      "External id": 216082,
      "cbid": 251, "correlation": 216082
    }
  },
  {
    "ph": "f", "id": 216082, "pid": 76337, "tid": -914061504, "ts": 1716454224752580,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752584, "dur": 0,
    "args": {
      "External id": 216084,
      "cbid": 251, "correlation": 216084
    }
  },
  {
    "ph": "f", "id": 216084, "pid": 76337, "tid": -914061504, "ts": 1716454224752584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224833428, "dur": 543,
    "args": {
      "External id": 216085,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216085,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 12,
      "warps per SM": 96,
      "grid": [40, 24, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 216085, "pid": 5, "tid": 7, "ts": 1716454224833428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752588, "dur": 14,
    "args": {
      "External id": 216085,
      "cbid": 211, "correlation": 216085
    }
  },
  {
    "ph": "s", "id": 216085, "pid": 76337, "tid": -914061504, "ts": 1716454224752588,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224833972, "dur": 124,
    "args": {
      "External id": 216093,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216093,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216093, "pid": 5, "tid": 7, "ts": 1716454224833972,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752653, "dur": 12,
    "args": {
      "External id": 216093,
      "cbid": 211, "correlation": 216093
    }
  },
  {
    "ph": "s", "id": 216093, "pid": 76337, "tid": -914061504, "ts": 1716454224752653,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224834098, "dur": 130,
    "args": {
      "External id": 216101,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216101,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216101, "pid": 5, "tid": 7, "ts": 1716454224834098,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752683, "dur": 8,
    "args": {
      "External id": 216101,
      "cbid": 211, "correlation": 216101
    }
  },
  {
    "ph": "s", "id": 216101, "pid": 76337, "tid": -914061504, "ts": 1716454224752683,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752760, "dur": 1,
    "args": {
      "External id": 216117,
      "cbid": 251, "correlation": 216117
    }
  },
  {
    "ph": "f", "id": 216117, "pid": 76337, "tid": -914061504, "ts": 1716454224752760,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224834229, "dur": 305,
    "args": {
      "External id": 216119,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216119,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 3,
      "warps per SM": 12,
      "grid": [5, 48, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216119, "pid": 5, "tid": 7, "ts": 1716454224834229,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752766, "dur": 12,
    "args": {
      "External id": 216119,
      "cbid": 211, "correlation": 216119
    }
  },
  {
    "ph": "s", "id": 216119, "pid": 76337, "tid": -914061504, "ts": 1716454224752766,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224834536, "dur": 27,
    "args": {
      "External id": 216127,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216127,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216127, "pid": 5, "tid": 7, "ts": 1716454224834536,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752808, "dur": 9,
    "args": {
      "External id": 216127,
      "cbid": 211, "correlation": 216127
    }
  },
  {
    "ph": "s", "id": 216127, "pid": 76337, "tid": -914061504, "ts": 1716454224752808,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224834564, "dur": 81,
    "args": {
      "External id": 216138,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216138,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216138, "pid": 5, "tid": 7, "ts": 1716454224834564,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752875, "dur": 12,
    "args": {
      "External id": 216138,
      "cbid": 211, "correlation": 216138
    }
  },
  {
    "ph": "s", "id": 216138, "pid": 76337, "tid": -914061504, "ts": 1716454224752875,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752938, "dur": 0,
    "args": {
      "External id": 216150,
      "cbid": 317, "correlation": 216150
    }
  },
  {
    "ph": "f", "id": 216150, "pid": 76337, "tid": -914061504, "ts": 1716454224752938,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752939, "dur": 0,
    "args": {
      "External id": 216151,
      "cbid": 203, "correlation": 216151
    }
  },
  {
    "ph": "f", "id": 216151, "pid": 76337, "tid": -914061504, "ts": 1716454224752939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752939, "dur": 0,
    "args": {
      "External id": 216152,
      "cbid": 205, "correlation": 216152
    }
  },
  {
    "ph": "f", "id": 216152, "pid": 76337, "tid": -914061504, "ts": 1716454224752939,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224834647, "dur": 22,
    "args": {
      "External id": 216156,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216156,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216156, "pid": 5, "tid": 7, "ts": 1716454224834647,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752955, "dur": 12,
    "args": {
      "External id": 216156,
      "cbid": 211, "correlation": 216156
    }
  },
  {
    "ph": "s", "id": 216156, "pid": 76337, "tid": -914061504, "ts": 1716454224752955,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize64x256x64_stage1_warpsize1x4x2_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224834671, "dur": 121,
    "args": {
      "External id": 216158,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216158,
      "registers per thread": 255,
      "shared memory": 40960,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216158, "pid": 5, "tid": 7, "ts": 1716454224834671,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752981, "dur": 7,
    "args": {
      "External id": 216158,
      "cbid": 211, "correlation": 216158
    }
  },
  {
    "ph": "s", "id": 216158, "pid": 76337, "tid": -914061504, "ts": 1716454224752981,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224834793, "dur": 23,
    "args": {
      "External id": 216160,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216160,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 48,
      "warps per SM": 384,
      "grid": [24, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216160, "pid": 5, "tid": 7, "ts": 1716454224834793,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224752992, "dur": 6,
    "args": {
      "External id": 216160,
      "cbid": 211, "correlation": 216160
    }
  },
  {
    "ph": "s", "id": 216160, "pid": 76337, "tid": -914061504, "ts": 1716454224752992,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224834817, "dur": 33,
    "args": {
      "External id": 216166,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216166,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216166, "pid": 5, "tid": 7, "ts": 1716454224834817,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753022, "dur": 9,
    "args": {
      "External id": 216166,
      "cbid": 211, "correlation": 216166
    }
  },
  {
    "ph": "s", "id": 216166, "pid": 76337, "tid": -914061504, "ts": 1716454224753022,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224834852, "dur": 26,
    "args": {
      "External id": 216174,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216174,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 96,
      "warps per SM": 384,
      "grid": [7680, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216174, "pid": 5, "tid": 7, "ts": 1716454224834852,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753055, "dur": 8,
    "args": {
      "External id": 216174,
      "cbid": 211, "correlation": 216174
    }
  },
  {
    "ph": "s", "id": 216174, "pid": 76337, "tid": -914061504, "ts": 1716454224753055,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::upsample_nearest2d_out_frame<c10::Half, &at::native::nearest_neighbor_compute_source_index>(c10::Half const*, c10::Half*, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)", "pid": 5, "tid": 7,
    "ts": 1716454224834880, "dur": 102,
    "args": {
      "External id": 216185,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216185,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 128,
      "warps per SM": 2048,
      "grid": [2, 4, 1280],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216185, "pid": 5, "tid": 7, "ts": 1716454224834880,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753117, "dur": 11,
    "args": {
      "External id": 216185,
      "cbid": 211, "correlation": 216185
    }
  },
  {
    "ph": "s", "id": 216185, "pid": 76337, "tid": -914061504, "ts": 1716454224753117,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753171, "dur": 0,
    "args": {
      "External id": 216195,
      "cbid": 317, "correlation": 216195
    }
  },
  {
    "ph": "f", "id": 216195, "pid": 76337, "tid": -914061504, "ts": 1716454224753171,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753172, "dur": 0,
    "args": {
      "External id": 216196,
      "cbid": 203, "correlation": 216196
    }
  },
  {
    "ph": "f", "id": 216196, "pid": 76337, "tid": -914061504, "ts": 1716454224753172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753172, "dur": 0,
    "args": {
      "External id": 216197,
      "cbid": 205, "correlation": 216197
    }
  },
  {
    "ph": "f", "id": 216197, "pid": 76337, "tid": -914061504, "ts": 1716454224753172,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224834983, "dur": 75,
    "args": {
      "External id": 216201,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216201,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216201, "pid": 5, "tid": 7, "ts": 1716454224834983,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753186, "dur": 11,
    "args": {
      "External id": 216201,
      "cbid": 211, "correlation": 216201
    }
  },
  {
    "ph": "s", "id": 216201, "pid": 76337, "tid": -914061504, "ts": 1716454224753186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224835059, "dur": 44,
    "args": {
      "External id": 216203,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216203,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 160,
      "warps per SM": 1280,
      "grid": [1, 20, 640],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216203, "pid": 5, "tid": 7, "ts": 1716454224835059,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753201, "dur": 6,
    "args": {
      "External id": 216203,
      "cbid": 211, "correlation": 216203
    }
  },
  {
    "ph": "s", "id": 216203, "pid": 76337, "tid": -914061504, "ts": 1716454224753201,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224835105, "dur": 4,
    "args": {
      "External id": 216205,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216205,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216205, "pid": 5, "tid": 7, "ts": 1716454224835105,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753211, "dur": 6,
    "args": {
      "External id": 216205,
      "cbid": 211, "correlation": 216205
    }
  },
  {
    "ph": "s", "id": 216205, "pid": 76337, "tid": -914061504, "ts": 1716454224753211,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753220, "dur": 0,
    "args": {
      "External id": 216206,
      "cbid": 51, "correlation": 216206
    }
  },
  {
    "ph": "s", "id": 216206, "pid": 76337, "tid": -914061504, "ts": 1716454224753220,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x64_ldg8_relu_f2f_exp_small_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224835110, "dur": 2235,
    "args": {
      "External id": 216207,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216207,
      "registers per thread": 242,
      "shared memory": 41472,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [96, 10, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216207, "pid": 5, "tid": 7, "ts": 1716454224835110,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753221, "dur": 5,
    "args": {
      "External id": 216207,
      "cbid": 211, "correlation": 216207
    }
  },
  {
    "ph": "s", "id": 216207, "pid": 76337, "tid": -914061504, "ts": 1716454224753221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224837347, "dur": 114,
    "args": {
      "External id": 216212,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216212,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216212, "pid": 5, "tid": 7, "ts": 1716454224837347,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753248, "dur": 9,
    "args": {
      "External id": 216212,
      "cbid": 211, "correlation": 216212
    }
  },
  {
    "ph": "s", "id": 216212, "pid": 76337, "tid": -914061504, "ts": 1716454224753248,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224837463, "dur": 169,
    "args": {
      "External id": 216221,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216221,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216221, "pid": 5, "tid": 7, "ts": 1716454224837463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753339, "dur": 14,
    "args": {
      "External id": 216221,
      "cbid": 211, "correlation": 216221
    }
  },
  {
    "ph": "s", "id": 216221, "pid": 76337, "tid": -914061504, "ts": 1716454224753339,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224837633, "dur": 128,
    "args": {
      "External id": 216241,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216241,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 216241, "pid": 5, "tid": 7, "ts": 1716454224837633,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753411, "dur": 11,
    "args": {
      "External id": 216241,
      "cbid": 211, "correlation": 216241
    }
  },
  {
    "ph": "s", "id": 216241, "pid": 76337, "tid": -914061504, "ts": 1716454224753411,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224837762, "dur": 5,
    "args": {
      "External id": 216253,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216253,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.375,
      "warps per SM": 3,
      "grid": [30, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 216253, "pid": 5, "tid": 7, "ts": 1716454224837762,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753433, "dur": 6,
    "args": {
      "External id": 216253,
      "cbid": 211, "correlation": 216253
    }
  },
  {
    "ph": "s", "id": 216253, "pid": 76337, "tid": -914061504, "ts": 1716454224753433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224837768, "dur": 160,
    "args": {
      "External id": 216256,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216256,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216256, "pid": 5, "tid": 7, "ts": 1716454224837768,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753451, "dur": 8,
    "args": {
      "External id": 216256,
      "cbid": 211, "correlation": 216256
    }
  },
  {
    "ph": "s", "id": 216256, "pid": 76337, "tid": -914061504, "ts": 1716454224753451,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224837930, "dur": 102,
    "args": {
      "External id": 216265,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216265,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 576,
      "warps per SM": 2304,
      "grid": [46080, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216265, "pid": 5, "tid": 7, "ts": 1716454224837930,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753492, "dur": 10,
    "args": {
      "External id": 216265,
      "cbid": 211, "correlation": 216265
    }
  },
  {
    "ph": "s", "id": 216265, "pid": 76337, "tid": -914061504, "ts": 1716454224753492,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753544, "dur": 0,
    "args": {
      "External id": 216275,
      "cbid": 317, "correlation": 216275
    }
  },
  {
    "ph": "f", "id": 216275, "pid": 76337, "tid": -914061504, "ts": 1716454224753544,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753545, "dur": 0,
    "args": {
      "External id": 216276,
      "cbid": 203, "correlation": 216276
    }
  },
  {
    "ph": "f", "id": 216276, "pid": 76337, "tid": -914061504, "ts": 1716454224753545,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753546, "dur": 0,
    "args": {
      "External id": 216277,
      "cbid": 205, "correlation": 216277
    }
  },
  {
    "ph": "f", "id": 216277, "pid": 76337, "tid": -914061504, "ts": 1716454224753546,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224838033, "dur": 111,
    "args": {
      "External id": 216281,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216281,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 288,
      "warps per SM": 2304,
      "grid": [96, 30, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216281, "pid": 5, "tid": 7, "ts": 1716454224838033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753562, "dur": 11,
    "args": {
      "External id": 216281,
      "cbid": 211, "correlation": 216281
    }
  },
  {
    "ph": "s", "id": 216281, "pid": 76337, "tid": -914061504, "ts": 1716454224753562,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224838146, "dur": 34,
    "args": {
      "External id": 216283,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216283,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 120,
      "warps per SM": 960,
      "grid": [1, 30, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216283, "pid": 5, "tid": 7, "ts": 1716454224838146,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753576, "dur": 5,
    "args": {
      "External id": 216283,
      "cbid": 211, "correlation": 216283
    }
  },
  {
    "ph": "s", "id": 216283, "pid": 76337, "tid": -914061504, "ts": 1716454224753576,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224838181, "dur": 3,
    "args": {
      "External id": 216285,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216285,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216285, "pid": 5, "tid": 7, "ts": 1716454224838181,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753586, "dur": 5,
    "args": {
      "External id": 216285,
      "cbid": 211, "correlation": 216285
    }
  },
  {
    "ph": "s", "id": 216285, "pid": 76337, "tid": -914061504, "ts": 1716454224753586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753594, "dur": 0,
    "args": {
      "External id": 216286,
      "cbid": 51, "correlation": 216286
    }
  },
  {
    "ph": "s", "id": 216286, "pid": 76337, "tid": -914061504, "ts": 1716454224753594,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224838186, "dur": 2023,
    "args": {
      "External id": 216287,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216287,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216287, "pid": 5, "tid": 7, "ts": 1716454224838186,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753595, "dur": 6,
    "args": {
      "External id": 216287,
      "cbid": 211, "correlation": 216287
    }
  },
  {
    "ph": "s", "id": 216287, "pid": 76337, "tid": -914061504, "ts": 1716454224753595,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224840210, "dur": 59,
    "args": {
      "External id": 216292,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216292,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216292, "pid": 5, "tid": 7, "ts": 1716454224840210,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753625, "dur": 8,
    "args": {
      "External id": 216292,
      "cbid": 211, "correlation": 216292
    }
  },
  {
    "ph": "s", "id": 216292, "pid": 76337, "tid": -914061504, "ts": 1716454224753625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224840270, "dur": 3,
    "args": {
      "External id": 216300,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216300,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216300, "pid": 5, "tid": 7, "ts": 1716454224840270,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753668, "dur": 9,
    "args": {
      "External id": 216300,
      "cbid": 211, "correlation": 216300
    }
  },
  {
    "ph": "s", "id": 216300, "pid": 76337, "tid": -914061504, "ts": 1716454224753668,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753732, "dur": 1,
    "args": {
      "External id": 216316,
      "cbid": 251, "correlation": 216316
    }
  },
  {
    "ph": "f", "id": 216316, "pid": 76337, "tid": -914061504, "ts": 1716454224753732,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753737, "dur": 0,
    "args": {
      "External id": 216318,
      "cbid": 251, "correlation": 216318
    }
  },
  {
    "ph": "f", "id": 216318, "pid": 76337, "tid": -914061504, "ts": 1716454224753737,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224840275, "dur": 11,
    "args": {
      "External id": 216319,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216319,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 216319, "pid": 5, "tid": 7, "ts": 1716454224840275,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753739, "dur": 11,
    "args": {
      "External id": 216319,
      "cbid": 211, "correlation": 216319
    }
  },
  {
    "ph": "s", "id": 216319, "pid": 76337, "tid": -914061504, "ts": 1716454224753739,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224840287, "dur": 5,
    "args": {
      "External id": 216321,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216321,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 216321, "pid": 5, "tid": 7, "ts": 1716454224840287,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753752, "dur": 6,
    "args": {
      "External id": 216321,
      "cbid": 211, "correlation": 216321
    }
  },
  {
    "ph": "s", "id": 216321, "pid": 76337, "tid": -914061504, "ts": 1716454224753752,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224840294, "dur": 53,
    "args": {
      "External id": 216331,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216331,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216331, "pid": 5, "tid": 7, "ts": 1716454224840294,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753811, "dur": 12,
    "args": {
      "External id": 216331,
      "cbid": 211, "correlation": 216331
    }
  },
  {
    "ph": "s", "id": 216331, "pid": 76337, "tid": -914061504, "ts": 1716454224753811,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224840348, "dur": 51,
    "args": {
      "External id": 216351,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216351,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 216351, "pid": 5, "tid": 7, "ts": 1716454224840348,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753877, "dur": 11,
    "args": {
      "External id": 216351,
      "cbid": 211, "correlation": 216351
    }
  },
  {
    "ph": "s", "id": 216351, "pid": 76337, "tid": -914061504, "ts": 1716454224753877,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224840401, "dur": 4,
    "args": {
      "External id": 216363,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216363,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216363, "pid": 5, "tid": 7, "ts": 1716454224840401,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753898, "dur": 6,
    "args": {
      "External id": 216363,
      "cbid": 211, "correlation": 216363
    }
  },
  {
    "ph": "s", "id": 216363, "pid": 76337, "tid": -914061504, "ts": 1716454224753898,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224840406, "dur": 56,
    "args": {
      "External id": 216366,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216366,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216366, "pid": 5, "tid": 7, "ts": 1716454224840406,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753918, "dur": 6,
    "args": {
      "External id": 216366,
      "cbid": 211, "correlation": 216366
    }
  },
  {
    "ph": "s", "id": 216366, "pid": 76337, "tid": -914061504, "ts": 1716454224753918,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224840463, "dur": 36,
    "args": {
      "External id": 216375,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216375,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216375, "pid": 5, "tid": 7, "ts": 1716454224840463,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224753958, "dur": 10,
    "args": {
      "External id": 216375,
      "cbid": 211, "correlation": 216375
    }
  },
  {
    "ph": "s", "id": 216375, "pid": 76337, "tid": -914061504, "ts": 1716454224753958,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754031, "dur": 0,
    "args": {
      "External id": 216385,
      "cbid": 317, "correlation": 216385
    }
  },
  {
    "ph": "f", "id": 216385, "pid": 76337, "tid": -914061504, "ts": 1716454224754031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754032, "dur": 0,
    "args": {
      "External id": 216386,
      "cbid": 203, "correlation": 216386
    }
  },
  {
    "ph": "f", "id": 216386, "pid": 76337, "tid": -914061504, "ts": 1716454224754032,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754033, "dur": 0,
    "args": {
      "External id": 216387,
      "cbid": 205, "correlation": 216387
    }
  },
  {
    "ph": "f", "id": 216387, "pid": 76337, "tid": -914061504, "ts": 1716454224754033,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224840501, "dur": 40,
    "args": {
      "External id": 216391,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216391,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216391, "pid": 5, "tid": 7, "ts": 1716454224840501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754049, "dur": 13,
    "args": {
      "External id": 216391,
      "cbid": 211, "correlation": 216391
    }
  },
  {
    "ph": "s", "id": 216391, "pid": 76337, "tid": -914061504, "ts": 1716454224754049,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224840542, "dur": 14,
    "args": {
      "External id": 216393,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216393,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216393, "pid": 5, "tid": 7, "ts": 1716454224840542,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754064, "dur": 5,
    "args": {
      "External id": 216393,
      "cbid": 211, "correlation": 216393
    }
  },
  {
    "ph": "s", "id": 216393, "pid": 76337, "tid": -914061504, "ts": 1716454224754064,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224840558, "dur": 3,
    "args": {
      "External id": 216395,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216395,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216395, "pid": 5, "tid": 7, "ts": 1716454224840558,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754073, "dur": 5,
    "args": {
      "External id": 216395,
      "cbid": 211, "correlation": 216395
    }
  },
  {
    "ph": "s", "id": 216395, "pid": 76337, "tid": -914061504, "ts": 1716454224754073,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754082, "dur": 0,
    "args": {
      "External id": 216396,
      "cbid": 51, "correlation": 216396
    }
  },
  {
    "ph": "s", "id": 216396, "pid": 76337, "tid": -914061504, "ts": 1716454224754082,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224840562, "dur": 704,
    "args": {
      "External id": 216397,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216397,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216397, "pid": 5, "tid": 7, "ts": 1716454224840562,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754083, "dur": 5,
    "args": {
      "External id": 216397,
      "cbid": 211, "correlation": 216397
    }
  },
  {
    "ph": "s", "id": 216397, "pid": 76337, "tid": -914061504, "ts": 1716454224754083,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224841267, "dur": 60,
    "args": {
      "External id": 216402,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216402,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216402, "pid": 5, "tid": 7, "ts": 1716454224841267,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754111, "dur": 9,
    "args": {
      "External id": 216402,
      "cbid": 211, "correlation": 216402
    }
  },
  {
    "ph": "s", "id": 216402, "pid": 76337, "tid": -914061504, "ts": 1716454224754111,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754169, "dur": 0,
    "args": {
      "External id": 216412,
      "cbid": 317, "correlation": 216412
    }
  },
  {
    "ph": "f", "id": 216412, "pid": 76337, "tid": -914061504, "ts": 1716454224754169,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754170, "dur": 0,
    "args": {
      "External id": 216413,
      "cbid": 203, "correlation": 216413
    }
  },
  {
    "ph": "f", "id": 216413, "pid": 76337, "tid": -914061504, "ts": 1716454224754170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754170, "dur": 0,
    "args": {
      "External id": 216414,
      "cbid": 205, "correlation": 216414
    }
  },
  {
    "ph": "f", "id": 216414, "pid": 76337, "tid": -914061504, "ts": 1716454224754170,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224841329, "dur": 3,
    "args": {
      "External id": 216418,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216418,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216418, "pid": 5, "tid": 7, "ts": 1716454224841329,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754186, "dur": 11,
    "args": {
      "External id": 216418,
      "cbid": 211, "correlation": 216418
    }
  },
  {
    "ph": "s", "id": 216418, "pid": 76337, "tid": -914061504, "ts": 1716454224754186,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754202, "dur": 0,
    "args": {
      "External id": 216419,
      "cbid": 51, "correlation": 216419
    }
  },
  {
    "ph": "s", "id": 216419, "pid": 76337, "tid": -914061504, "ts": 1716454224754202,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_256x128_ldg8_relu_filter1x1_stg8_interior_nchw_nn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224841334, "dur": 264,
    "args": {
      "External id": 216420,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216420,
      "registers per thread": 226,
      "shared memory": 49152,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [96, 3, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216420, "pid": 5, "tid": 7, "ts": 1716454224841334,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754203, "dur": 7,
    "args": {
      "External id": 216420,
      "cbid": 211, "correlation": 216420
    }
  },
  {
    "ph": "s", "id": 216420, "pid": 76337, "tid": -914061504, "ts": 1716454224754203,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224841599, "dur": 60,
    "args": {
      "External id": 216425,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216425,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216425, "pid": 5, "tid": 7, "ts": 1716454224841599,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754231, "dur": 8,
    "args": {
      "External id": 216425,
      "cbid": 211, "correlation": 216425
    }
  },
  {
    "ph": "s", "id": 216425, "pid": 76337, "tid": -914061504, "ts": 1716454224754231,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224841660, "dur": 50,
    "args": {
      "External id": 216433,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216433,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216433, "pid": 5, "tid": 7, "ts": 1716454224841660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754259, "dur": 8,
    "args": {
      "External id": 216433,
      "cbid": 211, "correlation": 216433
    }
  },
  {
    "ph": "s", "id": 216433, "pid": 76337, "tid": -914061504, "ts": 1716454224754259,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224841711, "dur": 35,
    "args": {
      "External id": 216441,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216441,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216441, "pid": 5, "tid": 7, "ts": 1716454224841711,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754288, "dur": 8,
    "args": {
      "External id": 216441,
      "cbid": 211, "correlation": 216441
    }
  },
  {
    "ph": "s", "id": 216441, "pid": 76337, "tid": -914061504, "ts": 1716454224754288,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224841748, "dur": 53,
    "args": {
      "External id": 216461,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216461,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 216461, "pid": 5, "tid": 7, "ts": 1716454224841748,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754371, "dur": 12,
    "args": {
      "External id": 216461,
      "cbid": 211, "correlation": 216461
    }
  },
  {
    "ph": "s", "id": 216461, "pid": 76337, "tid": -914061504, "ts": 1716454224754371,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224841803, "dur": 4,
    "args": {
      "External id": 216473,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216473,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216473, "pid": 5, "tid": 7, "ts": 1716454224841803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754393, "dur": 6,
    "args": {
      "External id": 216473,
      "cbid": 211, "correlation": 216473
    }
  },
  {
    "ph": "s", "id": 216473, "pid": 76337, "tid": -914061504, "ts": 1716454224754393,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224841808, "dur": 55,
    "args": {
      "External id": 216476,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216476,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216476, "pid": 5, "tid": 7, "ts": 1716454224841808,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754410, "dur": 7,
    "args": {
      "External id": 216476,
      "cbid": 211, "correlation": 216476
    }
  },
  {
    "ph": "s", "id": 216476, "pid": 76337, "tid": -914061504, "ts": 1716454224754410,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754468, "dur": 0,
    "args": {
      "External id": 216487,
      "cbid": 317, "correlation": 216487
    }
  },
  {
    "ph": "f", "id": 216487, "pid": 76337, "tid": -914061504, "ts": 1716454224754468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754468, "dur": 0,
    "args": {
      "External id": 216488,
      "cbid": 203, "correlation": 216488
    }
  },
  {
    "ph": "f", "id": 216488, "pid": 76337, "tid": -914061504, "ts": 1716454224754468,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754469, "dur": 0,
    "args": {
      "External id": 216489,
      "cbid": 205, "correlation": 216489
    }
  },
  {
    "ph": "f", "id": 216489, "pid": 76337, "tid": -914061504, "ts": 1716454224754469,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754497, "dur": 2,
    "args": {
      "External id": 216493,
      "cbid": 251, "correlation": 216493
    }
  },
  {
    "ph": "f", "id": 216493, "pid": 76337, "tid": -914061504, "ts": 1716454224754497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754500, "dur": 0,
    "args": {
      "External id": 216494,
      "cbid": 251, "correlation": 216494
    }
  },
  {
    "ph": "f", "id": 216494, "pid": 76337, "tid": -914061504, "ts": 1716454224754500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754501, "dur": 0,
    "args": {
      "External id": 216495,
      "cbid": 251, "correlation": 216495
    }
  },
  {
    "ph": "f", "id": 216495, "pid": 76337, "tid": -914061504, "ts": 1716454224754501,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754503, "dur": 0,
    "args": {
      "External id": 216496,
      "cbid": 251, "correlation": 216496
    }
  },
  {
    "ph": "f", "id": 216496, "pid": 76337, "tid": -914061504, "ts": 1716454224754503,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754504, "dur": 1,
    "args": {
      "External id": 216497,
      "cbid": 251, "correlation": 216497
    }
  },
  {
    "ph": "f", "id": 216497, "pid": 76337, "tid": -914061504, "ts": 1716454224754504,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754506, "dur": 1,
    "args": {
      "External id": 216498,
      "cbid": 251, "correlation": 216498
    }
  },
  {
    "ph": "f", "id": 216498, "pid": 76337, "tid": -914061504, "ts": 1716454224754506,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754508, "dur": 1,
    "args": {
      "External id": 216499,
      "cbid": 251, "correlation": 216499
    }
  },
  {
    "ph": "f", "id": 216499, "pid": 76337, "tid": -914061504, "ts": 1716454224754508,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754509, "dur": 1,
    "args": {
      "External id": 216500,
      "cbid": 251, "correlation": 216500
    }
  },
  {
    "ph": "f", "id": 216500, "pid": 76337, "tid": -914061504, "ts": 1716454224754509,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754511, "dur": 0,
    "args": {
      "External id": 216501,
      "cbid": 251, "correlation": 216501
    }
  },
  {
    "ph": "f", "id": 216501, "pid": 76337, "tid": -914061504, "ts": 1716454224754511,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224841863, "dur": 115,
    "args": {
      "External id": 216502,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216502,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 216502, "pid": 5, "tid": 7, "ts": 1716454224841863,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754514, "dur": 13,
    "args": {
      "External id": 216502,
      "cbid": 211, "correlation": 216502
    }
  },
  {
    "ph": "s", "id": 216502, "pid": 76337, "tid": -914061504, "ts": 1716454224754514,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224841979, "dur": 59,
    "args": {
      "External id": 216508,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216508,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216508, "pid": 5, "tid": 7, "ts": 1716454224841979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754549, "dur": 9,
    "args": {
      "External id": 216508,
      "cbid": 211, "correlation": 216508
    }
  },
  {
    "ph": "s", "id": 216508, "pid": 76337, "tid": -914061504, "ts": 1716454224754549,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 3, 64, 64>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 64, 64>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224842040, "dur": 483,
    "args": {
      "External id": 216517,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216517,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216517, "pid": 5, "tid": 7, "ts": 1716454224842040,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754633, "dur": 14,
    "args": {
      "External id": 216517,
      "cbid": 211, "correlation": 216517
    }
  },
  {
    "ph": "s", "id": 216517, "pid": 76337, "tid": -914061504, "ts": 1716454224754633,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224842524, "dur": 182,
    "args": {
      "External id": 216539,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216539,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 614.4,
      "warps per SM": 2457.6,
      "grid": [49152, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216539, "pid": 5, "tid": 7, "ts": 1716454224842524,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754691, "dur": 10,
    "args": {
      "External id": 216539,
      "cbid": 211, "correlation": 216539
    }
  },
  {
    "ph": "s", "id": 216539, "pid": 76337, "tid": -914061504, "ts": 1716454224754691,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754780, "dur": 1,
    "args": {
      "External id": 216550,
      "cbid": 251, "correlation": 216550
    }
  },
  {
    "ph": "f", "id": 216550, "pid": 76337, "tid": -914061504, "ts": 1716454224754780,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224842708, "dur": 197,
    "args": {
      "External id": 216551,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216551,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216551, "pid": 5, "tid": 7, "ts": 1716454224842708,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754785, "dur": 13,
    "args": {
      "External id": 216551,
      "cbid": 211, "correlation": 216551
    }
  },
  {
    "ph": "s", "id": 216551, "pid": 76337, "tid": -914061504, "ts": 1716454224754785,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754853, "dur": 1,
    "args": {
      "External id": 216562,
      "cbid": 251, "correlation": 216562
    }
  },
  {
    "ph": "f", "id": 216562, "pid": 76337, "tid": -914061504, "ts": 1716454224754853,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224842906, "dur": 189,
    "args": {
      "External id": 216563,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216563,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216563, "pid": 5, "tid": 7, "ts": 1716454224842906,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754857, "dur": 12,
    "args": {
      "External id": 216563,
      "cbid": 211, "correlation": 216563
    }
  },
  {
    "ph": "s", "id": 216563, "pid": 76337, "tid": -914061504, "ts": 1716454224754857,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754920, "dur": 1,
    "args": {
      "External id": 216574,
      "cbid": 251, "correlation": 216574
    }
  },
  {
    "ph": "f", "id": 216574, "pid": 76337, "tid": -914061504, "ts": 1716454224754920,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224843096, "dur": 187,
    "args": {
      "External id": 216575,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216575,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216575, "pid": 5, "tid": 7, "ts": 1716454224843096,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224754925, "dur": 12,
    "args": {
      "External id": 216575,
      "cbid": 211, "correlation": 216575
    }
  },
  {
    "ph": "s", "id": 216575, "pid": 76337, "tid": -914061504, "ts": 1716454224754925,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224843284, "dur": 18744,
    "args": {
      "External id": 216596,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216596,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 76.8,
      "warps per SM": 307.2,
      "grid": [96, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 216596, "pid": 5, "tid": 7, "ts": 1716454224843284,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755018, "dur": 14,
    "args": {
      "External id": 216596,
      "cbid": 211, "correlation": 216596
    }
  },
  {
    "ph": "s", "id": 216596, "pid": 76337, "tid": -914061504, "ts": 1716454224755018,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755117, "dur": 1,
    "args": {
      "External id": 216614,
      "cbid": 251, "correlation": 216614
    }
  },
  {
    "ph": "f", "id": 216614, "pid": 76337, "tid": -914061504, "ts": 1716454224755117,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224862030, "dur": 206,
    "args": {
      "External id": 216616,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216616,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [3, 384, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216616, "pid": 5, "tid": 7, "ts": 1716454224862030,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755122, "dur": 13,
    "args": {
      "External id": 216616,
      "cbid": 211, "correlation": 216616
    }
  },
  {
    "ph": "s", "id": 216616, "pid": 76337, "tid": -914061504, "ts": 1716454224755122,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224862237, "dur": 66,
    "args": {
      "External id": 216624,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216624,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216624, "pid": 5, "tid": 7, "ts": 1716454224862237,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755191, "dur": 12,
    "args": {
      "External id": 216624,
      "cbid": 211, "correlation": 216624
    }
  },
  {
    "ph": "s", "id": 216624, "pid": 76337, "tid": -914061504, "ts": 1716454224755191,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224862305, "dur": 97,
    "args": {
      "External id": 216632,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216632,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216632, "pid": 5, "tid": 7, "ts": 1716454224862305,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755230, "dur": 8,
    "args": {
      "External id": 216632,
      "cbid": 211, "correlation": 216632
    }
  },
  {
    "ph": "s", "id": 216632, "pid": 76337, "tid": -914061504, "ts": 1716454224755230,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224862403, "dur": 54,
    "args": {
      "External id": 216643,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216643,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216643, "pid": 5, "tid": 7, "ts": 1716454224862403,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755302, "dur": 13,
    "args": {
      "External id": 216643,
      "cbid": 211, "correlation": 216643
    }
  },
  {
    "ph": "s", "id": 216643, "pid": 76337, "tid": -914061504, "ts": 1716454224755302,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224862459, "dur": 92,
    "args": {
      "External id": 216665,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216665,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216665, "pid": 5, "tid": 7, "ts": 1716454224862459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755333, "dur": 8,
    "args": {
      "External id": 216665,
      "cbid": 211, "correlation": 216665
    }
  },
  {
    "ph": "s", "id": 216665, "pid": 76337, "tid": -914061504, "ts": 1716454224755333,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755418, "dur": 1,
    "args": {
      "External id": 216676,
      "cbid": 251, "correlation": 216676
    }
  },
  {
    "ph": "f", "id": 216676, "pid": 76337, "tid": -914061504, "ts": 1716454224755418,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224862553, "dur": 106,
    "args": {
      "External id": 216677,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216677,
      "registers per thread": 250,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216677, "pid": 5, "tid": 7, "ts": 1716454224862553,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755424, "dur": 12,
    "args": {
      "External id": 216677,
      "cbid": 211, "correlation": 216677
    }
  },
  {
    "ph": "s", "id": 216677, "pid": 76337, "tid": -914061504, "ts": 1716454224755424,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755497, "dur": 1,
    "args": {
      "External id": 216688,
      "cbid": 251, "correlation": 216688
    }
  },
  {
    "ph": "f", "id": 216688, "pid": 76337, "tid": -914061504, "ts": 1716454224755497,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755500, "dur": 0,
    "args": {
      "External id": 216689,
      "cbid": 251, "correlation": 216689
    }
  },
  {
    "ph": "f", "id": 216689, "pid": 76337, "tid": -914061504, "ts": 1716454224755500,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224862660, "dur": 10,
    "args": {
      "External id": 216690,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216690,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 216690, "pid": 5, "tid": 7, "ts": 1716454224862660,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755502, "dur": 13,
    "args": {
      "External id": 216690,
      "cbid": 211, "correlation": 216690
    }
  },
  {
    "ph": "s", "id": 216690, "pid": 76337, "tid": -914061504, "ts": 1716454224755502,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224862672, "dur": 5,
    "args": {
      "External id": 216692,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216692,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 216692, "pid": 5, "tid": 7, "ts": 1716454224862672,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755519, "dur": 7,
    "args": {
      "External id": 216692,
      "cbid": 211, "correlation": 216692
    }
  },
  {
    "ph": "s", "id": 216692, "pid": 76337, "tid": -914061504, "ts": 1716454224755519,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755581, "dur": 1,
    "args": {
      "External id": 216703,
      "cbid": 251, "correlation": 216703
    }
  },
  {
    "ph": "f", "id": 216703, "pid": 76337, "tid": -914061504, "ts": 1716454224755581,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755584, "dur": 0,
    "args": {
      "External id": 216704,
      "cbid": 251, "correlation": 216704
    }
  },
  {
    "ph": "f", "id": 216704, "pid": 76337, "tid": -914061504, "ts": 1716454224755584,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_f16_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224862678, "dur": 6,
    "args": {
      "External id": 216705,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216705,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 1.2,
      "warps per SM": 4.8,
      "grid": [8, 2, 6],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 8
    }
  },
  {
    "ph": "f", "id": 216705, "pid": 5, "tid": 7, "ts": 1716454224862678,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755586, "dur": 12,
    "args": {
      "External id": 216705,
      "cbid": 211, "correlation": 216705
    }
  },
  {
    "ph": "s", "id": 216705, "pid": 76337, "tid": -914061504, "ts": 1716454224755586,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, __half, __half, float, __half, true, false, false>(cublasSplitKParams<float>, __half const*, __half const*, __half*, float const*, float const*, __half const*, __half const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224862685, "dur": 3,
    "args": {
      "External id": 216707,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216707,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 216707, "pid": 5, "tid": 7, "ts": 1716454224862685,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755599, "dur": 5,
    "args": {
      "External id": 216707,
      "cbid": 211, "correlation": 216707
    }
  },
  {
    "ph": "s", "id": 216707, "pid": 76337, "tid": -914061504, "ts": 1716454224755599,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void attention_kernel_batched<AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true> >(AttentionKernel<cutlass::half_t, cutlass::arch::Sm70, true, 64l, 64l, true>::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224862690, "dur": 157,
    "args": {
      "External id": 216728,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216728,
      "registers per thread": 168,
      "shared memory": 17152,
      "blocks per SM": 38.4,
      "warps per SM": 153.6,
      "grid": [48, 8, 8],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 216728, "pid": 5, "tid": 7, "ts": 1716454224862690,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755672, "dur": 12,
    "args": {
      "External id": 216728,
      "cbid": 211, "correlation": 216728
    }
  },
  {
    "ph": "s", "id": 216728, "pid": 76337, "tid": -914061504, "ts": 1716454224755672,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755767, "dur": 2,
    "args": {
      "External id": 216746,
      "cbid": 251, "correlation": 216746
    }
  },
  {
    "ph": "f", "id": 216746, "pid": 76337, "tid": -914061504, "ts": 1716454224755767,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224862849, "dur": 107,
    "args": {
      "External id": 216748,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216748,
      "registers per thread": 142,
      "shared memory": 25088,
      "blocks per SM": 12,
      "warps per SM": 48,
      "grid": [5, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 216748, "pid": 5, "tid": 7, "ts": 1716454224862849,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755774, "dur": 13,
    "args": {
      "External id": 216748,
      "cbid": 211, "correlation": 216748
    }
  },
  {
    "ph": "s", "id": 216748, "pid": 76337, "tid": -914061504, "ts": 1716454224755774,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224862957, "dur": 35,
    "args": {
      "External id": 216756,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216756,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216756, "pid": 5, "tid": 7, "ts": 1716454224862957,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755842, "dur": 12,
    "args": {
      "External id": 216756,
      "cbid": 211, "correlation": 216756
    }
  },
  {
    "ph": "s", "id": 216756, "pid": 76337, "tid": -914061504, "ts": 1716454224755842,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224862993, "dur": 66,
    "args": {
      "External id": 216764,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216764,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216764, "pid": 5, "tid": 7, "ts": 1716454224862993,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755883, "dur": 9,
    "args": {
      "External id": 216764,
      "cbid": 211, "correlation": 216764
    }
  },
  {
    "ph": "s", "id": 216764, "pid": 76337, "tid": -914061504, "ts": 1716454224755883,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::vectorized_layer_norm_kernel<c10::Half, float>(int, float, c10::Half const*, c10::Half const*, c10::Half const*, float*, float*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224863060, "dur": 92,
    "args": {
      "External id": 216786,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216786,
      "registers per thread": 32,
      "shared memory": 24,
      "blocks per SM": 307.2,
      "warps per SM": 1228.8,
      "grid": [24576, 1, 1],
      "block": [32, 4, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216786, "pid": 5, "tid": 7, "ts": 1716454224863060,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224755935, "dur": 10,
    "args": {
      "External id": 216786,
      "cbid": 211, "correlation": 216786
    }
  },
  {
    "ph": "s", "id": 216786, "pid": 76337, "tid": -914061504, "ts": 1716454224755935,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756029, "dur": 1,
    "args": {
      "External id": 216802,
      "cbid": 251, "correlation": 216802
    }
  },
  {
    "ph": "f", "id": 216802, "pid": 76337, "tid": -914061504, "ts": 1716454224756029,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x128_ldg8_relu_f2f_stages_32x1_tn", "pid": 5, "tid": 7,
    "ts": 1716454224863154, "dur": 578,
    "args": {
      "External id": 216804,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216804,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 48,
      "warps per SM": 192,
      "grid": [20, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216804, "pid": 5, "tid": 7, "ts": 1716454224863154,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756036, "dur": 13,
    "args": {
      "External id": 216804,
      "cbid": 211, "correlation": 216804
    }
  },
  {
    "ph": "s", "id": 216804, "pid": 76337, "tid": -914061504, "ts": 1716454224756036,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::GeluCUDAKernelImpl(at::TensorIteratorBase&, at::native::GeluType)::{lambda()#2}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224863734, "dur": 244,
    "args": {
      "External id": 216812,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216812,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216812, "pid": 5, "tid": 7, "ts": 1716454224863734,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756101, "dur": 12,
    "args": {
      "External id": 216812,
      "cbid": 211, "correlation": 216812
    }
  },
  {
    "ph": "s", "id": 216812, "pid": 76337, "tid": -914061504, "ts": 1716454224756101,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > >(at::TensorIteratorBase&, at::native::BinaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> > const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224863979, "dur": 252,
    "args": {
      "External id": 216820,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216820,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 768,
      "warps per SM": 3072,
      "grid": [61440, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216820, "pid": 5, "tid": 7, "ts": 1716454224863979,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756133, "dur": 8,
    "args": {
      "External id": 216820,
      "cbid": 211, "correlation": 216820
    }
  },
  {
    "ph": "s", "id": 216820, "pid": 76337, "tid": -914061504, "ts": 1716454224756133,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756214, "dur": 1,
    "args": {
      "External id": 216836,
      "cbid": 251, "correlation": 216836
    }
  },
  {
    "ph": "f", "id": 216836, "pid": 76337, "tid": -914061504, "ts": 1716454224756214,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756219, "dur": 0,
    "args": {
      "External id": 216838,
      "cbid": 251, "correlation": 216838
    }
  },
  {
    "ph": "f", "id": 216838, "pid": 76337, "tid": -914061504, "ts": 1716454224756219,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_128x256_ldg8_relu_f2f_tn", "pid": 5, "tid": 7,
    "ts": 1716454224864233, "dur": 358,
    "args": {
      "External id": 216839,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216839,
      "registers per thread": 254,
      "shared memory": 65536,
      "blocks per SM": 3.6,
      "warps per SM": 28.8,
      "grid": [3, 96, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 0
    }
  },
  {
    "ph": "f", "id": 216839, "pid": 5, "tid": 7, "ts": 1716454224864233,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756221, "dur": 13,
    "args": {
      "External id": 216839,
      "cbid": 211, "correlation": 216839
    }
  },
  {
    "ph": "s", "id": 216839, "pid": 76337, "tid": -914061504, "ts": 1716454224756221,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224864592, "dur": 50,
    "args": {
      "External id": 216847,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216847,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216847, "pid": 5, "tid": 7, "ts": 1716454224864592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756264, "dur": 10,
    "args": {
      "External id": 216847,
      "cbid": 211, "correlation": 216847
    }
  },
  {
    "ph": "s", "id": 216847, "pid": 76337, "tid": -914061504, "ts": 1716454224756264,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1}>(at::TensorIteratorBase&, at::native::direct_copy_kernel_cuda(at::TensorIteratorBase&)::{lambda()#2}::operator()() const::{lambda()#10}::operator()() const::{lambda(c10::Half)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224864643, "dur": 159,
    "args": {
      "External id": 216858,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216858,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216858, "pid": 5, "tid": 7, "ts": 1716454224864643,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756332, "dur": 12,
    "args": {
      "External id": 216858,
      "cbid": 211, "correlation": 216858
    }
  },
  {
    "ph": "s", "id": 216858, "pid": 76337, "tid": -914061504, "ts": 1716454224756332,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756397, "dur": 0,
    "args": {
      "External id": 216870,
      "cbid": 317, "correlation": 216870
    }
  },
  {
    "ph": "f", "id": 216870, "pid": 76337, "tid": -914061504, "ts": 1716454224756397,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756398, "dur": 0,
    "args": {
      "External id": 216871,
      "cbid": 203, "correlation": 216871
    }
  },
  {
    "ph": "f", "id": 216871, "pid": 76337, "tid": -914061504, "ts": 1716454224756398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756399, "dur": 0,
    "args": {
      "External id": 216872,
      "cbid": 205, "correlation": 216872
    }
  },
  {
    "ph": "f", "id": 216872, "pid": 76337, "tid": -914061504, "ts": 1716454224756399,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756422, "dur": 1,
    "args": {
      "External id": 216876,
      "cbid": 251, "correlation": 216876
    }
  },
  {
    "ph": "f", "id": 216876, "pid": 76337, "tid": -914061504, "ts": 1716454224756422,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756424, "dur": 0,
    "args": {
      "External id": 216877,
      "cbid": 251, "correlation": 216877
    }
  },
  {
    "ph": "f", "id": 216877, "pid": 76337, "tid": -914061504, "ts": 1716454224756424,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756425, "dur": 0,
    "args": {
      "External id": 216878,
      "cbid": 251, "correlation": 216878
    }
  },
  {
    "ph": "f", "id": 216878, "pid": 76337, "tid": -914061504, "ts": 1716454224756425,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756426, "dur": 0,
    "args": {
      "External id": 216879,
      "cbid": 251, "correlation": 216879
    }
  },
  {
    "ph": "f", "id": 216879, "pid": 76337, "tid": -914061504, "ts": 1716454224756426,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756427, "dur": 0,
    "args": {
      "External id": 216880,
      "cbid": 251, "correlation": 216880
    }
  },
  {
    "ph": "f", "id": 216880, "pid": 76337, "tid": -914061504, "ts": 1716454224756427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756427, "dur": 0,
    "args": {
      "External id": 216881,
      "cbid": 251, "correlation": 216881
    }
  },
  {
    "ph": "f", "id": 216881, "pid": 76337, "tid": -914061504, "ts": 1716454224756427,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756428, "dur": 0,
    "args": {
      "External id": 216882,
      "cbid": 251, "correlation": 216882
    }
  },
  {
    "ph": "f", "id": 216882, "pid": 76337, "tid": -914061504, "ts": 1716454224756428,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756429, "dur": 0,
    "args": {
      "External id": 216883,
      "cbid": 251, "correlation": 216883
    }
  },
  {
    "ph": "f", "id": 216883, "pid": 76337, "tid": -914061504, "ts": 1716454224756429,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756431, "dur": 0,
    "args": {
      "External id": 216884,
      "cbid": 251, "correlation": 216884
    }
  },
  {
    "ph": "f", "id": 216884, "pid": 76337, "tid": -914061504, "ts": 1716454224756431,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "volta_fp16_s884gemm_fp16_64x128_ldg8_f2f_nn", "pid": 5, "tid": 7,
    "ts": 1716454224864803, "dur": 114,
    "args": {
      "External id": 216885,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216885,
      "registers per thread": 138,
      "shared memory": 24832,
      "blocks per SM": 14.4,
      "warps per SM": 57.6,
      "grid": [48, 3, 8],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 19
    }
  },
  {
    "ph": "f", "id": 216885, "pid": 5, "tid": 7, "ts": 1716454224864803,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756433, "dur": 12,
    "args": {
      "External id": 216885,
      "cbid": 211, "correlation": 216885
    }
  },
  {
    "ph": "s", "id": 216885, "pid": 76337, "tid": -914061504, "ts": 1716454224756433,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224864919, "dur": 60,
    "args": {
      "External id": 216891,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216891,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216891, "pid": 5, "tid": 7, "ts": 1716454224864919,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756468, "dur": 9,
    "args": {
      "External id": 216891,
      "cbid": 211, "correlation": 216891
    }
  },
  {
    "ph": "s", "id": 216891, "pid": 76337, "tid": -914061504, "ts": 1716454224756468,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224864980, "dur": 50,
    "args": {
      "External id": 216899,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216899,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216899, "pid": 5, "tid": 7, "ts": 1716454224864980,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756501, "dur": 8,
    "args": {
      "External id": 216899,
      "cbid": 211, "correlation": 216899
    }
  },
  {
    "ph": "s", "id": 216899, "pid": 76337, "tid": -914061504, "ts": 1716454224756501,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::CatArrayBatchedCopy<c10::Half, unsigned int, 4, 128, 1>(c10::Half*, at::native::(anonymous namespace)::CatArrInputTensorMetadata<c10::Half, unsigned int, 128, 1>, at::native::(anonymous namespace)::TensorSizeStride<unsigned int, 4u>, int, unsigned int)", "pid": 5, "tid": 7,
    "ts": 1716454224865031, "dur": 97,
    "args": {
      "External id": 216908,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216908,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 4,
      "warps per SM": 64,
      "grid": [160, 2, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216908, "pid": 5, "tid": 7, "ts": 1716454224865031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756540, "dur": 10,
    "args": {
      "External id": 216908,
      "cbid": 211, "correlation": 216908
    }
  },
  {
    "ph": "s", "id": 216908, "pid": 76337, "tid": -914061504, "ts": 1716454224756540,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224865129, "dur": 92,
    "args": {
      "External id": 216928,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216928,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 216928, "pid": 5, "tid": 7, "ts": 1716454224865129,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756609, "dur": 11,
    "args": {
      "External id": 216928,
      "cbid": 211, "correlation": 216928
    }
  },
  {
    "ph": "s", "id": 216928, "pid": 76337, "tid": -914061504, "ts": 1716454224756609,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224865223, "dur": 4,
    "args": {
      "External id": 216940,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216940,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 2,
      "grid": [20, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 216940, "pid": 5, "tid": 7, "ts": 1716454224865223,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756630, "dur": 75,
    "args": {
      "External id": 216940,
      "cbid": 211, "correlation": 216940
    }
  },
  {
    "ph": "s", "id": 216940, "pid": 76337, "tid": -914061504, "ts": 1716454224756630,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224865228, "dur": 109,
    "args": {
      "External id": 216943,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216943,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216943, "pid": 5, "tid": 7, "ts": 1716454224865228,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756716, "dur": 7,
    "args": {
      "External id": 216943,
      "cbid": 211, "correlation": 216943
    }
  },
  {
    "ph": "s", "id": 216943, "pid": 76337, "tid": -914061504, "ts": 1716454224756716,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224865339, "dur": 69,
    "args": {
      "External id": 216952,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216952,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 384,
      "warps per SM": 1536,
      "grid": [30720, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216952, "pid": 5, "tid": 7, "ts": 1716454224865339,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756758, "dur": 10,
    "args": {
      "External id": 216952,
      "cbid": 211, "correlation": 216952
    }
  },
  {
    "ph": "s", "id": 216952, "pid": 76337, "tid": -914061504, "ts": 1716454224756758,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756810, "dur": 0,
    "args": {
      "External id": 216962,
      "cbid": 317, "correlation": 216962
    }
  },
  {
    "ph": "f", "id": 216962, "pid": 76337, "tid": -914061504, "ts": 1716454224756810,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756811, "dur": 0,
    "args": {
      "External id": 216963,
      "cbid": 203, "correlation": 216963
    }
  },
  {
    "ph": "f", "id": 216963, "pid": 76337, "tid": -914061504, "ts": 1716454224756811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756811, "dur": 0,
    "args": {
      "External id": 216964,
      "cbid": 205, "correlation": 216964
    }
  },
  {
    "ph": "f", "id": 216964, "pid": 76337, "tid": -914061504, "ts": 1716454224756811,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224865410, "dur": 77,
    "args": {
      "External id": 216968,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216968,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216968, "pid": 5, "tid": 7, "ts": 1716454224865410,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756827, "dur": 11,
    "args": {
      "External id": 216968,
      "cbid": 211, "correlation": 216968
    }
  },
  {
    "ph": "s", "id": 216968, "pid": 76337, "tid": -914061504, "ts": 1716454224756827,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224865488, "dur": 24,
    "args": {
      "External id": 216970,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216970,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 80,
      "warps per SM": 640,
      "grid": [1, 20, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216970, "pid": 5, "tid": 7, "ts": 1716454224865488,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756841, "dur": 6,
    "args": {
      "External id": 216970,
      "cbid": 211, "correlation": 216970
    }
  },
  {
    "ph": "s", "id": 216970, "pid": 76337, "tid": -914061504, "ts": 1716454224756841,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224865513, "dur": 4,
    "args": {
      "External id": 216972,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216972,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216972, "pid": 5, "tid": 7, "ts": 1716454224865513,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756853, "dur": 6,
    "args": {
      "External id": 216972,
      "cbid": 211, "correlation": 216972
    }
  },
  {
    "ph": "s", "id": 216972, "pid": 76337, "tid": -914061504, "ts": 1716454224756853,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756861, "dur": 0,
    "args": {
      "External id": 216973,
      "cbid": 51, "correlation": 216973
    }
  },
  {
    "ph": "s", "id": 216973, "pid": 76337, "tid": -914061504, "ts": 1716454224756861,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224865518, "dur": 1369,
    "args": {
      "External id": 216974,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216974,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 216974, "pid": 5, "tid": 7, "ts": 1716454224865518,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756862, "dur": 5,
    "args": {
      "External id": 216974,
      "cbid": 211, "correlation": 216974
    }
  },
  {
    "ph": "s", "id": 216974, "pid": 76337, "tid": -914061504, "ts": 1716454224756862,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224866889, "dur": 60,
    "args": {
      "External id": 216979,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216979,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 216979, "pid": 5, "tid": 7, "ts": 1716454224866889,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756889, "dur": 9,
    "args": {
      "External id": 216979,
      "cbid": 211, "correlation": 216979
    }
  },
  {
    "ph": "s", "id": 216979, "pid": 76337, "tid": -914061504, "ts": 1716454224756889,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224866950, "dur": 4,
    "args": {
      "External id": 216987,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 216987,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 0.25,
      "warps per SM": 1,
      "grid": [20, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 216987, "pid": 5, "tid": 7, "ts": 1716454224866950,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224756933, "dur": 9,
    "args": {
      "External id": 216987,
      "cbid": 211, "correlation": 216987
    }
  },
  {
    "ph": "s", "id": 216987, "pid": 76337, "tid": -914061504, "ts": 1716454224756933,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757005, "dur": 1,
    "args": {
      "External id": 217003,
      "cbid": 251, "correlation": 217003
    }
  },
  {
    "ph": "f", "id": 217003, "pid": 76337, "tid": -914061504, "ts": 1716454224757005,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757011, "dur": 0,
    "args": {
      "External id": 217005,
      "cbid": 251, "correlation": 217005
    }
  },
  {
    "ph": "f", "id": 217005, "pid": 76337, "tid": -914061504, "ts": 1716454224757011,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cutlass::Kernel<cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8>(cutlass_70_wmma_tensorop_s161616gemm_f16_32x32_64x2_tn_align8::Params)", "pid": 5, "tid": 7,
    "ts": 1716454224866955, "dur": 11,
    "args": {
      "External id": 217006,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217006,
      "registers per thread": 96,
      "shared memory": 17408,
      "blocks per SM": 0.8,
      "warps per SM": 3.2,
      "grid": [8, 2, 4],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 5
    }
  },
  {
    "ph": "f", "id": 217006, "pid": 5, "tid": 7, "ts": 1716454224866955,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757012, "dur": 12,
    "args": {
      "External id": 217006,
      "cbid": 211, "correlation": 217006
    }
  },
  {
    "ph": "s", "id": 217006, "pid": 76337, "tid": -914061504, "ts": 1716454224757012,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void splitKreduce_kernel<32, 16, int, float, __half, float, __half, true, true, false>(cublasSplitKParams<float>, float const*, __half const*, __half*, float const*, float const*, __half const*, float const*, __half*, void*, long, float*, int*)", "pid": 5, "tid": 7,
    "ts": 1716454224866967, "dur": 5,
    "args": {
      "External id": 217008,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217008,
      "registers per thread": 32,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 2,
      "grid": [10, 1, 1],
      "block": [32, 16, 1],
      "est. achieved occupancy %": 3
    }
  },
  {
    "ph": "f", "id": 217008, "pid": 5, "tid": 7, "ts": 1716454224866967,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757026, "dur": 170,
    "args": {
      "External id": 217008,
      "cbid": 211, "correlation": 217008
    }
  },
  {
    "ph": "s", "id": 217008, "pid": 76337, "tid": -914061504, "ts": 1716454224757026,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224866974, "dur": 56,
    "args": {
      "External id": 217018,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217018,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217018, "pid": 5, "tid": 7, "ts": 1716454224866974,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757250, "dur": 12,
    "args": {
      "External id": 217018,
      "cbid": 211, "correlation": 217018
    }
  },
  {
    "ph": "s", "id": 217018, "pid": 76337, "tid": -914061504, "ts": 1716454224757250,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::RowwiseMomentsCUDAKernel<c10::Half>(long, c10::Half, c10::Half const*, c10::Half*, c10::Half*)", "pid": 5, "tid": 7,
    "ts": 1716454224867031, "dur": 52,
    "args": {
      "External id": 217038,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217038,
      "registers per thread": 25,
      "shared memory": 768,
      "blocks per SM": 3.2,
      "warps per SM": 51.2,
      "grid": [256, 1, 1],
      "block": [512, 1, 1],
      "est. achieved occupancy %": 80
    }
  },
  {
    "ph": "f", "id": 217038, "pid": 5, "tid": 7, "ts": 1716454224867031,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757316, "dur": 11,
    "args": {
      "External id": 217038,
      "cbid": 211, "correlation": 217038
    }
  },
  {
    "ph": "s", "id": 217038, "pid": 76337, "tid": -914061504, "ts": 1716454224757316,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::(anonymous namespace)::ComputeFusedParamsCUDAKernel<c10::Half>(long, long, long, c10::Half const*, c10::Half const*, c10::Half const*, c10::Half const*, at::AccumulateType<c10::Half, true>::type*, at::AccumulateType<c10::Half, true>::type*)", "pid": 5, "tid": 7,
    "ts": 1716454224867084, "dur": 4,
    "args": {
      "External id": 217050,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217050,
      "registers per thread": 20,
      "shared memory": 0,
      "blocks per SM": 0.125,
      "warps per SM": 1,
      "grid": [10, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217050, "pid": 5, "tid": 7, "ts": 1716454224867084,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757336, "dur": 6,
    "args": {
      "External id": 217050,
      "cbid": 211, "correlation": 217050
    }
  },
  {
    "ph": "s", "id": 217050, "pid": 76337, "tid": -914061504, "ts": 1716454224757336,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1}>(at::TensorIteratorBase&, at::native::(anonymous namespace)::GroupNormKernelImplInternal<c10::Half>(at::Tensor const&, at::Tensor const&, at::Tensor const&, long, long, long, long, c10::Half, at::Tensor&, at::Tensor&, at::Tensor&)::{lambda(c10::Half, float, float)#1} const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224867089, "dur": 56,
    "args": {
      "External id": 217053,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217053,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217053, "pid": 5, "tid": 7, "ts": 1716454224867089,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757355, "dur": 7,
    "args": {
      "External id": 217053,
      "cbid": 211, "correlation": 217053
    }
  },
  {
    "ph": "s", "id": 217053, "pid": 76337, "tid": -914061504, "ts": 1716454224757355,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2> >(int, at::native::(anonymous namespace)::silu_kernel(at::TensorIteratorBase&)::{lambda()#1}::operator()() const::{lambda()#3}::operator()() const::{lambda(c10::Half)#1}, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224867147, "dur": 37,
    "args": {
      "External id": 217062,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217062,
      "registers per thread": 24,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217062, "pid": 5, "tid": 7, "ts": 1716454224867147,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757396, "dur": 10,
    "args": {
      "External id": 217062,
      "cbid": 211, "correlation": 217062
    }
  },
  {
    "ph": "s", "id": 217062, "pid": 76337, "tid": -914061504, "ts": 1716454224757396,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757458, "dur": 0,
    "args": {
      "External id": 217072,
      "cbid": 317, "correlation": 217072
    }
  },
  {
    "ph": "f", "id": 217072, "pid": 76337, "tid": -914061504, "ts": 1716454224757458,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757459, "dur": 0,
    "args": {
      "External id": 217073,
      "cbid": 203, "correlation": 217073
    }
  },
  {
    "ph": "f", "id": 217073, "pid": 76337, "tid": -914061504, "ts": 1716454224757459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757459, "dur": 0,
    "args": {
      "External id": 217074,
      "cbid": 205, "correlation": 217074
    }
  },
  {
    "ph": "f", "id": 217074, "pid": 76337, "tid": -914061504, "ts": 1716454224757459,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224867184, "dur": 40,
    "args": {
      "External id": 217078,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217078,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217078, "pid": 5, "tid": 7, "ts": 1716454224867184,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757475, "dur": 12,
    "args": {
      "External id": 217078,
      "cbid": 211, "correlation": 217078
    }
  },
  {
    "ph": "s", "id": 217078, "pid": 76337, "tid": -914061504, "ts": 1716454224757475,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224867226, "dur": 14,
    "args": {
      "External id": 217080,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217080,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 40,
      "warps per SM": 320,
      "grid": [1, 10, 320],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217080, "pid": 5, "tid": 7, "ts": 1716454224867226,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757489, "dur": 5,
    "args": {
      "External id": 217080,
      "cbid": 211, "correlation": 217080
    }
  },
  {
    "ph": "s", "id": 217080, "pid": 76337, "tid": -914061504, "ts": 1716454224757489,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cask_cudnn::computeOffsetsKernel<false, false>(cask_cudnn::ComputeOffsetsParams)", "pid": 5, "tid": 7,
    "ts": 1716454224867242, "dur": 3,
    "args": {
      "External id": 217082,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217082,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 0.1625,
      "warps per SM": 1.3,
      "grid": [13, 1, 1],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 2
    }
  },
  {
    "ph": "f", "id": 217082, "pid": 5, "tid": 7, "ts": 1716454224867242,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757498, "dur": 6,
    "args": {
      "External id": 217082,
      "cbid": 211, "correlation": 217082
    }
  },
  {
    "ph": "s", "id": 217082, "pid": 76337, "tid": -914061504, "ts": 1716454224757498,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaMemsetAsync", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757507, "dur": 0,
    "args": {
      "External id": 217083,
      "cbid": 51, "correlation": 217083
    }
  },
  {
    "ph": "s", "id": 217083, "pid": 76337, "tid": -914061504, "ts": 1716454224757507,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "cudnn_volta_fp16_s884cudnn_fp16_128x128_ldg8_relu_f2f_exp_medium_nhwc2nchw_tn_v1", "pid": 5, "tid": 7,
    "ts": 1716454224867246, "dur": 702,
    "args": {
      "External id": 217084,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217084,
      "registers per thread": 254,
      "shared memory": 33280,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [192, 3, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217084, "pid": 5, "tid": 7, "ts": 1716454224867246,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757508, "dur": 5,
    "args": {
      "External id": 217084,
      "cbid": 211, "correlation": 217084
    }
  },
  {
    "ph": "s", "id": 217084, "pid": 76337, "tid": -914061504, "ts": 1716454224757508,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224867949, "dur": 60,
    "args": {
      "External id": 217089,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217089,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217089, "pid": 5, "tid": 7, "ts": 1716454224867949,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757535, "dur": 8,
    "args": {
      "External id": 217089,
      "cbid": 211, "correlation": 217089
    }
  },
  {
    "ph": "s", "id": 217089, "pid": 76337, "tid": -914061504, "ts": 1716454224757535,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamIsCapturing", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757592, "dur": 0,
    "args": {
      "External id": 217099,
      "cbid": 317, "correlation": 217099
    }
  },
  {
    "ph": "f", "id": 217099, "pid": 76337, "tid": -914061504, "ts": 1716454224757592,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaStreamGetPriority", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757593, "dur": 0,
    "args": {
      "External id": 217100,
      "cbid": 203, "correlation": 217100
    }
  },
  {
    "ph": "f", "id": 217100, "pid": 76337, "tid": -914061504, "ts": 1716454224757593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaDeviceGetStreamPriorityRange", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757593, "dur": 0,
    "args": {
      "External id": 217101,
      "cbid": 205, "correlation": 217101
    }
  },
  {
    "ph": "f", "id": 217101, "pid": 76337, "tid": -914061504, "ts": 1716454224757593,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nchwToNhwcKernel<__half, __half, float, false, true, (cudnnKernelDataType_t)0>(cudnn::ops::nchw2nhwc_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224868010, "dur": 75,
    "args": {
      "External id": 217105,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217105,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 192,
      "warps per SM": 1536,
      "grid": [96, 20, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217105, "pid": 5, "tid": 7, "ts": 1716454224868010,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757606, "dur": 12,
    "args": {
      "External id": 217105,
      "cbid": 211, "correlation": 217105
    }
  },
  {
    "ph": "s", "id": 217105, "pid": 76337, "tid": -914061504, "ts": 1716454224757606,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "sm70_xmma_fprop_implicit_gemm_f16f16_f16f32_f32_nhwckrsc_nhwc_tilesize128x128x32_stage1_warpsize2x2x1_g1_tensor8x8x4_t1r1s1_execute_kernel_cudnn", "pid": 5, "tid": 7,
    "ts": 1716454224868086, "dur": 209,
    "args": {
      "External id": 217107,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217107,
      "registers per thread": 252,
      "shared memory": 17408,
      "blocks per SM": 7.2,
      "warps per SM": 28.8,
      "grid": [3, 192, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 13
    }
  },
  {
    "ph": "f", "id": 217107, "pid": 5, "tid": 7, "ts": 1716454224868086,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757625, "dur": 8,
    "args": {
      "External id": 217107,
      "cbid": 211, "correlation": 217107
    }
  },
  {
    "ph": "s", "id": 217107, "pid": 76337, "tid": -914061504, "ts": 1716454224757625,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void cudnn::ops::nhwcToNchwKernel<__half, __half, float, true, false, (cudnnKernelDataType_t)0>(cudnn::ops::nhwc2nchw_params_t<float>, __half const*, __half*)", "pid": 5, "tid": 7,
    "ts": 1716454224868297, "dur": 39,
    "args": {
      "External id": 217109,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217109,
      "registers per thread": 32,
      "shared memory": 2112,
      "blocks per SM": 96,
      "warps per SM": 768,
      "grid": [96, 10, 8],
      "block": [256, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217109, "pid": 5, "tid": 7, "ts": 1716454224868297,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224757638, "dur": 493,
    "args": {
      "External id": 217109,
      "cbid": 211, "correlation": 217109
    }
  },
  {
    "ph": "s", "id": 217109, "pid": 76337, "tid": -914061504, "ts": 1716454224757638,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::elementwise_kernel<128, 4, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1}>(int, at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<c10::Half> >(at::TensorIteratorBase&, at::native::CUDAFunctor_add<c10::Half> const&)::{lambda(int)#1})", "pid": 5, "tid": 7,
    "ts": 1716454224868337, "dur": 59,
    "args": {
      "External id": 217115,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217115,
      "registers per thread": 16,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217115, "pid": 5, "tid": 7, "ts": 1716454224868337,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758153, "dur": 10,
    "args": {
      "External id": 217115,
      "cbid": 211, "correlation": 217115
    }
  },
  {
    "ph": "s", "id": 217115, "pid": 76337, "tid": -914061504, "ts": 1716454224758153,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3> >(int, at::native::CUDAFunctor_add<c10::Half>, at::detail::Array<char*, 3>)", "pid": 5, "tid": 7,
    "ts": 1716454224868398, "dur": 50,
    "args": {
      "External id": 217123,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217123,
      "registers per thread": 19,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217123, "pid": 5, "tid": 7, "ts": 1716454224868398,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758184, "dur": 9,
    "args": {
      "External id": 217123,
      "cbid": 211, "correlation": 217123
    }
  },
  {
    "ph": "s", "id": 217123, "pid": 76337, "tid": -914061504, "ts": 1716454224758184,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "ph": "X", "cat": "kernel", "name": "void at::native::vectorized_elementwise_kernel<4, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2> >(int, at::native::BUnaryFunctor<c10::Half, c10::Half, c10::Half, at::native::binary_internal::MulFunctor<float> >, at::detail::Array<char*, 2>)", "pid": 5, "tid": 7,
    "ts": 1716454224868449, "dur": 36,
    "args": {
      "External id": 217131,
      "queued": 0, "device": 5, "context": 1,
      "stream": 7, "correlation": 217131,
      "registers per thread": 17,
      "shared memory": 0,
      "blocks per SM": 192,
      "warps per SM": 768,
      "grid": [15360, 1, 1],
      "block": [128, 1, 1],
      "est. achieved occupancy %": 100
    }
  },
  {
    "ph": "f", "id": 217131, "pid": 5, "tid": 7, "ts": 1716454224868449,
    "cat": "ac2g", "name": "ac2g", "bp": "e"
  },
  {
    "ph": "X", "cat": "cuda_runtime", "name": "cudaLaunchKernel", "pid": 76337, "tid": -914061504,
    "ts": 1716454224758213, "dur": 29,
    "args": {
      "External id": 217131,
      "cbid": 211, "correlation": 217131
    }
  },
  {
    "ph": "s", "id": 217131, "pid": 76337, "tid": -914061504, "ts": 1716454224758213,
    "cat": "ac2g", "name": "ac2g"
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 76337, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 76337, "tid": 0,
    "args": {
      "labels": "CPU"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 76337, "tid": 0,
    "args": {
      "sort_index": 76337
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 0, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 0, "tid": 0,
    "args": {
      "labels": "GPU 0"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 0, "tid": 0,
    "args": {
      "sort_index": 16777216
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 1, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 1, "tid": 0,
    "args": {
      "labels": "GPU 1"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 1, "tid": 0,
    "args": {
      "sort_index": 16777217
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 2, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 2, "tid": 0,
    "args": {
      "labels": "GPU 2"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 2, "tid": 0,
    "args": {
      "sort_index": 16777218
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 3, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 3, "tid": 0,
    "args": {
      "labels": "GPU 3"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 3, "tid": 0,
    "args": {
      "sort_index": 16777219
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 4, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 4, "tid": 0,
    "args": {
      "labels": "GPU 4"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 4, "tid": 0,
    "args": {
      "sort_index": 16777220
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 5, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 5, "tid": 0,
    "args": {
      "labels": "GPU 5"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 5, "tid": 0,
    "args": {
      "sort_index": 16777221
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 6, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 6, "tid": 0,
    "args": {
      "labels": "GPU 6"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 6, "tid": 0,
    "args": {
      "sort_index": 16777222
    }
  },
  {
    "name": "process_name", "ph": "M", "ts": 1716454215749449, "pid": 7, "tid": 0,
    "args": {
      "name": "python"
    }
  },
  {
    "name": "process_labels", "ph": "M", "ts": 1716454215749449, "pid": 7, "tid": 0,
    "args": {
      "labels": "GPU 7"
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 7, "tid": 0,
    "args": {
      "sort_index": 16777223
    }
  },
  {
    "name": "thread_name", "ph": "M", "ts": 1716454215749449, "pid": 5, "tid": 7,
    "args": {
      "name": "stream 7 "
    }
  },
  {
    "name": "thread_sort_index", "ph": "M", "ts": 1716454215749449, "pid": 5, "tid": 7,
    "args": {
      "sort_index": 7
    }
  },
  {
    "ph": "X", "cat": "Trace", "ts": 1716454215749281, "dur": 11255661,
    "pid": "Spans", "tid": "PyTorch Profiler",
    "name": "PyTorch Profiler (0)",
    "args": {
      "Op count": 0
    }
  },
  {
    "name": "process_sort_index", "ph": "M", "ts": 1716454215749281,
    "pid": "Spans", "tid": 0,
    "args": {
      "sort_index": 536870912
    }
  },
  {
    "name": "Iteration Start: PyTorch Profiler", "ph": "i", "s": "g",
    "pid": "Traces", "tid": "Trace PyTorch Profiler", "ts": 1716454215749281
  },
  {
    "name": "Record Window End", "ph": "i", "s": "g",
    "pid": "", "tid": "", "ts": 1716454227005160
  }
  ],
  "traceName": "trace.json"
}